article
stringlengths 0
745k
| abstract
stringlengths 0
92.7k
|
|---|---|
quantum information is an important branch of quantum physics .
it includes mainly quantum communication and quantum computation @xcite . by far , many interesting quantum systems have been presented for quantum information processing , such as nuclear magnetic resonance @xcite , quantum dots @xcite , diamond nitrogen vacancy ( nv ) centers @xcite , photonic systems @xcite , circuit quantum electrodynamics ( qed ) @xcite , and so on . due to the good scalability @xcite and convenient operation on superconducting qubits , circuit qed has attracted much attention in recent years . composed of the superconducting circuit and the superconducting 1d resonator , circuit qed @xcite has some good characters for completing quantum information processing .
the superconducting circuit can act as a qubit perfectly .
the energy - level structure of the qubit can be divided into @xmath0 , @xmath1 , @xmath2 , and @xmath3 types @xcite which can not be found in atom systems .
a relative long life time of a superconducting qubit has been realized to reach @xmath4 ms @xcite . the strong coupling strength between a superconducting qubit and a superconducting resonator @xcite has been demonstrated in the experiment .
all these characters make circuit qed as a good platform for the quantum computation based on superconducting qubits . in 2009 , dicarlo _ et al .
_ demonstrated a two - qubit algorithms with a superconducting quantum processor @xcite . in 2012 , reed _
_ realized a three - qubit quantum error correction with superconducting circuits @xcite , and in the same year , lucero _
et al . _ computed the prime factors with a josephson phase qubit quantum processor @xcite in which they integrated five superconducting resonators and four superconducting qubits in a quantum processor .
a superconducting resonator can act as a cavity and a quantum bus , which can be coupled to the distant qubits .
the quality of the resonator can be reached to @xmath5 and even @xmath6 @xcite , that is , the superconducting resonators can also afford a powerful platform for quantum information processing . in 2007 ,
_ resolved the photon number states in a superconducting circuit @xcite . in 2010 , johnson _
realized a quantum non - demolition detection of single microwave photons in a circuit @xcite , and in the same year , strauch _
et al . _ presented a method to synthesize an arbitrary quantum state of two superconducting resonators @xcite . in 2012 , strauch proposed an all - resonant control of superconducting resonators with a drive field @xcite . in 2013
, we proposed a selective - resonance scheme to perform a fast quantum entangling operation for quantum logic gates on superconducting qubits @xcite , assisted by one or two superconducting resonators . by combination of the selective resonance and the tunable period relation between a wanted quantum rabi oscillation and an unwanted one besides the positive influence from the non - computational third levels of the superconducting qubits , these universal quantum gates are significantly faster than previous proposals and do not require any kind of drive fields .
recently , the generation of the noon state @xcite on two resonators attracted much more attention . in 2010 , strauch , jacobs , and simmonds @xcite proposed a scheme for completing the generation of the noon state on two resonators without using the third non - computational excited energy level .
the superconducting qubit was operated with a selective rotation by using a drive field whose amplitude should much smaller than the photon - number - dependent stark shifts on the qubit .
that is , the operation time of the qubit should be extended a little longer . in 2010 ,
merkel and wilhelm @xcite proposed a theoretic scheme for generating noon states on two resonators by using two superconducting qubits and three superconducting resonators . in 2011 , wang _ et al .
_ @xcite demonstrated merkel - wilhelm scheme in experiment . in ref.@xcite ,
a novel method was proposed to generate the noon state on two resonators by using a complicated classical microwave pulse and an all - resonant manipulation .
it can get a very high - fidelity noon state within a much shorter time without using the non - computational excited energy level . in 2013 ,
su _ et al . _
@xcite proposed an interesting scheme for the generation of the noon state on two resonators with the resonant operation between the transmon qubit and the superconducting resonator , assisted by the single - qubit rotation .
the scheme can be completed with @xmath7 steps , and in the first @xmath8 steps , the qubit should be maintained in the third - excited state corresponding to the case that the photon number in each resonator is zero . in this paper , we proposed a scheme to produce the noon state on two resonators in a quantum processor composed of two tunable superconducting resonators coupled to a tunable @xmath9-type three - energy - level superconducting qutrit .
our scheme requires two kind of quantum operations .
one is the resonant operation on the superconducting qutrit and the resonators .
the other is the single - qubit manipulation which can be completed by applying a drive field on the qutrit .
our scheme can be used to produce the noon state on two resonators effectively in a simple and fast way , compared with merkel - wilhelm scheme .
moreover , it does not require us to remain the qutrit in the third - excited state all the time , which relaxes largely the requirements of its implementation in experiment , compared with the previous work in ref .
and @xmath10 are the two microwave - photon resonators .
@xmath11 ( @xmath12 ) is the coupling strength between the resonator @xmath13 and the superconducting qutrit in the transition between the states @xmath14 and @xmath15 ( @xmath15 and @xmath16 ) .
, width=309 ] let us consider a quantum system composed of two superconducting resonators coupled to a superconducting qutrit , shown in fig .
[ fig1 ] ( a ) .
the energy - level structure of the qutrit is the @xmath9 type , which can be found in a superconducting charge qubit , shown in fig .
[ fig1 ] ( b ) . in order to construct the noon state on the two resonators @xmath17 and @xmath10 ,
we exploit the lowest three energy levels of the qutrit , denoted by @xmath18 , @xmath19 , and @xmath20 with the energy @xmath21 .
the hamiltonian of the system composed of the two resonators and the qutrit is ( under the rotating - wave approximation , and we choose @xmath22 below ) @xmath23 . \label{h}\end{aligned}\ ] ] here , @xmath24 and @xmath25 are the transition frequency and the creation operator of the resonator @xmath13 , respectively . @xmath26 and @xmath27 are the creation operators of the two transitions @xmath28 and @xmath29 of the qutrit , respectively .
@xmath30 is the coupling strength between the resonator @xmath13 and the qutrit in the two transitions @xmath31 and @xmath32 , and @xmath33 is the coupling strengths between the resonator @xmath13 and the qutrit in the two transitions @xmath15 and @xmath34 . in order to turn on or off the interaction between the resonators and the qutrit , on one hand , one can tune the transition frequency of the qutrit by using the external magnetic flux , or tune the transition frequency of the resonator to make them resonate or largely detune with each other . on the other hand , one can tune the coupling strength between the qutrit and the resonator .
it worth noticing that a tunable resonator @xcite and a tunable coupling qubit @xcite have been demonstrated in experiment .
the principle of our scheme for generating the noon state on two microwave - photon resonators efficiently is shown in fig .
[ fig1](a ) .
suppose the initial state of the system is @xmath35 here the subscripts @xmath36 and @xmath37 represent the two resonators @xmath17 and @xmath10 , respectively .
that is , the qutrit is in the state @xmath38 , and the resonators are in the state @xmath39 . here and below
, @xmath40 is the fock state of the resonator @xmath13 , which means there are @xmath41 microwave photons in the resonator @xmath13 ( @xmath42 ) . to generate the noon state @xcite @xmath43 on @xmath17 and @xmath10 ( @xmath44 is a special situation of the noon state )
, our scheme needs @xmath45 steps .
the first @xmath8 steps are described as follows .
step @xmath36 : by making both @xmath17 and @xmath10 detune largely with the qutrit , one can use a drive field with the frequency equivalent to the transition frequency @xmath46 of the qutrit to pump the state of the qutrit from @xmath47 to @xmath48 .
the amplitude of the drive field is chosen with a proper value for avoiding to pump the state from @xmath49 to @xmath47 . here
after the operation time @xmath51 ( @xmath52 is the proper amplitude of the drive field for pumping the qutrit from @xmath53 to @xmath48 ) , the state of the system evolves into @xmath54 subsequently , one can tune the transition frequencies of the qutrit and the two resonators to make @xmath17 resonate with the qutrit in the transition @xmath55 .
if the coupling strength between @xmath17 and the qutrit is tuned with a proper value before the resonance , one can neglect the interaction between @xmath17 and the qutrit in the transition @xmath56 .
meanwhile , @xmath10 and the qutrit detune largely with each other . after the interaction time @xmath57
, the state of the system becomes @xmath58 step @xmath59 ( @xmath60 ) : by repeating the operation of the step 1 for @xmath61 times and maintaining @xmath10 detuning largely with @xmath17 and the qutrit all the time , the state of the system is changed to be @xmath62 the whole operation time is @xmath63 here , @xmath64 is the rotated - operation time of the qutrit and @xmath65 is the resonated - operation time between the qutrit and the @xmath17 .
the details of the first @xmath8 steps have been described above .
the next @xmath66 steps are described as follows .
step @xmath67 : by making both @xmath17 and @xmath10 , detune largely with the qutrit , one can apply a drive field with the frequency equivalent to the transition frequency @xmath68 of the qutrit to rotate the states of the qutrit with @xmath56 . by choosing the proper amplitude of the drive field
, one can avoid to flip the qutrit with @xmath55 .
after the operation time @xmath69 , the state of the system evolves from eq.([n1 ] ) to @xmath70 applying a drive field with the frequency equivalent to the transition frequency @xmath46 of the qutrit , one can pump the state of the qutrit from @xmath19 to @xmath20 .
the amplitude of the drive field is chosen with a proper value for avoiding to pump the state from @xmath18 to @xmath19 .
after the operation time @xmath71 , the state of the system evolves into @xmath72 subsequently , one can tune the transition frequencies of the qutrit and the two resonators to make @xmath10 resonate with the qutrit in the transition @xmath29 .
if the coupling strength between t @xmath17 and the qutrit is tuned with a proper value before the resonance , one can neglect the interaction between @xmath10 and the qutrit in the transition @xmath73 .
meanwhile , @xmath17 and the qutrit detune largely with each other . after the interaction time @xmath74 , the state of the system becomes @xmath75 step @xmath76 @xmath77 : by repeating the operation of the step @xmath67 for @xmath78 times , and maintaining @xmath17 detuning largely with the qutrit all the time , the state of the system is changed to be @xmath79 the final step : applying a single - qubit operation to complete the rotations of the states @xmath80 and @xmath81 , the state of the system evolves into @xmath82 by resonating @xmath10 and the qutrit in the transition @xmath56 , and making @xmath17 detune largely with the qutrit , the state shown in eq.([m+14 ] ) is changed to be @xmath83 here , we have generated the noon state on two microwave - photon resonators efficiently .
the operation time of the second @xmath66 steps is @xmath84 @xmath85 is the rotation - operation time of the qutrit and @xmath86 is the resonance - operation time between the qutrit and @xmath10 .
in which , we neglect the operation time of the single - qubit operation in the final step for generating the noon state with large number of the @xmath8 and @xmath66 .
we have described the process of our scheme for generating the noon state on two superconducting resonators which are coupled to a @xmath9-type - energy - level structure superconducting charge qutrit .
it includes two kinds of quantum operations .
the first one is the resonant operation on the qutrit and the resonators .
the second one is the single - qubit operation on the qutrit .
they are the high - fidelity , high - efficiency , and simple quantum operations in experiment in circuit qed systems .
the whole operation time of our scheme for generating the noon state @xmath87 is @xmath88 in the calculation for the operation time in our scheme , we neglect the time for changing the transition frequencies of the superconducting qutrit and the superconducting resonator , and the operation time of the single - qubit operation in the final step . compared with the one in ref.@xcite ,
our scheme for generating the noon state on superconducting resonators is much faster as it is composed of the resonant controls . compared with the one in refs.@xcite , both the number of the resonators and that of the qutrits required in our scheme are much smaller as there are three superconducting resonators and two superconducting qutrits in the scheme in refs.@xcite , but only two superconducting resonators and a superconducting qutrit used in our scheme .
moreover , the single - qubit operation required in our scheme can be achieved with the simple classical drive field , and it is simpler than the one used in ref.@xcite as the amplitude of the drive field should be designed with a complex type and it is difficult to be realized in experiment in the latter . in ref .
@xcite , a similar method is used to generate the noon state on two resonators . in their work
, the transmon qutrit should be maintained in the first @xmath8 steps in the third excited state when there is no microwave photons in each resonators .
it worth noticing that the higher excited states lead to a lower fidelity operation @xcite .
luckily , our scheme does not require us to maintain the qutrit in its third excited state all the time , which relaxes the requirements of its implementation in experiment , compared with the one in ref.@xcite .
compared with a transmon qutrit , the level anharmonicity of a charge qutrit is larger and it is better for us to tune the different transitions of the charge qutrit resonant to the resonator @xcite . in summary , we have proposed an efficient scheme to generate the noon states on two superconducting resonators , assisted by a superconducting qutrit .
it requires some high - fidelity quantum operations , that is , the resonant operation on the qutrit and the resonator and the single - qubit operation on the qutrit .
our scheme is a fast and simple one .
moreover , it does not require to maintain the qutrit in the third excited state with a long time , which relaxes the requirements of its implementation in experiment .
this work is supported by the national natural science foundation of china under grant no .
11174039 and nect-11 - 0031 .
hu c y , young a , obrien j l , et al .
giant optical faraday rotation induced by a single - electron spin in a quantum dot : applications to entangling remote spins via a single photon .
phys rev b , 2008 , 78 : 085307 yang w l , yin z q , xu z y , et al .
one - step implementation of multiqubit conditional phase gating with nitrogen - vacancy centers coupled to a high - q silica microsphere cavity .
appl phys lett , 2010 , 96 : 241113 qian y , zhang y q , xu j b. amplifying stationary quantum discord and entanglement between a superconducting qubit and a data bus by time - dependent electromagnetic field .
chin sci bulletin , 2012 , 57 : 1637 - 1642 | we present an efficient scheme for the generation of noon states of photons in circuit qed assisted by a superconducting charge qutrit .
it is completed with two kinds of manipulations , that is , the resonant operation on the qutrit and the resonator , and the single - qubit operation on the qutrit , and they both are high - fidelity operations . compared with the one by a superconducting transmon qutrit proposed by su et al .
( sci .
rep . * 4 * , 3898 ( 2014 ) ) , our scheme does not require to maintain the qutrit in the third excited state with a long time , which relaxes the difficulty of its implementation in experiment . moreover , the level anharmonicity of a charge qutrit is larger and it is better for us to tune the different transitions of the charge qutrit resonant to the resonator , which makes our scheme faster than others .
+ * key words : * entanglement production , noon states , microwave - photon resonators , superconducting charge qutrit , circuit qed |
in our previous paper@xcite , we have provided a more detailed study on a new formulation of hqeft@xcite and applied it to evaluate the weak transition matrix elements between the heavy hadrons containing a single heavy quark .
consequently , the new formulation of hqeft has exhibited interesting features , such as : the luke s theorem comes out automatically without imposing the equation of motion @xmath10 ; the form factors at zero recoil are found to be related to the meson masses , so that the most important relevant form factors at zero recoil can be fitted from the ground state meson masses ; the number of universal form factors up to the order of @xmath11 are less than the one in the usual hqet .
it is of interest to apply the new formulation of hqeft@xcite to the inclusive decays of bottom hadrons .
the inclusive decays of bottom hadrons have been investigated in the recent years by several groups@xmath12 . while it is well known that in the usual hqet there are still some problems which are not yet well understood .
these problems mainly involve the following issues : firstly , the world average values for the lifetime ratios of bottom hadrons are @xcite @xmath13 while the usual hqet prediction leads to a uniform lifetime for all the bottom hadrons when the nonspectator effects are neglected .
the lifetime differences emerge from @xmath11 order which is found to be small .
one may expect that the @xmath14 terms become dominant and provide about @xmath15 contribution to the lifetime ratio @xmath16 ( or at least @xmath17 corrections to their total decay rates ) .
if it is the case , the heavy quark expansion seems to fail in the inclusive @xmath18 decays except there are some special reasons to explain why the @xmath14 terms become dorminant and the higher order terms @xmath19 are smaller .
the nonspectator effects have recently been considered in ref.@xcite and found to result in the following predictions @xmath20 where @xmath21 and @xmath22 characterize the nonfactorization effects .
such a confliction has received wide attention .
authors in refs.@xcite have also discussed the lifetime ratios and came to a similar result . though the prediction for @xmath23 agrees with the current world average , those of @xmath24 and @xmath25 deviate somewhat from the central values of the world average . in the usual hqet , eq.([bs ] ) is the final result because the only difference between @xmath7 and @xmath8 decays lies in the different ckm matrix elements related in the nonspectator effects . to understand the above problems , one of the attempts is to assume that the local duality may be violated in the nonleptonic inclusive decays .
it was suggested in ref.@xcite that a large @xmath26 order correction in nonleptonic inclusive decays may exist and be simply described by replacing the heavy quark mass by the mass of the decaying hadron in the @xmath27 factor , i.e. , @xmath28 this assumption was further discussed in refs.@xcite .
this simple ansatz could not only resolve the lifetime ratio problem , but also provide the correct decay widths for the @xmath9 baryon and the @xmath29 mesons .
but the charm counting may become much larger than the experimental data .
secondly , it seems to have difficulties to simultaneously explain the semileptonic branching ratio @xmath30 and the charm counting @xmath31 in @xmath7 decays . in general ,
when the charm counting is required to be near the experimental data , the predicted semileptonic branching ratio in the usual hqet is significantly larger than the experimental data .
it has been shown in ref.@xcite that for @xmath32 and @xmath33gev @xmath34 where the first result is for the os scheme and the second one for @xmath35 scheme .
after considering the spectator effects , the results has been found to be improved @xcite @xmath36 it is seen that the uncertainties in the two quantities are anti - correlated .
one may compare them with the world average@xcite @xmath37 thirdly , there is an inconsistent picture between the luke s
theorem@xcite for the exclusive heavy hadron decays and the chay - georgi - grinstein theorem@xcite for the inclusive heavy hadron decays .
luke s theorem tells us that the @xmath26 order corrections are absent if one uses the meson mass to normalize the weak matrix elements in the exclusive heavy to heavy transitions at zero recoil .
while according to the chay - georgi - grinstein theorem , @xmath26 order corrections are absent only when the quark mass is used .
consequently , the prediction of total decay width strongly depends on the value of quark mass .
these problems may arise from the simple treatment in the usual hqet , where the bound state effects and hadronization have not been taken into account .
this may be a strong indication for a necessity of developing a new formulation of effective theory to incorporate these effects .
for this purpose , we are going to devote in this paper the investigation of the inclusive bottom hadron decays and to provide better understanding on the problems mentioned above within the framework of the new formulation of hqeft@xcite .
in section ii , we briefly describe the framework of the new formulation of hqeft . in section iii , we apply it to the inclusive bottom hadron decays @xmath38 and present a general formulation for @xmath39 transitions via the heavy quark expansion . in section iv , we investigate the inclusive decays of @xmath7 and @xmath8 mesons as well as @xmath9 baryon by providing detailed numerical results for their semileptonic branching ratios and lifetime ratios as well as the charm countings @xmath31 .
it is interesting to see that the results obtained by using the new formulation of hqeft are consistent with the experimental data . in particular
, the new formulation of hqeft allows us to simply clarify the well known ambiguity of using the quark mass or hadron mass in the inclusive heavy hadron decays . as a consequence , the ckm matrix element @xmath2 is well determined from the semileptonic decay rate .
conclusions and remarks are presented in the last section .
for completeness , we present in this section a brief description on the framework of the new formulation of hqeft@xcite .
let us begin with the effective lagrangian for a heavy quark @xmath40 where @xmath41 and @xmath42 are defined as @xmath43 and @xmath44 denotes the leading term and @xmath45 represents the terms suppressed by the powers of @xmath26 .
and the representation of left - handed current @xmath46 is @xmath47 with @xmath48 and @xmath49 . in order to exhibit a manifest spin - flavor symmetry in the hqeft , it is useful to introduce a hadron state @xmath50 corresponding to the effective heavy quark field @xmath51 .
the hadron state @xmath50 is related to the state @xmath52 of heavy hadron @xmath53 by the following equation @xmath54 with @xmath55 being the mass of the heavy hadron @xmath53 and @xmath56 .
the factor @xmath55 comes from the standard normalization @xmath57 where the factor @xmath58 is introduced to make the normalization concerning the hadron state @xmath50 to be independent of the heavy flavor , i.e. , @xmath59 with @xmath60 expanding the right - hand side of eq.([hadronmatrix ] ) , to the order of @xmath26 , we have @xmath61 in general , a heavy quark in a hadron can not truly be on - shell due to strong interactions among heavy quark and light quark as well as soft gluons .
the off - shellness of heavy quark in the hadron is characterized by a residual momentum @xmath62 .
the total momentum @xmath63 of the heavy quark in a hadron may be written as : @xmath64 . in the usual hqet one
mainly deals with the heavy quark and treats the light quark as a spectator , which does not affect the properties of heavy hadrons to a large extent ( except the effects of weak annihilation and pauli interference ) .
however , since the light degrees of freedom affect the character of heavy hadrons , it may be this simple treatment that meets difficulties in explaining the lifetime differences between @xmath7 and @xmath8 as well as between @xmath9 and @xmath29 as mentioned in the introduction .
to take the effects of light degrees into account and not to deal with hadronization directly , we will adopt an alternative picture , namely the residual momentum @xmath62 of the heavy quark within a hadron is considered to comprise the contributions of the light degrees of freedom . with this picture the heavy quark may be regarded as a ` dressed heavy quark ' , thus the heavy hadron containing a single heavy quark may be more reliable to be considered as a dualized particle of a ` dressed heavy quark ' . thus the momentum @xmath65 of a hadron @xmath53 is decomposed into @xmath66 with @xmath67 being the momentum depending on the heavy flavor and suppressed by @xmath26 . with this picture ,
the momentum of the ` dressed heavy quark ' inside the hadron is given by @xmath68 hence @xmath69 then @xmath70 and @xmath71 in terms of the operator formulation , it implies that @xmath72 thus to simplify the evaluation of the matrix elements , one may approximately replace the propagator @xmath73 by @xmath74 @xmath75 where @xmath76 for pseudo scalar mesons , @xmath77 for vector mesons and @xmath78 for ground state heavy baryons . with this approximation , the mass formulae can be simply given in terms of the effective lagrangian @xmath45 @xmath79 which can be reexpressed as follows by using eq.([lambda ] ) @xmath80 it is useful to define the mass of ` dressed heavy quark ' as @xmath81 which can be expressed in terms of the hadron mass @xmath82
the techniques for inclusive decays of heavy hadrons in the framework of effective theory was developed in the early years of this decade . here
we shall extend the method of ref.@xcite to the framework of new formulation of hqeft .
let us first briefly recall the basic formulae for the description of inclusive semileptonic decay @xmath83 which is mediated by the effective hamiltonian @xmath84 to calculate the decay rates , it is useful to introduce the hadronic tensor @xmath85 which can be expanded in terms of five form factors when one averages the spin of initial and final states @xmath86 for massless lepton pair , @xmath87 and @xmath88 will not contribute to the decay width @xmath89 to evaluate the form factors @xmath90 , let us consider the time ordered product @xmath91 it is known that the form factors @xmath90 are related to @xmath92 via @xmath93 explicitly , the quark matrix element in eq.([tmunu ] ) is given by @xmath94 here @xmath95 is the momentum of charm quark . within the framework of new formulation of hqeft ,
as we have discussed above , the residual momentum of the bottom quark in the hadron is given by @xmath96 .
for ensuring the leading term to have the largest contribution , we may expand eq.([qm ] ) in terms of the small subtracted momentum @xmath97 here @xmath98 to evaluate the matrix element eq.([qm ] ) , we need to know the following relevant matrix elements @xmath99 up to the leading order in @xmath0
, we have , from eqs.([motion ] ) and ( [ lambda ] ) , the following results for @xmath100 and @xmath101 @xmath102 the matrix element concerning one gluon in eq.([tmunu ] ) has the form @xmath103 ^ 2 } \gamma^\tau b |h > . \end{aligned}\ ] ] expanding eqs.([qm ] ) and ( [ gm ] ) , one can extract the form factors @xmath92 . using the relation eq.([tiwi ] ) , one then obtains @xmath90 . here
we shall only list the final results for @xmath90 [ wi ] @xmath104 with @xmath105 . substituting eq.([wi ] ) into eq.([qxy ] ) and integrating out the variables @xmath106 and @xmath107 , one yields the inclusive lepton spectrum @xmath108 with @xmath109 where @xmath110 and @xmath111 are rescaled variables defined as @xmath112 the kinetic region for @xmath110 is @xmath113 treating charm quark @xmath114 as heavy quark and using @xmath115 , one has @xmath116 thus the total decay width for @xmath117 is found to be @xmath118 similarly , we find that @xmath119 transitions have the following general forms @xmath120 with @xmath121 and @xmath122 and @xmath123 .
the @xmath124 functions arise from qcd radiative corrections .
the one - loop results have been calculated in refs.@xcite .
here we will use the two - loop results in refs.@xcite and adopt the reference value @xmath125 , which leads to the following results @xmath126 , \\ \eta_{cc } & = & 1 - 0.21 \alpha_s(\mu ) + { \alpha_s(\mu)^2 \over \pi^2 } [ -9.99 + 7.17 \ln ( m_w / m_b ) + 4 \ln ( m_w / m_b)^2 ] .\end{aligned}\ ] ] the functions @xmath127 are phase space factors at tree level .
their explicit forms are @xmath128 \nonumber \\ & & + z\,{x^2}\ , { { ( -1 + y ) } ^2}\ , [ 3\ , { { ( -1 + y ) } ^2 } + 2\,x\ , ( 1 + 6\,y ) ] \nonumber \\ & & + { z^2}\,{x^2}\ , [ -15\ , { { ( -1 + y ) } ^2}\ , y + x\ , ( 2 + 8\,y - 18\,{y^2 } ) ] \ } \nonumber \\ & & + 72 x^2 y^2 \ln { ( 1-z - y+\sqrt{\kappa})^2 \over 4zy } , \\
i_2(x , y , z ) & = & -6(x - z)^2 [ z^2 + ( 1-y)^2 -2 z ( 1+y ) ] ^{3\over 2 } , \\ i_3(x , x,0 ) & = & \sqrt{1 - 4x } ( 1+{x \over 2}+3x^2)-3x(1 - 2x^2 ) \ln { 1+\sqrt{1 - 4x } \over 1-\sqrt{1 - 4x}},\end{aligned}\ ] ] with @xmath129 note that for @xmath130 , @xmath131 agrees with that in refs.@xcite . before ending this section , we would like to address the following points : firstly , the mass entering into the factor @xmath132 is neither the heavy quark mass @xmath133 nor the hadron mass @xmath55 , it is the so - called ` dressed heavy quark ' mass @xmath134 , which differs from the hadron mass by terms suppressed by @xmath0 .
secondly , paralleling to luke s theorem , there is no @xmath0 order correction when the hadron mass is used .
thus the uncertain parameters , i.e. , the bottom quark mass @xmath133 and binding energy @xmath135 , do not enter separately into the expression of decay widths . furthermore , one may notice that for the final charm quark , both charm quark mass @xmath136 and ` dressed charm quark ' mass @xmath137 have entered into the phace space factors of the general formulation of the decay rates . where the quark mass @xmath136 comes from the propagator of charm quark , and the ` dressed charm quark ' mass @xmath137 arises from the momentum of charm quark inside the hadron , i.e. , @xmath138 , for a parallel treatment to the bottom quark inside the hadron .
these features enable us to determine the lifetime ratios , semileptonic branching ratios and even the lifetimes of bottom hadrons in rather accuracy .
in our numerical calculations , we shall neglect the contributions from penguin diagrams and other rare decays and incorporate the two - loop radiative corrections in refs.@xcite to calculate the values of observable parameters in @xmath18 decays .
such an approximation causes errors no more than 2% .
thus the total decay width for bottom hadrons are given by @xmath139 the semileptonic branching ratios are defined as @xmath140 @xmath141 other two ratios concerning the charm counting @xmath142 and relative leptonic contributions between the @xmath5 decay @xmath143 and the @xmath6 decay @xmath144 @xmath145 note that the formula given above are general and can be applied to bottom hadrons by just taking different binding energies @xmath146 for different bottom hadrons .
we now discuss these observable quantities in detail below .
the basic parameters involved are @xmath147 , @xmath148 , @xmath149 and @xmath150 . using eq.([massl1l2 ] ) and
the measured masses of the ground state heavy mesons and heavy baryons , one can find that @xmath149 in the heavy baryon system is almost the same as the one in the heavy meson system .
thus besides the known masses@xcite @xmath151 there are only four parameters @xmath147 , @xmath148 , @xmath149 and @xmath150 in our calculations . here
@xmath149 could be computed by qcd sum rules@xcite or other phenomenological model@xcite .
it may also be extracted from fitting the meson spectra .
nevertheless , all the results remain suffering from large uncertainties . here
we shall use the most conservative range @xmath152 for @xmath150 , the value extracted from the known @xmath153 mass splitting is quite stable @xmath154 which is accurate up to power correction of @xmath155 .
the mass parameter @xmath147 arising from the propagator in eqs.([qm ] ) and ( [ gm ] ) shall be taken as an effective pole mass .
its value has a large range@xcite @xmath156 note that @xmath157 is different from the ` heavy quark mass ' in the lagrangian eq.([leff ] ) and the ` dressed heavy quark ' mass in eq.([addmass ] ) . in the numerical calculations
, we will restrict its range to be @xmath158 it is seen that , in the framework of new formulation of hqeft , only ` dressed bottom quark ' mass @xmath159 enters into the decay rate , but both charm quark mass @xmath147 and ` dressed charm quark ' mass @xmath160 appear in the phase space factors at tree level .
so the lifetime ratio is not merely given by the ratio @xmath161 ^ 5/[m_b+\bar{\lambda}(h^\prime)]^5 $ ] .
the @xmath148 , @xmath147 and @xmath149 dependences of the lifetime ratios @xmath3 and @xmath4 are plotted in fig .
it is seen that the ratios are not sensitive to the energy scale @xmath148 . for a large range of parameters @xmath147 and @xmath162 ,
the ratios only slightly change , but for large @xmath162 , the ratios become very sensitive to a large charm quark mass .
the ratios as functions of @xmath147 and @xmath162 also exhibit some minimal points around which they change slowly .
when taking @xmath163 and @xmath164 , we have @xmath165 which show a good agreement with the experimental data@xcite .
there is a puzzle in @xmath18 decays in the usual hqet that the predicted semileptonic branching ratio is significantly greater than the experimental data .
a large qcd enhancement of @xmath166 was expected to suppress the value of the ratio , but it will lead to a much larger charm counting than the world average . the theoretical predictions for @xmath30 and @xmath167 strongly depend on the energy scale @xmath148 , as shown in fig . 2 and fig .
3 . at @xmath168 and @xmath169 ,
we have @xmath170 their uncertainties caused by the charm quark mass and @xmath162 are anti - correlated
. one may compare it with the usual hqet prediction as shown in fig .
6a and 6b .
but their ratio @xmath171 is not sensitive to the choice of the scale @xmath148 as shown in fig .
4 . as the function of @xmath136 and @xmath162
, it shows a minimal point around which its variations become slow . for @xmath172 and @xmath164 ,
we have @xmath173 here the main uncertainties arise from the uncertainties of @xmath149 and @xmath147 .
one may also compare it with the usual hqet prediction as shown in fig .
6c . for @xmath174 gev ,
@xmath175 and @xmath176 , one has @xmath177 .
this shows one of the differences between the framework of new formulation of hqeft and the usual hqet . the cleo data for @xmath30 and @xmath167 is@xcite @xmath178 which leads to @xmath179 . in analogous to @xmath171
, @xmath31 is not sensitive to @xmath148 , but it strongly depends on @xmath147 and @xmath162 .
the uncertainty in the theoretical prediction mainly arises from that of the value @xmath147 as shown in fig .
5 . for @xmath180 and @xmath181 ,
the numerical result for @xmath31 is @xmath182 the uncertainties arising from the scale @xmath148 and @xmath183 are not more than 0.03 .
note that its @xmath147 dependence shows a minimal point around which its variation is relatively slow .
one may compare it with the usual hqet prediction as shown in fig .
6c . for @xmath184 , @xmath174 gev , @xmath175 and @xmath176
, one has @xmath185 . from figs . 2 - 5
, one can see that there exists a common region for parameters @xmath148 , @xmath147 and @xmath149 , so that all the quantities @xmath30 , @xmath167 , @xmath171 and @xmath31 are consistent with experimental data .
when taking the following interesting region for the parameters @xmath186 and normalizing @xmath187 and @xmath188 , we have @xmath189 which is consistent with the world average value .
note that a larger @xmath147 value leads to a lower @xmath171 and @xmath31 but to a larger @xmath30 and @xmath167 . in the usual hqet ,
up to @xmath11 order and neglecting the ` spectator effect ' , however , there does not exist a common region for any choice of parameters , which makes all the quantities to be consistent with the experimental data .
this can been explicitly seen from fig .
, where we have plotted four quantities @xmath30 , @xmath167 , @xmath171 and @xmath31 as the functions of the running energy scale @xmath148 and the mass ratio @xmath190 with @xmath191 ( note that in the usual hqet , the @xmath18 quark mass @xmath133 , instead of the ` dressed heavy quark ' mass @xmath159 , is the basic parameter ) .
it is clear from fig . 6
that there are two obstacles in the usual hqet to obtain a consistent fit .
firstly , a lower @xmath31 value requires a larger mass ratio @xmath192 and scale @xmath148 , while a higher scale will lead semileptonic branching ratio to be larger than the experimental bound , and a higher mass ratio will result in a lower value @xmath171 .
we now discuss more about @xmath8 and @xmath9 decay . in the usual hqet ,
all the above observable quantities in the @xmath7 , @xmath8 and @xmath9 decays get the same results up to @xmath193 order .
this is because the semileptonic decay rates of all the bottom hadrons are the same , thus the usual hqet leads to the following predictions up to @xmath193 order @xmath194 in the usual hqet , the lifetime ratio problems are expected to be solved through the non - leptonic decays .
if it is the case , one should have the following consequences for the ratios of the semileptonic branching ratios [ uhqet ] @xmath195 for the ratio @xmath171 one has @xmath196 one may compare it with the current experimental data@xcite @xmath197 in the framework of new formulation of hqeft , as we have discussed in the previous sections , the picture for heavy hadron containing a single heavy quark is such that the heavy quark in the hadron is off - shell , its off - shellness is characterized by the binding energy @xmath135 .
consequently , the so - called ` dressed heavy quark ' masses @xmath159 and @xmath160 enter into the phase space factors . for @xmath8 and @xmath9 decays , the differences of the phase space factors already appear at tree level .
it is not difficult to obtain the following relations [ hqeft ] @xmath198 compared eqs.([uhqet ] ) and eq.([r ] ) with eq.([hqeft ] ) , one may find the significant difference between the usual hqet and the new formulation .
the more and more accurate experiments in the future will provide results of @xmath199 and @xmath171 to test the usual hqet and the new formulation of hqeft .
when normalizing @xmath200 to @xmath201 and @xmath202 to @xmath203 , we have @xmath204 one of the most important applications is the determination of the ckm matrix element @xmath2 . from eq.([sl ] ) the semileptonic decay rate is given by @xmath205 where the ellipse denotes higher order perturbative and non - perturbative corrections .
it is explicitly seen that only the combination of @xmath135 and @xmath133 , i.e. , @xmath159 , appears in the above equation and there is no @xmath0 corrections . at the same time , the @xmath0 power corrections to the decay width @xmath206 is as small as @xmath207 .
thus all the parameters in eq.([vsl ] ) have clear physical meaning and the higher power corrections can be ignored .
this allows us to extract @xmath2 from fitting the experimental data .
the value of @xmath2 as function of @xmath147 and @xmath148 is shown in fig .
it is easy to find that the value of @xmath2 has only a weak dependence on the energy scale @xmath148 and the charm quark mass @xmath147 and @xmath162 cause the main uncertainties for the prediction .
there exist maximal points around which its variation as the function of @xmath147 and @xmath162 becomes slow .
the values of @xmath2 around those points should be more reliable .
it is interesting to note that once one normalizes the semileptonic branching ratio to the experimental data , @xmath2 exhibits an interesting correlation with the ratio @xmath171 .
its relation is shown in fig .
it is seen that the value of @xmath2 increases linearly as @xmath171 decreases . including two - loop qcd corrections and fitting @xmath208 to the measured data , we obtain @xmath209 and @xmath210 .
the experimental uncertainty comes from the errors of @xmath7 lifetime .
one may compare the above prediction for @xmath2 in the new formulation of hqeft with the one in the usual hqet which is plotted in fig .
we have presented , within the framework of new formulation of hqeft , a general formulation for the inclusive decays of bottom hadrons via the heavy quark expansion .
such a general formulation has exhibited interesting features : the @xmath211 corrections are absent and the mass preliminarily entered into the general formulation is neither the bottom quark mass @xmath133 nor heavy hadron mass @xmath212 , it is the so - called ` dressed bottom quark ' mass @xmath213 that is actually the heavy hadron mass in the infinite quark mass limit , i.e. , @xmath214 $ ] . consequently , the decay rates of the bottom hadrons can be reexpressed in terms of the bottom hadron mass with the @xmath211 corrections remaining absent .
this feature is now consistent with the expectation from luke s theorem , since according to that theorem the sum of all the channels of exclusive @xmath18 decays is free from the @xmath0 order corrections at zero recoil .
such a feature distinguishes from the one obtained by chay - georgi - grinstein theorem within the framework of the usual hqet .
the reason is that in the new formulation of hqeft , luke s theorem comes out automatically without the need of imposing the equation of motion @xmath215 . as a consequence , not only the lifetime ratio @xmath4 but
also the ratio @xmath3 is in good agreement with the experimental data . at the same time , the results for the inclusive semileptonic branching ratio @xmath30 , the ratio @xmath171 and the charm counting @xmath31 are also consistent with the experimental data .
in particular , the ckm matrix element @xmath2 has been nicely extracted from the inclusive semileptonic decay rate , and the result well agrees with the one from the exclusive decays@xcite . for @xmath216 and @xmath217 , we have latexmath:[\[\begin{aligned } { \tau(b^0_s ) \over \tau(b^0 ) } = 0.95 ; & & \quad { \tau(\lambda_b ) \over \tau(b^0 ) } = 0.77,\\ b_{sl}(b^0_s ) = 10.82\% ; & & \quad b_{sl}(\lambda_b ) = 11.32\%,\\ r(b^0 ) = 0.25 ; & \quad r(b^0_s ) = 0.27 ; & \quad r(\lambda_b ) = 0.32,\\ n_c(b^0 ) = 1.19 ; & \quad n_c(b^0_s ) = 1.17 ; & \quad n_c(\lambda_b ) = 1.13 , \\ if the nonspectator effects are taken into account , the charm counting will decrease further@xcite .
it is expected that more precise data for the @xmath7 decays and further test for the predictions on the @xmath8 and @xmath9 systems will provide a useful check for the framework of new formulation of hqeft at the level of higher order corrections .
wang , y.l .
wu and y.a .
yan , to be published in int .
a ; hep - ph/9906529 , 1999 .
a8 * , 819 ( 1993 ) .
j. chay , h. georgi , and b. grinstein , phys .
* b247 * 399 ( 1990 ) . i. bigi , n.g .
uraltsev , and a. vainshtein , phys . lett . *
b293 * 430 ( 1992 ) , erratum - ibid . * b297 * 477 ( 1993 ) ; i. bigi , m. shifman , and a. vainshtein , phys .
lett * 71 * 496 ( 1993 ) .
b. blok , l. koyrakh , m. shifman , and a.i .
vainshtein , phys . rev . *
d49 * 3356 ( 1994 ) ; erratum - ibid * d50 * 3572 ( 1994 ) .
a. manohar and m. wise , phys .
* d49 * 1310 ( 1994 ) .
t. mannel , nucl . phys . *
b413 * 396 ( 1994 ) .
m. neubert , phys . rev .
* d49 * 3392 ( 1994 ) ; ibid , 4623 .
a. falk , z. ligeti , m. neubert , and y. nir , phys .
* b326 * 145 ( 1994 ) , and reference therein . c.c.v .
chang , d. chang , w.f .
chang , h.n .
yu , phys .
* d58 * 094019 ( 1998 ) . for updated world averages of @xmath29 hadron lifetimes and lifetime difference ,
see j. alcaraz _ et al .
_ ( lep @xmath29 lifetime group ) , http://wwwcn.cern.ch/~claires/lepblife.html . h.y .
cheng and k.c .
yang , phys .
rev * d59 * 014011 ( 1999 ) .
m. neubert and c. sachrajda , nucl . phys . *
b483 * 339 ( 1997 ) ; m. neubert , hep - ph/9707217 . m. baek , j. lee , chun liu , and h.s .
song , phys .
* d57 * 4091 ( 1998 ) , and reference therein .
g. altarelli , g. martinelli , s. petrarca , and f. rapano , phys . lett . *
b382 * 409 ( 1996 ) ; g. martinelli , nucl .
instrum . methods .
. res . * a384 * 241 ( 1996 ) . i. bigi , hep - ph/9612293 .
cheng , phys .
* d56 * 2783 ( 1997 ) .
p. ball , m. braun , and p. gosdzinsky , nucl .
b432 * 3 ( 1994 ) ; phys . lett . *
b342 * 362 ( 1995 ) , erratum - ibid . *
b374 * 363 ( 1996 ) ; e. bagan .
p. ball , b. fiol , and p. gosdzinsky , phys . lett .
* b351 * 546 ( 1995 ) . m.e .
luke , phys .
b252 * , 447 ( 1990 ) .
pdg , eur .
j. * c3 * 1 ( 1998 ) .
m. voloshin , phys .
* d51 * 3948 ( 1995 ) .
a. czarnecki , m. jezabek , and j.h .
kuhn , phys .
b346 * 335 ( 1995 ) .
m. luke , m. savage , and m. wise , phys .
b343 * 329 ( 1995 ) ; * b345 * 301 ( 1995 ) .
m. lu , m. luke , m. savage , and b. smith , phys . rev . *
d55 * 2827 ( 1997 ) .
cleo collaboration , t.e .
coan et.al , phys .
* 80 * 1150 ( 1998 ) . | we apply the new formulation of heavy quark effective field theory ( hqeft ) to the inclusive decays of bottom hadrons .
the long - term ambiguity of using heavy quark mass or heavy hadron mass for inclusive decays is clarified within the framework of the new formulation of hqeft .
the @xmath0 order corrections are absent and contributions from @xmath1 terms are calculated in detail .
this enables us to reliably extract the important ckm matrix element @xmath2 from the inclusive semileptonic decay rates .
the resulting lifetime ratios @xmath3 and @xmath4 are found to well agree with the experimental data .
we also calculate in detail the inclusive semileptonic branching ratios and the ratios of the @xmath5 and @xmath6 decay rates as well as the charm countings in the @xmath7 , @xmath8 and @xmath9 systems . for @xmath7 decays ,
all the observables are found to be consistent with the experimental data .
more precise data for the @xmath7 decays and further experimental measurements for the @xmath8 and @xmath9 systems will be very useful for testing the framework of new formulation of hqeft at the level of higher order corrections . |
fig . 1 . : : \(a ) analyses of data for s + pb @xmath45 reaction by eq .
( [ eq : ratio ] ) .
( b ) the same but with coulomb corrections switched off for @xmath46 limit .
( results for @xmath47 pair production looks similar with only change being in the value of @xmath48 which depend on the value of radius parameter @xmath38 . ) fig . 2 . : : flow chart for our procedure of `` seamless fitting ( sf ) '' . fig . 3 . : : results of sf for bec for kaons produced in s + pb collisions ; ( a ) for @xmath47 pairs ; ( b ) for @xmath49 pairs . fig .
4 . : : results of sf for bec for kaons produced in @xmath1 + pb @xmath45 reaction .
5 . : : results of sf for bec for pions produced in s + pb @xmath50 reaction . | we applied an improved coulomb correction method developed by us recently to data on identical @xmath0-pairs production in s + pb and @xmath1 + pb reactions at 200 gev / c obtained by na44 collaboration . to analyse
the whole range of the momentum transfers measured the method of `` seamless fitting '' has been proposed and used together with the asymptotic expansion formula for the coulomb wave function .
we found that such coulomb corrections lead sometimes to different than previously reported ( by na44 collaboration ) interaction region and strongly influence the long range correlations .
+ preprint * dpsu-95 - 4 * ( july , 1995 ) = -45pt = 0 cm = 0 cm = 23.7 cm = 16 cm * introduction : * recently na44 collaboration has reported their data on the bose - einstein correlations ( bec ) of @xmath2 pairs produced in s + pb and @xmath1 + pb reactions at 200 gev / c @xcite . in our previous work
@xcite we have analysed these data by making use of the various source functions with the long range correlation ( without , however , invoking any sort of coulomb corrections ) .
the data @xcite have been corrected for coulomb interactions by applying only the gamow factor . as was pointed out some time ago by bowler
this is , however , not sufficient @xcite . in our recent works @xcite the still improved method of coulomb corrections was presented but not yet applied to any concrete data set . in the present letter , we apply it therefore to the analysis of data for @xmath2 pairs production in s + pb and @xmath1 + pb reactions mentioned above ( in the whole measured momentum transfer region ) and compare our results with those obtained before in @xcite . to be able to analyse the whole range of measured momentum transfer @xmath3 and to avoid wild oscillations developing at large @xmath3 s ( cf .
fig .
1a ) we have to use the the asymptotic expansion of the coulomb wave function together with the procedure of `` seamless fitting '' ( sf ) explained below ( cf . figs .
1b and 2 ) . +
* theoretical formula of bec with coulomb wave function : * to write down an amplitude @xmath4 satisfying bose - einstein statistics it is convenient to decompose the wave function of identical ( charged in our case ) bosons with momenta @xmath5 and @xmath6 into the wave function of the center - of - mass system ( c.m . ) with total momentum @xmath7 and the inner wave function with relative momentum @xmath8 .
it allows us to express @xmath4 in terms of the confluent hypergeometric function @xmath9 @xcite : @xmath10\ : , \label{eq : a}\\ \psi({\mbox{\bf q}},{\mbox{\bf r } } ) & = & \gamma(1+i\eta)e^{-\pi \eta/2 } e^{i{\mbox{\scriptsize \bf q}}\cdot{\mbox{\scriptsize \bf r } } } \phi(-i\eta;1;iqr(1 - \cos \theta))\:,\nonumber\\ \psi_s({\mbox{\bf q}},{\mbox{\bf r } } ) & = & \gamma(1+i\eta)e^{-\pi \eta/2 } e^{-i{\mbox{\scriptsize \bf q}}\cdot{\mbox{\scriptsize \bf r } } } \phi(-i\eta;1;iqr(1 + \cos \theta ) ) \:,\nonumber\end{aligned}\ ] ] where @xmath11 and the parameter @xmath12 .
assuming factorization in the source functions , @xmath13 ( here @xmath14 ) , one obtains the following expression for theoretical bec formula @xcite including the improved coulomb correction @xcite : @xmath15 \label{eq : bec}\\ & = & ( 1 + \delta_{{\mbox{\scriptsize 1c } } } ) + ( \delta_{{\mbox{\scriptsize ec } } } + e_{{\mbox{\scriptsize 2b}}}),\label{eq : result}\end{aligned}\ ] ] where @xmath16 denotes gamow factor and the first and the second parentheses in eq .
( [ eq : result ] ) correspond to the first and the second terms in eq .
( [ eq : bec ] ) and @xmath17 see @xcite .
] and @xmath18 to analyse data corrected only by the gamow factor using our formulae we should use the following ratio : @xmath19 ( 1 + \gamma q).\hspace{1 cm } \label{eq : ratio}\end{aligned}\ ] ] it should be noted that the normalization and an effective degree of coherence , i.e. , the denominator of the ratio @xmath20 , are related to each other .
notice also that other parameters like the additional normalization factor @xmath21 , the long range correlation @xmath22 and @xmath23 are introduced by hand .
+ * source function : * to obtain an explicit expression , we have to decide on some form of the source function . in the present letter , we shall use the gaussian source distribution , @xmath24 . for this type of source function
we have the following formulae for the elements of eqs .
( [ eq : result ] ) and ( [ eq : ratio ] ) @xcite : @xmath25
to analyze data corrected by the coulomb wave function as was done in @xcite , we should modify the formula ( [ eq : ratio ] ) replacing it by the following one : @xmath26 ( 1 + \gamma q ) .
\label{eq : a21}\end{aligned}\ ] ] for the sake of reference , we write down here also the conventional formula ( i.e. , a kind of standard formula without corrections due to the final state interactions ) : @xmath27 ( 1 + \gamma q ) .
\label{eq : a22}\ ] ] * asymptotic expansion of the coulomb wave function : * first of all , it should be noted that the expansion in eq .
( [ eq : result ] ) has to be the limited to @xmath28 only due to mathematical properties of the confluent hypergeometric function used @xcite .
this can be seen as wild oscillation developing in fig . 1 ( a ) where the eq . ( [ eq : ratio ] ) has been simply used .
if we , instead , set the coulomb correction to zero in the region @xmath29 limit , a small step appears as seen in fig . 1 ( b ) . therefore in order to analyse in a consistent way the whole region of the momentum transfer measured we have to use the following asymptotic expansion of the coulomb wave function : @xmath30 \ : , \label{eq : asym}\\ \psi^{{\mbox{\scriptsize asym } } } ( { \mbox{\bf q } } , { \mbox{\bf r } } ) & \!\!\ ! = & \!\!\ ! \exp\{i(qz - \eta\ln(r - z))\}\times \left(1+\frac{\eta^2 } { iq(r - z)}\right)\nonumber\\ & \!\!\ ! & \!\!\ !
+ f(\theta ) \frac{\exp\{i(qr - \eta \ln(2qr))\}}r\ : , \label{eq : asym1}\\ \psi_{{\mbox{\scriptsize s}}}^{{\mbox{\scriptsize asym } } } ( { \mbox{\bf q } } , { \mbox{\bf r } } ) & \!\!\ ! = & \!\!\ ! \exp\{i(-qz - \eta \ln(r+z))\}\times
\left(1+\frac{\eta^2}{iq(r+z)}\right ) \nonumber\\ & \!\!\ ! & \!\!\ !
+ f(\pi - \theta ) \frac{\exp\{i(qr - \eta \ln(2qr))\}}r\ : , \label{eq : asym2}\end{aligned}\ ] ] where @xmath31 and @xmath32 in analyses we should assure a smooth connection between both regions of @xmath33 . to avoid the divergence of denominators @xmath34 , we have to introduce a cutoff parameter @xmath35 ( of the order of @xmath36 ) such that @xmath37 always depends on the magnitude of the interaction region ( @xmath38 ) . ] .
this procedure , shown as a flow chart in fig . 2 ,
is called the `` seamless fitting ( sf ) '' .
+ * analyses of data by sf method : * our results obtained in terms of the new formula ( [ eq : ratio ] ) are shown in table i and figs . 3 and 4 . whereas the parameter @xmath22 ( representing here the influence of long range correlations ) increases now noticeably in comparison with that obtained previously in @xcite , we find that the interaction region represented by @xmath39 remains s + pb reactions almost the same . only in @xmath1 + pb collisions the estimated @xmath38 becomes larger with inclusion of coulomb correction than that obtained in @xcite .
these facts suggest that we should be careful in interpreting any data ( at least those for kaons ) which were corrected for coulomb interactions only by the gamow factor . + * concluding remarks : * we have proposed the possible method of applying the coulomb correction for the bec in the whole region of momentum transfer , the seamless fitting " ( sf ) . we confirm that this method works well when analysing data corrected only by the gamow factor .
+ our analyses of data of @xmath0 pairs in s + pb reaction @xcite shows ( cf . table i ) that @xmath40 = 4 fm and @xmath41 = 3 fm ( i.e. , they differ substantially ) , contrary to the estimation provided in @xcite that @xmath42 @xmath43 3 fm .
this is an important result for the study of signals of the quark gluon plasma ( qgp ) ( see refs .
@xcite ) .
moreover , we found that the long range correlations are strongly affected by coulomb corrections ( most probably because of the long range character of coulomb interactions ) .
+ for the sake of reference , we show also in table ii and fig .
5 results of our analysis of data for s + pb @xmath44 reaction @xcite ( which were corrected by the coulomb wave function method @xcite ) performed by using both eq .
( [ eq : a21 ] ) and the `` standard '' formula ( eq . ( [ eq : a22 ] ) ) . as one can see there are no significant differences between parameters estimated by means of these two formulae , in particular the magnitude of the interaction region is in both cases almost the same . +
* acknowledgements : * the authors would like to thank s. esumi , t. nishimura , and s. d. pandey ( members of na44 collaboration ) for useful conversations and correspondences .
this work is partially supported by japanese grant - in - aid for scientific research from the ministry of education , science and culture ( # 06640383 ) .
99 h. beker et al .
( na44 collaboration ) , _ z. phys .
_ * c 64 * ( 1994 ) 209 .
m. biyajima , t. mizoguchi , y. nakata and g. wilk , _ prog .
theor .
phys . _
* 92 * ( 1994 ) 1223 .
m. g. bowler , _ phys .
lett . _ * b270 * ( 1991 ) 69 .
m. biyajima and t. mizoguchi , _ coulomb wave function correction to bose - einstein correlations _ , preprint suldp-94 - 9 ( dec .
1994 ) .
m. biyajima , t. mizoguchi , t. osada and g. wilk , _ phys .
lett . _ * b353 * ( 1995 ) 340 .
h. bggild et al.(na44 collaboration ) , _ phys .
lett . _
* b302 * ( 1993 ) 510 .
l. i. schiff , _ quantum mechanics _ , 2nd ed .
( mcgraw - hill , new york , 1955 ) , p. 117 . s. nagamiya , _ nucl .
phys . _ * a544 * ( 1992 ) 5c .
s. pratt , t. csrg and j. zimanyi , _ phys .
rev . _
* c42 * ( 1990 ) 2646 .
.parameters of kaonic bec obtained from s + pb and @xmath1 + pb reactions . [ cols="^,^,^,^,^,^,^",options="header " , ] |
observational evidence @xcite suggests that approximately 70% of the energy density in the universe is in the form of an exotic , negative - pressure component , dubbed dark energy .
( for a recent review , see , e.g. , ref .
the observational bounds on the properties of the dark energy have continued to tighten .
taking @xmath2 to be the ratio of pressure to density for the dark energy : @xmath7 recent observational constraints are typically @xmath8 when @xmath2 is assumed constant ( see , e.g. , @xcite and references therein ) .
one possibility is that the dark energy is , in fact , merely a cosmological constant , with @xmath2 exactly equal to @xmath9 .
this is the standard @xmath10cdm model , in which the universe today is roughly 30% matter , and 70% vacuum energy , with the latter having a constant density . in this case , we expect future observations to continue to converge on this value of @xmath2 .
a second possibility is that @xmath2 is close to @xmath9 , but not exactly equal to it . in this case
, an alternative explanation for the dark energy is required .
one possible model , dubbed quintessence , assumes that the dark energy arises from a minimally coupled scalar field ; these models have been extensively explored .
( see , for instance , refs .
@xcite for some of the earliest discussions ) .
the equation of motion for the scalar field simplifies considerably in the limit where the background expansion is well - described by @xmath10cdm , which is the case whenever @xmath2 for the quintessence field is always close to @xmath9 .
this fact was exploited in ref .
@xcite , which considered quintessence models in a potential satisfying the slow - roll conditions , i.e. , @xmath11 and @xmath12 it was shown in this paper that the evolution of @xmath5 in these models , which we will call slow - roll quintessence , " approaches a single functional form , given by the present - day values of @xmath2 and @xmath13 .
( note that the term slow - roll quintessence " refers to the fact that the potential satisfies the slow - roll conditions ; the well - known slow - roll approximation for inflation can not be applied @xcite . for alternative approaches to this problem ,
@xcite ) . however , the slow - roll conditions given by eqs .
( [ slow1 ] ) and ( [ slow2 ] ) , while sufficient to give @xmath2 near @xmath9 today , are not necessary . in this paper
, we consider a second possibility : a scalar field rolling down near a local maximum in the potential .
we assume that we are sufficiently close to the maximum that eq .
( [ slow1 ] ) still applies , but we relax our assumption in eq .
( [ slow2 ] ) .
in analogy with recent discussions of similar models in inflation @xcite , we call these models hilltop quintessence .
" a particularly important model that is well - described by this methodology is the pseudo - nambu - goldstone boson ( pngb ) model @xcite . in the next section ,
we derive the version of the scalar field equation of motion that applies in the limit where the background expansion is well - approximated by a @xmath10cdm model . in sec .
iii , we expand the potential for hilltop quintessence as a quadratic function near its maximum and solve this equation to derive the evolution of the scalar field and thus , the evolution of its equation of state , for generic hilltop quintessence models .
we find that all of these models can be characterized by a single set of functional forms for @xmath5 that depends on the present - day values of @xmath13 and @xmath2 for the quintessence , along with a parameter ( @xmath14 ) that depends on the curvature of the scalar field potential at its maximum .
this result is given in eq .
( [ finalfinal ] ) .
our results are discussed in sec .
consider a minimally - coupled scalar field , @xmath1 , in a potential @xmath15 .
the density and pressure of the scalar field are given by @xmath16 and @xmath17 respectively , and the equation of state parameter , @xmath2 , is given by equation ( [ w ] ) .
the equation of motion for this field in an expanding background is @xmath18 where the hubble parameter @xmath19 is given by @xmath20 and we have assumed a flat universe . here @xmath6 is the scale factor , @xmath21 is the total density , and we work in units for which @xmath22 .
the evolution of the scale factor is further described by @xmath23 where @xmath24 is the total pressure .
we first eliminate the @xmath25 term in eq .
( [ motionq ] ) by making the change of variables @xmath26 which gives @xmath27u + a^{3/2 } v^\prime(u / a^{3/2})=0.\ ] ] applying the expressions for @xmath28 and @xmath29 from eqs .
( [ h ] ) and ( [ ddota ] ) gives @xmath30 this equation is as yet exact .
( [ exact ] ) takes a particularly simple form for two special cases : a matter - dominated universe ( for which @xmath31 ) and a @xmath10cdm universe ( for which @xmath24 is constant ) . in the former case , we obtain @xmath32 and it is easy to use this equation to derive , in a straightforward way , the results of refs . @xcite and @xcite . however , the subject of this paper is the second case .
we assume a universe containing matter and quintessence , but such that the quintessence always has @xmath2 near @xmath9 . in this limit , @xmath24 is well - approximated by a constant : @xmath33 , where @xmath34 is the nearly constant density contributed by the quintessence in this limit .
( [ exact ] ) becomes @xmath35 we expect eq .
( [ lambda_u ] ) to provide a good approximation to the evolution of any scalar field in the limit where the hubble parameter in eq .
( [ motionq ] ) is well - approximated by @xmath10cdm , i.e. , any model in which the kinetic term in eqs .
( [ density ] ) and ( [ pressure ] ) is dominated by the potential term , so that @xmath2 for the scalar field never evolves very far from @xmath9 .
( [ lambda_u ] ) applies , for example , to the limiting model examined in ref . @xcite
. we now apply it to a different class of models .
we are interested in models in which the scalar field evolves near a local maximum in the potential .
the most important model of this type is the pngb model @xcite , for which the potential is given by @xmath36,\ ] ] where @xmath37 and @xmath38 are constants .
( for recent discussions of the pngb model in the context of dark energy , see , e.g. , refs . @xcite and references therein ) .
other , less - well - motivated models with a local maximum in the potential include the gaussian potential , @xmath39 and the quadratic potential @xmath40 our purpose in this paper , however , is not to examine an exhaustive list of such models , but to show that they all converge to a common evolution under certain conditions .
we note that any model with a local maximum in the potential at @xmath41 can be expanded , near this maximum , in the form @xmath42 this expansion will be a good approximation for the pngb model when @xmath43 and for the gaussian model as long as @xmath44 .
it is exact for the quadratic potential . the evolution of @xmath1 for this expansion ( or , equivalently , for the quadratic potential of eq .
[ quadratic ] ) was previously examined in refs .
@xcite . the motivation in ref .
@xcite was similar to our own investigation , while refs .
@xcite were concerned with the future fate of the universe in such a model .
the exact solution to the scalar field equation of motion for this potential is given in ref .
@xcite for the matter - dominated case , while refs .
@xcite give the evolution for the scalar - field - dominated case ( appropriate for the far future of the universe ) .
here we solve for the general case in which both matter and the scalar field contribute to the evolution , since this is the relevant case at low redshift . substituting the expansion of eq .
( [ expand ] ) into eq .
( [ lambda_u ] ) ( and taking @xmath45 ) gives @xmath46u = 0.\ ] ] if we define the constant @xmath47 to be given by @xmath48 then the general solution to eq .
( [ eqfinal ] ) is @xmath49 now we assume that the scale factor is well - approximated by its value in the @xmath10cdm model , which is again a consequence of the assumption that @xmath2 is always close to @xmath9 : @xmath50^{1/3}\sinh^{2/3}(t / t_\lambda),\ ] ] where @xmath51 is the present - day value of @xmath13 , and @xmath52 at present .
the time @xmath53 is defined to be @xmath54 with this expression for @xmath55 , the general solution for @xmath56 is given by [ hilltop_phi ] ( t)=^1/2 we require that at @xmath57 , @xmath1 is equal to a fixed initial value , @xmath58 .
this forces @xmath59 and gives us the value of @xmath60 , so that [ phievol ] = . in the limit where @xmath61
, our solution reduces to the matter - dominated solution in ref .
@xcite , while for @xmath62 , we regain the scalar - field - dominated solution of refs . @xcite .
the equation of state parameter for quintessence is given by @xmath63 taking @xmath64 , eqs .
( [ phievol ] ) and ( [ state ] ) yield @xmath65 ^ 2.\ ] ] we can normalize this expression to the present - day value of @xmath2 , which we denote @xmath66 , and we can use eq .
( [ lcdm a ] ) to express @xmath2 as a function of the scale factor ( or the redshift ) rather than @xmath67 .
we obtain : [ final eos ] 1+w(a)=(1+w_0)a^-3 , where @xmath68 and @xmath69 can be derived from eq .
( [ lcdm a ] ) : @xmath70 and @xmath71 we now define the constant @xmath72 . in terms of the quintessence potential , @xmath14 is just @xmath73 thus , @xmath14 depends only on the value of the potential and its second derivative at its maximum .
note that @xmath74 , so @xmath75 .
( [ final eos ] ) can be written as @xmath76 ^ 2 } { [ ( \omega_{\phi0}^{-1/2}+1)^k(k-\omega_{\phi0}^{-1/2 } ) + ( \omega_{\phi0}^{-1/2}-1)^k ( k+\omega_{\phi0}^{-1/2})]^{2}},\ ] ] where @xmath77 is given by f(a ) = .
( note that @xmath78 , where @xmath79 is the value of @xmath13 as a function of redshift , so that @xmath80 . ) eq .
( [ finalfinal ] ) is our main result .
it shows that , in the limit where @xmath2 is close to @xmath9 ( i.e. , the scalar field potential energy dominates the kinetic energy ) , all of the hilltop quintessence models with a given value of @xmath66 form a single set of models parametrized by the value of @xmath14 , which depends only on @xmath81 . for the case of slow - roll quintessence models , a similar analysis yields only a single form for @xmath5 , once @xmath66 and @xmath51 are fixed @xcite . here
we have more complex behavior , since @xmath5 also varies with @xmath14 .
this behavior is illustrated in fig .
1 , where we fix @xmath82 and @xmath83 and then plot @xmath5 from eq . ( [ finalfinal ] ) as a function of @xmath6 . when @xmath84 , we see that the field evolution diverges significantly from slow - roll quintessence .
in particular , @xmath2 increases more slowly at high redshift , but much more rapidly at low redshift .
this is easily understood from the nature of these potentials : @xmath85 increases as @xmath1 rolls down the potential . in the limit where @xmath86 , corresponding to @xmath87 , both slow - roll conditions ( eqs .
[ slow1 ] and [ slow2 ] ) apply , and we expect the evolution to converge to the form derived in ref .
@xcite , namely @xmath88 \tanh^{-1 } \left[f(a)^{-1}\right ] \biggr]^2 } { \left[\omega_{\phi 0}^{-1/2 } - \left(\omega_{\phi 0}^{-1 } - 1 \right ) \tanh^{-1}\sqrt{\omega_{\phi 0}}\right]^{2}}.\ ] ] it is straightforward to show that eq .
( [ finalfinal ] ) reduces to eq .
( [ wpred ] ) in the limit where @xmath87 .
we also show this effect in fig .
1 : the curve corresponding to @xmath89 is indistinguishable from @xmath5 for slow - roll quintessence .
( [ finalfinal ] ) simplifies considerably for the case where @xmath14 is a small integer .
we have , for example , @xmath90 k=3 : 1+w = ( 1+w_0)[(1-_0)a^3 + _ 0a^6 ] , k=4 1+w = .
of course , it would require rather bizarre fine - tuning of the potential for @xmath14 to be exactly equal to one of these values , but these special cases provide some qualitative insight into the behavior of @xmath5 as a function of @xmath14 .
now we evaluate the accuracy of eq .
( [ finalfinal ] ) when applied to various models of interest .
consider the three hilltop potentials outlined above .
we have , for the quadratic potential , @xmath91 , for the pngb potential , @xmath92 , and for the gaussian potential , @xmath93 . in figs . 2 - 4 ,
we have plotted our expression for @xmath5 from eq .
( [ finalfinal ] ) against the exact numerical evolution for these three potentials , fixing @xmath94 .
( once the potential and @xmath51 are fixed , the value of @xmath58 is chosen to produce the desired value of @xmath66 ) .
the agreement , in all three cases , between eq .
( [ finalfinal ] ) and the exact numerical evolution is excellent , with errors @xmath3 for the quadratic potential , @xmath4 for the pngb potential , and @xmath95 for the gaussian potential . rather surprisingly , the errors for the quadratic potential ( which is the basis of our approximation ) are actually larger than for the other two potentials . however , this is due to the fact that we have made several approximations in our calculation ; these tend to cancel for the pngb and gaussian potentials . the agreement between eq .
( [ finalfinal ] ) and the exact evolution is particularly striking for the case of the pngb potential , which also happens to be the most interesting and well - motivated of the hilltop potentials .
we have therefore examined this potential in more detail . in fig . 5
, we extend our results to larger values of @xmath66 .
it is interesting to see that the agreement remains excellent for @xmath66 as large as @xmath96 .
thus , eq . ( [ finalfinal ] ) represents a nearly exact solution for @xmath5 for the pngb model within a wide range for @xmath66 . finally , in figs .
6 - 9 , we use eq . to construct a @xmath97 likelihood plot for @xmath66 and @xmath98 with @xmath99 , using the recent type ia supernovae standard candle data ( essence+snls+hst from @xcite ) . while none of these models is ruled out by current supernova data , it is interesting to note that the hilltop quintessence models ( @xmath100 ) produce a larger allowed region than the slow - roll quintessence model ( @xmath89 ) , and the allowed region increases with increasing @xmath14 .
our results indicate that hilltop quintessence models with @xmath2 near @xmath9 all produce a similar evolution for @xmath5 , given by eq .
( [ finalfinal ] ) .
the importance of this result lies in the fact that a very general set of models can be mapped onto a fairly constrained set of behaviors for @xmath5 .
note that in general , the evolution given by eq .
( [ finalfinal ] ) and shown in fig .
1 is _ not _ well - described by the popular linear parametrization , @xmath101 @xcite .
the one exception is the limiting case @xmath87 , where we regain the results of ref .
@xcite : in this limit the evolution is roughly linear for @xmath102 ( @xmath103 ) , with @xmath104 .
this investigation , along with ref .
@xcite , can also be thought of as a kind of taylor expansion of the potential about the initial value of @xmath1 .
the results of ref .
@xcite apply when the linear term dominates , while the results presented here assume a quadratic expansion .
the results of ref .
@xcite are then a special case ( albeit a very important special case ) of the results presented here .
of course , our results do not apply to all thawing quintessence models with @xmath2 near @xmath9 , but only those satisfying eq .
( [ slow1 ] ) .
a different approach was taken recently by cahn , et al .
they looked at the evolution of the scalar field at early times , when @xmath105 , while assuming nothing about the detailed nature of the potential .
this allows the derivation of a generic result for the evolution of the scalar field before it dominates the expansion ( see also ref .
@xcite for the special case of tracker fields ) .
our results for the particular class of potentials considered here agree with those given in ref .
@xcite ( as they should ) in the limit where @xmath106 . | we examine hilltop quintessence models , in which the scalar field is rolling near a local maximum in the potential , and @xmath0 .
we first derive a general equation for the evolution of @xmath1 in the limit where @xmath0 .
we solve this equation for the case of hilltop quintessence to derive @xmath2 as a function of the scale factor ; these solutions depend on the curvature of the potential near its maximum .
our general result is in excellent agreement ( @xmath3 ) with all of the particular cases examined .
it works particularly well ( @xmath4 ) for the pseudo - nambu - goldstone boson potential .
our expression for @xmath5 reduces to the previously - derived slow - roll result of sen and scherrer in the limit where the curvature goes to zero . except for this limiting case
, @xmath5 is poorly fit by linear evolution in @xmath6 . |
a breakthrough in our understanding of high - energy nuclear collisions is the recognition @xcite that quantum fluctuations in the wavefunctions of projectile and target , followed by hydrodynamic expansion , result in unique long - range azimuthal correlations between outgoing particles .
the importance of these fluctuations was pointed out in the context of detailed analyses of elliptic flow in nucleus - nucleus collisions @xcite .
it was later realized that fluctuations produce triangular flow @xcite , which has subsequently been measured in nucleus - nucleus collisions at rhic @xcite and lhc @xcite .
recently , fluctuations were predicted to generate significant anisotropic flow in proton - nucleus collisions @xcite , which quantitatively explains @xcite the long - range correlations observed by lhc experiments @xcite .
recently , the atlas and cms experiments reported the observation of a nonzero 4-particle cumulant of azimuthal correlations , dubbed @xmath6 , in proton - nucleus collisions @xcite .
the occurrence of a large @xmath6 in proton - nucleus collisions is not fully understood , even though it is borne out by hydrodynamic calculations with fluctuating initial conditions @xcite .
such higher - order cumulants were originally introduced @xcite to measure elliptic flow in the reaction plane of non - central nucleus - nucleus collisions , and isolate it from other , `` nonflow '' correlations .
it turns out that the simplest fluctuations one can think of , namely , gaussian fluctuations , do not contribute to @xmath6 @xcite . since flow in proton - nucleus collisions
is thought to originate from fluctuations in the initial geometry , one naively expects @xmath7 , _ even if _ there is collective flow in the system . in this paper , we argue that the values observed for @xmath6 in p - pb collisions are naturally explained by non - gaussian fluctuations , which are expected for small systems .
our explanation differs from that recently put forward by bzdak et al .
@xcite that it is due to symmetry breaking ( see eq .
( [ bg ] ) and discussion below ) .
as bzdak et al . , we assume that anisotropic flow @xmath8 scales like the corresponding initial - state anisotropy @xmath0 on an event - by - event basis .
this is known to be a very good approximation in ideal @xcite and viscous @xcite hydrodynamics .
thus flow fluctuations directly reflect @xmath0 fluctuations .
now , @xmath0 is bounded by unity by definition . on the other hand ,
gaussian fluctuations are not bounded , which is the reason why they fail to model small systems .
we propose a simple alternative to the gaussian parameterization which naturally satisfies the constraint @xmath9 .
we show that it provides an excellent fit to all monte - carlo calculations .
in each event , the anisotropy in harmonic @xmath1 is defined ( for @xmath10 ) by @xcite @xmath11 where @xmath12 is the initial transverse density profile near midrapidity in a centered polar coordinate system .
obtained in a monte - carlo glauber simulation of a p - pb collision at lhc , and fits using eqs .
( [ gaussian])-([powerlaw ] ) . ] fig .
[ fig : histogram ] displays the histogram of the distribution of @xmath13 in a p - pb collision at 5.02 tev obtained in a monte - carlo glauber calculation @xcite .
we use the phobos implementation @xcite with a gaussian wounding profile @xcite .
we assume that the initial density @xmath12 is a sum of gaussians of width @xmath14 fm , centered around each participant nucleon with a normalization that fluctuates @xcite .
these fluctuations , which increase anisotropies @xcite , are modeled as in ref .
we have selected events with number of participants @xmath15 , corresponding to typical values in a central p - pb collision .
we now compare different parameterizations of this distribution , which we use to fit our numerical results .
the first is an isotropic two - dimensional gaussian ( we drop the subscript @xmath1 for simplicity ) : @xmath16 where @xmath17 and the distribution is normalized : @xmath18 .
this form is motivated by the central limit theorem , assuming that the eccentricity solely originates from event - by - event fluctuations , and neglecting fluctuations in the denominator .
note that this distribution does not strictly satisfy the constraint @xmath19 , which follows from the definition ( [ defepsilon ] ) . when fitting our monte - carlo results , we have therefore multiplied eq .
( [ gaussian ] ) by a constant to ensure normalization between 0 and 1 .
the rms @xmath20 has been fitted to that of the monte - carlo simulation .
[ fig : histogram ] shows that eq .
( [ gaussian ] ) gives a reasonable approximation to our monte - carlo results , but not a good fit .
bzdak et al .
@xcite have proposed to replace eq .
( [ gaussian ] ) by a `` bessel - gaussian '' : @xmath21 this parameterization introduces an additional free parameter @xmath22 , corresponding to the mean eccentricity in the reaction plane in nucleus - nucleus collisions @xcite .
it reduces to ( [ gaussian ] ) if @xmath23 .
a nonzero value of @xmath22 is however difficult to justify for a symmetric system in which anisotropies are solely created by fluctuations . in fig .
[ fig : histogram ] , @xmath22 and @xmath24 have been chosen so that the first even moments @xmath25 and @xmath26 match exactly the monte - carlo results , as suggested in @xcite .
the quality of the fit is not much improved compared to the gaussian distribution , even though there is an additional free parameter .
note that the bessel - gaussian , like the gaussian , does not take into account the constraint @xmath19 .
we now introduce the one - parameter power law distribution : @xmath27 where @xmath28 .
( [ powerlaw ] ) reduces to eq .
( [ gaussian ] ) for @xmath29 , with @xmath30 . the main advantage of eq .
( [ powerlaw ] ) over previous parameterizations is that the support of @xmath31 is the unit disc : it satisfies for all @xmath28 the normalization @xmath32 . in the limit @xmath33 , @xmath34 .
( [ powerlaw ] ) is the _ exact _ @xcite distribution of @xmath13 for @xmath35 identical pointlike sources with a 2-dimensional isotropic gaussian distribution , with @xmath36 , if one ignores the recentering correction .
in a more realistic situation , eq .
( [ powerlaw ] ) is no longer exact .
we adjust @xmath37 to match the rms @xmath20 from the monte - carlo calculation .
[ fig : histogram ] shows that eq .
( [ powerlaw ] ) ( with @xmath38 ) agrees much better with monte - carlo results than gaussian and bessel - gaussian distributions .
cumulants of the distribution of @xmath20 are derived from a generating function , which is the logarithm of the two - dimensional fourier transform of the distribution of @xmath39 : @xmath40 where angular brackets denote an expectation value over the ensemble of events . if the system has azimuthal symmetry , by integrating over the relative azimuthal angle of @xmath41 and @xmath42 , one obtains @xmath43 where @xmath44 and @xmath17 .
the cumulant to a given order @xmath1 , @xmath45 , is obtained by expanding eq .
( [ defgbessel ] ) to order @xmath46 , and identifying with the expansion of @xmath47 to the same order .
this uniquely defines @xmath45 for all _ even _ @xmath1 .
one thus obtains @xcite @xmath48 , @xmath49 .
expressions of @xmath50 and @xmath51 are given in @xcite .
.values of the first eccentricity cumulants for the gaussian ( [ gaussian ] ) , bessel - gaussian ( [ bg ] ) and power law ( [ powerlaw ] ) distributions . [ cols="<,^,^,^",options="header " , ] expressions of the first four cumulants are listed in table [ table : models ] . for the power law distribution ( [ powerlaw ] ) , these results are obtained by expanding the generating function ( [ defgbessel ] ) : @xmath52\cr & = & \ln\left[\frac{2^{\alpha}\alpha!}{k^{\alpha}}j_{\alpha}(k)\right ] .
\end{aligned}\ ] ] general results have been obtained previously in the case of @xmath35 pointlike sources and in the large @xmath35 limit for @xmath53 @xcite and @xmath54 @xcite .
our results derived from eq .
( [ powerlaw ] ) are exact for a gaussian distribution of sources and therefore agree with these general results for @xmath55 .
similar results have also been derived for @xmath56 and @xmath57 @xcite , but not for cumulants of order 6 or higher . for the power law distribution ( [ powerlaw ] ) , where we have set @xmath58 .
] fig .
[ fig : gaussian ] displays the cumulants @xmath59 to @xmath51 as a function of @xmath35 , as predicted by eq .
( [ powerlaw ] ) for pointlike sources . by @xmath60 in the exact result of ref .
these results are similar to those obtained in full monte - carlo glauber calculations @xcite . in the limit @xmath55
, the power law distribution yields @xmath61 .
it thus predicts a strong ordering @xmath62 , unlike the bessel - gaussian which predicts @xmath63 . for fixed @xmath35
, however , the cumulant expansion quickly converges , as illustrated in fig .
[ fig : gaussian ] . in practice , for typical values of @xmath35 in p - pb collisions , one observes @xmath64 , in agreement with numerical findings of bzdak et al . @xcite .
this rapid convergence can be traced back to the fact that the generating function @xmath65 in eq .
( [ gpowerlaw ] ) has a singularity at the first zero of @xmath66 , denoted by @xmath67 .
this causes the cumulant expansion to quickly converge to the value @xcite @xmath68 this asymptotic limit is also plotted in fig .
[ fig : gaussian ] .
it is hardly distinguishable from @xmath50 and @xmath51 for these values of @xmath35 . versus @xmath59 .
the dashed line in both panels is eq .
( [ eps4vseps2 ] ) . left : p - pb collisions .
`` full '' refers to gaussian sources associated with each participant , and fluctuations in the weights of each source .
`` pointlike '' refers to pointlike identical sources .
dipsy results for p - p collisions are replotted from @xcite .
right : pb - pb collisions .
the dotted line is @xmath69 , corresponding to a nonzero mean eccentricity , and negligible fluctuations . ]
the power law distribution ( [ powerlaw ] ) predicts the following parameter - free relation between the first two cumulants : @xmath70 this relation can be used to test the universality of the distribution ( [ powerlaw ] ) . for p - pb collisions at 5.02 tev ,
we run two different types of monte - carlo glauber calculations : a full monte - carlo identical to that of fig .
[ fig : histogram ] , and a second one where fluctuations and smearing are switched off ( identical pointlike sources ) .
we calculate @xmath13 and @xmath71 for each event .
events are then binned according to the number of participants @xmath35 , mimicking a centrality selection . for p - p collisions at 7 tev , we use published results @xcite obtained with the event generator dipsy @xcite , which are binned according to multiplicity
results are shown in fig .
[ fig : e2e4 ] ( left ) .
each symbol of a given type corresponds to a different bin .
all monte - carlo results are in very good agreement with eq .
( [ eps4vseps2 ] ) . a closer look at the results show that the `` full '' monte - carlo glauber calculations are above the line by @xmath72 ( for both @xmath13 and @xmath71 ) , the `` pointlike '' results for @xmath71 by @xmath73 , and the `` pointlike '' results for @xmath13 ( where our result is exact , up to the recentering correction ) by @xmath74 .
dipsy results are above the line by @xmath75 . for pb - pb collisions at 2.76 tev ( fig .
[ fig : e2e4 ] right ) , we use the results obtained in ref .
@xcite using the monte - carlo glauber @xcite and monte - carlo kln @xcite models .
these results are in 5% centrality bins . for @xmath71 ,
both models are in very good agreement with eq .
( [ eps4vseps2 ] ) ( within @xmath76 or so ) .
note that pb - pb collisions probe this relation closer to the origin , in the large @xmath35 limit where more general results are available @xcite .
these general results predict @xmath77 for @xmath78 , but with a proportionality constant that depends on the density profile .
our results show that it is in practice very close to the value predicted by eq .
( [ eps4vseps2 ] ) , namely , @xmath79 .
monte - carlo results for @xmath13 in pb - pb differ from eq .
( [ eps4vseps2 ] ) .
this is expected , since @xmath13 in mid - central pb - pb collisions is mostly driven by the almond shape of the overlap area between colliding nuclei @xcite , not by fluctuations . in the limiting case where fluctuations are negligible , @xmath80 .
our results show that fluctuations dominate only for the most central and most peripheral bins .
we conclude that the power law distribution ( [ powerlaw ] ) is a very good approximation to the distribution of fluctuation - driven eccentricities , irrespective of the details of the model .
this could be checked explicitly with other initial - state models @xcite . as a function of the measured @xmath81 .
typical values for @xmath82 in pb - pb @xcite and @xmath83 in p - pb collisions @xcite are indicated by arrows . ]
we now discuss applications of our result . the distribution of @xmath0 is completely determined by the parameter @xmath37 in eq .
( [ powerlaw ] ) .
this parameter can be obtained directly from experimental data .
assuming that anisotropic flow is proportional to eccentricity in the corresponding harmonic , @xmath84 , which is proven to be a very good approximation for @xmath10 @xcite , one obtains @xmath85 the first equality has already been checked against monte - carlo models and experimental data @xcite .
the second equality directly relates the parameter @xmath37 in eq .
( [ powerlaw ] ) to the measured ratio @xmath81 .
this in turn gives a prediction for ratios of higher - order flow cumulants , which scale like the corresponding ratios of eccentricity cumulants .
these predictions are displayed in fig .
[ fig : predictions ] .
one can also directly obtain the rms eccentricity @xmath59 , which is a property of the initial state .
the ratio @xmath86 in pb - pb is close to 0.5 in mid - central collisions @xcite .
we thus predict @xmath87 and @xmath88 in the same centrality .
we also obtain @xmath89 , which is a typical prediction from monte - carlo models in the 10%-20% or 20%-30% centrality range @xcite .
similarly , the ratio @xmath90 measured in p - pb collisions @xcite implies @xmath91 and @xmath92 , that is , almost degenerate higher - order cumulants .
we obtain @xmath93 , in agreement with monte - carlo glauber models @xcite .
we have proposed a new parameterization of the distribution of the initial anisotropy @xmath0 in proton - proton , proton - nucleus and nucleus - nucleus collisions which , unlike previous parameterizations , takes into account the condition @xmath9 .
this new parameterization is found in good agreement with results of monte - carlo simulations when @xmath0 is created by fluctuations of the initial geometry .
our results explain the observation , in these monte - carlo models , that cumulants of the distribution of @xmath0 quickly converge as the order increases .
this is because the fourier transform of the distribution of @xmath0 has a zero at a finite value of the conjugate variable @xmath94 .
this , in turn , is a consequence of the fact that the probability distribution of @xmath0 has compact support ( that is , @xmath9 ) .
the consequence of this universality is that while the rms @xmath0 is strongly model - dependent @xcite , the probability distribution of @xmath0 is fully determined once the rms value is known in particular , the magnitudes of higher - order cumulants such as @xmath95 . assuming that anisotropic flow @xmath8 is proportional to @xmath0 in every event , we have predicted the values of @xmath96 and @xmath97 in pb - pb collisions , and the values of @xmath98 and @xmath99 in p - pb collisions .
if future experimental data confirm our prediction , these results will strongly support the picture that the long - range correlations observed in proton - nucleus and nucleus - nucleus collisions are due to anisotropic flow , which is itself proportional to the anisotropy in the initial state .
this picture , furthermore , will be confirmed irrespective of the details of the initial - state model .
jyo thanks art poskanzer for pointing out , back in 2009 , that bessel - gaussian fits to monte - carlo glauber calculations fail because they miss the constraint @xmath100 , larry mclerran for discussing ref .
@xcite prior to publication , christoffer flensburg for sending dipsy results , ante bilandzic and wojciech broniowski for useful discussions , and jean - paul blaizot and raju venugopalan for comments on the manuscript .
we thank the yukawa institute for theoretical physics , kyoto university .
discussions during the yitp workshop yitp - t-13 - 05 on `` new frontiers in qcd '' were useful to complete this work .
ly is funded by the european research council under the advanced investigator grant erc - ad-267258 .
99 b. alver _ et al . _
[ phobos collaboration ] , phys .
* 98 * , 242302 ( 2007 ) [ nucl - ex/0610037 ] .
b. alver and g. roland , phys .
c * 81 * , 054905 ( 2010 ) [ erratum - ibid .
c * 82 * , 039903 ( 2010 ) ] .
m. miller and r. snellings , nucl - ex/0312008 .
a. adare _ et al .
_ [ phenix collaboration ] , phys .
* 107 * , 252301 ( 2011 ) l. adamczyk _ et al .
_ [ star collaboration ] , phys .
c * 88 * , 014904 ( 2013 ) [ arxiv:1301.2187 [ nucl - ex ] ] .
k. aamodt _ et al . _
[ alice collaboration ] , phys .
lett . * 107 * , 032301 ( 2011 ) [ arxiv:1105.3865 [ nucl - ex ] ] .
s. chatrchyan _ et al . _
[ cms collaboration ] , eur .
j. c * 72 * , 2012 ( 2012 ) [ arxiv:1201.3158 [ nucl - ex ] ] . g. aad _ et al .
_ [ atlas collaboration ] , phys .
c * 86 * , 014907 ( 2012 ) [ arxiv:1203.3087 [ hep - ex ] ] .
p. bozek , phys .
c * 85 * , 014911 ( 2012 ) .
p. bozek and w. broniowski , phys .
b * 718 * , 1557 ( 2013 ) [ arxiv:1211.0845 [ nucl - th ] ] .
s. chatrchyan _ et al . _
[ cms collaboration ] , phys .
b * 718 * , 795 ( 2013 ) [ arxiv:1210.5482 [ nucl - ex ] ] .
b. abelev _ et al . _
[ alice collaboration ] , phys .
b * 719 * , 29 ( 2013 ) [ arxiv:1212.2001 ] .
g. aad _ et al . _ [ atlas collaboration ] , phys .
lett . * 110 * , 182302 ( 2013 ) [ arxiv:1212.5198 [ hep - ex ] ] .
g. aad _ et al .
_ [ atlas collaboration ] , phys .
b * 725 * , 60 ( 2013 ) [ arxiv:1303.2084 [ hep - ex ] ] .
s. chatrchyan _ et al . _
[ cms collaboration ] , phys .
b * 724 * , 213 ( 2013 ) [ arxiv:1305.0609 [ nucl - ex ] ] .
p. bozek and w. broniowski , phys .
c * 88 * , 014903 ( 2013 ) [ arxiv:1304.3044 [ nucl - th ] ] .
n. borghini , p. m. dinh and j. -y .
ollitrault , phys .
c * 63 * , 054906 ( 2001 ) [ nucl - th/0007063 ] .
n. borghini , p. m. dinh and j. -y .
ollitrault , phys .
c * 64 * , 054901 ( 2001 ) [ nucl - th/0105040 ] .
s. a. voloshin , a. m. poskanzer , a. tang and g. wang , phys .
b * 659 * , 537 ( 2008 ) [ arxiv:0708.0800 [ nucl - th ] ] .
a. bzdak , p. bozek and l. mclerran , arxiv:1311.7325 [ hep - ph ] .
z. qiu and u. w. heinz , phys .
c * 84 * , 024911 ( 2011 ) .
h. niemi , g. s. denicol , h. holopainen and p. huovinen , phys .
c * 87 * , 054901 ( 2013 ) .
d. teaney and l. yan , phys .
c * 83 * , 064904 ( 2011 ) .
m. l. miller , k. reygers , s. j. sanders and p. steinberg , ann .
nucl . part .
sci . * 57 * , 205 ( 2007 ) [ nucl - ex/0701025 ] .
b. alver , m. baker , c. loizides and p. steinberg , arxiv:0805.4411 [ nucl - ex ] .
m. alvioli , h. -j . drescher and m. strikman ,
b * 680 * , 225 ( 2009 ) [ arxiv:0905.2670 [ nucl - th ] ] .
m. rybczynski and w. broniowski , phys .
c * 84 * , 064913 ( 2011 ) [ arxiv:1110.2609 [ nucl - th ] ] .
m. rybczynski , g. stefanek , w. broniowski and p. bozek , arxiv:1310.5475 [ nucl - th ] .
a. dumitru and y. nara , phys .
c * 85 * , 034907 ( 2012 ) [ arxiv:1201.6382 [ nucl - th ] ] .
ollitrault , phys .
d * 46 * , 229 ( 1992 ) .
r. s. bhalerao and j. -y .
ollitrault , phys .
b * 641 * , 260 ( 2006 ) [ nucl - th/0607009 ] .
b. alver _ et al .
_ , phys .
c * 77 * , 014906 ( 2008 ) .
r. s. bhalerao , m. luzum and j. -y .
ollitrault , phys .
c * 84 * , 054901 ( 2011 ) [ arxiv:1107.5485 [ nucl - th ] ] .
r. s. bhalerao , n. borghini and j. y. ollitrault , nucl .
a * 727 * , 373 ( 2003 ) [ nucl - th/0310016 ] .
e. avsar , c. flensburg , y. hatta , j. -y .
ollitrault and t. ueda , phys .
b * 702 * , 394 ( 2011 ) . c. flensburg , arxiv:1108.4862 [ nucl - th ] .
r. s. bhalerao , m. luzum and j. -y .
ollitrault , phys .
c * 84 * , 034910 ( 2011 ) [ arxiv:1104.4740 [ nucl - th ] ] .
h. j. drescher and y. nara , phys .
c * 76 * , 041903 ( 2007 ) b. schenke , p. tribedy and r. venugopalan , phys .
c * 86 * , 034908 ( 2012 ) r. s. bhalerao , m. luzum and j. y. ollitrault , j. phys .
g * 38 * , 124055 ( 2011 ) [ arxiv:1106.4940 [ nucl - ex ] ] .
s. chatrchyan _ et al . _
[ cms collaboration ] , arxiv:1310.8651 [ nucl - ex ] .
e. retinskaya , m. luzum and j. -y .
ollitrault , phys .
c * 89 * , 014902 ( 2014 ) . | we show that the statistics of fluctuation - driven initial - state anisotropies in proton - proton , proton - nucleus and nucleus - nucleus collisions is to a large extent universal .
we propose a simple parameterization for the probability distribution of the fourier coefficient @xmath0 in harmonic @xmath1 , which is in good agreement with monte - carlo simulations .
our results provide a simple explanation for the 4-particle cumulant of triangular flow measured in pb - pb collisions , and for the 4-particle cumulant of elliptic flow recently measured in p - pb collisions . both arise as natural consequences of the condition that initial anisotropies are bounded by unity .
we argue that the initial rms anisotropy in harmonic @xmath1 can be directly extracted from the measured ratio @xmath2 : this gives direct access to a property of the initial density profile from experimental data .
we also make quantitative predictions for the small lifting of degeneracy between @xmath3 , @xmath4 and @xmath5 .
if confirmed by future experiments , they will support the picture that long - range correlations observed in p - pb collisions at the lhc originate from collective flow proportional to the initial anisotropy . |
galactic halos formed from lcdm initial conditions in n - body simulations have approximately 10% of their mass in orbiting sub - halos @xcite . the gravitational stirring and heating of a galactic disk and halo star clusters by dark sub - halos
is tempered by the large numbers , high random velocities and broad distribution of sub - halos throughout the overall dark halo . however ,
very low velocity dispersion star streams in galactic halos are sensitive to the degree to which the dark matter in the halo is sub - structured into thousands of orbiting sub - halos . the sub - halos fold and chop the star - streams and gradually increase the velocity dispersion around the mean motion to about @xmath115% of the halo circular velocity , typically 30 , over a hubble time @xcite .
consequently , cool star streams are sensitive indicators of the presence of the predicted dark matter substructure .
about half a dozen of the currently known milky way streams qualify as cool ( most confidently , pal 5 , gd-1 , orphan , archeron and styx ) that is , having local velocity dispersions below about 10 , or , width less than about 0.1 radian as seen from the center of the host galaxy .
a dark matter sub - halo crossing such a cool stream will lead to visible disturbances .
unfortunately , the star count data often do not yet have sufficient local numbers to allow statistically significant measurements of density variations relative to the galactic foreground and background @xcite .
the pan andromeda archeological survey ( pandas ) @xcite in one fell swoop provides deep and uniform data to a distance of about 150 kpc from m31 s center . the spectacular star stream north - west of m31 , more than 100 kpc long , was first displayed in its entirety in @xcite .
the great length of the stream , as well as being fairly distant from the disturbing effects of the main body of m31 , make the stream an exceptionally interesting case for analysis of density variations . and , the star counts have sufficient signal to noise to allow reliable local surface density measurements of the stream . multiple image gravitational lensing by a galaxy of a background quasars is another probe of sub - halos .
@xcite pointed out that the anomalous flux ratios relative to a smooth potential model that accurately predicts the locations of the images are the natural consequence of substructure in galaxy halos .
a statistical analysis of available strong lens systems @xcite assuming nfw density profile sub - halo profiles found that about 2% of the halo mass ( with a very large uncertainty ) is in substructure .
lensing is very sensitive to the central density profile which @xcite found to be close to an einasto profile , which is shallower than the nfw profile .
the theoretical situation has become somewhat unclear , since a projection of a sub - halo rich n - body galaxy simulation @xcite finds image distortions generally smaller than those seen in strong lens systems .
a better overall understanding and tighter limits requires better data on more strong lens systems @xcite .
in this paper we analyze m31 s nw star stream for the density variations that sub - halos are expected to induce .
we first trace the mean centerline of the stream .
the luminosity function and total luminosity of the main part of the stream are estimated as an indication of the progenitor system . the stream s width and deviations from the centerline
are measured .
the measurement of the density along the stream gives the data which we use to test whether the stream is smooth or lumpy .
an illustrative simulation is used to show the expected scale and degree of stream structure as the amount of substructure increases from the visible dwarfs to the thousands of halos that lcdm predicts .
pandas has acquired @xmath2 and @xmath3 band images covering nearly 400 square degrees of m31-m33 with the cfht megaprime system .
the images have a minimum s / n=10 at @xmath4 mag .
details of data reduction are in @xcite and @xcite .
we adopt the @xcite distance of 785 kpc ( dm of 24.47 mag ) for m31 , although we note a statistically consistent , but somewhat smaller , cepheid distance having a dm 24.32@xmath5 mag or 731 kpc @xcite .
the adopted distance indicates that the survey reaches to m@xmath6 mag at high signal - to - noise .
we use data up to and including january , 2010 . image extraction yields approximately 31 million stellar objects in the images . matching the colors and magnitudes of the stars to isochrones of varying metallicity at a fixed age of 12
gyr @xcite identifies m31 red giant branch ( rgb ) stars using the procedures of @xcite .
since we are seeing m31 through our own galaxy , it turns out that more than 80% of the stars do not match rgb isochrones at the adopted distance .
these stars are not included in the map , although all stars above the completeness limit are useful for the measurement of geometric biases of the images .
the stars are individually extinction corrected with galactic dust maps @xcite .
we are left with about 4.3 million nominal red giant branch stars to 24.5 mag which have been assigned metallicities in the range of [ fe / h ] = [ -3 , 0 ] . of these stars , approximately 25% are in the nw quadrant which is our area of interest .
the galactic stars that happen to fall into the range of colors and magnitudes that are appropriate for m31 red giants create a slowly varying foreground which will be removed in the analysis of the stream density .
note that a shorter distance would make the inferred luminosities smaller and cause the isochrone match to systematically yield larger metallicities . to a first approximation this simply offsets the metallicity range we analyze and does not otherwise affect the analysis .
star streams are identified as spatially coherent over - densities , over a limited range of metallicities . to create the map we project the rgb stars onto a tangent plane , for which we chose a center at ra=0h 24 m and dec = + 44 which is near the center of the nw stream .
this center is offset from m31 s center , to make the co - ordinates reasonably rectangular in the region of the stream analysis . without the rgb star selection no streams are readily visible in the halo .
figure [ fig_m31 ] plots the sky distribution ( in the tangent plane co - ordinates ) of the [ fe / h]=[-0.6 , -2.4 ] stars which we find below comprise the stream .
the same feature is clear in any map made with the low metallicity star sample .
the low metallicity map , figure [ fig_m31 ] , has a number of geometric sampling variations that must be controlled to make a reliable measurement of local stream density .
the simplest problem is that the ccd array has gaps between the individual chips of @xmath780 in the horizontal direction and @xmath713 in the vertical direction .
the images were dithered to cover the smaller gaps , but the resulting depth is shallower than the surrounding area and no simple count correction procedure that did not increase the noise was found .
these gaps are identified in the camera ccd co - ordinate system and masked out of the map . a more complicated geometric bias results from the catalog being built up from stacked local images in which the stellar photometry was done to create a local catalog .
these catalogs were then matched in the areas of overlap to eliminate duplicates and produce a global catalog .
the images are vignetted towards the edge of the array so the image depth drops at the field edge , however the vignetting is partially compensated in the catalog matching at the image edges .
that is , if the surface density is corrected with the inverse of the mean star density over the array then the map has an excess at the edges .
it is straightforward to devise a correction which flattens the average surface density .
the correction works well over a range of about 20% in surface density , but undesirably amplifies the noise if applied at lower completeness levels .
we apply a cut to the image where the local surface density in ccd image co - ordinates drops below 20% of the average .
the resulting masked out area is about 15% of the image , a fraction which is not very sensitive to the precise value of the cut .
the masked out regions are not included in the analysis and are readily visible in figure [ fig_m31 ] .
field to field photometric depth variations will introduce artificial variations in the density of stars in the map .
the depth of an image for a fixed exposure time depends on the sky transparency , brightness and image quality at the time of observation . below the completeness limit
there is substantial field to field variation that leads to a random checkerboard appearance , but the pattern essentially disappears above the completeness limit .
our measurement of the stream density subtracts the local mean background to reduce the problems of local depth variations . a second
, more difficult , form of depth variations is field to field variations in the zero points of the two filter bands .
we can set a limit on these by using the double measurements of star brightness in the overlap regions . for bright stars where the photon noise is small
the standard deviation is 0.037 mag in @xmath2 band and 0.052 mag in @xmath3 band , with no discernible pattern over the field .
the star catalog in our selected range of color and magnitude has nearly constant numbers with increasing depth near the completeness limit , so a 5% change in zero point leads to about a 5% change in numbers .
the 5% should be considered an upper limit since the edges of the images are vignetted and have significant local calibration variations .
across each 1 megaprime field there are a myriad of camera co - ordinate dependent photometric offsets @xcite which the current calibration does not fully take into account .
these offsets lead to a change in zero point within each one degree pointing that is an additional @xmath72 - 3% variation .
the masking procedure removes the worst of these variations , and the non - linear correction to flat star counts substantially reduces the variations within a field .
what remains is below the shot noise of the star counts . the north - west stream , readily visible in the star map of figure [ fig_m31 ] , extends more than 100 kpc from m31 in projection .
an ellipse is used to trace the stream , which corresponds to finding the lowest order fourier mode , that is , @xmath8 , rotated and centered for a best fit .
note that the angle @xmath9 is the angle once the ellipse is transformed into a sphere and is not the geometric angle except along the major and minor axes .
higher order fourier terms are not yet justified given the signal to noise of the stream within these data .
the ellipse parameters are varied , along with the stream width and the background region width to maximize the mean over - density within @xmath10 of the ellipse over the lower branch of the stream , taken as the angular range @xmath11 $ ] .
the stream region is shown in the mexican top - hat filtered figure [ fig_cutout ] . this filtered image , with an inner positive region of radius 0.4 and outer negative region of radius of 0.8 , with an area integral of zero , usefully illustrates the stream and noise properties near the filter length .
the analysis is done on an unfiltered high resolution version of figure [ fig_m31 ] with pixels of 18 , ten times smaller than in figures [ fig_m31 ] and [ fig_cutout ] .
we find that the best fit ellipse has a location of [ -1.25,+1.90 ] in the tangent plane co - ordinates of figure [ fig_m31 ] relative to our field center at [ 0h 24m,+44 ] .
there is no particular reason for this fitting procedure to be constrained to have an ellipse centered on m31 .
the ellipse has axes of @xmath12 ( or @xmath13 kpc ) at an angle of 42.25 w of n. the best fit occurs for an ellipse that is within the image data rather than for one that extends off the imaged area , even if the fitting procedure is started with a longer ellipse that extends out of the field .
rr & @xmath14 + [ fe / h ] - 0.2 & stars kpc@xmath15 + -0.60 & 0.05 + -0.80 & 0.16 + -1.00 & 0.41 + -1.20 & 0.53 + -1.40 & 0.27 + -1.60 & 0.28 + -1.80 & 0.06 + -2.00 & 0.17 + -2.20 & 0.08 + -2.40 & 0.01 + the straightforward consequences of projection are that the stream distance from m31 must be 117 kpc or larger , and , the ellipse must be at least as circular as we found above .
we have assumed that all the stars in the stream are at 785 kpc , the adopted distance to m31 .
however , the distances of stars around the stream could vary roughly 100 kpc closer or further away , depending on how the stream is projected .
such distance variations will cause a distance modulus error of nearly @xmath16 mag , which means that individual stars will be identified with isochrones of the incorrect metallicity .
this would be a significant problem if the stream were a narrow range in metallicity .
however , our unweighted selection of stars over the very broad range of [ fe / h]=[-0.6 , -2.4 ] corresponds to selecting a large rhombus in color - magnitude space that is 3.1 mag high in absolute magnitude .
the upper limit to the distance differences of the stream relative to the body of m31 would blur out our selection of stream stars about 10% or so , but it is only a small dilution of the stream density .
we measure the density perpendicular to the ellipse ridge line and display the result in figure [ fig_stream_width ] .
the stream has a half width at half of the peak density of about 2.5 kpc or a full width of 5 kpc .
the stream width is dominated by the physical width of the stream , but also includes a significant component from the stream location deviating slightly from our assumed elliptical shape .
the stream subtends an angle ranging from about 0.1 radian to 0.05 radian , as seen from the center of m31 , although the stream is physically wider than the cool , but relatively close , streams in our own galaxy .
the location of the centerline of the stream , calculated as the density weighted mean distance from the best fit ellipse , is shown in figure [ fig_stream_center ] .
the error flags are calculated from the statistics of the star counts and confirms the impression of figure [ fig_cutout ] that the stream center varies , sometimes discontinuously , from the best fit ellipse .
the variations are largest on the upper branch of the stream , @xmath17 , where it has large gaps . on the more continuous lower branch
there is a statistically significant offset from the mean centerline at @xmath18 .
the density is measured within a constant width region around the ellipse , @xmath19 ( @xmath20 kpc ) .
we subtract the foreground / background measured at the same angular bin within the two adjacent 0.2 strips on either side of the stream .
the range of metallicity in the stream is shown in table [ tab_slices ] in bins of width 0.2 in [ fe / h ] .
table [ tab_slices ] gives the top of the metallicity bin in column 1 , the mean number of stars per kpc@xmath21 in column 2 .
this table shows that the stream has its maximum density in the [ fe / h]@xmath22 bin . for our measurement of the stream density we will use the entire [ fe / h]=[-0.6,-2.4 ] range ,
uniformly weighted . in principle
somewhat better signal to noise would be obtained if we weighted the stars with the metallicity distribution function .
however , the flat distribution gives the most straightforward stream density measurement .
the stream is broadly distributed in brightness with the mean number of stars [ 21 - 22 , 22 - 23 , 23 - 24 , 2@xmath23(24 - 24.5 ) ] @xmath3-band mag in the [ fe / h]=[-0.6 , -2.4 ] range being [ 0.62 , 0.78 , 0.55 , 0.47 ] per magnitude per kpc@xmath21 over the -70 to -10 angle range .
the luminosity function of the stream is essentially flat over the absolute magnitude range @xmath24 as is usually the case for low metallicity rgb luminosity functions .
physical properties of the stream are derived below .
whether the nw stream is smooth or lumpy is the observational question of primary interest in this paper .
this measurement needs to be handled carefully since the stream over - density is only about 15% of the total nominal rgb star density .
the stream density will vary simply due to the @xmath25 fluctuations in the few thousand stars in each angular bin .
the density measurement is done within the region shown in figure [ fig_cutout ] .
the stream is defined as being a fixed width of the centerline and the local background is measured within the cutout region beyond the stream .
the analysis region of both the stream and local background is shown in figure [ fig_cutout ] .
the newly identified dwarf galaxy andxxvii @xcite , just above the upper branch of the stream in figure [ fig_m31 ] , is masked out to exclude it from the local background correction .
the density along the stream , within a half - width of 0.4 on the sky subtracting the local background in the adjacent 0.2 regions , is shown in figure [ fig_stream_density ] at intervals of 2 in the ellipse angle , @xmath9 .
the errors are computed from the @xmath25 of the star - counts in the stream and in the background region , added in quadrature .
the @xmath26 test estimates the probability that a set of data is consistent with a model , which in our case is a constant density . we need to be aware that sampling and calibration variations could artificially create density variations .
masking takes care of the ccd gaps and the low exposure regions of the images .
the field flattening largely corrects the counts over the rest of the image .
any residual photometric fluctuations will be measured in the control fields to estimate their size .
the @xmath26 statistic applied to the @xmath11 $ ] range indicates that the chance that the stream has a uniform density is less than about @xmath27 .
a binning angle of 2 corresponds to roughly 2 - 4 kpc , depending on location around the ellipse .
we have measured the stream density variations in bins ranging from 1 to 8 finding that the variations are present at all these scales .
since there are easily visible clumps and gaps for @xmath28 including the upper branch of the stream in the @xmath26 analysis would significantly increase @xmath26 , which of course could well be a real description of the larger stream .
we prefer to present the most conservative statistic on the visibly smoothest part of the stream .
we emphasize that even after restricting our statistical analysis to the lower branch , the segment is the longest stream segment available for study so far .
moreover its orbit appears to keep it well away from the potential disturbances of the structure in the disk of m31 .
rrrc & & + @xmath29 & @xmath30 & @xmath31 & @xmath32 + + + 1 & 2.18 & 59 & @xmath33 + 2 & 3.01 & 28 & @xmath34 + 4 & 3.58 & 14 & @xmath35 + + + 1 & 1.17 & 59 & 0.17 + 2 & 1.16 & 28 & 0.25 + 4 & 0.65 & 14 & 0.85 + + + 1 & 1.96 & 29 & 0.002 + 2 & 1.57 & 13 & 0.090 + 4 & 2.38 & 6 & 0.027
+ we have varied virtually all stream and image parameters to check the result that the density variations are truly those of the star stream .
reasonable changes in the path of the ellipse , its width and the local background make little difference , other than slightly reducing the mean density of the stream .
the photometric zero point variations that remain in this version of the data are readily estimated .
the mean background is approximately 12 stars kpc@xmath15 and the photometric variations are expected to produce 5% or smaller fluctuations coherent over scales of 1 degree in the image , or a standard deviation of about 0.6 stars kpc@xmath15 .
the measured density fluctuations have a standard deviation of @xmath12.0 stars kpc@xmath15 , at least three times the size of the variations that the photometric zero points are expected to cause .
a straightforward control sample is to measure the fluctuations at the same location in higher metallicity stars , [ fe / h]@xmath36 - 0.5 , in which the stream is not detected .
figure [ fig_zhi ] shows the mexican top - hat filtered image in the high metallicity stars , demonstrating that the background fluctuations have a significantly different pattern but have a similar amplitude .
the density around the stream is displayed in figure [ fig_density_zhi ] .
the @xmath26 test for the significance of the density variations is reported in table [ tab_chi2 ] .
no significant variations are found .
a second control sample is to simply offset the measurement region away from the stream .
we add 2 to the minor axis and 4 to the major axis and repeat the measurements , adjusting the @xmath9 range to approximately the same distance .
the statistics are reported in table [ tab_chi2 ] finding that the probability that the variations are due to chance to be about 5% . moving the region around find
finds that the @xmath26 values fluctuate a fair bit depending on position although nothing as significant as the nw stream itself ever emerges .
these fluctuations are likely the result of shorter and lower significance star streams that are real entities . whether the upper branch of the stream in fig .
[ fig_m31 ] or [ fig_cutout ] is the extension of the lower branch of stream is unclear in these data .
the upper branch certainly fits within the same ellipse and has a shared metallicity distribution at the precision we can measure it .
however , the upper branch is less a stream than three dominant segments separated by gaps of comparable size .
it is of great interest to note that the segments are tilted with respect to the general path of the stream .
simulations show that tilted segments are some of the dominant structures frequently seen in chopped streams , as first noted in @xcite , and as we show below .
the total luminosity of the lower branch of the stream is found by summing the over - densities in one magnitude bins within a range of [ -70,-10 ] ecliptic angle , to find a stream segment luminosity of @xmath37 to @xmath38 mag .
the luminosity is the sum over the rgb stars alone . to estimate the total luminosity we use the cumulative stellar luminosity function of a stellar population of similar age and metallicity , for which we chose the globular cluster m12 .
m12 is an intermediate metallicity cluster with a well studied stellar luminosity function @xcite . using their distance modulus of 14.0 mag and their i band luminosity function we find about
58% of the light above @xmath39 mag .
therefore the corrected total luminosity is @xmath40 .
the mean surface brightness is @xmath41 .
if we assume the stream is a uniform density cylinder of 5 kpc radius and a stellar mass - to - light ratio of , say , 3 in the @xmath3-band , then the volume stellar mass density is @xmath42 .
this can be compared to the mass density in the dark matter halo at say 90 kpc , which for our adopted halo parameters is @xmath43 , that is , about a factor of about 240 higher than the stellar stream mass density . at 30 kpc
the ratio is 3600 times more background halo dark matter than stream stellar mass .
if dark matter from the progenitor is mixed into the stream then the mass density of the stream would be proportionally increased , lowering the local mass ratios but the stream is unlikely to be significantly self - gravitating anywhere .
@xcite have raised the possibility that jeans instabilities can be the source of of density variations in the stream .
the jeans length in the stream , @xmath44 evaluated with the stellar mass density and assuming an internal velocity dispersion of 10 is about 400 kpc . on the other hand
if the velocity dispersion were as low as 1 then the formal jeans length would be 40 kpc , comparable to the longest scale of density variations in the stream .
very low velocity dispersions are expected for slowly tidally stripped globular clusters .
the presence of several globular clusters along the course of the stream @xcite and the possibility that andxxvii @xcite is the source of the stream suggests that the internal velocity dispersion is at the level that a dwarf galaxy would create , @xmath45 or more .
the resulting jeans length is then roughly 120 kpc , the size of the stream .
it is important to note that the stability analysis is far from complete .
first , the disrupted object may have its own dark matter which has identical kinematics to the visible stars and would increase the mass density .
second , the stability analysis needs to account for the very strong tidal field of the m31 dark halo .
strong tides generally act to suppress much instability , particularly in largely radially oriented streams , which is likely the case for the nw stream .
another mechanism that generates clumps in tidal streams is simply the pileup of stars at low velocity points of their epicyclic orbits @xcite .
the lumps appear at @xmath46 times the tidal radius .
since no progenitor object for the stream has been confidently identified , the tidal radius is very uncertain .
however if we take a low luminosity dwarf galaxy of stellar plus a relatively low dark mass for a total progenitor dwarf mass of @xmath47 , then at a galactic radius of @xmath48 kpc , the tidal radius , @xmath49 , is approximately 3.2 kpc .
the epicycle pileup separation will then be at a separation of 120 kpc . that scale compares to the size of the stream .
the stellar mass in the stream alone , @xmath50 for a fairly minimal assumed mass - to - light ratio of 3 , gives about 40 kpc epicyclic clumping scale , which is at the upper end of range of what we see , but still leaves the smaller scale variations to be explained .
we conclude that epicyclic pile - up is not a significant clumping factor for the nw stream .
if the progenitor system had large internal density variations in a highly order velocity field they could be fed out into the stream to create lumps .
however , this idea runs into trouble with our astrophysical knowledge of low mass galaxies .
dwarf spheroidal galaxies are comprised largely of stars on randomly oriented orbits and have essentially no substructure , which when tidally disrupted would lead to a smooth stream . dwarf irregular galaxies do have substantial substructure in their gas and young stellar populations .
the visible luminosity ( corrected for light below our observational limit ) in the lower branch of the stream is only @xmath51 .
dwarfs with such low luminosities usually have comparable rotational and random velocities which would quickly blur out stellar structures deposited into a stream . given that @xcite associate 3 or 4 globular clusters with the nw stream , which would lead to a unusually high ratio of globular cluster luminosity to progenitor luminosity , it it possible that the nw stream is only part of a larger stream that has yet to be clearly identified .
the structures within star forming dwarf galaxies are often low mass stellar associations and clusters that fairly quickly dissolve as a result of stellar mass loss and their own internal dynamics .
spiral patterns are at best very chaotic in low mass dwarf irregulars .
our assessment is that although carefully sub - structured young progenitor disk system could produce some of the stream lumpiness , appropriate astrophysical systems do not exist .
star streams normally originate from the tidal dissolution of either dwarf galaxies or globular clusters . in both cases tidal fields
pull off outer stars or sometimes dissolve the entire system and distribute the stars in nearby orbits around the galaxy .
the effect of adding randomly orbiting dark matter sub - halos on highly idealized star streams has been studied previously @xcite , finding that the streams were folded , chopped and heated to about 30 over a hubble time . for the examination of the expected degree of lumpiness of star streams we need a more realistic model stream .
the association of globular clusters with the nw stream @xcite indicates that a dwarf galaxy , likely andxxvii , is the source of the stream .
the study of the effects of dark matter sub - halos on the stream is inherently statistical since we are unsure to what degree the visible stars and gas give a complete census of the sub - halos , as highlighted in the celebrated missing satellite problem @xcite .
a detailed match to the orbit of the nw stream is not required for this very basic assessment of the action of sub - halos on a stream .
the details of the stream creation process have relatively little to do with the subsequent sub - halo interactions with the stream .
our test particle stream originates from particles orbiting in a constant mass plummer sphere to which particles in the core region are gradually given velocities that boost them out into the region where tides can carry them away .
future studies will report a full suite of self - gravitating simulations .
the simulation uses a galactic potential with a spherical nfw halo potential scaled to the results of the aquarius simulation @xcite .
the rotation curve of m31 is somewhat higher than the milky - way , but beyond 30 kpc mass modeling finds that the two galaxies are very similar @xcite . for simplicity and comparability
we therefore adopt the same disk - bulge potential for the disk and bulge as @xcite which has a miyamoto - nagai disk potential @xcite and a nuclear bulge modeled with a plummer sphere .
the sub - halos are modeled as a collection of plummer spheres following the distribution in masses , orbits and internal structure of the @xcite simulations .
the star stream particles are evolved in the combined potentials of the background galaxy and orbiting sub - halos .
we use time steps of 0.14 myr , or about @xmath52 in a hubble time , to ensure that the quickly moving particles are integrated properly over the small scale potentials of the sub - halos .
the simulations start with the plummer sphere @xmath53 progenitor located at @xmath54 100 kpc from the center in the plane of the disk , with @xmath55 .
the initial velocity is chosen to be entirely tangential perpendicular to the plane of the disk with an angular momentum of about 50% of the circular value at that radius . in the absence of sub - halos
there are no density variations in the simulated star streams .
m31 contains nearly 30 known dwarf galaxies . to what degree
could these dwarfs , along with their dark - halos , cause the density variations we are measuring ?
for instance , what is the effect of the heaviest 5 sub - halos alone ?
figure [ fig_nsub_xy ] shows the @xmath56 plane projection of the stream in the presence of 0 , 5 , 100 , and 1000 sub - halos as drawn from the @xmath57 mass distribution . including the 5 most massive sub - halos produces only large scale distortions of the stream relative to the smooth halo
. with 100 sub - halos we see significant small scale variations and going to 1000 sub - halos adds yet more small scale structure .
figure [ fig_nsub_den ] shows the density variations in angular bins assuming that the stream has a constant width , which is close to being equivalent to the m31 measurements shown in fig .
[ fig_stream_density ] . in this particular simulation
the progenitor remains bound , so there is a density peak near @xmath58 .
away from the progenitor there is a clear trend of increasing disturbance in the density variations as the number of sub - halos increases .
neither a smooth halo nor one with only the five most massive sub - halos , which roughly mimics the most massive dwarf galaxies , produce anywhere near the amount of substructure we see .
one hundred sub - halos do a reasonable job in the outer parts of the tail but do not do much damage to the inner part of the tail .
once we get to one thousand sub - halos we star seeing gaps and clumps essentially everywhere .
deviations from a mean centerline are also a signal of sub - halo interactions .
again the hundred halo simulation produces a few angled clumps but they are much more clearly present with the thousand halo simulation .
we will report on the statistics of a much larger set of simulations elsewhere .
in this illustrative study we find that of order of one thousand sub - halos are required to produce all the small scale structure on the scales seen in the m31 nw stream from an initially smooth stream .
@xcite report a more complete dynamical analysis focused on streams at smaller radii , but the dynamical outcomes they describe are quite generally applicable .
they also firmly conclude that a large population of dark matter sub - halos will induce gaps and clumps in streams .
a larger statistical study will be reported elsewhere .
a spectacular star stream in our own galaxy is the pal 5 stream @xcite which is kinematically cold and narrow @xcite . at a distance of 23.2 kpc
the visible part is only 4 kpc long .
it is very narrow , about 0.4 kpc , roughly 1/5 of the width of the m31 nw stream segments .
the source of the stream is the outer halo globular cluster pal 5 with the stream having the highest surface density at the location of its progenitor .
the density analysis of @xcite finds that there are density variations of the pal 5 stream above the shot noise in the northern part of the stream , but not the southern part .
the significance of the variation is not given , but there are 9 of 25 points more than @xmath59 from the mean trend over about 1 kpc of length .
this certainly constitutes the detection of a significant lump in the currently available 4 kpc length of the pal 5 stream , but on a smaller scale .
@xcite were inspired by the narrow velocity and geometric distribution of the 75 carbon stars of the sagittarius stream @xcite to consider the lumpiness of the milky way halo .
they develop a statistic based on the fifth through tenth fourier terms to compare smooth halo models to models with up to 256 sub - halos , being careful to consider models that allow for the main dwarf galaxies . when applied to the available sagittarius data they find that `` the degree of scattering is entirely consistent with debris perturbed by the lmc alone . ''
if applied to the 100 kpc nw m31 stream , the tenth fourier term would be a @xmath9=36 variation , which is roughly comparable to a 16 binning where there still is real structure relative to a constant , but only on yet larger scales .
@xcite considered kinematic data for the sagittarius stream in our galaxy and concluded that the halo could not be very lumpy , however we note that our simulations show that sub - halos tend to chop " streams leaving locally cold remnants with pieces at the same radius offset in velocity , which may be compatible with the @xcite figure 2 .
kinematic tests will become very powerful in testing for substructure once large velocity samples are available .
the 100 kpc nw stream in m31 is a coherent geometric structure whose orbit is well clear of the body of m31 making it a near ideal testing ground for the presence or absence of thousands of dark sub - halos predicted in lcdm n - body simulations .
although the stream appears to be nearly a half ellipse over a common range of metallicities the upper branch of the stream is less well defined and has a number of clearly visible gaps .
the lower branch is nearly complete and provides a much more conservative test for the presence of sub - halos .
the main result of this paper is that the stream has highly significant density variations on virtually all scales from 2 kpc , up to about 20 kpc .
the variation of density around the mean has a very low probability of being a chance statistical fluctuation , less than @xmath0 .
we have been careful to measure the lumpiness relative to an averaged local background and masked out gaps and regions where the photometric uncertainties add significantly to the variations . as a control sample
we take a higher metallicity set of stars at exactly the same location as the stream , which finds no significant density variations .
relative to other known cool star streams , the m31 stream stands apart for its length , distance from the disturbing effects of the m31 disk , a variety of scales of substructure and its high statistical significance .
it is interesting to note that m31 appears to have only the one well defined long stellar stream at large radius .
the nw stream has quite a low total luminosity so it should not particularly stand out relative to other stellar streams that were created at large radius over the buildup of the m31 dark halo .
several other stream fragments are visible in the full field map @xcite but none as long or coherent as the nw stream . if one assumed that other long streams likely formed over the lifetime of the halo then the absence of others at the present time indirectly suggests that they have been broken up by sub - halos @xcite .
the measured nw stream lumpiness essentially rules out the possibility that it is a low mass star stream orbiting in a smooth galactic potential of a disk plus bulge plus dark halo .
conversely , the nw stream density variations are compatible with the level of density changes that a large population of dark matter sub - halos induce .
the details of the density variation are sensitive to the statistical distribution of the sub - structure but are unlikely to be specifically modeled for a single stream . as more streams are observed statistical modeling of the degree of substructure present should become possible .
the present study is statistically consistent with a highly sub - structured dark halo , although we caution that the density data alone for a single stream is not a conclusive proof that halos are as sub - structured as lcdm simulations predict .
overall , the density variations of the nw stream are strong circumstantial evidence that the predicted thousands of dark matter sub - halos are present in m31 s dark halo . | the pan andromeda archeological survey ( pandas ) cfht megaprime survey of the m31-m33 system has found a star stream which extends about 120 kpc nw from the center of m31 .
the great length of the stream , and the likelihood that it does not significantly intersect the disk of m31 , means that it is unusually well suited for a measurement of stream gaps and clumps along its length as a test for the predicted thousands of dark matter sub - halos .
the main result of this paper is that the density of the stream varies between zero and about three times the mean along its length on scales of 2 to 20 kpc .
the probability that the variations are random fluctuations in the star density is less than @xmath0 . as a control sample
we search for density variations at precisely the same location in stars with metallicity higher than the stream , [ fe / h]=[0 , -0.5 ] and find no variations above the expected shot noise .
the lumpiness of the stream is not compatible with a low mass star stream in a smooth galactic potential , nor is it readily compatible with the disturbance caused by the visible m31 satellite galaxies .
the stream s density variations appear to be consistent with the effects of a large population of steep mass function dark matter sub - halos , such as found in lcdm simulations , acting on an approximately 10 gyr old star stream .
the effects of a single set of halo substructure realizations are shown for illustration , reserving a statistical comparison for another study . |
assembling a complex quantum optical information processor requires precise knowledge of the properties of each of its components , i.e. , the ability to predict the effect of the components on an arbitrary input state .
this gives rise to a quantum version of the famous black box problem
, which is addressed by means of _ quantum process tomography _ ( qpt ) @xcite . in qpt ,
a set of probe states is sent into the black box ( here an unknown completely - positive , linear quantum process @xmath0 over the set of bounded operators @xmath1 on a hilbert space @xmath2 ) and the output states are measured . from the effect of the process on the probe
states it is possible to predict its effect on any other state within the same hilbert space .
qpt exploits linearity of quantum process over its density operators .
if the effect of the process @xmath3 is known for a set of density operators @xmath4 , its effect on any linear combination @xmath5 equals @xmath6 . therefore ,
if @xmath4 forms a spanning set within the space @xmath7 of linear operators over a particular hilbert space @xmath2 , knowledge of @xmath8 is sufficient to extract complete information about the quantum process .
however , practical implementations of qpt become demanding especially for systems with large hilbert spaces . for @xmath9 , @xmath10 , which implies that at least @xmath11 unknown operators @xmath12 , each with @xmath11 unknown parameters , must be estimated .
this procedure requires preparation of at least @xmath13 states , subjecting each to the unknown process @xmath0 , and determining each element of @xmath14 through measurement ( each with @xmath11 unknown elements ) , thereby inferring an overall number of @xmath15 parameters .
furthermore , in order to build up sufficient statistics for reliable estimates of the output states , each measurement should be performed many times on multiple copies of the inputs .
thus , a large number of realizations and measurements are required for complete tomography of @xmath0 .
an additional complication , especially for qpt of quantum _ optical _ processes , is associated with preparation of the probe states .
typical optical qpt implementations deal with systems consisting of one or more dual - rail qubits @xcite , which implies that the probe states are highly nonclassical , hence difficult to generate .
these difficulties have been partially alleviated in the recently proposed scheme of coherent - state quantum process tomography " ( csqpt ) @xcite .
this scheme is based on the observation that the density operator @xmath16 of a generic quantum state of every electromagnetic mode can be expressed in the _ glauber - sudarshan representation _
@xcite , @xmath17 where @xmath18 is a quasi - probability distribution referred to as the quantum state s @xmath19-function and integrated over the entire complex plane @xcite .
linearity hence implies that measuring @xmath20 i.e. , determining the effect of the unknown process on _ all _ coherent states enables a prediction of its effect upon any generic state @xmath16 according to @xmath21 the implementation of csqpt is advantageous because ( i ) coherent states are readily available from lasers , ( ii ) coherent states of different amplitudes and phases can be produced without changing the layout of the experimental apparatus , and ( iii ) output - state characterization can be performed using optical homodyne tomography @xcite , which obviates the need for postselection and provides full information about the process in question .
moreover , csqpt has been tested experimentally on simple single - mode processes , such as the identity , attenuation , and phase shift operations @xcite .
furthermore , csqpt has been used to characterize quantum memory for light based on electromagnetically - induced transparency @xcite . an apparent obstacle to csqpt , however , is that the @xmath19 function for many nonclassical optical states exists only in terms of a highly singular generalized function @xcite .
a remedy therefor is provided by klauder s theorem @xcite , which states that any trace - class operator @xmath16 can be approximated , to arbitrary accuracy , by a bounded operator @xmath22 whose glauber - sudarshan function @xmath23 is in the schwartz class @xcite , so integration can be performed .
the klauder approximation can be constructed by low - pass filtering of the @xmath19 function , i.e. , by multiplying its fourier transform with an appropriate regularizing function equal to @xmath24 over a square domain of size @xmath25 and rapidly dropping to zero outside this domain .
@xcite employs this method to implement csqpt .
practical implementation of klauder s procedure is however complicated , because it requires finding the characteristic function of the input state and subsequently its regularized @xmath19 function .
this function features high - frequency , high - amplitude oscillations that limit the precision in calculating the output state .
furthermore , klauder s approximation is ambiguous in the choice of the particular filtering function as well as the cutoff parameter @xmath26 @xcite .
here we improve csqpt to overcome the above problems .
specifically , we develop a new method for csqpt that eliminates the explicit use of the glauber - sudarshan representation and thus removes the inherent ambiguity associated with employing klauder s approximation for csqpt . in sec .
[ subsec : csqpt - formalism ] , we obtain an expression for the process tensor in the fock ( photon number ) basis that can be directly calculated from the experimental data . using this tensor ,
the process output for an arbitrary input can be calculated by simple matrix multiplication rather than requiring integration and high - frequency cut - offs . in this way ,
transformations between the fock and glauber - sudarshan representations , which were necessary in ref .
@xcite , can be sidestepped .
using our new approach , we easily extend csqpt from its restrictive single - mode applicability to multi - mode processes and even to non - trace - preserving conditional processes .
these extensions are particularly relevant for quantum information processing circuits , whose basic components are inherently multi - mode and conditional @xcite .
process tomography is successful if , for every input state , the estimate for the process output closely approximates the actual process output state , and the worst - case error of this estimate , given by a distance between the actual and estimated process outputs , is less than a given tolerance . for states over infinite - dimensional hilbert spaces ,
this concept of error is however not meaningful because the finiteness of sampling implies that the process is necessarily under - sampled , hence can not be determined with bounded error .
instead we could consider the process estimation restricted to a finite - dimensional _ subset _ of @xmath27 .
this version of process tomography can always be successful with a sufficiently large amount of sampling . of particular practical interest
is the subspace @xmath28 defined by an energy cut - off , i.e. , estimating the process without accessing any information about its high - energy behavior .
this restriction is naturally consistent with our choice to work in the fock basis , because then the resulting process tensor is of finite size and with many practical settings ( e.g. quantum - information processing with photonic qubits ) . in sec .
[ subsec : erroranalysis ] , we provide process error estimates for several input state subsets that extend beyond @xmath29 . many interesting processes are phase symmetric ; that is , an optical phase shift of the input state results in the same phase shift of the output .
this property dramatically simplifies the experiment because one needs to collect data only for coherent states whose amplitudes lie on the real axis rather than the entire complex plane .
this prompts us to discuss , in sec .
[ sec : experimentalapplication ] , how to obtain the process tensor for phase - symmetric processes , which we test on the experimental data from ref .
next , in sec .
[ sec : examples ] , we illustrate our method by analytically deriving the superoperators for certain fundamental quantum optical processes using the fock basis .
the paper is concluded in sec .
[ conc ] and is supplemented with two appendices .
we study general quantum optical processes @xmath0 acting on quantum states of light and begin with the simplest case for which only a single electromagnetic field mode is involved .
an arbitrary quantum state @xmath16 can be expressed in the fock representation as @xmath30 subjecting this state to an unknown process @xmath0 , and imposing linearity , yields @xmath31 where @xmath32 is a rank@xmath33 tensor , hereafter referred to as the process tensor " ( superoperator ) .
thus , by expressing input and output states in the fock basis , a quantum process can be uniquely represented and characterized by its rank-@xmath33 tensor , which relates the matrix elements of the output and input states according to @xmath34_{jk}=\sum_{m , n\in\mathbb{n}_0 } \mathcal{e}^{mn}_{jk}\rho_{mn } , \ ] ] where @xmath35 .
below we show how to estimate process tensor elements @xmath36 for @xmath37 over a finite domain . because @xmath38 is in the schwartz class ,
the glauber - sudarshan @xmath19 representation @xmath39 is guaranteed to exist for any operator @xmath40 ( @xmath41 ) @xcite .
the @xmath19 function is @xmath42 for @xmath43 and @xmath44 and its complex conjugate @xmath45 treated as independent variables , and @xmath46 . by inserting representation ( [ nm- ] ) into eq .
( [ superoperator ] ) , and exploiting linearity of the process , we obtain the process tensor @xmath47 this expression can be simplified by using eq .
( [ tempered - distribution ] ) and performing integration by parts : @xmath48\nonumber\\ & = \frac{1}{\sqrt{m!n ! } } \partial_{\alpha}^m \partial_{\bar{\alpha}}^{n}\bigl [ e^{|\alpha|^2}\langle j\rvert \varrho_{\mbox{\tiny $ \mathcal{e}$}}(\alpha ) \lvert k\rangle \bigr]\big|_{\alpha=0 } .
\label{sup - op}\end{aligned}\ ] ] thus we have eliminated the need to make use of the glauber - sudarshan representation for quantum states .
the process tensor is found by taking partial derivatives ( with respect to @xmath44 and @xmath45 ) of the matrix elements of @xmath49 , which are estimated from experimental data and evaluated at @xmath50 .
the mathematical procedure defined by eq .
( [ sup - op ] ) is simpler and computationally faster ( see sec . [
sec : experimentalapplication ] ) than employing eq .
( [ sup ] ) with a regularized version of @xmath51 replacing the tempered distribution @xmath52 described in refs .
equation ( [ sup - op ] ) has been used to determine the fidelity of quantum teleportation of a single - rail optical qubit based on measurements performed on coherent states ( see supplementary material in ref .
@xcite ) .
generalization to the multi - mode case is straightforward . in the @xmath53-mode case ,
let us introduce the notation @xmath54 ( with @xmath55 ) for multi - mode fock states and @xmath56 ( with @xmath57 ) for multi - mode coherent states .
then the matrix elements of the output and input states with respect to the fock basis are related to one another by the rank@xmath58 tensor @xmath34_{\mbox{\scriptsize\boldmath$j$}\mbox{\scriptsize\boldmath$k$}}\equiv\langle \mbox{\boldmath$j$}|\mathcal e(\rho)|\mbox{\boldmath$k$}\rangle = \sum_{\mbox{\scriptsize\boldmath$m$},\mbox{\scriptsize\boldmath$n$}\in\mathbb{n}_0^m } \mathcal{e}^{\mbox{\scriptsize\boldmath$m$}\mbox{\scriptsize\boldmath$n$}}_{\mbox{\scriptsize\boldmath$j$}\mbox{\scriptsize\boldmath$k$}}\rho_{\mbox{\scriptsize\boldmath$m$}\mbox{\scriptsize\boldmath$n$ } } , \ ] ] where @xmath59 similarly to the single - mode case , we employ the glauber - sudarshan @xmath19 representation for the multi - mode operator @xmath60 , with the overall @xmath19 function being a product of the @xmath19 functions for the constituent modes : @xmath61 multiple integration by parts yields @xmath62 where @xmath63 equations ( [ sup - op ] ) and ( [ e2 ] ) complete our coherent - state tomography formalism and show that coherent states provide a complete set of probe states for characterizing quantum optical processes , insofar as the expression for @xmath64 completely determines the process tensor .
the above formalism is not restricted to trace - preserving quantum processes .
indeed , trace preservation was not required in the derivation of our results .
thus , our method is applicable to all quantum optical processes that are mathematically described by completely - positive maps , but may be trace - preserving , trace - reducing or even trace - increasing .
trace - nonpreserving quantum processes are either _ conditional processes _ or part of a larger process @xmath65 , which is trace - preserving as a whole , but whose components @xmath66 and @xmath67 may increase or decrease the trace , respectively .
a conditional process is a process that is conditioned on a certain probabilistic event ; it may be heralded if the event is observed .
one of the most notable examples of such a process is a probabilistic conditional - not gate ( cnot ) , which forms the basis for the knill - laflamme - milburn linear - optical quantum computing scheme @xcite .
other examples are photon - addition and photon - subtraction processes , whose superoperators are derived in sec .
[ sec : examples ] . in experimental csqpt , states @xmath49 are determined using homodyne tomography @xcite .
it is important to remember , however , that this procedure reconstructs a density matrix normalized to unity trace : @xmath68 $ ] . when measuring non - trace - preserving processes , one must recover the trace information contained in @xmath49
this is done by measuring the probability @xmath69 $ ] of the process heralding event for all @xmath44 s for which the measurements are performed .
the state to be used in eqs .
( [ sup - op ] ) and ( [ e2 ] ) in place of @xmath49 is then @xmath70 $ ] . an interesting feature of eqs .
( [ sup - op ] ) and ( [ e2 ] ) is that complete information about a quantum optical process is contained in its action on an infinitesimally small compact set of all probe coherent states in the immediate vicinity of the vacuum state . from a mathematical point of view , this feature can be understood by realizing that , for any @xmath71 , the matrix element @xmath72 is an _ entire function _
( see [ app : a ] ) , i.e. , a complex - valued function in the variables @xmath73 that is holomorphic over the whole complex plane , and so is its product with the exponential @xmath74 .
hence , each term @xmath75 is infinitely differentiable over the whole complex plane and is identical to its taylor series expansion in any point of @xmath76 .
moreover , eq . ( [ sup - op ] ) implies that the process tensor is determined by the corresponding taylor coefficients at @xmath50 .
the same conclusion applies to the multi - mode case , in which we deal with entire functions on @xmath77 . as discussed in sec . 1 ,
the incompleteness of the information acquired in the experiment is accommodated in csqpt by evaluating the process tensor over a restricted finite - dimensional subspace @xmath78 of the hilbert space @xmath2 with a fixed maximum number @xmath79 of photons .
the incurred expense is that , through this reduced tomography , only approximate information about the process will be inferred : for a given input state @xmath16 , the predicted output is not @xmath80 , but rather @xmath81 , where @xmath82}\ ] ] is the trace - normalized projection of @xmath16 onto @xmath83 and @xmath84 is the predicted output of the reconstructed process for input state @xmath85 . in eqs . and
, @xmath86 is the projection operator onto @xmath78 .
if the input state @xmath16 is outside @xmath83 , the process output estimation error @xmath87 ( where @xmath88 is the trace norm ) is generally unbounded .
however , it is possible to bound the error for certain practically important classes of input states and processes .
for example , all linear - optical processes involving only linear - optical elements ( interferometers , attenuators , conditional measurements ) do not generate additional photons , and thus map @xmath83 onto itself , so @xmath89 .
for such processes , the error for a particular input @xmath16 can be estimated according to @xmath90 , with the superoperator norm defined as @xmath91 @xcite . if the process is known to be trace - nonincreasing , we have @xmath92 @xcite so the error is bounded from above by @xmath93 note that the above result is not sufficient for evaluating the error for a general process , because this error is given by the deviation of @xmath80 from @xmath94 rather than from @xmath95 [ fig .
[ figure1 ] ] .
a further error bound can be obtained for the class of trace - preserving processes that do not increase the mean energy , acting on a set of input states whose mean energy does not exceed a certain value @xcite .
we illustrate this for a single optical mode @xmath96 with frequency @xmath97 and hamilton operator @xmath98 whose eigenvalues are denoted by @xmath99 .
suppose that the quantum states @xmath16 of interest satisfy @xmath100\le u$ ] . according to ref .
@xcite , if we choose the cutoff dimension @xmath101 such that @xmath102 for some ( small ) @xmath103 , the reconstructed process output errors are bounded from above as @xmath104 where @xmath105 conversely , if we want to achieve a certain upper bound @xmath106 on the error of approximation ( which corresponds to a lower bound on the desired accuracy of the process characterization ) , we first solve eq .
( [ epsilon = f(gamma ) ] ) for @xmath107 , and then find the minimum @xmath108 such that @xmath109 .
any cutoff dimension @xmath110 is then sufficient for our purpose .
for @xmath111 , @xmath112 , which yields @xmath113 this implies that the error of approximation scales as @xmath114 with the cutoff dimension @xmath115 .
for example , in order to achieve a 10% error in eq .
( [ ee " ] ) , we need @xmath116 and thus @xmath117 . for the input mean energy bound corresponding to one photon ( @xmath118 ) , the required cutoff is @xmath119 .
this calculation shows that the above error estimate is extremely conservative .
many practically relevant processes , including the single - mode processes studied in sec .
[ sec : examples ] , exhibit phase invariance .
if two input states are identical up to a shift by an optical phase @xmath120 , the process outputs for these states differ by the same phase shift : @xmath121=e^{i\hat{n}\phi}\mathcal{e}(\rho ) e^{-i\hat{n}\phi}.\ ] ] for such processes , it is convenient to express the probe coherent states in polar coordinates : @xmath122
. specifically , in these coordinates , we have @xcite @xmath123 and accordingly @xmath124\,\bigg|_{r=0}. \label{sup - op - polarform}\ ] ] hence eq . ( [ eq.shift ] )
can be expressed as @xmath125 and the superoperator @xmath0 [ eq . ( [ sup - op - polarform ] ) ] has the following explicit representation : @xmath126\,\bigg|_{r=0 } \delta_{m - j , n - k}.\ ] ] in experimental tomography of phase - invariant processes @xcite , it is sufficient to measure the process output for a discrete set of coherent states @xmath127 on the real axis of the phase space .
the matrix elements of the output states can then be interpolated as polynomial functions @xmath128 where @xmath129 is the degree of the polynomial ( which depends on the dimension of the truncated hilbert space ) and @xmath130 are its coefficients . furthermore , from eq .
( [ convergentseries ] ) together with eq .
( [ supopphaseinv ] ) , it follows that , for phase - symmetric processes , when @xmath131 is even or odd , @xmath132 and its analytic extension to negative values of @xmath133 are even or odd functions of @xmath133 , respectively . by taking into account the symmetric or antisymmetric property of this function
, we have additional information to be used in the interpolation procedure ; the constructed polynomial has to contain only even or odd powers of @xmath133 , respectively . in this way the precision of process estimation from the experimental data is substantially increased . with the knowledge of the coefficients @xmath130 , eq .
( [ supopphaseinv ] ) is further simplified to : @xmath134\bigg|_{r=0 } \delta_{m - j , n - k}\nonumber \\ & = \frac{\sqrt{m!n!}}{(m+n)!}\sum_{s=0}^{\infty}\sum_{l=0}^{q}\frac{\delta_{m+n,2s+l}(m+n)!}{s!}c_l(j , k ) \delta_{m - j , n - k}\nonumber \\ & = \sqrt{m!n!}\sum_{s=0}^{\lfloor(m+n)\rfloor/2}\frac{c_{m+n-2s}(j , k)}{s!}\delta_{m - j , n - k}\,.\end{aligned}\ ] ] the last result is significant in that one can obtain the process tensor directly from the experimentally reconstructed output states through simple summation . moreover , if the dimension of the truncated hilbert space is @xmath135 , from eq .
( [ sup - op2 ] ) it follows that only terms of power @xmath136 of the interpolation polynomial ( [ poly ] ) contribute to the process tensor .
we have tested this procedure on experimental data @xcite and calculated the process tensor in a few microseconds , which is a dramatic improvement in comparison to several hours required for the original procedure @xcite .
in this section , we illustrate our new method by applying it to some fundamental quantum optical processes , whose effects on coherent states are known . specifically , using eqs .
( [ sup - op ] ) or ( [ e2 ] ) , we analytically derive corresponding superoperator tensors @xmath137 in the fock basis .
the results are summarized in table [ p - tab ] . for the identity process ( @xmath138 ) ,
@xmath139 , the matrix elements of the output states are @xmath140 inserting these elements into eq .
( [ sup - op ] ) yields @xmath141 , as expected . for attenuation of light fields ( @xmath142 ) ,
the process s effect on single - mode coherent states is given by @xmath143 , where @xmath144 .
the matrix elements in the fock basis are @xmath145 from eq .
( [ sup - op ] ) , we obtain @xmath146\big|_{\alpha,\bar{\alpha}=0}\nonumber\\ & = \frac{\eta^{j+k}}{\sqrt{m!n!j!k!}}\partial^{m}_{\alpha}\partial^{n}_{\bar{\alpha } } \sum_{l=0}^{\infty}\frac{(1-\eta^{2})^{l}\alpha^{j+l}{\bar{\alpha}}^{k+l}}{l!}\big|_{\alpha,\bar{\alpha}=0}\nonumber\\ & = \sqrt{\frac{m!n!}{j!k!}}\frac{\eta^{j+k}(1-\eta^{2})^{m - j}}{(m - j)!}\delta_{m - j , n - k}\,,\label{eq : tensorforattenuation } \ ] ] which depends explicitly on @xmath147 .
photon subtraction is defined as a process that removes a single photon from the light field , whereas photon addition adds a single photon .
photon subtraction has been used by ourjoumtsev et al .
@xcite to generate optical schrdinger kittens ( coherent superpositions of low - amplitude coherent states ) from squeezed vacuum states for the purpose of quantum information processing .
single - photon - added coherent states can be regarded as the result of the most elementary amplification process of classical light fields by a single quantum of excitation ; being intermediate between single - photon fock states ( fully quantum - mechanical ) and coherent ( classical ) ones , these states have been demonstrated to be suited for the study of smooth transition between the particle - like and the wavelike behavior of light @xcite . here we discuss idealized single - mode photon subtraction and photon addition . both processes are non - trace - preserving .
for example , photon subtraction can be approximately realized @xcite by a highly - transmissive beam splitter , whose reflected mode is directed to a detector and whose transmitted mode constitutes the output , respectively , as illustrated in fig .
[ figure2]a .
any click in a detector implies extraction of photon(s ) from the input mode by the beam splitter .
as the beam splitter has low reflectivity , here single - photon extraction events are more likely than multi - photon events .
an approximate experimental realization of photon addition is illustrated in fig .
[ figure2]b .
the input quantum state @xmath16 enters the signal channel of a parametric down - conversion setup .
provided that detector dark counts are neglected , a photon detection in the idler mode heralds photon addition to the signal mode , which contains the output state of the process .
the effect of photon subtraction ( @xmath148 ) and addition ( @xmath149 ) on coherent states is given by @xmath150 and @xmath151 , respectively , where @xmath96 and @xmath152 are the photon annihilation and photon creation operators of a single mode , respectively .
the matrix elements of the output states in the fock basis are @xmath153 the process tensor is found to be @xmath154 where we have employed eq .
( [ sup - op ] ) .
the unitary evolution according to @xmath155 $ ] for @xmath156 , if applied to coherent states , generates schrdinger cat states ( hereafter denoted as @xmath157 ) @xcite @xmath158 with matrix elements @xmath159.\ ] ] the superoperator tensor for this non - gaussian unitary process obtained via eq .
( [ sup - op ] ) is @xmath160 interestingly , this process does not change the total particle number of any input state .
now let us consider the beam splitter as an example of a two - mode process .
the unitary beam splitter transformation is given by @xcite @xmath161 where @xmath162 is the parameter identifying how the beam splitter transmits or reflects beams .
specifically , its action on coherent state inputs @xmath163 and @xmath164 is given as @xmath165 and @xmath166 being the transmissivity and reflectivity , respectively . by knowing the effect of the process on two - mode coherent states
, we can calculate the corresponding tensor using eq .
( [ e2 ] ) , which yields @xmath167 as an explicit function of @xmath168 and @xmath169 .
another two - mode process of interest is parametric down - conversion ( pdc ) . in pdc ,
a crystal with an appreciably large second - order non - linearity is pumped by a laser field .
each of the pump photons can spontaneously decay into a pair of identical ( degenerate pdc ) or nonidentical photons ( nondegenerate pdc ) .
here we consider a nondegenerate pdc process @xmath170 induced by the transformation @xcite @xmath171 the effect of this unitary process on a two - mode coherent state is given by @xmath172 in [ app : b ] , we derive the process tensor in the fock basis .
the result can be expressed as : @xmath173 with @xmath174 the hypergeometric function , @xmath175 the pochhammer symbol and @xmath176 the gamma function @xcite .
@lllll operation @xmath0 & @xmath177 & process tensor @xmath178 + identity ( @xmath179 ) & @xmath180 & @xmath181 + & & + attenuation ( @xmath182 ) & @xmath183 + & & + photon addition ( @xmath149 ) & @xmath184 & @xmath185 + & & + photon subtraction ( @xmath148 ) & @xmath186 & @xmath187 + & & + cat generation ( @xmath188 ) & @xmath189 & @xmath190 + & @xmath191 & + + & & + beam splitter ( @xmath192 ) & @xmath193 & @xmath194 + & @xmath195 & @xmath196 + & & @xmath197 + & & @xmath198 + & & @xmath199 + & & + + & & + parametric down- & @xmath200 & @xmath201 + conversion ( @xmath202 ) & @xmath203 & @xmath204 + & & @xmath205 + & & @xmath206 + & & @xmath207 + [ p - tab ]
coherent states are easily generated probe states for tomography of unknown quantum - optical processes .
here , we have presented a new , more efficient data processing technique for estimating a quantum process from similar experimental procedure of ref .
the original formulation was based on regularization and filtering of the glauber - sudarshan representations for quantum states , which are cumbersome to implement numerically .
furthermore , ref .
@xcite introduces additional errors associated with regularization of the @xmath19 function .
in contrast , our new method to determine the process superoperator [ eq . ( [ sup - op ] ) or eq .
( [ e2 ] ) ] is mathematically simpler , computationally faster and unique up to the choice of the energy cutoff .
moreover , we presented straightforward generalizations of coherent state quantum process tomography to multi - mode and non - trace - preserving conditional processes .
we have illustrated the new framework through several examples ( summarized in table [ p - tab ] ) .
we have shown that it is straightforward to derive analytically exact and unique closed - form expressions for the superoperators for quantum optical processes whose effect on coherent states is known . for phase - invariant unknown processes , the formula to find the process tensor reduces to a simple summation of coefficients of a polynomial obtained from the experimentally reconstructed output states via interpolation . an interesting consequence implied by our formulation [ in particular , eqs .
( [ sup - op ] ) and ( [ e2 ] ) ] is that complete information about a quantum optical process is entirely captured by its effect on a compact set of all coherent states @xmath208 in the immediate vicinity of the vacuum state .
this is due to the entireness property of the image of processes on coherent states .
it thus appears sufficient to perform tomography experiments only for a range of coherent states whose mean photon number is much smaller than that required for the method of ref .
@xcite ( see the suppl . material therein ) .
however , coherent state quantum process tomography relies on the ability to approximately determine all the derivatives of a function which is obtained by interpolation from measured experimental data .
minimization of errors associated with this calculation imposes a lower bound on the phase space region over which the measurements need to be performed . for the time being , we have provided an evaluation of the error in the process estimation by introducing a truncation of the fock space . for the class of processes respecting a certain energy constraint ( which includes all processes that do not amplify the energy ) , we have determined ( i ) the cutoff dimension that is sufficient in order to achieve a certain degree of approximation accuracy , as well as ( ii ) the upper bound on the error of estimation for a given cutoff dimension .
* acknowledgments : * we acknowledge financial support by nserc , @xmath209core , mitacs , quantumworks and general dynamics canada .
ail is a cifar scholar , and bcs is a cifar fellow .
we would also like to thank connor kupchak for helpful discussions .
according to eq . ( [ sup - op ] ) , by knowing the complex - valued function @xmath210 ( of the variable @xmath211 ) for any @xmath212 and @xmath213 , one can determine the process tensor @xmath214 .
here we show that this function is an _ entire function _
so it can be represented as a power series that converges uniformly on any compact domain . as a completely - positive quantum operation , @xmath0 possesses a kraus decomposition @xmath215 , where @xmath216 and @xmath217 are some kraus operators on @xmath2 ( whose explicit form is not needed for our purpose ) .
hence we can rewrite the matrix elements of the output state as @xmath218 where @xmath219 is the dual or adjoint map @xcite .
the complex - valued function @xmath220 ( referred to as husimi function if @xmath221 is a density operator)where @xmath221 is any bounded operator on @xmath2is an entire function of the two variables @xmath44 and @xmath45 @xcite .
hence the right hand side of eq .
( [ invpro ] ) implies that the function @xmath210 is an entire function . by representing the coherent states in eq .
( [ invpro ] ) in the fock basis and using eq .
( [ superoperator ] ) , we obtain @xmath222 which is a power series of the complex variables @xmath44 and @xmath45 , hence convergent everywhere @xcite .
to obtain the fock representation of the pdc process , we first find the matrix elements of the output states in the fock basis : @xmath223 where @xmath224 and @xmath225 .
employing the relations @xmath226 and the binomial expansion , we have : @xmath227 using @xmath228 and @xmath229 we obtain : @xmath230 from the symmetry between @xmath231 and @xmath232 , and by replacing @xmath233 and @xmath234 by @xmath235 and @xmath236 , respectively , we also find : @xmath237 the fock representation of the superoperator for the pdc process is then given by @xmath238 which can also be expressed in terms of a product of values of the hypergeometric function @xmath239 , as given by eq .
( [ eq : pdc - tensor ] ) .
99 m. abramowitz and i. a. stegun , _ handbook of mathematical functions with formulas , graphs , and mathematical tables _
( dover , new york , 1972 ) .
d. kretschmann , d. schlingemann , and r. f. werner , ieee trans .
theory * 54 * , 1708 ( 2008 ) .
l. mandel and e. wolf , _ optical coherence and quantum optics _ ( cambridge university press , new york , 1995 ) . | we develop an enhanced technique for characterizing quantum optical processes based on probing unknown quantum processes only with coherent states .
our method substantially improves the original proposal [ m. lobino et al . , science * 322 * , 563 ( 2008 ) ] , which uses a filtered glauber - sudarshan decomposition to determine the effect of the process on an arbitrary state .
we introduce a new relation between the action of a general quantum process on coherent state inputs and its action on an arbitrary quantum state .
this relation eliminates the need to invoke the glauber - sudarshan representation for states ; hence it dramatically simplifies the task of process identification and removes a potential source of error .
the new relation also enables straightforward extensions of the method to multi - mode and non - trace - preserving processes .
we illustrate our formalism with several examples , in which we derive analytic representations of several fundamental quantum optical processes in the fock basis .
in particular , we introduce photon - number cutoff as a reasonable physical resource limitation and address resource vs accuracy trade - off in practical applications .
we show that the accuracy of process estimation scales inversely with the square root of photon - number cutoff . |
following the discovery of topological insulators@xcite , intense theoretical efforts have resulted in a good understanding of topological phases of free fermions , including a complete classification of such phases that are stable to disorder@xcite . in these phases
the bulk appears , to local probes , as a rather conventional gapped state .
the surface however is gapless unless one of the symmetries protecting the phase is broken .
in contrast our understanding of topological phases of interacting particles is much less complete .
the fractional quantum hall effect has inspired much work on phases with topological order@xcite .
these states have the remarkable property that the degeneracy of their ground states depends on the topology of the space on which they are defined .
while they are often associated with edge states , additionally , excitations with exotic statistics occur in the bulk .
moreover , these states are characterized by a topological entanglement entropy , implying long range quantum entanglement in the ground state .
it seems appropriate to exclude such phases in a minimal generalization of topological insulators to interacting systems .
we define a short range entangled ( sre ) state as a gapped state with a unique ground state on all closed manifolds i.e. no topological order . in the presence of interactions ,
do new sre phases appear that share the same symmetry , but differ at the level of topology ?
a possible distinction , for example , is the presence of protected states at the boundaries . in this paper
we study the physics of such phases in systems of interacting bosons . for bosons
the non - interacting limit is a simple condensate so that interactions are necessary to stabilize gapped phases .
thus we necessarily need to free ourselves from the crutch of free fermion hamiltonians and band topology upon which most current discussions of topological insulators are based . studying bosonic generaizations of topological insulators
is potentially a useful step toward the harder problem of interaction dominated fermionic topological insulators .
several examples of interacting boson systems exist .
a very natural realization of a strongly interacting boson system is a quantum magnet made up of localized quantum spins on a lattice . in that context
the phases we are interested in may be dubbed topological paramagnets " .
they have a bulk gap , no bulk topological order or fractional quantum numbers , but have protected surface states all in close analogy to electronic topological insulators .
the topological paramagnet should be distinguished from a more familiar ( and more exotic ) paramagnet - the quantum spin liquid state - which has been extensively discussed in the literature .
quantum spin liquid states either have bulk topological order or gapless excitations .
other realizations of strongly interacting bosons are ( of course ) provided by ultracold atoms in optical lattices .
a famous example of a topological paramagnet already exists : the spin-@xmath3 haldane ( or aklt ) chain .
this has a bulk gap , no fractionalization , but has dangling spin-@xmath4 moments at the edge that are protected by symmetry@xcite . using a powerful matrix product representation of gapped states@xcite topological phases in one dimension
are completely classified by the second group cohomology of symmetry group g. in 2d and 3d such rigorous results are not available .
however , it was pointed out by kitaev that 2d sre phases of bosons with chiral edge states are nevertheless possible @xcite .
recently , chen , liu , gu and wen@xcite have proposed that sre topological phases of bosons protected by symmetry , are captured by the higher dimensional cohomology groups .
while this provides an efficient mathematical scheme to enumerate phases , unfortunately their properties are not transparently obtained .
progress on clarifying the physics of these states was subsequently made by several authors .
levin and gu@xcite explicitly studied a specific example in 2d protected by z@xmath5 symmetry . later , 2d sre topological phases were described using a simpler chern - simons approach@xcite , which provides a field theory and explicit edge theories of these states .
it was found that bosons with a conserved global u(1 ) but no other symmetries can display an ` integer quantum hall state ' with quantized hall conductance predicted to be even integer multiples of @xmath6 , i.e. @xmath7 where @xmath8 is the elementary charge of the bosons@xcite .
a simple physical realization of such a phase of two component bosons in the lowest landau level has been provided@xcite .
other physical interpretations have also appeared @xcite .
thus , defining properties of several ` symmetry protected topological ' ( spt ) phases in 1d and 2d are now understood .
however in 3d , though the cohomology classification predicts various spt phases , including bosonic generalizations of topological insulators , their physical properties have not thus far been elucidated .
this will be our primary task .
potentially , our approach could also be used to classify 3d bosonic spt phases , which we leave to future work .
we note however that for all symmetry classes that we considered both with time reversal ( @xmath9 ) and without ( @xmath10 ) we identify candidate phases that exhaust those predicted by the cohomology classification@xcite .
further , we identify a possible topological phase that appears to be outside the classification of ref .. much of our discussion will focus on a theory of the novel surface states of these three dimensional bosonic topological insulators .
we construct effective landau - ginzburg " field theories of these surface states .
we also identify bulk field theories that correctly yield the proposed surface theory .
a key feature of this surface theory is that it does not admit a trivial gapped symmetry preserving surface phase .
the surface either spontaneously breaks symmetry , or if gapped , develops surface topological order ( even though there is no bulk topological order ) .
other more exotic symmetry preserving states with gapless excitations are also possible . in all these cases
the defining global symmetries are implemented in a way not allowed in strictly two dimensional systems . as
a specific example - consider insulating bosonic phases with the symmetries of a topological insulator , i.e. charge conservation and time reversal symmetry ( formally denoted as @xmath11 ) .
then , the surface could break a symmetry , e.g. by forming a surface superfluid .
alternately , the surface can remain insulating , while breaking time reversal symmetry .
the bulk is assumed to retain the symmetry .
then , these surface ordered phases can reflect their special origin by exhibiting features that are forbidden in purely two dimensional phases .
for example , the vortices of the surface superfluid above will be shown , in a precise sense , to be fermionic furthermore , the surface insulator with broken time reversal symmetry will be shown to have a hall conductance of @xmath12 ( in units of @xmath6 ) , in contrast to the 2d integer quantum hall phases of bosons which are only allowed to even integer hall conductance .
equivalently , the surface quantum hall effect may be considered a quantized three dimensional response , the magneto - electric polarizability @xmath0 .
recall , in the context of free fermion topological insulators with broken time reversal on the surface that gaps the surface states , have a quantized magneto - electric effect appears@xcite that is captured by the topological theta term : @xmath13 ( we have set @xmath14 and @xmath15 are applied electric and magnetic fields ) . for fermion
topological insulators @xmath16 @xmath17 , corresponding to a half integer hall effect on the surface .
the @xmath1 ambiguity in @xmath0 corresponds to the fact that one may deposit a fermionic integer quantum hall layer on the surface@xcite . here , for bosonic topological insulators
, @xmath0 is _ only _ defined modulo @xmath18 , and the topological phase corresponds to @xmath19 .
this implies , for example , the domain wall between opposite time reversal symmetry breaking regions on the surface induces a protected mode , the edge state corresponding to the @xmath20 quantized hall effect of bosons .
the @xmath18 ambiguity in @xmath0 corresponds to the fact that one may deposit an integer quantum hall layer of bosons on the surface , which must have an _ even _ integer hall conductance .
a symmetry preserving surface state may be accessed from the superfluid by condensing vortices which transform trivially under the symmetry .
however as the vortex of the surface superfluid has fermionic statistics it can not condense .
this precludes the possibility of a trivial gapped surface insulator .
the fermionic vortices can of course pair and condense .
however as is well known@xcite paired vortex condensation leads to a two dimensional state with topological order ( described in the present context by a deconfined @xmath21 gauge theory ) .
this surface topological order will be shown to realize symmetry in a manner not allowed in strictly two dimensional systems .
exotic gapless surface states that preserve all symmetries are also conceivable .
for example , the gapless surface state may intuitively be viewed as a quantum hall state that fluctuates between @xmath12 .
the theory of such a state is constructed using a network model that captures the quantum phase transition between distinct integer quantum hall states of bosons .
the same approach , when applied to fermionic topological insulators correctly yields the single dirac cone surface state .
the field theory thus obtained of the bosonic model poised at the transition , naturally leads to the required surface theory , which is closely related to the deconfined quantum critical theory@xcite , previously proposed in the context of frustrated quantum magnets . _
3d field theory : _ the general arguments above will be shown to be consistent with the following @xmath22 field theory : @xmath23 where the index @xmath24 refers to boson species , and bosons four currents are represented by @xmath25 while the curl of @xmath26 represents the vortex lines .
the first ` bf ' term@xcite just represents the @xmath1 phase factor of taking a particle around its vortex .
the key topological properties however are determined by the second term , which attaches quantum numbers to vortices . to avoid topological order at the surface , and ensure bosonic excitations , @xmath27 and diagonal entries are even integers . in most cases
we will take two species with @xmath28 . here
@xmath29 is assumed to lead to an equivalent theory that only differs in details of surface termination .
furthermore , time reversal symmetry constrains @xmath30 , the latter being the topological phase .
note , @xmath31 for the internal gauge fields @xmath26 should be distinguished from the @xmath0 for the external electromagnetic field we discussed above .
coupling to an external electromagnetic field allows one to obtain the quantized magneto electric effect discussed above .
related theories have appeared in the context of three dimensional topologically ordered phases and superconductors @xcite where only the first @xmath2 term in eqn.[bfoverview ] appears with a different coefficient . on the other hand
the field theory discussed in ref .
retains only the second term , which leads to gapless excitations in the bulk , as noted in ref . , which differ from the gapped phases of interest here .
thus , it will be important to combine both these terms .
recently , it was proposed that free _ fermion _ topological insulators are captured by similar theories , @xcite with a single component field and the first term of eqn . [ bfoverview ] along with coupling to the external field .
the surface states in such theories were argued to be bosonized dirac fermions@xcite .
while this is an intriguing idea , we point out certain problems ( see appendix [ appendixbfsurface ] ) with the identification of a metallic surface in ref . .
moreover , identification of a bulk fermionic operator is also problematic in this theory .
one approach , taken in a recent paper@xcite , is to regard this as a partial theory , that provides a purely hydrodynamic description that excludes fermionic excitations altogether .
thus , finding a complete effective field theory description for 3d fermionic topological insulators remains an open problem .
we also display continuum field theoretic models in @xmath32 dimension that realize some of the topological phases we describe .
these are obtained as perturbations of non - linear sigma models in the presence of a topological theta term .
the theta term has the effect of endowing topological defect configurations of the continuum fields with non - trivial global quantum numbers .
we show the connection to the topological bf theory and to the theory of protected surface states . _ possible phases with
half quantized surface thermal hall effect : _ interestingly , our general approach and the theory in eqn .
[ bfoverview ] predicts a new 3d spt phase , protected by time reversal symmetry , that is not obtained within the cohomology classification of chen et al.@xcite just as the quantized hall effect of @xmath33 sre bosons immediately constrained the physics in 3d , an analogous argument for thermal hall conductance can be made . in @xmath34 , this is quantized to 8 times the quantum of thermal conductance : @xmath35 , for a 2d sre phase of bosons .
a realization of @xmath36 is the kitaev @xmath37 state@xcite , with 8 chiral bosons at the edge .
therefore , a three dimensional phase protected by time reversal symmetry can be conceived , on whose surface a domain wall between opposite t breaking regions will host 8 chiral boson modes and is described by eqn .
[ bfoverview ] , with the 8 dimensional @xmath38 matrix , discussed in .
this putative phase lies outside the cohomology classification of chen et al.@xcite , which reports a single non - trivial topological phase with this symmetry , for which a different candidate , with @xmath28 is identified below .
as this paper is long and discusses bosonic spt phases from several points of view it is helpful to provide an overview .
we seek an effective landau - ginzburg " description of the surface of a 3d spt phase .
we employ various approaches , that , satisfyingly , lead to consistent conclusions . below we motivate these different approaches as preparation for the rest of the paper .
a key physical requirement of the surface theory is that there be no trivial gapped symmetry preserving surface phase .
this feature is reminiscent of the lieb - schutlz - mattis ( lsm ) theorem @xcite and its generalization@xcite to states of bosonic systems at a fractional filling on clean @xmath39 lattices . indeed in both cases either a symmetry must be broken or there is topological order , or more exotic long range entanglement possibly with protected gapless excitations .
there is however one important difference . for the surface states discussed in this paper , the trivial insulating phase does not exist _ even _ in the presence of disorder that breaks translation symmetry .
they are protected by internal symmetries rather than by lattice translation symmetry ( as is the case for lsm ) .
nevertheless we will exploit insights from existing effective field theories of clean lattice bosons that build in the lsm restrictions to construct the desired surface theory of the bosonic topological insulator .
we begin with a physical discussion of the constraints on quantized electrical and thermal response in sre 3d bosonic insulators , imposed by our knowledge of 2d sre phases ( * section * [ sec : gentrnsprt ] ) .
in particular we argue that the quantized magnetoelectric coupling for bosonic topological insulators ( @xmath19 ) is double that of the fermionic case ( @xmath16 ) , which parallels the doubling of hall conductance@xcite for 2d integer states of bosons compared to fermions .
next , we borrow an approach that is useful in classifying spt phases of 2d bosons . there ,
the edge states are typically described by a 1d luttinger liquid theory .
what makes them special is that symmetry acts on these edge states in a way that is impossible to realize in a purely 1d system@xcite .
this ensures that all perturbations that lead to a gapped phase also break symmetry . by analogy , to discuss a 3d spt phase we model the 2d surface by a conventional 2d theory of bosons ( e.g. an xy model ) .
however , we identify symmetry operations which are forbidden in a conventional 2d system , by demanding the 2d system can never enter a trivial gapped phase .
u(1 ) ) and time reversal symmetry ( @xmath40 ) .
the bulk is insulating and corresponds to a condensate of vortices of both species ( shown as black and white loops ) . in the topological phase
, the vortex line of one species that ends on a surface carries half charge of the other species .
such a surface , it may be argued , does not have a trivial gapped phase , where the symmetries are preserved .
pictorially , the vortex lines may be viewed in 1d haldane phases , with half charged end states . ] as a useful device consider an enlarged @xmath41 global symmetry , along with time reversal symmetry @xmath42 ( technically @xmath43 \rtimes z_2^t$ ] ) and eventually break it down to the symmetry of interest .
physically , this corresponds to separate conservation of two species of bosons species 1,2 . consider the surface of a 3d spt phase , with a broken symmetry eg . a condensate of species 1 bosons , that breaks the first u(1 ) symmetry .
this superfluid must also be unusual - in that it can not be connected to a fully symmetric insulator .
guidance from effective field theories of clean 2d lattice bosons at fractional filling suggests thinking in terms of the vortices of the superfluid .
since the insulating state is obtained by condensing vortices , we can ask what vortex properties provide the required obstruction . unlike a particle
, a vortex is a non - local object , and can transform projectively under the remaining @xmath44 and @xmath42 symmetry .
vortices with projectively realized symmetries provide an obstruction to realizing a trivial insulator , and can describe the surface of an spt phase . in this example , the projective transformation requires the two species of vortices ( @xmath45 ) carry half charge of species 2 , which are exchanged by time reversal symmetry .
a minimal theory of these vortices is obtained by representing the density of the condensed species 1 by the curl of a vector potential @xmath46 , which couple minimally to the vortices .
@xmath47 where we have inserted external electromagnetic fields @xmath48 that couple to the two conserved currents .
to obtain an insulating surface we must condense the vortices , but this inevitably breaks symmetry . since the vortices transform into one another under time reversal , one can not condense one species and not the other .
condensing them both however implies the breaking the other u(1 ) symmetry , since the vortices carry charge of species 2 . a third option is to condense pairs of vortices , but this results in z@xmath5 topological order@xcite .
this points to a schematic picture of the spt phase as shown in figure [ fig1 ] .
a bulk insulator may be regarded as a condensate of vortex loops . however , the special feature here is that the condensed vortex lines carry a half charge at their ends when they intersect the surface .
the vortex lines , viewed as one dimensional objects , are in a topological state analogous to the haldane phase , and thus carry edge states . note that if a single bosonic vortex species was present at low energies , we could perform a duality back to the usual boson phase variables .
however , having multiple vortex fields demands a dual description like the one above .
in fact an identical theory appeared in the discussion of deconfined quantum critical points@xcite , where however the symmetries included both internal and spatial symmetries . here
all symmetries are internal and hence are restricted to the surface of a 3d topological phase .
let us consider in more detail the insulating surface obtained on breaking time reversal symmetry .
this is obtained by condensing just one species of vortex say @xmath49 , in eqn.[deccpsurface ] which forces @xmath50 . when substituted into the action above this
yields the electromagnetic response @xmath51 .
if the separate species are now identified with a single conserved charge , we can set @xmath52 which yields @xmath53 on the surface , indicating a magneto electric response @xmath19 , as advertised . in the absence of time reversal breaking , the surface may be assumed to fluctuate between @xmath54 .
we provide an alternate derivation of the same surface theory eqn.[deccpsurface ] by modeling this by a network model , poised at the transition point between two bosonic integer quantum hall states .
this is analogous to obtaining the surface state of the 3d fermionic topological insulators , the single dirac cone , by a network model at the critical point between integer quantum hall plateaus@xcite in a clean system . in both cases ,
the time reversal symmetry automatically tunes the system to criticality . for the bosonic spt surface ,
resulting theory is an o(4 ) non - linear sigma model with a topological term as in the euclidean lagrangian below : @xmath55\end{aligned}\ ] ] where the o(4 ) vector has been written in terms of an su(2 ) matrix @xmath56 .
reassuringly , this has been argued in ref . to be equivalent to eqn .
[ deccpsurface ] once appropriate anisotropies are introduced .
these two descriptions of surface properties of spt phases are discussed in * section * [ sec : surface ] along with the connection between them .
we also emphasize the properties of a gapped symmetric surface state obtained by condensing paired vortices .
this state has surface topological order described by a deconfined @xmath21 gauge theory ( even though the bulk has no such order ) .
this state provides a particularly simple perspective on why a trivial gapped symmetric phase is forbidden at the surface .
indeed we will show that bosonic topological quasiparticles of this state carry fractional quantum numbers . destroying the topological order by condensing one of these quasiparticles necessarily breaks a symmetry .
not surprisingly we show that the implementation of symmetry in this surface topological ordered state is distinct from what is allowed in strict 2d systems . in the remaining sections
bulk theories that are consistent with the surface descriptions above are discussed in * section * [ sec:3d ] .
spt phases in quantum magnets - the ones we dubbed topological paramagnets - protected by other symmetries such as time reversal ( @xmath40 ) , or time reversal along with one component of spin rotation ( @xmath57 ) are described in * section * [ sec : othersymm ] .
an example without time reversal symmetry ( @xmath10 ) is discussed in the appendix .
finally , we discuss a new topological phase that is predicted by this approach , a 3d extension of kitaev s e@xmath58 state , along with miscellaneous comments in * section * [ sec : e8 ] .
we begin by considering a system of interacting bosons in @xmath59 space dimensions in the presence of time reversal and particle number conservation symmetries . specifically let us consider the situation where the boson field @xmath60 carries charge @xmath3 under a global @xmath44 symmetry and transforms as @xmath61 under time reversal .
the corresponding symmetry group is @xmath62 .
assume it is in a gapped insulating phase ( at least in the absence of any boundaries ) and that there is a unique ground state on topologically non - trivial manifolds .
for any such insulator in 3d , the effective lagrangian for an external em field obtained by integrating out all the matter fields will take the form @xmath63 the first term is the usual maxwell term and the second is the ` theta ' term in eqn .
[ ltopo ] : several properties of the theta term are well known . first under time reversal , @xmath64 .
next on closed manifolds , the integral of @xmath65 is quantized to be an integer so that the quantum theory is periodic under @xmath66 .
these two facts together imply that time reversal symmetric insulators have @xmath67 with @xmath68 an integer .
trivial time - reversal symmetric insulators have @xmath69 while free fermion topological insulators have @xmath70 .
if we allow for a boundary to the vacuum and further assume that the boundary is gapped ( if necessary by breaking time reversal symmetry ) , then the @xmath0 term leads to a surface hall conductivity of @xmath71 . to see this , assume a boundary ( say at @xmath72 ) , @xmath73 is zero for @xmath74 and constant @xmath0 for @xmath75 .
the action associated with the @xmath0 term is @xmath76 where @xmath77 is the external electromagnetic potential and @xmath78 .
this is a surface chern - simons term and leads to a hall conductivity @xmath79 . for fermion topological insulators
@xmath70 so that the surface @xmath80 .
if we shift @xmath81 , then the surface @xmath82 .
this corresponds to simply depositing an ordinary integer quantum hall state of fermions at the surface of this insulator - hence this should not be regarded as a distinct bulk state so that the only non - trivial possibility is @xmath83 .
now let us consider bosonic insulators .
again t - reversal and periodicity imply @xmath84 and a surface @xmath85 .
a crucial observation is that now @xmath86 must be regarded as * distinct * from @xmath69 . at @xmath86 the surface @xmath87 . but this can not be obtained from the surface of the @xmath69 insulator by depositing any 2d integer quantum hall state of bosons .
recent work@xcite has shown ( see ref . for a simple argument ) that 2d iqhe states of bosons necessarily have @xmath88 even .
thus the surface state of the @xmath86 boson insulator is not a trivial 2d state but rather requires the presence of the 3d bulk .
therefore @xmath86 necessarily corresponds to a non - trivial 3d bosonic ti .
@xmath89 is however trivial as then the surface state can be regarded as a 2d bosonic iqhe state .
one may still obtain a 3d topological phase , but the topology is not manifest in the electromagnetic response .
we can sharpen and generalize this result . under t - reversal as @xmath90 , @xmath91 . as the bulk state is t - reversal invariant
we require that the surface state at @xmath92 be obtainable from the surface state at @xmath93 by depositing a 2d iqhe boson state .
let us characterize the surface state by both its electrical and thermal hall conductivities @xmath94 . under t - reversal both
hall conductivities change sign .
the requirement above then means that @xmath95 must correspond to the allowed electrical / thermal hall conductivity of a 2d boson iqhe state . for @xmath86
, it follows that @xmath96 .
it is thus half " of the elementary 2d boson iqhe state . for @xmath70 ,
@xmath97 and this is not allowed for the 2d bosonic iqhe .
it follows therefore that 3d bosonic tis with no intrinsic topological order " can not have @xmath70 .
it is of course very easy to construct such states@xcite ( or other states with fractional @xmath0 ) if we allow for fractionalization of the boson but that violates our original assumption . a 2d iqhe state with @xmath98 is allowed and is discussed by kitaev .
thus a 3d boson ti with surface @xmath99 is allowed .
combining these two types of fundamental states generates the allowed thermal and electrical hall responses on the surface .
later on in the paper we will discuss how these results fit in with the formal classification of spt and other short ranged entangled phases in 3d .
for now we reiterate the crucial observation of this section : _ a state with em response of @xmath86 necessarily describes a topological insulator of @xmath100-reversal symmetric bosons while @xmath70 requires the presence of intrinsic topological order"_. in the next section we study the properties of this @xmath86 boson topological insulator in detail .
in this section we will derive the non - trivial surface theory of one example of a 3d bosonic spt phase .
we will soon specialize to the symmetries of the topological insulator : charge conservation and time reversal symmetry ( @xmath101 ) , and exhibit a nontrivial topological phase in three dimensions , built purely of bosons . to begin with we
will assume that there are two species of bosons whose numbers are separately conserved , and there is enlarged @xmath102 .
later we will break this to just @xmath103 symmetry by including inter - species boson mixing terms in the hamiltonian .
a similar construction @xcite has proven to be very powerful in @xmath34 .
we consider two approaches : the first approach exploits the fact that a bosonic spt phase in @xmath104 dimensions has surface states that correspond to a conventional theory of bosons in @xmath105 dimensions _ except _ in the way symmetries are implemented .
for example , the edges of spt phases of bosons in d=2 + 1 dimensions correspond to conventional 1d luttinger liquids , except for their unusual symmetry transformations @xcite . we therefore consider a 2d bosonic state to model the surface and assume the surface is a superfluid breaking one of the u(1 ) symmetries .
then , vortices of this condensate may transform under a projective representation of the remaining symmetry group . in a projective representation ,
even the identity element of the symmetry group induces a phase rotation .
hence local operators , that can be physically measured , must remain unchanged under the identity operation of the symmetry group , since this corresponds to ` doing nothing ' .
however , vortices , which are non - local objects , can transform projectively .
one may attempt to restore the u(1 ) symmetry by condensing vortices .
however , the projective transformation ensures that when vortices condense , they necessarily break another symmetry . in this way
both the boson and vortex condensates lead to symmetry breaking , in line with our general expectation for the surface of a 3d spt phase .
it is important that vortices transform projectively , so that they can not be screened by bosons to obtain a trivial representation of the symmetry group .
this is a generalization of the idea of quantum number fractionalization - for example a particle with half charge changes sign under the @xmath1 phase rotation of bosons , implying a projective representation .
clearly , a half charge can not be screened by any finite number of bosons .
projective representations were also recently used to classify spt phases in d=1 + 1where they correspond to the ends of gapped one dimensional topological phases .
for example , the half - integer spin edge states of spin-1 haldane chains furnishes a projective representation of the rotation group .
this suggests a physical picture of a 3d spt phase , in which the vortex line in the bulk is similar to a haldane chain type gapped phase , which necessitates low energy states on the surface where the vortex ends . in this section
we specialize to the symmetries of the topological insulator .
then this procedure explicitly produces a topological phase characterized by quantized magnetoelectric effect @xmath19 .
the second approach will be to directly implement the property discussed above in section [ sec : gentrnsprt ] that if the surface breaks @xmath100-reversal and is gapped then it has quantized hall transport . if @xmath100-reversal is not broken a powerful approach to obtain the surface theory is to start with the theory of the quantum phase transition point between the two bosonic quantum hall phases that correspond to the two @xmath100-broken surfaces . in the case of free fermion
topological insulators , a similar reasoning leads to the single dirac cone surface state that describes the transition between the @xmath106 states on the surface . for free fermions
the transition between these integer quantum hall states is described by a chalker - coddington network model@xcite . for the bosonic problem of interest here
we construct an analogous network model and show it leads to a sigma model with a topological term .
the results of these two approaches are readily seen to be connected . in both cases
the field theories we obtain for the surface have appeared previously in the context of deconfined quantum criticality .
we discuss the phase diagram of the surface states described by these field theories .
when inter - species tunneling is included the vortices of the two species of bosons get confined to each other .
the resulting single vortex no longer transforms projectively under the physical symmetries .
however we argue that it is most conveniently viewed as a fermion .
this precludes the possibility of obtaining a trivial insulating phase at the surface by condensing vortices .
consider a boson field at the surface with phase degree of freedom @xmath107 .
we assume the bulk is insulating and the surface is in the @xmath108 plane .
the surface theory could spontaneously break a global @xmath44 symmetry of boson number conservation ( a surface superfluid ) or stay insulating .
more precisely as the bulk is always assumed insulating , the vortex line loops have proliferated in the bulk .
these vortex lines penetrate the surface at points , which may be viewed as point vortices of the two dimensional surface theory , since there is no vortex line tension in the insulating bulk .
these point vortices are gapped when the surface is a superfluid .
if instead they are condensed the surface will be insulating . to describe vortices we go to a dual description@xcite , where we write the density and currents of the boson @xmath109 on the surface in terms of the field strengths of a gauge field @xmath110 .
in particular the density of bosons is : @xmath111 ( the reason for the subscript 2 on @xmath112 will soon be apparent ) .
the boson insertion operators @xmath113 correspond to monopole insertion operators , since they insert @xmath114 magnetic flux .
now , the vortices @xmath115 are particles that couple minimally to the gauge field @xmath116 . in general there will be multiple vortex species that transform into each other under the symmetry operation .
let us label them by @xmath117 , so @xmath118_i=\psi_{2i}$ ] .
all these fields couple minimally to the gauge field .
as argued above , one route to obtaining topological surface states is if the surface vortices transform under a projective representation of the remaining symmetry .
vortices that transform projectively under a global symmetry are actually not at all unfamiliar .
it describes the generic situation of two dimensional bosons on a lattice , say , at some commensurate filling@xcite .
these projective vortices play a crucial role in the theory of deconfined quantum criticality .
we will only need to consider two component vortex fields , @xmath121 for a variety of cases considered in this paper .
the gauge invariant combination @xmath122 then defines another bosonic field .
now , eqn . [ nccp1vortex ] closely resembles the action for a deconfined quantum critical point ( non - compact cp@xmath123 theory with easy plane anisotropy ) @xcite .
we will demonstrate how this emerges as the theory for the surface states , and also , in the next section , describe a three dimensional bulk theory that leads to this edge theory .
these are the symmetries of the topological insulator , a conserved @xmath44 charge and @xmath40 time reversal symmetry .
the semi - direct product appears so that the charge insertion operator @xmath125 is invariant under time reversal , which involves both @xmath126 and @xmath127 . here
we will construct the surface theory of a 3d topological phase with these symmetries .
let us begin with an enlarged symmetry , two species of bosons that are separately conserved . consider a condensate of one species @xmath109 .
vortices in this condensate are created by the field @xmath115 . we need to specify the projective representation for the vortices @xmath115 and the transformation of the bosons @xmath128 .
the remaining symmetry group @xmath129 has a single projective representation ( p1 ) which acts as follows . under a @xmath44 rotation by angle @xmath130 , the fields @xmath131 and under time reversal : @xmath40 : @xmath132 and @xmath133 . or more compactly : here the @xmath135 are the pauli matrices in the standard representation .
thus the vortices carry charge @xmath136 , of bosons of the other species .
the time reversal symmetry that interchanges the two vortex fields ensures that the vortex charge is fixed exactly at half .
it is impossible to ` screen ' this charge with regular integer charged bosons .
these transformation laws of course determine how the boson operator @xmath137 and their density @xmath138 transforms .
we also need to specify how the bosons @xmath139 and density @xmath140 transforms .
the symmetry transformations are : @xmath141 the conjugate boson numbers therefore transform as @xmath142 a necessary compatibility check is that eqn .
[ nccp1vortex ] is invariant under the symmetry operation , which can be verified for these transformations .
for example , time reversal symmetry is implemented via eqn .
[ p1p1vortex ] on the vortex fields which is compatible with @xmath138 ( and hence @xmath116 ) remaining invariant while @xmath127 under time reversal .
moreover , since the bosons carry charge , the monopole insertion operators are forbidden .
condensing single vortices will then break symmetry as described below .
this is best analyzed by assuming separate number conservation of each species of boson , @xmath143 , in which case they can be coupled to an _ external _ gauge potential @xmath144 .
then the effective lagrangian eqn[nccp1vortex ] reads : here @xmath146 .
the @xmath147 are vortex fields of @xmath109 which carry half charge of boson species 2 .
the flux of the gauge field @xmath148 is precisely the conserved density of species @xmath3 ; hence the last term in the above action where the external probe gauge field @xmath149 couples to this current .
this action will need to be modified by including all symmetry allowed perturbations .
we will do this and analyze the possible phases below , but as a preview , consider the effect of breaking time reversal symmetry by condensing just one species of vortex say @xmath150 . anticipating a single charge ,
consider an external field that couples equally to the two charge densities @xmath52 .
the vortex condensate forces @xmath151 , which when substituted into the action above yields the electromagnetic response @xmath152 , which yields @xmath53 on the surface , indicating a magneto electric response @xmath19 , as advertised .
we now change track and obtain the same surface theory from a very different point of view .
the general considerations of section [ sec : gentrnsprt ] showed that a time reversal symmetric boson insulator with electromagnetic response characterized by @xmath86 is in a topological insulator phase .
this key result relied on the observation that if @xmath153 is broken at the surface to gap it out then such a state has a quantized electrical hall conductivity @xmath154 , and a thermal hall conductivity @xmath155 .
what if t - reversal is not explicitly broken at the surface ?
the surface can then potentially be gapless .
what is the nature of the resultant theory ? to construct this theory it is extremely instructive to learn from the example of the free fermion @xmath100-reversal symmetric topological insulator . in that case
if @xmath100 is explicitly broken to gap out the surface , then we get @xmath106 .
when @xmath100 is unbroken , it is possible to get a single massless dirac cone which is exactly the low energy theory of the transition between two integer quantum hall plateaus of fermions in @xmath34 .
generically we can tune the chemical potential to move away from the dirac point to get a fermi surface which encloses the dirac point .
note that the quantized hall conductance jumps by 1 across the integer quantum hall plateau transition .
when applied to the surface of the 3d fermionic topological insulator , the transition connects the two possible @xmath100-breaking surface states that go into each other under time reversal .
thus each such surface state must be assigned @xmath106 and the critical theory itself is time reversal invariant . in contrast
when applied to strictly 2d systems the network model describes a transition between plateaus with @xmath156 and @xmath87 .
the familiar free fermion example gives us a crucial clue to construct the theory of the @xmath100-reversal symmetric surface state of the boson topological insulator .
first construct the low energy theory of the @xmath34 integer quantum hall state of bosons as a potential candidate for the gapless surface state of the @xmath157 topological insulator . then add perturbations allowed by symmetry to obtain the generic surface theory .
across the boson integer quantum hall transition , @xmath88 jumps by 2 . as for the fermionic example , when this transition is realized at the surface of the bosonic topological insulator ,
the two plateau states on ether side of the transition are related by @xmath100-reversal and must be assigned @xmath154 consistent with our earlier arguments .
we will study the phase transition using a network " model construction .
the idea is to start with the theory of the edge state and couple together opposite edges .
let us warm up with the familiar example of the iqhe transition of fermions from a state with @xmath87 to one with @xmath156 .
the model is defined by fig .
[ nwrkpic ] .
it is described by the euclidean action @xmath158 with @xmath159 with @xmath160 the first term is the sum of the actions of a single chiral edge mode of the @xmath87 fermion iqhe state taken to propagate in opposite directions for adjacent @xmath117 .
the second term describes electron hopping between opposite moving edge channels .
when @xmath161 all chiral edge channels are paired with partners and @xmath156 .
conversely if @xmath162 , then all edge channels get paired except the two end ones and we get the fermion iqhe state with @xmath87 .
the transition occurs at @xmath163 and is readily seen to yield a single massless dirac fermion in the continuum limit is taken in the @xmath117 direction .
let us now repeat this construction for the bosonic iqhe transition .
the edge theory for the boson iqhe state with @xmath164 has a pair of counterpropagating edge modes - one carries the charge and the other is neutral@xcite .
it is convenient to write the effective action of the edge as an @xmath165 wzw theory : @xmath166\ ] ] here @xmath56 is a @xmath167 matrix with entries @xmath168 .
the @xmath169 are the two physical boson fields which form the iqhe quantum hall state . a network model capable of describing the boson iqhe state may now be written down and is defined by fig [ nwrkpicb ] .
again we have an array of opposite edge channels which are coupled together by boson hopping @xmath170 .
the full effective action is then @xmath171 \\ s_t & = & - \sum_i t_i tr\left(g^\dagger_{i}g_{i+1 } + h.c\right)\end{aligned}\ ] ] with @xmath172 and @xmath173 as before . if @xmath174 we get the trivial insulator while if @xmath175 we get the boson iqhe state .
the transition occurs at @xmath163 .
a low energy theory of the transition is obtained by taking the continuum limit in the @xmath117 direction .
as the opposite moving edge channels have opposite wzw terms they nearly cancel , and it is necessary to carefully sum them .
fortunately precisely this sum was performed in ref . where the same model arose in a different context .
the result is the effective @xmath176 dimensional action @xmath177\ ] ] the second term is a @xmath0 term for the @xmath178 matrix valued field @xmath56 in @xmath179 dimensions corresponding to @xmath180 = z$ ] . in the present context our calculation has yielded this term at the value @xmath70 .
we of course do not have full @xmath178 symmetry rotating between @xmath109 and @xmath181 in the microscopic system .
for the time being let us assume that we have @xmath41 symmetry corresponding to separate conservation of the @xmath169 bosons .
further let us also assume there is a @xmath21 symmetry interchanging @xmath109 and @xmath181 .
later we will relax all these assumptions . then the results of ref .
show that the field theory above at @xmath70 maps to the self - dual easy plane non - compact @xmath182 ( nccp@xmath183 ) model .
equivalently it also maps onto a model of two species of spacetime loops with a phase @xmath184 associated with each linking of the two loop species .
the @xmath70 @xmath178 matrix field theory ( with the @xmath41 anisotropy ) or the equivalent easy plane nccp@xmath183 model arise also in the theory of deconfined quantum criticality in two space dimensions .
remarkably we see that the field theories describing the boson iqhe plateau transition ( and hence the surface states of the 3d boson topological insulator ) are closely related to the theory of deconfined quantum criticality . in the previous subsection we obtained this connection from a different point of view .
we will in the rest of the paper explore this connection in more detail and generality .
for now we merely point out that the results of ref .
( see also ref . )
show that the @xmath70 @xmath178 matrix field theory in two space dimensions does not have a trivial gapped disordered phase .
its phases either break symmetry , are gapless or have topological order .
this is a hallmark of the surface state of a symmetry protected topological phase - there is no trivial gapped phase that preserves all the symmetries .
* surface of the bosonic tis : field theories * we now exploit our intuition about deconfined criticality to obtain the theory of possible surface states of the 3d boson ti starting with the @xmath70 @xmath178 matrix field theory .
we first describe a number of equivalent field theoretic descriptions of the surface state paying particular attention to the realization of the physical @xmath103 symmetry .
first we note that the @xmath178 matrix @xmath56 is related to the physical boson fields @xmath185 through @xmath186 under the global @xmath44 symmetry , both bosons transform with charge @xmath3 , _
i.e _ @xmath187 we implement time reversal by simply requiring that @xmath188 in terms of the phases of the bosons , defined through @xmath189 , and the conjugate bosons densities @xmath190 the symmetry transformations are the same as in eqns .
[ phitransform],[ntransform ] so that we are indeed describing the same symmetry class in the two approaches .
we remind the reader that the total number @xmath191 of the two boson species is conserved due to the global @xmath44 symmetry but the relative number @xmath192 is in general not .
as promised before we will first analyse the theory in a limit where this relative number is also conserved ( so that there is @xmath41 symmetry ) and then include interspecies tunneling terms to recover the generic case .
.symmetry properties of the @xmath193 fields .
@xmath194 are the charges under the two @xmath44 symmetries associated with @xmath185 respectively .
@xmath195 are the vorticities in the phase of @xmath185 . they can also be viewed as the gauge charge for the coupling to the corresponding @xmath44 gauge fields . the last column gives the transformation under time reversal . [ cols="^,^,^,^,^,^",options="header " , ] as we emphasized in the main paper the realization of symmetry is such that these @xmath21 topologically ordered phases can not arise in strict two dimensional models with local action of the symmetry group .
this is readily seen from the general @xmath196-matrix classification of @xmath39 time reversal invariant gapped abelian phases in ref . .
here we will present an elementary analysis that is sufficient for the purposes of this paper . strictly
two dimensional systems admit an edge to the vacuum ( or equivalently a trivial gapped insulator ) . without changing the symmetry .
this is a key difference from the surface topological order of interest to us here where a domain wall with a trivial gapped insulator is not possible without breaking symmetry .
the analysis below , following ref . , relies crucially on analysing the symmetries of the edge lagrangian that describes a strictly 2d system .
hence it distinguishes between topological order that is allowed in strict 2d and ones that require a 3d bulk . for the case of @xmath21 topological order we may , as usual , take @xmath197
the corresponding chern - simons lagrangian is simply @xmath198 here @xmath199 are internal gauge fields and @xmath77 is an external ` probe ' gauge field .
the charge vector @xmath200 has integer components .
physically @xmath201 are @xmath202 of the @xmath203 and @xmath204 particles respectively . if both @xmath203 and @xmath204 carry global @xmath44 charge @xmath4 , then @xmath205 .
as we already noted such a charge assignment implies a non - zero hall conductivity ad hence can not describe a strict 2d system with time reversal invariance . thus the surface topological order described above in tables
[ stou1rz2 t ] and [ stou1z2tp2 ] can not occur in strict 2d systems .
it remains to discuss the other two surface topological orders . in both ( as well as the case in table .
[ stou1z2tp2 ] both @xmath203 and @xmath204 are kramers doublets .
we now show that this is not possible in any time reversal invariant strict 2d @xmath21 topologically ordered state .
the @xmath206 dimensional edge theory corresponding to the chern - simons lagrangian above is @xmath207 with @xmath208 .
demanding time reversal invariance of the edge lagrangian we see immediately that the edge densities @xmath209 must transform with opposite signs .
however if @xmath210 creates one member of a kramers doublet , it must transform as @xmath211 under time reversal .
equivalently @xmath212 so that the edge density @xmath213 .
thus if both @xmath203 and @xmath204 are kramers doublets then both corresponding edge densities must be even under time reversal .
but this is inconsistent with our deduction above from demanding time reversal invariance of the edge lagrangian .
we thus conclude that in strict 2d systems both @xmath203 and @xmath204 can not be kramers pairs in a time reversal invariant system .
it is however alowed to happen at the surface of the 3d spt phases described in this paper . in passing we
note that this precludes the possibility that strictly @xmath39 spin models have gapped @xmath21 topological phases where both non - trivial bosonic quasiparticles carry spin-@xmath4 ( _ i.e _ are spinons ) while the fermionic quasiparticle carries no spin .
99 m. z. hasan and c. l. kane , colloquium : topological insulators , " rev .
82 , 3045 ( 2010 ) .
qi and s .- c .
zhang , topological insulators and superconductors , " rev
83 , 1057 ( 2011 ) .
m. z. hasan and j. e. moore , three - dimensional topological insulators , " annu .
matter phys . 2 , 55 ( 2011 ) .
andreas p. schnyder , shinsei ryu , akira furusaki , andreas w. w. ludwig , classification of topological insulators and superconductors in three spatial dimensions , " phys . rev .
b 78 , 195125 ( 2008 ) .
alexei kitaev , periodic table for topological insulators and superconductors , " arxiv:0901.2686 .
t. senthil , a. vishwanath , leon balents , subir sachdev , m. p.
a. fisher,deconfined quantum criitcal points " , science 303,1490 ( 2004 ) ; t. senthil , leon balents , subir sachdev , ashvin vishwanath , and matthew p. a. fisher ,
quantum criticality beyond the landau - ginzburg - wilson paradigm , " phys .
b 70 , 144407 .
a. w. sandvik , evidence for deconfined quantum criticality in a two - dimensional heisenberg model with four - spin interactions , " phys .
98 , 227202 ( 2007 ) .
anders w. sandvik , continuous quantum phase transition between an antiferromagnet and a valence - bond solid in two dimensions : evidence for logarithmic corrections to scaling , " phys .
104 , 177201 ( 2010 ) .
a. banerjee , k. damle , and f. alet ,
impurity spin texture at a deconfined quantum critical point , " phys .
b 82 , 155139 ( 2010 ) .
r. g. melko and r. k. kaul , scaling in the fan of an unconventional quantum critical point , " phys .
100 , 017203 ( 2008 ) .
ribhu k. kaul and anders w. sandvik , lattice model for the su(n ) neel to valence - bond solid quantum phase transition at large n , " phys .
108 , 137201 ( 2012 ) . | we discuss physical properties of ` integer ' topological phases of bosons in d=3 + 1 dimensions , protected by internal symmetries like time reversal and/or charge conservation .
these phases invoke interactions in a fundamental way but do _ not _ possess topological order and are bosonic analogs of free fermion topological insulators and superconductors .
while a formal cohomology based classification of such states was recently discovered , their physical properties remain mysterious . here
we develop a field theoretic description of several of these states and show that they possess unusual surface states , which if gapped , must either break the underlying symmetry , or develop topological order . in the latter case ,
symmetries are implemented in a way that is forbidden in a strictly two dimensional theory . while this is the usual fate of the surface states , exotic gapless states
can also be realized .
for example , tuning parameters can naturally lead to a deconfined quantum critical point or , in other situations , a fully symmetric vortex metal phase .
we discuss cases where the topological phases are characterized by quantized magnetoelectric response @xmath0 , which , somewhat surprisingly , is an odd multiple of @xmath1 .
two different surface theories are shown to capture these phenomena - the first is a nonlinear sigma model with a topological term .
the second invokes vortices on the surface that transform under a _ projective _ representation of the symmetry group .
a bulk field theory consistent with these properties is identified , which is a multicomponent @xmath2 theory supplemented , crucially , with a topological term .
bulk sigma model field theories of these phases are also provided . a possible topological phase characterized by the thermal analog of the magnetoelectric effect
is also discussed . |
although the number of detected @xmath0-ray bursts has increased tremendously due to the batse instrument on board the compton @xmath0-ray observatory @xcite , unique identifications and quick follow - up observations of the celestial object that harbours the burst site are rare .
this issue has been successfully addressed by the bepposax satellite , that monitors a large fraction of the sky ( @xmath2 ) in the intermediate x - ray range @xcite and is able to derive burst localisations with an accuracy of @xmath3 . with such improved localisations ,
the beppo - sax team has succeeded in detecting x - ray afterglows @xcite , by which valuable insight into the emission mechanism was gained .
finding @xmath0-ray burst counterparts and performing spectroscopy at wavelengths other than the @xmath0-ray band during the active phase of a burst as well as providing good localisations for follow - up observations is the basic scientific motivation of the hete-2 mission .
this publication is structured as follows : after a brief description of hete-2 s instrumentation in section [ cont : instruments ] , the position reconstruction algorithm is outlined in section [ cont : algorithm ] .
the performance of the algorithm on monte carlo generated events as well as on data taken in astrometric observations of the x - ray source _
scorpius x-1 _ is presented in section [ cont : result ] . in section [ cont : psf ]
findings on the point spread function are shown . a summary in section [ cont : conclusion ]
concludes the paper .
the high energy transient explorer ( hete-2 ) is a dedicated mission for the localisation and spectroscopy of @xmath0-ray bursts .
details of the instrumentation of hete-2 and the mission are given in @xcite .
its scientific payload consists of three experiments : _ fregate _ , a scintillation crystal experiment , that is expected to provide triggers because of its high sensitivity and large sky coverage , _ sxc _ , a coded - mask imager based on a x - ray ccd chip design , which is able to localise @xmath0-ray bursts with very high spatial resolution , and the core experiment _
wxm_. the wide field x - ray monitor _ wxm _ consists of two perpendicularly oriented 1-dimensional coded mask cameras , in which photons are detected by position sensitive proportional counters .
the detectors , one pair for each of the two orthogonal systems , are filled with xenon gas at a pressure of @xmath4 with an admixture of @xmath5 carbon dioxide as a quenching gas .
each detector contains three carbon wire anodes in which the position of an absorbed photon is inferred by comparing the accumulated charges at both ends of a wire .
_ wxm _ achieves a position resolution of @xmath6 ( fwhm ) at @xmath7 .
a veto system below the counting wires reduces the background due to charged particles .
_ wxm _ is sensitive to photons with energies @xmath8 in the range @xmath9 with an relative energy resolution of @xmath10 at @xmath11 .
the mask pattern , identical for the @xmath12- and @xmath13-system , is placed @xmath14 above the detectors and is composed of 103 elements , each @xmath15 in width . on third of the elements
are open and randomly distributed .
the parameters have been optimised with respect to source localisation accuracy for the expected levels of signal and background photon count rates @xcite .
the combined @xmath12- and @xmath13-systems are monitoring a field of view of approximately @xmath16 .
figure [ fig : hete ] shows the actual mask pattern of a _ wxm _ camera .
further details of the _ wxm _ cameras can be found in @xcite . in order to provide other experiments with precise localisations of @xmath0-ray burst sites within seconds after burst onset
, an array of twelve burst alert stations has been installed below hete-2 s flight path .
once received , the information is relayed to the grb coordinate network ( gcn ) , from where it may be obtained by interested observers .
for the sake of readability , the algorithm is described for determining the angle of incidence @xmath17 in the @xmath12-detector , completely analogous formulae apply to the @xmath13-system . figure [ fig : angles ] provides the definition of all quantities involved .
the task of localising a single point source is performed by computing the correlation function @xmath18(x)$ ] of the mask pattern @xmath19 and the recorded intensity distribution @xmath20 ; the asterisk denoting complex conjugation : @xmath21(x)=\int m(\lambda)\cdot f^*(\lambda+x ) d\lambda\mbox{. } \label{eqn : correlation}\ ] ] the correlation function @xmath22(x)$ ] will peak at a value @xmath23 , indicating the distance @xmath20 is shifted with respect to @xmath19 .
this yields the angle of incidence @xmath17 by @xmath24 , where @xmath14 is the distance between mask and detector . in @xmath19 ,
open elements of the mask pattern have been assigned a value of @xmath25 , whereas closed elements correspond to a value of @xmath26 , where @xmath27 is the open fraction of hete-2 s imaging system .
this method is known as balanced correlation and removes the correlation background .
there exist more sophisticated correlation schemes such as unbiased balanced correlation ( e.g. @xcite , @xcite ) , that do take into account partial shadowing of the mask onto the detector by bursts in the periphery of the field of view and increased coding noise caused by steady strong off - axis sources . for hete-2 s flight algorithm we rely on basic balanced correlation for three reasons :
the width of the mask is almost twice as large as the active area of the detector and only bursts at angles @xmath28 off the optical axis are affected by partial shadowing .
furthermore , the running average of the mask s transparency is close to constant with deviations being very small , so that the modulation of the pattern is insensitive to angle of incidence . thirdly , in hete-2 s nominal survey mode , the satellite is rotated in such a way that strong x - ray sources are avoided and the assumption of isotropically incidenting background radiation is valid . in reality
the correlation is complicated by imperfections of the detector : the photons travel a finite path before interacting and the loci of photons interacting in the detector are measured with finite position resolution ( in the case of _ wxm _ to @xmath6 , subtending an angle of @xmath29 ) .
whereas the finite position resolution simply results in a blurring of the image , the penetration of energetic photons effectively induces an additional shift in the image , if the event is happening at nonzero angles of incidence , and causes the angles of incidence to be overestimated .
this effect increases in a complex fashion with distance from the center of the field of view and is of the order of @xmath30 at @xmath31 from the optical axis for a photon distribution following a power law spectrum with a spectral index of @xmath32 .
in contrast to a time consuming monte carlo simulation , in which one determines the image of the mask pattern under estimated angles of incidence , the algorithm presented in this article aims at modelling all effects involved in the delapidation of the mask pattern , enabling a much faster localisation of the @xmath0-ray burst site .
imaging aberrations may be described by applying suitable changes to the mask pattern @xmath19 prior to the calculation of the correlation function .
this can be archieved by convolving the mask pattern @xmath19 with integration kernels @xmath33 and @xmath34 , that describe the penetration of highly energetic photons and the detector resolution , respectively , @xmath35(x)=\int m(\lambda)\cdot p(x-\lambda ) d\lambda\mbox{. } \label{eqn : convolution}\ ] ] the convolution kernel @xmath33 that is used for describing the penetration of photons is given by equation [ eqn : penetrate ] : @xmath36\cdot x\right)\mbox { , } \label{eqn : penetrate}\ ] ] where @xmath37 is the heaviside function and _ sgn _ the signum function the convolution yields an altered mask pattern @xmath38(x)$ ] , in which the penetration of photons is incorporated .
the length scale @xmath39 of the exponential decay , given by equation [ eqn : ceqn ] , is equal to the spectrally averaged attenuation length @xmath40 of photons inside the detector , corrected by a projection factor : @xmath41 estimates for the angles of incidence @xmath17 and @xmath42 follow from a source localisation with the unmodified mask pattern @xmath19 in a first step .
the finite spatial resolution of the position sensitive proportional counter is described by convolution with a gaussian integration kernel as in equation [ eqn : resolution ] : @xmath43 the standard deviation corresponds to a value for full width half maximum of @xmath6 .
in contrast to @xmath40 , @xmath44 depends only very weakly on the spectral distribution of the incident photons and is considered to be constant .
after convolution with @xmath34 , the mask pattern has been modified to @xmath45(x)$ ] .
@xmath46 is the image of the mask pattern under ideal statistics and the source reconstruction with @xmath46 should not display any deviation from the ideal behaviour .
the correlation @xmath47(x)$ ] yields a corrected value @xmath48 for the angle of incidence .
due to their high numerical complexity it is favourable to perform both the convolutions and the image localisation by correlation in the fourier domain .
the nomenclature is such that lower case letters denote functions in real space and the corresponding upper case letters their fourier transforms : @xmath49=\frac{1}{2\pi}\int m(x)\exp(-ikx)dx\mbox{. } \label{eqn : fourierpair}\ ] ] convolutions and correlations reduce by virtue of equation [ eqn : fconvolve ] to mere multiplications in fourier space .
the asterisk denotes complex conjugation : @xmath50(x)&=&\mathcal{f}^{-1}\left[p(k)m(k)\right]\mbox{,}\\ \left[f\otimes m''\right](x)&=&\mathcal{f}^{-1}\left[f(k)^*m''(k)\right]\mbox{. } \end{array } \label{eqn : fconvolve}\ ] ] diagram [ dia : cc ] summarises all steps .
starting from the fourier - transform of the mask pattern @xmath51 $ ] , both convolutions and the correlation are carried out in fourier - space by determining the product @xmath52 .
inverse transformation yields the correlation function @xmath53 $ ] , from which the corrected angle of incidence @xmath48 is derived : @xmath54 in comparision to the _ wfc _ camera on board the bepposax satellite , the penetration effect is noticably more pronounced in the case of hete-2 .
evaluating the average attenuation length for a photon at the corner of the field of view projected onto the detector plane determines the angular deviation in the worst case .
this value for _ wfc _ is @xmath55 times smaller compared to _
wxm _ , because of the narrower field of view , the higher gas pressure , which results in a shorter average attenuation length and the larger distance between the mask and detector .
furthermore , being a genuine 2-dimensional camera , the aberration vector field of _ wfc _ is radial and does not have the complex angular dependence described by equation [ eqn : ceqn ] .
in order to investigate to which extend the algorithm is capable of correcting systematic deviations , a set of @xmath56 monte carlo images containing a large number of photons ( @xmath57 ) was generated , so that the statistical scatter in the source reconstruction is small .
the simulated burst positions were randomly distributed , the angles of incidence ranging from @xmath58 to @xmath59 .
the number of incident photons was corrected with a factor of @xmath60 for the purpose of coping for the decreasing projected area of the detector at increasing angles of incidence .
the spectral distribution of photons followed a power law with spectral index of @xmath32 , which is considered to be shallow , but not untypical for @xmath0-ray bursts @xcite . because of the flatness of the spectrum , the aberration effect due to penetrating photons is distinctive and easy to study .
localisation of photons by the proportional counter was modelled by employing detector response data taken prior to the launch .
figure [ fig : pure ] shows the deviation @xmath61 of the uncorrected reconstructed burst position from the nominal burst position as a function of angle of incidence @xmath62 .
the aberration increases with increasing distance @xmath62 from the center of the field of view .
each point corresponds to the localisation derived from a monte carlo image .
furthermore , one observes a widening of the scatter of @xmath61 with increasing distance from the optical axis .
this is due to the fact that the aberration , being proportional to the function @xmath63 to first order , is concave for @xmath64 and convex for @xmath65 .
the magnitude of this curvature increases with increasing distance from the origin . in order to disentangle the contributions from statistical scatter in the reconstruction from residual systematics ,
the dataset has been split into @xmath66 intervals in @xmath67 . for each interval
@xmath68 the mean value @xmath69 of @xmath70 as well as the standard deviation @xmath71 has been derived . because the intervals are chosen such that they cover an angular range of just @xmath72 , the scatter @xmath71 can be assumed to be dominated by statistics .
an increase in photon number @xmath73 per image would have reduced the statistical scatter , according to @xmath74.from the values @xmath69 we derive the quantity @xmath75 , which is the mean quadratic deviation from zero .
further on , from @xmath71 we determine the average statistical scatter @xmath76 . from those two quantities ,
we define the spot size @xmath77.figure [ fig : mcdata_depend ] illustrates the performance of the applied corrections : the mean quadratic deviation @xmath69 as well as the statistical error @xmath71 is plotted against the angle of incidence @xmath67 .
as the flatness indicates , @xmath69 is consistent with zero , i.e. the imaging aberrations have been properly corrected .
moreover , @xmath71 has become independent from the angle of incidence @xmath67 , which indicates that the second order influence of @xmath78 on the reconstruction has been properly accounted for .
residual deviations are caused by detector nonlinearities , especially in the charge division method by which the position of an absorbed photon is determined .
the mean statistical scatter has been determined to @xmath79 and the mean systematic deviation from zero to @xmath80 .
the algorithm thus fulfills the design requirements .
figure [ fig : mcdata_spot ] shows the result of combining the reconstructions in @xmath12- and @xmath13-direction .
the scatter plot is circularly symmetric and the width of the spot corresponds to @xmath81 . in the above analysis , the value for the average attenuation length @xmath40 has been optimised is such a way that @xmath82 assumed a minimal value .
the smallest deviations resulted in the choice @xmath83 , which is in agreement with the theoretical value @xmath84 .
the theoretical value was derived by determining the flux weighted average of tabulated values for the photon attenuation length in xenon , where the photon spectrum has been altered by the spectral photon absorption probability .
data from the x - ray source _
scorpius x-1 _ has been taken for calibration purposes .
due to hete-2 s antisolar pointing , any source will drift through the field of view at ecliptic rate , i.e. @xmath85 .
the orientation of the satellite was such that the _ wxm _ coordinate system was rotated @xmath86 relative to the celestial @xmath87 frame . _ scorpius x-1 _ was viewed at angles of incidence ranging from @xmath88 to @xmath89 and at @xmath90 fixed .
the spacecraft attitude was known to an accuracy surpassing @xmath1 .
as pointed out in @xcite , the spectrum of _ scorpius x-1 _ , being a low - mass x - ray binary , can be described by a multicolour blackbody spectrum . due to the integration time of @xmath91 , short term
spectral changes like quasi - periodic oscillations average out .
long term variabilities either happen on significantly longer time scales compared to the data taking , or they affect energies outside hete-2 s sensitivity interval .
determining the average attenuation length by a photon flux weighted integral with tabulated values for the attenuation length yielded @xmath92 .
thus , the spectrum of _ scorpius x-1 _ is much softer compared to the previous case .
the background subtracted images oconsisted roughly of @xmath57 photons .
the analysis [ ana : mc ] is repeated for the _ scorpius x-1 _ dataset . for each pointing @xmath69 and @xmath71
have been derived .
figure [ fig : scodata_depend ] illustrates the accuracy of the source localisation algorithm .
again , the values @xmath69 with their errors @xmath71 are shown as a function of angle of incidence @xmath67 . the average quadratic deviation from zero has been determined to @xmath93 which is slightly larger than in the previous case , but reflects inaccuracies in the satellite pointing .
furthermore , this value is probably underestimated because of the incomplete sampling of the field of view .
the statistical error is marginally smaller due to larger photon numbers and has been derived to be @xmath94 .
figure [ fig : scodata_spot ] shows the combined localisation in @xmath12- and @xmath13-direction , converted into the celestial coordinate frame .
the radius of the spot is @xmath95 .
the average attenuation length @xmath40 has been varied to yield the minimum value for @xmath82 . in the case of _
scorpius x-1 _ , we find the optimum value to be @xmath96 , which corresponds well to the theoretically expected value @xmath92 .
in order to assess the extend to which angular resolving power is affected by the source localisation algorithm , the point spread function of the imaging system has been derived . for a coded mask camera like _ wxm _ ,
the point spread function is equal to the autocorrelation @xmath97 of the mask pattern @xmath46 , i.e. @xmath98(x)$ ] . to be exact , @xmath97 describes the isotropised point spread function .
anisotropies provoke asymmetries in the central correlation peak of the function @xmath99(x)$ ] .
again , the determination of @xmath97 is carried out in fourier - space .
diagram [ dia : psf ] illustrates the derivation of @xmath100 $ ] : @xmath101 the width @xmath102 of the autocorrelation function @xmath97 has been determined by fitting a gaussian to the central peak .
the average width @xmath103 of the point spread function has been defined as the geometric mean of its extend in @xmath12- and @xmath13-direction ( @xmath104 ) , because the quantity of interest is the size of the spot a source is imaged onto .
the value of @xmath97 at the origin corresponds to the correlation strength @xmath105 .
accordingly , the average correlation strength @xmath106 has been defined to be equal to the arithmetic mean of @xmath105 and @xmath107 because of its analogy to brightness .
the angles of incidence have been restricted to @xmath108 , in order to ensure that the mask pattern is fully imaged onto the detector .
the source positions were arranged on a square lattice with mesh size of @xmath109 .
the average attenuation length @xmath40 has been set to , corresponding to a power law spectrum with spectral index @xmath32 .
the left plot in figure [ fig : psf ] illustrates , that the width @xmath103 of the autocorrelation function increases and how the correlation strength decreases correspondingly at larger distances from the optical axis . at increasingly larger distance from the optical axis , the functions @xmath33 and @xmath110 disperse the correlation peak .
because the area underneath the peak has to be conserved , the correlation strength @xmath106 drops accordingly .
thus , the angular resolution power has decreased thus by @xmath111 for a burst at @xmath112 compared to a burst on the optical axis . [ cols="^,^ " , ]
b.m.s . wishes to thank y. shirasaki for help on the monte carlo generator .
the support of the german national merit foundation is greatly appreciated .
we thank all members of the hete-2 team , who participated in the construction and operation of the satellite as well as the anonymous referee for valuable comments . 00
w. s. paciesas at al .
, preprint astro - ph/9903205 r. jager et al .
, preprint astro - ph/9706065 l. piro et al , astron
. astrophys .
329 , 906 - 910 ( 1998 ) e. costa et al . ,
preprint astro - ph/9796965 n. kawai et al . , astron .
138 , 563 - 564 ( 1999 ) j.j.m .
int zand et al . , proceedings of imaging in high - energy astronomy , capri ( 1994 ) | the high - energy transient explorer ( hete-2 ) , launched in october 2000 , is a satellite experiment dedicated to the study of @xmath0-ray bursts in a very wide energy range from soft x - ray to @xmath0-ray wavelengths . the intermediate x - ray range ( 2 - 30 kev ) is covered by the wide - field x - ray monitor _ wxm _ , a coded aperture imager . in this article , an algorithm for reconstructing the positions of @xmath0-ray bursts
is described , which is capable of correcting systematic aberrations to approximately @xmath1 throughout the field of view .
functionality and performance of this algorithm have been validated using data from monte carlo simulations as well as from astrometric observations of the x - ray source _
scorpius x-1_. epsf , x - ray astronomy , coded - mask imaging , @xmath0-ray bursts 95.55.ka,98.70.rz,95.75.mn |
it has been known since the advent of mathematical optics @xcite that lenses with certain curvature and/or a spatially varying refractive index allow for focussed images .
indeed , the simplest example of a convergent lens is a curved piece of silica which bends light according to the snell - descartes laws of refraction at curved air / glass interface .
however , such a lens is necessarily of a convex shape . on the contrary
, a medium with a continuously changing refractive index can act as a flat convergent lens : gradient - index optics is indeed the branch of optics covering optical effects produced by a gradual variation of the refractive index of a material with the famous paradigm of the maxwell fisheye , first published by james clerck maxwell in 1854 under the pseudonym of charles anthony swainson @xcite , but however still of current interest @xcite .
a continuously refracting index can also create optical illusions known as mirages : in the desert , these are due to the variation of refractive index between the hot , less dense air at the surface of the road , and the denser cool air above it .
the variation in temperature ( and thus density ) of the air actually causes a gradient in its refractive index , causing it to increase with height .
this index gradient causes anomalous refraction of light rays ( at a shallow angle to the road ) from the sky , bending them into eye of the viewer , with their apparent location being the road s surface .
similar effects occur at the surface of the sea , and are known as fata morgana , after the name of the fairy in the arthurian legend who lifted castles over lakes .
inverted optics include perfect lenses via negative refraction @xcite , that can also be designed using transformational optics by mapping the image plane onto the source plane , which causes a negative ( possibly spatially varying ) refractive index @xcite .
maybe the most stricking example of an optical illusion is the invisibility cloak independently proposed by pendry , schurig and smith , and leonhardt @xcite , which can steer light around an object which is then invisible to an external observer ( however , light emanating from this object is itself confined within the invisibility cloak ! ) .
the underlying mechanism of such cloaks relies upon a mirage effect , since a source located inside the coating seems to radiate from a shifted location @xcite .
this coating actually consists of a meta - material whose physical properties are deduced from a coordinate transformation in the maxwell system , and this viewpoint allows for a refreshed analysis of geodesics of light in complex media , since the maxwell equations retain their form under geometric changes and the only thing to do in the transformed coordinates is to replace the original medium ( often isotropic homogeneous ) by an anisotropic heterogeneous medium @xcite ( a meta - material ) .
such geometric transforms are very useful in computational electromagnetism to model unbounded @xcite and twisted @xcite domains in the context of optical fibres .
however , the physical interest in transformational optics has been fuelled by potential applications in the design of meta - materials , while the electrical engineering community has used geometric transforms as a mean to simplify numerical models involving large or thin domains for the past thirty years .
coming back to the invisibility cloak , the anisotropy and the heterogeneity of the parameters of its coat work as a deformation of the optical space around the object .
the first experimental validation @xcite of these theoretical considerations was given , a few months after the proposals of pendry et al . , and leonhardt were released in the press .
a copper cylinder was indeed made invisible to an incident plane wave at @xmath0 ghz as predicted by the numerical simulations .
this markedly enhance our capabilities to manipulate light , even in the extreme near field limit @xcite .
however , such cloaks suffer from an inherent narrow bandwidth as their transformation optics design leads to singular tensors on the frontier of the invisibility region which is a consequence of tearing apart the metric when one makes a hole in optical space ( obtained by blowing up a point @xcite ) .
physically , light has to curve its trajectory around the hole ( or invisibility region ) ; hence , to match the phase of a wave propagating in homogeneous space , it must travel faster . to remove the cloak s singularity , xiang et al . proposed to consider the blowup of a segment instead of a point @xcite , but this cloak only works for certain directions .
on the other hand , leonhardt and tyc considered a stereographic projection of a virtual hyper - sphere in a four dimensional space @xcite , which bears some resemblance with the construction of a maxwell fisheye .
as an alternative to non - singular cloaking li and pendry proposed a one - to - one geometric transform from a flat to a curved ground : their invisibility carpet @xcite is by essence non singular and thus broadband .
this proposal led to a rapid experimental progress in the construction of two - dimensional carpets approaching optical frequencies @xcite . another way to make
cloaks broadband is to approximate their parameters using a homogenization approach , which leads to nearly ideal cloaking @xcite , as it does not rely upon locally resonant elements . in 2008 ,
some of us @xcite demonstrated broadband cloaking of transverse electric waves using a micro - structured metallic cloak in the micro - wave regime .
this naturally prompts the question of whether , at optical frequencies , an object lying onto a metal film could be cloaked from a propagating surface plasmon polaritons ( spps ) .
the less than usual physics of the transmission of light through holes small compared with the wavelength is has given rise to a vast amount of literature following the 1998 paper by ebbesen et al .
@xcite , but some heralding earlier work combining both theory and experiment is less well known @xcite .
martin - moreno and garcia - vidal further showed in 2004 that one can manipulate surface plasmon ad libitum via homogenization of structured surfaces @xcite . in the same vein , pioneering approaches to invisibility relying upon plasmonic metamaterials have already led to fascinating results @xcite .
these include plasmonic shells with a suitable out - of - phase polarizability in order to compensate the scattering from the knowledge of the electromagnetic parameters of the object to be hidden , and external cloaking , whereby a plasmonic resonance cancels the external field at the location of a set of electric dipoles .
recently , baumeier et al . have demonstrated theoretically and experimentally that it is possible to reduce significantly the scattering of an object by a spp , when it is surrounded by two concentric rings of point scatterers @xcite . in the present paper
, we would like to extend the control of electromagnetic fields using tools of geometric transforms to the area of surface plasmons . to do this , we adopt the general form for the time - harmonic electric and magnetic field , @xmath1 and @xmath2 in @xmath3 .
@xmath4 where @xmath5 denotes the angular frequency and @xmath6 is the time variable .
we then consider the time - harmonic maxwell s equations in the initial space coordinates ( x , y , z ) : @xmath7 where the tensors of permittivity and permeability @xmath8 and @xmath9 describe the original ( possibly heterogeneous anisotropic ) medium . in the transformed coordinate system ( u , v , w )
, the structure of the maxwell equations is preserved : @xmath10 but we note that the space derivatives are now taken with respect to the transformed coordinates ( hence , the transformed gradient @xmath11 ) , and obviously the permittivity and permeability tensors @xmath12 and @xmath13 should have new expressions .
an elegant way to identify the tensors is to consider the jacobian matrix @xmath14 associated with the change of co - ordinate system : @xmath15 it follows that @xcite the transformation rule for the tensors is : @xmath16 with : @xmath17 where @xmath18 is the determinent of the jacobian matrix .
when the original permittivity and permeability matrices are scalar , their transformed counterparts are given by : @xmath19 we now wish to apply this general rule to the case of a transverse magnetic ( @xmath20-polarized ) spp propagating in the positive @xmath21 direction at the interface @xmath22 between metal ( @xmath23 ) and air ( @xmath24 ) : @xmath25 where @xmath26 is the speed of light in vacuum and @xmath27 has the usual drude form in the metal(@xmath23 ) , for which @xmath28 is the plasma frequency ( @xmath29 thz ) of the _ free electron gas _ and @xmath30 is a characteristic collision frequency of about @xmath31 thz @xcite .
moreover , the upper medium is described by spatially varying tensors of permittivity @xmath12 and permeability @xmath13 which for simplicity are assumed to be represented in a diagonal basis i.e. @xmath32 and @xmath33 .
the dispersion relation for the surface polariton at such an anisotropic interface takes the following form @xcite @xmath34 spps are bound to the interface , hence do not belong to the radiative spectrum ( unlike leaky waves ) .
in the present section we show numerically that one can control the spp field using transformational plasmonics .
a review of transformations leading to arbitrary carpets , concentrators and rotators is presented .
so far , such a mathematical derivation of the spp dispersion relation seems fairly abstract .
however , we now wish to analyse the interaction of this spp with a specific anisotropic heterogeneous structure , in the present case a three - dimensional invisibility carpet @xcite , deduced from the following geometric transformation : @xmath35 importantly , we note that this material is not only heterogeneous anisotropic but also magnetic , which seems a far technological reach .
however , these constraints can be further relaxed using some quasi - conformal grids in the spirit of li and pendry s work for two - dimensional carpets @xcite .
here @xmath36 is a stretched coordinate .
it is easily seen that this linear geometric transform maps the surface @xmath37 of the horizontal plan @xmath38 onto the surface @xmath39 , and it leaves the surface @xmath40 unchanged .
the surfaces @xmath41 and @xmath42 are assumed to be differentiable , and this ensures that the carpet wo nt display any singularity on its inner boundary . the symmetric tensors @xmath12 and @xmath13 are fully described by five non vanishing entries in a cartesian basis : [ fig1 ] [ fig2 ] @xmath43 where @xmath44 and with @xmath45 the jacobian matrix of the transformation .
furthermore , the derivatives of @xmath46 with respect to @xmath47 and @xmath48 are given by : @xmath49 the purpose of the next section is to show that such carpets work equally well for electromagnetic and plasmonic fields , as is intuitively the case when looking at the limit of ray optics , see figures [ fig1 ] and [ fig2 ] : this is due to the fact that the transformed medium is valid for any field solution of the maxwell s equations ( which is obviously the case for spps ) .
[ fig3a ] we now wish to apply the recipe for the design of 3d carpets to specific geometrical transformations in order to hide gaussian and conical shaped bumps on a metallic surface .
we first present the projection of the conical surface on the flat surface , as this case is more intuitive .
the geometric transform used in such a design is as follows : @xmath50 @xmath51 @xmath52 we report in figure [ fig3a ] the finite element computations which exemplify the role played by the carpet in the control of the spp wavefront .
[ fig3b ] we now present the projection of the conical surface on the flat surface , as this case is less intuitive . for this , we consider two surfaces @xmath41 and @xmath42 which describe the inner and outer boundaries of the carpet : @xmath53 explicit expressions for the partial derivatives are @xmath54 we report the finite element computations in figure [ fig3b ] .
we can see that control of electromagnetic and plasmonic fields is achieved for both the smooth ( gaussian ) and rough ( conical shaped ) bumps .
we emphasize that these numerical results are non - intuitive as the scattering by a conical shaped surface is one of the hardest cases to handle . in order to see this ,
let us remind some classical result of singularity theory for a sector ( the two - dimensional counterpart of a conical shaped domain ) .
let us consider the laplace problem in a semi - infinite sector bounded by two edges @xmath55 and @xmath56 : @xmath57 which we can express in polar coordinates : @xmath58 and let us assume that dirichlet boundary conditions on the edges @xmath59 .
using separation of variables , we find that @xmath60 it is well - known @xcite that if @xmath61 is square integrable , then @xmath62 is of finite energy , that is @xmath63 therefore , we need to ensure that the following expressions are finite : @xmath64 @xmath65 we note that @xmath66 is finite if and only if @xmath67 . this condition is not met for @xmath68 .
hence , all coefficients @xmath69 vanish in ( [ fousect ] ) .
the solution of the dirichlet problem is thus given by : @xmath70 from this expression , we can see that the field @xmath62 is singular near the tip of the sector i.e. when @xmath71 is close to zero .
similar expresions hold for a conical shaped region in the three - dimensional case .
however , when the sector / conical shaped region is filled with an heterogeneous anisotropic medium ( for instance deduced from a geometric transform ) , the laplace problem reads as : @xmath72 where @xmath73 is a matrix - valued function .
the solution @xmath62 is now very different from above , and its behaviour around the tip of the sector / conical shaped region therefore changes dramatically .
this simple analysis exemplifies the role of the transformed medium in the regularization of the field scattered by a conical shaped bump surrounded by a carpet .
the goal of this section is to show that we can also launch a spp on a cylindrical surface using the transformational optics approach in order to guide smoothly the spp which propagates on a flat metal surface to the top of a cylindrical bump lying on the surface .
more precisely , the transformed medium consists of two parts : on one hand , the quarter of a sphere ( shown in pink color in figure [ fig4 ] ) requires the following transform : @xmath74 on the other hand , the cylindrical part ( shown in yellow color in figure [ fig4 ] ) requires the following transform : @xmath75 we depict in figure [ fig4 ] the typical trajectory for a spp launched on the tubular surface surrounded by the carpet consisting of these two parts .
we report in figure [ fig5 ] the associated finite element computations .
[ fig4 ] [ fig5 ] in this section we consider an arbitrary shaped rotator described as follows : @xmath76 one can see in figure [ fig7](b ) that the spp is smoothly rotated within the rotator . moreover , the concentrator is itself invisible for spps as the wavefronts are unperturbed outside this metamaterial .
[ fig6 ] more precisely , we consider the geometric transform : @xmath77 + @xmath78 + @xmath79 + in this section we consider an arbitrary shaped concentrator described as follows : @xmath80 one can see in figure [ fig7](a ) that the spp is smoothly squeezed within the concentrator .
moreover , the concentrator is itself invisible for spps as the wavefronts are unperturbed outside this metamaterial .
[ fig7 ] more precisely , we consider the following geometric transform : @xmath81 @xmath82 @xmath83 : @xmath84 @xmath85 : @xmath86
[ fig8 ] in order to illustrate analytically the control of the evanescent component of the spp , we consider the case of a curved surface which is invariant along the @xmath21-direction , see figure [ fig8 ] .
following the paper by della valle and longhi @xcite where a detailed discussion of the role played by the geometric potential on the spp envelope in the direction of propagation led to the evolution equation : @xmath87 in the case of two isotropic media .
della valle and longhi used a two - scale asymptotic expansion in the electromagnetic field , assuming that the local curvature of the interface is small , i.e. @xmath88 with @xmath89 and @xmath90 is a small positive parameter .
importanttly , the asymptotic expansions are assumed as follows : @xmath91 with @xmath92 and @xmath93 i.e. for p- and s - field components .
moreover , the corrective terms are functions of the slow variable @xmath94 .
the former authors obtain a hierarchy of auxilliary equations at successive powers of @xmath90 . at the leading order @xmath95 , they retrieve the usual effective index of the spp : @xmath96 and proceed with the interface conditions for the field to obtain the asymptotic expansion of the decay length : @xmath97 where the corrective terms @xmath98 undergo some oscillations on the slow spatial scale @xmath99 .
regarding the penetration depth of the spp in the 2 media , della valle and longhi deduce the following expression : @xmath100 from the solvability condition : @xmath101 at order @xmath102 .
we performed a similar analysis for a transformed ( anisotropic heterogeneous ) medium above the curved surface .
the transformed medium is assummed to be described by diagonal tensors of permittivity and permeability @xmath103 where the geometric potential is given by : @xmath104 in the local coordinate basis , we obtain the same evolution equation , with a new effective index @xmath105 where the geometric potential is now given by : @xmath106
we have demonstrated the full control of surface plasmon polariton propagation by extending the transformational optics tools to the area of plasmonics .
we have shown that we can markedly reduce the scattering of spp on a bumped surface thanks to invisibility carpets whose design is the same as for an out - of - plane electromagnetic field incident upon the metal surface .
the versatility of the designs proposed in this paper illustrates the power of transformational plasmonics .
luneburg , mathematical theory of optics , ( university of california press , los angeles 1964 ) .
maxwell , `` problems , '' camb .
dublin math .
j. 8 , 188 - 189 ( 1854 ) .
u. leonhardt and t.g .
philbin , `` transformation optics and the geometry of light , '' prog .
53 , 70 - 152 ( 2009 ) .
s. guenneau , a. diatta and r.c .
mcphedran , `` focusing : coming to the point in metamaterials , '' j. mod .
optics , 57 , 511 - 527 ( 2010 ) .
pendry , negative refraction makes a perfect lens , phys .
85 , 3966 - 3969 ( 2000 ) .
ramakrishna , j.b .
pendry , m.c.k .
wiltshire and w.j .
stewart imaging the near - field , j. mod .
optics * 50 * , 1419 - 1430 , ( 2003 ) .
pendry and s.a .
ramakrishna , j. phys .
matter * 15 * , 6345 , ( 2003 ) .
s. guenneau , b. gralak and j.b .
pendry , perfect corner reflector , opt . lett .
* 30 * , 146 , ( 2005 ) . s. guenneau and s.a .
ramakrishna , negative refractive index , perfect lenses and checkerboards : trapping and imaging effects in folded optical spaces , comptes rendus - physique , doi : 10.1016/j.crhy.2009.04.002 j.b .
pendry , d. schurig , and d.r .
smith , `` controlling electromagnetic fields , '' science , * 312 * , 1780 ( 2006 ) .
a.j . ward and j.b .
pendry , refraction and geometry in maxwell s equations , j. mod . opt . * 43 * , 773 - 793 ( 1996 ) .
a. nicolet , j.f .
remacle , b. meys , a. genon , w. legros , `` transformation methods in computational electromagnetism , '' j. appl . phys . * 75 * , 6036 - 6038 , ( 1994 ) .
s. guenneau , a. nicolet , f. zolla , c. geuzaine et b. meys , a finite element formulation for spectral problems in optical fibers , international journal for computation and mathematics in electrical and electronic engineering compel * 20 * , 120 - 131 , ( 2001 ) .
a. nicolet , f. zolla et s. guenneau , `` a finite element modelling for twisted electromagnetic waveguides , '' european journal of physics , applied physics , vol .
28 , 153 - 157 , ( 2004 ) .
u. leonhardt , `` optical conformal mapping , '' science * 312 * , 1777 ( 2006 ) .
d. schurig , j.j .
mock , b.j .
justice , s.a .
cummer , j.b .
pendry , a.f .
starr , d.r .
smith,``metamaterial electromagnetic cloak at microwave frequencies , '' science * 314 * , 977 ( 2006 ) .
f. zolla , s. guenneau , a. nicolet , and j.b .
pendry , `` electromagnetic analysis of cylindrical invisibility cloaks and the mirage effect '' , opt .
lett . * 32 * , 1069 - 1071 ( 2007 ) .
a. greenleaf , m. lassas and g. uhlmann , `` on nonuniqueness for calderons inverse problem , '' math .
* 10 * , 685 - 693 ( 2003 ) .
u. leonhardt and t. tyc , `` broadband invisibility by non - euclidean cloaking , '' science * 323 * , 110 ( 2009 ) .
jiang et al .
`` invisibility cloak without singularity , '' appl .
. lett . * 93 * , 194102 ( 2008 ) .
g. i. stegeman , `` long range surface plasmons in birefringent media , '' applied optics * 22*(15 ) , 2243 - 2245 ( 1983 ) .
w. cai , u.k .
chettiar , a.v .
kildiev and v.m .
shalaev , `` optical cloaking with metamaterials '' , nature * 1 * , 224 - 227 ( 2007 ) .
m. farhat , s. guenneau , a.b .
movchan and s. enoch , `` achieving invisibility over a finite range of frequencies , '' opt .
express * 16 * , 5656 - 5661 ( 2008 ) a. greenleaf , y. kurylev , m. lassas and g. uhlmann , `` isotropic transformation optics : approximate acoustic and quantum cloaking , '' new j. physics * 10 * , 115024 ( 2008 ) j. li and j.b .
pendry , `` hiding under the carpet : a new strategy for cloaking , '' phys .
* 101 * , 203901 ( 2008 ) .
r. liu , c. ji , j.j .
mock , j.y .
chin , t.j .
cui and d.r .
smith , `` broadband ground - plane cloak , '' science * 323 * , 366 ( 2008 ) .
j. valentine , j. li , t. zentgraf , g. bartal and x. zhang , `` an optical cloak made of dielectrics . ''
nature mater .
* 8 * , 569 -571 ( 2009 ) .
gabrielli , j. cardenas , c.b .
poitras and m. lipson , `` silicon nanostructure cloak operating at optical frequencies , '' nature photonics ( 2009 ) .
doi:10.1038/nphoton.2009.117 i.i .
smolyaninov , v.n .
smolyaninova , a.v .
kildishev and v.m .
shalaev , `` anisotropic metamaterials emulated by tapered waveguides : application to optical cloaking '' , phys
lett . * 102 * , 213901 ( 2009 ) e. kallos , c. argyropoulos and y. hao , `` ground - plane quasicloaking for free space , '' phys .
a * 79 * , 063825 ( 2009 ) t. w. ebbesen , h. j. lezec , h. f. ghaemi , t. thio , p. a. woff , `` extraordinary optical transmission through sub - wavelength hole arrays , '' nature * 391 * , 667 ( 1998 ) .
p. j. bliek , r. deleuil , l. c. botten , r. c. mcphedran , and d. maystre , `` inductive grids in the region of diffraction anomalies - theory , experiment , and applications , '' ieee mtt * 28 * ( 10 ) , 11191125 ( 1980 ) .
pendry , l. martin - moreno and f.j .
garcia - vidal , `` mimicking surface plasmons with structured surfaces , '' science * 305 * , 847 ( 2004 ) .
nicorovici , r.c .
mcphedran and g.w .
milton,``optical and dielectric properties of partially resonant composites , '' phys .
b * 49 * , 8479 - 8482 ( 1994 ) .
a. alu and n. engheta ,
`` achieving transparency with plasmonic and metamaterial coatings , '' phys .
e * 72 * 016623 ( 2005 ) .
f.j . garcia de abajo , g. gomez - santos , l.a .
blanco , a.g .
borisov and s.v .
shabanov , `` tunneling mechanism of light transmission through metallic films , '' phys .
lett . * 95 * 067403 ( 2005 ) .
b. baumeier , t.a .
leskova and a.a .
maradudin , `` cloaking from surface plasmon polaritons by a circular array of point scatterers , '' phys .
103 * , 246809 ( 2009 ) .
guenneau and f. zolla , `` homogenization of three - dimensional photonic crystals , '' progress in electomagnetic research * 27 * , 91 - 127 ( 2000 ) .
h. chen and c. t. chan , appl .
* 91 * , 183518 ( 2007 ) .
cummer , b.i .
popa , d. schurig , d.r .
smith , j. pendry , m. rahm , and a. starr , phys .
* 100 * , 024301 ( 2008 ) .
zayats , i.i .
smolyaninov , a.a .
maradudin , `` nano - optics of surface plasmon polaritons , '' phys
* 408 * , 131 - 314 ( 2005 ) .
m. rahm , d. schurig , d.a .
roberts , s.a .
cummer , d.r .
smith and j.b .
pendry , `` design of electromagnetic cloaks and concentrators using form - invariant coordinate transformations of maxwell s equations , '' photon .
fundam appl .
* 6 * ( 1 ) , 87 - 95 ( 2008 ) .
y. luo , j. zhang , j. wu , bae - ian and h. chen , `` interaction of an electromagnetic wave with a cone - shaped invisibility cloak and polarization rotator , '' phys .
b * 78 * ( 12 ) , 125108 ( 2008 ) .
huidobro , m.l .
nesterov , l. martin - moreno and f.j .
garca - vidal , `` transformation optics for plasmonics , '' http://arxiv.org/abs/1003.1154 y. liu , t. zentgraf , g. bartal and x. zhang , `` transformational plasmon optics , '' http://arxiv.org/abs/1003.1326 j. renger , m. kadic , g. dupont , s. acimovic , s. guenneau , r. quidant and s. enoch , `` hidden progress : broadband plasmonic invisibility , '' opt .
express * 18*(15 ) , 15757 - 15768 ( 2010 ) . m. kadic , s. guenneau , and s. enoch ,
`` transformational plasmonics : cloak , concentrator and rotator for spps , '' opt . express * 18*(11 ) , 1202712032 ( 2010 ) .
palik , `` handbook of optical constants of solids , '' academic , london , ( 1985 ) g. dupont , s. guenneau and s. enoch , `` electromagnetic analysis of arbitrarily shaped pinched carpets , '' physical review a * 82*(3 ) , 033840 , ( 2010 ) .
a.b . movchan and n.v .
movchan , `` mathematical modelling of solids with nonregular boundaries , '' crc mathematical modelling series , ( 1995 ) .
g. della valle and s. longhi , `` geometric potential for plasmon polaritons on curved surfaces , '' j. phys .
b : at . mol .
43 ( 2010 ) . | transformational optics allow for a markedly enhanced control of the electromagnetic wave trajectories within metamaterials with interesting applications ranging from perfect lenses to invisibility cloaks , carpets , concentrators and rotators . here
, we present a review of curved anisotropic heterogeneous meta - surfaces designed using the tool of transformational plasmonics , in order to achieve a similar control for surface plasmon polaritons in cylindrical and conical carpets , as well as cylindrical cloaks , concentrators and rotators of a non - convex cross - section .
finally , we provide an asymptotic form of the geometric potential for surface plasmon polaritons on such surfaces in the limit of small curvature .
transformational optics , surface plasmon polariton , carpet , cloak , concentrator , rotator |
in an outreach paper written in the heyday of elemental fermiology , mackintosh remarked that while `` few people would define a metal as a ` solid with a fermi surface ' , this may nevertheless be the most meaningful definition of a metal.''@xcite . in opposition to a metal
, a semiconductor can be defined as a solid deprived of a fermi surface . doping a semiconductor@xcite ,
can eventually turn it to to metal .
this metal - insulator transition@xcite has been a central issue of the condensed - matter physics during the second half of the twentieth century .
the most - explored system has been silicon doped with phosphorus , its immediate neighbor to the right in the periodical table .
experiment has identified a sharp transition and has accurately measured the critical concentration of dopants for emergence of metallicity@xcite . above this threshold concentration ,
silicon becomes metallic : it conducts electricity at zero - temperature .
this is an alternative definition of a metal@xcite and more inclusive than the one mentioned above .
historically , the debate on metal - insulator transition have been focused on `` weak metallicity '' , which only requires the presence of mobile electrons at zero temperature and not `` strong metallicity '' , which implies the existence of a well - defined fermi surface . in a variety of insulators ,
the critical concentration of dopants for metal - insulator transition has been identified by experiment @xcite and the results match the expectations of the mott criterion for metal - insulator transition@xcite , which identifies the effective bohr radius of the host insulator as the key parameter in a given material .
[ adapted from ref.@xcite ) ] .
they attest to the existence of a well - defined fermi surface and quantify the amplitude of the fermi energy .
c ) the results imply that the chemical potential is 1mev above the bottom of the conduction band separated from the valence band by a gap of 3 ev .
d ) the seafloor is this shallow fermi sea is rough and the low - temperature mean - free - path of electrons is set by the average distance between islands or trenches , which are potential scattering centers.,width=529 ] according to the pauli exclusion principle , electrons of a solid which share the same discrete quantum numbers ( such as spin and valley ) can not have the same wave - vector .
they will fill k - states in a fermi - dirac distribution with an edge separating the filled and empty states .
the sharpness of this edge is set by the inverse of the mean - free - path , which defines the size of pixel in the k space .
a sharp fermi surface has an experimental signature other than a finite zero - temperature electric conductance .
landau quantification truncates a fermi surface to a number of concentric tubes leading to quantum oscillations in various physical properties@xcite . in many doped semiconductors , such oscillations
have been reported decades ago .
the list includes n - doped strontium titanate@xcite , doped iv - vi semiconductors such as pbs @xcite , pbte@xcite and snte@xcite .
this is also the case of heavily - doped germanium@xcite .
a recent study of nernst effect in srtio@xmath4 found quantum oscillations of a very large magnitude with a single and well - defined frequency ( see fig.1 ) .
this result implies that this wide - gap semiconductor , after loosing one out of its 10@xmath5 oxygen atoms , becomes a metal in the strongest sense of the word : a solid with a fermi surface . from this perspective ,
one unexpected finding of the recent research on topological insulators is instructive .
the quest to find ultra - mobile electrons on the surface protected from backscattering mostly documented _ bulk _
mobile electrons giving rise to easily - detectable quantum oscillations@xcite .
it is now clear that topological insulators are often on the metallic - side of the metal - insulator transition as a consequence of uncontrolled doping and a long bohr radius in the bulk .
the irony was not pursued however .
bulk electrons were often treated as a nuisance , impeding topological `` insulators '' to qualify as true insulators .
their proper mobility , with notable exceptions@xcite , was not put under scrutiny .
the subject of the present paper is the intrinsic limit to carrier mobility in a dilute metal made by the introduction of a random distribution of dopants in a semiconductor with sufficiently long bohr radius .
a pattern is visible in the available experimental data . in these systems , at identical carrier density , mobility of electrons is much higher than in metallic silicon or metallic germanium .
moreover , the variation with carrier concentration is quite distinct .
the main argument of this paper is that both the magnitude of the mobility and its density dependence can be understood by conceiving the simplest model for the probability of finding a scattering center in a shallow fermi sea .
the key parameter remains the bohr radius of the host insulator , which by setting the thomas - fermi screening length , shapes the local seafloor of the fermi sea .
1 occurs almost simultaneously . on the other hand , in pbte and srtio@xmath6 , which have both a long bohr radius , mobility enhances with decreasing carrier concentration and localization is easily avoided .
solid horizontal lines mark the required mobility to see detect quantum oscillations at a given magnetic field.,width=529 ] fig . 2
compares the hall mobility at liquid - helium temperature of of four different semiconductors as a function of carrier concentration between @xmath7 and @xmath8 .
the reported data for phosphorus - doped silicon@xcite , arsenic - doped@xcite , antimony - doped@xcite and gallium - doped@xcite germanium are presented together with those for p - doped@xcite and n - doped@xcite pbte , as well as for oxygen - deficient srtio@xmath4 @xcite .
the figure reveals several remarkable features .
.normalised electric permittivity , @xmath9 , the average effective , @xmath10 , and the effective bohr radius ( a@xmath11 ) , in four different semiconductors . [ cols="^,^,^,^,^",options="header " , ]
this paper puts under scrutiny an experimental fact already reported but rarely analyzed . in metallic semiconductors with a long bohr radius ,
mobility is large and continuously decreases with increasing carrier density .
both these features are absent in silicon or germanium . according to the argument proposed here ,
this happens because in these metallic semiconductors the global distribution of dopants shapes the local seafloor . for this to happen
, a threshold concentration of dopants is required .
this second critical concentration becomes attainable only when the bohr radius is long enough . with a set of most unsophisticated assumptions
, one can formulate an expression for the intrinsic mobility of a metallic semiconductor with random distribution of dopants . in spite of neglecting subtle details ,
the expression is in good agreement with the data on four different semiconductors .
this work is supported by _
agence nationale de la recherche _ through the superfield project .
99 r. a. mackintosh , scientific american , * 209 * , 110 ( 1963 ) .
b. i. shklovskii and a. l. efros , electronic properties of doped semiconductors , springer - verlag , berlin ( 1984 ) .
n. f. mott , metal - insulator transitions , second edition , taylor and francis , london ( 1990 ) .
t. f. rosenbaum , k. andres , g. a. thomas , and r. n. bhatt , phys .
* 45 * , 1723 ( 1980 ) .
p. p. edwards , m. t. j. lodge , f.hensel and r. redmer , phil .
a * 368 * , 941 ( 2010 ) .
p. p. edwards and m. j. sienko , phys . rev .
b * 17 * , 2575 ( 1978 ) .
n. f. mott , can .
* 34 * , 1356 ( 1956 ) .
n. f. mott , philos . mag . * 6 * , 287 ( 1961 ) .
d. shoenberg , magnetic oscillations in metals ( cambridge univ . press , 1984 ) .
b. gregory , j. arthur and g. seidel , _ phys .
* b * 19 , 1039 ( 1979 ) . h. uwe , r. yoshizaki , t. sakudo , a. izumi and t. uzumaki , jap . j. appl
* 24 * , suppl . *
242 * , 335(1985 ) .
x. lin , z. zhu , b. fauqu , and k. behnia , phys .
x , * 3 * , 021002 ( 2013 ) .
s. j. allen _ et al .
_ , phys . rev .
b * 88 * , 045114 ( 2013 ) .
x. lin , g. bridoux , a. gourgout , g. seyfarth , s. krmer , m. nardone , b. fauqu and k. behnia , phys .
* 112 * , 207002 ( 2014 ) .
p. j. stiles , e. burstein , d. langenberg , j. appl .
, suppl . * 32 * , 2174 ( 1961 ) .
r. burke , b. houston and h. savage , phys .
b * 2 * , 1977 ( 1970 ) .
j. jensen , b. houston and j. burke , phys .
b * 18 * , 5567 ( 1978 ) .
j. burke , r. allgaier , b. houston , j. babiskin , and p. siebenmann , phys . rev
* 14 * , 360 ( 1965 ) . h. t. savage , b. houston and j. r. burke , phys
b * 6 * , 2292 ( 1972 ) .
w. bernard , h. roth , w. d. straub , and j. e. mulhern , jr .
, phys . rev . *
135 * , a1386 ( 1964 ) .
m. nakamura , j. phys .
* 22 * , 830 ( 1967 ) .
g. carrre , j. p. ulmet , s. askenazy and r. barbaste , phys .
( b ) * 64 * , 729 ( 1974 ) .
n. p. butch , k. kirshenbaum , p. syers , a. b. sushkov , g. s. jenkins , h. d. drew , and j. paglione , phys .
b * 81 * , 241301(r ) ( 2010 ) .
h. cao , j. tian , i. miotkowski , t. shen , j. hu , s. qiao and y. p. chen , phys .
lett . * 108 * , 216803 ( 2012 ) . b. fauqu , n. butch , p. syers , j. paglione , s. wiedmann , a. collaudin , b. grena , u. zeitler and k. behnia , phys . rev .
b * 87 * , 035133 ( 2013 ) .
e. lahoud , e. maniv , m. s. petrushevsky , m. naamneh , a. ribak , s. wiedmann , l. petaccia , z. salman , k. b. chashka , y. dagan and a. kanigel , phys .
b * 88 * , 195107 ( 2013 ) . b. skinner , t. chen and b. i. shklovskii , phys . rev . lett .
* 109 * 176801 ( 2012 ) . c. yamanouchi , k. mizuguchi , w. sasaki , j. phys .
* 22 * , 859 ( 1967 ) .
m. j. katz , s. h. koenig and a. a. lopez , phys .
lett . * 15 * , 828 ( 1965 ) .
w. sasaki , suppl .
* 57 * , 225 ( 1975 )
. m. watanabe , y. ootuka , k. m. itoh and e. e. haller , phys .
b * 58 * , 9851 ( 1998 ) .
r. s. allgaier , and w. w. scanlon , phys . rev . * 111 * , 1029 ( 1958 ) .
h. p. r. frederikse and w. r. hosler , phys . rev . * 161 * , 822 ( 1967 ) . c. s. koonce , m. l. cohen , j. f. schooley , w. r. hosler and e. r. pfeiffer , phys
. rev . * 163 * , 380 ( 1967 ) .
a. spinelli , m. a. torija , c. liu , c. jan , and c. leighton , phys .
b * 81 * , 155110 ( 2010 ) .
x. lin , b. fauqu and k. behnia , science * 349*,945(2015 ) .
j. ziman , principles of the theory of solids , cambridge university press , cambridge ( 1960 ) .
n. w. ashcroft , j. non - crystalline solids * 156 - 158 * , 621 ( 1993 ) .
r. n. bhatt and t. m. rice , phys .
b * 23 * , 1920 ( 1981 ) .
r. n. bhatt , physica * 146 b * , 99 ( 1987 ) .
p. p. edwards , t. v. ramakrishnana avd c. n. r. rao , j. phys . chem . * 99 * , 5228 ( 1995 ) . i. f. newman and d. f. holcomb * 28 * , 638 ( 1983 ) .
g. a. thomas , m. capizzi , f. derosa , r. n. bhatt , and t. m. rice , phys .
b * 23 * , 5472 ( 1981 ) .
j. son , p. moetakef , b. jalan , o. bierwagen , n. j. wright , r. engel - herbert and s. stemmer , nature mat .
* 9 * , 482 ( 2010 ) .
m. novak , s. sasaki , k. segawa and y. ando , phys .
b * 91 * , 041203(r ) ( 2015 ) . s .- y .
xu , y. xia , l. a. wray , s. jia , f. meier , j. h. dil , j. osterwalder , b. slomski , a. bansil , h. lin , r. j. cava , and m. z. hasan , science * 332 * , 560 ( 2011 ) . m. capizzi and a. frova , phys .
lett . * 25 * , 1298 ( 1970 ) .
j. o. dimmock , i. melngailis , and a. j. strauss , phys .
* 16 * , 11193 ( 1966 ) . | several doped semiconductors , in contrast to heavily - doped silicon and germanium , host extremely mobile carriers , which give rise to quantum oscillations detectable in relatively low magnetic fields .
the small fermi energy in these dilute metals quantifies the depth of the fermi sea .
when the carrier density exceeds a threshold , accessible thanks to the long bohr radius of the parent insulator , the local seafloor is carved by distant dopants . in such conditions , with a random distribution of dopants , the probability of finding an island or a trench depends on on the effective bohr radius , a@xmath0 and the carrier density , n. this picture yields an expression for electron mobility with a random distribution of dopants : @xmath1(a@xmath2 n@xmath3 , in reasonable agreement with the magnitude and concentration dependence of the low - temperature mobility in three dilute metals whose insulating parents are a wide - gap ( srtio@xmath4 ) , a narrow - gap ( pbte ) and a `` zero''-gap ( tlbisse ) semiconductor . |
we consider the global existence of dynamics of classical electromagnetism for extended rigid charges .
to put our work into proper perspective we shall introduce a number of theories ml , ml - si , mld , wf , and @xmath0 , @xmath1 , and @xmath2 .
the former are theories for point - charges , the latter are the theories modified by smeared out charges , i.e. extended rigid charges indicated by the charge distribution @xmath3 .
it will become clear in a moment , why it is helpful to introduce these notations .
ml stands for maxwell - lorentz electrodynamics - the textbook electromagnetism @xcite .
it is the theory of interaction between electromagnetic radiation and charged matter .
the electrodynamic field is represented by an antisymmetric second - rank tensor field @xmath4 on four - dimensional minkowski space @xmath5 equipped with a metric tensor @xmath6 .
charged matter is described by the four - vector charge - current density @xmath7 on @xmath8 . for a prescribed current @xmath7
the time evolution of the electrodynamic field @xmath4 is ruled by the maxwell equations @xmath9 where we have used einstein s summation convention for greek indices , i.e. @xmath10 , and denote the partial derivative with respect to the standard unit vectors in @xmath11 by @xmath12 , @xmath13 . in turn , for a given electromagnetic field @xmath4 , the motion of @xmath14 point - like particles , which are represented by world lines @xmath15 , @xmath16 with labels @xmath17 , obey the lorentz force law @xmath18 here @xmath19 denotes the mass and @xmath20 is a coupling constant ( the charge ) of the @xmath21-th particle .
the overset dots denote differentiation with respect to the world line parametrization @xmath22 .
ml is defined by the system of equations ( [ eqn : original maxwell ] ) and ( [ eqn : lorentzforce ] ) coupled by @xmath23 where @xmath24 denotes the four - dimensional dirac delta distribution .
unfortunately and well known , ml is merely a formal set of equations .
the system has no solutions .
the reason is that the self - field , the field created by a point - charge and acting back on it , is infinite at the position of the point - charge . for reasons which become clear
soon we recall the nature of this singularity . due to the linearity of the maxwell equations
we may decompose the field @xmath4 into the sum of fields @xmath25 , @xmath26 , fulfilling @xmath27 and write ( [ eqn : lorentzforce ] ) as @xmath28 again by linearity the solutions @xmath25 can be decomposed into a special solution and an arbitrary homogeneous solution @xmath29 , i.e. a solution of the homogeneous maxwell equations ( [ eqn : cea m eq for ji ] ) for @xmath30 : @xmath31_{+}+f[z_k]_{-}\right).\end{aligned}\ ] ] the special solutions @xmath32 , f_{-}[z_k]$ ] are the well known advanced and retarded linard - wiechert fields of the @xmath33-th world line @xcite given by @xmath34^\mu_{\pm}(x):=e\frac{\dot z_i^\mu(\tau^\pm)}{(x - z_i(\tau^\pm))_\nu\dot z_i^\nu(\tau^\pm ) } , & & z^0_i(\tau^\pm)=x^0\pm\|{{\mathbf{x}}}-{{\mathbf{z}}}_i(\tau^\pm)\| , \end{aligned}\ ] ] where we use the notation @xmath35 for @xmath36 .
the square brackets emphasize that these fields are functionals of the world line @xmath37 ; note that @xmath38 in ( [ eqn : lw fields ] ) are implicitly defined . now
( [ eqn : lw fields ] ) shows that the @xmath39_+(x ) , f[z_k]_-(x)$ ] become infinite at @xmath40 .
but it is exactly there where the @xmath41 summand in the lorentz force ( [ eqn : cea ml+si lornetz force ] ) needs to be evaluated .
this divergence persists also in quantum field theories where it is referred to as _
uv divergence_. the simplest modification which avoids singular fields is @xmath0 , suggested by abraham and lorentz .
it is ml but with the point - charges replaced by extended charges ; cf . ( [ eqn : maxwell equations]),([eqn : lorentz force ] ) . however
, a rigid extension of the charge is for physical reasons unwanted @xcite , and furthermore ( even if correctly lorentz - boosted ) in violation with relativity @xcite .
the most familiar attempt to achieve a relativistic point - charge electromagnetism without singularities is the mass renormalization program of dirac @xcite .
in essence it is a point - charge limit procedure of @xmath0 .
dirac replaces the lorentz equations ( in a more or less ad hoc manner ) by @xmath42_{-}\right)+\frac{1}{2}\left(f[z_i]_{-}-f[z_i]_{+}\right)\right]^{\mu\nu}(z_i(\tau))\;\dot z_{i,\nu}(\tau).\end{aligned}\ ] ] these equations ( @xmath43 ) are called the lorentz - dirac equations ( ld ) .
note that the mass appearing on the left - hand side is the so called experimental mass @xmath44 ( see below ) . according to ld
the @xmath21-th particle feels a lorentz force due to an autonomous free field @xmath29 and due to the retarded fields @xmath39_{-}$ ] of all other charges .
furthermore , it feels the force term @xmath45_{-}-f[z_i]_{+}\right)$ ] which was interpreted by dirac as the radiation field produced by the charge itself and is to be held responsible for radiation damping .
he computed its principal value : @xmath46^{\mu\nu}_{-}-f[z_i]^{\mu\nu}_{+}\right)(z_i(\tau))=\frac{2}{3}e_i^2\left(\dddot z_i^\mu(\tau)\dot z_i^\nu(\tau)- \dddot z_i^\nu(\tau)\dot z_i^\mu(\tau)\right).\end{aligned}\ ] ] the third derivative is supposed to describe friction , hence the name _ radiation damping _ equation .
dirac s limit procedure can be reinterpreted as a renormalization procedure in which the so called bare masses @xmath47 approach @xmath48 , thereby subtracting the singular behavior of the fields @xmath49 when the charge extension goes to zero as well as yielding the observed experimental mass @xmath50 . in this respect
it may be worth noting that in @xcite it was observed that a negative bare mass @xmath47 causes the dynamics of @xmath0 to become unstable .
in any case it is well known that ld has unphysical solutions . already for @xmath51 and zero homogeneous fields all solutions except @xmath52 show run - away behavior , i.e. they approach the speed of light exponentially fast .
for a detailed analysis of the ld equation see @xcite .
mld is the theory defined by the maxwell equations ( [ eqn : cea m eq for ji ] ) coupled to ( [ eqn : cea effective lorentz ] ) via ( [ current ] ) .
the main aim of our research , of which we present results in this and the follow up paper @xcite , is in fact the description of electromagnetic phenomena without self - field divergences .
that is why we focus on another formulation of electrodynamics without self - fields which is suggested by the wheeler - feynman electromagnetism @xcite and discussed in @xcite .
the basic idea is that fields are only mediators of interaction _ between _ charges .
we consider the maxwell equations ( [ eqn : cea m eq for ji ] ) but we replace the lorentz force law by @xmath53 note that in contrast to ( [ eqn : cea ml+si lornetz force ] ) the self - field summand @xmath54 is excluded . we refer to this theory as ml - si ( maxwell - lorentz without self interaction ) . to connect this theory with mld we appeal to the observation done by wheeler and feynman , namely that to any solution of ml - si satisfying the extra constraint @xmath55_{-}-f[z_i]_{+}\right)(x)= 0 , \quad \forall x\in{{\mathbb{m } } } , \end{aligned}\ ] ] belong world lines of the charges which satisfy the ld equation .
one sees this by trivial manipulations of terms .
based on dirac s interpretation of the term @xmath56_{-}-f[z_i]_{+})$ ] as the field radiated by the @xmath21-th charge , ( [ eqn : absorber assumption ] ) states that the net radiation field is completely absorbed from which the name _ complete absorption condition _ ( ac ) is derived .
wheeler and feynman think of this condition as being satisfied for a thermal equilibrium distribution of a large number of charges and discuss it thoroughly in @xcite . stretching notations somewhat
we may rephrase the above in a formal way by @xmath57 where we understand the symbols now as sets of solutions : @xmath58 ( resp .
@xmath59 ) is the set of solutions of ml - si ( resp .
mld ) which fulfill ac . an important feature of ml - si is that it is very close to wf , the wheeler - feynman electromagnetism : wf contains no fields at all , only charges and is defined by @xmath60_++f[z_k]_-\right)^{\mu\nu}(z_i(\tau))\;\dot z_{i,\nu}(\tau)\end{aligned}\ ] ] where @xmath39_+$ ] and @xmath39_-$ ] are given by ( [ eqn : lw fields ] ) . due to the implicit definition of @xmath61 in ( [ eqn : lw fields ] ) these equations
involve advanced and retarded arguments and they belong mathematically to the class of neutral differential equations with unbounded delay .
the connection between wf and ml - si becomes manifest when we consider the case for which the homogeneous fields vanish : @xmath62 .
let us refer to this restricted theory as ml - si@xmath63 .
in view of ( [ eqn : forcedeco ] ) and ( [ eqn : cea ml - si lornetz force ] ) the world lines appearing in the solutions of ml - si@xmath63 are wf world lines , i.e. they fulfill ( [ eqn : wf eqn ] ) . in short : @xmath64it is important that the reader appreciates the difference between ml - si and wf .
there is no solution theory of wf , since the equations contain time - like advanced and retarded arguments .
the problem of existence of dynamics of wf is in fact famously difficult , since it is unclear how to even start a theory of solutions . on the other hand ,
ml - si is mathematically an initial value problem and at least the notion of local existence and uniqueness of solutions is clear .
now it seems that with ( [ eqn : crucial ] ) all is clear , because wf is simply ml - si with the homogeneous fields being zero .
but there is a catch
: one has no idea for which _ initial _ fields it is the case that the homogeneous field fulfills @xmath65 or equivalently @xmath66_++f[z_k]_-\right)$ ] . in other words
we do not know the initial conditions which define ml - si/(@xmath67 ) .
nevertheless ( [ eqn : crucial ] ) allows us to get a handle on the question of existence of solutions of wf which we present in the follow up paper @xcite . however , to be able to apply ( [ eqn : crucial ] ) to the wf problem we must be sure that the linard - wiechert fields ( [ eqn : lw fields ] ) generated by wf world lines are within the class of fields of ml - si/(@xmath67 )
. now some solutions of wf are known , namely the so called schild solutions @xcite which describe charges rotating around each other on stable orbits . such world lines with non - vanishing acceleration for large times generate linard - wiechert fields ( [ eqn : lw fields ] ) that are in general not square integrable .
we must therefore prove a general existence of dynamics result for ml - si where we allow fields which are not square integrable .
now that we have explained the role of the theory ml - si , which under the condition ac ( cf .
( [ eqn : absorber assumption ] ) ) describes the observed radiation phenomena , we must step back .
ml - si avoids the singular self - fields which make the dynamics ill - defined from the start .
but that does not mean that ml - si allows the existence of _ global _ solutions for _ all _ initial conditions .
in fact , to establish existence of global solutions some notion of _ typical _ initial conditions must be invoked , since ml - si ( for opposite charges ) is very analogous to masses interacting via gravitation , hence scenarios like explosions may be possible ; see for example @xcite . however , such considerations are at this early stage of research not in our focus and , for simplicity , we consider @xmath1 , and @xmath2 .
i.e. the theories with extended charges where singularities do not even occur when charges pass through each other .
we establish here the global existence and uniqueness of @xmath1 and by the same token that of @xmath0 .
the charge density @xmath3 we consider is rigid .
global existence and uniqueness of solutions of @xmath0 for one particle and square integrable initial fields , has been settled by two different techniques : while in @xcite one exploits the energy conservation to gain an priori bound needed for global existence , a grnwall argument was used in @xcite .
recent results are on the long - time behavior of solutions in @xcite and @xcite and on conservation laws in @xcite .
furthermore , a generalization to a spinning , extended charge was treated in @xcite .
for the mathematical analysis it is convenient to express @xmath1 and @xmath0 in non - relativistic notation using coordinates . the electric and magnetic field of each charge are defined by @xmath68 , @xmath69 , respectively .
the defining equations are @xmath70 together with @xmath71 \end{split}\end{aligned}\ ] ] for @xmath17
. the equations in the right - hand column of ( [ eqn : maxwell equations ] ) are also called the _
maxwell constraints_. we denote the partial derivative with respect to time @xmath72 by @xmath73 , the divergence by @xmath74 and the curl by @xmath75 .
vectors in @xmath76 are written as bold letters , e.g. @xmath77 . at time
@xmath72 the @xmath21-th charge is situated at position @xmath78 in euclidean space @xmath76 and has momentum @xmath79 .
it carries the classical mass @xmath80 .
the geometry of the @xmath21-th rigid charge is given in terms of a charge distribution ( or form factor ) @xmath81 which is assumed to be an infinitely often differentiable function with compact support , denoted by @xmath82 .
the factors @xmath83 denote the sign of the masses ( negative masses are useful to analyze dynamical instability when taking the point - particle limit to mld ) .
each charge is associated with an own electric and magnetic field @xmath84 and @xmath85 , which are @xmath76 valued functions on @xmath76 .
whereas in the classical literature one usually considers only one electric and magnetic field , we have given every charge its own field to allow exclusion of self - fields : the matrix coefficients @xmath86 for @xmath87 allow to switch on or off the coupling of the @xmath7-th field to the @xmath21-th particle .
this yields @xmath88 and @xmath89 for @xmath90 the corresponding system of equations formally define ml and ml - si , respectively .
the existence and uniqueness theory build in the following will neither depend on a particular choice of the coupling matrix @xmath91 , nor on the masses @xmath47 , nor on a particular choice of the charge distributions @xmath82 . for notational simplicity we shall now denote - in slight abuse of notation - the theory for any choices of the coupling matrix @xmath91 simply by @xmath0 .
we intend to arrive at a well - posed initial value problem for given positions and momenta @xmath92 as well as electric and magnetic fields @xmath93 at time @xmath94 for which we define the function space for the fields . as we remarked in the introduction
we wish to incorporate linard - wiechert fields produced by any time - like charge trajectory with uniformly bounded acceleration and momentum as cauchy data ( e.g. consider the bounded orbits of the schild solutions @xcite ) .
we shall show in the follow up paper @xcite that such fields decay as @xmath95 for @xmath96 .
hence , in general these fields are not in @xmath97 and that is why we establish the initial value problem for a bigger class of fields : [ def : weighted spaces ] we define the class of weight functions @xmath98 for any @xmath99 and open @xmath100 we define the space of weighted square integrable functions @xmath101 by @xmath102 for regularity arguments we need more conditions on the weight functions . for @xmath103
we define @xmath104 and @xmath105 the choice of @xmath106 is quite natural ( compare @xcite ) as for most estimates it allows to treat the new measure @xmath107 almost as if it were translational invariant .
clearly , @xmath108 is in @xmath106 . applying its definition twice we obtain for all @xmath99 the estimate @xmath109 which states that @xmath110 .
in particular , the weight @xmath111 is in @xmath106 because @xmath112 and therefore the desired linard - wiechert fields are in @xmath113 for @xmath114 . in the follow up paper @xcite it is shown that @xmath115 . with this we can define the space of initial values : [ def : phasespace ] we define @xmath116 any element @xmath117 consists of the components @xmath118 , i.e. positions @xmath119 , momenta @xmath120 and electric and magnetic fields @xmath121 and @xmath122 for each of the @xmath17 charges .
if not noted otherwise , any spatial derivative will be understood in the distribution sense , and the latin indices @xmath123 run over the charge labels @xmath124 .
we also need the weighted sobolev spaces @xmath125 for any @xmath103 .
we will rewrite @xmath0 using the following operators @xmath126 and @xmath127 : [ def : operator_a ] for a @xmath118 we defined @xmath128 and @xmath126 by the expression @xmath129 on their natural domain @xmath130 furthermore , for any @xmath131 we define @xmath132 [ def : operator_j ] together with @xmath133 we define @xmath134 by @xmath135 for @xmath118 .
note that @xmath127 is well - defined because @xmath82 . with these definitions , the lorentz force law ( [ eqn : lorentz force ] ) , the maxwell equations ( [ eqn : maxwell equations ] ) , while temporarily neglecting the maxwell constraints , take the form @xmath136 in the following we frequently use the notation @xmath137 to denote that @xmath138 is a continuous mapping depending non - decreasingly on all of its arguments .
the two main theorems are : [ thm : globalexistenceanduniqueness ] for @xmath139 , @xmath131 and @xmath140 the following holds : a. _ ( global existence ) _ there exists an @xmath141-times continuously differentiable mapping @xmath142 which solves ( [ eqn : dynamic_maxwell ] ) for initial value @xmath143 .
furthermore , it holds @xmath144 for all @xmath145 and @xmath146 , b. _ ( uniqueness and growth ) _ if any once continuously differentiable function @xmath147 for some open interval @xmath148 is also a solution to ( [ eqn : dynamic_maxwell ] ) with @xmath149 for some @xmath150 , then @xmath151 holds for all @xmath152 .
in particular , given @xmath153 , @xmath17 , there exists @xmath154 such that for all @xmath155 such that @xmath156 @xmath157}\|\varphi_t-\widetilde\varphi_t\|_{{{\mathcal{h}}}_w}\leq \constr{apriori lipschitz}(t,\|\varphi_{t_0}\|_{{{\mathcal{h}}}_w},\|\widetilde\varphi_{t_0}\|_{{{\mathcal{h}}}_w})\|\varphi_{t_0}-\widetilde\varphi_{t_0}\|_{{{\mathcal{h}}}_w } \end{aligned}\ ] ] holds .
furthermore , there is a @xmath158 such that for all @xmath153 , @xmath17 , and @xmath155 with @xmath156 one has @xmath159}\|\varphi_t\|_{{{\mathcal{h}}}_w } \leq \constr{apriori ml rho}\left(t,\|w^{-1/2}\varrho_i\|_{l^2},\|\varrho_i\|_{l^2_w } ; 1\leq i\leq n\right)\ ; \|\varphi^0\|_{{{\mathcal{h}}}_w}. \end{aligned}\ ] ] c. _ ( constraints ) _ if the solution @xmath160 obeys the maxwell constraints @xmath161 for one time instant @xmath145 , then they are obeyed for all times @xmath145 .
[ thm : regularity ] let @xmath162 , @xmath163 for @xmath164 and let @xmath165 be the solution to ( [ eqn : dynamic_maxwell ] ) for initial value @xmath166 .
then for all @xmath17 : a. it holds for any @xmath145 that @xmath167 . b. the electromagnetic fields regarded as mappings @xmath168 and @xmath169 are in @xmath170 and both have a representative in @xmath171 within their equivalence class .
c. for @xmath172 , @xmath173 , and every @xmath145 we have also @xmath174 and @xmath175 such that : @xmath176 the proofs will be given in sections [ sec : glob exist unique ] and [ sec : reg ] where we rely on tools for the study of @xmath113 and the associated weighted sobolev spaces discussed in section [ sec : l2w ] as well as on the abstract global existence and uniqueness theorem discussed in the appendix [ sec : abstractglobalexistandunique ] .
in the formulas we use `` @xmath177 '' to denote the right - hand side of the preceding equation or inequality .
theorem [ thm : globalexistenceanduniqueness ] permits us to define a time evolution operator : [ def : ml time evolution ] we define the non - linear operator @xmath178=\varphi_t = w_{t - t_0}\varphi^0 + \int_{t_0}^t w_{t - s } j(\varphi_s ) \end{aligned}\ ] ] which encodes @xmath0 time evolution from time @xmath179 to time @xmath72 .
we now collect all needed properties of the introduced weighted function spaces .
the following assertions , except theorem [ lem : sobolev ] , are independent of the space dimension .
that is why we often use the abbreviation @xmath180 and @xmath181 .
with @xmath99 the @xmath182 analogues of almost all results of the @xmath183 theory which do not involve the fourier transform can be proven with only minor modifications . for open @xmath100 , @xmath184 is clearly a linear space and has an inner product : @xmath185 as a standard result since @xmath186 for @xmath100 is a @xmath187-finite measure space : [ thm : l2w hilbert space cinftyc dense ] for any @xmath99 , open @xmath188 , @xmath184 with inner product ( [ eqn : sclarproduct ] ) is a separable hilbert space and @xmath189 lies dense .
we shall sometimes switch between the @xmath182 and the @xmath183 notation , i.e. in the case of the schwarz inequality for all @xmath190 we write @xmath191 .
we shall also need : for all @xmath99 , open @xmath100 and @xmath192 we define @xmath193 which are equipped with the inner products @xmath194 @xmath195 respectively .
we use the multi - index notation @xmath196 , @xmath197 , @xmath198 where @xmath199 denotes the derivative w.r.t . to the @xmath21-th standard unit vector in @xmath76 . in the following we use a superscript @xmath200 , e.g. @xmath201 , as a placeholder for @xmath33 , @xmath202 , and @xmath203 for any @xmath103 , respectively .
we shall also need the local versions , i.e. @xmath204 , where @xmath205 is short for compactly contained .
usually we abbreviate @xmath206 , use @xmath207 and drop the subscript @xmath208 if @xmath108 . by definition of the weak derivative , theorem [ thm : l2w hilbert space cinftyc dense ] implies also : [ thm : hnw_complete][thm : hcurlw_complete ] for any @xmath99
, @xmath201 is a separable hilbert space .
[ lem : local equivalence ] let @xmath99 . for an open @xmath209 one has @xmath210 and the respective norms are equivalent .
hence , a function @xmath211 is in @xmath212 for every open @xmath213 if and only if @xmath214 .
given @xmath99 , equation ( [ eqn : weight_relation ] ) ensures that there are two finite and non - zero constants @xmath215 , @xmath216 .
thus , we get @xmath217 .
a direct computations using definition ( [ eqn : weight function spaces ] ) gives : [ lem : dw bound ] let @xmath218 , then for every multi - index @xmath219 there also exists constants @xmath220 such that on @xmath76 @xmath221 [ thm : cinftyc dense in hcurl htriange dense ] in the case @xmath222 and @xmath223 let @xmath99
. in the case @xmath224 let @xmath162 .
then @xmath225 is a dense subset of @xmath226 .
the proof is similar in all cases .
only the case @xmath227 is a bit more involved as one e.g. needs to estimate the derivatives @xmath228 , @xmath229 , in terms of the laplacian .
let @xmath230 , then we need to show that for every @xmath231 there is a @xmath232 such that @xmath233 .
take a @xmath234)$ ] such that @xmath235 for @xmath236 . for @xmath237
the sequence @xmath238 converges to @xmath211 in @xmath182 by definition .
furthermore , we have @xmath239 again by definition of @xmath240 the first term on the right - hand side of ( [ eqn : triangle approx by bounded func ] ) goes to zero for @xmath241 . with @xmath242
we have @xmath243 so that also the second term goes to zero for @xmath241 .
furthermore , on open sets @xmath244 we have @xmath245 by lemma [ lem : local equivalence ] and therefore @xmath246 .
thus , we can apply partial integration and , using the abbreviation @xmath247 , yield @xmath248({{\mathbf{x}}})\right|.\ ] ] in the first terms on the right - hand side we apply the chain rule @xmath249 and integrate by parts again so that @xmath250({{\mathbf{x}}})\right| \leq \frac{1}{2 } \left\|\sqrt{|\triangle\omega_n|}\;{{\mathbf{f}}}\right\|^2_{l^2 } + \left|{\left\langle \sqrt{|\omega_n|}\;\triangle{{\mathbf{f}}},\sqrt{|\omega_n|}\;{{\mathbf{f } } } \right\rangle}_{l^2}\right|\ ] ] by lemma [ lem : dw bound ] we have @xmath251 on @xmath76 .
define @xmath252 , then @xmath253 and @xmath254 uniformly in @xmath141 on @xmath76 . finally , using schwarz s inequality we get @xmath255 going back to equation ( [ eqn : triangle approx by bounded func ] )
we then find that also the last term on the right - hand side goes to zero as @xmath241 .
it is straight - forward to apply this idea to also show @xmath256 as @xmath241 for @xmath257 ( note that also for @xmath257 the condition @xmath162 is sufficient ) .
hence , we conclude that there is an @xmath258 with compact support and @xmath259
. now let @xmath260 and define @xmath261 .
it is a standard analysis argument that @xmath262 for @xmath241 .
thus , due to lemma [ lem : local equivalence ] for @xmath141 large enough @xmath263 because , as @xmath264 and @xmath265 have compact support , @xmath266 . with that @xmath267 .
hence , @xmath268 is dense in @xmath227 . by the standard approximation argument this theorem allows us to make use of partial integration in the spaces @xmath226 .
[ thm : htrianglekw equals h2kw ] let @xmath162 , then for any @xmath103 we have @xmath269 and the respective norms are equivalent . first we prove @xmath270 and that their respective norms are equivalent
. by definition of the weak derivative and theorem [ thm : cinftyc dense in hcurl htriange dense ] it is sufficient to show that the norms @xmath271 and @xmath272 are equivalent on @xmath268 . by definition for @xmath232
we have @xmath273 using partial integration in a similar way as in the proof of theorem [ thm : cinftyc dense in hcurl htriange dense ] one yields the bounds @xmath274 thus , the claim is proven for @xmath275 .
clearly , @xmath276 implies @xmath277 . on the other hand ,
let us assume that for some @xmath103 and all @xmath230 also @xmath278 holds .
let @xmath279 then using the induction hypothesis @xmath280 now @xmath281 is in @xmath282 for @xmath283 which by @xmath270 and the equivalence of their respective norms implies @xmath284 .
we conclude @xmath269 and their respective norms are equivalent .
[ lem : sqrtw hkw is hk ] for @xmath222 , @xmath224 , and @xmath223 , set @xmath285 , @xmath286 and @xmath287 , respectively .
let @xmath288 , then @xmath289 and the respective norms are equivalent . due to theorem [ thm : htrianglekw equals h2kw ] we only need to show the cases @xmath223 and @xmath222 .
we only show the latter , the former can be derived similarly .
let @xmath290 then @xmath291 but as @xmath218 , there is some finite constant @xmath292 such that @xmath293 and , hence , @xmath294 this implies @xmath295 , i.e. @xmath296 . a similar computation using the estimate from lemma [ lem : dw bound ] yields @xmath297 , i.e. @xmath298 . [ lem : sobolev ] let @xmath173 , then : a. @xmath299 for an open @xmath209 and @xmath99 implies that there is a @xmath300 , @xmath301 , such that @xmath302 almost everywhere on @xmath303 .
b. @xmath299 for all @xmath209 and @xmath99 implies that there is a @xmath304 , @xmath301 , such that almost everywhere @xmath302 on @xmath76 .
c. @xmath305 and @xmath218 implies that there is a possibly @xmath33 dependent @xmath175 such that @xmath306 \(i ) for any open set @xmath209 , @xmath299 implies @xmath307 due to lemma [ lem : local equivalence ] .
sobolev s lemma ( * ? ? ?
* ix.24 ) states that there is then a @xmath300 for @xmath308 with @xmath302 almost everywhere on @xmath303 .
claim ( ii ) follows by ( i ) and continuity .
( iii ) for @xmath218 we know from lemma [ lem : sqrtw hkw is hk ] that @xmath309 .
applying sobolev s lemma as in ( i ) we yield the same result for @xmath310 which provides the conditions for morrey s inequality ( [ eqn : morrey ] ) to hold , see ( * ? ? ?
* chapter 8 , theorem 8.8(iii ) , p.213 ) .
note that theorem [ lem : sobolev ] is the only result that depends on the dimension of @xmath76 . assertion ( i ) and ( ii ) : we intend to use the abstract existence and uniqueness statement of theorem [ lem : aj_local_exist_uniqueness ] for @xmath311 . in order to do
so we need to show that the operators @xmath126 and @xmath127 from definitions [ def : operator_a ] and [ def : operator_j ] have the properties as given in definitions [ def : operatora ] and [ def : operatorj ] , respectively , and that the needed a priori bound for theorem [ lem : aj_local_exist_uniqueness](iii ) holds .
the needed properties of @xmath126 and @xmath127 are shown in lemma [ lem : operatora ] and lemma [ lem : operatorj ] .
the a priori bound is provided by lemma [ lem : a priori ] .
[ lem : operatora ] the operator @xmath126 introduced in definition [ def : operator_a ] on @xmath312 with weight @xmath139 fulfills all properties of definition [ def : operatora ] with @xmath311 , i.e. a. @xmath126 is closed and densely defined .
b. there exists a @xmath313 such that @xmath314 , the resolvent set of a. c. the resolvent @xmath315 of @xmath126 with respect to @xmath316 is bounded by @xmath317 , i.e. for all @xmath318 we have @xmath319 . in definition
[ def : operator_a ] the operator @xmath126 was given in terms of the operator @xmath128 on @xmath320 .
we abbreviate the hilbert space direct sum @xmath321 and express vectors @xmath322 componentwise as @xmath323 .
first , we prove that @xmath128 is closed and densely defined : according to theorem [ thm : hcurlw_complete ] , @xmath324 is a hilbert space so that @xmath325 is a banach space with respect to the norm @xmath326 .
this means any sequence @xmath327 in @xmath325 such that @xmath327 and @xmath328 converges in @xmath329 to @xmath330 and @xmath331 , respectively , converges also with respect to @xmath332 .
this implies @xmath333 and @xmath334 , i.e. @xmath128 is closed .
theorem [ thm : cinftyc dense in hcurl htriange dense ] ensures that @xmath268 is dense in @xmath324 .
hence , @xmath335 lies dense in @xmath329 .
thus , the operator @xmath128 is densely defined .
next we prove that there exists a @xmath336 such that @xmath337 which means that for all @xmath338 @xmath339 is a bijection : let @xmath340 denote the schwartz space of infinitely often differentiable @xmath76 valued functions on @xmath76 with faster than polynomial decay , and let @xmath341 denote the dual of @xmath340 .
on @xmath342 we regard in matrix notation @xmath343 for @xmath344 and @xmath345 . with the use of the fourier transformation @xmath346 and its inverse @xmath347 on @xmath341 we get @xmath348 = \begin{pmatrix } \lambda t_1[{{\mathbf{u } } } ] - \nabla\wedge t_2[{{\mathbf{u}}}]\\ \lambda t_2[{{\mathbf{u } } } ] + \nabla\wedge t_1[{{\mathbf{u } } } ] \end{pmatrix } = \begin{pmatrix } \widetilde{t_1}[\lambda\widehat{{{\mathbf{u } } } } ] - \widetilde{t_2}[{{\mathbf{k}}}\mapsto i{{\mathbf{k}}}\wedge\widehat{{{\mathbf{u}}}}({{\mathbf{k}}})]\\ \widetilde{t_2}[\lambda\widehat{{{\mathbf{u } } } } ] + \widetilde{t_1}[{{\mathbf{k}}}\mapsto i{{\mathbf{k}}}\wedge\widehat{{{\mathbf{u}}}}({{\mathbf{k } } } ) ] \end{pmatrix } = 0\ ] ] for all @xmath349 . here , e.g. @xmath350 $ ] denotes the evaluation of the distribution @xmath351 on test function @xmath352 . by plugging the second equation into the first for @xmath353 ,
one finds @xmath354=:r_1[\widehat{{{\mathbf{u}}}}]\ ] ] for all @xmath355 . however , for each @xmath356 we find a @xmath355 according to @xmath357 such that @xmath358=\widetilde{t_1}[\widehat{{{\mathbf{v}}}}]=r_1[\widehat{{{\mathbf{u}}}}]=0 $ ] , which means that @xmath359 , and hence , also @xmath360 on @xmath341 .
we have thus shown that @xmath361 since @xmath362 .
therefore , mapping ( [ eqn : resolvent map ] ) is injective for @xmath353 .
we shall now see that there exists a @xmath363 such that for all @xmath364 this mapping is also surjective , i.e. @xmath365 .
let @xmath366 . since @xmath268 is dense in @xmath324 , cf .
theorem [ thm : cinftyc dense in hcurl htriange dense ] , we may use partial integration from which we obtain @xmath367 for @xmath368 . on the other hand , we have shown that @xmath361 for all @xmath353 , hence @xmath369 must be zero which implies that @xmath370 since @xmath139 .
thus , @xmath371 is dense , so that @xmath372 .
as @xmath373 is bijective , we can define @xmath374 to be its inverse .
next , we show the boundedness of @xmath374 which implies the closedness of @xmath371 .
let @xmath375 , then there is a unique @xmath368 which solves @xmath376 .
the inner product with @xmath330 gives @xmath377 and with the schwarz inequality and the symmetry of the inner product it implies @xmath378 in the notation @xmath379 a partial integration yields @xmath380 in the last step we used schwarz s inequality and the constant @xmath381 coming from the bound on @xmath208 given by lemma [ lem : dw bound ] .
let us define @xmath382 .
hence , for @xmath364 we obtain the estimate @xmath383 from ( [ eqn : rlambda estimate ] ) . as @xmath371 is dense
, there is a unique extension of @xmath374 that we denote by the same symbol @xmath384 which obeys the same bound ( [ eqn : resolventestimate ] ) on whole @xmath329 .
next , we show that @xmath371 is closed .
let @xmath385 be a sequence in @xmath371 which converges in @xmath329 for @xmath364 .
define @xmath386 for all @xmath131 . by ( [ eqn : resolventestimate ] )
we immediately infer convergence of the sequence @xmath327 to some @xmath330 in @xmath329 .
thus , @xmath387 converge to @xmath388 in @xmath329 , and because @xmath128 is closed , @xmath368 and @xmath376 .
hence , @xmath375 and @xmath371 is closed . since we have shown that @xmath371 is closed , we have also @xmath365 .
hence , for all @xmath364 the mapping ( [ eqn : resolvent map ] ) is a bijection . finally , we show that @xmath126 inherits these properties from @xmath128 : let @xmath389 . by definition @xmath390
since @xmath128 is closed on @xmath391 , @xmath126 is closed on @xmath392 , and @xmath393 lies dense in @xmath394 .
this implies property ( i ) of definition [ def : operatora ] .
furthermore , as for @xmath395 @xmath396 as @xmath353 , @xmath397 is a bijection and for @xmath398 its inverse @xmath399 is given by @xmath400 therefore , @xmath401 is a subset of the resolvent set @xmath402 of @xmath126 .
this implies property ( ii ) of definition [ def : operatora ] .
finally , by ( [ eqn : resolventestimate ] ) we have the estimate @xmath403 which implies property ( iii ) of definition [ def : operatora ] and concludes the proof .
this lemma together with lemma [ lem : contraction_group ] states that @xmath126 on @xmath312 generates a @xmath404-contractive group @xmath405 which gives rise to the next definition : [ def : wt ] we denote by @xmath405 the @xmath404-contractive group on @xmath394 generated by @xmath126 on @xmath312 .
@xmath405 comes with a standard bound @xmath406 for all @xmath117 , see lemma [ lem : contraction_group ] . for the case
that @xmath208 is a constant , one finds @xmath407 and the whole proof above collapses into an argument about skew - adjointness on @xmath183 . in this case , @xmath405 is simply the unitary group generated by the skew - adjoint operator @xmath126 . for non - constant @xmath208 , @xmath405 does not preserve the norm .
for example , consider a weight @xmath208 that decreases with the distance to the origin .
then , any wave packet moving towards the origin while retaining its shape ( like e.g. solutions to the free maxwell equations ) has necessarily an @xmath182 norm of its fields that increases in time .
[ lem : operatorj ] the operator @xmath127 introduced in definition [ def : operator_j ] with a weight @xmath99 fulfills all properties of definition [ def : operatorj ] with @xmath311 , i.e. for an @xmath408 , @xmath127 mappings @xmath409 to @xmath410 and has the properties : a. for all @xmath411 there exist @xmath412 such that for all @xmath413 @xmath414 b. for all @xmath411 and @xmath415 , @xmath416 and any @xmath417 such that @xmath418 for @xmath419 , the operator @xmath127 fulfills for @xmath420 : a. @xmath421 and b. @xmath422 is continuous on @xmath423 .
furthermore , a. there exists a @xmath424 such that @xmath425 for any @xmath426 where @xmath427 and @xmath428 only depend on @xmath208 , cf .
( [ eqn : weightclass ] ) . as remarked below definition [ def : operator_j ] ,
@xmath127 is a well - defined mapping from @xmath394 to @xmath429 . assertion ( i ) : choose @xmath118 and @xmath430 in @xmath394 .
according to definition [ def : operator_j ] , for any @xmath131 we have @xmath431 and @xmath432 since @xmath433 , inequality ( [ eqn : an jdiff ] ) for @xmath434 gives @xmath435 .
therefore , it suffices to prove ( [ eqn : an jdiff ] ) .
the only involved case therein is @xmath436 as one needs to control the lorentz force on each rigid charge , which for @xmath437 is mapped to zero by any power of @xmath126 .
for @xmath436 we obtain : @xmath438 the following notation is convenient : for any function @xmath439 and @xmath440 we denote by @xmath441 the jacobi matrix of @xmath442 with entries @xmath443 for @xmath444 . furthermore ,
for any vector space @xmath445 with norm @xmath446 and any operator @xmath447 on @xmath445 we write @xmath448 .
recall also the coefficients @xmath19 , @xmath449 and @xmath86 for @xmath87 from definition [ def : operator_j ] and define @xmath450 . without loss of generality
, we may assume that @xmath451 , @xmath17 , with the possible signs being absorbed in @xmath91 . by the mean value theorem , for each index
@xmath21 there exists a @xmath452 $ ] such that for @xmath453 we obtain @xmath454 now with @xmath455 we have @xmath456 .
thus , it follows @xmath457 for @xmath458 so that @xmath459 next we must get a bound on the lorentz force . for @xmath460 and @xmath461
we define @xmath462 .
choose @xmath461 such that for @xmath17 it holds @xmath463 .
define @xmath464 , then @xmath465 let @xmath466 for each @xmath17 and @xmath467 $ ] , then @xmath468 now @xmath469 for @xmath470 so that @xmath471.\ ] ] in the following we denote the characteristic function of a set @xmath472 by @xmath473
. the following type of estimates will be used frequently : for @xmath474 it holds @xmath475^{\frac{1}{2 } } \left\|\frac{{\mathbbm{1}_{b_r(0)}}}{\sqrt w}\right\|_{l^2}\|{{\mathbf{f}}}\|_{l^2_w},\\ \int d^3x\ ; \varrho_i({{\mathbf{x}}}-{{\mathbf{q}}}_i)\|{{\mathbf{f}}}({{\mathbf{x}}})\|_{{{\mathbb{r}}}^3 } & \leq ( 1+\namer{cw}\|{{\mathbf{q}}}_i\|)^{\frac{\namer{pw}}{2 } } \left\|\frac{\varrho_i}{\sqrt w}\right\|_{l^2}\|{{\mathbf{f}}}\|_{l^2_w({{\mathbb{r}}}^3)}.\label{eqn : j_est2 } \end{aligned}\ ] ] the former inequality can be seen by schwarz s inequality : @xmath476 using the weight estimate ( [ eqn : weight_relation ] ) yields ( [ eqn : j_est1 ] ) . similarly the latter inequality can be seen by @xmath477 and again using the weight estimate ( [ eqn : weight_relation ] ) .
we abbreviate @xmath478^{\frac{1}{2 } } \left\|\frac{{\mathbbm{1}_{b_r(0)}}}{\sqrt w}\right\|_{l^2 } , & & g(x):=(1+\namer{cw}x)^{\frac{\namer{pw}}{2 } } \sum_{i=1}^n\left\|\frac{\varrho_i}{\sqrt w}\right\|_{l^2 } \end{aligned}\ ] ] so that ( [ eqn : j_est1 ] ) and ( [ eqn : j_est2 ] ) give @xmath479 we apply these estimates to the term @xmath480 and obtain @xmath481 for @xmath482 in the same way we estimate @xmath483 first we apply the mean value theorem to the velocities as we did before such that @xmath484 then again we rewrite the densities by the fundamental theorem of calculus and use @xmath485 in order to obtain @xmath486 finally we apply the two estimates ( [ eqn : j est1 short ] ) and ( [ eqn : j est2 short ] ) to arrive at @xmath487 thus , we obtain the estimate @xmath488 for @xmath489 it remains to estimate term .
we shall do this already for the general case of any fixed @xmath490 .
recall from equation ( [ eqn : anjformal ] ) that @xmath491 we begin with @xmath492 but before we continue we shall express these terms in a more convenient way . for all @xmath493 , @xmath494 and all @xmath495 ( using the notation @xmath496 )
the following identities hold : @xmath497,\\ ( \nabla\wedge)^{2m+1}({{\mathbf{v } } } h ) & = ( -1)^m[\nabla\triangle^m h]\wedge{{\mathbf{v}}}. \end{aligned}\ ] ] let us begin with term @xmath498 for odd @xmath141 . as before we write @xmath499 for each @xmath17 and @xmath467 $ ] so that for @xmath500^{\frac{1}{2 } } \end{aligned}\ ] ] we get @xmath501 \wedge { { \mathbf{v}}}({{\mathbf{p}}}_i)\big\|_{l^2_w}\nonumber\\ & \leq 4\pi \sum_{i=1}^n \big\|\int_0 ^ 1d\kappa\ ; d[\nabla\triangle^m\varrho_i({{\mathbf{x}}}-{{\mathbf{z}}}_i(\kappa))]\cdot({{\mathbf{q}}}_i-\widetilde{{{\mathbf{q}}}}_i)\big\|_{l^2_w}\leq 4\pi \namer{k rho }
\sum_{i=1}^n \|{\mathbbm{1}_{i_i}}\|_{l^2_w}\|{{\mathbf{q}}}_i-\widetilde{{{\mathbf{q}}}}_i\|_{{{\mathbb{r}}}^3}\nonumber\\ & \leq 4\pi \namer{k rho } { k_i\left(\|\varphi\|_{{{\mathcal{h}}}_w},\|\widetilde\varphi\|_{{{\mathcal{h}}}_w}\right)}\|\varphi-\widetilde\varphi\|_{{{\mathcal{h}}}_w}=:\constl{n:3.1}{(2m+1,\|\varphi\|_{{{\mathcal{h}}}_w},\|\widetilde\varphi\|_{{{\mathcal{h}}}_w})}\;\|\varphi-\widetilde\varphi\|_{{{\mathcal{h}}}_w}.\label{eqn : jest 2 for gronwall } \end{aligned}\ ] ] similarly the term @xmath498 for even @xmath141 gives @xmath502\right ) + \\ & \hskip 2 cm - { { \mathbf{v}}}(\widetilde{{{\mathbf{p}}}}_i)\triangle^m\left[\varrho_i(\cdot-{{\mathbf{q}}}_i)-\varrho_i(\cdot-\widetilde{{{\mathbf{q}}}}_i)\right]\big\|_{l^2_w}\\ & \leq4\pi \sum_{i=1}^n\bigg(\big\| { \mathbbm{1}_{i_i}}\int_0 ^ 1d\kappa\ ; d\left[d\left[\nabla\triangle^{m-1}\varrho_i({{\mathbf{x}}}-{{\mathbf{z}}}_i(\kappa))\right]\cdot({{\mathbf{q}}}_i-\widetilde{{{\mathbf{q}}}}_i)\right]\cdot{{\mathbf{v}}}({{\mathbf{p}}}_i)\big\|_{l^2_w } + \\ & \hskip 2 cm + \big\|{\mathbbm{1}_{i_i}}{{\mathbf{v}}}({{\mathbf{p}}}_i)\int_0 ^ 1d\kappa\ ; d\triangle^m\varrho_i({{\mathbf{x}}}-{{\mathbf{z}}}_i(\kappa))\cdot({{\mathbf{q}}}_i-\widetilde{{{\mathbf{q}}}}_i)\big\|_{l^2_w}\bigg ) . \end{aligned}\ ] ] again we estimate the coefficients of the jacobi matrices by @xmath503 obtaining a factor @xmath504 in the first summand such that @xmath505 the last term to be estimated for odd @xmath141 is : @xmath506 \big\|_{l^2_w}\nonumber\\ & \leq 4\pi\sum_{i=1}^n \left\|\nabla\triangle^m\varrho_i(\cdot-\widetilde{{{\mathbf{q}}}}_i)\right\|_{l^2_w } \namer{k vel } \|\varphi-\widetilde\varphi\|_{{{\mathcal{h}}}_w}\nonumber\\ & \leq 4\pi ( 1+\namer{cw}\|\widetilde \varphi\|_{{{\mathcal{h}}}_w})^{\frac{\namer{pw}}{2}}\sum_{i=1}^n\left\|\nabla\triangle^m\varrho_i\right\|_{l^2_w } \namer{k vel } \|\varphi-\widetilde\varphi\|_{{{\mathcal{h}}}_w}\nonumber\\ & = : \constl{n:3.2}{(2m+1,\|\varphi\|_{{{\mathcal{h}}}_w},\|\widetilde\varphi\|_{{{\mathcal{h}}}_w})}\;\|\varphi-\widetilde\varphi\|_{{{\mathcal{h}}}_w},\label{eqn : jest 4 for gronwall } \end{aligned}\ ] ] and for even @xmath141 : @xmath507\cdot\nabla\triangle^{m-1}\varrho_i(\cdot-\widetilde{{{\mathbf{q}}}}_i)\right ) - \left[{{\mathbf{v}}}({{\mathbf{p}}}_i)-{{\mathbf{v}}}(\widetilde{{{\mathbf{p}}}}_i)\right]\triangle^m\varrho_i(\cdot-\widetilde{{{\mathbf{q}}}}_i ) \big\|_{l^2_w}\nonumber\\ & \leq 4\pi \sum_{i=1}^n \bigg(\left({{\int d^3x\ ; } } w({{\mathbf{x}}})\|d\nabla\triangle^{m-1}\varrho_i({{\mathbf{x}}}-\widetilde{{{\mathbf{q}}}}_i)\|_{{{\mathbb{r}}}^3}\right)^{\frac{1}{2 } } + \|\triangle^m\varrho_i(\cdot-\widetilde{{{\mathbf{q}}}}_i)\|_{l^2_w}\bigg)\namer{k vel } \|\varphi-\widetilde\varphi\|_{{{\mathcal{h}}}_w}\nonumber\\ & \leq 4\pi n ( 1+\namer{cw}\|\widetilde \varphi\|_{{{\mathcal{h}}}_w})^{\frac{\namer{pw}}{2 } } \bigg(\left({{\int d^3x\ ; } } w({{\mathbf{x}}})\|d\nabla\triangle^{m-1}\varrho_i({{\mathbf{x}}})\|_{{{\mathbb{r}}}^3}\right)^{\frac{1}{2 } } + \|\triangle^m\varrho_i\|_{l^2_w}\bigg)\namer{k vel } \|\varphi-\widetilde\varphi\|_{{{\mathcal{h}}}_w}\nonumber\\ & = : \constr{n:3.2}{(2m,\|\varphi\|_{{{\mathcal{h}}}_w},\|\widetilde\varphi\|_{{{\mathcal{h}}}_w})}\|\varphi-\widetilde\varphi\|_{{{\mathcal{h}}}_w}.\label{eqn : jest 5 for gronwall } \end{aligned}\ ] ] collecting all estimates we finally arrive at the inequality ( [ eqn : an jdiff ] ) for @xmath508 which for fixed @xmath141 is a continuous and non - decreasing function in the arguments @xmath509 and @xmath510 , and hence , @xmath511 . assertion ( ii ) : for @xmath415 let @xmath512 be a mapping in @xmath513 such that for all @xmath419 and @xmath416 it holds that @xmath514 .
we have to show that for all @xmath420 , @xmath515 is continuous on @xmath423 and take values in @xmath516 . by formulas ( [ eqn : jformal ] ) and ( [ eqn : anjformal ] ) both properties are an immediate consequence of @xmath517 . in fact
, one finds that @xmath515 takes values in @xmath429 on @xmath423 . finally , we prove assertion ( iii ) , i.e. inequality ( [ eqn : j estimate for gronwall ] ) : in principle we could use ( [ eqn : j est main ] ) and the estimates ( [ eqn : vel est ] , [ eqn : jest 1 for gronwall ] , [ eqn : jest 2 for gronwall ] , [ eqn : jest 3 for gronwall ] , [ eqn : jest 4 for gronwall ] , and [ eqn : jest 5 for gronwall ] ) for @xmath434 so that we only had to treat the lorentz force .
however , this way we do not get an optimal dependence of the bounds on @xmath3 . therefore , we regard @xmath518=:\terml{j bound 1}+\terml{j bound 2}+\terml{j bound 3}. \end{aligned}\ ] ] the first term can be treated as before , cf .
( [ eqn : vel est ] ) , @xmath519 the second term @xmath520 can be bounded by @xmath521 using estimate ( [ eqn : j_est2 ] ) we find @xmath522 finally , for the last term we obtain @xmath523 hence , there is a @xmath524 for @xmath525 this concludes the proof .
note that both cases , @xmath0 ( [ eqn : ml+si ] ) as well as @xmath1 ( [ eqn : ml - si ] ) , are treated in estimate ( [ eqn : ml+-si ] ) because the summation goes also over @xmath526 .
this is only possible because of the smearing of @xmath153 . in the point - particle limit
this @xmath526 summand is the problematic term which blows up . on the contrary , in the ml - si case with an appropriate choice of norms on the field spaces and an a priori lower bound on the distance of the charges for all times ( in order to prevent singular situations like in classical gravitation @xcite ) , the point - particle limit bares no obstacles .
next , we prove the needed a priori bound ; the one assumed in ( [ eqn : apriori ] ) .
[ lem : a priori ] let @xmath527 be a solution to @xmath528 with @xmath529 . then there is a @xmath530 such that @xmath531}\|\varphi_t\|_{{{\mathcal{h}}}_w } \leq e^{\gamma t}(1+\constr{apriori ml } t e^{\constr{apriori ml } t } ) \|\varphi^0\|_{{{\mathcal{h}}}_w } \end{aligned}\ ] ] for @xmath532 . by lemma [ lem : operatora ]
we know that @xmath533 lemma [ lem : operatorj ] provides the bound to estimate the integrand @xmath534 for any @xmath535 .
moreover , as the velocities are bounded by the speed of light we get in addition @xmath536 hence , for some finite @xmath415 and @xmath537 we infer the following integral inequality @xmath538 for @xmath539 , according to which by grnwall s lemma @xmath540}\|\varphi_t\|_{{{\mathcal{h}}}_w } \leq e^{\gamma t}(1+\constr{apriori ml } t e^{\constr{apriori ml } t } ) \|\varphi^0\|_{{{\mathcal{h}}}_w } \end{aligned}\ ] ] holds with the parameter dependence of @xmath541 as claimed .
this concludes the proof .
now we have all ingredients to apply theorem [ lem : aj_local_exist_uniqueness ] .
lemma [ lem : operatora ] and lemma [ lem : operatorj ] prove the needed properties of the operators @xmath126 and @xmath127 .
lemma [ lem : a priori ] proves the needed a priori bound .
hence , theorem [ lem : aj_local_exist_uniqueness ] ensures global existence and uniqueness of solutions to ( [ eqn : dynamic_maxwell ] ) , to be precise , assertion ( i ) and ( ii ) except the bounds which we prove next : lemma [ lem : a priori ] proves claim ( [ eqn : apriori lipschitz no diff ] ) for @xmath542 while ( [ eqn : apriori lipschitz ] ) can be verified as follows .
let @xmath543 and @xmath544\to d_w(a)$ ] be solutions to ( [ eqn : dynamic_maxwell ] ) , then for @xmath545 $ ] we have @xmath546 by ( [ eqn : an jdiff ] ) .
now we use ( [ eqn : gronwall sup ] ) and find @xmath547}\constr{cj2}^{(1)}{(\|\varphi_s\|_{{{\mathcal{h}}}_w},\|\widetilde\varphi_s\|_{{{\mathcal{h}}}_w})}<\infty.\ ] ] hence , we can apply grnwall s lemma once again and find that ( [ eqn : apriori lipschitz ] ) holds for @xmath548 finally we prove assertion ( iii ) : we need to study whether solutions @xmath512 respect the constraints ( [ eqn : ml constraints ] ) . without loss of generality we may assume @xmath549 .
say we are given an initial value @xmath550 , then by part ( i ) and ( ii ) there exists a unique solution @xmath551 in @xmath552 of equation ( [ eqn : dynamic_maxwell ] ) . as before we use the notation @xmath553 for @xmath145 .
furthermore , let @xmath554 be chosen in such a way that @xmath555 and @xmath556 hold in the distribution sense .
we may write the divergence of the magnetic field of the @xmath21-th particle for each @xmath145 in the distribution sense as @xmath557 where we have used the equation of motion ( [ eqn : dynamic_maxwell ] ) and the assumption @xmath558 .
since @xmath559 , @xmath560 is in @xmath182 .
therefore , for any @xmath561 we find by fubini s theorem that @xmath562 as for any fixed @xmath72 @xmath563 } \|\nabla\wedge { { \mathbf{e}}}_{i , s}\|_{l^2_w } \leq \infty\end{aligned}\ ] ] holds .
the supremum exists because of continuity .
analogously , we find for the electric fields @xmath564 by the same argument as in ( [ eqn : div rot term ] ) the second term is zero .
we commute the divergence with the integration since @xmath78 , @xmath565 are continuous functions of @xmath72 and @xmath82 and find @xmath566 which concludes part ( iii ) and the proof
. assume the initial value @xmath567 for some @xmath164 . according to theorem [ thm : globalexistenceanduniqueness ] we know that there exists a unique solution @xmath160 which is in @xmath568 with @xmath143 .
the first aim is to see whether the fields @xmath569 are smoother than a typical function in @xmath324 .
we know that @xmath570 for any @xmath571 , but then @xmath572 in the distribution sense , where @xmath573 denotes the gradient of the divergence .
the same computation holds for @xmath85 . by inserting the constraints ( [ eqn : ml constraints ] ) we find : @xmath574 as @xmath517 we may conclude that for any fixed @xmath145 we have @xmath575 for @xmath571 and therefore @xmath576 which proves claim ( i ) .
in particular , for every open @xmath209 , @xmath569 are in @xmath577 which by theorem [ thm : htrianglekw equals h2kw ] equals @xmath578 . lemma [ lem : local equivalence ] then states @xmath579 .
this provides the necessary conditions to apply theorem [ lem : sobolev](ii ) which guarantees : in the equivalence class of @xmath84 as well as @xmath85 there is a representative in @xmath580 for @xmath581 .
we denote these smooth representatives by the same symbols @xmath84 and @xmath85 . moreover , for any @xmath582 the mapping @xmath583 , and hence , the mappings @xmath584 and @xmath584 are continuous .
hence , for any open @xmath585 and for @xmath419 the integrals @xmath586 are finite . applying sobolev s lemma in the form presented in (
* theorem 7.25 ) we yield that within the equivalence classes @xmath121 as well as @xmath122 there is a representative in @xmath171 which proves claim ( ii ) .
assume @xmath218 for @xmath173 .
then theorem [ thm : htrianglekw equals h2kw ] yields that also @xmath587 , and by theorem [ lem : sobolev](iii ) there is a constant @xmath138 such that ( [ eqn : fieldbound ] ) holds for every @xmath17 which proves claim ( iii ) and concludes the proof .
one is inclined to expect @xmath588\end{aligned}\ ] ] as the preserved energy . for the case
@xmath0 ( [ eqn : ml+si ] ) , i.e. @xmath589 for all @xmath87 , and initial values @xmath590 for weights
@xmath99 such that @xmath591 and @xmath592 $ ] this is indeed true . by computing the time derivative one can also see the mechanism of radiation damping as the particle and its own field exchange energy .
this is so beautiful that it is a pity it is ill - defined in the point - particle limit .
however , for weights @xmath99 such that @xmath593 for @xmath96 the integrals in the expression of @xmath594 diverge and the total energy is infinite . also in the case of @xmath1 ( [ eqn :
ml - si ] ) , i.e. @xmath595 , the energy ( [ eqn : energy ] ) is generically not conserved which can be understood as follows : in this case the time derivative of the electric field @xmath84 in ( [ eqn : maxwell equations ] ) depends on the position @xmath78 and velocity @xmath596 of the @xmath21-th charge which means that the charge can transfer energy by means of radiation to the field degrees of freedom . on the other hand the lorentz force law acting on the @xmath21-th charge ( [ eqn : lorentz force ] )
does not depend on the @xmath21-th field since @xmath597 .
therefore , the @xmath21-th charge can not be in turn decelerated whenever it radiates . this way
the charges can `` pump '' energy into their field degrees of freedom without `` paying '' by loss of kinetic energy .
hence , with respect to the @xmath1 expression ( [ eqn : energy ] ) is completely unnatural .
using a similar method introduced by @xcite in the context of the lorentz - dirac equations one can nevertheless define a variation of action principle to derive the @xmath1 equations of motion ( and also for the point - particle case ml - si ) from which all constants of motion can be inferred .
these are , however , more implicit as ( [ eqn : energy ] ) since they depend not only on data at one time instant @xmath72 but on whole intervals of the solution . in the special case of ( [ eqn : crucial ] ) these constants of motion
are discussed for point - particles in @xcite .
for this section let @xmath598 be a banach space with norm @xmath599 .
we consider two abstract operators @xmath126 and @xmath127 with the following properties : a. @xmath126 is closed and densely defined .
b. there exists a @xmath313 such that @xmath314 , the resolvent set of a. c. the resolvent @xmath315 of @xmath126 with respect to @xmath316 is bounded by @xmath317 , i.e. for all @xmath601 we have @xmath602 .
a. for all @xmath411 there exist @xmath605 such that for all @xmath413 @xmath606 b. [ rem : operatorj]for all @xmath411 and @xmath415 , @xmath416 and any @xmath417 such that @xmath418 for @xmath419 , the operator @xmath127 fulfills for @xmath420 : a. @xmath421 and b. @xmath422 is continuous on @xmath423 .
a. ( local existence ) for each @xmath607 with @xmath608 , there exists a @xmath415 and a mapping @xmath609 which solves the equation @xmath610 for initial value @xmath143 . furthermore , @xmath418 for @xmath419 and @xmath416 . b. ( uniqueness ) if @xmath611 for some @xmath612 is also a solution to ( [ eqn : abstract_partial_diff_eq ] ) and @xmath613 , then @xmath151 for all @xmath614 .
c. ( global existence ) assume in addition that for any solution @xmath615 of equation ( [ eqn : abstract_partial_diff_eq ] ) with @xmath616 and @xmath617 there exists a @xmath618 such that @xmath619}\|\varphi_t\|_{{{\mathcal{b}}}}\leq \constr{apriori}(t ) \end{aligned}\ ] ] then ( i ) and ( ii ) holds for any @xmath620 .
the proof of theorem [ lem : aj_local_exist_uniqueness ] is a generalization of the main proof in @xcite where a skew - adjoint operator @xmath126 was assumed .
however , for our purpose the skew - adjointness must be loosened and , here , is replaced by the conditions on @xmath126 given in definition [ def : operatora ] .
these conditions are required to apply the hille - yosida theorem @xcite to assure : [ lem : contraction_group ] operator @xmath126 introduced in definition [ def : operatora ] generates a @xmath404-contractive group @xmath405 on @xmath598 , i.e. a family of linear operators @xmath405 on @xmath598 with the properties that for all @xmath621,@xmath622 and @xmath623 : with the help of @xmath405 one can then show the existence and uniqueness of local solutions to the integral equation @xmath631 via banach s fixed point theorem which can be applied because of the convenient regularity condition imposed on operator @xmath127 .
one can show further that all these solutions are regular enough to solve the equation ( [ eqn : abstract_partial_diff_eq ] ) .
global existence is then achieved with the help of the a priori bound ( [ eqn : apriori ] ) .
see also @xcite for a beautiful exposition on the topic of time - evolution equations . 0
the proof of theorem [ lem : aj_local_exist_uniqueness ] can be found in @xcite .
\(i ) since we want to apply banach s fixed point theorem , we define a banach space on which we later define our self - mapping . for @xmath415
let @xmath632\to d(a^n)\;\big|\;&t\mapsto a^j\varphi_t\in{{\mathcal{c}}}^0([-t , t],d(a^n ) ) \text { for } j\leq n\\ & \text{and } \|\varphi\|_{x_{t , n } } : = \sup_{t\in[-t , t]}\sum_{j=0}^n\|a^j\varphi_t\|_{{{\mathcal{b}}}}<\infty\bigg\}. \end{aligned}\ ] ] @xmath633 is a banach space because it is normed and linear by definition , and complete because @xmath126 on @xmath409 is closed .
note that the mapping @xmath634 is an element of @xmath635 because for all @xmath145 and @xmath636 we have @xmath637 which is continuous and @xmath638 by lemma [ lem : contraction_group ] and because @xmath607 .
let @xmath639 which is a closed subset of @xmath635 .
next we show that @xmath640:= w_t\varphi^0 + \int_0^t w_{t - s}j(\varphi_s)\;ds \end{aligned}\ ] ] is a well - defined , contracting self - mapping provided @xmath447 is chosen sufficiently small .
the following estimates are based on the fact that for all @xmath641 we have the estimate @xmath642 for each @xmath643 $ ] .
let also @xmath644 , then the properties of @xmath127 , see definition [ def : operatorj ] , yield the helpful estimates for all @xmath645 $ ] : @xmath646 for @xmath647 hence , @xmath648 depend continuously and non - decreasingly on @xmath447 .
we show now that @xmath649 is a self - mapping .
since @xmath650 is in @xmath651 , it suffices to show that the mapping @xmath652 is @xmath653 valued , continuous and that its @xmath654 norm is finite for @xmath636 .
consider @xmath641 , so for some @xmath655 we get @xmath656{}0 \end{aligned}\ ] ] by continuity of @xmath657 , estimate ( [ eqn : j_estimate ] ) and properties of @xmath405 .
we may thus define @xmath658 as @xmath598 valued riemann integrals .
let @xmath659 be the corresponding riemann sums .
clearly @xmath660 since @xmath661 and @xmath662 @xmath663 for all @xmath145 and @xmath636 .
but @xmath126 is closed which implies @xmath664 and @xmath665 .
next we show continuity . with estimate ( [ eqn : j_estimate ] ) we get for @xmath416 : @xmath666 for @xmath667 the right - hand side goes to zero as the integrand of the second summand @xmath668 does , which by ( [ eqn : j_estimate ] ) is also bounded by @xmath669 so that dominated convergence can be used .
the self - mapping property is ensured by ( [ eqn : j_estimate ] ) : @xmath670-w_{(\cdot)}\varphi^0\|_{x_{t , n } } = \sup_{t\in [ -t , t]}\sum_{j=0}^n\left\|a^j\int_0^t w_{t - s } j(\varphi_s)\;ds\right\|_{{{\mathcal{b}}}}\\ & \leq e^{\gamma t}\sup_{t\in [ -t , t]}\sum_{j=0}^n\int_0^t \|a^jj(\varphi_s)\|_{{{\mathcal{b } } } } \;ds \leq t e^{\gamma t}\constr{clocalj1}(t)(n+1 ) .
\end{aligned}\ ] ] on the other hand for some @xmath671 we find @xmath670-s_{\varphi^0}[\widetilde\varphi_{(\cdot)}]\|_{x_{t , n } } = \sup_{t\in[-t , t]}\sum_{j=0}^n\|a^j\int_0^t w_{t - s } [ j(\varphi_s)-j(\widetilde\varphi_s)]\;ds\|_{{{\mathcal{b}}}}\\ & \leq e^{\gamma t}\sup_{t\in[-t , t]}\sum_{j=0}^n\int_0^t \|a^j[j(\varphi_s)-j(\widetilde\varphi_s)]\|_{{{\mathcal{b } } } } \;ds \leq t e^{\gamma t}\constr{clocalj2}(t)\|\varphi_{(\cdot)}-\widetilde\varphi_{(\cdot)}\|_{x_{t , n}}. \end{aligned}\ ] ] since @xmath672 and @xmath673 are continuous and non - decreasing , there exists a @xmath415 such that @xmath674<1 .
\end{aligned}\ ] ] thus , for this choice of @xmath447 , @xmath649 is a contracting self - mapping on the closed set @xmath651 so that due to banach s fixed point theorem @xmath649 has a unique fixed point @xmath675 .
next we study the differentiability of this fixed point , in particular of @xmath676 on @xmath423 for @xmath677 . as @xmath678 $ ] , definition ( [ eqn : local_contract ] ) , and @xmath607 we have @xmath679 by the properties of @xmath405 we know @xmath680 .
furthermore , @xmath681 for @xmath667 the first term on the right - hand side converges to @xmath682 because of @xmath683 and the continuity of @xmath684 in @xmath685 and @xmath686 . for @xmath667
the second term converges to @xmath687 by dominated convergence as the integrand converges to @xmath688 , and the following gives a convenient bound of it : @xmath689 collecting all terms , we have shown that @xmath690 note that the right - hand side is continuous because @xmath691 , @xmath641 and ( [ eqn : j_estimate ] ) .
hence @xmath692 and @xmath693 for all @xmath416 .
next we prove for @xmath416 and @xmath419 that @xmath609 , @xmath418 by induction .
we claim that @xmath694 holds , is continuous in @xmath72 on @xmath423 and in @xmath695 .
we have shown before that this holds for @xmath696 .
assume it is true for some @xmath697 .
we compute @xmath698 the first term on the right - hand side is continuous in @xmath72 on @xmath423 and in @xmath695 as shown before
. now definition ( [ def : operatorj])([rem : operatorj ] ) , where we have defined the operator @xmath127 , was chosen to guarantee that these properties hold also for the second term .
\(ii ) clearly , @xmath615 and @xmath699 are both in @xmath700 for any @xmath701 because they are at least once continuously differentiable
. since @xmath702 holds , we can choose @xmath703 sufficiently small such that @xmath615 and @xmath699 are also in @xmath704 and in addition that @xmath649 is a contracting self - mapping on @xmath704 .
as in ( i ) we infer that there exists a unique fixed point @xmath705 of @xmath649 which solves ( [ eqn : abstract_partial_diff_eq ] ) .
since @xmath615 and @xmath699 also solve ( [ eqn : abstract_partial_diff_eq ] ) , it must hold that @xmath706 on @xmath707 $ ] .
let @xmath708 be the supremum of all those @xmath351 and let us assume that @xmath709 .
we can repeat the above argument with e.g. initial values @xmath710 at time @xmath711 .
again we find a @xmath712 and a fixed point @xmath713 of @xmath714 so that @xmath715 on @xmath716 $ ] .
the same can be done for initial values @xmath717 at time @xmath718 .
this yields @xmath719 for @xmath720 $ ] and contradicts the maximality of @xmath708 .
hence , @xmath615 equals @xmath699 on @xmath721\cap[-\widetilde t,\widetilde t]$ ] .
\(iii ) fix any @xmath612 .
the a priori bound ( [ eqn : apriori ] ) tells us that if any solution @xmath722 with @xmath723 exists , then @xmath724}\|\varphi_t\|_{{{\mathcal{b}}}}\leq \constr{apriori}(\widetilde t)<\infty$ ] . by looking at equations ( [ eqn : banach time ] ) and
( [ eqn : banach consts ] ) we infer that there exists a @xmath725 such that for each @xmath726 $ ] the time span @xmath447 for which @xmath727 on @xmath728 fulfills @xmath729 .
let @xmath615 be the fixed point of @xmath649 on @xmath730 for @xmath703 , and let @xmath708 be the supremum of such @xmath351 .
assume @xmath731 . by taking an initial value @xmath732 for @xmath733 near to the boundary , ( i ) and ( ii )
extends the solution beyond @xmath734 and contradicts the maximality of @xmath708 .
g. bauer _ + fh mnster + bismarckstrae 11 , 48565 steinfurt , germany + _ d .- a .
+ department of mathematics , university of california davis + one shields avenue , davis , california 95616 , usa + _
institut der lmu mnchen + theresienstrae 39 , 80333 mnchen , germany | we establish global existence and uniqueness of the dynamics of classical electromagnetism with extended , rigid charges and fields which need not to be square integrable .
we consider also a modified theory of electromagnetism where no self - fields occur . that theory and our results are crucial for approaching the as yet unsolved problem of the general existence of dynamics of wheeler feynman electromagnetism , which we shall address in the follow up paper . +
* keywords : * maxwell - lorentz equations , wheeler - feynman equations , absorber electrodynamics , radiation damping , self - interaction , infinite energy solutions , weighted function spaces . +
* acknowledgments : * the authors want to thank martin kolb for his valuable comments .
d .-
a.d . gratefully acknowledges financial support from the _ bayefg _ of the _ freistaat bayern _ and the _ universit bayern e.v .
_ as well as the from the post - doc program of the daad
. 1 |
recently , babichev and fabbri @xcite have shown that the massive linearized equation around the schwarzschild black hole in both de rham , gabadadze , and tolley ( drgt ) theory @xcite and its bigravity extension @xcite gives rise to an instability of @xmath0-mode ( spherically symmetric mode ) with @xmath1 the spheroidal harmonic index .
this was done by comparing it with the four - dimensional linearized equation around the five - dimensional black string where the gregory - laflamme ( gl ) instability was found @xcite .
it turned out that the bimetric black hole is unstable provided a mass of @xmath2 satisfies a bound of @xmath3 with @xmath4 the horizon radius in the metric function @xmath5 .
we note that the limit of @xmath6 recovers the black hole in the drgt theory .
the black hole in the drgt theory is also unstable because @xmath7 satisfies a bound of @xmath8 .
in addition , the authors @xcite have confirmed this result by considering the schwarzschild - de sitter black hole and extending the @xmath9 mode to generic modes of @xmath10 .
these results may indicate an important fact that the static black holes do not exist in massive gravity theory . on the other hand ,
whitt @xcite has insisted thirty years ago that provided both massive spin-0 and spin-2 gravitons are non - tachyonic , the schwarzschild black hole is classically stable in a massive theory of fourth - order gravity when he uses the linearized - ricci tensor equation . in this case , one does not worry about the ghost instability arising from the fourth - order gravity theory because the linearized - ricci tensor satisfies a second - order tensor equation .
recently , the author has revisited this stability issue .
as expected , it was shown that the black hole in fourth - order gravity with @xmath11 is unstable provided the graviton mass of @xmath12 satisfies a bound of @xmath13 @xcite .
this was performed by comparing the linearized - ricci tensor equation with the four - dimensional metric perturbation equation around the five - dimensional black string @xcite . in this work ,
we wish to reexamine the classical stability of schwarzschild - ads ( sads ) black hole in einstein - weyl gravity which was known to be stable against the metric perturbation @xcite .
by contrast , it is shown that solving both the linearized - einstein tensor equation and the metric perturbation equation exhibit unstable modes featuring the gl instability of five - dimensional ads black string @xcite .
it confirms that the gl instability of the black hole in the einstein - weyl gravity is due to the massiveness but not a feature of fourth - order gravity giving ghost states .
taking into account the number of degrees of freedom ( dof ) , it is helpful to show why the sads black hole is physically stable in the einstein gravity @xcite , whereas the sads black hole is unstable in the einstein - weyl gravity .
the number of dof of the metric perturbation is 2 in the einstein gravity , while the number of dof is 5 in the einstein - weyl gravity .
the @xmath14-mode analysis of the massive graviton with @xmath15 dof shows the gl instability .
the @xmath14-mode analysis is relevant to the massive graviton in the einstein - weyl gravity but not to the massless graviton in the einstein gravity .
we start with the fourth - order gravity in ads@xmath16 spacetimes @xmath17 \label{action}\end{aligned}\ ] ] with two arbitrary parameters @xmath18 and @xmath19 .
although this theory is renormalizable in minkowski spacetimes @xcite , the massive spin-2 graviton suffers from having ghosts .
a massive spin-0 graviton is decoupled for the choice of @xmath11 , which leads to a critical gravity with @xmath20 @xcite . for a higher dimensional critical gravity ,
see a reference of @xcite .
the einstein - weyl gravity is defined under the condition of @xmath11 as @xmath21 \label{ewaction}\end{aligned}\ ] ] with @xmath22 here the last of gauss - bonnet term could be neglected because it does not contribute to equation of motion . from ( [ action ] )
, the einstein equation is derived to be @xmath23 where the einstein tensor is given by @xmath24 and @xmath25 takes the form @xmath26 it is wellknown that eq.([equa1 ] ) provides the sads black hole solution @xcite @xmath27 with the metric function @xmath28 here @xmath29 denotes the curvature radius of ads@xmath16 spacetimes .
we note that a mass parameter of @xmath30 is not the horizon radius @xmath31 which is obtained as a solution to @xmath32 .
hereafter we denote the background quantities with the `` overbar '' . in this case , the background ricci tensor is given by @xmath33 it is easy to show that the sads black hole solution ( [ sch ] ) to the einstein equation of @xmath34 is also the solution to the einstein - weyl gravity when one substitutes ( [ beeq ] ) together with @xmath35 into ( [ equa2 ] ) . to perform the stability analysis , we usually introduce the metric perturbation around the sads black hole @xmath36 then , the linearized einstein equation takes the form @xmath37\delta g _ { \mu\nu}&+&\alpha\big[\bar{\nabla}^2\delta g_{\mu\nu}+2\bar{r}_{\rho\mu\sigma\nu}\delta g^{\rho\sigma}-\frac{2\lambda}{3 } \delta r \bar{g}_{\mu\nu}\big ] \nonumber \\ & + & ( \alpha+2\beta)\big[-\bar{\nabla}_\mu\bar{\nabla}_\nu+\bar{g}_{\mu\nu}\bar{\nabla}^2 + \lambda \bar{g}_{\mu\nu}\big ] \delta r=0,\end{aligned}\ ] ] where the linearized einstein tensor , ricci tensor , and ricci scalar are given by @xmath38 with @xmath39 .
it is very difficult to solve the linearized equation ( [ lin - eq ] ) directly because it is a coupled second - order equation for @xmath40 and @xmath41 .
thus , we attempt to decouple @xmath41 from ( [ lin - eq ] ) . for this purpose ,
we take the trace of ( [ lin - eq ] ) which leads to @xmath42\delta r=0.\ ] ] it implies that the dalembertian operator could be removed if one chooses @xmath43 in this case , the linearized ricci scalar is constrained to vanish @xmath44 plugging @xmath45 into eq .
( [ lin - eq ] ) leads to the equation for the linearized einstein tensor solely @xmath46 this shows clearly why we consider the einstein - weyl gravity ( [ ewaction ] ) with @xmath11 instead of the fourth - order gravity action ( [ action ] ) with arbitrary @xmath18 and @xmath19 .
before we proceed , we wish to mention that the metric perturbation is not suitable for analyzing the sads black hole stability in the einstein - weyl gravity . for simplicity
, we consider the ads@xmath16 spacetimes background whose curvature tensor takes a simple form @xmath47 after choosing the transverse - traceless gauge ( ttg ) @xmath48 eq .
( [ slin - eq ] ) leads to a fourth - order differential equation @xcite @xmath49 which may imply a massless spin-2 graviton equation @xmath50 and a massive spin-2 graviton equation @xmath51 here the mass squared is given by @xmath52 in ads@xmath16 spacetimes , the stability condition is given by the absence of tachyonic instability ( @xmath53 ) @xcite , which implies that @xmath19 must satisfy @xmath54 in the massless case of @xmath55 , eq.([four - eq ] ) leads to that for a critical gravity @xmath56 however , it was shown that a general mode of @xmath57 suffers from negative norm states unless one truncates out the log - mode by imposing appropriate ads@xmath16 boundary conditions .
up to now , there is no consistent truncation mechanism to eliminate the log - mode .
we recall that this problem arises because we work with the fourth - order derivative equation ( [ four - eq ] ) for the metric perturbation . in this work
, we do not consider a new unitary gravity for @xmath58 @xcite because it has still a non - unitarity problem like the critical gravity .
going back to the sads black hole ( [ sch ] ) , we rewrite eq .
( [ slin - eq ] ) as a second - order equation for the linearized einstein tensor @xmath59 if one introduces the lichnerowicz operator @xmath60 the corresponding equation could be rewritten as @xmath61 taking into account the ttg ( [ ttg ] ) , the linearized einstein tensor reduces to @xmath62 then , one can rewrite ( [ g - eq ] ) as a fourth - order equation for @xmath63 @xmath64 which is similar to ( [ four - eq ] ) in ads@xmath16 spacetimes .
( [ four - eqq ] ) may imply a linearized massless equation around the sads black hole @xcite @xmath65 and a linearized massive equation for @xmath66 @xmath67 at this stage , we wish to point out the difference between ( [ se1m - eq ] ) and ( [ lem2-eq ] ) .
the former equation is a second - order equation for the linearized einstein tensor , whereas the latter is a suggesting second - order equation from the fourth - order equation ( [ four - eqq ] ) for the metric perturbation .
it is known that the introduction of fourth - order derivative terms gives rise to ghost - like massive graviton @xcite , which may imply an instability of a black hole even if a black hole solution exists .
hence , even though ( [ se1-eq])[([se2-eq ] ) ] were frequently used as a linearized massless [ massive ] equation around the ads@xmath16 spacetimes @xcite , their validity is not yet proved because they are free from ghost states . in order to check
whether ( [ se2-eq])[([lem2-eq ] ) ] are reliable or not , we note that our action ( [ ewaction ] ) reveals ghosts when we perform the metric perturbation @xmath63 around the minkowski spacetimes with @xmath68 @xcite .
( [ four - eq])[([four - eqq ] ) ] take the form in the minkowski background @xcite @xmath69 with an external source @xmath70 . replacing @xmath71 by @xmath72 ,
the metric perturbation is given by @xmath73 which the last term spoils the unitarity .
hence , splitting ( [ four - eq])[([four - eqq ] ) ] into two second - order equations ( [ se1-eq])[([lem1-eq ] ) ] and ( [ se2-eq])[([lem2-eq ] ) ] is dangerous because the ` @xmath74 ' sign in the front of ( [ se2-eq])[([lem2-eq ] ) ] is missed . as is shown in ( [ h - s ] ) , the ghost arises from this sign when one performs the partial fraction . to this end
, the authors in @xcite have found the two on - shell energies on the ads@xmath16 spacetime background @xmath75 when they compute each hamiltonian which satisfies ( [ se1-eq ] ) and ( [ se2-eq ] ) , respectively .
thus , for @xmath76 , ghost - like massive excitation is not avoidable . in order for the theory to be free from ghosts
, one needs to choose @xmath77 ) which corresponds to the critical gravity where a massive graviton becomes a massless graviton . because of a missing of ` @xmath74 ' sign
, we may insist that eq.([se2-eq])[([lem2-eq ] ) ] by itself do not represent a correct linearized equation for studying the stability of the sads black hole in the einstein - weyl gravity .
however , the overall ` @xmath74 ' sign in ( [ se2-eq])[([lem2-eq ] ) ] does not make any difference unless an external source is introduced in the right - hand side as eq.([source ] ) does indicate .
therefore , the fourth - order gravity does not automatically imply the instability of the black hole even if one uses ( [ lem2-eq ] ) .
hopefully , if one uses ( [ se1m - eq ] ) instead of ( [ lem2-eq ] ) , one is free from the ghost issue because ( [ se1m - eq ] ) is a genuine second - order equation .
in einstein gravity , the linearized equation around the schwarzschild black hole is given by @xmath78 with @xmath79 ( [ ricc - t ] ) .
then , the metric perturbation @xmath63 is classified depending on the transformation properties under parity , namely odd and even . using the regge - wheeler @xcite and zerilli gauge @xcite , one obtains two distinct perturbations : odd with 2 dof and even with 4 dof .
this implies that one starts with 6 dof after choosing the regge - wheleer gauge , leading to 2 dof ( 1 for odd and 1 for even ) for a massless spin-2 graviton propagation .
the schwarzschild black hole is stable against the metric perturbation @xcite .
performing the stability analysis of the sads black hole in einstein gravity , one has to use the linearized equation @xmath80 which was tuned out to be stable by following the regge - wheeler prescription @xcite . in these cases ,
the @xmath0-mode analysis is not necessary to show the stability of the schwarzschild and sads black holes because the massless spin-2 graviton requires modes with @xmath81 . however , the @xmath14-mode analysis is responsible for detecting an instability of a massive graviton propagating on the sads black hole in einstein - weyl gravity .
the even - parity metric perturbation is designed for a single @xmath14-mode analysis in the massive gravity and whose form is given by @xmath82 and @xmath83 as @xmath84 even though one starts with 4 dof , they are related to each other when one uses the ttg ( [ ttg ] ) .
hence , we expect to have one decoupled equation for @xmath85 . for a massive gravity theory in the minkowski background , there is correspondence between linearized ricci tensor @xmath86 and ricci spinor @xmath87 when using the newman - penrose formalism @xcite . here
the null real tetrad is necessary to specify polarization modes of a massive graviton , as the massive gravity requires null complex tetrad to specify six polarization modes @xcite .
this implies that in fourth - order gravity theory , one may take the linearized ricci tensor @xmath86 ( [ ricc - t ] ) with 6 dof as physical observables @xcite . requiring @xmath88 further
, the dof of @xmath86 is five which is the same dof for the metric perturbation @xmath63 in massive gravity theory . at this stage ,
we stress again that ( [ se1m - eq ] ) is considered as the second - order equation with respect to @xmath40 , but not the fourth - order equation ( [ four - eq ] ) for @xmath63 .
hence , we propose @xmath40 as physical observables propagating on the sads black hole background instead of @xmath89 on the schwarzschild black hole background .
also , we have the tracelessness of @xmath90 and the transversality of @xmath91 from the contracted bianchi identity
. then , @xmath40 describe exactly five dof propagating on the sads black hole background without ghosts . since eq.([lem2-eq ] ) is the same linearized equation for four - dimensional metric perturbation around five - dimensional black string , we follow the gl instability analysis in ads@xmath16 spacetimes @xcite .
eliminating all but @xmath85 , eq.([lem2-eq ] ) reduces to a second - order equation for @xmath85 @xmath92 where @xmath93 and @xmath94 were given by ( 20 ) in @xcite .
we stress again that the @xmath14-mode perturbation is described by single dof but not 5 dof .
the authors in @xcite have solved ( [ second - eq ] ) numerically and found unstable modes for @xmath95 .
1 that is generated from the numerical analysis . we note that @xmath96 correspond to @xmath97 , respectively . from the observation of fig . 1 with @xmath98 , we find unstable modes for @xmath99 with the mass @xmath100 similarly , we find eq.([lem2-eq ] ) when we replace @xmath101 by @xmath63 in ( [ se1m - eq ] ) .
hence , a relevant equation for @xmath102 takes the same form @xmath103 which shows the same unstable modes appeared in fig .
this implies that even if one uses ( [ lem2-eq ] ) as a linearized massive equation @xcite , our conclusion remains unchanged because ( [ lem2-eq ] ) and ( [ se1m - eq ] ) are the same equation for different tensors .
consequently , the instability arises from the massiveness ( @xmath104 ) but not from a feature of the fourth - order equation which gives the ` @xmath74 ' sign ( ghost = negative norm state ) when one splits it into two second - order equations .
this implies that static black holes in massive gravity theory do not exist and/or they do not form in the gravitational collapse . if a black hole was formed in the massive gravity theory , one may ask what is the end - state of such instability . for unstable black strings of sads@xmath105r with translational symmetry ,
there are some evidences that break - up occurs @xcite . however , we consider a spherically symmetric black hole in four dimensions
. a possible end - state may be a spherically symmetric black hole endowed with a graviton cloud @xcite finally , in the case of @xmath106 , the theory becomes massless and is stable against the einstein tensor perturbation .
however , this corresponds precisely to the critical gravity when one uses the metric perturbation .
here we have a non - unitarity issue due to the log - mode like ( [ crit - eq ] ) .
also , one finds that @xmath107(adt mass)=0 and @xmath108(wald s entropy)=0 at the critical point , leading to a vacuum but not a black hole @xcite .
e. babichev and a. fabbri , class .
grav . *
30 * , 152001 ( 2013 ) [ arxiv:1304.5992 [ gr - qc ] ] . c. de rham and g. gabadadze , phys .
d * 82 * , 044020 ( 2010 ) [ arxiv:1007.0443 [ hep - th ] ] .
s. f. hassan and r. a. rosen , jhep * 1202 * , 126 ( 2012 ) [ arxiv:1109.3515 [ hep - th ] ] .
v. cardoso and j. p. s. lemos , phys .
d * 64 * , 084017 ( 2001 ) [ gr - qc/0105103 ] .
a. ishibashi and h. kodama , prog .
phys . *
110 * , 901 ( 2003 ) [ hep - th/0305185 ] .
k. s. stelle , phys .
d * 16 * , 953 ( 1977 ) .
s. deser , h. liu , h. lu , c. n. pope , t. c. sisman and b. tekin , phys .
d * 83 * , 061502 ( 2011 ) [ arxiv:1101.4009 [ hep - th ] ] .
s. -j .
hyun , w. -j .
jang , j. -h .
jeong and s. -h .
yi , jhep * 1201 * , 054 ( 2012 ) [ arxiv:1111.1175 [ hep - th ] ] . | we investigate the classical stability of schwarzschild - ads black hole in a massive gravity theory of the einstein - weyl gravity .
it turns out that the linearized einstein tensor perturbations exhibit unstable modes featuring the gregory - laflamme instability of five - dimensional ads black string , in contrast to the stable schwarzschild - ads black hole in einstein gravity .
we point out that the instability of the black hole in the einstein - weyl gravity arises from the massiveness but not a feature of fourth - order derivative theory giving ghost states .
+ yun soo myung + institute of basic sciences and department of computer simulation , inje university gimhae 621 - 749 , korea + pacs numbers : 04.70.bw , 04.50.kd |
the 1-band hubbard model on a two - dimensional square lattice has the hamiltonian @xmath2 here , @xmath3 ( @xmath4 ) creates ( annihilates ) an electron with spin @xmath5 in a wannier orbital centered at lattice site @xmath6 .
the particle density at each site is given by @xmath7 .
the first sum for the kinetic energy is restricted to include only the hopping matrix element @xmath8 between next - nearest neighbor sites @xmath9 .
periodic boundary conditions are used throughout the following work .
the second sum describes for @xmath10 an on - site coulomb repulsion between particles of opposite spin that share the same lattice site . in the present work we restrict ourselves to @xmath11 .
the chemical potential @xmath12 in the third sum controls the occupation of the finite lattice in the finite - temperature grand canonical quantum - monte - carlo ( qmc ) simulation we preformed . at half - filling , particle - hole symmetry of
the kinetic and @xmath13-term implies @xmath14 .
the analytic continuation of the dynamic imaginary - times qmc data to the real frequency axis is performed with state - of - the - art maximum - entropy ( me ) techniques . for exhausting discussions concerning the qmc and me methods we refer the reader to @xcite . +
the 1-band hubbard model exhibits several energy scales : in the repulsive case the high - energy scale @xmath13 is important in determining the insulating gap at half - filling , @xmath15 .
an important low - energy scale is set by the exchange interaction @xmath16 : in second order perturbation theory two particles with different spins on neighboring lattice sites can exchange via a virtual double occupation .
this process is the source for the strong antiferromagnetic ( af ) correlations found near and at half - filling .
+ with the exception of some known symmetry properties like invariance under global spin - rotation , whose generators form the @xmath17algebra , and invariance under @xmath18-transformation , i.e. , charge - conservation , as well as the particle - hole transformation , no rigorous results are known for the 1-band hubbard model in two - dimensions @xcite .
the mermin - wagner theorem @xcite prevents a long - range ordered state in a two - dimensional system for finite temperatures , but it is commonly believed that the ground state of the spin-@xmath19 heisenberg antiferromagnet , i.e. the large-@xmath13 limit of the repulsive half - filled hubbard model @xcite , shows long - range nel order in two dimensions .
nel order results also in the weak - coupling limit @xcite , where the gap @xmath20 is due to a spin - density - wave ( sdw ) instability related to perfect nesting .
+ it might seem that due to the mermin - wagner theorem the physics of the ordered phase is out of reach for our numerical technique , which is limited to finite temperatures .
however , one may assume that the system ` effectively orders ' as soon as the spin - correlation length becomes comparable to the system size .
due to the periodic boundary conditions the spin - correlation function @xmath21 is a periodic function of @xmath22 and if the value of this function at the maximum value @xmath23 ( with @xmath24 @xmath25 cluster size ) is still appreciable , we may expect that the system is ` effectively ordered ' .
a rough measure would be the spin - correlation length @xmath26 , obtained by fitting @xmath27 to the form @xmath28 .
since the infinite system has the nel temperature @xmath29 , the spin - correlation length diverges as @xmath30 and we may expect that in a _
system @xmath26 becomes comparable to the cluster size @xmath24 at a _
finite _ temperature which depends on the lattice size .
below this temperature we then expect that the system resembles the ordered phase , although the spin - rotation symmetry persists even in this case . in that sense
, the finite size of the system creates an artificial ` nel temperature ' which , however , depends on @xmath24 and @xmath31 etc and has no real counterpart in the infinite system .
this has to be kept in mind when discussing the results .
+ we proceed by discussing some known approximations to the hubbard model .
the ` classical ' approximation to the hubbard model is the so - called hubbard - i approximation@xcite .
its essence is the splitting of the electron annihilation operator into the two ` eigenoperators ' of the interaction part : @xmath32 whence @xmath33 the physical content of the hubbard - i approximation , which neglects any spin - correlations , becomes clear by realizing @xcite that the equations of motion in this approximation are completely equivalent to an ` effective hamiltonian ' for double occupancy - like particles @xmath34 and hole - like particles @xmath35 : @xmath36 this hamiltonian contains terms which describe the pair creation of a hole and a double occupancy on nearest neighbors @xmath9 , terms which describe the propagation of these effective particles , and an additional energy of formation of @xmath13 for the double occupancy .
the matrix elements for the propagation are reduced by a factor of @xmath19 , because in an uncorrelated spin - background there is a probability of @xmath19 for the spin on a nearest neighbor to have the proper direction to allow for the hopping of an electron . solving ( [ hubeff ] ) by fourier and bogoliubov transform : @xmath37 yields the standard dispersion relation , which consists of the upper and lower hubbard bands : @xmath38 here @xmath39 denotes the free tight - binding dispersion @xmath40 . using ( [ interactioneigenoperators ] ) we also obtain the correct spectral weights of the two hubbard bands : @xmath41 as noted above , the key assumption in the hubbard i approximation is the neglect of spin - correlations .
therefore , we expect that this approximation will become inaccurate as soon as spin - correlations become sufficiently strong so as to appreciably influence the propagation of holes and double occupancies
. this effect will be strongest in the nel ordered phase believed to be realized in the ground state .
if we choose the ` spin - background ' , in which the holes and double occupancies propagate , to be the nel state , the double occupancy @xmath42 with spin @xmath43 can exist only on the @xmath44 sublattice and vice versa .
similarly , a hole @xmath45 can be created only on the @xmath44 sublattice and vice versa .
we , thus , expect that we have to modifiy the hubbard - i hamiltonian ( [ hubeff ] ) into : @xmath46 where @xmath47 ( @xmath48 ) denotes the @xmath44 ( @xmath43 ) sublattice and @xmath49 denotes the set of nearest neighbors of site @xmath6 .
note that the @xmath50 and @xmath51 propagation terms drop out ( because of the @xmath47 and @xmath48 sublattices ) , and the matrix element for pair creation are @xmath8 rather than @xmath52 - this takes into account the fact that the spins on neighboring sites are antiparallel with probability @xmath53 rather than @xmath19 as was the case in the paramagnetic state .
fourier and bogoliubov transformation of ( [ swham ] ) yields the dispersion @xmath54 and using @xmath55 ( where @xmath56 is within the af brillouin zone ) , we find the spectral weight of these bands : @xmath57 this is precisely what is obtained from the sdw mean - field treatment of the hubbard model by setting the staggered magnetization @xmath58 to a value of @xmath53 ( which is a good approximation in the limit of large @xmath13 ) . in general , the sdw mean - field approximation models the af nel state by assuming @xmath59 with af nesting vector @xmath60 and staggered magnetization @xmath61@xcite .
this results in the two - band dispersion @xmath62 and the spectral weight @xmath63 the gap parameter is @xmath64 and the staggered magnetization @xmath58 is determined self - consistently from the following equation : @xmath65 the solution of this self - consistency equation yields the value @xmath66 at @xmath11 , the value used in the present work .
if we set @xmath67 on the other hand , as would be apropriate for @xmath68 , we obviously recover the results from our hubbard - i - like hamiltonian ( [ swham ] ) .
+ in the two limiting cases of no spin - correlations and of perfect nel order we can thus treat the hubbard model in a quite analogous fashion and , as will be seen below , the hubbard - i results are indeed a good approximation to the actual spectral function in the limit of high temperature .
the main problem then is how to describe the effect of spin - fluctuations and how to manage the crossover from the completely disordered to the nel ordered phase .
below , we will adress this crossover by qmc simulations .
+ thereby we want to take advantage of the possibility to calculate the spectra of specifically designed ` diagnostic operators ' .
the first one of these is the _ `
shadow ' operator _ : @xmath69 we note first of all , that the fourier transform of this operator , @xmath70 has precisely the same quantum numbers as the ordinary electron operator @xmath71 : momentum @xmath72 , total spin @xmath19 , @xmath73-spin @xmath5 , and identical point group symmetry at high symmetry momenta .
it follows that the poles in its dynamical correlation functions @xmath74 originate from exactly the same final states @xmath75 as those of the photoemission spectrum .
it can happen only accidentally that a given state @xmath76 has an exactly vanishing weight in one of the correlation functions , but not the others . in this case , however , any arbitrarily small perturbation will remove the accidental vanishing of the peak .
the only thing that can and will be different in the spectra of the diagnostic operator , are the _ weights _ of the peaks , @xmath77 .
in fact , comparison of equations ( [ crep ] ) and ( [ hubihafwght ] ) shows immediately that both for the hubbard i approximation and for the sdw - approximation use of the operator @xmath70 instead of @xmath71 exchanges the weights of the two hubbard / sdw bands .
it follows that band portions which have a small spectral weight in the spectra of the ordinary electron operator will aquire a large spectral weight in the spectrum of @xmath70 and vice versa . therefore , this diagnostic operator is a useful tool to map out the ` shadowy ' parts of the spectra in the outer parts of the brillouin zone .
an additional benefit is that since the me technique resolves peaks with large spectral weight more reliably than those with small weight , we can get more precise information about the dispersion of these bands with weak intensity .
the second diagnostic operator that we will be using is the _ `
spin-@xmath19 string ' operator _
@xmath78 where the sum on the r.h.s .
runs over the nearest neighbors @xmath79 of site @xmath6 .
this operator is the clebsch - gordan contraction of the adjoint rank-@xmath53 spinor @xmath80 and the spin-@xmath53 vector operator @xmath81 into yet another adjoint rank-@xmath53 spinor .
it describes the annihilation of a ` dressed ' electron , i.e. , an electron with a spin - excitation on a nearest neighbor .
again , since the fourier transforms of the diagnostic operator , @xmath82 , agrees with the electron annihilation operator @xmath71 in all conceivable quantum numbers ( momentum , total spin , @xmath73-component of the spin , point group operations in the case of high symmetry momenta ) , it obeys precisely the same selection rules , whence it is just as good as the @xmath83-operator itself to map out the band structure .
+ one further point , which is important from the technical point of view , is the following : under the particle - hole transformation @xmath84 we have @xmath85 and @xmath86 .
this implies that at half - filling the spectra of the shadow and spin-@xmath19 string operator obey the same particle - hole - symmetry as those of the ordinary electron operator , i.e. @xmath87 in our maxent program particle - hole - symmetry is not implemented as an additional constraint , in other words : the maxent procedure does not ` know ' about this additional symmetry .
the degree to which ( [ phsym ] ) is obeyed in the final spectra thus gives a good check for the accuracy of the reconstructed spectra .
this is of particular importance in the case of the spin-@xmath19 string operator because the wick - contraction of this operator on any given time slice produces a total of @xmath88 products of noninteracting green s functions .
the computation is , therefore , much more prone to inaccuracies so that an additional check is desirable .
+ the present work is organized as follows :
first , we compare the temperature dependent dynamic single - particle properties of the hubbard model with the predicitions of the mean - field sdw and hubbard - i approximations .
in addition , we consider the temperature dependent two - particle excitations
. then , we will use our first diagnostic operator , the shadow operator @xmath70 , to show the existence of a total of four bands in the photoemission spectrum and to shed light onto the temperature dependent crossover from the sdw to the hubbard - i regime .
then we will investigate the 4-band structure in more detail : after a phenomenological fit , which is able to produce a total of four bands , we will consider the string picture which naturally leads to our second diagnostic operator , the spin-@xmath19 string operator @xmath82 . this operator will be used to ultimately reveal the underlying mechanisms behind the generation of the 4-band structure , namely the dressing of the photoholes by clouds of af spin - excitations . to resolve the ` af mirror image ' of the narrow quasiparticle spectral weight features between @xmath89 and @xmath90 around momentum @xmath91
we introduce also a spin-@xmath92 string operator .
finally , we will concentrate on the doped regime , thereby showing the violation of the luttinger theorem near half - filling , and discuss the hole concentration range in which these dressing effects dominante the low - energy physics .
we start in figure [ fig1 ] with the discussion of the angle - resolved single - particle spectral function @xmath93 for @xmath11 and various temperatures in the range from @xmath94 to @xmath95 . in the left column of the figure
the spectral functions are shown as ` grey - scale ' plots versus momentum @xmath96 and energy @xmath97 with dark ( light ) areas corresponding to large ( small ) spectral weight .
the same spectra are shown also in the right column , but now as line - plots at each momentum @xmath96 .
the qmc data in the figure are compared to the renormalized results of the hubbard - i approximation at high and medium temperatures and with the renormalized results of the mean - field sdw approximation at the lowest temperature , @xmath95 . `
renormalized ' means here that we have readjusted the parameters @xmath13 and @xmath8 in ( [ hubihafdisp ] ) and ( [ sdwdisp])so as to obtain an optimal fit to the ` bands ' of high spectral weight in the spectra .
these approximate dispersions are plotted as solid lines in the left column , while their spectra are shown in the right column as line - plots at each momentum @xmath96 .
thereby we have assumed a lorentzian lineshape with a suitably chosen temperature dependent width ( the hubbard - i approximation does not provide any information about linewidths ) .
+ starting at the highest calculated temperature , @xmath94 , we find that the hubbard - i approximation with practically unrenormalized parameters ( @xmath98 and @xmath99 ) fits the qmc spectral functions almost perfectly regarding both the general dispersion and the distribution of spectral weight ( quadratic deviation per degree of freedom @xmath100 ) .
this is not surprising since the hubbard - i approximation is derived for the paramagnetic state thereby neglecting all effects of spin - correlations , i.e. , the hubbard - i approximation essentially describes the interplay between itinerant electrons and strong on - site repulsion .
this should be a reasonable assumption at this high temperature since all relevant spin - degrees of freedom are thermally excited .
+ lowering the temperature to @xmath101 and further to @xmath102 , the data show that the hubbard - i approximation increasingly fails to reproduce the entire spectrum and is only able to fit the peaks with maximal spectral weight reasonably well . even then , in order to achieve these fits , one already has to renormalize the free parameters strongly .
the values we found are @xmath103 and @xmath104 with @xmath105 at @xmath101 and @xmath106 and @xmath107 with @xmath108 at @xmath102 .
the peaks that are missed by the hubbard - i approximation are the states that form the first ionization / affinity states around momentum @xmath109 on the photoemission / inverse photoemission side and two rather dispersionless bands at higher energies of @xmath110 .
the former states were previously resolved by preuss and co - workers @xcite .
alltogether one can distinguish a total of four ` bands ' in the single particle spectral density . as will be seen below , the temperature / doping regime where this 4-band structure is seen coincides with the regime where a collective low - energy mode with momentum @xmath111 in the spin - response exists .
inspite of this , however , we stress that the 4-band structure can not be explained by a backfolding of the band structure due to ordering effects since the spin - correlation length is @xmath112 lattice spacings at temperatures @xmath113 .
+ for the lowest temperature , @xmath95 , the qmc data are compared with the results from the af sdw approximation . as in the case of the hubbard - i approximation at medium temperatures , the lowermost spectra of figure [ fig1 ] show that the sdw approximation is only able to fit the peaks with large spectral weight .
again one has to renormalize the free parameters heavily to values of @xmath114 and @xmath115 with @xmath116 .
moreover , as was the case for the hubbard - i approximation at higher temperatures , the sdw approximation neither explains the states that form the first ionization / affinity states around momentum @xmath109 on the photoemission / inverse photoemission side , nor the two dispersionless bands at higher excitation energies which can be seen rather clearly in the spectra .
+ all in all , the overall distribution of spectral weight is roughly reproduced by the hubbard - i and sdw approximations as long as one forgets about the 4-band structure .
in fact it is well known that the _ integrated _ photoemission or inverse photoemission weight ( that means the electron momentum distribution ) at each @xmath96point is reproduced quite well by the hubbard - i approximation and the related @xmath117-pole approximation@xcite .
+ as already mentioned , the emerging of the 4-band structure in the photoemission somewhere in between @xmath101 and @xmath102 is closely related to a change in the spin - response : to illustrate this we consider figure [ fig2 ] , which shows the spin - correlation function , @xmath118 ( left column ) , and the charge - correlation function , @xmath119 ( right column ) , for different temperatures . whereas the spin - response is entirely incoherent at @xmath101 , with decreasing temperature it can be fitted increasingly well by the spin - wave dispersion @xmath120 this result is known from previous calculations @xcite , which demonstrated that the two - particle correlation functions like the spin - response can be described within the sdw approximation even for large values of the interaction @xmath13 .
the energy scale @xmath121 directly manifests itself in the spin - response since the spin - wave dispersion takes the value of @xmath122 at momentum @xmath90 .
the fit parameters are @xmath123 with @xmath124 at @xmath102 and @xmath125 with @xmath126 at @xmath127 .
the latter is already quite close to the strong coupling estimate @xmath128 .
furthermore , the figure shows that with decreasing temperature the spin - response concentrates its weight more and more at the af momentum @xmath60 ( as it is the case in af spin - wave theory ) and at a characteristic energy @xmath129 .
the latter decreases with decreasing temperature , i.e. , the spin - response comes closer and closer to the predictions of af spin - wave theory ( [ spinwave ] ) .
the spin - correlation length @xmath130 can be derived from a real - space fit of the qmc equal - imaginary - times spin - correlation function @xmath27 to the form @xmath131 thereby incorporating the effects of the periodic boundary conditions . while this is the best one can do on a finite lattice with periodic boundary conditions , the fit will only lead to roughly correct values due to the relative small system size of @xmath132 .
the values obtained for the spin - correlation length then are @xmath133 , @xmath134 , @xmath135 , @xmath136 and @xmath137 for @xmath101 , @xmath102 , @xmath138 , @xmath139 and @xmath95 , respectively .
we actually believe that the spin - correlation length reaches the system size already at a temperature of @xmath140 , because at this temperature the fit results in values between @xmath141 and @xmath142 ( with the exponent @xmath143 set to zero or not ) but always with error bars of roughly the system size . +
the charge - response @xmath119 , on the other hand , is rather broad in both momentum @xmath96 and energy @xmath144 for all temperatures studied .
furthermore , the charge - response is gapped for temperatures below @xmath145 and , therefore , can certainly not be responsible for any low - energy features of the single particle spectrum . + it is then quite obvious that at roughly the same temperature where the two narrow dispersive quasiparticle - like bands ( that can not be interpreted within the framework of the hubbard - i or sdw approximations ) appear in the single particle spectrum , the spin - response develops a sharp collective low - energy mode .
we conclude that the underlying mechanism behind the occurrence of the 4-band structure consists in dynamical magnetic correlation effects , which are beyond the scope of the hubbard - i and sdw approximations .
+ in the following , we want to explore the single particle spectrum by means of our _ diagnostic operators_. the first of them is the shadow operator @xmath146 of equation ( [ shadoweqt ] ) which will be used to transfer spectral weight from the inner parts of the brillouin zone to the outer ones in case of normal photoemission @xmath147 .
as already discussed above , this also improves the resolution of the me method in this region , since its resolution strongly depends on the spectral weight at a certain position .
nevertheless the spectrum of the shadow operator has to exhibit exactly the same peak positions as the normal photoemission spectrum .
+ figure [ fig3 ] shows the angle - resolved spectral function of the shadow operator for moderate and low temperatures .
as expected , the shadow operator has its main spectral weight near @xmath91 on the photoemission side and near @xmath89 on the inverse photoemission side .
furthermore , it s spectrum supports the existence of a total of four bands , because it resolves a group of peaks forming dispersionless bands at energies of @xmath148 , a region where the normal photoemission spectrum exhibits only some weak and smeared - out spectral weight .
these two dispersionless bands at energies of @xmath110 are inconsistent with the dispersions of the hubbard - i and sdw approximations of figure [ fig1 ] .
we will further address this topic later in this work .
+ next , we turn in more detail to the temperature dependence of the photoemission spectrum .
figure [ fig4 ] shows some close - ups of the normal photoemission spectrum and of the spectrum of the shadow operator at momentum @xmath91 and different temperatures . for the normal photoemission operator
these close - ups show a peak at @xmath149 , which would be consistent with hubbard - i ( see figure [ fig2 ] ) . in the spectrum of the shadow operator
this feature is visible as a single resolved peak only at the highest temperature , @xmath101 whereas for temperatures down to @xmath139 there is only some diffuse weight at this position . in the ordinary photoemisson spectrum the peak looses spectral weight with decreasing temperature .
it disappears completely at @xmath150 where the spin - correlation length @xmath130 reaches the system size ( see above ) .
thus , the temperature @xmath140 where we lose the hubbard - i - like peak at @xmath149 and @xmath91 coincides quite accurately with the temperature where ` effective ' long - range order sets in .
moreover , we find that as the normal photoemission spectrum loses the peak at @xmath149 and @xmath91 , the spectrum of the shadow operator gains weight at @xmath151 . thus , we expect that both features are closely related to the temperature development of the spin - correlation length @xmath130 .
we note , however , that the crossover in the shape of the dispersion from hubbard - i - like to sdw - like occurs in a quite unexpected way : the topmost band at @xmath111 does _ not _ deform into the sdw form in any continuous way , but simply ` fades away ' and eventually vanishes at the transition .
+ a further surprising result is the following : at @xmath152 neither the ordinary electron operator nor the shadow operator pick up the ` af umklapp band ' corresponding to the narrow dispersive band seen for example at @xmath153 for @xmath154 , i.e. there is no corresponding band at @xmath153 and @xmath155 .
note that in the framework if the sdw - approximation the shadow operator _ must _ reproduce this umklapp - band at @xmath111 with the _ same weight _ as the original band at @xmath156 in the ordinary photoemission spectrum - see the discussion in the first section . that this is
not the case shows that even at this lowest temperature , a simple sdw - like description of the band structure is invalid , in that the band structure can not be understood by simple backfolding of the spectrum obtained without broken symmetry . as we will see in the following the af sdw state provides only the ` background ' for the dressing of the photoholes with af spin - excitations which dynamically generate a total of four bands .
we return to the discrepancy between the 2-band dispersions of the hubbard - i and sdw approximations and the 4-band structure actually observed for example in the spectrum of the shadow operator of figure [ fig3 ] . + in order to generate a ` 4-band structure ' out of the two bands of the hubbard - i and sdw approximations we try as a phenomenological ansatz to mix the dispersions of the hubbard - i / sdw approximation with two dispersionless bands at energies of @xmath157 . in other words , for both the photoemission and inverse photoemission spectrum we diagonalize an ` effective ' @xmath158 hamilton matrix : @xmath159 and plot in figure [ fig5 ] the four bands obtained in this way on top of the spectral density obtained from qmc at @xmath160 and @xmath95 . for comparison
the figure also shows the original ( i.e. unhybridized ) hubbard - i bands plus the two phenomenological dispersionless bands at energies of @xmath161 .
the figure shows that the overall agreement between the qmc peak positions and the four bands generated by diagonalizing @xmath162 of equation ( [ mixeqt1 ] ) is surprisingly good , particularly so in view of the fact that for both , hubbard - i and sdw approximation , only _
unrenormalized _ parameters were used . in particular , the _ self - consistently _ determined value for the sdw gap @xmath20 of @xmath163 was used at @xmath95 .
the only ` external ' parameter in this figure is the mixing matrix element @xmath164 , which was set to a value of @xmath165 .
+ thus , we find in contrast to previous works @xcite , that the introduction of the dispersionless bands reproduces the sinlge - particle gap and the width of the quasiparticle band correctly without any renormalization of parameters .
rather , the narrowing of the quasiparticle band and the reduction of the hubbard gap as compared to the unrenormalized parameters is brought about by introduction of the dispersionless bands .
this naturally raises the question as to their physical origin . in the present paper
we restrict ourselves to a more phenomenological and ` numerics based ' approach . a complementary and more mathematical discussion is given in ref .
@xcite , where an equation of motion approach similar to hubbard s original work is pursued .
+ we consider the commutator of the creation operator for hole - like particles , @xmath166 , which annihilates a particle only on a singly occupied site , with the kinetic energy of the hubbard model and find@xcite : @xmath167 keeping only the first term in the square brackets of equation ( [ commute ] ) reproduces the hubbard - i approximation @xcite .
the second term in the square brackets describes the dressing of the created hole by a spin - excitation and is closely related to the spin-@xmath19 string operator of equation ( [ stringoperator ] ) .
the third term describes in an analogous fashion the coupling of the hole to a density fluctuations , whereas the fourth term describes the coupling to the @xmath168-excitation@xcite .
the two latter types of excitation are not important for a large positive @xmath13 near half - filling , @xmath15 , and will be neglected .
therefore , the operator @xmath169 is in this case the most important correction over the hubbard - i approximation .
as already stated , it describes a hole dressed by a spin - excitation : this operator not only creates a hole on site @xmath79 but dresses this hole with a spin - excitation on a neighboring site , which is exactly the idea behind the spin - bag@xcite or spin - polaron@xcite pictures known in the literature .
+ splitting this operator into eigenoperators of @xmath170 : @xmath171 we find @xmath172 = \frac{u}{2 } \hat{d}_{i , j,\sigma}$ ] and @xmath173 = - \frac{u}{2 } \hat{c}_{i , j,\sigma}$ ] .
assuming moreover that the mobility of these composite excitations is determined by the ` heavy ' spin - excitation , it seems quite resonable to assume that these ` particles ' are the source of the ( more or less ) dispersionless bands at @xmath174 .
+ finally , the commutation relation ( [ commute ] ) shows that the mixing matrix element between the @xmath175 and the new composite particles should be @xmath176 . based on these rough considerations
we might thus expect that the two string-@xmath53 ` effective particles ' ( [ eigy ] ) are excellent candidates for explaining the two dispersionless bands at @xmath177 required to upgrade the hubbard - i or sdw approximation so as to match the qmc data .
however , so far the above considerations are pure speculation and in the following we will turn to qmc - results to back up this hypothesis by numerical evidence .
+ before doing so , however , we want to illustrate the action of the ` string - operator ' in two extreme cases : an ideal nel state and a resonating valence - bond ( rvb ) state , i.e. a compact singlet covering of the plane@xcite ( see figure [ fig6 ] ) . in the nel state , a hole created initially on site @xmath6 can travel one place to a neighboring site @xmath79 thereby leaving behind a misaligned spin on the original site @xmath6 .
exactly this process is described by the second term , @xmath178 : it creates a hole of opposite spin on a neighboring site @xmath79 and flips the spin on the original site @xmath6 .
therefore , this process corresponds to the creation of a string of length @xmath53 .
in fact one might think about more sophisticated diagnostic operators incorporating the effects of longer - ranged strings @xcite .
indeed , dagotto and schrieffer@xcite and eder and ohta@xcite already measured the angle - resolved spectrum of a diagnostic operator containing strings with up to three lattice sites range by means of exact diagonalizations of the @xmath179 model .
as already mentioned in the introduction , in the qmc method each observable has to be expressed in terms of free single - particle greens functions on each time slice by the application of wicks theorem @xcite .
this results already in a quite large expression for the spin-@xmath19 string operator of equation ( [ stringoperator ] ) containing approximately @xmath180 contributions .
the implementation of even longer - ranged string operators therefore was not possible .
+ returning to the spin-@xmath19 string operator of equation ( [ stringoperator ] ) we note that the first term , @xmath181 will always annihilate a nel state .
this reflects the simple fact that spin - rotation symmetry is broken in the nel state .
this is not the case , however , in the fully rotationally invariant rvb state : again , creating a hole on site @xmath6 and allowing it to hop to site @xmath79 will produce a spin - excitation .
however , in the case of the rvb state , it produces the superposition of two states : in one case the dotted ellipse stands for the @xmath182 component of the triplet , and this state would again be created by the term @xmath178 .
there is , however , also a second state where the dotted ellipse corresponds to the superposition of a singlet and the @xmath183 component of the triplet .
this second state then would be created by the term @xmath181 , whereby the relative sign of the two terms in the string-1 operator makes sure that the two configurations are always produced with the proper phase . in both extreme cases , nel state and ` singlet soup ' , the string-@xmath53 operator thus creates a hole dressed with the proper spin - excitation : this scan be a spin wave ( i.e. a single inverted spin ) in the case of a nel state , or a singlet - triplet excitation in the case of an rvb state .
+ as a technical remark we still note that the excessive numerical effort which would have been necessary to compute spectra for the @xmath184 and @xmath185 ( which are products of @xmath186 fermion operators ) has made it impossible to compute the spectra of these operators - instead we have been using the ( fourier transform of ) the operator @xmath187 , defined in ( [ stringoperator ] ) . concerning the difference between this diagnostic operator and the operator @xmath188 obtained by commuting @xmath189 with the kinetic energy ( see the second term on the r.h.s . of equation ( [ commute ] ) ) , we note that their fourier transforms differ only by phase factors of the form @xmath190 .
both operators are indeed identical at momentum @xmath89 and differ at momentum @xmath91 only by a factor of @xmath191 since @xmath192 and @xmath193 are next - nearest neighbor lattice sites . at all other momenta
both operators have exactly the same peaks but will differ somewhat in their spectral weights .
+ after these remarks we discuss the angle - resolved spectral function of the spin-@xmath19 string operator @xmath194 , shown in figure [ fig7 ] .
the spectrum of the spin-@xmath19 string operator indeed picks up those band portions which , according to the rough hybridization scenario in figure [ fig5 ] , should have strong ` flat - band character ' , i.e. the part of the narrow low energy band between @xmath156 and @xmath195 and between @xmath156 and @xmath196 at energies @xmath153 . we note that these are precisely the positions where the two quasiparticle - like dispersive narrow bands occurred in the normal photoemission spectrum at @xmath102 .
in addition , the band portion at @xmath197 for momentum @xmath111 is also enhanced in the string-1 spectrum .
+ this , however , still leaves an important part of the band structure unexplained .
namely , the ` af umklapp band ' of the narrow quasiparticle band dispersing upward between @xmath156 and @xmath195 at @xmath153 still is not seen in any of the spectra , not even at the lowest temperature studied . on the other hand , in a state with true antiferromagnetically broken symmetry we know that this mirror image must exist due to the backfolding of the brillouin zone . to finally resolve this part
, we now introduce the last diagnostic operator of this work , which we call the spin-@xmath92 string operator : @xmath198 this describes again a composite object of a hole and a spin - excitation , but this time the two constituents are coupled to the total spin of @xmath92 .
we stress that this operator will detect states which can never be seen in an actual angle - resolved photoemission spectrum on a singlet ground state , because this is forbidden by the angular - momentum selection rule .
the angle - resolved spectral function @xmath199 of the spin-@xmath92 string operator is plotted in figure [ fig8 ] again for @xmath102 ( top ) and @xmath95 ( bottom ) .
it is then immediately obvious that it is this operator which resolves the ` missing piece ' of the af dispersion , i.e. the ` af mirror image ' of the narrow quasiparticle band .
it should also be noted that the spectrum of @xmath200 is remarkably independent of temperature , i.e. the states belonging to this ` spin-@xmath92 band ' persist irrespectively of whether there is long - range order or not .
+ combining the information obtained so far , suggests the following scenario for the crossover between the paramagnetic band structure at high temperature and the af band structure at low temperature : in the paramagnetic state at high temperatures ( such as @xmath102 ) , the spin is a good quantum number and the spin-@xmath92 ` band ' does exist but can not mix with any spin-@xmath19 band due to spin - conservation .
the band of spin-@xmath92 quasiparticles thus plays no role whatsoever in the actual photoemission spectrum , which is presumably the reason why the outer part of the spectrum is so remarkably ` invisible ' in actual arpes experiments , leading to the idea of a ` remnant fermi surface ' in the insulator@xcite . reaching this state by photoemission
would only be possible if the photohole is created in a thermally admixed state of at least spin @xmath53 . in an infinite system
this situation changes discontinuously at the transition to the true broken symmetry state : there the total spin ceases to be a good quantum number , and the spin-@xmath19 band in the interior of the af zone and the spin-@xmath92 band in the exterior now suddenly can mix with each other , thus leading to the familiar sdw dispersion .
we note that another way to generate a coupling between these two bands would be application of a magnetic field - this also would break spin - rotation invariance and hence enable the hybridization of a spin-@xmath19 and a spin-@xmath92 band . based on our results
we thus believe that a magnetic field could enhance the spectral weight of the ` shadow part ' of the band structure as seen in arpes .
summarizing our results so far , we may say that the hubbard - i approximation , slightly improved by the introduction of new quasiparticles corresponding to dressed holes provides a very good description of the spectral function for the case of half - filling , @xmath201 . in this section
we want to proceed to the doped case @xmath202 , which is of prime interest for cuprate superconductors . here
, an essential drawback of the qmc procedure is that reliable qmc simulations for lower temperatures are much more difficult or even impossible , since the absence of particle - hole symmetry away from half - filling introduces the notorious minus - sign problem into the algorithm .
truely low temperatures like @xmath203 , which in principle correspond to the physical temperature range , are therefore out of reach .
on the other hand in the study of the half - filled case we have seen that a major change takes place as the spin correlation length reaches the system size , whence ` effective long range order ' sets in . in the doped case the spin correlation length is expected to be short at any temperature , whence we may expect that the change of @xmath93 from high to small temperature is more smooth than at half - filling . in that sense ,
even @xmath93 data for the relatively high temperature @xmath102 are interesting to study .
moreover , we can at least try to elucidate trends with decreasing temperature and thus construct a reasonably plausible scenario .
+ at half - filling , we have seen that the ` approximation of choice ' for the paramagnetic case was the hubbard - i approximation .
this naturally poses the question as to how relevant the half - filled case is for the description of the doped case , i.e. how much of the hubbard - i physics remains valid for finite doping . at half - filling the two ` effective particles ' @xmath204 and @xmath205 ,
form the two separate hubbard bands .
the effect of doping would now consist in the chemical potential cutting progressively into the top of the lower hubbard band , in much the same fashion as in a doped band insulator . on the other hand , for finite @xmath31 the spectral weight along this band deviates from the free - particle value of @xmath53 per momentum and spin so that the fermi surface volume ( obtained from the requirement that the integrated spectral weight up to the fermi energy be equal to the total number of electrons ) is not in any ` simple ' relationship to the number of electrons - the luttinger theorem must be violated .
this is the major reason why the hubbard - i approximation has been dismissed by many authors as being unphysical .
+ we now wish to address the question as to what really happens if a paramagnetic ( i.e. not magnetically ordered ) insulator is doped away from half - filling , by qmc simulation .
we therefore choose @xmath206 and @xmath11 .
+ figure [ fig9 ] then shows the development of @xmath207 with doping .
it is quite obvious from this figure that initially the @xmath117 bands seen at half - filling in the photoemission spectrum ( i.e. @xmath208 ) persist with an essentially unchanged dispersion .
the chemical potential gradually cuts deeper and deeper into the topmost band , forming a hole - like fermi surface centered on @xmath111 , the top of the lower hubbard band .
the only deviation from a rather simple rigid - band behavior is an additional transfer of spectral weight : the part of the topmost band near @xmath111 gains in spectral weight , whereas the band with higher binding energy looses weight . in addition , there is a transfer of weight from the upper hubbard band to the inverse photoemission part below the hubbard gap .
this effect is actually quite well understood@xcite .
the band structure above the hubbard gap becomes more diffuse upon hole doping in that the rather clear two - band structure visible near @xmath111 at half - filling rapidly gives way to one broad ` hump ' of weight .
apart from the spectral weight transfer , however , the band structure on the photoemission side is almost unaffected by the hole doping - the _ dispersion _ of the quasiparticle band becomes somewhat wider but does not change appreciably . in that sense
we see at least qualitatively the behavior predicted by the hubbard i approximation .
+ next , we focus on the fermi surface volume .
some care is necessary here : first , we can not actually be sure that at the high temperature we are using there is still a well - defined fermi surface .
second , the criterion we will be using is the crossing of the quasiparticle band through the chemical potential .
it has to be kept in mind that this may be quite misleading , because band portions with tiny spectral weight are ignored in this approach ( see for example ref .
@xcite for a discussion ) .
when thinking of a fermi surface as the constant energy contour of the chemical potential , we have to keep in mind that portions with low spectral weight may be overlooked . on the other hand
the fact that a peak with appreciable weight crosses from photoemission to inverse photoemission at a certain momentum is independent of whether we call this a ` fermi surface ' in the usual sense , and should be reproduced by any theory which claims to describe the system .
it therefore has to be kept in mind that in the following we are basically studying a ` spectral weight fermi surface ' , i.e. the locus in @xmath96 space where an apparent quasiparticle band with high spectral weight crosses the chemical potential . with these _ caveats _ in mind , figures
[ fig10 ] and [ fig11 ] show the low - energy peak structure of @xmath207 for all allowed momenta of the @xmath209 cluster in the irreducible wedge of the brillouin zone , and for different hole concentrations . in all of these spectra
there is a pronounced peak , whose position shows a smooth dispersion with momentum . around @xmath111
the peak is above @xmath12 , whereas in the center of the brillouin zone it is below .
the locus in @xmath56-space where the peak crosses @xmath12 forms a closed curve around @xmath111 and it is obvious from the figure that the ` hole pocket ' around @xmath111 increases very rapidly with @xmath210 . to estimate the fermi surface volume @xmath211 we assign a weight @xmath212 of @xmath53 to momenta @xmath96 where the peak is below @xmath12 , @xmath213 if the peak is right at @xmath12 and @xmath214 if the peak is above @xmath12 .
our assignments of these weights are given in figures [ fig10 ] and [ fig11 ] .
the fractional fermi surface volume then is @xmath215 , where @xmath216 is the number of momenta in the @xmath209 cluster .
of course , the assignment of the @xmath212 involves a certain degree of arbitrariness .
it can be seen from figures [ fig10 ] and [ fig11 ] , however , that our @xmath212 would in any way tend to underestimate the fermi surface volume , so that the obtained @xmath211 data points rather have the character of a lower bound to the true @xmath211 .
even if we take into account some small variations of @xmath211 due to different assignments of the weight factors , however , the resulting @xmath211 versus @xmath210 curve never can be made consistent with the luttinger volume , see figure [ fig12 ] .
the deviation from the luttinger volume is quite pronounced at low doping .
@xmath211 approaches the luttinger volume for dopings @xmath217% , but due to our somewhat crude way of determining @xmath211 we can not really decide when precisely the luttinger theorem is obeyed .
the hubbard i approximation approaches the luttinger volume for hole concentrations of @xmath218% , i.e. the steepness of the drop of @xmath211 is not reproduced quantitatively .
the latter is somewhat improved in the so - called @xmath117-pole approximation@xcite .
for example the fermi surface given by beenen and edwards@xcite for @xmath219 obviously is very consistent with the spectrum in figure [ fig11 ] for @xmath220 .
+ we return to figure [ fig9 ] and discuss the entire width of the spectra , in particular the question of the fate of the 4-band structure in the doped system . for @xmath221
the different features that are seen at @xmath15 are still rather clearly visible , but for @xmath222 the low energy quasiparticle band at @xmath223 starts to disappear , and at @xmath224 the dominant ` band ' in the spectrum between @xmath225 and @xmath226 can be fitted by a sligtly renormalized free - electron band . as we have seen above
, the luttinger theorem also is valid in this case .
this suggests to classify the doping as ` underdoped ' for @xmath227 , where the luttinger theorem is invalid and the @xmath228-band structure known from half - filling persists , and ` overdoped ' where the luttinger theorem is valid and a renormalized free - electron band can be seen in the spectral function . following the convention for cuprate superconductors
, we call the doping where the crossover between the two regimes occurs the ` optimal ' doping .
+ next , the four plots of figure [ fig13 ] show the spectra at selected @xmath56-points .
the system size is only @xmath229 in this case because this allows for smaller error bars .
closer inspection , especially , of the peaks at momentum @xmath89 on the photoemission side ( plot ( a ) ) and at momentum @xmath91 on the inverse photoemission side ( plot ( d ) ) confirms , that with increasing hole concentration we are losing parts of the 4-band structure seen at half - filling .
+ to check the physics of the band structur in more detail , we again employ our diagnostic operators .
figure [ fig14 ] shows the angle - resolved spectral functions @xmath199 of the spin-@xmath19 and spin-@xmath92 string operators , @xmath187 and @xmath200 .
as was the case at half - filling , the spectrum of the spin-@xmath19 string operator highlights exactly those peaks that we associate with the dipsersionless ` dressed hole ' bands in figure [ fig5 ] .
the spectrum of the spin-@xmath92 string operator on the other hand has its peaks with maximal spectral weight around momentum @xmath91 , indicating that also in the doped case there is an ` antiferromagnetic mirror image ' of the quasiparticle band ( which , however , consists of spin-@xmath92 states ) .
again , coupling of photoholes to thermally excited spin excitations may make these states visible in arpes spectra , thus explaining the ` shadow bands ' seen in photoemission experiments by aebi _
et al._@xcite .
similarly as for half - filling one might speculate that a magnetic field , which would break spin symmetry and thus allow for a coupling of ` bands ' with different total spin , would enhance the spectral weight of these shadow bands .
+ all in all we have ssen that the ` band structure ' ( @xmath228-band structure , dispersion of regions of large spectral weight , ` character ' of the bands as measured by the diagnostic operators ) stays pretty much unchanged as long as we are in the underdoped regime . at half - filling the @xmath228-band structure
is closely related to the sharp low - energy mode in the dynamical spin correlation function , which naturally suggests to study the spin reponse also as a function of doping .
figure [ fig15 ] shows the spin - correlation function , @xmath230 ( left column ) , and the charge - correlation function , @xmath119 ( right column ) , for @xmath102 and densities @xmath221 ( underdoped ) , @xmath231 ( nearly optimally doped ) and @xmath224 ( overdoped ) .
the spin - response is sharply confined in both momentum @xmath91 and energy @xmath232 only in the underdoped region i.e. the regime where we also observe the features associated with spin excitations in the single - particle spectra .
as was the case at half - filling for temperatures below @xmath233 , the spin - response can be fitted by the af spin - wave dispersion ( [ spinwave ] ) in the underdoped regime .
on the other hand , as soon as the system enters the overdoped regime the spin - response is no longer sharply peaked at momentum @xmath91 and energy @xmath232 : it broadens in momentum and spreads in energy by an order of magnitude with the scale changing from @xmath16 to @xmath234 accompanied by a similar change in the bandwidth of the single particle excitations .
this result is already well known from previous qmc calculations @xcite and consistent with similar behaviour in the t - j model@xcite .
the charge - response , @xmath119 , is always broad in both momentum @xmath96 and energy @xmath144 for all densities studied .
it merely decreases its width from @xmath235 at @xmath221 to @xmath236 at @xmath224 .
+ although the minus - sign problem of the qmc algorithm prevents reliable simulations of large systems at low temperatures in the doped regime , we nevertheless studied the temperature evolution of the angle - resolved spectral function @xmath93 at density @xmath237 .
this was possible due to the relative small system size of @xmath229 , which alleviates the minus - sign problem as compared to @xmath132 at the @xmath138 .
figure [ fig16 ] shows the results from this analysis : the uppermost plot ( a ) compares the angle - resolved spectral functions @xmath93 at density @xmath238 for @xmath239 , @xmath102 and @xmath240 .
we stress that the simulation at @xmath240 suffers from minus - sign problems with a drastically reduced resolution . in the center plot ( b ) , the quasiparticle peak weights around momentum @xmath91 of the @xmath238 simulation are compared with the quasiparticle peak weight at momentum @xmath91 of a half - filled , @xmath241 , simulation for different temperatures . at half - filling , the hubbard - i - like quasiparticle peak @xmath91 and @xmath149 decreases in spectral weight with decreasing @xmath242 and disappears as the spin - correlation length ( which increases with decreasing temperature ) reaches the lattice size ( at @xmath140 ) . in the underdoped case for density
@xmath238 the weights of the corresponding peaks around momentum @xmath91 located also decrease with decreasing @xmath242 .
closer inspection of this peak ( see the inset of the center plot ( b ) ) reveals , that this peak even raises slightly in binding energy with decreasing temperature , very similar to the peak in the half - filled case . in a real photoemission experiment
this peak would have dropped below the typical resolution of roughly @xmath243 spectral weight @xcite at a temperature of @xmath244 .
the spin - correlation length ( again derived by a fit of the equal - imaginary - times spin - correlation function to a form @xmath245 ) also shows similar behavior in the underdoped and half - filled cases : the values for the spin - correlation length are @xmath246 , @xmath247 and @xmath134 in the underdoped case and @xmath248 , @xmath134 and @xmath249 at half - filling for @xmath239 , @xmath102 and @xmath240 , respectively .
the spin - susceptibility ( shown in plot ( c ) of figure [ fig16 ] ) also behaves very similar , but with changed magnitudes in the underdoped and in the half - filled cases .
these data suggest a similar temperature evolution of the band - structure of the hubbard model in the underdoped and half - filled cases driven by the temperature dependent spin - correlation length @xmath130 .
especially , we expect the hubbard - i - like quasiparticle peaks at energies of @xmath250 around momentum @xmath91 to vanish with decreasing @xmath242 in the underdoped case as the peaks at energies of @xmath149 around momentum @xmath91 do in the case of half - filling .
+ the latter observation suggests a profound change of the fermi surface with temperature : as seen above , it is precisely the hubbard - i - like band near @xmath111 which crosses the chemical potential and thus forms the fermi surface in the doped case .
it is then quite clear that the ` disappearance ' of this band with decreasing temperature must affect the fermi surface in some dramatic way .
studies at zero temperature are possible only by means of exact diagonalization .
analysis of the single particle spectrum shows the same ` rigid - band ' behaviour as at high temperatures@xcite and analysis of the momentum distribution @xmath251 suggest@xcite that the doped holes accumulate at the surface of the magnetic zone ( i.e. the line @xmath252 ) rather than around @xmath111 .
in the present work we have systemactically studied the temperature- and doping - dependent dynamics of the two - dimensional hubbard model by finite - temperature qmc simulations . comparing the qmc single particle spectral function , the dynamical spin response and the spectral functions of suitably chosen diagnostic operators , different physical regimes could be identified . in simplest terms
there are two quantities , which basically determine the single particle spectrum : the hole concentration and the the spin - response function , whereby there is a certain relationship between the two .
+ at half - filling and high temeratures ( @xmath253 ) , the combined photoemission and and inverse photoemission spectrum @xmath93 displays two dispersive features , the upper and lower hubbard band , roughly separated by @xmath13 ( @xmath254 , in our work ) . at these very high temperatures
the system is in a spin disordered state .
we have demonstrated here that the well - known hubbard - i approximation gives an excellent description of the single particle spectrum in this state , reproducing quantitatively both the single - particle dispersion and the distribution of spectral weight .
this is by no means trivial , since the hubbard - i approximation is dynamically equivalent to a simplified effective hamiltonian , which just contains hole - like ( @xmath255 ) and double occupancy - like particles in a simple bi - quadratic form .
+ at lower temperatures ( @xmath256 ) , the hubbard - i approximation needs to be improved ; this is to be expected , because it neglects all effects of spin - correlations .
in fact , the temperature where deviations from hubbard - i become strong , coincides fairly well with the transition from a spin response @xmath118 which is diffuse both in momentum and energy ( with a spread of order @xmath8 ) , to a more ` spin - wave - like ' response . in this regime @xmath118
displays the characteristic energy scale @xmath257 , with its spectral weight being concentrated at the af wave vector @xmath60 .
it should be noted that this ` spin - wave - like ' regime develops despite the fact that at @xmath102 the spin - correlations length @xmath258 is still short - ranged ( @xmath259 lattice spacings ) . only at
the lower temperature @xmath95 nel order spreads over the entire qmc block , creating an effective ( finite - size ) nel state .
+ it is well established by previous , in particular also qmc work , that in this temperature - regime ( @xmath256 ) new spectral features appear .
they have often been interpreted as four ` bands ' , two ` coherent ' bands forming the topmost valence and the lowest conduction band in the insulator plus two ` incoherent ' bands , i.e. the remaining upper and lower hubbard band features ( see for example @xcite ) .
our present work not only definitively identifies these four bands but also clarifies their physical origin and their connection to the spin - excitations . in simplest terms
the emerging spin waves at lower temperatures provide the excitations that can ` dress ' the hubbard quasiparticles , whence new bands corresponding to dressed holes / double occupancies appear in the single particle spectrum @xmath93 .
it has been shown in ref .
@xcite that the @xmath228-band structure which appears in @xmath93 at lower temperatures can be explained in this way , and our present numerical check by directly calculating the spectra of ` dressed electrons ' supports this interpretation .
+ this physical picture at half - filling can be extended into the underdoped regime .
this is most obvious in the single particle spectral function , which stays almost unchanged in the doped case ( i.e. the @xmath228-band structure and the ` character ' of the bands as measured by the diagnostic operators ) .
the main change in fact consists in the chemical potential cutting gradually into the ( top of the ) lower hubbard band , precisely as predicted by the hubbard - i approximation .
contrary to widespread belief the ` fermi surface ' , if determined by the fermi surface crossings of the dominant band through the chemical potential , does not satisfy the luttinger theorem .
rather , for small hole concentrations the fermi surface volume is considerable larger than that for a slightly less than half - filled free - electron band .
very similar conclusions have in fact been reached by a calculation of the electron momentum distribution in the 2d t - j model by puttika _
et al._@xcite .
their calculation actually was a high temperature series expansion plus a pad extrapolation to lower temperature , and it is encouraging that this method gives similar results as our qmc results which are performed at relatively high temperatures . in its range of applicability ,
i.e. in the absence of strong magnetic correlations and close to half - filling , the hubbard - i approximation thus works remarkably well , both at half - filling and in the doped case .
we stress that this has profound implications for the theoretical treatment of the model : perturbation expansions in @xmath13 or partial and self - consistent resummations thereof , may not be expected to give any meaningful results in this strong - coupling / low doping regime .
+ an interesting question is the possibility to verify our results experimentally .
as already mentioned above , a scan of the temperature development of @xmath207 shows that the part of the quasiparticle band near @xmath111 ( where the fermi surface is located ) is loosing weight with decreasing temperature .
in fact in arpes experiments on underdoped cuprate superconductors the ` hole - pocket ' around @xmath111 seen in our simulations ( and the expansion of puttika _ et al . _ ) is not observed , but rather a small ` fermi arc ' near @xmath195 , terminated by the ` pseudo gap ' around @xmath196 although our simulations do not allow to make statements about the truely low temperatures in the experiments , we believe that this suggests a strong temperature dependence of the single particle spectrum , with the temperature scale being set by the exchange constant @xmath121 ( which controls the degree of spin disorder ) .
the latter is rather large in copper oxides , so that the temperature regime studied in our simulations probably can not be accessed experimentally in these materials .
we note , however , that an arpes study for the 1d material @xmath260 which has a smaller exchange constant , has indeed provided evidence for a strong @xmath242-dependence of @xmath207@xcite .
clearly , it would be interesting to study the fermi surface evolution in a 2d material with lower exchange constant .
+ as was the case at half - filling , the dynamical spin - response plays an important role : throughout the hubbard - i phase at low doping , the spin response shows the sharp low - energy mode at @xmath111 .
the simultaneous disappearance of the @xmath228-band structure in @xmath207 and the low energy spin response with scale @xmath121 in the overdoped regime then show again the close relationship between the two .
the dressing of holes by spin excitations apparently remains the most important correction over hubbard - i . in the overdoped regime the spin response is spread out over an energy range of @xmath261 and thus becomes more similar to the charge response .
the single - particle spectral function is most consistent with a slightly renormalized free electron dispersion , and the luttinger theorem appears to be satisfied even at the relatively high temperature @xmath206 .
this is quite consistent with earlier results on the t - j model , which show that for hole concentrations @xmath262 % the spin and charge response can be approximated well by the self - convolution of the single particle spectral function@xcite .
this is essentially what is to be expected for a system of weakly interacting fermions , so that we conclude that in the overdoped regime we enter a new phase which most probably extends to the low - concentration limit where the nagaoka @xmath242-matrix approximation becomes exact .
finally we note that exact diagonalization studies at finite temperatures@xcite also show some evidence for a ` crossover ' between different physical regimes at a hole concentration of @xmath263% . +
this work was supported by dfn contract no .
tk 598-va / d03 and by bmbf ( 05sb8wwa1 ) .
computations were performed at hlrs stuttgart and hlrz jlich .
one of us ( w.h . ) acknowledges hospitality of the physics department in santa barbara and many useful discussions with d. j. scalapino .
e. loh and j. gubernatis , _ electronic phase transitions _
, w. hanke and y. v. kopaev , north holland , amsterdam ( 1992 ) .
j. e. hirsch , phys .
b * 31 * , 4403 ( 1985 ) .
m. jarrel , j. e. gubernatis , phys . rep . * 269 * , 133 ( 1996 ) .
e. fradkin , _ field theories of condensed matter systems _ , frontiers in physics , addison - wesley publishing company ( 1991 ) . n. d. mermin and h. wagner , phys . rev
. lett .
* 17 * , 1133 ( 1966 ) .
a. b. harris and r. v. lang , phys .
rev . * 157 * , 295 ( 1967 ) .
j. r. schrieffer , x .-
wen , and s .- c .
zhang , phys .
lett . * 60 * , 944 ( 1988 ) .
j. r. schrieffer , x .-
wen , and s .- c .
zhang , phys .
b * 39 * , 11663 ( 1989 ) .
l. n. bulaevskii , e. l. nagaev , and d. i. khomskii , sov .
jetp * 27 * , 638 ( 1967 ) ; s. a. trugman , phys .
b * 37 * , 1597 ( 1988 ) ; ibid .
* 41 * , 892 ( 1990 ) ; r. eder and k. w. becker , z. phys .
b * 78 * , 219 ( 1990 ) ; o. p. sushkov , solid state commun . * 83 * , 303 ( 1992 ) ; a. ramsak and p. horsch , phys . rev .
b * 48 * , 10559 ( 1993 ) ; g. reiter , phys .
b * 49 * , 1536 ( 1994 ) ; r. hayn , a. f. barabanov , j. schulenburg , and j. richter , phys . rev .
b * 53 * , 11714 ( 1996 ) .
j. hubbard , proc .
soc . a * 276 * , 238 ( 1963 ) .
a. dorneich , m. g. zacher , c. grber and r. eder , e - print cond - mat/ r. preuss , w. hanke , and w. v. d. linden , phys .
lett . * 75 * , 1344 ( 1995 ) .
b. mehlig , h. eskes , r. hayn , and m. b. j. meinders , phys .
b * 52 * , 2463 ( 1995 ) .
e. dagotto , f. ortolani , and d. scalapino , phys .
b * 46 * , 3183 ( 1992 ) . c. n. yang , phys . rev .
lett . * 63 * , 2144 ( 1989 ) ; s. c. zhang , phys .
lett . * 65 * , 120 ( 1990 ) .
e. dagotto and j. r. schrieffer , phys .
b * 43 * , 8705 ( 1991 ) .
r. eder and y. ohta , phys .
b * 50 * , 10043 ( 1994 ) .
a. l. fetter and j. d. walecka , _ quantum theory of many - particle systems _ , mcgraw - hill , new york ( 1971 ) .
f. ronning , c. kim , d. l. feng , d. s. marshall , a. g. loeser , l. l. miller , j. n. eckstein , i. bozovic , z .- x .
shen , science * 282 * , 2067 ( 1998 ) .
h. eskes and a. m. oles , phys .
73 * 1279 ( 1994 ) .
r. eder and y. ohta , phys .
* 72 * , 2816 ( 1994 ) .
j. beenen and d. m. edwards , phys .
b * 52 * , 13636 ( 1995 ) .
p. aebi , j. osterwalder , p. schwaller , l. schlapbach , m. shimoda , t. mochiku , and k. kadowaki , phys . rev . lett .
* 72 * , 2757 ( 1994 )
. l. m. roth , phys . rev . * 184 * , 451 ( 1969 ) .
g. geipel and w. nolting , phys .
b * 38 * , 2608 ( 1988 ) ; w. nolting and w. borgiel , phys . rev .
b * 39 * , 6962 ( 1989 ) .
r. preuss , w. hanke , c. grber , and h. g. evertz , phys .
lett . * 79 * , 1122 ( 1997 )
. s. haas , a. moreo , and d. dagotto , phys .
74 * , 4281 ( 1995 ) .
j. jaklic and p. prelovsek , phys .
74 * , 3411 ( 1995 ) .
r. eder , y. ohta , and t. shimozato , phys .
b * 50 * , 3350 ( 1994 ) .
s. nishimoto , y. ohta , and r. eder , phs .
b * 57 * , 5590 ( 1998 )
. w. o. putikka , m. u. luchini , and r. r. p. singh , phys .
lett . * 81 * , 2966 ( 1998 ) .
k. kobayashi , t. mizokawa , a. fujimori , m. isobe , y. ueda , t. tohyama , s. maekawa , phys .
* 82 * , 803 ( 1999 ) .
r. eder and y. ohta , phys .
b * 51 * , 11683 ( 1994 ) .
j. jaklic and p. prelovsek , phys .
lett . * 77 * , 892 ( 1996 ) . | we present results of a systematic quantum - monte - carlo study for the single - band hubbard model .
thereby we evaluated single - particle spectra ( pes & ipes ) , two - particle spectra ( spin & density correlation functions ) , and the dynamical correlation function of suitably defined diagnostic operators , all as a function of temperature and hole doping .
the results allow to identify different physical regimes .
near half - filling we find an anomalous ` hubbard - i phase ' , where the band structure is , up to some minor modifications , consistent with the hubbard - i predictions . at lower temperatures , where the spin response becomes sharp , additional dispersionless ` bands ' emerge due to the dressing of electrons / holes with spin excitatons .
we present a simple phenomenological fit which reproduces the band structure of the insulator quantitatively .
the fermi surface volume in the low doping phase , as derived from the single - particle spectral function , is not consistent with the luttinger theorem , but qualitatively in agreement with the predictions of the hubbard - i approximation .
the anomalous phase extends up to a hole concentration of @xmath0% , i.e. the underdoped region in the phase diagram of high - t@xmath1 superconductors .
we also investigate the nature of the magnetic ordering transition in the single particle spectra .
we show that the transition to an sdw - like band structure is not accomplished by the formation of any resolvable ` precursor bands ' , but rather by a ( spectroscopically invisible ) band of spin 3/2 quasiparticles .
we discuss implications for the ` remnant fermi surface ' in insulating cuprate compounds and the shadow bands in the doped materials . 2 |
over the years it has been found that there exist many two - dimensional classical spin models , discrete and continuous alike , whose ground - state manifolds are macroscopically degenerate and , more interestingly , also exhibit critical behaviours , i.e. , spin - spin correlation functions within the ground - state ensembles decay with distance as power laws .
the classification of universality class for these models has always been a challenging problem@xcite an earlier example of this kind is the antiferromagnetic ising model on the triangular lattice .
the exact solution for this model by stephenson@xcite showed that although this model remains paramagnetic at nonzero temperature , its ground state is critical .
later works by blte _
et al _ revealed yet another remarkable property of the ground - state ensemble of this model , namely , it permits a solid - on - solid ( sos ) representation in which spin fluctuations are subsequently described by the fluctuating interface in the sos model@xcite .
recent studies also demonstrated that this interfacial representation provides a valuable avenue for studying the ground - state ordering of quantum magnets@xcite and the ground - state roughness of oriented elastic manifolds in random media@xcite .
other recently studied models with critical ground states include three - state antiferromatic potts model on the kagom lattice@xcite , the @xmath3 model on the honeycomb lattice@xcite , the four - coloring model on the square lattice@xcite , and the square - lattice non - crossing dimer model and dimer - loop model@xcite . on the other hand , some very similar models with degenerate ground states exhibit long - range order , such as the constrained 4-state potts antiferromagnet@xcite .
in this article we study the ground - state properties of antiferromagnetic ising model of general spin on a triangular lattice which also belongs to the class of models mentioned above .
recent numerical studies of this model include monte carlo simulations@xcite and transfer matrix calculations@xcite . here
we revisit this model by performing monte carlo simulations .
the motivation of the present work is two - fold : ( 1)unlike previous simulations , we utilize the interfacial representation directly in analyzing the simulation results , for example , we compute the stiffness constant of the fluctuating interface which , in turn , yields more accurate critical exponents of various operators ; and ( 2 ) we also study the dynamical properties of this model for the first time making use of the interfacial representation .
the body of the this paper is organized as follows .
section [ model ] describes the model hamiltonian and maps it onto a spin-1 problem whose interfacial representation is then described . in section [ height - rep ] ,
we propose an effective continuum theory for the long - wavelength fluctuations of the interface . here
we also show how to relate scaling dimensions of various operators to the stiffness constant of the interface , and derive some other analytical results based on this `` height representation . ''
this allows analytical understanding of the phase diagram ( sec .
[ phasediag ] ) .
details of monte carlo simulations and numerical results on both dynamical and static properties are presented in section [ mc - results ] , including a comparison of the new and old approaches to determining the exponents . as a conclusion ,
the paper is summarized and various possible extensions are outlined , in section [ conc - disc ] .
the antiferromagnetic ising model of spin-@xmath0 on a triangular lattice can be described by the following hamiltonian : @xmath4 where the spin variable @xmath5 defined on lattice site @xmath6 of the triangular lattice can take any value from a discrete set @xmath7 $ ] , and the sum over @xmath8 runs over three nearest - neighbor vectors @xmath9 , @xmath10 and @xmath11 as shown in fig .
[ fig1 ] . here
the coupling constant @xmath12 is positive describing the antiferromagnetic exchange interaction between two nearest - neighbor spins : @xmath5 and @xmath13 .
one important reason for interest in this model is that the @xmath14 limit@xcite is the same as the ising limit of the ( classical or quantum ) heisenberg antiferromagnet on the triangular lattice with ising - like anisotropic exchange .
that model was shown to exhibit a continuous classical ground state degeneracy and unusual features of the selection by fluctuations of ground states@xcite .
the ground - state configurations of the above model given by eq .
( [ eq1 ] ) consist of entirely of triangles on which one spin is @xmath15 , another is @xmath16 , and the third can be anything in @xmath17 $ ] .
thus , if spin @xmath18 takes an intermediate value @xmath19 , this forces the six surrounding spins to alternate @xmath15 and @xmath16 ; exactly which intermediate value @xmath18 takes does not matter in determining whether a configuration is allowed .
therefore , this allows us to reduce each state @xmath20 to a state @xmath21 of a _
spin-1 _ model , by mapping @xmath22 to @xmath23 , @xmath24 to @xmath25 , and intermediate values @xmath26 to @xmath27 . in this _
spin-1 _ representation of the model , the rules for allowed configurations are exactly the same as for the @xmath28 model ; however instead of being equal , the statistical weights have a factor @xmath29 for each spin with @xmath27 .
it should be noted that in the @xmath30 case , @xmath31 simply maps to @xmath32 .
it can also be shown that the expectation of any polynomial in @xmath33 , in the ground - state ensemble of the spin-@xmath0 model , can be written in terms of a similar expectation in the spin-1 model .
specifically , one must simply replace @xmath34 , & m even \cr}\ ] ] where ( e.g. ) @xmath35 .
thus there is no loss of information in this mapping . indeed ,
in some sense , the extra freedom to have @xmath5 vary from @xmath36 to @xmath37 is trivial : once given that @xmath5 and @xmath38 are intermediate spin values , there is no correlation between these values .
so we henceforth restrict ourselves to the spin-1 mapped model whose partition function for its ground - state ensemble can be written as : @xmath39 where @xmath40 denotes the number of zero spins in a ground - state configuration @xmath41 . by varying the weight factor continuously in the spin-1 model , it would possible to give a precise meaning to _ any _ real value of @xmath0 , and to simulate such an ensemble . however ,
in this article we perform monte carlo simulations for an ensemble in which @xmath42 takes only integer values .
the spin-1 representation could be further reduced to a spin-1/2 representation @xmath43 as described in refs . .
they let @xmath44 here @xmath45 if @xmath32 and if @xmath27 , @xmath46 or @xmath47 according to whether the surrounding spins are @xmath48 or the reverse .
note this mapping is not invertible .
the spin-@xmath49 representation is less satisfactory in that is arbitrarily breaks the up - down symmetry of correlation functions , but it was desirable for the transfer - matrix calculations of lipowski _
et al_@xcite since it reduced the number of degrees of freedom .
we define a _ microscopic
_ , discrete - valued height function @xmath50 living on the vertex of the triangular lattice such that the step in @xmath50 between adjacent vertices is a function of the adjacent spins : @xmath51 where @xmath52 is the spin-1 operator and @xmath53 can be any of the three nearest - neighbor vectors @xmath54 .
it is easy to show that the total change in height function , when traversed along any smallest loop , i.e , an elementary triangle , is zero .
therefore , @xmath50 is well - defined everywhere for the ground - state configurations , but it is not well - defined in any excited state .
this prescription generalizes that originally introduced by blte et al for the case @xmath30@xcite ( the prescriptions agree in that case ) .
this type of height mapping differs from other sorts of mapping ( e.g. dualities ) in a crucial way : since the spin microstates of the spin-1 model are mapped essentially one - to - one to the height microstates , it is possible to perform monte carlo simulations and construct configurations @xmath50 after each time step .
we have found that analysis of the @xmath50 correlations is much more efficient for extracting critical exponents than analysis of the spin correlations directly as was done in previous monte carlo simulations@xcite .
in this section we propose an effective continuum theory which describes the long - wavelength fluctuations of the interface .
we also demonstrate how the critical exponents of various operators are determined by the stiffness constant of the interface . to describe the interface in the rough phase , we must define a smooth height field @xmath55 by coarse - graining the discrete field @xmath50 . as a first stage , on every triangular plaquette formed by sites @xmath56 , define a new discrete height @xmath57 where @xmath58 is the center of a triangle .
the possible values of the @xmath59 are @xmath60 , for any integer @xmath61 .
( for the case @xmath30 , the only possible values are integers . )
to each of these values corresponds a _
unique _ ground - state spin configuration of the spin-1 model on that triangle , i.e. , @xmath62 where @xmath63 is any vector from a site to the center of any adjoining triangle .
the mapping is many - to - one : the function @xmath64 has period 6 .
notice that the r.h.s . of eq.([eq5 ] ) turns out to be independent of @xmath65 , but the periodic dependence on @xmath66 is phase - shifted by a function @xmath67 which takes different values on each of the three @xmath68 sublattices .
essentially , we have mapped the @xmath69 ensemble of the spin-1 problem into an equivalent interface problem .
note that , given a configuration of @xmath70 , each @xmath52 is specified ( via eq .
( [ eq5 ] ) ) , once for each adjoining triangle .
the requirement that these six values of @xmath52 coincide translates to a somewhat complicated set of contraints between pairs @xmath59 and @xmath71 on adjoining triangles ; the difference @xmath72 may be 0 , @xmath73 , or @xmath74 , but some of these are disallowed ( depending on which @xmath75 values are integer or half - odd - integer , and on the orientation of @xmath76 ) .
the weight of each configuration is given , as in ( [ eq2 ] ) , by by @xmath77 .
[ fig1 ] shows the @xmath59 mapping explicitly where the spins @xmath52 take values from @xmath78 .
the twelve states are arranged in a circle because the pattern repeats when @xmath79 .
there are certain special `` flat states '' in which @xmath59 is uniform on all triangles .
each of these is periodic with a @xmath80 unit cell in effect it is a repeat of one of the triangles in fig .
we shall name these states by writing the spins on the three sublattices,``@xmath81 '' and `` @xmath82 '' ; here `` @xmath83 '' stands for @xmath84 .
it should be noted that there are two non - equivalent species of flat state corresponding to integer , and half - integer valued @xmath59 respectively .
they are non - equivalent in the sense that they are not related by _ lattice
_ symmetries .
one of the species that is favored by the _ locking potential _ ( see eq .
( [ eq - vlock ] ) below ) is what is previously called `` ideal '' states@xcite .
thus we can imagine that all states can be described as domains of uniform @xmath59 separated by domain walls .
finally , by coarse - graining @xmath59 over distances large compared to the lattice constant , one obtains @xmath55 which enters the conjectured continuum formula for the free energy , which is entropic in origin@xcite , @xmath85 where @xmath86 is the stiffness constant of the fluctuating interface .
a lattice shift by one lattice constant leaves the free energy invariant , but induces global shifts in height space @xmath87 ; hence the potential @xmath88 in ( [ eq6 ] ) must have period one .
it is typically approximated as @xmath89 such a periodic potential , usually referred as the _ locking term_@xcite , favors the heights to take their discrete values one of the two types of flat state , depending on the sign of @xmath90 . for large @xmath0
we expect @xmath91 , favoring the @xmath82 states , in view of the large entropy of flippable spins ; it is not so sure which state is favored at smaller @xmath0 , but this does not matter for the critical exponents ( see sec . [ scaling ] and [ operators ] , below . in the _ rough phase _ , by definition , the locking term is irrelevent , and so the long - wavelength fluctations of height variable @xmath55 are governed by the gaussian term of eq .
( [ eq6 ] ) : @xmath92 where we have performed the fourier transform .
hence by equipartition theorem , @xmath93 similarly , we can also measure the _ height - height difference function _ in the real space as : @xmath94 ^ 2\rangle \nonumber \\ & = & \frac{1}{2\pi k } \ln(\pi r / a ) + ... \;\ ; ( r \gg 1 ) \;\ ; , \label{eq9 } \end{aligned}\ ] ] where @xmath95 is the lattice spacing cutoff .
( [ eq9 ] ) , we can compute the scaling dimension @xmath96 of any _ local _ operator @xmath97 , which is defined as in the correlation function , @xmath98 by local operator , we mean that @xmath97 is a local function of spin operators in the vicinity of @xmath99 .
now , the same spin configuration is recovered when the height variable @xmath59 is increased by 6.@xcite thus any local operator @xmath97 is also a periodic function in the height space , and can consequently be expanded as a fourier series : @xmath100 where @xmath101 runs over height - space reciprocal - lattice vectors ( i.e. multiples of @xmath102 ) .
the last step of simplification in ( [ eq11 ] ) follows because the scaling dimension @xmath96 of the operator @xmath97 is determined by the leading relevant operator in the above expansion , i.e. , @xmath103 is the smallest @xmath101 with nonzero coefficient in the sum . inserting eq .
( [ eq11 ] ) into eq .
( [ eq10 ] ) and making use of eq .
( [ eq9 ] ) , we obtain the following : @xmath104 therefore , the critical exponent @xmath105 associated with the operator @xmath97 is given by : @xmath106 in this paper , besides the usual spin operator @xmath52 , we also study the bond - energy operator @xmath107 for the reason that will become clear in the next section : @xmath108 where @xmath8 denotes one of the three nearest - neighbor vectors as before . as discussed
already , the spin operator on a given site has a periodicity of @xmath109 in the height space , from which a simple inspection shows that the bond - energy operator is also periodic in the height space with a periodicity of @xmath110 .
therefore , the reciprocal lattice vectors of the most relevant operator in the fourier expansion in eq .
( [ eq11 ] ) are @xmath111 for spin and bond - energy operators respectively . if a magnetic field is implemented by adding a term @xmath112 to the hamiltonian , then our dimensionless uniform `` magnetic field '' is defined by @xmath113 .
the exponents associated with @xmath114 ( and with the uniform magnetic susceptibility ) , are easily related to the correlation exponents of the uniform magnetization operator , @xmath115 where @xmath58 is the center of a triangle formed by sites @xmath56 .
a simple inspection of fig . [ fig1 ] shows that such an operator has a periodicity of @xmath116 in the height space , thus yielding : @xmath117 observe that the microscopic height variable @xmath50 in any flat state is not uniform but is rapidly modulated with the wave vector @xmath118 .
the amplitude of modulation itself is a periodic function of the _ coarse - grained _ height field @xmath55 which in turn implies that the correlation function decays with distance as a power - law , and consequently that its structure factor has a power - law singularity at @xmath119 .
such a zone - corner singularity is also directly connected to the singularity in the structure factor of the bond - energy operator . to see this , recall that there is a linear relation between the microscopic height variables and the bond - energy operator given by eqs .
( [ eq3 ] ) and ( [ eq14 ] ) , i.e. , @xmath120 then it is interesting to note that the fourier transform @xmath121 of bond - energy operator given above turns out to be @xmath122 in other words , as a byproduct of measuring @xmath123 , we have at the same time measured the structure factor of , say , the bond - energy operator of the same orientation specified by the nearest - neighbor vector @xmath8 : @xmath124 we will utilize this relation in sec . [ structure ] to extract the exponent of bond - energy operator from the monte carlo simulations .
the @xmath30 triangular ising antiferromagnet is exactly soluble , by the same techniques which solve the ferromagnetic two - dimensional ising model , and was immediately recognized to have critical behavior as @xmath125 .
the spin and energy correlation functions were computed exactly by stephenson ; it transpires that @xmath126 and @xmath127 exactly , implying through the arguments of blte et al ( see sec . [ scaling ] and [ operators ] ) that the effective stiffness in eq .
( [ eq7 ] ) is @xmath128 exactly .
the exponents implied by the interface scenario@xcite in particular , the magnetic field exponent @xmath129 are fully confirmed by numerical transfer - matrix computations.@xcite the coulomb gas picture of kondev _
et al_@xcite , wherein the @xmath30 triangular ising antiferromagnet is viewed as a fully - packed loop model@xcite with fugacity 1 , also predicts the exact exponents .
in this section , we collect some consequences of the height representation for the phase diagram and the nature of the various phases within it.@xcite the locking potential @xmath88 in ( [ eq6 ] ) favors the flat states . in view of ( [ eq - vlock ] ) ,
its leading reciprocal - lattice vector is @xmath130 , corresponding to a scaling index @xmath131 for the corresponding conjugate field @xmath90 .
it is well known that if @xmath132 , then @xmath90 becomes relevant ( under renormalization and the interface locks into one of the flat states.@xcite since @xmath86 grows monotonically with @xmath0 , such a locking transition occurs at a critical @xmath133 where @xmath134@xcite . in this `` smooth '' phase ,
any spin operator @xmath97 has long - range order , by arguments as in sec .
[ scaling ] .
one of our aims in this paper was to pinpoint the locking transition @xmath133 , which demands that we have a criterion to distinguish these phases .
we must supplement eq .
( [ eq8 ] ) , which shows the expected qualitative behavior of height fluctuations @xmath135 in the rough phase , with a parallel understanding of the smooth phase . in the smooth state , the symmetry ( of height shifts )
is broken and a fully equilibrated system has long - range order , such that @xmath136 is well defined and uniform throughout the system .
fluctuations around this height , then , have at most short - range ( exponentially decaying ) correlations . thus we expect them to have a spatial `` white noise '' spectrum : @xmath137 for small @xmath138 .
a phase with `` hidden '' order was suggested by lipowski and horiguchi@xcite .
numerical transfer - matrix calculations@xcite using the spin-1/2 representation indicated @xmath139 for @xmath140 , which is impossible if the spin correlations are derived from height fluctuations,@xcite as we reviewed in sec .
[ height - rep ] .
an exotic phase to reconcile these facts was to postulate a phase in which the interface was smooth and @xmath141 , yet for the real spins @xmath142 as suggested by spin correlation measurements . what does this imply for our height variable @xmath59 , which has a one - to - one correspondence with the real spin configuration @xmath143 ?
if the interface is smooth , then the probability distribution of height values on a given plaquette , @xmath144 , is well defined . in order to `` hide '' the order
, it is necessary that @xmath145 correspond to zero expectations of the spins .
now , reversing @xmath5 on all three sites in the plaquette requires @xmath146 , as seen from fig .
one can convince oneself that , to have ensemble average @xmath147 , the distribution @xmath145 must be at least as broad as @xmath148 , with @xmath149 , implying the bound @xmath150 \equiv
\langle h({\bf r})^2 \rangle - \langle h({\bf r } ) \rangle ^2 \ge ( 3/2)^2 .
\label{eq - dhbound}\ ] ] at @xmath151 , plaquettes with non - minimal energy are present and they correspond to vortices in the function @xmath55 .
thus , unfortunately , the height approach of analyzing simulations more or less breaks down .
nevertheless , one can still predict the @xmath151 phase diagram from knowledge of the @xmath69 stiffness constant derived from our simulations .
the shape of this phase diagram has already been explained in ref .
@xcite ; here we note some additional interesting behaviors which can be predicted ( following ref .
( b ) ) using the exponents associated with vortices .
the other exponents in kosterlitz - thouless ( kt ) theory are associated with elementary defects ( often called vortices ) .
indeed , it is easy to check ( in this system ) that the excess energy of a non - ground - state plaquette is directly proportional to its vortex charge ( a burgers vector in height - space ) , so the effect of nonzero temperature is simply to make the vortex fugacity nonzero .
the vortex exponent is @xmath152 , so as usual the vortex fugacity becomes relevant and defects unbind , destroying the critical state , at the kt transition defined by a spin exponent taking the critical value @xmath153 .
if @xmath154 at zero temperature , i.e. @xmath155 , then defects unbind as soon as @xmath151 .
thus a zero - temperature kt transition occurs at @xmath156 defined by @xmath157.@xcite ref .
did not , however , address the critical exponents of the correlation length @xmath158 and the specific heat @xmath159 as a function of temperature , which are also controlled by vortex exponents .
naively , if the energy cost creating one vortex is @xmath160 , and if the minimum excitation is a vortex pair , then one would expect the low - temperature specific heat to behave as @xmath161 and at @xmath30 this is indeed true@xcite .
however , the renormalization group@xcite shows the singular specific heat behaves as @xmath162 where @xmath163 is the vortex fugacity ; consequently when @xmath164 , the true behavior is @xmath165 with @xmath166 .
( physically , part of the excitation energy is cancelled by the large entropy due to the many places where the vortex pair could be placed . )
this behavior has been observed in the 3-state potts antiferromagnet on the kagom lattice@xcite , and should occur in the present system for all @xmath167 .
it is interesting to consider the effect of a nonzero magnetic field @xmath114 .
it is known already that at @xmath30,@xcite such a field is an irrelevant perturbation , so that the system remains in a critical state , yet at sufficiently large @xmath168 it undergoes a locking into a smooth phase,@xcite approximated by any of the three symmetry - equivalent flat states of type `` @xmath169 '' with magnetization @xmath170 as also already noted@xcite , there is a critical value @xmath171 defined by @xmath172 , beyond which @xmath173 so that the system locks into long - range order as soon as @xmath114 is turned on . within this regime , there are still two subregimes with different behavior of @xmath174 near @xmath175 . for @xmath176 ,
the initial slope is zero , i.e. , the susceptibility is not divergent ; when @xmath177 , as occurs for @xmath178 , there is a divergent susceptibility and correspondingly there should be a singularity at @xmath179 in the spin structure factor @xmath180 .
what do we expect in the locked phase at @xmath181 ? here
the difference between the two kinds of flat states becomes crucial .
the @xmath114 field favors the @xmath169 type of flat state , but entropy favors the @xmath82 type of flat state .
thus we expect a transition to the @xmath169 state only at a nonzero critical field @xmath182 . on reducing @xmath114 through @xmath182 , a twofold symmetry breaking occurs ,
in which one of the @xmath183 sublattices becomes the @xmath184 ( disordered ) sublattice ; hence , this transition should be in the ising universality class .
presumably the line @xmath185 meets the @xmath186 axis at @xmath187 .
there must also be line of locking transitions @xmath188 , which terminates on the @xmath186 axis at @xmath171 . for @xmath30 ,
the effect of the magnetic field was confirmed numerically in ref . .
in this section we describe the implementation details of monte carlo simulations performed for spin-1 model in which @xmath42 takes only integer values from @xmath189 to @xmath190 . we then present numerical results for the relaxation times of slow modes in the monte carlo dynamics .
two different methods of compute the critical exponents of the spin , bond - energy , and uniform - magnetization operators are described in different sub - sections : one in terms of the extrapolated stiffness constants of the interface and the other in terms of the singularities of the corrsponding structure factors .
a spin is called _ flippable _ if its six surrounding nearest - neighbor spins alternate between @xmath191 and @xmath47 . clearly , changing the value of this flippable spin results in another new spin configuration in the ground - state ensemble , provided that we start with a spin configuration in the ensemble .
moreover , such an update maintains the global tilt of the interface due to the _ local _ nature of this update .
this update will be used as our monte carlo update in this paper .
two slightly different cases arise for different values of @xmath42 : ( 1 ) for @xmath192 , the local update is precisely equivalent to a spin flip i.e. , @xmath193 due to the absence of zero spin ; and ( 2 ) for all other values of @xmath42 , a random choice must be made in the local update : for example , @xmath194 or @xmath47 .
( recall @xmath0 denotes the spin magnitude of the original model . )
let @xmath40 and @xmath195 denote the number of zero - spins and flippable spins of configuration @xmath196 .
if an attempted single - spin update for @xmath196 results in a new configuration @xmath197 with @xmath198 and @xmath199 , then the transition probability @xmath200 in accordance with the detailed balance principle is : @xmath201 where @xmath202 denotes the _ bare _ transition probability : @xmath203 for @xmath192 , and @xmath204 for @xmath205 which reflects the random choice to be made in the local update as discussed above . with the transition probability given in eq .
( [ eq16 ] ) , it is straightforward to show that the detailed balance principle is satisfied , i.e. , @xmath206 , where @xmath207 denotes the probability for configuration @xmath196 to occur and @xmath208 since each spin configuration in the original spin-@xmath0 model has equal probability to occur .
note also that @xmath209 for large @xmath210 , so this rule is important only because of the finite system size . to implement in practice the transition probability given above
, we randomly select a site out of a list of the @xmath195 flippable sites , and randomly update this spin to one of the two possible new spin values if @xmath211 or simply flip this spin if @xmath192 .
the total numbers of zero spins @xmath198 and flippable spins @xmath199 in the resulting configuration are then computed .
this update is subsequently accepted with a probability : @xmath212 . a practical implementation of the transition probability given in eq .
( [ eq16 ] ) is thus achieved . throughout this paper , a unit time or one monte carlo sweep ( msc )
is defined such that there are @xmath213 attempts of updating within this unit of time ( or one attempt per spin on average ) . here
@xmath213 denotes the total number of spins in the simulation cell .
the simulation cell always contains @xmath214 spins in this paper unless explicitly mentioned otherwise .
periodic boundary conditions are adopted .
since we always start with a flat state , the simulations are thus performed in the sector with a zero global tilt of the interface .
we now discuss the correlations between the configurations generated sequentially in the monte carlo simulations by studying the relaxation time of the slow modes in the model , namely , the fourier modes @xmath216 which play the role of an order parameter@xcite .
the linear - response dynamics of such a mode is usually formulated as a langevin equation , @xmath217 where @xmath218 is the dissipation constant , and the static free energy functional @xmath219 is given by eq .
( [ eq6 ] ) .
here @xmath220 is a stochastic noise generated in the markov chain of monte carlo simulations .
as it is expected that the correlation time of the slow mode under consideration is much longer than that of the noise , and since the update steps are local and independent , it is proper to model @xmath220 as gaussian noise , uncorrelated in space or time : @xmath221 in which the choice of @xmath222 ensures that the steady - state of the interface under the langevin equation ( [ eq17 ] ) agrees with its equilibrium state under the free energy ( [ eq6 ] ) .
this linear stochastic differential equation can be solved easily by performing fourier transform .
( [ eq17 ] ) thus reduces to @xmath223 which implies an exponentially decaying correlation function of @xmath224 with the relaxation time @xmath215 given by @xmath225 therefore , the dynamical scaling exponent for the monte carlo dynamics , defined by @xmath226 , is always @xmath227 in the rough phase . to check this prediction on the dynamical scaling exponent in practice where the above continuum theory is regularized on a lattice , we compute
the following auto - correlation function @xmath228 of the _ microscopic _ height variable @xmath229 : @xmath230 here @xmath231 stands for the dynamical average , and the time @xmath232 is measured in unit of mcs . for each interger - valued @xmath233 , we perform @xmath234 mcs s with a flat initial configuration and compute the auto - correlation functions upto @xmath235 for modes that correspond to the five smallest @xmath236 values . in fig .
[ fig2 ] , we display the results so obtained for @xmath192 . other cases of @xmath42 are found to have very similar features .
it is clear from fig .
[ fig2 ] that @xmath237 can be fitted very well by @xmath238 where @xmath95 and the relaxation time @xmath215 are the fitting parameters . in other words ,
the relaxation is strictly exponential in all cases .
note that we used a cutoff @xmath239 in our fitting .
the same fitting procedure is carried out for other cases of @xmath42 .
the final results of the relaxation time @xmath215 as a function of @xmath236 for @xmath240 are shown in fig .
[ fig3 ] ; and for @xmath241 as an insert .
the fact that @xmath215 scales as @xmath236 for @xmath242 as indicated by the fitting in fig .
[ fig3 ] thus shows that the ground - state ensembles for @xmath242 are in the rough phase . on the other hand
, it is indeed clear from the insert that for @xmath243 and @xmath190 , @xmath215 curves downward as @xmath244 which is in sharp constrast to those of @xmath242 . from this , we conclude that ground - state ensembles for @xmath243 and @xmath190 are in the flat phase . as for @xmath245 , it is not conclusive from the data available whether @xmath215 scales as @xmath236 or curves downward as @xmath246 . nonetheless , the fact that the relaxation time of the slowest mode for @xmath245 is longer than for any smaller _ or larger _ value of @xmath0 , suggests that @xmath245 is very close to the locking transition .
further support for this phase diagram is also obtained by explicit calculations of stiffness constants and critical exponents which is discussed in the next section . as implied by sec .
[ fluctuations ] , the stiffness constant of the fluctuating interface can be directly measured by studying the long - wavelength fluctuations of the height variables , i.e. , their structure factor as given by eq .
( [ eq8 ] ) .
it should be noted that concerning the task of calculating the fourier components @xmath247 in eq .
( [ eq8 ] ) , it can be replaced by the approximation in terms of the _ microscopic _ height variables @xmath229 given by @xmath248 where @xmath99 labels a lattice site of the finite triangular lattice of total @xmath213 lattice sites used in the simulation . here
@xmath249 is the _ weight _ of a lattice site , i.e. , the area of its voronoi region , which is introduced so that the _ microscopic _ height variable @xmath229 coincides with the _ coarse - grained _ height variable @xmath247 in the long - wavelength limit ( @xmath250 ) .
but unlike @xmath247 , @xmath229 still contains features such as zone - corner singularities discussed in sec .
[ zone - corner ] that are only manifested in miscroscopic height variables .
starting with a flat state , we perform @xmath251 mcs s as the equilibrium time ; subsequent measurements of physical quantities are carried out at intervals of @xmath252 mcs s .
this separation is a compromise between the correlation times of small @xmath138 modes and of larger @xmath138 modes , which are respectively much longer and somewhat shorter than 20 mcs
see fig .
each run consisted of @xmath253 mcs , i.e. @xmath254 measurements were taken ; these were subdivided into @xmath252 independent groups for the purpose of estimating statistical errors .
the same procedure is used for all @xmath233 reported in this paper . in fig .
[ fig4 ] , we plot @xmath255 vs. @xmath256 for @xmath192 , including all @xmath257 in the first brillouin zone . from the plot
, we observe that @xmath255 is remarkably isotropic up to about @xmath258 .
this comes about because of the 6-fold rotational symmetry of the triangular lattice which ensures that _ anisotropy _ occurs only in @xmath259 and higher order terms , assuming that the function is analytic .
this is in constrast to other models defined on the square lattice where anisotropy already sets in at the order of @xmath260@xcite .
the lower envelope of the data points in fig .
[ fig4 ] corresponds to the line of @xmath261 in the @xmath262-vector space .
other cases of @xmath42 are found to have very similar features as illustated in the insert of fig .
[ fig4 ] where we plot the lower envelope for all @xmath233 .
the structure factor of the height variables appears to diverge in the long - wavelength limit @xmath263 for all @xmath0 values , even for the largest @xmath0 values .
( in the latter case , however , we believe one would see the plot asymptote to a constant value , in a sufficiently large system ; see below . )
two other interesting features of the structure factor are also revealed in the insert in fig .
[ fig4 ] : ( 1 ) for @xmath211 , it appears to indicate yet another singularity at the zone corner @xmath264 in the thermodynamic limit @xmath265 ; and ( 2 ) for @xmath192 , it approaches a constant instead .
as already discussed in sec .
[ zone - corner ] , the appearance of zone - corner singularities is expected , the precise nature of such singularities , however , is discussed in the next section . in the remaining of this section
, we analyze the zone - center singularity to check if height variables behave as required by eq .
( [ eq8 ] ) for the rough phase and consequently extract the stiffness constants . to further study the nature of zone - center singularity in terms of how @xmath123 scales as a function of @xmath256 in the long - wavelength limit , we show the log - log plot of @xmath266 vs. @xmath256 for @xmath267 in fig .
[ fig5 ] . comparing the simulation results for different systems sizes of @xmath268 , @xmath269 and and @xmath270
, we notice that data are well converged down to accessible small @xmath257 vectors except for the case of @xmath245 and @xmath271 , where the finite size effect is still discernible .
this is , of course , consistent with the fact that @xmath245 and @xmath271 are close to the locking transition where the correlation length diverges ; it is interesting , however , to notice that their finite - size trends are different . in the case
@xmath245 , the data plot for @xmath272 curves upwards less than that for @xmath273 , while in the case @xmath243 , the @xmath272 data show _ more _ upwards curvature than the @xmath273 data . by fitting @xmath266 to a function @xmath274 with @xmath275 being the fitting parameter , we obtain , using the data of system size @xmath272 and a cutoff @xmath276 , the exponent @xmath277 and @xmath278 respectively for @xmath279 , and @xmath109 .
apart from the case of @xmath245 , these values agree with @xmath280 as in the predicted @xmath281 power - law singularity of the structure factor in the rough phase , eq .
( [ eq8 ] ) . as for @xmath243 and @xmath190
, @xmath266 clearly deviates from a power - law scaling and instead curves upwards to level off , which indicates that models with @xmath243 and @xmath190 are in the smooth phases where @xmath123 remains finite as @xmath282 , as discussed in sec .
[ smooth ] .
this conclusion is in excellent agreement with that inferred from dynamical scaling analysis presented in sec .
[ dynamical ] .
it should be noted that in fig .
[ fig5 ] , as a general procedure adopted throughout this paper in extracting numerical values of some physical quantities , we have averaged the data corresponding to the same magnitude of @xmath236 to further reduce the effect due to statistical errors . the relative statistical error on each individual data point @xmath123 of small @xmath257 , which is measured directly from the variance among the 20 groups , is found to range from @xmath283 to @xmath284 .
this is indeed consistent with the estimates of such relative errors from the relaxation times of the slowest modes of models with different values of @xmath42 already given in sec .
[ dynamical ] .
it is perhaps also worth noting that another good check on the statistical errors on each data point is to compare the values of @xmath123 for three @xmath257 vectors which are related by @xmath285 rotations in reciprocal space , which ought to be equal by symmetry .
for example , in the case of @xmath192 , the values of @xmath123 for the three @xmath257 vectors of the same smallest magnitude @xmath286 of system size @xmath272 are , respectively , @xmath287 , @xmath288 , and @xmath289 , from which one thus also obtain the relative error of about @xmath283 .
this observation therefore motivates the averaging procedure used in this paper .
the stiffness constants can be subsequently determined by fitting @xmath290 to the function @xmath291 for the isotropic part of the data in which the stiffness constant @xmath86 and @xmath292 are the fitting parameters .
the final fitting on the averaged data is shown in fig .
[ fig6 ] where we used a cutoff @xmath276 in the fitting .
we also tried other different cutoffs of @xmath293 and @xmath294 , and found as expected that the stiffness is not sensitive to the value of cutoff as long as it falls into the isotropic part of the data .
for example , we obtain , in the case of @xmath192 , @xmath295 , and @xmath296 for cutoff @xmath297 , and @xmath298 respectively . therefore
, taking into account of the uncertainty introduced due to the cutoff , our final estimate for the stiffness constant is then @xmath299 which is in excellent agreement with the exact value @xmath300 .
similar procedure is carried out for other cases of @xmath42 and the results are tabulated in table i. in the same table , we also give the value for the critical exponents of spin , bond - energy and uniform magnetization operators which are obtained straightforwardly according to eqs .
( [ eq13 ] ) , ( [ eq15 ] ) and ( [ eq - gm ] ) .
the agreement of our @xmath301 values with the `` @xmath302 '' values from transfer - matrix eigenvalues ( see table i of ref . , is quite close and becomes better as @xmath0 grows ( until @xmath245 . ) as already discussed in sec .
[ finitet ] , a kosterlitz - thouless ( kt ) transition occurs at a critical value @xmath156 where @xmath153 , such that for @xmath303 algebraic correlations persist even at small finite temperatures .
it is clear from our data that @xmath304 . as for @xmath245 , the value of @xmath305 at the smallest nonzero @xmath306 is already larger than @xmath307 .
that is , even if the system may have a `` rough '' behavior at the length scales probed in the simulation , the stiffness constant is such that the locking potential is relevant and must dominates at sufficiently large length scales , as discussed in sec .
[ locking ] .
a similar observation has already been used to argue that the constrained potts antiferromagnet is in a smooth phase@xcite .
this fact together with the poor fitting using the formula suitable for the rough phase ( see the top curve of fig . [ fig6 ] ) leave us little choice but to conclude that the ground - state ensemble for @xmath245 also falls into the smooth phase , or possibly is exactly at the locking transition .
just as the finite - size effect for @xmath245 was severe both for the spin - spin correlations ( measured via monte carlo@xcite ) and also in spin - operator eigenvalues ( measured via tranfer - matrix,@xcite ) we similarly find it is severe for height fluctuations .
however , in view of the exponential relationship between the exponents and the stiffness constant , the latter measurements are much more decisive as to the true phase of the system . to sum up , based on the analysis on the nature of the singularity in the height structure factor at the long - wavelength limit and the numerical results on the stiffness constants , we thus conclude that the model exhibits three phases with a kt phase transition at @xmath1 and a locking phase transition at @xmath308 .
another more traditional approach@xcite in calculating the critical exponents of various operators is to compute the corresponding structure factors and analyze the power - law singularities at the appropriate ordering wave vectors .
namely , if the correlation function of an operator @xmath309 decays with distance as power - law ( thus critical ) @xmath310 then its structure factor near the ordering vector @xmath119 shows a power - law singularity @xmath311 from which the critical exponent @xmath312 can be numerically extracted . here in this section ,
we adopt this approach to calculate the critical exponents of spin , bond - energy , and uniform - magnetization operators so as to compare with those obtained from the stiffness constant . as given by eq .
( [ eq - se ] ) , @xmath313 . here
@xmath118 is the ordering vector of the bond - energy operator .
therefore the interesting feature of structure factor of height variables , namely , the appearance of zone - corner singularity as shown in fig . [ fig4 ] , is not only expected but also very useful in extracting the critical exponent @xmath314 . of course , such a zone - corner singularity can also be understood within the framework of interfacial representation , as in sec . [ height - rep ] , particularly subsec . [ zone - corner ] .
( similar zone - corner singularities have been studied in refs . and . ) finally , according to the exact result @xmath127 ( @xmath315 ) in the case of @xmath192 , i.e. , @xmath316 , the puzzling absence of the zone - corner singularity for @xmath192 as shown in fig .
[ fig4 ] is also resolved . in fig .
[ fig7 ] , we plot @xmath317 vs. @xmath318 where we have averaged data points with the same magnitude of @xmath319 .
fitting @xmath320 to the function @xmath321 where @xmath322 and @xmath323 are the fitting parameters , we obtain the critical exponents @xmath324 which are tabulated in table i. in practice , we used two different cutoffs in the fitting : @xmath325 and @xmath326 .
the fitting for the latter is shown in fig .
[ fig7 ] , and the final quoted errors take into account the uncertainty due to the cutoffs . similarly , we also computed the structure factor for the spin operator @xmath327 using fast fourier transform while computing the height - height correlation function within the same monte carlo simulations .
results are shown in fig .
[ fig8 ] and the extracted exponents are also tabulated in table i. fitting precedure used is exactly the same as that for bond - energy except that we fit @xmath328 to the function @xmath329 with @xmath292 and @xmath330 being the fitting parameters . from table
i , we note that the critical exponents extracted in this way are in good agreement with those obtained from stiffness constant utilizing the interfacial representation , however , the latter yields much better statistical errors by an order of magnitude using the same monte carlo simulation data .
this clearly demonstates the superiority of the interfacial representation in extracting critical exponents from numerical data .
similar points were made regarding other models , but based on much less extensive simulation data , in refs . and .
similar fits were attempted for @xmath245 , yielding @xmath331 and @xmath332 . while the statistical error on @xmath333 is too large to render the fitting meaningful , the increase in the value of @xmath334 when compared with @xmath335
is added evidence that @xmath245 is _ not _ in the rough phase ; if it were still rough at that value of @xmath0 , we would have expected a continuation of the decreasing trend of @xmath336 with @xmath0 . as for the cases of @xmath243 and @xmath190 ,
the structure factors of both the spin and bond - energy operators show _ weaker _ than power - law behavior as @xmath337 , as in figs .
[ fig7 ] and [ fig8 ] , but they increase to a larger value ( not seen in these logarithmic plots ) right _ at _ @xmath338 .
this is indeed consistent with the @xmath339-function singularity .
expected if these cases fall into the smooth phase with long - ranged order of the spin and bond - energy operators .
finally , we consider the uniform magnetization correlation exponent @xmath129 .
when @xmath340 , it can be predicted ( see @xmath341 in table i ) that @xmath342 , implying a divergent ( ferromagnetic ) susceptibility and a divergent structure factor @xmath343 as @xmath344 now , due to the linear relation ( [ eq - m ] ) between @xmath345 and @xmath143 , we immediately obtain @xmath346 near @xmath179 , just as @xmath347 near @xmath348 ( see sec .
[ zone - corner ] and eq .
( [ eq - se ] ) ) thus , a singularity at @xmath179 is expected in the structure factor of spin operator which is plotted in fig .
[ fig9 ] . from this figure
, it appears that only for @xmath349 , @xmath350 , and @xmath109 does @xmath343 show a power - law singularity indicated by a straight line in this log - log plot .
this confirms the prediction based on the stiffness constant ; however , the numerical values of @xmath129 extracted this way ( see table i ) differ considerably from those calculated from the stiffness constant in the case of @xmath351 and @xmath109 .
it is also apparent from table i that @xmath352 is systematically overestimated as compared with the more accurate value derived from height fluctuations .
we suspect that a similar overestimation affected the values of @xmath302 that were deduced from the finite - size scaling of the susceptibility of the staggered magnetization@xcite ( this obviously measures the same fluctuations seen in @xmath353 near @xmath338 . ) those data ( also quoted in ref . )
have quoted errors about four times as large as ours for @xmath301 .
their exponent values are all noticeably larger than the accurate value ( @xmath301 or @xmath354 from ref . ) becoming _ worse _ as @xmath0 grows ( for @xmath355 the difference is twice their their quoted error . ) clearly the systematic contribution to their errors was underestimated .
the transfer - matrix method@xcite ought to provide the effective exponent @xmath302 for spin correlations on length scales comparable to the strip width , and hence is likewise expected to overestimate @xmath302 ; indeed , every @xmath302 value found in ref . is slightly larger than our corresponding @xmath301 value .
which type of flat state is actually selected in the smooth phase ?
[ fig10 ] shows the measured expectation of @xmath40 , the number of zero spins in the spin-1 representation , for @xmath356 .
as @xmath0 grows , it is found that @xmath357 approaches its maximum allowed value @xmath358 as in the @xmath82 state , rather than zero , as in the @xmath169 state .
thus , the flat states with half - integer valued @xmath59 in fig . [ fig1 ] are being selected in the smooth phase .
translating back to the spin-@xmath0 model , this means that spins on two sublattices of the triangular lattice take the extremal values , @xmath15 and @xmath16 respectively , while spins on the third sublattice remain disordered .
it is perhaps more illuminating to study the distribution of height variables to probe the height fluctuations in the smooth phase . to this end
, we also show , in fig .
[ fig10 ] , the histogram of height variable @xmath59 in the cases of @xmath359 and @xmath360 , which is measured for a _ typical _ configuration generated in the monte carlo simulations.@xcite .
the broad distribution observed in the case of @xmath359 ( @xmath361 ) evolves to a narrowly peaked distribution in the case of @xmath360 ( @xmath362 ) .
( it decays as @xmath363 . )
this supports the intuitive picture presented in sec .
[ smooth ] .
furthermore , the center of this peaked distribution is half - integer valued .
( numerically , the mean is @xmath364 for the distribution plotted in fig .
[ fig10 ] . ) in other words , the locking potential @xmath365 favors the @xmath366 type of flat state , in which one sublattice is flippable , rather than the @xmath169 type of flat state .
( see fig .
[ fig1 ] ) .
this kind of flat state was also expected analytically in the limit of large @xmath0 @xcite .
we have also computed @xmath367 for each value of @xmath0 , in two ways .
first , @xmath368 is just normalization factors times @xmath369 , which we accumulated throughout the monte carlo run , as described earlier in this section ; then it can be shown that @xmath370 exactly . for @xmath371
this gives @xmath372 and @xmath373 for @xmath359 and @xmath360 , respectively , showing the contrast of the rough and smooth behavior .
secondly , we can compute @xmath374 directly from the histogram ( from one snapshot ) seen in fig .
[ fig10 ] ; this gives respective values @xmath375 and @xmath376 , in satisfactory agreement with the first method .
the exotic `` hidden order '' phase@xcite ( see sec . [ smooth ] ) can be ruled out on the basis of these data : according to eq .
( [ eq - dhbound ] ) the variance of @xmath59 should be at least @xmath377 in the hidden - order phase , while our measurements indicate it is at most only @xmath373 .
furthermore , for @xmath243 and @xmath190 , the structure factor @xmath378 at the zone - corner wave vector @xmath338 ( not plotted ) was much larger than at nearby @xmath138 ; that direct suggests a @xmath339-function singularity in the thermodynamic limit , i.e. , existence of long - ranged spin order in which @xmath379 on at least two of the sublattices . additionally , the spin structure factor @xmath353 near the zone - corner wave vector @xmath338 ( fig . [ fig8 ] ) showed a striking curvature in the `` smooth '' cases @xmath243 and @xmath190 , quite different from the behavior at smaller @xmath0 .
this makes it plausible that @xmath380 , so that spin fluctuations have short - range rather than power - law correlations for @xmath362 .
( it was not emphasized in ref . , but
power - law correlations are implied if one takes seriously their measured values @xmath381 for @xmath382 . )
we propose , then , that actually @xmath383 for @xmath384 , as in the simplest picture of the smooth phase , and that the observed nonzero values are simply finite - size effects due to the very slow crossover from rough to smooth behavior near a roughening transition ( see sec .
[ disc - crossover ] , below , for further discussion . )
to conclude , in this article , we have investigated the ground - state properties of the antiferromagnetic ising model of general spin on the triangular lattice by performing monte carlo simulations . utilizing the interfacial representation
, we extrapolated the stiffness constants by studying the long - wavelength singularity in the height variables , which in turn lead to straightforward calculation of critical exponents of various operators within the framework of height representation .
the results so obtained are further compared with those extracted from a more tranditional method , and demonstrate that the method in terms of height representation method is by far the preferable one for extracting the critical exponents .
we also analyzed both the dynamical and static properties of the model in order to map out the phase diagram which consists of three phases with a kosterlitz - thouless phase transition at @xmath1 and a locking phase transition at @xmath308 . even in the smooth state , analysis of the height fluctuations ( as in @xmath374 was helpful in resolving questions which are made difficult by the strong finite - size effects near the locking transition .
one of our initial motivations for this study was the possibility of finding rational exponents even for @xmath167 .
we believe the results in table i are the first which are accurate enough to rule out this possibility .
indeed , @xmath385 and @xmath386 , with differences similar to the error ( 0.001 ) .
but _ any _ random number differs from a rational number with denominator @xmath387 by the same typical error .
the exception is that @xmath388 is quite close to @xmath389 , but we have given other reasons to be suspicious of this value .
another intriguing question was whether the critical values @xmath391 and @xmath392 are exactly integers .
previous data@xcite suggested that @xmath393 exactly , and had large enough errors that @xmath394 could not be excluded .
since @xmath395 and @xmath396 , this question was answered by the preceding subsection : we find that definitely @xmath397 .
furthermore , we suspect @xmath398 as concluded in sec .
[ stiffness ] since the effective stiffness at the length scale we access is more than enough to drive the system to the locked phase .
the question of the value of @xmath133 suggests paying closer attention to the behavior of systems near the locking transition .
it has been noted previously how the locked phase tends to behave qualitatively like the rough phase in a finite - size system , since the crossover is a very slow function of size.@xcite this is consistent with the apparent power - law behaviors observed at @xmath399 in previous studies@xcite and with the tendency of those studies to overestimate the exponents @xmath400 and @xmath401 ( as compared with our more accurate estimates . )
this would suggest that , if extensive finite - size corrections were included in our analysis , they would reduce our estimate of @xmath390 a bit further , i.e. we would more definitely conclude that @xmath245 is in the locked phase .
our analysis near the locking transition at @xmath390 suffers from our ignorance of the expected functional form of the critical behavior as a function of @xmath402 .
a study of the roughening transition@xcite used the kosterlitz - thouless ( kt ) renormalization group to derive analytic approximations for the total height fluctuation ( closely analogous to @xmath374 in our problem ) , which made it possible to overcome very strong finite - size effects and fit the roughening temperature precisely .
use of kt finite - size corrections was also essential in extracting meaningful numbers from transfer - matrix calculations near the locking transition induced by a magnetic field in ref . .
thus , a similar adaptation of the kt renormalization group to give expressions for the behavior of @xmath403 , as a function of ( small ) @xmath404 and @xmath402 , or the functional form of @xmath405 near @xmath390 , could make possible a more conclusive answer as to whether @xmath406 exactly .
since the long - wavelength behavior in this model ( in its rough phase ) is purely gaussian with @xmath227 ( see sec . [ dynamical ] ) , the critical slowing down is particularly transparent .
it seems feasible to take advantage of the existence of a height representation to develop an acceleration algorithm .
for example , it might be possible to extend the cluster algorithms which are known for the @xmath30 triangular ising antiferromagnet.@xcite these are well - defined at @xmath151 , but their effectiveness seems to depend in a hidden fashion on the existence of the height representation when @xmath125 .
an intriguing alternative approach starts from the observation that at long wavelengths the system obeys langevin dynamics ( see sec .
[ dynamical ] and ref . ) .
fourier acceleration@xcite , a nonlocal algorithm ( efficient owing to use of the fast fourier transform algorithm ) , is known to be effective in such cases .
the key is to replace the uncorrelated noise function @xmath407 of eq .
( [ eq18 ] ) with a new correlated noise function having @xmath408 .
this might be implemented by first constructing a random function with such correlations , and then updating flippable spins with probabilities determined by that function , in such a fashion as to satisfy detailed balance .
additionally , it may be possible to analyze transfer - matrices using the height representation .
quite possibly this would yield an order - of - magnitude improvement in the accuracy of the numerical results , for the same size system , similar to the improvement in analysis of monte carlo data .
the transfer matrix breaks up into sectors corresponding to the step made by @xmath50 upon following a loop transverse to the strip ( across the periodic boundary conditions .
then the stiffness can be extracted directly from the ratio of the dominant eigenvalues of two such sectors ; such an analysis is already standard for quasicrystal random tilings , for which the long - wavelength degree of freedom is also an effective interface @xcite .
.stiffness constant and critical exponents .
here @xmath409 , @xmath410 and @xmath411 are the estimates for the critical exponents of spin and bond - energy operators calculated from the stiffness constant @xmath86 as done in sec .
v(c ) , while @xmath412 , @xmath413 , and @xmath414 stand for the same critical exponents , but extracted from the singularities of their respective structure factors in sec .
estimated errors are also given in the parenthesis . [ cols="^,^,^,^,^,^,^,^",options="header " , ] j. stephenson , j. math .
* 5 * , 1009 ( 1964 ) ; j. math .
* 11 * , 413 ( 1970 ) .
h. w. j. blte and h. j. hilhorst , j. phys .
a * 15 * , l631 ( 1982 ) ; ( b ) .
b. nienhuis , h.j .
hilhorst , and h.w .
blte , j. phys .
a * 17 * , 3559 ( 1984 )
. the limit of @xmath416 is well - defined and has been simulated by ref . and by y. yamada , s. miyashita , t. horiguchi , m. kang , and o. nagai , j. mag .
* 140 - 144 * , 1749 ( 1995 ) .
it is interesting as a model which has an infinite number of continuous degrees of freedom ( changes of @xmath417 for loose spins ) , under which the energy is degenerate ; in such systems it is expected that a particular average pattern of order is selected when fluctuations are taken into account , but it is nontrivial to compute the selecting terms of free energy , since the potential minimum is flat rather than quadratic as a function of the spins .
s. miyashita and h. kawamura , j. phys .
54 * , 3385 ( 1985 ) ; b. kleine , e. mller - hartmann , k. frahm , and p. fazekas , z. phys .
b * 87 * , 103 ( 1992 ) ; q. sheng and c. l. henley , j. phys .
cond . mat . * 4 * , 2937 ( 1992 ) .
g. murthy , d. arovas , and a. auerbach , preprint ( cond - mat/9607217 ) .
t. ogawa and y. nakajima , prog .
( suppl . ) * 87 * , 90 ( 1986 ) , independently identified the height representation , but considered the height as defined modulo 6 ( so roughness could not be clearly defined , and the connection to power - law correlations was not evident . ) although we only describe the interfacial representation for the spin-1 model , a very similar procedure can be carried out for the original spin-@xmath0 model straightforwardly .
contrary to what is suggested in ref .
, p. 3890
, it is also possible to construct @xmath50 directly from @xmath33 for general @xmath0 , _ without _ first mapping it to a spin-1/2 or spin-1 model .
the prescription is simply to replace @xmath418 on the right - hand side of ( [ eq3 ] ) .
the long wavelength height fluctuations would be the same under the new definition ; it would be useful only at nonzero temperature , to confirm that excitations have burgers vector the it had long been noted that the triangular ising model could be mapped apporoximately to a 6-state clock model model ( see e. domany , m. schick , j. s. walker , and r. b. griffiths , phys . rev .
b * 18 * , 2209 ( 1978 ) , particularly in connection with the effect of adding a second - neighbor ferromagnetic interactionn the height mapping of the _ microstates _ , is more powerful , in that it shows _ every _ excitation must correspond to vortex / antivortex pairs ; this is important in predicting the @xmath151 behavior .
j. kondev , j. de gier , and b. nienhuis , preprint cond - mat/9603170 ; j. kondev , to appear _ proc .
exactly soluble models in statistical mechanics _ ( march 1996 , northeastern university ) , cond - mat/9607181 .
p. chandra , p. coleman , and l. b. ioffe , phys .
b * 49 * , 12897 ( 1994 ) used a height representation to understand the phase diagram of another triangular antiferromagnet ; the behavior is reminiscent of the @xmath419 limit in our system ( discussed in footnote ) .
a plaquette has half - integer - valued @xmath59 if and only if it has a zero spin at exactly one corner ( fig .
[ fig1 ] ) . from this
it follows that the total weight of the half - integer bars in the in the histogram is @xmath420 ; this clarifies the relation between the upper and lower panels of fig .
[ fig10 ] .
h. g. evertz , m. hasenbusch , m. marcu , and k. pinn , physica a * 199 * , 31 ( 1993 ) . | we map the ground - state ensemble of antiferromagnetic ising model of spin-@xmath0 on a triangular lattice to an interface model whose entropic fluctuations are proposed to be described by an effective gaussian free energy , which enables us to calculate the critical exponents of various operators in terms of the stiffness constant of the interface .
monte carlo simulations for the ground - state ensemble utilizing this interfacial representation are performed to study both the dynamical and the static properties of the model .
this method yields more accurate numerical results for the critical exponents . by varying the spin magnitude in the model ,
we find that the model exhibits three phases with a kosterlitz - thouless phase transition at @xmath1 and a locking phase transition at @xmath2 .
the phase diagram at finite temperatures is also discussed . |
of all the iron - based superconductors @xcite , alkali iron selenides @xmath6fe@xmath1se@xmath2 ( @xmath7 k , rb , cs , tl ) @xcite are unique in that superconductivity in this class of materials always coexists with a static long - range antiferromagnetic ( af ) order with a large moment and high n@xmath8el temperature @xcite .
this is in contrast to iron pnictide superconductors @xcite where optimal superconductivity arises from the suppression of the static af order in their nonsuperconducting parent compounds @xcite .
an attempt to understand the coexisting static af order and superconductivity in alkali iron selenides @xcite has produced two proposed scenarios . in the first , superconductivity is believed to coexist microscopically with the static af order @xcite .
however , the af order in @xmath6fe@xmath1se@xmath2 forms a @xmath3 block af structure with an ordered moment of @xmath93.3 @xmath10 per fe as shown in fig .
1(a ) @xcite , making it unclear how superconductivity can survive such a large magnetic field background arising from the ordered moments @xcite .
alternatively , superconductivity in @xmath6fe@xmath1se@xmath2 may arise from a chemically separated superconducting phase in the matrix of the insulating block af phase .
although transmission electron microscopy ( tem ) @xcite , x - ray / neutron diffraction @xcite , scanning tunneling microscopy ( stm ) @xcite , m@xmath11ssbauer spectroscopy @xcite , muon spin relaxation @xcite , apertureless scattering - type scanning near - field optical microscopy @xcite , and nuclear magnetic resonance @xcite experiments have provided ample evidence for phase separation , where superconductivity comprises about 10 - 20% of the volume of @xmath6fe@xmath1se@xmath2 , there is currently no consensus on the chemical composition or crystal structure for the superconducting @xmath6fe@xmath1se@xmath2 .
for example , while some tem @xcite , x - ray scattering @xcite , and stm @xcite measurements suggest that the superconducting phase of @xmath6fe@xmath1se@xmath2 is @xmath12fe@xmath2se@xmath2 with the bafe@xmath2as@xmath2 iron pnictide crystal structure @xcite , other tem and stm measurements propose that the superconducting phase consists of a single fe vacancy for every eight fe - sites arranged in a @xmath13 parallelogram structure [ fig .
1(c ) ] @xcite .
in addition , single crystal neuron diffraction experiments indicate that superconductivity in @xmath6fe@xmath1se@xmath2 may arise from a semiconducting af phase with rhombus iron vacancy order [ fig .
1(b ) ] @xcite instead of the well - known the insulating @xmath3 block af phase @xcite .
fe@xmath1se@xmath2 phases : a ) main vacancy ordered phase with @xmath3 block af order @xcite .
the @xmath14 and @xmath15 signs indicate fe spin directions relative to the fese plane .
b ) proposed phase for semiconductor , where spin directions are marked as arrows @xcite .
c ) suggested superconducting phase with @xmath13 iron vancancy order @xcite .
d ) iron - disordered , partially occupied orthorhombic phase with bafe@xmath2as@xmath2 iron pnictide structure .
e - g ) schematic for phase separation through spontaneous nucleation in the temperature region of iron mobility . a disordered @xmath16 phase above @xmath17 . below @xmath18 , the @xmath19 symmetry insulating phase forms at random sites and spreads enriching the iron in remaining disordered phase until either full iron occupation of second phase ( slow cooling ) or temperature drops below zone of iron mobility ( quenching ) . ]
given the numerous proposed crystal structures for the superconducting @xmath6fe@xmath1se@xmath2 , a determination of the true chemical composition and relationship with the af insulating phase is essential to understand the bulk electronic properties @xcite .
in this article , we present systematic neutron powder diffraction measurements on superconducting and insulating k@xmath0fe@xmath1se@xmath2 . by carefully comparing rietveld refinements of neutron diffraction spectra of the superconducting and insulating k@xmath0fe@xmath1se@xmath2 in the high temperature paramagnetic phase
, we find that both materials are phase pure with the iron pnictide crystal structure ( space group @xmath16 ) @xcite and slightly more iron in the superconducting sample . on cooling to the low - temperature ground state ,
while the insulating k@xmath0fe@xmath1se@xmath2 remains a single phase , now of a @xmath3 iron vacancy structure with block af order @xcite , the superconducting k@xmath0fe@xmath1se@xmath2 becomes phase separated into the stoichiometric @xmath3 iron vacancy ordered k@xmath20fe@xmath21se@xmath2 and potassium deficient k@xmath4fe@xmath5se@xmath2 phase with the iron pnictide crystal structure @xcite .
the superconducting phase arises in the region of the excess iron between the low temperature @xmath3 iron vacancy ordered phase , and its volume fraction can be controlled through temperature cycling and quenching processes .
rietveld and pair distribution function ( pdf ) analysis of the data indicates that the superconducting phase can not be the stoichiometric @xmath22fe@xmath2se@xmath2 @xcite , @xmath13 @xcite or rhombus @xcite iron vacancy ordered phase .
therefore , the superconducting phase in @xmath6fe@xmath1se@xmath2 arises from a spontaneous chemical phase separation from the insulating @xmath3 iron vacancy ordered phase due to excess iron , fundamentally different from superconductivity in iron pnictides @xcite .
symmetry phase at 590 k and @xmath19 symmetry phase at 15 k. ( b inset ) the unfit peaks in the superconducting sample with only the @xmath19 symmetry are marked with arrows .
there is an unidentified impurity peak near 3.2 @xmath23 marked as @xmath24 in the superconducting sample .
the peak has no temperature dependence .
side cut of pure c)fese and d)kfe@xmath2se@xmath2 crystal structures .
darker atoms are placed in front of iron while lighter ones behind .
note inversion of middle layer necessary to form the bafe@xmath2as@xmath2 symmetry phase from fese .
e ) best fits using different proposed second phases .
fese can be interpreted as a k - intercalated sample with no k ordering .
f ) best fit using orthorhombic @xmath25 symmetry structure akin to that of bafe@xmath2as@xmath2 .
the inset shows clear broadening of the @xmath26 peak at base temperature ( blue dots ) compared to above @xmath17 ( line ) .
such broaden can be induced by a small orthorhombic lattice distortion .
g ) comparison of the fits using pure @xmath3 iron vacancy order without af order ( green ) , the effect of af order ( orange ) , the superconducting phase ( blue ) , and the combination of all phases ( red ) . at 15 k , the lattice parameters of the block af phase for the superconducting and insulating samples are @xmath27 , @xmath28 and @xmath29 , @xmath30 , respectively .
the lattice parameters for the k@xmath4fe@xmath5se@xmath2 phase are @xmath31 , @xmath32 , and @xmath33 . ]
symmetry phase ( inset , blue line ) .
the split in the fe - fe peak results from a contraction of [ fe - fe]@xmath34 and rotation of inner iron square .
b ) the superconductor requires an addition phase , here fit with @xmath25 symmetry ( inset , green line ) and disordered fe vacancies .
the fe - fe bond distance in the additional phase matches the [ fe - fe]@xmath34 distance , resulting in an increase of intensity relative to [ fe - fe]@xmath2 . ]
we carried out neutron scattering experiments using the neutron time of flight powder diffractometer npdf at the los alamos national laboratory .
this spectrometer has the advantage of a large @xmath35-range , allowing a determination of the average as well as local structures of the system through rietveld @xcite and pdf analysis of the diffraction spectra @xcite , respectively .
we grew several large single batches of k@xmath0fe@xmath1se@xmath2 via the self - flux method with nominal dopings of @xmath36 , @xmath37 and @xmath36 , @xmath38 for the insulating and superconducting samples , respectively .
since transport measurements are sensitive to paramagnetic tetragonal to iron vacancy ordered structural phase transition ( @xmath17 ) , af ordering ( @xmath39 ) , and superconductivity @xmath40 @xcite , samples from each batch were characterized by in - plane resistivity and magnetization to determine @xmath17 , @xmath39 , and @xmath40 to be approximately 540 k , 500 k and 32 k in the superconducting sample and 520 k and 500 k in the insulator . from previous neutron diffraction work @xcite , we know that structural and magnetic phase transitions in k@xmath0fe@xmath1se@xmath2 occur above the room temperature . choosing only crystals from a single batch to minimize stoichiometry differences , we ground several grams of each composition into a fine powder and divided it into two sets .
one set of powders was measured only at low temperatures while the other was measured to temperatures above @xmath17 to observe any hysteretical effects . for the neutron diffraction experiments , each sample
was effectively measured along a temperature loop beginning at the base temperature of 15 k and measured at intervals to above @xmath17 before returning to room temperature .
this produced two sets of data , one for superconducting and one for insulating k@xmath0fe@xmath1se@xmath2 , at identical temperatures to enable direct comparison .
following the completion of the neutron scattering measurements , exact crystal stoichiometry from small portions of both high and low temperature powder was measured via inductively coupled plasma ( icp ) analysis . symmetry unit cell .
b ) refined magnetic moment shows hysteresis and strengthens upon annealing .
c ) refined iron concentration for iron occupation by site also shows enhanced vacancy ordering below 450 k upon warming from a quenched state .
d ) iron site labels in the @xmath19 symmetry phase .
all arrows indicate direction in thermal cycle . ]
we carried out rietveld analysis to determine the average crystal structure and chemical compositions of the superconducting and insulating k@xmath0fe@xmath1se@xmath2 .
figure 2(a ) shows the intensity versus wave vector spectra and fits from the rietveld analysis for the insulating k@xmath0fe@xmath1se@xmath2 . in the paramagnetic tetragonal state ( top panel ) at @xmath41 k , the spectra can be well fit by the @xmath16 symmetry tetragonal unit cell ( green solid line ) suitable for the paramagnetic tetragonal state of bafe@xmath2as@xmath2 @xcite .
on cooling to @xmath42 k in the block af ordered state , we find that the spectra can still be well described by a single phase of k@xmath43fe@xmath44se@xmath2 , but with the space group @xmath19 and @xmath3 iron vacancy order [ fig .
1(a ) ] @xcite .
therefore , there is no phase separation in the insulating k@xmath43fe@xmath44se@xmath2 .
figure 2(b ) shows neutron diffraction spectra for the superconducting k@xmath0fe@xmath1se@xmath2 . in the paramagnetic tetragonal phase ( @xmath41 k )
, we find that a single phase tetragonal unit cell with the @xmath16 symmetry can fit the data well [ top panel in fig .
the refined stoichiometry indicates that the superconducting k@xmath0fe@xmath1se@xmath2 has more iron ( @xmath45 ) than that of the insulator ( @xmath46 ) . on cooling to @xmath42
k , we find that the @xmath19 space group suitable for the insulating k@xmath43fe@xmath44se@xmath2 in fig .
2(a ) can still fit majority of the bragg peaks very well .
however , there are additional bragg peaks occurring at positions not allowed by the @xmath19 symmetry [ bottom panel and arrows in the inset of fig .
2(b ) ] . .
phase 1 is the block af phase and phase 2 is the iron pnictide phase with k deficiency .
@xmath47 , @xmath48 , and @xmath49 are positions of atoms in the unit cell , and @xmath50 is the isotropic debye waller factor .
occ is the occupancy of the atomic site for different atoms . [
cols="^,^,^,^,^,^,^",options="header " , ] beginning with the assumption that the new phase is associated with superconductivity , there are many possible crystalline structures .
these include the rhombus iron vacancy order [ fig .
1(b ) ] @xcite , @xmath13 iron vacancy ordered phase [ fig .
1(c ) ] @xcite , iron disordered @xmath16 tetragonal symmetry [ fig .
1(d ) ] , expanded structure of fese [ fig .
2(c ) ] @xcite , and finally the bafe@xmath2as@xmath2 type iron pnictide structure [ fig .
2(d ) ] @xcite . to determine the space group and chemical composition of the new phase in the low temperature state , we plot in fig .
2(e ) the difference plot between the superconducting and insulating @xmath51fe@xmath1se@xmath2 , omitting regions of peak misalignment due to slightly different lattice parameters ( see fig .
2 caption ) .
we can then use different proposed models for the superconducting @xmath51fe@xmath1se@xmath2 to fit the observed spectra .
the yellow , red , and blue solid lines in fig .
2(e ) show rietveld fits of the data using the rhombus @xcite , @xmath13 @xcite iron vacancy ordered , and pure fese @xcite structures , respectively .
none of them can match allowed peak position and intensity properly .
figure 2(f ) shows rietveld analysis using the bafe@xmath2as@xmath2 structure . although the tetragonal @xmath16 symmetry suitable for the high temperature paramagnetic state fits the data fairly well ( black line )
, we find that a slight orthohombic distortion ( space group @xmath25 ) fits the data better as shown in solid green line of fig .
a direct comparison of the low ( blue circles ) and high temperature ( solid line ) data in the inset of fig .
2(f ) displays slight broadening of the low - temperature data , confirming the low - temperature orthorhombic symmetry . the best overall fit yielded 13.6% of the crystal volume in the second phase with a refined stoichiometry of k@xmath52fe@xmath53se@xmath2 , and the remainder in the block af phase .
the combined fit and bragg peak positions for different phases are shown in fig .
2(g ) and positions of all atoms in the unit cell are summarized in table i. using independently fit stoichiometry and phase fraction , we can compare a calculated overall crystal stoichiometry with the measured icp values .
doing so at base temperature results in an overall crystal stoichiometry of k@xmath54fe@xmath55se@xmath2 providing excellent agreement with icp measurements of k@xmath56fe@xmath57se@xmath2 .
b ) refined magnetic moment in insulating phase showing same behavior as pure insulator .
c ) refined iron occupation in insulating phase by site .
notable refinement of vacancy order upon temperature cycling .
d ) refined main phase fraction upon temperature cycling . a clear increase in main phase fraction upon warming to 450 k in a quenched sample .
solid lines are guides to the eye . ] in addition to carrying out rietveld analysis , we can also compare the low - temeprature local structural distortion of the superconducting and insulating @xmath51fe@xmath1se@xmath2 using the pdf calculation .
at base temperature , the @xmath3 iron vacancy order is already established and the square of iron in the center of the unit cell [ see inset in figure 3(a ) ] twists and contracts , leading to a slightly different intra- ( light ) and inter - square ( dark ) fe - fe bond distance @xcite . indeed , this is observed as two marked peaks near 2.7 angstroms in the pdf analysis of the insulating k@xmath0fe@xmath1se@xmath2 [ fig .
3(a ) ] , where the integrated peak intensity indicates the population density of atomic pairs in the sample .
figure 3(b ) shows the same pdf analysis for the superconducting @xmath51fe@xmath1se@xmath2 .
although we still see the same splitting of the two fe - fe bond distances , the intensity ratio of the split peaks differs between compositions . for a pure @xmath3 iron vacancy phase with @xmath19 symmetry
, one would expect integrated intensity of these peaks to be in a 2:1 ratio reflecting the number of such atomic pairs in the unit cell [ fig .
while a quantitative comparison is difficult , a qualitative comparison to the superconducting and insulating samples is still instructive . in figure 3(b ) for the superconducting sample , we see that the intensity of [ fe - fe]@xmath34 is substantially larger compared to the same peak in the insulator despite comparable intensity of the [ fe - fe]@xmath2 peak .
this discrepancy can only be understood assuming the presence of an additional phase in the superconducting sample with one fe - fe bond distance , consistent with rietveld analysis shown in fig .
2 . if we assume that the superconducting k@xmath0fe@xmath1se@xmath2 is phase pure in the high temperature paramagnetic state , it would be interesting to understand how it becomes phase separated at low temperatures .
in a recent tem paper @xcite , it was argued that the superconducting phase in @xmath51fe@xmath1se@xmath2 has a chemical composition of k@xmath58fe@xmath2se@xmath2 and originates from spinodal phase separation around @xmath59 k to form a archimedean solid - like framework embedded in the insulating k@xmath20fe@xmath21se@xmath2 matrix .
these results are consistent with previous x - ray diffraction findings which suggest phase separation @xcite .
while this scenario is not in conflict with our data , we present another possible picture in fig .
1(e)-1(g ) showing the temperature evolution of the crystal structures for superconducting k@xmath0fe@xmath1se@xmath2 . at high temperature ,
the entire sample is in the high symmetry @xmath16 phase . at this temperature ,
the iron atoms are extremely mobile resulting in a homogeneous iron deficient iron pnictide structure [ fig .
1(e ) ] . when cooled below @xmath17 , a @xmath3 iron vacancy ordered phase with chemical composition k@xmath60fe@xmath21se@xmath2 forms due to its lower energy @xcite .
the excess iron in the high temperature paramagnetic tetragonal phase are segregated from the iron vacancy ordered main phase , resulting in k@xmath60fe@xmath2se@xmath2 , which retains the iron pnictide structure [ fig .
1(f ) and 1(g ) ] . in this picture
, the rate of cooling and average fe concentration will determine the final phase fractions as well as second phase stoichiometry .
this picture is consistent with the hysteresis effects in superconducting volume fraction seen in the scanning electron microscopy measurements @xcite . to quantitatively determine the effect of annealing on the insulating and superconducting of k@xmath0fe@xmath1se@xmath2
, we carried out careful rietveld analysis of the temperature evolution of the neutron diffraction spectra .
figure 4 summarizes the outcome for the insulating k@xmath43fe@xmath44se@xmath2 . in an ideal @xmath3 block
af structure , there should be full occupancy at iron site 1 ( fe@xmath34 ) and full vacancy at iron site 2 ( fe@xmath2 ) as shown in fig .
4(d ) @xcite .
figure 4(a ) shows no hysteresis in temperature dependence of the lattice parameters across the block af ( @xmath19 ) to paramagnetic tetragonal ( @xmath16 ) structural phase transition .
figure 4(b ) shows temperature dependence of the ordered magnetic moment , which clearly increases at 300 k after warming up above @xmath17 . from fig .
4(c ) , we see that the significant iron occupation ( @xmath930% ) in the as - grown k@xmath43fe@xmath44se@xmath2 almost vanishes upon warming up to 450 k , suggesting a wide temperature region of iron mobility , enabling an ideal @xmath3 block af structure that increases the fe ordered moment .
figure 5 summarizes similar rietveld analysis for the superconducting k@xmath0fe@xmath1se@xmath2 .
the lattice parameters of the block af and iron pnictide phases merge together above @xmath17 , consistent with a pure paramagnetic tetragonal phase [ fig .
5(a ) ] . figure 5(b ) shows temperature dependence of the ordered moments for the block af structure , which again reveals a clear increase at 300 k after warming up to above @xmath40 similar to the insulating k@xmath43fe@xmath44se@xmath2 [ fig .
temperature dependence of the fe@xmath34 and fe@xmath2 occupancies for the block af phase is shown in fig .
5(c ) , behaving again similarly as that of the insulating k@xmath43fe@xmath44se@xmath2 [ fig .
figure 5(d ) plots the temperature dependence of the volume fraction for the block af phase .
we see that the annealing process slightly increases the insulating volume fraction at the expense of the superconducting phase in k@xmath0fe@xmath1se@xmath2 .
comparison of the lattice parameters of the superconducting k@xmath0fe@xmath1se@xmath2 with those of fese under pressure ( table ii ) reveals that the superconducting k@xmath0fe@xmath1se@xmath2 has lattice parameters similar to those of compressed fese @xcite .
the in - plane compression in the second phase and its corresponding @xmath40 fits well with the trend in fese .
this may explain the enhanced @xmath40 of k@xmath0fe@xmath1se@xmath2 compared with fese and suggests that superconductivity in these systems is related .
| c | c | c | c | c | + compound & pressure / temperature & @xmath61 ( ortho ) & @xmath62% & @xmath40 ( k ) + & 0.25 gpa & 5.2952 & 0.0% & 10 + & 4 gpa & 5.1749 & 2.3% & 28 + & 6 gpa & 5.1010 & 3.5% & 37 + & 9 gpa & 5.0835 & 4.0% & 23 + & 12 gpa & 5.0470 & 4.7% & 11 + & 590 k & 5.580 & 0.2% & - + & 15 k & 5.381 & 3.9% & 32 +
in summary , we have used neutron scattering to systematically study insulating and superconducting k@xmath0fe@xmath1se@xmath2 . by carefully comparing the observed diffraction spectra at different temperatures , we conclude that both the insulating and superconducting samples are phase pure in the paramagnetic state with the tetragonal iron pnictide structure .
while the insulating sample with the chemical composition of k@xmath43fe@xmath44se@xmath2 is still phase pure in the low temperature state and forms a @xmath3 block af structure , the superconducting k@xmath56fe@xmath57se@xmath2 phase separates into the insulating k@xmath63fe@xmath64se@xmath2 with block af structure and superconducting k@xmath52fe@xmath5se@xmath2 with a weakly orthorhombic iron pnictide structure .
the crystal structure of the superconducting phase in k@xmath0fe@xmath1se@xmath2 is consistent with the crystal structure of alkali - ammonia intercalated superconducting fese compounds if one neglects the ammonia - molecular constituent @xcite .
this suggests that the superconducting phase in k@xmath0fe@xmath1se@xmath2 and ammonia - intercalated fese compounds has the same crystalline structure .
our temperature dependent measurements of crystal structural and volume hysteresis suggest that phase separation is an intrinsic phenomenon in the iron rich k@xmath0fe@xmath1se@xmath2
. therefore , superconductivity appears in the k@xmath56fe@xmath57se@xmath2 phase with iron pnictide structure , and does not occur in the @xmath13 @xcite or rhombus @xcite iron vacancy ordered phase .
the work at rice university is supported by the us nsf dmr-1308603 and oise-0968226 .
the work at the university of virginia is supported by the u. s. doe , office of basic energy sciences , under contract number de - fg02 - 01er45927 .
this work has benefited from the use of npdf at the lujan center at los alamos neutron science center , funded by doe office of basic energy sciences .
los alamos national laboratory is operated by los alamos national security llc under doe contract de - ac52 - 06na25396 .
the upgrade of npdf has been funded by nsf through grant dmr 00 - 76488 .
work at ustc is supported by most and nsfc .
d. c. johnston , advnaces in physics * 59 * , 803 ( 2010 ) . g. r. stewart , rev . mod
. phys . * 83 * , 1589 ( 2011 ) .
p. c. dai , j. p. hu , and e. dagotto , nature phys . * 8 * , 709 ( 2012 ) .
j. g. guo , s. f. jin , g. wang , s. c. wang , k. x. zhu , t. t. zhou , m. he , and x. l. chen , phys
b * 82 * , 180520(r ) ( 2010 ) . a. krzton - maziopa , z. shermadini , e. pomjakushina , v. pomjakushin , m. bendele , a. amato , r. khasanov , h. luetkens , k. conder , j. phys . : condens .
matter * 23 * , 052203 ( 2011 ) . m. h. fang , h. d. wang , c. h. dong , z. j. li , c. m. feng , j. chen , h. q. yuan , europhys .
lett . * 93 * , 47004 ( 2011 ) .
a. f. wang , j. j. ying , y. j. yan , r. h. liu , x. g. luo , z. y. li , x. f. wang , m. zhang , g. j. ye , p. cheng , z. j. xiang , x. h. chen , phys .
b * 83 * , 060512 ( 2011 ) .
l. h@xmath65ggstr@xmath11 m _ et al .
_ , j. magn .
magn . mater . *
98 * , 37 ( 1991 ) .
e. dagotto , rev .
phys . * 85 * , 849 ( 2013 ) .
w. bao , q. huang , g. f. chen , m. a. green , d. m. wang , j. b. he , x. q. wang , y. qiu , chinese phys . lett . * 28 * , 086104 ( 2011 ) . v. yu .
pomjakushin , d. v. sheptyakov , e. v. pomjakushina , a. krzton - maziopa , k. conder , d. chernyshov , v. svitlyk , z. shermadini , phys .
b * 83 * , 144410 ( 2011 ) .
m. wang , m. y. wang , g. n. li , q. huang , c. h. li , g. t. tan , c. l. zhang , h. b. cao , w. tian , y. zhao , y. c. chen , x. y. lu , b. sheng , h. q. luo , s. l. li , m. h. fang , j. l. zarestky , w. ratcliff , m. d. lumsden , j. w. lynn , and p. c. dai , phys . rev .
b * 84 * , 094504 ( 2011 ) .
a. ricci , n. poccia , g. campi , b. joseph , g. arrighetti , l. barba , m. reynolds , m. burghammer , h. takeya , y. mizuguchi , y. takano , m. colapietro , n. l. saini , a. bianconi , phys .
b * 84 * , 060511(2011 ) .
a. charnukha , a. cvitkovic , t. prokscha , d. pr@xmath11pper , n. ocelic , a. suter , z. salman , e. morenzoni , j. deisenhofer , v. tsurkan , a. loidl , b. keimer , and a.v .
boris , phys .
lett . * 109 * , 017003 ( 2012 ) .
chen , c - c .
chang , h - h .
chang , a .-
fang , c .- h .
wang , w .- h .
chao , c .- m .
tseng , y .- c .
lee , y .- r .
wu , m .- h .
wen , h .- y .
tang , f .-
chen , m .- j .
wang , m .- k .
wu , and d. van dyck , pnas * 111 * , 63 ( 2014 ) . | we use neutron diffraction to study the temperature evolution of the average structure and local lattice distortions in insulating and superconducting potassium iron selenide k@xmath0fe@xmath1se@xmath2 . in the high temperature
paramagnetic state , both materials have a single phase with crystal structure similar to that of the bafe@xmath2as@xmath2 family of iron pnictides .
while the insulating k@xmath0fe@xmath1se@xmath2 forms a @xmath3 iron vacancy ordered block antiferromagnetic ( af ) structure at low - temperature , the superconducting compounds spontaneously phase separate into an insulating part with @xmath3 iron vacancy order and a superconducting phase with chemical composition of k@xmath4fe@xmath5se@xmath2 and bafe@xmath2as@xmath2 structure .
therefore , superconductivity in alkaline iron selenides arises from alkali deficient k@xmath4fe@xmath5se@xmath2 in the matrix of the insulating block af phase . |
during last few years many exciting experimental results and theory development are obtained in sidis on the transversely polarized target . up to now
only the measurements @xcite of sivers and collins asymmetries were performed by hermes and compass collaborations and together with data from belle @xcite they allow a first extraction the transversity and sivers transverse momentum dependent ( tmd ) distribution functions ( dfs ) and collins fragmentation function ( ff ) .
in addition to these , the general expression of sidis cross section @xcite contains six more target transverse polarization dependent azimuthal asymmetries . here
we present the preliminary results on these asymmetries for the first time measured by compass from the 2002 - 2004 data . in the following the notations of ref .
@xcite are used .
there are eight azimuthal modulations related to the target transverse polarization : @xmath2 where first two correspond to sivers and collins effects .
the expression for the cross section in interest can be represented as @xmath3 where @xmath4 is the target transverse polarization .
we factored out the explicitly calculable depolarization factors , @xmath5 , and defined the asymmetries as the ratios of corresponding structure functions to unpolarized one : @xmath6 where @xmath7 or @xmath8 corresponds to beam polarization dependent or independent part of asymmetry .
the depolarization factors entering in eq .
( [ eq : cros - sect - short ] ) depend only on @xmath9 and are given as @xmath10 the asymmetries extracted from the data as amplitudes of corresponding azimuthal modulations ( raw asymmetries ) are then given by @xmath11 where @xmath12 is the target polarization dilution factor . in the qcd parton model the asymmetries @xmath13 and @xmath14
are given by the ratio of convolutions of spin - dependent to spin - independent twist two dfs and ffs , for example @xmath15 and can be used for extraction of dfs @xmath16 and @xmath17 describing the quark longitudinal and transverse ( along quark transverse momentum ) polarization in the transversely polarized nucleon .
the other asymmetries can be interpreted as cahn kinematic corrections to spin effects on the transversely polarized nucleon @xcite , for example : @xmath18
the event selection and asymmetry extraction are done as described in @xcite .
the following kinematic cuts were imposed : @xmath19 ( gev / c)@xmath20 , @xmath21 gev , @xmath22 , @xmath23 gev / c and @xmath24 . in figs .
[ fig:1 ] and [ fig:2 ] for the first time we present six target transverse spin dependent asymmetries extracted from compass 20022004 data collected on deuterium target .
the estimated systematic errors are smaller than statistical
. all six newly measured asymmetries are compatible with zero within statistical errors .
v. y. alexakhin _ et al . _ [ compass collaboration ] , phys .
lett . *
94 * , 202002 ( 2005 ) [ arxiv : hep - ex/0503002 ] .
e. s. ageev _ et al .
_ [ compass collaboration ] , nucl .
b * 765 * , 31 ( 2007 ) [ arxiv : hep - ex/0610068 ] ; r. seidl _ et al .
_ [ belle collaboration ] , phys .
lett . * 96 * , 232002 ( 2006 ) [ arxiv : hep - ex/0507063 ] .
a. kotzinian , nucl .
b * 441 * , 234 ( 1995 ) [ arxiv : hep - ph/9412283 ] .
a. bacchetta , m. diehl , k. goeke , a. metz , p. mulders and m. schlegel , jhep * 0702 * , 093 ( 2007 ) [ arxiv : hep - ph/0611265 ] . | in semi - inclusive dis of polarized leptons on a transversely polarized target eight azimuthal modulations appear in the cross - section . within qcd parton model four azimuthal asymmetries
can be interpreted at leading order , two of them being the already measured collins and sivers asymmetries .
the other two leading twist asymmetries , related to different transverse momentum dependent quark distribution functions , and also additional four asymmetries which can be interpreted as twist - three contributions have been measured for the first time at compass , using a 160 gev / c longitudinally polarized ( @xmath0 ) muon beam and a transversely polarized @xmath1 target .
we present here the preliminary results from the 2002 - 2004 data . |
hawking radiation is the remarkable property of black hole : it occurs due to creation of a particle - antiparticle pair nearby the black hole edge @xcite .
if one of the pair constituents crosses the event horizon , it never returns , thus giving rise to the emission from the black hole , which in its turn leads to decreasing of black hole energy and mass ( for review see , e.g. , ref.@xcite ) .
nevertheless , the main difficulty in the experimental confirmation of hawking radiation is the big mass of known black holes in universe , and , hence , weak intensity of hawking radiation .
that is why the reproducing of black holes in laboratory conditions are of great interest .
from one side , small artificial black holes can be created in large hadron collider @xcite . from the other side ,
the phenomena , originated by gravity , are reproducible ( through the similarity of wave processes ) in number of other physical systems , called analogue gravity models @xcite .
in particular , black holes and hawking radiation can be obtained in electromagnetic wave waveguide @xcite , slow light in moving medium @xcite or in optical fibers @xcite ( actually , first experimental evidence of hawking radiation was observed in this system @xcite , although the nature of observed phenomena is also discussable @xcite ) , surface waves on moving water @xcite , the moving flow with a gradient from subsonic to supersonic flow ( so called acoustic or sonic black hole @xcite ) . sonic black holes , after being predicted in 1981 , later was implemented in quantum liquids like liquid he @xcite , atomic @xcite and polaritonic @xcite bose - einstein condensates ( bec ) .
moreover , reproducing in bec a system with inner horizon ( analogue of charged black hole @xcite ) , it is possible to achieve experimentally a black hole lasing , i.e. existence of self - amplifying hawking radiation @xcite .
one of factors , which determines a practical interest to bec ( subjected to linear periodic potential optical lattice(ol ) ] , is the possibility to use it as almost perfect testbench , which allows to reproduce ( without impurities and defects ) a lot of phenomena , known from condensed matter ( for review see , e.g. @xcite ) .
nevertheless , due to strong two - body interactions bec is essentially nonlinear system : the fact , which is hardly to ignore . a natural way to take into account the nonlinearity is to consider bec matter waves in the form of localized wavepackets solitons , which can propagate at long distances without loosing their shape .
moreover , loading the bec in ol allows one to control the dispersion ( normal and anomalous ) , and , as distinct from the case without ol , allows the existence of solitons either in the case of attractive as repulsive interaction ( for review see ref.@xcite ) . besides the intensive theoretical study ,
the creation of bright gap solitons in repulsively interacting bec was demonstrated experimentally @xcite .
binary mixtures of bose - einstein condensates ( becs ) can support vector ( multicomponent ) solitons with variety of interesting properties .
the possibility for two - component bec to be loaded into ol was experimentally demonstrated in ref.@xcite .
being loaded into the ol , two - component bec can sustain bright - bright and dark - bright stationary solitons @xcite , mixed - symmetry modes and breathers @xcite .
nevertheless , experimental study of vector solitons @xcite was restricted only to the case of dark - bright solitons in two - component bec without ol . in the present paper
we consider emission of matter wave solitons from two - component bec , loaded into parabolic trap with optical lattice .
we use periodic sign - varying linear coupling term to create vector soliton , which is characterized by first- and second - component chemical potentials located in the vicinity of the lower and upper edges of the ol spectrum first band , respectively . after being loaded into the parabolic trap ,
the first component of vector soliton ( characterized by positive effective mass nearby the bottom edge of first band ) oscillates periodically in the trap , while second component ( characterized by negative effective mass at top edge of the first band ) , is accelerated by parabolic trap and escapes from it . in general ,
the nature of investigated phenomena is similar to hawking emission , but at the same time has significant differences .
firstly , in our case the vector soliton with opposite masses of components ( analogue of particle - antiparticle pair ) is created by transferring atoms from first to second components of bec .
in conventional hawking emission , by contrast , the particle - antiparticle pair is formed from vacuum due to quantum fluctuations .
secondly , in our phenomenon the event horizon is defined somehow artificially ( as an edge of potential well ) .
that is why we prefer to call our phenomenon _ hawking - like _ emission ( in analogy with ref.@xcite ) .
the paper is organized as follows . in sec.[sec : model ] we expose and describe the model of our problem ( problem statement ) . in sec.[sec : gap - sol ] we represent the stationary state of bright - bright solitons in the two - component bec and propose a mechanism to create this state . in sec.[sec :
h - emis ] we describe the dynamics of two - component soliton in parabolic trap and the mechanism of hawking - like escape of the soliton from the trap .
to be specific , we consider a spinor bec composed of two hyperfine states , say of the @xmath0 and @xmath1 states of @xmath2rb atoms @xcite confined at different vertical positions by transverse parabolic traps and loaded into the optical lattice of cos - like shape .
additionally , a time - dependent external potential @xmath3 ( which is aperiodic in general case ) is applied to the condensate . at the same time
hyperfine states are coupled by a coordinate and time - dependent coupling field @xmath4 , which describes the possibility of conversion of atoms between states ( such kind of coupling can be originated by the external magnetic field ) .
we assume the condensate to be quasi - one - dimensional ( cigar - shaped ) .
then , in the mean - field approximation the system is described by the gp equations @xcite [ eq : tc ] @xmath5 equations ( [ eq : tc ] ) are written in dimensionless form : all energy values ( like ol amplitude @xmath6 ) are measured in recoil energy units of @xmath7 ( where @xmath8 is the atomic mass and @xmath9 is the ol period ) , while the coordinate @xmath10 and time @xmath11 are measured in units of @xmath12 and @xmath13 , respectively . at the same time
wavefunctions @xmath14 are measured in the @xmath15 units , where @xmath16 is the transverse [ in the @xmath17-plane ] oscillator length , @xmath18 is the inter - species s - wave scattering length [ thus , @xmath19 in ( [ eq : tc ] ) ] . in the absence of nonlinearity
( formally @xmath20 ) , additional external force @xmath21 and coupling term @xmath22 , periodicity of ol gives rise to band - gap structure of both spectrum @xmath23 and bloch functions @xmath24 of the linear problem @xmath25 , where @xmath26 is the band number and @xmath27 is the bloch wavenumber in the first brillouin zone , @xmath28 $ ] .
meanwhile both the spectrum and the bloch functions are periodic in the reciprocal space with period @xmath29 ( in the chosen units ) : @xmath30 , @xmath31 .
in further considerations we take into account the first band only ( @xmath32 ) , whose bottom and top correspond to @xmath33 and @xmath34 , respectively , so the index @xmath26 will be omitted .
[ cols="^ " , ] moreover , we can start from the situation when all the bec atoms are initially concentrated in the first component [ like eq.([eq : stsi ] ) ] , and then apply ( for the finite time ) the spatially - periodic linear coupling , which will transfer a portion of atoms to the second component .
in other words , we use the same method , as described in sec.[sec : gap - sol ] , but apply it not to stationary soliton , but to the soliton oscillating inside the finite - width potential well [ see fig.[fig : smooth](c ) ] . in this case atoms , transferred to the second component , will constitute the gap soliton with negative effective mass , which in its turn will escape from the finite - width potential well , as demonstrated in fig.[fig : smooth](d ) .
thus , we described a mechanism of stimulated emission of matter waves ( in form of bright solitons ) from the two - component bec , loaded into the ol , which is combined with the external parabolic potential .
the similarity between the hawking emission from the black hole and the soliton escape from the parabolic trap is defined by the fact , that we use the bright - bright soliton , where chemical potentials of first and second bec component lie nearby the opposite edges of the first band of ol spectrum . as a consequence , signs of the effective masses ( which characterize the bec first and second components )
are also opposite , and such type of low - amplitude soliton can be considered as an analogue of the particle - antiparticle pair in hawking emission .
we demonstrated , that this low - amplitude bright - bright two - component soliton can be created by partial transferring of atoms from one to another bec component , using spatially - periodic linear coupling term , whose period equals to the double ol period .
being loaded into the finite - width parabolic trap , one component of such bright - bright soliton ( characterized by the positive effective mass ) exerts periodic oscillations nearby the trap center , while another component ( with negative effective mass ) is gradually accelerated and moves in the direction of parabolic trap growth .
if soliton with negative effective mass passes the finite - width parabolic trap edge ( which is the event horizon analogue ) , it escapes from the trap and never returns to the initial point .
m.a.g .- n . thanks for the financial support of fondecyt project 11130450 . d. n. page , new j. phys . * 7 * , 203 ( 2005 ) s. dimopoulos , g. landsberg , phys . rev . lett . * 87 * , 161602 ( 2001 ) ; s c. park , prog . in particle and nuclear phys . *
67 * , 617 ( 2012 ) . c. barcel , s. liberati , m. visser , living rev . relativity * 14 * , 3 ( 2011 ) .
r. schtzhold , w. g. unruh , phys .
lett . * 95 * , 031301 ( 2005 ) .
d. faccio , _ et al .
_ , europhys .
lett . * 89 * , 34004 ( 2010 ) ; t. g. philbin , c. kuklewicz , s. robertson , s. hill , f. knig , u. leonhardt , science * 319 * , 1367 ( 2008 ) .
f. belgiorno , _ et al .
lett . * * 7**105 , 203901 ( 2010 ) ; e. rubino , _ et al .
_ , new j. phys . * 13 * , 085005 ( 2011 ) .
w. unruh , r. schtzhold , phys .
d * 86 * , 064006 ( 2012 ) .
g. rousseaux , c. mathis , p. massa , t. g. philbin , u. leonhardt , new j. phys .
* 10 * , 053015 ( 2008 ) ; s. weinfurtner , e. w. tedford , m. c. j. penrice , w. g. unruh , g. a. lawrence , phys .
. lett . * 106 * , 021302 ( 2011 ) .
l. j. garay , j. r. anglin , j. i. cirac , p. zoller , phys .
lett . * 85 * , 4643 ( 2000);. c. barcel , s. liberati , m. visser , class .
. grav . * 18 * , 1137 ( 2001 ) ; s. liberati , m. visser , s. weinfurtner , class .
* 23 * , 3129 ( 2006 ) ; a. recati , n. pavloff , i. carusotto , phys . rev .
a * 80 * , 043603 ( 2009 ) ; i. zapata , m. albert , r. parentani , f. sols , new j. phys .
* 13 * , 063048 ( 2011 ) ; o. lahav , _ et al .
lett . * 105 * , 240401 ( 2010 ) .
d. d. solnyshkov , h. flayac , g. malpuech , phys .
b * 84 * , 233405 ( 2011 ) .
s. corley , t. jacobson , phys .
d * 59 * , 124011 ( 1999 ) d. faccio , t. arane , m. lamperti , u. leonhardt , classical and quantum gravity * 29 * , 224009 ( 2012 ) . s. finazzi , r. parentani , new j. phys .
* 12 * , 095015 ( 2010 ) .
j. steinhauer , nat .
* 10 * , 864 ( 2014 ) . m. lewenstein , _ et al . _ ,
advances in physics , * 56 * , 243 ( 2007 ) .
v. a. brazhnyi , v. v. konotop , modern physics letters b * 18 * , 627 ( 2004);r .
carretero - gonzlez , d. j. frantzeskakis , p. g. kevrekidis , nonlinearity * 21 * , r139 ( 2008 )
. b. eiermann , _ et al .
lett . * 91 * , 060402 ( 2003 ) ; b. eiermann , _ et al .
lett . * 92 * , 230401 ( 2004 ) .
j. m. higbie , l. e. sadler , s. inouye , a. p. chikkatur , s. r. leslie , k. l.moore , v. savalli , and d.m .
stamper - kurn , phys .
lett . * 95 * , 050401 ( 2005 ) ; a. widera , f. gerbier , s. folling , t. gericke , o.mandel , and i. bloch , phys . rev . lett . * 95 * , 190405 ( 2005 ) .
e. a. ostrovskaya , y. s. kivshar , phys .
lett . * 92 * , 180405 ( 2004 ) .
h. cruz , v. a. brazhnyi , v. v. konotop , g. l. alfimov , m. salerno , phys . rev .
a * 76 * , 013603 ( 2007 ) . c. becker ,
_ et al . _ , nat .
* 4 * , 496 ( 2008);s .
middelkamp , _ et al .
a * 375 * , 642 ( 2011);c .
hamner , _ et al .
lett . * 106 * , 065302 ( 2011);c .
hamner , _ et al .
rev . lett . * 111 * , 264101 ( 2013 ) .
j. a. gonzlez , m. a. garca - ustes , , a. snchez , p. v. e. mcclintock , new j. phys . * 10 * , 113015 ( 2008 ) .
m. r. matthews , _ et al .
* 83 * , 3358 ( 1999 ) ; j. williams , et al .
a * 61 * , 033612 ( 2000 ) .
l. pitaevskii and s. stringari , _ bose - einstein condensation _
( oxford university press , 2003 ) . c. j. pethick and h. smith , _ bose - einstein condensation in dilute gases _ ( cambridge university press , cambridge , england , 2002 ) . v.
v. konotop and m. salerno , phys rev . a * 65 * , 021602 ( 2002 ) m. salerno , v.v .
konotop , and yu.v .
bludov , phys .
* 101 * , 030405 ( 2008 ) ; yu .
v. bludov , v. v. konotop , m. salerno , europhys . lett . * 87 * , 20004 ( 2009 ) . | we demonstrate , that bose - einstein condensate can escape from the trap , formed of combined linear periodic ( optical lattice ) and parabolic potentials , and the escaping mechanism is similar to hawking radiation from black hole .
the low - amplitude bright - bright soliton in two - component bose - einstein condensate ( where chemical potentials of the bec first and second components are located nearby the opposite edges of the first band of the optical lattice spectrum ) serves as an analogue of particle - antiparticle pair in hawking radiation .
it is shown that parabolic potential , being applied to such two - component bec , leads to spatial separation of its components : bec component with chemical potential located in semi - infinite gap exerts the periodical oscillations , while the bec component , whose chemical potential is in the first finite gap , escapes from the trap ( due to negative effective mass of gap soliton ) . we also propose a method for the creation of such bright - bright soliton transferring of atoms from one bec component to another by spatially periodic linear coupling term . |
it is believed that in spite of the recent potential importance of magnetic monopoles@xcite,@xcite,@xcite and dyons @xcite towards the quark confinement problem @xcite of quantum choromodynamics , possible magnetic condensation of vacuum @xcite , cp - violation @xcite , their role as catalyst in proton decay @xcite and the current grand unified theories @xcite , the formalism necessary to describe them has been clumsy and manifestly non - covariant .
keeping in view the potential importance of monopoles and the results of witten @xcite that monopoles are necessarily dyons , we @xcite have also constructed a self - consistent co - variant theory of generalized electromagnetic fields associated with dyons each carrying the generalized charge as complex quantity with its real and imaginary part as electric and magnetic constituents . on the other hand quaternions
were invented by hamilton @xcite to extend the theory of complex numbers to three dimensions .
maxwell s equations of electromagnetism were rewritten in terms of quaternions @xcite .
finklestein et al @xcite developed the quaternionic quantum mechanics and adler @xcite described the theory of the algebraic structure of quantum choromodynamics for strong interactions .
various aspects of quaternions are discussed by morita @xcite towards the kinematical structure of poincare gauge theory and the left - right weinberg - salam theory of quantum choromodynamics .
we have also studied @xcite the quaternionic formulation for generalized field equations of dyons in unique , simpler and compact notations .
quaternion non - abelian gauge theory has also been consistently discussed @xcite to maintain the structural symmetry between the theory of linear gravity and electromagnetism .
it is also shown that quaternion formalism characterises the abelian and non - abelian gauge structures @xcite of dyons in terms of real and imaginary constituents of quaternion basis elements .
altenatively , kravchenko @xcite and his coworkers have consistently analysed the maxwell s equations for time - dependent electromagnetic fields in homogeneous ( isotropic ) and chiral medium . extending this ,
we @xcite have also derived the generalised maxwells - dirac equation in the homogenous ( isotropic ) medium .
it has been shown that the field equations of dyons also remains invariant under the duality transformations in isotropic homogeneous medium and the equation of motion reproduces the rotationally symmetric gauge invariant angular momentum of dyons . in order to extend the theory of monopoles and dyons in isotropic medium and consequently the relevance of quaternion formalism of dyons , in the present paper
we have undertaken the study of the quaternion analysis of time dependent maxwell s equations in presence of electric and magnetic charges and the solution for the classical problem of moving sources are obtained in unique , simpler and consistent manner .
quaternion forms of potential current , field equation and equation of motion are developed in compact manner and it is imphasized that the quantum equations in terms of quaternions are invariant under quaternion , lorentz and duality transformations .
it has also been discussed that the quaternion analyticity of dyons in isotropic medium reproduces the results of kravchenko @xcite in the absence of magnetic monopole and accordingly this theory can be described symmetrically for pure monopole in the absence of electric charge or vice versa .
assuming the existence of magnetic monopoles , we may write the following form @xcite of symmetric generalized maxwell - dirac differential equations @xcite in free space in si units @xmath2@xmath3 as ; @xmath4 where @xmath5 and @xmath6 are respectively the electric and magnetic charge densities while @xmath7 and @xmath8 are the corresponding current densities , @xmath9 is electric induction vector , @xmath10 is electric field , @xmath11 is magnetic induction vector and @xmath12 is magnetic field .
here we asume the homogenous ( isotropic ) medium with the following definitions @xcite , @xmath13 and @xmath14 where @xmath15 the free space permitivity , @xmath16 is the permeability of free space and @xmath17 and @xmath18 are defined respectively as relative permitivity and permeability in electric and magnetic fields . on using equations ( [ eq:2 ] ) and ( [ eq:3 ] ) , equation ( [ eq:1],[eq:2],[eq:3],[eq:4 ] ) takes the following differential form @xcite , @xmath19 differential equations ( [ eq:1],[eq:2],[eq:3],[eq:4 ] ) are the generalised field equations of dyons in homogenous medium and the electric and magnetic fields are corresponding called generalised electromegnetic fields of dyons .
these electric and magnetic fields of dyons are expresed in following differential form in homogenous medium in terms of two potentials @xciteas , @xmath20 where @xmath21 and @xmath22 are the two four - potentials associated with electric and magnetic charges .
let us define the complex vector field @xmath23 in the following form @xmath24 equations ( [ eq:5],[eq:6 ] ) and ( [ eq:7 ] ) lead to the following relation between generalized field and the components of generalized four - potential as , @xmath25 where @xmath26 is the generalised four - potential of dyons in homogenous medium and defined as @xmath27 i.e. @xmath28 and @xmath29 maxwell s field equation ( [ eq:7],[eq:8],[eq:9],[eq:10 ] ) may then be written in terms of generalized field @xmath23as
@xmath30 where @xmath31 and @xmath32 are the generalized charge and current source densities of dyons in homogenous medium given by @xmath33 using equation ( [ eq:12 ] ) we introduce a new parameter@xmath34 ( i.e. the field current ) as @xmath35 where @xmath36is the dalembertian operator and expressed as @xmath37 where @xmath38 is the speed of electromagnetic wave in homogenous isotropic medium . in terms of complex potential the field equation is written as @xmath39 we write the following tensorial form of generalized maxwell s -dirac equations of dyons in homogenous medium as @xmath40 defining generalized field tensor of dyon as @xmath41 one can directly obtain the following generalized field equation of dyon in homogenous ( isotropic ) medium i.e. @xmath42 the lorentz four - force equation of motion for dyons in homogenous ( isotropic ) medium as @xmath43 where @xmath44 donetes real part , @xmath45 is the four - acceleration and @xmath46is the four - velocity of the particle and @xmath47 is the generalized charge of dyon in isotropic medium .
a quaternion is defined as @xmath48 where @xmath49 are real numbers and called the components of the quaternion @xmath50 and the quaternion units @xmath51 satisfy the following multiplication rules ; @xmath52 where @xmath53 and @xmath54 ( j , k , l= 1,2,3 and @xmath55 ) are respectively the kronecker delta and three - index levi - civita symbol .
the sum of two quaternions p and q is defined by adding the corresponding components , @xmath56 and the multiplication of two quaternions is defined as ; @xmath57 thus each quaternion @xmath50 is the sum of a scalar @xmath58 and a vector @xmath59 , @xmath60 where @xmath61 thus the product of two quaternions @xmath62 and @xmath50 is also written as @xmath63 where the dot and cross indicate , respectively , the usual three dimensional scalar and vector products . for any quaternion
, there exists a quaternion conjugate @xmath64 quaternion conjugate is an automorphism of ring of quaternion i.e. @xmath65 the norm of a quaternion is given as @xmath66 the inverse of a quaternion @xmath50 is also a quaternion @xmath67 let us define the quaternionic form of differential operator as@xcite , @xmath68 and @xmath69 where @xmath70 as such we can express the quantum equation associated with generalized four - potential , four current , electric field and magnetic field in terms of quaternionic analysis as @xmath71 operating equation ( [ eq:37 ] ) to equation ( [ eq:7 ] ) and using equation ( [ eq:27 ] ) , we get @xmath72 similarly we may operate equation ( [ eq:36 ] ) to equation ( [ eq:39 ] ) and on using equation ( [ eq:27 ] ) we get @xmath73 where we have used the following subsidiary condition @xmath74 and @xmath75 equation ( [ eq:44 ] ) is known as lorentz gauge condition while equation ( [ eq:45 ] ) is referred as continuity equation . @xmath23 and @xmath34 are defined as quaternion valued vector functions in the following manner @xmath76 similarly , we may obtain the quaternion conjugate field equations for dyons in homogenous ( isotropic ) medium as @xmath77 where @xmath78 and @xmath79 are the quaternion conjugates defined as @xmath80 hence the quaternion forms of equation ( [ eq:18],[eq:19 ] ) for generalized potential , equation ( [ eq:23],[eq:24 ] ) for generalized maxwell s dirac equation and lorentz force equation ( [ eq:25 ] ) for dyons in homogeneous medium may be expressed in terms of the following set of quaternion equations in simple , compact and consistent manner .
@xmath81 & = & j\label{eq:55}\\ { } [ \boxdot , j\ , ] & = & 0\label{eq:56}\\ { } [ u , g\ , ] & = & f\label{eq:57}\end{aligned}\ ] ] where @xmath82 in equation ( [ eq:57 ] ) , @xmath83 , @xmath84 and @xmath85 are the quaternionic forms of velocity , generalized field tensor and lorentz force associated with dyons in homogeneous medium .
the elements of g in equation ( [ eq:59 ] ) are the following quaternion forms , @xmath86 equations ( [ eq:55 ] , [ eq:56 ] ) and ( [ eq:57 ] ) may also be described as @xmath87 & = & j_{\mu}\label{eq:64}\end{aligned}\ ] ] and @xmath88 & = & f_{\mu}\label{eq:65}\end{aligned}\ ] ] where @xmath89 and @xmath90 are the four current and four force associated with generalized field of dyons in homogeneous ( isotropic ) medium .
let us factorize the wave operator as the combination of quaternion and its conjugate in the following manner@xcite , @xmath91 moreover , each solution of the wave equation @xmath92 can be written in a simple and compact form as the sum of two functions @xmath93 and @xmath94 , which reduces to following set of differential equations @xmath95 respectively for electric and magnetic charges of a particle described as dyons .
let us define the generalized charge of dyons moving in generalized electromagnetic field as ; where @xmath97 is electronic charge and @xmath98 is magnetic charge .
let us assume that a dyon is moving with a velocity @xmath99 .
thus , the electric charge density @xmath5 of a dyon may be expressed as , similarly as the case of electron , let us describe that a monopole constituent of a dyon is also moving with a velocity @xmath99 .
then the magnetic charge density @xmath6 and magnetic current density @xmath8 in view of duality transformations @xcite lead to the following expressions i.e. these expressions , associated with the solutions of generalized field equations of dyons , reduce to the solutions of usual elecrtic and magnetic field vectors in the absence of magnetic ( electric ) charge on dyons similar to those described by kravchenko @xcite or vice versa .
kravchenko , v.v . , " applied quaternionic analysis , research and exposition in mathematics " , heldermann press , germany , 28 ( 2003 ) and reference therein ; " quaternionic equation for electromagnetic fields in inhomogeneous media " , math - ph/0202010 ; sergei m. grudsky , kira v. khmelnytskaya , vladislav v. kravchenko , " on a quaternionic maxwell equation for the time - dependent electromagnetic field in a chiral media " , math - ph/0309062 . | quaternion analysis of time dependent maxwell s equations in presence of electric and magnetic charges has been developed and the solutions for the classical problem of moving charges ( electric and magnetic ) are obtained in unique , simple and consistent manner .
@xmath0department of physics + govt .
p. g. college + pithoragarh -(ua ) , india @xmath1department of physics + kumaun university + s. s. j. campus + almora-263601 ( ua ) , india email:-@[email protected] + @[email protected] + @xmath1 [email protected] . |
networks have attracted considerable recent attention in physics and other fields as a foundation for the mathematical representation of a variety of complex systems , including biological and social systems , the internet , the worldwide web , and many others @xcite .
a common feature of many networks is `` community structure , '' the tendency for vertices to divide into groups , with dense connections within groups and only sparser connections between them @xcite .
social networks @xcite , biochemical networks @xcite , and information networks such as the web @xcite , have all been shown to possess strong community structure , a finding that has substantial practical implications for our understanding of the systems these networks represent .
communities are of interest because they often correspond to functional units such as cycles or pathways in metabolic networks @xcite or collections of pages on a single topic on the web @xcite , but their influence reaches further than this .
a number of recent results suggest that networks can have properties at the community level that are quite different from their properties at the level of the entire network , so that analyses that focus on whole networks and ignore community structure may miss many interesting features .
for instance , in some social networks one finds individuals with different mean numbers of contacts in different groups ; the individuals in one group might be gregarious , having many contacts with others , while the individuals in another group might be more reticent .
an example of this behavior is seen in networks of sexual contacts , where separate communities of high- and low - activity individuals have been observed @xcite .
if one were to characterize such a network by quoting only a single figure for the average number of contacts an individual has , one would be missing features of the network directly relevant to questions of scientific interest such as epidemiological dynamics @xcite .
it has also been shown that vertices positions within communities can affect the role or function they assume . in social networks ,
for example , it has long been accepted that individuals who lie on the boundaries of communities , bridging gaps between otherwise unconnected people , enjoy an unusual level of influence as the gatekeepers of information flow between groups @xcite .
a surprisingly similar result is found in metabolic networks , where metabolites that straddle the boundaries between modules show particular persistence across species @xcite .
this finding might indicate that modules in metabolic networks possess some degree of functional independence within the cell , allowing vertices central to a module to change or disappear with relatively little effect on the rest of the network , while vertices on the borders of modules are less able to change without affecting other aspects of the cellular machinery .
one can also consider the communities in a network themselves to form a higher level meta - network , a coarse - grained representation of the full network .
such coarse - grained representations have been used in the past as tools for visualization and analysis @xcite but more recently have also been investigated as topologically interesting entities in their own right .
in particular , networks of modules appear to have degree distributions with interesting similarities to but also some differences from the degree distributions of other networks @xcite , and may also display so - called preferential attachment in their formation @xcite , indicating the possibility of distinct dynamical processes taking place at the level of the modules . for all of these reasons and others besides there has been a concerted effort in recent years within the physics community and elsewhere to develop mathematical tools and computer algorithms to detect and quantify community structure in networks .
a huge variety of community detection techniques have been developed , based variously on centrality measures , flow models , random walks , resistor networks , optimization , and many other approaches @xcite .
for reviews see refs .
@xcite .
in this paper we focus on one approach to community detection that has proven particularly effective , the optimization of the benefit function known as `` modularity '' over the possible divisions of a network .
methods based on this approach have been found to produce excellent results in standardized tests @xcite .
unfortunately , exhaustive optimization of the modularity demands an impractically large computational effort , but good results have been obtained with various approximate optimization techniques , including greedy algorithms @xcite , simulated annealing @xcite , and extremal optimization @xcite . in this paper
we describe a different approach , in which we rewrite the modularity function in matrix terms , which allows us to express the optimization task as a spectral problem in linear algebra .
this approach leads to a family of fast new computer algorithms for community detection that produce results competitive with the best previous methods .
perhaps more importantly , our work also leads to a number of useful insights about network structure via the close relations we will demonstrate between communities and matrix spectra .
our work is by no means the first to find connections between divisions of networks and matrix spectra .
there is a large literature within computer science on so - called spectral partitioning , in which network properties are linked to the spectrum of the graph laplacian matrix @xcite .
this method is different from the one introduced here and is not in general well suited to the problem of community structure detection .
the reasons for this , however , turn out to be interesting and instructive , so we begin our presentation with a brief review of the traditional spectral partitioning method in section [ specpart ] .
a consideration of the deficiencies of this method in section [ modsec ] leads us in sections [ specmod][multiway ] to introduce and develop at length our own method , which is based on the characteristic matrix we call the `` modularity matrix . '' sections [ negative ] and [ otheruses ] explore some further ideas arising from the study of the modularity matrix but not directly related to community detection . in section [ concs ]
we give our conclusions .
a brief report of some of the results described in this paper has appeared previously as ref .
there is a long tradition of research in computer science on graph partitioning , a problem that arises in a variety of contexts , but most prominently in the development of computer algorithms for parallel or distributed computation .
suppose a computation requires the performance of some number @xmath0 of tasks , each to be carried out by a separate process , program , or thread running on one of @xmath1 different computer processors .
typically there is a desired number of tasks or volume of work to be assigned to each of the processors . if the processors are identical , for instance , and the tasks are of similar complexity , we may wish to assign the same number of tasks to each processor so as to share the workload roughly equally .
it is also typically the case that the individual tasks require for their completion results generated during the performance of other tasks , so tasks must communicate with one another to complete the overall computation .
the pattern of required communications can be thought of as a network with @xmath0 vertices representing the tasks and an edge joining any pair of tasks that need to communicate , for a total of @xmath2 edges .
( in theory the amount of communication between different pairs of tasks could vary , leading to a _ weighted _ network , but we here restrict our attention to the simplest unweighted case , which already presents interesting challenges . ) normally , communications between processors in parallel computers are slow compared to data movement within processors , and hence we would like to keep such communications to a minimum . in network terms
this means we would like to divide the vertices of our network ( the processes ) into groups ( the processors ) such that the number of edges between groups is minimized .
this is the graph partitioning problem .
problems of this type can be solved exactly in polynomial time @xcite , but unfortunately the polynomial in question is of leading order @xmath3 , which is already prohibitive for all but the smallest networks even when @xmath1 takes the smallest possible value of 2 . for practical applications , therefore , a number of approximate solution methods have been developed that appear to give reasonably good results .
one of the most widely used is the spectral partitioning method , due originally to fiedler @xcite and popularized particularly by pothen _ et al .
_ @xcite .
we here consider the simplest instance of the method , where @xmath4 , i.e. , where our network is to be divided into just two non - intersecting subsets such that the number of edges running between the subsets is minimized .
we begin by defining the adjacency matrix @xmath5 to be the matrix with elements @xmath6 ( we restrict our attention in this paper to undirected networks , so that @xmath5 is symmetric . )
then the number of edges @xmath7 running between our two groups of vertices , also called the cut size , is given by @xmath8 where the factor of @xmath9 compensates for our counting each edge twice in the sum .
to put this in a more convenient form , we define an index vector @xmath10 with @xmath0 elements @xmath11 ( note that @xmath10 satisfies the normalization condition @xmath12 . )
then @xmath13 which allows us to rewrite eq . as @xmath14 noting that the number of edges @xmath15 connected to a vertex @xmath16also called the degree of the vertex is given by @xmath17 the first term of the sum in is @xmath18 where we have made use of @xmath19 ( since @xmath20 ) , and @xmath21 is 1 if @xmath22 and zero otherwise .
thus @xmath23 we can write this in matrix form as @xmath24 where @xmath25 is the real symmetric matrix with elements @xmath26 , or equivalently @xmath27 @xmath25 is called the laplacian matrix of the graph or sometimes the admittance matrix .
it appears in many contexts in the theory of networks , such as the analysis of diffusion and random walks on networks @xcite , kirchhoff s theorem for the number of spanning trees @xcite , and the dynamics of coupled oscillators @xcite .
its properties are the subject of hundreds of papers in the mathematics and physics literature and are by now quite well understood . for our purposes , however , we will need only a few simple observations about the matrix to make progress .
our task is to choose the vector @xmath10 so as to minimize the cut size , eq . .
let us write @xmath10 as a linear combination of the normalized eigenvectors @xmath28 of the laplacian thus : @xmath29 , where @xmath30 and the normalization @xmath12 implies that @xmath31 then @xmath32 where @xmath33 is the eigenvalue of @xmath25 corresponding to the eigenvector @xmath28 and we have made use of @xmath34 . without loss of generality , we assume that the eigenvalues are labeled in increasing order @xmath35 . the task of minimizing @xmath7 can then be equated with the task of choosing the nonnegative quantities @xmath36 so as to place as much as possible of the weight in the sum in the terms corresponding to the lowest eigenvalues , and as little as possible in the terms corresponding to the highest , while respecting the normalization constraint .
the sum of every row ( and column ) of the laplacian matrix is zero : @xmath37 where we have made use of .
thus the vector @xmath38 is always an eigenvector of the laplacian with eigenvalue zero .
it is less trivial , but still straightforward , to demonstrate that all eigenvalues of the laplacian are nonnegative .
( the laplacian is symmetric and equal to the square of the edge incidence matrix , and hence its eigenvalues are all the squares of real vectors . )
thus the eigenvalue 0 is always the smallest eigenvalue of the laplacian and the corresponding eigenvector is @xmath39 , correctly normalized . given these observations it is now straightforward to see how to minimize the cut size @xmath7 . if we choose @xmath40 , then all of the weight in the final sum in eq .
is in the term corresponding to the lowest eigenvalue @xmath41 and all other terms are zero , since @xmath38 is an eigenvector and the eigenvectors are orthogonal .
thus this choice gives us @xmath42 , which is the smallest value it can take since it is by definition a nonnegative quantity .
unfortunately , when we consider the physical interpretation of this solution , we see that it is trivial and uninteresting .
given the definition of @xmath10 , the choice @xmath40 is equivalent to placing all the vertices in group 1 and none of them in group 2 .
technically , this is a valid division of the network , but it is not a useful one . of course the cut size is zero if we put all the vertices in one of the groups and none in the other , but such a trivial solution tells us nothing about how to solve our original problem .
we would like to forbid this trivial solution , so as to force the method to find a nontrivial one .
a variety of ways have been explored for achieving this goal , of which the most common is to fix the sizes of the two groups , which is convenient if , as discussed above , the sizes of the groups are specified anyway as a part of the problem . in the present case ,
fixing the sizes of the groups fixes the coefficient @xmath43 of the @xmath44 term in the sum in eq . ; if the required sizes of the groups are @xmath45 and @xmath46 , then @xmath47 since we can not vary this coefficient , we shift our attention to the other terms in the sum .
if there were no further constraints on our choice of @xmath10 , apart from the normalization condition @xmath12 , our course would be clear : @xmath7 would be minimized by choosing @xmath10 proportional to the second eigenvector @xmath48 of the laplacian , also called the fiedler vector .
this choice places all of the weight in eq . in the term involving the second - smallest eigenvalue @xmath49 , also known as the algebraic connectivity .
the other terms would automatically be zero , since the eigenvectors are orthogonal .
unfortunately , there is an additional constraint on @xmath10 imposed by the condition , eq .
, that its elements take the values @xmath50 , which means in most cases that @xmath10 can not be chosen parallel to @xmath48 .
this makes the optimization problem much more difficult .
often , however , quite good approximate solutions can be obtained by choosing @xmath10 to be as close to parallel with @xmath48 as possible .
this means maximizing the quantity @xmath51 where @xmath52 is the @xmath16th element of @xmath48 . here
the second relation follows via the triangle inequality , and becomes an equality only when all terms in the first sum are positive ( or negative ) . in other words ,
the maximum of @xmath53 is achieved when @xmath54 for all @xmath16 , or equivalently when @xmath55 has the same sign as @xmath52 .
thus the maximum is obtained with the choice @xmath56 even this choice however is often forbidden by the condition that the number of @xmath57 and @xmath58 elements of @xmath10 be equal to the desired sizes @xmath45 and @xmath46 of the two groups , in which case the best solution is achieved by assigning vertices to one of the groups in order of the elements in the fiedler vector , from most positive to most negative , until the groups have the required sizes . for groups of different sizes
there are two distinct ways of doing this , one in which the smaller group corresponds to the most positive elements of the vector and one in which the larger group does . we can choose between them by calculating the cut size @xmath7 for both cases and keeping the one that gives the better result .
this then is the spectral partitioning method in its simplest form .
it is not guaranteed to minimize @xmath7 , but , particularly in cases where @xmath49 is well separated from the eigenvalues above it , it often does very well .
figure [ eppstein ] shows an example application typical of those found in the literature , to a two - dimensional mesh such as might be used in parallel finite - element calculations .
this particular mesh is a small 547-vertex example from bern _ et al . _
@xcite and is shown complete in panel ( a ) of the figure .
panel ( b ) shows the division of the mesh into two parts of @xmath59 and @xmath60 vertices respectively using the spectral partitioning approach , which finds a cut of size 46 edges in this case .
although the cut found in this example is a reasonable one , it does not appear at least to this author s eye that the vertex groups in fig .
[ eppstein]b constitute any kind of natural division of the network into `` communities . ''
this is typical of the problems to which spectral partitioning is usually applied : in most circumstances the network in question does not divide up easily into groups of the desired sizes , but one must do the best one can .
for these types of tasks , spectral partitioning is an effective and appropriate tool .
the task of finding natural community divisions in a network , however , is quite different , and demands a different approach , as we now discuss .
despite its evident success in the graph partitioning arena , spectral partitioning is a poor approach for detecting natural community structure in real - world networks , which is the primary topic of this paper .
the issue is with the condition that the sizes of the groups into which the network is divided be fixed .
this condition is neither appropriate nor realistic for community detection problems . in most cases
we do not know in advance the sizes of the communities in a network and choosing arbitrary sizes will usually preclude us from finding the best solution to the problem
. we would like instead to let the group sizes be free , but the spectral partitioning method breaks down if we do this , as we have seen : if the group sizes are not fixed , then the minimum cut size is always achieved by putting all vertices in one group and none in the other .
indeed , this statement is considerably broader than the spectral partitioning method itself , since any method that correctly minimizes the cut size without constraint on the group sizes is sure to find , in the general case , that the minimum value is achieved for this same trivial division .
several approaches have been proposed to get around this problem .
for instance , the ratio cut method @xcite minimizes not the simple cut size @xmath7 but the ratio @xmath61 , where @xmath45 and @xmath46 are again the sizes of the two groups of vertices .
this penalizes configurations in which either of the groups is small and hence favors balanced divisions over unbalanced ones , releasing us from the obligation to fix the group sizes .
spectral algorithms based on ratio cuts have been proposed @xcite and have proved useful for certain classes of partitioning problems .
still , however , this approach effectively chooses the group sizes , at least approximately , since it is biased in favor of divisions into equal - sized parts .
variations are possible that are biased towards other , unequal part sizes , but then one must choose those parts sizes and so again we have a situation in which we need to know in advance the sizes of the groups if we are to get the `` right '' results . the ratio cut method does allow some leeway for the sizes to vary around their specified values , which makes it more flexible than the simple minimum cut method , but at its core it still suffers from the same drawbacks that make standard spectral partitioning inappropriate for community detection .
the fundamental problem with all of these methods is that cut sizes are simply not the right thing to optimize because they do nt accurately reflect our intuitive concept of network communities .
a good division of a network into communities is not merely one in which the number of edges running between groups is small .
rather , it is one in which the number of edges between groups is _ smaller than expected_. only if the number of between - group edges is significantly lower than would be expected purely by chance can we justifiably claim to have found significant community structure .
equivalently , we can examine the number of edges _ within _ communities and look for divisions of the network in which this number is higher than expected the two approaches are equivalent since the total number of edges is fixed and any edges that do not lie between communities must necessarily lie inside them . these considerations lead us to shift our attention from measures based on pure cut size to a modified benefit function @xmath62 defined by @xmath63 this benefit function is called modularity @xcite .
it is a function of the particular division of the network into groups , with larger values indicating stronger community structure .
hence we should , in principle , be able to find good divisions of a network into communities by optimizing the modularity over possible divisions .
this approach , proposed in @xcite and since pursued by a number of authors @xcite , has proven highly effective in practice @xcite and is the primary focus of this article .
the first term in eq . is straightforward to calculate .
the second , however , is rather vague and needs to be made more precise before we can evaluate the modularity .
what exactly do we mean by the `` expected number '' of edges within a community ?
answering this question is essentially equivalent to choosing a `` null model '' against which to compare our network .
the definition of the modularity involves a comparison of the number of within - group edges in a real network and the number in some equivalent randomized model network in which edges are placed without regard to community structure .
it is one of the strengths of the modularity approach that it makes the role of this null model explicit and clear .
all methods for finding communities are , in a sense , assuming some null model , since any method must make a value judgment about when a particular density of edges is significant enough to define a community . in most cases , this assumption is hidden within the workings of a computer algorithm and is difficult to disentangle , even when the algorithm itself is well understood . by bringing its assumptions out into the open , the modularity method gives us more control over our calculations and more understanding of their implications .
our null model must have the same number of vertices @xmath0 as the original network , so that we can divide it into the same groups for comparison , but apart from this we have a good deal of freedom about our choice of model .
we here consider the broad class of randomized models in which we specify separately the probability @xmath64 for an edge to fall between every pair of vertices @xmath65 .
more precisely , @xmath64 is the expected number of edges between @xmath16 and @xmath66 , a definition that allows for the possibility that there may be more than one edge between a pair of vertices , which happens in certain types of networks .
we will consider some particular choices of @xmath64 in a moment , but for now let us pursue the developments in general form
. given @xmath64 , the modularity can be defined as follows .
the actual number of edges falling between a particular pair of vertices @xmath16 and @xmath66 is @xmath67 , eq . , and the expected number is , by definition , @xmath64 .
thus the actual minus expected number of edges between @xmath16 and @xmath66 is @xmath68 and the modularity is ( proportional to ) the sum of this quantity over all pairs of vertices belonging to the same community .
let us define @xmath69 to be the community to which vertex @xmath16 belongs .
then the modularity can be written @xmath70 \delta(g_i , g_j ) , \label{q1}\ ] ] where @xmath71 if @xmath72 and 0 otherwise and @xmath2 is again the number of edges in the network .
the extra factor of @xmath73 in eq .
is purely conventional ; it is included for compatibility with previous definitions of the modularity and plays no part in the maximization of @xmath62 since it is a constant for any given network . a special case of eq .
was given previously by the present author in @xcite and independently , in slightly different form , by white and smyth @xcite .
a number of other expressions for the modularity have also been presented by various authors @xcite and are convenient in particular applications . also of interest is the derivation of the modularity given recently by reichardt and bornholdt @xcite , which is quite general and provides an interesting alternative to the derivation presented here .
returning to the null model , how should @xmath64 be chosen ?
the choice is not entirely unconstrained .
first , we consider in this paper only undirected networks , which implies that @xmath74 .
second , it is axiomatically the case that @xmath75 when all vertices are placed in a single group together : by definition , the number of edges within groups and the expected number of such edges are both equal to @xmath2 in this case . setting all @xmath69 equal in eq .
, we find that @xmath76 = 0 $ ] or equivalently @xmath77 this equation says that we are restricted to null models in which the expected number of edges in the entire network equals the actual number of edges in the original network a natural choice if our comparison of numbers of edges within groups is to have any meaning . beyond these basic considerations
, there are many possible choices of null model and several have been considered previously in the literature @xcite .
perhaps the simplest is the standard ( bernoulli ) random graph , in which edges appear with equal probability @xmath78 between all vertex pairs . with a suitably chosen value of @xmath79 this model can be made to satisfy but , as many authors have pointed out @xcite ,
the model is not a good representation of most real - world networks .
a particularly glaring aspect in which it errs is its degree distribution .
the random graph has a binomial degree distribution ( or poisson in the limit of large graph size ) , which is entirely unlike the right - skewed degree distributions found in most real - world networks @xcite .
a much better null model would be one in which the degree distribution is approximately the same as that of the real - world network of interest . to satisfy this demand
we will restrict our attention in this paper to models in which the expected degree of each vertex within the model is equal to the actual degree of the corresponding vertex in the real network . noting that the expected degree of vertex @xmath16 is given by @xmath80
, we can express this condition as @xmath81 if this constraint is satisfied , then is automatically satisfied as well , since @xmath82 .
equation is a considerably more stringent constraint than in most cases , for instance , it excludes the bernoulli random graph but it is one that we believe makes good sense , and one moreover that has a variety of desirable consequences for the developments that follow . the simplest null model in this class , and the only one that has been considered at any length in the past , is the model in which edges are placed entirely at random , subject to the constraint .
that is , the probability that an end of a randomly chosen edge attaches to a particular vertex @xmath16 depends only on the expected degree @xmath15 of that vertex , and the probabilities for the two ends of a single edge are independent of one another .
this implies that the expected number of edges @xmath64 between vertices @xmath16 and @xmath66 is the product @xmath83 of separate functions of the two degrees , where the functions must be the same since @xmath64 is symmetric .
then eq .
implies @xmath84 for all @xmath16 and hence @xmath85 for some constant @xmath86 . and
. says that @xmath87 and hence @xmath88 and @xmath89 this model has been studied in the past in its own right as a model of a network , for instance by chung and lu @xcite .
it is also closely related to the configuration model , which has been studied widely in the mathematics and physics literature @xcite .
indeed , essentially all expected properties of our model and the configuration model are identical in the limit of large network size , and hence eq . can be considered equivalent to the configuration model in this limit
. and @xmath66 in the configuration model is also given by eq .
in the limit of large network size ; for smaller networks , there are corrections of order @xmath90 . ]
although many of the developments outlined in this paper are true for quite general choices of the null model used to define the modularity , the choice is the only one we will pursue here .
it is worth keeping mind however that other choices are possible : massen and doye @xcite , for instance , have used a variant of the configuration model in which multiedges and self - edges were excluded . and further choices could be useful in specific cases , such as cases where there are strong correlations between the degrees of vertices @xcite or where there is a high level of network transitivity @xcite .
once we have an explicit expression for the modularity we can determine the community structure by maximizing it over possible divisions of the network . unfortunately , exhaustive maximization over all possible divisions is computational intractable because there are simply too many divisions , but various approximate optimization methods have proven effective @xcite . here
, we develop a matrix - based approach analogous to the spectral partitioning method of section [ specpart ] , which leads not only to a whole array of possible optimization algorithms but also to new insights about the nature and implications of community structure in networks . as before ,
let us consider initially the division of a network into just two communities and denote a potential such division by an index vector @xmath10 with elements as in eq . .
we notice that the quantity @xmath91 is 1 if @xmath16 and @xmath66 belong to the same group and 0 if they belong to different groups or , in the notation of eq . , @xmath92 thus we can write in the form @xmath93 ( s_is_j+1 ) \nonumber\\ & = & { 1\over4 m } \sum_{ij } \bigl [ a_{ij } - p_{ij } \bigr ] s_is_j,\end{aligned}\ ] ] where we have in the second line made use of eq . .
this result can conveniently be rewritten in matrix form as @xmath94 where @xmath95 is the real symmetric matrix having elements @xmath96 we call this matrix the modularity matrix and it plays a role in the maximization of the modularity equivalent to that played by the laplacian in standard spectral partitioning : equation is the equivalent of eq . for the cut size and matrix methods
can thus be applied to the modularity that are the direct equivalents of those developed for spectral partitioning , as we now show .
first , let us point out a few important properties of the modularity matrix .
equations and together imply that all rows ( and columns ) of the modularity matrix sum to zero : @xmath97 this immediately implies that for any network the vector @xmath38 is an eigenvector of the modularity matrix with eigenvalue zero , just as is the case with the laplacian . unlike the laplacian however , the eigenvalues of the modularity matrix are not necessarily all of one sign and in practice the matrix usually has both positive and negative eigenvalues .
this observation and the eigenspectrum of the modularity matrix in general are , as we will see , closely tied to the community structure of the network . working from eq .
we now proceed by direct analogy with section [ specpart ] .
we write @xmath10 as a linear combination of the normalized eigenvectors @xmath98 of the modularity matrix , @xmath99 with @xmath100 .
then @xmath101 where @xmath102 is the eigenvalue of @xmath95 corresponding to the eigenvector @xmath98 .
we now assume that the eigenvalues are labeled in _ decreasing _ order @xmath103 and the task of maximizing @xmath62 is one of choosing the quantities @xmath36 so as to place as much as possible of the weight in the sum in the terms corresponding to the largest ( most positive ) eigenvalues . as with ordinary spectral partitioning , this would be a simple task if our choice of @xmath10 were unconstrained ( apart from normalization ) : we would just choose @xmath10 proportional to the leading eigenvector @xmath104 of the modularity matrix .
but the elements of @xmath10 are restricted to the values @xmath20 , which means that @xmath10 can not normally be chosen parallel to @xmath104 .
again as before , however , good approximate solutions can be obtained by choosing @xmath10 to be as close to parallel with @xmath104 as possible , which is achieved by setting @xmath105 this then is our first and simplest algorithm for community detection : we find the eigenvector corresponding to the most positive eigenvalue of the modularity matrix and divide the network into two groups according to the signs of the elements of this vector . ) .
the solid curve represents the division found by the modularity - based method of this section .
and the squares and circles represent the actual division of the network observed when the dolphin community split into two as a result of the departure of a keystone individual .
( the individual who departed is represented by the triangle . ) ] in practice , this method works nicely , as discussed in @xcite . making the choice for our null model ,
we have applied it to a variety of standard and less standard test networks and find that it does a good job of finding community divisions .
figure [ dolphins ] shows a representative example , an animal social network assembled and studied by lusseau _ et al .
_ @xcite .
the vertices in this network represent 62 bottlenose dolphins living in doubtful sound , new zealand , with social ties between dolphin pairs established by direct observation over a period of several years .
this network is of particular interest because , during the course of the study , the dolphin group split into two smaller subgroups following the departure of a key member of the population .
the subgroups are represented by the shapes of the vertices in the figure .
the dotted line denotes the division of the network into two equal - sized groups found by the standard spectral partitioning method .
while , as expected , this method does a creditable job of dividing the network into groups of these particular sizes , it is clear to the eye that this is not the natural community division of the network and neither does it correspond to the division observed in real life .
the spectral partitioning method is hamstrung by the requirement that we specify the sizes of the two communities ; unless we know what they are in advance , blind application of the method will not usually find the `` right '' division of the network .
the method based on the leading eigenvector of the modularity matrix , however , does much better .
unconstrained by the need to find groups of any particular size , this method finds the division denoted by the solid line in the figure , which , as we see , corresponds quite closely to the split actually observed all but three of the 62 dolphins are placed in the correct groups .
the magnitudes of the elements of the eigenvector @xmath104 also contain useful information about the network , indicating , as discussed in @xcite , the `` strength '' with which vertices belong to the communities in which they are placed . as an example of this phenomenon
consider fig .
[ books ] , which depicts the network of political books from ref . @xcite .
this network , compiled by v. krebs ( unpublished ) , represents recent books on us politics , with edges connecting pairs of books that are frequently purchased by the same customers of the on - line bookseller amazon.com . applying our method
, we find that the network divides as shown in the figure , with the colors of the vertices representing the values of the elements of the eigenvector .
the two groups correspond closely to the apparent alignment of the books according to left - wing and right - wing points of view @xcite , and are suggestively colored blue and red in the figure .
the most blue and most red vertices are those that , by our calculation , belong most strongly to the two groups and are thus , perhaps , the `` most left - wing '' and `` most right - wing '' of the books under consideration
. those familiar with current us politics will be unsurprised to learn that the most left - wing book in this sense was the polemical _ bushwacked _ by molly ivins and lou dubose .
perhaps more surprising is the most right - wing book : _ a national party no more _ by zell miller .
the algorithm described in the previous section has two obvious shortcomings .
first , it divides networks into only two communities , while real - world networks can certainly have more than two .
second , it makes use only of the leading eigenvector of the modularity matrix and ignores all the others , which throws away useful information contained in those other vectors .
both of these shortcomings are remedied by the following generalization of the method .
consider the division of a network into @xmath1 non - overlapping communities , where @xmath1 may now be greater than 2 .
following alpert and yao @xcite and more recently white and smyth @xcite , let us define an @xmath106 index matrix @xmath107 with one column for each community : @xmath108 .
each column is an index vector now of @xmath109 elements ( rather than @xmath50 as previously ) , such that @xmath110 note that the columns of @xmath107 are mutually orthogonal , that the rows each sum to unity , and that the matrix satisfies the normalization condition @xmath111 . observing that the @xmath112-symbol in eq .
is now given by @xmath113 the modularity for this division of the network is @xmath114 where here and henceforth we suppress the leading multiplicative constant @xmath73 from eq . , which has no effect on the position of the maximum of the modularity .
writing @xmath115 , where @xmath116 is the matrix of eigenvectors of @xmath95 and @xmath117 is the diagonal matrix of eigenvalues @xmath118 , we then find that @xmath119 again we wish to maximize this modularity , but now we have no constraint on the number @xmath1 of communities ; we can give @xmath107 as many columns as we like in our effort to make @xmath62 as large as possible .
if the elements of the matrix @xmath107 were unconstrained apart from the basic conditions on the rows and columns mentioned above , a choice of @xmath1 communities would be equivalent to choosing @xmath120 independent , mutually orthogonal columns @xmath121 .
( only @xmath120 of the columns are independent , the last being fixed by the condition that the rows of @xmath107 sum to unity . ) in this case our path would be clear : @xmath62 would be maximized by choosing the columns proportional to the leading eigenvectors of @xmath95
. however , only those eigenvectors corresponding to positive eigenvalues can give positive contributions to the modularity , so the optimal modularity would be achieved by choosing exactly as many independent columns of @xmath107 as there are positive eigenvalues , or equivalently by choosing the number of groups @xmath1 to be 1 greater than the number of positive eigenvalues .
unfortunately , our problem has the additional constraint that the index vectors @xmath122 have only binary @xmath109 elements , which means it may not be possible to find as many index vectors making positive contributions to the modularity as the set of positive eigenvalues suggests .
thus the number of positive eigenvalues , plus 1 , is an _ upper bound _ on the number of communities and again we see that there is an intimate connection between the properties of the modularity matrix and the community structure of the network it describes . in section [ singlevec ]
we maximized the modularity approximately by focusing solely on the term in @xmath62 proportional to the largest eigenvalue of @xmath95 .
let us now make the more general ( and often better ) approximation of keeping the leading @xmath79 eigenvalues , where @xmath79 may be anywhere between 1 and @xmath0 .
some of the eigenvalues , however , may be negative , which will prove inconvenient . to get around this we rewrite eq .
thus : @xmath123 \nonumber\\ & = & n\alpha + \sum_{j=1}^n \sum_{k=1}^c ( \beta_j-\alpha ) \biggl [ \sum_{i=1}^n u_{ij } s_{ik } \biggr]^2 , \label{alphaq}\end{aligned}\ ] ] where @xmath124 is a constant whose value we will choose shortly and we have made use of @xmath111 and the fact that @xmath125 is orthogonal .
now , employing an argument similar to that used for ordinary spectral partitioning in @xcite , let us define a set of vertex vectors @xmath126 , @xmath127 , of dimension @xmath79 , such that the @xmath66th component of the @xmath16th vector is @xmath128_j = \sqrt{\beta_j-\alpha}\,u_{ij}. \label{defsri}\ ] ] provided we choose @xmath129 , @xmath126 is guaranteed real for all @xmath16 . then , dropping terms in proportional to the smallest @xmath130 of the factors @xmath131 , we have @xmath132 ^ 2 \nonumber\\ & = & n\alpha + \sum_{k=1}^c \sum_{j=1}^p \biggl
[ \sum_{i\in g_k } \bigl [ { \mathbf}{r}_i \bigr]_j \biggr]^2 \nonumber\\ & = & n\alpha + \sum_{k=1}^c | { \mathbf}{r}_k |^2,\qquad { } \label{finalq}\end{aligned}\ ] ] where @xmath133 is the set of vertices comprising group @xmath134 and the community vectors @xmath135 , @xmath136 , are @xmath137 the community structure problem is now equivalent to choosing a division of the vertices into groups so as to maximize the magnitudes of the vectors @xmath135 .
this means we need to arrange that the individual vertex vectors @xmath126 going into each group point in approximately the same direction .
problems of this type are called vector partitioning problems . the parameter @xmath79 controls the balance between the complexity of the vector partitioning problem and the accuracy of the approximation we make by keeping only some of the eigenvalues
. the calculations will be faster but less accurate for smaller @xmath79 and slower but more accurate for larger .
for the special case @xmath138 where we keep all of the eigenvalues , eq .
is exact . in this case
, we note that the vertex vectors have the property @xmath139 it s then simple to see that eq .
is trivially equivalent to the fundamental definition of the modularity , so in the @xmath138 case our mapping to a vector partitioning problem gives little insight into the modularity maximization problem .
the real advantage of our approach comes when @xmath140 , where the method extracts precisely those factors that make the principal contributions to the modularity i.e .
, those corresponding to the largest eigenvalues discarding those that have relatively little effect . in practice ,
as we have seen for the single - eigenvector algorithm , the main features of the community structure are often captured by just the first eigenvector or perhaps the first few , which allows us to reduce the complexity of our optimization problem immensely .
the approach is similar in concept to the standard technique of principal components analysis ( pca ) used to reduce high - dimensional data sets to manageably small dimension by focusing on the eigendirections along which the variance about the mean is greatest and ignoring directions that contribute little . in fact , this similarity is more than skin - deep : the form of our modularity matrix is closely analogous to the covariance matrix whose eigenvectors are the basis for pca .
the elements of the covariance matrix are correlation functions of the form @xmath141 , where @xmath142 and @xmath143 denote measured variables in the data set .
thus the covariance is the difference between the actual value of the mean product @xmath144 of two variables and the value @xmath145 expected by chance for that product if the variables were uncorrelated .
similarly , the elements @xmath146 of the modularity matrix are equal to the actual number of edges @xmath67 between a given pair of vertices minus the number @xmath147 expected by chance , expressed in a product form . in a sense , our spectral method for modularity optimization can be regarded as a `` principal components analysis for networks . ''
this aspect of the method is clear , for instance , in the study of political books represented in fig .
[ books ] : the leading eigenvector used to assign the colors to the vertices in the figure is playing a role equivalent to the eigendirections in pca , defining a `` direction of greatest variation '' in the structure of the network .
the vertex vectors of eq .
are similarly analogous to the low - dimensional projections used in pca . or 3 could be used to define graph layouts for visualizing networks in 2 or 3 dimensions .
either the endpoints of the vectors could define vertex positions themselves or they could be used as starting positions for a spring embedding visualizer or other more conventional layout scheme . ] returning to our algorithm , let us consider again the special case of the division of a network into just two communities .
( multi - way division is considered in section [ multiway ] . ) since @xmath38 is always an eigenvector of the modularity matrix and the eigenvectors are orthogonal , the elements of all other eigenvectors must sum to zero : @xmath148_i = \sqrt{n}\,{\mathbf}{u}_1^t{\mathbf}{u}_j = 0.\ ] ] but eq .
then implies that @xmath149_j = \sqrt{\beta_j-\alpha } \sum_{i=1}^n u_{ij } = \sqrt{\beta_j-\alpha } \sum_{i=1}^n \bigl [ { \mathbf}{u}_j \bigr]_i = 0,\ ] ] and hence @xmath150 for any value of @xmath79 .
this in turn implies that the community vectors @xmath135 also sum to zero : @xmath151 and as a special case of this last result , any division of a network into two communities has community vectors @xmath152 and @xmath153 that are of equal magnitude and oppositely directed .
furthermore , the maximum of the modularity , eq . , is always achieved when each individual vertex vector @xmath126 has a positive inner product with the community vector of the community to which the vertex belongs . to see this ,
observe that removing a vertex @xmath16 from a community @xmath134 where @xmath154 produces a change in the corresponding term @xmath155 in eq . of @xmath156 similarly adding vertex @xmath16 to a community for which @xmath157 also increases @xmath155 .
hence , we can always increase the modularity by moving vertices until they are in groups such that @xmath157 .
taken together , these results imply that possible candidates for the optimal division of a network into two groups are fully specified by just the _ direction _ of the single vector @xmath152 .
once we have this direction , we know that the vertices divide according to whether their projection along this direction is positive or negative . alternatively , we can consider the direction of @xmath152 to define a perpendicular plane through the origin in the @xmath79-dimensional vector space occupied by the vertex vectors @xmath126 .
the vertices then divide according to which side of this plane their vectors fall on . finding the maximum of the modularity is then a matter of choosing this bisecting plane to maximize the magnitude of @xmath152 . for a small network with @xmath158 .
the dotted line represents one of the @xmath0 possible topologically distinct cut planes.,width=302 ] in general , this still leaves us with a moderately difficult optimization problem : the number of bisecting planes that give distinct partitions of the vertex vectors is large and difficult to enumerate as the dimension @xmath79 of the space becomes large . for the case @xmath158 , however , a relatively simple solution exists .
consider fig .
[ zachvecs ] , which shows a typical example of the vertex vectors .
in this two - dimensional case , there are only @xmath0 topologically distinct choices of the bisecting plane ( actually just a line in this case , denoted by the dashed line in the figure ) , and furthermore the divisions of the vertices that these choices represent change by only a single vertex at a time as we rotate the plane about the origin .
this makes it computationally simple to perform the rotation , keep track of the value of @xmath152 , and so find the maximum of the modularity within this approximation . evaluating
the magnitude of @xmath152 involves a constant number of operations each time we move the line , and hence the total work involved in finding the maximum is @xmath159 for all @xmath0 possible positions , which is the same as the @xmath159 operations needed to separate the vertices in the @xmath160 case . for @xmath161
, we do not know of an efficient method to enumerate exhaustively all topologically distinct bisecting planes in the vertex vector space , and hence we have to turn to approximate methods for solving the vector partitioning problem . a number of reasonable heuristics have been described in the past .
we have found acceptable though not spectacular results , for instance , with the `` melo '' algorithm of @xcite , which is essentially a greedy algorithm in which a grouping of vectors is built up by repeatedly adding to it the vector that makes the largest contribution to @xmath62 . before implementing any of these methods ,
a crucial question we must answer is what value we should choose for the parameter @xmath124 .
by tuning this value we can improve the accuracy of our approximation to @xmath62 as follows . by dropping the @xmath130 most negative eigenvalues ,
we are in effect making an approximation to the matrix @xmath162 in which it takes not its full value @xmath163 , but an approximate value @xmath164 , where @xmath165 and @xmath166 are the matrices @xmath117 and @xmath167 with the last @xmath130 diagonal elements set to zero .
we can quantify the error this introduces by calculating the sum of the squares of the elements of the difference between the two matrices , which is given by @xmath168 ^ 2 \nonumber\\ & = & { \mathrm{tr}}[({\mathbf}{d}-\alpha{\mathbf}{i})-({\mathbf}{d}'-\alpha{\mathbf}{i}')]^2 = \!\!\sum_{i = p+1}^n ( \beta_i-\alpha)^2,\qquad{}\end{aligned}\ ] ] where in the second line we have made use of the fact that @xmath125 is orthogonal . minimizing this error by setting the derivative @xmath169
, we find @xmath170 in other words , the minimal mean square error introduced by our approximation is achieved by setting @xmath124 equal to the mean of the eigenvalues that have been dropped .
the only exception is when @xmath138 , where the choice of @xmath124 makes no difference since no approximation is being made anyway . in our calculations we have used @xmath171 in this case , but any choice @xmath172 would work equally well .
implementation of the methods described in section [ specmod ] is straightforward .
the leading - eigenvector method of section [ singlevec ] requires us to find only the single eigenvector of the modularity matrix @xmath95 corresponding to the most positive eigenvalue .
this is most efficiently achieved by the direct multiplication or power method . starting with a trial vector
, we repeatedly multiply by the modularity matrix and unless we are unlucky enough to have chosen another eigenvector as our trial vector the result will converge to the eigenvector of the matrix having the eigenvalue of largest magnitude . in some cases
this eigenvalue will be the most positive one , in which case our calculation ends at this point .
in other cases the eigenvalue of largest magnitude may be negative . if this happens then , denoting this eigenvalue by @xmath173 , we calculate the shifted matrix @xmath174 , which has eigenvalues @xmath175 ( necessarily all nonnegative ) and the same eigenvectors as the modularity matrix itself
. then we repeat the power - method calculation for this new matrix and this time the eigenvalue of largest magnitude must be @xmath176 and the corresponding eigenvector is the one we are looking for . for the method of section [ multivec ] ,
we require either all of the eigenvectors of the modularity matrix or a subset corresponding to the @xmath79 most positive eigenvalues .
these are most conveniently calculated using the lanczos method or one of its variants @xcite .
the fundamental matrix operation at the heart of the lanczos method is again multiplication of the matrix @xmath95 into a trial vector .
efficient implementation of any of these methods thus rests upon our ability to rapidly multiply an arbitrary vector @xmath177 by the modularity matrix .
this presents a problem because the modularity matrix is dense , and hence it appears that matrix multiplications will demand @xmath178 time each , where @xmath0 is , as before , the number of vertices in the network ( which is also the size of the matrix ) . by contrast , the equivalent calculation in standard spectral partitioning is much faster because the laplacian matrix is sparse , having only @xmath179 nonzero elements , where @xmath2 is the number of edges in the network . for the standard choice , eq . , of null model
used to define the modularity , however , it turns out that we can multiply by the modularity matrix just as fast as by the laplacian by making use of the special structure of the matrix . in vector notation
the modularity matrix can in this case be written @xmath180 where @xmath5 is the adjacency matrix , eq . , and @xmath181 is the @xmath0-element vector whose elements are the degrees @xmath15 of the vertices .
then @xmath182 since the adjacency matrix is sparse , having only @xmath183 elements , the first term can be evaluated in @xmath183 time , while the second requires us to evaluate the inner product @xmath184 only once and then multiply it into each element of @xmath181 in turn , both operations taking @xmath159 time . thus the entire matrix multiplication can be completed in @xmath185 time , just as with the normal laplacian matrix .
if a shift of the eigenvalues is required to find the most positive one , as described above , then there is an additional term @xmath186 in the matrix , but this also can be multiplied into an arbitrary vector in @xmath159 time , so again the entire operation can be completed in @xmath185 time .
typically @xmath159 matrix multiplications are required for either the power method or the lanczos method to converge to the required eigenvalues , and hence the calculation takes @xmath187 time overall . in the common case in which the network is sparse and @xmath188 , this simplifies to @xmath178 .
while this is , essentially , the end of the calculation for the power method , the lanczos method unfortunately demands more effort to find the eigenvectors themselves . in fact , it takes @xmath189 time to find all eigenvectors of a matrix using the lanczos method , which is quite slow .
there are on the other hand variants of the lanczos method ( as well as other methods entirely ) that can find just a few leading eigenvectors faster than this , which makes calculations that focus on a fixed small number of eigenvectors preferable to ones that use all eigenvectors . in our calculations we have primarily concentrated on algorithms that use only one or two eigenvectors , which typically run in time @xmath178 on a sparse network .
the methods for spectral optimization of the modularity described in section [ specmod ] are only approximate .
indeed , the problem of modularity optimization is formally equivalent to an instance of the np - hard max - cut problem , so it is almost certainly the case that no polynomial - time algorithm exists that will find the modularity optimum in all cases .
given that the algorithms we have described run in polynomial time , it follows that they must fail to find the optimum in some cases , and hence that there is room for improvement of the results . in standard graph partitioning applications
it is common to use a spectral approach based on the graph laplacian as a first pass at the problem of dividing a network .
the spectral method gives a broad picture of the general shape the division should take , but there is often room for improvement .
typically another algorithm , such as the kernighan
lin algorithm @xcite , which swaps vertex pairs between groups in an effort to reduce the cut size , is used to refine this first pass , and the resulting two - stage joint strategy gives considerably better results than either stage on its own .
we have found that a similar joint strategy gives good results in the present case also : the divisions found with our spectral approach can be improved in small but significant ways by adding a refinement step akin to the kernighan lin algorithm .
as described in @xcite , we take an initial division into two communities derived , for instance , from the leading - eigenvector method of section [ singlevec ] and move single vertices between the communities so as to increase the value of the modularity as much as possible , with the constraint that each vertex can be moved only once . repeating the whole process iteratively
until no further improvement is obtained , we find a final value of the modularity which can improve on that derived from the spectral method alone by tens of percent in some cases , and smaller but still significant amounts in other cases . although the absolute gains in modularity are not always large , we find that this refinement step is very much worth the effort it entails , raising the typical level of performance of our methods from the merely good to the excellent , when compared with other algorithms .
specific examples are given in @xcite .
it is certainly possible that other refinement strategies might also give good results .
for instance , the `` extremal optimization '' method explored by duch and arenas @xcite for optimizing modularity could be employed as a refinement method by using the output of our spectral division as its starting point , rather than the random configuration used as a starting point by duch and arenas .
so far we have discussed primarily methods for dividing networks into two communities .
many of the networks we are concerned with , however , have more than two communities .
how can we generalize our methods to this case ?
the simplest approach is repeated division into two . that is , we use one of the methods described above to divide our network in two and then divide those parts in two again , and so forth .
this approach was described briefly in ref .
@xcite .
it is important to appreciate that upon further subdividing a community within a network into two parts , the additional contribution @xmath190 to the modularity made by this subdivision is not given correctly if we apply the algorithms of section [ specmod ] to that community alone .
that is , we can not simply write down the modularity matrix for the community in question considered as a separate graph in its own right and examine the leading eigenvector or eigenvectors . instead we proceed as follows .
let us denote the set of vertices in the community to be divided by @xmath191 and let @xmath192 be the number of vertices within this community .
now let @xmath107 be an @xmath193 index matrix denoting the subdivision of the community into @xmath1 subcommunities such that @xmath194 then , following eq .
, @xmath190 is the difference between the modularities of the network before and after subdivision of the community thus : @xmath195 s_{ik } s_{jk } \nonumber\\ & = & { \mathrm{tr}}({\mathbf}{s}^t{\mathbf}{b}^{(g)}{\mathbf}{s } ) , \label{deltaq}\end{aligned}\ ] ] where @xmath196 is an @xmath197 generalized modularity matrix with elements indexed by the vertex labels @xmath65 of the vertices within group @xmath191 and having values @xmath198 with @xmath199 defined by eq . .
equation has the same form as our previous expression , eq .
, for the modularity of the full network , and , following the same argument as for eqs . to , we can then show that optimization of the additional modularity contribution from subdivision of a community can also be expressed as a vector partitioning problem , just as before .
we can approximate this vector partitioning problem using only the leading eigenvector as in section [ singlevec ] or using more than one vector as in section [ multivec ] .
the resulting divisions can also be optimized using a `` refinement '' stage as in section [ refinement ] , to find the best possible modularity at each step . using this method
we can repeatedly subdivide communities to partition networks into smaller and smaller groups of vertices and in principle this process could continue until the network is reduced to @xmath0 communities containing only a single vertex each .
normally , however , we stop before this point is reached because there is no point in subdividing a community any further if no subdivision exists that will increase the modularity of the network as a whole .
the appropriate strategy is to calculate explicitly the modularity contribution @xmath190 at each step in the subdivision of a network , and to decline to subdivide any community for which the value of @xmath190 is not positive .
communities with the property of having no subdivision that gives a positive contribution to the modularity of the network as a whole we call indivisible ; the strategy described here is equivalent to subdividing communities repeatedly until every remaining community is indivisible . this strategy appears to work very well in practice .
it is , however , not perfect ( a conclusion we could draw under any circumstances from the fact that it runs in polynomial time see above ) . in particular , it is certain that repeated subdivision of a network into two parts will in some cases fail to find the optimal modularity configuration .
consider , for example , the ( rather trivial ) network shown in fig .
[ line8 ] , which consists of eight vertices connected together in a line . by exhaustive enumeration
we can show that , among possible divisions of this network into only two parts , the division indicated in fig . [ line8]a , right down the middle of the line , is the one that gives the highest modularity .
the optimum modularity over divisions into any number of parts , however , is achieved for the three - way division shown in fig .
[ line8]b .
it is clear that if we first split the network as shown in fig .
[ line8]a , no subsequent subdivision of the network can ever find the configuration of fig .
[ line8]b , and hence our algorithm will fail in this case to find the global optimum . nonetheless , the algorithm does appear to find divisions that are close to optimal in most cases we have investigated .
repeated subdivision is the approach we have taken to multi - community divisions in our own work , but it is not the only possible approach . in some respects a more satisfying approach would be to work directly from the expression for the modularity of the complete network with a multi - community division .
unfortunately , maximizing requires us to perform a vector partitioning into more than two groups , a problem about whose solution rather little is known .
some general observations are , however , worth making .
first , we note that the community vectors @xmath135 in the optimal solution of a vector partitioning problem always have directions more than @xmath200 apart .
to demonstrate this , we note that the change in the contribution to eq .
if we amalgamate two communities into one is @xmath201 which is positive if the directions of @xmath152 and @xmath153 are less than @xmath200 apart .
thus we can always increase the modularity by amalgamating a pair of communities unless their vectors are more than @xmath200 apart .
but the maximum number of directions more than @xmath200 apart that can exist in a @xmath79-dimensional space is @xmath202 , which means that @xmath202 is also the maximum number of communities we can find by optimizing a @xmath79-dimensional spectral approximation to the modularity .
thus if we use only a single eigenvector we will find at most two groups ; if we use two we will find at most three groups , and so forth .
so the choice of how many eigenvectors @xmath79 to work with is determined to some extent by the network : if the overall optimum modularity is for a division into @xmath1 groups , we will certainly fail to find that optimum if we use less than @xmath120 eigenvectors .
second , we note that while true multi - way vector partitioning may present problems , simple heuristics that group the vertex vectors together can still produce good results .
for instance , white and smyth @xcite have applied the standard technique of @xmath134-means clustering based on group centroids to a different but related optimization problem and have found good results .
it is possible this approach would work for our problem also if applied to the centroids of the end - points of the vertex vectors .
it is also possible that an intrinsically vector - based variant of @xmath134-means clustering could be created to tackle the vector partitioning problem directly , although we are not aware of such an algorithm in the current vector partitioning literature .
it is clear from the developments of the previous sections that there is useful information about the structure of a network stored in the eigenvectors corresponding to the most positive eigenvalues of the modularity matrix .
it is natural to ask whether there is also useful information in the eigenvectors corresponding to the negative eigenvalues and indeed it turns out that there is : the negative eigenvalues and their eigenvectors contain information about a nontrivial type of `` anti - community structure '' that is of substantial interest in some instances .
consider again the case in which we divide our network into just two groups and look once more at eq .
, which gives the modularity in this case .
suppose now that instead of maximizing the terms involving the most positive eigenvalues , we maximize the terms involving the most negative ones . as we can easily see from the equation ,
this is equivalent to _ minimizing _ rather than maximizing the modularity .
what effect will this have on the divisions of the network that we find ?
large negative values of the modularity correspond to divisions in which the number of edges within groups is _ smaller _ than expected on the basis of chance , and the number of edges between groups correspondingly bigger .
figure [ bipartex ] shows a sketch of a network having this property .
such networks are said to be bipartite if there are no edges at all within groups , or approximately bipartite if there are a few within - group edges as in the figure .
bipartite or approximately bipartite graphs have attracted some attention in the recent literature .
for instance , kleinberg @xcite has suggested that small bipartite subgraphs in the web graph may be a signature of so - called hub / authority structure within web communities , while holme _ et al . _
@xcite and estrada and rodrguez - velzquez @xcite have independently devised measures of bipartitivity and used them to analyze a variety of real - world networks .
the arguments above suggest that we should be able to detect bipartite or approximately bipartite structure in networks by looking for divisions of the vertices that minimize modularity . in the simplest approximation
, we can do this by focusing once more on just a single term in eq .
, that corresponding to the most negative eigenvalue @xmath173 , and maximizing the coefficient of this eigenvalue by choosing @xmath203 for vertices having a negative element in the corresponding eigenvector and @xmath204 for the others .
in other words , we can achieve an approximation to the minimum modularity division of the network by dividing vertices according to the signs of the elements in the eigenvector @xmath205 , and this division should correspond roughly to the most nearly bipartite division .
we can also append a `` refinement '' step to the calculation , similar to that described in section [ refinement ] , in which , starting from the division given by the eigenvector , we move single vertices between groups in an effort to minimize the modularity further . as an example of this type of calculation ,
consider fig .
[ words ] , which shows a network representing juxtapositions of words in a corpus of english text , in this case the novel _ david copperfield _ by charles dickens . to construct this network ,
we have taken the 60 most commonly occurring nouns in the novel and the 60 most commonly occurring adjectives .
( the limit on the number of words is imposed solely to permit a clear visualization ; there is no reason in principle why the analysis could not be extended to a much larger network . )
the vertices in the network represent words and an edge connects any two words that appear adjacent to one another at any point in the book .
eight of the words never appear adjacent to any of the others and are excluded from the network , leaving a total of 112 vertices .
typically adjectives occur next to nouns in english .
it is possible for adjectives to occur next to other adjectives ( `` the big green bus '' ) or for nouns to occur next to other nouns ( `` the big tour bus '' ) , but these juxtapositions are less common .
thus we would expect our network to be approximately bipartite in the sense described above : edges should run primarily between vertices representing different types of words , with fewer edges between vertices of the same type .
one would be hard pressed to tell this from fig .
[ words]a , however : the standard layout algorithm used to draw the network completely fails to reveal the structure present .
figure [ words]b shows what happens when we divide the vertices by minimizing the modularity using the method described above a first division according to the elements of the eigenvector with the most negative eigenvalue , followed by a refinement stage to reduce the modularity still further .
it is now clear that the network is in fact nearly bipartite , and the two groups found by the algorithm correspond closely to the known groups of adjectives and nouns , as indicated by the shapes of the vertices .
83% of the words are classified correctly by this simple calculation .
divisions with large negative modularity are like those with large positive modularity not limited to having only two groups . if we are interested purely in minimizing the modularity we can in principle use as many groups as we like to achieve that goal .
a division with @xmath134 groups is called @xmath134-partite if edges run only between groups and approximately @xmath134-partite if there are a few within - group edges .
one might imagine that one could find @xmath134-partite structure in a network just by looking for divisions that minimize the number of within - group edges , but brief reflection persuades us that the optimum solution to this search problem is always to put each vertex in a group on its own , which automatically means that all edges lie between groups and none within groups . as with the ordinary community structure problem , the way to avoid this trivial solution is to concentrate not on the total number of edges within groups but on the difference between this number and the expected number of such edges .
thus , once again , we are led naturally to the consideration of modularity as a measure of the best way to divide a network . one way to minimize modularity over divisions into an arbitrary number of groups
is to proceed by analogy with our earlier calculations of community structure and repeatedly divide the network in two using the single - eigenvector method above . just as before , eq .
gives the additional change @xmath190 in the modularity upon subdivision of a group in a network , and the division process ends when the algorithm fails to find any subdivision with @xmath206 .
alternatively , one can derive the analog of eq . and thereby map the minimization of the modularity onto a vector partitioning problem .
the appropriate definition of the vertex vectors turns out to be @xmath128_j = \sqrt{\alpha-\beta_{n+1-j}}\,u_{i , n+1-j } , \label{defsri2}\ ] ] where @xmath124 is a constant chosen sufficiently large as to make @xmath207 for all terms in the sum that we keep .
then the modularity is given by @xmath208 with the community vectors @xmath135 defined according to eq . .
one of the striking properties of the laplacian matrix is that , as described in section [ specpart ] , it arises repeatedly in various different areas of graph theory .
it is natural to ask whether the modularity matrix also crops up in other areas . in this section
we describe briefly two other situations in which the modularity matrix appears , although neither has been viewed in terms of this matrix in the past , as far as we are aware . for our first example , suppose we have a quantity @xmath209 defined on the vertices @xmath127 of a network , such as degrees of vertices , ages of people in a social network , numbers of hits on web pages , and so forth . and let @xmath177 be the @xmath0-component vector whose elements are the @xmath209 .
then consider the quantity @xmath210 where here we will take the same definition for our null model that we have been using throughout .
observing that @xmath211 , we can rewrite @xmath212 as @xmath213 x_ix_j \nonumber\\ & = & { \sum_{ij } a_{ij } x_ix_j\over\sum_{ij } a_{ij } } - \biggl [ { \sum_{ij } a_{ij } x_i\over\sum_{ij } a_{ij } } \biggr]^2.\end{aligned}\ ] ] note that the ratios appearing in the second line are simply averages over all edges in the network , and hence @xmath212 has the form @xmath214 of a correlation function measuring the correlation of the values @xmath209 over all pairs of vertices joined by an edge in the network
. correlation functions of exactly this type have been considered previously as measures of so - called `` assortative mixing , '' the tendency for adjacent vertices in networks to have similar properties @xcite .
for example , if the quantity @xmath209 is just the degree @xmath15 of a vertex , then @xmath212 is the covariance of the degrees of adjacent vertices , which takes positive values if vertices tend to have similar degrees to their neighbors , high - degree vertices linking to other high - degree vertices and low to low , and negative values if high - degree links to low .
equation is not just a curiosity , but provides some insight concerning assortativity .
if we expand @xmath177 in terms of the eigenvectors @xmath98 of the modularity matrix , as we did for the modularity itself in eq . , we get @xmath215 where @xmath102 is again the @xmath16th largest eigenvalue of @xmath95 and @xmath216 . thus @xmath212 will have a large positive value if @xmath177 has a large component in the direction of one or more of the most positive eigenvectors of the modularity matrix , and similarly for large negative values .
now we recall that the leading eigenvectors of the modularity matrix also define the communities in the network and we see that there is a close relation between assortativity and community structure : networks will be assortative according to some property @xmath142 if the values of that property divide along the same lines as the communities in the network .
thus , for instance , a network will be assortative by degree if the degrees of the vertices are partitioned such that the high - degree vertices fall in one community and the low - degree vertices in another .
this lends additional force to the discussion given in the introduction , where we mentioned that different communities in networks are often found to have different average properties such as degree .
in fact , as we now see , this is probably the case for any property that displays significant assortative mixing , which includes an enormous variety of quantities measured in networks of all types .
thus , it is not merely an observation that different communities have different average properties it is an expected behavior in a network that has both community structure and assortativity . for our second example of other uses of the modularity matrix
, we consider centrality measures , one of the abiding interests of the network analysis community for many decades . in section [ singlevec ]
we argued that the magnitudes of the elements of the leading eigenvector of the modularity matrix give a measure of the `` strength '' with which vertices belong to their assigned communities .
thus these magnitudes define a kind of centrality index that quantifies how central vertices are in communities .
focusing on just a single eigenvector of the modularity matrix , however , is limiting .
as we have seen , all the eigenvectors contain useful information about community structure .
it is useful to ask what the appropriate measure is of strength of community membership when the information in all eigenvectors is taken into account .
given eq .
, the obvious candidate seems to be the projection of the vertex vector @xmath126 onto the community vector @xmath135 of the community to which vertex @xmath16 belongs .
unfortunately , this projection depends on the arbitrary parameter @xmath124 , which we introduced in eq . to get around problems caused by the negative eigenvalues of the modularity matrix
this in turn threatens to introduce arbitrariness into our centrality measure , which we would prefer to avoid .
so for the purposes of defining a centrality index we propose a slightly different formulation of the modularity , which is less appropriate for the optimization calculations that are the main topic of this paper , but more satisfactory for present purposes , as we will see .
suppose that there are @xmath79 positive eigenvalues of the modularity matrix and @xmath217 negative ones .
we define two new sets of vertex vectors @xmath218 and @xmath219 , of dimension @xmath79 and @xmath217 , thus : @xmath220_j & = & \sqrt{\beta_j}\,u_{ij},\\ \bigl [ { \mathbf}{y}_i \bigr]_j & = & \sqrt{-\beta_{n+1-j}}\,u_{i , n+1-j}.\end{aligned}\ ] ] ( note that @xmath221 since there is always at least one eigenvalue with value zero . ) in terms of these vectors the modularity , eq . ,
can be written as @xmath222 ^ 2 \nonumber\\ & & { } - \sum_{k=1}^c \sum_{j=1}^q \biggl [ \sum_{i=1}^n \sqrt{-\beta_{n+1-j}}\,u_{i , n+1-j } s_{ik } \biggr]^2 \nonumber\\ & = & \sum_{k=1}^c \sum_{j=1}^p \biggl [ \sum_{i\in g_k } \bigl [ { \mathbf}{x}_i \bigr]_j \biggr]^2 - \sum_{k=1}^c \sum_{j=1}^q \biggl
[ \sum_{i\in g_k } \bigl [ { \mathbf}{y}_i \bigr]_j \biggr]^2 \nonumber\\ & = & \sum_{k=1}^c | { \mathbf}{x}_k |^2 - \sum_{k=1}^c | { \mathbf}{y}_k |^2 , \label{qxy}\end{aligned}\ ] ] where @xmath133 is once again the set of vertices in community @xmath134 and the community vectors @xmath223 and @xmath224 are defined by @xmath225 this reformulation avoids the use of the arbitrary constant @xmath124 , thereby making the vertex vectors dependent only on the network structure and not on the way in which we choose to represent it .
equation separates out the positive and negative contributions to the modularity , the positive contributions coming from vertices that have large corresponding elements in the eigenvectors with positive eigenvalues , and conversely for the negative contributions .
the two contributions correspond respectively to the traditional community structure of sections [ modsec ] and [ specmod ] , and to the bipartite or @xmath134-partite structure discussed in section [ negative ] .
it is important to notice that while obviously the overall modularity can only be either positive or negative , it is entirely possible for individual vertices to simultaneously make both large positive and large negative contributions to that modularity . upon reflection , this is clearly reasonable : there is no reason why a single vertex can not have more connections than expected within its own community _ and _ more connections than expected to other communities . in a sense , eq .
may be a more fundamental representation of the modularity than eq . because it makes this separation transparent , even if it is in practice less suitable as a basis for modularity optimization .
we can now define precisely the quantity that plays the role previously played by the elements of the leading eigenvector in the single - eigenvector approximation : it is the projection of @xmath226 onto the relevant community vector @xmath223 , as we can see by writing the magnitude @xmath227 in eq . as @xmath228 where @xmath229 is the unit vector in the direction of @xmath223 .
thus each vertex vector makes a contribution to @xmath227 equal to its projection onto @xmath223 . in the approximation where we ignore all but the leading eigenvector
, this projection reduces to the ( magnitude of ) the appropriate element of that eigenvector , as in section [ singlevec ] .
the projection specifies how central vertex @xmath16 is in its own community in the traditional sense of having many connections within that community .
if this quantity is large then we will lose a large positive contribution to the modularity if we move the vertex to another community , which is to say that the vertex is a strong member of its current community .
but there is also a second measure for each vertex , the projection of @xmath230 onto @xmath224 .
this projection corresponds to a more unusual sort of centrality which is high if vertex @xmath16 has many connections to others _
outside _ its community .
this `` outsider '' centrality measure could also be useful in certain circumstances to identify individuals with strong external connections .
these two projections , however , do not take precisely the form that we expect of a centrality measure because they are functions not only of the vertex itself ( via @xmath226 or @xmath230 ) but also of the community in which it is placed ( via @xmath223 or @xmath224 ) . instead , therefore , let us consider the projection in the form @xmath231 , where @xmath232 is the angle between @xmath226 and @xmath223 .
the two parts of this expression are both of interest .
the first , the magnitude @xmath233 , measures how large a positive contribution vertex @xmath16 can _ potentially _ make to the modularity .
the vertex only actually makes a contribution this large if the vertex vector is aligned with the community vector , i.e. , if the vertex is , in a sense , `` in the middle '' of the community to which it belongs .
even a vertex for which @xmath233 is large may in practice make a small positive contribution to the modularity if @xmath226 is almost perpendicular to @xmath223 , i.e. , if the vertex is `` on the edge '' of the community .
the second part of the projection , the @xmath234 , is a measure precisely of the vertex s position in the middle or on the edge of its community . in the parlance of social network analysis ,
the vertex is either in the core of its community ( @xmath234 near 1 ) or in the periphery ( @xmath234 nearer 0 ) .
the cosine is a property both of the vertex and of the community .
let us focus here on the vector magnitudes and define two centrality measures for vertices in a network equal to the magnitudes of the vertex vectors @xmath226 and @xmath230 .
( if we prefer , we could use @xmath235 instead , which is slightly easier to calculate .
if , as is sometimes the case with centrality measures , we only care about relative rankings of vertices , then the two are equivalent . )
these centralities are now properties of the vertices alone and are independent of the way the network is divided into communities .
we notice , however , that @xmath233 and @xmath236 are not independent since @xmath237 almost all networks considered in the literature are simple graphs , meaning , among other things , that they have no self - edges ( edges that connect vertices to themselves ) and hence @xmath238 for all @xmath16 .
if the expected number of self - edges @xmath239 is also zero ( as seems sensible ) , then @xmath240 and we have @xmath241 for all @xmath16 .
thus there is actually only one centrality for simple graphs , not two .
in fact , the choice for @xmath64 that we and other authors have mostly used does allow self - edges ( and is in this sense slightly unrealistic
see @xcite ) , but @xmath242 is typically small for most vertices if @xmath2 is large ( and indeed vanishes as @xmath243 if degrees are bounded ) , and hence it is still true to a good approximation that @xmath244 and there is only one centrality .
in other words , we come to the nontrivial conclusion that the vertices with the greatest capacity for making positive contributions to the modularity also have the greatest capacity for making negative contributions .
the fundamental meaning of this centrality measure is thus that there are certain vertices that , as a consequence of their situation within the network , have the power to make substantial contributions , either positive or negative , to the overall modularity of the network .
for this reason , we call this centrality measure community centrality .
we define it to be equal to the vector magnitude @xmath233 .
an alternative way to view the community centrality is to consider how a vertex @xmath16 is situated among the other vertices in its immediate vicinity
its neighborhood in the network .
if we were to artificially construct a community from the vertices of this neighborhood , then that community would presumably have a community vector @xmath223 with direction close to @xmath226 , and hence the magnitude @xmath233 would be a good measure of the actual strength with which vertex @xmath16 belongs the community . thus vertices with high community centrality are ones that play a central role in their local neighborhood , regardless of where the official community boundaries may lie .
conversely , even when considered as the `` center of its world '' in this way , vertex @xmath16 can never play a central role in its neighborhood in this sense if @xmath233 is small . as an example , consider fig .
[ collabs ] , which shows results for community centrality for a network of coauthorships between scientists , scientists in this case who are themselves publishing on the topic of networks .
the network is similar to the one presented in ref .
@xcite but is based on more recent data , including publications up until early 2006 .
the network has a total of 1589 scientists in it , from a broad variety of fields , but only the 379 falling in the largest connected component are shown in the figure .
the diameters of the vertices in the figure are proportional to their community centrality ( actually to @xmath235see above ) , and the ten vertices having the highest centralities are highlighted .
a couple of remarks are worth making about the results . without naming specific names , we observe that all of the highlighted authors are group leaders or senior researchers of groups working in this area .
thus community centrality appears to live up to its name in this admittedly anecdotal example : it highlights those vertices that are central in their local communities .
second , while the centrality is correlated with degree ( @xmath245see the inset figure ) , the two are not perfectly correlated and in particular some vertices have quite high centrality while having relatively low degree .
this emphasizes the point that high centrality is an indicator of individuals who have more connections _ than expected _ within their neighborhood ( and hence potentially make a large contribution to the modularity ) , rather than simply having a lot of connections .
in this paper , we have studied the problem of detecting community structure in networks .
there is already a substantial body of theory supporting the view that community structure can be accurately quantified using the benefit function known as modularity and hence that communities can be detected by searching possible divisions of a network for ones that possess high modularity .
here we have demonstrated that the modularity can be succinctly expressed in terms of the eigenvalues and eigenvectors of a matrix we call the modularity matrix , which is a characteristic property of the network and is itself independent of any division of the network into communities . using this expression
we have derived a series of further results including several new and competitive algorithms for identifying communities , a method for detecting bipartite or @xmath134-partite structure in networks , and a new community centrality measure that identifies vertices that play a central role in the communities to which they belong .
we have demonstrated a variety of applications of our methods to real - world networks representing social , technological , and information networks .
these , however , are intended only as illustrations of the potential of these methods .
we hope that readers will feel encouraged to apply these or similar methods to other networks of scientific interest and we look forward to seeing the results .
the author thanks luis amaral , alex arenas , roger guimer , edward ionides , and david lusseau for useful and enjoyable conversations , and valdis krebs and david lusseau for providing network and other data used in the examples .
this work was funded in part by the national science foundation under grant number dms0405348 and by the james s. mcdonnell foundation .
g. p. garnett , j. p. hughes , r. m. anderson , b. p. stoner , s. o. aral , w. l. whittington , h. h. handsfield , and k. k. holmes , sexual mixing patterns of patients attending sexually transmitted diseases clinics .
_ sexually transmitted diseases _ * 23 * , 248257 ( 1996 ) .
s. o. aral , j. p. hughes , b. stoner , w. whittington , h. h. handsfield , r. m. anderson , and k. k. holmes , sexual mixing patterns in the spread of gonococcal and chlamydial infections . _ american journal of public health _ * 89 * , 825833 ( 1999 ) . a. capocci , v. d. p. servedio , g. caldarelli , and f. colaiori , detecting communities in large networks . in s. leonardi ( ed . ) , _ proceedings of the 3rd workshop on algorithms and models for the web graph _ , number 3243 in lecture notes in computer science , springer , berlin ( 2004 ) . h. zhou and r. lipowsky , network brownian motion : a new method to measure vertex - vertex proximity and to identify communities and subcommunities . in _
lecture notes in computer science _ , volume 3038 , pp . 10621069 , springer , new york ( 2004 ) .
p. pons and m. latapy , computing communities in large networks using random walks . in _ proceedings of the 20th international symposium on computer and information sciences
_ , volume 3733 of _ lecture notes in computer science _ , pp . 284293 , springer , new york ( 2005 ) .
o. goldschmidt and d. s. hochbaum , polynomial algorithm for the @xmath134-cut problem . in _ proceedings of the 29th annual ieee symposium on the foundations of computer science _ , pp .
444451 , institute of electrical and electronics engineers , new york ( 1988 ) .
m. bern , d. eppstein , and j. gilbert , provably good mesh generation . in _ proceedings of the 31st annual ieee symposium on the foundations of computer science _ , pp .
231241 , institute of electrical and electronics engineers , new york ( 1990 ) .
wei and c .- k .
cheng , toward efficient hierarchical designs by ratio cut partitioning . in _ proceedings of the ieee international conference on computer aided design _ , pp .
298301 , institute of electrical and electronics engineers , new york ( 1989 ) .
p. k. chan , m. d. f. schlag , and j. y. zien , spectral @xmath134-way ratio - cut partitioning and clustering .
in _ proceedings of the 30th international conference on design automation _ , pp . 749754 , association of computing machinery , new york ( 1993 ) .
s. white and p. smyth , a spectral clustering approach to finding communities in graphs . in h. kargupta , j. srivastava , c. kamath , and a. goodman ( eds . ) , _ proceedings of the 5th siam international conference on data mining _ , society for industrial and applied mathematics , philadelphia ( 2005 ) .
t. uczak , sparse random graphs with a given degree sequence . in a.
m. frieze and t. uczak ( eds . ) , _ proceedings of the symposium on random graphs , pozna 1989 _ , pp .
165182 , john wiley , new york ( 1992 ) .
d. lusseau , k. schneider , o. j. boisseau , p. haase , e. slooten , and s. m. dawson , the bottlenose dolphin community of doubtful sound features a large proportion of long - lasting associations .
can geographic isolation explain this unique trait ?
_ behavioral ecology and sociobiology _ * 54 * , 396405 ( 2003 ) . c. j. alpert and s .- z .
yao , spectral partitioning : the more eigenvectors , the better . in b.
t. preas , p. g. karger , b. s. nobandegani , and m. pedram ( eds . ) , _ proceedings of the 32nd international conference on design automation _ , pp .
195200 , association of computing machinery , new york , ny ( 1995 ) . | we consider the problem of detecting communities or modules in networks , groups of vertices with a higher - than - average density of edges connecting them .
previous work indicates that a robust approach to this problem is the maximization of the benefit function known as `` modularity '' over possible divisions of a network .
here we show that this maximization process can be written in terms of the eigenspectrum of a matrix we call the modularity matrix , which plays a role in community detection similar to that played by the graph laplacian in graph partitioning calculations .
this result leads us to a number of possible algorithms for detecting community structure , as well as several other results , including a spectral measure of bipartite structure in networks and a new centrality measure that identifies those vertices that occupy central positions within the communities to which they belong .
the algorithms and measures proposed are illustrated with applications to a variety of real - world complex networks . |
it has been studied years ago that the supersymmetric generalizations of the bosonic cp(n-1 ) sigma model is automatically an @xmath2 supersymmetric theory by the virtue of that the target space cp(n-1 ) is a khler manifold @xcite .
recently edalati and tong suggested and built an @xmath3 cp(n-1 ) heterotic sigma model by study of the low - energy dynamics of vortex strings in @xmath4 four - dimensional gauge theories@xcite .
later shifman and yung formulated a geometric representation for this heterotic model with the cp(n-1 ) target space for bosonic fields and an extra right - handed fermion which couples to the fermion fields of the @xmath5 cp(n-1 ) model@xcite , and thus breaks the supersymmetry to @xmath3 . in this paper , we follow the geometric formulation to discuss the fermionic zero modes of @xmath3 heterotic cp(1 ) sigma model in the instanton background .
nontrivial homotopy group structure of target space cp(1 ) allows for the bosonic instanton background of the form of a holomorphic function @xmath6 , where @xmath7 and @xmath8 are instanton size and center respectively . under such background
, the fermionic zero modes can be obtained by acting supercharges and superconformal charges to the instanton solution . since the heterotic deformation of the standard cp(1 ) sigma model only preserves half of four original supercharges , the supersymmetries and superconformal symmetries are partly broken .
the remaining fermionic generators are not sufficient to generate all fermionic zero modes from the instanton background .
therefore we will find zero modes by solving dirac equations . in section 2
, we will calculate the chiral anomaly at first to give the number of zero modes . in section 3
we will explicitly solve dirac equations of zero modes and use the result of section 2 to pick up those linear independent modes . in the last section we generalize this result to cp(n-1 ) heterotic sigma models .
we start from the chiral anomaly of the standard cp(1 ) sigma model .
the notations are in accordance with the reference @xcite .
the lagrangian of cp(1 ) is given as follows : @xmath9\nonumber\\ & = & \frac{2}{g_{0}^2\chi^2}\partial_{\mu}\phi^{\dagger}\partial^{\mu}\phi+\frac{2i}{g_{0}^2}\bar{\xi}\gamma^{\mu}(\partial_{\mu}\xi - ia_{\mu}\xi)+\frac{2}{g_{0}^2}(\bar{\xi}\xi)^2\end{aligned}\ ] ] where @xmath10 , @xmath11 and @xmath12 for convenience to calculate the chiral anomaly .
it has a chiral symmetry and thus an axial current : @xmath13 the chiral anomaly is well - known@xcite as : @xmath14 this anomaly term corresponds to the tadpole diagram : + now we are going to consider the chiral anomaly of heterotic cp(1 ) sigma model .
the lagrangian of the heterotic model for bilinear fermion terms is deformed from eq.(1 ) by adding terms with an extra right - handed fermion field @xmath15 @xcite .
since only bilinear terms of fermions need to be considered when calculating one loop diagrams , we just write them up : @xmath16 + \frac{2i}{g_{0}^2\chi^2}\bar{\psi}\gamma^{\mu}(\partial_{\mu}\psi-\frac{2}{\chi}\phi^{\dagger}\partial_{\mu}\phi\psi)\nonumber\\ & = & i\bar{\zeta}\gamma^{\mu}\frac{1+\gamma_5}{2}\partial_{\mu}\zeta+(\bar{\zeta}\gamma^{\mu}\frac{1+\gamma_5}{2}b_{\mu}\xi+h.c.)+ \frac{2i}{g_{0}^2}\bar{\xi}\gamma^{\mu}(\partial_{\mu}\xi - ia_{\mu}\xi)\end{aligned}\ ] ] where we define @xmath17 and @xmath18 . it is easy to see @xmath19 and the lagrangian is invariant under @xmath20 and @xmath21
. therefore the corresponding axial current is : @xmath22 we will show that the chiral anomaly of the heterotic model will not be corrected by the extra @xmath23 terms , and thus prove that the number of zero modes is the same as the standard cp(1 ) model . actually , to calculate this anomaly one needs to consider diagrams as follows : + + the diagrams b and c cancel with each other because both of them are finite , and thus free to shift integral variables .
the diagrams d and e also cancel each other
. it can be proved by using momentum @xmath24 times the sum of these two diagrams , @xmath25 + + by multiplying @xmath26 , the original converged digrams d and e become two divergent parts respectively .
however they are all logarithmically divergent , and thus still free to shift integral variables to prove the cancellation .
therefore the anomaly of heterotic case is still only from the tadpole diagram , and is the same as in the standard cp(1 ) case , see equation ( 3 ) .
for the standard cp(1 ) model , @xmath27 has four zero modes , under the instanton background @xmath28@xcite .
since the chiral anomaly is not corrected in the heterotic case , the number of zero modes should be still four .
we now proceed to derive the dirac equations for @xmath27 and @xmath23 from the heterotic lagrangian and wick - rotate them to euclidean space by @xmath29 .
we only keep linear terms in the equations of motion and solve them .
we will see that the solutions are also satisfied with the equations of motion when we add nonlinear terms . since we have already shown that the number of independent zero modes is four , the solutions solved in the equations of motions up to linear terms are all zero modes . from eq .
( 4 ) and instanton background @xmath30 , in spinor components we have : @xmath31 notice that eqs .
( 8) ( 9 ) and ( 11 ) are the same as in the standard cp(1 ) case
. therefore the solutions to these three equations should be the same as before .
they give all four zero modes .
since we have argued that the total number of zero modes is four , there is no extra zero modes raised from field @xmath32 , although the equation of motion of @xmath32 ( 10 ) is deformed by the @xmath33 term .
therefore we have : @xmath34 by the same argument , fermionic field @xmath23 and @xmath35 can not provide extra zero modes . from eqs .
( 6 ) ( 7 ) and ( 10 ) , we get : @xmath36 at first sight , the solutions seem inconsistent with the anomaly calculation , which gives the number of the difference between left and right handed fermions under instanton background , @xmath25 @xmath37 from solutions ( 12)-(15 ) , it seems that the first two solutions are right handed fermions while the last two are left handed , thus the number of the difference is zero .
however , because we wick - rotated to euclidean space to perform the calculation , in which the complex conjugate is not a well - defined involution for chiral fermions , the field @xmath38 is thus totally independent of @xmath27 .
moreover in euclidean space , the lagrangian is @xmath39 invariant .
therefore the field @xmath38 transformed under @xmath39 is the same as @xmath40 .
bearing this in mind , we can be back to the lagrangian ( 4 ) : @xmath41 now it is clear that , when wick - rotated to euclidean space , the field @xmath42 and @xmath43 we solved in eqs .
( 10 ) and ( 11 ) are in fact the fields @xmath44 and @xmath45 respectively , while the field @xmath38 plays the same role as @xmath40 in euclidean space . therefore the four solutions are all right handed fermions .
it is consistent with the anomaly calculation . as for the field @xmath23 , solution
( 16 ) shows that the extra field @xmath23 mixes with field @xmath27 , however it is linearly dependent of field @xmath27 , and thus does not produce extra zero modes as what we also expected from the result of the chiral anomaly calculation ( 17 ) . at last ,
one more worthy to mention is the solutions of field @xmath27 .
they are of the same form as in the standard cp(1 ) zero modes .
it means that not only the number of zero modes does not change , the zero modes of @xmath27 also receive no correction in the heterotic model .
in fact it is easy to understand this result from equation ( 8) and ( 9 ) .
although the supersymmetries are partly broken by deformation in the heterotic case , the classical equation of motion of @xmath27 up to linear terms are still @xmath1 supersymmetic invariant . therefore in this sense
the zero modes of @xmath27 still can be generated by supercharges and superconformal charges from the instanton background , even though half of the symmetries are broken in the heterotic model .
in addition , one can check that these solutions are satisfied with the equations of motion when adding nonlinear terms of fermions .
our conclusion to the heterotic cp(1 ) model is that the zero modes of @xmath27 are the same as in the standard cp(1 ) case , while the extra fermion @xmath23 does not provide extra zero modes .
this result can be simply generalized to the heterotic cp(n-1 ) model .
for the heterotic cp(n-1 ) case , the lagrangian containing bilinear terms of fermions is@xcite :
@xmath46 + ig_{i\bar{j}}\bar{\psi}^{\bar{j}}\gamma^{\mu}(\partial_{\mu}\psi^{i}+\gamma^{i}_{lk}\partial_{\mu}\phi^{l}\psi^{k})\end{aligned}\ ] ] hence , the equations of motion up to linear terms under instanton backgroud @xmath47 are : @xmath48 it is not hard to see the solutions are : @xmath49 where @xmath50 is the khler potential . from these solutions
we see that the number of zero modes in the heterotic cp(n-1 ) model is still @xmath51 , while @xmath23 can be linearly expressed by the zero modes of @xmath27 and thus not extra zero mode .
i am grateful for m. shifman who led me to the subject , and a. vainshtein for very useful and fruitful discussions and reviewing the final version of the manuscript .
i also want to thank m. sasseville for proofreading the text .
+ b. zumino , phys . lett . *
b87 * , 203 ( 1979 ) m. edalati and d. tong , jhep * 0705 * , 005 ( 2007 ) m. shifman and a. yung , phys . rev . *
d77 * , 125016 ( 2008 ) v. a. novikov , m. a. shifman , a. i. vainshtein and v. i. zakharov , phys . rept . * 116 * , 103 ( 1984 ) m. shifman , a. vainshtein and r. zwicky , j. phys . *
a39 * , 13005 ( 2006 ) | we discuss instanton fermionic zero modes in the heterotic @xmath0 modification of the cp(1 ) sigma model in two dimensions . by calculating its chiral anomaly
we prove that the number of fermionic zero modes is same as in the standard @xmath1 cp(1 ) case , and determine their explicit form . |
this work was supported in part by doe grant de - fg02 - 91er-40690 .
j. d. bekenstein , phys .
d * 7 * , 2333 ( 1973 ) . j. d. bekenstein , lett .
nuovo cim . * 11 * , 467 ( 1974 ) . c. rovelli and l. smolin , nucl .
b * 442 * , 593 ( 1995 ) [ erratum - ibid .
b * 456 * , 753 ( 1995 ) ] [ arxiv : gr - qc/9411005 ] .
k. s. .
thorne , r. h. .
price and d. a. .
macdonald , _ new haven , usa : yale univ .
( 1986 ) 367p _ o. lunin and s. d. mathur , phys .
lett . * 88 * , 211303 ( 2002 ) [ arxiv : hep - th/0202072 ] ; o. lunin , j. m. maldacena and l. maoz , arxiv : hep - th/0212210 ; i. kanitscheider , k. skenderis and m. taylor , jhep * 0706 * , 056 ( 2007 ) [ arxiv:0704.0690 [ hep - th ] ] . s. giusto , s. d. mathur and a. saxena , nucl .
b * 710 * , 425 ( 2005 ) [ arxiv : hep - th/0406103 ] ; i. bena and n. p. warner , adv .
* 9 * , 667 ( 2005 ) [ arxiv : hep - th/0408106 ] ; v. jejjala , o. madden , s. f. ross and g. titchener , phys .
d * 71 * , 124030 ( 2005 ) [ arxiv : hep - th/0504181 ] .
v. balasubramanian , e. g. gimon and t. s. levi , arxiv : hep - th/0606118 ; j. de boer , f. denef , s. el - showk , i. messamah and d. van den bleeken , jhep * 0811 * , 050 ( 2008 ) [ arxiv:0802.2257 [ hep - th ] ] ; s. d. mathur , arxiv:0810.4525 [ hep - th ] .
s. d. mathur , arxiv:0706.3884 [ hep - th ] .
v. cardoso , o. j. c. dias , j. l. hovdebo and r. c. myers , phys .
d * 73 * , 064031 ( 2006 ) [ arxiv : hep - th/0512277 ] ; b. d. chowdhury and s. d. mathur , class .
* 25 * , 135005 ( 2008 ) [ arxiv:0711.4817 [ hep - th ] ] .
s. d. mathur , arxiv:0805.3716 [ hep - th ] .
s. d. mathur , int .
j. mod .
d * 18 * , 2215 ( 2009 ) [ arxiv:0905.4483 [ hep - th ] ] .
p. hayden and j. preskill , jhep * 0709 * , 120 ( 2007 ) [ arxiv:0708.4025 [ hep - th ] ] ; y. sekino and l. susskind , jhep * 0810 * , 065 ( 2008 ) [ arxiv:0808.2096 [ hep - th ] ] . j. m. maldacena , adv .
* 2 * , 231 ( 1998 ) [ int .
j. theor .
phys . * 38 * , 1113 ( 1999 ) ] [ arxiv : hep - th/9711200 ] .
m. parikh and f. wilczek , phys . rev .
d * 58 * , 064011 ( 1998 ) [ arxiv : gr - qc/9712077 ] ; t. shimomura , t. okamura and t. mishima , int .
j. mod .
d * 11 * , 789 ( 2002 ) ; p. kovtun , d. t. son and a. o. starinets , jhep * 0310 * , 064 ( 2003 ) [ arxiv : hep - th/0309213 ] ; v. cardoso , o. j. c. dias and l. gualtieri , int .
j. mod .
d * 17 * , 505 ( 2008 ) [ arxiv:0705.2777 [ hep - th ] ] ; m. natsuume , phys . rev .
d * 78 * , 066010 ( 2008 ) [ arxiv:0807.1392 [ hep - th ] ] ; n.
iqbal and h. liu , phys .
d * 79 * , 025023 ( 2009 ) [ arxiv:0809.3808 [ hep - th ] ] ; h. culetu , arxiv:0906.0920 [ gr - qc ] ; c. eling and y. oz , jhep * 1002 * , 069 ( 2010 ) [ arxiv:0906.4999 [ hep - th ] ] . | are there any degrees of freedom on the black hole horizon ? using the ` membrane paradigm ' we can reproduce coarse - grained physics outside the hole by assuming a fictitious membrane just outside the horizon . but to solve the information puzzle we need ` real ' degrees of freedom at the horizon , which can modify hawking s evolution of quantum modes .
we argue that recent results on gravitational microstates imply a set of real degrees of freedom just outside the horizon ; the state of the hole is a linear combination of rapidly oscillating gravitational solutions with support concentrated just outside the horizon radius .
the collective behavior of these microstate solutions may give a realization of the membrane paradigm , with the fictitious membrane now replaced by real , explicit degrees of freedom .
epsf = = = = = = = = ============= == = = membrane paradigm realized ?
+ + department of physics , + the ohio state university , + columbus , oh 43210 , usa + .2 in [email protected] 1.0 true in a longstanding puzzle in quantum gravity is the following : is there anything at the black hole horizon ?
there are many reasons to take the horizon seriously as a place where ` physics happens ' .
the bekenstein entropy @xmath0 of the hole is proportional to the area of the horizon @xcite .
this suggests that we attach one ` bit ' of information to each planck area of the horizon , and perhaps quantize the horizon area to give an integral number of bits @xcite . in loop quantum gravity
the entropy arises from spin degrees of freedom on links that intersect the horizon @xcite . in the schwarzschild
coordinate frame , all matter that ever fell towards the horizon appears to be frozen forever just outside the horizon .
this picture was concretized in the ` membrane paradigm ' @xcite , where one puts an artificial cutoff surface just outside the horizon , and finds boundary conditions at this surface which will reproduce correct mechanical and thermal effects in the region outside this surface .
these boundary conditions can be summarized by imagining a fictitious membrane placed at this surface , with appropriate mechanical and thermal properties . in particular , there is a very high temperature near the membrane ,
so one might imagine that an infalling observer would get burnt up and eventually reradiated out to infinity @xcite .
this would be good , because to avoid the information paradox we need to have some way of returning the information in infalling objects back to infinity .
but there turns out to be a flaw in these attempts to assign genuine dynamics to the horizon .
it is well known that the schwarzschild coordinate system fails at the horizon , and the regular coordinates there the kruskal coordinates show that _ nothing _ special happens at the horizon of the traditional black hole .
nothing can ` stay ' at the horizon ; all matter must fall in towards the singularity .
the high temperature in the schwarzschild frame disappears in the kruskal frame ; it is seen to be nothing but the rindler tempertaure observed in an accelerated system of coordinates .
careful renormalization of the stress tensor shows that there is no energy density at the horizon @xcite ; the energy of rindler quanta is cancelled by vacuum polarization energy , since the physical vacuum at the horizon is naturally described in kruskal rather than schwarzschild coordinates . what then , about the fate of information falling into the hole ? over the past decade , many physicists had settled down to the following set of beliefs ( i ) information is not lost ; it emerges in the hawking radiation from the hole ( ii ) nothing special happens at the horizon , because the kruskal description looks like the logical one ( iii ) the above two beliefs are consistent because we need only small delicate correlations among hawking quanta to carry the information , and subleading corrections to hawking s computation @xcite should be able to encode the required information in the quanta .
but if these beliefs were correct , one has to wonder why over the years no one produced a model showing how it was possible for information to get encoded in the radiation while preserving a near - semiclassical hamiltonian at the horizon .
the reason for this failure emerged last year , when it was proved that ( iii ) is in fact incorrect : it is _ not _ possible to make small corrections to hawking s leading order state and recover any significant information in the radiation @xcite .
more precisely , if matrix elements of the hamiltonian for low energy modes are within @xmath1 of their semiclassical values , then no more than a fraction @xmath2 of the information can be encoded in the radiation .
the proof of this theorem used a well known inequality from quantum information theory called ` strong subadditivity ' @xcite .
but no _ elementary _ proof of this inequality is known , so it is perhaps not surprising that its consequences were not built into our naive thinking about encoding of information in the radiation .
it would now seem that we are back to square one on the information puzzle .
must we reconcile ourselves to information loss ?
we will now argue that another set of results over the past few years gives us a picture of how there can be real dynamics at the horizon , which will lead to an escape of information in the radiation .
a black hole has given values of of its conserved quantum numbers : mass , charge and angular momentum . in earlier work ,
one assumed a simple ansatz for the metric .
thus for the nonrotating hole , one assumed independence of the angular directions @xmath3 for all metric components . in a higher dimensional gravity theory ,
the extra directions have to be assumed compact , and the ansatz assumed no dependence of the metric on the compact coordinates @xmath4 either .
einstein s equations then gave a unique solution , possessing a horizon and a singularity the traditional black hole .
but it has been now recognized that there are @xmath5 $ ] _ different _ solutions of einstein s equations with the _ same _ quantum numbers @xcite .
all are regular , with no horizon and no singularity .
none of them are spherically symmetric , and all involve the compact directions in a nontrivial way .
so it was the simplicity of our initial ansatz which led us to miss these solutions .
further , string theory has allowed us to list all possible states for simple sets of quantum numbers .
the regular states mentioned above map in a one - to one and onto way onto the set of all allowed states , while the spherical solution with horizon and singularity is not realized at all .
this situation gives us an appealing way out of the information problem : \(a ) the region far from the hole has an infinite dimensional phase space due to its infinite volume .
very close to the horizon , however , there is again a very large phase space @xcite , from the @xmath5 $ ] solutions termed microstates giving the entropy of the hole ( fig.[fone ] ) .
\(b ) most of these solutions have nontrivial structure only very close to the horizon .
this is because rapid variation of the metric with position costs a lot of energy , and the large redshift near @xmath6 is needed to lower this to the given total energy @xmath7 .
the detailed structure of the solutions can be inferred from the explicitly solutions which have been constructed .
the geometry has _ ergoregions _ , with the direction of frame dragging changing rapidly from place to place ( fig.[ftwo ] ) .
this rapid space - time time dependence is what keeps the geometrical structure from collapsing towards @xmath8 .
( for an analogy , see fig.[fthree ] ; a circular elastic band shrinks to a point , but an rapidly vibrating band stays open with a nontrivial average radius . )
\(c ) consider a spherically symmetric shell made of gravitational radiation , collapsing to make a hole .
the motion upto @xmath9 is described by traditional dynamics in the schwarzschild frame ( @xmath10 ) .
but at @xmath11 the wavefunction of the shell can spread over the @xmath5 $ ] alternative solutions of einstein s equations .
the amplitude for this spread can be estimated by computing the _ tunneling _
amplitude between the shell and one of the microstate solutions .
this amplitude @xmath12 $ ] is very _ small _ , since the involved objects are macroscopic and the classical action of tunneling @xmath13 is big . but
the _ number _ of states that we can tunnel to very _ large _
, given by @xmath5 $ ] .
it turns out that @xmath14 @xcite , so instead of making a long lived black hole , the shell state becomes a linear superposition of microstate solutions in a time which is much shorter than the hawking evaporation time @xcite .
\(d ) it has been checked that ergoregion emission from the microstate solutions reproduces exactly the hawking radiation rate expected from those solutions @xcite . but this emission does not involve information loss , since there is no horizon .
\(e ) the further evolution of the collapsing shell is given by the evolution of the linear superposition of microstates into which it has tunneled .
it is plausible that the short time evolution of the collective modes of these complicated microstates can be mimicked by evolution in the interior of the traditional black hole geometry .
this effective evolution would then corresponds to the interior of the black hole in the penrose diagram .
it is important to note here that the typical microstate is not a classical solution in which infalling quanta scatter around in a zig - zag path .
instead , the large phase space of fuzzball solutions creates a very large ` coordination number ' the number of states that can be reached from a given state by a small action @xmath15 .
( this may make the black hole a ` fast scrambler ' @xcite . )
an infalling heavy object excites a linear combination of fuzzball states , similar to the ` fermi - golden rule ' absorption encountered when we have a closely spaced band of energy levels .
different objects excite different degrees of freedom to leading order , since the number of excitable states is so large ; the small overlap between different sets of excited states gives the residual interaction that mimics the usual ( weak ) gravitational interaction between different quanta ( the dual cft description of this effect was noted in @xcite ) .
thus the effective dynamics of absorption into fuzzballs gets related to motion in gently curved space , in a manner similar to absorption by the large number of open string degrees of freedom on a stack of d - branes : the infalling quantum excites these open strings , different quanta typically excite different strings ( due to the large number of these strings ) , and the small overlap between the set of excited strings generates the effective gravitational interactions between infalling quanta .
thus in this picture we would think of both ads / cft duality @xcite and the above mentioned effective description of fuzzballs as arising from the same underlying physics : absorption into a dense set of levels with high coordination number .
with this picture we seem to have an explicit realization of the membrane paradigm . in several earlier works , effective gravitational dynamics or dual cft descriptions
have been used to infer an imaginary ` membrane ' that may sit at the horizon @xcite . by contrast
, we are now addressing how gravity manages to actually produce something at the horizon that is capable of sending information back out in the hawking radiation . in our description
the membrane has real degrees of freedom on it , given by the rapidly oscillating solutions of einstein s equations that correspond to the black hole microstates
. it would be exciting to explore if the collective dynamics of the microstates reproduces the very precise properties of the fictitious membrane of the membrane paradigm . |
the ability of polymers to considerably change the large - scale statistics of the advecting flow has important practical applications , drag reduction being one of the most relevant ones ( see * ? ? ?
polymers affect the dynamics of the advecting velocity field only if they are highly elongated . understanding how a single polymer chain is stretched by a random flow
is thus the first issue to address in the study of hydrodynamical properties of polymer solutions . at equilibrium ,
the radial shape of coiled polymers is spherical due to their entropy . when placed in a non - homogeneous flow ,
polymers are deformed and stretched by the gradients of the velocity .
the product of the longest relaxation time of polymers and the characteristic rate of deformation is called the weissenberg number @xmath0 . for small @xmath0
the entropic force prevails and polymers are in the coiled state .
when @xmath0 exceeds a critical value the molecules become highly elongated and their extension increases sharply .
this phenomenon is called coil stretch transition and the critical weissenberg number is known to be approximately one .
we investigate the statistics of polymer extension in the fene ( finite extensible nonlinear elastic ) dumbbell model ( see e.g. * ? ? ?
* for a review ) .
a polymer is described as two beads joined by an elastic spring .
the elastic force diverges as the elongation of polymers attains its maximum value , @xmath1 , and this gives a large - extension cutoff .
consequently the stationary regime exists however strong the velocity gradients are .
since attention is directed only to the dynamics of a single molecule , the feedback on the advecting flow is disregarded . to allow analytical progress ,
the random flow is chosen to have the batchelor - kraichnan statistics .
this means that the considered flow is gaussian , white in time and linear in space .
the batchelor - kraichnan model is a fully solvable model for passive turbulent transport which can provide useful connections between theory and real behaviours ( see * ? ? ?
* for a review on the applications to scalar and magnetic fields ) .
the results should be intended as a qualitative description of real polymer dynamics .
we derive the complete form of the stationary probability density function ( p.d.f . ) of polymer extension and describe how the statistics of polymer stretching changes with increasing velocity gradients . concerning the statistics at finite times , we compute the typical time needed for the system to reach the steady state and predict how it depends on the maximum length of polymers and the stretching by the flow .
our analysis shows that the coil
stretch transition of polymers is characterised by an anomalous dynamics in time .
stretch transition was predicted in 1974 for shear and hyperbolic flows @xcite and has been widely studied experimentally for such flows @xcite . in contrast ,
the experimental study of polymer dynamics in random flows is a very recent achievement .
this is due to the difficulty in generating a flow that is random at scales comparable with the size of polymers ( about 100 @xmath2 m ) .
this difficulty can be overcome thanks to the elastic turbulence discovered by @xcite ; the flow of a highly elastic polymer solution at low reynolds numbers , but large @xmath0 has all the main properties of fully developed turbulence .
therefore , in solutions of sufficiently elastic polymers it is possible to excite turbulent motion in exceedingly small volumes . exploiting elastic turbulence in polymer solutions
, @xcite thus investigated the stretching and the deformation of a single dna molecule in a three - dimensional random flow .
theoretical studies concerning the coil
stretch transition in random flows focused mainly on the hookean dumbbell model @xcite .
this model is suitable only for the coiled state ( @xmath3 ) since the linear force in principle allows infinite extensions and for large _ wi _ polymers can become more and more elongated under the action of velocity gradients . for @xmath4 a stationary p.d.f .
of the extension no longer exists and this behaviour was conjectured to coincide with the coil stretch transition . to overcome this oversimplification ,
the maximum length of polymers must be taken into account .
one possibility is to replace the hookean force by a nonlinear elastic force .
@xcite obtained the large - value tail of the stationary p.d.f .
of the extension for a general anharmonic force ; such approximate analysis was subsequently applied by @xcite to the fene model . here
we exactly derive the complete statistics of polymer stretching within the context of the fene dumbbell model at general _
wi_. concerning the statistics at finite times ,
preliminary results were obtained for the hookean model by @xcite . there , the relaxation time to the stationary regime could be defined only in the coiled state .
@xcite thus derived the behaviour of the relaxation time for very small _ wi _ and observed a divergence for @xmath5 ; this suggested a critical behaviour close to the coil stretch transition .
we present the first prediction of the complete dependence of the transient relaxation time on @xmath0 and @xmath1 with the more realistic fene model .
the rest of this paper is organised as follows . in
[ sec : main ] we introduce the model and present the main results . the stationary p.d.f . of the elongation and the transient relaxation time
are computed in
[ sec : staz ] and in [ sec : relax ] respectively . in [ sec : conclusions ] we discuss the relevance of our results for experiments and numerical simulations .
in elastic dumbbell models a polymer is described as two beads connected by a spring .
the beads represent the ends of the molecule and their separation is a measure of the extension . the beads experience : ( a ) a hydrodynamic drag force modeled by the stokes law ; ( b ) a brownian force due to thermal fluctuations of the fluid ; ( c ) an elastic force due to the spring connecting one bead to the other .
we consider two - dimensional and three - dimensional flows indifferently , the dimension of the flow being denoted by @xmath6 .
since in physical applications the elongation of polymers is always smaller than the viscous scale of the flow , the dumbbell is assumed to move in a linear velocity field @xmath7 .
inertial effects and hydrodynamic interactions between the beads are neglected .
consequently , the separation vector between the beads , @xmath8 , evolves according to the stochastic differential equation ( see * ? ? ?
* ) @xmath9 where @xmath10 is the equilibrium length of the polymer , @xmath11 is its relaxation time in the absence of flow , and @xmath12 is a @xmath6-dimensional brownian motion which accounts for thermal noise . in the fene dumbbell model , the elastic force @xmath13 takes the form @xmath14 , where @xmath1 denotes the maximum extension of the molecule . in physical applications
the ratio @xmath15 usually lies between 10 and 100 @xcite .
the length of the vector @xmath8 is a measure of the extension of the polymer . within the kraichnan model
@xmath16 is a statistically homogeneous gaussian field with zero mean and second - order correlation @xcite @xmath17 in the so - called batchelor regime the flow is assumed to be smooth in space .
if we further impose incompressibility and statistical invariance with respect to reflections and rotations , the tensor @xmath18 must take the form @xcite @xmath19,\ ] ] where @xmath20 represents the eddy diffusivity of the flow and @xmath21 determines the intensity of turbulent fluctuations . in random
flows the weissenberg number can be defined as @xmath22 , where @xmath23 is the maximum lyapunov exponent of the flow , that is the average logarithmic growth rate of nearby fluid particle separations .
the maximum lyapunov exponent of the batchelor - kraichnan flow has asymptotically a gaussian p.d.f . with mean value @xmath24 and variance @xmath25 @xcite .
the statistics of polymer elongation is described by the p.d.f . of the norm of @xmath8
averaged over velocity realizations , the average p.d.f .
of the elongation does not depend on the point of application of the vector @xmath8 . ] : @xmath26 , where @xmath27 denotes integration over angular variables . when the flow @xmath28 has the batchelor - kraichnan statistics @xmath29 obeys a one - dimensional fokker - planck equation with nontrivial drift and diffusion coefficients ( see [ sec : staz ] ) . under reflecting boundary conditions
( the probability does not flow outside the domain of definition ) the system reaches a steady state for all _ _ wi _ _ ( see * ? ? ? * ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
] . the stationary p.d.f . of the elongation , @xmath30 ,
has the form ( see [ sec : staz ] ) @xmath31 where @xmath32^{-1}$ ] and @xmath33 is the normalization coefficient ( see below ) .
the stationary p.d.f .
is shown in figure [ fig : staz ] for different @xmath0 . for small elongations compared to the equilibrium length
, @xmath34 scales as @xmath35 ; this result holds for a general elastic force since the left tail of @xmath36 comes from the events where the elastic force dominates and equation reduces to a @xmath6-dimensional langevin equation ( for physically meaningful elastic interactions @xmath37 should scale as @xmath8 for @xmath38 ) . for intermediate extensions , @xmath39 , the stationary p.d.f .
is proportional to the power law @xmath40 in accordance with the prediction of @xcite . for large elongations
@xmath34 scales as @xmath41 and vanishes for @xmath42 . in practical applications , @xmath43 ,
the exponent @xmath44 is approximatively @xmath45 , as predicted by @xcite . obviously , when @xmath46 and @xmath3 , @xmath47 tends to the stationary solution of the hookean model ( see * ? ?
( @xmath48).,height=188 ] the maximum of the p.d.f . ,
@xmath49 , determines the fraction of polymers which are highly stretched .
the graph of @xmath49 as a function of @xmath0 is shown in figure [ fig : max ] . when @xmath0 is smaller than one , @xmath49 is of the order of @xmath10 and most of polymers have the coiled equilibrium configuration . with increasing @xmath0
the most probable elongation @xmath49 grows slowly until @xmath0 exceeds @xmath50 .
then , a sharp transition occurs to a strongly elongated state .
this can be appreciated from the behaviour of the first - order derivative of @xmath49 as a function of @xmath0 ( figure [ fig : max ] ) . as @xmath0 becomes very large , @xmath49 approaches @xmath1 .
the same analysis holds for the average extension @xmath51 , apart from the fact that it starts increasing for a smaller @xmath0 and its limiting value is @xmath52 ( see figure [ fig : max ] ) .
it is worth noticing that the coil
stretch transition becomes sharper and sharper with increasing @xmath1 ( not plotted ) .
the normalized r.m.s .
value of the extension , @xmath53 , @xmath54 , is represented in figure [ fig : sigmaskew ] .
it increases at low @xmath0 until it reaches a maximum value ; then @xmath55 is compensated by the sharp increase in @xmath51 and at large @xmath0 the rescaled r.m.s .
eventually relaxes to the constant value @xmath56 .
the skewness @xmath57/\sigma^3 $ ] is positive for small @xmath0 and becomes negative at large @xmath0 ( figure [ fig : sigmaskew ] ) accordingly with the qualitative behaviour of the stationary p.d.f .
( figure [ fig : staz ] ) . the maximum of skewness in the neighbourhood of the coil
stretch transition can be easily understood as follows . at low @xmath0 the p.d.f .
is peaked at @xmath10 and the skewness is positive . with increasing @xmath0
the right tail starts raising , but @xmath51 is still of the order of @xmath10 : the skewness , therefore , increases and achieves is maximum value . beyond the coil
stretch transition @xmath51 starts moving towards the maximum extension and the skewness decreases until it becomes negative at large @xmath0 , that is when the p.d.f .
has a long left tail .
the limiting value of @xmath58 for @xmath59 is @xmath60 . and average rescaled extension @xmath61 as functions of the weissenberg number @xmath0 ( @xmath62 , @xmath63 , @xmath64 ) .
right : first derivative of @xmath65 with respect to @xmath0.,title="fig:",height=151 ] and average rescaled extension @xmath61 as functions of the weissenberg number @xmath0 ( @xmath62 , @xmath63 , @xmath64 ) .
right : first derivative of @xmath65 with respect to @xmath0.,title="fig:",height=151 ] as a function of the weissenberg number @xmath0 ( @xmath62 , @xmath63 , @xmath64 ) .
right : skewness @xmath58 _ vs _
@xmath0 for the same values of the parameters.,title="fig:",height=151 ] as a function of the weissenberg number @xmath0 ( @xmath62 , @xmath63 , @xmath64 ) .
right : skewness @xmath58 _ vs _
@xmath0 for the same values of the parameters.,title="fig:",height=151 ] we now turn to the time dependence of the p.d.f . of the elongation . starting from an initial condition peaked at @xmath10 ,
the system relaxes to the stationary regime described by .
the time needed to reach the stationary regime , @xmath66 , is solution of a transcendental equation which involves continued fractions ( see [ sec : relax ] ) . for small weissenberg numbers , @xmath67
, the transient relaxation time @xmath66 behaves according to the prediction of the linear model @xcite : @xmath68^{-1}\ ] ] independently of @xmath1 ( see inset in figure [ fig : tmax ] ) . in the proximity of the coil
stretch transition @xmath66 displays a maximum as a function of the @xmath0 .
the relaxation is exceptionally slow in this range of @xmath0 because the stationary regime results from the competition between the coiled state and the highly stretched state .
the position and the value of the maximum relaxation time @xmath69 depend on the cutoff @xmath1 ( figure [ fig : tmax ] ) . as the maximum allowed extension of polymers increases
, @xmath69 is closer and closer to @xmath5 and grows ; at large @xmath1 the fene model should indeed match the hookean model , where @xmath66 diverges as @xmath0 tends to one @xcite . for very large weissenberg numbers the stretching time is small compared to @xmath11 and the molecules are expected to rapidly reach the highly stretched configuration .
hence , @xmath66 vanishes as @xmath0 tends to infinity .
a numerical fit shows that @xmath66 scales as @xmath70 at large @xmath0 . in the next sections we explicitly derive and the equation for @xmath66 .
, as a function of the weissenberg number @xmath0 for three different values of @xmath15 ( @xmath62 ) .
the inset shows the linear - model approximation given by equation ( dashed line ) for @xmath48 : the agreement is good up to @xmath71 .
right : dependence of the maximum rescaled relaxation time @xmath72 on the maximum relative extension of polymers @xmath15.,title="fig:",height=170 ] , as a function of the weissenberg number @xmath0 for three different values of @xmath15 ( @xmath62 ) .
the inset shows the linear - model approximation given by equation ( dashed line ) for @xmath48 : the agreement is good up to @xmath71 .
right : dependence of the maximum rescaled relaxation time @xmath72 on the maximum relative extension of polymers @xmath15.,title="fig:",height=170 ]
for a fixed realization of the velocity field the p.d.f . of the end - to - end vector , @xmath73 , satisfies the fokker - planck equation associated with ( see e.g. * ? ? ?
* ) : @xmath74= \frac{r_0 ^ 2}{\tau}\ , \nabla^2_{\bm r } p.\ ] ] to obtain an equation for @xmath75 , we have to average the above equation over the velocity realizations and integrate the result over angular variables .
the terms of the type @xmath76 in general do not lead to a closed form for the mean p.d.f . and
a closed equation can not be deduced from .
the gaussianity and the @xmath77-correlation in time of the batchelor - kraichnan model provide an exact closure . exploiting the novikov - furutsu formula
( see e.g. * ? ? ?
* ) , we obtain : @xmath78 $ ] with @xmath79 $ ] .
we can thus derive from a one - dimensional fokker - planck equation for @xmath75 : @xmath80 + \partial^2_r\ , [ b(r){\cal p}(r , s)],\ ] ] where the time has been rescaled with @xmath11 , @xmath81 , and the drift and diffusion coefficients have the form @xmath82 the coefficients @xmath83 and @xmath84 are time independent due to the stationarity of the advecting flow . if @xmath10 is set to zero , then equation reduces to the approximate equation for the large - value tail of the p.d.f .
derived by @xcite . to solve
we impose reflecting boundary conditions , that is the probability current associated with the solution , @xmath85 $ ] , vanishes in @xmath86 and @xmath42 for all @xmath87 .
this means that there is no flow of probability through the boundaries of the domain . under these conditions the stationary p.d.f . of
the elongation takes the form @xcite @xmath88\ ] ] where the constant @xmath89 and the lower integration limit @xmath90 are fixed by the normalization condition .
the above formula holds for a general elastic force of the form @xmath91 . replacing the force of the fene model into
, we thus obtain with @xmath92 the function @xmath93 in the normalization coefficient denotes the hypergeometric function .
the time dependent solution of the fokker - planck can be obtained by separation of variables @xcite . in other words ,
@xmath94 can be sought in the form @xmath95 where the coefficients @xmath96 are fixed by the initial condition @xmath97 and @xmath98 , @xmath99 are respectively the eigenvalues and the eigenfunctions of the ordinary differential equation @xmath100- \frac{d}{dr}[a(r)\ , p_k(r ) ] + \mu_k \ , p_k(r)=0.\ ] ] the above equation should be solved with reflecting boundary conditions : @xmath101 , @xmath102 being the probability current associated with the eigenfunction @xmath103 . it can be shown that the eigenvalues @xmath98 are real and non - negative , @xmath47 belonging to the eigenvalue @xmath104 @xcite .
as we will see , the eigenvalues form a countable set and may be arranged in ascending order : @xmath105 .
the reciprocal of @xmath106 , therefore , is the time of relaxation to the stationary regime rescaled by @xmath11 .
equation is a second - order linear differential equation with four regular singularities in the complex plane . by the change of dependent and independent variable @xmath107 , @xmath108
, such equation can be transformed into a standard heun equation for the function @xmath109 ( see * ? ? ? * for a review ) : @xmath110 where @xmath111 \alpha= h+\frac{d}{4}-\frac{1}{4}\sqrt{d\left(d-\frac{4\mu_k}{\textit{wi } } \right ) } & & \beta= h+\frac{d}{4}+\frac{1}{4}\sqrt{d\left(d-\frac{4\mu_k}{\textit{wi } } \right ) } \\[0.3 cm ] \gamma = d/2 & \qquad \delta = h\qquad & \epsilon=1+h\end{aligned}\ ] ] with @xmath32^{-1}$ ] . reflecting boundary conditions for @xmath103 map into the following limiting conditions for @xmath112 :
@xmath113 the heun equation is the general fuchsian equation with four singularities . in the standard form
the singular points are @xmath114 .
let @xmath115 be a generic singularity of .
from the theory of fuchsian equations , the local behaviour of @xmath109 near @xmath115 is specified by the characteristic exponents @xmath116 , @xmath117 associated with @xmath115 ( see e.g. * ? ? ?
if @xmath118 is not integer , in a neighbourhood of @xmath115 which excludes the nearest other singularity @xmath109 can be written in the form @xmath119 , where @xmath120 , @xmath121 are constant and @xmath122 , @xmath123 are analytic functions such that @xmath124 , @xmath125 . if @xmath118 is integer and @xmath126 , the function @xmath123 can be no longer analytic in @xmath115 and involve the function @xmath127 .
the singularity @xmath128 has characteristic exponents 0 and @xmath129 ; the singularity @xmath130 has characteristic exponents 0 and @xmath131 . in physical applications we can exclude the situation where @xmath77 is integer . on the contrary
, @xmath129 is zero when @xmath132 .
consider first the case @xmath62 , where there are not logarithmic singularities in @xmath128 . to fulfill conditions
, @xmath112 must be simultaneously a local solution about @xmath128 and @xmath130 , in both cases belonging to the exponent 0 .
such a solution is called a heun function of class i relative to the points 0 and 1 , and exists only for a countable set of values of @xmath133 and hence of @xmath98 ( see * ? ? ?
the condition for the aforementioned heun function to exist leads to a transcendental equation for the eigenvalues @xmath98 @xcite : @xmath134 where @xmath135 l_i&=&q+ai(i+\omega)-\dfrac{\epsilon\ , i(i+\omega)(\gamma-\delta)+ [ i(i+\omega)+\alpha\beta][2i(i+\omega)+\gamma(\omega-1)]}% { ( 2i+\omega-1)(2i+\omega+1 ) } \\[0.2 cm ] m_i&=&\dfrac{(i+1)(i+\omega-\alpha+1)(i+\omega-\beta+1)(i+\delta)}% { ( 2i+\omega+1)(2i+\omega+2)}\end{aligned}\ ] ] with @xmath136 .
the rescaled relaxation time @xmath137 is then the reciprocal of the lowest non - zero solution of . in the case
@xmath132 the conclusions are unchanged since the solution involving a logarithm in the neighbourhood of @xmath128 should be discarded .
we solved numerically : the continued fraction was computed by the modified lentz s method and the first non - zero solution was evaluated by the root false position method ( see e.g. * ? ? ? * ) .
the goal of the paper was to investigate polymer stretching in a turbulent flow within the context of a fully solvable model .
the statistical features of the batchelor - kraichnan flow allow us to derive the complete form of the stationary p.d.f . of polymer elongation for a general elastic force . when specializing to finitely extensible polymers we recover the main properties of polymer dynamics in real turbulent flows and compute the time of relaxation to the stationary regime
it should be noted that the velocity field we consider is statistically isotropic .
together with the @xmath77-correlation in time , this is a key assumption in order to derive a fully analytical solution of the problem . in the experimental setup of @xcite the elastic turbulent flow
is superimposed to a mean shear flow .
the long - time statistics of polymer extension in the presence of a mean shear has been recently considered by @xcite .
the main result of our study is the behaviour of the time of relaxation to the steady state as a function of @xmath0 .
at low @xmath0 the transient relaxation time is an increasing function of @xmath0 , it is maximum close to the coil stretch transition , and eventually tends to zero with increasing @xmath0 . knowing the dependence of the transient time on _ wi _ is relevant both for numerical simulations and experiments .
for example , in the former case , the time required for uncorrelated polymer chains that are suddenly exposed to the same flow to correlate is ( implicitly ) related to the sharpness of stress gradients one can expect in the flow .
hence , the prediction of the transient time in our study is useful to estimate the required grid spacing to fully resolve those gradients ( l. collins 2005 , private communication ) . in the latter case , the fact that the transient relaxation time is especially long just below the coil
stretch transition implies that within such range of _ wi _ experimental measures are more sensitive to statistical fluctuations .
experiments concerning the transient relaxation to the stationary regime can investigate the time dependence of the conditional p.d.f .
@xmath138 , which corresponds to the initial condition peaked at the equilibrium size : @xmath139 .
such initial condition can be fixed experimentally as follows ( a. celani 2005 , private communication ) . the p.d.f . of the extension
is constructed by following the motion of different polymer molecules and collecting @xmath140 for each molecule : one should then start counting time only when the length of the corresponding polymer is approximatively @xmath10 .
this is equivalent to selecting the initial state where all molecules have the equilibrium extension , @xmath141 .
as for the transient relaxation time , this can be measured directly from the time behaviour of the conditional moments of the extension : @xmath142 , where @xmath143 is a positive integer . the conditional p.d.f
can indeed be expanded as in and the order of series and integral can be interchanged in the definition of @xmath144 due to the integrability of @xmath145 and the uniform convergence of series ( see e.g. * ? ? ?
* ) . therefore , all the moments of the extension converge to their stationary value with the same rate as the p.d.f . of the extension .
to conclude , we believe that the results obtained for the nonlinear dumbbell model are relevant for the comprehension of polymer dynamics in turbulent flows at any weissenberg number . moreover , we think that our study may stimulate new experiments directed to investigate the transient relaxation to the stationary regime .
we are deeply grateful to e. bodenschatz , a. celani , and l. collins for very useful suggestions .
we wish to thank c. m. casciola , m. chertkov , j. davoudi , e. de angelis , b. eckhardt , a. mazzino , s. musacchio , j. schumacher , and v. steinberg for illuminating discussions .
this research has been supported in part by the european union under contract no .
hprn - ct-2000 - 00162 , by the french - german program pai `` procope '' 07574 nm , and by the italian ministry of education , university and research under contract cofin .
2003 ( prot .
020302002038 ) .
babcock h. p. , teixeira r. e. , hur j. s. , shaqfeh e. s. & chu s. 2003 visualization of molecular fluctuations near the critical point of the coil
stretch transition in polymer elongation , _ macromolecules _ * 36 * , 4544 - 4548 . klyatskin v. i. , woyczynski w. a. & gurarie d. 1996 short - time correlation approximations for diffusing tracers in random velocity fields : a functional approach . in _ stochastic modeling in physical oceanography _ ,
39 of progress in probability , r. j. adler ed . , pp . 221 - 270 , birkhauser . | polymer stretching in random smooth flows is investigated within the framework of the fene dumbbell model .
the advecting flow is gaussian and short - correlated in time .
the stationary probability density function of polymer extension is derived exactly . the characteristic time needed for the system to attain the stationary regime
is computed as a function of the weissenberg number and the maximum length of polymers . the transient relaxation to the stationary regime
is predicted to be exceptionally slow in the proximity of the coil stretch transition . |
immiscible fluids and non - porous solids occur abundantly in both nature and industry .
the location at which two such fluids and a solid meet is termed a contact line , and the modelling of moving contact lines is consequently of central importance to the understanding of problems as diverse as how insects walk on water and the wetting properties of plant leaves to coating , inkjet printing , and oil recovery for reviews , see e.g. @xcite .
the main focus of this work is to compare a number of prevalent models proposed to resolve the _ moving contact line problem _ in the prototype system of quasistatic spreading of a thin two - dimensional droplet on a planar substrate .
two early papers sparked interest in the moving contact line problem .
the first by moffatt in 1964 @xcite considered the conceptually simple problem of a wedge of viscous fluid in contact with an inviscid fluid ( the interface kept flat ostensibly by the action of gravity ) , and also in contact with a solid substate moving with a constant velocity @xmath0 .
in this situation , moffatt noted that in the absence of fluid inertia , a fluid particle initially on the fluid interface will be moving towards the contact line at a speed independent of the distance away from it , and with a value less that @xmath0 .
the no - slip boundary condition will then enforce the speed of this fluid particle to instantaneously increase to @xmath0 at the contact line , where it ` turns the corner ' .
this would require an infinite acceleration caused by an infinite stress and pressure at the contact line , all of which are physically unacceptable .
this relatively brief glimpse into the moving contact line problem was brought into focus by the seminal work in 1971 of huh and scriven @xcite .
the authors extended the analysis of moffatt to include the situation of two immiscible viscous fluids of arbitrary viscosities and described in detail the _ nonintegrable shear - stress singularity _ alongside offering suggestions as to how it may be overcome .
huh and scriven highlighted the primary cause for the unphysical singularities as the no - slip condition at the wall . to resolve the moving contact line problem ,
huh and scriven suggested that some form of slip in the contact line vicinity could be allowed , such as the navier - slip condition initially proposed in the far earlier work by navier @xcite . whilst this slip boundary condition was not initially proposed with a physical basis
, it should be noted that in the last few decades there is evidence that slip can physically occur at the molecular scale , such as seen in the molecular dynamics simulations of @xcite , and there are numerous physical effects such as viscous heating , dissolved gas or polymer solutions , and surface roughness that may contribute to apparent slip @xcite . since these important works , including the work of dussan v. and davis @xcite who in 1974 clarified the moving contact line problem mathematically and with experiments , the volume of literature on the moving contact line problem has increased rapidly , as evidenced in the subsequent review articles @xcite .
the problem remains an active field of research , due in part to the wide variety of microscale physical effects which can resolve the problem .
further to the above , these include modelling a precursor film ahead of the contact line ( physically realised through the inclusion of a disjoining pressure , which accounts for intermolecular interactions near the contact line ) @xcite ; rheological ( i.e. non - newtonian ) effects @xcite ; treating surfaces as separate thermodynamic entities with dynamic surface tensions ( as in the interface formation model ) @xcite ; introducing numerical slip @xcite ; assuming the contact angle to be always @xmath1 at the solid substrate @xcite , including evaporative fluxes @xcite ; considering the interface to be diffuse @xcite with different models for liquid - gas @xcite and liquid - liquid ( binary ) fluids @xcite being adopted ; and modelling both molecular and hydrodynamic effects in hybrid models @xcite . in the present communication
we will compare a variety of models that are used to alleviate the non - integrable stress singularity at a moving contact line , concentrating primarily on slip and precursor film models ( including one derived from diffuse - interface theory ) as well as the aforementioned interface formation model . in slip models
, the no - slip boundary condition is relaxed to allow for fluid particles to move along the substrate . on the other hand ,
the no - slip condition is applied on the substrate for precursor film models , but has only an apparent contact line away from the substrate , thus removing the associated stress and pressure singularities that would have occurred had the contact line been sharp . as the substrate must be entirely covered by a film in these models a disjoining pressure
is included due to long - range intermolecular interactions ( such as from van der waals forces @xcite ) , with both repulsive and attractive contributions to this pressure balancing to create a stable thin - film .
details of the interface formation model will be given in sec .
[ sec : thinfilm ] where it is analysed . alongside the question of which microscale effects to include in a description of the moving contact line
, for many models there is another related question as to which additional boundary condition(s ) at the contact line should be imposed . for slip models , the boundary condition at the contact line usually takes the form of a prescription of the variation of contact angle with contact line velocity .
a simple , yet widely used condition is that the microscopic contact angle remains at its static value when the contact line is in motion , i.e. @xmath2 , where @xmath3 is the microscopic contact angle in any situation , @xmath4 is the dynamic contact angle , and @xmath5 is the static contact angle .
this is imposed in a number of studies involving slip models such as studies by hocking @xcite and cox @xcite , as well as a recent analysis providing model comparisons for 2d droplet spreading by savva and kalliadasis @xcite .
the analysis of other contact angle conditions and models to extend the work of @xcite , given in sec .
[ sec : thinfilm ] here , is the primary focus of this article .
two forms we consider here to allow for contact line velocity dependence on microscopic contact angle are @xmath6 where @xmath7 is a coefficient with dimensions of velocity , and @xmath8 is a dimensionless exponent . the parameters @xmath7 and @xmath8 could be determined from experiments ( see , for instance , @xcite for some experimental measurements ) but the question as to whether the contact angle varies dynamically _ at _ the substrate remains an open debate , given the limits of experimental resolution
see discussion relating to this in @xcite .
forms such as ( [ eq : introcarel ] ) are considered , with @xmath8 specified , for instance by greenspan @xcite , haley and miksis @xcite , hocking @xcite , ehrhard and davis @xcite , and wilson _ et al . _
@xcite . for the disjoining pressure models considered here , where there is an apparent contact line due to a precursor film , a different contact angle relation may be derived . for long - wave models ( where @xmath9 ) the contact angle is given by @xmath10 where @xmath11 is the height of the free surface
, @xmath12 is the precursor film thickness , @xmath13 is the fluid - fluid surface tension , @xmath14 is the interface potential for a flat film of thickness @xmath11 , and @xmath15 is the disjoining pressure , which here depends on the thin - film height .
this long - wave equilibrium contact angle relation ( [ eq : earlyangledp ] ) is derived through multiple integrations of the equilibrium thin - film equation for precursor film / disjoining pressure models ( e.g. by eggers @xcite ) , and is also used in the comparison study @xcite .
we note that ( [ eq : earlyangledp ] ) is not an imposed boundary condition , since the effect of the disjoining pressure creates a precursor film and prevents an actual contact line occurring .
the required boundary conditions are instead given by @xmath11 approaching the precursor film thickness @xmath12 at distances far from the bulk liquid region . before discussing these models and conditions in the thin - film ( long - wave ) approximation in sec .
[ sec : thinfilm ] , we revisit the early work of moffatt @xcite , which imposed no - slip , in sec .
[ sec : mofhuhscriv ] to uncover the nature of the moving contact line singularity .
we then extend the analysis to see how slip models overcome the problem in this simple situation , and to show that no essential physics is missing in the thin - film setting .
the results will be in agreement with the later thin - film analysis of sec .
[ sec : thinfilm ] .
whilst the problems considered by moffatt @xcite and huh and scriven @xcite are unphysical as in the imposed geometry the normal stress balance condition at the fluid - fluid free surface is not satisfied , it is instructive to revisit the situation and consider how slip conditions change the analysis . consider the motion of a contact line on a perfectly flat solid surface , with governing equations appropriately nondimensionalised .
it is convenient to work in a frame of reference where the contact line is fixed and the surface translates with steady ( nondimensional ) velocity @xmath16 .
the contact line is the location where the fluid free surface meets the solid , the free surface is assumed to be planar at an angle @xmath17 .
the situation is two - dimensional and we take the contact line to be the origin of a polar coordinate system .
the situation is shown in figure [ fig : fig1_hs_wedge ] .
, working in a frame of reference where the contact line is stationary and the solid is moving with nondimensional velocity @xmath16.,width=415 ] for rapid wetting processes inertial effects may be important , and have been studied analytically @xcite , and recently validated numerically @xcite . throughout this article
however , we will consider inertialess flow relevant for slow processes such as droplet spreading . in the moffatt wedge flow geometry , this means we consider the creeping ( or stokes , @xcite ) flow regime , where the nondimensional governing equations for inertialess incompressible newtonian flow are @xmath18 where @xmath19 is the fluid velocity , @xmath20 is the extra stress tensor , @xmath21 is the pressure , and @xmath22 is the stream function in polar coordinates . consider the behaviour as @xmath23 of @xmath24 , with @xmath25 an exponent to be determined . from this behaviour
we can determine the order - of - magnitude estimates @xmath26 so that the pressure order - of - magnitude is @xmath27 the momentum equation is easily reduced to the biharmonic equation @xmath28 , after introducing the stream function , and we note the form of the pressure and stresses in polar coordinates , using ( [ eq : genewtonian ] ) , may be found from @xmath29 following moffatt @xcite , we take a separable stream function @xmath30 we note here that the power - law form ( [ eq : newton : psib ] ) for the stream function is used , as in @xcite who looked at other more complex forms of slip .
this is in contrast to the analysis of hocking @xcite for navier - slip , who considered a form consisting of the no - slip solution modified by a logarithm . hocking s analysis gives a full matched asymptotic solution , which yields complex integral expressions that do not afford easy insight into the nature of the stress and pressure behaviours .
we instead focus our simple analysis on the behaviour at the contact line , essentially the limit as the contact line is approached from the hocking inner solution which becomes @xmath31 for the navier - slip model , as seen from equations ( 2.10 ) , ( 2.11 ) and ( 2.15 ) there .
returning to ( [ eq : newton : psib ] ) , on substituting this into the biharmonic equation we obtain the linear ordinary differential equation ( ode ) for @xmath32 of @xmath33 true for all @xmath25 , including ( in particular ) @xmath34 , @xmath35 , and @xmath36 .
this is a fourth order ode for @xmath32 , with solutions @xmath37+c_2\cos[(\gamma-2)\theta]+c_3\sin(\gamma\theta)+c_4\cos(\gamma\theta ) , & \mbox{for } \gamma \neq \{0,1,2\},\\ c_1+c_2\theta+c_3\sin(2\theta)+c_4\cos(2\theta ) , & \mbox{for } \gamma = \{0,2\},\\ c_1\sin\theta+c_2\cos\theta+c_3\theta\sin\theta+c_4\theta\cos\theta , & \mbox{for } \gamma = 1 , \end{array}\right.\label{eq : bhedeg}\end{aligned}\ ] ] noting the correction for the degenerate solution for @xmath34 ( in comparison to the erroneous form in @xcite ) .
the boundary conditions we apply in this ( artificial planar wedge ) configuration are of vanishing normal component of velocity at the solid surface and the fluid interface @xmath38 and the continuity of tangential stress at the interface @xmath39 where @xmath40 is the ( dimensionless ) surface tension , @xmath41 is the unit normal pointing out of the fluid , and @xmath42 , is a tangent vector on the interface .
the surface tension in the classical model is usually assumed to be a constant unless temperature variation or variable chemical composition are to be modelled . in this section
we will also assume constant surface tension , so that the boundary condition becomes @xmath43 but will return to discuss this assumption in appendix [ app : wedge ] .
the final condition is applied on the wall , where the various slip models may be introduced . in classical fluid mechanics problems ,
no - slip is imposed on an interface between a viscous fluid and a solid boundary . for a moving contact line , this leads to a singular stress and pressure at the contact line , as discussed .
navier - slip , discussed in the introduction , has been used extensively , and is of the form @xmath44 . in this representation
the variables are nondimensional , with @xmath45 the wall velocity and @xmath46 the nondimensional slip - length .
the possibilities for slip to overcome this problem we will investigate are : @xmath47 the final two models have the slip length related to the height of the fluid region , with @xmath48 , see figure [ fig : fig1_hs_wedge ] , and for these two models clearly we are restricted to contact angles @xmath49 .
whilst less used , the final two models do appear in the literature . in relation to
the droplet spreading problem which this work will focus on , noteworthy are works by greenspan @xcite who used a form of the ` inverse linear slip ' condition , and the work of haley and miksis @xcite , who employ all three above slip conditions . in the long - wave limit ,
the ` inverse linear slip ' boundary condition reduces to a slip condition proposed by ruckenstein and dunn @xcite , which scales with the pressure gradient , as shown in @xcite .
we now consider the asymptotic behaviour of the various slip models given in ( [ eq : slipmods ] ) .
this should then allow us to choose an appropriate form for the stream function , and thus the general solution of the biharmonic equation for each model .
we use the form of @xmath50 from ( [ eq : stressinpolar ] ) , with the stream function from ( [ eq : genewtonian ] ) , and the separable form of the stream function in ( [ eq : newton : psib ] ) to obtain @xmath51 consider ( [ eq : slipmodsr ] ) now to determine @xmath25 . in the no - slip case ,
the only appropriate balance is when @xmath35 , giving a stream function of @xmath52 . in all other cases ( i.e. the slip cases )
, there is an apparent choice of balances .
however , there is only one dominant balance
. the left hand side term of ( [ eq : slipmodsr ] ) which comes from @xmath53 clearly can not balance with the @xmath50 terms ( the terms involving @xmath54 on the right hand side of ( [ eq : slipmodsr ] ) ) , and attempting the possibility of balancing the @xmath55 term with the @xmath16 term at leading order would always leave the @xmath50 terms dominating .
thus the dominant balance is between the @xmath16 and @xmath50 terms , giving @xmath56 , \\
\mbox{navier - slip } & { \gamma=2 } , & \psi = r^2 \left [ c_1+c_2\theta+c_3\sin(2\theta)+c_4\cos(2\theta ) \right ] , \\
\mbox{general slip } & { \gamma=2+n } , & \psi = r^{2+n}\left[c_1\sin(n\theta)+c_2\cos(n\theta ) \right .
\\ \quad(n>0 ) & & \left.\qquad\qquad\qquad\qquad\quad+\ ; c_3\sin((2+n)\theta)+c_4\cos((2+n)\theta ) \right ] , \end{array}\right\ } \label{eq : psiforms}\end{aligned}\ ] ] where the appropriate form for @xmath32 has been used from ( [ eq : bhedeg ] ) , and the inverse linear and inverse quadratic slip models have been combined into a ` general slip ' model , with @xmath57 and @xmath58 corresponding to the inverse linear and inverse quadratic slip models respectively . using the appropriate stream function form from ( [ eq : psiforms ] ) , we may now use the boundary conditions ( [ eq : bcnonv ] ) , ( [ eq : bcts ] ) and the desired slip model from ( [ eq : slipmods ] ) to determine the stream function solution , and thus the velocity , stress and pressure forms . with the no - slip condition applied on the wall ,
the solution satisfying the boundary conditions is @xmath59 where @xmath60 . if we now consider the velocity on the free surface @xmath17 , it is easy to see that @xmath61 , as required , and that @xmath62},\end{aligned}\ ] ] giving the solution of moffatt @xcite see equation ( 2.3 ) there , and confirming that the free surface velocity is a monotonically decreasing function of @xmath63 $ ] , having values in the range @xmath64 to @xmath65 .
the @xmath66 singularities are clearly seen in ( [ eq : no - slip_sol ] ) in the shear stress and pressure , but this result is also of interest to motivate why taking the contact angle as @xmath1 ( not excluded for this no - slip case , and so that the liquid advances and rolls as would a caterpillar track ) could remove the singularity and be worthy of investigation at this order the above equations for @xmath67 give @xmath68 velocities but zero pressure and stresses .
analysis of this special case was considered by benney and timson @xcite in 1980 , but a sign error was later found @xcite .
the analysis has since been revisited by benilov and vynnycky in 2013 , @xcite , for the problem of two - dimensional couette flow with a free boundary .
however here we focus on small contact angles , which allow us to look into models applied for thin film equations .
the solutions for the different slip models in the moffatt wedge setting may be found in appendix [ app : wedge ] , however the important results are that the navier - slip model removes the singularity in the stress as the contact line is approached , but a logarithmic singularity in the pressure remains .
this is integrable , however , and as such is not as worrying as the @xmath69 singularity in the pressure and stress in the no - slip model . in comparison , the two cases @xmath57 and @xmath58 in the general slip model resolve both the pressure and stress singularities , but both give zero pressure and stresses as the contact line is approached the pressure and stresses being @xmath70 and @xmath31 in the two cases respectively . finally shown in appendix [ app : wedge ] is a solution with @xmath68 stresses , yet without the logarithmic singularity in the pressure , which is achieved through allowing for variable surface tension .
a useful and illuminating set of flows to consider the moving contact line in are that of thin - films @xcite . in the thin - film approximation
the characteristic scale in the direction normal to the substrate is much smaller than that of the direction tangential to it .
the small parameter of the system , which may be either purely the ratio of the normal to tangential scales , or based on ( for instance ) the smallness of the contact angle , then allows many of the terms in the full navier stokes equations and boundary conditions to be neglected , with analytical progress becoming more feasible . here
, we will consider contact line motion in the prototype setting of the spreading of a thin two - dimensional droplet , as in previous works @xcite , noting that the analysis holds also for receding droplets without modification , as seen in @xcite . for motion outside the validity of our quasistatic assumption , however , spreading and receding droplets may have to be treated differently due to the different asymptotics of the full inner - region equation @xcite . ideally , the full equations are considered in dimensional form and the small terms are neglected after a process of nondimensionalisation ( as performed in the thin - film reviews @xcite and also @xcite for slip and precursor film models , to which the reader is directed for full details ) , noting also a relatively new , yet elegant and powerful method based on gradient dynamics where a variational procedure allows the time evolution of the film height to be derived as a conserved order parameter @xcite .
slip is accounted for in the boundary condition at the wall , whereas intermolecular forces due to the wall are included through a disjoining pressure in the navier
stokes equations @xcite .
unlike the critique formulated in @xcite , the disjoining pressure does not ` double - count ' intermolecular forces of the fluid as it only accounts for the additional effects due to the presence of the wall - fluid interface .
we do note , however , that there remains some debate about the issue of including additional effects from the microscale in continuum mechanics equations , as also discussed in @xcite . for completeness here we will derive the evolution equation from the two - dimensional navier
stokes equations already reduced to the leading order long - wave form @xcite .
as mentioned , it is assumed that the characteristic scale in the direction normal to the substrate is much smaller than that of the direction tangential to it .
lubrication theory is known to be a good approximation for slowly moving , thin , viscous droplets @xcite .
the equations reduce to the continuity equation and simplified momentum equations where the pressure gradient ( with disjoining pressure gradient optionally included ) is non - zero in the @xmath71-direction only and balanced by the vertical shear of the horizontal velocity . thus the ( dimensional ) governing equations are @xmath72 with boundary conditions @xmath73 which are the thin - film forms of the tangential stress balance , slip law , no penetration condition , normal stress balance , and kinematic ( free - surface ) boundary condition , respectively.)([eq : lwbc2 ] ) we have included the disjoining pressure in the momentum equation , as in @xcite .
a derivation may be made with the disjoining pressure instead entering into the normal stress balance in eq .
( [ eq : lwbc2])(i ) , as e.g. @xcite , however the same evolution equation in ( [ eq : dimlw ] ) is then found .
] now , by using ( [ eq : lwgov])(ii ) and ( [ eq : lwgov])(iii ) , the velocity @xmath74 may be determined in terms of @xmath75 , @xmath76 and @xmath11 , with @xmath77 then determined through ( [ eq : lwgov])(i ) .
the three arbitrary functions involved in @xmath74 and @xmath77 are then determined with the three boundary conditions ( [ eq : lwbc1 ] ) , with the pressure then given using ( [ eq : lwbc2])(i ) .
finally , all of the results are substituted into the kinematic boundary condition ( [ eq : lwbc2])(ii ) yielding the evolution equation of the film height @xmath78 } } \right\}}=0 .
\label{eq : dimlw}\end{aligned}\ ] ] we note that ( [ eq : dimlw ] ) with @xmath79 is essentially the thin - film equation in the ` weak - slip regime ' of @xcite , which also considered a number of other nondimensionalisations for circumstances where slip plays a greater role . a number of other similar thin - film equations , predominately when the substrate is covered by a precursor film , have been considered using asymptotics in @xcite
. there will be three important quantities to consider at the contact line the pressure , the velocity parallel to the substrate , and the shear stress . in the long - wave regime
( and at @xmath80 ) , these correspond to @xmath81 the contact angle @xmath3 in general is given by @xmath82 , but as it must be small it may be used as the long - wave parameter , with the result @xmath83 .
we nondimensionalise eqs .
( [ eq : dimlw])([eq : dim_pvs ] ) through the scalings @xmath84 which gives the evolution equation @xmath85 } } \right\}}=0 , \label{eq : ndlw}\end{aligned}\ ] ] and @xmath86 } } \right|}_{z=0 } , \qquad { \left . { { \frac{\partial{u}}{\partial{z } } } } \right|}_{z=0 } = { \left .
{ { \left [ { 3 h { \frac{\partial^3{h}}{\partial{x}^3 } } } \right ] } } \right|}_{z=0}.\label{eq : nd_pvs}\end{aligned}\ ] ] for slip models here , we will restrict our attention to the three cases discussed by haley and miksis @xcite of @xmath87 . in @xcite ,
the authors described how the navier - slip and inverse linear slip conditions gave equivalent spreading dynamics as precursor film models , with two forms being investigated based on the models used by schwartz and eley @xcite , and by sharma @xcite .
these two disjoining pressure models correspond to @xmath88 } , \nonumber\\ \bar\pi(h ) & = \frac{a^*}{h^{*m^*}}{\left [ { { \left ( { \frac{h^*}{h } } \right)}^{m^ * } - { \mbox{e}}^{-(h - h^*)/(d^*h^ * ) } } \right]},\end{aligned}\ ] ] where @xmath89 is the hamaker constant of each respective model , @xmath90 and @xmath91 ( with @xmath92 ) are positive integers characterising the strength of the intermolecular interactions , and @xmath93 gives the lengthscale associated with the decay of the repulsive force . to find these disjoining potentials in nondimensional form
, we use the scaling in ( [ eq : ndscalsdp ] ) , and a version of ( [ eq : earlyangledp ] ) , to give a relationship between the hamaker constant and the contact angle ( which is scaled to unity in our choice of the nondimensionalisation here ) .
thus we use @xmath94 and find the two forms @xmath95 } , \nonumber\\ \pi(h ) & = \frac{m^*-1}{2\epsilon[1-(m^*-1)d^*]}{\left [ { { \left ( { \frac{\epsilon}{h } } \right)}^{m^ * } - { \mbox{e}}^{-(h-\epsilon)/(d^*\epsilon ) } } \right]},\end{aligned}\ ] ] where @xmath96 , is the small parameter associated with these disjoining pressure models ( analogous to @xmath54 for the slip models ) , see @xcite for further details .
the droplet spreading geometry is given in figure [ fig : drop ] , where the small parameter @xmath97 is used , and corresponds to either @xmath54 for slip models ( when @xmath98 ) , or @xmath99 for disjoining pressure models ( when @xmath100 .
we reiterate that these disjoining pressure models create a thin precursor film ahead of the contact line and extending to immediately cover the entire substrate ( although see discussion by de gennes about this point @xcite ) , so that there is no actual contact line and thus no need for slip to resolve an associated contact line singularity ( as e.g. @xcite ) .
other , more complex forms of disjoining pressure may be considered where actual contact lines exist and slip is necessary for the motion of contact lines , see e.g. @xcite .
inner regions near the contact lines , located at @xmath101.,width=453 ] another model has been proposed based on diffuse - interface theory . following @xcite
, we consider a liquid - gas diffuse - interface as the _ compressible _ navier stokes equations where the stress tensor is given by the compressible newtonian viscous stress tensor with a diffusion tensor added .
this diffusion tensor ( also known as the capillary or korteweg tensor ) incorporates terms which are associated with the pressure and surface tension , arising from the interaction of the molecules within the interface , and is given by @xmath102}\textbf{i } - k\bar{\bm{\nabla}}\bar\rho\otimes\bar{\bm{\nabla}}\bar\rho,\end{aligned}\ ] ] where @xmath103 is a gradient energy coefficient .
the fluid density @xmath104 acts as an order parameter that varies between the two equilibrium states of the fluid ( i.e. the bulk liquid and vapour densities @xmath105 and @xmath106 respectively ) .
the precise diffuse - interface governing equations , with no - slip applied on the wall , are given in @xcite .
the equations were studied in @xcite without making the long - wave assumption , and it was shown that the singularities associated with the contact line problem are resolved without requiring a precursor film ( i.e. the diffuse - interface model alone is sufficient mathematically ) .
this analysis was also extended to include general free - energy and viscosity forms in @xcite . however , if modelling a situation where a precursor film is desired , then a boundary condition @xmath107 ( where @xmath108 is a prescribed wall density ) may be applied .
pismen and pomeau @xcite considered this case , and found that in the thin - film approximation , the model reduces to the evolution equation ( [ eq : dimlw ] ) , with @xmath109 and disjoining pressure @xmath110},\end{aligned}\ ] ] where @xmath89 is now an effective hamaker constant , being the strength of the molecular interactions , @xmath12 is strictly now the width of the diffuse interface , rather than the exact thickness of the precursor film , and @xmath111 , is a small positive parameter being a measure of the density at the wall i.e. a measure of the wetting properties of the substrate . using the same nondimensionalisation procedure outlined above , we find @xmath112}.\end{aligned}\ ] ]
finally , we discuss the interface formation model , which was studied in the thin - film regime for droplet spreading in @xcite .
the interface formation model exploits distinct surface variables and dynamic surface tensions and is more closely linked to slip models than diffuse - interface methods , albeit in a rather complex and sophisticated form .
it was developed in 1993 by shikhmurzaev @xcite , ( thus predating diffuse - interface approaches for modelling contact lines , starting from the work of seppecher in 1996 , @xcite ) and has been used in a variety of wetting situations .
the interface formation model has attracted some controversy , however ( see the recent methodical analysis by the present authors on the model utilising the thin - film regime @xcite , and the monograph of shikhmurzaev @xcite ) .
briefly , the key physical ingredients of the interface formation model are that : * the interfaces between liquid and gas , and liquid and solid , have surface variables distinct from bulk values .
* the surface tension varies dynamically , and relaxes on a finite timescale @xmath113 .
the contact angle is also assumed to vary dynamically ( and will be given as a solution of the equations , rather than being prescribed as for the slip models ) , and a purely illustrative schematic is shown in figure [ fig : ifmschem ] .
satisfying no - slip , yet effective slip for bulk velocity @xmath19 .
a magnification of the liquid interface displays the density variation close to an interface , resulting in the surface tension where fluid particle a away from the interface is surrounded by other fluid particles , whereas the fluid particle b near the interface feels the presence of fewer fluid particles and is thus in an unfavourable state .
the surface tensions satisfy the young equation , and vary over a finite timescale @xmath113 .
this schematic is purely illustrative , as the interfaces are taken with zero thickness in the model.,width=453 ] at the contact line the interface formation model satisfies both the young equation @xcite , and conservation of mass ( which , unlike for slip models where this holds automatically , mass conservation at the contact line must be explicitly imposed in the model ) .
the young equation describes a force balance in the direction parallel to the solid surface . considering a static contact line region ,
then the force balance is @xmath114 where @xmath13 , @xmath115 and @xmath116 , are the surface tensions of the fluid 1-fluid 2 , fluid 1-solid and fluid 2-solid interfaces , and @xmath3 is the contact angle .
for the application of the interface formation model , the young equation ( [ eq : youngeq ] ) is assumed to hold dynamically ( see , for example sec .
4.3.5.2 of ref .
@xcite ) , with the microscopic contact angle @xmath3 changing from its static value @xmath117 to a dynamically varying angle @xmath118 .
this then requires at least one of the surface tensions to also change from its static value and provides the motivation for , and the distinguishing feature of , the interface formation model .
this is as inherent within the interface formation model is that the surface tensions of the interfaces dynamically vary on a finite timescale @xmath113 ( as discussed above ) .
the system of equations to solve is comparatively complex , being the usual incompressible navier
stokes equations in the bulk , but coupled to a system of six equations forming the boundary conditions at the solid - liquid interface , seven at the liquid - gas free surface and two further boundary conditions at the contact line ( from the young equation and from conservation of mass through the contact line ) . due to their abundance and complexity , we will not give details of the full equations here , but refer the reader to @xcite and to the monograph @xcite . after a process of nondimensionalisation ,
the thin - film equations of the interface formation model for the droplet spreading problem are ( see @xcite for details ) @xmath119 where @xmath120 and @xmath121 give effectively the shear stress and the slip velocity at the solid surface , respectively , and satisfy the two additional equations @xmath122 .\label{eq : chi0ab2}\end{aligned}\ ] ] these three governing equations contain the nondimensional parameters @xmath123 , @xmath124 , @xmath125 , @xmath126 , and @xmath127 , which correspond to nondimensional numbers characterising the magnitude of mass transfer between bulk and surfaces , an inverse slip coefficient , the nondimensional surface tension relaxation time , the ratio of the equilibrium solid and gas surface densities , and a nondimensional number combining darcy - type and slip length parameters to characterise the surfaces respectively further details again in @xcite .
finally , @xmath128 represents the sum of the dynamic surface densities and is given by @xmath129.\label{eq : chi0rhosg}\ ] ] these long - wave equations are substantially more complex than their slip or precursor film counterparts , but if @xmath130 so that the surface tension relaxes instantaneously i.e
. the surface tensions will be constant , then the navier slip model is recovered . at the contact line ( @xmath131 ) ,
the two boundary conditions imposed , related to the mass balance and the young equation are @xmath132 } - \frac{2\breve{\rho}\bar\alpha a}{\bar\beta } + b = \bar{u}_{cl } ( 1 + \breve{\rho } ) , \quad \mbox{and } \quad 1- { \left [ { { \frac{\partial{{{h}}}}{\partial{{{x } } } } } } \right]}^2 = 2\mathcal{d } , \end{aligned}\ ] ] where @xmath133 .
we now consider the prototype situation of quasistatic spreading of a thin two - dimensional droplet , with the geometry as shown in figure [ fig : drop ] .
the evolution of the droplet radius @xmath134 will be determined through the method of matched asymptotics . in the bulk of the droplet ( the outer region )
, the leading order behaviour is modelled with the thin - film evolution equation with negligible slip , disjoining pressure or other microscopic effects , whereas these play a significant role in the inner regions near the contact lines .
the outer region behaviour holds for all models considered in this work , i.e. for the slip models , the disjoining pressure models and the interface formation model , and arises from neglecting the microscopic effects in ( [ eq : ndlw ] ) or ( [ eq : ifmlw])([eq : chi0rhosg ] ) .
the inner region behaviours for the variety of slip models will then be considered in detail , with the remaining models then discussed and compared at the end of this section .
[ [ outer - region ] ] outer region : + + + + + + + + + + + + + the microscopic effects of slip , disjoining pressure etc . , are not significant in the outer region , and as such we take the limit @xmath135 ( and equivalently @xmath136 , with @xmath137 , for the interface formation model ) .
subsequent calculations are now given at leading order in an asymptotic expansion in @xmath97 .
we transform the symmetric droplet domain @xmath138 to @xmath139 via the mapping @xmath140 so that the governing equation in terms of the ( non - moving ) coordinate @xmath141 becomes @xmath142 = 0,\end{aligned}\ ] ] where @xmath143 is the spreading rate of the contact line . here
, we are considering slow spreading , and thus consider the quasistatic limit @xmath144 with expansions of the form @xmath145 where the time dependence of @xmath146 enters through @xmath147 and its time derivatives .
an assumption made in this procedure is that @xmath148 .
this will be found to be consistent as @xmath149 depends only logarithmically on @xmath97 [ see later in eq .
( [ eq : srr ] ) ] . at leading order in our quasistatic expansion we have @xmath150
these conditions correspond to zero height at the contact line , a symmetry condition such that the slope of the droplet is zero at its centre , and a normalization condition such that the nondimensional cross sectional area of the half - droplet is unity .
equations ( [ eq : outle ] ) are solved to find @xmath151 the equation for the next term is @xmath152 giving @xmath153-\frac{3y^2}{2 } + \frac{1}{2}\ln{\left ( { \frac{{\mbox{e}}^3}{16 } } \right ) } } \right ] } , \ ] ] having used @xmath154 , and where arbitrary constants have been determined from @xmath155 , @xmath156 , and @xmath157 . using these results for @xmath158 and @xmath159 , and transforming back to the original coordinate @xmath160 ,
the behaviour of the free surface slope in the outer region as the inner region is approached ( as @xmath161 ) is found as @xmath162.\label{eq : outbehav}\ ] ] a similar behaviour is found for @xmath163 for the left contact point , but the droplet symmetry allows us to consider only the right contact point in the remaining analysis .
[ [ inner - region ] ] inner region : + + + + + + + + + + + + + the behaviour from ( [ eq : outbehav ] ) must be matched to its inner region counterpart , where the microscopic effects are significant ( and as discussed only the inner region near @xmath131 being considered due to the droplet symmetry ) .
the inner region is defined by the change of variables @xmath164 which , for the slip models , yields the inner region evolution equation @xmath165 proceeding with a quasistatic expansion @xmath166 , then at leading order @xmath167 for @xmath168 , corresponding to navier - slip and inverse linear slip , the integration constant must be zero to satisfy the boundary condition of zero height at the contact line @xmath169 . in these cases ,
the nontrivial solution is given by @xmath170 , and on solving this subject to @xmath171 and @xmath172 as @xmath173 ( to be able to match to the outer region @xmath174 behaviour ) , then @xmath175 , with the arbitrary constant @xmath176 being fixed by the contact angle condition . for @xmath58 there will be other possible solutions to this leading order equation @xmath177/(\phi_0 ^ 3 + 1)$ ] , but the wedge solution @xmath175 holds .
as it would be expected that the leading order solution is unaffected by the slip model chosen , we assume this wedge solution is the physically relevant solution for the @xmath58 case also . at the next order ,
the equation to solve ( after integrating once ) is then @xmath178 and once again the integration constant must be zero for @xmath168 , but satisfies @xmath179 for @xmath58 .
it is found , however , that @xmath180 must vanish to prevent a singular velocity at the contact line for the @xmath58 slip model . for @xmath79 then , the navier - slip case , the solution satisfying @xmath181 and @xmath182 as @xmath173 is @xmath183 } , \label{eq : p1_0}\end{aligned}\ ] ] equivalently for @xmath57 , inverse linear slip , it is @xmath184 } , \label{eq : p1_1}\end{aligned}\ ] ] and for @xmath58 , the solution is @xmath185 , \label{eq : p1_2}\end{aligned}\ ] ] where in all three cases , another arbitrary constant @xmath186 will be fixed by the contact angle condition . alongside the crucial behaviour as @xmath173 to match to the outer region , there are three other important quantities to consider at the contact line @xmath187 , these being given in original nondimensional variables in ( [ eq : nd_pvs ] ) , and in inner variables ( with @xmath188 , @xmath189 , and @xmath190 ) becoming @xmath191 using these we find @xmath192 + o(\xi ) , \quad u_i|_{z_i=0 } = \dot{a } + o(\xi ) , \quad { \left . { { \frac{\partial{u_i}}{\partial{z_i } } } } \right|}_{z_i=0 } = 3 \dot{a } + o(\xi ) , \nonumber\\ n=1 : \quad & p_i = -\frac{3\dot{a}\pi}{2c_a } + o(\xi ) , \quad u_i|_{z_i=0 } = \dot{a } + o(\xi^2 ) , \quad { \left . { { \frac{\partial{u_i}}{\partial{z_i } } } } \right|}_{z_i=0 } = 3 c_a \dot{a } \xi + o(\xi^3 ) , \nonumber\\ n=2 : \quad & p_i = -\frac{2\pi\dot{a}}{\sqrt{3}c_a } + o(\xi^2 ) , \quad u_i|_{z_i=0 } = \dot{a } + o(\xi^3 ) , \quad { \left . { { \frac{\partial{u_i}}{\partial{z_i } } } } \right|}_{z_i=0 } = 3 c_a^2 \dot{a } \xi^2 + o(\xi^5),\end{aligned}\ ] ] where we see that the stress is regularised by all slip models , the velocity at the contact line is always @xmath149 as expected , and the logarithmic pressure singularity for navier - slip persists in this prototype system .
we note that for the cases @xmath193 , there is a constant term in the pressure due to the non - zero curvature of the free - surface at this order .
it is also of interest to compare to the moffatt wedge flow solutions , given in appendix [ app : wedge ] , eqs .
( [ appeq : ns])([eq : tocomp_p_thin ] ) .
accounting for the change of reference frame and the previous remark about non - zero curvature , we see agreement between order of magnitudes for all of these variables , noting that @xmath194 to compare between moffatt wedge and thin droplet settings . for the behaviour of the free - surface slope to match to the outer region , all three slip models have the _ same _ behaviour at leading order , that of @xmath195 } , \quad \mbox{i.e . }
\quad { \frac{\partial{\phi}}{\partial{\xi } } } \sim c_a + \dot{a}{\left [ { \frac{\ln\xi}{c_a^2 } + \frac{1+\ln c_a}{c_a^2 } + c_b } \right ] } , \quad \mbox{as } \xi\to{\infty } , \label{eq : innmatch}\end{aligned}\ ] ] and thus @xmath196 } , \quad \mbox{as } ( a - x)/\delta\to{\infty},\label{eq : innmatchh}\end{aligned}\ ] ] and , as performed in @xcite , matching the cube of the free - surface slope yields the evolution equation @xmath197 } = c_a^3 a^6 - 27 , \label{eq : srr}\end{aligned}\ ] ] with @xmath176 and @xmath186 still to be set by the imposed contact angle condition , which we will now consider .
the matching of the cube of the free surface slope may be justified through asymptotics , and is due to the specific structure of expressions ( [ eq : outbehav ] ) and ( [ eq : innmatchh])where for the logarithmic terms in @xmath198 to balance , a cubic power is necessary .
motivated by the asymptotic analysis of hocking , @xcite , where matching was performed between inner and outer regions through another intermediate region ( the purpose of the intermediate region purely to justify the matching of the cubes @xcite ) , we describe an intermediate region relevant for our analysis in appendix [ app : inter ] , for the interested reader .
[ [ contact - angle - condition - for - slip - models ] ] contact angle condition for slip models : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + as discussed in the introduction , a number of different contact angle conditions have been imposed in the literature for slip models .
the conditions discussed were that the dynamic contact angle always equals the static value , and two other relations in ( [ eq : introcarel ] ) , and in the thin - film regime here these correspond ( for the right contact point ) to @xmath199^m-1).\end{aligned}\ ] ] contact angle condition ( i ) for thin droplet spreading has been used in @xcite . in this instance
the asymptotic procedure follows as above , with , in the inner region @xmath200 .
this thus determines @xmath201 , and on using the solutions from ( [ eq : p1_0])([eq : p1_2 ] ) , we find @xmath202 , which must vanish for this boundary condition and hence @xmath203 .
in contrast , both contact angle conditions ( ii ) and ( iii ) can give a leading order solution for the spreading rate using only the outer solution , provided @xmath204 , as shown by hocking @xcite for a form of condition ( ii ) and based on work in @xcite .
in this case the evolution is of the form @xmath205 for ( ii ) and @xmath206^m - 1)$ ] for ( iii ) , forming very different evolutions .
such an imposed order - of - magnitude for @xmath207 , ostensibly just a constant of proportionality ( and thus perhaps expected to be @xmath68 ) , appears rather restrictive .
additionally , if no inner ( slip ) region is sought at leading order for the droplet evolution , care should be taken to ensure that conservation of mass for the droplet is not violated . if the asymptotic analysis as performed for contact angle condition ( i ) is considered , then contact angle conditions ( ii ) and ( iii ) in the inner region give respectively @xmath208 and thus provided @xmath209 , which is in particular true if @xmath210 , then @xmath201 .
as @xmath211 , then contact angle condition ( ii ) is only valid in this regime if @xmath212 , i.e. if @xmath213 , whereas ( iii ) is valid for any @xmath207 satisfying @xmath214 ( due to the requirement for @xmath201 ) . as @xmath202 for the slip models in ( [ eq : p1_0])([eq : p1_2 ] ) , then if @xmath215 in ( iii ) then @xmath203 as for contact angle condition ( i ) .
otherwise , for conditions ( ii ) and ( iii ) respectively @xmath216 we reiterate that if @xmath207 , which originally appeared as a constant of proportionality and most naturally would not be expected to depend on the spreading rate , is @xmath68 , then this asymptotic analysis fails for boundary condition ( ii ) unless @xmath217 . in this case , both conditions ( ii ) and ( iii ) are equivalent , and so for @xmath210 , boundary condition ( iii ) is preferred .
the slip models satisfy the spreading rate equation @xmath218 } = a^6 - 27.\label{eq : sr_slip}\end{aligned}\ ] ] we finally note that provided @xmath134 is far from the equilibrium value , @xmath219 , equation ( [ eq : sr_slip ] ) gives the radius behaviour of @xmath220 , where @xmath221 $ ] .
this logarithmic correction to a @xmath222 behaviour agrees with the rigorous result of @xcite , who showed that using a slip model for droplet spreading alters the naively expected spreading rate only logarithmically through deriving suitable estimates for integral quantities such as the free energy and the length of apparent support . [
[ precursor - film - models ] ] precursor film models : + + + + + + + + + + + + + + + + + + + + + + for the precursor film models , we follow @xcite and find the inner region behaviour when the outer region is approached ( and in outer variables ) as @xmath223 where @xmath224 and @xmath225 , which for the three disjoining pressure models ( schwartz and eley , sharma , and pismen and pomeau , respectively ) gives @xmath226 finally , matching the behaviour from ( [ eq : pfinbehav ] ) to the outer behaviour gives @xmath227 } = a^6 - 27 .
\label{eq : sr_pf}\end{aligned}\ ] ] [ [ comparisons ] ] comparisons : + + + + + + + + + + + + the last model we will discuss in this section is the interface formation model , whose analysis is substantially more involved . however , it was shown in @xcite that the equivalent spreading rate is given by @xmath228 } = a^6 - 27 , \label{eq : sr_ifm}\end{aligned}\ ] ] where @xmath229 was numerically determined from a seventh order system of odes , dependent upon the nondimensional numbers associated with the model@xmath123 , @xmath124 , @xmath125 , @xmath126 , and @xmath127 .
we are now able to draw comparisons between all of these models in this droplet spreading situation .
as we saw in ( [ eq : sr_slip ] ) , the three slip models ( i.e. navier slip , and inverse linear / quadratic slip ) all give equivalent spreading rates , but their behaviour depends on the contact angle condition through @xmath186 , which is zero for condition ( i ) , but nonzero and given in ( [ eq : cbvals ] ) for ( ii ) and ( iii ) , provided the coefficient of proportionality is in the regime considered . the spreading rates obtained through matching for these slip models become equal if the slip length for condition ( i ) , @xmath230 is given by @xmath231 , where @xmath232 .
the slip and precursor film models give equivalent spreading rates if @xmath233 ( seen from ( [ eq : sr_pf ] ) ) , and the individual precursor film models comparable , but reliant on numerically computing the integrals in ( [ eq : ell ] ) . finally , these models predict equivalent spreading rates to the interface formation model ( seen in ( [ eq : sr_ifm ] ) ) if @xmath234 , or @xmath235 .
thus , even though the original pdes for these models are different , we obtain nearly identical behaviours if the parameters ( e.g. parameters @xmath54 , @xmath186 , @xmath99 , and @xmath124 ) are chosen appropriately .
these results are summarised in table [ tab ] . as these results are for the prototype slow spreading thin droplet situation
, it would be of significant interest to see comparisons between the various models in situations where the underlying physics they capture is likely to be significant , such as for rapid wetting where dynamic contact angle variation could be more pronounced ( although noting that contact angles must remain small for the thin - film approximation to remain valid ) . the work of eggers @xcite , which extends the analysis beyond the quasistatic assumption for navier - slip and inverse linear slip models with bc ( i ) , provides some results of relevance in this pursuit
. cccc model & @xmath236 & @xmath237 & @xmath238 + interface formation & ' '' '' & @xmath239 & ' '' '' + slip with bc ( i ) & @xmath240 & @xmath54 & @xmath241 + slip with bc ( ii ) & @xmath242 & @xmath54 & @xmath243 , @xmath186 in ( [ eq : cbvals])(a ) + slip with bc ( iii ) & @xmath244^m-1)$ ] & @xmath54 & @xmath243 , @xmath186 in ( [ eq : cbvals])(b ) + precursor film & ' '' '' & @xmath99 & @xmath245 +
we have revisited the problem of thin two - dimensional droplet spreading , a useful prototype system involving moving contact lines , and extended the comparative analysis of savva and kalliadasis @xcite to include a disjoining pressure model arising from diffuse - interface theory , a variety of contact angle conditions applied to slip models [ specifically the extension to give the results for slip with bc ( ii ) and ( iii ) in table [ tab ] ] , and compared to results from sibley _
et al . _
on the interface formation model @xcite in this setting . a summary of the results may be found in table [ tab ] , where an effective redefining of the microscopic parameter of each model is shown , which will lead to each model giving the same spreading behaviour as the interface formation model .
it should be stressed that this work is for the slow spreading of thin two - dimensional droplets .
there remains debate in the literature about the best models to apply when moving contact lines are present , and further studies in other situations such as rapid / inertial motion @xcite are needed to clarify several open questions .
additionally , although there exists a wide range of models ( many of which have been discussed here ) there also remains the need for improved models based on first principles .
for example , progress has been made to more accurately capture effects such as long - range fluid - fluid interactions @xcite from the microscale , and density functional theory appears a promising route @xcite , but the determination of accurate coarse - grained approximations and associate numerical computations for contact lines especially in dynamic settings is ongoing .
this work compares three different slip models i.e. navier - slip and two nonlinear slip models , to three disjoining pressure models and the interface formation model .
the three slip models are based on those of navier @xcite for navier - slip , ruckenstein and dunn @xcite , and greenspan @xcite , for ` inverse linear ' slip and from haley and miksis @xcite for ` inverse quadratic ' slip . for the prototype quasistatic droplet spreading situation
we have investigated it was found that all of these slip models give the same equation for the rate of change of the droplet radius at leading order in the matched asymptotic analysis . as such
, all three slip models are described as ` slip ' in the summary table [ tab ] .
three boundary conditions for the contact angle are considered for these three slip models , being that the dynamic contact angle remains at its static value , and two conditions that perturb the dynamic contact angle from its static value through relationships with the contact line velocity .
the specific forms can be found in the second column of table [ tab ] .
the effect of including a relationship between dynamic contact angle and contact line velocity is to have an effective shift of the slip - length in the slip models , this shift being related to the specific form of the relationship and its associated proportionality constant and exponent .
it is interesting to note that dynamic variation of the contact angle has also been considered in the previous numerical simulations of @xcite ( of a form similar to bc ( ii ) in table [ tab ] here ) .
these simulations also saw an effective shift in slip - length suggesting the present findings may be robust beyond the thin - film limit .
the three disjoining pressure models are based respectively on van der waals interactions , a model introduced for polar fluids , and a liquid - gas diffuse - interface model .
these three models are used in the works of schwartz and eley @xcite , sharma @xcite , and pismen and pomeau @xcite , and all model the situation of an infinite precursor film extending away from the droplet . once again , these three models all lead to the same equation for the rate of change of the droplet radius in the matched asymptotic analysis , and hence are listed together as ` precursor film models ' in table [ tab ] .
we note that these direct comparisons have been made for droplet spreading on a homogeneous substrate . in the previous analysis @xcite
, it was shown that whilst the spreading rate equations from matching may agree , different dynamic behaviour may be realised for spreading on heterogeneous substrates .
this is due to the quite different forms of the pdes , and indeed the physical basis of the models . unlike for homogeneous substrates ,
for heterogeneous substrates the system may have multiple equilibria , so that any variation in the dynamics between models will occur most frequently for initial conditions that lie near boundaries separating the basis of attraction to these different equilibria .
the work of @xcite also included the effects of gravity and of substrate roughness , as mentioned , based on the works in @xcite .
other features for droplet spreading include chemically heterogeneous substrates @xcite , patterned substrates @xcite , random ( rather than deterministic ) substrates @xcite , and inclined substrates @xcite .
the understanding of heterogeneous substrates was also considered an important future issue for the contact line problem in the recent review @xcite , and to complement the body of work outlined above on chemical and topological substrate patterns , random disorders and inclinations , it would be of interest to further these analyses by considering how they are affected by the use of the dynamic contact angle models involving slip or interface formation discussed in the present communication .
finally , we note that the main results here were for two - dimensional droplets , however the main result of the comparability of the models should hold in three dimensions , as well as for other geometries .
this is as the asymptotic result relies on the behaviour in the inner asymptotic regions , which in 3d will be the same as for 2d , provided that the contact line variations occur at lengthscales longer than the microscopic lengthscale ( i.e. @xmath54 or @xmath99 ) .
we are grateful to the editor and all anonymous referees for the many useful comments and suggestions .
we acknowledge financial support from erc advanced grant no . 247031 , and imperial college london through a dtg international studentship .
the stream function , velocity , stress and pressure behaviours for the moffatt wedge geometry , for navier - slip in ( [ eq : psiforms ] ) , are found as @xmath246 these results show clearly how the navier - slip model removes the singularity in the stress as the contact line is approached , but a logarithmic singularity in the pressure remains .
similarly for the general slip case in ( [ eq : psiforms ] ) , we obtain @xmath247 + ( 2+n)[c_{3s}\cos((2+n)\theta ) + c_{2s}\sin((2+n)\theta)]\right ) , \nonumber\\ & v \sim -\frac{(2+n)\psi}{r } , \qquad p \sim 4(1+n ) r^n c_{1s } \frac{\cos(n(\theta_c-\theta))}{\cos(n\theta_c ) } , \qquad t_{\theta\theta } \sim -t_{rr } , \nonumber\\ & t_{rr } \sim 2(1+n ) r^n \left ( ( 2+n)[c_{3s}\cos((2+n)\theta ) + c_{2s}\sin((2+n)\theta ) ] + \frac{c_{1s}n\cos(n(\theta_c-\theta))}{\cos(n\theta_c ) } \right ) , \nonumber\\ & t_{r\theta } \sim 2(1+n ) r^n { \left [ { c_{2s}\frac{(2+n)\sin((2+n)(\theta_c-\theta ) ) } { \sin((2+n)\theta_c ) } -c_{1s } \frac{n\sin(n(\theta-\theta_c))}{\cos(n\theta_c ) } } \right ] } , \label{eq : tocomp_p_thin}\end{aligned}\ ] ] where @xmath248 we see from this solution that @xmath79 does not recover the navier - slip solution as this is a degenerate solution of the biharmonic equation .
however , the two cases considered by haley and miksis @xcite , corresponding to @xmath57 and @xmath58 here , will resolve both the pressure and stress singularities .
in contrast to the navier - slip model , they both give zero pressure and stresses as the contact line is approached the pressure and stresses being @xmath70 and @xmath31 in the two cases respectively .
there is , however a possibility of removing the logarithmic pressure singularity but retaining an @xmath68 stress behaviour .
previously in ( [ eq : presordwlog ] ) we suggested that for the stresses to be @xmath68 then the stream function must be prescribed by @xmath249 with @xmath250 to be determined .
we have also seen that this form , which corresponds to the navier - slip condition , yields @xmath251 pressure as @xmath23 due to the @xmath69 terms in the divergence of the stress .
however , it is also possible to cancel the logarithmic pressure terms by relaxing the assumption of constant surface tension , thus obtaining a finite pressure solution . more specifically , from ( [ eq : bctanst ] ) , where @xmath252 is the surface gradient of @xmath40 , and which given in terms of the usual gradient operator is @xmath253 , we can determine the dynamic boundary condition as @xmath254 now using the pressure equations from ( [ eq : stressinpolar ] ) with @xmath249 gives @xmath255 and thus ( at leading order in @xmath256 ) @xmath257 . to have a pressure without a singularity
requires @xmath258 , and this third order ode has the general solution @xmath259 where @xmath260 are constant coefficients . from ( [ eq : eqf ] ) and ( [ eq : bhedeg ] ) , we thus impose stream function form @xmath261 keeping two orders in @xmath256 so as to obtain the leading order term in the pressure behaviour .
we find after applying the boundary conditions ( including ( [ eq : bctsgradsig ] ) ) with navier - slip @xmath262 displaying the mathematical possibility of such a flow .
we note here as part of the result the form of the surface tension was forced to satisfy @xmath263 which gives a reasonable result for the surface tension @xmath40 , that it is a constant to leading order when @xmath264 , the regime in which we are interested .
this also gives a motivation as to why a model which allows for variable surface tension , i.e. surface tension relaxation , may be able to predict a flow which has no logarithmic pressure singularity , but finite and non - zero stresses at the contact line .
such surface tension relaxation is a feature of the interface formation model @xcite , where the surface tension is given as @xmath265 and the surface density is a function of position and time , i.e. @xmath266 .
motivated by the analysis of @xcite and the matching conditions ( [ eq : innmatch ] ) , we introduce the intermediate scalings @xmath267 where @xmath268 .
neglecting exponentially small terms , ( [ eq : ndlwin ] ) transforms to @xmath269 \right\}=0,\end{aligned}\ ] ] so that to leading order in @xmath270 we have @xmath271 where @xmath272 is a constant of integration . expanding these for @xmath273 and rewriting in inner variables , we have @xmath274},\end{aligned}\ ] ] which gives @xmath275 to match with ( [ eq : innmatch ] ) . to match this intermediate solution to the outer solution , ( [ eq : intmpower ] )
clearly suggests matching should be performed with the cube of the free surface slope , indeed we find from ( [ eq : intmpower ] ) using scalings ( [ eq : intscals ] ) ( with our result that @xmath275 ) that @xmath276 } } \right)},\end{aligned}\ ] ] which is the same result as would be found directly from the inner region behaviour ( [ eq : innmatch ] ) . from the outer region behaviour ( [ eq : outbehav ] ) we have @xmath277},\end{aligned}\ ] ] and thus the matching of these provides the spreading rate result given in ( [ eq : srr ] ) .
lauga , e. , brenner , m. , and stone , h. icrofluidics : the no - slip boundary condition . in tropea ,
c. , yarin , a. , and foss , j. , editors , _
springer handbook of experimental fluid mechanics _ , pages 12191240 .
springer berlin heidelberg ( 2007 ) .
sibley , d. n. , savva , n. , and kalliadasis , s. ( 2012 ) . slip or not slip ?
a methodical examination of the interface formation model using two - dimensional droplet spreading on a horizontal planar substrate as a prototype system .
, 24(8):082105 .
goddard , b. d , nold , a. , savva , n. , yatsyshin , p. , and kalliadasis , s. ( 2013 ) .
unification of dynamic density functional theory for colloidal fluids to include inertia and hydrodynamic interactions : derivation and numerical experiments .
, 25:035101 . | the motion of a contact line is examined , and comparisons drawn , for a variety of models proposed in the literature . pressure and stress
behaviours at the contact line are examined in the prototype system of quasistatic spreading of a thin two - dimensional droplet on a planar substrate .
the models analysed include three disjoining pressure models based on van der waals interactions , a model introduced for polar fluids , and a liquid - gas diffuse - interface model ; navier - slip and two nonlinear slip models are investigated , with three microscopic contact angle boundary conditions imposed ( two of these contact angle conditions having a contact line velocity dependence ) ; and the interface formation model is also considered . in certain parameter regimes
it is shown that all of the models predict the same quasistatic droplet spreading behaviour . _
keywords : _ contact line , slip , diffuse - interface , precursor film , disjoining pressure , interface formation . |
the properties of 4u0142 + 61 ( white et al .
1987 ) remained puzzling for a long time , owing to confusion with a nearby pulsating and transient be / neutron star system rxj0146.9 + 6121 ( motch et al . 1991 ; mereghetti et al . 1993 ) .
the 110 kev spectrum is extremely soft ( power law photon index of @xmath6 , white et al .
1987 ) and led to the initial classification of 4u0142 + 61 as a possible black hole candidate .
asca observations provide evidence for a @xmath7 kev blackbody component contributing @xmath8% of the 0.510 kev band x ray flux ( white et al .
the x ray luminosity of 4u 0142 + 61 has not shown substantial secular variations around an average value of @xmath9 erg s@xmath3 ( assuming a distance of 1 kpc ) . despite the small error box ( 5 radius ) , no optical or ir counterpart has yet been identified , down to @xmath10 , @xmath11 , @xmath12 and @xmath13 ( steinle et al .
1987 ; white et al . 1987 ; coe & pightling 1998 ) .
these limits rule out the presence of a massive companion . using data from the exosat archive , israel et al .
( 1994 ) discovered pulsations at 8.7 s , which were later confirmed with rosat ( hellier 1994 ) .
no delays in the pulse arrival times caused by orbital motion were found , with upper limits on @xmath14sin @xmath15 of about @xmath16 lt
s for orbital periods @xmath17 between 7 min and 12 hr ( israel et al .
tighter upper limits on the @xmath18 ( @xmath19 lt
s for 70s@xmath20 2.5days ) have been recently obtained with a rxte observation ( wilson et al . 1998 ) .
this yielded strong constraints on the orbital inclination and the mass of the possible companion star in the case of normal or helium main sequence star and giants with helium core .
a white dwarf companion would be compatible both with current optical photometric and pulse arrival time limits . [
cols="^,^,^,^,^,^,^,^ " , ] we also fit the lecs and mecs spectra of observation a with other two component spectral models , such as a power law with a cut off ( @xmath21/dof = 453.2/344 ) , two blackbodies ( @xmath21/dof = 503.1/344 ) and a broken power law ( @xmath21/dof = 400.6/344 ) .
all these models gave substantially worse fits than the power
law plus blackbody model .
similar results were obtained for the spectrum of observation d. since the blackbody peaks are close to the lower end of the mecs energy range we also checked these results by analysing the lecs data only ( which cover a wider spectral range than the mecs;see table 2 and discussion ) .
we also checked the stability of the results obtained by rebinning by a factor of @xmath03 the pha channels in the lecs and mecs spectra . again all the bins which were consistent with zero after background subtraction were rejected from the analysis .
no significant variation in the spectral parameters and uncertainties were found for either models .
the data from the high pressure gas scintillation proportional counter ( hpgspc ) and the phoswich detector system ( pds ) did not hold any useful information on 4u0142 + 61 .
in fact , due to the large fovs of these intruments and the steep spectrum of 4u0142 + 61 , the counts above 10 kev were likely dominated by the nearby source rx j0146.9 + 6121 .
the arrival times of the 0.510 kev photons from 4u0142 + 61 were corrected to the barycenter of the solar system and 1s binned light curves accumulated for each observation .
the average count rates are reported in table1 .
the mecs counts were used to determine the 4u0142 + 61 pulse period . the data from observations a , b and d were divided into 6 , 4 , and 5 time intervals , respectively and for each interval the relative phase of the pulsations was determined .
these phases were then fit with a linear function giving a best fit period of @xmath22 s for observation a ( see table3 ) . for observation b a period of @xmath23 s was obtained , while during observation c pulsations were not detected owing to poor statistics ( @xmath0 11000 photons ) to detect such a weak ( @xmath06% pulsed fraction ) signal .
finally for observations d we determined a period of [email protected] .
beppo_sax pulse period values are plotted in fig.4 together with the previous measurements ( see sect.3 for the rxte measurement ) .
the background subtracted light curves from observation a , folded at the best period in different energy ranges ( fig.3 ; first three panels ) show a double peaked profile ( see also white et al .
the pulsed fraction ( semiamplitude of modulation divided by the mean source count rate ) was [email protected]% , [email protected]% and 13@xmath242% in the 0.51.5 kev , 1.54.0 kev and 4.010 kev energy bands , respectively .
similar results are obtained for observations b and d. flux variations on several timescales are usually displayed by high accretion rate x ray pulsars ; we found no short or long term variations in any of the _ beppo_sax accumulated light curves .
the upper two panels of fig.5 show the pulsed fraction versus energy during observation a for the first two harmonics .
these values were obtained by fitting the corrisponding light curves with two sinusoidal functions .
the first harmonic shows a nearly constant value ( @xmath0 6% ) up to 4kev , while at higher energies increases to about 15% . a constant value of 57%
is inferred for the second harmonic .
we calculated also the root mean square variability ( rms ; defined as @xmath25 devided by the mean source count rate , where @xmath26 and @xmath27 are the observed and expected variance , respectively ) of the folded light curve at different energies ( third panel ) .
finally the ratio between the power law and the total flux for the blackbody plus power law spectral model is shown ( lowest panel ) . from the comparison of these quantites
we can infer that : ( i ) since the rms variability of the folded light curves is @xmath28 times the geometric sum of the amplitudes in all available harmonics ( mainly the 1@xmath29 and 2@xmath30 in the present case ) , it is apparent that the behaviour is consistent with the derived amplitudes of the 1@xmath29 and 2@xmath30 harmonics ; ( ii ) there is evidence for an increase of the pulsed fraction ( @xmath0 2@xmath31 confidence level ) at energies above 5 kev ; ( iii ) there is not a simple relationship between the pulsed fraction and the flux in either of the two spectral component adopted in our analysis .
a set of four phase
resolved spectra ( phase boundaries 0.06 , 0.26 , 0.5 , 0.78 ) were accumulated for observation a ( i.e. the observation with the highest number of source counts ) .
these were then fit with the power
law plus blackbody model described in sect.2.1 , with n@xmath32 fixed at the phase
averaged best
fit value .
initially , the blackbody temperature was fixed at its phase averaged best
fit value and only the power law parameters and blackbody normalisation were allowed to vary .
the fits were then repeated with @xmath4 fixed and blackbody parameters and power law normalisation free ( see fig.3 ) .
no significant changes were detected for @xmath4 and @xmath5 as a function of the pulse phase .
similar results were obtained for the fluxes of the two spectral component .
the 4u0142 + 61 position was included in a rxte observation pointed at the nearby high mass x ray binary pulsar rx j0146.9 + 6121 ( from 1996 march 28 11:16:48 to 22:09:20 ; @xmath020 ks of effective exposure time ) .
the results presented here are based on data collected in the so called good xenon " operating mode with the proportional counter array ( pca , jahoda et al .
the pca consists of 5 proportional counters operating in the 260 kev range , with a total effective area of approximately 7000 @xmath33 and a field of view , defined by passive collimators , of @xmath34 fwhm .
the data consist of the time of arrival and the pulse height of each count . in order to minimise the contamination from rxj0146.9 + 6121 ( a much harder spectrum x ray pulsar )
, we considered only photons in the 24 kev energy interval . in order to reduce the background ,
only the counts detected in the first xenon layer of each counter were used .
we obtained for 4u0142 + 61 a best pulse period of [email protected] ( see table3 and fig.4 ) .
figure6 shows the corresponding folded light curve . despite the uncertainties in the background subtraction ( the contribution from rxj0146.9 + 6121 is difficult to estimate )
, it is apparent that the rxte / pca pulse profile of 4u0142 + 61 is similar to that observed with _
the rxte value we obtained is consistent to within the statistical uncertainties with that of wilson et al .
( 1998 ; energy range 3.79.2 kev ) .
the somewhat different pulse shape is likely due to the different energy band used .
several models have been proposed in order to explain the nature of the `` anomalous '' x ray pulsars .
mereghetti & stella ( 1995 ) proposed that these sources form a homogeneous subclass of accreting neutron stars , perhaps members of low mass x ray binaries ( lmxbs ) , which are characterized by lower luminosities ( @xmath35 erg s@xmath3 ) and higher magnetic fields ( @xmath36 g ) than classical lmxbs .
however the lack of evidence for a binary nature from any of these systems ( mereghetti et al . 1998 ;
wilson et al . 1998 ) argues in favor of models in which the x ray emission originates from a compact object that is not in an interacting binary system .
van paradijs et al . (
1995 ) proposed that axps are young population i objects , which originate from the evolution of short orbital period high mass x
ray binaries ( hmxbs ) , following the expansion of the massive star and the onset of unstable roche lobe overflow , before central hydrogen is exhausted ( see also cannon et al .
the resulting common
envelope phase should cause the neutron star to spiral in and disrupt the companion star after the so called thorne zytkov stage .
therefore , these sources should consist of an isolated neutron star accreting matter from a residual disk with a mass in the @xmath37 m@xmath38 range .
the blackbody component in 4u0142 + 61 and , likewise , other axps has been interpreted as evidence for quasi
spherical accretion onto an isolated neutron star formed after common envelope and spiral in of a massive short period ( p@xmath39@xmath40 1yr ) x ray binary . in this case
the remains of the envelope of the massive star might produce two different types of matter inflow : a spherical accretion component with low specific angular momentum giving rise to the blackbody component and a high specific angular momentum component which forms an accretion disk and is likely responsible for the power
law emission ( ghosh et al . 1997 ) .
although some problems remain open in this scenario , we note that a post common envelope evolutionary scenario for 4u0142 + 614 is also suggested by the inference that about half of the x ray absorbing material is close to the source ( white et al . 1996 ) . in this context it is interesting to note that independent evidence supports the view that the binary x ray pulsars 4u162667 and hd49798 were formed following a common envelope and spiral in evolutionary phase ( see angelini et al . 1995 and israel et al .
1997 and references therein ) .
the different properties of the companion stars in these two cases may reflect the fact that unstable mass transfer set in at different evolutionary phases in the nuclear evolution of their progenitors ( in turn , reflecting different initial masses and orbital periods ; see also ghosh et al .
1997 ) .
thompson & duncan ( 1993 , 1996 ) proposed that the axps are `` magnetars '' , neutron stars with a superstrong magnetic field ( @xmath41 g ) .
this proposal is supported by the similarity in the pulse periods ( 8.05 , 7.5s and 5.16s ) and of p values measured in the soft @xmath42ray repeaters sgr052066 , sgr1806 - 20 and sgr1900 + 14 , respectively ( kouveliotou et al .
1998 ; hurley et al . 1998 ) .
if this connection proved correct , axps , would be quiescent soft @xmath42ray repeaters .
heyl & hernquist ( 1997 ) argued that their emission may be powered by the cooling of the core through a strongly magnetised envelope of matter ( made up mainly by hydrogen and helium ) .
pulsations would originate from a temperature gradient on the surface of the star .
moreover heyl & hernquist ( 1998 ) showed that in this scenario , spin down irregularities , observed in the period history of two `` anomalous '' x ray pulsars ( 1e2259 + 58 and 1e104859 ) , may be simply accounted for with glitches like those observed in young radio pulsars .
the 0.510 kev spectrum of 4u0142 + 61 is well modelled by the sum of an absorbed steep power
law and a low energy blackbody .
the latter component was introduced by white et al .
( 1996 ) based on the asca data ; the corresponding blackbody radius and flux were @xmath43 km ( at 1 kpc ) and @xmath040% of the total , respectively . neither the power
law photon index nor the blackbody temperature show evidence of changes across different _ beppo_sax observations .
we found marginal evidence ( at about 2@xmath31 confidence level ) for changes relative to previous observations .
a comparison of the asca and _
beppo_sax results show that : ( i ) the power
law photon index increases by about 4% ( observations a and d ) ; ( ii ) the blackbody radius decreased by about 37% ( a ) and 12% ( d ) ; ( iii ) the blackbody flux decreased to @xmath030% ( a ) and @xmath035% ( d ) of the total flux therefore showing a @xmath010%5% decrease with respect to that of the asca observation ; ( iv ) the 0.510 kev total flux decreased of about 1512% . by fixing the parameters
n@xmath32 , @xmath5 and @xmath4 to the values inferred by asca , the _ beppo_sax observation a spectrum gives a @xmath44/dof of 1.9/348 . in this case
the blackbody component accounts for @xmath030% of the total flux ( @xmath010% lower than that of asca ) . as an additional test
we also merged the data for observation a ( mecs 2 and 3 only ) and d and fit them with the power
law plus blackbody model ( see table2 ) .
while a small change in the spectral parameters is found relative to observations a and d , separately , the blackbody flux is @xmath030% of the total . to remove possible effects introduced by the vicinity of the lower end of the energy range of the mecs and the peak of the blackbody component , we also fitted the spectrum of observations a , d and a+d by using only the data as accumulated by the lecs , the energy band of which uninterruptely covers the range 0.59.0 kev and allows a better fit in the energy interval ( 14 kev ) where the power law and the blackbody components overlap .
again we obtained results similar to the lecs+mecs case , but with a larger uncertainty .
all these results point to a marginal variation of the spectral parameters of 4u0142 + 61 , if any .
the _ beppo_sax data suggest that the pulsed fraction for energies above 4kev decreased ( from 25%@xmath245% with asca to 13%@xmath242% with _ beppo_sax ; 90% uncertainties ) . for these energies
the counts are almost entirely dominated by the power law component ( in the spectral model assumed ) .
the pulse periods measured by _
beppo_sax show that 4u0142 + 61 has continued its secular spin down during 19961998 ( see fig.4 ) .
the period derivative inferred from the _
beppo_sax observations alone ( @xmath06.0@xmath45 @xmath110@xmath2 s s@xmath3 ) is consistent with the average spin down rate ( @xmath02@xmath110@xmath2 s s@xmath3 ) inferred over the 19yr span of the historical dataset ( note that in the period list we did not include the low significance detections inferred on 1985 november 11 and december 11 with exosat me , and on 1991 february 13 with rosat hri ; israel et al .
the inferred p for 4u0142 + 61 is consistent with the uncertainties of all pulse period measurements except , perhaps , one ( the 1979 _ einstein _
sss one ; see fig.4 ) implying that , so far , no `` glitches '' have been observed yet .
we would like to thank giancarlo cusumano for providing the mecs off axis matrices and the _ beppo_sax mission planning for their constant help .
the authors also thank k. long the comment of which helped to improve an earlier version of this paper .
this work was partially supported through asi grants .
angelini l. , white , n.e . , nagase f. , et al . , 1995 , apj 449 , l41 boella g. , chiappetti l. , conti g. , et al . , 1997 ,
a&as 122 , 327 cannon r.c . ,
eggleton p.p . ,
zytkov a.n . , et al . , 1992 ,
apj 386 , 206 coe m.j . , pightling s.l . , 1998 ,
mnras 299 , 223 corbet r.h.d .
, smale a.p . , ozaki m. , koyama k. , iwasawa k. , 1995 , apj 443 , 786 ghosh p. , angelini l. , white n.e . , 1997 ,
apj 478 , 713 gotthelf e.v . , vasisht g. , 1998 , new astronomy 3 , 293 hellier c. , 1994 , mnras 271 , l21 heyl j.s . ,
hernquist l. , 1997 , apj 489 , l67 heyl j.s . , hernquist l. , 1998 , apj 506 , l61 hurley k. , kouveliotou c. , murakami t. , et al . , 1998 iau circ .
7001 israel g.l . , mereghetti s. , stella l. , 1994 , apj 433 , l25 israel g.l . , stella l. , angelini l. , et al . , 1997 , apj 474
, l53 jahoda k. , swank j. h. , giles a. b. , et al . , 1996 ,
spie 2808 , 59 kouveliotou c. , dieters , s. , strohmayer t. , et al .
, 1998 , nat 393 , 235 mereghetti s. , stella l. , 1995 , apj 442 , l17 mereghetti s. , stella l. , de nile f. , 1993 , a&a 278 , l23 mereghetti s. , israel g.l . , stella l. , 1998 , mnras 296 , 689 motch c. , belloni t. , buckley d. , et al .
, 1991 , a&a 246 , l24 oosterbroek t. , parmar a.n .
, mereghetti s. , israel g.l . , 1998 , a&a 334 , 925 parmar a.n .
, martin d.d.e . , bavdaz m. , et al . , 1997 ,
a&as 122 , 309 parmar a.n . , oosterbroek t. , favata f. , et al .
, 1998 , a&a 330 , 175 steinle h. , pietsch w. , gottwald m. , graser u. , 1987 , ap.&ss 131 , 687 sugizaki m. , nagase f. , torii k. , et al . , 1997 ,
pasj 49 , l25 thompson c. , duncan r.c . , 1993 , apj 408 , 194 thompson c. , duncan r.c . , 1996 , apj 473 , 322 torii k. , kinugasa k. , katayama k. , et al . , 1998 , apj 503 , 843 van paradijs j. , taam r.e .
, van den heuvel e.p.j . , 1995 , a&a 299 , l41 vasisht g. , gotthelf e.v . , 1997 ,
apj 486 , l129 wilson c.a .
, dieters s. , finger m. , et al .
1998 , apj 513 , 464 white n.e . ,
mason k.o . ,
giommi p. , et al .
, 1987 , mnras 226 , 645 white n.e . ,
angelini l. , ebisawa k. , et al . , 1996 , apj 463 , l83 | the 8.7s x ray pulsar 4u0142 + 61 was monitored by _
beppo_sax between january 1997 and february 1998 .
this source belongs to the rapidly growing class of anomalous " x ray pulsars ( axps ) which have pulse periods in the 612s range and no plausible optical , infrared or radio counterparts .
the _
beppo_sax periods measurements show that 4u0142 + 61 continues its spin down at a nearly constant rate of p@xmath02@xmath110@xmath2 s s@xmath3 . the 0.510 kev pulse shape is double peaked .
the phase averaged spectrum can be described by a steep absorbed power
law ( @xmath4 @xmath0 4 ) plus a blackbody with @xmath5 @xmath0 0.4 kev .
4u0142 + 61 was also detected serendipitously in a march 1996 rxte observation pointed towards the nearby 1455s x ray pulsar rxj0146.9 + 6121 .
the timing analysis results are reported for this observation and the spin history of 4u0142 + 61 since 1979 is discussed . |
chiral perturbation theory ( chpt ) has been a very useful approach to the theory of the low momentum processes involving mesons and baryons , and has been used in various attempts to explain the baryon magnetic moments .
all the moments in the baryon octet except that of the @xmath2 have been measured .
the @xmath3 transition moment @xmath4 is also known . at the su(3)-symmetric tree level @xcite
, chpt parameterizes these eight measured quantities terms of the two parameters @xmath5 and @xmath6 in the effective lagrangian @xmath7 ) \ , \ ] ] and is able to fit them with an average error of about 0.24 nuclear magnetons @xmath8 . in this expression
, @xmath9 is the usual representation for the baryon - octet in flavor - space , @xmath10 is the quark matrix , @xmath11 is the nucleon mass , and @xmath12 is the electromagnetic field .
the study of the baryon moments beyond the tree level has a long history which can be found in @xcite .
the approaches and conclusions have varied .
caldi and pagels @xcite pointed out that the loop corrections lead to uniquely defined `` nonanalytic corrections '' to the moments proportional to the square root and logarithm of the symmetry breaking parameter @xmath13 of chpt .
in contrast , the analytic corrections involving integer powers of @xmath13 are cutoff dependent in perturbation theory .
the contributions left after the divergences are subtracted are not uniquely defined in the absence of further dynamical information , and mix with the contributions of higher chiral couplings . since these contributions are expected to be small , they have generally been ignored .
caldi and pagels @xcite calculated what were supposed to be the leading corrections to the baryon moments , those of order @xmath14 , and found that they are as large as the tree - level moments and fail to improve the fit to experiment .
these terms were recalculated by gasser _
@xcite , and the logarithmic terms of type @xmath13ln@xmath13 were analyzed by krause @xcite , but without detailed analyses of the moment problem .
the problem has been reexamined in recent years .
@xcite used heavy baryon chiral perturbation theory ( hbchpt ) @xcite to calculate the one - loop corrections to the octet baryon moments allowing both octet and decuplet intermediate states .
their formal analysis included the nonanalytic corrections of types @xmath14 and @xmath13ln@xmath13 and all the counterterms allowed to that order .
however , they assumed in fitting the moment data that the logarithmic corrections and the counterterms were small compared to the terms proportional to @xmath14 , and included only the last in their analysis . the resulting fit to the data was much worse than the tree - level fit if they used tree - level values for the axial coupling constants , but better if they used the axial coupling constants extracted from one - loop fits @xcite to other data , @xmath15 , @xmath16 , and @xmath17 , values about 20% smaller than those derived at tree level .
this sensitivity leaves the significance of the loop corrections unclear .
in contrast , meissner and steininger @xcite considered all octet terms ( analytic and nonanalytic ) which are of order @xmath18 or less in chpt .
they did not include the intermediate decuplet states directly , but argued that some decuplet contributions are included because of the way they determined the couplings .
with the counterterms included , they could fit the seven well measured octet moments precisely , and predicted a @xmath19 transition moment of @xmath20 ) to be compared with the experimental value , @xmath21 .
their principal conclusion was that the chiral expansion appeared to be converging rather well .
the problem has also been studied using the @xmath22 expansion by a number of authors @xcite who showed that the @xmath14 terms are leading order in the expansion , and obtained a number of interesting sum rules for the moments .
@xcite again kept only the leading nonanalytic corrections proportional to @xmath14 and the counterterms in their analysis of the moment data , and obtained a good overall fit , but with seven parameters .
finally , bos _ et al . _
@xcite considered the moments from the point of view of flavor su(3 ) breaking in the baryon octet using a chiral counting and treatment of meson masses different from that usually used and again obtained a successful parametrization using seven parameters to describe eight measured moments .
thus far , the numerically successful approaches to the baryon moments have all used seven parameters to fit eight experimental data .
two of these parameters are the tree - level moment parameters @xmath23 and @xmath24 .
the remaining five are the coefficients @xmath25 of the su(3 ) symmetry breaking counterterms defined as @xmath26 , { \cal m}]\ , + a_4 { \rm tr}\,\bar{b } \{[q,\sigma_{\mu\nu}b ] , { \cal m}\}\ , \nonumber\\ & & + a_5 { \rm tr}\bar{b } [ \{q,\sigma_{\mu\nu}b\ } , { \cal m } ] + a_6 { \rm tr}\bar{b}\{\{q,\sigma_{\mu\nu}b\ } , { \cal m}\}\\ & & + a_7 { \rm tr}\bar{b } \sigma_{\mu\nu}b \,{\rm
tr}{\cal m } q + \ , ... ) \;. \nonumber\end{aligned}\ ] ] here @xmath27 is a matrix used to introduce su(3 ) breaking introduced by the strange quark mass .
a factor @xmath13 has been absorbed in the coefficients of the lagrangian .
there are two terms with the same structures as those in the tree level moment operator ( [ lagt ] ) hidden in the ellipsis
. given the large number of parameters used in the fits , it is natural to ask if the loop corrections matter , and if one can show that the uniquely extractable nonanalytic corrections actually improve the tree - level fit to the baryon moments .
to answer these questions and investigate the expected cancellations and the convergence of the chiral expansion , we have redone the calculations of jenkins _ et al . _
@xcite , corrected several errors in the published work , missing in a contribution from the graph 2c and a factor @xmath28 missing from the decuplet magnetic moment operator .
other corrections are noted in the _ erratum _ to @xcite . ] and reanalyzed the problem of fitting the measured moments from a different point of view . we have followed jenkins _ et al .
_ in retaining only the nonanalytic contributions of types @xmath14 and @xmath13ln@xmath13 in our analysis , but including the effects of the octet - decuplet mass splitting throughout .
the graphs which produce these contributions are shown in figs . 1 and 2 , respectively .
the octet baryon moments are given in the one - loop approximation by @xmath29 where @xmath30 is the tree - level moment , @xmath31 contains the contributions from loops that involve only intermediate spin-1/2 octet baryons , and @xmath32 contains the loop contributions that involve the spin-3/2 decuplet baryons , @xmath33 \ , .\end{aligned}\ ] ] the functions @xmath34 , @xmath35 , and @xmath36 are defined as @xmath37 , & m \geq \delta , \\
\\ - 2\sqrt{\delta^2-m^2 } \ , { \rm ln } \ , [ ( \delta + \sqrt{\delta^2-m^2 } ) / m ] , & m < \delta , \end{array } \right . , \\
l_{(3/2)}(m , \delta , \mu ) & = & m^2 \ , { \rm ln } { m^2 \over \mu^2 } \ , + \ , 2\pi\delta \ , f(m , \delta , \mu ) , \\
l_{(3/2)}^{'}(m , \delta , \mu ) & = & m^2 \ , { \rm ln } { m^2 \over \mu^2 } \ ,
+ \ , { 2\pi \over 3\delta } \ , g(m , \delta , \mu ) , \\
( m , \delta , \mu ) & = & -\delta^3 \ , { \rm ln } { m^2 \over \mu^2 } \ , + \ , \pi m^3 \nonumber \\ & & + \left\ { \begin{array}{ll } 2(m^2-\delta^2)^{3/2 } \ , [ \pi/2 - { \rm arctan } \ , ( \delta / \sqrt{m^2-\delta^2 } ) ] , & m \geq \delta , \\ \\ - 2(\delta^2-m^2)^{3/2 } \ , { \rm ln } \ , [ ( \delta + \sqrt{\delta^2-m^2 } ) / m ] , & m < \delta,\label{g } \end{array } \right . .\end{aligned}\ ] ] the coupling coefficients @xmath38 , @xmath39 , @xmath40 , @xmath41 , @xmath42 are identical to those in @xcite where @xmath43 is denoted by @xmath44 and @xmath42 is denoted by @xmath45 .
the necessary corrections have been included .
the counterterm contributions follow from eq .
( [ counttt ] ) and are given in table i. before examining the loop contributions in detail , we note that the seven well - measured octet moments can be fitted exactly using the five counterterm couplings @xmath25 and the two tree - level parameters @xmath23 and @xmath24 , even in the presence of the loop corrections .. the alternative of making a weighted least - squares fit to all eight measured quantities leads to essentially identical results because of the low weight accorded @xmath46 in the fit .
the unweighted fits used in @xcite distribute the uncertainty in @xmath46 over all the moments without changing the one test of the theory , as will be seen below .
we believe that the present procedure is clearer . ] to see this , we define an @xmath47 matrix @xmath48 which has as its @xmath49th column the matrix elements of the corresponding operator in the physical baryon states @xmath50 and the @xmath3 transition matrix element as in table i. the most general column vector of magnetic moments consistent with this chiral structure is then of the form @xmath51 where @xmath52 is a seven - component column vector of coefficients .
the @xmath53 matrix @xmath54 obtained by omitting the @xmath3 row in @xmath48 is invertible , so coefficients @xmath55 can be found which give exact fits to either the seven well - measured baryon moments , @xmath56 , or to the difference between the measured moments and the loop corrections in this sector , @xmath57 . in the second case ,
the moments are again reproduced exactly when the loop corrections are added in .
the only remaining constraint on the theory is that provided by the @xmath3 transition moment . in the absence of loop corrections , @xmath46 is determined entirely by the seven coefficients @xmath52 through the product @xmath58 with the full matrix @xmath48 , or , equivalently , through the okubo sum rule @xcite @xmath59 the origin of the sum rule can be seen rather simply in the present context .
we can extend the matrix @xmath48 to an @xmath60 matrix @xmath54 by adding a final column with zero entries without affecting the seven baryon moments .
that is , @xmath61 with the same coefficients @xmath52 as determined above .
however , the extended matrix has one zero eigenvalue .
the sum rule is just the inner product of the left null eigenvector @xmath62 with @xmath63 , @xmath64 .
in particular , any set of moments which can be described by the initial seven - component chiral structure must satisfy the sum rule exactly .
the sum rule is , in fact , satisfied rather accurately by the measured moments .
the predicted value of @xmath46 is @xmath65 , to be compared with the measured value @xmath66 .
the discrepancy is at the level of 1.5@xmath67 or 8% of the measured transition moment .
the situation is potentially different when loop corrections are included . while the @xmath14 corrections can be described using only the chiral structures considered so far and therefore satisfy the okubo sum rule exactly @xcite , that is not the case for the corrections proportional to @xmath13ln@xmath13 , a point misstated by those authors .
the logarithmic corrections involve sums of terms of the form @xmath68 multiplied by coupling coefficients .
if the mass dependence of the logarithms could be neglected , the sum rule would hold exactly for external meson masses that satisfy the gell - mann okubo mass formula . as a result ,
the chiral cutoff @xmath69 drops out of the sum rule , and the violation depends only on the logarithm of the mass ratio , @xmath70 , hence would vanish for equal meson masses .
however , the sum rule is violated for @xmath71 , and the logarithmic loop corrections change the prediction for @xmath46 and introduce new structure in the @xmath60 moment space . of order @xmath72 which affects only the @xmath73 moment .
] to investigate this point in more detail , we have constructed exact fits to the seven octet moments using the results for the nonanalytic parts of the loop corrections given in eqs .
( [ loopsum]-[g ] ) using the corrected coupling coefficients from @xcite , and have then calculated @xmath46 . for illustration , we will present the results for the two cases considered in @xcite : ( a ) @xmath15 , @xmath16 , and @xmath17 , the one - loop axial vector couplings derived in @xcite ; and ( b ) @xmath74 , @xmath75 , and @xmath76 , the tree - level couplings . in both cases , we take the decuplet - octet mass difference as @xmath77 gev , and use @xmath78 gev , @xmath79 , @xmath80 , @xmath81 , and a chiral cutoff @xmath82 gev as in that reference . although the sum rule is violated analytically , we find that the violation is remarkably small numerically . in particular , @xmath46 predicted to be 1.508@xmath8 for the parameters in case ( a ) , rather the 1.483@xmath8 as required by the sum rule .
the difference is quite small compared to the experimental uncertainty in @xmath46 .
the results in case ( b ) are similar , with a predicted transition moment @xmath83 .
the chiral parameters for these two fits are given in table ii .
it is evident from table ii that the fitted parameters @xmath6 and @xmath5 are consistent with su(6 ) : @xmath5/@xmath6 @xmath84 3/2 .
that is expected in the usual quark model picture .
the axial coupling constants @xmath34 , @xmath85 , and @xmath86 also satisfy the su(6 ) relations .
the result that loop corrections lead to only small violations of the okubo sum rule appears to be quite robust . in particular , the changes in the loop corrections associated with different treatments of the meson - baryon couplings are largely absorbed by changes in the fitted counterterm parameters .
we conclude that the theory with all the counterterms present is not usefully predictive at the present time .
the results do not , for example , distinguish in a useful way between the standard chiral approach with the nonanalytic contributions singled out , the approach of meissner and steininger @xcite which includes analytic as well as nonanalytic terms , or the approach of bos _ et al . _
@xcite which allows only corrections that break su(3 ) symmetry linearly , fits the data with just the seven parameters above , and satisfies the sum rule exactly .
the insensitivity of the overall fits to loop effects raises the question of whether there is any evidence that these effects are actually important .
the difficulty is that the counterterm contributions are not calculable in the present theory , yet are essential for getting good overall fits to the data .
a breakdown of the detailed contributions to the moments for case ( b ) is given in table iii .
the results for case ( a ) are similar .
the structure of the loop contributions is of considerable interest .
we will examine this separately for the octet and decuplet cases . as shown in table iii ,
the contributions of type @xmath0 from the intermediate octet states are quite large , comparable to the tree - level contributions .
the @xmath87 corrections have the same general magnitude , but opposite signs . as a result
there are large cancellations between these two types of nonanalytic contributions .
this contradicts the arguments of chiral perturbation theory @xcite and the @xmath22 expansion @xcite that the @xmath88 corrections should be small , of the same order of magnitude as the counterterms and ignorable relative to the @xmath14 contributions .
in fact , the smallness of the characteristic ratio @xcite @xmath89 of the two types of nonanalytic terms , approximately -1/22 and -1/17 for @xmath48 a pion or a kaon , is offset by very large ratios of the effective couplings , e.g. , 16.1 and 12.7 for the pion and kaon contributions to the proton moment in case ( b ) , and -28.6 and 9.9 for the @xmath90 , and the large cancellations ensue .
there is no indication in this sector that the chiral perturbation series is converging .
the contributions from intermediate decuplet states introduce a new complication , namely the dependence of the results on the octet - decuplet mass difference @xmath91 .
the results are very sensitive to @xmath91 , which is a qcd parameter rather than a chiral parameter .
is described in the quark model in terms of a color hyperfine splitting between the spin-1/2 octet and the spin-3/2 decuplet states . ]
meissner and steininger @xcite argue that the decuplet contributions should be included in the chiral momentum expansion , expand in powers of @xmath92 , and evaluate their couplings to take the resulting effects into account .
unfortunately , this argument requires that @xmath91 be large on the scale of the momenta in the loop integrals .
it is not , with @xmath93 for typical mass splittings @xmath94 mev , and the heavy decuplet argument fails . we conclude that the decuplet states must be regarded as light , on the same mass scale as the octet states , to obtain a consistent picture .
their contribution to the moments is then given by the expressions in eqs .
( [ loopsum]-[g ] ) .
the approximation in @xcite becomes useful only for excited baryon multiplets , and can presumably be used to estimate those contributions .
jenkins _ et al .
_ included the effects of the mass splitting in the @xmath0 terms , but took @xmath95 in the logarithmic contributions .
this introduces very large changes in the latter , but does not affect their final conclusions because the differences are largely compensated through the counterterms . using the value @xmath96 mev
, we obtain the results for the decuplet contributions shown in table iii .
the logarithmic contributions associated with the decuplet are large and tend to cancel the logarithmic corrections from the octet as suggested in @xcite , but the cancellation is far from complete and all the logarithmic terms must be included .
if we look at table iii from the point of view of a loop expansion , we see that the total contributions from the octet loops in figs . 1 and 2 are considerably smaller than the individual @xmath0 and @xmath1 terms because of the cancellations .
this suggests that a loop expansion may converge more rapidly than the chiral expansion . however , the two types of terms tend to add rather than cancel for the decuplet , and turn out to dominate the final loop corrections .
the counterterms are also quite significant .
we see no convincing evidence that the chiral expansion with only the nonanalytic terms retained is under control .
our attempts to determine whether or not the loop corrections improve the tree - level fit to the moments are also inconclusive .
we have refit the data omitting the counterterms , but with the loop corrections included and @xmath5 and @xmath6 readjusted .
the quality of the fits is essentially unchanged with respect to tree level if all loop contributions are included .
_ find that the fits can be improved by omitting the logarithmic corrections and readjusting @xmath34 and @xmath85 in addition to the somewhat small values quoted for case ( a ) , but this procedure is not justified by the relative sizes of the terms retained and omitted ( see table iii ) .
the foregoing results suggest that the loop expansion may give a more useful approach to the calculation of the baryon moments than the chiral and @xmath22 expansions .
in particular , the results obtained from the different diagrams in figs
. 1 and 2 are all of comparable size when combined with the relevant coupling factors and must be treated together to get meaningful results .
however , loop effects are largely wiped out in the fits to the data when all the counterterms are treated as free parameters . neither the chiral nor loop expansion is usefully predictive in this setting given the experimental and theoretical uncertainties .
it is not possible , for example , to distinguish between the standard chiral expansion and the alternative approach proposed by bos _ et al .
we conclude that it will be necessary to develop a dynamical theory in which the baryons are treated as composite and the counterterms become calculable before the moment problem can be regarded as solved .
the authors would like to thank profs .
elizabeth jenkins , roxanne springer , and ulf - g .
meissner for useful correspondence about their published results .
this work was supported in part by the u.s .
department of energy under contract no .
de - fg02 - 95er40896 , and in part by the university of wisconsin graduate school with funds granted by the wisconsin alumni research foundation .
s. coleman and s.l .
glashow , phys .
* 6 * , 423 ( 1961 ) .
d.g . caldi and h. pagels , phys .
d * 10 * , 3739 ( 1974 ) .
j. gasser , m. sainio , and a. svarc , nucl .
* b307 * , 779 ( 1988 ) .
a. krause , helv .
acta * 63 * , 3 ( 1990 ) .
e. jenkins , m. luke , a.v . manohar and m. savage , phys .
b * 302 * , 482 ( 1993 ) ; ( e ) _ ibid . _ * 388 * , 866 ( 1996 ) .
r. dashen , e. jenkins , and a.v .
manohar , phys .
d * 49 * , 4713 ( 1994 ) .
e. jenkins and a.v .
manohar , phys .
b * 335 * , 452 ( 1994 ) .
luty , j. march - russell , and m. white , phys .
d * 51 * , 2332 ( 1995 ) .
j. dai , r. dashen , e. jenkins , and a.v .
manohar , phys .
d * 53 * , 273 ( 1996 ) .
meiner , s. steininger , nucl .
* b499 * , 349 ( 1997 ) .
j.w . bos , d. chang , s.c .
lee , y.c .
lin and h.h .
shih , chin .
( taipei ) * 35 * , 150 ( 1997 ) .
h. georgi , phys .
b * 240 * , 447 ( 1990 ) ; e. jenkins and a.v .
manohar , phys .
b * 255 * , 558 ( 1991 ) ; ucsd / pth 91 - 30 . s. okubo , phys .
lett . * 4 * , 14 ( 1963 ) .
e. jenkins and a.v .
manohar , in _ proc .
workshop on effective field theories of the standard model _ , edited by u. meissner ( world scientific , singapore , 1992 ) ; m.n .
butler , m.j .
savage , and r.p .
springer , nucl .
phys . , 69 ( 1993 ) .
.table of the coefficients of the counterterm contributions to the baryon magnetic moments .
this corresponds to the matrix @xmath48 defined in the text .
the notation is as follows : @xmath97 , @xmath98 , eq .
( 1 ) , while @xmath25 are the parameters are defined in eq .
the @xmath2 moment is not included , but is given by the su(2 ) relation @xmath99 . [ cols="^,^,^,^,^,^,^,^",options="header " , ] cddddddd case & @xmath100 & @xmath101 & @xmath102 & @xmath103 & @xmath104 & @xmath105 & @xmath106 + ( a ) & 3.183 & 1.992 & 0.269 & @xmath1070.215 & 0.534 & 0.224 & @xmath1070.392 + ( b ) & 3.946 & 2.353 & @xmath1070.001 & @xmath1070.172 & 0.569 & 0.694 & @xmath1071.165 + ( c ) & 2.568 & 1.809 & 0.591 & @xmath1070.253 & 0.497 & @xmath1070.506 & 0.621 cdddddddd @xmath108 & @xmath5 , @xmath6 & @xmath109 & @xmath110 & ln @xmath111 & ln @xmath112 & loops & ct & @xmath108 + p & 3.668 & @xmath1072.227 & @xmath1070.162 & 2.148 & @xmath1071.271 & @xmath1071.512 & 0.636 & 2.793 + n & @xmath1072.631 & 0.832 & 0.573 & @xmath1071.491 & 1.258 & 1.172 & @xmath1070.454 & @xmath1071.913 + @xmath113 & 3.668 & @xmath1072.843 & @xmath1070.474 & 2.829 & @xmath1071.111 & @xmath1071.598 & 0.388 & 2.458 + @xmath114 & @xmath1071.037 & 0.611 & @xmath1070.349 & @xmath1070.804 & 0.031 & @xmath1070.511 & 0.388 & @xmath1071.160 + @xmath2 & 1.315 & @xmath1071.116 & @xmath1070.411 & 1.012 & @xmath1070.540 & @xmath1071.055 & 0.388 & 0.649 + @xmath73 & @xmath1071.315 & 1.116 & 0.411 & @xmath1071.211 & 0.614 & 0.931 & @xmath1070.228 & @xmath1070.613 + @xmath90 & @xmath1072.631 & 2.288 & 0.698 & @xmath1072.650 & 0.739 & 1.075 & 0.306 & @xmath1071.250 + @xmath115 & @xmath1071.037 & 1.339 & @xmath1070.286 & @xmath1071.324 & 0.057 & @xmath1070.215 & 0.601 & @xmath1070.651 + @xmath3 & 2.278 & @xmath1071.157 & @xmath1070.496 & 1.680 & @xmath1070.771 & @xmath1070.745 & 0.000 & 1.534 + | we reexamine critically the chiral expansion for the baryon magnetic moments including the contributions from loops which involve intermediate octet and decuplet baryons .
we find that , contrary to some claims , the nonanalytic loop contributions of orders @xmath0 and @xmath1 are of the same general size because of large coupling factors for the latter , and that the decuplet contributions are as large as the octet contributions and must be included in a consistent calculation .
there is no clear evidence of the convergence of the chiral series .
the adequacy of the theory will not be established until dynamical models are able to calculate the contributions from the counterterms that largely hide the loop effects in fits to the data . |
globular clusters have dynamical properties that distinguish them from other stellar systems and allow one to test dynamical models .
they can be very long - lived , with typical ages for galactic globular clusters of 7 - 13 gyr @xcite , yet they are dynamically active in that their dynamical relaxation times are significantly less than a hubble time @xcite . for the well - observed galactic globular clusters ,
the ratio of the age to the relaxation time ( at the half - light radius ) lies in the range 2 - 40 ( a median of 7 - 8 ) , representing clusters that display modest dynamical evolution to those that have undergone core collapse .
binary stars are expected to play a central role in the dynamical evolution of globular clusters through the process of binary burning . in this process
, the dynamical interactions of binaries with other stars or binaries both hardens the binary and adds kinetic energy to the interacting star ( other binary ) , thereby slowing the contraction of the cluster , especially in the core region .
therefore , there are predicted relationships between the globular cluster binary properties and its dynamical state
. some of these predictions can be tested by measuring the binary fraction of globular cluster stars between systems in different dynamical states and as a function of radius within individual clusters .
the process of measuring binary fractions with globular cluster color - magnitude diagrams relies on accurate magnitude determinations of @xmath1 main sequence stars in crowded environments .
this approach is very challenging for ground - based observers , but it became feasible with the high angular resolution capabilities of the hubble space telescope @xcite .
hst has been used to observe globular clusters with the wfpc2 and more recently with the wfc / acs as part of a treasury project .
the results of these efforts on the topic of binary fractions were recently completed using these samples @xcite , with a variety of interesting results .
the methods used by @xcite are published in @xcite ( henceforth , paper i ) .
the present paper analyzes the trends of binary fraction with cluster properties and with radius .
both methods and analysis are contained within the thorough analysis of @xcite , and while there are a many similarities in the approaches of both efforts , there are a few differences , both technically and in sample selection .
several of the results by @xcite and @xcite are confirmed by us , but there are some important differences that we highlight in this paper .
the sample , which was frozen in 2010 , derives from the galactic globular cluster list @xcite for which there were sufficiently long wfpc2 or acs observations ( e.g. , snapshot observations were excluded ) .
we avoided clusters with high extinctions ( e(b - v ) @xmath2 0.4 ) to avoid the challenges associated with the variations in the extinction across the globular cluster ( paper i ) .
@xcite included more globular clusters with high extinction , for which they developed an approach to correct for small - scale extinction issues .
they also included other data sets , some taken after our cutoff date , so our sample contains 35 clusters @xcite while theirs contains 59 .
the earlier survey by @xcite examines 13 lower surface brightness globular clusters .
briefly , the color magnitude method of determining the main sequence binary fraction relies on the fact that when a second star is within the resolution element , the combination departs from the main sequence along a well - defined locus in color and magnitude .
this approach necessitates that the measurement error is small compared to the deviation from the main sequence of a typical binary , such as one with a mass ratio of 1:2 .
practically , this is best accomplished with a magnitude error near 0.01 mag , which can be challenging in crowded fields , but this challenge has been met through the development of point source extraction software that includes detailed treatments of the point spread function ( psf ) in such situations @xcite .
we calculate three measures of the binary fraction , two of which are non - parametric .
one approach is only to count stars with a mass fraction greater than q = 0.5 and form a binary fraction , f@xmath3(high q ) .
another method is to fit a gaussian distribution of the star density perpendicular to the straightened main sequence ( as a function of color ) , mostly using the blue side .
the fit is extended to the red side and subtracted from the total distribution , having made corrections for contamination by false binaries and field stars .
the remaining stars are assumed to be binaries whose mass ratios extend below q = 0.5 , typically reaching 0.3 ; the binary fraction is given by f@xmath3(count ) .
the final method is parametric in that we produce a best - fit using a binary fraction with a power - law dependence on mass ( for q @xmath4 0.3 ) , along with the gaussian described above and the contamination corrections .
this resulting binary fraction is denoted as f@xmath3(fit ) .
contamination is a critical issue because the chance superposition of two unrelated stars is indistinguishable from a true binary .
this problem is solved by simulations that lead to estimates for the number of chance superpositions as a function of magnitude and position in the cluster .
another type of contamination comes from foreground and background stars .
this is not a serious problem except at low latitudes , where we correct for it through galactic models that predict the contributions from such stars @xcite .
however , when the field is dense with foreground and background stars , these corrections are not always of sufficient precision .
a correction that is too large will produce a negative star density in the region that we are counting binaries , leading to a non - physical negative binary fraction .
this contamination most affects the @xmath5 method and we have two globular clusters where this occurred .
multiple stellar populations have been discovered in an increasing number of globular clusters @xcite and it is suspected that two generations of stars exist in globular clusters , formed typically within @xmath6 years of each other , a period of time far less than the age of these systems .
the evidence for the two ( or more ) populations comes from color magnitude diagrams that use ultraviolet and optical colors , and with some spectral data @xcite .
for example , in ngc 6352 @xcite , the width in color of the main sequence below the turnoff is 0.116 mag in @xmath7 , which is wider than the color range used for binary fitting . if this were the case for the cmd constructed with f606w and f814w data , determining binary fractions would be particularly challenging , if not impossible .
however , the spread of the main sequence is significantly less when using @xmath8 , because metallicity variations shift the relevant part of the main sequence in a vector nearly aligned with the main sequence ( paper i ) .
therefore , multiple populations can have a nearly degenerate color sequence on the lower main sequence .
for ngc 6352 , the distribution of the stars about the modal color value is largely fit by us with a gaussian with @xmath9 mag , well less than the @xmath10 mag range used for binary determinations . for ngc
6753 @xcite , populations a and c , which differ mainly in their @xmath11 enhancement and less so in the fe abundance , are distinct in uv colors and are successfully modeled with stellar atmosphere and isochrone codes . however , when using @xmath8 , the isochrones of the two populations are degenerate , based upon the stellar atmosphere models ( their figure 11 ) .
we can quantify the spread a bit better for the @xmath8 colors in the lower main sequence ( @xmath12 in their figure 6 ) , based on the models used for ngc 7089 @xcite ; this is an unusual globular cluster not included in our study . for fixed helium , age , and iron abundance ( y = 0.248 , 12 gyr , [ fe / h ] = -1.0 ) ,
a change in the alpha abundances from [ @xmath11/fe ] = 0.2 to 0.4 leads to a color shift in the main sequence of 0.008 mag .
the large difference in the metallicity from [ fe / h ] = -1.7 ( y = 0.246 , 13 gyr , [ @xmath11/fe ] = 0.4 ) to a slightly younger but more metal rich population with [ fe / h ] = -1.0 ( y = 0.248 , 12 gyr , [ @xmath11/fe ] = 0.4 ) leads to a shift of 0.029 mag .
such large metallicity differences are uncommon and more typical case is ngc 6656 @xcite , where the authors consider two populations with [ fe / h ] = -1.82 and -1.68 .
this difference leads to a measurable broadening near the turnoff , but further down the main sequence , we calculate that the broadening is 0.008 mag for the lower main sequence in the @xmath8 colors .
populations with enhanced he abundances are considered for ngc 7089 by @xcite , which lead to similar shifts , although these populations do not seem to account for much of the population on their lower main sequence or they fit poorly near the turnoff region ( their figure 6 ) .
also , there is no direct ( spectroscopic ) proof that the he abundance is enhanced by the large amounts they considered ( y = 0.33 ) .
multiple stellar populations certainly can add to the dispersion of the lower main sequence , but not necessarily to a degree that invalidates the determination of binary fractions .
a measure of the contamination is the breadth of the gaussian used to fit the central part of the aligned cmd in color space . for cases where the gaussian fit has @xmath13 ,
the main sequence is sufficiently wide to invalidate the cmd approach to determining the binary fraction , as in the case of ngc 2808 @xcite ; other objects were excluded from this study at the outset . however , when the main sequence is narrow , multiple main sequences are sufficiently degenerate to allow an estimation of the binary fraction .
the determination of @xmath14 is probably the most reliable , as it does not depend on the main sequence having a symmetric structure . to estimate the binary fraction to smaller values of @xmath15 ( to about 0.3 )
requires a fitting procedure that makes use of the entire distribution , but depends upon the assumption that most of an individual color distribution can be represented by gaussians .
while such fits lead to acceptable values of @xmath16 , small intrinsic asymmetries 2 - 3@xmath17 from the center of the distribution can lead to overestimates of the binary fraction .
many of the fitted binary fractions , @xmath18 are consistent with an extension of @xmath19 to @xmath20 , yet it is difficult to predict a - priori which are reliable and which are not .
therefore , we advise investigators to use @xmath18 with caution , with @xmath19 being more reliable .
addition discussion on these topics is found in paper i. analysis of the whole field of view statistically gives the most reliable results , as it includes the most number of stars .
this allows greater precision when comparing the binary fractions from different techniques . however , there are shortcomings to this approach .
due to the variation in distances , the sampling of each cluster extends to different radii .
for example , the half light radius of ngc 104 is not contained within the acs field . this issue would not be relevant except that there is a gradient in the binary fraction as a function of radius .
we do not include the region near the ccd edge and between the gap of the two chips because the photometry is not sufficiently accurate . in table [
tab : whole ] , we list the analysis results using different methods for the whole field of view . in this table , we give the region size in terms of their half mass radii in column 2 . in column 3 , 4 , and 5 , we list the binary fractions with 1 @xmath17 errors for the high mass - ratio ( @xmath5 ) method , the star counting method , and the fitting method , respectively .
column 6 and 7 are the fitted parameters for the third method , where power @xmath21 is the power index of the power - law function for the binary mass - ratio distribution , and @xmath22 is the minimum binary mass ratio we can get .
column 8 gives the @xmath23 and degrees of freedom for the fitting method .
the last column gives the binary fraction quality flags .
they indicate the main error sources . only the results including one quality flag can be used , such as g ( good estimate ) , d ( dense core ) , and n ( small number of stars ) .
results including flag f ( field stars ) or e ( large photometric errors ) should be used with caution , as the contaminations are asymmetric and with large uncertainties ( some fractions become negative values due to the asymmetric distribution of field stars ) . from the table
, we can calculate the mean binary fractions including 25 reliable binary fractions from the three methods : 5.2% ( high q ) , 6.3% ( counting ) , and 7.3% ( fitting ) , respectively . in figure
[ fig : compall ] , we show the comparisons of binary fractions obtained through different methods .
blue filled circles are for binary fractions obtained by high q method compared to the counting method .
red open circles are for binary fractions obtained by the fitting method compared to the counting method .
the black solid line shows where the two methods give the same results . from the figure
, we can see that the binary fractions obtained through the counting and fitting method are consistent with each other , while they are usually larger than the values obtained through the high mass - ratio method .
this is because the latter method does not include binaries with small mass ratio that are closer to the main sequence , while the former methods can statistically recover part of the signals from the small mass - ratio binaries . for the fitting method in table [ tab : whole ]
, the fitted power @xmath21 has a mean value of @xmath24 ( from the 25 reliable clusters ) , suggesting that most binaries tend to have small mass ratios .
the fitted minimum binary fraction @xmath22 has a mean value of @xmath25 , which is smaller than the cut - off ratio in the high mass - ratio method ( @xmath26 ) , indicating that the fitting method can recover part of mass - ratio binaries that are hidden in the main sequence .
the comparison between the minimum whole - field binary fractions with other efforts is not as good as one would expect , given that they are using the same data sets .
the most direct comparison uses the binary fraction above a certain value of @xmath15 , such as 0.5 for @xcite and this work , which was chosen because the distance of these binaries from the main sequence ridge line is generally above the 3@xmath17 uncertainties of a star .
@xcite also presents a binary fraction above some minimum binary ration , their @xmath22 , but they do not give values for @xmath22 , which may differ between clusters ( their table 3 ) .
for the 13 globular clusters considered by sollima , terzan 7 and palomar 12 have binary fractions that are significantly higher than the other clusters , a result confirmed by milone and this work , although the values differ between the groups . for the remaining 11 clusters ,
we do not find a meaningful relationship between the binary fractions from sollima and milone ( figure 2 ) .
the values from sollima are systematically larger than those of milone , but that may be due to a difference in how their binary fractions are calculated .
there is the same offset and lack of a correlation when comparing the sollima values to our values ( figure 2 ) .
when comparing our values with those of milone , there is no systematic offset and most of the values are consistent with each other , but there are a few significant discrepancies , which we discuss further below ( figure 3 ) .
although these various binary fractions are determined from the same data sets , the number of stars used is somewhat different and the photometric software is different .
however , each of the photometric methods has a heritage and is well - tested , although the method used by milone @xcite appears to be more efficient and more magnitudes are extracted for the same fields in approximately the same magnitude ranges .
one might expect that the different photometric methods would lead to different numbers of stars extracted but would not lead to systematically incorrect magnitudes on a scale greater than 3@xmath17 .
furthermore , the various groups performed simulations to demonstrate the reliability of recovering artificial stars that were placed into the very fields being analyzed . for most of these clusters ,
the other corrections are too small to account for the differences , such as overlapping stars and field star contamination .
the reasons for these differences remain unclear and a cross - method investigation is beyond the scope of this effort .
the analysis for the binary fractions within the half - mass radius has physical importance in that it provides a uniform basis to make comparisons between clusters and with theoretical models .
one difference with the fitting method is that we fixed the values of the power index @xmath21 and the minimum mass - ratio @xmath22 to the ones obtained by whole field analysis , because there are more stars to constrain the parameters using the whole chips .
the binary fractions are listed in table [ tab : rh ] , which is similar to table [ tab : whole ] , except that we omit the size column .
we also exclude the columns of power index @xmath21 and the minimum mass - ratio @xmath22 , as they are all fixed to the values in table [ tab : whole ] .
the average binary fractions within the half - mass radius from this table is 5.6% ( for the @xmath5 binary fractions ) , which includes 27 reliable clusters with flags of d , n , or g. there is a very tight relationship between the fitting and the counting method , so either are equivalently good for further analysis ) .
however , there is a poorer relationship between the high q binary fraction and the one from the counting method ( or from the fitting method ) .
this is also a poorer relationship than when using the whole fields .
based on the tests conducted in paper i , the weaker relationship can not be attributed to uncertainties in the methods unless there are mitigating factors , such as an incorrect estimate for the foreground / background stellar contamination , significant differential reddening , or multiple populations .
if these mitigating factors can be ignored , we would attribute the binary fraction difference to an increasing binary fraction in the range @xmath27 , which are included in the counting method but not in the high - q method . from the binary fractions that we calculated ,
we examine whether they are correlated with the physical properties of age , dynamical age ( age divided by the relaxation time at r@xmath28 ) , metallicity ( [ @xmath29 ) , and absolute magnitude m@xmath30 . to make the comparison , we assume that the binary fraction and above quantities are linearly related , as @xmath31 where y is the binary fraction and x is the different properties of globular clusters .
figure [ fig : rh_fb ] shows the relationships between the half - mass binary fractions and the ages , dynamical ages ( absolute ages divided by the relaxation time at half - mass radius ) , metallicity [ fe / h ] , and the absolute v magnitudes . the fitted line parameters and the non - parametric spearman rank coefficients are shown in table [ tab : fb_rh ] .
the strongest correlation is between the half - mass binary fractions and the cluster absolute v magnitudes , meaning that less luminous ( massive ) clusters have higher binary fractions .
this correlation is also confirmed by the spearman rank coefficient , with reliable significance .
this relationship was also discovered by @xcite . the lowest luminosity globular cluster , e3 , is not included in our work , but as evident in @xcite
, it supports the increase of binary fraction with decreasing luminosity .
the lower density end of this trend would correspond to the open clusters , where they usually have higher binary frequencies ( @xmath32 for hyades by @xcite , @xmath33 % for praesepe by @xcite , and 50% for m67 by @xcite ) .
a more recent study of 9 open clusters show 15%-54% of binary fraction @xcite . for field stars near the solar neighborhood
, the binary frequency can be as high as 50% @xcite . from this table
, we can see that the half - mass binary fractions appears to be anti - correlated with the cluster ages at above the 95% confidence level . however , two clusters have a significant effect on this result , terzan 7 and palomar 12 , which are both relatively young and have low absolute luminosities .
we return to this relationship when considering the core binary fractions , below .
we do not see a correlation of the half - mass radius binary fractions with either the dynamical time or the metallicity [ fe / h ] , which is the same result as that of @xcite , who used a larger sample but did the analysis with binary fractions within the core radius as well as between the core and the half light radius .
the binary fractions within the core regions are expected to respond most rapidly to dynamical interactions between stars . in our sample ,
not all the clusters are suitable for the core binary fraction analysis , because the core regions for some dense clusters are too crowded to recover stars or the core regions for some core - collapsed clusters are too small to include enough stars from which to measure a binary fraction . with such limitations ( and selection effects ) , we analyzed 25 clusters in our sample with the high mass - ratio method .
table [ tab : corefb ] shows the core binary fractions for 25 clusters with the high mass - ratio ( @xmath5 ) method .
the mean binary fractions within the core regions is about 7.0% ( excluding 8 clusters with field star contamination or due to large photometric errors and low star numbers ) .
the correlation of these binary fractions with those of milone is a good but not perfect correlation , with a few clusters in which our binary fractions are sytematically higher .
the relationship between the binary fraction and the physical properties of the clusters is similar to the results using the binary fractions within r@xmath28 , with one exception .
there is a strengthening of the inverse correlation between the binary fraction and the cluster age ( figure [ fig : corefb ] ) , and this relationship remains significant even when the two youngest clusters are excluded .
our ages @xmath34 , as reported in paper i , are mainly from @xcite , although a few additional ages were used from @xcite , @xcite , and @xcite .
the correlation coefficient of the correlation is -0.595 , a significance of 99.8% , and with the values @xmath35 and @xmath36 .
we can examine this relationship using a more uniform set of ages , where @xcite determined the absolute magnitudes of the main sequence turnoff , and using isochrones , obtained a relative age scale .
the relative age scale is normalized at the mean , so one can multiply it by the appropriate age ( e.g. , 12 gyr ) to obtain an absolute age .
the results are not strongly dependent on the isochrone library , so we use their d07@xmath37 values and find a similar anticorrelation ( figure [ fig : corefb2 ] ) with slope of @xmath38 ( for a mean age of 12 gyr , this would correspond to @xmath39 ) and intercept @xmath40 .
the spearman correlation indicates a significance at the 99.4% level , although if the points are normally distributed , we could use the pearson correlation test , which yields a higher levels of significance , @xmath41 . in this case , we included clusters for which the formal binary fraction is negative ( mainly systems with large numbers of foreground / background stars ) , except ngc 6656 , where the field stars are so dominant that the binary fraction is unreliable .
however , the result does not change significantly if those points are excluded ( or if ngc 6656 were included ) .
this result has implications that we comment upon below .
another important prediction of the models is that there will be a significant radial dependence on the binary fraction as well as on the stellar mass function .
this was explored by both sollima and milone , who both found radial gradients in the binary fraction .
we study the radial distribution of the binary fraction and also find evidence for a radial gradient .
in addition we investigate whether this is an intrinsic property or one that developed due to the dynamical interactions in the cluster .
in deriving the radial distribution of the binary fractions , we divide the cluster up into three radial bins of approximately equal numbers of stars , so that the uncertainties are comparable in the three bins .
the uncertainties are greater when compared to the whole field values , due to the fewer number of stars , and for the inner bin in clusters with dense central regions , the results have greater uncertainty . the methods that involve fitting , @xmath42 and @xmath18 have larger uncertainties than @xmath43 due to the additional components that must be constrained .
the results are given in table 4 .
due to the smaller uncertainties , we use @xmath43 to examine the radial binary dependence in figure [ fig : fb_r ] , we plot all the high mass - ratio ( @xmath5 ) binary fractions at different annular bins from all the clusters in our sample as a function of their radial distances from the cluster center .
the left panel is the high mass - ratio binary fractions normalized to their core fractions , and the distances are normalized to their core radii .
there are only 17 clusters with reliable core binary fractions measured in our sample . the straight line fitted results are shown in table [ tab : fb_r ] , first row , which shows a moderate correlation ( slope @xmath44 ) .
the spearman rank correlation coefficient , however , shows a significant correlation ( coefficient of @xmath45 , with a highly reliable significance ) .
the right panel in figure [ fig : fb_r ] and the second row in table [ tab : fb_r ] show the high mass - ratio binary fractions normalized to their half - mass fractions as a function of the distances normalized to their half - mass radii .
the straight line fit shows a strong correlation ( slope @xmath46 ) , with a spearman rank correlation coefficient of -0.61 , designating a highly significant correlation .
even though there can be significant differences between investigators for the value of a binary fraction for the same cluster , the radial distributions are very similar between investigators .
when a result is found , such as the radial decrease in the binary fraction , it could be due to evolutionary forces , but it could also be due to initial conditions the cluster was simply born with that property . to examine this issue
, we compare the six clusters where the ratio of the age to dynamical relaxation time ( at @xmath47 ) is less than four , ngc 104 , ngc 5053 , ngc 5272 ( m3 ) , ngc 5466 , ngc 5897 , and arp 2 .
if the radial decrease steepens with the number of elapsed relaxation times , these six clusters should show shallower radial binary fraction slopes .
of these six , only two show a radial decrease where the binary fraction drops by at least 20% . for the remaining 25 clusters for which we have good radial information
, only four fail to show a radial decrease of 20% or greater .
clusters with dynamical ages above four relaxation times show a radial decrease in 84% of the cases while those with younger ages show the decrease 33% of the time .
this implies that the decrease in the binary fraction with radius is due to evolution rather than birth .
the astrophysical question we sought to address is whether globular clusters follow the predictions of the dynamical models .
as a cluster progresses through several relaxation times , mass segregation should occur and soft binaries are rendered unbound .
mass segregation leads to binaries sinking deeper into the cluster , increasing the binary fraction , but strong dynamical interactions can destroy binaries , having the opposite effect . to understand which effect prevails
requires both theoretical and observational efforts , with the theoretical work indicating that the core binary fraction will rise with time prior to the core collapse period ( fregau 2009 ) .
another consideration in comparing theory to observation is that the initial properties of a cluster can mimic that expected from dynamical evolution .
for example , frank ( 2013 ) studied the globular cluster palomar 14 , which has a long half - mass relaxation time ( 20 gyr ) and lies in the outer - halo , where it does not feel strong tidal forces from the galaxy .
they find that palomar 14 exhibits significant mass segregation despite having an age less than even a single relaxation time .
the most likely conclusion is that it was born as a mass- segregated cluster , and if this formation history is common , it complicates the ability to compare data to models .
a final consideration is that the cluster - to - cluster dispersion may be larger than the effect one is trying to measure , so it is helpful to understand the range of initial conditions in a globular cluster .
a significant result is that the core binary fraction decreases inversely with the age of the globular clusters .
this was first found by @xcite also at a confidence level exceeding 99% ; both we and sollima used ages from @xcite .
however , when we used ages from @xcite , we obtain the same result , with only a slight decrease in significance .
when @xcite used the ages from @xcite , or from @xcite , they did not find a significant anticorrelation when using their entire sample .
restricted to just lower density clusters , they do not find an anticorrelation when using the relative ages from @xcite , but they appear to find an anticorrelation ( their figure b,10 ) when using the absolute ages from @xcite and @xcite . since the anticorrelation is with age and not dynamical age , it suggests that the younger clusters were born with a higher binary fraction . evidently , the conditions from which younger globular clusters formed were more conducive to binary formation . a examination of the binary fraction as a function of dynamical time can be used to obtain insight into the likely range of initial conditions . to examine this issue
, we use the core binary fraction from milone for @xmath48 in figure [ fig : binary_dyn_t ] , which leads to a more statistically significant result than using the binary fractions presented in this paper . at the shortest dynamical times , @xmath49 ,
the range in the binary fraction is about an order of magnitude .
this is probably representative of the range of initial binary fractions in clusters , although the absolute value of the binary fraction must have been larger for a few reasons .
binaries where one or both stars are remnants ( white dwarfs or neutron stars ) are recorded as single stars in the cmd method , while many of the soft binaries are destroyed even by the first relaxation time .
another impression one has from this figure is that the binary fraction lies within an envelope that is slowly decreasing with dynamical age , by about a factor of two over the range of the figure .
that is , above 10 relaxation times , there are no clusters with binary fractions above 6% and nearly half have binary fractions below 2% .
this is in contrast to the earlier times , where larger binary fractions are relatively more common
. it will be difficult to improve on this data set as the hst images are of excellent quality and the best globular clusters have been observed .
the radial decrease in the binary fraction is a robust result and one can account for an intrinsic dispersion in the binary fraction by dividing by the binary fraction either in the core or within half light radius .
this radial decrease is nearly always seen in clusters older than four dynamical times and seen less commonly for shorter dynamical times .
this implies that the radial binary fraction distribution is a result of relaxation and tidal effects rather than due to initial conditions .
the mass function for the binaries is still not very well known . for @xmath48
, one can simply count the stars in the color - magnitude diagram to obtain a binary fraction , but to extend this to lower mass ratios , a parametric fitting method is needed because the magnitude of the departure from the main sequence from photometry errors becomes comparable to that due to the pairing of two stars .
our fitting procedure finds that the number of binaries increases with decreasing mass ratio ( down to @xmath50 ) , although there is significant uncertainty in the value of the exponent .
this would be consistent with models in which binary stars are random pairs of stars , yet the work by milone indicates that the mass function is flat above @xmath51 .
we caution the reader that our fitting procedure depends upon an assumed gaussian symmetry in the width of the main sequence .
if this assumption is not generally valid , then our result of a rising binary mass distribution is thrown into question .
there are other avenues for studying binary populations in star clusters and one of the most effective is through spectroscopic programs .
a number of these clusters have stars bright enough for ground - based studies on large telescopes with multi - object capabilities .
sufficient spectroscopic monitoring can determine the period and velocity amplitude of the binary star system , which determines the mass function .
with the accurate photometric measurements from the hst data sets and a mapping between the absolute magnitude and stellar mass , the orbital parameters can be deduced .
such data provide information on the hardness of the binaries , essential data not provided by the cmd method used here .
spectroscopy will help in another important area in which a star has a degenerate partner , which will lead to a periodic shift in the stellar absorption lines .
this will help to give a fuller picture for the total number of stars in binaries , and these are probably binaries that formed relatively early . to help to facilitate such spectroscopic programs
, we compiled a list of 6421 binaries with @xmath48 , with ra , dec , magnitudes , and distance from the center of the globular cluster ( ji 2011 ; also available from the authors upon request ) .
the photometry for all stars in the fields is available from the acs globular cluster treasury program , with links from the mast website ( their magnitudes are slightly different from ours ) .
the authors would like to thank a.e .
dolphin for answering our many questions that arose when using the photometry package dolphot . for their many questions and suggestions
, we would like to thank from mario mateo , jon miller , eric bell , sally oey , patrick seitzer , and a very helpful referee .
we gratefully acknowledge financial support through hst programs 10939 and 11125 from nasa .
fan , x. , burstein , d. , chen , j. s. , zhu , j. , jiang , z. , wu , h. , yan , h. , zheng , z. , zhou , x. , fang , l. z. , chen , f. , deng , z. , chu , y. , hester , j. j. , windhorst , r. a. , li , y. , lu , p. , sun , w. h. , chen , w. p. , tsay , w. s. , chiueh , t. h. , chou , c. k. , ko , c. m. , lin , t. c. , guo , h. j. , & byun , y. i. 1996 , , 112 , 628 piotto , g. , milone , a. p. , anderson , j. , et al .
2012 , , 760 , 39 piotto , g. , bedin , l. r. , anderson , j. , king , i. r. , cassisi , s. , milone , a. p. , villanova , s. , pietrinferni , a. , & renzini , a. 2007 , apjl , 661 , 53 piotto , g. , milone , a. p. , bedin , l. r. , et al .
2015 , , 149 , 91 ccrrrrcrc ngc104 & 0.75 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 204.1/162 & d + ngc288 & 1.06 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 120.1/106 & g + ngc362 & 2.89 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 176.9/142 & d + ngc1851 & 4.64 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 126.4/136 & d + ngc2808 & 2.96 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 299.5/164 & d , p , e + ngc4590 & 1.57 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 141.6/ 96 & g + ngc5053 & 0.91 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 43.9/ 70 & g + m3 & 1.02 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 200.3/161 & d + ngc5466 & 1.03 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 34.6/ 63 & g + ngc5897 & 1.15 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 100.4/ 98 & g + ngc5904 & 1.34 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 113.6/122 & d + ngc5927 & 2.15 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 157.3/148 & f + ngc6093 & 3.88 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 207.6/125 & d + ngc6121 & 0.55 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 66.0/ 62 & f + ngc6101 & 2.25 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 105.8/100 & g + m13 & 1.40 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 147.1/145 & d + ngc6218 & 1.34 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 92.6/ 91 & g + ngc6341 & 2.32 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 247.0/146 & d + ngc6352 & 1.15 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 110.4/ 99 & f + ngc6362 & 1.15 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 108.2/ 95 & g + ngc6397 & 0.82 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 103.1/ 79 & g + ngc6541 & 2.23 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 227.7/140 & d , f + ngc6624 & 2.89 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 200.2/144 & d , f + ngc6637 & 2.82 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 142.3/135 & d , f + ngc6652 & 4.93 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 92.7/ 85 & d , f + ngc6656 & 0.70 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 488.0/124 & f + ngc6723 & 1.55 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 113.2/118 & g + ngc6752 & 1.24 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 124.2/ 98 & g + terzan7 & 3.07 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 40.5/ 63 & g + arp2 & 1.34 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 57.9/ 74 & g + ngc6809 & 0.84 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 64.8/ 77 & g + ngc6981 & 2.54 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 85.3/102 & d + ngc7078 & 2.37 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 229.2/160 & d , e + ngc7099 & 2.30 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 85.7/ 68 & g + palomar12 & 1.38 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 13.1/ 33 & n + crrrcc ngc104 & 3.03@xmath52 0.13 & [email protected] & [email protected] & 205.6/162 & d + ngc288 & 6.29@xmath52 0.32 & [email protected] & [email protected] & 96.9/ 99 & g + ngc362 & 5.69@xmath52 0.36 & [email protected] & [email protected] & 122.8/118 & d + ngc1851 & 3.89@xmath52 0.85 & [email protected] & [email protected] & 40.3/ 73 & d , e + ngc2808 & -12.09@xmath52 0.96 & [email protected] & [email protected] & 71.9/ 90 & d , e , p + ngc4590 & 6.24@xmath52 0.30 & [email protected] & [email protected] & 112.3/ 87 & g + ngc5053 & 5.57@xmath52 0.40 & [email protected] & [email protected] & 42.5/ 70 & g + m3 & 5.03@xmath52 0.17 & [email protected] & [email protected] & 197.4/161 & d + ngc5466 & 5.23@xmath52 0.35 & [email protected] & [email protected] & 33.4/ 63 & g + ngc5897 & 5.63@xmath52 0.28 & [email protected] & [email protected] & 82.5/ 96 & g + ngc5904 & 3.15@xmath52 0.15 & [email protected] & [email protected] & 111.6/122 & d + ngc5927 & 3.87@xmath52 0.28 & [email protected] & [email protected] & 107.8/127 & f + ngc6093 & 7.59@xmath52 0.58 & [email protected] & [email protected] & 80.8/ 98 & d + ngc6121 & 4.79@xmath52 0.48 & [email protected] & [email protected] & 46.7/ 40 & g + ngc6101 & 5.50@xmath52 0.38 & [email protected] & [email protected] & 105.7/100 & g + m13 & 3.46@xmath52 0.16 & [email protected] & [email protected] & 149.2/144 & d + ngc6218 & 3.21@xmath52 0.24 & [email protected] & [email protected] & 93.3/ 88 &
g + ngc6341 & 5.31@xmath52 0.29 & [email protected] & [email protected] & 151.0/136 & d + ngc6352 & -0.57@xmath52 0.88 & [email protected] & [email protected] & 104.7/ 98 & f + ngc6362 & 4.31@xmath52 0.29 & [email protected] & [email protected] & 99.9/ 94 & g + ngc6397 & 4.47@xmath52 0.49 & [email protected] & [email protected] & 100.7/ 79 & g + ngc6541 & 3.94@xmath52 0.28 & [email protected] & [email protected] & 184.6/123 & d + ngc6624 & 1.31@xmath52 0.57 & [email protected] & [email protected] & 63.7/ 93 & f , d + ngc6637 & 4.16@xmath52 0.40 & [email protected] & [email protected] & 102.5/103 & f , d + ngc6652 & 3.17@xmath52 0.78 & [email protected] & [email protected] & 42.6/ 41 & f , d + ngc6656 & -4.94@xmath52 0.28 & [email protected] & [email protected] & 486.9/124 & f + ngc6723 & 4.66@xmath52 0.21 & [email protected] & [email protected] & 101.3/115 & g + ngc6752 & 0.74@xmath52 0.16 & [email protected] & [email protected] & 118.3/ 96 & g + terzan7 & 17.15@xmath52 1.40 & [email protected] & [email protected] & 43.9/ 62 & g + arp2 & 8.05@xmath52 0.55 & [email protected] & [email protected] & 30.9/ 50 & g + ngc6809 & 3.31@xmath52 0.22 & [email protected] & [email protected] & 33.4/ 43 &
g + ngc6981 & 7.18@xmath52 0.38 & [email protected] & [email protected] & 85.7/102 & d + ngc7078 & 5.85@xmath52 0.32 & [email protected] & [email protected] & 231.0/160 & d + ngc7099 & 3.78@xmath52 0.44 & [email protected] & [email protected] & 103.6/ 62 & d + palomar12 & 13.36@xmath52 1.30 & [email protected] & [email protected] & 7.2/20.0 & g + [ tab : rh ] cccccc age & [email protected]&[email protected] & 1055.2/ 25&-0.392&0.043 + dynamical age & [email protected] & [email protected] & 1042.0/ 25&-0.068&0.737 + @xmath53$]&[email protected] & [email protected] & 1054.4/ 25&-0.016&0.936 + mv & [email protected] & [email protected] & 1017.0/ 25 & 0.490&0.010 + [ tab : fb_rh ] crrrrrc ngc104 & 0.00 - 0.29 & [email protected] & [email protected] & [email protected] & 144.7/130 & d + & 0.29 - 0.42 & [email protected] & [email protected] & [email protected] & 75.7/110 & d + & 0.42 - 0.56 & [email protected] & [email protected] & [email protected] & 104.7/100 & d + ngc288 & 0.00 - 0.43 & [email protected] & [email protected] & [email protected] & 64.0/ 63 & g + & 0.43 - 0.70 & [email protected] & [email protected] & [email protected] & 77.0/ 71 & g + & 0.70 - 1.05 & [email protected] & [email protected] & [email protected] & 38.8/ 67 & g + ngc362 & 0.00 - 0.97 & [email protected] & [email protected] & [email protected] & 127.7/118 & d + & 0.97 - 1.48 & [email protected] & [email protected] & [email protected] & 91.9/101 & d + & 1.48 - 2.16 & [email protected] & [email protected] & [email protected] & 89.0/ 87 & g + ngc1851 & 0.00 - 1.55 & [email protected] & [email protected] & [email protected] & 91.5/113 & d + & 1.55 - 2.34 & [email protected] & [email protected] & [email protected] & 76.0/ 87 & d + & 2.34 - 3.55 & [email protected] & [email protected] & [email protected] & 79.1/ 76 & d + ngc2808 & 0.00 - 1.35 & [email protected] & [email protected] & [email protected] & 186.9/141 & d , e , p + & 1.35 - 1.73 & [email protected] & [email protected] & [email protected] & 180.3/132 & d , e , p + & 1.73 - 2.26 & [email protected] & [email protected] & [email protected] & 204.4/130 & d , e , p + ngc4590 & 0.00 - 0.44 & [email protected] & [email protected] & [email protected] & 47.3/ 64 & g + & 0.44 - 0.74 & [email protected] & [email protected] & [email protected] & 38.4/ 55 & g + & 0.74 - 1.14 & [email protected] & [email protected] & [email protected] & 74.5/ 60 & g + ngc5053 & 0.00 - 0.36 & [email protected] & [email protected] & [email protected] & 22.2/ 43 & g + & 0.36 - 0.52 & [email protected] & [email protected] & [email protected] & 26.6/ 40 & g + & 0.52 - 0.68 & [email protected] & [email protected] & [email protected] & 25.1/ 41 & g + m3 & 0.00 - 0.35 & [email protected] & [email protected] & [email protected] & 112.2/135 & d + & 0.35 - 0.54 & [email protected] & [email protected] & [email protected] & 82.8/111 & d + & 0.54 - 0.76 & [email protected] & [email protected] & [email protected] & 128.6/102 & d + ngc5466 & 0.00 - 0.39 & [email protected] & [email protected] & [email protected] & 31.8/ 39 & g + & 0.39 - 0.58 & [email protected] & [email protected] & [email protected] & 24.6/ 38 & g + & 0.58 - 0.79 & [email protected] & [email protected] & [email protected] & 34.4/ 39 & g + ngc5897 & 0.00 - 0.44 & [email protected] & [email protected] & [email protected] & 41.0/ 63 & g + & 0.44 - 0.64 & [email protected] & [email protected] & [email protected] & 29.6/ 59 & g + & 0.64 - 0.85 & [email protected] & [email protected] & [email protected] & 32.6/ 61 & g + ngc5904 & 0.00 - 0.43 & [email protected] & [email protected] & [email protected] & 75.4/103 & d + & 0.43 - 0.68 & [email protected] & [email protected] & [email protected] & 75.9/ 78 & g + & 0.68 - 0.99 & [email protected] & [email protected] & [email protected] & 52.3/ 67 & g + ngc5927 & 0.00 - 0.75 & [email protected] & [email protected] & [email protected] & 117.0/112 & df + & 0.75 - 1.15 & [email protected] & [email protected] & [email protected] & 87.3/ 96 & f + & 1.15 - 1.64 & [email protected] & [email protected] & [email protected] & 122.6/ 94 & f + ngc6093 & 0.00 - 1.17 & [email protected] & [email protected] & [email protected] & 100.7/108 & d + & 1.17 - 1.82 & [email protected] & [email protected] & [email protected] & 89.0/ 85 & d + & 1.82 - 2.96 & [email protected] & [email protected] & [email protected] & 47.2/ 73 & g + ngc6101 & 0.00 - 0.84 & [email protected] & [email protected] & [email protected] & 73.0/ 69 & g + & 0.84 - 1.25 & [email protected] & [email protected] & [email protected] & 46.1/ 64 & g + & 1.25 - 1.72 & [email protected] & [email protected] & [email protected] & 46.2/ 61 & g + ngc6121 & 0.00 - 0.20 & [email protected] & [email protected] & [email protected] & 53.9/ 33 & n + & 0.20 - 0.30 & [email protected] & [email protected] & [email protected] & 29.0/ 35 & n + & 0.30 - 0.41 & [email protected] & [email protected] & [email protected] & 16.4/ 38 & n + m13 & 0.00 - 0.50 & [email protected] & [email protected] & [email protected] & 113.3/120 & d + & 0.50 - 0.75 & [email protected] & [email protected] & [email protected] & 84.0/102 & d + & 0.75 - 1.04 & [email protected] & [email protected] & [email protected] & 86.8/ 90 & g + ngc6218 & 0.00 - 0.47 & [email protected] & [email protected] & [email protected] & 51.6/ 59 & g + & 0.47 - 0.72 & [email protected] & [email protected] & [email protected] & 39.7/ 55 & g + & 0.72 - 1.00 & [email protected] & [email protected] & [email protected] & 54.4/ 57 & g + ngc6341 & 0.00 - 0.75 & [email protected] & [email protected] & [email protected] & 113.5/121 & d + & 0.75 - 1.18 & [email protected] & [email protected] & [email protected] & 76.3/102 & d + & 1.18 - 1.74 & [email protected] & [email protected] & [email protected] & 104.0/ 87 & g + ngc6352 & 0.00 - 0.43 & [email protected] & [email protected] & [email protected] & 36.8/ 57 & f + & 0.43 - 0.64 & [email protected] & [email protected] & [email protected] & 36.9/ 54 & f + & 0.64 - 0.86 & [email protected] & [email protected] & [email protected] & 34.6/ 55 & f + ngc6362 & 0.00 - 0.44 & [email protected] & [email protected] & [email protected] & 58.4/ 60 & g + & 0.44 - 0.65 & [email protected] & [email protected] & [email protected] & 28.4/ 56 & g + & 0.65 - 0.85 & [email protected] & [email protected] & [email protected] & 61.7/ 56 & g + ngc6397 & 0.00 - 0.28 & [email protected] & [email protected] & [email protected] & 36.4/ 46 & g + & 0.28 - 0.45 & [email protected] & [email protected] & [email protected] & 33.6/ 45 & g + & 0.45 - 0.61 & [email protected] & [email protected] & [email protected] & 43.3/ 49 & g + ngc6541 & 0.00 - 0.71 & [email protected] & [email protected] & [email protected] & 97.4/111 & d + & 0.71 - 1.14 & [email protected] & [email protected] & [email protected] & 96.3/ 93 & d + & 1.14 - 1.66 & [email protected] & [email protected] & [email protected] & 177.0/ 87 & g + ngc6624 & 0.00 - 0.91 & [email protected] & [email protected] & [email protected] & 64.2/ 86 & f + & 0.91 - 1.45 & [email protected] & [email protected] & [email protected] & 61.0/ 84 & f + & 1.45 - 2.13 & [email protected] & [email protected] & [email protected] & 73.2/ 90 & f + ngc6637 & 0.00 - 0.87 & [email protected] & [email protected] & [email protected] & 102.9/ 96 & f + & 0.87 - 1.38 & [email protected] & [email protected] & [email protected] & 67.2/ 82 & f + & 1.38 - 2.15 & [email protected] & [email protected] & [email protected] & 56.1/ 76 & f + ngc6652 & 0.00 - 1.34 & [email protected] & [email protected] & [email protected] & 34.5/ 50 & f + & 1.34 - 2.34 & [email protected] & [email protected] & [email protected] & 28.7/ 48 & f + & 2.34 - 3.77 & [email protected] & [email protected] & [email protected] & 36.2/ 47 & f + ngc6656 & 0.00 - 0.26 & [email protected] & [email protected] & [email protected] & 86.4/ 84 & f + & 0.26 - 0.39 & [email protected] & [email protected] & [email protected] & 126.0/ 82 & f + & 0.39 - 0.52 & [email protected] & [email protected] & [email protected] & 168.6/ 80 & f + ngc6723 & 0.00 - 0.50 & [email protected] & [email protected] & [email protected] & 93.7/ 92 & d + & 0.50 - 0.80 & [email protected] & [email protected] & [email protected] & 115.5/ 79 & g + & 0.80 - 1.16 & [email protected] & [email protected] & [email protected] & 132.6/ 72 & g + ngc6752 & 0.00 - 0.41 & [email protected] & [email protected] & [email protected] & 77.4/ 75 & g + & 0.41 - 0.66 & [email protected] & [email protected] & [email protected] & 70.2/ 65 & g + & 0.66 - 0.93 & [email protected] & [email protected] & [email protected] & 64.5/ 64 & g + ngc6809 & 0.00 - 0.32 & [email protected] & [email protected] & [email protected] & 39.3/ 49 & g + & 0.32 - 0.47 & [email protected] & [email protected] & [email protected] & 35.7/ 48 & g + & 0.47 - 0.60 & [email protected] & [email protected] & [email protected] & 44.3/ 48 & g + ngc6981 & 0.00 - 0.67 & [email protected] & [email protected] & [email protected] & 54.4/ 75 & d + & 0.67 - 1.12 & [email protected] & [email protected] & [email protected] & 41.2/ 66 & g + & 1.12 - 1.89 & [email protected] & [email protected] & [email protected] & 53.7/ 58 & g + ngc7078 & 0.00 - 0.85 & [email protected] & [email protected] & [email protected] & 163.3/140 & d , e + & 0.85 - 1.25 & [email protected] & [email protected] & [email protected] & 136.9/122 & d , e + & 1.25 - 1.78 & [email protected] & [email protected] & [email protected] & 252.2/107 & d , e + ngc7099 & 0.00 - 0.64 & [email protected] & [email protected] & [email protected] & 74.6/ 52 & g + & 0.64 - 1.11 & [email protected] & [email protected] & [email protected] & 43.0/ 42 & g + & 1.11 - 1.74 & [email protected] & [email protected] & [email protected] & 36.0/ 37 & g + arp2 & 0.00 - 0.49 & [email protected] & [email protected] & [email protected] & 13.0/ 28 & g + & 0.49 - 0.72 & [email protected] & [email protected] & [email protected] & 11.6/ 29 & g + & 0.72 - 1.02 & [email protected] & [email protected] & [email protected] & 9.1/ 27 & g + palomar12 & 0.00 - 0.40 & [email protected] & [email protected] & [email protected] & 2.2/ 13 & n + & 0.40 - 0.71 & [email protected] & [email protected] & [email protected] & 1.5/ 13 & n + & 0.71 - 1.05 & [email protected] & [email protected] & [email protected] & 3.6/ 13 & n + terzan7 & 0.00 - 0.86 & [email protected] & [email protected] & [email protected] & 13.0/ 28 & n + & 0.86 - 1.48 & [email protected] & [email protected] & [email protected] & 11.6/ 29 & n + & 1.48 - 2.35 & [email protected] & [email protected] & [email protected] & 9.1/ 27 & n + [ tab : radial2 ] ccc ngc104 & 4.26 @xmath52 2.68 & d , n + ngc288 & 5.64 @xmath52 0.41 & g + ngc4590 & 9.43 @xmath52 0.69 & g + ngc5053 & 5.67 @xmath52 0.40 & g + m3 & 5.42 @xmath52 1.07 & d + ngc5466 & 5.58 @xmath52 0.43 & g + ngc5897 & 5.08 @xmath52 0.34 & g + ngc5904 & 7.73 @xmath52 0.72 & d + ngc5927 & 5.41 @xmath52 0.82 & d , f + ngc6101 & 6.13 @xmath52 0.40 & g + ngc6121 & 4.94 @xmath52 0.63 & g + m13 & 6.17 @xmath52 0.47 & d + ngc6218 & 3.75 @xmath52 0.47 & g + ngc6341 & 1.20 @xmath52 2.47 & d , n + ngc6352 & 1.76 @xmath52 1.30 & f + ngc6362 & 4.81 @xmath52 0.44 & g + ngc6541 & -0.12 @xmath52 5.47 & d , n + ngc6637 & 8.51 @xmath52 1.27 & d , f + ngc6656 & -2.39
@xmath52 0.33&f + ngc6723 & 6.89 @xmath52 0.38 & g + ngc6752 & 5.27 @xmath52 2.47&d , n + ngc6809 & 3.43 @xmath52 0.24&g + ngc6981 & 11.56 @xmath52 0.81&g + arp2 & 7.42 @xmath52 0.69 & n + terzan7 & 20.02 @xmath52 2.10 & n + [ tab : corefb ] | binary stars are predicted to have an important role in the evolution of globular clusters , so we obtained binary fractions for 35 globular clusters that were imaged in the f606w and f814w with the acs on the hubble space telescope . when compared to the values of prior efforts @xcite , we find significant discrepancies , despite each group correcting for contamination effects and having performed the appropriate reliability tests .
the most reliable binary fractions are obtained when restricting the binary fraction to @xmath0 .
our analysis indicates that the range of the binary fractions is nearly an order of magnitude for the lowest dynamical ages , suggesting that there is a broad distribution in the binary fraction at globular cluster formation .
dynamical effects also appears to decrease the core binary fractions by a factor of two over a hubble time , but this is a weak relationship .
we confirm a correlation from previous work that the binary fraction within the core radius decreases with cluster age , indicating that younger clusters formed with higher binary fractions .
the strong radial gradient in the binary fraction with cluster radius appears to be a consequence of dynamical interactions .
it is often not present in dynamically young clusters but nearly always present in dynamically old clusters . |
although the presence of dark matter and dark energy is justified by all cosmological observations , their identification and properties are still far from being established .
the measurements of the fluctuations of the cmb are in this respect escpecially valuable because of the wealth and accurate information that can be extracted from them .
the most recent disclosed results of the planck mission contain issues like : the temperature power spectrum , gravitational lensing or integrated sachs - wolfe effect , up to sunyaev - zeldovich cluster counts , isotropy , nongaussianity or cosmic infrared background .
it seems that the old , unexpected features , beyond the @xmath0 + inflation model , persist in data and are even more highlighted : 1 .
large scale temperature power spectrum much lower than the @xmath0 prediction , not limited only to low quadrupole @xcite but to almost all multipole moments @xmath1 ( see fig .
37 in ref .
@xcite ) , 2 .
scalar power spectrum index less than 1 ( see table 8 in ref .
@xcite ) , 3 . violation of isotropy observed as hemispherical asymmetry , parity asymmetry , quadrupole - octopole alignment , cold spot , and dipolar asymmetry @xcite .
if the violation of isotropy will be confirmed by other complementary cosmic observations of radio galaxies @xcite , spiral galaxies @xcite , bulk flows of clusters @xcite or quasars @xcite , it will challenge cosmological principles and call for the new theoretical insights . assuming that the observed anomalies are real phenomena
, we try to understand and elucidate measured physical features by the einstein - cartan ( ec ) cosmology . incorporating rotating degrees of freedom of matter ( spin and angular momentum ) and spacetime ( torsion ) into the relativistic framework
, the ec cosmology appears as a nonsingular theory @xcite , the cosmic mass density can be fixed @xcite , the scalar power index can acquire the negative tilt @xcite , spin densities triger density fluctuations @xcite and the right - handed vorticity ( rotation ) of the universe @xcite resulting at later stages of the evolution in the nonvanishing total angular momentum of the universe @xcite .
the nonsingular ec cosmology is in conformity with the nonsingular theory of gauge interactions in particle physics @xcite that contains light and heavy majorana neutrinos as hot and cold dark matter particles @xcite , including other important implications of the perturbative and nonperturbative aspects of strong and electroweak interactions phenomenology @xcite . in this paper
we investigate and compare ec and @xmath0 cosmologies solving evolution equations for the scale dependent density contrasts , mass fluctuations , peculiar velocities , and integrated sachs - wolfe effect .
the next chapter contains description of the evolution equations , definitions and the introduction of our simple clustering model .
the concluding section deals with the numerical results of the computations , comparisons of the ec , and @xmath0 cosmologies and final remarks and hints for future research .
because any deviation from cosmic homogeneity and isotropy is very small , we limit our considerations to the homogeneous and isotropic geometry .
we start the evolution in the radiation era when clustering of dark and baryonic matter is negligible .
the evolution equations for matter density contrasts in fourier space are derived in @xcite @xmath2 we here use the notation @xmath3 $ ] , density contrasts are @xmath4 , k is the comoving wave number , @xmath5 , subscripts m , r,@xmath6 denote matter , radiation and the cosmological constant quantities , respectively , and v is a velocity .
all the quantites are functions of t and @xmath7
. these equations can be cast into the more suitable form by getting rid of @xmath8 and by changing the evolution variable to @xmath9 : @xmath10 our notation includes @xmath11 , @xmath12 , @xmath13 , @xmath14 , @xmath15 and @xmath16^{1/2 } s^{-1}$ ] .
the evolution equations for the ec cosmology ( neglecting small vorticity and acceleration @xmath17 ) are derived in @xcite ( eq . ( 14 ) ) : @xmath18^{1/2}. \nonumber\end{aligned}\ ] ] we assume that after redshift @xmath19 the nonlinear bound structures are formed in the form of stars , galaxies , and clusters .
the clustering of particles forming halos are described by a model with only two parameters @xmath20 and @xmath21 .
this is applied to both mass and angular momentum clustering : @xmath22 fourier transforms of the evolution equations ( 3 ) take now the following form with the evolution variable @xmath9 : @xmath23 the einstein - cartan field equations define the cosmic clocks ( see eq .
( 15 ) of @xcite ) as follows : @xmath24^{1/2 } , \nonumber \\ \frac{d\dot{a}}{d a}=h_{0}\frac{1}{2}[\omega_{m}a^{-1}+\omega_{\lambda}a^{2}-\frac{1}{3}a^{2}q^{2}]^{-1/2 }
[ -\omega_{m}a^{-2}+2 \omega_{\lambda}a+\frac{1}{3}aq^{2 } ] , \nonumber \\
\frac{\ddot{a}}{a}=\frac{1}{3}\kappa \lambda -\frac{1}{6}\kappa \rho + \frac{2}{3}q^{2}.\end{aligned}\ ] ] the following definitions and convolutions are used in eq .
( 6 ) : @xmath25 having all the evolution equations for the ec and @xmath0 cosmologies , we define initial conditions in the radiation era and choose the parameters of the models : @xmath26 our choice of the torsion ( angular momentum ) parameters is guided by the condition that at zero redshift @xmath27 @xcite ( at the redshifts @xmath28 the galaxy clusters emerge changing the total angular momentum contribution of the era @xmath29 ) , while at the earlier epoch @xmath30 our choice is guided by the condition to match roughly the correct cosmic clocks and the age of the universe : @xmath31^{-1/2 } , \\
\tau_{u}(\lambda cdm)=13.89 gyr,\ \tau_{u}(ec)=13.29 gyr , \\
k_{min}=10^{-3 } mpc^{-1},\ k_{max}=10^{2 } mpc^{-1 } , \ k_{g}=1 mpc^{-1},\ \sigma_{g}=0.25 mpc^{-1}.\end{aligned}\ ] ] we integrate the above evolution equations to the relative accuracy @xmath32 by lowering the integration steps until the required accuracy is reached .
equations ( 2 ) are solved for the evolution from @xmath33 up to @xmath34 and eqs .
( 6 ) are then solved from @xmath34 up to @xmath35 .
the adams - bashforth - moulton predictor - corrector method is used for differential equation integrations ( code of l. f. shampine and m. k. gordon , sandia laboratories , albuquerque , new mexico ) and cuba library for multidimensional integrations @xcite .
the next section is devoted to the detailed exposure of the numerical results and comparison between ec and @xmath0 models .
the relevance of the results for the planck data are given here as well .
because the best fit to the planck temperature power spectrum is dominantly performed by the multipoles @xmath36 , it is not a surprise that the concordance @xmath0 model is favoured , but at the expense of the wrong fit for low mutipoles ( large scales ) . by solving evolution equations for the ec and @xmath0 with the simple clustering model
, one can verify that at low redshifts these two models produce density contrasts that differ substantially at large scales , while being similar at smaller scales ( see fig .
1 ) . the conclusions are not sensitive to the reasonable choices of the parameters of the clustering model .
the integrated sachs - wolfe ( isw ) effect plays an important role at low redshifts in the evolution , if the mass density differs from one @xcite : the isw is positive ( negative ) for the @xmath0 ( ec ) cosmology ( see fig .
the structure of the ec isw curve around z=1 is just an artefact of our simple model for torsion with a nonanalytic behaviour at z=1 .
the compendium of all our results can be summarized as follows : ( 1 ) the @xmath0 model can not simultaneously fit the large and the small scale parts of the planck tt spectrum , while the ec can rectify this deficiency owing to the presence of the new rotational degrees of freedom ( torsion ) that partially cancels the large mass density ( @xmath45 ) when clustering matters , i.e. the rotation ( centripetal force ) acts opposite to the attractive force of gravity , ( 2 ) the presence of the isw effect is observed in planck data @xcite , but with the unknown sign ; the very small low multipoles of the planck tt spectrum imply the negative contribution of the isw @xcite , thus in accord with the ec model , ( 3 ) the peculiar velocities are larger at the galaxy and galaxy cluster scales for the ec than @xmath0 cosmologies at low redshifts .
these conclusions are robust and are qualitatively valid for the reasonable variation of the clustering model parameters @xmath20 and @xmath21 .
the two different analyses of the planck peculiar velocities of galaxy clusters @xcite are still not conclusive as to whether the data are consistent with the @xmath0 model or not .
our final remark is that the first planck results favour a description of the universe with the anisotropic models .
the @xmath0 and the inflationary paradigm can not fulfil severe phenomenological requirements . we show that the ec cosmology with the new rotational degrees of freedom can resolve almost all of the @xmath0 model deficiencies . however , the n - body numerical simulations , including the angular momenta of the cdm halos and their feedback onto the background cosmic geometry , have to be applied within the ec gravity . 100 c. l. bennett et al . , astrophys
. j. suppl .
ser . * 148 * , 1 ( 2003 ) .
planck collab .
, arxiv:1303.5075 .
planck collab .
, arxiv:1303.5083 .
p. birch , nature , * 298 * , 451 ( 1982 ) .
m. j. longo , phys .
b * 699 * , 224 ( 2011 ) .
a. kashlinsky , f. atrio - barandela , d. kocevski and h. ebeling , astrophys .
* 686 * , l49 ( 2008 ) .
d. hutsem@xmath46kers , astron . and astrophys .
* 332 * , 410 ( 1998 ) .
a. trautman , nature * 242 * , 7 ( 1973 ) .
d. palle , nuovo cim .
b * 111 * , 671 ( 1996 ) .
d. palle , nuovo cim .
b * 122 * , 67 ( 2007 ) .
d. palle , nuovo cim .
b * 114 * , 853 ( 1999 ) .
d. palle , entropy * 14 * , 958 ( 2012 )
. d. palle , eur .
j. c * 69 * , 581 ( 2010 ) .
d. palle , nuovo cim . a * 109 * , 1535 ( 1996 ) .
d. palle , nuovo cim .
b * 115 * , 445 ( 2000 ) ; ibidem * 118 * , 747 ( 2003 ) .
d. palle , hadronic j. * 24 * , 87 ( 2001 ) ; ibidem * 24 * , 469 ( 2001 ) ; d. palle , acta phys .
b * 43 * , 1723 ( 2012 ) ; ibidem *
43 * , 2055 ( 2012 ) ; d. palle , arxiv:1210.4404 .
p. j. e. peebles , _ the large - scale structure of the universe _
( princeton university press , new jersey , 1980 ) ; p. j. e. peebles , astrophys .
j. * 277 * , 470 ( 1984 ) .
t. hahn , comp .
. commun . * 168 * , 78 ( 2005 ) .
planck collab .
, arxiv:1303.5079 .
h. kodama and m. sasaki , prog .
. suppl . * 78 * , 1 ( 1984 ) .
planck collab . , arxiv:1303.5090 ; f. atrio - barandela , arxiv:1303.6614 ( to appear in astron . and astrophys . ) . | the first comprehensive analyses of planck data reveal that the cosmological model with dark energy and cold dark matter can satisfactorily explain the essential physical features of the expanding universe . however , the inability to simultaneously fit large and small scale tt power spectrum , scalar power index smaller than one and the observations of the violation of the isotropy found by few statistical indicators of the cmb , urge theorists to search for explanations .
we show that the model of the einstein - cartan cosmology with clustered dark matter halos and their corresponding clustered angular momenta coupled to torsion , can account for small scale - large scale discrepancy and larger peculiar velocities ( bulk flows ) for galaxy clusters .
the nonvanishing total angular momentum ( torsion ) of the universe enters as a negative effective density term in the einstein - cartan equations causing partial cancellation of the mass density .
the integrated sachs - wolfe contribution of the einstein - cartan model is negative , thus it can provide partial cancellation of the large scale power of the tt cmb spectrum .
the observed violation of the isotropy appears as a natural ingredient of the einstein - cartan model caused by the spin densities of light majorana neutrinos in the early stage of the evolution of the universe and bound to the lepton cp violation and matter - antimatter asymmetry .
+ pacs numbers : 98.80.es ; 12.10.dm ; 04.60.-m |
galaxy imaging with the _ hubble space telescope _ ( _ hst _ ) has revealed a population of compact , young star clusters in a variety of starburst environments ( holtzman et al .
1992 ; benedict et al . 1993 ; whitmore et al . 1993 ; conti & vacca 1994 ; hunter , oconnell , & gallagher 1994 ; oconnell , gallagher , & hunter 1994 ; oconnell et al .
1995 ; whitmore & schweizer 1995 ; meurer et al . 1995 ; barth et al.1995 ; maoz et al . 1996 ; holtzman et al . 1996
; miller et al .
1997 ; stiavelli et al .
1998 ; carlson et al . 1998 , 1999 ; whitmore et al . 1999 ; alonso - herrero et al .
2000 ; drissen et al .
2000 ; gallagher , homeier , & conselice 2000a ; gallagher et al .
2000b ; hunter et al .
these include cooling - flow galaxies , interacting / merging galaxies , `` amorphous '' peculiar galaxies , and galaxies with circumnuclear star - forming rings .
although such clusters were known previously from ground - based and infrared observations ( e.g. , arp & sandage 1985 ; tacconi - garman , sternberg , & eckart 1996 ) , has vastly increased the number of known objects and the wavelength range and distance to which they can be studied .
the resolution has also allowed measurements demonstrating the compactness of the clusters .
the more luminous among the clusters ( @xmath11 mag [ @xmath12 , according to some conventions , though the limit is somewhat arbitrary and wavelength dependent ) are sometimes termed `` super star clusters '' ( sscs ) .
the small radii , high luminosities , and presumably high masses of these clusters have led to suggestions that they may remain as bound systems and therefore could be young globular clusters .
this view has been reinforced by velocity - dispersion measurements of several bright sscs ( ho & filippenko 1996a , b ; smith & gallagher 2000 ) .
sternberg ( 1998 ) has questioned the ultimate globular cluster fate of some of these sscs .
more general debates of the relation between sscs and globular clusters based on their statistical properties can be found in van den bergh ( 1995 , 2000 ) , meurer ( 1995 ) , elmegreen & efremov ( 1997 ) , zhang & fall ( 1999 ) , and kroupa , aarseth , & hurley ( 2001 ) .
recently , new views of sscs are emerging from infrared ( ir ) and radio observations of dense regions in starburst galaxies that may be very young sscs , still embedded in their molecular clouds ( smith et al . 1999 ; kobulnicky & johnson 1999 ; turner et al .
1999 , 2000 ; beck , turner , & kovo 2000 ) . learning more about sscs can shed light on rapid star formation , globular - cluster formation , and the possible connections between them .
circumnuclear star - forming rings in barred spirals comprise possibly the most numerous class of nearby starburst regions , and offer us an opportunity to study starburst properties at a level of detail that would be impossible for more distant or more heavily obscured systems .
the large - scale morphologies and properties of many circumnuclear rings have been studied extensively from the ground ( e.g. , hummel et al.1987 ; prez - ramrez et al .
2000 ) , but the small sizes of the rings ( typically @xmath13 for nearby galaxies ) require the capabilities of _ hst _ for studies of their detailed structure . images with wfpc , wfpc2 , and foc have revealed large populations of sscs in several circumnuclear rings in nearby galaxies ( benedict et al . 1993 ; barth et al.1995 ; meurer et al . 1995 ; maoz et al .
1996 ; buta , crocker , & byrd 1999 ) .
these clusters , barely resolved by _ hst _ , have effective radii of only 24 pc , and luminosities that range as high as @xmath14 14 to 15 mag . as in other environments ,
if such luminous clusters survive as bound entities for @xmath15 years , they will have magnitudes comparable to those of old galactic globular clusters today .
circumnuclear rings represent the only known environment in morphologically normal galaxies in which sscs are found in abundance .
furthermore , analysis of ultraviolet ( uv ) _ hst _ images of five such circumnuclear rings indicates that much of the massive star formation in the starburst regions must occur within the compact clusters ( maoz et al .
1996 ) .
measuring the masses and ages of sscs is an important step toward understanding their future evolution and determining whether they are similar to young globular clusters .
earlier _ hst _ studies of star clusters in circumnuclear rings ( benedict et al . 1993 ; barth et al .
1995 ) were based mostly on single - filter images , and the resulting interpretation of the cluster properties was uncertain .
maoz et al .
( 1996 ) used an _ hst _ optical image of the circumnuclear ring in ngc 2997 , along with a uv ( 2300 ) exposure , to show that with two bands ( i.e. , a uv - optical color ) , meaningful lower limits on the masses and upper limits on the extinctions and ages of the clusters could be set .
recently , buta et al .
( 2000 ) presented wfpc2 images in five broad uv and optical bands and in of the circumnuclear ring in ngc 1326 , another nearby barred early - type spiral galaxy . comparing their data for the numerous individual detected clusters to spectral synthesis models , they derived a continuum of ages from 5 myr to 200 myr , but with 80% to 90% of the clusters being younger than 50 myr , and typically moderate visual extinctions of 0.6 - 0.9 mag .
the derived masses of the clusters go up to @xmath16 .
they also noted that many of the h ii regions were devoid of continuum sources , and concluded that the embedded clusters must have high extinctions , @xmath17 mag .
ngc 1512 and ngc 5248 ( hubble types sbab and sabbc , respectively ) are two circumnuclear ring galaxies that were included in the maoz et al .
( 1996 ) foc uv ( 2300 ) imaging survey .
distances to these nearby galaxies ( heliocentric velocities @xmath18 and 1156 km s@xmath19 , respectively ) are not accurately known , but are of order 10 mpc .
maoz et al .
detected 43 sscs in ngc 1512 and 46 in ngc 5248 , with similar apparent brightness distributions .
here we describe subsequent broad - band ( approximately @xmath20 , @xmath8 , @xmath21 , and @xmath1 ) and narrow - band ( h@xmath22+[n ii ] and pa@xmath22 ) images of these galaxies with wfpc2 and nicmos .
these observations probe for the first time the spectral energy distribution ( sed ) of sscs over such a wide wavelength interval .
apart from providing stronger constraints than before on the physical parameters of the clusters , they have the potential to reveal a large population of clusters that is obscured by dust .
such an obscured population may be younger than that which is detectable at shorter wavelengths , which select only the clusters that have blown off , or emerged from , their birth clouds .
the hydrogen emission - line observations provide a direct probe of the ionizing flux from massive stars , and the h@xmath22/pa@xmath22 ratio is a useful reddening diagnostic .
these observations can therefore give a more complete picture of ssc properties and evolution than previously possible .
both galaxies are at high galactic latitude and hence undergo little galactic reddening ( @xmath23=0.02 $ ] mag for ngc 1512 and 0.01 mag for ngc 5248 ; schlegel , finkbeiner , & davis 1998 ) ; we will neglect these small galactic extinction corrections . in 2 , below
, we describe the observations and data reduction . in 3 ,
we analyze the reduced images .
we discuss our results and summarize the main conclusions in 4 .
wfpc2 images were taken through the f336w , f547 m , f814w , and f658n filters ( which correspond loosely to @xmath20 , @xmath8 , @xmath21 , and h@xmath22+[n ii ] @xmath24 ) .
the center of each galaxy was positioned on the pc ccd .
the rings , which have diameters of @xmath25 and @xmath26 in ngc 1512 and ngc 5248 , respectively , fit well on the @xmath27 pc field of view .
integrations were split into sub - exposures to allow cosmic - ray event rejection .
the dates and total exposure times are listed in table 1 .
the individual images were processed by the standard space telescope science institute ( stsci ) pipeline ( biretta et al . 1996 ) , which included bias subtraction and flat - fielding .
we combined the sub - exposures using the stsdas crrej task in iraf to correct for cosmic - ray hits .
nicmos camera-2 images were taken through the f160w , f187w , and f187n filters .
the f160w filter has an effective wavelength of 1.6 @xmath2 m , at the minimum of the background due to zodiacal light and thermal emission from the telescope .
the line from these two galaxies falls well within the 1% width of the f187n filter .
f187w is a broad - band filter centered at 1.87 @xmath2 m and was used to provide a measurement of the continuum underlying .
the camera-2 field of view is @xmath28 , also accommodating the full diameter of the rings . at the wavelengths of the bandpasses we used
, the camera-2 resolution approximately matches the telescope diffraction limit of @xmath29 , corresponding to 7 pc at a distance of 10 mpc . for each galaxy - filter combination ,
the nicmos observations were split into four equal exposures , between which the telescope was dithered in a `` square '' pattern with a step size of @xmath30 , to allow subsequent elimination of the effects of bad pixels in the detector .
each individual exposure was read with 18 non - destructive readouts in multiaccum mode with the step64 temporal sampling sequence .
this sequence involves logarithmically growing readout intervals up to 64 s , followed by evenly spaced intervals of 64 s for the remainder of the exposure .
the multiaccum mode produces a large dynamic range and the ability to reject cosmic - ray events .
the nicmos images were also reduced by the stsci pipeline processing system , which includes bias - level subtraction , nonlinearity correction , dark - image subtraction , flat - fielding , cosmic - ray rejection , and combination of the individual dithered exposures .
the pipeline - reduced narrow - band f187n images of ngc 1512 have a low count rate and show the `` pedestal '' feature common in nicmos images , which results from improper bias - level subtraction .
we re - reduced these images with stsdas in iraf , including the biaseq and pedsky tasks to treat the pedestal feature .
although this improved the final results , some remaining background offset in one of the four nicmos quadrants was removed by adding a constant to the counts in that quadrant .
photometric calibrations for each camera and filter , as listed in table 1 , were used to convert count rates to physical flux units .
the inverse sensitivities for the broad and medium - band filters are the latest circulated by stsci , for a source that has a constant spectrum in @xmath31 .
the inverse sensitivities of the two narrow - band filters were obtained by multiplying the inverse sensitivities for a continuum source in these filters by their effective bandwidths .
the effective bandwidths were calculated by integrating over wavelength the filter / system throughput curves and dividing by the throughput at the redshifted wavelengths of the lines for each galaxy . for ngc 5248 ,
is on the falling part of the filter / system throughput curve , where the transmission is 15% to 35% lower than at peak , for line - emitting gas with velocity of @xmath32 km s@xmath19 in the galaxy s frame . as a result , its inverse sensitivity value is higher than that of ngc 1512 , and is uncertain , for a given line - emitting region , by @xmath33 .
the effective bandwidths are 39.5 for f658n , 213 for f187n in ngc 1512 , and 275 for f187n in ngc 5248 .
we note that , in general , the effective bandwidth is significantly larger than the full width at half maximum ( fwhm ) , or other filter widths appearing in the image headers and in stsci documentation . in our analysis , we also use the foc @xmath34/96 f220w images of these galaxies presented and analyzed by maoz et al .
f220w is a broad - band filter centered at 2300 .
see maoz et al .
( 1996 ) for full details of those data . to produce continuum - subtracted p@xmath22 images ,
we `` blinked '' the f187n and f187w nicmos exposures and verified that they were well aligned ( i.e. , the telescope pointing did not change between filters , to within the resolution limits ) . for each galaxy
, we then scaled the f187w image by the ratio of the inverse sensitivities of the f187n and f187w filters , and subtracted it from the f187n image .
the emission in these galaxies comes from relatively extended regions having weak continuum emission ( see below ) .
the central wavelengths of the two filters are the same , and the continuum makes only a small contribution to the f187n bandpass in these regions .
consequently , we find that the difference images are insensitive to the exact method of continuum subtraction , and there is no need to perform higher - order corrections ( e.g. , to account for the slope of the continuum by utilizing the f160w images ) .
we used combinations of scaled versions of the f547 m and f814w wfpc2 images to produce continuum images at 6580 for each galaxy , for subtraction from the f658n images .
the scaling and relative weights were calculated from the central wavelengths , the inverse sensitivities , and the relative exposure times in the three filters , and we found that this gives a good continuum subtraction , without artifacts such as conspicuous negative - flux regions .
as with the images , the good subtraction is a result of the mutual avoidance of stellar and emission - line light in these galaxies .
quillen & yukita ( 2000 ) recently presented an analysis of some of these data .
their work concentrates only on the emission - line data for just one of the galaxies discussed here ( ngc 1512 ) , and on the relations between reddening , gas density , and metallicity . instead of the effective badwidths
, they apparently used the plateau width for the f658n filter and the fwhm for the f187n filter in calibrating the narrow - band images of ngc 1512 .
their fluxes are therefore underestimated by 40% and their fluxes by 11% . in their reduction
, they used the f160w image for continuum subtraction .
apart from the relatively large wavelength difference , which may compromise the accuracy of their continuum subtraction , the mismatch between the point - spread functions ( psfs ) in the f160w and f187n bands is apparent from their figure . to enable tracking the sed of individual sources from the uv to the ir , and for constructing ratios of images in different bands
, we performed a careful alignment of the images .
all images of a galaxy were registered with the f547 m image .
this was done by `` blinking '' a given image and the f547 m image , and marking the coordinates of the brightest isolated clusters appearing in the two bands .
this proved to be fairly challenging for bands far from the f547 m band , due to the diverse colors of the clusters ( e.g. , some clusters that are bright in @xmath8 are barely visible at @xmath35 m , and difficult to identify among the myriad of ir sources that emerge at longer wavelengths ) .
the cluster coordinates were used to calculate a transformation between images using the geomap task in iraf , allowing for rotation , translation , and scaling , and the geotran task was used to apply the transformations to the images .
the emission knots in the image are extended and , as mentioned above , not coincident with the clusters seen in the continuum bands . however , since all the nicmos exposures of each galaxy were already well aligned ( including the f187n exposure , which includes continuum emission ) , the f160w transformation was applied to the continuum - subtracted image .
blinking between the optical and the transformed ir images shows that the registration is not perfect , due to local distortions in the nicmos images .
however , the mismatch is small enough to avoid any ambiguities in identifying a source in the various bands .
the transformations between the coordinates of the foc uv images of these galaxies and the wfpc2 images were similarly calculated , and the uv sources tabulated by maoz et al .
( 1996 ) matched with their optical counterparts .
to produce an image of the ( + [ n ii])/ratio , we first matched the angular resolution of the + [ n ii ] image and the image by convolving each image with the psf of the other image s bandpass .
the wfpc2 f658n and nicmos2 f187n psfs were determined using the tiny tim software ( krist & hook 1997 ) .
we then subtracted a median background level from each continuum - subtracted , smoothed , emission - line image , and divided the two .
figures 1 and 2 show mosaics of the final galaxy images in the different observed bands .
figures 3 and 4 show several color - composite views of each galaxy , combining data from several bands .
figures 5 and 6 illustrate the ( + [ n ii ] ) and images with the main line - emission complexes marked .
figures 7 and 8 show the + [ n ii ] emission as contour plots superimposed on the `` @xmath20-band '' f336w images .
figure 9 gives the ( + [ n ii])/ratio map for ngc 1512 . the large width of the ring in ngc 5248 ( see below ) results in a low emission - line surface brightness and a noisy ( + [ n ii])/ratio map of limited utility , which we therefore do not show .
although similar in most respects ( see below ) , the overall morphologies of the circumnuclear rings differ in ngc 1512 and ngc 5248 .
ngc 1512 has a highly ordered and narrow ring , with a radius - to - width ratio of @xmath36 .
nearly all of the line emission and the compact sources are concentrated in this ring . in ngc 5248
, the radius - to - width ratio is @xmath37 , and the h ii regions are distributed over a much larger area .
these differences may be related to the fact that ngc 1512 is a more strongly barred galaxy than ngc 1512 .
ngc 5248 also has a second , inner , ring of @xmath38-diameter major axis ( 120 pc , assuming a distance of 10 mpc ) , which is seen most clearly in its + [ n ii ] image . to our knowledge
, this feature has not been noted as such before , but it coincides with the inner part of the spiral found in continuum images by laine et al .
( 1998 ) .
comparison of the images of each galaxy in the various filters reveals a number of features .
the appearance of the galaxies changes dramatically when going from the uv to the ir , because of the growing prominence of the red populations of the central galactic regions . at first glance
, the appearance of the circumnuclear rings does not seem to change so drastically between bands , with the same major complexes of star formation being traceable from the uv to the ir .
closer examination shows , however , that the structure within individual complexes is strongly wavelength dependent . specifically , the individual stars and clusters have a large range in color , making the most prominent sources in one band inconspicuous or absent in another band .
a good example of this is the close pair of bright clusters in the southeast part of the ngc 1512 rings ( top of panel in figure 1 . )
the lower - left member of the pair is considerably fainter in the uv bands , but is brighter than the upper - right cluster at 1.6 @xmath2 m ( see also 3.3.5 ) .
similarly , a progressively larger number of sources become discernible as one goes to the ir , and their ir fluxes are comparable to the ir fluxes of the bright uv sources .
this large diversity in color and in the bands in which specific sources are above the detection threshold create the `` christmas tree '' appearance of the color composites in figures 3 and 4 .
this phenomenon is treated more quantitatively and physically in 3.3 .
the uv images of the galaxies show that the brightest regions follow a definite spiral , rather than ring , structure .
the optical and ir images prove , however , that this effect is produced by spiral dust structures that cross the rings .
as dust obscuration becomes less pronounced at ir wavelengths , the `` gaps '' between the ends of the spirals nearly close .
the true ring morphology of the underlying ionizing population is seen clearly in the + [ n ii ] and images , in which the spiral structure is no longer evident .
the continuum and emission - line images differ in appearance not only in terms of the spiral structure , but in other respects as well .
while the brightest continuum emission from the circumnuclear rings comes from numerous compact ( see below ) clusters and stars , the + [ n ii ] and emission occurs in more extended , resolved clumps encompassing entire complexes of clusters . in several regions ,
the line - emitting gas may actually have a bubble morphology , surrounding a group of young clusters . in the parts of the rings showing both strong continuum and line emission ( e.g. , the bright southern complex in ngc 1512 top of figure 7 ) , closer examination of the + [
n ii ] contour map overlaid on the @xmath20-band continuum image shows that the brightest emission - line clumps are at positions that are offset from the brightest continuum sources .
this suggests that expansion due to winds or ionization fronts around the clusters clears out the gas and the dust .
such shell - like or bubble - like morphologies around young clusters are also seen , for example , in ngc 1569 ( hunter et al .
2000 ) , ngc 5253 ( strickland & stevens 1999 ) , and ngc 4214 ( maz - apellniz et al .
1999 ) .
more remarkably , perhaps ( and as already noted above ) , in some regions of the rings the line emission and continuum emission are mutually exclusive .
this is conspicuous , for example , in ngc 5248 ( figure 8) , where clumps of + [ n ii ] emission trace ionizing sources that are not detected in any of the continuum bands .
this type of emission - line / continuum avoidance has been noted before in other starbursts ( e.g. , ravindranath & prabhu 1998 ; buta et al .
it can be seen clearly , for instance , in the composite and continuum image of the merging galaxy pair ngc 4038/4039 ( fig . 4 in whitmore et al .
1999 ) . to investigate this phenomenon quantitatively ,
we have measured the emission - line and continuum fluxes from the main h ii complexes along the rings .
table 2 lists the integrated + [ n ii ] , , and @xmath35 m fluxes , and their ratios , as measured in circular apertures roughly centered on the main emission - line complexes .
the labels and apertures for each complex are marked in figures 5 and 6 .
the radii of the apertures , also listed in table 2 , were approximately matched to the size of each emission complex . because of the spatial correlation between the h ii regions , the definition of what constitutes an individual region is not unique , and the choices of regions and radii are somewhat subjective .
these measurements nevertheless provide some rough estimates of the physical properties of the gas .
quillen & yukita ( 2000 ) have compiled a similar table for ngc 1512 .
apart from the incorrect calibrations , mentioned above , there seems to be an error in the offset coordinates of some of the h ii regions they list ( e.g. , the bright complexes 4 , 5 , 7 , and 8 are not on their list ) .
assuming approximate distances of 10 mpc to the galaxies , the typical diameters of the brightest h ii regions are in the range 20 - 50 pc , and the observed + [ n ii ] luminosities are @xmath39 to @xmath40 erg s@xmath19 .
as we will see below , based on the /intensity ratios , the typical extinction of is 12 mag , or more , if the gas and dust are mixed , as seems to be the case .
the extinction - corrected + [ n ii ] luminosities are therefore about an order of magnitude higher than the above numbers .
for comparison , the luminosity of the orion nebula is @xmath39 erg s@xmath19 , and that of 30 doradus is about @xmath41 erg s@xmath19 , the rough lower limit denoting `` giant h ii regions '' ( kennicutt & chu 1988 ) .
the total + [ n ii ] luminosities of the rings , obtained by integrating the fluxes over the entire rings , are @xmath42 erg s@xmath19 ( ngc 1512 ) and @xmath43 erg s@xmath19 ( ngc 5248 ) .
again , the extinction - corrected luminosities are likely an order of magnitude larger , or about @xmath44 erg s@xmath19 .
this is very similar to the luminosity derived by buta et al .
( 2000 ) for the starburst ring in ngc 1326 .
it implies an ionizing photon luminosity of @xmath45 s@xmath19 and a star - formation rate of about @xmath46 yr@xmath19 , assuming a continuous starburst with a salpeter initial mass function ( imf ) from 0.1 to 100 @xmath47 ( kennicutt 1998 ) .
@xmath2 m ) ratio , or the equivalent width of , is a good age indicator , since it basically measures the ratio of o stars to k and m giants and supergiants , and is insensitive to reddening because both measurements are at similar wavelengths . using spectral synthesis models
( see 3.3.3 for details ) for a starburst with a star - formation rate that exponentially decays on a timescale of 1 myr , we find that the /@xmath31(1.6 @xmath2 m ) ratio has a value of @xmath48 for the first few @xmath49
it then declines by 2 orders of magnitude and becomes of order 1 by @xmath50 years . the nebular continuum emitted by the gas
is included in this calculation .
the observed ratios in table 2 are actually lower limits to the intrinsic ratios for the starbursts producing the lines , because our integration includes continuum emission from unrelated foreground and background populations of stars . from the observed numbers , which are
roughly in the range 50500 , we see that the bright emission - line complexes have the ratios expected for a mixture of gas and an ionizing population that is a few million years old .
the ( + [ n ii])/ratios , also listed in table 2 , can give some indication of the amount of dust mixed with , or along the line of sight to , the gas .
as explained in more detail below , in the absence of reddening , the expected ( + [ n ii])/ratio in these galaxies is about 10 .
the ratios listed in table 2 are typically 2 to 5 times lower , and thus indicate some extinction in most of the complexes .
the ( + [ n ii])/ratio map of ngc 1512 in figure 9 reveals additional interesting information regarding the line emission .
the overlaid contours are the emission , marking the regions where there is significant detected flux in this line , above a level of @xmath51 erg s@xmath19 @xmath52 arcsec@xmath53 .
the inner contour corresponds to @xmath54 erg s@xmath19 @xmath52 arcsec@xmath53 .
the image is shallower than the + [ n ii ] image , because of the lower flux in , and the lower sensitivity , higher background , and poorer angular resolution in the ir .
the image therefore limits the regions where the ratio can be measured reliably ; outside the outer marked contours , the ratio is dominated by noise and background residuals .
the whitest regions correspond to a ratio of about 1 , while the darkest areas have a ratio of @xmath36 .
concentrating on the pixels within the contours of significant emission , we see that most of these have a greyish shade , which corresponds to ratios of about 4 .
the same is true for ngc 5248 ( ratio map not shown ) , but the low surface brightness of in that galaxy limits the measurement to a few small regions .
the dominant error in the ratio is caused by the uncertainty in the background levels that need to be subtracted from the + [ n ii ] and images prior to their division . to estimate this error
, we created ratio images using extreme plausible values for the backgrounds . we find that , within the marked contours , the ( + [ n ii])/uncertainty is @xmath55 , decreasing to @xmath56 in the regions of highest line emission .
this error is compounded by the @xmath57 calibration uncertainty in ngc 5248 due to the unknown line - of - sight velocity of each clump . the expected case b recombination value of /is about @xmath58 .
the heliocentric radial velocities of these galaxies ( 911 km s@xmath19 for ngc 1512 , and 1156 km s@xmath19 for ngc 5248 ) places both lines of the [ n ii ] @xmath59 doublet also within the f658n bandpass . for the nucleus of ngc 1512 ,
[ n ii ] @xmath60 is on the blue edge of the bandpass , at about 10% of the peak transmission , while [ n ii ] @xmath61 falls on the flat transmission plateau , along with . for ngc 5248 , the [ n ii ]
lines are on the rising and falling parts of the bandpass , each at about 50% of the peak transmission , and is on the plateau .
however , the individual h ii regions in the rotating rings can have line - of - sight velocities of order 100 km s@xmath19 relative to that of the nucleus ( e.g. , maoz et al . 1996 ) , shifting the lines by @xmath62 , and changing somewhat the relative contributions of each of the [ n ii ] lines to the counts in the bandpass . the nuclear spectrum of ngc 5248 presented in ho , filippenko , & sargent ( 1995 ) shows the [ n ii ] lines to be of comparable strength to .
however , in the circumnuclear h ii regions , the [ n ii ] lines , relative to , are usually much weaker ( e.g. , ho , filippenko , & sargent 1997 ) .
specifically , we have measured the relative line strengths of [ n ii ] and in two of the ring h ii regions that fall along the slit in the ho et al .
( 1995 ) spectrum of ngc 5248 .
assuming these two h ii regions are representative of the rest , and accounting for the f658n transmission at the systemic velocity , [ n ii ] contributes 14% to 18% of the counts in the bandpass .
the maximum [
n ii ] contamination , @xmath6323% to 28% , will be for a velocity of @xmath64 km s@xmath19 relative to systemic .
we do not have a spectrum of ngc 1512 , but assuming it is similar to ngc 5248 , at the systemic velocity [ n ii ] contributes 22% to 28% to the counts , and less than this for velocities of order @xmath32 km s@xmath19 .
from all these considerations , in the absence of reddening , we expect the observed ( + [ n ii])/flux ratio in both galaxies to be in the range of 10 to 12 , substantially above the /ratio .
the typical values integrated over complexes ( table 2 ) and observed in the ratio maps , in the range of 25 , again suggest that the line - emitting gas suffers from some dust extinction . assuming the galactic extinction curve of cardelli , clayton , & mathis ( 1989 ) , with @xmath65 , a moderate amount of extinction , @xmath66 1.53 mag , by a foreground dust screen would produce the observed values .
alternatively , the gas and dust may be mixed .
the observed line ratio is then just the intrinsic ratio times the ratio of the optical depths . for an infinite gas+dust slab ( @xmath67 )
the expected line ratio is about 2 , similar to what is seen in many of the regions .
it is likely that a combination of these two geometries is operating in practice .
this would explain the absence of continuum sources associated with some of the main emission - line complexes .
examining in closer detail the ratio maps , they do reveal some small , dark - coded clumps , often near the peaks of line emission , corresponding to higher ( + [ n ii])/ratios , of about 6 to 9
. these must be regions where the gas either undergoes less extinction , or has a lower ionization spectrum , with a higher [ n ii]/ratio , more similar to that of the nucleus .
this can arise through excitation by shocks , or photoionization by unusually hot stars , which produce a hard ionizing continuum ( shields & kennicutt 1995 ; ho et al .
similarly , some regions of the rings also have lower ( + [ n ii])/ratios .
for example , patch no .
6 in the west side of the ring in ngc 1512 has a fairly uniform value of ( + [ n ii])/@xmath68 12 .
the small ratio in this patch can also be appreciated by examining directly the + [ n ii ] and images in figure 5 . in ngc 1512
, we do not see any regions that have detected emission , but little or no + [ n ii ] .
ngc 5248 , on the other hand , does have some clumps that are faint or undetected in + [ n ii ] ( for example clumps no . 15 and 16 ) .
the average ( + [ n ii])/in these clumps is @xmath69 , with lower values in particular regions of a clump .
this indicates foreground dust , equivalent to the effect of @xmath66 34 mag of galactic dust , along the line of sight to some of the line - emitting gas in this galaxy .
we find additional evidence for substantial dust extinction by comparing the rings total ionizing photon luminosities of @xmath45 s@xmath19 , derived above based on the flux , to their uv luminosities .
maoz et al .
( 1996 ) measured total flux densities at 2300 for each of the rings , of @xmath70 erg s@xmath19 @xmath52 @xmath19 , or specific luminosities @xmath71 erg s@xmath19 @xmath19 at the assumed 10 mpc distances .
they found that @xmath6340% of this total uv flux comes from compact sources , and mostly from the few brightest ones .
as we will see below ( 3.3.4 ) , the seds of these bright clusters are like those expected from clusters that are 110 myr old undergoing typical attenuations at 2300 by factors of 4 ( in ngc 1512 ) to 18 ( in ngc 5248 ) .
the extinction - corrected luminosities are therefore @xmath72 erg s@xmath19 @xmath19 ( ngc 1512 ) and @xmath73 erg s@xmath19 @xmath19 ( ngc 5248 ) . from our spectral synthesis models ,
we find that a 1 myr - old starburst with the above uv luminosities produces @xmath74 s@xmath19 ionizing photons , while a 10 myr - old model with the same uv luminosity produces fewer ionizing photons by two orders of magnitude .
thus , under the extreme assumption that all the observed uv light is produced by a stellar population that is only 1 myr old ( or , equivalently , all the uv light is produced predominantly by o - type stars ) , the deduced ionizing photon flux is similar to that implied by the extinction - corrected flux . for a more conservative assumption about the spectrum of the integrated ionizing population ,
there is a deficit of ionizing photons by a factor of 1100 .
this again indicates that the blue sources we see are only a fraction of the ionizing sources , while some are inside the line - emitting gas , and extinguished by the dust mixed with it .
given the uncertainties in determining the emission - line extinction and the precise ages of the ionizing sources , it is difficult to say what the exact obscured fraction is . a peculiar , compact , emission - line source is found in the northern side of the ring in ngc 1512 [ pc position ( 452,251 ) , nicmos position ( 204,194 ) , labeled `` a '' in fig .
5 ] . in the + [ n ii ] image
it is one of the highest surface - brightness objects , but is very compact and round , with a fwhm of @xmath9 , and total + [ n ii ] flux of @xmath75 erg s@xmath19 @xmath52 , or an + [ n ii ] luminosity of @xmath76 erg s@xmath19 . in the image , it is barely detected , and we estimate the ( + [ n ii ] ) /ratio to be @xmath6325 . in the broad bands ,
this source is detected , if at all , as a faint , unresolved object .
it is undetected in the uv in the foc / f220w , with a limit of @xmath77 erg s@xmath19 @xmath52 @xmath19 , but detected with wfpc2 in f336w , with @xmath78 erg s@xmath19 @xmath52 @xmath19 .
it is only barely detected in the f547 m band , with @xmath79 erg s@xmath19 @xmath52 @xmath19 ,
i.e. , @xmath80 mag , or @xmath81 . at longer wavelengths , in the wfpc2 f814w and nicmos f187w bands
, it brightens again , with @xmath82 erg s@xmath19 @xmath52 @xmath19 , and @xmath83 erg s@xmath19 @xmath52 @xmath19 .
given the large equivalent width of + [ n ii ] ( about 7000 ) , it is possible that the counts in the broad bands are affected or dominated by other emission lines . in view of the discussion above , if the object has a recombination spectrum , then most of the flux in the + [ n ii ] band will be [ n ii]@xmath61 , with a ratio of [ n ii]@xmath61/of about 2 , or more if the line emission in reddened .
alternatively , other types of line excitation may be at work .
the properties of this source are similar to those of `` balmer - dominated '' supernova remnants ( snrs ) that are seen in the galaxy ( e.g. , tycho s snr ) and in the large magellanic cloud ( lmc ; smith et al .
the optical light from these young ( typically @xmath84 yr ) snrs consists almost exclusively of , which is thought to arise from a very fast shock that excites the line through collisional excitation of neutral hydrogen atoms and charge transfer with protons .
object a may be the first balmer - dominated snr found beyond the lmc .
another possibility is that object a is a supernova event similar to sn 1986j ( rupen et al .
1987 ; leibundgut et al .
1991 ) and sn 1988z ( aretxaga et al . 1999 , and references therein ) .
these objects last a long time , and have late - time optical spectra totally dominated by emission .
a primary goal of our work is to improve our understanding of the numerous discrete sources
compact clusters and individual stars that appear in the circumnuclear rings , by measuring their seds .
as seen in the figures above , the prominence of individual sources can change dramatically , with some that are bright in one band becoming faint or undetected in another .
we wish to obtain fluxes , or upper limits on fluxes , in every band for every source that appears in one or more bands .
we therefore created a summed `` uv - optical '' image consisting of the f220w , f336w , f547 m , and f814w images , which were first scaled to all have similar means .
the daofind task was applied to this combined image to create a list of discrete sources .
the complicated , uneven background and crowding of the sources mandated that the output of this task be edited manually , to delete false detections and append obvious misses .
such a procedure was found to be unavoidable also in previous work on these and other ring galaxies ( barth et al .
1995 ; maoz et al .
we verified that no obvious sources appearing in the individual optical - uv bands had been overlooked .
we attempted to locate all sources in the regions of the rings , but not in the more exterior regions , which are outside the fields of view of the foc and nicmos , nor interior to the rings , where the bright stellar background makes it progressively more difficult to detect sources at smaller radii and at longer wavelengths .
this procedure can not be considered objective , but our compilation should be complete for the brightest sources in each ring . in any event , there are few clusters outside or inside the rings .
altogether , 325 optical - uv sources were selected in ngc 1512 and 397 in ngc 5248 .
most of the sources in the nicmos 1.6 @xmath2 m exposures are not obviously detected in the optical images .
we therefore applied the same source detection procedure separately to the f160w images .
of the 221 ir sources in ngc 1512 , 150 were not selected in the uv - optical image . in ngc 5248 ,
110 of the 191 ir sources were not selected in the uv - optical image .
the combined uv - optical - ir lists of sources thus consist of 475 sources in ngc 1512 and 507 sources in ngc 5248 .
aperture photometry at the positions of the sources in the final list was carried out on the images in each individual band using the phot task in iraf . for the wfpc2 images , apertures of radii 1 , 2 , and 3 pixels
were used .
the 2-pixel - radius apertures are best suited for wfpc2 pc photometry , since they are large enough to minimize photometry errors due to pixelation and to centering errors , but small enough to minimize the effects of crowding on the photometry ( e.g. , barth et al .
the nicmos images have a larger psf and a coarser pixel scale , with one nicmos pixel corresponding to about 1.67 pc pixels . to avoid uncertainties in the photometry due to the resampling and registration of the nicmos images ,
nicmos photometry was performed on the original images .
the positions on the nicmos frames of all the uv - through - ir - selected sources were computed , based on the geometric transformations described above .
we used an aperture of radius 1.79 nicmos pixels , which is equivalent in angular scale to 3 wfpc2 pc pixels .
this was chosen as a compromise , given the constraints , on the one hand , to match approximately the pc apertures in order to measure the same objects in the optical and ir bands and avoid crowding effects while , on the other hand , to avoid overly small apertures that will be susceptible to pixelation and centering errors .
aperture corrections need to be applied to the measurements to account for the fact that not all the light of the psf is included within the aperture . as discussed in 3.3.7
, the individual sources in these galaxies appear to be unresolved or marginally resolved .
we therefore calculated the aperture corrections by generating artificial psfs for each of the wfpc2 and nicmos bands using the tiny tim software , and finding the ratio between counts within the radii used for the photometry and the radii corresponding to @xmath85 , which should include 99% of the light according to the stsci website .
photometry errors due to photon statistics and background uncertainties were calculated by the phot task . to this
, we added in quadrature a calibration uncertainty of 5% , which includes the effects of flat - fielding errors , charge - transfer efficiency and uv contamination variations in the wfpc2 detectors , uncertainties in aperture corrections , and uncertainties in the absolute sensitivities of each of the bands ( see the wfpc2 and nicmos data handbooks ) .
after aperture correction and photometric calibration , all of the photometry was consolidated into one list for each galaxy , including the 2300 fluxes from maoz et al .
tables 3 and 4 show the final photometry ordered by brightness in the f547 m filter , in decreasing order , for each of the galaxies .
the full tables for all the measured sources in both galaxies are available in the electronic version of the journal .
figure 10 is a color - magnitude diagram showing , for both galaxies , all of the sources detected in f547 m and f814w .
for ease of comparison with other work , the data in this figure have been converted to standard @xmath8 and @xmath21 magnitudes according to the transformations of holtzman et al .
the @xmath8 magnitudes also appear in tables 3 and 4 .
the figure shows that the brighest sources are blue , with @xmath86 .
the sources in ngc 5248 have a larger spread in color than those in ngc 1512 , and are typically 0.5 - 1 mag redder .
figures 1112 plot the seds of the brightest individual sources in each galaxy . for clarity ,
only the 25 brightest sources in each galaxy are shown , separated into different panels according to brightness in @xmath8 , and error bars are omitted .
missing measurements at 2300 and at 1.6 @xmath2 m are for objects outside the field of view of the foc or nicmos , or , for the fainter objects , are below the detection limits .
a cursory examination of the seds again shows that almost all of the bright sources are blue , with spectra that can be approximated as @xmath87 , with slopes @xmath88 of @xmath89 to @xmath90 .
these sources apparently undergo little reddening .
there are no bright sources that have seds which rise to the red .
the most luminous known red supergiants have @xmath91 mag , or , at 10 mpc , @xmath92 mag and @xmath93 erg s@xmath19 @xmath52 @xmath19 .
while many of the sources in the rings , including some of the more luminous ones , are only a few times brighter than this , their blueness excludes the possibility that they are individual red supergiants , rather than young clusters . as for individual blue supergiants
, they can have even higher luminosities , up to @xmath94 mag , but they are extremely short - lived and rare , and hence are not plausible candidates for explaining any but a handful of the observed sources . if , for the purpose of a rough appraisal of the content of the clusters , one assumes the uv is dominated by the light of main - sequence b0 stars , the observed fluxes ( uncorrected for extinction ) of the 25 brightest clusters correspond to 5 to 500 such stars . for reasonable imfs ,
the total masses of these clusters are of order @xmath95 to @xmath96 .
masses in this range also result from our detailed model fitting ( 3.3.6 ) .
it is only the sources with @xmath97 erg s@xmath19 that may , indeed , often be individual supergiant stars . however , some of these objects may also be reddened sscs and old , or low - mass , clusters . for reference , old globular clusters have @xmath91 mag , similar to the most luminous red supergiants . thus , some of the faint objects could also be old globulars from a previous starburst .
this possibility is examined in more detail below . to interpret the measured seds , we use the population and spectral synthesis code stars ( sternberg 1998 ; thornley et al .
2000 ) to generate model seds of clusters with a range of ages and assumed foreground extinctions . for the population synthesis , stars employs the geneva stellar evolutionary tracks ( schaller et al .
1992 ) . in the computations presented here ,
we assume solar metallicity tracks .
we also assume a time - independent power - law imf , @xmath98 , between a lower and an upper mass limit .
for the spectral synthesis of the cluster seds we adopt the procedures described in thornley et al ( 2000 ) .
briefly , the spectral energy distributions ( seds ) for individual stars are taken from a `` hybrid grid '' generated from two libraries of theoretical model atmospheres . for stars with effective temperatures
@xmath99 k we use the kurucz ( 1992 ) models which assume atmospheres in local thermodynamic equilibrium ( lte ) . for hotter stars ( @xmath100 k )
we adopt the more realistic non - lte models of pauldrach et al .
the seds for intermediate temperatures are obtained by interpolation between the kurucz and the pauldrach et al . models .
a more detailed discussion and comparison of the library seds is presented in thornley et al .
( 2000 ) .
we have compared some of our broad - band photometric computations with those of other codes , such as starburst99 ( leitherer et al . 1999 ) , and find good agreement .
for example , our computed @xmath8 magnitudes agree well ( generally to within 0.3 mag ) with those of leitherer et al . for both continuous and decaying bursts and a range of imf slopes and upper mass cutoffs . within the scope of the present study ,
we have limited ourselves to models with a salpeter ( 1955 ) imf with upper and lower mass limits @xmath101 and @xmath102 , solar metallicity , and an exponentially decaying star - formation rate with a characteristic timescale of 1 myr .
solar metallicity is a reasonable assumption for circumnuclear regions of barred spirals , where radial mixing occurs ( e.g. , roy & belley 1993 ; martin & belley 1997 ) , and , in any event , the broad - band properties are weakly dependent on metallicity ( e.g. , leitherer & heckman 1995 ) .
we find that the model calculations are only weakly affected if , instead of a salpeter slope of @xmath103 we assume an imf slope of @xmath104 , or if the burst timescale is 5 myr instead of 1 myr .
figure 13 illustrates the spectral evolution of a model cluster .
the five curves correspond to ages of 1 , 10 , 30 , 100 , and 300 myr , with the steepest sed corresponding to the youngest age , and then fading with age .
an arbitrary vertical shift has been applied to the models as a whole ( but not relative to each other ) .
a vertical shift is equivalent to a change in the mass of a cluster or in the distance to a galaxy .
the light from a 1 myr - old model cluster is dominated at all bands by main - sequence o and b stars , and by evolved massive stars .
the overall spectral slope is approximately a power law with slope @xmath105 . by 10 myr ,
the sed develops a change of slope longward of the balmer edge , where the light becomes dominated by red supergiants , but the slope at shorter wavelengths also flattens , as the most massive stars disappear .
the continued evolution of the cluster is characterized mainly by a slow overall fading , along with a strengthening of the balmer jump , and only mild changes of slope .
note that there is less than a factor of 10 difference in @xmath8 luminosity between the oldest and youngest model clusters we have considered .
therefore , coexisting clusters within this wide range of ages can be detected , in principle .
as we show below , however , the steep mass and luminosity functions of the clusters create an age bias that will cause young clusters to appear highly over - represented in a given sample .
figure 14 illustrates the effects of extinction by foreground dust on one of the models , and was calculated using the galactic extinction curve of cardelli et al .
( 1989 ) , with @xmath65 .
although many other extinction laws are possible , we will , again , limit our investigation to this one .
one way of comparing the models and the observations is by means of color - color diagrams , as has been done , for example , by whitmore et al .
( 1999 ) and buta et al .
a problem with this method is that , for particular choices of bands , the age trajectories of model clusters sometimes `` backtrack '' on themselves , making it difficult to uniquely determine the ages of some of the observed clusters .
furthermore , age and extinction may move points in the diagram in a similar direction , so that there is a degeneracy in these two parameters .
we have found that most of the color - color plots we can make with our available broad - band data suffer from the same problems .
however , the use of the uv ( 2300 ) band can partly solve this .
figure 15 shows the models and the data on a plot of @xmath106 $ ] vs. @xmath107 $ ] ( essentially , @xmath108 and @xmath109 , up to additive constants ) .
the model fluxes were calculated by integrating the model seds over a bandwidth equal to the filters fwhm .
cluster aging trajectories for several values of foreground extinction are shown .
ages , in myr , are marked in large numerals along the top trajectory .
one sees that , in this color - color space , the individual trajectories are well stretched out , partly solving the `` backtrack '' problem .
the reddening shifts the curves in a direction nearly orthogonal to the aging direction , partly alleviating the age - extinction degeneracy .
unfortunately , it is possible to place on this plot only the @xmath110 clusters in each galaxy that were detected in the relatively shallow pre - refurbishment foc uv images .
nevertheless , these plots are revealing .
figure 15 plots with progressively smaller symbols sources that are fainter in @xmath8 .
typical error bars are shown in the corner . to avoid cluttering ,
only the brightest clusters are labeled according to their designations in tables 3 and 4 .
it is clear that the brightest clusters are very young , with ages generally less than 5 myr for the 15 brightest clusters , and generally less than 10 myr for all the uv - detected clusters .
furthermore , the extinctions are quite moderate , with the brightest sources clustered around visual extinctions of @xmath111 mag in ngc 1512 and @xmath112 mag in ngc 5248 .
these correspond to the attenuations at 2300 by factors of 4 and 18 , respectively , already mentioned in 3.2 . to illustrate the form of a color - color diagram in another band , for which most of the sources do have measurements , figure 16 plots the same model trajectories in the plane of @xmath107
$ ] vs. @xmath113 $ ] ( essentially , @xmath109 and @xmath114 , up to additive constants ) . for clarity ,
only the models are shown in this plot .
the problems of backtracking and age - extinction degeneracy are evident . in figure 17
, the data for ngc 1512 are added to a similar plot , which shows additional extinctions , but omits some age labels and all cluster labels for clarity .
the 20 clusters brightest in @xmath8 are again shown as the largest points , and each set of the next 100 brightest sources is plotted with progressively smaller dots .
most of the clusters have separations from at least some model trajectories that are consistent with the statistical and systematic calibration uncertainties of the measurements , as represented by the typical error bars plotted . as in figure 15
, the diagram indicates that most of the bright clusters are young and only mildly reddened .
however there are two bright sources above , and to the left of , the others , that are simultaneously very blue in @xmath109 , but red in @xmath114 , and are about 0.5 mag from the nearest model .
( these are sources nos .
7 and 17 in table 3 . )
this `` ir - excess '' phenomenon is examined in more detail below . for most of the fainter sources ,
it is impossible to disentangle age and extinction in this plot , and the larger measurement errors exacerbate the problem .
some of the faintest sources are probably individual stars .
another method of estimating ages and extinctions , which exploits all the available broad - band information , is to directly compare the modeled and observed seds . to this end , we have generated a grid of models with ages between 1 myr and 300 myr , and extinctions between 0 and @xmath115 mag .
each model sed was multiplicatively scaled to best match each observed sed and the @xmath116 residual of the fit calculated .
figure 18 shows examples of three observed seds in ngc 5248 and their model fits . for the fainter cluster displayed
, two models are illustrated one that fits well , and one that is marginally consistent with the data .
this demonstrates the power of this diagnostic technique when measurements in numerous bands over a wide wavelength range are available . for most of the observed seds
we can find at least one model that provides an acceptable fit .
given the observational uncertainties in the seds and the remaining age - extinction degeneracy in some cases ( e.g. , for some clusters without a uv measurement ) , more than one model can fit most of the observed sources .
however , some sources are poorly fit by all models we have considered .
surprisingly , it is the brightest ( and bluest ) sources in both galaxies that often suffer from this problem , even though these are least susceptible to measurement errors due to , e.g. , crowding and background uncertainties .
figure 19 shows examples of seds and best - fit models for four such clusters in ngc 1512 ( nos . 2 , 3 , 4 , and 8 in table 3 ) . clearly , the poorness of the fits ( which are all formally rejected ) is due to the @xmath35 m measurements , which are 2 - 5 times higher than the predictions of the models that reproduce well the data in the other bands .
it is unlikely that this ir deviation in some of the bright sources is a calibration problem , since most of the sources in both galaxies _ are _ well fit in all bands , including the ir .
the phenomenon can actually be seen directly by examining , in fig .
1 , the two close bright clusters near the 12 oclock position . these are clusters nos . 1 and 5 in table 3 ( i.e. the brightest , and fifth brightest , in @xmath8 ) , and their seds appear in fig .
cluster no .
1 is several times brighter than cluster no . 5 in all bands , except in the ir , where clearly no . 5 overpowers no .
1 . cluster no .
1 is about as blue as a cluster can be , so if one changed the ir calibration to make the sed of no .
5 fit the models , cluster no
. 1 ( and many others , too ) would be too blue ( i.e. , too weak in the ir ) to fit any model .
this , again , argues against a calibration problem .
one could invoke a systematic error , such as saturation , that affects only the brighter sources .
however , saturation would make bright sources appear anomalously faint , rather than bright , and , in any case , no such effect has been reported with nicmos for sources of this brightness .
we believe that the observed ir excess is real , and signals some physical element that is absent from the models . as an example
, the dashed line in figure 19 shows the effect of adding to the lower model a @xmath117 k blackbody , scaled to reproduce the ir excess of the fainter cluster in the diagram .
the excess in this object is large enough that it seems to appear both in the @xmath21 and the @xmath35 m bands , permitting some constraints on its form .
if the excess is from dust which is heated by the cluster stars to @xmath117 k , its distance from the stars is @xmath118 stellar radii ; that is , the dust must be circumstellar .
tables 5 and 6 give the best - fitting model parameters for the sources in the two galaxies , with the full tables again provided in the journal s electronic edition .
figure 20 shows the ages and extinctions for the sources in the two galaxies , regardless of the acceptability of the fits .
given the ir - excess phenomenon described above , the best fits can still provide good estimates of the ages and the extinctions .
since the parameters of sources detected in only a few bands are poorly constrained , we fit only those sources detected in four or five bands ( 258 sources in ngc 1512 and 328 in ngc 5248 ) .
point sizes are proportional to @xmath8 brightness , and points have been given small arbitrary vertical and horizontal shifts to avoid overlap .
the vertical striping pattern in the plots is the result of the discrete grid of models used .
there is generally good agreement between the individual ages and extinctions deduced from the color - color diagrams in figure 15 ( for those clusters that have a uv measurement ) and those found from model - fitting , as listed in tables 5 and 6 .
the model - fitting , however , includes data from two additional bands , and is therefore sometimes more constrained .
in ngc 1512 , most of the sources are best fit by model clusters that are 20 myr or younger , and have low extinctions , @xmath119 mag . in ngc 5248 ,
the youngest and brightest sources are somewhat redder , with @xmath120 mag , as already concluded on the basis of color - color diagrams .
the remaining sources cover a larger range in age than in ngc 1512 , but here too , almost all have extinctions @xmath119 mag , and ages 40 myr or younger .
there is , however , a significant fraction of clusters in both galaxies that are best fit with older models , of ages @xmath121 myr . at first sight , this result is contrary to that expected , if the starbursts in the rings were continuous .
the clusters fade by a factor less than 10 over 300 myr ( see fig .
13 ) , and the massive ones thus remain well within our dynamic range for this entire age range . in a continuous burst scenario , one would therefore naively expect many more clusters to have large ages , and one would conclude from the figures that the starbursts in each of these galaxies are episodic events which last only several tens of myrs .
however , we show here that the scarcity of old clusters may be the result of an age bias which causes young clusters to be over - represented at every luminosity .
many studies of sscs in various environments have shown that the clusters follow a steep luminosity distribution , @xmath122 , with @xmath123 ( e.g. , meurer et al .
1995 ; barth et al . 1995 ; maoz et al . 1996 ; whitmore et al .
1999 ; carlson et al .
figure 21 shows the extinction - corrected @xmath8-band luminosity functions for the clusters in ngc 1512 and ngc 5248 , separated into young ( @xmath124 myr ) and old ( @xmath125 myr ) categories .
each cluster s age and extinction are based on its best - fit model .
we have also plotted on the diagrams suitably normalized @xmath126 power laws .
the figures show that the cluster @xmath8 luminosities in the two galaxies , when considering only the brightest clusters that are less susceptible to incompleteness , are consistent with an @xmath127 distribution , as has already been demonstrated for their 2300 luminosities by maoz et al .
more importantly , the figures show that the individual `` young '' and `` old '' categories appear to follow the same distributions , although the small number of old clusters makes this statement tentative . recalling the relatively small , but significant , fading of clusters with age , the evolution of the young stellar population can be simply described by a horizontal shift to the left of the luminosity functions shown in logarithmic coordinates in figure 21 .
because of their steep slopes , such effective age correction would make the old clusters intrinsically more abundant than the young ones , even for a luminosity shift by a factor of only a few .
thus , the fact that one sees relatively so many young clusters in these galaxies is not necessarily because the entire starburst is young .
instead , it could be the result of the fact that there are many more low - mass clusters than high - mass clusters . when the low - mass clusters are young , the moderate rise to a higher luminosity makes them more abundant than the older , higher - mass clusters that have the same luminosity .
the relative numbers of young and old clusters are thus consistent with continuous star formation .
if , on the other hand , the star formation were episodic , we would expect few or no old clusters to be detected at all .
turning around the question of a continuous versus an episodic starburst , we can then ask whether an episodic starburst can be ruled out .
one must recall that the best - fit models are not the _ only _ models that can fit each observed sed .
even with the considerable leverage afforded by the large wavelength range of our data , there is still degeneracy in assigning ages and extinctions to some clusters .
figure 22 shows , for every well - observed sed , the youngest model that is marginally consistent ( @xmath128 per degree of freedom ) .
these models are also listed in tables 5 and 6 .
the plots show that , if pushed , the data allow all but a handful of sources to be modeled as young clusters .
it is thus possible that there are actually few old clusters in the galaxies .
this could come about if the starbursts in the rings occur in @xmath5 myr - long episodes , with quiescent intervals of @xmath121 myr , or more .
the same phenomenon could also occur if most of the clusters do not survive for longer than @xmath5 myr . for completeness , tables 5 and 6 also list , for each well - observed sed , the _
oldest _ model that is marginally acceptable .
one sees that , in this case , there are many clusters that can not be older than a few myr , as already concluded based on the color - color diagrams . a total cluster mass can be associated with a model sed with a particular luminosity normalization .
the total mass is that which is reached asymptotically by the decaying burst of star formation , but without accounting for mass loss from the cluster due to stellar winds and supernovae , which is theoretically difficult to predict .
the total mass in an observed cluster is necessarily uncertain when derived based on photometric ( rather than kinematic ) data .
most of the mass is in low - mass stars on the main sequence , which contribute negligibly to the light at any band .
the total mass would increase by a factor of @xmath129 or decrease by a factor of @xmath130 if the low - mass cutoff were at @xmath131 or @xmath132 , respectively , instead of at @xmath46 . for young ( @xmath133 myr ) clusters ,
the total mass would increase by about 50% if the upper - mass cutoff were at @xmath134 , rather than at @xmath135 , because in the latter case , the very massive stars briefly dominate the luminosity , and fewer intermediate - mass stars are needed to explain the observed flux .
thus , compounding the theoretical uncertainties with the observational errors ( e.g. , the luminosity , and hence the mass , scales with the distance squared , and the distances to these galaxies are poorly known ) , the total masses we will derive below are uncertain to at least an order of magnitude , although this uncertainty may involve a systematic shift for all clusters , if all are similar in their star - formation parameters .
recently , zhang & fall ( 1999 ) have used age and extinction estimates to show that , in the merging galaxies ngc 4038/9 , the mass function of the clusters follows a power law similar to the luminosity function , @xmath136 , over the range @xmath137 .
the mass function is important for understanding the relation between sscs and old globular clusters .
our model fits to the clusters in ngc 1512 and ngc 5248 assign masses to each cluster , and allow extending this type of analysis to two more galaxies and a different starburst environment .
tables 5 and 6 list the masses for each of the fitted models .
figures 23 shows the mass functions for the clusters , based on the best - fit models . although we are limited by incompleteness at low masses ( low - mass clusters are faint ) and small numbers at high masses ( high - mass clusters are rare ) , the data are consistent with the same power - law mass distribution seen in ngc 4038/9 . the few , most massive , clusters in ngc 1512 and ngc 5248 have masses of @xmath7 .
zhang & fall ( 1999 ) found several tens of clusters in ngc 4038/9 with masses between @xmath138 and @xmath96 .
however , in making this comparison , two points need attention . first , the models of zhang & fall are generally similar to ours , but assume an imf with a low - mass cutoff of @xmath131 , and hence their mass estimates are systematically higher by a factor of 2 than ours .
second , ngc 4038/9 is a more extended starburst than those in the nuclear rings , and therefore includes more clusters zhang & fall derive masses for about 1600 clusters , whereas we do so for only about 300 in each galaxy .
given the steeply falling mass function , there is a low probability of finding in the ring galaxies clusters with the same masses as the few most massive ones in the merging galaxies . after accounting for these differences ,
the fraction of the cluster population above a given mass is similar in all three starburst systems . while some early studies with of sscs in starburst environments claimed that the clusters were resolved , with sizes of order 10 pc ( e.g. , whitmore & schweizer 1995 ) , recent studies of galaxies at @xmath36 mpc distances
have generally concluded that the clusters are unresolved , or marginally resolved with the pc on wfpc2 , corresponding to sizes of just a few pc or less .
specifically for the two galaxies examined here , maoz et al .
( 1996 ) examined the radial profiles of the clusters in aberrated foc f220w images .
they concluded that the flux profiles , if modeled as gaussians convolved with the psf , have a dispersion ( @xmath139 ) near or below the resolution limit , corresponding to a gaussian radius of @xmath140 pc .
we re - examine the size question in the present , post - refurbishment , images of the clusters in ngc 1512 and ngc 5248 , now taken in optical and ir bands .
we have carried out two tests to gauge the degree to which the sources are resolved .
first , we studied the azimuthally averaged radial profile of the brighter sources , as in maoz et al .
( 1996 ) , but starting from a radius of 1 pixel and outwards .
we attempted to fit the radial profiles with those of artificial psfs produced by tiny tim ( see 2 ) , or with those observed for some of the brighter isolated sources in the images themselves ( to see if there is a range of source sizes in the data ) .
we find that , as in the uv , the sources seem to be unresolved or marginally resolved in the optical and ir .
the result is not surprising in the ir , since the angular resolution is considerably worse than in the optical .
second , we compared the ratio of counts within 1-pixel and 3-pixel radii that we obtained for the clusters in the f547 m band ( which has the best angular resolution ) to the ratio obtained for artificial psfs .
various authors ( e.g. , barth et al .
1995 ; whitmore et al .
1999 ) have used this so - called @xmath141 test , or similar indicators , as a way of measuring sizes , with @xmath141 significantly larger than that of the psf indicating resolved sources .
tables 3 and 4 list @xmath141 for all sources .
we find that the tiny tim psf for f547 m has @xmath141=0.79 mag , in good agreement with the barth et al .
( 1995 ) value of 0.83 mag . to study the effect of the pc pixelation on @xmath141
, we created also a tiny tim psf in which the position of the source was on the corner of a pixel .
( this is not a standard tiny tim option , and required creating a subsampled psf , shifting and rebinning it , and convolving it with a kernel that simulates charge diffusion . ) for this psf , @xmath142 mag . to get a direct empirical evaluation of @xmath141 , one would ideally measure it for obvious foreground stars appearing on the same exposures , but there are only one or two such stars in these high galactic latitude exposures .
instead , we measured @xmath141 for the 100 brightest stars in an f547 m image of the globular cluster ngc 6441 .
the total counts for these stars are comparable to those of the brightest sources in the two galaxies , and we used only the less crowded half of the pc1 image of the globular cluster , where the crowding is similar to that in the circumnuclear rings .
figure 24 compares the distribution of @xmath141 for the globular cluster stars with the values for the 10 brightest clusters in each galaxy . also shown , with filled symbols ,
are the two measurements based on ideal artificial psfs .
the filled triangle represents a red , isolated , @xmath143 mag object in ngc 1512 ( pc1 coordinates 664 , 625 ) , possibly a foreground star . a similarly red object ( not plotted in fig .
24 ) is found in ngc 5248 ( pc1 coordinates 425 , 385 ) , in a high surface - brightness region , and has @xmath144 .
while the distribution of @xmath141 for the stars indeed peaks near 0.8 , there is a tail to higher values of @xmath141 , which encompasses many of the sscs .
we have repeated this experiment with the f547 m image of another globular cluster , ngc 5139 , which has a less crowded stellar field . in this case
, there appears to be less of a high-@xmath141 tail , but the number of stars is small .
examination of the cases with @xmath145 in the first globular cluster shows that these are stars projected close to other stars of comparable brightness , confirming that source crowding can have a major influence on @xmath141 .
the comparison with the globular clusters is not straightforward .
the globular clusters were observed about two years before the galaxies , when the telescope focus may have differed .
their exposures are only 80 s and 50 s , much shorter than those of the galaxies , and may hence suffer less from image - degrading effects such as jitter .
the stellar crowding in the globular clusters differs in its details from the crowding among the sources in the galaxies , which are also superimposed on a complex diffuse background . given these uncertainties
, we conclude that the bright clusters in ngc 1512 and ngc 5248 which have @xmath146 are possibly resolved , barring some of the possible systematics we can not address . for the remaining bright clusters , it is unclear whether the @xmath141 values , which are only slightly larger than those of isolated psfs , are due to crowding or to truly resolved sizes .
we suspect that various reports of resolved sscs in other starburst galaxies ( e.g. , carlson et al .
1999 ) based on comparison of @xmath141 analogs to artificial psfs , neglecting the various systematic effects , may have overestimated the ssc sizes . some of these other galaxies are significantly more distant than those studied here ( e.g. , @xmath147 mpc for ngc 3597 in the carlson et al .
1999 study ) . at such large distances
, there is a larger chance of size errors due to crowding and confusion , especially since the brighter clusters tend to be spatially correlated . in the nearest examples , such as ngc 1569 ,
the sscs appear extremely compact as well , suggesting a distance - related bias .
since our tests indicate that the sources are unresolved or marginally resolved , we will not attempt further to derive physical sizes .
instead we note that , assuming a generic distance of 10 mpc , the gaussian radii of all the discrete sources are @xmath140 pc .
we have analyzed broad - band and narrow - band images of two circumnuclear ring galaxies , obtained with the fine angular resolution of _ hst _ , and spanning a large continuum wavelength range , from 0.23 to 1.6 @xmath2 m , and two main emission lines , and .
the data provide a uniquely detailed look into the properties and interrelations of the gas , the dust , and the compact clusters .
our main findings are as follows .
there is only a loose spatial correlation between the observed young clusters and the ionized gas in the rings . in some h ii regions , no continuum sources
are seen , and the presence of an embedded young population is revealed only by means of the high equivalent width . in others ,
the ionized gas is in shells or bubbles surrounding complexes of clusters , which are apparently in the process of destroying their birth clouds .
the /ratios of the line - emitting gas generally indicate only moderate extinction by a foreground screen , or , more realistically , an optically thick mixture of gas and dust .
the latter option would explain the severe extinction , even at ir wavelengths , of the light from the young stars embedded in many of the clouds .
ionizing photon budget estimates also suggest that a large fraction of the ionizing stellar population is hidden .
unfortunately , because the line emission is probably also strongly extinguished by dust , it is difficult to say what this fraction is . comparison to other imaging studies of nearby starbursts , whether in rings or other configurations ( e.g. , whitmore et al .
1999 ; buta et al .
2000 ; lancon et al . 2001 ) , reveals a recurring pattern , whereby the continuum and line emission come from distinct sites .
although the visibility of the continuum sources depends on their ages ( it takes time for them to break out of their clouds ) , there seems to be a strong element of geometry , orientation , and variation from galaxy to galaxy in determining what fraction of the clusters in a starburst is visible .
for example , all the data point toward a generally dustier environment in ngc 5248 than in ngc 1512 .
this suggests that , on an individual galaxy basis , it is difficult to construct a coherent picture of stellar populations , gas content , and dust properties , based on measurements integrated over large areas of a starburst ( as is unavoidable when angular resolution is limited , or a galaxy is distant ) . this situation has led in the past to challenges in explaining self - consistently various subsets of spatially integrated measurements , such as the uv spectral slope , the /h@xmath88 intensity ratio , and the ir luminosity , both in nearby starbursts ( e.g. , calzetti , kinney , & storchi - bergmann 1994 , 1996 ; meurer et al .
1995 , 1997 ) , and in distant ones ( e.g. , pettini et al . 1998 ; meurer , heckman , & calzetti 1999 ) .
the common explanations have been that starbursts have a relatively grey extinction curve , or that the young stars and the gas have different spatial distribution .
takagi , arimoto , & vansevicius ( 1999 ) and gordon et al .
( 2000 ) have addressed this problem by considering also ir and sub - mm measurements , compared with stellar synthesis models that include the effects of mixed dust in various geometries .
charlot & fall ( 2000 ) have extended those studies by accounting for absorption by various interstellar components , and limiting the lifetimes of some of the components as a result of their destruction by the new stars .
they find they can reproduce the mean properties of starburst galaxies if the extinction law is grey , or if the extinction is steep but the dust has a patchy distribution .
the data we have presented clearly support the latter picture .
we have measured the properties of the compact clusters in the rings , and analyzed them by means of traditional color - color diagrams , but also using explicit model fitting that exploits all the available information over the five continuum bands we have sampled .
this improves our ability to separate the effects of age and extinction .
given the observational errors and the remaining age - extinction degeneracy in some cases , most of the seds are well fit by a range of spectral synthesis models .
however , some of the brightest clusters reveal an ir excess that is not reproduced by any model . in most objects ,
the excess shows up only in one band , at @xmath35 m , and hence it is difficult to constrain its properties . in one object , where the ir excess is particularly large , and there is some excess emission in the @xmath21 band as well , we have shown that the excess can be fit with a roughly 2000 k thermal spectrum .
since the bulk of the stellar radiation is from stars with effective temperatures of @xmath148 k , if the excess is the result of thermal reradiation by dust , the dust particles will be at no more than about 100 stellar radii , i.e. , they are circumstellar dust shells .
this result needs to be confirmed with additional observations , especially at longer ir wavelengths , which would probe the shape of the excess spectrum .
we find that the brightest clusters in both galaxies have ages of a few myr , and undergo only mild extinction .
the typical extinctions are @xmath111 mag in ngc 1512 and @xmath112 mag in ngc 5248 .
the extinction of clusters thus appears to be an `` on / off '' phenomenon , with the clusters either completely hidden in their birth clouds ( and revealed only via the strong line emission from the clouds ) , or almost completely exposed .
this shows that the stellar and supernova winds which presumably clear out the gas are extremely efficient in clearing out the gas on timescales @xmath69 myr .
the seds are well fit when assuming a steep , galactic extinction law , consistent with the conclusion , above , regarding the form of the extinction law .
our best - fit models to the observed cluster seds suggest the presence also of older clusters , aged up to 300 myr . however , given the uncertainties , almost none of these clusters can be ruled out from instead being 20 myr or younger .
the clusters have a @xmath8-band luminosity function of the form @xmath149 , similar to that previously measured in the uv for these galaxies and to what is generally found for clusters in starburst environments .
we have pointed out an `` age bias , '' where the steep luminosity function and the fading of clusters with time combine to make young clusters appear overabundant .
the small fraction of old clusters indicated by the best - fit models is consistent with expectations if the starbursts in the rings are actually continuous , rather than episodic .
evidence that other ring starbursts are episodic , rather than continuous , has been presented before , using similar age dating arguments ( e.g. , smith et al .
1999 ; prez - ramrez et al . 2000 ) .
buta et al .
( 2000 ) have emphasized the possible detection of _ some _ old ( @xmath150 myr ) clusters in the ring in ngc 1326 . from the theoretical perspective , elmegreen ( 1994 )
has shown how a long period of gas accretion onto the region of the ring can lead to a synchronized starburst when a critical gas density is reached , and which may last only briefly , until the accreted gas is depleted .
we caution that the age bias needs to be taken into account when deducing the star - forming history of a galaxy from its cluster age distribution . in the case of the galaxies analyzed here , because we can not rule out the near absence of old clusters , we can not exclude , based on the cluster ages , either of the possibilities , continuous or episodic , or the possibility that clusters are destroyed before they have a chance to age
. a statistical argument against episodic star formation can , however , be made .
the uv imaging survey of maoz et al .
( 1996 ) included an unbiased selection of 110 nearby galaxies from a complete sample . in the observed sample ,
71 galaxies were spirals of hubble type sc or earlier , and 52 were classified as barred ( sb or sab ) by de vaucoulers et al .
( 1991 ) . in this sample ,
maoz et al . found five examples of uv - bright circumnuclear rings .
accounting for incompleteness due to the limited field of view raised this number to seven , or about 10% of the sc , and earlier , spirals .
if circumnuclear starbursts are episodic , the models with the lowest acceptable ages suggest that the uv - bright phase lasts 10 - 20 myr , while the interval between successive bursts is @xmath151 myr .
this would mean that most , or all , early - type spiral galaxies undergo occasional circumnuclear starbursts .
while this possibility is not completely outlandish , it is contrary to the idea that circumnuclear rings require special conditions that are found only in some galaxies , which have strong bars and lindblad resonances that channel gas accretion onto the rings ( e.g. , elmegreen 1994 ) . the extensive broad - band data presented here , combined with our modeling , allow us to derive approximate masses and mass functions for the clusters
the mass function in the two galaxies is consistent with the @xmath152 form that has been found in the only other starburst system , the antennae galaxies , where this has been reliably measured ( zhang & fall 1999 ) .
this power - law distribution is distinct from the log - normal distribution observed for old globular clusters , which has a characteristic mass scale .
if sscs are the progenitors of globular clusters , some evolutionary process , such as evaporation or tidal disruption by galactic gravitational fields , must reshape the mass functions ( elmegreen & efremov 1997 ; fall & zhang 2001 ) . the most massive clusters in ngc 1512 and ngc 5248 have masses in the range @xmath95 to @xmath153 .
after accounting for statistics and differences in model parameters , this is similar to the masses in the antennae .
combined with the other ssc characteristics we have evaluated , there is a remarkable uniformity in the properties of young clusters in these two different starburst environments .
we have estimated the sizes of the brighter , isolated sscs , and argued that in some cases they are only marginally resolved , or unresolved , corresponding to gaussian radii smaller than a few pc .
some previous size measurements of sscs in other galaxies have probably not taken full account of the systematics which can artificially broaden the image of a point source .
it is also possible that sscs in different environments have different sizes .
it is easy to show ( e.g. , maoz et al .
1996 ) that the virial radius , @xmath154 , the mass , @xmath155 , and the age , @xmath156 , together determine whether or not a cluster is gravitationally bound , where , for a bound cluster , @xmath157 the parameters we have derived for most of the brighter clusters establish them as bound objects , unless our mass estimates are systematically high due to , e.g. , an extremely top - heavy stellar imf .
visual - band size measurements , such as above , are also an essential ingredient in the more robust kinematic estimates of ssc masses based on optical spectroscopy ( ho & filippenko 1996a , b ; smith & gallagher 2000 ; sternberg 1998 ) . 1 .
the young star clusters and the line - emitting gas have different spatial distributions .
the gas is mixed with dust , making it optically thick , and effective at completely obscuring some of the young clusters , which are revealed only by the line emission from the gas .
most of the clusters that are visible are only mildly reddened , suggesting that the processes that clear out the gas and dust are efficient and fast .
some of the brightest young clusters have an ir excess that may be thermal reradiation by circumstellar dust .
4 . the cluster seds are consistent with a range in ages , from 1 myr to 300 myr .
although only a minority appears to be old , an age bias that causes young clusters to be over - represented makes this fraction consistent with continuous star formation in the rings over the past @xmath4 myr . due to the uncertainties in dating the clusters , we can not rule out episodic , @xmath5 myr - long bursts of star formation , although the statistics of the occurrence of uv - bright nuclear rings in spiral galaxies argue against this possibility .
a third possibility is that old clusters are , in fact , absent , but due to the destruction of clusters after @xmath5 myr .
the sizes of the clusters are of order a few pc , or less .
considering also the range of ages and the large masses we have established for the clusters , this confirms that they are gravitationally bound objects .
the luminosity functions and the mass functions follow a power - law distribution with index @xmath6 , as seen in other starburst environments .
the lack of a mass scale means that subsequent evolution of the mass function is required , if some of the sscs are to evolve into globular clusters .
there is remarkable similarity in the properties of the gas , dust , and clusters in different starbursts that have been studied in detail , whether in circumnuclear rings or in merging galaxies .
8 . in ngc 5248 , there is a previously unknown , inner , emission - line ring of radius 60 pc . in ngc 1512
, we have found a peculiar compact emission - line source with @xmath10 + [ n ii ] equivalent width , which may be a young , balmer - dominated supernova remnant .
we thank a. pauldrach and r .-
kudritzki for providing us with their model atmospheres in advance of publication , and the anonymous referee for constructive comments .
this work was funded in part by grants go-6738 and go-7879 from the space telescope science institute , which is operated by aura , inc .
, under nasa contract nas 5 - 26555 .
d. m. and a. s. acknowledge support by a grant from the israel science foundation .
research by a. j. b. is supported by a postdoctoral fellowship from the harvard - smithsonian center for astrophysics .
a. s. acknowledges support by german - israeli foundation grant i-551 - 186.07/97 .
a. v. f. and l. c. h. acknowledge funding from nasa grant nag5 - 3556 .
wfpc2 & f336w & `` @xmath20''&@xmath158&2500&05/03/99&2300&30/01/99@xmath159 pix@xmath19 & f547 m & `` @xmath8''&@xmath160&900 & 05/03/99 & 900&30/01/99 & f814w & `` @xmath21''&@xmath161&1100&05/03/99&1100&30/01/99 & f658n & h@xmath22+[n ii]&@xmath162&5200&05/03/99&4800&30/01/99nicmos-2&f160w & `` @xmath1''&@xmath163&2560&29/07/98&2304&24/08/98@xmath164 pix@xmath19 & f187w & pa@xmath22 cont.&@xmath165&2560&29/07/98&2304&24/08/98 & f187n & pa@xmath22&@xmath166&2560&29/07/98 & & & f187n & pa@xmath22&@xmath167&&&2304&24/08/98 1&446.&549 . & 5.53 & @xmath1684.77&@xmath16968 & 179.6&44.4 & 31.0 & 4.0&143 2&405.&534 . & 3.66 & @xmath1685.44&@xmath16968 & 255.2&50.5 & 59.2 & 5.1 & 85 3&352.&515 . & 1.25 & @xmath1686.32&@xmath16968 & 289.9&58.8 & 39.5 & 4.9&149 4&299.&450 . & @xmath1682.49 & @xmath1685.58&@xmath16945 & 91.8&24.1 & 23.3 & 3.8&103 5&275.&434 . &
@xmath1683.80 & @xmath1685.72&@xmath16945 & 88.7&24.1 & 21.9 & 3.7&110 6&263.&352 . & @xmath1686.60 & @xmath1683.20&@xmath16923 & 16.9 & 9.8 & 3.6 & 1.7&271 7&324.&243 . & @xmath1687.64 & 2.39&@xmath16923 & 18.5 & 4.5 & 1.7 & 4.1&268 8&361.&234 . & @xmath1686.61 & 3.78&@xmath16945 & 70.3&24.1 & 7.7 & 2.9&313 9&574.&394 . &
5.50 & 4.38&@xmath16945 & 69.2 & 8.3 & 17.6 & 8.4 & 4710&577.&428 .
& 6.59 & 3.27&@xmath16945 & 46.5 & 8.3 & 13.2 & 5.6 & 6311&540.&536 .
& 8.44 & @xmath1681.58&@xmath16945 & 51.0&12.1 & 9.4 & 4.2&128&&&&&&&*ngc 5248 * & & & 1&310.&449 .
& 3.25 & 4.61&@xmath16932 & 30.6&10.7 & 9.6 & 2.9&111 2&333.&421 .
& 2.97 & 2.99&@xmath16918 & 16.9 & 8.8 & 5.9 & 1.9&149 3&277.&421 .
& 5.19 & 4.24&@xmath16918 & 1.2 & 4.9 & 1.3 & 0.2&361 4&269.&411 .
& 5.73 & 4.03&@xmath16918 & 1.9 & 6.8 & 1.3 & 0.3&515 5&272.&392 .
& 6.04 & 3.21&@xmath16918 & 17.0 & 6.8 & 1.9 & 2.5&349 6&342.&375 .
& 3.65 & 0.96&@xmath16927 & 35.2 & 8.8 & 18.6 & 4.0 & 47 7&329.&359 .
& 4.52 & 0.62&@xmath16932 & 35.8 & 9.7 & 23.9 & 3.7 & 41 8&342.&337 . & 4.50 & @xmath168.54&@xmath16923 & 27.1 & 3.9 & 14.3 & 7.0 & 27 9&228.&285 . & 10.18 & @xmath168.05&@xmath16918 & 7.8 & 6.8 & 0.7 & 1.1&94310&310.&289 . & 6.84 & @xmath1681.72&@xmath16918 & 12.1 & 4.9 & 3.3 & 2.5&14611&346.&285 . & 5.50 & @xmath1682.69&@xmath16927 & 17.5&11.7 & 13.1 & 1.5 & 8912&386.&277 . & 4.10 & @xmath1683.90&@xmath16945 & 104.7&25.3 & 29.1 & 4.1 & 8713&471.&266 . & 0.98 & @xmath1686.24&@xmath16932 & 21.8&13.6 & 8.0 & 1.6&17014&509.&346 .
& @xmath1682.32 & @xmath1683.93&@xmath16991 & 358.0&72.0&118.9 & 5.0 & 6115&543.&406 . &
@xmath1685.01 & @xmath1682.31&@xmath16936 & 33.8&20.4 & 16.8 & 1.7&12116&580.&416 . & @xmath1686.70 & @xmath1682.75&@xmath16936 & 8.8&21.4 & 3.8 & 0.4&56617&531.&481 . & @xmath1686.21 & 0.93&@xmath16936 & 45.2&11.7 & 20.3 & 3.9 & 5818&464.&489 . & @xmath1683.74 & 2.74&@xmath16923 & 21.8 & 6.8 & 11.7 & 3.2 & 5819&444.&504 . & @xmath1683.28 & 3.79&@xmath16923 & 18.7 & 4.9 & 10.9 & 3.8 & 4520&387.&564 . & @xmath1682.37 & 7.44&@xmath16918 & 13.6 & 3.9 & 2.1 & 3.5&18621&528.&561 . & @xmath1687.89 & 4.16&@xmath16918 & 33.6 & 6.8 & 8.5 & 4.9 & 80 rrrrrcccccccccccc 1 & 407.42 & 534.64 & 3.76 & -5.39 & 19.10 & 0.98 & 22000.0 & 4400.0 & 2957.5 & 149.8 & 755.9 & 38.9 & 237.0 & 12.4 & 39.9 & 4.0 2 & 350.21 & 512.03 & 1.10 & -6.26 & 20.22 & 1.14 & 3450.0 & 690.0 & 1212.1 & 61.6 & 269.3 & 14.0 & 85.2 & 4.6 & 13.8 & 1.2 3 & 444.00 & 549.25 & 5.47 & -4.84 & 20.31 & 1.05 & 1650.0 & 330.0 & 868.9 & 44.1 & 248.2 & 13.0 & 86.4 & 4.8 & 15.7 & 1.8 4 & 277.43 & 434.07 & -3.71 & -5.65 & 20.68 & 1.15 & 1360.0 & 272.0 & 555.7 & 30.6 & 176.4 & 9.7 & 70.8 & 4.1 & 14.4 & 2.5 5 & 398.78 & 528.47 & 3.28 & -5.43 & 20.74 & 1.33 & 670.0 & 134.0 & 458.1 & 28.2 & 167.2 & 10.2 & 79.7 & 5.0 & 59.0 & 4.3 6 & 250.51 & 414.17 & -5.23 & -5.74 & 21.07 & 0.86 & 210.0 & 42.0 & 177.6 & 9.6 & 122.9 & 6.6 & 62.5 & 3.3 & 16.5 & 1.6 7 & 405.49 & 531.97 & 3.62 & -5.35 & 21.19 & 3.09 & @xmath170 & & 488.4 & 29.8 & 110.3 & 8.6 & 50.9 & 4.1 & 22.6 & 2.9 8 & 444.72 & 545.52 & 5.38 & -4.69 & 21.26 & 1.57 & 1000.0 & 200.0 & 432.6 & 23.1 & 103.4 & 6.4 & 35.6 & 3.0 & 14.4 & 1.6 9 & 548.57 & 317.11 & 2.37 & 6.33 & 21.38 & 1.06 & 270.0 & 54.0 & 194.2 & 10.8 & 92.0 & 5.4 & 39.4 & 2.1 & 8.9 & 0.8 10 & 573.71 & 392.75 & 5.45 & 4.41 & 21.51 & 1.52 & 610.0 & 122.0 & 266.2 & 14.2 & 82.3 & 4.8 & 30.3 & 2.2 & 8.6 & 0.9 11 & 541.50 & 526.25 & 8.21 & -1.20 & 21.64 & 1.27 & 300.0 & 60.0 & 191.7 & 10.7 & 72.9 & 4.4 & 28.5 & 1.7 & 7.9 & 0.9 12 & 257.85 & 419.39 & -4.82 & -5.71 & 21.78 & 0.81 & 200.0 & 40.0 & 186.7 & 11.1 & 64.1 & 4.6 & 24.2 & 2.2 & 4.1 & 1.4 13 & 258.43 & 323.01 & -7.61 & -2.32 & 21.82 & 1.42 & @xmath170 & & 64.4 & 4.8 & 61.5 & 3.8 & 34.1 & 2.1 & 11.7 & 1.1 14 & 306.08 & 446.24 & -2.36 & -5.25 & 21.84 & 1.50 & 200.0 & 40.0 & 132.4 & 8.9 & 60.7 & 3.7 & 27.1 & 1.6 & 6.5 & 1.3 15 & 307.29 & 447.80 & -2.27 & -5.26 & 21.89 & 1.57 & 370.0 & 74.0 & 146.3 & 8.7 & 57.8 & 3.4 & 23.5 & 1.4 & 3.4 & 1.3 16 & 269.34 & 423.21 & -4.31 & -5.51 & 21.95 & 1.05 & 190.0 & 38.0 & 119.2 & 9.3 & 54.6 & 4.6 & 27.0 & 2.5 & 21.9 & 2.0 17 & 353.49 & 512.40 & 1.23 & -6.18 & 21.98 & 3.14 & @xmath170 & & 214.9 & 13.0 & 53.2 & 4.2 & 25.3 & 2.3 & 9.6 & 1.0 18 & 272.16 & 427.69 & -4.08 & -5.58 & 22.09 & 1.50 & 240.0 & 48.0 & 151.2 & 10.8 & 48.0 & 4.2 & 15.1 & 2.1 & 14.7 & 2.0 19 & 300.79 & 446.65 & -2.53 & -5.41 & 22.12 & 1.04 & 320.0 & 64.0 & 185.1 & 10.8 & 46.6 & 3.8 & 14.0 & 1.3 & 2.5 & 1.2 20 & 256.38 & 321.62 & -7.72 & -2.33 & 22.14 & 1.60 & @xmath170 & & 72.5 & 5.1 & 45.9 & 3.1 & 27.8 & 1.6 & 9.2 & 1.0 21 & 296.04 & 454.57 & -2.46 & -5.83 & 22.14 & 1.05 & 270.0 & 54.0 & 135.5 & 8.3 & 45.7 & 3.5 & 16.4 & 2.4 & @xmath171 & 22 & 439.91 & 196.02 & -4.95 & 7.40 & 22.16 & 0.85 & @xmath170 & & 305.2 & 15.8 & 45.3 & 2.6 & 17.1 & 1.0 & @xmath171 & 23 & 244.98 & 404.33 & -5.71 & -5.56 & 22.16 & 1.01 & @xmath170 & & 60.5 & 4.1 & 45.1 & 2.9 & 18.8 & 1.2 & 7.7 & 1.4 24 & 439.87 & 534.85 & 4.90 & -4.45 & 22.18 & 1.29 & @xmath170 & & 60.0 & 6.7 & 44.2 & 3.6 & 34.6 & 2.5 & 30.7 & 1.8 25 & 280.84 & 451.00 & -3.10 & -6.15 & 22.20 & 1.12 & 90.0 & 18.0 & 94.2 & 6.8 & 43.5 & 3.5 & 15.0 & 1.8 & 4.8 & 1.5 26 & 593.89 & 475.05 & 8.55 & 2.12 & 22.20 & 1.46 & 140.0 & 28.0 & 74.1 & 5.3 & 43.5 & 3.0 & 23.9 & 1.6 & 11.9 & 1.2 27 & 443.00 & 534.75 & 5.01 & -4.36 & 22.21 & 1.59 & @xmath170 & & 35.1 & 5.8 & 43.1 & 3.7 & 33.3 & 2.3 & 25.9 & 1.9 28 & 402.32 & 540.55 & 3.76 & -5.75 & 22.25 & 1.64 & @xmath170 & & 125.0 & 7.9 & 41.5 & 3.3 & 18.0 & 1.8 & 13.7 & 1.2 29 & 603.75 & 721.68 & 16.07 & -6.22 & 22.31 & 1.15 & & & 19.7 & 2.3 & 39.2 & 2.3 & 28.8 & 1.5 & & 30 & 597.08 & 511.50 & 9.72 & 0.94 & 22.31 & 0.95 & @xmath170 & & 5.7 & 1.3 & 39.1 & 2.4 & 40.0 & 2.1 & 21.2 & 1.2 rrrrrcccccccccccc 1 & 527.45 & 560.56 & -7.86 & 4.16 & 18.26 & 0.91 & 2810.0 & 562.0 & 3510.9 & 176.4 & 1641.9 & 82.5 & 649.2 & 32.7 & 84.7 & 4.7 2 & 302.41 & 311.83 & 6.63 & -0.65 & 19.64 & 0.84 & 300.0 & 60.0 & 621.6 & 31.8 & 460.1 & 23.8 & 209.9 & 11.0 & 28.5 & 3.8 3 & 330.29 & 312.58 & 5.51 & -1.25 & 19.87 & 1.10 & 1110.0 & 222.0 & 806.3 & 42.3 & 370.5 & 19.8 & 168.2 & 9.4 & 108.4 & 7.7 4 & 280.13 & 324.35 & 7.23 & 0.34 & 20.15 & 0.97 & 530.0 & 106.0 & 522.4 & 27.0 & 287.8 & 14.8 & 129.9 & 6.7 & 44.9 & 3.4 5 & 411.89 & 262.85 & 3.39 & -5.04 & 20.29 & 1.38 & 660.0 & 132.0 & 496.2 & 28.4 & 252.9 & 15.7 & 132.0 & 8.4 & 89.8 & 7.6 6 & 498.44 & 339.01 & -1.74 & -3.97 & 20.41 & 1.18 & 520.0 & 104.0 & 442.7 & 23.4 & 225.0 & 12.0 & 102.9 & 5.8 & 30.3 & 2.3 7 & 513.45 & 348.56 & -2.55 & -3.92 & 20.48 & 1.25 & 1220.0 & 244.0 & 733.9 & 38.8 & 211.1 & 12.1 & 72.0 & 4.2 & 17.1 & 2.3 8 & 554.06 & 478.15 & -7.06 & 0.30 & 20.66 & 1.01 & 770.0 & 154.0 & 511.7 & 26.9 & 179.2 & 9.9 & 61.1 & 3.6 & 8.4 & 1.8 9 & 454.36 & 520.05 & -4.05 & 4.19 & 20.71 & 1.02 & 430.0 & 86.0 & 447.5 & 23.4 & 171.5 & 9.5 & 63.2 & 3.9 & 15.3 & 2.6 10 & 502.67 & 343.70 & -2.02 & -3.88 & 20.77 & 1.01 & 230.0 & 46.0 & 337.3 & 19.2 & 162.2 & 9.6 & 69.4 & 4.9 & 9.8 & 1.6 11 & 346.77 & 268.92 & 5.83 & -3.34 & 20.82 & 1.16 & 260.0 & 52.0 & 294.2 & 15.6 & 154.8 & 8.8 & 73.8 & 4.5 & 34.3 & 3.8 12 & 338.38 & 187.17 & 8.00 & -6.39 & 20.88 & 1.01 & 410.0 & 82.0 & 327.8 & 17.3 & 146.2 & 7.8 & 58.3 & 3.2 & 18.4 & 1.4 13 & 344.61 & 375.92 & 3.52 & 0.94 & 20.93 & 1.23 & 350.0 & 70.0 & 333.2 & 17.7 & 140.3 & 8.5 & 52.1 & 4.0 & 5.7 & 1.7 14 & 515.11 & 349.60 & -2.64 & -3.92 & 20.96 & 1.92 & @xmath170 & & 422.3 & 24.2 & 136.3 & 9.1 & 55.1 & 4.8 & 18.5 & 2.7 15 & 504.04 & 342.91 & -2.05 & -3.94 & 21.10 & 2.26 & 230.0 & 46.0 & 242.4 & 14.8 & 119.2 & 7.8 & 53.6 & 3.9 & 12.9 & 1.7 16 & 496.68 & 336.63 & -1.62 & -4.02 & 21.17 & 1.96 & @xmath170 & & 218.2 & 12.2 & 112.3 & 6.4 & 62.1 & 3.8 & 16.2 & 2.1 17 & 535.25 & 481.75 & -6.40 & 0.86 & 21.18 & 1.60 & 240.0 & 48.0 & 233.5 & 13.5 & 110.7 & 7.2 & 41.2 & 3.6 & 13.3 & 1.9 18 & 341.92 & 538.23 & -0.01 & 7.43 & 21.26 & 1.13 & 240.0 & 48.0 & 191.3 & 10.6 & 103.0 & 6.0 & 56.4 & 3.3 & 35.5 & 2.5 19 & 531.00 & 570.30 & -8.22 & 4.46 & 21.28 & 1.00 & 320.0 & 64.0 & 231.4 & 15.1 & 101.7 & 8.0 & 42.6 & 3.9 & & 20 & 332.91 & 538.85 & 0.33 & 7.66 & 21.29 & 1.24 & 190.0 & 38.0 & 161.3 & 9.2 & 100.6 & 5.5 & 59.2 & 3.3 & 27.0 & 2.2 21 & 525.62 & 347.69 & -3.01 & -4.23 & 21.35 & 1.06 & 140.0 & 28.0 & 191.8 & 12.8 & 94.8 & 7.8 & 39.3 & 4.0 & 26.5 & 3.7 22 & 341.11 & 318.68 & 4.94 & -1.25 & 21.39 & 1.50 & @xmath170 & & 150.5 & 10.9 & 91.4 & 7.6 & 46.9 & 4.9 & 23.2 & 4.9 23 & 280.47 & 340.22 & 6.86 & 0.97 & 21.42 & 1.28 & 220.0 & 44.0 & 158.7 & 8.9 & 89.1 & 5.2 & 49.3 & 3.0 & 16.5 & 3.1 24 & 342.77 & 364.74 & 3.84 & 0.54 & 21.45 & 1.42 & @xmath170 & & 136.4 & 14.4 & 86.5 & 12.0 & 35.1 & 6.0 & 29.8 & 2.7 25 & 341.16 & 374.71 & 3.68 & 0.97 & 21.48 & 1.55 & 160.0 & 32.0 & 227.8 & 13.2 & 84.3 & 6.3 & 36.5 & 3.4 & 9.5 & 2.0 26 & 342.99 & 374.29 & 3.62 & 0.91 & 21.50 & 2.42 & 350.0 & 70.0 & 206.2 & 12.4 & 82.9 & 6.7 & 40.0 & 3.9 & 11.8 & 1.8 27 & 443.86 & 379.88 & -0.50 & -1.13 & 21.55 & 2.03 & @xmath170 & & 67.9 & 5.8 & 79.0 & 8.1 & 52.4 & 6.9 & @xmath171 & 28 & 431.58 & 376.18 & 0.07 & -1.00 & 21.56 & 1.72 & @xmath170 & & 63.1 & 4.9 & 78.2 & 6.6 & 58.2 & 6.4 & @xmath171 & 29 & 495.80 & 498.68 & -5.22 & 2.42 & 21.59 & 1.34 & 230.0 & 46.0 & 177.0 & 10.1 & 76.5 & 5.2 & 31.8 & 2.7 & 9.9 & 1.7 30 & 407.68 & 264.10 & 3.53 & -4.90 & 21.59 & 1.49 & @xmath170 & & 90.7 & 13.9 & 76.4 & 9.3 & 55.7 & 4.8 & 60.5 & 5.9 rrcccccccccccc 1 & 5 & 1 & 0.0 & 9.08&99.59 & & & & & & & & 2 & 5 & 3 & 0.2 & 0.72 & 8.55 & & & & & & & & 3 & 5 & 1 & 0.6 & 3.72 & 4.45 & & & & & & & & 4 & 5 & 5 & 0.1 & 0.31 & 2.52 & & & & & & & & 5 & 5 & 10 & 0.0 & 0.51&15.42 & & & & & & & & 6 & 5 & 20 & 0.6 & 1.35 & 1.61 & 20 & 0.6 & 1.35 & 1.61 & 20 & 0.6 & 1.35 & 1.61 7 & 4 & 3 & 0.6 & 0.49&17.59 & & & & & & & & 8 & 5 & 1 & 0.4 & 1.35&13.17 & & & & & & & & 9 & 5 & 5 & 0.6 & 0.24 & 2.14 & & & & & & & & 10 & 5 & 5 & 0.1 & 0.14 & 7.27 & & & & & & & & 11 & 5 & 20 & 0.1 & 0.55 & 3.22 & & & & & & & & 12 & 5 & 1 & 1.0 & 1.36 & 0.35 & 1 & 0.9 & 1.12 & 1.26 & 3 & 0.9 & 0.29 & 0.45 13 & 4 & 20 & 1.0 & 0.94 & 1.29 & 7 & 0.5 & 0.13 & 1.80 & 100 & 0.0 & 1.18 & 1.36 14 & 5 & 20 & 0.2 & 0.48 & 1.25 & 5 & 0.5 & 0.14 & 1.63 & 20 & 0.2 & 0.48 & 1.25 15 & 5 & 5 & 0.2 & 0.11 & 0.53 & 5 & 0.2 & 0.11 & 0.53 & 5 & 0.2 & 0.11 & 0.53 16 & 5 & 12 & 0.1 & 0.22 & 6.09 & & & & & & & & 17 & 4 & 3 & 0.8 & 0.28&20.73 & & & & & & & & 18 & 5 & 20 & 0.0 & 0.35 & 9.92 & & & & & & & & 19 & 5 & 3 & 0.5 & 0.15 & 0.75 & 1 & 0.5 & 0.60 & 0.80 & 3 & 0.4 & 0.12 & 1.24 20 & 4 & 7 & 0.3 & 0.10 & 0.46 & 7 & 0.2 & 0.09 & 1.09 & 70 & 0.0 & 0.68 & 1.34 21 & 4 & 1 & 0.6 & 0.61 & 1.29 & 1 & 0.6 & 0.61 & 1.29 & 5 & 0.2 & 0.08 & 1.61 23 & 4 & 20 & 0.6 & 0.47 & 0.97 & 20 & 0.5 & 0.42 & 1.72 & 40 & 0.0 & 0.43 & 1.95 24 & 4 & 12 & 0.6 & 0.33 & 1.21 & 12 & 0.6 & 0.33 & 1.21 & 12 & 0.6 & 0.33 & 1.21 25 & 5 & 1 & 1.2 & 1.00 & 1.17 & 1 & 1.1 & 0.84 & 1.77 & 3 & 1.1 & 0.21 & 1.41 26 & 5 & 15 & 0.1 & 0.20 & 1.83 & 15 & 0.1 & 0.20 & 1.83 & 15 & 0.1 & 0.20 & 1.83 27 & 4 & 10 & 0.9 & 0.28 & 1.73 & 10 & 0.9 & 0.28 & 1.73 & 10 & 0.9 & 0.28 & 1.73 28 & 4 & 20 & 0.0 & 0.35&21.46 & & & & & & & & 30 & 4 & 300 & 2.2&11.03 & 2.04 & & & & & & & & rrcccccccccccc 1 & 5 & 1 & 1.3&41.85 & 0.27 & 1 & 1.3&41.85 & 0.27 & 3 & 1.2 & 8.88 & 1.53 2 & 5 & 1 & 1.9&19.58 & 2.52 & & & & & & & & 3 & 5 & 40 & 0.1 & 4.89&13.01 & & & & & & & & 4 & 5 & 20 & 0.5 & 2.93 & 1.58 & 20 & 0.5 & 2.93 & 1.58 & 20 & 0.5 & 2.93 & 1.58 5 & 5 & 40 & 0.2 & 3.72 & 8.59 & & & & & & & & 6 & 5 & 20 & 0.4 & 2.14 & 1.41 & 20 & 0.4 & 2.14 & 1.41 & 20 & 0.4 & 2.14 & 1.41 7 & 5 & 3 & 0.6 & 0.75 & 4.54 & & & & & & & & 8 & 5 & 1 & 0.8 & 2.96 & 0.44 & 1 & 0.8 & 2.96 & 0.44 & 3 & 0.7 & 0.62 & 1.48 9 & 5 & 1 & 1.1 & 3.83 & 2.05 & & & & & & & & 10 & 5 & 3 & 1.4 & 1.15 & 0.03 & 1 & 1.4 & 4.59 & 0.19 & 3 & 1.3 & 0.98 & 0.98 11 & 5 & 20 & 0.6 & 1.87 & 3.59 & & & & & & & & 12 & 5 & 20 & 0.3 & 1.30 & 3.44 & & & & & & & & 13 & 5 & 1 & 1.1 & 2.99 & 0.23 & 1 & 1.0 & 2.49 & 1.84 & 3 & 1.0 & 0.63 & 0.81 14 & 4 & 5 & 0.1 & 0.25 & 7.16 & & & & & & & & 15 & 5 & 5 & 0.8 & 0.37 & 1.92 & 5 & 0.8 & 0.37 & 1.92 & 5 & 0.8 & 0.37 & 1.92 16 & 4 & 7 & 0.0 & 0.18 & 0.72 & 7 & 0.0 & 0.18 & 0.72 & 30 & 0.0 & 1.03 & 1.94 17 & 5 & 20 & 0.4 & 1.03 & 4.11 & & & & & & & & 18 & 5 & 40 & 0.3 & 1.70 & 8.86 & & & & & & & & 19 & 4 & 5 & 0.5 & 0.23 & 1.11 & 5 & 0.5 & 0.23 & 1.11 & 20 & 0.2 & 0.79 & 1.99 20 & 5 & 30 & 0.4 & 1.32 & 1.76 & 30 & 0.4 & 1.32 & 1.76 & 40 & 0.4 & 1.72 & 1.90 21 & 5 & 20 & 0.6 & 1.13 & 5.91 & & & & & & & & 22 & 4 & 40 & 0.1 & 1.13 & 0.64 & 7 & 0.0 & 0.14 & 1.76 & 70 & 0.0 & 1.37 & 1.58 23 & 5 & 30 & 0.2 & 0.93 & 1.31 & 20 & 0.5 & 1.03 & 1.70 & 40 & 0.2 & 1.22 & 1.50 24 & 4 & 12 & 0.0 & 0.30 & 2.83 & & & & & & & & 25 & 5 & 3 & 1.2 & 0.53 & 2.02 & & & & & & & & 26 & 5 & 20 & 0.1 & 0.64 & 1.85 & 20 & 0.1 & 0.64 & 1.85 & 20 & 0.1 & 0.64 & 1.85 29 & 5 & 20 & 0.3 & 0.71 & 2.55 & & & & & & & & 30 & 4 & 12 & 0.7 & 0.60 & 2.53 & & & & & & & & | we present broad - band ( @xmath0 and @xmath1 ) and narrow - band ( + [ n ii ] and ) images of the circumnuclear starburst rings in two nearby spiral galaxies , ngc 1512 and ngc 5248 , obtained with the wfpc2 and nicmos cameras on _ hst_. combined with previously published ultraviolet ( uv ) images at 2300 , these data provide a particularly wide wavelength range with which to study the properties of the stellar populations , the gas , and the dust in the rings .
the young star clusters and the line - emitting gas have different spatial distributions , with some large ( 50-pc scale ) line - emitting regions that have little associated continuum emission , but a equivalent width indicating a few - myr - old embedded stellar population .
the observed / intensity ratios suggest the gas is mixed with dust , making it effective at obscuring some of the young clusters .
we identify the major ( about 500 in each galaxy ) compact continuum sources ( super star clusters and individual stars ) and analyze their spectral energy distributions ( seds ) from 0.2 @xmath2 m to 1.6 @xmath2 m by fitting them with a grid of spectral synthesis models with a range of ages and dust extinction .
most of the visible clusters are only mildly reddened , with @xmath3 to 1 mag , suggesting that the processes that clear out the gas and dust of the stellar birth clouds are efficient and fast .
the patchiness of the dust distribution makes it difficult to reliably estimate the star formation rate , based on uv continuum slope or hydrogen emission - line ratios , in starbursts such as these .
the cluster seds are consistent with a range in ages , from 1 myr to 300 myr , but with only a minority older than a few tens of myr .
we point out an age bias , the result of the steep luminosity function of the clusters combined with the fading of clusters as they age , which causes young clusters to be over - represented at any luminosity . accounting for this bias , the fraction of old clusters is consistent with continuous star formation in the rings over the past @xmath4 myr . due to the uncertainties in dating the clusters
, we can not rule out episodic , @xmath5 myr - long bursts of star formation , but the presence of uv - bright rings in about 10% of spiral galaxies argues against this possibility .
although most of the observed seds are well fit by a range of models , some of the brightest young clusters have excess emission in the ir that is not predicted by the models and may be thermal re - radiation by circumstellar dust .
the cluster mass functions follow a power - law distribution with index @xmath6 , similar to that recently derived for the starburst in the merging antennae galaxies , and extending to @xmath7 .
the lack of a mass scale means that subsequent evolution of the mass function is required , if some of the sscs are to evolve into globular clusters .
the clusters are spatially unresolved or marginally resolved , corresponding to @xmath8-band gaussian radii of less than a few pc , at an assumed distance of 10 mpc . in ngc 5248 ,
we report a previously - unknown , 60-pc - radius , inner emission - line ring , and in ngc 1512 , a peculiar compact ( @xmath9 diameter ) source with an + [ n ii ] equivalent width of @xmath10 , which may be a so - called balmer - dominated supernova remnant . |
null dust represents the high frequency ( geometrical optics ) approximation to the unidirectional radial flow of unpolarized radiation .
this is a reasonable approximation whenever the wavelength of the radiation is negligible compared to the curvature radius of the background .
various exact solutions of the einstein field equations were found in the presence of pure null dust ( for reviews see @xcite , and more recently @xcite and @xcite ) . in some scenarios
even gravitation behaves as null dust .
price @xcite has shown that a collapsing spheroid radiates away all of its initial characteristics excepting its mass , angular momentum and charge .
( this result is known as the _ no hair _ theorem . )
the escaping radiation then interacts with the curvature of the background being partially backscattered .
both the escaping and the backscattered radiation can be modeled by null dust @xcite as the curvature radius of the background is larger than the wavelength of the radiation .
letelier has shown that the matter source composed of two null dust clouds can be interpreted as an anisotropic fluid @xcite , giving also the general solution for plane - symmetric anisotropic fluid with two null dust components .
later letelier and wang @xcite have discussed the collision of cylindrical null dust clouds .
the collision of spherical null dust streams was discussed by poisson and israel @xcite .
their analysis yielded the phenomenon of mass inflation .
however , no exact solution in the presence of two colliding null dust streams with spherical symmetry was known until now .
recently date @xcite tackled this problem under the assumption of staticity , integrating part of the einstein equations .
it is the purpose of the present paper to present the exact solution for the case of two colliding spherically symmetric null dust streams in equilibrium , in a completely integrated form .
the plan of the paper is as follows . in sec .
ii we derive the field equations and we integrate them . the emerging exact solution is written explicitly in suitable coordinates adapted to spherical symmetry and staticity .
the metric in radiation coordinates and in double null coordinates is also given .
we analyze the metric both analytically and by numeric plots . in sec .
iii we present various possible interpretations of the solution , including the anisotropic fluid picture , and dilatonic gravity .
finally sec .
iv contains the analysis of the interior and exterior matching conditions on junctions along timelike and null hypersurfaces , respectively .
we employ the matching procedure of barrabs and israel @xcite .
the interior junction fixes the parameters of the solution in terms of physical characteristics of the central star : the mass and energy density on the junction .
the exterior matching with incoming and outgoing vaidya solutions @xcite leads to the conclusion that no distributional matter is present at the junction .
this feature is in contrast with the exterior matching with the schwarzschild solution @xcite .
the general form of a spherically symmetric , static metric in a spacetime with topology @xmath0 is @xmath1 here @xmath2 is time , @xmath3 is the curvature coordinate ( e.g. the radius of the sphere @xmath2=const with area @xmath4 ) , @xmath5 is the square of the solid angle element .
the functions @xmath6 and @xmath7 are positive valued .
we introduce the local mass function @xmath8 , related to the gravitational energy within the sphere of radius @xmath3 @xcite : @xmath9 the energy - momentum tensor in the region of the cross - flowing null dust is a superposition of the energy - momentum tensors of the incoming and outgoing components : @xmath10 all energy conditions are satisfied for @xmath11 .
the same linear mass density function @xmath12 was chosen for both components as staticity requires no net flow in either of the null directions .
the vector fields @xmath13 are the tangents to the ( future - oriented ) outgoing and incoming null congruences , partially normalized such that @xmath14 .
similarly as for the one - component null dust , here both null congruences are geodesics @xcite . after eliminating second time derivatives from the nontrivial einstein equations we have the system : @xmath15 the solutions with @xmath12=const all reduce to @xmath16 , meaning vacuum .
they are either the schwarzschild or the flat solution , in accordance with the birkhoff theorem . for @xmath17const one of the equations
is a relation between the mass density @xmath18 and the metric function @xmath19 : @xmath20 where @xmath21 is a constant .
this relation can be deduced also from the energy - momentum conservation@xcite .
the remaining equations do not contain the constant @xmath22 : @xmath23 inserting ( [ mass ] ) in ( [ eq1 ] ) the mass is found to increase with the radius : @xmath24 .
now we solve the system ( [ eq1]),([eq2 ] ) .
following date@xcite , we eliminate @xmath25 from ( [ eq1 ] ) by its expression taken from ( [ eq2 ] ) : @xmath26 the resulting second order ordinary differential equation in @xmath27 can be integrated , finding : @xmath28 here @xmath29 is an integration constant . inserting this expression in ( [ f ] ) ,
an algebraic relation between the metric components emerges : @xmath30 where @xmath31 .
then we complete the integration of the system ( [ eq1]),([eq2 ] ) .
the key remark is that none of the equations in this system contains explicitly the independent variable @xmath32 .
thus one can pass to the new independent variable @xmath12 , in terms of which an ordinary first order equation can be written : @xmath33 introducing the new positive variables @xmath34 the equation ( [ eq3 ] ) takes the form of a first order linear ( inhomogeneous ) ordinary differential equation : @xmath35 the solution is found by integrating first the homogeneous equation , then varying the constant .
it is @xmath36 where @xmath37 is a third integration constant .
the function @xmath38 in ( [ int2 ] ) can be expressed either in terms of the error function or in terms of the dawson function : @xmath39 for properties of these transcendental functions see@xcite . from eqs .
( [ int1]),([newvar ] ) and ( [ int2 ] ) both the curvature coordinate @xmath3 and the metric functions @xmath12 and @xmath25 are found as functions of the radial variable @xmath40 : @xmath41 } .
\label{fl}\end{aligned}\ ] ] then the mass function @xmath42 is obtained from ( [ mass ] ) : @xmath43 it is easy to check that both @xmath3 and @xmath44 are monotonously increasing functions of @xmath40 : @xmath45 \geq 0
. \label{dml}\end{aligned}\ ] ] in the last relation the equality holds for @xmath46 .
now we have everything together to write the metric in terms of the new radial coordinate @xmath40 : @xmath47 \frac{dl^{2}}{c^{2}}+\left [ -e^{l^{2}}+2l\phi _ { b}(l)\right ] ^{2}\frac{d\omega ^{2}}{c^{2}}. \label{metrictl}\ ] ] there are three parameters in the solution , two of them restricted to be positive : @xmath22 and @xmath48 . without loss of generality
we can choose @xmath49 by rescaling the time coordinate .
the parameter @xmath48 provides some distance scale .
we comment on the third parameter in what follows . both @xmath50 and the energy conditions @xmath51
imply @xmath52 valid for all admissible values of @xmath40 .
the equality @xmath53 holds for @xmath54 corresponding to @xmath46 . as @xmath55
, the function @xmath56 is monotonously decreasing and the inequality ( [ bl ] ) will be satisfied for any @xmath57 .
thus @xmath37 gives the lower boundary @xmath58for the range of the radial coordinate @xmath40 .
next we plot numerically the functions @xmath18 , @xmath6 and @xmath59 for different values of the parameter @xmath37 ( figs.1 - 3 ) . as we see from fig .
3 , the mass function vanishes at some radius @xmath60 and takes negative values at @xmath61 .
we show here that such a positive @xmath60 exists irrespective of the choice of the parameters @xmath37 and @xmath48 . by combining ( [ rl ] ) and ( [ ml ] ) with the condition @xmath62
we find @xmath63 .
the constant @xmath48 has a simple scaling effect on the value of @xmath60 .
figure 3(b ) shows that in the domain @xmath64 the negative mass region can be extended by increasing @xmath37 .
we emphasize that the energy conditions are still satisfied in the negative mass regions .
the interpretation of the negative mass function is not immediate and depends on the measuring procedure of the mass in this asymptotically non - flat space - time . at the end of this section
we give the metric in double null and radiation coordinates .
this requires an additional integration .
we introduce a `` tortoise coordinate '' @xmath65 in the same manner it can be introduced in the schwarzschild space - time : @xmath66^{1/2 } } = \frac{\sqrt{2}}{c}\int^{l}\left [ -e^{y^{2}}+2y\phi _ { b}(y)\right ] dy .
\label{tortoise}\ ] ] the radial null geodesics are @xmath67 , with @xmath68 for incoming and @xmath69 for outgoing geodesics . introducing the null coordinates @xmath70 the metric takes the simple form : @xmath71 the radial null geodesics are now @xmath72=const . the functions @xmath73 and @xmath74 are contained in implicit form in eqs .
( [ g]),([tortoise ] ) , ( [ rl ] ) and ( [ bel ] ) .
finally we cast our static , spherically symmetric solution of crossflowing null dust in either the incoming or outgoing radiation coordinates @xmath75 .
these coordinates are like the eddington - finkelstein coordinates for the schwarzschild solution . if the solution was asymptotically flat , they would be the bondi - sachs coordinates .
@xmath76 + r^{2}d\omega ^{2}. \label{metricef}\ ] ] here @xmath77 for @xmath78 . in these
coordinates the radial null geodesics are given by one of the equations @xmath79=const and @xmath80 now it is evident that there is no apparent horizon : @xmath81 thus no event horizon either .
thus the singularity in the origin @xmath46 is naked .
this is very similar to the naked singularity of the _ negative mass _ schwarzschild solution @xcite .
in this subsection we present a dilatonic model which in 4d has the interpretation of a spherically symmetric gravitational field in the presence of two crossflowing null dust streams .
this dilatonic model emerges from the action @xmath82 + \frac{1}{2}g^{\alpha \beta } \nabla _ { \alpha } \ln \rho \nabla _ { \beta } \ln \rho + \frac{2}{\rho } \right ] - \frac{1}{2}\int d^{2}x\sqrt{-g}g^{\alpha \beta } \nabla _ { \alpha } \varphi \nabla _ { \beta } \varphi\ .
\label{dilaton}\ ] ] the first term is the einstein - hilbert action reduced by spherical symmetry ( a surface term was dropped ) . the second term in ( [ dilaton ] ) represents a 2d massless scalar field in minimal coupling .
the minus sign assures that all 4d energy conditions are satisfied .
@xmath83 is the dilaton and @xmath84 is the scalar field .
the conformal flatness of the @xmath85 metric @xmath86 is manifest in the corresponding 4d line element ( [ metricdn ] ) . @xmath87 and @xmath88 $ ] are the covariant derivative and curvature scalar , respectively associated with the metric @xmath89 .
although the equations emerging from this model are quite similar to the corresponding equations in the callan - giddings - harvey - strominger ( cghs ) model@xcite , they could not be exactly integrated in general . for our purposes
we take only the equation emerging from variation of the scalar field , in double null coordinates : @xmath90 here commas denote derivatives .
the equation has the dalembert solution @xmath91 showing that the scalar field behaves like our matter source of crossflowing null dust .
the 4d interpretation of the particular solution with either leftmoving or rightmoving matter is the vaidya solution@xcite .
recently mikovi has given a solution @xcite for the case where both components are present , in the form of a perturbative series in powers of the outgoing energy - momentum component .
our static solution ( [ metricdn ] ) represents the first explicit exact solution for this model , when none of the null dust components are neglected . following letelier
@xcite we reinterpret the energy - momentum tensor ( [ en - mom ] ) as describing an anisotropic fluid with a pressure component equaling its energy density : @xmath92 a straightforward comparison with ( [ en - mom ] ) gives @xmath93 thus the solution represents an anisotropic fluid at rest with the energy density @xmath94 and _ radial _ pressure @xmath95 no tangential pressures in the spheres @xmath3=const are present .
the fluid is isotropic only about a single point , the origin .
the solution can be interpreted as the outer region of a radiating star , receiving radiation from the surrounding region either .
if equilibrium is achieved between the two components , we have the static solution ( [ metrictl ] ) .
this was the initial interpretation proposed by date @xcite .
we write the energy - momentum tensor of the static solution in the double null coordinate system @xmath96 . inserting the null covectors @xmath97 in the covariant form of ( [ en - mom ] ) and taking account of ( [ g ] )
the energy - momentum tensor becomes @xmath98 a superposition of two cross - flowing null dust streams with equal and _ constant _ mass density functions .
however one can freely rescale the null vectors @xmath99 and @xmath100 to have _ arbitrary _ mass density functions either .
after such a rescaling the mass density function of the incoming null dust depends only on the outgoing coordinate @xmath101 and viceversa , a property pertinent to the mass functions of the vaidya solution @xcite , characterized by the energy - momentum tensor @xmath102 @xmath103 is the mass function and the coordinate @xmath104 is outgoing for incoming radiation and incoming for outgoing radiation . in the next section
we will study the interior junction with a static star , and the exterior junction with incoming and outgoing vaidya solutions .
we discuss the junction with an interior solution with accent on the anisotropic fluid interpretation given previously .
a similar treatment was given in @xcite .
both our analysis and the one in @xcite reveals that the junction with a static interior matter can be done without a regularizing thin shell .
our treatment is more general , however .
we formulate the junction conditions for two generic static spherically symmetric space - times , following the standard darmois - israel junction procedure @xcite and we establish a constraint on the matter pressures implied by the matching conditions . an other improvement over @xcite
is due to the fact that we dispose of the exact solution ( [ metrictl ] ) , thus the explicit computation of the the matching conditions with an arbitrary particular interior becomes possible . an orthonormal basis is given by the vectors @xmath105 and @xmath106 defined by eqs .
( [ u ] ) and ( [ chi ] ) together with the spacelike vectors @xmath107 any spherically symmetric static energy - momentum tensor has the form @xmath108 by inserting @xmath109 and @xmath94 in the form ( [ rho ] ) in the einstein equations for the metric ( [ metric ] ) we find @xmath110 the induced metric of the surface @xmath3=const is @xmath111 without loss of generality we can choose the time coordinates in both static space - times such that they are continuous on the junction .
then the continuity of the first fundamental form requires the metric function @xmath112 to be continuous .
the extrinsic curvature of the junction surface @xmath113 is defined as @xmath114 the nonvanishing components are : @xmath115 in the above expressions the derivatives were eliminated by use of eq .
( [ systfluid ] ) .
continuity of the extrinsic curvature across the junction hypersurface @xmath116 is achieved , provided that the metric and the function @xmath117 ( thus also @xmath118 ) are continuous .
we have proved the following result : _ any two spherically symmetric static solutions can be matched along hypersurfaces _ @xmath113 _ provided the radial pressures are continuous .
_ this is similar to the theorem given by fayos , jan , llanta and senovilla @xcite for the matching of the vaidya solution with a generic spherically symmetric solution along timelike hypersurfaces .
a generic discussion on matching spherically symmetric space - times along thin spherical timelike shells can be found in@xcite . for the double null dust solution @xmath119 .
in consequence the interior fluid should have a radial pressure equal to the energy density ( [ rho ] ) of the double null dust solution on the junction .
however , no conditions on the pressures tangent to the spheres emerge .
we see from eqs .
( [ bel ] ) , ( [ ml ] ) and ( [ rho ] ) that the integration constants @xmath37 and @xmath48 appear in the radial pressure and mass function of the static double null dust solution .
continuity of these functions on the junction fixes the value of the constants , once the interior solution is chosen . for a realistic star
, the mass @xmath120 should be positive .
this implies a lower boundary for the possible values of @xmath121 , as follows from fig .
let us illustrate the junction with the interior schwarzschild solution@xcite @xmath122 , with the energy density @xmath123=const and pressure @xmath124 given by @xmath125 several relations among the schwarzschild parameters @xmath126 , radius of the star @xmath127 and the parameters @xmath37 and @xmath48 emerge from the junction conditions : @xmath128 we have denoted by @xmath129 and @xmath130 the values of the functions @xmath131 and @xmath44 at the junction @xmath132 .
the first two relations ( [ params ] ) determine the constant @xmath37 and the value of the radial coordinate @xmath133 at the junction in terms of @xmath134 and @xmath135 , when eqs .
( [ rl ] ) , ( [ bel ] ) and ( [ ml ] ) are inserted .
eliminating @xmath129 from the last two relations of ( [ params ] ) , a constraint @xmath136 on the possible values of characteristics of the interior emerges .
finally ( [ rl ] ) implies @xmath137 . in the light of the above relations we see that after choosing some value for @xmath48 ( a scale ) , the constant @xmath37 is determined exclusively by the radius and density ( or mass ) of the star .
in this subsection we study the junctions with the incoming and outgoing vaidya solutions , which are at the exterior of the static double null dust solution .
there are only three points ( in fact spheres ) in common with exterior schwarzschild regions ( fig .
4 ) , and the matching can be done without introducing regularizing thin shells . the high - frequency approximation to the unidirectional radial flow of spherically symmetric unpolarized radiation , characterized by the energy - momentum tensor ( [ en - momva ] ) is represented by the vaidya solution@xcite : @xmath138 + r^{2}d\omega ^{2}. \label{va1}\ ] ] the radial null geodesics are the lines @xmath104=const and the curves given by @xmath139 we would like to match our static solution with incoming and outgoing vaidya solutions along the outgoing , respectively incoming radial null geodesics , e.g. along lines of constant incoming , respectively outgoing coordinate ( fig 4 . ) .
these are given by eq .
( [ geod ] ) in the vaidya solution and by eq .
( [ geodef ] ) in the static solution .
a convenient formalism for matching solutions along null surfaces , which does not require coordinates that match continuously on the shell , was developed by barrabs and israel@xcite .
their discussion on the particular case of spherical symmetry requires the metric in both space - times written in the form : @xmath140 + r^{2}d\omega ^{2}. \label{gensphe}\ ] ] here @xmath25 and @xmath141 depend on both @xmath3 and @xmath142 .
the null coordinate @xmath142 is always outgoing but it can be either increasing or decreasing with time
. then the junction is done along the null hypersurfaces @xmath142=const . to apply this formalism
it would be desirable to express the vaidya solution in the other set of radiation coordinates @xmath143 $ ] in which eq .
( [ geod ] ) takes the form @xmath105=const .
however , this is possible only when double null coordinates are found .
it was demonstrated in @xcite that double null coordinates exist when the mass is a _ linear _ or _ exponential _ function of the advanced or retarded time @xmath104 .
we argue in what follows that the mass function of that particular vaidya solution , which is matching continuously to the static solution has a more complicated dependence .
a further inconvenience is that there is no obvious choice for the intrinsic coordinates in terms of the space - time coordinates in the null junction hypersurface . for these reasons we proceed as follows .
first we find coordinates that match continuously on the junction and in terms of which the metric is continuous
. then we employ the barrabs - israel junction formalism to find the distributional stress - energy tensor on the junction . as a radial variable of both metrics
we choose @xmath40 by extending the expression ( [ rl ] ) of @xmath144 to the vaidya regions too .
an appropriate null coordinate @xmath142 is defined by the values of @xmath145 on the junction . in these
coordinates the junction hypersurfaces are characterized by the null geodesic equations @xmath146 in both space - times .
we proceed in deriving the expressions of the coordinates @xmath79 and @xmath147 and of the mass function @xmath148 in terms of the coordinate @xmath142 . identifying the corresponding part of eq .
( [ en - momdn ] ) with eq .
( [ en - momva ] ) , we find the relation between the null coordinates @xmath104 and @xmath79 , valid on the junction : @xmath149 then we extend this relation over both the static and the vaidya regions . the null geodesic equation ( [ geodef ] ) of the static solution , evaluated on the junction ( where @xmath146 ) gives @xmath150 inserting eq .
( [ zv ] ) in the null geodesic equation ( [ geod ] ) of the vaidya solution we find its mass function in terms of the new null coordinate @xmath142 : @xmath151 thus the mass function is continuous at the junction .
the geodesic equation ( [ zv ] ) together with ( [ mz ] ) gives the relation between the null coordinates @xmath104 and @xmath142 : @xmath152 dy}{\phi _ { b}(y)}. \label{vz}\ ] ] the relations ( [ mz ] ) and ( [ vz ] ) contain in implicit form the dependence of the mass function @xmath148 of the vaidya solution on the radiation coordinate @xmath104 . despite the complicated dependence @xmath103 it is straightforward to check that @xmath148 satisfies the required monotonicity condition : @xmath153 finally eqs .
( [ vv ] ) and ( [ vz ] ) give @xmath154 : @xmath155 dy . \label{vz}\ ] ] now we express both metrics in terms of the coordinates @xmath156 .
they take the form ( [ gensphe ] ) with @xmath157 \label{fpsi } \\
f_{v } & = & 1-\frac{2m\left ( z\right ) } { r\left ( l\right ) } \;,\qquad e^{\psi_{v}}=\frac{2}{c}\frac{e^{z^{2}}}{\phi _ { b}(z ) } \left [ -e^{z^{2}}+2z\phi_{b}(z)\right ] .
\label{fpsiva}\end{aligned}\ ] ] here the index @xmath104 refers to the vaidya solution and the expressions @xmath158 and @xmath159 are given by eqs .
( [ rl ] ) , ( [ ml ] ) and ( [ mz ] ) , respectively .
we have completed the task of writing both metrics in coordinates which are continuous on the junction and in terms of which the metric is continuous .
it is immediate to check the continuity of the induced metric given by @xmath160 .
the other junction condition is a somewhat subtle issue as the conventional extrinsic curvature tensor for null hypersurfaces carries no transversal information .
we define a pseudo - orthonormal basis @xcite @xmath161 , where @xmath162 and @xmath163 are given in eq .
( [ e3e4 ] ) and @xmath164 the vector @xmath165 is orthogonal ( and also tangent ) to the hypersurfaces @xmath166 , along which the two space - times are glued together .
the vector @xmath167 is the other radial null vector , transversal to these surfaces .
they are related to the previously introduced vectors @xmath168 and @xmath169 as follows . for the junction with the incoming vaidya region @xmath170 and @xmath171 while for the junction with the outgoing vaidya region @xmath172 and @xmath173 .
the projector to these null hypersurfaces with tangent space spanned by @xmath174 and @xmath175 is @xmath176 following @xcite we define the transverse or oblique extrinsic curvature tensor with the aid of the transverse vector @xmath177 : @xmath178 the only nonvanishing components of the transverse extrinsic curvature tensor by contracting with the three basis vectors @xmath174 and @xmath175 tangent to the hypersurface . ]
are @xmath179 thus _ when matching any two metrics of the form _
( [ gensphe ] ) _ on the null hypersurfaces _
@xmath180 , _ the jump in the extrinsic curvature is given by the jump in _
@xmath181 , _ provided the metric is continuous across the junction . _
a straightforward computation employing ( [ fpsi ] ) and ( [ fpsiva ] ) gives on the junction hypersurfaces of the static double null dust region with the vaidya regions @xmath182 } { c\left[-e^{l_{j}^{2}}+2l_{j}\phi_{b}(l_{j})\right ] } \ .\ ] ] in conclusion the extrinsic curvature is also continuous on the junction .
there is no need for a thin regularizing shell separating the two domains of the space - time , in contrast to the exterior junction proposed in @xcite .
we have integrated the einstein equations in the presence of crossflowing null dust under the assumptions of spherical symmetry and staticity and analyzed various aspects related to the properties of the emerging exact solution ( [ metrictl ] ) .
the solution has dilatonic gravity connections and it can be reinterpreted as an anisotropic fluid with radial pressure equal to its energy density and no pressures along the spheres @xmath3=const .
this can be a radiation atmosphere for a star with its radial pressure equal to the energy density of the athmosphere on its surface .
no constraint on the pressures along the spherical junction surface was found . on the exterior ,
the study of the matching conditions with the vaidya solution revealed no thin shells on the junction .
as a byproduct , we have derived general conditions for the junction of two spherically symmetric solutions .
matching of two static space - times ( [ metric ] ) along @xmath3=const hypersurfaces is assured by the continuity of the metric functions and of the radial pressure .
matching of generic spherically symmetric space - times ( [ gensphe ] ) along the null hypersurfaces @xmath166 is possible whenever the metric and @xmath181 are continuous .
the negativity of the mass function in some neighbourhood of the @xmath46 singularity raises the possibility of matching this solution to a negative mass core
. this may be difficult due to the nontrivial topology of the known negative mass black hole solutions@xcite . despite the lack of experimental evidence for negative mass objects , presumably of quantum origin@xcite
, their microlensing effect@xcite on radiation from active galactic nuclei was shown to produce features similar to some observed gamma ray bursts@xcite .
equally interesting would be to introduce dynamics in the picture by an interior matching with a collapsing star .
we defer this topic to a forthcoming study .
an intriguing open question remains whether exact solutions describing the collision of spherically symmetric null dust streams , which have not reached equilibrium , can be found . _ note added .
after the submission of this paper a relevant work in the subject was published by kramer @xcite . the solution presented there is the particular case of the metric ( [ metrictl ] ) with the parameter values @xmath183 and @xmath184 . our metric functions @xmath112 and @xmath25 , radial variable @xmath40 and transcendental function @xmath185 correspond to @xmath186 and @xmath187 , respectively of this paper , when the parameters values @xmath183 and @xmath184 are chosen . keeping the parameters arbitrary enabled us in sec .
iv a to match the solution ( [ metrictl ] ) with an interior schwarzschild solution with _ arbitrary _ mass and radius .
in contrast with our analysis relying on the barrabs - israel matching procedure , in @xcite the junction with the vaidya space - time was discussed by imposing the continuity of the four - metric on the junction .
the author is grateful to ji bik , gyula fodor , karel kucha and zoltn perjs for discussions on the subject and helpful references .
this work has been supported by otka no .
w015087 and d23744 grants .
the algebraic packages reduce and maplev were used for checking computations and numerical plots .
g. date , _ gen .
gravit._*29 * , 953 ( 1997 ) .
our notation @xmath188 corresponds to @xmath189 of this paper .
the scale factor @xmath190 does not appear in our equations and we have chosen the signature of the metric @xmath191 . | the einstein equations are completely integrated in the presence of two ( incoming and outgoing ) streams of null dust , under the assumptions of spherical symmetry and staticity .
the solution is also written in double null and in radiation coordinates and it is reinterpreted as an anisotropic fluid .
interior matching with a static fluid and exterior matching with the vaidya solution along null hypersurfaces is discussed .
the connection with two - dimensional dilaton gravity is established . |
the fano line shapes @xcite have played an important role in our understanding of the photoelectron spectra @xcite in atomic physics and more recently in the field of plasmonics @xcite . in atomic systems
the interference minimum in the fano line shape has been extensively used in extracting information on the relative strengths of the different decay channels which are responsible for interference effects .
the fano interference has been at the forefront in producing lasing without population inversion @xcite .
the fano line shapes have been considered as a probe of decoherence @xcite .
the fano interference is leading to other remarkable possibilities like the improvement in the efficiency of the heat engines @xcite . in plasmonics ,
interference effects play a very important role as photons can travel along different interfering paths and thus the fano line shapes become quite common .
the asymmetry of the fano line shape can be continuously tuned by controlling the rate of saturation of two interference paths @xcite .
such line shapes are also used in obtaining information on the interaction between nano particles with light @xcite .
interestingly enough the fano line shapes can be understood rather easily by using a two coupled oscillator model in which the two oscillators have quite different lifetimes @xcite .
thus the lifetimes of the modes are important in the realization of the fano line shapes .
in fact , plasmonic systems and metamaterial structures are designed keeping the lifetime issue in mind @xcite .
it turns out that the lifetime conditions are well satisfied for the optomechanical systems ( oms ) since the damping of the mirror is much smaller than the cavity damping . in the last few years the oms have shown the possibility of a variety of interference effects @xcite . clearly such a system , under certain condition , can exhibit fano line shape @xcite and in fact @xcite reports the observation of the fano line shape . in this paper
, we analyze the physics and mathematics behind fano resonances in optomechanics and report the possibility of double fano resonance in coupled oms .
the double fano resonances which are interesting in their own right can provide information on a variety of optomechanical couplings .
the fano minima can also be exploited for the successful transfer of states and transduction of photons .
the double cavity set up can lead to the tunability of the fano resonances , which in fact has been an important issue in other contexts @xcite .
the double - cavity oms are not the only ones for the observation of double fano resonances , for example , we can use electromechanical systems with several superconducting capacitors @xcite .
the outline of this paper is as follows : in sec .
[ analyze ] , we recall the fano line shapes and physically explain how these can arise in oms . in sec .
[ model ] , we introduce the model for oms and derive the line profiles under steady state conditions . in sec .
[ single ] , we examine the fano resonance in a single cavity oms and derive its resonance frequency center as well as its width .
we also quantify the fano shape by using the fano asymmetry parameter . in sec . [ double ] , we consider the double - cavity oms and show the existence of the double fano lineshape in different coupling regimes .
we explain how we can tune fano resonances by changing the coupling fields . in sec .
[ quantum ] , we calculate the quantum fluctuations of the fields in oms , and demonstrate the existence of fano resonances in the spectrum of quantum fluctuations .
in the classic work @xcite , fano considered photoionization process when a weakly bound state @xmath0 lies in the continuum @xmath1 .
the weakly bound state has a finite life time due to its coupling with the continuum .
thus , in the simplest case , there are two transition amplitudes leading to photoionization
one involves a direct transition to the continuum and the other involves transition via the autoionizing state to the continuum .
these two transition amplitudes interfere leading to the famous fano formula for the probability @xmath2 for the photoionization @xmath3 where @xmath4 is called the fano asymmetry parameter and @xmath5 is the width of the state @xmath0 .
the fano minimum occurs at @xmath6 .
the parameter @xmath4 depends on the relative strengths of the independent transitions to the state @xmath0 and @xmath1 .
if @xmath4 is large , interference disappears . clearly , to produce fano line shapes , we need two paths which interfere .
let us see , in terms of the physical processes , the possible physical mechanisms which lead to the building up of the cavity field in an oms as shown in fig .
[ fig1]a .
( color online ) ( a ) schematic of oms and ( b ) photon - phonon interaction.,scaledwidth=40.0% ] here the field @xmath7 with frequency @xmath8 is close to the cavity frequency @xmath9 and the field @xmath10 is red detuned from the cavity resonance by an amount close to the oscillating frequency of the mirror .
if the mirror position @xmath11 were fixed , the cavity field at @xmath8 will have the form @xmath12 $ ] , where @xmath13 is the rate at which the photons leak out of the left mirror .
if the mirror is active , due to the radiation pressure interaction , we have the possibility of stokes and antistokes processes .
the fields @xmath10 and @xmath7 can produce a coherent photon @xmath14 . once the coherent phonon is produced it can combine with @xmath10 to produce the cavity field at @xmath8 . in the case of a moving mirror , the two successive nonlinear frequency conversion processes between two light fields and a phonon as illustrated in fig .
[ fig1]b would produce the cavity field .
therefore , in oms , we have two coherent processes leading to the building up of the cavity field .
these are : ( a ) direct building up due to the application of @xmath7 ; and ( b ) building up due to the nonlinear processes in fig . [ fig1]b .
these two processes interfere leading to the fano resonances in oms .
the precise mathematical treatment is given in sec .
iv , where we introduce the analog of the fano asymmetry parameter @xmath4 . in atomic physics ,
fano profile have been extensively studied .
many generalizations to several resonances in continuum or interactions with different continua exist .
the different continua can be either radiative or ionizing .
radiative continuum tends to make the value of the minimum nonzero @xcite .
we study the system in fig .
[ fig2 ] , in which a mechanical resonator , coated with perfect reflecting films on both sides , is coupled to two cavities .
( color online ) schematic double - cavity oms.,scaledwidth=35.0% ] the mechanical resonator is modeled as a harmonic oscillator with mass @xmath15 , frequency @xmath16 and momentum decay rate @xmath17 . for each cavity , we denote its field by @xmath18 , frequency @xmath19 and decay rate @xmath20 , @xmath21 . the field annihilation and creation operators satisfy the commutation relation @xmath22=\delta_{ij}$ ] .
the probe laser with frequency @xmath8 , is sent into cavity @xmath23 .
other possible realizations would be using `` membrane in the middle '' setup @xcite .
the two cavities are coupled only via the oscillations of the mechanical mirror which are produced by the applied strong laser fields @xmath24 s .
further the two cavities can be in different frequency regimes .
we introduce normalized coordinates @xmath25 and @xmath26 for the mechanical oscillator with commutation relations @xmath27=[x , p]/\hbar = i$ ] .
the hamiltonian for this system is given by @xmath28 where @xmath29 is the detuning between the probe field and the coupling field in cavity @xmath23 ; and coupling coefficients are defined by @xmath30 with @xmath31 being the length of the @xmath32th cavity .
the @xmath33 terms in ( [ 1 ] ) denote the coupling of the cavity field to the applied laser fields .
the coupling and probe fields are related to the power of the applied laser fields via @xmath34 , and @xmath21 and @xmath35 , respectively . in eq .
( [ 1 ] ) , all the dissipative interactions are denoted by @xmath36 .
these include the leakage of photons from both cavities and the brownian motion of the mirror .
the hamiltonian ( [ 1 ] ) has been written by working in a picture so the very fast frequencies @xmath37 s are removed .
this results in detuning terms like @xmath38 . using eq .
( [ 1 ] ) we can derive the quantum langevin equations for the operators @xmath39 , @xmath40 , @xmath18 and @xmath41 .
however , for the purpose of this paper , we will work with semiclassical equations so that all operator expectation values are replaced by their mean values .
thus in the rest of the paper , all the quantities @xmath39 , @xmath40 , @xmath18 and @xmath41 will be numbers .
the equations of motion for @xmath39 , @xmath40 , @xmath18 and @xmath41 are found to be @xmath42 eqs .
( [ 2 ] ) involve periodically oscillating terms hence in the long time limit , any of the fields and the mechanical coordinates will have a solution of the form @xmath43 .
the @xmath44 s can be obtained by the floquet analysis .
we assume that the probe is much weaker than the coupling field , then @xmath44 s can be obtained perturbatively .
in particular , we find the steady state results to first order in @xmath45 @xmath46 where @xmath47 and @xmath48 are the detunings of the coupling lasers to the effective cavity frequencies .
@xmath49 , which is typically small , denotes the displacement of the mechanical resonator under radiation pressure .
the cavity field @xmath50 is the field in the @xmath32th cavity at the frequency of the coupling laser .
the coupling efficients are enhanced by the cavity photon number hence we define @xmath51 .
the fields @xmath52 s are the antistokes and stokes fields in the @xmath32th cavity .
the output fields from the two cavities are given by @xmath53 we now concentrate on the output fields from the cavity @xmath23 .
the output fields in the form of eq .
( [ 15 ] ) contain components at three different frequencies : the coupling frequency @xmath54 ; the antistokes frequency , which is also the probe frequency , @xmath55 ; and the stokes frequency @xmath56 . among these three components , we are most interested in the generated antistokes and stokes sidebands and we display them as the normalized quantities defined by @xmath57 and @xmath58 .
the actual normalized output field at the antistokes frequency from the cavity @xmath23 is given by @xmath59 , cf .
( [ 15 ] ) .
the antistokes field would be resonantly enhanced in the vicinity of the cavity frequency @xmath60 , when both the coupling fields are tuned by an amount close to the mechanical frequency below their corresponding cavity frequency , i.e. @xmath61 .
we work in the regime with cooperativity @xmath62 in which the oms is strongly coupled , then the antistokes and stokes fields in cavity @xmath23 are given by @xmath63 + \frac{g_1 ^ 2}{[\frac{\gamma_m}{2}-i(\delta-\omega_m ) ] + \frac{g_2 ^ 2}{[\kappa_2-i(\delta-\delta_2 ) ] } } } , \label{17 } \\
\mathcal{e}_{1s } & = \frac{-i\kappa_1/\omega_m}{1 + \frac{[\kappa_1+i(\delta-\delta_1)]}{g_1 ^ 2}\left\ { [ \frac{\gamma_m}{2}+i(\delta-\omega_m ) ] + \frac{g_2 ^ 2}{\kappa_2+i(\delta-\delta_2)]}\right\ } } .
\label{18}\end{aligned}\ ] ] similarly , the antistokes and stokes fields in cavity @xmath64 are found to be @xmath65 } \frac{g_2}{[\kappa_2-i(\delta-\delta_2)]}}{\frac{g_1 ^ 2}{[\kappa_1-i(\delta-\delta_1 ) ] } + \frac{g_2 ^ 2}{[\kappa_2-i(\delta-\delta_2 ) ] } + [ \frac{\gamma_m}{2}-i(\delta-\omega_m ) ] } , \label{19 } \\
\mathcal{e}_{2s } & = \frac{\kappa_2\frac{g_2}{\omega_m } \frac{g_1}{[\kappa_1-i(\delta-\delta_1)]}}{\frac{g_1 ^ 2}{[\kappa_1+i(\delta-\delta_1 ) ] } + \frac{g_2 ^ 2}{[\kappa_2+i(\delta-\delta_2 ) ] } + [ \frac{\gamma_m}{2}+i(\delta-\omega_m)]}. \label{20}\end{aligned}\ ] ]
we examine now fano resonances in the output fields . we have four different fields @xmath66 , @xmath67 , and @xmath21 .
we first examine @xmath68 if @xmath69 , i.e. the case of oms in fig .
[ fig1 ] . for this system ,
the antistokes field is @xmath70}{[\kappa_1-i(\omega_p-\omega_1)][\frac{\gamma_m}{2}-i(\omega_p-\omega_{c1}-\omega_m ) ] + g_1 ^ 2}.\ ] ] typically the mechanical damping is much smaller than the cavity damping , @xmath71 , and we work in the resolved sideband limit , @xmath72 .
we expect two resonances at @xmath73 , and at @xmath74 . in order to keep these two resonances distinct , we keep @xmath75 , i.e. @xmath76 . for clarity
, we show the relations between different frequencies in fig .
[ fig3 ] .
( color online ) schematic illustration of frequencies used in obtaining fano lineshapes .
fano asymmetry parameter @xmath4 is defined in terms of detuning @xmath77 .
the effective damping is defined by @xmath78$].,scaledwidth=40.0% ] the resonance at @xmath79 would be very narrow since @xmath71 .
the frequency offset factor @xmath80 plays important role in the production of the fano line shapes .
physically it means that the antistokes process is not resonant with the cavity frequency .
we examine the structure of @xmath81 near the resonance @xmath82 for a fixed value of @xmath80 , and we define @xmath83 . in the vicinity of this resonance , @xmath84 and eq .
( [ 21 ] ) can be approximated to @xmath85 and hence @xmath86 = \frac{2}{1+q^2}\cdot\frac{(\bar{x}+q)^2}{\bar{x}^2 + 1},\ ] ] where @xmath87 , @xmath88 , and @xmath77 .
the profile ( [ 23 ] ) has exactly the same form as the classic profile of fano resonance with maximum at @xmath89 and zero at @xmath90 .
the asymmetry parameter @xmath4 is related to the frequency offset @xmath80 .
keep in mind that this is derived in the vicinity of @xmath84 , i.e. @xmath91 . in order to see explicitly the nature of the output fields , we use the following set of experimentally realizable parameters @xmath92mhz , @xmath93mhz , @xmath94mhz , and @xmath95mhz .
we display the full profile of the output fields in fig .
[ fig4]a as a function of @xmath96 for a single cavity oms .
( color online ) the antistokes field @xmath97 ( a ) and the stokes field @xmath98 ( b ) as a function of frequency of the probe laser input @xmath8 for the oms in fig .
the black dotted , blue dashed and red solid curves are corresponding to @xmath99 , @xmath100 and @xmath101 , respectively.,scaledwidth=45.0% ] it shows the narrow fano profile as well as the relatively broad resonance near @xmath102 . for detuning @xmath103 ( dotted curve ) ,
we obtain the standard eit profiles @xcite .
as we increase the detuning , the fano resonance shifts away from the cavity resonance frequency and becomes asymmetric .
each of these fano lineshapes has a zero point exactly at the frequency @xmath90 or equivalently @xmath104 .
our approximation formula ( [ 23 ] ) and the numerical curves obtained directly from ( [ 17 ] ) agree well .
safavi - naeini et al @xcite have observed such profiles for a broad range of @xmath4 values . what we have demonstrated in this section
is how fano line shapes can arise in oms under the condition @xmath71 .
when @xmath17 starts increasing , the character of the line shape starts changing in a manner similar to changes in the fano line profiles when the radiative effects are included .
it is also noteworthy to study the stokes sideband generated by the coupling laser and the mechanical oscillator , although it is suppressed because it is an off - resonantly process . in fig .
[ fig4]b , we plot the stokes sideband .
the line shape is asymmetric though a good signature of interference is missing .
this is because fano resonance requires two coherent routes for building up the cavity field , which can interfere with each other , whereas the only route producing stokes sideband is via the combination of coupling field and the mechanical phonons .
recently double cavity configurations have attracted a lot of attention because of their wide applicability in photon switching @xcite , state transfer @xcite and transduction of photons @xcite .
we discuss yet another possibility , making use of double cavities to tune the fano resonances . in this section
, we will show how we can change and control the fano resonance by adding a second cavity in oms . when the coupling fields exist in both cavities , the denominator of eq .
( [ 17 ] ) becomes cubic and hence the number of roots increases from two to three .
this is because we have three coupled systems : two cavity modes and one mechanical mode . at the same time
, the numerator in ( [ 17 ] ) becomes a quadratic function of @xmath11 suggesting the possibility of two different minima in the output fields .
therefore , a single fano resonance goes over to a double fano resonance .
next we examine the quantitative features of the double fano resonance in oms .
the parameter space is large and therefore we begin by fixing the detuning in cavity @xmath23 as @xmath105 so that its fano asymmetry parameter @xmath106 , and we let the detuning of cavity @xmath64 arbitrary such that @xmath107 . in the vicinity of @xmath108 , the roots of the numerator in eq .
( [ 17 ] ) determine the existence of the fano minima .
we first discuss the case when @xmath109 .
then to first order in dampings , the roots are @xmath110 leading to the splitting of the fano resonances into two .
the power of the coupling field in cavity @xmath64 determines their frequency splitting .
( color online ) the antistokes fields @xmath97 in cavity @xmath23 ( a ) and in cavity @xmath64 ( b ) as a function of frequency of the probe laser input in a double - cavity oms . the thin black , blue dashed and red solid curves are corresponding to different coupling strength of cavity @xmath64 that @xmath111 , @xmath112 and @xmath113 , respectively .
we set @xmath114 and @xmath115.,scaledwidth=45.0% ] in fig .
[ fig5]a , we explicitly show the splitting of the fano resonance in the double - cavity oms using the same parameters for cavity @xmath23 and @xmath115 , @xmath114 for cavity @xmath64 with different coupling strengths . in fig .
[ fig5]a , the thin curve shows a single fano resonance when the coupling field in cavity @xmath64 is absent .
as we increase the coupling field in cavity @xmath64 , the fano resonance splits and the splitting increases linearly as we increase the coupling power .
apart from the splitting , their resonance frequency center is shifted by an amount @xmath116 .
the frequency splittings of the two fano resonance are @xmath117 and @xmath118 , which respectively equals to @xmath119 .
the frequency splitting is independent of the detuning of cavity @xmath64 , as long as it is close to @xmath16 .
therefore , one can always obtain the coupling strength , as well as the coupling power , by measuring the double fano resonances . in the figure
, the minimum values of the double fano resonances do not go to zero due to the finite values of @xmath120 and @xmath17 . in an oms with lower @xmath120
, we should be able to obtain a lower minimum and a higher maximum in the double fano resonances .
this is reminiscent of the result in the context of photoionization in which the value of the minimum depends on the radiative effects @xcite . in fig .
[ fig5]b , we plot the antistokes field in cavity @xmath64 in response to the probe laser input in cavity @xmath23
. we see asymmetric peaks generated around the frequency of the fano resonances in @xmath68 and their widths are similar to the corresponding fano resonances . both the peaks heights and peaks splitting increase with the increasing of the coupling power .
physically , this can be interpreted as the probe energy in cavity @xmath23 is transferred to cavity @xmath64 via the mechanical resonator .
the antistokes field in cavity @xmath64 shows anti - symmetric split fano resonances .
the characteristics of the double fano resonances are different in the weak coupling limit . when @xmath121 , the roots of the numerator in eq .
( [ 17 ] ) determining the fano minima are @xmath122 the root @xmath123 indicates a frequency shift of the fano resonance with an amount @xmath124 , and the root @xmath125 implies the emergence of a new fano resonance around @xmath126 besides the original fano resonance around @xmath84 . in fig .
[ fig6 ] , we illustrate both the antistokes and stokes field in cavity @xmath23 using the following parameters @xmath127 , @xmath128 , and @xmath129 , ( compared with @xmath69 for the single cavity case as dashed curves ) and parameters for cavity @xmath23 are identical to fig .
[ fig4 ] .
( color online ) the antistokes field ( a ) and stokes field ( b ) in cavity @xmath23 as a function of frequency of the probe laser input in a double - cavity oms . the solid curves : double - cavity oms with @xmath129 shows the emergence of a the second fano resonance , compared to the dashed curve : single cavity oms .
the inset in ( a ) shows the full profile in a large scale .
we set @xmath130 and @xmath127 .
, scaledwidth=45.0% ] using these parameters , the zero point frequency shift of the original fano resonance is calculated to be @xmath131 and the width increase to be negligible . in fig .
[ fig6]a , the new fano resonance emerges around @xmath132 which matches our calculation . in fig .
[ fig6]b , we also plot the stokes field in cavity @xmath23 .
it is interesting that a narrow dip is created inside the original single - peak lineshape when cavity @xmath64 is coupled to the oms .
the widths of the broad lineshape and the narrow dip are close to the widths of the original and newly - emerged fano resonances of the antistokes field in cavity @xmath23 , respectively .
the dip is caused by cavity @xmath64 adding an extra damping mechanism to the mechanical resonator and destructively interfering with the mechanical damping , so that it prevents the mechanical mode from aiding the generation of the stokes field in cavity @xmath23 .
in the previous sections , we studied the oms when the optical cavity is fed by both a detuned coupling field and a weak probe field , and found its output exhibits fano resonance .
now we will study the quantum fluctuation of the cavity field without any input probe field , as illustrated in fig .
[ fig7 ] .
( color online ) schematic double - cavity oms . here
@xmath24 s are coherent fields and @xmath133 s are the quantum vacuum fields .
@xmath134 is the brownian noise.,scaledwidth=35.0% ] the quantum fluctuation of the cavity fields arises ( i ) direct from the fluctuation of the vacuum input and ( ii ) from the process of photon creation via oscillating mirror subjected to thermal noise .
those two mechanisms can interfere destructively creating a zero amplitude in the fluctuations of the cavity field . in a double - cavity oms ,
the quantum langevin equations governing the operators @xmath39 , @xmath40 , @xmath18 and @xmath41 are given by @xmath135 where @xmath136 and @xmath137 are the input noise from cavity @xmath23 with correlation function @xmath138 and @xmath134 stems form the thermal noise of the mechanical resonator at finite temperature with correlation function in frequency domain @xmath139\delta(\omega+\omega),\ ] ] where @xmath140 is the boltzmann constant and @xmath141 is the temperature of the environment of the mirror .
( [ 40 ] ) are difficult to solve because they are nonlinear . however , considering that the quantum fluctuation values around their steady states are relatively small , we can adopt the standard linearization method by separating the fluctuations from their mean values , @xmath142 for @xmath21 . when expanding the products of two operators @xmath143 and @xmath144 , we can make the approximation @xmath145 so that quantum langevin equations are modified as @xmath146 the coupling fields are absorbed in the steady state mean values , so they do not show explicitly in eq .
( [ 42 ] ) . the @xmath50 s and @xmath147 s
are defined identically to sec .
[ model ] . in order to get
the spectra of the fluctuations in the quantities @xmath148 , @xmath149 , @xmath150 and @xmath151 , we fourier transform them into the frequency domain using @xmath152 . by solving them ,
we obtain the fields in cavity @xmath23 containing the signature of the quantum fluctuations @xmath153 and fluctuations @xmath154 in cavity @xmath64 can be calculated similarly using the symmetry property of the double - cavity configuration .
note that the term @xmath155 physically means that a noise photon at @xmath156 produces a photon at frequency @xmath156 where as the term @xmath157 corresponds to the four wave mixing process where a photon of frequency @xmath158 produces a photon of frequency @xmath159 .
similar interpretations apply to @xmath160 and @xmath161 .
thus @xmath155 and @xmath162 would have direct relation to the antistokes and stokes fields discussed in the earlier sections .
the mechanical noise can be suppressed by cooling down the environment temperature , though it is the dominant contribution to the fluctuations at high temperatures , and hence we omit the @xmath163 term . here
@xmath164 s , and @xmath165 s are the functions given by @xmath166 the quadratures of the field in cavity @xmath23 , which can be measured using homodyne detection , have the spectra defined as @xmath167 and @xmath168 with @xmath169 and @xmath170 .
now we calculate the fluctuation spectrum in the @xmath171 quadrature as @xmath172 we first study single - cavity oms when @xmath69 . the cavity field fluctuation in @xmath171 quadrature
is given as @xmath173 we do not show the expressions of fluctuations @xmath174 or @xmath175 since they do not exhibit fano minimum . equation ( [ 46 ] ) indicates a fano lineshape , which has a minimum at @xmath176 and a maximum at @xmath177 with width @xmath178 and asymmetry parameter @xmath77 .
to see the fano resonance , it is important to have @xmath179 and @xmath180 .
the spectra @xmath181 and @xmath182 of the output field are different from the cavity fields by an amount of @xmath136 using the input - output relation @xmath183 .
we illustrate the spectra of the quadrature @xmath184 , @xmath185 , @xmath181 , and @xmath182 for both single - cavity oms ( solid curves ) and double - cavity oms ( dashed curves ) in fig .
[ fig8 ] using parameters as in fig .
[ fig5 ] .
( color online ) the spectra of the quadratures for the cavity fields and output fields for both single - cavity oms ( solid curves ) and double - cavity oms ( dashed curves ) .
the parameters used are the same to fig .
, scaledwidth=48.0% ] from the solid curves in the figure , we see that the @xmath184 quadrature exhibits a clear fano resonance .
the fano resonance has a zero point at frequency @xmath186 and has width @xmath187 , both of which match our calculation .
the spectrum of the quadratures @xmath181 and @xmath182 of the output field also have typically asymmetric line shapes which are signatures of interferences .
these spectra have similarities to the spectra for the stokes field ( fig . [ fig4]b ) .
note that the formula like eq .
( [ 46 ] ) shows that the quadrature spectra are determined by the interference of the stokes and antistokes terms .
the reason is that in the region of interest in the spectrum @xmath184 , the term @xmath188 is approximately flat .
when the second cavity is coupled to the system , we expect a splitting of the fano lineshape in the spectrum of fluctuations following the the classical analysis of sec .
[ double ] .
the splittings of the resonances separated by @xmath189 appear in the dashed curves in fig .
the splittings are due to the enhanced coupling strength by the increasing photon number in the cavities , which induced normal mode splitting of the cavity states .
in conclusion , we have shown how the asymmetric fano line shapes can arise in optomechanics .
we identify interfering pathways leading to the fano resonances .
in contrast to atomic systems , the coupling field can be used to tune fano resonances using both the frequency and the power of coupling field .
in fact , as displayed in fig .
[ fig1]b the coupling field opens up another coherent path way .
we give explicit expressions for the width and the asymmetry parameter .
the fano resonances can be studied both via pump probe experiment and via the study of the quantum fluctuations in the output fields .
the fano minima are much more pronounced in the results of the pump probe experiments .
the double cavity oms produce double fano minima . 1 u. fano , phys . rev .
* 12 * 4 , 1866 ( 1961 )
. b. h. bransden , and c. c. jean joachain , physics of atoms and molecules , 2nd ed .
( addison - wesley , 2003 ) , chap . 4
. b. gallinet and o. j. f. martin , acs nano , * 5 * , 8999 ( 2011 ) ; b. gallinet and o. j. f. martin , phys .
b * 83 * , 235427 ( 2011 ) ; y. francescato , v. giannini , and s. a. maier , acs nano * 6 * , 1830 ( 2012 ) ; alp artar , ahmet ali yanik , and hatice altug , nano lett , * 11 * , 3694 ( 2011 ) ; r. taubert , m. hentschel , j. kstel , and h , giessen , nano lett . , * 12 * , 1367 ( 2012 ) . s. e. harris , phys . rev
. lett . * 62 * , 1033 ( 1989 ) ; v. g. arkhipkin and yu .
i. heller , phys . lett . * 98*a , 12 ( 1983 ) ; and o. a. kocharovskaya and ya .
i. khanin , pisma zh .
* 48 * , 581 ( 1988 ) ; m. o. scully , s. -y .
zhu , and a. gavrielides , phys .
62 * , 2813 ( 1989 ) ; e. s. fry , x. li , d. nikonov , g. g. padmabandu , m. o. scully , a. v. smith , f. k. tittel , c. wang , s. r. wilkinson , and s. -y .
zhu , phys .
lett . * 70 * , 3235 ( 1993 ) . a common source of decoherence would be radiative decay which is extensively discussed in g. s. agarwal , s. l. haan and j. cooper , phys .
a * 29 * , 2552 ( 1984 ) , see specifically eq .
( 4.8 ) which can be rewritten by introducing a complex fano asymmetry parameter , see also a. brnthaler , et al , phys .
105 * , 056801 ( 2010 ) .
m. o. scully , k. r. chapin , k. e. dorfman , m. b. kim and a. a. svidzinsky , pnas 108 , 15097 ( 2011 ) .
a. e. miroshnichenko , s. flach and y. s. kivshar , rev .
phys . * 82 * , 2257(2010 ) .
s. bandopadhyay , b. dutta - roy , and h. s. mani , am . j. phys .
* 72 * , 1501 ( 2004 ) ; c. l. garrido alzar , m. a. g. martinez , and p. nussenzveig , am . j. phys . * 70 * , 37 ( 2002 ) ; p. tassin , l. zhang , r. zhao , a. jain , t. koschny , and c. m. soukoulis , phys .
lett . * 109 * , 187401 ( 2012 ) .
g. s. agarwal and s. huang , phys .
a * 81 * , 041803(r ) ( 2010 ) ; s. weis , el al . , science * 330 * , 1520 ( 2010 ) ; q. lin , el al . , nat
. photon . * 4 * , 236 ( 2010 ) ; j. d. teufel , el al . ,
nature ( london ) , * 471 * , 204 ( 2011 ) .
a. h. safavi - naeini , el al . ,
nature ( london ) * 472 * , 69 ( 2011 ) .
f. elste , s. m. girvin , and a. a. clerk , phys .
lett . * 102 * , 207209 ( 2009 ) .
a. xuereb , r. schnabel , and k. hammerer , phys .
* 107 * , 213604 ( 2011 ) .
j. gu , r. singh , x. liu , x. zhang , y. ma , s. zhang , s. a maier , z. tian , a. k. azad , h .-
chen , a. j. taylor , j. han , and w. zhang , nat .
commun . * 3 * , 1151 ( 2012 ) .
f. massel , s. u. cho , juha - matti pirkkalainen , p. j. hakonen , t. t. heikkil , m. a. sillanp , nat . commun . * 3 * , 987 ( 2012 ) .
j. d. thompson , b. m. zwickl , a. m. jayich , florian marquardt , s. m. girvin and j. g. e. harris , nature ( london ) * 452 * , 900 ( 2008 ) .
a. h safavi - naeini and o. painter , new j. phys .
* 13 * , 013017 ( 2011 ) .
k. qu and g. s. agarwal , phys .
a * 87 * , 031802(r ) ( 2013 ) .
l. tian , phys .
108 * , 153604 ( 2012 ) .
wang and a. a. clerk , phys .
lett . * 108 * , 153603 ( 2012 ) .
barzanjeh , m. abdi , g. j. milburn , p. tombesi , and d. vitali , phys .
* 109 * , 130503 ( 2012 ) . | the fano line profiles , originally discovered in the context of photoionization , have been found to occur in a large class of systems like resonators , metamaterials , plasmonics .
we demonstrate the existence of such resonances in cavity optomechanics by identifying the interfering contributions to the fields generated at antistokes and stokes frequencies .
unlike the atomic systems , the optomechanical systems provide great flexibility as the width of the resonance is controlled by the coupling field .
we further show how the double cavities coupled by a single optomechanical mirror can lead to the splitting of the fano resonance and how the second cavity can be used to tune the fano resonances .
the fano resonances are quite sensitive to the decay parameters associated with cavities and the mechanical mirror . such resonances can be studied by both pump probe experiments as well as the spectrum of the quantum fluctuations in the output fields . |
let @xmath1 be a hyperbolic knot and @xmath2 a conway sphere .
namely @xmath3 intersects transversally @xmath4 in four points .
we write @xmath5 to denote any of the three involutions in figure [ fig : involutions ] , for @xmath6 .
the knot @xmath7 obtained by cutting along @xmath3 and gluing again after composing with @xmath8 is called the mutant knot .
we are interested in comparing invariants of @xmath4 and @xmath9 , thus we may assume that , if @xmath10 denotes the knot exterior , then the surface @xmath11 is essential in @xmath12 .
ruberman @xcite showed that @xmath9 is also hyperbolic , and that @xmath13 has the same volume as @xmath14 .
see @xcite for a recent account on invariants that distinguish or not @xmath4 from @xmath9 .
let @xmath15 be a lift of the holonomy . if @xmath16 is a meridian , then @xmath17 , and there are precisely two lifts of the holonomy up to conjugation , one with @xmath18 and another with @xmath19 .
by @xcite , @xmath20 is acyclic , namely the homology and cohomology of @xmath12 with coefficients twisted by @xmath20 vanish , hence the reidemeister torsion @xmath21 is well defined .
moreover as the dimension of @xmath22 is even , there is no sign indeterminacy , thus @xmath21 is a well defined nonzero complex number , independent of the conjugacy class of @xmath20 .
hence these torsions are two topological invariants of the hyperbolic knot .
[ thm : main ] let @xmath4 , @xmath12 , @xmath8 and @xmath23 be as above .
let @xmath15 and @xmath24 be lifts of the holonomy , with @xmath25 .
then @xmath26 this is not true for any representation of @xmath27 .
wada proved in @xcite that the twisted alexander polynomials could be used to distinguish mutant knots .
n. dunfield , s. friedl and n. jackson @xcite computed the torsion for the representation @xmath20 twisted by the abelianization map ( namely , the corresponding twisted alexander polynomials ) and proved that it distinguishes mutant knots .
however , the evaluation at @xmath28 of these polynomials gives numerical evidence of theorem [ thm : main ] .
in @xcite we prove that when we consider @xmath29 the @xmath30-dimensional irreducible representation , then the composition @xmath31 is acyclic , thus its torsion is well defined .
we have computed that the torsion of @xmath32 distinguishes the conway and the kinoshita - terasaka mutants , see section [ sec : concluding ] .
the paper is organized as follows . in section [ sec : mutation ] , we discuss the basic constructions for reidemeister torsion and representations of mutants , and we give a sufficient criterion in proposition [ prop : criterion ] for invariance of the torsion under mutation . the sufficient criterion of proposition [ prop : criterion ] is stated in terms of the action of @xmath8 on the cohomology of @xmath33 with twisted coefficients .
this is applied for the proof when @xmath34 in section [ sec : trace-2 ] .
the proof when @xmath35 in section [ sec : trace+2 ] is different , because the involved cohomology groups are different in each situation . in section [ section : example ] we compute an example , the kinoshita - terasaka and conway mutants , and section [ sec : concluding ] is devoted to further discussion .
let @xmath36 and @xmath37 denote the components of @xmath38 , so that the pairs @xmath39 are tangles with two strings .
the exterior of the knot is denoted by @xmath40 we also denote @xmath41 write a commutative diagram for the inclusions : @xmath42^{i_1 } \ar[d]_{i_2 } & b_1 \ar[d ] \\
b_2 \ar[r ] & s^3 } \qquad\textrm { and } \qquad \xymatrix { s \ar [ r]^{i_1 } \ar[d]_{i_2 } & m_1 \ar[d ] \\ m_2 \ar[r ] & m } \ ] ]
so that @xmath43 ) is an amalgamated product @xmath44 let @xmath45 , @xmath46 and @xmath47 denote the restriction of @xmath20 to @xmath48 , @xmath49 , and @xmath50 , respectively , so that @xmath51 using the notation @xmath52 there exists @xmath53 which is unique up to sign @xcite such that @xmath54 notice that @xmath53 corresponds to a rotation of order two in hyperbolic space , therefore , @xmath55 is conjugate to @xmath56 to construct the representation of @xmath57 , we also use the amalgamated product structure with the same inclusion @xmath58 , but with @xmath59 instead of @xmath60 .
the representation @xmath24 is then defined by @xmath61 because @xmath62 . to set notation , we recall the basic construction of cohomology with twisted coefficients .
let @xmath63 be a compact @xmath64-complex and @xmath65 a representation .
the cellular chains of its universal covering are denoted by @xmath66 , which is a chain complex of left @xmath67$]-modules of finite type .
the cochains with twisted coefficients are then @xmath68 } ( c_*(\tilde x ; \mathbf z ) , \mathbf c^2_{\rho } ) , \ ] ] where @xmath69 is viewed as a left @xmath67$]-module by the action induced by @xmath20 .
the corresponding cohomology groups are denoted by @xmath70 as they only depend on the underlying topological space @xmath71 of the @xmath64-complex @xmath63 .
we shall mainly work with aspherical spaces , in this case the homology or cohomology of @xmath63 with twisted coefficients is naturally isomorphic to the group cohomology of @xmath72 .
we shall also be interested in the de rham cohomology .
assuming @xmath73 is a smooth manifold , let @xmath74 denote the flat bundle with monodromy @xmath20 .
the space of @xmath75-forms valued on @xmath76 is @xmath77 .
the de rham cohomology of @xmath78 is naturally isomorphic to @xmath79 many properties of cohomology without coefficients hold true when we have twisted coefficients : mayer - vietoris , long exact sequence of the pair , etc .
poincar duality is discussed in section [ subsection : perfectpairing ] .
we will use mayer - vietoris for the pair @xmath80 to compute the torsion of @xmath12 and of @xmath23 .
for this , we need to compute some cohomology groups .
we start with the planar surface with four boundary components @xmath81 : [ lem : cohoms ] @xmath82 for @xmath83 and @xmath84 .
firstly @xmath85 is isomorphic to the subspace of @xmath22 of elements that are fixed by @xmath86 , hence it vanishes because @xmath45 is an irreducible representation . on the other hand , @xmath87 for @xmath88 , because @xmath33 has the homotopy type of a graph .
finally @xmath89
[ lem : cohomm_i ] for @xmath90 , @xmath91 for @xmath92 and @xmath93 . by mayer - vietoris , and using the fact that @xmath94 , then @xmath95 the lemma follows from lemma [ lem : cohoms ] , because @xmath96 .
mayer - vietoris for @xmath12 and @xmath23 give the isomorphisms : @xmath97 in order to relate @xmath98 and @xmath99 , we use the composition with @xmath55 , where @xmath100 denotes @xmath47 conjugated by @xmath55 .
namely , recall that @xmath101 } ( c_*(\widetilde m_2;\mathbf z ) , \mathbf c^2_{\rho_2}).\ ] ] define : @xmath102 it is straightforward to check that this defines an isomorphism of complexes .
thus @xmath103 is an isomorphism and we have a commutative diagram : @xmath104^{i_2^ * } \ar[d]_{a _ * } & h^1(s;\rho_2 ) \ar[d]^ { a_*\circ\tau^ * } \\
h^1(m_2;\rho_2^a ) \ar[r]^{(i_2\circ\tau)^ * } & h^1(s;\rho_0 ) } \ ] ] write @xmath105 . since @xmath106 ( see equation ( [ eqn : a ] ) ) and since @xmath107
, we have : @xmath108 let @xmath63 be a compact cw - complex equipped with a representation @xmath109 when @xmath110 , the reidemeister torsion can be defined , and it is an invariant of @xmath63 , up to subdivision , and the conjugacy class of @xmath20 .
we will not recall the definition , that can be found in @xcite , for instance .
there are two main issues for the torsion we are interested in .
firstly , the torsion is only defined up to sign , but since we consider a two dimensional vector space , it is sign defined , hence a nonzero complex number .
equivalently , any choice of homology orientation for turaev s refined torsion @xcite gives the same result .
secondly , since we are working with three and two - dimensional manifolds , the pl - structure is not relevant .
thus , for a two and three - dimensional manifold @xmath63 and an acyclic representation @xmath111 , its torsion is denoted by @xmath112 when @xmath20 is not acyclic , then we can also use the reidemeister torsion provided we specify a basis for @xmath113 . choose @xmath114 a basis for @xmath115 as @xmath116-vector space .
in particular @xmath117 is a basis for @xmath118 .
by milnor s formula @xcite for the torsion of a long exact sequence applied to ( [ eqn : mv ] ) and ( [ eqn : mvtau ] ) : @xmath119 here @xmath120 denotes the disjoint union of basis .
notice that milnor works with torsions up to sign in @xcite , but its formalism applies even with sign . since @xmath121 , we deduce @xmath122 namely , the determinant of the matrix whose entries are the coefficients of the basis @xmath123 with respect to @xmath124 .
the following is a sufficient criterion for invariance of torsion with respect to mutation .
[ prop : criterion ] if @xmath125 leaves invariant the image of @xmath126 , then @xmath127 .
since @xmath128 by formula ( [ eqn : tau2=-id ] ) , @xmath129 diagonalizes with eigenvalues @xmath130 . hence , assuming that @xmath129 leaves invariant the image of @xmath131 , the matrix in equation ( [ eqn : ratiotorsions ] ) is conjugate to @xmath132 hence it has determinant @xmath28 .
we discuss first the case where @xmath133 .
the proof has three parts . in subsection [
subsection : perfectpairing ] we consider a perfect pairing on @xmath134 ( pulling back the cup product on @xmath135 by the isomorphism @xmath136 ) .
we show that , for @xmath137 , @xmath138 , the images of @xmath139 are isotropic subspaces .
then in subsection [ subsec : quadric ] we analyze properties of isotropic planes of @xmath140 , that are viewed as lines in a ruled quadric in @xmath141 .
the properties of this ruled quadric are used in a deformation argument in subsection [ subsection : firstdfm argument ] to conclude the proof when @xmath133 . for a closed oriented @xmath142-manifold @xmath143 and a representation @xmath144 ,
there is a nondegenrate pairing : @xmath145 which is the composition of the cup product and the map induced from the @xmath20-invariant pairing @xmath146 this pairing is bilinear , nondegenerate ( poincar duality ) and natural .
in addition , when @xmath147 and @xmath137 it is symmetric ( because both the usual cup product and the determinant are antisymmetric ) .
in particular , we have two symmetric pairings , for @xmath90 : @xmath148 the following lemma assumes that @xmath34 , as the whole section , though it only requires @xmath149 .
[ lemma : spartial ] @xmath150 . since @xmath34
, @xmath151 has no nontrivial fixed vectors in @xmath22 .
thus @xmath152 and , by duality , @xmath150 .
[ lemma : spartialmk ] for @xmath90 , the inclusion map induces an isomorphism @xmath153 this follows from the mayer - vietoris sequence applied to @xmath33 and @xmath154 , which is the union of the two annuli around the arcs of @xmath155 that have the homotopy type of a component of @xmath156 .
hence by lemma [ lemma : spartial ] , @xmath157 .
pulling back the pairings ( [ eqn : pairingmk ] ) by the isomorphism ( [ eqn : isos ] ) , we obtain two symmetric perfect pairings @xmath158 [ lemma : cup1cup2 ] both pairings are the same : @xmath159 .
we use de rham cohomology for the proof .
we first claim that every cohomolgy class in @xmath160 is represented by a form compactly supported in the interior of @xmath33 , using that @xmath161 .
namely , let @xmath162 be any closed form and @xmath163 a tubular neighborhood of @xmath156 .
since @xmath161 , the restriction of @xmath164 to @xmath165 is exact : there exists a section @xmath166 satisfying @xmath167 where @xmath168 is the differential . consider a smooth function @xmath169 $ ] that extends smoothly to zero on @xmath170 and equals one in a smaller neighborhood of @xmath156 .
then the form @xmath171 is cohomologous to @xmath164 and is supported on a compact subset of the interior of @xmath33
. given two closed forms @xmath172 with compact support in the interior of @xmath33 , we may view them as smooth forms @xmath173 on @xmath174 by extending them trivially . as the isomorphism @xmath175 is induced by restriction , it maps the cohomology classes @xmath176 , [ \tilde \omega_2]\in h^1(\partial m_k ; \rho_k)$ ] to @xmath177 , [ \omega_2]\in h^1(s;\rho_0)$ ] .
thus the cup product @xmath178 of their cohomology classes is @xmath179\cup_k ' [ \omega_2]= [ \tilde \omega_1]\cup_k [ \tilde \omega_2]= \int_{\partial m_k } \det ( \tilde\omega_1\wedge\tilde \omega_2),\ ] ] where @xmath180 denotes the pairing associated to the determinant evaluated at the exterior product of the forms . by construction ,
the support of @xmath181 and @xmath182 is contained in @xmath33 , thus the previous integral can be evaluated on @xmath33 instead of @xmath174 , and in particular it does not depend on @xmath183 . by the previous lemma
, we may omit the subindex and just write @xmath184 [ lem : isotropic ] the image of @xmath185 is an isotropic plane for the product ( [ eqn : ss ] ) .
the fact that the image is a plane follows from lemma [ lem : cohomm_i ] and its proof .
then , by construction the lemma is equivalent to saying that the image of @xmath186 is an isotropic subspace .
this is well known @xcite , but we sketch the argument for completeness . using the long exact sequence of the pair @xmath187 , we have a commutative diagram with exact rows : @xmath188^{j_k^ * } \ar@{}[d]^*++{\!\!\!\!\times } & h^1(\partial m_k;\rho_k ) \ar[r]^{\delta } \ar@{}[d]^*++{\!\!\!\!\times } & h^2(m_k,\partial m_k;\rho_k ) \ar@{}[d]^*++{\!\!\!\!\times } \\ h^2(m_k,\partial m_k;\rho_k ) \ar[d ] & h^1(\partial m_k;\rho_k ) \ar[l]_{\delta } \ar[d ] & h^1(m_k;\rho_k ) \ar [ l]_{j_k^ * } \ar[d]\\ \mathbf c & \mathbf c & \mathbf c } \ ] ] where the columns denote the pairings .
this implies @xmath189 hence @xmath190 because @xmath191 , and we are done .
let @xmath192 denote the projective space on @xmath84 .
isotropic planes of @xmath160 are in bijection with projective lines in the quadric @xmath193 since @xmath194 is a nondegenrate paring , @xmath195 is the standard quadric , which is a ruled surface , with two rulings .
we recall next its basic properties .
[ prop : ruling ] there are two disjoint families of projective lines @xmath196 and @xmath197 in @xmath195 such that : * every line in @xmath195 belongs to either @xmath196 or @xmath197 . *
every point in @xmath195 belongs to precisely one line in @xmath196 and one in @xmath197 . * two lines in @xmath195 intersect if , and only if , one is in @xmath196 and the other one is in @xmath197 . * embedding those spaces of lines in the projective grassmannian , @xmath198 .
we shall also use the action of @xmath199 , the isometry group of @xmath160 equipped with the quadratic form @xmath194 in ( [ eqn : ss ] ) . recall the isomorphism : @xmath200 after projectivizing , this induces an isomorphism : @xmath201 [ prop : product ] the action of @xmath202 is equivalent to the product action on @xmath203 ( by an equivalence that preserves the product ) .
see @xcite for a proof .
it follows from this proposition that @xmath204 acts trivially on @xmath197 , and @xmath205 acts trivially on @xmath196 .
consider the involutions @xmath206 , @xmath207 and @xmath208 of the conway sphere @xmath3 , as in figure [ fig : involutions ] .
for @xmath6 let @xmath209 , satisfy @xmath210 , and define @xmath211 the corresponding actions on @xmath160 .
the induced maps @xmath212 on @xmath141 lie in one factor @xmath204 or @xmath205 .
in addition all @xmath213 , @xmath214 and @xmath215 lie in the same factor . by looking at the action on @xmath84
, we have @xmath216 , by ( [ eqn : tau2=-id ] ) .
this implies that @xmath212 projects to an involution of @xmath141 preserving @xmath194 , hence to an involution in @xmath202 . notice that if an involution of @xmath202 is nontrivial on each factor , it lifts to an element of @xmath217 whose square is @xmath218 , hence to an involution of @xmath219 . as @xmath220
, we deduce that each @xmath212 projects to an involution of one of the factors of @xmath202 , and is trivial on the other factor .
this proves the first assertion of the lemma . for the last assertion , just use that @xmath221 . hence , up to permuting @xmath197 and @xmath196 ,
we get : [ coro : tauasigmab ] * @xmath213 , @xmath214 and @xmath215 act trivially on @xmath197 .
* there is no point in @xmath196 fixed by all @xmath213 , @xmath214 and @xmath215 .
notice that , for assertion ( ii ) , we use that the subgroup of @xmath222 consisting of three involutions and the identity ( also called the 4-klein group ) has no global fixed point in @xmath223 .
let @xmath224 denote the image .
since @xmath225 , from proposition [ prop : ruling ] ( iii ) we deduce : [ cor : imi1i22 ] either both @xmath226 and @xmath227 belong to @xmath197 , or they both belong to @xmath196 .
either @xmath228 belongs to @xmath197 and , by corollary [ coro : tauasigmab ] , we may apply proposition [ prop : criterion ] , or @xmath228 belongs to @xmath196 . to get rid of this last case
we will use a deformation argument .
the idea is to deform the hyperbolic structure on @xmath229 , so that it matches with another tangle which is invariant under the involutions @xmath230 . by corollary [ coro : tauasigmab](ii ) , the tangle invariant by the involutions satisfies @xmath231 , hence @xmath232 for the deformed structure on @xmath229 , by corollary [ cor : imi1i22 ] .
then we shall use a continuity argument to have the same conclusion for the initial structure on @xmath229 .
next subsection is devoted to this deformation argument .
let @xmath233 be the pair of annuli , one around each arc of @xmath234 .
the pair @xmath235 is a _ pared _ manifold .
[ def : pared ] a _ pared _ manifold is a pair @xmath236 , where @xmath73 is a compact oriented 3-manifold , @xmath237 is a union of tori and annuli , such that * no two components of @xmath238 are isotopic in @xmath239 , * every abelian noncyclic subgroup of @xmath240 is conjugate to a subgroup of a component or @xmath238 , and * there are no essential annuli @xmath241,s^1\times \partial [ 0,1])\to ( m , p)$ ] . we say that a pared manifold @xmath236 is hyperbolic when the interior of @xmath73 admits a complete hyperbolic structure with cusps at @xmath238 .
the rank of the cusp is one for an annulus , and two for a torus .
[ lemma : tangle_hyp ] there exists a pared manifold @xmath242 , such that 1
. @xmath243 is obtained from a @xmath138-tangle : namely @xmath244 is the exterior of two properly embedded arcs in a 3-ball , @xmath245 are the annuli around the arcs of the tangle , and @xmath246 .
2 . for @xmath6
, @xmath247 extends to an involution of @xmath242 .
3 . the pared manifolds @xmath243 and @xmath248 are both hyperbolic .
we take @xmath244 to be the exterior of a simple 2-tangle that is symmetric with respect to @xmath206 , @xmath207 and @xmath208 . here
simple means that @xmath244 is irreducible , @xmath249-irreducible , atoroidal and anannular . in @xcite
, wu gives a criterion for deciding when a rational tangle is simple and provides examples of simple tangles with the required symmetries .
in particular , the pared manifold @xmath243 admits a hyperbolic structure with totally geodesic boundary in @xmath250 ( and rank one cusps in @xmath245 ) .
since @xmath243 is simple and @xmath235 hyperbolic , standard arguments in 3-dimensional topology prove that @xmath251 is irreducible , acylindrical , atoroidal and not seifert fibered ( @xmath33 should be horizontal in a seifert fibration ) , hence hyperbolic . the variety or representations of @xmath50 to @xmath0 is denoted by @xmath252 and it is an algebraic subset of affine space @xmath253 .
[ lemma : almostequal-2 ] if @xmath34 , then @xmath228 belongs to @xmath197 .
we connect @xmath254 to @xmath255 , a lift of the holonomy representation of @xmath229 that matches with the tangle @xmath244 of lemma [ lemma : tangle_hyp ] , which is a symmetric tangle .
namely we want to find a path or representations @xmath256&\to & r(m_2 ) \\ t & \mapsto & \varphi_t \end{array}\ ] ] that satisfies : * @xmath257 . *
@xmath258 $ ] , @xmath259 is the lift of the holonomy of a hyperbolic structure on @xmath229 , with rank one cusps at the arcs @xmath234 , and satisfying @xmath260 * @xmath258 $ ] , @xmath261 .
* @xmath262 is the lift of the holonomy of a hyperbolic structure on @xmath229 that matches with @xmath244 in lemma [ lemma : tangle_hyp ] .
assuming we have this path of representations , then since @xmath244 is @xmath206 and @xmath207-invariant , the image of @xmath263 is a subspace @xmath212-invariant .
hence the image of @xmath264 must be contained in @xmath197 , by corollary [ coro : tauasigmab ] ( ii ) .
this implies that for this hyperbolic structure @xmath265 by corollary [ cor : imi1i22 ] .
now , since there exists the path @xmath259 , the ruled quadric of @xmath266 is also deformed continuously ( notice that as @xmath267 is irreducible and @xmath259 of a meridian has trace @xmath268 , by ( iii ) , lemmas [ lem : cohoms ] , [ lemma : spartial ] , and [ lem : isotropic ] apply to @xmath266 ) .
hence along the deformation , the image of @xmath131 is contained in @xmath197 , as @xmath269 .
hence @xmath270 as claimed .
let us justify the existence of the path @xmath271 between @xmath47 and @xmath272 . if both @xmath273 and @xmath274 are geometrically finite , then they can be connected along the space of geometrically finite structures of the pared manifold , because by ahlfors - bers theorem this space is isomorphic to the teichmller space of @xmath33 , cf .
in addition , this is an open subset of the variety of representations , and since the dimension of de cohomology is upper semi - continuous ( it can only jump in a zariski closed subset ) , ( iii ) can be achieved by avoiding a proper zariski closed subset ( hence of real codimension @xmath275 ) .
if any of @xmath273 and @xmath274 is not geometrically finite , then it lies in the closure of geometrically finite structures ( cf .
@xcite though this is a particular case of the density theorem ) , thus there is still a path in the space of representations satisfying ( ii ) and ( iii )
. by lemma [ lemma : almostequal-2 ] , corollary [ cor : imi1i22 ] and proposition [ prop : criterion ] , @xmath276 we shall prove that there is also equality of signs : [ proposition : equal-2 ] if @xmath34 , then @xmath277 to remove the sign ambiguity , we use again the deformation @xmath259 of the proof of lemma [ lemma : almostequal-2 ] . since @xmath258 $ ] , @xmath259 satisfies the sufficiency criterion of proposition [ prop : criterion ] , the eigenvalues of @xmath212 restricted to the image of @xmath131 belong to @xmath278 , and they do not change as we deform @xmath279 , hence the determinant of @xmath212 restricted to the image of @xmath131 is @xmath280 , because this holds for @xmath281 ( as @xmath244 is @xmath230-invariant ) .
when @xmath282 , then lemma [ lemma : spartial ] does not apply , hence we can not use the argument of section [ sec : trace-2 ] . recall that @xmath283 denotes the variety of representations of @xmath27 in @xmath0 .
we will consider representations @xmath284 to which the arguments of section [ sec : trace-2 ] apply and such that @xmath285 converges to @xmath20 in @xmath286 , as @xmath287 .
let @xmath288 be a lift of the holonomy with @xmath282 .
by thurston s hyperbolic dehn filling and for @xmath289 large enough , the orbifold with underlying space @xmath290 , branching locus @xmath4 and ramification index @xmath142 is hyperbolic .
it induces a representation of @xmath27 in @xmath291 that lifts to @xmath284 .
the lift satisfies @xmath292 , and there is precisely one lift for every choice of sign .
we chose the lift satisfying @xmath293 .
[ prop : orbifold ] for @xmath289 large enough , there exist @xmath284 which is a lift of the holonomy of the orbifold with underlying space @xmath290 , branching locus @xmath4 and ramification index @xmath142 , so that @xmath294 .
these orbifolds can also be considered for the mutant knot , and there exist the corresponding mutant representations @xmath295 namely , the lifts of the holonomies of the orbifold structures on @xmath9 are the `` mutant representations '' of @xmath285 . moreover , @xmath296 . [
lem : orbif ] for @xmath289 large enough , @xmath297 and @xmath298 we use semi - continuity of cohomology to say that @xmath297 .
more precisely , the dimension of cohomology is an upper semi - continuous function on @xmath286 , cf .
@xcite . hence , as @xmath20 and @xmath299 are acyclic ,
then all representations in a zariski open subset containing @xmath20 and @xmath299 are acyclic , and so are @xmath285 and @xmath300 , as claimed . since @xmath285 and @xmath300 are acyclic , @xmath301 .
in addition , up to conjugation , @xmath302 hence @xmath22 has no @xmath303-invariant proper subspaces and lemma [ lemma : spartial ] applies .
thus the pairing ( [ eqn : ss ] ) , lemma [ lem : isotropic ] and all the results of section [ subsection : perfectpairing ] hold true for @xmath285 .
to conclude , for the deformation argument , we use that @xmath285 is the lift of the holonomy of an orbifold . instead of working with pared structures on @xmath235 , we work with orbifold structures with underlying space the ball @xmath37 , branching locus @xmath234 and branching index @xmath142 .
the results on the space of hyperbolic structures ( geometrically finite or infinite ) apply , and we may use the deformation argument of lemma [ lemma : almostequal-2 ] . by proposition [
prop : orbifold ] and lemma [ lem : orbif ] , by taking the limit when @xmath287 we get : [ cor : equal-2 ] if @xmath35 , then @xmath277 the proofs for @xmath282 and @xmath34 are quite different , because @xmath304 is non zero when the trace is + 2 , and vanishes when it is -2 .
the generic case is @xmath305 .
the proof of section [ sec : trace-2 ] applies to the following situation .
[ prop : generic ] let @xmath15 be a representation satisfying : 1 .
@xmath306 ; 2 . @xmath149
@xmath20 restricted to @xmath48 is irreducible ; 4 . the representation @xmath20 is in the same irreducible component of @xmath286 as some representation such that the image of @xmath131 is contained in @xmath307
. then @xmath308 .
[ cor : generic ] for a generic representation @xmath20 of the irreducible component of @xmath286 that contains a lift of the holonomy , @xmath309 .
the holonomy representation of a hyperbolic knot has two lifts to @xmath0 , each one with a different sign for the image of the meridian .
do they belong to the same irreducible component of the variety of representations ?
this happens to be true for instance if the component of the variety of representations contains a dihedral representation , as this is a ramification point or the map from the variety of representations in @xmath0 to those in @xmath222 .
let @xmath310 and @xmath3 be the kinoshita - terasaka knot and the conway knot respectively .
it is well known that they are mutant hyperbolic knots . using the snap program @xcite , based of j. weeks snappea @xcite
, we have obtained all the necessary information to compute their torsion .
the fundamental groups of these knots have the following presentations : @xmath311 as usual , capital letters denote inverse . the image of the holonomy representation is contained in @xmath312 where @xmath313 is the number field generated by a root @xmath164 of the following polynomial : @xmath314 the torsions then are elements of @xmath313 . in order to express elements in @xmath313 , we use the @xmath315-basis @xmath316 .
tables [ table : tra2 ] and [ table : tra-2 ] give the coefficients of the torsions of @xmath310 and @xmath3 with respect to this @xmath317-basis . on each table
, the first column gives the element of the basis .
we let @xmath142 denote the dimension of the irreducible representation of @xmath0 used to compute the torsion , and the tables show the values for @xmath147 ( i.e. the standard representation ) , but also @xmath318 and @xmath319 . in order to compare them ,
the coefficients of the torsion for kinoshita - terasaka ( @xmath310 ) and conway ( @xmath3 ) knots are tabulated side by side .
we give a table for each lift of the holonomy , one when the trace of the meridian is @xmath138 ( table [ table : tra2 ] ) and another when it is @xmath268 ( table [ table : tra-2 ] ) .
cc|r|r|r|r|r|r| & & & & + & & & & & & + & @xmath320 & @xmath320 & @xmath321 & @xmath321 & @xmath322 & @xmath323 + & @xmath324 & @xmath324 & @xmath325 & @xmath325 & @xmath326 & @xmath327 + & @xmath328 & @xmath328 & @xmath329 & @xmath329 & @xmath330 & @xmath331 + & @xmath332 & @xmath332 & @xmath333 & @xmath333 & @xmath334 & @xmath335 + & @xmath336 & @xmath336 & @xmath337 & @xmath337 & @xmath338 & @xmath339 + & @xmath340 & @xmath340 & @xmath341 & @xmath341 & @xmath342 & @xmath343 + & @xmath344 & @xmath344 & @xmath345 & @xmath345 & @xmath346 & @xmath347 + & @xmath348 & @xmath348 & @xmath349 & @xmath349 & @xmath350 & @xmath351 + & @xmath352 & @xmath352 & @xmath353 & @xmath353 & @xmath354 & @xmath355 + & @xmath356 & @xmath356 & @xmath357 & @xmath357 & @xmath358 & @xmath359 + & @xmath360 & @xmath360 & @xmath361 & @xmath361 & @xmath362 & @xmath363 + cc|r|r|r|r|r|r| & & & & + & & & & & & + & @xmath364 & @xmath364 & @xmath365 & @xmath366 & @xmath367 & @xmath368 + & @xmath369 & @xmath369 & @xmath370 & @xmath371 & @xmath372 & @xmath373 + & @xmath374 & @xmath374 & @xmath375 & @xmath376 & @xmath377 & @xmath378 + & @xmath379 & @xmath379 & @xmath380 & @xmath381 & @xmath382 & @xmath383 + & @xmath384 & @xmath384 & @xmath385 & @xmath386 & @xmath387 & @xmath388 + & @xmath389 & @xmath389 & @xmath390 & @xmath391 & @xmath392 & @xmath393 + & @xmath394 & @xmath394 & @xmath395 & @xmath396 & @xmath397 & @xmath398 + & @xmath399 & @xmath399 & @xmath400 & @xmath401 & @xmath402 & @xmath403 + & @xmath404 & @xmath404 & @xmath405 & @xmath406 & @xmath407 & @xmath408 + & @xmath409 & @xmath409 & @xmath410 & @xmath411 & @xmath412 & @xmath413 + & @xmath414 & @xmath414 & @xmath415 & @xmath416 & @xmath417 & @xmath418 + of course , for @xmath147 and for any lift of the holonomy , the torsion of @xmath310 and the torsion of @xmath3 is the same . notice that for the 4-dimensional representation , they are also the same for one lift but different for the other , and that they differ for both lifts when we use the 6-dimensional representation . as said in the introduction , when @xmath147 , these had been computed by dunfield , friedl and jackson in @xcite .
they computed numerically a twisted alexander invariant ( which are not mutation invariant ) for all knots up to 15 crossings , and the torsions computed here are just the evaluations at @xmath28 .
the mutation considered in this paper is called @xmath419-mutation , because the involved surface is planar and has 4 boundary components . by tubing along invariant arcs of the knot ,
this is a particular case of the so called @xmath420-mutation , namely , the mutation along a closed surface of genus 2 and using the hyperelliptic involution . in @xcite ruberman proved that a @xmath420-mutant of a hyperbolic manifold is again hyperbolic .
the behavior of invariants under @xmath420-mutation has been investigated by many authors , see @xcite for instance .
unfortunately , our arguments do not apply , as in section [ sec : trace-2 ] we require two involutions , and in genus two mutation we can use only the hyperelliptic involution .
so we arise : the three dimensional representation of @xmath0 is conjugate to the adjoint representation in the automorphism group of the lie algebra @xmath422 .
the representation @xmath423 is not acyclic , but a natural choice of basis for homology has been given in @xcite , hence its torsion is well defined .
moreover , we have : the proof is straightforward , as @xmath425 is the cotangent space to the variety of characters of @xmath33 , and the action of the hyperelliptic involution is trivial on the variety of characters of @xmath33 .
we have seen that if we compose the lift of the holonomy with the 6-dimensional representation of @xmath0 ( or the 4-dimensional one when the trace of the meridian is @xmath268 ) , the torsion is not invariant under @xmath420-mutation , as it is not invariant under @xmath419-mutation , see the example of the previous section . to conclude , we notice that our arguments do not apply if we tensorize @xmath15 with the abelianization map @xmath426 .
this torsion gives the twisted polynomial in @xmath427 $ ] studied in @xcite , where it is proved not to be mutation invariant .
if we could apply the arguments of this paper , then we would be in the generic situation of proposition [ prop : generic ] and section [ sec : trace-2 ] , because @xmath428 .
notice that two of the involutions in figure [ fig : involutions ] reverse the orientation of the meridian , and only one preserves them .
thus we can use only a single involution , and at least two involutions are required in our argument from section [ sec : trace-2 ] , more precisely in corollary [ coro : tauasigmab](ii ) . | given a hyperbolic knot , we prove that the reidemeister torsion of any lift of the holonomy to @xmath0 is invariant under mutation along a conway sphere . |
when a pure state , represented by a density matrix @xmath2 , is transmitted along a noisy channel , it is mapped into a mixed state @xmath3 .
the entropy of the initial pure state is necessarily zero , i.e. , @xmath4 since @xmath5 and so the only eigenvalues of @xmath2 are @xmath6 and @xmath7 .
however , the entropy @xmath8 $ ] of the mixed state which emerges need not be zero .
one seeks states @xmath2 which minimize the effect of the noise in the sense of minimizing the entropy @xmath8 $ ] of the state that emerges from the channel .
there are a number of reasons for studying such states , most notably the connection between minimizing entropy and maximizing channel capacity , which will be discussed in section [ subsect : prelim.cap ] . however , in this paper we focus attention on the entropy . the noise , which results from interactions between the states in a hilbert space @xmath9 and the environment , is represented by the action of a completely positive , trace - preserving map @xmath10 on the trace class operators in @xmath11 .
we use the term _ stochastic _ to describe such maps .
( following a similar use by alberti and uhlmann @xcite , this terminology was used by petz , @xcite and reflects the fact that @xmath10 is the non - commutative analogue of the action of a column stochastic matrix on a probability vector . )
we restrict attention to two - level quantum systems in which case @xmath12 or tensor products of @xmath1 .
a stochastic map @xmath10 acting on states in @xmath1 , can be naturally extended to tensor products , e.g. , @xmath13 acting on states on @xmath14 etc . , and leads to questions about the additivity of the minimal entropy and capacity of product channels .
in particular , shor has conjectured that the minimal entropy is additive .
we were led independently to this conjecture because it would imply additivity of channel capacity for _ unital _ stochastic maps , i.e. , maps which take the identity operator to itself so that @xmath15 .
although we have a convincing argument that our results for entropy , unlike those for channel capacity , extend to non - unital maps , we focus most of our attention on unital maps . in the last section we briefly consider non - unital maps .
recall that every completely positive map @xmath10 can be represented ( non - uniquely ) in the kraus form [ eq : kraus ] ( ) = _ k a_k^ a_k .
every map @xmath10 representing noisy evolution in a quantum channel must preserve the trace of @xmath2 , since @xmath3 is also a state . in terms of the kraus operators , the condition that @xmath10 be stochastic , that is completely positive and trace preserving , is [ eq : cond.tracepres ] ( ) = & & _
k=1^n a_k a_k^ = i .
the map @xmath10 is _ unital _ if @xmath15 , that is if @xmath10 maps the identity operator to itself . in terms of the kraus operators ,
the condition for @xmath10 to be completely positive and unital is [ eq : cond.unital ] ( i ) = i & & _ k=1^n
a_k^ a_k = i . a sufficient condition that a stochastic map be unital
is that the kraus operators are self - adjoint , i.e. , @xmath16 .
this condition is not necessary ; for example , a double - stochastic matrix which is not symmetric corresponds to a unital stochastic map which is not self - adjoint .
henceforth in this paper , `` unital map '' will mean `` unital stochastic map '' unless otherwise stated .
it is worth noting that the kraus operators are self - adjoint if and only if @xmath10 is self - adjoint with respect to the hilbert - schmidt inner product @xmath17
. the dual or adjoint map @xmath18 is then defined by the condition @xmath19^{\dagger } q = \tr p^{\dagger } \phi(q)$ ] .
it is easy to see that if @xmath10 is given by ( [ eq : kraus ] ) then @xmath20 .
in addition , since the dual of any trace - preserving map satisfies @xmath21 , any stochastic map ( considered as a linear map on the space of hilbert - schmidt operators ) has eigenvalue @xmath7 , and hence a fixed point @xmath22 such that @xmath23 . for unital maps ,
the identity is an eigenvector whose orthogonal complement is the set of operators with trace zero .
hence , a unital map also defines a linear map on the traceless part of a density matrix .
by contrast , a non - unital map is only affine on the set of traceless matrices .
this distinction is easily seen for the @xmath1 case when the bloch sphere representation is used as described in section [ sect : notation ] .
appendix c contains a list of examples of unital and non - unital maps .
recall that the entropy of a density matrix can be written in terms of its eigenvalues @xmath24 , namely @xmath25 .
the minimal entropy @xmath26 occurs if and only if one eigenvalue of @xmath2 is @xmath7 and all others @xmath6 ; the maximal entropy ( in @xmath27 dimensions ) of @xmath28 occurs if and only if all eigenvalues are @xmath29 so that @xmath30 .
thus , if @xmath31 , one must have one eigenvalue close to @xmath7 and the others near @xmath6 . hence , states with small entropy are those for which @xmath32 . thus ,
in seeking pure states @xmath2 which have minimal entropy @xmath8 $ ] after emerging from a noisy channel , we are led to seek states for which @xmath33 is maximal . in section [
sect : norm.bnd ] we give a precise definition of the maximal norm and show that it is multiplicative when ( at least ) one channel is a unital map on @xmath1 .
our results suggest that multiplicativity of the maximal norm may hold for general channels ; in fact , we can extend our result to some non - unital channels ( see remarks at the end of section [ sect : norm.bnd ] ) . in any case ,
our results provide strong support for the additivity of minimal entropy for unital channels .
another property of unital channels is that the entropy of a state is non - decreasing under the action of a unital stochastic map .
this follows easily from the fact that the relative entropy h(p , q ) = p [ p - q ] decreases under stochastic maps , i.e. , [ eq : relent.dec ] h[(p),(q ) ] h(p , q ) since @xmath34 if @xmath10 is unital , it follows from ( [ eq : relent.dec ] ) that @xmath8 \geq s(\rho)$ ] .
for a non - unital map , the entropy of a pure state can not decrease ; however , one can have mixed states for which the entropy actually decreases .
we now discuss the information capacity of a noisy quantum channel @xcite , @xcite used for what is sometimes called `` classical '' communication , i.e. , communications in which signals are sent using quantum particles but without additional or prior entanglement between sender and receiver . in the simplest case where no entanglement is used in either the transmission or the measurement , each letter
@xmath35 from the source alphabet is represented by a pure state which we represent by its density matrix @xmath36 on a quantum hilbert space . during transmission
the channel transforms this state into @xmath37 , where @xmath10 implements the noisy interaction between states and the environment .
the map @xmath10 is a completely positive , trace - preserving map on the set of states .
the resulting state @xmath38 is measured , and the outcome determines a letter from the output alphabet . in the general case
this measurement is effected by a positive operator - valued measurement ( povm )
namely , there is a positive operator @xmath39 assigned to each letter @xmath40 of the output alphabet , which together satisfy the constraint @xmath41 .
when the measurement is performed on a state @xmath2 , the result will be @xmath40 with probability @xmath42 .
several definitions of channel capacity have been proposed , corresponding to whether or not entangled states are used for transmission , and whether or not entangled measurements are made by the receiver .
bennett and shor @xcite identify four possibilities , which we denote @xmath43 and @xmath44 where the subscripts @xmath22 and @xmath45 refer to product and entangled processes respectively . in the process described above , with no entanglement at either end , the channel is equivalent to a classical noisy channel with transition probabilities @xmath46
. therefore its maximum rate of information transmission is given by [ shannon ] c_pp ( ) = c_shan ( ) = _ , , e _ i _ j _ i p_ij ( p_ij _ k _ k p_kj ) , where we take the @xmath47 over all probability distributions @xmath48 for the input alphabet , as well as all choices of input states and measurements .
we call this the _ shannon capacity _ of the channel , since it bears closest resemblance to the classical case .
it is reasonable to expect that by transmitting entangled states and by using entangled measurements it may be possible to exceed the shannon capacity for a noisy channel .
the holevo - schumacher - westmoreland theorem @xcite , @xcite provides a closed form expression for the capacity in the case where product states are transmitted at the input and entangled measurements of arbitrary length are allowed at the output : [ hsw ] c_pe = c _ ( ) . here
@xmath49 is the _ holevo capacity _ of the channel : [ holevo ] c _ ( ) = _ , ( s ( ) - _ i _ i s(_i ) ) , where @xmath50 and @xmath51 . the well - known holevo bound states that c _ ( ) c _ ( ) .
holevo @xcite provided examples of channels in which this inequality is strict , i.e. , @xmath52 .
furthermore , it has been shown @xcite that a necessary and sufficient condition for strict inequality is that the output states @xmath53 do not commute .
one important open question is whether or not the holevo capacity can be exceeded when entangled states are used at the input , that is whether @xmath44 exceeds @xmath54 .
this would be equivalent to the superadditivity of the holevo capacity . in this paper
we address this question in the case of messages which are entangled over _ two inputs _ only .
this is equivalent to the question whether @xmath55 exceeds @xmath56 .
holevo @xcite has shown that @xmath57 for the quantum binary channel , but to the best of the authors knowledge there is no known example of a superadditive channel for the holevo capacity .
bruss et al @xcite showed that @xmath58 for the depolarising channel , which is an example of a unital channel . as we will show
, our results strongly suggest that _ if _ the holevo capacity is superadditive then the channel must be non - unital .
we prove several theorems about the minimal entropy and the maximal norm for states of the form @xmath59 , where @xmath10 and @xmath60 are unital stochastic maps on @xmath61 and @xmath62 is an entangled state .
in addition , we explain how these results provide evidence for the conjecture that _ minimal entropy is additive _ for all stochastic maps on @xmath61 .
we also show that this conjecture has important implications for the capacity of unital quantum channels .
in particular , we show that the conjecture implies that if @xmath10 is unital , then the holevo capacity is additive over two inputs , that is @xmath63 . our first theorem concerns the maximal value of @xmath64 as @xmath2 varies over states on @xmath65 .
we will consider the general possibility of two stochastic maps @xmath10 and @xmath60 on @xmath61 and denote their maximal values by @xmath66 and @xmath67 respectively . in theorem [ thm : normbnd ] we prove , under mild conditions on one of these maps , that the maximal value of @xmath68 is @xmath69 , as @xmath62 varies over states on @xmath61 .
that is , the norm of @xmath70 achieves its maximal value on product states @xmath71 , rather than on entangled states . in theorem [ thm : mixing ] , we prove a similar , though slightly weaker , result for the minimal entropy of @xmath70 .
namely , we restrict @xmath62 to the family of entangled states whose reduced density matrices @xmath72 and @xmath73 are such that @xmath74 and @xmath75 have minimal entropy .
then we prove that the minimal entropy of @xmath76 , as @xmath62 varies over this family , is the sum of the minimal entropies of @xmath74 and @xmath75 .
that is , the entropy also achieves its minimal value on product states .
it seems extremely unlikely that we could further decrease the entropy of @xmath70 by using an entangled state whose reduced density matrices do not have minimal entropy themselves .
therefore we believe that the conclusion of theorem [ thm : mixing ] also holds as @xmath62 varies over all entangled states on @xmath61 .
in fact , similar arguments and numerical evidence support an even stronger conclusion , namely that minimal entropy is additive for _ all _ stochastic maps on @xmath61 .
this is the content of the conjecture below .
[ conj : minent ] if @xmath10 and @xmath60 are stochastic maps on @xmath61 , then _ s= _
s. we will discuss the evidence for this conjecture in detail in section [ sect : entanal ] and section [ subsect : nonu.min.ent ] .
we note that p. shor @xcite earlier made a similar conjecture , and together with j. smolin obtained numerical evidence which supports it .
the unital case is particularly important because it yields the following results as immediate corollaries . if @xmath10 is unital , then the holevo capacity is additive , i.e.,@xmath77 .
if @xmath10 is unital , then the holevo capacity can be achieved with orthogonal states .
if @xmath10 is unital , then @xmath78 and @xmath79 .
if @xmath10 is unital , then the shannon capacity is also additive , i.e. , @xmath80 . in section [ subsect : prelim.cap ] we will explain in detail how these corollaries follow if conjecture [ conj : minent ] holds for unital maps .
this paper is organized as follows . in section [
sect : prelim ] we introduce the notation we will use for the stokes parametrization for representing both states and maps in a basis consisting of the identity and pauli matrices , and show how the various conditions of unital , trace - preserving , and complete positivity can be expressed in this representation . with this background , we conclude section [ sect : prelim ] by presenting the arguments leading to the corollaries above .
in section [ sect : norm.bnd ] we prove the multiplicativity of the maximum value of @xmath33 .
section [ sect : entanal ] contains the heart of the paper in which we give the details of the proof of our theorem about additivity of minimal entropy . in section [
sect : non.unit ] we discuss some of the features of non - unital maps using a special subclass and then present the evidence for additivity of minimal entropy in general .
finally , in section [ sect : conc ] we summarize our results and discuss their implications for channel design .
we also include three appendices .
appendix a gives some important background on singular value decompositions and the details needed for the diagonal representation introduced in section [ sect : notation ] .
appendix b gives the details needed to verify the complete positivity conditions of section [ sect : singv.cond ] . in appendix c
we provide a number of examples of different types of channels and show how some familiar examples appear in the representation and notation we use .
recall that the identity and pauli matrices form a basis for @xmath81 so that any @xmath82 matrix @xmath83 can be written as @xmath84 where @xmath85 denotes the vector of pauli matrices and @xmath86 then for @xmath83 = @xmath84 * @xmath83 is self - adjoint @xmath87 is real , i.e. , @xmath88 and @xmath89 ; * @xmath90 ; and * @xmath91 thus , @xmath92 also form a basis for the _ real _ vector space of self - adjoint matrices in @xmath81 and every density matrix can be written in this basis as @xmath93 $ ] with @xmath89 and @xmath94 .
furthermore * @xmath2 is a one - dimensional projection ( or pure state ) @xmath95 every linear map @xmath96 can be represented in this basis by a unique @xmath97 matrix @xmath98 , and @xmath10 is trace - preserving if and only if the first row satisfies @xmath99 , i.e. , @xmath100 where @xmath101 is a @xmath102 matrix ( and @xmath103 and @xmath104 are row and column vectors respectively ) so that @xmath105 the matrix @xmath98 is self - adjoint if and only if @xmath10 is self - adjoint as an operator on @xmath81 with respect to the hilbert - schmidt inner product .
we are interested in those @xmath10 which map a subspace of self - adjoint matrices into itself , which holds if and only if @xmath106 is real .
the map @xmath10 is unital if and only if @xmath107 .
thus , any unital stochastic map @xmath10 acting on density matrices on @xmath1 can be written in the form [ eq : t3rep ] ( ) = , where @xmath101 is a real @xmath102 matrix . using the singular value decomposition
( see appendix a ) , we can write [ svd ] = r s where @xmath108 is a rotation and @xmath109 is self - adjoint . define the map @xmath110 by _ s ( ) = as explained in appendix a , the rotation @xmath108 defines a unitary operator @xmath111 such that for any state @xmath2 , [ eq : def.u ] ( ) = u u^ in this paper we are interested only in the critical values of certain functions of the spectrum of @xmath3 , as @xmath2 varies over the space of states the maximum value of the norm , the minimum value of the entropy .
since a unitary transformation leaves the spectrum unchanged , these are the same for @xmath10 and @xmath110 .
also , since @xmath109 is self - adjoint it can be diagonalized by a change of basis . hence without loss of generality
we need henceforth consider only unital stochastic maps whose matrix @xmath101 defined in ( [ eq : trep ] ) is diagonal , with eigenvalues @xmath112 . as a shorthand
, we will denote this diagonal map by @xmath113 $ ] .
the image of the set of pure state density matrices @xmath114 $ ] ( with @xmath115 ) under the action of @xmath113 $ ] is the ellipsoid [ eq : ellip.unit ] ( ) ^2 + ( ) ^2 + ( ) ^2 = 1 , and the image under the action of @xmath10 is obtained by a further rotation of the ellipsoid , corresponding to the operator @xmath111 in ( [ eq : def.u ] ) .
similar reasoning applies when @xmath10 is non - unital . using ( [ eq : t.svd.nonpos ] ) and ( [ phi.selfadj ] ) from appendix a
, the map @xmath10 can be written in the form @xmath116 where @xmath117 are unitary , @xmath118 is diagonal and @xmath119 is represented by the matrix [ eq : t.nonunit ] = the vector @xmath120 is equal to @xmath121 in the notation of ( [ eq : t.svd.nonpos ] ) . in this case , the image of the set of pure state density matrices @xmath114 $ ] ( with @xmath115 ) under the action of @xmath119 is the translated ellipsoid [ eq : ellip.nonunit ] ( ) ^2 + ( ) ^2 + ( ) ^2 = 1 , and again the image under @xmath10 is a rotation of this
. it will be useful to write out explicitly the action of the diagonal unital map @xmath113 $ ] on a density matrix in the form [ phi ] ( ) & = & + & = & .
the requirement that @xmath10 be stochastic imposes a number of constraints on the matrix @xmath98 .
we describe these in appendix b in which we give explicit formulas for the matrix elements of @xmath98 in terms of the stokes parameterization of the operators @xmath122 . these formulas in turn
imply constraints on the eigenvalues @xmath112 described in the previous section .
let @xmath123 denote the elements of @xmath106 using the convention that @xmath124 .
then the point with coordinates @xmath125 must lie inside a tetrahedron with corners at @xmath126 .
these conditions are equivalent to four linear inequalities which can be written compactly as @xmath127 . ) in the special case where @xmath10 is unital , ( [ eq : fuji.dcond ] ) implies that the eigenvalues ( which are necessarily real ) satisfy latexmath:[\ ] ] using convexity of the function @xmath358 , we can bound ( [ approx ] ) from below by @xmath359 , which demonstrates positivity for @xmath360 small .
similar results are obtained for the other cases . in this section
we extend theorem [ thm : main ] to pure states @xmath361 formed from any entanglement of states of minimal entropy . after a precise statement of this and proof of this extension , we discuss its interpretation and the evidence for more general validity of conjecture [ conj : minent ] .
let @xmath10 be a unital stochastic map .
as explained in appendix a , we can write @xmath362 where @xmath111 is a unitary operator and @xmath110 is self - adjoint .
let @xmath363 .
define [ def : l ] l ( ) = \ { = ( i + n ) | _ s(n ) = n } . in words , @xmath364 is the collection of density matrices which lie in the direction of the largest eigenvalue of @xmath110 . if this largest eigenvalue is non - degenerate , then @xmath364 is a line segment between antipodal points on the bloch sphere . in case of degeneracy
it may be a disk , or even the entire bloch sphere .
if @xmath62 is a density matrix on @xmath365 , we denote by @xmath366 and @xmath367 the reduced density matrices on @xmath1 obtained by taking the indicated partial traces .
[ thm : mixing ] let @xmath10 and @xmath60 be unital stochastic maps .
let @xmath62 be a density matrix on @xmath365 , such that @xmath368 lies in @xmath364 , and @xmath369 lies in @xmath370 . then s ( ) ( _ 12 ) _ s [ ( ) ] + _ s [ ( ) ] .
* proof * : wwe assume wlog that @xmath10 and @xmath60 are diagonal maps in the form @xmath371 $ ] and @xmath372 $ ] . furthermore we can arrange that @xmath373 is the largest eigenvalue , so that @xmath374 $ ] lies in @xmath364 .
similarly we can arrange that @xmath375 is the largest eigenvalue , so that @xmath376 $ ] also lies in @xmath370 .
wlog we can assume that @xmath377 is a pure state .
in the bases which diagonalize @xmath10 and @xmath60 , we have | = a_00 | 00 + a_01 | 01 + a_10 | 10 + a_11 | 11 define the matrix @xmath378 to be a = ( ) then as shown in appendix a , the reduced density matrices are _ 1 = a a^ , _ 2 = ( a^ a ) ^t we obtain the `` schmidt decomposition '' by applying the singular value decomposition to @xmath378 .
the result is a new basis in which @xmath379 has the diagonal form assumed in theorem [ thm : main ] .
this decomposition is obtained by finding unitary operators @xmath380 which diagonalize @xmath381 and @xmath382 respectively , so that @xmath383 is also diagonal . by assumption @xmath368
lies in @xmath364 , and hence so does @xmath381 . if @xmath199 is non - degenerate , then @xmath364 is the line segment consisting of the diagonal density matrices . therefore @xmath381 is also diagonal ,
so @xmath384 is equal to the identity , up to a phase .
if @xmath200 is also non - degenerate then @xmath385 is also proportional to the identity , and hence @xmath62 is already in diagonal form .
the result follows immediately by applying theorem [ thm : main ] . in general either or both of @xmath199 and @xmath200 may be degenerate .
for example , if @xmath199 is 2-fold degenerate , then @xmath364 is a disk containing the @xmath386-axis in the bloch sphere .
so @xmath381 lies in this disk , and hence it is diagonalized by a rotation of the bloch sphere which preserves this disk . by definition ,
such a rotation commutes with the action of @xmath10 , since the plane which contains this disk is an eigenspace of @xmath10 .
hence the unitary operator @xmath384 commutes with @xmath10 .
similarly if @xmath199 is 3-fold degenerate , then every unitary operator commutes with @xmath10 . to apply the argument to @xmath60 ,
note that by assumption @xmath369 lies in @xmath370 . if @xmath200 is non - degenerate , or is 3-fold degenerate , the same argument applies .
if @xmath200 is 2-fold degenerate , then @xmath370 is a disk .
the transpose operation on the bloch sphere is the reflection in the xz - plane , and this does not in general preserve a disk containing the z - axis .
however we have assumed that @xmath60 is diagonal , and hence @xmath370 either lies in the xz - plane or the yz - plane . in both cases
the transpose leaves @xmath370 invariant , and hence the same argument can be applied to deduce that @xmath385 also commutes with @xmath60 .
it follows that @xmath387 is unitarily equivalent to @xmath388 where @xmath389 has the form assumed in theorem [ thm : main ] , and hence the result follows .
* remarks : * 1 .
the set @xmath364 contains the states of minimal entropy for @xmath10 .
theorem [ thm : mixing ] shows that by entangling the minimal entropy states of the individual channels we can not decrease the entropy of the product channel .
since it seems unlikely that entangling states of higher entropy will improve the situation , we present this as strong evidence for our conjecture .
2 . we illustrate theorem [ thm : mixing ] in the case where @xmath113 $ ] is self - adjoint and diagonal , with @xmath390 .
if the @xmath391 is real , i.e. , the matrix @xmath392 is real , then the @xmath82 unitary matrices @xmath384 and @xmath385 which diagonalize @xmath381 and @xmath382 can be chosen real and orthogonal , in which case we emphasize this by writing them as @xmath393 and @xmath394 .
now suppose that our original orthogonal basis @xmath395 on @xmath1 corresponds to the eigenvectors of @xmath396 so that the corresponding pure state projections are @xmath397 $ ] , with @xmath398 corresponding to the `` north pole '' of a sphere .
each unitary @xmath82 matrix @xmath111 can be associated with a real orthogonal @xmath102 matrix , and the effect of @xmath111 on the basis vectors corresponds to a rotation on the sphere or the action of a real orthogonal @xmath102 matrix on @xmath399 when the original @xmath82 matrix is real orthogonal , the corresponding rotation on the sphere reduces to a rotation in the xz - plane .
+ if we now write the unital stochastic map @xmath10 in the form : = ( ) = , then the change of basis is equivalent to replacing @xmath400 by @xmath401 where @xmath402 denotes the @xmath102 orthogonal matrix associated with @xmath403 .
thus , for example , when @xmath404 , @xmath405
. if @xmath400 is diagonal with eigenvalues @xmath406 , then @xmath407 hence every state @xmath391 with real coefficients can be diagonalized , and the matrix @xmath400 is unchanged .
3 . suppose that @xmath10 has one very large singular value and two small ones .
then the unit sphere corresponding to the set of density matrices is mapped into an ellipsoid shaped like a football , and the states of minimal entropy lie at the ends of the football ( see figure [ fig : ellip1 ] )
. we can interpret theorem [ thm : main ] as saying that entangling these minimal entropy states will not decrease the entropy below the sum of the minimum entropies . + now suppose that we always keep two eigenvalues equal , but vary the parameters so that the ends of the football move in until it becomes a sphere and then a `` flying saucer '' ( see figure [ fig : ellip2 ] ) .
the ends of the football have moved in to states corresponding to maximal entropy .
the minimal entropy states now form a great circle . as we explained above , our special form for @xmath408 allows a general entanglement of states corresponding to these great circles of minimal entropy .
yet even this more general entanglement does not decrease the entropy below that of product states .
the discussion above shows that , at least in the case of unital maps , if conjecture [ conj : minent ] does not hold , then the entanglements which use states of higher entropy would achieve a lower entropy on the product space than entanglements of states of minimal entropy .
in addition , such entanglements would have to lower the entropy without increasing the largest eigenvalue of @xmath409 beyond the product value given in theorem [ thm : normbnd ] .
we do not find this plausible .
in this section we give some heuristic evidence to support conjecture [ conj : minent ] . before doing so
, we illustrate the differences between unital and non - unital maps by discussing some of the properties of a special class of maps on @xmath81 .
a non - unital map is one for which @xmath410 .
this means that it takes a randomly distributed alphabet to a non - random distribution .
one would intuitively expect that the maximum capacity would then be achieved for alphabets which are _ not _ evenly distributed .
although this is true classically , it need not be true for quantum stochastic maps as shown by the example below .
it follows from equation ( [ eq : trep ] ) that a unital stochastic map defines a linear map on the subspace of matrices with trace zero . however , a non - unital map is affine when restricted to this subspace .
we now consider non - unital maps which correspond , in the notation of section [ sect : notation ] to the matrix [ eq : phi.fuchs ] = with @xmath411 .
this is easily seen to yield the map [ eq : spec.nonunit ] ( ) = .
if @xmath412 and @xmath413 , this is equivalent to the `` splaying '' channel introduced by fuchs @xcite to demonstrate that there exist stochastic maps for which the holevo capacity ( [ holevo ] ) is achieved only by non - orthogonal states .
the case @xmath414 was considered in @xcite in a different context .
the case @xmath415 is essentially classical , i.e. , if @xmath2 is restricted to the subset of states of the form @xmath416 $ ] , the action of @xmath417 is equivalent to the action of the column stochastic matrix @xmath418 on the probability vector @xmath419 since equality in ( [ eq : af.nonunit ] ) holds for fuchs example , it can be regarded as an extreme case . because @xmath420 for this class of maps , they map the unit sphere of density matrices into the ellipse in the x - z plane satisfying the equation [ eq : ellipse ] + = 1 in the special cases , @xmath415 and @xmath414 , these ellipses become vertical and horizontal line segments respectively . in the classical case ( @xmath415 )
it is not hard to show that the maximal capacity is _ never _ achieved for @xmath421 .
one has only two pure states @xmath422 $ ] ( which _ are _ orthogonal ) for which @xmath423 $ ] is not identical .
by contrast , for the non - unital quantum case with @xmath424 , it appears , in general , that maximal capacity is achieved at @xmath421 and with non - orthogonal states .
moreover , these non - orthogonal states need not correspond to the minimal entropy states .
some insight into these observations can be obtained by looking at the ellipse ( [ eq : ellipse ] ) in figure [ fig : fuchs ] corresponding to fuchs channel .
( fuchs @xcite showed explictly that non - orthogonal states are required to achieve maximum capacity , and that the maximal capacity achievable with orthogonal states occurs for @xmath421 . )
the endpoints of the ellipse ( denoted @xmath425 ) correspond to @xmath426 \big ) $ ] and have entropy @xmath427 $ ] while the minimal entropy states ( denoted @xmath428 ) correspond to the images of @xmath429 \right)$ ] and have entropy @xmath430 $ ] .
note that this is the point at which the ellipse meets the circle @xmath431 , which is a level set for the entropy on the bloch sphere .
the states @xmath432 $ ] are the only pair of orthogonal states with identical entropy . if one tries to move from @xmath433 toward @xmath434 to lower the entropy from one of a pair of orthogonal states , the other orthogonal state must move along the ellipse away from @xmath435 , down and closer to the origin and hence has a higher entropy than @xmath435 .
explicit computation shows that the entropy price paid by moving away from @xmath435 is greater than that gained by moving @xmath433 toward @xmath434 . for any pair of states @xmath436 ,
the state @xmath437 which occurs in ( [ holevo ] ) lies along the line segment between @xmath438 and @xmath439 . for the points
@xmath425 it is easy to see that the maximum capacity will occur for @xmath421 .
if one does not require orthogonal states , then simultaneously moving both @xmath425 toward @xmath428 to decrease the entropy seems advantageous .
however , this will also decrease the entropy of the convex combination @xmath440 which _ increases _ the capacity .
hence , maximal capacity does not occur at the minimal entropy states but at states on the ellipse which lie between @xmath428 and @xmath425 .
( as long as the move is symmetric , maximal capacity will occur at @xmath421 .
symmetry suggests this is optimal , but that does not seem to have been proved explicitly . )
one expects similar behavior for any map @xmath10 of the form ( [ eq : phi.fuchs ] ) for which @xmath441 .
when @xmath442 the major axis of the ellipse lies along the x - axis , there is only one state of minimal entropy at the `` top '' of the ellipse , and one expects the channel to behave more and more like a classical channel as the ratio @xmath443 increases . when @xmath414 , the ellipse becomes a horizontal line so that the minimal entropy states and the endpoints of the ellipse coincide .
hence , the maximal capacity is again achieved for orthogonal states . in the limiting case @xmath444 ,
the endpoints of the ellipse lie on the unit circle of pure states and hence , have entropy zero
. however , the capacity does _ not _ achieve its maximum value of @xmath445 but instead has the value @xmath446 . as was noted earlier
, @xmath10 always has a fixed point . for channels of the form ( [ eq : phi.fuchs ] ) , this fixed point is at @xmath447 .
however , we have been unable to attach any significance to the fixed point .
( for fuchs channel , the fixed point is at @xmath448 $ ] ; however , this seems coincidental . ) at first glance , it might seem that the price paid for the versatility of non - unital channels is too great .
if we fix @xmath449 and @xmath280 , then when @xmath411 the requirement ( [ eq : cpmcond.nonunit ] ) implies that @xmath194 must be smaller ( i.e. , `` noisier '' ) than for the corresponding unital channel .
for example , using fuchs values @xmath450 , one finds that @xmath451 is optimal and corresponds to the least `` noisy '' direction .
however , if @xmath452 , one could increase this to @xmath453 with corresponding decrease in noise so that the minimal entropy ( which comes from the states @xmath454 $ ] ) is @xmath455 . for the non - unital channels these same states would yield an entropy of only @xmath456 .
however , the states @xmath457 $ ] will emerge with entropy @xmath458 .
the decrease in eigenvalues of the part of @xmath10 corresponding to the restriction to matrices of trace zero , is overcome by the contribution to the emerging state of the non - unital part of the map .
we expect that this behavior is generic for non - unital maps
. recall the discussion in section [ sect : notation ] . for unital maps ,
the generic situation is that there are two states of minimal entropy corresponding to the endpoints of the major axis . in the generic situation for a non - unital map ,
unless @xmath459 is perpendicular to the major axis , the ellipsoid in ( [ eq : ellip.nonunit ] ) will have _ only one _ state of minimal entropy .
hence _ any _ entanglement will require mixing with a state of higher entropy . in the case of a map ( such as the fuchs map ( [ eq : spec.nonunit ] ) with @xmath460 ) for which the translation @xmath459 is perpendicular to the ( non - degenerate ) major axis of the ellipsoid , there will be two states of minimal entropy .
however , because the translated ellipsoid is not centered at the origin , these two states will not be the images of two orthogonal states , but rather the images of two non - orthogonal pure states corresponding to vectors @xmath461 and @xmath462 .
this suggests that the best candidate for a minimal entropy entangled state should have the form @xmath463 .
however , after changing to orthogonal bases and using the svd decomposition ( [ eq : schmidt ] ) , this can be rewritten in the form @xmath464 where @xmath465 and , at most , only _ one _ state in each of the pairs @xmath466 and @xmath467 can equal either @xmath468 or @xmath469 .
hence any entanglement must include states which are mapped into states of higher entropy under the action of @xmath470 .
thus one expects states of minimal entropy under the action of @xmath471 to have the product form @xmath472 . in the case of two - fold degeneracy ( of the form @xmath473 )
, a shift orthogonal to both major axes will yield a circle of states of minimal entropy ; however , that circle will _ not _ correspond to a `` great circle '' but rather a circle of constant latitude .
such a circle never includes the image of two orthogonal states and hence the argument above still holds . in the case of three - fold degeneracy ,
the image will be a sphere and _ any _ shift will again yield only a single state of minimal entropy .
since the maps @xmath10 and @xmath470 differ only by rotations of the bloch sphere , the same conclusions hold for @xmath10 .
moreover , our analysis suggests that for _ any _ pair of maps @xmath10 and @xmath60 , the states which yield minimum entropy under @xmath474 will be simple product states , regardless of whether one or both are non - unital .
our main result in this paper is that channel capacity is additive for unital channels , where _ unital _ means that the channel maps the totally random input state ( whose density matrix is proportional to the identity ) into itself ( for example , every self - adjoint channel is unital ) . specifically , we present strong evidence that both the shannon capacity ( no entanglement for input states or output measurements ) and the holevo capacity ( inputs unentangled , but output measurements may be entangled ) are additive over two uses of a two - dimensional unital channel .
this is the first result that establishes additivity of channel capacity for a broad class of quantum channels .
we show that the problem reduces to finding the states of _ minimal entropy _ for two uses of the channel . if these minimal entropy states are _ product states _ , then this implies additivity of capacity .
we then prove that one of two possibilities occurs : either these minimal entropy states are product states , or else they are entangled states whose reduced density matrices have greater than minimal entropy .
we argue that the latter case is very unlikely ( numerical experiments confirm this ) , and so conclude that the former is true . as further supporting evidence , we prove that the maximal norm of states which emerge from the channel is multiplicative over two channel uses .
our results rely heavily on the stokes parametrization and the properties of the image of the bloch sphere under stochastic maps . in order to extend them to higher dimensions
, we would need an effective parameterization of the subspace of matrices of trace zero in higher dimensions .
one can always write a density matrix in @xmath475 as @xmath476 $ ] where @xmath477 . however , for @xmath478 , we do not know what the analog of the bloch sphere looks like .
we know only that its boundary corresponds to those @xmath479 with eigenvalues @xmath480 which is not the analogue of the surface of a sphere . without knowing the geometry of this region , we can hardly hope to answer the important question of how it transforms under maps of the form @xmath156 .
thus , we have been forced to use indirect methods to reach conclusions about the states of minimal entropy emerging from @xmath156 .
our results have implications for the design of communication channels .
ideally , one wants to eliminate all noise .
however , this will not be practical and one wants to know how best to allocate resources . in the case of unital channels ,
minimal entropy and maximal capacity are achieved if signal codes are chosen to correspond to the least noisy `` direction '' or polarization .
( here , we use `` direction '' in the sense of the bloch sphere or maximum @xmath24 in our notation .
this is unrelated to direction of signal transmission . ) hence , if maximizing capacity is the primary goal , then it would seem sufficient to minimize noise in only one direction .
even if the orthogonal directions are extremely noisy , signals sent using optimal codes will not be affected . however , in this case `` classical communication '' becomes truly classical . if codes are restricted to one direction , then one is back in the classical situation with one choice for encoding @xmath6 and @xmath7 .
one has effectively lost the versatility of rotating the code basis as a tool for such purposes as signal encryption .
non - unital channels have far more versatility , some aspects of which were discussed briefly in section [ sect : nonunit.sub ] . much more work needs to be done analyzing the properties of non - unital channels .
thus far , most authors have looked for examples of particular maps which illustrate particular facets of stochastic maps ( such as fuchs @xcite example demonstrating the possibility of maximizing capacity with non - orthogonal states ) .
our approach has been to try to find parameters which characterize subclasses of stochastic maps with certain properties . as summarized in appendix c ,
most of the known examples of noisy channels can easily be shown to belong to one of the groups discussed in this paper .
a complete analysis of non - unital maps would seem to require an extension of conditions of the type ( [ eq : af.nonunit ] ) to general maps of the form ( [ eq : t.nonunit ] ) with @xmath481 .
in this paper we make repeated , and sometimes subtle , use of the singular value decomposition ( svd ) of matrices . in view of this , and of some confusion in the literature about the connection between the svd and what is often referred to as the `` schmidt '' decomposition , we provide a brief summary and review of these closely related decompositions and their connection to the better known polar decomposition .
we begin with the polar decomposition which can be extended to bounded , and even some unbounded , operators @xcite on a hilbert space .
[ thm : polar ] _ ( polar decomposition ) _ any @xmath482 matrix @xmath378 can be written in the form @xmath483 where the @xmath184 matrix @xmath484 is positive semi - definite and the @xmath482 matrix @xmath111 is a partial isometry . the term _ partial isometry
_ means that @xmath485 ( or , equivalently , @xmath486 ) is a projection . in general , @xmath111 need not be unique but can be uniquely determined by the condition @xmath487 .
if @xmath378 is a square @xmath184 matrix , then @xmath111 can instead be chosen ( non - uniquely ) to be unitary .
since @xmath488 is self - adjoint , it can be written as @xmath489 where @xmath118 is a diagonal matrix with non - negative entries and @xmath490 is unitary . inserting this in theorem [ thm : polar ] with
@xmath111 chosen to be unitary yields the svd since @xmath491 is also unitary .
[ thm : svd ] _ ( singular value decomposition ) _ any @xmath184 matrix @xmath378 can be written in the form @xmath492 with @xmath493 unitary and @xmath118 a positive semi - definite diagonal matrix .
the non - zero elements of @xmath118 are called the _ singular values _ of @xmath378 .
they are easily seen to be the eigenvalues of @xmath488 and , hence , their squares yield the non - zero eigenvalues of @xmath494 . as an immediate corollary
, one finds that @xmath494 and @xmath495 are unitarily equivalent and that @xmath490 and @xmath496 are , respectively , the unitary transformations that diagonalize @xmath494 and @xmath495 .
these results can be extended to non - square matrices if the requirement that @xmath497 be unitary is relaxed to partial isometry . using the notation of section [ sect : notation ]
, we can apply the svd to the @xmath102 matrix @xmath101 which corresponds to the restriction of the stochastic map @xmath10 to the subspace of matrices with trace zero . because @xmath101 is real , the matrices @xmath497 can be chosen to be real orthogonal so that we can write [ eq : t.svd ] = o_1 _ 2^t where @xmath498 , @xmath499 are orthogonal and the superscript @xmath400 denotes transpose .
now , every @xmath102 orthogonal matrix is either a rotation , or the product of a rotation with the inversion @xmath500 .
hence we can rewrite ( [ eq : t.svd ] ) as [ eq : t.svd.nonpos ] = r_1 ( d ) r_2^t = ( r_1 r_2^t ) r_2 ( d ) r_2^t where @xmath501 and @xmath502 are rotations , and we conclude that @xmath101 can be written as [ eq : t.selfadj ] = r s where @xmath109 is self - adjoint and @xmath108 is a rotation . if @xmath10 is unital , define the map @xmath470 by [ eq : phi.selfadj ] _ s
( w_0 i + ) = w_0 i + s since every rotation is implemented by a unitary transformation on @xmath1 , there is a unitary operator @xmath111 such that for any state @xmath2 , ( ) = u _ s ( ) u^ for non - unital maps , a similar argument can be used to show that any stochastic map has the form ( [ eq : trep ] ) where the restriction of @xmath10 to the matrices with trace zero has the form ( [ eq : t.selfadj ] ) and @xmath503 , i.e. @xmath504 where [ phi.selfadj ] _ s ( w_0 i + ) = w_0 i + ( r^t * t * + s ) and @xmath108 is the rotation on @xmath505 corresponding to @xmath111 . by construction , either @xmath109 or @xmath506 is positive semi - definite . however , if @xmath507 , then composing @xmath508 with the diagonal map @xmath509 merely changes the signs of two of the diagonal elements of @xmath118 .
since @xmath510 is also a unitary map , we can drop the restriction that @xmath109 be semi - definite by modifying @xmath111 if necessary .
this is useful because , as we will see in the next section , the general conditions on the eigenvalues of a matrix @xmath109 corresponding to a self - adjoint map @xmath10 include the possibility of negative and positive eigenvalues .
we can summarize this discussion in the following [ thm : polar.phi ] any stochastic map @xmath10 on @xmath81 can be written in the form @xmath504 where @xmath111 is unitary and @xmath470 is a stochastic map whose restriction to matrices with trace zero is self - adjoint
. it may be worth noting that we can apply the polar decomposition theorem directly to a completely positive map @xmath10 .
if we use @xmath511 to denote the adjoint with respect to the hilbert - schmidt inner product , then theorem [ thm : polar ] implies that we can write [ eq : polar.phi ] = || where @xmath512 and @xmath513 is a partial isometry .
if @xmath10 takes an algebra ( e.g. , @xmath164 ) to itself , then @xmath514 can be chosen to be an isometry , i.e. , @xmath515 . however , neither @xmath516 nor @xmath513 need be stochastic in general ( even though their composition is ) .
on the contrary , if @xmath470 has an odd number of negative eigenvalues , then @xmath517 where @xmath518 changes the sign of an odd number of eigenvalues and , hence , is _ not _ a completely positive map .
( see appendix c for further discussion and explicit examples . )
what we have shown in the argument above is that for unital maps on @xmath81 , the isometry @xmath519 can always be implemented by a unitary transformation , possibly composed with a map @xmath518 that takes @xmath520 \raw [ w_0 i - \bw \dtsig ] $ ] , i.e. , there is a unitary matrix @xmath111 such that @xmath519 has the form @xmath521
u^{\dg}$ ] where one and only one sign holds . if @xmath10 is a non - unital stochastic map , then @xmath516 will not even be trace - preserving and the isometry @xmath519 will correspond to a change of basis that mixes the identity @xmath522 with the three @xmath523 matrices ( in contrast to the unital case in which the change of basis affects only the subspace of traceless matrices spanned by the three pauli matrices ) .
hence for non - unital maps , the full polar decomposition ( [ eq : polar.phi ] ) of a stochastic map @xmath10 may be less useful than the polar decomposition on the restriction to matrices of trace zero . for unital maps on @xmath164 with @xmath524 it would be interesting to know how much @xmath519 can differ from a map of the form @xmath525 where @xmath111 is an @xmath184 unitary matrix . in order to see the connection between the svd and the so - called `` schmidt decomposition '' , consider a wave function or vector of the form [ eq : wavefctn ] = _ jk a_jk _ j _ k with @xmath526 and @xmath527 orthonormal .
it is not hard to see that there is an isomorphism between such vectors and operators of the form [ def : kpsi ] k_= _ jk a_jk| _ j _ k | and that @xmath528 is a hilbert - schmidt operator if and only if @xmath529 is square - integrable ( in the case of wave functions ) . moreover ,
if @xmath530 , then _ 1 t_2 ( _ 12 ) & = & k_k_^ + _ 2 t_1 ( _ 12 ) & = & ( k_^ k_)^t where @xmath531 and @xmath532 are the reduced density matrices obtained by taking the indicated partial traces @xmath533 and @xmath534 .
the `` schmidt decomposition '' is an immediate consequence of the application of the svd to the operator @xmath535 given by ( [ def : kpsi ] ) which implies the following result . _
( schmidt ) _ any wave function of the form ( [ eq : wavefctn ] ) can be rewritten as [ eq : schmidt ] = _ k _ k _ k _ k where @xmath536 are the singular values of the matrix @xmath378 , the bases @xmath537 and @xmath538 are orthonormal and related by @xmath539 with @xmath528 given by ( [ def : kpsi ] ) .
it follows immediately that the reduced density matrices @xmath531 and @xmath532 have the same non - zero eigenvalues @xmath540 and @xmath537 and @xmath538 are the eigenvectors of @xmath531 and @xmath532 respectively .
there is an interesting history associated with both the svd and schmidt decompositions , as well as the attachment of schmidt s name to ( [ eq : schmidt ] ) in the physics literature . in chapter 3 of @xcite , horn and johnson
give a detailed account of the history of the svd which goes back to beltrami and jordan who independently obtained the svd for real @xmath184 matrices in the 1870 s . in 1902
autonne obtained the svd for general nonsingular complex @xmath184 matrices and later made explicit the straightforward generalization to singular matrices in a long paper in 1915 which seems to have been subsequently overlooked by many researchers .
independently , schmidt obtained analogous results in 1907 for operators associated with integral kernels .
when quantum chemists became interested in density matrices in the 1960 s , carlson and keller @xcite rediscovered some of his results .
however , john coleman @xcite soon pointed out the connection with schmidt s much earlier work .
coleman s observation probably made physicists and chemists aware of that work although he did not use the term `` schmidt decomposition '' which has recently become popular in the quantum computing literature .
physical chemists were initially interested in the reduced density matrices which arise from a multi - particle wave function of the form @xmath541 where @xmath542 denotes the space and spin coordinates associated with the i - th particle .
it was natural to decompose these coordinates into two subsets @xmath543 and @xmath544 so that the density matrices @xmath531 and @xmath532 actually correspond to p - th and ( n - p)-th order reduced density matrices of @xmath529 .
it is noteworthy that if the original wave function @xmath529 has some symmetry , then the functions @xmath537 and @xmath538 can always be chosen so that ( [ eq : schmidt ] ) has the same symmetry . in particular , if @xmath529 is antisymmetric , ( as required by the pauli exclusion principle for fermions ) then these bases can be chosen so that ( [ eq : schmidt ] ) is also antisymmetric ; i.e. , there is no need to apply an additional antisymmetrizer . in the case of multi - particle wave functions @xmath545
is readily interpreted as the kernel of an integral operator ( which corresponds to @xmath546 defined above ) acting on an infinite dimensional hilbert space .
hence , it is quite natural to attribute the results to schmidt when used in this context .
however , in quantum computation , one only considers the `` spin '' part of the wave function .
since this can always be represented by a finite dimensional matrix , the `` diagonalization '' of @xmath547 in ( [ eq : wavefctn ] ) can be obtained directly from the svd ; there is no need to detour into infinite dimensional hilbert spaces to use results for integral operators .
moreover , since the matrix form of the svd preceded schmidt s work , it seems natural to use the term svd decomposition .
to prove the eigenvalue conditions ( [ eq : fuji.cond ] ) , we rewrite the kraus operators defined by ( [ eq : kraus ] ) in the form a_k = v_k0 i + _ k where @xmath548 is a vector in @xmath155 .
we will let @xmath490 denote the @xmath549 matrix with elements @xmath550 , @xmath551 its columns as vectors in @xmath552 and @xmath553 . using the relation [ eq : matprod ] ( ai + ) ( bi + )
& = & ( ab + ) i + ( a + b + i ) one finds a_k a_k^ & = & _ j=0 ^ 3 a_k^ a_k & = & _ j=0 ^ 3 |v_kj|^2 i + ( v_k0 _ k + _ k0 _ k - i _ k _ k ) so that _ k a_k a_k^ & = & ( _ j=0 ^ 3 |v_j|^2 ) i + 2 ( v_0 , + _ k _ k _ k ) + _ k a_k^ a_k & = & ( _ j=0 ^ 3 |v_j|^2 ) i + 2 ( v_0 , - _ k _ k _ k ) [ eq : ident.form ] where @xmath554 denotes the standard inner product in @xmath552 and @xmath553 . hence if @xmath10 is _ either _ unital or trace - preserving , then [ eq : icond ] _
j=0 ^ 3 |v_j|^2 = v^ v = 1 .
in addition , it follows from ( [ eq : cond.tracepres ] ) and ( [ eq : cond.unital ] ) that [ eq : vcond.tracepres ] ( ) = & & v_0 , + _ k _ k _ k = 0 + ( i ) = i & & v_0 , - _ k _ k _ k = 0 [ eq : vcond.unital ] .
this implies that @xmath10 is both unital and trace - preserving if and only if , in addition to ( [ eq : icond ] ) , v_0 , & = & 0 [ iden2 ] + _ k _ k _ k & = & 0 [ iden3 ] where ( [ iden3 ] ) can be rewritten as [ iden4 ] v_j , v_k & = & 0 , jk 1,2,3 by defining @xmath555 to be the matrix obtained by replacing the first column of @xmath490 by @xmath556 , one can rewrite conditions ( [ iden2 ] ) and ( [ iden3 ] ) as the single requirement that that @xmath557 is diagonal or , equivalently , @xmath558 . we will now derive the general form of the real @xmath102 matrix @xmath101 defined by ( [ eq : t3rep ] ) so that [ eq : matrix.el ] _ jk = ( _ j ( _ k ) ) in general @xmath101 is not symmetric ; in fact , it is symmetric if and only if every operator @xmath122 is self - adjoint . one finds after straightforward calculation that [ diags ] t_jj & = & v_0,v_0 + v_j , v_j - _ i j v_i , v_i + [ offdiags ] t_ij & = & 2 v_i , v_j 2 v_0 , v_k where the @xmath559 holds in ( [ offdiags ] ) if @xmath560 is an even or cyclic permutation of @xmath561 and the @xmath562 sign if it is an odd permutation .
thus , for example , t_11 & = & v_0,v_0 + v_1,v_1 - v_2 , v_2 - v_3,v_3 + t_12 & = & 2 v_1,v_2 - 2 v_0,v_3 in the special case where @xmath400 is diagonal its eigenvalues can easily be obtained from ( [ diags ] ) .
let @xmath112 be the eigenvalues , and define @xmath563 .
then ( [ diags ] ) becomes _ 1 & = & q_0 + q_1 - q_2 - q_3 + _ 2 & = & q_0 - q_1 + q_2 - q_3 + _ 3 & = & q_0 - q_1 -q_2 + q_3 together with the condition ( [ eq : icond ] ) which can be written as 1 & = & q_0 + q_1 + q_2 + q_3 this implies that the point with coordinates @xmath112 must lie inside the tetrahedron with corners at @xmath126 .
furthermore by taking @xmath564 and choosing the vectors @xmath565 to be orthogonal we see that every point in this tetrahedron defines a triplet of eigenvalues which can arise from a unital stochastic operator .
these conditions are equivalent to four linear inequalities which must be satisfied by the eigenvalues , namely @xmath566 which is equivalent to the more compact ( [ eq : fuji.cond ] ) .
these conditions were obtained earlier by algoet and fujiwara @xcite .
the expression in ( [ diags ] ) and ( [ offdiags ] ) also hold for non - unital @xmath10 . to calculate the non - zero elements in the first column extend ( [ eq : matrix.el ] ) to @xmath567 and
observe that the trace - preserving condition ( [ eq : vcond.tracepres ] ) implies @xmath568 .
using this in ( [ eq : ident.form ] ) yields t_j = t_j0 = _ j ( i ) = 4 v_0 ,
we now give some examples of unital and non - unital maps which illustrate some of the features discussed earlier and show the correspondence between our parameterizations and some well - known examples which are usually described by their kraus operators .
following the notation of section [ sect : notation ] , we will let @xmath569 $ ] denote a diagonal unital map . before doing so
we note that the classic example of a linear , positivity preserving map which is _ not _ completely positive is the transpose , which corresponds to @xmath570 $ ] .
we also note that if @xmath508 denotes the maps @xmath507 , the composition @xmath571 $ ] changes the sign of the two eigenvalues whose subscript is not @xmath572 .
( for example @xmath573 = \phi[-\lambda_1,\lambda_2,-\lambda_3 ] $ ] . ) thus the map @xmath574 $ ] which takes @xmath575 \raw \half[i - \bw \dtsig]$ ] is also _ not _ completely positive .
* depolarizing channel : @xmath576 $ ] @xmath577 * bfs @xcite two - pauli channel : @xmath578 $ ] @xmath579 * phase - damping channel : @xmath580 $ ] @xmath581 $ ] * rotation : a single rotation is the simplest example of a unital , non - self - adjoint map . @xmath582 where @xmath111 is unitary and @xmath583 . +
a convex combination of such rotations using kraus operators @xmath584 where @xmath585 yields a more general example of a unital map which is not self - adjoint , namely , @xmath586 matrix @xmath587 where @xmath588 is the rotation corresponding to @xmath589 .
the bfs two - pauli channel was introduced by bennett , fuchs , and smolin @xcite to demonstrate that entangled states could reduce the probability of error .
it is worth noting that this map corresponds to the extreme points studied in detail in section [ sect : entanal ] .
hence , it is particularly noteworthy that our analysis provides particularly strong evidence that entanglements do _ not _ decrease the entropy for these maps .
it is also worth noting that when @xmath590 this map becomes @xmath591 $ ] .
thus , although @xmath592 $ ] and @xmath574 $ ] are _ not _ completely positive , both @xmath593 $ ] and @xmath594 $ ] _ are _ completely positive .
a. s. holevo , on the capacity of quantum communication channel " , _ probl .
peredachi inform .
_ , * 15 * , no . 4 , 3 - 11 ( 1979 ) ( english translation : _ problems of information transm . _ , * 15 * , no . 4 , 247 - 253 ( 1979 ) ) . *
* figure 1 . * the four eigenvalues of the product state of minimal entropy , with their direction of movement shown as @xmath599 increases away from @xmath6 . * * figure 2 . * the entropy difference @xmath600 $ ] for the two channels @xmath601 \otimes \phi[\mu , \mu , \mu]$ ] ( upper curve ) and @xmath601 \otimes \phi[\mu , -\mu , \mu]$ ] ( lower dashed curve ) , in the range @xmath339 .
note that the curves coincide over most the interval . *
* figure 3 . * the entropy difference @xmath600 $ ] for the three channels @xmath601 \otimes \phi[\mu , \mu , \mu]$ ] ( dots ) , @xmath601 \otimes \phi[\mu , 2\mu -1 , \mu]$ ] ( dashes ) , and @xmath602 \otimes \phi[\mu , 2\mu -1 , \mu]$ ] ( full ) in the range @xmath343 . * * figure 4 . *
the entropy difference @xmath600 $ ] for the channel @xmath603 $ ] in the range @xmath604 . *
* figure 5 .
* the bloch sphere and its image ( the ellipsoid ) under a unital map with one large and two small singular values .
the endpoints of the ellipsoid are shown separately
these are the unique states of minimal entropy . * * figure 6 . * the bloch sphere and its image under a unital map with two large and one small singular values .
the `` waistband '' of the ellipsoid is shown separately ; this entire circle consists of minimal entropy states . *
* figure 7 . * the bloch sphere ( the circle of radius 1 ) and its image ( the translated ellipse ) under the fuchs map , together with the circle of radius @xmath605 .
the endpoints of the ellipse are marked @xmath606 , and the points of minimal entropy are marked @xmath607 . | in this paper , we consider the minimal entropy of qubit states transmitted through two uses of a noisy quantum channel , which is modeled by the action of a completely positive trace - preserving ( or stochastic ) map .
we provide strong support for the conjecture that this minimal entropy is additive , namely that the minimum entropy can be achieved when product states are transmitted .
explicitly , we prove that for a tensor product of two unital stochastic maps on qubit states , using an entanglement that involves only states which emerge with minimal entropy can not decrease the entropy below the minimum achievable using product states .
we give a separate argument , based on the geometry of the image of the set of density matrices under stochastic maps , which suggests that the minimal entropy conjecture holds for non - unital as well as for unital maps .
we also show that the maximal norm of the output states is multiplicative for most product maps on @xmath0-qubit states , including all those for which at least one map is unital . for the class of _ unital _ channels on @xmath1
, we show that additivity of minimal entropy implies that the holevo capacity of the channel is _ additive _ over two inputs , achievable with orthogonal states , and equal to the shannon capacity .
this implies that superadditivity of the capacity is possible only for non - unital channels . _
key words : _ entangled state ; minimal entropy ; qubit ; stochastic map . |
many gregarious vertebrates are fusion - fission species , with frequent changes in size and composition of groups . in addition , the habitat they live in is generally heterogeneous , such that individuals alone or in groups may have to move among more or less distant areas in order to fulfil their basic vital requirements , e.g. drinking , resting or avoiding pblackators @xcite . even during feeding periods , animals have to search for available food , moving slowly and on short distances between feeding stations or more rapidly and on larger distances to exploit distinct feeding areas @xcite .
thus , groups either resting or feeding with animals motionless or moving slowly are frequently joined by incoming individuals but also submitted to departures of group members @xcite . when faced to multiple choices
, social organisms must reach consensus in order to maintain the cohesion of the group and the advantages linked to it @xcite .
the departure of one or few individuals from static groups as well as stops in moving groups compromise social cohesion @xcite .
this is particularly critical for small groups a scenario that applies to most gregarious animals @xcite despite the popularity and fascination that produce giant bird flocks or fish schools @xcite where group splitting represents a serious pblackatory risk @xcite .
what influences the individual decisions , _
i.e _ the interplay between external stimuli and internal state , and which decision - making processes occur to maintain social cohesion are among the most compelling questions in the study of collective animal behavior @xcite .
this goes beyond biology and concerns the broader question of how information at the individual level is evaluated , processed and distributed in the group @xcite .
recent experiments with primates and fish revealed that an individual spontaneously departing from a static group is likely to give up and return to the group when not followed @xcite .
it has been also observed that collective motion is promoted by already moving conspecifics @xcite .
finally , in some species , a quorum is requiblack to observe collective movement @xcite . on the other hand ,
how collective motion stops , remains largely unexploblack except for few exceptions @xcite . in summary , there is a lack of experimental and theoretical studies focusing on the decision - making processes that trigger and stop collective motion @xcite . here
, we use experimentally acquiblack data to address how groups of uninformed individuals respond to the departure and stop of an informed conspecific .
we show that the groups of uninformed individuals always reach a consensus : either all uninformed individuals follow the informed one or none does it . combining experiments where we control the stimulus , associated to the motion of the informed individual , and a mathematical model we unveil that the apparent collective decision - making process leading to an effective consensus results from a self - organized phenomenon resulting from the interplay of simple mimetic rules and the characteristic duration of the stimulus , with group size playing a central role .
fieldwork was carried out in the domaine du merle ( @xmath0 e , @xmath1 n ) in the south of france .
the field station is located in the crau region , a very flat area coveblack by a native steppe .
the experiments were performed within irrigated pastures providing homogeneous food resources . from the available flock of 1400 ewes , 30 of them
were randomly selected and allocated to the training set and a further 200 ewes to the naive set .
all ewes used were unrelated and were 3 to 5 year - old .
a number was painted on the back and fleece of each sheep for identification .
both sets were penned up in the same sheepfold during the evening and the night .
all the experiments were carried out in daylight ( from 10 am to 6 pm ) and the ewes were fed hay in the sheepfold in the morning and in the evening . the experiments rely on our capacity to trigger the movement of one sheep toward a fixed panel at the periphery of an arena at a desiblack time .
the protocol used to trained sheep was similar to the one used in previous experiments @xcite .
sheep to be trained were originally allocated at random in 6 groups of 5 animals which composition remained unchanged during the training period .
sheep were first habituated in the sheepfold to feed on corn and to receive simultaneously a vibration provided by a neck collar during 3 days .
then the training groups were introduced successively in one of two test arenas ( 50 x 50 m ) , for a period of 30 to 40 min , each animal wearing a vibrating collar .
ten minutes past the introduction , the collar was activated and one yellow panel ( 0.5 x 0.5 m ) was simultaneously raised delivering a handful of corn .
each group received four to eight stimulations , each separated by a period of at least 5 min during each training session .
past 14 days of training , we selected the 3 sheep with the best learning scores ( 100% of departure toward the panel following a vibration ) .
meanwhile , the set of naive sheep to be used in experiments was confronted to panel rising ( without corn delivery ) at the periphery of experimental arenas , at one - min interval during two sessions of 90 min . at the end of this habituation session ,
no naive sheep raised its head when raising the panel .
in addition , these two days allowed naive sheep to be familiarized with the experimental setup . the experimental setup consisted in two arenas ( 50 x 50 m ) delimited with sheep fences and surrounded by a visual barrier ( propylene net ) .
a 7m - high tower was placed at an equal distance ( 10 m ) apart from two next arenas corners .
yellow panels ( 0.5 x 0.5 m ) were hammeblack in the middle of each side for both arenas and were not visible to sheep ( ) . the tests consisted in introducing groups of 8 , 16 or 32 sheep within one arena , among which one trained sheep equipped with a vibrating collar .
the trained ewes were used no more than twice a day , but were implicated in all group sizes .
the naive sheep that composed the rest of the groups were selected randomly for each test . because of the large number of individuals needed to complete all replications , the naive sheep were used several times , except in groups of 8 .
thirty replications were performed for groups of 32 and 15 replications for groups of 8 and 16 individuals .
a test was conducted as follows : the group was introduced in the arena and sheep grazed spontaneously during 20 min .
then , one of the two panels closest to the tower was raised , waiting for all sheep grazing ( _ i.e _ head down ) .
simultaneously the vibrating collar of the trained sheep was activated for 2 sec .
past 10 min ( end of test ) , a new panel was raised ( one of the two farthest from the tower ) to reinforce the conditioning of trained ewes and avoid restricting their space use to the vicinity of panels closest to the tower .
the group was led back to the sheepfold shortly afterward .
the naive sheep that were not tested during one experimental day were introduced in distant pasture .
we never performed two trials in parallel .
we also carried out control experiments to be sure that naive ewes did not associate the panel rise and the food reward .
thereby , 6 tests before and 6 after the test series were conducted with groups of 32 naive ewes , using the same protocol as described before .
we found no movement of groups when raising the panel , almost all sheep continuing their spontaneous activity .
two digital cameras ( canon eos d50 ) were fixed on the tower , each one focusing on one arena .
fifteen minutes after the introduction of the groups , the digital camera was turned on , taking a picture of the entire arena every second and turned off five minutes after the panel was raised . for each replication
, we obtained a series of about 600 pictures . using a custom software developed by jg
, we were able to track on each picture the position and the orientation of animals by dragging a vector on their back , and identify the behavior of each individual , _
i.e _ grazing , standing head - up , moving and others .
we defined a departure of the trained sheep ( initiator ) , _
i.e _ initiation past experimental stimulation when it performed an uninterrupted walk towards the raised panel .
the following behavior , _
i.e _ a new departure , was defined as the movement of a naive ewe occurring after the trained sheep departure , without stop until joining the trained ewe near the panel .
the behavior of stopping was defined as an individual ceasing to walk and remains either stationary head - up or resumes grazing .
six replication in the groups of 32 were discarded , one because the initiator did not depart , two because the initiator stopped moving between the group and the target and three because the initiator showed a moving behavior not comparable to other trials ( going to a wrong target first and then joining the rewarded target ) .
thus we performed analyses on 15 trials for groups of 8 and 16 sheep and 24 trials for groups of 32 .
the level @xmath2 was fixed to 0.05 for statistical significance .
all analyses were conducted using r version 3.0.1 .
all the animals were maintained under routine husbandry conditions at a montpellier supagro research station ( domaine du merle , salon - de - provence , france ) with full approval of its director pierre - marie bouquet .
animal welfare requirements were fully respected in accordance with the european directive 2010/63/eu , with the rules of the european convention for the protection of vertebrate animals used for experimental and other scientific purposes and with the convention of the french comit national de rflexion thique sur lexprimentation animale .
no special authorization from the french ethical committee for animal experimentation ( commission nationale de lexprimentation animale ) was requiblack as no protected or endangeblack species was involved , as the experiments did not imply any invasive manipulation ( the experimental protocol consists in the observation of groups and the acquiblack data are only pictures of the animals ) and as sheep were conducted to the test arenas , as they are herded on a daily basis to the pastures .
all personnel involved had technical support from the employees of the research station as requiblack by the french ministry of research .
the experimental protocols included short test periods ( 20 minutes ) where sheep did not experience painful , stressful or unfamiliar situations .
the experimental procedures had no detrimental effect on the sheep and at the end of the experiment all the animals reintegrated the sheep herd of the breeding research station .
in our experiments , we work with groups of n = 8 , 16 and 32 sheep , among which 1 is a trained individual henceforth referblack to as initiator while the remaining n - 1 are uninformed / naive individuals .
the initiator is trained to move towards a target located at the periphery of the arena when a vibrating collar is activated by a remote control ( fig .
[ fig1]c ) .
the group is subject to the perturbation produced by the initiator : _
i.e. _ the sudden departure and stop of the initiator , which challenge the social cohesion of the group .
when the departure of the initiator triggers a collective response , we observe three distinct consecutive phases : departing , collective motion , and stopping as illustrated in fig .
the departing phase starts with the departure of the initiator and continues until the number n~m~ of moving individuals increases to match the group size . at this point ,
the collective motion phase begins with the group moving cohesively behind the initiator .
the stop of the initiator near the target marks the onset of the stopping phase , where n~m~ decreases until reaching 0 .
the behavior of an individual can be characterized by one of the following states : stopped at the start position ( s~s~ ) , moving ( m ) , or stopped at the target position ( s~t~ ) ( fig .
[ fig1]b ) . the whole process can be then described as a transition first from s~s~ to m , and then from m to s~t~. in all trials with 8 and 16 sheep , the departure of the initiator systematically triggers a collective motion ( fig .
[ fig1]d ) . in groups of 32
the departure of the initiator does not always leads to a collective motion of the naive group .
splitting of the naive group has not been observed ( see and for examples . ) we start our analysis by focusing first on the cases where the initiator successfully provokes a collective motion .
we are interested in quantifying the decision of sheep to switch from s~s~ to m and from m to s~t~. fig .
[ fig2]a ( respectively , fig .
[ fig2]c ) shows that from from s~s~ to m ( in fig .
[ fig2]c , from m to s~t~ ) increases with n~m~ ( with n~s~t~~ , the number of individuals in state s~t~ , in fig .
[ fig2]c ) .
[ fig2]b ( respectively , fig .
[ fig2]d ) indicates that the transition rate from s~s~ to m ( respectively , from m to s~t~ ) for a fixed value of n~m~ ( fixed value n~s~t~~ , for m @xmath3 s~t~ ) diminishes with n~s~s~~ ( n~m~ in fig .
[ fig2]d ) .
this indicates that both transitions s~s~ @xmath3 m and m @xmath3 s~t~ share similar features : they both exhibit a promoting component ( n~m~ in s~s~ @xmath3 m and n~s~t~~ in m @xmath3 s~t~ ) and an inhibiting component ( n~s~s~~ in s~s~ @xmath3 m and n~m~ in m @xmath3 s~t~ ) .
this is consistent with previous studies using smaller group sizes @xcite .
as mentioned above , the departure of the initiator does not always trigger collective motion in groups of 32 individuals .
importantly , the absence of collective follow is not related to a peculiar behavior of the initiator ( see for details ) .
these observations suggest a social effect linked to group size ( _ i.e _ to the number of uninformed individuals ) . for groups with n = 32 , the uninformed individuals , upon departure of the initiator , respond in an all - or - none way : either all follow the initiator or none of them does it , and thus no fission of the naive group is observed .
this phenomenon , which seems at first glance to require some sort of concertation among the naive individuals , can be understood by focusing on the behavior of the first potential follower .
our argument is based on the assumption that the initiator can stimulate a transition s~s~ @xmath4 m only when moving to the target position , _
i.e. _ while being in state m. this implies that if no naive sheep departs by the time the initiator reaches the target , no transition s~s~ @xmath3 m will ever occur . in consequence , the probability p~s~ that the initiator is still moving at time t provides a rough estimate of the probability that the stimulation is still present .
all this means that the problem can be blackuced to the competition between two probabilities : p~s~ and the probability p~f~ that the first follower departs before time t , as illustrated in fig .
[ fig3]a - c .
[ fig3]a and b indicate that for n = 8 and 16 , the transition s~s~ @xmath3 m for the first followers always occurs before the initiator stops near the target position .
on the other hand , fig .
[ fig3]c shows that the transition s~s~ @xmath3 m for the first follower is such that the initiator can stop at the target position before this transition has ever occurblack .
this provides a qualitative explanation of the remarkable group size effects observed in the experiments .
now , we go further in the quantitative analysis using a mathematical model , which has proved useful to analyze the experimental data and to test the various hypotheses formulated to interpret the observed phenomena . in particular
, we will see that ( i ) the departure and stop rates can be expressed as non - linear functions , with promoting and inhibiting components as proposed above , and ( ii ) that the hypothesis that the initiator can only induce a transition s~s~ @xmath3 m while being in state m is consistent with the experimental data . our first step is to formulate the ( individual ) rate @xmath5 associated to the transition s~s~ @xmath3 m as : @xmath6 with @xmath2 , @xmath7 and @xmath8 parameters modulating the effect of n~s~s~~ and n~m~ on @xmath5 .
notice that for simplicity we have assumed that every individual is able to perceive all individuals in the group .
this implies that each individual has a global perception of the group , that is a reasonable hypothesis for small groups of up to few dozens of individuals , but that becomes unrealistic in large herds with hundblacks of animals .
given expression , we can compute the mean time to depart t(n~m~ ) for the n~m~^th^ follower as @xmath9 , where @xmath10 is the departure transition rate at the group level . from the inverse function of this expression
we obtain n~m~ as a function of time ( fig .
[ fig3]e ) , _ i.e. _ from t(n~m~ ) we obtain n~m~(t ) .
notice that the departing phase is then given by @xmath11 , which means that the average collective motion phase is approximately @xmath12 , where @xmath13 is the time requiblack by the initiator to reach the target ( see ) .
in analogy to equation , we assume that the ( individual ) stop rate @xmath14 , related to the transition from m @xmath3 s~t~ , is given by @xmath15 with @xmath16 , @xmath17 and @xmath18 parameters modulating the effect of n~m~ and n~s~t~~ on @xmath14 . during the stopping phase , n~m~(t ) ) is obtained from the inverse function of @xmath19 with @xmath20 , where @xmath21 . as commented above
, it can occur that uninformed individuals do not follow the initiator in groups of n = 32 . using the proposed mathematical framework ,
we can account for such group size effect .
we recall that the absence of collective motion is associated with those situations where the initiator does not induce any transition s~s~ @xmath3 m of a naive individual during the time @xmath13 requiblack by the initiator to reach the target position .
mathematically , the problem blackuces to compute the probability c henceforth referblack to as _ commitment _ to observe a transition s~s~ @xmath3 m of a naive individual during @xmath13 , which takes the form : @xmath22 with @xmath23 , i the number of initiators ( i = 1 in our experiments ) and @xmath24 and @xmath25 , the minimum and maximum observed @xmath13 values respectively ( see for the derivation of expression ) .
the estimation of parameters is done in two steps .
the best fit of eq . to the experimental points in fig .
[ fig3]d provides a set of @xmath2 and @xmath8 values .
fixing these two parameters , the best fit on the individual rates is used to estimate the other parameter .
more details on the fitting procedure are given in .
the obtained values are @xmath26 , @xmath27 , @xmath28 , and @xmath29 , @xmath30 , @xmath31 . fig .
[ fig3]e shows that the dynamics resulting from equations and provides a qualitative description of the data ( _ cf . _
[ fig1]a ) , while fig .
[ fig3]d shows that the model also accounts for the experimentally observed commitment c. it worth noticing that given the nonlinear form of the transition rates and given the obtained parameter values , it turns out that after a first transition s~s~ @xmath3 m of a naive individual , a cascade of transitions s~s~ @xmath3 m will immediately follow .
the mathematical model allows exploring a large variety of hypothetical scenarios using conditions going beyond the ones used experimentally .
it is important to stress that model pblackictions are reliable as long as all hypotheses and assumptions remain valid .
ultimately , models pblackictions should be verified / falsified by performing further experiments . always according to our mathematical model ,
there exist three possible outcomes when the naive group faces the departure of one or more initiators : i ) all naive individuals follow the initiator(s ) , ii ) none of them does it , and iii ) only a fraction of naive individuals follow the initiator(s ) . the associated probabilities , p~af~ for i ) , p~nf~ for ii ) , and p~gs~ for iii ) ,
are shown in fig .
[ fig4 ] . while the first two scenarios , _
i.e. _ i ) and ii ) , imply a sort of consensus for the naive group , which assures group cohesion , the third scenario , _
i.e. _ iii ) , implies group fission .
although according to the mathematical model group splitting can occur , its probability is always small .
[ fig4]a shows that in groups of @xmath32 , the probability of observing group splitting is smaller than 5% .
given that with n = 32 we have performed @xmath33 experimental trials , one could expect in mean one group splitting event . in short ,
the experimental observations are consistent with model pblackictions .
in addition , the model pblackicts - for a fixed range of @xmath13 ( _ i.e. _ stimulation duration ) - that full collective motion decreases non linearly with group size n , while the probability p~nf~ that none follows increases non linearly ( black and black curves , respectively , in fig .
[ fig4]a ) .
the probability of observing splitting of the naive group ( green curve ) exhibits a non monotonic dependency with n with a maximum at n = 64 , where the probability is close to 20% .
this means that by performing similar experiments with @xmath34 naive individuals and one initiator - and performing a similar number of replications of the experiment - we should be able to observe group fission .
it is worth noticing that if the group size is larger than @xmath35 , the probability of splitting decreases again .
furthermore , the model pblackicts that for large group sizes the departure of the initiator can not induce naive individuals to move towards the target position : for large values of @xmath36 the naive group remains unresponsive to the departure of the initiator .
this observation is , however , only valid for a fixed range of @xmath13 .
if we imagine that the experiment is performed in larger arenas , in such a way that the target position is located farther away and @xmath13 is significantly larger , the departure of the initiator will again lead naive individuals to move towards the target ( fig . [
moreover , if the @xmath13 is large enough , the model pblackicts that the only possible outcome upon the departure of the initiator is a full collective follow .
accordingly , the only scenario where fission of the naive group can be observed is when the two resulting subgroups will be separated a relatively small distance one from the other , at which point we could ask ourselves whether such separation qualifies truly as group splitting .
these experimental and theoretical results are in full agreement with very recent findings showing that in animal groups , the persistence of the stimulus ( the behaviour of one or several initiators ) in time is a crucial parameter to trigger a consensus to either turn or depart cohesively @xcite
. this stimulus duration might be dependent of the travelled distance as shown in our results but also of the velocity of the initiator that is known to be a factor influencing group decisions @xcite .
finally , the model also allows us to explore situations where naive groups are subject to conflicting information cues .
for instance , let us imagine we trained a group of i individuals to move towards the target position and a group of r individuals to stay at the starting position .
we are interested in the behavior of the naive individuals , for whom there are always two mutually exclusive options either to stay or to go , which means that all probabilities , p~af~ , p~nf~ , and p~gs~ , are computed with respect to the naive group .
to avoid confusions , we clarify that this implies that p~af~ is the probability that all naive individuals follow , p~nf~ none of the naive individuals follows , and p~gs~ the probability that at the end of the process we find some of the naive individuals at the starting position and some at the target position .
we start out with a simple scenario where there is no trained individual to stay , _
i.e. _ r = 0 . by increasing the number i of individuals moving away ,
we find , not surprisingly , that p~af~ increases dramatically , to the point that the only possible outcome is a collective motion ( _ i.e. _ p~af~ @xmath4 1 , fig .
[ fig4]c ) .
similarly , we can fix i and increase the number r of individuals determined to stay .
we find that increasing r leads to a very weak increase of p~nf~ and p~gs~ ( fig .
[ fig4]e ) .
thus , there is a clear asymmetry in the role played by i and r. this is particularly evident by fixing the number of trained individuals and varying the relative weight between i and r. fig .
[ fig4]d corresponds to i + r = 3 and shows that i = 1 and r = 2 is remarkably different from i = 2 and r = 1 . in particular , while for i = 2 and r = 1 , p~af~ @xmath4 1 , for i = 1 and r = 2 , we do not find that p~nf~ @xmath4 1 .
i = 1 and r = 2 differs only slightly from i = 1 and r = 0 .
this means that if we have initially i = 1 and r = 1 , adding an extra trained individual determined to move towards the target ensures a full collective follow , while adding an extra individual determined to stay has a very weak effect on the collective outcome .
we have to make r close to n to assure that the trained individual will systematically fail to recruit any naive sheep . in short , a small variation of r has a weaker effect than a small variation of i. adding individuals determined to either stay or to leave does not produce , at the collective level , a symmetric bias , this is in sharp contrast with binary decision studies @xcite where mutually exclusive options are consideblack as symmetric .
this is particularly clear in flocking models @xcite where left - right choices are such that individuals determined to move towards the left exert the same social influence than individuals determined to move towards the right .
one important message we learn is that opposite alternatives are not necessarily symmetric .
the origin of such asymmetry may be related to the fact that at the individual level the decision whether to stay or to leave can be formulated as a decision whether to remain in the current state or to change it .
the obtained results , at both the experimental and theoretical level , suggest that _ behavioral change _ is strongly favoblack .
moreover , it seems that the individuals that initiate a change become as previously proposed @xcite incidental leaders , while those determined to remain in the current state , though playing an inhibiting role , exert a weaker influence on the naive group .
here , we have shown that simple mimetic responses as those described by equations and when combined with the characteristic duration associated to the stimulus here , the time requiblack to arrive at the target position act as an effective collective decision - making mechanism .
moreover , we showed that the proposed mechanism , whose derivation is intimately based on the presented experimental observations , allows a group of naive / uninformed individuals to solve a scenario with conflicting information in such a way that the most probable collective outcome is a consensus .
specifically , we analyzed a situation where the naive group faces a scenario where there is a subgroup decided to stay and subgroup decided to go , and showed that the two most probable outcome are that either all naive individuals follow or none does .
importantly , though group splitting can not be discarded , such event is , according to the proposed mechanism , unlikely . in summary
, the interplay between mimetic rules and the characteristic duration of the stimulus leads to a self - organized collective decision - making that does not requiblack neither explicit communication nor concertation among the group members .
finally , it is worth noticing that the effective decision - making mechanism described here applies to a specific social context : a group of nave individuals sharing an initial behavioral state , which is subject to the behavioral shift of one or several conspecifics .
one of the essential elements of the proposed mechanism is the presence of a discrete number of behavioral states , which has to be , necessarily , larger than two ( as _ e.g. _ s~s~ , m and s~t~ ) .
this means that collective decision - making models for groups in motion , where the group has to decide in which direction to move @xcite can not be used to model the specific social context addressed here .
such models have been designed to describe the navigation of a group , and not to describe behavioral shifts . at the mathematical level
the differences are evident . while navigation models associate a continuous variable to each individual , related to the moving direction of the individual , behavioral shift models deal with discrete variables associated to the possible behavioral states of the group members .
similarly , the social mechanism analyzed here can not be directly compablack with decision - making models designed to described dichotomic decisions as left and right moving direction @xcite .
the essential difference with such models is that .
in contrast to the social context we are interested in here where individuals initially share same behavioral state . in summary , the collective decision - making mechanism proposed here is fundamentally different , and not comparable to previous collective decision - making mechanisms , which have been designed to describe a different social context .
deeply rooted in the models proposed in previous works @xcite , the mechanism proposed here differs from these two by making use of three behavioral states , accounting simultaneously for the initiation and stop of the collective response , and with both processes modeled as a transition at the individual level .
given the simplicity of the proposed mechanisms here , we expect similar mechanisms to be at work in other animal systems .
we thank the staff of the domaine du merle , r. violleau and m .- h .
pillot for support during experiments , p. arrufat for technical help , dynactom team , r. fournier , f .- x .
dechaume - moncharmont , s. blanco and o. chepizhko for discussions .
this work was supported by one cnrs peps grant physique thorique et ses interfaces and one cnrs peps grant bio - maths - info .
st is supported by a phd grant of the french ministry of superior education and research
. 99 beauchamp g. .
animal behaviour .
2012 apr;83(4):11251129 .
conradt l , roper tj .
trends in ecology & evolution .
2005 aug;20(8):449456 .
fischhoff ir , sundaresan sr , cordingley j , larkin hm , sellier mj , rubenstein di . .
animal behaviour .
2007 may;73(5):825831 . king aj , cowlishaw g. communicative & integrative biology .
2009;2(2):147150 .
sueur c , king aj , conradt l , kerth g , lusseau d , mettke - hofmann c , et al . .
2011 sep;120(11):16081617 .
dumont b , boissy a. .
behavioural processes .
2000;49(3):131138 .
. 2000;59(2):403409 .
ramseyer a , boissy a , dumont b , thierry b. .
animal behaviour .
2009 jul;78(1):7178 .
ward ajw , herbert - read je , jordan la , james r , krause j , ma q , et al . .
the american naturalist .
2013 jun;181(6):748760 .
danchin , giraldeau la , valone tj , wagner rh . .
2004;305(5683):487491 .
galef jr bg , giraldeau la . .
animal behaviour .
2001 jan;61(1):315 .
couzin i d , krause j , james r , ruxton gd , franks nr .
journal of theoretical biology .
2002 sep;218(1):111 .
eftimie r , de vries g , lewis ma . .
proceedings of the national academy of sciences .
2007 apr;104(17):69746979 .
reiczigel j , lang z , rzsa l , tthmrsz b. .
animal behaviour .
2008 feb;75(2):715721 .
ballerini m , cabibbo n , candelier r , cavagna a , cisbani e , giardina i , et al . .
proceedings of the national academy of sciences .
2008 jan;105(4):1232 .
conradt l , roper tj .
behavioural processes .
2010 jul;84(3):675677 .
giardina i. .
hfsp journal .
2008 aug;2(4):205219 . king aj .
behavioural processes .
2010 jul;84(3):671674 .
petit o , bon r. behavioural processes .
2010 jul;84(3):635647 .
procaccini a , orlandi a , cavagna a , giardina i , zoratto f , santucci d , et al . .
animal behaviour .
2011 oct;82(4):759765 .
roberts g. animal behaviour
. 1997;54(6):15171522 .
ward ajw , sumpter djt , couzin i d , hart pjb , krause j. .
proceedings of the national academy of sciences . 2008;105(1518):69486953 .
bellman re , zadeh la . .
management science .
1970 dec;17(4):141164 .
dyer jrg , ioannou cc , morrell lj , croft dp , couzin i d , waters da , et al . .
animal behaviour .
2008 feb;75(2):461470 .
edwards w. psychological bulletin .
1954 jul;51(4):380 .
herrera f , martinez l , sanchez pj . .
european journal of operational research .
2005 oct;166(1):115132 .
lindley dv . making decisions ; 1985 .
petit o , gautrais j , leca jb , theraulaz g , deneubourg jl . .
proceedings of the royal society b : biological sciences .
2009 aug;276(1672):34953503 .
pratt sc . .
behavioral ecology .
2005 mar;16(2):488496 .
sumpter djt , krause j , james r , couzin i d , ward ajw . .
current biology .
2008 nov;18(22):17731777 .
bhattacharya k , vicsek t. .
new journal of physics .
2010 sep;12(9):093019 .
daruka i. proceedings of the royal society of london series b - biological sciences .
2009 mar;276(1658):911917 .
cresswell w , hilton gm , ruxton gd . .
proceedings of the royal society b : biological sciences .
2000 apr;267(1444):733737 .
pillot mh , deneubourg jl .
behavioural processes .
2010 jul;84(3 ) .
pillot mh , gautrais j , arrufat p , couzin i d , bon r , deneubourg jl . .
2011 jan;6(1):e14487 .
miller ny , garnier s , hartnett at , couzin i d . .
proceedings of the couzin i d , krause j , franks nr , levin sa . .
2005;433(7025):513516 .
conradt l , krause j , couzin i d , roper tj .
the american naturalist .
2009 mar;173(3):304312 .
pillot mh , gautrais j , gouello j , michelena p , sibbald am , bon r. . behavioural processes .
2010 feb;83(3):77 .
* one example of a trial in a group of 32 where the initiator triggers a collective motion . *
we compiled the snapshots at a rate of 2 frames per second ( 1 second in the video represents 2 s in real time ) . at 12
s , we added at the bottom left , for each frame a figure showing the locations of the group members across time .
the initiator is plotted in blue .
naive individuals not moving are plotted in black and the moving ones in black . * details on the behavior of sheep . * in trials with groups of 8 and 16 , the initiation always triggeblack the departure of all naives . in groups of 32 , only 75% of the trials ( 18 out of 24 ) displayed such collective response .
we investigated whether the lack of commitment of the naives could be explained by behavioral differences in initiations or by particular group configuration .
we used three different initiators in the experiments that were tested in all group sizes .
each failed to entrain naive sheep i.e. the cases where no collective departure was observed can not be due to any particular trained sheep .
when comparing trials with and without follower , no differences were found regarding initiators movement mean speed ( student _ t_-test : t = -1.08 , p = 0.3 ) or in terms of group density ( wilcoxon test : w = 66 , p = 0.9 ) .
thus , the lack of commitment can not be explained by any peculiar behavior or position of the initiators , indicating that social mechanisms are involved .
trials without collective motion occurblack on days 3 , 4 , 8 , 15 and 16 of the 17 days of experiments discarding any potential effect of habituation or change in motivation of the naives .
all naive sheep were tested in control trials in groups of 32 ( without trained individual ) , 6 before and 6 after the experiments .
these control groups were confronted to a panel rise in the same conditions as in the experimental trials ( except that no sheep was wearing a vibrating collar ) .
no naive sheep responded to the panel rise _ i.e. _ we did not record any movement nor any behavioral modification .
* statistics of the departing and stopping phases .
* we tested whether latencies of the first followers were affected by group size .
we found a significant effect of group size ( medians : 1 s , 1 s and 2 s for groups of 8 , 16 and 32 respectively ; kruskal - wallis test : @xmath37 = 8.67 , df = 2 , p = 0.01 ) .
furthermore , the first followers latencies were more variable in groups of 32 than in smaller ones ( bartlett - test : k2 = 42.97 , df = 2 , p @xmath38 0.0001 ; a ) .
the duration of the departing phase ( time elapsed between the first and the last following events ) increases with group size ( medians : 3.6 s , 7 s and 14.5 s respectively ; @xmath37 = 24.51 , df = 2 , p @xmath38 0.0001 ) but also varied highly in the larger group size ( @xmath39 = 20.81 , df = 2 , p @xmath38 0.0001 ; b ) . with the same logic , we tested the effect of group size on the latencies of first stoppers .
there is no significant effect of group size on the first stop latency ( medians : 1.3 s , 1.3 s and 3 s respectively for group sizes 8 , 16 and 32 ; @xmath37 = 4 , df = 2 , p = 0.13 ) , but stop latencies were more variable in groups of 32 ( @xmath39 = 10.14 , df = 2 , p = 0.006 ; a ) .
the duration of the stopping phase ( time elapsed between the first and the last stop ) increases ( medians : 6 s , 7.5 s and 11 s respectively ; @xmath37 = 12.28 , df = 2 , p = 0.002 ) but the variance was not affected by group size ( @xmath39 = 4.12 , df = 2 , p = 0.13 ; b ) .
* details on the parameter estimation .
* following pillot _
@xcite , we quantify the individual response function , i.e. the departure rate @xmath5 ( respectively the stop rate @xmath14 ) considering that sheep are stimulated to move ( respectively to stop ) by all moving ( stopped ) sheep but inhibited by all stopped ( moving ) ones with @xmath40 and @xmath41 previously , pillot _ et al .
_ never recorded cases where no collective departure was observed .
the parameter estimation we perform is of two types for departing and stopping phases because we want to get parameters values allowing to pblackict lack of collective departures ( which is highly dependent on the first followers latencies ) , but also to account for the dual combination of mimetic rules in order to get coherent description of the two phases when increasing the group size and the number of initiators . about the departing phase ,
the estimation of the parameters that best fit the experimental data is a non - trivial task .
we want equation to produce a value of commitment c close to the experimental data ( fig .
[ fig3]d ) , while equation to reproduce as faithfully as possible the temporal patterns of the departure and the stopping phases respectively ( fig .
[ fig3]e ) .
let s consider first the commitment c. equation which models c depends only on @xmath2 and @xmath8 for i = 1 .
we adjust @xmath2 and @xmath8 to obtain the best fitting of equation , and then we have only one free parameter to play with ( @xmath7 ) to tune equation .
we kept the values of @xmath2 and @xmath8 that minimize the error ( as calculated by the average of the squares of differences ) on c , and looked for a value of @xmath7 minimizing the error on the departure rates .
with respect to experimental departure rates , we modify the procedure followed by pillot _
_ in order to take into account both trials with and without collective departures .
the modified procedure computes the survival analysis on the distribution of the departure latency of the first followers including those censoblack by the time spent by initiators to reach the target in trials where they were not followed ( n = 6 ) . because the distribution of latencies with censorship follows an exponential function ,
we then are able to calculate the associated time constant and thus the estimated rate for first followers in groups of 32 ( estimated rate without censoring : 0.19 , with censoring : 0.08 ; time constant of the associated exponential without censoring : 5.22 s , with censoring : 11.56 s ; ) .
we use this corrected rate in figures and parameter estimation .
this process of parameter estimation gives the following values : @xmath42 , @xmath43 and @xmath44 . in the case of the stopping phase ,
we need to estimate parameters only for equation .
we fit the equation on the experimental stop rates , giving @xmath45 , @xmath46 and @xmath47 .
the range of values obtained for @xmath7 , @xmath8 , @xmath17 , and @xmath18 are larger than zero , which is in accordance with the hypotheses considering a promoting role of n~m~ ( n~s~t~~ ) and a inhibiting role of n~s~s~~ ( n~m~ ) for the transition from s~s~ to m and from m to s~t~ , respectively .
* details on the calculation of pblackicted n~m~(t ) . * given @xmath48 and @xmath49 , it is straightforward to obtain n~m~(t )
. let us initially focus on n~m~(t ) during the departure phase .
let t~1~ , the departure time of the initiator and t~2~ , t~3~ ... t~n~m~=n~ the departure times of the first , second ...
n-1^th^ followers .
assume that at time t~1~ = 0 , the initiator departs from the group .
the time t~2~ - t~1~ we have to wait to observe the departure of the first follower is exponentially distributed and characterized by an average time : * details on the calculation of the commitment ( equation ) . * the commitment is defined as the probability of observing a naive abiding by the departure of the initiator . here
, we assume a general case where we have an arbitrary number of initiators i. this means that we have n - i naive individuals .
the probability per time unit per naive individual to switch from s~s~ to m is given by @xmath5 .
since n - i individuals can potentially decide to follow the initiator(s ) , the probability per time unit ( _ e.g. _ , seconds ) to observe the first naive to depart is given by : from this expression we can obtain the probability that no naive will depart during time t and that it happens between t and t + dt .
this probability takes the form : @xmath57 . for a given @xmath13 ( time requiblack to arrive to the target ) , the probability that a following event occurs for t @xmath58
@xmath13 is simply : now , for simplicity , we assume that the distribution of time @xmath13 , denoted by @xmath61 is an uniform distribution between the experimentally observed @xmath24 and @xmath25 ( i.e. , @xmath62 ) and compute the average @xmath63 : this is the expression we use to calculate the expected commitment that we compare with the experimental observation . exploring the model allows to check the effect of adding initiators on the value of commitment as a function of group size ( a ) .
also , for a given group size ( here n = 100 ) , we see that the commitment value increases non - linearly with the number of initiators ( b ) . finally , we were able to compute the number of initiators i needed to recruit all group members ( c ) .
* experimental setup .
* the setup is composed of two arenas ( 50 m side ) delimited in native irrigated pasture and surrounded by a 1.2 m visual barrier ( propylene net ) .
observations were made possible thanks to an observation tower located nearby .
yellow panels placed in the middle of each side can be levelled up and used as targets for the trained sheep thanks to a remote control .
digital snapshots were taken at one - second interval with a camera fixed at the top of the tower .
* statistics of the departing phase . *
( a ) latency to depart of the first followers as a function of the group size .
( b ) duration of the departing phase i.e. time elapsed between the departure of the first and the last followers .
the bottom and top of `` boxes '' show the first and third quartile of data .
the thick lines show the median and the whiskers represent the minimum and maximum values of the distribution .
* statistics of the stopping phase .
* ( a ) latency to stop of the first stoppers as a function of the group size .
( b ) duration of the stopping phase i.e. time elapsed between the first and the last stopping events .
* survival curves of the latencies of the first followers in groups of 32 without and with censoblack data ( adding the times spent walking by initiators to reach the target in trials when they fail to be followed ) * dashed curves show the experimental data without ( black curve ) and with ( black curve ) the censoblack data . respectively
plain black ( @xmath65 ; time constant = 5.26 s ) and black lines ( @xmath66 ; time constant = 12.56 s ) show the exponential curves fitted to the corresponding experimental data .
black crosses : the times spent walking by initiators to reach the target in trials when they fail to be followed .
dotted lines : @xmath67 confidence interval .
* model pblackictions .
* probability c of triggering collective motion ( commitment ) as a function of group size for various numbers i of initiators ( a ) , and as a function of i for a fixed group size ( b ) .
( c ) the minimum number of initiators requiblack to always observe a follow ( commitment = 1 ) as a function of group size ( black line ) .
the grey area indicates the combination of initiators and group size where the commitment is less than 1 ( risk zone : none follows the initiator ) . | for group - living animals , reaching consensus to stay cohesive is crucial for their fitness , particularly when collective motion starts and stops .
understanding the decision - making at individual and collective levels upon sudden disturbances is central in the study of collective animal behavior , and concerns the broader question of how information is distributed and evaluated in groups . despite the relevance of the problem , well - controlled experimental studies that quantify the collective response of groups facing disruptive events are lacking .
here we study the behavior of groups of uninformed individuals subject to the departure and stop of a trained conspecific .
we find that the groups reach an effective consensus : either all uninformed individuals follow the trained one ( collective motion occurs ) or none does . combining experiments and a simple mathematical model we show that the observed phenomena results from the interplay between simple mimetic rules and the characteristic duration of the stimulus , here , the time the trained individual is moving away .
the proposed mechanism strongly depends on group size , as observed in the experiments , and though group splitting can occur , the most likely outcome is always a coherent collective group response ( consensus ) .
the prevalence of a consensus is expected even if the groups of naives face conflicting information , e.g. if groups contain two subgroups of trained individuals , one trained to stay and one trained to leave .
our results indicate that collective decision - making and consensus in ( small ) animal groups are likely to be self - organized phenomena that do not involve concertation or even communication among the group members . |
the so - called _ adiabatic piston _ is a long known problem in classical thermodynamics , which can be stated as follows @xcite .
an isolated cylinder of length @xmath0 , containing a gas , is divided by an adiabatic wall ( no internal degrees of freedom ) , _ the piston _ , into two compartments ( fig .
[ fig:1 ] ) .
the initial condition is prepared in the following way : the piston is kept fixed by a clamp at a given position @xmath1 ; the gases in the left ( @xmath2 ) and right ( @xmath3 ) compartments are in equilibrium defined by their pressure , temperature and volume : @xmath4 . by assuming that the two gases are perfect and composed by @xmath5 molecules with equal masses @xmath6 , the gas state equation @xmath7 holds in both chambers ( where the boltzmann constant is set to unity by rescaling the temperatures ) . being the piston adiabatic , the two subsystems are in equilibrium even if @xmath8 . at @xmath9
the clamp is removed and the piston is free to move without friction with the cylinder .
the nontrivial question is to predict the system evolution and the final position of the piston and of the thermodynamic quantities . and
@xmath3 indicate the left and right compartments , respectively . ] at the beginning of last century , the above setup was used as experimental device for measuring the ratio @xmath10 of the specific heat of gases @xcite , that is linked to the period of the piston oscillations . renewed interest on the problem has lead to recent experiments @xcite .
meanwhile , several attempts were made to predict the final equilibrium state by using the laws of thermodynamics only , ending in controversial answers .
a naive application of the first two laws of thermodynamics lead to the ( wrong ) conclusion that the equilibrium conditions is @xmath11 .
a more careful treatment @xcite shows that the correct answer is @xmath12 .
however , such a condition says nothing about the final position of the piston and gas temperatures , which remain undetermined .
therefore , equilibrium thermodynamics can not predict the final state . to shed light on the problem one has to cope with the non - equilibrium process that occurs after the clamp removal . from a microscopic point of view
the adiabatic piston problem for ideal gases ( non - interacting particles ) can be described in terms of a one dimensional model , where the piston is a heavy particle of mass @xmath13 much larger that the mass @xmath6 of the gas molecules , which collide elastically with the piston .
as argued by feynman , the system first converges toward a state of mechanical equilibrium with @xmath14 , consistently with the thermodynamic prediction .
then , the pressure fluctuations , which are asymmetric because @xmath15 , very slowly drive the system toward thermal equilibrium @xmath16 @xcite . in this way
the final position of the piston and the thermodynamic quantities are determined .
more recently , the problem was the subject of renewed attention , mainly stimulated by the talk of lieb at the statistical physics conference in 1998 @xcite , and by the connection of this problem with the physics of mesoscopic systems @xcite and brownian motors @xcite . among the first attempts to understand quantitatively the time evolution of the adiabatic piston we mention crosignani et al .
@xcite who introduced a set of ordinary differential equations for the macroscopic observables .
however , this model was only able to account for the position of the piston in the state of mechanical equilibrium and not for the final thermodynamic one . remaining in the framework of ideal gases , a systematic investigation in statistical mechanics terms , together with numerical simulations , has been carried on in the last decade by gruber and coworkers @xcite . in these works
the problem has been examined in several limits ( see ref .
@xcite for a review ) .
in particular , in the thermodynamic limit taken by letting the system size @xmath0 and the piston mass @xmath13 to go to infinity by holding fixed the ratios @xmath17 and @xmath18 , it has been shown that the system evolution can be reduced to a set of ordinary differential equations for the macroscopic observables ( i.e. the gas left / right temperatures , and the moments of the piston velocity ) .
these equations were obtained by using the liouville and boltzmann equations .
within such an approach it is possible to control the deviations from maxwell - boltzmann distribution for the gas velocities , observed in the simulations , and a whole hierarchy of equations can be written for all moments of the piston velocity .
remarkably , these equations describe not only the reaching of mechanical equilibrium , which comes from the treatment at zero order in @xmath19 @xcite , but also the final equilibrium state , which comes from the first order terms in @xmath19 @xcite .
these analytical results have been shown by the same authors to be in agreement with numerical simulations of the ideal gas piston problem . though it remains open the problem of a detailed description of the early stage of the dynamics in which the presence of shock waves has an important impact on the dynamics .
some recent attempts in this direction can be found in ref .
@xcite .
when the initial pressures are different , the system phenomenology can be described as follows @xcite . in a first stage , the piston oscillates driven by the pressure difference .
these oscillations are then damped till the `` mechanical equilibrium '' state , @xmath20 but @xmath21 , is reached .
then , as argued by feynman , it follows a regime controlled by the asymmetry in the fluctuations felt by the left / right walls .
this phase is characterized by a very slow approach to the thermodynamic equilibrium , @xmath20 and @xmath22 , with the piston position fluctuating around the middle of the cylinder . in the oscillatory phase , both experiments @xcite ,
numerical computations and analytical arguments @xcite have shown the existence of two different regimes : _ weak and strong damping _ , the relevant parameter being @xmath23 .
for @xmath24 the adiabatic oscillations of the piston are weakly damped , while for @xmath25 they are over - damped , @xmath26 being @xmath27 @xcite .
still in the context of ideal gases , it is worth mentioning some recent approaches based on dynamical systems theory that have been developed by chernov , lebowitz and sinai @xcite . in this context , also the case of gases starting from non - equilibrium conditions has been considered @xcite .
clearly , the ultimate goal would be to quantitatively understand the behavior of the system for an interacting gas , but this seems to be still too ambitious .
indeed only very few studies analyzed the case of gas composed by interacting particles @xcite . in this paper
, we consider a limiting case which has the advantage of being more tractable while displaying most of the non - trivial features of the problem .
the basic idea of our approach is to assume that the gases in the two compartments are composed of interacting molecules and thus characterized by a relaxation time toward the equilibrium state .
our main hypothesis is that this time is very short compared with all the other characteristic times of the system . in particular , we require that any fluctuation away from equilibrium ( which is characterized by homogeneously distributed gas molecules with a maxwell - boltzmann velocity statistics ) induced by the collision with the piston is re - adsorbed before the new collision with the piston walls . physically speaking , the efficient re - adsorption of the fluctuations means that the ( mechanical ) work done by the piston is immediately converted into heat ; an obvious consequence is that shock waves are ruled out . for the sake of simplicity
, we also assume that the gases follow the perfect gas law .
these hypothesis make the problem tractable while retaining the basic phenomenology of the original problem .
although a microscopic model of a gas able to fulfill the above requirements may sound rather artificial , at a practical level such a `` microscopic model '' can be easily implemented on a computer .
the basic idea is to start with an equilibrium configuration with temperatures @xmath28 for the gases , and then reinitialize the gas molecules as soon as one particle collides with the piston .
the temperatures are recomputed after the collision and used for extracting a new configuration of the gas molecules .
the procedure is then repeated . in the following
we shall call such a model
_ randomized gas_. even though , no actual interaction among the particles is actually considered , one can think that the re - generation of the gas configuration from an equilibrium one ( but with the new temperature ) is the result of such `` unresolved '' interactions . with the above assumptions for the gas
, we will derive a set of ordinary differential equations for the time evolution of the macroscopic quantities describing the state of the system .
indeed the fact that the gas is always homogeneous and following the maxwell - boltzmann distribution allows us to compute the joint probability density function that , in a given state of the system , the first colliding gas particle hits the piston in a time @xmath29 and with a velocity @xmath30 .
then , by averaging over this joint distribution the energy and momentum exchange due to the collisions with the piston , we derive the evolution of the macroscopic observables .
the minimal set of variables required to have a closed set of equations is made up of the gas temperatures , the mean piston position , the first and second moment of the piston velocity .
the second moment is required for accounting the piston fluctuations which , as argued by feynman , are crucial for recovering the correct thermodynamic equilibrium @xcite .
the equations are derived perturbatively up to the first order in @xmath19 .
as we will see , though with a different approach and assumption , these equations are very similar to those derived by gruber and coworkers @xcite , in particular at the zeroth order in @xmath19 they are identical .
we compare then the evolution of the system obtained by simulations of the ideal and randomized gas .
in particular , the agreement in the first ( mechanical ) regime is quantitatively perfect in the case of the randomized gas . while in the second regime , which is dominated by the fluctuations , the agreement seems to be only qualitative .
somehow surprisingly , we found that , in this regime , a better quantitative agreement seems to be possible disregarding some @xmath31 terms .
however , with such terms excluded the equipartition of energy at equilibrium is violated by the piston . some hints to explain these findings could come by higher orders terms in the expansion .
unfortunately , the computation of the higher order terms is very cumbersome .
since the newest aspects of our work is in the proposed derivation and in the introduction of the randomized gas model , we present in this paper the all approach up to first order in @xmath19 .
the paper is organized as follows . in sect .
[ sec:1 ] we present our approach based on the collision statistics and derive the equations for the macroscopic observables . in sect .
[ sec:2 ] we compare the results of the model with those obtained by simulations .
discussions and conclusions can be found in section [ sec:3 ] . in order to avoid long appendices , the technical material , with the detailed derivation and all the formulas needed to make explicit
the equations , is presented as _ electronic supplementary material _
the underlying idea of our approach is to derive a set of deterministic dynamical equations for the `` macroscopic '' variables describing the evolution of the thermodynamic state of the system under the assumption that , at any time , the gases in both chambers are perfect and at equilibrium . in other words
the gases are able to instantaneously dissipate the fluctuations induced by the collisions with the piston .
thus a maxwell - boltzmann equilibrium state holds always in both compartments but , in general , with different temperatures and volumes .
while the above hypothesis define the macroscopic state of the gas , for the piston the problem is more subtle .
we would like to describe its motion on times longer than the single collisions , that is to average its instantaneous position and velocity @xmath32 over the collisions so to obtain a deterministic ( macroscopic ) trajectory defined by the average position @xmath33 and velocity @xmath34 the symbol @xmath35}$ ] denotes the average over the collisions . as discussed in the introduction , it is crucial to account also for the fluctuations of the piston velocity . for this reason , the second moment of the piston velocity @xmath36
is included in the description . in the thermodynamic limit we will consider
, one can argue that the fluctuation of the piston position can be safely ignored .
this means that in the following we will consider the piston position as a deterministic quantity and we shall use only the mean piston position @xmath37 .
given the piston position , the gas is characterized by the temperature @xmath28 , and volume @xmath38 , with @xmath39 and @xmath40 ( we assume a 1d geometry for the sake of simplicity ) .
being perfect gases , the pressures are given by the equation of state @xmath41 . in the sequel
we show how to derive a set of differential equations for the evolution of @xmath37 , @xmath42 , @xmath36 and @xmath28 . in order to keep the presentation as simple as possible , here we shall sketch how the equations can be derived and the averages performed skipping all the algebra of the computation , which is detailed in @xcite . for the formal derivation of the deterministic equations , we only need the above discussed assumptions and a microscopic ingredient : the elastic collision rules @xmath43 primes denote postcollisional velocities , and @xmath30 the colliding gas particle velocity .
the quantities we are interested in are the time derivatives of the macroscopic observables @xmath44 where we set the boltzmann constant @xmath45 .
the time derivatives should be computed starting from the collision rules as the averages @xmath46\rangle_{l , r}= \overline{[\dots]}_{l , r}/\overline{\delta t}$ ] suggest , @xmath47 being the mean collisions time ( for a more precise and operative definition see sect [ subsec : average ] ) .
the subscripts @xmath48 denote averages performed over the collisions with particles residing on the left / right compartments .
it is useful to introduce @xmath49 which evolves as @xmath50 where we used eqs .
( [ eq : vx ] ) and ( [ eq : sigma ] ) .
it should be noted that , at this level , the piston is completely described by @xmath42 and @xmath51 .
this amounts to assume that its velocity distribution is gaussian @xmath52 since at the initial time @xmath53 , one starts with @xmath54 and @xmath55 , the above probability distribution is initially a @xmath56 function . plugging ( [ eqnurto ] ) into ( [ eq : vx]-[eq : temp ] ) , we obtain @xmath57 \label{eq1:sigma } \\
\frac{{\rm d}t_{l}}{{\rm d}t } & = & \frac{4mm}{(m\!+\!m)^2}\left[m\langle \sigma_v^2\rangle_l+m\langle v_x^2\rangle_l \right.\nonumber\\ & -&\left.(m\!-\!m)\langle v v \rangle_l -m\langle v^2\rangle_l \right ] \,,\label{eq : templ } \\
\frac{{\rm d}t_{r}}{{\rm d}t}&=&\frac{4mm}{(m\!+\!m)^2}\left[m\langle \sigma_v^2\rangle_r+m\langle v_x^2 \rangle_r\right.\nonumber\\ & -&\left.(m\!-\!m)\langle v v \rangle_r -m\langle v^2\rangle_r \right ] \,.\label{eq : tempr}\end{aligned}\ ] ] the equation for @xmath58 ( [ eq1:sigma ] ) is obtained from ( [ eq : dsigdt ] ) and ( [ eq : sigma ] ) by using the collision rules ( [ eqnurto ] ) . in ( [ eq1:vx ] ) and ( [ eq1:sigma ] )
, the prefactor @xmath59 appears as a result of a time rescaling , that sets the time unit to the average collision time , which is order @xmath60 .
said differently , the change of the gas temperatures due to the collision with the piston is order @xmath60 .
notice that , for reasons that will become clear in the following , the average of the type @xmath61 is different from @xmath62 .
we anticipate that this difference is not due to a breakdown of molecular chaos hypothesis ( as one may naively think ) but to the fact that the collision statistics depends on the instantaneous value of the piston velocity .
more explicitly , @xmath62 represents only the zeroth order term of @xmath61 , and a term coming from the fact that @xmath63 is a fluctuating quantity will also appear .
notice also that the above equations conserve the total ( gas plus piston ) energy @xmath64 the consistent ( first order in @xmath19 ) equations can then be obtained from ( [ eq1:vx]-[eq : tempr ] ) by expanding the various prefactors , performing the limit @xmath65 and suitably expanding around it .
since the procedure is delicate , we proceed step by step . first of all we have to specify the limiting procedure , that as explained in @xcite can be done in different ways .
we are interested in the limit @xmath66 in which we keep fixed @xmath67 and the nondimensional mass ratio @xmath68 .
we have now to expand around this limit retaining all terms which are first order in @xmath19 ( and consequently at the first order in @xmath60 ) . aiming to make explicit the zeroth and first order terms we formally write the averages as : @xmath69\rangle=\langle [ \dots]\rangle^{(0)}+\langle [ \dots]\rangle^{(1)}\,,\ ] ] where the first and second terms on the r.h.s are the zeroth and first order terms of the expansion . how to explicitly perform such an expansion
will be explained in the following subsections .
we warn the reader that for maintaining the notation as compact and explicit as possible in the following we adopt the convection to indicate with @xmath46\rangle^{(1)}$ ] all averages which are @xmath31 irregardless if this comes from the expansion of the average or from the averaged quantity .
for instance , by direct inspection of eq .
( [ eq1:sigma ] ) at equilibrium , one easily realizes that @xmath58 is @xmath31 .
therefore , we shall always indicate its average with @xmath70 .
finally , notice also that all the terms involving powers of @xmath71 vanish at the zeroth order . keeping in mind these simplifications ,
the ( expanded ) equations become : @xmath72 \label{eq2:vx } \\ \strut{\hspace{-.1truecm}}\frac{{\rm d}\sigma^2_v}{{\rm d}t}\!\ ! & = & \ ! \!4r\left [ -\langle \sigma^2_v \rangle^{(1 ) } + { \frac{m}{m}}\langle ( v - v_x)^2 \rangle^{(0 ) } + \langle ( v - v_x)v \rangle^{(1)}+\langle ( v - v_x)v_x \rangle^{(1)}\right ] \label{eq2:sigma } \\ \strut{\hspace{-.6truecm}}\frac{{\rm d}t_l}{{\rm d}t}\!\!&= & \!\!4m\ ! \left\{\ ! \langle v_x(v_x\!-\!v)\rangle_l^{(0 ) } \!+\ ! \left[\ ! \langle
v_x(v_x\!-\!v)\rangle_l^{(1 ) } \!\!\!-\!2{\frac{m}{m}}\langle v_x^2 \rangle_l^{(0 ) } \!\!+\ ! 3{\frac{m}{m}}\langle v_xv\rangle^{(0)}_l \!\!+\!\langle \sigma_v^2\rangle^{(1)}_l \right.\right . -\left.\left.\langle
( v\!-\!v_x)v\rangle^{(1)}_l \!\!-\!{\frac{m}{m}}\langle v^2\rangle^{(0)}_l \!\right]\!\right\ } \label{eq2:tl } \\ \strut{\hspace{-.6truecm}}\frac{{\rm d}t_r}{{\rm d}t}\!\!&=&\!\ ! 4m\ ! \left\{\ ! \langle
v_x(v_x\!-\!v)\rangle_r^{(0 ) } \!+\ ! \left[\ ! \langle
v_x(v_x\!-\!v)\rangle_r^{(1 ) } \!\!\!-\!2{\frac{m}{m}}\langle v_x^2 \rangle_r^{(0 ) } \!+\ ! 3{\frac{m}{m}}\langle v_xv\rangle^{(0)}_r \!\!+\!\langle \sigma_v^2\rangle^{(1)}_r \right.\right . -\left.\left.\langle
( v\!-\!v_x)v\rangle^{(1)}_r \!\!-\!{\frac{m}{m}}\langle v^2\rangle^{(0)}_r \!\right]\!\right\ } \,.\label{eq2:tr}\end{aligned}\ ] ] before sketching the way the above averages can be computed ( see sect .
[ subsec : average ] and @xcite ) we briefly discuss some properties of the above equations .
the first observation is that eqs .
( [ eq2:vx]-[eq2:tr ] ) ensure the energy conservation ( [ eq : energy_conservation ] ) both at the zeroth and first order , meaning that the expansion is consistent .
notice also that the relative importance of the various terms is not the same at all times .
as the system evolves , their relative weights change corresponding to the different stages of the evolution , briefly summarized in the introduction and detailed in the following .
for example , consider @xmath58 . at the beginning @xmath73
, then it grows until it reaches its equilibrium value .
consequently , the terms which involve the velocity fluctuations are not important at the beginning .
while they become @xmath31 and play a crucial role in the final stage of the system evolution .
the opposite is true for the terms involving the average drift @xmath42 , which is close to zero in the second ( brownian ) stage of the evolution , and large in the first ( mechanical ) part of the system evolution . among the terms involving the piston velocity fluctuations
, we should mention a special role played by those in which it does not ( explicitly ) appear @xmath58 .
these are the terms of the form @xmath74 and @xmath75 , as we will see they will be both proportional to @xmath58 and , as anticipated , find their origin in the way the fluctuations of @xmath63 affect the collision statistics ( see next subsections and the supplementary material @xcite ) .
however , a closer inspection shows that @xmath76 is very small at all times . in the first stage
the fluctuations are negligible while in the second stage the average drift is very small . though we retained this term in the equations and in the numerical simulations , one can show that they can be removed without problem .
differently the terms @xmath77 are very important in the final stage of the evolution and , as discussed below , for obtaining the correct equilibrium state .
we are still left with performing the infinite volume limit .
we anticipate here that all the terms appearing in the averages are proportional to either @xmath78 , when coming from a left - chamber average , or @xmath79 when coming from a right - chamber average .
there is no other dependence on @xmath37 or @xmath0 in equations ( [ eq : x]-[eq : temp ] ) and , consequently , eqs .
( [ eq2:vx]-[eq2:tr ] ) .
this implies that one can simply introduce the hydrodynamic time @xmath80 and the rescaled @xmath37 coordinate @xmath81 . with this rescaling
, @xmath0 does not appear anymore in the equations and there is no need to perform the limit , meaning that there are no corrections to the results due to finite size effects .
for a simpler comparison with the simulations , we will keep writing in the following the equations for finite values of @xmath0 ; the corresponding expressions in the hydrodynamic time can be simply obtained with the above substitution , that in practice corresponds to set @xmath82 .
let us start a closer inspection of the equations starting from the zeroth order terms , i.e. assuming @xmath84 .
the velocity fluctuations of the piston are ignored ( meaning the motion of the piston is purely deterministic ) and the final equilibrium position depends on the initial conditions .
the result is anyway nontrivial .
as discussed in @xcite , having considered the dynamics and not only the thermostatics ( which only tells us the equality of pressures ) , we can now determine the mechanical equilibrium position .
in the sequel , we shall show that at the zeroth order in @xmath19 we obtain , by using a different approach , the same equations of gruber and coworkers @xcite , and following them we sketch how the mechanical equilibrium point can be computed . for obtaining the equations at the @xmath83 order
, we need to set @xmath85 and to ignore all averages indicated with the superscript @xmath86 in ( [ eq2:vx ] ) , ( [ eq2:tl ] ) and ( [ eq2:tr ] ) . in other words
we only need the averages @xmath87 \label{eq : vmed}\\ \langle v \rangle_r^{(0 ) } & = & -\frac{t_r}{2(l - x ) m } \left[1+\mathrm{erf}\left(v_x\sqrt{\frac{m}{2t_r}}\ , \right)\right]\nonumber \,,\end{aligned}\ ] ] with @xmath88 , and the averages @xmath89\label{eq :
vxmed}\\ \langle v_x\rangle^{(0)}_{r}\!\!\!\ ! & = & \!\ ! \frac{v_x}{l\!-\!x}\!\ ! \left[\sqrt{\frac{t_r}{2 \pi m } } e^{-\frac{m v_x^2}{2t_r}}\!+\ ! \frac{v_x}{2}\!+\!\frac{v_x}{2}\mathrm{erf}\!\left(v_x\sqrt{\frac{m}{2t_r}}\,\right)\!\right]\nonumber\,.\end{aligned}\ ] ] see the supplements @xcite for the derivation of the above expressions . assuming @xmath90 ( which is reasonable for realistic values of the physical parameters ) , and expanding ( [ eq : vmed ] ) and ( [ eq : vxmed ] ) in @xmath42 , the equations ( [ eq2:vx ] ) and ( [ eq2:tl]-[eq2:tr ] ) reads @xmath91 where for the friction coefficients it holds @xmath92 ensuring energy conservation ( [ eq : energy_conservation ] ) at the @xmath83 order ; from ( [ eq : vmed]-[eq : vxmed ] ) , at the lowest order in @xmath42 one has : @xmath93 we remind that @xmath39 and @xmath40 . note that apparently , in the limit @xmath94 , @xmath95 , this is not the case if the hydrodynamic rescaling is properly applied .
indeed , taking the hydrodynamic limit the above expression remains unchanged , keeping in mind that in this case the `` hydrodynamic volumes '' are @xmath96 and @xmath97 .
in particular , the damping coefficients go to a finite value also in the hydrodynamic limit .
it is worth remarking , that the linearized equations ( [ eq : vxlin]-[eq : trlin ] ) coincide with those derived in ref .
@xcite with a different method . in the absence of friction
, one can easily see that they describe a purely adiabatic transformation of a one - dimensional perfect ( mono - atomic ) gas .
indeed the first term on the r.h.s . of eqs .
( [ eq : vxlin ] ) is simply the pressure difference on the two sides of the piston , while the first term of eq .
( [ eq : tllin ] ) and ( [ eq : trlin ] ) can be obtained by differentiating with respect to time the equation of an iso - entropic process , namely @xmath98 where @xmath99 is the specific heats ratio and the initial conditions fix @xmath100 . in the absence of the friction terms
this would give rise to periodic oscillations of the piston .
as discussed in @xcite , the friction terms are responsible for the irreversible evolution toward a state of mechanical equilibrium for which @xmath101 where the tilde indicates the mechanical equilibrium quantities .
notice that in this framework irreversibility naturally emerges as a result of the averaging over the collisions @xcite .
( [ eq : equalpress ] ) is also the result of thermostatics , but it is not enough to determine @xmath102 and @xmath103 .
indeed , the full dynamics given by ( [ eq : vxlin]-[eq : trlin ] ) is needed to predict such mechanical equilibrium point , as shown in the sequel where we briefly summarize the results first derived in gruber et al .
first , notice that ( [ eq : equalpress ] ) together with ( [ eq : energy_conservation ] ) tell us that @xmath104 , with @xmath105 .
so that the equilibrium pressure is @xmath106 .
second , defining @xmath107 and by using ( [ eq : vxlin]-[eq : trlin ] ) one can easily see that @xmath108 , which if @xmath109 means that @xmath110 is conserved .
@xmath111 provides the missing condition to determine the mechanical equilibrium .
the resulting equation for the equilibrium point is therefore @xcite : @xmath112 which should be solved for @xmath102 after plugging @xmath113 and @xmath114 . when the first order ( in @xmath19 ) terms are retained , the terms in @xmath51 ( among which , as discussed above , we have also to consider the terms @xmath116 and @xmath117 ) allows for energy exchange among the two compartments mediated by the fluctuation of the piston .
of course , such terms start to play a role once the mechanical regime ( described by the @xmath83 order terms ) is finished , i.e. when the fluctuations of the piston become relevant .
this regime driven by the fluctuations results from the expansions in @xmath19 and @xmath60 which , as it happens commonly in brownian motor - like systems @xcite , are intertwined and add new ( sometimes unexpected ) features to the dynamics .
in particular , in this case one can show that eq .
( [ eq2:vx]-[eq2:tr ] ) evolve toward a nontrivial stable fixed point corresponding to the thermodynamic equilibrium , i.e. : @xmath118 the last equality can not be explicitly seen from ( [ eq2:sigma ] ) , which simply states that at equilibrium @xmath119 .
as we discussed , @xmath120 and with the explicit computation at equilibrium @xcite one can see that @xmath121 , which is a pleasant result since it is in agreement with the condition of equipartition of energy .
notice that eq .
( [ fixedpoint ] ) suggests to interpret @xmath122 as the temperature of the piston .
we conclude this subsection mentioning that the above equations are similar with the ones obtained by gruber , pache and lesne @xcite . due to the very long expressions involved in the equations at the first order , we could not decipher whether they are exactly equal . at the end of next section
we shall discuss the possible source of differences .
however , we stress that in these two works a different approach and different assumptions were made on the gases .
in particular , gruber , pache and lesne derived the equations from an expansion of the boltzmann and liouvulle equations . in some sense ,
our and their different assumptions can be seen as two different ways to close the hierarchy of equations to the second order and one should expect the phenomenology of the two equations to be , at least , qualitatively similar . section [ sec:2 ] is devoted to compare the evolution of the macroscopic observables obtained by integrating ( [ eq2:vx]-[eq2:tr ] ) with that of the microscopic model . in the following subsection
we detail the procedure by which the averages can be computed . in order to finalize our program
we have now to make explicit the averages in ( [ eq2:vx]-[eq2:tr ] ) .
let us start by making explicit the formal expression of averages such as @xmath123\rangle_{l , r}$ ] , which should be interpreted as follows
. denote with @xmath124 the probability of having a left / right collision in a time @xmath29 with a velocity @xmath30 for the r / l - particle conditioned to a realization in which the piston has velocity @xmath63 , and indicate with @xmath125 their sum , i.e. @xmath126 , which is normalized to 1 . then performing the average @xmath127 of a generic function @xmath128
means : @xmath129 where @xmath130 is the mean collision time .
let us now derive @xmath131 and @xmath132 .
in particular , we shall compute them in the thermodynamic limit by explicitly considering the terms order @xmath60 and consequently @xmath19 . these terms are those entering the averages we indicated with @xmath46 \rangle^{(1)}_{l , r}$ ] .
we start by the equilibrium distribution of the gases , which is uniform in the particle positions @xmath133 and maxwell - boltzmann for the velocities @xmath30 : @xmath134 that , under our assumptions , describe the gases at all times .
note that the above distributions depend ( parametrically ) on the dynamical variables @xmath37 and @xmath28 . from the equilibrium joint distributions ( [ eq : equilibrio ] ) we can derive the probability density @xmath135 that a particle on the left / right collides in a time @xmath29 given its velocity @xmath30 and the macroscopic state of the system defined by the temperatures , the piston position @xmath37 and velocity @xmath63 and the equation of state .
of course , most of the weight to such an hitting probability comes from particles that are close to the piston and that have a large ( negative ) relative velocity with it .
these particles are far from the bulk of the gas and in this derivation we assume that we can use the evolution of free particles to compute their hitting time with the piston .
within such an assumption , it is easy to realize that the probability @xmath136 is simply obtained through a change of variables from ( [ eq : equilibrio ] ) @xmath137 @xmath138 being the unitary step function . note that in the second expression we ignored the second @xmath138 function .
this is justified by the fact that @xmath29 indicates the time between two consecutive collisions , which for @xmath139 is always much shorter than the time needed for a particle to travel along a whole chamber .
further , considering only positive times , the functions @xmath140 do not depend on @xmath29 .
we shall then use the compact notation @xmath141 .
notice also that we use the instantaneous piston velocity @xmath63 and not its average @xmath42 , while we ignore the fluctuations in position .
moreover , we neglected possible correlations between the velocities and positions , which amounts to implicitly assume molecular chaos . given @xmath135 , the joint probability of having an impact of a particle with @xmath30 in a time @xmath29 is : @xmath142 the probabilities @xmath143 that a left / right particle collides in a time @xmath144 is then given by @xmath145 which we rewrite as @xmath146 with @xmath147 in the following we will use the shorthand notation @xmath148 and @xmath149 . considering that we have @xmath59 particles on both the left and right ,
the probability densities that one of them on the left / right impacts the piston in @xmath37 with a velocity @xmath30 are given by @xmath150^{n\!-\!1}\!\left[1\!-\!f_r(t_{m})\right]^n\!f_l(v|v ) \nonumber\\ & & \strut\hspace{-1.05truecm}g_r(t_{m},\!v)\!=\!n\!\left[1\!-\!f_r(t_{m})\right]^{n\!-\!1}\!\left[1\!-\!f_l(t_{m})\right]^n \!f_r(v|v ) .
\label{step1}\end{aligned}\ ] ] we can now perform the limit @xmath151 and @xmath152 holding @xmath153 fixed .
noticing that @xmath154 and expanding @xmath155^{n}\!=\ !
\exp[n\ln(1\!-\!f_{l , r})]\approx \exp[n(-f_{l , r}-f_{l.r}^2/2)]$ ] and retaining terms only up to @xmath60 , we find @xmath156 \!\tilde{g}_{l , r}(\tau , v|v ) \label{step3}\end{aligned}\ ] ] with @xmath157 by recalling that @xmath158 , and aiming to retain only the first order @xmath19 terms ( [ step3 ] ) can be rewritten as @xmath159 where we substituted @xmath160 in place of @xmath161 and @xmath42 in place of @xmath63 in the part which is already at the first order in @xmath19 .
however another contribution to the @xmath162 term comes from the expansion of the first term in ( [ eq : gfinal ] ) .
we need now to further expand ( [ eq : gfinal ] ) in @xmath63 around @xmath42 , this can be accomplished by taylor expanding @xmath163 @xmath164 noticing that @xmath165 and that @xmath166 we can write , for example , the ( expanded ) average collision time : @xmath167.\nonumber\end{aligned}\ ] ] finally , we can write the correct expansion of ( [ eq : averages ] ) at the first order in @xmath19 , by plugging all the expanded terms in ( [ eq : averages ] ) to have the zero and first order terms of the average of a generic observable , @xmath168 .
the result is : @xmath169\ , a\!+\!\left[\ ! { \frac{m}{m}}\frac{h_{l , r}}{r(h_l+h_r)}\!-\!\sigma_v^2\right.\nonumber\\ & & \strut{\hspace{-.4 cm } } \left.\frac{(h_l+h_r)^2 } { 2 } \int \!\!\!\!\int\!\ !
\mathrm{d}t\ , \mathrm{d}v\ , t\ , \partial^2_v\tilde{g}(\tau,\!v|v_x)\right ] \int\!\!\ !
\mathrm{d}v \ , f_{l , r}(v|v_x ) a \label{eq : av1}\end{aligned}\ ] ] which can be used to compute all the averages in ( [ eq2:vx]-[eq2:tr ] ) .
however we mention that there are exceptions to the above recipe .
for instance , the average of @xmath170 , whose result is @xmath171 and not @xmath172 .
this is due to the convention adopted for the average : remember that we chose to write the superscript @xmath86 also when the @xmath19 order comes from the averaged quantity and not from the expansion of the collision distribution , like in this case .
notice that in the first term of the r.h.s of ( [ eq : av1 ] ) , as for the average collision time , we expanded @xmath163 with the aid of ( [ eq : expansion ] ) . depending on the observable @xmath173 , which may have or not a linear term in @xmath174 , also the first derivative of @xmath163 may appear .
in this section we compare the evolution of the macroscopic observables given by ( [ eq2:vx]-[eq2:tr ] ) with numerical simulations of the microscopic model .
we consider two kinds of microscopic simulations : the ideal and the randomized gas . the latter is meant to fulfill the assumptions we made on the gas in deriving the equations .
let us now better clarify how this is realized . from a computational point of view
, it is very easy to realize the randomized gas , the idea is to let the gases relax in an artificial manner through a randomization procedure .
more precisely , the simulations are performed in the following way : we generate an equilibrium configuration of the system , corresponding to given values of the macroscopic observables .
then the system is let to evolve up to the first collision of the piston _ without _ interactions among the gas particles .
after the collision , the energy of the gas containing the colliding particle changes ; the fast relaxation of the system is mimicked by updating the gas temperature ( corresponding to the new energy ) and redrawing an equilibrium configuration of the gases corresponding to the new temperatures ( and volumes ) .
then the process is iterated .
notice that in this way one keeps track of all the observables except the piston velocity fluctuation , which needs some kind of average to be defined / measured .
collisions are evaluated , as for the ideal gas , with an event driven algorithm . in figure
[ fig : evolution ] we show the evolution of the piston position by numerical simulations of the ideal and randomized gas for two different values of @xmath23 ( namely @xmath175 ( top ) and @xmath176 ( bottom ) . as discussed in the introduction
( see @xcite for a more detailed treatment ) these two choices correspond to the case of strongly and weakly damped oscillations , respectively . in both cases
one has two regimes : mechanical and brownian .
the former is characterized by damped oscillations of the piston , which in the strongly damped case are very few or nonexistent .
for the ideal gas , as discussed by gruber et al .
@xcite the detailed damping of the oscillations depends on the presence of shock waves , which are absent in the randomized gas .
in fact , the latter is damped much more efficiently than the former .
as shown in the bottom panel , for @xmath176 the evolution is weakly damped and many oscillations are observable . in this case , the piston is very slow and the gases perform quasi - adiabatic oscillations , which are damped more and more mildly as @xmath177 . again , also in this case the damping appear to be much more efficient in the randomized case . )
the dotted line indicates the adiabatic phase @xmath178 , while the solid straight line the isobar that characterizes the final phase ( brownian motor like regime ) .
the rightmost arrow indicates the initial state of the gas in left compartment @xmath179 @xmath180 which evolves ( red dots ) toward the final equilibrium in the middle .
the leftmost arrow indicate the initial state of the gas in left compartment @xmath179 @xmath181 which evolves ( blue dots ) toward the final equilibrium in the middle . the simulation is done with the ideal gas , for which the initial oscillations are much more evident that for the randomized one .
, scaledwidth=49.0% ] the brownian motor - like regime @xcite , occurs when the oscillations are completed damped , i.e. the mechanical equilibrium is realized with approximately equal pressures @xmath182 , which differ only for terms @xmath31 . from now on ,
both the ideal and randomized gas remain in a state of marginal equilibrium approximately along the isobar @xmath183 , which is the prediction of thermodynamics ( shown in fig .
[ fig : tv ] for the ideal gas only ) .
the brownian motor - like mechanism is responsible for heat transfer from the warmer to the colder chamber mediated _ via _ the piston fluctuations @xcite .
this stage occurs on a long timescale ( proportional to @xmath13 ) . for realistic values of the various parameters
, one can realize that the timescales necessary for reaching the thermodynamic equilibrium state are enormous .
it is therefore very difficult to observe this regime in experiments , and well controlled numerical simulations are mandatory . in the final state
the two chambers have the same temperatures and the piston fluctuates reaching equipartition with the same temperature .
it is worth noticing that for the ideal gas , since the molecules interact only through the collisions with the piston , the probability distribution of the velocities of the gas molecules may ( and actually do ) deviate sensibly from the maxwell - boltzmann distribution that is recovered only after the reaching of the final equilibrium state @xcite . by construction , this problem is not present in the randomized gas which is forced to remain maxwell - boltzmann . in the following
we shall compare the evolution of the macroscopic equations with the simulations .
we start the comparison by considering the oscillatory regime in the weakly damped case . here we know from previous studies @xcite that the model is able to correctly predict the period of the oscillations .
the period of the oscillations in the model can be estimated from the linearized dynamics at the zeroth order in @xmath19 , eqs .
( [ eq : vxlin]-[eq : trlin ] ) . as discussed in sect .
[ sec : expandedequations ] , the oscillations will occur around a mechanical equilibrium position @xmath102 defined by eq .
( [ eq : eqpoint ] ) . then to recover the period of the oscillations it is enough to linearize ( [ eq : vxlin]-[eq : trlin ] ) around the mechanical equilibrium state defined by @xmath102 , @xmath103 and @xmath184 .
linearizing ( [ eq : tllin]-[eq : trlin ] ) one can easily recognize that the equation define an iso - entropic process , i.e. ( [ eq : adiabatic ] ) , meaning that @xmath185 and @xmath186^{c_p / c_v-1}$ ] , which plugged in ( [ eq : vxlin ] ) and expanding in @xmath187 lead to the equation of a damped oscillator : @xmath188 where we recall that @xmath99 and @xmath106 ( see sec . [ sec : mecheq ] ) .
since the friction coefficient does not modify the period one immediately gets : @xmath189 this formula is at the basis of the measurement of the specific heat ratio in experiments @xcite . , @xmath190 , @xmath191 , @xmath192 , @xmath193 , @xmath194 corresponding to @xmath195 .
( top ) evolution of the piston position .
( bottom ) evolution of the temperatures .
blue open squares refer to the ideal gas and red filled circles to the randomized gas , the solid line is the model ( [ eq2:vx]-[eq2:tr ] ) . ] , @xmath190 , @xmath191 , @xmath192 , @xmath193 , @xmath194 corresponding to @xmath195 .
( top ) evolution of the piston position .
( bottom ) evolution of the temperatures .
blue open squares refer to the ideal gas and red filled circles to the randomized gas , the solid line is the model ( [ eq2:vx]-[eq2:tr ] ) . ] in fig .
[ fig : oscill - x ] , we compare the ideal gas simulations with that of the randomized gas and the numerical integration of the determinist equations ( [ eq2:vx]-[eq2:tr ] ) .
as one can see , the ideal and randomized gas have the same period , while the damping is different , and the model is in perfect agreement with the randomized gas simulations . integrating the equation at the zeroth order , we find @xmath196 which agrees with the predicted value ( [ eq : eqpoint ] ) and eq .
( [ eq : ruch ] ) predicts @xmath197 which is in very good agreement with the the measured period in both the randomized and ideal gas .
( top ) with superimposed the thick black line superimposing on randomized gas data ( blue empty circles ) is obtained by the numerical integration of ( [ eq2:vx]-[eq2:tr ] ) .
the agreement of the model with the randomized gas is perfect also for the temperatures ( not shown ) . ]
we mention that in ref .
@xcite a detailed study of the period of the oscillations was reported . since our linearized equations coincide with those of ref .
@xcite , we shall not repeat here this study .
it is however interesting to compare the first stage of the evolution of the ideal and randomized gas with the model in the case of strong damping . in fig .
[ fig : strong ] we show the leftmost plot of fig .
[ fig : evolution ] ( top ) , as one can see although the model is unable to reproduce the oscillation of the ideal gas , its evolution coincides with that of the randomized gas . as discussed in ref.@xcite ,
estimating the decay rate of the oscillations for the ideal gas is a nontrivial task , since it requires a detailed study of the dissipation mechanisms of the shock waves created by the piston motion .
these shock waves survive for a long time in the ideal gas , while they lifetime is expected to be shorter in the presence of interactions , which should be able to decrease their coherence .
the randomized gas represents a sort of limiting case of interaction in which the shock waves are completely absent .
most likely the absence of shock waves is at the origin of the faster damping of the oscillations in the randomized gas and of the very good agreement between its evolution and the one obtained from the macroscopic equations
. it would be interesting to investigate the transition between weak and strong damping in the randomized model .
this is far from the aim of the present paper , however the above results suggest that the critical value of @xmath23 for the transition will probably be smaller ( but of the same order ) of that of the ideal gas which is order unity .
as the mechanical equilibrium is reached @xmath182 , the system slowly evolves , along an approximate isobar , driven by fluctuations toward the thermodynamical equilibrium defined by eq .
( [ fixedpoint ] ) , which is the ( stable ) fixed point of the ordinary differential equations ( [ eq2:vx]-[eq2:tr ] ) . ,
@xmath190 and @xmath198 with @xmath199 and @xmath200 with @xmath201 : red triangles refer to the randomized gas and the blue squares to the ideal gas , respectively .
the dashed line is the prediction of the macroscopic equation , the solid line ( which perfectly superimposes on the randomized gas data ) is explained in the text .
the simulation data are obtained by performing an average over about @xmath202 realizations to reduce the fluctuations . ]
we shall now compare the evolution given by the macroscopic equations with that of the ideal and randomized gases in the brownian regime . to minimize the possible differences between the ideal and randomized gases we performed a simulation which starts in a mechanical equilibrium state having the gas molecules distributed according to the maxwell - boltzmann distribution .
for the ideal gas , this might be not the typical situation : usually when the system arrives to the mechanical equilibrium from a non - equilibrium state , it may be strongly non - maxwellian @xcite . as exemplified in fig .
[ fig : relax ] , where we show the deviation from the final equilibrium of the piston position , @xmath203 , the relaxation is exponential .
the randomized gas relaxes faster than the ideal one , likely because the latter develops slightly non - maxwellian distribution ( we observed that the difference in the relaxation time tends to diminish as the number of particles is increased ) .
note that the simulation has been performed with @xmath204 because the time scale for reaching the equilibrium is controlled by @xmath13 ; with @xmath201 , as here , working with @xmath205 would have implied the necessity to reach too large time scales to study the relaxation .
as shown in the figure , the macroscopic equations ( dashed line ) predict a relaxation slower than both the randomized and ideal gases .
this came out as a surprise for us , because we were encouraged by the very good agreement in the mechanical regime , discussed in the previous section .
with the aim of understanding such a difference we examined all the terms appearing in the macroscopic equations and we realized that the mismatch in the relaxation was due to the terms @xmath206 that , as discussed in sect . [
sec : brown ] , are those which ensures the equipartition of energy at equilibrium , i.e. the fact that @xmath207 . in particular , eliminating such terms from ( [ eq2:vx]-[eq2:tr ] ) ( note that this is not affecting the conservation of energy ) , we found a perfect agreement between the macroscopic equations , now modified , and the randomized gas , as shown by the solid line in the figure .
nevertheless , with such a modification the energy equipartition is not verified anymore and at equilibrium @xmath208 .
( bottom ) now plotted in log - lin scale .
the ( red ) symbols refer to the randomized gas simulations , the dashed curve to the evolution predicted by the macroscopic equations and the solid curve ( which perfectly superimposes on the simulation data ) is explained in the text .
( bottom ) same as ( top ) but for the evolution of the gas temperatures @xmath209 and @xmath210 as in label .
the two curved which start from zero show the evolution of the piston temperature @xmath211 as obtained from the macroscopic equations ( dashed curve which saturates to @xmath212 and as explained in the text the solid one which saturates ad @xmath213 . ]
( bottom ) now plotted in log - lin scale .
the ( red ) symbols refer to the randomized gas simulations , the dashed curve to the evolution predicted by the macroscopic equations and the solid curve ( which perfectly superimposes on the simulation data ) is explained in the text .
( bottom ) same as ( top ) but for the evolution of the gas temperatures @xmath209 and @xmath210 as in label .
the two curved which start from zero show the evolution of the piston temperature @xmath211 as obtained from the macroscopic equations ( dashed curve which saturates to @xmath212 and as explained in the text the solid one which saturates ad @xmath213 . ]
one may think that the agreement is incidental , and so we did a more severe test . in fig .
[ fig : relax2 ] we show the evolution of the piston position ( top ) and of the gas temperatures ( bottom ) with the simulations of the randomized gas previously shown in fig .
[ fig : evolution](bottom ) .
as one can see , the modified ( solid lines ) and original ( dashed lines ) macroscopic equations generate , a part from @xmath58 , two indistinguishable dynamics for @xmath37 and @xmath28 up to times for which the mechanical equilibrium is reached .
actually for such times the dynamics is essentially given by the zeroth order equations ( [ eq : vxlin]-[eq : trlin ] ) which are the same for both the original and modified equations : this explain their behavior in the mechanical regime .
as one can see the main difference in the dynamics is that retaining the terms @xmath214 leads the system to stay for a longer time interval in a state of approximate mechanical equilibrium and thus to a slower relaxation to equilibrium .
the evolution of the piston temperature @xmath215 shown in fig .
[ fig : relax2 ] ( bottom ) clearly show the difference in the two dynamics , and in particular the breaking of equipartition at equilibrium for the modified dynamics .
it should be stressed that the behaviors shown in the above figures is not related to the peculiar parameters choice as it has been verified in other simulations ( not shown ) .
the picture emerging from this comparison is that the model we introduced goes to the correct equilibrium state and respects the basic physics principles as conservation of energy and its equipartition . on the other hand ,
the relaxation is slower than the one of the microscopic model it should correspond to ( i.e. the randomized gas ) . as shown in the above
figures the terms responsible for this slowing down are the ones coming from the fluctuation of the piston velocity in the expansion of the function @xmath125 . at present
we dot not have a definite understanding either of this difference in the dynamics of the macroscopic equations and of the randomized gas , or of the very good agreement between the modified equations and the randomized gas .
we suspect that the contribution to the relaxation time of these terms might be counterbalanced by the resummation of higher order terms in the @xmath19 expansion .
however , carrying the analysis to higher order terms is very demanding due to the proliferation of terms in the expansion , and therefore we shall not discuss it here .
we conclude this section by mentioning that the equations derived by gruber and coworkers @xcite predict a reasonable relaxation time and equipartition at the same time .
we remark that going to higher order in their approach requires to include into the description also higher order moments of the piston velocity , namely @xmath216 ; while within our approach the description remains at the level of the second order moment because the statistics is constrained to remain gaussian ( maxwell - boltzmann ) for the gases and the piston
. this may be the origin of the difference between our and their derivation .
their equations at the first order might likely correspond to a resummation of higher order terms in our approach , meaning that their closure with the second moment of the velocity is exact while ours remains only approximate .
in the framework of kinetic theory , we derived a set of deterministic equations describing the evolution of the macroscopic variables in the adiabatic piston problem .
our basic assumptions are that at each time the gases in the two compartments are perfect , spatially homogeneous , and described by the maxwell - boltzmann statistics .
thus , at the level of simulations , a ( randomized ) gas model has been introduced with aim to have microscopic model respecting such assumptions .
we obtained a set of five ordinary differential equations for the variables that describe the macroscopic state of the system , namely the mean position of the piston , the average velocity of the piston , the temperatures of the gases in the two compartments and the second moment of the piston velocity .
the equations are derived up to the first order in @xmath19 . at the zeroth order
they describe a deterministic piston characterized by a velocity distribution collapsed on the mean , namely @xmath217 .
this is enough to solve the problem of finding the final state of mechanical equilibrium and the result coincides with that derived in ref .
@xcite by using a different approach . at the first order
the fluctuations of the piston velocity , now assumed to be gaussian , allow for recovering the correct final thermodynamic equilibrium .
although the evolution of the macroscopic observables provided by this set of equations is in good qualitative agreement with simulations of the randomized gas , we found some quantitative discrepancy for the relaxation timescales .
apart from the performance in comparing with the simulations , we would like to stress the conceptual aspects of the method we developed .
it allows for a transparent description of the macroscopic dynamics of a nontrivial non - equilibrium problem , similarly to how the perfect gas law can be derived from the microscopic collisions by using elementary kinetic theory .
we thanks e. caglioti , f. cecconi , b. crosignani , p. di porto , a. lesne , g.p .
morriss , and c. van den broeck for useful discussions , remarks and correspondence .
this work has been partially supported by prin2003 `` complex systems and many body problems '' , and prin2005 `` statistical mechanics of complex systems '' by miur .
s.p . has benefited from a mec - mius joint program ( italy - spain integral actions ) . | a simplified version of a classical problem in thermodynamics the adiabatic piston is discussed in the framework of kinetic theory .
we consider the limit of gases whose relaxation time is extremely fast so that the gases contained on the left and right chambers of the piston are always in equilibrium ( that is the molecules are uniformly distributed and their velocities obey the maxwell - boltzmann distribution ) after any collision with the piston . then by using kinetic theory we derive the collision statistics from which we obtain a set of ordinary differential equations for the evolution of the macroscopic observables ( namely the piston average velocity and position , the velocity variance and the temperatures of the two compartments ) .
the dynamics of these equations is compared with simulations of an ideal gas and a microscopic model of gas settled to verify the assumptions used in the derivation .
we show that the equations predict an evolution for the macroscopic variables which catches the basic features of the problem .
the results here presented recover those derived , using a different approach , by gruber , pache and lesne in _
j. stat .
phys .
_ * 108 * , 669 ( 2002 ) and * 112 * , 1177 ( 2003 ) . |
it has been widely speculated that the energy that heats the corona comes from the convective flows in the photosphere .
the energy is somehow transported into the coronal part through the magnetic field .
it is natural to think that alfvn waves channel the energy to the corona .
indeed , these waves have been observed in the solar atmosphere ( ulrich 1996 ) , and are ubiquitous in the extended corona the fast solar wind ( smith et al .
however alfvn waves are difficult to dissipate in collisionless plasmas .
hence , non - linear processes have been assumed to cascade the wave energy from low to high frequencies where wave dissipation is readily possible ( hollweg 1986 , li and habbal 2003 ) .
when alfvn waves propagate in a partially ionized plasma , neutral - ion collisions produce a channel for the wave dissipation ( de pontieu & haerendel 1998 ; de pontieu et al .
2001 ; leake et al . 2005 ) .
ions and neutrals can be collisionally coupled .
a slippage between the ion and neutral populations leads to the wave dissipation .
if we consider neutrals are constantly ionized at the chromosphere , we will show that these newly created he@xmath4 ions will be picked - up by the wave and will be energized . in this letter , a new scenario , ion pickup by an alfvn wave ,
is explored .
the pickup process can lead to the heating of ions , and it must also dissipate the wave . in section 2 ,
we discuss the ion pickup process by an instantly introduced wave ( or identically , the ions are created instantly ) . in section 3 ,
the pickup of continuously created ions by an alfvn wave is investigated .
finally , in section 4 we discuss possible applications of the ion pickup process in the solar atmosphere .
consider a parallel propagating monochromatic dispersionless alfvn wave with angular frequency @xmath5 and wave - number @xmath2 , and @xmath6(@xmath7 is the alfvn speed ) . the wave electromagnetic field @xmath8 and @xmath9 are @xmath10 , \\\delta { \mathbf e}_w= -v_a { \mathbf b}_0/b_0 \times \delta { \mathbf b}_w , \end{array } \right.\ ] ] @xmath11 and @xmath12 are unit vectors , @xmath13 the background magnetic field , and @xmath14 denotes the wave phase .
the motion of a particle is described by @xmath15 , { d z \over dt}= v_z.\ ] ] [ q1 ] let @xmath16 , @xmath17 , and @xmath18 , we have @xmath19 @xmath20 where @xmath21 , @xmath22 , and @xmath23 refers physical quantities of ion species @xmath23 . as a first order approximation , @xmath24 is a constant , where @xmath25 is the initial ion parallel velocity .
the approximation is valid when @xmath26 is small and the wave frequency is low so @xmath27 . with the initial condition @xmath28 and @xmath29 , the solution of eq .
( [ q2 ] ) for a low beta plasma is ( wu et al .
1997 , lu et al .
2006 ) @xmath30e^{-i \omega_0 t } -v_a{b_k \over b_0 } e^{-ik[v_at - z]}. \label{q4}\ ] ] where @xmath31 and we have adopted approximations @xmath32\approx \omega_0 $ ] , @xmath33 . the last term of @xmath34 is the perturbed ion velocity of an alfvn wave .
the first term is due to the gyromotion of the particle and the modification of the gyromotion due to the wave .
note , if @xmath35 and @xmath36 , the particle is surfing on the wave , there is no gyromotion .
now let us consider an ensemble of newly created particles with a maxwellian distribution .
the particles average parallel speed is zero but the transverse component is @xmath37 @xmath38 where @xmath39 , a constant , describes the degree that the newly created particles are settled in the wave field . if @xmath40 , the newly created particles are already settled .
@xmath41 denotes the random perpendicular thermal speed .
after time @xmath42 particles with different initial position @xmath43 will arrive at @xmath44 .
the average transverse velocity at @xmath44 is @xmath45 @xmath46 } e^{-i \omega_0 t } e^{-\left({v_\parallel(0 ) \over v_{th } } \right)^2 } dv_\parallel(0)\ ] ] @xmath47 where @xmath48 , @xmath49 , and @xmath50 is the initial temperature of species @xmath23 .
eq ( [ q6 ] ) illustrates the pickup of ions by the wave .
note @xmath51 when @xmath52 .
hence , regardless of the value of @xmath39 , these particles will be picked up by the alfvn wave : eventually @xmath53 is determined by the wave only . subtracting ( [ q6 ] ) from ( [ q4 ] ) ,
one finds the random perpendicular velocity at @xmath44 : @xmath54 } e^{-i\omega_0 t}\ ] ] @xmath55 eq .
( 8) describes the particles gyrospeed in the frame of @xmath53 .
when @xmath56 ( initially the newly created particles have zero flow speed ) , particles will be strongly scattered in the phase space .
the second term on the right hand side is crucial for our understanding the heating in the pickup process . in the phase space frame of @xmath53 ,
a randomization process is complete when particles with characteristic speed @xmath57 moving from @xmath43 to @xmath44 are scattered in the perpendicular direction and a phase shift @xmath0 in the gyrospeed relative to particles with @xmath36 is produced .
this translates to @xmath58 . for a finite amplitude alfvn wave ,
if @xmath59 , the distribution function will be a ring . hence particles are strongly heated . on the other hand , if @xmath40 , initially particles are already picked up " by ( surfing on ) the wave
. only the initial random motion survives , and there is no heating .
one notices that if @xmath60 , @xmath61 , there is also no heating .
hence , the heating is a warm plasma effect . with @xmath53
, the perpendicular temperature can be found @xmath62 @xmath63 , \label{q8}\ ] ] where @xmath64 is the plasma beta of species @xmath23 .
the heating is obviously mass - proportional . substituting eq .
( [ q4 ] ) into eq .
( [ q3 ] ) , and letting @xmath33 and @xmath65 \approx\omega_0 $ ] , one finds @xmath66\ } , \label{q9}\ ] ] using the same procedure to obtain @xmath53 and @xmath67 , we find the average parallel velocity and temperature : @xmath68 , \label{q10}\ ] ] @xmath69 @xmath70\right ] , \label{q11}\ ] ] where @xmath71 .
when @xmath72 , @xmath73 , and @xmath74 .
we then obtain asymptotic values : @xmath75 @xmath76,\ ] ] @xmath77 \label{q13}.\ ] ] subtracting ( [ q10 ] ) from ( [ q9 ] ) , when @xmath78 , one finds @xmath79 .
\label{q14}\ ] ] it is possible that the second term on the right hand side is larger than the initial random thermal velocity @xmath25 . in this case , the velocity distribution plotted in the @xmath80 plane may be a ring as well ( see fig.2 ) . to verify the above analysis ,
a test particle simulation was conducted by calculating the full dynamics of particles ( eq . [ q1 ] ) in the electromagnetic field of a given alfvn wave described by eq .
the equations were solved using the boris algorithm with time step @xmath81 ( @xmath82 is the proton gyrofrequency ) and periodic boundary conditions .
initially , 200000 particles with maxwellian velocity distribution were evenly distributed in a region with length @xmath83 ( 1000 cells ) .
the flow speed of these particles was zero .
the average parallel velocity , and the parallel and perpendicular temperatures were obtained by using the following procedure : we first computed @xmath84 , @xmath85 , and @xmath86 in every cell ( @xmath87 denotes an average over a cell ) .
these quantities were then averaged over all cells , so only random motions contribute to the temperatures .
fig.1 shows the time evolution of ( a ) the perpendicular temperature @xmath88 ( here 0 refers to the initial value ) and the parallel temperature @xmath89 , ( b ) the temperature anisotropy @xmath90 , and ( c ) the average parallel velocity @xmath91 of the test particle simulation .
the shaded areas " are due to rapid ion gyrations .
the results are consistent with our analytical predictions . at @xmath92 , an asymptotic stage
is reached : @xmath93 , @xmath94 and @xmath95 with the temperature anisotropy @xmath96 .
fig.2 displays scatter plots of the helium ions in the five cells around @xmath97 at @xmath98 , 2800 , 6600 and 11000 .
initially , helium ions satisfy the maxwellian distribution with @xmath99 and @xmath100 .
helium ions are dramatically scattered in the transverse direction in the phase space .
the ring velocity distributions are nicely shown in fig.2 .
the ring velocity distribution is unstable ( lu and wang 2005 ) .
the velocity distribution may eventually be thermalized by relevant microscopic instabilities of the ring distribution .
the bulk acceleration of particles along @xmath101 is obvious .
this is because in the frame of the alfvn phase speed , the kinetic energy of particles is conserved .
as particles are picked up by the wave , they gain a bulk acceleration along @xmath101 .
in a partially ionized plasma , the creation of new ions due to photoionization or collisional ionization is highly random . consider a region with a low plasma beta and where the amplitude of an alfvn wave is finite .
when ions are created from neutrals , they are subject to rapid gyromotions .
the continuous creation of ions represents a continuous phase shift in the ion gyrospeed .
this will naturally mix the phase of particles gyrospeeds . to explore the pickup process of continuously created ions ,
a new test particle simulation is conducted .
the alfvn wave and the simulation box are the same as in section 2 .
however , 1000 particles are deployed every @xmath102 at @xmath103 ( one particle per cell ) .
the beta value of the newly created he@xmath4 ions is 0.01 and they have zero flow speed .
the results of the simulation are shown in figs .
3 and 4 , which are plotted in the same way as figs . 1 and 2 .
note that in the @xmath104 plane , because the radius of the ring ( fig.3 ) is @xmath105 , the origin , where new particles are created around there , is always on the ring .
when new he@xmath4 ions are continuously created , the heating and scattering process are completed when they have filled the ring of radius @xmath105 in the @xmath106 plane .
this only takes one gyroperiod of he@xmath4 ions , or when @xmath107 , as shown in fig.3 and 4 .
hence the heating of these particles is extremely rapid .
the ion creation process was terminated at time @xmath108 , however no change is observed as @xmath109 .
we believe that the rapid heating due to alfvn waves reported by wang et al .
( 2006 ) is due to their deployment of particles within one gyroperiod , even though they used many wave modes in their computations .
the alfvn wave pickup of newly created ions may have applications at the top of the chromosphere or lower transition region .
let s consider the region with an ion cyclotron frequency much higher than coulomb collision frequencies , low plasma beta and finite amplitude alfvn waves . for neutral helium , when wave frequencies are higher than the ion neutral collision frequencies , the perturbed ion and neutral velocity will have a considerable difference between their phases and amplitudes ( de pontieu & haerendel 1998 ) . in coronal funnels ,
the plasma beta at the lower transition region is much smaller than unity ( li 2003 ) . for finite amplitude waves , the pickup process described above may operate and heat newly created he@xmath4 ions .
due to collisions , the temperature is not expected to reach the value given by ( [ q13 ] ) . instead , the energy will be transferred to protons and electrons via coulomb coupling .
obviously , the thermal energy of he@xmath4 gained through the pickup process will be passed to he@xmath110 later on when he@xmath110 are created . as the ionization continues to the transition region
( hansteen et al . 1997 ) , the pickup process will contribute to the plasma heating there .
the pickup of instantly created ions by an alfvn wave or the pickup of ions by an alfvn wave instantly introduced into a plasma is quite efficient even for low frequency waves in collisionless plasmas .
the pickup process is complete when @xmath111 or @xmath112 , where @xmath113 is the period of the wave .
transverse oscillations of active region coronal loops that suddenly appear with periods of 2@xmath11433 minutes have been frequently observed by trace ( nakariakov et al .
1999 ; aschwanden et al .
the oscillations usually just last several wave periods .
the relative amplitude @xmath115 of these waves can reach 0.05 ( aschwanden et al .
2002 ) . for typically observed 10@xmath116 k coronal loop with density of @xmath117 @xmath118 and magnetic field of 30 g ,
the plasma beta value is 0.00385 . due to the pickup process
, the perpendicular temperature of helium ions and oxygen ions may be increased to 3.6 and 11.4 times of their original values , respectively . | two scenarios of possible ion heating due to finite amplitude parallel propagating alfvn waves in the solar atmosphere are investigated using a 1d test particle approach .
1 .
a finite amplitude alfvn wave is instantly introduced into a plasma ( or equivalently , new ions are instantly created ) .
2 .
new ions are constantly created . in both scenarios ,
ions will be picked up by the alfvn wave . in case 1 ,
the wave scatters ions in the transverse direction leading to a randomization ( or heating ) process .
this process is complete when a phase shift of @xmath0 in the ion gyrospeed is produced between particles with characteristic parallel thermal speed and particles with zero parallel speed .
this corresponds to @xmath1 ( @xmath2 is the wavenumber and @xmath3 is the ion thermal speed ) . a ring velocity distribution can be produced for a large wave amplitude .
the process yields a mass - proportional heating in the transverse direction , a temperature anisotropy and a bulk flow along the background magnetic field . in case 2
, continuous ion creation represents a continuing phase shift in the ion gyrospeed leading to heating .
new particles are picked up by the alfvn wave within one ion gyroperiod .
it is speculated that the mechanism may operate in the chromosphere and active regions where transient events may generate finite amplitude alfvn waves . to appear in apj letters may 10 , 2007 . |
cortical activity manifests highly complex behaviors which is often strongly characterized by the presence of noise .
reliable responses to specific stimuli often arise at the level of population assemblies ( cortical areas or cortical columns ) featuring a very large number of neuronal cells presenting a highly nonlinear behavior and that are interconnected in a very intricate fashion . understanding
the global behavior of large - scale neural assemblies has been a great endeavor in the past decades .
most models describing the emergent behavior arising from the interaction of neurons in large - scale networks have relied on continuum limits ever since the seminal work of wilson and cowan and amari @xcite .
such models tend to represent the activity of the network through a macroscopic variable , the population - averaged firing rate , that is generally assumed to be deterministic .
many analytical and numerical results and properties have been derived from these equations and related to cortical phenomena , for instance for the problem of spatio - temporal pattern formation in spatially extended models ( see e.g. @xcite ) .
this approach tends to implicitly make the assumption that correlations in the activity do not modify the behavior of the system in large populations .
however , the stochasticity of the system , and in particular the presence of correlations in the activity may considerably affect the dynamics in such nonlinear systems , and these can be even more important as the network size increases .
these models therefore make the assumption that averaging effects counterbalance the prominent noisy aspect of _ in vivo _ firing , a feature that can dramatically affect finite - sized network activity , and , thus , that the emergent behavior of a cortical column is deterministic .
however , increasingly many researchers now believe that the different intrinsic or extrinsic noise sources are part of the neuronal signal , and rather than a pure disturbing effect related to the intrinsically noisy biological substrate of the neural system , they suggest that noise conveys information that can be an important principle of brain function @xcite . at the level of a single cell , various studies have shown that the firing statistics are highly stochastic with probability distributions close to poisson distributions @xcite , and several computational studies confirmed the stochastic nature of single - cells firings @xcite . how variability at the single neuron level affects dynamics of cortical networks is less established so far .
theoretically , the interaction of a large number of neurons that fire stochastic spike trains can naturally produce correlations in the firing activity of the population considered .
for instance power - laws in the scaling of avalanche - size distributions have been studied both via models and experiments @xcite . in these regimes
the randomness plays a central role .
a different approach has been to study regimes where the activity is uncorrelated .
a number of computational studies on the integrate - and - fire neuron showed that under certain conditions neurons in large assemblies end up firing asynchronously , producing null correlations @xcite . in these regimes , the correlations in the firing activity decrease towards zero in the limit where the number of neurons tends to infinity .
the emergent global activity of the population in this limit is deterministic , and evolves according to a mean - field firing rate equation .
however , these states only exist in the limit where the number of neurons is infinite , and such asynchronous states do not necessarily exist in cortical areas or in computational models , and even when they exist , they are not necessarily stable .
this raises the question of how the finiteness of the number of neurons impacts the behavior of asynchronous states , and how to study non - asynchronous behaviors .
the study of finite - size effects in the case of asynchronous states is generally not reduced to the study of mean firing rates and can include higher order moments of firing activity @xcite . in order to go beyond asynchronous states and take into account the stochastic nature of the firing and how this activity scales as the network size increases , different approaches have been developed , such as the population density method and related approaches @xcite .
most of these approaches involve expansions in terms of the moments of the resulting random variable , and the moment hierarchy needs to be truncated which is a quite hard task that can raise a number of technical issues ( see e.g.@xcite ) .
a recent approach overcomes this truncation difficulty by considering the mean - field behavior as a stochastic process and therefore treating the rigorous limit equation of the interaction of many stochastic neurons as a fixed point equation in the space of stochastic processes , making use of the powerful tools of stochastic limit theorems and large deviation techniques ( see @xcite ) . in order to study the effect of the stochastic nature of the firing in large network
, many authors strived to introduce randomness in a tractable form .
the models we will be interested in are based on the definition of a markov chain governing the firing dynamics of the neurons in the network , where the transition probability satisfies a differential equation called the _
master equation_. seminal works of the application of such modeling for neuroscience date back to the early 90s with the work of ohira and cowan @xcite , and have been progressively developed by different authors and today constitute a model of choice @xcite .
the present manuscript is based on the analysis of two recently developed stochastic models of the customary wilson and cowan equations.the global resulting activity is then readily derived from the activity of all neurons in the network .
the two models of interest here were instantiated specifically in order to be compared to standard wilson and cowan rate equations , and provide tractable stochastic extensions of the purely deterministic standard neural - field models @xcite .
the two approaches differ in the choice of the transition probability and on the variables considered , choices mainly based on the fact that they address two different regimes of network activity : buice , cowan and chow address the highly correlated poisson - like activity and bressloff , the asynchronous states
. the two approaches also differ in the way the equations are treated : buice , cowan and collaborators in @xcite derive moment equations by carrying out a loop expansion of a path integral representation of their master equation whose truncation is justified under the assumption that the network operates in a poisson - like regime .
bressloff rigorously justifies his moment truncation by exhibiting an expansion in powers of a small parameter @xmath0 where @xmath1 corresponds to the size of the network .
these different points of view lead the authors to define different variables and they obtain closely related yet different equations on the variables they define , derive equations for moments from the master equation which they then simplify and truncate to finite order . the present paper is organized as follows : in the first section
we describe the models we use .
both simplified models introduced involve coupling between mean firing rate and correlations between firing times , with finite - size correlations . in the second section ,
we are interested in the non - zero correlation solutions of the mean - field equations , and , in particular , focus on the qualitative similarities and differences between the solutions of these stochastic equations compared to the customary wilson and cowan behaviors . in the third section
, we turn to study the finite - size effects , namely the qualitative differences between the solutions of mean - field limits and wilson and cowan system on one hand , and the behaviors of large but finite networks on the other hand .
these finite - size solutions are compared back to the initial markovian model and the differential equations they produced .
in this section , we introduce and describe the master equation formalism used by buice , cowan and collaborators @xcite ( hereafter bcc ) and bressloff @xcite and extend these approaches to take into account different populations .
both models have been previously derived , so we quickly present their derivation and the resulting equations in this section , as our main purpose is to derive from these equations our model of interest , the _ infinite - size model _ , and to mathematically analyze it .
details on the derivation of these equations and on their theoretical justification can be found in @xcite . throughout manuscript
, we will be interested in a network structured into @xmath2 homogeneous neural populations .
each population @xmath3 is composed of @xmath4 neurons , and the total number of neurons is denoted by @xmath1 ( i.e. @xmath5 ) .
we are interested in the behavior of the network when the number of neurons @xmath1 tends to infinity . in this limit
, we assume that the proportion of neurons belonging to each population are non - trivial , i.e. @xmath6 .
each neuron of population @xmath7 interacts with all the neurons of the network , with an efficiency denoted @xmath8 . these weights are assumed to scale as @xmath9 , so that a given activity at the level of population @xmath7 will produce a bounded activity at the level of population @xmath10 .
we therefore define the effective interconnection weights @xmath11 such that @xmath12 .
each population is assumed to receive an input firing rate @xmath13 considered deterministic and constant in the rest of the paper .
the wilson and cowan ( wc ) model is based on the assumption that the synaptic input current to each population is a function of the firing - rate of the pre - synaptic population , and that the contributions of the different populations are linearly summed to produce the post - synaptic population firing rate .
it assumes moreover that the population - averaged firing rates @xmath14 are deterministic function of time that are governed by the equations : @xmath15 where @xmath16 is a function transforming an incoming current into an output firing rate and typically has a sigmoidal shape .
@xmath17 are the relaxation rates corresponding to the natural inactivation of each population when they receive no input .
this purely deterministic approach , often used for spatially extended neural networks , has proved efficient to model a large number of cortical phenomena , but fails to reproduce the stochastic behaviors that can appear at the level of population activity .
we now describe the stochastic framework used by @xcite to extend this framework in order to take into account the stochastic nature of the firing . in the markovian approach yielding bcc and bressloff models ,
each neuron can be either quiescent or active ( active meaning in the process of firing an action potential ) , and the state of the network is then described by the variable @xmath18 where @xmath19 is the total number of active neurons in population @xmath7 .
this chain is modeled as a one - step discrete - time markov process , i.e. the only possible transitions of this chain are assumed to be @xmath20 for @xmath21 .
each active neuron of population @xmath7 returns to a quiescent state with constant probabilistic rate @xmath17 and each quiescent neuron become active with a state - dependent rate @xmath22 depending on the inputs the neuron receives .
the probability of the network to be in a state @xmath23 at time @xmath24 therefore satisfies a differential equation called the _ master equation _ : @xmath25\end{gathered}\ ] ] from this equation the authors derive equations on the moments of the markov chain .
we propose in appendix [ append : rodigtuck ] an alternative derivation of the moment equations based on a langevin approximation of the rescaled markov chain and on the use of rodriguez and tuckwell expansions @xcite . in the present paper
we are particularly interested in a variable that would correspond in the markov setting to the proportion of active neurons in each population : @xmath26 for @xmath27 , which is _ a priori _ a well - behaved bounded stochastic process that takes values in the interval @xmath28 $ ] .
having defined this variable , we will be in a position to rescale bcc and bressloff variables , and obtain from their derivation the equation governing this variable .
in doing so , as further illustrated below , we will not need to derive new moment equations from the master equation , but it will appear as a change of variable in the set of ordinary differential equations these authors previously obtained .
in this view , the rescaled mean firing rate is defined as the instantaneous averaged proportion of active neurons in each population @xmath29 , and rescaled correlations as the average value of the correlations of the proportion of active neurons in each population @xmath30 , the two first moments of the process @xmath31 . the equations for the moments involve averaged values of @xmath32 which , since the activation is not polynomial , involve all the moments of the variable @xmath23 , and therefore need to be truncated in order to obtain a closed set of equations .
the choice of the function @xmath33 is so far unspecified and will be defined in order to recover the wc equations in the mean - field uncorrelated limit . with the aim of specifically investigating ( uncorrelated ) asynchronous states , bressloff truncates the moment series expansion up to second order , and considers the evolution of two coupled variables : the mean firing - rate @xmath34 and @xmath35 a correlation variable differing from our variable in that it is scaled by @xmath36 .
after taylor expansion and a van - kampen system - size expansion , a specific choice of the activation function @xmath33 is made so that one ends up with the following set of ordinary differential equations : @xmath37\delta_{ij } - ( \alpha_i+\alpha_j)\,n_{ij } \\ &
\qquad+ \sum_{k } [ f_i'(s_i ) w_{ik}\,n_{kj } + f_j'(s_j ) w_{jk}\,n_{ki } ] \end{cases}\ ] ] where @xmath34 corresponds to the firing rate of population @xmath7 : @xmath38 and @xmath39 are defined as the correlations @xmath40 . in the original formula , the number of neurons of each population is assumed to be equal to @xmath41 .
this quantity is bounded by application of kurtz theorem @xcite , and therefore the term @xmath42 tends to zero as @xmath1 tends to infinity , implying that the equation of the mean becomes independent of the correlations @xmath39 .
since we are interested in the statistics of the proportions of active neurons in each population , we are in a position to derive from the original bressloff equations the system of equations governing our variables @xmath34 and @xmath43 formally , and obtain in a multi - population case with distinct population sizes : @xmath44\delta_{ij } - ( \alpha_i+\alpha_j ) c_{ij } \\ & \qquad+ \sum_{k } [ f_i'(s_i ) w_{ik}\ , c_{kj } + f_j'(s_j ) w_{jk}\ , c_{ki } ] \end{cases}\ ] ] in both equations and , we denoted by @xmath45 the total instantaneous current received by population @xmath7 at time @xmath24 : @xmath46 we note that the system , though derived from bressloff s master equation and formalism , is equivalent to the original bressloff model as long as the total number of neurons @xmath1 is finite , since it was derived through an invertible change of variable which is invertible as long as @xmath1 is finite , but not invertible in the limit @xmath47 .
buice , cowan and collaborators @xcite ( bcc case ) , interested in studying the poisson - like firing modes of the network , transform the moment equations to derive equations of _ normal ordered cumulants _ measuring the deviations of the moments of the variable @xmath23 from pure poisson statistics .
the first normal ordered cumulant is equal to @xmath48 , the second ordered cumulant to @xmath49 .
the initial approach of buice , cowan and collaborators is not a finite - size expansion _
per se _ in the general case , hence the parameter according to which the expansion is performed is no longer @xmath0 as in the case of bressloff s van kampen expansion , but consists of a multiple time scale expansion where the small parameter is the decay time of normal order cumulants ( which applies far from bifurcation points ) .
however , in the fully connected case we are interested in in the present manuscript with weights scaling as @xmath0 , the expansion provided buice , cowan and collaborators method can be reduced to a finite - size expansion , and can be compared to bressloff s case . in terms of these variables after some approximations and moment truncation to the second - ordered cumulant , they end up with the following set of ordinary differential equations ( this step involves an instantiation of the activation functions @xmath33 different from the bressloff case ) : @xmath50 we note that the rescaling does not affect the form of the equations obtained in @xcite .
these equations are valid under our assumptions that the synaptic weights @xmath8 scale as @xmath0 in a fully connected network .
we emphasize the fact that both expansions can be rigorously derived far away from bifurcation points . in the manuscript
, we will however be interested in the bifurcations of the dynamical systems given by and .
indeed , these bifurcations produce changes in the number and in the stability of attractors ( fixed point and limit cycle ) which will be visible in non - trivial parameter intervals , and therefore will affect the behaviors of the system far away from bifurcations and will give us indications of the number and stability of attractors in those parameter regions .
these models formally appear to very similar .
however , it is important to note that the these two approaches are different : the master equation corresponds to different transition rates ( different choices of the functions @xmath33 as stated ) , and the variables considered differ . the two equations and constitute our starting point for a mathematical exploration of the solution , and in the rest of this paper , we mathematically analyze these equations .
the reader nevertheless needs to keep in mind the differences between the two approaches for interpreting the results . the equations and are _ a priori _ different .
first of all , they do not deal with the same quantities : equation couples the mean firing - rate and the correlations while equation the mean firing - rate and the first order cumulant .
however , in the limit @xmath51 , the first - order cumulants , @xmath49 , are simply equal to the correlation @xmath43 . moreover , in the rescaled models we introduced , we observe that when the number of neurons tends to infinity , both bressloff and bcc models converge to the same equations .
these equations will be referred to as the _ infinite size model _ and are given by the equations : @xmath52 \end{cases}\ ] ] we recall that although the interpretation of the variable @xmath53 differs in the general case , it is the not the case of the infinite model : in bcc model @xmath54 are the second ordered cumulant , while in bressloff model @xmath55 is the correlation in the firing activity , and in the infinite - size limit , the normal ordered cumulant is equal to the correlation .
however , when @xmath1 becomes finite , the reader needs to bear in mind that in the finite - size unfolding of the infinite - size equations in bcc case , @xmath53 represents the deviation of spike statistics from a poisson process and in bressloff model the covariance of the firing activity . when letting @xmath1 be finite ,
the two models unfold the behavior of the system in a different fashion .
we also note that if the variable @xmath56 is equal to zero , then the mean - firing rates @xmath14 satisfy the wc equations . in that view ,
bressloff and bcc models are generalizations of wc system that take into account the correlations in the firing . in this paper , we will first be interested in studying the mean - field limit solutions of the infinite size equations , and second in unfolding these mean - field behaviors when the number of neurons is large but finite .
our study will particularly focus on two main aspects : ( i ) how the stochastic nature of the firings affect the observed behaviors at the macroscopic level through the interplay of firing activity and the correlations , and ( ii ) how the finiteness of the number of neurons in the network disturbs these behaviors .
we also note that in view of kurtz theorem @xcite , our variable in the infinite - size system vanishes .
we formally consider these equations independently of this restriction . in other words ,
the infinite - size system is obviously not equivalent to the wc system , but bcc and bressloff systems should be equivalent , in this limit , to wc system .
the study of the infinite - size system will allow us to find what we call _ correlation - induced behaviors _ , which are qualitative distinctions between the behaviors of the solutions of the infinite - size system and of wilson and cowan system .
if these behaviors are unfolded non - trivially into solutions of the finite - size equations , these will evidence additional behaviors of the system presented by the finite - size bressloff and bcc equations that do not correspond to solutions of wilson and cowan system .
we will address these questions in in section [ sec : finitesize ] . besides our theoretical analysis on the equations
, we will also go back to the initial markovian model and compare ( with monte carlo simulations ) the stochastic behaviors it presents to the solutions of wc , bcc and bressloff models .
we have shown that in the limit where number of neurons is infinite , both bressloff and bcc models , in our particular rescaling , converge to the infinite - size model given by equations where the mean - firing rate is coupled to the correlations in the firing activity . when the correlations are equal to zero , the equation for the mean - firing rate is uncoupled from the one for the correlations @xmath53 and is identical to the wc system .
the question we address in this section is how the interaction between the mean activity and the correlations modifies the behavior of the global activity compared to wc system given by equations .
we address the following two questions : ( i ) are the solutions of wilson and cowan equations also solutions of the full systems including correlatons , and if the answer is yes , are the stability properties of these solutions conserved ? and ( ii )
do the correlations yield other behaviors ? in equation , it is easy to see that zero correlation : @xmath57 is always a fixed point of the correlation equations no matter what mean firing rates @xmath58 are . for zero correlations , the equations for the mean
firing rate @xmath59 are reduced to the classical wc equations . therefore , wc solutions define solutions for the infinite - size system with zero correlations .
let us now study the stability of the solutions defined by these uncorrelated wc behaviors in the infinite size system .
the infinite - size system involves a standard @xmath2-dimensional differential equation on the mean firing rates @xmath58 coupled to a matrix differential equation . in order to use customary approaches for dynamical systems ,
we transform the correlation matrix into a @xmath60 dimensional vector and express the equations on @xmath53 in this new format using kronecker products from linear algebra . to this end
, we start by defining the function @xmath61 transforming a @xmath62 matrix into a @xmath60-dimensional column vector , as defined in @xcite : @xmath63^t \end{cases}\ ] ] let us now denote by @xmath64 the kronecker product defined for @xmath65 and @xmath66 as the @xmath67 matrix : @xmath68 some definitions and identities in the field of kronecker products are reviewed in appendix [ append : kronecker ] , and a more complete discussion can be found in @xcite and references therein .
we recall here for the sake of completeness some well - known relationship that will be useful .
let @xmath69 , let @xmath70 be the @xmath71 identity matrix and @xmath72 or @xmath73 denote the standard matrix product .
we have : @xmath74 the relationship @xmath75 is called kronecker sum . in this framework , we have the following identity : [ lem : ckron ] let @xmath76 be the column vectorization of the matrix @xmath56 , and @xmath77 the matrix defined by @xmath78 where @xmath79 is the diagonal matrix with @xmath80 element @xmath81 and @xmath82 the matrix of general element @xmath83 .
the variable @xmath84 satisfies the differential equation in @xmath85 : @xmath86 the differential equation governing the evolution of the coefficients of the correlation matrix @xmath56 of the infinite size system can be easily reordered into : @xmath87 which , through straightforward linear algebra manipulations , can be written as : @xmath88 the linear operator @xmath89 for @xmath90 can be written in terms of kronecker product of matrix on the vectorized version @xmath84 of the matrix @xmath56 using the relationship given in and we have : @xmath91 now that we have defined a convenient framework to study the infinite size equations , we are in position to study the stability of fixed points and limit cycles of the wc system as solutions with zero correlations of the infinite size system .
we start by addressing the problem in the one population model ( @xmath92 ) , which allows straightforward calculations and will be helpful in our specific study of the one population infinite - size equation .
[ prop : onepop ] in the one population case , any fixed point of wilson and cowan equation is a zero - correlation fixed point of the infinite - size system , and with the same stability properties . in the one population case
, equations constitute a planar dynamical system that reads : @xmath93 where @xmath94 .
as stated , the solution @xmath95 is a fixed point of the correlation equation , and in that case , the activity satisfies the one - population wc equation : @xmath96 let @xmath97 be a fixed point of this system .
the jacobian matrix of the infinite - size system at the fixed point @xmath98 reads : @xmath99 as we can obviously see , the eigenvalues of this equilibrium are @xmath100 and @xmath101 , and therefore the stability of this fixed point is the same as the stability of the related fixed point in wc s system .
we therefore conclude that any stable solution of the one dimensional wc equation provides a zero correlation ( @xmath95 ) stable fixed point of the infinite - size system , and we observe that no stable solution of wc system is destabilized by the presence of correlations , and no unstable solution of the wilson and cowan equation will be stabilized .
therefore , zero - correlation fixed points of the infinite size system exists if and only if wilson and cowan equation has a fixed point , and the related fixed point of the infinite - size system has the same stability as wc s fixed point .
we now turn to demonstrating the related properties in arbitrary dimensions @xmath2 , for fixed points and cycles .
[ thm : wilsonbuicecowan ] solutions of the wc system provide solutions of the infinite - size system with zero correlations @xmath95 .
the stability of fixed points and limit cycles of the infinite size system depend on the stability of the solution for the wc system as follows : + _ i ) _ a fixed point of the infinite - size equation with null correlations is stable if and only if the value of related mean - firing rate is a stable fixed point of wc system ; + _ ii ) _ a cycle of the infinite size system with zero correlations @xmath102 is exponentially unstable if and only if @xmath58 is an unstable cycle of wc system .
stable cycles @xmath58 of wc s system define cycles of the infinite - size system with zero correlations with neutral linear stability .
the infinite size system does not present any stable cycle with null correlations . in order to prove the theorem ,
it is convenient to introduce the vector fields @xmath103 , @xmath104 and @xmath105 that govern the dynamics of the infinite size system : @xmath106 we have @xmath107^t$ ] , @xmath108 where @xmath109 is defined in lemma [ lem : ckron ] , and @xmath110 is the @xmath111 matrix such that : @xmath112^t\ ] ] let @xmath113 be a solution of the infinite - size system .
the jacobian matrix of the system evaluated on this solution has the form : @xmath114 where @xmath115 for a multidimensional vector @xmath116 and a multidimensional function @xmath117 denotes the differential of @xmath117 with respect to @xmath116 . since @xmath118 is exactly the vector field associated with wc system
, @xmath119 is an @xmath71 block corresponding to jacobian matrix of wc system at the point @xmath59 , which is equal to the matrix @xmath120 introduced in lemma [ lem : ckron ] . now that the jacobian matrix is identified , we prove the assertions of the theorem : + _ i ) _ let us consider a fixed point @xmath121 of the infinite size system .
necessarily , @xmath59 is a fixed point of wilson and cowan equation .
the stability of this fixed point depends on the spectrum of the jacobian matrix @xmath122 of the system at this point . since the jacobian matrix of the infinite - size system is block diagonal , its spectrum is therefore composed of the eigenvalues each diagonal block @xmath123 and @xmath124 .
the first block , @xmath120 , has exactly the eigenvalues of wilson and cowan system at its fixed point @xmath59 .
these eigenvalues are denoted @xmath125 .
the second diagonal block of the jacobian matrix @xmath126 is equal to the kronecker sum @xmath127 by application of lemma [ lem : ckron ] .
the eigenvalues of a kronecker sum are known to be all possible pairwise sums of all the eigenvalues of @xmath120 , viz .
@xmath128 ( see an elementary proof of this fact in proposition [ prop : kroneigen ] of appendix [ append : kronecker ] ) , hence the spectrum of the jacobian matrix of the infinite size system is composed of the eigenvalues : @xmath129 these eigenvalues depend in a simple fashion on the eigenvalues of the wc jacobian matrix at the fixed point @xmath59 , and it is easy to show that the fixed point @xmath130 in the the infinite system is : * exponentially stable if and only if the @xmath59 is an exponentially stable fixed point of wc system , i.e. if and only if all the eigenvalues @xmath131 have a strictly negative real part . * exponentially unstable if and only if the @xmath59 is an exponentially unstable fixed point of wc system , i.e. if and only if there exists an eigenvalue @xmath131 with strictly positive real part . *
neutrally stable if and only if @xmath59 is neutrally stable for wc system , i.e. if and only if all eigenvalues have non - positive real part and at least one eigenvalue have a null real part . in summary ,
the stability of the zero - correlation fixed point @xmath121 is exactly the same as the stability of @xmath59 as a fixed point of wc system . _
ii ) _ let us now address the case of cycles will null correlations . to this end , let us consider that @xmath132 is a periodic orbit of the infinite - size system ( i.e. @xmath58 is a periodic orbit of a wc system ) .
let us denote by @xmath133 the period of this cycle .
we show that the null correlations @xmath134 is an exponentially unstable solution if the cycle @xmath58 is exponentially unstable as a solution of wc system , and neutrally stable if the cycle @xmath58 is exponentially stable or neutrally stable as a solution of wilson and cowan system , which proves the theorem . given that the cycle @xmath58 is known , the correlations @xmath56 satisfy a linear equation : @xmath135 and the time - dependent matrix @xmath136 is @xmath137-periodic .
a basic result of floquet theory implies that the stability of the solution @xmath138 depends on the eigenvalues of the resolvant of the system at time @xmath137 , namely the matrix @xmath139 in @xmath140 where @xmath141 is defined by @xmath142 more precisely , the null fixed point is exponentially stable if and only if all eigenvalues of @xmath139 ( the _ multipliers _ ) are of modulus strictly smaller than @xmath143 , neutrally stable if all have modulus smaller or than equal to @xmath143 with at least one multiplier with modulus equal to @xmath143 , and exponentially unstable if there exists a multiplier with modulus larger than @xmath143 .
we therefore need to characterize the eigenvalues of @xmath139 in order to conclude on the stability of the solution in the infinite - size system . to this end , let us introduce @xmath144 the resolvent of wc system , i.e. the unique solution of the fundamental equation : @xmath145 we prove in theorem [ theo : psikron ] of appendix [ append : kronecker ] that @xmath146 .
let us denote by @xmath147 the eigenvalues of @xmath148 . because @xmath139 is the kronecker square of @xmath148 , its eigenvalues are are all pairwise products of the eigenvalues of @xmath148 , i.e. @xmath149 , as shown in proposition [ prop : kroneigen ] of appendix [ append : kronecker ] .
if the cycle @xmath58 is exponentially unstable as a solution of wc system , then floquet theory implies that there necessarily exists a multiplier of @xmath148 , @xmath150 , such that @xmath151 .
therefore the resolvant @xmath139 has a multiplier equal to @xmath152 which has modulus equal to @xmath153 and therefore this cycle is exponentially unstable as a solution of the infinite - size system . if @xmath58 is exponentially stable or neutrally stable as a solution of wc system , then a basic result of floquet theory states that the resolvant if the linearized system evaluated on the cycle at time @xmath137 , @xmath148 , has at least a multiplier equal to @xmath143 and all other multipliers with multipliers of modulus smaller than @xmath143 .
we now provide an elementary proof of the existence of the @xmath143 multiplier : let @xmath154 .
then since @xmath58 is @xmath137-periodic , so is @xmath155 , and furthermore we have : @xmath156 moreover , since @xmath58 is not a fixed point , @xmath157 .
we have shown @xmath155 is solution of the linearized equation and hence using its @xmath137-periodicity , we have @xmath158 , which proves that @xmath159 is eigenvector of @xmath148 associated with the eigenvalue @xmath143 .
therefore , the resolvent @xmath139 necessarily has an eigenvalue equal to one ( associated with the eigenvector @xmath160 ) , which implies that the null fixed point has a neutral linear stability as a solution of the infinite - size system .
we directly conclude from this result that the infinite size system does not features any exponentially stable cycle , which ends the proof .
we conclude that the solutions of wc system always provide solutions of the infinite - size system , and that all fixed points with null correlations in the infinite - size system are fixed points of wc system , with the same stability properties .
however , cycles of wc system lose exponential stability in the infinite - size system .
note that this does not necessarily implies that these cycles become unstable , and a nonlinear stability analysis is necessary .
but it implies that the transient phase of convergence towards the cycle or of repulsion from the cycle will be not be exponential . in the previous section
, we just proved that all the fixed points of the wc system are solutions of uncorrelated activity in the infinite size system .
we now investigate the existence of new solutions induced by the presence of non - null correlations in the infinite system , that are therefore not solutions of wc system .
such solutions will be referred to as _ correlation - induced behaviors_. the general resolution of this problem is quite hard .
for this reason , we provide a rigorous analysis of the one population case , and treat two - population case through numerical analysis and simulations .
we have seen in the one population case that any fixed point of wc system was solution of the infinite size system with zero correlations .
moreover , there is no possible cycle in the wilson and cowan equations in one dimension .
we investigate now the existence of cycles or additional fixed points in the infinite size system with one population .
we recall that acceptable solutions in the one population case necessarily have non - negative firing rate @xmath59 and @xmath53 . indeed , in the case of the infinite - size model arising in the limit of bressloff s rescaled model , the correlation @xmath53 needs to be positive as a limit of positive quantities ( and as the covariance of the markov process ) , and in bcc case @xmath53 is equal to the limit of the first order cumulant which differs from the correlation up to the coefficient @xmath161 that vanishes in the infinite - size limit . [
thm : onepopnosolution ] in the one population infinite size system , there does not exist any acceptable stable fixed point with strictly positive correlation @xmath53 .
+ let us assume that @xmath162 .
the fixed point equation of the infinite - size system : @xmath163 the first equation of the system is independent of @xmath53 and fixes possible values of @xmath59 . assuming that this equation has a solution @xmath97 and denoting @xmath164 , the jacobian matrix of the system at @xmath165 reads , using the property that @xmath97 solves the first equation of : @xmath166 such values of @xmath97 necessarily exist for some parameters , and
since the function @xmath117 is assumed to be sigmoidal , it can have multiple solutions .
moreover , because of the particular shape of the sigmoidal function @xmath117 , the second derivative of @xmath117 can vanish at a point corresponding to the inflection point of the sigmoid . in a very particular case ( for precisely tuned values of the parameters ) , one can hence have @xmath167 . in that case ,
fixed points only exist if @xmath168 , if this condition is satisfied , any value of the correlation @xmath53 provides a fixed point of the system .
this fixed point is never exponentially stable , since in that case the jacobian matrix has a zero eigenvalue ( obvious when using the expression of the jacobian matrix and using the fact that @xmath169 ) . in the general case where @xmath170 , the system has a fixed point with non - null correlations @xmath171 we therefore need to check whether if this fixed point is stable and acceptable ( i.e. @xmath172 ) .
the jacobian matrix at this point has the expression given by .
its determinant is equal to @xmath173 and has an opposite sign to @xmath174 .
therefore , for acceptable solutions with @xmath175 , the determinant of the jacobian matrix is strictly negative . we conclude that any fixed point of the infinite - size system with @xmath176 are saddle fixed points .
we conclude from this theorem that the system does not features any acceptable stable fixed point with non - zero correlations .
we can now easily conclude that there are no acceptable cycles .
consider , first , a limit cycle such that @xmath53 is strictly positive .
such a limit cycle must surround one or more fixed points whose poincare index must sum to + 1 @xcite .
however , every fixed point in the postive orthant is a saddle point and saddle points always have indices of -1 .
thus , no limit cycle can have a strictly positive @xmath177 any limit cycle must therefore be tangent to the @xmath95 line and the point of tangency can not contain a fixed point , so , again , the summed index of points in the limit cycle can not be + 1 . in conclusion , there are no acceptable limit cycles and all acceptable fixed points are saddles .
thus , the only stable behavior are the uncorrelated stationary states of the scalar wc equation
. this situation will be quite different in the multi - population case , as we now develop .
we now turn to study multi - population networks with the aim of identifying solutions of the infinite - size system that qualitatively differ from wc equation .
the situation in the multidimensional case will be quite different , since we have proven in theorem [ thm : wilsonbuicecowan ] that cycles of wc system lost exponential stability in the infinite - size system , allowing the appearance of distinct transient and/or asymptotic behaviors . to identify such correlation - induced behavior we numerically study two particular two population networks .
the first model ( model i ) is a two - population network including an excitatory and an inhibitory population known to produce oscillations , and the second example ( model ii ) builds on a famous model of binocular rivalry presenting bistability of fixed points .
[ [ sec : negfeedback ] ] perturbation of a wc cycle + + + + + + + + + + + + + + + + + + + + + + + + + + we proved in theorem [ thm : wilsonbuicecowan ] that any cycle of wc model lost exponential stability in the infinite - size system .
this loss of exponential stability does not necessarily mean that the cycle loses stability , and this property depends on the higher order terms of the nonlinear equation , but even if the cycle keeps stability , the transient convergence phase towards the cycle will be dramatically slowed . in this section ,
we choose to study a two - populations network featuring an excitatory population interconnected with an inhibitory one . because of the symmetry @xmath178 , the system is of dimension @xmath179 ( the two mean firing - rates and three correlation variables ) . to fix ideas ,
we denote by @xmath143 ( resp .
@xmath180 ) the excitatory ( resp . inhibitory ) populations .
we choose the same activation function for both populations equal to @xmath181 and the same inactivation rate @xmath182 , and are interconnected through the following connectivity matrix : @xmath183 each population receives different input currents @xmath184 and @xmath185 .
these parameters define a model called model i in the sequel . for these functions and parameters ,
the wc system features a cycle . when taking the input to the excitatory population @xmath186 small
, the system features a single stable fixed point , that loses stability as the input @xmath186 increases through a supercritical hopf bifurcation generating a family of stable limit cycles , that disappear through a homoclinic bifurcation ( see figure [ fig : cyclesmeanfield2pops](a ) ) , branching onto a high - state stable fixed point .
let us now analyze how this picture is modified by the presence of correlations .
first of all , from theorem [ thm : wilsonbuicecowan ] , it is clear that the parameter point where wc system undergoes the supercritical hopf bifurcation will be a very degenerate bifurcation point for the infinite - size system .
indeed , the hopf bifurcation is characterized by the presence of a pair of purely imaginay complex conjugate eigenvalues .
then lemma [ lem : ckron ] and the fact that the kronecker sum of two matrixes has all pairwise sums of eigenvalues of the matrixes in the sum ( see appendix [ append : kronecker ] ) directly implies that the jacobian matrix of the full system has the eigenvalues @xmath187 and the eigenvalue @xmath188 is of multiplicity one .
therefore at this point , the jacobian matrix of the infinite size system has its @xmath179 eigenvalues having a null real part , and thus at this point the system is very degenerate and in particular the bifurcation is not a generic hopf bifurcation .
it is quite difficult numerically and analytically to study the solutions emerging from this bifurcation point , and many solutions can appear at this point , including cycles and fixed points .
similarly , at the two saddle - node bifurcations of wc system , since the jacobian matrix of the infinite - size system has all the pairwise sums of the eigenvalues , it will have two null eigenvalues and therefore a degenerate bifurcation .
the behavior of the system around these degenerate points will appear more clearly in the finite - size unfolding of these degenerate bifurcation points , and we will observe in particular that the hopf bifurcation point corresponds to the merging of two generic hopf bifurcations and the saddle node bifurcations to the merging of two generic saddle - node bifurcations , see section [ sec : negfeedback ] and figure fig . [
fig : codim2bc2pops ] . in this section
we do not address here the problem of classifying all behaviors of the system around the bifurcation ( since as already mentioned , the finite - size unfolding will help answering these questions ) , and restrict the analysis to the question of whether the wc cycle keeps a nonlinear stability in the new system . to answer this question ,
we numerically perturb wc s cycle by adding a small initial condition in the correlations .
interestingly , we observe that a new cycle appears , that is attractive in a certain range of parameters ( see figure [ fig : cyclesmeanfield2pops ] )
. the wc cycle does not completely loses stability and still appears as the only behavior of the system in a certain range of input values ( e.g. @xmath184 as plotted in figure [ fig : cyclesmeanfield2pops](b ) ) .
but for smaller values of the parameters , it loses its nonlinear stability and the additional cycle appears ( e.g. @xmath189 and figure [ fig : cyclesmeanfield2pops](c ) and ( d ) ) .
this new cycle is totally different in its shape , has a period close to half the period of the cycle corresponding to wc system , and has non - zero correlations that vary periodically at the same frequency .
a continuation of this branch of cycles shows that it is stable in a significant range of parameters ( figure [ fig : cyclesmeanfield2pops](e ) ) .
it loses stability when @xmath190 decreases through a period - doubling bifurcation , and as @xmath190 increases through a neimark - sacker bifurcation .
this branch of limit cycles emerges from the very degenerate point corresponding to the hopf bifurcation of wc system , as we expected .
it disappears at a point corresponding to a subcritical hopf bifurcation with correlations ( on a branch of unstable fixed points that is not plotted in the diagram but that will be further investigated in section [ sec : twopopsfinite ] ) .
+ we conclude with a further observation on transient behavior .
we have seen that wc s cycle was neutrally stable , but in some regions of parameters , it has nonlinear stability . in the region of parameters where the wc cycle is nonlinearly stable ( white region in figure [ fig : cyclesmeanfield2pops](d ) ) , the transient phase of convergence towards this cycle takes a long time . in the green region where it is nonlinearly unstable
, the neutral stability of the cycle produce some very strange transient behaviors where the activity of the cortical column is trapped around the neutrally stable wc system for long times , before leaving the neighborhood of this solution and converging towards the only stable solution which is the new cycle .
[ [ sec : second2pop ] ] correlation - induced oscillations + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + correlations can have even more dramatic effects than modifying a periodic orbit as shown in model i , and in this section we describe a case where a periodic orbit arises for parameter values where wilson and cowan system only fixed - points .
to this end , we choose a two - population network known to have bistability .
this model consists of two - populations with inhibitory interactions and no self - interaction .
the connectivity matrix in that case reads : @xmath191 the inactivation constants @xmath17 are assumed constant equal to @xmath143 , and the inputs to the two populations are distinct and denoted @xmath186 and @xmath190 ( hence the system is not fully symmetrical ) .
the firing rate function are also identical and we make the usual choice @xmath192 for @xmath193 .
these parameters define a model called model ii .
we break the symmetries between the two populations by considering that each population can receive different input , we freeze the input @xmath190 and let @xmath186 vary . in that case , the standard wc system presents a bistable behavior between two stable fixed points as we illustrate in figure [ fig : binoc2pops ] .
the infinite size system driven by equations with these parameters has much richer dynamics , as illustrated in figure [ fig : binoc2pops ] .
we first set @xmath194 and study the bifurcation diagram as @xmath186 varies .
we observe that the infinite - size system features a stable limit cycle , which is a qualitatively nontrivial effect of the correlations in the mean - field limit .
all fixed - points behaviors in wc system are preserved , as predicted by theorem [ thm : wilsonbuicecowan ] ( black curves in figure [ fig : binoc2pops](b ) ) , but new fixed points with non - zero correlations appear ( blue lines in the diagram ) .
along one of the newly generated branch of fixed point , a supercritical hopf bifurcation appears , generating a family of stable limit cycles ( red circles in fig . [ fig : binoc2pops](b ) ) that further undergoes few consecutive period doubling bifurcations .
we note that the new branch of fixed points intersects the branch of fixed points of wilson and cowan system precisely at the saddle - node bifurcations of wilson and cowan system .
this fact is compatible with a partial result of the proof of theorem [ thm : wilsonbuicecowan ] : at the saddle - node bifurcation , instead of a single eigenvalue equalling zero , two eigenvalues are simultaneously equal to zero , and the saddle - node bifurcations of wc system appear as transcritical bifurcations in the infinite size system .
note that in that case , the wc behavior is nowhere the unique stable behavior of the system , and for any value of the input parameter @xmath186 , the infinite system will present attractive behaviors that are different from the wc behaviors .
in particular , it has stable fixed points with non - zero correlations and non - zero correlation oscillations .
however , we note that though these behaviors exist in the dynamics , the newly discovered fixed points do not constitute acceptable solutions since they are characterized by a negative firing - rate @xmath195 and a non - positive definite correlation matrix . however , the cycles constitute acceptable solutions that might have counterparts in the markov system .
+ we further explore this model by choosing a large value of the input parameter to the second population , @xmath196 .
for this level of input , the bifurcation diagram of wc system presents exactly the same features as in the case @xmath194 , namely two saddle - node bifurcations and a region of bistability of fixed points .
the infinite - size system , similarly to the case @xmath194 , also has additional branches of fixed points with non - zero correlations that are connected to wc bifurcation diagram at the saddle - node bifurcation points .
these saddle - nodes are degenerate and non - generic in the infinite size system from lemma [ lem : ckron ] and become transcritical bifurcations .
these additional non - zero correlation fixed points can be stable or unstable , depending on the parameter , and add additional possible behaviors . from one of these branches ,
the system undergoes a supercritical hopf bifurcation and a family of stable limit cycles appears .
this branch of limit cycles undergoes two period - doubling bifurcations , torus bifurcations and two fold of limit cycles , yielding additional oscillatory behaviors to the purely fixed - point structure of wc system . from all these observations on both models ,
we conclude that the infinite - size system , besides featuring the same solutions as wilson and cowan system as a subset of solutions with zero correlations , also has additional behaviors that qualitatively differ from the wilson and cowan system .
the mean - field limit including correlations is therefore a more complex system than wc system , in which correlations of the firing activity interact in a non - trivial way with the mean firing rate yielding complex behaviors .
these qualitative correlation - induced behaviors can therefore be a good indicator of which model better accounts for the behavior of cortical columns and cortical areas if these behaviors hold in the finite - size systems which are equivalent to the moment equations of the initial markov chain .
we now turn to investigating finite - size effects arising from the finiteness of the number of neurons in a network .
the bifurcation diagram displayed by the infinite - size system studied in the previous section is composed of some very degenerate points , and displays behaviors the wc system does not display .
the question that naturally arises at this point is to understand whether the infinite - size model or wc model really capture the qualitative behavior of large networks in the markovian framework through the bcc and bressloff finite - size equations . in this section
we first study how the infinite - size bifurcation diagram is unfolded in finite - size networks , before simulating the markov process and comparing its global behaviors to the behaviors of the finite - size systems .
we observed that the finite - size bcc and bressloff equations appeared as a perturbation of the infinite - size system , in which the size of the network appears as a parameter of the equations . in all of this study , instead of considering discrete population sizes
, we will consider the inverse of the total population size @xmath197 as a continuous parameter .
non - trivial distinctions between finite - size networks and infinite size equation will essentially appear at the bifurcation points .
indeed , since both the rescaled bressloff and bcc models vector fields are continuous with respect to the parameter @xmath23 , when the infinite - size system ( @xmath198 ) exhibits an hyperbolic fixed point , this fixed point will have a counterpart with the same stability for finite , sufficiently large networks , by application of the implicit functions theorem .
this is also the case of hyperbolic fixed points of wc system , since these are hyperbolic fixed points of the infinite - size system with zero correlations . in order to study similarities and qualitative differences between finite - size networks and the infinite size limit
, we will therefore focus on the bifurcations of both systems and numerically study the codimension two bifurcations in the one and two populations cases , with respect to the size of the network and to another parameter of interest ( for instance an input parameter ) .
we start by investigating a single - population case . in the mean - field limit
, we have shown that the infinite size system essentially exhibits the same behaviors as the wc model in terms of fixed points and limit cycles . but is this still the case where the number of neurons is finite ? at the bifurcations of the wc system , singular phenomena can occur .
consider for instance a saddle - node bifurcation point of wc system . at this point ,
the infinite size system will present a double zero eigenvalue , and the unfolding of this bifurcation can lead to different cases . as we show in this section , it can either unfold into a generic saddle - node bifurcation as we show in the first section , or it can generate an oscillatory behavior through a limit cycle , creating in the finite - size system oscillations which do not exist in the wc or in the infinite size one population equations .
we start by considering bcc model with a strictly positive voltage to rate function @xmath117 , and focus on acceptable solutions that have positive firing rates and correlations . in that case , we numerically show that there is no qualitative difference between the stable solutions of the finite - size bcc network on one hand , and the infinite size system and wc system on the other hand ( we showed that both models present the same qualitative behaviors in the one population case ) .
we address this problem numerically , with an inactivation constant @xmath182 ( without loss of generality , since modifying this value only amounts rescaling time ) , @xmath199 , and letting both the size of the network @xmath23 and the input parameter @xmath200 free .
the activity to rate function is chosen to be @xmath181 . with these parameters ,
wc system presents two saddle - node bifurcations when the input parameter @xmath200 varies , as shown in figure [ sfig : wc1pop ] .
the system therefore presents an interval of values of the parameter @xmath200 where the system presents two stable fixed points and an unstable fixed point , and outside this interval , the system has a unique fixed point which is stable .
similarly to what we showed in the previous section , the infinite - size network presents exactly the same features , and no cycle .
we now study bcc finite - size equations .
we set the size of the network to be @xmath201 .
bcc finite size equations present the same qualitative behaviors , and the same type of bifurcations : the system undergoes two saddle - node bifurcations as the input parameter value @xmath200 varies , and a bounded interval of input values @xmath200 where the system presents a bistable behavior .
the only qualitative distinction between finite - size networks and infinite size networks is the existence of a two unstable fixed points in the bistable zone and the fact that the system presents two distinct branches of fixed points , but these differences only affect unstable fixed points and do not produce generic qualitative differences in the behaviors .
the way this bifurcation diagram is transformed into the bifurcation diagram of the infinite system is very simple .
the two distinct branches of fixed point get increasingly close , and in the limit @xmath198 , the two branches are superimposed . for finite - size networks ,
we observe that the value of the correlations of the stable fixed points was non - zero ( see fig .
[ fig:1popcodim2bc ] ) .
these values were nevertheless quite small , and decrease towards zero as the size of the network as the number of neurons increases . when continuing the two identified saddle - node bifurcations as the network size increases , we observe that the bifurcation curve persists until the limit @xmath198 with no singularity . at the limit
however , the non - zero eigenvalue of the jacobian matrix at the saddle - node bifurcation point tangentially reaches the value @xmath188 yielding the singularity predicted in proposition [ prop : onepop ] .
+ + in this case , we therefore illustrated the fact that bcc system can smoothly unfold the wc system by introducing correlations .
these correlations vanish as the size of the population increases , yielding a purely poisson firing with firing rates given by the solution of wc equations . for the same parameters and functions ,
it is easy to show , following the same steps , that bressloff rescaled model also smoothly unfolds the wc system and that the finite - size system presents exactly the same type of bifurcations with two separated branches of fixed points each of which collapse in the limit where the number of neurons is infinite ( not shown ) , and the value of the correlations of the fixed points decreases towards zero when the number of neurons increases .
however in this case , one has to bear on mind that the interpretation of the result slightly differs : the limit behavior is now seen as a purely asynchronous state where all the neurons of the population fire independently in the limit @xmath51 .
these behaviors will closely match the markovian model s behavior as shown in section [ sec : onepop ] . in these cases
, wc system conveniently summarizes the behavior of large networks , and provides a simple model to study large networks .
however , there exist models for which this conclusion does not hold , and where the infinite - size system s behavior unfolds into more complex behaviors .
we now turn to study the possible singular unfolding of the infinite size system in one population . to this purpose
, we will be specifically interested in studying oscillatory behaviors in the finite - size systems that are not present in the limit where the number of neurons tends to infinity .
we will call these qualitative distinctions between finite - size and infinite - size systems singular unfolding . to this purpose
, we choose a framework where fixed points are easily described , and along the fixed point search for sufficient conditions to get a supercritical hopf bifurcation , yielding oscillatory behaviors in single population networks that are neither present in the infinite - size system nor in the wc system . to fix ideas , similarly to the previous case treated , we consider bcc one population model .
we consider here a non - physiological cases by allowing the firing rate variable to be negative .
we choose an activation function @xmath117 such that @xmath202 for some fixed @xmath200 , which can always be done with an unspecified sigmoid @xmath203 through the modification @xmath204 .
this quite unrealistic case will allow studying the qualitative differences that can appear between the finite and infinite size systems . however , the fixed point @xmath188 is quite particular , and we will see that this phenomenon is not generalized in a simple way when considering positive activation functions @xmath117 and positive firing rates ( the affine transform @xmath205 does not have a trivial effect on the nonlinear system ) .
the bcc single population model reads : @xmath206 therefore , under the assumptions made , it is easy to see that the null solution @xmath207 is always solution of the system .
the jacobian matrix @xmath122 at this point reads : @xmath208 where we denoted @xmath209 for @xmath210 .
the trace of the jacobian matrix at this point is equal to @xmath211 and the determinant is equal to @xmath212 .
the trace vanishes at the null fixed point for @xmath213 such that @xmath214 and therefore there exist a set of parameters for which the trace vanishes . in order to ensure that this null trace corresponds to the presence of a pair of purely imaginary eigenvalues , we further need to ensure that the determinant is strictly positive .
when the trace vanishes , the determinant simply reads @xmath215 .
since @xmath117 is a sigmoidal function , @xmath216 is always positive , and therefore the determinant is positive as soon as @xmath217 , which can be ensured for some specific firing rate functions . in that case
, we will denote @xmath218 .
therefore , under the assumptions that there exists @xmath200 such that @xmath202 and @xmath219 , there exist parameters @xmath220 and @xmath221 such that the jacobian matrix of the system at the null solution has a pair of purely imaginary eigenvalue .
let us now check that this point corresponds to a hopf bifurcation when varying the parameter @xmath220 .
to this purpose , we check transversality and genericity conditions of the hopf bifurcation at this point , as given for instance in @xcite . these calculations are provided in appendix [ append : calculations ] .
the transversality condition is easily checked .
it consists in showing that the differential real part of the eigenvalues of the jacobian matrix at this point , with respect to the parameter @xmath220 , does not vanishes .
let us denote by @xmath222 the real part of the eigenvalues .
since we are in a planar system , we have : @xmath223 therefore the transversality condition of the hopf bifurcation is satisfied at this point . in order to fully state that the system undergoes a hopf bifurcation ,
we need to show that the first lyapunov exponent of the system at this point does not vanish .
this lyapunov exponent , after some tedious calculations using the formula given in @xcite ( see appendix [ append : calculations ] ) , is shown to be of the same sign as : @xmath224.\ ] ] the sign of this expression is governed by the values of the differentials of @xmath117 at the point @xmath200 .
if the expression in the bracket is negative , then the hopf bifurcation is supercritical and the system has stable oscillations .
let us now investigate the dependence of this cycle on the network size .
we have shown that @xmath218 . therefore ,
if this cycle appears for some finite network , it will lose stability as the size of the population increases .
the period of the generated cycle is proportional to the inverse of the square root of @xmath1 , and therefore slowly tends to zero as the network size increases ( note that the cycle will lose stability as the network size increase , since for sufficiently large populations , the hopf bifurcation will be subcritical ) .
now that we identified simple conditions for the hopf bifurcation to take place , we exhibit a network that indeed has this bifurcation .
we choose for instance as activation function an hyperbolic tangent function @xmath225 with @xmath226 . in that case
, we have : @xmath227 therefore this function satisfies the assumptions done in the previous paragraph .
moreover , since @xmath228 and @xmath229 ( for @xmath1 small enough ) , the first lyapunov coefficient is negative , and therefore the system undergoes a supercitical hopf bifurcation and presents oscillations , as illustrated in figure [ fig : oscillations1pop ] .
therefore , we exhibited a case where the finite - size equations have important qualitative differences from the infinite - size equations . however this system involved non - physiological values of the parameters , in particular a non - positive activation function @xmath117 .
therefore , this behavior will never appear in the markov process that yielded bcc equations .
the question that then arises is whether such behaviors appear in plausible settings . in order to answer this question and search for a system with positive sigmoidal functions that indeed present oscillations in one - population networks , we continued the hopf bifurcation while continuously transforming the non - positive sigmoidal transform into a positive sigmoid .
for instance , we consider @xmath117 a non - positive sigmoidal function for which the hopf bifurcation exhibited here appears , and define @xmath230 a transformation of @xmath117 such that @xmath231 and @xmath232 is a positive sigmoid .
this homotopic transform smoothly maps the vector field presenting an attractive cycle onto a vector field with a positive activation function . since we have proved the genericity and transversality conditions on the hopf bifurcation identified , we are in position to continue it as the parameter @xmath233 varies together with @xmath220 in a codimension two bifurcation diagram ( by application of the inverse function theorem ) .
we observe in that case that the hopf bifurcation can not be continued when varying the homotopy parameter until reaching a positive sigmoid ( see figure [ fig : oscillations1pop](b ) for the example provided above ) .
we have tried numerous other sets of parameters and have never been able to continue periodic orbits to a regime where there is a non - negative activation function . therefore , though the one - population bcc equations can display oscillations for non biologically relevant parameters , we conjecture that there is no acceptable oscillatory solution for positive activation functions .
we have seen in the previous section that the one - population finite - size bcc model was accurately approximated by the infinite - size system ( and therefore the wc system from the result of a previous section ) for biologically realistic parameter , though oscillatory phenomena may occur for non - biologically relevant parameters that allow both the firing - rate and the correlations to be negative .
the picture will be quite different in multi - population networks .
indeed , in higher dimensions , the wc system can have hopf bifurcations that create very degenerate bifurcation points in the infinite - size system , which may non - trivially unfold in the finite - size systems . in this section , we consider the two population networks with excitatory self - interaction and negative feedback loop , called model i , introduced in section [ sec : negfeedback ] and address first the case of bressloff model , before showing that bcc model presents the same features in appendix [ appen:2popsbcfinite ] .
this study will allow addressing in particular the question of the existence of so - called _ quasicycles _ and also of aperiodic solutions in the system close to the hopf bifurcation of the mean - field wilson and cowan limit , addressed recently in the same model by bressloff in @xcite .
these quasicycles correspond to intrinsic noise - induced oscillations in parameter regimes where the deterministic mean - field limit ( in that case wilson and cowan system ) is below the hopf bifurcation , and where the system features a stable focus .
this phenomenon was also recently found in similar systems arising in biochemical oscillations of cellular systems , see e.g. @xcite .
our approach to these phenomena consists in unfolding the hopf bifurcation of the mean - field system to finite network sizes , in order to understand how the intrinsic noise ( that can be seen as through the langevin approximation of the system valid for large population sizes as used in appendix [ append : rodigtuck ] and in @xcite ) interacts with the hopf bifurcation . to fix ideas , we focus in this section on the particular network introduced in section [ sec : negfeedback ] , namely the two - population rescaled bressloff model . denoting @xmath234 , the equations read : @xmath235 - 2\alpha_1 c_{11}+2\,f'(s_{1})\,(w_{11}\,c_{11}+w_{12}\,c_{12})\\ c_{22}'&=\frac{1}{n_2 } [ \alpha_2 \nu_2 + f(s_2)]-2\alpha_2 c_{22}+2\,f'(s_{2})\,(w_{21}\,c_{12}+w_{22}\,c_{22})\\ c_{12}'&=-(\alpha_1+\alpha_2 ) c_{12}+f'(s_{1})\,(w_{11}\,c_{12}+w_{12}\,c_{22 } ) \\ & \qquad + f'(s_2)\,(w_{21}\,c_{11}+w_{22}\,c_{12})\\ \end{cases}\ ] ] where we denoted : @xmath236 for the set of parameters studied in this section and a number of neurons @xmath201 , bressloff finite - size system undergoes four hopf bifurcations labelled h1 through h4 and six saddle node bifurcations lp1 through lp6 .
it features two distinct families of limit cycles , one that undergoes two folds of limit cycles lpc1 and lpc2 , and one that undergoes a neimark - sacker ( torus ) bifurcation , as displayed in figure [ fig : bressbifs02](b ) , and are compared with the standard wc system that displays one hopf bifurcation point h and two saddle node bifurcations lpa and lpb .
the analysis of this bifurcation illustrates the fact that the small finite - size system studied have clear qualitative differences with the wc system .
however , because of the properties of the vector field , and in particular its smooth dependency with respect to the size of the network @xmath1 , the bifurcation diagram of fig .
[ fig : bressbifs02 ] smoothly connects to the degenerate bifurcation diagram of the infinite - size system presented in section [ sec : negfeedback ] .
let us now investigate how the bifurcation diagram of the infinite system unfolds into the bifurcation diagram of the finite - size system . to this end , we continue of the bifurcation diagram of figure fig .
[ fig : bressbifs02](b ) as the number of neurons increases , as shown in the codimension two bifurcation diagram of figure fig .
[ fig : codim2bc2pops ] .
the structure of the new bifurcation diagram appear way more intricate than wc s , and in particular , each bifurcation point of wc system , corresponding to degenerate dynamics of the infinite - size system , unfolds non - trivially into different generic bifurcations in the finite - size moment equations .
let us describe the diagram plotted in figure fig .
[ fig : codim2bc2pops ] .
we observe that the bifurcation diagram of the finite - size system , for any @xmath237 neurons will be similar to the one plotted in figure fig .
[ fig : bressbifs02 ] . as @xmath1 increases ,
the bifurcation diagram is smoothly transformed except at two particular codimension two bifurcation point : a cusp bifurcation at @xmath238 , from which emerge two additional saddle - node bifurcations denoted lp7 and lp8 .
the value of the correlation variables of these points progressively converge to zero as the network size increases .
these two saddle - node bifurcations correspond in the infinite - size limit to the two saddle nodes of wc system lpa and lpb , and are not present in the finite - size model for a small number of neurons . the second codimension two bifurcation appearing in the diagram correspond to the disappearance of the hopf bifurcation point h4 through a bogdanov - takens bifurcation with the saddle - node bifurcation manifold corresponding to lp8 . except from these two codimension two bifurcations , the continuation of all other bifurcations is smooth for any @xmath239 .
but in the limiting case @xmath198 , the system is highly singular and different bifurcation manifold collide , as we now discuss in more details . .
we observe that the fold of limit cycles merge with other bifurcations in the limit of infinitely many neurons ( see text ) . in the limit @xmath240 ( corresponding to @xmath198 ) we labelled the points h , lpa and lpb arising in wc system .
points with zero correlations are circled in red , all other points have non - zero correlations . ]
let us first follow the continuation the hopf bifurcation labelled h1 in diagram of fig . [
fig : bressbifs02](b ) as the network size increases .
we observe that this hopf bifurcation collides , in the limit @xmath240 , with the continuation of the hopf bifurcation h3 of the second branch of fixed point .
their collapse corresponds to the degenerate hopf bifurcation point of the infinite model , denoted h in fig .
[ fig : bressbifs02](a ) .
the cycle originating from h1 and the cycle originating from h3 will therefore emerge from this fixed same point , accounting for the result presented in section [ section : infinite ] and in particular figure fig .
[ fig : cyclesmeanfield2pops ] where we evidenced the presence of two different cycles emerging from the same bifurcation point h. moreover , we have seen that the corresponding branch of limit cycles in the infinite - size system bifurcation diagram fig .
[ fig : cyclesmeanfield2pops ] loses stability .
this loss of stability corresponds to the continuation of the neimark - sacker bifurcation ns of figure [ fig : bressbifs02 ] that can be continued in the mean - field limit .
however , an important distinction between the finite and infinite case is that this cycle undergoes an additional period - doubling bifurcation when the size of the network increases , as observed in the bifurcation diagram of the infinite - size system , and as shown in figure [ fig : poincarechaos ] . in that figure , we continued these cycles for one fixed value of the parameter @xmath184 , as the number of neuron increases .
the family of limit cycles continued is stable in a bounded interval of values for the parameter @xmath1 . for a number of neurons close to @xmath241 ( precisely @xmath242 ) , the branch of limit cycles undergoes a fold of limit cycle , and therefore the branch of limit cycles does not exists for smaller networks . for networks larger than @xmath241 neurons ,
the system presents a branch of stable and a branch of unstable limit cycles .
stable cycles persist up to networks as large as @xmath243 neurons ( precisely @xmath244 ) . at this size , where the branch stable limit cycles changes stability through a subcritical neimark saker ( torus ) bifurcation with no strong resonance . when increasing further the size of the network , the system shows the presence of a chaotic attractor which clearly can not exist in the two - dimensional wc system , yielding an important qualitative distinction between large scale networks and the relative mean - field limit ( see [ fig : poincarechaos ] ) .
this chaotic behavior exists until the infinite - size is reached .
let us now follow the branch of limit cycles arising from h3 as the network size varies .
we observe in our codimension three bifurcation diagram that the branch of hopf bifurcations h3 exists until the infinite - size is reached . in that limit , we observe that this branch of hopf bifurcation asymptotically meets the fold of limit cycles lpc2 and the curve of saddle - node lp7 and the saddle homoclinic curve arising from the bogdanov - takens bifurcations .
this point corresponds therefore to a very degenerate behavior of the system , and corresponds , in wc bifurcation diagram , to the saddle - node bifurcation lpa that also corresponds to the homoclinic bifurcation point .
regarding the fourth hopf bifurcation h4 , we already observed that it was continued until it disappeared , in finite size , through a bogdanov - takens bifurcation .
let us now consider the continuation of the saddle - node bifurcations .
the bifurcation diagram of wc presents two saddle - node bifurcations lpa and lpb that correspond to the two saddle - node arising from the cusp bifurcation cp of fig.[fig : codim2bc2pops ] that are labelled lp7 and lp8 . in the mean - field limit ,
the saddle - node bifurcations corresponding to the continuation of lp1 and lp4 merge in a cusp bifurcation and disappear . at this same point
, the fold of limit cycles lpc1 collapses and disappears in a homoclinic bifurcation .
the saddle - node bifurcations lp2 and lp6 disappear in the mean - field limit by being continued tangentially to the line @xmath198 .
the continuations of the saddle - node bifurcations lp3 and lp6 crosses the line @xmath198 with no singularity and therefore exists in the infinite size system , corresponding to a non - zero correlation saddle - node bifurcation and that do not correspond to any plausible behavior ( all plausible behaviors correspond to zero correlation points ) .
eventually , the saddle - node bifurcation lp7 connects with the fold of limit cycle , generating the homoclinic bifurcation of the infinite - size system .
we therefore observe that the bifurcation diagram of the finite - size bressloff model non - trivially unfolds the infinite - size system , and that at any finite scale the bifurcation diagram and the solutions it present differ significantly from the infinite - size system and from wc system .
in particular , it shows additional cycles , as observed in the infinite - size system previously , and a chaotic effect that is not present in the two - dimensional wc system .
this analysis also illustrates a mesoscopic - scale phenomenon in the presence of a periodic orbit that is stable only when the size of the network is comprised roughly between one hunded and one thousand neurons , a typical size of a cortical columns often modelled by wilson and cowan system .
these results remain valid for bcc model , as shown in appendix [ appen:2popsbcfinite ] .
the question that now arises is whether these behaviors indeed reflect behaviors of the markovian model . in the previous section
, we investigated the distinctions between the solutions of the finite - size bcc model and compared these with those of the infinite - size and of wc systems .
we now return back to the original markovian system that allowed derivation of bressloff and bcc equations , and compare simulations of this network with the finite - size bressloff model , composed of two differential equations arising from the moment truncation of this random variable .
we will perform this analysis in the two models with two populations investigated in the paper , model i and model ii .
this analysis is particularly interesting in regions of the parameter space where the system presents a fixed - point behavior , and where the moment expansion is a relevant representation of the system . in the periodic orbit regimes ,
we discuss a suitable methodology , beyond the scope of the present paper , that is suitable to analyze the system . in order to simulate the markov chain ,
we make use of doob - gillespie s algorithm @xcite . from the master equation
governing , e.g. bcc model : @xmath245\end{gathered}\ ] ] we derive the transition rates of the state variable @xmath23 .
we use these rates to simulate sample paths of the state variable @xmath23 and rescale these to compare to the solutions obtained for bcc or bressloff models and the infinite - size models . given the configuration @xmath246 of the network at time @xmath24 , we have in the bcc case ( bressloff case is straightforwardly treated using the same method ) : * the probabilistic intensity for the transition @xmath247 is equal to @xmath248 , * the probabilistic intensity for the transition @xmath249 is equal to @xmath250 .
given an initial condition @xmath251 , a simulation consists of computing all possible transitions probabilities @xmath252 and @xmath16 .
we then draw the time for the next transition as an exponential random variable with intensity @xmath253 .
this time provides the next event occurring in the chain . by the properties of exponential laws , the transition is @xmath247 ( resp .
@xmath249 ) for some @xmath254 with probability @xmath255 ( resp .
@xmath256 ) .
this simulation algorithm for the markov process @xmath246 is exact in law , and does not involve any approximation such as time - discretization , and therefore allows efficient simulation of the sample path and of its statistics .
these statistics are computed using a monte - carlo algorithm consisting of simulating many independent sample paths of the process and deriving from these the trajectory of the mean firing rate @xmath14 and the correlation in the firing activity @xmath257 through the empirical mean and correlations obtained . in the simulations we focus on bcc model .
the same numerical experiments can be performed in the bressloff case .
we start by considering the one - population model studied in section [ sec : onepop ] . in that case
we have shown that both wilson and cowan and the infinite - size systems had the same behavior , and that this behavior was regularly unfolded in the finite - size systems . in these cases , two saddle - node bifurcations defined a zone of bistability that depended on the size of the network as shown in figure [ fig:1popcodim2bc](d ) .
we simulate the network using doob - gillespie algorithm and identify the boundaries of the zones where bistability occurs by continuation of the fixed points obtained as the parameters increase or decrease .
the continuation is initialized by running the algorithm with parameters associated to a single attractive fixed point of wc system .
then the parameter @xmath200 is slowly varied and simulations are run with initial condition being the last values of the network state of the previous simulation .
we observe the hysteresis phenomenon and bistability , and the bistability zones closely matches wc system , as shown in figure [ fig : markovonepop ] .
we simulate the network for increasingly large networks ranging from @xmath201 neurons to @xmath258 neurons with @xmath259 simulated sample path from which we extract the empirical mean and standard deviations . from the curves obtained in that fashion
, we automatically identify zones where there is bistability .
these boundaries are plotted together with the values of the saddle - node bifurcation at these points arising from the codimension 2 bifurcation diagram of bcc finite - size model , and we observe that the curves closely match as soon as the network is larger than @xmath260 neurons . therefore in this model , bcc system closely reproduces the behavior of the finite - size markov model for sufficiently large networks .
a similar result is obtained in bcc case ( not shown ) .
we now turn to study the two - dimensional network model i. in that case we showed that depending on the input parameter @xmath186 , wilson and cowan system can either present a stable cycle or a stable fixed point .
we first chose a case where wc system converges towards a fixed point , and to this end will study the system for @xmath261 . in that case , both wc system , the infinite - size system , bcc and bressloff systems converge towards a fixed point .
the periodic regimes are addressed in section [ sec : cyclesmarkov ] . in the case where wc system presents a unique hyperbolic attractive fixed point
, we showed that the infinite - size presents an exponentially stable fixed point with the same values of mean firing rate and null correlations , which has a counterpart in the unfolding of bcc and bressloff model .
simulations of the markovian model illustrate the fact that all the sample paths of the process converge quite fast towards this fixed point with null correlations displaying stochastic variations , and each sample path presents a precise match with the solution of wc solution at this point . the stochastic variations around wilson and cowan trajectory arising from the randomness of the markov chain are averaged , and produce a fixed mean firing - rate corresponding to the fixed point of the infinite - size system , i.e. having null correlations and the same value of mean firing rates .
moreover , we observe that the correlations vanish , which implies that the state obtained is an asynchronous state , as expected from bressloff s expansion .
figure [ fig : markovfp ] represents simulations of model i with @xmath261 .
+ in panels ( a ) and ( b ) are plotted four sample paths of the process for @xmath259 neurons simulated over on a time interval equal to @xmath262 , and panels ( c ) and ( d ) the mean - firing rates averaged over @xmath263 realization of the process , and that s shows a very close agreement with the solution of wc system .
the maximal absolute value of the correlations ( not plotted ) are of the order of @xmath264 and are neglected as an effect of the finiteness of the sample path simulated for computing the mean and covariance .
the same observations hold for all the simulations performed for parameters corresponding to an hyperbolic fixed point of the infinite system .
note that the difference between mean - firing rates of the solution of wc system on one hand and of bcc system on the other hand is of the order of @xmath264 at this network scale , and therefore both models fairly approximate the stochastic markovian model .
we perform the same analysis in the case of the network model ii studied in section [ sec : second2pop ] with symmetric connectivity with null diagonal and inhibitory interaction in bcc case . in that case
, we showed that wc system presented bistability , and the infinite - size system presented a more complex behavior with multistability and different limit cycles . in the simulations of the markovian system , similarly to the case of model i , we observe that when wc system presents a single hyperbolic fixed point , the markovian system closely matches with the solution of wc system during the transient phase , and the stationary state randomly varies around the fixed point of wc system .
the mean of the firing rates over many realization of the process converges exactly towards the solution of wc system , and the correlations tend to zero , implying the system is in an asynchronous state and that its behavior is efficiently summarized by wilson and cowan system .
we note that in that case , the infinite - size system and the finite size systems both presented additional fixed points with non - zero correlations , that do not actually exist as a regime of the markov chain . for parameters associated with bistability in wc ,
the situation is similar , and we observe in our simulations that the wc system captures precisely the behavior of the markovian system and all additional solutions derived from the moment hierarchy truncation do not appear .
simulations of the markovian system show that depending on the initial condition , the system will converge towards one or the other fixed point .
albeit the stochasticity of the markov chain yielding the possibility to switch between the two fixed points , this phenomenon does not occurs often , and except for input parameters very close to the saddle - node bifurcations , no switching between fixed points is observed . in that case
, we note that the additional fixed point with non - zero correlations and the additional cycles evidenced in the infinite - size system and that are present in the finite - size moment equations do not appear in the simulations of the underlying markovian model .
we present in figure [ fig : network2 ] the values of taken by the mean firing rate and correlations for @xmath265 realizations of the markov chain over the @xmath266 last units of time for a simulation of @xmath267 neurons as a function of the input paramter @xmath186 .
we treat the cases @xmath194 and @xmath196 , the blue crosses are the continuation of the fixed points with @xmath186 increasing ( from @xmath268 to @xmath269 for @xmath194 and from @xmath270 to @xmath271 for @xmath196 ) and red circles for @xmath186 decreasing .
the continuation is initialized by running the algorithm with parameters associated to a single attractive fixed point of wc system .
then the parameter @xmath186 is slowly varied and simulations are run with initial condition being the last values of the network state of the previous simulation .
we observe the hysteresis phenomenon and bistability , and the bifurcation diagram obtained perfectly matches wc s , and no counterpart of any of the identified correlations effects are observed . as a function of @xmath186 ( see text ) .
continuation of equilibria by increasing @xmath186 ( blue crosses ) and by decreasing @xmath186 ( red circles ) , superimposed with wc bifurcation diagram ( black curve ) .
correlations are plotted with a different axis origin in cyan ( @xmath272 ) , magenta ( @xmath273 ) and yellow ( @xmath274 ) , keep very close to zero except at the precise value of the bifurcations , where the jump between the two fixed points produces a spurious standard deviation .
bcc case.,title="fig : " ] as a function of @xmath186 ( see text ) .
continuation of equilibria by increasing @xmath186 ( blue crosses ) and by decreasing @xmath186 ( red circles ) , superimposed with wc bifurcation diagram ( black curve ) .
correlations are plotted with a different axis origin in cyan ( @xmath272 ) , magenta ( @xmath273 ) and yellow ( @xmath274 ) , keep very close to zero except at the precise value of the bifurcations , where the jump between the two fixed points produces a spurious standard deviation .
bcc case.,title="fig : " ] it appears from this analysis that the finite - size moment equations present additional solutions , associated with large correlations , that do not exist in the initial markovian model .
we emphasize the fact that though the infinite - size model is not equivalent to the moment equations derived by both authors , the solutions evidenced have a counterpart in the finite - size moment equations corresponding to both systems .
these equations present solutions with large correlations ( that actually diverge as the number of neurons tend to infinity ) , which are not solutions of the initial markov chain , which naturally behaves as wc system for large population sizes , as expected from the moment equations by both authors . a complementary approach to the study of bistability in the markov framework allowing to quantify the probabilistic intensity of noise - induced transitions can be driven through the use of wentzel kramers brillouin ( wkb ) approximation ( see e.g. @xcite and references therein ) performed directly on the markovian system .
this method was used by bressloff in @xcite in the same context for a one - population model , and more generally allows studying noise - induced transitions between multiple metastable states where the mean - field theory based on the system - size or loop expansions break down .
it allows reducing the problem to a hamilton - jacobi equation from which one can derive the escape rate , namely the probability intensity that the solution escapes from the attraction bassin of a given equilibrium of the deterministic equation ( the wc system ) and enters the attraction bassin of the other equilibrium .
the models proposed in @xcite emphasize the role of the mean - firing rate and of the zero timelag correlation functions .
as illustrated in the previous section , these variables are suitable to model the system at fixed points .
however , it remains unclear how these variables characterize the behavior of the markovian system in periodic or chaotic regimes .
the analytical study of such regimes can be carried out in a satisfactory fashion either in the theory of master equations and in the theory of stochastic dynamical systems , using the langevin equation .
such studies have been addressed by various authors within the context of chemical master equations @xcite and recently bressloff applied the same tools to neural models , using langevin equations and floquet theory @xcite .
indeed , it appears that in regimes where the wilson and cowan system present a cycle , the markovian model is characterized by random fluctuations in the neighborhood of this attractor , and the system loses phase information over time .
these dynamics can be suitably studied in terms of a diffusive phase variable . in details , under the langevin approximation of the dynamics , a possible way to study the behavior of the solutions of the markovian system would be to use the stochastic phase reduction ( see @xcite ) .
in this section we do not make use of these more customary methods but rather investigate the relationship between behaviors of the finite - size moment equations and behaviors of the markovian system . for our choice of parameters in model
i , let us consider the parameter region @xmath275 $ ] where @xmath276 ( resp .
@xmath277 ) correspond to the value of @xmath186 at the point h ( resp .
lpa ) . in this region , the solutions of wc system are periodic orbits .
the question we address is whether the behavior of the finite - size ( moment equations and markovian ) systems is comparable to the behavior of wc solutions or to the behavior of the finite - size moment equations , which differs at any finite scale from wc s behavior as shown in section [ sec : twopopsfinite ] .
indeed , we observed in figure fig .
[ fig : codim2bc2pops ] that the hopf bifurcation of the deterministic wc system , which is a degenerate point of the moment equations , expands in the codimension two diagram into two branches of generic hopf bifurcations , that imply the presence of cycles in the finite - size system .
these hopf bifurcations either unfold a singularity corresponding to the actual eigenvalues @xmath278 of wc hopf bifurcation or to twice this value by application of the lemma [ thm : wilsonbuicecowan ] , and can be either stable or unstable . therefore the boundaries of the region where the system is in a periodic regime are modified as soon as the system size is finite .
the same phenomenon appears close to lpa , in which case the diagram unfolds into a hopf bifurcation , a saddle node bifurcation , a fold of limit cycles and a saddle homoclinic bifurcation .
we will observe that in both cases , the behavior of the markovian system actually reflects the properties of the finite - size moment equations , and that quasicycles will appear in the neighborhood of wc s hopf bifurcation .
let us first have a closer look to the behavior of the moment equations in the neighborhood of @xmath279 in the finite - size system .
wc s hopf bifurcation gives rise to two hopf bifurcations whose codimension two curves are tangent to the axis @xmath198 .
the curve h1 is related to a singular eigenvalue converging as @xmath1 goes to infinity to the value of wc hopf bifurcation , and gives rise to stable limit cycles with period close to the expected cycle in wc system .
the other branch , h3 , corresponds to faster cycles with period around half wc cycles period .
let us now fix a certain size of the network @xmath1 . in the moment equations , and let @xmath186 be a free bifurcation parameter . as @xmath186 increases , cycles will appear corresponding to the hopf bifurcation h1 at a value @xmath280 .
we observe that this value is increasing as the network size increases .
therefore , in regions where wc system has a stable focus , the finite - size moment equations show oscillations of frequency close to wc cycles frequency .
for instance for @xmath281 corresponding to the bifurcation diagram of figure fig .
[ fig : bressbifs02 ] , we have @xmath282 , defining a region of quasicycles @xmath283 $ ] .
this size of the zone where quasicycle appear decreases very fast ( as we can see from the shape of the graph of h1 in the codimension two plot fig .
[ fig : codim2bc2pops ] and for @xmath284 we have @xmath285 .
it is important to note that at a size @xmath201 , oscillations are very difficult to see because of the small system size .
moreover , in the particular case treated , the graph of the h1 bifurcation is almost a vertical line : there is almost no difference between the value of h1 across different system sizes . the same phenomenon is observed in the simulations : quasicycles close to the hopf bifurcation point barely exist , and the value of @xmath186 needs to be very close from @xmath276 to start observing these quasicycles .
wc s family of limit cycle lose stability through a saddle - node homoclinic bifurcation .
we observe that unfolding this bifurcation , we find out that the branch of limit cycles in the finite - size system undergo a fold of limit cycles corresponding to a decreasing value of @xmath186 , having the effect of reducing the amplitude of the interval of @xmath186 values related to a periodic orbit .
the curve of fold of limit cycles is no longer vertical in that case and a wide non - trivial region distinguishes the finite - size and wc s periodic orbit parameter regions .
we observe again in that case that the lpc2 line actually corresponds to a good estimate of the disappearance of cycles . indeed
, the codimension 2 bifurcation diagram shows that for @xmath286 oscillations should disappear at a system size close to 550 neurons . in figure fig .
[ fig : spectrum ] we observe that indeed , for @xmath287 neurons and @xmath286 , the system does not show any oscillations , and for @xmath288 neurons the system starts showing an oscillatory behavior .
moreover , in that case we observe that the amplitude of the cycles is not small , consistent with the cycles we observe in the finite - size system . the presence of a chaotic orbit as seen in fig .
[ fig : poincarechaos ] and the fact that indeed in such regions the behavior of the system is aperiodic suggests that in these regions close to a limit cycle of wilson and cowan system , the underlying markovian system can present a low - dimensional chaotic macroscopic behavior arising through the interaction of a large but finite number of neurons , a phenomenon that could be related to the work of el boustani and destexhe in @xcite , where intrinsically stochastic neurons are shown to yield low - dimensional chaotic macroscopic behaviors .
results are presented in figure fig .
[ fig : spectrum ] + + this particularly good correspondence between cycles of the moment equations and quasiperiodic orbits of the markov chain is a very promising result .
however , it is important to note that the additional cycles corresponding to the branches h2 and h4 do not appear in the markov simulations .
this might be linked with the stability of such orbits and the system size at which stable periodic orbit exist .
this study therefore indicates that further exploration is needed to assess more clearly a definitive relationship between quasicycles and actual orbits of the moment equations , and there is a good chance that these two phenomena are related .
this precise study is not within the scope of the present paper and we are currently actively working on this topic .
of high interest also is the fact that in some regions corresponding to the vicinity of a hopf bifurcation , the moment equations present a low - dimensional chaotic behavior .
such phenomena , of interest in neuroscience , have , to the best of our knowledge , not been shown , and we believe that if these phenomena indeed appear in the markovian system , this is of crucial interest in neuroscience and in the understanding of large - size stochastic systems .
in this paper , we studied the different solutions of newly developed models for describing the mean - field limit of markovian neural networks modeling large - scale cortical areas : bcc and bressloff models@xcite .
these equations were originally derived from a markovian implementation of the firing of each neurons in the population , whose transition probability satisfied a differential equation , the master equation of the process@xcite . both models studied drew on this framework but , since they addressed different neuronal regimes , dealt with different variables and different scaling , and after truncation of the moment hierarchy yielded sets two coupled equations , one governing the mean firing activity and the other governing the same - time correlations of the firing activity in the network ( or the ordered cumulant that can be deduced from the correlation and the mean firing rate ) .
the particular aim of both models was to recover , in the limit where the number of neurons is infinite , the standard wc equations .
we transformed these models in order to study the proportion of active neurons in each populations , in networks composed of different populations of neurons with different dynamics and composed of different number of neurons .
both models include finite - size corrections , and the size of the network appears simply in the equations as a parameter of the model .
the limit when the number of neurons tend to infinity appears in that view as a regular perturbation of the finite - size system in our particular choice of variables , and that both the bcc- and bressloff - derived models converged towards the same equations as the number of neurons tends to infinity , the infinite - size system .
we analyzed the solutions of the infinite - size system , and established the fact that all solutions of wc system were associated with solutions of the infinite - size system with null correlations .
however , we showed that the stability of the solutions obtained in that fashion did not necessarily have the same stability properties as in wc system : fixed points of the wilson and cowan system correspond to fixed points of the infinite - size system with zero correlations having the same stability , but all the cycles of the wc system are either unstable or neutrally stable in the infinite - size system .
this destabilization of cycles gives rise to new behaviors ( fixed points or periodic orbits ) with non - zero correlations , the correlation - induced behaviors .
when studying the finite - size equations and the way it relates to the properties of the infinite - size system , we presented evidence that the mean - field limit appears to be a singular limit , in the sense that for any finite - size of the network , the system presents behaviors the infinite - size system does not feature .
moreover , singular points appear in the limit of infinitely many neurons , and for infinitely many neurons , branches of bifurcations collapse and disappear . besides these qualitative distinctions between finite and infinite - size systems , we numerically found mesoscopic - scale effects , namely behaviors that are only present at a certain population scale , and disappear for smaller or larger populations .
the scale at which these behaviors took place coincided with the scale of typical cortical columns .
such behaviors are neither captured by the study of small - networks nor by the mean - field limit , stressing the importance of precise descriptions of these intermediate scales in tractable approaches .
we want to emphasize that the system - size expansion and the truncations are not valid at the bifurcation points . however , the bifurcation analysis we performed allowed us to uncover the number and stability of fixed points and limit cycles corresponding to different parameters .
the actual transition of the stochastic process is a very complex phenomenon which would be worth studying but that present very important technical intricacies .
we also note that the underlying markovian models are stochastic models .
if the moment equations present multiple attractors , the stochastic system will randomly switch between the different attraction basins . a large - deviation study in the case where there is a small parameter ( bressloff case and bcc case in the fully connected system treated in the manuscript ) can allow one to identify the attractor switching and the mean exit time of each attraction basin in order to more precisely study the behavior of the underlying stochastic system , that could possibly be performed using freidlin and wentzell theory in the large @xmath1 case @xcite . as already mentioned , paul bressloff opened the way to such studies by using wkb approximations allowing to compute the transition rates between metastable attractors in a one - population model @xcite .
alternative approaches could make use of the stochastic dynamical systems theory @xcite .
these approaches would allow directly studying the bifurcations of the initial stochastic system in the langevin approximation .
correlation - induced and finite - size behaviors were then compared to the initial markovian model .
we observed that when wc system presented an attractive fixed - point behavior , each sample path of the markov chain presented stochastic variations around these fixed points , the correlations tended towards zero and the mean coincided with wc fixed point . in that case , the network regime is therefore an asynchronous state , and is well approximated by wilson and cowan system .
the validity of the moment equations close to bifurcations was shown in the derivations to be mainly valid far away from bifurcations .
however , we showed that the codimension two bifurcations of the moment equations was closely matched with similar phenomena ( e.g. bistability ) observed in simulations of the markovian system .
however , we also observed that both bcc and bressloff models can present additional behaviors in these regimes that are not present in the simulations of the markov chain .
the moment expansion breaks down when considering periodic orbits of the wc system , corresponding to quasi - periodic stochastic solutions of the stochastic system , and different techniques such as the stochastic phase reduction , langevin approximation and floquet theory are very promising .
however , we provided numerical evidences showing that actual cycles in the moment equations could be related to the phenomenon of quasicycles in the markovian system .
these numerical observations strongly suggest further exploration along these lines .
the precise study of all these points will further allow understanding the behavior of large scale systems , and will particularly be interesting in the study of rhythms in the brain , and possible synchronization or desynchronization as a function of the network size and of the connectivity .
a particularly interesting point of the models studied in this paper is that they describe a stochastic process by a set of differential equations , where the size of the network appears as a perturbation parameter of a set of ordinary differential equations .
this approach can be promising in order to study bifurcations of a stochastic processes and qualitatively analyzing its behaviors and the dependency of its solutions as a function of the parameters .
we also noted that the behavior of each sample path of the process was possibly very different from the behavior of its first moments , which raises the question of the best variable to describe such stochastic processes .
the mathematical method developed to study the stability of the correlation equations , involving kronecker products and sum and the application of floquet theory is of potential importance in the study of networks with correlations , since the form of the correlation equation is quite standard in this kind of problems ( see e.g. @xcite in the case of small noise expansion ) .
this approach can also be applied to the study of the stability of solutions in networks involving hebbian learning , since the connectivity weights evolution depends on the firing correlations and the stability of the weights involves very similar equations as the ones we studied here .
the need for extensions of the present study to more realistic markovian models involving more relevant biologically realistic transitions is a direct consequence of the present work .
indeed , we have seen that the transitions chosen in the markovian model are only one - step jumps . in biological terms
, this assumption means that when @xmath19 neurons are active in population at time @xmath7 , at least @xmath289 neurons are active at time @xmath290 , i.e. keep firing , which is not biologically realistic since after firing , a neuron immediately returns to quiescence .
therefore extending the framework to a more complex state space of the markov chain would be of great interest for the development of these models towards more realistic biological grounds , and would probably allow a better assessment of the properties of cortical areas . finally , in order to go beyond the description of the moments of the markov chain , we believe that one shall turn towards alternative ways to derive mean - field equations for such systems .
a particularly interesting way to study such systems we are currently exploring builds upon recent methods developed for the study of markov chains , mainly in the field of queueing theory , such as fluid limits and large deviations approaches .
the fluid limit technique is based on a space - time proper rescaling depending on the process under consideration , that yields deterministic or stochastic limits of the activity of the network when the number of neurons tends to infinity .
large deviation techniques allow one to derive the finite - size corrections ( see e.g. @xcite in a case of stochastic ion channels ) and is a natural extension of our work .
* acknowledgments : * this work was supported by nsf dms 0817131 .
in this appendix we show that the moment equation corresponding to the markov chain governed by equations can be derived from a rodriguez - tuckwell moment expansion on the langevin approximation of the markov chain .
following kurtz approach in @xcite , we know that the dynamics of the our rescaled variables @xmath291 , which is a markov chain in the initial setting , approaches as @xmath1 goes to infinity a continuous diffusion process @xmath292 satisfying the equation : @xmath293 where @xmath294 . following the works of rodriguez and tuckwell @xcite ,
we derive from this equation the dynamical system governing the approximate moments of @xmath116 . denoting by @xmath295 the mean value of @xmath296 and by @xmath43 the correlation between @xmath297 and @xmath296 , a direct application of rodriguez and tuckwell formula applied to our particular form of dynamical system yields : @xmath298 \end{cases}\ ] ] where @xmath299 .
these equations therefore appear as a perturbation of the bcc equations with an additional nonlinear term in the dynamics or order two .
truncation at order @xmath143 in the small parameter @xmath0 yields exactly bcc equations .
moreover , it is interesting to note that these equations again correspond in the limit @xmath240 , to the infinite model described in section [ section : infinite ] .
in this appendix , we review and prove some useful properties of kronecker products of matrixes .
we recall the definition of the the function @xmath61 transforming a @xmath300 matrix into a @xmath301-dimensional column vector , as defined in @xcite : @xmath302^t \end{cases}\ ] ] let us now denote by @xmath64 the kronecker product defined for @xmath65 and @xmath66 as the @xmath67 matrix : @xmath68 for standard definitions and identities in the field of kronecker products , the reader is referred to @xcite .
we recalled in the main text the following identities for @xmath69 , @xmath70 be the @xmath71 identity matrix and @xmath72 or @xmath73 denote the standard matrix product : @xmath303 the relationship @xmath75 is called kronecker sum .
[ prop : kroneigen ] let @xmath304 and @xmath305 in @xmath306 , and assume that @xmath304 has the eigenvalues @xmath307 and @xmath305 the eigenvalues @xmath308
. then we have : * @xmath309 has the eigenvalues @xmath310 .
* @xmath311 has the eigenvalues @xmath312
. let @xmath131 ( resp.@xmath313 ) be an eigenvalue of @xmath304 ( resp .
@xmath305 ) with eigenvector @xmath314 ( resp .
@xmath315 ) , and define the matrix @xmath316 ( i.e. @xmath317 ) .
we have : @xmath318 which entails that @xmath319 is an eigenvector of @xmath127 associated with the eigenvalue @xmath320 .
similarly , we have for @xmath321 : @xmath322 the dimension of @xmath311 and @xmath309 is @xmath60 and there are exactly @xmath60 linearly independent matrices @xmath323 possible built in the proposed fashion , therefore we identified all possible eigenvalues .
[ theo : psikron ] let @xmath144 be the solution of the matrix differential equation : @xmath324 and @xmath325 the solution of : @xmath326 we have @xmath327 .
indeed , we have , using the basic properties of the kronecker product recalled in equation @xmath328 therefore @xmath329 satisfies the same differential equation as @xmath325 and moreover , @xmath330 , and therefore by existence and uniqueness of the resolvent , @xmath327 .
in this appendix we derive the expression of the first lyapunov exponent of the bifurcation , which proves that the existence of the hopf bifurcation exhibited in section [ ssect : onepophopf ] . in that section
, we derived the expression of the jacobian matrix at the considered fixed point : @xmath331 at the bifurcation point , we have @xmath332 , and therefore at this point the jacobian matrix reads : @xmath333 the eigenvalues of this matrix under the assumptions of section [ ssect : onepophopf ] are @xmath334 where @xmath335 .
we define @xmath336 the right eigenvector of @xmath337 associated with @xmath338 : @xmath339 and @xmath233 the right eigenvector of @xmath340 associated with the eigenvalue @xmath341 : @xmath342 for the sake of simplicity , we also name the components of the vector field of the system : @xmath343 following @xcite , we define @xmath344 and @xmath345 the second and third derivatives of the vector field , which are bi- and tri - linear forms .
we have the following expressions for these multilinear functions : @xmath346 we are now in position to compute the first lyapunov exponent @xmath347 using the formula : @xmath348 where @xmath349 denotes the complex inner product @xmath350 and the sum of three terms denoted @xmath351 , @xmath352 and @xmath353 are the real parts of the terms involved in the expression of the lyapunov exponent . after straightforward but tedious calculations ( that can be conveniently performed using a formal calculation tool such as maple ) ,
we obtain : @xmath354 which yields the expression for the lyapunov exponent : @xmath355\\ & = \frac{w^2\,n}{2\,\omega_0\,f'(i)^2 } \left [ f^{(3)}(i)f'(i ) \left(1+\frac { 1}{\omega_0}\right ) + f''(i)^2\left(\frac{2}{\omega_0 } - \frac{14}{3}\right)\right ] \end{aligned}\ ] ]
in this appendix , we study the two - populations bcc system corresponding to model i and show that the finite - size effects are closely related to what is observed in bressloff model as studied in section [ sec : twopopsfinite ] .
bcc finite - size equations read : @xmath356 where we denoted : @xmath357 similarly to bressloff case , bcc model features two families of limit cycles .
one of these branches corresponds exactly to the branch of limit cycles of wc system starting from a hopf bifurcation and disappearing through a homoclinic bifurcations .
two additional hopf bifurcations appear related to the family of periodic orbit corresponding to the correlation - induced cycle evidence in the analysis of the infinite - size system .
this branch of limit cycles exist whatever @xmath23 , and loses stability through a neimark - sacker bifurcation as the number of neurons increases
. this bifurcation generates chaos for large enough networks , which clearly does not exists in wc system .
this family of limit cycles presents stability for networks containing between 100 and 17.000 neurons , corresponding to a finite size effect that appears only in the region of interest of the cortical columns .
as the number of neurons increase , the system keep the same number and type of bifurcations , and the infinite model appears to be a singular case where different bifurcations meet ( see fig . [
fig : bccod2 ] ) , and very large networks lose the property of presenting a stable cycle , and present a chaotic behavior . , _ an effective kinetic representation of fluctuation - driven neuronal networks with application to simple and complex cells in visual cortex _ , proceedings of the national academy of sciences , 101 ( 2004 ) , pp .
77577762 . , _ brain dynamics at multiple scales : can one reconcile the apparent low - dimensional chaos of macroscopic variables with the seemingly stochastic behavior of single neurons ?
_ , international journal of bifurcation and chaos , 20 ( 2010 ) , pp . 116 . | the brain s activity is characterized by the interaction of a very large number of neurons that are strongly affected by noise .
however , signals often arise at macroscopic scales integrating the effect of many neurons into a reliable pattern of activity . in order to study such large neuronal assemblies ,
one is often led to derive mean - field limits summarizing the effect of the interaction of a large number of neurons into an effective signal .
classical mean - field approaches consider the evolution of a deterministic variable , the mean activity , thus neglecting the stochastic nature of neural behavior . in this article
, we build upon two recent approaches that include correlations and higher order moments in mean - field equations , and study how these stochastic effects influence the solutions of the mean - field equations , both in the limit of an infinite number of neurons and for large yet finite networks .
we introduce a new model , the infinite model , which arises from both equations by a rescaling of the variables and , which is invertible for finite - size networks , and hence , provides equivalent equations to those previously derived models .
the study of this model allows us to understand qualitative behavior of such large - scale networks .
we show that , though the solutions of the deterministic mean - field equation constitute uncorrelated solutions of the new mean - field equations , the stability properties of limit cycles are modified by the presence of correlations , and additional non - trivial behaviors including periodic orbits appear when there were none in the mean field .
the origin of all these behaviors is then explored in finite - size networks where interesting mesoscopic scale effects appear .
this study leads us to show that the infinite - size system appears as a singular limit of the network equations , and for any finite network , the system will differ from the infinite system . [ [ keywords ] ] keywords + + + + + + + + neural mass equations ; dynamical systems ; markov process ; master equation ; moment equations ; bifurcations ; wilson and cowan system . |
paper studies time - slotted communications over networks consisting of nodes and directed edges that connect the nodes , and the networks may contain cycles . each edge receives a symbol from a node and outputs a symbol to a node in each time slot , where the duration of a time slot is set to be the maximum propagation delays experienced by the edges . in practical communication networks ,
propagation delays of different links may vary significantly due to different distances and different transmission medium ( e.g. , optical fiber , air , water , etc . ) across different links .
for example , links with relatively short distances have shorter propagation delays compared to those with relatively long distances , and links established through the optical fiber medium generally experience negligible propagation delays compared to links established through the water medium . in order to characterize the scenario where the propagation delays experienced by some edges are negligible compared to the delays experienced by the other edges
, we allow the edges with negligible delays to be operated before the rest of the edges in each time slot .
since the symbols transmitted on earlier - operated edges may depend on the symbols output from latter - operated edges , we say that an edge @xmath2 terminating at node @xmath3 _ incurs zero delay _ on an edge @xmath1 originating from node @xmath3 if @xmath2 is operated before @xmath1 ; otherwise , we say @xmath2 _ incurs a unit delay _ on @xmath1 .
similarly , the network is said to _ contain zero - delay edges _ if there exists an edge that incurs zero delay on another edge ; the network is said to _ contain no zero - delay edge _ if every edge incurs a delay on every other edge . under the classical model , every discrete memoryless network ( dmn ) ( * ? ? ?
15 ) is assumed to contain no zero - delay edge because all the edges are operated at the same time .
a well - known outer bound on the capacity region of the dmn that contains no zero - delay edge is the _ classical cut - set bound _
it is easy to construct a network with zero - delay edges whose capacity region is strictly larger than the cut - set bound .
one such network is the binary symmetric channel with correlated feedback ( bsc - cf ) considered in ( * ? ? ?
vii ) , which will be introduced in the next subsection .
consider a network that consists of two nodes denoted by @xmath4 and @xmath5 respectively and two edges denoted by @xmath6 and @xmath7 respectively .
node 1 and node 2 want to transmit a message to each other .
this is a two - way channel @xcite . in each time slot ,
node 1 and node 2 transmit @xmath8 and @xmath9 respectively , and they receive @xmath10 and @xmath11 respectively .
all the input and output alphabets are binary , and the channel associated with edge @xmath6 is a binary symmetric channel ( bsc ) while the channel associated with edge @xmath7 is a discrete memoryless channel ( dmc ) whose output may depend on the output of channel @xmath6 . in this network ,
channel @xmath6 incurs zero delay on channel @xmath7 , i.e. , node @xmath5 can receive @xmath11 before encoding and transmitting @xmath9 .
we call this network the _ bsc with dmc feedback _ ( bsc - dmcf ) , which is illustrated in figure [ bscfb](a ) . when @xmath12 , the bsc - dmcf is also referred to as the bsc - cf in ( * ? ? ?
it has been shown in ( * ? ? ?
vii ) that the capacity region of the bsc - cf is strictly larger than the classical cut - set bound , where the classical cut - set bound is obtained under the assumption that the network contains no zero - delay edge while the capacity region of the bsc - cf is achieved when edge @xmath6 incurs zero delay on edge @xmath7 .
consequently , we have the following conclusion : ( @xmath13 ) : : _ the capacity region of some bsc - dmcf with zero - delay edges is strictly larger than the classical cut - set bound_. however , statement ( @xmath13 ) is based on an important assumption : the outputs of the two channels can have correlation given their inputs . in other words , the noises of the two channels can be correlated given the channel inputs .
if the noises are assumed to be independent given the channel inputs , i.e. , @xmath14 for all @xmath15 , then it is not clear whether statement ( @xmath13 ) still holds . to facilitate discussion
, we call the bsc - dmcf which satisfies the _ bsc with independent feedback _ ( bsc - if ) , which is illustrated in figure [ bscfb](b ) .
indeed , the bsc - if with zero - delay edges always lies within the classical cut - set bound due to the fact that the two channels are independent and the well - known fact that the presence of instantaneous feedback does not increase the capacity of a point - to - point channel ( * ? ? ?
* sec . 7.12 ) .
consequently , statement ( @xmath13 ) does not hold for the bsc - if .
this motivates us to investigate a general network with zero - delay edges under the assumption that the channels are independent , and compare its capacity region with the cut - set bound .
consider a two - relay network ( trn ) illustrated in figure [ 2relaynetwork ] , which consists of a source denoted by @xmath4 , two relays denoted by @xmath5 and @xmath16 respectively , and a destination denoted by @xmath17 .
the source wants to send information to the destination with the help of the two relays .
suppose all the input and output alphabets are binary . in each time slot , @xmath18 is transmitted on edge @xmath1 by node @xmath3 and @xmath19 is received from edge @xmath1 by node @xmath20 for each edge @xmath1 in the relay network .
in addition , suppose @xmath21 @xmath22 and @xmath23 where @xmath24 and @xmath25 are two independent bernoulli random variables with @xmath26 to facilitate discussion , we call the trn described above the _ two - relay network with correlated noises ( trn - cn)_. it can be easily seen that if edge @xmath6 incurs zero delay on edge @xmath27 and edge @xmath28 incurs zero delay on edge @xmath29 , then node 4 can receive one bit per time slot from node 1 with the help of nodes @xmath5 and @xmath16 sending @xmath24 and @xmath25 respectively ( cf . ) .
on the contrary , if either edge @xmath6 incurs a delay on edge @xmath27 or edge @xmath28 incurs a delay on edge @xmath29 , then node 4 can not receive any information from node 1 because the independent uniform bits @xmath24 and @xmath25 can not be completely cancelled simultaneously and @xmath25 generated in different time slots are assumed to be independent . ] .
consequently , the capacity of the trn - cn with zero - delay edges is strictly larger than the classical cut - set bound , i.e. , @xmath30 .
consider another trn with the same topology as illustrated in figure [ 2relaynetwork ] and specified by @xmath31 for all @xmath32 and @xmath33 where @xmath34 s are independent bernoulli random variables . to facilitate discussion ,
we call the trn described above the _ two - relay network with independent noises ( trn - in)_. it can be easily seen that the capacity of the trn - in , which is equal to the capacity of channel @xmath35 specified in , coincides with the cut - set bound even when zero - delay edges are present . in this paper , we consider the multimessage multicast network ( mmn ) ( * ? ? ?
18 ) consisting of independent channels , where the destination nodes want to decode the same set of messages transmitted by the source nodes .
two simple examples of the mmn with independent channel are the following two networks introduced in section [ sectionmotivating ] the bsc - if ( where both nodes want to decode all the messages ) and the trn - in , which belong to the class of mmns consisting of independent discrete memoryless channels ( dmcs ) @xcite .
note that the bsc - cf , unlike the bsc - if , does not belong to the class of mmns with independent dmcs because the forward and reverse channels of the bsc - cf are correlated ( cf .
figure [ bscfb](a ) ) .
similarly , the trn - cn , unlike the trn - in , does not belong to the class of mmns with independent dmcs because the noises among the channels of the trn - cn are correlated ( cf . figure [ 2relaynetwork ] , , and ) .
we propose an edge - delay model for the discrete memoryless mmn ( dm - mmn ) with independent channels and zero - delay edges .
in our model , each channel is associated with a set of directed edges ( e.g. , a channel characterized by @xmath36 is associated with @xmath37 ) and the channels are operated in a predetermined order so that the output random variables generated by earlier - operated channels are available for encoding the input random variables for latter - operated channels . therefore , an edge may incur zero delay on another edge under our model .
our edge - delay model can be used to investigate the practical situation when some edges with negligible propagation delays are allowed to be operated before the rest of the edges in each time slot ( for instance , in a cellular network , edges that connect the mobile users to their closest relays may experience negligible propagation delays compared to those that connect the relays to the base stations ) .
the channels of the mmn are assumed to be _ independent _ , meaning that the outputs among the channels are independent given their inputs , but the outputs within a channel are allowed to correlate with each other .
the main contribution of this paper is twofold : first , we establish an edge - delay model for the mmn consisting of independent channels which may contain zero - delay edges .
our model subsumes the classical model which assumes that every edge incurs a unit delay on every adjacent edge .
second , we prove that for each dm - mmn consisting of independent channels with zero - delay edges , the capacity region always lies within the classical cut - set bound despite a violation of the classical unit - delay assumption .
combining our cut - set bound result with existing achievability results from network equivalence theory @xcite and noisy network coding ( nnc ) @xcite , we establish the tightness of our cut - set bound under our edge - delay model for the mmn with independent dmcs , and hence fully characterize the capacity region . more specifically ,
we show that the capacity region is the same as the set of achievable rate tuples under the classical unit - delay assumption . the capacity region result is then generalized to the mmn consisting of independent additive white gaussian noise ( awgn ) channels with zero - delay edges . the mmn with independent dmcs has been investigated in @xcite under the _ node - delay _ model proposed in @xcite for general discrete memoryless networks .
the precise definition of zero - delay nodes in ( * ? ? ?
iv ) can be expressed under our edge - delay model as follows : a node _ incurs no delay _ if and only if every incoming edge of the node incurs no delay on every outgoing edge of the node . in @xcite
, it was shown that the capacity region of the mmn with independent dmcs is equal to that under the classical unit - node - delay assumption even when zero - delay nodes are present . under the node - delay model ,
all the outgoing edges of each zero - delay node must be operated simultaneously after all the incoming edges of the node have been operated ( cf .
* definitions 1 and 2 ) ) while under our edge - delay model , the incoming and outgoing edges of a node can be operated in some predetermined order ( cf . definition [ deforderedpartition ] , [ defdiscretenetwork ] and [ defchanneloperationsequence ] ) . for example
, for the trn - cn described in section [ subsectrn ] , edge @xmath28 can be operated before edge @xmath27 under our edge - delay model but not under the node - delay model . in this work ,
we strengthen the main result in @xcite under our edge - delay model for the mmn with independent dmcs and show that the capacity region remains unchanged even when zero - delay edges are present .
it was shown by effros @xcite that under the positive - delay assumption is a function of @xmath38 . ] in the classical setting , the set of achievable rate tuples for the mmn with independent channels does not depend on the amount of positive delay incurred by each edge on each other edge .
our capacity result for the mmn with independent dmcs ( as well as awgns ) complements effros s finding as follows : the set of achievable rate tuples for the mmn with independent dmcs ( as well as awgns ) does not depend on the amount of delay incurred by each edge on each other edge , even with the presence of zero - delay edges . from a practical point of view , the capacity region of the mmn with independent dmcs ( as well as awgns ) is not affected by the way of handling delays among the channels or how the channels are synchronized , even when zero - delay edges are present .
the rest of the paper is organized as follows .
section [ notation ] presents the notation .
section [ sectionformulation ] formulates our edge - delay model for the dm - mmn with independent channels and zero - delay edges and state our cut - set bound result , whose proof is contained in section [ sectioninnerouterbound ] . in section [ sectionmmnwithindepdmcs ] , we use our cut - set bound to prove the capacity region of the mmn with independent dmcs and zero - delay edges .
section [ sectiongaussianchannels ] generalizes our cut - set bound result to the mmn consisting of independent awgn channels with zero - delay edges and characterizes its capacity region .
section [ sectionconclusion ] concludes this paper .
the sets of natural , real and non - negative real numbers are denoted by @xmath39 , @xmath40 and @xmath41 respectively .
we use @xmath42 to denote @xmath43 .
we will take all logarithms to base 2 .
we use @xmath44 to represent the probability of an event @xmath45 .
we use an upper case letter @xmath46 to denote a random variable with alphabet @xmath47 , and use a lower case letter @xmath48 to denote the realization of @xmath46 .
we use @xmath49 to denote a random tuple @xmath50 , where the components @xmath51 have the same alphabet @xmath47 . for any arbitrary random variables
@xmath46 and @xmath52 , we let @xmath53 and @xmath54 denote the probability distribution of @xmath55 ( can be both discrete , both continuous or one discrete and one continuous ) and the conditional probability distribution of @xmath52 given @xmath46 respectively .
we let @xmath56 and @xmath57 be the evaluations of @xmath53 and @xmath54 respectively at @xmath58 . to avoid confusion ,
we do not write @xmath59 to represent @xmath56 unless @xmath46 and @xmath52 are both discrete . we let @xmath60 denote the joint distribution of @xmath55 , i.e. , @xmath61 for all @xmath48 and @xmath62 .
if @xmath46 and @xmath52 are independent , their joint distribution is simply @xmath63 .
we let @xmath64 denote the probability density function of a gaussian random variable whose mean and variance are @xmath65 and @xmath66 respectively , i.e. , @xmath67 . for any
random tuple @xmath68 distributed according to @xmath69 , we let @xmath70 and @xmath71 be the entropy of @xmath46 given @xmath72 and mutual information between @xmath46 and @xmath52 given @xmath72 respectively . if @xmath46 is continuous , we let @xmath73 be the differential entropy of @xmath46 given @xmath72 . if @xmath46 , @xmath52 and @xmath72 are distributed according to @xmath69 and they form a markov chain , we write @xmath74 . for simplicity
, we drop the subscript of a notation if there is no ambiguity .
the @xmath75-distance between two distributions @xmath76 and @xmath77 on the same discrete alphabet @xmath47 , denoted by @xmath78 , is defined as @xmath79 .
a multimessage multicast network ( mmn ) consists of @xmath80 nodes and @xmath81 directed edges . the mmn may contain cycles .
let @xmath82 be the index set of the nodes , and let @xmath83 be the index set of the edges .
let @xmath84 and @xmath85 be the sets of sources and destinations respectively , where each source in @xmath86 transmits one message and each destination in @xmath87 wants to decode all the messages transmitted by the sources in @xmath86 .
we call @xmath88 the _ multicast demand _ for the network .
the sources in @xmath86 transmit information to the destinations in @xmath87 in @xmath89 time slots ( channel uses ) as follows .
node @xmath3 transmits message @xmath90 for each @xmath91 and node @xmath20 wants to decode @xmath92 for each @xmath93 , where @xmath94 denotes the size of @xmath95 .
we assume that each message @xmath96 is uniformly distributed over @xmath97 and all the messages are independent . for each @xmath98 and each @xmath99
, node @xmath3 transmits @xmath100 and receives @xmath101 in the @xmath102 time slot , where @xmath103 is the symbol transmitted on edge @xmath104 by node @xmath3 and @xmath105 is the symbol received from edge @xmath106 by node @xmath20 .
the alphabet sets of @xmath107 and @xmath108 are denoted by @xmath109 and @xmath110 respectively for each @xmath99 , where @xmath109 and @xmath110 do not depend on the time index @xmath111 .
after @xmath89 time slots , node @xmath20 declares @xmath112 to be the transmitted @xmath96 based on @xmath113 for each @xmath114 .
to simplify notation , we use the following conventions for each @xmath115 : we let @xmath116 be the subtuple of @xmath117 , and let @xmath118 be the subtuple of @xmath119 .
for any @xmath81-dimensional random tuple @xmath120 , we let @xmath121 be its subtuple . similarly , for any @xmath98 and any random tuple @xmath122 , we let @xmath123 be its subtuple . [ deforderedpartition ] an @xmath124-dimensional tuple @xmath125 consisting of non - empty subsets of @xmath126 is called an _
@xmath124-partition of @xmath126 _ if @xmath127 and @xmath128 for all @xmath129 .
[ defdiscretenetwork ] the discrete network with independent channels consists of @xmath81 finite input sets @xmath130 , @xmath81 finite output sets @xmath131 and @xmath124 channels characterized by conditional distributions @xmath132 , @xmath133 , where @xmath134 is an @xmath124-partition of @xmath126 .
we call @xmath135 the _ edge partition _ of the network
. the discrete network is denoted by @xmath136 where @xmath137 . under our model ,
the discrete network is characterized by @xmath124 channels denoted by @xmath132 , @xmath138 as defined in definition [ defdiscretenetwork ] .
if @xmath139 , our model simplifies to the classical model which is characterized by only one channel denoted by @xmath140 .
the following example illustrates our model when @xmath141 .
[ example1 ] for the bsc - if described in section [ sectionmotivating ] , we have @xmath142 and @xmath143 , where @xmath144 . without loss of generality , we assume that @xmath145
. the edge partition is denoted by @xmath146 where @xmath147 , @xmath148 and @xmath149 . since @xmath150
, @xmath151 is a trivial channel which can carry no information .
the other two channels of the bsc - if are denoted by @xmath152 and @xmath153 respectively , which correspond to the bsc and the dmc in figure [ bscfb](b ) respectively .
@xmath154 [ defdelayprofile ] an _ edge - delay profile _ , also called _ delay profile _ for simplicity , is an @xmath155-dimensional tuple @xmath156 where @xmath157 represents the amount of delay incurred by edge @xmath0 on edge @xmath1 .
the delay profile is said to be _ positive _ if it is the all - one tuple @xmath158 . under the classical model , only the positive delay profile
is considered , i.e. , every edge incurs a unit delay on all the edges . in our model
, some elements of the delay profile can be zero , which indicates that some edge can incur zero delay on some other edges .
however , if the delay profile contains too many zeros , deadlock loops may occur which result in each node waiting for the other nodes to transmit first before encoding and transmitting its outgoing symbols .
the following two definitions enable us to formally define good " delay profiles that do not induce deadlock loops .
[ defchanneloperationsequence ] a _ channel operation sequence _ for the discrete network @xmath159 is some permutation of the tuple @xmath160 .
the set of all channel operation sequences ( permutations of @xmath161 ) is denoted by @xmath162 .
when we formally define a code on the discrete network later , a channel operation sequence @xmath163 together with a delay profile @xmath164 will be associated with the code , where the @xmath124 channels are operated in the order @xmath165 and @xmath166 represents the amount of delay incurred by edge @xmath0 on edge @xmath1 .
if @xmath167 , then node @xmath3 receives @xmath168 _ before _ encoding @xmath107 , and we say @xmath2 _ incurs zero delay _ on @xmath1 ; if @xmath169 , then node @xmath3 receives @xmath168 _ after _ encoding @xmath107 , and we say @xmath2 _ incurs a delay _ on @xmath1 . under the classical model , @xmath139 and the channel operation sequence can only be the tuple @xmath170 ( because there is only one channel ) , which implies that @xmath171 can only be positive , meaning that the amount of delay incurred between every pair of edges can only be positive .
in contrast , under our edge - delay model , the @xmath124 channels can be operated in different orders and some elements of @xmath171 can take @xmath30 as long as deadlock loops do not occur ( in a deadlock loop , every node wants to receive first and then encode , and hence no transmission can take place ) .
therefore our model is a generalization of the classical model .
the essence of the following definition is to characterize delay profiles which will not induce deadlock loops in the network .
[ deffeasible ] let @xmath159 be a discrete network , and let @xmath172 be a channel operation sequence . for each @xmath99 ,
let @xmath173 be the unique integer such that @xmath174 .
then , an edge - delay profile @xmath175 is said to be _ feasible with respect to @xmath176 _ if the following holds for each @xmath177 : if @xmath167 , then @xmath178 . under the classical model ,
the only possible delay profile is the positive delay profile , which is feasible with respect to any channel operation sequence according to definition [ deffeasible ] .
the following example illustrates a delay profile for the bsc - if which is feasible with respect to one channel operation sequence but not to another channel operation sequence .
[ example2 ] for the bsc - if as defined in example [ example1 ] , @xmath141 because it is characterized by three channels . since the third channel is trivial , we assume without loss of generality that it is always the last channel to be operated . then , the two channel operation sequences of our interest are @xmath179 and @xmath180 .
let @xmath181 for all @xmath177 , and let @xmath182 be the delay profile where only @xmath183 is @xmath30 .
then , it follows from definition [ deffeasible ] that @xmath184 is _ feasible with respect to the channel operation sequence @xmath179 _ but @xmath184 is _ not feasible with respect to @xmath180_. an interpretation for the feasibility of @xmath184 is given as follows : if the channel is operated according to the sequence @xmath179 , i.e. , @xmath185 is operated before @xmath186 , then @xmath11 will be generated before the activation of @xmath186 .
therefore , node @xmath5 can receive @xmath11 before encoding @xmath9 , and we say @xmath184 is feasible with respect to @xmath179 . if the channel is operated according to the sequence @xmath180 , then node @xmath5 will receive @xmath11 after the encoding of @xmath9 , and we say @xmath184 is not feasible with respect to @xmath180 .
we are ready to define codes that use the network @xmath89 times .
[ defcode ] let @xmath159 be a discrete network .
let @xmath88 be the multicast demand for the network .
in addition , let @xmath187 be a channel operation sequence , and let @xmath188 be a feasible delay profile with respect to @xmath176 .
a @xmath189-code , where @xmath190 , for @xmath89 uses of the network consists of the following : 1 .
a message set @xmath191 at node @xmath3 for each @xmath192 , where @xmath193 for each @xmath194 .
message @xmath95 is uniformly distributed on @xmath195 .
an encoding function @xmath196 for each @xmath99 and each @xmath197 , where @xmath198 is the encoding function for @xmath107 at node @xmath3 in the @xmath102 time slot such that @xmath199 .
3 . a decoding function @xmath200 for each @xmath201 , where @xmath202 is the decoding function for @xmath96 at node @xmath20 such that @xmath203 . given a @xmath189-code , it follows from definition [ defcode ] that for each @xmath204 , edge @xmath2 incurs a delay on edge @xmath1 if @xmath205 .
if @xmath206 , edge @xmath2 incurs zero delay on edge @xmath1 , i.e. , for each @xmath98 , node @xmath3 receives @xmath168 before encoding @xmath107 .
the feasibility condition of @xmath171 in definition [ deffeasible ] ensures that the operations of any @xmath207-code are well - defined for the subsequently defined discrete memoryless network ; the associated coding scheme is described after the network is defined .
[ defdiscretememoryless ] a discrete network @xmath159 , when used multiple times , is called a _ discrete memoryless multimessage multicast network ( dm - mmn ) _ if the following holds for any channel operation sequence @xmath208 and any @xmath189-code : let @xmath209 be the collection of random variables that are generated before the @xmath102 time slot . to simplify notation ,
let @xmath210 then , for each @xmath197 and each @xmath211 , @xmath212 for all @xmath213 , @xmath214 and @xmath215 .
following the notation in definition [ defdiscretememoryless ] , consider any @xmath189-code on the dm - mmn .
in the @xmath102 time slot , @xmath216 and @xmath217 are generated in the order @xmath218 by transmitting on the channels in this order @xmath219 using the @xmath207-code ( as prescribed in definition [ defcode ] ) .
specifically , @xmath220 and channel @xmath221 together define @xmath222 for each @xmath223 .
in addition , the feasibility condition of @xmath171 in definition [ deffeasible ] ensures that the encoding of @xmath220 specified in definition [ defcode ] is well - defined because @xmath220 depends on only the symbols generated before its encoding , which is formally shown in the following proposition .
[ propositionxfunctionofy ] let @xmath172 be a channel operation sequence .
fix any @xmath189-code and fix an @xmath223 . then , for each @xmath224 , @xmath107 is a function of @xmath225 for each @xmath197 .
let @xmath226 be a delay profile that is feasible with respect to @xmath176 and fix a @xmath189-code .
fix a @xmath197 and an edge @xmath224 . by definition [ defcode ]
, @xmath107 is a function of @xmath227 .
therefore , it suffices to show that @xmath228 is a function of @xmath229 .
since each @xmath166 is binary , it remains to prove the following statement : for each @xmath230 such that @xmath231 , @xmath232 is a function of @xmath233 . to this end
, we fix an @xmath230 that satisfies @xmath231 .
let @xmath234 and @xmath173 be the two unique integers such that @xmath235 and @xmath236 .
it then follows from definition [ deffeasible ] that @xmath237 , which implies that @xmath238 , which then implies that @xmath232 is a function of @xmath233 . under the classical model , @xmath139
( there is only one channel ) and @xmath216 and @xmath217 are generated in the order @xmath239 in each time slot @xmath111 .
the only channel operation sequence is @xmath170 and it follows from definition [ deffeasible ] the only feasible delay profile with respect to @xmath170 is the positive delay profile , which implies that every edge incurs a unit delay on all the edges .
consequently , the classical model for the dm - mmn is a special case of our edge - delay model when @xmath139 .
we are ready to define the capacity region of the dm - mmn using the following three definitions .
[ deferrorprobability ] for a @xmath189-code , let @xmath240 be the probability of decoding error .
[ defachievablerate ] let @xmath176 be a channel operation sequence and @xmath171 be a feasible delay profile with respect to @xmath176 .
a rate tuple @xmath241 , denoted by @xmath242 , is _
@xmath243-achievable _ for the dm - mmn if there exists a sequence of @xmath207-codes such that @xmath244 for each @xmath245 and @xmath246 .
a rate tuple is said to be _
@xmath176-achievable _ if it is @xmath243-achievable for some @xmath171 .
a rate tuple is said to be _ achievable _ if it is @xmath176-achievable for some @xmath176 . without loss of generality , we assume that @xmath247 and @xmath248 for all @xmath194 in the rest of this paper . [ defcapacityregion ]
the _ @xmath243-capacity region _ , denoted by @xmath249 , of the dm - mmn is the set consisting of every @xmath243-achievable rate tuple .
the _ @xmath176-capacity region _
@xmath250 is defined as @xmath251 the _ capacity region _
@xmath252 is defined as @xmath253 the following theorem is our main result .
[ thmcapacityregionmulticast ] let @xmath159 be a dm - mmn with independent channels and zero - delay edges under multicast demand @xmath88 , and let @xmath254{2.5in}{$\sum\limits_{i\in t}r_i \le i_{p_{x_{\mathcal{e } } , y_{\mathcal{e } } } } ( x_{t\times \mathcal{i}};y_{\mathcal{i}\times t^c}|x_{t^c\times \mathcal{i}})$ } \right . \right\}\label{rout}\end{aligned}\ ] ] be the classical cut - set bound for the dm - mmn with independent channels .
then for any channel operation sequence @xmath176 , we have @xmath255 hence , @xmath256 [ remark1 ] theorem [ thmcapacityregionmulticast ] subsumes the classical cut - set bound for @xmath139 , i.e. , @xmath257{2.7 in}{$\sum\limits_{i\in t}r_i \le i_{p_{x_{\mathcal{e}}}q^{(1)}_{y_{\mathcal{e } } | x_{\mathcal{e } } } } ( x_{t\times \mathcal{i}};y_{\mathcal{i}\times t^c}|x_{t^c\times \mathcal{i}})$}\!\ ! \right .
\right\}.\ ] ] [ remark2 ] recall that the bsc - if can be viewed as a multicast network as explained in section [ introduction ] . using the formulation of bsc - if in examples [ example1 ] and [ example2 ]
, we obtain by theorem [ thmcapacityregionmulticast ] that @xmath258 is bounded above by the capacity of the bsc and @xmath259 is bounded above by the capacity of the dmc regardless of what order the two channels are operated .
[ remark3 ] the cut - set bound in theorem [ thmcapacityregionmulticast ] is tight for the mmn consisting of independent dmcs @xcite .
this will be shown in section [ sectionmmnwithindepdmcs ] after the proof of theorem [ thmcapacityregionmulticast ] is presented in the next section .
[ remark4 ] the cut - set bound in theorem [ thmcapacityregionmulticast ] can easily be generalized for ( i ) any multiple multicast demand where each source multicasts a single message and the destinations want to decode different subsets of the messages ; ( ii ) any multiple unicast demand where each node transmits @xmath260 independent messages to the other @xmath260 nodes and each message is decoded by one node only .
fix a channel operation sequence @xmath176 .
let @xmath242 be a @xmath176-achievable rate tuple for the dm - mmn denoted by @xmath159 . by definitions [ defachievablerate ] and [ defcapacityregion ] , there exists a sequence of @xmath189-codes such that @xmath261 for each @xmath245 and @xmath262 since @xmath176 is a permutation of @xmath161 by definition [ defchanneloperationsequence ] and @xmath263 by definition [ deforderedpartition ] , it follows that @xmath264 fix @xmath89 and the corresponding @xmath189-code , and let @xmath265 be the probability distribution induced by the code . fix any @xmath266 such that @xmath267 , and let @xmath268 denote a node in @xmath269 . for the @xmath189-code , since the @xmath80 messages @xmath270 are independent , we have @xmath271 where the last inequality follows from fano s inequality ( cf . definition [ deferrorprobability ] ) .
following and omitting the subscripts for the entropy and mutual information terms , we consider @xmath272 where ( a ) follows from and .
following , we consider @xmath273 and @xmath274 for each @xmath197 and each @xmath275 that satisfies @xmath276 , where 1 .
follow from the fact by proposition [ propositionxfunctionofy ] that @xmath277 is a function of @xmath278 .
2 . is due to the following markov chain implied by in definition [ defdiscretememoryless ] : @xmath279 combining , , and and using the fact that @xmath176 is a permutation of @xmath161 , we obtain @xmath280 recalling that @xmath265 is the probability distribution induced by the code , we define the time - sharing distribution @xmath281 for all @xmath197 . in addition , we define @xmath282 for all @xmath197 , @xmath283 and @xmath284 . following , we consider the following chain of inequalities for each @xmath223 : @xmath285 where 1
. is due to the following fact by : @xmath286 for each @xmath111 , @xmath287 and @xmath288 .
2 . is due to the following markov chain implied by : @xmath289 combining and , we obtain @xmath290 where for each @xmath223 , @xmath291 is a distribution on @xmath292 that satisfies @xmath293 let @xmath294 be a subsequence of @xmath295 such that @xmath296 converges with respect to the @xmath75-distance for all @xmath223 , and we define the limit of @xmath296 as @xmath297 for all @xmath298 and all @xmath223 . since @xmath299 is a continuous functional of @xmath300 for each @xmath223 , we obtain from , , and that @xmath301 since @xmath302 for all @xmath303 by and , it follows from that @xmath304 define @xmath305 it then follows from that @xmath306 in order to simplify , we consider @xmath307 where ( a ) follows from the fact by that @xmath308 and @xmath309 are independent for all @xmath310 under the distribution @xmath311 . combining and , we obtain @xmath312 since @xmath311 satisfies and depends on only the sequence of @xmath207-codes but not on @xmath313 , holds for all @xmath266 that satisfies @xmath267 , which implies from and that @xmath314 .
this completes the proof .
the mmn consisting of independent dmcs @xcite consists of @xmath81 channels where each channel is associated with a directed edge .
define @xmath315 for each @xmath316 such that @xmath317 characterizes the dmc associated with edge @xmath1 .
then , this network can be viewed as a discrete network @xmath318 according to definition [ defdiscretenetwork ] , where @xmath319 and @xmath320 are defined as @xmath321 and @xmath322 respectively .
we assume that this network satisfies the discrete memoryless property stated in definition [ defdiscretememoryless ] .
let @xmath252 denote the capacity region of this network . to simplify notation
, we let @xmath323 and let @xmath324 be the capacity of channel @xmath325 for each @xmath326 . combining existing results with theorem [ thmcapacityregionmulticast ] , we fully characterize the capacity region of the mmn with independent dmcs and zero - delay edges in the following theorem .
[ thmmmnwithinddmcs ] for the mmn consisting of independent dmcs with zero - delay edges under multicast demand @xmath88 , we have @xmath327{1.6 in}{$\sum\limits_{i\in t}r_i \le \sum\limits_{(i , j)\in t\times t^c } c_{(i , j)}$ } \right . \right\},\ ] ] where @xmath252 and @xmath328 denote the capacity region and the @xmath329-capacity region respectively .
it was shown by effros @xcite that under the positive - delay assumption in the classical setting , the set of achievable rate tuples for the mmn with independent dmcs does not depend on the amount of positive delay incurred by each edge on each other edge .
therefore , theorem [ thmmmnwithinddmcs ] together with effros s result implies that the set of achievable rate tuples for the mmn with independent dmcs does not depend on the amount of delay incurred by each edge on each other edge even when zero - delay edges are present .
the achievability and converse proofs of theorem [ thmmmnwithinddmcs ] are presented in the next two subsections respectively .
define @xmath330{1.75 in}{$\sum_{i\in t}r_i \le \sum\limits_{(i , j)\in t\times t^c } c_{(i , j)}$ } \right . \right\}. \label{defr1ach}\ ] ] our goal is to prove @xmath331 the proof combines existing results from network equivalence theory in @xcite and noisy network coding ( nnc ) in @xcite . consider a deterministic counterpart of the mmn consisting of independent dmcs by replacing every dmc @xmath325 in the mmn with a noiseless bit pipe whose capacity is equal to @xmath332 ( cf . , and ) , and let @xmath333 be the capacity region of the deterministic counterpart network _ with unit - delay edges _ where @xmath329 denotes the all - one delay profile . letting @xmath328 denote the @xmath329-capacity region of the mmn consisting of independent dmcs , we conclude by using the network equivalence theory @xcite that @xmath334 , which then implies from definition [ defcapacityregion ] that @xmath335 by viewing @xmath336 as the cut - set bound for the deterministic counterpart network , it follows from the nnc inner bound in ( * ? ? ?
ii - a ) that @xmath337 , which then implies from that holds . we will prove @xmath338 by using theorem [ thmcapacityregionmulticast ] .
suppose @xmath339 is achievable .
it then follows from theorem [ thmcapacityregionmulticast ] , and that there exists a distribution @xmath340 such that for all @xmath341 that satisfies @xmath267 , we have @xmath342 consider the following chain of inequalities for each @xmath341 : @xmath343 where 1
. follows from the fact by that @xmath344 forms a markov chain for all @xmath99 .
2 . follows from the fact that for all @xmath345 , @xmath346 combining , and , we have @xmath347 .
this completes the proof of .
in this section , we generalize the result in the previous section to multicast networks consisting of independent additive white gaussian noise ( awgn ) channels with zero - delay edges .
the _ mmn consisting of independent awgn channels _ consists of @xmath81 channels , where the channel associated with edge @xmath1 is an awgn channel whose noise variance is denoted by @xmath348 for each @xmath99 . in each time slot @xmath111 , node @xmath3 transmits @xmath349 on edge @xmath1 and receives @xmath350 from edge @xmath0 .
in addition , we assume that the channel outputs are independent given their inputs , i.e. , the channel noises are independent .
each codeword transmitted on @xmath1 is subject to the power constraint @xmath351 for each @xmath99 , where @xmath352 denotes the average power available to edge @xmath1 .
similarly , each node @xmath3 is subject to the power constraint @xmath353 for each @xmath192 , where @xmath354 denotes the average power available to node @xmath3 .
the network is subject to the power constraint @xmath355 where @xmath356 denotes the total average power available for transmissions in the network . to facilitate discussion ,
we write @xmath357 and @xmath358 .
[ defmmnwithawgn ] the mmn with awgn channels , denoted by a positive - valued tuple @xmath359 , consists of @xmath81 awgn channels denoted by @xmath360 , where @xmath325 characterizes the awgn channel associated with the directed edge @xmath1 such that @xmath361 for all @xmath362 . in definition [ defmmnwithawgn ]
, we have not introduced the independence property among the channels yet .
we will define the independence among the awgn channels after we have specified codes that use the network multiple times . to facilitate discussion
, we define @xmath363 for each @xmath364 such that @xmath365 and @xmath366 for all @xmath367 . in other words
, @xmath368 is an @xmath81-partition of @xmath126 . as in definition
[ defchanneloperationsequence ] , we define a channel operation sequence @xmath176 to be a permutation of @xmath369 , and define @xmath162 to be the set of the permutations of @xmath369 . for a channel operation sequence @xmath176 and each delay profile @xmath171 that is feasible with respect to @xmath176 ( cf . definition [ deffeasible ] ) , we define a @xmath370-code similar to definition [ defcode ] with the additional power constraints , and .
we are ready to formally define the mmn with _
independent _ awgn channels . a mmn with awgn channels @xmath359 is called a _
mmn with independent awgn channels _ if the following holds for any channel operation sequence @xmath371 and any @xmath370-code : let @xmath209 be the collection of random variables that are generated before the @xmath102 time slot . to simplify notation ,
let @xmath372 and let @xmath373 be the unique edge such that @xmath374 then , for each @xmath197 and each @xmath211 , we have @xmath375 where @xmath376 for all @xmath377 ( channel @xmath378 was defined in ) . given a multicast demand @xmath88 , we define the _ @xmath243-achievability _ , the _ @xmath176-achievability _ and the of a rate tuple @xmath339 as in definition [ defachievablerate ] .
then , we define the _ @xmath243-capacity region _ denoted by @xmath249 , the _ @xmath176-capacity region _ denoted by @xmath250 and the capacity region @xmath379 as in definition [ defcapacityregion ] .
the following theorem fully characterizes the capacity region of the mmn with independent awgns and zero - delay edges .
the achievability and converse proofs of the theorem will be provided in the following two subsections respectively .
[ thmcapacitymmnwithawgn ] let @xmath359 be a mmn with independent awgns and zero - delay edges under multicast demand @xmath88 , and let @xmath380{2.1 in}{$ \sum_{i=1}^n\sum_{j=1}^n s_{(i , j)}\le p,\\ \sum_{j=1}^n s_{(i , j ) } \le p_i \text { for all $ i\in \mathcal{i}$ , } \\ s_{(i , j ) } \le p_{(i , j ) } \text { for all $ ( i , j)\in \mathcal{e}$ } $ } \right.\right\}\label{defsets}\ ] ] be a set of @xmath81-dimensional tuples which specify the power allocation for the edges in the network .
let @xmath381 denote @xmath382 and define @xmath383{2.3 in}{$\sum\limits_{i\in t}r_i \le \sum\limits_{(i , j)\in t\times t^c } \frac{1}{2}\log\big(1+\frac{s_{(i , j)}}{\sigma_{(i , j)}^2}\big)$ } \right . \right\}\label{routawgn}\end{aligned}\ ] ] to be the classical cut - set bound .
then , we have @xmath384 where @xmath252 and @xmath328 denote the capacity region and the @xmath329-capacity region respectively .
theorem [ thmcapacitymmnwithawgn ] generalizes the capacity result in theorem [ thmmmnwithinddmcs ] to the mmn consisting of independent awgns with zero - delay edges .
although the extra power constraints in , and are introduced in the awgn channels setting compared with the dmcs setting , we can still show in theorem [ thmcapacitymmnwithawgn ] that the set of achievable rate tuples for the mmn with independent awgns does not depend on the amount of delay incurred by each edge on each other edge even when zero - delay edges are present .
our goal is to prove @xmath385 since @xmath386 by , it suffices to show @xmath387 , which is equivalent to the following statement by : @xmath388{2.6 in}{$\sum\limits_{i\in t}r_i \le \sum\limits_{(i , j)\in t\times t^c } \frac{1}{2}\log\big(1+\frac{(s_{(i , j)}^*-\delta)_+}{\sigma_{(i , j)}^2}\big)$ } \right . \right\ } \label{goalinachawgn}\ ] ] holds for all @xmath389 and all @xmath390 . in order to show , we fix an @xmath389 and a @xmath390 .
suppose we use a random gaussian codebook with power @xmath391 for each edge @xmath99 so that the rate @xmath392 can be achieved for the awgn @xmath325 ( defined in ) as @xmath89 tends to infinity .
in addition , the power constraints , and hold with probability approaching @xmath4 by the weak law of large numbers due to the random gaussian codebooks and the fact that @xmath393 ( cf . ) . in the rest of the proof , we follow the network equivalence and nnc arguments in section [ subsectionachinddmcs ] which have been used for the achievability proof of theorem [ thmmmnwithinddmcs ] . consider a deterministic counterpart of the mmn consisting of independent awgns by replacing every awgn @xmath325 with a noiseless bit pipe whose capacity is equal to @xmath394 ( cf . ) , and let @xmath333 be the capacity region of the deterministic counterpart network _ with unit - delay edges _ where @xmath329 denotes the all - one delay profile .
letting @xmath328 denote the @xmath329-capacity region of the mmn consisting of independent awgns , we conclude by using the network equivalence theory @xcite that @xmath395 where ( a ) is not an equality because the capacity of every awgn @xmath325 is @xmath396 rather than @xmath394 .
define @xmath330{2.1 in}{$\sum\limits_{i\in t}r_i \le \sum\limits_{(i , j)\in t\times
t^c } \mathrm{c}(s_{(i , j)}^*-\delta)$ } \right . \right\ } \label{defr1achawgn}\ ] ] to be the cut - set outer bound for the deterministic counterpart network .
it follows from the nnc inner bound in ( * ? ? ?
ii - a ) that @xmath337 , which implies from that @xmath397 holds , which then implies from and that holds .
since holds for all @xmath389 and all @xmath390 , also holds .
our goal is to prove @xmath398 it suffices to show that @xmath399 holds for all channel operation sequence @xmath176 , which will then imply from that holds .
fix a channel operation sequence @xmath176 . in order to show
, we fix an arbitrary delay profile @xmath171 that is feasible with respect to @xmath176 and let @xmath339 be a @xmath243-achievable rate tuple . by definition
, there exists a sequence of @xmath400-codes such that @xmath401 for each @xmath245 and @xmath402 where @xmath403 is as defined in .
fix @xmath89 and the corresponding @xmath404-code , and let @xmath265 be the probability distribution induced by the code .
fix any @xmath266 such that @xmath267 .
following similar procedures for deriving in section [ sectioninnerouterbound ] , we obtain from fano s inequality , proposition [ propositionxfunctionofy ] and that @xmath405 following , we consider the following chain of equalities for each @xmath197 : @xmath406 where 1
. follows from the facts that @xmath407 is an @xmath81 partition of @xmath126 and each @xmath408 contains exactly one edge .
2 . follows from the fact that for each @xmath409 , @xmath410 let @xmath411 be the distribution of the zero - mean gaussian random variable @xmath412 whose variance is @xmath348 for each @xmath99 and each @xmath413 .
it then follows from and that @xmath108 has the same distribution as @xmath414 when @xmath108 and @xmath415 are distributed according to @xmath416 and @xmath417 respectively for each @xmath99 and each @xmath413 , which implies that @xmath418 following , we consider @xmath419+\var_{p_{z_{(i , j),k}}}\left[z_{(i , j),k}\right]\right ) } \notag\\ & \le \log \sqrt{2\pi e \left(\e_{p_{x_{(i , j),k}}}\left[x_{(i , j),k}^2\right]+\var_{p_{z_{(i , j),k}}}\left[z_{(i , j),k}\right]\right ) } \notag\\ & = \log \sqrt{2\pi e \left(\e_{p_{x_{(i , j),k}}}\left[x_{(i , j),k}^2\right]+\sigma_{(i , j)}^2\right ) } \label{cutsetstatement7awgn } % & \le \log \sqrt{2\pi e \left(\e_{p_{x_{(i , j),k}}}\left[x_{(i , j),k}^2\right]+\sigma_{(i , j)^2}\right ) } \end{aligned}\ ] ] and @xmath420}\notag\\ & = \log \sqrt{2\pi e \sigma_{(i , j)}^2 } \label{cutsetstatement8awgn } \end{aligned}\ ] ] for each @xmath99 and @xmath197 , where ( a ) follows from the fact that the differential entropy of a random variable @xmath46 is maximized by that of the zero - mean gaussian random variable whose variance is @xmath421 $ ] .
combining , , , and , we have @xmath422}{\sigma_{(i , j)}^2}\right ) \notag\\ & \stackrel{\text{(a)}}{\le } \sum_{(i , j)\in t\times t^c } \frac{1}{2}\log\left(1+\frac{\e_{p_{x_{(i , j)}^n}}\left[\frac{1}{n}\sum_{k=1}^n x_{(i , j),k}^2\right]}{\sigma_{(i , j)}^2}\right ) \label{cutsetstatement9awgn}\end{aligned}\ ] ] where ( a ) follows from jensen s inequality .
choose @xmath423 to be a subsequence of @xmath424 such that for all @xmath99 , @xmath425 $ ] converges to some @xmath426 , i.e. , @xmath427 . \label{cutsetstatement10awgn}\end{aligned}\ ] ] combining , , and , we obtain @xmath428 define @xmath429 to be an @xmath81 dimensional - tuple .
since @xmath430 does not depend on @xmath313 by , it follows from that @xmath431{2.3 in}{$\sum\limits_{i\in t}r_i \le \sum\limits_{(i , j)\in t\times t^c } \frac{1}{2}\log\big(1+\frac{s_{(i , j)}^*}{\sigma_{(i , j)}^2}\big)$ } \right . \right\}.\label{cutsetstatement12awgn } \end{aligned}\ ] ] on the other hand , we conclude from , , and that @xmath432 for each @xmath99 , @xmath433 for each @xmath192 , and @xmath434 combining , , and , we have @xmath435 consequently , follows from and .
since holds for all channel operation sequence @xmath176 , also holds .
we investigate the mmn consisting of independent channels and propose an edge - delay model which allows the presence of zero - delay edges in the network . under our model ,
the mmn with independent channels is characterized by multiple channels , which are operated in different orders so that an edge may incur zero delay on some other edges .
our model is a generalization of the classical model , under which the mmn with independent channels is characterized by a single channel and every edge incurs a unit delay on every adjacent edge .
a well - known outer bound on the capacity region under the classical unit - delay assumption is the cut - set outer bound . in this paper
, we prove that the mmn with independent channels and zero - delay edges lies within the classical cut - set bound despite a violation of the classical unit - delay assumption .
next , we use our outer bound to prove the capacity region of the mmn consisting of independent dmcs with zero - delay edges . more specifically
, we show that the capacity region is the same as the set of achievable rate tuples under the classical unit - delay assumption .
this capacity region result is then generalized to the mmn consisting of independent awgn channels with zero - delay edges .
consequently , the capacity regions of the two aforementioned mmns are not affected by the handling of delays among edges even when zero - delay edges are present .
it has been shown in section [ sectionmotivating ] that the capacity region of some two - node mmn with _
dependent _ channels and zero - delay edges is strictly larger than the set of achievable rate tuples under the classical unit - delay assumption every edge should incur a unit delay on all the edges .
future research may investigate the capacity regions for general mmns consisting of _ dependent _ channels with zero - delay edges .
another interesting direction for future research is extending the network equivalence theory @xcite ( which asserts the equivalence between the capacity region of any network with independent channels and the capacity region of the deterministic counterpart of the network ) for networks with unit - delay edges to networks with zero - delay edges . | we consider a communication network consisting of nodes and directed edges that connect the nodes .
the network may contain cycles .
the communications are slotted where the duration of each time slot is equal to the maximum propagation delay experienced by the edges .
the edges with negligible delays are allowed to be operated before the other edges in each time slot . for any pair of adjacent edges
@xmath0 and @xmath1 where @xmath2 terminates at node @xmath3 and @xmath1 originates from node @xmath3 , we say @xmath2 _ incurs zero delay _ on @xmath1 if @xmath2 is operated before @xmath1 ; otherwise , we say @xmath2 _ incurs a unit delay _ on @xmath1 . in the classical model ,
every edge incurs a unit delay on every adjacent edge and the cut - set bound is a well - known outer bound on the capacity region . in this paper , we investigate the multimessage multicast network ( mmn ) consisting of independent channels where each channel is associated with a set of edges and each edge may incur zero delay on some other edges .
our result reveals that the capacity region of the mmn with independent channels and zero - delay edges lies within the classical cut - set bound despite a violation of the unit - delay assumption .
multimessage multicast networks , zero - delay edges , independent channels , cut - set bounds |
using lichtenbaum s weight - two motivic complex @xmath2 ( a two - term complex derived from relative k - theory ) , m. spiess @xcite constructed a complex of tale sheaves @xmath3 ( definition [ definition : spiesscomplex ] ) on arithmetic surfaces @xmath4 over a dedekind domain @xmath5 and used it to prove a duality theorem of constructible sheaves . for @xmath4 over a perfect field @xmath6 of characteristic @xmath7 , t. moser @xcite studied gersten complexes of logarithmic de rham witt sheaves @xmath8 ( [ definition : mosercomplex ] ) and showed that , when @xmath6 is finite , @xmath9 is a dualizing complex for constructible @xmath10-sheaves . for regular semi - stable schemes
@xmath4 over certain dedekind domain @xmath5 ( see condition [ condition * ] ) , k. sato @xcite defined certain dualizing complex @xmath11 ( @xmath12 ) ( definition [ definition : satocomplex ] ) in derived category of tale sheaves and proved a duality theorem for @xmath10-sheaves as well . in more general situations , for instance , schemes over algebraically closed fields , finite fields , local fields and certain dedekind domains , t. geisser @xcite proved that the complex @xmath13 of tale sheaves ( see ( [ definition : blochcomplex ] ) ) is a dualizing complex for constructible sheaves . here
@xmath14 is bloch s cycle complex whose homology defines higher chow groups @xcite . in this paper , we answer the following questions of quasi - isomorphisms of these complexes : for @xmath4 a scheme separated and essentially of finite type over a perfect field @xmath6 of characteristic @xmath7 , and @xmath15 , there is a map @xmath16 which induces a quasi - isomorphism . here
@xmath17 .
let @xmath4 be a surface over a perfect field @xmath6 or a dedekind domain @xmath5 with perfect residue fields .
then @xmath18 is quasi - isomorphic to @xmath3 .
moreover , for any torsion sheaf @xmath19 , @xmath20 let @xmath21 be a prime number , @xmath4 be a scheme over a dedekind domain @xmath5 which satisfies condition ( [ condition * ] ) below . assume that @xmath22 , and the conjecture @xmath23 with @xmath10-coefficients ( conjecture [ conjecture : gersten ] ) holds for all @xmath24
. then @xmath25).\ ] ] let @xmath4 be a scheme of dimension @xmath26 over a perfect field of characteristic @xmath7 .
complex @xmath8 is the gersten complex of logarithmic de rham
witt sheaf @xmath27 , when @xmath4 is smooth . for general @xmath4 ,
@xmath28 is identified with kato s complex @xmath29 of milnor k - groups modulo @xmath0 ( see theorem [ thm : mktolrw ] ) . to prove theorem [ thm : cycletolrw ] , first
, we show that the niveau filtration of higher chow groups induces a canonical map @xmath30 the latter being the global sections of gersten complex of higher chow groups ( tale sheafified ) .
then we show that @xmath31 is isomorphic to @xmath32 , which provides us a map @xmath33 when composing with the isomorphism @xmath34 we obtain a map @xmath35 moreover , using a result of geisser levine ( * ? ? ?
* theorem 1.1 ) , we show that @xmath36 induces an isomorphism of cohomology groups .
hence we conclude that @xmath36 is a quasi - isomorphism . as another application of this method ,
we show that , for smooth and projective varieties over finite fields , the conjecture a(@xmath37 ) of geisser ( part of parshin s conjecture , see ( * ? ? ?
* proposition 2.1 ) ) is true , if and only if @xmath38 is a quasi - isomorphism .
let @xmath4 be a two dimensional scheme over a perfect field or a dedekind domain with perfect residue fields .
spiess complex @xmath3 is defined by connecting the complex @xmath39 with lichtenbaum s weight - two motivic complex @xmath40 of the function field @xmath41 ( see ( * ? ? ?
* definition 2.1 ) for @xmath42 ) . to compare it with bloch s complex
, we first define an intermediate complex @xmath43 . using the niveau spectral sequence ,
we show that @xmath43 is quasi - isomorphic to @xmath44 .
then we show that @xmath43 is quasi - isomorphic to @xmath45 .
a key ingredient in this step is the quasi - isomorphism in @xcite between a truncation of bloch s complex of @xmath41 and @xmath46 .
based on the above , we conclude that @xmath18 is quasi - isomorphic to @xmath3 .
let @xmath4 be a scheme over a dedekind domain satisfying condition ( [ condition * ] ) , and @xmath47 .
complex @xmath11 is defined to be a complex in the derived category of tale sheaves satisfying a distinguished triangle , with the other two terms the logarithmic de rham - witt sheaf on points of characteristic @xmath21 and @xmath48 on the open complement . in order to compare @xmath11 with @xmath49
$ ] , we show that @xmath49 $ ] satisfies a similar triangle ( the localization sequence of bloch s cycle complex ) , and that there is a map between the two triangles , which induces isomorphisms on the cohomologies up to degree @xmath37 .
the key point is to show that these isomorphisms are compatible .
using the gersten complexes of @xmath50 , cycle complex and logarithmic de rham witt sheaves , we replace maps of cohomology of schemes by those of function fields , whose compatibility is straightforward .
hence the two complexes are quasi - isomorphic up to degree @xmath37 . with the aid of the conjecture @xmath23 with @xmath10-coefficients , then they are actually quasi - isomorphic .
the paper is organized as follows : in section 2 , we recall the definitions of bloch s cycle complex @xmath14 , kato s complex @xmath51 of milnor k - theory and moser s complex @xmath52 , as well as the duality results of geisser and moser .
we also recall the construction of the niveau spectral sequence of higher chow groups . for the reader
s convenience , we include here a generalization of the beilinson lichtenbaum conjecture to general schemes and recall levine s proof of kummer isomorphism for regular schemes over dedekind domains ( * ? ? ?
* levine , theorem 12.5 ) , with the assumption of the conjecture @xmath23 . in section 3
, first we recall spiess complex @xmath3 , his duality results and the definition of lichtenbaum s weight - two motivic complex @xmath2 .
then we define the complex @xmath43 and compare it with @xmath53 and @xmath3 , respectively . in section 4 ,
we recall sato s definition of @xmath11 ( @xmath12 ) , and prove proposition [ prop : cycletosato ] and theorem [ thm : cycletosato2 ] .
throughout this paper , the concepts chain complex and cochain complex are used interchangeably . for instance , if @xmath54 is a chain complex , we think of it as a cochain complex by letting @xmath55 .
the convention for shift is : @xmath56^i = a^{i+n}$ ] . when talking about truncations , we mean truncations in cohomological degrees .
we use @xmath5 to denote a dedekind domain of characteristic @xmath57 with perfect residue fields , and @xmath6 to denote a perfect field of characteristic @xmath58 .
there is special assumption for @xmath5 in section 4 ( condition [ condition * ] ) .
all the schemes in this paper will be separated and essentially of finite type over @xmath59 with @xmath60 or @xmath61 . by variety we mean schemes separated and of finite type over fields .
the dimension of an irreducible @xmath59-scheme @xmath4 ( or dimension of @xmath4 over @xmath59 ) is defined as @xmath62 where @xmath63 is the image of the generic point of @xmath4 in @xmath59 , and @xmath41 is the function field of @xmath4 .
if @xmath4 is of finite type over @xmath59 , then @xmath64 , the krull dimension of @xmath4 .
if @xmath4 is spectrum of a field of transcendental degree @xmath65 over @xmath60 , then @xmath66 .
note that this definition of dimension is different from that of relative dimensions .
for instance , if @xmath67 and @xmath4 has relative dimension @xmath26 over @xmath59 , then @xmath68 we use @xmath69 ( resp .
@xmath70 ) to denote the set of dimension @xmath71 ( resp .
codimension @xmath71 ) points of @xmath4 . for @xmath24
, we use @xmath72 to denote the embedding of @xmath73 and @xmath74 the push - forward of ( tale ) sheaves .
the number @xmath37 is always less than @xmath22 .
i am grateful to my advisor thomas geisser for his instruction and help .
i am indebted to the referee for pointing out a mistake in a previous version .
i also want to thank enxin wu for helpful conversations .
_ bloch s cycle complex . _
let @xmath75/(\sum t_j-1).\ ] ] we define @xmath76 as the free abelian group generated by closed integral subschemes @xmath77 that intersect all the faces properly and @xmath78 .
then @xmath79 is a sheaf in the zariski and tale topology on @xmath4 .
the following complex is defined by bloch ( @xcite , or see @xcite for the notations ) : @xmath80 where @xmath81,\ ] ] with @xmath82 the closed integral subscheme generated by @xmath83 and @xmath84 $ ] the linear combination of irreducible components of @xmath85 with coefficients intersection multiplicities .
we define the complex @xmath86 to be the complex of sheaves in the topology @xmath87 with @xmath87=zar or t , and put @xmath88 in ( cohomological ) degree @xmath71 , i.e. , @xmath89 in this paper we are mostly interested in tale sheaves , so by @xmath14 we mean the complex of tale sheaves , unless otherwise stated .
note that @xmath90 satisfies zariski descent .
since @xmath14 is a complex of flat sheaves , the derived tensor product agrees with the usual tensor product of complexes . for an abelian group @xmath54 ,
define @xmath91 the complex @xmath92 is covariant for proper maps and contravariant for quasi - finite , flat maps .
let @xmath93 , and omit @xmath4 if there is no ambiguity .
define motivic borel
moore homology to be @xmath94 this definition of @xmath14 and @xmath95 depend on the base scheme @xmath59 .
unless otherwise specified , they will be defined over base @xmath59 and we will just write @xmath14 and @xmath96 , respectively . as an example , if @xmath97 has transcendental degree @xmath26 over the base @xmath6 , then @xmath98 for a scheme @xmath4 of pure dimension @xmath26 over @xmath59 , @xmath99 for a scheme @xmath4 smooth over over a perfect field @xmath6 and @xmath47 , @xmath100.\ ] ] here @xmath101 is the motivic complex defined by voevodsky , and the @xmath102 here is a quasi - isomorphism . in particular , for a field @xmath97 of transcendental degree @xmath26 over @xmath60 , @xmath103 the following three theorem state important property of cycle complex , and
we will use them in section 4 .
the first one is the localization property of cycle complex .
[ thm : localization ] for any @xmath4 , let @xmath104^j & x}$ ] be an open subscheme with closed complement @xmath105^i & x}$ ] . then there is a distinguished triangle of complexes of zariski sheaves @xmath106 & i_*{{{\mathbb{z}}}}^c_y(n)^{zar } \ar[r ] & { { { \mathbb{z}}}}^c_x(n)^{zar } \ar[r ] & j_*{{{\mathbb{z}}}}^c_u(n)^{zar } \ar[r ] & . }
\ ] ] we will need a weak form of the gersten conjecture of higher chow groups .
[ conjecture : gersten ] let @xmath107 be a regular local ring and @xmath108 be its function field .
then the canonical map @xmath109 is injective for any @xmath110 .
[ lemma : gerstenconjecture ] if @xmath4 is regular with dimension @xmath26 , and the conjecture @xmath23 is true for all the local rings of points @xmath24 , then the complexes @xmath111 and @xmath112 are acyclic at degree @xmath113 . from the conjecture , we see that for any @xmath24 with local ring @xmath114 and function field @xmath115 , the maps are injective @xmath116 but @xmath117 here @xmath118 is voevodsky s motivic cohomology , which vanishes for @xmath119 ( * ? ? ?
* theorem 3.6 ) .
hence , @xmath120 for @xmath119 .
it implies that @xmath121 for @xmath122 .
the same property holds for @xmath111 .
hence we finish the proof .
let @xmath123 be the map from tale site to zariski site . the beilinson
lichtenbaum conjecture claims that the following map @xmath124 is an quasi - isomorphism .
recall that the bloch
kato conjecture claims that for a field @xmath97 and @xmath125 , the galois symbol @xmath126 is an isomorphism .
since it was proved by voevodsky and rost recently , we will refer it as the rost voevodsky theorem .
it implies the beilinson lichtenbaum conjecture for smooth varieties over a field ( see @xcite if the field has characteristic 0 , and @xcite for positive characteristic case ) .
we generalize it to general schemes over @xmath59 , assuming the conjecture @xmath23 .
first we prove a lemma from homological algebra .
let @xmath127 be a left exact functor , and assume that there are enough injectives in @xmath128 .
assume that @xmath129 is a ( cochain ) complex .
then @xmath130 let @xmath131 be an injective resolution , i.e. , @xmath132 is a complex of injectives and is quasi - isomorphic to @xmath129 .
let @xmath54 be the kernel of @xmath133 then we obtain an injective resolution of @xmath54 as follows : @xmath134 & a \ar@{^(->}[r ] & i^n\ar[r]^{d_n } & i^{n+1 } \ar[r]^{c_{n+1 } } & j^{n+1 } \ar[r ] & . } \ ] ] then the new complex @xmath135 & i^{n-1 } \ar[r]^{d_{n-1 } } & i^n\ar[r]^{d_n } & i^{n+1 } \ar[r]^{c_{n+1 } } & j^{n+1 } \ar[r ] & } \ ] ] is an injective resolution of @xmath136 , which we denote as @xmath137 .
then for @xmath138 , @xmath139 hence we get the conclusion .
let @xmath4 be a pure dimensional scheme over @xmath59 with @xmath60 or @xmath61 , and @xmath140 .
then there is an quasi - isomorphism @xmath141 [ thm : blconjecturetorsion ] if @xmath4 is regular and the conjecture @xmath23 with @xmath142-coefficients is true for all the points @xmath24 , then the truncation in front of the first item can be removed .
if @xmath4 is smooth over @xmath60 , then the conclusion @xmath143 is implied by the rost - voevodsky theorem .
if @xmath4 is a general variety over @xmath144 we prove the conjecture by induction on the dimension of @xmath4 .
suppose that it is true for @xmath145 such that @xmath146 .
now suppose @xmath47 .
we can assume that @xmath4 is reduced .
let @xmath147 and @xmath148 be an smooth open subscheme with @xmath149 its complement . enlarge @xmath145 so that it has pure codimension @xmath150 . from theorem [ thm : localization ] , there is a distinguished triangle @xmath106 & i_*{{{\mathbb{z}}}}^c_z / m(n)^{zar } \ar[r ] & { { { \mathbb{z}}}}^c_x / m(n)^{zar } \ar[r ] & j_*{{{\mathbb{z}}}}^c_u / m(n)^{zar } \ar[r ] & .}\ ] ] apply the functor @xmath151 , then apply the functor @xmath152 , we obtain a distinguished triangle @xmath153 & r\epsilon_*(\epsilon^*{{{\mathbb{z}}}}^c_x / m(n)^{zar})\ar[r ] & r\epsilon_*(\epsilon^*j_*{{{\mathbb{z}}}}^c_u / m(n)^{zar } ) .}\ ] ] note that ( * ? ? ?
* proposition 2.2 ) @xmath154 by induction assumption , @xmath155 hence , by the five lemma , it suffices to show that @xmath156 that is , we only have to compare @xmath157 with @xmath158 for @xmath159 .
the former one is the zariski sheaf associated to the presheaf sending @xmath160 to @xmath161 while the latter one is the zariski sheaf associated to the presheaf sending @xmath160 to ( note that @xmath162 satisfies zariski descent ) @xmath163 since @xmath164 is smooth , @xmath165 satisfies the beilinson lichtenbaum conjecture , so the two sheaves are isomorphic . hence , @xmath166 if @xmath4 is a smooth scheme over @xmath67 with perfect residue fields , then the isomorphism @xmath143 is proved in ( * ? ? ?
* theorem 1.2(2 ) ) .
now let @xmath4 be a scheme flat over @xmath5 .
since this is a local problem , we can assume that @xmath5 is a discrete valuation ring .
let @xmath164 be the generic fiber and @xmath145 be the special fiber .
then @xmath167 .
so we have a localization diagram similar as above .
both @xmath164 and @xmath145 are varieties over fields , and @xmath164 is smooth , so similarly , we can prove the first part .
if the conjecture @xmath23 with @xmath142-coefficients is true , from lemma [ lemma : gerstenconjecture ] , we get the conclusion . for the reader s convenience , we recall levine s proof of a weak form of the kummer isomorphism for regular schemes over dedekind domains .
[ thm : kummerisomorphism ] if @xmath4 is a regular and pure - dimension @xmath26 scheme over @xmath67 or @xmath144 , @xmath65 is invertible in @xmath59 .
then the following map is an isomorphism @xmath168.\ ] ] if the conjecture @xmath23 with @xmath142-coefficients holds for all the points @xmath24 , then the truncation can be removed .
if @xmath4 is smooth over a field , then it is proved in ( * ? ? ?
* theorem 1.5 ) .
if @xmath4 is smooth over @xmath5 , then it is proved in ( * ? ? ?
* theorem 1.2(4 ) ) ( in this case the map @xmath169 $ ] is defined in @xcite ) .
the only case left is when @xmath4 is a regular scheme flat over @xmath5 , so @xmath170 .
this is a local problem , so assume that @xmath5 is a discrete valuation ring with perfect residue field of characteristic coprime to @xmath65 .
first we claim that , if @xmath4 is regular and admits an closed - open decomposition @xmath171 such that @xmath145 smooth and the conclusion holds for @xmath164 , then it is also true for @xmath4 . by localization sequence of cycle
complex and @xmath172 , there is a commutative diagram @xmath173 \ar[r ] & { { { \mathbb{z}}}}^c_x / m(n)\ar[d ] \ar[r ] & \epsilon^*j_*{{{\mathbb{z}}}}^c_u / m(n)^{zar}\ar[d]\\ i_*\mu^{\otimes b - n}_{m , z}[2b ] \ar[r ] & \mu^{\otimes d - n}_{m , x}[2d ] \ar[r ] & rj_*\mu^{\otimes d - n}_{m , u}[2d ] .}\ ] ] here the two rows are distinguished triangles , and @xmath174 .
the existence of the lower distinguished triangle is due to gabber , @xcite . then @xmath175)=\left\{\begin{array}{cc } \mu_{m , x}^{\otimes d - n}[2d ] , & s=-2d,\\ i_*\mu^{b - n}_{m , z}[2b ] , & s=-1 - 2b,\\ 0 , & \textrm{otherwise}. \end{array}\right.\ ] ] moreover , we have @xmath176).\ ] ] here the first isomorphism follows from the beilinson
lichtenbaum conjecture for regular @xmath164 ( theorem [ thm : blconjecturetorsion ] ) , and the second one follows from the assumption on @xmath164 . if @xmath177 , then since @xmath145 is smooth , we have isomorphism @xmath178.\ ] ] hence , the two distinguished triangles have isomorphic cohomology up to degree @xmath179 .
in particular , @xmath180)\cong \mu_{m , x}^{\otimes d - n}[2d].\ ] ] if @xmath181 , then @xmath182
. then @xmath183).\ ] ] but @xmath184 imply that @xmath185 so @xmath186)\cong r^{-2d}(j_*\mu_{m , u}^{\otimes d - n}[2d])\cong \mu_{m , x}^{\otimes d - n}[2d].\ ] ] this finishes the proof of the claim . now for @xmath4 regular over @xmath5 , we can find @xmath164 and a filtration of @xmath4 @xmath187 such that @xmath188 and @xmath164 are the special and generic fiber of @xmath4 , respectively , and @xmath189 is the singular locus of @xmath190
. then we have the following decompositions @xmath191
@xmath192 @xmath193 @xmath194 here all the @xmath195 and @xmath196 are smooth , and all the @xmath197 are regular .
from the first decomposition , we get the existence of the isomorphism for @xmath198 .
inductively , we get that the conclusion is true for @xmath199 . if the conjecture @xmath23 with @xmath142-coefficients is true , then the truncation on @xmath111 is not necessary .
the following are the duality results by geisser , @xcite .
if @xmath200 is separated and of finite type over a perfect field , then for any constructible sheaf @xmath19 on @xmath4 , there is a canonical quasi - isomorphism @xmath201
\(1 ) let @xmath59 be the spectrum of a number ring or a finite field , and @xmath19 be a constructible sheaf on @xmath4 , then there are perfect pairings of finite groups @xmath202 ( 2)let @xmath6 be an algebraically closed field and @xmath19 be a constructible sheaf on @xmath4 .
then there are perfect pairings of finitely generated groups @xmath203 _ niveau spectral sequence .
_ the method of this section depends highly on the niveau spectral sequence of higher chow groups , so let us recall its construction .
we adopt the notation in ( * ? ? ?
* proposition 2.1 ) .
we only define it for @xmath204-coefficients .
such spectral sequence with other coefficients is totally analogous .
let @xmath21 be the projection @xmath205 .
let @xmath206 be the subcomplex generated by cycles @xmath77 with @xmath207 , so @xmath208 there is a short exact sequence of complexes : @xmath209 which induces a long exact sequence of abelian groups @xmath210 moreover , by the localization property of higher chow groups , we have @xmath211 therefore , there is a convergent spectral sequence : @xmath212 this construction induces a filtration @xmath213 which is called the niveau filtration of higher chow groups .
note that @xmath214 for @xmath215 . from the niveau spectral sequence ( [ spectralsequence : niveau ] )
, we can define @xmath216 with @xmath217 and @xmath218 to be the corresponding complex with terms @xmath219 if @xmath4 is a scheme over a perfect field of characteristic @xmath7 , then @xmath220 and @xmath218 is exactly the complex @xmath221 modulo @xmath0 , so in this case our notation causes no ambiguity .
denote by @xmath222 the sheafification in topology @xmath87 of the complex of presheaves @xmath31 .
this complex @xmath223 is called the gersten complex of sheaf @xmath224 .
it is called the gersten resolution of @xmath224 if they are quasi - isomorphic via the canonical map @xmath225 here @xmath226 is denoted as the sheafification in topology @xmath87 , and @xmath227 is the direct image of presheaves along the embedding @xmath228 . note that @xmath229 if @xmath230 and @xmath74 is the direct image of tale sheaves .
there is a map of complexes @xmath231 which induces the edge morphisms of the spectral sequence .
more precisely , @xmath232 satisfies the following properties : \i ) if @xmath233 , then @xmath234 and @xmath235 ; similarly , if @xmath236 , @xmath235 as well .
\ii ) if @xmath237 , then @xmath238 is the composition : @xmath239 _ gersten complex of logarithmic de rham witt sheaves . _
let @xmath4 be a scheme essentially of finite type over @xmath6 of characteristic @xmath7 .
if @xmath240 , then define the logarithmic de rham
witt sheaf @xmath241 to be the tale subsheaf of @xmath242 generated by elements of the form @xmath243 with @xmath244 .
define @xmath245 if @xmath246 .
if @xmath247 is a field , we use the notation @xmath248 .
if @xmath4 is smooth , define @xmath8 to be the gersten complex of logarithmic de rham witt sheaves : @xmath249 the term indexed by @xmath250 is in ( cohomological ) degree @xmath71 , i.e. , @xmath251 define @xmath252 the differentials of @xmath8 are induced from the coniveau filtration of the sheaf @xmath27 , similar as the one for higher chow groups above .
if @xmath4 is not smooth , using theorem [ thm : mktolrw ] below , moser @xcite also defines @xmath28 by identifying it with @xmath253 ( see definition below ) .
the following is the duality result of moser .
let @xmath6 be a finite field of characteristic @xmath21 and @xmath4 be a @xmath6-scheme of pure dimension @xmath26 .
then for every @xmath254 and every constructible @xmath10-sheaf @xmath19 , there are perfect pairings of finite groups
@xmath255 _ kato s complex of milnor k - theory .
_ let @xmath4 be a scheme over @xmath59 .
define @xmath32 to be kato s complex of milnor k - theory ( cf .
k. kato , @xcite ) : @xmath256 the group @xmath257 is milnor k - group , and @xmath258 if @xmath246 .
the differential @xmath259 is defined as follows : for any @xmath260 and any @xmath261 , we take the normalization @xmath262 of @xmath263 with @xmath264 its generic point and define a map @xmath265 here the notation @xmath266 means that @xmath267 is in the fiber of @xmath268 , @xmath269 is the norm map of milnor k - theory ( see bass
tate , @xcite and kato , ( * ? ? ?
* section 1.7 ) ) , and @xmath270 is the tame symbol defined by @xmath271 . then @xmath272 note that the sum in @xmath259 is finite since elements in @xmath273 are represented by sums of tensors of @xmath274 elements in @xmath275 , and each element in @xmath276 has a finite number of poles and zeros . when applying the tame symbol , only a finite number of terms in the sum
are non - zero .
suppose that in ( cohomological ) degree @xmath71 , @xmath277 and define @xmath278 .
the complex @xmath32 is covariant for proper maps and contravariant for quasi - finite and flat maps ( see rost , ( * ? ? ?
* proposition 4.6(1),(2 ) ) ) .
note that all the three complexes , @xmath279 and @xmath8 are concentrated in cohomological degree @xmath280 $ ] when @xmath281 .
for @xmath4 a scheme separated and essentially of finite type over @xmath60 with @xmath6 perfect of characteristic @xmath7,[thm : mktolrw ] @xmath282 by bloch kato , ( * ? ? ?
* theorem 2.1 ) , for any field @xmath97 , there is an isomorphism @xmath283 sending @xmath284 to @xmath285 by ( * ? ? ?
* lemma 4.1 ) ( see jannsen - saito - sato , ( * ? ? ?
* theorem 2.1.1 and theorem 2.11.3(3 ) ) for a more detailed proof ) , this isomorphism respects the differentials in @xmath32 and @xmath286 hence it induces an isomorphism of complexes @xmath287 to relate @xmath31 with @xmath288 , we need the nesterenko suslin isomorphism .
[ isomorphism : nesterenkosuslin]let @xmath97 be a field of transcendental degree @xmath26 over a field @xmath6 , and @xmath289 . then there is an isomorphism @xmath290 the map @xmath291 is defined as follows : for any generator @xmath292 , @xmath293 . here
@xmath294 is the image of @xmath295 in @xmath296 , @xmath297 is the norm map of milnor k - theory , and @xmath298 with @xmath299 s the coordinates of @xmath295 in @xmath300 , i.e. , pull - back of @xmath299 on @xmath295 . since @xmath295 intersects all the faces properly , @xmath301 for all @xmath71 .
nestenrenko and suslin showed that @xmath291 is an isomorphism .
it is easy to see that this isomorphism does not depend on the choice of the base field @xmath6 . in particular , let @xmath302 , and @xmath303 .
then @xmath304 and we have isomorphism @xmath305 [ lemma : hcgtomk ] for @xmath4 separated and essentially of finite type over @xmath60 or @xmath67 , the nesterenko
suslin isomorphism induces an isomorphism of complexes @xmath306 if the base is @xmath144 with characteristic @xmath7 , then @xmath307 we have to show that the maps @xmath308 induces the following commutative diagram : @xmath309 ^ -{\chi}\ar[r]^-{d '' } & \displaystyle\bigoplus_{x_{(m - n-1)}}h_{m-1}^{{\operatorname{bm}}}(k(x ) , { { { \mathbb{z}}}}(n ) ) \ar[d]^-{\chi}\\ \displaystyle \bigoplus_{x_{(m - n ) } } k_{m-2n}^m(k(x ) ) \ar[r]^-{d ' } & \displaystyle\bigoplus_{x_{(m - n-1)}}k_{m-2n-1}^m(k(x ) ) .}\ ] ] it suffices to prove the conclusion for @xmath310 with @xmath22 , since commutativity of such diagram at the other degrees is similar .
let @xmath311 be the normalization of @xmath4 , and @xmath73 be a codimension 1 point of @xmath4 .
consider the following commutative diagram : @xmath312 ^ -{d''_{x ' } } \ar@{=}[d ] & \displaystyle\bigoplus_{x'|x}h_{d+n-1}^{{\operatorname{bm}}}(k(x'),{{{\mathbb{z}}}}(n))\ar[d]^-{\sum n}\\ h^{{\operatorname{bm}}}_{d+n}(k(x),{{{\mathbb{z}}}}(n ) ) \ar[r]^-{d''_x } & h_{d+n-1}^{{\operatorname{bm}}}(k(x ) , { { { \mathbb{z}}}}(n)).}\ ] ] here @xmath313 and @xmath314 are differentials in @xmath31 and @xmath315 , respectively , and @xmath316 is the push - forward of higher chow groups of finite field extensions .
this diagram is commutative by covariance of gersten complex .
so @xmath317 to finish the proof , consider the following diagram : @xmath318 ^ -{d''_{x ' } } \ar[d]^{\chi_{k(x ' ) } } & \displaystyle\bigoplus_{x'|x } h^{{\operatorname{bm}}}_{d+n-1}(k(x'),{{{\mathbb{z}}}}(n))\ar[d]^{\chi_{k(x')}}\ar[r ] & h^{{\operatorname{bm}}}_{d+n-1}(k(x),{{{\mathbb{z}}}}(n ) ) \ar[d]^{\chi_{k(x)}}\\ k_{d - n}^m(k(x ' ) ) \ar[r]^-{\partial_{x ' } } & \displaystyle\bigoplus_{x'|x } k_{d - n-1}^m(k(x'))\ar[r ] & k^m_{d - n-1}(k(x)).}\ ] ] the horizontal maps on the right hand square are norm maps . by definition ,
the composition of the maps at the bottom is the differential @xmath259 in @xmath32 . by the first part of the proof , the composition of the maps on the top is the differential @xmath319 in @xmath31 .
hence it suffices to prove commutativity of this diagram .
the right hand diagram commutes by ( * ? ? ?
* lemma 4.7 ) .
on the other hand , geisser and levine showed commutativity of the left hand square ( * ? ? ?
* lemma 3.2 ) ( even though their statement is for @xmath320-coefficients , their proof holds for @xmath204-coefficients as well ) .
hence we prove the first part .
the second part about the coefficients @xmath10 holds since both complexes are @xmath321 and @xmath31 modular @xmath0 .
[ definition:1 ] define @xmath322 explicitly , [ definition:2]denote @xmath323 . given a generator @xmath324
we define @xmath325 ( see ( [ filtration ] ) ) as follows : \1 ) if @xmath233 or @xmath236 , @xmath326 .
\2 ) if @xmath237 and @xmath327 , @xmath326 .
\3 ) if @xmath237 and @xmath328 , then @xmath145 is dominant over some @xmath329 . pulling back @xmath145 along @xmath330 we obtain @xmath331 , which is sent to @xmath332 by the quotient @xmath333 applying the nesterenko suslin isomorphism @xmath334 , we get @xmath335 since @xmath145 is dominant over @xmath73 , @xmath336 is a closed point in @xmath337 with residue field @xmath338 .
therefore , by definition of @xmath334 , @xmath339 with @xmath340 and @xmath341 note that in definition [ definition:2 ] , case 3 ) , @xmath342 where @xmath343 is the map from @xmath92 to @xmath344 . [ lemma : cyclecomplextomk ] let @xmath4 be separated and essentially of finite type over @xmath144 or @xmath61 .
the map @xmath345 defined above is a map of complexes , and it is functorial with respect to pullbacks defined by quasi - finite , flat maps and push - forwards defined by proper maps .
since @xmath346 , it is a map of complexes . for functoriality , it suffices to assume that @xmath347 .
first , we show that @xmath345 is compatible with pull - backs defined by quasi - finite , flat maps @xmath348 .
we have to prove that the following diagram is commutative : @xmath349^{f^ * } \ar[r]^-{\psi_y } & \displaystyle\bigoplus_{y_{(m - n)}}k_{m-2n}^m(k(y ) ) \ar[d]^{f^ { * ' } } \\ z_n(x , m-2n)\ar[r]^-{\psi_x } & \displaystyle\bigoplus_{x_{(m - n)}}k_{m-2n}^m(k(x ) ) .}\ ] ] here @xmath350 sends a generator @xmath351 to its cycle theoretic pull - back @xmath352 and @xmath353 is defined by the field extension @xmath354 with @xmath329 , @xmath355 and @xmath356 .
let @xmath357 be the projections .
if @xmath358 , then @xmath359 , so @xmath360 suppose that @xmath361 .
without loss of generality , replacing @xmath362 by @xmath363 and @xmath4 by @xmath364 , we can assume that @xmath362 is irreducible of dimension @xmath274 and @xmath145 is dominant over @xmath362 . since @xmath365 is quasi - finite and flat , @xmath4 is of equi - dimensional @xmath274 .
let @xmath366 , @xmath367 be irreducible components of @xmath4 and @xmath368 be the generic points of @xmath367
. then @xmath369 . therefore it suffices to prove commutativity of the following diagram : @xmath349^{f^*}\ar[r ] & h_m^{{\operatorname{bm}}}(k(y),{{{\mathbb{z}}}}(n ) ) \ar[d]\ar[r ] & k_{m-2n}^m(k(y ) ) \ar[d]^{f^ { * ' } } \\
z_n(x , m-2n)\ar[r ] & \displaystyle\bigoplus_{i}h_{m}^{{\operatorname{bm}}}(k(x_i),{{{\mathbb{z}}}}(n))\ar[r ] & \displaystyle\bigoplus_ik_{m-2n}^m(k(x_i ) ) } .\ ] ] the square on the left commutes by functoriality of higher chow groups with respect to flat pull back , the square on the right commutes since by definition , the nesterenko suslin isomorphism is covariant with respect to finite field extensions .
therefore @xmath345 commutes with quasi - finite and flat pull - backs .
if @xmath370 is a proper map , then @xmath345 is covariant for push - forwards . to see that , fix @xmath371 after replacing @xmath4 by the irreducible component of @xmath4 containing @xmath372 , it suffices to assume that @xmath4 is irreducible of dimension @xmath37 with function field @xmath108 and prove that @xmath145 is sent to the same element via the two paths in the following square : @xmath373^{g _ * } \ar[r]^{\psi_x } & k_{m-2n}^m(k ) \ar[d]^{g_*'}\\ z_n(y , m-2n ) \ar[r ] ^-{\psi_y } & \displaystyle\bigoplus_{y_{(m - n)}}k_{m-2n}^m(k(y)).}\ ] ] here @xmath374 is defined as follows : @xmath375 with @xmath376 $ ] , and @xmath377 to show that @xmath378 , there are three cases : \1 ) if @xmath379 , then @xmath380 .
moreover , @xmath381 hence @xmath382 .
\2 ) if @xmath4 is dominant over some @xmath383 and @xmath384 , then @xmath385 moreover , @xmath386 also implies @xmath387 , hence @xmath388
\3 ) if @xmath4 is dominant over some @xmath383 and @xmath389 , then @xmath390 therefore , @xmath145 is dominant over @xmath4 and @xmath391 is irreducible and dominant over @xmath268 .
we have a commutative diagram of field extensions : @xmath392 \\
k(g(z ) ) \ar[u ] & k(y)\ar[u]\ar[l].}\ ] ] then @xmath393 and @xmath394 since @xmath395 is the image of @xmath396 under the map @xmath397 , @xmath398 therefore @xmath399 . *
remark 1 . * in @xcite , langsburg defined a map from @xmath92 to @xmath32 exactly the same as the one in definition [ definition:2 ] , except in case 3 ) , instead of using @xmath395 , he used @xmath400 by multilinearity of milnor k - theory , it is easy to see that @xmath401 up to 2-torsion element . therefore , the map @xmath345 is equal to langdsburg s map up to a 2-torsion .
the advantage of using @xmath395 is that one can use the results in @xcite . on the other hand ,
using the idea in landsburg s proof of showing that his map is a map of complexes , together with properties of @xmath402 in @xcite , we can give a direct proof of showing that @xmath345 is a map of complexes .
this proof is lengthy , comparing to the one we give in theorem [ lemma : cyclecomplextomk ] .
( there is a small gap in landsburg s proof , as he only checks compatibility of his map with the differentials in the case of discrete valuation rings ) .
* remark 2 . *
it is easy to see that the map @xmath345 can be generalized to define a map from bloch s cycle complex @xmath92 to the corresponding cycle complex with coefficients in milnor k - groups or quillen k - groups defined by m. rost @xcite .
[ thm : cycletolrw ] for any @xmath4 separated and essentially of finite type over @xmath6 of characteristic @xmath7 , the map @xmath345 induces a quasi - isomorphism @xmath403 composing @xmath404 with the isomorphism @xmath405 we get a map of complexes @xmath406 to compare the cohomology of @xmath407 and @xmath8 , consider the niveau spectral sequence of higher chow groups : @xmath408 by ( * ? ? ?
* theorem 1.1 ) , this spectral sequence collapses to give edge isomorphisms @xmath409 composing with the isomorphisms @xmath410 and @xmath411 we get an isomorphism @xmath412since @xmath343 induces edge morphisms of the spectral sequence , and @xmath413 , we see that @xmath36 induces @xmath414 .
hence @xmath36 is a quasi - isomorphism .
_ @xmath415-coefficients .
_ let @xmath6 be a finite field with characteristic @xmath21 and @xmath4 be smooth and projective over @xmath6 .
let @xmath416 . in ( * ?
* proposition 2.1 ) , conjecture a(@xmath37 ) ( part of parshin s conjecture ) is equivalent to that , in the niveau spectral sequence @xmath417 @xmath418 for @xmath419 . in other words , it is equivalent to the existence of the following isomorphism : @xmath420 here @xmath421 is the gersten complex of higher chow groups with @xmath415-coefficients : @xmath422 similar to the case above for the @xmath10-coefficients , we obtain the following theorem : for smooth and projective varieties over finite fields , conjecture a(n ) is true if and only if @xmath343 induces a quasi - isomorphism from @xmath423 to the gersten complex of higher chow groups @xmath424 in @xmath415-coefficients .
in this section , we assume that @xmath60 or @xmath61 , and we will deal with the complex @xmath425 .
we assume that @xmath4 is of finite type over @xmath59 and @xmath426 .
spiess s dualizing complex @xmath3 of tale sheaves for surfaces uses the weight - two motivic complex defined by ( * ? ? ?
* definition 2.1 ) . for a regular noetherian ring @xmath54 ,
let @xmath427 , z={\operatorname{spec}}a[t]/t(t-1),\ ] ] and @xmath428 a finite set of exceptional units of @xmath54 ( i.e. , @xmath429 and @xmath430 are units ) .
let @xmath431/\prod_{i=1}^m(t - b_i).\ ] ] there is an exact sequence induced from relative k - theory @xmath432 here @xmath433 is the k - theory of the category of quasi - coherent sheaves .
taking limits among all the @xmath434 s , we obtain an exact sequence @xmath435 with @xmath436 if @xmath4 is a regular noetherian scheme , associating the group @xmath437 to each regular affine scheme @xmath438 which is tale over @xmath4 , we obtain a presheaf on @xmath439 .
denote by @xmath440 the associated tale sheaf .
then lichtenbaum s weight - two motivic complex @xmath441 is defined as the complex @xmath442 with the terms in cohomological degree 1 and 2 , respectively . if @xmath443 is a field , there is an exact sequence @xcite , @xmath444 [ spiess , ] define the complex of tale sheaves @xmath3 as follows : @xmath445 here @xmath446 , @xmath447 is the map sending a rational function to its associated divisors , and @xmath448 is the composition @xmath449 with @xmath450 the map in the gersten resolution of k - theory @xcite . the terms are put in ( cohomological ) degree -3 , -2 , -1 , 0 , respectively.[definition : spiesscomplex ] note that the assumption on degrees are different from that in @xcite . for @xmath4 an equidimensional surface over @xmath204 satisfying the ( nr ) condition and every constructible sheaf @xmath19 on @xmath4 , there are perfect pairings of finite groups @xmath451 here we say that @xmath4 satisfies the ( nr ) condition if _ @xmath452 is not formally real for every @xmath24 .
_ let @xmath6 be an algebraically closed field of characteristic @xmath21 , @xmath4 be an irreducible surface over @xmath6 and @xmath19 be an @xmath37-torsion constructible sheaf on @xmath4 , with @xmath453 .
let @xmath454 .
then there are perfect pairings of finitely generated groups : @xmath455 in @xcite , deninger considered a dualizing complex of tale sheaves @xmath456 for curve @xmath362 , and e. nart @xcite compared it with cycle complex @xmath457 by constructing a map from @xmath457 to @xmath458 which induces a quasi - isomorphism . here
we put the terms in ( cohomological ) degree -1 and 0 respectively . in what follows
we generalize this method and define a similar complex for surfaces .
consider the niveau spectral sequence of higher chow groups for surfaces @xmath459 only @xmath460,@xmath461 and @xmath462(@xmath463 ) are non - vanishing , @xmath464 for @xmath465 , and in the bottom of the spectral sequence , the only nonvanishing terms in @xmath466 are : @xmath467 with @xmath468 we define a cochain complex of abelian groups @xmath43 : @xmath469 here @xmath470 , @xmath471 is the map induced by @xmath472 , @xmath473 , and @xmath474 is the composition @xmath475^\pi \ar[dr]_{d_2 } & \displaystyle\bigoplus_{x_{(2)}}h_2^{{\operatorname{bm}}}(k(x),{{{\mathbb{z } } } } ) \ar[d]^f \\ & \displaystyle\bigoplus_{x_{(1)}}k(x)^*. } \label{definition : d_2}\ ] ] the ( cohomological ) degrees of the terms are -3 , -2 , -1 , 0 , respectively .
let @xmath4 be a surface over @xmath144 or @xmath61 , then there is a map @xmath476 which induces isomorphisms @xmath477 for @xmath478[theoremcycletruncated ] .
following the same idea of the definition of @xmath345 in 2 , we can define a map from @xmath44 to @xmath43 : @xmath479 ^ -d\ar[d]^-{\psi_3 ' } & z_0(x,2 ) \ar[r]^-d\ar[d]^-{\psi_2 ' } & z_0(x , 1 ) \ar[d]^-{\psi_1'}\ar[r]^-d & z_0(x,0 ) \ar[d]^-{\psi_0'}. \\
\displaystyle\bigoplus_{x_{(2 ) } } z_0(k(x),3)/i_x \ar[r]^-{d_3 } & \displaystyle\bigoplus_{x_{(2)}}z_0(k(x),2 ) \ar[r]^-{d_2 } & \displaystyle\bigoplus_{x_{(1)}}k(x)^ * \ar[r]^-{d_1 } & \displaystyle\bigoplus_{x_{(0 ) } } { { { \mathbb{z}}}}}\ ] ] here @xmath480 for @xmath481 , @xmath482 is defined as follows :
if @xmath483 , @xmath484 ; if @xmath485 , then @xmath145 is dominant over some @xmath486 , so pulling back @xmath145 along @xmath487 , we get an element @xmath488 . by similar argument as in 2 , we see that the diagram is commutative . moreover , the @xmath482 s induce the corresponding isomorphisms from the degeneration of the niveau spectral sequence : @xmath489 @xmath490 @xmath491 @xmath492 here the equality @xmath493 follows from the diagram ( [ definition : d_2 ] ) .
it remains to show that @xmath494 .
look at the diagram ( [ definition : d_2 ] ) .
there is an exact sequence @xmath495 since @xmath496 is surjective , @xmath497 , so @xmath498 . but @xmath499 .
hence @xmath500 * remark .
* in the above proof , @xmath501 is similar to the map defined by nart ( @xcite ) .
as noted by nart , there are only two types of generators in @xmath502 , the vertical ones and the horizontal ones .
the vertical ones are those @xmath145 s with @xmath503 , and the horizontal ones are those @xmath145 s with @xmath504 for some @xmath505 .
nart s map sends the first type to 0 and the second type to @xmath506 with @xmath507 the norm map of fields . in our situation , according to the nesterenko suslin isomorphism , @xmath501 sends the first type to 0 and the second type to @xmath508 .
so it agrees with nart s map up to 2-torsion .
[ thm : calccalk ] for a surface @xmath4 over @xmath144 or @xmath61 , @xmath509 in the derived category of tale sheaves . before proving the theorem
, we need the following lemma .
first , let us recall some notations from @xcite
. from now on until theorem [ thmcyclespiess ] , assume that @xmath97 is a field with @xmath510 .
let @xmath511 .
a closed subvariety @xmath512 is called a codimension @xmath513 face .
a closed subvariety @xmath514 is said to be in good position if @xmath515 has codimension @xmath516 in @xmath160 for any codimension @xmath513 face @xmath517 .
let @xmath518 denote the union of all codimension @xmath37 closed subvarieties of @xmath519 in good position . given a scheme @xmath4
, we write @xmath520 for some space , contravariant functorial in @xmath4 , whose homotopy groups calculate the quillen k - theory of @xmath4 . for @xmath521 a closed subset ,
we write @xmath522 for the homotopy fibre of @xmath523 .
this construction can be iterated .
given @xmath524 , we define the multi - relative k - space @xmath525 @xmath526 let @xmath527 @xmath528 where @xmath529 .
let @xmath530 or @xmath531 .
define @xmath532 if @xmath533 is an inclusion of closed subvarieties , then there is a canonical map @xmath534 hence we can define @xmath535 @xmath536 there are two isomorphisms from @xmath537 to @xmath538 in @xcite : the first one @xmath539 is induced by the edge morphisms of the spectral sequence @xmath540 proved in @xcite ( note that we have changed the notation in loc . it . to the motivic borel
moore homology of @xmath97 over base @xmath59 ) ; the second one @xmath541 is induced from the quasi - isomorphism between @xmath542 and @xmath543 ( loc.it . ,
7 ) . here
@xmath542 is the following truncation of cycle complex of fields : @xmath544 with @xmath545 hence @xmath546 . from ( [ weighttwo ] ) ,
we know that this quasi - isomorphism induces an isomorphism : @xmath547 [ lemma : bl ] ( 1 ) for any field @xmath97 , @xmath548 \(2 ) the following diagram is commutative @xmath549_{\chi } \ar[d]^{\theta } & \\
k_2(k ) \ar[r]_{s } & k_2^m(k ) .}\ ] ] ( 1)@xmath550(2 ) : by ( * ? ? ?
* proposition 3.3 ) , we know that @xmath551 . then from ( 1 ) , we have @xmath552 \(1 ) by the definitions of @xmath541 and @xmath539 ( * ? ? ? * and 7 ) , and
the following isomorphisms : @xmath553 we see that the two maps are induced by the two paths from @xmath554 to @xmath555 and to @xmath556 shown in the following diagram : @xmath557 \ar@{->>}[d ] & k_{0,{\mathcal v}^2}(\delta^2,\sum ) \ar[d]^{\cong } & k_{0,{\mathcal v}^2}(\delta^2,\partial ) \ar[l]^{\cong } \ar@{->>}[d]\\ k_{1,{\mathcal v}^1}(\delta^1,\partial ) \ar@{->>}[d ] & z_0(f,2 ) \ar@{->>}[d ] & k_{0,{\mathcal v}^1}(\delta^2,\partial)\ar[d]^{\cong } \\
k_{1}(\delta^1,\partial ) & h_{2}^{bm}(f,{{{\mathbb{z } } } } ) & k_0(\delta^2,\partial ) .}\ ] ] the map @xmath539 is the path from @xmath558 to @xmath559 and then to @xmath556 , while @xmath541 is the path from @xmath558 to @xmath555 via @xmath560 .
in order to show that @xmath561 , it suffices to show commutativity of the following diagram : ( starting from @xmath560 and end at @xmath556 ) @xmath562 ^ -{\cong}_-3 \ar@{->>}[r]^{4 } & k_{0,{\mathcal v}^1}(\delta^2,\partial ) \ar@{->>}[r ] & k_0(\delta^2,\partial ) \\ k_{1,{\mathcal v}^1-{\mathcal v}^2}(\delta^2,\sum ) \ar@{->>}[rr]^2\ar@{->>}[u]^1 & & k_{1,{\mathcal v}^1}(\delta^1,\partial ) \ar@{->>}[u]^5 \ar@{->>}[r ] & k_1(\delta^1 , \partial)\ar[u]^{\cong}}\ ] ] here the right hand square is induced from the embedding @xmath563 , so it is commutative .
therefore we have to show that the square on the left commutes , or more precisely , @xmath564 recall the long exact sequence of relative k - groups : @xmath565 that is the only type of long exact sequence involved in the diagrams above .
in particular , @xmath566 here map @xmath150 is for the embedding @xmath567 , map 4 is for the embedding @xmath568 , maps 2 , 3 and 5 are for the embedding @xmath569 ( embedding of the last face ) , and @xmath570 means that the map @xmath150 corresponds to type @xmath71 in the sequence ( [ sequence : relativektheory ] ) , etc . by functoriality of relative sequence of k - theory
, we obtain the commutativity .
[ proof of theorem [ thm : calccalk ] ] let us first define a map @xmath571 in the derived category .
for any @xmath486 , there is a quasi - isomorphism ( * ? ? ?
* theorem 7.2 ) in the derived category between the complex @xmath572 and the complex @xmath573 moreover , it induces an isomorphism between the cokernels of the two complexes @xmath574(see ( [ theta ] ) above ) .
recall that the complex @xmath43 ( resp . , @xmath45 ) is defined by connecting @xmath575 @xmath576 with @xmath577 via @xmath578 ( resp . , @xmath579 ) . since @xmath580
, it suffices to show that the following diagram is ( anti-)commutative : @xmath581 ^ -f \ar[d]^-{\theta_x } & \displaystyle\bigoplus_{x_{(1)}}k(x)^ * \ar@{=}[d ] \\
\displaystyle\bigoplus_{x_{(2)}}k_2(k(x))\ar[r]^{\partial ' } & \displaystyle\bigoplus_{x_{(1)}}k(x)^ * .}\ ] ] let @xmath582 and @xmath583 , consider the following diagram : @xmath549_{\chi}\ar[drr]^f \ar[dd]^{\theta } & & \\ & k_2^m(k)\ar[r]^-{\partial}\ar[dl]_s & \displaystyle\bigoplus_{x_{(1)}}k(x)^ * \\ k_2(k ) \ar[rru]_{\partial ' } & & .}\ ] ] here @xmath584 is the tame symbol and @xmath402 is the nesterenko
suslin isomorphism .
we have to show that the outside triangle is ( anti-)commutative , so it suffices to show that the three small triangles are ( anti-)commutative . the triangle on the top is commutative by lemma [ lemma : hcgtomk ] .
the one on the left commutes by lemma [ lemma : bl ] . for the triangle on the bottom , we have to show commutativity of the following diagram : @xmath585
^ -\partial \ar[d]^s & k^*\\ k_2(k)\ar[ru]_{\partial ' } & .}\ ] ] here @xmath6 is the residue field of a valuation @xmath586 of @xmath108 , with @xmath496 a prime element . from (
* definition 1.1 r3e ) , @xmath450 has the following property : for any @xmath587 with @xmath588 , @xmath589 , i.e. @xmath590 . here
@xmath591 is the residue class of @xmath592 in @xmath593 . but
@xmath594 by multilinearity of milnor k - theory , any element @xmath595 with @xmath596 can be decomposed into a product of elements of the form @xmath597 and @xmath598 with @xmath588 . therefore
, the diagram commutes up to sign .
so the triangle outside is commutative . in conclusion
, there is an morphism from @xmath43 to @xmath3 in the derived category of tale sheaves . now let us compare the ( co)homology groups .
it is clear that @xmath599 from ( * ? ? ?
* theorem 7.2 ) , @xmath600 to compare @xmath448 and @xmath474 , consider the diagrams defining them @xmath601^\pi \ar[dr]_{d_2 } & \displaystyle\bigoplus_{x_{(2)}}h_2^{{\operatorname{bm}}}(k(x),{{{\mathbb{z } } } } ) \ar[d]^f & \displaystyle\bigoplus_{x_{(2)}}c_{2,2}(k(x ) ) \ar[r]^ { \omega_{k(x)}}\ar[dr]_{c_2 } & \displaystyle\bigoplus_{x_{(2)}}k_2(k(x))\ar[d]^{\partial'}\\ & \displaystyle\bigoplus_{x_{(1)}}k(x)^ * & & \displaystyle\bigoplus_{x_{(1)}}k(x)^ * .}\ ] ] here @xmath496 and @xmath602 are both surjective onto groups that are connected via the isomorphism @xmath603 and @xmath604 , so @xmath605 therefore , @xmath606 similar to the proof of theorem [ theoremcycletruncated ] , we have that @xmath607 in the following theorem we will need to consider fields of dimension @xmath608 over the base @xmath59 .
if @xmath60 , a field @xmath108 of transcendental degree 2 over @xmath59 has @xmath609 . if @xmath67 , the function field @xmath108 of a scheme @xmath4 with @xmath610 will have @xmath609 .
[ thmcyclespiess ] suppose that @xmath4 is a two - dimensional scheme over @xmath60 or @xmath67 and the function field of @xmath4 is @xmath108 .
then there is a map in the derived category @xmath611 which induces a quasi - isomorphism @xmath612 assuming the beilinson soul conjecture for @xmath613 on fields of dimension @xmath608 over @xmath59 , then @xmath614 in the derived category . here
@xmath101 is voevodsky s motivic complex . from theorem [ theoremcycletruncated ] and [ thm : calccalk ] , we know that in the derived category , there is a map @xmath615
. moreover , @xmath616 the beilinson soul conjecture ( * ? ? ?
* conjecture 5 ) asserts that for a regular scheme @xmath4 and @xmath617 , @xmath618 for @xmath619 .
in particular , for @xmath613 on the function field @xmath108 of @xmath4 ( note that @xmath609 ) , it is equivalent to asserting that @xmath620 for @xmath621 . to prove that @xmath622 , we need to show that @xmath623 is acyclic at ( cohomological ) degree @xmath624 .
now we have two cases , _
( i ) _ @xmath4 over @xmath60 , or @xmath4 over @xmath67 but factors through some closed point , _ ( ii ) _
@xmath4 flat over @xmath67 .
_ case ( i ) _ this means that @xmath4 is a scheme over a perfect field @xmath6 .
let @xmath625 be any open subscheme of @xmath4 with complement @xmath626 .
enlarge @xmath145 so that it has dimension @xmath150 .
then there is a long exact sequence @xmath627 by ( * ? ?
* theorem 1.1 ) , @xmath628 is quasi - isomorphic to @xmath629 ( deninger s complex , see ( [ deningercomplex ] ) ) , so @xmath630 for @xmath631 .
hence @xmath632 for @xmath633 .
take limit among all the open subschemes over @xmath4 , we get that , for @xmath633 , @xmath634 hence @xmath635 _ case ( ii ) _ @xmath4 is flat over @xmath67 .
this is a local problem , so we can assume that @xmath5 is a discrete valuation ring .
let @xmath164 be the generic fiber and @xmath145 be the closed fiber .
then @xmath145 is a curve over the residue field of @xmath5 .
similar as in _
( i ) _ , we can get the conclusion .
[ corollaryrhom ] suppose that @xmath4 is a two - dimensional scheme over @xmath60 or @xmath67 .
then for any torsion sheaf @xmath19 on @xmath4 , there is an isomorphism @xmath636 first , we assume that the conclusion is true for constructible sheaves on surfaces smooth over @xmath60 or @xmath61 .
if @xmath4 is over @xmath60 and not smooth , and @xmath19 is constructible , choose an open smooth subscheme @xmath637 with complement @xmath638 of codimension 1 . by purity , @xmath639 ( * ? ? ?
* proposition 3.5 ) and @xmath640 ( * ? ? ?
* proposition 1.3.1 ( ii ) ) . from the short exact sequence @xmath641
we obtain a commutative diagram of distinguished triangles @xmath642\ar[d ] & { \operatorname{rhom}}_x({\mathcal f } , { { { \mathbb{z}}}}^c_x ) \ar[d]^{\epsilon_x}\ar[r ] & { \operatorname{rhom}}_x(j_!j^*{\mathcal f } , { { { \mathbb{z}}}}_x^c)\ar[d]\\ { \operatorname{rhom}}_x(i_*i^*{\mathcal f},{\mathcal k}_x ) \ar[r ] & { \operatorname{rhom}}_x({\mathcal f},{\mathcal k}_x ) \ar[r ] & { \operatorname{rhom}}_x(j_!j^*{\mathcal f } , { \mathcal k}_x ) . } \label{diagramlocalization}\ ] ] by adjointness and purity , the groups on the left are canonically isomorphic to @xmath643 respectively . hence , by nart @xcite , the left hand vertical map is an isomorphism . by adjointness , the groups on the right can be identified with @xmath644 respectively .
since @xmath164 is smooth , the right hand vertical map is an isomorphism by induction assumption .
therefore , the vertical map in the middle is an isomorphism .
let @xmath4 be a scheme over @xmath67 but not smooth and @xmath19 is constructible .
if @xmath645 factors through some closed point of @xmath59 , then it is indeed a variety over a perfect field , in which case it is proved as above . if @xmath645 does not factor through any closed point of @xmath59 , then it is flat over @xmath59
. it suffices to assume that @xmath5 is a discrete valuation ring .
let @xmath164 be the generic fiber and @xmath145 be the special fiber , then @xmath145 is a curve over a field . for the decomposition
@xmath646 we can apply the strategy as above to prove the conclusion for @xmath4 ( see ( * ? ? ?
* corollary 7.2 ) for purity of cycle complex ) . for general torsion sheaf @xmath19
, it can be written as @xmath647 with @xmath648 constructible ( * ? ? ?
* chapter ii , proposition 0.9 ) .
so @xmath649 since for any sheaf @xmath650 , @xmath651 the canonical map @xmath611 in the derived category induces a map @xmath652 hence a map of spectral sequences @xmath653\ar[d ] & { \operatorname{ext}}^{s+t}_x({\mathcal f},{{{\mathbb{z}}}}^c_x)\ar[d]^{\epsilon_x}\\ e_2^{s , t}=\lim^s{\operatorname{ext}}^t_x({\mathcal f}_i,{\mathcal k}_x ) \ar@{=>}[r ] & { \operatorname{ext}}^{s+t}_x({\mathcal f},{\mathcal k}_x).}\ ] ] note that these two spectral sequences are convergent . for the first one , it follows from (
* corollary 4.9 ) ( the statement in this corollary is for @xmath144-schemes , but the proof holds for @xmath61-schemes ) . for the second one , since @xmath654 , hence it is a bounded spectral sequence .
the two spectral sequences are isomorphic at the @xmath655-page via maps defined by @xmath611 , hence @xmath656 now let us prove the conclusion for constructible sheaves on surfaces smooth over @xmath59 .
again , we separate the cases : a ) @xmath60 with characteristic @xmath21 ; b ) @xmath67 .
\a ) let @xmath19 be a @xmath10-sheaf , with @xmath657 . by the niveau spectral sequence of higher chow groups with @xmath10 coefficients , for @xmath658 , @xmath659 by ( * ? ? ?
* theorem 1.1 ) , we see that @xmath660 for @xmath658 . therefore , @xmath661 in particular , for such @xmath19 , @xmath662 now let @xmath19 be an @xmath37-torsion sheaf , with @xmath453 .
let @xmath663 . since both @xmath53 and @xmath3 satisfy the kummer sequence ( see ( * ? ? ?
* theorem 1.5 ) and ( * ? ? ?
* proposition 1.5.1 ) , respectively ) : @xmath664 @xmath665 thus , @xmath666 \b ) let @xmath4 be surface smooth over @xmath61 .
let @xmath19 be an @xmath37-torsion .
if @xmath37 is invertible on @xmath4 , then @xmath53 and @xmath3 satisfy the kummer sequence ( * ? ? ?
* proposition 1.5.1 ) .
hence @xmath667 is an isomorphism .
if @xmath37 is not invertible on @xmath4 , consider the following decomposition @xmath668^{i}\ar[d ] & x \ar[d ] & v \ar[l]_{j}\ar[d ] \\ { \operatorname{spec}}d/(n ) \ar[r ] & s= { \operatorname{spec}}d & { \operatorname{spec}}d[1/n ] .\ar[l]}\ ] ] here the two squares are fiber products . then @xmath71 and @xmath669 are closed and open immersions , respectively . since @xmath4 is flat over @xmath59 , @xmath670 has codimension 1 .
similar as in ( [ diagramlocalization ] ) , there is a localization diagram and the vertical maps on the left and right are isomorphisms , which proves that @xmath671 is an isomorphism .
this finishes the proof .
_ sato s complex @xmath11 . _
k. sato @xcite defined a complex @xmath11 for certain schemes over dedekind domains .
let us first recall his notations and definitions . in this section ,
let @xmath672 be the set of closed points of @xmath67 of characteristic @xmath21 .
suppose that @xmath673 . throughout this section ,
the scheme @xmath4 over @xmath5 is assumed to satisfy the following condition : [ condition*]let @xmath4 be a regular , pure - dimensional scheme of dimension @xmath26 , flat and of finite type over @xmath59 . for any @xmath674 , each connected component @xmath675 of @xmath676 is a regular semi - stable family over @xmath677 or over the integral closure of @xmath677 in @xmath678 . here
@xmath677 is the hensenlization of @xmath5 at @xmath110 , and a scheme flat and of finite type over a discrete valuation ring is called a regular semi - stable family if it is regular , its generic fiber is smooth , and its special fiber is a normal crossing divisor .
[ definitionof nu]define @xmath682 to be the tale sheaf on @xmath362 @xmath683 here @xmath584 is the map in the gersten complex of logarithmic de rham - witt sheaves ( see @xcite or lemma [ thm : mktolrw ] above ) .
define @xmath684 on @xmath362 .
we recall sato s definition of the map @xmath685 since it is a map of sheaves , it can be defined on the stalks at points in @xmath362 , i.e. , points of characteristic @xmath21 .
let @xmath686 and @xmath674 be its image in @xmath59 . after replacing @xmath5 by @xmath687 and @xmath4 by @xmath688 ( and then replacing @xmath687 by its integral closure in one of the irreducible components of @xmath689 , if necessary , depending the condition ( [ condition * ] ) ) , we can assume that @xmath4 is a regular semi - stable family over an henselian discrete valuation ring @xmath690 with function field @xmath108 of characteristic @xmath57 and residue field @xmath6 of characteristic @xmath21 .
let @xmath691 be the embedding of points in @xmath692 and @xmath693 , respectively .
then there is a diagram : @xmath694 & \displaystyle\bigoplus_{x_k^{(0)}}\iota^*r^n ( i_{x})_*\mu_{p^r}^{\otimes n } \ar[d]^-{\partial_1}\ar[r]^-{\partial_2 } & \displaystyle\bigoplus_{x_k^{(1)}}\iota^*r^{n-1 } ( i_{x})_*\mu^{\otimes n-1}_{p^r}\ar[d]^-{\partial_3}\\ 0 \ar[r ] & \nu_{x_k , r}^{n-1 } \ar[r ] & \displaystyle\bigoplus_{x_k^{(0)}}i_{y*}\nu_{y , r}^{n-1 } \ar[r]^-{\partial } & \displaystyle\bigoplus_{x_k^{(1)}}i_{y*}\nu_{y , r}^{n-2}.}\ ] ] here all the @xmath695 s are boundary maps defined in @xcite ( or see ( * ? ? ?
* lemma 3.2.4 ) ) .
the upper row is part of the gersten complex of @xmath48 on @xmath4 .
the lower row is exact by definition .
the square is anti - commutative by kato , @xcite .
hence it induces a map @xmath696 define the map @xmath697 as the composition : @xmath698\overset{\sim}{\underset{\diamond}\longrightarrow}(\iota_*m_r^n)[-n ] \stackrel{\sigma^n_{x , r}}\longrightarrow \iota_*\nu^{n-1}_{y , r}[-n].\ ] ] here @xmath699 is the adjunction @xmath700 .
[ definition : satocomplex ] if @xmath701 , and let @xmath11 be a complex fitted into the following distinguished triangle @xmath702 \ar[r]^-g & { { \mathfrak i}}_r(n)_x \ar[r]^-t & \tau_{\le n}rj_*\mu_{p^r}^{\otimes n } \ar[rr]^{\sigma_{x , r}(n ) } & & \iota_*\nu^{n-1}_{y , r}[-n].}\]]then @xmath11 is concentrated in @xmath703 $ ] . moreover ,
the triple @xmath704 is unique up to a unique isomorphism and @xmath705 is determined by @xmath706 .
the following proposition and theorem compare @xmath11 with bloch s cycle complex , which partially proves a conjecture made by sato ( * ? ? ?
* conjecture 1.4.1 ) .
if @xmath4 is smooth over @xmath5 , the conclusion ( without truncation ) was proved by geisser in ( * ? ? ?
* theorem 1.3 ) .
[ lemma : hcgtosato ] let @xmath145 be a normal crossing variety over a perfect field of characteristic @xmath21 and @xmath714 , then @xmath715 admites gersten resolution , i.e. , the following sequence of tale sheaves is exact : @xmath716 & h_{d+n-1}({{{\mathbb{z}}}}^c_z / p^r(n ) ) \ar[r ] & \displaystyle\bigoplus_{z_{(d-1)}}r^{1-d - n}i_{z*}{{{\mathbb{z}}}}^c_z / p^r(n ) \ar[r ] & ... } \ ] ] @xmath717&\displaystyle\bigoplus_{z_{(n+1)}}r^{-2n-1}i_{z*}{{{\mathbb{z}}}}^c_z / p^r(n ) \ar[r ] & \displaystyle\bigoplus_{z_{(n ) } } r^{-2n}i_{z*}{{{\mathbb{z}}}}^c_z / p^r(n)\ar[r ] & 0.}\ ] ] the above complex ( without the first two terms ) is @xmath718 , the gersten complex of @xmath719 . by lemma [ lemma : hcgtomk ] and theorem [ thm : mktolrw ] , we know that @xmath720 is isomorphic to @xmath721 . for normal crossing variety @xmath145 ,
the latter is quasi - isomorphic to @xmath722 $ ] by sato , ( * ? ? ?
* corollary 2.5.2 ) .
hence , @xmath720 is exact except at degree @xmath723 . consider the niveau spectral sequence of higher chow groups in @xmath10-coefficients in ( [ definitionspectralsequence ] )
, we see that the cohomology of @xmath720 at degree @xmath723 is exactly @xmath719 .
[ lemma2:cycletosato]let @xmath97 be a discrete valuation field of characteristic 0 with perfect residue field @xmath6 of characteristic @xmath21 . then the following square is commutative : @xmath724^{\delta}\ar[d]^-{\kappa } & ch^{m-1}(k , m-1,{{{\mathbb{z}}}}/p^r)\ar[d]^-{\psi}\\ h^m(f , \mu^{\otimes m}_{p^r } ) \ar[r]^-{\partial } & \nu^{m-1}_{k , r}(k).}\ ] ] here the map @xmath725 is the map defined in @xcite , and @xmath726 is the defined in section 2 .
the map @xmath345 is defined as the composition : @xmath727 & k_{m-1}^m(k)/p^r \ar[r]^-{d\log}_{\cong } & \nu_{k , r}^{m-1}(k),}\ ] ] and @xmath584 is defined as the composition ( @xcite , ( 3.2.3 ) ) @xmath728 ^ -{{\mathfrak g}}_{\cong } \ar[r]^-{\partial^{tame } } k_m^m(f)/p^r & k_{m-1}^m(k)/p^r \ar[r]^-{d\log } & \nu^{m-1}_{k , r}(k).}\ ] ] here @xmath729 is the tame symbol of milnor k - theory , @xmath730 is the bloch
kato isomorphism , and @xmath731 is galois symbol . from the definition of the maps , the square in the lemma is actually formed from the following diagram @xmath732^{\kappa}\ar[r]^{\delta}\ar[d]^{\chi } & ch^{m-1}(k , m-1,{{{\mathbb{z}}}}/p^r)\ar[d]\ar[rd]^{\psi } & \\
h^m(f , \mu^{\otimes m}_{p^r } ) & k_m^m(f)/p^r\ar[l]^{\cong}_{{\mathfrak g } } \ar[r]^{\partial^{tame } } & k^m_{m-1}(k)/p^r \ar[r]^{d\log}_{\cong } & \nu^{m-1}_{k , r}(k).}\ ] ] the map @xmath402 is the nesterenko
suslin isomorphism .
so we have to show the trapezoid is commutative .
the triangle on the right commutes by definition .
the square in the middle commutes by geisser
levine , @xcite , lemma 3.2 .
so we only have to analyze the triangle on the left . since all the three maps @xmath733 respect the products of higher chow groups , milnor k - theory and galois cohomology , so it suffices to show that @xmath734 for @xmath735 .
but in this case , all the three groups are identified with @xmath736 , and the three maps are just identity .
hence we proved the lemma .
\(1 ) @xmath550 ( 2 ) : from theorem [ thm : blconjecturetorsion ] , we see that @xmath737)\cong \tau_{\le n}r\epsilon_*({{{\mathbb{z}}}}^c_x / p^r(d - n)[-2d])\cong \tau_{\le n}r\epsilon_*(\tau_{\le n}{{{\mathbb{z}}}}^c_x / p^r(d - n)[-2d]).\ ] ] hence , apply @xmath738 on the isomorphism in ( 1 ) , we obtain the one in ( 2 ) .
\(1 ) the idea of the proof is to show that @xmath739 $ ] satisfies a triangle similar to the one for @xmath11 , and then show that there is a map of these two triangles which induces isomorphisms on the cohomologies up to degree @xmath37 .
let @xmath740 , m_u={{{\mathbb{z}}}}_u^{c}/p^r(d - n)^{\text{zar}}[-2d ] , m_y={{{\mathbb{z}}}}_y^{c}/p^r(d - n)^{\text{zar}}[1 - 2d].\ ] ] then by theorem [ thm : localization ] , there is a distinguish triangle of complexes of zariski sheaves : @xmath741 since @xmath151 is exact and @xmath742 , we get a distinguish triangle of complexes of etale sheaves @xmath743 \to \epsilon^*j_*m_u \stackrel{\delta}\longrightarrow \epsilon^*\iota_*m_y .\label{distinguishtriangle1}\ ] ] consider the following diagram : @xmath744 & ( \iota_*\tau_{\le n}\widetilde{\nu}_{y , r}(d - n))[1 - 2d]\ar[l]^{\cong } \ar[r]^{\cong } & h^n(\iota_*\widetilde{\nu}_{y , r}(d - n)[1 - 2d])[-n]\\ \epsilon^*\iota_*m_y \ar[u]^{\cong } & \tau_{\le n}\epsilon^*\iota_*m_y \ar[r]\ar[l ] \ar[u]^{\cong } & h^n(\epsilon^*\iota_*m_y)[-n]\ar[u]^{\cong}\\ \epsilon^*j_*m_u\ar[u]^{\delta } \ar[d ] & \ar @ { } [ dr]|{\sharp } \tau_{\le n } \epsilon^*j_*m_u \ar[u ] \ar[l]\ar[r ] \ar[d]^{\cong } & h^n(\epsilon^*j_*m_u)\ar[u]^{\delta}\ar[d]^{\cong}[-n]\\ rj_*\epsilon^*m_u \ar[d ] & \tau_{\le n}rj_*\epsilon^*m_u \ar[d]^{\cong}\ar[r]\ar[l ] & ( r^nj_*\epsilon^*m_u)[-n ] \ar[d]_{\kappa}^{\cong}\\ rj_*\mu_{p^r}^{\otimes d } & \tau_{\le n}rj_*\mu_{p^r}^{\otimes d } \ar[r ] \ar[l ] & ( r^nj_*\mu_{p^r}^{\otimes d})[-n ] . } \ ] ] the maps on the left column are defined as follows : the first one is the map defined in theorem [ thm : cycletolrw ] , which is a quasi - isomorphism ( note that @xmath745 ) .
the second map is the boundary map of localization sequence of higher chow groups as above .
the third map @xmath746 is induced by universal property of the @xmath747tale sheafification @xmath151 .
the fourth map is defined in @xcite .
the vertical maps in the second and third column are induced by taking truncation and cohomology of maps in the first column , respectively .
hence all the squares commute .
the maps in the first row are isomorphisms , since it is the gersten resolution of @xmath748 $ ] on normal crossing varieties over perfect fields of characteristic @xmath21 ( * ? ? ?
* corollary 2.2.5(1 ) ) .
in particular , there is an isomorphism of sheaves @xmath749),\ ] ] which , composing with the isomorphism @xmath750),\ ] ] induces an isomorphism @xmath751 on the other hand , the beilinson
lichtenbaum conjecture ( theorem [ thm : blconjecturetorsion ] ) claims that @xmath752 therefore , @xmath753 i.e. , in the square @xmath754 , the vertical maps are isomorphisms . at last ,
the vertical maps in the lower right square are isomorphisms , by theorem [ thm : kummerisomorphism ] . in conclusion ,
we have the following diagram @xmath755 \ar[r ] & rj_*\epsilon^*m_u \ar[rr ] & & \epsilon^*\iota_*m_y \\ & \tau_{\le n}rj_*\epsilon^*m_u \ar[r]\ar[u ] \ar[d]^-{\cong } & ( r^nj_*\epsilon^*m_u ) [ -n ] \ar[r]^{\delta}\ar[d]_{\kappa}^{\cong } \ar @ { } [ dr]|{\diamondsuit } & h^n(\epsilon^*\iota_*m_y)[-n ] \ar[d]_{\psi}^{\cong}\ar[u]^{\cong}\\ { { \mathfrak i}}_r(n)_x \ar[r ] & \tau_{\le n}rj_*\mu_{p^r}^{\otimes n } \ar[r ] & ( r^nj_*\mu_{p^r}^{\otimes n})[-n]\ar[r]^{\partial } & \iota _ * \nu_{y , r}^{n-1}[-n ] . } \label{section3keysquare2}\ ] ] here @xmath584 is the map @xmath756 defined above .
the top row and bottom row ( with @xmath757 $ ] eliminated ) are distinguished triangles .
the upper square and lower left square commute as above .
if we can show that the square @xmath758 commutes , then it implies that there is a map @xmath759\ ] ] which induces isomorphisms on cohomologies of @xmath49 $ ] and @xmath11 up to degree @xmath37 , hence finishes the proof .
now let us prove that the square @xmath758 commutes . for @xmath760 ,
the right two terms vanish , hence the square @xmath758 commutes .
if @xmath761 , we can take the henselization @xmath762 of @xmath59 at @xmath110 , and prove the commutativity for all the irreducible components of @xmath763 over @xmath764 ( or replace @xmath765 by its integral closure in an irreducible component of @xmath766 , depending on the condition ( [ condition * ] ) ) . from now on ,
we assume that @xmath54 is a henselian discrete valuation ring with field of fraction @xmath108 of characteristic 0 and perfect residue field @xmath6 of characteristic @xmath21 , and @xmath4 is flat , regular semistable family over @xmath54 .
then @xmath767 is smooth over @xmath108 and @xmath768 is a normal crossing variety over @xmath6 .
since we are only left to show the commutativity of square @xmath758 in the diagram ( [ section3keysquare2 ] ) for points in @xmath145 , we can apply @xmath769 on the square , and obtain the following diagram : @xmath770^{\delta } \ar[d]^{\kappa } & h^n(\epsilon^*m_z ) \ar[d]^{\psi}\\
\iota^*r^nj_*\mu^{\otimes n}_{p^r } \ar[r]^{\partial } & \nu^{n-1}_{z , r}.}\label{section3keysquare3}\ ] ] to show that this square commutes , we will use gersten complex to replace all the maps by maps of functions fields of the schemes , and show compatibility of those maps .
first , we have the following isomorphism of complexes : @xmath771 \ar[d]_{\cong}^{\kappa } & \displaystyle\bigoplus_{v^{(0)}}\iota^*r^n(i_{x})_*(\epsilon^*m_x ) \ar[r]\ar[d]_{\cong } & \displaystyle\bigoplus_{v^{(1)}}\iota^*r^{n-1}(i_{x})_*(\epsilon^*m_x)\ar[d]_{\cong } \ar[r ] &
\\ \iota^*r^nj_*\mu^{\otimes n}_{p^r } \ar[r ] & \displaystyle\bigoplus_{v^{(0)}}\iota^*r^n(i_{x})_*\mu^{\otimes n}_{p^r } \ar[r ] & \displaystyle\bigoplus_{v^{(1)}}\iota^*r^{n-1}(i_{x})_*\mu^{\otimes ( n-1)}_{p^r } \ar[r ] & .}\ ] ] here @xmath772 is the embedding into @xmath4 , and @xmath773 $ ] if @xmath774 .
secondly , by @xcite , @xmath775 on normal crossing variety admits gersten resolution . by lemma [ lemma : hcgtosato ] ,
higher chow groups on normal crossing varieties admits gersten resolution as well .
more precisely , we have the following commutative diagram ( [ thm : cycletolrw ] ) : @xmath134 & h^n(\epsilon^*m_z ) \ar[r ] \ar[d]^{\psi}_{\cong } & \displaystyle\bigoplus_{z^{(0)}}r^ni_{z*}(\epsilon^*m_z ) \ar[r ] \ar[d]_{\cong } & \displaystyle\bigoplus_{z^{(1 ) } } r^{n-1}i_{z*}(\epsilon^*m_z)\ar[r]\ar[d]_{\cong}&\\ 0 \ar[r ] & \nu^{n-1}_{z , r } \ar[r ] & \displaystyle\bigoplus_{z^{(0)}}i_{z*}\nu_{r , z}^{n-1 } \ar[r ] & \displaystyle\bigoplus_{z^{(1 ) } } i_{z*}\nu_{r , z}^{n-2 } \ar[r ] & } \ ] ] such that the rows are exact . here
@xmath776 are points in @xmath145 , @xmath777 $ ] on @xmath778 with @xmath779 . in particular , the map @xmath780 is injective .
thirdly , by functoriality of cycle complex , the following diagram is commutative , @xmath781 \ar[d]^{\delta } & \displaystyle\bigoplus_{v^{(0)}}\iota^*r^ni_{x*}(\epsilon^*m_x ) \ar[r ] \ar[d ] & \displaystyle\bigoplus_{v^{(1)}}\iota^*r^{n-1}i_{x*}(\epsilon^*m_x ) \ar[d ] \\0 \ar[r ] & h^n(\epsilon^*m_z ) \ar[r ] & \displaystyle\bigoplus_{z^{(0)}}r^ni_{z*}(\epsilon^*m_z ) \ar[r ] & \displaystyle\bigoplus_{z^{(1 ) } } r^{n-1}i_{z*}(\epsilon^*m_z).}\ ] ] finally , the map @xmath584 is induced by the diagram in ( [ definition : sigma ] ) . to summarize the above analysis , let @xmath782 .
replacing @xmath4 by the component containing @xmath295 , we can assume that @xmath4 is irreducible , and @xmath783 is the embedding of the generic point . to prove the commutativity of square ( [ section3keysquare3 ] )
, it suffices to prove that the following square commutes ( here @xmath776 ) : @xmath784\ar[r ] & ( r^ni_{z*}\epsilon^*m_z)_{\bar{z } } \ar[d ] \\ ( \iota^*r^ni_{\eta*}\mu^{\otimes n}_{p^r})_{\bar{z } } \ar[r ] & ( i_{z*}\nu^{n-1}_{r , z})_{\bar{z } } .}\ ] ] since @xmath785 ) \stackrel{\natural}\cong h_{2d - n}^{{\operatorname{bm}}}(k(o^{sh}_{x,\bar{z } } ) , { { { \mathbb{z}}}}/p^r(d - n)),\ ] ] @xmath786 @xmath787)\stackrel{\natural}\cong h_{2d - n-1}^{{\operatorname{bm}}}(k(\bar{z } ) , { { { \mathbb{z}}}}^c / p^r(d - n)),\ ] ] @xmath788 here @xmath789 is the field of fraction of @xmath790 , and the isomorphisms @xmath791 follow from the beilinson
lichtenbaum conjecture ( theorem [ thm : blconjecturetorsion ] ) .
note that @xmath792 @xmath793 from lemma [ lemma2:cycletosato ] , we obtain the commutativity . from lemma [ lemma : gerstenconjecture ] , we see that the cycle complexes in the right hand side of the isomorphisms are acyclic at degree @xmath798 .
therefore , in the isomorphisms of theorem [ prop : cycletosato ] , the truncation in front of cycle complexes are not necessary . 99 hyman bass , john tate , _ the milnor ring of a global field , algebraic k - theory ii , _ lecture notes in math . 342 , spring 1972 , 349 - 466 .
spencer bloch , _ algebraic cycles and higher k - theory , _ adv . in math .
61 ( 1986 ) , no .
3 , 267 - 304 .
spencer bloch , arthur ogus , _ gersten conjecture and the homology of schemes _ , ann .
ser . 7 ( 1974 ) , 181 - 202 .
spencer bloch , kazuya kato , _ p - adic tale cohomology , _ inst .
hautes tudes sci .
63(1986 ) , 107 - 152 .
spencer bloch , stephen lichtenbaum , _ a spectral sequence for motivic cohomology , _ k - theory preprint , http://www.math.uiuc.edu/k-theory/0062/ jean - louis colliot - thelene , raymond hoobler , bruno kahn , _ the bloch ogus
gabber theorem _ , algebraic k - theory ( toronto , on , 1996 ) , 31 - 94 .
christopher deninger , _ duality in the tale cohomology of one - dimensional proper schemes and generalizations , _ math .
( 1987 ) , no . 3 , 529 - 541 .
kazuhiro fujiwara , _ a proof of the absolute purity conjecture ( after gabber ) _ , algebraic geometry 2000 , azumino ( hotaka ) , 153 - 183 .
thomas geisser , _ duality via cycle complexes , _ ann . of math .
172(2010 ) , no . 2 , 1095 - 1126 .
thomas geisser , _ motivic cohomology over dedekind rings , _ math
. z. 248 ( 2004 ) , no . 4 , 773 - 794 .
thomas geisser , _ motivic cohomology , k - theory and topological cyclic homology _
, handbook of @xmath108-theory .
vol . 1 , 2 , 193 - 234 , springer , berlin , 2005 .
thomas geisser , _
parshin s conjecture revisited , _
@xmath108-theory and noncommutative geometry , 413 - 425 , ems ser .
soc . , zrich , 2008 .
thomas geisser , marc levine , _ the bloch kato conjecture and a theorem of suslin voevodsky _ , j. reine angew .
( 2001 ) , 55 - 103 .
thomas geisser , marc levine , _ the k - theory of fields in characteristic @xmath21 , _ inv .
139 ( 2000 ) , 459 - 494 .
michel gros , noriyuki suwa , _
la conjecture de gersten pour les faisceaux de hodge - witt logarithmicique , _ duke math
. j. 57 ( 1988 ) , no . 2 , 615 - 628 .
uwe jannsen , shuji saito , kanetomo sato , _ tale duality for constructible sheaves on arithmetic schemes , _ k - theory preprint , http://www.math.uiuc.edu/k-theory/0946/ bruno kahn , _ algebraic k - theory , algebraic cycles and arithmetic geometry , _ handbook of @xmath108-theory .
vol . 1 , 2 , 351428 , springer , berlin , 2005 .
kazuya kato , _ a generalization of local class field theory by using k - groups ii ,
tokyo sect .
27(3 ) , 603 - 683 ( 1980 ) .
kazuya kato , _ milnor k - theory and the chow group of zero cycles , _ application of algebraic k - theory to algebraic geometry and number theory , contemp .
55 , part i , ameri . math .
soc . , providence , 1986 , 241 - 253 .
kazuya kato , _ a hasse principle for two - dimensional global fields _
( with an appendix by colliot - thlne , j .- l . ) , j. reine angew .
( 1986 ) 142 - 183 .
steven e. landsburg , _ relative chow groups , _
illinois j. math .
35 ( 1991 ) , no .
4 , 618641 .
marc levine , _ techniques of localization in the theory of algebraic cycles _ , j. alg .
10 , 299 - 363 ( 2001 ) .
marc levine , _ k - theory and motivic cohomology of schemes _ , k - theory preprint , http://www.math.uiuc.edu / k - theory/0336/. stephen lichtenbaum , _ the construction of weight - two arithmetic cohomology , _ invent .
88 ( 1987 ) , 183 - 215 .
james s. milne , _ etale cohomology _ , princeton math .
series 33 .
james s. milne , _ arithmetic duality theorems _ , perspectives in mathematics , 1 . academic press , 1986 .
thomas moser , _ a duality theorem for tale @xmath21-torsion sheaves on complete varieties over a finite field , _ composition math .
117 ( 1999 ) , no . 2 , 123 - 152 .
enric nart , _ the bloch complex in codimension one and arithmetic duality , _ j. number theory 32 ( 1989 ) , no . 3 , 321 - 331 .
p. nesterenko , andrei suslin , _ homology of the full linear group over a local ring , and milnor s k - theory , _ izv .
nauk sssr ser . mat .
53 ( 1989 ) , no . 1 , 121 - 146 .
daniel quillen , _
algebraic @xmath108-theory , i : higher @xmath108-theories , _
conf . , battelle memorial inst . ,
seattle , wash . , 1972 ) ,
lecture notes in math .
341 , springer , berlin 1973 .
markust rost , _ chow groups with coefficients , _ doc . math .
1(16 ) , 319 - 393 ( 1996 ) .
kanetomo sato , _ p - adic etale tate twists and arithmetic duality _
, with an appendix by kei hagihara .
sci . ecole norm .
( 4 ) 40 ( 2007 ) , no . 4 , 519 - 588 .
kanetomo sato , _ logarithmic hodeg
witt sheaves on normal crossing varieties _ , math . z. ( 2007 ) 257 : 707 - 743 .
michael spiess , _ artin - verdier duality for arithmetic surfaces , _ math .
( 1996 ) , no . 4 , 705 - 792 .
andrei suslin , vladimir voevodsky , _ bloch kato conjecture and motivic cohomology _ , the arithmetic and geometry of algebraic cycles ( banff , ab , 1998 ) , 117 - 189 .
carlo mazza , vladimir voevodsky , charles weibel,_lecture notes on motivic cohomology , _ clay mathematics monographs , 2 .
american mathematical society , providence , ri ; clay mathematics institute , cambridge , ma , 2006 . | we prove that there is a map from bloch s cycle complex to kato s complex of milnor k - theory , which induces a quasi - isomorphism from cycle complex mod @xmath0 to moser s complex of logarithmic de rham witt sheaves .
next we show that the truncation of bloch s cycle complex at @xmath1 is quasi - isomorphic to spiess dualizing complex . in the end , we prove that a weak form of the gersten conjecture implies that sato s dualizing complex is quasi - isomorphic to bloch s complex . |
one of the techniques to register ultra - high energy extensive air showers ( eas ) is measuring strength of radio pulse by antennas . unlike traditional techniques , including optic measurements of air shower propagation radio technique can operate in any atmospheric condition except during thunderstorm conditions for whole observation period , which dramatically increases effective time of air showers registration .
it is easier to use and much cheaper than other ground detectors in existing air showers array . the yakutsk array measured three components of air shower : the total charged component , the muon component and cherenkov radiation . from these components using average lateral distribution function ( ldf ) the integral characteristics of air shower , the total number of charged particles , the total number of muons and full flux of cherenkov light at the sea level are recovered .
all these shower characteristics are used for further model - free air shower energy estimation as shown in [ 1 ] .
cherenkov light registered at the sea level moreover is used to recover air shower longitudinal distribution and it characteristics , cascade curve and depth of maximum x@xmath4 [ 2][3 ] .
using this , in future is possible to find a relation between the characteristics of the radio emission and characteristics of the eas , including slope of the radio emission ldf with depth of maximum , as shown in [ 4 ] .
for the season 2008 - 2012 were recorded 600 air shower events with radio emission .
showers energy were above 3@xmath010@xmath1 ev , and zenith angle @xmath5 @xmath6 70@xmath7 .
for further analysis were selected only 421 showers , appropriate selection criteria of this paper .
therefore , for analysis at the yakutsk array we use following criteria : 1 .
the shower selected if adc prehistory contains radio pulse with amplitude 5 times more than noise level and pulse is localized within time gate equal to delay of `` master '' from small or large arrays . 2 .
extensive air shower axis must be within perimeter of central array with radius 600 m. zenith angle @xmath5 @xmath6 35@xmath7 .
azimuth angle @xmath8 chosen such a way as to exclude influence of polarization effect .
that is , the amplitude of the crossed antennas were equal or were nt go beyond limit ( 3 - 5 ) @xmath9 . with selected events
, we plotted dependence of maximum amplitude of radio pulse from zenith angle ( fig .
@xmath10 ) .
approximation curve is given by power function : @xmath11 in fig .
@xmath12 is shown dependence of maximum amplitude of radio signal from shower energy .
approximation is given by : @xmath13 as seen from fig .
@xmath14 , slope of the ldf changes with the distance . at large distance signal of radio emission attenuates
slowly . from equation ( 1 ) and ( 2 ) we derived formula for calculating energy in individual showers : 10@xmath2 ev , 4.38@xmath010@xmath2 ev and 1.32@xmath010@xmath3 ev.,scaledwidth=40.0% ] @xmath15 with : @xmath5 - zenith angle , r - distance of antennas to the shower axis , e@xmath16 - the primary particle energy . in fig .
@xmath17 dependence of radio emission ldf from depth of the air shower maximum .
the depth was determined by cherenkov detectors measurements of yakutsk array .
the slopes were determined from the ratio of the amplitudes , taken at 80 and 200 m ( fig .
@xmath14 ) for three energies : 1.73@xmath010@xmath2 ev , 4.38@xmath010@xmath2 ev and 1.32@xmath010@xmath3 ev . .
the slope was determined from the ratio of the amplitudes , taken at 80 and 200 m using averaged ldf and xmax from cherenkov light measurements .
, scaledwidth=40.0% ]
yakutsk array measurements are showing : a ) there is a correlation between measured maximum amplitude of radio signal and air shower energy determined measurements of the main components at observation level .
it follows from the formula ( 2 ) derived empirically from the joint consideration of radio signal amplitude and eas energy ; b ) the shape of ldf depends on the depth of air shower maximum x@xmath4 ( fig .
@xmath17 ) . | this paper presents the set of measurements of ultra - high energy air shower radio emission at frequency 32 mhz in period of 2008 - 2012 .
the showers are selected by geomagnetic and azimuth angles and then by the energy in three intervals : 3@xmath010@xmath1 3@xmath010@xmath2 ev , 3@xmath010@xmath2 6@xmath010@xmath2 ev and 6@xmath010@xmath2 5@xmath010@xmath3 ev . in each energy interval
average lateral distribution function using mathematically averaged data from antennas with different directions are plotted . in the paper , using experimental data the dependence of radio signal averaged amplitude from geomagnetic angle , the shower axis distance and the energy are determined .
depth of maximum of cosmic ray showers xmax for the given energy range is evaluated .
the evaluation is made according qgsjet model calculations and average lateral distribution function shape . |
charge symmetry is a special case of the general flavour symmetry of qcd , which at its simplest distinguishes the proton and the neutron ( or @xmath6 and @xmath7 quarks ) .
it is , of course , trivially broken by the electromagnetic interaction , notably the coulomb force in comparisons of the @xmath8 and @xmath9 systems and by the magnetic interaction in the @xmath3 system .
other well known sources are the @xmath3 mass difference and @xmath10- as well as @xmath11-meson mixing . these in turn may be related to the up- and down - quark mass difference - the microscopic flavour symmetry breaking in qcd .
one might consider remarkable the fact that , although the relative quark mass difference is large ( @xmath12 ) , the symmetry breaking at the observable hadron level is two orders of magnitude smaller . in the mirror system
@xmath8 vs. @xmath9 csb has been studied for many decades @xcite , while its appearance in the @xmath3 system was first seen only a decade ago @xcite as the difference @xmath13 of elastic analyzing powers .
different csb observables have been seen in calculations to be sensitive to different combinations of sources .
for example , in @xmath3 scattering above 300 mev the @xmath3 mass difference in ope dominates , while at and below @xmath14 200 mev @xmath11 meson mixing and the magnetic interaction become about equally important@xcite .
in contrast , the @xmath8 vs. @xmath9 difference is dominated by @xmath11 meson mixing @xcite .
the csb effects in the @xmath3 system change the total isospin of the two nucleons ( class iv in the terminology of ref .
@xcite ) , whereas in @xmath8 and @xmath9 the isospin must be conserved ( class iii ) .
it is easily seen that in the reaction @xmath15 the isospin change @xmath16 implies a csb asymmetry of the unpolarized cross section about 90@xmath17 and vice versa .
namely , due to the generalized pauli principle and conservation of the angular momentum and parity , with isospin one initial states for odd @xmath18 only singlet - even initial states are possible and for even @xmath18 only triplet - odd states .
the presence of some isospin zero component in the initial state introduces opposite spin - parity assignments and the initial spin states will then have both parities involved .
this minor asymmetry is being measured in an on - going experiment at triumf @xcite . since class iv forces mentioned above
change the isospin , as an initial state interaction they can quite naturally give rise to @xmath19 final states even from initial @xmath20 @xmath3 states .
further , although in a purely nucleonic basis @xmath10 mixing force conserves the isospin ( class iii ) , it can cause a @xmath20 @xmath21 admixture even in initial isospin zero states @xcite and thus contribute to pion production from these states .
in addition @xmath10 mixing can contribute very explicitly in the final actual production vertex in the form of production first of an isoscalar @xmath22 meson which then transforms into a pion . of traditional csb mechanisms in pion production @xmath10 mixing
is important and was seen to dominate at threshold @xcite , while the @xmath3 mass difference becomes more important at higher energies , where the two @xmath10 mixing mechanisms described above tend to cancel @xcite . exploiting the fact that chiral symmetry gives predictions for effects arising from the small but explicit breaking of this symmetry generated by the quark masses , ref . @xcite
employed a csb effective lagrangian based on chiral symmetry and including the @xmath6 and @xmath7 quark mass differences @xmath23 to describe pion - nucleon @xmath24-wave rescattering in @xmath25 .
this rescattering dominates isospin conserving production at threshold . here
@xmath26 represents the pauli matrices in isospin space , @xmath27 mev is the pion decay constant , and @xmath28 , though for simplicity @xmath29 was used .
it is important to note that the pion - nucleon interaction in the second term is linked by chiral symmetry to the first term , which in turn is directly related to the neutron - proton mass difference giving @xmath30 . here
@xmath31 is an electromagnetic `` hard photon '' exchange contribution to the neutron - proton mass difference also related to the isospin violating pion - nucleon interaction in the effective lagrangian @xmath32 this csb rescattering mechanism is presented in fig .
1a . above
, the quark mass difference is by no means small as compared with their sum but rather @xmath33 with @xmath34 , although the individual parameters @xmath35 and @xmath31 are not completely uniquely determined and remain model dependent .
their values have reasonable ranges 23 mev and ( 0.5)(1.5 ) mev , respectively . at the triumf experiment energy 279.5 mev the integrated forward - backward asymmetry @xmath36 } { \int_0^\pi d\omega \ ; \sigma(\theta)}. \label{asym}\ ] ]
was obtained in ref .
@xcite in terms of the above mass differences as @xmath37 where the conventional meson contributions in the first term have been taken from ref .
@xcite . due to the dominance of the @xmath10 mixing
, the uncertainties in its strength directly scale that term .
apart from minor pionic effects due to the @xmath3 mass difference , it is proportional to @xmath38 .
the values @xmath39 and @xmath40 mev@xmath41 have been used for this result .
( in addition , the analogous effect from @xmath42 with the same coupling and the mixing matrix element 5500 mev@xmath41 was icluded @xcite . )
since in eq .
( [ afb ] ) @xmath35 and @xmath31 add constructively , the latter term arising from the new effective lagrangian easily dominates the total @xmath43 making it change sign . with the given first term , the total @xmath43 would vary in the range 0.30.6% .
it is unlikely that the first term is an overestimate , so the range might rather be even more positive . with the anticipated experimental resolution of 0.12% @xcite this is a significant result .
if the uncertainties in the @xmath10 mechanisms ( notably the @xmath44 coupling constant ) can be solved , csb pion production could set constraints also on @xmath35 and @xmath31 .
since the new effective csb interaction gave such a large contribution in pion production , one is prone to worry how much this @xmath45 rescattering might contribute to elastic @xmath1 scattering possibly breaking the consensus there . in ref . @xcite the csb @xmath1 two - pion exchange interaction depicted in fig .
1b was shown in lowest order ( in the static approximation for baryons ) to be of the form of class iii @xmath46 [ \mu^2+({\bf k } -{\bf q}/2)^2]},\ ] ] where @xmath47 is the pion - nucleon coupling constant and @xmath48 the pion mass . a similar expression resulted for exchanges involving an intermediate @xmath49 isobar ( fig . 1c ) .
contrary to pion production , in this case , although the integrals and parameters are rather large , the two mass differences cancel to a large extent leaving only rather small contributions to the difference between @xmath9 and @xmath8 scattering . to see this , one need only compare the range of the factor @xmath50 [ + 0.8(-0.2 ) mev ] with the range of @xmath51 [ 2.23.4 mev ]
correspondingly , the contributions to the scattering length and effective range differences @xmath4 and @xmath5 vary between 0.17 fm and 0.05 fm , and 0.003 and 0.001 fm , respectively .
also the contribution from this source to the @xmath52h@xmath52he binding energy difference is expected to be small .
99 g. a. miller , b. m. k. nefkens and i. slaus , phys .
rep . * 194 * , 1 ( 1990 ) .
r. abegg _ et al .
lett . * 56 * , 2571 ( 1986 ) , phys .
d * 39 * , 2464 ( 1989 ) ; l. d. knutson _
lett . * 66 * , 1410 ( 1991 ) ; s. e. vigdor _
c * 46 * , 410 ( 1992 ) ; j. zhao _ et al .
c * 57 * , 2126 ( 1998 ) .
a. g. williams , a. w. thomas and g. a. miller , phys .
c * 36 * , 1956 ( 1987 ) ; m. j. iqbal and j. a. niskanen , phys .
c * 38 * , 2259 ( 1988 ) .
s. a. coon and r. c. barrett , phys .
c * 36 * , 2189 ( 1987 ) . g. f. de tramond , proc .
symposium / workshop on spin and symmetries , triumf , vancouver , june 30july 2 , 1989 , tri-89 - 5 , eds . w. d. ramsay and w. t. h. van oers , p. 235 .
e. m. henley and g. a. miller , in _ mesons and nuclei _ ,
i , eds . m. rho and d. h. wilkinson , north holland ( amsterdam 1979 ) .
triumf experiment e704 , spokespersons a. opper and e. korkmaz .
j. a. niskanen , few - body systems * 26 * , 241 ( 1999 ) .
j. a. niskanen , m. sebestyen , and a. w. thomas , phys .
c * 38 * , 838 ( 1988 ) .
j. a. niskanen , proc . of meson2000 workshop ,
cracow , may 2000 , acta physica polonica * b 31 * , 2683 ( 2000 ) .
u. van kolck , j. a. niskanen and g. a. miller , phys .
b 493 , 65 ( 2000 ) ; nucl - th/0006042 .
j. a. niskanen , phys .
c * 65 * , 037001 ( 2002 ) ; nucl - th/0108015 . | chiral effective field theory predicts a charge symmetry violating ( csb ) amplitude for pion - nucleon scattering .
this mechanism provides a very large contribution also to the csb forward - backward asymmetry in the angular distribution of the reaction @xmath0 .
this contribution was so large that it had a potential to cause a large effect also in csb elastic @xmath1 scattering and to disturb its present understanding .
however , it can be seen that , contrary to pion production , in this case the @xmath2-quark mass difference and electromagnetic contribution to the @xmath3-mass difference tend to cancel causing the total effect in the effective range parameters @xmath4 and @xmath5 to be relatively small . in the lowest order and within the static approximation for the nucleons csb pion - nucleon rescattering does not influence @xmath3 scattering .
= by -1 |
little has changed in our understanding of star formation in cooling flows since the the topic was last reviewed at the meeting on cooling flows at haifa in 1996 .
however , the nature of the so - called `` cooling flow problem '' has changed dramatically since then .
this controversial problem concerns the deficit between the large cooling rates of the kev gas in the centers of clusters and the much smaller star formation rates observed in central cluster galaxies located at the bases of cooling flows ( fabian 1994 ) .
while the cooling rates based on _ einstein _ and _ rosat _ x - ray observations ( but not asca , see makishima this conference ) were claimed to be tens to hundreds of solar masses per year , the star formation rates are generally between a few to several tens of solar masses per year . in spite of recent detections of molecular gas in cooling flows ( edge 2001 )
, this apparent violation of mass continuity can not be reconciled by a repository of molecular gas clouds or star formation with the local initial mass function in central cluster galaxies .
adherents of the cooling flow paradigm dodged the problem by appealing to a repository of elusive matter , such as low mass stars or a mist of cold clouds .
others proposed heating mechanisms ( e.g. , active nuclei , or agn , heat conduction , cosmic rays , mergers , magnetic reconnection , etc . ) with the potential to inject enough energy into the kev gas to balance radiative losses .
nevertheless , these proposals generally suffered various problems , such as the need for fine tuning , but mostly they lacked observational support .
two recent developments have dramatically changed our view of cooling flows .
first , xmm-_newton _ grating spectra of the critical soft x - ray band failed to detect the emission lines that dominate cooling below 2 kev at the predicted levels ( peterson et al .
2003 , & this conference ) .
although the spectra do not exclude cooling entirely , they limit the amount of gas cooling below x - ray temperatures ( where it is available to fuel star formation ) to be @xmath0 times less than the predicted levels .
the spectra imply that most of the cooling gas is maintained above @xmath1 kev ( i.e. , it is being reheated ) , or that it is cooling without an obvious spectroscopic signature ( fabian et al 2000b , peterson et al .
similar conclusions have been reached using moderate resolution ccd spectroscopy from _ chandra _ and xmm-_newton _
( mcnamara et al .
2000 , david et al .
2001 , molendi & pizzolato 2001 , bhringer et al .
2001 , blanton et al .
2003 ) .
secondly , strong interactions ( carilli et al . 1994 , bhringer et al . 1993 ) between radio sources and the intracluster medium are now commonly seen in _
chandra _ images ( mcnamara et al .
2000 , fabian et al .
2000 , and see mcnamara 2002 , and nulsen et al .
2003 for reviews ) .
these interactions are creating x - ray surface brightness depressions or cavities that , like bubbles in soda water , move buoyantly through the intracluster medium .
the bubbles in some ( but not all ) systems , contain enough energy to balance radiative losses emerging from the centers of clusters in the x - ray band .
this , along with the discovery of very short central cooling timescales , and at the same time , the lack of evidence for strong cooling below x - ray temperatures , have renewed interest in feedback - driven heating mechanisms capable of balancing radiation losses ( nulsen , this conference ) .
it is now an established fact that the kev gas in clusters with cooling times approaching 100 myr are frequently associated with the sites of star formation .
therefore , this star formation may have been fueled during the cooling phase of the feed - back loop .
i list below a few of the key observational facts about star formation in cooling flows .
( i use the term `` cooling flow '' to refer to clusters with short central cooling times . ) for more detailed discussions and additional references , see recent reviews by bhringer et al .
( 2001 ) , crawford ( 2003 & this conference ) , fabian ( 1994 ) , and mcnamara ( 1997 , 2002 ) . 1
. a trend exists between cooling flows and the occurrence and amplitude of blue color excesses associated with star formation in central cluster galaxies ( johnstone et al .
1987 , mcnamara & oconnell 1989 , cardiel et al .
1998 , crawford et al . 1999 ) .
_ chandra _ has shown that the regions of star formation are associated with bright lumps and filaments of gas whose radiative cooling times approach @xmath2 yr ( mcnamara et al . 2000 , fabian et al .
2001 , mcnamara et al .
2004 , blanton et al .
abell 1795 , shown in figure 1 , is a good example ( fabian 2001 ) .
2 . bright
, spatially extended nebular emission is seen preferentially in clusters ( e.g. perseus ) with central cooling times below @xmath3 gyr ( hu 1988 ; heckman et al .
recent comparisons between optical emission line maps and _ chandra _ x - ray maps have shown spatial correlations between nebular emission and bright lumps and filaments of gas ( figure 1 ) where the radiative cooling time approaches @xmath4 yr ( fabian et al . 2003 , blanton et al . 2001 ) .
the star formation histories vary between short duration bursts @xmath5 myr of age , or more extended episodes lasting between 0.1 - 1 gyr ; such histories are inconsistent with steady cooling and accretion that has endured for the ages of clusters ( allen 1995 , crawford et al 1999 , mcnamara 1997 , mcnamara et al . 2004 ) .
the short duration bursts are , in some cases , triggered by the central radio source ( mcnamara & oconnell 1993 ) .
a clean demonstration of equality between the level of cooling and star formation plus its associated gas would be a critical test of the new , self - regulated cooling paradigm .
more importantly , if it can be shown that the cooling upper limits are systematically below the star formation levels , a possibility that is now within reach , we would be in a position to reject the link between cooling and star formation , with some measure of confidence .
cooling rates have now been estimated for several clusters with _
chandra _ and xmm-_newton_. as i discussed above , they are systematically below the classical values .
this trend is shown in figure 2 , where i plot cooling rate ( @xmath6 ) versus star formation rate ( sfr ) , in solar masses per year , for five clusters .
the solid squares show the morphological cooling rates from the _ rosat / einstein _ era , calculated essentially by dividing the central gas mass by the cooling time .
the modern cooling rates fall well below these values and are shown along with their measurement uncertainties . only abell 1795 and abell 2597 have independent cooling measurements from _ fuse _ spectra of the o vi @xmath7 feature ( oegerle et al .
2001 ) . in both cases ,
the ultraviolet cooling rates , shown as filled dots in figure 2 , lie within the uncertainties of the x - ray cooling rates , represented as vertical bars .
the cooling rates for the remaining objects are presented as upper limits because the _ chandra_-acis spectral cooling rates represent the `` maximum cooling '' models consistent with the data ( wise et al .
2004 ) , and they should not be misconstrued as the unique spectral signatures of cooling below one or two kev . in other words , in spite of the very short cooling time of the kev gas , single temperature model plasmas provided acceptable fits to the data at each radius .
the star formation rates owe their uncertainties ( horizontal bars ) primarily to the difficult problem of measuring the intrinsic colors of young accretion populations against the background glare of the cd galaxy , mismeasuring extinction , and degeneracies in the population models themselves ( cf . , mcnamara et al .
2004 ) .
figure 2 shows that the cooling and star formation rates are generally converging , and in some cases they agree to within the ( substantial ) measurement errors .
this is evident in abell 1068 , abell 2597 , and abell 1795 .
while i do nt regard this as proof positive that the cooling flow problem has been solved , we are clearly on the right track . the challenges ahead include improving our understanding of the star formation rates and histories , and obtaining higher precision ( deeper ) x - ray spectra capable of turning the current upper limits into cooling detections , or excluding cooling entirely at levels below the the star formation rates .
essentially all indicators are telling us that the enormous radiation losses in cluster cores are balanced , or nearly so , by some form of heating .
clear - cut evidence now exists that interactions between radio sources and the surrounding gas are supplying some of this heat in many clusters and perhaps all of it in others ( see nulsen and blanton this conference ) .
cavities have been identified in nearly two dozen clusters over the past three years ( brzan et al .
the archetypes , hydra a ( mcnamara et al . 2000 ) and perseus ( bhringer et al . 1993 , fabian et al . 2000 ) , are typical of most systems : twin surface brightness depressions @xmath8 kpc in diameter lying at distances of @xmath9 kpc from the nucleus of the cd .
cavities have also been observed in giant elliptical galaxies , such as m84 ( finoguenov & jones 2001 ) , and groups , such as hgg 62 ( brzan et al .
cavity ages range between @xmath10 .
their enthalpy ranges between @xmath11 in isolated galaxies and groups to @xmath12 in rich clusters .
the total energy input from each agn outburst may be several times these figures when shocks are included ( fabian 2003 , forman et al .
the total energy deposited into the intracluster medium integrated over the lifetime of the agn can greatly exceed @xmath13 erg ( mcnamara et al .
2001 , and see nulsen et al .
2003 for a review ) .
nevertheless , while agn may be able to retard or quench cooling in many systems , they would do so with great difficulty in others .
only @xmath14 of clusters in the _ chandra _ archive have obvious cavity systems , and many of these systems contain too little energy to balance radiative losses at the current time ( brzan et al 2004 ) . other heating agents could be assisting the agn at balancing radiative losses . of these
, heat conduction between the cool cores and hot outer layers of clusters has received a good deal of attention ( voigt et al .
2002 ) . but
this model has problems . in order to work effectively
, most studies have concluded that conduction must be suppressed by several times the spitzer rate .
while this may be true in special cases , it can not be easily demonstrated observationally . in other systems ,
heat conduction acting alone at the spitzer rate is incapable of balancing radiative losses ( voigt et al .
2002 , zakamska & narayan 2003 , wise et al .
additional agents acting together to a greater or lesser degree , such as mergers ( motl et al .
2003 ) and supernovae associated with star formation ( mcnamara et al .
2004 ) , must be assisting the agn .
i have shown that the star formation rates estimated with optical , ultraviolet , and infrared observations are within factors of several of the new cooling limits , and in some cases , agree to within their errors . what we know about the star formation histories in cooling flows points to repeated bursts of star formation lasting @xmath15 yr .
in some cases , the starbursts are triggered by the central radio sources .
constant star formation over the ages of clusters , the history of star formation predicted by steady cooling flow models , is inconsistent with most data .
the overall picture of star formation in cooling flows is consistent with , and indeed has helped to shape , the emerging paradigm of self - regulated cooling in clusters . the x - ray , optical , and radio data taken together point to episodes of cooling and periodic reheating by radio outbursts and their associated bubbles .
this scenario ( e.g. , churazov et al . 2002 ) has received observational support from the newly - discovered trend between central x - ray luminosity and the instantaneous kinetic luminosity of bubbles in clusters ( brzan et al .
thermal conduction may be an additional element of the feed - back loop that regulates cooling ( ruszkowski & begelman 2002 , nulsen 2003 ) .
this emerging picture of cooling flows has broad implications for theories of structure formation and evolution .
feed - back processes may have been important during the early development of galactic bulges and their central black holes , and they may regulate the thermal balance of the hot gas in giant elliptical galaxies today ( nulsen , this conference ) . the possibility that such processes are active in large , bright , relatively nearby clusters provides a unique opportunity to test these models in detail .
i would like to acknowledge my colleagues michael wise , paul nulsen , craig sarazin , larry david , and chris carilli , and my graduate students laura brzan and david rafferty for their contributions to the work discussed here .
i would also like to thank the local organizing committee , thomas reiprich in particular , for hosting a great meeting , and noam soker for making it happen .
this research is supported by generous grants from nasa , the chandra x - ray center , the space telescope science institute , and the department of energy , including ltsa grant nag5 - 11025 and chandra general observer and archival research awards go0 - 1078a , ar2300 - 7x , and go1 - 2139x . | new x - ray observations from the _ chandra _ and xmm-_newton _ observatories have shown that cooling of the intracluster medium is occurring at rates that are now approaching the star formation rates measured in cd galaxies at the bases of cooling flows .
star formation proceeds in repeated episodes , possibly indicating an intermittent fuel supply . coupled with new evidence for
heating by agn , a new paradigm of self - regulated cooling and star formation in cluster cores is emerging . |
commutation relations of ( generalized ) pauli operators provide a skeleton for mutually unbiased bases , quantum entanglement and other conceptual ( or practical ) issues like quantum computing @xcite .
recently , an extensive study of commuting / non commuting rules has been undertaken , firstly in prime power dimensions @xmath11 of the hilbert space @xcite-@xcite , then in the smallest composite dimension @xmath12 @xcite .
commutation relations of two - qubit operators , and dually the incidence relations between maximal commuting sets of them , have been shown to fit the ( symplectic ) generalized quadrangle of order two , and several projective embeddings have been proposed @xcite . for higher - order pauli operators , the duality between the observables and their maximal commuting sets does not occur and the geometrical space of points / observables may exhibit several lines / sets passing through @xmath13-tuples of distinguished points @xcite . in this communication , one makes use of the maximal ideals of some ring @xmath14 ( possibly different from the modular ring @xmath3 ) as the gears of commutation relations . in particular , the incidence between the twelve lines of the sextit system fits the grid like structure of the projective line @xmath15 over the modular ring @xmath16 . in the higher composite dimensions explored so far @xmath17 , @xmath18 , @xmath19 and @xmath20 ,
the incidence of the maximal commuting sets is found to reproduce the projective line @xmath21 over rings @xmath22 ,
@xmath23 , @xmath24 and @xmath25 , respectively .
the unexpected irruption of the galois field of four elements @xmath26 , within the projective model of the two - qubit / qutrit system , seems to forbid an easy generalization to an arbitrary dimension @xmath1 .
there are indeed many ways of defining the quantum states ( let us call them qudits `` ) in a finite @xmath1-dimensional hilbert space .
one representation makes use of the unitary shift '' and clock "
operators @xmath27 and @xmath28 , with the actions @xmath29 , @xmath30 on the vectors @xmath31 of the hilbert space .
henceforth @xmath32 is a fixed @xmath1-th root of the unity . under matrix multiplication
, @xmath27 and @xmath28 generates the ( non - commutative ) pauli group @xmath33 from the basic relation @xmath34 . as a result
, elements of @xmath33 can be taken as @xmath35 , with @xmath36 , @xmath37 and @xmath38 in the ring @xmath3 @xcite-@xcite .
another representation of the pauli group is from tensor products of shift and clock actions in prime dimension @xcite .
the latter definition is favored in the theory of mutually unbiased bases @xcite and was used in our previous papers devoted to commutation relations @xcite-@xcite .
a condensation from the @xmath39 elements of the pauli group to @xmath0 pauli operators may also be achieved by taking the quotient of @xmath33 by its center @xmath40 ( the set of all operators which commute with every other one ) ref @xcite describes the commutation relations between operators in @xmath33 , and thus in @xmath41 , using vectors @xmath42 , their attached cyclic submodule @xmath43 and the points `` of the projective line @xmath44 an admissible vector @xmath45 is such that there exists another vector @xmath46 for which the matrix @xmath47 is invertible , which for a commutative ring is equivalent to have a determinant equal to a unit of the ring .
the equivalence class of @xmath45 is a free cyclic submodule @xmath48 , of order @xmath1 , and also a point '' of the projective line @xmath49 .
one reminds the geometrical structure of the projective line @xmath49 @xcite .
two distinct points @xmath48 and @xmath50 are called distant if @xmath51 equals a unit of the ring @xmath3 .
otherwise the two points belong to the same neighborhood .
another crucial concept organizes the vectors in @xmath52 : a perpendicular set @xmath53 is defined as @xmath54 in which two vectors @xmath45 and @xmath55 are perpendicular if @xmath56 .
note that two vectors within a cyclic submodule are mutually perpendicular . according to @xcite ,
operators in @xmath33 which commute with a fixed operator correspond to a perpendicular set - qubit systems ( see @xcite and sec 4.1 of @xcite ) . ] . using this analogy
, it seems natural to identify the elements of a free cyclic submodule , which are mutually perpendicular , with the maximal commuting sets of pauli operators , as we already did it implicitely in @xcite .
a posteriori one should not be surprised that the projective line @xmath57 fits the incidence relations between the maximal commuting sets of the sextit system . to complete the geometrical picture of commutation relations , one needs to identify the ( not necessarily admissible ) vectors of @xmath2 with the @xmath0 pauli operators .
let us summarize main results of @xcite : theorem 1 asserts that a free cyclic module @xmath50 containing a vector @xmath45 is contained in the perpendicular set @xmath53 .
only if @xmath45 is admissible the corresponding module equals @xmath53 .
it reinforces our interpretation that the maximal sets of mutually commuting operators [ corresponding to @xmath48 ] also define a base of operators [ corresponding to @xmath53 ] .
one immediate consequence concerns the application to mutually unbiased bases .
any two vectors in one base should be perpendicular , while any two vectors from distinct mutually unbiased bases should not . using two non - zero ( and admissible ) distinct vectors @xmath45 and @xmath60 , the two vector sets @xmath61 and @xmath62 are disjoint only if @xmath63 , i.e. if @xmath64 and @xmath45 , @xmath60 are not perpendicular .
this can not happen maximally since @xmath3 is a ring so that @xmath65 or @xmath66 may be zero divisors .
the maximal number of mutually unbiased bases in composite dimension may thus be reformulated as being the maximal number of such disjoint vector sets in the relevant ring .
if the dimension @xmath1 is the power of distinct primes @xmath67 , theorem 2 in @xcite provides quantitative results about ( a ) the number of points @xmath68 in which any vector @xmath45 lies , ( b ) the partitioning of @xmath53 as the corresponding set theoretic union of points @xmath48 and ( c ) the cardinality of @xmath53 .
one gets @xmath69 in which @xmath70 is a subset of the indices related to the decomposition of the entries of @xmath45 into their principal ideals .
the sextit system ( @xmath71 ) was investigated in our recent paper @xcite . in this dimension ,
the ( generalized ) pauli operators are defined as @xmath72 the orthonormal set of the qubits comprises the standard pauli matrices @xmath73 , where @xmath74 , @xmath75 , @xmath76 and @xmath77 , while the orthonormal set of the qutrits is taken as + @xmath78 , where @xmath79 is the @xmath80 unit matrix , @xmath81 , @xmath82 , @xmath83 , @xmath84 and @xmath85 .
the sextit operators can be conveniently labelled as follows : @xmath86 , @xmath87 , @xmath88 , @xmath89 , @xmath90 , @xmath91 , , @xmath92 , @xmath93 , ,
@xmath95 , in which we singled out the three reference points @xmath96 , @xmath97 and @xmath98 .
then one can use the strategy already described in @xcite for @xmath99-qudit systems .
the pauli operators are identified with the vertices of a ( pauli ) graph and the commuting operators are identified with the edges .
the maximal cliques of the graph correspond to the maximal sets of mutually commuting operators . for the sextit system one
gets the twelve sets @xmath100 as emphasized in @xcite , the incidence between the maximal commuting sets leads to a @xmath101 grid - like structure isomorphic to the projective line over the ring @xmath102 .
a subset of the commutation structure of the operators is illustrated in fig [ multi ] .
let us illustrate the relationship between the pauli graph of sextits and the fine structure of the projective line @xmath15 .
operators @xmath103 belonging to the maximal sets are of three distinct types includes the operator @xmath103 itself and the unity operator @xcite . but for maximal commuting sets one usually ignores the unity operator which commutes with every other operator @xcite ] ( see also fig [ multi ] ) \(i ) @xmath103 is one of the reference points @xmath96 , @xmath97 or @xmath98 , lies in four sets and the number of points commuting with @xmath103 is @xmath104 , \(ii ) @xmath105 lies in three sets and @xmath106 , \(iii ) otherwise @xmath103 lies in a single set and @xmath107 .
these results clearly fits ( [ quant ] ) with @xmath12 , @xmath108 and @xmath109 .
analogous results are indeed obtained for square free dimensions @xmath110 and @xmath18 , so far explored .
the qudit system in dimension @xmath111 contains the even square @xmath112 . in this dimension ,
the ( generalized ) pauli operators are defined as @xmath113 one proceeds as for the sextit system , one determines the pauli graph of the @xmath114-dit and one extracts the maximal cliques .
the incidence between the corresponding maximal commuting sets is found to reproduce the projective line over the ring @xmath115 , of order @xmath116 with @xmath117 and @xmath118 .
operators @xmath103 belonging to the maximal sets still are found to be of three distinct types \(i ) @xmath103 is one of the reference points ( it includes @xmath79 in its tensor product ) , then one finds that @xmath103 lies in @xmath119 sets and @xmath120 .
\(ii ) @xmath103 includes @xmath121 in its tensor product , lies in @xmath122 sets and @xmath123 , \(iii ) otherwise @xmath103 lies in @xmath124 sets and @xmath125 .
the commutation relations within a perpendicular set @xmath126 of type ( i ) are illustrated in fig [ para ] .
it comprises three bundles of four lines each , organized in a parachute like structure .
the lines of a specific bundle intersect at three distinguished points , each one of type ( i ) .
the qudit system in dimension @xmath127 contains the odd square @xmath128 . in this dimension ,
the ( generalized ) pauli operators are defined as @xmath129 again one determines the pauli graph of the @xmath5-dit and one computes the maximal cliques . the incidence between the corresponding maximal commuting sets is found to reproduce the projective line @xmath21 over the ring @xmath130 , of order @xmath131 with @xmath108 and @xmath118 .
operators @xmath103 belonging to the maximal sets are found to be of five distinct types \(i ) @xmath103 is one of the three reference points containing @xmath132 in the tensor decomposition , it lies in @xmath133 sets and @xmath134 , \(ii ) @xmath103 lies in @xmath119 sets and @xmath135 , \(iii ) @xmath103 lies in @xmath136 sets and @xmath137 , \(iv ) @xmath103 lies in @xmath124 sets and @xmath138 , \(v ) otherwise @xmath103 lies in a single set and @xmath137 . the perpendicular set attached to a point of type ( i )
is illustrated in fig [ paranew ] .
the fine structure of the bundles increases in complexity compared to fig [ para ] , each one comprising four lines intersecting at five points , one of type ( i ) , two of type ( ii ) and the remaining two of type ( iii ) .
it has been found that commuting operators associated to composite qudits in dimension @xmath1 correspond to perpendicular vectors within the symplectic module @xmath2 .
moreover the maximal commuting sets reflect the set - theoretic structure of free cyclic submodules defined over some commutative ring @xmath14 , possibly distinct from the modular ring @xmath3 as soon as @xmath1 contains squares in the prime number decomposition .
an admissible vector , which defines such a submodule , is of two types @xcite ( a ) either one ( at least ) of its entries is a unit of the ring @xmath14 , or ( b ) both of its entries are zero divisors , not in the same maximal ideal of @xmath14 . thus the maximal ideals underlie the projective line @xcite and the commutation structure of qudit operators . in fig [ ideals ] we give a sketch of the interaction between maximal ideals of the rings @xmath139 ( corresponding to the qubit / qutrit system ) , @xmath140 ( corresponding to the qubit / two - qutrit system ) and @xmath141 corresponding to the two - qubit / qutrit system ) .
to some extent one can identify the factors of the qudit system with the maximal ideals , and the peculiar set theoretic union / intersection of them governs the whole commutation structure .
the ideals themselves have a ring structure .
for example the three ideals in ( c ) are subsets isomorphic to @xmath139 , @xmath142 and @xmath143 respectively .
the corresponding projective lines are @xmath144 , @xmath145 and @xmath146 grids .
the last grid exhibits a maximum number of four distant points , corresponding to the maximum number of mutually unbiased bases in dimension twelve .
further work should clarify whether a ring @xmath14 is attached to any composite qubit .
this could have application not only to mutually unbiased bases , but to quantum chemistry @xcite , quantum channels @xcite , the non abelian hidden subgroup problem @xcite and other quantum information processing problems , as well .
the authors acknowledge hans havlicek , metod saniga , peter pracna and maurice kibler for helpful interactions during the workshop finite projective geometries in quantum theory `` [ http://www.ta3.sk/ msaniga / quantgeom.htm ] , held in tatranska - lomnica in august 2007 .
the work was supported under the eco - net project 12651nj geometries over finite rings and the properties of mutually unbiased bases '' and the cnrs - sav project 20246 projective and related geometries for quantum information " . | the @xmath0 pauli operators attached to a composite qudit in dimension @xmath1 may be mapped to the vectors of the symplectic module @xmath2 ( @xmath3 being the modular ring ) . as a result
, perpendicular vectors correspond to commuting operators , a free cyclic submodule to a maximal commuting set , and disjoint such sets to mutually unbiased bases . for dimensions
@xmath4 , and @xmath5 , the fine structure and the incidence between maximal commuting sets is found to reproduce the projective line over the rings @xmath6 , @xmath7 , @xmath8 , @xmath9 and @xmath10 , respectively . |
usually deterministic and stochastic analysis are not considered compatible . on one hand
we are often limited to statistical analysis when dealing with experimental time series . on the other hand
a purely stochastic approach to data analysis may lead to an erroneous interpretation about the genesis of the analyzed system . in this work
, we investigate , whether it is possible to substitute the lyapunov exponent with a statistical measure , at least in simple 1-d systems . this may be valuable in the case when we do not know what is the equation behind the process we observe .
it may also be useful when it is technically difficult to calculate the lyapunov exponent @xcite . in this paper
, we use the so - called general random walk as our main model @xcite .
we analyze the properties of this model when the noise components of the walk are generated by a stochastic process or a deterministic process ( the tent map ) . using detrended fluctuation analysis ( _ dfa _ ) @xcite we compare the time series obtained from the above implementations of the model .
the conclusion is that both implementations of the model lead to the same results . as a consequence of this observation
, we calculate the lyapunov exponent for the deterministic version of model .
the result is an equation relating the dfa exponents ( a statistical measure ) with the lyapunov exponent for our model , which is practically independent of the method used to generate noise . using the lyapunov exponent we may calculate the diffusion coefficient for a particle
, which movement is defined by the model . by doing this
we also doublecheck that the deterministic implementation of the model has the same statistical properties as the purely stochastic one .
detrended fluctuation analysis ( _ dfa _ ) has been originally applied to the dna walk @xcite . therefore in general one needs to define a walk that is related to the input series analogously @xcite . to do this ,
the input series @xmath0 of length @xmath1 is integrated after subtracting their average value .
the series @xmath2 is then : @xmath3 where @xmath4 is the @xmath5-th point of a discrete time series and @xmath6 is the average value of the data .
next , the integrated series is divided into subintervals of equal length @xmath7 , and for each subinterval a linear least squares fit to the @xmath8 , denoted @xmath9 , is made ( fig . 1 ) .
the rms fluctuation around the regression line is then given by the equation : @xmath10 ^ 2}\,.\ ] ] the dependence of @xmath11 on the length @xmath7 is examined via a plot of log @xmath12 versus log @xmath7 ( fig . 2 ) .
when scaling occurs , the overall slope of the line in the double - logarithmic scale is equal to the dfa exponent and denoted by @xmath13 where @xmath14 is the intercept of the approximated trend . generally , when a scaling @xmath15 is observed , the scaling exponent in the range @xmath16 indicates positive long - range power - law correlations ( in other words , persistence ) and @xmath17 infers anticorrelations ( antipersistence ) @xcite .
@xmath18 is obtained for the brownian walk . the exponent @xmath19 ( fig .
2a ) corresponds to uncorrelated data . if there are short - range correlations , the slope for low @xmath7 may differ from 0.5 but it will approach this value for large @xmath7 . a crossover occurs when for different ranges of @xmath7 we observe different slopes @xcite .
crossovers in the linear dependence of the exponent on window size n have been observed in detrended fluctuation analysis of complex data ( biological data ) and may be an important indicator characterizing the underlying process @xcite .
we define the model by the iterative equation : [ eq3 ] @xmath20 here , @xmath7 is the iteration index ( a natural number ) , @xmath21 is the variable value at the @xmath7-th iteration and @xmath22 is the noise term .
the model equation ( [ eq3a ] ) is equivalent to : @xmath23 as explicitly written in equation ( [ eq3b ] ) , our model is a system with memory , the range of which is defined by parameter @xmath24 .
this is an analogy to the general random walk process @xcite .
we use two processes to generate the noise @xmath22 - a stochastic and a deterministic one . in the _ first _ approach
we use a stochastic process that has a uniform distribution with @xmath25 and @xmath26 . as a _
second _ approach we generate noise using chaotic maps .
the following equation defines the tent map ( also known as the symmetric triangular map ) @xcite : @xmath27 it is well - known @xcite that iterating eq .
( [ tentb ] ) generates data which is equivalent to statistically uncorrelated noise with a uniform i.i.d .
distribution . in other words ,
the natural invariant density @xcite for the tent map ( [ tentb ] ) is equal to @xmath28 . in this paper
we also consider the generation of noise using the ulam map ( the logistic map at fully developed chaos ) @xcite : @xmath29 the main difference between the maps ( [ ulam ] ) and ( [ tentb ] ) is that the natural invariant density for the ulam map does not correspond to the uniform distribution @xcite , instead it is equal to : @xmath30 as we will show further , the differences between the maps ( [ tentb ] ) and ( [ ulam ] ) will not affect the results presented in this paper . to obtain a simple correspondence with the parameter @xmath21 we define @xmath22 as : @xmath31 what is important for our study , this approach does not alter the slope of the chaotic map therefore it preserves the value of the lyapunov exponent ( though it introduces singularities ) .
first , let us focus on the stochastic version of the model given by eq .
( [ eq3 ] ) . as it can be seen in fig .
2 we observe a crossover for series obtain with parameter @xmath32 .
the point of crossover is found to be dependent on the parameter @xmath24 .
the dependence is a power - law and for larger @xmath24 it occurs at larger @xmath7 ( fig .
a similar relation between the correlation range and the crossover point was assumed in @xcite ( without proof or reference to a model ) and applied to the analysis of the heart rate variability series .
we can find this dependence between the crossover point @xmath33 and parameter @xmath24 from numerical data ( fig .
3 ) : @xmath34 therefore , it is possible to obtain the parameter @xmath24 of the model ( [ eq3 ] ) _ only by analyzing the time series ( experimental data ) _ , using the statistical method dfa . for a pure random walk ( @xmath35 ) it is easy to find the diffusion coefficient @xmath36 , which in the general case is defined as @xcite : @xmath37 as @xmath35 , eq .
( [ eq3 ] ) becomes [ purerw ] @xmath38 and , equivalently @xmath39 therefore the diffusion coefficient of the pure random walk ( [ purerw ] ) is equal to : @xmath40 the dependence of the properties of chaotic maps on the control parameter is usually described by the lyapunov exponent , which is a measure of the memory of the initial conditions .
the lyapunov exponent for one - dimensional iterated maps is calculated as @xcite : @xmath41 where @xmath42 corresponds to the @xmath5-th iteration of the map @xmath43 ( in our case defined by eq . ( [ eq3 ] ) ) and
@xmath44 is the initial condition .
a negative value of @xmath45 indicates periodic states , a positive @xmath45 - chaotic states . in general the lyapunov exponent
is dependent of the initial condition @xmath44 , but in the case of the map defined by eq .
( [ eq3 ] ) @xmath46 . instead of calculating the lyapunov exponent using the long - time limit @xmath47 as in eq .
( [ lap ] ) , this exponent may be calculated as the average of all the possible values of the one - step finite - time lyapunov exponent @xmath48 @xcite : @xmath49 first , we will consider the case of the pure random walk , when @xmath35 .
the lyapunov exponent ( eq . ( [ lap ] ) ) for the above general walk can be easily found if we note , that the inversion of the tent map ( [ tentb ] ) gives two symmetric preimages @xmath22 for each @xmath50 @xcite . to calculate the exponent @xmath45 in our model we need to find the product of @xmath51 preimages . however , eq . ( [ lap1 ] ) is the average of one - step lyapunov exponents @xmath48 . due to the symmetry of the preimages of ( [ tentb ] ) , there are two equally probable ( in the statistical sense , as @xmath7 tends to infinity ) values of @xmath48 , namely @xmath52 and @xmath53 .
therefore the lyapunov exponent for the pure random walk is equal to : @xmath54 we obtain the same calculation in the case of the ulam map ( [ ulam ] ) .
now we will derive the relation between the lyapunov exponent and the memory parameter m for the general random walk . at first , this seems a difficult task , because this time we have to find the values of the derivatives of all the preimages of @xmath55 up to @xmath51 possible values of @xmath56 .
this step is a little similar to the construction of the julia set , in the way that one needs to find calculate all the preimages values ( compare @xcite , p.104 ) .
yet it is well - known that the julia set is quite complex . on the contrary , in this case , after simple calculations , a finite sum of a geometrical progression is obtained , and eq .
( [ lap1 ] ) yields : @xmath57 \label{mlap1 } \\\end{aligned}\ ] ] in the above derivation we use the fact , that in the long - time limit ( when averaging over infinitely many iterations ) all values of the one - step lyapunov exponents are equally probable . as we stated earlier ,
this is due to the symmetry of the two preimages in both the logistic map and the tent map .
concluding , we see that the lyapunov equation depends only on the parameter @xmath24 and can be calculated from the equation : @xmath58 .
\label{mlambda}\ ] ] note , that in the limit @xmath35 we obtain eq .
( [ lappurerw ] ) which was calculated earlier for more clarity .
the important observation now is that using the dfa method we have obtained the relation between the the crossover point @xmath33 and @xmath24 ( eq . ( [ logm ] ) , see fig .
this relation is valid for both the statistical model of eq .
( [ eq3 ] ) and the deterministic models of eqs .
( [ tentb ] ) and ( [ ulam ] ) .
this means that we can find the lyapunov exponent for our model using only statistical analysis by comparing eqs .
( [ mlambda ] ) and ( [ logm ] ) : @xmath59 .\ ] ] we obtained a quite complicated relation , therefore , to doublecheck its correctness we show that the lyapunov exponent approaches the limit : @xmath60 already for @xmath61 ( fig . 4 ) .
this is in agreement with the previous finding in eq .
( [ lappurerw ] ) obtained for pure random walk case .
the diffusion coefficient may be calculated using the second term of the cumulant expansion of the one - step lyapunov exponent @xcite : @xmath62 ^ 2 \right > \label{diffdet}\ ] ] equation ( [ diffdet ] ) for our model yields : @xmath63,\ ] ] where @xmath64{(9 - 2^{2 - 2m})(1 - 2^{2 - 2m})}$ ] .
the above equation in the case of the pure random walk ( @xmath35 ) yields : @xmath65 comparing the above result to eq .
( [ diffpurerw ] ) obtained for the stochastic model : @xmath66 thus , we obtained a value which is close to the stochastic diffusion coefficient ( [ diffpurerw ] ) . by comparing the basic statistical properties of stochastically and deterministically generated series we concluded earlier that the series are statistically equivalent .
( [ differ ] ) may indicate that although the diffusion coefficient is only a statistical measure ( as in eq .
( [ diff ] ) ) , it may be slightly sensitive to the dynamical genesis of the system . still , it would be very difficult to use such a method even numerically , as usually the statistical errors are greater than the difference found in eq .
( [ differ ] ) .
we studied the properties of the model defined by eq .
( [ eq3 ] ) when the noise generation term is either a stochastic process or a deterministic process .
we compared the time series obtained from the above implementations of the model by use of dfa .
the conclusion is that the two versions of the model are indistinguishable using such statistical methods .
the advantage of this observation is that we may calculate the lyapunov exponent for the model defined by eq .
( [ eq3 ] ) .
we obtain an equation relating the dfa exponents ( a statistical measure ) with the dynamical lyapunov exponent for such models . on the other hand , typical statistical properties
can be calculated in two ways : using statistical analysis or using equations of deterministic non - equilibrium mechanics . as an example , we calculated the diffusion coefficient ( [ dlambdarrow ] ) and obtained a value which is very close to the stochastic diffusion coefficient ( [ diffpurerw ] ) .
the dfa software used in the calculations , is generously available from the authors of the method at the physionet database world wide web site ( _ http://www.physionet.org _
@xcite ) . 10 christian beck and friedrich schlgl . .
cambridge university press , 1993 .
rdiger zillmer . .
phd thesis , universitt potsdam , potsdam , 2004 .
rainer klages . .
phd thesis , wissenschaft und technik verlag , berlin , 1996 .
nickolay korabel and rainer klages .
, 187:66 , 2004 .
jws , new york , 2nd edition , 1998 .
peng et al . , 49(2):1685 , 1994 .
peng et al . , 5(1):82 , 1995 .
p. talkner and r.o .
, 62:150 , 2000 .
edward ott . .
cambridge university press , 1993 .
schuster . .
vch , weinheim , 1988 . j. robert dorfman . .
cambridge university press , cambridge , 1999 .
e. aurell et al .
, 30(1):1 , 1997 .
a. l. goldberger et al .
( e - publication ) .
, 101(23):215 , 2000 .
1 . in detrended fluctuation analysis
the integrated series @xmath8 is linearly approximated in each window of size @xmath7 . here
, two cases of approximation are shown on a fragment of a test series : for @xmath67 ( continuous lines ) and @xmath68 ( dashed lines ) .
2 . dfa plots for time series from the model defined by eq . ( [ eq3 ] ) at different parameter @xmath24 values .
the length of the time series were @xmath69 . *
a ) * : @xmath70 , this is equivalent to uncorrelated noise series * b ) * : @xmath71 .
* c ) * : @xmath72 , a crossover occurs - we witness two slopes . *
d ) * : @xmath73 , the crossover is shifted in respect to @xmath72 . for window sizes
@xmath7 smaller than the crossover point @xmath74 the slope value is near to 1.5 ( corresponding to brownian noise ) , for window sizes larger than @xmath74 it approaches 0.5 ( white noise ) .
the equation ( [ logm ] ) obtained from numerical simulations .
4 . the dependence of the lyapunov exponent @xmath45 and its variance on the parameter @xmath24 . | a simple model of an irreversible process is introduced .
the equation of iterations in the model includes a noise generation term .
we study the properties of the system when the noise generation term is a stochastic process ( a random number generator ) or a deterministic process ( a chaotic map ) .
we compare the time series obtained from the above implementations of the model by use of statistical methods ( such as detrended fluctuation analysis ) .
the conclusion is that using statistical methods the two versions of the model are indistinguishable .
the advantage of this observation is that we may calculate the lyapunov exponent for the model . as a result
we obtain an equation relating the dfa exponents ( a statistical measure ) with the lyapunov exponent for such models . on the other hand , typical statistical properties
can also be calculated , as for example the diffusion coefficient for a particle , which movement is defined by the above model .
= by -1 |
as the most luminous persistent sources in the universe , active galactic nuclei ( agns ) , which are powered by the mass accretion of super - massive black holes ( smbhs ) , are widely believed to play important roles in the formation and evolution of typical massive galaxies ( for a recent review , see * ? ? ?
* ) . indeed , over the past sixteen years
, many studies have revealed that there are tight correlations between smbhs and the physical properties of host galaxies ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
for example , the smbh mass , @xmath12 , correlates well with the stellar mass of the bulge , @xmath13 .
the intrinsic scatter of the @xmath14 relation is found to be @xmath15 .
such a tight correlation might be explained by agn feedback ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . in this scenario , winds or jets launched from the accretion disk of the smbh heat the interstellar medium ( ism ) or clear the ism out of the host galaxy .
star formation is therefore quenched due to the lack of cold gas .
such a coupled agn triggering and feedback - regulated star formation could be responsible for the @xmath14 relation ( see , e.g. , * ? ? ?
* ; * ? ? ?
on the other hand , a non - causal origin of the @xmath14 relation has also been proposed by @xcite ( see also * ? ? ?
* ) . in this scenario , @xmath12 and @xmath13 are not initially correlated .
the @xmath14 relation is established simply by galaxy - galaxy mergers via the central limit theorem .
direct observational measurement of the evolution of the @xmath14 relation is crucial to investigate the formation and co - growth of smbhs and their hosts .
observationally , the evolution of the @xmath14 relation is still a controversial topic .
moreover , at high redshift ( e.g. , @xmath16 ) , it is very difficult to separate the bulge from the total stellar mass .
instead , most works explore the @xmath17 relation , where @xmath18 is the galaxy total stellar mass . for instance , @xcite find that at @xmath19 the @xmath17 relation is consistent with the local @xmath14 relation .
note that the host galaxies of the local sample of @xcite ( hereafter hr04 ) are bulge dominated ( i.e. , @xmath20 ) .
therefore , the @xcite result suggests little evolution of the @xmath17 relation to @xmath19 .
on the other hand , the hosts of the @xcite sample have substantial disks ( i.e. , @xmath21 ) .
that is , the @xmath14 relation does evolve as a function of redshift . making the local @xmath14 relation then , as pointed out by @xcite , requires a redistribution of stellar mass from disks into bulges in agn host - galaxies .
recently , the work of @xcite with a larger and less - biased sample also supported the results of @xcite . on the other hand , @xcite compiled a broad emission - line agn ( blagn ) sample with well - measured multi - band photometry from the cosmos survey . with this sample ,
they find that even the @xmath17 relation evolves ( at @xmath22 significance level ) over cosmic time ( i.e. , the smbh is over - massive at high redshifts ; see also * ? ? ?
however , as pointed out by @xcite , high - redshift samples are generally selected by agn activity ( which depends on @xmath12 ) and suffer from eddington bias .
the physical reasons are : ( 1 ) there is intrinsic scatter in the @xmath14 or @xmath17 relation ; ( 2 ) the galaxy stellar mass function is bottom heavy ( i.e. , the number density of galaxies steeply decreases to higher stellar mass ) .
hence , for these high - redshift samples , an agn with given @xmath12 is more likely to be found in less massive galaxies , resulting in an apparent evolution of the @xmath5 or @xmath17 relation ( also see section [ sec : bias ] ) .
this bias increases with redshift as at high redshifts we often only have access to very luminous agns ( very massive smbhs ) .
additionally , @xmath12 estimation itself is also biased for these high - redshift samples @xcite .
the reasons are similar : ( 1 ) @xmath12 is estimated from agn luminosity which also has intrinsic scatter ; ( 2 ) the agn luminosity function is also bottom heavy .
note that the sample of @xcite consists of mostly lower luminosity agns and hence suffers less bias than @xcite .
bias corrections must be made to compare the high - redshift @xmath17 relation with the local one ( for a more recent discussion on this topic , see , e.g. , * ? ? ?
in addition to measuring the @xmath17 relation at high redshift , it is also valuable to study its instantaneous evolution ; i.e. , the smbh and galaxy stellar mass growth rates . especially , will future smbh and galaxy mass increases help establish / maintain the @xmath17 relation ?
this question has been indirectly investigated .
for example , many works explored the possible connection between star formation rates ( sfrs ) and agn luminosities ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * and also see alexander & hickox 2012 for a review ) .
however , as pointed out by @xcite , agn variability can dilute the intrinsic correlation between star formation and agn activity .
therefore , this connection is expected to be observed only in an average sense , as individual galaxies will tend to show a much weaker ( or absent ) correlation .
the average smbh accretion rate traces the average sfr so well that the local @xmath17 relation can be established . simultaneously exploring both
the @xmath17 relation at high redshifts and its instantaneous evolution is even more valuable but is challenging .
@xcite explore this issue by measuring @xmath12 , @xmath18 , and their growth rates .
they find that star formation and agn activity can actually help reduce the scatter of the @xmath17 relation .
however , the sfr in their work is estimated by fitting a stellar population synthesis model to the optical - to - nir spectral energy distribution ( sed ) which has an uncertainty of @xmath23 . a more accurate determination of sfr ( e.g. , with _ herschel _ observations ) could better elucidate the coupled growth of agns and their host galaxies .
in this work , we uniquely explore the @xmath0 relation and its evolution , accounting for biases and using robust and uncontaminated estimates of smbh growth and galaxy star formation rate . in section [ sec : data ] , we describe our sample in detail . in section [ sec : property ] , we describe our methods for estimating @xmath12 , @xmath18 , @xmath24 , and @xmath25 . in section
[ sec : result ] we discuss selection biases and present our results . in section
[ sec : dis ] , we discuss implications of our results . finally , we summarize our main results and outline the possible future improvements in section [ sec : summ ] . throughout this work
we adopt a flat @xmath26cdm cosmology with @xmath27 , @xmath28 and @xmath29 .
the @xmath30-@xmath31 cosmic evolution survey ( cosmos , * ? ? ? * ) and @xmath32-@xmath33 chandra deep field - south survey ( cdf - s ; e.g. , * ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) both have well - calibrated multi - band photometry ( from the x - ray to radio bands ) and high - quality spectroscopic coverage .
hence , we selected our sample from the cosmos and cdf - s fields .
first , we selected blagns based on x - ray detection of agn activity ( for the advantages of using x - ray selection , see * ? ? ? * ) and subsequent optical spectroscopic identification of broad emission lines ( with coverage of either @xmath34 or ) .
second , we cross - matched our blagns with the _ herschel _ pep @xcite ( for the cdf - s field , we also used the goods_herschel _ catalogs , see * ? ? ? * ; * ? ? ? * ) and hermes @xcite catalogs .
third , the new sample was cross - matched with existing multi - band photometric catalogs to build the final sample . in the following subsections ,
we discuss our sample selection in detail ( see also table 1 ) .
the @xmath30-@xmath31 cosmos field has been surveyed in the x - ray by _ xmm - newton _ @xcite and partially ( 0.9 @xmath31 ) by _ chandra _ @xcite .
we first selected all @xmath35 blagns ( ensuring the coverage of either @xmath34 or ) from the public cosmos spectroscopy @xcite .
our blagns have public magellan / imacs @xcite , sdss @xcite , or vlt / vimos ( zcosmos ; * ? ? ?
* ) spectroscopic data which allow us to estimate @xmath12 via the single - epoch approach ( see also section [ sec : mbh ] ) .
the magellan / imacs spectroscopy is limited by @xmath36 @xcite . for consistency and ease of modeling the observational biases
, we impose the same @xmath36 limit for the combined cosmos and cdf - s sample .
we then cross matched them with x - ray catalogs to get the corresponding ( rest - frame ) @xmath37 luminosity .
for x - ray observations , we prefer _ chandra _ over _ xmm - newton _ and hard band over soft band ( i.e. , the _ chandra _ hard band has the highest priority , and the _ xmm - newton _ soft band has the lowest priority ) . as pointed out by @xcite and will be shown in figure [ fig : bolz ] , our agn selection is limited by the @xmath36 criteria and not the x - ray sensitivity ( whether _ xmm - newton _ or _ chandra _ ) . to get a robust estimate of sfr ,
we cross - matched our blagn sample with the _ herschel _ pep ( @xmath38 and @xmath39 ) and hermes ( @xmath40 ) catalogs .
note that for the hermes catalog the spire/_herschel _ observations have a large point - spread function .
we followed @xcite and chose a maximum matching radius of @xmath41 .
herschel _ pep catalogs have improved astrometry because sources were extracted using _
spitzer_/mips @xmath42 priors , and so we used a maximum matching radius of @xmath43 . for pep and @xmath44 for hermes .
( all but five objects have both pep and hermes data . )
the false - match rate is calculated by following the procedure described in section 3 of @xcite . ] as for multi - band photometry , we used the cosmos photometric redshift catalog fall 2008 @xcite to get the optical - to - nir broad - band photometry information ( @xmath45 , @xmath46 , @xmath47 , @xmath48 , @xmath49 of subaru ; @xmath50 , @xmath51 , @xmath52 of cfht ; and @xmath53 of ukirt ) for the selected sources ( see also * ? ? ?
* ; * ? ? ?
we also included four _ spitzer_/irac channels @xcite and _ galex_/near uv data when available .
zero - point corrections suggested in @xcite were applied .
the galactic extinction was also corrected using the galactic extinction map of @xcite .
note that for the optical - to - nir bands , the galactic extinction corrections are already included in the catalog .
the @xmath32-@xmath33 cdf - s field is covered by the deepest x - ray survey , with @xmath54 ms of _ chandra _ coverage @xcite .
we selected blagns by cross - matching the 4 ms cdf - s catalog with the optical spectroscopy of @xcite .
similar to the cosmos sources , we only selected blagns ( with @xmath36 ) whose host galaxies are also detected by the _
herschel _ pep or hermes ( the cross - matching criteria is the same as that of cosmos blagns ) . as for multi - band photometry
, we used the musyc 32-band ( including @xmath55 , @xmath56 , @xmath45 , @xmath46 , @xmath57 , @xmath58 , @xmath49 , @xmath53 , @xmath59 , @xmath60 bands ; 18 subaru medium bands ; and four _ spizter_/irac channels ) catalog ( * ? ? ? * and references therein ) .
the galactic extinction and zero - point corrections suggested in @xcite were adopted .
based on the above criteria , we started with a total number of 251 x - ray blagns at @xmath35 ( ensuring coverage of either @xmath34 or ) , with 221 from the cosmos field and 30 from the cdf - s field . when cross - matched with the _ herschel _ catalogs , 96 sources were left ( 85 from the cosmos field ; 11 from the cdf - s field ) .
69 of these 96 sources have high - quality spectroscopic observations and robust fir detections which enable us to measure @xmath12 , @xmath18 , @xmath24 , and sfr , reliably . for the other 27 sources , either the spectroscopic data ( nine objects ) or the fir observations ( 18 objects )
are not sufficient for a reliable measurement of @xmath12 or sfr ( see sections [ sec : mbh ] & [ sec : sfr ] ) .
note that due to the _ herschel _ sensitivity , @xmath61 of blagns are undetected .
we will consider the bias introduced by the _ herschel _
sensitivity in section [ sec : bias ] .
for the 69 sources in our sample , 62 are from the cosmos field and seven are from the cdf - s field .
note that their spectra are flux calibrated .
we compared the physical properties ( section [ sec : property ] ) of these seven cdf - s sources with the cosmos sources ( using the mann - whitney u test ) and found no statistical difference ( i.e. , the null probability @xmath62 ) .
we adopted single - epoch virial smbh mass estimators to estimate smbh masses . more specifically , h@xmath63 or line estimators were used depending on the redshift .
these smbh mass estimators use the correlation between the radius of the broad line region ( blr ) , @xmath64 , and the continuum luminosity , @xmath65 , @xmath66 , revealed by reverberation mapping observations of local agns @xcite .
if the blr is virialized , then the smbh mass can be estimated by @xmath67 , where @xmath68 is the full width at half maximum of the broad emission line and @xmath69 is related to the blr geometry ( calibrated from dynamical estimates of @xmath12 , see , e.g. , * ? ? ?
therefore , we can estimate the smbh mass from @xmath65 , and @xmath68 , @xmath70 where @xmath65 and @xmath68 are in units of @xmath71 and @xmath72 , respectively .
this relation has an intrinsic scatter of @xmath73 dex @xcite , which is due to the unknown blr geometry and other factors ( e.g. , ionization state ) .
note that for h@xmath63 , @xmath74 , @xmath75 , @xmath76 @xcite ; for , @xmath77 , @xmath78 , @xmath79 @xcite . both @xmath65 and @xmath68 can be measured by fitting the single epoch spectra .
we performed an iterative chi - squared minimization to fit the broad emission lines ( h@xmath63 or ) for each source and measured @xmath68 and @xmath65 .
the fitting procedures were similar to those of @xcite .
that is , a pseudo - continuum ( power - law continuum plus broadened fe template of * ? ? ?
* ) and one or two broad gaussian line profiles were ( simultaneously ) used to get a best fit of each spectrum . for h@xmath63 , we also added [ oiii]@xmath80 lines and a narrow h@xmath63 component to the fit and then subtracted them when fitting the broad emission lines since these narrow components come from the narrow - line region .
our criterion for the use of one or two broad gaussian line profiles was as follows : first , we used two gaussian components to fit the broad emission line and then we compared @xmath68 of both gaussian components . if at least one component had @xmath81 , we refitted the spectrum with only one gaussian component . note that if we use two gaussian line profiles for all sources , our @xmath12 estimates change by no more than @xmath82 dex , and this is small compared to the intrinsic uncertainty of @xmath83 . in figure
[ fig : fbh ] , we present examples of our fits to the broad emission lines .
the gray and red solid lines represent the observed spectra and our best fit , respectively .
red and orange dot - dashed lines represent the power - law continuum component and fe template , respectively .
the dotted lines correspond to minor features ( i.e. , [ oiii]@xmath80 lines , the narrow h@xmath63 line , which are excluded from the broad - line fit ) .
blue dashed lines represent best fits of broad emission lines .
we calculated @xmath68 as follows : if two broad gaussian profiles were used , we calculated @xmath68 from the emission profile which is the summation of the two broad gaussian components ; otherwise , we calculated @xmath68 from only the one broad gaussian profile .
we have visually inspected all fitting results and found nine sources whose spectroscopic line profiles do not allow reliable measurement of @xmath83 .
lines in their spectra ; i.e. , they are misclassified as blagns , and are instead type 2 or intermediate type ( type 1.x ) agns .
five of these nine sources show strong absorption features superposed on the broad emission lines .
the removal of these sources should not introduce a selection bias to our sample , since they are essentially random occurrences which are unlikely to be strongly correlated with eddington ratio or host properties .
] we excluded these nine sources from our sample .
meanwhile , @xmath65 was calculated from the power - law continuum component .
note that we also applied a small correction to account for the contamination of the stellar light of the host galaxy suggested in section [ sec : mgl ] .
a summary of smbh masses can be found in table 2 ( before compiling the table , we removed another 18 sources because of inadequate fir data , see section [ sec : sfr ] ) .
there are 24 ( 46 ) sources in table 2 whose @xmath83 were measured using h@xmath63 ( ) .
previous studies ( e.g. , * ? ? ?
* ) have revealed that the two @xmath83 estimators are consistent .
four sources in table 2 have spectroscopic data enabling us to measure @xmath83 using both h@xmath63 and . for three blagns ,
the two @xmath83 estimators are consistent with each other within @xmath82 dex . for the remaining blagn ( id-48 ) ,
the two @xmath83 estimators show a larger deviation ( @xmath84 dex ) but are still consistent within the @xmath85 uncertainty . in any case
, we preferred the h@xmath63 estimator over estimator .
we utilized an sed fitting technique to estimate @xmath18 .
since we were dealing with blagns , the seds ( from uv to near ir ) of the host galaxies are contaminated by agn emission .
we therefore followed the approach of @xcite ( see also * ? ? ? * ) which fits the observed seds as a composite of both agn and galaxy stellar emission .
that is , @xmath86 where @xmath87 and @xmath88 are the fluxes from the agns and host galaxies , respectively . for seds ,
fourteen bands ( observed - frame wavelength ranging from @xmath89 to @xmath90 ) are used for the cosmos sources and 32 bands ( observed - frame wavelength ranging from @xmath91 to @xmath90 ) are used for the cdf - s sources . for agn emission , we adopted the mean sed template of @xcite . for galaxy templates , we selected from an sed library constructed using the @xcite stellar population synthesis model .
we assumed an initial mass function ( imf ) of @xcite and adopted ten exponentially declining star formation histories ( sfhs ) , i.e. , sfr @xmath92 , with star formation timescale @xmath93 ranges from @xmath94 to @xmath95 , plus a sfh with constant sfr .
the galaxy age ranges from @xmath96 to @xmath97 with an additional constraint that the galaxy age should not be larger than the age of the universe at the redshift of the source .
the sed library was generated iterating over @xmath93 and the galaxy age .
when fitting the observed sed , we also took intrinsic extinction into consideration . for agns , we used a small magellanic cloud - like dust - reddening curve @xcite with @xmath98 .
for galaxies , we adopted the calzetti extinction curve @xcite with @xmath99 if @xmath100 , otherwise @xmath101 @xcite .
we additionally required that the agn continuum emission should be larger than @xmath102 of the galaxy continuum emission at @xmath103 ( if the virial mass estimator was used ) or @xmath104 ( if the h@xmath63 virial mass estimator was used ) .
this was simply motivated by the fact that our sources all have observed broad emission lines ( h@xmath63 or ) indicative of a significant agn contribution .
we chose @xmath102 based on the following considerations : ( 1 ) we performed a simple simulation by creating a mock spectrum with a young galaxy sed , an agn ( power - law ) component , a broad gaussian emission line with a typical equivalent width of @xmath105 ( @xmath106 ) for ( h@xmath63 ) ( e.g. , section 1.3.4 of * ? ? ?
* ) , and white noise ( assuming a signal - to - noise ratio of @xmath107 ) ; ( 2 ) we refitted the mock spectrum with a power - law and a gaussian emission line and found that we can not reliably fit the emission line if the agn emission is less than @xmath6 of galaxy emission .
furthermore , we have performed a monte carlo simulation to mimic our selection procedures ( see section [ sec : bias ] for more details ) . for our simulated sample
, we also found that there are a negligible number of sources with agn emission less than @xmath102 of the galaxy emission at @xmath103 ( or @xmath104 ) .
we fitted @xmath108 to the observed seds by performing iterative ( reduced ) chi - squared minimization .
figure [ fig : fgm ] shows examples of our two - component ( agn plus galaxy ) fitting to the observed seds .
the data are well described by a combination of agn and galaxy emission .
we also inspected the corresponding hst / acs images and found that generally our sed fits are consistent with the morphology information .
that is , when the sed fit suggests the agn emission dominates in the @xmath51 band , the hst / acs image also shows a point - like morphology ( the inverse is also true ) . from each best fit
, we determined both the normalization , @xmath109 , @xmath93 , and the sfh of our galaxy sed .
we then used them and the @xcite model to calculate the galaxy total stellar mass .
a summary of the galaxy total stellar masses can be found in table 2 .
note that the two - component sed fit can also tell us the agn contribution to either the @xmath110 or @xmath111 luminosity , and we can use this information to calculate the true agn contribution to @xmath65 measured in section [ sec : mbh ] .
this small correction ( accounting for the contamination of the stellar light of the host galaxy ) has been included when calculating @xmath65 from the continuum component in section [ sec : mbh ] .
to measure the accretion rate , @xmath24 , we estimated the bolometric luminosity , @xmath112 ( e.g. , * ? ? ? * ) .
the accretion rate is , @xmath113 where @xmath114 is the assumed radiative efficiency of the accretion disk , and @xmath115 is the speed of light .
we used two methods to calculate @xmath112 : ( 1 ) the rest - frame @xmath37 luminosity alone ; ( 2 ) the sed fit plus the rest - frame @xmath37 luminosity .
we first used the rest - frame @xmath37 luminosity as an estimator of @xmath112 .
the rest - frame @xmath37 luminosity is calculated from the observed - frame @xmath37 flux ( assuming a typical power - law photon index of @xmath116 ) .
note that there are four sources in our sample that are undetected in the observed - frame @xmath37 band . for them , we instead calculated the rest - frame @xmath117 luminosity from the observed - frame @xmath117 flux ( again with @xmath116 ) and used ( different ) bolometric corrections to estimate @xmath112 .
we adopted the @xcite luminosity - dependent bolometric corrections : @xmath118 where @xmath119 is the bolometric correction factor .
the constants @xmath120 are given by @xmath121 for @xmath122 , and @xmath123 for @xmath124 .
the uncertainty of @xmath119 is @xcite , @xmath125 with ( @xmath126 , @xmath63 , @xmath127 ) @xmath128 in the @xmath117 band and @xmath129 in the @xmath130 band . for sources in our sample ,
the median of the uncertainties is @xmath131 dex .
we also combined the sed fitting results presented in section [ sec : mgl ] with x - ray observations to calculate the bolometric luminosity .
this procedure was similar ( but not identical ) to that of @xcite .
first , we used the rest - frame @xmath37 luminosity to get the @xmath132 luminosity ( assuming @xmath116 and a cut - off at @xmath133 ) ; second , we integrated the big blue bump component ( from @xmath105 to @xmath134 ) in the best - fitted agn sed which should be responsible for the emission from the accretion disk ( note that we neglected the ir `` torus '' emission as it is probably reprocessed ) ; finally , we obtained the bolometric luminosity by summing the @xmath132 luminosity and the total big blue bump luminosity .
this approach gives a direct estimate of @xmath112 .
we compared the two bolometric luminosity estimates and found that the ratios of these two estimates follow a log - normal distribution .
this log - normal distribution has a mean value of @xmath135 and a standard deviation of @xmath136 dex .
this deviation should be caused by the uncertainties of the two bolometric luminosity estimators .
therefore , we conclude that the uncertainty of the bolometric luminosity estimated using the sed fit plus the x - ray luminosity is @xmath137 dex . in the following calculations ,
we use the bolometric luminosity calculated from sed fits plus x - ray observations to measure @xmath24 .
a summary of the bolometric luminosities can be found in table 2 .
figure [ fig : bolz ] plots @xmath138 as a function of redshift .
in addition , we also include the bolometric luminosity limits introduced by the @xmath36 spectroscopy limit and the x - ray sensitivity limits of cosmos ( soft bands of _ chandra _ and _ xmm - newton _ ) and cdf - s ( hard band of _ chandra _ ) .
the x - ray limits are significantly deeper than the optical spectroscopy limit , and the sample is essentially limited only by @xmath36 ( see also * ? ? ?
the four sources which are only detected in the observed - frame @xmath117 band are denoted as red points .
the redshifts of our blagns span @xmath1 with a median redshift of @xmath139 .
we also define a `` high - redshift '' sub - sample of the 45 sources at @xmath11 , and a `` low - redshift '' sub - sample of the 25 sources at @xmath140 .
figure [ fig : bolmbh ] plots the bolometric luminosity as a function of @xmath12 .
we also include the lines of @xmath141 , @xmath142 and @xmath143 , where @xmath144 .
as clearly seen from this figure , the eddington ratios of our blagns are typically spread from @xmath145 ( consistent with , e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
for cosmos sources , we also made a detailed comparison between eddington ratios of our sources and those of @xcite by checking the overlapping sources one by one .
we found that the median absolute deviation ( m.a.d . ) of the differences is @xmath82 dex , much smaller than the uncertainty we assigned to @xmath12 ( see section [ sec : err ] ) .
for the cdf - s field , @xcite found a broad distribution of eddington ratios , using host galaxy mass ( or velocity dispersion ) to estimate smbh mass .
of their sources with broad emission lines , 6/11 have eddington ratios between @xmath146 and 1 , and two others have eddington ratios consistent with @xmath146 .
only 3/11 have eddington ratios significantly smaller than @xmath146 . despite the very different methods of @xmath12 estimation , our results are broadly consistent with their results .
far - ir emission has long been argued to be an excellent sfr tracer as it is largely free of dust extinction .
we used the widely adopted @xcite relation ( with a modification to account for the imf of * ? ? ? * ) to derive the sfr : @xmath147 where @xmath148 is the total @xmath149 ir luminosity .
blagns make significant contributions in the near- to mid - ir bands .
the fir band is , however , known to be dominated by galaxy emission ( e.g. , * ? ? ?
* hereafter k12 ) . in this work , we used fir data from the _ herschel _ pacs / spire bands to normalize the @xmath16 star - forming galaxy sed template of k12 and integrated the normalized sed to estimate @xmath148 .
there are other popular galaxy ir sed templates ( e.g. , * ? ? ?
* ; * ? ? ?
however , the k12 template is directly derived from the high - redshift sources and therefore may be more relevant to our high - redshift sources . as stated before ,
we have additionally removed 18 sources which have @xmath11 but are detected only at the @xmath38 band .
the reason is that , under such circumstances , the @xmath38 band might be significantly contaminated by agn emission ( e.g. , k12 , * ? ? ? * ) .
such measurements of @xmath148 are therefore not reliable . for our 69 sources ,
26 are detected at @xmath38 , @xmath39 , and @xmath40 .
24 are detected in two bands .
the remaining 19 sources are detected only in one band .
a summary of @xmath148 can also be found in table 2 . to check the consistency of our @xmath148 estimation
, we took our 26 sources that are detected in three bands and compared @xmath148 values that are estimated from all three bands with those values from two bands or one band .
figure [ fig : fira ] shows the result . as seen from the figure , @xmath148 values estimated from three bands are consistent with those from only two bands or one band .
therefore , we can conclude that our method for estimating @xmath148 is self - consistent .
note that , even for @xmath11 sources , @xmath148 estimated from @xmath38 agrees well with that from three bands , which indicates that for these three - band detected sources agn emission does not strongly contaminate even the blue _ herschel _ fir band . for
the removed 18 sources , since we only have information on the bluest band ( i.e. , @xmath38 ) which corresponds to rest - frame @xmath150 ( for @xmath151 ) , we can not rule out the possibility that agn contamination may still be important . to quantify the star formation activity of our sample
, we compared the @xmath152 relation of our galaxies with the star - forming main sequence " ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the @xmath152 relation ( i.e. , the star - forming `` main sequence '' ) we adopted is from @xcite , @xmath153 with @xmath154 , @xmath155 .
this relation has a scatter of @xmath156 dex , roughly independent of @xmath49 and @xmath18 .
figure [ fig : sfrc ] plots the ratio of the sfr we measured to the one predicted by equation [ eq : ms ] . for the uncertainty of the ratio , we combined the uncertainties of @xmath18 , sfr , and the scatter of equation [ eq : ms ] .
our sources show higher star formation activity than that of the main sequence ( although they are not significantly offset by more than their uncertainties ; also the mean ratio is @xmath157 , unlike starburst galaxies whose ratios have @xmath158 offset from equation [ eq : ms ] ) .
this is driven mostly by the _
herschel _ flux limit , which lies around or above the main sequence ( see , e.g. , figure 4 of * ? ? ?
* ) and causes an eddington bias ( see section [ sec : bias ] ) .
the factor of two excess in the mean ratio is fully consistent with the expected eddington bias if the host galaxies of our blagns were drawn from the star - forming main sequence . in our sample , @xmath83 , @xmath18 , @xmath24 , and
sfr all have significant uncertainties .
let us first consider the uncertainty of @xmath83 estimation . for a fraction of our sources , @xmath12 has been carefully estimated by some previous studies ( e.g. , * ? ? ?
* ; * ? ? ?
therefore , we compared our @xmath12 estimations with these previous works .
we found that there is no global systematic offset and the standard deviation is no more than @xmath159 dex ( actually , when comparing with * ? ? ?
* we found that for all sources but one , the deviations are smaller than @xmath82 dex ; for the remaining one , the deviation is @xmath160 dex ) .
therefore , the error is dominated by the intrinsic error of the smbh mass single - epoch virial relation .
we then adopted @xmath136 dex , the intrinsic error , as the error of @xmath83 @xcite . for @xmath24
, we only considered the uncertainty from @xmath138 , which is @xmath161 dex ( see section [ sec : mgl ] ) .. note that our anti - correlations in section [ sec : dis2 ] do not significantly change if we instead assume the scatter of @xmath162 is @xmath163 dex ( e.g. , * ? ? ?
* ) ] for @xmath18 , we compared our results with those of @xcite ( cosmos sources ) and @xcite ( cdf - s sources ) .
we found the median and m.a.d . of the differences
is @xmath164 dex and @xmath82 dex , respectively .
note that there are a handful of sources in our sample showing relatively large deviations of @xmath18 with respect to those of @xcite ( although there is no global systematic offset ) .
this is likely caused by several factors , e.g. , different photometric data ( our data included the nuv data from @xmath165 while * ? ? ?
* on the other hand , took the @xmath166 data into consideration ) ; our additional requirement of agn contribution .
note that @xcite and our work both used the imf of @xcite , and the galaxy stellar masses from @xcite were also converted to those with the imf of @xcite .
the m.a.d .
we obtained should indicate this systematic / methodology uncertainty when estimating @xmath18 from seds .
this uncertainty is much larger than the uncertainty introduced by the photometry error .
therefore , we adopted the normalized m.a.d .
( nmad ) , @xmath167 dex , as the uncertainty of the galaxy total stellar mass @xcite .
sfr estimation is subject to two major uncertainties : ( 1 ) the errors of the fir fluxes ; ( 2 ) the intrinsic uncertainty of using the k12 template to estimate sfr is , as reported by k12 , @xmath168 dex .
we added these two uncertainties in quadrature to account for the full uncertainties of sfrs .
we now discuss biases due to selection effects .
first , our sources are selected based on agn luminosity ( as well as the luminosity contrast of the agn and host galaxy in the optical / uv ) and broad emission lines ( @xmath169 or ) . selecting galaxies around luminosity - limited agns
leads to an eddington bias in the @xmath0 relation , as detailed by @xcite ( also see section [ sec : intro ] ) . on the top of the @xcite bias
, there is another bias on @xmath12 for our luminosity - limited agn sample @xcite .
second , we required our sources be detected by _
the _ herschel _ detection limit effectively introduces a bias which acts in the opposite direction as the @xcite bias , since a sfr limit generally selects more massive host galaxies for an agn with given @xmath12 . to quantify selection biases in our sample , we begin by assuming that the scatter in each quantity has a log - normal distribution ( as commonly assumed in , e.g. , *
* ; * ? ? ?
according to the local smbh mass - galaxy stellar mass relation @xmath170 , the distribution of the smbh mass , @xmath171 , at fixed galaxy stellar mass , @xmath172 , is @xmath173 where @xmath174 is the intrinsic scatter .
we also assume that our blagn host galaxies are drawn from the star - forming main sequence in which sfr correlates well with galaxy stellar mass .
the motivations are as follows .
first , we demonstrate below that this is a good assumption , as the apparent factor of two excess of sfr seen in section [ sec : sfr ] and figure [ fig : sfrc ] is fully consistent with the bias from the _ herschel _ sensitivity limit .
moreover , @xcite also demonstrated that , after using simulations to model selection biases carefully , blagn hosts are consistent with being drawn from the star - forming main sequence relation .
note that blagns are also observed to be a factor of @xmath175 less common in quiescent galaxies than in blue star - forming hosts @xcite .
therefore , we can connect sfr with galaxy stellar mass by equation [ eq : ms ] .
the distribution of sfr ( @xmath176 ) for fixed @xmath177 is @xmath178 where @xmath179 , @xmath180 ( see equation [ eq : ms ] ) . the intrinsic scatter of the @xmath152 relation @xmath181 is roughly independent of both redshift and galaxy stellar mass ( e.g. , * ? ? ?
* ) . equations [ eq : gms ] and [ eq : msfr ] can be used to estimate the bias in the @xmath0 relation resulting from sample selections based on @xmath83 and sfr . while the sfr selection is directly related to the _ herschel _ sensitivity , the @xmath83 selection is a more complicated function of the agn luminosity limit and the broad - line width limit ( i.e. , @xmath182 ) .
motivated by @xcite , @xcite , and @xcite we related @xmath83 and @xmath138 by assuming our blagns have a log - normal distribution of the eddington ratio centered on @xmath94 with a scatter of @xmath183 dex and a cut off of @xmath184 and @xmath185 . using a different shape for the distribution ( e.g. , log - uniform )
turns out to have little effect on our bias estimate .
however , assuming a different mean eddington ratio for the distribution does have a small effect ; see section [ sec : mbias ] .
the fwhm of the emission line was related to @xmath83 and @xmath138 with a log - gaussian distribution centered on the value given by equation [ eq : mse ] with a scatter of @xmath186 @xcite . to mimic the @xmath12 bias
, we calculated the virial @xmath187 by using the distributions of @xmath138 ( converted to @xmath65 assuming a bolometric correction of @xmath188 ) and @xmath189 and equation [ eq : mse ] .
note that the scatter of the virial @xmath12 to true @xmath12 is @xmath190 dex @xcite .
we then apply the agn luminosity and sfr cuts to the distribution of @xmath191 to estimate the selection biases ( i.e. , a superposition of the lauer et al .
2007 and the @xmath83 biases ) .
we performed monte carlo simulations ( based on the previous arguments ) to estimate the selection biases in our sample .
we started by considering the galaxy stellar mass function of @xcite for star - forming galaxies : @xmath192 @xmath193 , @xmath194 , and @xmath195 ( at each redshift bin ) can be found from table 1 of @xcite ( note that the absolute value of @xmath195 is not important for the bias estimation ) . for each @xmath18 that was randomly drawn from the distribution of equation [ eq : smf ] ,
@xmath12 was randomly generated from the probability density function ( pdf ) of equation [ eq : gms ] with @xmath196 , @xmath197 , and @xmath198 @xcite .
in addition , for each @xmath18 , the sfr was randomly generated from the pdf of equation [ eq : msfr ] with @xmath199 @xcite .
the agn luminosity and line width were calculated as described in the previous paragraph by assuming @xmath200 and @xmath201 ( such that the uncertainty of the virial @xmath12 is @xmath73 dex ; we also tested other values of @xmath183 and @xmath186 and found that the bias is not sensitive to @xmath183 but depends more on the average value of the eddington ratio distribution see section [ sec : mbias ] for details ) .
we repeated the simulation @xmath202 times to get the simulated data , i.e. , our mock sample .
the pdf of these mock galaxies was estimated with kernel density estimation ( green contour in figure [ fig : bias ] ) .
we applied our sample selections to this mock sample , starting with the limits on agn luminosity and broad - line width . for agns in cosmos , which form the bulk of our sample , these limits are @xmath203 and @xmath204 @xcite
the resulting sub - sample is denoted as sample a. with these flux and width limits , our sample a is strongly affected by the @xcite bias , as seen by the yellow contour in figure [ fig : bias ] . in the left panel ( for @xmath205 ) , for example , we can clearly see a large offset between the @xmath0 relation of our sample a and that of our mock sample .
more quantitatively , @xmath206 for sample a. next , a redshift - dependent cut - off in sfr was also applied to sample a to mimic the _
detection limit .
the redshift - dependent sfr limit was estimated from the flux limit of _ herschel_/pacs in cosmos ( since most of our sources are from the cosmos field ; the limit is @xmath207 for @xmath205 , see , e.g. , * ? ? ?
the resulting sub - sample is denoted as sample b and is plotted as red points in figure [ fig : bias ] . for the left panel ,
the bias ( @xmath208 ) is much smaller compared with sample a. we also observed similar features in the right panel ( for @xmath209 ) : for sample a , @xmath210 ; for sample b , @xmath211 .
we also compared the @xmath152 relation of sample b with the star - forming main sequence and found that sources in sample b show higher star formation activity .
this result is simply due to selection effects : ( 1 ) there is intrinsic scatter in the `` main - sequence '' relation ; ( 2 ) the galaxy stellar mass function is bottom heavy .
that is , a star - forming galaxy with given sfr is more likely to be less massive , resulting in an apparent offset from the `` main - sequence '' relation . at @xmath16 , the median redshift for our sample , the @xmath212 of sample b is @xmath157 : the same
as the mean offset we obtained in section [ sec : sfr ] and figure [ fig : sfrc ] .
this result suggests that , after accounting for selection biases , our sources are consistent with being drawn from `` main - sequence '' galaxies .
therefore , it is valid to use the main - sequence relation in the simulations . for each redshift , we can perform similar simulations and therefore estimate the corresponding selection biases in sample b. through this procedure , we estimated the selection biases of sample b as a function of redshift .
we used this result to obtain the bias - corrected hr04 relation ( i.e. , a summation of the hr04 relation and the bias ) .
this bias - corrected relation will be included in section [ sec : msevol ] ( see the red solid line in figure [ fig : msr ] ) . the way we model the bias depends on the @xmath0 relation , the main - sequence relation , their intrinsic uncertainties , and the eddington - ratio distribution .
as will be shown in section [ sec : msevol ] , our measured uncertainty in the @xmath0 relation is @xmath160 dex , consistent with our assumed value .
the uncertainty of the star - forming `` main - sequence '' relation is not precisely known .
if we instead use @xmath213 as reported in @xcite for the `` normal '' star formation sequence , the bias of sample b at e.g. , @xmath205 decreases to @xmath214 ( the bias at @xmath215 remains almost the same ) .
we conservatively assume @xmath199 when estimating the bias of our sample .
the bias also depends on the connection between agn activity and star formation .
let us assume , for example , the agn luminosity is fully determined by galaxy properties ( e.g. , gas fueling ) and is not directly linked with @xmath12 . if so , there is no bias for the @xmath0 relation ( as already noted by * ? ? ?
the reason is simple : in this case , our sample would be based on galaxy properties rather than @xmath12 , just like the local sample .
some works suggest a link between agn activity and galaxy properties ( e.g , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* and this work ) . however , these agn galaxy relations may have large intrinsic scatter , perhaps making them inapplicable for modeling the bias in individual agns .
we also have limited knowledge of the distribution of the eddington ratio for blagns ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . in our simulations
, we also tried other eddington - ratio distributions , e.g. , a log - uniform distribution ( i.e. , close to the fiducial model proposed by * ? ? ? * ) or a log - normal distribution with a scatter of @xmath216 dex .
we additionally required these distributions to be truncated at both @xmath146 and @xmath217 .
these two cut - offs are motivated by both the theory of accretion disks ( see , e.g , * ? ? ?
* ; * ? ? ?
* ) and observations ( e.g. , * ? ? ?
* ; * ? ? ?
we found that all these distributions give similar results regarding selection biases as long as the mean value of these distributions is the same ( i.e. , @xmath94 ) . if we instead assume a smaller mean value for the eddington ratio , the selection biases increase ( and vice versa if we assume a larger mean value for eddington ratio ) .
this is simply due to the fact that the selection limit of @xmath12 increases with decreasing eddington ratio and the slope of the @xmath12 function becomes steeper with increasing @xmath12 . to illustrate its effect
, we adopted a power - law distribution of the eddington ratio with a slope of @xmath218 @xcite and a cut - off of @xmath219 and @xmath220 .
this distribution has a mean eddington ratio of @xmath221 .
the resulting selection bias on the @xmath17 relation is @xmath222 ( @xmath223 ) at @xmath205 ( @xmath209 ) .
future constraints on the eddington ratio distribution are crucial for a more reliable estimation of the biases .
regardless , we emphasize that our conclusions in section [ sec : msevol ] would not be significantly changed if the selection bias is essentially zero or as large as @xmath160 .
the detectability of a blagn actually depends on the luminosity contrast between the agn and the host galaxy , since this determines whether we can detect the broad h@xmath63 and/or emission lines ( and therefore measure @xmath83 ) .
this effect has not been discussed in previous simulations of the @xcite bias . to check the luminosity contrast in our simulated samples
, we used the 3d - hst catalogs @xcite .
the 3d - hst catalogs have redshift , stellar mass , age , sfr , and rest - frame colors for inactive galaxies and can be used to assign galaxy properties to our simulated samples . for each redshift bin , we started by dividing our mock sample into galaxy stellar mass bins .
as a second step , for each redshift and galaxy stellar mass bin , we selected main - sequence galaxies ( defined by equation [ eq : ms ] and its @xmath224 uncertainty ) from the 3d - hst catalogs whose redshifts and galaxy stellar masses lie within the bin .
then we used the selected sources to determine the distributions of the galaxy luminosity at ( rest - frame ) @xmath103 and the sdss@xmath47 band ( whose effective wavelength is @xmath225 , close to the wavelength of h@xmath63 ) for each bin .
finally , we assigned the ( rest - frame ) galaxy luminosity to our mock samples according to the distributions .
we again applied the limits in agn luminosity , fwhm , and sfr .
for the resulting samples , we calculated the distribution of the luminosity ratio of the agn to the host galaxy at @xmath103 ( for sources at @xmath226 , whose @xmath12 are measured using ) or the sdss-@xmath47 band ( for sources at @xmath227 whose @xmath12 are measured using h@xmath63 ) .
we found that for the mock sample a , there is a fraction of sources with @xmath228 at these wavelengths .
however , no source in the mock sample b ( i.e. , the one which suffers the same selection biases as our data ) has a luminosity ratio @xmath229 .
this suggests that our assumption of the agn contribution to the total sed in section [ sec : mgl ] is appropriate and does not introduce additional selection biases to our results . in order to measure the galaxy total stellar mass via two - component sed fitting ( see section [ sec : mgl ] ) , the galaxy must be detectable and not outshone by the agn in the @xmath60 band .
this requirement will introduce an upper limit to @xmath230 .
we calculated this limit for the galaxy emission to be at least @xmath102 of the agn emission in the @xmath60 band .
assuming an agn bolometric correction of @xmath107 at the @xmath60 band @xcite and an eddington ratio of @xmath94 ( see figure [ fig : bolmbh ] and e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , the @xmath60-band agn luminosity is @xmath231 meanwhile , the @xmath60-band galaxy luminosity can be written as @xmath232 where @xmath233 is the mass - to - light ratio .
we adopted @xmath234 for our @xmath16 star - forming galaxies ( * ? ? ? * with a correction for the imf of chabrier 2003 ) .
note that there is a weak dependence of @xmath233 on redshift , although this is not significant over the redshift range of our sources . with these parameters
the requirement of @xmath235 indicates that @xmath230 can not exceed @xmath236 .
this @xmath230 limit is more than two orders of magnitude larger than the mean value of the @xmath0 relation ( which suggests @xmath237 ) .
therefore , we expect this limit should not introduce significant biases to our data .
we also directly test the effect of the @xmath235 limit by assigning the rest - frame @xmath60-band luminosity to our monte carlo simulations of the @xmath0 relation , again finding that the galaxy detection limit does not introduce significant bias . in this subsection
, we explore the @xmath0 relation of blagns in star - forming galaxies .
the upper - left panel of figure [ fig : msr ] plots the distribution of our sources in the smbh mass galaxy total stellar mass plane . for comparison
, we also include the sample of local inactive galaxies from hr04 and the local smbh mass - bulge mass relation of hr04 ( see section @xmath238 in their paper ) .
however , our single epoch virial smbh mass estimators are calibrated by the local @xmath239 relation of @xcite which is consistent with the hr04 smbh mass - bulge mass relation .
that is , the geometry factor which converts the virial product into a smbh mass is derived from a @xmath239 relation that is consistent with the hr04 relation @xcite .
therefore we retain the older hr04 relation for consistency with the single epoch virial smbh mass estimates .
updating to the @xcite relation would require similar updating single epoch virial smbh estimates by ( roughly ) the same factor , resulting in no difference to our conclusions regarding evolution . ] and its @xmath7 uncertainty .
as seen from the upper - left panel of figure [ fig : msr ] , our sample is consistent with the hr04 relation .
in fact , only @xmath240 sources lie outside the @xmath7 uncertainty of the hr04 relation when considering error bars .
a mann - whitney u test of @xmath230 on our sample and the local inactive galaxies also indicates that the two samples are statistically consistent ( with a null probability @xmath241 ) .
the same conclusion holds for our low - redshift sub - sample , although considering the high - redshift subsample alone results in a small ( but statistically significant ) difference from the hr04 relation .
however , this is likely due to selection bias since the bias increases with redshift . indeed ,
if we take the selection bias ( see section [ sec : bias ] ) into consideration , the mann - whitney u test indicates that there is no significant difference between the high - redshift sub - sample and the local sample of hr04 ( with @xmath242 ) .
we also binned our sources by @xmath18 .
four bins were created : @xmath243 , @xmath244 , @xmath245 , and @xmath246 .
we calculated the median values of @xmath83 and @xmath18 and their @xmath7 uncertainties for each bin .
the upper - right panel of figure [ fig : msr ] plots the results . as a reference
, we again include the sample of local inactive galaxies from hr04 and the local smbh mass - bulge mass relation of hr04 and its @xmath7 uncertainty .
as seen from this panel , the binned data are consistent with the local smbh mass - bulge mass relation of hr04 .
we can also quantify possible evolution by testing the correlation between @xmath230 and redshift .
the lower panel of figure [ fig : msr ] plots this mass ratio and its uncertainty as a function of redshift . as a reference
, we also plot the local mass ratio ( black dashed line ) and its @xmath7 uncertainty .
for each redshift , we estimated the corresponding selection bias in section [ sec : bias ] .
the red solid line in the lower panel of figure [ fig : msr ] shows the expected hr04 relation after considering the selection bias .
we also calculated the mean value of @xmath230 and its @xmath85 uncertainty for sources with @xmath9 , @xmath247 , and @xmath248 ; red squares show the result . comparing the red squares with the red solid line
, we find our sample shares the same mean value of @xmath230 with the bias - corrected local one .
we also used the simulation - extrapolation ( simex ) technique @xcite to estimate the intrinsic scatter ( subtracting the measurement errors ) of the distribution of @xmath230 .
the simex estimation gives a scatter of @xmath249 dex for our sample .
this indicates that the intrinsic scatter of @xmath230 for our sample is not significantly larger than the local value .
our observed scatter in @xmath230 is somewhat affected by the biases , so we do not put too much emphasis on the similarity in scatter .
still , the similar scatter suggests that our estimation of selection bias is reliable , and that our sample is consistent with the hr04 relation ( also see section [ sec : dis1 ] ) .
this non - evolution result agrees well with , e.g. , @xcite , but it contrasts with some other works , e.g. , @xcite .
the latter claimed an evolution of @xmath250 at the @xmath251 significance level , although this is probably only due to the selection biases @xcite .
in the previous section , we find that the @xmath0 relation does not evolve with @xmath49 up to @xmath2 . in addition , the x - ray and fir data allow us to calculate the current _ mass growth rate _ of both agns and their host galaxies .
we use these mass growth rates to assess if our galaxies and agns will remain on or move away from the established @xmath0 relation .
we start by exploring the distribution of the specific smbh mass growth rate ( defined as @xmath252 ) and the specific galaxy total stellar mass growth rate ( defined as @xmath253 ) .
figure [ fig : gwr ] plots the specific mass growth rate distributions for the high - redshift and low - redshift sub - samples .
there are two points to emphasize : ( 1 ) in both sub - samples , the instantaneous specific smbh growth rate is not smaller than that of the galaxy ( especially for the low - redshift sub - sample , the instantaneous specific growth rate of the smbh is much larger than that of the host galaxy ) ; ( 2 ) the specific smbh growth rate does not evolve much ( the mean offset is @xmath131 dex ) while the specific galaxy growth rate decreases significantly ( the mean offset is @xmath254 dex ) from high to low redshift .
that is , our results indicate that the apparent @xmath255 seems to increase from high to low redshift .
this turns out to be largely due to selection effects ; after modeling the biases in section [ sec : dis ] , we find that the data are consistent with a uniform agn duty cycle at both high and low redshift .
in addition , for both sub - samples , the specific black hole mass growth rate distributions show negligible tails at the growth rate @xmath256 ( see figure [ fig : gwr ] ) , where @xmath257 .
these negligible tails agree with the conclusion of , e.g. , @xcite that blagns have a minimum eddington ratio of @xmath258 .
we use these specific growth rates to investigate the `` flow patterns '' ( following * ? ? ?
* ) of our sample in the @xmath0 plane
. figures [ fig : fpa ] and [ fig : fps ] plot the results .
the direction of the vector is defined by @xmath255 while the color map indicates the absolute specific growth rate ( i.e. , a summation of the specific smbh mass growth rate and the specific galaxy total stellar mass growth rate in quadrature ) along the vector ( which spans more than two orders of magnitude ) .
we again include the local hr04 relation and its @xmath7 uncertainty as the dashed line and the shaded region . for simplicity , these figures do not show the uncertainties of @xmath255 on the plot ( the uncertainties can instead be seen in figure [ fig : grms ] ) .
in general , agns and galaxies which are outliers in @xmath230 tend to have evolutionary vectors whose directions are anti - correlated with their positions : galaxies with over - massive smbhs tend to have lower @xmath255 , and galaxies with under - massive smbhs tend to have higher @xmath255 .
this anti - correlation behavior can be seen not only for the whole sample ( figure [ fig : fpa ] ) but also for each sub - sample ( the left and right panels of figure [ fig : fps ] ) . as will be shown in section [ sec : dis2 ] , this result is not solely due to selection effects . in sections [ sec :
dis2 ] and [ sec : dis3 ] , we demonstrate the anti - correlation between the ( instantaneous ) growth rate and mass ratios , if extended over some time with an agn duty cycle of @xmath94 , implies that agn activity and star formation could maintain the @xmath17 relation .
in section [ sec : msevol ] we showed that our sample exhibits no apparent redshift evolution in the @xmath0 relation .
therefore , our results along with , e.g. , @xcite and @xcite suggest that the @xmath0 relation does not evolve strongly with redshift . as stated before ,
the local inactive galaxies of hr04 are bulge - dominated ( @xmath259 )
. meanwhile , our sources ( at least a significant fraction ) are likely to have substantial disk components .
first , our sources show strong star formation activity which indicates non - negligible disk components ( for example , lang et al .
2014 suggest that star - forming galaxies are likely to have smaller bulge - to - total ratios ) .
in addition , for the x - ray luminosity and redshift ranges of our sources , a substantial fraction of agns reside in disk galaxies ( i.e. , @xmath260 ; see , e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
furthermore , we have also cross - matched our sample with that of @xcite ( for cosmos sources ) and @xcite ( for cdf - s sources ) and found that five out of the fourteen matched sources have srsic index much smaller than @xmath261 .
therefore , the @xmath5 relation is likely to evolve ( at least mildly ) with redshift . as suggested by @xcite , a redistribution of stellar mass from disk to bulge
is required to build the local @xmath5 relation .
our results suggest that in the next @xmath262 a typical galaxy in our sample would ( 1 ) increase its galaxy total stellar and smbh masses accordingly ; ( 2 ) transfer much of its stellar mass from its disk to its bulge .
the former indicates that gas fueling controls both agn activity and star formation ( see section [ sec : dis2 ] for more details ) . for the redistribution of stellar mass from disk to bulge ,
possible mechanisms are , e.g. , violent disk instabilities ( which are driven by cold gas , see e.g. , * ? ? ?
* ) and/or galaxy mergers @xcite . the former mechanism is proposed since our sources are likely to have substantial cold gas @xcite , although observational evidence of violent disk instabilities in agn hosts is mixed @xcite .
the latter mechanism is also attractive as it can not only redistribute stellar mass from disk into bulge but also help reduce the intrinsic scatter in the @xmath0 relation @xcite .
note that our agns are more likely fueled by cold gas rather than triggered by galaxy - galaxy mergers ( see , e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* and our next section ) .
therefore , if galaxy - galaxy mergers do play a role , they are likely to be dry mergers rather than gas - rich wet mergers .
currently , our estimation ( see section [ sec : msevol ] ) indicates that the intrinsic scatter of the @xmath0 relation is not significantly larger than that of the hr04 relation .
however , this estimation is also subject to selection biases . using our simulated samples in section [ sec : bias ]
, we found that a range of intrinsic scatter ( e.g. , @xmath263 dex ) can reproduce the observed scatter . on the other hand , our selection corrections in section [ sec : msevol ]
are made by assuming the @xmath0 relation has the same intrinsic scatter as the hr04 relation and our data are consistent with the bias - corrected hr04 relation .
this indicates that the intrinsic scatter of our sample should not be significantly larger than that of the hr04 relation ( i.e. , the intrinsic scatter is at most @xmath264 dex ) .
an average of one major merger for each of our galaxies is enough to reduce the intrinsic scatter by a factor of @xmath265 @xcite if the intrinsic scatter is indeed @xmath264 dex rather than @xmath160 dex . in section [ sec : tmevol ]
, we demonstrated that agns which are offset from the @xmath0 relation tend to have ( instantaneous ) evolutionary vectors which are anti - correlated with their mass offsets .
that is , galaxies with over - massive ( under - massive ) smbhs tend to have lower ( higher ) @xmath255 .
we quantify the anti - correlation between @xmath230 and @xmath255 by performing a linear fit ( via chi - squared minimization , taking the uncertainties into consideration ) .
results are plotted in the upper panels of figure [ fig : grms ] .
the red vertical dashed lines represent the local @xmath230 .
the red horizontal dashed lines correspond to @xmath266 .
there are more sources above the horizontal lines .
if the growth rate rates persist over some length of time , then this asymmetric distribution indicates the agn duty cycle is less than unity ( i.e. , the lifetime of an active smbh is smaller than the lifetime of star formation ) ; otherwise the smbh will become over - massive as we evolve our sources .
for example , if we assume our sources keep their current @xmath24 and sfr for some timescales @xmath267 and @xmath268 , we expect blagn hosts to evolve as @xmath269 . to maintain the @xmath0 relation , we expect that , on average , @xmath270 . in section [ sec : dis3 ] , we derive the appropriate agn duty cycle ( @xmath271 ) such that , under appropriate assumptions about the agn activity and star formation histories , the @xmath0 relation is maintained .
the red solid lines are for the best fits of the anti - correlation : for the high - redshift sub - sample , the best fit is @xmath272 ; for the low - redshift sub - sample , the best fit is @xmath273 . in each sub - sample , the slopes are consistent with @xmath274 and differ from zero at a high significance level ( for the high - redshift sub - sample , the null probability is @xmath275 ; for the low - redshift one , @xmath276 ) .
the anti - correlation between the _ specific _ growth rate ratio ( s@xmath24/ssfr ) and the mass ratio , with slope of @xmath274 , implies that the _ absolute _ growth rate ratio ( @xmath24/sfr ) is not correlated with the mass ratio and is a constant ( with small scatter ) .
we demonstrate this in the lower panels of figure [ fig : grms ] .
a linear fit ( via chi - squared minimization ) between @xmath277 and @xmath230 results in a slope of @xmath278 , i.e. , statistically consistent with zero .
the relatively narrow distribution of the @xmath24/sfr ratio is in agreement with @xcite , after accounting for the agn duty cycle ( see section [ sec : dis3 ] ) . to test whether this anti - correlation and the narrow distribution of @xmath24/sfr ratio are simply due to selection effects we again turn to the monte carlo simulations in section [ sec : result ] .
these simulations created mock samples with @xmath83 , @xmath18 , @xmath24 , and sfr following the same selection criteria as our data .
we tested three eddington - ratio distributions which are log - normal with scatter ( @xmath183 ) of @xmath216 , @xmath84 and @xmath136 , respectively .
meanwhile , we use the same @xmath152 and @xmath17 relations and scatter as in section 4 , and @xmath186 is set such that the uncertainty of the virial @xmath12 is @xmath279 dex .
we created 69 mock samples , one at each redshift of the 69 observed blagns , with @xmath280 sources in each mock sample . the initial mock sample ( i.e.
, the one without any selection cut - offs ) at each redshift does not show an anti - correlation between the ratio of the masses and their growth rates .
indeed , spearman s correlation test indicates a coefficient of @xmath281 for @xmath200 or weaker depending on @xmath183 ) .
we then applied the same ( redshift - dependent ) cut - offs presented in section [ sec : bias ] to these mock samples .
finally , we randomly selected @xmath282 sources from the 69 mock samples ( i.e. , selected one source per mock sample ) and calculated the corresponding @xmath283 ( for our blagns sample , @xmath284 ) and the scatter of the distribution of @xmath277 ( for our blagns sample , the scatter is @xmath136 ) .
note that , in this step , we added @xmath159 dex uncertainty to @xmath24 and @xmath168 dex uncertainty to sfr to mimic approximately the measurement error in each variable ( see section [ sec : err ] ) .
we repeated the selection @xmath202 times and constructed the distribution of @xmath283 .
we found that , of all the eddington - ratio distributions tested , the @xmath285 model comes closest to reproducing the observed anti - correlation between @xmath255 and @xmath17 , albeit with a probability of only @xmath286 .
the simulations also indicate that the probability to reproduce the observed scatter of the distribution of @xmath277 is only @xmath287 .
note that in the @xmath200 model , the scatter between agn luminosity and sfr is only @xmath288 dex . broader eddington - ratio distributions ( e.g. , uniform or shallow power - law ) decrease the likelihood of selection effects reproducing a constant @xmath24/sfr ratio , as do larger values of @xmath289 or @xmath290 .
moreover , we also performed the same simulations to the high- and low - redshift sub - samples and found the likelihood of selection effects also decrease . from the simulations
, we can conclude that the observed anti - correlation between specific growth ratio and mass ratio and/or the narrow distribution of @xmath291 are unlikely to be solely due to selection effects but instead support the idea that there is a ( instantaneous ) connection between star formation and smbh accretion activity ( see the lower panels of figure [ fig : grms ] ) .
however , this conclusion rests upon the ( rather uncertain ) assumed eddington - ratio distribution , as well as the similarly uncertain intrinsic scatter in @xmath12 and @xmath17 .
better estimates of these quantities and their scatter are necessary to fully confirm that our results are not purely a selection effect .
agn feedback may explain the anti - correlation between @xmath255 and @xmath292 and the constant ratio of @xmath24/sfr .
as has been long suggested , agn feedback may regulate star formation and therefore maintain a correlation between agn accretion and galaxy star formation ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . however , recent simulations with a more realistic ism structure suggest that agn feedback can inject substantial energy into the ism without having significant effect on cold gas in the galaxy ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . if so , our results are difficult to explain by agn feedback .
our results can also be understood in a scenario where agn activity and star formation are both governed by cold gas supply , without a need for agn feedback .
this scenario could lead to a connection between the agn luminosity and sfr or even an agn `` main sequence '' ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
for a host galaxy with an under - massive smbh , we would expect the specific smbh mass growth rate to be larger than that of the host galaxy since the mean ratio of the sfr to agn luminosity remains roughly constant with changing @xmath18 @xcite .
that is , an under - massive smbh has a preferentially steeper evolutionary vector than an over - massive smbh on the @xmath0 plane .
we are also interested in the agn duty cycle and star formation timescale , @xmath293 , that would maintain the @xmath0 relation as our sources evolve forward in time .
we started with a monte carlo simulation which re - sampled @xmath12 , @xmath18 , @xmath24 , and @xmath294 based on the observed values and errors for every source in our sample .
for each realization , we evolved every source in time and calculated their new positions in the @xmath0 plane by integrating @xmath295 and @xmath296 over a grid of agn duty cycles and @xmath293 .
for the high- ( low- ) redshift sub - sample , all sources were evolved from their current redshifts to @xmath205 ( @xmath297 ) .
@xmath295 was assumed to be @xmath298 , where @xmath299 was the re - sampled @xmath300 . as for @xmath296
, we assumed that , on average ( e.g. , over @xmath301 ) , @xmath302 .
that is , following similar derivations in the previous section , the new positions of our sources are ( @xmath303 , @xmath304 ) , where @xmath305 and @xmath306 ( as long as the evolution timescale is much larger than @xmath293 ) . @xmath307 and @xmath308 are the observed smbh mass and galaxy stellar mass , respectively .
we define the agn duty cycle as @xmath309 . assuming our agn hosts are drawn from the star - forming main sequence as suggested by @xcite this definition of agn duty cycle essentially measures the fraction of blagns among star - forming galaxies ( we will discuss this definition later ) .
we then determined the fraction ( @xmath310 ) of our sources whose new positions are consistent with the biased hr04 mass ratio ( see section [ sec : bias ] ) and its @xmath7 uncertainty for each combination of agn duty cycle and @xmath293 .
finally , we found the agn duty cycle and @xmath293 that result in the highest @xmath310 .
after @xmath311 re - sampled realizations , we obtained the distribution of the agn duty cycle and @xmath293 that would best maintain the @xmath0 relation .
the duty cycle used here is only equivalent to @xmath312 or @xmath313 if @xmath314 and @xmath315 are both very small ( i.e. , if @xmath316 and @xmath317 both are @xmath318 , then , @xmath319 ) .
this is true for our low - redshift sample .
if , however , @xmath316 and/or @xmath317 are @xmath320 ( e.g. , our high - redshift sample ) , the agn duty cycle is a complicated function of @xmath12 , @xmath18 , @xmath24 , and sfr . actually , for our assumed smbh accretion and star formation histories , our definition of the duty cycle measures the ratio of the lifetime of blagns to that of active star formation .
the two - dimensional pdf of the agn duty cycle and @xmath293 is plotted in figure [ fig : duty ] for the high - redshift ( left panel ) and low - redshift ( right panel ) sub - samples .
the blue contour in each panel represents the``@xmath7 '' scatter of the most likely agn duty cycle and @xmath293 .
that is , given the measurement errors of @xmath12 , @xmath18 , @xmath24 , and @xmath294 , there is a @xmath321 probability that agn duty cycles and @xmath293 which best maintain the @xmath0 relation are within the contour . to estimate the one - dimensional distribution of the agn duty cycle (
regardless of @xmath293 ) , we integrated the two - dimensional pdf along @xmath293 . from this distribution ,
the preferred agn duty cycle is @xmath322 ( agreeing with * ? ? ?
* ) in both the high- and low - redshift sub - samples ( @xmath323 at @xmath11 , and @xmath324 at @xmath9 , with errors defined by the 15.87 and 84.13 percentiles ) . in other words , our data favor a non - evolving ( by no more than a factor of @xmath325 ) agn duty cycle of @xmath6 for star - forming galaxies at both @xmath326 and @xmath327 , ruling out very high ( @xmath328 ) duty cycles at the @xmath329 confidence level .
the large confidence intervals for the agn duty cycle are partly due to the large uncertainties in the estimated masses and growth rates .
there may also be a large intrinsic scatter in agn duty cycle between different galaxies .
we note that these duty cycles are appropriate only for rapidly accreting smbhs ( i.e. , blagns or other agns with similar eddington ratios ) in star - forming galaxies .
lower - luminosity agns are likely to have higher duty cycles , while quiescent galaxies ( known to have lower agn fractions , e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) are likely to have lower duty cycles ( for example , as found by * ? ? ?
in this paper , we studied 69 blagns selected from the cosmos and cdf - s fields based on x - ray observations , fir observations , high - quality spectroscopy , and multi - band photometry . with these data , we simultaneously determined @xmath12 , @xmath18 , @xmath24 , and sfr and investigated the evolution of the @xmath0 relation .
our main conclusions are the following : \(1 ) our sample suggests that , up to @xmath215 , there is no evolution ( no more than @xmath131 dex ) of the @xmath17 relation .
the @xmath14 relation , however , is likely to evolve , simply because our galaxies are expected to be more disk - dominated than the local sample of hr04 .
see sections [ sec : msevol ] & [ sec : dis1 ] .
\(2 ) our data indicate an anti - correlation between @xmath330 and the ratio of the two specific mass growth rates .
such an anti - correlation suggests that the on - going agn activity and host - galaxy star formation are physically connected . under appropriate assumptions about the agn accretion and galaxy star formation histories , this anti - correlation can maintain the @xmath330 relation .
see sections [ sec : tmevol ] & [ sec : dis2 ] .
\(3 ) we also investigate the possible values of agn duty cycle which best maintain the non - evolving @xmath17 relation .
our data favor a non - evolving ( i.e. , a factor of @xmath331 ) agn duty cycle of about @xmath94 for rapidly accreting smbhs in star - forming galaxies ( @xmath323 at @xmath11 , and @xmath324 at @xmath9 ) . see section [ sec : dis3 ] .
our results can be advanced in several regards .
first , our sample size is significantly limited by the _ herschel _ sensitivity .
only @xmath332 of blagns are detected by _
herschel_. in this work we attempt to correct for this bias , but it would be better to simply have a less - biased sample .
future deeper rest - frame fir observations ( e.g. , with alma , scuba-2 , spica ) can enlarge the sample size and are crucial for reducing the uncertainty and improving our understanding of the connection between agn activity and star formation . as we have pointed out in section [ sec : bias ] , the biases we mentioned ( e.g. , the bias of @xmath83 alone and the bias of @xmath230 ) depend on this connection .
therefore , deeper fir surveys can also help us better address the selection biases in this work .
this improvement , in turn , providing benefits to both our understanding of the co - evolution of smbhs and their host galaxies ( e.g. , the formation of the local @xmath5 relation ) and smbh mass demographics . also , the large intrinsic scatter of @xmath83 prevents us from obtaining higher confidence level conclusions .
future improvements of the single - epoch virial estimators will be very helpful ( for example , with large multi - object reverberation mapping campaigns * ? ? ? * ) .
finally , more detailed observations and studies of physical properties ( other than @xmath18 ) of the host galaxies are also very important .
these properties ( e.g. , morphology and gas content ) can help place additional distinct constraints on the co - evolution path of smbhs and their host galaxies ( e.g , the bulge - to - total ratio and the agn / star formation triggering mechanism ) .
we would like to thank the anonymous referee for his / her helpful feedback that improved this work .
we thank angela bolzonella , michael eracleous , he gao , kazumi kashiyama , yuexing li , jessie runnoe , nicholas ross , yue shen , sydney sherman and guang yang for beneficial discussions . mys acknowledges support from the china scholarship council ( no . [ 2013]3009 ) , and the national natural science foundation of china under grant # 11222328 .
jrt acknowledges support by nasa through hubble fellowship grant # 51330 awarded by the space telescope science institute .
wnb and bl acknowledge support from _ chandra _ x - ray center grant ar3 - 14015 and nasa adp grant nnx10ac99 g .
dma gratefully acknowledges support from the science and technology facilities council through grant st / i001573/1 and the leverhulme trust .
yqx acknowledges support of the thousand young talents program ( kj2030220004 ) , the 973 program ( 2015cb857004 ) , the ustc startup funding ( zc9850290195 ) , the national natural science foundation of china ( nsfc-11473026 , 11421303 ) , and the strategic priority research program `` the emergence of cosmological structures '' of the chinese academy of sciences ( xdb09000000 ) .
this research has made use of data from the public data release of the pacs evolutionary probe pep @xcite and the goods_herschel _ programs @xcite as described in @xcite .
this research has also made use of data from the hermes project ( http://hermes.sussex.ac.uk/ ) .
hermes is a _
key programme utilising guaranteed time from the spire instrument team , esac scientists and a mission scientist .
the hermes data were accessed through the _ herschel _ database in marseille ( hedam - http://hedam.lam.fr ) operated by cesam and hosted by the laboratoire dastrophysique de marseille .
we gratefully acknowledge the contributions of the entire cosmos collaboration consisting of more than 80 scientists . more information on the cosmos survey is available at http://www.astro.caltech.edu/cosmos .
this work has made use of the nasa / ipac infrared science archive , which is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration .
this work has made use of catalogs from the 3d - hst treasury program ( go 12177 and 12328 ) using observations from the nasa / esa hst , which is operated by the association of universities for research in astronomy , inc .
, under nasa contract nas5 - 26555 .
this work has made use of astropy , a community - developed core python package for astronomy ( astropy collaboration , 2013 ) .
ccc total @xmath35 x - ray blagns & 221 & 30 + x - ray blagns with optical / uv and _ herschel _ photometry & 85 & 11 + final sample : x - ray blagns with optical / uv photometry , robust ir luminosity , good spectra & 62 & 7 + ccccccccccc 1 & 149.83875 & 2.67511 & 0.26 & s & cosmos & 7.07 & 10.88 & 44.51 & 44.93 & 0.06 + 2 & 150.15837 & 2.13961 & 1.83 & z & cosmos & 8.99 & 11.75 & 46.08 & 46.74 & 0.07 + 3 & 150.34592 & 2.14758 & 1.26 & z & cosmos & 8.28 & 11.23 & 45.38 & 46.17 & 0.08 + 4 & 150.0425 & 2.62917 & 1.57 & z & cosmos & 8.68 & 11.90 & 46.16 & 46.44 & 0.05 + 5 & 149.69879 & 2.44122 & 1.52 & i & cosmos & 8.74 & 10.52 & 45.77 & 46.17 & 0.07 + 6 & 149.881 & 2.45083 & 1.32 & z & cosmos & 8.01 & 11.05 & 45.28 & 46.11 & 0.02 + 7 & 149.99158 & 1.72428 & 1.62 & s & cosmos & 8.19 & 11.55 & 46.02 & 46.23 & 0.04 + 8 & 150.30954 & 2.39914 & 1.80 & s & cosmos & 8.81 & 11.49 & 46.46 & 46.35 & 0.04 + 9 & 150.62525 & 1.80289 & 0.63 & i & cosmos & 7.43 & 10.79 & 45.90 & 45.67 & 0.09 + 10 & 150.05871 & 2.47739 & 1.26 & z & cosmos & 8.58 & 11.52 & 45.66 & 46.03 & 0.03 + 11 & 149.70587 & 2.41975 & 1.12 & z & cosmos & 9.36 & 11.37 & 45.77 & 45.92 & 0.03 + 12 & 149.94171 & 2.79544 & 1.07 & s & cosmos & 8.67 & 11.38 & 46.19 & 45.90 & 0.07 + 13 & 149.58283 & 2.48433 & 0.34 & s & cosmos & 8.15 & 11.25 & 45.01 & 44.76 & 0.01 + 14 & 150.12508 & 2.86175 & 1.58 & i & cosmos & 8.30 & 11.80 & 46.08 & 45.90 & 0.06 + 15 & 150.32737 & 2.46094 & 1.05 & i & cosmos & 8.70 & 11.29 & 45.73 & 45.75 & 0.10 + 16 & 149.59246 & 1.75675 & 1.96 & s & cosmos & 8.90 & 11.23 & 47.15 & 46.66 & 0.26 + 17 & 150.14708 & 2.71747 & 1.18 & z & cosmos & 8.13 & 10.79 & 45.05 & 45.63 & 0.12 + 18 & 150.19558 & 2.00442 & 1.92 & z & cosmos & 9.02 & 11.14 & 46.20 & 46.22 & 0.11 + 19 & 150.63387 & 2.59369 & 0.66 & s & cosmos & 8.01 & 10.12 & 45.59 & 45.42 & 0.06 + 20 & 150.49567 & 2.41256 & 1.37 & z & cosmos & 8.43 & 11.19 & 45.37 & 45.64 & 0.16 + 21 & 150.57633 & 2.18142 & 0.55 & i & cosmos & 8.29 & 11.06 & 44.48 & 44.93 & 0.00 + 22 & 150.64521 & 2.71481 & 0.20 & s & cosmos & 6.55 & 10.63 & 43.81 & 44.07 & 0.04 + 23 & 150.19504 & 1.79383 & 1.87 & i & cosmos & 8.74 & 11.57 & 45.69 & 46.09 & 0.10 + 24 & 149.41992 & 2.03553 & 1.48 & i & cosmos & 9.12 & 11.49 & 45.74 & 45.84 & 0.10 + 25 & 150.58187 & 2.28769 & 1.34 & z & cosmos & 7.99 & 10.83 & 45.72 & 45.90 & 0.09 + 26 & 149.47792 & 2.64247 & 1.6 & s & cosmos & 8.64 & 11.77 & 46.26 & 45.94 & 0.22 + 27 & 149.86558 & 2.00306 & 1.25 & z & cosmos & 8.11 & 10.41 & 45.19 & 45.47 & 0.10 + 28 & 149.64183 & 2.74089 & 1.89 & s & cosmos & 8.72 & 10.18 & 46.45 & 46.03 & 0.24 + 29 & 150.658 & 2.7835 & 0.21 & i & cosmos & 6.95 & 10.66 & 44.85 & 44.29 & 0.04 + 30 & 150.55487 & 2.641 & 1.14 & z & cosmos & 8.51 & 10.76 & 45.33 & 45.14 & 0.35 + 31 & 149.58521 & 2.05111 & 1.35 & z & cosmos & 8.80 & 11.28 & 45.93 & 45.77 & 0.29 + 32 & 150.19975 & 2.19089 & 1.51 & z & cosmos & 8.89 & 11.18 & 45.56 & 45.57 & 0.16 + 33 & 150.11929 & 2.85353 & 0.77 & i & cosmos & 9.09 & 11.12 & 45.60 & 45.12 & 0.07 + 34 & 150.31367 & 2.80361 & 1.46 & i & cosmos & 9.06 & 10.80 & 45.69 & 45.58 & 0.25 + 35 & 149.91692 & 2.38522 & 1.13 & i & cosmos & 8.59 & 11.42 & 45.68 & 45.41 & 0.22 + 36 & 149.78971 & 2.32125 & 0.38 & i & cosmos & 7.39 & 10.55 & 44.63 & 44.14 & 0.01 + 37 & 150.6355 & 1.66917 & 1.79 & i & cosmos & 8.78 & 10.33 & 45.74 & 45.31 & 0.62 + 38 & 150.30004 & 2.70936 & 0.73 & i & cosmos & 7.61 & 9.33 & 44.50 & 44.77 & 0.09 + 39 & 150.13954 & 1.6365 & 0.52 & z & cosmos & 7.49 & 10.30 & 44.53 & 43.93 & 1.41 + 40 & 150.23083 & 2.57817 & 1.40 & z & cosmos & 8.87 & 11.35 & 45.87 & 46.02 & 0.20 + 41 & 150.80187 & 2.00061 & 1.78 & i & cosmos & 9.21 & 10.50 & 45.47 & 46.06 & 0.07 + 42 & 150.62771 & 2.741 & 0.82 & i & cosmos & 7.92 & 10.50 & 45.43 & 45.12 & 0.10 + 43 & 150.384 & 1.57247 & 1.36 & s & cosmos & 8.17 & 11.09 & 45.92 & 45.85 & 0.11 + 44 & 150.25267 & 2.48642 & 2.04 & i & cosmos & 9.06 & 11.39 & 45.89 & 46.16 & 0.13 + 45 & 150.24517 & 1.90008 & 1.56 & s & cosmos & 8.66 & 11.18 & 45.97 & 45.97 & 0.04 + 46 & 150.19467 & 2.06792 & 0.55 & i & cosmos & 7.29 & 11.17 & 44.93 & 44.77 & 0.08 + 47 & 150.13908 & 1.877 & 0.83 & i & cosmos & 8.12 & 11.07 & 45.18 & 45.09 & 0.01 + 48 & 150.006 & 2.81242 & 0.77 & s & cosmos & 8.38 & 10.77 & 45.59 & 45.10 & 0.10 + 49 & 149.86796 & 2.35192 & 0.35 & i & cosmos & 6.84 & 10.63 & 44.09 & 44.51 & 0.02 + 50 & 149.62429 & 2.18067 & 1.19 & i & cosmos & 8.43 & 10.99 & 45.55 & 45.57 & 0.13 + 51 & 149.4795 & 2.80183 & 1.11 & s & cosmos & 8.68 & 11.29 & 45.97 & 45.37 & 0.19 + 52 & 149.868 & 2.33075 & 1.49 & i & cosmos & 8.67 & 11.18 & 45.67 & 46.10 & 0.07 + 53 & 150.55046 & 1.709 & 0.37 & i & cosmos & 7.35 & 10.44 & 44.49 & 44.34 & 0.08 + 54 & 149.6692 & 2.07406 & 0.34 & z & cosmos & 8.31 & 10.67 & 44.42 & 44.17 & 0.10 + 55 & 150.10521 & 1.98117 & 0.37 & s & cosmos & 8.12 & 8.75 & 45.23 & 44.18 & 0.13 + 56 & 150.05379 & 2.58967 & 0.7 & s & cosmos & 8.33 & 11.21 & 45.78 & 44.88 & 0.11 + 57 & 150.1237 & 2.35825 & 0.73 & z & cosmos & 7.21 & 10.62 & 44.83 & 44.91 & 0.11 + 58 & 150.68317 & 2.57461 & 0.38 & z & cosmos & 7.75 & 10.31 & 44.78 & 44.07 & 0.12 + 59 & 150.2082 & 1.87536 & 1.16 & i & cosmos & 8.83 & 10.74 & 45.36 & 45.31 & 0.14 + 60 & 149.89484 & 2.17444 & 1.32 & z & cosmos & 8.57 & 10.90 & 45.19 & 45.75 & 0.09 + 61 & 149.76346 & 2.33411 & 1.13 & z & cosmos & 8.08 & 10.94 & 45.28 & 45.37 & 0.13 + 62 & 149.66362 & 2.08519 & 1.22 & z & cosmos & 8.19 & 10.80 & 45.54 & 45.48 & 0.13 + 63 & 53.22033 & -27.85556 & 1.23 & s04 & cdf - s & 8.25 & 10.74 & 45.06 & 44.96 & 0.07 + 64 & 53.11037 & -27.67658 & 1.03 & s04 & cdf - s & 7.80 & 10.64 & 45.77 & 45.55 & 0.02 + 65 & 53.10483 & -27.70525 & 1.62 & s04 & cdf - s & 8.74 & 11.00 & 45.64 & 46.15 & 0.04 + 66 & 53.07146 & -27.71761 & 0.57 & s04 & cdf - s & 7.65 & 10.35 & 44.72 & 45.33 & 0.14 + 67 & 53.0455 & -27.73756 & 1.615 & s04 & cdf - s & 8.84 & 11.06 & 45.58 & 45.58 & 0.01 + 68 & 53.25642 & -27.76183 & 0.62 & s04 & cdf - s & 7.96 & 10.82 & 44.72 & 44.85 & 0.03 + 69 & 53.12525 & -27.75656 & 0.96 & s04 & cdf - s & 7.43 & 10.67 & 44.75 & 45.59 & 0.02 + | we measure the location and evolutionary vectors of 69 _ herschel_-detected broad - line active galactic nuclei ( blagns ) in the @xmath0 plane .
blagns are selected from the cosmos and cdf - s fields , and span the redshift range @xmath1 .
black - hole masses are calculated using archival spectroscopy and single - epoch virial mass estimators , and galaxy total stellar masses are calculated by fitting the spectral energy distribution ( subtracting the blagn component ) .
the mass - growth rates of both the black hole and galaxy are calculated using _ chandra_/_xmm - newton _ x - ray and _ herschel _ far - infrared data , reliable measures of the blagn accretion and galaxy star formation rates , respectively .
we use monte carlo simulations to account for biases in our sample , due to both selection limits and the steep slope of the massive end of the galaxy stellar - mass distribution .
we find our sample is consistent with no evolution in the @xmath0 relation from @xmath2 to @xmath3 .
blagns and their host galaxies which lie off the black hole mass @xmath4 galaxy total stellar mass relation tend to have evolutionary vectors anti - correlated with their mass ratios : that is , galaxies with over - massive ( under - massive ) black holes tend to have a low ( high ) ratio of the specific accretion rate to the specific star formation rate .
we also use the measured growth rates to estimate the preferred agn duty cycle for our galaxies to evolve most consistently onto the local @xmath5 relation . under reasonable assumptions of exponentially declining star formation histories ,
the data suggest a non - evolving ( no more than a factor of a few ) blagn duty cycle among star - forming galaxies of @xmath6 ( @xmath7 range of @xmath8 at @xmath9 and @xmath10 at @xmath11 ) . |
electronic devices have been reduced more and more in size over the last decades .
furthermore it is now possible to place atoms or molecules accurately between macroscopic electrodes , hence experimental studies of the electronic transport for single atoms @xcite , molecules @xcite , and nanowires @xcite have become available .
one - dimensional structures are of particular interest due to their restricted transport channels , making them prototypical model systems .
for example , mono - atomic chains have been realized by molecular beam epitaxy @xcite .
using the tip of a scanning tunneling microscope , it has been possible to place a row of eight atoms on a nial substrate @xcite . from the theoretical point of view , transport through distorted one - dimensional systems has been addressed already in the early 1990ies by bosonization techniques @xcite .
transport properties of lattice models are currently investigated on several levels .
comparison of an exact treatment by the density matrix renormalization group @xcite with approaches using density functional theory @xcite shows that the latter approach often is sufficient to calculate the linear conductance , at least qualitatively .
the generalization of density functional theory to time - dependent potentials @xcite allows to study the propagation in time of the electronic states @xcite . to model the experimental setup more realistically ,
several methods have been developed for describing the transmission through nano - contacts .
tight - binding formulations have been applied to metallic nano - contacts @xcite , adding orbital information from chemical analysis ; however , these may fail in the contact regime .
nowadays , most apporaches rely on a combination of density functional theory and a scattering approach on the non - equilibrium green s function level , based on the landauer
bttiker scheme @xcite . in this article , the electronic structure and the transmission coefficient are determined for h and li chains with defects . in sec .
ii we present details of the calculational method and the structural setup , and discuss in sec .
iii some results for h chains and their dependence on the calculational parameters . in sec .
iv we turn to impurities in li chains , focusing on the interrelation between the energy level spectrum of the scattering region and the transport properties .
we calculate the transmission through nanowires using a combination of scattering theory and density functional theory @xcite , based on the smeagol and siesta @xcite codes . a single zeta basis set and the local density approximation for the exchange - correlation potential are used . in order to model the experimental situation ( without gate voltage )
, metallic leads are connected to a central scattering region . in our case , the leads are h chains , see fig . [ figstruct_asym ] . in the landauer
bttiker formalism , the self - energies of the left ( l ) and right ( r ) leads are calculated first .
the screening within the metallic leads ensures that effects of the contact region decay within a few nanometers .
since the leads are connected to the central scattering region ( molecule , nano - contact , or interface ) , an effective description of the central region ( c ) emerges which includes the properties of the leads . in linear response ,
the transmission coefficient is given by the retarded green s function @xmath0 of the central region and the lead self energies @xmath1 .
with @xmath2 $ ] we have @xcite @xmath3 \ ; .\ ] ] the conductance is given by @xmath4 , where the factor 2 accounts for the spin degeneracy @xcite .
the impurity models studied in sec .
iv are displayed in fig .
[ figstruct_asym ] .
the parent structure is a finite ( metallic ) li chain with three atoms , which is coupled to h - chain leads . by varying the distance between the last h and the first li atom
, we change the coupling from strong ( @xmath5 ) to weak ( @xmath6 ) .
the distance between the li atoms within the chain is fixed to @xmath7 .
the impurities are li atoms adjacent to the parent chain , thus breaking the rotational and/or inversion symmetry of the system . in configuration ( a )
an additional li atom is placed directly below a chain atom , compare @xcite .
the distance @xmath8 between the ad - atom and the chain is varied . a second atom on the other side of the chain , configuration ( b ) , restores the symmetry . an additional atom can also be placed below a li - li bond , configuration ( c ) , which also breaks the symmetry in the transport direction .
in addition , we study configuration ( d ) in which we place the ad - atom below the central bond , to obtain a system with inversion symmetry .
as the transmission coefficient shows resonance peaks at the energy levels of the scattering region , an analysis of this spectrum allows an explanation of the behavior of @xmath9 . for this purpose
, the energy levels can be discussed on the basis of a tight - binding approximation with nearest - neighbor hopping . with the hopping parameters indicated in fig .
[ figstruct_asym ] , the hamiltonian for a three - atom li chain is given by @xmath10 where the sites are numbered 1 , 2 , and 3 .
the hopping energy along the chain , @xmath11 , is of the order of 2 ev , as the bandwidth of a li chain with atomic spacing @xmath7 is @xmath12 ev .
the three - atom chain has the energy levels @xmath13 , and @xmath14 .
the coupling to the leads can be taken into account by an additional hopping to the left and to the right of the scattering region .
the hopping to the ad - atoms is denoted by @xmath15 , thus @xmath16 for configuration ( a ) , and @xmath17 for ( b ) . in case of configurations ( c ) and ( d ) ,
the hopping terms are given by @xmath18 , and @xmath19 , respectively . between the ad - atom(s ) and
the chain is varied .
the @xmath20-parameters refer to the tight - binding description of the central scattering region ; see main text.,scaledwidth=50.0% ]
we start our discussion with a review of results for h chains , which is the simplest case of one - dimensional scatterers , see the structural setup in fig .
he and coworkers @xcite have used a system of infinite h chains as leads and a finite six - atom h chain for their transport calculations , focusing on the influence of the exchange - correlation potential and the according self - interaction errors . in the following
we use the local density approximation , since it reproduces well the qualitative behavior of the transmission coefficient @xcite .
clearly , @xmath9 depends on the energy levels of the central h chain , and on the coupling strengths to the leads .
the number of atoms within the chain determines the number of energy levels , and hence the number of resonances , see fig .
a five - atom chain is found to be metallic , whereas a six - atom chain turns out to be insulating .
this odd - even behavior is reflected by a maximum or minimum in @xmath21 in fig .
moreover , the transmission peaks become broader with increasing coupling to the leads , and also shift in energy . considering in addition the bond length within the chain , @xmath22 , as a parameter
, we find , as expected , a decrease of the band width with increasing @xmath22 . for @xmath23
the bands are close to parabolic , and the unoccupied levels are more sensitive when decreasing the coupling than the occupied levels . with increasing @xmath22 ,
the dispersion becomes more and more symmetric and cosine shaped , consistent with the tight - binding description @xcite .
keeping the lead - chain distance fixed at , say , @xmath24 but increasing the bond length @xmath22 beyond this value , one arrives at a situation where the first and the last atom of the central region are effectively bound to the leads , i.e. , the scattering region is a diluted chain , however , with two atoms less @xcite . between the scattering region and the leads .
the bond lengths in the scattering region and in the leads are fixed at 1.0 .,title="fig:",scaledwidth=70.0% ] + between the scattering region and the leads .
the bond lengths in the scattering region and in the leads are fixed at 1.0 .,title="fig:",scaledwidth=45.0% ] between the scattering region and the leads .
the bond lengths in the scattering region and in the leads are fixed at 1.0 .,title="fig:",scaledwidth=45.0% ]
when a finite li chain is coupled to h chain leads , charge is transfered between the leads and the scattering region . for an even number of li atoms this charge transfer results in a level shift towards @xmath25 @xcite . extending the studies of @xcite
, we discuss in the following the influence of perturbations . for configuration ( a ) , with an additional li atom adjacent to the li chain ,
the tight - binding energy levels of the scatterer are given by @xmath26 , and @xmath27 .
the coupling to the leads lifts the degeneracy of the zero - energy level .
a comparison of the density of states ( dos ) obtained by siesta with @xmath9 is given in fig .
[ fig3](a ) for a projection of the dos onto the li atoms .
the distance of the ad - atom to the chain is @xmath28 .
the four energy levels of the scatterer are located at @xmath29 ev , 0 ev @xmath30 , 0.3 ev , and 1.7 ev .
the smaller peak at @xmath31 ev as well as the shoulder at 1.6 ev arise from hybridization with atoms in the leads . when the distance between the ad - atom and the li chain is increased , the splitting of the states near @xmath25 becomes more pronounced , see fig.[fig3](b )
. analyzing the spatial distribution of the charge for each energy level yields strong similarities between the tight - binding model and the dft calculation .
the lowest level corresponds to a homogeneous charge distribution over the scatterer .
the second level ( highest occupied molecular orbital , homo ) resides on li atoms 1 , 3 , and 4 , whereas the third level ( lowest unoccupied molecular orbital , lumo ) is dominated by the ad - atom , see fig .
[ fig3](c ) . in the range
@xmath32 a redistribution of weight from the homo to the lumo is apparent . for increasing distance @xmath8 ,
the charge becomes more localized on the ad - atom , resulting in a stronger transmission through the lumo .
the minimum of the total energy appears for @xmath33 ; then the energy increases almost linearly when increasing @xmath8 ( up to @xmath34 ) .
( b ) transmission coefficient for various distances between chain and impurity .
( c ) charge density iso - surfaces of the homo and the lumo .
( d ) zoom of ( b ) around @xmath25.,title="fig:",scaledwidth=45.0%](a ) .
( b ) transmission coefficient for various distances between chain and impurity .
( c ) charge density iso - surfaces of the homo and the lumo .
( d ) zoom of ( b ) around @xmath25.,title="fig:",scaledwidth=45.0%](b ) + .
( b ) transmission coefficient for various distances between chain and impurity .
( c ) charge density iso - surfaces of the homo and the lumo .
( d ) zoom of ( b ) around @xmath25.,title="fig:",scaledwidth=45.0%](c ) .
( b ) transmission coefficient for various distances between chain and impurity .
( c ) charge density iso - surfaces of the homo and the lumo .
( d ) zoom of ( b ) around @xmath25.,title="fig:",scaledwidth=45.0%](d ) for the symmetrical configuration ( b ) , see fig .
[ fig_confb ] , we find five energy levels in the tight - binding spectrum : bonding and antibonding levels @xmath35 as well as @xmath36 .
a next - nearest - neighbor hopping @xmath37 shifts one of the zero - energy levels , say , @xmath38 , to a finite value , given by @xmath39 ( which holds in the limit @xmath40 ) .
the coupling to the leads also lifts the threefold degeneracy of @xmath41 . as a result ,
a spectrum of five levels is obtained , with two nearly degenerate levels near the fermi energy , see fig . [ fig_confb ] .
level 4 , corresponding to @xmath42 , has a rather homogeneous charge distribution over the scattering region for a small distance , @xmath43 . with increasing distance , the contribution of atom 2 to the charge density decreases , and hence the transmission through this level .
the position of the fourth peak in @xmath9 depends strongly on the distance between the ad - atoms and the chain . for @xmath28 , for example
, it is located at about 0.5 ev , whereas it is found at 1.0 ev for @xmath44 (equilibrium distance ) .
the two peaks near @xmath25 can not be resolved on the left hand side of fig .
[ fig_confb ] due to the coupling to the leads . for weak coupling a three - peak structure around @xmath25
is resolved , again showing the redistribution of weight below and above @xmath25 , compare fig .
[ fig3](d ) with the right hand side of fig .
[ fig_confb ] .
, left ) and weak ( @xmath6 , right ) coupling to the leads.,title="fig:",scaledwidth=45.0% ] , left ) and weak ( @xmath6 , right ) coupling to the leads.,title="fig:",scaledwidth=45.0% ] configuration ( c ) corresponds to a scatterer which is asymmetric in both in - plane directions .
the tight - binding model yields the spectrum @xmath45 and @xmath46 provided @xmath47 is small @xmath48 , i.e. , when the distance between ad - atom and chain is large .
the first guess thus is that the system should be insulating , in contrast to the parent chain and to the configuration where the ad - atom is attached to a single chain atom . according to fig .
[ fig5 ] , however , @xmath9 does not show a minimum at @xmath25 : in fact , due to the charge transfer with the leads the homo is fixed near @xmath25 , as demonstrated in @xcite ( also within dft ) .
in addition , the lowest level shifts slightly in energy when varying @xmath8 .
note that for @xmath43 the distance between ad - atom and chain atoms is only about @xmath49 hence smaller than the intra - chain nearest - neighbor distance of @xmath44 .
nevertheless , @xmath9 for @xmath43 essentially is the same as the transmission for larger distances . for @xmath8 above 3.0 the homo and its transmission peak remain unchanged ; however , the lumo shifts to lower energy with increasing @xmath8 ( i.e. , for decreasing @xmath47 ) .
, left ) and weak ( @xmath6 , right ) coupling to the leads.,title="fig:",scaledwidth=45.0% ] , left ) and weak ( @xmath6 , right ) coupling to the leads.,title="fig:",scaledwidth=45.0% ] symmetry in the transport direction can be obtained by placing the ad - atom below the central bond of a four - atom li chain , see configuration ( d ) . in this case
the parent chain is insulating as @xmath50 . including the impurity yields @xmath51 , @xmath52 , and @xmath53 .
thus , the ad - atom again switches the electronic state of the chain , but now from insulating to metallic . if we couple the central chain to the leads , @xmath54 is shifted to higher energy , due to charge transfer , see fig .
[ fig6 ] . for @xmath55
the energy of the lumo and its transmission peak remain essentially unaltered , but the distance to the homo decreases as @xmath47 decreases , similar to the behavior of configuration ( c ) .
, left ) and weak ( @xmath6 , right ) coupling to the leads.,title="fig:",scaledwidth=45.0% ] , left ) and weak ( @xmath6 , right ) coupling to the leads.,title="fig:",scaledwidth=45.0% ] we next study the influence of an impurity different from the atoms in the central region , comparing ad - atom and intra - chain configurations .
as a prototypical example with high electronegativity we choose a fluor atom , which we place either adjacent to or incorporated into the li chain . in the former case
we expect , for an f atom attached to a single chain atom , that one li electron will be trapped in the li - f cluster , no longer contributing to transport . according to the left hand side of fig .
[ fig7 ] , the transmission spectrum of a seven - atom li chain with f ad - atom in the middle ( @xmath56 away from the chain ) consists of six peaks , confirming the above picture of one trapped electron .
we note that the chain remains metallic despite the impurity .
this can be understood by studying an equivalent system with a six - atom li chain and a weak link at the center . to model the weak link
, we increase the distance between the two central li atoms to 5 .
effectively , we obtain the transmission spectrum of two three - atom li chains , see the dashed curve on the left hand side of fig .
[ fig7 ] , i.e. , metallic behavior and a high transmission at @xmath25 .
the situation is different when the electronegative atom is placed within the chain , as shown on the right hand side of fig .
[ fig7 ] for a li - li - f - li configuration with a li - f distance of 1.5 .
we find that the f atom forms a site with large attractive potential , since one transmission peak appears at very low energy ( at about @xmath57 ev ) .
we note that the resonance level at @xmath25 does not attain a transmission of 1 . while an f ad - atom forms a neutral cluster with the adjacent li atom , an intra - chain f atom couples to both li neighbors and acts as strong attractive potential .
we have investigated the influence of impurities on the transmission of metallic and insulating monovalent li chains . we have demonstrated that an impurity connected to a single atom of the chain leads to a splitting of the zero - energy peak , causing a strong dependence of the conductance on structural details of the scatterer .
an impurity connected to two atoms of the chain with equal distance , on the other hand , switches the system from insulating to metallic behavior , or vice versa .
the transmission coefficient , however , in both cases shows a minimum at @xmath25 as a consequence of the charge transfer with the leads .
charge transfer with the leads prohibits a prediction of details of the transmission spectrum by tight - binding models of the scatterer .
electronegative ad - atoms localize electrons in the chain but do not decouple it , whereas electronegative impurities within the chain result in low - lying transmitting levels and , thus , shift the spectrum to higher energy .
our results demonstrate that switches based on mono - atomic chains show a high variability of possible modifications that can be used to tailor the switching behavior .
10 e. scheer , n. agrait , j. c. cuevas , a. l. yeyati , b. ludoph , a. martn - rodero , g. r. bollinger , j. m. van ruitenbeek , and c. urbina , nature * 394 * , 154 ( 1998 ) . k. hansen , s. k. nielsen , m. brandbyge , e. lgsgaard , i. stensgaard , and f. besenbacher , appl . phys
. lett .
* 77 * , 708 ( 2000 ) .
j. van ruitenbeek , e. scheer , and h. weber , _ contacting individual molecules using mechanically controllable break junctions _ in : g. cuniberti , g. fagas , and k. richter ( ed . ) , _ introducing molecular electronics _ ( springer , berlin , 2005 ) , pp .
253 - 274 . v. rodrigues , t. fuhrer , and d. ugarte , phys .
* 85 * , 4124 ( 2000 ) .
n. nilius , t. m. wallis , and w. ho , science * 297 * , 1853 ( 2002 ) .
c. l. kane and m. p. a. fisher , phys .
lett . * 68 * , 1220 ( 1992 ) .
i. safi and h. schulz , phys . rev .
b * 59 * , 3040 ( 1999 ) .
g. cuniberti , m. sassetti , and b. kramer , phys .
b * 57 * , 1515 ( 1998 ) .
p. schmitteckert and f. evers , phys .
lett . * 100 * , 086401 ( 2008 ) s. schenk , p. schwab , m. dzierzawa , and u. eckern , phys .
b * 83 * , 115128 ( 2011 ) .
e. runge and e. k. u. gross , phys .
lett . * 52 * , 997 ( 1984 ) .
s. kurth , g. stefanucci , c .- o .
almbladh , a. rubio , and e. k. u. gross , phys .
b * 72 * , 035308 ( 2005 ) .
l. chico and l. m. falicov , phys . rev .
b * 52 * , 6640 ( 1995 ) .
j. c. cuevas , a. l. yeyati , and a. martn - rodero , phys . rev .
* 80 * , 1066 ( 1998 ) .
m. bttiker , y. imry , r. landauer , and s. pinhas , phys .
b * 31 * , 6207 ( 1985 )
. a. r. rocha , v. m. garcia - suarez , s. w. bailey , c. j. lambert , j. ferrer , and s. sanvito , nature mater . * 4 * , 335 ( 2005 ) .
a. r. rocha , v. m. garcia - suarez , s. w. bailey , c. j. lambert , j. ferrer , and s. sanvito , phys .
b * 73 * , 085414 ( 2006 ) .
j. m. soler , e. artacho , j. d. gale , a. garca , j. junquera , p. ordejn , and d. snchez - portal , j. phys .
: condens .
matter * 14 * , 2745 ( 2002 ) .
y. meir and n. s. wingreen , phys .
68 * , 2512 ( 1992 ) .
a. chakrabarti , phys .
a * 366 * , 507 ( 2007 ) .
s. h. he , h. u. baranger , and w. yang , j. chem .
phys . * 126 * , 201012 ( 2007 ) .
t. wunderlich , diploma thesis , universitt augsburg ( 2009 ) , unpublished .
note that we use `` central region '' , `` scattering region '' , and `` scatterer '' synomymously . | we present electronic structure and transport calculations for hydrogen and lithium chains , using density functional theory and scattering theory on the green s function level , to systematically study impurity effects on the transmission coefficient . to this end
we address various impurity configurations .
tight - binding results allow us to interpret our the findings .
we analyze under which circumstances impurities lead to level splitting and/or can be used to switch between metallic and insulating states .
we also address the effects of strongly electronegative impurities . |
for decades , only about a dozen dwarf galaxies were known to orbit the milky way .
the majority of these systems corresponded to dwarf spheroidal ( dsph ) galaxies , the least luminous , but , by number , the dominant galaxy type in the present - day universe .
however , over the last five years , and thanks to the advent of the sloan digital sky survey ( sdss ; @xcite ) the field of dwarf galaxies in the milky way has been revolutionized .
to date , fourteen new systems have been detected as slight overdensities in star count maps using the sdss data ( @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) .
these recent discoveries have revealed a previously unknown population of `` ultra - faint '' systems which have extreme low luminosities , in some cases as low as @xmath14l@xmath15 @xcite , and in average comparable ( or lower ) to those of galactic globular clusters .
however , spectroscopic surveys of the majority of these systems reveal kinematics and metallicities in line with those of dwarf galaxies ( @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) .
dynamical mass estimates of the ultra - faint galaxies based on line - of - sight radial velocities indicate that these galaxies are extremely dark matter dominated , with central mass - to - light ratios ( m / l ) as high as @xmath16 in solar units .
these systems are thus good laboratories to constrain cosmological models ( e.g the missing satellites problem ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) and to study the properties of dark matter ( @xcite ; @xcite ; @xcite ) . however , these applications hinge critically on the assumption that the masses and density distributions in these systems are accurately known .
current mass estimates for the ultra - faint dwarfs are based on the assumption that the dynamical state of these systems have not been significantly affected by galactic tides , and therefore that they are near dynamical equilibrium .
there is circumstantial evidence for past tidal disturbance in a number of these satellites based on morphological studies .
@xcite for instance , found the hercules dsph to be highly elongated , with a major to minor axis ratio of 6:1 , larger than for any of the `` classical '' dsphs , and argued for a tidal origin of this elongation .
@xcite reported fairly distorted morphologies for several of the new ultra - faint dwarfs , although @xcite showed that results based on low number of star may not be statistically significant because they suffer from shot noise .
the strongest evidence for tidal interaction perhaps is for ursa major ii , which appears to be broken into several clumps , lies close to the great circle that includes the orphan stream ( @xcite ; @xcite ) and shows a velocity gradient along its major axis @xcite .
if dsphs galaxies are currently undergoing tidal stripping , or have in the past , then kinematical samples are expected to be `` contaminated '' with unbound or marginally bound stars .
this will impact subsequent dynamical modeling , often resulting in overestimated mass contents ( @xcite ; @xcite ) .
the degree of such contamination is a function of both the projection of the orbit along the line - of - sight and the strength of the tidal interaction .
therefore , even though tides do not affect appreciably the inner kinematics of dsphs until the latest stages of tidal disruption ( e.g. , @xcite ; @xcite ; @xcite ; @xcite ) , studies aimed at identifying the presence of tidal debris can help elucidate the dynamical state of these dwarf galaxies .
obvious tidal features around dsphs would indicate the presence of unbound stars and their effects on the derived masses would need to be investigated .
alternatively , a lack of clear detections in the plane of the sky would narrow the possibilities that kinematical samples suffer from contamination , although it would not automatically imply that an object has not been tidally affected .
tidal features could still exist but be aligned preferentially along the line - of - sight and thus be hard to detect .
ultra - faint dwarf galaxies are useful probes of galaxy formation on the smallest scales ( @xcite ; @xcite ) .
one of the outstanding questions related to their discovery is whether these galaxies formed intrinsically with such low luminosities , or whether they were born as brighter objects and attained their current luminosities through tidal mass loss .
current metallicity measurements ( e.g. , @xcite ) support the former scenario .
they show that the ultra - faint dwarfs are also the most metal poor of the dsphs , following the luminosity - metallicity trend found for the classical dsphs , a technique now known to fail at very low metallicities . ] . on the other hand ,
if tidal features around these objects are firmly detected , it would clearly support the latter hypothesis . given the importance of the questions at hand , it is essential that we investigate the dynamical state of these satellites . due to the extreme low luminosity of the ultra - faint dwarfs , and therefore low number of brighter stars available for spectroscopic studies ,
deep imaging is currently the only way to efficiently detect the presence of faint morphological features in the outskirts of these systems .
we expect any tidal debris to be of very low surface brightness ( @xcite ) so that detecting it via integrated light is virtually impossible
. however , these galaxies are sufficiently nearby to resolve individual stars .
thus , using matched - filter techniques it is possible to detect arbitrarily low surface brightness features ( @xmath17 mag arcsec@xmath4 , @xcite ; @xcite ) .
in this article we present the results of a deep , wide - field photometric survey of the coma berenices ( comber ) and ursa major ii ( umaii ) dsphs , claimed to be two of the most dark matter dominated dsphs @xcite , carried out with the megacam imager on the canada - france - hawaii telescope . in 2 , we present details about the observations and data reduction , as well as of artificial star tests carried out to determine our completeness levels and photometric uncertainties . in 3 ,
we recalculate the structural parameters of these systems using a maximum likelihood analysis similar to that of @xcite . in
4 we present the results of our morphological study .
we discuss the statistical significance of our density contour maps and the robustness of our result to the variation of different parameters .
we show that comber looks fairly regular in shape and find no signs of tidal debris down to a surface brightness limit of 32.4 mag arcsec@xmath4 , whereas umaii shows a significantly elongated and distorted shape , likely the result of tidal interaction with the milky way .
our discussion and conclusions are presented in 5 and 6 , respectively .
observations of both comber and umaii were made with the megacam imager on the canada - france - hawaii telescope ( cfht ) in queue mode .
megacam is a wide - field imager consisting of @xmath18 @xmath19 pixel ccds , covering almost a full @xmath20 deg@xmath21 field of view with a pixel scale of @xmath22/pixel .
umaii was observed on nights between february and march of 2008 , whereas comber was observed between april and may of 2008 . for each dsph , four different , slightly overlapping fields
were observed for a total area coverage of nearly @xmath23 deg@xmath21 in the case of comber , and @xmath24 deg@xmath21 in the case of umaii . in each field ,
the center of the dsph was placed in one of the corners so that , when combined , the galaxy is located at the center of the overlapping regions as shown in figure 1 . for each of the four fields
, we obtained eleven 270s dithered exposures in sloan @xmath25 and eleven 468s dithered exposures in sloan @xmath26 in mostly dark conditions with typical seeing of @xmath27 .
the dithering pattern was selected from the standard megacam operation options in order to cover both the small and large gaps between chips ( the largest vertical gaps in megacam are six times wider than the small gaps ) .
megacam data are pre - processed by the cfht staff using the `` elixir '' package ( @xcite ) prior to delivery .
the goal is to provide the user with frames that are corrected for the instrumental signature across the whole mosaic .
this pre - process includes bad pixel correction , bias subtraction and flat fielding . a preliminary astrometric and photometric solution is also included in the pre - processed headers .
the world coordinate system ( wcs ) information provided with the data is only approximate , and we refine it using the freely available scamp package as follows : terapix sextrator is run on all chips ( scamp reads in the output files generated by sextrator ) and output files are written in the fits - ldac format .
scamp is then run on all chips .
scamp uses the approximate wcs information in the frames headers as a starting point , and then computes astrometric solutions using , in our case , the sdss - data release 6 ( dr6 , @xcite ; @xcite ) reference catalog which is automatically downloaded from the vizier database .
typically , several hundred stars in common between each of our chips and sdss are used to compute the astrometry and final solutions . in comber , fields have typical global astrometric uncertainties of @xmath28 , while for fields in umaii they are slightly higher , with @xmath29 .
the output from scamp is a fits header file ( one per processed frame ) , which is then used to update the wcs information for that given chip .
the updated headers are then used to translate @xmath30 and @xmath31 positions into final equatorial coordinates . prior to performing point - source photometry on our data
, we split each mosaic frame into its 36 individual chips . to take full advantage of our 11 dithered exposures per field ( and per filter ) , we test two different methods for carrying out the photometry . for the first method , we carry out our photometry running first daophot / allstar on the individual , non - coadded frames and then running the allframe package on the resulting files as outlined in @xcite .
allframe performs photometry simultaneously on all 22 frames for a given field ( 11 per filter ) .
daophot / allstar are required prior to allframe to determine point spread function ( psf ) solutions for each chip as well as to generate starlists for them individually .
optimum starlists to be used as input by allframe are obtained by cross matching the daophot / allstar results for the individual frames using the daomatch and daomaster packages ( @xcite ) .
these packages also provide reasonably good estimates of the offsets between dithered individual exposures necessary to run allframe .
final output files from allframe are then combined into a single catalog . in our alternative reduction method ,
we coadd all 11 exposures per filter for each of the @xmath18 individual chips using the swarp package .
swarp uses the wcs information stored in the frames headers to correct for shifts , small rotation between chips and distortions .
however , since original wcs information in the headers is only approximate , we refine the astrometry ( as described in 2.2 ) for all frames prior to combining .
swarp then coadds the frames and the daophot ii / allstar package is used to do point - source photometry on the stacked images . using artificial star tests described below , we determine that the allframe method ( 2.3.1 ) yields slightly deeper and more accurate photometry and therefore we continue our data reduction and analysis using the catalog generated as described in the first part of this section .
finally , photometric calibration was carried out by comparison with the sdss - data release 7 ( dr7 , @xcite ) catalog .
we first match our final photometry with the sdss stellar catalog .
we typically found several hundred stars in common with sdss per chip . to determine zero point and color terms we only use stars with @xmath32 and @xmath33 .
the brighter limit is given by the saturation limit of our cfht data .
we then fit the equations : @xmath34 @xmath35 where @xmath36 and @xmath37 are the zero points and @xmath38 and @xmath39 are the respective color terms .
we do this for each chip individually in order to determine whether there are chip - to - chip variations . in all cases , we find that the chip - to - chip differences , for both zero points and color terms are lower than the uncertainties in the derived parameters , and therefore we combine the stars in all 36 chips to derive final zero point and color term values .
we do this via a linear least - squares fit weighting by the respective uncertainties in the photometric magnitudes ( as estimated by allframe ) and rejecting @xmath40 outliers .
we obtain zero point and color terms for each mosaic field independently ( four per dsph , eight total ) .
the resulting constants differ by less than 2% in all cases .
uncertainties in the zero points vary from @xmath41mag whereas uncertainties in the color terms are of the order of @xmath42 .
final photometric uncertainties and completeness levels are determined via artificial star tests .
we first generate a fake color - magnitude - diagram ( cmd ) from which we will randomly select artificial stars .
we populate the cmd with stars in the magnitude and color ranges of @xmath43 and @xmath44 respectively , with four times more stars in the @xmath45 range than in the brighter half of the cmd .
we then select one of the 11 dithered exposures as our reference frame . in all cases
we select the first exposure in the @xmath25 filter for this purpose .
the goal is to inject stars in all 22 frames ( per chip ) in the same fake ra and dec positions in order to mimic real observations .
artificial stars are then randomly selected from the fake cmd and are injected into the reference frame in a uniform grid with a spacing of @xmath46pixels in both the @xmath30 and @xmath31 directions . using the refined astrometric solution calculated by scamp we convert the @xmath30 and @xmath31 positions of the artificial stars into ra and dec coordinates which are then converted back to @xmath30 and @xmath31 positions , but in the reference frame of the other exposures .
care is taken so that all the stars fall in the common area between the 22 different exposures . in the end , nearly @xmath47 stars are introduced per chip .
we repeat this procedure ten times to improve our statistics , each time randomly offsetting the grid s zero - point position in @xmath30 and @xmath31 .
we then carry out photometry on the artificial stars using the allframe method in the exact same manner as we did for the science frames .
we perform this test on only one chip per mosaic frame due to computational constraints , for a total of four chips in each dsph . since each of these fields was observed under slightly different seeing and darkness conditions , their respective magnitude limits are also slightly different . thus , to set conservative overall completeness levels we select the values from our `` shallowest '' field in each dsph as our final numbers .
we find that this represents a better choice than determining our limits from our combined artificial star s photometry for all four chips . in the latter case ,
the deepest fields dominate the results .
the 50% and 90% overall completeness levels of our photometry thus correspond to @xmath48 and @xmath49 and @xmath50 and @xmath51 , respectively , for our comber fields , and @xmath52 and @xmath53 and @xmath54 and @xmath55 , respectively , for umaii .
figure 2 shows the resulting completeness levels . in order to clean our catalogs of galaxy interlopers and other , non - stellar detections ,
we apply cuts using daophot s sharpness ( @xmath56 ) and @xmath57 parameters . to define appropiate cuts
, we fit third - degree polynomials to the @xmath57 and @xmath56 distributions , as functions of @xmath25 magnitude , obtained from the artificial star tests . after applying these cuts to both our photometric catalog and artificial star photometry our 90% completeness levels drop to @xmath58 and @xmath59 for comber and @xmath60 and @xmath61 for umaii . in figures 3 and 4
, we show color - magnitude - diagrams ( cmd ) for the central regions of both galaxies , including 90% completeness levels . in both cases , we reach at least three magnitudes below the main sequence turn - off of these systems , improving by roughly an order of magnitude , with respect to the original sdss photometry , the number of likely members of each galaxy .
photometric uncertainties as a function of magnitude were derived by taking the difference between the artificial star s actual and measured magnitude and are shown as error bars in figures 3 and 4 .
our photometric catalogs for both comber and umaii contain an order of magnitude more stars than the sdss photometry , and thus provide us with an excellent opportunity to re - estimate the structural properties of these systems .
@xcite carried out a comprehensive analysis of the structural parameters of all the ultra - faints using the original sdss photometry . however , in the cases of comber and umaii , owing to their very low luminosities ( @xmath62 ) , only about @xmath63 and @xmath64 stars were found to belong to the respective galaxy and therefore the derived parameters ( and morphologies ) suffer from significant uncertainties due to low numbers statistics . for each galaxy
, we re - calculate the photometric center ( @xmath65 , @xmath66 ) , ellipticity ( @xmath67 ) , position angle ( @xmath68 ) , half - light radius and @xmath69 . ]
( @xmath70 ) and background density ( @xmath71 ) .
we try three different density profiles , exponential , plummer ( @xcite ) and empirical king ( @xcite ) , and obtain structural parameters for all three : @xmath72 @xmath73 @xmath74 where @xmath75 and @xmath76 are the exponential and plummer scale lengths and @xmath77 and @xmath78 correspond to the king core and tidal radii respectively .
the exponential scale length is related to the half - light radius by the relation @xmath79 while in the case of the plummer profile , @xmath76 is equivalent to @xmath70 .
we have followed a procedure similar to the one outlined in @xcite which relies on a maximum likelihood ( ml ) analysis of the data to constrain the structural parameters .
the basic idea of the method is as follows : we assume that the positions of the stars are well represented by a given density profile ( one of the three mentioned above ) which is , in turn , well described by a set of parameters @xmath80,@xmath81, ... ,@xmath82 .
we then maximize a function of the form : @xmath83 where @xmath84 is the probability of finding the datum @xmath85 given the set of parameters @xmath80 , @xmath81, ... ,@xmath82 .
in the case of an exponential profile , this function takes the form : @xmath86 where @xmath87 , @xmath88 and @xmath75 are expressed in terms of the structural parameters we want to determine . in practice
, we look for a global maximum @xmath89 by searching the @xmath90-dimensional parameter space . in our case ,
the parameter space is @xmath91-dimensional , with the free parameters @xmath92 , @xmath93 , @xmath67 , @xmath68 , @xmath94 and @xmath71 . in the case of a king profile
, we fix the background density using the value obtained for the plummer profile because of a degeneracy between @xmath71 and the king tidal radius . to find a solution , we use the method described in @xcite which relies on the amoeba simplex algorithm ( @xcite ) to search the parameter space .
this method is somewhat sensitive to the specified region of parameter space to be searched ( i.e. the initial guess and allowed range for the parameters ) but it runs considerably faster than using an iteratively refined grid . in order to derive uncertainties for the structural parameters , we carry out @xmath95 bootstrap ( resampling with replacement ) realizations of our data .
the distribution of a given parameter is well described by a gaussian with only minor deviations in some cases ( king core and tidal radii in particular ) , and therefore we fit gaussian functions and report their mean and standard deviation as the mean and 1-@xmath96 uncertainty for a given parameter . to facilitate comparison with previous studies , we first compare the results of our algorithm when applied to the sdss data with those derived by @xcite . to mimic their star selection procedure as closely as possible , we select stars within @xmath97 degree from the satellites centers , impose the conditions that @xmath98 , @xmath99 and , additionally , that stars live near the m92 fiducial line shifted to the distance of the respective dwarf satellite .
we perform the comparison only for the exponential profile .
for both galaxies , the resulting structural parameters are in very good agreement . however , since we determine our parameters via a bootstrap analysis , our derived uncertainties are significantly larger .
@xcite did not perform a bootstrap analysis of their data and estimated their uncertainties using the statistical properties of the @xmath100 function .
however , as shown in the same study , statistical fluctuations due to the low number of stars in these dsphs can dramatically affect the morphology as well as the best - fitting parameters and therefore we regard our uncertainties as more realistic .
we then apply the ml algorithm to our dataset .
we select stars above the 90% completeness levels and within a region around the main sequences of these objects ( in order to improve the signal to noise ratio of dsph versus milky way foreground stars ) . for this purpose ,
we use an isochrone for a population @xmath101gyr old and [ fe / h]@xmath102 ( @xmath103 ) from @xcite and define a region around the main sequences by the condition that stars have to lie within @xmath104mag from the isochrone .
we varied this distance between @xmath105@xmath106mag and found no significant changes in the final parameters .
we note that we do not match the isochrone to the blue edge of the main sequence as it is customary when isochrone matching is used to derive star cluster properties , but instead place the isochrone so that it goes through the middle of the main sequences . for comber , our ml method yields a @xmath107 arcmin , an ellipticity @xmath7 and a position angle @xmath108 for the case of an exponential profile with the parameters derived for the other profiles being very similar .
in the case of umaii we find @xmath109 arcmin , @xmath110 and @xmath111 .
the final sets of structural properties for both comber and umaii are presented in table 1 .
this exercise also allows us to compare the results obtained from both our cfht photometry and the shallower sdss data .
as it was pointed out by @xcite in their work on the hercules dsph , we find relatively good agreement between the results obtained with both datasets , but our photometry allows us to place much tighter constrains on the derived structural properties .
this is not surprising given that our deeper photometric database is less sensitive to shot - noise effects that plague the shallower sdss data .
figure 5 illustrates this point . in this figure
we show the distribution of four structural parameters for comber , @xmath92 , @xmath93 , @xmath70 and @xmath67 for all @xmath95 bootstraps and for both datasets where the smaller uncertainties derived with the cfht data are evident . in figures 6 and 7a
we show background subtracted density profiles for both satellites with the best exponential , plummer and king models overplotted .
these are not fits to the binned data points , but are models constructed with the best parameters found via the ml method . as it can be seen , in the case of comber , both the exponential and king profiles are adequate descriptions of the data , with the plummer profile being perhaps a less adequate one .
the case of umaii is quite different . neither profile does a good job matching the data .
in fact , as shown in figure 7b , the data are better matched by a shallower inner power law ( @xmath112 ) and a steeper one in the outer parts ( @xmath113 ) reminiscent of the density profile derived by @xcite for the botes iii stellar overdensity or the inner power law shown by the tidally stripped palomar 5 ( pal 5 ) globular cluster @xcite .
this result is consistent with a scenario wherein umaii has been significantly tidally stripped , as suggested by @xcite .
we will explore this possibility in more depth in 4 .
we estimate the absolute total magnitudes of comber and umaii following a procedure similar to that outlined by @xcite .
this method relies solely on the number of stars belonging to the galaxy , and not on their individual magnitudes . as pointed out by @xcite , @xcite and @xcite , the extreme low luminosity of the ultra - faint galaxies and therefore the low number of stars they contain ,
make traditional methods , like the addition of fluxes from individual stars , too sensitive to the inclusion ( or exclusion ) of potential members ( outliers ) . adding or subtracting a few red - giant - branch stars ( with absolute magnitudes as bright as @xmath114 for a metallicity of [ fe / h]@xmath115 ) can significantly alter the total luminosity measurement in these cases . we therefore follow the method briefly described below .
we first assume that the stellar populations of both systems are mainly comprised by old , metal - poor stars , and therefore are well described by a single population .
we can then model that population with a theoretical luminosity function to help estimate its luminosity .
for this exercise , we adopt luminosity functions from @xcite for a population @xmath101gyr old and with @xmath103 ( which corresponds to [ fe / h]@xmath102 ) .
this assumption is reasonable in light of the metallicity measurements by @xcite who show that both dwarf galaxies have mean metallicities around [ fe / h]@xmath116 .
in addition , we use theoretical luminosity functions computed using two different initial mass functions ( imf ) : a @xcite with cutoff at @xmath117m@xmath15 and a @xcite log - normal .
the theoretical luminosity function gives us the relative number of stars in magnitude bins , which can be integrated to obtain the total flux down to a given magnitude limit .
one of the parameters we determine using the ml algorithm is the background surface density @xmath71 which is related the number of stars @xmath118 that belong to the galaxies by : @xmath119 where @xmath120 is the total number of stars used to derive the structural parameters and @xmath121 represents the total area of our fields .
we use this @xmath118 to normalize the theoretical luminosity function . by integrating the luminosity function and correcting by this normalization factor
we obtain the actual flux corresponding to our galaxies down to the respective magnitude limits . to account for the light contributed by stars fainter than this limit
, we add the remaining normalized integrated flux . as mentioned above , this method differs from traditional ones , and our derived luminosities are conceptually different from those derived for brighter galaxies . in our method , two different galaxies comprised of the same population and with exactly the same number of stars will always have the same luminosity .
however , in practice this will most likely not be the case as one might expect galaxies with similar populations to show an intrinsic spread in luminosities even if they have the same number of stars . to account properly for this
we estimate uncertainties in our derived absolute magnitudes by carrying out a bootstrap analysis .
the procedure is as follows : we treat the theoretical luminosity function used to calculate the luminosities as a cumulative probability function ( down to our 90% completeness levels ) of the number of stars expected as a function of magnitude .
we then randomly draw a number @xmath118 of stars from the luminosity function and add their fluxes .
we do this 10,000 times . using this method , for the case of a salpeter imf we find , for comber , @xmath1 ( after using @xmath122 from @xcite ) . for umaii
we obtain @xmath2 . for the second choice of imf
we obtain @xmath1 for comber and @xmath123 for umaii .
the uncertainties quoted here do not include the uncertainties in the distance to this objects . to test further whether our method yields reliable values ,
we apply it to sdss data selected in the same way as described in 3.1 . for a choice of a salpeter or log - normal imf
our results are almost identical to those of @xcite but as in the case of deriving structural parameters our uncertainties are larger .
a main goal of this study is to re - assess to what extent galactic tides may be affecting the structure of comber and umaii .
ever since the early numerical simulations of disrupted satellites by @xcite and @xcite , the observed elongation of dsph galaxies have been associated with the degree of their tidal interaction with the milky way . in
3.1 we found that comber is slightly elongated ( @xmath124 ) roughly along the direction of the galactic center ( shown by the solid line in figure 8) .
umaii , on the other hand , is much more elongated , with @xmath125 but its elongation is in the east - west direction , nearly perpendicular to the direction of the galactic center ( umaii has galactic coordinates of [ l , b]=[152.5 , 37.5 ] ) .
while alignment with the galactic center is expected in intrinsically elongated satellites , it is not a sufficient condition to infer active tidal stripping in the case of comber .
tides can still affect a system without necessarily stripping stars off of that system ( c. simpson & k. johnston , in preparation ) .
the quality of our photometric data , which reach at least three magnitudes below the main sequence turn - off in both galaxies , allows us to address the tidal stripping question more directly by studying the morphology of these satellites , this time much more reliably than in previous studies . to look for potential tidal features ,
we create smoothed isodensity contour maps .
the photometric catalogs used to create these maps are the same as those we used to derive structural parameters , and include all star - like objects ( i.e. , after @xmath56 and @xmath57 cuts ) down to our 90% completeness levels that live in a region around the main sequence as defined in 3.1 . to make the maps ,
the positions of the stars are binned into @xmath126 bins which are subsequently spatially smoothed with an exponential filter of scale @xmath127 .
other reasonable bin sizes and exponential scales were tried as well , with no significant change in the overall results .
figure 8 shows the resulting map for comber .
this object shows fairly regular contours similar to those found using sdss photometry @xcite but at a much higher statistical significance .
we do not detect signs of potential tidal debris down to the @xmath40 isodensity contour level ( the lowest one shown in fig .
11 ) which corresponds to a surface brightness of @xmath128 mag arcsec@xmath4 .
the lack of significant elongation , clumpiness or irregularities in the morphology of comber indicate that it is unlikely that it is currently being significantly affected by tides .
this , coupled with the kinematics of comber measured by @xcite and the corresponding mass derivation , supports the interpretation that comber is in fact a stable dwarf galaxy .
even if comber is being tidally perturbed at levels below our detection limits , such low - level effects are not likely to alter its inner kinematics appreciably ( e.g. , @xcite ; @xcite ; @xcite ) .
we note that our countour maps do not show the actual smoothed surface brightness contours , but instead show `` significance contours '' , i.e , for each dwarf , we calculate the mean surface density and its standard deviation in areas away from the region dominated by the galaxy ( background density ) , and plot the background - subtracted local density divided by the standard deviation .
if the values of pixels in our surface density map are well described by a gaussian distribution , then the standard deviation used here would measure precisely how significant our features are in terms of @xmath96 values .
however , the distribution of pixels will , most likely , not be well described by a gaussian , and therefore we need an experiment to determine the true significance of features in our data . to do this
we first randomize the position of the stars , but leave the photometry untouched .
we then select stars photometrically and generate smoothed density maps in the same way as we do for the actual data .
when we perform this test , we find that @xmath40 overdensities randomly scattered across the field are not uncommon , but higher significance features are very rare . we therefore conclude that isolated @xmath40 features detected in our real maps are very likely random noise and do not reflect potential tidal features .
to assess both the robustness of the overall shape of comber and the significance of apparent substructure in its outer density contours , specifically a hint of elongation in the northeast - southwest direction , we carry out a different test . in this case
, we bootstrap our photometric samples ( as opposed to randomly assign positions to the stars ) and redo the maps .
figure 9 shows eight different bootstrap realizations of the data .
it can be inferred from the figure that hints of substructure in the outer parts of comber are not statistically significant , but the overall shape and structure of this system are well established with our data .
this is not surprising , given the improvement in the number of stars that belong to the satellite achieved with our photometry .
we conclude that comber is unlikely to be significantly affected by galactic tides and therefore it represents a solid case of a stable dwarf galaxy whose characteristic size is smaller than @xmath129pc .
in contrast to the regular morphology of comber , umaii looks entirely different .
figure 10 shows its isodensity contour map , where the @xmath40 contours are equivalent to @xmath130 mag arcsec@xmath4 . our photometry confirms previous findings that umaii is highly elongated and it shows that umaii is larger than previously reported , extending at least three times beyond its measured half - light radius , or nearly @xmath131pc ( @xmath132 ) on the sky .
more striking perhaps , is the fact that contours of umaii looks more like a boxy - like system than an elliptical one .
only the very inner parts of umaii resemble a spheroidal object . in figure 10
we also show the best - fit orbit derived for umaii by @xcite , based on the assumption that it is the progenitor of the `` orphan stream '' ( @xcite ; @xcite ) .
this shows that the observed east - west elongation of umaii does not match the predicted direction from the model , although this is not enough to rule out a connection between the two systems . to assess the statistical significance of our results we carry out the same tests we described in section 4.1 .
we find that remaking the surface density maps after randomly assigning coordinates to the stars yields identical results as in the comber case , namely , that isolated @xmath40 features are likely background noise but higher significance ones are real . likewise
, making contour maps after bootstrapping the data shows that the overall shape of umaii is fairly well established and is insensitive to resampling .
we illustrate this in figure 11 where we show isodensity contours for eight different bootstrap realizations of our umaii data .
@xcite found tentative evidence that umaii might be broken into several clumps ( see their figure 1 ) .
our contour map for this object does not show statistically significant substructure .
however , we are able to reproduce this results if we make contour maps use only stars brighter than @xmath133 .
we regard the presence of substructure in the inner parts of umaii as statistical fluctuations in its density due to low number of stars .
another of the ultra - faints that looks similarly elongated as umaii is the hercules dsph
. a recent study of this object by @xcite , who use deep photometry obtained with the large - binocular - telescope ( lbt ) , show that hercules has an ellipticity @xmath134 and extends at least @xmath135pc ( @xmath136 on the sky ) .
we measured an ellipticity @xmath125 for umai , but while this object resembles hercules in this regard , umaii s morphology is even more irregular .
our understanding of the ultra - faint milky way satellites hinge on having reliable morphologies and robust structural parameters .
tidal disruption can result in disturbed morphologies and tidal features which can be very low surface brightness , yet have a profound affect on the interpretation of a given object .
deep photometry is the only means to assess the presence or absence of faint tidal structure . in the absence of complete disruption , well determined sizes and luminosities are critical in calculating the total mass of a system . our main goal in this paper is to search for signs of tidal features and determine structural parameters around the milky way satellites comber and umaii .
our @xmath137 and @xmath25-band cfht photometry reaches three magnitudes below the main sequence turn - off in these systems , corresponding to stars which are two to three magnitudes fainter than currently accessible to spectroscopic studies .
we achieve similar surface brightness limits for both comber and umaii ( @xmath138 mag arcsec@xmath4 ) , yet find very different morphologies for these two ultra - faint satellites . as seen in figures 8 and 9
, comber is remarkably regular and devoid of potential tidal features , in stark contrast to the elongated and irregular structure seen in umaii . since low number of stars
can produce spurious features and/or shapes that can later be interpreted as signs of tidal stripping @xcite , we have outlined in 4.1 two tests which demonstrated the robustness of our maps .
we show that while the @xmath40 contours in these figures lack statistical significance , all higher @xmath96 contours are robust to resampling and should be regarded as a solid result .
comber is remarkably regular and devoid of morphological features potentially due to tidal debris down to a surface brightness level of @xmath3 mag arcsec@xmath4 .
we have found a slight elongation in the direction toward the galactic center but no discernible irregularities are visible .
we can not of course completely rule out the presence of unbound debris at fainter surface brightnesses , but , if present , it should reflect only mild , low - level tidal effects .
similarly , we can not rule out the presence of tidal tails along the line - of - sight . in the kinematic survey of comber , @xcite report a velocity dispersion for this object of @xmath139kms@xmath140 with only one possible interloper star beyond 3@xmath96 of the velocity distribution .
this suggests against the presence of tidal debris along the line of sight . in the absence of evidence to the contrary ,
we conclude that comber is in dynamical equilibrium , and therefore current mass determinations should be robust .
@xcite also report that radial velocity members of comber near the main sequence turn - off lie in a broader color region than the other ultra - faint dwarfs in their sample , which they attribute to the effects of multiple stellar populations of different ages and metallicities .
we have investigated this and find that this spread is a reflection of larger photometric errors in the shallower sdss data
. we will further investigate the stellar populations of the two objects presented here in a separate paper .
in contrast to comber , umaii shows signs of disequilibrium .
the contours of umaii are elongated , irregular and extend for at least @xmath141 , three times its @xmath70 .
only the very inner core of umaii looks somewhat spheroidal . unlike the case of most other dsphs ,
umaii s stellar density distribution is not well matched by any of the commonly used density profiles ( king , plummer or exponential ) , and instead it is better matched by two power laws .
several studies ( e.g. , @xcite ; @xcite ; @xcite ) show that a dwarf satellite initially in a plummer or king configuration develops a power law component in the outer parts as tidal debris is stripped by the milky way .
however , they also show that the inner parts of the satellite retain a core - like density until the very latest stages of tidal disruption . the lack of a proper core in the stellar density profile of umaii may be a indication that this system is in fact in the throes of destruction .
another way to assess qualitatively to what extent umaii may have been tidally affected , is by comparing the `` break '' in its density profile to those of other well - studied disrupting objects such as pal 5 or the sagittarius ( sgr ) dsph .
these objects show a clear `` break '' in their density distributions around the region where tidal debris is being stripped . in the case of pal 5
this happens at a surface density relative to the central one of @xmath142 whereas for sgr this value is closer to 200 . if for umaii we take the point where the inner power law changes slope as the `` break '' point , we obtain @xmath143 , higher than for pal 5 and sgr . in addition , this break in umaii occurs relatively closer to the center than in these two systems , at a break radius @xmath144 compared to @xmath145 for the latter .
all these observations further support a tidal scenario for umaii .
@xcite detect a velocity gradient along the major axis of umaii , measuring an @xmath146kms@xmath140 velocity difference between in eastern and western halves of this object in the same direction as the elongation seen in our deep photometry .
they also point out that the umaii s velocity dispersion of @xmath147 km s@xmath140 is an outlier in the observed trend of lower velocity dispersion with decreasing luminosity followed by other galactic dwarf galaxies ( see their figure 10a ) . given umaii s absolute magnitude of @xmath148
, a value of 34 km s@xmath140 would be more in line with the observed trend .
one possibility to explain this observation is that the velocity dispersion of umaii has been inflated by galactic tides .
we note however that these observations are limited to a small region inside @xmath70 of the system . despite the observational evidence discussed above in favor of the tidal stripping of umaii
, our current data set does not allow us to conclude that this object is completely unbound or out of dynamical equilibrium .
kinematical data in outer regions are required to determine more precisely the nature of umaii and to further explore its possible association with the orphan stream . in the context of the galaxy versus cluster issue raised by @xcite , our findings ( or lack thereof )
imply that comber is solidly situated in the gap ( in @xmath149 versus @xmath70 space ) between star clusters and dwarf galaxies , and likely can not be explained away as an evaporating cluster or dissolving dwarf galaxy .
metallicity measurements support this scenario .
while our photometric work presented here show that the stellar content of comber ( and umaii ) is consistent with being dominated by a very metal poor population of [ fe / h]@xmath150 ( see figs .
3 and 4 ) , a metallicity unusually low even for the most metal poor globular clusters , detailed spectroscopic studies reveal even more extreme numbers .
@xcite report a mean metallicity for both comber and umaii of [ fe / h]@xmath115 which falls squarely on the luminosity - metallicity relationship found for galaxies .
@xcite report high resolution abundances for three stars in comber with metallicities as low as [ fe / h]@xmath151 , which is more metal poor than any of the galactic globular clusters .
in addition , both studies show a metallicity spread of 0.6dex in this dsph , typical of a dwarf galaxies .
the same argument can be invoked to argue that umaii is a disrupting dwarf galaxy as opposed to a dissolving star cluster .
we have carried out a deep , wide - field photometric survey of the coma berenices and ursa major ii dwarf spheroidal galaxies using the megacam imager on cfht , reaching down to @xmath0 mag , more than three magnitudes below the main sequence turn - offs of these galactic satellites .
this increases roughly by an order of magnitude , with respect to the original sdss photometry , the number of stars that belong to the respective galaxies and that are available determination of their structural properties and for morphological studies .
we used a maximum likelihood analysis similar to the one used by @xcite and @xcite to calculate structural parameters for three different density profiles : king , plummer and exponential .
we find characteristic sizes of @xmath5pc ( @xmath6 ) and @xmath12pc ( @xmath13 ) for comber and umaii respectively ( from the exponential profile ) .
our results provide much tighter constraints on these structural parameters than possible with previous datasets , but are consistent with earlier determinations using sdss photometry .
we have re - calculated the total luminosities for both systems and find , for comber @xmath1 and for umaii @xmath2 ( for a choice of salpeter imf ) , in very good agreement with previous results , confirming that comber and umaii are among the faintest of the known dwarf satellites of the milky way .
we have also used a @xcite imf but the results remain virtually unchanged .
we have found that comber shows a fairly regular morphology with no clear detection of potential stripped material down to a surface brightness of @xmath3 mag arcsec@xmath4 .
additionally , its number density profile is reasonably well matched by a choice of either king , plummer or exponential profile .
we thus conclude that comber is likely a stable dwarf galaxy which would make it one of the most dark matter dominated of the dsph systems .
we have also studied the morphology of umaii and find that , unlike comber , it shows signs of being significantly disrupted .
umaii is larger than previously determined , extending at least @xmath9pc ( @xmath10 on the sky ) and it is also quite elongated .
its density profile and overall shape resemble a structure possibly in the latest stages of tidal destruction .
furthermore , its number density profile is not well matched by neither of the three profiles we tried and it is much better described by two power laws , further supporting a tidal scenario .
the overall 2d surface density distributions of both systems are not affected by shot - noise and are therefore robust .
we find no evidence for isolated tidal debris beyond the main bodies of comber and umaii to our surface brightness limits of @xmath3 and @xmath152 mag arcsec@xmath4 , respectively .
deep , wide - field imaging of the recently discovered ultra - faint galaxies currently lags behind spectroscopic observations of these objects .
we show in this paper that high quality , deep photometry is an equally important tool in studying the dynamical state of the ultra - faint dwarfs .
these data can also be used to constrain the star formation histories of comber and umaii which we will explore in a future contribution .
we acknowledge david sand , josh simon and gail gutowski for useful discussions and peter stetson for graciously providing copies of daophot and allframe .
m.g and b.w .
acknowledge support from the national science foundation under award number ast-0908752 .
lcccc @xmath153 ( h m s)&12:26:59.00&@xmath154 & 08:51:29.86 & @xmath155 + @xmath156 ( d m s)&+23:54:27.2&@xmath157 & + 63:07:59.2 & @xmath158 + @xmath159 ( arcmin ) & 5.8&@xmath160 & 14.1 & @xmath160 + @xmath159 ( pc ) & 74 & @xmath161 & 123 & @xmath162 + @xmath163 ( arcmin ) & 5.9&@xmath160 & 13.9 & @xmath160 + @xmath163 ( pc ) & 76&@xmath161 & 122 & @xmath162 + @xmath164&0.36&@xmath165 & 0.50 & @xmath166 + @xmath167 ( degrees)&@xmath16867.0&@xmath169 & @xmath16874.8 & @xmath170 + @xmath171 & 735&@xmath172 & 1335 & @xmath173 + @xmath174 ( arcmin)&4.2&@xmath175 & 10.8 & @xmath176 + @xmath174 ( pc)&54&@xmath177 & 94 & @xmath178 + @xmath179 ( arcmin)&27.9&@xmath180 & 64.2 & @xmath181 + @xmath179 ( pc)&355&@xmath182 & 560 & @xmath183 + @xmath149 ( salpeter ) & @xmath184 & @xmath185 & @xmath186 & @xmath175 + @xmath149 ( chabrier ) & @xmath184 & @xmath185 & @xmath187 & @xmath185 + | we present deep cfht / megacam photometry of the ultra - faint milky way satellite galaxies coma berenices ( comber ) and ursa major ii ( umaii ) .
these data extend to @xmath0 , corresponding to three magnitudes below the main sequence turn - offs in these galaxies .
we robustly calculate a total luminosity of @xmath1 for comber and @xmath2 for umaii , in agreement with previous results .
comber shows a fairly regular morphology with no signs of active tidal stripping down to a surface brightness limit of @xmath3 mag arcsec@xmath4 . using a maximum likelihood analysis
, we calculate the half - light radius of comber to be @xmath5pc ( @xmath6 ) and its ellipticity @xmath7 .
in contrast , umaii shows signs of on - going disruption .
we map its morphology down to @xmath8 mag arcsec@xmath4 and found that umaii is larger than previously determined , extending at least @xmath9pc ( @xmath10 on the sky ) and it is also quite elongated with an ellipticity of @xmath11 .
however , our estimate for the half - light radius , @xmath12pc ( @xmath13 ) is similar to previous results .
we discuss the implications of these findings in the context of potential indirect dark matter detections and galaxy formation .
we conclude that while comber appears to be a stable dwarf galaxy , umaii shows signs of on - going tidal interaction . |
random walks have been studied in abstract settings such as integer lattices @xmath7 or riemannian manifolds ( @xcite , @xcite , @xcite ) . in applied settings
there are many spatially explicit individual - based models ( ibms ) in which the behavior of the system is determined by the meeting of randomly moving agents .
the transmission of a pathogenic agent , the spread of a rumor , or the sharing of some property when randomly moving particles meet are examples that come to mind in biology , sociology , or physics ( @xcite , @xcite , @xcite , @xcite , @xcite ) . in many of these models the movement of agents
is conceptualized as discrete transitions between square or hexagonal cells ( @xcite ) .
however , such a stylized representation of individual movements may not always be entirely realistic .
although ibms are powerful tools for the description of complex systems , they suffer from a shortage of analytical results .
for example , if a susceptible and an infective agent move randomly in some bounded space , what is the probability of them meeting , and hence of the transmission of the infection ?
what is the average time until the meeting takes place ? in the present paper we begin to answer these questions by considering a random walk in a bounded region of the plane or on the sphere , which we denote by @xmath0 .
the model evolves in discrete time . at each time - step an agent leaves its current position at a uniformly distributed angle in the @xmath3 interval . on the plane
the agent moves a fixed distance @xmath8 in a straight line . on
a sphere the agent moves a fixed distance @xmath8 on a geodesic . here
we will consider two such agents who move different distances @xmath1 and @xmath2 at each time - step .
we assume that the agents initial positions are uniformly distributed in @xmath0 .
the paper s central results concern the probability that the paths of the two agents cross during the first time - step .
if @xmath0 is either a sufficiently large bounded region of the plane or a sphere , this `` first - step '' probability of intersection is close to @xmath6)$ ] where @xmath5 $ ] is the area of @xmath0 . in applied settings we are often interested in the _ long - run average rate _ at which the paths cross .
in order to extend results on the `` first - step '' probability of intersection and apply the law of large numbers we would need the following assumptions : * the positions of the two agents are uniformly distributed _ at every time step _ ( which is the case on the sphere but not on the plane because of reflection problems at the boundary of @xmath0 ) , * the crossing - path events are independent over time ( which is the case in neither setting because of the strong spatial dependence at consecutive time - steps ) .
numerous simulations have shown that despite marked departures from these assumptions the long - run rate at which the paths cross is also close to @xmath6)$ ] .
section [ mod ] contains the results both for the plane and the sphere .
section [ sim ] is devoted to the numerical simulations .
extensions are discussed in section [ disc ] .
three technical appendices can be found in section [ app ] . and @xmath9 with paths that crossed synchronously at @xmath10 for @xmath11 .
( we are not interested in the asynchronous crossing at @xmath12 which took place at different times for the two agents).,title="fig : " ] +
a bounded region of the plane is the most natural setting for agents moving in a 2d environment .
however we then need to specify how agents are reflected when they hit the boundary of the region .
there is no such problem on a sphere . in what follows @xmath0
is either a bounded region of the plane or a sphere .
the initial positions @xmath13 and @xmath14 of the two agents are assumed uniformly distributed in @xmath0 . at each time - step the two agents move distances @xmath1 and @xmath2 ( which are fixed positive parameters ) in a straight line ( or along a geodesic on a sphere ) .
they depart at random angles @xmath15 and @xmath16 that are uniformly and independently distributed over @xmath3 .
the endpoints after the @xmath17 time - step are @xmath18 and @xmath19 ( figure [ labelfig0 ] ) .
the definition of a meeting in such a model is tricky because the probability of the two agents being in exactly the same position at any given period is 0 .
there are however different ways of approximating such a meeting .
one could say that the agents meet if the distance between two points @xmath20 and @xmath9 is less than some @xmath21 .
in such a definition results would depend on @xmath21 , which is undesirable .
for this reason we choose to define a meeting during the @xmath17 time - step when paths cross between @xmath22 and @xmath23 .
this means that the segments ( or the `` geodesic arcs '' ) @xmath24 and @xmath25 intersect .
we recognize that @xmath20 and @xmath9 can then be close without the paths crossing , but at least the definition does not depend on some arbitrary @xmath21 .
the point @xmath10 in figure [ labelfig0 ] is an example of such a synchronous crossing of paths .
of course the agents are not at @xmath10 at the same time . in the figure the paths cross asynchronously at @xmath12 .
much will depend on whether @xmath20 and @xmath9 are uniformly distributed on @xmath0 for every @xmath22 .
this will be the case if @xmath0 is a sphere because @xmath13 and @xmath14 are themselves uniformly distributed .
if on the other hand @xmath0 is a bounded region of the plane , then the uniformity in the distributions of @xmath20 and @xmath9 is compromised for @xmath26 by the vexing problem of the behavior of the agents when they hit the boundary of @xmath0 .
for this reason we focus on the probability of paths crossing at the first time - step only .
we simplify notations by letting @xmath27 and @xmath28 be the initial positions of the two agents .
the corresponding endpoints are denoted @xmath29 and @xmath30 .
we next proceed with calculations when @xmath0 is a bounded region of the plane . in the plane the endpoints @xmath29 and @xmath30 are @xmath31 @xmath32 where @xmath15 and @xmath16 are polar angles that are uniformly distributed in @xmath3
. we will finesse the reflection problem at the boundary of @xmath0 by considering the possibility of intersection on the basis of eqs . - even if @xmath29 or @xmath30 is outside @xmath0 .
in such a case we examine first whether the intersection has occurred .
we then move , in some unspecified manner , the wayward point(s ) back inside @xmath0 .
we now define the _ feasible domain _ @xmath33 as the set of points @xmath34 that are within @xmath2 of the segment @xmath35 : @xmath36 where @xmath37 is the distance between a point @xmath34 and the segment @xmath35 . and @xmath38 with feasible domain @xmath39 .
inner and border regions @xmath40 and @xmath41 are at distances to the outside world that are larger than and less than the sum @xmath42 .
( the distance @xmath42 between the two boundaries is not to scale).,title="fig : " ] + the point @xmath43 must be in @xmath39 in order for the segments @xmath35 and @xmath38 to intersect - although the condition is not sufficient
. figure [ labelfig1 ] depicts a segment @xmath35 and the corresponding feasible domain @xmath39 .
this domain is bounded by dotted lines ( a rectangle with sides @xmath1 and @xmath44 with a half circle of radius @xmath2 at each end of the rectangle ) . in the figure
a point @xmath43 is in the feasible domain and the two segments @xmath35 and @xmath38 intersect . whether there is
an intersection depends on the angle @xmath16 .
the region @xmath0 is partitioned into an inner region @xmath40 and a border region @xmath41 characterized by a distance to the outside world that is either larger ( for @xmath40 ) or smaller ( for @xmath41 ) than @xmath42 .
( the point is that @xmath39 is entirely in @xmath0 if @xmath45 ) .
we define@xmath46 $ ] as the area of a bounded subset @xmath47 of @xmath48 . with @xmath49 and
@xmath50 uniformly and independently distributed on @xmath0 , we will now calculate the probability that @xmath35 and @xmath38 intersect .
this first - step probability of intersection depends only on @xmath51 and @xmath0 , and is noted @xmath52 .
( the subscript @xmath53 indicates a probability in the plane ) .
we first need the probability @xmath54 of the intersection conditionally on @xmath55 and @xmath56 ; @xmath54 is the probability ( denoted by @xmath57 below ) that @xmath43 falls in @xmath58 multiplied by the probability ( denoted by @xmath59 below ) of the segments intersecting conditionally on @xmath43 falling in @xmath58 .
( in the @xmath60 function we have replaced @xmath61 by its coordinates @xmath62 ) .
given the uniformity assumption on @xmath43 , the probability @xmath57 is then @xmath63}{a[r]}.\ ] ] in order to calculate @xmath59 we need to define the probability @xmath64 that @xmath65 intersects @xmath35 , conditionally on @xmath66 and the angle @xmath15 .
( this probability , derived in appendix a , is obtained by calculating the magnitude of the angle @xmath67 within which @xmath16 must fall for the intersection to occur , and then dividing by @xmath68 ) .
the conditional probability @xmath59 of the intersection is then @xmath69}.\ ] ] the probability @xmath70 is now the product @xmath71 which simplifies to @xmath72}.\ ] ] when @xmath55 is in @xmath40 ( i.e. the feasible domain is entirely in @xmath0 ) then the double integral on the right - hand side of eq .
is independent of @xmath61 and is noted @xmath73 , i.e. @xmath74 the quantity @xmath73 is an upper bound for the double integral in eq .
when @xmath55 is in the border region @xmath41 ( because the integration is then over an area smaller than the feasible domain @xmath75 ) .
we now have the following result on @xmath73 , @xmath54 , and @xmath52 . [ prop1 ]
we have @xmath76 and when @xmath55 is in @xmath40 , @xmath77}.\ ] ] when @xmath55 is in @xmath41 then @xmath78 } .\ ] ] the first - step probability of intersection @xmath52 satisfies @xmath79}{\pi a[r]^2 } \le p_p(d_1,d_2,r ) \le p_{hi } \overset{def . }
{ = } \dfrac{2d_1d_2}{\pi a[r]}\ ] ] which leads to the mid - point approximation @xmath80+a[r]}{\pi a[r]^2}.\ ] ] the absolute value of the maximum percentage error ( avmpe ) made with the approximation of ( [ p2200 ] ) is @xmath81-a[r_i]}{a[r]+a[r_i]}.\ ] ] see appendix a. _ remark .
_ the approximation of eq .
is of interest and the error of eq . is small only if @xmath5 $ ] is large enough in the sense that @xmath82 $ ] is _ relatively close _ to @xmath5 $ ] .
this means there is a large subset of @xmath0 within which the feasible domain @xmath83 is entirely in @xmath0 .
suppose for example that @xmath0 is a circle of radius @xmath84 , then @xmath85 which is approximately @xmath86 when @xmath42 is much smaller than @xmath87 . therefore if @xmath42 is one percent of the radius then the maximum error made with the estimate of eq .
is also approximately one percent .
we next turn our attention to the situation in which @xmath0 is a sphere .
when the domain @xmath0 is a sphere of radius @xmath88 we use the spherical system of coordinates where a point @xmath34 is defined by the triplet @xmath89 of radial , azimuthal , and zenithal coordinates .
given initial points @xmath90 each endpoint @xmath91 is on the circle at a geodesic distance @xmath92 from @xmath93 .
the position of @xmath91 on the circle is determined by an angle @xmath94 uniformly distributed in @xmath95 .
see appendix b for the exact derivation of the endpoints @xmath91 .
we let @xmath96 be the probability that the arcs @xmath35 and @xmath38 intersect .
because the initial points @xmath93 are uniformly distributed , the endpoints will also be uniformly distributed on the sphere .
therefore @xmath96 is the probability of paths crossing at every time - step . in order to calculate @xmath96 we
proceed as before except that there is no border area and the double integrals are calculated in spherical coordinates .
the feasible domain is denoted @xmath97 and is defined with geodesic distances .
the differential area element is now @xmath98 .
we let @xmath99 be the probability of intersection conditionally on the arc @xmath35 ( defined by @xmath100 and @xmath15 ) and on @xmath43 ( defined by @xmath101 and @xmath102 ) .
we also define the double integral @xmath103 the probability of the intersection conditionally on @xmath104 and @xmath15 is independent of @xmath105 and is denoted @xmath106 .
we then have @xmath107}\ ] ] where the area @xmath5 $ ] of the sphere appearing in the denominator is now @xmath108 .
the probability of paths crossing at each time - step is then @xmath109}\ ] ] @xmath110 ^ 2 } = = \dfrac { i_s(d_1 , d_2,\rho ) } { a[r ] } = \dfrac { i_s(d_1 , d_2,\rho ) } { 4\pi\rho^2 } .\ ] ] when @xmath111 the integral @xmath112 on the sphere approaches the corresponding integral @xmath113 on the plane ( eq . ) . in the next proposition
we show numerically that @xmath112 is extremely close to @xmath114 even when @xmath88 is not particularly large compared to @xmath1 and @xmath2 .
we will just assume that @xmath115 which insures that the feasible domain does not `` wrap around '' the sphere .
[ prop2 ] with ( [ g10uz ] ) we have @xmath116 and the probability of paths crossing at each time - step is @xmath117 } = \dfrac{d_1d_2}{2\pi^2\rho^2}.\ ] ] the differences between both sides of ( [ g10tt ] ) and ( [ ggb ] ) are so small that they are within the margins of error when calculating @xmath118 ( and @xmath119 ) numerically .
see appendix c. the next section is devoted to simulations aimed at assessing the quality of the approximations derived above .
we consider a region @xmath0 that is a circle of radius @xmath120 . at each time - step the two agents move distances @xmath121 and @xmath122 respectively . if an agent moves outside the circle we first check whether paths have crossed .
we then move the wayward agent to a point diametrically opposed to its current position , at a distance inside the circle equal to the distance between the circle and the current position .
this algorithm could be considered a 2d version of the wrapping around that takes place on a sphere .
the goal is to try to keep the distribution of the agents as uniform as possible .
we need this in order for the crossing - path probability to remain as close as possible to the `` first - step '' probability derived under the assumption that initial positions are uniformly distributed .
the bounds of and the approximation of are now used to calculate the low , high and mid - point approximations @xmath123 for the first - step probability of intersection @xmath124 : @xmath125}{\pi a[r]^2}= 0.0009772 , \bigskip p_{hi } = \dfrac{2d_1d_2}{\pi a[r]}=0.001418,\ ] ] @xmath126 + a[r_b]}{\pi a[r]^2 } = 0.001198\ ] ] which translate into a maximum error @xmath127 on @xmath128 of 18.42% ( eq . ) .
this error is relatively large because the sum @xmath42 is 1.7 , which is not particularly small compared to the radius 10 of the circle .
we let @xmath129 be the random variable equal to the average crossing path frequency over the first @xmath22 time - steps .
_ if _ the probability of paths crossing at every time - step were @xmath130 ( or @xmath131 ) and _ if _ paths crossing were independent events ( which they are not ) then the central limit theorem would insure that for large @xmath22 the corresponding `` hypothetical frequency '' @xmath132 would be approximately normally distributed with mean @xmath130 ( or @xmath131 ) and standard deviation @xmath133 ( or @xmath134 ) .
figure [ figsim ] depicts a simulated trajectory of the frequency @xmath129 .
we also plotted the hypothetical low and high intervals within which each @xmath132 would fall with probability @xmath135 .
these bands shed light on the expected fluctuations of the frequency , in the case of independent events taking place with probability @xmath130 ( or @xmath131 ) .
( over the first @xmath22 time - steps ) for two random walks in a circle of radius @xmath136 with @xmath121 and @xmath122 ( 15,000 time - steps ) . at each time
- step the intervals within which the `` hypothetical frequency '' @xmath132 falls with probability @xmath135 are given for the low and high approximations @xmath130 and @xmath131 of @xmath137 .
the mid - point approximation @xmath138 is also plotted . ]
this and other simulations suggest that @xmath128 gives at least an idea of the ( long - run ) crossing - path probability . there is less uncertainty on the sphere as we have shown that the one - step probability of intersection @xmath139 is extremely close to @xmath140 .
simulations performed on the sphere ( not shown ) yield long - run crossing - path rates that are close to @xmath140 even though the law of large numbers can not be invoked because of spatial dependence over time .
the expression of eq . for
the crossing - path probability in the plane can be relatively crude .
however it has the merit of simplicity and it improves if the area of the region @xmath0 increases . on the sphere
the crossing - path probability can be approximated very closely by the simple expression @xmath140 .
we derived this expression on the basis of an analytical result for the plane ( appendix a ) , expecting it to be a good approximation only for a sphere of infinitely large radius .
it is of some interest to note that because of the complexity of the multiple integrals in spherical coordinates we saw now way of obtaining this approximation from the calculations performed directly on the sphere ( appendix c ) . because of
marked departures from required assumptions the law of large numbers could be applied neither in the plane nor on the sphere . despite that
, the long - run average crossing - path probabilities appear to be close to the first - step probabilities .
this suggests that a weaker version of the law of large numbers may be applicable .
for example results on `` weakly dependent '' random variables ( i.e. variables that are `` m '' ( or `` @xmath141'')-dependent ( @xcite ) ) may provide more insights into the long - run behavior of the system . we note some obvious and some less obvious extensions that can be of use to population biologists ( and perhaps others ) : * if the crossing of paths takes place between an infected and a susceptible agent , then the transmission of the infection may occur with only a probability @xmath142 .
in such a case the crossing - path probabilities found here need simply be multiplied by @xmath142 in order to obtain the probability of transmission at each time - step .
* an important extension would have @xmath143 such infectives and @xmath47 such susceptibles .
epidemiologists would be keenly interested in _ analytical results _ on the rate at which the infection would then spread . *
the assumption that an agent moves at an angle that is uniformly distributed in ( 0,2@xmath144 ) may not be realistic .
for example animals may move only within a limited angle in the continuation of the previous direction .
preliminary investigations suggest that the results obtained here may still be applicable .
the results given in this paper are merely starting points for more in - depth theoretical investigations .
they also provide practitioners with some answers concerning the dynamics of a process that depends on randomly moving agents meeting in a spatially explicit environment .
the double integral @xmath146 in eq .
is calculated in the orthonormal coordinate system @xmath147 for which the segment @xmath35 lies on the @xmath145-axis and the origin is at the middle of the segment ( figure [ labelfig2 ] ) . in the new coordinate system , the probability @xmath64 that @xmath35 and @xmath38 intersect depends only on the components @xmath148 of @xmath43 .
if this probability is denoted @xmath149 , then @xmath146 is the probability @xmath149 is obtained by calculating the magnitude of the angle @xmath67 within which @xmath38 ( defined by the angle @xmath16 ) must fall and then dividing by @xmath68 ( see figure [ labelfig2 ] ) . in the figure
the point @xmath43 is in neither circle and the angle @xmath67 is the vertex angle in the isosceles triangle with apex @xmath43 and two sides of length @xmath2 .
therefore , when @xmath43 is in neither circle the probability @xmath149 of an intersection is given by the function @xmath151 similar geometric considerations show that if the two circles overlap ( i.e. @xmath152 ) then for a point @xmath153 in the intersection of the two circles , the probability of intersection @xmath154 is given by the function @xmath155 finally , if @xmath153 is in only one of the circles , then one of the vertices of the triangle with apex @xmath43 will be the center @xmath61 or @xmath29 of that circle .
the probability of intersection @xmath154 is then @xmath156 long but elementary calculations show that the innermost integral in , considered a function @xmath157 of @xmath158 , is now equal to @xmath159 we therefore have @xmath160 which is eq .
( [ g19 ] ) and yields eq .
( [ ggc ] ) . with @xmath61 and @xmath43 uniformly distributed on @xmath0 ,
the probability of intersection @xmath52 is now obtained by integrating @xmath161 over @xmath162 in @xmath163 and over @xmath164 in @xmath3 , and then dividing by @xmath165 $ ] : the first term is easily calculated while the second one is a complicated triple integral of a double integral that can only be calculated numerically .
the inequality in ( [ gga ] ) implies however that @xmath170}{\pi a[r]^2 } \le p_p(d_1,d_2,r ) \le \dfrac{2d_1d_2(a[r_i ] + a[r_b])}{\pi a[r]^2 } = \dfrac{2d_1d_2}{\pi a[r]}.\ ] ] which is ( [ p2100 ] ) .
the approximation of ( [ p2200 ] ) is simply the mid - point of the interval in ( [ p21 ] ) . the error term of eq .
( [ p24 ] ) is based on the low and high bounds of ( [ p21 ] ) . in order to find each endpoint @xmath91 ( @xmath172 ) we will first determine a particular point @xmath173 at a geodesic distance @xmath92 from @xmath93 .
each @xmath91 is then obtained by rotating @xmath173 by a uniformly distributed angle @xmath94 about the @xmath174 axis .
each @xmath173 is defined as having the same azimuthal coordinate @xmath175 as @xmath176 and a zenithal coordinate @xmath177 that puts @xmath173 at a geodesic distance @xmath92 from @xmath93 . to calculate @xmath173
we define the function we let @xmath183 denote the @xmath184 identity matrix .
we will also need the @xmath185 and @xmath186 functions that transform spherical coordinates into cartesian ones , and vice - versa , i.e. @xmath187 and we will derive an expression for @xmath112 in the spherical coordinate system in which the arc @xmath35 lies on the equator and its middle is at the point @xmath192 , i.e. @xmath193 the feasible domain now consists of two areas on the surface of the sphere .
first the `` spherical rectangle '' centered on @xmath192 with sides of geodesic lengths @xmath1 and @xmath44 ; and second at both ends of the rectangle the half - circles centered at @xmath61 and @xmath194 and of ( geodesic ) radius @xmath2 ( figure [ figde ] ) .
the feasible domain now depends only on @xmath51 and is denoted @xmath195 .
within which @xmath38 must fall for the arcs @xmath35 and @xmath38 to intersect .
( when @xmath43 is in the right circle ( resp . the left circle ) , @xmath196 ( resp .
@xmath197 ) will be at @xmath29 ( resp .
@xmath61 ) . ] * the geodesic distance function between two points @xmath199 and @xmath200 on the sphere : @xmath201 @xmath202.\ ] ] * if @xmath203 and @xmath204 are two vectors ( in cartesian coordinates ) on the sphere of radius @xmath88 , then the normed tangent vector at @xmath203 to the geodesic line between @xmath203 and @xmath204 : @xmath205 * the @xmath185 function that transforms spherical coordinates into cartesian ones ( eq . ( [ c24 ] ) ) .
given @xmath206 in the feasible domain @xmath195 , we let @xmath196 and @xmath197 be the two points on the equator ( on the right and on the left of @xmath43 ) that determine the magnitude of the angle @xmath67 within which @xmath38 must fall for the intersection to occur ( figure [ figde ] ) . in the new coordinate system ,
the probability of intersection @xmath99 conditionally on the arc @xmath35 and on @xmath43 depends only on the components @xmath212 of @xmath43 and is denoted @xmath213 .
given the uniformity assumption for the angle @xmath16 at which the second agent leaves @xmath43 to go to @xmath30 , the probability is equal to the magnitude of the angle @xmath67 within which the arc @xmath38 must fall for the intersection to occur , divided by @xmath214 .
the angle @xmath67 is the angle between the tangent vectors at @xmath43 in the directions of @xmath215 and @xmath216 .
the probability of intersection @xmath213 is therefore the double integral @xmath118 is now calculated in the new coordinate system by integrating @xmath213 over the feasible domain @xmath195 . because of symmetries the integral @xmath118 is the sum of four times the integral over the upper right quarter of the rectangular area of @xmath195 and of four times the integral over the upper half of the right circle .
we thus have @xmath220 over a wide range of values for @xmath51 and @xmath88 we found that the relative error made when approximating @xmath118 of eq .
( [ z400 ] ) as @xmath114 was of the order of @xmath223 to @xmath224 _
percent_. see table [ t1 ] for an example of this relative error for a range of values of @xmath2 and @xmath88 .
in fact the error is so small that it is within the margin of error when calculating @xmath118 numerically .
the fact that @xmath225 combined with eq .
( [ g10u ] ) yields the result of ( [ ggb ] ) . | we investigate crossing path probabilities for two agents that move randomly in a bounded region of the plane or on a sphere ( denoted @xmath0 ) . at each discrete time - step
the agents move , independently , fixed distances @xmath1 and @xmath2 at angles that are uniformly distributed in @xmath3 .
if @xmath0 is large enough and the initial positions of the agents are uniformly distributed in @xmath0 , then the probability of paths crossing at the first time - step is close to @xmath4)$ ] , where @xmath5 $ ] is the area of @xmath0 .
simulations suggest that the long - run rate at which paths cross is also close to @xmath6)$ ] ( despite marked departures from uniformity and independence conditions needed for such a conclusion ) . |
i - v characteristics of the high t@xmath0 superconductor bi@xmath1sr@xmath1ca@xmath2c@xmath1o@xmath3 ( bscco ) shows a strong hysteresis , producing multi - branches @xcite .
many experiments have been reported to investigate properties of hysteresis branches @xcite , and it has turned out that there is rich physics in properties of i - v characteristics .
substructures are found in hysteresis branches , and are discussed in relation with phonon effects @xcite .
the d - wave effect is also discussed in the higher voltage region , where the tunneling current becomes important @xcite .
it is well known in a single josephson junction that the application of a constant voltage - bias leads a rather simple phenomenon of the harmonic current - oscillation with the frequency proportional to the applied voltage @xcite .
this is due to the fact that the time - derivative of the phase difference is directly connected with the voltage , and that there is no room for the phase to move freely .
therefore , the most of the above mentioned experiments were performed with the current biased arrangement , aiming to investigate properties of i - v hysteresis branches , and , as long as we know , no systematic experiment with the voltage bias has been reported .
we will point out in this paper that anomalous behaviors are expected , theoretically , in the low voltage - biased region in the case of the multi - layered intrinsic josephson junctions of high t@xmath0 superconductors , and that the voltage - biased experiment gives additional information on behaviors of i - v characteristics . in the multi - layered intrinsic josephson junctions of high t@xmath0 superconductors , inter - layer couplings among superconducting phase differences become important due to the atomic scale of the thickness of the superconducting layer . in the ref .
@xcite the mechanism of the inter - layer coupling is proposed as the charging effect of superconducting layers ; the charging effect induces the variation of electric field and mediates the inter - layer coupling among superconducting phase differences .
it has been shown that , by this mechanism , a branch structure in the i - v characteristics is obtained as the intrinsic nature of the strong anisotropy of high t@xmath0 superconductors @xcite . in the previous paper @xcite
, we have investigated the origin of hysteresis branches , and have shown that hysteresis jumps are induced when solutions change in nonlinear - coupled equations of phase - differences .
namely , the number of rotating phases and its patterns classifies solutions . at a point of a hysteresis jump
, its number or distribution pattern is changed . the main difference between the cases of the single junction and the multi - junction lies in the fact that the applied voltage gives restriction only on the total voltage .
therefore , each junction may behave in a self - adjusted way , if an interaction effect allows it . in a simple array of @xmath4 single junctions ,
each junction feels only the @xmath4-th of the applied voltage , and one can not expect much difference from the case of the single junction .
however , in the intrinsic josephson array of the high t@xmath0 superconductors , the existence of an inter - layer coupling affects the collective motion of phase differences .
furthermore , the voltage control and current control lead different situations .
specially , in the voltage control case , one can reaches transient unstable states of the current biased case , since the applied voltage still enforces the total behavior of phase differences . in this paper
, we investigate i - v characteristics in voltage - biased cases of high t@xmath0 superconductors . since there is no experiment available to refer at present ,
several cases with possible surface conditions will be presented . in the next section ,
we summarize the formula for the voltage - biased case in the multi - josephson junction model . in sect .
3 , we present numerically simulated results , and present i - v characteristics of the voltage - biased case . it will be shown that certain anomalous behavior of i - v characteristics appears in the applied low voltage region .
we show that the voltage region is divided roughly into three regions , an anomalous periodic region , a chaotic region and a region of hysteresis branches . in an anomalous periodic region ,
the current shows a trigonometric functional increase and rapid drop . in a chaotic region , the total current behaves chaotically . in the last region ,
the behavior of the current is quasi - harmonic .
physical origin of these regions will be discussed .
is devoted to concluding remarks .
in this section , we rewrite the formula for the multi - josephson junction of ref .
@xcite suitable to the voltage - biased analysis .
the notations used here are same as those in ref .
namely , we consider the @xmath5 superconducting layers , numbered from @xmath6 to @xmath4 . we denote the gauge invariant phase difference of the @xmath7-th and @xmath8-th superconducting layer by @xmath9 , and voltage by @xmath10 .
the widths of the insulating and superconducting layers are denoted by @xmath11 and @xmath12 , respectively . at the edges , the effective width of the superconducting layer
may be extended due to the proximity effect into attached lead metals .
the widths of the @xmath6-th and @xmath4-th superconducting layers are denoted by @xmath13 and @xmath14 , respectively . in the voltage - biased case , the total voltage is equal to the applied external voltage @xmath15 , @xmath16 in order to perform a numerical simulation under the condition ( [ vcond ] ) , we rewrite the equation for the total current @xmath17 ( eq .
( 1 ) in ref .
@xcite ) and the equation relating the time - derivative of phase - difference and voltage ( eq . ( 3 ) in ref .
@xcite ) in the following way .
first , we note that the total current @xmath17 is expressed by the external voltage @xmath15 as @xmath18 the current @xmath17 is normalized by the critical current @xmath19 , @xmath20 with @xmath21 being the critical current for the @xmath8-th junction .
the time @xmath22 is normalized by the inverse of the plasma frequency , @xmath23 = @xmath24 , with @xmath25 being dielectric constant of the insulating layer .
the voltage @xmath10 is normalized by @xmath26 .
the parameter @xmath27 is given by @xmath28 .
we define the voltage difference @xmath29 as @xmath30 then we have the following coupled differential equations , @xmath31 a dissipation effect@xcitein superconducting layers will be neglected in this paper , and the matrix @xmath32 is given in eq . ( 4 ) in ref .
it should be noted that eq .
( [ dtv ] ) gives @xmath33 therefore if @xmath34 is satisfied at the initial time , it is satisfied in all time . in the next section , we present results of solving eqs .
( [ dtv ] ) and ( [ dtp ] ) by numerical simulation .
the restriction from the total voltage is achieved by putting the initial condition eq .
( [ cond ] ) . the equation for @xmath9 , eq .
( [ dtp ] ) , indicates that the phase differences move according to @xmath35 as average .
however , due to the mutual interaction , each @xmath9 shows a complicated behavior which affects the i - v characteristics and the behavior of the total current .
we have solved the coupled differential equations ( [ dtv ] ) and ( [ dtp ] ) by use of the fourth order runge - kutta method .
the average current @xmath17 is obtained by @xmath36 we choose the parameters as @xmath37 , @xmath38 and @xmath39 , which has been used in ref . @xcite . in the simulation
we have chosen the time step @xmath40 , @xmath41 , @xmath42 and the voltage step @xmath43 . as eq .
( [ dtp ] ) shows , the phase @xmath9 increases according to @xmath44 when @xmath15 is small .
when @xmath9 reaches the value @xmath45 , the phase goes into a rotating mode as can be seen from the analogy of the motion of a pendulum with a @xmath46-nonlinear term .
this indicates that the whole motions of phases @xmath9 are much affected by the phase with the lowest @xmath47
. it may be difficult to arrange two electrode at edges exactly in the same condition .
however , in order to show how surface conditions affect the behavior of i - v characteristics , we present the following three cases ; 1)@xmath21 are symmetric but widths of surface are asymmetric , 2 ) both @xmath21 and widths of surface symmetric , and 3)@xmath21 asymmetric and widths of surface symmetric .
the case with @xmath21 asymmetric and widths of surface asymmetric shows similar behaviors as the case 3 ) . it should be noted that , even if the asymmetry between @xmath13 and @xmath14 is introduced , the observed critical current is @xmath19 in the current - biased case .
we first assume that all @xmath48 are equal , @xmath49 and choose , for example , an asymmetric boundary condition @xmath50 and @xmath51 . in the present analysis ,
it is enough to have slight difference between @xmath13 and @xmath14 , the difference of proximity effect . in fig .
1(a ) , we show the overall behavior of i - v characteristics obtained from the adiabatic increase and decrease of the applied voltage ( circle points ) and those obtained from abrupt application of the voltage ( @xmath52 points ) .
we can see that the i - v characteristics show many hysteresis jumps , forming a branch structure . in the abrupt application of voltage ,
the data points in the low voltage region are scattered , although they lie on some linear i - v branches .
each i - v branch is classified by a number and pattern of rotating phases .
when the voltage is large enough , all junctions have the rotating phase and the single i - v characteristic is obtained . in fig .
1(a ) , we notice that there exists the region where the adiabatic and abrupt applications of voltage give the same results , in the low voltage region before the branch structure starts . in fig .
1(b ) , we expand the low voltage region
. we can identify the obvious branches labeled as i@xmath2 , i@xmath1 , ii , iii@xmath2 and iii@xmath1 .
the crosses ( @xmath52 ) are for the abrupt voltage application .
the lines i and ii are same for three cases of the adiabatic increase , decrease and abrupt application of the voltage .
the branches iii@xmath2 and iii@xmath1 are linear .
the calculation shows that the i - v relations for iii@xmath2 and iii@xmath1 are same as those obtained from the hysteresis branches in the i - v characteristics of the current - biased case . on the line iii@xmath2 , the phase @xmath53 is rotating and on the line iii@xmath1 the phases @xmath53 and @xmath54 are rotating .
other phases @xmath9 are oscillating .
the cross points are grouped as one rotating - phase and two - rotating phase .
the scattered points are identified by the distribution of the rotating - phases among @xmath4 junctions . in fig .
2 , we show the i - v characteristics , when we choose a symmetric boundary condition @xmath50 and @xmath55 .
the branch ii in fig.1 is separated into ii@xmath2 and ii@xmath1 .
the branch iii@xmath1 is linear and is exactly same line as iii@xmath1 in fig .
1 . although the branch ii@xmath1 is linear as average and lies on the same line as iii@xmath2 , the calculated points are largely scattered , indicating a certain kind of instability . in order to obtain fig .
2 , we have to ensure the symmetry , @xmath56 in each time step by taking the average of @xmath9 and @xmath57 , for example .
otherwise , a very small round error rapidly develops in the region ii , and we obtained the similar result of fig .
1(b ) , although we start with the symmetric boundary condition .
in order to see what kinds of solutions are realized in the regions i and ii , we plot , in fig .
3 , the time - dependence of the phase @xmath9 and the current @xmath17 for the case of fig .
3(a),(b ) are for the branches i@xmath2 , and fig .
3(c),(d ) are for ii@xmath2 .
3(b ) shows periodic current behaviors , while fig .
3(d ) shows chaotic current behaviors .
the results show the following .
all @xmath9 increase slowly in the same way according to the applied voltage .
when @xmath53 and @xmath54 at the edges reach the value @xmath45 , they increase rapidly in the phase - rotating mode . other @xmath9 s swing back . the phases
@xmath53 and @xmath54 damp their motion by the effect of resistance @xmath27 , and move slowly again in the same way as other phases . as the result
, the current behaves periodically with trigonometric functional increase and rapid drop .
the regions i@xmath2 and i@xmath1 are different in the values of phase rotation , @xmath58 and @xmath59 , before damping . in fig .
3(c ) , not only the phases at the edges but also other phases @xmath9 go into the rotating mode . since oscillation and rotation of phases are interface in a complicated way , the chaotic current behavior is produced .
the regions ii@xmath2 and ii@xmath1 are different in the values of phase rotation , @xmath58 and @xmath59 , at the edges .
the above results lead us the following understanding of solutions of the multi - josephson junction in the voltage - biased case .
the applied voltage @xmath15 forces the phase - differences @xmath9 of junctions to rotate .
when some @xmath9 reaches @xmath45 , it goes into the natural rotating mode with other phases swinging back .
how much it rotates depends on its speed and resistance , forming new non - ohmic branches .
when the rotation and oscillation of phase differences start to interact , solutions go into a chaotic region .
when the voltage becomes large enough for certain phase difference to keep rotating , the i - v characteristics become ohmic and form multi - branches . from the above understanding
, we can immediately see that the i - v characteristics may change when there exists a certain junction with weaker @xmath47 .
most likely , the surface layers may be damaged in the preparation process of electrodes .
in fact in the previous analysis @xcite , we have to choose reduced @xmath19 values at the edges in order to fit the data .
it was also reported that for thin film mesa structures the preparation process for the upper electrode can produce a cut of the first junctions , leading to an area about half size and thus half critical current for those junctions @xcite . as an example , we choose @xmath60 the result is presented in fig .
the chaotic region ii is missing . in the inset fig .
4(b ) , we show the enlarged figure of the region i. there are some drops and quasi - periodic structure in the i - v characteristics . each part has a different value of @xmath9-phase rotation , @xmath61 .
we note that , in ref .
@xcite , the appearance of certain periodicity in i - v characteristics has been reported , though it is for the sample of submicron junctions with no branch structures .
it is a future problem to see if there is some relation with the present analysis .
we also point out that , in the analysis of fig .
1 , the regions i - ii are obtained also when the external voltage is applied abruptly .
adiabatic and abrupt voltage application differ in the initial total energy .
therefore , the present result shows that the regions i - ii have the single stationary state .
in this paper , we have investigated the i - v characteristics of the voltage - biased case . differently from the case of the single junction ,
i - v characteristics shows the characteristic behavior originated from the mutual inter - layer coupling mediated by the charging effect .
although each junction feels as average the external voltage as @xmath35 , there is a still room for each phase to move self - consistently according to the interaction .
therefore , when the voltage is low , there is a competition of the voltage forced motion and the motion induced by the interactions .
adiabatic increase of the external voltage induces the following successive transition . in the low voltage ,
one has the region with a periodic oscillatory current of an asymmetric waveform , and then a possible region with a chaotic current oscillation follows . in these regions ,
non - ohmic branches are formed . in higher voltage
, the region with a stable harmonic oscillatory current appears , having ohmic i - v branches . when there is a junction with a weaker @xmath47 , the associated phase difference is easier to go into the rotating mode
. then the i - v characteristics have a quasi - periodic behavior in its non - ohmic region .
the present analysis shows that voltage biased i - v characteristics reveals the importance of inter - layer coupling more strongly , and gives additional information to the current biased case .
this work was supported by the special research grant in the faculty of engineering , seikei university ( h. m. and s. s. ) , and grant - in - aid for scientific research ( c ) from japan society for the promotion of science ( t. k. ) .
5 r. kleiner , f. steimmeyer , g. kunkel and p. m " uller , phys .
68 , 2394 ( 1992 ) .
r. kleiner and p. m " uller , phys . rev .
b 49 , 1327 ( 1994 ) .
r. kleiner , p. m " uller , h. kohlstedt , n. f. pedersen , and s. sakai , phys .
b 50 , 3942 ( 1994 ) .
s.sakai , p. bodin , and n. f. pederson , j. appl .
73 , 24 ( 1993 ) .
a. yurgens , d. winkler , n. v. zavaritsky , and t. claeson , phys . rev .
b 53 , r8887 , ( 1996 ) .
k. tanabe , y. hidaka , s. karimoto , and m. suzuki , phys .
b 53 , 9348 ( 1996 ) .
m. itoh , s. karimoto , k. namaekawa , and m. suzuki , phys .
b 55 , r12001 , ( 1997 ) . c. helm , ch .
preis , f. forsthofer , j. keller , k. schlenga , r. kleiner , and p. mller , phys .
79 , 737 ( 1997 ) .
m. sakai , a. odagawa , h. adachi , and k. setsune , physica c 299 , 31 ( 1998 ) .
k. schlenga , r. kleiner , g. hechtfisher , m. mle , s. schimit , p. mller , ch .
helm , , ch .
preis , f. forsthofer , j. keller , h. l. johnson , m. veith , and e. steinbei , phys . rev .
b 57 , 14518 ( 1998 ) .
m. tinkam,_introduction to superconductivity_,(mcgraw - hill , inc .
new york , 1996 ) .
t. koyama and m. tachiki , phys . rev .
b 54 , 16183 ( 1996 ) .
m. machida , t. koyama and m. tachiki , physica c 300 , 55 ( 1998 ) .
h. matsumoto , s. sakamoto , f. wajima , t. koyama and m. machida , phys .
b 60 no.5 ( to be published , 1999 ) .
d. a. ryndyk , pisma zh .
65 , 755 ( 1997 ) [ jetp lett .
65 , 791 ( 1997 ) ] .
d. a. ryndyk , phys .
letters 80 , 3376 ( 1998 ) .
p. seidel , a. pfuch , f. schmidl , j. scherbel and u. hbner , czech .
46 , suppl .
s3 , 1263 ( 1996 ) . yu .
i. latyshev , s. -j .
kim and t. yamashita , pisma zh .
eksp . teor
69 , 75 ( 1999 ) [ jetp letters b4 , 84 ( 1999 ) ] .
i - v characteristics with an asymmetric boundary condition + ( a ) for the wide range of @xmath62 and ( b ) for the expanded low voltage region .
the circles are for the adiabatic voltage change .
the crosses are for the abrupt voltage application .
parameters are @xmath38 , @xmath39 , @xmath37 , @xmath63 and @xmath64 . | by use of the multi - josephson junction model , we investigate voltage - biased i - v characteristics . differently from the case of the single junction ,
i - v characteristics show a complicated behavior due to inter - layer couplings among superconducting phase differences mediated by the charging effect .
we show that there exist three characteristic regions , which are identified by jumps and cusps in the i - v curve . in the low voltage region ,
the total current is periodic with trigonometric functional increases and rapid drops .
then a kind of chaotic region is followed . above certain voltage ,
the total current behaves with a simple harmonic oscillation and the i - v characteristics form a multi - branch structure as in the current - biased case .
the above behavior is the result of the inter - layer coupling , and may be used to confirm the inter - layer coupling mechanism of the formation of hysteresis branches . |
magnetars are neutron stars with super - strong magnetic fields .
the typical magnetic strength of magnetars is of the order of @xmath6 , which is higher than the critical magnetic field @xmath7 , and is @xmath8 times stronger than that of normal pulsars .
historically , they were discovered in two manifestations , i.e. anomalous x - ray pulsars ( axps ) and soft gamma - ray repeaters ( sgrs ) , even though later observations suggested that there seems no clear distinction between the two sub - categories .
sgrs are discovered in the hard x - ray / soft gamma - ray band characterized by repeating bursts .
eleven sgrs have been confirmed so far @xcite .
three of them each displayed one giant flare ( sgr 0526 - 66 on march 5 , 1979 ; sgr 1900 + 14 on august 27 , 1998 ; sgr 1806 - 20 on december 27 , 2004 ) .
these giant flares are characterized by an initial short - hard spike followed by a pulsating tail @xcite .
the initial spike has a duration of @xmath9 , and an isotropic energy of @xmath10 .
the pulsating tail , on the other hand , shows a decay pattern and is characterized by a typical duration of @xmath11 , an isotropic energy of @xmath1 , and a pulsating period of a few seconds ( the magnetar spin period ) .
the spectrum of the early spike is characterized by a temperature of @xmath12 , which is much harder than that of normal sgr bursts .
the spectrum of the pulsating tail is softer than that of the initial spike , which may be fitted with the superposition of a thermal component with @xmath13 and a non - thermal power law component .
the decay light curves of the pulsating tail typically show time - evolving , complex pulse profiles .
as pointed out by @xcite , the tail emission likely originates from the residual energy of the explosion that gives rise to the initial spike . owing to the large energy density and strong magnetic field pressure in the magnetar magnetosphere
, the residual energy is confined in the magnetosphere in the form of an optically thick photon - pair plasma named a `` trapped fireball '' .
such a trapped fireball would be the source of the thermal emission component observed in the pulsating tail .
the power law spectral component may be formed from the jets associated with the open - field lines that are outside the trapped fireballs @xcite .
the structure of the trapped fireball may determine the pulse profile of the pulsating tail .
in fact , the magnetars show more complicated pulse structure profiles during the giant flare decaying phase than during the quiescent phase , the latter of which typically shows a single , broad - peak , nearly sinusoidal pulse shape .
for example , the decay phase of the giant flare of sgr 0526 - 66 on march 5 , 1979 showed a double - peak pulse profile @xcite ; the giant flare of sgr 1900 + 14 on august 27 , 1998 showed a complex four - peak profile @xcite ; and the giant flare of sgr 1806 - 20 on december 27 , 2004 showed a roughly three - peak profile @xcite . on the other hand
, the temperature of the phase - resolved spectra during the decay phase shows a different evolution pattern from the light curve : for sgr 1900 + 14 , the phase - resolved temperature shows two peaks in each period , with the epoch of the highest thermal flux corresponding to the lowest count rate in the light curve @xcite ; for sgr 1806 - 20 , on the other hand , the evolution of temperature seems to track the photon count in the light curve @xcite .
therefore , the true structure of the trapped fireball is likely very complex .
the effective blackbody radius inferred from the data suggests that the radius of the trapped fireball is a few times of the magnetar radius @xcite , which suggests that the trapped fireball should be in the magnetosphere rather than being a hot spot on the magnetar surface .
a natural idea is that the trapped fireball is a closed - field region in the magnetosphere ( and the non - thermal emission comes from the open field line regions ) .
the multi - peak profile of the light curves might be the result of the inhomogeneous distribution of the plasmas within the trapped fireball and the non - thermal jets within the magnetosphere . due to the complexity of the problem itself , it is not easy to fully understand the mechanism of the giant flares
. nonetheless , a simple model that assumes that the trapped fireball is from a closed field region in the magnetosphere may catch the essence of the problem .
polarization is a key element to test theoretical models of the physical processes and geometric structure of magnetars .
unfortunately , none of the three giant flares were detected with polarimeters , so that their polarization properties remain a mystery .
thanks to technological development and funding support , a few x - ray/@xmath14-ray polarimeters are imminent , which would make ground - breaking discoveries to study the underlying physics in many astrophysical objects , such as isolated neutron stars , magnetars , x - ray pulsars , pulsar wind nebulae , rotation - powered pulsars , and millisecond x - ray pulsars @xcite .
for example , in the conventional x - ray band ( @xmath15 ) , polarimeters that are under construction include imaging x - ray polarimetry explorer ( ixpe , @xcite ) , polarimeter for relativistic astrophysical x - ray sources ( praxys , @xcite ) .
for hard x - rays and soft-@xmath14-rays ( above @xmath16 ) , several detector concepts have been proposed .
for example , there has been an extensive study of the poet ( polarimeters for energetic transients ) mission concept @xcite as well as its extension leap ( large area polarimeter ) . on the other hand , the polar detector @xcite
, a compton polarimeter designed to measure linear polarization of gamma - rays from gamma ray bursts ( grbs ) , is scheduled to be launched with the chinese space station tiangong-2 in 2016 .
polar has a large effective area of @xmath17 and a field - of - view covering about one third of the sky .
its energy band covers from @xmath18 to @xmath19 , which can cover part of the energy range of the pulsating tail of sgr giant flares . for a total flux larger than @xmath20
, polar can perform a measure of polarization degree with a relative systematic error @xmath21 @xcite .
it would be an ideal detector for giant flare polarization studies , should such a giant flare occurs during its mission period . among the available detectors , the reuven ramaty high energy solar spectroscopic imager ( rhessi ) is an instrument designed to study the brightest solar flares , which has the polarization detecting capability @xcite .
interestingly , the giant flare of sgr 1806 - 20 on december 27 , 2004 was detected with rhessi @xcite .
however , no polarization study of the pulsating tail emission of this event is published .
many authors have calculated the polarization properties from neutron stars .
@xcite calculated the polarization of thermal x - rays from isolated neutron stars with strong magnetic fields .
@xcite studied radiation spectra from the atmosphere of neutron stars and the effect of vacuum polarization . @xcite and @xcite suggested that the birefringence of the magnetized qed vacuum decouples the polarization modes up to a large radius , so that the surface polarization of neutron stars can be sufficiently large to be observed . @xcite and
@xcite calculated the soft x - ray polarization in thermal magnetar emission from a hot spot centered around the magnetic dipole axis .
@xcite studied the x - ray polarization signature of the magnetosphere of quiescent magnetars .
@xcite studied how future x - ray polarization measurements by the next generation detectors may be used to probe the magnetar magnetospheric structure . in this paper , we study the atmospheric structure and polarization properties of trapped fireballs during the decay phase of giant flares . since the radiation mechanism of the non - thermal component is still uncertain , we mainly consider the thermal emission component from the trapped fireball .
observationally , the thermal radiation component seems to dominate the observed emission during most phases in the decaying tail , and the flux level of the thermal component is higher than that of the power - law component between @xmath16 and @xmath22 @xcite . the paper is organized as follows . in sect .
2 , we introduce a simple model of the atmospheric structure of the trapped fireball . in sect .
3 , the photon polarization properties are calculated by considering dielectric tensor of the electron - positron pair plasma as well as the effect of vacuum polarization , and the scattering opacities for the two photon modes are presented . in sect .
we calculate the polarization degree of the trapped fireball using monte carlo simulations .
the results are summarized in sect .
5 with some discussion .
in this section , we calculate the atmospheric structure of the trapped fireball in a simplified model . we assume a dipolar configuration for the magnetar magnetic fields , and take the region enclosed by a closed field line ( not necessarily the last closed field line ) as the trapped fireball ( fig.1 ) .
thanks to the extreme large optical depth of the trapped fireball , we only need to solve the equations in a relatively thin layer ( the bisque layer shown in fig.1 ) , which corresponds to the radius range from an optical depth @xmath23 ( bottom of the atmosphere of the trapped fireball ) to @xmath24 .
the observed photons mostly come from this region .
the structure of this thin atmosphere can be solved via a set of equations , including the magnetic dipole field equation , the equations of state , the hydrostatic equilibrium equation , the energy transport equation , and the optical depth equation .
the total energy radiated in the pulsating tail of a giant flare is @xmath25 @xcite .
the requirement for the energy being trapped by magnetic fields demands @xmath26 where @xmath27 is the radius of the neutron star , and @xmath28 is the scale of the trapped fireball ( fig.1 ) .
the factor @xmath29 comes from converting the internal energy density ( both the thermal energy and the radiation energy ) of the trapped fireball to pressure . for a dipolar field ,
one has @xmath30 , where @xmath31 is the characteristic surface magnetic field strength of the neutron star .
observations show that the effective radius of thermal emission is a few times of the magnetar radius , @xmath32 @xcite .
we therefore require that the magnetic field strength satisfies @xmath33 which suggests that the neutron star is a magnetar . in the closed field line region ,
the electron - positron pair plasma is trapped , since transport of charged particles across field lines is suppressed by strong magnetic fields , and since gravity is no longer the dominant confining force .
the critical trapping luminosity @xmath34 can be estimated by the balance between the radiative energy density and the magnetic energy density @xmath35 , i.e. @xmath36 which is much larger than the luminosity of the giant flares . for a magnetic dipole field ,
the strength of the magnetic field at a point @xmath37 is @xmath38 where @xmath39 is the polar cap magnetic field strength at the surface .
the border of the closed magnetosphere field lines that enclose the trapped fireball is given by @xmath40 , where @xmath41 is the maximum radius .
next , we consider the equation of state in such a trapped fireball .
the deposition of @xmath42 in the magnetosphere is sufficient to generate a hot photon - pair plasma .
for the internal region of the trapped fireball , the electron - positron pairs are mildly - relativistic .
the total energy density of the photon - pair plasma is @xmath43 ( see eq .
( [ a8 ] ) in appendix ) .
the temperature in the internal region of the fireball is @xmath44 . for the atmosphere of the trapped fireball , on the other hand ,
observations show that its effective temperature is @xmath45 , which is much less than the first landau level energy @xmath46 .
therefore , this is a one - dimensional , magnetized , and non - relativistic electron - positron pair plasma , whose density is given by ( see eq .
( [ a9 ] ) in appendix ) @xmath47 we apply an ideal gas equation of state and assume that the gas is in local thermodynamic equilibrium with the radiation field .
the total pressure @xmath48 is given by the sum of the gas pressure @xmath49 , radiation pressure @xmath50 , and magnetic pressure @xmath51 , i.e. @xmath52 the giant flare pulsating tails are characterized by a duration of @xmath0 .
we assume that the trapped fireball is in hydrostatic equilibrium during the pulsating tail phase : @xmath53 where @xmath54 is the gravitational potential .
for the dipolar magnetic field , the magnetic stress in radial direction ( @xmath55 direction ) is given by @xmath56_{r}=-\frac{1}{2\pi}\left(\frac{b_p}{2}\right)^2\frac{r_s^6}{r^7}[1 + 5\cos^2(\theta)].\label{stress}\end{aligned}\ ] ] therefore one has @xmath57+f_b,\end{aligned}\ ] ] where @xmath58 is the neutron star mass ( with the trapped fireball mass neglected ) , and @xmath59 is the gravitational radius . here
, we neglect the term @xmath60 in the general oppenheimer - volkoff equation , because its contribution is always smaller than @xmath61 in our atmosphere model . in order to solve the atmospheric structure
, we also need the temperature gradient equation @xmath62 where @xmath63 is determined by the energy transport equation .
as pointed out by @xcite , convection is strongly suppressed by strong magnetic fields of magnetars .
we therefore only consider radiative transport in this paper .
for simplicity of treatment and without losing generality , we use the result from the radiative transport in spherical symmetry @xcite , which is given by @xmath64\nonumber\\ & \times&\left[1+\frac{p+1.5p_g+3p_r+p_b}{\rho c^2}\right]^{-1},\end{aligned}\ ] ] where @xmath65 is the luminosity of the photosphere ( i.e. the observed luminosity of the pulsating tail ) , and @xmath66 is the rosseland mean opacity , which is discussed in the next section . in order to obtain the scale length of the photosphere
, one also needs the optical depth equation @xmath67 finally , we introduce the boundary of the bottom of the atmosphere that is defined by @xmath68 , where @xmath69 is the maximum radius of the closed line of the atmosphere bottom , and @xmath70 is the temperature at the bottom of the atmosphere . using eqs.([dipole])-([pressure ] ) and eqs.([stress])-([optical ] ) , we can then solve for the atmospheric structure of the trapped fireball .
we assume @xmath71 , @xmath72 , @xmath73 , the rotation period of the magnetar @xmath74 , and the luminosity of the pulsating tail @xmath75 . since observations show that the effective radius of the thermal emission is a few times of the radius of the magnetar , we assume that the maximum radius of the bottom of the atmosphere is @xmath76 .
the corresponding temperature @xmath70 at @xmath69 should be less than the pair production temperature @xmath77 in order to reach a quasi - steady state solution . in our model ,
we take @xmath78 . for @xmath79 ,
the temperature is so low that the thermal electron - positron pair density decreases exponentially with temperature , leading to a very low optical depth . in fig .
[ fig2]-[fig5 ] , we plot the temperature , optical depth , number density of electron - positron pairs and pressure as a function of the radial distance from the bottom of the atmosphere @xmath80 , where @xmath81 is the radial distance from the neutron star center to the bottom of the atmosphere ( see fig .
[ fig1 ] , with @xmath82 denoted by dotted , dashed and solid lines ) . for the given boundary condition described above
, the atmosphere is very thin @xmath83 and the effective temperature at the photosphere ( @xmath84 ) is @xmath85 for @xmath86 .
this is consistent with the observed temperature of the pulsating tail .
the temperatures for different @xmath87 have little difference for @xmath88 , and the temperature near the magnetosphere equator is higher than that near the pole for @xmath89 . the optical depth from the direction near the pole is lower than that in the direction of the equator , which means that the flux near the pole might be higher . as shown in fig .
[ fig4 ] , we find that the electron - positron pair number density decreases rapidly when @xmath89 .
figure [ fig5 ] shows pressure as a function of the radial distance to the atmosphere bottom @xmath80 .
the electron - positron pairs , radiation and magnetic pressures are denoted by black , red and blue lines , respectively .
one can see that the total pressure is dominated by the magnetic pressure , so that the pair plasma is trapped in the magnetosphere .
near the equator , one has @xmath90 .
however , near the pole , the gas pressure could be larger than the radiation pressure at the atmosphere bottom .
the result suggests that the pressure at the pole decreases more rapidly than that of the equator , since the temperature at the pole decreases from a smaller radius @xmath80 .
@xcite calculated the photon polarization modes in a magnetized electron - ion plasma . in this section
, we calculate the photon polarization modes in a magnetized electron - positron pair plasma following the procedure introduced by @xcite . as shown below , because of the charge symmetry in a pair plasma , some noticeable differences from the case of an electron - ion plasma exist . for example
, the photons are completely linearly polarized , and there is no vacuum resonance point and ion cyclotron absorption in the case of the pair plasma . for the atmosphere of a trapped fireball , the pair plasma is non - relativistic ( @xmath91 ) . in the coordinate system @xmath92 with @xmath93 along @xmath94 , the dielectric tensor @xmath95 contributed by the plasma
is given by @xcite @xmath96_{\mathbf{z}^\prime=\mathbf{\hat{b}}}=\left ( \begin{array}{ccc } \varepsilon & ig & 0\\ -ig & \varepsilon & 0\\ 0 & 0 & \eta\\ \end{array } \right).\end{aligned}\ ] ] here @xmath97 where @xmath98 represents the charged particle species in the plasma , @xmath99 , and @xmath100 .
for the charged particle @xmath98 , @xmath101 is the cyclotron frequency , and @xmath102 is the plasma frequency .
the parameter @xmath103 delineates damping of the particle motion , where @xmath104 is the damping rate . in our calculation , we assume small damping ( @xmath105 ) . for electron - positron pairs , @xmath106 , we have @xmath107 where @xmath108 with @xmath109 and @xmath110 .
we note that the non - diagonal terms in the matrix are zero due to charge symmetry of pairs . on the other hand ,
in strong magnetic fields , vacuum polarization makes a corrected contribution to the dielectric tensor : @xmath111 where @xmath112 is the unit tensor and @xmath113 is the unit vector along @xmath93 .
similarly , vacuum polarization also modifies the magnetic permeability tensor @xmath114 as @xmath115 where @xmath116 and @xmath117 are the magnetizing field and the magnetic field of an electromagnetic wave , respectively .
the vacuum polarization coefficients are given by @xcite @xmath118 where @xmath119 is the relevant dimensionless magnetic field parameter , and @xmath120 is the fine - structure constant .
if @xmath121 ( @xmath122 ) , the vacuum polarization contribution , @xmath123 , could be added linearly to the dielectric tensor @xmath95 . in the frame with @xmath113 along @xmath94 ,
one has @xmath124_{\mathbf{z}^\prime=\mathbf{\hat{b}}}&=&\left
( \begin{array}{ccc } \varepsilon & 0 & 0\\ 0 & \varepsilon & 0\\ 0 & 0 & \eta\\ \end{array}\right)+\left ( \begin{array}{ccc } a-1 & 0 & 0\\ 0 & a-1 & 0\\ 0 & 0 & a+q-1\\ \end{array } \right)\nonumber\\ & = & \left ( \begin{array}{ccc } \varepsilon^\prime & 0 & 0\\ 0 & \varepsilon^\prime & 0\\ 0 & 0 & \eta^\prime\\ \end{array}\right),\end{aligned}\ ] ] where @xmath125 and @xmath126 . using the maxwell equations , the equation for a plane wave with @xmath127 can be expressed as @xmath128 which is @xcite , @xmath129\right\rbrace e_j=0 , \label{wave}\end{aligned}\ ] ] where @xmath130 is the refractive index . in the coordinate system @xmath131 with @xmath132 along @xmath133-axis and @xmath93 in the @xmath134 plane , the dielectric tensor is given by @xmath124_{\mathbf{z}=\mathbf{\hat{k}}}=\left ( \begin{array}{ccc } \varepsilon^\prime\cos^2\theta_b+\eta^\prime\sin^2\theta_b & 0 & ( \varepsilon^\prime-\eta^\prime)\sin\theta_b\cos\theta_b\\ 0 & \varepsilon^\prime & 0\\ ( \varepsilon^\prime-\eta^\prime)\sin\theta_b\cos\theta_b & 0 & \varepsilon^\prime\sin^2\theta_b+\eta^\prime\cos^2\theta_b\\ \end{array } \right),\end{aligned}\ ] ] where @xmath135 is the angle between @xmath132 and @xmath93 . according to the @xmath133-component of eq.([wave ] ) , one has @xmath136 then eq.([wave ] ) becomes @xmath137 where @xmath138 , and @xmath139.\end{aligned}\ ] ] there are two solutions to eq.([marix ] ) : ( i. ) if @xmath140 , then @xmath141 and @xmath142 .
the electric field vector of the wave is in the @xmath143 plane .
this is the ordinary mode ( o mode ) .
if @xmath144 , then @xmath145 and @xmath146 . the electric field vector of the wave is perpendicular to the @xmath143 plane .
this is the extraordinary mode ( e mode ) .
different from the electron - ion plasma , we find that for electron - positron pairs , one does not need to define the polarization ellipticity @xmath147 @xcite , and both modes are linearly polarized . in this case
, there is no vacuum resonance " point where e mode vs. o mode classification becomes ambiguous and can convert from one mode to the other .
the mode eigenvector in the coordinate system with the wave vector @xmath132 along @xmath133-axis is @xmath148 the scattering opacity depends on the polarization vector through its projection on the coordinate frame with the @xmath133-axis along the magnetic field @xmath93 direction @xcite . in this new
coordinate system with @xmath93 along @xmath133-axis , the eigenvector is given by @xmath149 the cyclic components of @xmath150 read and @xmath151 , and @xmath152 and @xmath153 are complex numbers , which are determined by the plasma medium properties .
thus , there is a plus or minus sign in @xmath154 in their notation . ]
@xmath155 corresponds to the e mode , and @xmath156 corresponds to the o mode .
specifically , for the e mode , the cyclic components are given by @xmath157 .
we note that the eigenvectors of the e mode are constants for the electron - positron plasma .
this is different from the case of electron - ion plasma .
the scattering opacity from the @xmath158 mode into the @xmath159 mode is given by @xcite @xmath160^{-1}|e_\alpha^j|^2a_\alpha^i,\end{aligned}\ ] ] where @xmath161 is the damping factor , and @xmath162 .
the electron scattering opacity from the mode @xmath158 ( into both model @xmath159 and mode @xmath158 ) is @xmath163^{-1}|e_\alpha^j|^2a_\alpha,\label{opacji}\end{aligned}\ ] ] where @xmath164 .
the scattering opacities for the two modes can be approximately derived @xcite . for the e - mode
, one has @xmath165 where the second equation is under the condition of @xmath166 ; for the o - mode , one has @xmath167 according to eq.([scae])-eq.([scao ] ) , the probability of e - mode photons converting into o - mode photons is @xmath168 , and the probability of o - mode photons converting into e - mode photons is @xmath169 . for the transverse - mode approximation ( @xmath170 )
, one has @xmath171 .
thus , the opacity is approximately @xmath172\end{aligned}\ ] ] for the e mode , and @xmath173\end{aligned}\ ] ] for the o model . for radiation frequencies below the cyclotron frequency ,
i.e. @xmath174 , and the damping factor @xmath175 , one has @xmath176 where @xmath177 ( @xmath178 is the thomson cross section , and @xmath179 is the electron mass ) . due to the absence of the ion cyclotron absorption and vacuum resonance absorption in the case of electron - positron plasma , the above equations could be approximately applied to calculate the opacities for @xmath174 .
we then finally obtain the rosseland mean opacity @xmath66 through @xmath180\bigg/\left[\int_0^\infty\frac{\partial b_\nu(t)}{\partial t}d\nu\right].\end{aligned}\ ] ] for electron energy @xmath181 , the contribution from electron - electron bremsstrahlung can be safely ignored @xcite .
we therefore ignore free - free absorption of electron pairs in the atmosphere of the trapped fireball .
we apply the full expressions of @xmath182 and @xmath183 in eq .
( [ opacji ] ) to perform the calculation .
the scattering opacities @xmath184 of the two modes as a function of energy for various angle @xmath135 , magnetic fields @xmath185 , and electron - positron pair density @xmath186 are presented in figures [ fig6 ] - [ fig8 ] , respectively .
different from the case of electron - ion plasma @xcite , for electron - positron pair plasma , there are no ion cyclotron absorption and vacuum resonance absorption , so that the opacities satisfy eq .
( [ opacity ] ) for @xmath187 . as shown in fig.[fig6 ] , for magnetic field @xmath188 and electron - positron pair number density @xmath189
, the o mode opacity exhibits an angle dependence for @xmath135 , i.e. @xmath190 . the e model opacity , on the other hand , is independent of @xmath135 ( see eq .
( [ base ] ) and eq .
( [ opacity ] ) ) .
as shown in fig [ fig7 ] , for @xmath189 and @xmath191 , the electron cyclotron peak @xmath192 is proportional to the strength of magnetic field @xmath185 and the e mode opacity satisfies @xmath193 for @xmath174 . as shown in fig .
[ fig8 ] , for @xmath188 and @xmath191 , the plasma peak @xmath194 is proportional to @xmath195 .
in this section , we conduct monte carlo simulations to calculate the polarization properties of the x - ray photons emitted from the trapped fireball . to calculate the polarization signals
, we set a coordinate system and @xmath196 introduced in this section are the global coordinate systems for the neutron star , which are very different from the local " coordinate systems @xmath197 and @xmath198 introduced in section 3 . in the following derivations ,
the coordinate symbols are still kept in lower case for these global coordinate systems . ]
@xmath199 with @xmath200-axis along the line of sight @xmath201 , as shown in fig .
the spin angular velocity vector @xmath202 is in the @xmath203 plane .
we define @xmath204 as the angle between the magnetic dipole axis @xmath114 and @xmath202 , @xmath205 as the angle between @xmath114 and the line of sight @xmath201 , and @xmath206 as the angle between @xmath201 and @xmath202 . for a dipolar field , in the coordinate system @xmath207 with @xmath114 along the @xmath208-axis ,
the magnetic field components in a spherical coordinate system are given by @xmath209 transforming from the spherical coordinate system to the cartesian coordinate system , we have @xmath210 where @xmath211 in the coordinate system @xmath199 with the line of sight along the @xmath200-axis , the magnetic field components can be calculated by @xmath212 and the coordinate system transformation is given by @xmath213 where @xmath214 with @xmath215 being the azimuthal angle , and @xmath216 being the rotation phase ( @xmath217 is the initial rotation phase ) .
we assume that the photon flux is same everywhere at the bottom of the atmosphere . in order to calculate the photon spatial distribution
, we need to know the surface area of the magnetosphere @xmath218 with @xmath87 .
because the border of the magnetosphere is given by @xmath40 , where @xmath41 is the maximum radius of the closed field lines , the length of the dipole field line is given by @xmath219 thus , the differential area of the magnetosphere is given by @xmath220 integrating over @xmath221 , we obtain the area of the closed magnetic field line within @xmath87 , i.e. @xmath222 . the total area is @xmath223 . using this information
, we can apply the monte carlo method to assign the initial spatial distribution of the photons . as we have discussed above ,
the observations show that the spectra of the pulsating tails are dominated by the thermal component .
therefore , in this paper , we only consider the thermal component contributed by the photon - pair trapped fireball .
because the optical depth of the trapped fireball is very large , the observed radiation comes from the atmosphere of the trapped fireball .
we apply the atmosphere model described in sec
. 2 . to calculate the photon polarization properties through monte carlo simulations .
the procedure of our calculations is the following : a. we assume that the boundary of the atmosphere of the trapped fireball satisfies the follow conditions : ( i. ) the bottom of the atmosphere is a closed magnetic field line surface defined by @xmath224 with a same temperature @xmath70 .
the radiative flux is the same everywhere at the bottom of the atmosphere . as a result , the initial photon spatial distribution can be simulated according to the coordinates @xmath225 , i.e. @xmath226 , @xmath221 is given by eq .
( [ ds ] ) , @xmath227 , with @xmath228 being a random number between ( 0,1 ) .
the initial directional distribution @xmath229 of the photons is isotropic , i.e. , @xmath230 and @xmath231 .
the initial photon number in the e mode is equal to that in the o mode at the bottom of the atmosphere .
however , the photons in one mode can be converted to the other mode due to the electron scattering in the atmosphere , leading to the change of the ratio between two modes .
b. we keep track the propagation and transition of each photon through the medium .
( i. ) at each time step , we grid the atmosphere of the trapped fireball in 3 dimensions .
the properties ( e.g. density , temperature , pressure ) in each cell are kept constant .
the scattering optical depth @xmath232 is given by @xmath233 .
each photon is allowed to travel along its trajectory in each time step . the optical depth in each cell
is calculated until @xmath234 .
note that we need to use different opacities for different photon modes in this process .
scattering would change the mode of the photons .
the probability for e mode converting into o mode is @xmath235 , whereas the probability for o mode converting into e - mode is @xmath236 .
due to the electrons thermal motion , scattering would be isotropic .
the direction @xmath225 of a scattered photon is given by @xmath230 and @xmath231 .
( v. ) we assume that the atmosphere satisfies local thermodynamic equilibrium ( lte ) .
the energy distribution of the photons depends on the energy distribution of the electron - positron pairs that have a thermal distribution .
( vi . ) we repeat the above procedure for each time step . based on the atmosphere model discussed above
, we find that the photons are almost no longer scattered when @xmath237 , since @xmath238 . c. we collect the photons along the light of sight within the solid angel @xmath239 , and calculate the directional angel @xmath240 for both photon modes at @xmath241 . for e - mode photons ,
one has @xmath242 , where @xmath243 and @xmath244 and @xmath245 are the magnetic field components perpendicular to the light of sight . for o mode photons ,
one has @xmath246 .
the stokes parameters near the photosphere are given by @xmath247 where the subscripts refer to different bases of the jones vector space : @xmath248 is the cartesian basis , @xmath249 is the cartesian basis rotated by @xmath250 , @xmath251 is the circular basis , which is defined as @xmath252 , and @xmath253 , @xmath254 , @xmath255 and @xmath256 are given by @xmath257 where @xmath159 presents each observed photon and @xmath258 presents the mode of the photon @xmath159 .
finally , the polarization degree near the photosphere is given by @xmath259 and the polarization angle is @xmath260 after a photon escapes the trapped fireball , it will propagate in the magnetic vacuum .
the dielectric and inverse permeability are @xmath261 and @xmath262 , respectively .
for a transverse electromagnetic wave emitted from a point in the magnetosphere atmosphere , it can be treated as the superposition of e - mode and o - mode photons , i.e. @xmath263 where @xmath264 with the refraction indices for different modes being @xmath265 one therefore has @xmath266 .
the evolution of the mode amplitudes is given by @xcite @xmath267 where @xmath268 represents the photon trajectory .
the condition for adiabatic evolution of the photon modes is @xmath269 so that the non - diagonal elements in the matrix of eq.([eq : lai - ho ] ) is essentially zero , and the normal modes do not mix @xcite . as the photon propagates in the magnetosphere , the qed correction to the dielectric tensor falls off as @xmath270 , and non - diagonal components would appear . within the neutron star atmosphere context ,
@xcite first defined the qed polarization limiting radius @xmath271 ( the distance from the star at which the adiabatic approximation breaks down ) , and found that the qed limit radius allows the neutron star surface polarization to be sufficiently large to be observed ( see also * ? ? ?
here we perform a similar treatment for the trapped fireball .
we approximately treat the adiabatic region to have a sharp edge at @xmath271 . based on the adiabatic condition ( [ eq : adiabatic ] ) , we calculate the polarization - limit radius , @xmath271 , which is set by the condition @xmath272 , i.e.@xcite @xmath273 where @xmath27 is the neutron star radius , @xmath274 is a slowly varying function of phase and is of the order of unity , @xmath275 is the photon energy , @xmath185 is the strength of magnetic filed , and @xmath276 is the spin angular velocity .
one can see that the adiabatic condition is safely satisfied in the trapped fireball and its atmosphere . beyond @xmath271 ,
the magnetic field strength has dropped enough so that it no longer affects the state of photon polarization , i.e. the photon polarization state is ` frozen ' . because @xmath277 , the electric vectors of all the photons with a certain energy @xmath275 emitted from the photosphere would rotate for a same angle @xmath278 , so that the polarization degree is no longer affected .
since magnetars are slow rotators , the effect of magnetar spin is not important in the polarization calculations . for simplicity , in the monte carlo simulations , we assume an aligned rotator , i.e. the spin angular velocity vector @xmath202 is parallel to the magnetic axis @xmath114 , so that @xmath279 .
all the photons with a certain energy have almost the same rotation angle @xmath280 under the adiabatic condition . in this case , the stokes parameter @xmath281 , so that the polarization degree is @xmath282 and the polarization angle is @xmath283 .
we consider two energy bands , i.e. a soft band ( @xmath3 ) and a hard band ( @xmath5 ) , and collect 5000 photons in each band within a solid angel @xmath239 along the light of sight . the e mode and o mode
are defined by the direction of the magnetic field at @xmath241 , where the adiabatic condition is satisfied .
our simulation results are displayed in table [ tab1 ] .
as shown in the table , there are more e - mode photons in the soft band than in the hard band , since the opacity of the e - mode photons @xmath284 .
the number ratio between the e - mode and o - mode photons , @xmath285 , is higher for @xmath286 than that for @xmath287 .
this is because the opacity of e - mode photons is suppressed in strong magnetic fields , i.e. @xmath288 , and @xmath185 is stronger near the polar . on the other hand ,
the polarization degree is close to zero for @xmath286 and reaches the maximum value at @xmath287 , i.e. 27.9% in the 1 - 30 kev band and 10.0% in the 30 - 100 kev band .
the result could be qualitatively explained from the direction of the magnetic field and the photon number ratio between e - mode and o - mode . as shown in the left panel of fig .
[ fig10 ] , if the light of sight is parallel to the magnetic axis , the directional angles of e - mode and o - mode are evenly distributed from 0 to @xmath289 . therefore , even though @xmath285 has the maximum value , the polarization degree is still zero . on the other hand , as shown in the right panel of fig .
[ fig10 ] , if the light of sight is perpendicular to the magnetic axis , the projected directions of the magnetic field lines almost have the same direction . in this case , the polarization degree depends on @xmath285 . since @xmath285 is of the order of unity , the polarization degree is mainly determined by the direction of the magnetic fields , so that an ordered @xmath185 field configuration naturally results in a relatively large polarization degree .
polarization observations provide the key information to probe the structure of the magnetar magnetosphere during giant flares . after the initial spike
, the residual energy would form an optically thick photon - pair plasma .
this photon - pair plasma is trapped in the magnetar magnetosphere , forming a `` trapped fireball '' .
observations show that the flux of the thermal component is much higher than that of the non - thermal component between @xmath16 to @xmath22 in the early and later stages @xcite . in this paper
, we assumed that the trapped fireball is enclosed by a set of closed - field lines in the magnetosphere , and calculated the atmospheric structure and polarization degree of the trapped fireball .
we have reached the following conclusions : a. regarding the atmosphere of the trapped fireball : ( i. ) the atmosphere of the trapped fireball is much thinner than the thickness of the trapped fireball .
the size of a trapped fireball is a few times of @xmath27 , which is determined by the requirement of the confinement of fireball energy within the magnetosphere , as shown in eq .
( [ trap ] ) .
observations show that the effective radius of the thermal component is a few times of the radius of neutron star , which is consistent with this requirement .
we calculated the structure of the trapped fireball and its atmosphere .
our results show that the thickness of the atmosphere is @xmath290 with a sharp temperature gradient .
( ii . ) in the atmosphere , the electron - positron pairs are non - relativistic , and the thermal pair density decreases with temperature ( see eq .
( [ eos ] ) ) . the pair gas pressure is lower than the radiative pressure at the photosphere , and the total pressure is dominated by the magnetic pressure .
the atmosphere could be still in hydrostatic equilibrium due to the magnetic confinement . in the trapped fireball ,
gravity is not the dominant confining force , and the plasma is trapped due to the suppression of the transport of charged particles across the magnetic field lines .
opacity plays an important role in the structure of the trapped fireball .
different from the surface of a neutron star @xcite , the trapped fireball is dominated by the electron - positron pairs @xcite .
we assume that the atmosphere is baryon free . for an electron - ion plasma
, there is a vacuum resonance point where e - mode and o - mode photons could convert from one mode to another , and there are two peaks in the opacity - energy relation , e.g. the vacuum resonance absorption and the proton cyclotron absorption .
however , for electron - positron pairs , both e - mode and o - mode photons are entirely linearly polarized . in this case
, there is no vacuum resonance point where they can change mode during their propagation in the medium .
rather , for electron - positron pairs , there is only one electron cyclotron absorption peak with energy @xmath291 , which is much larger than that of the observed effective temperature @xmath292 .
the atmosphere temperature distribution is determined by the structure of the magnetosphere . for a specific angle @xmath87 ,
the longer the distance from the magnetar , the lower the temperature . for a specific distance @xmath55 ,
the temperature of the equator is higher than that of the pole ( note that the abscissa in fig .
[ fig2]-[fig5 ] is the logarithm of @xmath80 rather than @xmath55 ) .
the pair density and opacity are determined by the temperature , so that the photosphere is no longer a sphere .
observationally one could only infer the effective size of the trapped fireball , but not the shape of the trapped fireball .
( v. ) our results show that the opacity of e mode is independent of @xmath135 for electron - positron pairs , which is different from the case of the electron - ion plasma @xcite .
b. regarding the polarization properties of the trapped fireball : ( i. ) the opacity of the e - mode photons scales as @xmath284 . as a result , the number of e - mode photons is larger than that of o - mode photons in softer bands .
conversely , o - mode photons become more dominant in harder bands . the number ratio between e - mode and o - mode photons , @xmath285 , is higher for @xmath286 than for @xmath287 , as a result of the suppression of the e - mode photon opacity in strong magnetic fields near the magnetic pole .
the polarization degree is zero for @xmath286 and reaches a maximum value for @xmath287 .
the maximum polarization degree from the trapped fireball is lower than that ( @xmath293 ) from a hot spot on the pole of the magnetar surface @xcite or in the quiescent magnetar magnetosphere @xcite .
this is mostly due to the extended thermal emission from a trapped fireball and the wider collected band , e.g. @xmath3 or @xmath5 considered in our work .
( iii . ) during the photon propagation in the magnetic vacuum , due to the adiabatic condition ( @xmath294 ) , the electric vector direction of the electromagnetic wave continues to change according to the direction of the local magnetic field . because the polarization - limit radius @xmath271 is much larger than that of the trapped fireball radius , the electric vectors of all the photons with a certain energy emitted from the photosphere rotate a same angle @xmath278 as they escape the magetosphere .
( iv . ) the polarization degree at @xmath295 can be as large as @xmath296 in 1 - 30 kev , and is @xmath297 in 30 - 100 kev .
since giant flares typically have large fluxes , such a polarization degree would be detectable future x - ray/@xmath14-ray polarimeters such as polar and leap , and may have been detected by the existing x - ray detectors such as rhessi .
we encourage a re - analysis of the possible polarization signal of sgr 1806 - 20 giant flare pulsating tail in the rhessi data @xcite . c. implications for detecting the predicted polarization signal :
our calculations are based on the magnetar conjecture and robust qed physics .
if the future observations confirm the polarization signal at the predicted level , it would suggest the following : ( i. ) the magnetar hypothesis is correct .
even if the magnetar model is the leading model to interpret sgrs , some other scenarios have been proposed to interpret sgrs without introducing strong magnetic fields .
for example , in the accretion models @xcite , the strength of the magnetic field is weak , i.e. @xmath298 . according to eq.([eq : qed - radius ] ) , the qed polarization limiting radius would be @xmath299 , which is not much larger than the scale of the trapped fireball
. this would give rise to a much smaller polarization degree .
qed is right : a high polarization degree suggests that vacuum polarization indeed makes a corrected contribution to the dielectric tensor and the magnetic permeability tensor , and that the qed vacuum is birefringent .
otherwise the polarization degree would be less than @xmath297 even for @xmath287 .
a polarization measurement would reveal the geometry of the system .
since @xmath285 is of the order of unity , the polarization degree is mainly determined by the magnetic field line directions .
therefore , a relatively large polarization degree corresponds to an ordered magnetic field configuration , and a relatively large angle between the magnetic pole and line of sight .
( iv . ) the idea of the pair - plasma trapped fireball is right : different from the electron - ion plasma , the pair plasma has no vacuum resonance point and ion cyclotron absorption .
both e - mode and o - mode photons are entirely linearly polarized .
the properties of pair opacity are also very different from that of electron - ion plasma .
therefore , an observed polarization degree at the predicted level would prove that the trapped fireball is indeed a photon - pair plasma . in our monte carlo simulations
, we have assumed an aligned rotator ( the rotation axis and magnetic axis overlap ) , so that the @xmath205 angle is constant for a particular observer . in reality , the two axes are mis - aligned .
the slow rotation of the magnetars would not affect the @xmath205-dependent polarization degrees calculated in this paper .
however , a certain observer would view the magnetosphere in a range of @xmath205 values .
the average polarization degree would be between the minimum and maximum @xmath300 values defined the maximum and minimum @xmath205 , which would usually be still @xmath297 in the 1 - 30 kev band ( or a few percent in 30 - 100 kev ) , unless the two axes are nearly aligned and the line of sight is also very close to the pole .
we thank the anonymous referee for valuable and detailed suggestions that have allowed us to improve this manuscript significantly .
this work is partially supported by national basic research program ( 973 program ) of china under grant no .
ypy is supported by china scholarship program to conduct research at unlv .
alpar , m. a. , 2001 , , 554 , 1245 boggs , s. e. , zoglauer , a. , bellm .
e. , hurley , k. , lin , r. p. , smith , d. m. , wigger , c. , & hajdas , w. 2007 , , 661 , 458 fernndez , r. , & davis , s. w. 2011 , , 730 , 131 feroci , m. , frontera , f. , costa , e. , amati , l. , tavani m. , rapisarda , m. , & orlandini , m. 1999 , , 515 , l9 feroci , m. , hurley , k. , duncan , r. c. , & thompson , c. 2001 , , 549 , 1021 feroci , m. & soffitta , p. , 2011
, in high - energy emission from pulsars and their systems , springer - verlag berlin heidelberg p585 harding , a. k. , & lai , d. 2006 , rep .
phys . , 69 , 2631 haug , e. 1975 , z. naturforsch . , 30 , 1099 heyl , j. s. , & shaviv , n. j. 2000 , mnras , 311 , 555 heyl , j. s. , & shaviv , n. , 2002 , phys .
d , 66 , 3002 heyl , j. s. , shaviv , n. , & lloyd , d. , 2003 , mnras , 342 , 134 ho , w. c. g. , & lai , d. 2001 , , 327 , 1081 ho , w. c. g. , & lai , d. 2003 , , 338 , 233 hurley , k. , et al . 1999 , , 397 , 41 hurley , k. , et al .
2005 , , 434 , 1098 jahoda k. m. , kouveliotou c. , kallman t. r. et al .
, 2015 , american astronomical society , aas meeting # 225 , # 338.40 lai , d. , & ho , w. c. g. 2002 , , 566 , 373 mazets , e. p. , golenetskii , s. v. , gurian , i. a. , & ilinskii , v. n. 1982 , ap&ss , 84 , 173 mcconnell , m. l. , ryan , j. m. , smith , d. m. et al . , 2007 , american astronomical society , aas meeting # 201 , # 93.01 mcconnell , m. l. , baring , m. , bloser , p. et al . , 2014 , proceedings of the spie , volume 9144 , i d .
91440o 8 pp .
mereghetti , s. , 2008 , a&arv , 15 , 225 mszros , p. , 1992
, high - energy radiation from magnetized neutron stars .
chicago press , chicago miller , m. c. , 1995 , , 448 , l29 olausen , s. a. , & kaspi , v. m. 2014 , , 212 , 6 paczyski b. , & anderson n. , 1986 , apj , 302 , 1 pavlov , g. g. , & zavlin , v. e. 2000 , , 529 , 1011 produit , n. , barao , f. , deluit , s. , hajdas , w. , leluc , c. , pohl , m. , rapin , d. , vialle , j. p. , walter , r. & wigger , c. 2005 , nimpa , 550 , 616 .
rajagopal , m. , & romani , r. w. , 1996 , , 461 , 327 suarez - garcia , e. et al .
, 2010 , nimpa , 624 , 624 taverna , r. , muleri , f. , turolla , r. , soffitta , p. , fabiani , s. , & nobili , l. 2014 , , 438 , 1686 taverna , r. , turolla , r. , gonzalez caniulef , d. , et al . , 2015 , arxiv:1509.05023 thompson , c. , & duncan , r. c. 1995 , , 275 , 255 thompson , c. , & duncan , r. c. 2001 , , 561 , 980 thorne , k. s. , 1977 , apj , 212 , 825 ulmer , a. , 1994 , , 437 , l111 van adelsberg , m. , & lai , d. , 2006 , , 373 , 1495 van adelsberg , m. , & perna , r. , 2009 , , 399 , 1523 ventura , j. 1979 , phys . rev .
d. , 12 , 1132 weisskopf , m. c. , bellazzini , r. , costa , e. et al . , 2014 , american astronomical society , head meeting # 14 , # 116.15 xiao , h. , hajdas , w. , wu , b. , & produit , n. , 2015 , arxiv:1507.04474 xiong , s. , produit , n. , & wu , b. , 2009 , nimpa , 606 , 552
in order to obtain the structure of a trapped fireball , we need to know the relation between the pair energy density and temperature . here , we summarize the basic properties of a free electron gas in strong magnetic fields . the results are consistent with @xcite .
consider an elementary cell in a phase space with a volume @xmath301 where @xmath302 and @xmath303 are the position and momentum vectors of the electrons . the electron distribution function as a function of energy @xmath275 is given by @xmath304 where @xmath305 is the degeneracy number and @xmath306 is the chemical potential .
for 3-d free electrons , the number density is given by @xmath307 in strong magnetic fields , the electron distribution is in 1-d ( due to the strong landau confinement ) , which is given by @xcite @xmath308 where @xmath309 is the cyclotron radius , and @xmath310 is the degeneracy at energy level @xmath311 ( @xmath312 and @xmath313 ) .
the landau level in a strong magnetic field is given by @xmath314^{1/2}.\end{aligned}\ ] ] the first landau level is @xmath315 ) for @xmath73 , @xmath76 and @xmath328 .
the dotted , dashed , and solid lines denote @xmath82 , respectively .
the electron - positron pairs , radiation and magnetic pressures are denoted by black , red and blue lines , respectively . ] as a function of energy for various angles between @xmath132 and @xmath93 , @xmath135 , for @xmath188 and @xmath189 .
the black curves represent the e model opacity , and the red curves represent the o mode opacity .
the dotted , dashed and solid curves denote @xmath329 , respectively . ] as a function of energy for different magnetic field strengths @xmath185 for @xmath189 and @xmath191 .
the black curves represent the e mode opacity , and the red curves represent the o mode opacity .
the magnetic field strengths @xmath330 are denoted by dotted , dashed and solid curves , respectively . ] as a function of energy for various electron - positron pair number density @xmath186 for @xmath188 and @xmath191 .
the black curves represent the e mode opacity , and the red curves represent the o mode opacity .
the pair number densities @xmath331 are denoted by dotted , dashed and solid curves , respectively . ] | three giant flares have been detected so far from soft gamma - ray repeaters , each characterized by an initial short hard spike and a pulsating tail .
the observed pulsating tails are characterized by a duration of @xmath0 , an isotropic energy of @xmath1 , and a pulse period of a few seconds .
the pulsating tail emission likely originates from the residual energy after the intense energy release during the initial spike , which forms a trapped fireball composed of a photon - pair plasma in a closed field line region of the magnetars .
observationally the spectra of pulsating tails can be fitted by the superposition of a thermal component and a power - law component , with the thermal component dominating the emission in the early and late stages of the pulsating tail observations . in this paper , assuming that the trapped fireball is from a closed field line region in the magnetosphere , we calculate the atmosphere structure of the optically - thick trapped fireball and the polarization properties of the trapped fireball . by properly treating the photon propagation in a hot , highly magnetized , electron - positron pair plasma , we tally photons in two modes ( o mode and e mode ) at a certain observational angle through monte carlo simulations .
our results suggest that the polarization degree depends on the viewing angle with respect to the magnetic axis of the magnetar , and can be as high as @xmath2 in the @xmath3 band , and @xmath4 in the @xmath5 band , if the line of sight is perpendicular to the magnetic axis . |
since the seminal work of gppert - mayer @xcite , two - photon decay rates of excited states in hydrogen - like atoms and ions have been the subject of intense experimental @xcite and theoretical @xcite studies . for many years , the investigations have dealt not only with the total decay rates but also with the energy and even angular distributions of the two emitted photons . by analyzing these ( two - photon ) properties ,
unique information has been obtained about the structural properties of one - electron systems including subtle relativistic effects as well as about the basic concepts of quantum physics such as , e.g. , the entanglement . even though large experimental and theoretical efforts have been undertaken in the past to understand various aspects of the two - photon decay of hydrogen - like atoms ,
the analysis of this process still raises a number of unresolved problems .
one of these problems , which currently attracts a lot of interest , concerns those two - photon transition from highly excited states to the ground state which pass through an intermediate state with a lower energy than the initial state of the two - photon process @xcite and can alternatively decay to the ground - state via two ( or more ) sequential one - photon emissions .
such a transition leads to resonance peaks in the energy spectrum of the coherently emitted photons from the two - photon decays which are located at the energies corresponding to the ( real ) intermediate states .
one of the most pronounced examples of such a situation is the @xmath0 two - photon @xmath1 transition for which the differential ( in energy ) emission rate has singularities at energies corresponding to the @xmath2 and @xmath3 cascades . a proper treatment of these singularities is obviously required for computing total decay rates obtained after an integration over the energies of the coherently emitted photons in the direct two - photon decay @xmath4 .
during the last two decades , the theoretical treatment of the resonances in the energy distribution of the emitted photons has been discussed in a number of places . in general , the decay rate @xmath5 of an initial state @xmath6 in a hydrogenlike atom is the sum of a one - photon decay rate @xmath7 and a two - photon contribution @xmath8 , @xmath9 the expression for @xmath8 as originally derived in @xcite is easily seen to involve an integral over the energies of the emitted photons , the sum of which has to be equal to the energy difference of the initial and final states , and a summation over all possible intermediate , virtual states . in order to avoid problems with non - integrable singularities ,
the authors of @xcite have attributed @xmath8 only to the so - called non - resonant intermediate transitions , in contrast to a summation over the complete intermediate - state spectrum .
the non - resonant transitions are given by intermediate states of energy higher than the energy @xmath10 of the initial state @xcite ( the `` resonant '' intermediate states , which are involved in the one - photon cascade processes , are explicitly excluded from the sum over intermediate states ) .
based on this assumption , the non - resonant contribution for the @xmath0 two - photon transition was calculated as @xmath11 for hydrogen .
later , this result has been also confirmed in the calculations by florescu and co - workers @xcite who used a different method for the summation over the `` non - resonant '' states .
although the results presented in refs . @xcite are in mutual agreement , they are both based on the interpretation of the two - photon decay rate @xmath8 as a rate generated only by non - resonant intermediate levels . in our manuscript
, we would like to propose an alternative way for the computation of the ( two - photon ) total decay rates which leads to a natural removal of the infinities otherwise introduced into the expression for the two - photon decay rate .
we apply here a fully relativistic , quantum electrodynamical approach to re - investigate the two - photon decay of highly - excited states of hydrogen - like atoms , paying special attention to a careful handling of the resonances infinitesimally displaced from feynman s photon integration contour ( these singularities exactly correspond to the problematic `` resonant '' intermediate states ) . by making use of this approach ,
we obtain finite , physically sensible results for the decay rates of the two - photon @xmath0 transitions in neutral hydrogen as well as in the various hydrogen - like ions .
apart of the leading , electric dipole ( @xmath1 ) transition , we also discuss the contributions from the higher multipole components to the total decay rate .
this paper is organized as follows : after a brief survey of the theoretical expressions used in our analysis ( sec . [ theory ] ) , we proceed by discussing the method of evaluation ( sec . [ evaluation ] ) as well as the numerical results obtained for the differential as well as the total two - photon decay rates ( sec .
[ results ] ) .
conclusions are given in sec .
[ conclusions ] .
within quantum electrodynamics , the ( negative ) imaginary part of the self - energy is just the @xmath12 , where @xmath13 is the decay width @xcite .
the one - photon decay @xmath14 rate is obtained from the imaginary part of the one - loop self - energy , while the two - loop self - energy gives rise to the two - photon decay rate @xmath15 . because the relativistic formulation of the two - loop self - energy problem has been discussed before in a number of places @xcite , we only mention here that by following a straightforward generalization of the standard procedure described for the non - relativistic framework in refs .
@xcite , we obtain the following expression for the two - photon decay rate ( @xmath16 ) , @xmath17 where @xmath18 with the initial and final state energies @xmath10 and @xmath19 , respectively .
@xmath20 is given by @xmath21 where in the second factor , the initial and the final state are exchanged , but the infinitesimal imaginary part in the denominators remains @xmath22 ( i.e. , does not change sign ) .
we here manifestly assume that @xmath23 and @xmath24 are the well - known solutions of the dirac hamiltonian for a single electron in the standard representation , describing an electron bound to a point - like nucleus with charge number @xmath25 .
for photons propagating with wave vector @xmath26 and unit polarization vector @xmath27 ( @xmath28 ) , moreover , the electron - photon interaction operator @xmath29 in the transition amplitude ( [ amplitude_general ] ) can be written in velocity gauge as : @xmath30 where @xmath31 is a normalization factor , @xmath32 are the standard dirac matrices , and @xmath33 denotes the _ helicity _ , i.e. the spin projection of the photon onto the direction of propagation @xmath34 .
it is important to note that even though the electron - photon interaction operator ( [ interaction_operator ] ) depends , of course , on the direction of the photon emission , one has to integrate over these directions in eq .
( [ gamma_2_qed ] ) in order to get the total decay rate .
the summation over the intermediate states in the amplitude ( [ amplitude_general ] ) runs over the _ complete _ one - particle spectrum @xmath35 , including a summation over the discrete part of the spectrum as well as an integration over the positive and negative - energy continuum of the dirac spectrum .
one has to use the full dirac coulomb green function which is not known in closed analytic form in order to perform this calculation consistently . in recent years
, the green s function method @xcite has been widely applied for the analysis of the total two - photon decay rates as well as the photon - photon angular correlation functions @xcite .
various possibilities for the numerical implementation of the relativistic green s function are known , among which we would like to mentionas ( i ) a well - known formulation in terms of whittaker functions @xcite and ( ii ) a sturmian decomposition in terms of laguerre polynomials as suggested by hylton and snyderman @xcite . for the dominant e1e1 component of the @xmath36 two - photon transitions in neutral hydrogen as well as in hydrogen - like ions with @xmath37 and @xmath38 , where @xmath39 is the energy sharing @xmath40 .
relativistic wave functions are used for the initial , intermediate and final states , but the electron - photon interaction has been restricted to electric dipole term ( @xmath41 term ) .
the resonance peaks in the decay rate correspond to the resonant @xmath2 and @xmath3 decay processes . ]
we use the latter representation and apply the techniques of racah s algebra to all spherical tensors and to the standard radial - angular representation of the wave functions , and to the dirac coulomb green function . for the interaction of electrons with the radiation field
, the spherical tensor components are obtained from the known standard multipole expansion of the photon operator [ see , e.g. , eq .
( 5 ) of ref .
@xcite ] , @xmath42 where @xmath43 denotes the electric ( @xmath44 = 1 ) and magnetic ( @xmath44 = 0 ) multipole fields , respectively .
the great advantage of the multipole decomposition ( [ photon_operator_decomposition ] ) is that is allows us to study the contributions to the total ( two - photon ) decay rate from the various _ allowed _ multipole combinations .
we use the integrand in the integral over @xmath45 in eq .
( [ amplitude_general ] ) as a measure of the differential decay rate ( where we can set explicitly @xmath46 for the differential rate ) .
the energy distributions of the two photons emitted in the @xmath0 decay of neutral hydrogen and hydrogen - like ions are calculated as a function of the energy sharing parameter @xmath47 . for an energy sharing in the range @xmath48 , the contributions to the energy distribution from the @xmath41 and @xmath49 multipole combinations
are displayed in figs . 1 and 2 , respectively .
as seen from these figures , the photon energy distributions for both multipole combinations exhibit sharp resonance peaks . as already mentioned , this behaviour is due to the fact that the summation in eq .
( [ amplitude_general ] ) includes also intermediate states @xmath50 having an energy @xmath51 with @xmath52 .
however , the intermediate states contributing to the peaks are not only defined by the ( symmetry of ) the initial @xmath53 and final @xmath54 states but also by the multipole components of the radiation field involved in the two - photon process and are different for @xmath41 as opposed to @xmath49 , and in addition , marked differences exist between the low-@xmath25 and the high-@xmath25 region .
e.g. , the fine - structure of the resonance in the @xmath41 energy spectrum grows with @xmath25 ( the contributing states are @xmath55 and @xmath56 ) . by contrast , no splitting is observed even for very heavy ions for the @xmath49 component of the @xmath0 decay .
only one intermediate state , namely @xmath56 , is allowed for @xmath49 . for the e1m2 component of the @xmath36 two - photon transitions in neutral hydrogen as well as in hydrogen - like ions with @xmath37 and @xmath38 ( @xmath39 is the fraction of energy carried by the e1 photon ) .
the @xmath49 term is treated relativistically .
the resonances in the decay rate exclusively correspond to the @xmath3 cascade , the @xmath55 state does not contribute . ]
our treatment of the intermediate state resonance peaks in the integration over the photon energy is dictated by an accurate analysis of eqs .
( [ gamma_2_qed])([amplitude_general ] ) as obtained form the relativistic two - loop self energy .
the general structure of the contribution to @xmath57 generated by virtual states with intermediate energies @xmath58 with resonance energy @xmath59 or @xmath60 is as follows : @xmath61 the integration of the first term can be carried out using the dirac prescription ( see , for example , ref .
the second term of eq .
( [ resonant_terms ] ) can be treated using the formula @xmath62 where we used an appropriate scaling of the photon energy integration variable in order to map the integration region to the interval @xmath63 .
it is important to note that eq .
( [ quadratic_term_integration ] ) holds strictly for @xmath64 , but the limit is not approached uniformly @xcite ; i.e. , it would be forbidden to exchange the sequence of the limit @xmath65 with the integration over @xmath66 . as usual in quantum electrodynamic processes ,
all regulators have to be kept up until the very end of the calculation . with these preparations ,
it is easy now to integrate over the photon energies ( see table 1 for the @xmath0 process ) . as seen from the table , the cross sections for the @xmath41 , @xmath49 and @xmath67 components components scale with the nuclear charge as @xmath68 , @xmath69 and @xmath69 , respectively .
as expected , this scaling behaviour is similar to the @xmath25-scaling of the multipole components in the @xmath70 transition .
.[table1 ] contributions from different combinations of multipoles to the integrated decay rate @xmath15 , in units of radians per second .
relativistic calculations have been performed for different hydrogen - like ions . [
cols="^,^,^,^ " , ] furthermore , as seen from table 1 and as implied by the non - uniform convergence of the integrals , the intermediate states with the energies @xmath71 lying between the energies of the initial and the final states give a _
finite _ contribution to the two - photon decay rate .
for the electric dipole ( @xmath41 ) transition in a neutral hydrogen atom , e.g. , a proper treatment of the intermediate @xmath55 and @xmath56 states leads to the decay rate of @xmath72 which is in agreement with the result of nonrelativistic calculations reported in ref .
however , when comparing our prediction with the theoretical data by cresser and co - workers @xcite a large discrepancy by about a factor of @xmath73 is observed .
the occurrence of the discrepancy is natural because the problematic virtual states with intermediate energies are treated differently in @xcite .
the two - photon decay of hydrogen - like ions has been re - investigated within the framework of relativistic quantum electrodynamics .
starting from first principles of this theory , we treat the total ( two - photon ) decay rate as the imaginary part of the relativistic two - loop self - energy .
the great advantage of this approach , which has its roots in field theory , is that it provides a simple and efficient route to handle the potentially problematic cases of those two - photon transitions from an excited into the ground state which pass intermediate states that can otherwise also be reached in one - photon cascades from the initial to the final states .
we found that those states with energies that lie between the energy of the initial and the final states , contribute a finite correction to the total two - photon decay rate .
taking into account this correction , we calculate the rates for the @xmath0 two - photon decay of neutral hydrogen as well as hydrogen - like ions .
our results are in a good agreement with nonrelativistic calculations for low @xmath25 ( see refs .
@xcite ) but show a significant deviation from the data by cresser and co - workers @xcite . our quantum electrodynamics approach , as discussed in the present paper , opens a way for a _
systematic _ theoretical analysis of the simultaneous , coherent two - photon emission from one - electron ( and many - electron ) atomic systems , even in cases where problematic intermediate states with an energy between the initial and final states give rise to resonance peaks in the photon energy distributions .
we stress here that a conceivable alternative approach to the removal of the formal infinities generated by the intermediate `` cascade '' states , which is based on the explicit removal of these states from the sums over @xmath74 and @xmath75 in eq .
( [ gamma_2_qed ] ) , gives rise to a number of conceptual problems , including gauge - noninvariance with respect to length and velocity gauges @xcite .
our approach is manifestly gauge invariant and also avoids problems connected with the identification of the infinitesimal parts @xmath76 in the propagator denominators in eq .
( [ gamma_2_qed ] ) with partial or total decay rates of the intermediate states : the @xmath77 parameters are free parameters which approach zero after all other operations , including the integrations over the photon energies , have been performed .
this operation leads to a finite result and corresponds , as explained in ref .
@xcite , to a _ partial _ removal of the problematic intermediate states from the sum over all virtual states involved in the two - photon process , albeit in a fully gauge - invariant manner .
in addition to its relevance for atomic physics , our approach may have a significant impact for astrophysical studies where a detailed knowledge of the ( properties of ) two - photon transitions is highly required for the analysis of cosmological hydrogen and helium recombination .
the contribution of two - photon processes to the recombination history represents an issue which has recently attracted substantial theoretical interest @xcite .
u.d.j . acknowledges support from deutsche forschungsgemeinschaft ( heisenberg program ) , and a.s .
acknowledges support from the helmholtz gemeinschaft ( nachwuchsgruppe vh
the authors acknowledge insightful discussions with z. harman regarding the sturmian decomposition of the dirac
coulomb green function as given in ref .
@xcite , and helpful discussions with p. indelicato on general aspects of the two - photon decay and associated resonances .
v. florescu , phys .
rev . a * 30
* , 2441 ( 1984 ) ; j. h. tung , x. m. ye , g. j. salamo , and f. t. chan , phys .
a * 30 * , 1175 ( 1984 ) ; v. florescu , s. patrascu , and o. stoican , phys . rev .
a * 36 * , 2155 ( 1987 ) . | based on quantum electrodynamics , we reexamine the two - photon decay of one - electron atoms .
special attention is paid to the calculation of the ( two - photon ) total decay rates which can be viewed as the imaginary part of the two - loop self - energy .
we argue that our approach can easily be applied to the cases with a virtual state having an intermediate energy between the initial and the final state of the decay process leading , thus , to the resonance peaks in the two - photon energy distribution . in order to illustrate our approach
, we obtain fully relativistic results , resolved into electric and magnetic multipole components , for the two - photon decay rates of the @xmath0 transition in neutral hydrogen as well as in various hydrogen - like ions . |
neural collective oscillations have been observed in very many context in brain circuits , ranging from ubiquitous @xmath1 oscillations to @xmath3 rhythm in the hippocampus .
the origin of these oscillations is commonly associated with the balance between excitation and inhibition in the network , while purely excitatory circuits are believed to lead to `` unstructured population bursts '' @xcite .
however , recent ` ex vivo' measurements performed on the rodent neocortex @xcite and hippocampus @xcite in the early stage of brain maturation reveal coherent activity patterns , such as _ giant depolarizing potentials_. these collective oscillations emerge despite the fact that the gaba transmitter has essentially an excitatory effect on immature neurons @xcite .
therefore , also in purely excitatory networks one can expect non trivial dynamics at a macroscopic level .
numerical and theoretical studies of collective motions in networks of simple spiking neurons have been mainly devoted to balanced excitatory - inhibitory configurations , e.g see @xcite and references therein .
only few studies focused on coherent periodic activity in fully coupled excitatory networks of leaky integrate - and - fire ( lif ) neurons .
these analysis revealed a regime characterized by a _ partial synchronization _ ( ps ) at the population level , while the single neurons perform quasi - periodic motions @xcite .
it has been shown that the ps regime is quite robust to perturbations , since it survives to moderate levels of noise or dilution @xcite . furthermore , in their recent study bonifazi et al .
@xcite found that the functional connectivity of developing hippocampal networks is characterized by a truncated power - law distribution of the out - degrees with exponent @xmath4 .
this scaling has been shown to hold over one / two decades , thus not ensuring a scale - free distribution for the links over all scales , but surely indicating the presence of a large number of hub neurons , namely cells characterized by a high connectivity . at early developmental stages of the brain , gabaergic hub interneurons ,
performing complex excitatory / shunting inibitory actions @xcite , seem to be responsible for the orchestration of the coherent activity of hippocampal networks @xcite .
the relevance of hubs in rendering a neural circuit extremely hyperexcitable has been also demonstrated in simulation studies of a realistic model of the epileptic rat dentate gyrus , even in the absence of a scale - free topology @xcite . motivated by these studies , but without attempting to reproduce the experimental results , we focus on a very preliminary question : to which extent is the macroscopic neural dynamics influenced by the average degree of connectivity of the neurons ?
our specific aim is to analyze the key ingredients leading to the onset of coherent activity , as opposed to asynchronous dynamics , for different network size and topology .
specifically , we consider the transition from an asynchronous regime to partial synchronization in the excitatory lif pulse coupled neural networks introduced by abbott and van vreeswijk @xcite . at variance with previous works ,
we consider dynamical evolution on random erds - renyi ( er ) networks with an average connectivity corresponding to that of a truncated power - law distribution with a decay exponent @xmath5 . this amounts to have an average connectivity which scales proportionally to @xmath6 with the number of neurons in the network . in the limit @xmath7 the massively connected network , where the connectivity is proportional to @xmath8 , is recovered ; while for @xmath9 a sparse network , where the average probability to have a link between two neurons vanishes in the thermodynamic limit @xcite , is retrieved . the topology of er networks is modified by varying the parameter @xmath1 in the interval @xmath10 $ ] ,
in particular as far as @xmath11 trees and cycles of any order are present in the network , while for @xmath7 complete subgraphs of increasing order appear in the system @xcite . in the paper
we will study how and to which extent these topological modifications influence the macroscopic dynamics of the network , with particular emphasis on the transition from asynchronous to ( partially ) synchronous collective dynamics .
the paper is organized as follows : the next section is devoted to the introduction of the neural model and of the indicators employed to characterize the dynamics of the network .
the phase diagram reporting the collective states emerging in our system is described in sect .
the influence of finite size effects on the coherent activity is analyzed in sect .
4 , while a characterization of the neural dynamics in term of maximal lyapunov exponent is reported in sect .
5 . a brief discussion of our results is outlined in sect . 6 .
we study a network of @xmath8 lif neurons with the membrane potential @xmath12 $ ] of the neuron @xmath13 evolving as : @xmath14 where @xmath15 is the suprathreshold dc current , @xmath16 the synaptic current .
whenever the neuron reaches the threshold @xmath17 , a pulse @xmath18 is instantaneously transmitted to all the connected post - synaptic neurons and the membrane potential of neuron @xmath13 is reset to @xmath19 .
the synaptic current can be written as @xmath20 , with @xmath21 representing the synaptic excitatory strength while the field @xmath22 is the linear superposition of the pulses @xmath18 received by neuron @xmath13 in the past , in formula @xmath23 where @xmath24 is the number of pre - synaptic neurons connected to the neuron @xmath13 ( i.e. the in - degree of neuron @xmath13 ) and @xmath25 the causal @xmath26 function .
the connectivity matrix @xmath27 appearing in eq .
( [ eq : e0 ] ) has entry @xmath28 ( resp . 0 ) depending if the pre - synaptic neuron @xmath29 is connected ( resp . not connected ) to neuron @xmath13 and in general it is not symmetric . following van vreeswijk @xcite we assume , for the single pulse emitted at @xmath30 , the shape @xmath31 . the explicit equation ( [ eq : e0 ] )
can be thus rewritten as an implicit ordinary differential equation : @xmath32 the continuous time evolution of the network can be transformed in a discrete time event - driven map by integrating eq .
( [ eq : e ] ) from time @xmath33 to time @xmath34 , @xmath33 being the time immediately after the @xmath35-th spike emission . following olmi et al .
@xcite , the event - driven map read as : @xmath36 where @xmath37 is an auxilary variable , @xmath38 is a nonlinear function and @xmath39 is the interspike time interval .
this can be determined by solving the following implicit relationship @xmath40 \ , \ ] ] where @xmath41 identifies the neuron which will fire next at time @xmath34 by reaching the threshold value @xmath42 .
the evolution of the system is now modeled with a discrete time map of @xmath43 variables , @xmath44 .
in fact the degree of freedom associated with the membrane potential of the firing neuron has been removed by the implementation of the poincar - section , leading to the event driven map .
more details on the model are reported in ref .
@xcite ; however at variance with that study the pulse amplitudes , appearing in eqs .
( [ eq : e0]),([eq : e ] ) and ( [ qq ] ) , are normalized by the in - degree @xmath24 of neuron @xmath13 and not by the total number of neurons @xmath8 .
the model parameters were fixed as @xmath45 , @xmath46 and @xmath47 , in order to ensure the emergence of a ps regime in the corresponding fully coupled network @xcite .
our choice for the connectivity matrix has been mainly motivated by the results obtained for `` ex vivo '' cells by bonifazi et al .
in particular , these authors have shown that in developing hippocampal networks the functional connectivity is distributed according to power - law distribution , at least over one / two decades . the distribution of links @xmath48 per neuron is thus given by @xmath49 , where @xmath50 is a normalization constant .
therefore , for a finite network made of @xmath8 neurons the average connectivity for the truncated ( power - law ) distribution is : @xmath51 \quad .
\label{kave}\ ] ] furthermore , bonifazi et al @xcite have measured quite low values for the exponent @xmath1 , namely @xmath52 , suggesting the existence of a large number of highly connected neurons , hubs , in the network .
the role of hub neurons in orchestrating the level of synchrony in living brain networks has been clearly demonstrated in @xcite . in spite of these findings
the underlying network topology is maybe less important than the actual number of highly connected neurons , as suggested by morgan and soltesz @xcite . in order to single out the effect of the average number of connections on the network dynamics we decided to limit our analysis to erds - renyi ( er ) random network @xcite , but with an average connectivity given by eq .
( [ kave ] ) .
in particular , we considered a directed er random graph , where the distribution of links is well approximated by a poisson distribution @xcite , namely : @xmath53 according to eq .
( [ poisson ] ) the degree distribution is completely defined once the value of @xmath0 is given .
as a matter of fact , by choosing for @xmath0 the expression ( [ kave ] ) , the probability of existence of an unidirectional link connecting neuron @xmath29 to @xmath13 ( i.e. the probability to have @xmath54 ) is : @xmath55 \quad .
\label{prob}\ ] ] in the limit @xmath7 the massively connected network is recovered @xcite , since the average connectivity @xmath56 is proportional to the system size and @xmath57 is , apart finite - size corrections , constant and coincident with @xmath50 . for @xmath58 ( resp .
@xmath59 ) the average number of synaptic inputs per neuron will grow ( resp .
decrease ) with @xmath8 , in the limiting case @xmath60 a sparse network will be essentially recovered @xcite since @xmath61 will vary in a limited manner with respect to the system size . indeed , by varying @xmath8 by three orders of magnitude from 100 to 100,000 the value of @xmath0 will modify from 3.7 to 9.2 with @xmath62 . in the following
, we studied networks of various sizes @xmath8 , ranging from @xmath63 to @xmath64 , for different @xmath1-values in the interval @xmath10 $ ] .
the value @xmath50 is usually fixed to 0.8 , apart for @xmath65 ( constant probability case ) where the dependence on @xmath50 is examined in details in section 3 .
let us stress that in the present study the distributions of pre - synaptic ( in - degree ) and of post - synaptic ( out - degree ) connections are identical , and this is guaranteed by the above outlined procedure to determine unidirectional links for er networks .
in what follows , we consider two different ways to select the random connectivity matrix : in the first case the synaptic connections are randomly chosen and are constant in time ( _ quenched disorder _ ) ; while for the second procedure the neurons receiving the excitatory pulse are randomly selected each time a neuron fires ( _ annealed disorder _ ) .
the latter choice can be justified from a physiological point of view by the fact that the synaptic transmission of signals is an _ unreliable _ process @xcite .
it should be noticed that in the annealed case , since the network modifies in time , the pulse amplitudes are normalized by the average in - degree @xmath0 and not by @xmath24 as in the quenched case . in order to perform a macroscopic characterization of the dynamical states of the network we exploit the average fields : @xmath66 where the fields have been also rescaled by the synaptic strength as done in ref .
@xcite . as a measure of the level of homogeneity among the neurons of the network , we consider the standard deviation @xmath67 of the fields @xmath68 acting on the single neurons @xmath69 for completely homogeneous systems , such as globally coupled networks , @xmath70 and @xmath71 . the degree of synchronization among the neurons is quantified by the order parameter usually employed in the context of phase oscillators @xcite @xmath72 where @xmath73 is the phase of the @xmath74 neuron , that can be properly defined as a ( suitably scaled ) time variable @xcite , @xmath75 , where @xmath76 indicates the time of the last spike emitted by the @xmath74 neuron , while @xmath77 is the @xmath35-th interspike interval associated to the neuron @xmath78 which was the last to fire in the network @xcite
. a non zero - value of @xmath79 represents an indication of _ partial _
synchronization among the neurons , while a vanishingly small @xmath80 is observable for asynchronous states in finite systems @xcite .
the dynamical microscopic instabilities of a system can be characterized in terms of the maximal lyapunov exponent @xmath81 : a positive @xmath81 being a measure of the degree of chaoticity of the considered system . in particular , we have employed the standard method developed in ref .
@xcite to estimate the maximal lyapunov exponent by following the evolution of an infinitesimal perturbation to a reference trajectory .
in this model two different macroscopic regimes can be observed : the asynchronous state ( as ) and the partial synchronization ( ps ) . as is characterized by an incoherent dynamics of the neurons in the network leading to a spot - like macroscopic attractor in the @xmath82 plane and an almost constant average field @xmath83 @xcite , while to the coherent ps regime corresponds a closed curve attractor @xcite and a periodic behavior of @xmath83 in time as shown in fig .
[ fig:1]a and b. the incoherent and coherent neural dynamics can be clearly appreciated , in the two regimes , also at microscopic level by examining the corresponding raster plots reported in fig .
[ fig:1]c and [ fig:1]d . as a first aspect
, we will investigate the occurrence of as and ps for finite networks and different average connectivity @xmath0 . in order to distinguish the two regimes we have examined the macroscopic attractor shape , the extrema values of the average field @xmath83 , and the synchronization parameter @xmath79 as a function of @xmath8
. plane .
panel b : the average field @xmath83 as a function of time .
panel c and d : raster plots .
the data refer to er networks with @xmath84 , quenched disorder , and @xmath85 , the black ( resp .
red ) symbols correspond to as observable for @xmath86 ( resp .
ps for @xmath87 ) . ,
title="fig:",scaledwidth=40.0% ] plane .
panel b : the average field @xmath83 as a function of time .
panel c and d : raster plots .
the data refer to er networks with @xmath84 , quenched disorder , and @xmath85 , the black ( resp .
red ) symbols correspond to as observable for @xmath86 ( resp .
ps for @xmath87 ) .
, title="fig:",scaledwidth=40.0% ] as an initial reference study we consider er networks with unidirectional links chosen at random with a constant probability ( cp ) @xmath50 for any network size @xmath8 .
this amounts to consider the limiting case @xmath7 and to have an average connectivity scaling linearly with the size , i.e. @xmath88 .
let us firstly examine the minima and maxima of @xmath83 of a network of size @xmath8 by varying the probability @xmath50 between 0 and 1 , as shown in fig .
[ fig:2]a for @xmath85 . at low probability
one has as while the ps regime emerges only for sufficiently large @xmath50 , both in the quenched and annealed case .
this result is related to the fact that the presence of noise reduces the coherence needed to observe the ps regime @xcite . as a matter of fact increasing @xmath50 ( i.e. diminishing the number of broken links in the network ) the attractor size increases and finally reaches the fully coupled result .
the degree of coherence can be measured in terms of the average synchronization indicator @xmath89 . as shown in fig .
[ fig:2]b the system coherence steadily increases with @xmath50 , except in the as regime where @xmath90 , apart for finite size fluctuations . as a function of @xmath50 for @xmath85 ,
the circles joined by solid lines refer to the annealed disorder , while the stars connected by the dashed line to the quenched case .
the dot - dashed ( red ) lines indicate the fully coupled results ( corresponding to @xmath91 ) .
b ) synchronization indicator @xmath89 averaged over time as a function of @xmath50 for the quenched case with @xmath92 .
the data refer to er networks with constant probability ( cp ) and have been estimated , after discarding a transient of @xmath93 spikes , by averaging over a train of @xmath94 spikes .
, title="fig:",scaledwidth=40.0% ] as a function of @xmath50 for @xmath85 , the circles joined by solid lines refer to the annealed disorder , while the stars connected by the dashed line to the quenched case .
the dot - dashed ( red ) lines indicate the fully coupled results ( corresponding to @xmath91 ) .
b ) synchronization indicator @xmath89 averaged over time as a function of @xmath50 for the quenched case with @xmath92 .
the data refer to er networks with constant probability ( cp ) and have been estimated , after discarding a transient of @xmath93 spikes , by averaging over a train of @xmath94 spikes .
, title="fig:",scaledwidth=40.0% ] let us now report the phase diagram for the macroscopic activity of the network in the @xmath95 plane , for both annealed and quenched disorder .
increasing the average connectivity , keeping the system size fixed , leads to a transition from as to ps regimes ( see fig . [ fig:3 ] ) .
the transition occurs at a critical average connectivity @xmath2 , indicates by the asteriskes connected by a black solid line in fig .
[ fig:3 ] , which for low @xmath8 increases steadily with @xmath8 , but eventually saturates for @xmath96 to an asymptotically constant value which depends on the noise realization : namely , @xmath97 for the quenched case and @xmath98 in the annealed one @xcite .
plane : a ) annealed disorder and b ) quenched disorder .
the ( black ) asteriskes connected by the solid ( black ) line correspond to the transition values @xmath2 from as to ps regime estimated for er networks with cp .
the other symbols refer to er with @xmath99 : solid ( resp .
empty ) symbols individuate as ( resp .
ps ) states . in particular , ( blue ) squares refer to @xmath100 , ( red ) circles to @xmath101 , ( green ) diamonds to @xmath102 , and ( magenta ) triangles to @xmath103 .
the reported data are relative to the state of the network after discarding transients ranging from @xmath104 spikes at the smaller sizes to @xmath105 spikes for the larger networks .
, title="fig:",scaledwidth=40.0% ] plane : a ) annealed disorder and b ) quenched disorder .
the ( black ) asteriskes connected by the solid ( black ) line correspond to the transition values @xmath2 from as to ps regime estimated for er networks with cp .
the other symbols refer to er with @xmath99 : solid ( resp .
empty ) symbols individuate as ( resp .
ps ) states . in particular , ( blue ) squares refer to @xmath100 , ( red ) circles to @xmath101 , ( green ) diamonds to @xmath102 , and ( magenta ) triangles to @xmath103 .
the reported data are relative to the state of the network after discarding transients ranging from @xmath104 spikes at the smaller sizes to @xmath105 spikes for the larger networks .
, title="fig:",scaledwidth=40.0% ] to verify the generality of these results we investigate er networks with @xmath1-dependent probability .
in particular , we have estimated , for system sizes in the range @xmath106 , the macroscopic attractors for various @xmath1-values ( namely , @xmath100 , 1.3 , 1.5 and 1.7 ) , after discarding long transient periods . for small system sizes
the network is in the as regime which is characterized by a spot - like attractor in the @xmath107-plane . for larger number of neurons , ps emerges in the system characterized by closed curve ( macroscopic ) attractors .
furthermore , similarly to the results reported in ref .
@xcite , increasing @xmath8 the curves tend to an asymptotic shape , corresponding to the fully coupled attractor , while fluctuations diminish . to exemplify this point various macroscopic attractors for @xmath101 and annealed disorder are reported in fig .
[ fig:3bis]a .
as already reported for the cp networks the systems with annealed disorder converge more rapidly with @xmath8 towards the asymptotic fully coupled attractor with respect to the quenched case , as shown in fig .
[ fig:3bis]b for @xmath100 , 1.3 , 1.5 .
increasing @xmath1 we observe that the transition from as to ps occurs at larger and larger system size , both for annealed and quenched disorder . the results for @xmath108
are not shown in fig .
[ fig:3bis]b , since for all the examined network sizes ( up to @xmath64 ) @xmath83 extrema coincide within the error bar , indicating that the system is in the as regime @xcite . as ( resp .
ps ) regimes are reported in the phase diagram displayed in fig .
[ fig:3 ] as filled ( resp .
empty ) symbols for the investigated @xmath1-values . the critical line @xmath109 ( indicated by the solid black line in fig . [ fig:3 ] ) denoting the transition from as to ps coincides with that determined previously for cp networks .
we can thus safely affirm that the dynamical regimes of the er networks depend , at a macroscopic level , simply on the average connectivity , once the system size @xmath8 is fixed .
this could be expected from the fact that for er networks the distribution of links per neuron is completely determined by @xmath0 ( see eq .
( [ poisson ] ) ) .
however , the independence of @xmath2 from @xmath8 at @xmath110 is unexpected and suggests that coherent behaviors , like ps regimes , can be observed in networks of any size for vanishingly small relative connectivity @xmath111 .
plane for a er networks with @xmath101 and annealed disorder , the curves from the interior to the exterior corresponds to increasing system sizes , from @xmath85 to 25,600 .
the most external ( black ) curve refers to a fully coupled network with @xmath92 .
b ) minima and maxima values of the average field @xmath83 as a function of @xmath8 for various @xmath1-values : namely , ( black ) circles @xmath100 , ( red ) triangles @xmath101 and ( blue ) squares @xmath102 .
the empty ( resp . filled ) symbols refer to annealed ( resp .
quenched ) disorder .
the dot - dashed ( magenta ) lines indicate the fully coupled values .
, title="fig:",scaledwidth=40.0% ] plane for a er networks with @xmath101 and annealed disorder , the curves from the interior to the exterior corresponds to increasing system sizes , from @xmath85 to 25,600 .
the most external ( black ) curve refers to a fully coupled network with @xmath92 .
b ) minima and maxima values of the average field @xmath83 as a function of @xmath8 for various @xmath1-values : namely , ( black ) circles @xmath100 , ( red ) triangles @xmath101 and ( blue ) squares @xmath102 . the empty ( resp
. filled ) symbols refer to annealed ( resp .
quenched ) disorder .
the dot - dashed ( magenta ) lines indicate the fully coupled values .
, title="fig:",scaledwidth=40.0% ]
we observe that for any considered @xmath1 the fields @xmath112 associated with the different neurons tend to synchronize for increasing @xmath8 .
therefore in the thermodynamic limit ( @xmath113 ) disordered networks tend to behave as fully coupled ones , where all the neurons are equivalent and a single field is sufficient to describe the macroscopic evolution of the system . in order to quantify the level of homogeneity among the various neurons , we measured the standard deviation ( eq .
( [ eq : sigma ] ) ) relative to the fluctuations of the different local fields @xmath114 with respect to the average field @xmath83 . in fig .
[ fig:4]a we plot the time average of the standard deviation , @xmath115 , for annealed disorder and various @xmath1 values .
we observe a power law decay @xmath116 , where the exponent @xmath117 depends on the @xmath1-parameter as @xmath118 ( see fig .
[ fig:4]b ) .
furthermore , for the liming case @xmath119 the decay of @xmath115 is consistent with a scaling @xmath120 as displayed in the inset of fig .
[ fig:4]a .
alltogether , the reported dependencies suggest the following relationship to hold : @xmath121 thus fields fluctuations are driven by the average in - degree value irrespectively of the total number of neurons .
these results confirm once more that in the limit @xmath113 the neural field dynamics converges to that of homogeneous networks , for both quenched and annealed disorder .
the relationship among @xmath115 and the average connectivity reported in eq .
( [ sigma_scal ] ) clearly indicates that for a sparse network , with constant average connectivity , @xmath115 will remain finite even in the thermodynamic limit .
therefore , we can conclude that a sparse network can not be ever reduced to a fully coupled one by simply rescaling the synaptic coupling as done in ref .
@xcite , even for very large systems . versus the system size @xmath8 for annealed disorder and various @xmath1-values :
@xmath100 ( blue ) squares , @xmath122 ( red ) circles , @xmath123 ( green ) diamonds and @xmath124 ( magenta ) triangles .
the dashed line represents best fits with a power - law @xmath125 to the reported data .
the data in the inset ( black asterisks ) refers to a @xmath119 , the dashed line is a guide for the eyes .
b ) power - law exponents @xmath117 for annealed ( black circles ) and quenched ( red triangles ) disorder in the network as a function of the parameter @xmath1 .
the dashed ( blue ) line refers to the linear law @xmath118 .
the reported data have been estimated by averaging over trains made of @xmath126 spikes , after discarding transients of @xmath127 spikes . , title="fig:",scaledwidth=40.0% ] versus the system size @xmath8 for annealed disorder and various @xmath1-values : @xmath100 ( blue ) squares , @xmath122 ( red ) circles , @xmath123 ( green ) diamonds and @xmath124 ( magenta ) triangles .
the dashed line represents best fits with a power - law @xmath125 to the reported data .
the data in the inset ( black asterisks ) refers to a @xmath119 , the dashed line is a guide for the eyes .
b ) power - law exponents @xmath117 for annealed ( black circles ) and quenched ( red triangles ) disorder in the network as a function of the parameter @xmath1 .
the dashed ( blue ) line refers to the linear law @xmath118 .
the reported data have been estimated by averaging over trains made of @xmath126 spikes , after discarding transients of @xmath127 spikes .
, title="fig:",scaledwidth=40.0% ]
homogeneous fully connected pulse coupled networks exhibit regular dynamics @xcite . in particular , for excitatory network and finite pulses the as becomes a _ splay state _ characterized by all neurons spiking one after the other at regular intervals with the same frequency and by a constant mean field @xmath83 , while the ps regime becomes perfectly periodic at a macroscopic level @xcite .
the introduction of disorder in the network leads to irregularity in the dynamics of the single neurons , which are reflected also at a macroscopic level .
this kind of deterministic irregular behavior has been identified as _ weak chaos _ whenever the irregularity vanishes for sufficiently large system size @xcite .
the chaotic motion can be characterized in terms of the maximal lyapunov exponent @xmath81 : regular orbits have non positive exponents , while chaotic dynamics are associated with @xmath128 . for various @xmath1-values .
the data have been obtained by discarding a transient of the order of @xmath129 spikes and then by following the dynamics in the real and tangent space for an equivalent duration , moreover the data have been averaged over 3 to 5 different network realization with quenched disorder . , scaledwidth=40.0% ] for finite size networks chaotic dynamics is observed both for annealed and quenched disorder . however , for all @xmath1-values examined in this work , @xmath81 tends to decrease for a sufficiently large number of neurons in the network .
therefore we can safely affirm that for any er network with average connectivity given by eq .
( [ kave ] ) the neuronal dynamics is weakly chaotic ; i.e. the evolution will become completely regular for infinite networks .
numerical results for quenched disorder are reported in fig .
[ fig:5]a for various @xmath1-values .
the maximal lyapunov exponent exhibits clear power - law decays @xmath130 , with @xmath131 decreasing and eventually vanishing for @xmath9 ( see table [ tab1 ] ) .
these results generalize previous indications reported in olmi et al . for a specific realization of diluted network @xcite .
.power law exponents giving the decay of the maximal lyapunov @xmath132 for the data reported in fig .
[ fig:5]a for quenched disorder .
[ cols= " < , < " , ] considering networks with annealed disorder it must be underlined that for sufficiently large number of neurons and @xmath58 , the maximal lyapunov reveals a tendency to decrease .
however , clear scaling laws can not be inferred for all the examined @xmath1-values on the range of affordable system sizes , namely @xmath133 .
larger network sizes are probably required with unreliable synapses to have clear scaling laws , due to the fact that the transition from as to ps occurs at critical system sizes within the investigated range .
on the contrary , in the quenched case once @xmath1 is fixed an unique regime is observable for almost all the investigated sizes . in particular , the network is always in an as for @xmath103 and 1.5 , while it is in the ps regime for @xmath100 and 1.3 for ( almost ) all the considered number of neurons .
these additional findings strengthen the above reported conclusions : finite size systems are weakly chaotic for any considered dynamical regime . according to the results reported above
, we expect that for sparse networks the maximal lyapunov exponent will eventually saturate to some constant value , apart for possible logarithmic corrections .
thus , sparse networks should remain chaotic for any system size , paralleling the behavior of the microscopic fluctuations shown in the previous section
. therefore microscopic inhomogeneities and chaotic behavior appear as deeply related .
collective periodic oscillations in excitatory er networks can be observed only above a critical average in - degree .
this latter quantity saturates to a constant value for networks with a sufficiently large number of neurons , thus suggesting that the key ingredient responsible for the emergence of collective behaviors is the number of pre - synaptic neurons both for massively connected networks as well as for sparsely connected ones .
this result confirms and generalizes previous findings on the stability of the complete synchronized state for pulse - coupled hindmarsh - rose neurons @xcite .
furthermore , our results indicate that the minimal network size required to observe collective oscillations diverges with the exponent @xmath1 ruling the scaling of the average connectivity with the number of neurons .
the presence of annealed disorder in the network ( corresponding to unreliable synapses ) favors the emergence of coherent activity with respect to the quenched case ( associated with reliable synapses ) , since in the first case the asymptotic average connectivity required to observe collective oscillations is much smaller .
this is probably due to the fact that in the annealed situation each neuron is on average subjected to the same train of stimuli , while with quenched disorder the dynamics of each neuron depends heavily on its neighbors .
furthermore , for sufficiently large networks the macroscopic behavior observable with reliable or unreliable synapses becomes identical .
this seems to indicate that at the level of population dynamics the reliability or unreliability in the synaptic transmission can be irrelevant .
the average in - degree @xmath0 also controls the fluctuations among different neurons , being the fluctuations proportional to the inverse of the square root of @xmath0 .
therefore , for er networks with average in - degree proportional to any positive power of @xmath8 , the fluctuations will vanish in the limit @xmath113 , leading to a homogeneous collective behavior analogous to that of fully connected networks .
on the contrary , inhomogeneities among neurons will persist at any system size in sparse networks .
recent experimental results on the intact neuronal network of the barrel cortex of anesthetized rats seems to clearly suggest that the dynamics of this system is chaotic @xcite .
this result poses severe questions about the possibility of reliable neural coding .
the evolution of our models is chaotic for any finite networks .
however , in presence of coherent periodic activity , the chaoticity present in the system is not so strong to destroy the average collective motion
. thus the trial - to - trial variability induced by chaos does not prevent the possibility of the network to encode information .
the information can be coded in some property associated to the global oscillations of the network which are robust to local fluctuations , and this _ collective coding _ can represent an alternative to the dilemma rate versus spike timing coding @xcite .
the level of chaos in the examined networks decreases with @xmath8 and the dynamics becomes regular in the thermodynamic limit .
however , the decrease of the maximal lyapunov exponent with @xmath8 slows down dramatically for @xmath9 , suggesting an erratic asymptotic behavior for sparse networks .
our results represent only a first step in the analysis of what matters in the network topology for the emergence of coherent neural dynamics . in future works
, our findings should be critically verified for more complex topologies , like scale - free and small - world , and the influence of other ingredients , like the asymmetry in the in - degree and out - degree distributions @xcite , should be also addressed .
we acknowledge useful discussions with a. politi , a. pikovsky , p. bonifazi and m. timme .
this research project is part of the activity of the joint italian - israeli laboratory on neuroscience funded by the italian ministry of foreign affairs and it has been partially realized thanks to the support of cineca through the italian super computing resource allocation ( iscra ) programme , project ecosfnn .
one of us ( lt ) thanks hpc - europa2 programme for supporting him during the period spent in the network dynamics group of the max planck institute for dynamics and self organization in gttingen ( germany ) .
allene c , cattani a , ackman jb , bonifazi p , aniksztejn l , ben - ari y and cossart r , sequential generation of two distinct synapse - driven network patterns in developing neocortex , the journal of neuroscience , 26 , 12851 - 12863 ( 2008 ) .
bonifazi p , goldin m , picardo ma , jorquera i , cattani a , bianconi g , represa a , ben - ari y and cossart r , gabaergic hub neurons orchestrate synchrony in developing hippocampal networks , science , 326 , 1419 - 1424 ( 2009 ) .
morgan rj and soltesz i , nonrandom connectivity of the epileptic dentate gyrus predicts a major role for neuronal hubs in seizures , proceedings of the national academy of sciences usa , 105 , 6179 - 6184 ( 2008 ) .
shimada i and nagashima t , a numerical npproach to ergodic problem of dissipative dynamical systems , prog . theor .
, 61 , 1605 ( 1979 ) ; benettin g , galgani l , giorgilli a and strelcyn jm , lyapunov characteristic exponents for smooth dynamical systems and for hamiltonian systems ; a method for computing all of them .
part ii : numerical application , meccanica , 15 , 21 ( 1980 ) . in the original definition reported in @xcite , @xmath134 ,
thus ensuring that the phase is bounded between 0 and @xmath135 .
however this definition requires to memorize all the spiking events within a certain time interval and then estimate the phases and @xmath136 over such time span .
our definition does not guaratee that the phase is strictly bounded within the @xmath137$]-interval , but it allows to estimate @xmath136 during the simulation run without memorizing any event .
indeed we have verified that , at least in the studied cases , no peculiar differences are observable by employing the two phase definitions and this is probably due to the fact that the dynamics of all the neurons in the networks are statistically equivalent .
we have verified that the above reported scenario is not modified by considering fully homogeneous connectivity with @xmath138 , the only difference concerns the asymptotic value @xmath139 at which the transition from as to ps occurs . as a matter of fact
@xmath139 is slightly smaller than for er distributions due to the absence of fluctuations among the in - degree values in the homogeneous case .
a simple estimation of the minimal network size needed to observe ps for generic @xmath1-distributions can be obtained by inverting eq .
( [ kave ] ) for the asymptotic values @xmath139 , namely @xmath140^{1/(2-\gamma ) } \quad ; \label{ncrit}\ ] ] for @xmath103 this leads to @xmath141 ( resp .
@xmath142 ) in the quenched ( resp .
annealed ) case .
this confirms that coherent activity is unobservable for @xmath103 with our computational resources for the chosen coupling parameter , namely @xmath46 . | in this article we investigate the role of connectivity in promoting coherent activity in excitatory neural networks . in particular , we would like to understand if the onset of collective oscillations can be related to a minimal average connectivity and how this critical connectivity depends on the number of neurons in the networks .
for these purpouses , we consider an excitatory random network of leaky integrate - and - fire pulse coupled neurons .
the neurons are connected as in a directed erds - renyi graph with average connectivity @xmath0 scaling as a power law with the number of neurons in the network .
the scaling is controlled by a parameter @xmath1 , which allows to pass from massively connected to sparse networks and therefore to modify the topology of the system . at a macroscopic level
we observe two distinct dynamical phases : an asynchronous state ( as ) corresponding to a desynchronized dynamics of the neurons and a partial synchronization ( ps ) regime associated with a coherent periodic activity of the network . at low connectivity
the system is in an asynchronous state , while ps emerges above a certain critical average connectivity @xmath2 . for sufficiently large networks , @xmath2 saturates to a constant value suggesting that a minimal average connectivity is sufficient to observe coherent activity in systems of any size irrespectively of the kind of considered network : sparse or massively connected .
however , this value depends on the nature of the synapses : reliable or unreliable .
for unreliable synapses the critical value required to observe the onset of macroscopic behaviors is noticeably smaller than for reliable synaptic transmission . due to the disorder present in the system , for finite number of neurons we have inhomogeneities in the neuronal behaviors , inducing a weak form of chaos , which vanishes in the thermodynamic limit .
in such a limit the disordered systems exhibit regular ( non chaotic ) dynamics and their properties correspond to that of a homogeneous fully connected network for any @xmath1-value .
apart for the peculiar exception of sparse networks , which remain intrinsically inhomogeneous at any system size . |
the dirac oscillator @xcite has attracted considerable attention both due to its simple formulation and its analytical solutions . yet
we may ask , if we can go beyond the single particle problem and conserve some of that elegance and simplicity . a plausible scenario in particle physics
is that of a dirac particle bound by an oscillator and interacting with an external ( free ) non - abelian field of mesons modeled as particles with finite radius @xcite . specifically , if the internal group of the interaction is chosen to be @xmath2 and the non - local potential dependence is linear in the position and momentum operators , the structure which leads to integrability is preserved .
this linearity can also be viewed as an approximation to a more complicated field .
the central subject of this paper will thus be to develop this scenario for one , two and three dimensional dirac oscillators . to achieve this
we shall reformulate moshinsky s approach in a matrix representation , where we shall see that a simple transformation will take the hamilonian formulation of the dirac oscillator to a block - diagonal form with @xmath0 blocks .
this representation is not manifestly covariant due to its hamiltonian form .
the equivalence which follows from the spectral theorem is confirmed by the fact that we recover the well - known solutions .
the coupling to the field , which carries isospin , will necessarily lead to @xmath3 blocks , which can still be solved exactly .
the solvable models are obtained by choosing the coupling in such a way , that the blocks remain uncoupled .
we shall show that several interesting models can be derived in this way , and that the two - dimensional case has a special role as compared to the one and three - dimensional cases . a covariant formulation of the different cases is also given .
once we obtain the exact eigen functions and eigenvalues we can study dynamics .
several cases of interest are studied by choosing particular values for the parameters of the model .
for instance , we can have a coupling to the field including explicit dependence on @xmath4-matrices ( pauli coupling ) .
the conservation of isospin + fermionic charge in the dynamics can be produced by restricting this example .
the case of vanishing coupling with the external field except for its static part is included as the simplest extension of the dirac oscillator .
the absence of explicit fermionic rest mass in the hamiltonian is also interesting in the sense of mass generation due to the external field . by way of example and as an extension to the fidelity studies presented for the dirac oscillator in @xcite we take advantage of the additional degree of freedom and consider the evolution of entanglement between the oscillator and the field or , equivalently , the decoherence of the oscillator due to the field .
in section two we will review the dirac oscilator in one , two and three dimensions in a matrix representation and show , that it is solved by casting it in block - diagonal form , the blocks being two - dimensional .
this will not only serve as a reminder of previous work and fix our notation .
it will also set the stage for the generalization given in the following section .
there we will implement the coupling to the isospin field and show that this can be done such that the block size is increased only to four .
this leaves us again with solvable models . in section four
we give the covariant formulation of these models and in section five an in - depth discussion of the one - dimensional case .
to get a feeling of the solvable dynamics we have found , we calculate in section six the evolution of entanglement between the field and the dirac oscillator .
we shall end by drawing some conclusions and giving an outlook .
the dirac oscillator @xcite was originally proposed as a way to introduce a linear potential @xcite which preserved the solvability of the resulting dirac equation .
we review some of the known results of this system by using a new notation .
the treatment to be presented is related to the algebraic structure of the dirac oscillator hamiltonians in 1 , 2 and 3 dimensions .
the relation of this structure with the solutions of the associated eigenvalue problem is discussed .
this should be helpful when we consider solvable extensions of this system .
let us start with the three dimensional case , originally considered by moshinsky and szczepaniak .
the hamiltonian is h= ( p + i ) + m [ 0.1 ] where we have adopted units such that the rest mass and the frequency satisfy @xmath5 and @xmath6 .
we use a representation of the dirac matrices given by = ( cc 0 & i + -i & 0 ) , = ( cc 1_2 & 0 + 0 & - 1_2 ) .
[ 0.2 ] defining the rising and lowering operators which act on big and small components of spinors in the form _ + = ( cc 0 & 1_2 + 0 & 0 ) = _ + 1_2 , _ - = ( _ + ) ^ , _ 3=[0.3 ] and using creation and anhilation operators @xmath7 for the oscillator variables , the hamiltonian ( [ 0.1 ] ) can be written in @xmath8 matrix form as h = ( cc 1_2 m & ^ + & - 1_2 m ) [ 0.4 ] or in algebraic form as h= _ + + _ - ^ + m _ 3 , [ 0.5 ] where the spin of the particle is given by @xmath9 . in ( [ 0.5 ] ) we have made explicit the separation of the spin and the generators @xmath10 of @xmath2 acting on big and small components of spinors ( or particle and anti - particle solutions under the appropriate rotation ) .
we shall refer to @xmath11 as @xmath12spin .
it is worth to mention that the projection of the @xmath12spin operator @xmath13 is not only related to quasi - relativistic and ultrarelativistic components of states .
when combined with oscillator operators , the @xmath12spin projection corresponds to states with positive and negative energies .
this can be seen from the foldy - wouthuysen transformation of this problem @xcite .
the solutions of the stationary schroedinger equation associated to ( [ 0.5 ] ) were originally obtained in @xcite by writing ( cc 1_2 m & ^ + & - 1_2 m ) ( c _ 1 + _ 2 ) = e ( c _ 1 + _ 2 ) [ 0.6 ] and recognizing that the second row yields @xmath14 .
this in turn can be replaced in the first row of ( [ 0.6 ] ) to give ( m-2e + [ ( m - e)/(m+e ) ] ( ) ( )^ ) _ 1 = 0 .
[ 0.61 ] all this can be derived in a simple way by squaring the hamiltonian in ( [ 0.6 ] ) , ( cc ( ^ ) ( ) + m^2 & 0 + 0 & ( ) ( ^)+ m^2 ) ( c _ 1 + _ 2 ) = e^2 ( c _ 1 + _ 2 ) [ 0.7 ] which explains the solvability of the problem , resulting in two decoupled equations where finite and infinite degeneracies appear . for a discussion of supersymmetry related to ( [ 0.6 ] ) , see @xcite .
we shall prefer , however , the hamiltonian in form ( [ 0.5 ] ) , since it is equally useful in deriving the solutions of ( [ 0.6 ] ) and it will prove helpful in further extensions of this system .
the dependence of @xmath15 on ladder operators as in ( [ 0.5 ] ) shows that @xmath16 is an invariant operator which commutes with the squared total angular momentum @xmath17 .
therefore , the eigenstates of @xmath18 with total angular momentum @xmath19 , projection @xmath20 and radial oscillator number @xmath21 can be used to evaluate the @xmath0 blocks of @xmath15 and obtain the stationary solutions .
we adopt the notation @xmath22 for oscillator states @xmath23 coupled to spin @xmath24 , where @xmath19 is the total angular momentum with projection @xmath20 , @xmath21 is the radial oscillator number , @xmath25 is the orbital angular momentum and @xmath26 its projection on the @xmath27 axis .
we denote @xmath12spinors by @xmath28 .
a pair of states with angular momentum @xmath19 and such that @xmath29 is given by another pair of states with the same angular momentum @xmath19 but with @xmath30 is the @xmath31 blocks of @xmath15 obtained from these states can be evaluated .
we denote their elements by @xmath32 for @xmath33 and @xmath34 for @xmath35 . from ( [ 0.7.1 ] ) and ( [ 0.7.2 ] ) one obtains the two following matrices h(j,2n+j-1)= ( cc -m & + & m ) , h(j,2n+j)= ( cc -m & + & m ) [ 0.7.3 ] leading to the well known energies @xmath36 and @xmath37 .
infinite and finite degeneracies come from these two blocks respectively . in connection with these degeneracies , it should be pointed out that a study of the spectrum of the dirac oscillator @xcite reveals a symmetry lie algebra given by a direct sum of @xmath38 ( a compact part with finite - dimensional unitary irreps ) and @xmath39 ( a non - compact part with infinite - dimensional unitary irreps ) .
the discussion on the algebraic structure above can be implemented directly in 1 and 2 spatial dimensions as well .
let us introduce the notation @xmath40 for the hamiltonian in @xmath41 dimensions .
take @xmath42 for the one - dimensional problem . for two dimensions ,
consider the chiral ladder operators a_r = a_1 + i a_2 , a_l = a_1 - ia_2 = ( a_r)^ * [ 0.8 ] with the properties @xmath43=[a_r,(a_l)^{*}]=0,[a_r , a^{\dagger}_r]=[a_l , a^{\dagger}_l]=4 $ ] .
the hamiltonians for one and two dimensional dirac oscillators in original form are h^(1)=_1 ( p+ix ) + m , [ 0.9 ] with @xmath44 and h^(2)= _ i=1,2_i(p_i+ir_i ) + m , [ 0.10 ] with @xmath45 .
these hamiltonians can be cast in algebraic form as h^(1)=_+ a + _ - a^ + m _ 3 [ 0.11 ] h^(2)=_+ a_r + _ - a^_r + m _ 3 [ 0.12 ] which exhibit some similarities with ( [ 0.5 ] ) .
both of them have a @xmath46 structure : the spin is absent in one spatial dimension and @xmath47 corresponds to @xmath12spin , while in two dimensions @xmath48 generates the @xmath49 spin .
the solvability can be viewed again as a consequence of the invariants i^(1)= a^ a + _ 3 [ 0.12.1 ] in one dimension , and i^(2 ) = a_r a^_r + _ 3 , j_3 = a_r a^_r - a_l a^_l + _ 3 [ 0.12.2 ] in two dimensions .
we must note , however , that the two dimensional case exhibits some peculiarities which do not appear in the other cases .
the conservation of angular momentum @xmath50 comes from the particular combination of @xmath51 and @xmath52 in @xmath53 , together with the absence of @xmath54 .
this absence is also responsible for the infinite degeneracy of all levels . on the other hand ,
the three dimensional example is manifestly invariant under rotations due to its dependence on @xmath55 and @xmath56 .
its infinite degeneracy comes from the interplay between the infinitely degenerate operator @xmath57 and @xmath12spin , dividing @xmath58 in two blocks as in ( [ 0.7 ] ) .
these observations will be used in the following section .
now that we have characterized the solvability of these systems as the reduction of @xmath40 to @xmath46 blocks , we may readily extend it to @xmath59 blocks by introducing appropriate potentials .
the corresponding lorentz invariance of such procedure will be discussed later .
consider a hermitean operator of the form @xmath60 as the potential to be introduced in the total hamiltonian .
one has @xmath61 , with @xmath62 given by the @xmath63dimensional dirac oscillator .
on physical grounds , this corresponds to a bound fermion perturbed by a momentum - dependent potential .
the scenario in which such type of potentials can arise is discussed with more detail in further sections .
for now , let us assume that @xmath64 may come from either a scalar theory interacting with the bound fermion through yukawa couplings , or as the scalar component of a vector potential introduced by the minimal coupling prescription .
these two possibilities are not equivalent
. we may introduce also an internal group for this field , for example the @xmath2 associated to isospin in the yukawa theory or as the gauge group of a non - abelian field . in either case we consider the fundamental representation ( doublets ) .
in the three dimensional case , let us impose full rotational invariance of this field in order to preserve integrability , @xmath65=0 $ ] .
in general , we shall see that a wide class of operators @xmath60 allows to retain not only integrability , but also solvability of the resulting model . to simplify the discussion , we linearize the field in the variables @xmath66 ( this approximation will be discussed in section 4 ) .
we simply have = _ 0 + _ 1 + ^_1 ^ [ 01.1 ] where @xmath67 are constant operators .
we denote by @xmath68 the vector of pauli matrices in isospin space and the corresponding ladder operators by @xmath69 . choosing @xmath70 properly , we may write = ( t_+ + t_- ^ + t_3 ) [ 01.2 ] with @xmath4 a real parameter . we may also write the more general = ( a+_3 b ) ( t_+ + t_-
^ + t_3 ) [ 01.3 ] parametrizing the two possibilities for @xmath64 mentioned above through the real constants @xmath71 . for @xmath72
we have a field which corrects the rest mass of the fermion , while @xmath73 represents the @xmath2 potential @xmath74 . see section 4 .
lower dimensional examples should follow the same pattern to preserve integrability . thus , we propose the extensions h^(1)= _ + a + _ - a^ + m _ 3 + ( a+_3 b ) ( t_+ a + t _ - a^ + t_3 ) [ 01.41 ] h^(2)= _ + a_r + _ - a^_r + m _ 3 + ( a+_3 b ) ( t_+ a_r + t_-
a^_r + t_3 ) [ 01.42 ] h^(3)= _ + + _ - ^ + m _ 3 + ( a+_3 b ) ( t_+ + t_- ^ + t_3 ) . [ 01.43 ] with these extensions , it is evident that the new invariants for one , two and three dimensions are i^(1)= a^ a + _ 3 + t_3 [ 01.5 ] i^(2 ) = a_r a^_r + _ 3 , j_3 + t_3 = a_r a^_r - a_l a^_l+ _ 3 + t_3 [ 01.6 ] i^(3)= ^ + _ 3 + t_3 , j = ^ + .
[ 01.7 ] before discussing the solutions of the schroedinger equations associated to ( [ 01.41],[01.42],[01.43 ] ) , our new hamiltonians deserve some comments .
it is important to note that the solvability of the resulting schroedinger equation for these hamiltonians resides completely on the existence of @xmath75 ( which in this case means superintegrability ) .
for instance , in the three dimensional case one may consider any potential of the form @xmath76 where @xmath77 admits a power expansion .
evidently , @xmath78=0 $ ]
. a suitable group of states can be used to evaluate the @xmath8 blocks of @xmath15 .
we describe this procedure in subsection 3.1 , restricting ourselves to the linear case for simplicity . despite the fact that the three extensions ( [ 01.41 ] , [ 01.42 ] , [ 01.43 ] ) exhibit a similar structure , the two dimensional case is , as before , special . in our strive to preserve integrability ,
we have introduced a field @xmath64 which depends only on @xmath52 and @xmath79 , leaving @xmath80 and @xmath81 out of the game .
the main consequence of this construction resides in the permanence of the infinite degeneracy of the system ( energies do not depend on @xmath82 ) , as well as a modification of the conserved angular momentum @xmath50 by the addition of isospin ( second equality in ( [ 01.6 ] ) ) .
the special nature of the original @xmath53 is responsible for restricting the type of linear extensions and the resulting spectra .
in contrast , the three dimensional extension breaks the infinite degeneracy through @xmath83 and conserves the angular momenta @xmath84 without modification . we discuss this through the eigenstates and energies of @xmath85 in the following section .
now we turn to the problem of finding the eigenstates of @xmath85 . for this
, we simply consider states which satisfy @xmath86 .
these are four in number
. then we proceed to evaluate the @xmath8 matrix @xmath87 .
introducing subscripts @xmath88 and @xmath89 for @xmath12spinors and isospinors , we write such states as @xmath21 is the oscillator radial number , @xmath19 is the total angular momentum and @xmath20 its projection in the @xmath27 axis .
these are eigenstates of @xmath90 with eigenvalue @xmath91 .
the resulting @xmath92 blocks of @xmath15 with elements @xmath93 are ( cccc -m-(a - b ) & ( a - b ) & - & 0 + ( a - b ) & -m+(a - b ) & 0 & + - & 0 & m-(a+b ) & ( a+b ) + 0 & & ( a+b ) & m+(a+b ) ) [ 02.2 ] and the secular equation @xmath94 can be solved explicitly using the formula for the roots of a quartic polynomial .
however , we shall contempt ourselves with the explicit solutions of the @xmath95 dimensional example in further sections , as the blocks of @xmath96 and @xmath85 are quite similar when @xmath97 .
it is also evident that the infinite degeneracy is now broken , since one can not reduce @xmath98 to smaller blocks where only @xmath21 appears .
the exception to this ocurrs when @xmath99 , which obviously recovers the usual dirac oscillator .
here we discuss the specific form of lorentz covariant external fields producing the solvable extension we have established . before proceeding , we should mention that the potential @xmath64 in one and two dimensions indicated in ( [ 01.41],[01.42 ] ) , is spin independent .
this makes our task easier , since @xmath64 can be directly related to the external field as shall be indicated .
the three dimensional case , however , should be treated separately due to the presence of @xmath100 in the potential @xmath64 considered in ( [ 01.43 ] ) . therefore , we shall consider three alternatives for generating @xmath64 : a parity violating coupling to an external non - local vector field ( mediated by @xmath101 ) , a pauli coupling to a non - local field tensor and a yukawa coupling to a non - local scalar field .
all of them shall have an internal symmetry group @xmath2 as indicated before .
the alternatives involving a non - local gauge field can be brought together when linear approximations of fields are considered , as we shall see .
here we use the metric tensor @xmath102 .
the original covariant formulation of the dirac oscillator @xcite considered a four vector @xmath103 such that , in some inertial frame , @xmath104 .
for the dirac oscillator of many particles , such vector was related to the center of mass of the system . in the present case , the meaning of @xmath103 is given through the introduction of the anomalous ( pauli ) coupling to an external field tensor .
the equation leading to the dirac oscillator hamiltonian ( [ 0.1 ] ) is = 0 , [ uno ] where @xmath105 and _ j = ( cc 0 & i _
j + i _ j & 0 ) , _ 0 = ( cc 1_2 & 0 + 0 & -1_2 ) .
[ uno.1 ] eq.([uno ] ) can be cast as = 0 [ uno.2 ] with @xmath106 and @xmath107 $ ] .
thus , the meaning of @xmath103 is obtained by means of the divergence of the faraday tensor as @xmath108 , @xmath103 is an external current .
it must be mentioned that another possibility to write a covariant form of the interaction comes through the dual tensor .
the anomalous coupling with an electric field in the form @xmath109 with @xmath110 , produces the same result . with the aid of the vector @xmath111
we can introduce more interactions in a covariant way . as announced before , a non - local , non
- abelian field tensor @xmath112 can be introduced in the equation by means of the pauli coupling .
the theory underlying this type of field can be found in many references @xcite , @xcite and was motivated by the possibility of introducing high order derivatives in field actions . from all these possibilities ,
we choose the path followed in @xcite by considering that the field acts on wave functions as a bilocal kernel , it can be represented as an operator depending on @xmath113 ( but not @xmath100 ) .
the integro - differential equations obeyed by @xmath114 are also in @xcite and can be used to elucidate the nature of the external source generating such field
. we shall proceed to find the external current @xmath115 producing such fields after we give the expressions for the interactions .
let us now consider a linearization of the field in the @xmath116 variables : an approximation which corresponds to the first corrections due to internal structure of particles described by the non - abelian field .
this means that both wavelength and particle structure are small compared to the scale of the system , which is dictated by the oscillator frequency .
the expansion of the field in @xmath117 can be truncated by keeping linear terms only .
we propose _
2^ = ^ u _ p _ + _ 3^ = 0 , [ dos ] for which the dirac equation reads = 0 , [ tres ] this equation reduces to the hamiltonian form given in ( [ 01.43 ] ) with @xmath118 , as can be verified by setting @xmath104 . to produce a term @xmath119 in the hamiltonian
, we simply recognize that such a potential corrects the rest mass of the dirac particle .
a non - abelian scalar field can be introduced for this purpose in analogy with the one and two dimensional cases explained below .
notably , one could also consider the dual tensor with components @xmath120 and an interaction of the form @xmath121 in order to produce ( [ 01.43 ] ) . the other case , @xmath122 , comes from considering again a parity violating coupling when we use the minimal coupling prescription .
such a form is encountered in field theories where chirality plays an important role @xcite , e.g. the electroweak model .
again , we need non - locality in our fields . by using a non - abelian 4-potential (
e.g. a current vector with isospin 1/2 ) of the form @xmath123 , where a^_1 = r _ , a^_2 = p _ , a^_3 = 0 , [ cuatro ] we can see that the equation = 0 [ cinco ] produces the desired potential in the hamiltonian for @xmath124 .
the term with @xmath125 does not appear in this case , but can be generated by the unitary transformation @xmath126 .
this leaves @xmath127 invariant , while the matrices @xmath128 in the hamiltonian simply rotate .
the nature of such field can be elucidated by inserting our @xmath129 in the corresponding non - local field equations and finding the external current producing it .
for instance , using @xmath130 one has @xmath131 and
its expression in terms of a 4-potential ( see ( 12 ) in @xcite ) reads ^ = i([p^,b^]- ) + [ b_,b _ ] .
[ cinco.1 ] this yields @xmath132 up to gauge transformations as b_=u _ ( r _ r^ _ t_1 + r _ p^ _ t_2 ) . [ cinco.2 ] the field equations give the current as j^ = i[p_,^ ] + [ b_,^ ] + = -u^ t_1 + p^ _ + ( p_^ , r _
_ - p_^ , r_^ r _ ) t_2 + = -u^ t_1 + p^ _ + .
[ cinco.3 ] so far , no approximations have been made .
if we focus on the independent and linear terms we find a ) a constant current @xmath111 as in the dirac oscillator case , b ) a spatial part given by the momentum of the particle , as if the imposed external current would follow the electron . in order to put the cases
@xmath133 together , we may consider a 4-potential given by the combination @xmath134 , with @xmath135 , @xmath132 given above .
the linearity in @xmath136 and @xmath137 invoked in our argumentation ( e.g. in the case where only the first non - local corrections are considered and field intensities are small compared to the oscillator frequency ) implies that the interaction @xmath138 can be approximated by @xmath139 when inserted in the dirac oscillator equation .
the anomalous coupling @xmath121 with the combination of @xmath135 and @xmath132 yields the one produced by @xmath132 when we ignore quadratic terms in @xmath116 and in the constants @xmath71 .
we would like to stress that despite the fact that the linear approximation is helpful to put both cases together , one may consider each case separately without approximations in the light of ( [ cinco.1],[cinco.2 ] ) .
the physical origin of our potentials , as we have seen , depends strongly on the validity of non - local theories and the possible existence of extended particles as discussed by yukawa ( composite particles described effectively by such fields is another possibility ) . here
we construct a field producing the two dimensional hamiltonian ( [ 01.42 ] ) .
greek indices now run from 0 to 2 .
since spin does not appear in the resulting potential @xmath64 , we identify it with a non - local scalar field , according to yukawa s formulation in @xcite , @xcite . in the presence of isospin @xmath68
, the potential has an expansion of the form = = _
j=1 ^ 3 t_j d^4k^(j ) d^4l^(j ) f_j(k , l ) e^ik^(j ) _ x^ e^il^(j ) _ p^ , [ a.2 ] where @xmath140 are 2 + 1 dimensional lorentz vectors representing the momentum of a scalar particle of mass @xmath141 , while the @xmath142 vectors represent the transversal _ radius .
_ we shall consider only one fourier component . according to yukawa @xcite ,
the corresponding 4-vectors obey k^(1 ) _
k^(1 ) = k^(2 ) _
k^(2 ) = m^2 + l^(1 ) _
l^(1 ) = l^(2 ) _
l^(2 ) = -^2 + k^(1 ) _
l^(1 ) = k^(2 ) _
l^(2 ) = 0 .
[ seis ] where @xmath143 is the radius of the scalar particle .
again , linearization in @xmath136 will be necessary and it is related to the first corrections to a scalar particle ( or a doublet of particles , since we use @xmath2 as isospin ) with small @xmath143 .
now , we can choose a frame of reference where the set of vectors above is given by k^(1 ) _ = ( , k^(1),0 ) l^(1 ) _ = ( 0,0 , ) + k^(2 ) _ = ( , 0,k^(2 ) ) l^(2 ) _ = ( 0,,0 ) . [ siete ] and define ^(1 ) _ = ( ) l^(2 ) _
^(2 ) _ = ( ) l^(1)_. [ ocho ] finally , we give the field @xmath144 as _ 1 = ^(1 ) _ r^ - l^(1 ) _ p^ , _ 2 = ^(2 ) _ r^ - l^(2 ) _
p^ , _ 3 = .
[ nueve ] the equation = 0 , [ diez ] with @xmath145 , yields the two dimensional hamiltonian in ( [ 01.42 ] ) for @xmath146 .
the frame of reference where this is possible , is dictated by the unitary time - like vector @xmath103 chosen as @xmath147 .
the case @xmath148 is obtained by considering a non - abelian gauge field whose scalar component is written in the same spirit , = 0 + a_= u _ [ once ] we proceed along the same line by introducing a non - local scalar field .
greek indices now run from 0 to 1 and dirac matrices are @xmath149 .
we use the momentum and radius vectors @xmath150 in a frame of reference such that k_=(m,0 ) , l_=(0 , ) , [ doce ] define @xmath151 and set the field components as _ 1 = _
r^ , _ 2 = l _ p^ , _ 3 = .
[ trece ] using the same yukawa coupling and choosing @xmath152 we obtain the desired potential in the hamiltonian for @xmath72 . the case @xmath73 is obtained as before , by using a gauge potential @xmath153 .
now we concentrate on the @xmath95 dimensional case for simplicity .
furthermore , we shall find a similarity between the block ( [ 02.2 ] ) for @xmath97 and the corresponding block for the one dimensional problem .
let us write the dirac equation in hamiltonian form by choosing @xmath154 .
it is useful to introduce a coupling constant @xmath155 for the field components @xmath156 in ( [ trece ] ) .
the resulting schroedinger equation in terms of ladder operators reads = i , [ 6 ] for which the usual yukawa coupling is obtained when @xmath157 and @xmath73 , while for @xmath72 , @xmath158 we obtain the zero component of the non - local potential .
now we proceed to find the solutions of the stationary version of ( [ 6 ] ) by noting that the operator i = a^a + ( _ 3 + t_3 ) - 1 [ 7 ] satisfies @xmath159=0 $ ] , @xmath18 is a conserved quantity and it gives the set of states for which @xmath15 is reduced to a block diagonal form .
such states are formed by harmonic oscillator states @xmath160 and states @xmath161 for isospin projection @xmath162 and @xmath12spin projection @xmath163 .
the states are given by |_1^n = |n+2 | satisfy i |_i^n = n |_i^n , i=1, ... ,4 .
[ 9 ] the hamiltonian @xmath15 in ( [ 6 ] ) has a matrix representation h= ( cccc h_0&0&0 &
+ 0&h_1&0 & + 0&0&h_2 & + & & & ) , [ 10 ] where the diagonal blocks @xmath164 have matrix elements @xmath165 .
they have the form h_n= ( cccc -m-(b - a)&(b - a)&&0 + ( b - a)&-m+(b - a)&0 & + & 0&m-(b+a)&(b+a ) + 0&&(b+a)&m+(a+b ) ) , [ 11 ] + except for the lowest states forming a singlet and a triplet which we now indicate .
the state @xmath166 forms a singlet of @xmath18 , the only state with eigenvalue equation @xmath167 .
its energy is given by @xmath168 trivially .
the states @xmath169 form a triplet of @xmath18 with eigenvalue @xmath170 , @xmath171 ( similarly for the other two ) .
they give a block of the form h_-1 ( ccc -m-(b - a)&(b - a)&1 + ( b - a)&-m+(b - a)&0 + 1&0&m-(b+a ) ) [ 11.1 ] leading to the lowest energy levels as solutions of a cubic polynomial .
these two cases represent corrections to the vacuum energies , as can be seen by choosing vanishing interaction parameters @xmath71 . in the following ,
we shall concentrate on general states described by ( [ 8 ] ) , states satisfying ( [ 9 ] ) for values @xmath172 .
the spectrum can be obtained from the @xmath173 dimensional secular equation @xmath174 using ( [ 11 ] ) , with @xmath175 the @xmath8 identity matrix .
this , of course , lies on the fact that the formulae for the roots of a fourth degree polynomial are known @xcite .
eigenvectors can be written explicitly as well .
for the sake of simplicity we study particular cases and discuss their physical importance .
this case corresponds to gauge fields acting on the dirac particle .
conversely , one could put @xmath176 , with analogous solutions except for the presence of some signs in the parameters @xmath177 .
the situation is then quite general from the point of view of the solutions .
the resulting eigenvalues ( [ ener ] ) exhibit a dependence on @xmath21 such that the factors in front of it get corrected by the presence of @xmath177 .
the oscillator vacuum energy @xmath178 is modified as well .
therefore , both frequency and rest mass acquire corrections due to the interaction , as expected .
the blocks of @xmath15 are given by h_n= ( cccc -m-&&&0 + & -m+&0 & + & 0&m- & + 0&&&m+ ) .
[ block ] the following definitions are helpful p= & ( ( 2 n+1 ) ( 1+^2)+2 ^2 + 2m^2 ) + q=&16 m^2(m^2-(1+^2))+ ( 1 + 4 ^2)^2 + 2(1 + 4 m^2 - 8 ^2 ) ^2 + ^4- + & 16(1+m^2 ( ^2 - 2)+^2-^2 ( 1 + 2 ^2)+^4)(n+2)+ + & 16(1- ^2 + ^4 ) ( n+2)^2 + r=&108 ( ( ( n+2)(1- ^2)+m^2-^2 ) ( ( n+1)(1-^2)+m^2-^2)p + & - p^3 + ( m ^2+)^2 ) + s= & ( ) ^1/3 + ( ) ^1/3 . the eigenvalues can be expressed as : e_i(n)=\ { l + , i=1,2 + + -+ , i=3,4 .
[ ener ] we denote the eigenvectors of ( [ block ] ) without normalization by @xmath179 such that @xmath180 .
their components @xmath181 for @xmath182 can be written explicitly as v_1i= & ( ( n+1)(1-^2)+(m - e_i)^2-^2 ) + v_2i= & ( m+ ) ( ( n+2)(1-^2)+^2+m^2-^2)- + & ( ( n+2)(1+^2)-^2+(m+)^2 ) e_i+ ( -m ) e_i^2+e_i^3 + v_3i=&((2n+1)e_i - m- ) + v_4i= & ( ( n+2)(^2 - 1)+(+e_i)^2-m^2 ) .
we put this case separately , although it can be reached from the solutions in the last subsection .
for this set of parameters , it is evident that a new symmetry arises in the form of permutational invariance @xmath184 and conservation of the total spin magnitude @xmath185 . in the context of particle physics
, such a conservation law indicates that isospin @xmath186 fermion charge is a good quantum number .
the characteristic polynomial is factored in qubic and linear expressions as a result of the antisymmetric singlet and the symmetric triplet of the total spin @xmath187 .
in fact , the cubic part is a _
depressed _ cubic .
the spin zero state is infinitely degenerate with zero energy .
we use the states @xmath188 are states of the total spin @xmath187 with @xmath189 and its projection @xmath190 .
it is straightforward to compute @xmath191 .
we obtain h_n= ( cccc 0&0&0&0 + 0&2&&0 + 0&&0 & + 0&0&&-2 ) [ 17 ] this yields eigenvalues which are similar to those for the previous case , but with @xmath192 . as a simple application of our relativistic model
, one can eliminate the presence of the field , except for its static part .
for a perturbed confined particle , this case only takes into account a static correction to its rest mass and zero point energy .
as we shall see , the oscillator number is not affected by factors involving @xmath4 , leaving the dirac oscillator frequency untouched .
one could state that the simplest case of mass generation ( even at the level of classical fields ) is thus achieved .
the matrix becomes h_n= ( cccc -m-(b - a)&0&&0 + 0&-m+(b - a)&0 & + & 0&m-(b+a)&0 + 0&&0&m+(a+b ) ) , [ 17.2 ] with eigenvalues e_1(n ) = -b - , e_2(n ) = -b
+ + e_3(n ) = b - , e_4(n ) = b + .
[ 17.3 ] the eigenvectors of @xmath194 are given by the rows of the following matrix v^t= ( cccc & 0&1&0 + & 0&1&0 + 0&&0&1 + 0&&0&1 ) .
[ 17.4 ] this corresponds to a confined massless fermion perturbed by a field with no static part .
although one could eliminate direct corrections to the rest mass by setting @xmath73 in the hamiltonian , the energies for this case contain terms analogous to the rest energies due to @xmath155 and with a two - fold splitting .
the hamitonian is now independent of @xmath196 and is invariant under the map - , - , x -x , p -p .
[ 17.5 ] the characteristic polynomial of each block @xmath194 is now biquadratic .
the four eigenvalues are found by choosing one of the alternating signs in the following expression e(n)= [ 17.6 ] where a(n ) = \ { ( n+2)^2 - ( n+2 ) .
^2 } ^1/2 + b(n ) = 2n + 3 .
[ 17.7 ] eigenvectors can be obtained easily .
as an example for dynamics in this system we shall now study the evolution of entanglement of the dirac oscillator and the field .
we assume our dirac particle to be initially in some eigenstate of the dirac oscillator , e.g. of positive energy @xcite given by a_n^(-)&= & + _ n&= & the complete system is considered to be initially in a factorized state of the form dirac oscillator @xmath197 isospin , @xmath198 , @xmath199 being a pauli spinor for isospin .
we put in general @xmath200 and examine two different angles @xmath201 .
we shall see that the particular choice of @xmath202 will not be relevant to what we want to show .
the initial condition can be written as linear combinations of the states ( [ 8 ] ) with coefficients @xmath203 , @xmath204 .
such states , in turn , are expressed as superpositions of eigenstates of the system @xmath205 .
this results in & & k_2 |_n+2 ^ 2 + k_4 |_n+2
^ 4 + & = & k_1 _ j=1 ^ 4 d_1,j^(n+1 ) |_n+1^j + k_3 _ j=1 ^ 4 d_3,j^(n+1 ) |_n+1^j + & + & k_2 _ j=1 ^ 4 d_2,j^(n+2 ) |_n+2^j + k_4 _ j=1 ^ 4 d_4,j^(n+2 ) |_n+2^j , [ 4.2 ] where @xmath206 are coefficients obtained by inverting the expansion decomposition of the propagator to ( [ 4.2 ] ) we obtain @xmath207 . now that we know the explicit form of the time - dependent state , we are in the position to compute all quantities related to it . since we are interested in the way in which the bound fermion state entangles with the external degrees of freedom , we take a pure state density operator @xmath208 of the entire system and compute purity @xmath209 and entropy @xmath210 of the dirac oscillator subsystem : p(t)&= & tr_n , ( ( tr _ ( t ) ) ^2 ) + s(t)&= & - tr_n , ( tr _ ( t ) log ( tr _ ( t ) ) ) , + [ 4.3 ] where @xmath211 is the trace with respect to oscillator and @xmath212-spin degrees of freedom , while @xmath213 is the trace with respect to isospin .
results for ( [ 4.3 ] ) are depicted in figure ( [ fig2 ] ) as functions of parameters @xmath214 and time .
( abscissa ) and static field @xmath4 ( ordinate ) .
parameters are @xmath215 , @xmath216 .
the left column corresponds to @xmath217 and the right column to @xmath218 .
we take @xmath178 for the dirac oscillator state .
the structure at @xmath219 is present in all density plots and is persistent at all times . in this region
we observe an entropy increase and a purity decrease .
this is an indication of maximal entanglement between the external field and the positive and negative energy states of the confined fermion .
the situation corresponds to a regime in which the field intensity is comparable to the rest energy : particles can be created in the context of quantum field theory.,title="fig : " ] ( abscissa ) and static field @xmath4 ( ordinate ) .
parameters are @xmath215 , @xmath216 .
the left column corresponds to @xmath217 and the right column to @xmath218 .
we take @xmath178 for the dirac oscillator state .
the structure at @xmath219 is present in all density plots and is persistent at all times . in this region
we observe an entropy increase and a purity decrease .
this is an indication of maximal entanglement between the external field and the positive and negative energy states of the confined fermion .
the situation corresponds to a regime in which the field intensity is comparable to the rest energy : particles can be created in the context of quantum field theory.,title="fig : " ] ( abscissa ) and static field @xmath4 ( ordinate ) .
parameters are @xmath215 , @xmath216 .
the left column corresponds to @xmath217 and the right column to @xmath218 .
we take @xmath178 for the dirac oscillator state .
the structure at @xmath219 is present in all density plots and is persistent at all times . in this region
we observe an entropy increase and a purity decrease .
this is an indication of maximal entanglement between the external field and the positive and negative energy states of the confined fermion .
the situation corresponds to a regime in which the field intensity is comparable to the rest energy : particles can be created in the context of quantum field theory.,title="fig : " ] ( abscissa ) and static field @xmath4 ( ordinate ) .
parameters are @xmath215 , @xmath216 .
the left column corresponds to @xmath217 and the right column to @xmath218 .
we take @xmath178 for the dirac oscillator state .
the structure at @xmath219 is present in all density plots and is persistent at all times . in this region
we observe an entropy increase and a purity decrease .
this is an indication of maximal entanglement between the external field and the positive and negative energy states of the confined fermion .
the situation corresponds to a regime in which the field intensity is comparable to the rest energy : particles can be created in the context of quantum field theory.,title="fig : " ] one finds two regimes of maximal entanglement as expected .
the first one is naturally related to the possibility of tuning the external field to the frequency of the oscillator , @xmath220 .
the second regime appears when the strength of the coupling is comparable with the rest mass of the dirac particle , @xmath219 .
this shows that the structure of the external field is capable of probing the region where particles can be created .
it is well known that under these circumstances the external field creates particles and the quantization of the dirac field becomes important @xcite . at the level of classical fields ,
the results show that this particle creation is accompanied by a surge in the entanglement between the dirac oscillator and the external field .
it is not surprising , that purity and entropy yield quite similar information .
we display both , as entropy is more popular but purity can be obtained analytically .
we refer to @xcite , @xcite for a discussion of the difference of the two quantities .
by considering a bound fermion under the influence of a non - local external field , we have constructed a generalization of the dirac oscillator in which interactions represent the influence of extended particles in the yukawa formalism or the non - local gauge field theory .
the relativistic model has been shown to be solvable in one , two and three dimensions by constructing the corresponding invariants in a specific frame of reference . by covariance
, the solutions in arbitrary inertial frames can be reached by lorentz transformations .
we gave spectra and eigenstates explicitly for several cases according to different choices of parameters .
these include explicit expressions for positive and negative energy ( particle and antiparticle ) states as well as corrections to the rest mass of the bound fermion .
an interpretation of entanglement of the field with positive and negative energy states as a particle creation effect has been possible .
it is worth to mention that our solvable problem has been formulated in a way which allows an immediate identification with quantum optics . the remarkable analogy between the dirac oscillator and the jaynes - cummings hamiltonian @xcite ,
@xcite has been pointed out before @xcite , @xcite , @xcite with the aim of producing such a system in a quantum optical experiment .
our work may extend the analogy further by identifying our hamiltonians with a jaynes - cummings model of two atoms .
the @xmath12spin and isospin of the field correspond to the operators describing two atoms of two levels , while the harmonic oscillator operators @xmath221 correspond to one mode of the electromagnetic field .
an experimental realization for this system in the context of quantum optics can be proposed as well . in @xcite , we have given a dynamical study in this context .
specifically , we may consider the following setup for the case studied in section 5.1 , where the parameters are quite general .
we require two atoms of different species trapped in an electromagnetic cavity .
both atoms must be nearly in resonance with only one mode of the eletromagnetic field .
therefore , a sufficiently large spacing between adjacent modes is needed .
the coupling constant @xmath155 defined in relation with @xmath222 ( fixed as unity ) , can be adjusted by placing the atoms in regions with different field intensties . moreover ,
a large distance between the atoms is needed in order to ignore direct interaction terms .
the case 5.2 is a particular restriction for which the atoms are of the same kind .
finally , the evolution of entanglement studied in section 6 can be realized by preparing the initial state in ( [ 4.2 ] ) . in principle
, one could prepare it by trapping one atom in the cavity and measuring the energy of the total system ( this corresponds to a dressed state of the jaynes - cummings model with one atom ) .
after this is achieved , the second atom can be introduced in the trap .
this setup thus emulates our model and makes it experimentally accessible .
seligman and e. sadurn dedicate this work to the memory of their teacher and friend marcos moshinsky .
we acknowledge support under the grants papiit in112507 and conacyt 57334 | the dirac oscillator coupled to an external two - component field can retain its solvability , if couplings are appropriately chosen .
this provides a new class of integrable systems .
a simplified way of solution is given , by recasting the known solution of the dirac oscillator into matrix form ; there one notices , that a block - diagonal form arises in a hamiltonian formulation .
the blocks are two - dimensional . choosing couplings that do not affect the block structure ,
these blow up the @xmath0 matrices to @xmath1 matrices , thus conserving solvability .
the result can be cast again in covariant form . by way of example
we apply this exact solution to calculate the evolution of entanglement . |
laminar flow is the case of a flow that persists in a stationary fashion at each point .
such flow is either static or found in the case of uniformly driven boundary conditions or external forces .
the well - definedness of the term is a bit in doubt in the case of time varying boundaries .
if the boundaries move in a periodic fashion and the flow does as well , it might be considered to be laminar even though the flowlines are not time independent .
an unambiguous notion of unsteady flow might be given by the notion of bifurcation as we see in the case of the van karman vortex street .
these flows around obstacles give a set of counterrotating vortices that grow and advance behind the body .
the reason we call this a bifurcation is that , in the eulerian point of view , the flow velocity at each point undergoes a simple period of evolution at each point .
given the values of @xmath3 for a particular point @xmath4 , we can find the neighboring @xmath5 values from the function @xmath6 by integration .
are not arbitrary but must satisfy the frobinius integration condition @xmath7 . ]
the pressure can be found from the the equation of state after integrating the density @xmath8 with @xmath9 .
this particular case gives a foliation of solutions @xmath10 indexed by the one parameter set of value of @xmath5 at @xmath4 .
the navier - stokes equations also require second derivative information in the velocity @xmath11 to relate this to the n - s equations requires us to give a time derivative of each one of these functions that preserves the foliation and advances through it at a rate indicated by the relative density @xmath12 at each point .
let these solutions be indicated by @xmath13 , @xmath14 where @xmath15 is a compact parameter .
we can write the n - s equations , assuming that the fluid is incompressible enough during its motions that bulk viscous contributions are negligible , as @xmath16 our parameterized set of solutions then evolves along the parameter space by a function @xmath17 .
we can rewrite the equations of motion in terms of the local functions @xmath18 @xmath19 in general , we would expect closure relations to be necessary for the evolution of @xmath20 and @xmath21 but , for such a simple flow , these are constrained uniquely by the values of @xmath5 and @xmath22 at each point .
this then becomes an implicit equation for @xmath23 . in terms of our statistical variables
, we see that for each point @xmath24 , the values of @xmath25 , @xmath26 , @xmath22 and @xmath27 have only one solution for each choice of @xmath5 obtained at that point .
this means that the probability function @xmath28 condenses to the form @xmath29 where the functions @xmath30 are assumed known .
the one parameter set of solutions at each point will impose self consistency conditions on such functions since they are not independent of the functions @xmath31 and @xmath32 .
we have chosen not to utilize the pressure function as a fundamental variable and instead to keep track of the density @xmath22 with an implicit equation of state @xmath33 . in the case of a constant density fluid
we can derive pressure from an elliptic constraint equation of the flow but , for our new statistical variables , some correlation data will almost certainly enter and this becomes intractable for our distribution function in the form @xmath34 .
as the flow gets more complicated we can not assume that the single valued nature of the variables is preserved so that knowing @xmath35 is sufficient probability to determine the entire solution .
in fact , we know from the case of the lorenz attractor that the distribution function may not even be quasicontinuous as this grows in complexity . however , we can be sure that , if we drive the flow uniformly with stationary boundary conditions that the flow will converge to a long time average distribution on some time scale even if it is turbulent .
our `` solution '' in eqn .
[ formal ] was rather formal and seems to buy us no new information .
let us now use it as a template to describe more complicated flows .
as in the case of nonlinear differential equations , it seems likely that the local flow data at each point undergo a set of bifurcations as the flow becomes more extreme .
specifically , the laminar flow leads to oscillations so @xmath5 and @xmath22 at each point move in a bounded cycle so that the probability @xmath36 at each point gives an orbit in 4-d space .
this orbit can become very convoluted and never self intersect .
as long as the distribution @xmath37 is such that only one choice of values of @xmath22 , @xmath25 and @xmath26 can correspond to each @xmath5 , the distribution function in eqn . [ dist ] decomposes and the evolution of the stable distribution(s ) is simple . when this simple periodicity no longer holds , one can look for higher dimensional correlation functions like @xmath38 so that a particular choice of @xmath39 at any given point are not sufficient to determine all the required derivatives in the n - s equations uniquely .
thus , instead of a closure scheme , we are confronted with a pde with boundary conditions where we seek a static solution to a distribution where the dimensionality of the solution may become dependent on the complexity of the flow .
observations of turbulent flow give some qualitative observations on the size and strength of vortices so that knowing one component of the velocity at a point puts strong limits on the other directions of flow and its gradients .
turbulence gives a range of vortex sizes and energies but the observation of scaling laws of kolmogorov and richardson @xcite suggests some ordering in the energy of vortices based on size . because of this and the above bifurcation arguments , we begin with the assumption that the time averaged filling of the local jet spaces for unsteady flows , even turbulent ones , is rather sparse . specifically , given a probability function @xmath36 , once @xmath5 and @xmath22 are specified there is only one local flow profile corresponding to it .
this seems like a large assumption but in the case of the vortex street , it is clearly so
. consider such a density function found by a long time observation .
not just any such function is meaningful .
these are found by averaging over many global flows . as such
, we expect some consistency constraint on such functions .
when we start with @xmath40 at a particular point @xmath4 this should correspond to one global flow for a particular @xmath39 there .
this means we should be able to extend our solution outwards from this point by looking for similar @xmath12 weights in the neighboring points . of course , if @xmath12 is constant in space this is not sufficiently restrictive .
we know that linear shear flow between two parallel plates under high @xmath41 numbers is a solution of n - s but not a stable one .
it is reasonable to expect physical solutions will involve @xmath12 functions of variability in all directions in the space @xmath42 .
an immediate problem we face is that restricting @xmath36 to a constant value in a neighborhood of @xmath43 does not uniquely fix @xmath5 and @xmath22 in it .
therefore , we will need a collection of such functions .
a set that meets this need is the topic of the next section .
our goal is to find one ( or at least a sparse set of ) solutions corresponding to the stationary driving boundary conditions for a flow . to this end
we introduce a time dependence for @xmath12 as @xmath44 reminiscent of the ensemble approaches in statistical physics .
however , we are not attempting to imply some sort of local , temporal or ensemble averaging of questionable meaning in this case . we know how n - s evolves the implied solutions and therefore how such a distribution should evolve . by using some sort of annealing procedure
we hope to find stationary physical @xmath12 functions that give the distribution of the flows over long times .
these distributions will degenerate to particular values at the boundaries or other locations where driving conditions are present .
we have seen that there is reason to suspect that the statistical distributions from long time averages of a flow with stationary driving conditions will tend to be fairly sparse .
however there is some problem with getting the relations between the spaces at neighboring points to relate to the n - s equations from a single correlated probability function @xmath36 .
let us begin with the notion that the probability of a given @xmath45 at point @xmath46 tends to a distribution that we can use to bootstrap our way to correlated probabilities of the @xmath47 , @xmath48 and @xmath22 components .
the field gradients are to be determined by the self consistency condition that the probability functions evaluated at constant value on a neighborhood induce a region of a true solution in time evolution of the system regardless of the initial data . to this end
introduce the following probability functions at each point : @xmath49 to extract the local information about the solution with value @xmath50 at point with coordinates @xmath4 we evaluate the value of @xmath51 at @xmath52 and find the values of @xmath45 that force @xmath51 to be constant in the neighborhood of @xmath4 .
the dimensionality of the argument implies that there is , generally , one such solution in this neighborhood so that we can expand @xmath53 where @xmath54 is a function of a small @xmath55 direction vector defined by @xmath56 . to linear approximation , the constant value @xmath51-sheet is locally defined by @xmath57 so that @xmath58 and @xmath59 however , we need second order information to utilize the viscosity terms in the n - s equations .
the conditions to second order can be found by realizing that @xmath60 can be expanded as @xmath61 requiring the probability function to be constant locally gives @xmath62 or @xmath63 choosing the branch of solutions connected to the above linear one we have @xmath64 from which we determine that @xmath65 and @xmath66 substitution in the n - s equations gives us equations for @xmath67 in terms of @xmath5 , @xmath68 , @xmath69 , and @xmath70 .
this still leaves the missing parts @xmath47 , @xmath48 , @xmath22 and @xmath71 .
interestingly , the n - s equations are linear in the velocities in the viscous term but not in the advective term @xmath72 .
we now note that we have a local second order representation for @xmath45 we can use @xmath73 and substitute this result to give @xmath74 . by iterating the above process
we obtain equations for @xmath47 . similarly , @xmath75 and
@xmath76 give the missing @xmath48 and @xmath22 information .
the equations for @xmath77 and @xmath78 arise in the same fashion .
the conservation law specifies @xmath79 .
these are all in terms of our probability functions , @xmath80 , and gradients of them .
the above repeated iteration allows us to bootstrap our way to the second order derivatives of all the velocity functions @xmath81 and @xmath22 while allowing correlations among them .
our goal is not now utilize these the accelerations @xmath82 from the n - s equations to specify the time derivatives of the probability functions as a closed system . to accomplish this for @xmath51
we need to transfer the information from each solution patch and its evolution to obtain the time dependence of the probability function @xmath83 .
the n - s solution tells us how @xmath5 is locally changing through @xmath84 . to map this to an action on our probability function
we view this as a push - forwards of the distribution function @xmath85 at each point .
since @xmath86 is a density the transformation @xmath87 is induced by the shift in the coordinates and a factor of the measure of the transformation .
the jacobian of the transformation is @xmath88 where the location @xmath43 is fixed .
the function @xmath89 is a function of second spatial derivatives of @xmath51 but now we interpret it as vector field in the local space labeled by @xmath45 so the derivatives wrt @xmath45 require no higher order derivatives of @xmath51 .
this measure factor ensures that our normalization condition is preserved and @xmath90 .
the evolution of @xmath51 is then given by the infinitesimal evolution of the distribution @xmath91 @xmath92 where @xmath93 is given indirectly by the n - s equations through @xmath94 .
we have taken some liberty with the single derivative `` dot '' symbol here not to imply @xmath51 is a function of one variable but to indicate that the function @xmath51 lives on the fixed coordinate space @xmath95 so none of the arguments evolve hence no chain rule factors arise .
we can iteratively apply this argument to get equations for the time derivatives of each of the distribution functions to give a set of four coupled partial differential equations . for the case of
@xmath96 we have @xmath97 where the jacobian of the transformation is @xmath98 is the determinant of a 2x2 matrix in @xmath24 and @xmath99 components .
the results for @xmath100 and @xmath101 are similar , though of increasing complexity due to the iterations .
the calculational flow for these repeated steps can be summarized in fig .
[ flow ] yields a unique corresponding local velocity field about @xmath43 that is evolved by n - s then mapped to back to a flow on the local probability space.,width=384 ] the physical solution is the case where the time derivatives vanish as we have argued that this is the limiting time average of the stationarily driven flow .
certainly such a limit exists .
it is not clear that it is unique or that our solution could not have combined multiple distinct results . in nonlinear dynamics of odes
one can have multiple separate attractors .
certainly we can have distinct unstable flows as in the laminar case . in the case of turbulence or randomly perturbed situations , it is unknown if such a situation can exist .
for now we assume that our flows have no such complications . a natural approach to solving such
a system is to begin with some sufficiently nontrivial initial data so that unique local solution patches can be inferred from our constant probability conditions then to `` anneal '' the distributions by removing probability from the rapidly changing parts of each probability function at each @xmath43 and renormalizing so that the norm of each is preserved .
the motivations for this approach to hydrodynamic flow came from investigations into thermal systems , specifically , the thermal quantum state .
although physicists have grown more comfortable with a deterministic many body wavefunction as the universal descriptor for all time in nature , the extension of this to thermodynamics has been especially troublesome .
the main problems with quantum mechanics as a deterministic system stem from discrete measurement actions .
this is presumed to be solved by decoherence by many , often by updating the very definition of decoherence to meet new challenges and problems . in the case of quantum statistical mechanics ,
ensembles are used rather formally@xcite .
the boltzmann counting is used among eigenstates near a given energy to give density of states function from which all thermodynamic properties are derived @xcite .
this is problematic in the case of discrete states where such a quasicontinuous averaging is ambiguous but more so in the nonequilibrium case .
one can make even more severe objections to the fact that one should be able to have constant energy states that are made from states nowhere near the same eigenstate energies . in the case of classical kinetics ,
the constant energy surface constrains the gas in phase space but such a space in the quantum hilbert space is far larger than that limited by the quantum ensemble . to overcome this problem and find a description that is more compatible with the actual many body wavefunction of a particular system that is overwhelmingly likely to give statistical behavior , in the spirit of classical kinetics , we can attempt a time averaged state in many body space @xmath102 we can give two functions @xmath103)$ ] and @xmath104,\im[\psi(\tilde{x})])$ ] .
equivalently , one could use the many body density @xmath105 and current @xmath106 with the proviso that the current be finite and irrotational except at @xmath107 dimensional subspaces where the density vanishes .
the problem of deriving quantum kinetics from first principles persists despite the frequent usefulness of the kubo formula .
it is clear that one must have a spread in eigenstate energies to get any fluctuations . on the other hand ,
if one needs a particular distribution of energies then the property of self equilibration of a system , as we generally see in nature , would not be present .
a local statistical distribution made from the long time averages might provide some universality for a broad class of physical initial data . in a more classical direction , hydrodynamics of classical gases
can be derived to lowest order from the maxwell - boltzmann equations themselves @xcite .
however , attempts at higher order corrections typically lead to divergences .
the problem seems likely to be in the local averaging we use to define our thermodynamic variables .
the density and velocity fields require some space and/or time averaging to get smoothed results . when mean - free - path scale effects become important such averaging is no longer valid .
fluctuations on the small scale are suspected to be important @xcite and this is when it is said that hydro `` breaks . ''
a possible approach that does not rely on ensembles or such regional averaging is to give nonlocal variables that measure , at each instance , the weighted distribution of particle neighbor location about every point ( independent of wether there is a particle at the point or not ) . by extending to many particle separations
, one can get a rather smooth distribution of quasilocal information that is associated with that point . by finding a self consistency condition for
such measures one might evolve statistical measure of such distributions using annealing methods to reach a time independent result corresponding to the long time average of the classical kinetic system .
the use of ensembles in physics have many conceptual difficulties and are introduced `` formally '' so often that these problems are now rarely known well even to practitioners .
this collection of probability functions is introduced in such a way as to avoid these .
the only physical meaningful solutions are the final time independent solutions with the time dependent ones only transitional tools to arrive at them .
the boundary conditions are specified on the distribution function by no - slip conditions so that the distribution degenerates to a single value of @xmath5 there .
the pressure is not necessarily constant but the perpendicular acceleration of @xmath5 should remain zero which induces an indirect constraint on the density there . even if this method should solve many problems of turbulent flow they still have obvious limitations .
we often care about transitions to turbulence and cases where the external forces and boundaries are not constant .
one might be tempted to utilize these probability functions in such cases but care here is essential .
the meaning of such a distribution is not clear even if it should produce a set of desired boundary stresses and internal dynamics and energy distribution though some relaxation .
the use of statistical methods in physics has been a powerful tool since maxwell introduced them for the theory of gases . in hydrodynamics
, flows can have energy distributed on so many scales that numerical simulations rapidly run into speed and data storage limits .
an analytic approach to turbulence based on statistical data has evident advantages even if it is only a stepping stone to more condensed numerical methods . in the case here we have a reasonably compact set of functions with no arbitrary freedom in the evolution equations .
the biggest assumption is in the sparseness of the distribution so we can use local statistical data to match with a local hydrodynamic patch of data to evolve and induce a closure in the evolution of the distribution functions themselves .
it is quite possible that this condition fails at some level of turbulence .
extensions of this method could involve some additional local differential data in the distribution functions to obtain the desired sparseness or a kind of statistical `` connection '' function that joins such local data and also requires some self consistent evolution .
p. kovtun , g. d. moore , and p. romatschke,``stickiness of sound : an absolute lower limit on viscosity and the breakdown of second - order relativistic hydrodynamics '' , phys .
d , * 84 * , 2 , 025007 ( 2011 ) | through a discussion of some typical unsteady hydrodynamic flows , we argue that the time averaged hydrodynamic functions at each point give a rather sparse filling of the local jet space .
this situation then suggests a set of time dependent probability functions that are shown to give evolution uniquely defined by the navier - stokes equations through a set of `` differential distribution equations . ''
the closure relations are therefore unique and have no ad hoc characteristics .
annealing methods are proposed as a way to arrive at the stable stationary solutions corresponding to time averaged fluid flow with constant driving forces and fixed boundary conditions .
some applications of this method to quantum statistical mechanics and kinetic theory to higher orders are suggested .
the treatment of turbulence and otherwise unsteady flow in hydrodynamics is an old problem with enduring relevance @xcite . given that most large energetic fluid systems have enormous reynolds numbers , turbulence is an essential part of the process of conveying strain , heat and material across them . even as a microscopic physical understanding of fluidity and the liquid state remains elusive
, so fluids have enduring mysteries at large scales . computational power for simulation grows while analytic approaches and closure schemes for them proliferate yet our confidence in the results is often undermined by their ad hoc nature .
anticipating that turbulence would be adequately characterized by a mean flow and some fluctuations about it , reynolds introduced an averaging scheme for the velocity @xmath0 .
this gives a mean motion driven by the mean pressure and mean viscosity but with a density dependent `` stress '' term @xcite .
how to treat this stress term is the main complication in the analysis of hydrodynamics of turbulent flow .
analogously , in kinetic theory , the collision term in the evolution of the distribution function @xmath1 is handled by the assumption of `` molecular chaos . '' in the case of kinetics , this works very well to low orders in the flow gradients and there are quantitive bounds on why this should be so @xcite . for fluid dynamics we do not have such a powerful condition suggesting
the problem is either much harder or we have not chosen the best variables . in this article
we will introduce a new set of variables to give a statistical analysis of unsteady flow that is analytical in nature .
specifically , we introduce a probability function , normalized to one at each point , @xmath2 corresponding to the long time average of the local flow and its correlations .
of course , truncation will be necessary as will justifications for why and where this is to be done .
most importantly , we will be interested in stationary boundary conditions and the resulting flows . since ensembles play a standard role in the derivation of hydrodynamics from statistical mechanics and we are interested in a statistical treatment of flows , it is appropriate to give a brief discussion on the relevance and limitation of ensembles to physics in general .
classically , a gas is a particular set of particle positions and motions that we can view as a single point in phase space .
the conserved quantities of energy , momentum
constrain the gas to move about in a reduced dimensional foliated subset of this manifold .
although accurate , this does not give a very useful description of a gas .
we can only measure large scale averaged quantities like pressure , density , velocity and such so we could imagine considering an equivalence class of all such states and evolving .
fluctuations and small scale variations can grow and lead to large scale deviations so this method has its limitations but , the implicit assumption of statistical mechanics is that this not a problem for most purposes . for static (
i.e. flow free ) gases this is valid on long time scales set by entropy considerations .
for dynamic flows , especially driven ones , this is not nearly as good an assumption .
a second consideration is the validity of scales of parcel averaging
. one can average over finite volumes at one time but then this places limits on the size of flow variation we can consider . alternately , one can average over smaller volumes over finite times but this places other constraints on validity as flows change and such a consideration is limited to lagrangian parcel averages .
subcritical fluctuations can exist and these can grow to give large scale differences in turbulent flows . even if we are only interested in local fluctuations , these can give important contributions to macroscopic observables like the viscosity @xcite .
it is hard to see how the kinetic ensemble distribution functions can capture this reality . |
stochastically nanostructured materials exhibit optical , electronic , and thermal properties which can differ drastically from those of ordered nanostructures and of homogeneous samples of the same material . in nanostructured aggregates ,
a combination of strong broadband optical absorption and fast local thermal response gives rise to a strong photoacoustic effect , in which amplitude - modulated light is transduced to sound at the frequency of amplitude modulation . in this work
we theoretically and experimentally investigate the photoacoustic response of nanostructured aggregates in air , and demonstrate that such materials enable the construction of acoustic sources including an optically - addressed photoacoustic phased array .
the flexibility of such acoustic sources could enable low - cost generation of arbitrary broadband acoustic and ultrasonic waveforms , with potential applications ranging from subdiffraction air - coupled acoustic sources to materials characterization .
the study of photoacoustic phenomena in bulk materials has an illustrious history @xcite , and the effect has proven useful as a characterization tool in a variety of contexts @xcite .
photoacoustic material response has also been explored as an optically - controlled source of sound waves : for example , photoacoustic sources have been fabricated from lithographically - defined nanostructures , thin films , multilayer polymer - metal composites , nanoparticle - filled epoxies , and carbon nanotubes @xcite . in this work
, we discuss a simple and low - cost technology for air - coupled photoacoustic ultrasound generation based on dla films .
the technology is within the capabilities of nearly any laboratory , requiring only very moderate ( few torr ) vacuum conditions , inexpensive metal sources , and modulated leds .
we illustrate potential applications by constructing an optically - addressed photoacoustic phased array capable of producing patterned ultrasonic fields at far lower complexity and cost than a conventional multiple - piezoelectric - source phased array .
highly disordered nanostructured films of many evaporable substances can be simply and reproducibly prepared by thermal evaporation in the presence of a low - pressure background of inert gas . in films prepared using this procedure , a complex granular structure results from the randomization of atomic velocities before aggregation in the gas phase @xcite .
such films often but not always resemble the product of diffusion - limited aggregation .
diffusion - limited aggregation occurs when particles executing a random walk collide and stick together ; this process occurs in a wide variety of physical contexts , and is expected to produce a treelike fractal structure @xcite . as a concrete example , we find that nanostructured bismuth films are reliably produced in a standard thermal evaporator by running 115 a of current through a 9 m@xmath0 alumina - coated molybdenum evaporation boat for 4 minutes , in a background pressure of 2 torr of argon .
the nanostructured films are deposited on any surface above the boat , although for surfaces too close to the boat they tend to melt and coarsen .
the properties of the film do not depend sensitively on deposition parameters .
[ sems],title="fig : " ] m high . [ sems],title="fig : " ] m high .
[ sems],title="fig : " ] m high .
[ sems],title="fig : " ] although much of the research into such films has used gold as the evaporant @xcite , we find that dla films are easily prepared from virtually any evaporable substance . in a sense , this method achieves its generality and relative material - independence by using classical mechanics rather than chemistry to produce structures at the nanoscale .
we have used this technique to prepare dla films of copper , silver , aluminum , iron , gold , indium , nickel , bismuth , tin , and silicon monoxide .
films prepared using copper and bismuth appear to be stable over several years when stored in air at room temperature .
elevated temperatures or very high optical intensities ( at least two orders of magnitude higher than those used to obtain the results presented below ) can rapidly decrease the optical absorptivity of the films , presumably via oxidation or coarsening of the nanostructure . at somewhat lower optical intensities of roughly 0.1 - 1 kw
/ m@xmath1 bismuth films display photo - induced aging over the course of weeks , during which the absorptivity and photoacoustic efficiency are gradually reduced , though not to zero .
we expect that this behavior is strongly material - dependent .
[ sems ] compares electron micrographs of samples produced with this technique to numerically - generated diffusion - limited aggregates .
the qualitative agreement between the observed structure and the predicted morphology of diffusion - limited aggregates is quite good for materials like aluminum , copper , and bismuth .
the most striking visual aspect of films prepared in this way is their extreme blackness : as shown in fig .
[ blackness ] , they typically absorb 99% of incident light across a broad range of wavelengths @xcite .
if the light is amplitude - modulated , the films also exhibit a strong photoacoustic response .
although the photoacoustic phenomena modeled and presented here all occur in air , fluid - coupled ultrasound is a possible future direction of study .
for this reason we note that dla films of copper and bismuth , though they are somewhat fragile and ( in the case of copper ) extremely hydrophobic , can easily be immersed in fluids without damage . in the next section ,
we discuss a simple model of the air - coupled photoacoustic properties of a nanostructured granular material .
our qualitative model of the air - coupled photoacoustic response of dla films assumes that the films absorb nearly all light incident upon them , converting the energy into heat .
we consider the effect of this absorption on a single grain of material at the smallest ( material - dependent ) length scale which characterizes the structure . because of the diffuse fractal nature of the material , individual grains are very weakly thermally coupled to the rest of the film .
this means that the grains , which have a small heat capacity due to their small size , can change their temperature substantially and rapidly in response to sudden illumination .
when the illumination stops , our model assumes that the grains lose heat by conduction into the surrounding gas .
this is in contrast to a three - dimensional bulk metal , in which conductive losses into the rest of the material would generally be more important .
amplitude - modulated light can thus excite a large - amplitude oscillation in the temperature of the gas near individual grains at the surface , which leads to a pressure oscillation and the emission of sound waves into the gas .
this process should occur for essentially any material , and in the 1 khz range we have observed audible responses from other black materials such as carbon soot .
however , the efficiency of photoacoustic generation by this mechanism will fall off rapidly above some maximum frequency which depends on thermal time constants of the specific nanostructure . for dla films , both this maximum frequency and the expected amplitude of temperature changes are large , leading to a strong photoacoustic response in air in the ultrasound regime . ] to quantitatively elucidate this simple model , we consider how quickly an individual grain in a dla film can thermally equilibrate with its surroundings , assuming that the dominant thermal relaxation mechanism is conductivity into the neighboring gas .
the rate of thermal energy loss from a grain of metal due to conduction into the neighboring gas is @xmath2 where @xmath3 is the heat capacity of the grain , @xmath4 is a numerical factor parametrizing the degree of thermalization of a gas atom after a single collision with a surface , @xmath5 is the collision rate of gas atoms with the surface , and @xmath6 is the temperature difference between the grain and the gas .
the collision rate @xmath7 is the number density of the gas @xmath8 times the rms velocity @xmath9 of a gas atom times the area @xmath10 of the grain , which for a 10-nm - radius grain at stp is about 4 thz . the heat capacity @xmath3 depends on the substance and grain size ; for a 10-nm bismuth grain it is around 5@xmath11j/@xmath12k .
there is therefore a material - dependent upper bound on the frequency at which such grains can change their temperature , given by @xmath13 .
this result suggests that the bismuth grain can change its temperature at rates greater than 1 mhz .
this model suggests directions for further optimization .
a 1 nm grain would respond 10 times faster than a 10 nm grain ( 100@xmath14 less area , 1000@xmath14 less heat capacity ) .
as shown in fig .
[ sems ] , the grain size is material - dependent ; aluminum , for example , produces much smaller grains than bismuth under similar deposition conditions .
operating in helium instead of air would decrease @xmath15 by a factor of 3 due to the higher mean velocity , and operating at higher pressure would decrease @xmath15 linearly . at frequencies below @xmath16 , the amplitude of the photoacoustic response depends upon the amplitude of the light - induced temperature oscillation . for light with intensity @xmath17 and a grain with surface area @xmath18 ,
the rate of change of temperature is @xmath19 in the steady state the optical heating rate balances the cooling rate . for square - wave light modulation at frequencies far below @xmath16
, we observe that the films respond with a sharp positive and negative pressure pulse at the light switching times , as predicted by the model .
this simple model does not predict the fraction of optical energy converted to acoustic energy , which we experimentally estimate to be on the order of @xmath20 .
for reference , this is about 3.5 orders of magnitude greater than the photoacoustic efficiency reported for a metal - coated silicon wafer in air @xcite .
higher efficiencies are attainable in water , although our efficiency is comparable to that reported for metallic films @xcite and nanotube - based absorbers @xcite in water .
our model also ignores effects which are certainly present at some level , such as radiative loss or conductive loss into the material @xcite .
m was deposited on a plano - concave lens with a 150-mm radius of curvature and back - illuminated with amplitude - modulated white leds at an optical intensity of around 140 w / m@xmath1 .
acoustic intensity was measured with narrowband ultrasonic transducers .
points show measured acoustic intensity versus position .
lines are gaussian curves with a frequency - dependent diffraction - limited width @xmath21 , scaled only vertically and centered on the observed spot .
left panel is a 40khz focus , and right panel is a 200 khz focus .
these data were obtained by summing 2d acoustic intensity scans in one direction . ]
the photoacoustic response of all dla films we have produced is strong enough that illumination of a flat film with an ordinary led flashlight which uses amplitude modulation in the acoustic range produces a response audible to the human ear .
more quantitatively , the strength of the photoacoustic response at 40 khz is more than an order of magnitude stronger than that of a black graphite surface .
the simplicity of dla - based photoacoustic sources enables the construction of flexible and powerful acoustic devices .
the simplest incarnation of a directional dla - based photoacoustic source is a plano - concave spherical lens with a nanostructured film deposited on the concave side .
in such a device , the acoustic wavefronts generated by the dla film inherit the curvature of the lens surface , leading to a converging acoustic beam @xcite . as a proof of concept
, we deposited nanostructured bismuth - black on a plano - concave lens and back - illuminated it with amplitude - modulated light to create a focused acoustic source .
a scanning piezoelectric transducer was used to measure the resulting sound field .
results for modulation frequencies of 40 khz and 200 khz are presented in fig .
[ focusedsound ] , and compared to calculated diffraction - limited intensity curves with @xmath22 radius @xmath23 , where @xmath24 is the speed of sound , @xmath25 is the focal length , @xmath26 is the frequency , and @xmath27 is the lens diameter .
the foci are found to be diffraction - limited ; as expected , the 200 khz spot is 5 times smaller than the 40khz one . to further explore the flexibility of acoustic sources based on dla films
, we constructed an optically - addressed acoustic phased array , diagrammed in fig .
[ papadiagram ] . a field - programmable gate array in combination with an array of fet - based current switches
is used to individually control the phase of amplitude modulation for each high - power led in an 8-by-8 square array with a pitch of 12 mm .
the array is mounted directly behind a glass panel covered with nanostructured bismuth .
this configuration allows for virtually arbitrary optical control of the acoustic wavefronts generated at the dla film . like the recently - proposed metascreen - based passive phased array @xcite
, this device avoids the complexity and cost of a multiple - acoustic - source phased array , but in an entirely different manner . here
the technical advantages arise from the comparative ease and flexibility of generating amplitude modulated light , as compared to direct generation of ultrasound .
figure 6 shows a few intensity patterns created by the photoacoustic phased array and measured with a scanning acoustic transducer . with an appropriately chosen phase configuration
, the device can mimic the photoacoustic lens , creating a single focused acoustic intensity maximum as shown in the upper - left panel of the figure . by adding a phase offset which depends linearly on position , this virtual lens can be dynamically tilted , allowing for full spatiotemporal control of the acoustic intensity maximum .
such a translated focus is shown in the upper - middle panel of fig . [ focusing ] .
slightly more complex phase configurations can create multiple foci ( upper - right panel of fig .
[ focusing ] ) or lines of maximum acoustic intensity ( lower - left panel of fig . [ focusing ] ) .
arbitrary spatial variation of the phase pattern can be used to create more complex waveforms , with a precision limited by acoustic diffraction and the spatial period of the led array .
the fact that the 12 mm pitch is larger than half a wavelength of 40 khz ultrasound also limits the photoacoustic efficiency due to radiation of power into side lobes .
time - varying phase configurations allow for `` painting '' of time - averaged acoustic intensity patterns , for example by scanning a focused spot .
the `` ucsb '' images in fig . [ focusing ] were generated in this way , at a repetition rate greater than 200 hz .
of white light was used to back - illuminate bismuth - black films with a thickness of about 3 @xmath28 m .
the acoustic frequency is 40 khz , and all images are 30 mm across except the `` ucsb '' images , which are 45 mm across .
all images were taken at a distance of 70 mm from the dla film .
colorbar indicates the measured pressure for the upper - right figure .
this pressure ( 0.1 pa rms ) is representative of the typical range , though colormaps on the other subplots vary for visibility . ]
we have discussed , modeled , and demonstrated air - coupled photoacoustic sources based on nanostructured diffusion - limited aggregates .
these sources are inexpensive and simple to fabricate and are capable of broadband ultrasound generation . as an illustration of the device possibilities , we designed and constructed a dla - based optically - addressed ultrasonic phased array and used it to create virtually arbitrary acoustic intensity patterns .
the authors thank thomas witten for a useful and interesting conversation ; carl felten , daeseong kim , and andrew williams for experimental assistance in the construction of the acoustic phased array driver ; and the hellman family faculty fellowship for support . | metallic diffusion - limited aggregate ( dla ) films are well - known to exhibit near - perfect broadband optical absorption .
we demonstrate that such films also manifest a substantial and relatively material - independent photoacoustic response , as a consequence of their random nanostructure .
we theoretically and experimentally analyze photoacoustic phenomena in dla films , and show that they can be used to create broadband air - coupled acoustic sources .
these sources are inexpensive and simple to fabricate , and work into the ultrasonic regime .
we illustrate the device possibilities by building and testing an optically - addressed acoustic phased array capable of producing virtually arbitrary acoustic intensity patterns in air . |
in recent years , kaluza - klein black hole solutions have been studied by many authors in the context of string theory .
since kaluza - klein black hole solutions have compactified extra dimensions , the space - time effectively behaves as four - dimensions at the infinity .
the first kaluza - klein black hole solutions with an extra twisted s@xmath2 were found by dobiasch and maison @xcite as vacuum solutions to the five - dimensional einstein equation , and the features of the black hole were investigated by gibbons and wiltshire @xcite .
the static charged kaluza - klein black holes were also found in the five - dimensional einstein - maxwell theory @xcite and were generalized to the rotating case as the solutions in the five - dimensional einstein - maxwell theory with a chern - simons term @xcite .
supersymmetric rotating kaluza - klein black hole solutions were found by gaiotto _
et al _ @xcite and elvang _
et al _ @xcite .
supersymmetric static multi - kaluza - klein black hole solutions were also constructed @xcite .
these solutions were constructed on the self - dual euclidean taub - nut space in the framework of gauntlett _ et al _ s classification of the five - dimensional supersymmetric solutions @xcite . in these solutions , at the infinity , the space - times asymptote to a twisted @xmath3 bundle over the four - dimensional minkowski space - time .
exact kaluza - klein black hole solutions which asymptote to the direct product of the four - dimensional minkowski space - time and an @xmath3 were also constructed @xcite .
the squashing transformation@xcite is a useful tool to generate kaluza - klein black hole solutions from the class of _ cohomogeneity - one _ black hole solutions with the asymptotically flatness .
actually , wang @xcite regenerated the five - dimensional kaluza - klein black hole solution found by dobiasch and maison @xcite from the five - dimensional myers - perry black hole solution with two equal angular momenta @xcite . applying the squashing transformation to the charged rotating black hole solutions with two equal angular momenta @xcite in the five - dimensional einstein - maxwell theory with a chern - simons term
, the present authors obtained the new kaluza - klein black hole solution @xcite in the same theory .
this is the generalization of the kaluza - klein black hole solutions in ref .
@xcite and it describes a non - supersymmetric ( non - bps ) black hole boosted in the direction of the extra dimension .
one of the interesting features of the solution is that the horizon admits a prolate shape in addition to a round @xmath1 by the effect of the rotation of black hole . in the previous work
@xcite , applying this squashing transformation to non - asymptotically flat kerr - gdel black hole solutions @xcite , we also constructed a new type of rotating kaluza - klein black hole solutions to the five - dimensional einstein - maxwell theory with a chern - simons term . though the kerr - gdel black hole solutions have closed timelike curves in the region away from the black hole , the squashed kerr - gdel black hole solutions have no closed timelike curve outside the black hole horizons .
in addition , the solution has two kinds of rotation parameters in the same direction of the extra dimension .
these two independent parameters are associated with the rotations of the black hole and the universe . in the absence of a black hole ,
the solution describes the gross - perry - sorkin ( gps ) monopole which is boosted in the direction of an extra dimension and has an ergoregion by the effect of the rotation of the universe . in this paper , taking a limit of parameters in the charged version of squashed kerr - gdel black hole solutions introduced in the appendix of ref @xcite , we construct new supersymmetric rotating kaluza - klein black hole solutions to the five - dimensional einstein - maxwell theory with a chern - simons term .
these can be regarded as solutions generated by the squashing transformation of the supersymmetric kerr - newman - gdel black hole solutions @xcite .
like the squashed kerr - gdel black holes in @xcite , these kaluza - klein black hole solutions have no closed timelike curve outside the black hole horizons .
the space - time is asymptotically locally flat , i.e. , at the infinity , the space - time approaches a twisted @xmath3 fiber bundle over a four - dimensional minkowski space - time .
the horizons are a round @xmath1 , unlike known supersymmetric rotating kaluza - klein black hole solutions , where they are the squashed @xmath1 .
we also generalize these solutions to multi - black hole solutions .
in particular , we study two - black hole solutions . as will be shown later
, each horizon admits various lens space topologies @xmath0 ( @xmath4 : natural numbers ) in addition to an @xmath5 and ergoregions have rich structures .
the rest of this paper is organized as follows .
first , following the results of classification of solutions of the five - dimensional minimal supergravity @xcite , in section [ sec : solution ] , we construct general solutions on the gibbons - hawing space , which is the taub - nut space in the special case . in section [ sec :
single ] , we present a new supersymmetric single black hole solution on the taub - nut space . in section
[ sec : multi ] , we study the multi - black hole solutions , in particular two black holes case .
we conclude our article with a discussion in section [ sec : summary ] .
we consider the five - dimensional einstein - maxwell system with a chern - simons term .
the action is given by @xmath6 , \label{action}\end{aligned}\ ] ] where @xmath7 is the five - dimensional scalar curvature , @xmath8 is the 2-form of the five - dimensional gauge field associated with the gauge potential 1-form @xmath9 and @xmath10 is the five - dimensional newton constant .
varying the action ( [ action ] ) , we can derive the einstein equation @xmath11 and the maxwell equation @xmath12 we construct rotating multi - black hole solutions satisfying the equations ( [ eineq ] ) and ( [ maxeq ] ) . the forms of the metric and the gauge potential one - form are @xmath13 ^2 + h ds^2_{\rm gh } , \label{metric } \\
\bm a & = & \frac{\sqrt 3}{2 } h^{-1 } \left [ dt + \alpha v ^\beta \left ( d\zeta + \bm \omega \right ) \right ] , \label{gauge } \end{aligned}\ ] ] where the function @xmath14 and the metric @xmath15 are given by @xmath16 respectively , where @xmath17 is a metric on the three - dimensional euclid space , @xmath18 , and @xmath19 denotes a position vector on @xmath18 .
the function @xmath20 is a harmonic function on @xmath18 with point sources located at @xmath21 , where the killing vector field @xmath22 has fixed points in the base space .
the one - form @xmath23 , which is determined by @xmath24 has the explicit form @xmath25 where @xmath26 and @xmath27 are constants , @xmath28 , and @xmath29 . the base space ( [ ghmetric ] ) with the equations ( [ nut ] ) and ( [ omega ] ) is often called the gibbons - hawking space . in particular , the gibbons - hawking space with @xmath30 , @xmath31 and @xmath32 @xmath33
is the self - dual euclidean taub - nut space , and the space with @xmath34 , @xmath31 and @xmath32 @xmath33 is the four - dimensional euclid space . the solutions - coincide with several known black hole solutions . for example , multi - black hole solutions on the multi - centered - taub - nut space @xcite are obtained by @xmath35 and @xmath36 .
multi - black hole solutions on the multi - centered eguchi - hanson spaces @xcite are obtained by setting @xmath37 , @xmath38 and @xmath36 . restricting the solutions to the case with @xmath39 and @xmath30
, we can obtain new black hole solutions . to avoid the existence of singularities and closed timelike curves outside the black hole horizons , we choose the parameters such that @xmath40 see appendix [ pr ] about the detail discussion .
first , we study the case of a single black hole , i.e. , the case of @xmath41 , @xmath42 and @xmath43 @xmath44 . in the single - black hole case ,
the metric and the gauge potential one - form are expressed in the form @xmath45 ^2 \notag \\ & & + h \left [ v^{-1 } \left ( dr^2 + r^2 d\omega _ { s^2 }
^2 \right ) + v \left ( d\zeta + n\cos \theta d\phi \right ) ^2 \right ] , \label{metric0 } \\ \bm a & = & \frac{\sqrt 3}{2 } h ^{-1 } \left [ dt + \alpha v \left ( d\zeta + n\cos \theta d\phi \right ) \right ] , \label{gauge0 } \end{aligned}\ ] ] where the functions @xmath14 and @xmath20 can be written as @xmath46 respectively .
@xmath47 denotes the metric of the unit two - sphere .
the coordinates run the ranges of @xmath48 and @xmath49 . from the requirements for the absence of naked singularities and closed timelike curves ( ctcs ) outside the black hole horizon , the parameters are restricted to the region @xmath50 the constant @xmath51 is related to the size of the compactified radius @xmath52 at the infinity by @xmath53 where @xmath4 is a natural number . when the parameter @xmath54 vanishes , the metric and the gauge potential one - form coincide with the extreme case of the static charged kaluza - klein black hole solution @xcite .
it should be noted that this solution with @xmath55 coincides with a limiting solution of the squashed kerr - newman - gdel black hole solution @xcite by putting the parameters as @xmath56 and by identifying the coordinates and parameters as @xmath57 about the explicit form of the squashed kerr - newman - gdel black hole solution , readers should see appendix [ kn ] in this article .
introduce a new coordinate @xmath58 with the periodicity of @xmath59 .
in the neighborhood of the infinity , @xmath60 , the metric behaves as @xmath61 where we introduced the following coordinates @xmath62 @xmath63 which are chosen so that they are in the rest frame at the infinity .
the asymptotic structure of the solution is an asymptotically locally flat , i.e. , the metric asymptotes to a twisted constant @xmath64 fiber bundle over the four - dimensional minkowski space - time and the spatial infinity has the structure of an @xmath3 bundle over an @xmath65 such that it is the lens space @xmath0 . the komar mass associated with the timelike killing vector field @xmath66 at the infinity , @xmath67 , the charge at the infinity , @xmath68 , and the angular momenta associated with the spacelike killing vector fields @xmath69 and @xmath70 at the infinity , @xmath71 and @xmath72 , can be obtained as @xmath73 } { 8 \pi \sqrt { 1-\alpha ^2 } g_5 } \frac{{\cal a } _ { s^3}}{n } , \label{sbhmass } \\ q & = & \frac{\sqrt 3 l m } { 2 \pi g_5 } \frac{{\cal a } _
{ s^3}}{n } , \label{sbhcharge } \\
j_{\phi } & = & 0 , \\ j_{\tilde \psi } & = & - \frac{\alpha l ^2 \left ( 3 m + \alpha ^2 n \right ) } { 8 \pi g_5 } \frac{{\cal a } _ { s^3}}{n } \label{sbhangmom},\end{aligned}\ ] ] where @xmath74 denotes the area of a unit @xmath1 .
a black hole horizon exists at the position of the source for the harmonic functions @xmath14 and @xmath20 , i.e. , @xmath75 . in the coordinate system @xmath76 ,
the metric diverges apparently at @xmath75 . in order to remove this apparent divergence
, we introduce a new coordinate @xmath77 such that @xmath78 then , near @xmath79 , the metric behaves as @xmath80+{\cal o}(r ) .
\label{metric0h}\end{aligned}\ ] ] this metric well behaves at the null surface @xmath75 .
the killing vector field @xmath81 becomes null at @xmath75 and @xmath82 is hypersurface orthogonal from @xmath83 at the place .
therefore the hypersurface @xmath79 is a killing horizon . in the coordinate system @xmath84 ,
each component of the metric is analytic in the region of @xmath85 . hence the space - time
has no curvature singularity on and outside the black hole horizon .
the induced metric on the three - dimensional spatial cross section of the black hole horizon located at @xmath75 with the timeslice is obtained as @xmath86 = 2 l m n d\omega _ { \rm s^3/{\mathbb z}_n } ^2 , \end{aligned}\ ] ] where @xmath87 denotes the metric on the lens space @xmath0 with a unit radius . in particular ,
in the case of @xmath55 , the shape of the horizon is a round @xmath1 in contrast to the gaiotto _ et al _ s supersymmetric black holes @xcite . here
, we investigate the number and the structure of the ergoregions . as is discussed in the ref .
@xcite , the space - time admits a considerable rich structure by two kinds of rotations of black holes and the background . for the solution ,
the ergosurfaces are located at @xmath7 satisfying the equation , @xmath88 , where the explicit form of the function @xmath89 is given by @xmath90 r^3 \notag \\ & & + \left [ 3 m \alpha ^2 \left ( m+n \right ) -n^2 \left ( 1-\alpha^2 \right)^2 \right ] r^2+m^2 \left ( m+3 n \right ) \alpha ^2 r+m^3 n \alpha ^2 .
\label{defergor}\end{aligned}\ ] ] note that @xmath91 and @xmath92 for the regions of parameters .
hence there always exists an ergoregion around the black hole horizon .
furthermore , within the parameter region @xmath93 satisfying the inequalities , @xmath94 and @xmath95 , which is shown in fig.[2ergos ] , there are two disconnected ergoregions , inner ( @xmath96 ) and outer @xmath97 ergoregions , where @xmath98 , @xmath99 and @xmath100 are three different positive roots of @xmath101 , two different positive roots of @xmath102 and four different roots of @xmath103 , respectively .
the inner ergoregion is inside a sphere which contains the black hole horizon .
the outer ergoregion has a shape of shell , _ ergoshell _ , which is disjoint from the inner ergoregion .
there exists a normal region between the inner and the outer ergoregions .
the angular velocities of the locally non - rotating observers are obtained as @xmath104 } \alpha .
\end{aligned}\ ] ] from these equations , two disconnected ergoregions of the solution always rotate in the same direction in contrast to the squashed kerr - gdel black hole solutions @xcite , which two ergoregions can also rotate in the opposite directions .
the angular velocities of the horizon @xmath75 are @xmath105
for simplicity , we restrict ourselves to the two - black holes case , i.e. , @xmath106 . without loss of generality
, we can put the locations of two point sources as @xmath107 and @xmath108 , where the constant @xmath109 denotes the separation between two black holes . in this case , the metric is given by @xmath110 ^2 + h \left [ v^{-1 } \left ( dr^2 + r^2 d\omega _ { s^2 } ^2 \right ) + v \left ( d\zeta + \bm \omega \right ) ^2 \right ] , \label{2metric } \end{aligned}\ ] ] where the functions @xmath111 and the one - form @xmath112 are @xmath113 the constants @xmath114 @xmath115 are related to the size of the compactified radius @xmath52 at the infinity by @xmath116 where @xmath117 @xmath115 are the natural numbers .
the metric diverges at the locations of two point sources , i.e. , @xmath118 and @xmath119 .
we make the coordinate transformation so that @xmath120 and @xmath121 .
then , the functions @xmath122 and the one - form @xmath23 in the metric and the gauge potential take the following forms @xmath123 respectively . in order to remove this apparent divergence at @xmath124
, we introduce new coordinates @xmath125 such that @xmath126 dr , \\ d\zeta ' & = & d\zeta + n_2 d \phi .
\end{aligned}\ ] ] then , near @xmath79 , the metric behaves as @xmath127 \notag \\ & & + \left [ \frac{3 n_1 ( 2d+m_2 ) ( n_1 ( 2d+m_2 ) + 2 m_1 ( 2d+n_2 ) ) - m_1 ^2 ( 2d+n_2 ) ^2 } { 16 d^2 m_1 n_1 } \right . \notag \\ & & \left .
- \frac{3 m_2 n_1 + m_1 n_2}{4d^2 } \cos \theta \right ] dr^2 . \end{aligned}\ ] ] this
metric well behaves at the null surface @xmath75 . in this case ,
@xmath128 and @xmath129 also hold at @xmath75 .
therefore , the killing vector field @xmath130 becomes null and is hypersurface orthogonal at this place .
so the hypersurface @xmath79 is a killing horizon . from the same discussion , the other point source @xmath131 also corresponds to a killing horizon .
note that @xmath132 .
so the periodic coordinate @xmath133 has the same periodicity as @xmath134 .
then , the induced metric on the @xmath135-th horizon @xmath136 is @xmath137 , \end{aligned}\ ] ] where @xmath138 .
hence the horizon is topologically the lens space @xmath139 . in the neighborhood of the infinity , @xmath60 , the harmonic functions @xmath14 and @xmath20 behaves as ones with a single point source , i.e. , @xmath140
then , the one - form @xmath23 is asymptotically @xmath141 hence the metric asymptotically behaves as @xmath142 where we introduced the following coordinates @xmath143 @xmath144 which are chosen so that they are in the rest frame at the infinity .
the asymptotic structure of the solution is an asymptotically locally flat , i.e. , the metric asymptotes to a twisted constant @xmath64 fiber bundle over the four - dimensional minkowski space - time and the spatial infinity has the structure of an @xmath3 bundle over an @xmath65 such that it is the lens space @xmath145 , where @xmath146 is the natural number . from the asymptotic behavior of the metric , we can obtain the komar mass , the charge and the komar angular momenta at the spatial infinity as @xmath147 } { 8 \pi \sqrt { 1-\alpha ^2 } g_5 } \frac{{\cal a}_{\rm s^3 } } { n } , \\ q & = & \frac{\sqrt 3 l \sum _
{ 2 \pi g_5 } \frac{{\cal a}_{\rm s^3 } } { n } , \\ j_{\phi } & = & 0 , \\ j_{\tilde \psi } & = & - \frac{\alpha l ^2 \left ( 3 \sum _
i m_i + \alpha ^2 \sum _
i n_i \right ) } { 8 \pi g_5 } \frac{{\cal a}_{\rm s^3}}{n}.\end{aligned}\ ] ] for simplicity , assume that two black holes have the equal mass and the horizon topology of @xmath1 , which correspond to the choice of the parameters @xmath148 and @xmath149 .
the ergosurfaces are located at @xmath7 satisfying the equation @xmath150 introduce the coordinates @xmath151 defined by @xmath152 and @xmath153 .
fig.[fig : rot1 ] and fig.[fig : rot2 ] show how the ergoregions change the shapes in a @xmath151-plane as the rotation parameter @xmath27 varies with the other parameters fixed in the cases of the separation parameter @xmath154 and @xmath155 , respectively . here
the horizontal axis and the vertical axis denote the @xmath156-axis and the @xmath157-axis , respectively . in these figures ,
the shaded regions denote the ergoregions , i.e. , the regions such that the left hand side of eq.([eq : ergo_2 ] ) is positive .
two black holes are located at @xmath158 in this plane .
first , see fig.[fig : rot1 ] , the case of @xmath154 .
there exists an ergoregion around each rotating black hole when @xmath27 is small .
when @xmath159 , a new ergoregion in the shape of a shell enclosing two black holes appears far away from them .
there is an inner normal region between the ergoshell and the two inner ergoregions .
as the value of @xmath27 gets larger , the outer ergoshell becomes thick .
when @xmath160 , the inner normal region becomes disconnected .
when @xmath161 , the outer ergoshell and the two inner ergoregions merge and the inner normal region disappears .
see also fig.[fig : rot2 ] , the case of @xmath155 . in this case , when @xmath159 , in contrast to the previous case , two disconnected outer ergoshells appear around each black hole .
when @xmath162 they merge together . .
@xmath154 and @xmath163 . , width=472 ] .
@xmath155 and @xmath164 .
, width=472 ] next , fig.[fig : sep ] shows how the shapes of ergoregions change with varying the separation parameter @xmath165 , where the other parameters are kept unchanged . when the separation is large enough ,
there exist two disconnected outer ergoshells and two inner ergoregions around each black hole .
when two black holes become closer , two outer ergoshells are connected with each other into a single large outer ergoshell .
there are two inner normal regions around each black hole .
when the separation becomes smaller , the inner normal regions join together .
finally , two inner ergoregions around two black holes also join together . .
@xmath166 and @xmath167 . , width=472 ]
we have considered the limiting case given by ( [ limit ] ) in the kaluza - klein - kerr - newman - gdel black hole solutions @xcite , and have presented supersymmetric kaluza - klein - kerr - newman - gdel multi - black hole solutions , in the five - dimensional einstein - maxwell theory with a chern - simons term .
the new solutions have no closed timelike curve everywhere outside the black hole horizons . at the infinity ,
the metric asymptotically approaches a twisted @xmath3 bundle over the four - dimensional minkowski space - time .
though the kerr parameter and the gdel parameter are related , each black hole can have an inner ergoregion and an outer ergoshell depending on the parameter .
we have explicitly presented the various shapes of ergoregions in the case of two - black holes . the solutions ( [ metric])-([omega ] ) can be easily generalized to solutions with a positive cosmological constant @xmath168 . in this
solutions , the harmonic function ( [ h ] ) is replaced by @xmath169 where the constant @xmath170 is related to the cosmological constant by @xmath171 .
in particular , in the case with @xmath172 and @xmath173 , the solutions coincide with five - dimensional kastor - traschen solutions @xmath174 @xcite or klemm - sabra solutions @xmath175 @xcite , which describes the coalescences of black holes with the horizon topologies of @xmath1 into a single black hole with the horizon topology of @xmath1 . in the case of two black holes with @xmath176
, @xmath177 , @xmath172 and @xmath173 , the solution describes the coalescence of two rotating black holes with the horizon topologies of @xmath1 into a single rotating black hole with the horizon topology of the lens space @xmath178 @xcite .
cosmological non - rotating multi - black hole solutions on the multi - centered - taub - nut space are obtained by setting @xmath179 and @xmath30 @xcite .
this solution is not static even in a single black hole case . in the case of @xmath180 , the solution with the harmonic function ( [ eq : cos ] )
would also describe the coalescence of black holes .
we leave the analysis for the future .
h.i . is supported by the grant - in - aid for scientific research no .
is supported by the jsps under contract no . 20 - 10616 .
here we show the inequality @xmath181 under @xmath182 and @xmath183 is the necessary and sufficient condition for the absence of closed timelike curves outside the horizons .
assume that all point sources are located at the @xmath184-axis , i.e. , @xmath185 on the three - dimensional euclid space in the gibbons - hawking space .
in this case , the one - form @xmath186 is proportional to @xmath187 .
the condition of the absence of closed timelike curves outside the horizons is equivalent to the condition that the two - dimensional @xmath188-part of the metric , @xmath189 is positive - definite , where @xmath190 this metric is positive - definite if and only if the following two - dimensional matrix is positive - definite @xmath191 therefore , noting that @xmath192 , we obtain the condition @xmath193 as a result , it is enough to prove the @xmath194 . at the infinity @xmath195
, the function behaves as @xmath196 hence @xmath181 is necessary .
next we show that if @xmath181 is satisfied with @xmath182 and @xmath183 , there is no closed timelike curve outside the horizons . noting that @xmath197 and @xmath198 under the conditions @xmath182 and @xmath183 , the inequality @xmath199 holds everywhere outside the horizons .
the metric and the gauge potential of the squashed kerr - newman - gdel black hole solution @xcite is given by @xmath200,\end{aligned}\ ] ] and @xmath201,\end{aligned}\ ] ] respectively , where the metric functions are @xmath202}{r^2}\nonumber\\ & & + \frac{2(m - q)a^2+q^2(1 - 16ja-8j^2(m+3q))}{r^4},\\ k(r)&=&\frac{v(r_\infty)r_\infty^4}{(r^2-r_\infty^2)^2}.\end{aligned}\ ] ] in the limit of @xmath203 , i.e. , @xmath204 , the solution coincides with the kerr - newman - gdel black hole solution in ref.@xcite . | we obtain new five - dimensional supersymmetric rotating multi - kaluza - klein black hole solutions with the gdel parameter in the einstein - maxwell system with a chern - simons term .
these solutions have no closed timelike curve outside the black hole horizons . at the infinity ,
the space - time is effectively four - dimensional .
each horizon admits various lens space topologies @xmath0 in addition to a round @xmath1 .
the space - time can have outer ergoregions disjointed from the black hole horizons , as well as inner ergoregions attached to each horizon .
we discuss the rich structures of ergoregions . |
from the simple pendulum to more complex oscillators , the unavoidable coupling with the environment generates dissipation .
even if the loss of energy is balanced by a permanent input , the hamiltonian structure is lost .
the description of open systems , i.e. when the coupling with the environment can not be neglected , has been explored in the framework of self - sustained oscillators .
a theoretical approach was first developed by van der pol @xcite to describe the spontaneous regular heartbeats @xcite .
this oscillator is characterized by an amplitude - dependent friction term : the oscillation is amplified for a small amplitude and damped for a larger one .
this equation has been widely studied as a rare example of nonlinear equations that can be solved analytically .
+ self - propelled entities also result from a coupling with their environment , from the motion of living animals to artificial self - propelled systems .
the efficiency of the propulsive mechanism depends on the permanent input of energy and the specificity of the dissipative process .
the origin of the propulsion is diverse @xcite , from low reynolds propulsion of a flagellum @xcite to the bird flight , from vertically vibrating anisotropic disks @xcite to motile colloids @xcite .
however , all these situations can be described by a speed - dependent friction term , as initially introduced by lord rayleigh @xcite .
+ a simple oil drop can be propelled as well . set on a vertically oscillating liquid bath , a millimetric drop does not necessarily coalesce due to the presence of an air film between it and the liquid surface @xcite .
the coupling with the environment arises from the faraday waves .
they appear spontaneously at the surface of a vibrated liquid bath by a parametric instability mechanism @xcite , for an acceleration amplitude larger than the faraday threshold @xmath0 . just below this threshold ,
the drop impacts excite slowly damped faraday waves @xcite . as a result ,
the drop bounces on a surface perturbed by the previously generated waves .
a horizontal momentum can be transmitted to the bouncing drop which propels it @xcite .
the resulting entity , called a walker , is thus formed by a localized drop and an extended accompanying wave packet .
this association is a rare example of a macroscopic dual object formed by a particle and its associated wave .
walkers display various quantum like effects : diffraction through a slit @xcite , tunneling effect @xcite , zeeman - like splitting @xcite , landau level analog @xcite , and surprising behaviors in cavity @xcite .
recently , the confinement of walkers in a two - dimensional harmonic potential leads to a set of discrete eigenstates for the possible trajectories and their associated wave field @xcite . in spite of the similarity between walkers and some quantum systems @xcite ,
the dynamics of a walker is far from the quantum world in many aspects . by its periodic interaction with the liquid surface ,
a walker is an open system , in which the interaction with the environment plays a crucial role
. in particular , the underlying hydrodynamic equations governing the bouncing states @xcite are not conservative , and the generation of the capillary waves dissipates energy .
this energy loss is compensated by the propulsion from the standing faraday waves at each impact .
therefore the walking state is the result of an energy balance between propulsion and dissipation as any self - propelled entity .
+ in this paper , we identify the consequences of the self - propulsion mechanism on the horizontal walker dynamics .
we investigate its horizontal dynamics with a model description of the walker propulsion based on a rayleigh - type friction law .
we focus on a walker trapped by a harmonic potential well .
the study is also restricted to the short memory regime , i.e , when the damping time @xmath1 of the faraday waves is much shorter than the period of rotation of the walker in the harmonic well .
section [ sec : ii ] describes the experimental results in the harmonic well , in the limit of short memory .
section [ sec : iia ] is devoted to the experimental set up .
section [ sec : iib ] presents the general features of the walker motion in this harmonic well . in sec .
[ sec : iic ] , the transients behaviors are investigated .
based on this description , we construct in sec . [ sec : iii ] a general structure for the equation governing the propulsion of the walker at low memory .
the simplified equations of motion are derived in sec .
[ sec : iii]a .
then , the theoretical model is validated with the experimental data in sec .
[ sec : iiib ] .
section [ sec : iiic ] deals with an energetic approach for analyzing the convergence towards the attractors of the dynamics . finally a linear stability analysis close to these attractor points is presented in sec .
[ sec : iiid ] . as a conclusion ,
we highlight some general consequences of the self - propulsion process for the walker dynamics .
as sketched in fig . [ fig1](a ) , a bath of silicon oil of viscosity @xmath2 cp is set into an oscillatory motion vertically at a frequency of @xmath3 hz and at an acceleration amplitude @xmath4 . a stationary wave pattern called faraday waves appear spontaneously at the surface above an acceleration threshold @xmath5 where @xmath6 denotes the acceleration of gravity .
the appearance of faraday waves is triggered by a parametric instability at half the forcing frequency so the wave pattern oscillates vertically at a period @xmath7 s @xcite .
the usual dispersion relation of surface waves prescribes the corresponding wavelength @xmath8 mm .
a submillimetric drop is placed on the vibrating bath tuned slightly below the faraday threshold . at a typical bath acceleration of @xmath9 ,
the drop bounces at twice the period of the bath , which corresponds to the faraday period @xmath10 , so that the drop becomes a faraday wave exciter . at each impact , the drop generates a capillary front which leads to the formation of a circular standing faraday wave pattern centered at the impact point .
the complete mechanism of the wave generation is detailed in @xcite .
the amplitude of the emitted wave decreases in time as the bath acceleration amplitude @xmath4 is chosen slightly below the instability threshold .
the dimensionless parameter @xmath11 is a measure of the persistence time .
it compares the damping time @xmath1 of the faraday waves to the time @xmath10 between two successive drop bounces . for a memory parameter larger than 1
, the faraday wave emitted by each bounce is still surviving during at least the next bounce .
the liquid surface will be thus perturbed by the waves generated during the @xmath12 previous impacts .
the memory parameter @xmath12 can be tuned through the difference between the bath acceleration @xmath4 and the faraday threshold @xmath0 .
indeed , close to the faraday threshold , the damping time diverges as @xmath13 @xcite .
the propulsion mechanism is a direct consequence of this memory effect .
as soon as the decay time of faraday waves is larger than the faraday period ( @xmath14 ) , the drop hits a surface perturbed by the previous impacts .
since the drop bounces on an inclined surface , there is a transfer of horizontal momentum to the drop .
it has been shown that this situation leads to a pitchfork bifurcation at @xmath15 @xcite .
the vertical bouncing state becomes unstable and the drop starts moving , propelled by the waves emitted in its near past . in the present study , the acceleration amplitude is set at about 90% of the faraday threshold @xmath0 , corresponding to a decay time @xmath16 .
for such an acceleration , bouncing drops of diameter @xmath17 @xmath18 m moves at a horizontal velocity @xmath19 averaged on the vertical period of motion between 5 and 15 mm / s .
the exact velocity value depends on the drop size and the bath acceleration and is fixed for a given drop @xcite . for each drop
, the decay time can be converted to a memory length @xmath20 .
this length @xmath21 , sketched in fig .
[ fig1](b ) , corresponds to the typical distance along which the faraday wave sources are still active behind the drop . in this article , we consider the low memory regime , i.e. , when the memory length @xmath21 is smaller than the typical curvature radius of any curved trajectory followed by the walker .
one important consequence is that the wave sources contributing to the propulsion can be considered as mainly aligned on a straight line behind the drop .
the force exerted by each wave on the drop is thus mainly tangential to the trajectory .
consequently the global resulting force is mainly tangential to the trajectory . in the absence of other external force
, a drop bounces along a straight line until it reaches the boundary of the container .
a more interesting experimental case is obtained when the walker is confined .
this situation is obtained by trapping the walker in a harmonic potential well as sketched in fig . [ fig1](a ) . for this purpose ,
the millimetric drop is loaded with a small amount of ferrofluid and is exposed to a magnetic field .
two coils in a helmholtz configuration provide a uniform magnetic field in the region of interest and induces a magnetic moment @xmath22 to the drop .
an additional non uniform magnetic field is generated by a cylindrical magnet with a diameter of @xmath23 mm and a height of @xmath24 mm placed above the liquid bath .
the interaction between the external magnetic field and the drop magnetic dipole generates an external force on the magnetized drop . when the vertical distance @xmath25 between the magnet and the liquid surface is larger than the magnet diameter , the potential well can be considered as harmonic @xcite . for a drop of mass @xmath26 , the dynamics is then reduced to the motion of a bouncing particle propelled by a wave with an external forcing @xmath27 .
@xmath28 is a spring constant which can be tuned by changing the distance @xmath25 .
the vector @xmath29 denotes the distance from the drop center to the symmetry axis of the magnet .
additional experimental details have been given in the supplemental material of a previous work @xcite .
when the walker is released from the edge of the cell , the magnetic force pulls it towards the center of the harmonic potential .
as the drop is self - propelled , it will never stop at the center but orbits around it .
this trapping mechanism is represented on a chronophotograph as seen from above [ see fig . [ fig2](a ) ] .
the position is sampled every @xmath30 s , corresponding to the time interval of two vertical periods @xmath10 . in a few oscillations
, the trajectory converges to a circular motion in which the walker orbits around the center at its free velocity @xmath19 .
such a trajectory can also be computed from the numerical model developed by e. fort @xcite to mimic the walker dynamics ; [ see fig . [ fig2](b ) ] . the same quantitative behavior is recovered with this numerical model in fig .
[ fig2](b ) ( solid black line ) . finally , balancing the centripetal acceleration @xmath31 with the spring force @xmath32 , scales the final orbit radius @xmath33 as @xmath34 where @xmath35 corresponds to the natural frequency of the magnetic potential well , ( see figure [ fig2]c ) .
the coefficient of proportionality @xmath36 differs slightly from @xmath37 as investigated by oza _
_ @xcite and interpreted as a contribution of the surface waves in the radial mechanical balance . for a hamiltonian system
, one would expect a more diverse set of trajectories in a two dimensional harmonic well .
but among all possible solutions , only the circular motion enables the particle to move at a constant speed .
the selection of particular trajectories is a consequence of the dissipative nature of the walker dynamics . in dissipative systems ,
any volume of the phase space can contract in time and additional measurements have been carried out to study the transient regime .
one way to proceed , is to study the transition between circular motions of different radius by abruptly changing the external force . a walker
is then prepared on a circular orbit and the current inside the coils is switched between two values @xmath38 and @xmath39 .
the consequence is an abrupt change of the characteristic frequency value from @xmath40 to @xmath41 .
a typical experimental signal is shown in fig . [ fig3](a ) .
we observe a transition between two different circular motions of radii @xmath42 and @xmath43 corresponding to the two values of the spring constant . after each switch
, the walker follows a reproducible transient regime and converges to the new circular attractor .
this transition is repeated ten times with a duration of 10 s between each switch , larger than the observed transient duration . for small orbits ,
the radius decreases slowly to the final state [ see fig . [ fig3](b ) ] . for larger orbits ,
the radius increases quickly , crosses the final value and follows damped oscillations around the final state [ see fig . [ fig3](c ) ] .
the transient behavior depends only on the final state which can be reached either from a smaller or a larger orbit .
we thus find for the evolution of the instantaneous radius @xmath44 , typical transient behavior of a second - order oscillator , in which the quality factor strongly depends on the orbiting period @xmath45 of the final orbit .
the qualitative change of the transient behavior shows that at least two time scales control the transition .
depending on their ratio , the transient will evolve from oscillations without overshoot ( small orbits ) to damped oscillations with overshoot ( large orbits ) .
the first time scale @xmath46 is imposed by the external force , and should be compared to another one .
where does this second time scale come from ?
we address this question in the next paragraph through a theoretical approach .
the second time scale relies on the dissipative nature of the motion . in the present case ,
the energy exchange comes from the coupling of the drop to the bath in two ways .
first , the drop loses energy at impact by viscous dissipation but also because a new propagating capillary wave is generated .
second , when the drop hits a non horizontal part of a wave , its horizontal momentum is changed .
once the walker moves at its free velocity , i.e. , in a permanent regime , these two terms balance one another .
a phenomenological description of the propulsive force has been introduced by a. boudaoud @xcite , e. fort _
@xcite , and extended by oza _
experiments show that the fluctuations of speed are typically smaller than @xmath2% .
this speed constraint enables us to introduce a simplified expression to model the net accelerating force @xmath47 .
we derive it through symmetry arguments and show _ a posteriori _ that such an expression fits the experimental results .
as we are in the short memory regime , the force @xmath47 is mainly tangential to the trajectory , i.e. , @xmath48 .
@xmath49 must only depend on the amplitude of the speed @xmath50 and not on its direction .
an exchange @xmath51 should only change the sign of @xmath49 .
consequently @xmath49 must be an odd function in @xmath50 , vanishing for a velocity equal to the equilibrium one @xmath19 .
the simplest expression satisfying these requirements is @xmath52 with @xmath53 a constant homogeneous to a friction coefficient . for @xmath54
the force is propulsive whereas it acts as a friction term for @xmath55 .
this force vanishes for @xmath56 .
such a term is called a rayleigh - type friction @xcite and is commonly used to model the motion of active particles .
it is the lowest order expansion of a force which depends on the velocity and vanishes for an equilibrium velocity .
alternatively , eq . [ forcepropulsion ] can be derived by a direct calculation detailed in the appendix .
this form may also be derived from the integrodifferential formulation developed by oza _
we state that @xmath47 is mainly tangential , but the corrective radial term could also be added .
@xcite shows with a higher order expansion from an integrodifferential form that this extra term is responsible for the prefactor @xmath57 discussed in sec .
[ sec : iib ] . note that the various expressions of this velocity - dependent term introduced in previous studies @xcite are all equivalent to the rayleigh friction provided that the fluctuations in speed remain small , i.e. , less than 20% of the free velocity @xmath19 .
the rayleigh - type friction retains the minimal complexity required to describe the walker propulsion theoretically .
however , our expression is not intended to describe the consequence of an abrupt change of velocity .
such cases have been encountered for very specific situations , like highly disordered trajectories @xcite , where velocity fluctuations become larger than @xmath58% . in the present article
, we do not observe significant discrepancies between our simplified model and the experimental data .
the horizontal dynamics can then be rationalized by a continuous equation as the bouncing period @xmath59 is much smaller than the period of rotation @xmath60 s @xcite . using the rayleigh - type friction law and the continuous limit ,
the simplified equation of motion for a walker in a harmonic potential becomes : @xmath61 with @xmath62 a characteristic time .
the one - dimensional case is equivalent to a van der pol oscillator for the velocity , signature of the hamiltonian structure breaking . in spite of its apparent simplicity
, only a few articles deal with the two - dimensional case .
@xcite studied this equation in two dimensions including noises , indicating that circular motions are solutions .
however , the two dimensional rayleigh equation has not been solved analytically .
this can be attributed to the friction term , which involves the norm of the speed .
this term provides a coupling between the motions in the two directions of space .
one particularity of this equation is the presence of two different time scales .
the external force provides the period of oscillation while the rayleigh type friction introduces another time scale @xmath63 , the typical time needed for a walker to recover its free velocity .
it should not depend on the external force and is thus a physical parameter for a given bouncing drop at a given memory @xmath12 . even if symmetry arguments enable us to obtain the simplified equation [ equadiff1 ] , they do not provide the value of this characteristic time @xmath63
. it will be determined by comparing the model with the experimental data .
+ at this step , we introduce the dimensionless position @xmath64 and speed @xmath65 while the time is scaled by the natural frequency @xmath66 .
the equation of motion in cartesian coordinates yields @xmath67 here the upper dots indicate a derivative with respect to the dimensionless time .
@xmath68 is the natural dimensionless parameter comparing the two relevant time scales .
+ as the full time - dependent resolution is not of particular importance for the current study , a numerical implementation of this system of equations has been performed .
the solution of the rayleigh equation can be compared to both experiments and simulations from fort s model .
the transients for @xmath69 are compared in fig .
[ fig2](b ) and shows a good agreement .
the theoretical final radius can also be compared with the experimental and numerical results [ see fig . [ fig2](c ) ] and are also in good agreement .
+ the convergence to the final radius , oscillating or not [ see figs [ fig3](c ) and [ fig3](b ) ] can be approximated by an exponential law and we define a converging time @xmath70 corresponding to this exponential decay time . thus , this converging time can be evaluated directly from the time evolution of the radius both for experimental and numerical data .
figure [ fig4 ] indicates by red circles , the experimental evolution of the inverse of the converging time @xmath70 with the orbital frequency @xmath71 .
we also indicate by dashed black lines , the prediction of eq .
[ equationdyninertia ] as we increase @xmath72 , i.e. , by decreasing @xmath73 .
@xmath63 can be seen as a free parameter obtained from the best matching between the experimental data and the theory .
it enables us to compare the experimental data of convergence with the theoretical predictions and gives @xmath74 s. the existence of an optimal value of convergence for @xmath75 s.rad@xmath76 or equivalently @xmath77 can be seen in fig .
[ fig4 ] . for @xmath78 ( or @xmath79 ) , the nonlinear terms become small and the dissipation averaged on one period is limited : the time of convergence will naturally increase . in the opposite case @xmath80 ( or @xmath81 ) , the two characteristic time scales differs strongly so that the nonlinear term is not efficient to reach the equilibrium speed at @xmath82 . in both cases ,
the consequence is an increase of the convergence time .
we note that the convergence time depends slightly on the initial conditions as soon as the initial motion is not radial . in this purely radial case ,
the drop starts oscillating along a straight line passing close to the center .
but the nonlinear coupling @xmath83 makes this regime unstable to transverse perturbations and the dynamics converges finally to a circular motion .
+ we have investigated the behavior of the walker in a short memory limit and have shown that the dynamics is well approximated by a two - dimensional rayleigh oscillator . we have demonstrated that there are two regimes of convergence depending on the relative magnitude of the two time scales of the dynamics . by comparing experimental results and model calculations ,
we have been able to obtain the value of the second characteristic time @xmath63 .
we now turn to the question : which role does each mechanism play to stabilize onto the circular attractors ? as shown in the previous part , the transient regime of a walker can present either oscillations of radial position or oscillations of speed .
the time of convergence is minimal when their relative amplitude of oscillations is comparable .
@xmath84 can be seen as the value of @xmath73 which optimizes the exchange between potential and kinetic energy . for all other values ,
the mismatch of the two timescales increases the time of convergence .
+ the oscillations of both position and velocity with variable amplitudes suggest studying the stability through an energy approach . for this purpose
we denote the dimensionless mechanical energy @xmath85 .
it evolves in time since the system is not conservative .
equations [ equationdyninertia ] yields @xmath86 the energy evolves as a function of the speed only , and its three fixed points are @xmath87 and @xmath88 . only the solution @xmath89 and the related condition @xmath90 defines an attractive limit cycle as sketched in fig .
this equilibrium corresponds to an equipartition of the total energy into kinetic @xmath91 and potential @xmath92 terms .
the equipartition of the energy is not necessarily a property of a non - hamiltonian system .
in the plane @xmath93 , two different transient regimes are represented in figure [ fig5]b for two different values of @xmath73 and the related paths in fig .
the case @xmath94 ( solid black line ) presents limited oscillations in speed while the case @xmath95 corresponds to oscillations in speed and position ( dashed blue line ) .
the energetic approach proves the existence of only two stable limit cycles @xmath88 , but it does not provide any information about the time evolution .
the latter can be analyzed in principle by seeking the response to a perturbation in the neighborhood of the fixed points @xmath96 . as the amplitude of the nonlinear term vanishes as @xmath97 close to the stable limit cycle
, its influence can be probed by a perturbative approach . introducing a perturbation as @xmath98 and
@xmath99 leads to @xmath100 the left - hand side corresponds to a damped harmonic oscillator but the meaning of the parameter @xmath68 differs from a simple damping ratio .
indeed , fig .
[ fig3]b indicates over - damped relaxation with @xmath101 while fig .
[ fig3]c shows under - damped oscillatory motions with @xmath102 .
the right - hand side is the first order expansion of the nonlinear coupling of the rayleigh equation and introduces time - dependent coefficients with a amplitude proportional to the parameter @xmath73 .
this term provides both a periodical forcing on the velocity and a coupling with the velocity along the other directions of space .
the solution @xmath103 can be represented in a four - dimensional phase space with a time evolution @xmath104 prescribed by @xmath105 @xmath106 the adequate theoretical framework was developed by floquet @xcite and indicates that the solution must be of the form @xmath107 where @xmath108 is a function of time and @xmath18 the floquet coefficients .
the @xmath109 periodicity of the elements of @xmath105 implies that @xmath108 satisfies the condition @xmath110 and that it is also expandable in fourier series .
here we are only interested in the values of the floquet coefficients @xmath18 and not the full solution that would be given by the fourier decomposition of @xmath108 . a way to find the asymptotic ( @xmath111 ) values of @xmath18 , corresponding to the experimental situation is to note that the coupling term in the right hand side of eq .
[ equationdyninertiaordre1 ] induces a shift of @xmath112 in the frequency of oscillation of @xmath113 and @xmath114 .
an asymptotic solution can be sought under the form @xmath115 once inserted in eq .
[ equationdyninertiaordre1 ] , @xmath113 and @xmath114 are solutions only for four particular values of @xmath18 .
the corresponding frequencies are @xmath116 the floquet coefficients can be found numerically @xcite by first solving @xmath117 , with @xmath118 the @xmath119 identity matrix and then finding the eigenvalue of @xmath120 , @xmath121 . as observed in figs [ fig6](c ) and [ fig6](d ) , @xmath122 is found to be always stable as its related floquet coefficient has a negative real part .
the floquet coefficients of asymptotic values @xmath123 giving way to the frequency @xmath124 , are stable and are asymptotically ( @xmath81 ) neutrally stable [ see fig . [
fig6](c ) and [ fig6](d ) ] .
the fourth floquet coefficient , signature of the rotational invariance of the steady state @xcite , is a pure imaginary number , and corresponds to a neutrally stable solution . for this particular case
, the linear approach reaches its limit for analyzing the stability of the rayleigh oscillator .
let us mention that we truncated the initial equation of motion with symmetry arguments ; a linear perturbative approach of the complete dynamical equation could recover a linear stability at short memory arising from terms of higher symmetry .
the values of these particular frequencies should also appear in any transient to the circular motion since they represent a signature of the nonlinear terms when the dynamics spreads from the limit cycle .
this behavior can be observed in the transient regime .
figure [ fig6](a ) shows the spectrum density of @xmath125 during a transient as a function of the normalized frequency @xmath126 with @xmath127 the fundamental frequency .
they have been computed from an experimental transient , the numerical model , the rayleigh equation and the linearized system of equations .
the four predicted frequencies , @xmath128 , @xmath127 , @xmath129 , @xmath130 can be well identified .
for both the numerical model and the rayleigh equation resolution , others frequencies appear for @xmath131 and @xmath132 .
these higher order frequencies are of small amplitude , almost undetectable looking at the amplitude spectrum , but have a well defined signature in the phase spectrum @xmath133 [ see fig . [ fig6](b ) ] . in the perturbative development of the rayleigh equation , i.e. , eq .
[ equationdyninertiaordre1 ] , these frequencies do not arise .
they correspond to higher order terms , which can not be revealed by the first order expansion . a development at higher orders of @xmath134 and @xmath135 would give the whole set of eigenfrequency @xmath136 +
for the short memory regime , we show that a walker placed in a two - dimensional harmonic potential well converges to a circular motion .
the mechanism of convergence involves a dissipation with two features .
first , the kinetic energy converges to the equilibrium kinetic energy .
second , the oscillations of the velocity close to its mean value relax with another time scale .
the first effect limits the time of convergence for large orbits , whereas the second one is dominant for small orbits .
the time corresponding to the transition between these two regimes is typically the time needed to travel a faraday wavelength .
the fluctuations over a distance smaller than the faraday wavelength will then be guided by this relaxation of speed oscillations .
these two transient regimes can be described by a two dimensional rayleigh oscillator .
this proves that the complex underlying hydrodynamic description is here reduced to a standard nonlinear system . from the theoretical point of view
, we highlight the stability of the circular attractor by an energetic argument where the linear stability analysis reaches its limit . only the nonlinear terms
are responsible for this stability .
nevertheless the linear expansion predicts the correct set of eigenfrequencies @xmath137 , potentially arising in any perturbation of the walker trajectory . + this approach isolates the effect of the propulsion of the walker dynamics from the other contributions , particularly a complex feedback from the memory effects . in the case of a harmonic well
, we have thus studied the low memory dynamics near @xmath56 . for the long memory regime
, the trajectories would be much more complex .
however , the fluctuations of the norm of the speed are limited so that the symmetry arguments remain relevant in the tangential direction .
consequently the dominant terms of the tangential component of the propulsive force must be of the form of the eq .
[ forcepropulsion ] , even when the memory increases . in this sense
, the rayleigh equation is a key ingredient for understanding any fluctuations in velocity as observed in several experiments @xcite .
the propulsion force can be alternatively derived from mechanical arguments . in the short memory regime ,
only the very last rebounds contribute significantly to the surface wave field @xmath138 . for the sake of simplicity
, we only retain the contribution of the last bounce .
the surface wave field @xmath138 can be expressed as @xmath139 with @xmath140 the amplitude of the field @xcite , @xmath141 the faraday wave vector and @xmath142 the bessel function of order @xmath143 .
the associated wave force is related to the local slope of the field , @xmath144 , with @xmath145 a coupling constant calculated in @xcite .
the surface force can be expressed as @xmath146 and be expanded as @xmath147 and rewritten @xmath148 the total propulsion force @xmath47 results from the loss of energy at the surface and the propulsion from the wave and can be expressed as @xmath149 @xmath150 being the apparent friction calculated in @xcite .
this expression takes the form @xmath151 with @xmath152 and @xmath153 . beyond the details of the coefficients ,
the form of this expression is the same as eq .
[ forcepropulsion ] .
this mathematical expression is relevant if we only consider one last rebound and remains a good approximation in the low memory regime . | a bouncing droplet on a vibrated bath can couple to the waves it generates , so that it becomes a propagative walker .
its propulsion at constant velocity means that a balance exists between the permanent input of energy provided by the vibration and the dissipation .
here we seek a simple theoretical description of the resulting non - hamiltonian dynamics with a walker immersed in a harmonic potential well .
we demonstrate that the interaction with the recently emitted waves can be modeled by a rayleigh - type friction .
the rayleigh oscillator has well defined attractors .
the convergence toward them and their stability is investigated through an energetic approach and a linear stability analysis .
these theoretical results provide a description of the dynamics in excellent agreement with the experimental data .
it is thus a basic framework for further investigations of wave - particle interactions when memory effects are included . |
much has been discussed about the observed spectral x - ray properties of black hole candidates ( bhcs ) and numerous models that have been developed to account for these properties , during this workshop and in the literature . however , the spectral properties are only part of what we can learn from x - ray observations of these objects .
bhcs are known to exhibit x - ray variability on a wide range of time scales from milliseconds to years ( see , e.g. , asm / rxte light curves on http://space.mit.edu/xte/asm_lc.html ) .
the temporal properties contain information on some of the same physical properties of inverse comptonization processes ( hua & titarchuk 1996 ; cui et al .
1997 ; kazanas et al .
1997 ) , which are believed to be responsible for producing hard x - ray emission from black hole candidates ( thorne & price 1975 ; coe et al .
1976 ; shapiro et al .
in addition , they add new information about the size of comptonizing regions ( hua & titarchuk 1996 ; cui et al .
1997 ) , and perhaps their density structures ( kazanas et al .
1997 ; hua et al . 1999 ; bttcher & liang 1998 ) .
unfortunately , little has been done to _ quantitatively _ model the temporal properties of black hole candidates until very recently .
i will review , in this paper , recent progress ( both observational and theoretical ) on this subject , with a focus on short - term variabilities ( @xmath0 hundreds of seconds ) which seem common among bhcs .
the discovery of rapid x - ray variability in cyg x-1 ( oda et al .
1971 ) quickly led to the speculation that such variability might be a signature for bhcs .
we now know that similar phenomenon has also been observed in some of the x - ray bursters that are known to contain a weakly magnetized neutron star ( van der klis 1995 and references therein ) . nonetheless , rapid x - ray variability is universal and typically very strong in bhcs .
an interesting question is what is the shortest time scale on which x - ray variability has been _ measured _ with high significance .
this is a very important question because the answer to it immediately constrains the dimension of the x - ray emitting region . in practice
, however , the answer depends on how well we can determine the level of poisson noise due to photon counting statistics , which in turn depends on how well we understand the physical processes that cause `` dead time '' in the electronics that collects and processes the data .
the latter is usually poorly understood , so empirical models are often constructed , based on simulations , to estimate the instrumental noise level .
for instance , the poisson noise level can be determined to within a fraction of one percent of the true value with this approach for the pca aboard rxte ( e.g. , cui et al .
no claim should , therefore , be made of any detection of significant power within this level of uncertainty .
the best studied bhc is cyg x-1 . even in this case
, significant variability has reliably been detected only up to about 200 hz ( or 5 ms ; cf .
van der klis 1995 ; see also belloni et al . 1996 , cui et al .
1997a , 1997c , and nowak et al .
1998a for results from recent rxte observations ) .
however , any bursts or flares of low duty cycles would be absent from power spectral analyses . in a time series analysis ,
the estimation of the significance of an event on a short time scale is always complicated by the effects of intrinsic variability on much longer time scales ( press & schechter 1974 ) . with this in mind , careful analyses have been conducted for cyg x-1 , and similar conclusions have been reached : significant variability is confirmed by data only for time scales longer than several milliseconds ( meekins et al .
it is , therefore , safe to conclude that rapid x - ray variability has only been _ observed _ in bhcs on time scales much longer than dynamical time scales in the immediate vicinity of central black holes ( assuming radiation field plays a negligible role in gas dynamics ) .
for bhcs , the shape of power spectral density ( psd ) correlates strongly with spectral states .
such correlation is illustrated in fig .
[ fig-2 ] . in the low ( or hard ) state ,
the psd can typically be characterized by a white noise at low frequencies that breaks into roughly a power law ( with an index @xmath1 @xmath2 ) beyond certain characteristic frequency .
the cutoff frequency , @xmath3 , typically lies in the range 0.040.4 hz .
the level of the white noise can vary while the power - law portion remains steady .
belloni & hasinger ( 1990 ) noticed an interesting anti - correlation between the white noise level and @xmath3 for cyg x-1 : as the white noise level drops , @xmath3 increases .
various subtle features of the power - density spectrum have also been reported ( cf .
van der klis 1995 ) , such as another break at a higher frequency ( a few hz for cyg x-1 in the low state ; belloni & hasinger 1990 ) which might be related to the spatial distribution of `` seed photons '' in comptonization processes ( d. kazanas , private communication ) . during a transition to the high state ,
several things seem to happen ( cui et al .
1997a , 1997b , 1997c ) : the low - frequency red noise ( or the `` 1/f '' noise ) strengthens ; the level of white noise drops ; @xmath3 increases to a few hz ; and a quasi - periodic oscillation ( qpo ) feature appears . as the high state is finally reached , the psd becomes dominated by the 1/f noise with an apparent break at high frequencies ( which might be similar to the second break in the low state ) .
it has also been noted that in some cases the psd for the very - high state bears remarkable similarities to that for transitional periods , as do the spectral properties ( cui et al .
1997c ; rutledge et al .
1998 ) .
there are exceptions to the canonical correlation .
most notable are a subclass of bhcs that never seem exhibit high - state x - ray properties ( zhang et al . 1997b ) , even during an x - ray outburst ( van der klis 1995 ; cui et al .
1997d ) , which is generally considered to be the high state for transient sources in terms of mass accretion rate .
the cause is still not clear ( see zhang et al .
1997b for a possible explanation on the spectral properties ) .
there is a general lack of systematic investigations on the energy dependence of broad - band x - ray variability for bhcs ( see a brief summary by van der klis 1995 ) . with limited statistics
, the dependence seems to be quite weak for both the low and high states .
again use cyg x-1 as an example , the recent rxte observations of the source show that the psd varies little with photon energy , both in terms of the shape and the integrated fractional amplitude , for the low state ( nowak et al .
1998a ) and high state ( cui et al .
if anything , the overall fractional amplitude appears to decrease slightly toward high energies . during the transitional period between the two states
, the energy dependence grows stronger ( cui et al .
1997a , 1997c ) : the 1/f noise strengthens toward high energies , accompanied by a drop of the white noise level .
it is worth noting that in this case the cutoff frequency ( @xmath3 ) is rather insensitive to photon energy . for the very high state
, the fractional amplitude seems to increase with photon energy ( miyamoto et al .
1991 ; rutledge et al .
1998 ) , although the energy dependence of the psd shape seems to be rather complicated ( rutledge et al . 1998 )
. the majority of bhcs are transient x - ray sources . for these sources ,
spectral state transitions are likely to be driven by an increase in the overall rate of mass accretion ( van der klis 1995 ) . typically , the x - ray intensity is more variable ( measured in terms of fractional amplitudes ) in the low state ( i.e. , low mass accretion rate ) than in the high state ( van der klis 1995 ) .
the same is also true for cyg x-1 , a persistent source , although in this case the bolometric x - ray luminosity ( i.e. , the total mass accretion rate ) varies little during the state transition ( zhang et al .
the rate of mass accretion through the disk is , however , likely to be higher in the high state , as indicated by the sudden jump in the soft x - ray flux as the source goes from the low state to the high state ( e.g. , cui et al .
1997a ; zhang et al .
therefore , the state transition in cyg x-1 is perhaps driven by a change in the relative importance between energy dissipation processes in the disk and through other channels ( such as the stellar wind ; zhang et al .
1997 ) . for several sources in the low state
, there appears to be an anti - correlation ( or correlation ) between the white noise level ( or @xmath3 ) and the _ disk _ accretion rate ( as measured by soft x - ray fluxes ) ( van der klis 1995 ) , perhaps implying the disk origin of noise processes .
it has been well established in bhcs that high - energy x - ray photons systematically lag low - energy photons ( van der klis 1995 ) .
tentative evidence for such lags was initially found in the form of asymmetries in cross - correlation functions between x - ray light curves in two energy bands for cyg x-1 ( priedhorsky et al .
1979 ) . at the time
, the importance of inverse comptonization processes was already realized and fully appreciated : the hard power law tail on the observed energy spectrum for sources like cyg x-1 was thought to be the results of soft photons being upscattered by hot electrons in an optically thin region ( thorne & price 1975 ; coe et al .
1976 ; shapiro et al .
1976 ) . in the context of comptonization models ,
the hard time lags are expected , since more scatters are required for photons to reach higher energies .
consequently , the observed hard lags were conveniently interpreted this way . to investigate hard lags for variation on different time scales , miyamoto et al .
( 1988 ) computed cross power spectrum in fourier space , using data from ginga observations of cyg x-1 in the low state .
as they derived the hard lag for each fourier frequency from the cross power spectrum , they were surprised by the strong frequency - dependence of the measured lags , which is entirely _ incompatible _ with the proposed comptonization models that had been otherwise successful in producing the observed power - law energy spectrum .
it can be shown that comptonization processes in a uniform electron gas only produce a uniform shift between light curves in different energy band ( miyamoto et al .
1988 ) ; a frequency - dependent lag would , on the other hand , require a differential `` stretching '' of light curves .
this result led miyamoto et al . to question the applicability of the proposed models for sources like cyg x-1 .
it is truly astonishing that this extremely important observational fact seems to have been completely forgotten or most likely ignored when models are developed to account for the observed _ spectral _ properties
. numerous variations of comptonization models have been proposed ever since .
success is often claimed as soon as a model seem to fit the energy spectrum , without any mention of its incompatibility with the observed frequency - dependence of hard lags ; such a model is , of course , not useful .
systematic investigations of this issue have just begun ( kazanas et al .
1997 ; hua et al . 1999
; bttcher & liang 1998 ) .
these studies have already made us realize that the incompatibility arises from the basic assumption of a uniform distribution of comptonizing electrons in nearly _ all _ models . also found is an intimate relationship between the frequency - dependence of hard lags and the density distribution ( spatial ) of hot electrons .
more specifically , the observed hard lags can be accounted for with the proper choice of a density profile for the hot electron corona .
these non - uniform corona models will be discussed in more detail in 4.1 .
hard time lags and their dependence on fourier frequency appear to be a common phenomenon for bhcs ( see review by van der klis 1995 and a summary of recent results by hua et al .
the frequency dependence is in general different for different sources , perhaps implying different density spatial profiles in the context of the non - uniform corona model ( hua et al .
in fact , there are cases where no clear dependence of lags on frequency is detected ( miyamoto et al . 1992 ) .
hard time lags have also been observed for different states .
there is evidence that the frequency dependence of lags can be different in different states for some sources ( hua et al .
1999 ) .
although the measured time lags show a general logarithmic dependence on photon energy ( i.e. , @xmath4 , where @xmath5 and @xmath6 are the effective energies of two bands between which the lag , @xmath7 , is measured ; cui et al .
1997c ; crary et al .
1998 ; nowak et al .
1998a ) , as expected from comptonization processes ( e.g. , kazanas et al .
1997 ) , the proportionality constant is dependent of fourier frequency , as noted by nowak et al .
( 1998a ) for cyg x-1 in the low state .
i have carried out similar analyses for the high state and the transitional periods , using public data from rxte observations of cyg x-1 during its 1996 state transition . for completeness ,
i have also analyzed two public rxte observations of cyg x-1 in the low state ( taken in june and october of 1997 , respectively ) .
the results are summarized in fig .
[ fig-6 ] .
the energy - dependence of the slope is apparent for all states , as clearly shown in the figure .
it is worth noting that although the hard lag is , on average , smaller in the high state than that in the low state the difference is not as much as previously thought ( cui et al .
also , the lag seems much larger during the transitions than that in either state
. this might be artificial due to the variable nature of the low state and the randomness in our selection of observations for this state for comparison
. however , our results for the low state are in good agreement with those of nowak et al .
( 1998a ) whose observations were taken shortly after the state transition .
this seems to imply the stability of underlying physical processes operating in the low state .
these results , therefore , add further complications for theoretical models to reconcile with ( a detailed discussion of these results will be presented elsewhere ; cui et al . in preparation ) .
complementary to hard lags , coherence functions are also of importance ( vaughan & nowak 1997 ) .
they measure the degree of linear correlation between x - rays at different energies in fourier domain , normalized in practice such that a value of 1 stands for perfect linear correlation and 0 for no correlation . if the measured value is less than 1 , the coherence function would not appear interesting , since numerous processes can be thought of to account for the loss of coherence .
observations show , however , that coherence functions are generally near unity for bhcs in either low or high states ( vaughan & nowak 1997 ; cui et al .
1997c ; smith et al . 1997
; nowak et al .
1998a ) , which put useful constraints on x - ray production processes and x - ray emitting environment ( hua et al .
moreover , it was found that the loss of coherence is significant during state transitions , perhaps implying the variable nature of the comptonizing region on a time scale less than the duration of an observations ( @xmath81 hour ; cui et al .
a qpo manifests itself in a well - localized broad peak in the psd .
the fractional rms amplitude is customarily quoted to specify its strength , with respect to the average x - ray intensity .
often , qpos represent transient events . while in some cases qpos are present in a well - defined range of mass accretion rate ( or x - ray flux ) or a well - defined spectral state , their presence is entirely unpredictable in other cases .
there is a variety of qpos in bhcs , in terms of their basic characteristics ( frequency , width , and strength ) as well as the dependence of these characteristics on photon energy and x - ray intensity . as an example , we show in fig .
[ fig-7 ] the power - density spectrum for grs 1915 + 105 from a public rxte observation made in may 5 , 1996 ( see also morgan et al .
a sharply peaked qpo is prominent at 0.067 hz , along with its 3 harmonics ; the qpo at 0.8 hz is rather much broader and weaker ; and even weaker is a qpo at 67 hz .
the frequency of qpos spans three decades for this source ! it should be emphasized that not all these qpos are present all the time ; a different set of qpos may be revealed by a different snapshot ( see morgan et al .
1997 for details ) .
grs 1915 + 105 has really pushed the study of x - ray variability in bhcs to an extreme . along with its cousin , gro j1655 - 40 ( mndez et al . 1998 ; remillard et al . 1998 ) ,
the two microquasars have begun to revolutionize our view of qpos in bhcs .
the study of qpos in bhcs is also one of the most important areas that rxte has contributed significantly .
prior to the launch of rxte , the qpos were observed only at very low frequencies ( @xmath9 hz ; van der klis 1995 and references therein , although rutledge et al .
1998 recently reported the detection of qpos of higher frequencies upon a re - analysis of ginga data ) , except for the very - high state qpos ( at a few hz ) .
the rxte observations have now pushed the discovery space up to a few hundred hz ( remillard et al .
of course , rich qpo phenomena are not limited to microquasars ; they have also been observed in `` normal '' bhcs , such as cyg x-1 ( cui et al .
1996 , 1997c ) , grs 1758 - 258 and 1e 17040.7 - 2942 ( smith et al .
1997 ) , 4u 1630 - 47 ( dieters et al .
1998 ) , and newly - discovered bhcs xte j1748 - 2848 and xte j1550 - 564 ( cui et al .
, in preparation ) .
the very low - frequency qpos often manifest themselves prominently in x - ray light curves , sometimes associated with `` dipping '' activities ( see fig .
[ fig-7 ] for examples ; also morgan et al .
1997 ) .
it appears that for bhcs most qpos strengthen ( in terms of fractional rms amplitude ) toward high energies ( van der klis 1995 ; cui et al . 1997c ; morgan et al .
1997 ; remillard et al .
1998 ) , although some follow the opposite trend ( morgan et al . 1997
; see also remillard et al .
1998 and rutledge et al . 1998 for other possible cases ) . in some cases
, there is a correlation between the qpo properties and the hardness of x - ray emission ( cui et al . 1997c ; remillard et al . 1998 ) .
these facts might be the key to understanding the origin(s ) of the qpos in bhcs , as will be discussed in 4.2 .
investigations of the dependence of qpos on x - ray intensity are often complicated by the ambiguity in `` labeling '' and tracking a particular qpo over a wide intensity range . with very limited statistics ,
we seem to have already seen all possibilities : some qpos show no correlation with x - ray fluxes ( cui et al .
1997c ) ; the frequency of some increases with x - ray fluxes , while that of others decreases ( remillard et al .
1998 ; rutledge et al . 1998 ) ; and , finally , certain qpos seem to maintain a stable frequency as x - ray fluxes vary ( morgan et al .
1997 ; remillard et al .
1998 ) , i.e. , the so - called `` stable qpos '' ( cui et al .
therefore , it is almost certain that no single process can account for all qpos seen in bhcs . like the broad - band variability , qpos also exhibit hard time lags ( van der klis 1995 ; cui et al . 1997c ) .
a systematic study of qpo lags has become possible only recently with the rxte discovery of a variety of qpos in bhcs . for illustration ,
[ fig-8 ] shows the measured hard lag of the 67 hz qpo in grs 1915 + 105 ( see fig .
[ fig-7 ] ) . at the qpo position , the time lag peaks at @xmath83 ms .
the hard lags have also been measured for the low - frequency qpos in grs 1915 + 105 .
[ fig-9 ] shows the measured lags as a function of photon energy for the 0.133 hz qpo ( see fig .
[ fig-7 ] ) .
it is clear that in one source the time lag associated with one qpo can be much larger than that with the others . in this case , the time lag scales roughly logarithmicly with photon energy , as would be expected from comptonization processes
. more complicated energy dependence of qpo lags have also been reported ( van der klis 1995 ) .
it appears to be generally true , however , that the larger the separation between two energy bands , the more hard x - ray photons lag soft ones .
although shot - noise models ( e.g. , terrell 1972 ; sutherland et al .
1978 ) have been popular in describing broad - band noise properties , they have shed little light on the physical origin(s ) of underlying noise processes .
the strength of such models lies in the fact that they can mimic _ mathematically _ the observed light curves for bhcs .
however , it has been noted that any stochastic time series can be simulated by a shot noise , with the proper choice of the shot profile , up to second - order statistics such as psd , auto- and cross - correlation functions ( doi 1978 ) . often , a distribution of shot profiles are required to account for subtle details ( canizares & oda 1977 ; ogawara et al .
1977 ; nolan et al . 1981 ; meekins et al . 1984 ; belloni & hasinger 1990 ; lochner et al . 1991 ; negoro et al .
for instance , the psd of cyg x-1 in the low state can drop rapidly above the cutoff frequency , following roughly a power law ; and the power - law index varies in the range @xmath2 to @xmath1 . since the shots with a profile of instantaneous rise and exponential decay can only account for a power - law index @xmath1 , a distribution of lifetimes are often required to explain the slower drop - off . then
, a mechanism would be required to filter out shots with long or short lifetimes outside of the desired range , in order to model one observed psd ; a different range would be needed to model a different psd , to an extreme that in some cases ( e.g , cui et al .
1997a , 1997c ) only shots with one particular lifetime would survive the filtering process .
this illustrates the ad - hoc nature of the models .
interesting suggestions have been made on the nature of shot - filtering mechanisms , such as a possible application of self - organized criticality ( takeuchi et al .
1995 ) , but little physics has so far been included along this line of research .
other models employing different statistical techniques of analyzing stochastic time series suffer from similar drawbacks ( e.g. , pottschmidt et al .
progress has also been made recently on shot - production mechanisms ( mantomo et al .
1998 ) , in the context of `` advection - dominated accretion flow '' models ( e.g , narayan & yi 1995 ) .
since the 1/f noise is ubiquitous in nature , it should not be surprising to see it in the process of mass accretion . for bhcs
, the noise appears to be associated with the accretion disk , because it strengthens as the _ disk _ accretion rate increases ( i.e. , the high state is approached ; cui et al .
1997a , 1997b , 1997c ) .
one model suggests that the low - frequency 1/f noise in bhcs might be attributed to the fluctuation in mass accretion rate caused by the fluctuation in local viscosity in the outer portion of the disk ( kazanas et al .
1997 ) . for bhcs in a luminous state ,
the accretion disk is likely to extend very close to the central black hole ( tanaka & lewin 1995 and references therein ) , where dynamical time scales are very short .
thermodynamic fluctuation in accretion flows in the inner region of the disk could , therefore , manifests itself in the form of white noise within limited frequency bands covered by current instrumentation ( cui et al .
the disk - origin of white noise seems to be supported by the observed anti - correlation between the relative strength of white noise and the accretion rate through the disk ( belloni & hasinger 1990 ; also see discussion in 2.2 ) .
the power spectral break can then be naturally attributed to the `` low - pass filtering '' effects of comptonization processes , using an analogy in electronics , as soft ( disk ) photons being upscattered by energetic electrons in the surroundings to produce the observed hard power - law tail ( cui et al .
in the context of this model , the cutoff frequency is determined by the characteristic escape time of soft photons from the comptonizing region , and thus provides a measure of the size of such region ( cui et al .
1997a ; kazanas et al .
1997 ) . for cyg x-1
, the model would imply that , in the context of thermal comptonization models , the hot electron corona is much smaller in the high state than in the low state , because of the much larger cutoff frequency in the high state ( cui et al .
1997a ) , which is consistent with the smaller hard lags ( cui et al .
1997c ) and softer energy spectrum in the high state , if both can also be attributed to comptonization processes . a smaller corona in the high state makes physical sense , because compton cooling power is expected to increase in this state , due to an increase in the supply of soft seed photons ( from the disk ) .
despite early doubts from miyamoto et al .
( 1988 ) , the comptonization process is still a viable and likely mechanism to account for hard lags in bhcs .
it has been realized that the frequency - dependence of the lags can be attributed to the non - uniformity of the comptonizing region ( kazanas et al .
1997 ; bttcher & liang 1998 ) .
this model has been _ quantitatively _ applied to bhcs , and the agreement with data can be quite good ( hua et al .
1999 ) . as an example , while the observed x - ray spectrum of cyg x-1 in the high state can be fit nearly equally well by models assuming three different density profiles ( uniform , @xmath10r , and @xmath11 ) for the comptonizing corona , which shows the degeneracy intrinsic to usual spectral analyses , the observed frequency - dependence of the hard lags can be fit only by assuming the @xmath12 density profile ( hua et al .
to my best knowledge , this study represents the first attempt to _ simultaneously _ model the spectral and temporal properties of a bhc .
the main drawback of such models is the requirement of maintaining a high temperature ( @xmath8100 kev ) throughout a large corona ( roughly one light second across ) .
it is not clear at present whether such conditions can be satisfied physically .
also proposed is the possibility that the hard lags are associated with the frequency - dependent propagation times of some waves from the soft x - ray emitting region to the hard x - ray emitting region ( miyamoto et al .
while the physical nature of such waves are not known , kato ( 1989 ) suggested that low - frequency corrugation waves in a thin accretion disk might be a possibility .
however , it seems nearly impossible for such models to account for the lags associated with very high - energy photons ( e.g. , grove et al . 1998 ) .
this difficulty can be alleviated by noting that the inner portion of the thin disk may be `` evaporated '' to form a geometrically thick , optically thin hot `` corona '' ( see , e.g , narayan & yi 1995 ) , which can produce high - energy photons via inverse comptonization .
however , in order to explain large lags at low frequencies , the required propagation speed seems too small to be attainable in the context of these models ( nowak et al .
1998b ) ; neither is it clear whether near perfect coherence can be preserved by the superposition of randomly - phased waves . while it is clear that qpos constitute a heterogeneous class of phenomena , their origins are hardly known .
often , qpos are attributed to processes in the accretion disk , but evidence for such a disk origin is scarce and inconclusive . in some cases ( cui et al .
1997c ) , the properties of qpos seem to correlate much better with those of hard x - rays ( perhaps of a comptonizing corona ) than those of soft x - rays ( perhaps of a disk ) .
it is , therefore , certainly premature to think of qpos as signatures for accretion disks . with these in mind
, i will list ( by no means exhaustively ) in the following some of the ideas that have been proposed over the years ( see also review by van der klis 1995 ) .
although the low - frequency qpos ( @xmath13 100 hz ) are too slow to be associated with the keplerian motion of x - ray emitting `` blobs '' in the inner portion of an accretion disk , they might be caused by some obscuring structures in the outer part of the disk ( ebisawa et al .
we would then expect to see absorption dips in x - ray light curves that are associated with the qpos ( in practice , only for very low - frequency qpos ) . while we do see intensity dips , as shown in fig .
[ fig-7 ] , the dips do not always have a harder x - ray spectrum ( due to absorption ) compared to the non - dip emission ( e.g. , greiner et al .
in fact , some of the dips in grs 1915 + 105 are thought to be related to the ejection of matter from the disk due to thermal - viscous instabilities ( belloni et al .
alternatively , the low - frequency qpos might be related to certain oscillation modes of long wavelength in the accretion disk ( kato 1990 ) .
again , the spectral properties need to be taken into account to assess the applicability of such models .
the high - frequency qpos , such as the stable qpos in grs 1915 + 105 and gro j1655 - 40 , can conceivably be associated with the keplarian motion at the last stable orbit around central black holes ( morgan et al .
1997 ) , but at least for gro j1655 - 40 , such an interpretation is incompatible with the observed spectral properties ( cui et al .
a proposal has been made to associate the stable qpos to epicyclic oscillation modes ( nowak et al .
1997 ) . not only can such modes be supported by accretion disks , they also become trapped in the innermost portion of the disk , purely due to relativistic effects ( kato & fukue 1980 ; nowak & wagoner 1992 , 1993 ; perez et al . 1997 ) .
it has also been realized that g - mode oscillations affect the largest area of a region in the disk where most x - rays are emitted and are , therefore , perhaps most relevant to observations ( nowak et al .
this model can explain the observed 67-hz qpo in grs 1915 + 105 and 300 hz qpo in gro j1655 - 40 , but does not seem to be consistent with the spectral properties either , at least for grs 1915 + 105 ( cui et al .
also proposed is that the stable qpos might originate in the gravitomagnetic precession of accretion disks surrounding spinning black holes in microquasars ( cui et al .
1998a ) . from such a model , solutions
have been derived to the spin of black holes in microquasars that are also consistent with the spectral properties ( cui et al .
finally , the high - frequency qpos might also be attributed to an inertial - acoustic instability in the disk ( chen & taam 1995 ) .
a detailed comparison of these models is beyond the scope of this review , and will be given elsewhere ( cui et al .
. certain qpos might be the observational manifestation of the oscillatory nature of comptonizing regions ( cui et al .
1997a ; kazanas et al . 1997
; titarchuk et al . 1998 ) .
evidence lies in the correlation between the properties of qpos and those of hard x - ray emission for some cases ( e.g , cui et al .
the oscillations might be shock - induced , and comptonization processes simply take place in the post - shock region ( titarchuk et al .
1998 ) . in the very high state ,
the mass accretion rate is thought to be near the eddington limit , so radiation drag might play a critical role in gas dynamics around central black holes .
the qpos observed in this state appear to be relatively stable in frequency ( van der klis 1995 ) , similar to the normal - branch qpos for z sources , so they might also be caused by radiative - driven oscillations in a quasi - spherical accretion flow ( fortner et al . 1989 ) .
most qpos are transient events , as already mentioned .
the ever - improving quality of observational data has begun to allow us to quantify the conditions in which qpos are observed .
such information might shed more light on the origin(s ) of qpos . unfortunately , none of the above mentioned models are very explicit or quantitative about actual x - ray modulation processes or excitation and damping mechanisms for oscillation or precession modes , although some makes definitive predictions on the fundamental frequency of interest .
more work is clearly required to make progress on this subject .
clearly , our knowledge about x - ray variability in bhcs is still severely limited by the lack of high - quality data , which is partly due to the transient nature of many phenomena of interest . a concerted effort among observers is urgently needed to present a complete , uniform set of _ reliable _ results , which would include such basic properties as light curves in different energy bands , auto- and cross - correlation functions , psd , hard time lags and coherence functions as a function of fourier frequency , and so on , for any particular bhc at different x - ray fluxes ( i.e. , mass accretion rates ) or in various spectral states .
not only would such results allow detailed studies of the temporal properties and their dependence on photon energy and mass accretion rate for individual sources , they would also form a valuable database for carrying out systematic investigations of bhcs as a class , from which statistically significant conclusions can be drawn on the common properties and ultimately shed light on the origins of phenomena like broad - band noise processes and qpos .
too often , only a small subset of temporal properties are presented for a bhc , despite the fact that more sophisticated and complete analyses are warranted by the quality of data .
the conclusions based such limited studies can be quite biased .
for example , there is simply not sufficient evidence to support the disk origin of _ all _ qpos .
progress can be made by correlating qpo properties with the characteristics of accretion disks ( i.e. , soft x - rays ) and comptonizing coronae ( i.e. , hard x - rays ) .
better statistics is also imperative for us to gain confidence in some of the correlations that have been presented in this review .
as the quality of data is ever improving , thanks to dedicated x - ray missions like rxte , more emphasis should also be made on the _ reliability _ of results .
this will almost certainly require significant effort to learn about the instruments with which data are taken and to understand such important instrumental effects as dead time .
for instance , as discussed in 2.1 ,
the shortest time scale on which significant x - ray variability can be detected in an observation depends critically on our ability to estimate the poisson noise level , besides other obvious factors such as the statistical quality of the data .
the estimation must take into account effects of dead time , which are usually poorly understood . however , uncertainty can usually be estimated based on simulations or comparison between the model and real data .
the uncertainty should always kept in mind when results are presented . also helpful
is the presentation of enough information so that others can evaluate the results objectively . with the accumulation of high - quality data
, theoretical effort should be focused more on modeling the observed temporal x - ray properties in bhcs .
desperately needed are detail models , like the non - uniform corona model ( kazanas et al .
1997 ) , that make _ definitive _ predictions about the properties of x - ray variability .
quantities , such as light curves , auto- and cross - correlation functions , psd , hard lags , and coherence functions , can then be computed from the model and be directly compared to data , similar to techniques in modeling spectral properties .
models will be accepted or rejected based on the goodness of the fit .
good models can then provide quantitative information on x - ray production processes and environment ( e.g. , hua et al .
1999 ) .
although certain points are worth debating , the non - uniform corona model ( kazanas et al . 1997 ) represents the first attempt to describe , in a quantitative way , the observed temporal properties in bhcs . not only
has it overcome some of the difficulties with earlier comptonization models for bhcs ( as discussed in
2.4 ) , it also sheds light on the dynamics of mass accretion in these sources ( kazanas et al .
1997 ; hua et al .
1998 ) . also attempted , for the first time , was to fit the observed spectral and temporal x - ray properties of cyg x-1 in the high state _ simultaneously _ with this model ( hua et al .
although it was still only an eyeball fit , it was the first step in the right direction . because of the degeneracy intrinsic to usual spectral analyses , simultaneous spectral and temporal modeling represents the ultimate approach to studying mass accretion processes in bhcs .
i would like to thank drs .
d. kazanas and wan chen for informative discussions and helpful comments on the manuscript .
it is a pleasure to acknowledge stimulating conversations with many participants at the workshop .
this work is supported partially by nasa through contract nas5 - 30612 and grants nag5 - 7424 and nag5 - 7484 .
belloni , t. , & hasinger , g. 1990 , , 227 , l33 belloni , t. , et al .
1996 , , 472 , l107 belloni , t. , et al .
1997 , , 488 , l109 bttcher , m. , & liang , e. p. 1998
, , in press canizares , c. r. , & oda , m. 1977 , , 214 , l119 chen , x. , & taam , r. e. 1995 , , 441 , 354 coe , m. j. , engel , a. r. , & quenby , j. j. 1976 , nature , 259 , 544 crary , d. j. , et al .
1998 , , 493 , 71 cui , w. 1996 , iau circ .
6404 cui , w. , focke , w. , & swank , j. 1996 , iau circ .
6439 cui , w. , et al .
1997a , , 474 , l57 cui , w. , et al . 1997b , proc .
second integral workshop , eds : c. winkler et al .
malo , france ) , esa sp 382 , 209 ( astro - ph/9610072 ) cui , w. , et al . 1997c , apj , 484 , 383 cui , w. , et al . 1997d , , 487 , l73 cui , w. , zhang , s. n. , & chen , w. 1998a , , 492 , l53 cui , w. , chen , w. , & zhang , s. n. 1998b , proc .
third william fairbank meeting on the lense - thirring effect , eds .
l. - z. fang , & r. ruffini ( rome , italy ) dieters , s. , belloni , t. , & kuulkers , e. 1998 , iaucirc .
6823 doi , k. 1978 , nature , 275 , 197 ebisawa , k. , mitsuda , k. , & inoue , h. 1989 , , 41 , 519 fortner , b , lamb , f. k. , & miller , g. s. 1989 , nature , 342 , 775 greiner , j. , morgan , e. h. , & remillard , r. 1996 , , 473 , l107 grove , j. e. , et al .
1998 , , 502 , 45 hua , x .-
titarchuk , l. 1996 , , 469 , 280 hua , x .- m . , kazanas , d. & titarchuk , l. 1997 , , 482 , l57 hua , x .- m . , kazanas , d. & cui , w. 1999 , , 512 , in press kato , s. 1990 , , 42 , 99 kato , s. , & fukue , j. 1980 , , 32 , 377 kazanas , d. , hua , x .-
& titarchuk , l. 1997 , , 480 , 735 lochner , j. c. , swank , j. h. , & szymkowiak , a. e. 1991 , , 376 , 295 mantomo , t. , et al .
1998 , , 464 , l135 meekins , j. f. , et al . 1984 , , 278 , 288 mndez , m. , belloni , t. , & van der klis , m. 1998 , , 499 , l187 miyamoto , s. , et al .
1988 , nature , 336 , 450 miyamoto , s. , et al .
1991 , , 383 , 784 miyamoto , s. , et al .
1992 , , 391 , l21 morgan , e. h. , remillard , r. a. , & greiner , j. 1997 , , 482 , 993 narayan , r. , & yi , i. 1995 , , 452 , 710 negoro , h. , miyamoto , s. , & kitamoto , s. 1994 , , 423 , l127 nolan , p. l. , et al .
1981 , , 246 , 494 nowak , m. a. , & wagoner , r. v. 1992 , apj , 393 , 697 nowak , m. a. , & wagoner , r. v. 1993 , apj , 418 , 187 nowak , m. a. , et al .
1997 , , 477 , l91 nowak , m. a. , et al . 1998a , , in press nowak , m. a. , et al .
1998b , proc .
`` the active x - ray sky '' ( rome , italy ) , in press ( astro - ph/9712106 ) oda , m. , et al .
1971 , , 166 , l1 ogawara , y. , et al .
1977 , nature , 270 , 154 perez , c. a. , et al .
1997 , apj , 476 , 589 remillard , r. a. , et al .
1998 , , in press ( astro - ph/9806049 ) pottschmidt , k. , et al .
1998 , , in press priedhorsky , w. , et al . 1979 , , 233 , 350 rutledge , r. e. , et al .
1998 , , in press ( astro - ph/9807235 ) shapiro , s. l. , lightman , a. p. , & eardley , d. m. 1976 , , 204 , 187 smith , d. m. , et al .
1997 , , 489 , 51 sutherland , p. g. , weisskopf , m. c. , & kahn , s. m. 1978 , , 219 , 1029 takeuchi , m. , mineshige , s. , & negoro , h. 1995 , , 47 , 617 tanaka , y. , & lewin , w. h. g. 1995 , in `` x ray binaries '' , eds . w.
h. g. lewin , j. van paradijs , & e. p. j. van den heuvel ( cambridge u. press , cambridge ) p. 126 terrell , n. j. 1972 , , 174 , l35 thorne , k. s. , & price , r. h. 1975 , , 195 , l101 titarchuk , l. , lapidus , i. , & muslimov , a. 1998 , , 499 , 315 van der klis , m. 1995 , in `` x ray binaries '' , eds .
w. h. g. lewin , j. van paradijs , & e. p. j. van den heuvel ( cambridge u. press , cambridge ) p. 252
vaughan , b. a. & nowak , m. a. 1997 , , 474 , l43 press , w. h. , & schechter , p. 1974 , , 193 , 437 zhang , s. n. , et al .
1996 , iau circ .
6462 zhang , s. n. , et al . 1997a , , 477 , l95 zhang , s. n. , cui , w , & chen , w. 1997b , , 482 , l155 | i review the observed temporal x - ray properties of galactic black hole candidates and the status of theoretical modeling of these properties . like spectral x - ray properties , which has so
far attracted most of the attention , the temporal properties , such as power spectral density , hard x - ray time lags , and coherence functions , contain information on some of the same physical properties of inverse comptonization processes , which are generally believed to be responsible for producing hard x - ray emission in black hole candidates .
in addition , the temporal properties add new information about the size of the comptonizing region and perhaps its density structure . therefore , the spectral and temporal properties taken together may help break the degeneracy , intrinsic to most spectral studies , in determining the temperature and optical depth of hot electrons .
moreover , significant insights might be gained into the dynamics of mass accretion in black hole candidates from the inferred density structure of accreted gas ( from the temporal properties ) . unfortunately , until very recently little has been done to _ quantitatively _ model the observed temporal properties of black hole candidates .
i will discuss the recently proposed models , which have been formulated in the context of thermal comptonization , and compare them to the data .
emphasis will be made on simultaneous studies of the observed spectral and temporal properties , the ultimate approach to studying x - ray production mechanisms in galactic black holes . |
in recent years the cooperative interaction of @xmath0 two level radiators with cavity electromagnetic field ( emf ) was in the attention of many experimental and theoretical researches @xcite .
this it is connected with the big application of two - level system as a q - bit in quantum processing of information . in many cases , distinguished ensembles of q - bits
are used for the realization of quantum registers @xcite . according to the principle of indistinguishable between the radiators @xcite ,
@xmath1 states of @xmath0 two - level atoms can be reduced to @xmath2 states in the processes of coherent excitation . as a particular case ,
the cooperative effect between two undistinguished radiators occurs , due to the fact that two states of the single excitation ( the atom @xmath3 in an excited state and @xmath4 in the ground and respectively atom @xmath3 in the ground state and @xmath4 in the excited state ) are considered as a same collective state .
it is not difficult to observe that this number of collective excitation drastically is reduced with increasing the number of atoms .
it is attractive from physical points of view to apply this principle to the atoms placed in coupled cavities .
combination of atom - cavity physics and photon systems offers new opportunities in this field for a better device functionality and for probing of emulators of condensed - matter systems . in ref .
@xcite a single - polariton approximation was proposed for the study of the periodical photonic systems . within this approximation
the authors of this paper have applied bloch states to the uniformly tuned jaynes - cummings - hubbard model to analytically determine the energy - band structure . in the work @xcite it was studied the photon - blockade - induced mott transitions and xy spin models in coupled cavity arrays .
it was found that a range of many - body effects such as a mott transition for polaritons obeying mixed statistics could be observed in the optical systems of individual addressable coupled cavity arrays interacting with two - level systems .
exploited first in systems like photons , atoms , or ions , the entanglement is becoming more and more attainable in condensed matter physics .
as follows from the distinctive description of doped cavities @xcite,@xcite , the analytical descriptions of the quantum systems become complicated due to the increasing number of degrees of freedom of such systems by increasing the number of cavities in the system . in this case the number of degrees of freedoms is connected with the numbers of coupled quantum oscillators : atoms and cavity - modes . considering that the excitations of atoms and cavity modes become indistinguishable , we can use the symmetry transformation of such system considering the invariance of the states after the actions of the operation of corresponded symmetry group .
for example , the wave function of the single excitation of the radiator or mode inside of one of the three coupled cavities must remain in the same quantum state after the rotation around the axis of symmetry with angle @xmath5 and @xmath6 .
as it is observed , the number of degrees of freedom of the system formed from three coupled cavities is reduced as in the dicke problem for three radiators situated in the volume with the dimension less than the radiation wavelength . in sec.[sec:2
] we have proposed to use the symmetry principle for reducing the number of degrees of freedom of two and three coupled cavities . due to this
, we have obtained the exact solution for the schrodinger equation in interaction picture . introducing the losses from the cavities , the master equation for undistinguished excitation of two and more cavities
is studied in the sec.[sec:3 ] .
in this section we apply the symmetrical propriety of undistinguished atoms and photons packed in coupled cavities . for comparison
we consider the system of atoms situated at a distance less than the emission wavelength as in the dicke model @xcite . for simplicity
we consider the system of two atoms .
these distinguished atoms are characterized by four states : @xmath7 ; @xmath8 ; @xmath9 and @xmath10 . here
@xmath11 and @xmath12 are excited and ground states of @xmath13 atom , @xmath14 . considering that one of the atoms from the ensemble is in excited state we construct the collective excitation state @xmath15 which together with other two states @xmath16 and @xmath17 form three atomic collective states .
it is important to emphasize , that the rotation of this ensemble at angle @xmath18 these collective states remain the same . for three atoms situated in vertexes of equidistant triangle , we can obtain four collective states @xmath19 , @xmath20 , @xmath21 and @xmath22 from @xmath23 distinguished @xmath24 , @xmath25 , @xmath26 , @xmath27 , @xmath28 , @xmath29 , @xmath30 and @xmath31 . as follows from this description of two collective states @xmath20 , and @xmath21
, we observe that they can be regarded as a superposition of three distinguished states @xmath32 , @xmath33 , and are invariant under the angle rotations @xmath34 and @xmath35 .
let us return to the coupled ensemble of n doped cavities . in comparison with
the above situation we have two localized excitations in each cavity .
first of them is connected with atomic excitation and second with mode excitation in each cavity . in order to apply the proposed symmetry we introduce the hamiltonian of such coupled cavities in the interaction with the external electromagnetic field . indeed considering
the @xmath36 single mode coupled cavities doped with radiators as this is represented in the fig .
[ figure ] ) , we can introduce the following hamiltonian @xmath37 in which the free part describes the energies of the atomic inversion , single - mode emf of each cavity @xmath38 the interaction hamiltonian contains the following couplings : between the atom and the mode of each cavity , between the cavities [ figure ] @xmath39 here @xmath40 is the resonant frequency between two level radiators and modes of the electromagnetic field , @xmath41and @xmath42 are the creation and annihilation boson operators of emf , @xmath43 , @xmath44 and @xmath45 are the excitation , de - excitation and inversion atomic operators for the @xmath46-cavity .
these operators satisfy the commutation relation @xmath47 = 2\delta _ { i , j}r_{zi } $ ] , @xmath48 = \pm r_{i}^{\pm } $ ] .
the coupling between the cavities is achieved by the second term of the interaction hamiltonian described by the exchange parameter @xmath49 .
following the idea , that the system formed from @xmath36 cavities must remain in the same state after the application of rotation symmetry to regular polygon in the vertexes of which are situated the coupled cavities , we introduce the wave functions of the hamiltonian ( [ eq : h0 ] ) .
let us neglect in first approximation the coupling of the cavities with external emf . solving the problem in higher q cavity limit we will introduce the losses from the cavities in the next section .
let us below apply these symmetry proprieties of two and three cavities .
_ two cavities situation .
_ in the good cavity limit , for single excitation in both cavities it is easy to observe @xmath50 wave functions for disinfested atoms and cavity modes described by free hamiltonian ( [ eq : h0 ] ) : @xmath51 , @xmath52 , @xmath53 , @xmath54 .
for undistinguished situation we obtain two wave functions @xmath55 for two excitations in two cavities instead of the @xmath23 wave functions @xmath56 , @xmath57 , @xmath58 , @xmath59 ; @xmath60 ; @xmath60 ; @xmath60 ; @xmath53 we obtain @xmath61 undistinguished @xmath62 ; \nonumber \\
\left\vert \psi _ { 21}\right\rangle & = & \left\vert e,0\right\rangle _ { 1}\left\vert e,0\right\rangle _ { 2};\ \
\left\vert \psi _ { 25}\right\rangle = \left\vert g,1\right\rangle _ { 1}\left\vert g,1\right\rangle _ { 2 } ; \label{eq : wf22}\end{aligned}\ ] ] here @xmath63 for symmetrical functions and @xmath64 for anti - symmetrical case , @xmath65 is the state of @xmath66 degenerate level of energy , where @xmath67 is the number of excitations in the system , @xmath68 and @xmath69 indicate the atomic and field excitations in the @xmath46 cavity respectively .
this wave functions ( [ eq : wf12 ] ) and ( [ eq : wf22 ] ) are invariant under the rotation of the system with angle @xmath70 . for the rotation of the system of two cavities at angle @xmath70 we must replace the indexes of the cavities between them @xmath71 .
it is easy to observe that for one quanta of energy we have and other superposition functions for two coupled cavities ( one photon in first and the atom - excitation in second and vice versa ) . but
these wave functions represent a superposition of collective excitations of atomic and field states of both cavities .
let consider that the degenerate states for single ( [ eq : wf12 ] ) and two excitations ( [ eq : wf22 ] ) are weak split by the interaction hamiltonian ( [ eq : hi ] ) , so that this splitting is less than the energy distance between first and second excited states of the cavities . representing the non - stationary wave function through the superpositions of the single ( [ eq : wf12 ] ) or two excitations ( [ eq : wf22 ] ) states @xmath72=@xmath73\left\vert \psi_{j}\right\rangle $ ] , we can find the new dressed states of the system solving the stationary schrodinger equation in the interaction picture @xmath74 .
the wave functions for single collective excitations are @xmath75|\psi _ { 1j}\rangle $ ] . here
@xmath46 indicate the new energy level , the eigenvalues are @xmath76 and the superposition coefficients @xmath77 are labeled so that the first index indicates the new level of energy , the index in parentheses indicate the number of excitations from the cavities . for two excitations in the system
we have the wave function @xmath78|\psi _ { 2j}\rangle $ ] . here
the solutions of characteristic equations are @xmath79 and the new coefficients for two collective excitations can be found from the normalization condition @xmath80 here indexes of @xmath81 : @xmath82 , @xmath36 and @xmath46 indicate the number of quanta of energy in the system , the new quasi - energy level number corresponding to the energy @xmath83 and the state @xmath84 used in superpositions .
the expressions ( [ eq : l1 ] ) , [ eq : l2 ] ) and ( [ eq : c1 ] ) ( [ eq : c2 ] ) represent the frequencies of new quasi - levels of energy and wave function coefficients for single and two excitations . _ three cavity case . _ for three coupled doped cavities we can obtain in the similar way the wave function of the single excitation .
this methodology can be extended for larger numbers of excitations representing the eigenstates and wave functions numerically . taking this into account
it is studied only the master equation for one excitation in system , for which is possible to obtain analytical solutions .
the degenerate states for one excitation are @xmath85 we observe and other superpositions functions of the single excitation in the three cavities .
these are the superposition with one photon in one of them and atomic excitation in another .
similarly , it is founded the eigenstates and eigenvalues of interaction hamiltonian @xmath86 so , the analyzed interaction allows us to get two new quasi - energy levels . for two excitations
we observe five collective states @xmath87 and same number of the solutions for characteristic equation for @xmath88 are @xmath89 , @xmath90 , @xmath91 , @xmath92 , @xmath93 . here
for three excitation , from 38 states we obtain 10 collective states .
it is clear that for two or more excitations in three cavities we can solve the schrodinger equation only numerically . _
n cavity case .
_ let us considerate that we have n cavities placed at the vertices of a regular polygon with n equal sides . if the number of cavities is large as this is represented in fig.[figure ] , according to the interaction hamiltonian ( [ eq : hi ] } ) , for one undistinguished excitation of atoms or photons we observe the following degenerates wave functions @xmath95is not difficult to observe that the actions of the interaction hamiltonian on these functions are @xmath96 and @xmath97 . in this case , the eigenvalues and eigenfunctions of interaction hamiltonian are equivalent to the three cavities case @xmath98 and @xmath99,@xmath100for@xmath101 . in conclusion , we observe the following behavior of the excited atoms in two or more cavities .
two excitations in two cavities are described by 5 collective states . in three cavities ,
three excitations are described by 10 collective states .
so we can observe that for small number of excitations can be applied the following expression for the number of the collective states @xmath102 , where @xmath103 is the number of excitations in the system .
this behavior is applicable for number of excitations @xmath103 less than the number of cavities .
this law is violated beginning with four excitations .
indeed for four excitation it is not difficult to observe that we can construct 19 collective states . according to our representation this corresponds to the number @xmath104 . in our opinion
this modification is connected with the packing of the cavities in the space , the atomic and photon excitations between cavities .
for example four cavities can be placed in the circle and , of course , these cavities can be packed in the vertices of an regular tetrahedron .
so we have two different types of symmetry , two cases with different number of collective states for four excitations .
let us obtain now the master equation for collective excitations of two radiators placed in two coupled cavities with the losses .
in order to find the master equation for such system let us introduce in the hamiltonian ( [ eq : hi ] ) the coupling of cavity modes with vacuum of emf @xmath105 . here
the photon losses @xmath106 indicate the transformation of cavity photons in the free emf photons described by the new annihilation and creation boson operators @xmath107 and @xmath108 . using the collective excitation states ( [ eq : c1 ] , [ eq : c2])obtained in the first section we can represent the cavity field operators through the new collective states . considering that the energy is measured from the relatively level @xmath109 ( see fig.[figure ] )
, the hamiltonian for two coupled cavities can be represented as @xmath110 @xmath111 , where @xmath112 here @xmath113 are the hilbert vectors of collective excitation , it is introduced re - normalized frequencies @xmath114 , where @xmath115 and @xmath116 , @xmath117 . in hamiltonian
we have considered that the action of the creation operators in the state with maximal number of excitations @xmath118 is neglected .
@xmath119 due to the fact that such an excitation is impossible in the system in interaction with the external vacuum field . considering the large number of degrees of freedom of free em vacuum
, we can eliminate from the density matrix equation the boson operators of external emf .
in the interaction picture the density matrix equation @xmath120 $ ] where @xmath121\hat{h}_{i}\exp [ -i\hat{h}_{0}t/\hbar ] $ ] is the hamiltonian in interaction picture . using the method of the projection operator on the vacuum field @xmath122
we can represent the density matrix through slower @xmath123 and rapidly oscillating @xmath124 parts respectively , @xmath125 .
it can be shown that @xmath126 and @xmath127 . following the well known procedure of the elimination of the rapid oscillatory part of the density matrix @xcite
, one can obtain the equation of the density matrix @xmath128 for the cavity excitations @xmath129
\nonumber \\ & = & -\gamma \lbrack \check{w}(t)\left ( c_{1}\check{u}_{0}^{1,1}(t)+c_{2}\check{u}_{0}^{1,-1}(t)\right ) , c_{1}\check{u}_{1,1}^{0}(t)+c_{2}\check{u}_{1,-1}^{0}(t ) ] \nonumber \\ & + & h.c .
\label{4}\end{aligned}\ ] ] here @xmath130 @xmath131 , @xmath132 is the ladder operator which indicate the transition between the new quasi - levels for a single cavity - atom excitation .
these operators satisfy the commutation relation @xmath133=\delta _ { \beta , \mu } \check{u}_{\upsilon } ^{\alpha } -\delta _ { \alpha,\upsilon } \check{u}_{\beta } ^{\mu } $ ] . the generalized equation ( [ 4 ] ) takes into consideration the cooperative decay processes of two cavities in the interaction with the external vacuum field .
let us study the decay rate of single collective excitation placed in the two or three cavities .
considering that the new collective states @xmath134 and @xmath135 can be easily prepared , we found the following closed system of equations for the population and correlation of photons and radiators @xmath136 u(\tau ) -\frac{p}{2}(x(\tau ) + y(\tau ) ) , \\
\frac{dw(\tau ) } { d\tau } & = & 2qu(\tau ) -\left [ p^{2}+1\right ] w(\tau ) .\end{aligned}\]]here @xmath137 , @xmath138 , @xmath139 , @xmath140 , @xmath141 , @xmath142 and @xmath143 . for simplicity
, we have represent the relaxation times of two transitions from two upper to the ground states through the constants @xmath144 .
the evolution of the population of new quasi - levels of energy is studied considering the situation when only the state @xmath134(@xmath145 ) is populated while the second excited state @xmath135 is un - populated .
it is remarkable for physical point of view the partially transfer of excitation from the state @xmath134 to the state @xmath135 .
this transfer is represented in the fig .
2 : in which figure @xmath146represent the nutation of the excitation between the upper states @xmath134 and @xmath135 .
this nutation is accompanied with the transfer of the excitation to the state @xmath135 which in the processes of interaction with vacuum field becomes entangled with the state @xmath134 . during the decay time , both states @xmath147 and @xmath135 become unpopulated .
[ figure_2 ] in order to obtain the entropy of excitations in the cavities let us represent the density matrix through the excitation correlations @xmath148the entropy of the cavity excitations is described by the expression @xmath149 where @xmath150 @xmath151and @xmath152 . from the time - dependence of entropy and inversion , we observe what then the population of excited levels becomes equal to the population of the ground state , the entropy takes the maxim value .
this maximum corresponds to the creation of an entangled state between excited and ground states(see fig.[figure_4 ] ) .
another interesting effect is the decay process of excited states .
here is possible two situations , first of them corresponds to the preparation of excited states in the superposition of the states @xmath134 and @xmath135 . in this case
the entropy become increases from non zero value @xmath153 . preparing the system in only one excited state
, we can observe , during the process of spontaneous emission that the excitation partially `` jump '' in the second excited state .
@xmath134 when the population of another state @xmath154 . in this case
, after the decay process of spontaneous emission , the electron jump on the unpopulated state .
similar effect is known in the literature and experimentally observed in three levels systems @xcite .
in this paper we have introduced the collective excitations of emf of coupled cavities in interaction with two level radiators placed in each cavity ( [ eq:3c ] , [ eq:3cv ] ) . of course
, with increasing the number of excitations , the problem becomes difficult in the analytical representation ( see two cavities with two excitations ( [ eq : l2 ] , [ eq : c2 ] ) ) . unlike the traditional approach refs @xcite it examines the behavior of two collective excitations of coupled cavities taking into consideration local symmetry of this system , ( see ( [ eq : l2 ] ) and ( [ eq : c2 ] ) ) .
taking into consideration the undistinguished principle between the atomic and photon excitations this paper proposes to introduce the collective excitations for the description of the interaction q - bits . for this we proposed the exact solutions of two excitations in two cavities ( [ eq:3c ] ) and ( [ eq:3cv ] ) . considering that
the energy of collective dressed states of atom + cavity field is larger than the decay rates from the cavities we have introduced the new interaction hamiltonian of collective excitations with external vacuum field ( [ eq : him ] ) . from analytical and numerical solutions follows that the solution of a damped system of equations depends on the initial preparation of the collective states @xmath155 and @xmath156 . in comparison with other approaches
we observe the damped oscillations of the population inversions between two excited states as it is represented in the fig.[figure_2 ] .
it is observed two limits for this association .
first limit @xmath157 corresponds to the bad coupling between the cavities and second limit @xmath158 is connected to high coupling between the cavities . | the exact solution of the system consisted from two or three q - bits doped in coupled cavities is discussed .
the problem of indistinguishable between the excited radiators and photons is analyzed using the intrinsic symmetry of the system .
it is demonstrated that the solution is drastically simplified when the radiators and photons are considered as a new polariton excitations .
the exact solution of schrodinger equation is obtained for single and two excitations in each cavity taking into consideration the indistinguishable principle .
this approach opens new possibilities in the interpretation of quantum entangled states in comparison with the traditional distinctive situation ( see for example @xcite ) due to the decreasing of the number of degrees of freedoms in the system . considering that the energies of coupling between the radiators and photons is larger than the coupling with external vacuum field
, we have found the master equation for the dumping of collective excitations of the system of coupled radiators through the cavity fields .
the time - dependence of population for new dressed quasi - levels of energy is obtained solving analytically and numerically the master equation . |
brown dwarfs substellar objects with masses below the hydrogen burning limit of 0.08@xmath1 are ideal to test the mass dependence of critical parameters in stellar evolution .
one example for such a parameter is the lifetime of circumstellar disks , which is an important constraint for the core - accretion scenarios for planet formation .
the disk lifetime is affected by a number of physical processes , e.g. disk ionisation by the central object and cosmic rays , accretion , grain growth ( e.g. @xcite ) .
our understanding of the relative importance of these processes and how they change with object mass is still incomplete , i.e. observational guidance is important to advance the theory .
the clear majority of low - mass stars lose their disk within less than 5myr @xcite .
maybe the best test for the longevity of disks is the upper scorpius ob assocation ( upsco in the following ) , the oldest nearby star forming region with a substantial number of brown dwarfs .
upsco is often assumed to have an age of 5myr @xcite but recently @xcite have derived an older age of 10myr for this region . in upsco , @xcite derived a disk frequency of @xmath2% for f and g stars and @xmath3% for k0-m5 stars ( with 1@xmath4 binomial confidence intervals ) . the brown dwarfs in this region exhibit a disk fraction of @xmath5% , based on an examination of 35 objects @xcite .
thus , the spitzer data tentatively shows that the disk fractions _ in upsco increase monotonically with decreasing object mass for early f- to late m - type objects_. this would imply a mass dependence in the disk evolution , resulting in long - lived disks in the substellar regime .
so far , the brown dwarf disk frequency in upsco is affected by low number statistics . here
we set out to test previous findings based on a much enlarged number of brown dwarfs in upsco , identified from ukidss . to identify the objects in our sample that have a disk ( class ii objects ) and those that do not ( class iii objects ) , we use data from the wide - field infrared survey explorer ( wise)@xcite . as will be shown , with improved statistics we find a disk fraction for brown dwarfs that is consistent with the value published for low - mass stars in this region . | we present a census of the disk population for ukidss selected brown dwarfs in the 5 - 10 myr old upper scorpius ob association . for 116 objects originally identified in ukidss ,
the majority of them not studied in previous publications , we obtain photometry from the wise database . the resulting colour - magnitude and
colour - colour plots clearly show two separate populations of objects , interpreted as brown dwarfs with disks ( class ii ) and without disks ( class iii ) .
we identify 27 class ii brown dwarfs , 14 of them not previously known .
this disk fraction ( 27 out of 116 or 23% ) among brown dwarfs was found to be similar to results for k / m stars in upper scorpius , suggesting that the lifetimes of disks are independent of the mass of the central object for low - mass stars and brown dwarfs .
5 out of 27 disks ( 19% ) lack excess at 3.4 and 4.6@xmath0 m and are potential transition disks ( i.e. are in transition from class ii to class iii ) .
the transition disk fraction is comparable to low - mass stars .
we estimate that the timescale for a typical transition from class ii to class iii is less than 0.4myr for brown dwarfs .
these results suggest that the evolution of brown dwarf disks mirrors the behaviour of disks around low - mass stars , with disk lifetimes on the order of 5 - 10myr and a disk clearing timescale significantly shorter than 1myr .
[ firstpage ] techniques : photometric techniques : brown dwarfs open clusters and associations : individual : upper scorpius infrared : stars . |
the current observed neutrino mixing @xcite suggests around the maximal 2 - 3 mixing angle and zero 1 - 3 mixing angle : @xmath1 , @xmath2 .
in such a symmetric limit where both @xmath3 and @xmath4 vanish , the resulting @xmath5 effective majorana neutrino mass matrix forms in the flavor basis as @xcite ( ccc x & c & c + c & a & b + c & b & a ) .
[ m - z2 ] this matrix has an exact symmetric form under a @xmath6 symmetry , i.e. the 2 - 3 ( @xmath7-@xmath8 ) permutation , and is diagonalized by the unitary matrix : u_z_2 = ( ccc _ 12 & -_12 & 0 + _ 12/ & _ 12/ & -1/ + _ 12/ & _ 12/ & 1/ ) , with remaining the solar mixing angle @xmath9 arbitrary and the entry @xmath10 of ( [ m - z2 ] ) is determined as x = a + b + .
now we well know two special values for @xmath9 which give typical mixing matrices ; one is bimaximal and the other is tri - bimaximal mixing matrices . in a limit of bimaximal mixing @xcite where @xmath11 , resulting mns matrix forms u_bm = ( ccc 1/ & -1/ & 0 + 1/2 & 1/2 & -1/ + 1/2 & 1/2 & 1/ ) , with x = a + b. for the case @xmath12 with so - called tri - bimaximal mixing which is proposed by harrison , perkins and scott , then we have the hps type matrix @xcite : u_hps = ( ccc 2/ & 1/ & 0 + -1/ & 1/ & -1/ + -1/ & 1/ & 1/ ) , where @xmath9 is fixed by as well x = a + b + c is also derived .
note that the tri - bimaximal structure is consistent with current experimental data , where @xmath9 is not maximal .
so far , the discrete symmetry @xmath0 is successfully applied for leptons .
namely , the tri - bimaximal mixing pattern can be realized naturally in a number of specific models @xcite .
however , it is not easy to have small quark mixing angles in naive and straight way . in such applications
the generic prediction @xcite is that @xmath13 , the quark mixing matrix becomes just the unit matrix . starting with @xmath14 ,
the realistic small quark mixing angles can be generated by extending interactions beyond those of the standard model , such as in supersymmetry @xcite or breaking @xmath0 symmetry explicitly @xcite .
our study is aimed to obtain realistic quark masses and mixing angles entirely within the @xmath0 context @xcite .
it is worthy of mention that the other types of unified models for quarks and leptons with the @xmath0 symmetry @xcite and models which predict tri - bimaximal mixing matrix with @xmath15 symmetry @xcite have been also studied .
@xmath0 is the symmetry group of the tetrahedron and the finite groups of the even permutation of four objects .
it has twelve elements which are derived into four equivalence classes : [ @xmath16 : ( 1234 ) , [ @xmath17 : ( 2143 ) , ( 3412 ) , ( 4321 ) , [ @xmath18 : ( 1342 ) , ( 4213 ) , ( 2431 ) , ( 3124 ) and [ @xmath19 : ( 1423 ) , ( 3241 ) , ( 4132 ) , ( 2314 ) , corresponding to its four irreducible representations we call three one - dimensional representations ( singlets ) as @xmath20 , @xmath21 , @xmath22 , and one three - dimensional representation ( triplet ) as @xmath23 , respectively .
the @xmath0 is the smallest discrete group which includes the three - dimensional irreducible representation .
the presence of the three - dimensional irreducible representation might be ideal for describing three families of quarks and leptons .
the character table of four representations is shown in table [ table : ct ] . here
@xmath24 is the order of each element , @xmath25 is the number of elements and the complex number @xmath26 is the cube root of unity : = ( 2 i/3 ) = - + i , 1 + + ^2 = 0 . .character table of @xmath0 .
[ table : ct ] [ cols="^,^,^,^,^,^,^",options="header " , ] in minimal su(5 ) , there is just one @xmath27 representation of higgs bosons , yielding thus only two invariants , i.e. @xmath28 ( that is for up quark mass matrix ) and @xmath29 ( for charged lepton and down quark mass matrices ) . the second invariant implies @xmath30 at the unification scale which is phenomenologically desirable . we follow the usual strategy of using both @xmath31 and @xmath32 representations of higgs bosons , so that one linear combination couples to only leptons , and the other only to quarks .
both transform as @xmath23 under @xmath0 .
there are also two @xmath27 representations transforming as @xmath21 and @xmath22 under @xmath0 which couple only to up type quarks and @xmath22 , other patterns can be assigned with different prediction . ] . with this @xmath0 assignment for higgs doublets ,
the relevant yukawa couplings linking @xmath33 with @xmath34 are given by & & h_1 d_1 ( d^c_1 ^0_d1 + d^c_2 ^0_d2 + d^c_3 ^0_d3 ) + h_2 d_2 ( d^c_1 ^0_d1 + d^c_2 ^0_d2 + ^2 d^c_3 ^0_d3 ) + & & + h_3 d_3 ( d^c_1 ^0_d1 + ^2 d^c_2 ^ 0_d2 + d^c_3 ^0_d3 ) , resulting in the @xmath5 down quark mass matrix : m_d = ( ccc h_1 & 0 & 0 + 0 & h_2 & 0 + 0 & 0 & h_3 ) ( ccc 1 & 1 & 1 + 1 & & ^2 + 1 & ^2 & ) ( ccc v_1 & 0 & 0 + 0 & v_2 & 0 + 0 & 0 & v_3 ) , where @xmath35 , @xmath36 are three independent yukawa couplings , and @xmath37 are the vacuum expectation values of @xmath38 . to get quark mass hierarchy , @xmath39 should be satisfied contrary to the lepton sector ( @xmath40 ) .
by contrast , the higgs doublets transforming as @xmath21 and @xmath22 linking @xmath41 with @xmath42 give , as following the manner of eq .
( [ 1x1x1 ] ) , the @xmath5 symmetric up quark mass matrix : m_u = ( ccc 0 & _ 2 & _
3 + _ 2 & m_2 & 0 + _ 3 & 0 & m_3 ) , where @xmath43 , @xmath44 come from @xmath45 and @xmath46 , @xmath47 from @xmath48 and three of them can be taken real in general . in the limit @xmath49 and @xmath50
, we obtain the three eigenvalues of @xmath51 as m_t |m_3| , m_c |m_2| , m_u |+ | , with mixing angles v_uc , v_ut , v_ct 0 . in the down sector
, we note that m_dm_d^= ( ccc y |h_1|^2 & z^h_1 h_2^ & z h_1 h_3^ + z h_1^h_2 & y |h_2|^2 & z^h_2 h_3^ + z^h_1^h_3 & z h_2^h_3 & y |h_3|^2 ) , where y & = & |v_1|^2 + |v_2|^2 + |v_3|^2 , + z & = & |v_1|^2 + |v_2|^2 + ^2 |v_3|^2 .
its eigenvalue @xmath52 satisfies the equation & & ^3 - y(|h_1|^2 + |h_2|^2 + |h_3|^2)^2 - ( y^3+z^3+z^3 -3y|z|^2)|h_1|^2|h_2|^2|h_3|^2 & & + ( y-|z|^2)(|h_1|^2|h_2|^2 + |h_1|^2|h_3|^2 + |h_2|^2|h_3|^2)= 0 . if @xmath53 as well as assumed in the charged lepton case , then @xmath54 , @xmath55 and three eigenvalues are simply @xmath56 .
we choose them instead to be different , but we still assume @xmath57 . in that case , we find m_b^2 & & y|h_3|^2 , + m_s^2 & & ( ) |h_2|^2 , + m_d^2 & & ( ) |h_1|^2 , and the mixing angles are given by v_sb & & ( ) , + v_db & & ( ) , + v_ds & & ( ) , thereby requiring the condition |
| | | . using current experimental values for the left - hand side
, we see that quark mixing in the down sector alone can not explain the observed quark mixing matrix @xmath13 . taking into account @xmath58
, we then have v_ckm = v_u^v_d .
hence v_us & & v_ds - v_uc ( ) - , [ vus ] + v_cb & & v_sb ( ) , [ vcb ] + v_ub & & v_db - v_uc v_sb - v_ut ( ) - ( ) - .
[ vub ] it is noted that our up and down quark mass matrices are restricted by our choice of @xmath0 representations to have only five independent parameters each . in the up sector , we have three real and one complex parameters ( for example we choose here @xmath43 , @xmath46 and @xmath47 to be real with @xmath44 complex ) .
the five independent parameters can be chosen as the three up quark masses , one mixing angle and one phase . in the down sector ,
the yukawa couplings @xmath59 can all be chosen real , @xmath60 is just an overall scale , and @xmath61 is complex .
the five independent parameters can be chosen as the three down quark masses and two mixing angles .
now we have ten parameters in these two matrices except their overall normalizations or magnitudes of yukawa couplings .
since we also have ten observables for six quark masses , three angles and one phase , it may appear that a fit is not so remarkable .
however , the forms of the mass matrices are very restrictive , and it is by no means trivial to obtain a good fit . indeed
, we find that @xmath62 is strongly correlated with the cp phase @xmath63 which is one of angles of the unitary triangle .
if we were to fit just the six masses and the three angles , the structure of our mass matrices would allow only a very narrow range of values for @xmath63 at each value of @xmath64 .
this means that future more precise determinations of these two parameters will be a decisive test of this model .
cp violation is also predicted in our model .
the jarlskog invariant @xcite is given by j_cp ( ) ( 1 + ) , [ jcp ] which is mainly comes from the down sector and the up sector only affects as correction terms . in order to fit the ten observables ( six quark masses , three ckm mixing angles and one cp phase )
, ten parameters of our model have been generated numerically .
we choose the parameter sets which are allowed by the experimental data .
we show the prediction of @xmath64 verses @xmath63 in fig .
[ f1 ] , with the following nine experimental inputs @xcite : & & m_u = 0.9 ~2.9 ( mev ) , m_c = 530 ~680 ( mev ) , m_t = 168 ~180 ( gev ) , & & m_d = 1.8 ~5.3 ( mev ) , m_s = 35 ~100 ( mev ) , m_b = 2.8 ~3 ( gev ) , & & |v_us| = 0.221 ~0.227 , |v_cb| = 0.039 ~0.044 , j_cp = ( 2.75 ~3.35)10 ^ -5 , & & [ input1 ] which are given at the electroweak scale .
we see that the experimental allowed region of @xmath63 ( @xmath65 radian at @xmath66 c.l . )
@xcite corresponds to @xmath64 in the range @xmath67 , which is consistent with the experimental value of @xmath68 .
thus our model is able to reproduce realistically the experimental data of quark masses and the ckm matrix .
plane , where the value of @xmath63 is expressed in radians .
the horizontal and vertical lines denote experimental bounds at @xmath66 c.l .
( [ input1 ] ) . ]
precisely measured heavy quark masses and ckm matrix elements are expected in future experiments and precise light quark masses are expected in future lattice evaluations . if the allowed regions of the current data shown in eq .
( [ input1 ] ) are reduced , the correlation between @xmath64 and @xmath63 will become stronger .
we show in fig .
[ f2 ] the case where the experimental data are restricted to some very narrow ranges about their central values : & & m_u = 1.4 ~1.5 ( mev ) , m_c = 600 ~610 ( mev ) , m_t = 172 ~176 ( gev ) , & & m_d = 3.4 ~3.6 ( mev ) , m_s = 60 ~70 ( mev ) , m_b = 2.85 ~2.95 ( gev ) , & & & & [ input2 ] here we use the tighter constraints on the mass ratios of light quarks , i.e. @xmath69 and @xmath70 , consistent with the well - known successful low - energy sum rules @xcite : clearly , future more precise determinations of @xmath64 and @xmath63 will be a sensitive test of our model .
plane , where input data are restricted in the narrower regions shown in eq ( [ input2 ] ) . ]
a comment is in order .
our quark mass matrices are in principle given at the su(5 ) unification scale .
however , the @xmath0 flavor symmetry is spontaneously broken at the electroweak scale .
therefore , the forms of our mass matrices are not changed except for the magnitudes of the yukawa couplings between the unification and electroweak scales .
hence , our numerical analyses are presented at the electroweak scale .
we should also comment on the hierarchy of @xmath36 and @xmath37 .
the order of @xmath36 are fixed by the quark mixing ( [ vus ] ) , ( [ vcb ] )
. the ratios of h_1/h_2 ( 0.22 ) , h_2/h_3 ^2 , are required by @xmath71 and @xmath72 , respectively .
once @xmath36 are fixed , quark masses determine the hierarchy of @xmath37 as follows : v_1/v_3 ^2 , v_2/v_3 ~^1/2 . these hierarchies of @xmath36 and @xmath37 are also consistent with the magnitude of @xmath73 given in eq .
( [ jcp ] ) .
the @xmath0 family symmetry which has been successful in understanding the mixing pattern of neutrinos ( tri - bimaximal mixing ) is applied to quarks , motivated by the quark - lepton assignments of su(5 ) .
realistic quark masses and mixing angles are obtained entirely with the @xmath0 context , in good agreement with data . in particular , we find a strong correlation between @xmath62 and the cp phase @xmath63 , thus a decisive future test of this model can be allowed .
it is one of a powerful guideline to find the constraints from neutrinos to grand unification models .
discrete symmetries are suitable to decode the flavor problem : they can accommodate maximal 2 - 3 mixing and explain zero 1 - 3 mixing
. moreover they can be the origin of texture zeros or equalities .
i would like to thank y. koide for his great hospitality and organization of this stimulating workshop in shizuoka .
i am also grateful to e. ma and m. tanimoto for their collaboration . 99 m. maltoni , t. schwetz , m.a .
tortola and j.w.f .
valle , new j. phys .
* 6 * ( 2004 ) 122 .
fogli , e. lisi , a. marrone and a. palazzo , prog . part .
* 57 * ( 2006 ) 742 .
see for example , w. grimus , a.s .
joshipura , s. kaneko , l. lavoura , h. sawanaka and m. tanimoto , nucl .
b * 713 * ( 2005 ) 151 , and references therein .
barger , s. pakvasa , t.j .
weiler and k. whisnant , phys .
b * 437 * ( 1998 ) 107 .
harrison , d.h .
perkins and w.g .
scott , phys .
b * 530 * ( 2002 ) 167 ; p.f .
harrison and w.g .
scott , phys .
b * 535 * ( 2002 ) 163 .
see also , z.z .
xing , phys .
b * 533 * ( 2002 ) 85 ; x.g .
he and a. zee , phys .
b * 560 * ( 2003 ) 87 .
e. ma and g. rajasekaran , phys .
d * 64 * ( 2001 ) 113012 .
babu , e. ma and j.w.f .
valle , phys .
b * 552 * ( 2003 ) 207 .
babu , b. dutta and r.n .
mohapatra , phys . rev .
d * 60 * ( 1999 ) 095004 .
e. ma , mod .
a * 17 * ( 2002 ) 627 ; x.g .
he , y.y . keum and r.r .
volkas , jhep * 0604 * ( 2006 ) 039 .
see for example , m. hirsch , j.c .
romao , s. skadhauge , j.w.f .
valle and a. villanova del moral , phys .
d * 69 * ( 2004 ) 093006 ; e. ma , phys .
d * 70 * ( 2004 ) 031901r ; phys .
d * 72 * ( 2005 ) 037301 ; mod .
lett . a * 20 * ( 2005 ) 2601 ; phys .
d * 73 * ( 2006 ) 057304 ; g. altarelli and f. feruglio , nucl .
b * 720 * ( 2005 ) 64 ; nucl .
b * 741 * ( 2006 ) 215 ; nucl .
b * 724 * ( 2005 ) 423 ; k.s .
babu and x.g .
he , hep - ph/0507217 ; a. zee , phys .
b * 630 * ( 2005 ) 58 ; b. adhikary , b. brahmachari , a. ghosal , e. ma and m.k .
parida , phys .
b * 638 * ( 2006 ) 345 ; l. lavoura and h. kuhbock , mod .
a * 22 * ( 2007 ) 181 ; g. altarelli , f. feruglio and y. lin , hep - ph/0610165 .
e. ma , h. sawanaka and m. tanimoto , phys .
b * 641 * ( 2006 ) 301 .
e. ma , mod .
a * 21 * ( 2006 ) 2931 .
s.f . king and m. malinsky , phys .
b * 645 * ( 2007 ) 351 .
w. grimus and l. lavoura , jhep * 0601 * ( 2006 ) 018 .
n. haba , a. watanabe and k. yoshioka , phys .
* 97 * ( 2006 ) 041601 .
mohapatra , s. nasri and h.b .
yu , phys .
b * 639 * ( 2006 ) 318 .
mohapatra and a.y .
smirnov , ann .
nucl . part .
* 56 * ( 2006 ) 569 .
g. altarelli , hep - ph/0611117 .
h. fritzsch and z.z .
xing , prog .
* 45 * ( 2000 ) 1 .
particle data group , w.m .
yao _ et al .
_ , j. phys .
g * 33 * ( 2006 ) 1 .
ckm fitter group , j. charles , hep - ph/0606046 .
c. jarlskog , phys .
* 55 * ( 1985 ) 1039 .
h. leutwyler , phys .
b * 378 * ( 1996 ) 313 ; hep - ph/9609467 . | realistic quark masses and mixing angles are obtained applying the successful @xmath0 family symmetry for leptons , motivated by the quark - lepton assignments of su(5 ) .
the @xmath0 symmetry is suitable to give tri - bimaximal neutrino mixing matrix which is consistent with current experimental data .
we study new scenario for the quark sector with the @xmath0 symmetry .
* quark and lepton mass matrices + with @xmath0 family symmetry * + + |
a small number of close binary stars are thought to dominate the dynamic evolution of many globular star clusters ( hut et al .
1992 , bailyn 1995 ) , yet classes of such objects which are relatively easily found in the field have proven frustratingly difficult to discover in clusters .
a prime example is cataclysmic variables ( cvs ) , which call attention to themselves via large amplitude light outbursts , and peculiar , ultraviolet - excess colors in quiescence . with quiescent absolute magnitudes @xmath2
, modern ground - based photometric techniques should easily uncover such objects in clusters with typical distances of @xmath3 , even with modest telescopes , unless all such objects are lost to the crowded cores . yet
prior to the launch of _ hubble space telescope _ ( _ hst _ ) , we are aware of only two candidate identifications of cvs in globular clusters , m5 v101 ( margon et al . 1981 ) , whose classification and cluster membership seems secure ( naylor et al .
1989 , shara et al .
1990 ) , and m30 v4 ( margon & downes 1983 ) , whose membership is unclear ( shara et al . 1990 , machin et al .
1991 ) .
observations from _ hst _ have certainly improved the situation , although perhaps not as much as many would have expected .
a few clusters are now known to have a handful of spectroscopically - confirmed cvs ; a recent review is given by grindlay ( 1999 ) . however despite intensive photometric and color - selected searches deep into the cores of a number of clusters , few outbursting objects are found , and most authors believe there is a serious discrepancy with theoretical predictions ( shara & drissen 1995 ; livio 1996 ; shara et al .
1996 , hereafter s96 ) .
whether the problem lies with formation / destruction rates , or some unique property of cluster cvs , remains to be clarified , and will surely require a larger sample of objects .
here we discuss _ hst _ observations of the cluster ngc6624 using the _ space telescope imaging spectrograph _ ( stis ) .
this cluster contains near its center a highly luminous bursting x - ray source with an 11-minute period , the shortest - known binary star ; the system is thought to be a double - degenerate ( stella et al .
1987 , king et al .
1993 , anderson et al .
another object in the cluster has also been suggested by s96 as a candidate cv , although we discuss its nature further in 3 . as part of a program to study the central bright x - ray source
, we obtained deep stis spectroscopic exposures of this object , and these results will be discussed elsewhere ( deutsch et al .
2000 ) . here
we report the completely serendipitous discovery of @xmath1 emission line object , which fell by good fortune in the stis slit , and has the properties of a classical cv .
this object deepens the mystery of the cv content of clusters .
s96 report a multi - epoch sensitive photometric search of the core of ngc6624 which identifies only one candidate cv , and suggest these objects are very rare , at least in this cluster .
yet we have found another such an object completely by accident , in the first long - slit ultraviolet spectral exposure made of the cluster .
on 1998 march 14 we obtained 12150 s of integration in the center of ngc6624 over 5 orbits with the _ hst _ stis fuv - mama , using a @xmath4 wide , @xmath5 long slit .
the spectra are reprocessed and extracted using calstis 2.0 and the latest calibration files available as of 1999 april 20 , where nearly half of the calibration files are updated from the original pipeline processing . in fig .
1 we show a @xmath6 long region of the cluster spectrum , which has been background subtracted for display purposes . the primary target , the x - ray source hereafter denoted star k , is clearly visible at the top .
two other objects , labeled stars 1 and 2 , are visible .
no other objects are discernible in the entire long - slit spectrum , including the region not shown in fig
the extracted spectrum of star 1 , binned @xmath7 to 1.17 per bin , is displayed in fig .
strong , broad emission lines of n v @xmath81238,1242 , si iv @xmath81394,1403 and/or o iv ] @xmath91401 , c iv @xmath81548,1550 , and he ii @xmath91640 are clearly detected .
the o i @xmath91304 feature is most likely imperfectly subtracted geocoronal emission .
the location of a few other common emission lines which may contribute to this spectrum are also marked .
the continuum is very weakly detected at @xmath10 erg @xmath11 s@xmath12 @xmath12 .
this spectrum is that of a classical cataclysmic variable in a quiescent state ( e.g. , wu et al . 1992 ) .
we show in 3 that both the optical and uv magnitudes of the object agree with those expected for a cv at the distance of the cluster , and the probability of a chance superposition of a non - member within @xmath13 of the cluster core is small .
we conclude that we have serendipitously discovered a cv in the cluster .
we have attempted to independently verify cluster membership via radial velocity measurements of the spectrum , but the results are inconclusive .
we were forced to use an undesirably large slit width due to potential problems in target acquisition in this complex field , and the resulting configuration is poorly suited for absolute velocity determinations . if we accept the nominal calstis wavelength calibration , the inferred radial velocity , obtained from several strong emission lines , is @xmath14 km s@xmath12 .
the uncertainty in this value is however dominated by the zero point error introduced by the unknown location of the object within the slit , which could be as large as @xmath15 km s@xmath12 , so the constraint is not meaningful .
further , we expect the radial velocity of the star to vary , with unknown period and probably large amplitude , and so can not make a cogent estimate of the mean velocity after one observation . we have also searched for radial velocity variations from orbit to orbit , again with inconclusive results due to large uncertainties in the subsets of the data .
the spectrum of star 2 is a flat continuum of @xmath16
erg @xmath11 s@xmath12 @xmath12 , and featureless except for possible detections of the common interstellar absorption lines si ii @xmath17 , c ii @xmath81334,1335 , and c iv @xmath81548,1551 .
when dereddened , the spectrum and photometry are reasonably well described by a kurucz model ( kurucz 1992 ) of t@xmath18,000 k and @xmath19 .
this is consistent with the properties of the non - flickering `` nf '' objects discussed by cool et al .
( 1998 ) and edmonds et al . ( 1999 ) .
we attempt to determine the precise location of the objects seen in our spectroscopic observations by examining archival _
we find f140w and f430w ( pre - costar ) foc images taken on 1992 august 13 which cover this field , as well as wfpc2 images taken on 1994 april 17 and 1994 october 15 , obtained with a variety of filters .
there exist two additional epochs of wfpc2 data , but as this field falls on the lower resolution wide field camera ccds in those observations , they provide no additional useful information
. given the crowding in the cluster core , determination of which objects are responsible for the observed spectra is not trivial .
based on information in the stis header , the distance of stars 1 and 2 from the bright star k is @xmath20 and @xmath21 , respectively , at a position angle of @xmath22 . in fig .
3 we show 5@xmath235@xmath24 regions of the various foc and wfpc2 images . overlaid are two arcs indicating the calculated radii of these objects from star k ( not present in this field ) .
the nearly horizontal lines indicate the edges of a @xmath4 slit .
therefore , we expect to find the objects responsible for the spectra near the intersections of the arcs and the slit .
residual uncertainty in the slit angle is less than 1@xmath25 and thus not important at this scale .
finally , the figure assumes that star k is centered on the slit .
if star k is in fact shifted slightly in the slit , then the location of the slit we have drawn may be displaced up or down by up to @xmath26 ; the target acquisition scenario employed should place star k very near the center of the slit , however . in the 1400 foc image , we find a faint object at the expected location of star 2 , and label it star b ( recognizing that the association is hardly guaranteed . ) we find no evidence for any detection at the expected location of star 1 in this f140w image .
examining successively longer wavelength images , we find that star b exhibits uv - excess between the f336w and f439w images , but is hopelessly contaminated by light from neighboring stars at longer wavelengths . at the expected location of star 1 ,
we find a faint star which we label a at f439w and longer wavelengths , although there is no evidence of it at f336w and f140w .
star a is positively detected at multiple epochs as well as multiple wavelengths , although it falls just outside the foc f430w image .
there is also another faint object @xmath27 nw of star a , which is also a plausible candidate .
no small adjustments in slit angle or slit shift yield a better alignment . in the astrometric frame of _ hst
_ image u2as0101 t , the coordinates for star a are @xmath28@xmath29 , @xmath30@xmath31 .
although internally precise , this position has a probable uncertainty of @xmath32 with respect to external frames . based on our estimate of the cluster center in the same image , we find that star a is @xmath33 , or 1.8 core radii , from the center of ngc6624 ( assuming @xmath34 from harris 1996 ) . a detailed discussion of a precise measurement of the cluster center , not relevant to our analysis , is given by king et al .
two radio pulsars are located within @xmath35 of star a ( biggs et al .
1994 ) , but the positions are definitely disjoint , given the quoted uncertainties . in a cluster of @xmath34 , it is perhaps not surprising that multiple interesting but unrelated objects are in such proximity . in table 1
we present photometry for selected objects in the field .
we use a combination of profile - fitting and aperture photometry to derive these magnitudes . for wfpc2 images , aperture corrections
are taken from table 2(a ) in holtzman et al .
( 1995b ) .
the photometric measurements have not been corrected for geometric distortions , nor is any correction for charge transfer efficiency losses ( holtzman et al .
1995b ) applied ; for most of the images , these effects should contribute errors of only a few percent .
we use the photometric zero points for the stmag system from table 9 ( @xmath36 ) in holtzman et al .
( 1995a ) .
approximate @xmath37 measurement uncertainties are also provided in the table .
systematic errors for all magnitudes due to uncertainties in detector performance and absolute calibration are @xmath385% .
magnitude measurements presented here are denoted @xmath39 , where @xmath9 is the filter designation , approximately indicating the central wavelength of the filter in nanometers . in the stmag system ,
zero points are set such that a flat spectrum ( @xmath40 ) source will have identical magnitudes at all wavelengths . for the foc measurements , we use the calibration provided in the header of the images .
a stsdas _ synphot _ calculation yields a calibration value which differs by @xmath41% .
we use a tiny tim ( krist 1993 ) synthetic profile to determine an aperture correction .
since the f140w images were obtained in an unusual mode and far uv calibration is often difficult in any case , we suspect that the absolute calibration may not be accurate to better than a factor of two . in order to further check that the associations between stars 1 and a , and stars 2 and b , are plausible , we estimate @xmath42 for stars 1 and 2 at the time of the stis observations by convolving the spectra with the f140w bandpass and integrating the observed flux .
we estimate @xmath43 for star 1 , and @xmath44 for star 2 , and estimate uncertainties of 0.2 mag principally because the spectra do not fully cover the f140w bandpass .
given the various uncertainties in this estimate and foc calibration , the agreement between the stis and foc ( table 1 ) magnitude estimates is excellent . aside from the observed level of excitation in the spectrum ,
the lack of any brighter and/or spatially extended image near the correct location for star 1 provides confirmation that the observed emission spectrum can not be due to a cluster planetary nebula or the background superposition of a low redshift agn .
we conclude that star a is the likely source of the emission line spectrum .
we point out in passing yet another unusual object in the same field .
star c ( table 1 , fig .
3 ) is extremely bright in these f140w images , but a rather faint object in longer wavelength passbands .
when dereddened , the @xmath42 , @xmath45 , and @xmath46 measurements are well fit by a kurucz model of t@xmath47,000 k. this temperature is most sensitive to the f140w measurement for which the calibration is poorest . if the foc calibration is adjusted by 0.4 mag such that the foc observed and stis convolved magnitudes are equal for star b , the implied temperature for star c is t@xmath48,000 k. this extraordinary uv excess is noteworthy but otherwise not relevant to the present discussion , except as further evidence of a multitude of exotic objects near the center of this cluster .
our photometry in 2.2 together with the known distance and reddening of the cluster permit a comparison of the luminosity of this object with that of the far better - studied field cvs . adopting @xmath49 and @xmath50 ( harris 1996 )
, our measured @xmath51 implies @xmath19 .
there are now four trigonometric parallaxes for classical field cvs ( harrison et al .
1999 , mcarthur et al .
1999 ) , which collectively imply @xmath52 , similar to the @xmath53 often quoted from much larger samples ( warner 1995 ) . given the uncertainties in our photometry for this faint object in a very crowded field , and our single - epoch measurement of an undoubtedly variable star , the agreement of the inferred luminosity of the new ngc6624 cv with those in the field is gratifying . if the system is strongly magnetic , a possibility we consider below , our observed magnitude is perhaps somewhat brighter than expected from polars in the field , but , again given the uncertainties , not alarmingly so .
we are aware of few if any ultraviolet spectra of globular cluster cvs .
edmonds et al .
( 1999 ) display a low signal - to - noise spectrum of an object in ngc 6397 , where heii @xmath91640 is termed marginally detected " in emission by those authors . through absolutely no credit to the current authors , our spectrum is far better exposed .
the prominent heii @xmath91640 emission in our spectrum is deserving of comment .
this line is normally not strong in classical cvs , but is seen in polars .
our spectrum is indeed quite similar to that of am her in this wavelength range ( greeley et al .
unfortunately the simplest defining spectroscopic characteristic of am her stars , very strong heii @xmath94686 emission , is not accessible to us .
although little can be inferred from one object , we note that grindlay et al .
( 1995 ) have suggested that magnetic white dwarfs are preferentially produced in globular clusters ( see also grindlay 1999 and references therein ) .
these authors also discuss the issue of possible confusion of the spectra of cluster cvs with those of quiescent low - mass x - ray binaries ( lmxbs ) , and those considerations also apply here .
our spectrum does not unambiguously distinguish between the two cases ; for example , the quiescent lmxb cen x-4 ( v822 cen ) shows quite weak heii @xmath94686 ( mcclintock & remillard 1980 ) and little or no @xmath91640 emission ( blair et al . 1984 ) , but aql x-1 ( v1333 aql ) displays strong heii @xmath94686 in quiescence ( garcia et al .
1999 ) . as more cvs than quiescent lmxbs
are known in clusters , it seems most conservative , but still uncertain , to continue the discussion of cvs . as this object lies only a few arcsec from the brightest x - ray source in any globular cluster , its x - ray properties are as yet unknown , but may be accessible to observations from _
chandra_. although our detection of the ultraviolet continuum is weak , it is of interest to ask if the observed ultraviolet flux agrees with that expected from the field objects , as we have determined above is the case for the visible band .
if we assume @xmath54 ( cardelli et al .
1989 ) , our observed @xmath55 erg @xmath11 s@xmath12 @xmath12 corresponds to an extinction - corrected monochromatic luminosity of @xmath56 erg s@xmath12 @xmath12 . for the quiescent u gem , for example , @xmath57 erg @xmath11 s@xmath12 @xmath12 ( wu et al .
1992 , long et al . 1994 , long & gilliland 1999 ) ; this ratio of 30,000 in observed fluxes speaks well to advances in ultraviolet spectroscopic capabilities from the _ international ultraviolet explorer _
( _ iue _ ) to the _ hst _ era . adopting @xmath58 pc ( harrison et al . 1999 ) and @xmath59 ( panek and holm 1984 , long & gilliland 1999 ) for u gem , we infer @xmath60 erg s@xmath12 @xmath12 for that object .
while this excellent level of agreement is almost surely fortuitous , it reaffirms that the cv in ngc6624 , with the exception of its unusual environment , seems a totally normal cataclysmic .
as various abnormal mass transfer scenarios have been at times invoked to explain the lack of cluster cv outbursts , this agreement is interesting .
there is one previous candidate for a cv in ngc6624 , as discussed by s96 .
the identification was based on the presence of the object on two consecutive _ hst _ images , but its absence in all other observations .
we have recovered the candidate in the archival _ hst _ images , and derive its position as @xmath28@xmath61 , @xmath30@xmath62 using the astrometric header information in _ hst _ image u2kl0406 t .
this position is @xmath63 from star a , and thus this object can not be responsible for our spectrum .
moreover , upon close examination of these data , we find that the radial profile of this object is not compatible with a stellar one .
we suggest it is an image artifact of uncertain origin . in fig . 4 , we show a surface plot of the pixels around this object and a nearby , typical star of similar total counts .
the s96 object has a profile much too sharp as compared with this and other stars .
it is indeed detected in two frames , and thus is not a charged particle hit , but rather probably a group of hot pixels or calibration file defect .
we stress however that the removal of the s96 object from the list of cluster cvs merely _ strengthens _ the primary scientific conclusion of that paper , namely that erupting cvs are remarkably rare in clusters .
we have serendipitously discovered a mass - transfer close binary , presumably with a degenerate companion , close to the core of ngc 6624 .
it is likely that the system is a classical cataclysmic variable , although a quiescent lmxb can not be ruled out .
the situation is a curious inversion of the normal problem in globular cluster observations , where one finds an interesting star but has difficulty obtaining the spectrum due to severe crowding .
we have easily obtained the spectrum , but are uncertain of which object is responsible for the emission , although we advance as a reasonable candidate star a. the uncertainty in the exact coordinates of the system does not however change the surprising result that an object supposedly so rare has been found accidentally .
the object was found in a single stis long - slit exposure encompassing @xmath64 arcsec@xmath65 at an arbitrary position angle passing quite close to the cluster center .
a hypothetical observing program that included all possible position angles with this slit , thus completely mapping the cluster center to @xmath66 , would cover @xmath67 more area .
tempered by the usual uncertainties of _ a posteriori _ statistics , it seems quite likely that the center of ngc6624 contains several , and possibly many , more cvs of the type we have discovered here . if we now accept that these systems most certainly are present
as has indeed been theoretically expected for some time then one or more mechanisms which suppress the expected outbursts and/or odd colors , and thus hide globular cluster cvs , must certainly be operative . | despite indications that classical cataclysmic variable ( cv ) stars are rare in globular clusters in general , and in the cluster ngc6624 in particular , we have serendipitously discovered such a star @xmath0 from the cluster center .
a _ hubble space telescope _
spectrum of the @xmath1 object shows strong , broad emission lines typical of numerous field cvs , and the inferred optical and uv luminosity are also similar .
our accidental observation also provides the first high - quality ultraviolet spectrum of a globular cluster cv . that we have detected such an object in an observation that includes just a few percent of the central area of the cluster
may indicate that cluster cvs are more common than previously thought , at least near the core .
-0.6 in 9.6 in 0.210 in accepted for publication in the astronomical journal + to appear in the 1999 december issue , volume 118 + _ received 1999 july 28 ; accepted 1999 august 31 _ |
the investigation on how purely self gravitating matter evolves on long time scales has played and still plays a central role in cosmology , in astrophysics and in statistical mechanics . from the statistical point of view , the main obstacle towards the establishment of a general picture is the lack of simple additivity due to the long range interaction .
this prevents a system , even if composed by a macroscopic number of `` particles '' , to reach thermodynamic equilibrium , so that the usual laws of equilibrium thermodynamics do not apply @xcite .
conversely , it is well known that there exist infinitely many stable equilibrium phase space configurations to choose from as @xmath17 @xcite .
still , relaxation of some sort must occur , since self gravitating systems of many different kinds and sizes appear to be in ( quasi ) stationary states @xcite .
likewise , numerical @xmath18body simulations of such systems show that , from a coarse grained point of view , some type of equilibration does occur @xcite .
the central issue is then how to properly identify the quasi stationary states corresponding to given initial conditions . as detailed below , such identification is the aim of this work in the specific case of single warm dark matter ( wdm ) halos , although the implications of our findings probably extend to wider domains . indeed , in the cosmological context , important questions concern the properties of dm halos , which are the cradles of galaxy formation @xcite . in turn , the halos are formed through gravitational collapse triggered by tiny perturbations over the uniform dm background which dominates the total matter contribution @xmath19 to the energy density of the universe ( baryons contributing only a smaller fraction @xmath20 ) . in this scenario ,
the counting and the properties of galaxies with any given mass should be traced back to the primordial power spectrum of dm perturbations , which statistically determines the initial conditions , and to the quasi stationary state ( or states ) to which gravity drives the various local matter overdensities .
for example , in a universe dominated by cold dark matter ( cdm ) the density perturbations are gravitationally unstable down to mass scales much smaller than those of galaxies and extensive numerical simulations ( see _ e.g. _ @xcite and references therein ) have shown that structure formation proceeds bottom up with the early collapse of very small regions , followed by larger and larger ones , with a complex merging histories that produce successively more or less relaxed objects of increasing mass , like dwarf satellites , massive galaxies , groups and clusters .
thus the properties of a single dm halo depend on the full history of a much larger region of comoving space , hindering a clear cut analytic modeling of its quasi stationary state .
however , this cdm scenario has ( at least ) two distinguishing features contradicted by observations : it exhibits an overabundance of small mass virialized halos @xcite and cuspy density profiles @xcite where observations point to cored ones @xcite .
the latter problem , the so called cusp core puzzle of cdm , apparently fades away upon including baryon feedback in certain simulations @xcite , but not in others @xcite . structure formation based on wdm is in principle protected against these cdm difficulties , while reproducing all the observed large scale ( @xmath21mpc ) properties as cdm does .
in fact wdm is characterized by the proper amount of primordial velocity dispersion , with associated free streaming , to smooth out initial density fluctuations over scales below the mpc @xcite .
this velocity dispersion should also help to better match the observed density profile of collapsed halos . in brief
, wdm reconciles , or tend to reconcile with observations several small to middle scale astrophysical and cosmological features such as , for example , the halo concentration and satellite counts in @xmath18body simulations @xcite , the phase space density of dwarf spheroids in a model
independent wdm treatment @xcite , the mass function in the halo model @xcite , the galaxy luminosity functions and the stellar mass distributions in semi
analytic models @xcite .
many other investigations @xcite confirm that wdm agrees with observations better than cdm .
still , one observational fact that many numerical simulations on wdm fail to correctly reproduce is the size of the halo cores @xcite . indeed , a well known theoretical argument , rooted on the tremaine
gunn bound @xcite , implies that wdm halos are cored @xcite .
it is based on the so called ( pseudo ) phase space density @xcite , @xmath22 , where @xmath23 is the characteristic mass density and @xmath24 the characteristic one dimensional velocity dispersion of the core
. one can argue @xcite ( see @xcite for more punctual analysis that however do not alter significantly the conclusions ) that @xmath25 , where the space
constant @xmath26 is computed at wdm decoupling , in any given particle model .
then , _ assuming _ a halo model ( _ e.g _ an isothermal sphere as in ref . @xcite or a pseudo - isothermal one as in ref .
@xcite to relate @xmath24 to @xmath23 , one obtains a bound on the core size .
wdm @xmath18body simulations apparently produce cores which almost saturate this bound @xcite .
the problem is that these cores are resolved , and found with size comparable to the observed ones , only for initial velocity dispersions so large that the corresponding free streaming would have erased at the linear level the fluctuation seeds of the halo themselves ( the _ catch 22 _ of ref .
viceversa , using the theoretical bound to extrapolate the size of resolved cores to velocity dispersions small enough to allow the fluctuation seeds , yields cores two order of magnitude smaller than the observed ones .
if this were indeed the situation , wdm structure formation would need help from baryon feedback ( if that really can help ) almost as badly as cdm does .
alternatively , in could be that current @xmath18body simulations , plus extrapolations based on equilibrium assumptions , do not provide a quantitatively sound description of wdm gravitational collapse below the kpc scale .
here we address the second possibility only and give an affirmative answer , by showing that wdm is an effective working hypothesis for explaining the size of observed dm cores . in this work ,
we report the results of extensive high resolution direct numerical integrations of the spherically symmetric vlasov poisson equation for a @xmath27kev wdm thermal relic .
the vlasov poisson equation ( [ eq : vp ] ) , often called collisionless boltzmann poisson equation , describes the non dissipative incompressible @xmath28dimensional phase
space flow of self gravitating continuous matter .
it is what @xmath18body simulations attempt to approximately solve with fictitious particles moving in @xmath29dimensional configuration space . our findings , outlined in the next subsection and described in more detail in sec .
[ sec : results ] , provide an overall picture of wdm halo cores quite different from the commonly expected one , which is based on some apriori assumptions and the ( extrapolation of the ) results of @xmath18body simulations .
a common expectation about cored dm halos , as those of wdm , is that their mass density is a monotonically decreasing function of @xmath30 which has some core radius @xmath4 as unique length scale in the core region .
this is indeed the situation in collisionless self gravitating systems at equilibrium , that are described by stationary and stable ergodic phase
space distributions @xcite .
the above expectation is then extended to quasi stationary states , with the understanding that they do not differ much from truly equilibrium states .
the property that should guarantee this proximity is core virialization , which can be described as follows .
let us consider a spherical , purely self gravitating system .
thanks to gauss law , one can unambiguously compute the gravitational potential energy @xmath31 within the sphere of radius @xmath30 , regardless of the system configuration for @xmath32 .
hence one can define the @xmath33dependent virial ratio @xmath34 , where @xmath35 is the kinetic energy within the sphere .
suppose the system has relaxed to some quasi stationary state with a core of size @xmath4 .
for @xmath30 large enough we expect virialization , that is @xmath36 .
for @xmath37 instead , we have @xmath38 , since @xmath39 while @xmath40 as @xmath41 in a core that is in hydrostatic equilibrium . in between @xmath42
will decrease in some system specific way .
if @xmath43 for @xmath44 , we can say that also the core is virialized .
since @xmath4 is the only scale in the core region , also @xmath42 is a function only of @xmath45 , which behaves as @xmath46 for @xmath47 small enough and monotonically decreases in the interval @xmath48 .
this is exactly what happens for instance in the isothermal sphere and many other systems with ergodic phase
space distributions .
cdm halos do not have cores but are quite virialized , so they have only the system specific decrease . for example , if we rather crudely assume @xmath49 _ la _ nfw @xcite and @xmath50 @xcite , then @xmath51 , in any case much slower than within a core .
wdm halos are cored and the @xmath52 region can be theoretically and/or numerically investigated , provided the resolution is high enough . according to the common lore just described , the inner part of a halo should be in some quasi stationary state not too different from stable equilibrium states .
the halo core would then be virialized in the sense defined above .
on the contrary , the main result of our investigation can be stated as follow : _ the spherical collapse of wdm yields cores that are hollow and not virialized_. details are provided in sec .
[ sec : results ] ( see figs . [
fig : sixp ] , [ fig : rho ] and [ fig : w ] in particular ) .
here we just make few simple observations to support these numerical findings .
let us assume for the moment that the inner part of the halo can indeed be modeled by some equilibrium type cored configuration .
then for @xmath53 we have @xmath54 , where @xmath4 is in the kpc scale to match , for instance , the dm halos of dwarf galaxies .
consider now the uniform isotropic universe at some time @xmath55 when the energy contribution of matter inhomogeneities is negligible and choose , in order to evaluate the initial value @xmath56 of the virial ratio , the center of a future collapse ( of what type of dm does not matter yet ) as origin of the comoving coordinates .
the result is ( see sec . [
sec : w ] for details on the elementary derivation ) : @xmath57 where @xmath58 is the scale factor , @xmath59 is the hubble parameter and @xmath60 is the dm ( comoving ) velocity dispersion . in the derivation we assumed that dm is non relativistic at the time @xmath61 considered . in the case of cdm , we have @xmath62 and the gravitational collapse will need to reduce the nearly constant @xmath63 by just a factor @xmath64 to virialize the halo . instead , in the case of wdm particles , that decouple while ultra
relativistic but are already non relativistic deep in the rd era , we obtain @xmath65 where we used @xmath66 , with @xmath67 the radiation fraction at time @xmath55 , to recognize in @xmath68 a ( very crude ) estimation of the free streaming length of the wdm particles .
therefore @xmath69 indeed , by construction @xmath70 .
for instance , thermal relics with a mass of @xmath27kev/@xmath1 ( and two internal degrees of freedom ) have @xmath71km / s , so that @xmath72kpc since @xmath73 by the time only neutrinos and photons contribute to radiation .
hence , throughout the region of the future core , kinetic energy initially exceeds by three orders of magnitudes the typical value it would eventually have in a few kpc virialized core . and the smaller the core , the higher the kinetic energy excess , since @xmath68 is fixed way before the collapse started .
of course this is an idealized scenario , since the real collapse will not be spherical , there could be mergers and so on .
but the free streaming of a kev - sized wdm has also the merit of smoothing out the matter fluctuations on scales of several hundreds kpc .
then the possibility of a nearly spherical and almost undisturbed collapse is certainly not a far fetched idealization as it would be for cdm .
hence the kinetic energy excess in the initial conditions w.r.t .
the virialized core after the collapse holds beyond the approximation of spherical symmetry . during the collapse , phase mixing and violent relaxation @xcite alone might still bring the system to a quasi stationary state but , lacking more efficient energy transfer mechanisms such as radiation or dissipation , they need not be able to reduce the virial ratio by the thousands in the finite amount of time available .
this is just what we observe by numerically integrating the vlasov
poisson equation .
a complementary viewpoint is based on the conservation of angular momentum , which is the same in physical and comoving coordinates . in a spherically symmetric system
, the @xmath74particle angular momentum is conserved also if the potential is not constant .
of course , the mean vector angular momentum vanishes by symmetry , but the squared angular momentum @xmath75 does not .
matter which is initially at a distance @xmath76 but will eventually fall in the core , has a conserved @xmath75 of order @xmath77 . like the skate dancer that closes her arms
, this matter will spin faster in the core , sustaining the original excess of core kinetic energy . the net result we observe and describe in sec .
[ sec : results ] , is the formation of halos with a hollow core , that is with a non monotonically decreasing mass density which develops its maximum @xmath78 at some @xmath79 , which provides a first natural definition of core radius . throughout the core and beyond a large excess of kinetic energy is trapped mostly in tangential motions .
the core is approaching a dynamic rather than hydrostatic equilibrium and the mass density @xmath80 features a roughly parabolic rise till @xmath81 , where @xmath78 is almost three order of magnitude larger than the density deep into the core . for @xmath82 , @xmath80 quite rapidly settles to the @xmath83 decrease that implies a constant circular velocity .
clearly the crossover at @xmath84 provides another length scale beside @xmath81 .
we find that these properties are common to all the 111 halos of our sample .
the halos are obtained starting from different initial overdensity profiles , which are in turn constructed , according to the procedure described in sec .
[ sec : sphinit ] , by angle averaging over the peaks of a random realization of the fluctuation field proper of a @xmath27kev wdm thermal relic .
the overall shape of @xmath80 is almost independent of the initial conditions and of the location or value of the maximum itself ( see figs . [
fig : sixp ] , [ fig : rho ] and [ fig : highres ] ) .
these latter quantities instead , which fix the overall scale of @xmath80 , do depend on the details of the the initial fluctuation profile , such as height and size .
the hollow core shape allows a natural definition of the core radius @xmath4 and the core density @xmath5 slightly different from @xmath85 : the pair @xmath86 define the point where the density @xmath80 should be cut to replace the hollow core with a constant density core ( see sec .
[ sec : rc ] ) .
@xmath87 is therefore slightly larger than @xmath81 . in our sample
we find values of @xmath4 ranging from @xmath2 to @xmath3kpc and values of @xmath5 ranging from @xmath88 to @xmath89pc@xmath90 . very interestingly , the core radius @xmath4 is nearly inversely proportional to the core density @xmath5 leading to an nearly constant value of @xmath91 around @xmath92pc@xmath9 ( see sec .
[ sec : rc ] for more details ) , in remarkable agreement with the observations reported in refs .
however , while the near constancy of @xmath93 could be rooted in the core hollowness ( a constant surface density means a constant core mass per unit area , a property certainly more appropriate to hollow cores than to bulky ones ) , the quantitative proximity to the observational value @xmath94/pc@xmath9 is rather puzzling ( and quite intriguing ) , since the estimation of @xmath95 is based on the hypothesis that the dm density has a bulky burkert profile @xcite , with its own definition of core radius @xmath96 and core density @xmath97 .
when a burkert fit is performed on the hollow cores for @xmath98 ( see sec . [
sec : rc ] for the details ) , one obtains values of @xmath99 around @xmath100pc@xmath9 , roughly four times larger than @xmath95 .
the typical @xmath4 we find is more than one order of magnitude larger than what expected for a wdm particle of mass @xmath0kev/@xmath1 according to the @xmath101based theoretical bound and extrapolations from resolved cores of @xmath18body simulations @xcite . but this type of arguments rely on the apriori assumption of a _
thermalized _ , and _ a fortiori _ virialized , core and therefore is not applicable to the wdm cores found in our simulations .
see sec .
[ sec : rc ] for the explicit comparison of the hollow core with that of the isothermal sphere , which shows in detail how the isothermal extrapolation fails ; see sec .
[ sec : sigma ] for our technical explanation of why it fails . on the other hand , at odd with @xmath18body simulations , our halos feature a diffuse part with a @xmath83 tail which is too long .
the density is still decreasing almost as slowly as @xmath83 where it tales values from 100 to 200 times larger than the critical density .
one possible cause is a bias in our construction of spherically symmetric overdensity profiles for the initial conditions ( see sec .
[ sec : sphinit ] ) , which does not allow enough matter outflow from the collapsing halo .
this makes it awkward to build a reliable @xmath87 _ vs. _ @xmath102 relation , for any common definition of halo mass @xmath103 .
hence , as in observations , the best way to quantify the mass content of a halo is to consider the maximal circular velocity @xmath6 . in our sample
we find that @xmath104 takes values between @xmath105 to @xmath106km / s , characteristic of small disk galaxies , whereas @xmath4 and @xmath5 are typical of dwarf spheroids .
this concentration excess is properly measured by means of the burkert fit for @xmath107 .
we find that @xmath108 is roughly twice @xmath4 , while @xmath109 , the maximal circular velocity of the burkert fit , is always very close to @xmath6 .
hence , apart from the mass deficit of the hollow core w.r.t .
the burkert core for @xmath110 , our halos differ from realistic dm halo only because the surface density @xmath11 is four times larger than the observed value . in other words , the region @xmath111 of our hollow
core halos is more concentrated than real dm halos by a factor roughly equal to @xmath112 . in sec .
[ sec : sigma ] we estimate that this factor grows to @xmath113 when the mass of the wdm particle is raised up to @xmath114kev@xmath115 , the value that now seems most favored @xcite , and reaches @xmath116 when @xmath117kev@xmath115 , the lower bound from lyman@xmath118 and hydrodynamical simulations of ref .
@xcite .
this quantitative analysis is the least restrictive , since the burkert fit is performed only for @xmath119 , the only region of the inner halo where the fit can be accurate .
since the bulky burkert profile provides a very good fit to real dm halos down to few percents of @xmath96 , the hollow core is most likely ruled out observations .
hence whatever improvement , modification and/or enhancement that might succeed in reducing @xmath11 should also eliminate , or strongly reduce , the hollowness . an obvious addition could be baryons , completely neglected in the zeroth order approach of this work and possible hollowness reducer thorough adiabatic compression .
another could be quantum corrections , to be briefly motivated in the next section . at any rate , since we find cores with an inner mass deficit rather than excess , with a surface density nearly constant and relatively close to the observed value , the starting point provided by our vp approach appears by far better than in cdm only @xmath18body simulations or in the too - small core wdm simulations cited above . in the next section we outline some improvements for our wdm - only vp simulations and we also question the soundness of another common apriori assumption on dm dynamics .
the first , obvious and major improvement would be to go beyond the approximation of spherically symmetric collapse and tackle the full @xmath28dimensional vlasov poisson problem . a promising low resolution attempt in this direction is reported in ref .
@xcite , but the computational resources required for a high resolution cosmological simulation appear at the moment prohibitively large , at least for the author . moreover , it is very unlikely that the problem @xmath120 is entirely due to the approximation of spherical symmetry .
remaining in the more tractable framework of spherical symmetry , we can consider a few interesting directions for future improvement : 1 .
a problematic aspect of our halo sample is the small variation of the basic parameters @xmath5 , @xmath4 and @xmath6 .
the most likely cause is a bias in our construction of spherically symmetric overdensity profiles for the initial conditions ( see sec .
[ sec : sphinit ] ) , which does not allow enough matter outflow from the collapsing halo .
a more sophisticated selection procedure could be implemented to better mimic , even within the approximation of spherical symmetry , the effects of nearby collapsing halos on the halo of interest .
this could at the same time increase the parameter variations and reduce the too long @xmath83 tail of the halos .
it is also conceivable that a shorter @xmath83 tail could improve the value of the surface density @xmath11 , although probably less than necessary .
2 . the large initial virial ratio and the crucial role played by angular momentum imply that the halo collapse might be sensitive to details of the freezed out velocity distribution , which in our simulation we fixed to the simplest fermi dirac form proper for thermal relics ( see sec .
[ sec : vpdm ] ) .
indeed we find that the core properties strongly depend on @xmath60 ( see sec .
[ sec : sigma ] .
hence other wdm models , which are already known to yield different cutoff effects on the primordial power spectrum @xcite , might also lead to noticeable effects on the core properties . in this respect ,
accurate analysis at the linear level also of wdm velocities , as that in ref .
@xcite , could play an important quantitative role in the collapse dynamics through a better determination of the initial conditions .
3 . another common apriori assumption on dm dynamics , besides the core virialization discussed above , is that quantum effects are fully negligible .
the large density variation in the hollow core , the crucial role played by angular momentum in our wdm collapse and the profound quantum modifications to the theory of angular momentum , suggest instead that quantum mechanics could have a deep role in the shaping of wdm cores , provided its effects were non negligible in the primordial universe .
we elaborate this point below .
the relevance of quantum mechanics for ultra compact dwarf galaxies with wdm halos , within an equilibrium approach , was discussed already in ref .
@xcite . for a particle of mass @xmath12 ,
the quantum unit of phase
space density is @xmath121 strictly speaking it should just be @xmath122 , the inverse of the volume of a quantum cell in the phase space with the conventional dimensions of ( length @xmath123 momentum)@xmath90 .
the extra factor of @xmath124 comes from the normalization to a mass density , rather than a number density , and to the use of velocity rather than momentum as coordinate in phase space .
if @xmath125 denotes the number of non translation degrees of freedom of the particle , such as spin and/or internal symmetry quantum numbers , it follows that ( notice the @xmath126 difference w.r.t .
@xcite ) @xmath127 where @xmath128 is the characteristic de broglie wavelength and @xmath129 is the mean interparticle distance .
thus @xmath130 provides a semiclassical measure of how much particles in a non relativistic gas with density @xmath23 and velocity dispersion @xmath24 are packed w.r.t .
the reference quantum packing fixed by heisenberg s indetermination principle .
values @xmath131 of @xmath130 then indicate that quantum effects could be important , as in the standard non relativistic free fermi gas . since @xmath132 largely decreases during the classical gravitational collapse ( we verified this also in our simulations , see sec . [
sec : phiq ] ) , we have to compare it with @xmath133 at the beginning , when wdm is already non
relativistic but still nearly homogeneous . for thermal relics
the issue is solved in the simplest possible way , since in that case the primordial value of @xmath130 is a pure number that does not depend on cosmological parameters nor on the actual value of the mass : @xcite : @xmath134 thus the classical framework might not be fully adequate for wdm , after all .
if so , also the history before should be revisited to determine more appropriate initial conditions with the necessary quantum corrections , namely , the classical framework ignores from the start the fermionic quantum pressure which , in view of the above value of @xmath26 , could significantly contrast the gravitational pull and flatten the halo cores , as advocated in refs . @xcite .
in the mean field approximation , the phase
space one particle distribution function @xmath135 of a purely self gravitating system of @xmath136 identical particles evolves according to the vlasov
poisson ( vp ) equation @xmath137 f = 0\;,\\ & \nabla_r^2\phi = 4\pi g\,\rho \;,\quad \rho({{{\boldsymbol{r}}}},t ) = \int d^3v\,f({{{\boldsymbol{r}}}},{{{\boldsymbol{v}}}},t)\;. \end{split}\ ] ] here @xmath23 represent the mass density of the system , obviously normalized as @xmath138 where @xmath139 is the particle mass . strictly speaking then , it is @xmath140 that plays the role of one particle distribution function .
field approximation means that two body or higher correlation functions are considered always negligible .
then the specific value @xmath12 of the particle mass drops completely out of the game if @xmath141 is normalized as in eqs .
( [ eq : vp ] ) , which now describe the evolution of a non dissipative incompressible fluid in phase space subject only to the self consistent gravitational force it generates .
in fact , the first equation in ( [ eq : vp ] ) is nothing but liouville s equation for free streaming in the `` external '' field @xmath142 . it is commonly believed ( and in a certain special sense rigorously demonstrated @xcite ) that in the limit @xmath143 the mean field description becomes exact . but from the more practical point of view of understanding the dynamics of the @xmath144 ( or more ) particles of a dm halo , for example , there can be little doubt that the self gravitating fluid description implied by the mean field approximation is more accurate than any @xmath145body dynamics with @xmath146 at least @xmath147 times smaller than @xmath136 and simulation `` particles '' as massive as @xmath148 gev/@xmath1 . due to
the absence of dissipative phenomena in the collisionless mean field dynamics , it is legitimate to have doubts also on the hydrodynamic approximations to eqs .
( [ eq : vp ] ) , where the moment expansion in velocity space is closed at second order by some heuristic equation of state .
the problem , from a numerical point of view , is that simulations of the six dimensional continuous system eqs .
( [ eq : vp ] ) , without any symmetry to help , are much more demanding than @xmath145body or @xmath29dimensional hydrodynamical simulations . since the vlasov equation is formally time
reversible , the system retains full memory of its initial conditions , albeit dispersed on smaller and smaller scales over phase space .
it is commonly believed that any reasonably coarse grained version of the exact solution @xmath135 does relax to some form of equilibrium , or quasi stationary states ( qss ) , a typical feature of systems with long - range interactions .
bell theory of violent relaxation @xcite provides a rather general framework for this , although only for very special situations it is has been possible to verify that the qss to which the system relaxes does indeed maximize the lynden bell coarse grained entropy @xcite .
in other words , little is known on the class of initial conditions that do lead to qss , or , if the system does relax to qss , on the map from the space of initial conditions to the space of qss and on the relative time scales .
however , this fact is not a major issue in the cosmological applications of the vp equation , since the initial conditions in the linear regime of the gravitational clustering are rather well known .
then the real problem is to solve as accurately as possible the vp equation and see what really happens in the finite cosmic time available , regardless of any assumed relaxation to a qss . in the cosmological frw spacetime , well after matter - radiation equilibration , the non relativistic dm fluid evolves according to vp system ( [ eq : vp ] ) of newtonian equations , with @xmath149 identified with the cosmic time . but in a cosmological context one can not ignore the ( accelerated ) universe expansion even for a single dm halo .
if the phase space coordinates @xmath150 are interpreted as physical coordinates for a dm particle , then no change is needed in eqs .
( [ eq : vp ] ) to implement a decelerated expansion , since the latter affects only the initial conditions through the assumption of the proper hubble flow . on the other hand ,
the accelerated expansion caused by the cosmological constant @xmath151 requires to introduce antigravity , by appropriately changing the source term in poisson s equation , that is @xmath152 where @xmath153 is the energy density due to @xmath151 , written in terms of the present fraction @xmath154 and of the critical density or , equivalently , of the present hubble parameter @xmath155 .
here we neglect the other smaller source of newtonian potential , namely baryons , so that @xmath156 and @xmath157 if @xmath158 is the total matter fraction .
actually , a more convenient framework is obtained by using comoving coordinates and the superconformal time @xmath159 in place of the cosmic time @xmath149 .
we first rename @xmath150 to @xmath160 and @xmath23 to @xmath161 ; then we set @xmath162 where @xmath58 is the scale factor , solution of the acceleration equation of uniform expansion @xmath163 and upper dots denote as usual derivative w.r.t .
@xmath149 . in terms of the comoving @xmath150 and the superconformal time @xmath159
the vp system of equations now reads @xmath164 f = 0\;,\ ] ] where @xmath165 is @xmath166 times the potential due solely to fluctuation over the uniform background , that is @xmath167\;. \end{split}\ ] ] the time dependence of the scale factor if fixed by eq .
( [ eq : a ] ) , or by the equivalent friedmann equation in superconformal time @xmath168 we see that , apart from the different symbol interpretation , the liouville part of the vp system is the same as before and the universe expansion is equivalent to introducing a negative matter background and time dependency on the gravitational coupling .
the background due to the cosmological constant drops out of the game and @xmath154 directly affects only the evolution of the scale factor . to complete this dynamical setup
, we have to provide the initial conditions for eqs .
( [ eq : vpcomov ] ) and ( [ eq : af ] ) .
we assume that at @xmath169 the scale factor is @xmath170 , with the initial redshift @xmath171 large enough so that dm is well within the linear regime and the distribution function has to a very good approximation the factorized form @xmath172\ , f_{\rm i}(|{{{\boldsymbol{v}}}}+ \nabla \psi({{{\boldsymbol{r}}}})|)\ ] ] here the overdensity fluctuation field @xmath173 is a random realization of the gaussian process with the matter power spectrum at redshift @xmath171 , while @xmath174 is the freezed out distribution , that is the unit normalized space uniform isotropic velocity distribution at decoupling which would remain constant in the uniform universe , that is if @xmath175 .
@xmath176 is the potential for the initial average velocity field ( the zeldovich velocity ) and satisfies @xmath177 in order to fulfill to first order in @xmath173 the mass continuity equation @xmath178 at the initial time .
the specification of the matter power and of the freezed out distribution @xmath179 characterizes the type of dm . in the case of cdm
the power spectrum has a slow power
like falloff at small scales while @xmath180 is for all practical purposes a delta function at @xmath181 .
in the case of wdm the power spectrum is more or less ( depending on the specific wdm model ) sharply cut off at scales smaller than the dm free streaming length , while @xmath180 is a ( highly model dependent ) isotropic distribution . here
we consider only a specific model of wdm , namely fermionic thermal relics that decoupled at thermal equilibrium while ultrarelativistic , so that , if @xmath182 denotes the dimensionless fermi dirac distribution , we have @xcite @xmath183 where @xmath184 is the kinetic energy , @xmath185 is the ( comoving ) decoupling temperature and @xmath125 is the number of non translational degrees of freedom of the dm particle , such as spin and/or internal symmetry quantum numbers .
for example , @xmath186 or @xmath187 , depending on the specific model , in the case of a serious spin1/2 candidate for wdm such as the sterile neutrino @xcite . from eq .
( [ eq : f0fd ] ) we can derive the following expression for the freezed out velocity distribution ( @xmath188 and @xmath189 is riemann s @xmath190function ) : @xmath191^{1/2 } = 2.0768098\ldots\ ; , \end{split}\ ] ] with the identifications @xmath192 thus , recalling that @xmath193 , we obtain a rewriting of the relation of refs .
@xcite between the mass and the velocity dispersion of the wdm particle ( notice that @xmath194 is most often quoted in the literature ) @xmath195 for what concerns the explicit choice of the wdm power spectrum , details are provided in section [ sec : sphinit ] .
the solution of the full six - dimensional vp equation presents a major challenge , both from the analytical and numerical point of view .
analytical results are restricted to perturbation theory and partial resummation thereof ( see _ e.g. _ ref .
@xcite and references therein ) , while numerical results are only preliminary @xcite .
the numerical study of the collapse of a single spherically symmetric system is much more tractable , while retaining a great interest both theoretically and from the point of view of its applications to concrete physical contexts .
if the dm distribution function is rotational invariant , we can use as independent phase
space variables @xmath196 , @xmath197 and the conserved squared angular momentum ( per unit mass ) @xmath198 .
then @xmath199 and the vp equations read @xmath200 f = 0 \quad,\qquad \phi'(r , s ) \equiv \partial_r \phi(r , s ) = \frac{g\,a}{r^2 }
\left[m(r , s ) - \frac{4\pi}3 { { \rho_{\raisebox{-1.2pt}{\scriptsize\rm m}}}}r^3\right]\ ] ] where @xmath201 is the total dark mass within a sphere of radius @xmath30 .
no derivative w.r.t .
@xmath75 may appear in the left hand of eq .
( [ eq : vpsph ] ) because @xmath75 is a conserved quantity in this collisionless dynamics . in this spherically symmetric framework , the moment expansion of the vp equation
is obtained by multiplying it by integer powers of @xmath202 and @xmath75 and then integrating over @xmath202 and @xmath75 . using the standard notation of expectation values @xmath203 and after one integration by parts in @xmath202 , one obtains @xmath204 with the convention that @xmath205 .
when @xmath206 one has the mass continuity equation with spherical symmetry @xmath207 while if @xmath208 , @xmath209 and the continuity equation is used , one obtains euler ( or momentum ) equation @xmath210 where @xmath211 are the squared radial and tangential velocity dispersions , respectively . in the cartesian frame
the pressure tensor reads @xmath212 and the vectorial form of eq .
( [ eq : euler ] ) can be recovered by setting @xmath213 . in a quasi stationary state with slowly varying hydrodynamic variables
@xmath23 and @xmath214 , euler equation ( [ eq : euler ] ) implies that wherever @xmath215 we must also have @xmath216 which is known as jeans equation in the astrophysical context . from the fluid point of view , the condition @xmath217 is the marker of hydrostatic equilibrium . on the other hand , for the gas of individual particles orbiting in the slowly varying potential @xmath165 , the definition of _ dynamic
_ equilibrium is perhaps more appropriate . only in the isotropic limit @xmath218 , @xmath219 ,
jeans equation reduce to the equation of simple hydrostatic equilibrium @xmath220 . to be useful , the infinite hierarchy in eq .
( [ eq : moments ] ) needs to be approximately closed at some order , but with no obvious physical basis for collisionless dm . on the other hand ,
if effective collisional terms were added on some physical grounds , the induced dissipation would allow to write an effective equation of state to relate the pressure ( then quickly rendered isotropic by dissipation ) to the density and perhaps the local entropy .
dm is collinsionless and only more complex mechanisms , such as the effects of localized dm clumps acting like macro particles , could allow reliable closures at higher order such as those in ref .
numerous clumps at many small scales are indeed present in cdm bottom up clustering , but they most likely do not play an important role for wdm , due to the much smoother initial conditions . tackling the full hierarchy , that is the original vp equation , appears then mandatory for wdm . the initial conditions in eq .
( [ eq : init ] ) take now the form @xmath221\ , f_{\rm i}\big(\sqrt{[u - u_{\rm i}(r)]^2+\ell^2/r^2}\,\big ) \;,\ ] ] where @xmath222 is the initial infall velocity @xmath223 written in terms of the overmass @xmath224 we see that the spherical symmetry has allowed to reduce the problem to three dimensions . of these , two dimensions correspond to the radial phase space with coordinates @xmath225 .
the third coordinate @xmath75 enters the vp equation in ( [ eq : vpsph ] ) as a constant parameter , but is to be integrated over to obtain the density as in eq .
( [ eq : fnu ] ) .
let us consider the vlasov equation ( [ eq : vpcomov ] ) regarding first @xmath226 as an external time
independent acceleration field .
this is just the liouville equation for the incompressible streaming of any conserved local quantity in the phase
space of a single particle .
the formal solution can be written @xmath227 where @xmath228 is the so called liouvillian operator .
of course @xmath229 and @xmath230 do not commute and @xmath231 can not be factorized into the product of exponentials of @xmath229 and @xmath230 .
but we can certainly write @xmath232^n \;,\quad s = n\tau\ ] ] and for @xmath233 @xmath234 the cubic order or the approximation can be checked by brute force power expansion of the exponentials or , more simply , by noticing that @xmath235^{-1 } = e^{-\tau l_v/2 } e^{-\tau l_r } e^{-\tau l_v/2 } = e^{-\tau \tilde l(-\tau)}\ ] ] implies @xmath236 furthermore it should be noticed that @xmath237^n e^{\tau l_v/2}\ ] ] and therefore one needs to alternatively apply many times @xmath238 and @xmath239 , while @xmath240 only once at the beginning an the end .
notice also that we may exchange the role of @xmath229 and @xmath230 in @xmath241 without any problem .
this yields a second approximate evolution whose proximity to the first one can be used to check the accuracy of both methods .
the promotion of @xmath226 to a time dependent acceleration field that depends on @xmath141 itself is rather straightforward : we need just to calculate @xmath226 from the poisson equation just before every time @xmath238 is applied . in the application to the cosmological context of eqs .
( [ eq : vpcomov ] ) and ( [ eq : phi ] ) the gravity strength is also growing in time with the scale factor @xmath58 .
we found that better stability in the evolution is obtained by choosing non uniform @xmath242steps corresponding to uniform steps in @xmath58 , as can easily be determined from eq .
( [ eq : af ] ) . the great advantage of the well known operator splitting @xcite defined in eq .
( [ eq : osplit ] ) is that the half - step evolution operators , @xmath238 or @xmath243 , are just @xmath244dependent translations in @xmath245 or @xmath246dependent translations in @xmath247 . with the
so called finite volume methods @xcite these translations ( which correspond to the so called _ advection _ equation ) can be implemented very accurately on uniform as well as non uniform grids on phase space , in such a way to ensure machine precision local conservation of the @xmath141 values at each update . that is , the @xmath248 norm of @xmath141 in any portion of the grid changes only because of the flows at its boundaries . on the other hand , the @xmath249 norms with @xmath250 , which are all exactly conserved in the continuum , are only approximately conserved on any finite grid , unavoidably implying a numerical coarse graining . in the continuum also
the the squared modulus of the @xmath141 fourier transforms in @xmath247 and/or @xmath245 , is conserved .
this ceases to apply after discretization , unless one makes use of the non local translation algorithm based on the discrete fast fourier transform .
this algorithm , however , is more demanding in terms of computer power and is essentially restricted to uniform grids ; a serious limitation , as we shall see , in the present case of gravitational collapse . generally speaking all vlasov solvers , that is numerical methods that solve directly the vlasov equation on a phase
space grid , are free of the noise inherent to @xmath18body simulations but suffer to one degree or another of numerical artifacts , like diffusion and dissipation . in ref .
@xcite a comparison of different solvers was performed in the case of a two dimensional phase space .
no method was found as a clear winner , especially when the dependence of the acceleration on the density is considered , as necessary in the vp system ( in @xcite the context is that of plasma physics , but there a close similarity to the cosmological setup in comoving coordinates ) .
in fact , the semi - lagrangian methods with large integration time steps , which are certainly less diffusive of those based on operator splitting plus finite volumes , are not allowed when the acceleration field changes with time in a way that depends on the solution itself .
altogether , provided diffusion and dissipation are under control , the simplicity and stability of operator splitting and finite volume methods makes them particularly suited for the purposes of this work , all numerical results reported in the sequel as based on such a method . in particular , we have chosen a high resolution finite volume advection solver based on the piecewise linear reconstruction with total variation decreasing slope limiters @xcite .
a similar choice was made in ref .
@xcite , where the first ( low resolution ) simulations in the full six dimensional phase space are reported . as already stated above , in this work
we restrict our attention to the spherical symmetric case of section [ sec : sph ] .
the phase space is then two dimensional and , together with the angular momentum @xmath75 , we have a three dimensional setup .
thus high resolutions can be achieved . to obtain spherically symmetric initial conditions , we perform averages over angles of the initial distribution @xmath251 of eq .
( [ eq : init ] ) around suitably chosen points as discussed in more detail below .
some caution is necessary in the treatment of the squared angular momentum @xmath252 , which of course needs to be discretize too .
the actual values used , and in particular their total number , must be chosen wisely depending on the @xmath225 phase space grid ( see subsection [ sec : angmom ] ) .
the latter can not really be a static uniform grid , since the gravitational collapse spans too many dynamical scales . in this case , where the collapse center is fixed beforehand , it is simpler to opt for a static non uniform grid which gets finer and finer near @xmath253 and @xmath254 .
of course , in a higher dimensional situation when the collapse centers are not known apriori , the most sensible choice would be to use some adaptive mesh refinement scheme .
high resolution requires indeed non uniform grids .
the radial density grows by several order of magnitudes near the origin and the detailed evolving structure of this peak is just the subject of this study . at the same time
radial velocities becomes very large near the collapse center and the grid must allow for them since the only conceivable boundary conditions in velocity space are those of free outflow and no inflow .
a grid with a too small velocity cutoff would lead to unphysical mass loss .
a non uniform grid may allow for a large velocity cutoff while keeping within reason the total number of cells in the @xmath202 direction .
we used non uniform phase space grids which have @xmath255 cells , with @xmath256 ranging from 400 to 700 and @xmath257 from 600 to 1000 , to select the best compromise between resolution and speed .
these cells have increasing width both for increasing @xmath30 and @xmath258 , with the exponential laws @xmath259 for @xmath260 and @xmath261 , @xmath262 , both variable parameters in search of the best compromise between resolution and speed .
the narrowest cells in @xmath30 , which are adjacent to @xmath253 , have typical widths around 2 pc and the narrowest cells in @xmath202 , which are adjacent to @xmath263 , have typical widths around 2 m / s .
the grid setup in the @xmath264direction is described next .
the conservation of angular momentum plays a crucial role in the spherically symmetric collapse and must be handled very carefully .
the angular momentum content of the system is fixed once and for all by the initial distribution @xmath265 in eq .
( [ eq : sphinit ] ) and we need to numerically perform the integral over @xmath75 in eq .
( [ eq : vpsph ] ) with a level of accuracy which is consistent with the chosen @xmath266grid .
since @xmath267 dies exponentially fast for large @xmath75 , the integral can be cutoff at some finite value maintaining the desired accuracy .
much trickier is the discretization near @xmath268 . indeed , from eq .
( [ eq : fnu ] ) and eq . ( [ eq : f0fd ] ) , we see that @xmath269 while @xmath267 dies exponentially fast in @xmath270 for any @xmath271 .
still , the condition ( [ eq : cond ] ) should be fulfilled as closely as possible on the cell centers of the @xmath33grid when the integral over @xmath202 and @xmath75 are replaced by finite sums .
moreover , one should worry also about the higher @xmath272moment versions of ( [ eq : cond ] ) . the higher @xmath264moment versions are not as worrisome , since the various @xmath264 components are coupled only through their sum . in our mass
conserving finite volume setup , the integral over @xmath202 is naturally replaced by the sum over @xmath272cells properly weighted by the cell widths .
for the sum @xmath75 we have more freedom but , for obvious computational reasons , this sum should be as small as possible . to reach a convenient compromise
, we adopted a gaussian quadrature scheme with few free parameters which were optimized to fulfill eq .
( [ eq : cond ] ) to a given order for all cell centers of the @xmath33grid . for grids with 400 @xmath33cells with the first cell , adjacent to @xmath253 , of width
@xmath273 pc , uniform agreement to order @xmath274 requires a costly @xmath264grid with several thousands of points , where @xmath275 pc@xmath9 ( m / s)@xmath9 is the smallest value , that is few times smaller than the angular momentum discretization scale corresponding to the @xmath266grid .
fortunately , the @xmath264grid can be substantially coarsened by the following simple strategy : as far as the initial conditions are concerned , we can relax the accurate fulfillment of eq .
( [ eq : cond ] ) and start with a coarser @xmath264grid by including in eq .
( [ eq : init ] ) a suitable compensating factor @xmath276 ; @xmath276 can not anyway be too different from unity , since it allows to preserve the initial overmass , but can not prevent the inaccuracy in higher @xmath272 and @xmath264moments of @xmath179 .
of course this trick is not possible during the evolution , since we do not know beforehand the r.h.s . in eq .
( [ eq : cond ] ) when @xmath277 replaces @xmath265 in the l.h.s .
; we just have to run few prototype cases with finer and finer @xmath264grids , with correspondingly @xmath276 closer and closer to unity , until the evolution stabilizes .
typically , this happens more or less when the number of @xmath75 values reaches that of @xmath30 values .
the corresponding @xmath276 exceeds unity by few percents for all but the first and/or second leftmost @xmath33grid center , where it can reach values up to 10 . in spite of this
we could verify in the prototype cases , by measuring against long runs with thousands of @xmath75 values , that the numerical evolution remains quite accurate even on those cells .
boundary conditions must be set also at @xmath253 and at some finite value @xmath278 of @xmath30 . on the line @xmath253
the correct condition is of reflecting type , with each half @xmath33line at a given @xmath279 and the corresponding half @xmath33line at @xmath280 `` glued '' together so that @xmath33 translations act smoothly at @xmath253 .
finally , at @xmath281 we assume free outflow , while the inflow can be null or free .
the last two alternatives are actually quite different , since with no inflow the total mass can only decrease with time , while with free inflow it might go both ways depending on the initial conditions . in the cosmological context free inflow appears more natural , since the system under study is just a very small portion of an initially nearly uniform universe .
on the other hand , owing to the presence of the background mass in eq .
( [ eq : vpsph ] ) , it is possible to essentially decouple the dm halo from the rest of the universe , thus rendering the two alternatives of null or free inflow almost indistinguishable . to this end
, it is sufficient to set up the initial fluctuation profile @xmath282 in eq .
( [ eq : sphinit ] ) in such a way that the overmass @xmath283 of eq .
( [ eq : om ] ) vanishes at some point @xmath284
. then the initial infall velocity @xmath222 also vanishes at @xmath285 and in the first stages of evolution dm flow away from @xmath286 in both directions . for the rest of the evolution till redshift zero , a small but rather stable outward flux for @xmath287
is maintained by the gravitational push in the under dense region and by the free outflow conditions at @xmath288 .
this prevents in a physical way any uncontrolled inflow at @xmath289 when inflow is allowed by the boundary conditions .
as anticipated at the end of section [ sec : vpdm ] , we assume for the initial velocity distribution @xmath290 the fermi
dirac form of eq .
( [ eq : f0 ] ) proper for fermions . that decoupled at equilibrium while ultrarelativistic .
@xmath290 parametrically depends only on the initial one dimensional velocity dispersion @xmath60 given by eq .
( [ eq : sigma ] ) .
typically , in our simulations we made the most straightforward reference choice @xmath291 , @xmath292 , @xmath186 and @xmath0kev , so that @xmath293 m/s . to complete the determination of @xmath294 in eq .
( [ eq : sphinit ] ) , we need instances of the initial overdensity profile @xmath282 , from which also the initial infall velocity @xmath222 can be determined according to eq .
( [ eq : u0 ] ) .
one possibility is to make use of the reknown results of ref .
@xcite on the local peak statistics of random gaussian fields .
we prefer here to proceed in a purely numerical fashion , to avoid the averaging of local peak profiles over all random field realizations . 1 .
compute the wdm power spectrom @xmath297 for fermionic thermal relics of the given mass ( @xmath298 kev/@xmath1 ) at redshift @xmath299 , using ( a slight modification of ) the 2011 camb package @xcite .
analytic expressions like those in refs .
@xcite could be used with almost unnoticeable effects on the simulations .
2 . extract a random gaussian field @xmath300 over a cubic lattice ( we used two lattices , one with @xmath301 points and a lattice spacing of 45 kpc , the other with @xmath302 points and a spacing of 22 kpc ) , with zero mean and unit ( in the sense of kronecker s delta ) variance .
fast - fourier transform @xmath300 , multiply the result by @xmath297 and inversely fast - fourier transform this product to obtain one realization of the fluctuation field @xmath303 on the original lattice ; 4 . repeat the previous step multiplying the fourier transform of @xmath300 by @xmath304 , where @xmath305 is the fourier transform of the sharp window function @xmath306\theta(r - r)$ ] , namely @xmath307 and @xmath308 is chosen with an eye to the characteristic mass @xmath309 of collapsed halos ( but see below for more ) ; then inversely fast - fourier transform to obtain a smoothed @xmath310 realization of the fluctuation field .
obviously @xmath303 and @xmath310 differ significantly only if @xmath308 is significantly larger than the wdm free streaming length , which for a 1 kev@xmath115 thermal relic can be calculated to be 185 kpc @xcite .
select local maxima of @xmath310 which are _ prominent _ , that is , which are higher than all other peaks within a distance @xmath311 to be properly selected ( see below ) .
perform a spherical average of the unsmoothed @xmath303 around these prominent maxima and then interpolate over the chosen @xmath312grid to obtain the radially symmetric peaked profiles @xmath282 to use in eq .
( [ eq : sphinit ] ) and eq .
( [ eq : u0 ] ) .
some comments are required on the procedure just outlined .
first of all let us stress that the aim is not really at an accurate halo statistics , for which more sophisticated approaches are needed .
rather , we try and produce several initial spherically symmetric peak profiles that are at least compatible with the true initial conditions , which are certainly not spherically symmetric . in this respects ,
certain common interpretations require some adjustment , especially because of the higher smootheness of the wdm fluctuation fields @xmath303 w.r.t . those of cdm .
for instance , in our approach the relation between the smoothing scale @xmath308 and the mass @xmath313 of collapsed halos is just marginal . only after the collapse , at redshift @xmath314
, we can measure the actual mass of each dm halo and this typically turns out to be larger than @xmath315 . in fact , to trust the spherical symmetry approximation , it is necessary that a collapse center be sufficiently isolated , bringing in the other distance scale @xmath311 characterizing the prominence as in point 5 above . by its definition , @xmath311 must be chosen quite larger than @xmath308 , but not necessarily as proportional to @xmath308 .
the mass of a collapsed halo is in between @xmath315 and @xmath316 and depends on specific features , as shown in sec .
[ sec : profiles ] , of the initial overdensity profile @xmath282 .
the chosen value of @xmath308 , which must be in any case larger than the free streaming length , sets an approximate lower bound on the halo mass and allows for collapse centers that are not maxima of the unsmoothed fluctuation field @xmath303 .
notice indeed that the initial profiles @xmath282 are often not monotonically decreasing nor maximal at @xmath253 , although a common feature of all profiles is their overall decrease toward zero ( which is the average of the fluctuation field ) at larger distances , as dictated by the spherical average and the prominence requirement .
the latter also implies that quite often @xmath282 becomes negative sufficiently far away . in turns ,
this often causes the corresponding overmass @xmath283 to vanish at some slightly larger distance . in the upper panels of fig .
[ fig : sixp ] we show a small subsample of initial fluctuation peaks obtained by the procedure outlined above , together with the corresponding overmass . in the lower panel
we anticipate the form of the collapsed halos .
the distance @xmath286 at which @xmath283 vanishes determines @xmath317 , the total amount of dm mass potentially involved in the collapse , which in turns provides an upper limit to the mass of the collapsed halo for whatever definition of such a mass one may choose .
sometime the overmass does not vanishes within the simulation interval @xmath318 , in which case we modify by hand the far end of the @xmath282 , through the addition of a negative gaussian peak centered near @xmath278 , in order to determine a value of @xmath286 close to @xmath278 . as explained in sec .
[ sec : bc ] the vanishing of @xmath319 allows to decouple the collapsing halo from the rest of the universe , preventing on physical basis a uncontrolled inflow rather than just setting the inflow to zero as a sharp ( and unphysical ) boundary condition . on the other hand ,
all initial peak profiles for which the arrangement by hand is necessary have essentially the same value of @xmath320 .
this feature , together with selection of peaks by prominence ( and probably several other reasons ) , quite likely makes our sample of initial conditions statistically biased . named after courant , friedrichs , and lewy , this is a necessary condition of stability ( and hence convergence ) for any numerical method that aims at solving partial differential equations ( pde ) .
it can be formulated as `` _ a numerical method can be convergent only if its numerical domain of dependence contains the true domain of dependence of the pde _ '' @xcite . in the case of linear as well as
non linear transport equations , this requirement is often translated into a bound of the form @xmath321 on the so called _ courant number _ @xmath322 .
here @xmath323 represents the cell width of a one dimensional grid with a generic @xmath47 coordinate , @xmath324 is the integration timestep and @xmath325 is the advection , or transport , velocity . to be more precise , there are in general as many bounds like the one above as there are cells , and with time dependence too , since @xmath323 , @xmath324 and @xmath325 might all have a local nature , due to non uniformity of the grid , adaptive integration stepping and most notably dependence of @xmath325 on the pde unknowns when the transport equation is non linear . in our case
@xmath47 is either @xmath30 or @xmath202 , @xmath325 is one of the corresponding advection velocities in eq .
( [ eq : vpsph ] ) @xmath326 while @xmath324 is to be identified with the superconformal timestep @xmath327 of eq .
( [ eq : tau ] ) . then the bound in eq .
( [ eq : cfl ] ) becomes prohibitively stringent at small values of @xmath30 and/or large values of @xmath202 , imposing values too small for @xmath327 , with uncontrollable diffusion and dissipation .
this would indeed constitute a major obstacle to the whole approach if the bound on the courant number was regarded as necessary for the cfl condition to hold . in reality , however , the bound of eq .
( [ eq : cfl ] ) applies only to local updating algorithms , that is methods in which the updated value of the variable of interest in a given cell depends on the old values in the same cell and in a small , fixed number ( which depends on the specific method ) of nearby cells ( the so called algorithm stencil ) .
the bound does not apply to non local algorithms where the stencil grows with the advection velocity .
the simplest example is uniform advection in one dimension . in this case advection
reduces to rigid translation ; then , if the lattice is uniform , by splitting the translation parameter into an integer multiple of the lattice spacing @xmath323 plus the fractional part , one can advect in a single step by amounts @xmath328 much larger than @xmath323 , making the courant number @xmath322 arbitrarily large . using piecewise linear reconstruction ,
it is not difficult to extend this method also to non uniform lattices . finally , since as explained above , the operator
splitting vlasov solver reduces to intertwined translations , we can forget about the bound of eq .
( [ eq : cfl ] ) on the courant number and freely choose the superconformal timesteps in order to minimize diffusion and dissipation while maintaining accuracy . to check our algorithms against numerical artifacts we performed several stability and accuracy tests on the core advection solver as well as on the full program . for brevity
, we report here only the main results of three full program tests .
indeed , the advection solver was adapted from quite standard , widely tested and employed algorithms ( see _ e.g. _ ref .
@xcite and reference therein ) . as accuracy quantifiers we considered the relative @xmath329norm , that is @xmath330 where @xmath331 is either the distribution @xmath141 itself or
@xmath332 ( mass ) , @xmath35 ( kinetic energy ) , @xmath31 ( potential energy ) computed for the sphere of radius @xmath30 .
@xmath333 stands for the measured value while @xmath334 stands for the one expected on theoretical grounds .
the sum runs over all indices of the observable , that is all grid indices in case of @xmath141 or only the @xmath30 indices in case of @xmath335 , @xmath336 and @xmath337 .
the weights @xmath338 are the cell volumes or widths in the @xmath33direction , respectively .
the data reported in table [ tab : tests ] refer to a @xmath339 grid with @xmath340 , @xmath341 and @xmath342 .
similar grids were used for most collapse simulations .
.results of the test simulations described in sec .
[ sec : tests ] . for each quantifier
the two column report the values at @xmath343 and @xmath344 , except for test 3 , for which they correspond to redshift @xmath345 and @xmath314 . [ cols="^,^,^,^,^,^,^,^,^ " , ] let us first notice that using the @xmath346 relation , eq . [ eq : sigma ] , the value @xmath347km / s corresponds to @xmath348kev/@xmath1 . in turns , this mass corresponds to the @xmath18body simulation wdm3 of ref .
one can check that their wdm3 density profile agrees quite well with our hollow @xmath80 for @xmath349pc , on the right of the maximum .
this shows that , on the common domain of validity , @xmath18body simulations and direct vp integration agree , as far as the mass density is concerned .
[ fig : sigma ] and table [ tab : sigma ] show that the overall size of the core increases with @xmath60 , as is natural to expect .
maybe less obvious is the peak broadening , that progressively makes the notion of hollowness less appropriate and harder to detect with low
resolution means . in other words ,
the second length scale of the hollow core , the hollowness scale , grows to values comparable to the first scale @xmath87 .
notice that @xmath4 , as defined in sec [ sec : rc ] , does not grow very much , while @xmath96 , the core radius of the burkert fit , grows almost linearly with @xmath12 . moreover
, both @xmath7 and @xmath350 considerably decreases , roughly maintaining their ratio constant . from the data in table [
tab : sigma ] we can write to a good approximation @xmath351 and @xmath352 thus , baldly extrapolating to larger values of the wdm mass , we obtain @xmath353 since @xmath94/pc@xmath9 , we see that the core of h1 would have a surface density ( or surface gravity acceleration , since the two are proportional ) roughly 6 times larger than real dm halos when @xmath354 . taking into account
that this applies to pure wdm collapse ( that is with no baryon feedback of any type ) and within a purely classical approach ( probably a more relevant limitation ) , we consider this a very good starting point for the improvements outlined in the introduction .
the last row in table [ tab : sigma ] shows values of @xmath96 , @xmath11 and @xmath109 in very good agreement with observations .
the corresponding mass , @xmath355 , is however too small w.r.t .
the constraints from large scale structure . as a matter of fact
, it is also too small w.r.t .
the value @xmath356 used in the matter power spectrum to generate the fluctuation field from which the initial profile was drawn . we can not pretend to realistically lower the surface density of one given halo by just raising @xmath60 , while keeping fixed the initial overdensity profile .
larger @xmath60 implies more free streaming in the linear regime of gravitational clustering , with the associated small scale depression of the matter power spectrum .
thus , a fluctuation profile that was generic at @xmath357km / s , could be practically impossible at @xmath347km / s .
this so called _ catch 22 _ of wdm , already mentioned in the introduction , was pointed out already in ref .
@xcite , on the basis of the results of @xmath18body simulations .
however , in ref .
@xcite , the scaling of the core size with the mass was determined to be @xmath358 , while we find the relation ( [ eq : r0sig ] ) for the burkert core radius and an even slower decrease for @xmath4 . once extrapolated to @xmath356
, the scaling @xmath359 leads to core sizes around @xmath360pc instead of several hundreds , as we find .
the incorrect scaling @xmath358 can be traced to the underestimation of the relaxation factor @xmath361 , that is the overestimation of the pseudo phase
space density @xmath362 .
the smallest value @xmath363 in the second column of table table [ tab : sigma ] , which corresponds to a value of the mass quite close to simulation wdm3 of ref .
@xcite , implies that @xmath364 will be even closer to unity , its theoretical minimum , for smaller masses . indeed the wdm @xmath18body simulations of ref .
@xcite could resolve the core only for such masses and found @xmath132 in the core very close to its theoretical maximum .
but the explosive growth of @xmath364 in the second column of table table [ tab : sigma ] shows that their extrapolation of @xmath132 to larger masses was ill founded .
the author is deeply indebted to h.j .
de vega and n.g .
sanchez for many illuminating discussions and for the kind hospitality in the observatoire de paris and the chalonge school events , were basic motivations were laid down and the development of fruitful ideas was initiated . | we report the results of extended high resolution numerical integrations of the vlasov
poisson equation for the collapse of spherically symmetric wdm halos . for thermal relics with mass @xmath0kev/@xmath1
, we find collapsed halos with cores of size from @xmath2 to @xmath3kpc .
the typical core is hollow , with the mass density decreasing towards the core center by almost three orders of magnitude from its maximum near the core radius @xmath4 .
the core is in equilibrium with the diffused part of the halo but far from virialization .
these properties are rooted in the conservation of the squared angular momentum and in the original excess , proper of wdm initial conditions , of kinetic energy in the core region . in a sample of more than one hundred simulated collapses ,
the values of @xmath4 and of the core density @xmath5 are in the range typical of dwarf spheroids , while the maximal circular velocities @xmath6 are proper of small disk galaxies .
the product @xmath7 takes values between @xmath8pc@xmath9 and @xmath10pc@xmath9 , while the surface density @xmath11 , as determined from a burkert fit , is roughly three times larger . from these data and data obtained at smaller values of @xmath12
, we extrapolate for one particular halo @xmath13pc@xmath9 and @xmath14pc@xmath9 at @xmath15kev/@xmath1 , to be compared with the observed value @xmath16/pc@xmath9 . in view of the many improvements and enhancements available ,
we conclude that wdm is a viable solution for explaining the presence and the size of cores in low mass galaxies . |
the galaxy luminosity function ( glf hereafter ) is a powerful cosmological tool ( see e.g. the review by binggeli , sandage @xmath3 tammann 1988 ) . it is a good indicator of the formation history of both galaxies and clusters of galaxies , and is also directly related to the mass distribution function of structures in the universe ( press @xmath3 schechter 1974 ) . excluding environmental effects as a first approximation , hierarchical models predict a mass distribution characterized by an exponential cut - off beyond a characteristic mass m@xmath4 and a power - law for low mass structures .
although a detailed analysis presented by binggeli et al . ( 1988 ) shows that individual galaxy types actually have a gaussian distribution , except for the dwarf ellipticals ( probably the dominant type among the faint coma galaxies ) , the glf is now commonly described with a schechter function ( see e.g. lumsden et al .
1997 or rauzy et al .
1998 for recent statistical applications ) .
it is possible to find a simple linear relation between the shape of the power spectrum ( fourier transform of the mass function ) and the shape of the glf ( see lobo et al .
1997 ) .
a determination of the glf in rich clusters such as the coma cluster ( see biviano 1998 for a review including the article by godwin , metcalfe @xmath3 peach , gmp hereafter ) is interesting as it can provide information about the evolution and formation of small galaxies in rich clusters as well as to place limits to the amount of mass in the cluster in the form of faint galaxies that might have been missed in previous surveys .
it is also interesting to ask if the luminosity function varies within projected distance from the core of rich clusters ( cf .
secker et al 1998 and adami et al 1998 : a98 hereafter ) and to determine if the glf differs in the field versus that in rich clusters ( cf valotto et al 1997 or gaidos 1997 ) .
both of these questions relate to how the environment of rich clusters affects and is related to galaxy formation and evolution . in order to address some of these issues several groups
have acquired and analyzed deep images of the coma cluster to extend the glf to as faint values as possible ( cf .
bernstein et al , 1995 , b95 hereafter , ulmer et al 1996 and references therein ) .
b95 concluded that the luminosity function turned up at the faint end as also suggested for the case of virgo by bingelli et al .
phillipps et al .
( 1998 ) confirm this trend with a very deep survey of the virgo cluster ( @xmath5 as faint as @xmath613.5 ) .
a98 made direct redshift measurements of coma that tentatively suggested a similar turn up , but a98 did not have enough redshifts to draw a firm conclusion . in order to redress the issue of too few redshifts , we requested and obtained more observing time at the cfht . with these data and a compilation of redshifts from the literature
, we now have 88 redshifts that cover the region surveyed by b95 .
we can , therefore , make a more definitive statement about the shape of the luminosity function of coma at the faint end .
we also used our data to compare with indirect methods of determining cluster membership , i.e. , statistical inferences of the number of galaxies in the cluster ( b95 ) and the use of the color magnitude relation ( e.g. biviano et al .
1995 ) . in section 2 of the present paper
, we describe our sample of 88 redshifts . in section 3
, we discuss the implications of these results .
finally , in section 4 we discuss the glf .
we will use h@xmath7=100 [email protected]@xmath8 and q@xmath7=0 throughout the paper to be coherent with the previous analysis of a98 .
we used the photometric catalogue of b95 .
all the details of this catalogue are described in b95 and a98 .
briefly , it covered a [email protected] arcmin@xmath10 area centered on coordinates ( 12h 57 m 17s , 28@xmath110935 " ) ( equinox 1950 hereafter ) .
this field is close to the cluster center , located very near the two giant dominant galaxies of the coma cluster .
each galaxy is characterized by its position and its r and b@xmath12 magnitudes .
this new spectroscopic follow - up was done during 1 night ( 11/12 april 1999 ) using the cfht / mos multi - slit spectrograph with the v150 cfht grism with a resolution of 10 px@xmath8 .
we selected as potential targets all the galaxies with magnitudes [email protected] and resolution parameter ( see definition in b95 ) between 0.85 and 2.2 .
this range of resolution parameters corresponds to the objects securely classified as galaxies .
the galaxy selection was confirmed by the very low contamination rate we found with our spectroscopic observations : less than 5@xmath13 of the targets were found to be stars .
the limiting magnitude of [email protected] is the faintest possible magnitude for a redshift determination in a non - prohibitive exposure time of about 2 hours . in a98 , we used no blocking filters ( wavelength range of @xmath146000 and about 30 slits per mask ) . in the current run
, we have chosen to use a 4400 wide filter ( cfht filter # 4610 ) which allows a greater number of slits .
the redshift range is lower but is still acceptable ( z in the [ 0;0.9 ] range for both the h@xmath3k lines and the g band within our spectral range ) .
we have shown in a98 that the galaxies in this area of the sky and with magnitudes brighter than 22.5 are unlikely to be more distant than z=0.9 .
due to bad weather , only 50@xmath13 of the night was available for spectroscopy .
this resulted in a 2 hour exposure with 45 slits ( mask 1 ) and a second 40 minute exposure ( mask 2 ) with 40 slits . therefore , mask 2 produced a low success rate of redshift determination of about 15@xmath13 ( the mask 1 success rate is about 60@xmath13 ) .
the spectra were reduced in exactly the same manner as in a98 i.e. we used both eso midas and geotran / rvsao iraf .
we assumed the derived redshifts were valid only when the absorption lines provided a rvsao correlation coefficient greater than 3 or when there were more than two well defined and self - consistent emission lines .
when both emission ( at least 2 ) and absorption lines were detected , the redshift was computed with the emission lines .
table 1 summarizes all the information we have for the galaxies we used here to derive a glf ( except those of a98 ) . in order to increase the number of redshifts in the coma core area , we have also searched for velocities available in the literature in the area surveyed by b95 and with a measured magnitude .
the compilation of a. biviano ( private communication ) gives 23 objects .
these 23 objects are also summarized in table 1 ( objects with a gmp number and without a `` * '' ) .
the colors of these galaxies are from gmp and we have diminished the b@xmath15-r@xmath16 of gmp by 0.55 in order to match with our system of magnitudes , but we note this is only an approximation ( the error is about 0.2 magnitude on this number ) .
the objects from biviano are significantly brighter than the galaxies we have observed , but provide complementary data for the bright portion of the luminosity function .
we note that the objects in table 1 with a gmp number and a `` * '' are found both in b95 and gmp .
we had previously described in a98 the data of secker et al .
these data are not yet public , but they cover a field very close to the b95 area .
the galaxies of this sample are in the magnitude range r=[15.5;20.5 ] and only 4 are in the coma cluster .
these data are not included in the sample used in this paper .
we simply use the secker et al .
( 1998 ) results for comparison in section 3 .
cccccc z & err@xmath17 & r & b@xmath12-r & x ( ) & y ( ) + 0.0162 ( 3400 ) & 1 10@xmath18 & 13.82 & 1.33 & 3.08 & -0.10 + 0.0221 ( 3201 ) & 1 10@xmath18 & 14.01 & 1.36 & 0.06 & 1.57 + 0.0207 ( 3313 ) & 2 10@xmath18 & 14.20 & 1.28 & 1.74 & -3.61 + 0.0227 ( 3423 ) & 1 10@xmath18 & 14.30 & 1.40 & 3.41 & -2.16 + 0.0271 ( 3296 ) & 1 10@xmath18 & 14.38 & 1.41 & 1.53 & 1.25 + 0.0266 ( 3178 ) & 3 10@xmath18 & 14.68 & 1.29 & -0.29 & -1.76 + 0.0263 ( 3068 ) & 1 10@xmath18 & 14.97 & 1.38 & -2.62 & 2.62 + 0.0229 ( 3222 ) & 3 10@xmath18 & 14.97 & 1.20 & 0.53 & 2.30 + 0.0123 ( 3262 ) & 3 10@xmath18 & 15.27 & 1.27 & 1.03 & -1.88 + 0.0328 ( 3133 ) & 3 10@xmath18 & 15.73 & 1.37 & -1.17 & 2.29 + 0.0214 ( 3339 ) & 3 10@xmath18 & 16.04 & 1.23 & 2.11 & -1.35 + 0.0263 ( 3126 ) & 3 10@xmath18 & 16.05 & 1.27 & -1.37 & -3.20 + 0.0207 ( 3205 ) & 4 10@xmath18 & 16.11 & 1.28 & 0.13 & -1.11 + 0.0215 ( 3034 ) & 2 10@xmath18 & 16.56 & 1.15 & 3.26 & 3.25 + 0.0228 ( 3376 ) & 3 10@xmath18 & 16.74 & 1.00 & 2.78 & 2.10 + 0.0564 ( 3225 ) * & 1 10@xmath18 & 16.75 & 1.23 & 0.54 & -1.63 + 0.0153 ( 3383 ) & 2 10@xmath18 & 17.00 & 1.31 & 2.86 & -1.50 + 0.0140 ( 3340 ) & 5 10@xmath18 & 17.04 & 1.38 & 2.13 & 2.94 + 0.1782 & 3 10@xmath18 & 17.11 & 1.46 & 1.25 & -0.82 + 0.0128 ( 3275 ) & 1 10@xmath18 & 17.16 & 1.02 & 1.16 & -1.91 + 0.0204 ( 3311 ) * & 4 10@xmath18 & 17.54 & 1.23 & 1.69 & -0.93 + 0.1715 ( 3149 ) * & 5 10@xmath18 & 17.41 & 1.67 & -0.87 & -2.30 + 0.0187 ( 3424 ) & 2 10@xmath18 & 17.48 & 1.81 & 3.41 & 3.37 + 0.1954 ( 3095 ) * & 5 10@xmath18 & 17.62 & 1.69 & -2.17 & -2.17 + 0.0242 ( 3080 ) * & 6 10@xmath18 & 17.96 & 1.19 & -2.41 & 1.89 + 0.0298 ( 3325 ) & 5 10@xmath18 & 18.16 & 1.45 & 1.94 & 1.20 + 0.2748 ( 3150 ) * & 2 10@xmath18 & 18.31 & 1.66 & -0.87 & 1.02 + 0.1774 ( 3284 ) & 3 10@xmath18 & 18.50 & / & 1.29 & -0.88 + 0.1922 & 3 10@xmath18 & 18.55 & 1.27 & -3.32 & -1.64 + 0.6330 ( 3353 ) * & 6 10@xmath18 & 19.43 & 1.63 & 2.29 & 1.36 + 0.0273 ( 3370 ) & 3 10@xmath18 & 19.63 & 1.53 & 2.69 & -2.07 + 0.1776 ( 3137 ) & 3 10@xmath18 & 19.63 & 1.02 & -1.17 & 2.64 + 0.4680 & 3 10@xmath18 & 19.97 & 2.17 & 2.57 & -0.04 + 0.3162 & 2 10@xmath18 & 20.02 & 1.54 & -2.07 & -1.14 + 0.5060 & 4 10@xmath18 & 20.43 & 1.06 & 2.14 & 2.06 + 0.2438 & 4 10@xmath18 & 20.43 & 2.46 & 0.43 & -1.78 + 0.6142 & 2 10@xmath19 & 20.54 & 2.30 & -1.82 & -3.53 + 0.2415 & 3 10@xmath18 & 20.55 & 1.23 & 3.67 & 1.41 + 0.4606 & 3 10@xmath18 & 20.66 & 1.33 & -0.11 & 3.33 + 0.2523 & 4 10@xmath18 & 21.08 & 1.12 & 2.97 & -2.99 + 0.4229 & 2 10@xmath18 & 21.13 & 1.42 & 3.41 & 1.82 + 0.6262 & 9 10@xmath18 & 21.16 & 1.03 & 1.19 & 2.87 + 0.2368 & 4 10@xmath18 & 21.20 & 1.10 & -3.62 & 0.65 + 0.4035 & 3 10@xmath18 & 21.28 & 1.97 & 0.60 & 2.13 + 0.3127 & 3 10@xmath18 & 21.37 & 0.97 & -3.34 & -0.76 + 0.6018 & 4 10@xmath18 & 21.41 & 2.03 & 3.04 & 2.95 + 0.4236 & 3 10@xmath18 & 21.63 & 1.48 & 1.48 & 2.48 + 0.0934 & 3 10@xmath18 & 21.95 & 1.23 & -0.22 & 0.46 + 0.1622 & 6 10@xmath18 & 22.02 & 1.12 & 1.71 & -0.56 + 0.5509 & 2 10@xmath18 & 22.13 & 0.89 & -3.15 & 0.54 + 0.4665 & 3 10@xmath18 & 22.23 & 0.58 & -0.48 & 3.50 + 0.5188 & 4 10@xmath18 & / & / & 5.12 & 0.62 + 0.2827 & 8 10@xmath18 & / & / & 4.56 & 0.53 + 0.0926 & 7 10@xmath18 & / & / & 4.38 & -2.49 + 0.0619 & 4 10@xmath18 & / & / & 3.89 & -2.41 + 0.2134 & 3 10@xmath18 & / & / & 4.44 & 0.55 + [ ] we show in fig .
1 the distribution of our 88 redshifts along the b95 line of sight ( a98 : 32 z , this paper : 33 z , literature : 23 z ) . we clearly see the coma cluster at [email protected] and some other structures along the line of sight ( @xmath20 hereafter ) ( including the distant cluster detected in a98 ) . the coma cluster comes mainly from the literature . in order to estimate
how complete our sample is in apparent magnitude , we compare the number of redshifts we produced to the number of available targets in the magnitude range ( r=[16.5;22.23 ] ) in b95 ( we ignore the very bright galaxies taken from the literature and not sampled by b95 ) .
we find that we covered about 23@xmath13 of all the available galaxies in this magnitude bin .
2 shows the magnitude histogram of the galaxies of our sample superimposed on that of all the b95 galaxies in this magnitude range .
we see that our completeness level drops abruptly above r=21.5 . below this value ,
our completeness level is 33@xmath13 .
we have also tested the spatial representativeness of our sample .
we have used the same bidimensional kolmogorov - smirnov test as in a98 to compare the spatial distribution of the galaxies for which we have a redshift and the spatial distribution of all the galaxies available in b95 in the range r=[16.5;22.23 ] .
we find that the probability that the two spatial distributions statistically differ is less than 0.1@xmath13 .
we conclude , therefore , that our sample is a statistically representative sample of the coma luminosity function and that it is valid to apply corrections to the b95 luminosity function based on the data we use here .
finally , we note that we are not likely to be strongly biased by non - detected very low surface brightness galaxies .
even the lowest surface brightness galaxies of our sample ( described in fig . 5 of ulmer et al .
1996 ) with magnitudes brighter than r=22 ( @xmath21 in fig .
5 of ulmer et al .
1996 ) have a central surface brightness brighter than 26.5 mag.arcsec@xmath22 . on the same figure ,
the galaxies studied by ferguson @xmath3 binggeli ( 1994 ) for a similar magnitude range have the same central surface brightness .
we are , therefore , not biased more than ferguson @xmath3 binggeli ( 1994 ) are .
we concluded in a98 that the use of the cmr to find the galaxies inside the coma cluster for the faint magnitudes ( typically fainter than r=19 ) was inefficient .
we confirm this statement here . by using all the galaxies for which we have a b@xmath23r color
, we see that some of the galaxies beyond the cluster are included in the envelope expected for coma ( see fig .
3 ) , while for the bright galaxies the relation is very good .
this contamination is likely due to distant intrinsically blue galaxies with a k - dimming that makes them mimic coma cluster members .
yet , the two galaxies detected inside the coma cluster in our spectrocopic survey ( the two faintest filled circles , see table 1 for relative coordinates ) are redder than this envelope ( they are not low surface brightness galaxies and they have a mean surface brightness brighter than 23.5 mag.arcsec@xmath22 ) .
we note however that the redshift of the first of these two galaxies is based on only one emission line ( see a98 ) . in order to check this tendency with a larger sample ,
we have plotted on fig .
3 the running mean ( thin solid line with a running window of 10 galaxies ) of the mean color of all the galaxies of the gmp catalog belonging to the coma cluster ( with a known redshift ) , not only in the small field of b95 ( redshifts from biviano et al .
1995 and included in [ 3300;10500]km.s@xmath8 ) . except for a few outliers ( a few percent ) , the coma galaxies brighter than r=17.25
are in good agreement with the mean position of the cmr . for the fainter gmp galaxies
the agreement is also good globally , but the percentage of red outliers increases ( 25@xmath13 ) in the data ( according to the envelope of the cmr , see mazure et al .
finally , the two objects we observed are clearly redder than the cmr envelope .
this is consistent with the suggestion of a98 that there exist high metallicity dwarf galaxies in the coma cluster core .
we do not generalize this hypothesis to all the dwarf galaxies because our statistics are very poor .
different hypotheses can explain , however , the existence of red dwarfs .
for example , we can assume that the intra - cluster medium confines these faint galaxies ( when they are not destroyed by tidal forces ) , thereby keeping a high fraction of heavy elements . in order to reconstruct the glf of the coma cluster core
, we need to know the number of galaxies per magnitude bin belonging to the cluster .
one of the classical methods to obtain this information is to observe comparison fields ( as in b95 ) and to make differential counts between these fields and the cluster field .
we plot in fig .
4 the percentage of galaxies along the @xmath20 which are in the coma cluster according to the b95 field counts ( dotted line with the @xmath24 1@xmath25 error bars ) .
the error bars are the rms deviation on the b95 field counts and are described in b95 section 3.1 .
we also plot on the same figure this percentage according to the redshifts we have ( solid line ) .
this curve was computed using a running mean .
this running mean was created with an average redshift in each bin of 10 galaxies .
the bins were generated by shifting from low to high redshift by just one galaxy .
we used this running mean in order to reduce the noise due to our small sample .
we see that the spectroscopic and the b95 field counts are inconsistent at the 1@xmath25 level for r fainter than 18.5 .
we note that for the two faintest bins , the b95 error bars are large
. we also plot the same percentage , but with all the gmp galaxies with a redshift and not only in the b95 field . for magnitudes brighter than r=17.75 ,
the gmp spectroscopic counts are higher than ours indicating a higher percentage of galaxies belonging to the coma cluster if we consider the whole cluster instead of the small b95 field .
they agree well with the field counts of b95 . for the faintest magnitudes of gmp , the gmp counts agree better with our counts than with the b95 counts .
we also note that our estimates are the same as those of secker et al .
these authors have measured 17 redshifts in the magnitude range [ 15.5;20.5 ] and find only 4 galaxies in the coma cluster ; this gives a percentage of 24@xmath13 .
if we make the same estimate for the same magnitude bin with the data of this paper , we find that 25@xmath13 of the galaxies are in the coma cluster .
the expected absolute magnitudes for typical cluster galaxies range from @xmath26 to @xmath27 ( see e.g. patterson @xmath3 thuan 1996 ) . at the coma cluster redshift , we should expect a significant number of galaxies for apparent magnitudes r close to 2324 .
the observational evidence is that we do not detect such a significant number in the coma cluster core .
secker et al .
( 1998 ) and a98 have suggested that the faint galaxies are disrupted by the giant ellipticals in the center of the clusters . this explanation is theoretically supported by the simulations of merritt ( 1984 ) .
he predicted a tidal radius ( i.e. the lowest possible value for the size of a non tidally disrupted galaxy ) of about 15 kpc in the core of clusters .
the size of dwarf galaxies is significantly lower .
more recent simulations by moore et al .
( 1998 ) also show a similar scenario for the disruption of faint galaxies .
if these galaxies are really destroyed by tidal forces , their luminosities can be transferred from the galaxy " matter phase to the diffuse component " phase .
gregg & west ( 1998 ) propose such an explanation for the coma cluster , estimating that 20% of the diffuse luminosity of a cd galaxy may come from these disruptions .
more quantitatively , b95 show that the diffuse light profile is strongly peaked at the center of the coma cluster and ulmer et al .
( 1996 ) reported that the color of the diffuse light is b@[email protected] , similar to the expected color for dwarf galaxies inside the coma cluster ( slightly redder ) .
these two facts support the disruption hypothesis .
some faint low surface brightness galaxies can also be produced by this process ( ulmer et al . 1996 ) to fit with our explanation of the death of faint galaxies in the coma core ; however , the overall distruction rate must be faster than the production rate .
there are several other explanations besides tidal disruption for the scarity of dwarfs in the core of rich clusters .
for example , as suggested by the referee , the formation of faint dwarf galaxies could have been suppressed due to heating of the baryons in the relatively high density fluctuation that caused coma to form . in this case
, such high temperatures could have prevented faint dwarfs from forming .
five comparison fields ( b95 ) were observed under identical conditions ( same telescope , same instrument , same filter ) as for the b95 coma field .
they were selected at random and chosen to be free of bright stars or galaxies . the flat - fielding and image
combining techniques were the same as for the b95 coma field .
all the details are given in b95 .
the errors on the b95 counts come from a calculation based on the variations of the counts in these five fields .
the question is why do we see a high number of faint galaxies ( in apparent magnitude ) along the coma cluster core @xmath20 compared to these standard empty fields ?
the only explanation we have is an accidental overdensity of structures along this @xmath20 which artificially increases the counts .
one of these structures has been detected by a98 : a cluster at [email protected] . moreover , using the compilation by a. biviano of all the literature redshifts for the coma cluster @xmath20 ( not only the b95 @xmath20 ) , we have detected 8 other structures beyond coma .
the detection in the redshift space is based on the method used for the enacs clusters of galaxies , described in katgert et al .
each structure is sampled with more than 5 redshifts and the estimation of the velocity dispersion @xmath25 ( biweight estimators ) ranges from 170 to 500 km s@xmath8 .
these values are consistent with groups of galaxies ( see e.g. barton et al .
1998 ) . in order to estimate the contribution of these structures to the counts
, we have first searched for the groups that overlap with the b95 field @xmath20 . assuming that these eight groups are all potentially compact groups embedded in a dense environment
, we assumed that their contribution is significant up to 1 mpc in radial projection from the center of the mass concentration , as suggested by barton et al .
5 shows the location of the b95 field and the 2 mpc @xmath9 2 mpc boxes for all these groups .
we see that we have three overlapping groups , that contribute 7 @xmath13 , 4 @xmath13 and 20 @xmath13 of their area to the b95 @xmath20 .
they are described in table 2 .
ccccc & group1 & group2 & group3 & dist .
cluster + mean z & 0.0907 & 0.0970 & 0.2753 & @xmath10.5 + @xmath25 ( km.s@xmath8 ) & 510 & 235 & 170 & 660 + [email protected] & 0 or 0 & 0 or 0 & 1 or 1 & 0 + 16.5@[email protected] & 1 or 1 & 0 or 0 & 3 or 4 & 1 + 19.5@[email protected] & 1 or 1 & 0 or 0 & 6 or 6 & 6 + 20.5@[email protected] & 2 or 1 & 0 or 0 & 11 or 7 & 17 + 21.5@[email protected] & 4 or 2 & 1 or 0 & 23 or 10 & 60 + [ ] we now estimate the contribution of these three overlapping groups and of the distant cluster ( detected along the @xmath20 in a98 ) to the counts . in order to find an upper limit for the effect
, we assume that the distant cluster is a very rich one , with the richness of a3112 ( the richest cluster in the enacs sample of rauzy et al .
1998 ) and we assume the luminosity function of rauzy et al .
( 1998 ) : @xmath28 and m@xmath29@xmath4@xmath30 . for the groups , we have used the relation @xmath25=5.28@xmath9n+106 given by barton et al .
( 1998 ) to evaluate their richness n to a given limiting magnitude . for their luminosity function
, we have both used the rauzy et al .
( 1998 ) cluster function and the muriel et al .
( 1998 ) group function ( @xmath31 and m@xmath29@xmath32 ) .
we have then predicted the number of galaxies along the coma cluster core @xmath20 coming from these groups and from the distant cluster .
subtracting these numbers to the b95 field counts , we plot the corrected values in fig .
4 as filled symbols : triangles when we used the rauzy et al . ( 1998 ) function for the groups and squares when we used the muriel et al .
( 1998 ) function .
we see that we are able to produce values in agreement with those derived spectroscopically in the two faintest bins . with these corrections , the [ 17.5;19.5 ] and [ 19.5;20.5 ] bins also become consistent with the spectroscopic results .
we therefore conclude that the presence of groups and of the distant cluster along the line of sight can explain the difference between the b95 and the spectroscopic counts .
however , we have used an extreme hypothesis for the richness of the groups and of the distant cluster .
we note that such superpositions are found in the enacs clusters as shown in katgert et al .
however , this study also shows that the contamination level spreads over a wide range and that all the clusters are not subject to such superposition effects .
we show in fig .
6 the glf of b95 ( rebinned to conform better to our sample ) and the glf constructed with our own spectroscopic counts .
we produced the brightest two points of our glf by using the lobo et al .
( 1997 ) photometric data included in the b95 field and by applying the spectroscopic percentages of galaxies belonging to the cluster ( see fig . 4 : @xmath1100@xmath13 of the galaxies in the coma cluster ) .
we produced the last five points of our glf by using the spectroscopic counts computed with the b95 data and by applying the spectroscopic percentages of galaxies belonging to the cluster ( instead of the field counts ) .
while b95 quotes a best fit to the r@xmath1415 - 24 glf of [email protected] , the glf in the r@xmath1415 - 22 range is consistent with a schechter function with a slope of @xmath33=-1 .
after our corrections are applied , the glf seems to be flat after r = 13 and may actually be begin to turn over at r@xmath1417 or higher .
we note , however , that this turn - over is not statistically significant given the size of the error bars and the turn - over ( if there is one ) could take place anywhere between r@xmath34 and r@xmath35 .
these errors are difficult to estimate due to our limited number of redshifts , but we have assumed that they have the same values as those of b95 . given these uncertainties , we conclude that the glf is most likely flat ( flatter than deduced by b95 ) , though a moderate rise or decrease are not excluded .
the glf slope ( using dwarf galaxies ) is the same as in loveday et al .
( 1992 : @[email protected] in a magnitude range b@xmath12=[-22;-15 ] ) for normal " elliptical field glf and we have overplotted this function in fig .
6 . their turn - over
, however , begins around b@xmath36 ( r@xmath37 ) , sooner than for our glf as we can see on fig .
we therefore conclude that the glf of the coma cluster core is not in agreement with the field elliptical glf . however , the coma cluster core glf is also significantly different from the glf for the whole coma cluster of lobo et al .
( 1997 ) or smith et al .
( 1997 ) ( the faint ends of the glf are respectively at @xmath38 and @xmath39 ) , or of phillipps et al .
( 1998 ) for the virgo cluster .
these authors find a slope for the glf of @xmath40 or greater .
the conclusion is that the coma cluster core glf has a peculiar shape that was probably induced by environmental effects .
this shape is flat for the faint magnitudes , similarly to what is observed in the field ( e.g. lin et al .
1996 without morphological distinctions ) . however , we stress here that any general conclusion concerning the similarity between the coma inner core glf and the glf in the field is not valid because we are very likely to have strong environmental effects in the inner core of the coma cluster .
a lack of faint galaxies is likely to be due in a large part to the tidal forces created by the two giant dominant galaxies , and an interesting question is : what is the variation of the efficiency of this disruption versus the distance from the center ?
do environmental effects dominate the shape of the faint end of the glf at larger radii ? if this phenomenon is localized to the few central 100 kpc of the cluster , the effect on the global glf of the entire cluster will be limited to a correction of a few percent .
another question to address is do we see the same tendency in the clusters where we have only one giant dominant galaxy or none ? in order to answer to these questions ,
more spectroscopic data are required with at least a similar magnitude depth over a larger area of the coma cluster and other clusters need to be studied as well .
we have used a sample of 88 redshifts along the b95 @xmath20 to investigate the glf in the coma cluster core .
we find that the only two faint coma galaxies in our sample ( r fainter than 19 ) , are redder than the envelope of the cmr in the coma cluster core
. the explanation could be the confinement of the metal - rich gas of these galaxies by the intra - cluster medium .
alternatively , these two galaxies can also have formed from the recent disruption of old infalling galaxies .
we find that there are only a few galaxies in the coma cluster with magnitudes fainter than r=18.5 .
the field counts of b95 predicted significantly more galaxies in the coma cluster velocity field , but this can be explained by accidental superposition effects of 3 groups and 1 distant cluster along the b95 field @xmath20 . using the spectroscopic counts to correct the glf of b95
, we find a nearly flat shape for this function for the magnitudes fainter than r=17 and possibly a turn - over in the glf between r@xmath1417 and 21 .
this is likely due to environmental effects leading to the disruption of most of the faint galaxies by the two dominant galaxies of the coma cluster core .
ac thanks the staff of the dearborn observatory for their hospitality during his postdoctoral fellowship .
bh would like to acknowledge support from the following : nsf ast-9256606 , nasa grant nag5 - 3202 , nasa go-06838.01 - 95a , and the center for astrophysical research in antartica , a national science foundation science and technology center .
the authors thank the referee , n. trentham , for very useful comments . | our previous study of the faint end ( [email protected] ) of the galaxy luminosity function ( glf ) was based on spectroscopic data in a small region near the coma cluster center . in this previous study adami et al .
( 1998 ) suggested , with moderate statistical significance , that the number of galaxies actually belonging to the cluster was much smaller than expected .
this led us to increase our spectroscopic sample . here , we have improved the statistical significance of the results of the coma glf faint end study ( [email protected] ) by using a sample of 88 redshifts .
this includes both new spectroscopic data and a literature compilation . the relatively small number of faint galaxies belonging to coma that was suggested by adami et al .
( 1998 ) and secker et al .
( 1998 ) has been confirmed with these new observations .
we also confirm that the color - magnitude relation is not well suited for finding the galaxies inside the coma cluster core , close to the center at magnitudes fainter than r@xmath119 .
we show that there is an enhancement in the coma line of sight of field galaxies compared to classical field counts .
this can be explained by the contribution of groups and of a distant @xmath2 cluster along the line of sight .
the result is that the coma glf appears to turn - over or at least to become flat for the faint galaxies .
we suggest that this is due to environmental effects .
galaxies : clusters : individual : coma cluster - galaxies : distances and redshifts - galaxies : luminosity function , mass function - cosmology : observations |
the ability to control individual spins in semiconductors nanostructures is an important issue for spintronics and quantum information processing .
the control of single spins in solids is a key but challenging step for any spin - based solid - state quantum - computing device @xcite .
thanks to their expected long coherence time , localized spins on magnetic atoms in a semiconductor host could be an interesting media to store quantum information in the solid state .
optical probing and control of the spin of individual or pairs of mn atoms ( s=5/2 ) have been obtained in self - assembled ii - vi @xcite and iii - v @xcite semiconductor quantum dots ( qds ) .
recent studies of the spin dynamics of mn atoms in self - assembled cdte / znte qds have shown that a strain induced magnetic anisotropy ( _ i.e. _ crystal field splitting ) , changing from dot to dot , blocks the electronic mn spin along the qd growth direction @xcite : the precession of the mn spin in the hyperfine field of its nuclei ( i=5/2 ) or in a weak transverse magnetic field is quenched .
this is at the origin of the mn spin memory observed at zero magnetic field in ii - vi qds @xcite . as one aims at performing a fast optical coherent control of the spin of a magnetic atom , using for example the optical stark effect @xcite in a weak transverse magnetic field , the strain induced magnetic anisotropy of the mn should be suppressed .
strain free mn - doped qds would be an interesting model system to probe the coherent spin dynamics of the magnetic atom in a weak magnetic field and would offer the possibility to externally tune the electron nuclei flip - flops with an applied magnetic field to prepare the nuclear spin of the mn atom . to obtain strain free magnetic qds , we developed thin unstrained cdte / cdmgte quantum wells ( qws ) doped with a low density of mn atoms and lattice matched on a cdte substrate . in these structures ,
localization of the carriers in the quantum well plane is achieved thanks to thickness fluctuations of the quantum wells at the monolayer scale @xcite .
a low density of mn atoms is introduced during the quantum well growth and interface fluctuation islands containing individual mn spins are obtained and optically probed using micro - spectroscopy techniques . a large exciton - mn ( x - mn ) exchange interaction is obtained and a valence band mixing induced by the elongated shape of the confinement potential is observed both in the emission of neutral and charged magnetic qds .
the large exchange induced splitting of the exciton leads to an efficient thermalization of the x - mn complex . to analyze the spin relaxation channels within the x - mn complex in this new magnetic qd system
, we performed resonant photoluminescence ( pl ) under excitation on the s - shell of the mn - doped qds .
these measurements reveal a fast exciton spin relaxation and some slower spin - flips of the mn during the lifetime of the exciton . in these strain free structures ,
we show that there is no optical pumping of the mn spin at zero magnetic field .
the optical pumping is restored under a weak magnetic field in the faraday configuration .
the magnetic field dependence of the resonant pl intensity is well described by a model including the mn hyperfine coupling and a residual crystal field splitting likely due to alloy fluctuations around the mn atom .
the dynamics of the coupled electronic and nuclear spins of the mn atom produces a large photon bunching revealed by the second order correlation function of the resonant pl signal .
the rest of the paper is organized as follows : in sec .
ii , we describe the sample structure , the experimental techniques used and the basic optical properties of strain free singly mn - doped qds . in sec .
iii , we analyze the influence of the valence band mixing on the spin properties of neutral and charged strain free mn - doped qds . in sec .
iv , we discuss the spin relaxation channels in the exciton - mn complex . in sec .
v , we show that the optical pumping of the coupled electronic and nuclear spins of the mn is controlled by a weak magnetic field applied along the qd growth axis . finally , in sec .
vi , we present auto - correlation of the resonant pl of x - mn that reveals the dynamics of the coupled electronic and nuclear spins of the mn atom .
the studied sample consists of a 4 monolayers thick ( 1.6 nm ) @xmath0 qw lattice matched to a ( 001 ) cdte substrate .
a low density of mn atoms is introduced during the growth of the thin cdte qw that is realized by atomic layer epitaxy ( ale ) at 280@xmath1c . both @xmath0 barriers ( 40 nm thick below the qw and 90 nm above the qw )
are grown by molecular beam epitaxy .
the ale growth allows for a smoothing of the surface which leads to the presence of terraces with monomolecular steps @xcite that are the source of in - plane localisation of the carriers .
non - magnetic and magnetic qds containing a small number of mn atoms ( 1,2,3 ... ) are then formed .
mg@xmath2te qw grown on a cdte substrate .
( b ) calculated energy levels of heavy hole ( hh ) , electron ( e ) and electron - hole transitions in thin cdte / cd@xmath2mg@xmath2te qw .
( c)-(h ) pl of singly mn - doped strain free qds with increasing carrier - mn overlap from qd1 to qd6 observed on the low energy side of the pl of a 4 monolayer cdte / cd@xmath2mg@xmath2te qw . the emission of qd2 is dominated by the contribution of a charged exciton , for the other qds the neutral exciton is presented.,width=336 ] optical addressing of individual qds is achieved using micro - spectroscopy techniques . a high refractive index hemispherical solid immersion lens is mounted on the surface of the sample to enhance the spatial resolution and the collection efficiency of single dot emission in a low - temperature ( @xmath3=5k ) scanning optical microscope or in an he bath cryostat equiped with superconducting coils for magneto - optics measurements in faraday geometry .
individual qds are excited with a tunable continuous wave laser tuned to an excited state of the dots or on resonance with the qd s - state @xcite .
the resulting collected pl is dispersed and filtered by a 1 @xmath4 double monochromator before being detected by a cooled back - illuminated si charge - coupled device camera .
for measurements of the second order correlation function of the resonant pl , the circularly polarized collected photons are sent in a hanbury brown and twiss ( hbt ) setup with a time resolution of about 0.7 ns . under our experimental conditions with counts rates of a few khz , the photon pair distribution measured with the hbt setup yields the intensity second order correlation function g@xmath5(@xmath6 ) .
the pl of a 4 monolayers thick mn - doped @xmath0 qw is presented in fig .
[ fig1](a ) .
the thickness fluctuations , and possible inter - diffusion of cd and mg , lead to the inhomogeneous broadening of the pl line of the qds ensemble . the emission of non - magnetic and magnetic individual qds is spectrally isolated on the low energy side of the broad pl line .
the emission of six different strain free mn - doped qds observed on such structure are presented in fig .
[ fig1](c)-(h ) for an increasing value of the exciton - mn exchange interaction . a x - mn overall
splitting as large as 3 mev can be obtained in some qds ( fig.1(h ) ) .
this value of splitting is similar to the maximum splitting that one can observe in magnetic self - assembled cdte / znte qds suggesting that in these two qds systems the carrier - mn overlap is mainly controlled by the confinement along the qd growth axis . in most of strain
free mn - doped qds , the bright exciton presents a linearly polarized structure and a significant pl of the dark excitons is usually observed on the low energy side of the neutral exciton emission @xcite .
both characteristics are a consequence of the presence of valence band mixing in a low symmetry confinement potential @xcite .
to extract the value of the carrier - mn exchange interaction and analyze the presence of valence band mixing in these strain free qds , the experimental pl spectrum are compared with the exciton - mn energy levels obtained from the diagonalization of a spin effective hamiltonian similar to the one used to model the emission of mn - doped self - assembled qds @xcite .
the spin interacting part of the hamiltonian of a mn - doped qd describing the coupling of the electron spin @xmath7 , the hole spin @xmath8 , and a mn spin @xmath9 reads : @xmath10 where the hole spin operators , @xmath8 represented in the basis of the two low energy heavy - hole states , are related to the pauli matrices @xmath6 by @xmath11 and @xmath12 with @xmath13 .
@xmath14 is the coupling energy between heavy - holes and light - holes split by the energy @xmath15 and @xmath16 the angle relative to the [ 110 ] axis of the principal axis of the anisotropy responsible for the valence band mixing @xcite .
i@xmath17 ( i@xmath18 ) is the exchange integral of the hole ( electron ) with the mn atom .
it strongly depends on the position of the mn atom within the carrier wave function .
i@xmath19 is the electron - hole exchange interaction which splits the bright and dark excitons and mixes the two bright excitons in the presence of valence band mixing in an anisotropic potential .
the last term , @xmath20 , describes the perturbation of the exciton wave function by the hole - mn exchange interaction @xcite .
this perturbation depends on the value of the mn spin s@xmath21 and it can be represented , using perturbation theory , by an effective spin hamiltonian @xmath22 with @xmath23 .
the perturbation of the wave function is responsible for the irregular energy spacing of the x - mn and x@xmath24-mn lines observed in self - assembled mn - doped qds @xcite .
= 10k is used to reproduce the pl intensity distribution . for the magnetic field dependence of qd5 we use g@xmath25=-0.55 , g@xmath26=0.75 and a diamagnetic shift coefficient @xmath27=1.25@xmath28.,width=336 ] a comparison between the calculated x - mn spectra and the experimental pl data
is given in figure [ fig2 ] for two qds ( qd3 and qd5 ) .
the main features of the experimental spectra can be well reproduced by this model . a determination of the hole - mn ( @xmath29 )
, electron - mn ( @xmath30 ) and electron - hole ( @xmath31 ) exchange interaction energies can be obtained thanks to the observation of both bright and dark excitons levels at zero magnetic field ( fig.[fig2](a ) and fig.[fig2](b ) ) .
the values of these exchange parameters can be confirmed by the pl spectra obtained under a magnetic applied along the qd growth direction .
the experimental magnetic field dependence of the pl of qd5 and the corresponding model are presented in fig.[fig2](c ) and fig.[fig2](d ) respectively .
the exchange parameters , scattering parameter and valence band mixing amplitude estimated for the six strain free singly mn - doped qds discussed in this paper are listed in table [ table1 ] .
.values of the parameters used for the modelling of the six qds discussed in this paper ( i@xmath18 , i@xmath17 , i@xmath19 and @xmath32 are expressed in @xmath33 ) .
for all the qds , we use @xmath16=0 . [
cols="<,<,<,<,<,<,<",options="header " , ] to explain the dark exciton emission in the x - mn spectra , a small valence band mixing term has to be introduced in the model . in strain
free qds , the mixing can only be induced by a shape anisotropy of the confinement potential .
for an elongated qd , a hole band mixing appears through the non diagonal term of the kohn - luttinger hamiltonian @xcite . in the presence of such valence band mixing ,
the h - mn exchange interaction couples bright and dark excitons and some pl from the dark excitons can be observed @xcite .
in addition , the e - h exchange interaction can now couple bright exciton states and induce some linear polarization @xcite .
-mn ) and negatively ( x@xmath34-mn ) charged excitons coupled to a mn spin can be observed .
the corresponding intensity maps of the linear polarization dependence of the pl of the different excitonic species are shown in ( c ) and ( d ) .
they are compared to the results of the spin effective model in ( e ) , ( f ) and ( g ) .
parameters used in the model are listed in table [ table1],width=336 ] .
[ fig3 ] as shown in figure [ fig3 ] , the linear polarization induced by the valence band mixing can be observed in the pl of the neutral and charged mn - doped qds .
the central lines of the charged excitons coupled to the mn are linearly polarized .
this can be explained by the spin flip interaction between the mn and the hole induced by the valence band mixing @xcite . as presented in the bottom panel of fig .
[ fig3 ] , the main characteristics ( number of emission lines , intensity distribution and linear polarization structure ) of the linearly polarized emission spectra can be well reproduced by the spin effective model with a small valence band mixing coefficient and scattering parameter ( see table [ table1 ] for the parameters used in the model of qd1 and qd2 ) .
this confirms the validity of this spin effective model to describe strain free qds doped with single mn atoms .
-mn in qd6 and pl - excitation ( ple ) detected on the low energy line of x - mn .
( b ) power dependence of the pl of x - mn and x@xmath24-mn in qd6 .
( c ) temperature dependence of the pl of x - mn and x@xmath24-mn in qd6 .
( d ) model of the emission of x - mn and x@xmath24-mn including a thermalization with t@xmath35=5k ( red : bright exciton and biexciton , black : dark exciton ) compared with the experimental spectra ( bleu).,width=336 ] the impact of @xmath20 can be emphasized in qds with a large carrier - mn overlap as for example qd6 in fig .
for this qd , a large asymmetry appears in the emission intensity distribution at zero magnetic field .
this asymmetry decreases with the increase of the lattice temperature ( fig .
[ fig4](c ) ) .
this is a signature of an efficient thermalization on the x - mn split states . the significant transfer of population towards the low energy x - mn levels results from efficient spin - flips within the x - mn system during the lifetime of the exciton .
this thermalisation process also affects the biexciton - mn emission ( fig .
[ fig4](b ) ) .
the perturbation term @xmath20 , used in the model of neutral and charged excitons , induces a splitting of the biexciton - mn states according to s@xmath36 .
the combination of a large carrier - mn exchange interaction and of a small energy splitting between the @xmath37-state and the first excited states ( see figure [ fig4](a ) ) in interface fluctuation qds leads to a large perturbation of the excitonic wave functions by the hole - mn exchange interaction . in the case of qd6 ,
a perturbation term @xmath32=20@xmath38ev is used to model the irregular energy spacing of the x - mn pl lines ( fig .
[ fig4](d ) ) .
this perturbation is two times larger for the biexciton and induces an overall x@xmath24-mn splitting of 240 @xmath33 ( energy splitting between s@xmath39 and s@xmath40 ) .
the carrier relaxation mechanisms responsible for the thermalization of x - mn also takes place within x@xmath24-mn .
the thermalization produces a decrease of the intensity of the central lines of x@xmath24-mn ( _ i.e. _ corresponding to s@xmath21=@xmath411/2 ) .
these intensity distributions in the pl of x - mn and x@xmath24-mn can be well reproduced by the spin effective model ( fig .
[ fig4](d ) ) with an effective mn spin temperature t@xmath35=5k .
the analysis of the optical emission of these strain free magnetic qds demonstrates that they can be used at low temperature ( typically below 20k , see fig .
[ fig4](c ) ) to optically probe and address any spin state of a mn atom in a strain free environment .
we will show in the following how we can use such qds to optically pump the coupled electronic and nuclear spins of a mn atom .
an significant exciton relaxation is observed within the x - mn complex under non - resonant excitation .
more detailed information about the spin - relaxation mechanisms of the exciton exchange coupled with a mn spin can be extracted from the energy and intensity of the pl signal obtained under resonant excitation on the ground state of the x - mn complex .
the spectral distribution of the pl under resonant excitation results from spin - flips processes within the 24 x - mn levels and a simple thermalization should give rise to a thermal distribution on the intensities of the resonant pl spectrum . here , we report resonant pl results for two qds presenting a very different carrier - mn overlap ( qd3 and qd6 ) .
in contrast to strained cdte / znte mn - doped self - assembled qds , a large pl signal is observed under resonant excitation on x - mn in these strain free qds .
we will see in the next section that this results from the absence of optical pumping of the mn spin at zero magnetic field .
this is particularly interesting for a study of the x - mn dynamics : the optical transition that is resonantly excitated is always absorbent and the energy of the re - emitted photons reveals the fastest spin relaxation channels .
figure [ fig5 ] presents the pl excitation spectra ( ple ) detected on the low energy lines of x - mn while the excitation laser is scanned on the high energy side of the spectrum .
corresponding resonant pl spectra obtained under excitation on the high energy line are also displayed and compared with non - resonant pl spectrum presented at the bottom of each panel . from these ple and resonant pl spectra
, it follows that the most efficient spin - relaxation channels within the x - mn system corresponds to a spin - flip of the exciton with a conservation of the mn spin .
for instance , a resonant excitation on the high energy line in @xmath42 polarisation ( excitation of s@xmath21=+5/2 ) gives a maximum of emission in the low energy line in opposite circular polarization ( detection of s@xmath21=+5/2 ) . in this spin - flip process
, the exchange interaction with the magnetic atom can be considered in first approximation as a source of an effective quasi - static magnetic field giving rise to a splitting of the exciton .
the relaxation of the bright excitons is then dominated by the interaction with acoustic phonons .
like in a large magnetic field , this spin relaxation process is enhanced by the large exciton splitting induced by the exchange interaction with the mn spin @xcite . less efficient but still significant excitation transfers between levels corresponding to different spin states of the mn
are also observed especially in qds with a large x - mn splitting ( qd6 in fig .
[ fig5 ] ) .
these transfers under resonant excitation corresponds to the spin - flip channels required for a resonant optical pumping of the mn spin .
the very weak value of the crystal field splitting of the mn expected in these strain free qds is revealed by the longitudinal magnetic field dependence of the resonant pl intensity . as presented in figure [ fig6 ] , for a cross circularly polarized excitation and detection on the high and low energy lines of x - mn ( see inset of fig .
[ fig6](c ) for the excitation / detection configuration ) , a significant intensity of resonant pl is only observed for a longitudinal magnetic field lower than a few tens of mt . as a magnetic field
is applied in the faraday configuration , the resonant pl intensity abruptly decreases ( fig .
[ fig6](a ) ) . a clear asymmetry is also observed in the magnetic field dependence of this resonant pl signal around zero tesla . under @xmath43 excitation on the high energy line and @xmath42 detection on the low energy line ( _ i.e. _ excitation and detection of s@xmath21=-5/2 ) ,
the maximum of resonant pl intensity is slightly shifted towards positive magnetic fields and the decrease of the intensity is faster for negative magnetic fields ( black curve in fig .
[ fig6](b ) ) .
the situation is reversed for swapped excitation / detection circular polarizations ( red curve in fig .
[ fig6](b ) ) . excitation / @xmath42 detection ( black ) and @xmath42 excitation / @xmath43 detection ( red ) .
( c ) ple detected on the low energy line for a @xmath42 excitation / @xmath43 detection and different transverse magnetic fields ( the curves are horizontally and vertically shifted for clarity ) .
the inset presents the non - resonant pl ( red ) , the resonant pl for an excitation on the high energy line ( blue ) and the ple detected on the low energy line ( black ) of qd4 at b=0t.,width=336 ] this magnetic field dependence of the resonant pl is a consequence of the fine and hyperfine structure of the mn atom on its optical pumping . to model
this behaviour we consider the dynamics of the coupled electronic and nuclear spins of the mn including a possible weak residual crystal field splitting .
the fine and hyperfine hamiltonian of a mn atom in a cdte layer grown along [ 001 ] axis is known from magnetic resonance measurements @xcite and reads : @xmath44 \nonumber\\ + { \cal d}_0[s_z^2-\frac{1}{3}s(s+1)]+e[s_x^2-s_y^2 ] \nonumber\\ + g_{mn}\mu_b\overrightarrow{b}.\overrightarrow{s}\end{aligned}\ ] ] where @xmath45 is the hyperfine coupling ( @xmath46 ) @xcite between the electronic ( s=5/2 ) and nuclear ( i=5/2 ) spins . the second term
, @xmath47 @xcite comes from the cubic symmetry of the crystal field and mixes different @xmath48 of the mn spin .
the deviation from the cubic symmetry due to strains in the qd or the presence of any other anisotropy at the mn location are described by the additional crystal field terms @xmath49 and @xmath50 .
@xmath49 is the dominant term for a mn atom in a highly strained self - assembled qd .
it splits the mn spin states according to s@xmath36 and it is responsible for the long mn spin life time at zero field @xcite .
different s@xmath21 of the mn are coupled by the non - diagonal terms of @xmath51 .
for instance , the hyperfine terms @xmath45 couples two consecutive mn spin states through an electron - nuclei flip - flop .
an anisotropy of the mn site also couples mn spin states s@xmath21 separated by two units through the crystal field term @xmath52 . in the absence of magnetic anisotropy ( weak value of d@xmath53 ) , all these coupling terms at zero magnetic field prevent the optical pumping of the electronic mn spin .
the dynamics of the coupled electronic and nuclear spins of the mn is controlled by the time evolution of @xmath51 and the electronic mn spin relaxation time @xmath54 .
the nuclear spin relaxation is considered to be longer than the timescale of all the other spin relaxation mechanism discussed here .
to describe the magnetic field dependence of the resonant pl , we extend here the model presented in reference @xcite . the exciton states @xmath55 are laser coupled to the mn states
for a general description , valid from the low to the high optical excitation intensity regime , we consider that the resonant laser field induces a coherent coupling between the ground mn spin state and the exciton state described by a rabi energy @xmath57 . a pure dephasing time
@xmath58 ( _ i.e. _ not related to an exchange of energy with a reservoir ) is introduced for the exciton state .
the x - mn complex can relax its energy along a mn spin conserving channel with a characteristic time @xmath59 ( optical recombination of x ) or along channels including a relaxation of the mn spin with a spin flip time @xmath60 .
the parameter @xmath60 is a simplified effective way to describe the complex x - mn spin dynamics at the origin of the optical pumping mechanism @xcite . with this level scheme
, we can compute the steady state population @xcite of the resonantly excited exciton level @xmath61 .
this population is proportional to the intensity measured in the resonant pl experiment .
the evolution under longitudinal magnetic field of the calculated population of the exciton states @xmath55 is presented in fig . [ fig7 ] for different values of the crystal field anisotropy d@xmath53 and with e = d@xmath53/3 .
we chose e = d@xmath53/3 to obtain the maximum rhombic splitting and no preferential axial symmetry for the mn environment@xcite .
the main feature of the magnetic field dependence of the resonant pl observed experimentally ( _ i.e. _ the width , the asymmetry and shift of the maximum from b=0 t ) can be well reproduced with d@xmath53 in the 1 - 2@xmath33 range ( fig .
[ fig7](a ) ) . for a qualitative understanding of the behaviour of the resonant pl , we can look at the magnetic field dependence of the fine and hyperfine structure of the mn atom ( fig . [ fig7](c ) ) . for a negative magnetic field ,
the state s@xmath21=-5/2 is shifted away from the other mn spin states : the electron - nuclei flip - flops are blocked , the optical pumping is restored within a few mt and the resonant pl signal decreases abruptly . for a positive magnetic field of a few mt ,
the state s@xmath21=-5/2 is pushed to lower energy across the other mn spin states .
all the non - diagonal terms of @xmath51 ( in particular @xmath62 and @xmath50 terms ) mixes the different s@xmath21 and prevent the optical pumping of the electronic spin of the mn . to restore the optical pumping , a larger positive magnetic field is needed .
this explains that for a resonant excitation on s@xmath21=-5/2 , the drop of the resonant pl intensity is slower for positive magnetic fields .
the magnetic field dependence is reversed for an optical excitation of s@xmath21=+5/2 ( _ i.e _ opposite circular polarization for the excitation ) . for large positive or negative magnetic fields ,
the zeeman splitting of the mn ( controlled by g@xmath63=2 ) dominates the hyperfine coupling and all the crystal field terms .
the mn spin is quantized along the direction of the applied magnetic field , the optical pumping efficiency increases and the resonant pl vanishes . under resonant excitation for different values of the crystal field terms d@xmath53 and e = d@xmath53/3 ( the curves are vertically shifted for clarity ) .
the population of this level is proportional to the resonant pl signal .
the parameters used in the calculation are @xmath64 , @xmath65 , @xmath66 , @xmath67 , @xmath68 , @xmath69 , @xmath70 and @xmath71 .
( b ) calculated resonant pl signal for d@xmath53=2@xmath33 , e=0 ( black ) and e = d@xmath53/3 ( red ) .
( c ) calculated fine and hyperfine structure of a mn spin with @xmath69 , @xmath70 , @xmath71 , d@xmath53=1.5@xmath33 and e = d@xmath53/3 .
the black rectangle highlights the magnetic field range where the state s@xmath21=-5/2 significantly interacts with the other mn spin states .
( d ) experimental magnetic field dependence of the resonant pl signal in qd4 for different excitation intensities and @xmath43 excitation / @xmath42 detection on the high and low energy lines respectively .
the curves are normalized and vertically shifted for clarity.,width=336 ] the mn atoms are introduced in a thin cdte layer that is coherently grown on a cdte substrate . the first expectation would be that d@xmath53=0
. however , a residual crystal field in the @xmath33 range has to be introduced in the model to explain the experimental resonant pl data . for an axial crystal field ( _ i.e. _ e=0 ) ,
the calculated magnetic field dependence of the resonant pl is narrower and less asymmetric than in the experiment ( fig .
[ fig7](b ) ) .
this suggests that the remaining small crystal field felt by the mn atom in our structure has no axial symmetry but arises from a more disordered environment . this crystal field could arise from an alloy composition fluctuation around the mn atom
one should then keep in mind that the mn atoms are located in a cdte layer that is only 4 monolayers thick and surrounded by cd@xmath2mg@xmath72te barriers .
a mn atom interacting with a confined exciton is always close to the barrier and the substitution of some cd atoms by mg atoms in the vicinity of the mn reduces the symmetry of the local electric field and can induce weak crystal field terms d@xmath53 and e @xcite .
while further investigations would be needed to fully understand this point , we observe that d@xmath53 is positive as one could expect from the lattice parameter difference between mgte and cdte : mgte has a smaller lattice parameter than cdte , so the presence of a mg atom in the vicinity of the mn induces a local reduction of the lattice parameter and then a positive crystal field term @xcite . as presented in fig .
[ fig7](d ) , the magnetic field dependence of the resonant pl is observed in a broad range of excitation intensities .
however , the width of the resonant pl peak and the residual background of resonant pl increase with the increase of the excitation intensity .
this reflects a reduction of the optical pumping efficiency at high excitation power .
this reduction at high excitation power is likely due to a perturbation of the mn spin by the additional non - resonant injection of free carriers that decreases the mn spin memory @xcite .
the dynamics resulting from the coupling of the electronic and nuclear spins of the mn in this weak crystal field can be observed in the second - order correlation function of the resonant pl signal of x - mn .
such experiment in strain free mn - doped qds is made possible by the absence of optical pumping of the mn at zero magnetic field and the resulting large resonant pl intensity .
we performed auto - correlation of the resonant pl using an hbt set - up @xcite with a time resolution of about 0.7 ns . in these start - stop experiments ,
the detection of the first photon indicates by its energy and polarisation that the mn spin has a given orientation .
the detection probability of a second photon with the same energy and polarisation is then proportional to the probability of conserving this spin state .
the time evolution of this intensity correlation signal is a probe of the spin dynamics of the mn atom and of the x - mn complex . ) .
the black curve corresponds to the pl decay of the biexciton in qd6 where no contribution of the dark exciton at long time delay is observed.,width=336 ] examples of excitation power dependence of the auto - correlation of the resonant pl are given in figure [ fig8 ] for two qds with a very different exciton - mn overlap ( qd3 and qd6 ) .
the data are obtained under cross - circular excitation an detection on the high and low energy lines of x - mn respectively ( experimental configuration illustrated in the inset of fig . [ fig6](c ) ) . with this configuration ,
we resonantly excite and detect the same spin state of the mn ( either s@xmath21=+5/2 for a @xmath42 excitation or s@xmath21=-5/2 for a @xmath43 excitation ) .
the auto - correlation signal presents first a dip near zero delay characteristic of a single - photon emitter .
the width of this antibunching signal is given by the lifetime of the emitter ( inset of fig .
[ fig8](b ) ) and the generation rate of excitons .
its depth is limited here by the time resolution of the hbt setup .
in addition , the large bunching at short delays reveals a significant pl intermittency . the width and
the amplitude of the bunching decreases with the increase of the excitation intensity .
the decrease of the amplitude of this bunching signal in a transverse magnetic field of a few tens of mt ( fig .
[ fig9](a ) ) confirms that the pl intermittency results mainly from fluctuations of the mn spin . to model this bunching signal
, we have to take into account the spin dynamics of the mn and x - mn . in a simplified level scheme ,
we neglect the hyperfine splitting within the x - mn complex and we consider that a single exciton state , @xmath73 , is laser coupled to the mn spin states @xmath56 with a generation rate @xmath74 .
the dynamics of the mn spin in an empty qd is described by the time evolution of @xmath51 and the relaxation time @xmath54 describing the coupling by spin - flips of one unit the different spin states @xmath48 .
the 24 x - mn levels are obtained from the diagonalization of @xmath75 .
x - mn can optically recombine along a mn spin conserving channel with a lifetime @xmath59 for the bright exciton or @xmath76 for the dark exciton .
x - mn can also relax along channels involving a change of the spin by one unit with a characteristic time @xmath77 .
this relaxation path , combined with an optical recombination of the exciton , permits a transfer of population from the exciton state @xmath78 to any other spin state of the mn @xmath48 . to describe the main x - mn spin relaxation channel revealed by the resonant pl spectra ( _ i.e _ relaxation of the exciton with conservation of the mn spin observed in fig.[fig5 ] ) , we consider a transfer rate between bright excitons with a characteristic time @xmath79 . at finite temperature , the intraband relaxation rates @xmath80 between the different states of the x - mn complex depend on their energy separation @xmath81 .
here we use @xmath80=1/@xmath82 if @xmath83 and @xmath80=1/@xmath84 if @xmath85 ( @xmath86 corresponding either to @xmath77 or @xmath79 ) @xcite .
this describes a partial thermalization among the 24 x - mn levels during the lifetime of the bright or dark excitons , as observed experimentally ( see fig .
[ fig4 ] ) . in this model
we also consider that for a magnetic field lower than a few hundreds mt , the zeeman energy of the exciton can be neglected since it is much smaller than the x - mn exchange interaction .
we only take into account the effect of the magnetic field on the mn in the empty qd ( last term of @xmath51 ) . using this level scheme
, we can calculate the time evolution of the 60x60 density matrix describing the population and the coherence of the 36 states of the mn alone ( empty qd described by @xmath51 ) and the 24 x - mn states described by @xmath75 . to model the auto - correlation of the resonant pl , the initial state of the system is set to @xmath56 and one monitors the time evolution of the population of @xmath87 ( low energy line ) for an excitation of @xmath78 ( high energy line ) with a generation rate @xmath74 .
when normalized to one at a long time delay , this time evolution accounts for the auto - correlation function of the transition associated with the level @xmath87 . .
the curves are obtained after convolution with the response of the hbt setup ( @xmath88 ) .
the inset shows the calculated autocorrelation at short time delay before convolution with the response set - up .
the parameters used in the calculation are : @xmath89 , @xmath65 , @xmath90 , @xmath66 , @xmath91 @xmath67 , @xmath69 , @xmath70 , @xmath71 , d@xmath53=1.5@xmath33 , e = d@xmath53/3 and the qd parameters of qd3 ( see table [ table1]).,width=336 ] auto - correlation signals calculated with and without transverse magnetic field are presented in fig .
[ fig9](b ) . in a weak transverse magnetic field ( inset of fig .
[ fig9](b ) ) , the auto - correlation displays small oscillations at short delay due to the precession of the mn electronic spin . the limited time resolution of the hbt setup prevents the observation of these oscillations . to take into account this time resolution ,
the calculated curves are convoluted with the response of the system ( fig .
[ fig9](b ) ) . as in the experiment
, the calculated auto - correlation presents a large bunching at short time delay with a typical half width at half maximum of about 4 ns . this short mn spin memory is due to its precession in the hyperfine field of its nuclei @xcite . at low excitation intensity and at zero magnetic field ,
the width of the experimental photon bunching is well reproduced by the model . in a weak transverse magnetic field ,
the amplitude of the bunching signal decreases significantly .
the transverse magnetic field dependence of the shape of the calculated bunching signal ( width and amplitude ) is in qualitative agreement with the experiment ( fig .
[ fig9 ] ) .
a magnetic field of a few tens of @xmath92 has no effect on the x - mn system and do not change the contribution of the x - mn dynamics on the bunching amplitude .
the observed and calculated decrease of the bunching amplitude is then induced by the precession of the mn spin in the transverse magnetic field when the qd is empty .
this confirms that in our experiments the bunching signal at zero field is mainly controlled by the dynamics of the coupled electronic and nuclear spins of the mn in a weak residual crystal field .
the width of the bunching signal measured at low excitation power is in good agreement with the calculated mn spin dynamics . however , at high excitation intensity a reduction of the width and the amplitude of the bunching is observed ( fig .
[ fig8 ] ) .
this reduction can not be explained by the presented model and it likely comes from a perturbation of the mn spin by non - resonantly injected free carriers in the vicinity of the qd .
a similar increase of the mn spin relaxation rate was already observed for much confined cdte / znte qds under non - resonant excitation @xcite .
we have incorporated and optically probed individual mn atoms in strain free cdte qds formed by interface fluctuations in thin quantum wells . despite the weak lateral confinement and a possible valence band mixing induced by an anisotropic shape of the confinement potential in these natural qds
, we have shown that an individual mn atom can be optically addressed by a resonant optical excitation . magnetic field dependence and auto - correlation of the photoluminescence signal obtained under resonant excitation of the qd
ground state show that , in these strain free structures , an isolated mn atom experiences a residual weak crystal field splitting likely due to local alloy fluctuations .
an efficient optical pumping the mn spin is obtained under a weak magnetic field in the faraday geometry . despite this residual crystal field ,
strain free magnetic qds are promising systems to probe the coherent dynamics of the coupled electronic and nuclear spins of a mn atom in zero or weak magnetic field .
these strain free mn - doped qds could also be included in photonic structures such as optical planar micro - cavities to increase the number of collected photons and their interaction with light .
this would open the possibility to realize an optical coherent control of an individual mn spin in a weak transverse field using the optical stark effect or to probe and control the dynamics of the nuclear spin of the mn atom .
p. m. koenraad and m. e. flatte , nature materials * 10 * , 91 ( 2011 ) .
l. besombes , y. leger , l. maingault , d. ferrand , h. mariette and j. cibert , phys . rev .
lett . * 93 * , 207403 ( 2004 ) .
l. besombes , c.l .
cao , s. jamet , h. boukari and j. fernandez - rossier , phys .
b * 86 * , 165306 ( 2012 ) .
a. kudelski , a. lemaitre , a. miard , p. voisin , t.c.m .
graham , r.j . warburton and o. krebs , phys .
lett . * 99 * , 247209 ( 2007 ) .
o. krebs and a. lemaitre , phys .
lett . * 111 * , 187401 ( 2013 ) . c. le gall , l. besombes , h. boukari , r. kolodka , j. cibert and h. mariette , phys .
102 * , 127402 ( 2009 ) .
m. goryca , t. kazimierczuk , m. nawrocki , a. golnik , j. a. gaj , p. kossacki , p. wojnar and g. karczewski , phys .
* 103 * , 087401 ( 2009 ) . c. le gall , r. s. kolodka , c. l. cao , h. boukari , h. mariette , j. fernandez - rossier and l. besombes , phys .
b * 81 * , 245315 ( 2010 )
reiter , t. kuhn and v.m .
axt , phys .
b * 85 * , 045308 ( 2012 ) .
l. besombes , k. kheng and d. martrou , phys .
* 85 * , 425 ( 2000 ) .
d. martrou , j. eymery and n. magnea , phys .
* 83 * , 2366 ( 1999 ) .
j. fernandez - rossier , phys .
b * 73 * , 045301 ( 2006 )
glazov , e.l .
ivchenko , l. besombes , y. leger , l. maingault , h. mariette , phys . rev .
b * 75 * , 205313 ( 2007 ) .
y. leger , l. besombes , l. maingault , and h. mariette , phys .
b * 76 * , 045331 ( 2007 ) . c. l. cao , l. besombes and j. fernandez - rossier , phys . rev .
b * 84 * , 205305 ( 2011 ) .
l. besombes , y. leger , l. maingault , d. ferrand , h. mariette , j. cibert , phys .
b * 71 * , 161307(r ) ( 2005 ) .
anna h. trojnar , marek korkusinski , udson c. mendes , mateusz goryca , maciej koperski , tomasz smolenski , piotr kossacki , piotr wojnar , pawel hawrylak , phys .
b * 87 * , 205311 ( 2013 ) .
liao , c - c .
liao , c - h .
ku , y - c .
chang , s - j .
cheng , m. jo , t. kuroda , t. mano , m. abbarchi , and k. sakoda , phys .
b 86 , 115323 , ( 2012 ) .
u. bockelmann and g. bastard , phys .
b * 45 * 1688 ( 1992 ) .
y. leger , l. besombes , j. fernandez - rossier , l. maingault , h. mariette , phys .
lett . * 97 * , 107401 ( 2006 ) .
e. tsitsihvili , r.v .
baltz , h. kalt , phys .
b * 67 * , 205330 ( 2003 ) .
a.v . khaetskii and y.v .
nazarov , phys .
b * 64 * 125316 ( 2001 ) .
l. m. woods , t. l. reinecke , y. lyanda - geller , phys .
b * 66 * , 161318(r ) ( 2002 ) .
z. wang , w. zheng , j. van tol , n. s. dalal , g. f. strouse , chemical physics letters * 524 * , 73 ( 2012 ) .
l. besombes , y. leger , j. bernos , h. boukari , h. mariette , j.p .
poizat , t. clement , j. fernandez - rossier and r. aguado , phys . rev .
b * 78 * 125324 ( 2008 ) .
m. qazzaz , g. yang , s.h .
xin , l. montes , h. luo and j.k .
furdyna , solid state communications * 96 * , 405 ( 1995 ) .
causa , m. tovar , s.b .
oseroff , r. calvo and w. giriat , phys .
* a77 * , 473 ( 1980 ) . l. cywinski , phys .
b * 82 * , 075321 ( 2010 ) .
a. o. govorov , a. v. kalameitsev , phys .
b * 71 * , 035338 ( 2005 ) .
s. jamet , h. boukari and l. besombes , phys .
b * 87 * , 245306 ( 2013 ) .
s. cronenberger , m. vladimirova , s. v. andreev , m. b. lifshits and d. scalbert , phys .
lett . * 110 * , 077403 ( 2013 ) .
d. gatteschi , r. sessoli , j. villain , molecular nanomagnets , oxford university press ( 2006 ) . d.l . griscom and r.e .
griscom , the journal of chemical physics * 47 * , 2711 ( 1967 ) . | we report on the spin properties of individual mn atoms in ii - vi semiconductor strain free quantum dots .
strain free mn - doped cdte quantum dots are formed by width fluctuations in thin quantum wells lattice matched on a cdte substrate .
these quantum dots permit to optically probe and address any spin state of a mn atom in a controlled strain environment .
the absence of strain induced magnetic anisotropy prevents an optical pumping of the mn spin at zero magnetic field .
thus , a large photoluminescence is obtained under resonant optical excitation of the exciton - mn complex .
an efficient optical pumping of the coupled electronic and nuclear spins of the mn is restored under a weak magnetic field .
the observed reduction of the resonant photoluminescence intensity under magnetic field is well described by a model including the hyperfine coupling and a residual crystal field splitting of the mn atom .
finally , we show that the second order correlation function of the resonant photoluminescence presents a large photon bunching at short delay which is a probe of the dynamics of coupled electronic and nuclear spins of the mn atom . |
motivated by works on lifshitz scaling in condensed matter physics , it is possible to take different lifshitz scaling for space and time , and the resultant theory is called horava - lifshitz gravity @xcite . in the ir limit
this gravity reduces to general relativity , and so it can be considered as a uv completion of general relativity . motivated by this work on horava - lifshitz gravity , different lifshitz scaling for space and time have been considered for type iia string theory @xcite , type iib string theory @xcite , ads / cft correspondence @xcite , dilatonic black branes @xcite , and dilatonic black holes @xcite . another uv completion theory of general relativity which reduces to general relativity in the ir limit is called gravity s rainbow @xcite .
in fact , it has been demonstrated that gravity s rainbow is related to horava - lifshitz gravity @xcite .
this is because both of these theories are based on the modifying the usual energy - momentum dispersion relation in the uv limit such that it reduces to the usual energy - momentum dispersion relation in the ir limit .
it may be noted that such modification of the usual energy - momentum has also been obtained in discrete spacetime @xcite , spacetime foam @xcite , spin - network in loop quantum gravity ( lqg ) @xcite , ghost condensation @xcite and non - commutative geometry @xcite .
the non - commutative geometry occurs due to background fluxes in string theory @xcite , and is used to derive one of the most important rainbow functions in gravity s rainbow @xcite .
it may be noted that the uv modification of the usual energy - momentum relation implies the breaking of the lorentz symmetry in the uv limit of the theory .
the spontaneous breaking of the lorentz symmetry can occur in string theory because of the existence of an unstable perturbative string vacuum @xcite .
it is possible for a tachyon field to have wrong sign for its mass squared in string field theory , and this causes the perturbative string vacuum to become unstable .
the theory becomes ill - defined if the vacuum expectation value of the tachyon field is infinite .
it is also possible for the vacuum expectation value of the tachyon field to be finite and negative . in this case , the coefficient of the quadratic term for the massless vector field is nonzero and negative , and this breaks the lorentz symmetry .
the spontaneous breaking of the lorentz symmetry in string theory has also been investigated using the gravitational version of the higgs mechanism @xcite .
this has been done for the low energy effective action obtained from string theory .
the lorentz symmetry breaking has been also studied using black brane in type iib string theory @xcite . in this analysis ,
moduli stabilization was studied using a kklt - type moduli potential in the context of type iib warped flux compactification .
it was demonstrated that a higgs phase for gravity will exist if all moduli are stabilized .
another study regarding the breaking of lorentz symmetry was done by using compactification in string field theory @xcite .
it may be noted that various other approaches to quantum gravity also indicate that the lorentz symmetry might only be an effective symmetry which occurs in the ir limit of some fundamental theories of quantum gravity @xcite .
hence , there is a good motivation to study the uv deformation of geometries that occur in string theory .
in fact , motivated by lifshitz deformation of such geometries , and the relation between horava - lifshitz gravity and gravity s rainbow @xcite , recently rainbow deformation of geometries that occur in string theory has been performed .
thus , the modifications of the thermodynamics of black rings has been analyzed using gravity s rainbow @xcite .
it has been observed that a remnant exists for black rings in gravity s rainbow .
it has also been argued that a remnant might exist for all black objects in gravity s rainbow @xcite .
this has been explicitly demonstrated for kerr black holes , kerr - newman black holes in de sitter space , charged ads black holes , higher dimensional kerr - ads black holes and black saturn @xcite .
this was done by generalizing the work done on the thermodynamics of black holes in gravity s rainbow @xcite . as the usual uncertainty principle still holds in gravity s rainbow @xcite , and it is possible to obtain a lower bound on the energy @xmath0 , using the usual uncertainty principle .
this energy can be related to the energy of a particle emitted in hawking radiation .
furthermore , the value of the uncertainty in position can be equated to the radius of the event horizon , @xmath1 this energy can be related to the energy at which spacetime is probed , and hence it describes the energy @xmath2 in gravity s rainbow .
this is because effectively this particle emitted with the energy @xmath2 can be viewed as a probe which is probing the geometry of the black hole .
this consideration modifies the temperature of the black hole @xcite .
the entropy and heat capacity of black hole in gravity s rainbow can be calculated using this modified temperature .
an interesting consequence of this modified solution is that it predicts the existence of remnants for the black hole .
thus , the temperature of the black hole reduces to zero , when black hole has a small but finite size . at this size
, the black hole does not emit any hawking radiation .
the existence of black hole remnants can be used as a solution for the information paradox @xcite .
furthermore , it also solves problem related to the existence of naked singularity at the last stage of the evaporation of a black hole . as in this picture ,
a black hole does not evaporate completely producing a naked singularity , but rather a remnant is produced at the last stage of the evaporation of the black hole .
the existence of a remnant also has phenomenological consequences .
this is because it is not possible to produce black holes smaller than these remnants .
this increases the energy at which mini black holes can be produced at the lhc @xcite .
recently , a lot of interest has been generated in gravity s rainbow @xcite .
it may be noted that the rainbow functions have been constrained from experimental data @xcite .
black hole solutions in gravity s rainbow with nonlinear sources have been investigated in @xcite .
in addition , the hydrostatic equilibrium equation for this gravity was obtained in ref .
@xcite . as there is a strong motivation to study rainbow deformation of geometries that occur in string theory , so we analyze the rainbow deformation of charged dilatonic black holes in this paper . it may be noted that dilaton gravity arises as a low - energy effective field theory of string theory @xcite .
the dilaton field is also a candidate for dark matter @xcite .
in fact , in order to have better picture of nature of the dark energy , a new scalar field is added to the field content of the original theory @xcite .
black objects in presence of dilaton gravity , have also been investigated @xcite .
recently , dilaton field has been used for analyzing compact objects and hydrostatic equilibrium of starts @xcite .
the evaporation of quantum black holes has also been investigated using two dimensional dilaton gravity @xcite .
motivated by these applications , we analyze dilaton field using the formalism of gravity s rainbow .
thermodynamical aspects of black holes have been of a great interest ever since of pioneering works of hawking and beckenstein @xcite . the idea that geometrical aspects of black holes could be interpreted as thermodynamical quantities provides a deep insight into the connection between gravity and quantum mechanics .
on the other hand , introduction and developments in gauge / gravity duality highlighted the importance of black holes thermodynamics @xcite .
in addition , hawking and page showed the existence of a phase transition for asymptotically anti de - sitter black holes @xcite .
this phase transition was reconsidered through the use of ads / cft correspondence by witten @xcite .
these works motivated a large number of researches to be conducted in context of black holes thermodynamics , stability and their phase transitions @xcite .
recently , it has been demonstrated that , it is possible to treat the cosmological constant as the thermodynamic pressure in extended phase space .
there are several reasons for such consideration which among them one can point out the existence of second order phase transition for black holes , van der waals like liquid / gas behavior in phase diagrams and formation of the triple point @xcite .
the consideration of the cosmological constant as a thermodynamical variable could be supported by studies that are conducted in context of ads / cft @xcite .
in addition , it was shown that a case of ensemble dependency exists for charged @xmath3-dimensional black holes which could be removed by considering cosmological constant as a thermodynamical variable @xcite .
the thermodynamical critical behavior of black holes in presence of different matter fields and gravities has been investigated in literature @xcite another interesting method of studying thermodynamical structure of the black holes is through the use of geometry .
it is proposed that one can build a phase space of the black holes by employing one of thermodynamical quantities of the black holes as thermodynamical potential and its corresponding extensive parameters as components of the phase space .
the information regarding phase transitions of the black holes is within the singularities of ricci scalar of the constructed phase space . in other words , the divergencies of the ricci scalar of thermodynamical metric
are representing bound and phase transition points .
the thermodynamical potential for this method could be mass which is used in weinhold @xcite , quevedo @xcite and hpem @xcite metrics or entropy which is employed in ruppeiner @xcite metric .
it was pointed out that ruppeiner and weinhold methods are related to each other with temperature as conformal factor @xcite .
it was shown that for specific cases of black holes , weinhold , ruppeiner and quevedo metrics may fail to provide consistent results regarding phase transitions while the hpem metric is proven to be successful one @xcite . in what follows ,
we use all of these methods to study phase transitions of the black holes .
this paper is organized as follows .
we obtain the charged black hole solutions in dilaton gravity s rainbow and analyze their properties .
this will be done by making the metric of charged black hole solutions in dilaton gravity depends on the energy .
we also examine the first law of thermodynamics for this solution .
next , we study the stability of such solutions in gravity s rainbow and phase transition of these black holes through heat capacity , geometrical thermodynamics and the analogy between cosmological constant and thermodynamical pressure .
finally , we obtain critical pressure and horizon radius through another method .
last section is devoted to conclusion .
in this section , we obtain charged black hole solutions in dilaton gravity s rainbow and investigate their properties . this will be done by writing an energy dependent version of the metric for dilaton - maxwell gravity . it may be noted that gravity s rainbow is based on the generalization of doubly special relativity @xcite , and so it is not possible for a particle to attain energy greater than the planck energy in gravity s rainbow .
this is because gravity affects particles of different energies differently , and so the spacetime is represented by a family of energy dependent metrics in gravity s rainbow @xcite .
the gravity s rainbow can be constructed by considering the following deformation of the standard energy - momentum relation @xmath4 where energy ratio is @xmath5 , in which @xmath2 and @xmath6 are , respectively , the energy of test particle and the planck energy . the functions @xmath7 and @xmath8 are required to be constrained in such a way that the standard energy - momentum relation is obtained in the infrared limit .
thus , we require @xmath9 it may be noted that the spacetime is probed at the energy @xmath2 , and by definition this can not be greater than the planck energy . @xmath10 and @xmath11 are called the rainbow functions and their functional forms are phenomenologically motivated .
now it is possible to define an energy dependent deformation of the metric @xmath12 as @xcite @xmath13 where @xmath14 here @xmath15 and @xmath16 refer to the energy independent frame fields .
the @xmath17-dimensional action of charged dilaton gravity is @xcite @xmath18 , \label{action}\]]where @xmath19 is the ricci scalar curvature , @xmath20 is the dilaton field and @xmath21 is a potential for @xmath20 .
the electromagnetic field is @xmath22 in which @xmath23 is the electromagnetic potential .
in addition , it should be pointed out that @xmath24 is a constant which determines the strength of coupling of the scalar and electromagnetic field . due to the fact that we are looking for the black hole with a radial electric field ( @xmath25 )
, the electromagnetic potential will be in the following form @xmath26 using variational principle and varying eq .
( [ action ] ) with respect to the gravitational field @xmath27 , the dilaton field @xmath20 and the gauge field @xmath23 , we can obtain the following field equations @xmath28@xmath29@xmath30 in this paper we are attempting to obtain dilaton - maxwell rainbow solutions .
to do so , one can employ following static metric ansatz @xmath31 , \label{metric}\]]where @xmath32 and @xmath33 are radial dependent functions which should be determined , and @xmath34 represents the line element of a @xmath35 dimensional hypersurface with the constant curvature @xmath36 and volume @xmath37 .
we should note that the constant @xmath38 indicates that the boundary of @xmath39 and @xmath40 can be a positive ( elliptic ) , zero ( flat ) or negative ( hyperbolic ) , constant curvature hypersurface with following explicit forms @xmath41 using eq .
( [ maxwell equation ] ) , one can obtain electromagnetic tensor as @xmath42where @xmath43 is an integration constant which is related to the electric charge of the black hole . here
, in order to find consistent metric functions , we use a modified version of liouville - type dilation potential with following form @xmath44where @xmath45 and @xmath46 is a free parameter which plays the role of the cosmological constant .
it is worthwhile to mention that for the case of @xmath47 , one obtains @xmath48which is the usual liouville - type dilation potential that is used in the context of friedman - robertson - walker scalar field cosmologies @xcite and einstein - maxwell - dilaton black holes @xcite .
next , we employ an ansatz , @xmath49 , in the field equations .
the motivation for considering such an ansatz is due to black string solutions of einstein - maxwell - dilaton gravity which was first introduced in ref .
now , we are in a position to obtain metric functions .
it is a matter of calculation to show that by using eq .
( [ ftr eq ] ) , the metric ( [ metric ] ) and the mentioned ansatz for @xmath33 , we have following solutions for the field equations , ( eqs .
( [ dilaton equation(i ) ] ) and ( [ dilaton equation(ii ) ] ) ) @xmath50@xmath51where @xmath52 is an arbitrary constant and @xmath53 . in the above expression
, @xmath54 is an integration constant which is related to the total mass of the black hole .
it is notable that , in the absence of a non - trivial dilaton ( @xmath55 ) , the solution ( f(r ) ) reduces to @xmath56which describes a @xmath17-dimensional asymptotically ads topological charged black hole in gravity s rainbow with a positive , zero or negative constant curvature hypersurface . in order to confirm black hole interpretation of the solutions , we look for the curvature singularity .
to do so , we calculate the kretschmann scalar .
calculations show that for finite values of radial coordinate , the kretschmann scalar is finite . on the other hand , for very small and very large values of @xmath57
, we obtain @xmath58 equation ( [ rr0 ] ) confirms that there is an essential singularity located at @xmath59 , while eq .
( [ rrinf ] ) shows that for nonzero @xmath24 , the asymptotical behavior of the solutions is not ads .
it is easy to show that the metric function may contain real positive roots ( see fig .
[ metricfig ] ) , and therefore , the curvature singularity can be covered with an event horizon and interpreted as a black hole .
now , we are in a position to calculate thermodynamic and conserved quantities of obtained solutions and examine the validity of the first law of thermodynamics . in order to obtain the temperature
, we use the concept of surface gravity to show that the temperature of these solutions has following form @xmath61 . \label{temp}\ ] ] on the other hand , one can use the area law for extracting modified version of the entropy related to the einsteinian class of black objects with the following structure @xmath62 in which by setting @xmath63 and @xmath64 , the entropy of einstein - maxwell - dilation black holes is recovered . in order to find the total electric charge of the solutions
, one can use the gauss law .
calculating the flux of electric field helps us to find the total electric charge with the following form @xmath65 next , we are interested in obtaining the electric potential . using following standard relation , one can obtain the electric potential at the event horizon with respect to the infinity as a reference @xmath66 finally , according to the definition of mass due to abbott and deser ad1,ad2,ad3 , the total mass of the solution is @xmath67 it is worthwhile to mention that for limiting case of @xmath68 and @xmath63 , eq .
( [ mass ] ) reduces to mass of the einstein - maxwell black holes @xcite .
in addition , in obtained conserved and thermodynamical quantities , only the electric potential remains unaffected by considering gravity s rainbow .
now , we are in a position to check the validity of the first law of thermodynamics .
to do so , first , we calculate the geometrical mass , @xmath54 , by using @xmath69 . then by employing obtained relation for geometrical mass and eq .
( [ mass ] ) for total mass of the black holes , we find @xmath70where@xmath71 .\ ] ] it is a matter of calculation to show that @xmath72 therefore , we proved that the first law is valid as @xmath73
in this section , we study thermal stability of the solutions in context of the canonical ensemble .
the stability conditions in the context of canonical ensemble are determined by the sign of the heat capacity . in other words , the positivity / negativity of the heat capacity
is denoted as the black object being in stable / unstable state .
therefore , in order to study the stability of the charged black holes in dilatonic gravity s rainbow , we study the changes in the sign of corresponding heat capacity . it is worthwhile to mention that investigating the behavior of the heat capacity enables one to obtain the phase transitions of the solutions at the same time .
the root and divergence point of the heat capacity are denoted as bounded point ( @xmath74 ) and second order phase transition point ( @xmath75 ) , respectively . bounded point is related to the root of temperature and the sign of @xmath76 changes at @xmath74 , while we expect to obtain positive temperature at @xmath75 .
the system in canonical ensemble is considered to be in fixed charge .
therefore , we have @xmath77 considering the mentioned bounded and phase transition points , one can obtain @xmath78 there are three valuable known cases for the rainbow functions which are characteristics of the rainbow solutions .
these three cases are arisen from different phenomenological origins with an upper limit for considering the energy of test particle @xmath2 @xmath79 the first case is originated from loop quantum gravity and non - commutative geometry . in this case , we have following relations for rainbow functions of metric @xcite @xmath80 the other case is constructed by considering the hard spectra from gamma - ray bursts which leads to @xcite @xmath81 interestingly , opposite to previous case , in this one the effect of @xmath82 which is coupled to spatial coordinates of the metric is vanished , whereas in the case one which is related to loop quantum gravity the effect of coupling term for time component of the metric is vanished . finally , in the case three , the choices of rainbow functions are due to constancy of the velocity of the light @xcite
@xmath83 using first law of thermodynamics , one can rewrite the relation for heat capacity into @xmath84 now , by employing eqs .
( [ temp ] ) and ( [ entropy ] ) with ( [ cqq ] ) , one can show that heat capacity is@xmath85 considering obtained relation for the heat capacity ( eq.([heat ] ) ) and three mentioned cases for the rainbow functions of the metric ( eqs .
( loop-[light ] ) ) , we study the stability of the solutions . in case of horizon flat ( @xmath86 ) , one can find that the root(s ) of the heat capacity and divergence point(s ) are given by following relations @xmath87@xmath88 interestingly , for horizon flat , the root of the heat capacity , hence bound point is independent of the dilaton parameter , @xmath89 while divergency of the heat capacity is a function of this parameter . on the other hand , in order to have positive real valued divergency in ads spacetime , we have @xmath90 restriction for dilaton parameter .
this condition for ds spacetime is opposite .
in other words , the real valued divergency is obtained if @xmath91 .
( see eqs .
( [ root - flat ] ) and ( [ div ] ) for more details ) .
in order to charged black holes in dilatonic gravity s rainbow with flat horizon be stable , following conditions must be hold @xmath92 as for the cases of @xmath93 , it was not possible to obtain the analytical relations for the root and divergence point of the heat capacity .
therefore , we employ numerical method for studying the properties of the heat capacity for spherical and hyperbolic horizons . as for the stability conditions ,
there are different orders of the horizon radius for each term .
these terms will have dominant effect in specific regions of horizon radius and other parameters .
considering the effectiveness of these terms , the stability conditions will vary from one case to another one .
also , these effective behaviors may present different regions of stability and instability . taking a closer look at different terms
, one can see that the most effective parameter in stability conditions which modifies the exponent of horizon radius highly and changes the positivity and negativity of each term , is dilaton parameter , @xmath94 .
in other words , considering different values of @xmath94 , stable / unstable regions will be modified highly .
this highlights the effect of dilaton field on thermodynamical behavior of the solutions . in order to give a better insight regarding the thermodynamical behavior of these black holes , we study the behavior of the temperature
the reason is the fact that negativity of the temperature represents nonphysical systems which are not of our interest .
therefore , we study the conditions for positivity / negativity of the temperature . considering eq .
( [ temp ] ) , there are three terms which are related to electric charge , cosmological constant and topological structure of the metric .
the effectiveness of each term is a function of their factors .
therefore , considering different values for these factors may lead to one of the following scenarios : one root and temperature is an increasing function of horizon radius ( left panel of fig .
[ fig1 ] ) , no root and temperature is negative with one maximum ( right panel of fig .
[ fig1 ] ) , two roots with one region of positivity and two regions of negativity ( fig .
[ fig2 ] ) , one root and temperature is a decreasing function of the horizon radius ( fig .
[ fig3 ] ) . in general ,
charged term is always negative in eq .
( [ temp ] ) .
if one considers ads solutions , the second term will be positive .
as for the last term , if spherical solution is chosen , then in order for topological term be positive , the dilaton parameter must be @xmath95 .
whereas in case of hyperbolic horizon the condition will be modified into @xmath96 . on the other hand , if ds solutions are considered , the second term will be negative .
therefore , the possibility of having positive temperature depends on topological term with mentioned conditions for spherical and hyperbolic cases .
it is worthwhile to mention that horizon flat of ds solutions has negative temperature .
therefore , it is not physical .
considering the mentioned changes in different terms of temperature , depending on the dominant regions of the different each term , the temperature will have positive / negative value with different behavior ( which were pointed out in plotted diagrams ) .
@xmath97 @xmath98 @xmath99 @xmath100 in order to elaborate the effects of the dilation field on thermal stability and mentioned behaviors for the temperature , we plot various diagrams using the first model of rainbow functions of gravity s rainbow .
it is evident that in case of temperature being an increasing function of horizon radius , for positive temperature , we have stable black holes and heat capacity is an increasing function of @xmath101 ( fig .
[ fig1 ] left panel ) .
increasing the dilaton parameter will lead to formation of two stable and unstable states where both of these states are in negative temperature ( fig .
[ fig1 ] right panel ) . on the other hand , by increasing dilaton parameter , the temperature will have two regions of negativity and one positivity . in the positive region , a phase transition takes place between unstable larger state to smaller stable state .
this phase transition point is represented by a divergency of the heat capacity .
increasing dilaton parameter leads to decreasing the place of divergency of the heat capacity and increasing the region in which we have unstable physical solutions ( fig .
[ fig2 ] ) . the larger root of the temperature is highly sensitive to variation of the @xmath24 ( see fig .
[ fig2 ] ) . for sufficiently large values of the dilaton parameter ,
the temperature is a decreasing function of the horizon radius with one root .
the heat capacity in the region of the positive temperature is negative , therefore in this case rainbow solutions are unstable ( fig .
[ fig3 ] ) .
it is worthwhile to mention that the root in this case is an increasing function of dilaton parameter ( compare two diagrams of fig .
[ fig3 ] ) .
in this section , we will conduct a study regarding the phase transition points of the obtained black holes through the use of geometrical thermodynamics concept . in this concept
, one can construct the thermodynamical structure of the black hole through the use of thermodynamical variables .
in other words , by using one of the thermodynamical variable as potential and its corresponding extensive parameters , it is possible to build phase space .
the singularities of the ricci scalar of this phase space , marks two different properties of the solutions : one is a bound point which is related to the root of the temperature and marks the physical and non - physical solutions . the other one is related to singularities of the heat capacity which are marking the points which system goes under phase transition .
considering such property for the constructed phase space , a valid approach of the geometrical thermodynamics produces a ricci scalar which has singularities that cover both of the mentioned points . depending on the thermodynamical potential
, the extensive parameters would be different for each phase space .
one of these potentials could be entropy which could be used to construct ruppeiner phase space .
an other potential could be mass which is employed to build weinhold , quevedo and hpem phase spaces . the mentioned phase spaces have following forms @xcite @xmath102where their corresponding denominator of their ricci scalars are @xmath103 in which @xmath104 , @xmath105 , @xmath106 and @xmath107 .
now , by using eqs .
( [ entropy ] ) , ( [ q ] ) , ( [ mass2 ] ) and ( [ met ] ) , one can construct mentioned phase spaces and calculate their corresponding curvature scalar for these black holes .
due to economical reasons , we will not present obtained relation for ricci scalar but demonstrate results in plotted diagrams ( see figs .
[ fig1]-[fig3 ] for hpem metric and fig .
[ fig33 ] for other metrics ) . fig .
[ fig33 ] shows that for specific values one can find cases in which weinhold ( left panel of fig .
[ fig33 ] ) , ruppeiner ( middle panel of fig .
[ fig33 ] ) and quevedo ( right panel fig .
[ fig33 ] ) will not produce suitable divergencies in their ricci scalar to cover mentioned points .
in other words , the divergencies of their ricci scalar may not coincide with root and divergencies of the heat capacity . on the other hand
, it is seen that all the divergencies of the curvature scalar of the hpem metric match with bound and phase transition points of the heat capacity .
the nature of the behavior of the ricci scalar around each one of these divergencies enables one to recognize whether it is a bound point or a divergence point in heat capacity @xcite .
in this section , we investigate the existence of second order phase transition through the analogy between negative cosmological constant and thermodynamical pressure . the usual relation for pressure and cosmological constant
is give by @xcite @xmath108 it was shown that the gravitational theory under consideration may affect this relation and modifies it @xcite . in calculations of the conserved and thermodynamical quantities , we found that these quantities were modified due to existence of gravity s rainbow and dilaton field .
it is natural to question whether the usual relation between cosmological constant and thermodynamical pressure could be modified in presence of the dilaton field as well as rainbow functions . to investigate such modification
, we use the right hand side of the eq .
( [ dilaton equation(i ) ] ) .
it is a matter of calculation to show that ( after removing parts related to electromagnetic field ) @xmath109 obtained relation indicates that although both dilaton field and rainbow functions modified thermodynamical quantities , only the dilatonic part has direct effect on the relation between cosmological constant and pressure .
therefore , we use following analogy for studying the critical behavior of the system @xmath110 the conjugating quantity related to pressure is obtained through the use of enthalpy @xmath111 since consideration of the cosmological constant extends our thermodynamical phase space , the mass term plays the role of enthalpy .
therefore , by using eqs .
( [ mass ] ) , ( [ p ] ) and ( [ v ] ) , one can find modified volume of these dilatonic black holes as @xmath112 clearly , the volume of these black holes is a function of both rainbow functions and dilaton parameter . in other words ,
contrary to some specific modified gravities , in this gravity , the volume of the black hole is affected by the presence of rainbow and dilaton gravities . here , in order to have a positive and non - zero volume , we find a restriction @xmath113 . since we are not interested in negative values of the @xmath94 , we restrict ourselves to @xmath114 .
it should be pointed out that due to the relation between volume of the black hole and horizon radius , one is able to introduce specific volume for these black holes which enables us to use horizon radius instead of volume in following calculations .
so , the pressure is given by @xmath115 } { 8{r}_{+}^{4}{\pi } \mathcal{k}_{-1,1}\mathcal{k}% _ { 3,1}}\left ( { \frac{b}{r_{+}}}\right ) ^{-2\gamma } .
\label{pressure}\ ] ] in order to find a relation for calculating critical volume , hence critical horizon radius , we use the concept of inflection point . in this method ,
one uses @xmath116to find critical horizon radius which in case of this thermodynamical system is @xmath117which will lead to following critical temperature and pressure @xmath118 it is worthwhile to mention that the restriction that was observed was originated only from dilatonic part of the solutions . in other words , we have no restriction on values that charge and rainbow functions can acquire and our system is only thermodynamically restricted by dilaton parameter . using obtained critical values , one can find the following critical ratio @xmath119which shows that this critical ratio was modified due to the presence of dilaton field as well as rainbow functions .
it is worthwhile to mention that critical horizon radius depends only on one of the rainbow functions whereas the other critical values and also the ratio @xmath120 are functions of both of them . in addition , it is notable that in the absence of dilaton field ( @xmath121 ) and low energy limit ( @xmath122 ) , eq .
( [ ratio ] ) reduces to the usual universal ratio in four dimensional einstein gravity @xcite .
next , using the renewed role of the total mass of the black holes , we have gibbs free energy as @xmath123which by using eqs .
( [ temp ] ) , ( [ entropy ] ) , ( [ mass ] ) and ( [ p ] ) will be @xmath124 in order to see whether obtained critical values represent a second order phase transition , we study phase diagrams ( @xmath125 , @xmath126 and @xmath127 diagrams ) in figs .
[ fig4]-[fig6 ] .
@xmath128 @xmath129 @xmath130 it is evident that for specific values of different parameters , a second order phase transition is observed for obtained critical values ( see figs .
[ fig4 ] and [ fig5 ] ) . the critical pressure ( left panels of figs . fig4 and [ fig5 ] ) , temperature and subcritical isobars ( middle panels of figs .
[ fig4 ] and [ fig5 ] ) , energy of different phases and size of swallow - tails ( right panels of figs .
[ fig4 ] and [ fig5 ] ) are functions of gravity s rainbow and dilaton parameter .
the effects of rainbow functions and dilaton parameter on critical values are different from each other ( compare figs .
[ fig4 ] with [ fig5 ] ) .
interestingly , for a set of values , it is possible to obtain positive critical pressure and horizon radius whereas the temperature is negative .
the plotted diagrams for these cases show a normal critical behavior in @xmath131 diagram ( left panel of fig .
[ fig6 ] ) whereas in @xmath127 diagram an abnormal behavior is observed ( right panel of fig .
[ fig6 ] ) . in case of @xmath132 diagram ,
also normal critical behavior is observed except that this behavior is located in negative temperature .
in this section , we will obtain critical points through a method which was developed in ref .
@xcite . in this method ,
the denominator of the heat capacity is employed to obtain an explicit relation for thermodynamical pressure .
obtained relation may yield a maximum(s ) for pressure which is(are ) critical pressure(s ) in which a second order phase transition takes place .
this critical pressure is exactly the same as that of obtained through the use of phase diagrams . using eqs .
( [ heat ] ) and ( [ p ] ) and solving the denominator with respect to thermodynamical pressure will lead to following explicit relation for pressure @xmath134 g^{2}(\varepsilon ) } { 8\pi r_{+}^{4}\mathcal{k}_{3,1}% \mathcal{k}_{1,1}}\left ( \frac{b}{r_{+}}\right ) ^{-2\gamma } .
\label{pr}\ ] ] it is evident that this relation is different from previously obtained relation for pressure ( eq . [ pressure ] ) .
now , by using values that are employed for plotting phase diagrams ( figs . [ fig4]-[fig6 ] ) , we plot following diagrams ( fig .
[ fig7 ] ) .
a simple comparison shows that the maximums of the plotted diagrams are exactly where corresponding critical pressure and horizon radius are located in phase diagrams .
this shows that these two approaches yield consisting picture regarding the critical behavior of these black holes . on the other hand ,
plotted diagram which corresponds to one with abnormal behavior ( middle panel of fig .
[ fig7 ] ) also represents the characteristic behavior of the phase transition point .
therefore , in case of these black holes , a phase transition occurs in the mentioned critical point .
in this paper , we studied @xmath17-dimensional charged dilatonic black holes in gravity s rainbow and their thermal stability conditions .
we obtained thermodynamical quantities such as temperature , electric charge , entropy and total mass of the black holes .
these quantities were modified in gravity s rainbow and became energy dependent .
next , we conducted a study regarding physical / nonphysical black holes ( positivity / negativity of temperature ) and thermal stability of the solutions .
it was pointed out that dominant factor in studying these properties is the dilaton parameter . in other words , these properties were highly sensitive to variation of @xmath94 . due to different factors of dilaton parameter ,
different types of behavior were observed for temperature which put restrictions on the solutions being physical . observed behaviors for the temperature were : a ) two roots with maximum , b ) an increasing ( a decreasing ) function of horizon radius with one root c ) a negative definite function with one maximum .
the analyzed behaviors were , increasing function of horizon radius with one root , increasing and decreasing function of horizon radius with two roots , decreasing function of the @xmath101 with one root located and being negative with one maximum located at negative temperature .
as for the stability and phase transition , we found depending on the behavior of the temperature , heat capacity could have phase transition and stable state for larger values of horizon radius . in case of two roots for temperature ,
interestingly , a phase transition of larger / smaller black hole was observed .
finally , as for temperature being decreasing function of @xmath135 , for physical solutions ( positive temperature ) , unstable solutions were observed . in other words , in this case , physical solutions are unstable . next
, geometrical approach was employed to study the bound and phase transition points of these black holes .
it was demonstrated that the ricci scalars of the phase spaces of weinhold , ruppeiner and quevedo metrics , have divergencies which do not match with mentioned points while the singular points of curvature scalar of the hpem coincide with roots and divergence points of the heat capacity .
we also , studied the critical behavior of these black holes in extended phase space .
it was shown that the usual relation between cosmological constant and thermodynamical pressure was modified due to existence of dilaton gravity whereas such modification was not seen for gravity s rainbow .
on the contrary , it was shown that volume of the black holes depends on both of these modifications .
then , we showed that the critical values are related to the rainbow functions as well as the dilaton parameter . in order to have positive critical temperature , we found restrictions which were purely dilatonic dependant . therefore , one is not free to choose any value for dilaton parameter . in studying phase diagrams ,
two different behaviors were observed for different diagrams , especially in @xmath127 diagrams .
interestingly , although we observed an anomaly in these diagrams , other corresponding phase diagrams presented usual thermodynamical behavior around critical points . in other words , the observed abnormal behaviors in phase diagrams present the existence of a second order phase transition for these black holes .
next , we used a new method which was introduced in ref .
@xcite , for studying the critical behavior of the system .
this method is based on obtaining an explicit form for thermodynamical pressure from denominator of the heat capacity .
it was seen that the maximum maximum of this relation is located at the critical pressure and horizon radius in which second order phase transition takes place .
it was shown that in this method , for irregular behavior which was observed in phase diagrams , also a second order phase transition occurs .
this indicates that these points are phase transition point despite their abnormal behavior . finally , it is worthwhile to think about the physical interpretation of abnormal behavior which was seen in this paper .
it is notable that , one can generalize obtained linear solutions in this paper to nonlinear case of electrodynamics and investigate the effects of nonlinearity @xcite .
in addition , one may investigate the extended phase space and thermodynamic criticality in higher order lovelock - maxwell gravity s rainbow as well as lovelock - nonlinear electrodynamics @xcite .
these subjects are under examination . | in this paper , we present charged dilatonic black holes in gravity s rainbow .
we study geometric and thermodynamic properties of black hole solutions .
we also investigate the effects of rainbow functions on different thermodynamic quantities for these charged black holes in dilatonic gravity s rainbow .
then , we demonstrate that first law of thermodynamics is valid for these solutions . after that
, we investigate thermal stability of the solutions using canonical ensemble and analyze the effects of different rainbow functions on thermal stability . in addition , we present some arguments regarding the bound and phase transition points in context of geometrical thermodynamics .
we also study the phase transition in extended phase space in which cosmological constant is treated as the thermodynamic pressure .
finally , we use another approach to calculate and demonstrate that obtained critical points in extended phase space are representing a second order phase transition for these black holes . |
one requirement for measuring the hubble constant using sne ia as standard candles is a sample of well - measured distant supernovae .
they should be distant enough that their measured redshifts are dominated by the hubble flow , but not so distant that the still - uncertain dynamics associated with deceleration due to the mass density of the universe and possible acceleration due to conjectured repulsive forces are important .
a broad range of redshift @xmath7 from 0.01 to 0.2 is suitable , with @xmath7 about 0.05 being optimum . for this purpose
the carefully measured and uniformly analyzed caln - tololo collection of 29 sne ia ( hamuy et al .
1996 ) covering the range @xmath8 is very well - suited .
the collection has a spread in relative luminosity covering more than 1.5 magnitudes , but it has been shown that a two - parameter luminosity correction using the color and decline rate of each supernova is both necessary and sufficient to standardize them to a common luminosity ( tripp 1998 ) . a second requirement for measuring @xmath0 is to establish the absolute luminosity of sne ia that have been standardized in the above way . at the present time
there are seven nearby ( typically an order of magnitude closer ) galaxies that have hosted sne ia whose galaxy distances have been determined using cepheid variables measured by the hubble space telescope .
this includes the recently discovered sn 1998bu in the leo i group galaxy ngc 3368 whose cepheid distance had already been measured .
the cepheid - determined distances each have a typical accuracy of about 5% , plus an additional overall uncertainty of about 7% associated with the absolute distance scale for cepheid variables .
these seven galaxies have hosted eight recorded sne ia .
the oldest , sn 1895b , was observed photographically in only one color so can not be used here .
three of the others preceded the use of ccds in astronomy and , despite being well - measured by the techniques of the day , are not of the quality that can now be achieved .
thus we also consider an expanded list containing three more recent sne ia which , although not found in cepheid - calibrated galaxies , are very near on the sky to such a galaxy or are generally considered to be within groups of galaxies that have at least one member that has been so calibrated .
results are presented both with and without these additional supernovae .
an extensive bibliograpy on the use of type ia supernovae for measuring the hubble constant can be found in the recent review by branch ( 1998 ) .
we treat the 29 distant sne of hamuy et al .
( 1996 ) following the method of tripp ( 1998 ) , except that we allow the absolute magnitude to vary in a simultaneous fit with the cepheid - calibrated supernovae .
we calculate the hubble constant @xmath0 for each distant supernova using @xmath9 here @xmath10 and @xmath11 are the apparent and absolute blue magnitudes at maximum light , while @xmath12 is the deceleration parameter and @xmath7 is the measured red shift . for the 29 distant sne , @xmath10 and @xmath7
are tabulated by hamuy et al .
these supernovae are not so remote @xmath13 that the still uncertain value of @xmath12 leads to a significant uncertainty . conforming to the current evidence of riess et al .
( 1998 ) and perlmutter et al .
( 1998 ) for a negative @xmath12 , we fix it at -0.45 ( see section 4).as shown previously ( tripp 1998 ) , in order to fit the data of hamuy et al .
( 1996 ) within their quoted errors , the above absolute magnitude @xmath11 for each supernova must be adjusted according to both its rate of decline @xmath14 and its color @xmath15 . here
@xmath10 and @xmath16 are the maximum apparent blue and visual magnitudes .
empirically , linear dependences prove to be more than adequate for both corrections .
we therefore write for @xmath11 in equation ( 1 ) @xmath17 the parameter @xmath18 incorporates both intrinsic color differences between the sne and any reddening caused by dust in the host galaxies .
these two effects are often difficult to distinguish , particularly for the more distant supernovae .
the reddening parameter due to dust alone , usually denoted by @xmath19 , should be about 4 if the dust surrounding extragalactic supernovae is similar to dust in the milky way .
values @xmath20 due perhaps to a smaller average grain size .
since sne ia are not closely associated with star formation it may well be that cirrus values are relevant for them . in any event
, all sne with evidence of strong dust obscuration are removed from the sample for the case where we apply a color selection to the data . ] the weighted average value @xmath21 in equation ( 2 ) is obtained from the cepheid - calibrated sne in the following manner .
we use a similar dependence to standardize each of them to the value it would have if @xmath22 and @xmath23 for that supernova .
thus the corrected absolute magnitude becomes @xmath24 a least - squares fit of the cepheid - calibrated sne gives the weighted average @xmath21 appearing in equation ( 2 ) , along with the @xmath3 for the fit .
both are functions of @xmath25 and @xmath18 .
the value 1.05 in equations ( 2 ) and ( 3 ) is the average @xmath2 for 13 sne ia compiled by branch et al .
( 1996 ) . chosen for convenience ,
the choice is arbitrary and has no effect on @xmath0 , although it will directly affect @xmath21 .
the same remarks apply to the arbitrary choice of a @xmath26 subtraction in the color term . for each caln - tololo distant supernova , we use equation .
( 1 ) to evaluate @xmath0 using the value of @xmath11 from equation ( 2 ) with @xmath25 and @xmath18 as parameters .
the uncertainty in @xmath0 for each supernova is obtained by combining in quadrature the quoted errors @xmath27 , @xmath28 , and @xmath29 with an uncertainty in luminosity distance due to possible peculiar motion @xmath30 km / s of the host galaxy with respect to the hubble flow . neglecting correlations between errors ( since they are not reported )
, we have : @xmath31 + \left [ \left ( { 1 \over z } + { { 1- q_0 } \over 2 } \right ) \delta z \right ] ^2 } \ ] ] a weighted average of the 29 values of @xmath32 then gives a least - squares value for @xmath0 along with its uncertainty .
the @xmath33 for this fit of @xmath0 is added to the @xmath33 for the cepheid- calibrated sne fit of @xmath21 , both with @xmath25 and @xmath18 as parameters .
these are then varied to minimize the overall @xmath33 . when the parameters to be varied enter into the evaluation of the uncertainty @xmath6 and thus into the @xmath33 of the fit , as they do here , there is a bias favoring larger values of @xmath25 and @xmath18 leading to a larger @xmath6 and thus to a lower @xmath33 .
we try to eliminate this by temporarily fixing @xmath25 and @xmath18 for the evaluation of @xmath6 in equation ( 4 ) at the first solution values , then reminimizing and again fixing them at the new values .
after a few such iterations , stable values of @xmath25 and @xmath18 are found which are also @xmath3 minima . in the end , this results in a small decrease in @xmath25 , a substantial decrease in @xmath18 ( by about 0.6 unit ) , and a consequent increase in @xmath0 by about 1.3 units .
in table 1 , we list the seven sne from cepheid - calibrated galaxies and the three from cepheid - calibrated galaxy groups , along with the measured values used in our fits ; these are the distance modulus @xmath34 , the measured apparent magnitude @xmath10 and the resulting absolute magnitude @xmath35 appearing in equation ( 3 ) , the decline in @xmath10 magnitude , @xmath2 , during the first 15 days after maximum , and the color , taken to be @xmath36 .
where appropriate , the distance modulus incorporates the hst long - exposure correction of 0.05 as well as an estimate of the effect of host - galaxy absorption on the cepheid distance determination . conforming to the procedure used by hamuy et al .
( 1996 ) for their more distant sne , both @xmath10 and @xmath1 are corrected for galactic absorption using the estimates of burstein & heiles ( 1984 ) , but no correction is made for absorption within the host galaxy .
our method accomodates this host - galaxy absorption as well as any intrinsic reddening with the parameter @xmath18 used in the fits .
thus the generally uncertain dust absorption is effectively corrected for by this procedure .
for the 29 caln - tololo sne , we use values of red - shift , @xmath10 , @xmath37 , and @xmath2 given in table 1 of hamuy et al.(1996 ) .
table 2 presents the results of our joint fitting of the cepheid - calibrated ( cc ) sne + caln - tololo ( ct ) sne for two data selections . shown
are the number of fitted supernovae in each category , the best - fit values of @xmath0 , @xmath21 , @xmath25 , and @xmath18 , and the individual confidence levels for the best joint fit of the two subsets of data . in the first row we limit the sample to the six directly calibrated cc sne that also satisfy the color selection @xmath38 ( vaughan et al . , 1998 ) and to the 26 color - selected ct sne .
in the second row , the full nearby sample of 10 cc sne are jointly fitted with all 29 of the ct sne . both yield , fortuitously , identical values for @xmath39 .
it can be seen from table 2 that the ct data alone fit extremely well in both cases even though the @xmath25 and @xmath18 parameters are optimized to fit the combined ( ct and cc ) data .
the confidence levels for the ct data are in all cases considerably higher than the most likely value of 0.5 .
this reflects the presumed overestimate of the errors in the ct data set described in tripp ( 1998 ) where these data alone lead for the 29 ( 26 ) sne to an unrealistically high confidence level of 0.98 ( 0.97 ) for @xmath40 and @xmath41 . in both cases the joint confidence levels are acceptable due in large measure to these overly good fit of the ct data .
for the case of six cc sne , good fits are also realized for the cc subsample .
but this is not the case for the 10 cc sne where their confidence level falls below 1% .
this is due to a conflict between the two most recent cc sne : 1991 t and 1998bu .
they are both somewhat reddened slow decliners , with the first being superluminous and the second being subluminous according to this analysis , so that no @xmath25 and @xmath18 corrections can adequately reconcile the two and at the same time fit the ct data . and @xmath42 . alternatively
, if one or the other of the two conflicting sne is eliminated , then ( cc+ct ) fits can be found that are also good cc fits .
thus , making the color cut , thereby discarding 98bu and retaining 91 t , yields @xmath43 with a cc confidence level of 0.36 , while selecting just the directly calibrated sne and making no color cut discards 91 t and retains 98bu , yielding @xmath44 with a cc confidence level of 0.30 .
it is to be expected from the extreme nature of these two sne , lying , as they do , far out on either side of the @xmath33 distribution , that when one or the other is eliminated it will substantially impact @xmath0 .
it has been suggested ( fisher et al .
1999 ) that sn 1991 t forms a separate class of relatively rare superluminous objects not seen among the ct sample .
it is best explained as the result of a merger of two white dwarfs leading to a super - chandrasekhar explosion .
sn 1998bu is a normal supernova , but with considerable dust obscuration .
making the conventional color cut @xmath45 eliminates it .
since there is no evidence coming from interstellar absorption lines for strong dust obscuration among the 29 ct sne , whereas all three of the 10 cc sne falling outside this cut show clear evidence for this , in order to minimize bias between the samples the safest procedure is to apply the color cut to both groups as we do in the cc6/ct26 fit . ]
we list in table 3 the values of @xmath46 for each of the cepheid - calibrated sne along with their residuals , i.e. , @xmath47 \slash \delta m_b^0 $ ] , for both fits of table 2 . in figure 1 , we show the corrected values ( filled circles ) and the uncorrected values @xmath11 ( open circles ) for the full sample ( cc10/ct29 ) .
these are displayed as a function of @xmath2 in figure 1a and as a function of b - v in figure 1b .
the strong dependences of the uncorrected open circles on @xmath2 and @xmath1 are mostly removed by the best - fit parameters @xmath48 and @xmath49 , as seen in the corrected full circles .
however , as noted above , this fit to a common value of @xmath50 has an unsatisfactory @xmath33 with cl = 0.009 due to the conflicting demands of sn 1991 t and sn 1998bu .
> from the confidence levels in table 2 , it is evident that the combined > data from the caln - tololo and the cepheid - calibrated supernovae can be simultaneously fit in a satisfactory manner . however , the good confidence level , primarily the result of exceedingly good fits of the ct sne by themselves , disguises a significant difference in the data sets .
this is apparent in figure 2 , which shows plots of @xmath37 vs. @xmath14 for the 29 ct sne ( figure 2a ) and the 10 cc sne ( filled circles in figure 2b ) .
the striking difference between the two groups is the complete absence of unreddened @xmath51 events beyond @xmath52 among the 10 cc sne and their abundance among the 29 ct sne .
since @xmath37 and @xmath2 are both independent of distance , we may include in figure 2b six additional nearby sne in the virgo and fornax clusters ( open circles ) whose distances are still uncertain .
this inclusion makes the two samples more compatible , suggesting that the void in the nearby sample of 10 may be , in part , due to a statistical fluctuation .
thus , while the 29 ct sne in figure 2a display very little correlation between @xmath53 and @xmath2 , within the limited statistics of the 10 cc sne there is a strong correlation , but one which is much reduced by the inclusion of the six distance - uncalibrated sne .
a possible bias affecting any measurement of the hubble constant , both in our two parameter fit and previous one and zero parameter fits , may arise because the nearby and distant samples come from galaxies of somewhat different morphologies .
since a galaxy must contain a population of young stars in order to produce cepheid variables , the cc sne are generally found in spiral galaxies .
( an exception to this is ngc5253 , the parent of sne 1972e and 1895b , which is often classified as an e / so peculiar galaxy .
a putative explanation for its star forming regions is that they are the result of a galactic merger . ) .
thus , apart from sn 1972e , our cc sample comes from spiral galaxies , whereas the ct sample is about equally divided between spirals and e , e / so , and so galaxies . eliminating from ct26 all but
the 13 spirals gives for cc6/ct13 , after reminimizing @xmath33 for this smaller sample , a value of @xmath54 for @xmath0 compared to @xmath55 for the full sample ct6/ct26 .
since the former is presumably freer of potential bias we take @xmath4 as our best estimate of the hubble constant and the associated value of -19.46 for @xmath56 .
we now discuss the uncertainty in @xmath0 arising from a variety of sources . the 1-@xmath57 statistical error in @xmath6 is found to be @xmath58 . in addition , there is uncertainty in @xmath25 and @xmath18 which we display in figure 3 where @xmath0 , as a function of @xmath25 and @xmath18 , is superposed on the 1- , 2- , 3- , and 4-@xmath57 contours of the cc6/ct26 fit .
> from this we obtain the 1-@xmath57 error from the @xmath25 and @xmath18 uncertainty > of @xmath59 . throughout this analysis
we have fixed the insensitive deceleration parameter at @xmath60 , found by making a three parameter @xmath61 fit of the 26 ct sne jointly with the 9 cosmological sne of riess et al .
( 1998 ) for which @xmath1 colors were available .
if we assign an uncertainty of @xmath62 to this value of @xmath12 , this leads to an uncertainty @xmath63 . combining these three sources in quadrature yields @xmath64 . to these uncertainties arising from the sne ia analysis
must be added a larger uncertainty coming from the calibration of the cepheid variables and their possible metallicity dependence .
the madore and freedman ( 1991 ) distance scale in current use fixes the distance modulus to the large magellanic cloud ( lmc ) to be 18.50 mag ( 50.1 kpc ) and scales more distant cepheid - based galaxy distances to this value . because of the importance of this number for the determination of the hubble constant , much attention has been devoted to methods for obtaining a more accurate value .
a recent review from a post - hipparcos perspective ( walker 1998 ) recommends a mean modulus of @xmath65 , found using a variety of distance indicators . however , even more recent analyses of an eclipsing binary in the lmc ( guinan et al .
1998 ) , found during the ogle microlensing search , yield a value as low as @xmath66 mag ( udalski et al .
1998 ) . for this and other reasons evident in the walker(1998 ) summary
, we use without change the distance moduli of cepheid - calibrated sne ia found by the various observers which are all based on the 18.50 mag lmc value .
we assign to this value a 1-@xmath57 error of @xmath67 mag , with the usual statistical view that a true distance modulus to the lmc differing by twice that value would be surprising and one differing by 3-@xmath57 would be very unlikely . using @xmath68 obtained from equation ( 1 )
, this uncertainty leads to a 1-@xmath57 error of @xmath69 .
for some years there has been concern that metallicity differences between the lmc and the cepheid- calibrated galaxies can alter the derived distances to these galaxies . as the most easily measured proxy for metallicity , kochanek ( 1997 )
has collected the logarithmic abundance ratio @xmath70 $ ] for a number of the cepheid - calibrated galaxies relative to that of the lmc . for galaxies hosting sne ia ,
these cover a range of @xmath70 $ ] from -0.35 to + 0.69 .
we have used these to alter the measured distance moduli @xmath34 by means of the expression @xmath71 $ ] , where the parameter @xmath72 is varied to obtain a best fit . for cc6/ct26 , the data set for which five of the cepheid - calibrated galaxies have measured values of @xmath70 $ ] , we find @xmath73 , in agreement with other recent findings which fall between 0.14 and 0.31 ( kochanek ( 1997 ) , kennicutt et al .
( 1998 ) , nevalainen and roos ( 1998 ) ) , but showing that this data set has little sensitivity to @xmath72 . if we fix @xmath74 and follow the previous procedure of varying @xmath25 and @xmath18 to find a @xmath33 minimum , then @xmath0 increases by 0.8 km s@xmath5 mpc@xmath5 . since the question of a metallicity dependence is still in dispute ( see beaulieu et al .
1997 , saha et al .
1997 , and saio and gautschy 1998 for recent differing views ) , we retain the value of @xmath0 without correction but assign an uncertainty in @xmath72 of @xmath75 resulting in @xmath76 due to this source . combining all these errors in quadrature yields @xmath77 for the overall uncertainty . despite significant differences with other analyses
, our value of @xmath78 falls squarely in the middle of the range spanned by recent determinations using sne ia .
these values of @xmath0 , all using the distance modulus of 18.50 to the lmc , range between 55 and 69 km s@xmath5 mpc@xmath5 .
one difference between our analysis and others is that they all seem to use the nearby approximation to equation ( 1 ) , which is tantamount to setting @xmath79 . if instead they were to use equation ( 1 ) with a negative @xmath12 , as is now apparently required by the cosmological data , then their values would each increase by about 1.7 units in cases when the ct sne are used for the hubble flow sample .
the other major difference is that only our analysis uses two independent parameters ( @xmath25 and @xmath18 ) to standardize the luminosity of the supernovae ; this is required in order to obtain a fit to the ct data with an acceptable @xmath33 ( tripp 1998 ) . among the recent analyses using a collection of nearby calibrating supernovae , saha et al .
( 1997 ) , employing seven cepheid - calibrated sne ( 1895b , 1937c , 1960f , 1972e , 1981b , 1989b , 1990n ) and no standardizing parameters , find @xmath80 .
suntzeff et al .
( 1998 ) , using five cc sne ( 1937c , 1972e , 1981b , 1990n , 1998bu ) along with the hubble flow ct sne and with a one parameter correction for @xmath2 , obtain @xmath81 .
the difference in @xmath0 found in these two analyses lies primarily in the introduction of the @xmath25 parameter .
our analysis , involving both @xmath25 and @xmath18 , reduces @xmath25 and leads to an intermediate result .
this is immediately apparent from figure 3 .
as can be discerned from the figure , the saha et al.(1997 ) @xmath82 fit should yield about 60 for @xmath0 .
this is equivalent to their 58 after correcting for the above effect of @xmath12 .
likewise the suntzeff et al .
( 1998 ) fit with only @xmath25 as a free parameter should , after the @xmath12 correction , yield about 66 , as can be seen from the figure for @xmath83 .
apparently , different selections of cepheid - calibrated sne have only a minor effect on @xmath0 .
thus future augmentations to the present small number of cepheid - calibrated sne will probably have little impact on @xmath0 .
assuming that there are no significant observational selection differences between the ct and cc samples , only a revision of the cepheid distance scale will be capable of altering @xmath0 by more than a few units .
this work was supported in part by the director , office of energy research , office of high energy and nuclear physics , division of high energy physics of the u.s .
department of energy under contract ac03- 76sf00098 and ( for d.b . ) an nsf grant ast 9417102 .
cllllll 1937c&ic4182&28.36(9 ) ( 1)&8.80(9 ) ( 6)&-19.56(13)&0.87(10 ) ( 6)&-0.02(6 ) ( 6)1960f&n4496a & 31.13(10 ) ( 1)&11.60(10 ) ( 1)&-19.53(14)&1.06(8 ) ( 11)&0.09(8 ) ( 1)1972e&n5253&27.94(8 ) ( 1,2)&8.30(14 ) ( 1)&-19.64(16)&0.87(10 ) ( 6)&-0.05(8 ) ( 6)1974g&n4414&31.41(23 ) ( 3)&12.48(5 ) ( 7)&-18.93(24)&1.11(6 ) ( 7)&0.18(5 ) ( 7)1981b&n4536 & 31.10(13 ) ( 4)&12.03(3 ) ( 6)&-19.07(13)&1.10(7 ) ( 2)&0.10(4 ) ( 6)1986g&n5128 & 28.13(46 ) ( 1)&11.96(7 ) ( 8)&-16.17(46)&1.73(7 ) ( 12)&0.83(10 ) ( 8)1989b&n3627 & 30.28(26 ) ( 1)&12.29(6 ) ( 9)&-17.99(27)&1.31(7 ) ( 12)&0.34(5 ) ( 9)1990n&n4639 & 32.03(22 ) ( 1)&12.71(3 ) ( 10)&-19.32(22)&1.07(5 ) ( 2)&0.05(4 ) ( 10)1991t&n4527 & 31.11(10 ) ( 1 ) & 11.70(2 ) ( 10)&-19.41(10)&0.94(5 ) ( 10)&0.19(3 ) ( 10)1998bu&n3368&30.37(16 ) ( 5,2)&12.10(3 ) ( 2)&-18.27(16)&1.01(5 ) ( 2)&0.30(4 ) ( 2 ) lrrrrr 1937c&-19.44 ( 18)&-0.10&&-19.41 ( 20)&0.151960f&-19.70 ( 21)&-1.36&&-19.75 ( 24)&-1.271972e&-19.46 ( 23)&-0.18&&-19.42 ( 26)&0.901974g&-19.30 ( 26)&0.47&&-19.39 ( 27)&0.181981b&-19.28 ( 16)&0.87&&-19.34 ( 17)&0.631986g*&&&&-18.54 ( 52)&1.731989b*&&&&-18.95 ( 30)&1.671990n&-19.39 ( 24)&0.11&&-19.45 ( 24)&-0.031991t*&&&&-19.80 ( 13)&-2.831998bu&&&&-18.97 ( 19)&2.49 | in this paper , we make a comprehensive determination of the hubble constant @xmath0 by using two parameters the @xmath1 color and the rate of decline @xmath2to simultaneously standardize the luminosities of all nearby cepheid - calibrated type ia supernovae ( sne ia ) and those of a larger , more distant sample of 29 sne ia . each group is treated in as similar a manner as possible in order to avoid systematic effects .
a simultaneous @xmath3 minimization yields a standardized absolute luminosity of the cepheid - calibrated supernovae as well as the hubble constant obtained from the more distant sample .
we find @xmath4 km s@xmath5 mpc@xmath5 and a standardized absolute magnitude of -19.46 . the sensitivity of @xmath0 to a metallicity dependence of the cepheid - determined distances is investigated . the total uncertainty @xmath6 , dominated by uncertainties in the primary cepheid distance indicator , is estimated to be 5 km s@xmath5 mpc@xmath5 . |
let us consider a system consisting of a waveguide connecting 2 ports as illustrated in fig .
a resonator supporting a mode with frequency @xmath21 is coupled to this waveguide at @xmath19 independent coupling points , such that each coupler is spatially separated on the waveguide by an identical length @xmath20 . for simplicity , we assume that these couplers are all connected to the resonator at anti - nodes of equal phase for the resonant mode at @xmath21 . using the framework of temporal coupled mode theory @xcite , we will first derive an effective coupling coefficient for this system , and we will show later that the effective coupling is sensitive to the direction of wave propagation when spatiotemporal modulation is applied to the individual coupling coefficients .
the use of coupled mode theory generalizes the analysis allowing for implementation with electromagnetic , optical , or acoustic waves without changing the underlying systemic principles .
we can describe the complex - valued excitation amplitude @xmath16 of the resonant standing wave mode ( traveling modes are discussed in the supplementary information section s1 ) through equation ( [ eqin ] ) , @xmath22 where @xmath21 and @xmath23 describe the frequency and decay rate of the resonance , and @xmath24 is the analysis frequency of the waves traveling into the ports .
waves incident on port 1 , described by @xmath25 , are coupled to the resonator with effective coupling @xmath26 , which is the sum of contributions from all @xmath19 couplers @xmath27 here @xmath28 describes the coupling strength of the @xmath29 coupler , numbered such that @xmath30 is nearest to port 1 and @xmath31 is nearest to port 2 .
waves incident on port 2 ( @xmath32 ) are likewise coupled to the resonator through the effective coupling strength @xmath33 we here assume that losses in the waveguide can be neglected and propagation is solely described by the waveguide phase constant @xmath34 . in general , the spatial separation of the multiple couplers causes the effective couplings @xmath26 and @xmath35 to be sensitive to the direction of wave propagation . although this directional sensitivity does not in general lead to nonreciprocity , it is a crucial part of the nonreciprocal coupling theory that we develop in this paper . as a practical consideration
we note that the waveguide propagation constant @xmath34 will be a function of frequency due to the waveguide dispersion , and will result in some deviation from this theory .
for simplicity we will ignore the effects of waveguide dispersion .
for a 2-port system to be nonreciprocal it is sufficient that the transmission coefficients @xmath36 and @xmath37 are not equal .
these steady - state transmission coefficients are defined as @xmath38 ( with @xmath39 ) and @xmath40 ( with @xmath41 ) where @xmath42 and @xmath43 ( the outgoing wave amplitudes at port 2 and port 1 ) are defined by equation ( [ eqout ] ) @xmath44 for the simple case where each coupler has the same coupling strength @xmath45 , we can find a compact form of the transmission coefficients @xmath46 ~,\ ] ] where we have introduced a dimensionless periodic function @xmath47 that describes the squared efficiency of the effective coupling .
@xmath47 is defined as @xmath48 and is described in detail in the supplementary information .
this function is symmetric about @xmath49 and periodic such that @xmath50 for integer @xmath51 , and concisely describes the effective excitation of the resonant mode resulting from interference due to the multiple couplers .
the zeros of @xmath47 arise when the interference is completely destructive ( i.e. the resonator and waveguide are phase mismatched ) and the resonant mode is not excited at all . note that the resonator decay rate @xmath23 can be defined as @xcite @xmath52 where @xmath53 @xmath54 denotes the time - average .
therefore @xmath23 can also be expressed in terms of @xmath55 , making the efficiency function extremely useful in simplifying the system equations . here
@xmath56 describes the intrinsic decay rate of the resonator .
let us now consider the case where the coupling strength at each individual coupling point is sinusoidally modulated with amplitude @xmath57 and frequency @xmath17 around a static value @xmath58 , where @xmath59 .
additionally , we apply a discretely spaced phase shift @xmath60 to the modulation term for the @xmath61 coupler such that @xmath62 .
this phase shift creates a discretized traveling modulation wave , or in other words a ` synthetic phonon ' having a wave - vector controlled by the relative phase @xmath63 .
we can apply this spatiotemporally varying coupling to equations ( [ eqin ] , [ eqout ] ) to find the steady - state transmission coefficients @xmath64 e^{-j \beta \ell ( n-1 ) } ~ \textrm { , and } \end{split}\ ] ] @xmath65 e^{-j \beta \ell ( n-1 ) } .
\end{split}\ ] ] equations ( [ eqf ] , [ eqb ] ) define transmission for incident and outgoing waves of the same frequency @xmath24 and do not explicitly describe outgoing waves at other frequencies .
recalling from equation ( [ eqn : gamma ] ) that @xmath23 is dependent on the squared magnitude of the effective coupling we note that resonator decay rate will increase when this type of modulation is applied , which is in fact required for energy conservation .
the above transfer functions reveal distinct transmission dips at the original resonance @xmath21 and at sidebands @xmath66 .
notably , in the sidebands the @xmath47 function is offset by the modulation phase difference @xmath63 , with the sign of this offset determined by the direction of wave propagation in the waveguide . in the general case we have @xmath67 implying that the sideband transmission coefficients @xmath36 and @xmath37 are unequal and the system is nonreciprocal .
this nonreciprocity condition is always satisfied if neither @xmath68 nor @xmath63 is an integer multiple of @xmath69 .
the induced nonreciprocity in this system is clearly of the canonical form @xmath70 , since a transformation of @xmath71 is equivalent to @xmath72 in the equation for @xmath28 . , [ eqb ] ) ) through the system of fig .
[ fig1]*b * at the upper sideband frequency @xmath73 , where @xmath17 is the frequency of the applied coupler modulation .
rows indicate the case of two ( @xmath74 ) and three ( @xmath75 ) coupling points , corresponding to the simplest possible implementations of nonreciprocal coupling .
for each figure , the x - axis represents the phase - separation ( @xmath68 ) between adjacent couplers , while the y - axis represents the relative phase setting ( @xmath63 ) of the applied modulation at each coupler . in a typical system @xmath76
will correspond to the physical design while @xmath63 is a dynamically adjustable parameter .
low transmission corresponds to stronger coupling , since the incoming waves are absorbed by the resonator .
as anticipated , the effective total coupling differs for forward ( @xmath36 ) and backward ( @xmath37 ) propagating waves .
the transmission coefficients are mirrored about @xmath77 , demonstrating the relationship @xmath78 .
large contrast can be achieved in a system of 3 coupling points designed with @xmath79 having a relative phase setting of @xmath80 .
, scaledwidth=75.0% ] figure ( [ fig2 ] ) shows theoretically calculated transmission at the upper sideband frequency @xmath73 for systems having two and three couplers ( @xmath81 ) , corresponding to the simplest possible implementations .
the transmission at the lower sideband can be found by simply reversing the sign of @xmath63 , as seen in equations ( [ eqf ] , [ eqb ] ) .
this figure is simplified such that @xmath34 does not vary with frequency , a useful approximation if the variation is small over the desired frequency range .
we find that the largest transmission dips are centered on the zeros of @xmath55 , i.e. the values of @xmath68 where the original resonant mode at @xmath21 is totally uncoupled from the waveguide .
the sidebands can still couple at these points with a coupling efficiency described by @xmath82 , but the overall decay rate of the mode ( @xmath23 ) is minimized because there is no decay at the center frequency .
the minimized decay rate means the quality - factor of the resonance is maximized and the dip is transmission as deep and narrow as possible . by extension of this idea , for a given sideband coupling rate the transmission dip in that sideband will increase in depth and decrease in width as coupling to the original resonance decreases .
figure [ fig2 ] illustrates that for the @xmath74 system the maximum transmission dip occurs at @xmath83 , a symmetry point of @xmath84 that unfortunately exhibits a reciprocal transmission response .
this symmetry problem is easily overcome in a @xmath85 system , since zeros of @xmath84 will then occur away from @xmath83 .
notably , for @xmath75 the transmission is maximally nonreciprocal for @xmath86 and a relative phase setting of @xmath80 .
we additionally note that more couplers will also increase the total effective coupling to the mode as evidenced by the @xmath87 term in equations ( [ eqf ] , [ eqb ] ) , implying greater transmission contrast if implemented .
we note that the derivation presented here describes a standing wave resonator , where incident waves from both ports are coupled to the same resonant mode .
as we discuss in the supplementary information , the method we propose for creating nonreciprocal transmission can also be applied to a traveling wave resonator having intrinsic directionality to its resonant modes .
in fact , since the effective coupling itself is direction sensitive , our approach is generally independent of the type of resonator used .
however , care must be taken when using a standing wave resonator since simultaneous incident waves from both ports can interfere in the resonator .
this problem is automatically resolved if waves in one direction are completely uncoupled from the resonance , which is also necessary for a maximum nonreciprocal response and lossless propagation in the uncoupled direction .
a common goal of nonreciprocal networks is to create a strong directional effect , such as complete isolation or a unidirectional phase - shift . in order to create these effects through nonreciprocal coupling
, it is useful if waves propagating in one direction are strongly coupled to the resonant mode while counter - propagating waves are completely uncoupled .
this special case occurs when @xmath88 and @xmath89 , and when applied to equations ( [ eqf ] , [ eqb ] ) we see that transmission in the uncoupled direction becomes lossless ( @xmath90 ) .
for example , this occurs for @xmath91 when @xmath92 ( see fig .
s2 in the supplementary information ) . when the additional condition that @xmath93 is met the system is critically coupled and behaves as an ideal isolator at the sideband frequency , with unity transmission magnitude in one direction and no transmission in the other .
note that is also applies if the sign of @xmath63 is flipped .
a network that imparts a unidirectional @xmath69 phase - shift , often called a gyrator , is an important nonreciprocal component that can be used to create circulators and more complex nonreciprocal circuits .
our system can behave as a gyrator if certain conditions are met .
specifically , waves from one direction must be uncoupled from the resonance while counter - propagating waves are strongly coupled at one of the sidebands ( @xmath88 and @xmath89 ) .
simultaneously , waves traveling in both directions must be uncoupled at the original resonance and the other sideband .
this can be achieved with a waveguide that is dispersive or has a bandgap . in this configuration
an ideal gyrator can be approached as @xmath94 .
the conditions leading to ideal isolation and gyration are thus fundamentally possible to achieve with careful system engineering , allowing our simple nonreciprocally coupled waveguide - resonator system to be used as a building block for more complex nonreciprocal circuits .
we implemented nonreciprocal waveguide - resonator systems having @xmath74 and @xmath75 couplers using microwave frequency microstrip circuits @xcite , shown in fig .
we also introduce a time - modulated coupling with controllable phase shift at each coupling point in order to generate the requisite ` synthetic phonon ' wave ( see methods ) . coupling points .
* c , d * photograph and schematic of a system with @xmath75 coupling points .
each microstrip ring resonator is coupled to the waveguide through multiple varactor diodes that are separated on the waveguide by phase @xmath76 , with the dc and time - modulated biased controlled using voltages @xmath95 . for the 3-coupler circuit , biasing to the central varactor is provided through the resonator .
, scaledwidth=75.0% ] circuit are presented at both sidebands using 100 mhz modulation applied to the couplers .
data for forward transmission ( @xmath36 , red ) and backward transmission ( @xmath37 , blue ) are presented for both @xmath96 ( dashed lines ) and @xmath97 ( solid lines ) . *
b , c * data present the line - center forward ( red , x ) and backward ( blue , o ) transmission coefficients at the lower ( upper ) sideband frequency @xmath98 ghz ( @xmath99 ghz ) for selected values of @xmath63 .
we observe nonreciprocal coupling to the resonance while using coupler phase settings of @xmath100 in both upper and lower sidebands of the resonance .
, scaledwidth=60.0% ] circuit are presented at both sidebands using 100 mhz modulation applied to the couplers .
data for forward transmission ( @xmath36 , red ) and backward transmission ( @xmath37 , blue ) are presented for both @xmath96 ( dashed lines ) and @xmath101 ( solid lines ) . *
b , c * data present the line - center forward ( red , x ) and backward ( blue , o ) transmission coefficients at the lower ( upper ) sideband frequency @xmath102 ghz ( @xmath103 ghz ) for selected values of @xmath63 .
again we observe nonreciprocal coupling when @xmath100 .
, scaledwidth=60.0% ] the fabricated resonator system with two couplers ( @xmath74 ) shown in fig . [ fig3]*a , b * has a loaded resonance frequency @xmath104 ghz .
figure [ fig4]*a * shows the measured lower and upper transmission sidebands in forward and reverse directions when @xmath105 modulation is applied to the couplers , with relative phase setting of @xmath96 ( dashed lines ) and @xmath97 ( solid lines ) . in figure
[ fig4]*b , c * we quantify the measured transmission at the lower sideband resonance ( frequency @xmath106 ) and upper sideband resonance ( @xmath73 ) respectively , as @xmath63 is swept from @xmath107 to @xmath69 .
as predicted , we observe reciprocal transmission at @xmath108 .
however , when @xmath63 is not equal to an integer multiple of @xmath69 the coupling in the sidebands diverges depending on wave directionality , resulting in appreciable nonreciprocal transmission . in the lower sideband
there is minimum forward coupling at @xmath97 and minimum reverse coupling at @xmath109 , confirming that the fabricated circuit has @xmath110 at the lower sideband @xmath111 . on the other hand
, we see that upper sideband coupling is minimized at @xmath112 ( positive sign corresponds to minimized forward coupling ) indicating that @xmath113 at @xmath114 .
the large difference in @xmath68 for these two frequencies can be attributed to the dispersion characteristics of the waveguide which are non - negligible in our experiment .
we additionally note that the upper sideband is much smaller and misshapen compared to the lower sideband , which is due to additional undesired resonant modes and reflections in our circuit implementation . despite these experimental difficulties ,
we observe strong nonreciprocal coupling and over 5 db of directional contrast in this initial experiment .
our test system with three coupling points ( @xmath75 ) has a loaded resonance frequency @xmath115 ghz and is presented in fig .
[ fig3]*c , d*. figure [ fig5]*a * shows the lower and upper sideband measurements for @xmath96 ( dashed lines ) and @xmath101 ( solid lines ) for an applied modulation frequency of @xmath116 . in fig .
[ fig5]*b , c * we present the measured transmission at the lower and upper sideband resonances as @xmath63 is swept from @xmath107 to @xmath69 . as predicted in fig .
[ fig2 ] , this three coupler case yields two distinct dips in transmission ( one large , one small ) as the phase setting @xmath63 is varied .
these two dips in transmission correspond to two local maximums in the coupling efficiency function @xmath117 . in the lower sideband
the large transmission dip occurs at @xmath118 ( minus sign corresponding to minimized forward transmission ) corresponding to @xmath119 .
figure [ fig5]*a * shows simultaneous coupling of both forward and backward waves to the resonance at the lower sideband .
this occurs since forward waves couple to the resonator due to the small local maximum of @xmath120 , while backward waves couple strongly at the large local maximum .
these two local maxima occur for the same modulation phase @xmath63 due to the waveguide dispersion .
the upper sideband exhibits similar behavior with one large local coupling maximum at @xmath121 and a smaller local coupling maximum at @xmath122 , corresponding to @xmath123 . again
, this large difference in @xmath68 between the sidebands can be attributed to non - idealities in the circuit implementation .
figure [ fig5]*c * shows that on the upper sideband we are able to achieve strong coupling for backward waves and essentially zero coupling for forward propagating waves , demonstrating over @xmath124 db of contrast with @xmath125 .
a key advantage of employing generalized nonreciprocal couplers is that the nonreciprocal transfer function is not limited to the response of simple resonator .
this possibility arises since the shape of the transfer functions @xmath36 and @xmath37 are determined by the denominator of equations ( [ eqf ] , [ eqb ] ) , while the nonreciprocity generated by the coupling , is contained within the numerator . to illustrate ,
let us consider an arbitrary shunt impedance that is coupled capacitively to a waveguide and generates an elaborate transfer function , such as that of a multi - band filter .
use of a nonreciprocal coupler instead of the linear coupler between the waveguide and this shunt impedance will cause the transfer function to appear only for reverse propagating waves at the sideband frequencies as illustrated in fig .
[ figpre]*a*. a specific example that can be generated using an impedance composed of only a few coupled resonators is a nonreciprocal band - reject filter , as illustrated in fig .
[ figpre]*b*. this capability has not been demonstrated in other approaches to creating nonreciprocal responses , as many of these approaches rely on exact resonance conditions @xcite and do not allow for arbitrary impedance networks .
that is nonreciprocally coupled to a waveguide can generate an elaborate nonreciprocal transfer function .
the capacitor with an arrow is used here as a symbol representing nonreciprocal coupling which only couples waves traveling in the direction of the arrow . *
b * a flatband nonreciprocal band - pass filter can be easily produced using three resonators and one nonreciprocal capacitive coupler .
the simulated @xmath37 curve is evaluated with @xmath126 , @xmath127 , @xmath128 , @xmath129 . here , @xmath130 is the decay rate associated with the @xmath29 resonator and @xmath131 is the coupling rate between the resonators . , scaledwidth=60.0% ] [ t ! ] circuit used to demonstrate nonreciprocal coupling to a split resonance .
the circuit is functionally identical to the previous experiment except for an additional resonator coupled to the first .
an applied voltage @xmath132 is used to tune the center - frequency of the second resonance .
* b * measured nonreciprocal transmission coefficients @xmath36 ( red ) and @xmath37 blue at the upper sideband ( @xmath133 ghz ) when the resonance frequencies of the coupled resonators are far apart , similar to previous experiments .
* c * measured isolation ratio and transmission coefficients ( numbered 1 - 3 ) at the upper sideband when the second resonance ( @xmath134 ) is tuned across the first ( @xmath135 ) .
we observe that backward propagating waves couple effectively to the split resonance in this sideband although both modes are at a much lower frequency , clearly demonstrating that the nonreciprocity is caused by the modulated coupling .
an increase in the directional contrast ( isolation ) bandwidth due to the broadening of the mode is also evident .
, title="fig : " ] to experimentally verify this concept of nonreciprocal coupling to an arbitrary resonant state , we produced the simple two - resonator circuit shown in fig .
[ fig6]*a*. this microwave circuit is a direct extension of the @xmath91 implementation ( fig .
[ fig3]*c , d * ) in which a second resonator ( @xmath134 ) has been coupled to the first resonator ( @xmath135 ) through a fixed capacitor .
the resonance frequency of this added resonator can be capacitively tuned using a voltage controlled varactor diode at the top of the ring .
the nonreciprocal coupler placed between the waveguide and resonator is modulated using @xmath136 with a phase setting of @xmath101 .
figure [ fig6]*b * presents the measured nonreciprocal coupling occurring between the first resonator and the waveguide at the upper sideband frequency @xmath137 ghz , when the frequencies of the two coupled resonances are far apart . the second resonance at @xmath134 is then tuned across the first resonance at @xmath135 , resulting in an avoided crossing and lifted degeneracy of the hybrid modes of the system .
the avoided crossing is easily observed in our measurements of the @xmath138 isolation contrast in fig .
[ fig6]*c * at the sideband frequency .
this simple example showcases that nonreciprocal transfer functions that are more complex than simple resonances @xcite can be obtained quite readily by this approach .
in this study we have presented a theoretical description and experimental validation of employing multiple spatiotemporally modulated couplers to create nonreciprocal coupling in a very simple 2-port system .
we have also discussed how such a system could theoretically achieve unidirectionally lossless nonreciprocal transmission , as well as how nonreciprocal coupling could be used to create arbitrary nonreciprocal transfer functions .
the methods presented are independent of the type of resonator used and may be applied more broadly to optical and mechanical systems .
presently , we were unable to observe lossless nonreciprocal transmission due to several non - ideal factors in the design .
specifically , in circuit implementation of a waveguide and ring resonator coupled with discrete capacitors there are several effects that are not accounted for in the temporal coupled - mode theory the most important of which are reflections from the capacitive couplers that can significantly distort the resonance shape and the slightly non - linear response of the varactor diodes . due to this effect
we observe scattering from the modulated couplers even far from the resonance , which effectively creates extra loss in the system .
this behavior could be overcome in circuit implementations by lowering the intrinsic losses in the resonator and using a smaller capacitance for coupling .
these modifications will decrease the distortion of the resonance shape as well as the non - linearity introduced by the varactor diodes .
many experiments on nonreciprocal systems created through spatiotemporal modulation @xcite can be seen through the lens of unidirectional inter - modal scattering or a momentum bias applied to the medium @xcite .
it is also possible to relate our work to this concept , with the momentum bias applied to the coupling coefficients instead of the waveguide or resonator components .
however , major features of our demonstrated approach are that it allows the modulated region to be sub - wavelength , and that high - quality conventional ( static ) waveguides and resonators can be employed without the complexities of time - modulation .
further , the addition of a lossy element ( the resonant mode ) enables true signal isolation functionality where waves incident in one direction are dissipated as heat .
most importantly , we have experimentally demonstrated that nonreciprocal coupling enables the synthesis of arbitrary nonreciprocal transfer functions using conventional components and enables the use of well - known impedance networks .
such design capability has not been previously achieved in other approaches since they are often limited to exact resonant responses .
nonreciprocity achieved through nonreciprocal coupling is thus a powerful technique that can offer greater reconfigurability and customization of the system dynamics and signal processing functions .
our microwave circuits are fabricated on rogers 6006 substrate with a 1 oz copper conductor , and consist of a microstrip waveguide coupled to a ring resonator by @xmath19 varactor diodes ( skyworks smv1275 ) that act as variable capacitors .
the coupling strength @xmath28 of each capacitive coupler is an approximately linear function of the applied voltage @xmath95 for small changes in voltage , allowing a modulation of @xmath28 that is proportional to a modulation of @xmath95 .
we first apply a dc bias of 3 volts to each varactor diode with a dc power supply ( agilent e3631a ) .
we modulated this bias using a sine - wave from a signal generator ( hp-8647b ) , which is set at a frequency of 100 mhz and a signal amplitude of 2 volts .
the signal is split ( minicircuits za3cs-400 - 3w - s ) into three lines that each connect to a variable phase shifter ( minicircuits jsphs-150 + ) so that the phase shift between each signal can be independently controlled .
after the splitting and taking into account the loss of the phase shifters and cables , we estimate about 1 volt ac bias at 100 mhz is applied to each varactor diode .
the dc bias and 100 mhz modulation are combined through a bias tee ( minicircuits zfbt-4r2gw - ft+ ) connected directly to the circuit . on each circuit ,
butterfly band - pass filters were incorporated along with a another low - frequency biasing tee ( johanson technology l-14c10n - v4 t 10 nh inductor and johanson technology r14s 12 pf capacitor ) as shown in fig .
[ fig3 ] to isolate the lower frequency bias modulation from the higher frequency resonant circuit . the applied 3 volt dc bias results in an average varactor capacitance of 1.87 pf , which is then modulated by the 1 volt ac signal from 1.58 pf to 2.32 pf .
this is a slightly nonlinear voltage - capacitance relationship , but the non - linearity is small enough to not drastically impact the results . for large modulation amplitude
we observed a shift in the phase @xmath63 at which the maximum and minimum coupling occur , due to interference between the applied biases and non - linearity in the phase shifters .
we measured the transmission parameters ( @xmath36 , @xmath37 ) of the circuit using a keysight e5063a vector network analyzer .
the network analyzer measurement was calibrated to the ends of the sma cables that connected to the surface - mount sma connectors on the circuit board .
thus , the data presented throughout this paper only reflect the s parameters of the circuits that we have developed , while eliminating any parasitic effects from the cables and supporting systems .
we acknowledge funding support from the us office of naval research ( grant n00014 - 16 - 1 - 2830 ) , the national science foundation emerging frontiers in research and innovation newlaw program ( grant efma-1627184 ) , and a national science foundation graduate research fellowship .
c.w.p . and g.b .
conceived the idea and with s.k . developed the theory .
designed and carried out the experiments .
j.t.b . provided material support and guidance for the experiments .
all authors analyzed the data and co - wrote the paper .
g.b . supervised all aspects of this project .
+ in this supplement we explicitly derive the nonreciprocal scattering parameters for a simple waveguide - resonator system with multiple spatiotemporally modulated couplers .
the system under consideration , shown in figure [ figs1 ] , consists of two ports connected by a waveguide with input and output fields described by @xmath139 and @xmath140 ( where @xmath141 is the port number ) and a resonant mode with angular frequency @xmath21 and amplitude @xmath16 .
the mathematical derivations in this paper are an extension of temporal coupled - mode theory @xcite modified to incorporate time - varying couplers .
we begin with the differential equation describing the resonant field @xmath16 of a standing - wave resonator in the time - domain with inputs @xmath25 and @xmath32 that are time - harmonic with frequency @xmath24 . @xmath142
this equation can be modified for a traveling wave resonator by simply splitting @xmath16 into two orthogonal modes which are each only coupled to a single port .
@xmath143 we can express the output fields @xmath42 , @xmath43 as @xmath144 where the subscript of @xmath16 is used only for the traveling wave resonator case . in these equations @xmath21 is the resonance frequency , @xmath23 is the decay rate of the mode , and @xmath145 is the effective coupling between port @xmath141 and the resonant mode .
we assume the waveguide is matched to the ports and is lossless with a propagation constant @xmath34 .
the couplers are spatially separated in the waveguide by a length @xmath20 , which leads to the term @xmath146 to account for the phase associated with a wave traveling between ports . at steady state
the solution to equation [ s1 ] ( a standing wave resonator ) is is described by @xmath139 and @xmath140 respectively and the resonant field is described by @xmath16 .
the ports are coupled directly via a waveguide and are also coupled to a resonant modes using a system of @xmath19 spatiotemporally modulated capacitive couplers with coupling rate @xmath28 .
the individual couplers are spatially distributed over the waveguide with subwavelength spacing @xmath20 in order to generate sensitivity to wave propagation direction . ]
@xmath147 we can then rewrite equation [ s2 ] as @xmath148 in the system under consideration , shown in fig .
[ figs1 ] , the effective coupling rates @xmath26 and @xmath35 can be written as @xmath149 where we modulate the coupling rate of each individual coupler @xmath28 with the function @xmath62 . under this modulation
the effective coupling rates @xmath26 and @xmath35 can be expressed as @xmath150 for simplicity we assume a resonant mode with only one input ( @xmath151 , which could be a standing wave or traveling wave resonator .
we later discuss the effects of @xmath32 on the standing wave case and the difference in the decay rate @xmath23 .
the steady - state field in the resonator is thus @xmath152 the time - varying terms here represent sidebands of the resonance created by the modulated coupling , which allow waves with a frequency offset @xmath17 from the resonance frequency to couple into the resonance .
we can now find the steady - state fields @xmath42 and @xmath43 ( also with @xmath39 ) as @xmath153 \\
+ a_1 e^{\omega t } \bigg [ & - \frac{d_0 d_m}{2 } \frac { \sum_{n=1}^n e^{-j \beta \ell ( n-1 ) } \sum_{m=1}^n e^{j \beta \ell ( m - n ) } e^{- j \phi ( m-1 ) } } { j ( \omega - \omega_0 ) + \gamma } \\ &
- \frac{d_0 d_m}{2 } \frac{\sum_{n=1}^n e^{-j \beta \ell ( n-1 ) } e^{-j \phi ( n-1 ) } \sum_{m=1}^n e^{j \beta \ell ( m - n ) } } { j ( \omega + \omega - \omega_0 ) + \gamma } \bigg ] \\ + a_1 e^{-j \omega t }
\bigg [ & - \frac{d_0 d_m}{2 } \frac { \sum_{n=1}^n e^{-j \beta \ell ( n-1 ) } \sum_{m=1}^n e^{j \beta \ell ( m - n ) } e^{j \phi ( m-1 ) } } { j ( \omega - \omega_0 ) + \gamma } \\ &
- \frac{d_0 d_m}{2 } \frac{\sum_{n=1}^n e^{-j \beta \ell ( n-1 ) } e^{j \phi ( n-1 ) } \sum_{m=1}^n e^{j \beta \ell ( m - n ) } } { j ( \omega - \omega - \omega_0 ) + \gamma } \bigg ] \\ + a_1 e^{2 \omega t }
\bigg [ & - \frac{d_m^2}{4 } \frac { \sum_{n=1}^n e^{-j \beta \ell ( n-1 ) } e^{- j \phi ( n-1 ) } \sum_{m=1}^n e^{j \beta \ell ( m - n ) } e^{- j \phi ( m-1 ) } } { j ( \omega + \omega - \omega_0 ) + \gamma } \bigg ] \\
+ a_1 e^{-j 2 \omega ) t } \bigg [ & - \frac{d_m^2}{4 } \frac
{ \sum_{n=1}^n e^{-j \beta \ell ( n-1 ) } e^ { j \phi ( n-1 ) } \sum_{m=1}^n e^{j \beta \ell ( m - n ) } e^{j \phi ( m-1 ) } } { j ( \omega - \omega - \omega_0 ) + \gamma } \bigg ] ~ .
\end{aligned}\ ] ] from the output field @xmath42 we find the transmission coefficient @xmath38 ( with @xmath39 ) .
@xmath154 \\
+ e^{j \omega t } \bigg [ & - \frac{d_0 d_m}{2 } \frac { \sum_{n=1}^n e^{-j \beta \ell ( n-1 ) } \sum_{m=1}^n e^{j \beta \ell ( m - n ) } e^{- j \phi ( m-1 ) } } { j ( \omega - \omega_0 ) + \gamma } \\ &
- \frac{d_0 d_m}{2 } \frac{\sum_{n=1}^n e^{-j \beta \ell ( n-1 ) } e^{-j \phi ( n-1 ) } \sum_{m=1}^n e^{j \beta \ell ( m - n ) } } { j ( \omega + \omega - \omega_0 ) + \gamma } \bigg ] \\
+ e^{-j \omega t } \bigg [ & - \frac{d_0 d_m}{2 } \frac { \sum_{n=1}^n e^{-j \beta \ell ( n-1 ) } \sum_{m=1}^n e^{j \beta \ell ( m - n ) } e^{j \phi ( m-1 ) } } { j ( \omega - \omega_0 ) + \gamma } \\ & - \frac{d_0 d_m}{2 } \frac{\sum_{n=1}^n e^{-j \beta \ell ( n-1 ) } e^{j \phi ( n-1 ) } \sum_{m=1}^n e^{j \beta \ell ( m - n ) } } { j ( \omega - \omega - \omega_0 ) + \gamma } \bigg ] \\ + e^{j 2 \omega t } \bigg [ & - \frac{d_m^2}{4 } \frac
{ \sum_{n=1}^n e^{-j \beta \ell ( n-1 ) } e^{- j \phi ( n-1 ) } \sum_{m=1}^n e^{j \beta \ell ( m - n ) } e^{- j \phi ( m-1 ) } } { j ( \omega + \omega - \omega_0 ) + \gamma } \bigg ] \\ + e^{-j 2 \omega t } \bigg [ & - \frac{d_m^2}{4 } \frac { \sum_{n=1}^n e^{-j
\beta \ell ( n-1 ) } e^ { j \phi ( n-1 ) } \sum_{m=1}^n e^{j \beta \ell ( m - n ) } e^{j \phi ( m-1 ) } } { j ( \omega - \omega - \omega_0 ) + \gamma } \bigg ] ~ .\end{aligned}\ ] ] the reverse transmission coefficient @xmath37 can be found as @xmath40 ( with @xmath41 ) .
we can thus understand the system as taking one input and splitting it into five frequencies separated by the modulation frequency @xmath17 .
presently , we wish to only consider the solution where output terms are the same frequency @xmath24 as the input .
these are the terms that are measured on a typical network analyzer , while terms that are frequency offset can be dismissed as noise produced by the system .
we then obtain @xmath155 ~ .
\end{split}\ ] ] using this frequency limited output we can rewrite @xmath36 as @xmath156 ~ .
\end{split}\ ] ] we have introduced the coupling efficiency function @xmath157 for @xmath158 . ] the function @xmath47 is a periodic even function that is symmetric about @xmath49 for integer @xmath51 as illustrated in fig .
[ figs2 ] for selected values of @xmath19 .
while the above derivation assumes @xmath39 , this system can be readily solved for simultaneous inputs @xmath25 and @xmath32 as written in equations [ s1 ] and [ s2 ] .
it is clear from equation [ s1s ] that for a standing wave resonator the resonant field @xmath16 is simply a superposition of the contribution from both inputs , therefore in general the outputs @xmath43 and @xmath42 depend on both inputs .
this dependence is obviously undesirable for the functionality of a system but is resolved in the case where waves propagating in one direction are completely uncoupled from the resonance ( completely unidirectional coupling ) . as discussed in the manuscript
this condition is also important to other functions of the system such as lossless functionality and isolation .
the dependence of the outputs on both inputs is automatically resolved for a traveling wave resonator since the orthogonal resonant fields each only depend on one of the inputs .
as previously shown @xcite , the decay rate @xmath23 for a lossless resonator is defined as @xmath159 which can be extended to lossy resonators through the addition of the intrinsic decay rate @xmath56 .
we note that for a traveling wave resonance only the @xmath13 term coupling the resonant field to the output field @xmath15 ( in equation [ s2 ] ) is used .
our system has time - varying @xmath145 , so is it necessary to slightly modify this definition to @xmath160 where @xmath53 @xmath54 represents the time - average .
this approximation is valid because @xmath161 and the field in the resonator decays very slowly compared to the modulation .
the expression for @xmath23 can then be simplified to @xmath162 where the frequency dependence of @xmath34 is explicitly written for clarity .
the decay rate @xmath23 defines the resulting modal linewidth and is present in the denominator of all the s - parameter equations for our nonreciprocal time - varying coupler . | breaking reciprocity associated with wave propagation is a fundamental challenge across a wide range of physical systems in electromagnetics , optics , and acoustics .
in contrast to established magneto - optic methods , recent efforts have focused instead on inducing nonreciprocity using time - varying biases such as synchronized spatiotemporal modulation of resonators and waveguides . to date , the elements that couple these time - varying resonator and waveguide modes are always considered to be reciprocal , and the nonreciprocal transfer functions achieved by these systems have been mostly limited to narrow - band resonances . here
we demonstrate that transferring the time - varying bias to the often overlooked coupling between waveguides and resonators can permit the synthesis of arbitrary nonreciprocal transfer functions . using simple microwave circuits
we experimentally demonstrate dynamically reconfigurable nonreciprocal responses , and the possibility of arbitrary function synthesis , through the use of this nonreciprocal coupler .
this unique method of breaking reciprocity is applicable to essentially all wave phenomena , opening the door for the creation of a broad class of nonreciprocal systems .
reciprocity is a fundamental property of wave propagation in linear time - invariant ( lti ) media @xcite that implies invariance of a system when its inputs and outputs are exchanged and arises due to the symmetry of the medium under time - reversal . in classical systems reciprocity is best captured through network theory , where a reciprocal network is defined by a symmetric scattering matrix @xmath0 @xcite .
the off - diagonal elements of this matrix describe the transmission of waves between any two ports of the network .
since the reciprocity property derives from time - reversal symmetric media , nonreciprocal networks can be created by applying a time - varying bias that generates the property @xmath1 , where @xmath2 is the time variable of the bias .
under such a bias the scattering matrix will become asymmetric and nonreciprocal , but will retain the property that @xmath3 .
historically , linear nonreciprocal networks have been predominantly based on the magneto - optic effect in gyrotropic materials @xcite and can be understood by the relationship @xcite @xmath4 under influence of magnetic field @xmath5 .
magnetically induced nonreciprocity may be also considered a subset of the time - variation approach since magnetic fields are quantifiable by the time parameter of an equivalent current loop .
recent efforts on achieving linear nonreciprocal effects without magnetic fields have focused on using spatiotemporal biases in optical , rf , and acoustic domains @xcite .
for instance , the scattering of light from a propagating acoustic wave in a medium depends on the relative momentum vectors of the photons and phonons and the available density of states in each domain .
such a traveling acoustic wave bias can thus be exploited to generate nonreciprocal effects @xcite .
most systems of this type can be generalized to a combination of ports , waveguides , and resonances ( fig .
[ fig1]*a * ) .
generally , we can describe these systems using coupled mode theory through the equations @xcite @xmath6 where @xmath7 are complex valued vectors respectively describing the field in the resonances , into the ports , and out of the ports as illustrated .
four matrices , @xmath8 , represent the system parameters : resonance frequencies , resonance decay rates , coupling between the ports and the resonances , and coupling between ports respectively .
the scattering matrix @xmath9 , which is only dependent on the four matrices that describe the system , can be obtained through the relationship @xmath10 .
past research has shown that a nonreciprocal ( asymmetric ) scattering matrix can be obtained by applying a spatiotemporal bias to coupled resonant modes ( @xmath11 matrix ) @xcite or the waveguide ( @xmath12 matrix ) @xcite but the coupling between these waveguides and resonances ( @xmath13 matrix ) is always assumed to be linear time invariant . in this work
, we invert the situation and demonstrate instead an approach to achieving nonreciprocal propagation in lti waveguides and resonators through the use of spatiotemporally modulated coupling ( @xmath13 matrix ) while keeping resonances and waveguides unchanged .
as we will show , the power of using this approach is that highly tailorable nonreciprocal transfer functions can be arranged without needing sophisticated modulation schemes involving multiple synchronized resonances or waveguides @xcite .
the use of nonreciprocal coupling also opens new and convenient paths to produce a variety of nonreciprocal devices using a single waveguide and resonator .
we discuss how ultra low loss isolators could be created through nonreciprocal critical coupling and gyrators ( nonreciprocal phase - shifters ) could be created nonreciprocal over - coupling to resonators . , and can also couple to resonant modes via the coupling matrix @xmath13 .
the inputs and outputs at the ports are described by the vectors @xmath14 and @xmath15 respectively .
the resonant field vector @xmath16 is described through the frequency matrix @xmath17 and the decay matrix @xmath18 . *
b * in a simple system with one resonance and two ports , nonreciprocity can be synthesized through the coupling matrix @xmath13 using a system of @xmath19 spatiotemporally modulated capacitive couplers .
the individual couplers are spatially distributed over the waveguide with subwavelength spacing @xmath20 in order to generate sensitivity to wave propagation direction . for mathematical simplicity , we assume that each coupler is connected to an isophase point of the resonant mode .
sinusoidal time - modulation of the individual coupling rates at a relative phase set by their spatial location generates significant nonreciprocal effects that we see in this work .
, scaledwidth=60.0% ] in this paper we first develop a theory for directly generating a nonreciprocal coupling matrix @xmath13 using spatiotemporally modulated coupling elements . using this approach ,
we experimentally demonstrate significant nonreciprocal responses in simple resonant microwave circuits . we discuss how nonreciprocal coupling automatically creates networks that are theoretically lossless in the uncoupled direction and how it can be applied to arbitrary resonant structures and impedance networks . |
parallel linear algebra algorithms have been developed for many hardware architectures with the aim of accelerating routine calculations @xcite . in this paper
we report on our experiences with parallel calculation of determinants of large nearly singular matrices with very high accuracy .
the need to solve such a problem came from research on the famous riemann s _
zeta function_. the distribution of zeroes of zeta function has puzzled mathematicians for over a century .
the famous riemann hypothesis @xcite , which was included by d. hilbert at the very end of xix century as part of his 8-th problem , and which is also one of the clay institute seven millennium problems @xcite , is to prove or disprove that all non - real zeroes of @xmath0 lie on the critical line @xmath1 . among other things , riemann s zeta function
is closely related to the distribution of prime numbers , whose fundamental role in mathematics is well known .
recently , the second author has proposed what he called _
artless method _ for studying zeroes of the zeta function ( see @xcite and visit @xcite ) .
it consists in approximating zeroes by using a special interpolant built on already known zeroes of zeta function . in the process of building the interpolant ,
it is necessary to compute with high accuracy the determinant of a large matrix , as a function of a parameter in its last column , and hence the signed minors corresponding to the last column are needed .
the matrix is nearly singular , and it was required to perform calculations with the accuracy of ten thousand decimal places .
furthermore , certain patterns in the behaviour of these minors as a function of matrix size @xmath2 needed to be investigated @xcite , and therefore the sequence of arrays of minors for matrix sizes @xmath3 was required .
all those special requirements prevented us from mapping directly the problem at hand to one of the standard parallel linear algebra solutions and using existing tools such as scalapack or plasma @xcite .
in this work we describe our approaches to solving the problem of calculating the required sequence of minors with high precision on various architectures .
initially we attempted using graphics processing units ( gpus ) for this task , and designed and implemented a parallel algorithm for this architecture , using quad - precision ( 256-bits ) and arbitrary precision libraries @xcite .
our initial experiments , though , indicated that gpu architecture delivered about the same efficiency as one single core of a modern cpu when using arbitrary precision libraries ( with over 1000 bits accuracy ) .
subsequently we implemented a parallel algorithm for shared memory multicore architecture using ` pthread ` library , which exhibited linear reduction of cpu time when using up to 16 cores .
however , it soon became clear that only a sufficiently large cluster would be capable of solving the problem with ten thousand decimal places ( 32kbits accuracy ) for @xmath2 in the order of 10 thousand .
indeed , just to store the elements of one matrix of that size , 400 gb of ram was needed .
therefore we designed and implemented an mpi ( message - passing interface ) algorithm and verified it on several computer clusters , running on up to 200 cores .
we dealt appropriately with the issue of load balancing , and at the end observed nearly linear gain in performance with the number of nodes used .
we successfully verified the algorithm on the computations with two types of matrices needed for the artless method .
this paper is structured as follows . in section [ sec1 ]
we will formally describe the problem and give a brief introduction to some elements of the artless method . in section [ sec2 ] we present computational algorithm and its various parallelizations for gpu - cuda , pthread , openmp and mpi platforms . in section [ sec3 ]
we detail some of the numerical experiments and quantify the performance of our algorithm , including the loss of accuracy .
section [ sec_conc ] concludes .
zeta function is defined in the complex plane by the analytic continuation of the series @xmath4 which converges for @xmath5 .
it satisfies the functional equation @xmath6 riemann also defined a symmetric version of the above , using function @xmath7 , @xmath8 which yields a simpler functional equation @xmath9 the _ trivial zeroes _ of @xmath10 are negative even integers , and they are its only real zeroes .
the other , _ non - trivial zeroes _ can only be found in the strip @xmath11 .
the zeroes of the function @xmath7 are exactly the non - trivial zeroes of the function @xmath10 . also following riemann
, we make a change of variables @xmath12 and define @xmath13 the functional equation implies that @xmath14 is even function , @xmath15 . with this notation , the riemann hypothesis states that all zeroes of @xmath14 are real numbers .
an important relation of zeta function to prime numbers was given by von mangoldt .
let chebyshev function @xmath16 be @xmath17 then for non - integer @xmath18 greater than @xmath19 , by von mangoldt s theorem @xmath20 can be expressed as @xmath21 the first sum runs over the trivial zeroes of zeta , and the second sum runs over the non - trivial zeroes .
in fact , knowing zeroes of zeta function , one could compute primes by merely looking at the graph of the right - hand side of , and identifying the powers of primes by jumps in the graph . assuming the riemann hypothesis and that all zeroes are simple ,
let us denote by @xmath22 the real zeroes of @xmath14 listed by increasing of absolute values .
further , let @xmath23 thanks to the functional equation we formally have @xmath24 we define the _ interpolating determinant _ as @xmath25 clearly , the determinant @xmath26 vanishes as soon at @xmath27 is equal to @xmath28 , because for such a @xmath27 there are two equal columns in .
the surprising observation is that @xmath26 vanishes also at certain points extremely close to a number of the next zeroes @xmath29 .
the larger is @xmath2 , the more subsequent zeroes approximately coincide with the zeroes of @xmath30 .
for example determinant @xmath31 has zeroes having more than 500 common decimal places with @xmath32 .
the determinant @xmath26 can be expanded into the linear combination of functions @xmath33 , @xmath34 where @xmath35 the object of interest are normalized minors @xmath36 in particular , function @xmath37 has the same zeroes as @xmath26 .
the second matrix of interest had a slightly different structure @xmath38 where @xmath39 are the non - trial zeroes of @xmath10 and @xmath40 .
similarly , zeroes of @xmath41 for @xmath42 have more than 1000 common decimal places with @xmath43 .
the second author observed several patterns of behaviour of the sequences of coefficients @xmath44 for various @xmath2 @xcite .
these patterns have clear number - theoretical meaning , but in order to discover and assess such patterns , which have very fine structure , one has to compute the minors given in ( [ minors ] ) with extremely high accuracy to avoid numerical artifacts .
further , the interpolating matrices are nearly singular , and the values @xmath45 approach zero very rapidly with @xmath2 ( e.g. @xmath46 ) , so one expects very large losses of precision in numerical calculation .
all this dictates the need to use very high accuracy in calculations , of order of ten thousand decimal places . at this point
we abandon the topic of riemann s zeta function , and focus squarely on calculating the determinants @xmath47 ( or @xmath48 ) and the corresponding normalized minors , as the main numerical complexity of the artless method comes from these calculations .
there is a number of ways determinants can be evaluated numerically .
matrix factorisation is a standard method , although not necessarily the best when computations are parallelised .
our initial thought was to look at parallel versions of the condensation method .
this condensation method was inspired by the celebrated dodgson s condensation method @xcite . in dodgson
s method , an @xmath49 matrix determinant is condensed " into an @xmath50 determinant by calculating @xmath51 connected subdeterminants of sizes @xmath52 . this algorithm is trivially parallel , yet it suffers from the requirement of having no zeroes in the interior of the matrix .
inspired by the parallel nature of the condensation steps , the authors of @xcite proposed a variant of the condensation method , in which the determinant of an @xmath49 matrix @xmath53 is replaced by the determinant of an @xmath50 matrix @xmath54 , calculated as follows @xmath55 for @xmath56 , and @xmath57 for @xmath58 , where @xmath59 is the smallest column index of a non - zero element in the first row of @xmath53 .
the relation between @xmath53 and @xmath54 is that @xmath60 after @xmath61 condensation steps the determinant of @xmath53 reduces to a single number divided by a product of powers of pivot factors .
the condensation method from @xcite was taken up by @xcite , where it was slightly modified to avoid division by @xmath62 , by factoring out the pivot element @xmath63 from the first column of @xmath53 ( hence getting a modified matrix @xmath64 ) , so that the relation between the determinants of the matrices becomes @xmath65 with @xmath66 obtained from @xmath64 in ( [ condensation ] ) .
successive condensation steps produce the array of pivot elements , whose product is the determinant of @xmath53 .
this modification of the condensation method is data parallel as elements of @xmath54 are computed independently of each other .
the authors of @xcite developed a parallel version of their algorithm suitable for gpus and assessed its performance .
we started with formula ( [ condensation ] ) and also developed our own gpu parallel algorithm , differing from that of @xcite in that all computations were performed in place ( instead of having two rotating matrices for input and output ) , and that we used high precision arithmetic for all steps ( we used quad - precision qd library for gpus and gpuprec arbitrary precision libraries for gpus @xcite ) .
we soon realised though , that condensation steps in ( [ condensation ] ) were exactly the steps of gaussian elimination , and hence the version of condensation algorithm in @xcite and @xcite was in fact not different from gaussian elimination without pivoting .
optional pivoting ( as it was done in @xcite ) can be trivially added . knowing the equivalence of with the gaussian elimination steps , calculations could be streamlined , and computational complexity can be confirmed to be @xmath67 .
however , the question of calculation of minors corresponding to the last column of the matrix @xmath53 was still outstanding .
the naive approach by calculating @xmath2 determinants ( [ minorsdef ] ) was not appealing because of computational cost .
it is known that one can compute a determinant of size @xmath2 and all its @xmath68 minors in the same time @xmath67 .
it is based on the result from @xcite which establishes that the complexity of evaluation of all partial derivatives of certain functions ( in particular multivariate polynomials ) is a constant multiple of the complexity of evaluating the function itself .
since all the @xmath68 minors are partial derivatives of the determinant , the result follows . additionally , would it be possible to compute not just one array of signed minors @xmath45 for a particular @xmath2 , but the whole series of such arrays for @xmath3 in one non - redundant computation ? to answer these questions we used the following trick
let us construct the augmented matrix @xmath69,\ ] ] where @xmath70 is the identity matrix of size @xmath2 .
perform gauss elimination on @xmath71 . for completeness
, the algorithm is shown on fig .
[ alg : gauss ] .
the last @xmath2 elements of the last row will contain the signed minors of @xmath53 corresponding to the last column , divided by @xmath72 . to see this , recall
that gauss elimination ( without pivoting ) of a non - singular matrix @xmath53 is equivalent to multiplying it by a lower triangular matrix @xmath73 from the left , @xmath74 where @xmath75 is in row - echelon form .
the same operations applied to @xmath70 result in @xmath76 .
now , the product of the last row of @xmath73 and the last column of @xmath53 yield @xmath77 , since @xmath75 is in row - echelon form . if we now multiply @xmath78 by @xmath72 , we obtain the result , @xmath79 .
we can see that all the required signed minors can be computed automatically using gaussian elimination ( the determinant of @xmath53 is found from the product of diagonal elements of @xmath73 ) with the help of additional row operations .
the computational cost of it is twice that of gaussian elimination , and computational complexity remains @xmath67 .
storagewise , the matrix @xmath73 can be stored in the lower triangular part of @xmath53 , because once the row @xmath80 is processed in gaussian elimination , the elements @xmath81 below diagonal with @xmath82 are no longer required ( indeed they are zero in row - echelon form ) . on the other hand , at step @xmath80 of the elimination process , only the elements @xmath83 with @xmath82 and the main diagonal are different from zero .
therefore , nonzero elements @xmath83 can be stored below the diagonal of @xmath53 , and the only extra storage required is for the overlapping diagonals of @xmath53 and @xmath73 .
therefore all calculations can be done in memory @xmath84 , as only one extra array of size @xmath2 is required .
next we turn to the second question of whether it is possible to compute the whole series of the minors , for different sizes of the matrices @xmath3 .
the answer here is also affirmative , and in fact no extra work or storage is required at all . to see this
, note that at step @xmath80 of gaussian elimination , the row @xmath80 of matrix @xmath73 contains the desired ( up to a factor ) minors , corresponding to the @xmath80th column of @xmath53 .
we apply the same reasoning as before to show this is the case : note that after step @xmath80 , the diagonal element of the row echelon form @xmath85 , which is the product of the @xmath80th row of @xmath73 and @xmath80th column of @xmath53 .
to put it into the context of our motivating problem , all normalised minors @xmath45 ( [ minors ] ) , @xmath86 , @xmath87 can be computed with cpu cost @xmath88 and storage @xmath89 using gaussian elimination .
there are methods for matrix inversion and calculation of determinant based on fast matrix multiplication , e.g. schnhage - strassen @xcite and coppersmith - winograd algorithms @xcite , which have complexity @xmath90 and @xmath91 respectively . through the result of @xcite ,
all @xmath68 minors can also be computed in the same time , although we were unable to find implementations of such algorithms in the literature .
however , it appears that the fast matrix multiplicaiton algorithms are not structured to facilitate computation of the whole sequence of arrays of minors for @xmath87 in one run .
therefore we settle on gaussian elimination as the most efficient way of computing all the normalised minors @xmath45 .
the question of available ram becomes important for computations with larger @xmath2 of order of 10000 . as we mentioned previously , to store a matrix of that size , with 10 thousand decimal places accuracy
, 400 gb of ram is required .
one approach we pursued is to partition the matrix into blocks and use memory paging to fit the blocks into available ram .
let us partition the matricies into @xmath92 blocks of size @xmath93 , with @xmath94 .
each block is stored on a hard disk and is loaded to ram when needed .
as soon as certain blocks are in ram , we perform all possible steps of gaussian elimination with data in these blocks . at each point in time at most four blocks are needed in ram , hence ram required is @xmath95 .
however , now we do not use shared space to keep both matrix @xmath53 and @xmath73 , so memorywise the algorithm uses @xmath96 space .
the operations within blocks were executed in parallel in different threads on a single host , using openmp library , however distinct sets of blocks could be processed in parallel on different hosts in a specified order but otherwise asynchronously . this way a cluster of hosts can be used for parallel execution
@xmath97 , @xmath98 ( identity matrix ) 2 . for @xmath99
@xmath100 2 .
for @xmath101 do : 1 . @xmath102 2 . for @xmath103 do : + @xmath104 3 .
@xmath105 4 .
for @xmath106 do : + @xmath107 3 .
return @xmath108 and @xmath73 .
we implemented a variant of gaussian elimination with high accuracy on four platforms : 1 . on gpu using cuda and cump @xcite arbitrary precision and gqd @xcite quad - double ( 256 bits ) precision libraries , connected to a single host ; 2 .
on a cluster of multicore cpus using ` openmp ` for multithreading and gmp @xcite arbitrary precision library , partitioning the matrix into blocks and using memory paging ; 3 . on a multicore cpu using ` pthread ` for multithreading and gmp @xcite arbitrary precision library ; 4 . on a computer cluster using mpi , and using gmp , mpfr @xcite and mpigmp @xcite libraries .
the following parallelisation strategies were adopted . on gpu , at every step @xmath109 of the gauss elimination algorithm , we spawned @xmath2 threads , and each thread @xmath110 was performing calculations in the @xmath110th column of matrix @xmath53 starting from row @xmath109 .
while this was efficient in terms of coalescent global memory access pattern for data type ` double ` and to some degree for ` quad - double ` , for arbitrary precision numbers the way gpu threads accessed elements of the matrix @xmath53 did not matter , as the data were stored in non - sequential locations anyway , and hence was misalligned .
we did not observe any differences in cpu time when parallelisation was done columnwise , rowwise or elementwise . when we used memory paging and matrix partitioning in the second approach , parallelization was performed in two ways .
first , we created a queue of tasks , from which tasks were taken by idle hosts based on certain pre - conditions .
each task consisted of a block in which gaussian elimination was performed , and pre - conditions ensured that a block was processed only when all the blocks containing rows and columns with smaller indices have been processed .
second , within each block the computations were parallelized with openmp . common to the remaining implementations
, we parallelised the algorithm rowwise , that is , each thread was performing calculations in a specific row , or a group of rows .
the simplest way is to break the matrix into @xmath111 blocks of rows sequentially , where @xmath111 is the number of threads , and let each thread process its own block .
some load balancing strategy is needed , as the first thread will complete processing of its rows before the rest and will be then idle , then it will be the second thread and so on . to avoid this , on shared memory architectures we reassigned the blocks to each thread every now and then , so that at all steps the sizes of the blocks were approximately the same .
this was easily done by simply changing the pointers to blocks of rows each thread was responsible for , hence no copying of data took place , and overheads were negligible . on distributed memory architectures , though , such as clusters running mpi ,
this method was not suitable as it would involve copying large chunks of data between the hosts .
instead , we interleaved the rows of the matrix , i.e. , thread @xmath110 was processing rows @xmath110 , @xmath112 , @xmath113 , ... and so on , @xmath114 .
the threads processed the pivot rows of the matrix @xmath53 in turns , in round robin fashion .
this way load balancing was implicit , and all threads had equal job to do until the very end of the computations .
each pivot row was broadcast to all the threads at every step @xmath109 of the algorithm , hence @xmath2 rows were broadcast altogether , which is the smallest number when the matrix is partitioned among the threads , and complexity of the data transfers was therefore @xmath115 ( we remind that broadcast in mpi has logarithmic time ) .
the row being broadcast was packed into a binary array using ` mpigmp ` library @xcite , and then unpacked at the receiving end .
here we make an observation regarding the value of partitioning the matrix into blocks and using paged memory as opposed to storing the whole matrix in ( combined ) ram .
it has two main advantages : 1 ) less ram is needed , and 2 ) the upper limit on the matrix size @xmath116 needs not be fixed in advance : we can gradually get determinants of sizes @xmath117 .
extra blocks can be added later on . on the other hand ,
less ram means that some blocks were processed sequentially rather than in parallel .
the cost of data transfer between hard disk and ram was not significant compared to the cost of the actual computations .
however the overall storage requirement was 50% higher , as the matrix @xmath73 could not share the same space as @xmath53 .
we opted for the fourth method in our final computations , based on mpi and using combined ram of the available hosts , for the following reasons : 1 ) cpu and ram resources available to us were sufficient to perform computations for @xmath118 with 10 thousand decimal places accuracy , and in fact the amount of combined ram exceeded our permanent storage quota ; 2 ) coding the algorithm was more straightforward ; 3 ) at that moment in time , we did not need to increase @xmath116 beyond 12000 , because our computations have shown that the loss of accuracy during numerical computations was such that the results were unreliable beyond that number .
we compared the speed of determinant calculations on various platforms using the following hardware : tesla c2070 gpus with 6 gb of ram , and clusters of intel e5 - 2670 nodes with @xmath119 gb of ram , connected by 4x qdr infiniband interconnect , running centos 6 linux .
the hardware was provided by the vpac and monash e - research centre http://www.vpac.org , http://www.monash.edu.au / eresearch/. while tesla gpu was quite efficient in evaluating determinants with gqd package , when we passed to much higher accuracy ( from 256 to 8192 bits and more ) , gpu performance became equivalent to that of one cpu core , see table [ tab1 ] .
the computing time grew with the accuracy at the rate of @xmath120 , @xmath121 being the number of bits used for multiprecision float numbers .
this is consistent with the schoolbook multiplication algorithm employed in cump library .
subsequently we implemented a parallel shared memory algorithm using ` pthread ` library , and ran it on up to 16 intel cpu cores ( on the same host ) .
the efficiency increased linearly with the number of cores employed , as no significant interprocess communication took place , see table [ tab2 ] . therefore the cpu time has decreased 16-fold compared to running the algorithm on a gpu .
furthermore , since gmp library @xcite uses a faster karatsuba multiplication algorithm with complexity @xmath122 , the benefit was even greater .
as we mentioned previously , current shared memory architecture has a limitation of the total ram that could be used and the limited number of cores .
therefore we opted to parallelising the algorithm for an mpi cluster .
here we observed linear gain in performance as evidenced by table [ tab3 ] .
this is consistent with the fact that the dominating term in the overall complexity of the algorithm @xmath67 is due to computations , and the complexity of data transfers @xmath115 is comparatively small . given that the matrices involved in the computations are large and ill conditioned , a valid question of the accuracy of the end result arises . to estimate the loss of accuracy we performed computations with various precisions : 8 kbits , 32 kbits , and to a limited extent , 64 kbits .
the input data , the zeroes of riemann s zeta function were taken with just over 32 kbits accuracy ( 10000 decimal places ) @xcite . by comparing the results calculated with different accuracies
, we could estimate empirically the rate of accuracy loss during gaussian elimination .
figures [ accuracy64 ] , [ accuracy8 ] present graph of the number of coinciding decimal places in the normalised minors @xmath123 as a function of matrix size @xmath2 .
we compared both the average and the smallest number of coinciding decimal places for @xmath124 , which turn out to be very close ( difference @xmath125% ) .
what we observe is that the accuracy decays linearly .
the equation of regression line in figure [ accuracy64 ] is @xmath126 .
this indicates that the accuracy of calculations with 32 kbits accuracy is predicted to have 1400 correct decimal places . on the other hand , calculations with 8
kbits accuracy were valid only up to @xmath127 , and after that point were completely unreliable .
therefore the choice of 10000 decimal places accuracy was sufficient , and also necessary for calculations with matrices up to @xmath128 in size .
the coefficient @xmath129 indicates the number of decimal places lost in each row operation .
it corresponds to @xmath130 bits average accuracy loss per one row operation .
this result also shows that if the input values were calculated with 10000 decimal places , then it only makes sense to increase @xmath2 to 13500 , because after that size the loss of accuracy will make the results unreliable .
the above analysis was performed for determinants @xmath48 in ( [ betanew ] ) .
a similar analysis was performed for @xmath131 in ( [ det ] ) , which yielded a slightly different regression lines , e.g. the loss of accuracy at @xmath132 kbits was approximated by @xmath133 .
the coefficient @xmath134 indicates that for that matrix , the loss of accuracy was @xmath135 bits for each row operation .
hence it appears that matrix in ( [ det ] ) is slightly better conditioned than that in ( [ betanew ] ) . after completing initial evaluations , we performed two production runs of the algorithm on vpac ( http://www.vpac.org ) and massive ( http://www.massive.org.au ) clusters , using up to 168 processes .
we used the upper limit on the size of the matrix of @xmath118 and accuracy of 32768 bits .
one cluster was used for @xmath136 from ( [ det ] ) and the second for @xmath137 from ( [ betanew ] ) .
the accurate values of riemann s zeroes were precomputed @xcite .
the algorithms ran for 5 and 7 days respectively .
the results of our computations constituted almost 700 gb of compressed high accuracy data .
our results helped to confirm earlier calculations in @xcite for @xmath136 up to @xmath138 , which were performed with block partitioning of the matrix @xmath53 .
however we noticed some deviations from the earlier results for larger @xmath2 .
it was later confirmed that our current results were correct , hence these calculations were valuable in correcting previous computational errors .
the results with the determinant @xmath137 from ( [ betanew ] ) were new , and they helped establish various patterns in the values of the normalised minors @xmath45 , in particular their fine structure related to prime numbers , as presented recently in @xcite .
as we discussed earlier , the loss of accuracy in arithmetical operations was quite significant due to nearly degenerate matrices .
in addition , matrix entries themselves were computed up to 10 thousand decimal places ; that was the accuracy of zeroes of zeta function we started with .
however , our analysis is empirical , it provides only an indication , but not a guarantee , that the results are accurate to the indicated number of decimal places .
interval arithmetic can be employed to obtain rigorous results . at every operation ,
the upper and lower bounds on the result can be computed . a library ` mpfi ` @xcite is available for such calculations with arbitrary precision .
however the cost of such computations is doubling the cpu time and memory requirements .
in addition , interval computations provide pessimistic error bounds , much larger than the actual errors .
a posteriori interval analysis @xcite provides more realistic error bounds , but it requires @xmath67 memory , and hence bears prohibitive cost in our case .
it might be possible to reduce its cost by not keeping all intermediate values , and in addition it may help save cpu time , as due to the inevitable loss of accuracy , later iterations of the algorithm need not be performed with full precision .
we leave this analysis for future work .
calculation of matrix determinant is one of the standard linear algebra operations needed for many computational tasks .
determinants of ill - conditioned matrices are a particular challenge because of rapidly degrading accuracy , hence very high precision calculations are needed .
we have presented an algorithm based on gaussian elimination which computes not only the determinant but a series of determinants and minors corresponding to one column or row for matrix sizes @xmath87 in one run .
the algorithm has @xmath88 cpu and @xmath139 memory complexity .
we parallelized and implemented this algorithm on four different parallel architectures , including gpu and mpi - based clusters , using arbitrary precision arithmetics .
we confirmed that the cpu wall time decreases linearly with the number of cpu cores used ( for fixed matrix size and accuracy ) .
our production runs involved up to 168 cores , 400 gb combined ram and determinants of up to @xmath118 in size with 10 thousand decimal places accuracy .
we empirically estimated the losses of accuracy during gaussian elimination and found that the accuracy used was appropriate for computations with that matrix size .
we applied the algorithm to studies of the riemann zeta function as part of the 8-th hilbert problem .
the results helped the second author to observe various theoretically interesting patterns in the values of the normalized minors related to prime numbers @xcite .
the authors wish to acknowledge support by victorian partnership in advanced computing ( vpac ) and monash e - research centre for providing computing resources at their clusters , and specifically mr .
s. michnowicz for his help in developing mpi parallelization code .
part of the calculations was performed on the `` chebyshev '' supercomputer of moscow state university supercomputing center .
the work of the second author was partly supported by the programme of fundamental research
modern problems of theoretical mathematics of the mathematics branch of the russian academy of sciences .
e. agullo , j. demmel , j. dongarra , b. hadri , j. kurzak , j. langou , h. ltaief , p. luszczek , and s. tomov . numerical linear algebra on emerging architectures : the plasma and magma projects .
, 180:012037 , 2009 .
l. s. blackford , j. choi , a. cleary , e. dazevedo , j. demmel , i. dhillon , j. dongarra , s. hammarling , g. henry , a. petitet , k. stanley , d. walker , and r. c. whaley . .
society for industrial and applied mathematics , philadelphia , pa , 1997 .
new conjectures about zeroes of riemann s zeta function .
research report ma12 - 03 of the department of mathematics of the university of leicester , available online http://www2.le.ac.uk/departments/mathematics/research/research-reports-2/reports-2012/ma12_03matiyasevich.pdf and from @xcite , 2012 .
calculation of riemann s zeta function via interpolating determinants .
preprint 2013 - 18 of max planck institute for mathematics in bonn , available online http://www.mpim-bonn.mpg.de/preblob/5368 and from @xcite , 2013 .
matiyasevich and g. beliakov .
zeroes of riemann s zeta function on the critical line with 20000 decimal digits accuracy , research data australia , available online http://hdl.handle.net/10536/dro/du:30051725 , 2013 . t. nakayama and d. takahashi .
implementation of multiple - precision floating - point arithmetic library for gpu computing . in _ proc .
23rd iasted international conference on parallel and distributed computing and systems ( pdcs 2011 ) _ , pages 343349 .
iasted , 2011 . | we present a parallel algorithm for calculating very large determinants with arbitrary precision on computer clusters .
this algorithm minimises data movements between the nodes and computes not only the determinant but also all minors corresponding to a particular row or column at a little extra cost , and also the determinants and minors of all submatrices in the top left corner at no extra cost .
we implemented the algorithm in arbitrary precision arithmetic , suitable for very ill conditioned matrices , and empirically estimated the loss of precision .
the algorithm was applied to studies of riemann s zeta function .
* keywords * determinant , linear algebra , parallel algorithms , message passing interface , gpu , riemann s zeta function . |
the spectra of hot stars show often excess emission at ir and radio wavelengths that can be ascribed to free - free and free - bound ( ff - fb ) emission from their wind zones ( see e.g. panagia & felli @xcite ) .
waters & lamers ( @xcite ) have investigated this excess emission for @xmath4 m and winds with a @xmath1-law velocity distribution of varying @xmath1 , pointing already to the sensitivity of the wind emission to the chosen velocity distribution . over the last few years
, two major effects have become obvious that both strongly influence the wind continuum emission : ( i ) the winds of hot stars seem to be clumped , and ( ii ) many ob supergiants have winds with @xmath5 ( see table [ kraus : tab1 ] ) .
we investigate the wind continuum emission of ob supergiants especially at optical wavelengths .
first , the influence of high @xmath1 values is discussed , and later on the effects of clumping are studied .
the calculation of the continuum emission of a typical ob supergiant is performed in three steps : ( i ) we first calculate the stellar emission of the supergiant with no stellar wind , ( ii ) then we calculate the emission of the wind with the stellar parameters ( @xmath6 ) as boundary conditions , ( iii ) and finally we combine the two continuum sources whereby the stellar emission still has to pass through the absorbing wind . to simulate a typical ob supergiant
we adopt the following set of stellar parameters : @xmath7k , @xmath8 , @xmath9 , and @xmath10 . with these parameters
we compute the stellar continuum emission with the code of kubt ( @xcite ) , suitable for the calculation of nlte spherically - symmetric model atmospheres .
.range of @xmath1 values found for ob supergiants in the galaxy ( markova et al .
@xcite = ma ; crowther et al .
@xcite = cr ; kudritzki et al .
@xcite = ku ) and the magellanic clouds ( evans et al .
@xcite = ev ; trundle & lennon @xcite = tl ; trundle et al .
@xcite = tr ) .
[ cols="<,^,<",options="header " , ] the spherically symmetric wind is assumed to be fully ionized , isothermal , and in lte .
this reduces the problem to a pure 1d treatment of the simplified radiation transfer ( e.g. panagia & felli @xcite ) .
the electron temperature is fixed at 20000k , and the density distribution in the wind follows from the equation of mass continuity , relating the density at any point in the wind to the mass loss rate and the wind velocity . the velocity of hot star winds can be approximated with a @xmath1-law @xmath11 where @xmath1 describes the steepness of the velocity increase at the base of the wind , and @xmath12 defines the velocity on the stellar surface . a more detailed description of the calculations will be given elsewhere .
the range in @xmath1 found for galactic and magellanic cloud ob supergiants is listed in table[kraus : tab1 ] . increasing @xmath1 means that the wind is accelerated more slowly .
consequently , the density in these regions is enhanced because @xmath13 . these density peaks close to the stellar surface with respect to the wind density with @xmath14
are shown in fig.[kraus : dens ] .
even though these density peaks are rather narrow in radius , they strongly influence the optical depth and therefore the emission of the ff and especially the fb processes .
now , the wind can become ( at least partially ) optically thick even at optical wavelengths .
this leads to an enhanced wind emission meanwhile the stellar flux suffers from the simultaneously increasing wind absorption .
our test supergiant is assumed to have a wind with @xmath15yr@xmath16 , @xmath17kms@xmath16 , and we calculate the continuum emission for @xmath18 , and 3 .
the results are shown in fig.[kraus : beta ] .
it is obvious that with increasing @xmath1 the wind creates a near - ir excess , absorbs more of the stellar emission , and contributes to the total emission even at optical wavelengths . )
that grow with increasing @xmath1.[kraus : dens ] ] hillier et al .
( @xcite ) introduced a formalism to account for the presence of wind clumping , and in our calculations we use their filling factor defined by @xmath19 and setting @xmath20kms@xmath16 , and @xmath21 .
since @xmath22 depends on the velocity distribution , this filling factor is a function of radius as well , constructed such that it quickly reaches its terminal value ( top panel of fig.[kraus : dens_cl ] ) .
this way of clumping introduction requests , however , that in order to maintain the same radio flux , the mass loss rate has to be decreased , i.e. @xmath23 , while the absorption coefficient of the ff - fb processes increases , @xmath24 , due to its density squared dependence .
our clumped models automatically account for this mass loss reduction . at those positions in the wind
where @xmath25 there is no difference between the clumped and the unclumped wind .
however , in those regions where @xmath26 , which are also the regions where @xmath1 has its strongest influence , the wind opacity is sensitive to the clumping .
but while @xmath1 enhances the density , clumping ( for the same input @xmath27 ) reduces the density again . a wind with high @xmath1 and clumping
will therefore have a different opacity in the innermost wind region than a wind with low @xmath1 and clumping , and the clumped wind will have a different opacity than the smooth wind .
this is shown in the lower panel of fig.[kraus : dens_cl ] where we plotted the opacity ratio of the clumped with respect to the smooth wind for different values of @xmath1 .
the higher @xmath1 , the stronger the effect . in fig.[kraus : clumped ] we compare the continuum of a clumped with an unclumped wind for @xmath28 . .
shown are the stellar emission having passed through the absorbing wind ( dotted ) , the ff - fb emission from the wind ( dashed ) and the total continuum ( solid ) .
[ kraus : beta ] ] .
bottom : opacity ratio between clumped wind model and unclumped wind model .
[ kraus : dens_cl ] ] it is obvious that the clumped model generates less wind emission for @xmath29 m . for a better visualization we calculated the flux ratios between unclumped and clumped models ( fig.[kraus : ratio ] ) .
they show a maximum of up to 30% at near - ir wavelengths .
continuum observations at these wavelengths are therefore an ideal tool to discriminate whether the winds of ob stars with @xmath30 are clumped . .
the clumped model produces less wind emission for @xmath31 m , resulting in a lower total near - ir flux .
[ kraus : clumped ] ]
for ob supergiants with high @xmath1 values the ff and especially the fb emission can strongly influence the total continuum , even at optical wavelengths .
clumping , introduced by the filling factor approach , also influences the wind opacity and therefore the continuum emission . whether the wind of an ob supergiant is clumped or not can be checked based on continuum observations in the optical and near - ir region .
especially winds with high @xmath1 are found to have fluxes that differ by about 30% ( see fig.[kraus : ratio ] ) .
the optical and near - ir continuum are therefore ideal to discriminate between clumped and unclumped winds .
crowther , p. a. , lennon , d. j. , & walborn , n. r. 2006 , a&a , 446 , 279 evans , c. j. , crowther , p. a. , fullerton , a. w. , & hillier , d. j. 2004 , apj , 610 , 1021 hillier , d. j. , lanz , t. , heap , s. r. , et al .
2003 , apj , 588 , 1039 kubt , j. 2003 , in modelling of stellar atmospheres , ed . n. piskunov , w. w. weiss , & d. f. gray ( san francisco : asp ) , iau symp . , 210 , a8 kudritzki , r. p. , puls , j. , lennon , d. j. , et al . 1999 ,
a&a , 350 , 970 markova , n. , puls , j. , repolust , t. , & markov , h. 2004 , a&a , 413 , 693 panagia , n. , & felli , m. 1975 , a&a , 39 , 1 trundle , c. , & lennon , d. j. 2005 , a&a , 434 , 677 trundle , c. , lennon , d. j. , puls , j. , & dufton , p. l. 2004 , a&a , 417 , 217 waters , l. b. f. m. , & lamers , h. j. g. l. m. 1984 , a&as , 57 , 327 | the influence of the wind to the total continuum of ob supergiants is discussed . for wind velocity distributions with @xmath0 ,
the wind can have strong influence to the total continuum emission , even at optical wavelengths .
comparing the continuum emission of clumped and unclumped winds , especially for stars with high @xmath1 values , delivers flux differences of up to 30% with maximum in the near - ir .
continuum observations at these wavelengths are therefore an ideal tool to discriminate between clumped and unclumped winds of ob supergiants . * wind emission of ob supergiants + and the influence of clumping * + _ * m .
kraus@xmath2 , j. kubt@xmath2 & j. krtika@xmath3 * _ + _ @xmath2astronomick stav , akademie vd esk republiky , ondejov , czech republic _
+ _ @xmath3stav teoretick fyziky a astrofyziky pf mu , brno , czech republic _ 2 |
is a novel transient radio source @xcite , whose notable properties have included `` bursts '' with approximately 1 jy peak flux density lasting approximately 10 min .
each and occurring at apparently regular 77 min.intervals .
we first identified from archival 330 mhz ( 90 cm ) observations taken with the very large array ( vla ) on 2002 september 30 .
we then detected a single burst from the source with the giant metrewave radio telescope ( gmrt ) in archival observations from 2003 september 28 at the same frequency .
is located about 1.25 south of the galactic center just outside ( in angular distance ) the shell - type supernova remnant @xcite . the environment of the source is discussed further in @xcite .
the source is notable because it is one of a small number of _ radio - selected _ transients .
moreover , with only a few exceptions @xcite such as electron cyclotron masers from flare stars and the planets , plasma emission from solar radio flares , pulsar radio emission , and molecular - line masers , most radio transients are incoherent synchrotron emitters . for an incoherent synchrotron emitter ,
the energy density within the source is limited to an effective brightness temperature of roughly @xmath3 k by the inverse compton catastrophe @xcite .
the properties of suggest strongly that its brightness temperature exceeds @xmath3 k by a large factor and that it is a member of a new class of coherent emitters @xcite .
the two previous detections of were based on vla and gmrt 330 mhz observations from two different epochs , but from which similar source properties were observed .
this _ letter _
reports on a third 330 mhz archival observation of , made with the gmrt in 2004 , in which a single , much fainter and shorter burst is detected in contrast to the burst properties observed in 2002 and 2003 .
in addition , the burst is found to have a very steep spectrum , as expected for a coherent emitter , providing another important clue to understanding the nature of this enigmatic source .
possible models for include nearby objects such as a flaring brown dwarf , flare star , or extra - solar planet , although , as discussed in @xcite , the properties of the source do not easily fit these classes .
alternative models proposed for thus far are a nulling pulsar ( @xcite ) , a double pulsar ( @xcite ) , a transient white dwarf pulsar ( @xcite ) , and a precessing radio pulsar ( @xcite ) .
as noted in the latter paper , the double pulsar model predicts a @xmath03-yr bursting activity period which is not supported by the redetections in 2003 and 2004 .
furthermore , in @xcite we presented tentative evidence suggesting that the 2003 burst was an isolated one , and not one of several bursts emitted at 77 min .
intervals , as observed in 2002 .
this evidence could be problematic for some or all of these models .
for example , the proposed double pulsar would link the radio emission to a 77 min .
orbital period . in this model ,
coherent radio emission is triggered by the shock formed through the interaction of the wind of the more enigmatic pulsar with the magnetosphere of the companion pulsar .
on the other hand , in analogy to the psr b1259 - 63 system which consists of a pulsar and be star companion , it is also possible that the magnetosphere of the companion is not constant , and therefore that the radio bursts are not always triggered every orbit . indeed ,
the much fainter detection from reported in this _ letter _ could be evidence of variable conditions in the environment around a companion star . similarly , the precessing radio pulsar and transient white dwarf pulsar models would also require one or more types of nulling effects to explain the occurrence of isolated bursts in the short term , as well as the low duty cycle in the long term .
a few radio pulsars are known to have a very large nulling fraction .
psr b1931 + 24 remains in an off state for @xmath090% of the time , and it emits bursts quasi - periodically at @xmath040 per day @xcite . such a high nulling fraction may be consistent with the measured duty cycle estimated for .
the new , serendipitous detection reported in this paper is derived from 330 mhz gmrt galactic center observations obtained by two of us ( s. roy and s. bhatnagar ) as part of an unrelated project and not included in @xcite .
one of the observations , from 2004 march 20 - 21 , is pointed 0.5 from and consists of eleven 10 min .
scans spread over six hours .
the observations were carried out using the default observing mode with a bandwidth of 16 mhz in each of the two available sidebands .
the sources b1822 - 096 and b1714 - 25 were used as secondary calibrators .
the gmrt does not measure the system temperature ( @xmath4 ) , and the increase in @xmath4 from the calibrator field to the target source affects the source visibility amplitudes in the default observing mode ( i.e. , the automatic level control [ alc ] in the system is turned on ) .
we employed the following method to correct for the @xmath4 variation .
as the system gain does not change with the alc off , we observed 3c48 and b1822 - 096 once with the alc off and determined the flux density of b1822 - 096 to be 10.8 jy using the known flux density of 3c48 from @xcite . also with the alc off , we determined the ratio of the total power on the target source to that of b1822 - 096 to be 1.8 .
since this ratio was quite similar ( within 10% ) for almost all the antennas , rather than correcting the antenna - based gains , we multiplied the final map of the source intensity distribution by this value .
we estimate the overall calibration uncertainty to be within @xmath515% .
the initial images were improved by phase and later amplitude and phase self - calibration .
to produce the final image , separately self - calibrated data from both the sidebands were combined to improve the uv - coverage .
we detect a @xmath010@xmath6 burst from in the middle of the first scan beginning at approximately 21:31:00 ( iat ) on march 20 and lasting for approximately 2 min .
the source is unresolved and has a flux density of [email protected] mjy . as a check on the calibration
, we find the flux densities of several bright sources in the field - of - view to be consistent with those obtained in the 330 mhz vla survey of the galactic center by @xcite . over the past several years
, we have conducted a 330 mhz search for radio transients in the galactic center @xcite using archival observations made between 1989 and the present , and monitoring observations beginning in 2002 .
@xcite summarize the vla and gmrt observations and data reduction through 2005 .
most of these observations are pointed towards sgr a * where the large primary beam at 330 mhz ( 2.2and 1.4 ) and high stellar density subtended optimize the likelihood of detecting a transient , including .
our 2006 and 2007 monitoring program includes observations at 330 mhz ( vla ) and simultaneous observations at 235 and 610 mhz ( gmrt ) ; these will be described in a subsequent paper .
the flux density of the detected 2004 burst is much fainter than the @xmath01 jy peak flux density measured for the five bursts detected in 2002 @xcite and for the one burst detected in 2003 @xcite . due to the relatively short and infrequent scans in the 2004 observations , we detect only a single burst , as in the 2003 epoch .
we can not rule out the possibility that additional bursts were emitted at the same 77 min .
interval observed in 2002 .
we note , however , that nondetections in additional scans made at multiples of 77 min . after the detected 2003 burst , but in which is located well beyond the half power point of the primary beam , suggest that the 2003 detected burst is actually an isolated one @xcite .
figure [ fig : lightcurve04 ] shows the light curves for all seven bursts detected to date . while only the @xmath02 min .
decay portion of the 2003 burst was detected due to its coincidence with the beginning of a scan , the 2004 burst is sampled completely within one scan .
it is detected for only @xmath02 min .
compared with the @xmath010 min .
duration observed for the bursts in 2002 .
is unresolved in all three epochs .
the resolution for the 2004 epoch is approximately 20 @xmath1 10 , comparable to the 2003 epoch , and approximately a factor of 2 better than in the 2002 epoch .
a gaussian fit to the 2004 detection yields a position of ( j2000 ) right ascension @xmath7 @xmath8 509 ( @xmath9 ) , declination @xmath10 09 56.4 ( @xmath11 ) , which is consistent with the 2003 and 2002 positions and is approximately 2.5@xmath1 and 5@xmath1 more accurate , respectively .
the source position and uncertainty cited above include a correction for ionospheric refraction which is prevalent in low frequency observations and discussed in @xcite and @xcite .
separate images were made for the upper ( 333 mhz ) and lower ( 317 mhz ) sidebands of the observations and yield flux densities of [email protected] mjy and [email protected] mjy , respectively .
no significant differences are found in the shapes of the separate light curves generated for each sideband .
figure [ fig : spectrum04 ] shows the spectrum of obtained by imaging pairs of adjacent frequency channels across the two sidebands .
a power - law fit yields a very steep spectrum of @xmath12 for .
an identical analysis of the data for the nearby strong source [email protected] yields a spectral index of @xmath14 , consistent with the determination of @xcite who found a spectral index of @xmath15 between 330 and 1400 mhz .
a monte - carlo simulation was conducted to assess the confidence level of the steep spectrum obtained for the 2004 detection of .
first , a spectrum was generated based on the fitted spectral index of @xmath16 and normalized to the observed flux density at a particular channel .
one thousand spectra were simulated by randomly adding gaussian noise to the flux densities of the normalized spectrum , based on the observed noise level obtained for the individual channel images
. a flat spectrum fit to the data is ruled out with a confidence level of 99.7% .
unfortunately , however , due to limitations in the analysis of the 2002 and 2003 observations , we were not able to detect reliable evidence of a steep spectrum for those bursts .
no emission is detected from when imaging the 2004 observation at times when the burst is not occurring . from these 2004 observations
, we are able to improve the ( 5@xmath6 ) upper limit for quiescent 330 mhz emission between bursts to 6 mjy , as compared to earlier limits of 75 mjy and 25 mjy from the 2002 and 2003 observations , respectively .
the upper limit on quiescent emission during periods of no burst activity is 15 mjy at 330 mhz @xcite .
in addition , we find an upper limit of 50% circular polarization for the 2004 burst , while an upper limit of 15% was previously determined for both the 2002 and 2003 bursts .
in @xcite we crudely estimated the activity of the bursting behavior of by comparing the time during which the source is observed to be active to the total amount of observing time . with the addition of @xmath040 hr of 330 mhz vla observations analyzed in 2006 , with no detection ,
the total observing time is @xmath0120 hr .
the 2002 september 30 - october 1 bursts lasted for at least 6 hr , and since only a single burst was detected on 2003 september 28 and on 2004 march 20 , we assume that the source was active for @xmath01 hr in each of these epochs . thus , has exhibited bursting activity approximately 7% of the time it was observed .
the 2002 and 2003 detections of occured in late september and the 2004 detection occurred in late march , suggesting a roughly 6-month interval between active periods . although we find no detections in our 1998 , 2002 , 2005 , and 2006 observations in march and september , it is still possible that bursting activity did occur in these months but was either too faint or lasted too short a time to be detected .
similarly , while our database has many observations in 2002 , 2003 , and 2006 , we did not monitor for transients nearly as often in 2004 and 2005 , and useful archival data is very limited in those years .
since we have now detected in three significantly different observed states ( @xmath01 jy versus @xmath050 mjy burst strengths , @xmath010 min . versus @xmath02 min .
burst durations , and regularly repeating versus isolated bursts ) , future detections may yet exhibit additional properties ( e.g. , emission in other frequency bands and polarized emission ) that will lead to a definitive understanding of the nature of .
given the high variability in flux density already detected , and the very steep spectrum of the 2004 burst , it is even conceivable that bursts much stronger than @xmath01 jy will be detected at 330 mhz and lower frequencies , although it is also possible that the bursts are now continually decreasing in strength at all frequencies .
we note that all three detections suggest strongly that is a coherent emitter , as first indicated in @xcite , and now even more so by the very steep spectrum reported here .
although the 2004 burst is much weaker than the previous bursts it also appears to have much shorter rise and decay times , @xmath17 , as seen in figure [ fig : lightcurve04 ] .
if we constrain the emitting region to be less than c@xmath17 , with @xmath17 @xmath01-min , then the brightness temperature of is @xmath18 , where @xmath19 is the distance to the source .
if the source is at the galactic center , @xmath08.5 kpc distant , its brightness temperature then still far exceeds the @xmath3k upper limit for an incoherent synchrotron emitter . in summary ,
the new detection of consists of a single @xmath050 mjy burst lasting only 2-min and exhibiting a very steep spectrum ( @xmath20 ) .
the burst is significantly weaker and shorter than the @xmath01 jy and 10-min .
bursts detected in 2002 and 2003 . like the 2003 detection
, the single burst detected in 2004 appears to be an isolated one , although the sparse sampling of the observation does not rule out the possibility that additional bursts were emitted at the same 77 min .
period observed in 2002 @xcite . given the @xmath0120 hrs of 330 mhz observations searched
, we estimate that the source is in a detectable bursting state @xmath07% of the time .
further , multi - wavelength observations are needed to better constrain the physical nature of .
together with continued radio monitoring for transient or weaker pulsed emission ( e.g. pulsar - like ) , additional desired observations include searches for quiescent or variable infrared or x - ray counterparts .
the national radio astronomy observatory is a facility of the national science foundation operated under cooperative agreement by associated universities , inc .
we thank the staff of the gmrt that made these observations possible .
gmrt is run by the national centre for radio astrophysics of the tata institute of fundamental research .
is supported by funding from research corporation and sao _ chandra _ grants go6 - 7135f and go6 - 7038b .
basic research in radio astronomy at the naval research laboratory is supported by 6.1 base funding . | is a transient bursting radio source located in the direction of the galactic center .
it was discovered in a 330 mhz vla observation from 2002 september 30october 1 and subsequently rediscovered in a 330 mhz gmrt observation from 2003 september 28 by hyman et al . here
we report a new radio detection of the source in 330 mhz gmrt data taken on 2004 march 20 .
the observed properties of the single burst detected differ significantly from those measured previously , suggesting that was detected in a new physical state .
the 2004 flux density was @xmath00.05 jy , @xmath010@xmath1 weaker than the single 2003 burst and @xmath030@xmath1 weaker than the five bursts detected in 2002 .
we derive a very steep spectral index , @xmath2 , across the bandpass , a new result previously not detectable due to limitations in the analysis of the 2002 and 2003 observations .
also , the burst was detected for only @xmath02 min .
, in contrast to the 10 min .
duration observed in the earlier bursts . due to sparse sampling , only the single burst was detected in 2004 , as in the 2003 epoch , and we can not rule out additional undetected bursts that may have occurred with the same @xmath077 min .
periodicity observed in 2002 or with a different periodicity .
considering our total time on source throughout both our archival and active monitoring campaigns , we estimate the source exhibits detectable bursting activity @xmath07% of the time . |
the human eye is distinguished from commercially available electronic cameras by virtue of having much better resolution in the fovea near the optical axis .
the eccentricity of the retina , which produces a large field of view and high acuity at the fovea , causes the human visual system to have quick jerky eye movements - i.e. , saccades .
it was shown that an artificial visual system equipped with saccades could reduce the number of processing units required for pattern recognition @xcite and could greatly simplify the calculations in low - level vision systems @xcite .
the saccades involve two subtasks ; i.e. , a location task and an identification task .
the location task fixates the location of a pattern in a wide field of view using _ a priori knowledge of the pattern class of interest . on the other hand
, the identification task determines the class of a pattern under the assumption that the location of the pattern can be fixated .
these subtasks are therefore cooperative .
recently , hopfield and his colleagues have shown that a symmetric interconnected neural network ( a hopfield neural network ) can perform error corrections in associative retrieval @xcite and lead to appropriate solutions in optimization problems @xcite . in this letter
, we treat the saccades as the optimization problem and apply the hopfield network to the emulation of saccades .
computer simulation confirms that the network performs these subtasks cooperatively .
it also suggests that the network is applicable to shift - invariant pattern recognition . _
the network ( shown in fig.[figure01 ] ) comprises three blocks of neurons called the saccades ( s ) block , hidden ( h ) block , and output(o ) block , respectively .
its energy function is defined as follows : @xmath0 where @xmath1 and @xmath2 where @xmath3 is a gaze position , @xmath4 a pixel image of an input pattern , @xmath5 a window function , @xmath6 and activity of the @xmath7th neuron of @xmath8 block , @xmath9 the output of the @xmath7th neuron of @xmath8 block ( @xmath10 ) , @xmath11 the @xmath12th template embedded in the weights between the @xmath13 and @xmath14 blocks , @xmath15 positive constants .
the @xmath16 block is a matrix representation of the location of an input pattern relative to a gaze position represented as the @xmath17th neuron .
the most - activated neuron in the @xmath14 block indicates the class of the input pattern whose location would be represented by the @xmath16 block .
the location and identification tasks are obtained by minimizing the first and second terms in the energy function , respectively .
the third and fourth terms represent winner - take - all constraints in the @xmath16 and @xmath14 blocks , respectively .
a gaze position is updated by the following rule : @xmath18 where @xmath19 is an index of the most - activated neuron in the @xmath16 block . when the gaze position is unchanged by the updating rule , the tasks are complete . and @xmath14 block .
fully interconnected between the @xmath16 and @xmath13 block and between the @xmath13 and @xmath14 block . ]
the first simulation confirmed that the network can perform cooperative subtasks in the saccades .
positive constants in the energy function and the number of neurons of each block were chosen to be @xmath20 , @xmath21 , @xmath22 , @xmath23 and @xmath24 , @xmath25 , @xmath26 , respectively . four templates ( @xmath27 binary - valued pixels ) are depicted in fig.[figure02 ] .
the stable states of the network are illustrated in fig.[figure03](a)-(d ) , where one side of each square represents an output level of the corresponding neuron , an empty circle in @xmath4 a gaze position , and a directed line in the s block a saccadic vector .
it is also seen that the activation pattern of the @xmath28 block resembles the template of the class of the target pattern .
the network was next applied to shift - invariant pattern recognition .
positive constants in the energy function and the number of neurons of each block were chosen to be @xmath29 , @xmath21 , @xmath30 , @xmath31 and @xmath32 , @xmath33 , @xmath34 , respectively . ten templates ( @xmath35 binary - valued pixels )
are depicted in fig.[figure04 ] . in the experiment ,
160 input patterns shifted in any direction by one pixel and deformed by no more than eight hits ( hamming distance ) , including the patterns shown in fig.[figure05 ] , were recognized perfectly .
the behavior of the network is illustrated in fig.[figure06 ] .
figure [ figure06](a ) and ( b ) show the stable states before and after a saccade by one pixel to the upper left , respectively .
we can see the identification task was completed by the saccade that resulted from the location task .
a hopfield neural network for saccades has been proposed .
computer simulations confirmed that the network performs the location and the identification tasks in a cooperative fashion .
it has also been demonstrated that the network can be applied to shift - invariant pattern recognition . since
out network is a standard hopfield - type neural network , it is suitable for vlsi and optical implementations and therefore can speed up saccadic tasks and shift - invariant pattern recognition significantly . used in the first simulation . ]
used in the second simulation . ]
the author thanks the reviewer who originally recommended the publication of this letter and dr . y. tomita of the university of electro - communications for his helpful comments .
ballard , `` behavioral constraints on animate vision , '' image and vision computing , vol.7 , no.1 ( 1989 ) .
j. j. hopfield , `` neural networks and physical systems with emergent collective computational abilities , '' proc .
sci . u.s . , vol.79 ( 1982 ) .
k. nakano , m. niizuma , and t. omori , `` model of neural visual system with self - organizing cells , '' biolog .
cybern . , vol.60 ( 1989 ) .
m. takeda and j. w. goodman , `` neural networks for computation : number representations and programming complexity , '' appl .
vol.25 , no.18 ( 1986 ) .
d. w. tank and j. j. hopfield , `` simple neural optimization networks : an a / d converter , signal decision circuit , and a linear programming circuit , '' ieee trans . circuits and syst . ,
vol.cas-33 , no.5 ( 1986 ) | human eye movement mechanisms ( saccades ) are very useful for scene analysis , including object representation and pattern recognition . in this letter , a hopfield neural network to emulate saccades
is proposed .
the network uses an energy function that includes location and identification tasks .
computer simulation shows that the network performs those tasks cooperatively .
the result suggests that the network is applicable to shift - invariant pattern recognition . |
quantum phase transitions ( qpts ) and related collective quantum phases represent one of the most exciting research topics in condensed matter physics @xcite .
contrary to thermal phase transitions , qpts occur at zero temperature upon tuning a parameter of the system hamiltonian .
the emergence of quantum collective phenomena opens the path towards quantum phases that do not admit any classical counterpart .
quantum magnets provide a large showcase of materials in which qpts have been experimentally demonstrated .
quantum fluctuations , driving the system through a qpt , can be continuously tuned by _
e.g. _ applying a magnetic field or by exerting pressure on the sample to control the magnetic couplings among the spins .
a well - known example is represented by magnetic bose - einstein condensation ( bec ) in spin - gap materials @xcite , such as systems of weakly coupled @xmath0 dimers .
these magnetic insulators show a paradigmatic quantum - disordered ground state , namely a total singlet state with a gap to all triplet excitations .
application of a magnetic field can close the triplet gap , inducing the appearance of bosonic spin triplets in the ground state of the system , forming a bose condensate of magnetic quasiparticles .
this condensed state corresponds to a magnetically ordered state with spontaneous appearance of a staggered magnetization transverse to the field . a completely different route towards qpt transitions in quantum magnets
is represented by the effect of non - magnetic doping .
starting from a magnetically ordered state , a simple route towards disordering the system is by diluting the lattice via site or bond removal . a genuine quantum phase transition is realized if quantum fluctuations lead to the loss of magnetic order at @xmath1 before the lattice reaches the percolation threshold .
experimental studies on model magnets , such as non - magnetically doped la@xmath2cuo@xmath3
@xcite , supported by extensive numerical studies @xcite show that fundamental quantum models like the two - dimensional quantum heisenberg antiferromagnet do not develop strong enough quantum fluctuations for a quantum phase transition to occur . yet , increasing quantum fluctuations _ ad hoc _ , _
e.g. _ by explicit lattice dimerization @xcite or by anisotropic bond dilution @xcite , can lead indeed to disorder - induced quantum phase transitions ( see fig.[f.qperc ] ) . and
what better strategy could one envision to tune quantum fluctuations than to consider a system which already exhibits a quantum phase transition in the clean limit ?
this idea brings naturally to the investigation of the effect of disorder on quantum critical phenomena in magnetic systems , and in particular on magnetic bec as a paradigmatic magnetic quantum phase transition @xcite .
introducing disorder in the magnetically ordered phase immediately close to the condensation qpt corresponds to exposing the dilute gas of spin triplets induced by the field to a random potential : as in any system of weakly interacting quantum particles , this leads to anderson localization for moderate ( or even infinitesimal ) disorder strengths @xcite .
this property translates into the possibility of disrupting the field - induced magnetic order by diluting the lattice well below the percolation threshold , accomplishing in this way a genuine disorder - induced quantum phase transition .
a side effect of doping weakly coupled dimer systems is the appearance of unpaired @xmath0 local moments , which are coupled through virtual excitations of the intact dimers @xcite ( see fig.[f.dimers](a ) ) .
the resulting random network of interacting local moments can support long - range antiferromagnetic order in zero field , giving rise to a counter - intuitive phenomenon of order by disorder ( obd ) .
this ordered state is found to be disrupted by the application of a moderate magnetic field @xcite .
the resulting disordered phase shows short - range antiferromagnetic correlations surviving on localized regions , which can be mapped onto anderson localized quasiparticles .
hence we see in general that disorder - induced quantum phase transitions drive the system to novel quantum disordered phases with rather exotic properties .
a common denominator of these phases is a gapless spectrum associated with the appearance of low - energy localized excitations , associated with exponentially rare regions in the disordered lattice . if the energy cost of the localized excitation decreases algebraically with the inverse size of the region that hosts it , all conventional response functions remain non - singular in this phase down to @xmath1 . on the other hand ,
if the same energy cost decreases exponentially when the size of the region increases , this leads to quantum griffiths singularities @xcite with a non - universal , disorder - dependent divergent behavior in measurable quantities such as the uniform susceptibility . .
( b ) disorder - induced quantum phase transition : at a critical dilution @xmath4 the magnetic order disappears , leaving space to a novel quantum disordered phase.,scaledwidth=80.0% ] in this paper we review our recent progress in studying disorder - induced qpts as well as the effect of disorder on field - induced qpts in two - dimensional quantum magnets .
this topic is particularly challenging from a technical point of view .
indeed conventional perturbation methods are generally doomed to fail at quantum critical points ; renormalization group schemes , while generally successful for the study of critical points in clean systems , prove to be particularly hard for critical points in presence of disorder @xcite .
a third way is provided by numerically exact methods , which are particularly well developed for unfrustrated quantum spin systems . in particular we make use of the stochastic series expansion quantum monte carlo method @xcite which proves to be a very powerful quantum monte carlo technique to investigate the low - temperature properties of quantum spin systems on a large scale .
the non - local nature of the update algorithm makes it possible to circumvent critical slowing down and to investigate quantum critical phenomena in a very accurate way .
while disorder imposes the extra cost of averaging over disorder statistics , this task can be accomplished on modern supercomputers .
the structure of the paper is as follows . in section [ s.bond ]
we show how the interplay between anisotropic geometric randomness and quantum fluctuations gives rise to a new class of percolative quantum phase transitions in an inhomogeneous bond - diluted antiferromagnet in two dimensions ( 2d ) .
this transition opens a novel quantum disordered phase which exhibits quantum griffiths singularities . in section [ s.bg ] , we analyze the very rich phase diagram of site - diluted weakly coupled dimer systems in a magnetic field , showing the occurrence of and extended magnetic bose glass phase . , and
the inter - dimer bonds are shown in light color with coupling @xmath5 . both @xmath6 and @xmath5 are antiferromagnetic .
( b ) site dilution of the dimer singlet ground state releases random magnetic local moments which are coupled through long - range coupling @xmath7 .
the circles refer to the non - magnetic impurities , and arrows correspond to local moments near the impurity sites .
( c ) inhomogeneous bond dilution of the same lattice : the dimer bonds ( @xmath6 ) are populated with probability @xmath8 , while the inter - dimer bonds ( @xmath5 ) are populated with probability @xmath9.,scaledwidth=40.0% ]
our starting point is the quantum heisenberg antiferromagnetic model defined on a regular bipartite lattice , as illustrated in fig .
[ f.dimers](a ) , with hamiltonian @xmath10 where @xmath11 are @xmath0 spin operators . in the most general case ,
two sets of bonds with different local strengths ( @xmath12 and @xmath13 ) have been singled out . in the clean case one
has @xmath14 for all @xmath15 and @xmath16 for all @xmath17 . in the following
we will focus our attention on 1 ) the case of a planar array of dimers in a square lattice ( as specifically illustrated by fig .
[ f.dimers](a ) ) and 2 ) the case of a bilayer system , in which the bonds of strength @xmath5 form two square lattices which are connected by the bonds of strength @xmath6 . in the case 1 ) of a simple square lattice , the choice @xmath18 reproduces the well - known limit of the two - dimensional quantum heisenberg antiferromagnet ( 2dqhaf ) . in this section
we will consider the case of zero applied magnetic field @xmath19 , while the effects of a finite field will be accounted for in the next section . in the clean limit , the 2dqhaf is well known to support long - range nel order@xcite .
as mentioned in the introduction , a central question of quantum magnetism is the response of magnetically ordered systems to the dilution of their magnetic lattice , either in the form of bond or site dilution . from a geometric point of view , bond or site dilution
reduce the connectivity of the lattice , ultimately leading to a percolative phase transition @xcite beyond which the system is broken up into finite clusters . in a classical spin system , this percolation transition is coupled to a magnetic transition with the same critical exponents , since spontaneous magnetic order can not survive beyond the percolation threshold . in a quantum spin system , on the other hand , a progressive reduction of the lattice connectivity enhances quantum fluctuations in a continuous fashion , raising the possibility of quantum destruction of magnetic order _ before _
the percolation threshold is reached .
recently , the evolution of the magnetic state of the 2dqhaf under site or bond dilution has been studied extensively both in experiment and theory .
experimentally , the effect of site dilution has been probed in la@xmath2cu@xmath20(zn , mg)@xmath21o@xmath3 , in which magnetic cu@xmath22 ions are replaced randomly by non - magnetic zn@xmath22 or mg@xmath22 ions @xcite .
the fundamental result of this study is that magnetic order at low temperature disappears at a critical dilution @xmath23 which coincides with the percolation threshold @xmath24 @xcite .
this has been further confirmed by extensive numerical simulations both in the case of site and bond dilution @xcite , showing that the percolating cluster supports long - range order up to the percolation threshold ; this means that magnetism can only be discarded geometrically by fragmenting the percolating cluster beyond @xmath25 .
antiferromagnet on the square lattice .
the colored area indicates the quantum - disordered region in which the system has developed an infinite percolating cluster , but the magnetization @xmath26 of the system vanishes because of quantum fluctuations.,scaledwidth=65.0% ] an alternative scenario to the above classical percolation picture is offered by _ quantum percolation _ ,
in which the geometric transition and the magnetic one are decoupled by quantum fluctuations .
this scenario invokes the fact that spins involved in locally strongly fluctuating quantum states , such as dimer singlets and resonating valence bond states , are weakly correlated with the remainder of the system . in a random network of spins ,
the local strongly fluctuating states create _ weak links _ with small spin - spin correlations
. if these weak links are part of the backbone of the percolating cluster , they can prevent the percolating cluster from developing long - range order . therefore , if lattice dilution favors the local formation of such states , it is possible to drive the system towards a quantum disordered state before the percolation threshold is reached , decoupling percolation from magnetic ordering .
this quantum percolation scenario has been recently demonstrated for the @xmath0 heisenberg model on the bilayer lattice under dimer dilution @xcite . in this
model quantum fluctuations can be arbitrarily tuned by increasing the strength of the inter - layer coupling @xmath6 with respect to the intra - layer one @xmath5 : indeed , even in the clean model , for @xmath27 the spins on each inter - layer dimer form a singlet and long - range order is lost @xcite . randomly diluting the bilayer by taking away a percentage @xmath28 of _ whole dimers _ at once leads to the percolation of the bilayer lattice for @xmath29 , analogously to what happens on a simple square lattice .
for @xmath30 the bilayer shows long - range order up to the percolation threshold @xcite .
the tuning knob of quantum fluctuations offered by the inter - layer coupling allows to destroy long - range order on the percolating cluster for @xmath31 : beyond this value , geometric percolation of the lattice is no longer a sufficient condition for long - range magnetic ordering , due to the extremely strong quantum fluctuations on dimers with lower local connectivity .
the quantum disordered phase appearing on a percolated lattice of dimers still lacks a complete characterization , but it is legitimate to suspect that , for @xmath32 , rare but arbitrarily large regions which are devoid of dimer vacancies can support locally magnon - like excitations similar to those of the clean system in its magnetically ordered phase .
hence the spectrum of the system is expected to be _ gapless _ , but not leading to a singular contribution to the response functions ( as shown by the regular behavior of the uniform susceptibility in ref .
a very similar picture of quantum percolation has been found by two of us in the anisotropic @xmath33 heisenberg antiferromagnet with site dilution @xcite : here the quantum disordered phase has been identified with a _
mott glass _ , namely a phase with a gapless spectrum and a vanishing compressibility . in ref .
@xcite we have shown that a similar quantum percolation scenario can be achieved in the _ standard _
2dqhaf under _
inhomogeneous _ bond dilution . in our model
all bonds are of equal strength @xmath34 , and quantum fluctuations are instead enhanced in a purely _ geometrical _ fashion by the lattice randomness . as illustrated in fig .
[ f.dimers](a ) , a square lattice can be geometrically decomposed into dimers and ladders ( made of inter - dimer bonds ) in such a way that there are no two adjacent dimers or ladders .
the inhomogeneous bond dilution is realized by assigning different occupation probabilities to intra - dimer and inter - dimer bonds .
explicitly @xmath35 @xmath36 as it is well - known , @xmath0 spin ladders have a quantum disordered singlet ground state @xcite , and the same applies to weakly connected dimer lattices , such as the comb lattice .
this special property of the @xmath0 heisenberg model on low - dimensional lattices is the key of our dilution scheme .
in fact , _ inhomogeneous _ bond dilution @xmath37 ( see fig.[f.dimers](c ) ) favors the appearance of ladder segments ( @xmath38 ) or of weakly connected dimers ( @xmath39 ) , both of which have the tendency to support locally a strongly fluctuating quantum state with a significant singlet component , and effectively decoupled from the rest of the percolating cluster .
when such geometrical structures appear on the backbone of the percolating cluster , they can effectively break it into sub - clusters from the point of view of magnetic correlations , leading to a quantum - disordered ground state .
the phase diagram of this inhomogeneous bond dilution model , as resulting from an extensive classical and quantum monte carlo study @xcite , is presented in fig .
[ f.phdbonddilute ] . when the inhomogeneity is weak ( @xmath40 )
, the magnetic transition is found to coincide with the percolation one , both in terms of the location of the transition and in terms of its critical exponents .
this is in agreement with the findings in the homogeneous limit @xmath41 already investigated in ref .
however , for strong enough inhomogeneity the magnetic transition deviates from the geometric one , turning into a quantum phase transition beyond two multi - critical points .
the critical exponents , extracted from a finite - size scaling analysis of the correlation length , confirm this crossover from _ classical _ to _ quantum _ percolation : @xmath42 and @xmath43 ( the fractal dimensions of the percolation cluster at threshold @xcite ) are found for weak to intermediate inhomogeneity , while @xmath44 and @xmath45 are found for sufficiently strong inhomogeneity @xcite .
[ width=5.5cm]diluteladderfig16.eps , in the quantum - disordered ( ladder - like ) regime .
the solid lines are power - law fits of the form @xmath46 , and the resulting fit coefficients @xmath47 are indicated .
the percolating cluster is picked as the largest cluster in a @xmath48 lattice .
the non - universal @xmath47 values indicate the existence of griffiths singularities in this quantum - disordered phase.,title="fig:",width=321 ] as seen in fig .
[ f.phdbonddilute ] , the bifurcation of the magnetic transition line with respect to the percolation transition one at strong enough inhomogeneity opens up two quantum disordered phases , characterized by the existence of an infinite percolating clusters but zero staggered magnetization @xmath26 .
real - space spin - spin correlations in this phase are short - ranged @xcite . but the perfectly correlated disorder in imaginary time allows for long - range correlations in this extra dimension , giving rise to so - called quantum griffiths singularities @xcite . as already mentioned in the introduction , the dilution of the lattice introduces local @xmath0 moments . if one dilutes a lattice developing a spin - gapped singlet ground state with a finite correlation length @xmath49 ( as it will be the case in the following section ) , or if dilution induces the local formation of such singlet states ( as it is the case in this section ) , the resulting local moments develop mutual effective interactions via the exchange of virtual gapped excitations of the intermediate regions .
the effective couplings , @xmath50 , decay exponentially with the distance , as a result of the finite mass " of the exchanged excitation . for two sites
@xmath51 and @xmath52 with positions @xmath53 , @xmath54 , we have that @xcite @xmath55 it is important to notice that the staggering prefactor eliminates any frustration effect . at variance with site dilution ( compare fig.[f.dimers](b ) ) , for bond dilution these local @xmath0 moments always occur in pairs . the effective interaction between two closeby moments is the strongest ( @xmath56 ) , so that the lowest - energy excitations are not associated with exciting a pair of neighboring moments , but rather exciting two pairs of such moments lying far apart from each other by rotating one pair with respect to the other ( see fig . [ f.chiutgriffiths](a)-(b ) ) .
according to eq .
[ e.jeff ] , this excitation has an energy scaling exponentially to zero with the inter - pair distance @xmath57 , but the probability of existence of two pairs separated by a large distance is conditioned by the presence of a clean region in between them , whose probability in a diluted system is also exponentially suppressed with its size . in the case depicted in fig .
[ f.chiutgriffiths](a)-(b ) , this probability is @xmath58 ( where @xmath59 is the characteristic number of bonds present in a region of linear size @xmath57 ) .
the presence of an exponentially rare local excitation with exponentially small energy leads to a cancellation of the two exponentials in the calculation of the contribution of such excitations to fundamental response functions , as , _
e.g. _ , the uniform susceptibility .
the result is a paradigmatic quantum griffiths effect @xcite .
indeed the uniform susceptibility of the percolating cluster in the quantum disordered phase displays a non - universal power - law divergence when @xmath60 , as presented in fig .
[ f.chiutgriffiths](c ) .
it is important to stress that here we only consider percolating clusters with an _ even _ number of particles , which means that the ground state on a finite - size system is expected to be a total spin singlet .
hence the presence of a non - universal divergence of the uniform susceptibility is not due to the presence of paramagnetic dangling spins , but it is a genuine result of the fact that the spectrum over the singlet ground state is gapless . from an experimental point of view , bond dilution can be considered as a limit of bond disorder introduced by doping the non - magnetic atoms / ions lying on the super - exchange paths responsible for the occurrence of antiferromagnetic couplings .
hence inhomogeneous bond disorder can be introduced in antiferromagnets in which super - exchange paths are mediated by different atomic species along different spatial directions .
this is the case of many spin - ladder compounds , such as _ e.g. _
( c@xmath61h@xmath62n@xmath2)@xmath2cubr@xmath3 , in which intra - ladder and inter - ladder couplings between cu@xmath22 ions are mediated by chemically different non - magnetic ions . from a more fundamental point of view , our results , along with those on dimer - diluted lattices @xcite and on diluted anisotropic @xmath33 models @xcite show that disorder - induced magnetic transitions can be pushed arbitrarily far from percolation thresholds .
in the previous section , we have seen that lattice dilution can locally enhance quantum fluctuations in a quantum magnet , possibly discarding long - range order . in this section the critical role of quantum fluctuations will be tuned arbitrarily by driving the system close through a quantum phase transition in presence of disorder .
as already mentioned in the introduction , the quantum phase transition in question is represented by magnetic bose - einstein condensation in a system of weakly coupled @xmath0 dimers in a magnetic field . as in the previous section ,
we consider two different geometries leading to essentially the same physics : a bilayer geometry with interlayer couplings @xmath6 and intra - layer couplings @xmath5 ; a planar dimer array ( compare fig .
[ f.dimers](a ) ) , with dimer couplings @xmath6 and inter - dimer couplings @xmath5 . in absence of lattice dilution , for both geometries a ratio of couplings @xmath63 overcoming a critical value @xmath64 ( @xmath65 for the bilayer system , @xmath66 for the planar dimer system @xcite ) stabilizes a quantum disordered ground state with dimer singlets on the strongest magnetic bonds . in both systems , the application of a magnetic field @xmath67 ( in reduced units ) leads to condensation of triplet excitations at a lower critical field @xmath68 and to the appearance of staggered magnetic order transverse to it @xcite . increasing the field up to an upper critical field @xmath69 leads to full polarization of the spins and to the destruction of spontaneous order . in the field - dilution plane .
ordered phases are indicated in orange , gapped disordered phases are indicated in green , and gapless disordered phases in white and grey.,scaledwidth=65.0% ] doping the system induces a very rich phase diagram at @xmath1 , as illustrated in fig .
[ f.phdbg ] .
if the precise location of the phase boundaries depends on the model parameter @xmath63 as well as on the lattice geometry , the qualitative features of this phase diagram are actually independent of the value of @xmath63 , as long as @xmath70 , and of the specific geometry of the dimer array ( either bilayer or planar array or other unfrustrated lattices ) .
for the specific case shown in fig .
[ f.phdbg ] ( bilayer system with @xmath71 ) , the clean limit @xmath72 features bose - einstein condensation transitions at @xmath73 and @xmath74 . introducing site dilution of the bilayer lattice leads to an upward shift of the lower critical field @xmath75 with respect to its clean value @xmath68 , and a downward shift of the upper critical field .
this means that diluting the system at fixed field starting from its ordered phase @xmath76 leads to a disorder - induced transition into a novel quantum disordered phase .
this all happens well away from the geometrical percolation threshold @xmath25 , which means that the disorder - induced transition is a genuine quantum phase transition .
on intact dimers in a 40x40x2 bilayer with @xmath71 , dilution @xmath77 and at inverse temperature @xmath78 , for @xmath79 ( a ) and @xmath80 ( b ) .
the radius of the dots is proportional to the dimer magnetization .
the magnetization of unpaired spins is omitted for clarity .
the most visible localized states are highlighted in ( a ) , while the backbone of the percolating magnetized network is highlighted in ( b ) .
( c ) superfluid density ( spin stiffness ) as a function of the field for the specific sample considered.,scaledwidth=65.0% ] the novel phase that opens up at the disorder - induced transition can be fully characterized in terms of bosonic quasiparticles . in the case of @xmath81 these quasi - particles ( qps )
are represented by the dilute triplet gas induced by the field , condensing in the magnetically ordered phase .
the case @xmath82 is the ( approximately ) particle - hole symmetric one , in which singlet quasi - holes ( qhs ) form a dilute gas that bose condenses .
the introduction of site dilution in the system creates an effective random potential for qps ( analogous considerations apply to qhs ) .
indeed site dilution leads to the disappearance of whole dimers hosting a qp , or to the appearance of dangling spins ( local moments ) which are essentially all polarized in the field range @xmath76 , and hence act as impenetrable barriers to qps ( see below for a detailed discussion of the physics of local moments ) . therefore the problem of the response of the magnetic system to site dilution is analogous to that of the evolution of a bose - einstein condensate upon increasing the strength of a random potential in which the condensate is immersed . for a strong enough random potential
the bosonic system will be _ fragmented _ into anderson - localized states , sitting in the rare regions which are devoid of vacancies .
this phase is called _ bose glass _ in the literature of disordered interacting bosons @xcite .
given that we are effectively working in the grand - canonical ensemble , the transition from a condensed phase to a bose glass is accompanied by a decrease in the population , due to the fact that disorder lowers dramatically the local chemical potential in the regions close to impurities .
this same localization - condensation transition can be probed by varying the field at fixed disorder strength : in this case the transition is clearly understood as a reduction of the density of the qps / qhs , induced by the decrease / increase of the magnetic field . when traversing this transition in the reversed sense , by driving it with either a field or with disorder , we observe that the region occupied by the qps / qhs undergoes a quantum percolation transition from a localized disordered phase , in which only disconnected rare clean regions host qps / qhs , to a percolated ordered phase , in which the regions hosting qps / qhs connect to form a percolating network .
this geometric transition can be directly visualized by calculating the local magnetization profile , as shown in fig .
[ f.lmbg ] .
the novel , percolative nature of this transition with respect to the condensation in the clean case reveals itself in the critical exponents , which can be determined numerically via finite - size scaling @xcite .
the resulting values are quite different from those of the transition in the clean system , and in particular it is found that the dynamical critical exponent @xmath83 equals the spatial dimension @xmath84 , which is consistent with an early theoretical prediction @xcite . and @xmath85 , from roscilde _ et al .
_ , @xcite.,scaledwidth=65.0% ] from the point of view of macroscopic observables , the bose glass is fundamentally dominated by the absence of a gap in its excitation spectrum . indeed in this phase the system is fragmented into droplets of qps / qhs that are hosted on rare , but arbitrarily large regions , and which can therefore support arbitrarily low - energy excitations , in the form of localized magnon excitations analogous to those appearing in the ordered phase in the clean system .
this means that that the system displays a finite response to an applied magnetic field , namely its uniform susceptibility remains finite ( corresponding to a finite compressibility of the qps / qhs ) even in absence of long - range coherence .
this is clearly illustrated in fig .
[ f.fsbg ] .
so far we have worked under the assumption that dangling spins left unpaired by site dilution are fully polarized by the field .
actually the evolution of their magnetic state represents another very fascinating aspect of the physics of these systems .
indeed , as already mentioned in the previous sections , in zero field dilution liberates local @xmath0 moments ( lms ) which are exponentially localized close to the site of an unpaired spin and which interact with each other via the effective couplings of eq .
[ e.jeff ] .
these couplings , although weak , are sufficient for the lms to order antiferromagnetically at experimentally relevant temperatures @xcite . hence
, at any finite dilution concentration smaller than the classical percolation threshold , the zero - field ground state displays long - range antiferromagnetic order ( order - by - disorder phase ) .
yet the long - range order appearing in this phase can be easily destroyed by a small field in a quite peculiar way @xcite .
indeed the system of lms features highly inhomogeneous couplings , and hence the response to the field is equally diverse . in the case of low dilution ( to which this discussion is restricted )
most of the lms are weakly coupled to the other nearby lms which sit at a distance of the order of the average inter - vacancy spacing @xmath86 ; but a minority of the lms might be involved in a lm dimer ( with probability @xmath87 ) , in a trimer ( with probability @xmath88 ) , and so on .
this means that when the field is strong enough to overcome the _ typical _ coupling energy between two lms , it polarizes a majority of lms destroying thereby their long - range magnetic order . yet , in this _ disordered lm _ phase , a minority of strongly coupled lms resist polarization , and they host locally one or more spin which is polarized opposite to the field . in particular , a mapping of the localized @xmath0 spins which remain antiparallel to the field onto hardcore bosons , reveals for the disordered lm phase a clear nature of a bose glass , characterized by fragmentation of a bose gas into disconnected localized states .
this can be quantitatively confirmed by a finite - size scaling study of the transition between the order - by - disorder phase and the disordered lm phase : the extracted critical exponents turn out to be fully consistent with those of the 2d superfluid - to - bose - glass transition @xcite . in particular , as in a bose glass , the disordered lm phase is not fully polarized , and it remains _ gapless _ , given that all different clusters of strongly coupled lms see a different magnetic environment and hence exhibit a different local gap to further polarization .
nonetheless the local gaps are distributed around some distinct values associated with lm dimers , trimers , etc .
these values manifest themselves in a special form of the magnetization curve , exhibiting _ pseudo - plateaus _ which mark the accomplished polarization of one class of lm clusters . and
dilution @xmath89 , showing a characteristic succession of pseudo - plateaus.,scaledwidth=65.0% ] if the coupling ratio @xmath63 is large enough , or the disorder concentration is weak enough , upon further increasing the field the lms can be fully polarized at a field @xmath90 , namely before the magnetization process of the intact dimers begins . under this assumption , a gapped _
phase ( characterized by a marked plateau of the uniform magnetization at the value @xmath91 ) appears , completely separating the physics of the lms from that of the intact dimers . on the other hand , for smaller @xmath63 ratios or strong dilution the plateau phase might be completely washed out , and dimers start being magnetized in the bose glass phase when lms have not yet been fully polarized .
in this case disordered lm phase and bose glass phase coexist : indeed the bose glass phase will show both localized triplet qps in rare regions devoid of impurities , and localized anti - parallel spins in the complementary regions exhibiting clustering of impurities .
in conclusion , large - scale quantum monte carlo simulations give access to the vast wealth of novel quantum phases appearing in quantum magnets when quantum fluctuations are tuned to a critical strength in presence of disorder in the magnetic lattice .
these novel phases are the result of the separation , induced by quantum fluctuations , between the geometric percolation of the lattice and the loss of long - range magnetic order .
their common denominator is a rich structure of the low - energy spectrum , with absence of a gap and a possibly anomalous role of the low - energy excitations in the response functions .
this rich physics enjoys the availability of a large family of @xmath0 compounds exhibiting strongly fluctuating quantum magnetism , and which are amenable to dilution of the magnetic lattice by non - magnetic doping . in particular candidate systems
are spin - ladder materials such as cu@xmath62(c@xmath61h@xmath62n@xmath2)@xmath2cl@xmath3 @xcite and ( c@xmath61h@xmath62n@xmath2)@xmath2cu br@xmath3 @xcite , planar dimer systems such as sr@xmath2cu(bo@xmath92)@xmath2 @xcite , magnetic bilayer compounds as bacusi@xmath2o@xmath93 @xcite . in all these compounds the @xmath0 spins
is carried by a cu@xmath22 ion , which can be replaced by non - magnetic mg@xmath22 and zn@xmath22 to give site dilution of the lattice .
alternatively bond disorder can be realized by chemical substitution of the non - magnetic ions bridging the magnetic couplings @xcite . for
what concerns specifically the physics of magnetic bec in presence of disorder , analogous phenomena to those observed in @xmath0 compounds can be recovered also with @xmath33 haldane chains , or with @xmath33 antiferromagnets with strong single - ion anisotropy @xcite .
we therefore believe that quantum magnets under static doping offer the possibility of deepening tremendously our understanding of quantum many - body physics in a controlled random environment .
we thank omid nohadani for useful discussion . this work was supported by doe through grant no .
de - fg02 - 06er46319 .
the numerical computations were carried out on the university of southern california high - performance computer cluster .
s. l. sondhi , s. m. girvin , j. p. carini , and d. shahar , rev .
1997 , * 69 * , 315 .
giamarchi t. , regg c. , tchernyshyov o. , nat .
phys . , 2008 , * 4 * , 198 .
vajk o.p . , _
et al . _ , science , 2002 , * 295 * , 1691 .
sandvik a.w .
b , 2002,*66 * , 024418 .
vajk o.p .
, greven m. , phys .
lett . , 2002 , * 89 * , 177202 .
sandvik a.w .
lett . , 2002 , * 89 * , 177201 .
sknepnek r. , vojta t. , vojta m. , phys .
lett . , 2004 , * 93 * , 097201 .
yu r. , roscilde t. , haas s. , phys .
lett . , 2005 , * 94 * , 197204 .
roscilde t. , haas s. , phys .
, 2005 , * 95 * , 207206 .
roscilde t. , haas s. , j. phys .
b 2006 , * 39 * , s153 ( 2006 ) .
yu r. , roscilde t. , haas s. , new j. phys .
, 2008 , * 10 * , 013034 .
lee p. a. , and ramakrishnan t. v. , rev .
phys . 1985 , * 57 * , 287 .
sigrist m. and furusaki a. , j. phys .
soc . jpn .
1996 , * 65 * , 2385 .
mikeska h .- j . , ghosh a. , and kolezhuk a. k. , phys .
2004 , * 93 * , 217204 .
griffiths r. b. , phys .
1969 , * 23 * , 17 ; mccoy b. m. , phys . rev . lett .
1969 , * 23 * , 383 ; mccoy b. m. , phys . rev . 1969 , * 188 * , 1014 .
weichman p. b. , arxiv:0810.3263 2008 .
syljusen o.f .
, sandvik a.w .
e , 2002 , * 66 * , 046701 .
manousakis e. , rev .
phys . , 1991 , * 63 * , 1 .
stauffer d. , aharony a. , introduction to percolation theory .
taylor and francis , london , 1994 .
sandvik a. w. and scalapino d. j. , phys .
1994 , * 72 * , 2777 .
roscilde t. , haas s. , phys .
lett . , 2007 ,
* 99 * , 047205 .
e. dagotto and t. m. rice , science 1996 , * 271 * , 618 .
yu r. , roscilde t. , haas s. , phys . rev .
b , 2006 , * 73 * , 064406 .
matsumoto m. _ et al .
_ , phys . rev .
b 2001 , * 65 * ,
014407 . fisher m.p.a .
b , 1989 , * 40 * , 546 . azuma m. _ et al . _ ,
phys . rev .
b , 1997 , * 55 * , r8658 . chaboussant g _ et al . _ , eur . j. phys .
b 1998 , * 6 * , 167 .
watson b. c. _ et al .
rev . lett .
2001 , * 86 * , 5168 .
s. e. sebastian _ et al .
b 2005 , * 71 * , 212405 .
s. e. sebastian _ et al .
b 2005 , * 72 * , 100404 .
oosawa a. , tanaka h. , phys .
b , 2002 , * 65 * , 184437 .
masuda t. , zheludev a. , uchinokura k. , chung j.h .
, park s. , phys .
lett . , 2004 , *
93 * , 077206 .
a. paduan - filho _ et al .
b 2004 , * 69 * , 020405 ; v. zapf _ et al .
_ 2006 , phys .
lett . * 96 * , 077204 . | quantum magnets represent an ideal playground for the controlled realization of novel quantum phases and of quantum phase transitions .
the hamiltonian of the system can be indeed manipulated by applying a magnetic field or pressure on the sample . when doping the system with non - magnetic impurities , novel inhomogeneous phases emerge from the interplay between geometric randomness and quantum fluctuations .
in this paper we review our recent work on quantum phase transitions and novel quantum phases realized in disordered quantum magnets .
the system inhomogeneity is found to strongly affect phase transitions by changing their universality class , giving the transition a novel , quantum percolative nature .
such transitions connect conventionally ordered phases to unconventional , quantum disordered ones - quantum griffiths phases , magnetic bose glass phases - exhibiting gapless spectra associated with low - energy localized excitations .
heisenberg antiferromagnets , quantum disorder , geometric randomness , percolation , bose glass |
many solar filaments ( or prominences when they are observed above the solar limb ) end their life with a sudden rapid rise called an eruption .
sometimes a filament rises like an enlarging loop lying in a plane containing filament endpoints anchored in the chromosphere and the centre of the sun @xcite .
a famous example suggestive of such behaviour is the eruption of the giant prominence on 28 june 1945 ( granddaddy ) observed at the high altitude observatory .
some eruptive prominences deviate significantly from this plane and move in a non - radial direction @xcite
. moreover , the loop may be not flat but the apex exhibits writhing motion as it rotates about the direction of ascent @xcite .
there are also partial filament eruptions when only a section of a long filament starts to ascend while other parts of the filament are observed unchanged @xcite .
usually a prominence stretched along the limb consists of a number of arches with feet ( called also barbs ) connected to the chromosphere like a long road bridge with several spans . after the start of ascending ,
the feet successively break one after another except the filament endpoints for a full eruption or feet of undisturbed filament sections for a partial eruption .
of course , in real solar environment , events are often asymmetric .
one leg of an eruptive prominence may be fixed to the chromosphere at the prominence endpoint while the other point connecting the prominence with the chromosphere changes its position following after successive breaking of intermediate feet .
@xcite identified two types of asymmetric filament eruptions : whipping - like , where the active leg whips upward , occasionally extending high into the corona ; and zipping - like , where the visible end of the active leg moves along the polarity inversion line ( pil ) like the unfastening of a zipper .
it should be noted that the visibility of a filament depends on the loading of its magnetic skeleton with dense plasma . during the asymmetric filament eruption ,
the active leg can either whip upward , if it is anchored at the location where the eruption initiates ; or zip away from the visible end of the active leg , where the eruption initiates , toward the invisible end of the active leg .
the invisible end later becomes visible during the zipping process with mass draining down along axial filament field lines .
image at 16:15 ut ( courtesy of the meudon observatory ) .
the red line shows the position of the east limb for _ stereo a _ , the green line indicates the position of the west limb for _ stereo b_. the green circle shows the final position of the new filament end region before the filament faded .
bottom : h@xmath0 image at 16:34 ut ( courtesy of the kanzelhoehe solar observatory ) overlaid with line - of - sight field strength from _
soho_/mdi ( only field strengths @xmath1 gauss are shown ; red negative ; blue positive ( courtesy of the _ soho_/mdi consortium ) ) .
, width=317 ] while , falling back to the chromosphere , eruptive prominence material can move along pre - existing magnetic flux tubes , possibly highly stretched and deformed , which belong to the prominence magnetic skeleton , there are also indications that sometimes prominence plasma returns to the chromosphere along trajectories that were formed by reconnection of the prominence magnetic field with the ambient coronal magnetic field .
@xcite reported an explosive filament eruption on 13 july 2004 in which one part of the ejecta escaped as a cme , and another one fell back onto the sun .
the latter part consisted of fragments of the filament dispersed into a cloud covering almost the whole nw quadrant of the solar disk .
obviously , some areas where filament fragments landed had not been connected by field lines with the filament . on 7 june 2011 ,
an active region filament near the west solar limb rose and erupted , hurling an enormous amount of material into the solar atmosphere @xcite .
the diagonal scale of the ejecta appears at least an order of magnitude larger than the initial foot - point separation , and suggests the filament carried a very large amount of mass .
a significant fraction of the prominence mass was observed falling back to the solar surface along newly reconfigured magnetic field lines .
@xcite consider this event as clear evidence that large - scale re - configuration of the coronal magnetic field takes place during solar eruptions via the process of magnetic reconnection .
@xcite reported a disk event on 25 october 1994 , which provided evidence for a large - scale magnetic reconnection occurring between the expanding twisted loops and overlying transequatorial loops that interconnect quiet solar regions . in this work , we analyze observations of the eruptive filament on 30 april - 1 may 2010 , which shows reconnection of one filament leg with a region far away from its initial position .
in contrast to above - mentioned examples of the coronal reconnection , the erupting filament loop does not disintegrate but keeps its shape of a rather thin loop even after a fast jump of the endpoint over a distance of 0.7 solar radius @xmath2 .
we calculated parameters of coronal potential magnetic field and found that the eruption began with instability not in the vertical direction , as typical for eruptive filaments , but after violation of the horizontal equilibrium .
a quiescent filament , located close to noaa active region 11064 to the north - west from it ( fig .
1 ) , started to rise rapidly after 23 ut on 30 april 2010 .
the eruption was observed on the disk in h@xmath0 line at the mauna loa solar observatory with a cadence of 3 minutes ( fig .
2 ) and culgoora solar observatory with a cadence of 1 minute as well as on the eastern limb by the _ solar terrestrial relations observatory - ahead ( stereo a ) _ and close to the western limb by _ stereo b ( behind ) _ with a cadence of 10 minutes ( fig . 3 and fig .
the event was also observed on the disk with the _
proba2_/swap euv solar telescope in a spectral bandpass centered on 174 with a cadence of about 1 minutes @xcite . at first
, it looked like a typical eruption of a filament showing an expanding loop with anchored endpoints ( movie 1 ) . at 23:46 ut
, the top of the loop folded over in images obtained by sun earth connection coronal and heliospheric investigation ( secchi ) euvi @xcite onboard _ stereo a _ , showing writhing of the filament axis .
the apex of the eruptive filament ( prominence ) deflected to the south during the ascending . at 00:06 ut , it was over the southern endpoint of the prominence .
this endpoint became wider and consisting of several strands of threads after 00 ut . the southernmost strand faded out after 00:30 ut , while the northern strand became narrower with fine threads .
the size of each frame is .5 @xmath2 @xmath3 .5 @xmath2 .
( courtesy of the mauna loa solar observatory , operated by the high altitude observatory , as part of the national center for atmospheric research ( ncar ) .
ncar is supported by the national science foundation.),width=317 ] the behaviour of the northern endpoint of the filament was more dramatic .
before the eruption , the filament axis entered the chromosphere at this endpoint inclined slightly to the south from the vertical . from 23:30 ut to 00 ut , the axis was nearly vertical .
then , it inclined to the south . about 00:30 ut , the northern leg of the erupting prominence loop jumped laterally to the position further south than the southern endpoint of the prominence .
the loop proceeded with expanding and ascending , however the endpoints reshuffled their positions in _ stereo a _ images .
the originally southern endpoint is now on the north , while the originally northern endpoint is on the south . further evolution of the eruptive prominence in _ stereo a _ images looks like typical prominence eruption with anchored in the chromosphere endpoints , if one forgets that the endpoints have exchanged the roles .
the top of the prominence leaved the _
stereo a _ field - of - view at 00:40 ut , and after 01:40 ut all the structure faded and flew away .
after half an hour acceleration ( @xmath4 23 ut - 23:30 ut ) , the top of the prominence rose with an approximately constant speed of 90 km s@xmath5 ( fig .
@xmath3 1.1 @xmath2 .
( courtesy of the _ stereo_/secchi consortium , nasa / sdo and the aia science team.),width=317 ] the eruption of the prominence was followed by a coronal mass ejection ( cme ) , which was observed from all three viewpoints ( fig .
6 ) . due to the geometrical factor
, it was brightest in the field - of - view of the _ stereo a _ cor2 coronagraph ( howard et al . , 2008 ) .
at the late stage , the legs of the cme connecting the core with the sun show noticeable twisted structure ( not visible in the difference image in fig . 6 ) . the large angle and spectrometer coronagraph ( lasco ) c2 @xcite onboard the _ solar and heliospheric observatory ( soho ) _ and the _ stereo b _ cor2 coronagraph registered only faint features , which became perceptible enough only in difference images .
the cme propagated in the see direction for _ stereo a _ and soho , while in the sww direction for _ stereo b_. the bright core in the _ stereo a _
cor2 field - of - view moved with a speed of about 100 km s@xmath5 . the frontal cme structure , of course , moved faster . according to the _ soho_/lasco cme catalog ( http://cdaw.gsfc.nasa.gov/cme_list/ )
, the cme appeared first in the field - of - view of c2 at 07:12 ut on 1 may at a polar angle of 111@xmath6 , had an angular width of 108@xmath6 , and reached the final speed of 380 km s@xmath5 . in h@xmath0 images obtained with groundbased telescopes ,
the filament rotates counterclockwise about its southern end point like a nearly straight structure ( fig .
2 , see also movie 2 ) .
it passed over ar 11064 in projection and faded when it stretched in nearly longitudinal direction .
two short faint flaring ribbons appeared at 23:50 ut at the location of the most curved and widest section ( an elbow in fig .
1 ) of the pre - eruptive filament ( fig .
2 ) . brightening below the ascending filament in _ stereo b _
images arose at 23:36 ut soon after the start of the eruption ( fig .
4 , movie 3 ) .
later on , two discontinuous ribbons became very prominent in images and a system of post - eruptive loops was observed in 171 channels of both _ stereo a _ and _ stereo b _ spacecraft .
the beginning of the event in _ stereo b _ observations also looks like a typical filament eruption .
between 00:26 ut and 00:36 ut , the attachment of the filament to the northern endpoint tore , the filament end jumped a large distance both in latitude and longitude , and found new more or less stable position in projection on the disk . in order to understand , what happened with the connectivity of the filament , what was a real displacement , and what was a projection effect , let us examine in details the position of the filament endpoints .
we used images in jpeg format from websites of the _ stereo _ mission and ground observatories .
all images are rotated in such a way that the projection of the solar rotation axis is vertical and the heliographic north is on the top of the images . in the vertical plane containing the line of sight
, the rotation axis is inclined by an angle @xmath7 from the sky plane .
usually @xmath7 is indicated as the heliolatitude of the centre of the solar disk in an image of the sun .
we will use the heliocentric earth equatorial coordinate system ( heeq ) with @xmath8 directed along the solar rotation axis , @xmath9 pointed to the intersection of solar equator and solar central meridian as seen from earth , and @xmath10 completing the right - handed system .
after measurements of the coordinates @xmath11 ( horizontal ) and @xmath12 ( vertical ) relative the to disk centre in units of the solar radius @xmath2 ( since the scale is different in different observational data ) , we should rotate the @xmath13 coordinate system by the angle @xmath7 around the @xmath14 axis and transform the cartesian coordinates @xmath15 into spherical heliocentric coordinates @xmath16 and @xmath17 assuming that all points of interest are located on the spherical surface @xmath18 .
@xmath19 @xmath20 @xmath21 @xmath22 @xmath23 the results of measurements are shown in fig . 7 and fig
the red curves and symbols correspond to the southern endpoint , the blue curves and symbols correspond to the northern endpoint .
triangles ( squares ) show data from _ stereo a ( stereo b ) _ images , circles represent data from h@xmath0 observations . in general , data from all points of view are in good agreement with each other . uncertainties and discrepancies appear due to difficulties of identification of the same features in different projections because of the complicated internal structure of the filament .
in particular , bifurcation of the southern filament leg leads to appearing of an additional ( higher - latitude ) branch in the southern endpoint position data of _ stereo a _
( fig . 7 ) . on the late phase of the eruption
, the originally southern endpoint can not be recognized in _ stereo - b _ images . only material projected on the sky plane above the limb as a prominence indicates the position of the originally southern section of the filament .
the lowest part of the prominence , indicated by the magenta arrow in fig .
9 , is the nearest to the southern endpoint visible section of the filament . the magenta curve in fig . 7 shows the latitude of the crossing point of the prominence leg with the _ stereo - b _ limb . due to the curved shape of the erupting filament , the latitude of the crossing point is greater than the latitude of the southern endpoint derived from _ stereo a _ observations ( short red curve with triangle symbols in fig . 7 ) but the difference is not very significant . there is a wide jump of more than 30@xmath6 in the northern endpoint latitude during changing of its connectivity. longitude of the northern endpoint also changes about 30@xmath6 ( fig .
. the exact position of this endpoint in longitude is less definite than in latitude because the northern section of the filament has nearly longitudinal orientation ( fig .
1 ) , while this section is rather faint in h@xmath0 images and its visibility varies in time ( compare fig . 1 and fig .
this is the main reason for a more steep decrease of the northern endpoint longitude derived from the h@xmath0 data .
we fix the visible northern end of the filament in h@xmath0 images , however this is not the endpoint of the flux rope anchored in the photosphere but the middle part of the moving flux rope with the northern endpoint still anchored at the same place ( compare fig . 2 or better movie 2 with the difference images at 23:56 ut - 00:06 ut , 00:26 ut - 00:16 ut in fig .
3 , fig . 4 and movie 3 ) . formulas ( 4 ) - ( 5 ) and results shown in fig . 7 and fig .
8 assume that selected points are located on the solar surface .
this is correct for the southern endpoint which is observed connected with the chromosphere on the limb till the last phase of the event .
situation with the northern endpoint is more intricate .
we have no information about the height of the filament loop end above the chromosphere observed by _
stereo b _ .
since the final position of the moving loop end is not far from the disk centre in the _ stereo b _ images , the real coordinates more or less correspond to the values shown in fig . 7 and fig . 8 .
we can only estimate an upper bound on the height of the filament end point seen by _
stereo b_. the difference _
stereo a _ image in fig .
9 shows that at 01:06 ut the former northern ( now eastern ) end point must have a height lower than the line of sight tangent to the limb for _ stereo a _ because the prominence material extends down to the limb in the _ stereo a _ view ( see also movie 3 ) . at this time , the eastern ( former northern ) endpoint in _ stereo b _
images finds its final position .
@xmath3 .7 @xmath2.,width=317 ] the height of the line - of - sight above a point on the solar surface with coordinates @xmath16 and @xmath17 ( or the length of the radial segment from a given point to the intersection with a tangent to the limb ) is @xmath24 where @xmath25 is the angle between the radius passing through the given point and the sky plane .
this angle is given by @xmath26 for _ stereo a _ images , we should use a coordinate system related to them . for this purpose ,
the heeq system should be rotated by the separation angle between the earth and _ stereo a _ of 70@xmath6 around the @xmath8 axis and by the angle of inclination of the solar rotational axis from the sky plane @xmath7 = [email protected] around the @xmath10 axis . according equations ( 1 ) - ( 5 ) @xmath27 @xmath28 the approximate coordinates of the eastern endpoint @xmath16 = 8@xmath6 ,
@xmath17 = -52@xmath6 in the heeq system transform to @xmath29 = 10@xmath6 , @xmath30 = -121@xmath6 .
then , @xmath31 = 31@xmath6 and @xmath32 = 116 mm .
this height is much less than the height of the prominence top of about 500 mm at this time ( fig .
therefore , the shape of the prominence axis at the later stage of the eruption is a loop with low endpoints .
the eastern ( former northern ) filament extremity is located definitely below @xmath4 100 mm and possibly finds connection with the chromosphere at a new point after coronal reconnection .
it is widely accepted now that filaments represent cold dense plasma contained within magnetic flux ropes embedded into coronal magnetic field . a flux rope can exist in equilibrium in an ambient magnetic field for a rather long time before an eruption .
@xcite showed first that there is a critical height for stable flux rope equilibria above which the background coronal magnetic field decreases faster than the inverse height .
the transition from stability to instability was named later catastrophic loss of equilibrium and was assumed to be the cause of sudden eruptive events @xcite .
@xcite modeled a flux rope with the magnetic field created by a straight line current .
if a flux rope is curved , an additional force called the lorentz self - force or hoop force is present @xcite .
it is directed away from the curvature centre . in the presence of an ambient magnetic field
, the curved flux rope can be both in stable or unstable equilibrium depending on properties of the external field .
@xcite called the related instability torus instability and showed , following @xcite , that it occurs when the background magnetic field decreases along the major radius @xmath33 of the expanding flux rope faster than @xmath34 . @xcite
carefully compared the two types of models and came to conclusion that the same physics is involved in the instabilities of circular and straight current channels .
the stability of the flux - rope equilibrium in both models depends on the rate of the background field decrease , quantified by the so - called decay index , @xmath35 where @xmath36 is the horizontal magnetic - field component perpendicular to the flux rope axis and @xmath32 is the height above the photosphere .
@xcite pioneered in applying this index to the analysis of filament stability .
@xcite found that for the typical range of current - channel thickness expected in the corona and used in many mhd simulations @xcite , and for a current channel expanding during an upward perturbation , a critical decay index @xmath37 has similar values for both the circular and straight current channels in the range 1.1 - 1.3 . to analyze the equilibrium conditions of the flux rope associated with the filament we calculated the shape of pils and the distribution of the decay index of the potential magnetic field at different heights in a region surrounding the filament ( see @xcite for details ) .
figure 10(a ) represents a fragment of the magnetogram taken by the michelson doppler imager ( mdi ; @xcite ) onboard _ soho _ on 30 april 2010 at 22:27 ut , which was used as a boundary condition for the potential magnetic field calculations . in fig .
10 , thin lines show isocontours of @xmath38 = 0.5 , 1 , 1.5 , while thick red lines indicate the positions of pils at respective heights . areas , where @xmath39 , are shadowed .
a pil is a place where a coronal electric current ( a flux rope ) can find the horizontal equilibrium . the thick green line in fig .
10(b ) shows the section of the pil occupied by the filament as it is seen in h@xmath0 line .
the filament lays over the pil separating the area of predominantly negative network polarity in the upper right corner of the frame from active region 11064 near the centre of the frame .
the left part of the magnetogram contains a mixture of faint small - scale opposite polarities producing numerous closed pil contours ( fig .
10(b ) - 10(e ) ) . the height of the prominence in a stable state before the eruption is 17 mm ( fig . 5 ) . at this height ,
the filament is within the area of stability @xmath40 ( figures 9(b ) - 9(c ) ) .
the location of the filament is stable in the vertical direction even at much greater heights , but at a height of 20 mm the pil occupied by the filament touches the pil surrounding the negative polarity of ar 11064 and reconnects with it .
a newly formed pil ( fig .
10(d ) ) protrudes far to the east from the initial filament position .
since the flux rope can find the horizontal equilibrium only on the pil , it will be pushed to the east by the lorentz force , if it reaches a height greater than 20 mm . moreover , some sections of higher - altitude pils are located within unstable ( grey ) areas .
so , the flux rope , if it arrives upon these sections , will be forced to rise .
pils associated with small - scale magnetic sources disappear above a height of 60 mm .
only a small annular pil in the centre of the domain remains at a height of 140 mm and it also disappears at a height of 160 mm .
it is widely believed that the most probable initial magnetic configuration , which accumulates dense prominence plasma and later hurls a cme into the interplanetary space , is a flux rope consisted of helical field lines @xcite .
an alternative configuration of a cme source region is a sheared arcade @xcite , which is converted into a flux rope structure due to reconnection in the course of the eruption .
usually , initiation of filament eruptions is associated with the instability of the flux - rope equilibrium in the vertical direction .
our prominence was rather low before the eruption , and it starts to ascend rapidly from a height of about 20 mm .
the maps of the decay index distribution ( fig .
10 ) show that the vertical equilibrium of a flux rope at the location of the filament can be stable even at several times greater heights .
however , the topology of the magnetic field varies rapidly with height . at a height of about 20 mm the pil , where the flux rope sits , touches the pil surrounding the negative polarity of ar 11064 and reconnects with it .
the horizontal equilibrium of the flux rope can easily be disturbed due to the proximity of another pil . in principle
, it could find new stable equilibrium over the changed pil , but some sections of this pil are unstable for vertical displacements , which leads to the filament eruption .
this is unusual initiation of an eruption with combination of initial horizontal and vertical flux - rope displacements , to the author s knowledge , never mentioned before .
initial motion is slow , so it is not so easy to catch the beginning of the displacements in horizontal and vertical directions . taking into account changes in seeing conditions , some lateral displacement of the filament can be recognized between 22:40 ut and 22:58 ut in mlso h@xmath0 images .
however , internal motions within the filament can mask or emulate the real displacement of the filament as a whole .
the same changes can be found in _ proba2_/swap 174 images .
small changes in a height of the prominence are noticeable between 22:36 ut and 22:56 ut in _ stereo a _ 304 images .
it is also difficult to distinguish slow prominence rising from effects related to the solar rotation and prominence shape changes .
thus , it can be only stated that both motions start practically simultaneously .
we can compare finite displacements of the filament in horizontal and vertical directions in the beginning of the event over a period of time from 23:06 ut to 23:46 ut . since the contrast of the filament is low in the h@xmath0 images ( fig .
2 ) , we draw in fig .
11 filament spines with the help of the time sequence ( movie 2 ) .
the horizontal distance is about 150 mm , while the change in height as evident from fig .
3 and fig . 5 is about 150 mm too .
so , the deviation of the trajectory of the eruptive filament from the vertical line is near 45@xmath6 .
image at 23:06 ut . the blue arrow shows the horizontal displacement of about 150 mm .
the size of the frame is .5 @xmath2 @xmath3 .5 @xmath2 .
( courtesy of the mauna loa solar observatory , operated by the high altitude observatory , as part of the national center for atmospheric research ( ncar ) .
ncar is supported by the national science foundation.),width=317 ] the behaviour of the filament endpoints is also unusual .
the northern end of the filament during the eruption suddenly changes its anchoring to a position about 40@xmath6 from the starting position .
the process of the filament endpoint change does not look very similar to whipping - like or zipping - like asymmetric filament eruptions .
the latter would imply the existence of a long flux rope partly loaded with the filament mass . during the eruption ,
internal plasma motions along the flux - rope axis would reveal the previously invisible far endpoint of the flux rope . in our case
, it is difficult to admit the existence of such a long flux rope reaching the disk centre in the _ stereo b _ images .
the pil , which the filament is associated with , runs to the north - east at all heights below 40 mm .
there is a high pil stretched from the filament position to the south - east ( fig .
10(f ) ) but we have no manifestations of the presence of the flux rope there . in principle
, we can imagine the existence of an invisible ( no mass load ) flux rope with one endpoint anchored at the bottom - left corner of the studied region , which extends at a height more than 40 mm to the upper - right corner where it is lowed down to a height of 20 mm and turns to the bottom of the region providing conditions for the filament formation . only in this case
, the observed prominence behaviour could be interpreted as the asymmetric eruption .
another possibility is reconnection of the flux - rope field lines somewhere along the northern half of rising filament loop with ambient coronal magnetic field lines rooted in a quiet region to the south - east from the filament .
in contrast to examples of the coronal reconnection , where the filament body disintegrates , the studied erupting filament loop keeps its shape of a rather thin loop .
reconnection of an erupting flux rope with ambient coronal flux has been seen in several numerical simulations of eruptions @xcite .
we found from comparison of _ stereo a _ and _ stereo b _ images that the observed final height of the originally northern leg ( marked with the blue arrows in fig .
9 ) is at least lower than 100 mm while the height of the prominence top at this time is about 500 mm .
thus , the shape of the erupting filament axis is a loop with the high summit and low ends .
we studied the filament eruption on 30 april - 1 may 2010 , which shows the reconnection of one filament leg with the region far away from its initial position .
observations from three viewpoints were used , namely on - disk h@xmath0 observations by the mauna loa solar observatory , culgoora solar observatory together with euv observations by _
proba2 _ and on - limb observations by _ stereo a _ and _ stereo b_. at the beginning of the event , the eruptive prominence looked like a typical one , that is as an expanding loop with anchored endpoints .
then the top of the loop folded over showing writhing of the filament axis and deflected to the south during the ascending .
a little later , the endpoints reshuffled their positions in the limb view .
the northern leg of the erupting prominence loop jumps laterally to the latitude lower than the latitude of the former southern endpoint .
this behaviour could be interpreted as the asymmetric zipping - like eruption , although it does not look very likely .
although there is a pil stretched at a height above 40 mm from the former filament position to the new eastern end point ( fig .
10(f ) ) , no manifestations of the presence of the flux rope were observed there and any flare - ribbon - like brightening along the path of this pil was absent .
hence , observations do not support the interpretation of the event as a whipping - like or a zipping - like asymmetric filament eruption .
more probable seems reconnection of the flux - rope field lines in its northern leg with ambient coronal magnetic field lines rooted in a quiet region far from the filament .
the eruption of the prominence was followed by a cme . due to the geometrical factor
, it was brightest in the field - of - view of the _ stereo a _ cor2 coronagraph . at the late stage ,
the legs of the cme connecting its core with the sun show noticeable twisted structure .
we calculated parameters of coronal potential magnetic field and found that the eruption is likely to begin with an instability not in the vertical direction , as typical for eruptive filaments , but after the violation of the horizontal equilibrium .
observations show that the trajectory of the eruptive filament deviates from the vertical line by an angle of about 45@xmath6 .
this is unusual initiation of an eruption with combination of horizontal and vertical initial flux - rope displacements showing new unexpected possibility for the start of eruptive events .
the author acknowledge the mauna loa solar observatory , culgoora solar observatory , _ stereo _ , _ soho _ , and _ proba2 _
teams for the high - quality data supplied .
the author thank the referee for critical comments and useful suggestions . this work was supported in part by the russian foundation for basic research ( grants 12 - 02 - 00008 , 14 - 02 - 92690 ) .
99 amari t. , luciani j. , mikic f.z . , linker j. , 2000 , apj , 529 , l49 antiochos s.k . , devore c.r .
, klimchuk j.a . , 1999 ,
apj , 510 , 485 bateman g. , 1978 , mhd instabilities , massachusetts institute of technology , cambridge , ma .
brueckner g.e .
1995 , sol .
, 162 , 357 carlyle j. , williams d.r , van driel - gesztelyi l. , innes d. , hillier a. , matthews s. , 2014 , apj , 782 , 87 chen j. , 1989 , apj , 338 , 453 choe g.s .
, lee l.c . , 1996 ,
apj , 472 , 360 d@xmath41moulin p. , aulanier g. , 2010 , apj , 718 , 1388 fan y. , 2010 , apj , 719 , 728 filippov b. , 2013 , apj , 773 , 10 filippov b.p . , den o.g . , 2000 , astron .
, 26 , 322 filippov b.p . ,
den o.g . , 2001 , j. geophys .
res . , 106 , 25177 filippov b.p , gopalswamy n. , lozhechkin a.v . , 2001 , sol .
, 203 , 119 filippov b.p . , gopalswamy n. , lozhechkin a.v . , 2002 ,
, 46 , 417 forbes t.g . ,
isenberg p.a .
, 1991 , apj , 373 , 294 howard r.a .
et al . , 2008 , space sci .
, 136 , 67 gibson s.e . , fan y. , 2006 , apjl , 637 , l65 gibson s.e . , fan y. , 2008 , jgr , 113 , a09103 gilbert h.r . , inglis a.r .
, mays m.l . , ofman l. , thompson b.j . , young p.r . , 2013 , apj , 776 , l12 gopalswamy n. , hanaoka y. , 1998 , apjl , 498 , l179 gopalswamy n. , hanaoka y. , hudson h.s . , 2000 ,
space res .
, 25 , 1851 grechnev v.v . ,
uralov a.m. , slemzin v.a .
, chertok i.m . ,
kuzmenko i.v .
, shibasaki k. , 2008 , sol .
phys . , 253 , 263 green l.m .
, kliem b. , t@xmath42r@xmath42k t. , van driel - gesztelyi l. , attrill g.d.r .
, 2007 , sol .
phys . , 246 , 365 halain j .- p .
et al . , 2013 ,
phys . , 286 , 67 innes d.e . , cameron r.h .
, fletcher l. , inhester b. , solanki s.k . , 2012 ,
a&a , 540 , l10 ji h. , wang h. , schmahl e.j . , moon y .- j .
, jiang y. , 2003 , apj , 595 , l135 kliem b. , t@xmath42r@xmath42k t. , 2006 , phys .
lett . , 96(25 ) , 255002 kliem b. , t@xmath42r@xmath42k t. , thompson w.t . , 2012 , sol .
phys . , 281 , 137 lin j. , forbes t.g . ,
isenberg p.a . , d@xmath41moulin p. , 1998
, apj , 504 , 1006 liu r. , alexander d. , gilbert h.r . , 2009 , apj , 691 , 1079 low b.c . , 2001 , j. geophys . res . , 106 , 25141 lugaz n. , downs c. , shibata k. , roussev i.i . , asai a. , gombosi t.i . , 2011 , apj , 738 , 127 manoharan p. k. , van driel - gesztelyi l. , pick m. , p. d@xmath41moulin p. , 1996
, apjl , 468 , l73 moore r.l .
, roumeliotis g. , 1992 , in svestka z. , jackson b.v .
, machado m.e .
, eds , proc . iau colloquium 133 , eruptive solar flares , lecture notes phys . vol .
399 , springer , berlin , p. 69 muglach k. , wang y. , kliem b.,2009 , apj , 703 , 976 priest e.r . , forbes t.g . , 1990 ,
phys . , 126 , 319 romano p. , contarino l. , zuccarello f. , 2003 , sol .
phys . , 214 , 313 seaton d.b .
et al . , 2013 ,
phys . , 286 , 43 scherrer p.h .
1995 , sol .
phys . , 162 , 129 schmieder b. , d@xmath41moulin p. , aulanier g. , 2013 , adv .
space res . , 51(11 ) , 1967 schrijver c.j .
, elmore c. , kliem b. , t@xmath42r@xmath42k t. , title a.m. 2008 , apj , 674 , 586 sun x. , hoeksema j.t . , liu y. , chen q. , hayashi k. , 2012 , apj , 757 , 149 titov v.s . , d@xmath41moulin p. , 1999 ,
a&a , 351 , 707 t@xmath42r@xmath42k t. , kliem b. , 2007 , astronomische nachrichten , 328 , 743 tripathi d. , gibson s.e . , gilbert h.r . , qiu j. , fletcher l. , liu r. , mason h.e . , 2009 , a&a , 498 , 295 van driel - gesztelyi l. et al . , 2014 , in schmieder b. ,
malherbe j .-
, wu s.t , eds , proc . iau symposium 300 , nature of prominences and their role in space weather , cambridge university press , cambridge , uk , p. 502
van tend w. , kuperus m. , 1978 , sol .
phys . , 59 , 115 williams d.r . , t@xmath42r@xmath42k t. , d@xmath41moulin p. , van driel - gesztelyi l. , kliem b. , 2005 , apj , 628 , l163 wuelser j .-
et al . , 2004 ,
spie , 5171 , p. 111
zhou j. et al . , 2006 ,
apj , 651 , 1238 zuccarello f. , meliani z. , poedts s. , 2012 , apj , 758 , 117 | filament eruption on 30 april - 1 may 2010 , which shows the reconnection of one filament leg with a region far away from its initial position , is analyzed .
observations from three viewpoints are used for as precise as possible measurements of endpoint coordinates . the northern leg of the erupting prominence loop jumps laterally to the latitude lower than the latitude of the originally southern endpoint .
thus , the endpoints reshuffled their positions in the limb view .
although this behaviour could be interpreted as the asymmetric zipping - like eruption , it does not look very likely .
it seems more likely to be reconnection of the flux - rope field lines in its northern leg with ambient coronal magnetic field lines rooted in a quiet region far from the filament . from calculations of coronal potential magnetic field
, we found that the filament before the eruption was stable for vertical displacements , but was liable to violation of the horizontal equilibrium .
this is unusual initiation of an eruption with combination of initial horizontal and vertical flux - rope displacements showing a new unexpected possibility for the start of an eruptive event .
[ firstpage ] sun : activity sun : filaments , prominences sun : magnetic fields infrared : stars . |
the fundamental interactions of quarks and gluons , especially these with large momentum transfer , studied in lepton - hadron and hadron - hadron collisions , are successfully described by quantum chromodynamics ( qcd ) .
a challenge for this theory is the description of larger and more complex systems , necessary for understanding the properties of the matter created in the early stage of the big bang .
conditions similar to that in the early stage of the universe are created in the collisions of heavy nuclei in the relativistic heavy ion collider ( rhic ) at brookhaven national laboratory . in this accelerator the beams of heavy ions , up to au , can be accelerated to the momentum 100 gev / c per nucleon and collide with the highest energy available currently in the laboratory ( @xmath1 = 200 gev ) .
the system created in the collision of two au nuclei reaches , after approximate equilibration at @xmath2 2 fm / c , the energy density of at least 3 gev/@xmath3 .
this energy density is about 6 times larger than that of the proton under normal conditions @xcite .
qcd based calculations using numerical techniques of lattice gauge theory suggest that under such conditions a different type of matter , quark - gluon plasma ( qgp ) , should appear . in the following sections the properties and the evolution of the system created in the heavy nuclei collisions are described .
the dependence on the centrality of the collision ( i.e. the geometrical overlap between nuclei in the plane transverse to their direction of flight ) represented by the number of binary collisions between nucleons , @xmath4 , and the number of nucleons participating in the collision , @xmath5 , is extensively studied .
the early prediction for the hot and dense nuclear matter suggested weak interactions between partons and creation of a gas of quarks and gluons , but experimental results did not confirm these expectations @xcite .
the first evidence of strong interactions between partons came from the observation of high-@xmath0 particles , which originate from jets created in relatively rare hard collisions of quarks or gluons .
a proper measure of the relative yields of such particles is the nuclear modification factor : + @xmath6 + which normalizes production in @xmath7 collisions by elementary @xmath8 yields and accounts for increased probability of hard processes by dividing by @xmath4 . in the absence of any nuclear effects @xmath9=1 , and
this value is obtained for large @xmath0 in @xmath10 collisions @xcite , where no such effects are expected . in the most central @xmath11 collisions
@xmath9 drops below 0.4 for @xmath12 3 gev / c ( fig . [ figraa ] ) .
this effect is explained by strong interactions in the medium causing that only partons produced near the surface and emitted in the outside direction do not lose their momentum and can fragment into high-@xmath0 particle(s ) . for smaller systems like @xmath13 or for less central @xmath11 collisions
the suppression is also present , but its magnitude is smaller @xcite .
large acceptance of the phobos detector allows to study the effects of parton absorption in the sqgp .
recently , we have measured in @xmath11 collisions the correlation between a high-@xmath0 trigger particle ( @xmath12 2.5 gev / c ) with the remaining particles @xcite .
it is known that in the elementary @xmath8 interactions such high-@xmath0 particle is accompanied by close partners , but correlated particles are also present at opposite azimuthal angles . in the case of @xmath11 collisions production of correlated particles is enhanced ( fig . [ figtrighist ] ) .
in contrast to the elementary interactions the correlation in the `` near side '' is present also at large distances in pseudorapidity , @xmath14 2 , which may be related to the longitudinal expansion of the system .
equally interesting is the study of very low-@xmath0 particles production .
theory predicted strong increase of their yields in the case of quark - gluon gas creation .
the phobos experiment is capable to register and identify charged particles with very low momenta , starting from @xmath0= 30 mev / c .
measured yields ( fig .
[ figlowpt ] ) do not show any significant enhancement , they agree with the extrapolations of the fits ( performed in the higher @xmath0 range ) of the models assuming expansion of the system , for more details see @xcite .
as it was already shown before , the system created in the @xmath7 collision is expanding .
more detailed information on the system evolution is obtained in the study of the collective elliptic flow .
the elongated shape of the overlap area of the nuclei is reflected in the anisotropy in azimuth of the particle momenta distribution and measured as the second coefficient , @xmath15 , in the fourier expansion of @xmath16 distribution ( relative to the reaction plane , which is defined by the direction of the beam and the impact parameter vector ) @xcite .
as expected , largest flow is observed in peripheral collisions and decreases with centrality ( fig .
[ figv2npart ] ) . for comparison of the flow with collision geometry the eccentricity , @xmath17 , of the interaction area was used .
the phobos experiment has proposed to introduce participant eccentricity @xcite , @xmath18 , basing on positions of nucleons taking part in the collision ( as obtained from glauber monte carlo simulations ) .
this definition accounts for the fact that the axis of the interaction area defined by the nucleons usually is rotated with respect to the reaction plane .
dividing of @xmath15 for @xmath11 and @xmath13 collisions by eccentricity allows to obtain a universal dependence on @xmath5 only if @xmath19 is used ( fig .
[ figv2ecc ] ) .
one step further goes the analysis of the elliptic flow fluctuations @xcite .
the relative @xmath15 fluctuations measured by phobos are large , they exceed 30% .
the comparison with the fluctuations of participant eccentricity presented in fig .
[ figv2fluct ] shows that the fluctuations present at the very beginning of collisions are not significantly modified during the expansion of the system .
= 5.6 cm = 5.6 cm details of the expansion of the system and it s hadronization determine also two - particle correlations , which are measured by phobos in the full azimuthal angle and a wide pseudorapidity range , @xmath20 3 .
earlier study of forward - backward multiplicity correlations has already shown strong short - range correlations , not explained by the models of @xmath7 collisions @xcite .
they can be interpreted and parameterized by a cluster model @xcite , which assumes that the particles are produced in two steps : first some unstable objects , clusters , are produced , later they decay into observed particles . using the correlation functions obtained from the experimental data two basic parameters of the clusters can be extracted @xcite : @xmath21 - the effective number of particles forming a cluster and @xmath22 - the width of the 2-particle correlation in @xmath23 which is directly connected with the cluster width .
the values of these parameters obtained for @xmath8 , @xmath13 and @xmath11 collisions are presented in fig .
[ figkeffdelta ] .
width of the clusters in the @xmath7 collisions seems to be larger than for @xmath8 .
the cluster size , @xmath21 , decreases with centrality of @xmath7 collision .
the values of @xmath21 are large , especially as they were not corrected for acceptance effects ( which would increase them ) and that the neutral particles , which are not detected , belonging to the clusters are not counted .
interestingly , the centrality dependence for @xmath11 and @xmath13 becomes almost identical if events with similar shapes of nuclei overlap area are compared @xcite .
in the collisions of heavy nuclei at very high energies extremely high energy density is reached and a new phase of matter , sqgp , is created @xcite .
strong interactions in the quark - gluon matter are visible as the suppression of high-@xmath0 particles , the lack of enhancement in low-@xmath0 particles production and the presence of collective effects leading to elliptic flow .
the yields of low-@xmath0 particles and the elliptic flow are consistent with the assumption of expansion of the system , as predicted by hydrodynamic model for an almost perfect fluid .
global observables in the collisions of nuclei , not discussed in this paper , reveal unexpected simple relations : scaling of multiplicity with @xmath5 and extended longitudinal scaling of @xmath24 and @xmath15 @xcite .
in the recent studies the phobos experiment measured correlations of associated particles with a high-@xmath0 trigger particle , which extend at least 4 units in pseudorapidity . such long range correlations may be a sign of longitudinal expansion of the system .
new results on the yields of low-@xmath0 particles support the hypothesis of a radial component of such expansion .
the shape of the interaction area , characterized by eccentricity @xmath19 , determines the value of elliptic flow , @xmath15 .
fluctuations of elliptic flow are of the same size as the fluctuations of eccentricity at the very beginning of the collision .
strong short range correlations are observed in the a+a collisions , large size of clusters can not be explained by low mass resonances .
15 b.b . back _ et al .
_ ( phobos ) , nucl . phys . * a757 * ( 2005 ) 28
. m. gyulassy , l. mclerran , nucl . phys .
* a750 * ( 2005 ) 30 .
b.b . back _ et al .
_ ( phobos ) , phys . rev
* 91 * ( 2003 ) 072302 .
b. alver _ et al . _
( phobos ) , phys .
* 96 * ( 2006 ) 212301 .
b.b . back _ et al .
_ ( phobos ) , phys .
b * 578 * ( 2004 ) 297 .
e. wenger _
( phobos ) , arxiv : 0804.3038v3 [ nucl - ex ] ( 2008 ) .
adler _ et al . _
( phenix ) , phys .
c * 69 * ( 2004 ) 034909 .
t. gburek _ et al . _
( phobos ) , arxiv : 0804.4270v1 [ nucl - ex ] ( 2008 ) .
b.b . back _ et al .
_ ( phobos ) , phys . rev .
c * 72 * ( 2005 ) 051901 .
b. alver _ et al . _
( phobos ) , phys .
* 98 * ( 2007 ) 242302 .
s. manly _ et al .
_ ( phobos ) , nucl . phys . * a774 * ( 2006 ) 113
. b. alver _ et al .
_ ( phobos ) , arxiv : 0804.4297v1 [ nucl - ex ] ( 2008 )
. b. alver _ et al . _
( phobos ) , phys .
c * 77 * ( 2008 ) 014906 .
b.b . back _ et al .
_ ( phobos ) , phys . rev .
c * 74 * ( 2006 ) 011901 .
a. morel and g. plaut , nucl
* b78 * ( 1974 ) 541 .
b. alver _ et al . _
( phobos ) , phys .
c * 75 * ( 2007 ) 054913 .
w. li _ et al .
_ ( phobos ) , arxiv : 0804.2471v2 [ nucl - ex ] ( 2008 ) . | the phobos experiment at the relativistic heavy ion collider ( rhic ) has collected a large dataset of au+au , cu+cu , d+au and p+p collisions in the center of mass energy range spanning from 19 gev / nucleon to 200 gev / nucleon .
the almost full angular coverage of the phobos detector allows the study of particle production over 10 units pseudorapidity .
the unique design of the spectrometer enables reconstruction and identification of charged particles down to very low transverse momenta . in this
paper properties of the strongly interacting quark - gluon plasma ( sqgp ) created in the nucleus - nucleus collisions at the highest energy available in laboratory are discussed .
results from the phobos experiment on jet suppression , very low @xmath0 particles production and elliptic flow are shown . in more details
are presented the most recent studies of the correlations of charged particles with respect to a high-@xmath0 trigger particle , elliptic flow fluctuations and two particle correlations . |
bosonization analyses@xcite together with finite size numerical work using the dmrg@xcite and other methods@xcite have given a clear understanding of the behavior of the 2-leg hubbard and @xmath0 ladder models .
the hubbard hamiltonian is written : @xmath3 here @xmath4 destroys an electron on rung @xmath5 and leg @xmath6 with spin @xmath7 .
@xmath8 is the electron number operator .
the @xmath0 hamiltonian is : @xmath9 where @xmath10 and the hilbert space now excludes all states with doubly - occupied sites .
we will generally set @xmath11 in what follows .
both models are expected to be in a `` c1s0 '' phase , over a wide range of parameters , in which the low energy degrees of freedom consist of a single free massless charge boson , whose excitations carry even multiples of the electron charge . this is often characterized as a `` d - wave superconductor '' based on the nature of the power law decay of pair correlations which have a positive sign for singlet rung - rung or leg - leg correlations and a negative sign for rung - leg correlations .
both bosonization and dmrg results exist on multi - leg ladders with analytical and numerical uncertainty which increase with the number of legs .
most of the dmrg data is obtained with open boundary conditions along the legs of the ladder . unlike with periodic boundary conditions , this generally leads to charge density oscillations in the groundstate .
if these oscillations persist at the center of the chain for arbitrarily long chain length then they correspond to a charge density wave ( cdw ) .
more commonly , they decay away from the boundaries with a power law , in the limit of an infinite chain . in this case , we may think of them as generalized friedel oscillations where the chain ends themselves act as impurities which induce gradually decaying density oscillations . unlike in a fermi liquid
, these generalized friedel oscillations in a luttinger liquid decay with an exponent which depends on the interaction strength , and which is simply related to the density correlation exponent.@xcite furthermore , the wave - vector of the oscillations itself can be changed by the interactions .
these density oscillations , for the largest systems studied , up to eight legs , have been identified with `` stripes''.@xcite an understanding of the occurance of stripes in these models , whether or not they require long range coulomb interactions to exist and their connection with superconducitivity are important open problems .
since any ladder system is ultimately one - dimensional if the number of legs is held fixed and the length taken to @xmath12 , the friedel oscillations /cdw analysis may provide the appropriate description of stripe behavior , at least in this limit and in cases where the stripe wave - vector is parallel to the chains . in sec .
ii we briefly review the bosonization picture of the 2-leg @xmath0 ladder .
we also present a technique for extracting the correlation exponent purely from data with open boundary conditions which has not , to our knowledge , been used previously on ladders , although it has been used on chains .
this method is used to obtain values for the charge velocity and exponent . in sec .
iii we review the bosonization treatment of commensurate charge density waves and the related behavior of the correlation exponent , and analyse data on the 2 leg tj model at electron density @xmath13 , showing that a cdw may occur for weak enough @xmath14 . in sec .
iv we review friedel oscillations in luttinger liquids and apply this analysis to data on the single chain hubbard and @xmath0 model and the 2 leg @xmath0 model .
a convenient starting point for bosonization is the weak coupling hubbard model version of the 2-leg ladder , eq .
( [ hhubb ] ) with @xmath15 small .
it is then far from obvious that this analysis will apply to the infinite coupling limit , corresponding to the @xmath0 model of eq .
( [ htj ] ) so that comparisons with numerical results is important . in the weak coupling limit
we may start by diagonalizing the non - interacting problem , giving symmetric and anti - symmetric electron operators @xmath16 where @xmath17 , @xmath18 labels even and odd channels .
we pass to the continuum limit by introducing left and right moving fields : @xmath19 with @xmath20 and @xmath21 the fermi wave vectors for the two bands . in the usual way
, we represent the left and right moving fermion fields by left and right moving boson fields : @xmath22 and then , introducing the dual canonical bose fields , @xmath23 it is then convenient to introduce spin and charge bosons , for each channel : @xmath24 and then , finally , switch to the two linear combinations of the even and odd bosons : @xmath25 with similar relations for @xmath26 , @xmath27 , and @xmath28 .
actually , this last transformation is not canonical when the even and odd bosons have different velocities .
however , we will follow the standard practice @xcite of assuming that this velocity difference is irrelevant .
a renormalization group analysis , based on the weak coupling hubbard model , suggests that the cosine interactions `` pin '' the bosons @xmath29 and @xmath30 , introducing excitation energy gaps for these bosons and leaving @xmath31 as the only massless boson which thus describes the low energy excitations .
all interactions involving @xmath31 , @xmath32 are irrelelvant in this phase .
the low energy effective hamiltonian can be written : @xmath33,\label{ham}\ ] ] where @xmath34 is the momentum density variable canonically conjugate to @xmath32 , @xmath35 is the velocity of the corresponding gapless low energy excitations and the parameter @xmath36 controls the correlation exponents .
( our definition of @xmath36 corresponds to that of schulz@xcite and hayward and poilblanc@xcite but is the inverse of the parameter with the same name in balents and fisher.@xcite ) the hamiltonian may be equally well written in terms of the other boson field @xmath31 and its conjugate momentum using : @xmath37 the two parameters , @xmath35 and @xmath36 are generally difficult to calculate analytically and are extracted from numerical data .
the long - range behaviors of various correlation functions are calculated straightforwardly by expressing the corresponding fermionic operators in terms of the bosons @xmath38 , @xmath39 and their duals .
exponentials of the pinned bosons can be replaced by their groundstate expectation values but exponentials of the duals of pinned bosons lead to exponentially decaying factors in correlation functions .
exponentials of the gapless boson , @xmath32 and its dual give power law decaying factors .
for instance , to calculate the uniform part of the pair correlation function we bosonize the pair operator : @xmath40 we may replace the exponentials of @xmath30 and @xmath41 by a constant factor leaving simply the operator @xmath42 .
the correlation function for this operator decays as @xmath43 .
the same result is obtained for the correlation function of a pair of electrons in the odd channel , @xmath44 . on the other hand ,
the correlation function @xmath45 is the same except for a factor of @xmath46 .
this is expected to be @xmath47 , based on the sign of the cosine interaction which pins the @xmath30 boson to @xmath48 , so this pair correlation function has the opposite sign , corresponding to `` d - wave pairing '' . the @xmath49 part of the ( even , spin - up ) density operator is : @xmath50 the correlation function of this operator has exponential decay due to the @xmath51 factor .
exponential decay is also obtained for all other terms in the @xmath49 part of the density operator . on the other hand ,
if we consider the correlation function for the square of the density operator , we get power - law decay for the @xmath52 part .
this arises from terms of the form : @xmath53 we may replace @xmath54 by its expectation value leaving only the gapless @xmath32 field which gives power law decay : @xmath55\over will pick up a @xmath52 term proportional to this and also exhibit the same type of @xmath52 power law decay .
this phase with one gapless charge mode and no gapless spin modes , labeled `` c1s0 '' is expected to be the generic phase in the 2 leg @xmath0 or hubbard ladder . whether this phase is better thought of as almost a d - wave superconductor , or almost an incommensurate cdw depends on the magnitude of @xmath36 .
the pairing correlation function decays less rapidly than the density correlation function in the case @xmath56 .
schulz has shown@xcite that @xmath57 as the density , @xmath58 , approaches @xmath59 . at @xmath60 ,
the ladder is a spin gapped mott insulator .
thus it is of interest to calculate @xmath36 .
this can be done from the finite size spectrum using either periodic or open boundary conditions .
the latter are often used with dmrg so we discuss both cases .
while , as we discuss below , it is enough to measure the excitation energies of only 3 states to extract @xmath36 ( and the velocity @xmath35 ) we give formulas for an infinite number of low lying excitations .
this allows for a more extensive check of the bosonization results with dmrg . due to the somewhat conjectural nature of the extrapolation of weak coupling hubbard results to the @xmath0 model
this would be a worthwhile check .
we discuss only the groundstate and low - lying excitations , whose excitation energies all scale to 0 as @xmath61 , where @xmath62 is the length of the system .
these can be simply calculated from the free boson hamiltonian of eq .
( [ ham ] ) when proper account is taken of the boundary conditions on the boson field , @xmath32 .
we will generally set @xmath63 , restoring it by dimensional analysis when needed .
we first consider the case of periodic boundary conditions ( b.c.s ) .
clearly periodic b.c.s on the fermions translate into periodic b.c.s on the boson fields but we must take into account the fact that the boson fields are actually phase fields ( only their exponentials and derivatives occur as physical local operators ) .
since all physical operators in the low energy effective theory involve integer powers of @xmath64 $ ] and @xmath65 $ ] , as in eqs .
( [ pair ] ) and ( [ density ] ) , we see that : @xmath66 where @xmath67 and @xmath68 are arbitrary integers . to obtain the complete low energy spectrum we simply write a mode expansion for the boson fields , consistent with these b.c.s and the canonical commutation relations .
this gives : @xmath69 here the quantum numbers @xmath67 and @xmath68 are integer valued .
( [ pi ] ) we see that the zero wave - vector mode of the conjugate momentum , @xmath70 , has eigenvalues @xmath71 .
thus the wave - functions for the zero - momentum modes are @xmath72 $ ] .
these wave - functions are invariant under @xmath73 , as required by the angular nature of @xmath74 .
( the same result holds for the dual zero wave vector field , @xmath75 and its conjugate momentum . )
the @xmath76 and @xmath77 operators annihilate left and right moving harmonic boson modes .
we may give a physical interpretation to the quantum number @xmath67 using the bosonization formula for the total charge density : @xmath78 thus the total charge , relative to that of the groundstate , is : @xmath79 only excitations of even charge occur in the c1s0 phase .
the other quantum number , @xmath67 , measures the `` chiral charge '' which has a less obvious physical interpretation as the difference of charges of left and right movers . inserting eq .
( [ mode ] ) into the hamiltonian of eq .
( [ ham ] ) and using the identity , eq .
( [ pi ] ) we can immediately read off the spectrum : @xmath80 . \label{es}\end{aligned}\ ] ] here @xmath81 and @xmath82 are the occupation numbers for the left and right moving states of momentum @xmath83 .
@xmath84 is the groundstate energy for a given density , @xmath58 , and is non - universal .
this formula gives the excitation energy for low - lying excitations with all quantum numbers @xmath85 .
the parameters @xmath86 ( chemical potential ) @xmath35 and @xmath36 all depend on density .
the parameters @xmath36 and @xmath35 can be determined by measuring the excitation energies of three states .
generally one calculates the excitation energies of the states with @xmath87 ( and all other quantum numbers set to zero ) to determine the ratio @xmath88 , using : @xmath89 the compressibility for a 2-leg ladder is generally defined as @xmath90 where @xmath58 is the density . from eqs .
( [ q ] ) and ( [ es ] ) , this has the value @xmath91 .
the velocity may be measured separately from the excitation energy of the lowest state of momentum @xmath92 : @xmath93 thus measuring three excitation energies , for fixed , large @xmath62 , allows a determination of the correlation exponents .
using a large @xmath62 is important because corrections to eq .
( [ es ] ) are only down by additional powers of @xmath61 .
these corrections become especially large at commensurate filling near a transition to a cdw .
the corresponding exponent will be given in sec .
iii . while these three measurements are enough to determined the critical exponents , measurements of energies of additional states and a study of the @xmath62-dependence provide additional confirmation of the predictions of the rg and bosonization .
as mentioned above , dmrg works much more efficiently with open boundary conditions .
we now discuss the finite size spectrum in that case .
if we number the sites from 1 to l then we have a free boundary condition at @xmath94 and @xmath95 .
a free boundary condition on the fermion field at @xmath94 is equivalent to a vanishing boundary condition at @xmath96 .
[ this follows from adding an extra `` phantom site '' at @xmath96 with associated hopping and exchange terms but then making these terms vanish by imposing @xmath97 . ] in terms of left and right movers this free boundary condition becomes : @xmath98 upon bosonizing we obtain : @xmath99 or @xmath100 ( a determination of this constant involves consideration of some subtle commutators
. we will not bother to keep track of it in what follows . )
clearly this boundary condition remains the same when expressed in the eventual basis @xmath101 , @xmath102 .
the only important boundary condition for the low energy excitations is : @xmath103 the same b.c .
is obtained at @xmath104 except that the constant will generally be different .
taking into account the periodic nature of @xmath74 , eq .
( [ bc ] ) , and the fact that @xmath105 must also vanish at the boundaries , we see that the mode expansions become : @xmath106 here @xmath107 is proportional to the difference of constants appearing in the b.c.s at @xmath108 and @xmath109 . substituting into the hamiltonian
, we now obtain the finite size spectrum : @xmath110.\label{esopen}\end{aligned}\ ] ] we see that eq .
( [ detk ] ) for @xmath88 remains true with free b.c.s as we might expect due to the relation with the compressibility .
we may now determine the velocity independently ( and hence determine @xmath36 ) by measuring the gap to the first excited state with the same charge as the groundstate : @xmath111 note that this gap has half the value of the gap to the lowest energy state of momentum @xmath92 in the case of periodic b.c.s , eq .
( [ detv ] ) .
this is just a consequence of the familiar result that the spacing of wave - vectors for open b.c.s is @xmath112 , half the spacing for periodic b.c.s .
previous measurements of the parameter @xmath36 , and its analogue in other chain and ladder systems , have used periodic boundary conditions .
exact diagonalization , or `` modified lanczos '' methods work efficiently to determined the lowest energy state of given quantum numbers .
thus the energy differences of eqs .
( [ detk ] ) and ( [ detv ] ) can be measured from finding the lowest energy states with various charges and with momenta @xmath113 and @xmath92 .
an alternative approach is to meaure the dependence of the groundstate energy ( for a fixed charge ) on an applied flux , i.e. a twist in the boundary conditions on the fermion fields : @xmath114 this corresponds to putting a twist into the boundary condition on @xmath31 in eq .
( [ bc ] ) , as can be seen from the fact that all the fermion fields ( @xmath62 or @xmath115 ) contain a phase factor of @xmath116 $ ] .
@xmath117 from eq .
( [ es ] ) we see that the groundstate energy is increased by : @xmath118 hence measuring the flux dependence of the groundstate energy determines @xmath119 , while measuring the compressibility determines @xmath88 .
hayward and poilblanc@xcite measured @xmath36 in this way for the 2-leg ladder using a maximum system size of @xmath120 .
another approach was taken by siller et al.@xcite they modeled the 2-leg tj model as a single chain model of bosonic hole pairs .
the parameters in the effective hamiltonian for hole pairs were determined from dmrg calculations with open boundary conditions on system sizes up to @xmath121 for systems containing 2 holes or 4 holes only .
the resulting bosonic model was then studied using exact diagonalization ( lanczos ) methods and periodic boundary conditions . since the number of effective bosons ( of density @xmath122 where @xmath123 ) is small near 1/2-filling , it was possible to diagonalize large systems ( up to 220 sites in the case of only 2 bosons ) .
the approach advocated here , working exclusively with open boundary conditions , has been used previously for spin chains@xcite but not , as far we we know , for tj ladders .
it allows us , using dmrg , to study much larger system sizes , up to @xmath124 , than are accessible with lanczos in the usual formulation.@xcite it furthermore avoids making any assumptions which are necessary in treating the hole pairs as bosons , such as the particular form of the boson - boson interaction and the absence of three - boson terms , which could induce some density dependence in the interaction .
however , the dmrg calculations with the accuracy and system length required to determine @xmath36 for these ladder systems were surprisingly difficult . in order to minimize the effect of correction terms to the asymptotic formulas ,
long systems were needed . in these systems , one is probing competing pairing and charge fluctuations at very low energies and large distances , which is difficult in dmrg . typically more than 2000 states were kept in the calculations and more than a dozen sweeps were performed .
some of the calculations were more accurate and reliable than others , however .
in fig . [ figdencgall ] we show a plot of the left - hand side of eq .
( [ detk ] ) versus @xmath61 for three values of @xmath125 , showing linear behavior and allowing us to extract @xmath88 .
these calculations require only the ground state energy , for which an extrapolation in the energy versus the truncation error is very reliable .
thus this determination of @xmath88 is the most reliable and accurate ingredient in the determination of @xmath36 , and , in fact , the results shown could be improved with modest additonal effort . )
are 2.70 for j=.5 , n=0.875 , 6.75 for j=.5 , n=0.75 and 7.68 for j=.35 , n=0.75.,width=302 ] in fig .
[ figcgall ] we plot the left - hand side of eq .
( [ detvopen ] ) versus @xmath61 , again obtaining linear behavior .
these calculations were difficult .
it is necessary to target two states simultaneously , and extrapolation is not very useful .
note that , if @xmath62 is not large enough and the ( infinite @xmath62 ) spin gap is not small enough , @xmath126 , then the first excited state with the same quantum numbers might actually have @xmath127 , @xmath128 and correspond to a spin excitation , rather than the desired neutral excitation of the @xmath129 field .
( it might also be a neutral excitation of one of the other gapped boson fields . ) in order to check the former possibility , we have calculated the spin gap , i.e. the excitation energy for the lowest state with @xmath130 .
we find that the very nonlinear behavior for @xmath131 , @xmath1 shown in fig .
[ figcgall ] is due to a small spin gap in this system , of order 0.04 .
thus it was necessary to study a @xmath132 system , for which we kept 4000 states , and performed a dozen iterations , to clearly see the required gap .
we believe this data is reliable , but due to the large numerical work required we have only studied three different values of @xmath125 . combining the results of figs . [ figdencgall ] and [ figcgall ] allows a determination of @xmath35 and hence @xmath36 .
the resulting values are given in table [ table : v , k ] . and same electron density .
the spin gap is about 0.14 for j=0.5 , n=0.875 , about 0.10 for j=.5 , n=0.75 and about 0.07 for j=.35 , n=0.75 . the breakdown of the asymptotic linear behavior is clearly visible in the third case , at energies of order the spin gap .
the straight lines are linear fits to the data .
the slopes of the fits ( which asymptotically yield @xmath133 ) are 1.631 for n=0.875 , j=0.5 ; 3.03 for n=0.75 , j=0.5 ; and 1.6368 for n=0.75 , j=0.35.,width=302 ] .results for the charge velocity @xmath134 and luttinger liquid parameter @xmath135 .
@xmath136 comes from using eqs .
( [ detk ] ) and ( [ detvopen ] ) , while @xmath137 is determined using the decay of the friedel oscilations in the center of the system as a function of the system length .
[ cols= " > , > , > , > " , ] [ table : v , k ]
while a c1s0 phase is expected over most of the parameter range in the 2-leg @xmath0 model , a completely gapped , charge density wave ( cdw ) phase may occur at special commensurate filling factors , depending on the value of @xmath138 ( and other possible interaction parameters).@xcite in this section we review the conditions on the critical exponent parameter , @xmath36 for a cdw to occur and discuss numerical work on the @xmath0 model at electron densities @xmath13 and @xmath139 , corresponding to fillings of 3/8 and 1/4 respectively .
the low energy effective hamiltonian only contains fourier modes of the electron fields within a small momentum range , @xmath140 of @xmath141 .
umklapp type interaction terms , which do not conserve separately the number of left and right movers , are generally accompanied by rapidly oscillating phase factors .
they therefore do not appear in the low energy effective hamiltonian since the electron fields vary slowly and hence these rapidly oscillating factors cause these interactions to average to zero .
however , at special filling factors , corresponding to special values of the fermi wave - vector , the oscillating factors become constants and these operators then appear in the effective hamiltonian . whether or not they produce a gap and a cdw depends on whether or not they are relevant operators in the rg sense . in a 1-dimensional relativistic quantum field theory
an operator is relevant if it has a scaling dimension , @xmath142 .
this scaling dimension determines the exponent , @xmath143 with which the correlation function of this operator decays , with : @xmath144 using the free boson hamiltonian of eq .
( [ ham ] ) , the scaling dimension of any operator is easily obtained .
for example , the hubbard or @xmath145 interaction leads to an interaction term of the form given in eq .
( [ density ] ) .
we see that this oscillates at wave - vector @xmath146 .
only when this wave - vector is a mutiple of @xmath147 will this operator appear in the low energy effective hamiltonian .
furthermore , any operators whose correlation functions decay exponentially are irrelevant .
the number of electrons , @xmath148 in each band is : @xmath149 ( the factor of 2 arises from spin . ) thus : @xmath150 where : @xmath151 is the electron density . while the value of @xmath20 and @xmath21 may be renormalized by interactions ( and may , in fact , not really be well - defined in the interacting model ) their sum is `` protected '' by the one - dimensional version of luttinger s
theorem@xcite so that eq .
( [ k - n ] ) is expected to be exact .
thus we see that the umklapp operator of eq .
( [ density ] ) can only occur in @xmath152 for @xmath60 ( half - filling ) .
this operator has scaling dimension : @xmath153 at 1/4 filling , corresponding to @xmath139 , the product of this operator times itself with spin up replaced by spin down , can occur , giving rise to an operator containing 8 electron operators . under bosonization
this is proportional to @xmath154 $ ] and has dimension @xmath155 . at a filling of 3/8 ( or 1/8 ) corresponding to n=3/4 ( or 1/4 ) the fourth power of the basic umklapp operator in eq .
( [ density ] ) containing 16 electron operators can occur .
it has dimension @xmath156 .
the condition on the rg scaling dimension , @xmath157 , for an operator to induce a cdw is somewhat subtle , and depends on whether the density or @xmath138 is varied .
if we vary @xmath138 , holding the density fixed at the commensurate value , then the system will remain in the gapless c1s0 phase as long as @xmath158 , so that the operator is irrelevant .
a transition to a gapped cdw phase occurs at the value of @xmath138 where @xmath157 becomes @xmath159 . in the cdw
phase the value of @xmath36 becomes undefined since all correlation functions decay exponentially ( or go to constants ) .
we may also consider what happens as we vary the density in the vicinity of the commensurate value .
if the system is not in the cdw phase at a commensurate filling , then we expect @xmath157 to vary smoothly and , of course , to satisfy @xmath158 at the commensurate point . on the other hand ,
if the system _ is _ in the cdw phase at a commensurate filling , then we expect @xmath160 as the commensurate filling is approached .
this follows from a general and clever argument by schulz@xcite based on `` refermionizing '' the single boson hamiltonian .
the relevant operator always behaves like a fermion mass term , of scaling dimension @xmath161 very close to the commensurate filling , whenever a cdw occurs at that point .
it follows that , @xmath157 must vary rapidly in the vicinity of the critical value of @xmath138 where the cdw transition occurs near a commensurate filling . on the cdw side of the transition
@xmath157 will be close to 2 away from the commensurate filling but then drop abruptly to @xmath59 as that filling is approached .
we emphasize that schulz s argument , which was formulated in terms of the single boson sine - gordon theory , is very general and does nt depend on the underlying microscopic model . similar behavior can occur for a @xmath0 or hubbard ladder with any number of legs .
such rapid variation of @xmath135 in the vicinity of commensurate filling for a system with a cdw was observed by schulz for the single leg hubbard model near half - filling , by obtaining @xmath135 from the bethe ansatz and similar behavior was also observed near 1/2-filling ( @xmath60 ) for the 2-leg @xmath0 model by siller et al . now consider a density @xmath139 corresponding to 1/4 filling .
if we sit at this density and vary @xmath138 then a cdw transition occurs when @xmath162 , with the gapless phase having @xmath163 . on the other hand ,
if @xmath138 is such that the system is in a cdw phase at @xmath139 , then @xmath164 as @xmath165 . as a function of @xmath138 and @xmath58 .
a cdw occurs along the dashed line at @xmath166 so @xmath36 is not well - defined there .
as this line is approached by varying @xmath58 , @xmath167 . on the other hand , at @xmath166 and
larger @xmath138 there is no cdw and @xmath36 _ is _ well defined , with a value approaching @xmath168 at the cdw critical point.,width=302 ] similarly , a cdw at 3/8 filling , @xmath13 , is signalled by @xmath169 , as we vary @xmath138 at fixed @xmath58 or @xmath167 as we vary the density at fixed @xmath138 in the cdw phase .
[ fig : krho ] . the connection between the relevance of the multiple umklapp interaction and the presence of a cdw is easily established .
when the multiple umklapp term ( for @xmath139 or @xmath170 ) is relevant it pins the @xmath32 boson . from eqs .
( [ density ] ) and ( [ k - n ] ) we see that the @xmath52 term in the density operator then has a non - zero groundstate expectation value , so that @xmath171 where @xmath107 is a constant . thus the density oscillations should have wavelength 2 for n=1/2 and wavelength 4 for n=3/4 . in general , it is sometimes convenient to introduce @xmath172 the density of holes measured from half - filling , sometimes called the density of `` holons '' and proportional to the doping parameter in the cuprates .
we see that @xmath58 can be replaced by @xmath173 in eq .
( [ densosc ] ) so that the wavelength of the cdw oscillations is @xmath174 . for a 2-leg ladder , this is the average horizontal separation of _ hole pairs_. ( `` horizontal '' refers to the direction along the legs of the ladder . )
the cdw corresponds to a broken translational symmetry , so that there are two different groundstates for n=1/2 and four different groundstates for n=3/4 ( differing only by translation by 1 site ) . for a finite ladder with periodic b.c.s
we expect quantum tunnelling between these groundstates to occur so that no density oscillations exist in the finite system groundstate . on the other hand , with open b.c.s one or the other of the groundstates
gets picked out by the boundary conditions so that the cdw is directly observable
. we now return to the point raised in sec .
ii , the higher order corrections in @xmath61 to the finite size energy gaps given in eqs . ( [ es ] ) and ( [ esopen ] ) .
the leading correction is determined by the leading irrelevant operator . at a commensurate filling , when the system is close to having a cdw , the multiple umklapp operator discussed above has a dimension @xmath157 only slightly greater than 2 , corresponding to being barely irrelevant . in this case the higher order terms in the energy formula are of @xmath175 .
this will make it difficult to determined @xmath36 reliably from finite size gaps in the vicinity of a cdw , complicating the determination of the cdw phase boundary .
previously reported work@xcite based on the bosonic hole pairs approximation together with dmrg and lanczos , on the 2-leg @xmath145 model , with @xmath176 , found a smooth behavior of @xmath36 in the vicinity of @xmath13 with a value at a commensurate density @xmath13 of approximately @xmath177 . using the direct drmg approach ,
we have found @xmath178 .
both results are far above the critical values ( @xmath168 and @xmath179 ) , indicating that a cdw does _ not _ occur at @xmath13 for @xmath176 .
the charge gap for @xmath166 with @xmath180 and @xmath181 is plotted versus @xmath61 in fig .
[ figdencg.35.25 ] .
there is , indeed , no evidence for a charge gap at the larger value of @xmath14 , consistent with the absence of a cdw . on the other hand , there _ is _ evidence for a charge gap at @xmath182 .
note that these calculations involve only ground state energies and thus are very reliable .
although the results are extrapolated to zero truncation error , the extrapolation is quite small in magnitude , of order the symbol size in the figure , and the estimated error in the extrapolation is very small , as shown by the error bars .
these results provide some evidence that there is a criticial value of @xmath14 between 0.25 and 0.35 such that a charge gap and cdw occur for smaller @xmath14 .
as reviewed in the next section , if there is no cdw , and the system in is the c1s0 phase , we expect the density oscillations at the center of the chain to decay with chain length as @xmath183 with @xmath184 . comparing the oscillation amplitude for @xmath166 , @xmath181 and the two longest lengths studied , l=96 and l=192 , we find a ratio of oscillation amplitudes of .915 which would give an exponent of @xmath185 .
this is slightly larger than the critical value of 1/8 . on the other hand ,
it is so close to the critical value that we might wonder if the density oscillation amplitude would eventually approach a constant with still longer chain lengths corresponding to a true cdw .
, showing the occurance of a cdw at @xmath166 for @xmath181 but not @xmath180.,width=302 ] system as a function of @xmath138.,width=302 ] another important question is the behavior of the spin gap in the putative cdw phase . since the umklapp term which drives the cdw does not contain the spin bosons , it is natural to assume that the gapping of the charge boson , corresponding to the cdw , occurs without any major effects on the other bosons which could thus remain gapped
however , it is possible that one ( or more ) of the other bosons , such as a spin boson , becomes gapless at the same transition .
indeed , the occurence of the cdw may be related to the close proximity of a phase with no spin gap . in fig .
[ spingap16x2 ] , we show the spin gap on a @xmath186 system for a wide range of @xmath14 .
the data suggests that in the thermodynamic limit , the spin gap vanishes near @xmath182 . to verify this result
, one must perform a finite size study .
however , the spin density pattern on the @xmath186 system for @xmath187 has the largest values of @xmath188 concentrated near the ends of the system , indicating that an edge excitation has the lowest energy . in order to ensure that we have a bulk spin excitation in the calculation , we next studied systems where we increased the rung exchange interaction to @xmath189 on the first and last rungs of a ladder .
this drives up the energy of the edge excitation , and one finds that the resulting spin pattern is concentrated in the central region of the ladder . in fig .
[ spingapvsl ] , we show the spin gap as a function of @xmath61 for several systems . the results for @xmath187 , @xmath1 are consistent with a gapless state , although one can never rule out a very small gap .
similarly , for @xmath182 , @xmath1 no evidence for a spin gap on the largest system , a @xmath190 ladder . for @xmath187 , @xmath191
we see a clear spin gap of order 0.013 . for several systems .
the values of the exchange on the first and last rungs have been altered to avoid edge excitation.,width=302 ] the proximity to a phase with vanishing spin gap may be related to the occurence of the cdw . as the spin gap gets smaller , and the corresponding length scale larger , we may expect the size of the 2-hole pairs to get larger .
this is naturally associated with a growing scattering length for the effective boson system , which enhances the formation of a cdw . a cdw with vanishing spin gap at @xmath13 ,
can be easily understood heuristically .
one can imagine one hole localized on every second rung , with the electrons on doubly occupied rungs forming spin singlets , as illustrated in fig .
[ fig : per2 ] , giving an effective @xmath192 heisenberg chain with lattice spacing 2 which has vanishing spin gap . on the other hand , in the spin - gapped phase we can think of pairs of holes localized on every fourth rung , as in fig .
[ fig : per4 ] .
thus we expect the groundstate density oscillations to have wavelength 2 in the cdw with vanishing spin gap but wavelength 4 in the gapped cdw .
the connection between wavelength doubling and the spin gap is related to the lieb - schultz - mattis theorem.@xcite if the wavelength is not doubled to 4 , then gapless excitations are expected .
more precisely , we can prove the existence of gapless excitations at n=3/4 if the groundstate is invariant under translations by 2 and under site parity : @xmath193 and is not ferromagnetic .
( note that _ link _ parity , @xmath194 _ is _ spontaneously broken by the wavelength 2 groundstate illustrated in fig .
[ fig : per2 ] but site parity is not . )
this is proven by considering a long ladder of even length @xmath62 with _ periodic _ b.c.s .
let @xmath195 be a groundstate .
we then consider the variational state obtained by acting on the @xmath195 with the unitary operator : @xmath196,\ ] ] where @xmath197 is the number operator for spin up electrons on rung @xmath198 ( summing over both sites on the rung ) .
it is straightforward to calculate @xmath199 and show that it is @xmath200 provided that the following term vanishes : @xmath201.\ ] ] we can show that this vanishes , for example , by using the assume site parity symmetry to show that : @xmath202 to complete the proof of a low energy state we must prove that @xmath203 does not become @xmath195 in the limit @xmath204 .
we prove this by considering the behavior of @xmath205 under translations by 2 sites , the assumed symmetry of the groundstate .
we fine that under this translation : @xmath206 where @xmath207 is the density of spin up electrons , @xmath208 , where @xmath209 is the total number of spin up electrons and @xmath210 is the number sites on a 2-leg ladder of length @xmath62 .
if we further assume that @xmath211 , i.e. that the groundstate is not ferromagnetic , then we may replace the exponential factor in eq . ( [ utr ] ) by @xmath212 . for a density , @xmath13 , this factor is -1 .
this proves orthogonality of @xmath203 with @xmath195 and hence the existence of a low energy excitation . on the other hand , if the groundstate is only invariant under translations by 4 sites then , repeating the argument with a translation by 4 sites changes the exponential factor to @xmath213 which is 1 for n=3/4 .
thus we can not prove that @xmath203 is orthogonal to the groundstate in this case ; it may approach it in the limit @xmath214 , so the proof collapses .
unfortunately , this does not prove that there is a gap when the groundstate has wavelength 4 , only that the gap neccessarily vanishes when it has wavelength 2 .
however , it often appears to be the case that when the lsm theorem fails , a gap appears .
for instance , for s=1/2 spin chains , a gap is expected whenever the groundstate has wavelenth 2 , as in a dimerized state .
thus , the nature of the density oscillations provides additional evidence for which phase the system is in . to see what fourier components are present in the density oscillations ,
we fourier transform the density as a function of rung position and plot the power spectrum . in order to avoid spurious edge effects ,
a windowing function is used . if the original density is @xmath215 , we fourier transform @xmath216 , where @xmath217 is a smooth windowing function .
@xmath217 is chosen to vanish at @xmath161 and @xmath108 , and to be unity for the middle third of the @xmath157 values , with a smooth continuation in between ; the particular choice we use is given in ref .
( ) . before fourier transforming
, we calculate the sum @xmath218 of @xmath219 from @xmath59 to @xmath62 , and the sum @xmath14 of @xmath217 , and then subtract from @xmath219 the function @xmath220 .
this removes a large peak centered at @xmath221 which is uninteresting .
the lattice spacing is set to 1 , so that the exponential in the ft is @xmath222 , where j runs over integer lattice sites . with this approach
there is no restriction on the allowed values of @xmath223 ; the finite value of @xmath62 instead leads to finite widths for the peaks .
the density oscillations for @xmath224 and also @xmath225 , with @xmath166 have a very strong component at a wavelength of 4 , corresponding to a spin - gap , as shown in figure [ 192j.25 ] .
corresponding to a wavelength of 4.,width=302 ]
in this section we briefly discuss a possible phase diagram and the qualitative behavior of @xmath36 as a function of @xmath14 and @xmath58 for the 2-leg @xmath0 model .
we are interested in the region @xmath226 where both the even and odd parity bands have carriers in the weak coupling hubbard limit . at still lower doping the system
is expected to enter a c1s1 phase . in the region
@xmath227 our phase diagram is similar to that proposed by hayward and poilblanc,@xcite and is based on their exact diagonalization results , earlier dmrg results , our new dmrg results and some general results which follow from bosonization and rg .
this is sketched in fig .
[ fig : phase ] . here
the two ordered cdw phases at n=0.75 and n=0.5 are indicated by solid cuts for small values of j / t and a phase separation region occurs for large j / t values .
we conjecture that the remaining region is in a c1s0 ( d - wave-4k@xmath228 cdw ) phase . as discussed in sec .
ii , this phase has the @xmath30 and @xmath29 bosons pinned and is characterized by power law d - wave pairing and 4k@xmath228 cdw correlations .
when @xmath229 , the pairing correlations are dominant . supporting this conjecture , dmrg calculations for j / t and n values near
the phase separation boundary show clear power law d - wave like pairing correlations .
the phase diagram shown in fig .
[ fig : phase ] for the @xmath0 ladder differs in several ways from that expected for the 2-leg repulsive u hubbard model .
firstly , it is generally believed that the hubbard ladder does nt exhibit phase separation and the ordered cdw phases are absent . secondly ,
weak coupling renormalization - group studies , bosonization [ 4 ] and dmrg calculations suggest that @xmath230 for the repulsive u hubbard model .
for an extended hubbard model with additional interactions one can have @xmath231 ; however orignac and giamarchi@xcite argue that when @xmath232 , the d - wave-4k@xmath228 cdw sector gets replaced by a c1s0 s - wave pairing or an orbital antiferromagnetic phase .
this raises questions regarding the phase diagram show in fig .
[ fig : phase ] .
we believe that while alternate c1s0 phases are certainly logical possibilities , they do not occur in the @xmath0 ladder .
the argument that they occur whenever @xmath232 appears to depend upon a weak coupling analysis which may not be applicable to the @xmath0 ladder .
in this analysis , the phase boundaries between the various c1s0 phases is determined by the equality of 2 small marginal coupling constants in the hamiltonian .
these coupling constants also determine @xmath36 , in the weak coupling limit , in such a way that when they are equal , @xmath233 .
( 35 ) of ref .
( ) . ] for stronger coupling , and , in particular , for the @xmath0 model , the possible phase boundary to a phase with orbital antiferromagnetism
need not be related in any general way to the value of @xmath36 and the dmrg results suggest that orbital antiferromagnetism does not occur in the ordinary @xmath0 model.@xcite the phase separation boundary was obtained in a previous drmg study.@xcite as this boundary is approached the compressibility diverges and @xmath234 .
we have taken into account the behavior of @xmath36 near commensurate fillings in drawing fig .
[ fig : phase ] . in particular , at @xmath235 ,
the @xmath0 model reduces to the 2-leg heisenberg spin ladder and has an infinite gap for all charge excitations and presumably a gap of @xmath236 for spin excitations .
the general arguments of schulz@xcite then imply that @xmath57 along the line @xmath60 .
we have also assumed that a cdw occurs at n=.75 and n=.5 for small enough @xmath14 .
as discussed in the previous section , we have found some evidence for such a phase at @xmath181 and n=.75 .
the corresponding behavior of @xmath36 , reviewed in the previous section is incorporated in fig .
[ fig : phase ] .
note that , at small @xmath14 , @xmath36 apparently varies rapidly between .0625 for n=.75 and 1 for n=1 . as we also discussed in the previous section ,
it is possible that the spin gap vanishes at commensurate fillings for some range of @xmath14 .
we do not attempt to indicate this possibility in fig .
[ fig : phase ] . for values of @xmath138 around .35 , which has been argued to be a reasonable value for modeling the cuprates , @xmath237 near n=.75 but increases towards 1 as n=1 is approached .
thus there is only a narrow window of doping near 1/2-filling where d - wave pairing correlations dominate .
ladder as a function of @xmath138 and the electron density @xmath58 in the region @xmath58 greater than about .5.,width=302 ]
at incommensurate filling , or in general when no cdw occurs , there will still be density oscillations produced by the boundaries of an open ladder which only decay slowly ( with a power law ) into the ladder .
detailed predictions can be made about these using bosonization generalizing the approach of ref .
( ) for the spin chain and hubbard chain to the case of the ladder .
these oscillations provide a further check on the theory .
thus consider again the @xmath52 term in the density operator , of eq .
( [ density ] ) : @xmath238 where @xmath239 is a constant into which we have adsorbed @xmath240 . as we discussed in sec .
ii , this @xmath52 term in the continuum representation of @xmath241 , determines the leading behavior of the density - density correlation function at long distances since any @xmath49 terms decay exponentially . for an infinite system , the correlation function of the exponential operator appearing in eq .
( [ density2 ] ) , behaves as : @xmath242 where @xmath243 is a constant , with dimensions of ( length)@xmath244 . decomposing @xmath245 , we may write this correlation function as a square of two identical factors , the correlation functions of @xmath246 $ ]
. it thus follows that the oscillating term in the density correlation function is : @xmath247 now consider the semi - infinite system ( @xmath248 ) with one free boundary condition . as remarked in sec .
ii , this corresponds to a b.c . on the boson field given in eq .
( [ obc ] ) , or equivalently : @xmath249 now using the fact that @xmath250 is a function of @xmath251 only and @xmath252 a function of @xmath253 only , we see that this equation implies that we may regard @xmath250 as the analytic continuation of @xmath252 to the negative axis : @xmath254 thus the expectation value of @xmath255 reduces essentially to the square root of the correlation function @xmath256 : @xmath257 the wave - vector of the friedel oscillations is the same as the wave - vector governing the long - distance oscillations in the correlation function .
no friedel oscillations occur at wave - vector @xmath258 as they would for the non - interacting system or more generally in phases where , for example , all 4 bosons are gapless .
the exponent governing the decay of the friedel oscillations is @xmath36 , 1/2 the exponent governing the decay of density correlations .
furthermore , the amplitude is simply the square root of the amplitude of the density correlation function ( multiplied by @xmath259 ) .
finally , we may readily generalize eq .
( [ frie - inf ] ) to the case of a finite chain of length @xmath62 , by a standard conformal transformation : @xmath260^{k_{+\rho}}}.\label{fried - fin}\ ] ] the decay of friedel oscillations allows a way of determining the critical exponent @xmath36 which is alternative to measuring directly the long distance behavior of the density correlations . indeed this latter measurement becomes quite difficult with open b.c.s due to the necessity of eliminating boundary effects .
we emphasize that the wave - vector of these friedel oscillations @xmath261 , or equivalently @xmath262 where @xmath123 is the hole density , is a characteristic feature of the particular c1s0 phase that we are assuming .
more correctly , this is the _ minimum _ friedel oscillation wave - vector since higher harmonics are also expected to occur .
this minimum oscillation wave - vector would be different in a different phase .
this wave - vector corresponds to 2 holes per wave - length and is the same wave - vector that would occur for a 1-component spinless hard core bose gas , which is an approximate description of the c1s0 phase in which hole pairs are assumed to form tightly bound rung singlets.@xcite an immediate consequence of eq .
( [ fried - fin ] ) is that the amplitude of the density oscillations near the center of a finite chain scales as @xmath183 .
we show a log - log plot of this amplitude versus length in fig .
[ figampall ] . fitting to a straight line
allows a determination of @xmath36 .
the corresponding values of @xmath36 determined in this way are shown in table [ table : v , k ] .
we see that they are roughly comparable to the values obtained from the finite size spectrum . , of the friedel oscillations in the center of the system , as a function of the length , @xmath62 .
the resulting slopes are @xmath263 and are given in table [ table : v , k ] .
the triangles show results for @xmath264 as a function of the number of block states kept , @xmath67 , for @xmath180 , @xmath166 .
one can see the very slow convergence of the amplitude with the number of states kept .
, width=302 ] in figs .
[ fig : n_j.5 ] and [ fig : n_j.35 ] we show the friedel oscillations in the density at site @xmath265 , for two different values of @xmath266 , fitted to eq .
( [ fried - fin ] ) with @xmath36 taken as a free parameter .
the agreement is fair although the presence of corrections due to irrelevant operator effects is evident . in both cases
the value of @xmath36 so determined is in rough agreement with the values in the third column of table i , determined from the finite size spectrum .
we emphasize that these formulas are true very generally for luttinger liquids with a single gapless charge boson .
in particular , they apply to the spinless single chain model . in the non - interacting case
we may readily find the exact formula for the friedel oscillations .
for @xmath271 electrons on @xmath62 sites : @xmath272 \over 2(l+1)\sin [ ( \pi j/(l+1)]}.\end{aligned}\ ] ] for large @xmath271 and @xmath62 this can be approximated : @xmath273 this has the expected form of eq .
( [ fried - fin ] ) with @xmath274 and @xmath275 . on the other hand ,
the @xmath49 part of the density correlation function at long distances is : @xmath276 which has the form of eq .
( [ dens - corr ] ) with the same value of @xmath275 .
we see that , not surprisingly , when the charge density correlations drop off slowly , so do the friedel oscillations .
in particular , this makes it difficult to determine numerically whether or not a cdw occurs at @xmath13 , for example , by measuring density oscillations .
if @xmath138 is such that the system almost has a cdw then @xmath36 will be only slightly greater than 1/8 .
the extremely slow decay of the friedel oscillations will be difficult to distinguish from a true cdw , where the oscillation amplitude goes to a non - zero constant far from the boundaries .
we now consider the single chain hubbard and @xmath145 models .
the hubbard model is known to be in a c1s1 phase for all densities ( except @xmath60 ) and all @xmath277 .
thus the low energy effective hamiltonian contains both spin and charge bosons .
( we use the same notation as for the 2-leg ladder except that the @xmath278 labels are no longer needed since there is only one type of charge boson and one type of spin boson . )
the charge boson hamiltonian is written exactly as in eq .
( [ ham ] ) , in terms of the parameter @xmath279 .
the spin boson hamiltonian also has exactly this form but with @xmath280 , as follows from @xmath281 spin rotation invariance .
the pairing , @xmath49 and @xmath52 density operators now take the form : @xmath282 note that , in this case , we have formed a @xmath52 operator in which the spin boson does nt appear .
it then follows that the pair correlations , @xmath49 density correlations and @xmath52 density correlations decay with power law exponents : @xmath283 open b.c.s imply : @xmath284 it thus follows that the density will exhibit both @xmath49 and @xmath52 friedel oscillations with different exponents : @xmath285 where we have expressed the oscillation wave - vectors in terms of the electron density using the free electron result for electrons with spin on a single chain : @xmath286 the amplitudes , @xmath239 and @xmath287 are proportional to the square roots of the corresponding terms in the density correlation function and depend on density and @xmath205 ( as does @xmath288 .
it is interesting to consider the limit @xmath289 where the no double occupancy constraint is present .
if we consider some initial configuration of holons and spins , then the order of the spins along the chain can never change under nearest neighbor hopping processes consistent with no double occupancy .
it follows that the charge dynamics must be identical to those of a system of _ non - interacting _ spinless fermions with the same density .
that is to say , the charge dynamics is `` spin blind '' .
it follows that the density oscillations should not have the @xmath290 term in this limit but only the @xmath291 term as for spinless fermions [ eq .
( [ free ] ) ] ; i.e. that @xmath292 at @xmath293 .
this argument also determines @xmath294 and @xmath295 in this limit .
we remark that this argument is special to the single chain model with only nearest neighbor hopping . on a ladder , or even on a chain with next nearest neigbor hopping , spin rearrangement is possible without double occupancy by moving the electrons around or past each other . in fig .
[ figallu ] we plot the fourier transform of the density for the hubbard model , at 7/8 filling , for various values of @xmath205 .
the same smoothing procedure was used that we mentioned in sec .
iii . in this case
@xmath296 and @xmath297 or equivalently @xmath298 .
both @xmath49 and @xmath52 components are clearly visible and it is evident that the @xmath49 component vanishes at large @xmath205 . in the single chain @xmath145 model at 1/4 filling for a ) @xmath176 and b ) j=1.0 , showing friedel oscillations at @xmath299 and @xmath300.,title="fig:",width=302 ] in the single chain @xmath145 model at 1/4 filling for a ) @xmath176 and b ) j=1.0 , showing friedel oscillations at @xmath299 and @xmath300.,title="fig:",width=302 ] what does this imply about the ( single chain ) @xmath0 model ?
the @xmath301 limit of the @xmath0 model is identical to the @xmath293 limit of the hubbard model .
therefore , in that limit , the density oscillations have vanishing @xmath49 component .
it is not completely obvious what happens at non - zero @xmath14 .
insofar as this is the same as large but finite @xmath205 , we would expect to have a small @xmath49 component to the density oscillations .
note however , that the no double occupancy constraint is enforced exactly in the @xmath0 model , by construction , whereas it is not in the finite @xmath205 hubbard model .
thus we must understand whether it is the no double occupancy which is responsible for the vanishing @xmath49 oscillations or whether the vanishing of all spin rearrangement processes is neccessary .
the argument in the previous paragraph seems to rely on the vanishing of spin rearrangement processes so we suspect that the friedel oscillations should exhibit a @xmath49 component for any finite @xmath14 .
of course , for small @xmath138 this is expected to be small . in fig .
[ fig:1legtj ] we show the fourier transformed density for the single leg @xmath0 model at density @xmath302 for two values of @xmath138 .
note that now @xmath299 and @xmath300 . for smaller @xmath138
the @xmath52 oscillations clearly dominate although a small @xmath49 part is observed . at larger @xmath138 ,
the oscillations at @xmath49 are larger than those at @xmath52 .
thus we expect @xmath145 models to exhibit generic friedel oscillations in all cases except for the special case of a single leg in the limit @xmath303 where the absence of spin rearrangement processes eliminates the lowest wave - vector component of the oscillations .
the friedel oscillation wave - vector can be a useful diagnostic of which phase a particular multi - leg @xmath145 model is in .
for instance , the minimum oscillation wave - vector , @xmath304 , is a characteristic of the particular c1s0 phase of the 2 leg ladder reviewed in sec .
ii ( in which the @xmath29 and @xmath30 fields are pinned ) .
other phases exhibit other oscillation wave - vectors .
we expect such an analysis of dmrg results to be useful in determining the phase diagram of multi - leg ladders .
djs would like to thank t. giamarchi for helpful discussions .
djs and ia acknowlege the hospitality of the itp , university of california , santa barbara where this work was initiated .
srw acknowledges support from the nsf under grant no .
dmr98 - 70930 ; ia acknowleges support from the nsf grant no .
phy99 - 07949 ( itp ) and nserc of canada ; djs acknowledges support from the nsf under grant no .
phy99 - 07949 ( itp ) and grant no . drm98 - 17242 ( djs ) .
999 m. fabrizio , phys . rev . *
b37 * , 325 ( 1993 ) .
schulz , phys .
* b53 * , 2959 ( 1996 ) .
l. balents and m.p.a .
fisher , phys . rev . *
b53 * , 12133 ( 1996 ) .
e. orignac and t. giamarchi , phys . rev .
* b56 * , 7167 ( 1997 ) .
noack , s.r .
white and d.j .
scalapino , phys .
lett . * 73 * , 882 ( 1994 ) .
t. siller , m. troyer , t.m .
rice and s.r .
white , phys . rev . *
b63 * , 195106 ( 2001 ) .
hayward and d. poilblanc , phys . rev . * b53 * , 11721 ( 1996 ) . s. eggert and i. affleck , phys
lett . * 75 * , 934 ( 1995 ) .
r. egger and h. grabert , phys .
lett . * 75 * , 3505 ( 1995 ) .
s.r . white and d.j .
scalapino , phys . rev . *
b61 * , 6320 ( 2000 ) .
schulz , phys .
* b59 * , 2471 ( 1999 ) .
schulz , phys . rev . *
b22 * , 5274 ( 1980 ) . , t. giamarchi , physica * b230 * , 975 ( 1997 ) .
s. eggert and i. affleck , phys . rev . * b46 * , 10866 ( 1992 ) .
m yamanaka , m. oshikawa and i. affleck , phys .
* 79 * , 1110 ( 1997 ) .
m. vekic and s.r .
white , phys .
* 71 * , 4283 ( 1993 ) ; see eq . ( 4 ) and ( 10 ) .
scalapino , s.r .
white and i. affleck , phys . rev .
* b64 * , 100506 ( 2001 ) .
j.b . marston and a. sudbo , cond - mat/0103120 .
s. rommer , s.r .
white and d.j .
scalapino , phys . rev . * b61 * , 14701 ( 1997 ) . | the density matrix renormalization group method for ladders works much more efficiently with open boundary conditions .
one consequence of these boundary conditions is groundstate charge density oscillations that often appear to be nearly constant in magnitude or to decay only slightly away from the boundaries .
we analyse these using bosonization techniques , relating their detailed form to the correlation exponent and distinguishing boundary induced generalized friedel oscillations from true charge density waves .
we also discuss a different approach to extracting the correlation exponent from the finite size spectrum which uses exclusively open boundary conditions and can therefore take advantage of data for much larger system sizes
. a general discussion of the friedel oscillation wave - vectors is given , and a convenient fourier transform technique is used to determine it .
dmrg results are analysed on hubbard and @xmath0 chains and 2 leg @xmath0 ladders .
we present evidence for the existence of a long - ranged charge density wave state in the @xmath0 ladder at a filling of @xmath1 and near @xmath2 . |
the morse potential has raised a great deal of interest over the years and has been one of the most useful models to describe the interaction between two atoms in a diatomic molecule .
it is known that the radial schrdinger equation for this potential can be solved exactly when the orbital angular quantum number @xmath0 is equal to zero @xcite . on the other hand
, it is also known that for @xmath1 , one has to use some approximations to find analytical or semi - analytical solutions .
several schemes have been presented for obtaining approximate solutions @xcite . among these approximations ,
the most widely used and convenient one is the pekeris approximation @xcite , which is based on the expansion of the centrifugal barrier in a series of exponentials depending on the internuclear distance up to the second order .
other approximations have also been developed to find better analytical formulas for the rotating morse potential .
however , all these approximations other than the pekeris one require the calculation of a state - dependent internuclear distance through the numerical solutions of transcendental equations @xcite
. in this respect , the rotating morse potential has so far been solved by the super - symmetry ( susy ) @xcite , the nikiforov - uvarov method ( nu ) @xcite , the shifted and modified shifted 1/n expansion methods @xcite as well as the variational method @xcite using pekeris approximations for @xmath2 .
it is also solved by using the hypervirial perturbation method ( hv ) @xcite with the full potential without pekeris approximation . in this paper ,
our aim is to solve the rotating morse potential using a different and more practical method called , the asymptotic iteration method ( aim ) @xcite within the pekeris approximation and to obtain the energy eigenvalues and corresponding eigenfunctions . in the next section , the asymptotic iteration method ( aim )
is introduced .
then , in section [ apply ] , the schrdinger equation is solved by the asymptotic iteration method with the non - zero angular momentum quantum numbers for the rotating morse potential : the exact energy eigenvalues and corresponding wave functions are calculated for the @xmath3 , @xmath4 , @xmath5 and @xmath6 diatomic molecules and aim results are compared with the findings of the susy @xcite , the hypervirial perturbation method ( hv ) @xcite , the nikiforov - uvarov method ( nu ) @xcite and the shifted and modified shifted 1/n expansion methods @xcite as well as with the variational method @xcite .
finally , section [ conclude ] is devoted to the summary and conclusion .
we briefly outline the asymptotic iteration method here and the details can be found in references @xcite .
the asymptotic iteration method is proposed to solve the second - order differential equations of the form @xmath7 where @xmath8 and s@xmath9(x ) , @xmath10(x ) are in c@xmath11(a , b ) .
the variables , s@xmath9(x ) and @xmath10(x ) , are sufficiently differentiable .
the differential equation ( [ diff ] ) has a general solution @xcite @xmath12\ ] ] if @xmath13 , for sufficiently large @xmath14 , we obtain the @xmath15 values from @xmath16 where @xmath17 the energy eigenvalues are obtained from the quantization condition .
the quantization condition of the method together with equation ( [ iter ] ) can also be written as follows @xmath18 for a given potential such as the rotating morse one , the radial schrdinger equation is converted to the form of equation ( [ diff ] ) .
then , s@xmath19 and @xmath20 are determined and s@xmath21 and @xmath22 parameters are calculated .
the energy eigenvalues are determined by the quantization condition given by equation ( [ kuantization ] ) .
however , the wave functions are determined by using the following wave function generator @xmath23
the motion of a particle with the reduced mass @xmath24 is described by the following schrdinger equation : @xmath25+v(r ) \right)\psi_{n\ell m}(r,\theta,\phi)=e\psi_{n\ell m}(r,\theta,\phi)\label{sch}\ ] ] the terms in the square brackets with the overall minus sign are the dimensionless angular momentum squared operator , @xmath26 . defining @xmath27 , we obtain the radial part of the schrdinger equation : @xmath28u_{n\ell}(r)+\frac{2 \mu e}{\hbar^{2}}u_{n\ell}(r)=0 \label{radialr}\end{aligned}\ ] ] it is sometimes convenient to define @xmath29 and the effective potential as follows : @xmath30 since @xmath31 the radial schrdinger equation given by equation ( [ radialr ] ) follows that @xmath32r_{n\ell}(r)=0 \label{factorschrodinger}\ ] ] instead of solving the partial differential equation ( [ sch ] ) in three variables @xmath33 , @xmath34 and @xmath35 , we now solve a differential equation involving only the variable @xmath33 , but dependent on the angular momentum parameter @xmath0 , which makes the solution of this equation difficult for @xmath1 or sometimes impossible within a given potential .
the morse potential we examine in this paper is defined as @xmath36 with @xmath37 and @xmath38 . here ,
@xmath39 and @xmath40 denote the dissociation energy and morse parameter , respectively .
@xmath41 is the equilibrium distance ( bound length ) between nuclei and @xmath42 is a parameter to control the width of the potential well . for the @xmath3 diatomic molecule , the effective potential , which is the sum of the centrifugal and morse potentials ,
is shown in figure [ pot ] for various values of the orbital angular momentum .
the superposition of the attractive and repulsive potentials results in the formation of a potential pocket , whose the width and depth depend on the orbital angular momentum quantum number for a given molecular potential . the potential pocket becomes shallower as the orbital angular momentum quantum number @xmath0 increases , which also indicates that the number of states supported by the potential decreases .
this pocket is also very important for the scattering case due to the interference of the barrier and internal waves , which creates the oscillatory structure in the cross - section .
the effect of this pocket can be understood in terms of the interference between the internal and barrier waves that corresponds to a decomposition of the scattering amplitude into two components , the inner and external waves @xcite .
the effective potential together with the morse potential for @xmath1 can be written as , @xmath43 it is known that the schrdinger equation can not be solved exactly for this potential for @xmath1 by using the standard methods such as susy and nu . as it is seen from equation ( [ veff ] ) ,
the effective potential is a combination of the exponential and inverse square potentials , which can not be solved analytically .
therefore , an approximation has to be made : the most widely used and convenient one is the pekeris approximation .
this approximation is based on the expansion of the centrifugal barrier in a series of exponentials depending on the internuclear distance , keeping terms up to second order , so that the effective @xmath0-dependent potential keeps the same form as the potential with @xmath0=0 @xcite .
it should be pointed out , however , that this approximation is valid only for low vibrational energy states . in the pekeris approximation , by change of the coordinates @xmath37 , the centrifugal potential is expanded in a series around @xmath44 @xmath45 where @xmath46 .
taking up to the second order degrees in this series and writing them in terms of exponentials , we get @xmath47 in order to determine the constants @xmath48 , @xmath49 and @xmath50 , we also expand this potential in a series of @xmath51 @xmath52 comparing equal powers of equations ( [ expand ] ) and ( [ exp ] ) , we obtain the constants @xmath48 , @xmath49 and @xmath50 as , @xmath53 now , the effective potential with pekeris approximation becomes , @xmath54 instead of solving the radial schrdinger equation for the effective potential given by equation ( [ veff ] ) , we solve the radial schrdinger equation for the new effective potential given by equation ( [ veffpekeris ] ) obtained by using the pekeris approximation .
inserting this effective potential equation ( [ veffpekeris ] ) into equation ( [ factorschrodinger ] ) and using the following _ ansatzs _ @xmath55 the radial schrdinger equation takes the following form : @xmath56 if we rewrite equation ( [ radyal ] ) by using a new variable of the form @xmath57 , we obtain @xmath58r_{n\ell}(y)=0\ ] ] in order to solve this equation with aim for @xmath59 , we should transform this equation to the form of equation ( [ diff ] ) .
therefore , the reasonable physical wave function we propose is as follows @xmath60 if we insert this wave function into the equation ( [ trans ] ) , we have the second - order homogeneous linear differential equations in the following form @xmath61 which is now amenable to an aim solution . by comparing this equation with equation ( [ diff ] )
, we can write the @xmath62 and @xmath63 values and by means of equation ( [ iter ] ) , we may calculate @xmath64 and @xmath65 .
this gives ( the subscripts are omitted ) : @xmath66 combining these results with the quantization condition given by equation ( [ kuantization ] ) yields @xmath67 when the above expressions are generalized , the eigenvalues turn out as @xmath68 using equation ( [ ansatz ] ) , we obtain the energy eigenvalues e@xmath69 , @xmath70^{2}+\gamma c_{0}\ ] ] as it is seen that the energy eigenvalue equation is easily obtained by using aim .
this is the advantage of the aim that it gives the eigenvalues directly by transforming the radial schrdinger equation into a form of @xmath71 = @xmath72 . in order to test the accuracy of equation ( [ energyeigenvalues ] )
, we calculate the energy eigenvalues of the @xmath3 , @xmath4 , @xmath5 and @xmath6 diatomic molecules .
the aim results are compared with those obtained by susy method @xcite using original pekeris approximation , the hypervirial perturbation method ( hv ) @xcite , the shifted 1/n and modified shifted 1/n expansion methods @xcite for the @xmath3 diatomic molecule in table [ table1 ] . in table [ table2 ] , we show the same comparison for the @xmath4 diatomic molecule . furthermore , the aim results are compared with those obtained by nu method @xcite , shifted 1/n and modified shifted 1/n expansion methods @xcite for the @xmath5 and @xmath6 diatomic molecules in tables [ table3 ] and [ table4 ] , respectively . as it can be seen from the results presented in these tables that the aim results are in good agreement with the findings of the other methods . after we find the energy eigenvalues , the following wave function generator can be used to find @xmath73 functions by using aim @xmath74where @xmath75 represents the radial quantum number and @xmath14 shows the iteration number .
below , the first few @xmath76 functions can be seen @xmath77 it can be understood from the results given above that we can write the general formula for @xmath73 as follows , @xmath78 thus , we can write the total radial wave function as below , @xmath79 when the hypergeometric function is written in terms of the laguerre polynomials , we get @xmath80 where @xmath81 is the normalization constant and can be obtained from @xmath82^{2}dy=1 $ ] as below , @xmath83 where @xmath84 .
we have shown an alternative method to obtain the energy eigenvalues and corresponding eigenfunctions of the rotating morse potential using pekeris approximation within the framework of the asymptotic iteration method .
the main results of this paper are the energy eigenvalues and eigenfunctions , which are given by equations ( [ energyeigenvalues ] ) and ( [ wave1 ] ) respectively .
the energy eigenvalues are obtained for the @xmath3 , @xmath4 , @xmath5 and @xmath6 diatomic molecules .
our aim results are compared with the findings of the other methods such as the susy @xcite , the hypervirial perturbation method ( hv ) @xcite , the nikiforov - uvarov method ( nu ) @xcite and the shifted and modified shifted 1/n expansion methods @xcite as well as the variational method @xcite in tables [ table1 ] , [ table2 ] , [ table3 ] and [ table4 ] .
the advantage of the asymptotic iteration method is that it gives the eigenvalues directly by transforming the radial schrdinger equation into a form of @xmath71 = @xmath72 .
the wave functions are easily constructed by iterating the values of @xmath85 and @xmath86 .
the method presented in this study is a systematic one and it is very efficient and practical .
it is worth extending this method to the solution of other interaction problems .
this paper is an output of the project supported by the scientific and technical research council of turkey ( tbitak ) , under the project number tbag-2398 and erciyes university ( fba-03 - 27 , fbt-04 - 15 , fbt-04 - 16 ) .
authors would also like to thank dr .
aye boztosun for the proofreading as well as dr .
hakan ifti for useful comments on the manuscript .
99 p. m. morse , phys . rev . *
34 * ( 1929 ) 57 .
d. a. morales , chem .
letters * 394 * ( 2004 ) 68 . c. l. pekeris , phys
* 45 * ( 1934 ) 98 .
s. flgge , ` practical quantum mechanics ` vol .
i , springer , berlin , ( 1994 ) .
depristo , j. chem . phys . * 74 * ( 1981 ) 5037
elsum and g. gordon , j. chem . phys . *
76 * ( 1982 ) 5452 .
morales , chem .
* 161 * ( 1989 ) 253 .
et al . _ ,
phys . rev .
* a46 * , ( 1992 ) 9 .
f. cooper , a. khare and u. sukhatme , phys .
* 251 * ( 1995 ) 267 .
a. f. nikiforov and v. b. uvarov , ` special functions of mathematical physics ` , birkhuser , basel , ( 1988 ) . c. berkdemir and j. han , chem .
letters , * 409 * ( 2005 ) 203
. t. imbo and u. sukhatme , phys .
. lett . * 54 * ( 1985 ) 2184 .
e. d. filho and r. m. ricotta , phys . lett . * a269 * ( 2000 ) 269 .
killingbeck , a. grosjean and g. jolicard , j. chem . phys .
* 116 * ( 2002 ) 447 .
h. ciftci , r. l. hall and n. saad , j. phys . a : math
* 36 * ( 2003 ) 11807 .
h. ciftci , r. l. hall and n. saad , phys . lett . * a340 * ( 2005 ) 388 .
lee , nucl . phys . *
a311 * ( 1978 ) 518 .
brink and n. takigawa , nucl .
a279 * ( 1977 ) 159 .
i. boztosun , phys . rev .
c * 66 * ( 2002 ) 024610 .
ccccccccccc & & & & & & & + & & & & & & & + 0 & 0 & -4.47601 & -4.47601 & -4.47601 & -4.4758 & -4.4760 & -4.4749 + & 5 & -4.25880 & -4.25880 & -4.25901 & -4.2563 & -4.2590 & -4.2590 + & 10 & -3.72193 & -3.72193 & -3.72473 & -3.7187 & -3.7247 & -3.7247 + + 5 & 0 & -2.22052 & -2.22051 & -2.22051 & - & -2.2205 & -2.2038 + & 5 & -2.04355 & -2.04353 & -2.05285 & - & -2.0530 & -2.0525 + & 10 & -1.60391 & -1.60389 & -1.65265 & - & -1.6535 & -1.6526 + + 7 & 0 & -1.53744 & -1.53743 & -1.53743 & - & -1.5374 & -1.5168 + & 5 & -1.37656 & -1.37654 & -1.39263 & - & -1.3932 & -1.3887 + & 10 & -0.97581 & -0.97578 & -1.05265 & - & -1.0552 & -1.0499 + ccccccccccc & & & & & + & & & & & + 0 & 0 & -4.4356 & -4.4360 & -4.4355 & -4.4352 + & 5 & -4.3968 & -4.3971 & -4.3968 & -4.3967 + & 10 & -4.2941 & -4.2940 & -4.2940 & -4.2939 + + 5 & 0 & -2.8051 & - & -2.8046 & -2.7727 + & 5 & -2.7721 & - & -2.7718 & -2.7508 + & 10 & -2.6847 & - & -2.6850 & -2.6712 + + 7 & 0 & -2.2570 & - & -2.2565 & -2.2002 + & 5 & -2.2263 & - & -2.2262 & -2.1874 + & 10 & -2.1451 & - & -2.1461 & -2.1194 + ccccccccccccc & & & & & & & & & & & & + & & & & & & & & & & & & + 0 & & 0 & & -11.0915 & & -11.091 & & -11.093 & & -11.092 & & -11.091 + & & 5 & & -11.0844 & & -11.084 & & -11.085 & & -11.084 & & -11.084 + & & 10 & & -11.0653 & & -11.065 & & -11.066 & & -11.065 & & -11.065 + + 5 & & 0 & & -9.7952 & & -9.795 & & - & & -9.795 & & -9.788 + & & 5 & & -9.7883 & & -9.788 & & - & & -9.788 & & -9.782 + & & 10 & & -9.7701 & & -9.769 & & - & & -9.770 & & -9.765 + + 7 & & 0 & & -9.2992 & & -9.299 & & - & & -9.299 & & -9.286 + & & 5 & & -9.2925 & & -9.292 & & - & & -9.292 & & -9.281 + & & 10 & & -9.2745 & & -9.274 & & - & & -9.274 & & -9.265 + ccccccccccccc & & & & & & & & & & & & + & & & & & & & & & & & & + 0 & & 0 & & -2.4289 & & -2.4287 & & -2.4291 & & -2.4280 & & -2.4278 + & & 5 & & -2.4013 & & -2.4012 & & -2.4014 & & -2.4000 & & -2.3999 + & & 10 & & -2.3288 & & -2.3287 & & -2.3287 & & -2.3261 & & -2.3261 + + 5 & & 0 & & -1.6477 & & -1.6476 & & - & & -1.6402 & & -1.6242 + & & 5 & & -1.6238 & & -1.6236 & & - & & -1.6160 & & -1.6074 + & & 10 & & -1.5607 & & -1.5606 & & - & & -1.5525 & & -1.5479 + + 7 & & 0 & & -1.3776 & & -1.3774 & & - & & -1.3682 & & -1.3424 + & & 5 & & -1.3550 & & -1.3549 & & - & & -1.3456 & & -1.3309 + & & 10 & & -1.2958 & & -1.2957 & & - & & -1.2865 & & -1.2781 + | for non - zero @xmath0 values , we present an analytical solution of the radial schrdinger equation for the rotating morse potential using the pekeris approximation within the framework of the asymptotic iteration method .
the bound state energy eigenvalues and corresponding wave functions are obtained for a number of diatomic molecules and the results are compared with the findings of the super - symmetry , the hypervirial perturbation , the nikiforov - uvarov , the variational , the shifted 1/n and the modified shifted 1/n expansion methods . |
lattice gauge field theory brings the representation theory of an underlying manifold and its quantum invariants into the same setting .
consider the case of a cylinder over a compact , oriented surface with boundary .
a lattice model of the surface determines an algebra of gauge invariant fields ( i.e. observables ) . in the classical case , based on a connected , simply connected lie group @xmath0 , observables are the characters of the fundamental group of the lattice represented in @xmath0 .
wilson loops can be understood as traces of conjugacy classes in the fundamental group of the lattice .
for the theory based on a drinfeld - jimbo deformation of a simple lie algebra @xmath1 , the observables are a deformation quantization of the @xmath0-characters of the surface with respect to the standard poisson structure @xcite . in the case of @xmath2
this , together with the classical isomorphism @xcite , allowed us to prove that the algebra of observables is the kauffman bracket skein algebra of the surface , completed as an algebra over formal power series .
in this paper we return to an analytic setting in which the deformation parameter is any complex number other than a root of unity . the analogous theorem relating observables and the kauffman bracket skein algebra is again true . the proof is based on the combinatorial equivalence between the temperley - lieb algebra and the quantized invariant theory of @xmath3 ; it does not explicitly use the relationship with surface characters .
the paper is organized as follows .
section [ kauffman ] recalls definitions and associated formulas of the kauffman bracket skein algebra .
section [ reps ] summarizes , for the generic parameter , the construction of a quasi - triangular matrix model of quantum @xmath3 .
section [ lgft ] outlines basic definitions and constructions of lattice gauge field theory .
section [ temperley ] describes the correspondence between skeins and intertwiners in the verlinde algebra for quantum @xmath3 .
finally , section [ observable ] contains a proof of the main theorem .
let @xmath4 be a complex number that is neither @xmath5 nor a root of unity .
suppose that @xmath6 is a compact , oriented surface with boundary and @xmath7 is a closed interval .
denote by @xmath8 the set of isotopy classes of framed links in @xmath9 , including the empty link .
let @xmath10 be the vector space with basis @xmath8 .
define @xmath11 to be the subspace of @xmath10 spanned by all expressions of the form @xmath12 and @xmath13 , where the framed links in each expression are identical outside the balls pictured in the diagrams .
the kauffman bracket skein algebra @xmath14 is the quotient @xmath15 .
multiplication is given by laying one link over the other .
more precisely , if @xmath16 and @xmath17 are in @xmath8 , isotop them so that @xmath16 lies in @xmath18 , and @xmath17 in @xmath19 $ ] . then @xmath20 is the union of these two links in @xmath21 $ ] .
extend linearly to a product on @xmath10 . since @xmath11 is an ideal
, the product descends , making the skein module into a skein algebra .
since the algebra structure depends on the specific product structure of @xmath9 , rather than its topological type , we use the notation @xmath22 .
we use the standard convention of modeling a skein in @xmath22 on a framed , admissibly colored , trivalent graph .
an admissible coloring is an assignment of a nonnegative integer to each edge so that the colors at each vertex form admissible triples .
a triple @xmath23 is admissible if @xmath24 , @xmath25 , @xmath26 and @xmath27 is even .
the corresponding skein in @xmath22 is obtained by replacing each edge labeled with the letter @xmath28 by the @xmath28-th jones wenzl idempotent ( see @xcite , or @xcite ) , and replacing trivalent vertices with kauffman triads ( see ( * ? ? ?
* fig . 14.7 ) ) . if @xmath29 is a trivalent spine of @xmath6 then the set of skeins carried by admissible colorings of @xmath29 forms a basis @xmath30 for @xmath22 .
if @xmath6 is an annulus , @xmath30 consists of skeins obtained by labeling the core with a jones wenzl idempotent .
one may think of the core as a `` trivalent '' spine with one vertex , whose admissible labels are @xmath31 .
the space @xmath22 also has a basis @xmath32 consisting of all links with simple diagrams on @xmath6 , i.e. with no crossings and no trivial components .
the details of the following can be found in @xcite .
let @xmath33 be the unital hopf algebra on @xmath34 , @xmath35 , @xmath36 , @xmath37 , with relations : @xmath38 let @xmath39 denote the irreducible @xmath40-dimensional representation of @xmath33 .
fixing an ordered basis for @xmath39 we define linear functionals @xmath41 to be the coefficient in the @xmath42-th row and @xmath43-th column in the representation @xmath39 . the @xmath44 form a basis for the stable subalgebra @xmath45 of the hopf algebra dual @xmath46 .
( here @xmath47 . )
define @xmath48 and give it the product topology .
a typical element of @xmath49 is a sequence of arbitrarily chosen matrices in which the @xmath42-th term is an @xmath50 matrix .
let @xmath51 be the homomorphism corresponding to the representation @xmath39 .
the homomorphism @xmath52 given by @xmath53 , is injective and its image is dense in @xmath49 ( see @xcite ) . the algebra @xmath49 is the completion of @xmath33 by equivalence classes of cauchy sequences in the weak topology from @xmath45 .
it has the structure of a topological ribbon hopf algebra .
the projection of @xmath49 onto its @xmath40-st factor is an irreducible representation of @xmath49 , also denoted by @xmath39 . composing @xmath44 with this projection yields a function on @xmath49 , also called @xmath44 .
thus @xmath45 is understood to lie in @xmath54 .
in this section we recall basic definitions and constructions of lattice gauge field theory .
details can be found in @xcite .
let @xmath55 be an oriented , ciliated graph ; i.e. the edges are oriented and the vertices carry a linear ordering of the adjacent edges .
one can think of this as instructions for building a strip and disk model of an oriented surface @xmath6 having @xmath55 as a strong deformation retract .
vertices correspond to the disks , edges to strips , and the ciliation determines how to glue the strips to the disks . the surface @xmath6 is called the envelope of @xmath55 .
define a space of connections : @xmath56 and define an algebra of fields : @xmath57 = \bigotimes_{\text{edges $ e$ } } ( { { _ qsl_2}})_e.\ ] ] note that fields are functions on connections in the obvious way .
the connections form a coalgebra with comultiplication as defined in @xcite .
multiplication of fields is the convolution product dual to comultiplication of connections .
there is an action of the gauge algebra , @xmath58 on the space of connections , and adjointly on fields .
the invariant part of the gauge fields under this action is called the observables , @xmath59 .
the multiplication of fields restricts to make @xmath59 into an algebra , which is a deformation of the @xmath60-characters of @xmath61 .
let @xmath62 be a representation of the hopf algebra @xmath49 , that is @xmath62 is a finite dimensional left module over @xmath49 .
the dual vector space to @xmath62 carries two distinct @xmath49-module structures . when @xmath49 acts on the left , the dual module is denoted by @xmath63 .
the action is : @xmath64 for any @xmath65 , @xmath66 and @xmath67 in the dual vector space to @xmath62 .
when @xmath49 acts on the right , denote the dual by @xmath68 with @xmath69 there is an alternate description of observables in terms of `` colorings '' of the lattice . assigning a representation @xmath70 to each positively oriented edge @xmath71 of the lattice
@xmath55 determines a map @xmath72 this yields , at each vertex , a tensor product of representations coming from the edges adjoining that vertex taken in the order given by the ciliation .
use the representation @xmath70 for the edges @xmath71 starting at a vertex and the dual @xmath73 for the edges terminating there . the resulting representation at a vertex @xmath74 is denoted by @xmath75 .
finally , choose @xmath76 for each vertex @xmath74 . the element @xmath77 is evaluated on a connection @xmath78 by mapping @xmath78 to @xmath79 and then re - parsing to an element of @xmath80 . by (
* corollary 1 ) every observable is a linear combination of observables of this form .
now assume that @xmath55 is a trivalent lattice colored by irreducible representations .
the coloring is admissible if , at each vertex , the integers corresponding to the colorings of the incident edges form an admissible triple . [ basis ] suppose that @xmath55 is an admissibly colored trivalent lattice . for each @xmath75 there
exists a non - zero dual element invariant under the right action of @xmath49 .
the tensor product of these invariants over all vertices defines an observable .
the set of such observables , one for each admissible coloring , is a basis for @xmath81 .
let @xmath82 be an admissible coloring .
note that admissibility implies a @xmath83-dimensional invariant subspace in each @xmath84 .
hence there is a non - zero observable @xmath85 .
since @xmath86 is nonzero , there exists @xmath87 so that @xmath88 .
let @xmath89 be the `` parsing '' map .
let @xmath90 be the map @xmath91 where the inclusion is given by forming a sequence that has all zero matrices except for the @xmath40-st entry corresponding to @xmath39 .
define @xmath92 to be the connection @xmath93 .
clearly @xmath94 . since @xmath95
unless @xmath96 , @xmath92 is annihilated by all observables constructed from colorings different than @xmath97 . from this
we conclude that any set of observables constructed from distinct colorings is independent .
by @xcite they span .
there is a map @xmath98 that assigns to each framed link a wilson operator .
classically , a wilson operator is the trace of the holonomy of a connection along some fixed loop .
the construction in the quantum setting is described in @xcite . in a theory based on the drinfeld - jimbo @xmath2 and its hopf algebra
dual , the observables are isomorphic to the kauffman bracket skein algebra of the lattice envelope . since @xmath2 is a complete topological algebra over @xmath99 $ ] , it was necessary to complete the skein algebra as well .
the isomorphism was inferred from the classical isomorphism , the agreement of poisson brackets , and the @xmath100-adic completion . in section
[ observable ] we will show directly that the map @xmath101 is an isomorphism .
in this section we recall the correspondence between skeins and intertwiners in the category of representations @xmath39 of @xmath49 .
consider a rectangle @xmath102 with @xmath103 distinguished points ; @xmath104 of them on @xmath105 and @xmath104 on @xmath106 .
take the space of blackboard framed tangles with @xmath104 arc components ending at the distinguished points .
its quotient by the kauffman bracket skein relations is denoted by @xmath107 .
this quotient has an algebra structure given by placing the bottom of a rectangle on the top of another in such a way that the distinguished points meet .
[ rect ] the algebra @xmath107 is isomorphic to @xmath108 , the space of @xmath49-linear maps of @xmath109 to itself . in the basis @xmath110 of @xmath111 ,
the isomorphism is given by a tangle functor which makes the following assignments .
a local maximum is sent to the morphism @xmath112 defined by @xmath113 @xmath114 a local minimum is associated to the morphism @xmath115 given by @xmath116 [ joneswenzl ] the isomorphism takes the @xmath104-th jones - wenzl idempotent to the intertwiner that projects @xmath109 onto its highest weight invariant subspace .
let @xmath117 be the closed upper half plane .
for any @xmath104 , choose @xmath103 distinguished points on the @xmath118-axis , @xmath119 , and form a space of blackboard - framed tangles with @xmath104 arc components ending at the distinguished points .
the quotient of this space by the kauffman bracket skein relations is denoted @xmath120 .
@xmath121 the isomorphism is given by the same tangle functor as for fact [ rect ] .
an admissible triple @xmath122 determines a skein in @xmath123 consisting of a kauffman triad with all three legs attached to the @xmath118-axis .
fact [ joneswenzl ] gives a canonical inclusion of @xmath124 into @xmath125 .
[ triad ] an admissible triple @xmath122 equivalently , a kauffman triad corresponds to a nonzero vector in the @xmath83-dimensional space @xmath126 .
our goal is to prove a theorem analogous to ( * ? ? ?
* theorem 10 ) , but replacing power series by complex numbers . [ iso ]
let @xmath55 be a lattice and let @xmath6 be its envelope .
assume that @xmath127 is not a root of unity .
the algebra of observables of lattice gauge field theory on @xmath55 based on @xmath128 is isomorphic to @xmath129 .
the following description of @xmath101 is implicit in @xcite .
since @xmath59 is a homeomorphism invariant of @xmath6 , we can assume that the lattice @xmath55 comes from giving orientation and ciliation to a trivalent spine @xmath131 of @xmath6 .
let @xmath132 be an element of the basis @xmath32 of @xmath22 ( i.e. , a link with a simple diagram ) .
choose an orientation of @xmath132 . to compute the image of @xmath132 under @xmath101 ,
first perform the composition : @xmath133 here @xmath134 is the number of strands of @xmath132 running along the edge @xmath71 of @xmath55 , and @xmath135 means comultiplying @xmath136 times in the factor corresponding to @xmath71 .
third , treat each ciliated vertex as a half plane with the cilium at infinity .
up to isotopy , the link @xmath132 now appears as a collection of oriented caps in each half plane .
the cap pictured on the left of figure [ caps ] is associated with the map @xmath140 where @xmath141 , and @xmath142 is given by @xmath143 for the cap on the right .
notice that @xmath147 sends the switch map of @xcite to the map @xmath148 and sends multiplications to contractions . hence our description of @xmath144 for @xmath149 coincides with the wilson operator .
it follows from @xcite that @xmath101 does not depend on the choice of orientation of @xmath132 . extend it linearly to all of @xmath22 .
it is a homomorphism of algebras @xmath22 and @xmath59 . in order to prove that @xmath101 is an isomorphism we factor it into two maps which are isomorphisms on the level of vector spaces .
the first map , expressed in the basis @xmath32 , is given by a diagonal matrix with @xmath83 s and @xmath150 s on the diagonal .
the second map , @xmath151 does not require a choice of orientation of a link @xmath132 . to compute the image of @xmath149 under @xmath152 first
perform the composition ( [ phi ] ) .
second , apply the map @xmath153 to each copy of @xmath154 .
third , treat each vertex as a half plane and associate the map @xmath155 from equation ( [ max ] ) to the ( unoriented ) caps .
by fact [ triad ] , the map @xmath152 takes an element of the basis @xmath30 ( i.e. , a skein obtained by an admissible coloring of @xmath131 ) to an observable coming from coloring the edges of @xmath55 with corresponding irreducible representations of @xmath49 .
thus , by proposition [ basis ] , the map @xmath152 takes the basis @xmath30 of @xmath22 to a basis of @xmath59 . | we prove that the kauffman bracket skein algebra of a cylinder over a surface with boundary , defined over complex numbers , is isomorphic to the observables of an appropriate lattice gauge field theory . |
the phenomena of synchronization of several dynamical variables or oscillators is important for many physical and biological systems @xcite .
some important physical systems showing synchronization are arrays of lasers @xcite , microwave oscillators and superconducting josephson junctions @xcite while some important biological systems are synchronous firing of neurons @xcite , networks of pacemaker cells in the heart @xcite , metabolic synchrony in yeast cell suspensions @xcite , congregations of synchronously flashing fireflies @xcite , and cricket that chirp in unison @xcite .
coupled oscillators were first studied by winfree @xcite and kuramoto @xcite .
the kuramoto model describes a large population of coupled limit cycle oscillators with random frequencies .
if the coupling strength exceeds a critical threshold , the system exhibits a phase transition to a synchronous state where several oscillators synchronize and lock to a common frequency @xcite .
two important developments have regenerated interest in the study of synchronization of dynamical variables .
first is the recognition that chaotic dynamical systems can show exact or phase synchronization @xcite .
second is the observation that several natural systems have an underlying geometric structure which can be described by complex networks @xcite .
this has opened up the possibility of discovering new interesting phenomena in coupled dynamical systems on complex networks .
we discuss some of these aspects in this article .
several networks in the real world consist of dynamical elements interacting with each other and the interactions define the links of the network .
several of these networks have a large number of degrees of freedom and it is important to understand their dynamical behaviour .
a general model of coupled dynamical systems on networks will consist of the following three elements . 1 .
the evolution of uncoupled elements . 2 .
the nature of couplings .
the topology of the network .
most of the earlier studies of synchronized cluster formation in coupled chaotic systems have focused on networks with large number of connections ( @xmath0 ) @xcite or nearest neighbour couplings on lattice models .
recently , we have considered complex networks with number of connections of the order of @xmath1 @xcite .
this small number of connections allows us to study the role that different connections play in synchronizing different nodes and the mechanism of synchronized cluster formation .
the study reveals two interesting phenomena .
first , when nodes form synchronized clusters , there can be some nodes which show an intermittent behaviour between independent evolution and evolution synchronized with some cluster .
secondly , the cluster formation can be in two different ways , driven and self - organized phase synchronization @xcite .
the connections or couplings in the self - organized phase synchronized clusters are mostly of the intra - cluster type while those in the driven phase synchronized clusters are mostly of the inter - cluster type .
we will briefly review these features in this article .
synchronization of different dynamical variables can be defined in several ways .
exact synchronization corresponds to the situation where the dynamical variables have identical values , i.e. two dynamical variables @xmath2 and @xmath3 are exactly synchronized if @xmath4 @xcite .
generalised synchronization is defined by some functional relation between the dynamical variables @xcite .
several other types of synchronization such as phase synchronization @xcite , lag synchronization @xcite , anticipatory synchronization @xcite etc . have been defined .
kuramoto model of coupled oscillators can be introduced through the evolution equation @xcite @xmath5 where @xmath6 is the phase of oscillator @xmath7 , @xmath8 is its natural frequency and @xmath9 is the coupling strength .
phase synchronization of the oscillators can be studied using the complex order parameter defined as @xmath10 where @xmath11 measures the phase coherence , and @xmath12 is the average phase .
clearly , @xmath11 vanishes for a uniform distribution of phases @xmath6 , and @xmath13 if all @xmath6 are equal . the evolution equation for the phases
can be written as , @xmath14 this is a mean field form of the equation . for @xmath9 less than some threshold @xmath15 ,
the oscillators show unsynchronized behaviour .
but , when @xmath16 , the oscillator population splits into two groups , the oscillators near the center of the frequency distribution lock together at some frequency @xmath17 to form a synchronized cluster , while those in the tail retain their natural frequencies and drift relative to the synchronized cluster .
this state is often called partially synchronized state . with further increase in @xmath9 ,
more and more oscillators are recruited into the synchronized clusters @xcite .
kuramoto s model was originally motivated by biological oscillators @xcite , but it has found applications in many diverse systems such as flavour evolution of neutrinos @xcite , arrays of josephson junction @xcite , semiconductor lasers @xcite and in several other systems @xcite .
pecora and carroll showed that two identical chaotic systems can synchronize if approrpiate driving mechanisms are introduced @xcite .
let @xmath18 be an @xmath19 dimensional dynamical variable evolving as ( drive system ) @xmath20 divide the dynamical variables into two parts , @xmath21 , a drive part and a response part .
consider another dynamical system ( response system ) given by @xmath22 where the drive variables @xmath23 are obtained from eq .
( [ dyn - nd ] ) . under suitable conditions ,
i.e. the conditional lyapunov exponents are negative , the response variable @xmath24 synchronize with those of the drive system , @xmath25 .
the conditional lyapunov exponents are obtained by considering the subspace of response variables .
the important interesting feature is that the variables synchronize even when they are evolving chaotically .
thus we do not have frequency locking , but we can have phase synchronization if a suitable phase variable can be defined @xcite .
such phase synchronization is observed in several biological systems @xcite several complex systems have underlying structures that are described by networks or graphs @xcite .
recent interest in networks is due to the discovery that several naturally occurring networks come under some universal classes and they can be simulated with simple mathematical models , viz small - world networks @xcite , scale - free networks @xcite etc .
consider a network of @xmath1 nodes and @xmath26 connections ( or couplings ) between the nodes .
let each node of the network be assigned an @xmath27-dimensional dynamical variable @xmath28 .
a very general dynamical evolution can be written as @xmath29 here , we consider a separable case and the evolution equation can be written as , @xmath30 where @xmath31 is the coupling constant , @xmath32 is the degree of node @xmath7 , and @xmath33 is the set of nodes connected to the node @xmath7 .
a sort of diffusion version of the evolution equation ( [ evol - cont ] ) is @xmath34 discrete versions of eqs .
( [ evol - cont ] ) and ( [ evol - cont - diff ] ) are @xmath35 and @xmath36 in numerical studies , for the discrete evolution we can use any descrete map such as logistic or circle maps while for the continuous case we can use chaotic systems such as lorenz or rssler systems .
as noted before synchronization of coupled dynamical systems @xcite is manifested by the appearance of some relation between the functionals of different dynamical variables . when the number of connections in the network is small ( @xmath37 ) and when the local dynamics of the nodes ( i.e. function @xmath38 ) is in the chaotic zone , and we look at exact synchronization , only few synchronized clusters with small number of nodes are formed . however , when we look at phase synchronization , synchronized clusters with larger number of nodes are obtained .
we consider some general properties of synchronized dynamics
. they are valid for any coupled discrete and continuous dynamical systems .
also , these properties are applicable for exact as well as phase or any other type of synchronization and are independent of the type of network .
as the network evolves , it splits into several synchronized clusters . depending on their asymptotic dynamical behaviour
the nodes of the network can be divided into three types .
+ ( a ) _ cluster nodes _ : a node of this type synchronizes with other nodes and forms a synchronized cluster .
once this node enters a synchronized cluster it remains in that cluster afterwards .
+ ( b ) _ isolated nodes _ : a node of this type does not synchronize with any other node and remains isolated for all the time .
+ ( c ) _ floating nodes _ : a node of this type keeps on switching intermittently between an independent evolution and a synchronized evolution attached to some cluster . of particular interest are the floating nodes and we will discuss some of their properties afterwards .
the study of the relation between the synchronized clusters and the couplings between the nodes represented by the adjacency matrix @xmath39 shows two different mechanisms of cluster formation @xcite .
+ ( i ) self - organized clusters : the nodes of a cluster can be synchronized because of intra - cluster couplings .
we refer to this as the self - organized synchronization and the corresponding synchronized clusters as self - organized clusters .
+ ( ii ) driven clusters : the nodes of a cluster can be synchronized because of inter - cluster couplings . here
the nodes of one cluster are driven by those of the others .
we refer to this as the driven synchronization and the corresponding clusters as driven clusters . in numerical studies it is possible to observe ideal clusters of both the types , as well as clusters of the mixed type where both ways of synchronization contribute to cluster formation .
[ scale - x - clus ] shows some examples of ideal as well as mixed clusters in coupled map networks @xcite .
in general we find that the scale free networks and the caley tree networks lead to better cluster formation than the other types of networks with the same average connectivity . , the local dynamics is given by the logistic @xmath40 and the coupling function is linear .
the solid circles show that the two corresponding nodes are coupled ( i.e. @xmath41 ) and the open circles show that the corresponding nodes are phase synchronized . in each case
the node numbers are reorganized so that nodes belonging to the same cluster are numbered consecutively and the clusters get displayed in decreasing sizes .
( a ) figure shows turbulent phase for coupling constant @xmath42 .
( b ) an ideal self - organized phase synchronization for @xmath43 .
( c ) mixed behavior for @xmath44 .
( d ) an ideal driven phase synchronization for @xmath45.,width=529 ] geometrically the two mechanisms of cluster formation can be easily understood by considering a tree type network .
a tree can be broken into different clusters in different ways .
+ ( a ) a tree can be broken into two or more disjoint clusters with only intra - cluster couplings by breaking one or more connections .
clearly , this splitting is not unique and will lead to self - organized clusters .
figure [ tree - clus](a ) shows a tree forming two synchronized clusters of self - organized type .
this situation is similar to an ising ferromagnet where domains of up and down spins can be formed .
+ ( b ) a tree can also be divided into two clusters by putting connected nodes into different clusters .
this division is unique and leads to two clusters with only inter - cluster couplings , i.e. driven clusters .
figure [ tree - clus](b ) shows a tree forming two synchronized clusters of the driven type .
this situation is similar to an ising anti - ferromagnet where two sub - lattices of up and down spins are formed .
+ ( c ) several other ways of splitting a tree are possible .
e.g. it is easy to see that a tree can be broken into three clusters of the driven type .
this is shown in figure [ tree - clus](c ) .
there is no simple magnetic analog for this type of cluster formation .
it can be observed close to a period three orbit .
we note that four or more clusters of the driven type are also possible . as compared to the cases ( a ) and ( b ) discussed above which are commonly observed , the clusters of case ( c ) are not so common and
are observed only for some values of the parameters .
a suitable network to study the stability of self - organized synchronized clusters is the globally coupled network .
the stability of globally coupled maps is well studied in the literature @xcite . an ideal example to consider the stability of the driven synchronized state is a complete bipartite network .
a complete bipartite network consists of two sets of nodes with each node of one set connected with all the nodes of the other set and no connection between the nodes of the same set .
let @xmath46 and @xmath47 be the number of nodes belonging to the two sets .
we define a bipartite synchronized state as the one that has all @xmath46 elements of the first set synchronized to some value , say @xmath48 , and all @xmath47 elements of the second set synchronized to some other value , say @xmath49 .
all the eigenvectors and the eigenvalues of the jacobian matrix for the bipartite synchronized state can be determined explicitly .
the eigenvectors of the type @xmath50 determine the synchronization manifold and this manifold has dimension two .
all other eigenvectors correspond to the transverse manifold .
lyapunov exponents corresponding to the transverse eigenvectors for eq .
( [ evol - disc - diff ] ) with one dimensional variables and @xmath51 are @xmath52 and @xmath53 and @xmath54 are respectively @xmath55 and @xmath56 fold degenerate @xcite . here ,
@xmath57 and @xmath58 are the derivatives of @xmath38 at @xmath59 and @xmath60 respectively .
the synchronized state is stable provided the transverse lyapunov exponents are negative . if @xmath61 is bounded then from eqs .
( [ lya - driven - nlarge ] ) we see that for @xmath31 larger than some critical value , @xmath62 , bipartite synchronized state will be stable .
note that this bipartite synchronized state will be stable even if one or both the remaining lyapunov exponents corresponding to the synchronization manifold are positive , i.e. the trajectories are chaotic .
the linear stability analysis for other type of couplings and dynamical systems can be done along similar lines .
we had noted earlier that the nodes can be divided into three types , namely cluster nodes , isolated nodes and floating nodes , depending on the asymptotic behavior of the nodes .
here , we discuss some properties of the floating nodes which show an intermittent behavior between synchronized evolution with some cluster and an independent evolution .
let @xmath63 denote the residence time of a floating node in a cluster ( i.e. the continuous time interval that the node is in a cluster ) .
figure [ freq - floating ] plots the frequency of residence time @xmath64 of a floating node as a function of the residence time @xmath63 . a good straight line fit on log - linear plot shows an exponential dependence , @xmath65 where @xmath66 is the typical residence time for a given node .
we have also studied the distribution of the time intervals for which a floating node is not synchronized with a given cluster .
this also shows an exponential distribution . of a floating node in a cluster as a function of the residence time @xmath63 . a good straight line fit on log - linear plot shows exponential dependence.,width=377 ] several natural systems show examples of floating nodes , e.g. some birds may show intermittent behaviour between free flying and flying in a flock .
an interesting example in physics is that of particles or molecules in a liquid in equilibrium with its vapor where the particles intermittently belong to the liquid and vapor . under suitable conditions it is possible to argue that the residence time of a tagged particle in the liquid phase should have an exponential distribution @xcite , i.e. a behavior similar to that of the floating nodes ( eq .
( [ exp - dist - floating ] ) ) .
we have briefly discussed synchronization properties of coupled dynamical elements on complex networks . in the course of time evolution these dynamical elements form synchronized clusters . in several cases
when synchronized clusters are formed there are some isolated nodes which do not belong to any cluster .
more interestingly there are some floating nodes which show an intermittent behavior between an independent evolution and an evolution synchronized with some cluster .
the residence time spent by a floating node in the synchronized cluster shows an exponential distribution .
we have identified two mechanisms of cluster formation , self - organized and driven phase synchronization . for self - organized clusters
intra - cluster couplings dominate while for driven clusters inter - cluster couplings dominate .
a. pikovsky , m. rosenblum and j. kurth , _ synchronization : a universal concept in nonlinear dynamics _
( cambridge university press , 2001 ) .
z. jiang , m. mccall , _ j. opt .
* 10 * , 155 ( 1993 ) .
p. a. robinson , j. j. wright and c. j. rennie , _ phys .
lett . _ * 57 * , 4578 , ( 1998 ) . c. s. peskin , _ mathematical aspects of heart physiology _ ( courant institute of mathematical science publication , new york , 1975 , pp .
j. aldridge , e. k. pye , _ nature _ * 259 * , 670 ( 1976 ) .
j. buck , e. buck , _ scientific am . _ * 234 * , 74 ( 1996 ) .
t. j. walker , _ science _ * 166 * , 891 ( 1969 ) .
a. t. winfree , _ the geometry of biological time _ ( springer - verlag , new york , 1980 ) .
y. kuramoto , _ chemical oscillations , waves , and turbulence _ ( springer- verlag , berlin , 1984 ) .
a. t. winfree , _ j. theoret
. biol . _ * 16 * 15 ( 1967 ) .
l. m. pecora and t. l. carroll , _ phys .
lett . _ * 64 * , 821 ( 1990 ) .
s. h. strogatz , _ nature _ * 410 * , 268 ( 2001 ) and references therein .
r. albert and a. l. barabsi , _ rev .
* 74 * , 47 ( 2002 ) and references therein .
k. kaneko , _ physica _ * d124 * ( 1998 ) 322 .
s. jalan , r. e. amritkar , _ phys .
* 90 * ( 2003 ) 014101 .
h. fujisaka and t. yamada , _ prog .
phys . _ * 69 * , 32 ( 1983 ) .
h. f. rulkov , m. m. sushchik , l. s. tsimring and h. d. i. abarbanel , phys .
e * 51 * , 980 ( 1995 ) .
m. g. rosenblum , a. s. pikovsky , and j. kurth , _ phys .
lett . _ * 76 * , 1804 ( 1996 ) .
m. g. rosenblum , a. s. pikovsky , and j. kurth , _ phys .
lett . _ * 78 * , 4193 ( 1997 ) .
h. u. voss , phys .
e * 61 * , 5115 ( 2000 ) . j. d. crawford , _ j. statist . phys .
_ , * 74 * , 1047 ( 1994 ) ; j. d. crawford and k. t. r. davies , _ physica d _ * 125 * 1 ( 1999 ) . s. h. strogatz , _ physica d _ , * 143 * 1 ( 2000 )
. j. pantaleone , _ phys .
d _ * 58 * , 3002 ( 1998 ) .
k. wiesenfeld , p. colet , s. h. strogatz , _ phys .
* 57 * , 1563 ( 1998 ) .
g. kozyreff , a. g. vladimirov and p. mandal , _ phys .
lett . _ * 85 * , 3809 ( 2000 ) .
y. maistrenko , o. popovych , o. burylko and p. a. tass , _ phys .
lett . _ * 93 * , 841021 ( 2004 ) m. g. earl and s. h. strogatz , _ phys .
e _ * 67 * , 036204 ( 2003 ) .
d. j. watts and s. h. strogatz , collective dynamics of small world networks, _ nature ( london ) _ * 393 * , 440 ( 1998 ) .
barabsi , r. albert , ` emergence of scaling in random networks , ' _ science _ * 286 * , 509 ( 1999 ) . | we review some recent work on the synchronization of coupled dynamical systems on a variety of networks . when nodes show synchronized behaviour , two interesting phenomena can be observed .
first , there are some nodes of the floating type that show intermittent behaviour between getting attached to some clusters and evolving independently .
secondly , two different ways of cluster formation can be identified , namely self - organized clusters which have mostly intra - cluster couplings and driven clusters which have mostly inter - cluster couplings . |
experimental realization of trapped bose - einstein condensates ( bec ) in alkali metal atoms has triggered immense interest in understanding the various properties of ultra cold matter @xcite .
these include four wave mixing , formation of vortices , bright solitons , dark solitons , gap solitons , interference patterns and domain walls in binary becs @xcite .
an interesting dynamical feature in the context of bose - einstein condensation is the formation of matter wave solitons @xcite .
for example , dark solitons @xcite and gap solitons @xcite were experimentally observed in becs with repulsive interactions , whereas bright solitons were demonstrated in systems with attractive interactions @xcite .
the recent experiments at heidelberg and hamburg universities have shown the formation of dark solitons , their oscillations and interaction in single component becs of @xmath0 rb atoms with confining harmonic potential @xcite .
these experimental results have been augmented by extensive theoretical work including predictions on bright and dark matter wave solitons @xcite , lattice solitons @xcite , and soliton trains @xcite .
the dynamics of a bec at absolute zero temperature is usually described by the mean - field gross - pitaevskii ( gp ) equation @xcite , which is a generalized form of the ubiquitous nls equation@xcite , for the wave function of the condensate .
the nls equation and its generalizations are widely used in the nonlinear optics literature @xcite to describe the evolution of coherent light in a nonlinear kerr medium , envelope dynamics of quasi - monochromatic plane wave propagation in a weakly dispersive medium , high intensity pulse propagation in optical fibers and so on . as is well known the nls equation in one space and one time dimension
is a completely integrable soliton system @xcite : it possesses bright solitons in the focusing case , while dark solitons exist in the defocusing case . on the other hand ,
the gp equation which is a three space and one time dimensional nonlinear evolution equation , is not integrable in general , but can admit exact solutions only in very special cases .
in general , the gp equation has to be analyzed numerically or by approximation methods . for the cigar - shaped trap potential
when the transverse linear oscillator length is much smaller than the longitudinal length , the 3d gp equation can be reduced to the effective 1d gp equation by assuming the kinetic energy of the longitudinal excitations and the two - body interaction energy of the atoms are of the same order , while both are much less than the kinetic energy of the transverse excitations @xcite . to obtain bright and dark soliton solutions ,
its quasi-1d version has to be analyzed essentially numerically .
interestingly , some of the soliton solutions in the presence of regular harmonic confinement , with appropriately tailored gain or loss , have been shown to exhibit collapse and revival phenomena , with an increase in amplitude .
for example , hulet and his co - workers observed the growth of a condensate of trapped @xmath1li atoms with attractive interaction and subsequent collapse @xcite , and miesner et al .
measured the rate of growth of a @xmath2na condensate fed by a thermal cloud @xcite .
such a growth of a bec from thermal vapor was also experimentally realized for atoms of @xmath0rb @xcite .
several methods have been introduced theoretically to account for the growth of bec @xcite .
the gp equation with the complex linear gain or loss term , often referred to as gp gain equation , was used to model this type of bec dynamics by drummond and kheruntsyan @xcite .
it has been shown that , as the condensate grows , the center of mass oscillates in the trap .
the gp gain equation is of special importance in the field of atom laser , which is a device that produces an intense coherent beam of atoms by stimulated process @xcite . in recent years
, several attempts have been made to understand the dynamics of bec by studying the gp gain equation @xcite .
further , there has been an increased interest during recent times in studying the properties of becs with time varying control parameters . in particular ,
investigations have been made over a decade or so to understand the properties of condensates under 1 .
the temporal variation of atomic scattering length which can be achieved through feshbach resonance @xcite , 2 .
inclusion of appropriate time - dependent gain or loss terms which can be phenomenologically incorporated to account for the interaction of atomic cloud or thermal cloud @xcite , 3 .
periodic modulation of trap frequencies@xcite and so on .
earlier studies on becs have shown that a suitable periodic modulation of scattering length can stabilize the condensate against collapse @xcite . however , there are certain occasions where a sign alternating nonlinearity and increase in the frequency of oscillations result in the acceleration of collapse @xcite . in this connection ,
a systematic study on the condensates under the influence of time - dependent parameters which can be achievable through experiments is of considerable importance .
most of the theoretical studies on the matter wave solitons in becs have been carried out using either numerical or certain approximation ( perturbation ) methods @xcite .
for example , ruprecht et al @xcite have shown the existence of bright soliton solutions in weakly interacting 1d and 2d becs by solving the corresponding gp equations numerically . in an
another context , prez - garca et al @xcite have shown the existence of bright soliton solutions to the gp equation with negative scattering length in highly asymmetric traps by using the multiple - scale approximation method , while busch and anglin @xcite have shown the motion of dark solitons in trapped effective 1d bec by using a multiple time scale boundary layer theory . on the other hand
, there are only a few exact analytical methods available in the literature @xcite to deal with the gp equation . in this connection
it is worth mentioning that serkin et al @xcite have constructed exact soliton solutions by relating the effective 1d gp equation to the compatibility condition of a lax pair which is in turn related to that of the lax pair of the 1d integrable nls equation using an inverse scattering transform ( ist ) method while atre et al @xcite transformed the gp equation to the unforced duffing equation and thereby derived soliton solutions from the elliptic function solutions of the duffing equation .
also , perez - garcia et al .
@xcite obtain a similarity transformation connecting the nls equation with time varying coefficients and the autonomous cubic nls equation .
in addition , there exists several other interesting works in the literature to construct exact soliton solutions of the 1d gp equation : specifically , for certain time - dependent atomic scattering lengths ( nonlinearity ) with ( i ) time independent homogeneous , linear and quadratic potentials without gain / loss term @xcite , ( ii ) time independent expulsive parabolic potential without gain / loss term @xcite , ( iii ) time independent expulsive parabolic potential and arbitrary constant gain / loss term @xcite , ( iv ) tanh type modulated harmonic trap frequency without gain / loss term @xcite , explicit growing / decaying type solutions have been constructed . in the optical soliton context , in ref .
@xcite exponentially growing solutions have been constructed for either constant atomic scattering length or zero potential case with a specific tanh type dissipation term . however , a more detailed analysis for bright and dark soliton solutions in quasi-1d bec with time - dependent atomic scattering length , harmonic trap frequency and gain / loss term is of considerable present interest from an experimental point of view . in this paper
we bring out a large class of bright and dark soliton solutions by simply mapping the effective 1d gp equation onto the standard nls equation , following the earlier works in the literature @xcite .
in particular we show that the effective 1d equation resulting from the gross - pitaevskii ( gp ) equation in confining as well as expulsive potentials with time varying parameters , namely the scattering length , gain / loss term and trap frequency , can be transformed into the standard soliton ( bright / dark ) possessing , completely integrable 1d nls equation by effecting a change of variables of the coordinates and the wave function under certain condition , which is related to the scattering length , gain / loss term and the trap frequency .
we consider both confining and expulsive harmonic trap potentials separately and treat the atomic scattering length , gain / loss term and trap frequency as experimental control parameters by modulating them as a function of time .
we have deduced different kinds of bright / dark soliton solutions for different forms of the trap frequency by suitably tailoring the atomic scattering length and atom gain / loss term such as 1 .
periodic oscillating solitons , collapse and revival of condensate , and snake - like solitons for time independent confining harmonic potential , 2 .
stable solitons , soliton growth and decay , and formation of two - soliton bound state for time independent expulsive harmonic potential , 3 .
oscillatory solitons , phenomena of collapse and revival of two - soliton like bound state formation of the condensate for double modulated periodic potential and 4 .
stable solitons , soliton growth and decay for step - wise modulated potential .
we also make comparisons of our results with the experiments of khaykovich et al @xciteand strecker et al @xcite obtained for @xmath3li atoms and for @xmath0rb atoms @xcite .
further , we also compare our results with the theoretical results of atre et al @xcite and serkin et al @xcite .
it may be further noted that in experiments the matter wave solitons @xcite have been observed within few micrometer ( @xmath4 m ) lengths and times up to the order of few ms . in the present paper
we choose the soliton parameters such that the length ( @xmath4 m ) and time ( ms ) scales are well within the experimental regime .
further , we have confined ourselves to the issue of stabilizing solitons by choosing the forms of modulated atomic scattering length and atom gain / loss term so as to balance each other .
we have also considered here only the two - body interactions and neglected the three - body recombination loss which will introduce a quintic term in the gp equation which does not in general reduce to exactly integrable standard nls equation .
such a case may be studied by the numerical and approximation methods .
another interesting aspect of the temporal modulations is to look into the problem of parametric amplification , which however falls outside the scope of the present analysis and should be dealt with separately .
it may also be noted that in the context of optical systems with varying coefficients , only picosecond optical solitons have been observed @xcite . this paper is organized as follows . in sec .
[ sec : gpe ] , we discuss the method of transforming cylindrically symmetric gp equation with time varying parameters such as the atomic scattering length , gain / loss and with both confining and expulsive harmonic potentials into an effective 1d gp equation with time varying parameters .
then , we map the 1d gp equation to the standard nls equation by making use of suitable transformations , subject to a constraint on the forms of scattering length , gain / loss term and the trap frequency . next , in sec . [ sec : solitons ] we deduce the generalized bright and dark matter wave soliton solutions of the gp equation from the soliton solutions of standard nls equation .
we show the existence of various forms of bright and dark soliton solutions , their collapse and revival , and decay and growth for suitable choices of the condensate parameters with time independent confining and expulsive harmonic potentials in secs .
[ ind : con ] and [ ind : exp ] , respectively .
this is followed by a demonstration of various forms of bright and dark matter wave soliton solutions for temporally modulated potentials in sec . [ de ] .
finally , in sec .
[ con ] , we present a summary of the results and conclusions . in appendix
[ app : validity ] , we have shown the validity criteria of the 1d gp equation for various forms of control parameters studied in this paper . in appendix
[ app : solve ] , we give a brief description on the solution procedure of solving the restrictive condition for which the effective 1d gp equation can be mapped on to the standard nls equation .
we also tabulate the possible solvable choices of the modulated trap frequency forms from the ricatti equation which specifies the restrictive condition in appendix [ app : ricatti ] and provide a table of physical meaning of symbols used in the text in appendix [ app : symbol ] .
in the present study , we consider the gp equation with the inclusion of a complex gain or loss term of the form @xmath5 \psi({\mathbf r},t ) , \label{3dgpe}\end{aligned}\ ] ] where @xmath6 , @xmath7 is the time varying s - wave scattering length , @xmath8 , @xmath9 is a cylindrical time independent harmonic trap , @xmath10 is a time - dependent harmonic trap along the @xmath11 direction which can be either confining or expulsive , @xmath12 depending on whether the interaction is attractive @xmath13 or repulsive @xmath14 , and @xmath15 is the gain / loss term which is phenomenologically incorporated to account for the interaction of atomic cloud or thermal cloud .
the plus sign of the @xmath15 term leads to the mechanism of loading external atoms ( thermal clouds ) into the bec by optical pumping while the minus sign of @xmath15 describes a bec that is continuously depleted ( loss ) of atoms .
if the condensate is fed by a surrounding thermal cloud , then the condensate undergoes a periodic growth and collapse .
the above equation has gained much importance in recent times to understand the growth and decay of the condensate due to time - dependent parameters @xcite .
however , one has to note that the parameters are varied sufficiently slowly so as to satisfy the adiabaticity criteria so that the wave - packet continues to evolve in the same quantum eigenstate as a function of time .
a possible condition for the validity of the adiabaticity criteria @xcite is that the characteristic time scale , say @xmath16 , of the variation of the modulated parameters is slow enough when compared with the instantaneous nonlinear eigenvalue @xmath17 of the nonlinear equation at time @xmath18 , such that @xmath19 .
we also note that for the gp equation ( [ 3dgpe ] ) the normalization condition in the presence of the gain / loss term @xmath15 becomes @xmath20,\end{aligned}\ ] ] where @xmath21 is number of the condensate atoms .
there are typical characteristic length scales to specify the condensate density .
the confining frequencies of the harmonic trap potential ( @xmath22 ) , that is @xmath23 and @xmath24 , set the length scale for spacial size which are the transversive oscillator length @xmath25 and the longitudinal length @xmath26 .
in addition , the effective mean - field nonlinearity introduces the length scale , namely the healing length @xmath27 , which is the distance over which the kinetic energy and the interaction energy balance @xcite . here
@xmath28 is the mean density of the condensate .
the presence of the nonlinear term in the above gp equation ( [ 3dgpe ] ) complicates its solution procedure .
one should note that it is very difficult to obtain closed form solution of eq .
( [ 3dgpe ] ) as such . in most cases
the above equation can be studied either by approximation methods or by suitable numerical methods . however , under certain circumstances , eq .
( [ 3dgpe ] ) can be reduced to a simpler form by appropriate transformation whose solutions are known .
one way here is to reduce or relate the above eq .
( [ 3dgpe ] ) to the standard nls equation in ( 1 + 1 ) dimensions so that the soliton solutions can be readily obtained . in the following
we shall describe a method of reducing eq .
( [ 3dgpe ] ) to the nls equation and obtaining the consequent spatiotemporal patterns .
first let us reduce the gp equation ( [ 3dgpe ] ) to an effective 1d equation which deals with the cigar - shaped trap potential when the transverse linear oscillator length @xmath29 satisfies the following criteria : @xmath29 is much smaller then the longitudinal length @xmath30 @xcite , that is @xmath31 the above criteria essentially means that only for an axially open trap ( @xmath32 ) one obtains a 1d gp equation . in this case
we assume that the condensate wave function can be taken as @xmath33 here the width and length of the condensate are of the order of @xmath29 and @xmath34 , respectively , and hence @xmath35 is the mean density of the condensate .
note that the kinetic energy of the longitudinal excitations [ @xmath36 and the the two - body interaction energy of atoms [ @xmath37 are of the same order , while both are much less than the kinetic energy of the transverse excitations ( @xmath38 ) @xcite . from this
, we get following criteria for the validity of the reduction of the 3d gp equation to the effective 1d gp equation in the _ absence _ of the gain / loss term @xmath15 @xcite , @xmath39 however , in the presence of the gain term in eq .
( [ 3dgpe ] ) , and in view of the normalization condition ( [ norm ] ) , the above criterion may be modified by replacing @xmath21 with @xmath40 $ ] as @xmath41}{a_z(t ) } \ll 1 .
\label{cre1 } \end{aligned}\ ] ] note that due to the above condition , for a given @xmath21 , the functional form of the scattering length @xmath7 , gain / loss term @xmath15 and the longitudinal(axial ) length @xmath30 are to be chosen appropriately as a function of time so that the above condition can be valid for sufficiently long times , which we will demonstrate in appendix [ app : validity ] for various forms @xmath7 , @xmath15 and @xmath30 studied in this paper . substituting the above form of the wave function ( [ trans1 ] ) in eq . ( [ 3dgpe ] ) ,
one obtains the effective 1d gp equation with modulational gain / loss term , atomic scattering length and trap frequency as @xmath42 where @xmath43 , @xmath44 , @xmath25 , @xmath45 and @xmath46 is the bohr radius .
note that in eq ( [ 1dgpe ] ) , the variable @xmath11 actually represents @xmath47 , due to the transformation ( [ trans1 ] ) .
similarly @xmath18 stands for @xmath48 . using the relation ( [ trans1 ] )
, we can rewrite the above condition as @xmath49.\label{norm1}\end{aligned}\ ] ] in addition , one can also transform away the gain ( or loss ) term @xmath50 in eq .
( [ 1dgpe ] ) by applying the following transformation @xcite @xmath51 q(z , t ) , \label{phi}\end{aligned}\ ] ] so that eq .
( [ 1dgpe ] ) can be rewritten in terms of the new variable @xmath52 as @xmath53 where @xmath54 is a time - dependent parameter and can be written as a function of @xmath55 and @xmath50 in the following way @xmath56 r(t ) \label{rtilde}.\end{aligned}\ ] ] it has been shown by grses @xcite recently that the above eq .
( [ nnlse ] ) can be mapped onto the standard nls equation under the following , transformation , @xmath57 where the new independent variables @xmath16 and @xmath58 are chosen as functions of the old independent variables @xmath18 and @xmath11 as @xmath59 while @xmath60 is a function of @xmath18 and @xmath11.it may also noted that similar mapping has been done by serkin et al .
@xcite and prez - garca et al .
@xcite using the lens transformation . applying the above transformation ( [ qtrs ] ) , so that @xmath61 one can reduce eq .
( [ nnlse ] ) to the standard nls equation of the form @xmath62 subject to the conditions that the functions @xmath63 , @xmath64 , @xmath65 , @xmath66 and @xmath67 should satisfy the following set of equations , @xmath68 [ eq : cond ] the above nls equation ( [ nlse ] ) is widely used in the nonlinear optics to describe the evolution of coherent light in dispersive medium , high intensity pulse propagation in optical fibers and so on , which is completely integrable equation and admits soliton , elliptic and trigonometric function solutions @xcite . in order to solve the unknown functions
@xmath63 , @xmath64 , @xmath65 in the above equations ( [ eq : cond ] ) we assume the polar form@xmath69,\label{lambda}\end{aligned}\ ] ] one can immediately check from the relations ( [ eq : cond : c ] ) that @xmath70 is a function of @xmath18 only , @xmath71 , since @xmath65 and @xmath67 are functions of @xmath18 only . then from eqs .
( [ eq : cond ] ) one can easily deduce the transformation function @xmath63 given by ( [ lambda ] ) and the transformations @xmath72 and @xmath73 through the following relations ( details are given in appendix [ app : solve ] ) : @xmath74 [ solv ] here @xmath75 , @xmath76 are arbitrary constants , and @xmath67 and @xmath77 should be related by the following condition ( see appendix [ app : solve ] , eq .
( [ sol_con : b3 ] ) for made details ) @xmath78which is a riccati type equation with dependent variable @xmath79 and independent variable @xmath18 .
note that here @xmath80= \frac{2 a_s(t)}{a_b } \exp \left [ \int \gamma(t ) dt \right]$ ] is a quantity representing the modulated scattering length and gain / loss term and @xmath81 represents the time - dependent trap frequency .
( [ con : eqn ] ) implies a specific constraint between these quantities and for a suitable choice of them satisfying the equation ( [ con : eqn ] ) , the effective 1d gp equation ( [ 1dgpe ] ) and so eq . ( [ nnlse ] ) can be mapped onto the standard nls equation ( [ nlse ] ) .
it may be noted that a transformation similar to the above transformation has also been reported by serkin et al .
@xcite by relating eq .
( [ nnlse ] ) to the compatibility condition of the related lax pair and prez - garca et al .
@xcite using the lens transformation .
however , we will make use of the relations ( [ qtrs])-([con : eqn ] ) as they are more straightforward .
also we note that the transformation given in eq .
( 18 ) is more general than the one given in refs .
@xcite because of the additional term proportional to the arbitrary constant @xmath75 in eqs .
( 18 b ) and ( 18 c ) .
one can consider three possible approaches to the solutions to the present riccati equation constraint ( [ con : eqn ] ) : 1 .
given @xmath82 , obtain possible explicit forms of @xmath54 by solving ( [ con : eqn ] ) , 2 .
given @xmath54 , deduce the form of the trap frequency @xmath83 as can be written down from ( [ con : eqn ] ) straightforwardly 3 .
given @xmath82 , obtain numerical solution for @xmath54 when the riccati equation can not be explicitly solved for @xmath54 firstly , there are numerous forms of @xmath82 for which explicit solutions for @xmath54 can be obtained [ see for example ref .
@xcite ] by solving the riccati equation ( [ con : eqn ] ) .
most important of them are given in the form of table [ table1 ] in appendix [ app : ricatti ] .
next , for a given form of @xmath54 , the corresponding form of the trap frequency @xmath83 can be straightaway calculated from ( [ con : eqn ] ) as @xmath84 finally , for those forms of @xmath82 for which no exact solutions can be found by solving the riccati equation ( [ con : eqn ] ) , numerical solutions can always be found for @xmath79 , from which @xmath67 can be calculated .
one can make use of all the above three approaches to obtain the matter wave solutions .
the nls equation ( [ nlse ] ) has been known for quite some time for its novel soliton solutions .
it admits jacobian elliptic function solutions , bright and dark solitary wave solutions and trigonometric function solutions @xcite , depending on the sign of @xmath85 .
for instance , soliton solutions of the standard nls equation ( [ nlse ] ) were given by zakharov and shabat by solving the cauchy initial value problem through the ist method for @xmath86 ( focusing or attractive case ) @xcite and for @xmath87 @xcite ( defocusing or repulsive case ) .
the corresponding soliton solutions are called bright ( @xmath86 ) and dark ( @xmath87 ) solitons , respectively . in the following ,
we will make use of the one soliton solutions only and higher order soliton solutions will be discussed separately .
we obtain the one soliton solutions of the effective 1d gp equation ( [ 1dgpe ] ) , by making use of the one soliton solutions of nls equation ( [ nlse ] ) through the relations ( [ phi ] ) , ( [ qtrs ] ) and ( [ solv ] ) .
the bright soliton solution of ( [ nlse ] ) for the case of @xmath86 takes the following form @xcite ( leaving aside the unimportant phase constants ) @xmath88 { \;{\rm sech}}\left[\frac{a}{\sqrt{2 } } ( z - c t)\right ] .
\label{bright}\end{aligned}\ ] ] here the real parameters @xmath89 and @xmath90 correspond to the amplitude and velocity , respectively , of the bright soliton envelope and plus sign in the superscript represents the @xmath86 case . correspondingly for the gp equation([1dgpe ] ) , using the relations ( [ phi ] ) , ( [ qtrs ] ) and ( [ bright ] ) , the bright soliton solution can be written as @xmath91 q^+(z , t ) = \exp\left[\int \frac{\gamma(t)}{2 } dt \right ] \lambda(z , t ) q^+(z , t ) .
\label{phi+}\end{aligned}\ ] ] now substituting the forms of @xmath92 , @xmath58 and @xmath16 from eqs .
( [ fg ] ) , ( [ lambda ] ) and ( [ solv ] ) in the above eq .
( [ phi+ ] ) , we get a generalized expression for the bright soliton as [ eq:1d : bright ] @xmath93 \exp\left[i\eta^+(z , t)\right ] , \label{bright } \end{aligned}\ ] ] here the time - dependent amplitude of the generalized soliton ( [ bright ] ) is given by @xmath94 which depends on the form of the atomic scattering length @xmath55 and gain / loss term @xmath50 , and the generalized wave variable @xmath95 , which specifies the width and velocity of the bright soliton as a function of time and longitunumberdinal distance , is given by @xmath96 \label{width : bright}\end{aligned}\ ] ] so that the width is proportional to @xmath97 and the velocity is a function of time
. also the generalized phase of the bright soliton is given by the expression @xmath98\int \tilde{r}^2 dt.\end{aligned}\ ] ] note that the amplitude of the generalized bright soliton @xmath99 given by eq.([amp : bright ] ) can vary as a function of time depending upon the forms of the scattering length @xmath55 and gain / loss term @xmath50 .
consequently one can have growth or decay or constancy ( of the amplitude ) of the bright soliton for appropriate choices of @xmath55 and @xmath50 .
in particular for the specific choice @xmath100 $ ] , the amplitude @xmath101 and the soliton amplitude remains constant in time , which has not been noted in the earlier literature @xcite .
however note that the corresponding width ( @xmath102 ) as given ( [ width : bright ] ) need not be constant which can increase or decrease as the case may be . in the following sections 4 - 6
, we will discuss different kinds of interesting generalized soliton solutions for different forms of unmodulated / modulated harmonic potentials by suitably tailoring the atomic scattering length and gain / loss term . finally the normalization condition ( [ norm1 ] )
can be rewritten in terms of the soliton parameters as @xmath103 = \frac { 2 n a_b}{a_{\perp}}\exp\left[\int \gamma(\tilde{t } ) d\tilde{t}\right ] , \label{norm21}\end{aligned}\ ] ] which allows one to fix the constant @xmath76 in terms of the experimental bec parameters as @xmath104 on the other hand , when @xmath87 , one may write the dark one soliton solution in the form @xcite @xmath106 \ , \exp\left [ -\frac{i}{2}(c^2 + 4 \beta^2 ) t \right ] , \label{dark}\end{aligned}\ ] ] where the parameters @xmath107 , @xmath90 correspond to the depth and velocity of the dark soliton , respectively . here
the minus sign in the superscript means the @xmath87 case .
using eqs.([phi ] ) , ( [ qtrs ] ) and ( [ dark ] ) , one can write down the dark soliton solution of ( [ 1dgpe ] ) as @xmath108 q^-(z , t ) = \exp\left[\int \frac{\gamma(t)}{2 } dt \right ] \lambda(z , t ) q^-(z , t ) \label{phi-}\end{aligned}\ ] ] now substitute @xmath92 , @xmath58 and @xmath16 from eqs .
( [ fg ] ) , ( [ lambda ] ) and ( [ solv ] ) in the above eq .
( [ phi- ] ) , and we get [ eq:1d : dark ] @xmath109\right\ } \left\{c-2
i \beta \tanh\left [ \xi^-(z , t)\right]\right\ } \exp\left[i\eta^-(z , t)\right ] , \label{dark } \end{aligned}\ ] ] here the generalized wave variable @xmath110 which specifies the width and velocity of the dark soliton as a function of time and longitudinal distance is given by the expression @xmath111 , \end{aligned}\ ] ] and the generalized wave phase of the dark soliton is expressed as @xmath112 also the depth of the generalized dark soliton is of the form @xmath113\right\}=
\left\ { 2 \beta r_0 \sqrt{r(t ) } \exp\left[\int \gamma(t ) dt\right]\right\ } \end{aligned}\ ] ] which depends on @xmath55 and @xmath50 . as in the case of the bright soliton solution we can get stable dark soliton solutions by choosing the forms of @xmath55 and @xmath50 to balance each other so as to achieve a constant amplitude / depth . in this case also , the width depends on @xmath55 and @xmath50 and can increase / decrease or remain constant with time .
again the normalization condition ( [ norm1 ] ) of the dark soliton solution for the gp equation leads us to the condition @xmath114 dz & = 4 \beta r_0 \exp \left[\int \gamma(t ) dt \right ] = \frac { 2 n a_b}{a_{\perp}}\exp\left[\int \gamma(\tilde{t } ) d\tilde{t}\right ] , \label{norm22}\end{aligned}\ ] ] so that the parameter @xmath76 can be fixed in this case as @xmath115 now making use of the forms of @xmath67 for given @xmath82 , as presented in table [ table1 ] in appendix [ app : ricatti ] or through a detailed numerical analysis , into the above bright or dark soliton solutions ( [ bright ] ) or ( [ dark ] ) , respectively , one can analyze the nature of the bec matter wave properties . in table
[ table1 ] , we list out a chosen set of @xmath116 different choices of @xmath83 and the corresponding forms of @xmath117 for which the riccati type equation ( [ con : eqn ] ) exhibits closed form solutions . in the following sections , we shall analyse the nature of the bright and dark bec solitons , their growth , decay , collapse and revival for a select set of physically interesting choices of @xmath83 and @xmath117 from table [ table1 ] . let us first focus our attention on the case of time - independent harmonic potential for which @xmath82 in ( [ 1dgpe ] ) is a constant ( @xmath118 ) . in this case
, we shall identify the different soliton solutions for both confining , @xmath119 , as well as expulsive , @xmath120 , potentials for various forms of scattering length @xmath54 and gain / loss term @xmath50 .
it is of relevance here to point out that in recent experiments on the formation of bright matter wave solitons in an ultracold @xmath3li gas , both the confining as well as expulsive potentials have been used @xcite . in addition , such a formation of matter wave bright soliton has been achieved by temporal modulation of atomic scattering length due to feshbach resonance @xcite .
further , as pointed out earlier , we note that the formation of dark solitons , their oscillations and interaction in single component becs of @xmath0 rb atoms with confining harmonic potential have been demonstrated in recent experiments at heidelberg and hamburg universities @xcite .
motivated by these experiments , we fix the parameters such as the trap frequencies ( @xmath23 and @xmath121 ) , number of atoms ( @xmath21 ) and scattering length in the transverse direction , @xmath29 , to confirm to the experimental values . in particular we choose the parameters as @xmath122hz , @xmath123 , and @xmath124 @xmath4 m .
the trap frequency in the axial direction is chosen as @xmath125hz for the case of the confining potential , while we choose @xmath126hz for the expulsive potential .
correspondingly @xmath127 takes the value @xmath128 so that @xmath129 is taken as @xmath130 in our analysis in the following . accordingly , in the entire manuscript the unit of time is taken as milliseconds and the length is expressed in micrometers .
also it is not difficult to provide a simple estimation of the validity of the reduction of 3d gp equation to the 1d gp equation given by eq .
( [ cre ] ) in terms of dimensional units .
for example , when one considers a bec of @xmath0 rb atoms with @xmath131 , for @xmath132 m eq .
( [ cre ] ) gives the estimate @xmath133 nm , which is experimentally realizable .
note that , in the presence of the gain / loss term the criteria for the validity of 1d approximation becomes ( see eq .
( [ cre1 ] ) ) @xmath134 \ll 1 $ ] .
in the present analysis , we suitably tune @xmath55 , @xmath83 and @xmath50 to satisfy the above criterion , which may be still realizable ( which will be shown in the appendix [ app : validity ] for various forms ) . in the next section
we shall discuss the formation of both bright and dark matter wave solitons in time independent confining harmonic potential for different choices of temporal modulation of the atomic scattering length and gain / loss term .
this is followed by a demonstration of the solitons in time independent expulsive harmonic potential case in sec .
[ ind : exp ] . finally , in sec . [ de ]
, we discuss the formation of both bright and dark matter wave solitons in time - dependent confining potential such as double modulated periodic potential and step - wise modulated potential .
we consider the time independent confining potential with the ratio of trap frequencies obeying the relation @xmath118 , where @xmath129 is a constant .
using @xmath118 in eq .
( [ con : eqn ] ) , we obtain a particular solution for @xmath67 as , [ see table [ table1 ] ] @xmath135 since @xmath136 , one can identify different types of soliton solutions of eq .
( [ 1dgpe ] ) for appropriate forms of @xmath137 and @xmath138 .
accordingly , we will choose these forms and investigate the nature of matter waves . when the gain term is absent in eq .
( [ 1dgpe ] ) so that for @xmath139 the atomic scattering length can be written , using eqs .
( [ rtilde ] ) and ( [ eq : rtilde_1 ] ) , as @xmath140 .
accordingly , substituting the form ( [ eq : rtilde_1 ] ) into ( [ bright ] ) the magnitude of the bright soliton solution of ( [ 1dgpe ] ) is given by @xmath141 .
\label{bright_81}\end{aligned}\ ] ] it may appear that the presence of @xmath142 in the the amplitude of the above bright soliton will lead to singularities at finite time .
however , at the same time when @xmath143 the envelope term [ @xmath144 also goes to zero sufficiently fast so as to keep the overall amplitude finite .
note that the width of the soliton is proportional to @xmath145 and hence it is varying periodically with time .
[ figure1](a ) depicts the spatiotemporal pattern of condensate wave function in a confining oscillator potential for a typical choice of parameters .
it is clear from fig .
[ figure1](a ) that the bright matter wave = @xmath127 , @xmath139 , and @xmath146 and ( b ) the corresponding contour plot .
the parameters are chosen as @xmath147 , @xmath148 , @xmath149 , @xmath150 , and @xmath151 in eq .
( [ bright_81 ] ) . ]
soliton wave for @xmath140 and @xmath139 shows periodic oscillation in amplitude due to the presence of the periodic terms @xmath152 and @xmath153 in the right hand side of eq .
( [ bright_81 ] ) .
[ figure1](b ) shows the periodic oscillation of the width .
in this case the corresponding dark soliton solution can be written as @xmath154\right\}^{\frac{1}{2}}. \label{dark_81}\end{aligned}\ ] ] however , in the absence of a suitable gain / loss term ( @xmath139 ) , the above expression becomes singular at finite values of @xmath18 and so is not physically interesting .
a physically interesting case can be realized by choosing the gain term as @xmath156 , ( @xmath157 , @xmath158 and @xmath159 are parameters ) , so that @xmath160 \label{r_4.1.3}.\end{aligned}\ ] ] the above form of @xmath50 was suggested recently by atre , panigrahi and , @xmath155 , @xmath146 with @xmath161 , @xmath162 , @xmath163 , @xmath164 , @xmath165 , @xmath166 , @xmath167 , @xmath168 in eq .
( [ bright_82 ] ) and ( b ) the corresponding contour plot .
note the periodic collapse and revival of the wave with increase in amplitude over time . ]
agarwal @xcite . in this case , the bright soliton solution of ( [ 1dgpe ] ) can be written as @xmath169 \notag \\ & \ , \times{\;{\rm sech}}\left[\left(\frac{n a_b}{a_{\perp } } \sec(\omega_0 t + \delta)\right)z -\left(\frac{n^3 a_b^3 } { a^2 a_{\perp}^3 } b+ \frac{n^2 a_b^2 } { \sqrt{2 } a a_{\perp}^2}c\right ) \frac{\tan(\omega_0 t+\delta)}{\omega_0}\right ] .
\label{bright_82}\end{aligned}\ ] ] the amplitude of the soliton is @xmath170 $ ] and the width is proportional to @xmath145 , indicating collapse and revival of the condensate . fig .
[ figure2 ] illustrates the collapse and revival of the condensate wave function with an overall increase in the amplitude .
the collapse and revival of atomic condensate and amplification [ see fig . [ figure2 ] ] through periodic exchange of atoms with the background is obviously due to the partially sinusoidal nature of the gain function , @xmath155 .
we also note here that , strecker et al .
have experimentally observed the formation of the bright soliton of @xmath3li atom in quasi-1d optical trap , by magnetically tuning the interactions in a stable bec from repulsive to attractive @xcite .
theoretically , atre , panigrahi and agarwal have shown that by suitably tailoring the gain profile in the same range of the experimental parameters , the matter wave soliton exhibits dramatic collapse and revival of condensate with an increase in the amplitude @xcite . the corresponding expression for dark soliton
can be written as @xmath171 \notag \\ & \ , \times \left\{c^2 + 4 \beta^2 \tanh^2\left [ \left(\frac{n a_b}{a_{\perp } } \sec(\omega_0 t + \delta)\right)z -\left(\frac{n^3 a_b^3 } { 2 \beta^2 a_{\perp}^3}b + \frac{n^2 a_b^2 } { 2 \beta a_{\perp}^2}c\right ) \frac{\tan(\omega_0 t+\delta)}{\omega_0}\right]\right\}^{\frac{1}{2}}. \label{dark_4.1.4}\end{aligned}\ ] ] the above expression , as in the case of the bright soliton ( [ bright_82 ] ) , clearly shows that a periodic collapse and revival of the matter wave dark , @xmath155 , @xmath146 with @xmath172 , @xmath162 , @xmath163 , @xmath164 , @xmath165 , @xmath166 , @xmath167 , @xmath168 in eq .
( [ dark_4.1.4 ] ) and ( b ) the corresponding contour plot , again portraying a collapse and revival with increase in amplitude over time . ]
soliton can also be realized for the choice of @xmath55 as given by ( [ r_4.1.3 ] ) .
the dark soliton ( [ dark_4.1.4 ] ) is depicted in fig .
[ figure3 ] .
finally we consider the case with another form of gain term as @xmath173 so that @xmath174 $ ] which can be obtained from eqs .
( [ rtilde ] ) and ( [ eq : rtilde_1 ] ) . in this case , the bright soliton solution of ( [ 1dgpe ] ) can be written as @xmath175 .
\label{bright_83}\end{aligned}\ ] ] one may note that in this case the amplitude of the soliton ( @xmath176 ) remains constant in time .
however , the corresponding width ( @xmath177 ) is varying periodically in time , leading to a snake - like effect . and
@xmath173 , @xmath178 $ ] .
( b ) the corresponding contour plot .
the other parameters are @xmath147 , @xmath150 , @xmath179 , @xmath166 , and @xmath148 in eq .
( [ bright_83 ] ) . ] figs .
[ fig4 ] shows the bright solitary wave ( [ bright_83 ] ) exhibiting a snake - like effect .
it is interesting to point out that a similar snake - like effect has been demonstrated by serkin and hasegawa in the context of dispersion managed optical soliton with sinusoidally varying dispersion coefficient , nonlinearity , and gain or absorption term @xcite . and @xmath173 , @xmath178 $ ] case .
( b ) the corresponding contour plot .
the other parameters are @xmath180 , @xmath150 , @xmath179 , @xmath166 , and @xmath148 in eq .
( [ dark_83 ] ) . ] the dark soliton solution , in this case , is given by @xmath181\right\}^{\frac{1}{2}}\label{dark_83}\end{aligned}\ ] ] as pointed out earlier in sec .
[ sec4:case5 ] for the bright soliton , the above dark soliton ( with constant amplitude and periodically varying width ) also exhibits a snake - like effect .
[ fig5 ] depicts the propagation of snake - like dark solitary wave .
one may note that a similar type of snake - like dark matter wave solitons and their collisions have been observed in the recent experiments of @xmath0rb condensates @xcite with confining potential . in this context
, it is quite interesting to have an exact solution of the form ( [ dark_83 ] ) for the snake - like dark soliton solutions for confining potential with time varying scattering length and atom gain / loss .
in this section , we consider the case of time - independent expulsive potential @xmath184 . for this choice , substituting @xmath184 in the integrability condition ( [ con : eqn ] )
, one may identify three independent particular solutions of physical interest for @xmath67 @xcite , namely ( i ) @xmath185 , ( ii ) @xmath186 , and ( iii ) @xmath187 [ see table [ table1 ] ] .
however , the stabilization of bright and dark matter wave solitons is solely dependent on the constancy of the product term @xmath188= \sqrt{r(t ) } \exp[\int \gamma(t ) dt]$ ] .
the corresponding restriction on the parameters is called safe range parameters . when the parameters are chosen outside the safe range
, the bright and dark solitons may become unstable , that is , one may witness either growth or decay , as studied for ( i ) @xmath139 and @xmath189 in ref .
@xcite and ( ii ) @xmath190 constant and @xmath191 in ref . @xcite .
in fact , stable solitons can be formed only if we choose the form of the scattering length @xmath55 and the gain / loss term @xmath50 as @xmath100 $ ] , so that the soliton amplitude remains constant in time . in the following , we discuss various kinds of soliton solutions of eq .
( [ 1dgpe ] ) for different forms of @xmath50 and @xmath55 both within as well as outside the safe range parameter values in all the above three cases of @xmath67 . for @xmath186 case , we study the solution of eq .
( [ 1dgpe ] ) for both the negative ( loss ) and positive ( gain ) values of @xmath138 . in this case , we find stable solitons for the safe range value @xmath193 and growth of matter wave soliton for the out of safe range value @xmath194 , where @xmath195 .
first let us consider the inclusion of depletion / loss term in ( [ 1dgpe ] ) , that is , @xmath197 . in this case
@xmath198 $ ] , obtained from eqs .
( [ rtilde ] ) and ( [ eq : rtilde_1 ] ) , implying that the atomic scattering length is exponentially growing with time .
accordingly the bright soliton solution of ( [ 1dgpe ] ) is given by @xmath199 , \label{bright3_1}\end{aligned}\ ] ] the solitary wave solution occurs due to the balance between the negative gain term and the exponential growth of the atomic scattering length .
it may be noted that , here , the overall amplitude remains constant ( @xmath176 ) .
[ fig6](a ) shows the bright solitary wave ( [ bright3_1 ] ) for @xmath204 and @xmath198 $ ] .
however , the width of the wave ( @xmath205 ) decreases exponentially with time as seen from fig .
[ fig6](b ) . for the safe range parameter @xmath196 the corresponding dark soliton solution of ( [ 1dgpe ] ) is given by the expression @xmath206
\right\}^{\frac{1}{2}}. \label{dark3_1}\end{aligned}\ ] ] one may again note that the amplitude remains constant ( @xmath207 ) in the above expression . figs .
[ fig7 ] illustrates the dynamics of the dark solitary wave . once again , as in the case of bright soliton , the width of the wave is decreasing exponentially with time [ see fig .
[ fig7](b ) ] , while the amplitude remains constant . for positive gain term @xmath209 case
, one finds that @xmath137 is independent of time , which indicates that the atomic scattering length is constant . in this case
, the bright soliton solution of ( [ 1dgpe ] ) can be written as one may note that , in the above expression , the amplitude of bright soliton grows exponentially ( @xmath213 ) due to the presence of the gain term ( parameters in the out of safe region ) , while the width ( @xmath214 ) decreases exponentially with time .
[ fig8](a ) depicts the exponential growth of bright solitary wave for @xmath215 and @xmath216 constant . here
the width of the wave decreases with respect to time exponentially as seen in fig .
[ fig8](b ) . next , in the case of @xmath217 similar type of stable solitary waves and growth / decay of solitary waves can be identified for suitable forms of @xmath55 and @xmath50 .
also one may note here that the width of the wave will be increasing exponentially with time ( i.e , proportional to @xmath218 ) in contrast to the case discussed in sec .
[ + omega ] .
since the details are similar we do not present them here separately .
[ [ gamma0-decay - of - brightdark - soliton - and - two - soliton - bound - state ] ] @xmath139 , decay of bright / dark soliton and two - soliton bound state : ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ first we choose @xmath222 .
consequently , @xmath137 takes the form @xmath223 as deduced from eqs .
( [ rtilde ] ) and ( [ eq : rtilde_1 ] ) .
the bright soliton solution of ( [ 1dgpe ] ) for @xmath222 can be written as in the present case , another interesting structure similar to a bright two - soliton bound state [ shown in fig . [ figure22 ] ] occurs when the propagation term is absent in the above soliton solution ( i.e , velocity is zero ) when the parameters @xmath75 and @xmath90 in eq .
( [ bright_61 ] ) are chosen as @xmath164 , @xmath165 .
one may note that a similar kind of two - soliton like bound state has been recently identified by atre , panigrahi and agarwal using a different approach @xcite .
next , the dark soliton solution of ( [ 1dgpe ] ) for the present case can be written as @xmath226\right\}^{\frac{1}{2}}. \label{dark_61}\end{aligned}\ ] ] figs .
[ figure15 ] shows the decaying nature of the dark solitary wave solution for @xmath227 and @xmath228 , where the width also increases exponentially in time .
similar to the case of bright soliton solution , a dark two - soliton like bound state can arise when @xmath164 , @xmath165 in eq .
( [ dark_61 ] ) .
[ figure23 ] shows such a dark two - soliton like bound state for @xmath164 , @xmath165 in eq .
( [ dark_61 ] ) .
next let us choose the form of the gain term as @xmath231 $ ] .
then the form of the atomic scattering length becomes @xmath232 , as obtained from eqs .
( [ rtilde ] ) and ( [ eq : rtilde_1 ] ) .
consequently , the bright soliton solution of ( [ 1dgpe ] ) can be written as @xmath233 , \label{bright_62}\end{aligned}\ ] ] fig . [
figure16](a ) shows the corresponding bright solitary wave .
one may note that the amplitude of the wave is constant ( @xmath234 ) over all time due to the balance between the gain and atomic scattering length but the width of the wave increases with respect to time ( proportional to @xmath235 ) as shown in fig .
[ figure16](b ) .
next , the corresponding dark soliton solution of ( [ 1dgpe ] ) can be written as @xmath236\right\}^{\frac{1}{2 } } \label{dark_62}\end{aligned}\ ] ] which is shown in figs .
[ figure17 ] . here
the amplitude , @xmath237 , remains constant while the width increases with time .
[ [ gamma - a_1-t - a_2-sinkappa - t - collapse - and - revival - of - the - brightdark - solitary - waves ] ] @xmath238 , collapse and revival of the bright / dark solitary waves : ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ finally , we consider the case with periodically varying gain term of the form @xmath239 . in this case
, one may obtain the expressions for the atomic scattering length from eqs .
( [ rtilde ] ) and ( [ eq : rtilde_1 ] ) as @xmath240 { \;{\rm sech}}(\omega_0 t+ \delta).\end{aligned}\ ] ] here again we observe the periodic collapse and revival of the bright and dark solitary waves , similar to the case of the confining potential as shown in fig .
[ figure2 ] . however , the amplitude is decaying with time ( @xmath241 @xmath242 $ ] ) while in fig . [ figure2 ]
it increases with time ( @xmath243 @xmath242 $ ] ) .
however , we do not present the further details as the qualitative features are similar to the case [ case3_fig2 ] .
in the above we have identified various interesting soliton solutions that exist in the case of time - independent harmonic potentials .
a summary of these results are tabulated in table [ table3 ] for easy reference .
it is also of wide interest at present to look at the nature of the solutions in the presence of time varying harmonic trap potentials as experiments in bec with time varying harmonic trap potentials are also possible . in the following section , we consider certain interesting soliton solutions that exist in a select set of time - dependent traps .
j. r. ensher , d. s. jin , m. r. matthews , c. e. wieman , e. a. cornell , bose - einstein condensation in a dilute gas : measurement of energy and ground - state occupation , phys .
77 ( 1996 ) 49844987 .
b. p. anderson , p. c. haljan , c. a. regal , d. l. feder , l. a. collins , c. w. clark , e. a. cornell , watching dark solitons decay into vortex rings in a bose - einstein condensate , phys .
86 ( 2001 ) 29262929 .
j. denschlag , j. e. simsarian , d. l. feder , c. w. clark , l. a. collins , j. cubizolles , l. deng , e. w. hagley , k. helmerson , w. p. reinhardt , s. l. rolston , b. i. schneider , w. d. phillips , generating solitons by phase engineering of a bose - einstein condensate , science 287 ( 2000 ) 97101 .
k. bongs , s. burger , d. hellweg , m. kottle , s. dettmer , t. rinkleff , l. cacciapuoti , j. arlt , k. sengstock , w. ertmer , motion of dark solitons in trapped bose - einstein condensates , j. opt .
b. 5 ( 2003 ) s124 . c. becker , s. stellmer , p. soltan - panahi , s. dorscher , m. baumert , e .-
richter , j. kronjger , k. bongs , k. sengstock , oscillations and interactions of dark and dark - bright solitons in bose - einstein condensates , nature physics 4 ( 2008 ) 496501 .
a. weller , j. p. ronzheimer , c. gross , j. esteve , m. k. oberthaler , d. j. frantzeskakis , g. theocharis , p. g. kevrekidis , experimental observation of oscillating and interacting matter wave dark solitons , phys .
( 2009 ) 130401 .
s. stellmer , c. becker , p. soltan - panahi ,
e. m. richter , s. drscher , j. k. m. baumert , k. bongs , k. sengstock , collisions of dark solitons in elongated bose - einstein condensates , phys .
( 2008 ) 120406 .
j. stenger , s. inouye , m. r. andrews , h .- j .
miesner , d. m. stamper - kurn , w. ketterle , strongly enhanced inelastic collisions in a bose - einstein condensate near feshbach resonances , phys
82 ( 1999 ) 24222425 .
s. k. adhikari , stabilization of bright solitons and vortex solitons in a trapless three - dimensional bose - einstein condensate by temporal modulation of the scattering length , phys .
a 69 ( 2004 ) 063613 .
z. x. liang , z. d. zhang , w. m. liu , dynamics of a bright soliton in bose - einstein condensates with time - dependent atomic scattering length in an expulsive parabolic potential , phys .
94 ( 2005 ) 050402 .
b. li , x .- f .
zhang , y .- q .
li , y. chen , w. m. liu , solitons in bose - einstein condensates with time - dependent atomic scattering length in an expulsive parabolic and complex potential , phys .
a 78 ( 2008 ) 023608 . v. n. serkin , a. hasegawa , exactly integrable nonlinear schrdinger equation models with varying dispersion , nonlinearity and gain : application for soliton dispersion , ieee selected topics in quant .
( 2002 ) 418 431 . c. h. tenorio , e. v. vargas , v. n. serkin , m. a. granados , t. l. belyaeva , r. p. moreno , l. m. lara , dynamics of solitons in the model of nonlinear schrdinger equation with an external harmonic potential : ii .
dark solitons , quant .
35 ( 2005 ) 929937 .
a. grlitz , j. m. vogels , a. e. leanhardt , c. raman , t. l. gustavson , j. r. abo - shaeer , a. p. chikkatur , s. gupta , s. inouye , t. rosenband , w. ketterle , realization of bose - einstein condensates in lower dimensions , phys .
87 ( 13 ) ( 2001 ) 130402 . | we investigate the exact bright and dark solitary wave solutions of an effective one dimensional ( 1d ) bose - einstein condensate ( bec ) by assuming that the interaction energy is much less than the kinetic energy in the transverse direction .
in particular , following the earlier works in the literature prez - garca et al .
@xcite , serkin et al .
@xcite , grses @xcite and kundu @xcite , we point out that the effective 1d equation resulting from the gross - pitaevskii ( gp ) equation can be transformed into the standard soliton ( bright / dark ) possessing , completely integrable 1d nonlinear schrdinger ( nls ) equation by effecting a change of variables of the coordinates and the wave function .
we consider both confining and expulsive harmonic trap potentials separately and treat the atomic scattering length , gain / loss term and trap frequency as the experimental control parameters by modulating them as a function of time . in the case
when the trap frequency is kept constant , we show the existence of different kinds of soliton solutions , such as the periodic oscillating solitons , collapse and revival of condensate , snake - like solitons , stable solitons , soliton growth and decay and formation of two - soliton like bound state , as the atomic scattering length and gain / loss term are varied .
however when the trap frequency is also modulated , we show the phenomena of collapse and revival of two - soliton like bound state formation of the condensate for double modulated periodic potential and bright and dark solitons for step - wise modulated potentials . , , and bose - einstein condensates ; gross - pitaevskii equation ; nonlinear schrdinger equation ; soliton solutions .
05.45.-a , 03.75.lm , 03.75.nt , 67.85.jk |
networks in particular large networks with many nodes and links are attracting widespread attention .
the classic reviews @xcite with their primary focus on structural properties have been followed up by more recent ones addressing the role of dynamics , such as spreading and synchronisation processes on networks , as well as the role of weights and mesoscopic structures , i.e. cliques ( fully connected subgraphs ) and communities ( groups of densely interconnected nodes ) , within networks @xcite .
although several different measures for characterising networks have been presented , for example in a recent survey @xcite , the simple concept of vertex degree remains unrivalled in its ability to capture fundamental network properties .
when comparing the degrees of connected vertices , however , one often finds that they are correlated , a quality that gives rise to a rich set of phenomena @xcite .
degree correlations constitute a central role in network characterisation and modelling but , in addition to being important in their own right , also have substantial consequences for dynamical processes unfolding on networks . given the increasing current interest in network dynamics , understanding structural correlations remains important and timely .
in this paper we provide a detailed mathematical formalism for modelling networks with correlations .
it is built around a statistical description of inter - node linkages as opposed to single - node degrees .
most works devoted to analytical calculations of correlations in models have been performed only for particular cases @xcite .
we provide a brief overview of degree correlations for network structure and dynamics in section [ sec : correlations ] and discuss how the work presented relates to existing studies .
section [ sec : notation ] contains the notation used throughout the rest of the paper .
we highlight the importance of acknowledging these correlations in section [ sec : saramaki ] through reviewing a model of network growth proposed by saramki and kaski @xcite ( and subsequently by evans and saramki @xcite ) which employs random walkers to grow a network .
the _ link - space _ formalism , which lies at the core of the paper , is introduced in section [ sec : linkspace ] .
the formalism comprises a master equation description of the evolution of a very specific matrix construction which we term the _ link - space matrix_. the formalism can be implemented in a number of ways . to demonstrate its use , we apply it to two well - known , non - equilibrium examples , namely random - attachment and barabsi - albert ( ba ) preferential - attachment networks @xcite in section [ sec : linkspace ] and solve the so - called link - space matrix analytically for these models . in section [ sec : linkspace ]
, we also apply it to the classical equilibrium random graph of erds and rnyi @xcite ( er ) using some of the ideas discussed in @xcite which , interestingly , requires a full , time - dependent solution of the link - space master equations . the cumulative link - space introduced in section [ sec : linkspace ] aids comparison between simulated and analytic link - space matrices and could ,
in principle , be applied to empirical networks to better ascertain their degree correlations .
analytic derivation of link - space matrices allows detailed analysis of the degree - degree correlations present within these networks as is demonstrated in section [ sec : degrees ] . within section [ sec : degrees ]
, we also show how , for arbitrary degree distribution , a link - space matrix can be derived which has no correlations present , i.e. the link - space is representative of a perfectly non - assortative network with that degree distribution .
we then consider the counter - intuitive prospect of finding steady states of decaying networks in section [ sec : decay ] . in section [ sec : redner ] , we discuss the growing network with redirection ( gnr ) model of krapivsky and redner @xcite . in section [ sec : model ] , we introduce a simple one - parameter network growth algorithm with a similar redirection process which is able to produce a wide range of degree distributions .
the model is interesting in its own right in the sense that it makes use of only local information about node degrees . here
the link - space formalism allows us to identify the transition point at higher exponent degree distibutions switch to lower exponent distributions with respect to the ba model .
we conclude in section [ sec : lsconclusion ] .
a discussion of correlations might naturally start with the phenomenon of _ clustering_. the highly influential paper of barabsi and albert and many subsequent papers focussed on the modelling and analysis of models which replicate power - law degree distributions observed in many real - world systems @xcite .
however , the ba preferential attachment mechanism lacks certain key features observed in many of those systems , one such feature being _
clustering_. clustering is an important local statistic in many networks and reflects the connectedness of neighbours of a node . in a social context , if person @xmath3 is friends with both @xmath4 and @xmath5 , one might expect some kind of link between @xmath4 and @xmath5 .
the measurement of the connectedness of neighbours of nodes is the clustering coefficient .
this is often averaged over all nodes and a single value is given as the clustering coefficient of the network @xcite .
an explicit example is given in figure [ fig : clustering1 ] .
the node labelled @xmath1 has @xmath6 neighbours ( in this case @xmath7 ) . between these
there could be a possible @xmath8 undirected links of which @xmath9 exist .
for this example , @xmath10 links exist out of the possible @xmath11 between the neighbours of node @xmath1 .
the clustering of node @xmath1 is given by @xmath12 and would be @xmath13 for our example .
the clustering nature of a network can also be expressed as the average over all nodes of degree @xmath14 giving a clustering distribution ( or _ spectrum _ ) , @xmath15 .
node @xmath1 has @xmath7 neighbours which could potentially share 10 links between them .
only three out of these @xmath11 possible links exist between the neighbours of @xmath1 .
its clustering coefficient is then expressed @xmath16.,scaledwidth=55.0% ] as noted by klemm and eguluz , in a ba network , the average clustering of a given node is independent of its degree @xcite , in contrast to the findings of fronczak _ et al . _
@xcite , and tends to zero in the large - size ( thermodynamic ) limit .
a network model which does exhibit high clustering is that of watts and strogatz @xcite in which a random rewiring process is carried out on an initially regular lattice .
the networks generated by such a process feature short average shortest path lengths between node pairs ( the _ small - world _
however , these networks do not exhibit a power - law degree distribution .
subsequently , there have been many efforts to build models which can encompass both of these features such as the non - equlibrium growing network model proposed by klemm and eguluz which features deactivation of a nodes availibility to be connected to @xcite .
other models such as that of holme and kim @xcite and that introduced by toivonen _
_ @xcite modified the original ba preferential attachment mechanism , allowing further links between the new node and neighbours of the preferentially selected node .
the social network model of toivonen _ et al . _ produced communities with dense internal connections .
_ formulated a scaling assumption and a mean - field theory of clustering in growing scale - free networks and applied it to the holme and kim mechanism @xcite .
as discussed by bogu and pastor - satorras , clustering in networks is closely related to degree correlations @xcite .
infact , based on the work of szab _ et al .
_ , barrat and pastor - satorras introduced a framework for computing the rate equation for two vertex correlations in the continuous degree and continuous time approximation @xcite .
we shall now describe these degree correlations .
vertex degree correlations are measures of the statistical dependence of the degrees of neighbouring vertices in a network . in general , @xmath17-vertex degree correlations , or @xmath17 point correlations , can be fully characterised by the conditional probability distribution that a vertex of degree @xmath18 is connected to a set of @xmath17 vertices with degrees @xmath19 .
two and three point correlations are of particular interest in complex networks as they can be related to network assortativity and clustering respectively . more specifically , two vertex degree correlations ( two point correlations ) can be expressed as conditional probability @xmath20 that a vertex of degree @xmath14 is connected to a vertex of degree @xmath21 .
similarly , three vertex degree correlations ( three point correlations ) can be fully characterised by the conditional probability distribution @xmath22 that a vertex of degree @xmath14 is connected to both a vertex of degree @xmath21 and a vertex of degree @xmath23 .
this implies that the degrees of neighbouring nodes are not statistically independent .
reliable estimation of @xmath20 and @xmath22 requires a large amount of data and , in practice , one often resorts to related measures . instead of @xmath20 , the average degree of nearest neighbours , @xmath24 ,
is often measured .
this can be formally related to @xmath20 @xcite . if @xmath24 increases with @xmath14 ,
high degree vertices tend to connect to high degree vertices .
a network with this property would be described as being _ assortative _ or displaying positive degree correlations .
if @xmath24 decreases with @xmath14 , high degree vertices tend to connect to low degree vertices ( _ disassortative _ or negatively correlated ) @xcite .
an alternative to @xmath24 is to use a normalised pearson s correlation coefficient of adjacent vertex degrees providing a single number measure of assortativity as suggested by newman to further classify networks @xcite .
these approaches are discussed in more detail in section [ sec : degrees ] .
to characterise three point correlations , instead of using @xmath22 , one can employ the clustering spectrum @xmath15 , which can be related to @xmath25 @xcite .
in many real world networks such as the internet @xcite , the clustering spectrum is a decreasing function of degree and while this is sometimes interpreted as a signature of hierarchical structure in a network , soffer and vzquez suggested that this is a consequence of degree - degree ( two point ) correlations that enter the definition of the standard clustering coefficient @xcite .
the authors introduced a different definition for the clustering coefficient that does not have the degree - correlation ` bias ' , i.e. a three - point correlation measure that filters out two - point correlations .
following the suggestion of maslov et al .
@xcite that these phenomenon might arise from topological constraints rather than evolutionary mechanisms , park and newman demonstrated that dissasortative degree correlations observed in the internet could be explained via the restriction of there being no double edges between nodes @xcite .
in contrast , social networks have been found to be assortative @xcite .
similarly , catazaro _ et al . _ observed that the network of scientific collaborations was assortative and presented a model to reproduce this feature @xcite .
functional processes occurring on networks are influenced by degree correlations , highlighting the importance of their role in complex networks .
eguluz and klemm considered highly clustered scale - free networks and showed that correlations play an important role in epidemic spreading @xcite .
the time average of the fraction of infected individuals in the steady - state undergoes a phase transition at a finite critical infection probability .
they related this critical threshold to the transmission probability and the mean degree of nearest neighbours of all nodes in the system @xmath26 , the conjectured criterion for epidemic spreading being related to the product of the two .
the value @xmath26 scales with system size in their highly clustered scale - free network slower than in the random scale - free which is a byproduct of the dissasortativity of the system .
consequently , whereas in random scale - free networks in which viruses with extremely low spreading probabilities can prevail , the absence of connections between highly connected nodes in highly clustered scale - free networks protects the system against epidemics @xcite .
interestingly , bogn _ et al .
_ assert that any scale - free network of appropriate exponent will have diverging @xmath26 in the thermodynamic ( infinite size ) limit , resulting in no threshold properties for epidemic spreading regardless of the correlations within the network @xcite in contrast to the earlier suggestion of bogn and pastor - satorras @xcite .
brede and sinha induced correlations to erds - rnyi and scale - free networks by rewiring them appropriately to examine their dynamic stability @xcite .
each node had an associated variable who s evolution was governed by some non - linear function of the variables associated with neighbouring nodes .
they mapped the adjacency matrices into jacobian matrices and examined the largest eigenvectors , reflecting the decay rates of perturbations of the nodes variables about their equilibrium states .
they found that positive correlations within the network reduced the dynamical stability @xcite .
similarly , bernardo _ et al .
_ induced negative degree correlations in scale - free networks whose links couple non - linear oscillators . through analysis of the eigenratio ( ratio of the eigenvalues ) of the laplacian of such networks , they found that network synchronisability improved as the network was made more dissassortative @xcite .
based on this result , they conjectured that negative degree correlations may emerge spontaneously as the networked system attempts to become more stable @xcite .
the same authors found similar results to hold also for weighted networks @xcite .
maslov and sneppen found that in regulatory networks , links between highly connected proteins were relatively rare increasing the robustness of the network to perturbations through localising deletous perturbations " @xcite .
this is consistent with the findings of berg _ et al .
_ whose model of the evolution of protein interaction networks exhibited disassortativity consistent with their empirical findings @xcite .
krapivsky _ et al .
_ used a master equation method , in which rate equations for the densities of nodes of a given degree are employed , to investigate the steady state of the ba preferential attachment mechanism @xcite .
this is a general method which can be applied to various growing network models . to extend the method to encompass two - point correlations , krapivsky and redner applied the approach to the number @xmath27 of nodes with total degree @xmath14 connected to ancestor nodes of total degree @xmath28 in a _ directed _ network @xcite .
they solved analytically the master equations for the steady - state of a specific directed network growth model , namely the growing with redirection algorithm @xcite which is further discussed in section [ sec : redner ] and is similar to the mixture model introduced in section [ sec : model ] .
bogu and pastor - satorras considered a link orientated description of two point correlations . for
_ undirected _ graphs , they introduced a symmetric matrix whose elements @xmath29 represent the number of links connecting nodes of degree @xmath14 to nodes of degree @xmath28 @xcite .
bogu and pastor - satorras related this matrix to the joint degree distribution @xmath30 but noted that finite size effects make empirical evaluation of the matrix difficult , suggesting the use of the spectrum @xmath31 as a more suitable observable . here
, we introduce a similar matrix construction which can be applied to both the undirected _ and _ directed scenarios .
we call this the _ link - space _ matrix .
we show that it is possible to construct master equations to model the evolution of this matrix which can be applied to a wide variety of network evolution algorithms retaining important degree correlations which can be critical to the network s development .
this framework is termed the _ link - space formalism _ and is introduced in section [ sec : linkspace ] . in certain cases , these master equations can be solved analytically providing a full time - dependent solution or a steady - state solution of the link - space matrix of the network . from these analytically derived link - space matrices both the degree and joint degree distributions
can be obtained allowing accurate analysis of degree correlations .
the formalism also allows the derivation of the form of networks of predetermined correlation properties such as a perfectly non - assortative network ( section [ sec : degrees ] ) and the derivation of the steady - state solutions to network decay algorithms ( section [ sec : decay ] ) .
the formalism can also be employed in its iterative guise to provide approximations to the steady - state when an analytic solution is not possible such as for the mixture model of section [ sec : model ] .
one and two - point correlations have natural physical counterparts within the network , specifically the nodes and links .
the one - point correlations @xmath32 are simply related to the fraction of nodes in the network with degree @xmath14 . as described in detail within this paper ,
two - point correlations @xmath30 are related to the fraction of links within a network connecting nodes of degree @xmath14 with nodes of degree @xmath21 , hence the term _ link - space_. one can extend the analysis to three - point correlations , @xmath33 .
this would be related to the number ( fraction ) of pairs of links within a network sharing a common node of degree @xmath14 , the remaining link ends being connected to nodes of degrees @xmath21 and @xmath23 and all possible ` open triangles ' within the network would have to be considered .
clearly , the process can be extended to arbitrary @xmath17-point correlations although the physical interpretation of the appropriate measurable quantities will become increasingly obscure .
throughout the subsequent discussions , we shall use the following notation : @xmath34
the use of random walkers to generate networks was introduced by saramki and kaski @xcite . in their model , as in many non - equilibrium evolving networks , one new node is attached to the existing network at each timestep with @xmath35 undirected links .
the simplest implementation of their growth algorithm attaches this new node to the existing network component with @xmath36 links as follows : + 1 .
select a node at random within the existing network .
2 . perform a two - step random walk .
3 . establish ( undirected ) links between the nodes arrived at within the existing network and the new node .
note that a link is not established between the new node and the initial node of the random walk .
the generalisation is clear for @xmath37 links per new node attached . to summarize , a random walk of length @xmath35 steps
is performed within the existing network from a randomly chosen start point , and a link between the new node and each of the nodes reached in the random walk is formed with the exception of the start point .
+ the degree distribution of an example network using the saramki / kaski algorithm grown to @xmath38 nodes.,scaledwidth=65.0% ] certainly this yields results comparable to the scale - free preferential attachment of the ba model as shown in figure [ fig : saramaki ] . however , the analytic approximations derived by saramki and kaski in reference @xcite do not hold in all cases .
their analysis is as follows .
consider choosing some vertex labeled @xmath39 initially ( at random ) within the existing network comprising @xmath40 nodes .
this occurs with probability : @xmath41 we now wish to consider the probability of moving to a neighbour vertex labeled @xmath42 after one step of the random walk . using bayes theorem , @xmath43 where @xmath44 denotes arriving at node @xmath42 after one step , @xmath45 denotes the probability of arriving at at @xmath42 if node @xmath39 is the initial vertex chosen and @xmath46 is the conditional probability that , given we have arrived at @xmath42 , we originated at @xmath39 .
the value @xmath45 can be trivially written @xmath47 where @xmath48 is the degree of vertex @xmath39 .
so far so good
. however , the next justification is somewhat misunderstood .
it is often thought @xcite that the conditional probability @xmath46 can be written @xmath49 which combined with the above yields : @xmath50 this compares to the the ba preferential attachment probability for attaching to a particular node of degree @xmath1 : @xmath51 such that the probability of attaching to a node is proportional to its degree . a trivial example network.,scaledwidth=65.0% ] to understand the problems with this analysis ,
consider a real system of nodes , as described in figure [ fig : simplenet ] .
we can write the probability associated with arriving at node @xmath42 after a random walk of one step in terms of the initial node selected being one of the neighbours of @xmath42 and the subsequent probability of moving to node @xmath42 : @xmath52 clearly this is not the same as the earlier expression ( [ eqn : wrong ] ) . nor
should it be .
the earlier expression would yield different results depending on which node was labeled @xmath39 which is clearly a nonsense .
another interesting assumption often made is that because the initial node is chosen at random , its vertex degree will be distributed similarly to that of the entire network @xcite .
if this is so , ( [ eqn : wrong ] ) can be written in terms of arriving at a particular of node of degree @xmath1 after randomly choosing an initial vertex and moving at random to one of its neighbours : @xmath53 where @xmath54 is the mean degree of the existing network component . even averaging over all nodes of degree @xmath1 , and ( incorrectly ) assuming that on average the neighbours of these nodes are distributed identically to the network as a whole
, this would have implied the entirely different result @xmath55 we will discuss a similar expression in section [ sec : model ] and the mean degree of nearest neighbours in section [ sec : degrees ] .
in fact , the inherent degree - degree correlations are such that they can not be ignored .
we can see this by looking at an example network grown to @xmath56 nodes using the simplest form ( @xmath36 ) of the saramki algorithm outlined earlier .
having generated such a network , we shall investigate the probability of arriving at a node of degree @xmath1 after a random walk of one and two steps and compare these to the preferential attachment probability .
transforming the adjacency matrix of the network generated by the algorithm into a markov - chain transition matrix @xcite ( dividing the columns by their sums ) we can look at all possible random walks from all possible starting points for a number of steps .
this is simply acheived by premultiplying a vector with all elements @xmath57 by the transition matrix .
the result is a vector containing the probabilities of arriving at each node after a random walk of one step from random initial node . multiplying this vector by the transition matrix again , provides the probabilities of arriving at each node in the network after a two - step random walk . knowing the degrees of all the nodes in the network , it is then straight - forward to look at the probability of arriving at a degree @xmath1 node after a one or two step random walk . whilst the degree distribution is described by @xmath58 which is the same as @xmath59 for selecting nodes at random ,
the ba preferential attachment probability is described in terms of individual node probabilities @xmath60 .
we use the first two parts of ( [ eqn : picalc ] ) to infer the probability of arriving at a specific node of degree @xmath1 after one or two steps .
the preferential attachment probability would correspond to an infinite random walk .
the comparison is illustrated in figure [ fig : correlation1 ] . clearly , a one - step random walk is biased towards high degree nodes in comparison to preferential attachment whilst a two step random walk is biased towards low degree nodes .
the probability of arriving at a specific node of degree @xmath1 after a random walk of one and two steps compared to the preferential attachment probability .
this analysis has been carried out on a network comprising @xmath38 nodes grown using the saramki and kaski algorithm.,scaledwidth=65.0% ] to understand why this model yields the scale - free degree distributions equivalent to preferential attachment , we must look at the expected degree of the nodes in the existing network to which the new node is to be attached . using the same transition matrix process ,
the mean degree of nodes reached after a number of steps of random walk can be calculated for a given network ( here @xmath38 nodes ) for all possible random walks from random starting positions .
the expected degree of the first node in the walk is much higher than the convergent ( preferential attachment ) value , and the second lower as shown in figure [ fig : correlation2 ] . by interpolating between these two , an approach to preferential attachment
has been achieved for this network growth algorithm , even for @xmath36 links added ( 2 step random walk ) per new node .
the expected degree of the node arrived at after a number of steps in a random walk from random initial node .
this is performed on a realisation of the saramki - kaski network , @xmath36 for @xmath38 nodes.,scaledwidth=65.0% ] we will make use of these correlations in our proposed model ( see section [ sec : model ] ) .
we now introduce the link - space formalism which is based upon the master equation method .
consider a simple , growing , non - equilibrium network in which one node is added to a network at each timestep and this node is connected to the existing network with exactly @xmath35 undirected links .
the process is governed by an attachment probability _ kernel _ @xmath61 , defined as the probability that a specific , newly - introduced link attaches to any node of degree @xmath2 within the existing network . at some time @xmath62 there exist @xmath63 nodes of degree @xmath1 and we wish to compute the expected number of nodes with degree @xmath1 at time @xmath64 .
the _ node - space master equations _ can be expressed in terms of the attachment kernels and are written @xmath65 since a new node of degree @xmath35 is added to the existing network at each timestep and there are no nodes with degree less than @xmath35 .
so far we have said nothing about the attachment mechanism , and have made the easily geneneralisable restriction that only one node is being added per timestep with undirected links .
we now follow a similar analysis , but retain the node - node linkage correlations that are inherent in many real - world systems @xcite . consider any link in a general network we can describe it by the degrees of the two nodes that it connects .
hence we can construct a matrix @xmath66 such that the element @xmath67 is equal to the number of links from nodes of degree @xmath1 to nodes of degree @xmath2 for @xmath68 at some time @xmath62 . to ease the mathematical analysis below
, the diagonal element @xmath69 is defined to be _ twice _ the number of links between nodes of degree @xmath1 for the undirected graph , a mathematical convienience also observed by bogu and pastor - satorras @xcite . for undirected networks @xmath66 is symmetric and @xmath70 , twice the total number of links @xmath71 in the network which is simply @xmath72 when introducing @xmath35 links per timestep .
an explicit example is given for the simple network discussed in section [ sec : saramaki ] which is shown in figure [ fig : simplenet ] .
this network will also be further analysed in section [ sec : model ] . the link - space matrix for this network
is then expressed as in ( [ eqn : explicit ] ) .
note the double counting of the @xmath73 link ( @xmath74 ) . @xmath75
the matrix , @xmath76 , represents a surface describing degree - degree correlations and is called the _ link - space _ matrix .
consider one of the newly introduced links in an evolving network , one end of which is attached to the new node .
the probability of selecting any node of degree @xmath77 within the existing network for the other end to attach to is given by the attachment probability @xmath78 .
suppose an @xmath77 node is selected .
the fraction of nodes of degree @xmath77 that are connected to nodes of degree @xmath2 is @xmath79 the expected increase in links from nodes of degree @xmath1 to nodes of degree @xmath2 , through the attachment of the new node to a node of degree @xmath77 , is given by @xmath80 since each link has two ends , the value @xmath81 can increase by a connection to an @xmath82-degree node which is in turn connected to a @xmath2-degree node , or by connection to an @xmath83-degree node which is in turn connected to an @xmath1-degree node .
we write master equations governing the evolution of the link - space matrix as the evolution of the expected number of links from @xmath1 to @xmath2 degree nodes , i.e. the number of @xmath84 links .
this is the _ link - space formalism _ and is written for the ( generalisable ) case of adding one new node with @xmath35 undirected links to the existing network as @xmath85 there are a variety of ways in which both the node - space and link - space master equations can be approached .
for example , a full , time - dependent solution could be investigated as in section [ subsec : er ] or , using appropriate initial conditions , the equations can be iterated over the required timescale as in section [ sec : model ] .
we can also investigate the possibility of a steady state of the algorithm under scrutiny . to do so
, we assume that there exists a steady state in which the degree distribution remains static limit but simply that the solution is static with respect to the growth algorithm . in practice
, however , a master equation approach often yields good analytic agreement with even single realisations . ] and investigate a solution . under this assumption
, the fraction of nodes @xmath86 which have a given degree remains constant such that @xmath87 when one new node is added per timestep ( @xmath88 ) .
it is also assumed that in the steady state , the attachment kernels are static too .
we drop the notation ` @xmath89 ' to indicate the steady - state and can rewrite ( [ eqn : nodexmaster ] ) as @xmath90 the fraction of links between nodes of degree @xmath1 and nodes of degree @xmath2 can be expressed as the normalised link - space matrix , @xmath91 , which sums to @xmath92 in the undirected case . in the steady state
we can assume that these values are static and can rewrite the link - space master equation ( [ eqn : lspace ] ) as @xmath93 the notation ` @xmath89 ' has again been dropped to indicate the steady - state and here the generalisable situation of adding one new node per timestep is considered .
to apply the link - space formalism one starts by specifying the model dependent attachment kernel @xmath94 . to investigate a time - dependent solution ,
the attachment kernel is substituted into ( [ eqn : nodexmaster ] ) and ( [ eqn : lspace ] ) . to investigate a steady - state solution ,
the kernel is substituted into the ( [ eqn : nsmaster ] ) and ( [ eqn : linkspacemaster ] ) to yield recurrence relations for @xmath58 and @xmath0 respectively which can be solved analytically in some cases .
the number of @xmath1-degree nodes is given by @xmath95 which allows us to retrieve the degree distribution from the normalised link - space matrix : @xmath96 degree distributions of empirically observed or simulated networks typically become dominated by noise at large degrees reflecting the small probabilities associated with these values occurring . in the link - space , the situation is exacerbated and the high @xmath97 limit reflects connections between these high degree nodes .
this is , of course , rarer than the existence of nodes of either degree .
following the conventional approach which is applied to degree distributions @xcite , we can use a cumulative representation of the link - space to address this issue . the use of a cumulative binning technique averages over stochasticity in the system . with degree distributions , regression techniques ( curve fitting ) applied to the cumulative distribution is used to obtain a more accurate description of the actual degree distribution than would be obtained from fitting to the empirical distribution itself @xcite .
the process can be similarly applied in the link - space .
surface fitting could be applied to the cumulative link - space obtained from an empirical network . from the cumulative fitted surface , a more accurate representation of the actual link - space
could be obtained for the empirical network , from which a better representation of the network s correlations could be obtained .
this process also allows for comparison between simulated and analytically - derived link - space matrices .
we define the cumulative link - space matrix , @xmath98 to be @xmath99 note that we have not lost generality in that given a cumulative linkspace matrix , the actual link - space can be derived using variations on the following : @xmath100 the computation involved when evaluating @xmath98 can be cut down considerably by first evaluating @xmath101 , which in the undirected graph is equal to @xmath92 . the leftmost column ( or top row ) for @xmath102 can then be evaluated as @xmath103 note that the second term on the right hand side is a row sum of the normalised link - space matrix and , hence , quickly evaluated .
indeed this row sum can be related to the degree distribution @xmath104 from ( [ eqn : dist ] ) .
subsequent elements can be evaluated using the following simple identity , @xmath105 we will now demonstrate the use of the link - space formalism to study a random attachment model and the barabsi - albert ( ba ) model using steady - state solutions and the classical erds and rnyi random graph using a time - dependent solution . consider first a random - attachment model in which at each timestep a new node is added to the network and connected to an existing node with uniform probability without any preference ( ` model a ' in @xcite ) with one undirected link such that @xmath106 and @xmath107 .
we assume a steady - state solution and the attachment kernel is @xmath108 . substituting into ( [ eqn : nsmaster ] )
, we obtain the recurrence relation @xmath109 , which yields the familiar degree distribution @xmath110 .
substituting into ( [ eqn : linkspacemaster ] ) yields the recurrence relation @xmath111 with @xmath112 .
it is easy to populate the link - space matrix numerically , just from the recurrence equations ( [ eqn : ramaster2 ] ) and the degree distribution . at first glance ,
the solution to the master equation ( [ eqn : ramaster2 ] ) would be of the form @xmath113 however , the boundary conditions ( which could be interpreted as influx of probability into the diffusive matrix ) are such that this does nt hold . the paths of probability flux from @xmath114 influencing element @xmath0 .
each arrow ( step ) represents a further factor of @xmath115.,scaledwidth=60.0% ] we can actually solve the link - space for this model exactly . consider the values @xmath58 as being influxes of probability into the top and left of the link - space matrix .
we can compute the effects of such influx on an element in the normalised link - space matrix , @xmath0 .
each step in the path of probability flux reflects an extra factor of @xmath115 .
first , we will consider the influx effect from the top of the matrix as in figure [ fig : radiag ] .
the total path length from influx @xmath114 to element @xmath0 is simply @xmath116 .
the number of possible paths between @xmath114 and element @xmath0 can be expressed @xmath117 using the conventional binomial combinatorial or ` choose ' coefficient .
we can similarly write down the paths and lengths for influxes into the left hand side of the matrix . the first row ( and column )
can be described as @xmath118 subsequent rows can be similarly described .
so , for @xmath119 an element can be written @xmath120 this expression is compared to a numerically populated link - space matrix in figure [ fig : raanalvexact ] .
comparison of the numerically derived normalised link - space matrix and the analytic solution for random attachment whereby one new node is added with one undirected link per timestep .
the numerical implementation populates the link - space matrix directly from the degree distribution and the link - space recurrence relations .
the first twenty rows ( @xmath121 ) of the link - space matrices are illustrated.,scaledwidth=70.0% ] we shall make use of this solution to investigate the correlations of such a network in section [ sec : degrees ] .
this normalised link - space matrix is illustrated in fig .
[ fig : ra ] along with a comparison to simulated networks using the cumulative link - space matrix . left : the analytically - derived , normalised link - space matrix for the random attachment growth algorithm .
this could be filled to arbitrary size but is truncated to maximum degree of @xmath122 here .
right : comparison of the cumulative link - space matrices for the analytic solution and a simulation of the algorithm .
the simulation comprises an ensemble average of @xmath123 networks grown to @xmath124 nodes .
the maximum degree obtained was @xmath125 .
the first @xmath126 rows ( @xmath1 values ) of the cumulative link - space are illustrated and finite - size effects are noticable at high @xmath2 .
, title="fig:",scaledwidth=45.0% ] left : the analytically - derived , normalised link - space matrix for the random attachment growth algorithm .
this could be filled to arbitrary size but is truncated to maximum degree of @xmath122 here .
right : comparison of the cumulative link - space matrices for the analytic solution and a simulation of the algorithm .
the simulation comprises an ensemble average of @xmath123 networks grown to @xmath124 nodes .
the maximum degree obtained was @xmath125 .
the first @xmath126 rows ( @xmath1 values ) of the cumulative link - space are illustrated and finite - size effects are noticable at high @xmath2 . , title="fig:",scaledwidth=45.0% ] in the ba model @xcite , at each timestep a new node is added to the network and connected to @xmath35 existing nodes with probabilities proportional to the degrees of those nodes , i.e. @xmath127 yielding @xmath128 we consider the scenario when the new node is added with one undirected link , @xmath106 , and the attachment kernel is well approximated by @xmath129 . substituting this into ( [ eqn : nsmaster ] )
yields the recurrence relation @xmath130 , whose solution is @xmath131 .
this corresponds well to an actual network grown using this algorithm . using the same substitution ,
the link - space master equations yield the recurrence relations : @xmath132 again , it is easy to populate the matrix numerically just by implementing the known degree distribution and the link - space recurrence equations but we can solve the link - space analytically . at first glance , the solution to the master equation ( [ eqn : sfmaster3 ] ) would be of the form @xmath133 where @xmath134 is a constant .
this would imply for the degree distribution @xmath135 summing over the entire node - space would give @xmath136 . however , the boundary conditions ( which could be interpreted as influx of probability into the diffusive matrix ) are such that this solution does nt hold .
we can obtain an exact ( although somewhat less pretty ) solution by tracing fluxes of probability around the matrix and making use of the previously derived degree distribution .
the steady - state link - space master equation for general attachment kernel is written @xmath137 we can rewrite ( [ eqn : linkspacemastersf ] ) in terms of vertical and horizontal components : @xmath138 where @xmath139 the components of flux of probability around the normalised link - space matrix .
this is general to any attachment kernel .
an element within the link - space matrix can be built up from contributing elements and the appropriate factors.,scaledwidth=80.0% ] by considering the probability fluxes as shown in figure [ fig : sfexact ] , we can write the individual elements in the link - space matrix as @xmath140 note that we have yet to introduce the attachment probability kernels and the analysis so far is general . using the preferential attachment probability , @xmath141 we can write our component - wise factors for the master equation @xmath142 substituting ( [ eqn : components ] ) into ( [ eqn : elements ] ) and using the previously derived degree distribution yields for the first row @xmath143 subsequent rows can be described : @xmath144 rewriting ( [ eqn : working1 ] ) gives @xmath145 where the meaning of @xmath146 and @xmath147 is transparent .
clearly , we can write ( [ eqn : working2 ] ) for @xmath148 , @xmath149 substituting ( [ eqn : working3 ] ) into ( [ eqn : working1 ] ) yields @xmath150 in order to solve this recurrence relation we define an operator for repeated summation , @xmath151 such that @xmath152 the subscripts denote the initial variable to be summed over and the final limit .
a few examples of this operation clarify its use : @xmath153 we can use this operator in our expression for the element @xmath0 in ( [ eqn : working4 ] ) and expand to the value @xmath154 , @xmath155 we can express @xmath154 in terms of the operator @xmath156 too , @xmath157 the element @xmath0 can be expressed as @xmath158 this form is somewhat obscure as the calculation of the operator values is less than obvious .
however , we can transform to a more easily interpreted operator @xmath159 analogous to @xmath156 but with different limits ( which enables further analysis ) such that @xmath160 whilst , at first glance , it looks like little progress has been made , for this analysis we only need evaluate the repeated operation on initial function @xmath161 .
this is exactly solvable and as such , we can rewrite in terms of a function @xmath162 dropping the superfluous @xmath163 subscripts , @xmath164 the inductive proof associated with ( [ eqn : inductive ] ) can be understood by path counting for some repeating binomial process and is an intrinsic property of the combinatorial choose coefficient .
consider a repeated coin toss over @xmath163 steps .
the number of ways of achieving @xmath165 successes after these @xmath163 iterations is @xmath166 .
now , the occurrence of this last success could have happened on the @xmath167th iteration or the following one , or any of the subsequent iterations till the @xmath163th one .
for the last successful outcome to occur on the @xmath35th step , @xmath17 successful outcomes must have occurred in the previous steps .
the number of ways this could have occurred is @xmath168 .
clearly , summing over all possible @xmath35 values , the total possible paths resulting in @xmath165 successes must equate to @xmath166 . for clarity
, this is depicted in figure [ fig : choosediag ] . to reach point @xmath169 from @xmath170 in the binomial process ,
one of the steps @xmath171 or @xmath172 must be traversed , after which there is only one route to @xmath169 .
the number of paths between @xmath170 and @xmath169 utilising step @xmath134 is the same as the number of paths between @xmath170 and @xmath173 .
similarly , the number of paths between @xmath170 and @xmath169 utilising step @xmath163 is the same as the number of paths between @xmath170 and @xmath174 and so on .
the number of paths between @xmath170 and @xmath169 can be expressed as the sum of the paths @xmath175 , @xmath176 .
@xmath177 and @xmath178 . a binomial process over seven steps .
the number of paths between @xmath170 and @xmath169 can be expressed as the sum of the paths @xmath175 , @xmath176 , @xmath177 and @xmath178 .,scaledwidth=70.0% ] we incorporate this behaviour into our proof for the solution of @xmath179 , @xmath180 as @xmath181 , this inductive proof holds for all @xmath17 and the following relations hold @xmath182 we can now write the element in our link - space matrix for preferential attachment exactly as our function @xmath162 is easily evaluated , @xmath183 and the first few rows of our link - space can be written @xmath184 a comparison of the numerically populated link - space matrix and the analytic solution for the ba preferential attachment model with @xmath106 is illustrated in figure [ fig : sfanalvexact ] .
comparison of the numerically derived normalised link - space matrix and the analytic solution for the steady state for preferential attachment whereby one new node is added with one new link .
the first twenty rows ( @xmath121 ) of the link - space matrices are illustrated.,scaledwidth=70.0% ] we shall use this solution to investigate the correlations with this network in section [ sec : degrees ] .
this normalised link - space matrix is illustrated in fig .
[ fig : ba ] along with a comparison to simulated networks . left : the analytically - derived , normalised link - space matrix for the preferential attachment growth algorithm .
this could be filled to arbitrary size but is here truncated to maximum degree of approximately @xmath38 .
right : comparison of the cumulative link - space matrices for the analytic solution and a simulation of the algorithm .
the simulation comprises an ensemble average of @xmath185 networks each grown to @xmath186 nodes .
the maximum degree obtained was @xmath187 .
various rows ( @xmath1 values ) from @xmath188 to @xmath189 of the cumulative link - space are illustrated and the effects of the finite nature of the simulation are apparent . , title="fig:",scaledwidth=45.0% ] left : the analytically - derived , normalised link - space matrix for the preferential attachment growth algorithm .
this could be filled to arbitrary size but is here truncated to maximum degree of approximately @xmath38 .
right : comparison of the cumulative link - space matrices for the analytic solution and a simulation of the algorithm .
the simulation comprises an ensemble average of @xmath185 networks each grown to @xmath186 nodes .
the maximum degree obtained was @xmath187 .
various rows ( @xmath1 values ) from @xmath188 to @xmath189 of the cumulative link - space are illustrated and the effects of the finite nature of the simulation are apparent . , title="fig:",scaledwidth=45.0% ] the erds and rnyi ( er ) classical random graph is the quintessential equilibrium network model @xcite .
however , to employ the link - space formalism , which tracks the evolution of a network s correlation properties , we must model it as an evolving , non - equilibrium network .
the classical random graph can be modelled as a constantly accelerating network with random attachment . at each timestep
, we add one new node to the existing network .
all possible links between the new node and _ all _ existing nodes are considered and each is established with probability @xmath190 .
that is , a biased coin toss ( bernoulli trial ) is employed for every node within the existing network to decide whether a link is formed between it and the new node .
conseqently the expected number of new , undirected links with which the new node connects to the existing network is @xmath191 and this would be described as an accelerating network @xcite .
no new links are formed between existing nodes .
this model subsequently produces a network in which the probability of a link existing between any pair of nodes is simply @xmath190 and is thus representative of the er random graph @xcite .
a network grown to @xmath62 nodes will have mean degree @xmath192 which will also be the expected degree of all nodes in the network . whilst the random attachment and preferential attachment models both have steady - state assymptotic behaviour ,
clearly this model does nt and a full , time - dependent solution is required . using the random attachment probability kernel @xmath193 , the node - space master equation for the number of nodes of degree @xmath1
can be written @xmath194 the time - dependent solution of the degree distribution for this model is simply @xmath195 which is what we would expect for the random graph with mean degree of @xmath196 in the large size limit @xcite .
we can similarly write the link - space master equation for this model : @xmath197 the last two terms reflect the expected number of new @xmath198 links being formed between the new node and existing nodes . recalling that the normalised link - space matrix is given by @xmath91 where the number of links , @xmath71 , at time @xmath62 can be approximated as @xmath199 and making use of the previously derived , time - dependent degree distribution such that @xmath200 , ( [ eqn : lsacc ] ) can be rewritten as @xmath201 the solution to ( [ eqn : lsacc2 ] ) is found by means of ansatz (
the choice of which is explained in section [ sec : degrees ] ) and is @xmath202 this is the normalised link - space matrix for a classical random graph with a mean degree of @xmath196 in the large size limit and this is illustrated in fig [ fig : er ] where a comparison to a simulation is also made
. left : the analytically - derived , normalised link - space matrix for the erds and rnyi ( er ) random graph with mean degree equal to @xmath7 .
this could be filled to arbitrary size but is here truncated to maximum degree of @xmath126 .
right : comparison of the cumulative link - space matrices for the analytic solution and a simulation of the algorithm .
the simulation comprises an ensemble average of @xmath203 networks each consisting of @xmath186 nodes .
the probability that a link exists between any pair of nodes is @xmath204 and the maximum degree obtained was @xmath205 .
the first @xmath206 rows ( @xmath1 values ) of the cumulative link - space are illustrated and , again , the finite nature of the simulation is apparent for high @xmath2.,title="fig:",scaledwidth=45.0% ] left : the analytically - derived , normalised link - space matrix for the erds and rnyi ( er ) random graph with mean degree equal to @xmath7 .
this could be filled to arbitrary size but is here truncated to maximum degree of @xmath126 .
right : comparison of the cumulative link - space matrices for the analytic solution and a simulation of the algorithm .
the simulation comprises an ensemble average of @xmath203 networks each consisting of @xmath186 nodes .
the probability that a link exists between any pair of nodes is @xmath204 and the maximum degree obtained was @xmath205 .
the first @xmath206 rows ( @xmath1 values ) of the cumulative link - space are illustrated and , again , the finite nature of the simulation is apparent for high @xmath2.,title="fig:",scaledwidth=45.0% ]
as discussed in section [ sec : correlations ] , calculation of the mean degree of nearest neighbours of a node as a function of the degree of that node , @xmath207 presents an indication as to the degree correlations of the network .
this calculation is straightforward if one has the link - space matrix for a particular network , @xmath208 the mean degree of the nearest neighbours @xmath207 as a function of @xmath1 for both the preferential and random attachment algorithms and the calssical random graph described in section [ subsection : nonassort].,scaledwidth=70.0% ] this is illustrated in figure [ fig : assort2 ] . if the average nearest neighbour degree @xmath207 is constant , non - assortativity is implied .
certainly , the preferential attachment curve in figure [ fig : assort2 ] appears to asymptote to a constant value ( as was noted in @xcite ) , whilst that of random attachment continues to increase , implying positively assortative mixing . in search of a single number representation , newman @xcite
further streamlined the measure by simply considering the correlation coefficient of the degrees of the nodes at each end of the links in a given network , although he subtracted one from each first to represent the remaining degree . whilst this might be of interest in some situations ,
the correlation of the actual degrees is arguably a more general measure although the two do not differ much in practice . using a normalised pearson s @xcite correlation coefficient , the measure of assortativity
is expressed in terms of the degrees ( @xmath1 and @xmath2 ) of the nodes at each end of a link , and the averaging is performed over all links in the network . @xmath209 the normalisation factor represents the correlation of a perfectly assortative network , i.e. one in which nodes of degree @xmath1 are only connected to nodes of degree @xmath1 .
this would be of the same degree distribution as the network under consideration and has the same number of nodes and links .
the physical interpretation is less than obvious especially as we are considering trees .
however , for this assortative network , @xmath210 , i.e. all off - diagonal elements are zero . whilst this assortativity measure can be expressed as summations over all links if calculated for a specific realisation of a network , we can express it in terms of the previously derived link - space correlation matrix , namely @xmath211 it is interesting that the normalisation factor is degree - distribution specific . evaluating this measure numerically ,
we obtain for the preferential attachment algorithm the value @xmath212 and for the random attachment algorithm , @xmath213 . so according to this measure of degree mixing ,
the randomly grown graph is highly assortative and the preferential attachment graph is not .
the link - space formalism allows us to address two - vertex correlations in a more powerful way in that we can calculate the conditional , two - point vertex degree distribution , @xmath214 which has been traditionally difficult to measure @xcite .
we can write this joint probability , i.e. the probability that a randomly chosen edge is connected to a node of degree @xmath2 given that the other end is connected to a node of degree @xmath1 , in terms of the link - space matrix ( normalised or not ) as @xmath215 . for total number of edges aproximately equal to the total number of nodes ,
this can be approximated as @xmath216 .
this is illustrated in fig .
[ fig : assort1 ] for the scenarios of random and preferential attachment .
consider selecting an edge at random in the network .
if we select one end of the edge at random , the probability that this node has degree @xmath2 will be proportional to @xmath2 since higher degree nodes have , by definition , more links connected to them than low degree nodes .
consider now only a subset of all edges with one end attached to a node of degree @xmath1 . if there were no correlations present , the probability that the other end is attached a node of degree @xmath2 is again proportional to @xmath2 .
this is the criterion of non - assortativity @xcite and for total number of edges @xmath217 total number of nodes can be expressed @xmath218 this is also the criterion that a one step random walk replicates preferential attachment . the conditional probability @xmath219 that one end of a randomly selected link has degree @xmath2 given that the other end is of degree @xmath1 over the range @xmath220 .
the left plot is for the random attachment algorithm and the right is preferential attachment .
the criterion of non - assortativity of ( [ eqn : assort1 ] ) is denoted with red circles.,title="fig:",scaledwidth=45.0% ] the conditional probability @xmath219 that one end of a randomly selected link has degree @xmath2 given that the other end is of degree @xmath1 over the range @xmath220 .
the left plot is for the random attachment algorithm and the right is preferential attachment .
the criterion of non - assortativity of ( [ eqn : assort1 ] ) is denoted with red circles.,title="fig:",scaledwidth=45.0% ] it is interesting to ask whether a perfectly non - assortative network can be generated .
assuming that such a network will not have equal numbers of nodes and links , we must rewrite the conditions of non - assortativity accordingly for total number of edges @xmath221 : @xmath222 recalling our definition of the normalised link - space matrix that @xmath223 , we can express this conditional probability @xmath214 in terms of this matrix as @xmath224 as such , we can now write for @xmath0 @xmath225 that is , for _ any _ degree distribution , a normalised link - space matrix can be found which is representative of a perfectly non - assortative network .
can be simply obtained through the addition of fully connected components of @xmath226 nodes .
similarly an off - diagonal element @xmath0 can be generated through suitable rewiring of @xmath1-degree and @xmath2-degree cliques .
] this might provide an alternative to the network randomisation technique of maslov and sneppen to provide a
null model network " @xcite .
the expression of ( [ eqn : assort5 ] ) provides the ansatz solution for the normalised link - space matrix to the master equations for the er random graph in section [ subsec : er ] .
although somewhat counter - intuitive , it is possible to find steady states of networks whereby nodes and/or links are removed from the system . aside from the obvious situation of having no nodes or edges left , we would like to investigate the possible existence of a network configuration whose link - space matrix and , subsequently , degree distribution are static with respect to the decay process .
the concept that decay processes are highly influential on a network s structure has been considered before although this has typically only been investigated in conjunction with simultaneous growth @xcite . here
we shall employ the link - space formalism to examine the effect of some simple , decay - only scenarios , specifically the two simplest cases random link removal and random node removal .
consider an arbitrary network .
we select @xmath134 links at random and remove them .
we shall implement the link - space to analyse the evolution of the network .
consider the link - space element @xmath67 denoting the number of links from nodes of degree @xmath1 to nodes of degree @xmath2 .
clearly this can be decreased if an @xmath198 link is removed . also , if a @xmath227 link is removed and that @xmath1 node has further links to @xmath2 degree nodes , then those that were @xmath198 links will now be @xmath228 , similarly for @xmath229 links being removed .
however , if the link removed is a @xmath230 link and that @xmath231 degree node is connected to a @xmath1 degree node , then when the @xmath231 node becomes an @xmath2 node , that link will become an @xmath198 link increasing @xmath81 as shown in figure [ fig : decay1 ] .
let us assume that we are removing links at random from the network comprising @xmath232 vertices and @xmath71 links .
a non - random link selection process could be incorporated into the master equations using a probability kernel .
the master equation for this process can be written in terms of the expected increasing and decreasing contributions : @xmath233 this simplifies to @xmath234 this expression can be simply understood by considering figure [ fig : decay1 ] . for each @xmath235 link
( labelled @xmath39 in the figure ) , there are @xmath2 links ( @xmath236 ) whose removal would make link @xmath39 an @xmath198 link
. this would be the same for all of the @xmath235 links , of which there are @xmath237 in the network . the expected increase in @xmath81 through this process corresponds to the second term on the right hand side of ( [ eqn : rlr2 ] ) .
the other terms are equally simply obtained , with the last referring to the physical removal of an @xmath198 link .
removal of any of the links @xmath236 will result in the link @xmath39 becoming an @xmath198 link.,scaledwidth=70.0% ] to check that all is well , we can explicitly look at the evolution of the link - space matrix for the simple example network discussed in section [ sec : saramaki ] and shown again in figure [ fig : simplenet4 ] .
a simple network which we shall examine to check the validity of the master equation ( [ eqn : rlr2 ] ) for the evolution of the link - space matrix for the random link removal process.,scaledwidth=65.0% ] prior to removing any links , the link - space can be simply written as in ( [ eqn : simplel ] ) . @xmath238 the expected link - space matrix after removing a link at random
can be interpreted as the ensemble average of all possibilities . by removing a single link from the original network , the resulting link - space matrix for the new network
can be easily generated . replacing that link and performing this process for all links and simply averaging all the resulting matrices
, we can generate ( albeit somewhat laboriously ) the desired expected link - space matrix for the network at the next timestep .
this yields the result of ( [ eqn : simplelt1 ] ) which represents the average over all possible resulting networks after one link has been removed at random .
this concurs with the master equation for this process given in ( [ eqn : rlr2 ] ) .
@xmath239 to investigate the possibility of a steady state solution , we make a similar argument as before for growing networks but this time for the process of removing @xmath134 links per timestep . @xmath240
the expression of ( [ eqn : rlr2 ] ) can be reduced to @xmath241 the number of @xmath242 links does not reach a steady state .
this might be expected , as the removal process for such links requires them to be physically removed as opposed to the process by which the degree of the node at one end of the link is reduced .
subsequently , the value @xmath243 would increase in time .
however , we can investigate the properties of the rest of the links in the network which do reach a steady state by neglecting these two node components . in a similar manner to @xcite we can make use of a substitution to find a solution to this recurrence equation , namely for @xmath244 .
@xmath245 and we obtain @xmath246 this has a trivial solution @xmath247 the link - space for this system can then be written for @xmath244 @xmath248 the shape of the steady state link - space for random link removal from an infinite network numerically populated using ( [ eqn : rlrfull ] ) for @xmath244.,scaledwidth=70.0% ] the summation over this link - space matrix does not converge , i.e. @xmath249 . however , we can still derive the degree distribution and normalise this such that @xmath250 .
we can use the form of ( [ eqn : rlrfull ] ) to infer the shape of the degree distribution . recalling that @xmath251 we can write for the degree distribution , neglecting the two node components in the network , @xmath252 @xmath253 the equation ( [ eqn : rlrdegdist1 ] )
can be easily solved by considering the maclaurin series ( taylor series about zero ) of the function @xmath254 ( as studied by mercator as early as 1668 ) and evaluating for @xmath255 , @xmath256 rearranging ( [ eqn : rlrdegdist1 ] ) and letting @xmath257 , @xmath258 for nodes with degree greater than one , the solution of ( [ eqn : rlrdegdist2 ] ) requires a simple proof by induction .
consider the function @xmath259 defined for positive integer @xmath260 as @xmath261 where @xmath262 denotes the conventional binomial coefficients .
we can write @xmath263 with similar ease : @xmath264 as such , @xmath265 a quick substitution of @xmath266 leads to @xmath267 rearranging ( [ eqn : rlrdegdist2 ] ) and using some simple substitutions , @xmath268 and @xmath269 we can derive the following @xmath270 we can normalise this degree distribution for the network without the two node components ( the @xmath242 links ) : @xmath271 although difficult to compare to simulations , we can populate a link - space matrix numerically using ( [ eqn : rlrfull ] ) as illustrated in figure [ fig : rlr ] . with this
, we can compare the degree distributions using ( [ eqn : ci3 ] ) and those of ( [ eqn : analrlr ] ) as shown in figure [ fig : rlrdegs ] .
comparison of the degree distributions from the numerically populated link - space matrix and the analytic results of ( [ eqn : analrlr ] ) for the random link removal model .
the two node components have been neglected in both the numerical matrix population and analytically derived expression.,scaledwidth=70.0% ] in a similar manner to section [ subsec : rlr ] we will now discuss the possibility of creating such a steady state via a process of removing nodes ( and all of their links ) from an existing network at the rate of @xmath134 nodes per timestep . clearly , removing some node which has a link to an @xmath272-degree node which in turn has a link to a @xmath2-degree node can increase the number of @xmath198 links in the system .
the other processes which can increase or decrease the number of links from @xmath1 degree nodes to @xmath2 degree nodes can be similarly easily explained .
we consider that we select a node of some degree @xmath14 for removal with some probability kernel @xmath273 ( as in the growing algorithms of section [ sec : linkspace ] ) .
the master equation for such a process can thus be written for general node selection kernel : @xmath274 we now select a kernel for selecting the node to be removed , namely the random kernel although the approach is general to any node selection procedure .
for this process @xmath275 as a node is selected purely at random .
the steady state assumptions must be clarified slightly .
as before , @xmath276 however , because we can remove more than one link ( through removing a high degree node for example ) we must approximate for @xmath71 . using the random removal kernel , we can assume that on average , the selected node will have degree @xmath277 and we can use this to calculate the number of remaining links in the network .
@xmath278 making use of the link - space identities and ( [ eqn : rnrsteady ] )
, the master equation ( [ eqn : noderemovemaster ] ) can be written in simple form : @xmath279 clearly , this is identical to ( [ eqn : lsrlr ] ) for the random link removal model of section [ subsec : rlr ] and consequently , the analysis of the degree distribution will be the same too . by
considering the average degree of the neighbours of nodes of degree @xmath1 , denoted @xmath207 and making use of ( [ eqn : karnn ] ) in section [ sec : degrees ] we can see that both the random link removal and the random node removal algorithms generate highly assortative networks as evidenced in figure [ fig : kbarnnrnr ] . the mean degree of the neighbours of nodes of degree @xmath1 for the random link and random node removal algorithms.,scaledwidth=70.0% ]
in @xcite , krapivsky and redner suggested a model of network growth without prior global knowledge using directed links . within the same paper , they also provide analysis for the evolution of the joint degree distribution of simpler , directed growing networks ( @xmath280 ) , which is similar to the link - space analysis outlined above .
the _ gnr _
algorithm is implemented as follows : 1 .
pick a node @xmath281 within the existing network at random .
2 . with probability @xmath282
make a directed link pointing to that node or 3 .
else , pick the ancestor node of @xmath281 ( the node which its out - link is directed to ) and make a directed link pointing to that node . the growing network with redirection . a random initial node is selected .
the new node links to ( and points at ) this node with probability @xmath282 or else is redirected to and attaches to the ancestor of the node.,scaledwidth=65.0% ] we note that because each new node is attached to the existing network with only one link , each node has only one out link .
when redirection occurs , the destination is predetermined .
the degree distribution for this network is easily obtained .
consider some node of total ( in and out ) degree @xmath1 labeled @xmath39 .
its out - degree is @xmath188 so its in - degree is simply @xmath77 .
thus , it must have @xmath77 ` daughters ' .
if any of these @xmath77 daughter nodes is selected initially in the _ gnr _ algorithm _ and _ redirection occurs , then the new node will link to and point at @xmath39 .
it is straightforward to construct the evolution of the node - space for the total degree of nodes within the system .
the attachment probability kernel for the new node linking to any node of degree @xmath1 within the existing network can be written as @xmath283 the first term on the right hand side corresponds to initially selecting a node of degree @xmath77 and linking to it .
the second term reflects redirection from any of the daughters of any of the degree @xmath1 nodes .
the static equation for the steady state degree distribution becomes @xmath284 for @xmath285 , the familiar recurrence relation of preferential attachment is retrieved and the distribution of total degree of nodes will subsequently be the same too .
in this section we introduce a model that makes use of only local information about node degrees as microscopic mechanisms requiring global information are often unrealistic for many real - world networks @xcite .
it therefore provides insight into possible alternative microscopic mechanisms for a range of biological and social networks .
the link - space formalism allows us to identify the transition point at lower power - law exponent degree distributions switch to higher exponent power - law distributions with respect to the ba preferential attachment model .
while similar local algorithms have been proposed in the literature @xcite , the strength of the approach followed here is the ability to describe the inherent degree - degree correlations @xcite .
it is well known that a mixture of random and preferential attachment in a growth algorithm can produce power - law degree distributions with exponents greater than three , @xmath286 .
it has often been assumed that a _ one step _
random walk replicates linear preferential attachment @xcite .
this is _ not _ true .
a one step random walk is in fact more biased towards high degree nodes than preferential attachment ( as discussed in section [ sec : saramaki ] ) as can be easily seen by performing the procedure on a simple hub and spoke network . in this case
, the probability of arriving at the hub tends to one for increasingly large networks rather than a half as would be appropriate for preferential attachment .
we can use this bias to generate networks with degree distributions that are overskewed ( lower power - law exponent ) than the preferential attachment model . a mixture of this approach with random attachment results in a simple model that can span a wide range of degree distributions .
this one - parameter network growth model , in which we simply attach a single node at each timestep , does not require prior knowledge of the existing network structure .
explicitly , the external nearest neighbour algorithm proceeds as follows : 1 .
pick a node @xmath281 within the existing network at random .
2 . with probability @xmath282
make a link to that node or 3 .
else , pick any of the neighbours of @xmath281 at random and link to that node .
hence this algorithm resembles an object or ` agent ' making a short random - walk .
this is very similar to the growing network with redirection model discussed in section [ sec : redner ] except that the model here employs undirected links which necessitates a random ( as opposed to deterministic ) walk .
clearly , this algorithm generates trees , some examples of which are illustrated in figure [ fig : exnn ] .
external nearest neighbour : some examples of the graphs generated for 300 nodes . left is @xmath287 , center is @xmath288 and right is @xmath289 .
there is no spatial relevance in these plots which are drawn with pajek @xcite .
, title="fig:",scaledwidth=30.0% ] external nearest neighbour : some examples of the graphs generated for 300 nodes . left is @xmath287 , center is @xmath288 and right is @xmath289 .
there is no spatial relevance in these plots which are drawn with pajek @xcite .
, title="fig:",scaledwidth=30.0% ] external nearest neighbour : some examples of the graphs generated for 300 nodes . left is @xmath287 , center is @xmath288 and right is @xmath289 .
there is no spatial relevance in these plots which are drawn with pajek @xcite .
, title="fig:",scaledwidth=30.0% ] the ( cumulative ) degree distributions for the evolving networks grown to @xmath38 nodes , ensemble averaged over 100 networks per parameter value .
these have been initiated from a 2 node , 1 link component.,scaledwidth=95.0% ] the resulting distributions of vertex degree from implementing this growth algorithm are shown in figure [ fig : results ] .
these are ensemble averages over 100 networks per parameter value .
interestingly , @xmath289 yields a network that is dominated by hubs while @xmath290 yields the random attachment network .
intermediate values of @xmath282 yield networks which are neither too ordered nor too disordered . for @xmath291 ,
the algorithm generates networks whose degree - distribution closely resembles the ba preferential - attachment network .
our analysis of the algorithm starts by establishing the attachment probability @xmath94 , which in turn requires properly resolving the one - step random walk .
the link - space formalism provides us an expression for the probability @xmath292 associated with performing a random walk of length one and arriving at any node of degree @xmath1 .
@xmath293 we can reconcile this with the alternative expression @xmath294 where the average is performed over the neighbours of nodes of degree @xmath1 . with proper normalisation , @xmath295 which clearly yields the same results .
note that this quantity does _ not _ replicate preferential attachment , in contrast to what is commonly thought @xcite .
we shall perform this process explicitly for the example ( @xmath296 , @xmath297 ) network of figure [ fig : simplenet4 ] with the link - space matrix @xmath298 let us consider the probability of arriving at any node of degree @xmath188 by performing a random walk of length one from a randomly chosen initial vertex , @xmath299 this is clearly consistent with the analysis of ( [ eqn : neighbour ] ) . + we can now begin to formulate the master equations for this model . defining @xmath300
as @xmath301 the attachment kernel for the external nearest neighbour algorithm becomes @xmath302 substituting ( [ eqn : xnnkernel ] ) into ( [ eqn : nsmaster ] ) we obtain for the steady state node degree @xmath303 substituting into ( [ eqn : linkspacemaster ] ) we get @xmath304 the non - linear terms resulting from @xmath305 mean that a complete analytical solution for @xmath0 is difficult . we leave this as a future challenge .
however , the formalism can be implemented in its non - stationary form numerically ( iteratively ) with reasonable efficiency . as described in section [ sec : linkspace ] , the master equations for the network growth in the link - space are given for @xmath306 as @xmath307 and for @xmath308 as @xmath309 computationally , it is useful to rewrite these equations ( [ eqn : lspace3 ] ) and ( [ eqn : lspace4 ] ) in terms of matrix operations .
we define the following matrices and operators : @xmath310 @xmath311 @xmath312 armed with these useful building blocks , we can represent the link - space master equations , ( [ eqn : lspace ] ) as @xmath313 note that we have not lost generality in specifying the attachment kernel @xmath314 . although ( [ eqn : matrix ] ) looks somewhat intractable , the terms can be easily explained .
premultiplying by @xmath315 has the effect of moving all the elements down one place .
postmultiplying by @xmath316 moves the elements right .
the last two terms represent the @xmath317 term in ( [ eqn : lspace4 ] ) .
we can now tailor for the three systems .
for random attachment we have @xmath318 for preferential attachment we have @xmath319 things start to get a little trickier for the external nearest neighbour algorithm yielding @xmath320 recall from ( [ eqn : beta ] ) that @xmath321 actual computation is somewhat more straight - forward than the mathematics suggests and is based upon the assumption that we can interchange @xmath322 with @xmath323 to enable a recursive process for the evolution of the link - space matrix . whilst this would not be strictly apppropriate to model the evolution of a single realisation of the algorithm or even an ensemble average ( each realisation would have a path dependent evolution ) , if a steady state is reached ( albeit approximately ) then this will be a steady state of the algorithm .
this is a similar assumption to that made in the node - space master equation approach to non - equilibrium networks in section [ sec : linkspace ] .
we seed with @xmath324 reflecting two connected nodes ( such that @xmath325 and total links , @xmath326 ) . to account for the finite size of the matrices involved , the link - space is normalised to @xmath327 at each iteration although in practice , the very edges of the matrix where numbers might fall off have very small values .
we retrieve the degree distribution @xmath328 from the link - space matrix , @xmath329 comparison of the matrices @xmath330 and @xmath66 or the vectors @xmath331 and @xmath332 gives an idea of the proximity to the steady state .
the results can be seen in figure [ fig : lsresults ] and appear ( qualitatively ) comparable to the numerical numerical simulations of figure [ fig : results ] .
degree distributions generated using the link - space analysis of our one - step algorithm .
dashed line : results for the barabsi - albert ( ba ) @xcite preferential - attachment algorithm .
our one - step algorithm ressembles the ba results with parameter value @xmath333 ( see inset).,scaledwidth=70.0% ] we can now use the link - space formalism to deduce the parameter value at which our algorithm approximates the ba degree distribution . at this value @xmath334 , the node - degree distribution goes from higher to lower power - law exponent in relation to the ba model . for this parameter value ,
the attachment probability to nodes of various degrees is approximately equal for both our mixture algorithm and the ba model .
recall from section [ subsec : ba ] and ( [ eqn : xnnkernel ] ) @xmath335 for large @xmath1 we have @xmath336 we can use the exact solution of the link - space for the preferential attachment algorithm ( or read off the graph in figure [ fig : betasf ] ) to infer @xmath300 in the high @xmath1 limit .
using the first two terms of ( [ eqn : sfexact ] ) we obtain @xmath337 ( the graph asymptotes to @xmath338 ) and substituting into ( [ eqn : a ] ) gives @xmath339 .
it is important to note that @xmath300 is a function of @xmath1 ( as is evident in figure [ fig : betasf ] ) which in turn is non - trivially dependent on @xmath340 .
consequently , it is unlikely for ( [ eqn : criticalc ] ) to hold for all values of @xmath1 . whilst the mixture model approximates the preferential attachment distribution at this value ,
the actual distributions are not identical . the value of @xmath300 for the preferential attachment algorithm.,scaledwidth=70.0% ]
in conclusion , we have developed a new formalism which accurately accounts for the node - node linkage correlations in networks .
we have employed the formalism to produce analytic solutions to the link - space correlation matrix for the random attachment and ba models of network growth as well as the classical random graph .
we have also derived the form of a perfectly non - assortative network for arbitrary degree distribution and illustrated the possiblility of a steady - state degree distribution and link - space for decaying networks .
we have used the formalism to dispel the myth that a one - step random walk from random initial position in an arbitrary network is the same as selecting a node with probability proportional to its degree . by definition , only a perfectly non - assortative network , as derived in section [ subsection : nonassort ] , would satisfy this condition .
the link - space formalism has allowed us to describe a simple one - parameter network growth algorithm which is able to reproduce a wide variety of degree distributions without any global information about node degrees .
there are clearly many alternative yet similar algorithms which could be designed which might also be analysed with the link - space formalism
. these might provide insight into certain real world networks which not only exhibit similar scaling behaviour but might also have similar constraints in their microscopic growth rules .
erds , e. , and rnyi , a. , on random graphs " , publicationes mathematicae debrecen * 6 * , 290297 ( 1959 ) .
eguluz , v.m . and klemm , k. , `` epidemic threshold in structured scale - free networks '' , phys .
lett . * 89 * , 108701 ( 2002 ) .
szab , g. , alava , m. and kertsz , j. , structural transitions in scale - free networks " , phys .
e * 67 * 056102 ( 2003 ) .
toivonen , r. , onnela , j .-
p . , saramki , j. , hyvnen , j. and kaski , k. , a model for social networks " , physica a * 371 * , 851860 ( 2006 ) . | many networks contain correlations and often conventional analysis is incapable of incorporating this often essential feature . in arxiv:0708.2176
, we introduced the link - space formalism for analysing degree - degree correlations in evolving networks . in this extended version , we provide additional mathematical details and supplementary material .
we explore some of the common oversights when these correlations are not taken into account , highlighting the importance of the formalism .
the formalism is based on a statistical description of the _ fraction of links _
@xmath0 connecting nodes of degrees @xmath1 and @xmath2 . to demonstrate its use
, we apply the framework to some pedagogical network models , namely , random - attachment , barabsi - albert preferential attachment and the classical erds and rnyi random graph . for these three models
the link - space matrix can be solved analytically .
we apply the formalism to a simple one - parameter growing network model whose numerical solution exemplifies the effect of degree - degree correlations for the resulting degree distribution .
we also employ the formalism to derive the degree distributions of two very simple network decay models , more specifically , that of random link deletion and random node deletion .
the formalism allows detailed analysis of the correlations within networks and we also employ it to derive the form of a perfectly non - assortative network for arbitrary degree distribution . |
in the past @xmath015 yr the study of cooling neutron stars in transiently accreting x - ray binaries has emerged as a new approach to understanding the structure of neutron stars .
@xcite discussed how the neutron - star crust in transiently accreting systems could be driven out of thermal equilibrium with the neutron - star core .
some of the heat that is generated by non - equilibrium nuclear reactions deep in the neutron - star crust during an outburst @xcite is radiated away in quiescence and this should result in observable cooling of the neutron - star surface , especially in systems that have undergone long - duration outbursts @xcite . by following the detailed evolution of the surface temperature during quiescence one can then extract information on the properties of the neutron - star crust . the neutron - star transient ks 1731260 provided the first opportunity to test this when it returned to quiescence following an outburst of more than 12.5 yr @xcite . during its first year in quiescence
ks 1731260 indeed showed significant cooling @xcite , and this continued during the next decade @xcite .
following ks 1731260 , cooling of neutron stars heated by transient accretion has been studied in a handful of systems : mxb 165929 @xcite , xte j1701462 @xcite , exo 0748676 @xcite , and igr j174802446 @xcite .
these systems had outburst durations ranging from @xmath011 weeks to @xmath024 yr , and the outburst - averaged luminosities ranged from a few percent of the eddington luminosity ( @xmath5 ) to @xmath0@xmath5 .
the ( long - term ) cooling seen in these systems has been modeled by a variety of cooling curves from increasingly sophisticated thermal evolution codes ( see , e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
together this has resulted in valuable insights into the thermal conductivity of the neutron - star crust , the location of heat sources in the crust , and , in some cases , the cooling mechanisms at work in the neutron - star core and crust ( e.g. , * ? ? ?
* ) .
the temperatures of the neutron star in maxi j0556332 during its first @xmath0500 days in quiescence are extraordinarily high compared to those measured for other cooling neutron stars ( see figure [ fig : sources ] ) . even at 500 days into quiescence its temperature is higher than the maximum quiescent temperature seen in the other five cooling systems .
it likely is also hotter than the neutron star in sax j1750.82900 , which was claimed by @xcite to be the hottest quiescent neutron star in an x - ray binary system , although maxi j0556332 is expected to continue cooling for a while .
the obtained temperatures for maxi j0556332 are sensitive to the assumed distance ( see table [ tab : fits ] ) , but even for distances well below our estimated distance range ( e.g. , 20 kpc ) we find temperatures ( 134218 ev for model i and 131195 ev for model ii ) that are substantially higher than those observed in xte j1701462 during its first @xmath0500 days ( 125163 ev ) .
the short cooling timescale observed in maxi j0556332 implies a high thermal conductivity of the crust , similar to the other cooling neutron stars that have been studied .
given the similarities between the outbursts of maxi j0556332 and xte j1701462 , it is interesting to compare these two systems in more detail , as it may help us understand what causes the neutron - star crust in maxi j0556332 to be so hot .
maxi j0556332 was in outburst for @xmath0480 days with a time - averaged luminosity of @xmath01.7@xmath9 ergs@xmath10 , while xte j1701462 was in outburst for @xmath0585 days with a time - averaged luminosity @xmath02.0@xmath11 ergs@xmath10 @xcite .
the total radiated energies of the maxi j0556332 and xte j1701462 outbursts are therefore 7.1@xmath12 erg and 1.0@xmath13 erg , respectively .
despite the fact that the radiated energies and time - averaged luminosities of the two systems are comparable , the initial luminosity of the thermal component ( which reflects the temperature at shallow depths in the crust at the end of the outburst ) is an order of magnitude higher in maxi j0556332 than in xte j1701462 .
this suggests the presence of additional shallow heat sources in the crust of maxi j0556332 and/or that the shallow heat sources in maxi j0556332 were more efficient per accreted nucleon .
the high observed temperatures are difficult to explain with current crustal heating models . bringing the initial temperatures down to those seen in xte j1701462 requires a distance of @xmath01015 kpc ( depending on the assumed model ) .
such distances are problematic for several reasons .
first it implies that z source behavior in maxi j0556332 is observed at much lower luminosities ( by factors of nine or more ) than in other z sources .
second , fits to the quiescent spectra with such a small distance are of poor quality .
finally , a smaller distance does not solve the fact that crustal heating appears to have been much more efficient per accreted nucleon than in other sources .
a reduction in distance by a factor of three results in a reduction in luminosity and presumably then , by extension , the total mass accreted onto the neutron star and total heat injected into the crust by a factor of nine . given that we inferred @xmath030% less mass accreted onto the neutron star in maxi j0556332 during its outburst than in xte j1701462 for our preferred distance of @xmath045 kpc
, this would mean @xmath012 times less mass accreted onto maxi j0556332 than xte j1701462 yet similar initial temperatures .
the nsa model that we used to fit the thermal emission from the neutron star in maxi j0556332 did not allow us to explore values of the neutron - star parameters other than @xmath14 and @xmath15 km , as these parameters are advised to remain fixed at those values @xcite .
while other neutron - star atmosphere models allow for changes in @xmath16 and @xmath17 , none of the available models are able to handle the high temperatures observed during the first @xmath0200 days of quiescence .
it is , of course , possible that the properties of the neutron star in maxi j0556332 are significantly different from those in the other cooling neutron - star transients that have been studied .
lower temperatures would be measured if one assumed a lower @xmath16 and/or a larger @xmath17 . to estimate the effects of changes in neutron - star parameters we used the nsatmos model to fit the spectrum of observation 11 , initially assuming @xmath14 and @xmath1810 km . while keeping the distance from this fit fixed , and changing @xmath16 to 1.2@xmath2 and @xmath17 to 13 km ( values that are still reasonable ) , the measured temperature was reduced by only @xmath010% .
such changes are not large enough to reconcile the temperatures measured in maxi j0556332 with those of the other sources . an alternative explanation for the high inferred temperatures could be that part of the quiescent thermal emission is caused by low - level accretion .
indications for low - level accretion in early quiescence were seen in the form of small flares in the _
swift_/xrt light curve ( figure [ fig : curves]b ; see below for a more detailed discussion ) .
the level of accretion in between those flares , where most of our observations were done , is unclear however .
@xcite suggest that low - level accretion onto neutron stars can result in thermal emission , although it would be hard to distinguish it from that of a cooling neutron star @xcite .
an additional hard x - ray component could be produced by emission from a radiatively inefficient ( advection - dominated ) accretion flow @xcite and/or by energetic protons illuminating the neutron - star surface @xcite . as shown in section [ sec : spectra ] , we find that the addition of a power - law to our spectral model ( to fit the hard , non - thermal emission ) presents a minor , but statistically significant improvement to our fits of the quiescent spectra of maxi j0556332 .
we also find that , when adding a power - law , the fluxes of the power - law and thermal components are strongly correlated ( see table [ tab : bestfit2 ] ) , which one would expect if low - level accretion played a significant role .
however , the index of the power - law is unusually low and when fixing it to commonly observed values ( @xmath19=1.52 ) the addition of a power - law no longer presents a statistical improvement .
it is possible that the nsa model simply does not present a good - enough description of the thermal emission from a non - accreting neutron star and that the ( shallow ) power law is merely fitting the resulting discrepancies ; this would likely also result in correlated behavior between the nsa and power - law fluxes , as observed .
@xcite showed that xte j1701462 likely had not yet reached a steady - state temperature profile at the end of its 19-month outburst . as a result of this
, the crust temperature profile had a dip at intermediate densities , leading to a temporary plateau in the temperature decay followed by late - time cooling , both of which have now been observed in xte j1701462 ( j.k .
fridriksson et al . , in preparation ) .
given the outburst similarities between xte j1701462 and maxi j0556332 , we expect the cooling curve of maxi j0556332 to eventually show similar features , although their onset may occur at different times than in xte j1701462 . in observations
4 and 6 large increases in the neutron - star temperature and luminosity and power - law luminosity were seen with respect to the overall trend .
this behavior is similar to that of two flares seen during the quiescent phase of xte j1701462 @xcite .
additional flares can be seen in the _ swift _ light curves shown in figure [ fig : curves]b .
significant flare - like variations in the quiescent luminosity of other ns - lmxbs have been reported as well : e.g. , a strong @xmath2016 d flare in sax j1750.82900 @xcite , a similarly strong ( @xmath03@xmath21 ergs@xmath10 ) @xmath204 day flare in ks 1741293 @xcite , and a @xmath01 week @xmath09@xmath21 ergs@xmath10 flare in grs j17412853 @xcite .
variations on timescales of hours to years have also been observed in the quiescent emission of cen x-4 @xcite , aql x-1 @xcite , and sax j1808.43658 @xcite .
the origin of these flares and variations are almost certainly accretion related .
the decay timescale of the second flare observed in xte j1701462 is similar to that of the main outburst decay @xcite , suggesting a common ( possibly viscous ) timescale .
the decay of the first of the flares seen in maxi j0556332 with _ swift _ , a @xmath010 d flare which ended just before our first _ chandra _ observation ( see figure [ fig : curves]a ) , is reasonably well resolved with _ swift _ and has a decay timescale of a few days , also similar to that of the outburst decay . moreover
, the power - law flux increases strongly during the flares and such strongly variable power - law components are not expected from a cooling neutron star , whereas they are commonly associated with low - level accretion , both onto neutron stars and black holes .
finally , the fact that the thermal and non - thermal component are correlated during the flares suggests that the accretion flow is reaching the neutron - star surface .
for this reason , the flare observations are not considered in our modeling of the cooling curves .
in addition to the early flaring activity , maxi j0556332 showed a much stronger rebrightening event , starting around day @xmath0170 of quiescence ; it lasted for @xmath060 days , reached a luminosity of @xmath08@xmath22 ergs@xmath10 , and had an average luminosity of @xmath04.4@xmath23 ergs@xmath10 . although the fluence of this event was a factor of @xmath033 smaller than that of the main outburst , it is only @xmath030% smaller than , e.g. , the 2010 outburst of igr j174802446 , in which moderate heating occurred @xcite .
however , our observations after the large rebrightening event did not suggest strong deviations from the ongoing cooling trend .
the distance to maxi j0556332 remains an uncertain factor in our fits to the quiescent spectra and modeling of the cooling curve , and this complicates the comparison with other sources .
we have employed two methods to estimate the distance to the source : one was a comparison of the cd / hid tracks with those of other z sources and the other the quiescent spectral fits themselves . in the other cooling neutron - star transients that have been studied ,
quiescent spectral fits suffer in general from a high degree of degeneracy between the normalization ( which depends on the distance and neutron - star radius ) and the temperature of the neutron - star atmosphere spectral component .
due to the low effective temperature , the small effective area of the detectors at the lowest energies , and the effects of interstellar absorption , it is primarily the high - energy tail of the thermal component that is being fitted , and usually fits of similar quality can be obtained for a rather wide range of distance ( and/or radius ) values .
this degeneracy is broken to a significant extent in maxi j0556332 due to the very low absorption towards the source and the ( presumably ) much higher temperatures exhibited by it .
the fits are quite sensitive to the value of the normalization and it was not possible to achieve fits of adequate quality except for large distance values .
our confidence in such a large value for the distance to the system is strengthened by the fact that the fits with these large distance values also give @xmath24 values that are consistent with other measurements , whereas significantly smaller distance values do not , and the fact that the best - fit distance ( @xmath045@xmath13 kpc for @xmath17 = 10 km ) is nicely consistent with the range we estimate from the cd / hid tracks of the source during outburst ( @xmath046@xmath115 kpc ) .
our distance estimates put maxi j0556332 far into the galactic halo .
while such a location is not common for x - ray binaries , there are several other systems whose properties also suggest very large distances : e.g. , gs 135464 has a distance greater than 27 kpc @xcite and igr j170913624 could be as distant as 35 kpc if its mass and luminosity are similar to grs 1915 + 105 @xcite , although both sources lie close to the galactic plane
. an optical spectrum of the companion star in quiescence might provide a better handle on the distance to maxi j0556332 than the above methods .
we are grateful to harvey tananbaum , belinda wilkes , and the _ chandra _ team for allowing us to use the remaining time of a previously triggered target of opportunity program .
we also thank norbert schartel for approving two _ xmm - newton _ director s discretionary time observations , and neil gehrels and his team for the flexible scheduling of our _ swift _ observations .
j.h . would like to thank the members of sron utrecht , where part of this work was done , for their hospitality .
this research has made use of data , software , and/or web tools obtained from nasa s high energy astrophysics science archive research center ( heasarc ) , a service of goddard space flight center and the smithsonian astrophysical observatory .
support for this work was provided by the national aeronautics and space administration through _ chandra _ award number go213052x issued by the _
chandra _
x - ray observatory center , which is operated by the smithsonian astrophysical observatory for and on behalf of the national aeronautics space administration under contract nas8 - 03060 , and through _ swift _ award number nnx12ad56 g . r.w . was supported by an european research council starting grant .
finally , we thank the referee for his / her constructive comments . | we present _ chandra _ , _ xmm - newton _ , and _ swift _ observations of the quiescent neutron star in the transient low - mass x - ray binary maxi j0556332 .
observations of the source made during outburst ( with the _ rossi x - ray timing explorer _ ) reveal tracks in its x - ray color color and hardness
intensity diagrams that closely resemble those of the neutron - star z sources , suggesting that maxi j0556332 had near- or super - eddington luminosities for a large part of its @xmath016 month outburst .
a comparison of these diagrams with those of other z sources suggests a source distance of 46@xmath115 kpc .
fits to the quiescent spectra of maxi j0556332 with a neutron - star atmosphere model ( with or without a power - law component ) result in distance estimates of 45@xmath13 kpc , for a neutron - star radius of 10 km and a mass of 1.4 @xmath2 .
the spectra show the effective surface temperature of the neutron star decreasing monotonically over the first @xmath0500 days of quiescence , except for two observations that were likely affected by enhanced low - level accretion .
the temperatures we obtain for the fits that include a power - law ( @xmath3=184308 ev ) are much higher than those seen for any other neutron star heated by accretion , while the inferred cooling ( @xmath4-folding ) timescale ( @xmath0200 days ) is similar to other sources .
fits without a power - law yield higher temperatures ( @xmath3=190336 ev ) and a shorter @xmath4-folding time ( @xmath0160 days ) .
our results suggest that the heating of the neutron - star crust in maxi j0556332 was considerably more efficient than for other systems , possibly indicating additional or more efficient shallow heat sources in its crust . |
the investigation of the formation and the stationary properties of random structures has been an exciting topic in statistical physics for many years .
since it provides an intuitively appealing and transparent model of the irregular geometry emerging in disordered systems , _ percolation _ has provided a leading paradigm for random structures .
_ bond percolation _ constitutes perhaps the simplest percolation problem .
two fundamental variants of bond percolation have been introduced in the past : in the _ isotropic percolation _ ( ip ) problem the bonds connecting the sites of a regular lattice ( in @xmath2 spatial dimensions ) are randomly assigned to be open ( with probability @xmath3 ) or blocked ( with probability @xmath4 ) , and an agent may traverse an open bond in either direction .
in contrast , in the _ directed percolation _
( dp ) problem the open bonds can be passed only from one of the two connecting sites , whence the allowed passage direction globally defines a preferred direction in space , see fig .
[ dirper ] . in these terms
, percolation can be viewed as the passage of an agent through an irregularly structured medium , in the sense that the agent can propagate through certain regions , whereas it can not traverse other areas .
though percolation represents one of the simplest models for random systems , it has in fact many applications .
moreover , it yields a prototypical _ non - equilibrium phase transition _ : for small values of @xmath3 , regions with open bonds form disconnected clusters of typical linear extension @xmath5 , wherein the agent becomes localized or trapped . for @xmath6 , a critical threshold value for the frequency of open bonds ,
the probability for the existence of an infinite connected accessible cluster becomes non - zero .
therefore , an agent on an infinite cluster is not confined to a finite region of space any more , but may percolate through the entire system .
this percolation transition defines a genuine _ critical phenomenon _ as encountered in equilibrium statistical mechanics , with @xmath5 providing the divergent length scale as @xmath7 is approached .
it turns out that many , microscopically quite different percolation - type systems share their critical properties either with ip or dp .
thus , the phase transitions in ip and dp , the latter distinguished from the former through the broken spatial isotropy , define genuine _
universality classes_. for reviews on isotropic and directed percolation the reader is referred to refs .
critical ip and dp clusters are shown in fig .
[ perccl ] .
= 6 truecm = 2.5 truecm broadbent and hammersley @xcite first directed attention to the _ dynamics _ of percolation processes .
in contrast to random diffusion , in percolation processes the spreading of agents is mainly governed by the disordered structure of the medium .
consequently , they define stochastic growth processes where agents randomly generate offspring at neighboring sites , with rates determined by the susceptibility of the medium .
the agents themselves decay spontaneously , thereby producing debris . in this way
the medium becomes exhausted , and its susceptibility is randomly diminished .
hence , static dp in @xmath8 spatial dimensions can be directly interpreted as a markov process in @xmath2 spatial and @xmath9 time dimension .
the preferred ( propagation ) direction defines the temporal axis , and the history of the process yields the debris that forms the ( @xmath8)-dimensional spatially anisotropic _ directed _ percolation clusters .
in contrast , in a _ dynamical isotropic percolation _ ( dip ) process the agents grow from a seed in an expanding stochastic annulus in @xmath2 spatial dimensions leaving behind the debris in their wake .
this debris then forms @xmath2-dimensional _ isotropic _ percolation clusters .
remarkably , the renormalized field theory of dp has appeared originally in elementary particle physics in the guise of _ reggeon field theory _
grassberger et al .
@xcite pointed out that reggeon field theory is not a hamiltonian field theory but constitutes a stochastic process for which they coined the name _
gribov process_. it represents a stochastic version of schlgl s ` first reaction ' @xcite .
subsequently , the formal connection of reggeon field theory to dp was explicitly demonstrated @xcite , and janssen and grassberger stated the _ dp conjecture _ : the critical behavior of an order parameter field with markovian stochastic dynamics , decoupled from any other slow variable , that describes a transition from an active to an inactive , _ absorbing state _
( where all dynamics ceases ) should be in the dp universality class @xcite .
the field theory of dip was initiated by grassberger s formulation of the _ general epidemic process _ ( gep ) on a lattice , and his statement that near criticality this process produces isotropic percolation clusters @xcite .
this insight led to the construction of the renormalized field theory for the dip universality class @xcite .
it is tempting to express percolation processes in the language of an epidemic disease such as blight in a large orchard , the spreading of bark beetles in a forest , or proliferation of a forest fire , as depicted in fig . [ waldbrand ] . here
, the susceptible individuals or healthy trees form the medium , and the disease , blight , beetles , or fire represent the agent .
the sick individuals , the befallen or burning trees may randomly infect neighboring individuals .
the sick individuals are allowed to recover , becoming susceptible anew .
this behavior defines the so - called _ simple epidemic process _ ( sep ) , also known as _ epidemic with recovery _
properties of the sep near the ensuing critical point separating the endemic and pandemic phases ( whose precise location depends on the susceptibility of the individuals ) shares the universal properties of dp . in contrast , in a situation where the sick individuals die out or become immune as opposed to susceptible again , the medium is eventually exhausted .
this scenario defines the _ general epidemic process _ ( gep ) , also termed _
epidemic with removal _ @xcite .
its universal properties are governed by the dip universality class .
the statistical properties of the debris clusters that are left behind after the disease is extinguished are described by static ip . in the recent physical literature ,
basically two different methods have been employed to construct field theories of percolation processes .
the first approach starts from a specific microscopic stochastic reaction - diffusion model formulated in terms of a master equation on a lattice .
basically it consists of a description of the random walks generated by diffusing and reacting particles .
a field theory is then obtained through a representation of the master equation by means of a bosonic creation / destruction operator formalism and introduction of coherent - state path integrals , followed by a naive continuum limit @xcite .
the second method , to be applied in the first part of this article , takes a more universal , phenomenological perspective , and can be viewed as a dynamical analog of the established landau - ginzburg - wilson approach to static critical phenomena .
it is based on formulating the fundamental principles shared by all systems in a given universality class , and directly employs a _
stochastic continuum theory .
of course , one has to a - priori define the slow dynamic fields in the theory , typically local densities of the order parameter and any conserved quantities , as characterizing the universality class under consideration .
this approach assumes that on a mesoscopic scale the motion of the fast microscopic variables can be absorbed into the statistical properties of the stochastic equations of motion for the relevant slow variables .
the corresponding stochastic processes should be markovian provided all slow variables are retained .
taking into account the general ( symmetry , conservation , etc . )
properties that define the universality class , the stochastic field theory is constructed using small - density and gradient expansions , i.e. , long - wavelength and low - frequency approximations that focus on the infrared ( ir ) properties of the system . for percolation processes , the fundamental statement regarding the slow variables consists in the assertion that percolation near the critical point can be described by a markov process that only depends upon the density @xmath10 of activated sick parts of the medium and , in the case of dip , upon the density of the debris @xmath11 . in terms of the field
@xmath12 the density of the debris is given by @xmath13 , where @xmath14 denotes an appropriate kinetic coefficient .
indeed , this assumption that percolation can be described solely in terms of @xmath12 becomes manifest in typical lattice simulations of such processes .
near the critical point one observes a patchwork of regions consisting of _ correlated _ active neighboring sites and vacuum regions devoid of any activity .
in stark contrast to this picture , simulations for branching and annihilating random walks show anticorrelating behavior , namely widely separated single active sites propagating on isolated paths and reacting only after encounter @xcite .
we now proceed to formulate four principles that allow the explicit construction of a mesoscopic stochastic description of percolation processes .
we shall describe the universal aspects of the dynamics of @xmath12 beginning with the dp universality class : 1 .
the susceptible medium becomes locally infected , depending on the density @xmath12 of neighboring sick individuals .
the infected regions recover after a brief time interval .
the state with @xmath15 is absorbing .
this state is equivalent to the extinction of the disease .
3 . the disease spreads out diffusively via the short - range infection ( i ) of neighboring susceptible regions .
microscopic fast degrees of freedom may be captured as local noise or stochastic forces that respect statement ( ii ) , i.e. , the noise alone can not regenerate the disease . for the dip universality class ,
we need to modify principles ( i ) and ( ii ) by 1 .
the susceptible medium becomes infected , depending on the densities @xmath12 _ and _ @xmath16 of sick individuals and the debris , respectively . after a brief time interval , the sick individuals decay into the immune debris .
the debris ultimately stops the disease locally by exhausting the supply of susceptible regions .
the states with @xmath15 and any spatial distribution of @xmath16 are absorbing .
these states are equivalent to the extinction of the disease .
it is important to realize that the mechanism ( i ) introduces _ memory _ into the stochastic process .
note that we do not explicitly assume diffusive motion of the sick individuals .
all that is needed is a spreading mechanism for the agent , here the disease itself : @xmath17 with a spreading probability @xmath18 that is assumed short - range , hence allowing to approximate its propagation by simple diffusion .
however , this scenario by no means excludes reaction - diffusion systems where the `` infected '' particles themselves spread out diffusively , if otherwise the general percolation principles are satisfied .
we proceed to set up the stochastic equations of motion for percolation processes according to the general principles formulated in the preceding section , @xmath19 or , in a discretized version ( in the it sense ) @xmath20 with @xmath21 , @xmath22 here , @xmath23 denotes a random density that has yet to be constructed , and the second equations for @xmath16 only appear for the dip process . for simplicity we shall suppress spatial arguments in the following considerations .
the markovian assumption implies that the stochastic variables @xmath24 are uncorrelated at different times , and that the statistical properties of @xmath24 depend solely upon @xmath25 and @xmath26 . as a consequence
, the generating function of the cumulants of @xmath27 ( i.e. , the laplace transform of the corresponding probability distribution ) must have the general form @xmath28 \big ) \
, , \ ] ] where the overbar denotes the statistical average over the fast microscopic degrees of freedom .
the @xmath29 constitute independent new variables . of course
, in the dp process the cumulants @xmath30 are independent of the debris @xmath16 .
the statistical properties of the stochastic process are fully encoded in the simultaneous probability density of the history @xmath31 : @xmath32 } \\ & = \int \prod_k \frac{d\tilde{n}(t_{k+1})}{2\pi i } \
\overline{\exp \big\ { \sum_k \tilde{n}(t_{k+1 } ) \bigl [ \delta \mathcal{v}(t_k)+n(t_k)-n(t_{k+1 } ) \bigr ] \big\ } } \ , . \nonumber\end{aligned}\ ] ] upon inserting ( [ kum - v ] ) and reverting to continuous time , we may then formally write @xmath33 as a path integral over functions @xmath34 : @xmath35 \ , \exp \int \ ! dt \ , \biggl\ { \ , \sum_{l=1}^\infty \frac{\tilde{n}(t)^l}{l ! } \ , k_l[n(t),m(t ) ] - \tilde{n}(t ) \ , \partial_t n(t ) \ , \biggr\ } \ , .\ ] ] this expression is always to be interpreted through the preceding prepoint ( it ) discretization ; we note that this specifically implies @xmath36 for the heaviside function that enters the dynamic response functions .
the integration over the ` response variable ' @xmath37 runs along the imaginary axis from @xmath38 to @xmath39 .
however , this path can be analytically deformed in finite regions .
therefore @xmath37 can attain real contributions .
all cumulants @xmath30 are subject to a fundamental constraint as a consequence of the existence of the absorbing state(s ) : they must vanish for @xmath40 . assuming that the functionals @xmath30 can be expanded in powers of @xmath12 and @xmath16
, this implies that all @xmath41 or higher powers of @xmath12 .
moreover , as we shall demonstrate below , the higher cumulants @xmath30 with @xmath42 are irrelevant in the rg sense .
thus , the remaining task is to find the functional forms of the mean - field part @xmath43 and the gaussian stochastic force correlator @xmath44 . after reducing the statistical properties of @xmath27 to these two terms , the stochastic equation of motion for @xmath12
may be written in the usual langevin form ( in the it interpretation ) [ langevgl ] @xmath45 + \zeta(t ) \
, , \label{langevgl1 } \\
\overline{\zeta(t ) \ , \zeta(t^\prime ) } & = k_2[n(t),m(t ) ] \ , \delta(t - t^\prime ) \ , . \label{langevgl2}\end{aligned}\ ] ] note that @xmath46 since our theory is based on a real field @xmath12 .
this is in contrast to the path integral representation of the theory of random walks subject to pair annihilation which may formally lead to a negative second cumulant @xcite .
statistical averages of any functional of the slow variable @xmath12 can now be performed by means of path integrals over the fields @xmath12 and @xmath37 with the weight @xmath47 @xcite ; reinstating the spatial arguments , the integral measure here is @xmath48 = \prod_{\mathbf{r},t } \ , d\tilde{n}(\mathbf{r},t ) \ , dn(\mathbf{r},t ) / 2\pi i$ ] . retaining only @xmath43 and @xmath44 , the _ dynamic response functional _
@xmath49 from eq .
( [ simprob ] ) becomes @xmath50 \bigr ] - \frac{1}{2 } \
, \tilde{n}(t)^2 \ , k_2[n(t),m(t ) ] \ , \big\ } \ , .\ ] ] here integrations over @xmath2-dimensional space are implicit
. fundamental principles ( ii ) and ( ii ) classify dp and dip as absorbing state systems : any finite realization will reach an absorbing state in a finite time @xmath51 with probability @xmath9 .
however , in the active pandemic state above the critical percolation threshold @xmath7 , @xmath51 grows exponentially with the system size @xmath52 .
it is thus appropriate to set @xmath53 in the thermodynamic limit @xmath54 .
the assertion that principles ( i)(iv ) lead to the dp universality class is known in the literature as the dp conjecture @xcite .
the diffusive spreading of the disease renders the cumulants @xmath43 and @xmath44 _ local _ functionals of @xmath12 and @xmath16 , whence a gradient expansion is appropriate .
the absorbing state condition then inevitably implies @xmath55 where the ellipsis denotes higher - order gradient terms . in the same manner , @xmath56 \ , \delta(\mathbf{r}-\mathbf{r}^\prime ) \ , .\ ] ] the contributions with kinetic coefficients @xmath57 and @xmath58 can be interpreted as purely diffusional noise if @xmath59 . yet here
these terms simply arise from the gradient expansion .
we shall demonstrate below that these as well as higher - order terms are in fact irrelevant for the critical properties of percolation .
naturally , the dependence of @xmath16 is absent for dp .
furthermore , @xmath12 and @xmath16 are small quantities near the critical point .
this allows for a low - density expansion @xmath60 with @xmath61 for dp . to ensure stability ,
the coupling constants @xmath62 are assumed to be positive ; otherwise one has to extend the expansion to the subsequent powers .
note that there is no general reason for @xmath63 to vanish as is the case for interacting random walks @xcite . discarding the irrelevant higher - order expansion terms ,
indicated by the ellipsis in eqs .
( [ k1])([dens_exp ] ) , and inserting into eq .
( [ j1 ] ) , we at last arrive at the fundamental dynamic response functional for percolation processes : @xmath64 n - q \ , \tilde{n } \big\ } \ , .\ ] ] here we have introduced an additional external source @xmath65 for the agent . specifically , a seed inserted at the origin @xmath66 at time @xmath67 in order to initialize a spreading process is modeled by @xmath68 . the evaluation of the path integrals is always restricted by the initial and final conditions @xmath69 , respectively . after eliminating the fast microscopic degrees of freedom , and thereby focusing on the ir properties of the relevant slow variables , the wave vectors @xmath70 of the fluctuations of the fourier transformed fields @xmath71 and @xmath72
is clearly limited by a momentum cutoff @xmath73 . correspondingly , those fluctuations are effectively set to zero for @xmath74 .
an alternative approach to constructing a field theory representation for a stochastic system starts from the classical master equation that defines the process microscopically . for reaction - diffusion systems
, one may formulate the reactions in a straightforward manner in terms of ( bosonic ) creation and annihilation operators , and therefrom via coherent - state path integrals proceed to a field theory action @xcite .
this method invokes no phenomenological assumptions or explicit coarse - graining , but does rely on a the validity of a ` naive ' continuum limit .
one may thus view this procedure in analogy to the derivation of the @xmath75 theory from a soft - spin ising model on a lattice by directly taking the limit of zero lattice constant .
universal features do not immediately become apparent in such a treatment , whereas the approach laid out in the preceding section focuses on general principles and in this sense rather corresponds to the landau - ginzburg - wilson model for second - order equilibrium phase transitions .
the phenomenological construction certainly requires considerable a - priori insight on the defining universal features of a given system , but its benefit is considerable predictive power .
yet there are cases , such as pure pair annihilation processes @xcite or branching and annihilating random walks with even offspring @xcite , for which as yet only the master equation formulation has led to successful field theory representations . in these processes
the random - walk properties of individual particles govern the essential physics ; hence the microscopic picture is essential and a mesoscopic description simply in terms of the particle density would be erroneous . on the other hand ,
proceeding from the microscopic model via taking a naive continuum limit definitely fails , e.g. , for the pair contact process with diffusion ( pcpd ) @xcite .
we shall consider here a specific reaction - diffusion model that displays an active to absorbing state phase transition which , according to our general principles , should belong to the percolation universality classes . yet reaction - diffusion systems do of course not provide an exact description of , e.g. , directed bond percolation or the contact process since they differ manifestly with respect to the microscopic details .
we note that one may also construct a direct field theory representation for the pair connectivity of bond - percolating system @xcite . as usual in reaction - diffusion models , the activated particles both represent the diffusing agents and may locally create new offspring .
we introduce the following reaction scheme : [ reakt ] @xmath76 supplemented with hopping of the agents @xmath77 to nearest - neighbors on a @xmath2-dimensional lattice with sites @xmath78 with diffusion constant @xmath14 . the back - reaction in ( [ reakt1 ] )
may also be viewed as effectively capturing mutual exclusion on the same site .
the reactions in ( [ reakt2 ] ) are specific to dip where the presence of the produced debris @xmath79 suppresses the agent @xmath77 .
the reaction rules ( [ reakt ] ) are reformulated in terms of a master equation that describes the time dependence of the probability @xmath80 for a given configuration of site occupation numbers @xmath81 and @xmath82 of the agent @xmath77 and the debris @xmath79 , respectively .
the configuration probability is encoded in the state vector @xmath83 in a bosonic fock space spanned by the basis @xmath84 .
these vectors as well as the stochastic processes in the master equation are then expressed through the action of bosonic creation and annihilation operators @xmath85 and @xmath86 , respectively , which are defined via @xmath87 , and @xmath88 , etc .
subsequently , the master equation can be written in the form @xmath89 with an appropriate non - hermitean pseudo - hamilton operator @xcite .
for example , the reaction scheme ( [ reakt ] ) leads to @xmath90 with [ h ] @xmath91 \ , .
\label{h - react}\end{aligned}\ ] ] here @xmath92 denotes a pair of neighboring sites . in order to compute statistical averages
it is necessary to introduce the projection state @xmath93 . using the identity @xmath94
one easily finds the expectation value of an observable @xmath95 at time @xmath96 : @xmath97 where in the last expression @xmath12 and @xmath16 are replaced with the operators @xmath98 and @xmath99 , respectively .
the formal solution of the equation of motion ( [ ham ] ) reads @xmath100 . following standard procedures @xcite the expectation value ( [ expv ] )
can be expressed as a path integral @xmath101 \ ,
a(\ { \hat{a } a , \hat{b } b \ } ) \ , \exp ( -s[\hat{a},\hat{b},a , b ] ) \ , , \ ] ] with an exponential weight that defines a field theory action . after applying a ( naive ) continuum limit ,
the action becomes @xmath102 \ , .
\label{s1}\end{aligned}\ ] ] here , the fields @xmath103 , @xmath104 , @xmath105 , and @xmath106 correspond to the coherent - state eigenvalues of the bosonic creation and annihilation operators .
the integrations are subject to @xmath107 , @xmath108 , and the final conditions are @xmath109 .
thus often a field shift according to @xmath110 , @xmath111 is useful .
the new variables @xmath112 then appear only linearly in the action ( [ s1 ] ) . hence , they can be integrated out , which results in the differential equation constraint @xmath113 and the new action @xmath114 a \ , , \ ] ] where @xmath115 is given by the solution of eq .
( [ db ] ) .
if we now discard all fourth - order terms which should become irrelevant after the application of coarse - graining near the critical point , the action ( [ s2 ] ) attains the same form as the response functional ( [ j2 ] ) , with the correct sign of the coupling constants .
physically , however , the fields are of different origin : whereas @xmath12 in the action @xmath49 represents the fluctuating density of the active medium , the field @xmath116 in ( [ s1 ] ) or ( [ s2 ] ) results from the bosonic annihilation operators for the random walks ; therefore , @xmath116 generally is a complex - valued quantity , and only @xmath117 has to be real and non - negative .
thus , the action ( [ s2 ] ) derived above really is akin to the original reggeon field theory @xcite .
this can be remedied by means of a quasi - canonical transformation to proper density variables @xmath118 , @xmath119 , @xmath120 , and @xmath121 .
after integrating by parts the action ( [ s1 ] ) then becomes @xmath122 \nonumber \\ & \qquad\qquad\quad + \big [ 1 - \exp(-\tilde{n } ) \big ] \big [ \sigma - \rho \exp(\tilde{n } ) + \kappa n\big ] n \nonumber \\ & \qquad\qquad + \tilde{m } \ , \partial_t m + \mu \big [ 1 - \exp(\tilde{m}-\tilde{n } ) \big ] n + \nu \big [ 1 - \exp(-\tilde{n } ) \big ] m \ , n \big\ } \ , .
\label{s3}\end{aligned}\ ] ] integration over @xmath123 , followed by the expansion of the exponentials , and dispensing with fourth - order terms finally leads to @xmath124 \nonumber \\ & \qquad\qquad\quad + \tilde{n } \big [ ( \sigma + \mu - \rho ) + \kappa \ , n + \nu \ , m - \frac{\rho+\sigma}{2 } \
, \tilde{n } \big ] n \big\ } \ , , \label{s4}\end{aligned}\ ] ] with @xmath125 .
this action again acquires the same form as the response functional ( [ j2 ] ) . aside from the different meaning of the fields , it is reminiscent of the action ( [ s2 ] ) without the fourth - order contributions
however , irrelevant diffusional noise arises in the action ( [ s4 ] ) , and the noise term @xmath126 comes with a slightly different coupling constant that is invariably negative , even in the `` free '' case with @xmath127 . hence , as remarked above , the stochastic processes described by the actions @xmath128 and @xmath129 are in fact distinct , yet their _ universal _ features are identical .
they will however usually differ with respect to non - asymptotic , non - universal details .
the first approximation in the evaluation of path integrals generally consists of a gaussian truncation in the action @xmath49 with respect to the fluctuations about the maximum of the statistical weight @xmath47 .
the extrema are in turn determined from the saddle - point ( mean - field ) equations [ spe ] @xmath130 n - q \ , , \label{spe1 } \\ 0 = \frac{\delta\mathcal{j}}{\delta n } & = \big [ - \partial_t + \lambda ( \tau - \nabla^2 ) + \lambda ( 2 g_1 \ ,
n + g_2 \ , m - g_3 \ , \tilde{n } ) \big ] \tilde{n } \nonumber \\ & \qquad\qquad\qquad\qquad + \lambda \ , g_2 \int_t^\infty \ !
dt^{\prime } \ , n(t^{\prime } ) \ , \tilde{n}(t^{\prime } ) \ , .
\label{spe2}\end{aligned}\ ] ] for @xmath131 , the stable homogeneous stationary solutions in the case of homogeneous vanishing source @xmath132 are @xmath133 , and @xmath40 if @xmath134 .
for @xmath135 we obtain @xmath136 for @xmath61 ( dp ) , and @xmath40 , @xmath137 for @xmath138 ( dip ) .
this demonstrates that within the mean - field approximation the critical point is located at @xmath139 , and in its vicinity on the active side the order parameter vanishes as @xmath140 with critical exponent @xmath141 .
now let us scale spatial distances @xmath142 by a convenient mesoscopic length scale @xmath143 .
consequently we find the naive scaling dimensions @xmath144 , @xmath145 , and @xmath146 .
hence , in the asymptotic long - time and large - distance limit , @xmath147 and @xmath148 constitute _ relevant _ parameters , flowing to @xmath149 under successive rg scale transformations .
the last relation shows that the coupling @xmath150 becomes irrelevant , provided @xmath151 ( dip ) .
since the action @xmath49 is dimensionless , we find from eq .
( [ j2 ] ) that @xmath152 .
it is characteristic of the field theory for spreading phenomena with an absorbing state that the dynamic response functional @xmath49 contains a _ redundant _
parameter @xcite that must be eliminated by a suitable rescaling @xmath153 with an amplitude @xmath154 that generally carries nonvanishing scaling dimension . because both the lowest - order non - vanishing coupling constants in the mean - field part of the action and the stochastic noise strength are clearly required for a meaningful non - trivial perturbation expansion , it is convenient to choose @xmath154 such that the corresponding couplings attain the same scaling dimensions . thus we set @xmath155 in the case of dp , and @xmath156 for dip , where @xmath157 parametrizes , e.g. , the line of dp transitions in the domany - kinzel automaton @xcite .
we note that this amplitude characterizes the non - universal `` lacunarity '' property of percolating clusters which tends to zero in the case of compact percolation .
the dynamic response functional now assumes the two distinct forms [ j3 ] @xmath158 s - \lambda h \ , \tilde{s } \big\ } \ , , \label{jdp } \\
\mathcal{j}_{\rm dip } & = \int \ !
d^dr \ , dt \ , \big\ { \tilde{s } \big [ \partial_t + \lambda ( \tau - \nabla^2 ) + \frac{\lambda g}{2 } \ , ( 2s - \tilde{s } ) \big ] s - \lambda h \ , \tilde{s } \big\ } \ , , \label{jdip}\end{aligned}\ ] ] for dp and dip , respectively , where @xmath159 .
hence , a seed at the origin , i.e. , a sick individual at @xmath160 , is represented by @xmath161 .
after fixing the redundancy in this manner the naive scaling of the fields and couplings is uniquely given by [ skal ] @xmath162 from the scaling dimensions of the couplings @xmath163 we infer the upper critical dimensions @xmath0 for dp , and @xmath164 for dip .
at this point it is also straightforward to show that all higher - order terms in the gradient and density expansions that we had neglected before in fact acquire negative scaling dimensions near @xmath165 .
this proves that those terms are indeed irrelevant for the asymptotic ir scaling behavior .
note that relevance and irrelevance are assigned here with reference to the gaussian theory . after discarding the irrelevant terms , the dynamic response functionals ( [ skal ] ) display duality invariance with respect to time inversion , @xmath166 however , in general these symmetries only hold asymptotically . the important goal of statistical theories is the determination of correlation and response functions ( generally called green s functions ) of the dynamical variables as functions of their space - time coordinates , as well as of the relevant control parameters . in a compact form , one attempts to determine the _ cumulant generating functional _
@xmath167 = \ln \int \mathcal{d}[\tilde{s},s ] \ , \exp \big [ - \mathcal{j}[\tilde{s},s ] + ( h , s ) + ( \tilde{h},\tilde{s } ) \big ] \ , .\ ] ] functional derivatives with respect to the sources @xmath168 and @xmath169 , @xmath170 [ \delta\tilde{h}^{\tilde{n } } ] } \right\vert_{h=\tilde{h}=0 } = \langle [ s^n][\tilde{s}^{\tilde{n } } ] \rangle^{\rm ( cum ) } = : g_{n,\tilde{n } } \ , , \ ] ] define the _ green s functions _ ( here we suppress all space and time coordinates ) . in general , an exact expression for @xmath171 $ ] can not be found and one must resort to a perturbational evaluation .
perturbation theory is developed starting from the gaussian contribution @xmath172 to the weight , and the subsequent expansion of the remainder @xmath173 , where @xmath174 .
the different contributions to the perturbation series are graphically organized in successive order of closed loops in terms of linked diagrams which , in translationally invariant theories , can be decomposed into one - line irreducible amputated feynman diagrams that represent the building blocks for the _ vertex functions_. the generating functional for the vertex functions @xmath175 $ ] is related to the cumulant generating functional via the legendre transformation [ vert - kum ] @xmath176 & + \mathcal{w}[h,\tilde{h } ] = ( h , s ) + ( \tilde { h},\tilde{s } ) \ , , \label{vert - kum-1 } \\ \text{with } \ & \ s = \frac{\delta\mathcal{w}}{\delta h } \
, , \quad \tilde{s } = \frac{\delta\mathcal{w}}{\delta\tilde{h } } \ , , \label{vert_kum-2}\end{aligned}\ ] ] and vice versa .
it is easy to see that to zero - loop ( ` tree ' ) order @xmath175 $ ] is identical to @xmath177 $ ] .
the vertex functions are defined via the functional derivatives @xmath178 } { [ \delta^{\tilde{n}}\tilde{s}][\delta s^n ] } \right\vert_{\tilde{s}=s=0 } = : \gamma_{\tilde{n},n } \
, , \ ] ] and they are represented by irreducible diagrams if @xmath179 .
note that all the @xmath180 necessarily vanish as a consequence of causality .
the gaussian parts of the response functionals eqs .
( [ j3 ] ) define the _ propagator _ [ prop ] @xmath181 where the momentum integral is limited by the ultraviolet ( uv ) cutoff @xmath73 .
we have within the tree approximation @xmath182 the non - linear contributions from @xmath183 yield the vertices of the field theory .
as is common with quantum and statistical field theories , the naive perturbation expansion , here based on our response functionals ( [ j3 ] ) , is problematic for several reasons @xcite . to be specific , consider the inverse susceptibility @xmath184 that describes the linear response to a stationary homogeneous external source
. we shall be interested in how the structure of the theory depends on the spatial dimension @xmath2 as encoded in the parameter @xmath185 .
first , the perturbation series in @xmath186 in general is divergent .
although it is characterized by a vanishing radius of convergence , it presumably constitutes an asymptotic series and should be resummable .
yet more severe problems arise from the massless critical limit .
the critical point @xmath187 , defined implicitly by @xmath188 , may by dimensional arguments be written in the form @xmath189 however , the function @xmath190 is not calculable by means of perturbation theory , since it diverges in the limit @xmath191 for all positive @xmath192 .
the limit @xmath193 does exist for @xmath194 , and an analytic continuation to @xmath192 defines the dimensionally regularized symanzik function @xmath195 .
but now @xmath195 displays simple ir poles at all @xmath196 with positive integers @xmath12 .
note that at these discrete points @xmath197 .
let us now introduce the new variable @xmath198 .
the vertex functions @xmath199 , regarded as functions of @xmath200 instead of @xmath147 , are free of ir singularities for @xmath201 , but they require the perturbationally inaccessible function ( [ tau_c ] ) . in order to eliminate the ir poles completely from the calculation in any dimension @xmath202 , it is possible at this stage to change variables to the correlation length squared @xcite , @xmath203 since the function @xmath5 , which does not display ir poles either , should be a monotonic function of @xmath200 , it can be inverted to @xmath204 with @xmath205 .
finally one may substitute @xmath200 by @xmath206 and thus arrive at vertex functions written in terms of @xmath5 , i.e. , @xmath207 , which are now calculable in perturbation theory and are devoid of ir singularities for @xmath208 and @xmath209 .
we return to the consideration of the analytical properties of the response function .
dimensional analysis yields @xmath210 with the perturbational expansion @xmath211 for @xmath208 , the functions @xmath212 are divergent in the critical limit @xmath213 , a direct consequence of the fact that @xmath214 for @xmath215 with a critical exponent @xmath216 if @xmath202 .
hence , a series in @xmath217 appears to be produced by the perturbational expansion which needs to be properly resummed .
wilson s momentum - shell renormalization procedure @xcite fully avoids the ir problems of the naive perturbation expansion .
furthermore , in contrast to renormalized field theory , wilson s approach does not require the elimination of the ir - irrelevant couplings prior to deriving the rg flow equations , and so pre - asymptotic critical as well as crossover properties are calculable . however , even though wilson s rg procedure in that sense is superior to the field - theoretic method , it is generally not advisable for a systematic calculation of universal properties to higher then one - loop order .
in contrast , the field - theoretic rg method @xcite does not proceed by successive elimination of short - wavelength degrees of freedom and a rescaling of parameters , but instead exploits the uv - renormalizability of the perturbation expansion for @xmath218 to enable a mapping from the critical to a non - critical region in parameter space wherein perturbational calculations are unproblematic .
both methods are fully equivalent with respect to describing asymptotic and universal features .
we rename the original bare fields and parameters according to @xmath219 , @xmath220 , @xmath221 , etc . in accord with the symmetries ( [ zeitsp ] ) we choose the following multiplicative renormalizations [ ren - z ] @xmath222 where @xmath223 is a convenient amplitude .
@xmath224 represents the dimensionless coupling constant , and @xmath139 at the critical point .
the renormalization constants @xmath225 can be chosen in a uv - renormalizable theory in such a way that [ ren - gg ] @xmath226 , \label{ren - gamma}\\ & \mathring{g}_{n,\tilde{n}}(\{\mathbf{r},t\},\mathring{\tau } , \mathring{g},\mathring{\lambda},\lambda ) = z^{n/2 } \tilde{z}^{\tilde{n}/2 } g_{n,\tilde{n}}(\{\mathbf{r},t\},\tau , u,\lambda,\mu ) \big [ 1 + o\big ( ( \lambda\xi)^{-\delta } \big ) \big ] , \label{ren - greens}\end{aligned}\ ] ] with @xmath227 . within each successive order of the perturbation expansion ,
the renormalized vertex functions @xmath228 are thereby rendered finite and well - defined .
note that the physical bare and the renormalized vertex function display the same dependence on the variables @xmath229 in the limit @xmath230 up to nonuniversal amplitudes . in principle
, the infinite cutoff limit is unphysical and only employed here to develop a systematic rg mapping that works effectively to higher orders in the loop expansion .
the theory becomes only uv renormalizable at the upper critical dimension @xmath165 ( super - renormalizable below @xmath165 ) because we have previously eliminated all ir - irrelevant couplings and shifted the relevant control parameters ( here , @xmath147 ) to be zero at the critical point .
indeed , the problematic uv and ir singularities are linked precisely at the critical dimension @xmath231 . for @xmath232
the field theory is ir - finite and uv - infinite , whereas conversely for @xmath202 it is ir - infinite and uv - finite . for @xmath218 ,
the renormalization factors diverge in three distinct limits the first two represent uv divergences , while the third one constitutes an ir singularity : @xmath233 if 1 .
@xmath234 , @xmath235 , @xmath224 fixed ; 2 .
@xmath236 , @xmath237 , @xmath224 fixed ; 3 .
@xmath238 , @xmath239 , @xmath208 fixed . here
@xmath240 denotes the first nontrivial zero of the gell - mann low function @xmath241 to be introduced below , and @xmath242 is the solution of the rg flow equation @xmath243 .
hence , the determination of the @xmath244 factors from the uv divergences provides us at the same time with important information on the critical ir singularities and thereby on critical exponents .
this observation lies at the heart of the field - theoretic rg method .
explicit calculations of the renormalization constants are facilitated if one first takes the continuum limit @xmath245 with @xmath208 ( _ dimensional regularization _ ) together with the requirement that the @xmath244 factors absorb just the @xmath246 poles ( _ minimal subtraction _ ) .
notice that this minimal dimensional regularization does not at all require an @xmath246 expansion .
yet the continuum limit raises subtle problems in the realm of statistical physics @xcite .
for example , one may conclude from the above remarks that the full range from zero to infinity of the bare coupling constant @xmath247 is mapped only to the interval @xmath248 $ ] which lies between the uv - stable fixed point @xmath249 and the ir - stable fixed point @xmath250 .
values of @xmath224 larger than @xmath240 are excluded . on the other hand , there exist many pre - asymptotic simulational and experimental results , e.g. , for the three - dimensional ising model near the critical point and some polymer systems , which are linked to this strong - coupling region that does not include an uv - stable fixed point . in wilson s
rg approach this regime can be reached by a choosing suitable initial values of the relevant and irrelevant couplings , and a finite uv cutoff @xmath73 .
the uv - renormalized field theory should therefore with @xmath245 should therefore be accepted as an _ effective _ field theory where the influences of irrelevant couplings and the finite cutoff are implicitly encoded in the full range of the coupling @xmath224 also above the fixed point .
next , we compare renormalized vertex functions from the same bare theory , but which are renormalized at two different external momentum scales @xmath251 and @xmath252 , respectively .
the bare vertex functions are held fixed , and we shall henceforth neglect the @xmath253 corrections .
then it follows from eq .
( [ ren - gg ] ) that [ mapping ] @xmath254 where @xmath255 and @xmath256 this is the desired mapping .
if @xmath257 , the left - hand site of eq .
( [ mapping ] ) is ir - problematic , whereas for an appropriate @xmath258 the vertex function on the right - hand site is unproblematic if we choose @xmath259 .
it is expected that @xmath260 , and the leading critical properties are transformed to the functions @xmath261 and @xmath262 . casting this mapping in differential form yields the _ renormalization group equation _ ( rge ) .
we let @xmath245 and define the rg functions as logarithmic derivatives with respect to the normalization scale @xmath251 , holding bare parameters fixed , [ rg - fu ] @xmath263 consider the renormalized green s functions @xmath264 \ , ; \ ] ] using the definitions ( [ rg - fu ] ) it is straightforward to derive the rge @xmath265 \ , g_{n,\tilde{n}}(\{\mathbf{r},t\},\tau , u,\lambda,\mu ) = 0 \ , .\ ] ] the solution of this partial differential equation is provided by the method of characteristics , which solve the ordinary differential equations [ char ] @xmath266 with the initial conditions @xmath267 , @xmath268 , @xmath269 , and @xmath270 .
this yields the general expression @xmath271 in the critical limit @xmath272 the characteristic equations provide us with the _
asymptotic scaling laws_. from eqs .
( [ char ] ) we learn that @xmath273 flows to a stable fixed point @xmath240 , given as a zero of the rg beta function @xmath274 , provided the first derivative @xmath275 . with the definitions @xmath276 and employing dimensional scaling , we obtain from eq .
( [ rg - sol ] ) the asymptotic form for the green s functions @xmath277 here , @xmath77 , @xmath278 , @xmath279 , and @xmath280 represent four _ non - universal _ amplitudes .
the symmetries ( [ zeitsp ] ) provide us with _ scaling relations _ between the critical exponents . for dp , eq .
( [ ren - z3 ] ) implies @xmath281 , whereas @xmath282 for dip .
thus we obtain @xmath283 the scaling functions @xmath284 are universal , and the exponents @xmath285 are given by [ delta ] @xmath286 consequently , there are only _ three _ independent scaling exponents @xmath287 , @xmath288 , and @xmath289 . in principle , a similar reduction may be found also for the number of independent non - universal amplitudes .
however , the rescaling ( [ redscal ] ) introduces an additional amplitude @xmath154 related to @xmath77 and @xmath278 by @xmath290 . in summary
, we have established that percolation processes near the critical point are asymptotically described by universal scaling functions with three scaling exponents , but four non - universal amplitudes .
the remaining task is to explicitly calculate the @xmath244 factors and the universal scaling functions in the non - critical region by means of perturbation theory .
there are two methods to avoid the non - calculable symanzik function .
either one changes variables from the parameter @xmath147 to the correlation length @xmath5 , see eq .
( [ korr - l ] ) , which is invariant under renormalization : @xmath291 , and then calculates perturbationally the derivative of @xmath147 by @xmath292 which eliminates @xmath187 @xcite .
alternatively , one may apply , in addition to the perturbation expansion , a dimensional @xmath246 expansion which formally sets @xmath187 to zero , see eq .
( [ tau_c ] ) .
directed percolation constitutes perhaps the simplest model of a strictly non - equilibrium system that displays a continuous phase transition . in this respect
its role is comparable with the paradigmatic ising model for equilibrium critical phenomena .
correspondingly , the dp field theory as given by the response functional ( [ jdp ] ) can be regarded as the non - equilibrium analog of the @xmath293 field theory . in the following sections we shall consider the renormalized dp field theory and the ensuing asymptotic scaling of important quantities in detail . as stated before , the perturbation expansion is arranged loop - wise with respect to the harmonic part of the response functional ( [ jdp ] ) . in the momentum - time representation , the propagator ( [ prop2 ] ) reads @xmath294 \ , , \ ] ] where the heaviside step function is defined with @xmath295 .
this follows from the it discretization of the path integral and ensures causality .
the anharmonic coupling terms in @xmath296 define the elements of the graphical perturbation expansion , depicted in fig .
[ elemente ] : an arrow marks a @xmath297-``leg '' , and we conventionally draw diagrams with the arrowss always directed to the left ( i.e. , we employ ascending time ordering from right to left ) . the perturbation series of the translationally invariant field theory can be analyzed through calculation of the vertex functions @xmath298 that correspond to the one - particle irreducible `` amputated '' graphs .
we are now ready to consider the renormalization of the dp field theory . to this end , we evaluate all diagrammatic contributions to the vertex functions to a given loop order by means of dimensional regularization , and subsequently absorb the uv divergences order by order in the loop expansion into a suitable renormalization of the fields and model parameters .
we choose the scheme ( [ ren - z ] ) and determine the renormalization constants in the minimal subtraction prescription . absorbing the @xmath246 poles of the naively divergent vertex functions into z factors in fact renormalizes the full theory .
the naively divergent vertex functions carry non - negative scaling dimensions . in the case of dp , these are @xmath299 and @xmath300 , schematically illustrated in fig .
[ vertfu ] . and @xmath301 in the dp field theory.,width=340 ]
we will now explicitly determine the renormalizations to one - loop order .
the primitively divergent one - loop graphs are shown in fig .
[ 1-loop ] .
we begin by expressing the one - loop contribution to the propagator self - energy , fig . [
1-loop](a ) , as a function of external momentum @xmath70 and frequency @xmath302 : @xmath303 here we have retained only terms linear in @xmath302 and @xmath304 .
these are the contributions that display poles in @xmath305 .
the first term in the brackets has an ir singularity at @xmath306 ( @xmath307 ) .
this pole can be eliminated by changing from the variable @xmath147 to the correlation length @xmath5 as an independent parameter . here , for simplicity , we just employ the @xmath246 expansion .
for the extraction of the divergences of the vertex function @xmath308 , the external momenta and frequencies may be set to zero .
the contribution from the diagram in fig .
[ 1-loop](b ) then becomes @xmath309 ^ 2 } = \frac{g_\varepsilon}{\varepsilon } \
, \mathring{\tau}^{-\varepsilon/2 } \ , \mathring{\lambda } \mathring{g}^{\,3 } \ , .\ ] ] combining the zero - loop expressions with the results of this short calculation , we obtain the one - loop vertex functions to the desired order in @xmath302 and @xmath304 : @xmath310 and @xmath311 in order to absorb the @xmath246 poles into renormalization factors , we employ the scheme ( [ ren - z ] ) , ( [ ren - gamma ] ) using @xmath312 . to one - loop order
we arrive at @xmath313 + \lambda \ , \mathbf{q}^2 \big [ z_\lambda - \frac{u}{8 \varepsilon } \left ( \frac{\mu}{\tau } \right)^{\varepsilon/2 } \big ] \nonumber \\ & \quad + \lambda \ , \tau \big [ z_\tau - \frac{u}{\varepsilon ( 2-\varepsilon ) } \left ( \frac{\mu}{\tau } \right)^{\varepsilon/2 } \big ] + \ldots \ , , \end{aligned}\ ] ] and @xmath314 \ , .\ ] ] therefore ,
the @xmath246 poles are eliminated by the minimal choices @xmath315 , @xmath316 , @xmath317 , and @xmath318 .
the two - loop graphs for the propagator self - energy and the three - point vertex are depicted in fig .
( [ 2-loop1 ] ) and ( [ 2-loop2 ] ) , respectively .
the ensuing explicit z factors read @xcite [ z - fakt ] @xmath319 the rg functions then follow directly from eqs .
( [ rg - fu ] ) , [ rg - fudp ] @xmath320 u \ , .
\label{beta - dp}\end{aligned}\ ] ] the ir - stable fixed point @xmath240 is determined via @xmath321 by means of the @xmath246 expansion , @xmath322\ , .\ ] ] as a consequence , the general asymptotic scaling results ( [ fp - werte ] ) and ( [ as - skal ] ) are finally found to carry the @xmath246-expanded special dp critical exponents [ dp - exp ] @xmath323 \ , , \label{eta - dp}\\ z & = 2 - \frac{\varepsilon}{12 } \big [ 1 + \left ( \frac{67}{288 } + \frac{59}{144 } \ln\frac{4}{3 } \right ) \varepsilon + o(\varepsilon^2 ) \big ] \ , , \label{z - dp}\\ \nu & = \frac{1}{2 } + \frac{\varepsilon}{16 } \big [ 1 + \left ( \frac{107}{288 } - \frac{17}{144 } \ln\frac{4}{3 } \right ) \varepsilon + o(\varepsilon^2 ) \big ] \ , , \label{nu - dp}\\ \beta & = \nu \ , \frac{d+\eta}{2 } = 1 - \frac{\varepsilon}{6 } \big [ 1 - \left ( \frac{11}{288 } - \frac{53}{144 } \ln\frac{4}{3 } \right ) \varepsilon + o(\varepsilon^2 ) \big ] \ , , \label{bet - dp}\end{aligned}\ ] ] with @xmath324 .
equipped with the important general asymptotic scaling results for the dp green s functions , which followed directly from the knowledge of the z factors , we are now in the position to determine the critical properties of dynamic as well as static observables .
using the asymptotic scaling form ( [ as - skal ] ) , the number @xmath325 of _ active particles _ generated by a seed at the origin is given by @xmath326 here we have introduced the exponents @xmath327 and the non - universal amplitude combinations @xmath328 @xmath329 is a universal scaling function normalized to @xmath330 .
as shown in ref .
@xcite , the _ survival probability _ @xmath331 of an active cluster emanating from a seed at the origin is asymptotically given by @xmath332 with @xmath333 , and where the green s function @xmath334 is calculated with the response functional @xmath335 .
note that because of the duality invariance ( [ zeitsp ] ) the survival probability @xmath331 is in fact related to the mean asymptotic density of the active particles @xmath336 for a process which starts with a homogeneous active density @xmath337 by @xmath338 recalling the scaling form ( [ as - skal ] ) of @xmath334 we deduce that @xmath339 with @xmath340 again , @xmath341 represents a universal function normalized to @xmath342 .
@xmath341 tends to zero exponentially for @xmath343 , and asymptotically to @xmath344 for @xmath345 , where @xmath346 is a universal constant .
thus , we find the percolation probability , for @xmath135 , @xmath347 the extension of a active cluster at time @xmath96 generated by a seed at the origin is measured by the _ radius of gyration _
@xmath348 of the active particles , as defined via @xmath349 from the asymptotic scaling law ( [ as - skal ] ) we infer @xmath350 whence for the radius of gyration we obtain the asymptotic scaling form @xmath351 with @xmath352 where again @xmath353 is a universal function normalized to @xmath354 .
the four quantities @xmath355 ,
@xmath356 , @xmath357 , and @xmath358 define a measurable complete set of non - universal amplitudes for any system belonging to the dp universality class . once these amplitudes are determined , all other observables attain universal values .
the last dynamic observable we consider here is the _ active density _ @xmath336 for finite initial @xmath337 , @xmath359 the initial density @xmath337 is introduced into the response functional ( [ jdp ] ) via a source @xmath360 .
no new initial renormalization @xcite is involved since the perturbation theory is based solely on the causal propagator ( the correlators do not enter ) , and initial correlations are irrelevant .
hence , according to the scheme ( [ ren - z ] ) , the renormalization of @xmath337 is given by @xmath361 which leads to an additional derivative term @xmath362 in the rge ( [ rge ] ) .
thus a new dependence on @xmath363 and @xmath364 arises in eqs .
( [ rg - sol ] ) and ( [ as - skal ] ) , respectively , which leads to @xmath365 where @xmath366 is universal with @xmath367 and @xmath368 .
the non - universal amplitudes are @xmath369 and the initial scaling exponent is @xmath370 a steady state with dp dynamics can be generated by introducing a homogeneous and time - independent external source of activity @xmath371 in the response functional ( [ jdp ] ) .
in such a steady state one can then measure static , time - independent observables , such as the mean density of activated particles and their fluctuations @xcite .
recall , however , that dp defines genuine non - equilibrium systems ; hence , there is no fluctuation - dissipation theorem that would relate the correlations of the order parameter to its response to an external conjugated field .
dynamic correlation and response functions therefore constitute independent observables . in order to obtain the _ equation of state _ ,
i.e. , the order parameter @xmath372 as a function of @xmath147 and @xmath371 in the steady state , we perform the variable shift @xmath373 and determine @xmath374 by the ` no - tadpole ' requirement @xmath375 . after this shift ,
the linear and harmonic part of the response functional become @xmath376 \tilde{s } \big\ } \ , .\end{aligned}\ ] ] as a consequence , aside from the propagator the perturbation expansion is now based on a correlator induced by the noise vertex @xmath377 .
we will present no details of the calculation here , but merely state the equation of state in a parametric form , as derived in a two - loop approximation in ref .
@xcite : @xmath378 where @xmath116 and @xmath115 are non - universal amplitudes that can be related to the previously introduced fundamental four amplitudes , and the exponent @xmath379 is given by @xmath380 the parameter @xmath381 , which measures the distance to the critical point , and the parameter range @xmath382 are required to describe the entire phase diagram in the critical region . as a simple application of this parametric representation we briefly discuss the susceptibility @xmath383 , which satisfies a power law for @xmath384 , @xmath385 with amplitudes @xmath386 and @xmath387 that correspond to the cases @xmath388 and @xmath389 , respectively .
we obtain @xmath390 therefore the universal amplitude ratio @xmath391 can be expressed to order @xmath392 in terms of the order parameter exponent @xmath393 as @xmath394 \ , .\ ] ] the susceptibility @xmath395 may also be represented by the integral over space and time of the green s function @xmath396 : @xmath397 in contrast , the mean - square fluctuation @xmath398 of the activated particle number are given by @xmath399 where the exponent @xmath400 is defined by eq .
( [ skalexp-1 ] ) and differs from @xmath401 .
a one - loop calculation yields @xmath402 with the non - universal parameter @xmath403 and @xmath404 . above the upper critical dimension @xmath0 , where mean - field theory is applicable , the coupling constant @xmath163 tends to zero under the renormalization group transformation .
however , @xmath163 represents a _ dangerously irrelevant variable _ here , since it scales various observables , and setting @xmath405 rigorously leads either to zero or infinity for relevant quantities . the twofold nature of @xmath163 as both a relevant scaling variable and an irrelevant loop - expansion generating parameter is lucidly exposed by writing the generating functional for the vertex functions ( [ vert - kum ] ) in the form @xmath406 = g^{-2 } \ , \phi[g \tilde{s},g s;\tau , u ] \ , .\ ] ] the expansion of the functional @xmath407 $ ] into a series with respect to @xmath408 yields the loop expansion .
the zeroth - order term @xmath409 $ ] is just the response functional ( [ j3 ] ) itself , i.e. , @xmath49 represents the mean - field contribution to the dynamic ` free energy ' @xmath410 .
the scaling form of the generating functional ( [ kumgen ] ) for the cumulants that corresponds to ( [ phi ] ) reads @xmath411 = g^{-2 } \omega[g h , g \tilde{h};\tau , u ] \ , .\ ] ] to leading order in the logarithmic corrections , we may neglect the dependence of @xmath412 and @xmath413 on @xmath224 .
( for the next - to - leading corrections see ref . @xcite . ) solving the characteristic equation ( [ char ] ) for @xmath414 at the upper critical dimension ( @xmath235 ) yields to leading order for @xmath272 : @xmath415 where we introduce the taylor expansion @xmath416 for any of the rg flow functions @xmath417 .
thus , from eq .
( [ beta - dp ] ) , @xmath418 for dp .
the remaining characteristics are all of the same structure , namely @xmath419 here , @xmath420 stands for @xmath262 , @xmath421 , and @xmath422 , respectively , whereas @xmath423 represents either @xmath401 , @xmath424 , or @xmath425 , as defined in eq .
( [ rg - fu ] ) . to leading order , eq .
( [ char4 ] ) is solved by @xmath426 combining everything we obtain asymptotically @xmath427 \simeq |\ln l| \nonumber
\\ & \qquad \times \omega\big [ |\ln l|^{-5/12 } h , |\ln l|^{-5/12 } \tilde{h } ; |\ln l|^{-1/4 } \tau,0 ; |\ln l|^{1/12 } \lambda , l \mu \big ] \ , , \end{aligned}\ ] ] where the non - universal amplitudes have been absorbed into the variables . taking the required functional derivatives at @xmath428 , @xmath429 and
employing dimensional scaling then yields the asymptotic green s functions with the logarithmic scaling corrections at @xmath0 : @xmath430 from this general result we find for the dynamical observables [ log - korr - dyn ] @xmath431 the logarithmic correction for the percolation probability becomes @xmath432 and the equation of state and the fluctuations in the steady state read to leading order [ log - korr - stat ] @xmath433 \ , \big\vert \ln\sqrt{\tau^2 + b h } \big\vert^{1/3 } \
, , \label{log - korr -eqs}\\ \chi^\prime & = a^\prime \bigg [ 1 - \frac{\tau}{\sqrt{\tau^2 + b h } } \bigg ] \ , \big\vert \ln\sqrt{\tau^2 + b h } \big\vert^{1/6 } \ , , \label{log - korr - fluc}\end{aligned}\ ] ] where @xmath77 , @xmath434 , @xmath79 , and @xmath435 are non - universal amplitudes .
we end this section by discussing a method that allows the explicit calculation of _ finite - size effects _ in dp @xcite .
we will consider the model defined by the action ( [ jdp ] ) , in a finite cubic geometry of linear size @xmath436 with periodic boundary conditions .
expanding the fields in fourier modes @xmath437 where each component of @xmath70 only assumes discrete values , namely multiples of @xmath438 , it is clear that the @xmath439-mode can not be treated perturbatively at the critical point @xmath440 , since the propagator displays an isolated pole at @xmath441 .
therefore , in order to evaluate finite - size effects , one has to construct an effective dynamic response functional for the @xmath439-mode ( which subsequently has to be treated non - perturbatively ) by tracing out all modes with @xmath442 . upon decomposing the fields [ sep ] @xmath443 one can separate the @xmath439-modes @xmath444 and @xmath445 from their orthogonal complements @xmath446 and @xmath447 .
after performing this decomposition in the action @xmath296 we obtain [ decomp ] @xmath448 \phi - \lambda \ , h \ , \tilde{\phi } \biggr\ } \ , , \label{j-0}\\ \mathcal{j}_{\rm dp}^{(1 ) } & = \int \ ! dt \ , d^dr \,\biggl\ { \tilde{\varphi } \bigg [ \partial_t + \lambda ( \tau - \nabla^2 ) + \frac{\lambda g}{2 } ( \phi - \tilde{\phi } ) \bigg ] \varphi + \frac{\lambda g}{2 } \bigl ( \tilde{\phi } \varphi^2 - \phi \tilde{\varphi}^2 \bigr ) \biggr\ } \ , , \label{j-1}\\ \mathcal{j}_{\rm dp}^{(2 ) } & = \frac{\lambda g}{2 } \int \ ! dt \ , d^dr \ , \tilde{\varphi } \bigl ( \varphi - \tilde{\varphi } \bigr ) \varphi \ , .
\label{j-2}\end{aligned}\ ] ] integrating over all modes with @xmath442 we arrive at the _ effective action _
@xmath449 \exp\bigl(-\mathcal{j}_{\rm dp}^{(0)}- \mathcal{j}_{dp}^{(1)}-\mathcal{j}_{\rm dp}^{(2 ) } \bigr ) = \mathcal{j}_{\rm dp}^{(0)}[\tilde{\phi},\phi ] + \sigma[\tilde{\phi},\phi]\ , .\ ] ] the contribution @xmath450 $ ] can now be analyzed perturbatively by means of a double expansion in powers of the fields @xmath451 and @xmath413 that arise in the vertices of @xmath452 , and in the number of loops due to insertion of extra vertices originating from @xmath453 . up to and including terms of third order in @xmath451 , @xmath413 , and after renormalization
, one finds @xcite @xmath454 \phi - \lambda \ , h \ , \tilde{\phi } \biggr\ } \ , .\ ] ] in dimensions @xmath455 , the functions @xmath456 , @xmath457 , and @xmath458 display scaling properties that follows from the rge . for @xmath459 ,
the contributions stemming from @xmath450 $ ] can be neglected asymptotically , whence @xmath460 , @xmath461 , and @xmath462 .
we proceed by rescaling the fields and the time scale according to [ fp - skal ] @xmath463 and thereby obtain for the effective response functional @xmath464 m - b \
, \tilde{\phi } \biggr\ } \ , , \ ] ] where the two parameters @xmath116 and @xmath115 are given by [ fp - parame ] @xmath465 and @xmath466 for @xmath459 , we have @xmath467 and @xmath468 .
hence , besides the finite - size scaling of the control parameter @xmath147 and the source @xmath371 , an additional shift @xmath469 results from the elimination of the @xmath470-modes .
however , above four dimension this shift is proportional to @xmath471 and represents the leading correction .
the green s functions for the spatially averaged variables @xmath444 and @xmath445 are now calculated with the response functional @xmath472 $ ] , eq .
( [ funk - fp ] ) .
consequently @xmath473^n [ \tilde{\phi}]^{\tilde{n } } \rangle^{({\rm cum } ) } = \alpha^{n+\tilde{n } } f_{n,\tilde{n}}(\ { 2s \},a , b ) \ , , \
] ] and above four dimension , i.e. , in the mean - field region , we obtain asymptotically @xmath474 the scaling functions @xmath284 can be evaluated by an alternative approach . via the equivalence of the response functional ( [ funk - fp ] ) with the it - langevin equation , which in turn is equivalent to a corresponding fokker - planck equation for the probability density @xmath475 of the stochastic variable @xmath372 , we obtain @xmath476 \mathcal{p}(m , s ) \big\ } + \frac{\partial^2}{\partial m^2 } \big [ m \mathcal{p}(m , s ) \big ] \ , .\ ] ] in particular , the stationary solution @xmath477 for @xmath478 is easily found , @xmath479 \ , , \ ] ] where @xmath435 represents a finite normalization constant .
all the moments of the averaged agent density @xmath480 can now be expressed in terms of parabolic cylinder functions . in particular , for @xmath459 , and at criticality ( @xmath481 ) , using @xmath482 , one arrives at @xmath483 where @xmath484 and @xmath485 are non - universal amplitudes , and @xmath486 denotes euler s gamma function .
specifically , binder s cumulant @xmath420 is given by the simple expression @xmath487 this function has been measured to high precision in recent simulations @xcite .
we consider now the field theory of dynamic isotropic percolation .
we are interested in the dynamic as well as static ( emerging as @xmath131 ) properties of isotropic percolation .
as mentioned earlier , the static debris clusters after the activation process have ceased are described by the usual static percolation theory near the critical threshold .
of course , we could base our entire rg analysis on the full dynamic functional @xmath488 as given in eq .
( [ jdip ] ) , and extract the static from the full dynamic behavior by taking the limit @xmath131 in the end .
this would imply , however , that we would have to determine all the required renormalizations from the dynamic feynman graphs , composed of the diagrammatic elements encoded in the action @xmath488 .
fortunately , there is a considerably more economic approach which is based on taking the so - called _ quasi - static limit_.
imagine that we initiate a process through external activity sources @xmath489 , localized in time at @xmath67 ; but we are interested only at green s functions of the debris at @xmath490 .
we will see shortly that the perturbation expansion simplifies tremendously in this limit .
all renormalization factors but one can be calculated directly using this much simpler method .
thus , only for the single remaining renormalization do we have to resort to the full dynamic response functional @xmath488 .
taking the quasi - static limit amounts to switching the fundamental field variable from the agent density to the final density of debris @xmath491 that is ultimately left behind by the epidemic , and the associated response field @xmath492 .
in particular , the green s functions corresponding to the correlation functions @xmath493 will turn out to be important for our analysis , since they actually encode the static properties of the debris percolation cluster emanating from a seed that is localized at the origin at time @xmath67 . following up on the previous remarks , we now proceed to formally take the quasi - static limit of the dynamic functional for dip .
the structure of @xmath494 allows us to directly let @xmath495 this procedure leads us from the action ( [ jdip ] ) directly to the _ quasi - static hamiltonian _ with source @xmath496 @xmath497 \varphi - \ , k \tilde{\varphi } \big\ } \ , .\ ] ] by just considering the involved time integrations from ` left ' ( i.e. , the largest time involved ) to ` right ' ( the smallest time ) in any graph , it is easy to see that @xmath498 in fact generates all diagrams contributing to @xmath499 . by itself , however , the hamiltonian ( [ hamilt ] ) is insufficient to describe the static properties of isotropic percolation ( ip ) . as a remnant of its dynamical origin ,
@xmath498 must be supplemented with the causality rule that forbids closed propagator loops .
the propagator of the quasi - static theory immediately follows from the hamiltonian , @xmath500 the vertices contained in eq .
( [ hamilt ] ) are actually identical to those of the dp field theory , whence the elements of the perturbation expansion coincide with those depicted in fig .
[ elemente ] ( with the changed propagator ( [ statprop ] ) and @xmath501 ) . correspondingly , determining the renormalization constants of the quasi - static theory proceeds as familiar from dp , and is based on the same diagrams , see figs .
[ 1-loop ] , [ 2-loop1 ] , and [ 2-loop2 ] .
noticing that our conventions ( [ ren - z ] ) imply that @xmath502 and @xmath503 , we explicitly find to two - loop order [ z - ip ] @xmath504 these @xmath244 factors coincide with those calculated in ref .
@xcite for the potts model in the single - state limit .
the renormalization constants are known to three - loop order @xcite .
the rg functions appearing in the rge read to two - loop order @xcite [ rg - fudip ] @xmath505 u \ , .
\label{beta - dip}\end{aligned}\ ] ] the ir - stable fixed point , determined as zero of @xmath241 , is @xmath506 \ , .\ ] ] thus we recover the well - known critical exponents for isotropic percolation : [ dip - exp ] @xmath507 \ , , \label{eta - dip}\\ \nu & = \bigl [ 2 - \kappa(u_\ast ) \bigr]^{-1 } = \frac{1}{2 } + \frac{5\varepsilon}{84 } \big [ 1 + \frac{589}{2205 } \ , \varepsilon + o(\varepsilon^2 ) \big ] \ , , \label{nu - dip}\\ \beta & = \nu \ , \frac{d-2+\eta_p}{2 } = 1 - \frac{\varepsilon}{6 } \big [ 1 + \frac{61}{1764 } \
, \varepsilon + o(\varepsilon^2 ) \big ] \ , .
\label{bet - dip}\end{aligned}\ ] ] let @xmath508 be the measure for the probability that the cluster mass of the debris generated by a seed at the origin is between @xmath128 and @xmath509 .
we obtain @xmath510 \right\rangle \ , .\ ] ] for the probability density @xmath511 of large clusters with @xmath512 we may use the expansion of the exponential to first order ( higher orders asymptotically only lead to subleading corrections ) and find @xmath513 the _ percolation probability _
@xmath514 is defined as the probability for the existence of an infinite cluster generated from a single seed .
hence @xmath515 \right\rangle \ , . \end{aligned}\ ] ] via expanding @xmath516 $ ] we arrive at the asymptotic form @xcite @xmath517 where @xmath518 .
the virtue of this formula is that it relates the percolation probability in an unambiguous manner to an expression accessible by field theory . for actual calculations the term
@xmath519 needs to be incorporated into the quasi - static hamiltonian ; i.e. , we must replace the original @xmath498 with @xmath520 here , @xmath521 plays the role of a source conjugate to the field @xmath522 .
whereas in general @xmath523 by causality if @xmath524 , the limit @xmath525 leads to a non - vanishing order parameter @xmath514 in the active phase with spontaneously broken symmetry . in terms of averages
@xmath526 with respect to the new hamiltonian ( [ newhamilt ] ) , we may now write @xmath527 with the aid of eq .
( [ perkprob1 ] ) and the scaling form ( [ as - skal ] ) we readily obtain that @xmath528 in order to examine the scaling behavior of @xmath511 , we consider its moments @xmath529 using eq .
( [ p - as ] ) , our general scaling result ( [ as - skal ] ) implies @xmath530 which tells us that @xmath531 where @xmath532 is the number of clusters of size @xmath128 per lattice site .
these _ cluster numbers _
@xmath532 play an important role in percolation theory , which conventionally employs the scaling exponents @xmath533 and @xmath534 as defined in eq .
( [ pskal ] ) . in terms of our earlier critical exponents , these
are @xmath535 it follows from eq .
( [ moments ] ) that the mean cluster mass @xmath536 ( taken as average of the finite clusters ) scales as @xmath537 next we consider correlation functions restricted to clusters of given mass @xmath128 . in terms of the conventional unrestricted averages with respect to @xmath498 , these restricted correlation functions can be expressed for large @xmath128 as @xmath538 their scaling from can again be read off from eq .
( [ as - skal ] ) , @xmath539 by means of these restricted green functions we can write the _ radius of gyration _ ( i.e. , the mean - square cluster radius ) of clusters of size @xmath128 as @xmath540 whence with eq .
( [ korrelfu ] ) @xmath541 with the _ fractal dimension _
@xmath542 we conclude this section by considering the scaling behavior of the debris statistics if the initial state is prepared with a homogeneous seed density @xmath543 .
as discussed above , at the level of the quasistatic hamiltonian @xmath498 such an initial state translates to a further additive contribution @xmath544 .
our general scaling form ( [ as - skal ] ) implies that the correlation functions of the densities @xmath545 for the case of a homogeneous initial condition behave as @xmath546 it is obvious that the initial seed density @xmath371 plays the role of an ordering field .
it corresponds to the ghost field of conventional percolation theory @xcite .
hence , the green functions do not display critical singularities as long as @xmath547 . for a homogeneous initial condition
the appropriate order parameter is given by the debris density @xmath548 in the non - percolating phase ( @xmath134 ) , the order parameter @xmath372 is linear in @xmath371 for small seed density , with a susceptibility coefficient that diverges as @xmath549 , @xmath550 with the susceptibility exponent @xmath401 given in eq .
( [ meanmass ] ) . at criticality ( @xmath139 )
the order parameter @xmath372 tends to zero for @xmath384 as @xmath551 with the exponent @xmath552 finally , in the percolating phase ( @xmath135 ) the order parameter becomes independent of the initial seed density in the limit @xmath384 and tends to zero with @xmath147 according to @xmath553 which of course is just eq .
( [ meanmass ] ) .
( [ probinf ] ) and ( [ pdichte ] ) establish explicitly that the two distinct order parameters , namely the debris density @xmath372 and the percolation probability @xmath514 , both scale with the identical exponent @xmath393 near criticality . at this stage
we are in a position to discuss _ crossovers _ between the two fundamental percolation transitions @xcite .
we have established earlier that the introduction of the memory term @xmath554 in the response functional ( [ j2 ] ) changes scaling behavior from dp to dip .
hence , memory is clearly a relevant perturbation ( in the rg sense ) in dp and the process described by @xmath296 asymptotically crosses over to the process described by @xmath488 , both in @xmath2 spatial and the single temporal dimension .
we have also argued above that the static asymptotic behavior of isotropic percolation in @xmath2 dimensions is described by the quasi - static hamiltonian @xmath498 .
if we now imagine a forest fire as a special realization of isotropic percolation and introduce a strong wind blowing in a preferred direction @xmath555 , such a directional disturbance may be described by an additional term @xmath556 in eq .
( [ hamilt ] ) . upon comparing with the longitudinal part @xmath557 of the laplacian contribution
, we see that this additional directional term is more relevant for the ir behavior .
that is , if we scale the transverse part of the laplacian as @xmath558 , the ir - scaling dimensions become @xmath559 and @xmath560 . as a consequence ,
the contribution @xmath561 is asymptotically irrelevant , and the quasistatic hamiltonian becomes @xmath562 \varphi - k \
, \tilde{\varphi } \big\ } \ , , \ ] ] where @xmath563 is a new parameter that describes the directionality of the percolation in @xmath2 dimension .
it is now straightforward to show by means of simple rescaling that by identifying the longitudinal spatial direction with ` time ' and the @xmath564 transverse subspace directions with a new space , @xmath565 transforms into the response functional for dp , namely @xmath296 , in @xmath564 spatial and @xmath9 temporal dimension .
thus , we have established the following schematic crossover scenarios : @xmath566{lllllll } \
\text{dp } & & \ \text{dip } & & \ \text{ip } & & \ \text{dp } \\
\mathcal{j}_{\rm dp } & \quad\underrightarrow{\text{memory } } \quad & \mathcal{j}_{\rm dip } & \quad\underrightarrow { \begin{array}[c]{c } \text{quasistatic } \\ \text{limit } \end{array } } \quad & \mathcal{h}_{\rm ip } & \quad\underrightarrow{\text{directionality } } \quad & \mathcal{j}_{\rm dp } \\
d+1 & & d+1 & & \ d & & ( d-1)+1 \end{array } $ ] we note that one has to apply some care when trying to capture these crossover scenarios by means of the dynamic rg , since both the transitions from dp to dip as well as from ip to dp involve different upper critical dimensions .
therefore a mere @xmath246 expansion about the @xmath165 of either theory can not possibly access the opposite scaling limit .
one can , however , work out the rg flow functions that incorporate the entire crossover regimes at fixed dimension @xmath2 , provided one employs the correlation length @xmath5 as independent variable rather than @xmath147 in order to eliminate ir singularities . for more details on this method to describe the crossover from isotropic to directed percolation ,
see refs .
@xcite .
in order to investigate scaling properties of genuinely dynamical observables of dip , one must resort to the full response functional ( [ jdip ] ) .
for the determination of the last independent renormalization factor @xmath244 to two - loop order , one needs to evaluate self - energy diagrams such as fig . [ 1-loop](a ) and fig .
[ 2-loop1 ] , but now with one temporally delocalized leg of the vertex corresponding to the coupling @xmath567 in @xmath488 .
we thus obtain from the renormalization of the derivative @xmath568 @xcite @xmath569 in combination with eqs .
( [ z - ip1 ] ) and ( [ ren - z ] ) , eq .
( [ rg - fu ] ) then yields the additional rg function @xmath570 inserting the fixed point value ( [ fixpt - dip ] ) into eq .
( [ fp - werte ] ) , we find for the dynamic exponent @xmath571 \frac{\varepsilon^2}{36 } + o(\varepsilon^3 ) \ , .\ ] ] the scaling form of the survival probability is equal to the expression ( [ p(t ) ] ) , with the spreading exponent @xmath572 and the longitudinal correlation exponent @xmath573 that follow from the fundamental exponents listed in eqs .
( [ dip - exp ] ) and ( [ z - dip ] ) .
likewise , the radius of gyration is given by eq .
( [ r(t ) ] ) with the spreading exponent @xmath574 \frac{\varepsilon^2}{144 } + o(\varepsilon^3 ) \ , .\ ] ] remarkably , this two - loop approximation for @xmath575 differs only by @xmath9-@xmath576 from simulation results obtained in the physical dimensions @xmath577 and @xmath578 @xcite .
the number of active particles generated by a seed becomes @xmath579 the active density @xmath580 at time @xmath96 , initialized by a finite homogeneous density @xmath337 , is @xmath581 with an initial scaling exponent @xmath582
in this overview , we have studied the field theory approach to percolating systems . based on the fundamental and universal features of the simple and general epidemic processes ,
we have constructed a mesoscopic description in terms of stochastic equations of motion , which we subsequently represented through a path integral with the dynamic response functional serving as the appropriate effective action . we have also commented on a more microscopic representation that starts from the classical master equation of a specific realization of such processes . in the bulk of this paper , we have provided a detailed description of the analysis of the ensuing stochastic field theories , from basic scaling properties to the perturbation expansion and uv renormalization , and explained why and how one may therefrom infer the correct asymptotic ir scaling behavior .
we have derived a number of scaling relations for the critical exponents of directed ( dp ) and dynamic isotropic percolation ( dip ) , and explicitly demonstrated how these are linked to the large - scale , long - time properties of numerous static and dynamic observables . for the case of dip
, we have also derived the effective quasistatic field theory , which yields the scaling exponents of isotropic percolation .
we remark that non - perturbative rg methods have recently been applied to the dp field theory as well @xcite . naturally , there are various possible extensions of the above models , some of which lead to novel critical properties .
we end this review with a brief discussion of some interesting modifications of the standard percolation processes . in the standard version of
the percolation processes in the language of an infectious disease , the susceptible individuals can become contaminated by already sick _ neighboring _ individuals . at the same time sick individuals are subject to spontaneous healing or immunization . in more realistic situations ,
however , the infection could be also long - ranged . as an example , envision the spreading of a disease in an orchard where flying parasites contaminate the trees practically instantaneous in a widespread manner , provided the time scale of the parasites flights is much shorter than the mesoscopic time scale of the epidemic process itself . thus following a suggestion by mollison @xcite ,
grassberger @xcite introduced a variation of the epidemic processes with infection probability distributions @xmath583 which decay with the distance @xmath584 according to a power law .
we will refer to such long - range distributions as _ lvy flights _ , although a true lvy flight is defined by the fourier transform , @xmath585 @xcite , and only lvy exponents in the interval @xmath586 give rise to positive distributions @xcite .
the spreading probability in this situation is rendered non - local , @xmath587 after fourier transformation of this equation , and after applying a small momentum expansion , we arrive at @xmath588 n(\mathbf{q},t ) \
, , \ ] ] where the analytical terms stem from the short - range part of @xmath589 , whereas the non - analytical contributions arise from the power - law decay .
the constant @xmath590 is included in the reaction rate as a negative ( `` birth '' ) contribution to @xmath147 , while @xmath591 represents a diffusional term . in order to decide which of the terms in eq .
( [ fouinfr ] ) are relevant , one has to compare with the scaling behavior of the fourier - transformed susceptibility @xmath592 , where @xmath593 denotes the anomalous field dimension within the short - range spreading theories defined by the response functionals ( [ j3 ] ) , i.e. , @xmath593 is given by eqs .
( [ eta - dp ] ) and eq .
( [ eta - dip ] ) , respectively .
if @xmath594 , the parameter @xmath595 constitutes a relevant perturbation and must be included in a renormalization group procedure . upon taking the leading non - analytical term in eq .
( [ fouinfr ] ) into account , the harmonic part of the response functionals ( [ j3 ] ) changes to @xmath596 = \int \ ! d^dr \ , dt \
, \big\ { \tilde{s } \big ( \partial_t + \lambda \big [ \tau - \nabla^2 + f \bigl ( -\nabla^2 \bigr)^{\sigma/2 } \big ] \big ) s \big\ } \ , .\ ] ] the mathematical treatment of this field theory with both gradient terms is somewhat delicate .
therefore , dp and dip with long - range spreading were originally studied via wilson s momentum shell rg method , since this approach can handle relevant and irrelevant contributions on equal footing @xcite . within a one - loop calculation
, it was shown that the spreading is dominated by the lvy flights for @xmath594 , and the scaling exponents change continuously to their short - range counterparts at @xmath597 .
the application of renormalized field theory has to distinguish between two separate cases .
if @xmath598 , one must apply a double expansion in @xmath246 and @xmath599 , following the work by honkonen and nalimov @xcite .
thereby the renormalizations were obtained to two - loop order @xcite , and the one - loop results indeed corroborate the findings within the momentum shell approach with respect to the crossover to short - range percolation . in the case
@xmath600 , the diffusional term becomes ir - irrelevant , and must be removed in order to obtain a uv - renormalizable field theory in the infinite - cutoff limit .
then , by rescaling of time , @xmath601 can be set to @xmath9 . the usual scaling @xmath602 yields @xmath603 , and @xmath604 , where @xmath605 with @xmath606 for dp and @xmath607 , @xmath608 with @xmath609 for dip .
thus we infer @xmath610 and @xmath611 , respectively , to be the upper critical dimensions .
the propagator is now @xmath612 $ ] .
the vertex functions are analytical functions of the external momenta and frequencies as long as @xmath134 . thus , the non - analytic lvy flight term in eq .
( [ harm - lr ] ) proportional to @xmath613 does not require renormalization , whence besides @xmath312 as usual we find for dp that @xmath614 exactly , while for long - range dip @xmath615 .
it turns out that counterterms are only needed for vertex functions with zero external momenta .
the following identity @xmath616 where @xmath617 , is useful for the explicit computation of the @xmath244 factors , from which subsequently the rg functions ( [ rg - fu ] ) are found .
to one - loop order for long - range dp , the result is [ rg - fulrdp ] @xmath618 u \
, , \label{rg - fulrdp1 } \\
\kappa(u ) & = \frac{u}{2 } + o(u^2 ) \ , , \quad \tilde{\gamma}(u ) = \gamma(u ) = \zeta(u ) = - \frac{u}{4 } + o(u^2 ) \ , , \label{rg - fulrdp2}\end{aligned}\ ] ] whereas for long - range dip [ rg - fulrdip ] @xmath619 u \ , , \label{rg - fulrdip1}\\ \kappa(u ) & = u + o(u^2 ) \ , , \quad \tilde{\gamma}(u ) = 0 \ , , \quad \gamma(u ) = 2 \zeta(u ) = - \frac{3u}{2 } + o(u^2 ) \ , .
\label{rg - fulrdip2}\end{aligned}\ ] ] at the long - range dp and dip fixed points @xmath620 and @xmath621 , respectively , we then get the critical exponents [ lr - exp ] @xmath622 within the field theory formulation , one can also investigate the influence of a spatial boundary on critical behavior @xcite .
generally , for percolation processes four possible scenarios can be envisioned : the boundary remains inactive , whereas the bulk is critical ( _ ordinary _ transition ) , the boundary is active , the bulk is critical ( _ extraordinary _ transition ) , the boundary is critical , but the bulk inactive ( _ surface _ transition ) , or both boundary and bulk are critical , which represents a multicritical point ( the _ special _ transition ) @xcite .
let us consider a semi - infinite geometry , with a boundary plane at @xmath623 ; in this situation we need to supplement the dynamic response functionals ( [ j3 ] ) with the surface action @xmath624 and impose the boundary condition @xmath625 .
naive power counting yields that the new parameter @xmath626 is relevant , whence its rg fixed points are either @xmath627 or @xmath628 .
the ordinary transition scenario corresponds to @xmath629 and the special transition to @xmath630 .
the presence of the boundary implies a different scaling behavior of the fields near the surface as compared to the bulk .
for example , the surface order parameter acquires new critical exponents [ surfexp ] @xmath631 one may readily generalize the previous mesoscopic description of the simple epidemic process to multiple activity carriers @xmath632 in order to capture , say , a variety of population species near an extinction threshold . in the spirit of sec .
2.1 one thus arrives at the coupled dp langevin equations @xcite [ multlange ] @xmath633 + \zeta_\alpha \ , , \label{multl1}\\ r_\alpha[\ { n_\alpha \ } ] & = - \lambda_\alpha \
, \bigl ( \tau_\alpha + \frac{1}{2 } \sum_\beta g_{\alpha \beta } n_\beta + \ldots \bigr ) \ , , \label{multdens}\end{aligned}\ ] ] with the stochastic noise correlations @xmath634 although this coupled multispecies systems appears to be very rich , it turns out that in fact all the ensuing renormalizations are given precisely by those of the single - species process , and hence the critical behavior at the extinction threshold is quite remarkably just that of dp again @xcite .
in addition , this generically universal model displays an instability that asymptotically leads to unidirectionality in the interspecies couplings .
a special situation arises when several control parameters @xmath635 vanish simultaneously , which implies _ multicritical _ behavior of unidirectionally coupled dp processes @xcite . in the active phase ,
one then finds a hierarchy of order parameter exponents @xmath636 with @xmath637 in addition , one can show that the crossover exponent associated with the multicritical point is @xmath638 to all orders in perturbation theory @xcite
. there remains , however , an unresolved technical issue originating from the emergence of a relevant coupling that enters the perturbation expansion @xcite .
quite analogous features also characterize multispecies generalizations of the general epidemic process or coupled dip processes .
several years ago kree , schaub , and schmittmann @xcite introduced a model that consists of the two - species reaction - diffusion system @xmath639 , @xmath640 , and @xmath641 with unequal diffusion constants species @xmath79 and @xmath435 .
the active particles @xmath79 , whose density we set proportional to @xmath642 , become poisoned by a diffusing conserved quantity @xmath435 with density @xmath643 . in a mesoscopic description ,
this kss - model is described by the coupled langevin equations [ beweg ] @xmath644 with positive parameters @xmath14 , @xmath401 , and @xmath163 .
the stochastic forces @xmath645 must respect the absorbing state condition as well as the conservation property . hence their gaussian fluctuations are given by [ lang ] @xmath646
more recently , van wijland , oerding , and hilhorst studied the two - species reaction - diffusion system @xmath647 and @xmath648 with unequal diffusivities @xcite . here , the total number of @xmath77 and @xmath79 particles constitutes a conserved quantity @xmath435 .
elimination of the @xmath77 density in favor of the density of @xmath435 then recovers the stochastic equations of motion ( [ beweg ] ) , ( [ lang ] ) , yet now with a cross - diffusion term in eq .
( [ beweg_c ] ) : @xmath649 where @xmath650 is proportional to the difference of the diffusion constants of the @xmath79 and @xmath77 species .
both the kss and the woh model represent generalizations of dp via including the influence of a non - critical conserved quantity in the critical dynamics of the agent , akin to the generalization of the relaxational model a of near - equilibrium critical dynamics to model c @xcite .
therefore we propose the label dp - c for this modification of the dp universality class .
following the same lines that resulted in the response functional ( [ j2 ] ) , and omitting ir - irrelevant terms , we obtain the renormalizable response functional of the dp - c class corresponding to the above langevin equations , @xmath651 s \nonumber\\ & \qquad\qquad\quad + \tilde{c } \bigl [ \partial_t c - \gamma \nabla^2 ( c - \sigma s ) \bigr ] - \tilde{\gamma } \ , ( \nabla \tilde{c})^2 \biggr\ } \ , .\end{aligned}\ ] ] our first observation concerns the stability of dp - c .
consider the mean - field approximation for the stationary state .
( [ cross - diff ] ) implies that @xmath652 . including the constant @xmath653 in the variable @xmath147
, we arrive at @xmath654 which demonstrates that the stable homogeneous solution @xmath655 for positive @xmath147 and @xmath656 becomes unstable for @xmath657 .
higher orders of @xmath658 should then be included in the density expansion , and one would expect the transition to become dicontinuous or first - order .
a continuous second - order transition therefore requires that the constraint @xmath659 be satisfied . an other qualitative view on this instability is illustrated by the following consideration .
assume @xmath660 in the following , i.e. , the density @xmath563 operates as an inhibitor .
if now an enhancement of @xmath658 is created by a fluctuation in some region of space the current contribution @xmath661 shows that the inhibitor flows into this region if @xmath662 and subsequently reduces the fluctuation .
however , if @xmath663 , the inhibitor is reduced by the flow out of this region , and the fluctuation of @xmath658 becomes increasingly enhanced .
rescaling the fields @xmath563 , @xmath664 , and the parameter @xmath650 , we may set @xmath665 .
the response functional ( [ jdp - c ] ) possesses the following symmetries under three transformations that involve constant continuous parameters @xmath666
: [ symm - dpc ] @xmath667 moreover , @xmath49 is invariant under the inversion @xmath668 in the particular case @xmath669 , the time reflection @xmath670 yields an additional discrete symmetry transformation .
the invariance with respect to the symmetry v distinguishes the special kss model from the general dp - c class .
symmetry i results from the conservation property of the field @xmath563 .
symmetries iii and iv show that dimensionless invariant coupling constants and parameters are defined by @xmath671 , @xmath672 , @xmath673 , and the ratio of the kinetic coefficients @xmath674 with @xmath305 .
symmetry iii implies that the response functional @xmath49 again contains a redundant parameter which can be fixed in different ways .
dimensional analysis and the scaling symmetry iii applied to the green functions @xmath675^n [ \tilde{s}]^{\tilde{n } } [ c]^m [ \tilde { c}]^{\tilde{m } } \rangle$ ] gives @xmath676 wherein the uv singularities and critical properties reside in the functions @xmath677 and @xmath678 , respectively . if @xmath669 , only eq .
( [ skalg ] ) can be used , and it is natural to apply a rescaling that leads to @xmath679 , which is fixed also under renormalization by the time inversion symmetry v. however , this is not the case in the general situation @xmath680 , whence it is more natural to absorb @xmath650 into the fields @xmath658 and @xmath297 .
in other words , one may then hold @xmath681 constant under renormalization .
it is easily seen that loop diagrams do not contribute to the vertex functions @xmath682 with @xmath683 , whence we infer @xmath684 , @xmath685 , @xmath686 , and @xmath687 .
we define the remaining renormalizations via [ ren - dpc ] @xmath688 where the last setting can be implemented through an appropriate choice for @xmath689 . for @xmath680 , the non - renormalization of @xmath658 ( @xmath690 )
follows from @xmath681 .
if @xmath669 , we have @xmath691 .
symmetry ii in connection with the trivial renormalization of @xmath563 shows that @xmath601 is renormalized with the same @xmath244 factor as @xmath147 : @xmath692 .
the rge ( [ rge ] ) now contains four gell - mann low functions @xmath693 corresponding to the four dimensionless parameters @xmath224 , @xmath694 , @xmath695 , and @xmath480 .
a one - loop calculation gives [ beta - dpc ] @xmath696 u \nonumber\\ & \qquad + \big [ \frac{u}{1 + \rho } - \frac{4v}{\rho ( 1 + \rho ) } + \frac{2w}{\rho ( 1 + \rho ) } + o(\text{2-loop } ) \big ] w \ , , \label{beta - dpc - u}\\ \beta_v & = ( - \varepsilon + 2 \kappa ) v \nonumber\\ & = \big [ - \varepsilon + \frac{3u}{4 } - \frac{4v}{(1 + \rho)^3 } + \frac{(9 + 8 \rho + 3 \rho^2 ) w}{2 ( 1 + \rho)^3 } + o(\text{2-loop } ) \big ] v \ , , \label{beta - dpc - v}\\ \beta_w & = \big [ - \varepsilon + u - \frac{2 ( 3 + 2 \rho + \rho^2 ) v}{(1 + \rho)^3 } + \frac{(5 + 5 \rho + 2 \rho^2 ) w}{(1 + \rho)^3 } + o(\text{2-loop } ) \big ] w \ , , \label{beta - dpc - w}\\ \beta_\rho & = - \zeta \rho = \big [ \frac{u}{8 } - \frac{2v}{(1 + \rho)^3 } + \frac{(7 + 4 \rho + \rho^2 ) w}{4 ( 1 + \rho)^3 } + o(\text{2-loop } ) \big ] \rho \ , .
\label{beta - dpc - rho}\end{aligned}\ ] ] the rg flow functions leading to the anomalous dimensions of the fields @xmath658 and @xmath297 are found to be [ gam - dpc ] @xmath697{lll } \tilde{\gamma } \quad & \text{if } & \ \sigma = 0 \\ 0 & \text{if } & \
\sigma \neq 0 \end{array } \right . \ , , \label{gam - dpc1}\\ \gamma + \tilde{\gamma } & = \big [ - \frac{u}{2 } + \frac{4v}{(1 + \rho)^2 } - \frac{(3 + \rho ) w}{(1 + \rho)^2 } \big ] + o(\text{2-loop } ) \ , .
\label{gam - dpc2}\end{aligned}\ ] ] at a non - trivial ir - stable fixed point with all @xmath698 and @xmath699 and @xmath700 both different from @xmath627 and @xmath149 , one finds from eqs .
( [ beta - dpc - v ] ) and ( [ beta - dpc - rho ] ) the exact statements @xmath701 and @xmath702 , and eq .
( [ as - skal ] ) yields the asymptotic scaling properties of the green s functions , @xmath703 consequently , the dp - c universality is characterized by the exact critical exponents @xmath704 in eq .
( [ gskal ] ) we have [ beta_exp - dpc ] @xmath705 d/2 \ , , \\
\beta & = \left\ { \begin{array}[c]{lll } \beta^{\prime } \quad & \text{if } & \
\sigma = 0 \\ 1 & \text{if } & \ \sigma \neq 0 \end{array } \right . \ , , \\
\beta^{\prime } & = ( d+\eta^{\prime } ) / d \ , , \qquad \eta^{\prime } = \tilde{\gamma}_{\ast } \ , .\end{aligned}\ ] ] thus , there is merely one unknown independent critical exponent @xmath706 in the dp - c universality class . to order @xmath246
, one finds the following ir fixed points as zeros of the gell - mann low functions @xmath693
: [ fix - dpc ] @xmath707 the run - away solution in the case @xmath663 eventually violates the stability bound @xmath708 in the course of the flow , which indicates the emergence of a first - order transition is this situation . in the other cases , we find for the remaining exponents to one - loop order : [ eta - dpc ] @xmath709 all scaling laws now follow from the scaling of the green s functions ( [ gskal ] ) . as important examples we note the initial slip exponent @xmath710 and @xmath711 for @xmath712 and @xmath713 , respectively .
according to the equation of motion ( [ beweg_s ] ) , the dp - c processes can be understood as models of systems where the critical control parameter @xmath147 itself becomes a dynamical variable @xmath714 subject to a diffusive process . for the case that a dp process evolves in a _
disordered medium _ , the field @xmath563 represents a static but spatially inhomogeneous random quantity with short - range gaussian correlations @xmath715 .
it is easily seen that the influence of the disorder on the other parameters of the model and beyond - gaussian correlations are ir - irrelevant .
in contrast to equilibrium systems , where the normalization factor of the probability distribution , namely the partition function , is itself a functional of the disorder , the description of dynamics by means of an exponential weight @xmath716 with a response functional @xmath717 that depends on the disorder field @xmath718 does not require a normalization factor .
hence , averaging over @xmath563 can be easily performed directly on the statistical weight directly , @xmath719 , without invoking any additional procedures such as , e.g. , replication .
one thereby obtains the following effective action for the evaluation of path integrals averaged over the randomness @xcite : @xmath720 s - \frac{\lambda^2}{2 } \ , f \big [ \int \ ! dt \ , \tilde{s } s \big]^2 \big\ } \ , .\ ] ] from here , the calculation of the renormalizations and the rg flow functions proceeds in the usual manner . to one - loop order ,
one finds the gell - mann low functions @xmath721 where @xmath694 is the renormalized dimensionless coupling corresponding to @xmath601 .
it is now easily seen that the flow equations ( [ beta - rand ] ) allow only for _ runaway _ solutions as @xmath272 in the physical region @xmath722 , @xmath723 .
the fixed point of the pure system @xmath724 , @xmath725 is unstable , as is the non - physical fixed point @xmath726 , @xmath727 .
there is a stable fixed point , namely @xmath728 , @xmath729 , but it is in the non - physical region as well .
the runaway rg trajectories render the perturbation expansion useless and , perhaps , indicate a more complicated critical behavior than just simple power laws .
indeed , the simulations by moreira and dickman @xcite show logarithmic critical spreading instead of power laws , and seem to display violation of simple scaling @xcite .
apparently the continuum field - theoretic treatment is not capable of capturing the strong disorder limit of dp , which may be dominated by rare localized excitations .
a promising alternative approach has recently been developed that employs a real - space rg framework specifically tailored to strongly disordered systems @xcite .
we have benefitted from fruitful collaborations and insightful discussions with numerous colleagues . specifically
, we would like to thank j. cardy , o. deloubrire , h.w .
diehl , v. dohm , g. foltin , e. frey , y. goldschmidt , p. grassberger , h. hilhorst , h. hinrichsen , m. howard , r. kree , . kutbay , s. lbeck , k. oerding , b. schaub , b. schmittmann , f. schwabl , o. stenull , b. vollmayr - lee , f. van wijland , and r. zia for their valuable contributions .
u.c.t . gratefully acknowledges support from the national science foundation under grant no .
dmr-0308548 . for introductions to percolation in general , see , e.g. , d. stauffer and a. aharony , _ introduction to percolation theory _ , 2nd ed .
( taylor and francis , london , 1994 ) ; a. bunde and s. havlin , in _ fractals and disordered systems _ , 2nd ed . , eds . a. bunde , s. havlin ( springer , berlin , 1996 ) .
janssen , z. phys .
b : cond . mat .
* 23 * , 377 ( 1976 ) ; r. bausch , h.k .
janssen , and h. wagner , z. phys .
b : cond . mat .
* 24 * , 113 ( 1976 ) ; h.k .
janssen , in _ dynamical critical phenomena and related topics _
( _ lecture notes in physics _ , vol .
104 ) , ed .
enz , ( springer , heidelberg , 1979 ) .
note that the frequently cited original source for the two - loop results , j.b .
bronzan and j.w .
dash , phys .
b * 51 * , 496 ( 1974 ) , phys .
d * 10 * , 4208 ( 1974 ) , is incomplete , contains a numerical mistake , and is thus erroneous ; for a revision see j.b .
bronzan and j.w .
dash , phys .
d * 12 * , 1850 ( 1975 ) .
f. van wijland , k. oerding , and h. j. hilhorst , physica a * 251 * , 179 ( 1998 ) ; k. oerding , f. van wijland , j. p. leroy , and h. j. hilhorst , j. stat . phys .
* 99 * , 1365 ( 2000 ) ; h.k .
janssen ( 1989 ) , unpublished . | we review the field theory approach to percolation processes . specifically , we focus on the so - called simple and general epidemic processes that display continuous non - equilibrium active to absorbing state phase transitions whose asymptotic features are governed respectively by the directed ( dp ) and dynamic isotropic percolation ( dip ) universality classes .
we discuss the construction of a field theory representation for these markovian stochastic processes based on fundamental phenomenological considerations , as well as from a specific microscopic reaction - diffusion model realization .
subsequently we explain how dynamic renormalization group ( rg ) methods can be applied to obtain the universal properties near the critical point in an expansion about the upper critical dimensions @xmath0 ( dp ) and @xmath1 ( dip ) .
we provide a detailed overview of results for critical exponents , scaling functions , crossover phenomena , finite - size scaling , and also briefly comment on the influence of long - range spreading , the presence of a boundary , multispecies generalizations , coupling of the order parameter to other conserved modes , and quenched disorder .
+ percolation , epidemic processes , directed percolation , dynamic isotropic percolation , active to absorbing phase transitions , renormalization group theory , dynamic critical phenomena , crossover + 64.60.ak , 05.40.-a , 64.60.ht , 82.20.-w |
a wire made of the normal metal is called mesoscopic if the wire length ( @xmath7 ) is smaller than the electron coherence length @xcite .
it is called quasi - one - dimensional ( q1d ) , if @xmath7 is much larger than the width ( @xmath8 ) and thickness ( @xmath9 ) of the wire @xcite .
fabrication of the q1d wires from such metals like au , ag , cu , etc . , usually involves techniques like the electron beam lithography , lift - off , and metal evaporation .
these techniques always provide wires with disorder due to the grain boundaries , impurity atoms and rough wire edges @xcite .
disorder scatters the conduction electrons and limits the electron mean free path ( @xmath10 ) in the wires to @xmath11 nm @xcite . of fundamental interest
are the wires with @xmath8 and @xmath9 as small as @xmath12 nm . in this work the electron transport in metallic q1d wires is studied theoretically .
we compare an impurity - free wire with rough edges with a smooth wire with impurity disorder ( a wire with grain boundaries will be studied elsewhere ) .
we study the q1d wires made of a two - dimensional ( 2d ) conductor ( @xmath13 ) of width @xmath8 and length @xmath14 .
our results are representative for wires made of a normal metal as well as of a 2d electron gas at a semiconductor heterointerface
. and length @xmath7 . the figure on the left depicts the wire with impurities positioned at random with random signs of the impurity potentials .
the figure on the right depicts the impurity - free wire with rough edges , where @xmath15 and @xmath16 are the @xmath17-coordinates of the edges at @xmath18 and @xmath19 , respectively , randomly fluctuating with @xmath20 .
the roughness amplitude is @xmath21 and the step @xmath22 also means the correlation length ( see the text ) . ]
first we review the basic properties of the q1d wires . consider the electron gas confined in the two - dimensional ( 2d ) conductor depicted in the figure [ fig:1 ] . at zero temperature ,
the wave function @xmath23 of the electron at the fermi level ( @xmath24 ) is described by the schrdinger equation @xmath25 with hamiltonian @xmath26 where @xmath27 is the electron effective mass , @xmath28 is the potential due to the impurities , and @xmath29 is the confining potential due to the edges . following the figure [ fig:1 ] , the confining potential in a wire with smooth edges can be written as @xmath30 while in a wire with rough edges it has to be modified as @xmath31 where @xmath15 and @xmath16 are the @xmath17-coordinates of the edges .
the potential of the impurities , @xmath32 , is usually assumed to be a white - noise potential @xcite .
the simplest specific choice is @xmath33 where one sums over the random impurity positions @xmath34 $ ] with a random sign of the impurity strength @xmath35 ( see fig .
[ fig:1 ] ) .
similar models of disorder like in the figure [ fig:1 ] are commonly used in the quantum transport simulations @xcite the disordered q1d wire is connected to two ballistic semiinfinite contacts of constant width @xmath8 , as shown in the figure [ primka ] . in the contacts the electrons obey the schrdinger equation @xmath36 \varphi(x , y)=e_f\varphi(x , y ) , \label{schrodka}\ ] ] where @xmath37 is the confining potential given by equation . solving equation
one finds the independent solutions @xmath38 with the wave vectors @xmath39 given by equation @xmath40 where @xmath41 is the energy of motion in the @xmath17-direction and @xmath42 is the wave function in the direction @xmath17 .
the vectors @xmath39 in are assumed to be positive , i.e. , the waves @xmath43 and @xmath44 describe the free motion in the positive and negative direction of the @xmath20-axis , respectively .
the energy @xmath45 is called the @xmath46-th energy channel .
the channels with @xmath47 are conducting due to the real values of @xmath39 while the channels with @xmath48 are evanescent due to the imaginary @xmath39 .
the conducting state @xmath49 in the contact @xmath50 impinges the disordered wire from the left .
it is partly transmitted through disorder and enters the contact @xmath51 in the form @xmath52 where @xmath53 is the probability amplitude of transmission from @xmath46 to @xmath27 . at zero temperature ,
the conductance of the disordered wire is given by the landauer formula @xcite @xmath54 where we sum over all ( @xmath0 ) conducting channels .
we note that @xmath55 is the transmission probability through disorder for the electron impinging disorder in the @xmath46-th conducting channel .
the amplitudes @xmath56 have to be calculated for specific disorder by solving the equation with the asymptotic condition .
in the ensemble of macroscopically identical wires disorder fluctuates from wire to wire and so does the conductance .
hence it is meaningful to evaluate for the ensemble of wires and to study the ensemble - averaged conductance @xmath57 , variance @xmath58 , resistance @xmath59 , etc .
we now discuss a few important results of such studies .
for simplicity we use the variables @xmath60 and @xmath61 . , @xmath62 coming in the wire and the amplitudes @xmath63 , @xmath64 coming out the wire . ]
first we discuss the smooth q1d wires with disorder due to the white - noise potential @xmath32 .
for @xmath65 the formula gives the ballistic conductance @xmath66 .
as @xmath7 increases , the formula first shows the classical transmission law @xcite @xmath67 where @xmath68 and @xmath69 is the q1d localization length . if @xmath70 , the wire is in the diffusive regime . for @xmath71 and @xmath72
we obtain from the standard expression @xmath73 where @xmath74 is the diffusive conductivity , @xmath75 is the 2d fermi wave vector , and @xmath76 is the 2d electron density . however , the mesoscopic diffusive conductance is also affected by weak localization .
hence , one in fact obtains from a slightly modified version of , namely @xcite @xmath77 where the term @xmath78 is the weak localization correction typical of the q1d wire .
the mean free path @xmath10 in the above formulae coincides with the mean free path derived from the semiclassical boltzmann transport equation , i.e. , the quantum conductance captures the boltzmann transport limit exactly .
the mean of the two - terminal resistance , @xmath79 , shows in absence of weak localization the diffusive behavior @xcite @xmath80 where @xmath81 is the contact resistance and @xmath82 is the diffusive resistivity .
the diffusive resistance and diffusive conductance thus coexist in a standard way : @xmath83 for @xmath84 .
if we include the weak localization by means of , then @xmath85 the conductance fluctuates in the diffusive regime as @xcite @xmath86 where the numerical factor @xmath87 is typical of the q1d wire .
finally , as @xmath7 exceeds @xmath88 , the mesoscopic q1d wire enters the regime of strong localization , where @xmath89 decreases with @xmath7 exponentially while @xmath90 shows exponential increase @xcite .
the formulae - hold for the wires with impurity disorder .
do they hold also for the wires with rough edges ?
in our present work we address this question from the first principles : we calculate the amplitudes @xmath56 by the scattering - matrix method @xcite for a large ensemble of macroscopically identical disordered wires , we evaluate the landauer conductance , and we perform ensemble averaging .
in fact , a few serious differences between the wires with rough edges and wires with impurity disorder were identified prior to our work . in the wire with impurity disorder
the channels are equivalent in the sense that @xmath91 @xcite .
for instance , in the diffusive regime @xmath92 for all channels @xcite .
in the impurity - free wires with rough edges @xmath93 decays fast with rasing @xmath46 , because the scattering by the rough edges is weakest in the channel @xmath94 and strongest in the channel @xmath95 @xcite .
this is easy to understand classically : in the channel @xmath94 the electron avoids the edges by moving in parallel with them , while in the channel @xmath95 the motion is almost perpendicular to the edges , resulting in frequent collisions with them . as a result ,
@xmath93 shows coexistence of the quasi - ballistic , diffusive , and strongly - localized channels @xcite . due to this coexistence , the work @xcite reported absence of the dependence @xmath96 , suggesting that the wire with rough edges does not exhibit the diffusive conductance .
however , according to @xcite , the wire with rough edges seems to exhibit the diffusive resistance . in our present work
these findings are examined again , but for significantly larger @xmath0 as in previous works .
first we study the impurity - free wire whose edges have a roughness correlation length comparable with the fermi wave length . for @xmath97 we observe the quasi - ballistic dependence @xmath98 , where @xmath99 is the quasi - ballistic resistivity .
as @xmath7 increases , we observe crossover to the diffusive dependence @xmath100 , where @xmath101 and @xmath102 is the effective contact resistance due to the @xmath103 open channels .
we find the universal results @xmath104 and @xmath105 for @xmath68 .
as @xmath7 exceeds the localization length @xmath88 , the resistance shows onset of localization while the conductance shows the diffusive dependence @xmath106 up to @xmath107 and the localization for @xmath108 only .
finally , we find @xmath109 the fluctuations differ from and were already reported in the past @xcite . for the smooth wires with impurities
our calculations confirm the formulae - .
moreover , we derive the wire conductivity from the semiclassical boltzmann equation @xcite , and we compare the semiclassical mean - free path with the mean - free path obtained from the quantum resistivity @xmath110 . for the impurity disorder we find that the semiclassical and quantum mean - free paths coincide , which is a standard result .
however , for the edge roughness the semiclassical mean - free path strongly differs from the quantum one , showing that the diffusive transport in the wire with rough edges is not semiclassical .
we show that it is semiclassical only if the roughness - correlation length is much larger than the fermi wave length . for such edge roughness
the conductance behaves like the conductance of the wire with impurities ( formulae [ land - ballist - dif ] - [ varga ] ) , also showing the fluctuations .
the next section describes the scattering - matrix calculation of the amplitudes @xmath56 for the impurity disorder and edge roughness . in section iii ,
the impurity disorder and edge roughness are treated by means of the boltzmann equation and the semiclassical q1d conductivity expressions are derived . in section
iv we show our numerical results . moreover ,
the crossover from @xmath98 to @xmath100 in the wire with rough edges is derived by means of a microscopic analytical theory .
the theory neglects localization but nevertheless captures the main features of our numerical results .
in particular , we obtain analytically the universal results @xmath111 and @xmath112 .
a summary is given in section v.
consider the q1d wire with contacts @xmath50 and @xmath51 , shown in the figure [ primka ] . the wave function @xmath23 in the contacts can be expanded in the basis of the eigenstates .
we introduce notations @xmath113 and @xmath114 , where @xmath115 and @xmath116 are the amplitudes of the waves moving in the positive and negative directions of the @xmath20 axis , respectively . at the boundary @xmath117 @xmath118 \chi_n ( y ) , \label{rozvojdrs1 0}\ ] ] while at the boundary @xmath119 @xmath120 \chi_n ( y ) , \label{rozvojdrs2 l}\ ] ] where @xmath121 is the considered number of channels ( ideally @xmath122 ) .
we define the vectors @xmath123 and @xmath124 with components @xmath125 and @xmath126 , respectively , and we simplify the notations @xmath123 and @xmath124 as @xmath127 and @xmath128 .
the amplitudes @xmath127 and @xmath128 are related through the matrix equation @xmath129 \left ( \begin{array}{c } { \textbf{\textit{a } } } ^+ \\ { \textbf{\textit{b } } } ^- \\ \end{array } \right ) , \quad s \equiv \left [ \begin{array}{cc } r & t ' \\ t & r ' \\ \end{array } \right ] \label{smatrixrovnica}\end{aligned}\ ] ] where @xmath130 is the scattering matrix .
its dimensions are @xmath131 and its elements @xmath132 , @xmath133 , @xmath134 , and @xmath135 are the matrices with dimensions @xmath136 . physically , @xmath132 and @xmath134 are the transmission amplitudes of the waves @xmath137 and @xmath138 , respectively , while @xmath133 and @xmath135 are the corresponding reflection amplitudes .
the matrix elements of the transmission matrix @xmath132 are just the transmission amplitudes @xmath56 which determine the conductance .
consider two wires @xmath50 and @xmath51 , described by the scattering matrices @xmath139 and @xmath140 .
the matrices are defined as @xmath141 , \quad s_2 \equiv \left [ \begin{array}{cc } r_2 & t'_2 \\ t_2 & r'_2 \\
\end{array } \right ] .
\label{s - matica 1 and 2}\ ] ] let @xmath142 \label{s - matica12}\ ] ] is the scattering matrix of the wire obtained by connecting the wires @xmath50 and @xmath51 in series .
the matrix @xmath143 is related to the matrices @xmath139 and @xmath140 through the matrix equations @xcite @xmath144^{-1}t_1 , \\
r_{12 } = r_1 + t'_1r_2[i - r'_1r_2]^{-1}t_1 , \\
t'_{12 } = t'_1[i + r_2[i - r'_1r_2]^{-1}r'_1]t'_2 , \\
r'_{12 } = r'_2 + t_2[i - r'_1r_2]^{-1}r'_1t'_2 , \end{array } \label{skladka}\ ] ] where @xmath145 is the unit matrix .
the equations are usually written in the symbolic form @xmath146 consider the wire with impurity potential .
between any two neighboring impurities there is a region with zero impurity potential , say the region @xmath147 , where the electron moves along the @xmath20 axis like a free particle .
the wire with @xmath46 impurities contains @xmath148 regions with free electron motion , separated by @xmath46 point - like regions where the scattering takes place .
as illustrated in figure [ prim ] , the scattering matrix @xmath130 of such wire can be obtained by applying the combination law @xmath149 where @xmath150 is the scattering matrix of free motion in the region @xmath147 and @xmath151 is the scattering matrix of the @xmath152-th impurity .
the symbols @xmath153 mean that the composition law is applied in step by step : one first combines the matrices @xmath154 and @xmath155 , the resulting matrix is combined with @xmath156 , etc .
the scattering matrix @xmath150 can be expressed as @xmath157 , \label{matrixoffree motion}\ ] ] where @xmath158 is the @xmath136 matrix with zero matrix elements and @xmath159 is the @xmath136 matrix with matrix elements @xmath160 finally , for a @xmath161-function - like impurity the scattering matrix @xmath162 \label{s_i - matica}\ ] ] is composed of the matrices @xcite @xmath163^{-1 } k , \label{tt}\\ r = r ' = - [ k + i\gamma]^{-1 } i \gamma , \label{rr}\end{aligned}\ ] ] where @xmath164 and @xmath165 are the @xmath136 matrices with matrix elements @xmath166 concerning the value of @xmath121 , we use @xmath167 chosen in such way @xcite , that in the diffusive regime our simulation reproduces the boltzmann - equation results .
the electrons in the impurity - free wire with rough edges are described by the schrdinger equation with hamiltonian without the impurity potential , but with the confining potential @xmath29 [ equation ] including the edge roughness .
we specify the edge roughness as follows .
we define @xmath168 , where @xmath169 and @xmath22 is a constant step . for @xmath20 between @xmath170 and @xmath171 ,
the smoothly varying functions @xmath29 , @xmath16 , and @xmath15 in equation are replaced by constant values @xmath172 , @xmath173 , and @xmath174 , respectively .
we obtain the equation @xmath175 we assume that @xmath176 and @xmath177 vary with varying @xmath178 at random in the intervals @xmath179 and @xmath180 , respectively .
this is depicted in the figure [ fig:1 ] , where @xmath16 and @xmath15 fluctuate with varying @xmath20 by changing abruptly after each step @xmath22 .
the wire width fluctuates with varying @xmath20 as well .
however , for @xmath20 between @xmath170 and @xmath171 we have the constant width @xmath181 consequently , the electron wave function for @xmath20 between @xmath170 and @xmath171 can be expressed in the form @xmath182 \chi^j_n ( y ) , \label{j rozvojdrs}\ ] ] where @xmath183 is the considered number of channels ( @xmath184 in the ideal case ) , the wave vectors @xmath185 are given by equations @xmath186 @xmath187 , & d_j < y < h_j \\ 0 , & \mbox{elsewhere } \end{array } \right . \end{array } \label{j subbandfunction}\ ] ] are the wave functions for the @xmath17-direction , and the index @xmath178 means that the above equations hold for @xmath20 between @xmath170 and @xmath171 .
the equations and are just the equations and , respectively , modified for the wire with rough edges . in practice
we choose @xmath183 by means of the relation @xcite @xmath188 i.e. , the ratio of the channel numbers in the @xmath178-th region and contact regions is the same as the ratio of their widths .
we choose @xmath167 and we check , that the calculated conductance does not depend on the choice of @xmath121 .
for @xmath167 the relation ensures @xmath189 , where @xmath190 is the number of the conducting channels in the @xmath178-th wire region .
this makes the calculation reliable also when @xmath190 fluctuates with varying @xmath178 , which happens for @xmath21 larger than the fermi wave length .
as shown in figure [ skoken ] , the scattering matrix @xmath130 of the wire with rough edges is again given by the combination law , where @xmath191 is the scattering matrix of free motion in the region @xmath192 and @xmath193 is the scattering matrix of the @xmath178-th edge step .
the scattering matrix @xmath191 is given by equation with the matrix elements modified as @xmath194 finally , we follow @xcite to specify the scattering matrix @xmath193
. consider the wire shown in figure [ 1skok ] . a single edge step at @xmath117 divides the wire into the region a ( @xmath195 ) and region b ( @xmath196 ) .
the widths of the regions a and b are @xmath197 and @xmath198 , respectively .
we consider the case @xmath199 and we assume , that the wire cross - section @xmath200 includes the wire cross - section @xmath201 ( see the figure ) .
in this situation , the confining potential is simply @xmath202 .
the symbols in the figure are discussed in the text.,width=302 ] so we can use and write the wave function at @xmath203 as @xmath204 \chi^a_n ( y ) , \quad \epsilon \rightarrow 0 , \label{edge rozvojdrs1 minusepsilon}\ ] ] @xmath205 \chi^b_n ( y ) , \quad \epsilon \rightarrow 0 , \label{edge rozvojdrs2 plusepsilon}\ ] ] where @xmath206 , @xmath207 , and @xmath208 and @xmath209 are the channel numbers in the regions a and b. the continuity equation @xmath210 takes the form @xmath211 where @xmath212 is the matrix with the matrix elements @xmath213 and dimensions @xmath214 .
similarly , the continuity equation @xmath215 can be written in the form @xmath216 where @xmath217 and @xmath218 are matrices with the matrix elements @xmath219 and dimensions @xmath220 and @xmath221 , respectively . combining and one finds the matrix equation with the scattering matrix @xmath222 composed of the matrices @xmath223 , \\ r ' = m n , \\
\end{array } \label{trtr}\ ] ] where @xmath224^{-1 } , \\
n = i^b + h_{ba } c(b \rightarrow a ) , \end{array } \label{ihmn}\ ] ] with @xmath225 and @xmath226 being the unit matrices and @xmath227 being the matrix obtained by transposition of the matrix @xmath228 .
the dimensions of the matrices @xmath132 , @xmath133 , @xmath134 and @xmath135 are @xmath229 , @xmath220 , @xmath214 and @xmath230 , respectively . proceeding in a similar way one can derive @xmath193 for the situation @xmath231 , with the cross - section @xmath200 included in the cross - section @xmath201 . in this case @xmath193
is composed of the matrices @xmath232 , \\ r = m n,\\ t ' = -2 m h_{ab},\\ r ' = -2 c(a \rightarrow b ) m h_{ab } - i^b , \\
\end{array}\ ] ] where @xmath233 , @xmath234 , and @xmath121 are the matrices with the index @xmath235 replaced by @xmath236 and vice versa .
the conductance needs to be evaluated for a large ensemble of macroscopically identical wires because it fluctuates from wire to wire .
to evaluate the ensemble - averaged results like @xmath89 , @xmath237 , @xmath238 , etc . , we need to perform averaging typically over @xmath239 samples with different microscopic configurations of disorder .
moreover , the ensemble averages are studied in dependence on the wire length . especially for long wires ( @xmath240 ) with a large number of conducting channels ( @xmath241 ) already the scattering - matrix calculation of a single disordered sample takes a lot of computational time . to decrease the total computational time substantially ( say a few orders of magnitude ) , we introduce a few efficient tricks , partly motivated by the work @xcite .
we recall that two scattering matrices , @xmath139 and @xmath140 , can be combined by means of the operation , written symbolically as @xmath242 .
this operation is associative , i.e. , @xmath243 the property allows us to proceed as follows .
we can construct the disordered wire of length @xmath7 by joining a large number of short wires ( building blocks ) , where each block has the same length ( @xmath244 ) and contains the same number ( @xmath245 ) of scatterers ( impurities or edge steps ) . the scattering matrix of such wire can be expressed as @xmath246 where @xmath247 is the scattering matrix of the first block , @xmath248 is the scattering matrix of the second block , etc . in the simulation
, we can create the @xmath249 different blocks ( typically @xmath250 ) so that we select at random the positions of the @xmath245 scatterers in a given block .
then we evaluate the scattering matrices ( @xmath251 ) of all @xmath249 blocks .
we can now readily study the wire lengths @xmath252 by applying one of the two approaches described below .
a single sample of length @xmath253 can be constructed by joining @xmath178 blocks , where each block is chosen at random from the @xmath249 blocks .
clearly , one can in principle construct @xmath254 different samples of length @xmath255 , but in practice a much smaller number of samples ( @xmath256 ) is sufficient .
the @xmath130-matrix of each sample can be evaluated by means of and ensemble averaging can be performed .
already this approach works much faster than the approach which combines in each sample the @xmath130-matrices of all individual scatterers .
a further significant improvement is achieved as follows . as before ,
we evaluate the scattering matrices @xmath251 for all @xmath249 blocks , but we apply a more sophisticated algorithm : ( i ) we choose at random a single @xmath251-matrix describing a specific sample of length @xmath257 .
( ii ) choosing at random another @xmath251 and combining it with the previous one by means of we obtain the @xmath130-matrix of a specific sample of length @xmath258 .
( iii ) choosing at random another @xmath251 and combining it with the @xmath130 matrix obtained in the preceding step we obtain the @xmath130-matrix of a specific sample of length @xmath259 . in this way
we obtain a set of the landauer conductances @xmath260 for a set of the specific samples with lengths @xmath261 . repeating the algorithm
again we obtain a new set @xmath260 .
repeating the algorithm say @xmath262 times we obtain @xmath262 various sets of @xmath260 and we perform ensemble averaging separately for each @xmath178 .
this approach saves a lot of time because the @xmath130 matrix of the sample of length @xmath263 is created by combining a single @xmath251-matrix with the @xmath130 matrix of a sample of length @xmath264 .
in reality the positions of all scatterers differ from sample to sample . if we take this into account in our simulation , the ensemble - averaged results are the same ( within statistical noise ) as those obtained by means of our trick , but the ensemble averaging takes far much computational time . owing to our trick
we can analyze much larger systems , which is essential for a successful observation of our major results .
in this section , the semiclassical q1d conductivity expressions are derived both for the impurity disorder and edge roughness .
precisely , we assume that the electron motion is semiclassical along the direction parallel with the wire and quantized in the perpendicular direction .
our approach is technically similar to the previous studies of the q1d wires @xcite and q2d slabs @xcite .
the semiclassical conductivity of the q1d wire is given as @xmath265 where @xmath266 is the electric field applied along the wire , @xmath267 is the electron distribution function in the @xmath46-th conducting channel and the factor of 2 includes two spin orientations .
similarly to the usual textbook approach , we express @xmath267 as @xmath268 where @xmath269 + 1)$ ] is the equilibrium occupation number of the electron state with energy @xmath270 and @xmath271 is the relaxation time . setting into we obtain at zero temperature @xmath272 where @xmath39 is the fermi wave vector in the channel @xmath46 [ equation [ spasy ] ] .
the function obeys the linearized boltzmann equation @xmath273 , \label{boltz lin}\end{aligned}\ ] ] where @xmath274 is the fermi - golden - rule probability of scattering from @xmath275 to @xmath276 , with @xmath277 being the scattering - perturbation potential and @xmath278 being the unperturbed electron state .
the index @xmath279 means that @xmath280 is averaged over different configurations of disorder . from
we find ( see the appendix a ) the relaxation time @xmath281 where @xmath164 is the matrix with matrix elements @xmath282 \label{kmatkaa 1}\end{aligned}\ ] ] the boltzmann q1d conductivity thus reads @xmath283 if we set for @xmath277 the impurity potential , the matrix elements can be expressed ( see the appendix b ) in the form @xmath284 where @xmath285 using we obtain from the q1d conductivity @xmath286 for @xmath68 the sum in converges to @xmath287 and converges to the 2d limit @xmath288 derivable from the 2d boltzmann equation . to evaluate the conductivity for the wire with rough edges , we need to determine the perturbation potential @xmath277 produced by the edge roughness potential in the figure [ fig:1 ] , to set the resulting @xmath277 into the right hand side of , and to evaluate @xmath289 .
all this is performed in the appendix c. the result is @xmath290 - n^2 { n'}^2 \left [ \mathcal{f}(|k_{n}-k_{n'}| ) - \right . \nonumber \\ & - & \mathcal{f}(|k_{n}+k_{n'}| ) ] \bigg ] , \label{ker}\end{aligned}\ ] ] where @xmath161 is the root mean square of the fluctuations of the edge coordinates @xmath15 and @xmath16 ( see below ) and @xmath291 is the fourier transform of the roughness correlation function @xmath292 . to specify @xmath161 , @xmath292 , and
@xmath291 we recall ( see figure [ fig:1 ] ) , that @xmath15 and @xmath16 fluctuate with varying @xmath20 in the intervals @xmath179 and @xmath180 , respectively , by changing their values abruptly after constant steps @xmath22 .
obviously , the values of @xmath15 are distributed in the interval @xmath179 with the box - shaped distribution .
using such distribution we find , that @xmath293 where the symbol @xmath161 labels the root mean square of @xmath15 .
the correlation function @xmath292 is defined as @xmath294 we use the box distribution and we take into account that the step @xmath22 plays the role of the correlation length .
we obtain @xmath295 the fourier transform of the last equation is @xmath296 the same results as for @xmath15 hold also for @xmath297 , because the roughness of both edges is the same . as a function of the wire length @xmath7 .
note that @xmath90 is reduced by the contact resistance @xmath81 and @xmath7 is scaled by the mean free path @xmath10 .
the left panel shows the mean resistance of the wire with impurity disorder , the right one shows the mean resistance of the wire with rough edges .
the wire width @xmath8 is fixed to the value @xmath298 , which means that @xmath299 .
the parameters of disorder are listed in the figure together with the resulting values of @xmath10 .
determination of @xmath10 is demonstrated in the next figure .
note that for the au wire ( @xmath300 and @xmath301 kg ) we have @xmath302 nm and @xmath303 nm , with the smallest @xmath21 being only @xmath304 nm .
such small @xmath21 is obviously not realistic , but it is used to emulate the weak roughness limit . ]
in subsection a , the quantum transport in the wires with impurity disorder and wires with rough edges is simulated in the quasi - ballistic , diffusive , and localized regimes . in subsection
b the crossover from the quasi - ballistic to diffusive regime is explained by means of an intuitive model based on the concept of the open channels .
a microscopic analytical theory of the crossover is given in subsection c. in subsection d , the diffusive mean - free path obtained from the landauer conductance and mean - free path from the boltzmann theory are compared for both types of disorder . in subsection
e the wire with rough edges is studied for large roughness - correlation lengths . in principle , all our transport results can be expressed and presented in dependence on the dimensionless variables @xmath305 , @xmath306 , @xmath307 , etc .
, with the fermi wave length @xmath308 being the length unit .
nevertheless , in a few cases we also use the normal ( not dimensionless ) variables and we present the results for the material parameters @xmath301 kg and @xmath309ev ( @xmath302 nm ) , typical of the au wires .
we mostly simulate the wires with the number of the conducting channels being @xmath299 .
this number emulates the limit @xmath68 without spending too much computational time .
whenever needed we also use much larger @xmath0 , the largest one being @xmath310 .
we calculate the landauer conductance for @xmath311 wires and we evaluate the means .
we start with discussion of the mean resistance @xmath90 . in the figure [ fig-5 ] the mean resistance of the wire with impurity disorder
is compared with the mean resistance of the wire with rough edges for various parameters of disorder .
the data obtained for various parameters tend to collapse to a single curve when plotted in dependence on the ratio @xmath312 .
hence it is sufficient to discuss only the data for one specific choice of parameters . , while the dashed line in the right panel is the linear fit @xmath313 , with @xmath314 being the effective contact resistance . in both cases
the resistivity @xmath110 is a fitting parameter and @xmath10 is extracted from the drude formula @xmath315 . to determine @xmath314 , the dashed line in the right panel
is extrapolated to @xmath316 .
we find @xmath317 while @xmath318 .
inset shows the full line from the right panel for @xmath319 , where the transport is quasi - ballistic .
the dotted lines show the linear fit @xmath320 , where the quasi - ballistic resistivity @xmath99 is a fitting parameter .
the quasi - ballistic mean free path @xmath321 is given as @xmath322 . ]
the figure [ fig-55 ] shows the selected numerical data ( full lines ) from the figure [ fig-5 ] . in accord with the textbooks @xcite , the mean resistance of the wire with impurities follows for @xmath323 the standard diffusive dependence @xmath324
however , the mean resistance of the wire with rough edges shows a more complex behavior .
for @xmath97 it follows the linear dependence @xmath325 , where @xmath99 is the quasi - ballistic resistivity . only for large enough @xmath7
it shows crossover to the diffusive dependence @xmath326 , where the resistivity @xmath110 is much smaller that @xmath99 and the effective contact resistance @xmath327 strongly exceeds the fundamental contact resistance @xmath328 .
in other words , the wire with rough edges shows two different linear regimes ( the quasi - ballistic one and the diffusive one ) separated by crossover , while the wire with impurities shows a single linear regime for the quasi - ballistic as well as diffusive transport .
the figure [ fig-55 ] also shows that the mean resistance of the wire with impurities increases for @xmath329 slightly faster than linearly , which is due to the weak localization .
however , the mean resistance of the wire with rough edges increases linearly even for @xmath330 .
the origin of this difference will become clear soon .
, mean resistance @xmath90 and mean conductance @xmath89 versus @xmath331 .
the results for the wire with impurity disorder ( left panels ) are compared with the results for the wire with rough edges ( right panels ) .
the calculations were performed for various sets of the parameters , listed in the figure [ fig-5 ] .
the results for various parameter sets collapse to a single curve shown in a full line ( see the remarks in the text ) .
the dashed lines in the top panels show the fit @xmath332 , from which we determine the localization length @xmath88 .
the resulting values of @xmath88 are shown . ]
the figure [ fig-2 ] shows the numerical results ( full lines ) for the the typical conductance @xmath333 , mean resistance @xmath90 , and mean conductance @xmath89 in dependence on the ratio @xmath331 .
the calculations were performed for various sets of the parameters , shown in the figure [ fig-5 ] .
the results for various sets tend to collapse to a single curve ( full line ) when plotted in dependence on @xmath331 .
we see for both types of disorder , that the numerical data for @xmath333 approach at large @xmath7 the dependence @xmath332 .
this is a sign of the localization @xcite .
fitting of the numerical data provides the values of @xmath88 shown in the figure . in the wire with impurities
we find the result @xmath334 , which agrees with the theoretical @xcite prediction @xmath335 and with numerical studies @xcite . in the wire with rough edges we find @xmath336 .
this does not contradict the work @xcite , which reports @xmath337 , but @xmath338 is @xmath339 times larger than our @xmath10 .
finally , due to the localization also @xmath90 and @xmath89 depend on @xmath331 exponentially at large @xmath331 .
, inverse mean conductance @xmath340 , and conductance fluctuations @xmath341 .
the dashed lines in the top panels are the linear fits @xmath324 ( left panel ) and @xmath313 ( right panel ) .
the dashed lines in the middle panels are the linear fits @xmath342 ( left panel ) and @xmath343 ( right panel ) .
onset of strong localization is marked by arrows at the points , where the numerical data start to deviate from the linear fit remarkably .
the relative deviation from the linear fit is shown in a separate figure [ figure ] .
the dotted line in the bottom panels shows the theoretical value ( @xmath87 ) of the conductance fluctuations , predicted in the limit @xmath344 for the white - noise disorder . ]
the figure [ fig-3 ] show the mean resistance @xmath90 and inverse mean conductance @xmath340 , now in the linear scale .
concerning the mean resistance , the wire with rough edges and wire with impurities behave similarly : @xmath90 rises with @xmath331 linearly on the scale @xmath345 , as is typical for the diffusive regime .
however , both types of wires show a quite different @xmath340 . in the wire with impurities
@xmath340 rises with @xmath331 linearly in the interval @xmath345 , while in the wire with rough edges @xmath340 shows the linear rise with @xmath331 in the interval as large as @xmath346 .
the @xmath96 dependence is a sign of the diffusive conductance regime , which now persists up to @xmath347 and which was not observed in @xcite due to the too narrow length window ( as explained by the authors ) .
further , the slope of the dashed lines is larger for the edge roughness than for the impurities .
this can be understood if we write the equation @xmath348 and we realize that the ratio @xmath349 is larger for the edge roughness .
the figure [ fig-3 ] also shows the conductance fluctuations .
the fluctuations in the wire with impurities approach the universal value @xmath87 , derived @xcite in the limit @xmath344 for the white - noise disorder .
it is remarkable that the fluctuations in the wire with rough edges show a length - independent universal value ( of size @xmath350 ) just in the interval @xmath346 , in which we see the linear rise of @xmath340 .
coexistence of the universal conductance fluctuations with the conductance @xmath351 is typical of the diffusive conductance regime @xcite . in the figure [ fig-3 ] ,
onset of strong localization is visible on the first glance at the points ( marked by arrows ) , where the numerical data for @xmath90 and @xmath340 start to deviate from the linear dependence remarkably . for the edge
roughness the inverse conductance shows onset of localization at @xmath107 : note that the corresponding conductance fluctuations are not universal just for @xmath108 ( they decay with @xmath7 ) . and
@xmath340 from the preceding figure , presented as a relative deviation from the linear fit .
the dashed line shows the weak - localization - mediated relative deviation @xmath352 , shifted by replacement @xmath353 to obey the limit @xmath354 . ]
the figure [ fig-333 ] shows the relative deviation from the linear fit , obtained from the numerical data in figure [ fig-3 ] .
also shown is the relative deviation @xmath352 , obtained from the formula .
as expected , the inverse conductance of the wire with impurities exhibits for @xmath355 the deviation close to @xmath352 .
this is evidently not the case for the wire with rough edges .
first , if @xmath356 , both @xmath90 and @xmath340 show a large negative deviation due to the crossover from the quasi - ballistic to diffusive regime .
second , if @xmath357 , then @xmath340 shows the deviation as small as @xmath358 and almost no deviation for @xmath359 . in other words
, @xmath340 exhibits up to @xmath107 the linear diffusive behavior with a minor nonlinear deviation .
finally , @xmath90 shows a steeply increasing deviation at @xmath360 due to the localization .
we have sofar discussed the numerical data for the wire parameters listed in figure [ fig-5 ] .
apart from small differences due to the statistical noise , these data collapse almost precisely to the same curve , when plotted in dependence on @xmath331 .
in fact , such single - parameter scaling ( dependence solely on @xmath331 ) holds exactly for weak disorder @xcite .
if disorder is not weak , the data can deviate from the single - parameter scaling and the question is whether our findings hold generally . for instance , already in the figure [ fig-5 ]
we do not see for various parameters exactly the same curves . however , the difference between various curves is so small , that the findings extracted from one of these curves ( see figure [ fig-55 ] ) are obtainable with a minor quantitative change from other curves .
so we expect that the same findings hold for any ( reasonable ) choice of the wire parameters .
a strong support for this expectation is that the parameters used in the figure [ fig-5 ] are very different : the mean free paths @xmath10 range from @xmath361 up to @xmath362 . , inverse mean conductance @xmath340 , and conductance fluctuations @xmath341 for various wire parameters . ]
the figure [ fig-33 ] shows again the numerical data for @xmath90 , @xmath340 , and @xmath341 , but for various sets of the wire parameters .
obviously , the data for various sets do not collapse exactly to the same curve .
this may be due to the fact that disorder is not weak , however , the resulting curves are also sensitive to how accurately we determine @xmath88
. we could improve proximity of the curves in the figure [ fig-33 ] by simulating a larger ensemble of samples and a larger wire length ( in order to obtain a more accurate @xmath88 ) .
however , the presented proximity is quite sufficient in the sense that each of the curves allows to obtain the results very similar to those in figures [ fig-3 ] and [ fig-333 ] .
proximity of the curves is satisfactory also with regards to the fact that the values of @xmath88 obtained for various parameters in the figure [ fig-33 ] vary in the range of five orders of magnitude . in this respect
we can also say , that the figure [ fig-33 ] confirms universality of the conductance fluctuations in the wire with rough edges : they are of size @xmath363 , reported by others @xcite .
let us examine the crossover from @xmath325 to @xmath326 , observed in the figure [ fig-55 ] .
such crossover was not observed in the works @xcite , where a similar situation was studied numerically .
therefore , we first analyze the conditions of observability .
we define the effective number of the open channels , @xmath364 , and we evaluate @xmath103 numerically ( by means of the same procedure as in the figure [ fig-55 ] ) for various wire parameters . , versus the roughness correlation length @xmath22 for various @xmath0 and various roughness amplitudes @xmath21 .
the parameters of the edge roughness are scaled as @xmath306 and @xmath365 .
the results for @xmath366 are shown by open circles , the dashed and full lines show the results for @xmath367 and @xmath368 , respectively . ]
the figure [ fig-44 ] shows @xmath103 in dependence on the roughness - correlation length @xmath22 for various @xmath0 and various @xmath21 .
we see that @xmath103 reaches for @xmath369 a minimum value which is roughly @xmath370 and which depends , within our numerical accuracy , neither on @xmath0 nor on @xmath365 . in other words , @xmath103 approaches for @xmath371 the universal value @xmath6 .
the calculations in @xcite were performed for @xmath372 , but only for @xmath373 .
this value is obviously too small for noticing the existence of @xmath374 .
hence the work @xcite reported the diffusive dependence @xmath375 rather than the dependence @xmath376 .
finally , for @xmath377 we see that @xmath103 approaches @xmath0 .
that limit is studied in the last subsection .
further , we look at the numerical data for @xmath93 , shown in the figure [ fig-4 ] . the theory based on the white - noise disorder predicts , that the conducting channels are equivalent @xcite in the sense that @xmath91 . in the figure [ fig-4 ] , this equivalency is reasonably confirmed for the wire with impurity disorder but not for the wire with rough edges . in the wire with rough edges
@xmath93 decays fast with rasing @xmath46 which is easy to understand classically : in the channel @xmath94 the electron avoids the edges by moving in parallel with them , while in the channel @xmath95 the motion is almost perpendicular to the edges , resulting in frequent collisions with them . as a result , the @xmath93 dependence in the right panels of figure [ fig-4 ] shows for @xmath360 the coexistence of the quasi - ballistic , diffusive , and strongly - localized channels , already reported in previous works @xcite . concerning the coexistence ,
two comments are needed .
evidently , the coexistence is not in contradiction with the fact that all @xmath55 decay in semilogarthmic scale linearly with a single parameter @xmath331 , when @xmath378 ( see also @xcite ) .
further , it is clear that if the diffusive regime means @xmath379 in all @xmath0 channels at the same value of @xmath7 , then there is no diffusive regime but only crossover from the quasi - ballistic regime to the localization regime @xcite . in the preceding text we were speaking about the diffusive regime in the sense @xcite that @xmath380 and @xmath381 . versus @xmath331 for the channel indices @xmath382 , where @xmath299 .
for @xmath46 ordered increasingly , the resulting curves are ordered decreasingly : the top curve shows @xmath383 , the bottom one shows @xmath384 .
the bottom panels show @xmath93 versus @xmath46 for various @xmath331 . ] for the uncorrelated impurity disorder , the transmission @xmath93 in absence of the wave interference takes the form @xcite @xmath385 where @xmath386 is the characteristic length . for the edge roughness we can adopt as an ansatz .
we will prove later on that the ansatz is indeed correct for the uncorrelated roughness .
_ in what follows we combine the ansatz with the concept of the open channels and we explain all major features of the crossover from the quasi - ballistic regime to the diffusive one , albeit the formula @xmath387 can not capture the fact that besides the quasi - ballistic and diffusive channels there are also the localized ones .
_ we asses @xmath386 from the numerical data in figure [ fig-4 ] . from the figure [ fig-4 ] it is obvious , that @xmath386 in the channels with @xmath388 is much smaller than @xmath386 in the channels with @xmath389 .
we emulate these findings by a simple model .
we introduce the number @xmath390 and we assume , that the channels @xmath391 have the characteristic length @xmath392 , while the rest of the channels has the characteristic length @xmath393 . by means of the above model
, we can estimate the mean resistance @xmath90 and mean conductance @xmath89 for the wire lengths @xmath394 . from the figure [ fig-3 ] we see that @xmath395 we therefore rely on the equation @xmath396 in the quasi - ballistic limit ( @xmath397 )
the first term in the denominator of is simply @xmath103 and the second term can be evaluated by means of .
for @xmath398 we find @xmath399 where the right hand side holds for @xmath390 . in the diffusive regime ( @xmath355 )
we evaluate the denominator of by means of and we neglect the second term in the denominator assuming that @xmath400 .
we get @xmath401 where the right hand side is the limit @xmath402 .
_ this means that we assume @xmath403 for all @xmath103 channels , i.e. , we ignore that the channel @xmath94 is almost quasi - ballistic even at @xmath360 ( see figure [ fig-4 ] ) . nevertheless , we succeed to obtain all major features of the crossover from the quasi - ballistic regime to the diffusive one ( see mainly the next subsection ) . _
if we compare the formulae and with the formulae @xmath325 and @xmath326 , we obtain @xmath404 we have assumed above that @xmath400 .
this means that @xmath405 .
indeed , we will see that @xmath406 the formula holds only for @xmath407 , as the equation holds for @xmath407 .
however , the formula @xmath408 holds for @xmath409 as we know from our numerical data .
( the fact that the conductance behaves diffusively up to @xmath107 has previously been recognized from the conductance distribution @xcite .
we do not show here the conductance distributions as they are similar to those in @xcite . )
the formulae and show , that the crossover from @xmath325 to @xmath326 is due to the channel non - equivalency . in
what follows , the formulae and will be derived from the first principles , without using the parameters @xmath392 and @xmath244 .
we express the transmission probability @xmath93 as @xmath410 where @xmath411 is the probability that an electron impinging the disordered region in the @xmath27-th channel is reflected back into the @xmath46-th channel . in
the work @xcite the wire with rough edges was analyzed in the quasi - ballistic limit and the reflection probability @xmath412 was derived by means of the first order perturbation theory .
the result is @xmath413 where @xmath414 and the factor of @xmath51 accounts for two edges .
( in fact , the result given in @xcite involves a missprint .
the result can also be extracted from the backscattering length reported in @xcite . )
the mean conductance in the quasi - ballistic limit reads @xmath415 = n_c \left [ 1 - \frac{l}{\frac{\pi}{2}l_{qb } } \right ] , \label{gqb}\ ] ] where @xmath321 is the quasi - ballistic mean free path : @xmath416^{-1 } , \label{lqb}\ ] ] where @xmath417 .
note that the quasi - ballistic limit contains only the backscattering contribution @xmath418 , while in the diffusive regime ( equation [ ker ] ) also the forward - scattering contribution @xmath419 is present .
the mean resistance in the quasi - ballistic limit is @xmath420 in the figure [ lnmvsef ] the expression is compared with @xmath321 determined numerically by means of the approach discussed in figure [ fig-55 ] ( the right panel and inset to the right panel ) .
the formula agrees with our numerical data if the roughness amplitude @xmath21 is small .
this is what one expects , because the perturbation expression is exact in the limit @xmath421 and our scattering matrix calculation is ( in principle ) exact for any @xmath21 . as @xmath21 increases
, the result fails to agree with our numerical data because the scattering is not weak @xcite . in the au wire as a function of the wire width @xmath8 for various roughness amplitudes @xmath21 .
the roughness correlation length is fixed to @xmath422 , the au material parameters are @xmath301 kg and @xmath309ev .
the circles show our numerical data and the squares ( connected by a full line ) are the data points obtained from the expression .
bottom panel : the same calculations as in the top panel , but @xmath8 is fixed to @xmath4 nm and the fermi energy is varied . to explore the small @xmath21 limit reliably
, we have to use the values of @xmath21 which are too small to be realistic . ]
simple formulae can be derived for @xmath321 and @xmath10 if the roughness is uncorrelated ( @xmath423 ) .
we start with @xmath10 . if @xmath423 , the correlation function is simply @xmath424 .
consequently , the backscattering and forward - scattering terms in become the same and the matrix reduces to the diagonal form @xmath425 where the factor of @xmath51 is just due the equal contribution of the backward and forward scattering .
we set into the boltzmann conductivity and we extract from the diffusive mean free path .
it reads @xmath426^{-1}. \label{lunco}\end{aligned}\ ] ] for @xmath68 the summations in can be approximated as @xmath427 and the semiclassical diffusive mean - free path becomes @xmath428 now we evaluate for @xmath423 the quasi - ballistic mean free path .
we rewrite into the form @xmath429^{-1}.
\label{lqblmn}\ ] ] where @xmath430 setting into the formula @xmath424 we obtain @xmath431 combining with and using we obtain @xmath432 we recall that @xmath321 is limited exclusively by backscattering . and diffusive mean free path @xmath10 in the au wire with rough edges , calculated as a function of the wire width @xmath8 and fermi energy @xmath24 .
the thick lines show the formulae and , derived for the uncorrelated roughness .
the thin lines show the formulae valid for an arbitrary correlation length @xmath22 , namely the quasi - ballistic result and the semiclassical @xmath10 extracted from the boltzmann conductivity .
the roughness amplitude @xmath21 and roughness correlation length @xmath22 are fixed to @xmath433 nm and @xmath434 nm ( the limit @xmath435 is fulfilled for all considered data ) . the fermi energy used in the top panels is @xmath309ev , the wire width in the bottom panels is @xmath436 .
the oscillations with sharp minima appear whenever the fermi energy approaches the bottom of the energy subband @xmath437 . ] the formulae and hold for @xmath423 . in figure [ lnmvsef1 ] we compare them with the original formulae valid for any @xmath22 .
we can see that the major difference is absence of the oscillating behavior in the formulae and .
finally , the ratio @xmath438 can be expressed as @xmath439 the result is universal - independent on the wire parameters and parameters of disorder .
the figure [ l pomer lqb ] shows that the universal relation @xmath440 is confirmed by our exact quantum - transport calculation .
obviously , since the formula is the semiclassical boltzmann - equation result and formula is the weak - scattering limit , the formula can not reproduce the exact quantum results quantitatively . , where @xmath10 is the diffusive mean free path and @xmath321 is the quasi - ballistic mean free path , evaluated as a function of @xmath0 for various roughness amplitudes @xmath21 .
the roughness correlation length @xmath22 is fixed to the value @xmath441 which reasonably emulates the uncorrelated roughness .
the open symbols show the results of our quantum transport simulation , where @xmath10 and @xmath321 are calculated as in the figure [ fig-55 ] . ]
the diffusive mean - free path contains both the backward and forward scattering .
it is therefore interesting , that the same result can be obtained when the quasi - ballistic ( backward - scattering - limited ) resistance is extrapolated into the diffusive regime by means of the ansatz .
we start from @xmath442^{-1 } \label{rdifkot}\ ] ] and we use the ansatz . in the quasi - ballistic limit ( @xmath443 ) we obtain from the formula @xmath444 .
we set this formula into and we compare with the quasi - ballistic expressions and .
we find that @xmath445^{-1}. \label{elenka}\ ] ] in the diffusive limit ( @xmath446 ) we obtain from the formula @xmath447 and from the diffusive expression @xmath448^{-1}l .
\label{rdifkolim}\ ] ] we compare this expression with @xmath449 , where @xmath315 .
we find the mean free path @xmath450^{-1}. \label{ldiflmn}\ ] ] if we set into the uncorrelated limit , we obtain again the boltzmann mean - free path .
this is the proof that the ansatz works correctly for the uncorrelated roughness .
now it is useful to make two remarks .
first , our characteristic length @xmath386 should not be confused with the often used @xcite attenuation length @xmath386 .
our @xmath386 is defined by the ansatz and we have just seen , that the expression gives for such @xmath386 the mean free path coinciding with the mean free path obtained from the boltzmann equation .
this is the momentum - relaxation - time - limited mean - free path .
however , if one sets into the attenuation length ( equation ( 5.2 ) in @xcite ) , one obtains from the scattering - time - limited mean - free path , i.e. , the mean distance between two subsequent collisions . for the uncorrelated roughness
the latter is exactly twice shorter than the former one .
second , we note that the ansatz does not work for arbitrary roughness .
indeed , we can set into the formula the more general expression and we can compare with the boltzmann mean free path valid for an arbitrary correlation length @xmath22 . in such case
the formula fails to reproduce the boltzmann - equation result . assuming the uncorrelated roughness
, we are ready to derive analytically the effective number of the open channels , @xmath103 .
the expression can be formally written as @xmath451 where the symbol @xmath103 is defined as @xmath452^{-1 } \label{neff1}\ ] ] in order to obtain again .
we now express the transmission @xmath453 by means of the formula and the mean free path @xmath10 by means of the formula .
we obtain @xmath454^{-1}. \label{neff2}\ ] ] in the diffusive limit ( @xmath455 ) we obtain after some algebraic manipulations the equation @xmath456 the expression no longer depends on @xmath7 and evidently represents the effective number of the open channels , if the resistance is considered in the diffusive limit .
we set into the formulae and .
we obtain @xmath457 ^ 2 \left [ \sum_{m=1}^{n_c } \frac{m^2}{k_m } \right]^{-2}}{\left [ \sum_{n=1}^{n_c } \frac{k^2_n}{n^4 } \right ] \left [ \sum_{m=1}^{n_c } \frac{m^2}{k_m } \right]^{-2}}. \label{neffin1}\ ] ] for @xmath68 the first sum in the denominator of becomes @xmath458 and other sums in are already known [ see equations ] .
we arrive to the result @xmath459 which implies that @xmath103 is a universal number depending neither on the roughness amplitude @xmath21 nor on the number of the conducting channels , @xmath0 .
all this agrees with our quantum - transport calculation in the figure [ fig-44 ] , except that the result underestimates the numerical value @xmath105 about twice .
we however recall that the result relies on the formulae and which are not exact . in this subsection ,
the mean free path obtained from the semiclassical boltzmann equation is compared with the mean free path obtained from the quantum resistivity @xmath110 .
for the impurity disorder we see a standard result @xcite : the semiclassical and quantum mean - free paths coincide for weak impurities . on the contrary , for the edge roughness we find ,
that the quantum mean - free path differs from the semiclassical one even if the roughness amplitude @xmath21 is vanishingly small . in the au wire with impurity disorder as a function of the wire width @xmath8 , calculated for three different sets of the impurity parameters ( listed in the bottom panel ) .
the semiclassical mean free path obtained from the boltzmann q1d conductivity is shown in a full line : all three sets of parameters are intentionally chosen to give the same semiclassical result .
the data shown by symbols are the quantum - transport results obtained from the quantum resistivity @xmath110 ( c.f .
figure [ fig-55 ] ) .
bottom panel : the same calculations as in the top panel , but @xmath8 is fixed to @xmath4 nm and the fermi energy is varied . ] in the figure [ fig-9 ] the quantum and semiclassical mean free paths are compared for the impurity disorder .
the semiclassical data are shown in a full line : three different sets of parameters are intentionally chosen to provide the same semiclassical result . the full lines exhibit oscillations with sharp minima , appearing whenever the fermi energy approaches the bottom of the energy subband @xmath437 .
evidently , the oscillating curve intersects the quantum results ( the data shown by symbols ) , albeit they are slightly affected by statistical noise .
both the semiclassical and quantum results follow the trend predicted by the semiclassical 2d limit ( equation [ sigmaimp ] ) , namely that @xmath10 is proportional to @xmath460 and independent on @xmath8 . in summary , in the q1d wire with impurity disorder
the semiclassical and quantum mean - free paths coincide .
this is a standard result @xcite , known from the theory based on the white - noise disorder @xcite ( see however @xcite ) . in the au wire with rough edges as a function of the wire width @xmath8 , calculated for various roughness amplitudes @xmath21 .
the left panel shows the same calculation , but @xmath8 is fixed to @xmath4 nm and the fermi energy is varied .
the roughness correlation length is kept at the value @xmath434 nm , which means that @xmath461 for most of the presented data .
the full lines show the semiclassical mean free path obtained from the boltzmann q1d conductivity ( equations [ sigmab ] and [ ker ] ) .
the open circles show the quantum mean free path extracted from the quantum resistivity @xmath110 ( c.f .
figure [ fig-55 ] ) .
the open squares show the mean free path obtained by means of the classical scattering - matrix calculation ( see the text ) , in which the localization is absent . to explore the small @xmath21 limit reliably
, we have to use the values of @xmath21 which are too small to be realistic . ]
as shown in the figure [ fig-934 ] , in the wire with rough edges the situation is different .
we again compare the semiclassical mean free path ( full lines ) with the quantum mean free path ( open circles ) .
as before , the full lines exhibit oscillations with sharp minima , appearing whenever the fermi energy approaches the bottom of the energy subband @xmath437 .
evidently , the full line and open circles show a quite different behavior if the roughness amplitude @xmath21 is large ( notice the results for @xmath433 nm ) .
it is tempting to ascribe this difference to the weak - perturbation approximation involved in the boltzmann equation , and similarly , it is tempting to expect that the full line will intersect the open circles for sufficiently small @xmath21 .
however , this is not the case : even for the smallest considered @xmath21 the open circles are systematically a factor of @xmath462 below the full line . in summary , in the wire with rough edges the quantum mean - free path differs from the semiclassical one ( by the factor of @xmath462 ) even if @xmath21 is vanishingly small . to understand the origin of this difference
, we now exclude from our quantum - transport calculation the wave interference .
we recall ( see subsection ii.c ) , that in the quantum - transport calculation the total @xmath130-matrix of the disordered wire is obtained by combining at random the scattering matrices ( @xmath251 ) of the building blocks , where the @xmath131 matrix @xmath251 is composed of the complex amplitudes @xmath463 , and @xmath464 . to exclude the wave interference , we proceed as follows @xcite .
first , we consider only the conducting channels and we completely neglect the evanescent ones : this reduces the size of the matrix @xmath251 to @xmath465 .
second , instead of the complex matrix @xmath251 we use the real one , in which the complex amplitudes @xmath466 , and @xmath464 are replaced by the real probabilities @xmath467 , and @xmath468 , respectively .
of course , the wave interference is excluded completely , if the length of the building block , @xmath244 , coincides with the length of the edge step , @xmath22 .
fortunately , in practical calculations the wave interference is negligible already for @xmath469 .
if we combine the resulting real matrices @xmath251 by means of the same combination law as before ( equation [ skladka ] ) , we obtain the classical transmission probability @xmath470 and eventually the classical landauer conductance @xmath471 finally , we perform ensemble averaging over many samples . of the wire with rough edges in dependence on the ratio @xmath331 .
the thin full line is the result of the quantum scattering - matrix calculation ( the same data as in the top right panel of figure [ fig-3 ] ) , now labeled as @xmath472 .
the thick full line is the result of the classical scattering - matrix calculation , labeled as @xmath473 .
the dashed lines show the linear fits @xmath313 , from which we determine the effective contact resistance @xmath314 and the diffusive mean free path @xmath474 .
the resulting @xmath314 are shown in the figure .
inset shows the classical result again , but for larger @xmath331 in order to stress the linear raise and absence of the localization . ] in the figure [ fig-8 ] the classical scattering - matrix calculation is compared with the quantum one .
the mean resistance due to the quantum calculation is labeled as @xmath472 to distinguish from the classical @xmath473 .
it can be seen that both @xmath472 and @xmath473 exhibit the crossover to the linear diffusive dependence @xmath313 , but the classical result shows a smaller value of @xmath110 and a larger value of @xmath314 .
the value @xmath475 is already close to our theoretical result @xmath112 ( equation [ neffuni ] ) , where the wave interference is excluded as well . the smaller value of @xmath110 is due to the absence of the localization .
let us return to the figure [ fig-934 ] .
the squares show the mean free path extracted from @xmath473 .
these data overestimate the quantum results ( circles ) systematically by a factor of @xmath462 .
compare now the classical scattering - matrix results ( squares ) with the semiclassical boltzmann results ( full lines ) .
the full lines intersect the squares only if @xmath21 is vanishingly small .
this is due to the fact that the boltzmann results rely on the weak - perturbation theory while the classical scattering - matrix calculation still involves the exact quantum scattering by a single edge step .
the roughness - limited conductivity derived from the boltzmann equation ( usually further simplified by using the fuchs - sondheimer model @xcite ) is often compared with the transport measurements of the metallic nanowires @xcite .
however , such comparison can not verify the boltzmann result unambiguously due to the presence of other processes ( like the scattering by impurities and grain boundaries ) and also due to the fact , that the roughness parameters @xmath21 and @xmath22 are not known _ apriori_. our quantum calculation shows unambiguously , that the semiclassical boltzmann approach describes the wires with rough edges reasonably , only if @xmath21 _ is too small to be realistic ( in the au nanowires)_. in practice , the metallic nano - wires are usually fabricated by advanced lift - off techniques @xcite , which hardly allow to suppress the edge roughness to the value @xmath476 nm . in the table
[ tabula ] , the quantum and semiclassical mean - free paths are compared for a more realistic @xmath21 and @xmath8 like in the figure [ fig-934 ] . we see that the quantum result can exceed the semiclassical boltzmann result one order of magnitude .
one might argue that the quantum result holds only in the coherent regime while decoherence is present at any nonzero temperature .
we expect that decoherence will tend to destroy the wave interference and to restore the resistance @xmath473 resulting from the classical scattering - matrix approach
. the resulting classical mean - free path would be twice larger than the quantum one [ see figure [ fig-934 ] ] , i.e. , it would exceeds the boltzmann mean - free path in the table [ tabula ] by another factor of two .
.the mean - free path @xmath10 in the au wire obtained from the quantum - transport simulation is compared with the semiclassical boltzmann mean - free path @xmath477 ( equations [ sigmab ] and [ ker ] ) for various values of the wire width @xmath8 , roughness amplitude @xmath21 and roughness - correlation length @xmath22 .
each set of the parameters considered in the table represents strong disorder , for which @xmath477 strongly differs from @xmath10 .
[ cols="^,^,^,^,^",options="header " , ] the wires with rough edges have sofar been discussed for the roughness - correlation lengths @xmath478 . the only exception was the figure [ fig-44 ] where the effective number of the open channels , @xmath103 , was calculated from @xmath479 up to @xmath480 .
we have seen , that the value of @xmath103 approaches for @xmath480 the value of @xmath0 , which suggests that the diffusive dependence @xmath481 approaches the standard form @xmath482 , seen in the wire with impurities . in this subsection
we focus on the limit @xmath480 . , except that the edge roughness is considered in the limit @xmath377 . specifically , the data obtained for the wire with impurities ( left column ) are the same as those in the left column of figure [ fig-3 ] , while for the edge roughness ( right column ) we show the data obtained for @xmath483 and @xmath484 ( the data for @xmath485 and @xmath486 , not shown , look similarly except for statistical noise ) .
we recall that the localization length @xmath88 is determined from the dependence @xmath332 . for the edge roughness
we now obtain @xmath487 , which is close to the impurity case : @xmath488 ] the figure [ fig-323 ] shows the same calculations as the figure [ fig-3 ] , except that the wire with rough edges is now simulated in the limit @xmath480 . unlike the figure [ fig-3 ] , the results for the wire with rough edges are now close to the results for the wire with impurity disorder . first , the localization length in the wire with rough edges fulfills the relation @xmath487 , which is close to the relation @xmath488 found for the impurity disorder .
this is why the resistance @xmath90 and inverse conductance @xmath489 in the figure [ fig-323 ] show for both types of disorder the same linear slope with @xmath331 .
moreover , the same linear slope means a single linear regime for the quasi - ballistic as well as diffusive transport , i.e. , the crossover between the quasi - ballistic and diffusive regimes , observed for @xmath478 , tends to disappear .
( in the figure [ fig-3 ] a remarkably larger slope is seen for the edge roughness because @xmath336 . )
second , the inverse conductance in the figure [ fig-323 ] shows for both types of disorder the linear diffusive regime for @xmath490 and onset of localization for @xmath491 .
( in the figure [ fig-3 ] the inverse conductance of the wire with rough edges shows onset of localization for @xmath347 . ) third , the wire with rough edges now exhibits essentially the same universal conductance fluctuations as the wire with impurities .
the size of these fluctuations is in accord with the value @xmath492 , predicted for the white - noise - like disorder . to see such conductance fluctuations for the edge roughness
is surprising : sofar only the value @xmath363 has been reported @xcite . , except that the edge roughness is considered in the limit @xmath377 [ we show the data for @xmath483 ] . ] to provide insight , it is useful to look at the transmission probabilities @xmath93 .
the figure [ fig-4434 ] shows the same calculation as the figure [ fig-4 ] , but with the edge roughness considered in the limit @xmath377 .
it can be seen that the figure [ fig-4434 ] differs from the figure [ fig-4 ] in the following respect : except for a few channels with lowest values of @xmath46 , the wire with rough edges and wire with impurities show for the rest of the channels a very similar @xmath93 , which is clearly not the case in figure [ fig-4 ] .
this similarity is responsible for similarity of the transport results in the figure [ fig-323 ] .
the observed similarity could be further improved by simulating the wires with @xmath0 larger than @xmath1 , but this is beyond the scope of this work .
finally , the figure [ fig-13 ] shows for @xmath480 , what happens in the wire with rough edges with the diffusive mean free path .
specifically , the semiclassical boltzmann result ( full lines ) is compared with the quantum ( open circles ) and classical ( open squares ) scattering - matrix calculations . in the limit @xmath480
all three calculations tend to give the same mean free path .
this means that the interference effects observed for @xmath478 are no longer important , or in other words , the diffusive transport in the wire with rough edges is semiclassical .
this is again similar to the wire with impurities . for @xmath480
it is easy to show analytically , that the semiclassical boltzmann mean free path rises with @xmath22 linearly .
the correlation function @xmath291 contains the function @xmath494 .
this function oscillates with @xmath495 at random and the oscillations become very fast for @xmath480 . consequently , the correlation function @xmath291 oscillates fast around the value @xmath496 .
if we ignore these fast oscillations , the diffusive mean free path can be written as @xmath497 where @xmath498 \right .
\nonumber \\
+ \frac{n^4}{2k_n^2 } \bigg ] - ( 1 - \delta_{n n ' } ) n^2 { n'}^2 \left [ \frac{1}{(k_{n'}-k_n)^2 } - \frac{1}{(k_{n'}+k_n)^2 } \right ] .
\nonumber \\ \label{kersl}\end{aligned}\ ] ] it can be seen that @xmath499 .
the @xmath499 dependence resembles the wire with impurities , where @xmath10 rises linearly with the average distance between the impurities .
the formula is plotted in the figure [ fig-13 ] in a dashed line .
it indeed agrees with the numerical data for @xmath480 . on the other hand ,
the formula , derived in the opposite limit , reasonably agrees with the numerical data for @xmath479 , but the dependence @xmath500 has no similarity with the impurity case . finally , as also pointed out in the figure caption , the figure [ fig-13 ] shows that the semiclassical boltzmann result ( full line ) reproduces the classical scattering - matrix calculation ( squares ) for any @xmath22 only if @xmath21 is too small to be realistic .
this is another indication that the boltzmann - equation theory of the edge - roughness - limited transport should be used with caution at least in the au nanowires . in the wire with rough edges as a function of the roughness - correlation length @xmath22 for various roughness amplitudes @xmath21 , with @xmath10 , @xmath22 and @xmath21 scaled by @xmath308 .
the wire width @xmath8 is fixed to the value @xmath298 , which means that @xmath299 ) .
the full lines show the semiclassical mean - free path obtained from the boltzmann conductivity ( equations [ sigmab ] and [ ker ] ) .
the dashed lines for @xmath461 show the uncorrelated limit , the dashed lines for @xmath501 show the correlated limit .
the circles show the quantum mean - free path extracted from the quantum resistivity @xmath110 ( c.f .
figure [ fig-55 ] ) .
the squares show the mean - free path obtained by means of the classical scattering - matrix calculation in which the localization effect is absent ( see subsection iv.d ) .
note that for the au wire ( @xmath300 and @xmath301 kg ) we have @xmath302 nm ,
i.e. , the wire width is @xmath502 nm and the smallest @xmath21 is only @xmath304 nm . such small @xmath21 is obviously not realistic , which demonstrates an important finding : the semiclassical boltzmann result ( full line ) reproduces the classical scattering - matrix calculation ( squares ) for any @xmath22 only if @xmath21 is too small to be realistic . ]
we have studied quantum transport in quasi - one - dimensional wires made of a two - dimensional conductor of width @xmath8 and length @xmath14 .
the main purpose of our work was to compare an impurity - free wire with rough edges with a smooth wire with impurity disorder .
we have calculated the electron transmission through the wires by the scattering - matrix method , and we have obtained the landauer conductance / resistance for a large ensemble of macroscopically identical disordered wires . we have first studied the impurity - free wire whose edges have a roughness correlation length comparable with the fermi wave length . the mean resistance @xmath90 and inverse mean conductance @xmath340 have been evaluated in dependence on @xmath7 .
for @xmath97 we have found the quasi - ballistic dependence @xmath98 , where @xmath503 is the fundamental contact resistance and @xmath99 is the quasi - ballistic resistivity . for larger @xmath7
we have found the crossover to the diffusive dependence @xmath100 , where @xmath110 is the resistivity and @xmath102 is the effective contact resistance corresponding to the @xmath103 open channels .
we have found the universal results @xmath504 and @xmath105 for @xmath68 .
approaching the localization regime , we have demonstrated the following numerical finding : as @xmath7 exceeds the localization length @xmath88 , the resistance shows onset of localization while the conductance shows the diffusive dependence @xmath106 up to @xmath107 and the localization for @xmath108 only . on the contrary , for the impurity disorder
we have found a standard diffusive behavior , namely @xmath505 for @xmath506 .
we have also seen that the impurity disorder and edge roughness show the universal conductance fluctuations of different size , as already reported in previous works @xcite .
we have then attempted to interpret our quantum - transport results in terms of an approximate but microscopic analytical theory . in particular , the crossover from @xmath98 to @xmath100 in the wire with rough edges has been derived analytically assuming the uncorrelated edge roughness and neglecting the localization effects . in spite of this approximation
, our analytical results capture the main features of our numerical results .
specifically , we have derived the universal results @xmath111 and @xmath112 .
we have also derived the wire conductivity from the semiclassical boltzmann equation , and we have compare the semiclassical mean - free path with the mean - free path obtained from the quantum resistivity @xmath110 . for the impurity disorder we have found , that the semiclassical and quantum mean - free paths coincide , which is a standard result . however , for the edge roughness the semiclassical mean - free path strongly differs from the quantum one , showing that the diffusive transport in the wire with rough edges is not semiclassical .
we have found that it becomes semiclassical only if the roughness - correlation length is much larger than the fermi wave length . in such case
the resistance and conductance tend to scale with @xmath331 like in the wire with impurity disorder , also showing the universal conductance fluctuations of similar size .
we end by a remark about our edge - roughness model .
it is the same model as in the previous simulations @xcite .
we believe that most of the results obtained within the model are model - independent .
for example , we have tested the correlation function @xmath291 of gaussian shape and we have seen that the resulting transmissions remain quite similar to those in the figure the only exception is the correlated limit , which predicts the dependence @xmath499 . this dependence is the artefact of our @xmath291 choice .
for example , for the gaussian correlation function , the dependence @xmath499 is replaced by a much faster increase with @xmath22 . on the other hand
, the uncorrelated limit can be equally well derived for the gaussian correlation function or for the exponential one , i.e. , the result is very general .
we thank for the grant vvce-0058 - 07 , vega grant 2/0633/09 , and grant apvv-51 - 003505 .
we thank peter marko for useful discussions in the initial stage of this work and for helpful comments to the manuscript .
at zero temperature the distribution reads @xmath507 and the boltzmann equation can be written as @xmath508 .
\label{boltz 1}\end{aligned}\ ] ] we set and into .
we obtain @xmath509 .
\label{boltze dosadene}\end{aligned}\ ] ] we proceed similarly to the q2d theory @xcite .
first , we multiply both sides of by @xmath510 and sum over @xmath510 .
second , on the left hand side we replace @xmath511 by @xmath512 and integrate .
third , on the right hand side we replace @xmath513 by @xmath514 and @xmath515 by @xmath516 , which is justified due to the presence of @xmath517 .
we obtain the equation @xmath518 \label{boltzes 1}\end{aligned}\ ] ] and from @xmath519 eventually @xmath520 where @xmath521 is defined by equation . from
we obtain the relaxation time .
if the scattering potential @xmath277 is given by the impurity potential , then @xmath522 since the impurities are positioned at random , equation can be readily simplified as @xmath523 and the right hand side of can be easy evaluated : @xmath524_{\mbox{\fontsize{8}{8}\selectfont av } } \nonumber \\ = \frac{\gamma^2 n_i}{l w } \left [ 1 - \frac{\delta_{n n'}}{2 } \right ] , \nonumber \\ \label{scatimp}\end{aligned}\ ] ] where @xmath525 is the sheet impurity density . to obtain the right hand side of
, we have replaced the term @xmath526 by integral @xmath527 . setting into we get the expression .
assume first that the electrons are confined in the wire by the potential barriers of finite hight .
specifically , if the wire edges are smooth , the electron hamiltonian reads @xmath528 where @xmath529 and @xmath530 is the hight of the potential barrier at @xmath531 and @xmath532 , respectively , and @xmath533 is the heaviside step function . if the wire edges are rough , the hamiltonian reads @xmath534 , where @xmath535 + \nonumber \\ & + & v_{+}\left [ \theta(y - h(x ) ) - \theta(y - w ) \right ] \label{hsrpo}\end{aligned}\ ] ] is the perturbation potential due to the edge roughness , with @xmath15 and @xmath297 being the fluctuations of the edges ( see figure [ fig:1 ] ) .
now we also assume that @xmath536 and @xmath537 and we approximate as @xmath538 \delta(y - w ) .
\label{hsrpoap}\ ] ] then @xmath539 where @xmath540 , @xmath541 , @xmath542 , and @xmath543 .
if we integrate in the equation over the variable @xmath17 and then use the limit @xmath544 both terms in the sum on the right hand side of can be rewritten into the form @xmath545 where the brackets @xmath546 symbolize the ensemble averaging ( instead of the symbol _ av _ ) and @xmath547 is the roughness - correlation function .
the correlation function depends only on the distance @xmath548 , so we rewrite it as @xmath549 where @xmath550 is the normalized correlation function and @xmath551 and @xmath552 are the root - mean squares of the randomly fluctuating functions @xmath553 and @xmath554 , respectively .
we introduce the variable @xmath555 and simplify as @xmath556 to obtain the second line in , we have replaced the limits @xmath557 and @xmath558 in the first line by @xmath559 and @xmath560 , respectively . except for very small @xmath561 , such replacement is justified because @xmath562 decays with increasing @xmath563 to zero on the distance ( correlation length ) much smaller than @xmath7 .
the equation can now be expressed as @xmath564 where @xmath565 is the fourier transform of @xmath566 .
assuming the same randomness at both edges we can put @xmath567 i.e. , we skip the indices @xmath568 .
we set into and we replace the symbol @xmath569 in by the factor of 2 . on the first glance
it might seem that the scattering has to be weak if @xmath7 is sufficiently small .
however , the quasi - ballistic ( not ballistic ) regime is a meaningfull concept only if the length @xmath7 involves at least one edge step .
even one edge step can scatter strongly if the step size is large .
all this is correct for weak impurities .
if the impurities are not weak , the semiclassical boltzmann result has to be modified by treating the single - impurity scattering non - perturbatively .
see e.g. m. moko and p. vagner , phys .
b * 59 * , r10445 ( 1999 ) . h. s. lee and j. b. yoon , j. micromech . microeng . * 15 * , 2136 ( 2005 ) .
a. voigt , m. heinrich , k. hauck , r. mientus , g. gruetzner , m. topper , and o. ehrmann , microelectronic engineering * 89 * , 503 ( 2005 ) .
m. t. bryan , d. atkinson , and r. p. cowburn , journal of physics : conference series * 17 * , 40 ( 2005 ) . | we study electron transport in quasi - one - dimensional metallic wires .
our aim is to compare an impurity - free wire with rough edges with a smooth wire with impurity disorder .
we calculate the electron transmission through the wires by the scattering - matrix method , and we find the landauer conductance for a large ensemble of disordered wires .
we first study the impurity - free wire whose edges have roughness with a correlation length comparable with the fermi wave length . simulating wires with the number of the conducting channels ( @xmath0 ) as large as @xmath1 - @xmath2
, we observe the roughness - mediated effects which are not observable for small @xmath0 ( @xmath3 - @xmath4 ) used in previous works .
first , we observe the crossover from the quasi - ballistic transport to the diffusive one , where the ratio of the quasi - ballistic resistivity to the diffusive resistivity is @xmath5 independently on the parameters of roughness .
second , we find that transport in the diffusive regime is carried by a small effective number of open channels , equal to @xmath6 .
this number is universal - independent on @xmath0 and on the parameters of roughness .
third , we see that the inverse mean conductance rises linearly with the wire length ( a sign of the diffusive regime ) up to the length twice larger than the electron localization length .
we develop a theory based on the weak - scattering limit and semiclassical boltzmann equation , and we explain the first and second observations analytically . for impurity disorder
we find a standard diffusive behavior . finally , we derive from the boltzmann equation the semiclassical electron mean - free path and we compare it with the quantum mean - free path obtained from the landauer conductance .
they coincide for the impurity disorder , however , for the edge roughness they strongly differ , i.e. , the diffusive transport in the wire with rough edges is not semiclassical .
it becomes semiclassical only for roughness with large correlation length .
the conductance then behaves like the conductance of the wire with impurities , also showing the conductance fluctuations of the same size . |
flares are the energetic explosions in the stellar atmosphere and are thought to occur by intense releases of magnetic energy stored around starspots , like solar flares ( e.g. , @xcite ) .
the total energy released in the largest solar flares is estimated to be of the order of @xmath3 erg ( e.g. , @xcite ; @xcite ) .
superflare " events have been recently found by using the kepler data ( @xcite ; @xcite ) .
superflares are gigantic flare events that are 10-@xmath4 times more energetic ( @xmath5 erg ; @xcite ) than the largest solar flares on the sun ( @xmath6erg ) .
we have analyzed the data from kepler space telescope @xcite , and discovered 365 superflare events on 148 solar - type ( g - type main - sequence ) stars from the data of 83,000 solar - type stars observed for the first 120 days of the kepler mission @xcite .
we here define solar - type stars as the stars that have a surface temperature of @xmath7k and a surface gravity of @xmath8 . extending the study of @xcite
, we found 1547 superflare events on 279 solar - type stars by using the kepler data of a longer period ( @xmath9500 days ) @xcite .
kepler is very useful for detecting small increases in the stellar brightness caused by stellar flares , because kepler realized high photometric precision exceeding 0.01% for moderately bright stars , and obtained continuous time - series data of many stars over a long period @xcite .
+ + the analyses of kepler data enabled us to discuss statistical properties of superflares since a very large number of flare events were discovered .
@xcite confirmed that the occurrence rate ( @xmath10 ) of the superflare versus the flare energy ( @xmath11 ) has a power - law distribution of @xmath12 , where @xmath13 , and this distribution is roughly similar to that for the solar flare .
@xcite also estimated that a superflare with an energy of @xmath14 erg occurs once in 800 - 5000 years in sun - like stars . sun - like stars " are here defined as solar - type stars with an effective temperature of @xmath15k , a surface gravity of @xmath8 , and a rotation period ( @xmath16 ) longer than 10 days .
+ + considering these results , we have performed more analyses in order to investigate whether properties of superflare stars can be explained by applying our current physical understanding of the sun .
solar flares are the intense releases of magnetic energy stored around sunspots , as mentioned above .
if we assume superflares are the similar magnetic energy releases , large starspots are necessary to explain their large amount of energy .
many of the superflare stars show quasi - periodic brightness variations with a typical period of from one day to a few tens of days .
the amplitude of these brightness variations is in the range of 0.1 - 10% @xcite , and is much larger than that of the solar brightness variation ( 0.01 - 0.1% ; e.g. , @xcite ) caused by the existence of sunspots on the rotating solar surface .
many ordinary solar - type stars observed by kepler also show such brightness variations and these variations are thought to be caused by the rotation of a star having starspots ( e.g. , @xcite ; @xcite ) .
@xcite showed that the above brightness variations of superflare stars can be well explained by the rotation of a star with fairly large starspots , taking into account the effects of inclination angle and the spot latitude .
in other words , the existence of large starspots is also supported by the brightness variations .
+ + @xcite compared the superflare energy and frequency with the rotation period , assuming that the brightness variation period corresponds to the rotation period .
they then found slowly rotating stars can still produce as energetic flares as those of more rapidly rotating stars although the average flare frequency is lower for more slowly rotating stars .
@xcite also clarified that the superflare energy is related to the total coverage of the starspots , and that the energy of superflares can be explained by the magnetic energy stored around these large starspots .
in addition , @xcite suggested , on the basis of theoretical estimates , that the sun can generate large magnetic flux sufficient for causing superflares with an energy of @xmath17 erg within one solar cycle ( @xmath911yr ) .
+ + the results described above are , however , only based on kepler monochromatic photometric data . we need to spectroscopically investigate whether these brightness variations are explained by the rotation , and whether superflare stars have large starspots . the stellar parameters and the binarity of the superflare stars are also needed to be investigated with spectroscopic observations in order to discuss whether the sun can really generate superflares .
+ + we have then performed high dispersion spectroscopy of 50 superflares stars .
we have already reported some results of our spectroscopic observations of three superflare stars in @xcite and @xcite .
in particular , @xcite found that spectroscopic properties ( @xmath18 , @xmath19 , [ fe / h ] , and rotational velocity ) of the two superflare stars kic9766237 and kic9944137 are very close to those of the sun . in (;
hereinafter referred to as paper i ) , we have already conducted measurements of stellar parameters for the sake of detailed discussions in this paper . in paper
i , we investigated the binarity of the target stars , and found that more than half ( 34 stars ) of 50 target superflare stars have no evidence of binarity .
we then estimated temperature ( @xmath18 ) , surface gravity ( @xmath19 ) , metallicity ( [ fe / h ] ) , and projected rotational velocity ( @xmath20 ) of these 34 single " superflare stars .
the accuracy of our stellar atmospheric parameters ( @xmath18 , @xmath19 , and [ fe / h ] ) is higher than that of kic ( kepler input catalog ; @xcite ) values , and the differences between our values and kic values ( @xmath21k , @xmath22 dex , and @xmath23})_{\rm{rms } } \sim 0.46 $ ] dex ) are comparable to the large uncertainties and systematic differences in kic values reported by the previous researches ( e.g. , @xcite ) .
we confirmed that the estimated temperature and @xmath19 values of these 34 superflare stars are roughly in the range of solar - type ( g - type main sequence stars ) stars .
in particular , the temperature , surface gravity , and the brightness variation period ( @xmath24 ) of 9 stars including two stars ( kic9766237 and kic9944137 ) reported in @xcite are in the range of sun - like " stars ( @xmath15k , @xmath254.0 , and @xmath2610 days ) .
we also found that five of the 34 target stars are fast rotators ( @xmath27 km s@xmath28 ) , while 22 stars have relatively low @xmath0 values ( @xmath29 km s@xmath28 ) .
these results suggest that stars whose spectroscopic properties similar to the sun can have superflares , and this supports the hypothesis that the sun might cause a superflare . + + on the basis of the above stellar parameters derived in paper i , in this paper we investigate whether the brightness variation is explained by the rotation , and whether superflare stars have large starspots . in this process
, we use @xmath0 " estimated in paper i. moreover , we use the intensity of ca ii infrared triplet ( ca ii irt , @xmath30 ) lines and h@xmath1 line , which we have already used for investigating chromospheric activity in @xcite and @xcite .
ca ii irt and h@xmath1 lines are well - known good indicators of the chromospheric activity of g - type stars ( e.g. , @xcite ; @xcite ; @xcite ) .
+ + in section [ sec : obsdata ] , we briefly explain the observational data used in this paper . in section [ subsec : caiiha ]
, we show the measurement results of the intensity of ca ii irt and h@xmath1 lines .
we then indirectly estimate the mean intensity of stellar magnetic field using ca ii 8542 line on the basis of our solar spectroheliographic observation in section [ subsec : hida ] . in section [ subsec :
inc ] , we compare @xmath20 " with the period of the brightness variation , and consider whether the brightness variation is explained by the rotation . in section [ subsec : activity - spots ] , we discuss stellar chromospheric activities , including the starspot sizes and rotational velocity of superflare stars .
the details of our spectroscopic observations are described in section 2 of paper i. our observations have been performed by using the high dispersion spectrograph ( hds : @xcite ) at the 8.2-m subaru telescope .
the names of the above 34 superflare stars , which show no evidence of binarity in paper i , and 10 comparison stars are shown in table [ tab : r0caha ] .
in addition to the above 10 comparison stars , moon was also observed as a comparison star in this observation , as briefly mentioned in section 2.2 of paper i. the observed spectra of these stars around ca ii 8542 and h@xmath1 6563 are shown in figures [ fig : sg1ca ] and [ fig : sg1 ha ] , respectively .
we did not use ca ii 8542 line of 7 target superflare stars ( kic7093547 , kic8359398 , kic8547383 , kic9459362 , kic10252382 , kic10387363 , and kic11818740 ) since the resultant s / n ratio around ca ii irt lines is low , as we can see in figure [ fig : sg1ca ] .
we show these stars with a comment low s / n " in this figure .
examples of spectra around photospheric lines , including fe i 6212 , 6215 , 6216 , 6219 are shown in figure [ fig : sg1fe ] . as we have already mentioned in paper i
, we observed 13 superflare stars among these 34 superflare stars , and 3 comparison stars ( 59 vir , 61 vir , and 18 sco ) multiple times .
we made co - added spectra of these 16 stars in total , which have s12b - comb " or
s13a - comb " in supplementary table 2 of paper i. in figures [ fig : sg1ca ] , [ fig : sg1 ha ] , and [ fig : sg1fe ] , co - added spectra of these 16 stars are used .
only the co - added spectra are used in this paper when we analyze the spectral data of these 16 stars .
in order to investigate the chromospheric activity of the target stars , we measured @xmath31(8498 ) , @xmath31(8542 ) , @xmath31(8662 ) , and @xmath31(h@xmath1 ) index , which are the residual core flux normalized by the continuum level at the line cores of the ca ii irt and h@xmath1 , respectively . as we have already introduced previous researches in detail in section 3.3 of @xcite ,
these indexes are known to be good indicators of stellar chromospheric activity ( e.g. , @xcite ; @xcite ) . as the chromospheric activity
is enhanced , the intensity of these indicators becomes large since a greater amount of emission from the chromosphere fills in the core of the lines .
the values of @xmath31(8498 ) , @xmath31(8542 ) , @xmath31(8662 ) , and @xmath31(h@xmath1 ) indexes of the target single superflare stars and comparison stars are listed in table [ tab : r0caha ] .
we were not able to measure the values of ca ii irt indexes ( @xmath31(8498 ) , @xmath31(8542 ) , and @xmath31(8662 ) ) of 7 target superflare stars ( kic7093547 , kic8359398 , kic8547383 , kic9459362 , kic10252382 , kic10387363 , and kic11818740 ) because of low s / n ratios around ca ii irt lines . for reference
, we also measured the emission flux of ca ii irt and h@xmath1 lines in appendix [ sec : excessflux ] .
+ + as mentioned above , ca ii lines are good indicators of stellar chromospheric activity , and the intensity of them well correlates with the intensity of stellar magnetic field .
@xcite derived a rough empirical relation between the emission core flux of ca ii k line and the average photospheric magnetic field , on the basis of spectroheliographic observations of solar active regions . in this paper
, we extended this relation to the emission core flux of ca ii 8542 ( @xmath31(8542 ) ) and the average stellar magnetic field by newly carrying out spectroheliographic observations of solar active regions .
we summarized the method and results in the following of this section .
+ + in order to derive an empirical relation , we compared @xmath31(8542 ) values obtained from spectroheliographic observation of the active region noaa11654 , with the photospheric magnetic field strength data of the same region on the same observation date . the spectroheliographic observation was conducted using 60 cm domeless solar telescope ( dst ) with the horizontal spectrograph of the hida observatory , kyoto university @xcite .
the observation date was 2013 january 11th ( jst ) .
the photospheric magnetic field strength data ( @xmath2 ; @xmath32 " is a filling factor ) that we used here were taken by hmi ( @xcite ; @xcite ) on the solar dynamics observatory ( sdo ; @xcite ) on the same day .
+ + adjusting the spatial scale of both the dst and hmi data to @xmath33 , we plotted @xmath31(8542 ) value of each data point in the target active region as a fuction of @xmath2 ( mean intensity of the photospheric magnetic field ) of the same point .
as shown in figure [ fig : hida - r0fb ] ( a ) , a rough positive correlation is seen between @xmath31(8542 ) and @xmath2 .
this figure means that @xmath31(8542 ) index roughly reflects the average intensity of the stellar magnetic field ( @xmath2 ) , though this correlation is not so good especially for less active stars ( low @xmath2 ) .
using the least - square method , we conduct a linear fit in the double logarithmic graph @xmath34(8542 ) vs. @xmath35 ( figure [ fig : hida - r0fb ] ( b ) ) . as a result
, we derived a very rough empirical relation between @xmath31(8542 ) and @xmath2 : @xmath36 we then roughly estimate @xmath2 values of the target stars , by using equation ( [ eq : r0fb ] ) and @xmath31 ( 8542 ) values .
the estimated @xmath2 values are listed in table [ tab : r0caha ] .
the error value of each @xmath2 value in table [ tab : r0caha ] is estimated from the standard deviation of @xmath2 distribution as a function of @xmath31(8542 ) in figure [ fig : hida - r0fb ] .
in addition , if we apply equation ( [ eq : r0fb ] ) to solar @xmath31(8542 ) value estimated in @xcite ( @xmath31(8542)=0.193 ) , @xmath2 is estimated to be @xmath37 [ gauss ] .
this value is consistent with the values of solar mean magnetic field ( a few gauss ) within its error range , though the error range is large .
as mentioned above , we reported the values of projected rotational velocity ( @xmath0 ) , stellar radius ( @xmath38 ) , and the brightness variation period ( @xmath24 ) of the 34 single " target superflare stars in paper i. the values of @xmath0 and @xmath24 are listed again in table [ tab : r0caha ] .
assuming that the brightness variations of these stars are caused by the rotation of the stars with starspots , we can estimate the rotational velocity ( @xmath39 ) from @xmath24 and @xmath38 by using @xmath40 as we have already mentioned in section 5.2 of @xcite .
the resultant values of @xmath39 are listed in table [ tab : r0caha ] .
we here consider that the typical error of @xmath39 is about @xmath4125% , assuming that the upper limit of errors of @xmath38 and @xmath24 are about 20% and 10% , respectively , on the basis of paper i. + + in figure [ fig : vlc - vsini ] , we plot @xmath0 as a function of the @xmath39 .
some data points in figure [ fig : vlc - vsini ] show differences between the values of @xmath39 and @xmath0 .
the projected rotational velocity ( @xmath0 ) tends to be smaller than @xmath39 .
such differences should be explained by the inclination effect , as mentioned in the previous studies ( e.g. , @xcite ; @xcite ) . on the basis of @xmath0 and @xmath39 , the stellar inclination angle ( @xmath42 )
can be estimated by using the following relation : @xmath43 in figure [ fig : vlc - vsini ] , we also show four lines indicating @xmath44 ( @xmath45 ) , @xmath46 , @xmath47 , and @xmath48 .
this figure shows two following important results .
first , for almost all the stars ( 33 stars ) except for kic11764567 , the relation
@xmath49 " is satisfied .
this is consistent with our assumption that the brightness variation is caused by the rotation since the inclination effect mentioned above can cause the relation @xmath49 " if @xmath39 values really correspond to the rotational velocities ( i.e. @xmath50 ) .
this is also supported by another fact that the distribution of the data points in figure [ fig : vlc - vsini ] are not random .
their distribution is expected to be much more random if the brightness variations have no relations with the stellar rotation .
second , stars that are distributed in the lower right side of figure [ fig : vlc - vsini ] are expected to have small inclination angles and to be nearly pole - on stars . in this figure , we distinguish such five stars with especially small inclination angle ( @xmath51 ) ( kic3626094 , kic4742436 , kic6503434 , kic6934317 , and kic9412514 ) from the other stars , using filled triangle data points .
+ + we can confirm the above inclination effects from another point of view , as already mentioned for kic6934317 in figure 7 of @xcite .
figure [ fig : spotene ] is a scatter plot of the flare energy of superflares and solar flares as a function of the spot coverage .
the data used in this figure are the same as in figure 10 of @xcite .
the spot coverage of superflare stars are calculated from the amplitude of stellar brightness variations , and the energy of superflares are estimated by using the superflare amplitude and the duration time @xcite .
thick and thin solid lines correspond to the analytic relation between the spot coverage and the flare energy , which is obtained from equation ( 14 ) of @xcite in case of @xmath44 for @xmath52=3,000 g and 1,000 g .
the thick and thin dashed lines correspond to the same relation in case of @xmath53 ( nearly pole - on ) for @xmath52=3,000 g and 1,000 g , assuming that the brightness variation becomes small as a result of the inclination effect .
these lines are considered to give an upper limit of superflare energy for each inclination angle @xcite .
considering these things , the superflare stars located in the upper left side of this figure are expected to have a low inclination angle .
open triangles in figure [ fig : spotene ] correspond to superflare stars whose inclination angle is especially small ( @xmath54 ) on the basis of figure [ fig : vlc - vsini ] , while open circles represent the other stars ( @xmath55 ) .
this classification is the same as that in figure [ fig : vlc - vsini ] .
all of the five triangle data points are located above the thick solid line ( @xmath44 and @xmath52=3000 g ) .
this means that these stars are also confirmed to have low inclination angle on the basis of figure [ fig : vlc - vsini ] . as a result
, these two figures ( figure [ fig : vlc - vsini ] and [ fig : spotene ] ) are confirmed to be consistent , and in other words , the stellar projected rotational velocity spectroscopically measured is consistent with the rotational velocity estimated from the brightness variation .
the fact indicates that the brightness variation of superflare stars is caused by the rotation .
in figure [ fig : tvr0 ] ( a ) , we plot @xmath31(8542 ) as a function of @xmath18 of the target superflare stars .
the data of ordinary solar - type stars reported in @xcite are also plotted in this figure .
the variability range of @xmath31(8542 ) between the solar maximum and solar minimum is only [email protected] @xcite .
this figure shows that almost all the target superflare stars are more active compared to the sun from the viewpoint of the @xmath31(8542 ) index . in other words ,
the mean magnetic field strength of the target stars can be higher than that of the sun ( see section [ subsec : hida ] ) .
in addition , 4 target superflare stars ( kic8429280 , kic9652680 , kic10528093 , and kic11610797 ) show much higher chromospheric activity ( @xmath31(8542)@xmath560.6 ) compared to the other solar - type stars in this figure .
+ + in figure [ fig : tvr0 ] ( b ) , we plot @xmath31(8542 ) as a function of @xmath0 of the target superflare stars .
the data of ordinary solar - type stars reported in @xcite(8542 ) due to the blurring effect caused by an increase of @xmath0 is small compared to the real changes of the stellar activity level .
then we here consider that this effect does not have any essential effects on the interpretation of figure [ fig : tvr0 ] ( b ) . ]
are also plotted for reference in this figure .
there is a rough positive correlation between @xmath31(8542 ) and @xmath0 , and more rapidly rotating stars have higher chromospheric activity .
all of the 5 stars with extremely high @xmath0 value ( @xmath27 km s@xmath28 ) compared to the sun ( @xmath572 km s@xmath28 ) show relatively high @xmath31(8542 ) values ( @xmath58 ) .
this is consistent with many previous observational results of ordinary main - sequence stars ( e.g. , @xcite ; @xcite ) . in addition ,
about 6 superflare stars with relatively small @xmath0 values ( @xmath59 km s@xmath28 ) in this figure ( kic4831454 , kic6865484 , kic8802340 , kic11140181 , kic11303472 , and kic12266582 ) also have relatively high chromospheric activity ( @xmath58 ) compared to the data of ordinary solar - type stars taken from @xcite .
+ + the @xmath31(8542 ) values are plotted in figure [ fig : ampr0fb ] ( a ) as a function of the amplitude of stellar brightness variation ( @xmath60bvamp@xmath61 ) .
we calculated @xmath60bvamp@xmath61 listed in table [ tab : r0caha ] by taking the average of the amplitude values of q2@xmath9q16 data .
the amplitude value in each quarter is available in supplementary data of this paper .
we did not use q0 , q1 , and q17 data since the duration of these three quarters is short ( @xmath6230 days ) compared to those of the other 15 quarters ( @xmath990 days ) @xcite .
the errors of @xmath60bvamp@xmath61 in table [ tab : r0caha ] correspond to the maximum and minimum of the amplitude values of q2@xmath9q16 data .
the solar @xmath31(8542 ) value in @xcite ( @xmath63 ) is also plotted for reference . the solar brightness variation amplitude in this figure
is calculated from the sunspot coverage in the observation date ( 2007 february 2 ) of @xcite , and its maximum and minimum value shown here correspond to the typical sunspot coverage of the solar maximum and minimum , respectively . in figure
[ fig : ampr0fb ] ( a ) , there is a rough positive correlation between @xmath31(8542 ) and @xmath60bvamp@xmath61 . assuming that the brightness variation of superflare stars is caused by the rotation of a star with starspots , the brightness variation amplitude ( @xmath60bvamp@xmath61 ) corresponds to the starspot coverage of these stars .
then , we can say that there is a rough positive correlation between the starspot coverage and chromospheric activity level ( @xmath31(8542 ) ) .
this rough correlation shows us that all the target stars expected to have large starspots on the basis of their large amplitude of the brightness variation show high magnetic activity compared to the sun . in other words , our assumption that the amplitude of the brightness variation correspond to the spot coverage is supported , since high magnetic activity , which are confirmed by using @xmath31(8542 ) values , are considered to be caused by the existence of large starspots .
+ + in figure [ fig : ampr0fb ] ( b ) , we also plot @xmath2 values that we estimated from @xmath31(8542 ) in section [ subsec : hida ] , as a function of @xmath60bvamp@xmath61 . with this figure
, we can confirm the same conclusion as we did with figure [ fig : ampr0fb ] ( a ) , though the errors of @xmath2 values are a bit large especially for less active stars . in appendix
[ sec : excessflux ] , we also investigated the emission flux of ca ii irt and h@xmath1 lines for reference , and confirmed basically the same conclusions as we did with @xmath31(8542 ) index in this section .
+ + in addition to the analyses in this paper , we plan to discuss the li abundances and stellar age of superflare stars in another future paper ( honda et al . in preparation ) .
as we have introduced previous researches in section 4.5 of @xcite , li abundance can provide some constraints on the age of g - type stars ( e.g. , @xcite ; @xcite ) .
information of stellar age is important since the stellar activity has a deep relation with the stellar age ( e.g. , @xcite ) , but this point is beyond the scope of this paper .
we have performed high dispersion spectroscopic observations of 50 superflare stars with subaru / hds .
more than half ( 34 stars ) of these 50 target superflare stars show no evidence of binarity , and we measured stellar parameters of these 34 stars in our previous study ( paper i ) . in this paper , we investigated whether the quasi - periodic brightness variation of superflare stars is explained by the rotation , and whether superflare stars have large starspots , by using stellar parameters derived in paper i. first , the value of @xmath0 measured from spectroscopic results is consistent with the rotational velocity estimated from the period of the brightness variation .
next , we measured the intensity of ca ii irt and h@xmath1 lines , which are good indicators of stellar chromospheric activity , and compared them with other stellar properties .
the intensity of ca ii irt lines indicates that the mean magnetic field strength ( @xmath2 ) of the target superflare stars can be higher than that of the sun .
we found a correlation between the amplitude of the brightness variation and the intensity of ca ii irt .
all the targets expected to have large starspots because of their large amplitude of the brightness variation show high chromospheric activity compared to the sun .
these results support that the brightness variation of superflare stars is explained by the rotation with large starspots .
+ + + + this study is based on observational data collected with subaru telescope , which is operated by the national astronomical observatory of japan .
we are grateful to dr .
akito tajitsu and other staffs of the subaru telescope for making large contributions in carrying out our observations .
we would also like to thank dr .
yoichi takeda for his many useful advices on the analysis of our subaru / hds data , and for his opening the tgvit and sptool programs into public .
when we conducted our solar observation at hida observatory mentioned in section [ subsec : hida ] , we received important advices and kind supports from dr .
satoru ueno , dr . tetsu anan , dr .
reizaburo kitai , dr . ayumi asai , and other members of kwasan and hida observatories , kyoto university .
kepler was selected as the tenth discovery mission .
funding for this mission is provided by the nasa science mission directorate .
the kepler data presented in this paper were obtained from the multimission archive at stsci .
this work was supported by the grant - in - aids from the ministry of education , culture , sports , science and technology of japan ( no .
25287039 , 26400231 , and 26800096 ) .
alonso , a. , arribas , s. , & martinez - roger , c. 1996 , , 313 , 873 anderson , r. i. , reiners , a. , solanki , s. k. 2010 , , 522 , a81 basri , g. , et al .
2011 , , 141 , 20 brown , t.m . ,
latham , d. w. , everett , m. e. , & esquerdo , g. a. 2011 , , 142 , 112 candelaresi , s. , hillier , a. , maehara , h. , brandenburg , a. , & shibata , k. 2014 , , 792 , 67 emslie , et al .
, 2012 , , 759 , 71 foing , b. h. , crivellari , l. , vladilo , g. , rebolo , r. , & beckman , j. e. 1989 , , 80 , 189 frasca , a. , & catalano , s. 1994 , , 284 , 883 frasca , a. , frhlich , h .- e .
, bonanno , a. , catanzaro , g. , biazzo , k. , & molenda - akowicz , j. 2011 , , 532a , 81f gray , d. f. 2005 , the observation and analysis of stellar photospheres , 3rd ed .
( cambridge : cambridge university press ) hall , j.c .
1996 , pasp , 108 , 313 hirano , t. , sanchis - ojeda , r. , takeda , y. , narita , n. , winn , j. n. , taruya , a. , & suto , y. 2012 , , 756 , 66 koch , d. g. , et al .
2010 , , 713 , l79 lanza , a. f. , rodon , m. , pagano , i. , barge , p. , & llebaria , a. 2003 , , 403 , 1135 linsky , j. l. , hunten , d. m. , sowell , r. , glackin , d. l. , & kelch , w. l. 1979 , , 41 , 481 livingston , w. , wallace , l. , white , o. r. , & giampapa , m. s. 2007 , , 657 , 1137 maehara , h. , et al .
2012 , 485 , 478 martnez - arniz , r. , lpez - santiago , j. , crespo - chacn , i. , & montes , d. 2011 , , 414 , 2629 [ erratum : , 417 , 3100 ] nakai , y. , & hattori , a. 1985 , memoirs faculty of sciences university of kyoto , 36 , 385 nogami , d. , notsu , y. , honda , s. , maehara , h. , notsu , s. , shibayama , t. , & shibata , k. 2014 , , 2014 , 66 , l4 noguchi , k. , et al .
2002 , , 54 , 855 notsu , s. , honda , s. , notsu , y. , nagao , t. , shibayama , t. , maehara , h. , nogami , d. , & shibata , k. 2013a , , 65 , 112 notsu , y. , et al .
2013b , 771 , 127 notsu , y. , honda , s. , maehara , h. , notsu , s. , shibayama , t. , nogami , d. , & shibata , k. 2015 , , in press ( paper i ) noyes , r. w. , hartmann , l. w. , baliunas , s. l. , duncan , d. k. , & vaughan , a. h. 1984 , , 279 , 763 pesnell , w. d. , thompson , b. j. , & chamberlin , p. c. 2012 , , 275 , 3 priest , e. r. 1981 , solar flare magnetohydrodynamics ( new york ; gordon and breach science publishers ) reinhold , t. , reiners , a. , & basri , g. 2013 , , 560 , a4 schaefer , b. e. , king , j. r. , & deliyannis , c. p. 2000 , , 529 , 1026 scherrer , p. h. , et al .
2012 , , 275 , 207 schou , j. , et al .
2012 , , 275 , 229 schrijver , c. j. , cote , j. , zwaan , c. , & saar , s. h. 1989 , , 337 , 964 sestito , p. , & randich , s. 2005 , , 442 , 615 shibata , k. , & magara , t. 2011 , living rev .
phys , 8 , 6 shibata , k. , et al .
2013 , , 65 , 49 shibayama , t. , et al .
2013 , , 209 , 5 soderblom , d. r. , duncan , d. k. , & johnson , d. r. h. 1991 , , 375 , 722 soderblom , d. r. , jones , b. f. , balachandran , s. , stauffer , j. r. , duncan , d. k. , fedele , s. b. , & hudon , j. d. 1993 , , 106 , 1059 soderblom , d. r. , stauffer , j. r. , hudon , j. d. , & jones , b. f. 1993 , , 85 , 315 takeda , y. , honda , s. , kawanomoto , s. , ando , h. , & sakurai , t. 2010 , , 515 , a93 thompson , s. e. , et al .
2013 , kepler data release 23 notes ( ksci-19063 - 001 ) ( 70mm,70mm)shsg1-ca8542.eps ( 70mm,70mm)shsg2-ca8542.eps ( 70mm,70mm)shsg3-ca8542.eps ( 70mm,70mm)shsgh - ca8542.eps ( 70mm,70mm)shsg1-ha6563.eps ( 70mm,70mm)shsg2-ha6563.eps ( 70mm,70mm)shsg3-ha6563.eps ( 70mm,70mm)shsgh - ha6563.eps ( 70mm,70mm)shsg1-fe6212.eps ( 70mm,70mm)shsg2-fe6212.eps ( 70mm,70mm)shsg3-fe6212.eps ( 70mm,70mm)shsgh - fe6212.eps ( 120mm,120mm)hida - r0fb-2.4.eps ( 120mm,120mm)loghida - r0fb-2.4.eps ( 150mm,150mm)vlc - vsini - log_q17_bw1.eps ( 130mm,130mm)fig10-rkic - bw.eps ( 100mm,100mm)tca_wtakeda2010.eps ( 100mm,100mm)vsini - r0-bw.eps ( 105mm,105mm)bvamp - r0wcap - bw - ye.eps ( 105mm,105mm)bvamp - fbwcap - bw.eps lccccccccc starname & @xmath64 & @xmath39 & @xmath24 & @xmath60bvamp@xmath61 & @xmath65 & @xmath2 & @xmath66 & @xmath67 & @xmath68 + & [ km s@xmath28 ] & [ km s@xmath28 ] & [ day ] & [ % ] & & [ gauss ] & & & + + + + + + kic3626094 & 2.9 & 100.6 & 0.7 & @xmath70 & 0.24 & 0.6@xmath416 & 0.33 & 0.25 & 0.21 + kic4742436 & 3.6 & 28.3 & 2.3 & @xmath71 & 0.30 & 3@xmath416 & 0.40 & 0.32 & 0.29 + kic4831454 & 2.1 & 8.6 & 5.2 & @xmath72 & 0.50 & 68@xmath4141 & 0.58 & 0.44 & 0.32 + kic6503434 & 5.5 & 39.5 & 3.9 & @xmath73 & 0.27 & 1@xmath416 & 0.35 & 0.27 & 0.27 + kic6504503 & 1.6 & 1.4 & 31.8 & @xmath74 & 0.25 & 0.8@xmath416 & 0.35 & 0.26 & 0.21 + kic6865484 & 2.7 & 4.5 & 10.3 & @xmath75 & 0.48 & 52@xmath4141 & 0.56 & 0.42 & 0.40 + kic6934317 & 1.9 & 19.7 & 2.5 & @xmath76 & 0.33 & 5@xmath4124 & 0.43 & 0.37 & 0.42 + kic7093547 & 2.1 & 3.2 & 14.2 & @xmath77 & - & - & - & - & 0.32 + kic7354508 & 3.2 & 4.5 & 16.8 & @xmath78 & 0.40 & 16@xmath4124 & 0.42 & 0.33 & 0.24 + kic7420545 & 6.3 & 4.8 & 36.2 & @xmath79 & 0.40 & 16@xmath4124 & 0.50 & 0.37 & 0.32 + kic8359398 & 1.8 & 2.9 & 12.7 & @xmath80 & - & - & - & - & 0.34 + kic8429280 & 37.1 & 31.6 & 1.2 & @xmath81 & 0.68 & 489@xmath41173 & 0.80 & 0.72 & 0.93 + kic8547383 & 4.4 & 4.0 & 14.8 & @xmath82 & - & - & - & - & 0.38 + kic8802340 & 4.3 & 4.4 & 10.3 & @xmath83 & 0.49 & 60@xmath4141 & 0.57 & 0.46 & 0.31 + kic9412514 & 8.1 & 38.9 & 3.7 & @xmath84 & 0.22 & 0.4@xmath416 & 0.32 & 0.22 & 0.19 + kic9459362 & 7.0 & 9.3 & 12.6 & @xmath85 & - & - & - & - & 0.46 + kic9583493 & 7.2 & 9.1 & 5.5 & @xmath86 & 0.51 & 77@xmath41123 & 0.61 & 0.49 & 0.38 + kic9652680 & 34.2 & 29.0 & 1.5 & @xmath87 & 0.68 & 489@xmath41173 & 0.76 & 0.69 & 0.60 + kic9766237 & 2.1 & 4.2 & 14.2 & @xmath88 & 0.26 & 1@xmath416 & 0.40 & 0.28 & 0.23 + kic9944137 & 1.9 & 3.7 & 12.6 & @xmath89 & 0.27 & 1@xmath416 & 0.41 & 0.29 & 0.22 + kic10252382 & 5.5 & 6.1 & 16.8 & @xmath90 & - & - & - & - & 0.52 + kic10387363 & 1.5 & 1.4 & 29.9 & @xmath91 & - & - & - & - & 0.31 + kic10471412 & 2.7 & 3.2 & 15.2 & @xmath92 & 0.35 & 7@xmath4124 & 0.48 & 0.34 & 0.28 + kic10528093 & 11.4 & 8.5 & 12.2 & @xmath93 & 0.67 & 445@xmath41173 & 0.75 & 0.66 & 0.51 + kic11140181 & 3.6 & 3.7 & 11.5 & @xmath94 & 0.56 & 141@xmath41123 & 0.63 & 0.45 & 0.36 + kic11197517 & 0.9 & 3.2 & 14.3 & @xmath95 & 0.19 & 0.1@xmath416 & 0.27 & 0.21 & 0.22 + kic11303472 & 2.1 & 2.8 & 13.5 & @xmath96 & 0.49 & 60@xmath4141 & 0.56 & 0.39 & 0.33 + kic11390058 & 3.5 & 4.2 & 12.0 & @xmath97 & 0.32 & 4@xmath4124 & 0.42 & 0.30 & 0.29 + kic11455711 & 7.9 & 7.6 & 13.9 & @xmath98 & 0.40 & 16@xmath4124 & 0.59 & 0.44 & 0.30 + kic11494048 & 3.4 & 3.4 & 14.8 & @xmath99 & 0.32 & 4@xmath4124 & 0.45 & 0.31 & 0.23 + kic11610797 & 22.5 & 38.4 & 1.6 & @xmath100 & 0.63 & 300@xmath41173 & 0.70 & 0.61 & 0.46 + kic11764567 & 14.7 & 3.6 & 22.4 & @xmath101 & 0.49 & 60@xmath4141 & 0.57 & 0.44 & 0.39 + kic11818740 & 3.0 & 2.7 & 15.4 & @xmath102 & - & - & - & - & 0.35 + kic12266582 & 4.8 & 6.1 & 6.9 & @xmath103 & 0.52 & 88@xmath41123 & 0.61 & 0.48 & 0.37 + 59vir & 6.1 & - & - & - & 0.40 & 16@xmath4124 & 0.50 & 0.38 & 0.27 + 61vir & @xmath1041.0 & - & - & - & 0.19 & 0.1@xmath416 & 0.30 & 0.20 & 0.18 + 18sco & 2.0 & - & - & - & 0.20 & 0.2@xmath416 & 0.30 & 0.21 & 0.18 + hd163441 & 2.5 & - & - & - & 0.23 & 0.5@xmath416 & 0.34 & 0.25 & 0.19 + hd173071 & 2.7 & - & - & - & 0.25 & 0.8@xmath416 & 0.34 & 0.25 & 0.21 + hip100963 & 2.4 & - & - & - & 0.22 & 0.4@xmath416 & 0.34 & 0.22 & 0.20 + hip71813 & 2.3 & - & - & - & 0.20 & 0.2@xmath416 & 0.30 & 0.20 & 0.19 + hip76114 & 1.8 & - & - & - & 0.21 & 0.3@xmath416 & 0.32 & 0.20 & 0.19 + hip77718 & 2.5 & - & - & - & 0.23 & 0.5@xmath416 & 0.34 & 0.23 & 0.20 + hip78399 & 2.5 & - & - & - & 0.21 & 0.3@xmath416 & 0.32 & 0.21 & 0.18 + moon & 2.4 & - & - & - & 0.20 & 0.2@xmath416 & 0.33 & 0.22 & 0.19 +
we measured the values of the @xmath31 of ca ii irt and h@xmath1 in section [ subsec : caiiha ] , and discussed chromospheric activity of superflare stars by using the @xmath31(8542 ) index in section [ subsec : activity - spots ] .
this index is an indicator of chromspheric activity as described in the above sections , but it also depends on @xmath0 ( e.g. , figure 5(a ) of @xcite ) .
a large value of @xmath0 can indeed increase the residual flux , mimicking the effect of filling the line core with chromospheric emission . because of this , we here roughly estimated the emission flux of the ca ii 8542 and h@xmath1 lines in order to remove this influence of @xmath0 .
the ways of estimation is basically the same as done in @xcite , and we summarize the method in the following .
+ + we used the spectral subtraction technique ( e.g. , @xcite ; @xcite ; @xcite ) . with this subtraction process
, we can subtract the underlying photospheric contribution from the spectrum of the star , and can investigate the spectral emission originating from the chromosphere in detail @xcite .
we used the spectrum of 61vir obtained in this observation as an inactive template to be subtracted from the spectrum of the target stars .
this is because 61vir is a slowly rotating and non - active early g - type main sequence star ( e.g. , @xcite ) .
we measured the excess equivalent width ( @xmath105 ) of the ca ii 8542 and h@xmath1 lines by using the residual spectrum around the line core resulting from this subtraction process .
we then derived emission fluxes ( @xmath106 ) of ca ii 8542 and h@xmath1 lines from @xmath105 of these lines by the following relation , @xmath107 ( @xcite ) , where @xmath108 is the continuum flux around the wavelength of each line .
we calculated @xmath108 by using the the following empirical relationships between @xmath108 and color index ( @xmath109 ) derived by @xcite : @xmath110\pm 0.033 , \ + 0.0\leq b - v \leq + 1.4,\ ] ] @xmath111\pm 0.043 , \
-0.1\leq b - v \leq + 0.22,\ ] ] @xmath112\pm 0.055 , \ + 0.22\leq b - v \leq + 1.4.\ ] ] for deriving @xmath109 values , we used the following empirical relations of low - mass main - sequence stars ( f0v - k5v ) among @xmath109 , @xmath18 , and [ fe / h ] derived by @xcite : @xmath113 } \nonumber \\ -0.0085\rm{[fe / h]}+0.0199\rm{[fe / h]}^{2 } , \ \ \ \sigma(b - v)=0.039 \rm{mag}~,\end{aligned}\ ] ] where @xmath114 .
+ + the estimated values of the emission flux of ca ii 8542 ( @xmath115 ) and h@xmath1 ( @xmath116 ) are listed in supplementary table 1 of this paper .
the excess equivalent width values ( @xmath117 and @xmath118 ) are also listed in this table . as we have already mentioned in section [ subsec : caiiha ]
, we did not use the values of ca ii 8542 of 7 target superflare stars ( kic7093547 , kic8359398 , kic8547383 , kic9459362 , kic10252382 , kic10387363 , and kic11818740 ) , which are marked in the 8th column of supplementary table 1 .
this is because the exposure time for such stars was not sufficient and the resultant s / n ratio around ca ii triplet lines is low .
+ + in figure [ fig : flux ] , the @xmath119 and @xmath116 are plotted as a function of @xmath31(8542 ) and @xmath31(h@xmath1 ) , respectively .
the error values of @xmath119 and @xmath116 in figure [ fig : flux ] are plotted in the following method . roughly assuming the upper limit of subtraction error
, we here consider the typical error value of the emission flux ( @xmath120 ) is roughly @xmath121% , if excess equivalent width ( @xmath122 ) is larger than 30m ( sufficiently high compared to 61vir ) . on the other hand , if the excess equivalent width ( @xmath122 ) is less than 30m ( not so high compared to 61vir ) , we consider that the typical error value of the emission flux ( @xmath120 ) is about @xmath123% .
we can see a positive correlation between emission flux and @xmath31 index in this figure , and all of the superflare stars with high @xmath31(8542 ) and @xmath31(h@xmath1 ) values show high @xmath119 and @xmath116 values .
this positive correlation show us that by using values of the emission flux , we can conclude the basically same conclusions as we did with @xmath31(8542 ) index in section [ subsec : activity - spots ] . | we conducted high dispersion spectroscopic observations of 50 superflare stars with subaru / hds .
these 50 stars were selected from the solar - type superflare stars that we had discovered from the kepler data .
more than half ( 34 stars ) of these 50 target superflare stars show no evidence of binarity , and we estimated stellar parameters of these 34 stars in our previous study ( @xcite , hereafter called paper i ) . according to our previous studies using kepler data , superflare stars show quasi - periodic brightness variations whose amplitude ( 0.1 - 10% ) is much larger than that of the solar brightness variations ( 0.01 - 0.1% ) caused by the existence of sunspots on the rotating solar surface . in this study , we investigated whether these quasi - periodic brightness variations of superflare stars are explained by the rotation of a star with fairly large starspots , by using stellar parameters derived in paper i. first , we confirmed that the value of the projected rotational velocity @xmath0 is consistent with the rotational velocity estimated from the period of the brightness variation .
next , we measured the intensity of ca ii infrared triplet lines and h@xmath1 line , good indicators of the stellar chromospheric activity , and compared them with other stellar properties .
the intensity of ca ii infrared triplet lines indicates that the mean magnetic field strength ( @xmath2 ) of the target superflare stars can be higher than that of the sun .
a correlation between the amplitude of the brightness variation and the intensity of ca ii triplet line was found .
all the targets expected to have large starspots because of their large amplitude of the brightness variation show high chromospheric activities compared to the sun .
these results support that the brightness variation of superflare stars is due to the rotation with large starspots . |
atomic qubits encoded in hyperfine ground states are one of several approaches being developed for quantum computing experiments@xcite .
single qubit rotations can be performed with microwave radiation or two - frequency laser light driving stimulated raman transitions .
the microwave approach , while simpler in implementation , does not directly yield site resolved operations in a multi - qubit array .
single site selectivity can be achieved with microwaves using a magnetic field gradient@xcite , or tightly focused stark shifting beams@xcite , or with focused two - frequency raman light@xcite . in order to achieve as high a fidelity as possible for qubit rotations
the field strength must be precisely controlled at the location of the atom .
finite temperature position fluctuations or jitter in the alignment of optical beams lead to gate errors due to variation of the optical intensity interacting with the atom .
there is also a second cause of gate errors due to variations in the differential stark shift experienced by an atom as a function of location in an optical trap .
both the spatially varying trap intensity , and the spatially varying raman intensity lead to a variable stark shift , and hence position dependent qubit detuning errors .
while these errors can be minimized by cooling to the motional ground state of the trap , ground state cooling tends to take longer than cooling to a thermal state , and any heating sources lead to motional excitation . in light of these effects
it is of interest to design the experimental control system to minimize errors due to atomic motion or beam misalignment .
it is known that using beams with more uniform intensity profiles can lead to improved fidelity of rabi oscillations@xcite and that specially shaped beams have the potential for reduced crosstalk@xcite . in this paper
we present a detailed analysis of the fidelity of qubit rotations driven by two - frequency raman light .
we investigate how modifying the beam shape from the typical gaussian profile of a laser beam to a super gaussian with a top - hat " like profile can reduce the effects of intensity variations and simultaneously minimize crosstalk to neighboring qubit sites .
numerical simulations are used to quantify the effects of the thermal motion of trapped atoms as well as radial and axial control beam misalignment on the fidelity of rabi oscillations with finite temperature atoms .
we show that the super gaussian beam can dramatically reduce both motional dephasing and crosstalk to neighboring sites .
while we specifically analyze the case of two - frequency raman light we expect that our results will also be applicable to localized gates that rely on microwaves with focused stark shifting beams . in sec .
[ sec : theory ] , we describe how to calculate the three - dimensional ( 3d ) intensity profiles of gaussian and super gaussian beams , define the density weighted intensity variance , discuss the optical trap geometry used in our simulations and review how the two - photon rabi transitions in @xmath3cs depend on the position of the atoms due to the intensity profiles of the trapping and control lasers . in sec .
[ sec : calcs ] , we present our calculation results for the 3d intensity profile of super gaussian beams as well as the effects of atom temperature and control beam misalignment on the density weighted intensity variance and the rabi oscillations of the atomic qubits , before summarizing our conclusions in sec . [
sec : conclusions ] .
in this section , we theoretically investigate the effects of using a super gaussian beam to address the atoms instead of a tem@xmath4 gaussian beam .
first , we define what constitutes a super gaussian beam for this investigation
. then we present a model to calculate the propagation of the super gaussian beam in order to have a complete three - dimensional ( 3d ) mapping of the electric field and intensity distributions . in sec .
[ sec : calcs ] , we use the 3d intensity distributions to quantitatively compare the spatial intensity variations between gaussian and super gaussian beams and their subsequent effects on rabi oscillations of trapped atoms to compare the evolution of the qubit states for super gaussian addressing beams to tem@xmath4 addressing beams . the position - dependent variations in the differential
ac stark shifts result in reduced flopping amplitudes , faster decay of the rabi oscillations , and changes to the rabi frequency due to position - dependent variations in the electric field of the light experienced by the atom .
the dominant contribution to the differential ac stark shift at the location of each trap site is that of the addressing laser beam .
thus , to eliminate the position - dependent differential ac stark shifts of the atomic states , one would ideally use an addressing beam with a uniform intensity over the entire trap volume ; i.e. , a flat - top beam profile .
however , for qubit operations tight focusing is required to address a single qubit without crosstalk at neighboring qubit sites . also , the sharp spatial features required for flat - top beams in the focal plane result in undesired spatial oscillations away from the focal plane . it is therefore necessary to consider both the uniformity of the intensity and spatial crosstalk when selecting an optimized beam profile . if the addressing beam is a tem@xmath4 gaussian beam then the electric field of the beam can be analytically modeled in three dimensions using @xcite @xmath5 \nonumber \\ & & \times \exp \left ( { i { \displaystyle \left [ \frac{k}{2 } \frac{\left ( x^2 + y^2 \right)}{r \left ( z \right ) } -\tan^{-1 } \frac{z}{z_0 } \right ] } } \right ) e^{i\left(kz-\omega t\right ) } , \label{eq : tem00scalare}\end{aligned}\ ] ] where @xmath6 is the amplitude of the electric field at the center of the focal plane , or @xmath7 . because @xmath8 , the corresponding three - dimensional intensity distribution is @xmath9 , \label{eq : igaussxyz}\ ] ] where @xmath10 is the intensity at the center of the focal plane , the gaussian width ( radius where the intensity is @xmath11 of the peak value ) as a function of axial distance from the plane of the beam waist , @xmath12 , is @xmath13 where @xmath14 is the rayleigh range , and @xmath15 is the gaussian beam waist in the focal plane , and the beam radius of curvature is @xmath16 .
super gaussian beams are light patterns whose intensity profiles reside in the regime between smoothly propagating tem@xmath4 gaussian beams and pure flat - top beams . a super gaussian beam is defined here as one whose intensity profile at the beam waist follows the mathematical function @xcite @xmath17 and @xmath18 is the order of the super gaussian .
we also define the phase front in the @xmath19 plane to be planar and perpendicular to the propagation direction . for super gaussian beams
the axial location of the peak intensity , or the axial location of the narrowest beam distribution , is not the same as the location of the planar wave front which we refer to as the `` beam waist '' , the focal plane , or the @xmath19 plane ; as is discussed further in sec . [ sec : sgcalcs ] .
if the super gaussian has order 2 then the beam profile is that of a tem@xmath4 gaussian beam in the focal plane , and the beam propagates exactly as a tem@xmath4 gaussian beam would for axial locations away from the beam waist plane . as the order @xmath18 of the super gaussian increases ,
the effect on the beam waist profile is to widen and flatten the central intensity peak of the beam while increasing the rate of change of the intensity of the sides of the beam as illustrated in fig .
[ fig : sgvsn ] . as we show in
the following a super gaussian with @xmath1 provides significant improvement of gate fidelity and reduction of crosstalk .
although the intensity difference between the @xmath1 beam and @xmath20 or @xmath21 is not more than @xmath22 the performance improves by a much larger factor .
it is therefore important to accurately prepare the desired beam profile .
we discuss possible methods for doing so in sec .
[ sec : conclusions ] . 2 , 4 , 6 , 8 , and 10.,width=264 ] unlike tem@xmath4 gaussian beams where the electric field at any point in space can be analytically determined using eq .
( [ eq : tem00scalare ] ) , an analytical form for super gaussian beams in all space does not exist . the analytical form given in eq .
( [ eq : sggeneral ] ) is only valid for points within the plane containing the beam waist , or @xmath23 . for any points outside of the beam waist plane the electric field at the point of interest
must be determined using beam propagation methods and numerical integration .
scalar diffraction theory can be used to calculate the propagated electric field behind the focal plane .
the model chosen here for determining the electric field at the point of interest is the rayleigh - sommerfeld diffraction integral @xcite .
rayleigh - sommerfeld diffraction is a scalar diffraction model which integrates the known field values over an input plane , @xmath24 , and propagates the field to a particular point of interest , @xmath25 using @xmath26 where @xmath27 is the wave number , and @xmath28 is the distance from the integration point to the point of interest , or @xmath29 in this investigation , it is assumed that the input plane for the rayleigh - sommerfeld diffraction integral coincides with the axial location of the beam waist , and that the phase front of the electric field is planar with the direction of energy propagation to be the @xmath30-direction for all points within the beam waist plane . using these assumptions , the known electric field term in eq .
( [ eq : rsdiffintegral ] ) can be simply written as @xmath31 substitution of eq .
( [ eq : sge0 ] ) into eq .
( [ eq : rsdiffintegral ] ) and numerically integrating over the input plane for a desired point of interest with @xmath32 one obtains the scalar value of the electric field at the point of interest . if the intensity of the addressing beam is not uniform over the entire volume of the trap site then as the atom moves within the trapping volume it samples different intensity values resulting in position - dependent variations in the differential ac stark shifts between atomic states and rabi frequency of atomic transitions . in order to help quantify the total variation of intensities an atom at a particular temperature would be exposed to we can calculate a density weighted intensity variance ( which we will subsequently refer to as the `` intensity variance '' ) , or @xmath33 .
we quantitatively define the intensity variance to be @xmath34 ^ 2 , \label{eq : abivgeneral}\ ] ] where the integration is performed over all space and @xmath35 is the normalized position - dependent intensity of the addressing beam , @xmath10 is the peak intensity in the input beam waist plane , and @xmath36 is the normalized wavefunction of the harmonically trapped atom with temperature - dependent widths @xmath37 , @xmath38 , and @xmath39 , as discussed in the next section . by inspection , we see that if the intensity distribution , @xmath40 , is a uniform constant for all space where the wavefunction @xmath36 is non - zero then @xmath33 would be zero .
thus , the higher the variation of intensity the atom samples during the time which it is exposed to the addressing beam , the larger the value of @xmath33 . for the detailed analysis of the intensity variance , crosstalk , and rabi oscillations
we have assumed blue - detuned optical traps using the 49 site gaussian beam array introduced in @xcite .
each trapping site is formed by the intensity overlap of four gaussian beams , each of power @xmath41 , in a unit cell of the array with area @xmath42 .
the trapping potential is @xmath43 where @xmath44 is the atomic polarizability , and @xmath45 is the intensity found by adding the contributions from the four beams forming a unit cell of the array . following the analysis in @xcite we find the trap spring constants @xmath46 [ eq : kappaxyz ] here @xmath47 with @xmath48 the array period , @xmath15 the waist of each trapping beam , @xmath49 , and @xmath50 is the average intensity of each unit cell .
the corresponding trap frequencies are @xmath51 , where @xmath52 is the atomic mass , and @xmath53 . using the relationship @xmath54 , the time - averaged position variances are @xmath55 [ eq : sigmaxyz ] equations ( [ eq : kappaxyz ] ) and ( [ eq : sigmaxyz ] ) are the same as eqs .
( 10 ) and ( 11 ) from @xcite , except for corrections to eqs .
( 10c ) and ( 11c ) .
the parameters @xmath56 and @xmath57 are defined as @xmath58 where @xmath59 is boltzmann s constant and @xmath60 is the atom temperature .
for our calculations , we used a trap spacing of @xmath61 m , a @xmath62 trap wavelength , a laser beam waist of 1.73 @xmath63 m ( making the normalized array period @xmath64 ) , and a laser power of 3 w , split into 64 equal laser beams so @xmath65 .
the polarizability @xmath66 for the 780-nm trap light was calculated using a standard sum over states expression@xcite including the @xmath67 and @xmath68 levels in cs . using numerical values from @xcite for the transition wavelengths and dipole matrix elements
we find @xmath69 in cgs units and in si units @xmath70 .
a two - level atom interacting with a monochromatic field undergoes rabi oscillations between its two levels .
we will designate the two states as ground ( lower ) state @xmath71 and excited ( upper ) state @xmath72 with energies @xmath73 and @xmath74 , respectively .
writing the state vector as @xmath75 the schrdinger equation in the rotating wave approximation takes the form @xmath76 [ eq : coupledrabi ] with @xmath77 @xmath78 is the optical frequency , @xmath79 , the rabi frequency is @xmath80 where @xmath81 is the matrix element of the dipole operator @xmath82 , and @xmath83 is the elementary charge . solving this system of coupled differential equations results in the solution @xmath84 where @xmath85 is the initial time and @xmath86 & i e^{\imath \frac{\delta ( t+t_0)}{2}}\ , \frac{\omega^*}{\omega'}\sin\left ( \frac{\omega ' ( t - t_0)}{2 } \right ) \cr
i e^{-\imath \frac{\delta ( t+t_0)}{2}}\ , \frac{\omega}{\omega'}\sin\left ( \frac{\omega ' ( t - t_0)}{2 } \right ) & e^{-\imath \frac{\delta ( t - t_0)}{2}}\left [ \cos\left ( \frac{\omega ' ( t - t_0)}{2 } \right ) + i \frac{\delta}{\omega'}\sin\left ( \frac{\omega ' ( t - t_0)}{2 } \right ) \right ] \end{matrix } \right)\nonumber\\ \label{eq : rotationmatrix } \end{aligned}\ ] ] with the effective off - resonance rabi frequency @xmath87 . when the atom is initially in the ground state , the time dependent probabilities to be in the ground and excited states
are found to be @xmath88 we see that the ground and excited state probabilities undergo rabi oscillations with the effective off - resonance rabi frequency @xmath89 . in this work , we investigate the rabi oscillations due to a two - photon stimulated raman transition from the @xmath90 state of @xmath3cs , corresponding to the lower state @xmath71 from sec .
[ sec : rabioscillations ] , to @xmath91 , corresponding to the upper state @xmath72 from sec .
[ sec : rabioscillations ] via the @xmath92 state driven by a pair of 459 nm raman laser beams .
in order to calculate the evolution of the @xmath93 population with time , we need to use the appropriate on - resonance rabi frequency @xmath94 and detuning @xmath95 for this two - photon transition .
as shown in the appendix , the two - photon rabi frequency for a @xmath96 type coupling scheme with resolved excited state hyperfine structure is @xmath97 with @xmath98 where @xmath99 refers to raman laser beams 1 and 2 , respectively , @xmath100 is the detuning of the first raman laser from the @xmath101 ( fine structure level ) transition , and @xmath102 and @xmath103 are the hyperfine shifts from the @xmath92 fine structure level to the @xmath104 and 4 hyperfine states , respectively .
@xmath105 refers to the electric field amplitudes of raman lasers 1 and 2 , respectively .
@xmath106 is the reduced dipole matrix element for the @xmath107 transition in @xmath3cs @xcite , and @xmath108 is the bohr radius . as we discuss in the next section , the detuning , @xmath95 , for our calculation will be due to changes in the differential ac stark shift between the @xmath109 and @xmath93 hyperfine ground states .
there are two contributions to the differential stark shift , one from the raman addressing beams , @xmath110 , and one from the trap light , @xmath111 , so @xmath112 since the raman laser beams are near resonant to the @xmath107 transition , the differential ac stark shift due to the raman beams is ( see the appendix for the derivation ) @xmath113 where @xmath114 ghz is the ground state hyperfine splitting in @xmath3cs . here
, we have assumed that both raman beams have the same power , waist , and alignment , so that @xmath115 .
for the differential stark shift due to the far - detuned 780-nm trap laser , we use the expression @xmath116 where @xmath44 is the polarizability of the @xmath117 state in @xmath3cs in si units and @xmath118 is the electric field amplitude of the trap light .
@xmath119 is the effective detuning of the trap laser from the d1 and d2 transitions in @xmath3cs , given by @xcite @xmath120 with @xmath121 and @xmath122 .
thus far , we have assumed that the intensity of the light field and the atomic energy levels are uniform throughout space . however , the electric field amplitudes @xmath123 and @xmath124 , depend on the position of the atom in the raman beams and in the trap , so @xmath125 therefore , the effective on - resonance rabi frequency @xmath94 and the differential ac stark shifts @xmath110 and @xmath111 due to the raman lasers and the trap light , respectively , depend on the position of the atom . consequently ,
as the atom moves through the trap , @xmath94 and @xmath95 will change with time .
thus , to calculate the rabi oscillations of an atom in our system , eq .
( [ eq : statevector ] ) must be solved numerically in small time steps to account for these changes with time as the atoms are moving .
the electric field strength @xmath126 is related to the light intensity @xmath127 at the location of the atom by @xmath128 for gaussian addressing beams of waist @xmath15 , this intensity is calculated from the laser power @xmath41 at the atoms as @xmath129 with @xmath130 and @xmath131 .
\nonumber%\label{eq : normintgauss}\ ] ] for the super gaussian addressing beams , we used the same maximum intensity , @xmath10 , as for the gaussian beams , multiplied by the numerically calculated normalized intensity profiles , @xmath132 , as described in sec .
[ sec : sgbeams ] . in the calculations , we used raman laser beams of identical power and waists that are perfectly aligned with each other , so @xmath133 . from eqs .
( [ eq : rabifrequency ] , [ eq : ixyznormint ] ) we thus find the position - dependent on - resonance rabi frequency for evaluating eq .
( [ eq : statevector ] ) to be @xmath134 because the atoms are moving , for each time step the position of the atom must be calculated and the on - resonance rabi frequency for that atom determined in order to evaluate eq .
( [ eq : statevector ] ) . by the same means ,
( [ eq : diffstarkraman ] ) , we find that the differential stark shift due to the raman laser beams is @xmath135 the position - dependent differential stark shift due to the trap light can be obtained using @xmath136 and eq .
( [ eq : diffstarktrape ] ) as @xmath137 leading to a total differential stark shift of @xmath138 as for the detuning @xmath95 from eq .
( [ eq : rotationmatrix ] ) , we tune the raman lasers exactly to the @xmath139 to @xmath93 hyperfine ground state transition , taking into account the ac stark shifts from the trap and raman lasers for an atom at the center of the trap and assuming the raman laser beams are perfectly centered
. therefore , @xmath95 will be the difference of the differential stark shift of an atom at position @xmath140 from that at the center of the trap and addressing laser beams .
since the atoms are moving , this needs to be evaluated at each time step .
in this manuscript we want to duplicate the experimental setup used in ref .
@xcite and investigate the effects of using super gaussian beams for the addressing laser instead of a tem@xmath4 gaussian beam .
therefore , the parameters used for the computations for the tem@xmath4 gaussian beam largely match those used in the experiment . for this reason ,
the width of the gaussian beam , @xmath141 , for this investigation was chosen to be 2.30 @xmath142 . as observed in fig .
[ fig : sgvsn ] , when the super gaussian order @xmath18 increases both the region of uniform intensity increases as well as the magnitude of the ramp rate of the intensity outside of the central region . for radial values of @xmath143
the intensity of the beam increases with the order @xmath18 , and for radial values of @xmath144 the intensity of the beam decreases as the order @xmath18 increases .
there is a trade - off to be considered when deciding what width to use for each super gaussian beam order : a larger width increases the size of the central uniform intensity volume , but may increase the crosstalk intensity at the location of a neighboring site in the case of radial misalignment .
all experimental beams have radial and axial misalignment , or jitter , due to experimental conditions ; i.e. , mechanical vibrations in the optical equipment , air currents in the lab room , etc .
if the beam widths for @xmath145 super gaussian beams are chosen such that the crosstalk intensity is equal to that of @xmath2 at @xmath146 , then all super gaussian beams will have much higher variations in the crosstalk intensity due to radial jitter and the steeper intensity ramp rates , an undesired outcome .
thus , we have set the widths of the super gaussian beams such that the crosstalk intensity for each value of @xmath18 is less than , or equal to , that of a centered tem@xmath4 gaussian beam up to a maximum acceptable value of radial jitter , @xmath147 ; i.e. , the worst - case crosstalk scenario for a misaligned super gaussian beam is equal to the best - case scenario of a centered gaussian beam , or @xmath148 or , using eq .
( [ eq : sggeneral ] ) , @xmath149 where @xmath15 is the gaussian beam waist , @xmath48 is the distance to the nearest neighbor , and the maximum acceptable radial jitter , @xmath147 , is chosen to be 150 nm . this value was chosen based on estimates of day to day misalignment observed when taking data in a 2d qubit array@xcite .
equation ( [ eq : wn ] ) yields the width , @xmath150 , for each super gaussian beam .
all of the parameters used for the various addressing beams , as well as the crosstalk intensity values , are summarized in table [ tab : parameters ] .
.list of parameters used for the various possible addressing laser beams .
the width of the super gaussian beams is chosen such that the crosstalk intensity , @xmath40 , at a neighboring trap site is lower than , or equal to , that of a gaussian beam even if the super gaussian beam has a radial jitter up to a maximum allowable value of @xmath147 .
@xmath10 is the intensity at the center of the focal plane of the gaussian beam .
[ tab : parameters ] [ cols="^,^,^,^ " , ] for a gaussian beam , we find that as the axial misalignment is increased to 2.5 @xmath63 m and then 5 @xmath63 m , there is a small drop in differential stark shift and rabi frequencies , which can be explained by the small drop in intensity of the gaussian beam in this axial range ( see fig . [ fig : gsgonaxis ] ) .
the increase in intensity variance ( see fig . [
fig : variancevsz ] ) is so gradual that its effects on the decay time and the @xmath93 populations at @xmath151 and 3@xmath151 times are small to insignificant .
we conclude that gaussian addressing beams are not very sensitive to axial misalignment . for super gaussian raman beams ,
the results follow the same trends as those of the perfectly aligned case for the same reasons . as shown in fig .
[ fig : gsgonaxis ] , the @xmath20 super gaussian has a higher intensity at @xmath152 m than at the focal plane , leading to a larger differential stark shift and increased rabi frequencies , both compared to the gaussian beam and the aligned case .
the @xmath1 super gaussian beam has a flat axial intensity profile , resulting in differential stark shifts and rabi frequencies closest to those for an atom at the center of the trap for aligned raman laser beams .
the @xmath21 super gaussian axial intensity profile is oscillating about the focal plane intensity , leading to results very close to those for an atom at the center for aligned addressing beams .
the decay time of the rabi oscillation amplitude and the @xmath93 populations for orders @xmath153 and @xmath154 match within their uncertainties , while those for @xmath1 are significantly larger .
the decay time for @xmath1 is approximately ten times longer than those for the other orders , stemming from its lower intensity variance as shown in fig .
[ fig : variancevsz ] and thus narrower range of rabi frequencies .
overall , these results indicate , that for the cases explored here , the @xmath1 super gaussian beam would be the ideal choice for raman addressing beams , least sensitive against radial and axial misalignment , providing the most even intensity profile for atoms moving around in the gaussian beam array trap , resulting in rabi oscillations with long decay times and little spread in @xmath93 population .
we also repeated the simulations for raman laser beams with a 100 ghz detuning from the @xmath155 transitions , respectively , and five times the laser power .
this keeps the on - resonance rabi frequency approximately the same while reducing the differential ac stark shift due to the raman lasers by about a factor of six .
the differential ac stark shifts , rabi frequencies , rabi oscillation amplitude decay times , @xmath93 populations at @xmath151 and 3@xmath151 times , and spread in @xmath93 population follow the same trends with temperature and axial misalignment as the 20 ghz results .
however , the effects of radial misalignment on the decay time and the spread of the @xmath93 population were significantly reduced .
this confirms that the spread in rabi oscillation amplitudes ( leading to population spread ) and frequencies ( resulting in rabi oscillation decay ) is caused by the spread in the differential ac stark shift , since that is lower by a factor of six in the 100 ghz case .
this means that an increase in laser power and corresponding increase in laser detuning is also a way to improve the consistency of the resulting quantum operations .
however , laser power is limited , so increasing the detuning may not be possible in a given experimental setup .
thus far , we have assumed that the raman laser beams are free from any laser power noise .
we repeated each simulation for 2@xmath156 laser power noise on the raman laser beams . for each atom in the simulation , a laser power number in the @xmath157 range is randomly generated and stays constant through the whole simulation .
we thus are assuming that the laser noise varies slowly compared to the timescales investigated here ( 0.15 ms ) .
we found that laser power noise has little to no effect on the differential ac stark shift due to the raman lasers , the rabi frequencies , and the @xmath93 populations at @xmath151 and 3@xmath151 pulse times .
however , in some cases the decay times were reduced by factors of two or three with stronger effects for cases with larger decay times .
since the most strongly affected cases were those with large decay times , this reduction does not have a significant effect on the overall performance of quantum operations on the @xmath151 to 3@xmath151 pulse time scales .
thus , the effects of noise are negligible compared to the effects of atom temperature and beam misalignment .
overall , we found that while atom temperature and laser noise impact the rabi oscillations , the most significant factor influencing the effective off - resonant rabi frequency , rabi oscillation amplitude decay time , and @xmath93 population spread is the raman beam alignment .
super gaussian laser beams have a smaller intensity variance , which significantly reduces the sensitivity of the quantum operation performed to alignment .
therefore , the use of super gaussian raman addressing beams may help reduce downtime for repeated beam alignment and improve the consistency of quantum operations .
specifically , the flat radial and axial intensity profile of the @xmath1 super gaussian beam make it ideal as an addressing laser beam for quantum operations on atomic qubits .
we have presented a detailed parametric study of the effects of finite temperature , radial and axial beam misalignment , and laser noise on the fidelity of stimulated raman single qubit gates .
our results show that a high order super - gaussian beam provides more uniform intensity , and greatly increased coherence of rabi oscillations , as presented in tables [ tab : rabitdepgsg]-[tab : rabimisalignaxialgsg ] , together with less crosstalk for optically trapped atomic qubits .
the optimal choice of the super gaussian index @xmath18 will depend on details of the experimental environment .
while the super gaussian beams provide lower intensity variance and less crosstalk for small axial or radial misalignment they perform worse than a gaussian beam for large misalignment . for radial misalignment of not more than @xmath158 and axial misalignment of not more than @xmath159
we find a @xmath1 super gaussian to be near optimal .
these misalignment values are realistic estimates based on recent experiments@xcite . for larger misalignment values a lower order should be chosen as can be seen from fig .
[ fig : gsgxtalkrandz ] . on the other hand ,
if it is possible to ensure very small beam misalignment then the lowest possible intensity variance , and best possible rabi oscillation fidelity , is obtained for @xmath160 as is shown in table [ tab : variancealignednear6 ] .
variations of @xmath18 by @xmath161 about this value have only a small effect on the intensity variance .
it is apparent from our results that the super gaussian beam has only a minor effect on the amplitude achieved in a @xmath151 or @xmath162 pulse , but has a large influence , by up to a factor of 60 , on the decay time of the rabi oscillations .
very high fidelity gates as needed for scalable quantum computation will likely be based on composite pulse techniques@xcite to minimize sensitivity to imperfect control of the pulse area .
since composite pulses imply larger total pulse areas the increase of the decay time afforded by the super gaussian pulse will be particularly useful in conjunction with composite pulses .
we have studied the super gaussian beam profile because it has a compact analytical form and is readily visualized .
this leaves open the question of what the global optimal beam shape might be .
the choice of an optimal shape will strongly depend on the experimental conditions .
if there is no experimental beam misalignment and the atoms are very cold and well localized then a high order beam profile with excellent uniformity is to be preferred .
allowing for misalignment and atomic motion a lower order profile with smoothly varying intensity gives better performance .
the specific choice can not be predicted in general and will depend on actual parameters . in order to reap full advantage of using a specific beam profile
it is of course necessary to generate such a profile experimentally .
there are several methods available for beam shaping of top hat , or similar profiles , including aspherical lenses@xcite , diffractive optical elements@xcite , and computer controlled spatial light modulators ( slms)@xcite . although arbitrary beam shapes with desired amplitude and phase profiles can in principle be produced using a slm in holographic mode there remains the experimental challenge of compensating for imperfections due to optical elements and vacuum windows .
a notable advance was reported recently in @xcite where a slm together with a shack - hartmann wave - front sensor was used to correct for imperfections in the optical train and obtain a uniform array of intensity spots with a standard deviation of only 1.4% .
based on these techniques it should be possible to create essentially any desired beam profile with accuracy at the 1% level . such accuracy together with
the weak sensitivity of the atom averaged intensity variance with super gaussian index presented in table [ tab : variancealignednear6 ] , suggests that performance comparable to that calculated here will be possible in real experiments .
ms and mjp were supported by the iarpa mqco program through aro contract w911nf-10 - 1 - 0347 .
the rabi oscillations investigated in this work are driven via a raman process from the @xmath163 hyperfine ground state of the @xmath117 manifold in @xmath3cs to its @xmath164 hyperfine ground state via the @xmath92 manifold using two laser beams . to treat this kind of rabi oscillation , we repeat the steps from sec . [
sec : rabioscillations ] for a @xmath96 type three - level system with two lasers tuned to the two transitions of the raman process . for detunings large enough so that the excited state population is small
, we can adiabatically eliminate the @xmath92 state , resulting in an effective two - level rabi oscillation with an on - resonance rabi frequency of @xmath165 where @xmath166 are the single photon on - resonance rabi frequencies for the @xmath167 and @xmath168 transitions , respectively .
@xmath100 is the detuning of the first raman laser beam from the @xmath169 ( fine structure level ) transition , and we have assumed that the detuning of the second raman laser from the @xmath170 transition is the same . equation ( [ eq:2photonrabif ] ) is valid for two - photon resonance or when the departure from two - photon resonance is small compared to @xmath100 .
taking into account the hyperfine splitting of the @xmath92 level , we have to sum over all possible intermediate states , resulting in @xmath171 where @xmath172 are the total angular momentum quantum numbers of the intermediate , initial , and final states , respectively , @xmath173 and @xmath174 are the single photon on - resonance rabi frequencies for the @xmath175 and @xmath176 transitions , respectively , and @xmath177 is the detuning of the first raman laser beam from the @xmath175 transition . in @xmath3cs , we have @xmath178 , @xmath179 , and @xmath180
. we thus find for the two - photon on - resonance rabi frequency @xmath181 where we used primes to indicate quantum numbers pertaining to the excited states .
the detunings from the @xmath92 hyperfine states are @xmath182 and @xmath183 . here ,
@xmath102 and @xmath103 are the hyperfine shifts from the @xmath92 fine structure level to the @xmath180 hyperfine states , respectively .
the one - photon rabi frequencies are @xmath184 with @xmath185 , where @xmath186 is the electric field amplitude of raman laser @xmath99 , @xmath187 is the reduced dipole matrix element for the @xmath188 transition , @xmath83 is the elementary charge , and @xmath189 for a raman absorption and @xmath190 for a stimulated raman emission . here ,
@xmath191 and @xmath192 are clebsch - gordan coefficients , and @xmath193 for our specific transitions in @xmath3cs , @xmath194 and @xmath195 , the relevant quantum numbers are the initial and final total electron angular momentum quantum numbers @xmath196 , the nuclear spin quantum number @xmath197 , the initial magnetic quantum number @xmath198 , and we use circularly polarized raman laser beams such that the z - components of the angular momentum of the absorbed and emitted photons are @xmath199 . with these
, we find @xmath200 | we study the fidelity of single qubit quantum gates performed with two - frequency laser fields that have a gaussian or super gaussian spatial mode .
numerical simulations are used to account for imperfections arising from atomic motion in an optical trap , spatially varying stark shifts of the trapping and control beams , and transverse and axial misalignment of the control beams .
numerical results that account for the three dimensional distribution of control light show that a super gaussian mode with intensity @xmath0 provides reduced sensitivity to atomic motion and beam misalignment . choosing a super gaussian with @xmath1 the decay time of finite temperature rabi oscillations
can be increased by a factor of 60 compared to an @xmath2 gaussian beam , while reducing crosstalk to neighboring qubit sites . |
despite of the increasing success of qcd in describing a large variety of phenomena , both in the perturbative as well as in the non - perturbative regimes , some fundamental questions remain unsolved .
prominent examples are the very nature and detailed properties of the strongly coupled quark gluon plasma which is the conjectured state of qcd matter at temperatures comparable and larger than the qcd energy scale @xmath0 .
furthermore , the nature and properties of the chiral and deconfinement phase transition as well as the position of a conjectured critical point ( cp ) in the qcd phase diagram are among the still challenging issues @xcite .
to answer these questions experimentally , a number of large scale experiments is currently running ( alice , atlas and cms at the lhc , and star and phenix at rhic ) , planned ( mpd at nica ) or under construction ( cbm and hades at fair ) .
on the theory side , lattice qcd yields a smooth crossover from the hadronic phase to the quark - gluon phase for small chemical potential at temperatures of about @xmath1 . at sufficiently large net baryon density
the crossover may turn into a first order phase transition at a cp . at non - zero net densities ( _ i.e._non - zero quark chemical potential )
there is no first - principle approach to the phase diagram .
therefore , one has to rely on effective models or on truncation schemes .
nevertheless , many of these approaches seem to point to a first order phase transition connected to the spontaneous breaking of chiral symmetry at densities a few times the nuclear density . the end point of this transition line has interesting properties on its own . from macroscopic examples , the phenomenon of critical opalescence , _
i.e._the diverging scattering strength of transparent media in the vicinity of a critical point , has been known for a long time @xcite .
quite common also is the phenomenon of critical slowing down , _
i.e._the diverging relaxation time at criticality @xcite .
these two examples as well as most of the special properties of critical points have their reason in the diverging correlation length making the system scale free . to understand the mass generation connected to chiral symmetry breaking , several effective models have been constructed with the nambu - jona - lasinio ( njl ) model @xcite and the linear sigma model ( l@xmath2 m ) @xcite being the most prominent ones .
our present investigation is motivated by the question whether penetrating probes reflect directly the phase structure of strongly interacting matter .
we focus here on real photons and select the @xmath3 to mimic the above anticipated phase structure .
the @xmath3 contains quark and meson ( pion and sigma ) fields as basic degrees of freedom , where the fluctuations of the latter ones are accounted for in linear approximation , as in @xcite and the photon field is minimally coupled to the strongly interacting components of the @xmath3 .
there is a large difference in the time scales concerning the strong and the electromagnetic interactions , respectively .
this makes possible separating the two interactions involved .
the strong interaction is responsible for the relaxation towards a local thermal equilibrium as well as to the mass generation via the spontaneous breaking of chiral symmetry .
the electromagnetic interaction with a perturbative radiation field contributes little to this , because its effects are @xmath4 suppressed .
therefore we might calculate the thermodynamics without regarding electromagnetism and use the thermodynamic properties as well as the effective masses of the dressed quarks and mesons later on in the photon emission calculations .
the @xmath3 is a widely used effective model of qcd and has been applied often for studying various aspects of thermodynamics of strongly interacting matter .
it was suggested by gell - mann and levy in 1960 @xcite for studying chiral symmetry breaking . in absence of an explicit symmetry breaking term
, the model has a symmetry and therefore belonging to the same universality class as @xmath5 qcd in the chiral limit @xcite .
this symmetry present at high temperatures is spontaneously broken to a residual @xmath6 symmetry with the three pseudoscalar @xmath7 mesons being the goldstone modes . breaking chiral symmetry explicitly ,
the pions acquire non - zero masses .
besides these satisfying properties there is a close connection to the non - linear @xmath2 model , which in turn is equivalent to leading order chiral effective field theory of qcd .
compared to the njl model the @xmath3 has the advantage of including the mesons directly as dynamic field quanta , making it easier to address their properties .
the @xmath3 lagrangian reads @xmath8 where the dirac field @xmath9 describes a doublet of quarks , @xmath2 corresponds to an iso - scalar and lorentz - scalar field , and @xmath10 describes an iso - vector and lorentz - pseudoscalar field , the latter ones conveniently interpreted as the @xmath2 and @xmath7 mesons . from the lagrangian the thermodynamic potential @xmath11 is constructed via the path integral of the exponential of the euclidean action and evaluated following the procedure described in @xcite for including linearized fluctuations .
first , one integrates over the fermionic fields @xmath9 and @xmath12 .
the remaining path integral corresponds to a purely mesonic theory with a complicated interaction potential , which is approximated by a quadratic one to account for small fluctuations .
the parameters of this quadratic potential are identified with the masses and thermodynamic averages of the meson fields .
this leads to self consistency relations for the masses .
the parameters are fixed by the following requirements : the mass of the pions is set to @xmath13 in vacuum ( @xmath14 ) and the sigma meson mass to @xmath15 .
the effective quark mass in the vacuum is fixed to one third of the nucleon mass @xmath16 , and the parameter @xmath17 is identified with the pion decay constant in vacuum , @xmath18 . with these parameters one obtains the results depicted in fig .
[ fig_thermodyn ] . + figs .
[ subfig_m_pi_lf]-[subfig_m_q_lf ] show contour plots of the masses over the phase diagram .
one notes that the pion mass ( fig .
[ subfig_m_pi_lf ] ) increases with temperature and chemical potential with the strongest change at the phase boundary . the sigma meson mass ( fig .
[ subfig_m_si_lf ] ) on the other hand exhibits a valley of low mass values around the phase boundary and with a global minimum at the critical point . the quark mass plotted in fig .
[ subfig_m_q_lf ] drops from its vacuum value to about @xmath19 . the most drastic change , again , is at the phase boundary , signaling that the mechanism for mass generation is indeed the spontaneous breaking of chiral symmetry within the @xmath3 . because the chiral symmetry is also explicitly broken by a nonzero @xmath20 in the lagrangian
, the quark mass does not drop to zero , but stays finite in the high temperature phase . comparing the meson masses ( _ cf._figs .
[ subfig_m_pi_lf ] and [ subfig_m_si_lf ] ) , one realizes that they are degenerate above the 1st order phase transition curve and the crossover region , respectively , but very different below .
this behavior of the mass difference of these chiral partners is another sign of the chiral symmetry breaking and restoration . for quantifying the size of the critical region the quark number susceptibility
is chosen , since the susceptibility scales with the correlation length whose divergence causes many of the special features of a cp . in fig .
[ subfig_chi_lf ] , @xmath21 is normalized to the susceptibility @xmath22 of a massless ideal fermion gas to scale out trivial contributions .
for calculating photon emission rates , the @xmath3 lagrangian is extended by an electromagnetic sector coupled minimally ( _ cf._@xcite ) to the strongly interacting part .
@xmath23 where @xmath24 is the free photon lagrangian and @xmath25 denotes the photon field .
photon emission rates are , in a kinetic theory approach , convolutions of squared matrix elements @xmath26 and phase space distribution functions @xmath27 , the latter ones explicitly depending on @xmath28 and @xmath29 . superimposed
are implicit @xmath28 and @xmath29 dependencies from the effective masses of the involved fields , as displayed in . given the marked variations of these masses one can expect an pronounced impact on the emission rates @xmath30 owing to the weakness of the electromagnetic interaction we restrict the calculations to first order in the electromagnetic coupling .
since we expect to have captured the dominant part of the strong interaction in the calculation of the thermodynamic potential and the effective masses , the residual interaction is expected to be relatively weak .
therefore we restrict our calculation to 1st order processes in the quark - meson coupling . within this approximation
the contributing processes are the tree - level processes in the @xmath31 , @xmath32 and @xmath33 channels . in
, four of the nine integrations can be carried out exactly applying the delta distribution .
another ( angular ) integration drops out by symmetry reasons , so one is left with four integrals , which have to be executed numerically resulting in the rates depicted in fig .
[ fig_rates ] . + when photon energies @xmath34 are much larger than the respective masses , it is not expected to see much of the details of the phase structure .
contrary , at lower energies there are huge differences in available phase space and matrix elements squared leading to pronounced patterns which reflect phase diagram features , in particular the effective masses . for this reason
@xmath35 is chosen .
figure [ fig_rates ] shows contour plots of the photon rates for the different contributing processes over the phase diagram . in figs .
[ subfig_rate_qp_gq_omega=0010 ] and [ subfig_rate_qs_gq_omega=0010 ] we see an enhancement in the crossover region and in fig .
[ subfig_rate_qs_gq_omega=0010 ] a global maximum in the critical region . in figs .
[ subfig_rate_qq_gp_omega=0010 ] and [ subfig_rate_qq_gs_omega=0010 ] one notices large rates in the chirally restored phase and much less photon emission in the chirally broken phase , which in case of [ subfig_rate_qq_gs_omega=0010 ] is superimposed by an island of enhanced rates for @xmath36 and @xmath37 .
figures [ subfig_rate_qq_gp_omega=0010 ] and [ subfig_rate_qp_gq_omega=0010 ] show photon rates from processes involving pions . pions exhibit a large mass difference between the two phases , but contrary to the sigma meson whose mass has a global minimum at the cp the pion mass does not show special features at this point .
this leads to a large difference in the emissivity between the phases but no features characteristic for the cp itself . for the pion - involving compton process ( fig . [ subfig_rate_qp_gq_omega=0010 ] )
there is an enhancement in the crossover region .
this is probably due to a combination of phase space effects and the ( comparatively ) large probability for the internally propagating pion to get on - shell .
a better channel for obtaining signatures of a cp are sigma involving processes .
this is expected , since the sigma meson is precisely the mode getting massless at the cp making long range interactions possible and thus driving the critical processes .
unfortunately , the inclusion of linearized fluctuations increases the sigma mass , so it is not clear whether the endpoint of the 1st order phase transition shows correctly the critical behavior .
but linearizing fluctuations anyhow restricts to small fluctuations making it not adequate very near the cp .
nevertheless the sigma mass drops to small values in the critical region , which has a notable effect on the corresponding processes , _ e.g._the excess of the photon rate in the critical region in fig .
[ subfig_rate_qs_gq_omega=0010 ] .
there is a large difference in the rates for the processes under consideration , even between the corresponding compton ( figs .
[ fig_rates](b ) and [ fig_rates](d ) ) and annihilation ( figs .
[ fig_rates](a ) and [ fig_rates](c ) ) processes . these can be understood in terms of available phase space in combination with thermal suppression . within a boltzmann approximation two of the remaining integrals in can be solved to obtain @xmath38 the difference between the minimal kinematically allowed value of the center of mass energy @xmath39 for the different processes , together with a small value of @xmath34 leads to the huge thermal suppression at small @xmath28 seen in figs .
[ fig_rates ] ( a ) and ( d ) .
focusing on soft - photon emission rates we demonstrate that some features of the phase diagram provided by the linear sigma model are nicely mapped out . being aware of some limitations , such as the restriction to linearized fluctuations ( _ cf._@xcite for a proper account of fluctuations ) and the need to implement more complete rates in a model of space - time evolution of the matter
, we hope that improved calculations can provide useful complementary information on strongly interacting matter produced in the course of relativistic heavy - ion collisions at various energies , system sizes and centralities . | soft - photon emission rates are calculated within the linear sigma model .
the investigation is aimed at answering the question to which extent the emissivities map out the phase structure of this particular effective model of strongly interacting matter . |
the possibility of manipulating magnetization by spin - currents in a very efficient way is a key requirement for the design of novel spintronic devices @xcite , which promise to change the way digital information is processed and stored . in particular ,
the advantageous scaling of current - induced spin manipulation compared to the oersted field - induced switching allows for lower power operation at small design rules .
the driving mechanism behind current - induced magnetization dynamics pioneered for use in metallic ferromagnets in the last twenty five years has been the spin - transfer torque @xcite .
however , recently a novel and possibly more efficient approach to current - driven magnetization manipulation has been developed .
in particular , very efficient current - induced spin dynamics has been observed in multilayer systems with an ultra - thin ferromagnetic layer sandwiched between two different non magnetic materials @xcite .
these new current - induced torques originate from spin - orbit effects ( at least one of the two non - magnetic layers consists of an heavy metal ) and so they are referred to as spin - orbit torques ( sots ) @xcite .
two different origins have been suggested for the sots .
one is the spin accumulation induced at the [ heavy metal]@xmath0ferromagnet interface due to the bulk spin - hall effect ( she ) in the heavy metal layer @xcite .
a second possible origin of the sots is the inverse spin - galvanic effect ( isge ) @xcite , which generates a non - equilibrium spin - density at both the top and the bottom interfaces of the ferromagnet .
both effects are expected to induce an effective torque whose strength is a function of the ferromagnetic layer thickness .
+ in the same kind of materials stacks the presence of topologically non - trivial spin textures , such as homo - chiral domain walls has also been observed @xcite .
chiral domain walls are reported to be very stable against annihilation @xcite and to exhibit interesting transport properties @xcite , thus being very promising for technological applications .
the origin of these chiral spin structures is the interfacial dzyaloshinskii - moriya interaction ( dmi ) @xcite .
the dmi is expected to primarily originate from the interface between the heavy metal and the ferromagnet , predicting a direct scaling of its effective strength with the inverse of the ferromagnetic layer thickness @xcite . accordingly , both the dmi and the sots are expected to depend strongly on the materials system as well as on the layers thickness .
so , the ferromagnetic layer thickness dependence of dmi and sots is directly linked to their origin . a pure interface - like effect on one side and a more complicated mechanism which includes bulk - like processes on the other
are expected to scale differently with the thickness of the ferromagnetic layer @xcite . in particular , there are predictions that the dmi and the isge can have a common origin , which should result in a similar dependence on the thickness @xcite .
so , studying the sign and amplitude of dmi and sots as a function of the ferromagnetic layer thicknesses is a key necessity to reveal their origins .
+ materials stacks with a strong dmi as well as large sots are particularly promising , due to their rich physics as well as the possibility to be used in novel spintronic devices @xcite .
one of the most promising of those materials stacks is the trilayer pt@xmath0co@xmath0alo@xmath1 @xcite .
several manuscripts have reported the characterization of the dmi @xcite and of the sots @xcite in this system .
however , reports from different groups for nominally identical samples have often shown contradicting results possibly originating from different thicknesses and growth conditions and no systematic study has been provided to date . in particular , varying the co thickness is a key challenge as it potentially allows to distinguish between effects from the interface with the pt and the alo@xmath1 and from the bulk of the materials .
accordingly , a systematic and combined study of both the dmi and sots as a function of co thickness is the key step needed to understand and tailor the spin - orbit effects in this system .
+ here we report on the characterization of the dmi as well as the sots in identical samples of pt@xmath0co@xmath0alo@xmath1 .
we extract the sign and the magnitude of dmi and sots as a function of the co layer thickness combining two key techniques , namely : current - induced dw motion ( cidwm ) and second harmonic hall measurements .
a detailed study of cidwm in magnetic tracks is presented , which allows us to characterize the thickness dependence of the dmi and of the effective torque driving the dw motion .
this is complemented by an in depth characterization of the sots employing second harmonic measurements in a hall bar geometry .
comparing the thickness dependence of the dmi and the torques allows us to draw conclusions about their origin .
2 we describe the experimental techniques and the corresponding set - ups used in our study . in sec .
3 we report the characterization of current - induced dw motion in magnetic nanowires . sec .
4 reports the dmi values extracted from the analysis of this dw motion data .
5 presents the measured thickness dependence of sots . in sec .
6 we present a discussion of the main results , where we discuss the ferromagnetic thickness dependence of the dmi and the sots and compare to literature .
the material system that is employed for the patterning of the magnetic devices is the multilayer : ta(4.0)@xmath0pt(4.0)@xmath0co(0.81.8)@xmath0alo@xmath1(2.0 ) ( all thicknesses in nm ) .
the stack was deposited by magnetron sputtering technique on a si@xmath0sio@xmath2 substrate .
the magnetic layer consists of a wedged co layer , with a nominal increase in thickness of 1 nm over a length of 2 cm in one of the two in - plane directions ( @xmath3-direction in the following ) .
the wedged co layer is obtained by an _ in - situ _ moving shadow mask @xcite . +
before patterning , the material stack is characterized by brillouin light scattering ( bls ) technique @xcite and magneto - optic kerr effect ( moke ) magnetometry .
the stack exhibits perpendicular magnetic anisotropy ( pma ) , with a spontaneous magnetization pointing along the out - of - plane ( oop ) direction ( see fig .
[ fig_characterization ] ) .
the observed pma is induced by a strong interface anisotropy at the pt@xmath0co interface @xcite . the effective anisotropy energy density , @xmath4 , is found to decrease with increasing co thickness ( see fig . [ fig_k_eff ] ) , consistent with the presence of an interface anisotropy of @xmath5 mj / m@xmath6 and a saturation magnetization of @xmath7 a / m , both obtained by bls measurements @xcite .
this is consistent with the results of magneto - optic kerr effect ( moke)-magnetometry measurements shown in fig .
[ fig_moke_ptcoalox ] .
hysteresis loops for an applied oop magnetic field , @xmath8 , are measured at different positions on the surface of the material stack with different @xmath9 .
the coercive field is observed to decrease with increasing @xmath9 , in agreement with previous observations @xcite .
the loss of pma with increasing @xmath9 explains the decreasing coercive field as a function of an increasing co thickness visible in fig .
[ fig_moke_ptcoalox ] . after magnetic characterization
, the wedge sample is patterned into several devices with different thicknesses of the co layer .
the patterned devices consist of an array of several nanowires ( nws ) in parallel ( 1.5 - 2.0 @xmath10 m in width and 25 - 28 @xmath10 m in length , see fig .
[ fig_cidwm_protocol ] ) used for current - induced domain wall motion ( cidwm ) experiments , and in hall - crosses ( 1 - 2 @xmath10 m in width and 50 @xmath10 m in length , see fig .
[ fig_sh_setup ] ) used for the measurements of effective spin - orbit fields by the second harmonic ( @xmath11 ) technique @xcite .
the devices are patterned by electron - beam lithography and ar - ion milling , at different positions on the sample surface corresponding to different @xmath9 .
+ in both experimental setups a 50 @xmath12-resistor is used to terminate the circuit to ground .
an oscilloscope is used for measuring the pulse waveform , across its 50 @xmath12-internal resistance , @xmath13 .
the total current flowing through the device is obtained by measuring the voltage across @xmath13 . for the evaluation of the current density , @xmath14 ,
the nominal thicknesses of the layers are used . in these type of thin film systems , the resistivity of the ta layer
is known to be around 4 - 5 times larger than is for pt , while the co layer and the pt layer have a similar resistivity value @xcite .
accordingly , the calculated current densities are obtained considering the 4 nm - thick ta bottom layer equivalent to a 1 nm - thick pt layer .
the conventional current density @xmath15 is taken to be positive when it flows in the @xmath16-direction ( see fig . [ fig_setups ] ) , corresponding to an electron current density @xmath17 flowing in the @xmath18-direction .
+ concerning the cidwm experiment , the magnetic configuration of the wires is imaged by a wide field kerr microscope in the polar configuration @xcite .
a magnetic coil is used for the generation of an external in - plane magnetic field .
the experiments are carried out at t=300 k. the second harmonic measurements are carried out in hall - crosses ( see fig .
[ fig_sh_setup ] ) .
a small - amplitude sinusoidal ac current is applied with a frequency ( @xmath19 ) of 13.7 hz .
this induces periodic oscillations of the magnetization about its equilibrium direction .
these periodic oscillations can be attributed to the effective fields generated by the injected current .
the periodic oscillation of the magnetization results in an oscillation of the hall resistance .
the resulting hall voltage has a second harmonic component ( @xmath11 ) that relates directly to the current - induced fields @xcite .
the measurement is performed in two schemes : longitudinal ( x - axis ) and transverse ( y - axis ) . in both schemes ,
the hall voltage is measured during a sweep of the in - plane magnetic field ( -400 mt to + 400 mt ) . in the longitudinal ( transverse ) scheme
the direction of the magnetic field is applied along ( perpendicular to ) the direction of the injected current .
it is important to note that the hall voltage also includes contributions from the planar hall effect @xcite , nernst effect and joule heating @xcite .
these effects are taken into account to extract artifact - free current induced effective fields .
the first type of experiment reported here is the study of current - induced dw motion ( cidwm ) .
this allows us to establish if either the standard spin - transfer torque or the sots are the main driving mechanisms behind dw motion in this materials system and gauge the strength .
furthermore , the dw motion study allows one to determine if chiral dws are present in the system , to extract their chirality and finally to obtain the sign and magnitude of the dmi for each investigated device .
current - induced dw motion experiments are carried out in four different devices .
the nominal thickness of the co layer in the different devices is : 0.93 nm , 1.31 nm , 1.37 nm and 1.43 nm .
the measurement protocol is described in fig .
[ fig_cidwm_protocol ] .
first , the magnetic wires are saturated in the ( @xmath20 , @xmath21 ) or ( @xmath22 , @xmath23 ) magnetization state by an external out - of - plane magnetic field .
second , the magnetization is reversed in all the nws by the switching process presented in previous papers @xcite , where each combination ( parallel / anti - parallel ) of field , @xmath24 , and current , @xmath14 , corresponds to a specific final state of the magnetization in the nws . as a result ,
a dw ready to be displaced is obtained in each nw .
the differential kerr microscopy images at the bottom left ( right ) of fig .
[ fig_cidwm_protocol ] show the cidwm experiment for @xmath25- ( @xmath26- ) dws . at @xmath27 dws
are nucleated . at @xmath28 the same dws
are moved by the injection of a train of current pulses .
+ the average dw velocity is extracted as the ratio between the total displacement , @xmath29 , visible in the differential kerr microscopy images and the total pulsing time , @xmath30 .
the total pulsing time is given by the total number of pulses , @xmath31 , times the single pulse length , @xmath32 ( full width at half maximum , 10 - 15 ns ) : @xmath33 .
two consecutive current pulses are separated by 1 ms , in order to fully magnetically relax the domain wall in between pulses to have reproducible conditions . the average dw velocity , @xmath34 , as a function of the current density , @xmath14 , is reported in fig .
[ fig_vdw - ja ] , for the four different devices .
the measurement is carried out for both @xmath25-dws ( red dots ) and @xmath26-dws ( blue dots ) , for positive and negative @xmath14 .
+ in all the four devices , @xmath25- and @xmath26-dws are observed to move in the same direction of the conventional current , @xmath15 , at approximately the same speed ( within the error bars ) . the critical current density for inducing significant dw motion in each device is observed to be in the range @xmath35 a / m@xmath6 , in line with what has been reported in previous works on the same materials system @xcite . from the result that both domain wall types ( @xmath25 and @xmath26 ) move against the electron flow
, we can conclude that sots are the dominating torques responsible for the dw displacement and that the dws are all homo - chiral due to a finite dmi .
next , we use cidwm to determine the dmi in these samples . to this end
, the dw velocity is measured as a function of an applied magnetic field along the wire axis ( @xmath36 ) for a fixed current density .
the measurement protocol is the following .
first , one type of dw ( @xmath25 or @xmath26 ) is nucleated in each pre - saturated nw by current - induced magnetization switching @xcite . a typical switching pulse used in the experiment has a current density amplitude of @xmath37 a / m@xmath6 and a time duration of @xmath38 ns , assisted in the switching process by a fixed external longitudinal field of about 50 mt .
once the dws are nucleated , they are displaced by the injection of a burst ( n=1 - 50 ) of current pulses with a measured duration of 15 ns . + the motion of @xmath25-dws in the presence of an applied longitudinal field
is shown in fig .
[ fig_vdw - hx_protocol ] .
the total dw displacement is strongly affected by the presence of a finite @xmath24 .
a positive field makes the @xmath25-dws move slower , while a negative field makes them move faster . .
the velocity of the dws is observed to be reduced for @xmath39 and increased for @xmath40 , for both @xmath41 and @xmath42 .
the gold dashed lines indicate the initial position of the dws .
the gold arrows indicate the dw direction of motion .
the injected current density is @xmath43 @xmath44 , while the longitudinal field amplitudes are : @xmath45 mt , @xmath46 mt , @xmath47 mt .
the yellow numbers indicate how many pulses are used to generate the shown dw displacements .
the different dw displacements in the two wires reflect the statistical spread of our observations .
the images show current - induced motion of @xmath25-dws in the device with @xmath48 nm.,width=302 ] the measured average dw velocities as a function of the longitudinal field , @xmath49 , for all the devices are reported in fig .
[ fig_vdw - hx ] ( symbols ) .
red ( blue ) symbols refer to @xmath25- ( @xmath26- ) dws , while squares ( stars ) refer to @xmath41 ( @xmath42 ) . + as visible in fig .
[ fig_vdw - hx_protocol ] and fig .
[ fig_vdw - hx ] , while at zero - field the velocity of both types of dws is the same , in the presence of a finite longitudinal field the two types of dws move at different velocities .
the change in the field amplitude affects differently the velocity of the two types of dws , making it possible to obtain @xmath25-dws and @xmath26-dws moving in opposite directions , when the field amplitude is large enough .
a symmetric behavior is observed for the velocity of the two dw types with respect to @xmath24 , which can be described as : @xmath50 .
+ considering an @xmath25-dw , a sufficiently large positive @xmath24 slows them down , while a negative @xmath24 speeds the walls up . for very large positive @xmath24 the @xmath25-dws
are also observed to change their direction of motion
. however , differently from the case of the ta@xmath0cofeb@xmath0mgo system @xcite , it is not possible in this case to access the regime of fast dw motion in the reversed propagation direction . indeed , for in - plane fields larger than the ones shown in the graphs , local spontaneous magnetization reversal events @xcite start to occur , making the dw motion measurement not possible anymore .
the amplitude @xmath24 at which domains start to nucleate is observed to decrease with an increasing @xmath9 .
this results in a reduction of the range of in - plane fields which can be used for the investigation of dw motion , moving from the thinnest to the thickest device .
+ there is no single simple reason why the nucleation probability is observed to increase with increasing @xmath9 , but different effects can play a role .
first , an increasing @xmath9 results , for ultra - thin layers like the ones under investigation , in an increasing conductivity of the ferromagnetic layer @xcite . accordingly
, the observed increasing nucleation probability could be linked with the increase of the current flowing in the ferromagnetic layer , due to an increasing joule heating produced directly in the ferromagnetic material @xcite . secondly , as reported in fig .
[ fig_characterization ] , the magnetic anisotropy of the magnetic layer decreases as @xmath9 increases .
thus , the required amplitude of the in - plane field , at a fixed current density , for driving the magnetization reversal in the nws is expected to decrease with an increasing thickness @xcite .
the nucleation process is observed to be particularly prevalent in the thickest investigated device , where already at applied fields of 10 - 20 mt reverse domains start to appear in the nws after the current pulse injection .
this results in the impossibility of observing any reversal in the dw motion direction , as shown in fig .
[ fig_vdw - hx_5 ] .
+ all that has been described above concerning @xmath25-dws is equally valid for @xmath26-dws for a symmetric reversal of the field @xmath24 . in this case , a positive in - plane field speeds up dw motion , while a negative field slows them down ( see fig .
[ fig_vdw - hx ] ) .
+ these observations suggest strong spin - orbit torques acting in the materials stack , in combination with the presence of an interfacial dzyaloshinskii - moriya interaction @xcite .
having established that both sots and dmi are present , the key task is to determine the sign and the strength of dmi and sots as a function of the co - layer thickness , as reported next .
we use the cidwm results to determine the dmi present in the different devices . in order to do so , the so - called stopping fields for the @xmath25-dw and @xmath26-dw need to be extracted .
the stopping field is the external longitudinal field which needs to be applied in order to make the dw stop moving @xcite .
this can be extracted from the graphs shown in fig .
[ fig_vdw - hx ] .
+ it is known that dmi stabilizes nel dws with the same chirality @xcite .
this can be interpreted as due to the presence of an effective dmi field , @xmath51 , along the longitudinal direction .
accordingly , the stopping field is the field needed to counteract the dmi field and turn the dw spin texture into the bloch configuration , making the spin - orbit torques acting on the dw zero .
+ the common approach used so far in order to extract the stopping fields from the cidwm graphs is based on a linear fitting of the high velocity data points ( for positive and negative @xmath24 ) , for each @xmath52 combination @xcite .
the crossing point between the zero - velocity axis and the fitting curve would define the stopping field .
however , it is not possible to use this approach in the present case .
the reason is the lack of high velocity data points for one of the two field signs ( positive , for @xmath25-dws ; negative , for @xmath26-dws ) in the reported graphs , as discussed above .
a simple linear fitting of the high velocity data points ( see fig .
[ fig_vdw - hx_linfit ] ) results in the extraction of stopping field values that are outside the observed pinning ranges in fig .
[ fig_vdw - hx ] . according to that , an alternative approach to extract the stopping fields is used .
-dw with respect to @xmath24 , for the four studied devices .
the reported data points refer to measurements with @xmath53 , while @xmath24 is parallel to the intrinsic nel - component of the dw internal magnetization .
the lines are the linear fitting curves for the experimental data .
the slope is found to increase with increasing @xmath9.,width=302 ] + the range of in - plane magnetic fields where the dws move at low velocities and become pinned is centered around the stopping field .
indeed , the damping - like sot driving the dw motion is proportional to the nel component of the dw , which in turn results from the interplay of the external field and the dmi field .
so , the stopping field where the dw does not feel a force is equal in magnitude and opposite in sign compared to the dmi field @xcite . as a consequence , the pinning range can be expressed as : @xmath54 $ ] ; where @xmath55 is the dmi effective field for @xmath56 dws , @xmath57 the pinning range of longitudinal magnetic fields .
accordingly , the dmi effective field can be extracted as the center of the observed pinning ranges of @xmath24 .
finally , the strength of the dmi can be obtained by the equation @xmath58 @xcite .
+ to determine the two stopping fields , @xmath59 and @xmath60 , we first ascertain the magnetic field value , @xmath61 , to which corresponds the highest measured @xmath34 in the reversed direction .
this velocity is used as the threshold velocity , @xmath62 , which defines the boundary between the low velocity regime ( pinning regime ) and the high velocity regime . secondly , the data point with the same absolute value of the velocity , but with opposite sign is determined , together with the magnetic field at which it occurs , @xmath63 .
these two values of the applied magnetic field are used as extremes of the pinning range , and the corresponding stopping field , @xmath64 , is calculated as the arithmetic mean of the two .
two stopping fields are calculated for each type of dw ( when this is possible ) , corresponding to @xmath53 ( @xmath65 ) and @xmath66 ( @xmath67 ) , and the average stopping field for each dw type is obtained as @xmath68 . + the described protocol for the extraction of the effective dmi field is now applied to the experimental data reported in fig .
[ fig_vdw - hx_1 ] , referring to the magnetic device with @xmath48 nm . the extracted stopping field for the @xmath25-dw is @xmath69 mt ( error defined as @xmath70 ) ; while for the @xmath26-dw @xmath71 mt . since @xmath72 , an effective dmi field of @xmath73 mt is obtained .
+ in order to extract the effective dmi coefficient @xmath74 , the dw width parameter needs to be calculated first .
in the case of the thinnest device , @xmath75 nm ( where the used value of the exchange stiffness for the co layer @xmath76 j / m is chosen as reported in literature @xcite ) .
accordingly , the effective dmi results to be @xmath77 mj / m@xmath6 .
+ the same process is repeated for @xmath78 nm and @xmath79 nm , extracting the respective effective dmi field and dmi strength , which are reported in fig .
[ fig_dmi_1 ] and fig .
[ fig_dmi_2 ] , respectively .
for the device with thickness of @xmath80 nm it is not possible to extract a reliable dmi value .
the lack of observations of dw motion in the pinning regime , does not allow to employ the described dmi extraction method for that specific device .
the injection of an in - plane current through a nm@xmath0fm hetero - structure generates two different types of torques : a damping - like ( dl ) torque and a field - like ( fl ) torque , corresponding to two current - induced effective fields : @xmath81 and @xmath82 , respectively .
these torques are responsible for the efficient current - induced domain wall motion and magnetization switching observed in those hetero - structures @xcite . here the thickness dependence of sots is obtained by two complementary approaches : on the one hand , the effective field moving the dws is extracted by analyzing the cidwm data by a collective - coordinate model ; on the other hand , both the dl - field and the fl - field are characterized by second harmonic hall measurements . to determine the spin - orbit torques acting on the dw , a collective - coordinate model ( ccm ) @xcite based on the extension of the one - dimensional model ( 1 dm ) is employed to reproduce the experimental observations reported in fig .
[ fig_vdw - hx ] ( see appendix a for more details ) . in the framework of the ccm
, the dw dynamics is described by three degrees of freedom : the position of the dw in the track , @xmath83 , the in - plane angle of the dw magnetic moment with respect to the @xmath3-axis , @xmath84 , and the angle defined by the normal to the dw surface with respect to the @xmath3-axis , @xmath85 , describing the tilt of the dw plane . + the action of the sots is equivalent to the presence of two effective magnetic fields : @xmath81 and @xmath82 @xcite . in the present work the fl - field
is defined as @xmath86 , with @xmath87 being the proportionality factor between the two effective fields . however , in this section only the zero fl - field scenario ( @xmath88 ) is discussed , due to the fact that the dl - sot is the main source of the domain wall motion . in particular , we find that the same final dw velocity is predicted by the ccm calculations with or without the inclusion of a finite fl - field into the calculations ( more details can be found in appendix b ) .
+ the resulting ccm fitting curves are shown in fig .
[ fig_vdw - hx ] ( solid lines ) .
as it can be seen , they reproduce the experimental data well .
the effective field ( @xmath81 ) is used as a free parameter in the calculations , while the dmi values used for the fitting procedure are the experimentally extracted ones . by using the best fitting curves ,
the amplitude and sign of the effective sot - field is extracted , for each device . the corresponding current - field efficiency , @xmath89 , is reported in fig . [ fig_1dm_sots ] as a function of the co thickness .
the sign of the effective field is in agreement with a positive spin - hall angle , if the she is assumed as the main source of the observed sot .
this is in agreement with what previously reported in literature for pt - based systems @xcite . + as shown in fig .
[ fig_1dm_sots ] , the extracted effective field is observed to increase for a co thickness between 0.9 and 1.4 nm and then to level off around @xmath90 [ mt/(@xmath91a / m@xmath6 ) ] for larger thicknesses .
this indicates that the effective sot - field generating the observed dw motion is initially scaling with the ferromagnetic layer thickness and then becomes independent of it .
, as extracted by the 1 dm calculations .
the values of @xmath89 shown here are the ones used for the generation of the fitting curves reported in fig .
[ fig_vdw - hx].,width=302 ] next , we use the established @xmath11 technique @xcite to study the current - induced sots in pt@xmath0co@xmath0alo@xmath1 hall cross devices . in fig .
[ fig_sots ] the measured current - induced effective fields along the longitudinal ( fig .
[ fig_sl - sot ] ) and transverse ( fig .
[ fig_fl - sot ] ) direction are shown with respect to the injected current , for one of the studied devices .
both effective fields are found to scale linearly with the current amplitude .
+ in order to learn more about the microscopic origins of sots in the present materials system , their ferromagnetic thickness dependence is studied .
accordingly , the current - induced sots are measured for different devices with a co thickness of : 0.93 nm , 0.99 nm , 1.31 nm 1.37 nm and 1.43 nm . + the measured effective fields as a function of the co thickness are reported in fig .
[ fig_sot - vs - tco ] .
first of all , the two effective fields are found to have the same qualitative dependence on the co thickness , suggesting a possible common origin for such a dependence .
second , the two effective fields are observed to first increase and then decrease with the co thickness , clearly indicating a more complicated character than a simple interface - like one .
indeed , in the pure interface - like case we would observe a simple 1/t dependence , which is clearly not the case here .
third , the dl - field is found to be always larger than the fl - field , for all the investigated co thicknesses . in general , in [ heavy metal]@xmath0ferromagnet systems the primary origin of the dl - torque is usually attributed to the she due to the large soc characterizing the heavy metal @xcite .
accordingly , the spin - hall effect is most probably an important source of both sots measured here , where the weaker fl - field is generated by the precession , around the exchange field , of the itinerant spins diffusing in the ferromagnetic layer @xcite .
this interpretation is also in agreement with the observed non - monotonous thickness dependence .
the she - induced spin - current diffuses in the ferromagnet and interacts with the local magnetization by generating the two sots .
the length scale defining the thickness dependence of the corresponding effective fields is the transverse spin diffusion length in co @xcite , reported to be around 1.2 nm @xcite .
indeed , after diffusing across the ferromagnet for a distance equal to the transverse spin - diffusion length , the spin - current is absorbed and no further effect on the magnetization beyond this thickness is produced .
accordingly , in the first spin diffusion length the effective fields build up , then decay with a further increase of the co thickness . while this qualitatively describes the measurements , we can not rule out further origins that would be influenced by an identical effect , resulting in the similar thickness dependence reported here . ,
black squares ) and transverse ( @xmath82 , red dots ) effective fields as a function of the ferromagnetic thickness.,width=302 ]
in multilayer systems like the one discussed here , the dmi is predicted to originate from the interface between the heavy metal ( pt ) and the ferromagnetic material ( co ) @xcite .
accordingly , the dmi is expected to be an interface - like effect , where its effective strength scales with the inverse of the ferromagnetic thickness . in fig .
[ fig_dmi_1 ] the measured dmi fields ( blue spheres ) are shown to be proportional to @xmath92 .
furthermore , the extracted values of @xmath93 are in agreement with what has already been reported in literature for cidwm experiments @xcite .
safeer et al .
@xcite reported a 100 mt stopping field for dw motion parallel to the current flow in a pt(3 nm)@xmath0co(0.6 nm)@xmath0alo@xmath1(2 nm ) sample .
ryu et al .
@xcite reported a 140 mt stopping field in pt(1.5 nm)@xmath0[co(0.3 nm)@xmath0ni(0.7 nm)@xmath0co(0.15 nm ) ] nws .
accordingly , the range of stopping fields reported here is in line with previously reported values for other material systems with a pt buffer layer , even though in those previous reports no systematic thickness dependence , as we provide here , is given .
[ fig_dmi_1 ] we also show the calculated dw width parameter , @xmath94 ( red pentagons ) , as a function of @xmath92 . using the extracted @xmath93 and the calculated @xmath95 for calculating the dmi strength , results in a dmi that scales with the inverse ferromagnetic thickness ( see fig . [ fig_dmi_2 ] ) .
this is in agreement with an interfacial dmi scenario , where the effective dmi strength is expected to be proportional to @xmath96 : @xmath97 , with @xmath98 being the pure interfacial dmi .
however , a linear fitting of the data points ( not shown here ) does not generate a linear curve that crosses the origin of the axes .
the crossing for @xmath99 would happen at a finite value of the ordinate axis .
these observations are in agreement with what reported in ref .
@xcite for a pt@xmath0co system , and in ref .
@xcite for a pt@xmath0ni@xmath100fe@xmath101 system .
cho and co - authors @xcite observed a linear dependence of the effective dmi strength on @xmath92 for @xmath102 nm@xmath103 ( similar to the present study ) , and a rapid drop of it for @xmath104 nm@xmath103 , going towards zero at @xmath99 .
nembach and co - authors @xcite obtained similar results and attributed their observations to a non - trivial dependence of the interfacial dmi strength , @xmath98 , on @xmath105 . in the case
reported here , devices with thicknesses larger than the studied ones exhibit a weak pma and so the formation of a multi - domain state .
this prevents the study of cidwm and so the extraction of dmi , so that here we do not probe the expected faster decay of the effective dmi with respect to the inverse ferromagnetic thickness in thicker devices .
+ moving the analysis from a qualitative to a quantitative level , we compare now the dmi strengths extracted here with the values obtained by kim and co - authors @xcite , who carried out bls measurements on the very same material stack used for the patterning of the devices reported in this manuscript . in fig .
[ fig_dmi_total ] the dmi values by kim et al .
@xcite ( squares ) and our values ( dots ) are reported in the same graph . from the graph it is clearly visible that the values of @xmath106 ( solid dots ) obtained by cidwm are in quantitative disagreement with the ones obtained by bls .
the two differently extracted values are off by about a factor 3
. indeed , if we multiply the extracted dmi values by a factor @xmath107 ( arbitrary choice , any value between 3 and 4 could have been chosen ) , the values of @xmath108 ( open dots ) are very close to the bls values .
( solid dots ) , is compared with the values obtained by bls by kim et al .
@xcite ( open squares ) .
the latter result to be about 3 times larger then the former .
the open dots are the values of an arbitrarily defined @xmath108.,width=302 ] this could indicate that the procedure of comparing the values of dmi extracted by different techniques needs to be adjusted .
if , on the one hand , the @xmath106 obtained here are in good agreement with the outcome of other cidwm - based measurements @xcite , on the other hand @xmath109 is in good agreement with the values obtained by bls measurements on the same sample @xcite .
these results clearly motivate a theoretical analysis about the comparability of dmi measurements obtained with different experimental techniques .
similar differences were also found by soucaille et al .
@xcite , where the dmi values extracted by dw creep motion and bls studies are compared , for several materials stacks with low gilbert damping ( @xmath110 ) .
+ one possible interpretation proposed by soucaille et al .
for the different values of dmi extracted by the two experimental techniques is the gilbert damping experienced by the dw during the creep motion .
they observed a larger discrepancy between the two techniques results for materials stacks with a smaller damping factor , suggesting a better agreement for systems with large damping .
however , despite the very large damping factor usually reported for pt@xmath0co@xmath0alo@xmath1 ( @xmath111 @xcite ) we do observe a quantitative discrepancy between the dmi values extracted by cidwm and the ones obtained by bls .
this rules out the small damping as the key factor generating the discrepancy between the cidwm- and the bls - extracted dmi values .
+ a second possible interpretation of such a discrepancy relies on the different type of physical process probed by the two techniques . in the bls measurements
spin waves with wavelengths of hundreds of nanometers are probed on an area of tens of microns ( defined by the laser spot ) @xcite . on the other hand , in cidwm experiments
the object at the center of the study is the dw , whose width is usually @xmath112 nm in pma systems .
accordingly , the dw motion is much more sensitive to the local variations in magnetic properties such as anisotropy , magnetization and dmi due to grain boundaries , interface roughness etc . , while the spin waves probed in the bls measurements are affected by the average values of those physical properties .
indeed , it has been suggested @xcite that pinning sites could be characterized by a weaker dmi compared to the defect - free regions of the sample , which might help to explain the lower dmi probed by the dw compared to thermal magnons . +
a final possible interpretation for the lower dmi extracted by the cidwm study relies on the presence of a non - negligible stt , even though for a related system no significant stt was claimed @xcite .
it is known that the standard stt drives the dw motion along with the electron flow , unless exotic materials properties like a negative non - adiabaticity parameter characterize the system under study .
however , to the best of our knowledge , no experimental observations of a negative non - adiabaticity parameter , in pt@xmath0co@xmath0alo@xmath1 and related systems , has ever been reported so far .
accordingly , the presence of a finite stt would result in the observation of a stopping field that is smaller than the actual dmi field ( as previously reported by torrejon et al .
nevertheless , if this was the main origin of the observed discrepancy , the strength of the stt needed to generate such effect would be @xmath113 times the actual sot s strength .
this would correspond to a @xmath114 underestimation of the actual strength of the sot by the used 1d model calculations . in the scenario of a simple model based on the she - induced sot , where @xmath115 , this would result in the extraction of a max sot efficiency of @xmath116 .
such a large sot efficiency has never been reported before in the study of cidwm , and thus , most likely , this can not explain our results .
so the reported discrepancy between cidwm- and bls - extracted dmi values might be understood if a combination of all the three interpretations reported above is taken into account .
however , it also shows that care needs to be taken when comparing values of the dmi determined by different techniques .
furthermore , this should encourage future comparative studies of different systems using different approaches and in particular theoretically analyze the different probed properties to identify the origin of the discrepancies , which will be reserved for a future study . when comparing the effective sot - fields ( @xmath81 ) extracted by the ccm on the one side and the @xmath11 technique on the other side , two major observations can be made .
first , the effective fields obtained by the two techniques are of comparable magnitude .
second , while the effective fields extracted by the dw motion study are all contained in a range of values as large as the 25% of the largest extracted value ( see fig .
[ fig_1dm_sots ] ) , for the effective fields obtained by hall measurements the largest value is 4 times as large as the smallest one ( see black squares in fig .
[ fig_sot - vs - tco ] ) .
furthermore , while the effective field increase with increasing thickness is seen by both techniques , the second harmonics technique detects also a stronger decrease for large thicknesses that is not directly visible in the dw motion results .
+ there may be several possible explanations for the different thickness dependence of the dl - field extracted by the two experiments , and here we discuss two of them .
the first explanation is based on the symmetry of the sots .
it is well known that sots depend on the polar angle ( angle between the magnetic moment and the z - axis ) @xcite , and the thickness dependence of the same sot ( dl - torque in this case ) does not need to be necessarily the same for different polar angles . furthermore
, the magnetic moment in a dw is characterized by a polar angle of 90 deg ( contained in the x - y plane of the sample ) , while the magnetization in the hall measurements is almost collinear to the z - axis ( @xmath20 or @xmath22 state ) .
accordingly , the observed different results obtained by cidwm and @xmath11 measurements can be interpreted as a polar angle - dependent scaling of the dl - field with the ferromagnetic thickness .
a second possible interpretation is based on the different spin structures that are probed in the two experiments . in the dw motion study , the effective sot - field acts on a highly non - collinear spin texture , while in the hall measurements the probed magnetic state is a single domain .
accordingly , the spin - transfer process in the two cases can be different , resulting in the extraction of different effective field amplitudes .
furthermore , the dmi present in this system plays a key role in the dw dynamics @xcite , while it does not affect the current - induced dynamics of the saturated magnetic state .
finally , the 1d - model used to analyze the torques exerted in the domain wall motion is limited by the virtue that it does not capture the evolution of the structure of the domain walls during their motion and this could lead to a limited understanding of the precise magnitude of the torques . in a future work , using independent determination of the damping and more refined modeling might lead to a more quantitative agreement .
+ in conclusion , we observe that the analysis of the results of @xmath11 and dw motion experiments provide us with effective sot - fields that are of similar magnitude . however a detailed correlation between the trend of the effective fields measured by the two methods calls for further investigation , which is beyond the scope of this work .
a detailed characterization of the thickness dependence of dzyaloshinskii - moriya interaction ( dmi ) and spin - orbit torques ( sots ) in pt@xmath0co(t)@xmath0alo@xmath1 was obtained by mean of two different experimental techniques : current - induced domain wall motion and second - harmonic hall measurements . the sign and strength of the dmi
are extracted by measuring the domain wall motion stopping fields .
a negative dmi , corresponding to the presence of left - handed homo - chiral dws , is observed to decrease in strength with an increasing co thickness .
this confirms that the dmi originates from the pt@xmath0co interface and its measured effective strength decreases for thicker ferromagnetic layers .
the extracted dmi strengths are in agreement with values reported in previous current - induced dw motion studies on similar materials stacks , however they are quantitatively different from the values extracted for the very same sample by brillouin light scattering ( bls ) measurements .
the dmi values presented here are about a factor 3 smaller than the values extracted by bls .
this quantitative disagreement highlights that care has to be taken when comparing the results of the two techniques and the differences possibly originate from the different length scales and physical processes probed in the two experiments . + from the analysis of the dw motion using a collective - coordinates model we extracted the driving sot - field , while the symmetry and magnitude of both damping - like and field - like sot - fields acting on a magnetic single domain state were extracted by @xmath11 hall measurements .
the effective damping - like sot - field driving the dw motion is observed to increase with an increasing ferromagnetic layer thickness up to @xmath117 nm . from the @xmath11 measurements both effective fields acting on the homogeneous spin structure are found to initially increase up to a co thickness of about 1.3 nm , and then to decrease with a further increase of the ferromagnetic thickness .
the differences in the thickness dependence of the effective fields obtained by the two techniques can have different origins : the non - ability of the 1d - model to capture the real dynamics of the dw internal magnetization during its motion , or the different parts of the polar angle dependence of the torques probed in the two experiments .
+ the similar non - monotonic thickness dependence observed for the damping - like and field - like effective fields suggests a possible common origin for the two fields : the spin - hall effect in the pt bottom layer . in this scenario ,
the final effective fields are defined by the transverse spin diffusion length in co , which is expected to be around 1.2 nm .
accordingly , for co thicknesses smaller than the transverse spin diffusion length the effective fields are observed to increase , while for co thicknesses larger than the spin diffusion length the effective fields are observed to decrease .
however , while qualitatively the behavior fits a dominating she origin we can not rule out additional effects that affect the two torques similarly leading to a similar thickness dependence .
finally , the qualitatively different thickness dependence of dmi and sots shows that both effects clearly do not have a common origin in our investigated system .
we acknowledge support by the graduate school of excellence materials science in mainz ( mainz ) gsc 266 , staudinger weg 9 , 55128 , germany ; the dfg ( kl1811 , sfb trr 173 spin+x ) ; the eu ( ifox , nmp3-la-2012 246102 ; maspic , erc-2007-stg 208162 ; multirev erc-2014-poc 665672 ; wall , fp7-people-2013-itn 608031 ) and the research center of innovative and emerging materials at johannes gutenberg university ( cinema ) .
e. m. acknowledges the support by project mat2011 - 28532-c03 - 01 from spanish government and project sa163a12 from junta de castilla y leon . c .- y .
you acknowledges the support by nrf - dfg collaborative research program ( no .
2014k2a5a6064900 ) .
kim acknowledges the support by the leading foreign research institute recruitment program ( no .
2012k1a4a3053565 ) through the national research foundation of korea .
in the framework of the collective - coordinates model ( ccm ) the dw dynamics is described by three degrees of freedom : the position of the dw in the track , @xmath83 , the in - plane angle of the dw magnetic moment with respect to the @xmath3-axis , @xmath84 , and the angle defined by the normal to the dw surface with respect to the @xmath3-axis , @xmath85 , describing the tilt of the dw surface .
the system of equations governing the dw dynamics reads : @xmath118 \\ & + \alpha\gamma_{0}[h_{pin}(q)+q\frac{\pi}{2}h_{dl}cos\phi ] , \end{split}\ ] ] @xmath119 \\ & + q\gamma_{0}[h_{pin}(q)+q\frac{\pi}{2}h_{dl}cos\phi ] , \end{split}\ ] ] @xmath120\frac{dq}{dt } & = -[\frac{2k_{eff}}{\mu_{0}m_{s}}+q\frac{\pi}{2}h_{d}cos(\phi-\chi)+h_{k}cos^{2}(\phi-\chi ) \\ & -\frac{\pi}{2}h_{x}cos\phi-\frac{\pi}{2}h_{fl}sin\phi]tan\chi - q\frac{\pi}{2}h_{dmi}sin(\phi-\chi ) \\
& -\frac{\pi}{2}h_{k}sin2(\phi-\chi ) , \end{split}\ ] ] where @xmath121 , with @xmath122 ( @xmath123 j / m @xcite ) being the dw parameter at rest , @xmath124 , @xmath125 and @xmath126 .
@xmath127 mj / m@xmath6 is the pma anisotropy constant , @xmath128 the ferromagnetic layer thickness and @xmath129 a / m the saturation magnetization @xcite .
@xmath130 is the dw shape anisotropy field with @xmath131 being the demagnetization field factor @xcite .
a large gilbert damping parameter @xmath132 is used as taken from literature @xcite .
@xmath133/@xmath134 for the @xmath25/@xmath26 configuration of dw .
the dmi is modeled as an effective field along the @xmath3-axis with its amplitude given by @xmath135 , where @xmath74 is the dmi parameter @xcite .
@xmath24 is the applied longitudinal field , @xmath136 is the pinning field
. the spatially - dependent pinning field accounts for local imperfections ( such as edge and surface roughness or defects ) , and can be derived from an effective spacial - dependent pinning potential , @xmath137 , as @xmath138 ( @xmath139 is the width of the magnetic wire ) .
a periodic potential is employed to describe the experimental results , @xmath140 , where @xmath141x@xmath142 j is the energy barrier of the pinning potential and @xmath143 is its periodicity .
[ 1dm_1 ] , [ 1dm_2 ] , and [ 1dm_3 ] are numerically solved by means of a 4th runge - kutta algorithm with a time step of 0.1 ps over a temporal window of 100 ns .
the analysis based on the 1 dm calculations allows us to gauge the role of the fl - torque on the dw motion . in fig .
[ fig_1dm_fl - sot ] the experimental results ( open dots ) for the device with @xmath48 nm ( a ) and ( b ) , and @xmath79 nm ( c ) and ( d ) are reported together with several fitting curves obtained considering different @xmath82 amplitudes ( different @xmath87 values ) but always the same @xmath81 used to generate the fitting curves in fig .
[ fig_vdw - hx ] . in fig .
[ fig_1dm_posfl_d1 ] and [ fig_1dm_posfl_d4 ] the case @xmath88 ( solid squares ) is compared with the case @xmath144 ( open diamonds ) . while , in fig . [ fig_1dm_negfl_d1 ] and [ fig_1dm_negfl_d4 ] the comparison is between @xmath88 ( solid squares ) and @xmath145 ( open diamonds ) .
+ on the one hand , as visible in every graph reported in fig .
[ fig_1dm_fl - sot ] , the different fitting curves almost overlap each other .
no net difference is visible between the curves for the high dw velocity regime , demonstrating the negligible role played by the fl - torque in the definition of the final dw velocity .
both fitting curves manage to reproduce the experimental data points with the same accuracy .
this supports the initial choice of not including any fl - torque in the extraction procedure of the effective sha . on the other hand ,
some differences between the fitting curves can be observed at the pinning regions .
the curves obtained by including a finite @xmath82 with different sign predict a different size of the pinning range ( compare between fig .
[ fig_1dm_posfl_d1 ] and [ fig_1dm_negfl_d1 ] , and between fig .
[ fig_1dm_posfl_d4 ] and [ fig_1dm_negfl_d4 ] ) .
this seems to suggest that the fl - torque can play a role in the depinning process of the dw , facilitating the dw depinning or making it harder , according to its sign with respect to the dl - torque .
i. m. miron , k. garello , g. gaudin , p .- j .
zermatten , m. v. costache , s. auffret , s. bandiera , b. rodmacq , a. schuhl , and p. gambardella , perpendicular switching of a single ferromagnetic layer induced by in - plane current injection , nature * 476 * , 189 ( 2011 ) .
i. m. miron , t. a. moore , h. szambolics , l. d. buda - prejbeanu , s. auffret , b. rodmacq , s. pizzini , j. vogel , m. bonfim , a. schuhl , and g. gaudin , fast current - induced domain - wall motion controlled by the rashba effect , nat . mater . * 10 * , 419 ( 2011 ) .
k. garello , i. m. miron , c. o. avci , f. freimuth , y. mokrousov , s. blgel , s. auffret , o. boulle , g. gaudin , and p. gambardella , symmetry and magnitude of spin orbit torques in ferromagnetic heterostructures , nat . nanotechnol .
* 8 * , 587 ( 2013 ) .
j. torrejon , j. kim , j. sinha , s. mitani , m. hayashi , m. yamanouchi , and h. ohno , interface control of the magnetic chirality in cofeb / mgo heterostructures with heavy - metal underlayers , nat .
commun . * 5 * , 4655 ( 2014 ) .
r. lo conte , e. martinez , a. hrabec , a. lamperti , t. schulz , l. nasi , l. lazzarini , r. mantovan , f. maccherozzi , s. s. dhesi , b. ocker , c. h. marrows , t. a. moore , and m. klui , role of b diffusion in the interfacial dzyaloshinskii - moriya interaction in ta / co@xmath101fe@xmath146b@xmath101/mgo nanowires , phys .
b * 91 * , 014433 ( 2015 ) .
m. j. benitez , a. hrabec , a. p. mihai , t. a. moore , g. burnell , d. mcgrouther , c. h. marrows , s. mcvitie , magnetic microscopy and topological stability of homo - chiral nel domain walls in pt / co / alo@xmath1 trilayer , nat .
* 6 * , 8957 ( 2015 ) .
v. kashid , t. schena , b. zimmermann , y. mokrousov , s. blgel , v. shah , and h. g. salunke , dzyaloshinskii - moriya interaction and chiral magnetism in @xmath147 zigzag chains : tight - binding model and @xmath148 @xmath149 calculations , phys .
b * 90 * , 054412 ( 2014 ) .
h. t. nembach , j. m. shaw , m. weiler , e. ju , and t. j. silva , linear relation between heisenberg exchange and interfacial dzyaloshinskii - moriya interaction in metal films , nat . phys . * 11 * , 825 ( 2015 ) .
j. kim , j. sinha , m. hayashi , m. yamanouchi , s. fukami , t. suzuki , s. mitani , h. ohno , layer thickness dependence of the current - induced effective field vector in ta@xmath150cofeb@xmath150mgo , nat .
* 12 * , 240 ( 2013 ) .
m. belmeguenai , j .- p .
adam , y. roussign , s. eimer , t. devolder , j .- v .
kim , s. m. cherif , a. stashkevich , and a. thiaville , interfacial dzyaloshinskii - moriya interaction in perpendicularly magnetized pt / co / alo@xmath1 ultrathin films measured by brillouin light spectroscopy , phys .
b * 91 * , 180405 ( 2015 ) .
j. cho , n .- h .
kim , s. lee , j .- s .
kim , r. lavrijsen , a. solignac , y. yin , d .- s .
han , n. j. j. van hoof , h. j. m. swagten , b. koopmans , and c .- y .
you , thickness dependence of the interfacial dzyaloshinskii - moriya interaction in inversion symmetry broken systems , nat .
* 6 * , 7635 ( 2015 ) .
kim , d .- s .
han , j. jung , j. cho , j .- s .
kim , h. j. m. swagten , and c .- y .
you , improvement of the interfacial dzyaloshinskii - moriya interaction by introducing a ta buffer layer , appl .
* 107 * , 142408 ( 2015 ) .
kim , j .- s .
kim , y. yin , j .- w .
koo , j. cho , s. lee , m. klui , h. j. m. swagten , b. koopmans , and c .- y .
you , asymmetric hysteresis for probing dzyaloshinskii - moriya interaction , nano lett . * 16 * ( 7 ) , 4438 ( 2016 ) .
l. liu , o .- j .
lee , t. j. gudmundsen , d. c. ralph , and r. a. buhrman , current - induced switching of perpendicularly magnetized magnetic layers using spin torque from the spin hall effect , phys .
lett . * 109 * , 096602 ( 2012 ) .
n. nakajima , t. koide , t. shidara , h. miyauchi , h. fukutani , a. fujimori , k. iio , t. katayama , m. nvlt , and y. suzuki , perpendicular magnetic anisotropy caused by interfacial hybridization via enhanced orbital moment in co / pt multilayers : magnetic circular x - ray dichroism study , phys .
lett . * 81 * , 5229 ( 1998 ) .
u. h. pi , k. w. kim , j. y. bae , s. c. lee , w. j. cho , k. s. kim , and s. seo , tilting of the spin orientation induced by rashba effect in ferromagnetic metal layer , appl .
phys . lett . *
97 * , 162507 ( 2010 ) .
lee , k. lee , j. cho , y .- h .
choi , c .- y .
you , m .- h .
jung , f. bonell , y. shiota , s. miwa , and y. suzuki , spin - orbit torque in a bulk perpendicular magnetic anisotropy pd / fepd / mgo system , sci . rep .
* 4 * , 6548 ( 2014 ) . c. o. avci , k. garello , i. m. miron , g. gaudin , s. auffret , o. boulle , and p. gambardella , magnetization switching of an mgo / co / pt layer by in - plane current injection , appl .
. lett . * 100 * , 212404 ( 2012 ) .
r. lo conte , a. hrabec , a. p. mihai , t. schulz , s .- j .
noh , c. h. marrows , t. a. moore , and m. klui , spin - orbit torque - driven magnetization switching and thermal effects studied in ta@xmath0cofeb@xmath0mgo nanowires , appl .
. lett . * 105 * , 122404 ( 2014 ) .
t. a. moore , i. m. miron , g. gaudin , g. serret , s. auffret , b. rodmacq , a. schuhl , s. pizzini , and m. bonfim , high domain wall velocities induced by current in ultrathin pt / co / alox wires with perpendicular magnetic anisotropy , appl .
* 93 * , 262504 ( 2008 ) .
e. martinez , s. emori , n. perez , l. torres , and g. s. d. beach , current - driven dynamics of dzyaloshinskii domain walls in the presence of in plane fields : full micromagnetic and one - dimensional analysis , j. appl . phys . * 115 * , 213909 ( 2014 ) .
s. emori , e. martinez , k .- j .
lee , h .- w .
lee , u. bauer , s .-
ahn , p. agrawal , d. c. bono , and g. s. d. beach , spin hall torque magnetometry of dzyaloshinskii domain walls , phys .
b * 90 * , 184427 ( 2014 ) .
a. hrabec , n. a. porter , a. wells , m. j. benitez , g. burnell , s. mcvitie , d. mcgrouther , t. a. moore , and c. h. marrows , measuring and tailoring the dzyaloshinskii - moriya interaction in perpendicularly magnetized thin films , phys .
b * 90 * , 020402 ( 2014 ) . c. k. safeer , e. jue , a. lopez , l. buda - prejbenau , s. auffret , s. pizzini , o. boulle , i. m. miron , and g. gaudin , spin - orbit torque magnetization switching controlled by geometry , nat
. nanotechnol .
* 11 * , 143 ( 2016 ) .
r. soucaille , m. belmeguenai , j. torrejon , j .- v .
kim , t. devolder , y. roussigne , s. m. cherif , a. a. stashchevich , m. hayashi , j .-
adam , probing the dzyaloshinskii - moriya interaction in cofeb ultrathin films using domain wall creep and brillouin light spectroscopy , arxiv:1604.05475 ( 2016 ) .
a. j. schellekens , l. deen , d. wang , j. t. kohlhepp , h. j. m. swagten , and b. koopmans , determining the gilbert damping in perpendicularly magnetized pt / co / alo@xmath1 films , appl .
. lett . * 102 * , 082405 ( 2013 ) .
x. qiu , p. deorani , k. narayanapillai , k .- s .
lee , k .- j .
lee , h .- w .
lee , h. yang , angular and temperature dependence of current induced spin - orbit effective fields in ta / cofeb / mgo nanowires , sci .
rep . * 4 * , 4491 ( 2014 ) . | we report the thickness dependence of dzyaloshinskii - moriya interaction ( dmi ) and spin - orbit torques ( sots ) in pt@xmath0co(t)@xmath0alo@xmath1 , studied by current - induced domain wall ( dw ) motion and second - harmonic experiments . from the dw motion study ,
a monotonous decay of the effective dmi strength with an increasing co thickness is observed , in agreement with a dmi originating at the pt@xmath0co interface .
the study of the ferromagnetic thickness dependence of spin - orbit torques reveals a more complex behavior . the effective sot - field driving the dw motion
is found to initially increase and then saturate with an increasing ferromagnetic thickness , while the effective sot - fields acting on a saturated magnetic state exhibit a non - monotonic behavior with increasing co - thickness .
the observed thickness dependence suggests the spin - hall effect in pt as the main origin of the sots , with the measured sot amplitudes resulting from the interplay between the varying thickness and the transverse spin diffusion length of the co layer . |
although the nature is still not fully understood at present , active galactic nuclei ( agns ) are widely believed to co - evolve with their host galaxies ( see kormendy & ho 2013 for a recent review ) .
the concept of co - evolution mainly stems from two observational facts : 1 ) the tight correlations between the mass of the central supermassive blackhole ( smbh ) and several properties of the bulge of the host galaxy , including the velocity dispersion , luminosity and mass of the bulge ( e.g. , magorrian et al . 1998 ; gebhardt et al .
2000 ; merritt & ferrarese 2001 ; mclure & dunlop 2002 ; tremaine et al .
2002 ; haring & rix 2004 ; ferrarese & ford 2005 ; aller & richstone 2007 ; gultekin et al . 2009 ; woo et al .
2010 ) ; 2 ) both agn s accretion and star formation have a peak at similar redshifts of @xmath5 .
( e.g. , ueda et al . 2003 ; croom et al . 2004 ; hasinger et al . 2005 ; nandra et al . 2005 ; silverman et al .
2008 ; shankar et al .
2009 ; aird et al . 2010 ; assef et al .
2011 ) .
what is associated with the gas fall onto a smbh is the agn s feedback that results in an interaction between the energy released in the accretion and the gas in the host galaxy ( see a review in fabian 2012 ) . in the local universe , a kpc - scale outflow has been frequently identified by spatially resolved spectroscopy for both ionized ( e.g. , holt et al .
2008 ; fu & stockton 2009 ; rupke & veilleux 2011 ; westmoquette et al .
2011 ) and molecular ( e.g. , feruglio et al .
2010 ; alatalo et al .
2011 ) gas .
recent observations point out that the interstellar medium ( ism ) throughout the host can be photoionized and kinematically disturbed by agn s feedback ( e.g. , fu & stockton 2009 ; greene et al .
in fact , involving a feedback is useful for solving the `` over cooling '' problem in the @xmath6 cold dark matter ( @xmath6cdm ) galaxy formation model in which the predicted massive galaxies are much more numerous than the observed ones ( e.g. , ciotti & ostriker 2007 ; somerville et al .
2008 ; hirschmann et al .
2013 ) , and for reproducing the observed @xmath7 relation , luminosity functions of quasars and normal galaxies ( e.g. , haehnelt et al .
1998 ; silk & rees 1998 ; fabian 1999 ; kauffmann & haehnelt 2000 ; granato et al .
2004 ; springel et al .
2005 ; di matteo et al .
2005 , 2007 ; croton et al .
2006 ; hopkins et al . 2007 , 2008 ;
khalatyan et al .
2008 ; menci et al .
2008 ; somerville et al . 2008 ; power et al .
2011 ; scannapieco et al .
dehnen & king ( 2013 ) recently proposed a new scenario of smbh growth in which the smbh accretion disk is formed because of the feedback .
the gas that is swept - up by the feedback finally falls towards the smbh on near - parabolic orbit when the feedback weakens .
it has been known for a long time that a large fraction of [ oiii ] doublet show blue asymmetry and bulk blueshift with respect to the systemic velocity ( e.g. , heckman et al .
1981 ; veron - cetty et al . 2001 ; zamanov et al . 2002 ; marziani et al . 2003
; aoki et al . 2005 ; bian et al . 2005
; boroson 2005 ; komossa et al .
2008 ; xu & komossa 2009 ; zhang et al .
the observed [ oiii ] line profile is generally explained by an interaction between nlr clouds and wind from central agns , which is supported by the spatially resolved spectroscopic observations of a few nearby seyfert 2 galaxies , although the origin of the wind is still an open issue .
the observations carried out with hubble space telescope indicate a nlr kinematics with a radial outflow acceleration of form @xmath8 at a radius @xmath9 and a deceleration of @xmath10 beyond the turnover radius @xmath11 ( e.g. , fischer et al . 2013 and references therein ) . by using the two dimensionless shape parameters ( i.e. , skewness and kurtosis ) to quantify the [ oiii]@xmath15007 line shape deviation from a pure gaussian function in narrow emission - line galaxies , wang et al .
( 2011 ) indicate a trend that agns with stronger blue asymmetries tend to be associated with younger stellar populations .
the authors additionally argued that the trend is likely driven by the co - evolution between agn s eddington ratio ( @xmath4 , where @xmath12 is the eddington luminosity ) and host galaxy , when @xmath13})/\sigma_*^4 $ ] is used as a proxy of @xmath4 .
the argument is mainly based upon the correlation between @xmath13})/\sigma_*^4 $ ] and line skewness .
however , it is generally accepted that a combination of a narrow core gaussian profile and an additional blueshifted , broad gaussian component is required to reproduce an observed asymmetric [ oiii]@xmath15007 line profile .
there is therefore a circular argument on which the measured @xmath13})$ ] contains the contributions from both components . in order to avoid the `` circular reasoning '' issue , this paper re - studies the evolution of [ oiii ] line profile by focusing on partially obscured agns that are excluded in our previous study .
here , the partially obscured agns refer to seyfert 1.8/1.9 galaxies and composite galaxies / h@xmath14 versus [ nii]/h@xmath2 diagnostic diagram between the empirical and theoretical demarcation lines that separate agns from star - forming galaxies .
see section 4.1 for the details . ] with a seyfert 1.8/1.9-like spectrum .
in addition to the [ oiii ] line profile and host stellar population , the spectra of these objects allow us to directly estimate @xmath4 from their broad h@xmath2 emission .
the partially obscured agns are excluded in wang et al .
( 2011 ) because of the contamination from the central agn s continuum , which generally results in an underestimation of the modeled stellar population age . in this paper , the underestimation is alleviated by properly removing the underlying agn s continuum from each observed spectrum . the paper is organized as follows .
the sample selection and spectral analysis are presented in section 2 and section 3 , respectively .
the results are shown in section 4 , and the implications are discussed in section 5 . a @xmath6cdm cosmology with parameters @xmath15 , @xmath16 , and @xmath17 ( spergel et al . 2003 )
is adopted throughout the paper .
a sub - sample of partially obscured agns has already been identified by wang et al .
( 2011 ) from the value - added sdss data release 7 ( abazajian et al . 2009 ) max - planck institute for astrophysics / johns hopkins university ( mpa / jhu ) catalog ( see heckman & kauffmann 2006 for a review ) .
the sample totally contains 229 broad - line seyfert galaxies / liners and broad - line composite galaxies after removing the duplicates .
their redshifts range from 0.0111 to 0.025 .
we refer the readers to section 2 and 3.4 in wang et al .
( 2011 ) for the details of the sample selection .
briefly , the sample requires : 1 ) a spectrum has a median signal - to - noise ratio ( s / n ) per pixel of the whole spectrum s / n@xmath1820 ( number of entries : 135,912 ) ; 2 ) all the emission lines used in the traditional baldwin - phillips - terlevich ( bpt ) diagnostic diagram ( e.g. , veilleux & osterbrock 1987 ) are detected with a significance level of at least 3@xmath19 ( number of entries : 39,384 ) .
3 ) the [ oiii]@xmath15007 emission line has s / n@xmath1830 ( number of entries : 10,913 ) ; 4 ) the [ oiii ] line width has @xmath20 , where @xmath21 is the instrumental resolution of the sdss spectroscopic survey ( number of entries : 3,478 ) ; 5 ) redshift does not lie within the range from 0.11 to 0.12@xmath15577 at the observer frame . ]
( number of entries after removing the duplicates : 2,616 ) ; 6 ) a broad h@xmath2 component can be identified in the spectrum by means of its blue high - velocity wing after the stellar features are removed from the spectrum .
the broad component is automatically selected by the criterion @xmath22 , where @xmath23 is the specific flux of the line wing averaged within the wavelength range from 6500 to 6350 in the rest frame and @xmath24 the standard deviation of the continuum flux within the emission - line - free region ranging from @xmath15980 to @xmath16020 .
the 1-dimensional spectra of the partially obscured agns are analyzed by the iraf package .
the spectral analysis includes galactic extinction correction , transformation to the rest frame , starlight component removal and emission line profile measurement .
the galactic extinction is corrected for each spectrum by the color excess @xmath25 taken from the schlegel , finkbeiner , and davies galactic reddening map ( schlegel et al .
1998 ) , by assuming an @xmath26 extinction law of the milkyway ( cardelli et al . 1989 ) .
the spectrum is then * shifted to the rest frame using the redshift provided by the sdss pipelines . the reduced spectrum in the rest - frame of a partially obscured agn sdssj095156.71 + 023602.0 is shown in figure 1 as an illustration . * the continuum of the partially obscured agns is dominated by the starlight component emitted from the host galaxies .
we model the stellar absorption features from each rest - frame spectrum as the sum of the first seven eigenspectra .
the eigenspectra are built through the principal component analysis ( pca ) method ( e.g. , francis et al .
1992 ; hao et al . 2005 ; wang & wei 2008 ) from the standard single stellar population spectral library developed by bruzual & charlot ( 2003 ) . a galactic extinction curve with @xmath27
is adopted in the modeling to account for the intrinsic extinction due to the host galaxy .
a @xmath28 minimization is performed for each of the spectra over the rest - frame wavelength range from 3700 to 8000 , except for the regions with strong emission lines .
the modeling and subtraction of the starlight component is illustrated in figure 1 for sdssj095156.71 + 023602.0 .
the modeled starlight component is used to measure host stellar population age .
the 4000 break index @xmath29 ( bruzual 1983 ; balogh et al .
1999 ) defined as @xmath30 is generally used as an excellent mean age indicator of the stellar population of the bulge of a galaxy until a few gyr after the onset of a star formation activity ( e.g. , kauffmann et al .
2003 ; heckman et al . 2004
; ; kewley et al . 2006 ; kauffmann & heckman 2009 ; wild et al . 2007 , 2010 ; wang & wei 2008 , 2010 ; wang et al . 2011 ,
2013 ) .
the contamination due to the underlying agn s continuum is an issue and must be considered in the stellar population synthesis of partially obscured agns .
the observed broad h@xmath2 emission demonstrates the existence of an underlying agn s continuum leaking from the edge of the torus according to the unified model ( e.g. , antonucci 1993 ) .
the leaking powerlaw continuum impacts the spectral shape at the blue end , which always results in an underestimate of @xmath3 index , and consequently a relatively younger stellar population if the contamination is ignored ( e.g. , cid fernandes & terlevich 1995 ; storchi - bergmann et al .
2000 ; cid fernandes et al .
2004 ; wang et al .
2013 ) . strictly speaking , the best way to remove
the underlying agn s continuum is to model the observed continuum by a linear combination of a power - law continuum and the used eigenspectra .
this is , however , a hard task according to our experiments because of the strong degeneracy between the agn s continuum and the blue spectra of hot stars .
the degeneracy typically results in a significant overestimate of the amplitude of the powerlaw component ( e.g. , cid fernandes et al .
2004 ; wang et al .
2013 ) .
an alternative approach is adopted here to remove the underlying agn s continuum .
a scaled powerlaw ( i.e. , @xmath31 ) is subtracted from each modeled starlight spectrum that is dereddened by the modeled local extinction in advance .
the amplitude of the powerlaw is estimated from the tight @xmath32-@xmath33 relation ( eq .
the relation finally yields a specific flux of the powerlaw at 5100 : @xmath34 , where @xmath35 is the measured line flux of broad h@xmath2 emission , and @xmath14 accounts for the effect on the flux due to reddening .
we at first checked the effect of different powerlaw index @xmath2 on the measured @xmath3 value .
the variation of powerlaw shape accounts for not only the intrinsic scatter of @xmath2 around its typical value of 1.7 ( e.g. , vanden berk et al . 2001 and references therein ) , but also the uncertainty due to reddening of agn s continuum .
when @xmath2 varies from 1.5 to 1.9 , a tiny change @xmath36 can be obtained for @xmath3 , which is significantly smaller than the dynamical range ( @xmath37 ) of the measured @xmath3 .
this result is not unreasonable because the specific flux ratio between the two powerlaws with @xmath38 and 1.9 has a shape @xmath39 , which is very close to a flat spectrum within the narrow wavelength ranges defining @xmath3 .
the left panel in figure 2 compares the @xmath29 index measured with the powerlaw correction ( @xmath40=[1.7,1.0]$ ] ) with that without the correction .
one can see clearly from the comparison : 1 ) as expected , an enhancement of @xmath29 value after the correction of powerlaw ; 2 ) the enhancement moderately increases with the measured @xmath29 .
the increase could be understood as follows . in a spectrum with larger 4000 break ,
the relative flux level contributed by a powerlaw has a more significant difference between the wavelength ranges blueward and redward of 4000 .
the right panel shows the same comparison but in the case of @xmath40=[1.7,2.0]$ ] is used by taking into account of the factor of @xmath41 obtained in wang & wei ( 2008 ) .
the factor @xmath42 represents an upper limit on the reducing of broad h@xmath2 lines emission due to reddening . ] .
as expected , a stronger agn s continuum results in a larger @xmath3 enhancement and a more significant increasing of the enhancement with @xmath3 value .
moreover , a larger scatter could be seen from the comparison for a stronger agn s continuum . without a further statement ,
the @xmath3 value measured after the powerlaw correction with @xmath40=[1.7,1.0]$ ] is used throughout the subsequent analysis . the effect on the results by different powerlaw level
is discussed in section 5.2 . after the removal of the starlight component , the emission - line profiles are at first modeled by the specfit task ( kriss 1994 ) in the iraf package in the emission - line isolated spectra for both h@xmath2 and h@xmath14 regions component is required to be separated from the observed line profile for the estimation of both smbh accretion properties and level of the underlying agn s continuum . ] .
each line is modeled by a linear combination of a set of several gaussian profiles .
the intensity ratios of the [ oiii ] and [ nii ] doublets are fixed to their theoretical values .
the flux of the [ oi]@xmath16300 emission line is measured through direct integration by the splot task in the iraf package . by following the method in wang et al .
( 2011 ) , the two high - order line shape parameters @xmath43 ( skewness ) and @xmath44 ( kurtosis ) are used to quantify the line shape deviation from a pure gaussian profile ( binney & merrifield 1998 ) .
the parameters are defined as @xmath45 , where @xmath46 is the k - order moment of line and @xmath19 the second - order moment in units of @xmath47 ( i.e. , @xmath48 in eq .
@xmath49 is the specific flux density of the continuum - subtracted emission line , and @xmath50 the line centroid ( i.e. , the first moment ) of the emission line .
a value of @xmath51 corresponds to a red asymmetry , and @xmath52 a blue asymmetry .
the emission line with a value of @xmath53 has a peak profile superposed on a broad base , while the line with @xmath54 has a `` boxy '' line profile and @xmath55 that are derived from the duplicate observations .
we emphasize that the sample selection in the current study is the same as that in wang et al .
( 2011 ) , except for the broad h@xmath2 emission . ] .
figure 3 presents the @xmath43 versus @xmath44 diagram for the partially obscured agns .
the figure shows a sequence starting from the pure gaussian region ( i.e. , @xmath56 , @xmath57 ) to the upper left corner for both seyfert and transition galaxies with broad balmer emission , which is highly consistent with that in wang et al .
( 2011 ) for narrow emission - line galaxies .
the [ oiii]@xmath58 line bulk relative velocity shift is calculated as @xmath59 , where @xmath60 and @xmath61 are the rest - frame wavelength of the [ oiii]@xmath15007 emission line and the wavelength shift with respect to the narrow h@xmath14 line , respectively .
the narrow h@xmath14 line shows a very small velocity shift relative to the galaxy rest frame ( e.g. , komossa et al .
@xmath62 is based on the measured line centroids ( @xmath63 ) and on the wavelengths in a vacuum of both h@xmath14 and [ oiii ] lines .
the calculated @xmath62 is compared with the @xmath43 ( the left panel ) and @xmath44 ( the right panel ) parameters in figure 4 .
the trends plotted in the two diagrams are very close to that for narrow emission - line galaxies shown in wang et al .
a spearman rank - order test yields a correlation between @xmath62 and @xmath43 with a correlation coefficient of @xmath64 and a probability of null correlation of @xmath65 corresponding to a z - value of 4.53 .
one can see from the left panel that there is a small fraction of the objects that deviate from the correlation at the large blueshift end .
the right panel indicates that these objects are dominated by the `` boxy '' , rather than the peaked , [ oiii ] line profiles , which is again in agreement with that in wang et al .
the authors have argued that a `` boxy '' line profile could be reproduced by the sum of two or more ( distinct ) peaks with comparable fluxes and line widths , which results in a shift of line centroid .
the two basic parameters ( i.e. , @xmath4 and @xmath66 , where @xmath67 ) of smbh accretion are estimated from the agn s broad h@xmath2 line emission that is obtained through our spectral profile modeling .
greene & ho ( 2007 ) provided an updated estimator of the mass of central smbh @xmath68^{2.06\pm0.06}m_\odot\ ] ] by combining the revised luminosity - radius relation ( with a scatter of 30 - 50% of the measured radius ) reported in bentz et al .
( 2006 ) and the @xmath32-@xmath33 relation in greene & ho ( 2005 ) @xmath69 where @xmath70 is the intrinsic luminosity of the h@xmath2 broad component corrected for local extinction .
the luminosity relation has a rms scatter around the best - fit line of 0.2dex .
the extinction is obtained from the narrow - line ratio h@xmath2/h@xmath14 for each object , assuming the balmer decrement for standard case b recombination and a galactic extinction curve with @xmath27 .
the agn s bolometric luminosity is then derived from the calibration @xmath71 ( kaspi et al . 2000 ) .
this section presents the statistical study on the partially obscured agns based on the spectral measurements described above .
the spectral properties of these objects enable us to determine the agn s accretion properties directly from their broad h@xmath2 emission instead of from other proxy .
this is essential for getting rid of the `` circular reasoning '' in our previous study of narrow emission - line galaxies ( wang et al .
2011 ) . in that study ,
both bolometric luminosity of agn and line profile parameters are derived from [ oiii]@xmath15007 emission line . in total ,
15 objects are excluded from the subsequent statistical analysis either because of their poor starlight removal or because of their poor s / n of the broad h@xmath2 emission .
figure 5 displays the three bpt diagrams for the final 170 broad - line seyfert galaxies and 44 broad - line transition galaxies to @xmath72 , * this range indicates that some of these objects are agns ( e.g. , those having a luminosity above @xmath73 ) . * _ ] .
all the line ratios are calculated from the narrow balmer line emission that is obtained through our profile modeling .
the [ oiii]@xmath15007 line flux used in the line ratio calculation is the sum of both narrow and broad ( mostly blueshifted ) components because these two components are required to properly model the observed [ oiii ] line profile .
the diagrams are a powerful tool in determining the dominant powering source in narrow emission - line galaxies through their emission - line ratios .
the solid lines in the three panels show the theoretical demarcation lines separating `` pure '' agns from star - forming galaxies ( kewley et al .
the long - dashed and dotted line in the left panel mark the empirical demarcation line proposed in kauffmann et al .
( 2003 ) and the theoretical one proposed by stasinska et al .
( 2006 ) , respectively . based upon their new photoionization models , stasinska et al .
( 2006 ) claimed that their demarcation line is slightly restrictive in separating `` pure '' starforming galaxies than the empirical one .
as shown by the diagrams , almost all the objects listed in the sample are located above these demarcation lines , and our sample is strongly biased against liners that are typical of low [ oiii]/h@xmath14 ratio .
the bias is caused by our sample selection that requires a high s / n ratio ( and consequently a large flux ) of [ oiii ] emission line .
the measured @xmath3 value is plotted against @xmath43 ( the left panel ) and @xmath62 ( the right panel ) in figure 6 . at first ,
the diagrams confirm the results obtained in the previous studies ( e.g.,kewley et al . 2006 ; schawinski et al . 2007 ; wang & wei 2008 ; wang et al . 2011 ) :
the transition galaxies are clustered in the region with younger stellar populations than do seyfert galaxies , which implies that the transition galaxies are at an intermediate evolutionary phase in the context of the co - evolution of agns and their host galaxies ( see discussion and references in wang et al .
2011 ) . secondly ,
although no evident correlation can be identified from the two diagrams , a clear difference can be obtained if the sample is separated into two parts according to their stellar population ages .
the separation is marked by a dashed line with @xmath74 in each panel .
this value is usually used to define young / old stellar population according to the @xmath3 measurement ( e.g. , kauffmann et al . 2003 ) .
the partially obscured agns associated with young stellar populations show a wide range in their [ oiii ] line profiles that vary from a blue asymmetrical shape to a gaussian function . on the contrary ,
a symmetric line profile is typical for the agns associated with old stellar populations .
a similar trend can be also found in the @xmath62 case .
agns with significant [ oiii ] blueshift ( @xmath75 ) can be only observed in the galaxies with young stellar populations with @xmath76 .
the left panel in figure 7 plots @xmath4 as a function of stellar population age .
a significant correlation can be identified between agn s accretion power and host galaxy stellar population age : larger the @xmath4 , younger the host stellar population will be ( @xmath77 , @xmath78 and @xmath79 for the spearman rank - order test ) .
the strong correlation confirms and reinforces the results that obtained by many authors in the previous studies ( see discussion for more details ) .
we parametrize the relationship as the best fit of @xmath80 where @xmath81 , the relationship is over plotted by a dashed line in the diagram .
@xmath4 is plotted against @xmath43 in the middle panel of figure 7 .
similar as the @xmath3 versus @xmath43 diagram , there is no evident relationship between the two parameters at first glance .
however , one can identify a trend that the objects are avoid to occupy the left - bottom corner with low @xmath4 and strong [ oiii ] blue asymmetry .
to reveal this trend more clearly , we separate the sample into two parts basing upon @xmath82 .
figure 8 compares the distributions of @xmath4 of the two sub - samples . when compared with the agns with weaker [ oiii ] blue asymmetry , the distribution of the ones with stronger asymmetry is slightly shifted towards the higher @xmath4 end .
a two - sided kolmogorov - smirnov test yields a marginal difference between the two distributions at a significance level of @xmath83 with a maximum absolute discrepancy of 0.23 in logarithm .
mullaney et a. ( 2013 ) compares the stacking [ oiii ] line profiles with various agn s parameters for 24,264 optically selected agns .
their panel b in figure 3 indicates a trend in which the strength of the [ oiii ] blue wing increases with eddington ratio , which is consistent with the result reported here .
the right panel in figure 7 presents an anti - correlation between @xmath4 and @xmath62 , which is failed to be identified for narrow emission - line galaxies in wang et al .
the correlation indicates that the bulk velocity blueshift of [ oiii ] emission line increases with @xmath4 .
a spearman rank - order test returns a correlation coefficient @xmath84 at a significance level of @xmath85 ( @xmath86 ) .
zhang et al . (
2011 ) recently analyzed a large and homogeneous sample of radio - quiet seyfert 1 galaxies and quasars selected from sdss .
their analysis indicates that the bulk velocity shift of [ oiii ] is found to be strongly related with @xmath4 : the large velocity shift generally occurs in the agns with high @xmath4 ( see also in e.g. , boroson 2005 ; bian et al .
2005 ) . as a rare population ,
the `` blue outliers '' are the objects with strong [ oiii ] blueshifts larger than @xmath87 ( e.g. , zamanov et al .
2002 ; zhou et al .
spectroscopic observations point out that these objects occupy the high @xmath88 end in the eigenvector - i ( ei ) space , and are exclusively population a objects and large rfe ) and b ( having @xmath89 and small rfe ) , where rfe is defined as the feii to broad h@xmath14 line ratio . generally speaking ,
* population a agns have higher @xmath4 than do population b agns ( see citations in the main text ) .
population a agns contain narrow - line seyfert 1 galaxies that are believed to * be at early evolution stage ( e.g. , mathur 2000 ) . * * * ] that have small broad h@xmath14 line widths ( @xmath90 ) and strong feii complex emission ( e.g. , zamanov et al . 2002 ; marziani et al .
2003 ; komossa et al .
2008 ; marziani & sulentic 2012 ) . compared with the seyfert galaxies
, one can learn from figure 7 that the composite galaxies in the current sample have extreme properties with younger host stellar population , higher eddington ratio , less significant [ oiii ] blue wing and smaller bulk [ oiii ] line shift .
as discussed in wang & wei ( 2008 ) , although @xmath70 is an acceptable measurement of the accretion power in partially obscured agns , a systematical underestimate of @xmath4 and @xmath66 can not be avoided both because of the obscuration of the torus and because of the reddening in the broad - line region .
our previous study obtained an upper limit of underestimate of @xmath91 for @xmath4 and @xmath92 for @xmath66 .
the underestimate of @xmath70 could be learned in the left panel of figure 9 , which compares the distribution of @xmath70 between the used partially obscured agns and typical type i agns in greene & ho ( 2007 ) .
the difference between the peaks of the two distributions is about 0.5dex .
this difference is likely caused by the reddening and obscuration of the broad h@xmath2 emission .
the middle panel compares the calculated @xmath66 between the two samples .
a strong bias against small @xmath66 could be identified in the comparison .
the bias is caused by our broad h@xmath2 selection .
the selection is based on the high velocity h@xmath2 line wing over the wavelength range from 6500 to 6530 ( i.e. , blueward of [ nii]@xmath93 line ) in the rest frame , which naturally results in a significant loss of the objects with small @xmath94 .
a similar comparison is shown in the right panel for the calculated @xmath4 .
compared with the type i agns , the @xmath4 distribution of the partially obscured agns is systematically shifted to low @xmath4 end by an amount of @xmath951.0dex , although both distributions have roughly identical dynamical range of @xmath952.0dex . this systematical shift could be explained by the combination of the underestimate of @xmath70 and the selection bias of broad h@xmath2 line width .
comparing with our previous studies , the current measurement of @xmath3 index is improved by subtracting an underlying agn s continuum from the observed spectrum according to the broad h@xmath2 emission . as stated in section 3.1 , there is an uncertainty for the moved powerlaw continuum due to the uncertainties of reddening and obscuration and the intrinsic scatter of the powerlaw index .
we have demonstrated that the measured @xmath3 is more sensitive to reddening factor @xmath14 than to powerlaw index . a monte - carlo simulation with 1,000 random experiments
is then performed to quantitatively examine the effect on the @xmath4-@xmath3 correlation caused by the uncertainty of removed powerlaw continuum level . for each object , a random @xmath3
is produced by a random sampling of @xmath14 in the range from 1 to 2 .
a spearman rank - order test is then performed for each of the 1,000 random samples .
figure 10 displays the distribution of the simulated correlation coefficient @xmath96 .
the distribution has a mean value of @xmath97 and a standard deviation of @xmath98 .
the simulation indicates that the relationship between @xmath4 and @xmath3 is not sensitive to the uncertainty of the amplitude of the used powerlaw .
] , although the relationship can be slightly degraded by the uncertainty of the level of the removed powerlaw continuum .
the relationship between @xmath4 and stellar population age of host galaxy has been established by different approaches in past decade .
this relationship allows us to establish a co - evolution scenario in which agns evolve from a high-@xmath4 state to a low-@xmath4 state as the circumnuclear stellar population continually ages . in a wide range of smbh masses
, the majority of local smbh growth is found to occur not only in the high accretion phase but also in the galaxies with young stellar populations and with ongoing or recent star formations ( goulding et al .
2010 ; heckman & kauffmann 2006 and references therein ) .
wang et al . ( 2006 ) reported a relation between the ei space and middle - far - infrared color @xmath99 by performing a pca analysis on a sample of _ iras - selected seyfert 1.5 galaxies .
this relation naturally indicates a relation between @xmath4 and host galaxy stellar population , both because the ei space is commonly accepted to be driven by @xmath4 ( e.g. , boroson 2002 ) and because the color @xmath99 addresses the relative importance of agn activity and starburst activity .
a direct correlation between @xmath4 and host stellar population age is established either by using @xmath13})/\sigma_*^4 $ ] as a proxy of @xmath4 in type ii agns ( e.g. , kewley et al . 2006 ; wild et al . 2007 ; kauffmann et al . 2007 ) or by estimating @xmath4 directly from the balmer broad lines in partially obscured agns ( wang et al . 2008 , 2010 , and this paper ) .
the role of @xmath4 in the coevolution issue is additionally revealed from agn s hard x - ray emission , basing upon an identified correlation between agn s hard x - ray spectral index and stellar population age of the host galaxy ( wang et al .
_ the evolutionary scenario mentioned above describes how agns ( strictly speaking , which parameters of agns ) co - evolve with their host galaxies .
this phenomenological scenario , however , does not involve the background physical process that produces a self - regulated growth of smbh mass and host galaxy star formation .
one can see from our statistical study that the feedback is a potential `` bridge '' that links and regulates the growth of smbh and host galaxy star formation . according to the aforementioned results ,
a strong agn - driven feedback is required for an agn in its early gas - rich phase with both rapid smbh growth ( i.e. , high @xmath4 ) and young host stellar population , although the physical process that drives the outflow is still an open question .
in contrast , a weak feedback is sufficient to regulate smbh mass growth and host star formation in the late gas - poor phase .
the available models include agn s wind ( e.g. , crenshaw et al .
2003 ; pounds et al . 2003
; ganguly et al . 2007 ; reeves et al . 2009 ; dunn et al . 2010 ; tombesi et al .
2012 ) , radiation pressure ( e.g. , alexander et al . 2010 ) and radio jets ( e.g. , rosario et al . 2010 ) .
in the current sample , the non - detection of the connection between radio emission and [ oiii ] line profile ( see section 5.4 ) allows us to believe that the radiation pressure is probably the initial driver of the outflows .
matsuoka ( 2012 ) recently identified a deficit of extended emission - line region in the agns with high @xmath4 .
the deficit could be explained by the agn s feedback that blows the gas around smbh away .
moreover , the spectroscopic observations using integral - field unit suggest the similar signatures of outflow in high - redshift galaxies / quasars ( e.g. , nesvadba et al . 2008 ; cano - diaz et al .
2012 ; harrison et al .
2012 ) , where the outflow from agn is predicted to be much stronger than in local universe according to the model of cosmic galaxy formation ( see a more extended discussion on this issue at the end of this subsection ) .
the revealed dependence between feedback and agn - host co - evolution provides an evidence that supports the hypothesis that the host star formation is regulated by smbh growth through the information provided by agn - driven feedback . both agn - suppressed and agn - induced host star formation have been proposed in past decades , although we can not say which one occurs in agns based on our current results .
a commonly accepted mechanism that the feedback works is that the feedback from central agn sweeps out circumnuclear gas , which suppresses star formation and results in a self - regulated smbh growth and host star formation in both galaxies merger and secular evolution scenarios ( e.g. , alexander & hickox 2012 ; fabian 2012 ; kormendy & ho 2012 ; ) . in the color versus absolute magnitude diagram
, there is an excess of agns in the `` green valley '' between the blue cloud and red sequence ( e.g. , sanchez et al .
2004 ; nandra et al .
2007 ; schawinski et al . 2009 ; treister et al . 2009 ; georgakakis et al . 2009 ; cisternas et al . 2011 ; mullaney et al . 2012 ; rosario et al .
the excess implies that agns are at a transition from the blue cloud to the red sequence , which is likely resulted from a suppression of star formation due to the feedback from agns emission line and infrared emission indicate that these agns are in reddened star - forming galaxies , rather than in a transition phase ( e.g. , silverman et al .
2009 ; mullaney et al .
2012 ; rosario et al .
xue et al . ( 2010 ) indicates that agn hosts and non - agn galaxies occupy the same region in the color - magnitude diagram by using mass - matched samples . ]
( e.g. , nandra et al .
2007 ; hickox et al .
2009 ; hopkins & elvis 2010 ) .
zubovas et al .
( 2013 and references therein ) recently performed a study on positive feedback through well - resolved numerical simulations of agn s feedback process . at early gas - rich phase , the energy - driven feedback compresses the cold gas in host galaxy into dense shells and clumps , which triggers star formation if the shocked gas cools rapidly .
a negative feedback that quenches the star formation , however , occurs at late gas - poor phase .
the model in king ( 2003 , 2005 ) indicates that the @xmath100 relation is established through the momentum - driven flow with a mildly relativistic velocity of @xmath101 , where @xmath102 is the accretion efficiency .
however , a energy - driven outflow at large scale is expected after the @xmath100 relationships established .
the outflow velocity is @xmath103 ( king et al .
2011 , zubovas & king 2012 ) , where @xmath104 is the velocity dispersion in bulge , @xmath105 eddington ratio of central smbh , @xmath106 the cosmology value of the baryon - to - dark - matter density fraction , and @xmath107 the gas fraction in bulge that is of the same order of @xmath108 .
in the current sample , we have the averaged values of @xmath109 and @xmath110 . with these values
, the outflow velocity at large scale is expected to be @xmath111 , which is close to the velocity shift measured from the [ oiii ] emission lines .
large - scale outflow is expected to be strong in high redshift ( @xmath112 ) agns where the peaks of both agn s activity and star formation occur roughly coincident ( e.g. , ishibashi et al . 2013 ) .
searching for the relationship between the strength of agn s feedback and effect on host properties in high@xmath113 universe is important for understanding the impact of the feedback on the galaxy formation and evolution . in spite of the debates ( e.g. , santini et al . 2012 ; harrison et al .
2012 ) , page et al .
( 2012 ) claims a suppressed star formation in powerful agns in the redshift range from @xmath114 to 3 basing upon the herschel spire observations in sub - millimeter .
the interaction between radio jet and ism is another possible mechanism that drives outflow in galactic - scale ( e.g. , holt et al .
2008 ; nesvadba et al .
2008 ; guillard et al .
an outflow of cold , neutral gas due to jet - ism interaction is identified by many observations ( e.g. , morganti et al .
2005 , 2007 ; holt et al .
2011 ; mahony et al .
there is an alignment between radio jet and structure in ( extended ) nlr in many objects ( e.g. , whittle & wilson 2004 ; rosario et al .
strong blue wings of [ oiii]@xmath15007 lines have been identified by brotherton ( 1996 ) in a sample of radio - loud quasars .
the [ oiii ] line width is found to be correlated with radio luminosity at 1.4ghz ( @xmath115 ) for flat - spectrum radio galaxies in early studies ( e.g. , heckman et al .
1984 ; whittle 1985 ) . by using a large sample of agns ( both type i and type ii ) from sdss , mullaney et al .
( 2013 ) claimed that the [ oiii ] line width is more strongly related to @xmath115 than the other agn s parameters .
the partially obscured agns listed in the current sample are cross - matched with the first survey catalog ( becker et al .
2003 ) to examine the connection between their radio emission and [ oiii ] line profiles .
the cross - match returns 69 seyfert galaxies and 26 composite galaxies with detected radio flux exceeding the first limiting flux density ( 5@xmath19 ) of 1mjy .
their radio fluxes range from 1.0 to 734mjy .
all these objects have a radio emission with s / n@xmath116 . an underlying contribution from host starburst
is a possible explanation of the high fraction of radio detection in current sample ( @xmath117% ) than in typical type - i agns ( @xmath118% , e.g. , rafter et al .
2009 ; mullaney et al . 2013 ) .
the most radio - luminous starburst galaxies ( @xmath119 ) have a radio luminosity of @xmath120 ( smith et al .
the high detection rate could be alternatively attributed to our sample selection criterion , which excludes the majority of less luminous agns , such as liners , with faint [ oiii ] line emission .
the luminosity at 1.4ghz is calculated from @xmath121 , where @xmath122 is the integrated flux density at 1.4ghz , @xmath123 is the luminosity distance , and @xmath124 ( e.g. , ker et al .
2012 ) is the spectral slope defined as @xmath125 .
@xmath115 is plotted against @xmath43 and @xmath62 in the left and right panel in figure 11 , respectively .
the calculated radio luminosity ranges from @xmath126 to @xmath127 .
one can see from the figure that there is no evident connection between the radio emission and the [ oiii ] line profile asymmetry and bulk velocity blueshift in the current sample .
the strongest radio emission ( i.e. , @xmath128 ) in fact occurs in a few of objects with nearly gaussian profile and small velocity shift with @xmath129 at most . the lack of numbers in our partially obscured agn sample ( 95 vs. 1,988 ) is a possible explanation for none detection of the influence by radio emission seen in mullaney et al .
( 2013 ) .
the spectral analysis in this study allows us to compare the [ oiii ] line profile between the partially obscured seyfert galaxies and the seyfert 2 galaxies that are studied in wang et al .
the distributions of line width , @xmath43 and @xmath62 are compared in figure 12 .
the three panels show that the partially obscured agns tend to have wider [ oiii ] line , stronger blue asymmetry and larger bulk velocity blueshift than do the seyfert 2 galaxies .
these results are in agreement with previous studies , and are easily understood in the context of the agn s unified model ( e.g. , antonucci 1993 ) .
vaona et al .
( 2012 ) performed a comprehensive comparison of narrow emission line between seyfert 2 galaxies and intermediate - type seyfert galaxies .
the authors identified a higher percentage of strong [ oiii ] line blue asymmetry and large line width in the intermediate - type seyfert galaxies than in the seyfert 2 galaxies .
a similar conclusions are recently obtained in a comparison between type i and type ii seyfert galaxies in mullancy et al .
the [ oiii ] line asymmetry and bulk velocity blueshift are widely used to characterize agn s outflows that are initially launched from accretion disc or dusty torus ( e.g. heckman et al .
1981 ; barth et al . 2008 ; nesvadba et al . 2008 ; greene et al . 2011 ) . in the context of the unified model , the observed properties of the outflow inferred from [ oiii ] emission depend on the orientation of the agn s ionization cone with respect to the line - of - sight of an observer .
a more asymmetrical [ oiii ] line profile and a larger bulk velocity are expected in partially obscured agns than in seyfert 2 galaxies if the outflow is along the axis perpendicular to the plane of torus . a majority difference between the current study and wang et al .
( 2011 ) is the method used to estimate @xmath4 .
@xmath4 is obtained through [ oiii ] line luminosity and stellar velocity dispersion in wang et al .
( 2011 ) , and through broad h@xmath2 emission in this paper . figure 13 compares the different methods in the current partially obscured agn sample .
the left panel shows a tight correlation between the luminosities of the two emission lines .
one can see clearly a large scatter , a systematical offset and different dynamical ranges when the eddington ratios estimated from h@xmath2 ( @xmath4 ) and [ oiii ] lines ( @xmath105 ) are compared with each other in the middle panel .
our examination indicates that the large scatter and the systematic are mainly caused by the poor relationship between the @xmath66 determined by h@xmath2 line and stellar velocity dispersion .
the role of feedback in the agn - host coevolution issue is re - studied here by focusing on the [ oiii ] emission line profile in partially obscured agns selected from sdss dr7 .
the broad h@xmath2 emission lines enable us to 1 ) directly determine the agn s accretion properties ; 2 ) subtract an underlying agn s continuum from the observed spectrum . with these improvements ,
our analysis indicates ( confirms ) that the agns associated with young stellar population tend to be of large @xmath4 , strong [ oiii ] blue asymmetry and large [ oiii ] bulk velocity blueshift , which implies that the feedback from central agn plays an important role in linking as self - regulated smbh growth and host star formation .
_ acknowledgments + we thank the anonymous referee for his / her helpful suggestions for improving the manuscript .
this study uses the sdss archive data that was created and distributed by the alfred p. sloan foundation .
the study is supported by the national basic research program of china ( 973-program , grant no .
2014cb845800 ) and by the national science foundation of china under grant 11473036 . _
abazajian , k. n. , adelman - mccarthy , j. k. , agueros , m. a. , et al .
2009 , apjs , 182 , 543 aird , j. , et al .
2010 , mnras , 401 , 2531 alatalo , k. , et al .
2011 , apj , 735 , 88 alexander , d. m. , & hickox , r. c. 2012 , newar , 56 , 93 aller , m. c. , & richstone , d. o. 2007 , apj , 665 , 120 antonucci , r. r. j. 1993 , ara&a , 31 , 473 aoki , k. , kawaguchi , t. , & ohta , k. 2005 , apj , 618 , 601 assef , r. j. , et al .
2011 , apj , 728 , 56 balogh , m. l. , et al .
1999 , apj , 527 , 54 baldwin , j. a. , phillips , m. m. , & terlevich , r. 1981 , pasp , 93 , 5 barth , a. j. , greene , j. e. , & ho , l. c. 2008 , aj , 136 , 1179 becker , r. h. , helfand , d. j. , white , r. l. , gregg , m. d. , & laurent - muehleisen , s. a. 2003 , vizier online data catalog , 8071 , 0 bentz , m. c. , peterson , b. m. , pogge , r. w. , vestergaard , m. , & onken , c. a. 2006 , apj , 644 , 133 bian , w. h. , yuan , q. r. , & zhao , y. h. 2005 , mnras , 364 , 187 binney , j. , & merrifield , m. 1998 , galactic astronomy ( princeton , nj : princeton univ . press ) boroson , t. a. 2002 , apj , 565 , 78 boroson , t. a. 2005 , aj , 130 , 381 brotherton , m. s. 1996 , apjs , 102 , 1 bruzual , a. g. 1983 , apj , 273 , 105 bruzual , g. , & charlot , s. 2003 , mnras , 344 , 100 cano - diaz , m. , maiolino , r. , marconi , a. , netzer , h. , shemmer , o. , & cresci , g. 2012 , a&a , 537 , l8 cid fernandes , r. jr . , & terlevich , r. 1995 , mnras , 272 , 423 cid fernandes , r. , gu , q. , melnick , j. , terlevich , e. , terlevich , r. , kunth , d. , rodrigues lacerda , r. , & joguet , b. 2004 , mnras , 355 , 273 ciotti , l. , & ostriker , j. p. 2007 , 665 ,
1038 cisternas , m. , et al .
2011 , apj , 726 , 57 crenshaw , d. m. , et al .
2003 , apj , 594 , 116 croom , s. m. , smith , r. j. , boyle , b. j. , shanks , t. , miller , l. , outram , p. j. , & loaring , n. s. 2004 , mnras , 349 , 1379 croton , d. j. , springel , v. , white , s. d. m. , et al .
2006 , mnras , 365 , 1 di matteo , p. , combes , f. , melchior , a .-
l . , & semelin , b. 2007 , a&a , 468 , 61 di matteo , t. , springel , v. , & hernquist , l. 2005 , nature , 433 , 604 dehnen , w. , & king , a. 2013 , apjl , 777 , 28 dunn , j. p. , et al .
2010 , apj , 709 , 611 fabian , a. c. 1999 , mnras , 308 , l39 fabian , a. c. 2012 , ara&a , 50 , 455 ferrarese , l. , & ford , h. 2005 , ssr , 116 , 523 feruglio , c. , maiolino , r. , piconcelli , e. , menci , n. , aussel , h. , lamastra , a. , & fiore , f. 2010 , a&a , 518 , l155 fischer , t. c. , crenshaw , d. m. , kraemer , s. b. , & schmitt , h. r. 2013 , apjs , 209 , 1 francis , p. j. , hewett , p. c. , foltz , c. b. , & chaffee , f. h. 1992 , apj , 398 , 476 fu , h. , & stockton , a. 2009 , apj , 690 .
953 ganguly , r. , brotherton , m. s. , cales , s. , scoggins , b. , shang , z. , & vestergaard , m. 2007 , apj , 665 , 990 gebhardt , k. et al . 2000 , apjl , 539 , 13 georgakakis , a. , clements , d. l. , bendo , g. , rowan - robinson , m. , nandra , k , & brotherton , m. s. 2009 , mnras , 394 , 533 goulding , a. d. , alexander , d. m. , lehmer , b. d. , & mullaney , j. r. 2010 , mnras , 406 , 597 granato , g. l. , de zotti , g. , silva , l. , bressan , a. , & danese , l. 2004 , apj , 600 , 580 greene , j. e. , & ho , l. c. 2005 , apj , 630 , 122 greene , j. e. , & ho , l. c. 2007 , apj , 670 , 92 greene , j. e. , zakamska , n. l. , ho , l. c. , & barth , a. j. 2011 , apj , 732 , 9 guillard , p. , et al .
2012 , apj , 747 , 95 gultekin , k. , et al .
2009 , apj , 698 , 198 haehnelt , m. g. , natarajan , p. , rees , m. j. 1998 , mnras , 300 , 817 hao , l. , et al .
2005 , aj , 129 , 1795 haring , n. , & rix , h. w. 2004 , apjl , 604 , 89 harrison , c. m. , et al .
2012 , 426 , 1073 hasinger , g. , miyaji , t. , & schmidt , m. 2005 , a&a , 441 , 417 heckman , t. m. , & kauffmann , g. 2006 , newar , 50 , 677 heckman , t. m. , miley , g. k. , & green , r. f. 1984 , apj , 281 , 525 heckman , t. m. , miley , g. k. , van breugel , w. j. m. , & butcher , h. r. 1981 , apj , 247 , 403 hickox , r. c. , et al .
2009 , apj , 696 .
891 hirschmann , m. , dolag , k. , saro , a. , borgani , s. , & burkert , a. 2013 , astro - ph / arxiv1308.0333 holt , j. , tadhunter , c. n. , & morganti , r. 2008 , mnras , 387 , 639 holt , j. , tadhunter , c. n. , morganti , r. , & emonts , b. h. c. 2011 , mnras , 410 , 1527 hopkins , p. f. , bundy , k. , hernquist , l. , & ellis , r. s. 2007 , apj , 659 , 976 hopkins , p. f. , cox , t. j. , keres , d. , & hernquist , l. 2008 , apjs , 175 , 390 hopkins , p. f. , & elvis , m. 2010 , mnras , 401 , 7 ishibashi , w. , fabian , a. c. , & canning , r. e. a. 2013 , mnras , 431 , 2350 kaspi , s. , smith , p. s. , netzer , h. , maoz , d. , jannuzi , b. t. , & giveon , u. 2000 , apj , 533 , 631 kauffmann , g. , & haehnelt , m. 2000 , mnras , 311 , 576 kauffmann , g. , & heckman , t. m. 2009 , mnras , 397 , 135 kauffmann , g. , heckman , t. m. , white , s. d. m. , et al .
2003 , mnras , 341 , 33 kauffmann , g. , et al .
2007 , apjs , 173 , 357 ker , l. m. , best , p. n. , rigby , e. e. , rttgering , h. j. a. , & gendre , m. a. 2012 , mnras , 420 , 2644 kewley , l. j. , dopita , m. a. , sutherland , r. s. , heisler , c. a. , & trevena , j. 2001 , apj , 556 , 121 kewley , l. j. , groves , b. , kauffmann , g. , & heckman , t. 2006 , mnras , 372 , 961 khalatyan , a. , cattaneo , a. , schramm , m. , gottlber , s. , steinmetz , m. , & wisotzki , l. 2008 , mnras , 387 , 13 king , a. 2005 , apjl , 596 , 27 king , a. 2005 , apjl , 635 , 121 king , a. , zubovas , k. , & power , c. 2011 , mnras , 415 , l6 kormendy , j. & ho , l. c. 2013 , ara&a , 51 , 511 komossa , s. 2008 , revmexaa conf .
ser . , 32 , 86 komossa , s. , & xu , d. 2007 , apj , 677 , 33 komossa , s. , xu , d. , zhou , h. , storchi - bergmann , t. , & binette , l. 2008 , apj , 680 , 926 kriss , g. 1994 , adass , 3 , 43 magorrian , j. , et al .
1998 , aj , 115 , 2285 mahony , e. k. , morganti , r. , emonts , b. h. c. , oosterloo , t. a. , & tadhunter , c. 2013 , mnras , 435 , l58 marziani , p. , & sulentic , j. w. 2012 , new astronomy review , 56 , 49 marziani , p. , sulentic , j. w. , zamanov , r. , et al .
2003 , apjs , 145 , 199 mathur , s. 2000 , mnras , 314 , l17 matsuoka , y. 2012 , apj , 2012 , 750 , 54 mclure , r. j. , & dunlop , j. s. 2002 , mnras , 331 , 795 merritt , d. , & ferrarese , l. 2001 , mnras , 320 , l30 menci , n. , fiore , f. , puccetti , s. , & cavaliere , a. 2008 , apj , 686 , 219 morganti , r. , holt , j. , saripalli , l. , oosterloo , t. a. , & tadhunter , c. n. 2007 , a&a , 476 , 735 morganti , r. , oosterloo , t. a. , tadhunter , c. n. , van moorsel , g. , & emonts , b. 2005 , a&a , 439 , 521 mullaney , j. r. , et al . 2012 , apjl , 753 , 30 mullaney , j. r. , alexander , d. m. , fine , s. , goulding , a. d. , harrison , c. m. , & hickox , r. c. 2013 , mnras , 433 , 622 murray , n. , quataert , e. , & thompson , t. a. 2005 , apj , 618 , 569 nandra , k. , laird , e. s. , & steidel , c. c. 2005 , mnras , 360 , l39 nandra , k. , et al .
2007 , apjl , 660 , 11 nesvadba , n. p. h. , lehnert , m. d. , de breuck , c. , gilbert , a. m. , & van breugel , w. 2008 , a&a , 491 , 407 page , m. j. et al . , 2012 , nature , 485 , 213 pounds , k. a. , king , a. r. , page , k. l. , & obrien , p. t. 2003 , mnras , 346 , 1025 rafter , s. e. , crenshaw , d. m. , & wiita , p. j. 2009 , aj , 137 , 42 reeves , j. n. , sambruna , r. m. , braito , v. , & eracleous , m. 2009 , apjl , 702 , 187 rosario , d. j. , et al .
2012 , a&a , 535 , 45 rosario , d. j. , et al .
2013 , apj , 763 , 59 rosario , d. j. , shields , g. a. , taylor , g. b. , salviander , s. , & smith , k. l. 2010 , apj , 716 , 131 rupke , d. s. n. , & veilleux , s. 2011 , apjl , 729 , 21 sanchez , s. f. , et al .
2004 , apj , 641 , 586 santini , p. , et al .
2012 , a&a , 540 , 109 schawinski , k. , thomas , d. , sarzi , m. , maraston , c. , kaviraj , s. , joo , s. , yi , s. k. , & silk , j. 2007 , mnras , 382 , 1415 schawinski , k. , virani , s. , simmons , b. , urry , c. m. , treister , e. , kaviraj , s. , & kushkuley , b. 2009 , apjl , 692 , 19 shankar , f. , bernardi , m. , & haiman , z. 2009 , apjs , 694 , 867 silk , j. , & rees , m. j. 1998 , a&a , 331 , l1 silverman , j. d. , et al .
2008 , apjs , 679 , 118 silverman , j. d. , et al .
2009 , apjs , 696 , 396 spergel , d. n. , et al .
2003 , apjs , 148 , 175 springel , v. , white , s. d. m. , jenkins , a. , et al .
2005 , nature , 435 , 629 somerville , r. s. , hopkins , p. f. , cox , t. j. , robertson , b. e. , & hernquist , l. 2008 , mnras , 391 , 481 stasinska , g. , cid fernandes , r. , mateus , a. , sodre , l. , & asari , n. v. 2006 , mnras , 371 , 972 storchi - bergmann , t. , raimann , d. , bica , e. l. d. , & fraquelli , h. a. 2000 , apj , 544 , 747 tombesi , f. , cappi , m. , reeves , j. n. , & braito , v. 2012 , mnras , 422 , l1 tremaine , s. , gebhardt , k. , bender , r. , et al .
2002 , apj , 574 , 740 treister , e. , et al . 2009,apj , 706 , 535 ueda , y. , akiyama , m. , ohta , k. , & miyaji , t. 2003 , apj , 598 , 886 vanden berk , d. e. , et al .
2001 , aj , 122 , 594 vaona , l. , ciroi , s. , di mille , f. , cracco , v. , la mura , g. , & rafanelli , p. 2012 , mnras , 427 , 1266 veilleux , s. , & osterbrock , d. e. 1987 , apjs , 63 , 295 veron - cetty , m .- p . , veron , p. , & goncalves , a. c. 2001 , a&a , 372 , 730 wang , j. , & wei , j. y. 2008 , apj , 679 , 86 wang , j. , & wei , j. y. 2010 , apj , 719 , 1157 wang , j. , mao , y. f. , & wei , j. y. 2011 , apj , 741 , 50 wang , j. , wei , j. y. , & he , x. t. 2006 , apj , 638 , 106 wang , j. , zhou , x. l. , & wei , j. y. 2013 , apj , 768 , 176 westmoquette , m. s. , smith , l. j. , & gallagher , j. s. iii .
2011 , mnras , 414 , 3719 whittle , m. 1985 , mnras , 213 , 33 whittle , m. , & wilson , a. s. 2004 , aj , 127 , 606 wild , v. , heckman , t. m. , & charlot , s. 2010 , mnras , 405 , 933 wild , v. , kauffmann , g. , heckman , t. , et al .
2007 , mnras , 381 , 543 woo , j. h. , et al .
2010 , apj , 716 , 269 xu , d. w. , & komossa , s. 2009 , apj , 705 , 20 xue , y. q. , et al .
2010 , apj , 720 , 368 zamanov , r. , marziani , p. , sulentic , j. w. , et al .
2002 , apj , 576 , 9 zhang , k. , wang , t. , gaskell , c. m. , & dong , x. 2013 , apj , 762 , 51 zhou , h. y. , wang , t. g. , yuan , w. m. , lu , h. l. , dong , x. b. , wang , j. x. , & lu , y. j. 2006 , apjs , 166 , 128 zubovas , k. , & king , a. r. 2012 , 426 , 2751 zubovas , k. , nayakshin , s. , king , a. , & wilkinson , m. 2013 , mnras , 433 , 3079 measurements after the correction of an underlying agn s continuum and the ones without the correction . _ left panel : for @xmath130 ; _ right panel : for @xmath131 .
see text for the definition of the parameter @xmath14 .
the seyfert galaxies and composite galaxies are shown by the red - solid and blue - open circles , respectively.__,height=188 ] @xmath15007 line parameter @xmath44 ( kurtosis ) is plotted against @xmath43 ( skewness ) for the partially obscured agns .
the seyfert galaxies and composite galaxies are shown by the red - solid and blue - open squares , respectively.,height=188 ] of [ oiii]@xmath15007 line is plotted as a function of the parameters @xmath43 ( left panel ) and @xmath44 ( right panel ) for the partially obscured agns .
the symbols are the same as in figure 3 .
the dotted line in the left panel shows the best fit of the relationship.,height=188 ] index ) plotted against the [ oiii]@xmath15007 line shape parameter @xmath43 and bulk velocity shift @xmath62 for the partially obscured agns .
the symbols are the same as in figure 3 .
the horizontal dashed lines in both panels mark the value of @xmath74 that is used to separate the sample into two parts ( i.e. , with young and old stellar population ) .
the two sub - samples show a significant difference in the distributions of their [ oiii ] line shape and bulk velocity shift .
agns with old host stellar populations are associated with both symmetric [ oiii ] line profile and small bulk velocity shift ( @xmath132 ) .
on the contrary , significant blue wings and bulk blueshift velocities can be only identified in the agns with young stellar populations.,height=188 ] plotted as a function of stellar population age index @xmath3 . the best fit nonlinear relationship ( see eq . 5 in the main text )
is over plotted by the dash line . _
middle and right panels : @xmath4 plotted against @xmath43 and @xmath62 . the vertical dashed line at @xmath82 in the middle panel is used to separate the sample into two sub - samples ( see figure 8) . *
a best fit by the linear least square regression is shown by the dashed line in the right panel .
the symbols are the same as in figure 3.*__,height=188 ] obtained from broad h@xmath2 emission line for the two sub - samples with different [ oiii ] line shape parameter @xmath43 .
the solid and dashed lines represent the objects with @xmath133 and @xmath134 , respectively.,height=188 ] @xmath15007 line width is compared between the partially obscured seyfert galaxies ( blue - dashed line , this paper ) and the seyfert 2 galaxies ( red - solid line ) studied in wang et al .
( 2011 ) . _
middle and right panels : the same as the left panel but for @xmath43 and @xmath62 , respectively.__,height=188 ] versus @xmath135}}$ ] for the partially obscured agns used in the current study .
_ right panel : y - axis denotes the @xmath4 in logarithm estimated from broad h@xmath2 , and * x - axis the eddington ratio @xmath136 estimated from [ oiii ] luminosity and stellar velocity dispersion.*__,height=188 ] | partially obscured agns within a redshift range @xmath0 are used to re - study the role of feedback in the agn - host coevolution issue in terms of their [ oiii]@xmath15007 emission line profile .
the spectra of these objects enable us to determine the agn s accretion properties directly from their broad h@xmath2 emission .
this is essential for getting rid of the `` circular reasoning '' in our previous study of narrow emission - line galaxies , in which the [ oiii ] emission line was used not only as a proxy of agn s bolometric luminosity , but also as a diagnostic of outflow . in addition , the measurement of @xmath3 index is improved by removing an underlying agn s continuum according to the corresponding broad h@xmath2 emission . with these improvements ,
we confirm and reinforce the correlation between @xmath4 and stellar population age .
more important is that this correlation is found to be related to both [ oiii ] line blue asymmetry and bulk blueshift velocity , which suggests a linkage between smbh growth and host star formation through the feedback process .
the current sample of partially obscured agns shows that the composite galaxies have younger host stellar population , higher eddington ratio , less significant [ oiii ] blue wing and smaller bulk [ oiii ] line shift than do the seyfert galaxies .
galaxies : nuclei - galaxies : evolution - quasars : emission line |
globular clusters are densely packed , gravitationally bound systems of several thousands to about one million stars .
the dimensions of the globular clusters are small compared to their distance from us : half of the light is generally emitted within a radius of less than 10 pc , whereas the closest globular cluster has a distance of 2 kpc and 90% lie more than 5 kpc away .
we can thus safely assume that all stars within a globular cluster lie at the same distance from us . with ages in the order of @xmath0 years globular clusters
are among the oldest objects in our galaxy .
as they formed stars only once in the beginning and the duration of that star formation episode is short compared to the current age of the globular clusters the stars within one globular cluster are essentially coeval .
in addition all stars within one globular cluster ( with few exceptions ) show the same initial abundance pattern ( which may differ from one cluster to another ) .
globular clusters are thus the closest approximation to a physicist s laboratory in astronomy . the horizontal branch , which is the topic of this article , was discovered by @xcite , when he used shapley s ( * ? ? ? * ( 1915 ) ) data on m 3 and other clusters to plot magnitude versus colour ( replacing luminosity and spectral type in the hertzsprung - russell diagram ) and thus produced the first colour - magnitude diagrams ( `` farbenhelligkeitsdiagramme '' ) . in these colour - magnitude diagrams ( cmd s ) ten
bruggencate noted the presence of a red giant branch that became bluer towards fainter magnitudes , in agreement with @xcite .
in addition , however , he saw a horizontal branch ( `` horizontaler ast '' ) that parted from the red giant branch and extended far to the blue at constant brightness . as more cmd s of globular clusters
were obtained it became obvious that the relative numbers of red and blue horizontal branch stars ( i.e. the horizontal branch morphology ) varied quite considerably between individual clusters , with some clusters showing extensions of the blue horizontal branch ( so - called `` blue tails '' ) towards bluer colours and fainter visual magnitudes , i.e. towards hotter temperatures to temperature on one hand and by the increasing bolometric correction for hotter stars ( i.e. the maximum of stellar flux is radiated at ever shorter wavelengths for increasing temperatures , making stars fainter at @xmath1 ) on the other hand . ] .
@xcite noted a correlation between the metal abundance and the horizontal branch morphology seen in globular cluster cmd s : the horizontal branch ( * hb * ) became bluer with decreasing metallicity .
about 25 years after the discovery of the horizontal branch @xcite were the first to identify horizontal branch stars with post - red giant branch stars that burn helium in the central regions of their cores .
@xcite managed for the first time to compute zero age horizontal branch ( * zahb * ) models that qualitatively reproduced the observed trend of hb morphology with metallicity without taking into account any mass loss but assuming a rather high helium abundance of y = 0.35 .
@xcite , however , found that they could _ `` account for the observed spread in colour along the horizontal branch by accepting that there is also a spread in stellar mass along this branch , bluer stars being less massive ( on the average ) and less luminous than redder stars . '' _ comparing hb models to observed globular cluster cmd s @xcite found that an hb that _
`` is made up of stars with the same core mass and slightly varying total mass , produces theoretical c - m diagrams very similar to those observed .
a mass loss of perhaps 0.2 m@xmath2 with a random dispersion of several hundredths of a solar mass is required somewhere along the giant branch . '' _ the assumption of mass loss on the red giant branch diminished the need for very high helium abundances .
thus our current understanding sees horizontal branch stars as stars that burn helium in a core of about 0.5 m@xmath2 and hydrogen in a shell and evolve to the asymptotic giant branch , when the helium in the core is exhausted ( * ? ? ?
* ( for a review on hb evolution see sweigart 1994 ) ) .
the more massive the hydrogen envelope is the cooler is the resulting star at a given metallicity .
the masses of the hydrogen envelopes vary from @xmath30.02 m@xmath2 at the hot end of the horizontal branch ( about 30,000 k ) to 0.30.4 m@xmath2 for the cool hb stars at about 40005000 k ( * ? ? ?
* ( depending on metallicity , e.g. dorman et al .
the stable red and blue hb stars are separated by the variable rr lyrae range at about 65007500 k. this article deals with blue hb stars , which at effective temperatures of about 8,000 k to 20,000 k show spectra rather similar ( at moderate resolution ) to main sequence stars of spectral types a and b and are therefore called hba and hbb stars . in the field of the milky way such stars
are often denominated by fhb ( field hb star ) and used as tracers for halo structure .
already early studies of hba and hbb stars in globular clusters showed discrepancies between observational results and theoretical expectations : @xcite mentioned that the comparison of @xmath4 vs. @xmath5 for 50 blue hb stars in ngc 6397 to models from @xcite indicated low surface gravities and a mean mass of 0.3m@xmath2 ( 0.4m@xmath2 ) for solar ( negligible ) helium abundance , assuming @xmath6 12.0 and e@xmath7 = 0.16 .
later spectroscopic analyses of hb stars ( see cited papers for details ) in globular clusters reproduced this effect ( cf .
fig . [ ag_plottga1 ] ) : ( * ? ? ?
* crocker ( 1988 , m 3 , m 5 , m 15 , m 92 , ngc 288 ) ) , ( * ? ?
? * de boer ( 1995 , ngc 6397 ) ) , ( * ? ? ?
* ; * ? ? ?
* moehler ( 1995 , 1997a , m 15 ) ) , ( * ? ? ?
* moehler ( 1997b , ngc 6752 ) ) . the zero age hb ( zahb ) in fig .
[ ag_plottga1 ] marks the position where the hb stars have settled down and started to quietly burn helium in their cores .
the terminal age hb ( tahb ) is defined by helium exhaustion in the core of the hb star ( @xmath8 ) . for
temperatures between 12,000 k and 20,000 k the observed positions in the ( logg , t@xmath9)-diagram fall mostly above the zahb and in some cases even above the tahb .
this agrees with the finding of @xcite that field hbb stars show a larger scatter away from the zahb in t@xmath9 , logg than the hotter subdwarf b stars with t@xmath10 20,000 k. knowing the atmospheric parameters of the stars and the distances to the globular clusters we can determine masses for the stars ( * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
moehler 1995 , 1997b , de boer 1995 ) ) .
while the stars in m 3 , m 5 , and ngc 6752 have mean masses consistent with the canonical values , the hot hb stars in all other clusters show masses that are significantly lower than predicted by canonical hb evolution even for temperatures cooler than 12,000 k. also some uv observations suggest discrepancies between theoretical expectations and observational results : the iue ( international ultraviolet explorer ) and hut ( hopkins ultraviolet telescope ) spectra of m 79 ( * ? ? ?
* ; * ? ? ?
* ( altner & matilsky 1993 , dixon 1996 ) ) suggest lower than expected gravities and higher than expected metallicities for hot hb stars ( * ? ? ?
* ( but see vink 1999 , who do not need low surface gravities to fit the hut data ) ) .
@xcite find from ultraviolet imaging telescope ( uit ) photometry of m 79 that stars bluer than @xmath11 = @xmath120.2 lie above the zahb , whereas cooler stars scatter around the zahb .
uit data of m 13 ( * ? ? ?
* ( parise 1998 ) ) find a lack of stars close to the zahb at a colour ( temperature ) range similar to the low logg range shown in fig .
[ ag_plottga1 ] .
@xcite on the other hand find good agreement between uit photometry of blue stars in ngc 6752 and a standard zahb .
it has been realized early on that the blue hb and blue tail stars in globular clusters show weaker helium lines than field main sequence b stars of similar temperatures : ( * ? ? ?
* searle & rodgers ( 1966 , ngc 6397 ) ) ; ( * ? ? ?
* greenstein & mnch ( 1966 , m 5 , m 13 , m 92 ) ) ; ( * ? ? ?
* sargent ( 1967 , m 13 , m 15 , m 92 ) ) .
@xcite already suggested diffusion to explain this he deficiency .
@xcite performed the first theoretical study of diffusion effects in hot horizontal branch stars . using the evolutionary tracks of @xcite they found for the metal - poor models that _ `` in most of each envelope , the radiative acceleration on all elements ( i.e. c , n
, o , ca , fe ) is much larger than gravity which is not the case in main - sequence stars . ''
_ the elements are thus pushed towards the surface of the star .
turbulence affects the different elements to varying extent , but generally reduces the overabundances .
models without turbulence and/or mass loss ( which may reduce the effects of diffusion ) predict stronger he depletions than observed .
a weak stellar wind could alleviate this discrepancy ( * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ( heber 1986 , michaud 1989 , fontaine & chayer 1997 , and unglaub & bues 1998 discuss this effect , albeit for hotter stars ) ) .
the extent of the predicted abundance variations varies with effective temperature , from none for hb stars cooler than about @xmath13k ( due to the very long diffusion timescales ) to 2 4 dex in the hotter stars ( the hottest model has t@xmath9 = 20,700
k ) and also depends on the element considered .
observations of blue hb and blue tail stars in globular clusters support the idea of diffusion being active above a certain temperature : abundance analyses of blue hb stars cooler than 11,000 k to 12,000 k in general show no deviations from the globular cluster abundances derived from red giants , while for hotter stars departures from the general globular cluster abundances are found , e.g. iron enrichment to solar or even super - solar values and strong helium depletion ( see moehler 2001 for references and more details ) .
this agrees with the finding of @xcite and @xcite that solar metallicity model atmospheres are required to fit the uv spectra of m 79 .
all this evidence supports the suggestion of @xcite that the onset of diffusion in stellar atmospheres may play a rle in explaining the jump along the hb towards brighter @xmath14 magnitudes at effective temperatures of about 11,500 k. this jump in @xmath15 is seen in all cmd s of globular clusters that have strmgren photometry of sufficient quality .
the effective temperature of the jump is roughly the same for all clusters , irrespective of metallicity , central density , concentration or mixing evidence , and coincides roughly with the onset of the `` low logg problem '' seen in fig . [ ag_plottga1 ] at t@xmath9 @xmath1611,000 k to 12,000 k. this in turn coincides with the region where surface convection zones due to hydrogen and hei ionization disappear in hb stars ( * ? ? ? * ( sweigart 2002 ) ) .
radiative levitation of heavy elements decreases the far - uv flux and by backwarming increases the flux in @xmath14 .
@xcite show that the use of metal - rich atmospheres ( [ fe / h ] = @xmath17 for scaled - solar atlas9 kurucz model atmospheres with @xmath18 ) improves the agreement between observed data and theoretical zahb in the @xmath15-cmd at effective temperatures between 11,500 k and 20,000 k , but it worsens the agreement between theory and observation for hotter stars in the strmgren cmd of ngc 6752 ( see their fig .
8) . thus diffusion may either not be as important in the hotter stars or the effects may be diminished by a weak stellar wind .
the gap at @xmath19 discussed by @xcite is not directly related to the @xmath14-jump as it corresponds to an effective temperature of about 9000 k and is also not seen in every cluster ( which would be expected if it were due to an atmospheric phenomenon ) .
the gap at t@xmath9 @xmath16 13,000 k seen in the @xmath20 diagram of field horizontal branch stars ( * ? ? ?
* ( newell 1973 , newell & graham 1976 ) ) may be related to the @xmath14-jump as the @xmath21 index contains @xmath14 . the abundance distribution within a stellar atmosphere influences the temperature stratification and thereby the line profiles and the flux distribution of the emergent spectrum .
a deviation in atmospheric abundances of hb stars from the cluster metallicity due to diffusion would thus affect their line profiles and flux distribution .
model atmospheres calculated for the cluster metallicity may then yield wrong results for effective temperatures and surface gravities when compared to observed spectra of hb stars .
this effect could explain at least part of the observed discrepancies ( see sect . [ hb_final ] for more details ) .
self - consistent model atmospheres taking into account the effects of gravitational settling and radiative levitation are , however , quite costly in cpu time and have started to appear only quite recently for hot stars ( * ? ? ?
* ; * ? ? ?
* ( dreizler & wolff 1999 , hui - bon - hoa 2000 ) ) .
* ; * ? ? ?
* ; * ? ? ?
* peterson ( 1983 , 1985a , 1985b ) ) found from high - resolution spectroscopic studies that clusters with bluer hb morphologies show higher rotation velocities among their hb stars .
however , the analysis of @xcite shows that while the stars in m 13 ( which has a long blue tail ) rotate on average faster than those in m 3 ( which has only a short blue hb ) , the stars in ngc 288 and m 13 show _ slower _ rotation velocities at _ higher _ temperatures .
these results are consistent with those reported by @xcite , who determined rotational velocities for stars as hot as 19,000 k in m 13 .
they found that stars hotter than about 11,000 k have significantly lower rotational velocities than cooler stars and that the change in mean rotational velocity may coincide with the gap seen along the blue hb of m 13 ( * ? ? ?
* ( see ferraro 1998 for an extensive discussion of gaps ) ) .
also the results of ( * ? ? ?
* cohen & mccarthy ( 1997 , m 92 ) ) and ( * ? ? ?
* behr ( 2000b , m15 ) ) show that hb stars cooler than @xmath1611,000 k to 12,000 k in general rotate faster than hotter stars . the studies by @xcite and (
* ; * ? ? ?
* recio ( 2002 , 2004 ) ) , which both consider several clusters , confirm the trends mentioned above .
below the diffusion threshold about 20% to 30% of the blue hb stars show rotation velocities of @xmath22 km / s , whereas there are no fast rotators among the hotter stars . from observations of field hb stars @xcite and @xcite note that red and cool blue hb stars show similar distributions of rotational velocities ( after accounting for the larger radii of the red hb stars ) , whereas field rr lyrae stars show no evidence for rotation .
@xcite study theoretical models for the rotation of hb stars and find that the observed rotation of cool blue hb stars in m 13 can be explained if their red giant precursors have rapidly rotating cores and differential rotation in their convective envelopes and if angular momentum is redistributed from the rapidly rotating core to the envelope on the horizontal branch .
if , however , turn - off stars rotate with less than 4 km / s , a rapidly rotating core in the main - sequence stars ( violating helioseismological results for the sun ) or an additional source of angular momentum on the red giant branch ( * ? ? ?
* ( e.g. mass transfer in close binaries or capture of a planet as described by soker & harpaz 2000 ) ) are required to explain the rotation of blue hb stars .
@xcite speculate that such differential rotation could be understood within the framework of internal gravitational waves .
the change in rotation rates towards higher temperatures is not predicted by the models of @xcite but could be understood as a result of gravitational settling of helium , which creates a mean molecular weight gradient , that then inhibits angular momentum transport to the surface of the star .
@xcite suggests that the weak stellar wind invoked to reconcile observed abundances in hot hb stars with diffusion calculations ( cf .
[ sec - diff ] ) could also carry away angular momentum from the surface layers and thus reduce the rotational velocities of these stars .
@xcite argue that the distribution of rotational velocities along the hb can be explained by spin - up of the progenitors due to interaction with low - mass companions , predominantly gas - giant planets , in some cases also brown dwarfs or low - mass main - sequence stars ( esp .
for the very hot hb stars ) . the slower rotation of the hotter stars in their scenario is explained by mass loss _ on _ the hb , which is accompanied by efficient angular momentum loss .
this scenario , however , does not explain the sudden change in rotational velocities and the coincidence of this change with the onset of radiative levitation .
analysis of a larger sample of hot hb stars in ngc 6752 ( * ? ? ?
* ( moehler 2000 ) ) showed that the use of model atmospheres with solar or super - solar abundances removes much of the deviation from canonical tracks in effective temperature , surface gravity and mass for hot hb stars discussed in sect .
[ other ] .
however , some discrepancies remain , indicating that the low logg , low mass problem can not be completely solved by scaled - solar metal - rich atmospheres ( * ? ? ? * ( which _ do _ reproduce the @xmath14-jump reported by grundahl 1999 ) ) .
as @xcite noted diffusion will not necessarily enhance all heavy elements by the same amount and the effects of diffusion vary with effective temperature .
elements that were originally very rare may be enhanced even stronger than iron .
the question of whether diffusion can fully explain the `` low gravity '' problem can not be answered without detailed abundance analyses to determine the actual abundances and subsequent analyses using model atmospheres with non - scaled solar abundances ( * ? ? ?
* ( like atlas12 , kurucz 1992 ) ) .
additional caution is also recommended by the results of @xcite on m 13 , where even the use of the metal - rich model atmospheres does not eliminate the problem of the low masses or the low gravities for effective temperatures between 12,000 k and 16,000 k. in that cluster we also found significant disagreement between atmospheric parameters derived from strmgren photometry or from balmer line fitting for stars cooler than about 9,000 k , for which we found no explanation .
still unexplained are also the low masses found for cool blue hb stars ( which are not affected by diffusion ) in , e.g. , ngc 6397 and m 92 . for those stars a longer distance scale to globular clusters would reduce the discrepancies ( * ? ? ?
* ( moehler 1999 ) ) .
such a longer distance scale has been suggested by several authors using hipparcos results for metal - poor field subdwarfs to determine the distances to globular clusters by fitting their main sequence with the local subdwarfs .
one should note , however , that @xcite , report that hipparcos parallaxes for field hba stars still yield masses significantly below the canonical mass expected for these objects .
@xcite present an extensive and excellent discussion of various globular cluster distance determinations and the zoo of biases that affect them , while @xcite concentrate on the error budget of distances from main sequence fitting in their paper . altner b. , matilsky t.a . , 1993 , apj 410 , 116 behr , b.b .
2003a , apjs 149 , 67 behr , b.b .
2003b , apjs 149 , 101 behr b.b . , djorgovski s.g .
, cohen j.g . , 2000a , apj 528 , 849 behr b.b . ,
cohen j.g . ,
mccarthy j.k . , 2000b , apj 531 , l37 caloi v. , 1999 , a&a 343 , 904 carney , b.w . , latham , d.w . ,
stefanik , r.p . ,
laird , j.b . , morse , j.a .
2003 , aj 125 , 293 carretta e. , gratton r. , clementini g. , fusi pecci f. , 2000 , apj 533 , 215 charbonneau p. , michaud g. , 1988 , apj 327 , 809 cohen j.g .
, mccarthy j.k . , 1997 , aj 113 , 1353 crocker d.a . , rood r.t . ,
oconnell r.w . , 1988 , apj 332 , 236 de boer k.s . , schmidt j.h.k . , heber u. , 1995 , a&a 303 , 95 de boer k.s . , tucholke h
.- j . , schmidt j.h.k . , 1997 ,
a&a 317 , l23 dixon w.v . ,
davidsen a.f . , dorman b. , ferguson h.c . , 1996 , aj 111 , 1936 dorman b. , rood , r.t . , oconnell , w.o . , 1993 , apj 419 , 596 dreizler s. , wolff b. , 1999 , a&a 348 , 189 faulkner j. , 1966 , apj 144 , 978 ferraro f.r . , paltrinieri b. , fusi pecci f. , dorman b. , rood r.t . , 1998 , apj 500 , 311 fontaine g. , chayer p. , 1997
, in _ the 3@xmath23 conf . on faint blue stars
philip , j. liebert & r.a .
saffer ( l. davis press , schenectady ) , p. 169
graham j.a . ,
doremus c. , 1966 , aj 73 , 226 gratton r.g .
, bragaglia a. , carretta e. , clementini g. , desidera s. , grundahl f. , lucatello s. , 2003 , a&a 408 , 529 greenstein g.s . ,
mnch g. , 1966 , apj 146 , 518 greenstein g.s .
, truran j.w . , cameron a.g.w . , 1967 ,
nature 213 , 871 grundahl f. , catelan m. , landsman w.b . , stetson p.b . , andersen m. , 1999 , apj 524 , 242 heber u. , 1986 , a&a 155 , 33 hill r.s . , cheng k .- p . ,
smith e.p . ,
1996 , aj 112 , 601 hoyle f. , schwarzschild m. , 1955 , apjs 2 , 1 hui - bon - hoa a. , leblanc f. , hauschildt p. , 2000 , apj 535 , l43 iben i.jr . , rood r.t . , 1970 ,
apj 161 , 587 kurucz r.l . , 1992 , in _ the stellar populations of galaxies _ , eds .
b. barbuy & a. renzini , iau symp .
149 ( kluwer : dordrecht ) , 225 landsman w.b .
, sweigart a.v .
, bohlin r.c . , 1996 , apj 472 , l93 michaud g. , 1982 , apj 258 , 349 michaud g. , vauclair g. , vauclair s. , 1983 , apj 267 , 256 michaud g. , bergeron p. , heber u. , wesemael f. , 1989 , apj 338 , 417 mihalas d.m .
, 1966 , apjs 13 , 1 moehler s. , 1999 , reviews in modern astronomy , ed .
r. schielicke , vol .
12 , p. 281 moehler s. , 2001 , pasp 113 , 1162 moehler s. , heber u. , de boer k.s . , 1995 , a&a 294 , 65 moehler s. , heber u. , durrell p. , 1997a
, a&a 317 , l83 moehler s. , heber u. , rupprecht g. , 1997b , a&a 319 , 109 moehler s. , sweigart a.v .
, landsman w. , heber u. , 2000 , a&a 360 , 120 moehler s. , landsman w.b . , sweigart a.v .
, grundahl f. , 2003 , a&a 405 , 135 newell e.b . , 1973 ,
apjs 26 , 37 newell e.b .
, graham j.a . , 1976 ,
apj 204 , 804 parise r.a . , bohlin r.c . ,
neff s.g . , 1998 ,
apj 501 , l67 peterson r.c . , 1983 , apj 275 , 737 peterson r.c .
, 1985a , apj 289 , 320 peterson r.c . , 1985b , apj 294 , l35 peterson r.c . , rood r.t . , crocker d.a . , 1995 , apj 453 , 214 recio - blanco , a. , piotto , g. , aparicio , a. , renzini , a. 2002 , apj 527 , l71 recio - blanco , a. , piotto , g. , aparicio , a. , renzini , a. 2004 , a&a 417 , 597 rood r.t . , 1973 , apj 184 , 815 saffer r.a . , keenan f.p
, hambly n.c .
, dufton p.l . , liebert j. , 1997 , apj 491 , 172 sandage a.r . , wallerstein g. , 1960 , apj 131 , 598 sargent w.l.w . , 1967 ,
apj 148 , l147 searle l. , rodgers a.w . , 1966 , apj 143 , 809 shapley h. , 1915 , contr .
wilson 116 sills a. , pinsonneault m.h . , 2000 , apj 540 , 489 soker n. , harpaz a. , 2000 , mnras 317 , 861 sweigart a.v .
, 1994 , in _ hot stars in the galactic halo _
s. adelman , a. upgren , c.j .
adelman , ( cambridge university press , cambridge ) , p. 17
sweigart a.v . , 2002 ,
highlights of astronomy , vol .
h. rickman ( asp ) , p. 292
sweigart a.v . ,
gross p.g . , 1976 ,
apjs 32 , 367 talon s. , charbonnel c. , 2004 , a&a 418 , 1051 ten bruggencate p. , 1927 , _ sternhaufen _ ( julius springer verlag , berlin ) unglaub k. , bues i. , 1998 , a&a 338 , 75 vink j.s . ,
heap s.r . ,
sweigart a.v .
, lanz t. , hubeny i. , 1999 , a&a 345 , 109 | globular clusters offer ideal laboratories to test the predictions of stellar evolution .
when doing so with spectroscopic analyses during the 1990s , however , the parameters we derived for hot horizontal branch stars deviated systematically from theoretical predictions .
the parameters of cooler , a - type horizontal branch stars , on the other hand , were consistent with evolutionary theories . in 1999 , two groups independently suggested that diffusion effects might cause these deviations , which we verified subsequently .
i will discuss these observations and analyses and their consequences for interpreting observations of hot horizontal branch stars . |
let @xmath0 be a prime number , @xmath1 its power .
let @xmath2 be a natural number prime to @xmath0 .
we consider the family of curves over @xmath3 defined by an equation of the form [ as]y^p - y = f(x)=a_dx^d+ ... +a_1xwith @xmath4 and @xmath5 for all @xmath6 divisible by @xmath0 ( every curve defined by an equation of the form @xmath7 with @xmath8 $ ] of degree @xmath2 is a twist of a curve of the form satisfying this condition , see section [ geom ] ) .
we call such curves artin - schreier curves , or a - s curves for short , and the corresponding family the a - s family ( with parameter @xmath2 ) .
denote by @xmath9 the set of nontrivial additive characters of @xmath10 .
it is known that the l - function of the normalisation of the projective closure of a curve defined by factors into primitive l - functions as follows : [ pr]l_f(z)=_l_f,(z)with [ deflpsi]l_f,(z)=_r=1^__q^r__q^r/_pf().each of the @xmath11 factors in is a polynomial of degree @xmath12 with all zeroes having absolute value @xmath13 due to the riemann hypothesis for curves over finite fields ( see section [ lfunc ] ) .
denote by @xmath14 the set of polynomials @xmath15 of the form in satisfying the stated conditions .
we denote @xmath16 where @xmath17 are the normalised zeroes of @xmath18 counting multiplicity .
the zeroes are normalised as follows : @xmath19 , where @xmath20 are the zeroes of @xmath18 .
we have @xmath21 .
note that the normalised zeroes are proportional to the inverse zeroes of @xmath22 and we preserve this normalisation convention for zeroes of l - functions throughout the paper .
the quantity @xmath23 is the trace of the @xmath24-th power of the frobenius element corresponding to the l - function @xmath18 . for any finite set @xmath25 and a function @xmath26 ( we denote by @xmath27 the set of integers , rational , real and complex numbers respectively ) we denote by @xmath28 the average of @xmath29 as @xmath30 runs uniformly through @xmath25 , in other words @xmath31 denote @xmath32 it does not depend on the choice of @xmath33 , see section [ asfam ] . our main result is the following [ thm1]@xmath34 where [ e ] e_p , r=(the implicit constant is absolute , i.e. does not depend on @xmath35 ) .
note that the error term in theorem [ thm1 ] is small when @xmath36 for any fixed @xmath37 . for @xmath38 theorem [ thm1 ] suggests that for @xmath39 the zeroes of all the l - functions in this family when taken together are distributed quite uniformly on the unit circle .
we conjecture that the average number of zeroes of @xmath18 contained in an arc of length @xmath40 on the unit circle as @xmath15 varies uniformly through @xmath14 tends to the length of the arc divided by @xmath41 as @xmath39 . for arcs of length
@xmath42 s.t .
@xmath43 this has recently been proved in @xcite .
the conjecture is related to the random unitary matrix model for a - s l - functions which we present in section [ rmm ] .
we are only able to obtain a weaker result with the arc replaced by a window function with bounded frequency .
denote by @xmath44 the space of smooth complex - valued functions on the real line with all derivatives decaying faster than any power of @xmath45 at infinity ( the schwartz space ) .
[ thm2 ] assume @xmath38 .
let @xmath46 be a function s.t .
its fourier transform @xmath47 is supported on the interval @xmath48 .
denote @xmath49 this function has period @xmath41 .
let @xmath50 be any real number .
denote @xmath51 where @xmath52 are the normalised zeroes of @xmath18 , @xmath53 being real numbers well defined modulo @xmath41 .
then @xmath54 as @xmath39 and @xmath55 may vary as we please , i.e. there is a bound on the decay rate of the @xmath56 term which depends on @xmath2 and @xmath57 but not on @xmath55 ( in fact the error becomes smaller as @xmath55 grows ) .
we conjecture that this in fact holds for any @xmath46 , but with the presently existing methods it seems difficult to prove .
we also consider some nonlinear statistics of the zeroes of l - functions in the a - s family .
let @xmath58 be a two variable window function and @xmath59 the periodic window function associated with @xmath57 by @xmath60 let @xmath61 be the normalised zeroes of @xmath18 for @xmath62 , @xmath52 and let @xmath50 be some fixed real number .
we consider the 2-level density function ( at @xmath50 ) : @xmath63 for a function @xmath58 we define its fourier transform by @xmath64 in section [ nonlinear ] we will prove the following [ thm2p]assume @xmath38 .
let @xmath58 be a window function s.t .
its fourier transform @xmath65 is supported on the set @xmath66 .
then for all @xmath50 , @xmath67 as @xmath39 .
the connection with the random unitary matrix model is discussed in section [ nonlinear ] .
we also consider the subfamily of a - s curves defined by with odd polynomial @xmath68 .
for an odd natural number @xmath2 denote by @xmath69 the subset of @xmath62 s.t .
@xmath70 . in section [ odd ] we prove the following [ oddthm ] assume @xmath38 and @xmath2 is odd .
there exists a positive constant @xmath71 ( in fact any @xmath72 will do ) s.t .
@xmath73 provided that @xmath74 and @xmath75 ( @xmath76 is defined by ) .
this result agrees with a random symplectic matrix model for the l - zeroes in the family @xmath69 ( see section [ odd ] ) .
we consider a more general type of families of l - functions corresponding to dirichlet characters and show that the a - s family ( as well as the a - s family with odd polynomials ) is a special case .
we also indicate how our results for the a - s family can be generalised to such families .
this occupies section [ secdirg ] .
we also consider a related problem - the distribution of the number of points of a curve varying uniformly in a family of a - s curves .
the proofs of our results are presented in section [ numpoints ] .
this part is independent of the rest of our work and the interested reader may skip to section [ numpoints ] after section [ back ] .
we consider the distribution of the number of points on the curve @xmath77 as @xmath15 varies uniformly through the a - s family and @xmath39 . here
we denote by @xmath77 the normalisation of the projective closure of the curve defined by for @xmath78 $ ] .
we define @xmath79 to be the set of all
monic degree @xmath2 polynomials in @xmath80 $ ] . for the problem of the distribution of the number of points it is more convenient to consider the family of a - s curves defined by the polynomials in @xmath79 .
it is not difficult to adjust the statements and proofs for the case of the family @xmath14 . for the rest of this section let @xmath24 be a fixed natural number .
we will see in section [ geom ] that the number of @xmath81-rational points on an a - s curve @xmath77 always equals 1 modulo @xmath0 and so we denote @xmath82 .
our results concern the distribution of @xmath83 as @xmath15 varies uniformly in the family @xmath79 .
we denote by @xmath84 the number of monic degree @xmath85 irreducible polynomials in @xmath86 $ ] .
we denote by @xmath87 the bernoulli random variable which assumes 1 with probability @xmath45 and 0 with probability @xmath88 .
for two random variables @xmath89 we write @xmath90 if they have the same distribution .
[ t1 ] let @xmath91 be fixed .
for each @xmath92 let @xmath93 be random variables with all @xmath94 independent .
then for @xmath95 the following holds : if @xmath96 , the distribution of @xmath83 equals that of @xmath97 in particular the mean value of @xmath83 is @xmath98 .
if @xmath99 then the distribution of @xmath100 equals that of @xmath101 in particular the mean value of @xmath83 is @xmath102 we also consider what happens when @xmath91 are allowed to vary .
[ t2 ] let @xmath103 both tend to infinity and @xmath104 .
then @xmath83 converges in distribution to the poissonian distribution with mean 1 , i.e. @xmath105 [ t3 ] let @xmath0 be fixed and @xmath106 .
if @xmath96 or @xmath38 then @xmath107 converges in distribution to the gaussian distribution with mean 0 and variance 1 . if @xmath108 and @xmath24 is even then @xmath109 converges in distribution to the gaussian distribution with mean 0 and variance 1 .
[ t4 ] let @xmath103 both tend to infinity and assume that @xmath110 ( not necessarily constant ) or @xmath111 .
then @xmath112 converges in distribution to the gaussian distribution with mean 0 and variance 1 .
the paper is organised as follows : in the next section we review related work dealing with similar problems for other families of l - functions . in section [ back ]
we provide the necessary background on a - s curves and l - functions . in section [ rmm ]
we describe the random unitary matrix model for the a - s family of l - functions . in section [ pmain ]
we prove theorem [ thm1 ] and derive theorem [ thm2 ] from it . in section [ nonlinear ]
we consider the 2-level density statistics of the l - zeroes in the a - s family and obtain results which agree with the random unitary matrix model . in section [ secdirg ]
we reformulate our main results in terms of dirichlet l - functions over @xmath80 $ ] and generalise them to suitable families of dirichlet l - functions . in section
[ odd ] we consider the family of a - s curves defined by with @xmath15 odd , for which we formulate conjectures corresponding to a random symplectic matrix model and provide evidence for them in the form of theorem [ oddthm ] . in section [ p2 ]
we will discuss the situation with @xmath108 . in section [ numpoints ]
we prove our results on the distribution of the number of points on curves in the a - s family .
the main inspiration for the present work is the paper @xcite , which studies similar questions and obtains similar results for an ensemble of hyperelliptic curves .
we briefly present the content of that work .
one considers the family of curves over @xmath113 , with @xmath55 odd , defined by equations of the form @xmath114 , with @xmath15 monic squarefree of degree @xmath2 , with @xmath2 odd .
denote by @xmath115 the set of all such polynomials @xmath15 . for @xmath116
the curve defined by @xmath114 has an l - function which is a polynomial with integer coefficients , which can be written as @xmath117 where @xmath17 are the normalised zeroes of @xmath118 satisfying @xmath21 .
denote @xmath119 .
it is shown in @xcite that @xmath120 where @xmath76 is given by and the error term @xmath121 satisfies @xmath122 for @xmath123 ( if @xmath124 an additional summand of @xmath125 appears ) .
this provides evidence in favor of the random symplectic matrix model for the l - functions of hyperelliptic curves because @xmath126 this result is used to obtain a result about the average number of zeroes in short intervals , which again agrees with the random symplectic matrix model .
namely , let @xmath57 be as in theorem [ thm2 ] but with fourier transform supported in @xmath127 and define @xmath128 as in theorem [ thm2 ]
. denote @xmath129 , where @xmath130 are the normalised zeroes of @xmath131 ( @xmath132 are real numbers well defined modulo @xmath41 ) and @xmath133 , where @xmath134 are the eigenvalues of a matrix @xmath135 .
then @xmath136 as @xmath39 and @xmath55 is fixed . note that @xmath115 has on the order of @xmath137 elements while @xmath138 , which means that the a - s family is sparser and we get less averaging . as a result
we get large errors in our estimate for @xmath139 already for @xmath140 , while in the hyperelliptic case it is possible to obtain small errors for @xmath141 .
much more is known about the statistics of zeroes for various families of l - functions over finite fields if the degree of the family is held constant while @xmath142 .
for an l - function of the form @xmath143 we attach the class of unitary matrices with eigenvalues @xmath144 . for many families of l - functions (
e.g. the hyperelliptic family , the family of dirichlet characters and families similar to our a - s family ) it was shown by katz and sarnak ( see @xcite ) that as @xmath142 and @xmath145 is fixed the classes corresponding to the l - functions of the objects in the family become equidistributed in the set of conjugacy classes of a suitable compact group of matrices ( endowed with the measure induced by the haar measure on the group ) , usually @xmath146 , the orthogonal group or some similar group , called the symmetry type of the family . these results can not be extended to the case @xmath147 ( there is no meaning to equidistribution in a varying space ) but the symmetry types observed for constant @xmath145 can be used to give random matrix models to families of l - functions .
the model for our family of a - s l - functions is presented in section [ rmm ] .
the distribution of the number of points on curves in various families has been studied extensively in recent years .
it follows from , below that the number of points on a curve @xmath71 with normalised l - zeroes @xmath148 ( @xmath149 is the genus of the curve ) is @xmath150 .
the distribution of @xmath151 as @xmath71 varies through some family of curves over @xmath113 with genus @xmath149 as @xmath152 is considered in @xcite for the hyperelliptic family , in @xcite,@xcite for the family of trigonal covers of @xmath153 ( with further generalisation to @xmath42-gonal covers , @xmath42 a prime dividing @xmath154 ) and in @xcite for the family of plane curves .
a family of curves in higher dimensional projective spaces has been studied in @xcite .
the distribution of the number of points for the family of a - s curves is considered in section [ numpoints ] .
results similar to theorem [ thm2 ] have previously been obtained for families of l - functions over number fields .
the family of quadratic dirichlet l - functions ( with varying modulus ) is considered in @xcite and the family of all dirichlet l - functions with given modulus is considered in @xcite
. for example let @xmath55 be a prime number and @xmath155 a dirichlet character modulo @xmath55 .
let @xmath156 be the corresponding l - series and @xmath157 its sequence of nontrivial zeroes ordered by increasing absolute value . for simplicity
we assume the generalised riemann hypothesis ( although it is not assumed in @xcite ) so that @xmath158 .
denote @xmath159 .
denote by @xmath160 the set of nontrivial dirichlet characters modulo @xmath55 . the average number of l - zeroes satisfying @xmath161 as @xmath155 varies uniformly in @xmath160 , @xmath162 is fixed and @xmath142 is known to be @xmath163 so we normalise @xmath164 ( now we expect on average one zero with @xmath165 ) .
let @xmath46 be a window function , @xmath166 .
it is shown in @xcite that if @xmath167 is supported on @xmath168 then @xmath169 as @xmath170 .
studying nonlinear statistics they obtain agreement with a random unitary matrix model ( with matrix size around @xmath171 ) for restricted classes of window functions .
the material reviewed in this section can be found in @xcite,@xcite,@xcite .
the notation and conventions introduced in this subsection apply to the entire paper , including the introduction .
when we use the @xmath172-notation ( asymptotic bound ) the implicit constant is absolute , except when the bounded quantity depends on a window function @xmath57 , in which case it may depend on @xmath57 . if there are additional parameters upon which the bound depends we write them explicitly as a subscript ( e.g. @xmath173 ) .
the @xmath174-notation is always used for @xmath39 and @xmath175 implies that @xmath176 as @xmath39 regardless of how the other parameters on which @xmath177 depend vary , except possibly for a window function @xmath57 which is always assumed to be fixed . for a pair of integers @xmath178
we denote by @xmath179 their greatest common divisor . for a pair of polynomials @xmath177
, @xmath180 denotes their greatest monic common divisor and @xmath181 denotes the residue class of @xmath15 modulo @xmath149 . for a finite set @xmath182
we denote by @xmath183 its number of elements .
let @xmath184 be a field of characteristic @xmath185 .
an artin - schreier ( a - s shortly ) curve over @xmath184 is the normalisation of the projective closure of the affine curve defined by an equation of the form @xmath7 with @xmath186 $ ] a polynomial of degree @xmath187 .
we denote this curve by @xmath77 . if @xmath188 then @xmath77 is geometrically irreducible @xcite ( however this condition is far from necessary for geometric irreducibility ) .
we assume throughout that @xmath188 .
the affine part of @xmath77 is smooth , as @xmath189 never vanishes .
the curve @xmath77 has exactly one point ( always @xmath184-rational ) outside its affine part and the genus of @xmath77 is @xmath190 ( this follows from the material of @xcite ) .
let @xmath191 be an affine point on the curve @xmath77 , possibly defined over the algebraic closure of @xmath184
. then @xmath192 are all points on @xmath77 and these are all the points of @xmath77 with abscissa @xmath193 .
we see in particular that if @xmath194 is a finite field then the number of points on the affine part of @xmath77 is divisible by @xmath0 and the total number of points ( including the single point at infinity ) equals 1 modulo @xmath0 .
if @xmath195 for some @xmath196 then @xmath197 is @xmath184-isomorphic to @xmath77 via the substitution @xmath198 . if @xmath184 is a finite field then every element of @xmath184 is a @xmath0-th power and so every curve @xmath77 with @xmath199 is isomorphic to a curve @xmath197 s.t .
@xmath200 with @xmath201 for all @xmath202 .
now assume that @xmath203 is a finite field .
an element @xmath196 can be written as @xmath204 iff @xmath205 ( this follows from the hilbert 90 theorem or more simply by noting that the map @xmath206 is linear with one - dimensional kernel @xmath10 , while its image is contained in the kernel of the trace map ) . if @xmath207 with @xmath196 satisfying @xmath205 then @xmath197 is isomorphic to @xmath77 via @xmath208 , where @xmath209 . even if @xmath210 the curves @xmath197 becomes isomorphic to @xmath77 over @xmath211 ( because @xmath212 ) .
if @xmath213 we say that @xmath197 is a twist of @xmath77 .
let @xmath214 be a finite field .
let @xmath71 be a smooth projective curve over @xmath113 with genus @xmath149 . denote by @xmath215 the number of @xmath81-points on @xmath71 .
the l - function of @xmath71 is defined by the power series [ defl]l(z)=_r=1^z^r.it turns out that @xmath216 is a polynomial of degree @xmath217 and in fact we may write [ zeros]l(z)=_i=1 ^ 2g(1-_iq^1/2z).the normalised zeroes @xmath17 come in conjugate pairs and they all satisfy @xmath21 ( the riemann hypothesis for curves over a finite field ) . for all these properties of the l - function of a curve over a finite field
see @xcite , @xcite , @xcite .
let @xmath218 be a primitive @xmath0-th ( complex ) root of unity .
we define an additive character @xmath219 by @xmath220 ( this is well defined ) .
all the nontrivial additive characters of @xmath10 are of this form and there are @xmath11 characters corresponding to the @xmath11 roots of unity .
we denote by @xmath221 the trace map from the field with @xmath222 elements to the field with @xmath45 elements , provided @xmath222 is a power of @xmath45 .
now let @xmath186 $ ] be a nonconstant polynomial of degree @xmath2 ( we always assume @xmath188 ) and @xmath77 the corresponding a - s curve .
let @xmath24 be a natural number and @xmath223 .
any element @xmath224 can be written as @xmath225 with @xmath226 iff @xmath227 ( see previous subsection ) . applying this to @xmath68 we see that @xmath77 has an affine point with abscissa @xmath193 iff @xmath228 , in which case it has exactly @xmath0 such points ( namely @xmath229 ) . using the orthogonality relation for the additive characters of
@xmath10 this can be restated as follows : the number of @xmath230-points on @xmath77 with abscissa @xmath231 equals @xmath232 and so the total number of points on @xmath77 ( including the infinite point ) is [ numpts]n_r(c_f)=q^r+1+__q^r_(_q^r / pf ( ) ) .
define @xmath233 it follows from and that the l - function of @xmath77 can be written as a product @xmath234 each function @xmath18 turns out to be a polynomial of degree @xmath12 with constant term @xmath235 , see @xcite , @xcite .
we can write @xmath236 where the @xmath237 are all the normalised zeroes of @xmath118 and they satisfy @xmath238 . to understand the behaviour of the zeroes of @xmath118 it is enough to study the zeroes of the individual @xmath18 and the relationship between the l - functions corresponding to different characters .
these functions are called primitive l - functions or the primitive factors of @xmath131 .
note that for two conjugate characters @xmath239 the two l - functions @xmath240 are conjugate and so are their zeroes . thus for @xmath38 the primitive factors come in conjugate pairs ( for @xmath108 there is only one nontrivial character and @xmath131 itself is a primitive l - function ) .
let @xmath241 $ ] be polynomials of degree @xmath2 s.t .
@xmath242 for some @xmath196 and natural @xmath6 .
then for all @xmath243 we have @xmath244 , since @xmath245 therefore @xmath246 .
this agrees with the fact that @xmath247 are isomorphic and so @xmath248 .
now assume that @xmath207 for some @xmath249 . from we
see that @xmath250 and denoting by @xmath251 the normalised roots of @xmath252 respectively we see that @xmath253 . if @xmath254 then @xmath246 and as we have seen in the previous subsection the curves @xmath247 are isomorphic in this case
. denote @xmath255 we refer to @xmath256 as the artin - schreier ( a - s in short ) family of curves with parameter @xmath2 over @xmath113 and to @xmath257 the as the a - s family of l - functions with parameter @xmath2 over @xmath113 .
note that the latter does not depend on the choice of @xmath33 because for @xmath258 we have @xmath259 ( this follows from ) , so replacing @xmath260 with @xmath261 permutes the family of l - functions .
fix some element @xmath262 with @xmath263 .
for @xmath264 denote @xmath265 . by the observations in the end of the previous subsection we have
@xmath266 so the statistics of zeroes of the l - functions of @xmath267 is essentially the same as that of @xmath62 .
the size of the family @xmath14 and each @xmath268 is @xmath269 .
denote by @xmath270 the set of all degree @xmath2 polynomials in @xmath86 $ ] .
define the map @xmath271 by @xmath272 this map is precisely @xmath273 to one .
it follows from the observations in the end of the previous subsection that for any @xmath24 and any @xmath243 we have @xmath274 and therefore @xmath275 . we conclude that studying the statistics of zeroes of l - functions of @xmath276 reduces to studying it for @xmath277 , which in turn reduces to studying it for the family @xmath14 .
henceforth we only consider the family @xmath14 .
in recent decades it has been suggested that the zeroes of l - functions ( of all sorts ) behave as the spectra of matrices drawn randomly ( with some natural measure ) from some classical ensemble of matrices .
we will illustrate this approach on our example of the a - s l - functions .
for the rest of this section assume @xmath38 .
denote by @xmath278 the group of @xmath279 unitary matrices .
this is a compact lie group and so it has a haar measure .
we may draw a random matrix @xmath280 uniformly w.r.t . the haar measure and ask about the statistics of its spectrum .
the eigenvalues of @xmath281 lie on the unit circle , just like the zeroes of @xmath18 .
we take @xmath282 ( recall @xmath283 ) and we model the set of roots of @xmath18 for fixed @xmath260 and `` random '' @xmath15 ( i.e. @xmath15 may vary through some large family , e.g. @xmath14 with either @xmath142 or @xmath39 , or both ) by the spectrum of a random matrix from @xmath284 .
this model is suggested by the result due to n. katz and p. sarnak for the case of fixed @xmath2 and @xmath142 , stating that the sets of normalised zeroes of the l - functions in the a - s family become equidistributed in the space of sets of eigenvalues of matrices in @xmath284 with the measure induced from the haar measure on @xmath285 .
see theorem 3.9.2 in @xcite . to model the @xmath11 sets of zeroes of @xmath286 jointly for @xmath38
we may take @xmath287 independent random matrices from @xmath284 and their conjugates ( see section [ lfunc ] ) .
now we formulate some conjectures on the statistics of the zeroes of @xmath288 , of which our main theorems are special cases . in all that follows
assume @xmath188 .
we are interested in the case where @xmath39 and @xmath55 may be fixed or vary as we please .
first we consider the linear statistics - the number of zeroes in short intervals and the related quantity of the average of powers of the zeroes . since multiplying a matrix @xmath280 by a scalar matrix @xmath289 rotates the eigenvalues by an angle of @xmath50 it is obvious that the average number of eigenvalues contained in an arc of length @xmath42 is @xmath290 ( as @xmath281 is drawn uniformly at random w.r.t .
the haar measure )
. we will be interested in @xmath291 as @xmath39 , the so - called local regime ( it is easier to obtain results for larger arcs ) .
using our model we formulate [ conj1 ] take any @xmath33 .
let @xmath292 be a constant .
for every natural number @xmath2 let @xmath293 be any arc on the unit circle of length @xmath294 .
then the average number of zeroes of @xmath18 contained in @xmath293 as @xmath15 is chosen uniformly at random from @xmath14 is @xmath295 as @xmath39 . instead of just looking at arcs we may take a smooth window function to count the zeroes .
let @xmath296 .
the function @xmath297 is well - defined and periodic with period @xmath41 .
it can be viewed as a function on the unit circle .
we say that @xmath298 is the periodic window function associated with @xmath299 with scaling parameter @xmath2 . for every @xmath300 with @xmath301 and real number @xmath50 the value of @xmath302 is well - defined .
denote @xmath303 , where @xmath17 are the normalised zeroes of @xmath18 .
conjecture [ conj1 ] is equivalent to the following statement : the average of @xmath304 as @xmath15 is chosen uniformly at random from @xmath14 is @xmath305 as @xmath39 for any @xmath46 and @xmath306 ( because the indicator of an interval can be approximated by a window function in @xmath44 and any window function can be approximated by a superposition of interval indicators ) .
now we consider the quantity @xmath307 , where as usual @xmath308 are the normalised zeroes of @xmath18 .
the uniform distribution of the l - zeroes on the unit circle suggests the following [ conj3 ] take any @xmath33 . for every @xmath37
the average of @xmath23 where @xmath15 is drawn uniformly at random from @xmath14 is @xmath309 as @xmath39 and @xmath310 ( @xmath55 may vary with @xmath2 as we please ) .
it can be shown by a standard argument that conjecture [ conj3 ] combined with theorem [ thm1 ] implies conjecture [ conj1 ] .
see the proof of theorem [ thm2 ] in section [ pthm2 ] for this kind of argument .
we remark that conjecture [ conj3 ] would follow from a function field analogue of a conjecture of h. montgomery about the distribution of primes in arithmetic progressions ( see @xcite ) . at this point a simpler model for the zeroes of l - functions in the a - s family would be just @xmath12 independent random points on the unit circle ( with uniform distribution ) , which is also consistent with conjectures [ conj1],[conj3 ] and theorems [ thm1],[thm2 ] .
however in section [ nonlinear ] we study nonlinear statistics of the zeroes which show agreement with the random unitary matrix model and disagreement with the independent random points model .
finally we note that for @xmath108 we need a different model , namely a random symplectic matrix model .
see section [ p2 ] for a description of this model and some partial results .
we keep the notation of the previous section .
let @xmath311 be as in section [ intro ] , @xmath33 , @xmath2 a natural number satisfying @xmath188 .
first we need a lemma [ lem1 ] let @xmath33 be a character , @xmath78 $ ] of degree @xmath2 .
let @xmath312 be the normalised zeroes of @xmath18 and @xmath313 .
then @xmath314 this is a well known fact that follows directly from and .
see @xcite for details . for the rest of this section fix @xmath33 . to prove theorem 1 we need to estimate the average of the sum @xmath315 as @xmath15 varies uniformly through @xmath14 . recall that @xmath14 consists of the degree @xmath2 polynomials @xmath316 with @xmath317 and @xmath201 for all @xmath318 .
we begin with a simple observation that establishes a weak form of theorem 1 , namely with @xmath319 .
[ lem0 ] assume @xmath319 .
let @xmath243 be an element . then [ elem0](_q^r / pf())_f= if @xmath320 the assertion is clear since @xmath321 and so + @xmath322 for all @xmath62 .
if @xmath99 and @xmath323 then for all @xmath62 we have @xmath324 , so @xmath325 and @xmath322 .
now assume that @xmath326 and if @xmath99 then @xmath327 .
this means that the minimal polynomial @xmath328 of @xmath329 over @xmath113 satisfies @xmath330 .
denote @xmath331|\deg f = d , f(0)=0\}$ ] .
recall the definition of the map @xmath332 in section [ asfam ] .
since the @xmath333 is precisely @xmath334 to one and preserves @xmath335 so we may replace @xmath14 by @xmath336 in
. let @xmath337 $ ] be the minimal polynomial of @xmath329 over @xmath3 .
since @xmath338 the map @xmath339/h\cong { \f_{q^r}}$ ] defined by @xmath340 is exactly @xmath341 to one and since @xmath193 is invertible modulo @xmath328 ( as @xmath326 ) so is the map @xmath342/h$ ] defined by @xmath343 .
thus each value of @xmath344/h$ ] is obtained equally many times as @xmath15 ranges through @xmath336 . since @xmath330 the value of @xmath345 is uniformly distributed among the @xmath0-th roots of unity as @xmath346 ranges through @xmath347 , which proves the assertion of the lemma .
the following corollary establishes theorem [ thm1 ] for @xmath319 .
[ cor1]assume @xmath319
. then @xmath348 by lemma [ lem1 ] we have [ erld]m_d^r = t^r_f,_f =- q^-r/2__q^r(_q^r / pf())_f.now using lemma [ lem0 ] we see that the rhs of equals @xmath349 if @xmath96 ( only @xmath320 contributes 1 to the sum ) or @xmath350 if @xmath99 ( each @xmath243 contributes 1 to the sum ) . to go
further we need some lemmata .
[ lem2 ] let @xmath243 have monic minimal polynomial @xmath351 $ ] of degree @xmath24 .
assume that for some @xmath352 with @xmath353 we have @xmath354 .
then for the minimal such @xmath6 we have @xmath355 and for all @xmath356 we have @xmath357 denote by @xmath358 the conjugates of @xmath329 over @xmath3 .
denote by @xmath359 the degree @xmath360 elementary symmetric polynomial in @xmath24 variables .
we have @xmath361 for all @xmath362 .
denote @xmath363 .
newton s identity ( see @xcite ) states that for all @xmath364 we have [ newton]m_m=(-1)^m+1s_m-_i=1^m-1 s_i_m - i.we show by induction that for @xmath365 we have @xmath366 . if @xmath367 then the case @xmath368 is clear as by assumption @xmath369 .
assume that @xmath370 and that @xmath371 holds for all @xmath372 .
by we have @xmath373 ( the other terms in the identity are zero by the induction hypothesis ) . now
if @xmath360 is not divisible by @xmath0 the assumption on @xmath6 implies that @xmath374 and if @xmath360 is divisible by @xmath0 we still have @xmath375 ( as we are in characteristic @xmath0 ) .
this completes the induction .
now again we see from that @xmath376 as required .
[ mainlem ] let @xmath243 have monic minimal polynomial @xmath351 $ ] of degree @xmath24 .
assume that either @xmath319 or @xmath310 and for some @xmath377 with @xmath353 we have @xmath354 . then @xmath378 .
if @xmath379 and there is no such @xmath6 then if @xmath380 we have @xmath378 and if @xmath381 we have @xmath382 .
the case @xmath383 follows from lemma [ lem0 ] , so we assume @xmath384 .
first assume there exists @xmath377 s.t .
@xmath353 and @xmath354 .
let @xmath6 be minimal with this property .
by the previous lemma @xmath385
. therefore there exists @xmath386 s.t .
@xmath387 now the set @xmath14 can be partitioned into subsets of the form @xmath388 note however that @xmath389 because @xmath390 is a primitive @xmath0-th root of unity .
since @xmath14 is partitioned into sets of the form @xmath391 we get the first claim of the lemma .
now assume that for all @xmath392 we have @xmath393 .
take some @xmath243 of degree @xmath24 . by the previous lemma we get that @xmath394 for @xmath395 and so @xmath396 for all @xmath386 and by the second part of the lemma we have @xmath397 and so @xmath398 for every @xmath386 .
thus for every @xmath399 we have @xmath400 . if @xmath380 then @xmath401 .
assume @xmath381 .
the leading coefficient of @xmath62 is distributed uniformly in @xmath402 and so is @xmath403 .
we have @xmath404 because as @xmath30 ranges through @xmath405 each nonzero value of @xmath406 occurs @xmath407 and zero occurs @xmath408 times , so @xmath409 .
this concludes the proof of the lemma . for @xmath410 denote by @xmath411 the number of monic irreducible polynomials @xmath412 s.t .
@xmath393 for all @xmath413 with @xmath353 .
denote by @xmath414 the number of such polynomials with @xmath380 ( if @xmath415 we define @xmath416 ) .
[ irr ] @xmath417 by lemma [ lem1 ] we have [ irr1]m^r_d = q^-r/2__q^r(_q^r / pf())_f.every @xmath243 has degree @xmath418 over @xmath3
. first let @xmath418 be such that @xmath419 .
for @xmath329 of degree @xmath222 ( over @xmath3 ) and @xmath62 we have @xmath420 and so @xmath421 by lemma [ mainlem ] applied to @xmath422 instead of @xmath423 we see that the contribution of all @xmath326 of degree @xmath222 to the rhs of is @xmath424 since each irreducible polynomial @xmath425 of degree @xmath426 has @xmath222 roots , each contributing 1 to the sum if @xmath427 and @xmath428 otherwise ( elements of degree @xmath429 contribute nothing by lemma [ lem0 ] ) . if @xmath96 we obtain the assertion of the lemma .
it remains to consider the contribution of @xmath320 and @xmath323 in case that @xmath99 .
since for @xmath320 and @xmath323 we have @xmath430 , this contribution is obviously @xmath431 .
theorem 1 follows at once from proposition [ irr ] .
indeed the total number of polynomials of the form @xmath432 with @xmath433 and @xmath434 is at most @xmath435 and therefore @xmath436 note that the assertion of the theorem is interesting only for @xmath437 , because by the definition of @xmath438 ( see lemma [ lem1 ] ) we have @xmath439 for @xmath62 and so @xmath440 .
now we deduce theorem [ thm2 ] from theorem [ thm1 ] . for this subsection
we assume @xmath38 .
let @xmath46 be a fixed window function . from a window function @xmath441 we may construct a periodic window function with parameter @xmath2 ( natural number ) as follows : @xmath442 we also define @xmath443
. the function @xmath444 has period @xmath41 and as @xmath39 it becomes `` localised '' at points of the form @xmath445 .
the fourier transform of @xmath299 is given by @xmath446 and the @xmath24-th fourier coefficient of @xmath444 is @xmath447 a simple calculation shows that [ fourlem]_d,(r)=. for @xmath62 denote @xmath448 where @xmath52 are the normalised zeroes of @xmath18 ( @xmath53 are real numbers well defined modulo @xmath41 ) . then @xmath449 since @xmath46 the function @xmath444 is smooth and so for @xmath301 we have @xmath450 applying this to @xmath451 , noting that @xmath452 , using and summing over @xmath360 we obtain the assertion of the lemma .
@xmath453 just average the previous lemma over @xmath62 and note that @xmath454 by proposition [ irr ] .
now we are ready to prove theorem [ thm2 ] .
assume that @xmath455 is supported in @xmath456 .
there exists @xmath37 s.t .
@xmath457 for all @xmath458 . using the last corollary and theorem [ thm1 ]
we obtain @xmath459 as @xmath39 ( note that we used the fact that @xmath460 since @xmath46 is fixed ) .
it is now enough to notice that @xmath461
. * remark . * as can be seen from the above proof , the @xmath56 term in theorem [ thm2 ] can be replaced with @xmath40 . if @xmath108 the assertion of theorem [ thm2 ] does not hold and has to be modified .
the random matrix model needs to be modified as well .
indeed there is only one character @xmath33 and the l - function @xmath462 for @xmath62 is already primitive . since it has real coefficients it is obvious that the set of normalised zeroes consists of conjugate pairs and can not be `` random '' in the space of eigenvalue sets of unitary matrices . instead we should consider a random unitary symplectic matrix @xmath135 ( thus we denote the group of @xmath463 unitary symplectic matrices ) .
in fact for @xmath108 the curves in the a - s family are hyperelliptic and the usual model for families of hyperelliptic curves is the random symplectic matrix model , see for example the work mentioned in section [ hyp ] .
for the unitary symplectic group the following holds : @xmath464 see @xcite . from this and theorem [ thm1 ] we conclude that for @xmath319 we have @xmath465 from this one can derive using the method of section [ pthm2 ] the following let @xmath46 be a window function , @xmath466 for @xmath467 denote @xmath468 , where @xmath52 are the normalised zeroes of @xmath131 . similarly for a matrix @xmath135 with eigenvalues
@xmath52 denote @xmath469 .
assume that @xmath455 is supported on @xmath470 .
then @xmath471 as @xmath472 .
theorems [ thm1 ] and [ thm2 ] suggest that the zeroes of a random l - function from the a - s family are ( at least on average ) rather uniformly distributed on the unit circle , but this seems like a weak confirmation of the random unitary matrix model . a simpler model would be @xmath12 independent random points on the circle ( with uniform distribution ) . in this section
we study more delicate statistics of the zeroes and show agreement with the random unitary matrix model and disagreement with the independent random points model .
we preserve the notation of the previous sections .
let @xmath473 be a two - variable window function , @xmath474 a natural number and @xmath475 the periodic window function associated with @xmath57 by @xmath476 let @xmath477 be @xmath478 points on the unit circle , @xmath52 .
finally let @xmath50 be some fixed real number .
we consider the 2-level density function ( at @xmath50 ) : @xmath479 for a matrix @xmath480 with eigenvalues @xmath477 we denote @xmath481 and for an a - s l - function @xmath18 with normalised zeroes @xmath61 we denote @xmath482 it is easy to see that if @xmath477 are selected uniformly and independently on the unit circle , then the average of @xmath483 tends to @xmath484 as @xmath485 ( for any @xmath50 ) . on the other hand we have the following result ( see ( * ? ? ? * ad.2 ) ) : [ upair]s^2_(u)_u(n)v(t , u)1-^2tu.as @xmath486 ( the average is taken w.r.t . the haar measure ) , for any @xmath50 . theorem
[ thm2p ] , which we prove in the present section , provides evidence for the random unitary matrix model . for the rest of section [ nonlinear ]
we assume that @xmath38 .
just as we used the quantities @xmath487 to study the linear statistics of the zeroes of @xmath18 we introduce the quantities @xmath488 ( again @xmath489 does not depend on the choice of @xmath33 ) .
we also define @xmath490 for any integer @xmath24 ( possibly negative ) by the same expression @xmath491 ( where @xmath492 are the normalised zeroes of @xmath18 ) and extend the definition of @xmath493 to all integers @xmath494 .
note that @xmath495 .
good estimates for these quantities provide good estimates for the quadratic statistics of the l - zeroes , such as the square of the number of points in short intervals and the 2-level density .
[ trs ] for @xmath496 we have @xmath497 @xmath498 this follows immediately from lemma [ lem0 ] .
[ rsd ] assume @xmath496 , @xmath499 .
let @xmath500 be nonzero elements with monic minimal polynomials @xmath501 over @xmath113 respectively .
for any natural @xmath145 denote @xmath502 we have [ rsd1](_q^r / pf()-_q^s / pf())_f=[rsd2](_q^r / pf()+_q^s / pf())_f= we prove , being similar . if @xmath503 then @xmath504 are conjugate over @xmath113 and so are @xmath505 , so @xmath506 , where @xmath507 .
we have @xmath508 and so @xmath509 if @xmath510 then @xmath511 is divisible by @xmath0 and so @xmath512 and @xmath513 for all @xmath62 and of course it also holds on average . if @xmath514 then denoting @xmath515 we get from lemma [ lem0 ] applied to the nontrivial character @xmath516 instead of @xmath260 that the lhs of [ rsd1 ] equals 0 . for @xmath517
we have @xmath430 because if @xmath99 we have @xmath518 and if @xmath519 then @xmath320 and @xmath520 for all @xmath62 . similarly if @xmath521 then @xmath522 . we see that if @xmath523 then @xmath513 for all @xmath62 and the same holds for the average .
now assume that @xmath521 but @xmath524 .
then for all @xmath62 we have @xmath525 and so @xmath526 by lemma [ lem0 ] , since @xmath527 and @xmath524 .
the case @xmath528 is treated similarly .
finally assume that @xmath529 and @xmath530 , i.e. @xmath531 , @xmath532 , where @xmath533 ( we have @xmath534 ) . as in the proof of lemma [ lem0 ] we may average over @xmath535|\deg f = d , f(0)=0\}\ ] ] instead of @xmath14 ( this does not change the average ) .
now since @xmath536 , the map @xmath537/gh$ ] defined by @xmath340 is exactly @xmath538 to one and since @xmath193 is invertible modulo @xmath539 ( as @xmath531 ) so is the map @xmath540/gh$ ] defined by @xmath541 .
however by the chinese remainder theorem we have @xmath80/gh\cong{\f_q}[x]/g\times{\f_q}[x]/g\cong{\f_{q^u}}\times{\f_{q^v}}$ ] ( direct product of rings ) with the isomorphism given by @xmath542 . we conclude that as @xmath15 ranges over @xmath336 each pair @xmath543 is obtained equally many times .
however since @xmath544 the map @xmath545 defined by @xmath546 also assumes every value equally many times and the same goes for @xmath547 on @xmath548 .
we conclude that as @xmath15 ranges over @xmath336 each @xmath0-th root of unity occurs equally many times as @xmath549 and since averaging over @xmath14 is equivalent to averaging over @xmath336 we obtain
. now denote by @xmath550 the number of monic irreducible polynomials in @xmath80 $ ] with degree @xmath145 .
[ mdrs1]assume @xmath496 and @xmath499 .
then @xmath551 @xmath552 we prove , begin similar .
by lemma [ trs ] we have [ mrs]m^r ,- s_d = q^-(r+s)/2__q^r(_q^r / pf()-_q^s / pf())_f.now we can use lemma [ rsd ] to evaluate this expression .
denote by @xmath501 the monic minimal polynomials ( over @xmath113 ) of @xmath553 respectively .
first we count the contribution of those @xmath504 for which @xmath503 ( i.e. they are conjugate ) , @xmath554 and @xmath555 ( and consequently @xmath556 ) , so @xmath524 and @xmath557 .
each polynomial @xmath149 has exactly @xmath558 roots and the number of pairs @xmath504 with minimal polynomial @xmath149 is @xmath559 .
the contribution of all such @xmath149 to the sum in is @xmath560 it remains to evaluate the contribution of the pairs @xmath523 ( see the notation in the previous lemma ) .
if @xmath561 then every pair @xmath562 contributes 1 to the sum ( by lemma [ rsd ] ) and the number of such pairs is @xmath563 . if @xmath99 but @xmath564 then the pairs @xmath565 ( and only them ) contribute 1 , so we get a total contribution of @xmath566 .
the case @xmath567 is treated similarly .
if @xmath568 then only @xmath569 adds 1 to the sum . in any case
we get the value stated in the proposition . to prove we proceed similarly using .
[ thm1p ] assume @xmath570 and @xmath499 .
then @xmath571 where @xmath572 we also have @xmath573 by the proposition [ mdrs1 ] we have @xmath574 it is well known that @xmath575 ( see @xcite ) . if @xmath576 then @xmath577 ( note that except for @xmath578 and possibly @xmath579 the other terms are negligible ) , which implies the assertion of the theorem for the case @xmath576 . if @xmath580 then @xmath581 and @xmath582 which implies the assertion of the theorem for @xmath583 .
the second part of the theorem follows similarly from the second part of proposition [ mdrs1 ] ( note that this time only @xmath145 s.t .
@xmath584 contribute to the sum ) .
the last result should be compared with the following ( see ( * ? ? ?
2 ) ) : @xmath585 @xmath586 ( for @xmath496 ) . for simplicity
we will prove theorem [ thm2p ] for @xmath587 , the proof of the general case proceeds with only slight modifications .
we will write @xmath588 instead of @xmath589 for the 2-level density function .
@xmath590 we have [ breaks]s^2(f,)=_j , k=1^d-1v_d-1(_j,_k)-_j=1^d-1v_d-1(_j,_j).the fourier series coefficients of the bi - periodic function @xmath591 are given by @xmath592 the derivation is by a standard calculation , similar to that of [ fourlem ] . since @xmath591 is smooth the following holds for @xmath593 : @xmath594 now if @xmath61 are the normalised zeroes of @xmath18 and @xmath52 then @xmath595 and @xmath596 combining , and we obtain the statement of the lemma .
[ corp]@xmath597 just average the previous lemma over @xmath14 .
we need one more lemma : [ flem ] let @xmath58 be a window function , @xmath598 its fourier transform
. denote @xmath599 .
we have @xmath600 define @xmath601 .
we have @xmath602 .
it is easy to see from the definitions and fubini s theorem that the fourier transform of @xmath603 is @xmath604 .
the fourier transform of the function @xmath605 is @xmath606 , so we have by plancherel s theorem @xmath607 now we are ready to prove theorem [ thm2p ] .
assume @xmath38 .
let @xmath58 be s.t .
@xmath455 is supported on @xmath66 . by corollary [ corp ]
we have [ msum]s^2(f,)_f=_r , s , m^r , s_d - m^r+s_d .
first we bound the contribution to the sum of @xmath494 s.t .
@xmath608 . by theorem [ thm1 ]
we have @xmath609 ( note that @xmath610 ) , since @xmath611 decreases geometrically in @xmath24 for @xmath612 .
similarly , by theorem [ thm1p ] we also have @xmath613 as @xmath614 for @xmath583 decreases geometrically in @xmath615 .
the overall contribution to the rhs of is @xmath40 ( note that @xmath598 is bounded ) .
it remains to estimate @xmath616 ( note that @xmath617 ) . invoking theorem [ thm1p ] and noting that the error terms accumulate to at most @xmath618 ( the error term for @xmath619 in theorem [ thm1p ] decreases geometrically in @xmath24 ) we see that @xmath620 as @xmath472 by the definition of the riemann integral ( we used the fact that @xmath621 is supported on @xmath622 ) . now using lemma [ flem ]
we obtain the assertion of theorem [ thm2p ] .
in the present section we will see that the family of l - functions @xmath288 is actually a special case of a family of dirichlet l - functions corresponding to multiplicative characters of @xmath86 $ ] and generalise our main results to such families .
we briefly recall the basic properties of dirichlet characters and l - functions over @xmath80 $ ] . for details see @xcite .
let @xmath623 $ ] be a monic polynomial of degree @xmath145 and let @xmath624/q\rb^\times\to\c^\times$ ] be a character of the multiplicative group of residues modulo @xmath625 .
we may extend @xmath155 to @xmath80 $ ] by @xmath626 the map @xmath627\to\c$ ] thus defined is called a dirichlet character .
the character @xmath155 is called primitive if there is no proper divisor @xmath628 of @xmath625 s.t .
@xmath629 for @xmath149 prime to @xmath625 only depends on @xmath630 .
it is called trivial if it takes the value 1 on all polynomials prime to @xmath625 .
it is called even if it takes the value 1 on constants and odd otherwise .
we denote @xmath631 denote by @xmath632 the set of monic polynomials in @xmath80 $ ] and by @xmath633 the set of monic irreducible polynomials in @xmath80 $ ] .
the l - function corresponding to the character @xmath155 is defined as follows : [ dirl]l_(z)=_g(g)z^g=_h(1-(h)z)^-1.it turns out that for a primitive character @xmath155 modulo @xmath625 the function @xmath634 is a polynomial of degree @xmath635 if @xmath155 is even and @xmath12 if @xmath155 is odd .
further it factors as follows : @xmath636 with @xmath21 .
the @xmath17 are called the normalised zeroes of @xmath634 .
let @xmath2 be a natural number and @xmath33 a nontrivial additive character of @xmath10 .
let @xmath637 be a polynomial .
we will attach a dirichlet character @xmath155 modulo @xmath638 to @xmath639 .
let @xmath640 $ ] be a polynomial .
if @xmath641 we define @xmath642 . otherwise we may write [ gpr]g(x)=_i=1^k(1-_ix),where @xmath643 are the roots of @xmath149 in the algebraic closure of @xmath113 .
first we observe that for every natural @xmath644 the quantity @xmath645 lies in @xmath113 and depends only on the coefficients @xmath646 ( in other words it depends only on @xmath647 ) . as in section [ pthm1 ] we denote by @xmath648 the order @xmath644 elementary symmetric function in the variables @xmath649 . by we have @xmath650 .
using newton s identity [ newton ] recursively we can show that @xmath651 can be expressed as a polynomial in @xmath652 with integer coefficients .
therefore @xmath645 can be expressed as a polynomial in @xmath653 with integer coefficients and so it must lie in @xmath113 and depends only on + @xmath653 ( and therefore only on @xmath647 ) .
it follows from the above that the quantity @xmath654 is in @xmath113 and depends only on @xmath655 .
now we define [ dir](g)=_q / p_i=1^k f(_i).by what we have seen @xmath629 is well defined and depends only on @xmath647 .
further it is multiplicative , because the set of zeroes ( counting multiplicity ) of a product of two polynomials is the union of their sets of zeroes .
therefore @xmath155 is a dirichlet character modulo @xmath638 . obviously for constant @xmath149
we have @xmath656 , so @xmath155 is even . by [ dir ]
we also have that @xmath657 is trivial , so @xmath155 is an order @xmath0 character modulo @xmath638 .
we fix @xmath33 and denote by @xmath658 the character constructed above for a given @xmath62 . for any @xmath62 the character @xmath658 is primitive .
the characters @xmath659 are all distinct and any primitive character @xmath155 modulo @xmath638 with @xmath657 trivial is of the form @xmath660 for some @xmath62 .
first we show that for @xmath62 the character @xmath660 is primitive . for this
it is enough to show that for some @xmath262 we have @xmath661 .
write @xmath662 .
it is easy to see from the definition that @xmath663 . since @xmath188 and @xmath317 there exists @xmath262 s.t .
@xmath664 .
now let @xmath665 s.t .
@xmath666 . we will show that @xmath667 . by the definition of @xmath14 there exists some @xmath668 s.t .
@xmath669 . let @xmath670 $ ] be some polynomial prime to @xmath193 .
write it as @xmath671 with @xmath672 for some @xmath24 .
we have @xmath673 it is enough to show that for some @xmath149 we have @xmath674 .
taking @xmath675 we have ( as above ) @xmath676 where @xmath677 and using the fact that @xmath678 and @xmath679 we can pick @xmath680 so that @xmath681 .
we have seen that the correspondence @xmath682 is one - to - one from @xmath14 to the set of order @xmath0 primitive characters modulo @xmath638 .
to show that it is onto it is enough to show that these sets are identical in size .
recall that @xmath683 . for
a finite abelian group @xmath684 denote by @xmath685 its dual and by @xmath686 $ ] the @xmath145-torsion of the group @xmath684 .
the groups @xmath687 are always isomorphic .
take @xmath688/x^{d+1})^\times$ ] .
first we compute @xmath689=\#g[p]$ ] , which is the number of all order @xmath0 characters modulo @xmath638 .
each element of @xmath684 is represented uniquely by a polynomial @xmath690 .
we have @xmath691 and @xmath692 iff @xmath693 and @xmath694 . therefore @xmath689=\#g[p]=q^{d-\floor{d / p}+1}$ ] and this is also the number of order @xmath0 characters modulo @xmath638 . by the same reasoning
the number of order @xmath0 characters modulo @xmath695 is @xmath696 ( since @xmath188 ) and so the number of primitive characters modulo @xmath638 equals @xmath697 . [ lemm ] for the character @xmath155 defined above we have @xmath698 .
in particular the normalised nontrivial zeroes of @xmath699 coincide with the normalised zeroes of @xmath18 . since both @xmath634 and @xmath700 have constant coefficient 1 it it enough to show that @xmath701 let @xmath24 be a natural number , @xmath243 an element with minimal polynomial @xmath328 ( over @xmath113 ) of degree @xmath418 .
denote by @xmath702 the roots of @xmath328 in @xmath230 .
we have @xmath703 and so for @xmath326 we have @xmath704 ( recall that @xmath705 where @xmath706 and it has roots @xmath707 ) .
for @xmath320 we have @xmath708 . using the above and [ dirl ] we obtain @xmath709 ( we used the fact that the operation @xmath710 permutes @xmath711 , that @xmath712 and that @xmath713 for @xmath62 ) .
we see that the family of l - functions @xmath714 coincides with the family of l - functions @xmath715 where @xmath155 ranges over the order @xmath0 primitive characters modulo @xmath638 .
next we generalise theorems [ thm1 ] and [ thm2 ] to more general families of dirichlet characters , obtaining a new ( but essentially equivalent ) proof of our results for a - s l - functions . for any finite abelian group @xmath25
we denote by @xmath717 its dual group .
let @xmath623 $ ] be a monic polynomial of degree @xmath145 .
denote by @xmath684 the group of characters modulo @xmath625 , which we will also identify with @xmath716/q)^{\times*}$ ] , i.e. we view the elements of @xmath684 also as characters of @xmath716/q)^\times$ ] .
let @xmath718 be a subgroup of @xmath684 .
for any @xmath719 we denote by @xmath720 the subgroup of @xmath718 consisting of the characters which have period @xmath721 ( or a divisor of @xmath721 ) .
denote by @xmath722 the set of primitive characters in @xmath718 .
denote by @xmath723 the subgroup of @xmath716/q)^\times$ ] consisting of elements @xmath149 s.t .
@xmath656 for all @xmath724 .
it is the subgroup of @xmath716/q)^\times$ ] orthogonal to @xmath725/q)^{\times*}$ ] and its order is @xmath726 ( this relation holds for any finite abelian group ) .
the following orthogonality relation holds for @xmath670 $ ] : [ ort ] _
h(g)=we denote by @xmath727 the set of monic irreducible polynomials @xmath328 s.t . @xmath728 .
let @xmath155 be an even primitive character modulo @xmath625 . recall that its l - function can be factored as [ dirlzeros]l_(z)=(1-z)_i=1^m-2(1-q^1/2_i),with @xmath21 ( @xmath17 are the normalised zeroes of the l - function ) .
[ 1]let @xmath24 be a natural number .
@xmath729 by [ dirl ] and @xmath730 we have @xmath731 comparing coefficients at @xmath732 we obtain the statement of the lemma .
we denote @xmath733 . for a group of dirichlet characters
@xmath734 modulo @xmath625 and a natural number @xmath222 denote @xmath735 . for a nonzero polynomial
@xmath736 $ ] with factorisation @xmath737 into irreducibles we denote @xmath738 ( this is the mbius function on @xmath80 $ ] ) . for
an abelian group @xmath25 and natural number @xmath6 denote by @xmath739 the subgroup of @xmath6-th powers in @xmath25 .
finally denote by @xmath740 the set of primitive characters in @xmath718 and @xmath741 .
the following proposition is a generalisation of proposition [ irr ] to an arbitrary family of dirichlet characters corresponding to a subgroup @xmath718 of characters modulo @xmath625 .
[ dirprop ] @xmath742 where @xmath743 denotes summation over monic divisors @xmath721 of @xmath625 .
first of all it follows from the inclusion - exclusion principle that for any map @xmath744 we have [ ie]_h^x()=_q|q(q / q)_h_qx().next , by lemma [ 1 ] and the orthogonality relation for any @xmath719 we have @xmath745 ( we used the fact that @xmath746 iff @xmath747 ) . combining this with we obtain the statement of the proposition .
the last proposition can be used to obtain bounds on @xmath748 .
for example assume that @xmath625 is irreducible .
then it follows from the proposition that for @xmath749 we have @xmath750 , since the total number of monic polynomials @xmath328 with @xmath751 is @xmath752 . for @xmath753
we have @xmath754 . for another example take @xmath755 and @xmath756/q)^\times\rb^*[p]$ ] ( the group of order @xmath0 characters modulo @xmath625 ) .
we have seen in section [ diras ] that @xmath757 , so @xmath758 .
proposition [ irr ] now follows from proposition [ dirprop ] .
indeed the only divisors @xmath719 s.t .
@xmath759 are @xmath760 and @xmath761 consists of the ( invertible ) @xmath0-th powers modulo @xmath721 , which are represented by polynomials of the form @xmath762 where @xmath763 . combining this with the fact that @xmath764 if @xmath353 and @xmath765 is trivial if @xmath766 the formula in proposition [ dirprop ] translates into the one in proposition [ irr ] ( note that for @xmath527 we have @xmath767 , since
the classes in @xmath723 are represented by @xmath0-th powers ) .
our main theorems follow from the latter proposition .
throughout this section @xmath38 . let @xmath2 be odd and as usual @xmath188
. denote by @xmath69 the subset of odd polynomials in @xmath14 , i.e. polynomials @xmath62 satisfying @xmath768 , in other words only odd powers of @xmath193 appear in @xmath15 .
we call @xmath69 as well as the corresponding family of curves and l - functions the odd a - s family . as with @xmath14
the family @xmath769 does not depend on the choice of @xmath33 . in the present section
we formulate conjectures for @xmath69 analogous to our main results for @xmath14 based on a random symplectic matrix model
. however we will only be able to prove a very weak result in this direction .
for @xmath770 we have @xmath771 $ ] .
by is would suffice to show that @xmath772 for every natural @xmath24 . since @xmath15 is an odd polynomial we have @xmath773 for @xmath243 and so @xmath774 and partitioning @xmath775 into pairs @xmath776 we obtain @xmath777 .
the latter fact suggests that we model the set of normalised l - zeroes of a random @xmath467 by the set of eigenvalues of a random matrix @xmath778 ( we denote thus the unitary symplectic group ) , taken uniformly w.r.t . the haar measure .
note that the characteristic polynomial of a unitary symplectic matrix has real coefficients .
this is the model usually used for the l - zeroes of a family of curves over a finite field , provided that the corresponding l - functions do not split into primitive l - functions with non - real coefficients , as it happens for the entire a - s family if @xmath38 . a more compelling reason for considering the random symplectic matrix model is an equidistribution result due to n. katz and p. sarnak for similar ( and more general ) families of l - functions with fixed @xmath2 and @xmath142 , see theorem 3.10.7 in @xcite . for the unitary symplectic group
the following holds : @xmath779 see @xcite .
we conjecture the following [ thm1o ] there exists a constant @xmath780 such that for any @xmath37 we have @xmath781 from this one can derive using the method of section [ pthm2 ] the following let @xmath46 be a window function , @xmath466 for @xmath467 denote @xmath468 , where @xmath52 are the normalised zeroes of @xmath18 . similarly for a matrix @xmath778 with eigenvalues @xmath17 denote @xmath782 .
then @xmath783 as @xmath472 . a possible approach to estimating @xmath784 is to reformulate the problem in terms of a family of dirichlet characters and use proposition [ dirprop ] .
we keep the notation of section [ secdir ] .
we take @xmath755 .
recall that to any polynomial @xmath8 $ ] with @xmath785 we can attach a character @xmath658 modulo @xmath638 defined by , .
it is primitive iff @xmath786 .
we also have @xmath787 if @xmath788 .
denote @xmath789/q)^\times\rb^*$ ] . as usual
we identify @xmath684 with the group of characters modulo @xmath625 .
denote @xmath790,\deg f\le r , f(-x)=-f(x)\}$ ] . by the above remarks this is a subgroup of @xmath684 .
we have @xmath791 . by lemma [ lemm ]
we have @xmath792 the only monic divisor @xmath719 s.t .
@xmath759 is @xmath793 which we denote by @xmath628 .
obviously for any natural @xmath418 we have @xmath794 .
also we have @xmath795 .
note also that @xmath796 since @xmath718 consists of order @xmath0 characters .
proposition [ dirprop ] now implies [ oddprop ] @xmath797 [ hort]the group @xmath723 consists of residues of the form @xmath798 where @xmath799},(g_i , x)=1 $ ] .
the same holds for @xmath800 and @xmath628 respectively .
take any @xmath467 and @xmath801},(g_1,x)=1 $ ] . since @xmath658 is an order @xmath0 character we have @xmath802 .
now take @xmath803},(g_2,x)=1 $ ] .
we may assume @xmath804 since @xmath658 is even .
the inverse roots of @xmath805 come in pairs @xmath806 and since @xmath15 is odd by [ dir ] we have @xmath807 .
it remains to note that the group of residues of the form @xmath808 has order @xmath809 and so does @xmath718 .
the same argument works for @xmath810 .
now the problem of estimating @xmath784 reduces to estimating the number of monic irreducible polynomials of degree @xmath24 ( and @xmath811 if @xmath24 is even ) which can be written in the form @xmath812 for some @xmath799},(g_i , x)=1 $ ] ( and the same for @xmath638 ) .
conjecture [ thm1o ] follows from heuristics about the number of irreducible polynomials of given degree falling in the subgroups @xmath813 modulo @xmath814 respectively . the following conjecture ,
if proven , would settle the case @xmath75 : [ niceconj ] let @xmath55 be a power of a prime @xmath38 , @xmath2 a natural number .
let @xmath815}$ ] be prime to @xmath193 , @xmath816 and @xmath75 .
assume that there exist @xmath817}$ ] s.t . @xmath812 .
then for all sufficiently large @xmath2 there in fact exist @xmath818}$ ] s.t .
@xmath819 . in particular if @xmath328 is irreducible then @xmath820 .
we will give some evidence for conjecture [ niceconj ] , namely we will show that in holds for @xmath821 with any constant @xmath822 . first we show how conjecture [ niceconj ] implies conjecture [ thm1o ] for @xmath75 .
first assume that @xmath24 is odd .
then by conjecture [ niceconj ] and lemma [ hort ] we have @xmath823 and so by proposition [ oddprop ] we have @xmath824 .
now assume that @xmath24 is even .
consider first @xmath825 . by conjecture [ niceconj ] and lemma [ hort ]
any irreducible polynomial of degree @xmath24 the residue of which modulo @xmath625 is orthogonal to @xmath718 is of the form @xmath826 with @xmath827}$ ] and @xmath828 .
the polynomial @xmath826 is irreducible iff @xmath149 is irreducible and any root of @xmath149 in @xmath829 is not a square in @xmath829 .
the number of such ( monic ) @xmath149 is easily seen to be @xmath830 , so @xmath831 .
we also see from conjecture [ niceconj ] that @xmath832 and @xmath833 , so by proposition [ oddprop ] we obtain @xmath834 .
we see that to establish theorem [ oddthm ] it suffices to prove conjecture [ niceconj ] in the case @xmath835 .
let @xmath55 be a power of a prime @xmath38 and @xmath319 natural numbers .
let @xmath815}$ ] be a polynomial prime to @xmath193 with @xmath816 .
suppose that @xmath328 can be written in the form @xmath836 .
since the polynomials of the form @xmath837 modulo @xmath695 ( prime to @xmath193 ) form a group we may also write @xmath838 ( for a different choice of @xmath839 ) .
write @xmath840 ( for sufficiently large @xmath360 we have @xmath841 ) .
we may assume @xmath842 , otherwise it is easy to see that for @xmath843 to be of the form @xmath844 modulo @xmath695 the polynomial @xmath328 itself must be even . now take any @xmath72 and assume that @xmath74 .
for sufficiently large @xmath2 we have @xmath845 . by the pigeonhole principle for some @xmath846 we have @xmath847 for @xmath848 , with @xmath849 having the same parity .
now denote by @xmath850 the infinite power series @xmath851 the first @xmath852 coefficients of @xmath850 coincide with those of @xmath839 and then the sequence of coefficients continues periodically with period @xmath853 .
the coefficients of @xmath854 coincide with those of @xmath843 up to the @xmath855-th coefficient , after which they continue periodically with period @xmath856 .
this is because each coefficient depends on at most @xmath857 consecutive coefficients of @xmath850 ( and the coefficients of @xmath328 ) .
when we multiply @xmath328 by @xmath858 the first @xmath2 coefficients are zero for odd powers and the same holds for the first @xmath855 coefficients of @xmath854 , after which it continues to hold by periodicity ( since the period @xmath856 is even ) .
thus the power series @xmath854 can be written in the form @xmath859 for some power series @xmath860 .
but @xmath850 is periodic and so must be @xmath861 .
we then have two rational functions @xmath862 the @xmath193-adic expansions of which are @xmath863 respectively and we must have @xmath864 . now by the unique factorisation property in @xmath865}$ ] we see that @xmath328 can be written in this form @xmath866 with @xmath867 $ ] .
in this section we consider the distribution of the number of points on the curve @xmath77 as @xmath15 varies uniformly through the family @xmath79 of all degree @xmath2 monic polynomials in @xmath80 $ ] and @xmath39 and prove theorems [ t1],[t2],[t3],[t4 ] . throughout this section
@xmath24 is a fixed natural number .
the number of @xmath230-rational points on @xmath868 equals the number of solutions to @xmath869 over @xmath81 plus one . from the hilbert 90 theorem or elementary linear algebra it follows that for @xmath224 the equation @xmath870 is solvable in @xmath230 iff @xmath871 , in which case it has exactly @xmath0 solutions , here by @xmath872 we denote the trace map from @xmath230 to @xmath10 .
thus for a given @xmath231 the equation @xmath869 is solvable iff @xmath873 and in this case it has exactly @xmath0 solutions ( see section [ geom ] ) .
we denote by @xmath874 the number of solutions to @xmath873 in @xmath230 .
it is enough to study the distribution of @xmath874 ( the number of points on the curve is then @xmath875 ) . from now on
we fix @xmath24 .
let @xmath337 $ ] be an irreducible polynomial or degree @xmath92 .
its splitting field is the subfield @xmath876 which is isomorphic to @xmath86/h$ ] .
if @xmath224 is a root of @xmath328 and @xmath78 $ ] then we denote @xmath877 ( it does not depend on the choice of the root @xmath30 ) .
the value of @xmath878 only depends on the residue @xmath879 .
denote @xmath880 thus we have [ 2]n_r(f)=_e|de_h = e_h(f),where the inner sum is over monic irreducible @xmath328 ( henceforth @xmath328 will always denote an irreducible polynomial in @xmath86 $ ] and summation over @xmath328 will be always understood in this sense ) .
if @xmath0 divides @xmath881 then @xmath882 for all @xmath78 $ ] .
otherwise exactly @xmath883 of the residues modulo @xmath328 satisfy @xmath882 , because @xmath884/h\to{\f_p}$ ] is a nonzero @xmath885-linear map .
[ l1]let @xmath886 $ ] be distinct monic irreducible polynomials , @xmath887 .
suppose @xmath888 .
let @xmath889 be chosen uniformly at random .
then the values @xmath890 are distributed uniformly in @xmath891/h_i$ ] .
the values @xmath892 depend only on the residue of @xmath15 modulo @xmath893 .
if @xmath888 then @xmath79 can be divided into @xmath894 complete systems of residues modulo any polynomial of degree @xmath2 .
denote @xmath904 .
assume that there are sequences of real numbers @xmath905 s.t .
the distribution of @xmath906 weakly converges to a distribution @xmath907 as @xmath147 .
assume further that @xmath907 is uniquely determined by its moments .
then @xmath908 converges in distribution to @xmath907 .
it is enough to show that for each @xmath6 the @xmath6-th moment of @xmath908 equals the @xmath6-th moment of @xmath906 for @xmath901 .
this follows immediately from the assumed properties , the definition of the @xmath6-th moment and the multiplicativity of expectation on independent variables .
for monic irreducible @xmath337 $ ] and @xmath78 $ ] denote @xmath909 lemma [ l1 ] shows that for any distinct @xmath910 the variables @xmath911 ( @xmath15 chosen uniformly from @xmath79 ) are independent for @xmath912
. we have @xmath913 if @xmath914 and @xmath915 otherwise
( recall that by @xmath87 we denote the bernoulli random variable taking the value 1 with probability @xmath45 and 0 with probability @xmath88 ) .
if @xmath99 then @xmath916 iff @xmath917 and we have @xmath918 for any @xmath15 . now taking all the monic irreducible @xmath328 s.t
@xmath919 and noting that the sum of their degrees is @xmath920 we obtain theorem 1
. now denote by @xmath921 $ ] the sequence of all monic irreducible polynomials of degree @xmath85 . given a sequence @xmath922 denote by @xmath923 a set of independent random variables with @xmath924 if @xmath925 and @xmath926 if @xmath99 .
taking @xmath927 , we see from lemma [ l1 ] that the conditions of lemma [ l2 ] are satisfied for @xmath928 . thus to establish theorems [ t2],[t3],[t4 ] we only need to show that @xmath929 converges in distribution to the limit stated in the theorems . for the rest of this section we omit @xmath145 from the notation ( it is implicit ) . in the setting of theorem [ t2 ]
we get ( for @xmath930 ) a sum of @xmath0 independent random variables distributed as @xmath931 , which converges to the poissonian distribution with mean 1 . in the setting of theorem [ t3 ]
it is enough to consider the variables corresponding to the irreducible polynomials of degree @xmath24 and @xmath811 ( if the latter is integral ) , because the number of polynomials of degree @xmath92 and @xmath932 is @xmath933 .
if @xmath24 is odd then there are @xmath934 variables @xmath935 distributed like @xmath936 , with mean @xmath937 and variance @xmath938 .
the conclusion now follows from the central limit theorem
. if @xmath24 is even and @xmath38 then there are @xmath939 variables @xmath935 with mean @xmath937 and variance @xmath940 and @xmath941 variables @xmath942 with mean @xmath943 and variance @xmath944 , which together gives the same result as for odd @xmath24 . for @xmath108 and @xmath24
even there are @xmath945 variables @xmath935 distributed as @xmath936 and @xmath946 variables @xmath947 and again the conclusion follows from the central limit theorem . in the setting of theorem [ t4 ] we can see as above that for odd @xmath24 the limit distribution of our sum is the same as for @xmath948 .
this is a sum of @xmath934 independent random variables with distribution @xmath936 .
we can group each @xmath0 such variables into a single variable with some distribution @xmath949 .
the distribution @xmath949 has mean @xmath24 , variance @xmath950 and a bound on the third moment independent of @xmath0 ( since the variable itself is bounded ) .
we have a sum of @xmath951 variables with distribution @xmath949 , so invoking the effective version of the central limit theorem ( see @xcite ) we obtain the conclusion of theorem [ t4 ] .
* acknowledgments . *
the author would like to thank zev rudnick for suggesting the problems studied in the present work and many helpful discussions and suggestions in the course of research and writing the present paper .
the present work is part of the author s m. sc .
thesis written under the supervision of zev rudnick at tel - aviv university .
the author would also like to thank the authors of @xcite for pointing out a small error in an earlier version of this paper .
finally the author would like to thank the anonymous referee of this paper for their thorough review and many helpful suggestions for improving the exposition . | we study the distribution of the zeroes of the l - functions of curves in the artin - schreier family .
we consider the number of zeroes in short intervals and obtain partial results which agree with a random unitary matrix model . |
recent experiments have demonstrated an impressive level of control over micro- and nanomechanical oscillators .
one key achievement is the engineering of mechanical modes with extremely low dissipation , providing remarkable isolation from the environment @xcite .
a second cornerstone is the detection and manipulation of micro- and nanomechanical motion .
both in solid - state based systems @xcite and cavity optomechanical settings @xcite , techniques have been perfected to monitor position fluctuations with an imprecision down to the standard quantum limit @xcite . moreover ,
sophisticated techniques for cooling and coherent excitation of individual mechanical modes have been developed @xcite , providing access to the quantum ground state @xcite . in the light of these achievements ,
mechanical oscillators appear as promising systems for the study of quantum physics .
the motivation is manifold .
firstly , manipulations on the quantum level represent the ultimate control over a system .
practically , this would have implications for the achievable sensitivity e.g. in force sensing @xcite , gravitational wave sensing @xcite , or nuclear spin imaging @xcite .
secondly , quantum control over massive objects could enable new tests of non - standard decoherence theories @xcite and illucidate the crossover from microscopic quantum systems to macroscopic classical systems .
finally , mechanical oscillators coherently coupled to other well - controlled quantum systems are examples of _ hybrid quantum systems _ , which are of interest in the context of quantum engineering , quantum information networking , and quantum information processing @xcite .
the different systems that are investigated have specific advantages and drawbacks .
optomechanical systems so far mostly realize a linear coupling of mechanical motion to a coherently excited optical cavity mode . in this limit
, the coupling can be very strong , and e.g. a mechanical analog of the autler - townes splitting in the energy spectrum of a coupled resonator - cavity system was observed @xcite ( similarly in an electromechanical system @xcite ) .
theoretical proposals have shown that the coupled system could generate mechanical squeezing @xcite , entanglement between the light field and the resonator @xcite , or schrdinger cat states of the light field @xcite .
however , the creation of mechanical fock states or more general quantum superpositions is not possible by a linear coupling to a classical field .
different approaches have been developed to overcome this limitation .
for example , dispersive measurements that rely on a quadratic coupling to the light field could permit the observation of quantum jumps between fock states @xcite .
alternatively , non - classical light states such as single photon states could be employed to generate superpositions of position states @xcite . yet
, single photons lead only to a minute coupling , and enhancement of the interaction by additional classical fields is required to achieve strong coupling with current experimental parameters @xcite .
a promising way to achieve full quantum control of a mechanical oscillator is to couple it to another well - controlled quantum system , preferably a two - level system .
solid - state two - level systems are very attractive in this context , and a variety of scenarios has been discussed @xcite .
mechanical systems have already been successfully coupled to a single electron spin in a solid @xcite or to a cooper pair box @xcite , and the recent demonstration of quantum control of a mechanical oscillator coupled to a superconducting qubit @xcite has set the pace in the field .
the latter experiment displays the advantages of entirely solid - state based approaches , being in essence a high coupling rate and a natural compatibility with cryogenic environments . yet , the coherence times of solid - state based two - level systems as used so far reside in the nanosecond to microsecond regime , which leads to the requirement of an exceptionally high coupling rate in the mhz range and consequently also very high mechanical resonance frequencies .
furthermore , the mechanical oscillator is typically integrated into a complex electronic environment and carries electrodes , making it difficult to independently optimize the mechanical quality factor .
coupling ultracold atoms to mechanical oscillators creates a novel , qualitatively different setting .
both optically or magnetically trapped clouds of ultracold atoms as well as trapped ions represent systems where quantum control has been achieved over all degrees of freedom @xcite .
one of their outstanding features is the possibility to achieve exceptionally long coherence times .
superpositions of internal states have been demonstrated to preserve coherence for several seconds @xcite .
remarkably , for neutral atoms this is still true in close proximity to chip surfaces @xcite .
in addition , an elaborate toolbox for the quantum manipulation of the atomic state is at hand @xcite , offering control over internal and motional degrees of freedom .
atoms can be regarded as mechanical oscillators themselves .
acting as a dispersive medium inside an optical cavity , an interesting variant of dispersive cavity optomechanics can be realized @xcite .
atoms provide both the continuous degree of freedom of collective atomic motion as well as a discrete set of internal levels that can be reduced to a two - level system .
furthermore , a unique feature of ultracold atoms is that dissipation can be tuned , e.g. by switching on / off laser cooling @xcite , and that self - interactions can be tuned and used for the generation of entanglement @xcite .
the challenge in the present context is to find a suitable coupling mechanism that provides controlled and sufficiently strong interactions of the atoms with a single mechanical mode of a micro- or nanostructured oscillator . in this article
we discuss different approaches to engineer a coupling between atomic systems and micro- and nanomechanical oscillators .
we focus initially on the most straightforward approach , the direct mechanical coupling of the oscillator s motion to the motion of trapped atoms .
the coupling results from a force that depends on the distance between the two systems , such that the force gradient converts an oscillation of the position of one system into a modulation of the force experienced by the other .
energy can be coherently exchanged when both systems oscillate with about the same frequency , i.e. for ( near-)resonance conditions .
to realize a direct mechanical coupling , it has been suggested to use electrostatic forces to couple to trapped ions @xcite or polar molecules @xcite . with neutral atoms ,
a first experiment was already performed @xcite , exploiting attractive surface forces between a micromechanical oscillator and a bose - einstein condensate ( bec ) , see sec .
[ sec : surfacecoupling ] .
a beneficial variant of this approach is to use the mechanical oscillator as an integral part of the trap itself , either as an electrode of an ion trap @xcite , current carrying wire of a magnetic trap @xcite , or as a mirror generating a one - dimensional optical lattice @xcite .
we discuss the direct mechanical coupling in detail in sec .
[ sec : directcoupling ] , where we show that the large mass difference between a single atom or ion and a micro- or nanostructured mechanical oscillator results in an impedance mismatch that limits the coupling to rather small values .
we point out that the coupling can not be simply increased beyond a certain level by increasing the strength of the coupling force , because a significant distortion of the atom trap would result .
however , one can think of several promising strategies to enhance the coupling , and we present them in the context of specific implementations in the subsequent chapters .
we first discuss the possibility to compensate the trap deformation in order to achieve stronger coupling in the case of an ion trap ( sec .
[ sec : ioncoupling ] ) .
second , a way to minimize the impedance mismatch is to use molecular - scale oscillators with low mass .
we discuss this perspective for surface - force coupling between a bec and a carbon nanotube ( sec .
[ sec : cntcoupling ] ) .
alternatively , one can increase the effective mass of the atoms by using a large number of them .
such collective enhancement is a promising route to achieve strong coupling mediated by an optical lattice ( sec .
[ sec : latticecoupling ] ) . for the perspective of creating a coupled quantum system where the mechanical oscillator is `` macroscopic '' ,
it is highly desirable to find coupling mechanisms where the impedance mismatch does not play a role .
indeed this is possible in several schemes : a powerful method is to use a high - finesse optical cavity that incorporates both the mechanical oscillator and the atoms , such that the two systems are coupled via the intracavity light field .
possible schemes rely either on a dispersive optomechanical coupling @xcite or on resonant interaction @xcite .
we discuss one particular scenario of this kind @xcite ( sec .
[ sec : cavitycoupling ] ) , where strong coupling may be achievable even for a single atom and a rather massive membrane oscillator .
coupling the oscillator motion to the internal degrees of freedom of atoms is another way to avoid the impedance mismatch .
this can be achieved by a magnet on the oscillator tip which transduces the mechanical motion into magnetic field oscillations , thereby establishing a coupling to the atomic magnetic moment @xcite .
we discuss this scheme in sec .
[ sec : magneticcoupling ] .
it could be used for quantum state tomography of the mechanical oscillator via an atom @xcite , enable the generation of entanglement between two nanomechanical oscillators @xcite , and it could be scaled up by the use of atoms in an optical lattice coupling to an array of oscillators @xcite .
magnetic coupling was the first mechanism that was studied experimentally , using an atomic gas in a vapor cell @xcite .
we note that magnetic coupling is also considered for the coupling of mechanical oscillators to color centers in diamond @xcite . finally , the mechanical oscillator could also act on the polarization state of light , which then couples to internal states of atoms @xcite .
the remarkable feature of this scheme is that long - distance entanglement between the mechanical oscillator and the atoms can be created by means of a projective measurement even for a thermal oscillator .
let us consider a system of two harmonic oscillators of different mass that are coupled by a spring .
this is a general model for various proposed methods to directly couple the motion of trapped atoms or ions to the vibrations of a mechanical oscillator via a distance - dependent force .
it allows us to make general statements about the achievable coupling strength . for simplicity
, we initially restrict the discussion to two undamped harmonic oscillators , one being an atom of mass @xmath0 trapped in a conservative potential with trap frequency @xmath1 , and the other being a mechanical oscillator with effective mass @xmath2 and eigenfrequency @xmath3 , see fig .
[ fig : couplingpot ] ( a ) .
furthermore , the two oscillators should have similar frequencies @xmath4 to achieve resonant coupling .
( a ) direct mechanical coupling of two harmonic oscillators of different mass . the two oscillators with respective masses
@xmath0 and @xmath5 interact via a potential @xmath6 that depends on their relative distance .
the interaction leads to shifted eigenfrequencies @xmath7 and @xmath8 and an equilibrium distance @xmath9 ( see sec . [
sec : trapdeform ] ) .
( b ) coupling parameters as a function of the normalized curvature of the coupling potential .
we show the relative change in trap frequency @xmath10 , the coupling strength parameter @xmath11 which compares the curvature of the coupling potential to the curvature of the trap , and the coupling strength @xmath12 normalized to the value @xmath13 where @xmath14 ( see text ) .
, title="fig:",scaledwidth=49.0% ] ( a ) direct mechanical coupling of two harmonic oscillators of different mass .
the two oscillators with respective masses @xmath0 and @xmath5 interact via a potential @xmath6 that depends on their relative distance .
the interaction leads to shifted eigenfrequencies @xmath7 and @xmath8 and an equilibrium distance @xmath9 ( see sec . [
sec : trapdeform ] ) .
( b ) coupling parameters as a function of the normalized curvature of the coupling potential .
we show the relative change in trap frequency @xmath10 , the coupling strength parameter @xmath11 which compares the curvature of the coupling potential to the curvature of the trap , and the coupling strength @xmath12 normalized to the value @xmath13 where @xmath14 ( see text ) . , title="fig:",scaledwidth=33.0% ] the hamiltonian of the coupled system is given by @xmath15,\ ] ] where @xmath16 ( @xmath17 ) is the momentum of the atom ( mechanical oscillator ) , @xmath18 ( @xmath19 ) its position , and @xmath20 ( @xmath21 ) the equilibrium position in the absence of the coupling .
we consider the case where the coupling can be described by a potential @xmath22 $ ] that depends only on the relative distance @xmath23 between the two oscillators ( i.e. @xmath6 is translationally invariant ) .
for example , the coulomb interaction between two charged oscillators or the magnetic interaction between two oscillators with a magnetic moment is of this kind .
because @xmath6 will in general shift the equilibrium positions ( see below ) , we define the excursions @xmath24 ( @xmath25 ) with respect to the equilibrium positions @xmath26 in the presence of @xmath6 . for small excursions @xmath27 , where @xmath28 is the equilibrium distance ,
@xmath6 can be expanded as @xmath29\approx u_c[d ] + u_c'[d](\delta z_m-\delta z_a ) + \frac{1}{2}u_c''[d](\delta z_m^2 - 2\delta z_m\delta z_a+\delta z_a^2).\ ] ] the coupling potential adds to the potentials of the atom and the mechanical oscillator and gives rise both to a deformation of the trap and to the desired mechanical coupling . the first term in eq .
( [ eq : uc ] ) introduces an energy offset and does not influence the dynamics .
the second term , proportional to the gradient @xmath30 , is responsible for the shift of the equilibrium positions from @xmath31 to @xmath32}{m\omega_{m,0}^2 } \quad \textrm{and } \quad z_a = z_{a,0}+\frac{u_c'[d]}{m\omega_{a,0}^2}.\ ] ] because of the inverse depencence on the mass , the position shift of the mechanical oscillator is a factor @xmath33 smaller than for the atom and can be neglected in most cases . for the atom ,
the shift should be small compared to the trap size in order not to expel the atom from the trap .
the third term of eq .
[ eq : uc ] is the most important one and has three contributions : the two terms quadratic in @xmath34 change the trap frequencies from the unperturbed values @xmath35 to @xmath36}{m } \quad \textrm{and } \quad \omega_a^2 = \omega_{a,0}^2+\frac{u_c''[d]}{m}.\ ] ] the frequency shift of the mechanical oscillator is suppressed again by the factor @xmath33 compared to the atom , and is usually negligible .
the change of the atomic trap frequency , on the other hand , can be substantial . in the extreme case where
@xmath37 , the trap is entirely generated by @xmath6 and the trap frequency is given by @xmath38 .
if , on the other hand , @xmath39 , the trap vanishes for @xmath40 , see fig .
[ fig : couplingpot ] ( b ) .
if higher derivatives of @xmath6 do not vanish , the coupling potential introduces anharmonicity to the atomic trapping potential .
this results in a dependence of the oscillation frequency @xmath41 on the atomic oscillation amplitude @xmath42 , which can be well approximated by @xcite @xmath43}{8m}\ ] ] for small amplitudes .
anharmonicity leads to dephasing and excitation of higher modes , especially for collective oscillations of atomic ensembles @xcite . the remaining mixed term in eq .
[ eq : uc ] gives rise to the coupling and can be interpreted as the interaction part of the hamiltonian , @xmath44\delta z_a \delta z_m = \epsilon m \omega_a^2 \delta z_a \delta z_m.\ ] ] it is important to realize that the coupling scales with the same curvature @xmath45 that also causes the change in trap frequencies in eq .
( [ eq : freqshift ] ) .
we therefore introduce the coupling strength parameter @xmath46}{m\omega_a^2 } = \frac{u_c''[d]}{m\omega_{a,0}^2+u_c''[d]},\ ] ] which is the ratio of @xmath45 and the curvature of the overall atomic trapping potential .
we now quantize the system by promoting the amplitudes @xmath34 to operators @xmath47 and @xmath48 , with the ground state amplitudes @xmath49 and @xmath50 and the bosonic creation and annihilation operators satisfying the usual commutation relations @xmath51=1 $ ] and @xmath52=1 $ ] .
this casts eq .
( [ eq : hint ] ) into the form @xmath53 this is a linear coupling in the amplitude , which simplifies for near - resonant coupling @xmath54 and in the rotating wave approximation to @xmath55 with the single - phonon coupling constant @xmath56 the linear coupling is in contrast to the dispersive opto- or electromechanical schemes , where the coupling is quadratic in the cavity field amplitude @xmath57 , @xmath58 . in optomechanics ,
a large frequency difference between the cavity and the mechanical oscillator can be tolerated , since the cavity can be strongly driven and the linearized dynamics in a frame rotating at the drive frequency is considered . for the direct linear coupling between two mechanical oscillators considered here ( undriven or
only weakly driven ) , significant energy transfer occurs only near resonance .
let us discuss some of the implications of the obtained coupling constant .
it contains a disadvantageous term @xmath59 that reflects the impedance mismatch between the two oscillators . for a single atom with @xmath60 g and micro- or nanostructured oscillators with @xmath61 g @xcite , we obtain @xmath62 , resulting in a strong reduction of the coupling rate .
next , @xmath12 scales with the resonance frequency , which is limited by the experimentally achievable curvature of the atomic trapping potential .
while magnetic microtraps and optical lattices can reach @xmath63mhz @xcite , ion traps have reached up to 50 mhz @xcite .
finally , @xmath12 contains the coupling strength parameter @xmath11 as defined in eq .
( [ eq : epsilon ] ) . in most cases ,
this parameter will be small , @xmath64 , to avoid strong distortion of the trapping potential by @xmath45 .
if on the other hand the trap is designed such that @xmath45 itself provides the atomic confinement , the parameter saturates at @xmath65 .
it is thus not possible to increase @xmath11 at will by simply increasing @xmath45 .
for @xmath66 it is actually possible to reach @xmath67 , but in this regime @xmath7 decreases with increasing @xmath68 and the trap becomes very shallow and strongly distorted , see fig .
[ fig : couplingpot ] ( b ) . as long as no other measures such as compensation of the trap distortion are taken ( see below ) , we thus have @xmath69 overall , @xmath70 is typically in the regime of @xmath71 hz for the direct mechanical coupling of a single atom or ion to a nanostructured oscillator . to put the coupling rate into context
, one has to compare it to the dissipation rates in the system .
there , the question about the mechanical quality factor of the atom arises . since levitated systems such as ultracold atoms have no direct contact with the environment , there are only minute damping mechanisms present ( e.g. the friction caused by collisions with the background gas , by resistive damping of moving mirror charges in an ion trap , or by damping of induced currents from moving paramagnetic atoms ) .
larger effects on the motion come from trap anharmonicity and residual heating due to noise in the structures generating the trap ( fluctuating currents , voltages , magnetization ) . in particular , slow drifts of trap parameters lead to drifts of the resonance frequency and cause dephasing .
while experiments with collective motion of ultracold atoms have achieved quality factors @xmath72 @xcite limited by the atomic trap lifetime and rather low trap frequencies , trapped ions can achieve @xmath73 in cryogenic penning traps @xcite .
however , @xmath74 is in many cases not the relevant parameter , and particularly in the context of quantum engineering , the decoherence rate @xmath75 is the key figure of merit . in ion traps , nonclassical motional states with coherence lifetimes up to @xmath76 ms have been reported @xcite . for neutral atoms , nontrivial motional quantum states have been studied in optical lattices by populating excited lattice bands @xcite , and in magnetic microtraps in the context of trapped atom interferometers @xcite , where decoherence lifetimes up to 200 ms have been demonstrated @xcite .
furthermore , for some systems it is possible to map motional states to internal states which have much longer coherence lifetimes .
this could be used to preserve quantum states for several seconds .
on the other hand , decoherence of the mechanical oscillator is typically fast despite a high mechanical quality factor , since the coupling to the environment introduces a dependence on the temperature @xmath77 of the bath , @xmath78 for standard cryogenic cooling with @xmath79k and good quality factors @xmath80 one ends up with @xmath81mhz .
however , for state of the art cryogenics with @xmath82mk and exceptional quality factors @xmath83 as demonstrated in @xcite one may achieve @xmath84hz , sufficient to achieve the strong coupling regime where @xmath85 ) . while the large mechanical decoherence rate is a general challenge for all experiments aiming at the observation of quantum behaviour of mechanical oscillators ,
it is particularly severe in the presence of low coupling rates .
strategies to increase the coupling beyond the values given above are thus highly desired . a natural way to increase
the coupling rate is to couple many atoms simultaneously to the oscillator .
coupling to the atomic center of mass mode @xmath86 of an ensemble of @xmath87 atoms , one obtains a collectively enhanced coupling strength @xmath88 .
we note that this result can be obtained by simply replacing @xmath0 with the mass @xmath89 of the whole ensemble in eq .
( [ eq : coupling ] ) . the coupling rate increases with @xmath90 , and for large atomic ensembles with @xmath91 , a collectively enhanced coupling of @xmath92 could be reached .
a system benefitting from this approach is discussed in sec .
[ sec : latticecoupling ] .
furthermore , the coupling strength parameter @xmath11 can be increased beyond the limit given in eq .
( [ eq : couplingcondition ] ) if the trap deformation is compensated with the help of additional potentials .
this could allow one to increase @xmath12 by about two orders of magnitude .
we give an example in sec .
[ sec : ioncoupling ] .
an alternative approach is to focus on molecular - scale oscillators such as carbon nanotubes . due to the ultimately low effective mass @xmath93
g , both @xmath12 and @xmath94 can be increased by about two orders of magnitude compared to the more massive oscillators discussed above .
this scenario is further discussed in sec .
[ sec : cntcoupling ] .
finally , the coupling can be significantly enhanced with the help of high - finesse optical cavities .
the cavity acts as a lever and provides a coupling that is independent of the mass ratio .
this is possible because the cavity breaks the translational invariance , i.e. the interaction potential is no longer of the form @xmath22 $ ] considered above .
this scheme provides much stronger coupling of up to @xmath95 and we discuss it in sec .
[ sec : cavitycoupling ] .
we point out that even if the absolute coupling rate is moderate , the normalized coupling rate , i.e. @xmath12 compared to the oscillator frequency , is rather large for coupled atom - oscillator systems .
in optical cavity qed experiments , by comparison , the highest normalized coupling rates achieved so far amount to @xmath96 @xcite , with @xmath97 being the optical resonance frequency of the cavity . only recently
, solid - state based cavity qed experiments in the microwave regime could enter the ultrastrong coupling limit with larger normalized coupling strengths of up to @xmath98 @xcite . in this regime ,
the counter - rotating terms of the coupling hamiltonian become relevant and lead to new effects , which would be interesting to study for mechanical oscillators .
remarkably , in the first experiment demonstrating strong coupling of nanomechanical motion to a superconducting qubit , @xmath99 was achieved @xcite .
trapped ions are one of the best controlled systems with elaborate techniques for quantum state engineering . in particular , ion motion can be controlled on the single quantum level by addressing motional sidebands of optical transitions with lasers @xcite , and quantum states of motion are routinely employed in experiments e.g. on quantum computing @xcite .
coherence of motional quantum states has been maintained for more than hundred milliseconds @xcite , and thus , ions are a particularly promising coupling partner for micro- and nanomechanical oscillators .
a further advantage is the availability of a strong coupling force : an electrically charged mechanical oscillator can exert a substantial force on the ion , allowing to couple the mechanical motion to the ion motion .
several proposals @xcite have considered such a coupling .
notably , this coupling has been recently demonstrated at the single quantum level for an ion and the smallest mechanical oscillator available , namely another ion @xcite .
we want to estimate the achievable coupling rate for an ion - oscillator system similar to ref .
the coupling constant can be written in the form of eq .
( [ eq : coupling ] ) , where the details about the coupling potential are absorbed in @xmath11 .
more explicitly , a charged nanoscale oscillator would provide a coupling via the coulomb potential @xmath100=1/(4\pi\epsilon_0)\times eq / d$ ] , where @xmath101 is the elementary charge of a singly charged ion , and @xmath102 the charge on the tip of the oscillator . as discussed above , in order to avoid a strong distortion of the trap , the position shift of the ion due to the gradient @xmath103 $ ] has to be small compared to @xmath9 , which yields the condition @xmath104 , and the curvature @xmath105 $ ] has to be smaller than the curvature of the ion trap , or equivalently @xmath106 , giving essentially the same limit . while the coupling rate scales with @xmath107 , the maximum achievable trap frequency in an ion trap scales with @xmath108 , such that lighter ions achieve a larger coupling . for a @xmath109be@xmath110 ion in a trap with @xmath111mhz and a low - mass oscillator with @xmath112 g one obtains a coupling constant @xmath113hz for @xmath114 .
one can see that even though the high achievable trap frequency is advantageous , the coupling remains modest . to achieve @xmath115 for this case
, we assume a nanoscale oscillator with a conductive sphere at the tip with a capacity @xmath116 , proportional to the sphere radius @xmath117 . for @xmath118
one has to charge the sphere with a voltage @xmath119 , which for @xmath120 m and @xmath121 nm amounts to @xmath122v . to benefit from the excellent motional control , coupling to a single or very few ions
is preferable , and collective enhancement is probably not the way to go . to achieve stronger
coupling one can increase @xmath105 $ ] further , e.g. by decreasing @xmath9 or increasing @xmath123 , if one compensates for the resulting distortion of the ion trap .
additional trap electrodes , not mounted on the oscillator , allow one to compensate the static electric field , gradient , curvature , and perhaps also higher - order derivatives of @xmath6 , while the oscillating component is not reduced . with this approach , the distorting effect of the coupling force can be reduced and @xmath11 can be made larger than unity .
note that this requires precisely tuned compensation voltages about a factor @xmath11 larger than the voltages that produce the unperturbed trap , and also the voltage charging the oscillator has to be larger by the same factor .
this is experimentally challenging , and compensation may allow one to increase the coupling by 1 - 2 orders of magnitude , such that @xmath124 .
similar compensation techniques could also be employed in the case of magnetic coupling , see sec .
[ sec : magneticcoupling ] .
a ) schematic overview of the experiment .
a micromechanical cantilever is integrated on an atom chip with microfabricated gold wires for magnetic trapping .
a bec is created and positioned close to the cantilever in order to sense mechanical oscillations .
figure adapted from @xcite .
b ) calculation of the undisturbed magnetic potential @xmath125 ( red , dashed ) and the resulting trapping potential ( black solid line ) destorted by the casimir - polder potential @xmath126 ( blue , solid ) for small atom - surface distance .
it leads to a barrier of height @xmath127 , a shift of the trap minimum and a change in trap frequency .
parameters are @xmath128khz and @xmath129
nm . the chemical potential @xmath130 of a bec with @xmath131 atoms is shown for comparison .
oscillations of the cantilever modulate the trapping potential ( light blue , grey ) .
right panel : enlarged view of the potential close to the trap minimum ( indicated by the box ) .
modulation of the trap minimum position with an amplitude @xmath132 leads to excitation of collective motion .
, title="fig:",scaledwidth=36.0% ] a ) schematic overview of the experiment .
a micromechanical cantilever is integrated on an atom chip with microfabricated gold wires for magnetic trapping .
a bec is created and positioned close to the cantilever in order to sense mechanical oscillations .
figure adapted from @xcite .
b ) calculation of the undisturbed magnetic potential @xmath125 ( red , dashed ) and the resulting trapping potential ( black solid line ) destorted by the casimir - polder potential @xmath126 ( blue , solid ) for small atom - surface distance .
it leads to a barrier of height @xmath127 , a shift of the trap minimum and a change in trap frequency .
parameters are @xmath128khz and @xmath129 nm .
the chemical potential @xmath130 of a bec with @xmath131 atoms is shown for comparison .
oscillations of the cantilever modulate the trapping potential ( light blue , grey ) .
right panel : enlarged view of the potential close to the trap minimum ( indicated by the box ) .
modulation of the trap minimum position with an amplitude @xmath132 leads to excitation of collective motion .
, title="fig:",scaledwidth=60.0% ] the motion of neutral atoms can be coupled to mechanical oscillators via a distance - dependent potential , similar to the case of ions . in the following
we summarize an experiment where such direct motional coupling was demonstrated for the first time @xcite .
the experiment studies the coupling of a mechanical oscillator to collective oscillations of a bose - einstein condensate ( bec ) in a trap .
the fact that the atoms are bose - condensed is important because of the small spatial extension of the condensate compared to a thermal gas .
while in the present experiment the quantum nature of the bec is not directly exploited , it is an important ingredient of future work since it provides the initialization of the atoms in the motional ground state .
the employed oscillator is a commercial micromechanical cantilever with dimensions @xmath133 m , an effective mass of @xmath134ng , a fundamental resonance frequency @xmath135khz , and a quality factor @xmath136 .
the coupling is realized via attractive surface forces which are of fundamental origin and do not require functionalization of the mechanical oscillator .
surface forces typically decay over a few micrometers distance and thus require exceptional control over the position of the atoms . in the experiment ,
magnetic microtraps on an atom chip are used to create , hold and position becs with a few nanometers precision close to an on - chip integrated microcantilever , see fig .
[ fig : surfforceschem ] ( a ) . for small atom - surface distance ,
the magnetic potential @xmath125 used to trap the atoms is strongly deformed by the surface potential @xmath137 , where @xmath138 denotes the casimir - polder ( cp ) potential @xcite , and @xmath139 includes possible additional potentials due to stray fields . in the direction perpendicular to the surface ,
the combined potential is ( see fig . [
fig : surfforceschem ] b ) @xmath140 & = & u_m + { u_\mathrm{cp}}+ { u_\mathrm{ad}}+ u_\mathrm{grav } \nonumber\\ & = & \frac{1}{2}m\omega_{a,0}^2(z - z_{a,0})^2 -\frac{c_4}{(z - z_m)^4 } + { u_\mathrm{ad}}[z - z_m ] + m g z.\end{aligned}\ ] ] here , @xmath141 is the position of the cantilever surface , @xmath142 the cp - coefficient for a perfect conductor , @xmath20 the unperturbed minimum position of the magnetic trap , @xmath1 the unperturbed trap frequency , @xmath0 the atomic mass , and @xmath143 the acceleration in the gravitational potential @xmath144 . ( a ) coupling strength parameter @xmath11 for different surface potential strength @xmath145 as a function of atom - surface distance and ( b ) barrier height .
chosen parameters are @xmath146khz and @xmath147 with @xmath142 the cp - coefficient for a perfect conductor .
stronger surface potential with large @xmath148 can arise e.g. from adsorbates or surface contaminants , weaker potentials arise e.g. for molecular - scale oscillators .
, scaledwidth=77.0% ] to estimate the achievable coupling strength for this kind of coupling potential , we calculate the coupling strength parameter @xmath149 for a pure cp potential .
in many cases , additional ( surface ) potentials originating e.g. from adsorbates , stray charges , or static external charging , have a very similar distance dependence as @xmath138 .
[ fig : epsvsdist ] shows @xmath11 for the case of a harmonic trap with @xmath146khz and attractive coupling potentials of the form @xmath100=-\beta c_4/d^4 $ ] for several coefficients @xmath148 .
one finds that the trap vanishes for small atom surface distance , where the coupling strength parameter reaches its maximum @xmath150 .
the rapid decay of @xmath11 with the distance can be mitigated by stronger coupling potentials , where @xmath151 can be achieved also for deep traps ( see fig .
[ fig : epsvsdist ] b ) with up to 10 bound levels . for a typical experimental value @xmath152 , the coupling rate for a single atom amounts to @xmath153hz in this system , too small for the coupling to be detectable on the single - phonon level .
however , for a coherently driven cantilever , the atoms can be used as a probe to readout the oscillator vibrations . in the experiment , coherently driven cantilever oscillations are coupled to the motion of trapped ultracold atoms nearby .
the atoms are ramped to a typical distance of @xmath154 m from the cantilever , where the magnetic trap is markedly affected by the static surface potential and one reaches values of @xmath152 for a trap depth @xmath155 . to achieve resonant interaction , the trap frequency
is tuned to the eigenfrequency of the cantilever s fundamental mode . when the cantilever is excited with a piezotransducer at different frequencies around its resonance , mechanical oscillations
excite the center of mass ( c.o.m . )
mode of the condensate , leading to trap loss over the potential barrier . in this way
, the resonance spectrum of the cantilevers fundamental mode can be probed by the atoms ( see fig . [ fig : dynloss ] ( a ) ) .
remaining atom number in a resonant trap for varying cantilever excitation frequency @xmath156 .
the dark ( light ) blue circles correspond to a cantilever amplitude @xmath157 nm ( @xmath158 nm ) on resonance .
solid lines : lorentzian fits with @xmath159hz fwhm , corresponding to the width of the cantilever resonance .
the remaining atom number with ( @xmath160 ) and without ( @xmath161 ) resonant piezo excitation of the cantilever is indicated .
figure adapted from @xcite .
b ) contrast @xmath162 for the c.o.m
. mode as a function of the cantilever amplitude @xmath163 . for measurements on the metallized side ( dark blue ) a @xmath164 times larger coupling signal than on the dielectric backside ( light blue ) is found . for comparison ,
the contrast for an off - resonant trap with @xmath165khz is shown ( red ) .
the dotted line indicates the rms noise of the measurement .
the red solid line is a 1d numerical simulation with no free parameters .
, title="fig:",scaledwidth=59.0% ] remaining atom number in a resonant trap for varying cantilever excitation frequency @xmath156 .
the dark ( light ) blue circles correspond to a cantilever amplitude @xmath157 nm ( @xmath158 nm ) on resonance .
solid lines : lorentzian fits with @xmath159hz fwhm , corresponding to the width of the cantilever resonance .
the remaining atom number with ( @xmath160 ) and without ( @xmath161 ) resonant piezo excitation of the cantilever is indicated .
figure adapted from @xcite .
b ) contrast @xmath162 for the c.o.m
. mode as a function of the cantilever amplitude @xmath163 . for measurements on the metallized side ( dark blue ) a @xmath164 times larger coupling signal than on the dielectric backside ( light blue ) is found .
for comparison , the contrast for an off - resonant trap with @xmath165khz is shown ( red ) .
the dotted line indicates the rms noise of the measurement .
the red solid line is a 1d numerical simulation with no free parameters .
, title="fig:",scaledwidth=39.0% ] the sensitivity of this readout method can be inferred by measuring the coupling signal on resonance as a function of the cantilever amplitude .
such measurements can be performed both on the metallized frontside and on the dielectric backside of the cantilever .
figure [ fig : dynloss ] b ) shows measurements of the contrast @xmath166 for the c.o.m . mode with @xmath161 ( @xmath160 ) the remaining atom number off ( on ) resonance as a function of the cantilever amplitude . for comparison , also the signal for an off - resonant trap with @xmath165khz is shown .
the measurements yield a minimum resolvable r.m.s .
cantilever amplitude of @xmath167 nm for unity signal to noise ratio without averaging .
the contrast for a given amplitude is different for the front and the backside of the cantilever . with additional measurements and simulations , it is possible to quantify the difference in the strength of the surface potential ( see also fig .
[ fig : cntcoupstr ] ) .
it was found that on the metallized side @xmath168 while on the dielectric side it is consistent with a cp potential within the errorbars .
a likely origin of the stronger potential is the presence of surface adsorbates which cause electric stray fields @xcite .
the potential strength can be quantitatively reproduced when considering the becs deposited on the surface during operation of the experiment .
the larger electric dipole moment of adsorbates and the larger number of experiments performed on the metallized side are consistent with the stronger potential .
two effects influence the observed sensitivity .
first , the finite lifetime of the atoms in the trap due to parasitic loss limits the coupling duration and thus the achievable excitation of the atoms for small amplitudes .
second , trap anharmonicity leads to dephasing of the cloud oscillations and thereby to a maximum cloud amplitude for a given cantilever amplitude .
more insight into the dynamics of the cloud and the influence of the trap anharmonicity can be obtained from a 1d numerical integration of the time dependent gross - pitaevskii equation .
the measured data is well reproduced by the simulation where only the inferred strength of the surface potential and measured parameters of the experiment enter ( fig .
[ fig : dynloss ] b ) . in the measurements ,
trap loss was used as the simplest way to detect the bec dynamics induced by the coupling .
this implies that in order to be observable , the excitation has to lead to large amplitude collective oscillations , such that a part of the cloud spills over the barrier and is lost from the trap .
by contrast , bec amplitudes down to the single phonon level could be observed by direct imaging of the motion .
a coherent state @xmath169 of the c.o.m
. mode of @xmath170 atoms with @xmath171 released from a relaxed detection trap with @xmath172 hz has an amplitude of @xmath173 nm after @xmath174 ms time - of - flight .
this is about @xmath175 of the bec radius and could be resolved by absorption imaging with improved spatial resolution . from the numerical simulation of the cloud excitation
we estimate that @xmath176 nm would excite the bec to @xmath171 within an interaction time of @xmath177 ms and could thus be detected .
this should allow detection of the thermal amplitude @xmath178 nm of the used cantilever .
top graph : bec response as a function of @xmath7 , for fixed resonant piezo excitation ( @xmath179 nm ) ( dark blue ) .
datapoints are connected with a line to guide the eye .
light blue : reference measurement without piezo excitation .
two major resonances at @xmath180 ( c.o.m .
mode ) and @xmath181 ( parametric breathing mode ) and up to four smaller ones ( green arrows ) are observed . due to cantilever aging ,
@xmath182khz in this measurement .
bottom graph : set values of @xmath9 , chosen to keep @xmath161 sufficiently constant and to avoid saturation of @xmath160 .
figure adapted from @xcite.,scaledwidth=65.0% ] so far , the described measurements were performed for resonant coupling , where the trap frequency and the mechanical resonance frequency match .
both single atoms as well as ensembles of atoms furthermore show parametric resonances , and the latter also exhibit shape resonances .
the spectrum of collective mechanical modes is influenced by atomic collisions , which leads to additional , shifted resonance frequencies compared to the harmonic spectrum of a non - interacting gas @xcite .
when anharmonicity is present , resonance broadening and nonlinear mode mixing can further modify the spectrum .
studying the atomic mode spectrum can give insight about which modes can be excited efficiently , how sensitive the atomic response depends on the trap frequency , and to which extent the strong anharmonicity of the coupling trap affects the situation . with a measurement of the dependence of the atomic response on the trap frequency @xmath7 , a spectroscopy of the strongest cloud excitations can be performed .
it is found that the cantilever can be coupled selectively to different , spectrally narrow bec modes , see fig . [
fig : cantres ] .
the sharp resonances indicate coherent excitation and provide a means to control the coupling by changing the trap frequency .
coupling ultracold atoms and mechanical oscillators via surface forces has a central advantage : the employed coupling force is of fundamental origin , and functionalization of the oscillator , e.g. by the fabrication of mirrors , magnets or electrodes on the oscillator tip , is not required .
this is in contrast to most of the theoretical proposals considering atom - resonator coupling so far .
coupling via surface forces could thus be used to couple atoms to molecular - scale oscillators . due to their ultimately low mass and the high achievable quality factor , single wall carbon nanotubes ( cnt )
are particularly promising in this context . considering a suitable geometry with a nanotube diameter @xmath183 nm and length @xmath184
m we obtain @xmath185khz and an effective mass of @xmath186 kg .
this leads to a room temperature thermal motion amplitude of @xmath187 , much larger than typical atom - surface distances in the previously described experiment .
even more impressive , the ground state amplitude is almost macroscopic , @xmath188 .
these numbers are about four orders of magnitude larger than those of the cantilever used in the experiment above .
if the readout scheme of direct imaging of the cloud excitation is employed , the expected sensitivity should permit measurements close to the ground state of the nanotube . however , one central concern is whether the strength of the surface potential of such nanoscale objects is sufficient for coupling experiments . from an estimation of the casimir - polder potential of a cnt @xcite
we infer a potential @xmath189 in the relevant distance range , where @xmath138 is the potential of a bulk conductor .
the achievable coupling strength parameter for a resonant trap with sufficient trap depth in this case is of the order @xmath190 .
stronger coupling could be realized e.g. by static charging of the cnt .
electric fields @xmath191 $ ] polarize neutral atoms via their static polarizability @xmath192 , and a potential @xmath100=-(\alpha_0/2)|e[d]|^2 = -(\alpha_0/2)(q/4\pi\epsilon_0d^2)^2\propto -1/d^4 $ ] arises , with the same distance dependence as the cp potential . due to the small atom - surface distance , the electric field and
also the field gradient will be large even for small charging voltages .
since cnts can be fabricated in metallic or semiconducting configuration , charging does not require deposition of electrodes on the nanotube , and electric contacting is a standard technique that is compatible with high quality factors @xcite . by charging the nanotube , a potential much stronger than the cp potential
could be generated , and coupling strength parameters of @xmath193 might be possible . figure [ fig : cntcoupstr ] shows the coupling strength parameter @xmath11 as a function of the strength of the surface potential .
the observed @xmath11 at the both sides of the cantilever used in the described experiment is shown together with the expected value for a cnt .
variation of the coupling strength parameter @xmath11 as a function of surface potential strength normalized to the cp potential of a perfect conductor , calculated for a trap depth of @xmath194 ( solid line ) .
in addition , the evaluated @xmath11 above the metallized and dielectric side of the cantilever used in the experiments and the expected value for a cnt are shown.,scaledwidth=55.0% ] finally , coupling schemes that do not rely on surface forces could be employed .
one possibility is to use a doubly clamped , suspended cnt as current carrying wire to generate a magnetic trap .
several papers have investigated this situation theoretically @xcite and studied its feasibility for atom trapping . in this configuration ,
mechanical oscillations of the nanotube translate directly into trap oscillations and provide a coupling strength parameter @xmath118 . in combination with compensating fields , even values of @xmath195 could be possible .
we now estimate the achievable coupling parameters for a cnt coupled either to a small bec or to a single atom . for an atom cloud ,
three - body collisional loss is the dominating dissipative process , which restricts the choice of the trap frequency . for @xmath196 atoms in an elongated trap with aspect ratio @xmath197 that is resonant with the cnt fundamental mode considered above with frequency @xmath198khz
, we obtain an atomic loss rate of @xmath199hz which we assume to be also the value of the atomic decoherence rate @xmath200 . for a quality factor of @xmath201 and @xmath82mk
we have a decoherence rate @xmath202hz .
this is to be compared to the single - phonon single - atom coupling rate , which amounts to @xmath203hz , and for an ensemble of atoms , the coupling is collectively enhanced according to @xmath204 . on the other hand , for a single atom in a magnetic trap ,
the only dissipative processes are collisions with the background gas , technical current noise induced heating , or johnson noise induced spin - flip loss .
these effects are small or can be controlled , and we assume an atomic decoherence rate @xmath205hz . here , the trap frequency can be made as large as possible , and we choose @xmath206khz in a magnetic microtrap . for a cnt with @xmath207 m such that @xmath180 we find a coupling rate @xmath208hz , and for the same quality factor as above we again obtain @xmath209hz . comparing the coherent coupling @xmath210 and dissipative rates for both cases we obtain @xmath211 this is sufficient to enter the strong coupling limit both for a single atom and collectively for an atom cloud with comparable rates .
furthermore , the strong coupling condition is met for a feasible coupling strength parameter down to @xmath193 .
however , cryogenic ground state cooling will not be feasible for the low resonance frequency considered here , and the mechanical mode will still be thermally populated with a mean phonon number @xmath212 .
additional cooling e.g. via radiation pressure @xcite in an optical cavity could be a way to initialize the oscillator in the ground state .
coupling many atoms collectively to a mechanical oscillator is another way to achieve strong coupling . in this section
we discuss a scheme where an optical lattice provides a long - distance coupling between a micromechanical membrane oscillator and @xmath87 laser cooled atoms in the lattice @xcite .
consider a laser beam with wave vector @xmath213 that is retro - reflected on a membrane oscillator to form a 1d standing - wave optical potential of the form @xmath214 with the potential depth @xmath127 , suitable to trap laser cooled atoms .
movements of the membrane will move the standing wave , such that the membrane oscillations are directly transferred to an oscillation of the trapping potential of the atoms .
this is a direct mechanical coupling with a coupling parameter @xmath118 .
if the membrane oscillation frequency is resonant with the trap frequency of the atoms along the lattice direction , resonant excitation of the collective c.o.m . mode of the atoms leads to an energy transfer from the membrane to the atom cloud .
figure [ fig : latticecoupschem ] depicts the situation .
schematic coupling mechanism : a lattice laser is retro reflected on a membrane oscillator to form a standing wave dipole trap .
oscillations of the membrane shake the lattice and excite c.o.m .
motion of atoms trapped in the lattice .
laser cooling of the atoms provides sympathetic cooling of the oscillator.,scaledwidth=65.0% ] on the other hand , there is also a back - action of the atoms onto the membrane : the restoring force @xmath215 that attracts an atom displaced by @xmath216 towards an anti - node of the standing wave arises from a coherent redistribution of photons between the two @xmath213-vectors components that make up the lattice @xcite . for an oscillating atom
, this leads to a modulation of the power of the beam impinging on the membrane @xmath217 which results in a modulation of the radiation pressure force @xmath218 experienced by the membrane due to the reflection of the light with power reflectivity @xmath219 .
if the membrane is oscillating , it excites the atoms to collective motion through the coupling , and there is a fixed phase relation between atomic and membrane motion .
the resulting radiation pressure modulation at the membrane will have the right phase lag to damp the membrane oscillation .
furthermore , one can apply laser cooling to the trapped atoms and thereby continuously extract their gained energy .
it was found in @xcite that the resulting cooling rate of the membrane is given by @xmath220 with the mechanical energy damping rate @xmath221 , the atomic cooling rate @xmath222 , and the coupling rate @xmath223 .
this leads to a steady state occupation of the membrane mode of @xmath224 where @xmath225 is the thermal occupation of the membrane .
the motion of the membrane can in this way be sympathetically cooled via laser cooling of the atoms .
this scheme has several remarkable advantages .
most prominently , the lattice provides a coupling that can bridge large distances ( in principle hundreds of meters ) , limited only by retardation of the light field and by the frequency stability of the laser .
this relaxes the technical effort considerably , as the mechanical oscillator does not have to be integrated into an ultracold atom vacuum setup .
it also simplifies the implementation of a cryogenic environment for the oscillator . on the atomic side ,
it is sufficient to superimpose laser cooling and an optical lattice , and no positioning is necessary .
this simplifies and shortens the experimental cycle considerably .
furthermore , laser cooling of the trapped atoms can be switched on and off , thereby giving access to both sympathetic cooling and the regime of coherent interaction .
finally , the coupling strength can be collectively enhanced and thus made large by loading a large ensemble of atoms into the lattice . as an example , for a sin membrane of dimensions @xmath226 m with resonance frequency @xmath227mhz and effective mass @xmath228 g one can expect @xmath83 at @xmath229k @xcite and a power reflectivity @xmath230 for @xmath231 nm . for an ensemble of @xmath232 atoms @xcite one obtains a collectively enhanced coherent coupling rate @xmath233khz , larger than all dissipative rates . furthermore ,
if raman sideband cooling is applied to the atoms , cooling factors of the order of @xmath234 can be achieved , sufficient for ground state cooling of a membrane precooled to below @xmath235mk .
already the simple picture given here shows the remarkable aspect of an asymmetric coupling : while the atoms experience a force @xmath236 that leads to a coupling constant @xmath237 as derived in eq .
( [ eq : coupling ] ) , the membrane feels a force @xmath238 which is reduced by the power reflectivity @xmath219 , and thus also the coupling constant is reduced to @xmath239 . a rigorous treatment of this effect as well as several dissipative processes such as radiation pressure noise call for an elaborate quantum description of the system .
such a description is given in @xcite and it shows that the coupled atom - oscillator system considered here is an example of a cascaded quantum system .
the use of a high - finesse optical cavity to mediate the coupling represents an approach that does not require collective enhancement to achieve strong coupling . in @xcite
, a scheme was elaborated that achieves a coupling that does not depend on the mass ratio @xmath240 , thereby enabling strong coupling between a single atom and a micromechanical membrane oscillator .
the system combines an optomechanical setup with a membrane inside a cavity @xcite and a cavity qed system with a single trapped cs atom @xcite , fig .
[ fig : cavitycouplingschem ] ( a ) shows the setting .
( a ) coupling a single atom and a membrane via an optical cavity .
( b ) cavity transmission spectrum and frequencies of two detuned lasers that mediate the coupling .
( c ) relevant level scheme of cs with the two laser frequencies , both red detuned from the atomic transitions to provide attractive dipole potentials .
( d ) the atom and the membrane are placed at specific positions within the spatial mode structure of the two standing wave cavity fields .
( e ) small displacements of the membrane shift the cavity resonances and thus the relative intensity of the two fields ( see ( b ) ) , thereby shifting the trap potential of the atom .
figure adapted from @xcite .
, scaledwidth=58.0% ] the cavity light field interacts with the membrane via radiation pressure while it serves as a standing wave dipole trap for the atom . on the other hand , atom and membrane act as dispersive media that shift the cavity resonance depending on their position in the cavity light field . to realize an interaction that is linear in the amplitudes of the atom and the membrane , two different laser fields ( @xmath241 )
are required with opposite detunings from the cavity resonance ( see fig . [
fig : cavitycouplingschem ] b ) , but both red detuned from atomic transitions to provide an attractive potential for the atom ( see fig .
[ fig : cavitycouplingschem ] ( c ) ) . in this configuration ,
the motion of the atom or the membrane leads to a change in intracavity intensity , proportional to the gradient of the intensity at their respective positions . to achieve the desired coupling
, the membrane has to be placed at a point of maximal intensity gradient with similar slope for both cavity fields , while the atom is trapped at a position where both fields have close to maximum but opposite slopes of the intensity ( see fig .
[ fig : cavitycouplingschem ] ( d ) ) . if the membrane is displaced from its equilibrium position , the blue detuned laser comes closer to resonance and generates a deeper lattice potential , while the red detuned laser becomes further detuned and thus weaker .
this gives rise to a shift of the atomic trapping potential ( fig .
[ fig : cavitycouplingschem ] ( e ) ) and results in the desired coupling of the type @xmath242 .
a feasible set of parameters including a membrane of dimensions @xmath243 m with @xmath244mhz and @xmath83 at @xmath229k , coupled to a single cs atom through a cavity of finesse @xmath245 , achieves a coupling rate @xmath246khz and decoherence rates a factor 10 smaller , firmly entering the strong coupling regime .
this scheme achieves by far the largest coupling rate , even though the mechanical oscillator is rather large and the mass of the atom differs by 12 orders of magnitude .
since the required parameters have already been achieved in individual experiments @xcite , the scheme appears as a promising , yet technologically challenging way to achieve strong coupling between a single atom and a mechanical oscillator at the quantum level .
coupling to internal states of the atoms has the advantage that high - frequency mechanical oscillators can be employed , as the oscillator vibrations now should be resonant e.g. with zeeman transitions that can have frequencies in the mhz to ghz range .
additionally , the internal levels can be chosen and modified ( e.g. by state - selective microwave dressing @xcite ) such that the atoms act as effective two - level systems .
this is important for the generation and readout of quantum states , because in a two - level system , single excitation quanta and coherent superposition states can be prepared by driving rabi oscillations with classical fields .
furthermore , the quantum control of internal states is more advanced than for motional states ( at least for neutral atoms ) , and one can benefit from the exceptionally long coherence lifetimes of internal states .
we consider the scenario of a nanomechanical cantilever with a nanomagnet on the tip , whose magnetic field couples to the spin of a trapped atom or bec and drives zeeman transitions .
this coupling has been studied both theoretically @xcite and in a first experiment involving atoms in a vapor cell @xcite ( see below ) .
figure [ fig : magcoupschem ] shows the situation considered in @xcite with the transitions of @xmath247 atoms involved .
the magnetic field @xmath248 of the nanomagnet transduces mechanical oscillations of the cantilever into magnetic field oscillations at the position of the atoms .
the amplitude of the field oscillations is linked to the cantilever oscillation @xmath249 via the field gradient @xmath250 , @xmath251 the interaction of the atomic spin with @xmath252 is described by the zeeman hamiltonian @xmath253 where @xmath254 is the magnetic moment and @xmath255 the atomic spin .
the field direction of @xmath248 is chosen orthogonal to the field in the trap center @xmath256 set by an external source , such that the oscillations can drive spin - flip transitions between neighboring zeeman sublevels ( see fig .
[ fig : magcoupschem ] b ) .
the spin - flip transition rate becomes sizable when the field oscillations at frequency @xmath8 are resonant with the larmor frequency @xmath257 of the atoms in the trap , set by the value of @xmath258 in the trap center .
since @xmath258 can be easily varied experimentally , the resonance condition @xmath259 can be fulfilled for a large range of cantilever frequencies .
furthermore , by rapidly changing @xmath258 , the coupling can be switched on and off .
( a ) schematic of the magnetic coupling of a nanomechanical resonator to a single atom or a bec .
a nanomagnet on the tip of the cantilever transduces mechanical motion into magnetic field oscillations .
figure adapted from @xcite .
( b ) ground state levels of @xmath260rb , magnetically trappable states are indicated in red .
magnetic field oscillations at the larmor frequency @xmath261 lead to a coupling @xmath143 between different zeeman sublevels .
( c ) coupling via two - photon transitions with an additional microwave field permits to connect clock states with identical ( dashed ) or vanishing ( solid ) magnetic moment .
, scaledwidth=90.0% ] the magnet can be either a single domain ferromagnet with the magnetic moment @xmath262 spontaneously oriented along its long axis due to the shape anisotropy , or a grain of a permanent magnet .
approximating the magnet by a magnetic dipole , we have @xmath263 . by changing @xmath9 and the magnet dimensions
, @xmath250 can be adjusted .
equation ( [ eq : hz ] ) suggests that the strength of the atom - nanoresonator coupling can be maximized by increasing @xmath250 as much as possible .
however , the atoms experience a force in this field gradient , and an excessively large @xmath250 would strongly distort the trapping potential , leading to a similar limitation as discussed for the direct mechanical coupling in sec .
[ sec : directcoupling ] . to mitigate distortion ,
two compensation magnets can be placed next to the coupling magnet with identical direction of magnetization .
this reduces the static field gradient at the location of the atoms , while the oscillatory field @xmath252 remains unaffected as the compensation magnets do not oscillate .
it was already demonstrated @xcite that preparation of arbitrary quantum states in a harmonic oscillator can be achieved by a switchable , linear coupling to a two - level system with the law - eberly protocol @xcite . in the present scheme ,
a single atom is an effective two - level system if other transitions are sufficiently detuned .
this can be achieved intrinsically via the quadratic zeeman effect , or controllably via state - selective microwave dressing @xcite .
transitions between two states @xmath264 and @xmath265 of the atom can be described by creation and annihilation operators @xmath266 and @xmath267 , and with the linear coupling to the mechanical oscillator , the system is analog to the well - known jaynes - cummings model described by the hamiltonian @xcite @xmath268 where @xmath269 is the single - atom single - phonon coupling constant for the state pair @xmath270 . in the field of cavity qed , eq .
( [ eq : hjc ] ) usually describes the coupling of an atom to the electromagnetic field of a single mode of an optical or microwave cavity @xcite . here , it describes the coupling to the phonon field of a single mode of mechanical oscillations . in this sense , this is a mechanical cavity qed system . in the case of an ensemble of atoms coupling collectively to the oscillator
, the description is given by the tavis - cummings model @xcite .
this system can achieve the strong coupling regime both for a single atom and for a small cloud of atoms .
a single @xmath247 atom trapped at a distance @xmath271 nm from a cantilever with @xmath272 m and @xmath273mhz carrying a single domain co nanomagnet of dimensions @xmath274 nm achieves a coupling rate @xmath275hz .
the expected decoherence rate of the atoms depends strongly on the involved magnetic levels . for coherence times on the order of seconds , magnetic sublevels with identical ( @xmath276 ) or vanishing ( @xmath277 ) magnetic moment
have to be used @xcite , see fig .
[ fig : magcoupschem ] c. the latter state pair also avoids trap distortion due to the coupling magnet , but it requires an optical or electrodynamical trap to hold the atom @xcite . to connect states in different hyperfine manifolds ,
a two - photon process is required with an additional strong microwave field at @xmath278 .
if the parameters are adjusted properly , the two - photon coupling reduces the effective coupling rate only by a factor of @xmath279 . a sufficiently low decoherence rate of the cantilever to achieve the strong coupling condition @xmath280 again requires a low bath temperature @xmath82mk and a very high quality factor @xmath83 .
similar to the previous discussion for the nanotube in sec .
[ sec : cntcoupling ] , cryogenic ground state cooling of the cantilever is not available .
further cooling could be achieved by a state swap with a bec of @xmath87 atoms collectively coupled to the oscillator .
excitations of the oscillator can be coherently transferred to the atoms prepared in state @xmath281 , such that after an interaction time @xmath282 the oscillator is in a state with reduced mean phonon number . alternatively , by making use of atomic transitions between the two hyperfine manifolds @xmath283 and @xmath284 with a transition frequency @xmath285ghz , one could employ ghz mechanical oscillators , for which cryogenic ground state cooling can be achieved @xcite .
based on the magnetic coupling mechanism described above , a scalable experimental setup was proposed in @xcite , where the atoms are trapped in an optical lattice and each lattice site couples to an individual mechanical oscillator .
this could be used for atomic state manipulation with single lattice site resolution , for the operation of quantum gates , or for the entanglement of distant mechanical oscillators @xcite .
a magnetic interaction was also used in the first experiment that demonstrated a coupling between a mechanical oscillator and an atomic system @xcite . there , a microfabricated cantilever with a magnetic tip was excited to coherent motion and coupled to a gas of rb atoms in a vapor cell .
the oscillating magnetic field of the tip causes spin precession of the atoms , which can be detected as an intensity modulation of a detection laser .
this scheme can serve for magnetic field sensing , and a sensitivity of @xmath286nt@xmath287 was demonstrated .
finally , magnetic interactions can also be used to couple a mechanical oscillator to a solid - state `` artificial atom '' such as a nitrogen - vacancy center ( nv ) in diamond @xcite .
the advantage compared to ultracold trapped atoms is that a large coupling strength @xmath288khz can be achieved because the distance between the magnetic tip and the nv center can be made very small ( @xmath289 nm ) and distortion of trapping potentials is not an issue . to compete with the typically faster decoherence of the nv ground states ,
level dressing was proposed to tune the coupled system to a sweet spot with strongly suppressed magnetic field sensitivity and thus reduced decoherence rate , sufficient to achieve the strong coupling regime @xcite .
ultracold atoms coupled to mechanical oscillators represent a novel type of hybrid mechanical system that can benefit from the exceptional coherence properties and the versatile techniques for quantum manipulation of trapped atoms .
this could enable readout , cooling , and coherent manipulation of mechanical systems via the well - developed tools of atomic physics .
a multifaceted set of theoretical proposals shows that several different approaches have the potential to achieve strong coherent coupling between atoms and micro- or nanostructured oscillators .
the central challenge is that the overall dynamics of the system is relatively slow , so that exceptional coherence properties of the mechanical oscillator are required .
however , in light of the recent experimental progress with high - quality mechanical oscillators at low temperatures it seems feasible that this challenge can be met . experimental work on coupled atom - oscillator systems is still at an early stage , but first experiments have already demonstrated different coupling mechanisms @xcite . moreover , some of the proposed schemes can be implemented with moderate technical effort @xcite and enable the observation of backaction of the atoms onto the mechanical oscillator already in room temperature setups .
coupling schemes based on high - finesse optical cavities promise access to a regime of strong coherent coupling with state - of - the - art experimental parameters @xcite , and thus offer a realistic perspective for experiments on the quantum physics of massive mechanical objects .
we acknowledge fruitful collaborations and discussions with j. reichel , j. p. kotthaus , d. knig , k. hammerer , m. wallquist , c. genes , k. stannigel , p. zoller , m. ludwig , f. marquardt , h. j. kimble , j. ye , p. rabl , m. lukin , and d. leibfried .
this work was supported by the excellence cluster nanosystems initiative munich ( nim ) and by the european community s seventh framework programme through the integrating project aqute . | in this article we discuss and compare different ways to engineer an interface between ultracold atoms and micro- and nanomechanical oscillators .
we start by analyzing a direct mechanical coupling of a single atom or ion to a mechanical oscillator and show that the very different masses of the two systems place a limit on the achievable coupling constant in this scheme .
we then discuss several promising strategies for enhancing the coupling : collective enhancement by using a large number of atoms in an optical lattice in free space , coupling schemes based on high - finesse optical cavities , and coupling to atomic internal states . throughout the manuscript
we discuss both theoretical proposals and first experimental implementations .
= 1 ultracold atoms , micro- and nanomechanical oscillators , hybrid quantum systems , cavity optomechanics , bose - einstein condensate , ultracold ions , optical lattice , optical cavity |
although the classical random walk problem is rather well understood by now @xcite , some of its modifications , related , for example , to disordered systems , are less clear @xcite .
departure from the classical random walk manifests often as an anomalous diffusion that appears whenever the typical square displacement of a tracer particle @xmath6 does not obey the normal diffusion law @xmath7 with @xmath8 .
deviations include numerous examples of subdiffusion ( @xmath9 ) such as , for example , photoconductivity of amorphous materials @xcite , diffusion in convective rolls @xcite , or a random walk on the percolation cluster @xcite . on the other hand
, there is much rarer superdiffusive ( @xmath10 ) behavior , of which the richardson diffusion in turbulent fluids is a classical example @xcite .
some macromolecules or biological systems also exhibit anomalous dynamics with superdiffusive behavior , as reported , e.g. , for micelles @xcite or living cell migrations @xcite .
recently , a number of studies have indicated that foraging or hunting strategies of some animals also show superdiffusive behavior and examples include ants @xcite , fruit flies @xcite , or wandering albatrosses @xcite .
the prevailing understanding is that while searching for food these animals move in steps of varying length with fat power - law distribution consistent with the lvy flights .
however , some other works show that this hypothesis should be taken with care . indeed , re - analyzing the data for the wandering albatrosses , edwards _ et al .
_ concluded that the previously suggested lvy - flight scenario appeared due to a spurious inclusion of very long flights @xcite .
it has also been claimed that the idea of animals using stochastic but essentially blind search strategies does not seem to be viable since it neglects the role of animals intelligence and experience in guiding them @xcite .
nevertheless , some more extensive recent studies for 14 species of open - ocean fish of prey seem to support the lvy - flight scenario @xcite . in the present article
we show that a simple search strategy , i.e. : go to the nearest ( intelligence ) and not yet visited ( experience ) place , leads to an interesting diffusive behavior .
in particular , a normal diffusion that takes place for @xmath1 and 4 systems is modified by some logarithmic corrections in the @xmath4 case , which suggests that this is the critical dimension .
the distribution of distance does not have fat power - law tails , however , when the search exploits all available places , much broader and possibly lvy - like distributions can be found .
we consider a traveling salesman @xcite that has to visit cities which are uniformly distributed in the mensional space of linear size @xmath11 .
we assume a unit density of cities , i.e. , @xmath12 , where @xmath13 is the number of cities .
the salesman is not interested in finding the minimal tour but visits cities according to the following greedy algorithm : in each step , from the last visited city it moves to the nearest city which has not been visited yet . for most simulations ,
the starting point is the city placed in the center of the space , i.e. , of the cartesian coordinates @xmath14 .
the quantity of main interest is the average square distance @xmath2 from the starting point and especially its asymptotic form after many steps @xmath3 .
an example of a typical trajectory in the two - dimensional case is shown in fig .
let us emphasize that for a given starting point and a distribution of cities , the route of the salesman in our model is completely deterministic even though it resembles a random walk trajectory .
our model might be considered a version of the so - called deterministic random walk @xcite .
cities ( dots ) and duration @xmath15 .
[ conf],width=340 ] to examine the asymptotic behavior of @xmath2 , it is essential to average over many configurations with large number of cities @xmath13 so that a relatively large part of the trajectory of the salesman was not affected by the boundaries .
because of these finite - size effects , we terminate the generation of trajectories after number of steps that is much smaller than @xmath13 . as a result ,
the @xmath3-th city on a trajectory is typically a close neighbour of the ( @xmath31)-th city . instead of searching for the nearest
not yet visited city in the entire configuration , it would be much faster to restrict the search to the cities which are in the vicinity of the last visited city
. such a modification can be implemented by means of the sectorization of the space .
in addition to coordinates , we store the information about a sector to which a given city belongs .
then , to find the @xmath3-th city , we search the sectors surrounding the ( see fig . [ sectors ] ) .
let us emphasize that the above described sectorization does not introduce any approximation as for a given configuration of cities the algorithm generates exactly the same trajectory as the one without sectorization . in our simulations we used such mesh that each sector contained approximately 10 cities .
( in the two - dimensional case , where the longest trajectories had to be generated , the sectorization reduced the simulation time by the factor @xmath16 . ) , the program searches for available cities within @xmath17 .
( i ) if it finds some , then to find the nearest city it is sufficient to examine @xmath17 and its neighbouring sectors as indicated in ( a ) .
this is because any city outside these sectors is further from any point in @xmath17 than the length of the diagonal of @xmath17 ( which is the largest distance within @xmath17 ) .
( ii ) if there are no available cities in @xmath17 , the program searches for them in a larger area ( the bold square in ( b ) ) .
if it finds some , then to find the nearest city it is sufficient to examine sectors as indicated in ( b ) .
( iii ) if no cities are found , the program examines even larger areas ( until some cities are found ) .
[ sectors],width=340 ] for several values of @xmath13 indicated in the right bottom .
the straight line corresponds to the linear increase @xmath18 . for each @xmath13 ,
the plotted results are averages over at least @xmath19 independent configurations .
[ walklog],width=340 ]
using the above algorithm , we simulated systems containing up to @xmath20 cities .
results for the two - dimensional case are shown in fig .
[ walklog ] .
since the asymptotic slope seems to be bigger than unity , these data show a slightly superdiffusive behavior , but more quantitative analysis of the slope would be desirable .
one should take into account that the asymptotic regime in the numerical data will be visible only in an intermediate interval since both the initial transit as well as large-@xmath3 data should be neglected ( for long trajectories @xmath2 is strongly affected by finite - size effects ) . to estimate @xmath21 , we first examine the time - dependent quantity @xmath22 defined as the local exponent estimated from data spanning over one decade and centered at @xmath3 ( i.e. , to estimate @xmath22 we used @xmath2 calculated for @xmath23 such that @xmath24 ) . calculated in the two - dimensional case for several values of @xmath13 .
[ wykl2],width=340 ] the obtained results are presented in fig .
[ wykl2 ] .
as expected , for large @xmath3 all curves bent downward and in the middle range a plateau emerges .
our estimate based on data around @xmath25 suggests that @xmath26 .
however , such an estimation is very close to unity and in addition it seems to slowly decrease for increasing @xmath3 .
such a slow decrease suggests that asymptotically @xmath21 might be equal to unity but the power - law scaling is affected by logarithmic corrections . to examine such a possibility
, we plotted @xmath27 vs. @xmath28 $ ] ( fig .
[ walkcorr ] ) .
the asymptotic linearity of the data confirms the presence of the logarithmic corrections and , since the slope is close to unity , we obtain @xmath29 . vs. @xmath28 $ ] calculated in the two - dimensional case .
the dashed straight line has a unit slope .
[ walkcorr],width=340 ] calculated in the three - dimensional case for several values of @xmath13 .
[ wykl3],width=340 ] we used the same technique in the three- and four - dimensional versions , and the final results are shown in fig .
[ wykl3 ] and fig .
[ wykl4 ] , respectively . in these higher - dimensional cases ,
the linear size @xmath11 increases slower with the number of cities @xmath13 , and that is why the curves bent downward for smaller values of @xmath3 .
the emerging plateaux indicate that @xmath21 is very close to unity both for @xmath1 and 4 .
let us also notice that in the @xmath30 case the salesman , possibly after some initial switches , will stick to one side and then move ballistically in the same direction , trivially implying @xmath31 .
calculated in the four - dimensional case for several values of @xmath13 .
[ wykl4],width=340 ] of intervals @xmath32 calculated for paths of the length @xmath33 .
the inset shows that the variability of the average interval @xmath34 ( at time @xmath3 ) during such runs is rather small .
[ distlog],width=340 ] for the @xmath4 and 3 cases , we also analyzed the distributions of the length @xmath32 between successive cities on a trajectory .
one should take into account , however , that these distributions are not entirely stationary and to some extent vary in time @xmath3 ( in general , very long distances appear in the late - time evolution and they could be of the order of the system size @xmath11 in the final stage ) . to reduce this variability , we generated relatively short trajectories ( most likely none of them reached the boundary of the system ) .
the logarithmic plot of these distributions ( fig . [ distlog ] ) shows that , in the @xmath4 version , the distribution is much broader than in the @xmath1 version , and perhaps of a qualitatively different large-@xmath32 tail .
let us notice that in both cases @xmath34 is similar and of the order of unity ( inset in fig .
[ distlog ] ) but the recorded fluctuations of @xmath32 are several times larger in the @xmath4 case .
let us also observe that large-@xmath32 tails of these distributions do not seem to have a power - law form .
this indicates that the logarithmic corrections seen in the @xmath4 case are not due to sufficiently broad ( and perhaps lvy - like ) distributions of lengths @xmath32 but rather due to the development of long - time correlations .
we will return to this point in the final part of the article . as in fig .
[ distlog ] but in the log - log scale .
the inset shows the log - log plot of @xmath35 for runs where all cities are visited ( @xmath36 ) .
the dotted line has the slope -3 ( @xmath37 ) .
[ distloglog],width=340 ] the log - log plot of these distributions confirms the absence of power - law tails for large @xmath32 ( fig . [ distloglog ] ) .
it turns out that the other side of these distributions , namely the small-@xmath32 tails , can be understood with some simple arguments .
indeed , these tails do not differ much from the small-@xmath32 tails in the initial distribution of cities . for simplicity ,
let us examine first the @xmath30 case , where @xmath131 cities are uniformly distributed on the interval of the length @xmath38 and one city is placed in the center of the interval ( i.e. , it has the coordinate @xmath39 ) .
calculating the probability density @xmath40 that exactly one city is at the distance @xmath41 from the central city ( to the right or left ) and @xmath132 cities are further than @xmath41 from it , in the limit @xmath42 , we obtain @xmath43 . generalizing the above calculations to the @xmath4 and @xmath1 cases , we obtain @xmath44 and @xmath45 , respectively . as can be seen in fig .
[ distloglog ] , the small-@xmath32 tails are indeed close to the tails of the initial distributions both for @xmath4 and 3 .
one can also notice that for large @xmath41 the initial distributions @xmath46 and @xmath47 decay very fast ( faster than exponentially ) while the corresponding distributions @xmath35 decay slower ( see fig . [ distlog ] ) .
thus , the traveling salesman generates paths where short distances appear with approximately the same distribution as in the initial distribution but long distances are generated much more frequently .
it is perhaps interesting , especially in the context of modeling animal search strategies , that the distributions @xmath35 become much broader when calculations are made on paths that visit all the cities ( inset in fig .
[ distloglog ] ) .
this is because in the final part of the tour the salesman has to visit all the `` leftovers '' that are typically quite distant from each other .
these numerical data suggest that in the two - dimensional data the tail of the resulting distribution might even have the power - law decay , @xmath48 and @xmath49 , and thus could be consistent with the limiting lvy distribution .
let us emphasize , however , that , for paths that visit all the cities , the generated sequence of lengths is nonstationary ( early- and late - stage statistics differ very much ) , and assembling the data into a single distribution is rather questionable .
let us also note that there are some claims that the lvy distributions in the foraging data appeared most likely due to a spurious inclusion of late - stage flights @xcite .
these flights , e.g. , of albatrosses returning to their nests , were obviously very long but should not be considered a part of the searching pattern .
in this section we briefly summarize the overall behavior of our model and comment on possible relations with some other statistical - mechanics models .
the memory of the salesman restricts its movements and effectively pushes the salesman out of regions that have already been visited . for @xmath30
, this strategy trivially leads to the superdiffusive behavior with @xmath31 . for @xmath1 and 4
, the long - time memory turns out to be basically irrelevant and an ordinary diffusion ( @xmath8 ) is restored .
the dimension @xmath4 seems to be the critical dimension separating these two regimes and the logarithmic corrections are an expected feature in such a case .
let us note that a similar memory mechanism generates a superdiffusive behavior in the so - called self - avoiding walk problem @xcite .
however , this model of a polymer is known to be superdiffusive for @xmath30 ( @xmath31 ) , @xmath4 ( @xmath50 ) , and @xmath1 ( @xmath51 ) , with the normal diffusion resuming for @xmath52 .
our model seems to be related to yet another model of the self - avoiding walk , introduced by amit _
@xcite , where the walker tries to avoid lattice sites that it has already visited .
renormalization - group arguments show that for such a model the critical dimension equals to 2 and the logarithmic corrections are present @xcite in the critical dimension .
it is likely that , at the coarse - grained level , our model is described by the same field theory as the model of amit _ et al . _
its direct applicability to search strategies is rather limited but our model might indicate that the very dimension of the search space plays an important role .
in particular , it suggests that much broader distributions should appear in the two - dimensional rather than in the three - dimensional search spaces . since
a horizontal span of the search space for albatrosses or sharks is much larger than a vertical one , perhaps this space could be considered as effectively two - dimensional and , as our work suggests , broad distributions might appear in such a case .
thus it would be interesting to examine species for which a search space is less anisotropic .
g. h. weiss , _ aspects and applications of the random walk _ , ( north holland 1994 ) .
bouchaud and a. georges , phys .
rep . * 195 * , 127 ( 1990 ) . h. pfister and h. scher , adv . phys . * 27 * , 747 ( 1978 ) .
y. pomeau , a. pumir , and w. r. young , phys .
fluids a * 1 * , 462 ( 1989 ) .
s. havlin and d. ben avram , adv . phys . * 36 * , 695 ( 1987 ) .
l. f. richardson , proc .
london a * 110 * , 709 ( 1926 ) .
r. metzler and j. klafter , phys .
* 339 * , 1 ( 2000 ) .
a. ott , j. p. bouchaud , d. langevin , and w. urbach , phys .
* 65 * , 2201 ( 1990 ) .
r. ganapathy , a. k. sood , and s. ramaswamy , europhys . lett . * 77 * , 18007 ( 2007 ) .
p. dieterich , r. klages , r. preuss , and a. schwab , proc .
105 , 459 ( 2008 ) .
m. f. shlesinger , in _ on growth and form _
( eds . h. e. stanley and n. ostrowsky ) ( nijhoff , dordrecht , 1986 ) .
b. cole , j. anim .
* 50 * , 1317 ( 1995 ) .
g. m. viswanathan , v. afanasyev , s. v. buldyrev , e. j. murphy , p. a. prince , and h. e. stanley , nature * 381 * , 413 ( 1996 ) .
a. m. edwards _ et al .
_ , nature * 449 * , 1044 ( 2007 ) .
j. travis , science * 318 * , 742 ( 2007 ) .
n. e. humphries _ et al .
_ , nature * 465 * , 1066 ( 2010 ) .
d. l. applegate , r. e. bixby , v. chvtal and w. j. cook , _ the traveling salesman problem : a computational study _ , princeton series in applied mathematics .
( princeton university press 2007 ) .
l. a. bunimovich , physica d * 187 * , 20 ( 2004 ) .
b. li , n. madras , and a. d. sokal , j. stat . phys .
* 80 * , 661 ( 1995 ) .
d. j. amit , g. parisi , and l. peliti , phys . rev .
b * 27 * , 1635 ( 1983 ) .
s. p. obukhov and l. peliti , j. phys .
a * 16 * , l147 ( 1983 ) . | using monte carlo simulations we examine the diffusive properties of the greedy algorithm in the @xmath0-dimensional traveling salesman problem .
our results show that for @xmath1 and 4 the average squared distance from the origin @xmath2 is proportional to the number of steps @xmath3 . in the @xmath4 case
such a scaling is modified with some logarithmic corrections , which might suggest that @xmath4 is the critical dimension of the problem .
the distribution of lengths also shows marked differences between @xmath4 and @xmath5 versions .
a simple strategy adopted by the salesman might resemble strategies chosen by some foraging and hunting animals , for which anomalous diffusive behavior has recently been reported and interpreted in terms of lvy flights .
our results suggest that broad and lvy - like distributions in such systems might appear due to dimension - dependent properties of a search space . |
natural processes proceed simultaneously on many different scales interacting with each other .
this is especially important in astrophysics , where an astronomical observation or a hypothesized system can only be understood by modeling processes on scales spanning many orders of magnitude .
thus , also nuclear physics and astrophysics are closely entwined .
nuclear reactions power quiescent burning of stars and cause the most powerful explosions known .
they not only release or transform energy but also change the composition of the matter in which they occur and thus are responsible for the range of chemical elements found on our planet and throughout the universe . finally , to understand the properties of matter at extreme density and/or density , the nuclear equation of state has to be known .
it determines the properties of neutron stars and the latest stages of the life of stars with more than @xmath1 which end their lives in a supernova explosion .
it is evident that nuclear physics input is essential for many astrophysical models and this fact is represented in the field of nuclear astrophysics .
there are different perceptions on how to define this field .
some limit it to the application of nuclear physics to reactions of astrophysical interest .
a more comprehensive , and perhaps more adequate , definition includes the more astrophysical aspects in the investigation of nuclear processes and nucleosynthesis in astrophysical sites through reaction networks . in any case , the interests of astrophysics emphasize different aspects than those of basic nuclear physics and this makes nuclear astrophysics a distinct research area . from the nuclear point of view , astrophysics involves low - energy reactions with light projectiles on light , intermediate , and heavy nuclei .
although nuclear physics has moved to higher energies in the last decades , low - energy reactions are not well enough explored and still offer considerable challenges to both experiment and theory , even for stable target nuclei .
explosive conditions in astrophysics favor the production of nuclei far off stability and reaction rates for these have to be predicted across the nuclear chart .
this proves very difficult , especially for first - principle methods , due to the complexity of the nuclear many - body problem .
the calculation of astrophysical reaction rates also includes special requirements and processes not studied in nuclear physics so far . among these
are effects appearing in low energy , subcoulomb reactions and reactions on excited states of target nuclei .
depending on the conditions , plasma effects also have to be considered because these alter the reaction rates .
this includes , e.g. , the shielding of coulomb barriers through electrons , pycnonuclear burning in the lattice of a high - density plasma , and the modification of nuclear partition functions at very high plasma temperatures .
this paper focuses on nuclear reactions in astrophysics and , more specifically , on reactions with light ions ( nucleons and @xmath0 particles ) proceeding via the strong interaction .
it is intended to present the basic definitions essential for studies in nuclear astrophysics , to point out the differences between nuclear reactions taking place in stars and in a terrestrial laboratory , and to illustrate some of the challenges to be faced in theoretical and experimental studies of those reactions .
the sensitivity of the reaction rates to the uncertainties in the prediction of various nuclear properties is explored and some guidelines for experimentalists are also provided .
the discussion revolves around the relevant quantities for astrophysics , which are the astrophysical reaction rates .
the impact of using different models or data is always presented with respect to the possible modification of the reaction rates , not the cross sections . at first , the basic equations through which nuclear processes enter astrophysical models are introduced in sec .
[ sec : netrate ] .
the astrophysically relevant energy ranges are defined in sec .
[ sec : energies ] .
the thorough discussion of the special stellar effects affecting reactions in sec .
[ sec : stellareffects ] is the heart of this review .
stellar and effective cross sections are derived in sec .
[ sec : stellar ] , then the relation between laboratory reactions and those occurring in a stellar plasma is investigated in sec .
[ sec : stellarexp ] .
as it is important to realize when individual reactions are important and when they are not , reaction equlibria are introduced in sec . [
sec : equilibria ] . cross sections in a plasma
are affected by the free electrons present which shield the nuclear charge .
this is explained in sec .
[ sec : electro ] .
finally , astrophysically relevant reaction mechanisms are reviewed in sec .
[ sec : mech ] and reactions through isolated resonances ( sec .
[ sec : reso ] ) , in systems with high nuclear level density ( sec .
[ sec : statmod ] ) , and direct reactions ( sec .
[ sec : direct ] ) are discussed separately .
this includes a detailed discussion of the sensitivities of the rates on the required input in secs .
[ sec : relevance ] and [ sec : directsensi ] .
section [ sec : conclusion ] presents a brief conclusion .
nuclear reactions are the engine of stellar evolution and determine the overall production of the known chemical elements and their isotopes in a variety of nucleosynthesis patterns .
a detailed understanding of the characteristic production and depletion rates of nuclei within the framework of a nucleosynthesis process is crucial for reliable model predictions and the interpretation of the observed abundances . instead of the number @xmath2 of nuclei of a given species per volume @xmath3 ( the number density @xmath4 ) , it is advantageous to use a quantity independent of density changes : the abundance @xmath5 , where @xmath6 is the plasma density and @xmath7 denotes avogadro s number .
the change of abundances @xmath8 with time due to nuclear processes is traced by coupled differential equations . due to the nature of the involved reactions and the vastly different timescales appearing , the equation system is non - linear and stiff .
in addition , for complete solubility of the coupled equations the number of equations @xmath9 has to equal the number of involved nuclei acting as reaction partners and thus an equation matrix of size @xmath10 has to be solved .
nucleosynthesis processes include thousands of nuclides and tens of thousands reactions .
this still makes it impossible to fully couple such a reaction network to a full set of hydrodynamic equations as would be required for a complete modeling of nucleosynthesis in a given astrophysical site .
a reaction network generally can be written as @xcite @xmath11 where @xmath12 numbers the nucleus , @xmath13 is the @xmath14th rate for destruction or creation of the @xmath15th nucleus without a nuclear projectile involved ( this includes spontaneous decay , lepton capture , photodisintegration ) , and @xmath16 is the rate of the @xmath14th reaction involving a nuclear projectile and creating or destroying nucleus @xmath15 .
similarly , we have three - body reactions where nucleus @xmath15 is produced or destroyed together with two other ( or similar ) nuclei .
reactions with more participants ( denoted by @xmath17 above ) are unlikely to occur at astrophysical conditions and are usually neglected .
the quantities @xmath18 , @xmath19 , and @xmath20 are positive or negative integer numbers specifying the amount of nuclei @xmath15 produced or destroyed , respectively , in the given process . as shown below , the rates @xmath21 , @xmath22 , and @xmath23 contain the abundances of the interacting nuclei .
rates of type @xmath21 depend on one abundance ( or number density ) , rates @xmath22 depend on the abundances of two species , and rates @xmath23 on three . the rates @xmath13 appearing in the first term of eq .
( [ eq : network ] ) are reactions per time and volume , and only contain the abundance @xmath24 .
for example , @xmath25 is simply @xmath26 for @xmath27-decays .
the factor @xmath28 is the usual decay constant ( with the unit 1/time ) and is related to the half - life @xmath29 of the decaying nucleus @xmath14 .
it has to be noted that some decays depend on the plasma temperature and thus @xmath30 is not always constant , even for decays ( see eq .
( [ eq : stelldecay ] ) in sec .
[ sec : stellar ] ) .
two - body rates @xmath22 include the abundances of two interacting particles or nuclei .
in general , target @xmath31 and projectile @xmath32 follow specific thermal momentum distributions @xmath33 and @xmath34 in an astrophysical plasma . with the resulting relative velocities @xmath35 ,
the number of reactions per volume and time is given by @xmath36 and involves the reaction cross section @xmath37 as a function of velocity , the relative velocity @xmath35 and the thermodynamic distributions of target and projectile @xmath33 and @xmath34 .
the evaluation of this integral depends on the type of particles ( fermions , bosons ) and distributions which are involved . however , many two - body reactions can be simplified and effectively expressed similarly to one - body reactions , only depending on one abundance ( or number density ) .
if reaction partner @xmath32 is a photon , the relative velocity is always @xmath38 and the quantities in the integral do not depend on @xmath34 .
this simplifies the rate expression to @xmath39 where @xmath40 stems from an integration over a planck distribution for photons of temperature @xmath41 .
this is similar to the decay rates introduced earlier and therefore we replaced @xmath22 by @xmath21 in our notation and can include this type of reaction in the first term of eq .
( [ eq : network ] ) .
a similar procedure is used for electron captures by protons and nuclei . because the electron is about 2000 times less massive than a nucleon , the velocity of the nucleus is negligible in the center - of - mass system in comparison to the electron velocity ( @xmath42 ) .
the electron capture cross section has to be integrated over a fermi distribution of electrons .
the electron capture rates are a function of the plasma temperature @xmath41 and the electron number density @xmath43 . in a neutral , completely ionized plasma ,
the electron abundance @xmath44 is equal to the total proton abundance @xmath45 and thus @xmath46 again , we have effectively a rate per target @xmath47 ( with unit 1/time ) similar to the treatment of decays earlier and a rate per volume including the number density of only one nucleus .
we denote the latter by @xmath21 and use it in the first term of eq .
( [ eq : network ] ) .
this treatment can be applied also to the capture of positrons , being in thermal equilibrium with photons , electrons , and nuclei .
furthermore , at high densities ( @xmath48g@xmath49 ) the size of the neutrino scattering cross section on nucleons , nuclei , and electrons ensures that enough scattering events occur to lead to a continuous neutrino energy distribution
. then also the inverse process to electron capture ( neutrino capture ) can occur as well as other processes like , e.g. , inelastic scattering , leaving a nucleus in an excited state which can emit nucleons and @xmath0 particles .
such reactions can be expressed similarly to photon and electron captures , integrating over the corresponding neutrino distribution . in the following ,
we focus on the case of two interacting nuclei or nucleons as these reactions will be extensively discussed in the following sections .
( we mention in passing that eq .
( [ eq:2body ] ) can be generalized to three and more interacting nuclear species by integrating over the appropriate number of distributions , leading to rates @xmath23 and higher order terms in eq .
( [ eq : network ] ) . ) the velocity distributions in the rate definition in eq .
( [ eq:2body ] ) can be replaced by energy distributions .
furthermore , it can be shown that the two distributions can be replaced by a single one in the center - of - mass system.@xcite then the two - body rate @xmath22 is defined as an interaction of two reaction partners with an energy distribution @xmath50 according to the plasma temperature @xmath41 and a reaction cross section @xmath51 specifying the probability of the reaction in the plasma : @xmath52 the factor @xmath53 with the kronecker symbol @xmath54 is introduced to avoid double counting .
the nuclear cross section @xmath55 is defined as in standard scattering theory by @xmath56 however , in an astrophysical plasma reactions not only proceed on the ground state of a nucleus but also from excited states .
this is implied in the notation for the modified cross section @xmath57 , contrasting the usual laboratory cross section ( denoted by @xmath55 or @xmath58 ) for reactions acting only on the ground state of the target nucleus .
the implications of using such a cross section modified in the stellar plasma , instead of the usual laboratory one , will be discussed in sec .
[ sec : stellar ] . it should be noted that @xmath59 may be a function not only of energy but also of plasma temperature .
the distribution of kinetic energies of nuclei in an astrophysical plasma with temperature @xmath41 follows a maxwell - boltzmann distribution ( mbd ) @xmath60 and we obtain finally : @xmath61 here , @xmath62 denotes the reduced mass of the two - particle system and @xmath63 is the reaction rate per particle pair or _
reactivity _ under stellar conditions .
the angle brackets stand for the appropriate averaging , i.e. integration , over the energy distribution .
for the remainder of the paper we will be concerned with the determination of this reactivity and the involved cross sections , respectively .
before we proceed to the details of the determination of the reaction cross sections , it is instructive to further investigate the rate equation and to derive the relevant energies at which the nuclear reaction cross sections have to be known .
although the integral in eq . ( [ eq : rate ] ) runs to infinity , the mbd folded with the cross section selects a comparatively narrow energy range with non - negligible contributions to the total value of the integral .
historically , this energy range is called the gamow window because gamow realized early on the astrophysical relevance of the fact that if the energy dependence of the cross section is dominated by the coulomb barrier between the projectile and the target the integrand in eq .
( [ eq : rate ] ) can be factorized as@xcite @xmath64 where @xmath65 is the astrophysical @xmath65-factor @xmath66 which is assumed to be only weakly dependent on the energy @xmath67 for non - resonant reactions .
the second exponential in eq .
( [ eq : penet ] ) is called the gamow factor and contains an approximation of the coulomb penetration with the sommerfeld parameter @xmath68 where @xmath69 , @xmath70 are the charges of projectile @xmath32 and target @xmath31 , respectively , and @xmath62 is their reduced mass . while the first exponential ( the tail of the mbd ) decreases with increasing energy , the gamow factor increases , leading to a confined peak of the integrand , the so - called gamow peak .
the location of the peak @xmath71 is shifted to higher energies with respect to the maximum of the mbd at @xmath72 ( @xmath73 mev when @xmath74 is the plasma temperature in gk ) .
the width of the peak gives the astrophysically relevant energy range in which most of the reactions will take place . in absence of a coulomb barrier the energy dependence of the non - resonant cross section
is roughly given by the one of the wave number of the particle ( @xmath75 ) folded with the angular momentum barrier .
this does not , however , lead to a relevant shift of the peak of the integrand compared to the peak of the mbd .
thus , the effective energy window for neutrons is simply the peak of the mbd .
the above considerations concerning the location and size of the energy window have given rise to simple approximation formulae extensively used by experimentalists to estimate the energies of interest .
for instance , with a charged projectile the location @xmath76 and width @xmath77 of the gamow window is often computed from @xmath78 which is derived from eq .
( [ eq : penet ] ) assuming a gaussian shape of the peak and using appropriate numerical constants , yielding @xmath76 and @xmath77 in mev.@xcite this approximation of the gamow window @xmath79 is valid for some but not all cases because it is oversimplified .
the above factorization with the given sommerfeld parameter including the charges of projectile @xmath32 and target
@xmath31 implicitly assumes that the energy dependence of the cross section is given by the coulomb penetration in the entrance channel @xmath80 of the reaction @xmath81 .
it has been realized , however , that sometimes resonances below the gamow window derived with the above approximation significantly contribute to the reaction rate for certain capture reactions.@xcite in those considered cases the energy dependence of the cross section is dominated by the energy dependence of the @xmath82 width in the exit channel instead of the charged particle width in the entrance channel . this can be generalized@xcite and leads to the important realization that the energy dependence of the integrand @xmath83 has to be numerically examined in order to derive reliable energy windows .
this has been generally and extensively studied in ref . . here
, only a few examples are shown .
figure [ fig : sn112pa ] shows a comparison between the actual @xmath83 and the integrand assumed with the standard approximation . in this case , the relative shift of the energy window is to higher energy because of the higher coulomb barrier in the exit channel .
in other cases , the shifts can also be to much lower energy than predicted by the standard approximation
. for a detailed understanding of these differences one has to realize that both resonant and hauser - feshbach cross sections ( see secs . [ sec : reso ] , [ sec : statmod ] ) can be expressed as@xcite @xmath84 with @xmath85 being either breit - wigner widths or averaged hauser - feshbach widths , depending on the context .
the width of the entrance channel is given by @xmath86 , the one of the exit channel by @xmath87 , and the total width including all possible emission channels from a given resonance or compound state with spin @xmath88 by @xmath89 it has become common knowledge that a cross section of the form shown in eq . ( [ eq : cs ] ) is determined by the properties of the smaller width in the numerator if no other channels than the entrance and exit channel contribute significantly to @xmath90 ( see also sec . [ sec : sensigeneral ] ) .
then @xmath90 cancels with the larger width in the numerator and the smaller width remains .
( the effect is less pronounced and requires a more detailed investigation when other channels are non - negligible in @xmath90 . ) in consequence , the energy - dependence of the cross section will then be governed by the energy dependence of this smallest @xmath91 . only if this happens to be the charged - particle ( averaged ) width in the entrance channel , the use of the standard formula for the gamow window ( eqs .
[ eq : e0approx ] , [ eq : deltaapprox ] ) will be justified .
since @xmath92 and @xmath93 have different energy dependences , it will depend on the specific energy ( weighted by the mbd ) which of the widths is smaller .
therefore , the above approximation should not be applied blindly but rather the actual gamow windows have to be determined from the true energy dependence by inspection of the integrand in eq .
( [ eq : rate ] ) .
extensive tables of revised effective energy ranges for astrophysics from such a numerical inspection are given in ref . .
these ranges can be shifted by several mev to higher or lower energy compared to the ones obtained with the standard formula .
furthermore it is found that the assumption of a gaussian shape of @xmath83 is untenable for the majority of cases with intermediate and heavy target nuclei .
rather , the integrand @xmath83 may show a pronounced asymmetry around its maximum value .
therefore , the energy of the maximum alone is not sufficient to determine the astrophysically relevant energy range . although derived from cross sections of a specific model prediction , the energy windows given in ref .
are supposed to be robust .
this can be understood by realizing that they mainly depend on the relative energy dependence of the acting reaction channels and not the absolute value of the cross sections .
this dependence is governed by the relative energy and the coulomb barrier seen in each reaction channel .
therefore , the limits of the energy windows are set by the knowledge of the charges of the nuclei involved in the different reaction channels , and the reaction @xmath94-values
. only the latter may be unknown for nuclei far off stability and mass measurements may have an impact .
another important consequence of using the correct energy dependence is that different reactions may not have necessarily the same effective energy window , even when projectile and target nucleus are the same .
this is immediately seen when considering the case of a reaction with a positive @xmath94-value for one channel but a negative one for another reaction channel .
the astrophysically relevant energy window of the exothermic reaction may lie below the threshold of the other reaction channel .
obviously , the relevant energy window for the endothermic channel can not open below the threshold energy and thus has to lie at higher energies than the one for the capture .
a randomly chosen example for such a reaction pair would be @xmath95pd(p,@xmath82)@xmath96ag and @xmath95pd(p , n)@xmath95ag , with the energy windows at @xmath97 mev and @xmath98 mev , respectively , for a plasma temperature of 2 gk.@xcite below the ( p , n ) threshold at 5.06 mev , the ( p , n ) cross section is zero and does not give a contribution to the integral in eq .
( [ eq : rate ] ) .
this also further illustrates the limitation of the standard approximation which yields identical energy windows for reactions with identical entrance channels but different exit channels .
it has to be noted , though , that the effective energy windows only point out the energy range contributing mostly to the rate integral at given stellar temperature but do not make a statement on the size of the rate or its astrophysical relevance .
it has already been mentioned that eq .
( [ eq : cs ] ) applies to reactions either exhibiting isolated resonances treatable by a breit - wigner resonance formula or smooth cross sections stemming from an averaging over a large number of narrowly spaced and unresolved resonances .
therefore the derived energy windows are also applicable to obtain the relevant energy ranges in which narrow resonances have to be considered .
they do not , however , specify the relative strengths of the resonances within a given window . definition ( [ eq : sfactor ] ) only makes sense when using the laboratory cross section @xmath99 ( see eq .
( [ eq : labcs ] ) in the following section ) .
strictly speaking , with the laboratory cross section the energy windows apply to laboratory measurements only , i.e. to the determination of the ground state component of the actual stellar cross section . for low stellar temperatures and positive @xmath94-values ( see the discussion in sec .
[ sec : stellar ] ) , however , this will dominate the stellar cross section .
relevant energy windows can also be derived numerically for stellar cross sections ( see next section ) , of course , in the same manner . since these can not be measured ( yet ) , they may be of limited use , though . a general scrutiny of the astrophysically relevant energy windows up to a plasma temperature of 5 gk ( above this temperature , reaction equilibria are established which do not require the knowledge of individual rates ; see , e.g. , refs .
) shows that the appearing interaction energies are small by nuclear physics standards . for neutron - induced reactions ,
the encountered maximum energies are a few hundred kev , depending on the examined nucleosynthesis process ( e.g. , in the @xmath100 process they are more like 8 - 60 kev).@xcite the relevant energy windows are shifted to higher energies for charged reactants , with a few mev for reactions with protons and several mev up to about 10 mev for reactions involving @xmath0 particles .
the formulae given in eqs .
( [ eq : e0approx ] ) , ( [ eq : deltaapprox ] ) are inadequate for the determination of these energies and should not be used anymore .
in an astrophysical plasma , nuclei quickly ( on the timescale of nuclear reactions and scattering ) reach thermal equilibrium with all plasma components .
this allows thermal excitation of nuclei which follows a boltzmann law and gives rise to the _ stellar _ cross section @xmath101 where the sum runs over all excited states @xmath102 of the target nucleus @xmath31 ( for simplicity , here we assume the projectile @xmath32 , i.e. the second reaction partner , does not have excited states ) with spin factor @xmath103 and excitation energy @xmath104 .
thus , the stellar cross section is the sum of cross sections @xmath105 ( evaluated at their respective center - of - mass energies @xmath106 with @xmath107 when @xmath108 ) for reactions on a nucleus in excited state @xmath102 , weighted by the population coefficient @xmath109 .
the above relation can be derived from a saha equation@xcite .
it is to be noted that the stellar cross section depends on energy _ and _ temperature , contrary to the usual cross section which only is a function of energy .
the quantity @xmath110 is the partition function of the nucleus .
often , the partition function normalized to the ground state @xmath111 \end{split } \label{eq : partfuncint}\ ] ] is used ( the ground state is labeled as @xmath112 , the first excited state as @xmath113 , ) . equation ( [ eq : partfuncint ] ) shows how the computation can be extended beyond the energy of the highest known discrete energy level @xmath114 by using an integration of a nuclear level density @xmath115 over a range of excitation energies @xmath116 .
likewise , the sums appearing in ( [ eq : csstar ] ) can be amended with a supplemental integration over the level density above the last discrete level used .
although the product of the boltzmann factor @xmath117 and @xmath115 does not have a trivial energy dependence , it has been shown that for the application of ( [ eq : partfuncint ] ) at temperatures @xmath118 gk it is sufficient to integrate only up to @xmath119 mev.@xcite temperatures above 10 gk are encountered in some explosive astrophysical events , in accretion disks , and in the formation of neutron stars and black holes .
nuclear transformations in such environments are described in reaction equilibria between several or all possible reactions ( with the exception of reactions mediated via the weak interaction because they are too slow in most cases ) , replacing full reaction networks by simplified abundance equations ( see , e.g. , refs . for details ) but still containing the partition functions . at such high temperatures ,
a straightforward application of ( [ eq : partfuncint ] ) would overestimate the partition function because continuum effects have to be taken into account .
these can be treated by approximated correction factors to @xmath115 and extending the integration to @xmath120 mev.@xcite a more rigorous treatment of the correction would be desireable but the sheer number of involved nuclei proves prohibitive for fully microscopic approaches .
transitions ) in a compound reaction involving the nuclei a and f , and proceeding via a compound state ( horizontal dashed line ) with spin @xmath121 and parity @xmath122 in the compound nucleus c. the reaction @xmath94 values for the capture reaction ( q@xmath123 ) and the reaction a@xmath124f ( q@xmath125=q@xmath126 ) are given by the mass differences of the involved nuclei .
above the last state , transitions can be computed by integrating over nuclear level densities ( shaded areas ) .
[ fig : cnscheme],scaledwidth=80.0% ] the use of stellar cross sections in the calculation of reaction rates assures an important property , the reciprocity of forward and reverse rate .
a scheme of the energetics and the transitions between nuclear levels in the involved nuclei is shown in fig . [
fig : cnscheme ] . for a reaction @xmath127 involving only one initial level @xmath102 in nucleus @xmath31 and one final level @xmath128 in nucleus @xmath129 ( this can also be the compound nucleus @xmath130 if the ejectile @xmath131 is a photon ) the well - known reciprocity relation between forward ( @xmath132 ) and reverse ( @xmath133 ) reaction cross section is@xcite @xmath134 where @xmath135 , @xmath136 are the reduced masses , @xmath137 , @xmath138 the center - of - mass energies relative to the levels @xmath102 , @xmath128 , respectively , and @xmath139 the spin factors as before .
this relation connects one initial with one final state and therefore is not applicable to the regular laboratory reaction cross sections which connect one initial state ( usually the ground state @xmath112 ) of the target nucleus with a number of possible final states @xmath140 in order to obtain a cross section obeying reciprocity one has to construct a theoretical quantity called _ effective cross section _
the effective cross section is a sum over all energetically possible transitions of initial levels to final levels ( capture : in nuclei @xmath31 and @xmath130 , otherwise : in nuclei @xmath31 and @xmath129 ; as indicated in fig . [
fig : cnscheme ] ) , applying to compound reactions as well as to direct reactions ( see sec . [
sec : mech ] ) .
it includes all the transitions shown by arrows in fig .
[ fig : cnscheme ] and therefore sums over all final levels @xmath128 _ and _ initial levels @xmath102 @xmath142 as before , the relative center - of - mass energy of a transition proceeding from level @xmath102 is denoted by @xmath143 , and @xmath144 for @xmath108 . the first summand ( @xmath112 ) in the sum over @xmath102 is just the laboratory cross section @xmath145 .
note that the effective cross section is only a function of energy , like the usually defined cross section , and does not depend on temperature , contrary to the stellar cross section .
when interchanging the labels @xmath102 and @xmath128 , a similar quantity @xmath146 is obtained for the reverse direction commencing on levels @xmath128 .
it is straightforward to show that the two effective cross sections obey the reciprocity relations @xmath147 which are identical to the one in ( [ eq : reci_single ] ) for a single transition between two states in two nuclei .
the relative energies of the transitions proceeding on the ground states of the two target nuclei for forward and reverse reaction are denoted by @xmath148 and @xmath149 , respectively .
the effective cross section is , of course , unmeasureable .
its usefulness becomes apparent when we combine definition ( [ eq : csstar ] ) of the stellar cross section with the definition ( [ eq : rate ] ) for the astrophysical reaction rate .
since excited states of target nuclei are populated in a stellar plasma according to ( [ eq : csstar ] ) , we have to sum over the rates for reactions from each level and weight each summand with the population factor @xmath150 this means that projectiles with mb distributed energies are acting on each level @xmath102 separately .
insertion of definition ( [ eq : csstar ] ) for the population factor @xmath109 leads to @xmath151 where @xmath152 is the normalized partition function as defined in ( [ eq : partfuncint ] ) . in order to obtain an expression similar to the original single mbd , the integral
can be transformed by replacing @xmath153 , with @xmath154 and this yields @xmath155 } e^{-\frac{e_0^a}{kt}}\,de_0^a } = \nonumber \\ & = \frac{1}{g^a_0 } \sum_\mu { \int \limits_{0}^\infty \frac{g_\mu^a}{g_0^a } \sigma^\mu_{aa}(e_\mu^a ) \left[e_0^a - e_\mu^\mathrm{x}\right ] e^{-\frac{e_0^a}{kt}}\,de_0^a } \quad .
\label{eq : stellrate3}\end{aligned}\ ] ] in the last line above , the lower limit of the integration was reset to zero .
this is allowed because cross sections at negative energies do not give any contribution to the integral .
it has been pointed out in ref .
that it is mathematically equivalent when sum and integral are exchanged , leading to@xcite @xmath156 e^{-\frac{e_0^a}{kt}}\,de_0^a } = \nonumber \\ & = \left(\frac{8}{m_{aa } \pi}\right)^{1/2 } ( kt)^{-3/2}\frac{1}{g^a_0 } \int \limits_{0}^\infty { \left\ { \sum_\mu { \frac{g_\mu^a}{g_0^a } \sigma^\mu_{aa}(e_\mu^a ) e_\mu^a } \right\ } e^{-\frac{e_0^a}{kt}}\,de_0^a } = \nonumber \\ & = \left(\frac{8}{m_{aa } \pi}\right)^{1/2 } ( kt)^{-3/2}\frac{1}{g^a_0 } \int \limits_{0}^\infty { \sigma^\mathrm{eff}_{aa } e_0^a e^{-\frac{e_0^a}{kt}}\,de_0^a } = \frac{\langle \sigma^\mathrm{eff } v\rangle_{aa}}{g^a_0 } \quad .
\label{eq : effrate}\end{aligned}\ ] ] the last line was obtained by realizing that the expression in the curly brackets is identical to @xmath157 , with the effective cross section from ( [ eq : effcs ] ) .
thus , the weighted sum over many mbds acting on the thermally populated excited states is reduced to a single mbd acting on an effective cross section and divided by the normalized partition function . in terms of relevant physics , this means that the boltzmann factor in the population probability is offset by shifting down each mbd to the same relative energy.@xcite now @xmath59 in ( [ eq : rate ] ) can be identified as @xmath158 , introducing a temperature dependence while @xmath141 is conveniently independent of @xmath41 .
it may be confusing that the stellar reactivity frequently is written as @xmath159 instead of @xmath160 .
this is not meant to imply that the stellar cross section as defined in ( [ eq : csstar ] ) is inserted in a single integral over a mbd as shown in ( [ eq : rate ] ) .
rather , the angle brackets imply a separate integration for each populated state as performed in ( [ eq : stellrate1 ] ) and ( [ eq : stellrate2 ] ) in this case .
equation ( [ eq : effrate ] ) not only simplifies the numerical calculation of the stellar rate but also allows to better understand certain details .
for instance , it can immediately be seen that stellar rates obey a reciprocity relation because the effective cross sections do .
a similar expression has to hold for the reverse reactivity @xmath161 as for the forward reactivity @xmath162 , being @xmath163 this expression is derived in the same manner as ( [ eq : effrate ] ) but by starting from thermally populated excited states in the final nucleus . with the help of ( [ eq : reci_eff2 ] )
we can express the reverse reactivity of ( [ eq : revreac ] ) in terms of the forward reactivity:@xcite @xmath164 where @xmath165 is the reaction @xmath94-value of the forward reaction .
the reciprocity relation ( [ eq : reci_eff2 ] ) applies to photodisintegration and captures as well . in relating the photodisintegration rate @xmath21 to the capture rate
, however , it has to be assumed that the denominator @xmath166 of the planck distribution for photons appearing in the photodisintegration reactivity @xmath167 can be replaced by @xmath168 , similar to the one of a mbd with the same temperature @xmath41 . with this approximation and realizing that @xmath169 , one obtains@xcite @xmath170 in the same manner as ( [ eq : revrate ] ) .
using the approximation of the denominator resulting in ( [ eq : revphoto ] ) is very important for the application in reaction networks .
employing the expressions ( [ eq : revreac ] ) and ( [ eq : revphoto ] ) avoids numerical inconsistencies in network calculations which may arise when forward and reverse rates are calculated separately ( or even from different sources ) .
the proper balance between the two reaction directions can only be achieved in such a treatment .
furthermore , simplified equations for reaction equilibria ( see sec . [
sec : equilibria ] ) can be derived which prove important in the modeling and understanding of nucleosynthesis at high temperature .
how large is the error stemming from the approximation involved in the derivation of ( [ eq : revphoto ] ) ?
although mathematically unsound , it turns out that setting @xmath171 is a good approximation for the calculation of the rate integrals and introduces an error of less than a few percent for astrophysically relevant temperatures and rate values.@xcite in other words , the contributions to the integral in ( [ eq : lambda ] ) are negligible at the low energies where planck and mbds differ considerably ( see also fig . 3.5 in ref . , and ref . ) .
this is assured by either a sufficiently large and positive @xmath172 , which causes the integration over the planck distribution to start not at zero energy but rather at a sufficiently large threshold energy , or by vanishing effective cross sections at low energy due to , e.g. , a coulomb barrier .
the assumption may not be valid for s - wave neutron captures with very small ( of the order of @xmath173 ) or negative @xmath94-values , but the required correction still is only a few % as can be shown in numerical comparisons between photodisintegration rates calculated with the two versions of the denominator .
such a comparison was performed with the code smaragd ( version 0.8s ; see sec .
[ sec : codes ] ) and the results are shown in figs . [
fig : mbcomp_proton ] , [ fig : mbcomp_alpha ] , and [ fig : mbcomp_neut ] .
generally , larger errors appear at lower temperature .
this results in astrophysical irrelevance of the errors in many cases because either the rates are too slow ( especially for rates involving charged projectiles ) or the target nuclei in question are so short - lived that they will never be produced at low plasma temperature .
the largest error found was between 50 and 100% for a few heavy nuclei at the driplines for proton- or @xmath0-capture at @xmath174 gk . for neutron captures , the errors when applying the standard approximation for the reverse rate were never larger than 10% at any investigated temperature , even at the driplines . in the figures , errors of 5% and smaller
are not presented in detail because they are assumed to be negligible , especially given the remaining uncertainties in the prediction of the rates far from stability .
figure [ fig : mbcomp_proton ] shows target nuclei for proton capture where errors reach 510% at @xmath175 gk when computing the photodisintegration rate from ( [ eq : revphoto ] ) .
as expected , this occurs close to the dripline where the reaction @xmath94-value is small or negative . at @xmath176
gk , however , the errors for all nuclei with mass number @xmath177 are below 5% already and thus negligible .
figure [ fig : mbcomp_alpha ] shows target nuclei for @xmath0 capture where errors maximally reach 550% at @xmath175 and 510% at @xmath178 gk , respectively , when computing the photodisintegration rate from ( [ eq : revphoto ] ) . again , this occurs for some ( but not all , depending on the energy - dependence of the effective cross section ) @xmath0 captures with small or negative @xmath94-values . at @xmath178
gk even fewer rates are affected and the maximal error is below 10% .
the shown rates have little or no astrophysical impact because those nuclei can only be reached at higher temperatures .
for instance , the @xmath82-process ( p - process ) significantly photodisintegrates nuclei close to stability at @xmath179 gk and the rp - process also requires such high temperatures and probably does not proceed beyond @xmath180.@xcite figure [ fig : mbcomp_alpha ] shows target nuclei for neutron capture where errors reach 510% at any plasma temperature .
as expected , captures with low @xmath94-value , either at the dripline or in the vicinity of closed shells , exhibit the largest errors but those do not exceed 10% .
nuclei so far from stability are expected to be synthesized only at higher temperature . for @xmath181
gk a smaller number of nuclei shows errors above 4% .
in fact , for most reactions across the chart the error introduced in the reverse rate due to the approximation of the planck distribution is less than 1% .
having assured that the approximation of the planck distribution in the derivation of ( [ eq : revphoto ] ) does not introduce a considerable error in the obtained rates , it is important to realize that further conditions have to be fulfilled to allow the application of the relations discussed above . we started from introducing a thermal population @xmath109 in ( [ eq : csstar ] ) above and consequently all derivations up to here depend on the assumption that the excited states in all participating nuclei are occupied according to this population factor .
this is a valid assumption for most astrophysical plasmas and most nuclei reach thermal equilibrium very rapidly through collisions and interactions with photons and other plasma components .
however , there are some nuclei which exhibit long - lived isomeric states ( well - known examples are @xmath182al , @xmath183lu , and @xmath184ta)@xcite with such spins that they can not be easily excited or de - excited through electromagnetic transitions . at sufficiently high temperature they may still get into equilibrium , sometimes through couplings to intermediate states , but the relevant transitions have to be carefully studied .
this can be achieved through an internal reaction network , not connecting different nuclei but rather including the different levels within one nucleus.@xcite levels not being in thermal equilibrium can be included in regular networks in such a manner as if they were a different nucleus .
also in this case , however , the populating and depopulating reactions have to be known explicitly .
this also applies when ensembles of excited states are in equilibrium but the different ensembles within a nucleus are not .
then each ensemble can be treated as a separate species in a reaction network and the reactions connecting the ensembles have to be included explicitly .
although not discussed in further detail here , it is worth mentioning that also weak interactions are affected by the thermal population of excited states .
for example , the @xmath27-decay half - life @xmath185 of a nucleus will be changed relative to its ground state half - life when the ground state becomes depopulated and excited states with different decay half - lives are populated .
thus , the decay `` constant '' @xmath186 is actually temperature - dependent @xmath187 where @xmath188 and @xmath189 is the decay lifetime of the excited state .
similar considerations also apply to other processes , such as electron capture and neutrino - induced reactions .
using the reciprocity relations derived above allows to simplify the full reaction networks as defined in ( [ eq : network ] ) .
such simplifications are instructive because they enable us to study nucleosynthesis properties which are independent of details in the hydrodynamic evolution of the system or even , as shown below , independent of individual reaction rates .
they usually go along with a restriction to only types of reactions in the network which are actually necessary instead of blindly evolving a large system of differential equations .
such an approach is not always feasible but considerable understanding of nucleosynthesis has been gained in the past through such means by circumventing the necessity of computationally intensive calculations .
such simplifications remain important today because it is still impossible to couple multi - dimensional hydrodynamic simulations to full reaction networks .
furthermore , restriction to the essential often provides a much better insight into the physical processes than a brute - force full network calculation .
here we are concerned with the simplifications because it has to be understood when it is necessary to know astrophysical reaction rates and when not . setting the abundance change @xmath190 on the left - hand side of ( [ eq : network ] )
implies that the sum of all rates destroying the nuclear species @xmath15 is exactly balanced by all production rates and the net change in abundance is therefore zero , leaving the abundance constant .
this is called _ steady flow equilibrium_. it is especially useful with reaction chains where most rates ( perhaps except one ) are in steady flow equilibrium
. then the slowest reaction sets the timescale of the reaction flow and all other reactions adjust .
as long as steady flow is upheld , no full reaction network has to be solved .
rather , the ratios of the steady flow abundances of the involved nuclei are related by the ratios of their net destruction rates ( or , equivalently , production rates as these have to be the same ) . for illustration ,
let us assume a chain of reactions @xmath191 connecting nuclei through reactions with the same projectile , where all net reactions are in steady flow and therefore the same , except for the one starting at nucleus @xmath31 . then @xmath192 and @xmath193 the slowest rate sets the abundance of @xmath194 through @xmath195 .
the use of a complete set of coupled differential equations is not required anymore but the important rates still have to be known .
steady state considerations are helpful when investigating hydrostatic hydrogen burning of stars through the pp - chains and the cno cycles.@xcite in the past they have also been used for sequences of neutron captures in the s - process on nuclei in between magic numbers.@xcite the fact that separate steady flows can be assigned to each mass region between closed shells has been termed _
local approximation _ in s - process studies .
a slightly different concept is to assume equilibrium between a forward and its reverse rate .
this is the case when the two rates are equal or , in practice , very close . since stellar rates obey the simple reciprocity relations ( [ eq : revrate ] ) and ( [ eq : revphoto ] ) , respectively , it is trivial to show that @xmath196 for a reaction @xmath197 and @xmath198 for a reaction @xmath199 .
the individual rates do not appear anymore in the relation between the abundances .
note that this does not imply that the abundances remain constant , they still depend on @xmath41 which may vary with time as well as on @xmath200 and @xmath201 .
depending on the plasma density reaction @xmath202c is very sensitive to the density and will not get into equilibrium for @xmath203 g/@xmath204 . ] , above @xmath205 gk all reactions ( with the exception of the weak interaction ) achieve equilibrium .
it can be shown that the equilibrium abundance of a nucleus @xmath31 can be calculated from a set of three equations@xcite @xmath206 where @xmath207 is the mass number , @xmath208 , @xmath209 are the abundances of the free neutrons and protons , respectively , and @xmath210 the nuclear mass unit .
the binding energy of the nucleus with neutron number @xmath211 and proton number @xmath212 is denoted by @xmath213 .
the sums run over all species of nuclei in the plasma , including neutrons and protons .
equation ( [ eq : masscons ] ) expresses mass conservation and ( [ eq : chargeconv ] ) is the charge conservation .
the unknown abundances @xmath214 , @xmath208 , and @xmath209 are obtained with the above equation set . note that reactions mediated by the weak interaction are not included in the equilibrium and @xmath44 may be time - dependent .
again , individual rates are not required to determine the abundances . when all abundances in the network obey the above relations , full _ nuclear statistical equilibrium _
( nse ) is achieved . in this case
, no reaction rates have to be known . in realistic cases ,
more or less extended groups of nuclei are in statistical equilibrium and the relative abundances within a group can be described by equations similar to ( [ eq : nse ] ) .
the different groups are connected by comparatively slow reactions not being in equilibrium , which determine the abundance level of one group with respect to another group similar to what was shown above for the steady state equilibrium .
the rates of these slow , connecting reactions have to be known explicitly .
this is called _ quasi - statistical equilibrium _ ( qse ) .
it appears in various kinds of high - temperature burning , such as hydrostatic oxygen and silicon burning in massive stars and different explosive scenarios .
a special kind of qse is the ( n,@xmath82)@xmath215(@xmath82,n ) equilibrium or _ waiting point approximation _ , often used in r - process calculations.@xcite this is nothing else than a qse within an isotopic chain , where neutron captures and ( @xmath82,n ) reactions are in equilibrium under very neutron - rich conditions ( @xmath216 @xmath49 ) and @xmath217 gk . for the r - process the network is reduced to neutron captures and their inverse reactions , and to @xmath218 decays ( with possible subsequent neutron emission ) .
the decays are not in equilibrium and determine the timescale with which matter is processed from small @xmath212 to the heaviest nuclei .
when ( n,@xmath82)@xmath215(@xmath82,n ) equilibrium is achieved , the abundances of nuclei within an isotopic chain are connected by @xmath219 which connects the abundance of nucleus @xmath31 ( mass number @xmath220 ) with the one of nucleus @xmath221 ( mass number @xmath222 ) .
there is an exponential dependence on the neutron separation energy @xmath223 of @xmath221 .
also in this type of equilibrium there is no dependence on the individual capture or photodisintegration rates .
the r - process flow to higher elements , however , depends on the @xmath218-decay rates which connect the isotopic chains and are not in equilibrium .
they are very slow compared to the rates in equilibrium and that is why `` waiting points '' are established , which are just the nuclei ( usually only one or two within a chain ) with the highest abundances according to ( [ eq : nggn ] )
. the r - process can not proceed until they decay and their decay rates have to be known . a similar equilibrium , but between proton captures and ( @xmath82,p ) ,
is reached in the late phase of the rp - process on the surface of mass accreting neutron stars.@xcite there , the waiting points are established close to the proton dripline .
nucleosynthesis under extreme conditions , such as encountered in some explosive scenarios , involves exotic nuclei far from stability .
according to the above , reaction rates are not needed for all of them because reaction equilibria are established at such extreme conditions . required
are nuclear masses ( to determine @xmath94-values or binding energies ) as well as spectroscopic information and nuclear level densities ( entering the calculation of the partition functions @xmath110 ) .
but although nse , qse , and the waiting point approximation do not contain the rates explicitly , they implicitly depend on them because they determine whether nuclei are participating in the equilibrium or not .
the higher the rates , the lower the temperature at which equilibrium is reached . with a time - dependent @xmath41 evolution
, this means that the rates determine whether equilibrium is reached earlier with increasing @xmath41 or the freeze - out happens later with decreasing @xmath41 . in principle , stellar cross sections @xmath224 as defined in ( [ eq : csstar ] ) correspond to physically measurable quantities , contrary to the purely theoretical effective cross sections introduced in ( [ eq : effcs ] ) . in practice ,
cross sections @xmath225 measured in terrestrial laboratories do not include thermal effects . therefore the rates derived from them do not obey reciprocity relations . to derive astrophysical rates appropriate for the utilization in reaction networks , laboratory cross sections almost always have to be supplemented by theory to account for the additional transitions not included in the measurement .
the energy- and temperature - dependent stellar enhancement factor for the cross sections ( again obtained via theory ) @xmath226 compares the stellar cross section including reactions from thermally populated excited states to the cross section obtained with reactions proceeding from the ground state of the target nucleus only .
another definition of a stellar enhancement factor involves the rates or reactivities , @xmath227 in fact , @xmath228 is the astrophysically interesting quantity because it shows the deviation introduced in the _ rate _ at a chosen stellar temperature @xmath41 when using the laboratory cross section instead of the stellar cross section .
then the fraction of the ground state contribution to the stellar rate is @xmath229 . on the other hand , @xmath230 is supposed to provide information for the experimentalist on how much off the measured @xmath231 is in comparison to @xmath224 at each energy for a given stellar temperature . to this end
it was quoted in literature occasionally .
however , a much more useful definition is @xmath232 which weights the excited states appropriately for a straightforward comparison .
this is the same as @xmath228 only if @xmath233 is independent of @xmath67 .
this is not necessarily fulfilled because the energy - dependence of laboratory and effective cross section may be different and this would lead to a different evaluation of the reaction rate integral .
( it turns out that in practice @xmath233 is often more slowly varying with energy than the cross section across the relevant energy window ( see sec .
[ sec : energies ] ) and that it can be approximated by an energy - independent factor in this case . )
it is recommended to use only either @xmath233 or @xmath228 , depending on whether an energy - dependent measure is desired or one independent of interaction energy . as has become clear from the derivation of the stellar rate in sec .
[ sec : stellar ] there is no simple relation between @xmath228 and @xmath230 .
it should be noted that while the stellar population factors @xmath109 are normalized to unity and the normalized partition functions @xmath234 can not become smaller than unity , the three types of stellar enhancement factors defined above can assume any positive value , _ larger or smaller _ than unity . in which case
do we have to expect large deviations @xmath235 from the laboratory value ? ( the above definition assures @xmath236 ; without thermal effects @xmath237 . ) again , a scrutiny of the effective cross section appearing in ( [ eq : sefrate ] ) helps to understand the various dependences .
let us start with the dependence on stellar temperature .
naively , one would assume that the higher the temperature , the larger the stellar enhancement will become .
indeed , this is often the case but there are further intricacies .
there are two dependences on @xmath41 appearing in ( [ eq : sefrate ] ) , one in the rate in the numerator and the other in the normalized partition function in the denominator .
the latter is monotonically increasing with increasing @xmath41 .
the integration of the product of the effective cross section and the mbd may show a different @xmath41-dependence , however , which even may not prove to be monotonic .
this can be understood by combining the knowledge of the relevant energy windows from sec .
[ sec : energies ] with definition ( [ eq : effcs ] ) of the effective cross section .
larger @xmath41 shifts the energy window to higher relative energy , both in the entrance channel @xmath238 with the relevant @xmath148 becoming larger and in the exit channel @xmath239 ( or @xmath240 ) with the relevant @xmath241 becoming larger ( see fig . [
fig : cnscheme ] ) .
since the effective cross section sums over transitions with relative energies @xmath242 and @xmath243 , it becomes obvious that the higher the relevant energy window , the more transitions are included .
this does not necessarily result in an increased rate although it often will .
it is also conceivable that the additionally included transitions at large @xmath41 have small cross sections and do not provide a considerable increase in the rate .
depending on the type of reaction , cross sections for transitions already included at small @xmath41 may also decrease with increasing relative energy . in these cases , @xmath228 will be reduced at larger @xmath41 because @xmath234 is always increasing . the stellar `` enhancement '' may even become smaller than unity and not live up to its name anymore .
this behavior should not be viewed as monotonic , either , because as additional transitions become accessible at even larger @xmath41 the stellar rate again may increase faster than @xmath234 .
it is to be expected , however , that @xmath228 is decreasing to very small values for very large @xmath41 after having reached a maximum value , i.e. , for very high ( but not necessarily astrophysically important ) temperatures @xmath244 will always be achieved .
this is because with increasing relevant energy @xmath245 other reaction channels become increasingly important , reducing the cross sections of the individual transitions .
additionally , more unbound levels will be included in the effective cross section .
these may lose particles to other reaction channels ( e.g. , through pre - equilibrium emission ; see also sec . [
sec : statmod ] ) and also not contribute to the effective cross section anymore .
it was suggested in ref . to include only bound states in the definition of the effective cross section .
although this may be a good approximation , it may neglect some transitions which still can contribute to the effective cross section at high temperature .
note , however , that unbound states may not be in thermal equilibrium although the timescale for reaching equilibrium under stellar burning conditions is short compared to the one of a nuclear reaction.@xcite a more suitable cutoff , if required , would be the energy @xmath246 appearing as cutoff in the calculation of the partition function in ( [ eq : partfuncint ] ) , although this may include already too many transitions with negligible cross sections .
examples for the above considerations are shown in figs .
[ fig : ossef1 ] , [ fig : ossef2 ] for neutron - induced reactions on os isotopes .
neutron capture on @xmath247os is important to understand the re - os cosmochronometer which can be used for age determinations in our galaxy.@xcite modern measurements of neutron capture in the astrophysically relevant energy range have reached a precision that requires the inclusion of the @xmath228 correction , even when it is only a few tens of percent.@xcite due to low - lying levels ( @xmath248 at 9.75 and 74.3 kev , @xmath249 at 75 kev , and @xmath250 at 100 kev ) in @xmath247os , @xmath228 is higher for @xmath247os(n,@xmath82)@xmath251os at low temperature than for neutron captures on neighboring isotopes .
the enhancement factors of the reactions @xmath247os(n,@xmath82 ) and @xmath251os(n,@xmath82 ) both rise to a maximum and decline from there , soon reaching values below unity and thus not being `` enhancements '' anymore . while the @xmath228 for both capture reactions stay below a factor of two ( and reach only a few tens of percent at the s - process temperature relevant for the cosmochronometer ) , the @xmath228 of @xmath252os(n,@xmath0)@xmath253w is also shown here to give an example for larger values ( see below for the even larger factors encountered for photodisintegration reactions ) .
having realized the importance of constraining @xmath148 by the relevant energy window at a given temperature and the definition of the range of energies @xmath254 of relevant excited states , we can arrive at a more general understanding of when @xmath255 will considerably differ from unity . in the assessment of which excited states in the target are contributing to the stellar rate , it is incorrect to directly use the boltzmann population factors @xmath109 from eq .
( [ eq : csstar ] ) !
these are only appropriate when each state is bombarded by its own mbd of projectiles .
they can not be used when we calculate the rate as usual , integrating over just one mbd with the energy scale being relative to the ground state . for levels at excitation energy @xmath256 and zero spin .
the weights are independent of stellar temperature but their slopes depend on the maximal transition energy @xmath76 as indicated by the labels.[fig : excweights1],scaledwidth=64.0% ] as shown in eqs .
( [ eq : stellrate2])([eq : effrate ] ) the transformation to a single mbd is possible because the boltzmann weight @xmath257 in the population factor offsets the exponentials in the mbd for each state .
this offset is not a complete cancellation but results in a transformation of the population @xmath109 .
the individual weights are transformed to linearly declining functions of the excitation energy @xmath258 this is also readily seen in the weighting of the reactions from excited states in the definition of the effective cross section by eq .
( [ eq : effcs ] ) . these relative _ effective weights _
@xmath259 are the ones to be employed when using the standard definition of the rate as in ( [ eq : rate ] ) , instead of the relative weights @xmath260 .
contrary to the @xmath260 , the relative effective weights @xmath259 are not explicitly temperature dependent anymore .
however , they depend on the energy @xmath76 of the transitions to the ground state which are the transitions with the highest possible relative energy . in the application to astrophysical reaction rates the range of @xmath76
is given by the relevant energy windows as discussed in sec .
[ sec : energies ] .
it is a relatively narrow range , depending on stellar temperature and the type of reaction ( and thus introducing an implicit temperature dependence ) . due to the linearity in excitation energy @xmath104
contrary to the @xmath260 which show an exponential decrease
almost all states up to @xmath76 contribute . neglecting the spin weights @xmath261 , one has to consider levels up to @xmath262 to include 90% of the levels with non - negligible weight . depending on the nucleus and considered reactions , some levels with large spin values and/or large reaction cross sections may require the inclusion of all levels up to very close to @xmath263 .
the relative effective weights @xmath259 as the product of the population factors with their respective mbds are shown in fig .
[ fig : excweights1 ] . understanding the @xmath259 we are ready to make some general statements on the magnitude of @xmath255 in various reactions in a range of astrophysically relevant temperatures .
for example , neutron captures have their relevant energies around the maximum of the mbd , @xmath264 .
this is between a few kev up to several tens of kev for the s - process@xcite ( see also sec . [
sec : energies ] ) .
therefore only ( exothermic ) captures on nuclei with low - lying excited states below several tens of kev will exhibit @xmath265 . for light and intermediate mass nuclei ,
the average level spacings typically are larger than @xmath266 kev and thus @xmath255 remains close to unity , with a few notable exceptions .
the r - process@xcite involves neutron captures at temperatures of @xmath267 gk which translates to a location of the mbd peak at @xmath268 kev and we will expect slightly larger @xmath255 than for the s - process on average . and @xmath269 in the initial and final nucleus , respectively , under the assumption of a positive reaction @xmath94-value @xmath270 . in both nuclei a few low - lying states
are explicitly numbered , above them nuclear level densities are indicated ( shaded areas ) .
[ fig : excweights2],scaledwidth=80.0% ] reactions having their relevant energy window determined by charged particle widths have considerably larger @xmath76 , in the range of about @xmath271 mev , depending on the stellar temperature and the charges of the interacting particles .
accordingly , @xmath255 can already become large at rather low temperature because transitions from many excited states have to be considered . only in light ,
strongly bound nuclei with level spacings of several mev above the ground state , @xmath255 may still remain close to unity .
the relative effective weights @xmath272 derived in ( [ eq : effweights ] ) also allow to understand a rule - of - thumb which has been known , but never quantified , for quite some time ( see , e.g. , ref . ) .
the rule states that exothermic reactions ( @xmath273 ) usually have smaller @xmath255 than their endothermic inverses . defining the forward reactions by the exothermic reaction and the reverse reaction by its endothermic counterpart ,
this means that @xmath274 and frequently @xmath275 .
this is not immediately comprehensible upon inspection of the boltzmann weights @xmath260 as these seem to act similarly in the initial and final nuclei . as stated above , however , it is a mistake to straightforwardly use the @xmath260 together with a single mbd .
the rule becomes obvious when using the appropriate @xmath259 .
the situation is sketched in fig .
[ fig : excweights2 ] where the behavior of the weights is shown for two nuclei @xmath31 and @xmath129 being the target and the final nucleus of a reaction , respectively . for the relevant interaction energy set to @xmath76
the maximum relative transition energy is @xmath148 in the target nucleus and @xmath276 in the final nucleus .
provided that @xmath277 the energy range of possible transitions @xmath278 in the final nucleus is larger than in the target nucleus . in consequence , the relative effective weights @xmath279 in nucleus @xmath129 decline slower with increasing excitation energy and thus a larger range of levels contributes .
this assumes similar spin structure in the two nuclei , of course , and the rule may not work at small @xmath280 when @xmath259 , @xmath279 are similar and the spin factors @xmath261 , @xmath281 dominate .
the largest reaction @xmath94-values are encountered in capture reactions .
for instance , neutron captures close to stability exhibit @xmath94-values of the order of @xmath282 mev , for highly proton - rich nuclei they can reach @xmath283 mev .
similar values are found for proton captures around stability and on the neutron - rich side of the nuclear chart . in the light of the above it
is not surprising that endothermic photodisintegration reactions exhibit very large @xmath228 of the order of several hundreds to thousands as illustrated in table [ tab : photosefs ] .
an exception to the above rule @xmath274 , however , has recently been discovered.@xcite although the effective weight @xmath284 may be slowly decreasing with increasing excitation energy @xmath104 of a level , the corresponding cross section @xmath105 may decrease much faster , even exponentially .
this is because with increasing @xmath104 the relative interaction energy in that channel @xmath143 is reduced .
if the cross section @xmath105 is strongly decreasing with decreasing energy as it is the case in the presence of a coulomb barrier or at high relative angular momentum
transitions on excited states will cease to importantly contribute to the effective cross section even when being strongly weighted .
this is the reason why charged particle reactions show only moderate values of @xmath255 at low @xmath41 even though their relevant energy window may be high above the reaction threshold .
this , of course , acts in the entrance channel of a reaction as well as in its exit channel . the point , however , is to realize that the barriers may be different in the two channels , leading to a different suppression of contributions .
for instance , a ( n , p ) or ( n,@xmath0 ) reaction has a coulomb barrier only in the exit channel , ( p,@xmath0 ) reactions have different barriers in entrance and exit channel .
this has the consequence that low - energy transitions in the exit channel of an exothermic reaction may be suppressed in such a manner as to yield @xmath285 . whether this is the case
will strongly depend on the @xmath94-value because it determines the range of transition energies to be efficiently suppressed .
the higher the barrier , the larger @xmath94-value is allowed while still permitting to suppress most transitions from excited states .
the stellar temperature also plays a role but its impact is smaller . a global search across the chart of nuclides for reactions exhibiting @xmath285
was performed in refs . .
the study focused on identifying cases which are interesting for experiments by requiring @xmath286 and the stellar enhancement factors for forward and reverse reaction to differ by more than 10% . at stellar temperatures @xmath287 gk about 1200 reactions were found but not all of them are astrophysically interesting.@xcite figure [ fig : qsupp ] shows the dependence of the @xmath94-values on the charge of the target nucleus of two types of reactions selected from the total set .
the envelope from the maximal @xmath280 appearing for each type at each charge @xmath288 illustrates the action of the increasing coulomb barrier .
it can be clearly seen that larger maximal @xmath289 is allowed with increasing charge @xmath288 .
the increase with increasing charge is steeper for ( @xmath0,n ) reactions than for ( p , n ) reactions due to the higher coulomb barrier for @xmath0 particles .
the scatter at a fixed charge number @xmath288 is due to the range of @xmath94-values found within an isotopic chain . summarizing the above , comparing the position of the relevant energy window with the average level spacing in a nucleus already gives an estimate of the magnitude of the stellar enhancement to be expected . when measuring reaction cross sections for astrophysics it is desireable to perform the experiment in the reaction direction showing smallest @xmath255 in order to stay as close as possible to the actually required stellar cross section .
this favored direction turns out to be the exothermic one in the vast majority of cases , i.e. , a reaction with positive @xmath94-value , with comparatively few exceptions ( compared to the total number of conceivable reactions ) as discussed above .
it should be remembered that the reverse rate can always be calculated using the reciprocity relations ( [ eq : revrate ] ) and ( [ eq : revphoto ] ) . , appearing in ( [ eq : revrate ] ) , ( [ eq : revphoto ] ) , when @xmath290 . ] finally , it is worth mentioning that similar considerations for the transformation of effective weights and deviations @xmath255 from the ground state values also apply to other types of reactions , not only the ones discussed here . for instance , also for electron captures and other reactions mediated by the weak interaction an effective cross section , effective weights , and relevant energy windows can be derived , applying the appropriate energy distributions .
they may become more complicated , however , than the ones obtained for the mbd because an explicit dependence on the chemical potential is required in the general case .
moreover , neutrinos are not in thermal equilibrium with nuclei and thus their temperature will enter as an additional parameter .
the plasma temperature in the astrophysical sites where nuclear reactions occur is so high that the nuclei are fully ionized and embedded in a cloud of free electrons .
this situation affects reactions and decays and has to be considered when preparing a rate to be used in astrophysical reaction networks .
similar to the treatment of stellar enhancement factors , these corrections also have to be modeled theoretically and experimental rates have to be corrected for them . for completeness , some of the effects are discussed briefly without going into detail here .
decays ( and electron captures ) are affected not only by the thermal excitation of the nucei as shown in ( [ eq : stelldecay ] ) but also by the electrons surrounding the nuclei . a nucleus in an atom
can be converted by , e.g. , capture of an electron from the atomic k - shell . in the plasma ,
electrons are captured from the electron cloud and this may alter the half - life of a nucleus considerably . a well - known example for this is the decay of @xmath291be .
its lifetime under central solar conditions ( @xmath292 d ) is almost double the one under terrestrial , non - ionized conditions ( @xmath293 d ) .
also the @xmath218 decay lifetimes are modified by a change in the electron emission probability .
the plasma electrons reduce the phasespace available for emission to the continuum and thus increase @xmath294 . on the other hand , bound - state decay , i.e. the placement of the emitted electron into a low - lying atomic shell , becomes possible even when it were forbidden in an atom because of the occupation of available electron shells .
similar considerations apply to charged - current reactions with electron neutrinos .
these effects act in addition to the alteration of the lifetimes through thermal excitation of the nucleus and depend on the plasma temperature , density , and composition , which may all affect the distribution of the electrons throughout the plasma .
nuclear reactions with charged nuclei are affected by _
electron screening_. electrons in the vicinity of the nucleus shield part of its charge and thus effectively lower its coulomb barrier .
theoretical predictions of cross sections and rates always assume bare nuclei without any electrons and therefore have to be corrected for screening .
the magnitude of the screening is strongly dependent on the temperature @xmath41 , density @xmath6 , and composition @xmath295 of the plasma . at high density
, electron screening can increase the rate by several orders of magnitude and lead to _ pycnonuclear _ burning at much lower temperatures than otherwise needed to ignite nuclear burning .
moreover , dynamical effects due to the fast movement of electrons may also play a role , although there is discussion on whether this is important when all reactants are in thermodynamic equilibrium.@xcite it is apparent that the theoretical treatment of screening is complicated and we are far from complete microscopic descriptions.@xcite the proper inclusion of screening effects remains a challenging problem in plasma physics . under most conditions
, the screened reactivity can be decomposed into the regular stellar reactivity and a screening factor@xcite @xmath296 this implies that the coulomb potential @xmath297 seen by the reaction partners can be described by the bare coulomb potential and an effective screening potential @xmath298 @xmath299 then the screening factor acquires the form @xmath300 .
the challenge is in the determination of @xmath301 depending on the plasma conditions . an often used static approximation , being appropriate for early burning stages of stars , is _ weak screening _ in the debye - hckel model.@xcite weak screening assumes that the average coulomb energy of each nucleus is much smaller than the thermal energy , i.e. @xmath302 .
then a nucleus will be surrounded by a polarized sphere of charges , with a radius @xmath303 with @xmath304 where the sum runs over all charged plasma components .
the screening factor @xmath305 is transformed to @xmath306 with @xmath307 . for _ strong screening _ in high density plasmas
it is more appropriate to use the ion - sphere model instead of the debye - hckel approximation.@xcite the ion - sphere model is equivalent to the wigner - seitz model used in condensed matter theory .
another type of screening is observed in nuclear experiments in the laboratory .
there , nuclei are present in atoms , molecules , or metals , each with specific electron charge distributions around the nucleus .
although completely different from plasma screening , this type of screening has to be understood because it is especially important at the low interaction energies of astrophysical relevance . the _ measured _ reaction cross sections have to be corrected to obtain the bare cross section which can be compared to theory or used to determine the rate .
atomic screening can be treated in the adiabatic approximation , leading to @xmath308 with the sommerfeld parameter @xmath309 from ( [ eq : sommerfeld ] ) .
the screening potential @xmath310 in this approximation is given by the difference in the electron binding energy of the target atom and the atom made from target atom plus projectile @xmath311 . in light systems ,
the velocity of the atomic electrons is comparable to the relative motion between the nuclei .
therefore a dynamical model is more appropriate.@xcite however , the adiabatic approximation provides an upper limit on the expected screening effect on the cross section .
there seem to be discrepancies between theory and laboratory determinations of @xmath310 , the latter often yielding much larger values of @xmath310 .
some of them have been resolved through improved stopping powers used in the determination of the experimental cross sections,@xcite while others remain puzzling , especially regarding cross sections of nuclei implanted in metals.@xcite thus , the laboratory screening seems to be less understood than the stellar screening .
having discussed the special requirements of astrophysics for the determination of the stellar rates in the preceding sections , we turn to the question of how to obtain the cross sections @xmath312 required in eqs .
( [ eq : rate ] ) , ( [ eq : csstar ] ) and ( [ eq : effcs ] ) at the energies of astrophysical relevance .
it has become apparent that astrophysical rates include more transitions than usually obtained in straightforward laboratory reaction cross section measurements .
cleverly designed experiments may study some of them but especially at larger stellar temperatures , as they are typical for explosive burning , theory will be indispensable for providing the appropriate _ stellar _ reaction rates . even at stability , the small cross sections of charged - particle reactions at astrophysically relevant energies ( see sec . [
sec : energies ] ) pose a considerable challenge for current measurements and future experiments have to employ novel techniques or new facilities to address this problem .
moreover , hot astrophysical environments produce highly unstable nuclei which can not be studied in the laboratory , yet .
reaction cross sections of nuclei far from stability at astrophysical energies probably will never be experimentally determined .
therefore reaction networks for explosive nucleosynthesis have to include the majority of their reaction rates from theoretical predictions although experiments may help to determine nuclear properties and some of the transitions required for the calculation of effective cross sections .
although reaction theory dates back as far as the 1950s , the special requirements of astrophysics and the need for cross section predictions of nuclei far from stability provide an interesting and stimulating environment for the application and further developments of different approaches .
the challenges are manyfold . on one hand ,
astrophysical energies are very low and , as we will see below , different reaction mechanisms may contribute or interfere .
on the other hand , even if the reaction mechanism is unique and well understood , nuclear properties entering the reaction model have to be predicted for nuclei far off stability .
this proves challenging even for modern nuclear structure calculations .
although a fully microscopic treatment is preferrable , good parameterizations and averaged quantities are still necessary in many cases due to the sheer number of reactions and involved nuclei , especially for intermediate and heavy nuclei consisting of more than @xmath313 nucleons and thus not allowing the application of few - body models . finally , the interpretation of experiments has to be supported by theory .
this latter case may involve different methods than the one dealing with the prediction of astrophysical rates because experiments may be conducted at higher energies and theory is needed to extract the information to be included in the rates.@xcite for example , the properties of excited states and their spectroscopic factors can be studied by ( d , p ) reactions at comparatively high energy which are not directly relevant in astrophysics . in
the following i focus on theory for the prediction of astrophysical reaction rates .
theoretical models can be roughly classified in three categories:@xcite 1 .
models involving adjustable parameters , such as the @xmath314-matrix@xcite or the @xmath315-matrix@xcite methods ; parameters are fitted to the available experimental data and the cross sections are extrapolated down to astrophysical energies .
these fitting procedures , of course , require the knowledge of data , which are sometimes too scarce for a reliable extrapolation .
ab initio " models , where the cross sections are determined from the wave functions of the system .
the potential model@xcite , the distorted wave born approximation ( dwba)@xcite , and microscopic models@xcite are , in principle , independent of experimental data .
more realistically , these models depend on some physical parameters , such as a nucleus - nucleus or a nucleon - nucleon interaction which can be reasonably determined from experiment only .
the microscopic generator coordinate method ( gcm ) provides a basic " description of a nucleonic system , since the whole information is obtained from a nucleon - nucleon interaction . since this interaction is nearly the same for all light nuclei , the predictive power of the gcm is high for such nuclei .
the above models can be used for low level - density nuclei only .
this condition is fulfilled in most of the reactions involving light nuclei ( @xmath316 ) .
however when the level density near threshold is large ( i.e. more than a few levels per mev ) , statistical models , using averaged optical transmission coefficients , are more suitable ( see sec . [ sec : statmod ] ) .
the nuclear level density ( nld ) at the compound nucleus excitation energy corresponding to the astrophysical energy window determines which reaction mechanism is applicable and which model to choose . the compound formation energy @xmath317 is given by the astrophysical energy @xmath76 relative to the ground state of the target nucleus and the separation energy of the projectile @xmath318 .
around stability , @xmath318 usually is high and dominates @xmath319 .
statistical models will then be applicable for intermediate and heavy nuclei with sufficiently high nld . even at stability
, however , the nld may not be high enough at nuclei with closed shells ( see sec .
[ sec : statmod ] ) . approaching the driplines ,
the neutron- or proton - separation energies strongly decrease , resulting in low @xmath319 for neutron- or proton - induced reactions , respectively .
this leads to low nld at @xmath319 even for intermediate and heavy nuclei.@xcite isolated resonances but also direct reactions will become important.@xcite in the following , reactions at intermediate ( sec .
[ sec : reso ] ) , high ( sec .
[ sec : highnld ] ) , and low ( sec .
[ sec : direct ] ) compound nld are discussed separately although there may be contributions from several reaction mechanisms simultaneously , especially in systems with low and intermediate nld .
the discussion will focus on models more or less applicable for large - scale predictions across the nuclear chart .
further models are presented in , e.g. , refs . .
resonances in reaction cross sections are important for the majority of nuclei .
depending on the number of nucleons in the target nucleus resonances appear in the reaction cross section at lower or higher energy and their average spacing also depends on the structure of the nucleus .
astrophysical energy windows cover regions of widely spaced , isolated resonances to regions of a large number of overlapping , unresolved resonances .
accordingly , different approaches have to be combined .
the latter region is more important for reactions between charged reactants because the coulomb barrier shifts the relevant energy window to higher energy compared to reactions where neutrons determining the location of the window ( see sec .
[ sec : energies ] ) .
isolated resonances in the low and intermediate nld regimes can be treated in the @xmath314-matrix@xcite or the @xmath315-matrix@xcite approaches or by applying simple single - level or multi - level breit - wigner formulae . in all these methods ,
the resonance properties ( resonance energy , spin , partial and total widths ) have to be known .
often , an inverse approach is used and the resonance properties are derived from experimental data by , e.g. , @xmath314-matrix fits .
where this is impossible , nuclear theory has to be invoked to predict the required quantities .
this remains problematic , however , because the reaction cross sections are very sensitive to the resonance properties .
the astrophysical rates are also sensitive but since their calculation involves an integration over an energy range , only strong resonances truly contribute and others may be averaged out .
nevertheless , there are large uncertainties in reaction rates off stability due to the unknown resonance contributions ( see also sec . [ sec : sensi ] ) .
cluster models ( see sec .
[ sec : optmod ] and ref . ) have been successful in describing resonant cross sections in light nuclei but can not be easily applied to nuclei at intermediate and heavy mass .
although resonances with the same spin @xmath320 interfere and single resonances may also show interference with a direct reaction ( sec . [
sec : direct ] ) , the single - level breit - wigner formula ( bwf ) is often used:@xcite @xmath321 it is quoted here despite its restrictions because it allows to demonstrate some important principles important for the relation between stellar and laboratory rates .
equation ( [ eq : breit ] ) gives the bwf for @xmath322 non - interfering resonances .
the wave number is denoted by @xmath323 .
the total width @xmath324 of a resonant state @xmath15 in the compound nucleus is the sum over the widths of the individual decay channels @xmath325 , also including transitions to other reaction channels beyond the exit channel @xmath239 .
the widths of the individual decay channels are summed over transitions to all possible final states in the channel .
thus , @xmath326 .
figure [ fig : cnscheme ] shows the energy scheme and the contributing transitions in each channel . when the resonance energy @xmath327 is known , the widths @xmath328 and @xmath329 can be calculated from the transmission coefficients obtained by solution of a schrdinger equation in the optical model ( see sec .
[ sec : optmod ] ) and a spectroscopic factor ( see sec . [
sec : direct ] ) . both resonance energies and spectroscopic factors should , in principle , be predictable in the shell model ( see , e.g. , refs . ) or other microscopic theories but this is currently not feasible for all nuclei across the nuclear chart .
different approaches yield results which differ more than it is tolerable in the calculations of astrophysical reaction rates.@xcite according to ( [ eq : labcs ] ) , we take @xmath112 for the usual laboratory cross section .
it is interesting to note that also resonances located below the reaction threshold may contribute due to their finite width reaching above the threshold .
these are called _ sub - threshold resonances_.@xcite in reactions with a large , positive @xmath94-value the energy dependence of the partial width in the exit channel @xmath330 can be neglected .
this is not true when the @xmath94-value is small or negative . as has become obvious in the discussion of the stellar cross section in sec .
[ sec : stellar ] , for the astrophysical reaction rate , the effective cross section has to be employed in the integration for the reaction rate and thus a weighted sum over excited target states @xmath102 has to be performed and we obtain @xmath331 with the effective weights @xmath332 taken from ( [ eq : effweights ] ) and @xmath143 .
this can be simply achieved by replacing @xmath333 in ( [ eq : breit ] ) by @xmath334 ( summing over all possible transitions in the entrance channel ) and dividing the resulting cross section by the normalized partition function of the target nucleus @xmath335 .
this can easily be shown when combining definitions ( [ eq : effcs ] ) and ( [ eq : breit ] ) for a single resonance with spin @xmath320 , @xmath336 this results in @xmath337 for the stellar rate .
some simplifications can be made depending on the resonance widths . for simplicity ,
the derivations are given for a single resonance with spin @xmath320 , for @xmath338 the contributions can be added .
a frequently used simplification is the one for narrow resonances , assuming that the widths @xmath329 in the numerator in the integral and the boltzmann factor @xmath339 do not change across the width of the resonance.@xcite then their values can be taken at the resonance energy and the integration can be performed analytically ( see also eq .
( [ eq : hfaverage ] ) ) , yielding @xmath340 where the resonance energy @xmath341 is given relative to the ground state of the target nucleus .
how does this compare to the ground state rate usually measured in the laboratory ?
the stellar rate can be recast as @xmath342 with @xmath343 the second line in ( [ eq : narrowcontrib ] ) was obtained by neglecting the energy dependence of @xmath344 and @xmath345 .
this is a valid assumption provided the reaction has a sizeable , positive @xmath94-value . as can be seen from ( [ eq : narrowcontrib ] ) , the contributions from excited states vanish quickly because often @xmath328 is strongly energy dependent and vanishes fast with decreasing energy ( remember that @xmath143 ) .
this is certainly true for reactions between charged particles and low resonance energies .
nevertheless , the resonant transitions from excited states may dominate a resonant stellar rate.@xcite for broader resonances the above approximation can not be used and the rate has to be determined by numerical integration of eq .
( [ eq : rate ] ) .
the wings of broad resonances can also contribute significantly to the rate even when the resonance energy is outside the relevant energy window for the rate .
sometimes the values for @xmath346 , @xmath344 , and @xmath347 are known experimentally at the resonance energy .
then a frequently used approach in experimental nuclear physics is to assume that @xmath344 and @xmath347 are approximated by energy - independent values and the energy - dependence of @xmath346 is only due to a barrier penetration factor derived from an optical model .
even if the energy dependence of @xmath344 , @xmath347 is accounted for explicitly , this type of extrapolation does not include the stellar enhancement and is only valid for laboratory cross sections .
the stellar rate must be calculated from a weighted sum of resonant contributions , as shown in ( [ eq : stellbw ] ) , both for the value at the resonance energy and in the extrapolation .
it follows from ( [ eq : sumgamma ] ) that the only difference in the energy dependence , however , stems from the width in the entrance channel where @xmath346 has to be replaced by @xmath348 .
the additional transitions to excited states in the target nucleus can be measured in principle .
if they are not available , @xmath348 at the resonance energy has to be predicted from theory .
also the extrapolation is more involved because @xmath348 will have a different energy dependence than @xmath346 .
the same methods can be used as in the extrapolation of @xmath346 but they have to be applied to all contributing transitions separately . the discussion of electron screening in the stellar plasma in sec .
[ sec : electro ] applied to nonresonant rates .
it can be shown that the same screening corrections can be applied for resonant rates when @xmath349.@xcite a more complicated form arises for @xmath350 , see refs . and for further details .
microscopic models are based on basic principles of quantum mechanics , such as the treatment of all nucleons , with exact antisymmetrization of the wave functions .
the hamiltonian of an @xmath220-nucleon system is @xmath351 where @xmath352 is the kinetic energy and @xmath353 a nucleon - nucleon interaction.@xcite the schrdinger equation associated with this hamiltonian can not be solved exactly when @xmath354 .
the quantum monte carlo method represents a significant breakthrough in this direction , but is currently limited to @xmath355.@xcite in addition its application to continuum states is not feasible for the moment ( it has been applied to the d(@xmath0,@xmath82)@xmath356li reaction but the @xmath0+d relative motion is described by a nucleus - nucleus potential).@xcite in cluster models , it is assumed that the nucleons are grouped in clusters and internal wave functions describing the relative cluster motions are generated.@xcite the main advantage of cluster models with respect to other microscopic theories is its ability to deal with reactions , as well as with nuclear spectroscopy .
over the past years , much work has been devoted to the improvement of the internal wave functions : multicluster descriptions@xcite , large - basis shell model extensions@xcite , or monopolar distortion@xcite .
the main limitation arises from the number of channels included in the wave function , which reduces the validity of the model at low energies .
also large nlds require many channels in the wave functions .
therefore the application of cluster models is limited to light nuclei . due to the complexity of the nucleon - nucleon ( nn ) interaction , one often resorts to working with effective interactions instead of solving microscopic models based on nn potentials .
widely used in calculating different reaction mechanisms is the optical model . in that model , the complicated many - body problem posed by the interaction of two nuclei is replaced by the much simpler problem of two particles interacting through an effective potential , the so - called optical potential.@xcite such an approach is usually feasible only with few contributing channels .
always included is elastic scattering .
that is why optical potentials can be derived from elastic scattering data .
the time - independent radial schrdinger equation is numerically solved with an optical potential which provides a mean interaction potential , averaging over individual nn interactions . in consequence
, single - particle resonances can not be described in such a model .
however , resonances stemming from potential scattering can still be found .
elementary scattering theory yields expressions for the elastic cross section and the reaction cross section .
the latter includes all reactions and inelastic processes which cause loss of flux from the elastic channel . with the diagonal element @xmath357 of the s - matrix ( sometimes also called scattering matrix or collision matrix ) , the reaction cross section for spinless particles is then given by@xcite @xmath358 this can be generalized to other outgoing channels @xmath27 , not just the elastic one .
the elements of the s - matrix are complex , in general , and related to the scattering amplitude @xmath359 of the outgoing wave function @xmath360 which , in turn , is nothing else than the transition amplitude @xmath361 connecting entrance channel @xmath0 and exit channel @xmath27 , with @xmath362 being the reduced mass in the entrance channel and @xmath363 a legendre polynomial .
the imaginary part of the optical potential gives rise to an absorption term in the solution of the schrdinger equation , thus removing flux from the considered channels .
therefore , the matrix element @xmath357 is also related to the transmission coefficient @xmath364 which describes the absorption of the projectile by the nucleus .
important for practical application is that the phase shifts @xmath365 can be derived from elastic data .
the optical model is well suited for describing transitions between states of intermediate and heavy nuclei .
it has been and is still used also to treat reactions with light nuclei although other methods exist for these . the optical model can be used to compute the widths ( @xmath328 , @xmath329 , @xmath345 ) appearing in the bwf for resonances , see ( [ eq : breit ] ) above . as mentioned before , the relevant energy windows for astrophysics also include compound nucleus excitation energies with such high nld that individual resonances can not be separated because the average resonance width @xmath366 becomes larger than the average level spacing @xmath367 . in fact , this is the case for the majority of reactions included in astrophysical reaction networks . instead of explicitly dealing with a large number of unknown resonances , one moves to averaged resonance properties .
starting with the bwf given in ( [ eq : breit ] ) , the sum of individual resonances can be replaced by an average over an energy interval @xmath368 using the mathematical relation@xcite @xmath369 here , the angle brackets denote the average as defined by the above equation .
note that the approximation for narrow resonances , as also used in ( [ eq : narrowres ] ) and ( [ eq : narrowgs ] ) , was applied to arrive at the last line of ( [ eq : hfaverage ] ) .
this is obviously allowed because of the assumption of a large number of narrowly spaced resonances . with this we rewrite the sum over resonances in the bwf as @xmath370 the number of resonances @xmath371 within an energy interval @xmath368 was replaced by the nld @xmath115 in the last line .
the averaged widths , the nld , and the @xmath372 are energy - dependent , of course .
the width fluctuation coefficients @xmath372 account for the different averaging in the last line @xmath373 in terms of physics , they describe non - statistical correlations between the widths in the channels @xmath238 and @xmath239 . in practice , they differ from unity only close to channel openings.@xcite making use of the relation between transmission coefficients obtained from the solution of the schrdinger equation with an optical potential and the averaged widths @xmath374 the cross section for the statistical model of compound reactions can be written as @xmath375 the summation in the denominator runs over all channels @xmath38 leading to the same compound nucleus , not only ( but including ) @xmath238 and @xmath239 .
thus , this sum is equivalent to @xmath376 .
also , the sums over channel spins @xmath14 and partial waves @xmath377 are explicitly written to emphasize that the transmission coefficients must include these quantum numbers .
each transmission coefficient includes transitions from states at the compound energy @xmath378 ( @xmath379 being the separation energy of the projectile in the compound nucleus ) . while @xmath380 only includes those to the state @xmath102 in the target nucleus , the @xmath381
include all transitions allowed by energetics and quantum selection rules . comparing ( [ eq : hf ] ) with ( [ eq : breit ] ) and ( [ eq : narrowgs ] ) it is readily seen that the statistical model cross section is an averaged breit - wigner cross section for narrow resonances , when @xmath382 .
completely equivalently to the breit - wigner case , it can be shown that for the calculation of the stellar rate it is sufficient to replace @xmath380 by @xmath383 in ( [ eq : hf ] ) and divide the resulting integral by the normalized partition function @xmath335 , giving@xcite @xmath384 the total transmission coefficients @xmath385 in each channel @xmath386@xmath387 , @xmath388 , include a sum over final states @xmath389 in that channel .
similar to the treatment of the partition functions in ( [ eq : partfuncint ] ) , the sum over discrete states can be extended by an integration over a level density above the energy @xmath390 of the last discrete state included , @xmath391 the integration is over the nld @xmath392 in the channel @xmath393 , i.e. in the target nucleus @xmath31 for channel @xmath387 , in the final nucleus @xmath129 for channel @xmath388 , and so on .
the transmission @xmath394 is the same as @xmath395 , only that it is a transition to an artifical state with given @xmath396 and weighted by the nld @xmath397 .
the relative transition energy in channel @xmath393 is @xmath398 , where @xmath399 is the channel separation energy .
the reader is advised to consult fig . [
fig : cnscheme ] to get an overview of the included transitions and their relative energies .
particle transmission coefficients have to obey spin selection rules and thus @xmath400 here the angular momentum @xmath401 and the channel spin @xmath402 are connected by @xmath403 including the particle spin @xmath100
. each @xmath404 can be directly obtained from the solution of the ( time - independent , radial ) schrdinger equation at the energy @xmath405 with an appropriate optical potential .
the calculation of radiative transmission coefficients proceeds equivalently to ( [ eq : parttrans ] ) but electromagnetic selection rules ( see , e.g. , appendix b of ref . )
have to be obeyed .
the parities @xmath406 , @xmath407 and the angular momentum @xmath377 select the type of allowed electromagnetic transition ( e1 , e2 , m1 , m2 , etc . ) and accordingly the appropriate description has to be invoked to calculate the transition strength @xmath404 .
to phenomenologically account for pre - equilibrium particle emission at higher compound excitation energy ( see sec . [
sec : statmodmod ] ) , sometimes the integration in ( [ eq : tottrans ] ) is only carried out to a cut - off energy @xmath408 , with an appropriately chosen @xmath409 ( e.g. , the energy at which the @xmath82 transmission exceeds a certain fraction of the total transmission ) for the @xmath82 transmission coefficient appearing in the numerator of ( [ eq : hfrate ] ) .
the total transmission coefficient @xmath410 in the denominator , however , always has to include the full integration up to @xmath319 .
the statistical model of compound reactions was initially developed by bohr , who conceived the independence hypothesis.@xcite it states that the projectile forms a compound system with the target , shares its energy among all of the nucleons , and finally the compound nucleus decays by emitting photons or particles independently of the formation process .
this implicitly requires long reaction timescales as the compound nucleus has to live long enough to establish complete statistical equilibrium among the nucleons .
compared to the direct mechanism ( sec.[sec : direct ] ) the timescale is about @xmath411 orders of magnitude longer and includes many degrees of freedom . in the independence hypothesis
, the ( laboratory ) cross section can be factorized into two terms @xmath412 the formation cross section @xmath413 and a branching ratio describing the probability for decay to the observed channel @xmath388 . an early implementation of this was the weisskopf - ewing theory.@xcite since then , the hauser - feshbach approach has been widely used , which also incorporates conservation of angular momentum partially lifting the independence assumption but thus being more realistic.@xcite equation ( [ eq : hf ] ) is the cross section from the full hauser - feshbach formalism . nevertheless , although too simplified , eq . ( [ eq : independence ] ) is sometimes useful when estimating the relative feeding of different reaction channels .
although it might seem tempting to conclude that the cross section of a reaction proceeding through the compound mechanism should be smooth because it is formed from the superposition of amplitudes from a very large number of states with random phases , this is a wrong assumption.@xcite it was first shown in ref . that the cross sections can continue to show large fluctuations .
the usual hauser - feshbach equations do not account for these fluctuations .
therefore , a meaningful comparison to experimental data is only possible after averaging the data over a sufficiently wide energy range , comparable to the average resonance widths . when using the statistical model to compute astrophysical reaction rates ( or when deriving rates experimentally directly ) this is taken care of automatically .
however , when using beams with a very narrow energy spread it should be noted that the results can not be directly compared to calculations.@xcite it is worthwhile to point out that the reaction rate is rather `` forgiving '' to deviations around a `` true '' cross section value , provided the deviations go both ways and can cancel within the integration in ( [ eq : rate ] ) .
therefore , the statistical model approach may even be applicable in the presence of small but isolated resonances as long as their average contribution is correctly accounted for .
this is closely connected to the question of the applicability of the statistical model .
this does not necessarily mean the actual validity of the model but rather its suitability to calculate astrophysical reaction rates . as pointed out above
, the model cross sections may deviate from the actual ones with large consequences for the rate .
the point is that the cross section _ averaged _ within the relevant energy window is described correctly so that the evaluation of the integral yields the `` true '' value .
this is a different and less stringent criterion than one asking for a close reproduction of experimental cross sections .
the rule - of - thumb@xcite of 10 levels within the relevant energy window has been found quite sound on average in simple numerical tests.@xcite it has to be emphasized that this is an average value .
sometimes even one or a few resonances , if broad enough to almost fill the energy window , may be sufficient . on the other hand , it is always desireable to get as much experimental information as possible for any nucleus but especially for those with low nld in the relevant energy window .
is in gk ) : @xmath414 ( 0 ) , @xmath415 ( 1 ) , @xmath416 ( 2 ) , @xmath417 ( 3 ) , @xmath418 ( 4 ) , @xmath419 ( 5 ) , @xmath420 ( 6 ) , @xmath421 ( 7 ) , @xmath422 ( 8) , @xmath423 ( 9 ) .
[ fig : pgapplic],scaledwidth=80.0% ] using the criterion of 10 levels per energy window , applicability maps for neutron- , proton- , and @xmath0-induced reactions were shown in ref .
, showing the lowest stellar temperature at which the statistical model is applicable to predict the rate .
it is important to note that those maps were derived with the approximations ( [ eq : e0approx ] ) and ( [ eq : deltaapprox ] ) for the gamow windows and not with the correct energy windows as explained in sec .
[ sec : energies ] . within the accuracy with which the plots can be read , this may not impact capture reactions too much , however , especially the energy windows for ( n,@xmath82 ) reactions have not been changed.@xcite for comparison , fig .
[ fig : pgapplic ] shows the minimal temperatures required for ( p,@xmath82 ) ( although on a coarser temperature grid ) for a region of the nuclear chart when applying the correct energy windows .
it has to be kept in mind , though , that reactions with different exit channels do not exhibit the same relevant energy window , as explained in sec .
[ sec : energies ] .
the general picture arising is the same as described in ref . .
the statistical model can be applied to the majority of neutron capture reactions , with exceptions close to magic neutron numbers or low neutron separation energy , both leading to a low nld at the compound formation energy @xmath319 .
since the relevant energy windows for charged - particle capture are shifted to higher energy with respect to the ones for ( n,@xmath82 ) , the applicability is even broader .
this may also apply to reactions with a charged particle in entrance or exit channel but no general statement can be made because it depends on which width determines the location of the energy window .
finally , endothermic reactions always require higher temperature to have appreciable rates but that does not necessarily mean that the compound nucleus is formed at high excitation energy .
therefore , they may require even higher temperature until the statistical model becomes applicable . in the preceding section
, the applicability of the statistical model ( hauser - feshbach model , hfm ) due to the required average nld has been discussed .
modern reaction theory knows a multitude of models , each suited to a particular reaction type and mechanism .
these all imply certain approximations . of course , nature is continuous , many types of reactions occur simultaneously and we have to choose an approximation suited to describe the dominant effects .
astrophysical energy windows prefer low projectile energies but still may include transitional regions between several reaction mechanisms .
discussed below are two types of modifications to the rate calculations : 1 ) accounting for additional reaction mechanisms , and 2 ) modifying the hfm itself to provide a smooth transition to low nld regimes . the inclusion of isospin conservation is a further modification which is discussed in a separate subsection of sec . [ sec : relevance ] .
direct reactions are known to be relevant at high projectile energies but were also found to significantly contribute at low energies in nuclei with low nld .
they are further discussed in sec .
[ sec : direct ] .
semi - direct@xcite and multistep reactions@xcite also occur on a faster timescale than equilibrated compound reactions .
they become important at several to several tens of mev projectile energy and are thus outside the astrophysically relevant energy range .
the capture of @xmath0-particles on intermediate and heavy nuclei may barely reach such energies but only at high plasma temperature .
it was shown that semi - direct capture is negligible even for very neutron - rich nuclei at r - process conditions.@xcite when assessing the importance of additional mechanisms it is essential to consider the change they bring about for the astrophysical rates . even if the ground state transitions ( as usually explored in theoretical and experimental nuclear physics )
may barely reach the required high energy for additional processes , such as semi - direct reactions , this may not affect the reaction rates much .
this is because they also include transitions from excited states according to ( [ eq : effrate ] ) and ( [ eq : effweights ] ) , which proceed at lower relative energy and thus remain unaffected by modifications of the cross sections at higher energy .
the hfm assumes that the compound nucleus has sufficient time to distribute the energy gained through the interaction with the projectile among all nucleons of the compound system before it decays . at a high compound formation energy @xmath319 of several tens of mev transitions occurring before this energy - equilibration lead to multistep reactions and pre - equilibrium emission of particles and
photons.@xcite the required compound excitation energy is too high to be of astrophysical relevance .
the second type of modifications introduced above is discussed in the following , focussing on two ideas .
the first is that subsequent emission of several particles may also occur when the final nucleus is produced at an energy above a further particle emission threshold .
astrophysically this can be relevant in nuclei close to the driplines with low separation energies .
it can be treated approximately by applying an iterative application of ( [ eq : hf ] ) , ( [ eq : tottrans ] ) , and ( [ eq : parttrans ] ) to transitions to states in each system formed in each emission process .
this has been used , for example , for calculating neutrino - induced particle emission which is relevant to the construction of neutrino detectors.@xcite in such calculations , the formation transmission coefficients @xmath424 in ( [ eq : hf ] ) are replaced by neutrino transmission coefficients @xmath425 describing the population of compound states by neutrino interactions ( using , e.g. , the random phase approximation ) .
neutrino reactions select high excitation energies and thus multi - particle emission will also be important at stability .
the same approach can be used to determine @xmath27-delayed particle emission when @xmath27-decay produces a daughter nucleus in an excited state above a particle - separation energy .
it can also be used to study @xmath27-delayed fission .
both processes are important in the r - process.@xcite the second idea revolves around the fact that regular hfm calculations assume a compound formation probability independent of the compound nld at the compound formation energy .
therefore the sum in ( [ eq : cs ] ) runs over all @xmath426 pairs ( a high - spin cutoff is introduced in practical application of the model because spin values far removed from the spins appearing in the initial and final nuclei do not contribute significantly to the transmission coefficients ) .
the availability of compound states and doorway states defines the applicability of the hfm.@xcite relying on an average over resonances , the hfm is not applicable with a low nld at compound formation .
furthermore , not all spins and parities will be available with equal probability at each @xmath319 , especially at low nld .
on average the hfm will then overpredict the resonant cross section ( unless single resonances dominate ) because it will overestimate the compound formation probability .
this can be treated by introducing a modification of the formation cross section which includes the compound nld dependence .
the summands of ( [ eq : cs ] ) will then be weighted according to the available number of states with the given @xmath426 .
( formally this is the same as assuming @xmath426 dependent potentials for particle channels . ) in the most general case this will require a dependence on a @xmath426-dependent compound nucleus nld at @xmath319 . note that the standard hfm only includes the nld of the compound nucleus in the photon transmission coefficients ( see sec .
[ sec : relevance ] ) to determine the endpoints of the @xmath82-transitions .
the assumption that all spins and parities are available can be lifted in several steps .
a parity - dependent , global nld @xmath392 was used in the calculation of the transmission coefficients for nuclei without experimentally determined excited states ( see sec . [ sec : relevance]).@xcite shortly thereafter , the parity - dependence of the compound formation was implemented in a modified hfm in version 4.0w of the code non - smoker@xmath427 ( see also sec . [
sec : codes]).@xcite a discussion of the implications of a parity - dependent compound formation for astrophysical neutron capture is given in ref . .
although the parities are not equidistributed up to sizeable excitation energies , the impact on _ stellar _ rates remains small ( in comparison to other uncertainties ) because the effective cross section ( [ eq : effcs ] ) also includes transitions from excited states which washes out the selectivity on parity .
this gives rise to factors of two modifications close to the neutron dripline ( but see sec .
[ sec : equilibria ] ) .
additionally , non - smoker@xmath427 offered the option of weighting the hfm cross section by a function depending on the total nld since version 4.0w.@xcite an improved implementation with @xmath426 dependent weighting of the summands , thus implicitly accounting for a low nld at the compound formation energy , is introduced in the smaragd code and will be used for a future update of large - scale reaction rate predictions.@xcite preliminary results with this modification are shown in fig .
[ fig : avdc ] in sec .
[ sec : avdc ] . obviously , these modifications of the hfm depend on the nld treatment to obtain the spin- and parity distribution .
see the paragraphs on the nld in sec .
[ sec : relevance ] for further details .
there are different ingredients required to calculate the hfm cross section with the formulas given in the preceding sections . which ingredients impact different parts of the calculation in what manner is discussed below , but how a change in the transmission coefficients ( of certain or all included transitions ) affects the resulting cross section
can only be understood with the help of ( [ eq : cs ] ) .
similar to the determination of the energy - dependence of the cross section ( which is crucial in the derivation of the relevant energy windows in sec .
[ sec : energies ] ) , the sensitivity of the cross section and rate to a change in the nuclear properties of the participating nuclei depends on which transmission coefficient ( or averaged width ) actually affects the cross section while the others cancel from ( [ eq : cs ] ) . as already pointed out in sec .
[ sec : energies ] , the discussion applies to both the bwf and the hfm .
thus , the sensitivities are rather well known in certain parts of the experimental community studying resonances and in the field of nuclear data evaluation . since they do not seem to be so well known in nuclear astrophysics and their implications also have to be interpreted in terms of stellar cross sections and astrophysical reaction rates , it is helpful to outline briefly the main points here . by comparison
to ( [ eq : hf ] ) and ( [ eq : hfrate ] ) we find that ( [ eq : cs ] ) transforms to @xmath428 for each @xmath429-dependent summand . in laboratory cross sections ( usually with @xmath112 ) only few @xmath429-summands contribute due to the spin selection rules , and so the application of @xmath430 to determine the sensitivities to changes in the widths is straightforward .
the situation is different for stellar rates and ratios @xmath431 because transitions from excited target states additionally contribute and more terms in the sum may be relevant , depending on the nucleus and the plasma temperature .
it is interesting to note that @xmath410 includes @xmath432 in both cases . depending on whether the average entrance transmission @xmath433 significantly defines the size of @xmath432 , a variation of the entrance transmission will affect the total transmission more or less
this can lead to a different sensitivity of laboratory cross sections to the entrance channel than to the transmission in the exit channel , even if everything else is comparable .
the interpretation of experimental results concerning the impact on stellar rates has also to proceed carefully .
for example , if a strong dependence is found on the entrance transmission coefficient and some deficiencies when comparing the model to experimental data , this does not necessarily mean that this is of relevance to the astrophysical rate . or
strength functions @xmath434 can be determined experimentally , it is more useful to obtain the latter because they are directly proportional to the transmission coefficients @xmath381 without an additional dependence on the compound nld.@xcite ] since the astrophysical rate involves @xmath432 also in the numerator , it may more often cancel with the denominator , even if it would not for @xmath433 in the numerator . only for rates with low sef @xmath435 .
this shows again that stellar rates and laboratory cross sections do not have a one - to - one correspondence and additional , mostly theoretical considerations have to be included
. several special cases can appear in ( [ eq : fraction ] ) : ( i ) the larger of the two transmissions in the denominator also dominates the numerator ; then the cross section or rate will change similarly to a change of the smaller transmission coefficient in the numerator and be oblivious to any others ; ( ii ) three channels will be important when neither of the transmission coefficients in the numerator dominates the total transmission ; then any change in the two transmissions in the numerator will be translated into an equal change in the cross section but a change in the one determining the denominator will result in an inverse proportional change in the cross section . in the general case when @xmath410 is not dominated by a single channel
the situation is more complicated as any change in a transmission coefficient will not fully affect the cross section at a similar level .
a helpful visual aid to estimate the relative importance of the different channels is a sensitivity plot .
i define the sensitivity @xmath436 as a measure of a change in the cross section @xmath437 as the result of a change in a transmission coefficient ( or width ) by the factor @xmath438 , with @xmath439 when no change occurs and @xmath440 when the cross section changes by the same factor as the transmission coefficient ( or width ) : @xmath441 plotting @xmath436 as a function of the c.m .
energy yields a plot like the example shown in fig .
[ fig : sensi ] for the reaction @xmath442ru(p,@xmath82)@xmath443rh .
its astrophysically relevant energy window is @xmath444 mev for the typical p - process temperature @xmath445 gk.@xcite it can clearly be seen in fig .
[ fig : sensi ] that the sensitivities are very different at lower and higher energies .
for example , a measurement closely below the neutron threshold would be in a region where @xmath436 is largest for the @xmath82 transmission coefficient ( or width ) but smallest for the proton transmission , just the opposite of what is found in the astrophysically relevant energy region . above the neutron
threshold the situation is even more complicated because there is additional sensitivity to the neutron channel ( dominating @xmath446 ) , although not as large as to @xmath447 . if any discrepancy between measured and predicted cross sections was found , it would be hard to disentangle the different contributions .
in this example , no information on the astrophysically important proton transmission coefficient can be extracted from a measurement at higher energies .
the above example also shows that some of the standard assumptions usually used in nuclear physics experiments do not apply .
for instance , it is usually assumed that the @xmath448width is always smaller than particle widths and therefore a capture reaction will always only be sensitive to the @xmath448width .
many , if not most , astrophysical reactions with charged particles , however , proceed close to or below the coulomb barrier .
this leads to very narrow charged particle widths and they may well become smaller than the @xmath448widths . even neutron widths close to a neutron threshold become smaller than photon widths ( which are relatively independent of energy compared to the particle widths ) . therefore it is always important to closely inspect the widths and to perform a thorough sensitivity study when investigating the astrophysical impacts of changes in the hfm inputs . from the general considerations above it also follows that it is advantageous to use reactions with neutrons in one channel for investigating the sensitivity to the charged - particle optical potential , i.e. , using ( @xmath0,n ) , ( n,@xmath0 ) , ( p , n ) , or ( n , p ) reactions . except within a few kev above the neutron threshold
, the neutron width will always be much larger than the charged - particle width , even at higher than astrophysical energies .
therefore it will cancel with the denominator in eq . and leave the pure energy dependence of the charged - particle width . on the other hand
, the neutron channel may not be open at the energies required to study the astrophysically important charged particle width and extrapolations have to be performed .
furthermore , this shows that it is difficult to obtain information on the neutron potential from reactions .
but this also implies that the sensitivity of astrophysical rates to the neutron optical potential is not high .
nuclear properties and how they affect the calculation of transmission coefficients ( or averaged widths ) are briefly discussed in the following .
this is by no means meant to be an exhaustive listing and discussion but rather some exemplary points are taken to explain the special requirements in the calculation of astrophysical rates and to point out the challenges .
further information on the nuclear input used in statistical model calculations for nuclear physics and nuclear astrophysics can be found , e.g. , in refs . and references therein .
while the hfm has been used extensively to study reaction data , its application in nuclear astrophysics has a slightly different focus .
standard nuclear physics investigations use the hfm by including measured or known properties of nuclei .
then reaction cross sections can be reproduced with high accuracy .
stellar cross sections not only require the modification of the hfm shown above but also the inclusion of mostly unknown further input because the reactions proceed at much lower energy and/or involve unstable nuclei .
the required properties have to be predicted _ globally _ for a large number of nuclei and it has to be realized which properties are important in which astrophysical process .
this is where the challenge lies and where basic research is necessary , going beyond a mere application of a seasoned model .
although intellectually the most satisfying , modern microscopic models still are not applicable for large - scale predictions and do not produce all the necessary data with sufficient reliability .
this calls for clever combinations of microscopic models and parameterizations ( which can also use dependences derived from microscopic calculations for a limited range of nuclei ) .
furthermore , additional information from experiments is required to constrain such models .
[ [ masses ] ] masses : + + + + + + + nuclear masses or , rather , mass differences determine separation energies and reaction @xmath94-values . in this way
, they also determine the range of transition energies to be considered in each reaction channel and through this the relative importance of a channel . except in the nse equations ( [ eq : nse ] )
, masses always appear in mass differences .
this poses a potential difficulty when a certain mass region is not fully explored experimentally yet .
care has to be taken to avoid artifical breaks and structures in the mass surface when calculating mass differences from a mix of experimental and theoretical masses .
this is usually considered in codes especially written for astrophysical applications . on the other hand , it is expected that mass differences can be measured as well as predicted with higher accuracy than single masses .
this seems reassuring for the calculation of astrophysical rates far from stability .
a change in the mass of a nucleus impacts the rates in two different manners .
first , the separation energies are altered in the reaction channels including the nucleus .
this leads to a change in the transition energies in these channels ( see fig . [
fig : cnscheme ] ) .
if the change is large , also more or fewer transitions may become possible .
although charged - particle transitions sensitively depend on the interaction energy , it has to be realized that even in this case the change in @xmath94-value _ plus _ ejectile energy has to be considerable to have a sizeable impact on the rate .
the entrance channel is not affected .
secondly , a change in the @xmath94-value changes the relation between forward and reverse rate as shown in ( [ eq : revrate ] ) and ( [ eq : revphoto ] ) .
it is highly sensitive to a small change in @xmath94-value due to the exponential dependence .
this impacts the temperature at which the two rates become comparable and at which equilibria are reached . on the other hand , if forward and reverse rates are different by many orders of magnitude ( i.e. for large @xmath280 ) , there may not be much astrophysical impact , nevertheless . [ [ properties - of - ground - and - excited - states ] ] properties of ground and excited states : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + it is standard procedure to include spin , parity , and excitation energy of low - lying discrete levels when calculating the transmission coefficients .
information about discrete states comes from experiments or from nuclear theory ( single particle states and shell model states ) .
close to stability , a large number of excited states are well known . however , often excited states of nuclides not produced in reactions on stable target nuclei are only partially known experimentally , even if the nuclide itself is stable .
obviously , the situation worsens further off stability .
therefore , nuclear spectroscopy is important to provide the database for reaction modelling .
discrete states are not only important in the hfm but also for resonant reactions treated in an @xmath314-matrix or bwf approach ( unbound states ) and even more for the treatment of direct reactions ( see sec . [ sec : direct ] ) ( bound states ) .
it is very important to have a complete set of excited states because a large number of missing levels at and below a given excitation energy would lead to an incorrect prediction .
therefore it is important to set a cut - off in the excitation energy above which no states are included ( even if there are some data ) and below which the assumption of a complete level set holds . above the cut - off ,
a nld is employed ( see below ) .
it is difficult to define a useful cut - off energy but blindly including all existing data leads to worse results than neglecting too many , provided a reliable nld description is used .
usually , an educated guess has to be made by comparing the level data to nld predictions .
[ [ nuclear - level - density ] ] nuclear level density : + + + + + + + + + + + + + + + + + + + + + + it should be noted that the level density @xmath115 used throughout this paper follows the `` experimental '' definition of number of observed levels per energy interval @xmath449 , where @xmath450 is the observed number of levels with spin @xmath320 and parity @xmath406 in a small energy interval around an excitation energy @xmath67 .
it is not to be confused with a state density @xmath451 appearing in microscopic nuclear theory , such as the shell model .
the terms `` level density '' and `` state density '' are used inconsistently in the literature . for the relation of the two types of density ,
see , e.g. , refs . . in the regular hfm
, the nuclear level density only enters in the calculation of the transmission coefficients ( eq . [ eq : tottrans ] ) when there are no or not enough discrete states known and therefore its importance depends on how many low - lying , discrete states were included ( see above ) . because of the energetics connected to particle emission ( see fig .
[ fig : cnscheme ] ) , usually only a small fraction of the particle width is due to transitions calculated with a level density at astrophysically relevant projectile energies as long as discrete excited states are known .
this is especially true for neutron - capture reactions due to their low interaction energies .
reactions with charged particles in entrance or exit channel prefer somewhat higher projectile energies due to the location of the relevant energy window ( sec .
[ sec : energies ] and thus may show slightly larger sensitivity to the level density in the target and final nucleus .
( these sensitivities are different from those encountered in usual nuclear reaction studies where projectile energies are much larger . ) on the other hand , there is a larger range of energies available for @xmath82-transitions ( see fig .
[ fig : cnscheme ] ; grey areas signify transitions calculated by integration over a level density , as shown in the second term on the right hand side of eq .
[ eq : tottrans ] ) and thus the impact of the level density in the compound nucleus will be largest . in most cases , it is accurate to assume that a variation of the level density will mostly ( or only ) affect the @xmath82-widths . generally , the impact of a change in the nld will be larger for nuclides with fewer known discrete states , i.e. far from stability . on the other hand ,
the @xmath94-values of the astrophysically relevant reactions become lower and thus ground state transitions to the final nucleus may again dominate .
it is instructive to remember which excitation energies are the most important ones . for particle transmission coefficients the transitions with highest relative energy are most important , i.e. to the ground state and the lowest excitation energies in both target and final nucleus ( similar considerations apply to direct reactions ) .
discrete levels or a nld have to be known there . for electromagnetic transitions ,
the relevant excitation energy is around @xmath452 in the compound nucleus ( see the discussion of electromagnetic transmission coefficients below and ref . for a detailed explanation ) . due to the low excitation energies
involved it is important to a ) include the correct ground state spin and parity , and b ) to account for a possible parity dependence of the nld at low excitation .
both is automatically ensured when experimental information is available up to sufficiently high excitation energies .
otherwise , ground state properties and nlds have to be predicted .
this introduces additional uncertainties in rates far from stability . until recently ,
astrophysical rate predictions made use of equally distributed parities @xmath453.@xcite modern rate predictions include parity - dependent nlds in various manners . either , microscopic nlds are directly utilized in calculations@xcite or a parity - dependence is applied to a total nld.@xcite the advantage of the latter approach is that it can conveniently be applied to total nlds from different sources and for a large number of nuclei .
the total nld @xmath115 is not changed but the parities are redistributed according to excitation energy .
this approach was used to study the impact of a parity - dependence across the nuclear chart . since
capture reactions mainly populate higher lying states , for which an equipartition of parities already is a good assumption , the impact of a parity - dependent level density is small , unless very low @xmath319 is encountered due to low projectile separation energies and low plasma temperatures.@xcite according to the discussion above , however , the impact is larger in particle emission channels and for direct reactions .
a modified hfm is introduced in sec .
[ sec : statmodmod ] above , which accounts for the relative level distribution at the compound formation energy @xmath319 .
again , the impact of a parity - distribution is small for sufficiently large @xmath319 and only becomes important far from stability at low separation energies ( in the r - process).@xcite however , it remains doubtful whether this actually is of astrophysical relevance as it is not clear whether the involved nuclei can be produced outside of equilibria ( see sec .
[ sec : equilibria ] ) and whether the uncertainty introduced by using the statistical model for nuclei with such low nld at @xmath319 is not much larger than the impact of the parity - dependence . on the other hand ,
a dependence on the total nld and on the @xmath320 distribution in the further modified hfm may be more important at a larger range of excitation energies and thus also for nuclei closer to stability ( see sec .
[ sec : statmodmod ] ) . in any case , the uncertainties introduced by the predicted nlds in rates far from stability are overall much smaller than the uncertainties stemming from other input to the hfm calculations , such as optical potentials and photon transmission coefficients . [
[ optical - potentials ] ] optical potentials : + + + + + + + + + + + + + + + + + + + optical potentials are required in the solution of the radial schrdinger equations to determine the particle transmission coefficients as shown in ( [ eq : parttrans ] ) .
together with the electromagnetic transitions strengths ( see below ) , the unknown optical potentials at low energy give rise to the largest uncertainties in astrophysical reaction rate predictions .
there is a combination of two problems involved in the determination of appropriate optical potentials for astrophysical applications : the prediction of optical potentials for highly unstable nuclei and the extension to the astrophysically relevant energies .
there is a large amount of reaction and scattering data along the line of stability and many parameterizations ( usually of the saxon - woods shape ) , partly mass- and/or energy - dependent , are available .
however , most scattering experiments to derive optical potential parameters have been performed at several tens of mev , far above the astrophysically relevant energy window ( see sec . [
sec : energies ] ) . even at stability
, there are almost no data ( not even reaction data ) for charged - particle reactions at astrophysical energies .
measuring low - energy cross sections for charged - particle reactions is especially problematic due to the coulomb barrier causing the astrophysically relevant cross sections to be tiny . even more problematic
is the standard way to obtain information on optical potentials through elastic scattering experiments .
the scattering cross section at low energy becomes indistinguishable from rutherford scattering . in the optical model , an imaginary part of
the potential appears whenever there is loss of flux from the elastic channel due to any kind of inelastic process .
the reaction cross section is sensitive to both real and imaginary part of the optical potential because they determine the relation of real and imaginary part of the nuclear wavefunction and thus the transmission coefficient.@xcite microscopic approaches to derive the optical potential are preferrable over parameterizations when predicting rates far from stability , especially because the available parameterizations were derived at far too high projectile energies .
especially the imaginary part of the potential may vary strongly with energy , due to the energy - dependence in the available reaction channels included in the imaginary part .
nevertheless , any more sophisticated theoretical approach also includes some parameters which have to be constrained by comparison with experiment and so even in this case there may be uncertainties at low energy ( see below and the further discussion of how additional reaction mechanisms impact the optical potential in sec . [
sec : directintro ] ) .
an optical potential widely used for interactions of nuclei with neutrons and protons uses the brckner - hartree - fock approximation with reid s hard core nucleon - nucleon interaction and adopts a local density approximation.@xcite a low - energy modification of this potential was provided specially for astrophysical applications.@xcite the latter has become the standard potential in predictions of astrophysical rates and is generally very successful when compared to the scarce experimental data at low energy . as stated in sec .
[ sec : sensigeneral ] , astrophysical rates are rather insensitive to the neutron potential . regarding rates involving charged particles in entrance or exit channel , an important difference to cross sections at higher energies is in the fact that astrophysical cross sections are mostly sensitive to the charged particle widths instead of the @xmath82- or neutron widths .
a series of ( p,@xmath82 ) and ( p , n ) reactions was measured close to astrophysical energies recently ( see , e.g. , refs . , and references therein ) .
the latter reactions are especially useful for testing the proton potential because the neutron width will ( almost always ) be larger than the proton width at all energies . despite of the overall good agreement when using the standard potential ,
certain systematic deviations at low energy were found recently ( see , e.g. , figs . [
fig : pg1 ] , [ fig : pg2 ] , more examples are shown in ref . ) .
it was found that an increase in the strength of the imaginary part at low energies considerably improves the reproduction of the data ( denoted by `` mod jlm '' in figs .
[ fig : pg1 ] , [ fig : pg2 ] , [ fig : dens1]).@xcite an increased absorption is permitted within the previous parameterization because the isoscalar and especially the isovector component of the imaginary part is not well constrained at low energies.@xcite therefore , the change has to be energy - dependent ( i.e. acting only at low energy ) .
more low - energy data is required to obtain a better picture .
figures [ fig : pg1 ] , [ fig : pg2 ] also shows results obtained with another recent , lane - consistent new parameterization of the jlm potential.@xcite although it showed improved performance at higher energy , it yields worse agreement at astrophysically low energy .
this is understandable as neither does it include the additional modifications of ref . ,
nor can it constrain well the low energy part because it was fitted to data at higher energy.@xcite similar considerations apply to another recent reevaluation of the standard potential.@xcite required input to the calculation of this type of optical potentials is the nuclear density distribution @xmath454 .
figure [ fig : dens1 ] shows results when employing a droplet model density@xcite and one from a hartree - fock - bogolyubov model ( hfb-2).@xcite for the reactions considered here , the droplet description yields better agreement to the data in both absolute scale and energy dependence of the theoretical s - factor . for comparison ,
[ fig : dens2 ] also shows the results when employing the optical potentials of refs . with both densities .
in the original work , hfb densities were employed.@xcite for further information on how nuclear density distributions affect the astrophysical rates see the subsection on density distributions below , including figs .
[ fig : pgdens1]@xmath215[fig : ngdens2 ]
. the standard potential for neutrons and protons seems to work very well compared to the situation encountered when exploring the adequacy of @xmath0+nucleus optical potentials for astrophysics .
global parameterizations describing scattering , reaction , and decay data have been notoriously hard to find for @xmath0-particle potentials .
somewhat surprisingly , a mass- and energy - independent potential of saxon - woods type has been quite successful and is widely used to evaluate reaction data and also for astrophysical applications.@xcite the potential was fitted to scattering data at 26.7 mev @xmath0-energy for a wide range of nuclei .
however , it became obvious early on that it may be impossible to find a global potential with a predictive power comparable to those for nucleonic projectiles , especially at low energy.@xcite the number of optical potential parameters can be reduced by using folding potentials @xmath455 for the real part,@xcite @xmath456 in this expression @xmath22 is the separation of the centers of mass of the two interacting nuclei , @xmath457 and @xmath458 are their respective nucleon densities and @xmath21 is an adjustable strength factor .
the factor @xmath21 may differ slightly from unity because it accounts for the effects of antisymmetrization and the pauli principle .
the effective nucleon - nucleon interaction @xmath459 for the folding procedure is usually of the ddm3y type.@xcite density distributions have to be taken from experiment or theory ( see the section on matter density distributions below ) .
a global parameterization of the real part with such folding potentials was found based on extensive scattering data.@xcite unfortunately , there is no simple description for the imaginary part , for which shape and strength have to be energy - dependent .
correlations with the compound nld , different parametrizations for the energy - dependence of the strength ( fermi - function , brown - rho dependence ) , and an energy - dependence in the relative strength of volume and surface imaginary parts have been suggested.@xcite again , those extrapolations to low energy are only loosely constrained due to the lack of scattering data . based on apparently different potentials required for the description of @xmath0-particles in entrance and exit channel and the fact that a potential fitted to reaction data is able to describe a number of reactions but does not reproduce scattering data ,
it was suggested that there may be some dependence on nuclear temperature and that absorption and emission potentials may be different from scattering potentials.@xcite moreover , at energies close to or below the coulomb barrier , so - called `` threshold anomalies '' have been observed , a rapid variation of optical potential parameters with energy.@xcite it has been shown that the dispersion relation connecting real and imaginary part of the optical potential is essential to describe these.@xcite there is a large literature on different local and global parameterizations of @xmath0+nucleus potentials , mostly at high energy , underlining the lack of a coherent nuclear physics treatment .
a complete review can not be provided here but see , e.g. , refs . and references therein for further details . for astrophysical applications ,
early reaction rate predictions made use of an equivalent square well potential.@xcite later on , the potential of ref . was used . with accumulating evidence that low - energy data deviates from the predicted cross sections more complicated parameterizations were tried , based on various combinations of scattering and reaction data , but these did not lead to a consistent picture .
on the other hand , the potential by ref .
( see also refs . )
was fitted to simultaneously reproduce low - energy reaction cross sections of @xmath460nd(n,@xmath0)@xmath461ce , @xmath462sm(n,@xmath0)@xmath463nd , and @xmath463sm(@xmath0,@xmath82)@xmath464gd .
although the potential does not describe scattering data ( at higher energy ) it was found to work surprisingly well for low - energy cross sections for target nuclei across a large mass range @xmath465.@xcite sharing the same imaginary part with the potential of ref . , it predicts systematically lower cross sections at low energy due to a shallower real part .
although not fully satisfactory yet , this shows that the main change required by a global @xmath0-potential is to reduce the predicted low - energy cross sections by factors of about @xmath466 .
the only and remarkable exception to these factors known so far is in the comparison to experimental data for @xmath463sm(@xmath0,@xmath82)@xmath464gd.@xcite this reaction is important in the astrophysical context not just generally in p - process calculations@xcite but specifically in deriving astronomical timescales from the nd / sm abundance ratios measured in meteoritic inclusions or , vice versa , to determine production ratios of these elements in supernovae.@xcite the astrophysical s - factors were measured from @xmath467 mev , the astrophysically relevant energy window extends from 9 mev downwards . above 11.5 mev the s - factors obtained with the potential of ref .
are too high by a factor of three but the data is well described with the potential of ref . and
a potential derived from scattering data at 20 mev.@xcite however , below 11.5 mev the energy - dependence changes and requires further modifications of the potentials . at the lowest measured energy of 10.2 mev ,
the measured s - factor differs by a factor of 10 from the one predicted with the potential of ref . . in ref .
a potential with energy - dependent imaginary part was fitted to describe the data .
it predicts s - factors at and below 9 mev which are orders of magnitude lower than those predicted with the global potentials .
although the original problem was a too low calculated nd / sm ratio as compared to what is found in meteorites , such low s - factors yield much too high ratios and additional astrophysical dilution effects have to be invoked in an ad hoc manner.@xcite the s - factors are shown in fig .
[ fig : sm144a ] , where also new results obtained with the recent potential of ref . are included .
it should be noted that this reaction may be a special case because it is strongly endothermic .
nevertheless , it was shown that the stellar enhancement factor is lower in the capture direction ( see also sec . [
sec : stellarexp]).@xcite since the extrapolation to astrophysical energies strongly hinges on the data points at the lowest energies , an independent remeasurement of this reaction at comparable or lower energies is highly desireable . further issues in the determination of s - factors below the coulomb barrier are illustrated in figs .
[ fig : sm144a]@xmath215 [ fig : sm144rcoul2 ] .
figure [ fig : sm144a ] shows the results obtained when using the routine of ref .
( as also used in refs . ) for the solution of the schrdinger equation and the determination of the @xmath0-particle transmission coefficients
. it does not work properly when applied at energies at or below the coulomb barrier as a comparison to results obtained with a modern routine shows ( fig .
[ fig : sm144b ] ) .
the discrepancy between predictions and measurement is even enhanced when using an appropriate method . since the s - factors are very sensitive also to the wavefunction far outside the nuclear center , they show a strong dependence to the shape and width of the effective barrier ( determined by the sum of nuclear potential and coulomb potential ) .
results obtained with the optical potentials of refs . and with fine - tuned energy - dependences of the real and/or imaginary part also have a strong sensitivity to the coulomb radius parameter , as shown in figs .
[ fig : sm144rcoul1 ] and [ fig : sm144rcoul2 ] .
this is often overlooked because the sensitivity is much lower at higher energies . also , when using the potentials of refs .
there is no sensitivity to the coulomb radius in the investigated energy range .
these additional complications show that any extrapolation to subcoulomb energies has to be performed very carefully and that it is difficult to construct a global potential .
further experimental data ( on scattering and reactions ) are especially in demand for improving the optical potentials ( see also the discussion of additional reaction mechanisms contributing to the absorptive part of the potential in sec . [
sec : directintro ] ) . currently , progress is hampered by the lack of systematic reaction ( and scattering ) data at astrophysically relevant energies , even at stability . [
[ electromagnetic - transitions ] ] electromagnetic transitions : + + + + + + + + + + + + + + + + + + + + + + + + + + + + first , a few words on the energies of the emitted @xmath82-rays and their significance for changes in @xmath82-ray strength functions ( drawn from experiment or from theory ) are in order . as shown in fig . [
fig : cnscheme ] , the energies of emitted @xmath82-rays are in the range @xmath468 .
therefore , the behavior of @xmath448strength functions at low energy have to be known . since the strength of the @xmath82-transition scales with some power of @xmath469 , @xmath82-transitions with higher energies are favored .
on the other hand , the number of available endpoints of the transitions increases with increasing excitation energy of the nucleus because the nld increases rapidly .
this competition between transition strength and nld gives rise to a peak in the @xmath82-emission energies as shown in fig .
[ fig : gammapeak ] .
this peak is fragmented when certain transitions to discrete excited states are dominating .
this is mainly the case far off stability for captures with low @xmath94-values , forming a compound nucleus with low nld.@xcite figures [ fig : gammaenergies1 ] , [ fig : gammaenergies2 ] show examples for the @xmath82-energies which maximally contribute to the reaction rate integral .
interestingly , it can be seen that for astrophysically relevant projectile energies , the @xmath82-energies with the strongest impact are between @xmath470 mev unless the level density is so low that only few transitions are allowed ( usually for highly unstable nuclei).@xcite then the relevant @xmath471 is @xmath472 ( in the figures this is the ground state with @xmath473 ) instead of the almost constant value below @xmath319 .
the @xmath82-emission peak defines the range of @xmath82-energies at which changes in the strength function have largest impact as well as the excitation energies at which the nld is most important .
this also holds for the reverse reaction ( photodisintegration ) under stellar conditions because the additional , linear weight @xmath474 from ( [ eq : effweights ] ) has a much weaker @xmath471 dependence then both the nld and the @xmath82-strength . in consequence , changes in the strength function around this energy have the largest impact . testing strength function models outside the energy range will not be relevant to astrophysics .
unfortunately , such low energies can not be probed by photodisintegration experiments because they are below the particle separation energies , at least close to stability .
such experiments would allow to study strength functions in the most direct way ( but they can not test the astrophysically relevant rates , either , see sec . [
sec : stellarexp ] ) .
other types of experiments are complicated by the fact that the observables are generated by a convolution of different nuclear properties ( such as the dependence on the nld or different spin selectivities of transitions ) which have to be known and disentangled .
at least the dominant @xmath82-transitions ( e1 and m1 ) have to be included in the calculation of the total photon width for astrophysics .
some codes offer the possibility of including higher order transitions .
there are two issues involved : 1 ) obtaining , understanding , and modeling photon strength functions ( psf ) at stability ; 2 ) predicting strength functions far from stability by using parameterized or microscopic models , predicting the nuclear properties ( e.g. , deformation ) entering the descriptions . despite of decades of experience in studying nuclear reaction data , the understanding of the electromagnetic transitions between nuclear states
is limited and predictions are subject to considerable uncertainties , even at stability.@xcite the paradigm in the field of studying electromagnetic transitions is the validity of the reciprocity theorem ( [ eq : reci_single ] ) also when applied to photon emission and absorption , and the independence of the psfs from the nuclear structure of the initial and final states ( except for spin and parity , selecting the allowed multipolarity of the radiation).@xcite this is called the brink hypothesis.@xcite this is also the basis for the construction of the effective cross section ( [ eq : effcs ] ) , the expression for the reverse rate ( [ eq : revphoto ] ) , and the introduction of equilibrium abundances ( sec .
[ sec : equilibria ] ) .
the brink hypothesis has been studied extensively in experiments and its violation would have grave consequences not only in nuclear reaction theory but also for astrophysical reaction rates and network calculations . among the collective modes of nuclei the electric dipole ( e1 )
excitation has the special property that most of its strength is concentrated in the isovector giant dipole resonance ( gdr ) .
macroscopically , this strong resonance is described as a vibration of the charged ( proton ) matter in the nucleus against the neutral matter ( neutrons ) .
the transmission coefficient in a nucleus with charge number @xmath212 , neutron number @xmath211 , and mass number @xmath475 can be parameterized as@xcite @xmath476 here , @xmath477 is the proton mass , @xmath478 accounts for the neutron - proton exchange contribution,@xcite and the summation over @xmath15 includes two terms which correspond to the split of the gdr in statically deformed nuclei , with oscillations along ( @xmath479 ) and perpendicular ( @xmath480 ) to the axis of rotational symmetry . in this deformed case , the two resonance energies are related to the mean value calculated by the relations@xcite @xmath481 the deformation parameter @xmath309 is the ratio of the diameter along the nuclear symmetry axis to the diameter perpendicular to it , and is obtained from the experimentally known deformation or from mass model predictions .
many microscopic and macroscopic models have been devoted to the calculation of the gdr energies @xmath482 and widths @xmath483
. of special interest here is the low - energy tail of the gdr .
it is a long - standing question of nuclear physics to specify how much of the e1 strength is still present at energies far below the gdr maximum , which also encompasses the astrophysically relevant energy region .
theoretically it has been shown that it is justified to describe the gdr by a lorentzian also below the particle emission thresholds.@xcite various experimental attempts to determine the low - energy extension of the gdr for heavier nuclei have led to conflicting results .
neutron - capture experiments often have indicated an overshoot of the lorentzian over the observed e1 strength at the low - energy tail of the gdr.@xcite on the basis of these data theoretical explanations have been proposed to explain the differences.@xcite photon - scattering experiments , however , are in many cases in good agreement with the lorentzian extrapolation.@xcite unfortunately , they do not access the astrophysically relevant energies . also other experiments are used to extract psfs , e.g. , @xmath484he - induced reactions.@xcite recent investigations have shown that the e1 strength can be described by lorentzians for a large range of nuclei.@xcite it was also shown that indirect determinations of pdfs are prone to large uncertainties due to the experimental difficulties and some claims of enhancement@xcite at low @xmath471 have been premature.@xcite however , in some nuclei extra strength at low energy with respect to the smooth lorentzian was found consistently and denoted as `` pygmy dipole resonance '' ( pdr).@xcite the pdr has experimentally been studied so far in spherical nuclides around @xmath485 , @xmath486 , and in the doubly magic @xmath487pb.@xcite theoretical approaches describe the pdr as caused by an oscillation of excessive neutrons against the symmetric proton - neutron system ( see , e.g. , refs . ) .
other oscillation modes were also proposed ( e.g. , scissor modes ) which may also add strength beyond the lorentzian tail of the gdr . in any case , extra strength in the low - energy tail only has an astrophysical impact if it is within the relevant energy range defined above .
although a pdr may lead to an increase by several orders of magnitude in the astrophysical capture rate,@xcite this depends sensitively on its location and width .
different models give varying predictions of these crucial properties .
depending on the microscopic model used , the pygmy resonance is sometimes predicted at too high or too low an energy as to have any astrophysical consequence.@xcite further investigation of this issue is required .
it is important to note that the uncertainties in predicting the pdr enter additionally to the general uncertainties still present in the prediction of gdr energies and widths . together with the predictions of optical potentials for the particle transmission coefficients , these are among the largest uncertainties in the determination of astrophysical rates .
psfs of higher multipole order are even less studied due to their small contributions to cross sections .
there are several descriptions available for m1 transitions , starting from psfs from the simple single particle approach to more sophisticated ( but less thoroughly tested ) models.@xcite unless energy - independent psfs are employed , a relevant energy window similar to the one for e1 transitions will arise .
[ [ isospin ] ] isospin : + + + + + + + + isospin conservation restricts transitions to certain final states with the same isospin as the initial and compound states , i.e. @xmath488 .
isospin conservation is not absolute and cross section measurements of isospin - forbidden reactions give an estimate of the size of the isospin breaking ( or isospin mixing ) .
internal isospin mixing due to the coulomb interaction and external mixing via the other reaction channels have to be distinguished .
the hfm equation as shown in ( [ eq : hf ] ) with the transmission coefficients ( [ eq : tottrans ] ) does not account for isospin conservation unless it is included in the transmission coefficients .
in other words , complete isospin mixing is assumed .
the calculation of the transmission coefficients can be generalized to explicitly treat the contributions of the dense background states with isospin @xmath489 and the isobaric analog states with @xmath490.@xcite in reality , compound nucleus states do not have unique isospin and for that reason an isospin mixing parameter @xmath491 was introduced which is the fraction of the width of @xmath492 states leading to @xmath493 transitions.@xcite for complete isospin mixing @xmath494 , for pure @xmath493 states @xmath495 . in the case of overlapping resonances for each involved isospin
, @xmath496 is directly related to the level densities @xmath497 and @xmath498 , respectively .
isolated resonances can also be included via their internal spreading width @xmath499 and a bridging formula was derived to cover both regimes.@xcite in order to determine the mixing parameter @xmath500 , experimental information for excitation energies of @xmath492 levels can be used where available.@xcite experimental values for spreading widths are also tabulated.@xcite inspection of the tables shows that internal mixing dominates and that the associated spreading width is nearly independent of mass number and excitation energy , facilitating the extrapolation to unstable nuclei.@xcite similarly to the standard treatment for the @xmath493 states ( the regular transmission coefficients as shown above ) , a nld can be invoked above the last experimentally known @xmath492 level . since the @xmath492 states in a nucleus ( @xmath212,@xmath211 ) are part of multiplet , they can be approximated by the levels ( and nld ) of the nucleus ( @xmath212@xmath2151,@xmath211 + 1 ) , only shifted by a certain energy @xmath501 .
this displacement energy can be calculated and it is dominated by the coulomb displacement energy @xmath502.@xcite thus , the uncertainties involved are the same as in the prediction of the nld and discrete excited states .
the inclusion of the explicit treatment of isospin has two major effects on statistical cross section calculations in astrophysics which will be discussed below:@xcite the suppression of @xmath82-widths for reactions involving self - conjugate nuclei and the suppression of the neutron emission in proton - induced reactions .
non - statistical effects , i.e. the appearance of isobaric analog resonances , can be included in the treatment of the mixing parameter @xmath496 but will not be further discussed here.@xcite the isospin selection rule for e1 transitions is @xmath503 with transitions @xmath504 being forbidden.@xcite an approximate suppression rule for @xmath488 transitions in self - conjugate nuclei can also be derived for m1 transitions.@xcite in the case of ( @xmath0,@xmath82 ) reactions on targets with @xmath505 , the cross sections will be heavily suppressed because @xmath506 states can not be populated due to isospin conservation . a suppression will also be found for capture reactions leading into self - conjugate nuclei , although somewhat less pronounced because @xmath506 states can be populated according to the isospin coupling coefficients .
this cross section suppression can be implemented as a suppression of the photon transmission coefficient .
some older reaction rate calculations treated this suppression of the @xmath82-widths completely phenomenologically by dividing by rather arbitrary factors of 5 and 2 , for ( @xmath0,@xmath82 ) reactions and nucleon capture reactions , respectively.@xcite this can be improved by explicitly accounting for population and decay of @xmath507 and @xmath508 states , and considering isospin mixing by the parameter @xmath496.@xcite an astrophysically important reaction with pronounced isospin suppression effect is the reaction @xmath509ca(@xmath0,@xmath82)@xmath510ti which is responsible for the production of the long - lived @xmath510ti in supernovae.@xcite decay @xmath82-emission of @xmath510ti is observed in supernova remnants and can be used to test supernova models.@xcite furthermore , assuming incomplete isospin mixing , the strength of the neutron channel will be suppressed in comparison to the proton channel in proton - induced reactions.@xcite this leads to a smaller cross section for ( p , n ) reactions and an increase in the cross section of ( p,@xmath82 ) reactions above the neutron threshold , as compared to calculations neglecting isospin ( i.e. implicitly assuming complete isospin mixing with @xmath494 ) .
the isospin mixing parameter was varied in the theoretical investigation of a @xmath511v(p,@xmath82)@xmath512cr experiment.@xcite it was found that complete isospin mixing closely reproduced the measured cross sections when width fluctuation corrections were considered .
width fluctuation corrections affect the ( p,@xmath82 ) cross sections above as well as below the neutron threshold , whereas incomplete isospin mixing only reduces the cross sections above the threshold .
thus , the two corrections can be discriminated . mainly from this result
it was concluded that
contrary to width fluctuation corrections isospin can be neglected . however
, a closer investigation of the @xmath492 levels in @xmath512cr ( using results from refs . ) shows that isospin mixing should be rather complete already at the neutron threshold ( since the first @xmath492 state is almost 1 mev below the threshold).@xcite this is also true for lighter targets . for reactions on heavier nuclei ( @xmath513 ) , however , the neutron and proton threshold , respectively , will still be in a region of incomplete isospin mixing and therefore isospin effects should be detectable there . on the other hand , this effect is not as important in the calculation of astrophysical reaction rates as the suppression of the @xmath82-width because of the averaging over an energy range in the calculation of the rate , washing out the cusp effect .
[ [ nuclear - matter - density - distribution ] ] nuclear matter density distribution : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the density distribution @xmath454 of neutrons and protons inside a nucleus is needed to calculate the optical potentials for some choices of potentials ( see above ) .
there are charge density distributions available from electron scattering experiments@xcite on stable nuclei but the majority of density distributions for application across the nuclear chart comes from theoretical predictions .
rates are mostly sensitive to optical potentials , and thus densities , at large radii because of the low astrophysical energies .
density distributions are required in the determination of optical potentials and the potentials used in direct capture calculations .
some optical potentials rely on nuclear density distributions from a certain model which were used when fitting the remaining open parameters to experimental data.@xcite although modern microscopic models have considerably improved in predicting nuclear masses and radii , the differences between different models still are large .
because of the sensitivity of the transmission coefficients to the optical potentials , one would expect that small differences in the nuclear density distributions can give rise to large differences in the rates , especially at low plasma temperatures . as an example , figs .
[ fig : pgdens1 ] , [ fig : pgdens2 ] show comparisons of ( p,@xmath82 ) rates obtained when using densities from hartree - fock - bogolyubov ( hfb ) models and from the droplet model in the jlm potentials while leaving all other input unchanged.@xcite densities from hfb-2 were included in the recommended input parameter library for hauser - feshbach calculations ripl-2 ( ref . ) and the hfb-14 densities are in its successor ripl-3 ( ref . ) .
the left panel shows the comparison of rates obtained within the `` family '' of hfb models . the right panel shows a comparison of rates obtained with droplet model and hfb-14 densities . while the ratios stay well within a factor of two even with the seasoned droplet densities , it is interesting to note that they reach unity when approaching the proton dripline .
figures [ fig : ngdens1 ] , [ fig : ngdens2 ] are the same as above but for ( n,@xmath82 ) rates at @xmath175 gk with stable and neutron - rich target nuclei . here
, the largest ratios appear for very neutron - rich nuclei but again recede to unity when approaching the neutron dripline .
overall , the maximal ratios are higher than for proton capture but stay within a factor of ten .
this is partly due to the fact that nuclei further from stability are involved where the disagreement between different models becomes larger .
the main reason , however , is the lower relevant temperature because the differences in the rates due to the density distributions decrease with higher @xmath41 , as the average transmission coefficients become less sensitive to the nuclear surface region .
a further example of the impact of nuclear density distributions on calculated reaction @xmath65-factors is shown in figs .
[ fig : dens1 ] , [ fig : dens2 ] in the section on optical potentials above .
[ [ deformation ] ] deformation : + + + + + + + + + + + + deformations are implicitly present when taking excited states from experiments or theory , or nuclear density distributions from theoretical models .
there the problem lies in the fact that most results of microscopic models available for large - scale calculations assume sphericity .
the hfm using transmission coefficients as described in ( [ eq : tottrans ] and ( [ eq : parttrans ] ) can not describe reactions on deformed nuclei because it assumes @xmath377 to be a good quantum number .
the coupled - channel model has to be invoked for a rigorous treatment.@xcite it is computationally very expensive and thus not suited for large - scale calculations .
furthermore , the effective cross sections ( [ eq : effcs ] ) require to include even more transitions than in standard , laboratory nuclear reactions .
fortunately , it has been shown that experimental data can be well described in a spherical hfm using an effective optical potential which is obtained by averaging over nuclear orientation.@xcite this leads to a spherical potential with larger diffuseness . to compute this modified potential , the explicit inclusion of the nuclear deformation is required and is usually taken from microscopic or macroscopic - microscopic models .
a deformation parameter may also enter the description of the nld employed in the hfm calculation .
but certain nld descriptions include the deformation implicitly , like the one of ref . where the deformation is contained in the microscopic correction .
the splitting of the gdr in deformed nuclei can also be accounted for phenomenologically by a dependence on a deformation parameter .
this has the largest impact on the rates among the possibilities of the inclusion of deformation discussed here .
finally , fission transmission coefficients ( see below ) are also very sensitive to deformation parameters . usually , the deformation is already included in an effective fission barrier , leading to double - humped fission barriers.@xcite [ [ width - fluctuation - corrections ] ] width fluctuation corrections : + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the width fluctuation coefficients ( wfc ) defined in ( [ eq : wfc ] ) impact the reaction cross section only closely around channel openings , with a few kev to tens of kev .
contrary to isospin competition cusps ( see above ) , the modify the cross section above and below the channel threshold.@xcite generally , they enhance the elastic channel and reduce the other open channels accordingly to obey flux conservation . above the neutron threshold the behavior may be not so obvious because of the dominance of the neutron channel with respect to other channels .
the transfer of strength from the dominant neutron channel to the elastic channel results in a marked reduction in competition with other exit channels and actually increases the strengths in those other channels .
there are different ways to implement the calculation of the wfc , depending on the assumptions taken.@xcite an explicit form for the @xmath372 , requiring the knowledge of the level width distribution in the compound nucleus , is obtained by assuming that all transmission coefficients are small.@xcite this approach has been widely used , even when the transmission coefficients were not small.@xcite an alternative approach is to recognize that the main effect of the correlations is in the elastic channel , with smaller effects on the other channels .
this leads to a modification of the transmission coefficients and an additional factor applied in the elastic channel only .
this is the more general hrtw method which has been used in most astrophysical applications.@xcite another implementation of the hrtw method gives a general formula without restricting assumptions for the additional factor but is complicated to apply.@xcite overall , in astrophysical rates the impact of the wfc is small because of the low energies encountered . even when the relevant energy window covers the neutron threshold , the difference when choosing one or the other description is barely noticeable because of the energy averaging taking place in the rate integration .
the wfc are important , however , when trying to compare theoretical and experimental cross sections close to channel openings .
[ [ fission ] ] fission : + + + + + + + + neutron - induced ( and @xmath218-delayed ) fission of extremely neutron - rich nuclides is important in determining the endpoint of the r - process and the amount of intermediate nuclei produced by the fission process.@xcite it can be included in the hfm by using an additional exit channel in ( [ eq : tottrans ] ) describing the fission process .
the fission transmission coefficient is calculated from the penetration probability through a fission barrier . since most of the astrophysical fission occurs at energies below or close above the barrier , the resulting rates are very sensitive to the height and width of the fission barrier .
barrier predictions from various models show large differences and thus there are considerable uncertainties ( reaching several orders of magnitude ) in the resulting fission rates .
these uncertainties have been explored in ref . and recommendations for comparative rate sets were given .
the prediction of fission barriers remains a challenge to current microscopic models .
another issue concerns the fission fragment distribution .
earlier studies of the r - process have used simple fission barriers and assumed symmetric fission.@xcite recent years have seen the advent of improved predictions using more sophisticated statistical models and their results are being included in rate calculations.@xcite as for the barriers , considerable uncertainties may exist for extremely neutron - rich nuclei , though .
on one hand , fission determines how far the r - process can synthesize elements and whether it could reach the region of long - lived superheavy nuclei . on the other hand ,
the fission rate together with the fragment distribution impacts the abundances of intermediate and heavy r - process nuclei .
intermediate r - process nuclei ( including the rare earth peak ) have a contribution from fission fragments.@xcite if the fission timescale is short compared to the r - process timescale , fission cycling can occur whereby the fission fragments capture neutrons and follow an r - process evolution until they fission again.@xcite this exponentially enhances the final r - abundances .
the number of possible fission cycles depends on the fission rates and thus is very sensitive to the fission barriers .
the final abundance level is less sensitive to the fragment distribution but the distribution will determine the details of which nuclei receive contributions from fission . in principle , one would think that any code implementing the hfm should give the same result .
this is obviously true when using the same implementation of the hfm _ and _ the same descriptions of the required input .
however , both may differ among different codes . when quoting results it is therefore essential to not only always specify the exact version of the used program but also what selections regarding the properties described above have been made .
there is a variety of codes which have been and are used in data evaluation .
these usually focus on higher energies than astrophysically relevant .
they also may include further reaction mechanisms which may not be relevant for astrophysical application ( see sec .
[ sec : statmodmod ] ) .
finally , they use experimental knowledge ( either directly or by renormalizing theoretical results ) or local parameterizations of nuclear properties . this way , high accuracy may be achieved locally for one or a few nuclei but no global prediction , essential for astrophysics , can be made .
astrophysical codes are especially written for global predictions of reaction rates and thus focus on low - energy cross sections .
this includes using different internal numerics ( see also figs . [
fig : sm144a]@xmath215[fig : sm144rcoul2 ] for issues regarding the calculation of subcoulomb s - factors ) but also different choices of the used input .
global treatments by global parameterizations or ( semi-)microscopic models are preferred .
experimental information may be used where available to test those global approaches and , of course , to locally improve astrophysical reaction rates . however , a fair comparison of the global predictions of different astrophysical and other codes is only possible when a similar philosophy is used in determining their input ( and by using the same experimental input where unavoidable ) .
most importantly , however , astrophysical codes directly account for the additional transitions required for the calculation of the true astrophysical reaction rate with thermally excited nuclei in the stellar plasma .
they properly include the effective cross section defined in ( [ eq : effcs ] ) and thus implicitly use the correct weighting factors of transitions from excited states as derived in ( [ eq : effweights ] ) .
early large - scale reaction rate and cross section predictions made use of a code developed at caltech.@xcite the original code@xcite was developed further@xcite but no further tables of reaction rates were published , ready for use in astrophysical reaction networks ( see also ref . for a comparison of these early codes ) .
the work of refs .
was not only important for astrophysical modeling but also in nuclear physics .
prior to these calculations , all hitherto experimentally studied reactions had featureless excitation functions and tests of the statistical theory were hindered by this .
the large - scale calculations allowed to identify the cases suited to study competition between different channels.@xcite another influential development is the one of the smoker code.@xcite it went beyond the previous codes by including a more sophisticated calculation of the transmission coefficients in all channels by explicitly solving the schrdinger equation with optical potentials and including several new global parameterizations of the further inputs .
a set of neutron - capture rates for r - process nucleosynthesis was published in ref . and
complete sets of neutron - induced as well as charged - particle induced reactions calculated with this code became included in the first version of reaclib , a library of theoretical and experimental cross sections which can directly be used in astrophysical reaction networks.@xcite the reaclib format , using fits of reaction rates to a function of 7 temperature - dependent terms ( see , e.g. , ref . ) , has become a standard widely used in the astrophysics community.@xcite a series of codes some closely related , other only loosely based on smoker has appeared since then .
the code non - smoker extended the functionality by not only updating the input data but also including an improved , global nld description and isospin effects.@xcite although it already included many possibilities for descriptions of nuclear properties to use , often the term `` non - smoker calculations '' is used synonymously for the extended tables in refs .
calculated with a chosen input set . in a parallel development ,
the code most also updated the input physics and provided a different selection of treatments of nuclear properties.@xcite tables of reaction rates ( but no fits in reaclib format ) were provided online for several versions of the code.@xcite the newly written , but still closely related , non - smoker@xmath427 code included several changes.@xcite apart from several updates of included nuclear data , also the internal numerical calculations were modified , the isospin suppression was improved , additional choices for microscopic and macroscopic predictions of nuclear properties were offered .
the innovative web interface allows access to the code from anywhere through a simple web browser .
additional switches can be set and different nuclear properties provided in an optional input file which is uploaded to the server running the code .
the resulting cross sections and reaction rates are immediately displayed .
certain nuclear properties , e.g. optical potentials , can not only be uploaded as data but also as formulae because the code includes a simple equation parser .
non - smoker@xmath427 has been used for the astrophysical analysis of a large number of experimental results , especially for p - process nucleosynthesis ( see , e.g. , refs . and references therein ) .
its development has been frozen at version v5.8.1w but it is still available and used for calculations .
most recently , the code smaragd ( statistical model for astrophysical reactions and global direct reactions ) continues and extends the development initiated with non - smoker@xmath427 of a user - friendly , easily extensible code tailored for astrophysical reaction rates.@xcite the code is written completely in fortran90/95 ( with exception of the routines handling the web interface and the function parser , which are written in c ) and has a modular structure , making changes easy .
the latest nuclear data can also swiftly be included through files or web downloads .
internally , the numerics and solvers for the schrdinger equation have been improved to be more accurate at low , subcoulomb energies and to consistently also calculate direct processes as required for astrophysics ( see sec .
[ sec : directintro ] ) .
recent developments in parameterized or microscopic predictions of global nuclear properties ( masses , nlds , density distributions , optical potentials , photon strength functions ) have been included .
it also uses the modified hfm discussed in sec .
[ sec : statmodmod ] .
reaction rates are provided in tabular form as well as in the reaclib format through an automated fit routine .
future versions will include multiple particle emission , follow @xmath82-cascades explicitly , and allow the calculation of fission rates .
code versions below v1.0s are not public , those below v2.0s do not include direct reactions , yet .
direct reactions are included in a number of ways , as discussed in sec .
[ sec : direct ] below . at later stages of the code development
it is planned that users may upload modules providing nuclear data , numerical methods to compute required properties , or even altering the functionality of the program . however , before these improvements in versatility are made accessible , a new large - scale calculation will provide a new set of published reaction rates between the driplines , intended to improve on and supersede the non - smoker rates@xcite which currently are used by the majority of astrophysical modelers worldwide.@xcite in its most general definition , the term `` direct reaction '' includes all processes directly connecting the initial and final states of a nuclear reaction without formation of an intermediate compound system .
this includes elastic scattering as described in the optical model , and inelastic scattering which predominantly excites collective states.@xcite the latter includes coulomb excitation which has been found to be important in heavy ion collisions due to the high coulomb barriers involved.@xcite in astrophysically relevant reactions , especially with @xmath0-particles , energies may also be close to or below the coulomb barrier and coulomb excitation may also become important , depending on the structure of the target nucleus.@xcite here , we focus on direct reactions when some ( if it is a stripping reaction ) or all ( if it is a capture or charge - exchange reaction ) nucleons of the projectile are incorporated in the target nucleus . in a pick - up reaction ,
one or more nucleons from the target nucleus are added to the projectile to form the ejectile , again in a direct manner .
pick - up and stripping reactions are subsumed under the term `` transfer reactions '' .
in contrast to the hfm model , direct reactions excite only few degrees of freedom because most of the nucleons included in the system of target nucleus plus projectile remain spectators .
a nucleon of the projectile reaches its final state without sharing any energy with any of the other nucleons present and the excess energy is emitted as a discrete photon carrying the energy difference between initial and final state .
direct reactions can be identified experimentally because of their angular dependence of the differential cross sections , being peaked in forward direction .
direct processes are also faster by at least 5 orders of magnitude than compound reactions , with reaction timescales of the order of @xmath514 s. this is comparable to the time the projectile requires to cover a distance of the size of a nucleus
. therefore , direct reactions are important at high projectile energies when compound formation is disfavored .
although the notion of direct processes was inspired originally by angular distributions of low - energy reactions , it was assumed for a long time that higher energies are the domain where they are dominating.@xcite in resonant reactions at lower energy , it is sometimes necessary to include a non - resonant background ( which may show interference with resonances ) but experimentally it is often difficult to distinguish between a direct component and contributions from tails of resonances .
however , in systems with low nld , and thus widely spaced resonances , direct reactions become important even at astrophysically low interaction energies because compound formation is suppressed.@xcite this even applies to intermediate and heavy nuclei far off stability , e.g. , for neutron capture in nuclei with low neutron separation energy.@xcite the direct capture cross section can become considerably larger than the compound cross section .
figure [ fig : dchf ] compares ( n,@xmath82 ) cross sections at 30 kev from direct capture and from hfm for a number of isotopes . with decreasing neutron separation energy ,
the direct component plays an increasingly important role because the compound nucleus is formed at lower excitation energy and thus also at lower nld .
elements with inherently low nld , such as sn , show a larger direct contribution for all isotopes . also for nuclei at shell closures
the nld is low and the importance of direct reactions enhanced relative to the hfm.@xcite similar considerations may apply to proton - induced reactions on proton - rich nuclei .
direct processes are not only important to be included in reaction rate predictions .
all possible processes ( elastic and other direct ones , compound - elastic and compound nucleus reactions ) have to be taken into account in the analysis of scattering and reaction data when extracting optical potentials .
the elastic scattering cross section @xmath515 ( which is a direct process ) , the one for reactions @xmath516 , and the total @xmath517 are related by @xmath518 where we distinguished between elastic scattering at the optical potential @xmath519 and compound - elastic scattering @xmath520 .
the cross section @xmath521 includes all inelastic processes , i.e. reactions .
often , one reaction mechanism is dominating by far and then @xmath521 can be identified with the cross section for that mechanism , e.g. the compound cross section @xmath520 for resonant processes ( including the hfm ) as discussed in secs .
[ sec : reso ] and [ sec : statmod ] .
but this is not always the case as depending on the nucleus and the projectile energy
there may be additional mechanisms contributing to @xmath521 in some cases .
therefore , using an optical potential derived from scattering implies that the absorption term is due to some reaction(s ) but does not define the reaction mechanism(s ) .
using such a potential in a pure hfm implicitly assumes that the missing flux from the elastic channel is due to the compound mechanism only .
this may not be appropriate when direct processes are non - negligible ( this comprises direct reactions as discussed below but also coulomb excitation at low energy ) and will require a modification of the optical potential depending on which mechanism is to be studied .
this is also implicitly contained in the idea of the modified hfm briefly discussed in sec .
[ sec : statmodmod ] .
earlier calculations of direct neutron capture have made use of a simple hard sphere capture model ( see sec .
[ sec : dc ] ) in a combination of direct and hfm capture , not just to simplify calculations but also because it also allows the assumption that the contribution to direct capture potential absorption due to the tail of distant resonances is already included in the statistical model averaging.@xcite astrophysical rates can be calculated from cross sections by applying ( [ eq : distrirate ] ) , ( [ eq : rate ] ) , and ( [ eq : effrate ] ) , regardless of the reaction mechanism . since each discrete transition appearing in a direct process obeys the reciprocity relation ( [ eq : reci_single ] ) , a similar effective cross section ( [ eq : effcs ] ) can be derived as for the compound case , resulting in the same weighting factors ( [ eq : effweights ] ) of transitions from the excited states . also the same reciprocity relations ( [ eq : revrate ] ) and ( [ eq : revphoto ] ) apply , provided that thermal population of the states in all participating nuclei is valid . in the following only a few methods
are outlined which have been used to calculate direct reactions for astrophysics . only the basic equations for the reaction from one initial state to one final state
are given but the actual rate equations can straighforwardly be obtained by using the methods described in secs . [
sec : stellar ] and [ sec : stellarexp ] . it should be noted that here not only light targets are implied but also intermediate and heavy ones ( see fig . [
fig : dchf ] ) , for which microscopic models are not feasible ( see sec .
[ sec : mechintro ] ) .
furthermore , low projectile energies are implied as these are required for calculating astrophysical reaction rates .
nevertheless , direct reactions at higher energies can be used to extract certain properties , such as spin assignments and spectroscopic factors ( see , e.g. , ref . ) , of stable and unstable nuclei which are required for the calculation of the cross sections and the rates .
direct transfer reactions can be treated by solving the time - independent schrdinger equation with optical potentials in the entrance and exit channels .
a simple implementation of this is the distorted wave born approximation ( dwba ) .
the differential cross section for the one - nucleon or cluster transfer @xmath522 with @xmath523 , @xmath524 for @xmath525 and @xmath526 or 3 is given in zero - range dwba by@xcite @xmath527 with the zero - range normalization constant @xmath528 .
the reduced cross section without spin - orbit coupling is given by @xmath529 with the reduced transition amplitude @xmath530^ * \chi_{aa}^{(+)}\left(k_{aa},r \right)\ , dr \quad.\ ] ] as before , the quantities @xmath62 , @xmath531 and @xmath532 , @xmath533 are the reduced masses and wave numbers in the entrance and exit channel , respectively .
the orbital angular momentum quantum number @xmath377 , the spin quantum number @xmath100 , and the total angular momentum quantum number @xmath14 refer to the nucleon or cluster @xmath534 bound in the residual nucleus @xmath194 .
the spectroscopic factor and the isospin clebsch - gordan coefficient for the partition @xmath535 are given by @xmath130 and @xmath536 , respectively .
the optical wavefunctions in the entrance and exit channels are given by @xmath537 and the time - reversed solution @xmath538 .
the bound state wave function is denoted by @xmath539 and the @xmath540 are the usual spherical harmonics .
expressions similar to the above are obtained when the finite range of the interaction potential is taken into account.@xcite important for the successful application is to keep the number of open parameters as small as possible .
for this reason , folding potentials ( see the paragraphs on optical potentials in sec . [
sec : relevance ] ) were used in many astrophysical applications of the model , with @xmath21 either determined from scattering data or from global dependences.@xcite this leaves the spectroscopic factor @xmath541 which is usually determined by simply comparing the calculated magnitude of the differential cross section to measurements .
this is mainly done with ( d , p ) or ( d , n ) reactions at energies above the astrophysically relevant ones.@xcite the partial width @xmath542 appearing in ( [ eq : breit ] ) can be related to spectroscopic factors for a particle @xmath393 in a state @xmath102 by@xcite @xmath543 the single particle width @xmath544 can be derived from scattering phase shifts and this offers a different experimental access to spectroscopic factors ( see sec . [
sec : avdc ] for further methods to determine spectroscopic factors).@xcite further required input includes , of course , masses or separation energies and nuclear spectroscopic information which have a similar importance as for the hfm , discussed in sec .
[ sec : relevance ] . in the absence of experimental data
, the spectroscopic factor can be calculated microscopically from the overlap between initial and final state wave functions , e.g. , in the shell model.@xcite however , there is some ambiguity because this overlap is not well defined in different microscopic approaches.@xcite it has been shown that the spectroscopic factors for depositing or picking up a single nucleon are related to the occupation factors of the participating quasi - particle states ( see also sec .
[ sec : avdc]).@xcite as in the case of the hfm , many codes have been used for dwba calculations over the years , especially for the analysis of data at intermediate and high energies . the code tetra has been written especially for application at astrophysically relevant low energies.@xcite it has been applied successfully to astrophysically relevant reactions with light and intermediate target nuclei ( see , e.g. , refs . , and references therein ) .
is it necessary to go beyond the dwba ?
there are three fundamental assumptions contained in the dwba treatment:@xcite 1 .
the reaction proceeds directly from initial to final state and all particles except the transferred one(s ) remain unaffected spectators .
the wave function for the relative motion between the reactands is assumed to be correctly described by the optical potential .
the reaction is assumed to be sufficiently weak to be treated in lowest order . to relax the first two assumptions , the coupled - channel born approximation ( ccba ) was introduced.@xcite under rare circumstances
the transfer amplitudes may be large and the third assumption has to be relaxed .
this leads to a full coupled - channels treatment for the reaction.@xcite one has to be aware of the fact that the relevant energies remain low for astrophysical reaction rates , also because transitions from excited states contribute considerably .
this is contrary to what one is used to in the investigation of reactions proceeding at several tens of mev . due to the low energies involved , the reaction channel is weak ( compared to , e.g. , elastic scattering ) and
the third assumption is valid .
the usual concern with the second assumption is that the optical potential also has to describe well the wave function even deep in the nuclear interior .
the deep region , however , is crucial for reactions at higher energy whereas at the low astrophysical energies most contributions to the overlap integrals stem from regions close to the surface of the nucleus or even from outside of the nuclear radius .
as long as these regions are described well by the optical potentials , the dwba should work .
the first assumption implies that either no indirect processes exist or that they can be treated separately ( incoherently ) as was suggested by the above , separate discussion of compound reactions and other mechanisms . in the hfm
it is assumed that interference terms cancel and thus also interference with direct reactions should cancel on average .
interference with isolated resonances can be treated explicitly by adding an interference term , e.g. , between the s - factor of the direct reaction @xmath545 and the one of a breit - wigner resonance @xmath546 ( see sec .
[ sec : reso ] ) @xmath547 where @xmath548 is the energy - dependent , relative phase shift.@xcite a potential model can also be used to calculate direct capture ( dc ) .
although microscopic models are an alternative for light systems ( see sec .
[ sec : mechintro ] ) , a dc potential model has the advantage that it can be applied also to heavier nuclei .
the dc cross section for a particular transition is determined by the overlap of the scattering wave function in the entrance channel , the bound - state wave function in the exit channel , and the electromagnetic multipole transition operator .
the dc cross section is then given by@xcite @xmath549 the polarisation @xmath37 of the electromagnetic radiation can be @xmath550 .
the wave number in the entrance channel and for the emitted radiation is given by @xmath532 and @xmath551 , respectively .
the multipole expansion of the transition matrices @xmath552 including electric dipole ( e1 ) and quadrupole ( e2 ) transitions as well as magnetic dipole ( m1 ) transitions is given by @xmath553 the rotation matrices depend on the angle between @xmath554 and @xmath555 which is denoted by @xmath556 , where @xmath557 . defining @xmath558
we can write for the transition matrices for the electric dipole ( e@xmath559 = e1 ) or quadrupole ( e@xmath559 = e2 ) transition @xmath560 in the above expressions the quantum numbers for the channel spin in the entrance channel and for the transferred angular momentum are denoted by @xmath561 and @xmath562 , respectively .
the quantities @xmath563 , @xmath564 and @xmath565 ( @xmath566 , @xmath567 and @xmath568 ) are the spins ( magnetic quantum numbers ) of the target nucleus @xmath31 , residual nucleus @xmath194 and projectile @xmath32 , respectively . for magnetic dipole transitions ( m@xmath559 = m1 ) we obtain @xmath569 \nonumber\\ & - ( l_a\ , 0\ , s_a\ , m_a \mid j_a\ , m_a ) ( j_a\ , m_a\ , i_a\ , m_b\ ! -\ ! m_a \mid i_b\ , m_b ) \nonumber\\ & \times ( i_a\ , m_b\ ! -\ !
m_a\ , 1\ , \delta\ , \mid i_a\ , m_a ) \nonumber \\ & \times \mu_a \ , \delta_{j_a j_b } \sqrt{\ , ( i_a+1)/i_a } \biggr\ } \biggl\ { \frac{\hbar c}{2 m_{\rm p } c^2 } \biggr\}\ , \delta_{l_a l_b } \ : \hat{l}_a\ : i_{l_b j_b i_b ; l_a j_a}^{\rm m1 } \quad,\label{eq : kim4}\end{aligned}\ ] ] where @xmath570 is the racah coefficient , the @xmath571 are the magnetic moments and @xmath572 is the mass of the proton .
the overlap integrals in ( [ eq : kim3 ] ) and ( [ eq : kim4 ] ) are given as @xmath573 for the electric dipole ( e@xmath559 = e1 ) or quadrupole ( e@xmath559 = e2 ) transition , and by @xmath574 for the magnetic dipole transition ( m@xmath559 = m1 ) .
the radial part of the bound state wave function in the exit channel and the scattering wave function in the entrance channel is given by @xmath575 and @xmath576 , respectively .
the radial parts of the electromagnetic multipole operators are@xcite @xmath577 \quad , \nonumber\\ { \cal o}^{\rm e1}(r)&= \frac{3}{\hat{\rho}^3}\left [ ( \hat{\rho}^2 - 2 ) \sin \hat{\rho } + 2 \hat{\rho } \cos \hat{\rho } \right ] r \quad , \nonumber\\ { \cal o}^{\rm e2}(r)&= \frac{15}{\hat{\rho}^5}\left [ ( 5 \hat{\rho}^2 - 12 ) \sin \hat{\rho } + ( 12 - \hat{\rho}^2 ) \hat{\rho } \cos \hat{\rho } \right ] r^2 \quad .\label{11}\end{aligned}\ ] ] in the long wave - length approximation applicable as long as @xmath578 these quantities reduce to @xmath579 usually , only the dominant e1 transitions have to be taken into account .
possible exceptions are captures far from stability with very low reaction @xmath94-values because for these cases no final states may be energetically accessible through e1 transitions .
however , because the astrophysical reaction rate involves summing over transitions originating from excited states , a larger spin range may be available and e1 ( from excited target states ) may again dominate . for e1 transitions ,
the above expressions reduce to@xcite @xmath580 the coefficients @xmath581 are calculated in ls coupling to @xmath582 in the above expressions , the energy of the emitted photon is @xmath469 .
the orbital and total angular momentum quantum numbers of the nuclei in the entrance and exit channels are @xmath583 , @xmath584 , @xmath585 and @xmath586 , respectively .
the spin quantum number , the orbital and total angular momentum quantum numbers are characterized by @xmath65 , @xmath47 and @xmath587 , respectively , with indices @xmath32 , @xmath31 and @xmath194 corresponding to the projectile , target and residual nucleus , respectively .
the notation @xmath588 stands for the @xmath589 symbol .
the radial wave functions in the entrance and exit channels are given by @xmath590 and @xmath591 , respectively .
the spectroscopic factor and the isospin clebsch - gordan coefficient for the partition @xmath592 are given by @xmath130 and @xmath593 , respectively .
the dc potential model has been successfully applied to many reactions with light target nuclei at astrophysical energies , e.g. , for @xmath291be(p,@xmath82)@xmath594b and @xmath291li(n,@xmath82)@xmath594li.@xcite for further calculations see , e.g. , refs . and references therein .
many calculations were performed with the dc potential model code tedca , which is tailored to treat low - energy reactions of astrophysical interest.@xcite again , folding potentials ( see sec .
[ sec : relevance ] ) were the key to reduce the number of parameters .
spectroscopic factors were taken from experiment or shell model calculations .
it was also used to extend calculations to intermediate and heavy target nuclei.@xcite other astrophysical calculations have not made use of the full dc equations but used simplifying assumptions .
resonant and dc rates on the proton - rich side based on hard sphere scattering wavefunctions in the entrance channel with resonance properties , final states , and spectroscopic factors taken from shell model calculations were provided for target nuclei with mass number @xmath595.@xcite in these calculations , however , the resonant part exceeds the direct part of the cross section by several orders of magnitude .
astrophysical neutron capture on stable and neutron - rich nuclei was calculated in the hard - sphere model for e1 capture by refs . .
in this model , the e1 neutron capture cross section can be written as@xcite @xmath596 where @xmath597 is the hard sphere radius and the multiplicity @xmath389 is the number of incident channel spins which can lead to the same final state with spin @xmath598 .
it is @xmath599 for @xmath600 , or for @xmath601 and @xmath602 .
the value @xmath603 applies for @xmath601 and @xmath604 .
the quantity @xmath605 is the orbital angular momentum of the final bound state @xmath128 and @xmath606 is the spectroscopic factor of this state .
the dimensionless parameter @xmath607 is given by @xmath608 the advantage of this approach is that no explicit wave functions for scattering and bound states are required .
on the other hand , the correct overlap of the wave functions may yield more accurate cross sections , especially for low projectile energies when considerable contributions to ( [ eq : overlape ] ) are coming from far outside the nuclear radius .
ref . showed the importance of direct neutron capture in the r - process based on the hard - sphere model .
for the validity of the low - order potential model approach for dc while neglecting higher - order processes , similar arguments can be made as were presented for the dwba towards the end of sec .
[ sec : dwba ] . for astrophysics ,
a large number of rates for highly unstable nuclei have to predicted .
similar to the uncertainties of the hfm discussed in sec .
[ sec : relevance ] the dwba and dc predictions are sensitive to certain inputs , such as masses ( @xmath94-values ) , excited and bound state properties , spectroscopic factors , optical and bound state potentials .
when folding potentials are used , a ( weak ) dependence on nuclear matter density distributions appears additionally .
the discussion of these quantities in sec .
[ sec : relevance ] also applies here .
however , some of the quantities appearing in the hfm are sums over individual transitions and averaged quantities whereas individual transitions are determining the direct reaction cross sections .
thus , direct reactions are more sensitive to nuclear properties impacting those individual transitions , including spins , parity , energy of bound and excited states , @xmath94-values , and spectroscopic factors .
due to the angular momentum barrier and the low projectile energies , low partial waves contribute most to the e1 cross section , i.e. s - waves when initial and final states have differing parity , p - waves when they have the same parity . assuming equal spins and parities
, low - lying states are contributing more than higher ones . the strong sensitivity to individual transitions , however
, may be reduced in the astrophysical rate , when using the weighted sums ( [ eq : rate ] ) , ( [ eq : effrate ] ) , ( [ eq : effweights ] ) over transitions from excited target states to available final states . for neutron - rich sn and pb isotopes , predictions for 30 kev neutron capture employing input from different microscopic or semi - microscopic approaches
were compared in detail.@xcite the dc calculation was tested in comparison to experimental data for @xmath487pb(n,@xmath82)@xmath609pb . in this case , a discrepancy between the data obtained in an activation measurement and the data from a high - resolution resonance counting experiment was resolved by showing that the difference is due to the dc contribution to the cross section which is only included in the cross section from the activation measurement . for very neutron - rich isotopes of sn and pb it was found that the resulting cross sections differ by orders of magnitude with the different inputs ( see figs . [
fig : dccompsn ] , [ fig : dccomppb ] ) .
this is mainly due to the sensitivity of the cross section to the predicted location of the low - spin bound states with respect to the neutron separation energy.@xcite the dc calculation can be tested for @xmath610sn(n,@xmath82 ) because there is experimental information on the bound states in @xmath611sn.@xcite this nucleus is predicted to be close to or directly in the r - process path.@xcite an independent calculation confirmed the original work.@xcite fortunately for astrophysics , very similar cross sections ( within a factor of 3 ) are computed for this reaction with input from the different microscopic approaches .
unfortunately for nuclear physics , the reaction is not a good case to select a preferred microscopic model for the same reason .
spectroscopic data for neutron - richer isotopes would be necessary .
furthermore , spectroscopic factors were set to unity in these calculations .
this is a good assumption for the states in @xmath611sn ( and was recently confirmed by ref . ) but is not valid in mid - shell .
however , the uncertainty introduced by the predicted bound state energies exceeds by far the one introduced by the spectroscopic factors when only particle states are considered and hole states are neglected.@xcite the latter have very small spectroscopic factors and are negligible because they involve a reordering process in the final nucleus ( see also ref . for another example of a reordering process in @xmath612s and its dependence on deformation ) .
regular dc cross sections are obtained by summing over all allowed transitions to energetically accessible final states , @xmath613 where each summand contains the appropriate spectroscopic factor . following the derivations in sec .
[ sec : stellar ] , the astrophysical rate will contain an effective cross section and a partition function .
similarly to the hfm , it was suggested ( refs . )
that the sum over final states may be ( partially ) replaced by an integration over the nld in the final nucleus @xmath614 , @xmath615 + \nonumber \\ & + \int_{e^\mathrm{x}_{\nu^f}}^{s_\mathrm{n } } \sum_{j^f\pi^f } \rho_f(e^f , j^f,\pi^f ) \sigma_\mathrm{dc}^{\mu\rightarrow f}(e , e^f , j^f,\pi^f,\bar{\mathcal{s}}_{e^fj^f\pi^f } ) \ , de^f \quad .
\label{eq : avdcmu}\end{aligned}\ ] ] similarly , the summation over initial states @xmath102 in ( [ eq : effcs ] ) can be ( partially or fully ) replaced by an integration over the nld in the target nucleus @xmath616 , @xmath617 + \nonumber \\ & + \int_{e^\mathrm{x}_{\mu^f}}^{e_\mathrm{proj}(t ) } \sum_{j^i\pi^i } \rho_i(e^i , j^i,\pi^i ) \bar{\sigma}_\mathrm{dc}^{i\rightarrow ( \nu , f)}(e , e^i , j^i,\pi^i,\bar{\mathcal{s}}_{e^ij^i\pi^ie^fj^f\pi^f } ) \ , de^i \quad .
\label{eq : avdceff}\end{aligned}\ ] ] through the nld , the cross sections @xmath618 and @xmath619 not only include transitions to discrete final states but also `` average '' transitions to states described by the nld .
therefore i call this _ averaged direct capture _ ( adc ) . as in the hfm , the adc cross section may only include the ground state transitions and nlds above the ground state when all excited state properties are unknown .
the advantage of this is that the sensitivity to the location of discrete states relative to the projectile separation energy ( as seen in ref . )
is washed out and the change in cross section from one isotope to the next is smoother .
of course , this may not properly describe nuclear structure details in the cross sections when only few states are available but it can give a more appropriate estimate of the magnitude of the rate for astrophysics ( which involves an averaging over a relevant energy range and includes more transitions than the laboratory rate ) than relying on just a single ( semi-)microscopic approach .
obviously , the adc cross section will be sensitive to the nld .
contrary to the hfm , where the nld mostly impacts the @xmath82-widths and has its strongest effect at several mev excitation energy , captures to low - lying states ( assuming spins and parities are favorable ) are dominating and thus the nld at low excitation energy will be relevant .
this is why it is important to use a proper spin- _ and _ parity - dependent nld description .
it was found recently , however , that thermal excitation of target nuclei reduces the sensitivity to the parity dependence in the nld for astrophysical rates in the hfm.@xcite we can expect a similar effect for direct capture , although the number of possible transitions is more limited.@xcite appropriate spectroscopic factors are a further important ingredient in adc calculations . in the integral of ( [ eq : avdcmu ] ) these
are _ averaged _ spectroscopic factors @xmath620 , describing the average overlap between the initial state @xmath102 and the final bound states with given spin and parity at an excitation energy @xmath621 .
the doubly averaged spectroscopic factors @xmath622 appearing in the integrand of ( [ eq : avdceff ] ) are even more complicated , as they involve the average overlap between all initial states with given spin and parity at excitation energy @xmath623 and all final states . currently , the only spectroscopic factors to be found in literature ( from experiment or theory ) are for transitions connecting the ground state @xmath112 of the target nucleus with the final states .
spectroscopic factors for transitions from excited states have yet to be calculated .
they are needed not just for the averaged dc model employing nlds but also for the regular dc model when applied to compute astrophysical reaction rates ( see also the discussion of the stellar enhancement and the effective weights in sec . [
sec : stellarexp ] ) .
spectroscopic factors for one - nucleon capture ( or transfer ) on a target in the ground state can easily be computed from the occupation numbers @xmath624 as calculated , e.g. , from bcs or lipkin - nogami pairing.@xcite then the spectroscopic factor for putting a nucleon in state @xmath14 with spin @xmath625 is just @xmath626 for a target nucleus with an even number of nucleons of the same type as the projectile .
the occupation probability is the one of the target nucleus . in a chain of linked reactions ,
the total processing efficiency is given by the slowest reaction(s ) .
therefore , when considering sequences of capture reactions , e.g. , in the s- , r- , rp - processes , the rates on such target nuclei with even nucleon number will be the slowest ( and their reverse photodisintegrations the fastest ) and thus they will have the largest astrophysical impact . when the target nucleus has an odd number of nucleons of the projectile type , the spectroscopic factor is @xmath627 the occupation probability @xmath628 is always taken in the nucleus with even number of nucleons . in ( [ eq : occodd ] )
this is the final nucleus of the reaction .
the expressions for the extraction of a nucleon from a given state follow from the fact that the time - reversed reactions have to show the same spectroscopic factor .
the occupation probabilities can be calculated from microscopic theory , e.g. , using bcs or lipkin - nogami pairing on a single - particle basis.@xcite in the absence of calculated spectroscopic factors several different approaches have been used in the past . often , spectroscopic factors were set constant to 1.0 ( e.g. , refs . ) or to 0.1 ( e.g. , refs . ) .
these values can already be seen as averaged spectroscopic factors @xmath620 for low - lying particle - states .
useful for the application to the adc approach is the construction of an excitation energy - dependent , averaged spectroscopic factor to be employed along with the nld . as an example for this , in studying neutron capture on the astrophysically important nucleus @xmath510ti the dc component was estimated implementing a distributed spectroscopic strength.@xcite in this case , transitions to @xmath629 and @xmath630 states in @xmath612ti are dominating .
a distribution @xmath631 of the @xmath629 and @xmath630 strengths was assumed , reaching from the ( experimentally known ) location of the lowest @xmath629 and @xmath630 state , respectively , to the neutron separation energy . with this
`` smearing out '' of the states and of the strength and due to the @xmath632 dependence of the e1 transition probability , then the calculation is reduced to computing a transition to an effective bound state with full spectroscopic strength at an energy of @xmath633{\widetilde{e^3_\gamma}}\quad,\ ] ] where the average transition energy is given by @xmath634 such an approach accounts for the uncertainties in spectroscopic strength and location of excited states and can be viewed as a zeroth approximation to the adc .
comparison between experimental and averaged ( eq .
[ eq : holzerspectfact ] ) spectroscopic factors as function of excitation energy for @xmath635xe+n.@xcite ] another suggestion for the functional form of the energy - dependence of the average spectroscopic factors was made in refs . .
the spectroscopic factors describe the overlap between the antisymmetrized wave functions of target+nucleon and the final state .
the number of final state configurations increases with increasing excitation energy @xmath636 and the overlap of initial and final state wavefunctions decreases .
thus , also the spectroscopic factor decreases . in a simple approximation
, the energy dependence of the spectroscopic factor for single nucleon transfer can be parameterized by a fermi function with @xmath637 and the parameters @xmath638 .
this is motivated by the excitation - energy dependence of the occupation probabilities .
figure [ fig : avdcspect ] shows how well averaged spectroscopic factors with the functional dependence ( [ eq : holzerspectfact ] ) compare to experimental ones for @xmath635xe+n.@xcite there is also a connection between single - particle spectroscopic factors @xmath639 and the partial resonance widths @xmath640 appearing in the bwf ( see sec .
[ sec : reso ] ) @xmath641 where @xmath130 is the isospin clebsch - gordan coefficient , @xmath362 the reduced mass of the system nucleus+particle , and @xmath642 is the dimensionless single - particle reduced width.@xcite the penetrability @xmath363 for the relative angular momentum @xmath377 can be expressed in terms of the regular and irregular coloumb wavefunctions @xmath643 and @xmath644 @xmath645 in literature , a value of @xmath646 is often assumed for an average single - particle reduced width . comparing ( [ eq : spwidth ] ) and ( [ eq : reduwidth ] ) , it can be seen that an average @xmath642 can be calculated from solutions of the radial schrdinger equation ( giving @xmath328 ) with an optical potential.@xcite since @xmath647 , there is a direct connection to the average transmission coefficients @xmath648 appearing in the hfm ( see sec .
[ sec : statmod ] ) which can be used to consistently derive the average combined value @xmath649 .
the averaged spectroscopic factors for transitions from excited target states in ( [ eq : avdceff ] ) can be estimated in this way .
sums of the contributions to the total direct capture cross sections with the averaged dc model and the standard dc for @xmath635xe(n,@xmath82 ) as function of excitation energy of the final nucleus.@xcite ] the integrals appearing in ( [ eq : avdcmu ] ) and ( [ eq : avdceff ] ) always contain products @xmath650 of nld and averaged spectroscopic factors . with the relation between strength function @xmath651 and reduced width ( see , e.g. , ref . )
we can derive @xmath652 and thus ( partially ) eliminate the nld within the integrals as it is implicitly contained in the @xmath653 or @xmath654 .
the strength function @xmath653 is only defined for resonance states above the projectile separation energy . for estimating transitions to bound states it was suggested to construct an `` internal strength function '' @xmath655 , extending the regular strength function below the separation energy.@xcite this
is motivated by the relation ( [ eq : avspect ] ) and the fact that the spectroscopic factors describe the structure of bound states as well .
the adc was investigated in ref . for the reactions @xmath635xe(n,@xmath82)@xmath656xe and @xmath463sm(n,@xmath82)@xmath657sm , using ( [ eq : lanelynn ] ) with an internal strength function . the adc cross section ( for the target in the ground state )
was compared to a calculation summing over transitions to bound states in the standard potential model ( as described in sec .
[ sec : dc ] ) and with experimental data .
the energy - dependence of @xmath655 was chosen very similar to the one found for the average spectroscopic factors @xmath658 , with independent parameters @xmath659 , @xmath660 , @xmath661 .
the parameters were determined by requiring @xmath662 at the neutron separation energy .
simultaneously , it was required that the adc cross section integrated up to the excitation energy of the last included state yields the same cross section as obtained with the standard potential model .
although this is not suited for a prediction , it can be used to assess the validity of the assumptions . including only s - wave neutrons ,
the experimental cross sections were reproduced within 25% for @xmath635xe(n,@xmath82 ) and 2% for @xmath463sm(n,@xmath82 ) .
figure [ fig : holzeravdc ] displays a comparison between the results from the averaged direct neutron capture and standard direct neutron capture for @xmath635xe . for predictions across the nuclear chart
, the spectroscopic factors and/or internal strength functions can be obtained from optical model single - particle states and/or occupation numbers of quasi - particle states as shown in ( [ eq : occeven ] ) and ( [ eq : occodd ] ) . for simpler application and to increase computational speeds , these can be parameterized according to ( [ eq : holzerspectfact ] ) .
spectroscopic factors for transitions from excited target states or the doubly averaged factors required in ( [ eq : avdceff ] ) remain an open problem . the code smaragd ( see sec . [
sec : codes ] ) will also include a global dc treatment using an adc model and energy - dependent spectroscopic factors , making use of a combination of the above approaches intended to also yield consistency with the hfm.@xcite such an adc approach aims at providing robust predictions despite of considerable differences between microscopic predictions.@xcite preliminary results for this adc are shown in fig .
[ fig : avdc ] , along with results from the hfm ( sec .
[ sec : statmod ] ) and a modified hfm ( sec .
[ sec : statmodmod ] ) . the final rate ( or cross section ) is the sum of the modified hfm value and the adc one .
interestingly , for the isotopes shown here ( except for @xmath663 ) this sum is approximated by the unmodified hfm result within a factor of 10 .
this is in accordance with ref .
( see figure 3 therein ) , where the dc contribution also almost replaces the standard hfm values .
this shows that it seems justified to use unmodified hfm rates as crude estimate of the total rates for exotic nuclei .
there is no fast highway to improvements in predictions of reaction rates , not even a wide road .
historically , the fields of standard nuclear physics and nuclear data for applications have taken another direction . therefore instead of following a beaten track rather a new , narrow path has to be driven step by step through a jungle of complications and details .
this is only possible in a concerted effort of theory and experiment , the latter involving both large - scale rare - isotope production sites and smaller facilities .
there is not one `` most important '' nucleus or `` most important '' reaction in nuclear astrophysics .
which nuclei , nuclear properties , and reactions are at the center of attention depends on the astrophysical process studied .
therefore , systematical studies are needed as well as information on specific nuclei and reactions . on the theory
side large - scale studies of general trends and dependences are required as well as detailed predictions of individual nuclear properties . as shown above
, the calculation of astrophysical rates , even when experimental information is present , involves a number of specialities not encountered in usual nuclear physics investigations .
thus , nuclear astrophysics is heavily relying on advances in nuclear theory and experiment but also requires its own special developments in theory and experiment which justify its existence as a separate field .
nuclear physicists working at the boundary to astrophysics have to be aware of these special requirements and it is one of the aims of this work to having outlined a number of them .
a further noteable fact is that it is necessary to not only point out special effects or important possible improvements but to actually apply them across the nuclear chart and produce large - scale sets of reaction rates which can be readily used by astrophysicists .
this implies that they are made accessible in a form suitable to be implemented in astrophysical models .
it should not be forgotten that the other essential aspect of nuclear astrophysics is the astrophysical modeling using reaction networks . only in conjunction with this part of the field progress
can be made .
the models set the stage and define the ranges of conditions within which nuclear processes occur .
nevertheless , it can be treacherous to rely too strongly on a certain model .
reliable nuclear models and astrophysical reaction rates should cover a large range of possibilities and provide a sound base for pinpointing the sites of nucleosynthesis processes or even for discovering new types of nucleosynthesis in different astrophysical models .
moreover , the modifications of the cross sections and rates in a stellar plasma are an interesting topic in itself and warrant an independent study even without connection to a specific astrophysical site .
the prediction of astrophysical reaction rates takes nuclear physics in a new direction and tests nuclear theory at the limits . due to the finite number of nuclei
, however , this is a finite task .
this is also why parameterizations or phenomenological models may still have their justification , if designed in an appropriate manner .
we do not have to extrapolate to infinity but within rather limited ranges of nucleon numbers .
it is reassuring that the overall abundance distributions obtained by combining several postulated nucleosynthesis processes are already closely resembling what we find in nature , even when explosive events and highly unstable nuclei are involved .
this tells us that the most important properties seem to be described acceptably well .
there are remarkable exceptions , however , both in explaining abundance distributions ( e.g. , of the p - nuclei , the light s - process elements , isotopic anomalies in meteorites , the heaviest nuclei at the endpoint of the r - process , and many more ) and in assigning astrophysical sites ( e.g. , to the r - process , probably also to parts of the p - process ) . a detailed understanding and reliable reaction rates are also essential for using nuclear cosmochronometry to determine astronomical timescales .
to go beyond previous estimates and reach a new level of detail , however , requires a large , dedicated effort of both experiment and theory .
this work was supported by the swiss national science foundation , grant 200020 - 105328 , and by the european commission within the ensar / thexo project .
k. wildermuth and y. c. tang , in _ a unified theory of the nucleus , clustering phenomena in nuclei _ , vol . 1 ,
k. wildermuth and p. kramer , a unified theory of the nucleus , clustering phenomena in nuclei , vol .
1 , ed . k. wildermuth and p. kramer ( vieweg , braunschweig , germany , 1977 ) .
a. j. koning , s. hilaire and m. c. duijvestijn , _ proceedings of the international conference on nuclear data for science and technology , nice , 2227 april 2007 _ , eds .
o. bersillon , f. gunsing , e. bauge , r. jacqmin , s. leray ( edp sciences , les ulis , 2008 ) pp .
211214 ; online available at http://www.talys.eu f .- k .
thielemann , m. arnould and j. w. truran , in _ advances in nuclear astrophysics _ ,
e. vangioni - flam , j. audouze , m. cass , j .-
chize , j. tran thanh van ( editions frontires , gif - sur - yvette , 1986 ) , p. 525 .
t. rauscher , w. bhmer , k .-
kratz , w. balogh and h. oberhummer , _ proc .
nuclei and atomic masses _ , eds .
m. de st .
simon , o. sorlin ( editions frontires , gif - sur - yvette , 1995 ) , p. 683
. w. balogh , r. bieber , h. oberhummer , t. rauscher , k .-
kratz , p. mohr and g. staudt , _ proc .
workshop on heavy element nucleosynthesis _ ,
e. somorjai , zs .
flp ( atomki , debrecen , 1994 ) , p. 67 ; arxiv : nucl - th/9404010 .
h. herndl , r. hofinger and h. oberhummer , _ proc .
origin of matter and evolution of galaxies _ , eds .
s. kubono , t. kajino , k. i. nomoto , i. tanihata ( world scientific , singapore , 1998 ) , p. 233 ; arxiv : nucl - th/9803012 .
j. dobaczewski , h. flocard and j. treiner , _ nucl .
phys . _ * a422 * ( 1984 ) 103 .
j. dobaczewski , i. hamamoto , w. nazarewicz and j.a .
sheikh , _ phys .
_ * 72 * ( 1994 ) 981 .
m. m. sharma , m. a. nagarajan and p. ring , _ phys .
b _ * 312 * ( 1993 ) 377 .
m. m. sharma , g. a. lalazissis and p. ring , _ phys .
b _ * 317 * ( 1993 ) 9 .
p. mller and j. r. nix , _ nucl .
phys . _ * a361 * ( 1981 ) 117 .
p. mller , j. r. nix , w. d. myers and w. j. swiatecki , _ at .
data nucl .
data tables _ * 59 * ( 1995 ) 185 .
p. mller , j. r. nix and k .- l .
kratz , _ at .
data nucl .
data tables _ * 66 * ( 1997 ) 131 . | this review focuses on nuclear reactions in astrophysics and , more specifically , on reactions with light ions ( nucleons and @xmath0 particles ) proceeding via the strong interaction .
it is intended to present the basic definitions essential for studies in nuclear astrophysics , to point out the differences between nuclear reactions taking place in stars and in a terrestrial laboratory , and to illustrate some of the challenges to be faced in theoretical and experimental studies of those reactions .
the discussion revolves around the relevant quantities for astrophysics , which are the astrophysical reaction rates .
the sensitivity of the reaction rates to the uncertainties in the prediction of various nuclear properties is explored and some guidelines for experimentalists are also provided .
_ caminante , son tus huellas el camino , y nada ms ; + caminante , no hay camino , se hace camino al andar .
+ al andar se hace camino y al volver la vista atrs + se ve la senda que nunca se ha de volver de pisar .
+ caminante , no hay camino sino estelas en la mar . + antonio machado _ |
the field of heavy ion collisions is a very active one , straddling high energy and nuclear physics . at the upper energy limit of this flourishing area of research
, the goal is to eventually produce and study a new state of matter in the laboratory : the quark - gluon plasma ( qgp ) .
that strongly interacting matter in conditions of extreme energy densities undergoes a phase transition is in fact a prediction of qcd @xcite . to confirm whether the phase transition indeed occurs in relativistic heavy ion collisions and
that the qgp is formed , one needs a clear signal as a signature for the qgp .
many approaches have been suggested to elucidate the existence of this elusive state of matter , but unfortunately , no single measurement can be singled - out as a `` smoking gun '' candidate .
instead , it appears that many complementary experimental data will require simultaneous analysis @xcite .
one class of observables that appears especially attractive is that of electromagnetic signals .
this owes first to the fact that such probes essentially suffer little or no final state interactions and thus constitute reliable carriers able to report on the local conditions at their emission site .
the calculated emission rates for photons and lepton pairs have been shown to strongly depend on the local density and temperature .
those facts were established some time ago and several heavy ion experiments specializing in the measurement of electromagnetic radiation are either running right now or being planned . in this talk , we specialize on the topic of dilepton measurements and calculations . with respect to photons ,
the extra degree of freedom associated with a variable invariant mass constitutes an advantage that is clear when one considers annihilation reactions , for example . with this in mind , a pioneering experiment was that of the dls @xcite , at the disassembled bevalac at lbnl .
dileptons were also measured by the ceres @xcite and helios @xcite collaborations at the sps at cern , and will be measured in the phenix experiment @xcite at rhic at bnl , and in the hades experiment at gsi @xcite .
the emission of lepton pairs from high temperature matter formed in high energy nucleus@xmath2nucleus collisions is of great theoretical and experimental interest .
it can signal a change in the properties of hadrons ( more precisely , correlation functions ) in hot hadronic matter as it approaches a chiral symmetry or a quark deconfinement phase transition or rapid crossover .
vector mesons are prominent in these studies because of their coupling to the electromagnetic current .
most studies have focussed on the lighter ones , @xmath3 , @xmath0 , and @xmath1 , and on the heavier j/@xmath4 .
the best signals are those that have a characteristic shape or structure .
the @xmath3 meson is very broad in vacuum and undoubtedly gets even broader at finite temperature .
the j/@xmath4 is narrow ; it has a whole literature of its own within the field .
the @xmath0 and @xmath1 mesons are rather narrow , but not so narrow that they will all decay after the hot matter has blown apart in a high energy collision .
this makes them good candidates to study .
in general , one can show that the rate for emitting a dilepton pair ( or a real photon , for that matter ) is related to the retarded photon self - energy , at finite temperature @xcite : @xmath5 im\ , \pi_{\mu \nu}^r ( k ) \frac{1}{e^{\beta \omega } - 1}\ , \end{aligned}\ ] ] where the photon energy is @xmath0 , its three - momentum is @xmath6 , and @xmath7 . a plausible philosophy then consists of performing an order - by - order loop expansion of the finite - temperature photon self - energy in the physical fields relevant for the problem at hand .
one can then establish a correspondence between this approach and that consisting of simply using kinetic theory .
it can be shown that the difference in the dilepton rate between those two scenarios is indeed small up to reasonably high temperatures @xcite .
systematic evaluation of this sort were carried out for the theoretical evaluation of real and virtual photon emission rates from a hot gas of mesons @xcite .
we first concentrate on the rates for low mass dileptons and recall the assumptions inherent to most `` first generation '' estimates .
we do not attempt to do justice to the many calculations in this area : we apologize in advance . in the following , we highlight some general aspects .
having in mind the energy regime germane to the cern sps , a reasonable working hypothesis is to first assume that the hadronic medium is meson - dominated .
this statement is experimentally supported @xcite . also , if thermal equilibrium is indeed established this would influence the soft sector .
this defines a theoretical framework where finite - temperature hadronic field theory can be used to evaluate the photon self - energy .
its imaginary part can then be related to a dilepton emission rate . in practical reality , however , the above - mentioned connection with kinetic theory is in fact what is often used .
our `` first generation '' dilepton emission rate then contains the reactions and decays constructed from an ensemble of the lightest pseudoscalar @xmath8 and vector mesons @xmath9 .
those are @xmath8 : @xmath10 , @xmath11 , and @xmath9 : @xmath12 , and @xmath13 .
the processes included in our rate calculation are then : @xmath14 , @xmath15 , @xmath16 , and @xmath17 .
the respective sizes of those different contribution to the lepton spectrum are compared and discussed in ref .
@xcite ( see also ref .
note that only two - body reactions were first considered , because of phase - space arguments .
we shall return to this point later .
effective chiral lagrangians were used in this calculation , coupled with vector meson dominance @xcite whenever experimental time - like form factors were unavailable . a dynamical model is then needed to translate the rates into actual yields , effectively performing a space - time integration .
the rates obtained by the procedure described above were integrated in a bjorken model @xcite , and in an hydrodynamical model @xcite . running the results thus obtained through the detector acceptances
, one can compare with the measured data .
this has been done by several groups now , and the results are nicely summarized in ref .
@xcite . concentrating on the ceres invariant mass coverage ,
most calculations are seen to underestimate the data in the low mass sector ( @xmath18 300 mev ) .
one must nevertheless point out the remarkable agreement between the different theoretical estimates .
the latter were obtained within different dynamical frameworks ( see @xcite and references therein for details ) and this convergence of theoretical results has been claimed to show the insensitivity of the ceres results to dynamical modeling .
going back to our rates and recalling that those included many reaction channels ( even though a few actually dominate ) , it appears that an independent verification of those estimates is desirable . in this context , we focus on two calculations .
first , a calculation of photon and dilepton production rates from a gas of hot mesons was performed using chiral reduction formul and a virial expansion @xcite .
the results can be expressed in terms of vacuum current - current correlation functions , which can be experimentally constrained .
second , using spectral functions extracted from @xmath19 data one can thus calculate finite - temperature dilepton emission rates @xcite . a comparison between the meson reactions / decays approach outlined earlier and those two calculations appears in fig . 1
. = 150 mev calculated using experimentally - extracted spectral functions @xcite ( open circles ) , chiral reduction formul @xcite ( solid line ) , and meson reactions and decays @xcite ( dashed line).,width=377 ] it is important to point out that ref .
@xcite will not have the equivalent of the radiative decay channels , as the analysis is done starting from @xmath19 data .
strictly speaking , it will then be relevant just below the @xmath3 pole , where the pion channels become large . also , ref .
@xcite does not have the @xmath1 in its formalism : this explains the absence of the peak .
this being said , those three very different approaches exhibit striking agreement .
bear in mind that _ no _ arbitrary normalization enters _ any _ of those calculations .
this is not to say that improvements ca nt be introduced but the unity of the displayed results is nevertheless satisfying .
returning to the underestimation of the low mass data by the majority of the theoretical efforts , several alternative explanations have so far emerged . since it is not the point of this talk to address this specific issue with the detail it deserves , we simply enumerate some of them .
the early suggestion that the ceres data could be accounted for by a dropping of the in - medium @xmath3 mass , has aroused immense theoretical interest .
this conjecture can be incorporated in a model that fits the data @xcite .
another line of reasoning consists of trying to make more complete dilepton rate assessments .
this , combined with a kinetic theory constrained by the available data on spectral and multiplicity distributions , appears to be a step towards a resolution of the discrepancy @xcite .
note that those rates will have to be reconciled with the alternate approaches described on fig .
1 . finally , in - medium baryon effects on the @xmath3 spectral function were shown to produce promising results , in connection with the ceres measurements @xcite .
the last two schools of thought do not invoke dropping @xmath3 masses .
it is obviously too early to pin down the theoretical explanation which will stand the test of time .
however , the community is eagerly awaiting the dilepton transverse momentum spectra which should help to discriminate between theories .
along with the tantalizing low - mass dilepton excess reported by ceres , an excess in the intermediate mass region ( @xmath20 ) has also been seen @xcite . before attempting to make phenomenological estimates , it is best to exercise caution and to tackle this issue purely theoretically .
it has been shown in the past that , right past its threshold , the reaction @xmath21 constituted a sizeable fraction of the lepton pair spectrum @xcite .
this exercise was carried out using an effective chiral lagrangian .
however , the procedure to derive those is unfortunately not unique @xcite . now , because we are no longer in the soft sector , our concern is that the non - uniqueness of the available chiral lagrangians will manifest itself , principally through off - shell effects . with this in mind , we report here on our findings only briefly , leaving a more detailed discussion to be published @xcite . here
, we therefore carry out a mostly pedagogical exercise : what are the sizes of the possible off - shell effects inherent to effective lagrangians , when used to calculate lepton pair production ?
the definite example we have in mind is again @xmath22 , evaluated with the help of vmd .
we are therefore concerned by the different possible manifestations of @xmath23 .
the source of our concern is illustrated as follows .
the most general vertex for @xmath24 decay can be written as @xmath25 where @xmath26 is a form factor , with @xmath27 and @xmath28 being the the @xmath29 and @xmath3 four - momenta . for on - shell @xmath29 decay ,
three of the above terms are identically zero , leaving only two .
the form factors are then chosen such that the experimental width , @xmath30 , is reproduced .
the extra experimental constraint is then the ratio of @xmath31 to @xmath32 wave in the final state , which is measured experimentally @xcite .
it is probably fair to say that the importance of this hadronic constraint has not quite penetrated the heavy ion community , as revealed by a literature survey . when using the above vertex for dilepton production ,
the virtual @xmath3 is pushed off - shell and one has no direct control over such extrapolations .
it turns out that for the case at hand , the uncertainty actually resides in the form factors @xcite .
we calculate the dilepton rate from the reaction above with : * ( 1 ) * = 150 mev using the following interactions : * ( a ) * and * ( b ) * ko and rudaz s chiral lagrangian @xcite ( the difference between * ( a ) * and * ( b ) * is explained in the main text ) ; * ( c ) * the effective chiral lagrangian used by song , ko , and gale @xcite;*(d ) * lagrangian of janssen , holinde , and speth @xcite ; * ( e ) * lagrangian used by xiong , brown , shuryak @xcite
. also shown is the rate calculated from experimentally - fitted spectral functions @xcite ( open circles).,width=377 ] an effective chiral lagrangian constructed by ko and rudaz based on the @xmath33 linear @xmath34 model @xcite ; * ( 2 ) * an effective chiral lagrangian where the vector mesons are introduced as massive yang - mills fields of the chiral symmetry @xcite ; * ( 3 ) * an effective lagrangian , previously used by xiong , shuryak and brown @xcite in connection with photon emission rates ; * ( 4 ) * an effective lagrangian developed by janssen , holinde , and speth @xcite , to address the issue of form factors in the bonn potential .
the dilepton rates evaluated with those interactions are displayed in fig .
the first striking feature of this plot is that the dilepton rates span three orders of magnitude .
off - shell effects are indeed important , as also pointed out by a calculation done along similar lines @xcite . for additional help
, we can however rely on the rates calculated through the experimentally - extracted spectral functions @xcite used previously .
the @xmath35 reaction should be a large ( if not the largest ) contributor to the dilepton signal in the invariant mass region of interest . with the available information we can start our examination of the models .
table 1 summarizes their empirical predictions , along with the actually measured quantities .
.the phenomenology of different models . [ cols="^,^,^,^,^,^,^ " , ] parameter set i for the interaction of ko _ et al .
_ is taken from ref .
@xcite . concerning the observables in table 1
, this parameter set does not actually represent the lowest @xmath36 : parameter set ii satisfies this criterion .
it is interesting to notice that a modest variation in @xmath36 for the fit of the on - shell empirical observables translates into a large variation of the predicted dilepton rate .
2 also contains the rate deduced from the experimentally - fitted spectral functions @xcite .
because the process discussed here is actually included in the spectral function analysis along with other channels ( which are known to interfere constructively ) , the open circles represent a strong upper limit for the calculations discussed here .
this upper bound is strongly violated by our parameter set ii for the lagrangian of ref .
@xcite . to a lesser extent
this is also the case for the results of calculations with the interaction of ref .
@xcite . in the remaining models in table 1 ,
the lowest @xmath36 is obtained with the lagrangian of ref .
parameter set i of ko _ et al .
_ has as good a @xmath36 , but the slope of the dilepton spectrum departs considerably from that of the spectral function analysis , with a violation of the upper bound just below @xmath37 = 2 gev .
the lagrangian of xiong _ et al .
_ produces a high @xmath36 , with a much flatter slope .
we could deem here the interaction of janssen _ et al . _ to be the most satisfactory .
it can also be made to saturate the spectral function curve with a proper choice of form factors , without spoiling its on - shell properties .
however , the spectral function analysis can be made to be specific to the @xmath38 channel and the qualitative aspects discussed here can be set on a quantitative footing ; we stick here to our pedagogical goal and postpone this discussion @xcite .
it is clear which caveats will then await us when we embark on the phenomenological analysis of the experimental data .
the purpose of this exercise was to show that using all of the available empirical constraints , one could indeed restore some certainty to an apparently ambiguous situation .
there are thus reasons to be optimistic about this type of analyses .
the experimentally - extracted spectral function prove to be helpful , and serve here as baseline calculations .
then , the effective lagrangian theory can be used to extrapolate into the realm of finite - temperature and many - body effects .
a last application of effective hadronic lagrangians will cover here the mixing of the @xmath0 and @xmath1 vector mesons . in vacuum
the @xmath1 meson is almost entirely @xmath40 in its valence quark content while the @xmath0 meson is almost entirely nonstrange .
there is a small mixing as evidenced by the observed decay mode @xmath41 and by previous studies using effective hadronic lagrangians .
we will study the change in masses and mixing angles of the @xmath0 and @xmath1 at moderate temperatures using only conventional ideas .
we consider our study a simple extrapolation of known physics which may help in deciphering future experiments .
specifically , we study the scattering of these mesons from thermal pions which are the most abundant mesons at temperatures below 100 mev or so .
we use effective lagrangian techniques to model the relatively soft interactions coupling pions , @xmath0 , and @xmath1 mesons , fitting the parameters to known physical quantities .
the resulting scattering amplitudes are used in a virial expansion to compute the vector meson properties at finite temperature . a survey of the literature on effective hadronic lagrangians and the review of particle physics @xcite suggests that the interactions of relevance involve the 3point vertices @xmath42 , @xmath43 , @xmath44 , and @xmath45 .
the vector self@xmath2energies are obtained by computing one loop diagrams at finite temperature @xcite .
the interaction involving the @xmath3 meson is described by the wess
zumino term @xcite .
@xmath46 the octet and singlet fields , @xmath47 and @xmath48 , are expressed in terms of the ( vacuum ) physical fields with a mixing angle @xmath49 . @xmath50
ideal mixing is defined such that the physical @xmath1 meson would not couple to nonstrange hadrons .
this corresponds to @xmath51 .
the real world is not far from that .
durso @xcite , for example , fits 39.2@xmath52 , while the review of particle physics quotes 39@xmath52 based on the gell
mann okubo mass formula .
the coupling constant @xmath53 is related to the coupling @xmath54 used by gomm , kaymakcalan , and schechter @xcite by @xmath55 .
durso fits @xmath56 mev@xmath57 with a pseudoscalar mixing angle @xmath58 = @xmath59 .
we will insure that our parameters reproduce the @xmath60 decay rate .
accepting durso s value of @xmath53 we fit @xmath61 .
the @xmath62 has a branching ratio of more than 50% into @xmath63 and less than 1.5% into @xmath64 .
thus , the @xmath65 meson is important for the mass shift of the @xmath0 meson at finite temperature , as noticed by shuryak @xcite .
the interactions are assumed to be su(3 ) symmetric with su(3 ) broken only by mass terms .
the lagrangian is @xcite @xmath66 the two coupling constants can be inferred from the decay rate @xmath67 and from the ratio of the @xmath31 wave content of the decay amplitude to its @xmath32 wave content .
the review of particle physics gives the width 142@xmath688 mev and the measured @xmath69 ratio [email protected] .
fitting to central values determines the most likely values of the coupling constants to be @xmath70 gev and @xmath71 gev@xmath72 .
the location of the poles and the mixing angle at finite temperature are obtained by finding the zeros of the inverse propagator in the ( vacuum ) physical @xmath0@xmath1 basis .
@xmath74 is the 2@xmath752 mass matrix at zero temperature .
it is diagonal with components @xmath76 and @xmath77 .
the self@xmath2energy has contributions from the @xmath3 meson and from the @xmath65 meson .
the first was calculated by haglin and gale @xcite .
the second is readily calculated by the usual finite - temperature rules @xcite .
we restrict our attention to the finite - temperature contribution .
we consider only @xmath0 and @xmath1 mesons at rest in the many body system ( @xmath78 ) since that is where temperature will have its maximum impact . as @xmath79 increases ,
many@xmath2body effects will decrease , and in the limit @xmath80 they will disappear altogether .
finally , the imaginary part of the self@xmath2energy is small compared to the real part ( inclusive of @xmath74 ) and so we do not include it in our present calculation .
we find @xmath81 where @xmath82 measures the deviation from ideal mixing .
and @xmath1 mesons as functions of temperature due to scattering from thermal pions .
multiple pion scattering and scattering from other thermal mesons should become important above 100 mev temperature.,width=302 ] the scalar functions @xmath83 and @xmath84 are given in ref .
@xcite .
after diagonalization one finds the mass shifts as functions of the temperature .
they are displayed in fig .
although the results are plotted up to a temperature of 150 mev to see the effect , other scattering processes will come into play above 100 mev , as mentioned earlier .
the @xmath0 mass goes down and a shift of roughly 13 mev is reached at a temperature of 140 mev .
the @xmath1 mass increases monotonically to about 0.6 mev above its vacuum value at a temperature of 140 mev .
the mixing angles remain uninterestingly small at all temperatures @xcite , but the mass shift of the @xmath0 is potentially detectable in future experiments .
in conclusion , in our brief description of calculations for the low mass sector we have shown agreement between different theoretical approaches to dilepton rate calculations . at higher invariant masses , the considerable uncertainties associated with off - shell effects have been confirmed .
nevertheless , we argue that the available phenomenology helps considerably in narrowing down the possibilities .
finally , we compute the finite - temperature change in the masses of the @xmath0 and @xmath1 mesons and their mixing angle .
i am happy to acknowledge the collaborators involved with some aspects of this work .
i am grateful to j. steele and z. huang for sharing with me their results .
this work was supported in part by by the natural sciences and engineering research council of canada , in part by the fcar fund of the qubec government , and in part by the us department of energy under grant de - fg02 - 87er40328 .
a. drees , nucl .
phys . * a 610 * , 536c ( 1996 ) ; see also these proceedings . james v. steele , hidenaga yamagishi , and ismail zahed , phys.lett . * b 384 * , 255 ( 1996 ) .
z. huang , phys .
* b 361 * , 131 ( 1995 ) . k. haglin , proceedings of the int / rhic workshop on the electromagnetic probes of the quark - gluon plasma , seattle , january 24 - 27 1996 ; j. murray , w. bauer , and k. haglin , hep - ph/9611328 .
r. rapp , g. chanfray , and j. wambach , nucl .
phys . * a 617 * , 472 ( 1997 ) .
g. q. li , c. m. ko , g. e. brown , and h. sorge , nucl . phys . * a 611 * , 539 ( 1996 ) .
c. song , c. m. ko , and c. gale , phys . rev .
* d 50 * , r1827 ( 1994 ) ; the lagrangian parameters for the dilepton calculation reported on here are from c. song , phys .
* c 47 * , 2861 ( 1997 ) .
see ulf - g .
meissner , phys .
rep . * 161 * ( 1988 ) 213 , and references therein .
s. gao and c. gale , submitted to phys .
d. h. albrecht _
et al . _ , z. phys . * c 58 * , 61 ( 1993 ) .
p. ko and s. rudaz , phys . rev . * d50 * , 6877(1994 ) .
l. xiong , e. shuryak and g. e. brown , phys . rev . * d46 * , 3798 ( 1992 ) .
g. janssen , k. holinde and j. speth , phys . rev . *
c49 * , 2763 ( 1994 ) .
j. k. kim , p. ko , k. y. lee , and s. rudaz , phys . rev . *
d 53 * , 4787 ( 1996 ) .
j. i. kapusta , _ finite temperature field theory _ , cambridge university press , cambridge , england ( 1989 ) .
charles gale , david seibert , and joseph kapusta , phys .
d 56 , 508 ( 1997 ) ; see also erratum , in print . | we first review the calculations of low invariant mass dilepton production in relativistic heavy ion collisions , using effective hadronic lagrangians .
we go through some of the theoretical techniques used in this kinematical region and we consider some of the appropriate experimental measurements . moving up to the intermediate invariant mass region , we point out some of the uncertainties that show up in theoretical estimates .
those originate mainly from off - shell effects . finally , as an application of hadronic chiral lagrangians at finite temperature , we compute the mass shifts and mixing of the @xmath0 and @xmath1 mesons due to scattering from thermal pions . |
electromagnetic radiation represents penetrating probes mapping out the full space - time history of strongly interacting matter in the course of heavy - ion collisions .
particularly interesting is the stage of maximum temperature , where a quark - gluon plasma might be formed .
various properties of the quark - gluon plasma are calculable from first principles exploiting numerical lattice qcd techniques .
the results , such as thermodynamical properties , can be interpreted within quasi - particle models @xcite .
other quantities , such as the electromagnetic emission rates , are accessible within perturbation theory @xcite .
the motivation to measure electromagnetic signals in ultralrelativistic heavy - ion collisions is based on the hope to have a direct access to the hottest and densest stages of matter ( cf .
@xcite ) . according to the present understanding of the dynamics of central heavy - ion collisions at cern - sps , the temperature range covered is @xmath0 mev ( where @xmath1 mev is the confinement temperature ) . in this regime
the electromagnetic emission rates from deconfined and confined hadronic matter are fairly similar ( cf .
@xcite ) . indeed , taking into account the empirical electromagnetic formfactors from the reaction @xmath2 and including vector - axial vector mixing @xcite one gets a dilepton rate which coincides with the lowest - order @xmath3 rate for invariant masses
@xmath4 gev @xcite . below 1.1 gev the dilepton spectrum of hadronic matter
is determined by the sharp @xmath5 and @xmath6 peaks and the broad @xmath7 peak .
hadronic reactions broaden the @xmath7 peak so much that the net dilepton spectrum again resembles that from the @xmath3 rate down to @xmath8 mev @xcite .
it is therefore tempting to entirely describe the spectra at @xmath9 mev by the lowest - order @xmath3 rate , thus treating it as a convenient parameterization .
analog arguments are not available for the real photon rate .
therefore , we use as a working hypothesis the lowest - order born rates of the processes @xmath10 and @xmath11 for thermal photons .
many hadron observables at cern - sps energies can be described by thermal models . in line with this finding
we will describe the secondary real and virtual photons by a thermal source which adds to the drell - yan - like yields and hadron decay contributions ( hadronic cocktail and correlated semi - leptonic decays of open charm mesons ) .
by now the following experimental data are available : ( i ) lead beam at 158 a@xmath12gev and au , pb targets : ceres ( low - mass @xmath13 ) , na50 ( intermediate - mass @xmath14 ) , wa98 ( direct photons ) , ( ii ) lead beam at 40 a@xmath12gev and au target : ceres ( low - mass @xmath13 ) , ( iii ) sulfur beam at 200 a@xmath12gev and au , u , w targets : ceres ( low - mass @xmath13 ) , na38 ( intermediate - mass @xmath14 ) , helios/3 ( low- and intermediate - mass @xmath14 ) , wa80 ( upper bounds on direct photons ) .
in addition , data with proton beams at these energies and also 450 gev for various target nuclei are at our disposal .
the lowest - order rates are in boltzmann approximation @xmath15 , \label{rates_1}\ ] ] where @xmath16 ( @xmath17 ) is the virtual ( real ) photon s 4-momentum ; effects of a finite chemical potential @xmath18 can be absorbed in an additional normalization factor .
note the necessary occurrence of the medium s
4-velocity @xmath19 to build up lorentz - invariant rates .
to reduce the number of parameters in the space - time integration , we replace in ( [ rates_1 ] ) @xmath20 , @xmath21 and @xmath22 yielding @xmath23 thus not relying on a particular model for @xmath24 , @xmath25 and @xmath26 ( for the explicit form of ( [ rates_2 ] ) cf .
this approach is in the spirit of the usual parameterization of the transverse momentum spectra of hadrons by an exponential with a slope parameter and a normalization factor . as an example
we show in figure 1 our results for the invariant mass spectra of dileptons and the momentum spectrum of photons for the experiments of group ( i ) ( cf .
@xcite for more details ) .
also the transverse momentum spectra of dileptons are described very well @xcite .
the na50 acceptance is incorporated only in an approximate way here ; the use of the correct filter reproduces the data perfectly @xcite .
a comparison of our model with the preliminary ceres data @xcite of the experiment group ( ii ) is shown in figure 2 for @xmath27 mev and a fairly large value of @xmath28 .
deliver a better description of the spectral shape but need even larger @xmath28 . ] since the experiments of group ( iii ) cover a large rapidity range an additional gaussian smearing of the source in ( [ rates_2 ] ) needs to be included @xcite . in figure 3 the strength distributions of the thermal dilepton source without and with smearing
are displayed .
as shown in @xcite the shape of the spectra can be well described by @xmath29 mev , however , different values of @xmath28 are needed .
this might be attributed to an inadequate rapidity distribution of our thermal source and different centrality selections in the experiments .
( -3.5,-2.45)(-3.55,-1.55 ) ( -2.9,-2.6)thermal ( -3.0,-2.2)(-3.0,-1.9 ) ( -2.7,-2.3)cocktail ( -2.3,-1.4)(-2.88,-1.68 ) ( -1.5,-1.4)drell - yan ( -3.1,-2.9)(-2.76,-2.66 ) ( -3.6,-2.9)charm ( -2.85,-3.41)(-2.65,-3.35 ) ( -3.5,-3.4)thermal ( -2.6,-2.77)(-2.,-2.2 ) ( -2.5,-2.9)thermal ( -2.5,-3.25)(@xmath30 ) ( -3.3,-2.3)(-2.95,-1.28 ) ( -3.3,-2.5)pqcd + ( 2.,1.7)(2.0,2.3 ) ( 2.,1.5)cocktail ( 2.8,3.56)(2.33,3.23 ) ( 3.5,3.6)thermal + + by using a model for @xmath31 and @xmath32 and neglecting spatial gradients one can equally well describe the dilepton spectra of the ceres and na50 collaborations .
as shown in @xcite , there is virtually no difference with the results displayed in figure 1 .
the needed initial temperature is 210 mev , for a given freeze - out temperature of 120 mev .
while the analysis of the low - mass dileptons uses a hadronic cocktail and is normalized to the mean multiplicity of charged hadrons , the analyses of the intermediate - mass dileptons and photons need careful estimates of the hard drell - yan - like yields and the open charm contribution .
we used the pythia event generator with charm quark mass of 1.5 gev , intrinsic parton transverse momentum distribution width of 0.8 gev and photon cut - off parameter of 1 gev .
the parameter adjustments are done by comparing with @xmath33 and @xmath34 data , and then scaling to heavy - ion collisions .
the possible subtleties of this procedure are described in @xcite .
in summary we have demonstrated that the superposition of background contributions and a thermal source of real and virtual photons describes the shape of all data from central heavy - ion collisions at cern - sps .
we find evidence for a large and long - lived thermal source in lead beam experiments at 158 a@xmath12gev .
the normalization factor @xmath28 translates into a lifetime of 23 fm / c , in accordance with @xcite , when assuming a spatial volume of the fireball 1440 @xmath35 as found for a radius of 7 fm .
the space - time averaged temperature of @xmath36 mev coincides with both the confinement temperature and the temperature needed to describe the hadron multiplicity ratios in thermal models . using a model for the temperature and volume evolution of the fireball we find a maximum temperature of 210 mev indicating an initial state in the deconfinement region .
the sulfur beam data point to a similar value of @xmath37 mev , while the recent preliminary lead beam data at 40 a@xmath12gev can be described by @xmath38 mev .
the present analysis strategy needs to be extended to understand the centrality dependence of the intermediate - mass dileptons as measured by the na38/50 collaborations @xcite .
the results of the high - statistic data sampling of dileptons in the lead beam induced reactions at 158 a@xmath12gev by the ceres collaboration is eagerly waited for to decide whether a smooth @xmath3-like rate is still sufficient .
we used it here as a convenient parameterization , and this is not a substitute to a more detailed microscopic calculation .
the consistent description of the real and virtual photon data makes an interpretation of the na50 data by an anomalous open charm enhancement unlikely .
nevertheless , the explicit measurement of the charm yield , as envisaged by the na60 collaboration , is needed to arrive at a better understanding of the various dilepton sources .
9 a. peshier , b. kmpfer , o.p .
pavlenko , g. soff , phys . rev .
d 54 ( 1996 ) 2399 , + a. peshier , b. kmpfer , g. soff , phys .
c 61 ( 2000 ) 045203 .
p. aurenche , f. gelis , r. kobes , h. zaraket , phys .
d 60 ( 1999 ) 076002 .
k. gallmeister , b. kmpfer , o.p .
pavlenko , prog .
42 ( 1999 ) 333 .
j. kapusta , p. lichard , d. seibert , phys .
d 44 ( 1991 ) 2774 .
li , c. gale , phys .
c 58 ( 1998 ) 2914 , phys .
81 ( 1998 ) 1572 .
z. huang , phys .
b 361 ( 1995 ) 131 .
r. rapp , j. wambach , hep - ph/9909229 , adv .
phys . , in press .
k. gallmeister , b. kmpfer , o.p .
pavlenko , phys .
c 62 ( 2000 ) 057901 .
k. gallmeister , b. kmpfer , o.p .
pavlenko , c. gale , hep - ph/0010332 , nucl .
a , in press . l. capelli ( na50 ) , this volume ; ph.d .
thesis , univ .
de lyon 2001 .
b. lenkeit ( ceres ) , nucl . phys . a 661 ( 1999 ) 23c ; ph.d .
thesis , heidelberg 1998 .
e. scomparin ( na50 ) , nucl .
phys . a 610 ( 1996 ) 331c , j. phys .
g 25 ( 1999 ) 235c .
aggarwal et al .
( wa98 ) , phys .
85 ( 2000 ) 3595 .
s. damjanovic , k. filimonov ( ceres ) , poster p084 at qm2001 .
k. gallmeister , b. kmpfer , o.p .
pavlenko , phys .
b 473 ( 2000 ) 20 . k.o .
gallmeister , ph.d .
thesis , tu dresden 2001 .
. hung , e.v .
shuryak , phys .
c 57 ( 1998 ) 1891 . | a unique parameterization of secondary ( thermal ) dilepton and photon yields in heavy - ion experiments at cern - sps is proposed . adding those thermal yields to background contributions the spectral shapes of the ceres / na45 , na38 , na50 , helios/3 and wa98 data from experiments with lead and sulfur beams can be well described . |
in many open quantum systems , the coupling between the system and the bath can be considered as a small parameter . in this case
the application of second order perturbation leads to a master equation of redfield or lindblad type @xcite .
their numerical implementation is straightforward and not computationally expensive .
however , for many physical systems of current interest it has been shown that the weak coupling approximation is not justified .
one example is the energy transfer process in photosynthetic complexes where the magnitude of the system - bath coupling is comparable to the electronic couplings @xcite .
there are only a few non - perturbative techniques to obtain the numerically exact dynamics ; examples include the hierarchy master equation @xcite , the quasi - adiabatic propagator path integral ( quapi ) @xcite , and the multiconfiguration time - dependent hartree ( mctdh ) @xcite approach .
however , these methods are computationally demanding and also not trivial to implement .
recently , a polaron transformed second order master equation @xcite and its variational form @xcite have been derived to study the dynamics of open quantum systems at strong coupling .
this approach transforms the total hamiltonian into the polaron frame such that the system hamiltonian is dressed by a polaron . the master equation
is then obtained by applying perturbation theory to the transformed system - bath interaction term .
this approach extends the regime of validity of the master equation to stronger system - bath coupling , provided that the electronic couplings ( or tunneling matrix elements ) are small compared to the typical bath frequency .
when this condition is not fulfilled the polaron is too sluggish to accurately follow the system motion and the polaron transformation may perform worse than the standard master equation approach . in order to partially overcome this difficulty ,
the variational method has been developed as a generalization of the polaron transformation @xcite .
instead of performing the full transformation , the variational polaron approach seeks for an optimal amount of transformation , depending on the properties of the bath .
thus it is able to interpolate between the strong and weak coupling regimes and to capture the correct behavior over a much broader range of parameters .
both the polaron and variational master equations have the attractive feature of being computationally economic ( they have the same computational complexity as the redfield equation ) and are therefore suitable for studying large systems . however , a thorough assessment of the accuracy of second order perturbation theory in the polaron and variational polaron frames is still lacking .
it is not exactly clear how the accuracy depends on the properties of the bath , namely the bath relaxation time and the coupling strength .
one of the main goals of this work is to provide such a benchmark . instead of studying the dynamics ,
here we focus on the equilibrium density matrix .
focusing on this quantity offers two key advantages .
firstly , in the equilibrium case the second order perturbation is the only approximation involved . in the derivation of second order master equations , additional approximations generally must be invoked , such as factorized initial conditions , the born - markov approximation , the rotating wave approximation , etc .
these additional restrictions prevent a clear assessment of the isolated role of second order perturbation theory and the merits of the polaron transformation . thus studying the equilibrium density matrix offers a direct comparison of the various perturbation methods .
a second advantage of studying the equilibrium state is that it is much easier to obtain numerically exact results .
therefore we are able to systematically explore a large range of the parameter space that is often not possible with other exact treatments of the dynamics . in the next section ,
the details of the spin - boson model used in the remainder of the text are outlined . following this ,
the polaron transformation and its variational extension are applied to the hamiltonian in sec .
[ sec : transformations ] . in sec .
[ sec : pt ] , the second - order corrections to the equilibrium reduced density matrix are derived in the original , polaron and variational polaron frames . in the ensuing section ,
results for the various perturbation theories are compared with exact numerical results from path integral calculations over a broad range of the parameter space .
it is found that the second order results in the original frame are accurate for small system - bath coupling , the full polaron results are accurate in the opposite regime of strong coupling , and the variational method is capable of interpolating between these two extremes .
all three approaches become less accurate for slow baths .
the spin - boson model is a paradigm for the study of quantum dissipative systems .
it has been used to investigate the energy transfer in light harvesting systems @xcite , the problem of decoherence in quantum optics @xcite , tunneling phenomena in condensed media @xcite , and quantum phase transitions @xcite .
the spin - boson model consists of a two - level system coupled to a bath of harmonic oscillators .
its hamiltonian can be written as ( we set @xmath0 ) @xmath1 where @xmath2 ( @xmath3 ) are the usual pauli matrices , @xmath4 is the energy splitting between the two levels , and @xmath5 is the tunneling matrix element .
the bath is modeled as a set of harmonic oscillators labeled by their frequencies , @xmath6 , and couplings to the two - level system denoted by @xmath7 .
the properties of the harmonic bath are completely determined by the spectral density , @xmath8 . throughout the paper
, we use a super - ohmic spectral density with an exponential cut - off , @xmath9 where @xmath10 is the system - bath coupling strength and has the dimension of frequency .
the cut - off frequency is denoted by @xmath11 , and its reciprocal governs the relaxation time of the bath , @xmath12 .
the polaron transformation is generated by the unitary operator @xmath13,\end{aligned}\ ] ] which displaces the bath oscillators in the positive or negative direction depending on the state of the two - level system .
the parameter @xmath14 determines the magnitude of the displacement for each mode .
setting @xmath15 corresponds to the full polaron transformation whereas @xmath16 corresponds to no transformation .
the variational method allows us to determine an optimal value of @xmath14 that lies in between these two limits , @xmath17 , making the transformation valid over a wider range of parameters . applying the transformation to the total hamiltonian , we have @xmath18 where the total free hamiltonian is @xmath19 .
the transformed system hamiltonian is given by @xmath20 and the bath hamiltonian remains unaffected , @xmath21 .
the transformed interaction hamiltonian becomes @xmath22 where @xmath23 and @xmath24 is the product of displacement operators , @xmath25 .
the tunneling rate is renormalized by the expectation value of the bath displacement operators , @xmath26 , where @xmath27 } { \mbox{tr}_b[\mbox{e}^{-\beta { \hat}{h}_b } ] } , \\ & = & \mbox{exp}\big[-2\sum_k\frac{f_k^2}{\omega_k^2 } \coth(\beta\omega_k /2)\big].\end{aligned}\ ] ] note that the interaction term is constructed such that @xmath28 .
following silbey and harris @xcite , we use the bogoliubov variational theorem to determine the optimal values for the set @xmath29 .
we first compute the bogoliubov - feynman upper bound on the free energy , @xmath30 @xmath31 + \langle { \hat}{h}_i \rangle_{{\hat}{h}_0}.\end{aligned}\ ] ] since @xmath32 by construction , the upper bound is solely determined by the free hamiltonian . the variational theorem
states that @xmath33 where @xmath34 is the true free energy .
therefore , we want to make this bound as small as possible by minimizing @xmath30 with respect to @xmath29 , i.e. @xmath35 . the minimization condition leads to @xmath36^{-1 } , \label{f - omega}\end{aligned}\ ] ] where @xmath37 . in the continuum limit
, the renormalization constant can be written as @xmath38 .
\label{self - consistent}\end{aligned}\ ] ] since @xmath39 is also a function of @xmath40 , the above equation must be solved self - consistently .
this section is dedicated to finding the second order correction to the equilibrium state of the system .
the exact equilibrium reduced density matrix can be formally written as @xmath41}{\mbox{tr}_{s+b}[\mbox{e}^{-\beta { \hat}{h}_{tot}}]}. \label{exact_dm}\end{aligned}\ ] ] expanding the operator @xmath42 up to second order in @xmath43 , we have @xmath44 . \nonumber \\\end{aligned}\ ] ] the above expansion is similar to the dyson expansion , with @xmath45 treated as imaginary time . since @xmath46 ,
the leading order correction to @xmath47 is of second order in @xmath43 . inserting the above expression into eq .
( [ exact_dm ] ) and keeping terms up to the second order in @xmath43 , the system equilibrium state can be approximated as @xcite @xmath48 where @xmath49 , \,\,\,\,\,\ , \ , z^{(2)}_s= \mbox{tr}_s[{\hat}{a}],\end{aligned}\ ] ] and @xmath50 is the bath correlation function in imaginary time , @xmath51}{\mbox{tr}_b[\mbox{e}^{-\beta { \hat}{h}_b}]}.\end{aligned}\ ] ] the non - vanishing bath correlation functions are @xmath52 \frac{\sinh\big(\frac{1}{2}(\beta -2\tau ) \omega \big)}{\sinh(\beta \omega/2)},\\ c_{xx } ( \tau ) & = & \frac{\delta_r^2}{8 } ( \mbox{e}^{\phi(\tau ) } + \mbox{e}^{-\phi(\tau ) } -2),\\ c_{yy } ( \tau ) & = & \frac{\delta_r^2}{8 } ( \mbox{e}^{\phi(\tau ) } - \mbox{e}^{-\phi(\tau ) } ) , \\
c_{zz } ( \tau ) & = & \int^\infty_0 \frac{d\omega}{\pi } j(\omega)[1-f(\omega)]^2
\frac{\cosh\big(\frac{1}{2}(\beta -2\tau ) \omega \big)}{\sinh(\beta \omega/2 ) } , \label{eq : correlation_functions}\end{aligned}\ ] ] where @xmath53 it is useful at this point to analyze the behavior of the perturbation theory at strong coupling in the polaron frame . as seen from eq .
( [ self - consistent ] ) , when @xmath54 then @xmath55 and the system becomes incoherent since the coherent tunneling element vanishes . at the same time , @xmath56 as @xmath57 so that all of the above correlation functions vanish , and hence also the second order correction to @xmath58 .
therefore in this limit , the equilibrium density matrix is only determined by the energy splitting of the two levels , @xmath59 .
the full polaron result can be conveniently obtained by setting @xmath60 ; the only non - vanishing correlation functions in this case are @xmath61 and @xmath62 .
the opposite limit of @xmath63 corresponds to performing no transformation and @xmath64 is the only non - zero correlation function . for comparison , below we will also include the results from these two limiting cases .
in this section , we compare the results from second order perturbation theory ( 2nd - pt ) in the original [ @xmath63 ] , the polaron [ @xmath60 ] , and the variational polaron [ @xmath39 as in eq .
( [ f - omega ] ) ] frames with those from numerically exact imaginary time path integral calculations .
we compute the expectation value , @xmath67 , since it is not affected by the transformation , @xmath68 .
therefore this quantity allows us to make a direct comparison between the path integral results , which provide the density matrix in the original frame , and the ( variational ) polaron results .
results from the transformed zeroth order density matrix , @xmath69 , which depends only on the renormalized system hamiltonian @xmath70 , are also included .
we first calculate @xmath67 as a function of the dissipation strength for fast , slow and adiabatic baths , assessing the accuracy of 2nd - pt for different bath cut - off frequencies .
we then conclude this section by presenting phase diagrams of the relative errors of the various methods as functions of the dissipation strength and the bath cut - off frequency .
this allows us to establish the regimes of validity of each approach across the entire range of bath parameters . throughout the paper
, we set @xmath71 and @xmath72 .
+ the value of @xmath67 is plotted as a function of the dissipation strength , @xmath10 , in fig .
[ fig : fast - bath ] for a fast bath , @xmath73 .
firstly , it can be seen that the result from the usual 2nd - pt in the original frame ( dashed line ) is linearly dependent on @xmath10 .
while this approach is accurate at small @xmath10 , it quickly degrades as the coupling increases . on the other hand , the results from 2nd - pt in the polaron ( empty circles ) and the variational polaron ( solid line ) frames are in excellent agreement with the exact path integral result ( solid dots ) over the entire range of dissipation .
the zeroth - order result for @xmath67 in the the polaron frame ( crosses ) tends to overestimate the correction , whereas the variational frame result ( diamonds ) provides at least a qualitatively correct description . .
comparison of the approximate methods with the exact path integral results as a function of the dissipation , @xmath10 , for @xmath74 , and @xmath75 . plotted
are the values of @xmath67 from the zeroth order density matrix in the polaron frame ( light blue crosses ) and the variational polaron frame ( dark blue diamonds ) , as well as from the density matrix in the original frame with the second order correction ( red dashed line ) , the polaron frame ( purple empty circles ) and the variational polaron frame ( solid green line ) .
the exact path integral results are shown as filled ( black ) dots .
, width=360 ] fig .
[ fig : slow - bath ] displays the opposite case , when the bath is slow as compared to the tunneling rate , @xmath76 . at small and intermediate @xmath10
, it can be seen that the polaron method with 2nd - pt fails to predict the correct behavior , while the usual 2nd - pt result in the original frame agrees well with the exact result . as with the case of the fast bath , at large @xmath10 , the polaron method provides an accurate description , while the original frame 2nd - pt breaks down . the variational polaron method ( with 2nd - pt ) interpolates between these two methods , providing accurate results over a large range of the dissipation strength .
the failure of the full polaron method is due to the fact that the bath oscillators are sluggish and are not able to fully dress the system .
therefore the full polaron displacement is no longer appropriate .
it can also be seen that the second order correction in the full polaron frame is huge at small @xmath10 ( the difference between the crosses and open circles ) .
this should cast doubt on the validity of 2nd - pt in this case since the perturbative correction should be small .
it can also be observed that there is a discontinuity in the variational result ( both with and without 2nd - pt ) at the critical point of @xmath77 .
the variational approach exhibits a rather abrupt transition from a small transformation [ @xmath78 to the full polaron transformation [ @xmath79 .
this discontinuity , which is an artifact of the transformation rather than a physical phase transition , has been predicted by silbey and harris @xcite for a slow bath .
the discontinuity comes from solving the self - consistent equation in eq .
( [ self - consistent ] ) . over a certain range of dissipation strengths , there exist multiple solutions to the self - consistent equation . according to the variational prescription
, the solution with the lowest free energy is selected .
this causes a `` jump '' in the solution , as depicted in fig .
[ fig : self - consistent ]
. it is also observed in fig .
[ fig : slow - bath ] that the variational polaron result is least accurate around the transition point .
therefore , it will be worthwhile to look for a better variational criterion that removes this discontinuity , which can hopefully provide a uniformly accurate solution . .
comparison of the approximate methods with the exact path integral results as a function of the dissipation , @xmath10 , for @xmath80 , and @xmath81 . plotted
are the values of @xmath67 from the zeroth order density matrix in the polaron frame ( light blue crosses ) and the variational polaron frame ( dark blue diamonds ) , as well as from the density matrix in the original frame with the second order correction ( red dashed line ) , the polaron frame ( purple empty circles ) and the variational polaron frame ( solid green line ) .
the exact path integral results are shown as filled ( black ) dots . ,
width=360 ] .
the solutions to the self consistent equation eq .
( [ self - consistent ] ) are the points when @xmath82 . at @xmath83
there is only one solution to the self - consistent equation . at @xmath84 ,
multiple solutions start to develop , but the solution with the lowest free energy ( denoted by the empty circle ) is chosen . at the critical point of @xmath85 , there is a `` jump '' in the lowest free energy solution which causes a discontinuity in the transformation . , width=360 ] .
comparison of the approximate methods with the exact result [ from eq .
( [ adiabatic ] ) ] as a function of the dissipation , @xmath10 , for @xmath87 , and @xmath88 .
plotted are the values of @xmath67 from the density matrix with the second order correction in the original frame ( dashed red line ) , the polaron frame ( empty blue circles ) and the variational polaron frame ( solid green line ) .
the exact results are shown as filled ( black ) dots.,width=360 ] in the adiabatic limit ( @xmath89 ) the exact solution to the equilibrium state of the system can be obtained analytically . in this regime , the partition function is given by @xcite @xmath90\big]\big\ } , \label{adiabatic}\end{aligned}\ ] ] where @xmath91 is the bath correlation function [ in the original frame with @xmath63 ] in the adiabatic limit , @xmath92 .
the expectation value , @xmath67 , can be obtained from the partition function via the following relation @xmath93 in the regime where @xmath94 , the transition from @xmath63 to @xmath60 in the variational method is sharp , as seen in fig .
[ fig : adiabatic ] . before the transition
, the variational polaron result coincides with the exact result and that of perturbation theory in the original frame .
the full polaron result fails to give the correct results , and even predicts the wrong limiting behavior as @xmath95 .
after the transition , the variational result deviates from the exact result and becomes essentially the same as the full polaron result . as @xmath10 increases , results from both methods approach the exact result while the untransformed 2nd - pt breaks down as seen before . to get a better perspective of how the accuracy of 2nd - pt in different frames depends on the properties of the bath
, we calculate the relative errors over the entire range of the bath parameters .
the relatives errors are defined as @xmath96 where the subscripts `` pert '' and `` pi '' denote the perturbative calculation and path integral calculation respectively .
[ fig : contour ] displays the respective errors for the three methods as a function of the cut - off frequency and the coupling strength . as seen in fig .
[ fig : contour](a ) , the usual 2nd - pt without transformation breaks down at large @xmath10 .
it is also less accurate when the cut - off frequency is small , which corresponds to a highly non - markovian bath . on the other hand
, the 2nd - pt in the full polaron frame fails at small @xmath10 and @xmath11 [ see fig .
[ fig : contour](b ) ] .
these two approaches provide complementary behavior as a function of the coupling strength ; the polaron method is essentially exact for large @xmath10 , while the usual 2nd - pt is exact for small @xmath10 .
the variational calculation is valid over a much broader range of parameters [ see fig .
[ fig : contour](c ) ] , and essentially combines the regimes of validity of the full polaron result and 2nd - pt in the original frame .
it is only slightly less accurate in the slow bath regime around the region where the discontinuity appears that was discussed above . , the relative errors of the second order perturbation theory as defined in eq . ( [ eq : error ] ) in ( a ) the original frame , ( b ) the full polaron frame , and ( c ) the variational polaron frame.,title="fig:",width=288 ] , the relative errors of the second order perturbation theory as defined in eq .
( [ eq : error ] ) in ( a ) the original frame , ( b ) the full polaron frame , and ( c ) the variational polaron frame.,title="fig:",width=288 ] , the relative errors of the second order perturbation theory as defined in eq .
( [ eq : error ] ) in ( a ) the original frame , ( b ) the full polaron frame , and ( c ) the variational polaron frame.,title="fig:",width=288 ]
in conclusion , we have provided a thorough assessment of the accuracy of the polaron and variational polaron methods .
we compared the second order perturbation results in the polaron and variational polaron transformed frames with numerically exact path integral calculations of the equilibrium reduced density matrix . focusing on the equilibrium properties allowed us to systematically explore the whole range of bath parameters without making any additional approximations as is generally required to simulate the dynamics . as a function of the system - bath coupling , it is found that the standard perturbation result without the polaron transformation is accurate for small coupling , while the polaron result is accurate in the opposite regime of strong coupling .
the variational method is capable of interpolating between these two limits .
it is valid over a much broader range of parameters and is only slightly less accurate around the region where the discontinuity appears .
as the relaxation time of the bath becomes longer leading to more non - markovian character , all three of the perturbation methods are seen to be less accurate .
the centre for quantum technologies is a research centre of excellence funded by the ministry of education and the national research foundation of singapore .
j. moix and j. cao acknowledge support from nsf ( grant no .
che-1112825 ) and darpa ( grant no . n99001 - 10 - 1 - 4063 ) .
j. cao is supported as part of the center for excitonics , an energy frontier research center funded by the us department of energy , office of science , office of basic energy sciences under award no .
de - sc0001088 .
c. k. lee is grateful to leong chuan kwek and jiangbin gong for reading the manuscript and providing useful comments .
for hamiltonians such as the spin - boson model where the bath is harmonic , the trace over the bath degrees of freedom in eq .
( [ exact_dm ] ) may be performed analytically . in the path integral formulation
this procedure leads to the well - known feynman - vernon influence functional.@xcite using a hubbard - stratonovich transformation,@xcite it was shown that the influence functional may be unraveled by an auxiliary stochastic field . the ensuing imaginary time evolution
may then be interpreted as one governed by a time - dependent hamiltonian .
explicitly , the spin - boson model may be equivalently expressed as @xmath97 all of the effects of the bath are accounted for by the colored noise term , @xmath98 , which obeys the autocorrelation relation , @xmath99 where @xmath100 is the correlation function given in eq .
( [ eq : correlation_functions ] ) with @xmath63 .
the trace over the bath that was present in the original path integral formulation now corresponds to averaging the imaginary time dynamics over realizations of the noise .
the auxiliary field is simply an efficient method of sampling the influence functional . in practice
, a sample of the reduced density matrix is propagated to the imaginary time @xmath45 , where the time steps , @xmath101 , are determined by @xmath102 \hat \rho_s(\tau ) \ ; , \label{eq : schrodinger}\ ] ] with the initial condition , @xmath103 .
the primary benefit of this approach is that it generates the entire reduced density matrix from a single monte carlo calculation .
additionally , any form for the spectral density of the bath , @xmath104 , may be used . in our calculations , @xmath105 ( at small @xmath10 ) to @xmath106 ( at large @xmath10 ) monte carlo samples
are needed to achieve convergence . | in the study of open quantum systems , the polaron transformation has recently attracted a renewed interest as it offers the possibility to explore the strong system - bath coupling regime . despite this interest , a clear and unambiguous analysis of the regimes of validity of
the polaron transformation is still lacking . here
we provide such a benchmark , comparing second order perturbation theory results in the original untransformed frame , the polaron frame and the variational extension with numerically exact path integral calculations of the equilibrium reduced density matrix .
equilibrium quantities allow a direct comparison of the three methods without invoking any further approximations as is usually required in deriving master equations .
it is found that the second order results in the original frame are accurate for weak system - bath coupling , the full polaron results are accurate in the opposite regime of strong coupling , and the variational method is capable of interpolating between these two extremes . as the bath becomes more non - markovian ( slow bath ) , all three approaches become less accurate . |
the notion of a projective quantum measurement was introduced by von neumann@xcite in his treatise on the mathematical foundations of quantum mechanics .
the idea is the following : ( i ) every quantum measurement yielding the data `` yes '' or `` no '' to an experimental question is described by the projection operators @xmath0 ( for `` yes '' ) or @xmath1 ( for `` no '' ) , @xmath2 ( ii ) if the initial state of a quantum object _ before _
a measurement is @xmath3 , and if the measurement yields the experimental answer `` yes '' to a question @xmath4 , then the state of the quantum object _
after _ the measurement @xmath5 is constructed according to the _ collapse _ of the quantum state rule @xmath6 eq.(2 ) is part of a calculation scheme which is _ not _ a unitary development of the quantum state _ during a measurement_. the unitary behavior ( implicit in the schrdinger equation for a quantum object ) was thought by von neumann to hold only _ between measurements_@xcite . to see how this works ,
consider a beam of photons which has passed through an `` upwards '' polarizer as in figs.1(a ) and ( b ) .
if a second polarizer has an axis at an angle of @xmath7 with respect to the upward direction , then one writes the photon wave function @xmath8 in the basis @xmath9 the probability to pass through the second polarizer is then @xmath10 the cases shown in fig.1 below work as follows : in fig.1(a ) , the second polarizer is set orthogonal to the first polarizer so that no photons will pass ; i.e. @xmath11 this probability can be increased , as in fig.1(b ) below , by placing an intermediate polarizer between the first and last orthogonal polarizers . in fig.1(b ) , a fraction @xmath12 of the @xmath8 photon beam will pass through the intermediate polarizer and _ collapse into the state @xmath13_. then , a fraction @xmath14 of the @xmath13 photon beam will pass through the final polarizer and again _ collapse into the state @xmath15_. the above two _ collapse of the quantum state processes _ dictate that one multiply probabilities .
the final probability that a photon in the upward polarized beam passes through both the intermediate and final polarizers is then given by @xmath16 that an intermediate measurement can _ increase _ the probability of passing through a polarizer stack , say @xmath17 , is an example of what is thought to be a quantum zeno effect @xcite a textbook quantum zeno effect problem in quantum mechanics@xcite is shown in fig.2 .
a beam of photons in state @xmath8 is incident on a stack of @xmath18 polarizers each of which has a polarization axis rotated by an angle of @xmath19 relative to the previous polarizer . according to the von neumann projection postulate@xcite , the quantum state of each photon will _ collapse _ @xmath18 times , yielding a probability @xmath20 of passing through the total stack as pictured in fig.2 .
note , in the formal limit of an _ infinite number of quantum state collapse events _ , @xmath21 theoretically _ every photon passes through the stack_. in real experiments , there is always some attenuation of a photon beam passing through many polarizers .
although the large @xmath18 quantum zeno effect does appear to be verified in the laboratory@xcite , we have _ a very uneasy feeling _ about attributing such observations to the rule of _ collapsing quantum states_. after all , long before von neumann considered projective measurements , maxwell had developed a reliable electromagnetic radiation theory which perfectly well determined how an electromagnetic wave will ( or will not ) pass through stacks of polarizers . _ and not even once _ did maxwell feel obliged to discuss the collapse of the electromagnetic radiation field functions . for that matter ,
neither bohr nor einstein@xcite had _ ever _ felt obliged to discuss the so - called wave function collapse .
our purpose is to discuss why the projection postulate ( collapse of the quantum state ) , _ certainly may and probably should _ be eliminated from ones thinking in quantum mechanics . the application of this view towards photon polarization measurements is typical of a more general state of affairs .
if one returns to the original copenhagen interpretation of the quantum mechanical formalism , then the quantum state describes an _ ensemble of experiments_. each photon certainly does _ not _ carry its own wave function which might then undergo a collapse whenever a photon disappears ( e.g. whenever a photon is absorbed in a polarizer ) . the wave function ( in reality ) _ describes an ensemble _ of many photons .
the collapse of the quantum state whenever a photon is absorbed closely resembles the collapse of a probability table ( at the automobile insurance company ) every time a car crashes .
actually , the probability table at the insurance company stays intact during a car crash .
it is the _ automobile _ and _ not _ the probability that collapses .
this constitutes an altogether _ different _ pile of scrap metal , even if the cross section for two automobiles to scatter off one another were computed with a quantum mechanical amplitudes to include the small diffraction effects . in sec.ii
, it will be shown how the schrdinger equation for photons in the vacuum is equivalent to the vacuum maxwell field equations . in sec.iii , the schrdinger equation for photons propagating in continuous media ( e.g. polarizers ) will be explored .
the photon wave function follows from the maxwell electromagnetic theory in continuous media . however , the electromagnetic wave attenuates if the continuous media exhibits dissipation in the form of finite conductivity .
the complete quantum system ( polarizers plus photons ) obeys _ a big monster _
schrdinger equation with a huge number of degrees of freedom and _ does not at all require the projective measurement _ formalism of eq.(2 ) .
this feature of quantum mechanics is valid for many systems as will be discussed in sec.iv . in sec.v
we compute , _ for a realistic model of polarizers _ , the maxwell equation formulation of the zeno effect .
the _ classical maxwell wave _ computation ( without projection operators ) yields the same result as _ the quantum photon schrdinger amplitude _
( without projection operators ) . in an appropriate regime
, the projection postulate does indeed give an accurate answer , but this may only be proved by solving the full maxwell theory without the projection postulate .
one might be led to ask why should one _ ever _ add to the quantum theory a projection theory of measurements . a hint as to why the projection postulate is so popular
might be connected to the fact that the maxwell theory computation ( to be presented ) required a fast computer .
more than mere speed , we required algorithms sufficient to maintain @xmath22 significant figure accuracy during the entire calculation of the photon transmission probability . without recently developed precise numerical algorithms , one might be coerced into employing the projection postulate . in the concluding sec.vi , we question the utility of replacing the wisdom of experimentalists and engineers who design and build laboratory equipment with a _ merely _ philosophical view of projective measurements .
it is somewhat unreasonable to represent real laboratory equipment with mathematical projection operators .
for a spin @xmath23 particle@xcite one requires three hermitian matrices @xmath24 obeying @xmath25=i\epsilon_{ijk}s_k , \ \
s_1 ^ 2+s_2 ^ 2+s_3 ^ 2=s(s+1).\ ] ] the photon is a spin one ( @xmath26 ) particle so we may choose the spin matrices @xmath27 and @xmath28 with @xmath29 , the vacuum schrdinger equation for free photons reads @xmath30 writing @xmath31 allows one to cast the photon schrdinger eq.(12 ) into the ( perhaps ) more simple vector form @xmath32 the vector field @xmath33 should be constrained to obey @xmath34 eq.(15 ) is equivalent to only allowing two independent spin states ( say superpositions of only two transverse polarizations ) , thereby _ eliminating _ a third ( longitudinal polarized ) `` zero energy '' mode .
finally , eqs.(14 ) and ( 15 ) take on a clearly recognizable form if one defines the real and imaginary parts of @xmath33 to be , respectively , the electric field @xmath35 and
the magnetic field @xmath36@xcite@xcite@xcite ; i.e. @xmath37 the photon schrdinger eq.(12 ) then appears equivalent to the vacuum maxwell theory ; i.e. @xmath38 and @xmath39 with the vacuum gauss law eq.(15 ) constraints @xmath40 it is of interest to discuss the relativistic lorentz symmetry of the photon schrdinger eq.(12 ) .
the proof of lorentz symmetry is actually well known since we have already proved equivalence with the maxwell theory .
however , the discrete lorentz symmetries are still worthy of further discussion@xcite .
if we consider the classical lorentz force on a charge equation @xmath41 , and then allow the charge to change sign @xmath42 , then the force @xmath43 will remain the same if we also allow @xmath44 and @xmath45 .
thus , the electromagnetic field eq.(16 ) is odd under charge conjugation , @xmath46 under time reversal , @xmath47 and @xmath45 so that @xmath48 finally , under parity @xmath44 and @xmath49 so that @xmath50 note @xmath51 as would be expected from relativistic wave equations .
the maxwell theory for an electromagnetic disturbance moving through continuous media@xcite may be formulated starting from @xmath52 and @xmath53 @xmath54 it is sufficiently general to supplement eqs.(24 ) , ( 25 ) and ( 26 ) by employing a linear causal relationship between the current and the electric field , i.e. @xmath55 if , for a complex frequency @xmath56 where @xmath57 ,
@xmath58 @xmath59 eqs.(27 ) , ( 28 ) and ( 29 ) serve to define the non - local conductivity @xmath60 , i.e. @xmath61 where @xmath62 some workers prefer to complete the maxwell equations in continuous media by employing a constitutive equation relating the maxwell displacement field @xmath63 to the electric field @xmath35 . in the frequency domain ,
the dielectric response @xmath64 of the continuous media is defined as @xmath65 just so long as the various response functions are maintained as non - local ( in space and time ) , which of the many equivalent response functions are actually used in a calculation remains a matter of convenience .
for example , eqs.(30 ) and ( 32 ) are equivalent if we choose @xmath66 the dielectric , magnetic , or conductivity response functions conventionally used to describe continuous media now make an appearance into the schrdinger equation for a photon moving through the continuous media via the photon `` self - energy '' insertion . to see what is involved ,
we may write the maxwell eqs.(24 ) and ( 25 ) as @xmath67 and @xmath68 where eq.(16 ) has been invoked .
employing eqs.(21 ) , ( 27 ) and ( 35 ) , we find our central result for the photon schrdinger equation in continuous media @xmath69 @xmath70 where the photon self - energy part is defined as @xmath71 using eqs.(13 ) and ( 36 ) , one may write the photon schrdinger equation in the form @xmath72 @xmath73 in principle the calculation of the self energy part @xmath74 is equivalent to the calculation of the conductivity @xmath60 as is evident from eqs.(21 ) , ( 31 ) , and ( 37 ) . in a practical calculations involving the photon schrdinger equation in a polarizer
, one should use the conductivity and dielectric response functions as those of the optical engineers who designed the device@xcite .
finally , the non - local form of eq.(38 ) is shared by many other quantum mechanical systems as we shall now briefly explore .
although we do not adhere to the non - unitary collapse of the quantum state eq.(2 ) as postulated by von neumann , we do find the notion of projection operators ( also introduced into quantum mechanics by von neumann ) to be more than just convenient .
eq.(1 ) will thereby be invoked in what follows . for a macroscopic laboratory quantum system
we here insist that a _ monster _ schrdinger equation with a huge number of degrees of freedom still holds true , @xmath75 we do not contemplate the _ ultimate monster wave function _ of the _ whole universe _
, not only because we are not smart enough to find it , but because we do not know how to build an experimental _ ensemble of universes _ to test a hypothetical quantum state of _ our universe_. we leave such considerations to people more wise than are we .
we are happy to say something meaningful about a very small physical piece of a laboratory quantum state ; e.g. a `` projected piece '' @xmath76 we leave the rest of the monster wave function , @xmath77 to the engineers who design the experiment . with the little piece of the wave function eq.(40 ) and the big piece of the wave function eq.(41 ) , we rewrite eq.(39 ) as @xmath78 where @xmath79 , @xmath80 , @xmath81 and @xmath82 .
we can only hope to find @xmath83 which obeys @xmath84 so we solve for the monster piece @xmath85 subject to the causal boundary condition @xmath86 from eqs.(43 ) and ( 45 ) we find the general non - local schrdinger equation for the piece of the wave function that is of interest ; it is @xmath87 where @xmath88 the full monster schrdinger eq.(39 ) for complete wave function @xmath89 reduces to the non - local in time schrdinger eq.(46 ) for the small subspace piece of the wave function @xmath90 .
a particular example of eq.(46 ) is the photon schrdinger eq.(38 ) .
the point is that when a photon propagates inside of a continuous media device , say a polarizer , the _ monster wave function _
includes _ both _ the photon _ and _ the polarizer degrees of freedom . in a one photon projected subspace ,
the quantum state of the polarizer is fixed .
in addition to the polarizer , there is but one propagating photon .
if the photon is absorbed , then an electronic excitation is produced which heats up the polarizer .
the piece of the monster quantum state representing this absorbed photon situation is not contained in the single photon subspace .
the single photon subspace wave function is then attenuated due to photon absorption in the polarizer .
all of this ( including the attenuation of the maxwell wave in the medium ) is described by the one photon schrdinger equation , _ without _ recourse to the projection postulate .
maxwell was so lucky to be able to write down the schrdinger equation for the photon in matter , including attenuation , _ without _ knowing anything about quantum states ( be they monster or otherwise ) .
it is of interest to look for a fixed energy solution of the subspace wave function eq.(46 ) of the form @xmath91 eqs.(46 ) , ( 47 ) and ( 48 ) imply @xmath92 where @xmath93 that the effective energy dependent hamiltonian on the left hand side of eq.(49 ) @xmath94 is not hermitian , is due to the `` fermi platinum rule '' transition rates to leave the small subspace and wander into the remaining monster subspaces .
the fermi platinum rule @xmath95 is just a little bit more valuable than the more usual perturbative fermi golden rule , since eq.(52 ) is rigorously true to all orders of perturbation theory .
the fixed energy @xmath96 case corresponds to having a fixed photon frequency @xmath97 for the electromagnetic problem at hand .
let us return to this problem .
the maxwell theory of a transverse wave moving through a polarizer may be formulated following the discussion of born and wolf@xcite .
( i ) if a transverse wave propagates in the @xmath98-direction , then the wave may be described by four electromagnetic field components , say @xmath99 if the polarizer is oriented by a rotation angle @xmath7 , as in fig.1(b ) , then the dielectric response tensor will have the form @xmath100 where we have employed the @xmath101 pauli matrices @xmath102 and @xmath103 the complex principal eigenvalues of the dielectric tensor at frequency @xmath104 have the form @xmath105 the real parts of the dielectric eigenvalues @xmath106 determine the velocities of the waves polarized in the principal directions . the conductivities @xmath107 determine the dissipative attenuation of the waves polarized in the principle directions .
for a well designed polarizer , a wave with an allowed polarization direction will propagate unattenuated while a wave with a disallowed polarization will be strongly attenuated .
_ it is the anisotropic dissipative conductivity that is often modeled by the quantum projective postulate for the case of photon polarization_. on the other hand , conductivity is evidently _ classical _ since both ohm and maxwell were dead and gone _ before _ quantum mechanics had been invented . for the classical maxwell calculation ,
the wave at the beginning @xmath108 of the polarizer is related to the wave at the end @xmath109 of the polarizer by a @xmath110 transfer matrix @xmath111 @xmath112 is defined in eq.(53 ) , and in partitioned matrix form @xmath113 @xmath114 for a stack of @xmath18 polarizers , each of width @xmath115 and oriented at angles @xmath116 , one may repeat the above process going from the beginning @xmath108 of the stack to the end @xmath117 of the stack @xmath118 one multiplies the transfer matrices for each polarizer in the stack @xmath119 once the total @xmath110 transfer matrix is computed , one may find the transmission amplitudes @xmath120 and reflection amplitudes @xmath121 for the two polarization directions @xmath122 by solving the linear algebra problem @xmath123 for an incident photon in the allowed @xmath124-polarization direction , the probability that the photon exits a stack of @xmath18 polarizers ( rotated into the allowed @xmath125-polarization direction ) is given by the absolute square of the transmission amplitude , @xmath126 from eqs.(54 ) , ( 59 ) , and ( 61)-(63 ) , with each of the angles in @xmath116 set equal to @xmath127 , one may compare the the projection postulate @xmath128 in eq.(7 ) to the maxwell theory , or equivalently the photon schrdinger eq.(63 ) , without the projection postulate .
the matrix multiplications in eq.(61 ) and the linear algebra in eq.(62 ) are best left to a computer .
the results of such a computation are shown in fig.3 .
the solid curve in fig.3 is the prediction of the projection postulate eq.(7 ) . for the maxwell theory with a reasonable stack of laboratory polarizers
, there may be @xmath129 wavelengths of light in each polarizer in the stack .
thus we chose the parameter @xmath130 for purpose of illustration .
we further choose @xmath131 to describe the polarization direction for which the polarizer would be transparent and the complex @xmath132 to describe the polarization direction for which the polarizer would be strongly attenuated ( if not absolutely opaque ) .
the resulting computed points are the filled circles of fig.3 . for this case ,
the projection postulate yields a @xmath128 which is quite close to the full maxwell theory . for some regimes ( perhaps more close to the laboratory situation ) , the maxwell theory begins to diverge from from the projection postulate for large @xmath18 . in the laboratory
all polarization directions are subject to small ( but certainly finite ) attenuation .
to test this situation , we computed the maxwell theory prediction when the allowed polarization has a weak attenuation response @xmath133 while the disallowed polarization has a stronger attenuation response @xmath134 .
the results are the filled squares of fig.3 . for @xmath135
, there is a zeno effect in which @xmath128 increases .
however for @xmath136 , the is an attenuation induced decrease in @xmath128 , contrary to the `` zeno effect expectation '' .
however , one might regard this attenuation of the allowed polarization direction to be merely an indication of poor experimental design .
some final comments about the calculations producing fig.3 are worthy of note . the transfer matrix formalism used to calculate the transmission probability @xmath128 was tedious , but straight forward .
however , the numerical computations required to produce fig.3 were not entirely trivial .
the first three numerical commercial software programs that we employed for the calculation produced complete nonsense@xcite .
the physical problem was that the maxwell waves reflect back and forth within the polarizers .
hence , there exist _ exponentially attenuated _ terms as well as _ exponentially amplified _ terms in the transfer matrix .
high accuracy was needed for the matrix multiplications in eq.(61 ) . only with high accuracy at intermediate stages of the calculation , could eqs.(62 ) and ( 63 ) be reliably computed . for realistic laboratory polarizers ,
the required accuracy turns out to be very high indeed .
fortunately , aberth@xcite has developed precise numerical algorithms together with publicly available programming code which allows for @xmath137 significant figure accurate computations .
much of the required code for the standard functions and standard linear algebra was thus available .
we used this code .
however , we required @xmath22 significant figure computations to produce the reliable plots shown in fig.3 .
in discussions of the foundations of quantum mechanics there has been a tendency to regard all interference of waves as explicitly due to quantum mechanics .
this is only sometimes true , since `` waves '' as well as `` particles '' also have classical limits . in many ( perhaps most ) of the electromagnetic experiments and applications ,
the maxwell wave itself is classical .
the _ distinction _ between a classical wave , a classical particle , and a quantum particle has been discussed simply , _ yet clearly _ , by feynman@xcite .
\(i ) a quantum particle is _ certainly not _ a classical particle .
quantum particle dynamics dictates amplitude superposition , whereas classical dynamics dictates probability superposition .
( ii ) a quantum particle is _ certainly not _ a classical wave .
a classical wave is `` smooth '' while a quantum particle comes in `` lumps '' .
photons are lumpy . photon counters yield one click whenever a photon is detected .
_ one _ photon can never make _ two _ different counters each give you _
one half _ of a click .
that is much too smooth for quantum mechanics . in quantum mechanics ,
each photon counter gives one click or no clicks .
the photon is _ certainly not _ a classical wave .
the photon is _ certainly not _ a classical particle .
the photon is _ certainly not _ a little bit classical wave and a little bit classical particle .
( iii ) the photon is _ neither _ a classical wave _ nor _ a classical particle _ nor _ a little bit of both . electromagnetic waves ,
as described by maxwell , are classical waves .
if a radio station broadcasts a signal , then the signal is ( for all practical purposes ) smooth .
your radio gets some of the signal , and our radio gets some of the signal .
there is enough smoothly distributed classical wave to go around to all of the radios .
the classical maxwell wave is smooth .
one detector can not get a fraction of a photon .
remember , photons come in lumps .
one photon can get destroyed in at most one radio receiver .
one detector can get any fraction of a classical maxwell wave .
remember , classical waves are smooth . yet
the schrdinger equation for photon is the same ( in a mathematical formalism ) as the maxwell equations for classical electromagnetic waves .
this is because the photon has no mass and because the photon is a boson .
when very many photons are bose condensed into a single quantum state , the result is a maxwell wave which is classical for all practical purposes .
the above picture was not all that clear to einstein ( by his own admission ) , and we do not claim to know more about photons than did einstein . according to einstein , a maxwell wave hits a half silvered mirror and some of the wave goes to one detector and the rest of the wave goes to a second detector .
if the wave represents one photon , then only one counter goes click .
the other counter gets whacked with a maxwell wave .
yet this counter counts nothing .
when there is but one photon , only one counter can speak for her .
some maxwell wave parts turn out to have have a photon , and some maxwell wave parts turn out not have a photon .
some counters go click and some counters remain mute . you never know . and photons are lumpy .
it is a very strange shell game according to einstein@xcite .
however , the zeno effect formally attributed to the quantum mechanical projection postulate has here been argued to be merely a classical wave effect .
maxwell would not have had to increase his knowledge of electromagnetic theory by even one iota in order to totally understand the polarization version of the zeno effect .
no quantum mechanics is required , not to even speak of the projection postulate .
99 j. von neumann , _ mathematical foundations of quantum mechanics _ , princeton university press , princeton ( 1996 ) . r.l .
stratonovich and v.p .
belavkin , _ international journal of theoretical physics _ , _ 35 _ , 2215 - 28 ( 1996 ) . g.s .
agarwal and s.p .
lett _ , * 185 * , 139 - 142 ( 1994 ) m.a.b .
whitaker , _ interaction - free measurement and the quantum zeno effect _
, _ physics lett a _ , * 244 * , 502 - 6 , ( 1998 ) .
o. alter and y. yamamoto , _ phys .
rev . a _ * 55 * , r2499 - 2502 , ( 1997 ) .
a. beige and g.c .
hegerfeldt , _ j. of phys .
* 30 * , 1323 - 34 , ( 1997 ) .
a. beige , g.c .
hegerfeldt , and d.g .
sondermann , _ foundations of physics _ , *
27 * , 1671 - 88 ( 1997 ) .
s. pascazio , h. nakazato , and m. mamiki , _ j of the phys .
soc of japan _ , * 65 * , 285 - 8 , ( 1996 ) .
j. pan , y.zhang , and q. dong , _ phys lett a _ , * 238 * , 405 - 7 ( 1998 ) . l.s .
schulman , _ phys rev a _ , * 57 * , 1509 - 15 ( 1998 ) .
s. pascazio , _ foundations of physics _ , * 27 * , 1655 - 70 ( 1997 ) .
b. nagels , l.j.f .
hermans , and p.l .
chapovsky , _ phys rev lett _ , * 79 * , 3097 - 100 ( 1997 ) . h. nakazato , m. namiki , s. pascazio , and h. rauch , _ phys lett a _ , * 199 * , 27 - 32 ( 1995 ) . j. s. townsend , _ a modern approach to quantum mechanics _ ,
p.62 , mcgraw - hill , new york ( 1992 ) .
stancil and g.e .
copeland , _ j. of phys .
b _ , * 27 * , 2801 - 7 ( 1994 ) .
p. kwiat , h. weinfurter , and a. zeilinger , _ quantum seeing in the dark , scientific american _ 72 , nov .
( 1996 ) ; p. kwiat , h. weinfurter , t. herzog , a. zeilinger and m.a .
kasevich , _ phys .
lett . _ * 74 * , 4763 ( 1995 ) .
n. bohr , _ discussions with einstein on epistemological problems in atomic physics _ in _ albert einstein philosopher scientist _ , edited by p. a. schilpp , open court , la salle ( 1997 ) .
m. ban , _ phys rev a _ , * 49 * , 5078 - 85 ( 1994 ) .
gurvitz , _ phys .
b _ , * 56 * , 15215 - 23 ( 1997 ) .
y. japha , a.g .
kofman , and g. kurizki , _ acta physica polonica a _ , * 93 * , 135 - 45 ( 1998 ) .
s. machida and a. motoyoshi , _ foundations of physics _ , * 28 * , 45 - 57 ( 1998 ) .
m. bordag , k. kirsten , and d.v .
vassilevich , _ j of phys a _ , * 31 * , 2381 - 9 ( 1998 ) .
f. mihokova , s. pascazio , and l.s .
schulman , _ phys rev a _ , * 56 * , 25 - 32 ( 1997 ) .
n.l . manakov and v.d .
ovsiannikov , _ j of phys b _ , * 30 * , 2109 - 17(1997 ) . m. ya .
agre and l.p .
rapoport , _ optics and spectroscopy _ , * 75 * , 622 - 3n ( 1993 ) .
evans , _ foundations of phys .
lett , _ , * 7 * , 67 - 74 ( 1994 ) . s.l . braunstein and j.a .
smolin , _ phys .
a _ * 55 * , 945 - 50 , ( 1997 ) .
gillespie , _ phys .
a _ , * 49 * ( 1994 ) .
guo , h. ebert , w.m .
temmerman , and p.j .
durham , _ phys .
b _ , * 50 * , 3861 - 8 , ( 1994 ) .
s.p . vyatchanin and e.a .
zubov , _ optics and spectroscopy _ , * 74 * , 544 - 6 , ( 1993 ) .
v. sochor , _ proceedings of the spie - the international society for optical engineering _ , * 2799 * , 185 - 7 ( 1996 ) .
horvath , r.c .
thompson , and p.l .
knight , _ contemporary physics _ ,
* 38 * , 25 - 48 ( 1997 ) .
m. namiki , s. pascazio and c. schiller , _ phys lett a _ , * 187 * , 17 - 25 , ( 1994 ) .
manakov , a.v .
meremianin , a.f .
starace , _ phys rev a _ , * 57 * , 3233 - 44 ( 1998 ) .
schulman , _ foundations of physics _ , * 27 * , 1623 - 36 , ( 1997 ) .
m. kitano , _ optics communications _ , * 141 * , 39 - 42 ( 1997 ) .
d. home and m.a.b .
whitaker , _ annals of phys _ , * 258 * , 217 - 85 ( 1997 ) .
gagen , h.m .
wiseman , and g.j .
milburn , _ phys rev a _ , * 48 * , 132 - 42 ( 1993 ) .
phipps jr . , _ physics essays _ , * 11 * , 155 - 63 , ( 1998 ) .
m. born and e. wolf , _ priciples of optics _ , chapt.xiv , pergamon press , oxford ( 1980 ) .
o. aberth , _ precise numerical methods using c++ _ , academic press , san diego ( 1998 ) .
r. p. feynman , r.b .
leighton , and m. sands , _ the feynman lectures on physics _ , * iii * , chapt.1 , addison wesley , reading ( 1965 ) .
a. fine , _ the shaky game ; einstein realism and the quantum theory _ , the university of chicago press , chicago ( 1996 ) . | we consider a photon beam incident on a stack of polarizers as an example of a von neumann projective measurement , theoretically leading to the quantum zeno effect .
the maxwell theory ( which is equivalent to the single photon schrdinger equation ) describes measured polarization phenomena , but without recourse to the notion of a projective measurement . |
suppose that @xmath3 and @xmath4 are ( continuous ) distribution functions of two default times .
it is well known from coupling theory ( see @xcite ) that there exists a maximal coupling , i.e. a two - dimensional distribution function @xmath8 with marginals @xmath3 and @xmath4 such that for the case of @xmath9 the probability of a joint default @xmath10 is maximal ( within the class of all two - dimensional distribution functions having @xmath3 and @xmath4 as marginals ) .
translating to the class of copulas ( see @xcite and section [ mksklar ] ) , maximizing the probability of a joint default means calculating @xmath11 for @xmath5 \rightarrow [ 0,1]$ ] being defined by @xmath12 , @xmath13 denoting the quasi - inverse of @xmath3 , @xmath14 the graph of @xmath7 , @xmath15 the family of all two - dimensional copulas and @xmath16 being the doubly stochastic measure corresponding to the copula @xmath17 . as pointed out in @xcite
there is a ( not necessarily unique ) copula @xmath18 with @xmath19 that can even be computed in closed form .
considering @xmath20 and setting @xmath21 , the pair @xmath22 has marginal distribution functions @xmath3 and @xmath4 and maximizes the joint default probability . in the current paper
we tackle the closely related problem of maximizing @xmath23 , the probability that @xmath0 does not default before @xmath1 , and solve it in a definitive manner .
we first translate the maximization problem to the copula setting and prove the existence of a ( mutually ) completely dependent copula @xmath24 with @xmath25 where @xmath26 and @xmath27 ^ 2 : y \leq t(x)\}$ ] denotes the endograph of @xmath7 .
afterwards we study the situation of not necessarily monotonic @xmath5 \rightarrow [ 0,1]$ ] and prove a simple and easily calculable formula for @xmath28 only involving the distribution function of @xmath7 . as in the monotonic case it is possible to construct a completely dependent copula maximizing the mass of @xmath6 .
finally , using the just mentioned results we derive an equally simple formula for @xmath29 and show that there are situations where the infimum is not attained .
the rest of the paper is organized as follows : section 2 gathers some preliminaries and notations . in section 3
we prove the markov kernel version of sklar s theorem and then apply it to show that @xmath30 holds , where @xmath31 is an arbitrary measurable transformation , @xmath22 has marginals @xmath32 and copula @xmath33 , and @xmath7 is defined by @xmath34 .
all aforementioned maximization results are collected in section 4 .
section 5 presents an alternative proof of the main result and derives some useful consequences .
for every @xmath35-dimensional random vector @xmath36 on a probability space @xmath37 we will write @xmath38 if @xmath36 has distribution function ( d.f . , for short ) @xmath3 and let @xmath39 denote the corresponding distribution on the borel @xmath40-field @xmath41 of @xmath42 . for every univariate distribution function @xmath3 we will let @xmath13 denote the quasi - inverse of @xmath3 , i.e. @xmath43 . note that for every @xmath44 we have @xmath45 if and only if @xmath46 , that for @xmath47 and @xmath3 continuous we have @xmath48 and that the random variable @xmath49 coincides with @xmath0 with probability one .
for further properties of @xmath13 we refer , for instance , to @xcite
. given univariate distribution functions @xmath3 and @xmath4 , we will let @xmath50 denote the frchet class of @xmath3 and @xmath4 , i.e. the family of all two - dimensional distribution functions having @xmath3 and @xmath4 as marginals ; @xmath51 will denote the corresponding class of probability measures on @xmath52 .
@xmath53)$ ] and @xmath53 ^ 2)$ ] denote the borel @xmath40-fields on @xmath54 $ ] and @xmath54 ^ 2 $ ] , @xmath55 and @xmath56 the lebesgue measure on @xmath53)$ ] and @xmath53 ^ 2)$ ] respectively . for every measurable transformation @xmath5 \rightarrow [ 0,1]$ ] the push - forward of @xmath55 via @xmath7 will be denoted by @xmath57 . as already mentioned before
, @xmath15 will denote the family of all two - dimensional _
copulas_. for background on copulas we refer to @xcite .
@xmath58 and @xmath59 will denote upper and the lower frchet - hoeffding bound , @xmath60 the product copula .
@xmath61 will denote the uniform distance on @xmath15 ; it is well known that @xmath62 is a compact metric space and that @xmath61 is metrization of weak convergence in @xmath15 . for
every @xmath17 @xmath16 will denote the corresponding _ doubly stochastic measure _ defined by @xmath63\times [ 0,y])=a(x , y)$ ] for all @xmath64 $ ] , @xmath65 the class of all these doubly stochastic measures . a _ markov kernel _ from @xmath66 to @xmath67 is a mapping @xmath68 $ ] such that @xmath69 is measurable for every fixed @xmath70 and @xmath71 is a probability measure for every fixed @xmath72 .
given real - valued random variables @xmath2 on @xmath73 , a markov kernel @xmath74 $ ] is called a _ regular conditional distribution of @xmath1 given @xmath0 _ if for every @xmath70 @xmath75 holds @xmath76-a.s .
it is well known that for each pair @xmath22 of real - valued random variables a regular conditional distribution @xmath77 of @xmath1 given @xmath0 exists , that @xmath77 is unique @xmath78-a.s .
( i.e. unique for @xmath78-almost every @xmath72 ) and that @xmath77 only depends on the distribution @xmath79 . hence , given @xmath9 , we will denote ( a version of ) the regular conditional distribution of @xmath1 given @xmath0 by @xmath80 and refer to @xmath80 simply as _
markov kernel of @xmath8 _ or _ markov kernel of @xmath22_. note that for every two - dimensional distribution function @xmath8 , its markov kernel @xmath80 , and every borel set @xmath81 the following _ disintegration _ formula holds ( @xmath82 denoting the @xmath83-section of @xmath4 for every @xmath72 ) @xmath84 for @xmath17 we will directly consider the corresponding markov kernel @xmath85 to be defined on @xmath54 \times \mathcal{b}([0,1])$ ] . considering that
in this case eq . ( [ disallg ] ) implies that @xmath86 } k_a(x , f)\ , d\lambda(x ) = \lambda(f)\ ] ] holds for every @xmath87)$ ] , and that , additionally , every markov kernel @xmath88\times \mathcal{b}([0,1 ] ) \rightarrow [ 0,1]$ ] fulfilling eq .
( [ bound2 ] ) obviously induces a unique element @xmath89 , it follows that there is a one - to - one correspondence between @xmath15 and the family of all markov kernels @xmath88 \times \mathcal{b}([0,1 ] ) \rightarrow [ 0,1]$ ] fulfilling eq .
( [ bound2 ] ) .
notice that for @xmath90 eq .
( [ bound2 ] ) also implies that @xmath91 holds for @xmath55-almost every @xmath92 $ ] , so it is always possible to choose a ( version of the ) kernel fulfilling @xmath91 for every @xmath92 $ ] . for more details and properties of conditional expectation , regular conditional distributions , and disintegration
see @xcite and @xcite , various results underlining the usefulness of the markov kernel perspective can be found in @xcite and the references therein
. in the sequel @xmath93 will denote the class of all @xmath55-preserving transformations @xmath94 \rightarrow [ 0,1]$ ] , @xmath95 the subset of all bijective @xmath96 , and @xmath97 the subset of all piecewise linear , bijective @xmath96 .
a copula @xmath98 will be called _
completely dependent _ if and only if there exists @xmath96 such that @xmath99 is a regular conditional distribution of @xmath33 ( see @xcite for equivalent definitions and main properties ) .
for every @xmath96 the induced completely dependent copula will be denoted by @xmath100 throughout the rest of the paper , @xmath101 will denote the family of all completely dependent copulas . following @xcite , for every @xmath96 and every copula @xmath90 we will let @xmath102 denote the ( generalized ) _ @xmath103-shuffle of @xmath33 _ , defined implicitly via the corresponding doubly stochastic measures by @xmath104 for all @xmath105)$ ] .
notice that @xmath106 is a shuffle in the sense of @xcite if @xmath107 , and that it is a shuffle in the sense of @xcite ( to which we will refer as classical shuffle in the sequel ) if @xmath108 .
suppose now that the vector @xmath109 has distribution function @xmath110 with @xmath32 continuous . according to sklar s theorem ( see @xcite ) there exists a unique copula @xmath17 such that @xmath111 holds for all @xmath112 .
translating this to the markov kernel setting we get the following result describing how to construct a kernel of @xmath8 given the kernel of @xmath33 : [ katokh ] suppose that @xmath32 are continuous distribution functions , that @xmath22 has d.f .
@xmath113 and copula @xmath33 , and let @xmath85 denote a markov kernel of @xmath33 fulfilling @xmath91 for all @xmath92 $ ] . then setting @xmath114\big ) : = k_a\big(f(x),[0,g(y ) ] \big)\ ] ] for all @xmath112 defines a markov kernel @xmath77 of @xmath115 . *
proof : * * ( i ) * we need to show that @xmath116)$ ] is a distribution function for every fixed @xmath72 : the fact that @xmath116)$ ] is non - decreasing is trivial . if @xmath117 is monotonically decreasing with limit @xmath118 then , using continuity of @xmath4 and the fact that @xmath85 is a markov kernel , we have @xmath119=[0,g(y)]$ ] as well as @xmath120)&=&\lim_{n \rightarrow \infty } k_a(f(x),[0,g(y_n ) ] ) = k_a(f(x),[0,g(y ) ] ) = k(x,(-\infty , y])\end{aligned}\ ] ] since both @xmath121)=0 $ ] and @xmath122)=1 $ ] follow in the same manner , @xmath116)$ ] is a distribution function and we can extend @xmath123 from the generator @xmath124 : y \in \mathbb{r}\}$ ] to a probability measure on @xmath67 in the standard way ( @xcite ) . + * ( ii ) * measurability of @xmath125)$ ] for every fixed @xmath118 is a direct consequence of measurability of @xmath126 and the fact that @xmath85 is a markov kernel . considering that @xmath127 is a dynkin system , that @xmath128 is closed w.r.t
. intersection , and that @xmath129 , it follows that @xmath130 ( see @xcite ) , implying that @xmath77 is indeed a markov kernel . + * ( iii ) * it remains to show that @xmath77 is a markov kernel of @xmath22 .
let @xmath44 be arbitrary but fixed . then setting @xmath131 } k(t,(-\infty , y ] )
\,d \mathbb{p}^x(t)$ ] and using disintegration , continuity of @xmath3 , change of coordinates and sklar s theorem we get @xmath132\big ) \,d \mathbb{p}^x(t ) \\ & = & \int_{\mathbb{r } } \mathbf{1}_{(-\infty , q)}(f(t ) ) k_a\big(f(t),[0,g(y)]\big ) \,d \mathbb{p}^x(t ) \\ & = & \int_{[0,1 ] } \mathbf{1}_{(-\infty , q)}(z ) k_a\big(z,[0,g(y)]\big ) \,d \mathbb{p}^{f \circ x}(z ) \\ & = & \int_{[0,1 ] } \mathbf{1}_{[0,q)}(z ) k_a\big(z,[0,g(y)]\big ) \,d \lambda(z ) = \int_{[0,q ] } k_a\big(z,[0,g(y)]\big ) \,d \lambda(z ) \\ &
= & a(q , g(y ) ) = a(f \circ f^-(q),g(y ) ) = h(f^-(q),y).\end{aligned}\ ] ] this shows that we have @xmath133 } k(t,(-\infty , y ] ) \,d \mathbb{p}^x(t)\ ] ] for all @xmath118 and all @xmath134 of the form @xmath135 for some @xmath44 . in case of @xmath136 and
@xmath137 we can use completely the same arguments to show that @xmath138 } k(t,(-\infty , y ] ) \,d \mathbb{p}^x(t)= h(f^-(1),y)=g(y)\ ] ] holds for every @xmath118 .
the extension to full @xmath139 is now straightforward : let @xmath112 be arbitrary , set @xmath140 and @xmath141 . if @xmath44 then @xmath142 as well as @xmath143)=0 $ ] follow and we get @xmath144 } k(t,(-\infty , y ] ) \,d \mathbb{p}^x(t ) \\ & = & \int_{(-\infty , x ] } k(t,(-\infty , y ] ) \,d \mathbb{p}^x(t).\end{aligned}\ ] ] in case of @xmath136 we have @xmath145 , so using eq .
( [ marginal ] ) and @xmath146 we get @xmath147 } k(t,(-\infty , y ] ) \,d \mathbb{p}^x(t ) = \int_{(-\infty , x ] } k(t,(-\infty , y ] ) \,d \mathbb{p}^x(t),\end{aligned}\ ] ] and in case of @xmath148 it follows that @xmath149 } k(t,(-\infty , y ] ) \,d \mathbb{p}^x(t).\end{aligned}\ ] ] altogether we have shown that @xmath150 } k(t,(-\infty , y ] ) \,d \mathbb{p}^x(t)$ ] holds for all @xmath112 , so , extending in the standard way from @xmath151 to @xmath52 ( see @xcite ) we get that @xmath77 is a markov kernel of @xmath22 .
@xmath152 + proceeding analogously to the proof of lemma [ katokh ] we can show the following result , describing how to construct a kernel @xmath85 of the copula @xmath33 if the kernel @xmath80 of @xmath22 is known : [ khtoka ] suppose that @xmath32 are continuous distribution functions , that @xmath22 has d.f . @xmath113 and copula @xmath33 , and let @xmath80 denote a markov kernel of @xmath8 with @xmath153 for every @xmath72 . then setting @xmath154 for all @xmath155 and @xmath156 $ ] defines a markov kernel @xmath77 of @xmath17 .
suppose now that @xmath157 is an arbitrary borel - measurable mapping .
in the sequel we will let @xmath158 and @xmath159 denote the graph and the endograph of @xmath160 respectively , i.e. @xmath161 lemma [ katokh ] allows to express @xmath162 as well as @xmath163 in terms of @xmath32 and the underlying copula @xmath33 . in order to prove a more general result and to simplify notation , given ( continuous ) @xmath32 and ( measurable ) @xmath160 we will write @xmath164 in the sequel . in general @xmath7
is only well - defined on @xmath165 - we will however , directly consider it as function on @xmath54 $ ] by setting @xmath166 and @xmath167 .
[ translate ] suppose that @xmath2 are random variables on @xmath37 with joint distribution function @xmath8 , continuous marginals @xmath3 and @xmath4 and copula @xmath33 .
furthermore let @xmath157 be an arbitrary borel - measurable mapping and define @xmath7 according to eq .
( [ st ] ) .
then the following identities hold : @xmath168 * proof : * using the fact that @xmath169 , disintegration and lemma [ katokh ] the second identity can be proved as follows : @xmath170
\big ) \,d\mathbb{p}(\omega ) \\ & = & \int_\omega k_a \big(f \circ x(\omega),[0,g \circ s \circ f^- \circ f \circ x ( \omega ) ] \big ) \,d\mathbb{p}(\omega ) \\ & = & \int_\mathbb{r } k_a \big(z,[0,g \circ s \circ f^- ( z ) ] \big ) \,d\mathbb{p}^{f \circ x}(z ) \\ & = & \int_{[0,1 ] } k_a \big(z,[0,t(z ) ] \big ) \,d\lambda(z ) = \mu_a(\gamma^{\leq}(t)).\end{aligned}\ ] ] working with @xmath171 instead of @xmath172 \big)$ ] the first identity @xmath173 follows in the same manner .
for the special case of @xmath174 calculating @xmath175 corresponds to finding ( joint ) distributions of @xmath22 for which @xmath23 is as big as possible - interpreting @xmath2 as lifetimes or default times of financial institutions this translates to maximizing the probability that @xmath0 does not die or default before @xmath1 .
notice that , setting @xmath176 and considering the pair @xmath177 , this maximization problem can be considered a special case of the much more general situation studied in @xcite .
theorem [ translate ] implies that the problem can be reduced to the family of copulas , i.e. we have @xmath178 obviously the same is true when considering minimal probabilities , i.e. @xmath179 holds .
taking into account that in case of @xmath174 the mapping @xmath34 according to theorem [ translate ] is non - decreasing , it is actually possible to calculate @xmath180 and even construct a dependence structure for which @xmath23 coincides with @xmath180 .
the following result holds : [ solutionmax ] suppose that @xmath5 \rightarrow [ 0,1]$ ] is non - decreasing
. then we have @xmath181 } ( t(x)-x).\ ] ] additionally , setting @xmath182 and letting @xmath183 denote the rotation @xmath184 , we have @xmath185 . * proof : * considering @xmath186 \times [ 0,t(x ) ] \ , \cup \ , [ x,1 ] \times [ 0,1]$ ] it follows that @xmath187 holds for every @xmath92 $ ] and every @xmath17 , which implies that the left - hand side of ( [ maxendo ] ) is smaller than or equal to the right - hand side . +
to prove the other inequality set @xmath188 } \big(t(x)+1-x \big)$ ] . in case of
@xmath189 we have @xmath190 for every @xmath83 so considering @xmath191 we are done .
suppose now that @xmath192 .
compactness of @xmath54 $ ] implies the existence of a sequence @xmath193 and a point @xmath194 $ ] such that @xmath195 and @xmath196 .
using @xmath197 we get @xmath198 and , using monotonicity of @xmath7 it follows that @xmath199 . letting @xmath200 \rightarrow [ 0,1]$ ] denote the rotation mentioned in the theorem , obviously @xmath183 holds . considering that for every @xmath201 $ ]
we have @xmath202 it follows immediately that @xmath203 } \big(t(x)+1-x \big ) , \end{aligned}\ ] ] which completes the proof .
@xmath152 considering that continuity of @xmath7 plays no role in theorem [ solutionmax ] , that @xmath7 has ( as non - decreasing function ) at most countably many discontinuities , and that @xmath204)=0 $ ] for every countable set @xmath205 and @xmath17 we may , w.l.o.g . , assume that @xmath7 is left continuous , in which case the infimum in eq .
( [ maxendo ] ) is a minimum .
suppose that @xmath2 are random variables with continuous distribution functions @xmath3 and @xmath4 respectively , set @xmath206 and @xmath207 } ( t(x)-x)$ ] , define @xmath200 \rightarrow [ 0,1]$ ] by @xmath208 , and let @xmath209 denote the completely dependent copula induced by @xmath210 .
then for @xmath211 with @xmath212 we have @xmath213 . [ exponential_lifetimes ]
suppose that the default times @xmath0 and @xmath1 are exponentially distributed with parameters @xmath214 and @xmath215 respectively .
it is straightforward to verify that in this case @xmath206 is given by @xmath216 , where @xmath217 . for the case of
@xmath218 we have @xmath219 for every @xmath92 $ ] , so @xmath220 . remarkably , for the case of @xmath221 the maximal mass of the endograph of @xmath222 and the maximal mass of the graph of @xmath222 coincide .
in fact , applying theorem [ solutionmax ] , on the one hand we get @xmath223 and on the other hand , according to theorem 3 and theorem 4 in @xcite ( also see @xcite ) we have @xmath224 } \big(\mathbf{1}_{[0,1 ] } ( f \circ t_\theta ) + \frac{1}{f\circ t_\theta } \mathbf{1}_{(1,\infty)}(f\circ t_\theta)\big ) \,d\lambda\ ] ] where @xmath225 denotes the density of @xmath226 . since for @xmath227 the latter
is given by @xmath228 , we get @xmath229 and eq .
( [ maxgraphjap_exponential ] ) calculates to @xmath230 } \frac{1}{\frac{1}{\theta } ( 1-x)^{1-\theta } } d\lambda(x ) + 1-\big(1- \theta^{\frac{1}{1-\theta}}\big ) = 1 - \theta^{\frac{\theta}{1-\theta } } + \theta^{\frac{1}{1-\theta}}.\\ & = & \sup_{a \in { \mathcal{c } } } \mu_a\big(\gamma^{\leq}(t_\theta ) \big).\end{aligned}\ ] ] for the special case of @xmath231 , which is depicted in figure [ grx2 ] , we get @xmath232 [ exgegen ] based on example [ exponential_lifetimes ] it might seem natural to conjecture that the equality @xmath233 holds for a much bigger class of non - decreasing transformations @xmath7 fulfilling @xmath234 for every @xmath92 $ ] . since counterexamples are easily constructed for the case where @xmath7 is singular ( @xmath235 for some @xmath236)$ ] with @xmath237 ) and the case where @xmath7 has discontinuities , the conjecture reduces to strictly increasing , continuous transformations @xmath7 . for every @xmath238 the transformation @xmath239 \rightarrow [ 0,1]$ ] , defined by @xmath240 \\ \frac{x}{2 } + \frac{x}{2 } \sqrt[n]{4x-2 } & \textrm{if } x \in ( \frac{1}{2},\frac{3}{4 } ) \\
-1 + 2x & \textrm{if } x \in [ \frac{3}{4},1 ] \end{array } \right.\ ] ] is easily verified to be homeomorphism with @xmath241 for every @xmath92 $ ] ( see figure [ homgegenbsp ] for the case @xmath242 ) .
applying theorem [ solutionmax ] we get @xmath243 , however , either by graphical arguments or by using theorem 3 and theorem 4 in @xcite it is straightforward to verify that @xmath244 , so the conjecture is wrong .
although monotonicity is crucial in the proof of theorem [ solutionmax ] it is even possible to calculate @xmath245 for the case of arbitrary measurable ( not necessarily monotonic ) transformations @xmath246 . letting @xmath5 \rightarrow [ 0,1]$ ] denote an arbitrary measurable transformation , we will now directly concentrate on the quantity @xmath247 , defined by @xmath248 and prove a simple formula for @xmath247 only involving the d.f .
@xmath249 \rightarrow [ 0,1]$ ] of @xmath7 , defined by @xmath250)=\lambda(t^{-1}([0,x])).\ ] ] we start with two simple lemmata that will be used in the proof of the main results - the first one contains an alternative simple formula for @xmath247 involving @xmath251 which will be key in the proofs of the main results , the second one gathers two properties describing how much @xmath247 may change if @xmath7 changes . [ lemupper ] suppose that @xmath5 \rightarrow [ 0,1]$ ] is measurable .
then we have @xmath252 } \big(t(x)-f_t\circ t ( x)\big ) = 1 + \inf_{y \in [ 0,1 ] } \big(y - f_t(y)\big ) = 1 + \min_{y \in [ 0,1 ] } \big(y - f_t(y)\big)\ ] ] if @xmath7 is non - decreasing then we have equality in ( [ maxendogen ] ) .
* proof : * considering @xmath253 \times [ 0,t(x ) ] \ , \cup \ , t^{-1}((t(x),1 ] ) \times [ 0,1]$ ] and using @xmath254)=1-f_t \circ t(x ) $ ] we get @xmath255 holds for every @xmath92 $ ] and every @xmath17 , from which the first inequality follows immediately . + to prove the second part of ( [ maxendogen ] ) it suffices to show that for every @xmath156 $ ] we have @xmath256 } \big(t(x)-f_t\circ t ( x)\big ) \leq y - f_t(y).\ ] ] in the following @xmath257 will denote the range of @xmath7 , @xmath258 its topological closure .
it is easy to see that the left hand - side of ineq .
( [ eqtemp ] ) can not exceed zero : in fact , setting @xmath259 there are two possibilities : if @xmath260 for some @xmath92 $ ] then @xmath261 . and if @xmath262 then @xmath263 , so @xmath264 is a continuity point of @xmath251 and , by construction , we can find a monotonically increasing sequence @xmath265 converging to @xmath264 , implying @xmath266 .
+ for @xmath267 and , using the previous paragraph , for @xmath268 and for @xmath269 ineq .
( [ eqtemp ] ) is trivial .
the inequality is also clear for @xmath270 since in case of @xmath271 we have @xmath267 .
suppose now that @xmath272 and that @xmath273
. then obviously @xmath274 , i.e. @xmath275 is a continuity point of @xmath251 .
consequently , if @xmath276 then there exists a sequence @xmath265 converging to @xmath275 , so @xmath277 } \big(t(x)-f_t\circ t ( x)\big)$ ] . considering that @xmath278 } t(x)$ ] implies @xmath269 , whence ineq .
( [ eqtemp ] ) , it remains to prove the inequality for the case that @xmath279 , @xmath280 } t(x)$ ] and @xmath281 .
setting @xmath282 we have @xmath283 as well as @xmath284 , so @xmath285 .
since the construction of @xmath286 implies @xmath287 the proof of ineq .
( [ eqtemp ] ) is complete .
+ proving the existence of @xmath288 $ ] fulfilling @xmath289 } \big(y - f_t(y)\big ) = y^\star - f_t(y^\star)\big)$ ] can be done as follows : for every @xmath238 we can find @xmath290 $ ] with @xmath291 .
compactness of @xmath54 $ ] implies the existence of a subsequence @xmath292 and some @xmath288 $ ] with @xmath293 . if @xmath294 we are done since @xmath295 .
suppose therefore that @xmath296 and let @xmath297 $ ] be arbitrary .
then there exists an index @xmath298 such that @xmath299 , hence @xmath300 , holds for all @xmath301 . considering @xmath302 yields @xmath303 , hence , using right - continuity of @xmath251 we get @xmath304 .
+ finally , suppose that @xmath7 is non - decreasing .
we want to show that @xmath305 } ( t(x)-f_t\circ t ( x ) ) = \inf_{x \in [ 0,1 ] } ( t(x)-x)\ ] ] it follows directly from the construction that @xmath306 holds for every @xmath92 $ ] , so the left - hand side of ( [ mono2 ] ) can not be greater than the right - hand side . additionally , it is straightforward to verify that @xmath307 holds if and only if there exists @xmath308 with @xmath309 .
hence in case of @xmath310 , setting @xmath311 , @xmath312 follows and , using @xmath313 , we get that @xmath314 } ( t(x)-x)\leq t(x_0)-f_t\circ t ( x_0)$ ] , which completes the proof .
@xmath152 [ sensitiv ]
suppose that @xmath315 \rightarrow [ 0,1]$ ] are measurable transformations .
then the following two assertions hold : 1 . for @xmath316:t(x)\not = t'(x)\}$ ]
we have @xmath317 .
if @xmath318 and @xmath319 , then @xmath320 follows .
* proof : * to prove the first assertion set @xmath321 and @xmath322 .
considering that obviously @xmath323 as well as @xmath324)=\lambda(d)$ ] holds for every @xmath17 , we get @xmath325 for every @xmath17 . having this
, the desired inequality follows immediately .
+ to prove the second assertion let @xmath326 \rightarrow [ 0,1]$ ] be defined by @xmath327 and fix @xmath17 . since obviously @xmath328 , defining @xmath329 yields a doubly stochastic measure @xmath330 which corresponds to a copula @xmath331 ( that , in turn , is easily seen to be the transpose of the @xmath332-shuffle @xmath333 of @xmath33 ) .
defining @xmath334 \rightarrow [ 0,1]$ ] by @xmath335 , @xmath336 follows and , using disintegration , we get @xmath337 ) } k_{a_\delta}\big(x,[0,t(x)-\delta]\big ) d\lambda(x ) \\ & = & \int_{t^{-1}([\delta,1 ] ) } k_{a}\big(x,[\delta , t(x)]\big ) d\lambda(x ) \\ & = & \int_{[0,1 ] } k_{a}\big(x,[0,t(x)]\big ) d\lambda(x ) - \int_{t^{-1}([0,\delta ) ) } k_{a}\big(x,[0,t(x)]\big ) d\lambda(x ) \\ & & \quad - \,\int_{t^{-1}([\delta,1 ] ) } k_{a}\big(x,[0,\delta)\big ) d\lambda(x ) \\ & \geq & \mu_a(\gamma^{\leq}(t ) ) - \int_{[0,1 ] }
k_{a}\big(x,[0,\delta)\big ) d\lambda(x ) = \mu_a(\gamma^{\leq}(t ) ) - \delta.\end{aligned}\ ] ] since @xmath17 was arbitrary it follows immediately that @xmath320 .
@xmath152 + we now tackle the calculation of @xmath247 for arbitrary measurable @xmath7 in two steps - we first prove the result for continuous @xmath7 and then extend it via lusin s theorem ( see @xcite ) and some compactness arguments to the general case . since the proof for riemann - integrable @xmath7 is only slightly more complicated than that for continuous @xmath7 we directly focus on riemann - integrable transformations @xmath5 \rightarrow [ 0,1]$ ] .
[ riemann ] suppose that @xmath5\rightarrow [ 0,1]$ ] is riemann - integrable .
then we have @xmath338}(x - f_t(x)).\ ] ] * proof : * let @xmath339 denote the oscillation of @xmath7 at the point @xmath92 $ ] , i.e. @xmath340 where @xmath341 : \vert z - x\vert < r\}$ ] .
it is well known ( @xcite ) that @xmath342 is upper semincontinuous and that @xmath83 is a continuity point of @xmath7 if , and only if @xmath343 .
+ in what follows let @xmath344 be arbitrary but fixed .
riemann - integrability ( @xcite ) of @xmath7 implies that @xmath345 : w_t(x ) \geq \varepsilon\}$ ] is compact and fulfills @xmath237 so we can find open intervals @xmath346 such that @xmath347 and @xmath348 holds .
\setminus \bigcup_{i=1}^n u_i$ ]
. for every @xmath350 we have @xmath351 and , using compactness of @xmath352 , we can find pairwise disjoint intervals @xmath353 such that @xmath354 and @xmath355 holds for every @xmath356 . defining a step function @xmath357 \rightarrow [ 0,1]$ ] by @xmath358 we have @xmath359 for every @xmath350 . applying lemma [ sensitiv ] yields @xmath360 . proceeding in completely the same manner we can construct a sequence @xmath361 of step functions such that @xmath362 and @xmath363 holds for every @xmath238 , which implies @xmath364 . for each @xmath365 we can reorder the intervals on which @xmath366 is constant in such a way that the resulting step function @xmath367 is monotonically increasing . working with classical shuffles
it is straightforward to verify that @xmath368 holds . according to lemma [ lemupper ] we have @xmath369 } ( y - f_{t_n}(y))$ ] , so taking into account @xmath370 altogether we have already shown @xmath371 } ( y - f_{s_n}(y))\big ) = \overline{m}_t,\ ] ] and we are done if we can prove that @xmath372 } ( y - f_{s_n}(y))}_{=:i_n}\big ) = \underbrace{\min_{y \in [ 0,1 ] } ( y - f_t(y))}_{=:i}.\ ] ] considering @xmath362 we have @xmath373 , so @xmath374 .
the construction of @xmath366 implies @xmath375 , so ( see @xcite ) there exists a subsequence @xmath376 converging @xmath55-a.e . to @xmath7 .
using the fact that almost sure convergence implies weak convergence ( see @xcite ) it follows that @xmath377 holds for every point @xmath156 $ ] at which @xmath251 is continuous .
choose @xmath378 in such a way that @xmath379 } ( y - f_{s_{n_i}}(y))=y_{n_i}-f_{s_{n_i}}(y_{n_i})$ ] .
( consider another subsequence if necessary ) assume that @xmath380 converges to some @xmath288 $ ] .
for @xmath294 we have @xmath381 , from which @xmath382 follows .
suppose therefore that @xmath383 and let @xmath384 denote a continuity point of @xmath251 .
then there exists an index @xmath385 such that for all @xmath386 we have @xmath387 , so , in particular , @xmath388 . since the latter implies @xmath389 , taking into account that the set of all continuity points of @xmath251 is dense , we finally get @xmath390 , which completes the proof of the theorem
. @xmath152 [ main ]
suppose that @xmath5\rightarrow [ 0,1]$ ] is measurable .
then we have @xmath338}(x - f_t(x)).\ ] ] * proof : * for every @xmath238 , lusin s theorem ( see @xcite ) implies the existence of a compact set @xmath391 $ ] and a continuous ( hence riemann - integrable ) function @xmath367 such that @xmath392 for all @xmath393 and @xmath394 . applying lemma
[ sensitiv ] and theorem [ riemann ] immediately yields @xmath395}(x - f_{t_n}(x))\big)\ ] ] and the theorem is proved if we can show that @xmath396 } ( x - f_{t_n}(x))}_{=:i_n}\big ) = \underbrace{\min_{x \in [ 0,1 ] } ( x - f_t(x))}_{=:i}.\ ] ] for @xmath238 and arbitrary @xmath92 $ ] we get @xmath397 ) = \lambda\big(\{z \in e_n : t_n(z ) \leq x \}\big ) + \lambda\big(\{z \in e_n^c : t_n(z ) \leq x \}\big ) \nonumber \\ & = & \lambda\big(\{z \in [ 0,1 ] : t(z ) \leq x \}\big ) - \lambda\big(\{z \in e_n^c : t(z ) \leq x \}\big ) + \lambda\big(\{z \in e_n^c : t_n(z ) \leq x \}\big ) \nonumber \\ & = & f_t(x ) + \delta\end{aligned}\ ] ] for some @xmath398 . since @xmath83 was arbitrary this implies that @xmath399 converges uniformly to @xmath251 , based on which it is straightforward to prove eq .
( [ lebesgue_final ] ) : ( i ) if @xmath400 for some @xmath194 $ ] then eq .
( [ tempunif ] ) and the definition of @xmath401 yield @xmath402 from which @xmath403 follows immediately .
( ii ) to prove the opposite inequality , for every @xmath238 choose @xmath404 $ ] such that @xmath405 .
applying eq .
( [ tempunif ] ) yields @xmath406 from which @xmath407 follows .
theorem [ main ] can be proved in a different way by using lemma [ lemupper ] and results from @xcite .
in fact , slightly modifying the ideas in the first section of @xcite it can be shown that for each measurable @xmath5 \rightarrow [ 0,1]$ ] there exists a non - decreasing function @xmath408 \rightarrow [ 0,1]$ ] ( called the non - decreasing rearrangement of @xmath7 ) and a @xmath55-preserving transformation @xmath409 \rightarrow [ 0,1]$ ] such that @xmath410 holds . having this , letting @xmath411 denote the operator
studied in @xcite and implicitly defined via @xmath412 and using disintegration and change of coordinates we get that @xmath413 } k_{\mathcal{u}_\varphi(a)}(x,[0,t(x ) ] ) d\lambda(x ) = \int_{[0,1 ] } k_{a}\big(\varphi(x),[0,t^\star \circ \varphi(x)]\big ) d\lambda(x ) \nonumber \\ & = & \int_{[0,1 ] } k_a\big(z,[0,t^\star(z)]\big ) d\lambda(z ) = \mu_a(\gamma^{\leq}(t^\star))\end{aligned}\ ] ] holds for every @xmath17 , implying @xmath414 . again using @xmath415 and
the fact that @xmath416 is @xmath55-preserving , it is straightforward to verify that @xmath7 and @xmath417 have the same d.f . , i.e. @xmath418 holds .
therefore , applying lemma [ lemupper ] yields @xmath419}(x - f_{t}(x))=1+\min_{x \in [ 0,1]}(x - f_{t^\star}(x ) ) = \overline{m}_{t^\star } \leq \overline{m}_t \leq 1+\min_{x \in [ 0,1]}(x - f_{t}(x)),\ ] ] from which the desired equality @xmath420 follows immediately .
although this alternative proof is shorter we opted for the one presented in the previous section since , firstly , it is self - contained and , secondly , lemma [ sensitiv ] is interesting in itself and will also be used in the sequel when deriving some corollaries . according to theorem [ solutionmax ] the completely dependent copula @xmath421 fulfills @xmath422 , so eq . ( [ koopman ] ) implies @xmath423 . by definition of @xmath424
we have @xmath425 so @xmath426 is completely dependent and the following corollary holds : having found a simple and easily computable formula for the maximal mass of @xmath429 we now derive the analogous result for the minimal mass and set @xmath430 given the aforementioned results , the subsequent corollary does not come as a surprise : * proof : * we first concentrate on the _ strict endograph _
@xmath432 , defined by @xmath433 ^ 2 : y < t(x)\big\}.\ ] ] defining @xmath239 \rightarrow [ 0,1]$ ] by @xmath434 for every @xmath92 $ ] and @xmath238 we obviously have that @xmath435 is monotonically increasing and that @xmath436 .
lemma [ sensitiv ] yields @xmath437 and corollary [ supattained ] implies the existence of a copula @xmath438 with @xmath439 .
altogether we get @xmath440 so considering @xmath441 shows that @xmath442 . having this , eq .
( [ eqminmass ] ) is a straightforward consequence since @xmath443}(x - f_{1-t}(x ) ) = \max_{x \in [ 0,1]}(x - f_{t}(x ) ) .
\,\blacksquare \end{aligned}\ ] ] we close the paper with two examples - the first one shows that @xmath444 is not necessarily attained whereas the second one considers a non - monotonic transformation for which copulas attaining @xmath444 and @xmath247 can easily be constructed . for @xmath445
, considering rotations @xmath332 and the corresponding shuffles @xmath446 , it follows immediately that @xmath447 .
there is , however , no copula @xmath33 fulfilling @xmath448 , i.e. contrary to @xmath247 , there are situations , in which @xmath444 is not attained for any copula .
suppose , on the contrary , that @xmath17 fulfills @xmath448 .
then , defining @xmath107 by @xmath449 and setting @xmath450 , we have @xmath451 , so , @xmath452 holds for every @xmath453 $ ] .
the latter implies @xmath454 , which is a contradiction .
[ examplefinal ] for @xmath455 it is straightforward to find mappings @xmath417 and @xmath416 such that eq .
( [ arrangementt ] ) holds .
in fact , defining @xmath456 \rightarrow [ 0,1]$ ] by @xmath457\\ -1 + 2x & \textrm{if } x \big(\frac{1}{2},1\big ] \end{array } \right.\ ] ] and setting @xmath458 we immediately get @xmath459 . using eq .
( [ cdmax ] ) , and setting @xmath460 , it follows that @xmath461 is @xmath55-preserving and that @xmath427 fulfills @xmath462 . considering that for @xmath463 we obviously have @xmath464
, we get @xmath465 which coincides with @xmath466}(x - f_{t}(x))$ ] .
figure [ exfinal ] depicts the supports of the copulas @xmath100 and @xmath463 as well as the endograph of @xmath7 .
mai , m. scherer : simulating from the copula that generates the maximal probability for a joint default under given ( inhomogeneous ) marginals , _ topics from the 7th international workshop on statistical simulation _ ed . v. melas , s. mignani , p. monari , and l. salmaso , springer proceedings in mathematics & statistics * 114 * , pp .
333 - 341 , 2014 . | we solve the problem of maximizing the probability that @xmath0 does not default before @xmath1 within the class of all random variables @xmath2 with given distribution functions @xmath3 and @xmath4 respectively , and construct a dependence structure attaining the maximum .
after translating the maximization problem to the copula setting we generalize it and prove that for each ( not necessarily monotonic ) transformation @xmath5 \rightarrow [ 0,1]$ ] there exists a completely dependent copula maximizing the mass of the endograph @xmath6 of @xmath7 and derive a simple and easily calculable formula for the maximum .
analogous expressions for the minimal mass are given .
several examples and graphics illustrate the main results and falsify some natural conjectures .
copula , dependence , coupling , endograph , markov kernel + 60e05 , 28a50 , 91g70 [ thm]lemma [ thm]proposition [ thm]corollary [ thm]remark [ thm]example [ thm]definition |
supergiant fast x ray transients ( hereafter sfxts ; smith et al .
2006a ) are a new class of hard x ray sources mostly discovered by the @xmath2 satellite ( negueruela et al .
2005a , sguera et al . 2005 ) .
they are transient sources which seem to emit x
rays only during `` short '' outbursts ( few hours , as observed with @xmath2 or @xmath3 ) and their optical counterparts are blue supergiant stars .
their x ray spectra are similar to those of accreting pulsars , thus it is likely that they are high mass x ray binaries ( hmxbs ) hosting neutron stars . in two sfxts ( among about twenty sources , comprising candidate sfxts ) x
ray pulsations have been indeed observed ( igr j18410 - 0535/ax j1841.0 - 0536 , p@xmath4=4.74 s , bamba et al .
2001 ; igr j112155952 , p@xmath5s , smith et al . 2006b , swank et al . 2007 ) .
sfxts reach x ray luminosities up to a few 10@xmath6 erg s@xmath7 , while the quiescent level ( @xmath810@xmath9 erg s@xmath7 ) has been observed only in igr j17391@xmath13021/xte j1739@xmath1302 ( smith et al . 2006a ; sakano et al .
2002 ) , igr j17544@xmath12619 ( int zand 2005 ) and probably igr j18410@xmath10535/ax j1841.0@xmath10536 ( halpern et al .
2004 ; bamba et al .
2001 ) .
it is important to note that none of the sfxt sources has ever been caught to undergo a transition from quiescence to outburst and then back to quiescence in a few hours .
the quiescent emission had always been observed well far away from the outbursts , except in one case : only int zand ( 2005 ) did observe the transition from quiescence to outburst with @xmath10 ( in igr j17544@xmath12619 ) , but the observation finished before the start of the declining phase to quiescence .
thus the real duration of this outburst could not be measured .
the so - called `` short '' duration ( a few hours ) of the outbursts from sfxts is indeed based on observations with instruments not sensitive enough to detect the quiescence level .
the instruments onboard @xmath3 and @xmath2 could only observe the brightest fast flaring activity ( lasting a few hours , less than one day ) reaching a few 10@xmath6 erg s@xmath7 . hence , the definition of sfxts as transient sources displaying `` short '' x ray outbursts lasting only a few hours is strongly biased .
this has been observationally demonstrated by our recent deep campaign with @xmath0 ( romano et al .
2007 , hereafter paper ii ) of the outburst from the unique sfxt displaying `` periodic outbursts '' , igr j112155952 ( sidoli et al .
2006 , hereafter paper i ) .
these very sensitive observations showed that the accretion phase in sfxts lasts longer than what previously thought : a few days instead of only hours . with these new observations at hand
, we report on an alternative model to explain the outbursts from this new class of sources , based on @xmath0 monitoring observations of igr j112155952 during the last two outbursts ( starting on february 9 and july 24 , 2007 ; sidoli et al .
2007 , hereafter paper iii ) .
is an x ray transient discovered by during a fast outburst in april 2005 ( lubinski et al .
the optical counterpart is a b1 supergiant , hd 306414 , located at a distance of 68 kpc ( negueruela et al .
2005b , masetti et al .
2006 , steeghs et al .
2006 ) . from the analysis of observations of the source field
, we discovered ( paper i ) that the outbursts are equally spaced by @xmath8330 days ( although a half of this period could not be excluded , due to a lack of observations ) .
this periodicity was later confirmed in march 2006 ( outburst after 329 days ; smith et al .
2006c ) during a monitoring with @xmath11 , and was related in a natural way to the orbital period of the system , with the outbursts triggered at ( or near to ) the periastron passage ( paper i ) .
based on this known periodicity , a new outburst was expected for 2007 february 9 and we planned a monitoring campaign with @xmath0 , starting on 2007 february 4 ( romano et al .
a second monitoring campaign was performed with @xmath0 in july 2007 , in order to monitor the quiescent level and the epoch of the supposed apastron passage ( based on the 329 days period ; romano et al .
these observations led to the detection of a new unexpected outburst starting on 2007 , july 24 , which reached roughly the same flux as during the february 2007 outburst ( paper iii ) .
details of the @xmath0 data analysis and spectral / timing results are reported in paper ii and paper iii . here
we concentrate on the shape of the x ray lightcurve in order to understand the physical mechanism which produces the outbursts .
the igr j112155952 lightcurve observed during the february 2007 outburst represents the most complete set of observations of a sfxt outburst ( fig . [
fig : ecc ] , black curve ) . the first important result of these observations is that the whole outburst phase lasts longer than what previously thought , based on less sensitive instruments : a few days , instead of a few hours .
only the brightest part of the outburst is short ( lasts less than 1 day ) and would have been seen by the integral instruments .
intense flaring activity is also present , both during the bright peak and the declining phase of the outburst , with each single flare lasting minutes or a few hours .
it is natural to associate the clock responsible for the outbursts with the orbital periodicity of the binary system . since displayed a new outburst after about a half of the 329 days period ( paper iii ; romano et al .
2007c ) , it is possible that 164.5 days is indeed the real orbital period which escaped detection up to now .
in both cases ( p@xmath12=329 days or 164.5 days ) , the system is a wide binary where the blue supergiant does not fill its roche lobe , and the system is very likely wind - fed .
applying the bondi - hoyle wind accretion scenario , where the neutron star accretes from the wind of the supergiant at different rates depending on the wind density and relative velocity along the orbit , and assuming reasonable parameters for the b - supergiant , we obtain that the observed x ray lightcurve is always too narrow and steep to be explained with accretion from a spherically symmetric wind , even adopting extreme eccentricities for the binary system ( see fig .
[ fig : ecc ] ) .
this result led us to suggest that the wind from the b supergiant is not spherically symmetric .
the alternative viable explanation we propose for the sharpness of the observed x
ray lightcurve is that in the supergiant wind has a second component ( besides the polar spherically symmetric one ) , in the form of an `` equatorial disk '' , inclined with respect to the orbital plane ( see fig .
[ fig : geom ] for an artistic view of the geometry of the system ) . the short outburst is then produced when the neutron star crosses this equatorial wind component , denser and slower than the polar one .
deviations from spherical symmetry in hot massive star winds are also suggested by optical observations ( e.g. prinja 1990 , prinja et al .
2002 ) and the presence of equatorial disk components , denser and slower with respect to the polar wind , also results from simulations ( ud - doula et al .
2006 ) .
the thickness _ h _ of the densest part of this supergiant equatorial wind can be calculated from the duration of the brightest part of the outburst ( which lasts less than 1 day , time needed for the neutron star to cross it ) and the neutron star velocity , 100200 km s@xmath7 : _
cm , where @xmath16 is the inclination angle of the equatorial wind with respect to the orbital plane ( with @xmath16=90@xmath17 if the disk is perpendicular to the orbital plane ) .
the model we are proposing can explain also the short flares from all the other sfxts where a clear periodicity in the outbursts recurrence has not been found yet , if a different geometry of the equatorial wind component with respect to the orbital plane is assumed : in , where the outbursts are equally spaced and occur with a fixed periodicity , the inclined equatorial disk wind component should intersect the neutron star at the periastron ( or very close to it , see the left panel in fig .
[ fig : geom ] ) and can intersect the neutron star orbit once or twice depending both on the extension of the wind disk and on the orbital eccentricity . instead , it is possible that in the other sfxts the inclined disk wind intersects twice a wide and highly eccentric orbit , not at the periastron ( see the right panel in fig . [
fig : geom ] ) , leading to a double periodicity ( one shorter than the other ) which has not been found yet _ only _ because of a lack of a continuous monitoring .
this model can also explain the x ray emission from the persistently accreting hmxbs , if we admit that in this case the neutron star is always moving inside the equatorial wind component which lies on the orbital plane . in this framework , the sharp x
ray lightcurve observed from can be modelled with different wind parameters ( for both polar and equatorial components ) depending on the orbital period ( 164.5 days or 329 days ) and the eccentricity of the binary .
we assume a blue supergiant with a mass of 39 m@xmath18 and radius of 42 r@xmath18 , and a polar wind component with a terminal velocity of 1800 km s@xmath7 .
the x ray lightcurve observed with @xmath0 is better reproduced assuming a `` polar wind '' mass loss rate of 5@xmath1310@xmath19 m@xmath18 yrs@xmath7 ( for a p@xmath12 of 164.5 days and an eccentricity of 0.4 ) and 9@xmath1310@xmath20 m@xmath18 yrs@xmath7 ( for a p@xmath12 of 329 days and a circular orbit , which is required by the fact that the two consecutive outbursts from reached roughly the same peak flux ) .
the equatorial wind component should have a variable velocity ranging from 750 km s@xmath7 to 1400 km s@xmath7 ( for p@xmath12=164.5 days ) , and from 850 km s@xmath7 to 1600 km s@xmath7 ( for p@xmath12=329 days ) , and a density about 100 times higher than the polar wind component .
note however that since the x ray luminosity expected for the wind accretion is proportional to @xmath21@xmath22@xmath23 ( where @xmath21 is the wind mass loss rate , and @xmath22 is the relative velocity of the wind with respect to the neutron star ) , different combinations of wind density and velocity in the equatorial component can reproduce the x ray lightcurve as well . in conclusion ,
in our model we explain the short recurrent flares if the neutron star intersects an inclined equatorial wind component ( once or twice ) during its orbit .
a different particular geometry and inclination of this equatorial wind with respect to the orbital plane can account for the whole phoenomenology of both sfxts and persistently accreting hmxbs in general .
both the orbital eccentricity and no - coplanarity can be explained by a substantial supernova kick at birth .
this could indicate that sfxts are likely young systems , probably younger than persistent hmxbs .
based on observations obtained with xmm - newton , an esa science mission with instruments and contributions directly funded by esa member states and the usa ( nasa ) . based on observations with integral , an esa project with instruments and the science data centre funded by esa member states ( especially the pi countries : denmark , france , germany , italy , switzerland , spain ) , czech republic and poland , and with the participation of russia and the usa .
we thank the , , and @xmath24 teams for making these observations possible , in particular the duty scientists and science planners .
pr thanks inaf - iasfmi for their kind hospitality . dg acknowledges the french space agency ( cnes ) for financial support .
this work was supported by contract asi / inaf i/023/05/0 . | the physical mechanism responsible for the short outbursts in a recently recognized class of high mass x - ray binaries , the supergiant fast x - ray transients ( sfxts ) , is still unknown .
recent observations performed with @xmath0 , and of the 2007 outburst from igrj11215@xmath15952 , the only sfxt known to exhibit periodic outbursts , suggest a new explanation for the outburst mechanism in this class of transients ; the outbursts could be linked to the possible presence of a second wind component in the supergiant companion , in the form of an equatorial wind .
the applicability of the model to the short outburst durations of all other supergiant fast x - ray transients , where a clear periodicity in the outbursts has not been found yet , is discussed .
the scenario we are proposing also includes the persistently accreting supergiant high mass x ray binaries . |
Subsets and Splits