Вы находитесь на странице: 1из 57

Home

Search

Collections

Journals

About

Contact us

My IOPscience

Ginzburg-Landau theory for superconductors

This article has been downloaded from IOPscience. Please scroll down to see the full text article. 1973 Rep. Prog. Phys. 36 103 (http://iopscience.iop.org/0034-4885/36/2/001) View the table of contents for this issue, or go to the journal homepage for more

Download details: IP Address: 131.247.152.4 The article was downloaded on 15/03/2012 at 16:07

Please note that terms and conditions apply.

Ginzburg-Landau theory for superconductors


M CYROT
Laboratoire de Physique des Solides, Biitiment 510, UniversitC Paris-Sud, 91 Orsay, France

Abstract
This review article describes how the simple Ginzburg-Landau approach lies in the heart of the general theory of superconductors. I t introduces the reader to the handling of the theory and to the numerous possibilities of applications. A general free energy functional for a superconductor is given and we study the different cases where it can be reduced to a Ginzburg-Landau form or to a simple generalization of this form. We emphasize that applications are not restricted to thermodynamical ones as the Ginzburg-Landau approach can be used in the calculation of dissipative phenomena. T h e possibility of an extension to the time-dependent phenomena is discussed in detail to present the difficult problems which arise in that case. T h e gapless regime of type I1 superconductors is given as an example. This review was completed in July 1972.

Rep. Prog. Phys. 1973 36 103-158


5

104

M Cyrot

Contents 1. Introduction . 2. Introductory survey . 2.1. Basic phenomena . 2.2. London equations. 2.3. T h e Ginzburg-Landau theory , . 2.4. T h e BCS theory and the energy gap 2.5, T h e fall of the hallmarks of superconductivity . . 2.6. T h e universal characteristic of superconductivity 3. T h e Ginzburg-Landau equations-phenomenological approach . 3.1. T h e postulated free energy . 3.2. Ginzburg-Landau equations . 3.3. T h e two fundamental lengths . 3.4. Use of the Ginzburg-Landau equations . 3.5. Applications. 4. NIicroscopic derivation of the Ginzburg-Landau equations . 4.1. T h e BCS hamiltonian . 4.2. General free energy functional for a superconductor . 4.3. Derivations of the Ginzburg-Landau equations. 4.4. Applications. 5. Validity and breakdown of the Landau theory . 5.1. T h e Ginzburg criterion . 5.2. Fluctuations above T, . 6. Extension of the Ginzburg-Landau equations . 6.1. Preliminary remarks . 6.2. Small order parameter . 6.3. T h e slow variation approximation . 6.4. Applications. 6.5. Other extensions . 7. Time-dependent Ginzburg-Landau equations . 7.1. Position of the problem . 7.2. Nonequilibrium thermodynamics of superconductors . 7.3. Possibility of a linear time-dependent Ginzburg-Landau equation 7.4. T h e linear time-dependent Ginzburg-Landau equations . 7.5. Applications of the linear time-dependent Ginzburg-Landau . equation 7.6. T h e nonlinear term . 7.7. Applications. Acknowledgments . References .
Page 105 106 106 107 108 109 110 112 112 112 114 115 116 117 122 123 124 126 130 131 131 133 135 135 135 138 141 142 142 142 143 145 146

148 149 154 156 156

Ginxburg-Landau theory f o r superconductors

105

1. Introduction
T h e superconducting state is due to an instability of the normal state in the presence of an attraction between electrons. This attraction is made possible by an interaction through the lattice. Pairs of electrons of opposite spin, or Cooper pairs, are then formed and a new state is built up from these pairs. This state cannot be obtained by perturbation theory and a self-consistent approach is used. One assumes a local pair potential which is roughly proportional to the number of Cooper pairs and determines the motion of the electrons and pairs in this potential. This pair potential is then determined self-consistently. T h e mathematics consists of solving two coupled equations with an unknown pair potential and then its determination by a self-consistent equation. T h e complexity of these two steps prevents solutions of the problems in most cases, and the whole apparatus tends to hide the physical simplicity of the problem. Long before the microscopic theory, a phenomenological approach to superconductivity had been proposed by Ginzburg and Landau. A priori it has nothing to do with the microscopic theory and is based on the idea that the normal superconducting transition is, in the absence of a magnetic field, a thermodynamical second-order transition. Thus, one can apply to it the general theory of secondorder transitions and define an order parameter # which is zero in the disordered state (the normal metal) and finite in the ordered state (the superconducting metal). From general considerations an equation for the order parameter can be given, this is a second-order differential equation which has to be solved withsome boundary conditions. T h e two approaches, microscopic and phenomenological, remained completely separate up to the time when Gorkov showed that, in some limiting cases, the order parameter of the phenomenological equations and the pair potential of the microscopic theory are proportional. This gave a considerable interest to this approach. Firstly, the complicated mathematical apparatus of the microscopic theory is replaced by a simple differential equation for the order parameter and secondly, the comparison between the two approaches threw light on the basic physics of superconductors, that is, the existence of an order parameter in a thermodynamical sense, even if it has a quantum nature. Superconductivity is just a case of long-range order. Thus, when derived microscopically the Ginzburg-Landau approach brings two simplifications: (i) a conceptual one: the theory is recast in a very simple form based on physical intuition; (ii) a mathematical one: one is left with just one differential equation. T h e first question which arises is: when is this simplification possible ? At first, the equivalence was demonstrated only for a narrow range of temperature near the critical temperature T, in the absence of a magnetic field. However, this was greatly generalized especially near the second-order transition at H,, for dirty typeI1 superconductors and in some other cases of less generality. T h e second question one must put is: what kind of problems can be handled by this approach ? When first proposed the Ginzburg-Landau approach was used mainly for problems of a thermodynamic nature : critical fields, critical currents, etc. However, one must keep in mind that we can go back to the microscopic theory and now knowing the order parameter calculate the energy spectrum and other properties when we need

106

M Cyrot

the exact knowledge of this spectrum. T h e third question that we will envisage concerns the validity of the approach for time-dependent problems. For instance, if we apply an external magnetic field which varies very slowly one can assume that the order parametei will be given by the same equation as in the static case where the time is a parameter. On the other hand, if the field varies very rapidly with time we can think that the superconductor will respond to an average of the field as it often happens in other systems in physics for too rapid perturbations. Is it possible to describe all these behaviours by a time-dependent Ginzburg-Landau equation for the order parameter without going back to the complicated apparatus of nonequilibrium statistical thermodynamics ? T h e problem turns out to be far more complicated than the static one. This is due to the fact that a time-dependent perturbation can break Cooper pairs. So there is a possibility of local conversion of the normal excitations to superfluids and vice versa. Happily, it will appear that there exist situations of considerable interest where such a simplification is possible. These are strongly connected with gapless superconductivity. I n this article, an introductory survey gives the main features of superconductivity in connection with the history of the Ginzburg-Landau theory. I t is written for the fresh reader in superconductivity and can be omitted by readers with a very general background in this field. Then we give the original phenomenological approach of Ginzburg and Landau. We show what kind of problems this theory can handle independently of any microscopic theory and review some applications. T h e third part contains the microscopic theory and the Gorkov derivation of the Ginzburg-Landau equation. We emphasize that this derivation finds its interest not only in the justification of the phenomenological approach but also in a new way to handle many basic problems. T h e fourth part is devoted to an extension of the Ginzburg-Landau theory. We try to answer our first question on the possibility of a Landau approach and review the two main lines which have been studied. T h e last part gives the recent time development of the theory. It has both considerable practical and theoretical interest but is still an active domain of research. One just begins to understand how these problems can be handled but controversy still exists. We try to make clear the state of affairs and list a number of unsolved problems.

2. Introductory survey
We intend to survey the basic phenomena in superconductivity and develop what the Ginzburg-Landau approach brings to the theory. The conclusion will be that the characteristic feature of superconductivity is the existence of the order parameter introduced by Landau and Ginzburg.

2.1. Basic phenomena


Superconductivity was discovered in 1911 by Kamerlingh Onnes in Leiden. He observed that the electrical resistance of various metals, such as Pb, Hg, Sn, AI and In, disappears completely below a critical temperature T, (figure 1) characteristic of the material. T h e highest critical temperature known up to now is of the order of 20 K. T h e complete disappearance of resistance is most sensitively demonstrated by experiments on persistent currents in superconducting rings. Once set up, such currents have been observed to flow without measurable decrease for

Ginxburg-Landau theory for superconductors

107

years. Thus perfect conductivity was the first traditional hallmark of superconductivity. Even now it is the major attraction of superconductivity. T h e next important step was made in 1933 by Meissner and Ochensfeld. They found that a

Figure 1. Schematic plot of the resistivity of a superconductor near Tc.

magnetic field is expelled when a material becomes superconducting (figure 2) and also excluded when a magnetic field is set up on an already superconducting material, that is, the magnetic field inside a superconductor always vanishes. T h e second possibility can be explained by perfect conductivity but not the first one. Perfect conductivity would tend to trap flux inside. This is a completely new feature of superconductivity whose most important consequence is that the superconducting transition can be handled by thermodynamics. This has been emphasized

Figure 2. Magnetization curve of a long cylinder of type I material.

by F and H London. There is a superconducting thermodynamic state which does not depend on the magnetic history of the specimen. T h e first immediate consequence is that there exists a thermodynamical critical magnetic field H, above which the superconducting state is destroyed. It is related to the free energy difference between the normal and superconducting state in zero field. H, is determined by equating the energy H2/8.rrper unit volume associated with holding the field out against magnetic pressure with the difference in Helmoltz free energy

where n and s refer to the normal and superconducting state. The critical field at zero temperature for metals such as Al, Zn and Pb is lo2 to 103 G.

2.2. London equations T h e two basic electrodynamic properties were well described by H and F London in 1935 (London 1950). They proposed to add to Maxwells equations the

108

M Cyrot

following two relations in order to treat the electrodynamic properties of a superconductor:

E
rot J, where

4 ~ - J, c2

~2

at

-H 4 .A2 %
mc2 4nn, e2

(3)

A2 =

h is a phenomenological parameter and has the dimensions of length. T h e first equation is only a rewriting of the equation of motion for a free electron of mass m and charge e in the presence of an electric field E

dv mdt

eE.

(4)

n, is a kind of number density of superconducting electrons which is expected and found to vary continuously from zero at T, to the density of conduction electrons at zero temperature?. For one electron per atom A has a value of the order of cm. T h e second equation has been guessed so that the field is expelled over a distance A. As a matter of fact, application of Maxwells equation to equation (3) leads to

T h e field decreases from the surface exponentially over distance A. T h e existence of the penetration depth A was fully confirmed afterwards. I t is worth noting that both equations follow from

if a suitable gauge choice (the London gauge) is chosen for the vector potential A. London equations gave a semiquantitative description of the electrodynamic phenomena. Except for accumulation of experimental data, no further important step was taken before the appearance of the Ginzburg-Landau theory.

2.3. The Ginxburg-Landau theory Ginzburg and Landau (1950) pointed out that the London phenomenological theory is unsatisfactory from the two points of view. Firstly, it does not allow one to determine the surface tension at the boundary between the normal and the superconducting states. T h e surface energy connected with the field and supercurrent, as predicted from the solution of the equation of London, is negative, contrary to the observed positive surface energy. So, in order to obtain the positive surface energy, it is necessary to introduce a surface energy of nonmagnetic origin
f This is a two-fluid model assumption. We assume that there exist normal and superfluid electrons. The superfluid electron can transport current without dissipation.

Ginxburg-Landau theory f o r superconductors

109

which is greater than the one connected with the field distribution. T h e theory also does not enable the destruction of superconductivity by a current to be considered, as the number of superconducting electrons n, is just a function of temperature. T h e aim of Ginzburg and Landau was to construct a theory free from these defects. I n so doing they introduced entirely new features to the theory which permitted major new developments in the ensuing decade. I n the absence of a magnetic field the transition into a superconducting state at the critical temperature T, is of the second order. I n the general theory of such transitions (Landau and Lifshitz 1950) some parameter always enters which differs from zero in the ordered phase and which equals zero in the disordered phase. For example, in ferroelectrics the spontaneous polarization plays this role and similarly in ferromagnetics the spontaneous magnetization. Ginzburg and Landau introduced an order parameter # to describe the phenomena of superconductivity. I n the superconducting phase, which is the ordered phase, # # 0 while for temperatures above T,, # = 0 in the state of thermodynamic equilibrium. It was a long time before the exact significance of this order parameter and all its implications were understood. However, Ginzburg and Landau started from the idea that ) t represents some effective wavefunction of the superconducting electrons such that the density of superconducting electrons in the London equations was given by the local relation Then they assumed an expansion of the free energy in powers of # and V#. Minimization of the free energy with respect to the order parameter leads then to a differential equation for # which is analogous to a Schrodinger equation for a free particle but with a nonlinear term. T h e major early results of the theory were in handling two features beyond the scope of the London theory, namely the spatial variation of n, and the nonlinear effects in fields or currents strong enough to change n,. I n spite of these successes, when first proposed the theory appeared phenomenological and was little appreciated.

2.4. The BCS theory and the energy gap I n the fifties the third hallmark of superconductivity was established. This was the discovery of the existence of an energy gap between the ground state and the quasiparticle excitations of the system. T h e existence of such a gap was firmly established by a large number of experiments and became a keystone for the microscopic theory. Superconductivity was then characterized by the three basic phenomena : infinite conductivity, perfect diamagnetism and existence of an energy gap. Any theory had first to explain these three features. I n 1957 Bardeen, Cooper and Schrieffer (BCS) produced their pairing theory of superconductivity. An energy gap was predicted to increase from zero at T, to
2A = 3*52kBT,
at zero temperature. This result agrees with the data and the other BCS predictions were fully confirmed. This microscopic theory which seemed to end half a century of theoretical research was, however, only the beginning of our theoretical understanding of superconductivity.

110

M Cyrot

2.5. The fall of the hallmarks of superconductieity The most exciting developments came from the Russian workers following the Ginzburg-Landau theory. They were the successive fall of the three traditional hallmarks of superconductivity. This has been made possible by the microscopic derivation of the Ginzburg-Landau theory by Gorkov (1959). One year before, he had been able to generalize the BCS theory to deal with spatially varying situations (Gorkov 1958) and from this he showed that the Ginzburg-Landau theory was a limiting form of the microscopic theory valid near T,. Thus a firm basis was given to the existence of this order parameter. 2.5.1. Perfect diamagnetism. T h e prediction of superconductors in which flux enters did not even wait for a microscopic theory. It was made by Abrikosov (1957) but was overlooked at that time mainly because he used the Ginzburg-Landau theory. The theory introduces two characteristic lengths : the penetration depth X

Figure 3. Magnetization curve of a long cylinder of type I1 material.

characterizes the distance over which the magnetic field can vary and the coherence length 4 chacterizes the distance over which the order parameter can vary. T h e ratio of these two lengths defines the Ginzburg-Landau parameter.
K = -

x
4

For typical pure superconductors A < 4 and this was the reason for a positive surface energy associated with a normal superconducting boundary. Roughly speaking, one loses an energy [H,2/87r for the variation of 1c, from its superconducting value to zero while one gains hHc2/87rin reducing the magnetic energy, thus

Abrikosov investigated what would happen if K were large, that is, 6< A. This leads to a negative surface energy which tends to subdivide the material into domains characterized by the length 5. He called these materials type I1 superconductors and showed that the exact critical value for negative surface energy is K = JJ. For these materials there is a continuous increase in flux penetration, starting at a first critical field H,, and reaching B = H at a second critical field HOz,where the material becomes normal (figure 3). Because of this partial flux penetration the energy of holding the field out is less and H,, is greater than H,

HCz= K H , J ~ .

Ginxburg-Landau theory for superconductors

111

Abrikosov showed that the flux penetrates in a regular array of flux tubes or vortices each carrying a quantum of flux (figure 4)

- -= 2x - 2e

hc

lo-' Gcm2,

These type I1 superconductors are now known to be the more common in nature (any type I material when alloyed becomes type 11) and are used more in applications of superconductors. Much of the following work is devoted to these materials.

Figure 4. The Abrikosov array of flux lines.

2.5.2. The energy gap. T h e second hallmark to fall was the existence of an energy gap, Abrikosov and Gor'kov ( 1 9 6 9 studying the destruction of superconductivity by magnetic impurities, found that there exists a concentration range where the material remains superconducting but no gap exists. T h e general feeling that the order parameter is directly related to the energy gap and that perfect conductivity is more or less related to the energy gap took a long time to disappear. For superconductors with magnetic impurities the order parameter $ is different from zero and there is no gap. T h e comprehension of gapless superconductivity increased when de Gennes, deriving a Ginzburg-Landau equation for type I1 superconductors near Hc2,calculated the density of states and showed the absence of a gap. Direct experimental evidence came from tunnelling ; superconductivity was proved not to be connected with an energy gap. One must, however, note that the Australian school (Blatt 1964) had largely recognized this fact long before this discovery but at that time an energy gap was a keystone for a new theory. 2.5.3. Perfect conductivity. T h e final hallmark was also lost when it was demonstrated that a voltage appears when a current flows in the mixed state of a type I1 superconductor. T h e vortex lines experience a Lorentz force J x @,/c tending to make them move sideways. T h e motion of flux creates an electric field, which, in the presence of a current, creates a dissipation. This problem has great importance in the application of superconductivity and is linked with the last development of the Ginzburg-Landau theory. Recently many attempts have been made to study

112

M Cyrot

how the order parameter can vary with a time-dependent external perturbation and how dissipation is linked with the time dependence. This is the subject of the time-dependent Ginzburg-Landau equation.

2.6. The universal characteristic of superconductinity Faced with this experimental situation one might ask what is the characteristic of superconductivity. I n a masterstroke of physical intuition Ginzburg and Landau have given the answer. This is the existence of an order parameter $ ( r ) having a nonzero value. This does not explain what this order parameter means physically. Gorkov, when deriving the Ginzburg-Landau equation microscopically from the suitably generalized BCS theory, threw light on this problem. Cooper had shown that in the presence of an attractive interaction a pair of electrons outside the normal Fermi sea will form a bound state no matter how weak the interaction. It had been shown by Frohlich and by Pines and Bardeen that an interaction between electrons by exchange of virtual phonons does provide such an interaction. These works give the background for the development of the BCS theory which was recast by Gorkov in a many-body formalism. Gorkov showed that the order parameter can be defined as a thermal average ($r(r)$I(~)) where $J~(Y) destroys an electron of spin up at point r . Physically we can think of $ as being the amplitude probability for finding a Cooper pair at point Y . It has a quantum-mechanical nature and has no classical analogue. Superconductivity is defined by a nonzero value of this aver age.

3. The Ginzburg-Landau equations-phenomenological approach


I n this section we review the phenomenological approach of Ginzburg and Landau. We then describe what kind of problems can be handled independently of the microscopic theory. We emphasize the most remarkable application which is the prediction of type I1 superconductors in which flux penetrates in a regular array of flux lines.

3.1. The postulated free energy Superconductivity should be regarded as simply one of the many examples of phase transitions. Thus the primary observations of Ginzburg and Landau were that the state of order (the superconducting state) must be defined by an order parameter $) (. T h e order parameter is defined in order to be zero for a normal Y region and must characterize the degree of superconductivity at various points. I n a two-fluid model it must be related to the number of superconducting electrons. However, to allow for supercurrent flow, $(r) must be a complex function. Thus Ginzburg and Landau started from the idea that + ( r ) must be analogous to a wavefunction for the superconducting electrons. I n that case its absolute square can be identified with the density of superconducting electrons :
ndr) = I 12* Landau also pointed out that the equation for $ should be simpler if we consider only temperatures close to the transition point T,. Near a second-order transition the order parameter is small and its spatial variation is always slow. Using the general theory of second-order transitions (Landau and Lifshitz 1950) they constructed the free energy of a homogeneous superconducting material.

Ginxburg-Landau theory f o r superconductors

113

(i) First consider the case when $ is constant and there is no magnetic field. T h e free energy density depends only on the density of superconducting electrons and is as shown in figure 5 . We represent it by an expansion in powers of I $I2.

01

a( T -

T,) P

finite and positive constant.

This expansion assumes that F, is analytic near I z,b 1 = 0 and T = T,. I n fact this is not verifiable for most second-order transitions but is generally correct for superconductors. We shall justify this statement later on.

Figure 5 . Schematic plot of the free energy difference between the normal and superconducting states as a function of the order parameter.

(ii) If we allow $ to vary in space, we must add a term to prevent too rapid changes of 1 $ 1 which are physically not possible. This is due to the fact that in general the order parameter at one point is given as a function of the order parameter at all other points due to the finite range of the forces. Thus we have

#(Y)

[ K ( r- Y) $(r)d3r.

If we assume that $ is slowly varying near T,

$Y ()

= Jdr3K(r-r){z,b(r)+4(r.V)2z,b(r)+ ...}
21

hz,b(r)+ XR2V2$(r)

where h and AR2 are the zeroth and second moment of the kernel. This nonlocal correction contributes to the free energy density a term

$hR2 lV$12,

R representing an effective force range. T h e term h ($I2 has already been included in the 01 term. T h e normalization of the coefficient in the free energy is just a matter of convention. We will take it, following the general use, so that the coefficient is of IO$/* P / 2 m in order to make a close analogy with a Schrodinger equation. (iii) Finally we add the magnetic field Ha. This first increases F, by the amount ( 1 / 8 ~1 H ( r )- H a 1 2 . However F, must be independent of the gauge chosen for the ) vector potential A. As $ must behave somewhat like a wavefunction its phase can

114

M Cyrot

depend on the gauge. Thus, to ensure the invariance of F,, we replace

T h e presence of a charge 2e is physically very important. It expresses the fact that +(r) is an amplitude for a pair of electrons as me pointed out in the first part. However, in the original paper a factor e was used as a guess. The necessity of a charge 2e was pointed out after the appearance of the microscopic theory by Gor'kov (1959). So we write the free energy of a superconductor as

3.2. Ginzburg-Landau equations T h e total free energy is given by

9 J F r F,. =
Minimizing the free energy with respect to $ and A we arrive at the two GinzburgLandau equations

2eA
m

-ihv--)

$+cc).

T o give a suggestive form, we interpret them in a two-fluid hydrodynamic model. There is a quantum fluid of superconducting electrons and a quantum wavefunction +(r) associated with it. T h e number of superconducting electrons is
ndr) =

$(y>

12,

the current is from equation (10)

j = 2en,v,(r)
where o, is the superfluid velocity m
C

T h e free energy (equation (8)) has a part which depends only on the density n,(r) and another one connected with the spatial variation of the order parameter which can be written

1 1 - n, mo,2+ -(hO
2

2m

I + 1)Q.

(14)

T h e first term is the kinetic energy of the current and the second term prevents any rapid variations in the superfluid density. T o the coupled set of equations (9) and (10) we must add the boundary conditions. T h e most general linear boundary condition on ensuring that no current flows through the boundary surface is of

Ginxburg-Landau theory for superconductors


the form

115

where the subscript n designates the component normal to the surface and 6 has dimensions of length. I n their original paper Landau and Ginzburg assume

From the microscopic theory, one can show that equation (16) is correct only for a superconductor-insulator interface. T h e most general condition (15) must be used for a superconductor-normal metal interface (de Gennes 1964b, Zaitsev 1965).

3.3. The two fundamental lengths T h e set of equations (9) and (10) permits one to define two fundamental lengths. T h e first characteristic length is the penetration depth for a very weak field which London introduced in his equation. I n zero field I$I is everywhere equal to the value #o
$b02

-- > 0.

This value is not modified by the weak field.* Then we have

4e2 --l$o[2A. mc

This is the form of the London equation which leads to a penetration depth (cf equation ( 5 ) )

(17)
T h e above expression is similar to Londons one if n, = 4i,ho2. For T+ T,, $02 is going to zero so X+m. T h e other characteristic length is obtained by considering, for example, a onedimensional situation with no magnetic field. We obtain

K2 d2$ +a*+p*3 2m dx2


This equation defines a length

= 0.

This we call the temperature-dependent coherence length. I t defines the distance over which the order parameter can vary. Both lengths become large simultaneously as T+T,. This is a justification for the slow variation approximation which was used to derive the equation. Notice that the ratio of the two lengths
- = K

is temperature independent. We call the ratio

the Ginzburg-Landau parameter.

116

M Cyrot

3.4. Use of the Ginzburg-Landau equations


There are two unknown parameters a and /3 in the Ginzburg-Landau theory. They can easily be replaced by quantities accessible to measurements. Firstly, in the absence of an external field the free energy difference between the two phases is given by

Recalling that this difference equals the magnetic energy (equation ( 1 ) ), we have
4rra2

Hc2 =

P*

Next the penetration depth is given by

So we can eliminate the unknown parameter in favour of measurable quantities. T h e dimensionless parameter K is given by

Only the difference of K from one material to another prevents the equations from scaling perfectly into a single law of corresponding states valid for all superconductors. Indeed let us introduce a system of dimensionless quantities

=-

4rrF HG2

T h e dimensionless free energy is given by

F;

= F+q-l+12+~l+/4+l

(GV-A) 1

#I

+Arz.

Dimensionless units correspond to measuring lengths in units of A, magnetic fields in units of HcJ2, the vector potential in units of HcAJ2, the order parameter in units of $,, and the energy density in units of Hc2/4rr. These reduced units are

Ginxburg-Landau theory for superconductors

117

often used in applications (Saint-James et a1 1968) but we will not use them in the following. 3.5. Applications It is not our purpose to develop in detail applications of the Ginzburg-Landau equations. A whole volume will not be sufficient and excellent books (de Gennes 1966, Saint-James et al 1968) and review articles (Parks 1969) exist on the subject. We just wish to suggest the kind of problems which can be handled. Once the parameter K is determined by specifying H, and X no further physical knowledge is needed before solving a purely mathematical problem. I n fact the majority of the most interesting cases were solved before the advent of the microscopic theory. Due to the origin of the theory, these problems are of a thermodynamic nature. T h e first one is calculation of the critical field in a variety of situations, that is, the magnetic field below which nucleation of superconducting regions can occur in the bulk (Abrikosov 1957), on a surface (Saint-James and de Gennes 1963), in thin films or in sandwiches of normal and superconducting films (Parks 1969). We give an example in $3.5.1. T h e second type of application that we can mention is calculation of critical currents. T h e existence of a superfluid velocity reduces the amplitude of the order parameter. I n fact, assuming constant, equation (8) gives
a+p1$12+$m322.'2=

where is the superfluid velocity so I is reduced compared to its value in the absence of o. From equation (12) one can deduce the superfluid current and the critical value at which the transition occurs. I n both applications the presence of free surfaces or boundaries between normal and superconducting metal is taken into account by the boundary conditions, equations (15) and (16). These boundary conditions also enable one to handle the Josephson effect (de Gennes 1966). We notice that no transport properties can be handled using this simple approach. I t does not describe the normal excitations of the superfluids so we are unable to calculate properties which depend on these excitations as for instance thermal conductivity, nuclear relaxation, ultrasonic attenuation, etc. However, it does not mean that the Ginzburg-Landau equations are of no use for these properties. O n the contrary, the Ginzburg-Landau solution can be put in the microscopic theory so as to be able to solve the coupled equations for the excitations. Examples will be given in the next section. We now turn to the most famous application of the Ginzburg-Landau theory. 3.5.1. Type I1 superconductors. As we mentioned earlier, the theory was constructed, in particular, to explain the positive surface energy of the usual superconductors (now known as type I). However for K 2 & the theory predicts a negative surface energy and it is energetically favourable to make the flux enter the specimen, Abrikosov (1957) has studied this possibility in detail. T h e importance of this study stems from the fact that most superconductors are type I1 superconductors, in particular all high field or high critical current superconductors important for practical applications. Let us consider a superconducting metal in a sufficiently high field for superconductivity to be destroyed. If we.decrease the field, at a certain field H = HC2

#I2

118

M Cyrot

superconducting regions begin to nucleate. H,, does not coincide with the thermodynamical field H,. I n the region of nucleation the order parameter is small. T h e Ginzburg-Landau equation can be linearized to give

This equation is formally identical to a Schrodinger equation for a particle of charge 2e and mass m in a uniform magnetic field. I n an infinite medium the lowest eigenvalue is @ U , = eH/mc. I t corresponds to the highest field where superconductivity can occur
---a: =

eAH,, -

mc

Recalling the definition of H, (equation (19) ) we have

H,,

H,J2.

When K > J3, H,, He, an ordered phase ($# 0) will appear in the sample. This > phase cannot correspond to a complete exclusion of flux which is energetically unfavourable at H > H, and for fields greater than He we obtain the mixed state described by Abrikosov. T h e detailed study of this phase needs the nonlinear term of the Ginzburg-Landau equation in order to calculate the amplitude of $. The solution for the order parameter will be a linear superposition of the degenerate solutions of equation (22). Wre expect that I +I has a periodic structure in x and y. Abrikosov chose the form:
n

I n order that $ should be periodic, it is necessary to impose a condition on the Cn. For instance, the condition would be for a square lattice:

c n

Cn+,,

T o determine the normalization coefficient C we follow the method given by de Gennes (1966). T h e free energy must be stationary if the order parameter $ is changed to (1 + E) $. T o first order in E the variation in the free energy is

T h e stationary condition is d F = 0.

In the equation A is the vector potential of the field which exists in the sample. It is different from the vector potential A, which exists in the presence of the field H,, because (i) we assume that the external field H, is slightly smaller than H,, (ii) there exist supercurrents which also contribute to the field. Let and

A,+A,.

Assuming that A, is small we can expand to first order in A , using the fact that $ satisfies the linear Ginzburg-Landau equation for A, and the stationary condition

Ginzburg-Landau theory f o r superconductors


then becomes

119

where

J, represents the current associated with the unperturbed solution. Integration by parts gives 1 ~d3v(Pl+(r)~4---h,.ho) = 0.

4%h, and h, are the magnetic fields corresponding to the vector potential A, and the current J, respectively. By definition, we have h1 = Ha - H,,

+ ho.

Now we use the following theorem: if is the eigenfunction for the lowest eigenvalue of equation (22),the lines of current associated with J, coincide with the lines I+/ = Ct. This is a simple consequence of the fact that, if

equation (22) can be written as

rI+rIwhere II=
=

IIz 5 iny. T h e eigenfunction for the lowest eigenvalue is such that


U-#
=

0.

This can be rewritten

ek a Joy +- - / + I 2 = m ax
or equivalently

The stationary condition becomes

This solves our problem because only the averages of / + I z and /+I4 occur. T h e ratio of these two averages is only a function of the lattice. For instance, one can compute for a square lattice the ratio of the average of I +I4 to the square of the average of I l2 to be

120

M Cyrot

Thus, from the stationary condition we obtain the average of

I# I z

This gives the value of C in a straightforward manner. We are, in fact, mainly interested by the thermodynamic properties of this phase. We first compute B by the average of the magnetic field.

= -/d3r(Ha+ho(r)) V

= Ha-- mc [l#12d3r.

4re%

T h e free energy F, which is a function of B, is given by

Here we have introduced K given by equations (19), (20) and (21). From F we can deduce the magnetic field and all the thermodynamic properties. For instance,

H = 477'so the magnetization is

dF dB

We now quote some consequences of Abrikosov's calculation which are all in very good agreement with experiments. (i) T h e magnetization increases linearly with the magnetic field, (ii) T h e most favourable lattice will minimize PA. I n fact numerical computation of PA shows that the triangular lattice is more stable because ,BA = 1.16. (iii) I n the Ginzburg-Landau regime the slope of the magnetization is given as a function of K which can be determined independently through the ratio H,,/H,. (iv) T h e magnetic field inside the superconductor varies periodically.

3.5.2. Macroscopic quantum phenomena. We now turn to some fundamental consequences of the existence of a complex order parameter. It will throw light on the complete significance of the phase of the order parameter and on the second Ginzburg-Landau equation for the current (equation (10)). Since $ can be thought of as a kind of wavefunction for the superconducting electrons, superconductivity is a macroscopic quantum state. This leads to macroscopic quantum phenomena that we describe now and which are consequences of the extension of the wavefunction over macroscopic distances. T h e first unusual effect arises naturally from the topology of the macroscopic quantum state. From equations (12) and (13) the phase gradient is given by

Integrating any gradient on a closed contour gives zero for a singly connected path. However, if we allow multiply connected regions, we just require that the wavefunction # is a single-valued function. Thus the integral of the gradient can be a multiple of 277'. Consider a hollow superconducting cylinder. As from the London

Ginxburg-Landau theory for superconductors

121

equations currents are restricted in the region of thickness X near the surface, one can choose a circuit distant enough from the surface so that j = 0. We have

2n-n = $/:.dl

= -# 2e
C7i

where n is an integer and # is the magnetic flux inside the cylinder. Thus the flux is quantized in units of CO = clz/2e. This remarkable consequence of the existence of a macroscopic quantum state was first predicted by London (1950) and confirmed experimentally later on. Further examination of this quantum state needs some form of quantum phase detector to probe the wave nature of the system. This is realized in Josephson functions. We now describe properties of weakly connected superconductors. We have mentioned that the surface can be taken into account through boundary conditions, equations (15) and (16), when no current flows through the boundary surface. We now consider a situation where two pieces of superconductor are coupled by some means which permits electrons to flow from one piece to the other. For instance, we intercalate a thin insulating barrier, approximately 20 A thick, in order that the electrons can tunnel from one side to the other. T h e boundary conditions must be changed in that case. T h e value of the order parameter and its derivative at the surface on one side of the barrier will not be independent of their value at the surface on the other side. We can expect a more general relationship (de Gennes 1964b).

where + and - represent the two planes of the insulating barrier which are perpendicular to the x axis. We do not introduce higher order derivatives because they can be expressed in terms of the zeroth and the first derivatives by the GinzburgLandau equation. T h e four unknown coefficients are not independent but are related to each other by one relation expressing current conservation through the boundary. T h e supercurrent crossing a unit area of the wall is given by

T h e current conservation gives

Mll Mzz - Ml2 MZl


and the current I, is

I,

- 2e7i = -(*?*+-cc)

mM12

where # is the phase of the order parameter. Therefore a supercurrent must flow through the junction (Josephson 1962). If the two pieces of superconducting materials were related by a superconducting junction the phase would be identical

122

M Cyrot

i n the two pieces. However, if there is a weak link between them we expect a correlation between them. T h e phase difference is very sensitive to the presence of currents and magnetic fields and nothing can be said a priori. If no magnetic field is present the phase difference is fixed by the current which flows in the junction. It follows immediately that there exists a critical current given by

2eK
IC

I +-I I + I .

Higher current cannot flow without a voltage across the barrier. For a symmetric junction, if we assume that the absolute values of the order parameter are only slightly modified by the presence of the superconducting current (typically small and of order 1 A cm-,), we find that I,cc T,- T as the order parameter varies as (T,- T)''z. This must be contrasted with the (T,- T)3'2 dependence of the critical current in a bulk superconductor. One also expects that M,, varies inversely with the transmission coefficient of the barrier for the electrons. Microscopic calculation of this coefficient gives (Josephson 1965)

M I , goes to infinity as the transmission coefficient vanishes. T o make the fundamental importance of the phase of the order parameter more sensitive, we describe what can be called a superconducting interferometer. An interferometer is an instrument for comparing phases at two points. Implicit is the assumption that the wave is coherent. We discuss the maximum supercurrent of two Josephson junctions in parallel as a function of the applied magnetic field. T h e overall supercurrent is the sum of the two currents flowing in the two junctions

Imax{sin($! - $5) sin (4; - $?)}.

There exists a relationship between the phases due to the second Ginzburg-Landau equation. We consider a closed circuit in the loop formed by the two junctions. As the current flows on a sheet X near the surface, in the bulk of the superconductor j = 0. Integrating the gradient of the phase over the circuit, we obtain

where is the total flux in the loop which is not quantized as the superconducting is the quantum of flux. Thus we obtain loop is broken by the two junctions and
# J ~

T h e maximum current is a periodic function of $ or H. T h e experiment has been successfully performed by Jaklevic et a1 (1964). This is a striking consquence of the quantum nature of the order parameter.

4. Microscopic derivation of the Ginzburg-Landau equations


We first present the simplified hamiltonian used in the microscopic theory of superconductivity. It takes into account only terms which are basic to the theory and neglects terms which are only occasionally of interest. For instance we

Ginzburg-Landau theory for superconductors

123

completely neglect strong-coupling effects and the BCS model hamiltonian that we use cannot describe such effects. Then we derive a free energy functional in some generality and consider the Ginzburg-Landau one and all the other extensions that we will describe in $ 5 as special cases. T h e derivation of the Ginzburg-Landau theory is just an expansion in powers of the order parameter valid near T,. This microscopic derivation finds its interest, firstly, in the justification of the phenomenological approach and in the interpretation of the order parameter and, secondly, i n all the applications which mix the phenomenological approach and the microscopic theory.

4.1. The BCS hamiltonian T h e basic interaction responsible for superconductivity is an attractive interaction between electrons by means of an exchange of virtual phonons. T h e discovery of the isotope effect (or the change of the critical temperature with isotope) had clearly indicated that the interaction must be one between the electrons and the lattice. Cooper (1956) showed that in the presence of an attractive interaction, no matter how weak, the Fermi sea becomes unstable. Two electrons of opposite momentum near the Fermi surface tend to pair in a bound state of lower energy. One year later, Bardeen et a1 (1957), using this idea, were able in a simple model for the interaction to describe the principal features of superconductivity. They calculated only the energy difference between the normal and the superconducting eV atom-I) compared to the Fermi energy state which is very small ( HO2/8.rr (10-20 eV). Out of all the matrix elements of the interaction between electrons, the only one contributing to this difference is the term describing the transition of a pair (k?,- A l ) to the state ( I t , -Il)
N

All the other terms will only renormalize the normal ground state which cannot be known with the desired precision needed for the energy difference, so, in this respect, they are irrelevant. T h e attempts at a detailed calculation of the interaction have shown that the interaction can be attractive only near the Fermi surface. Bardeen et al made the further simplifying assumption that V,, is isotropic and constant for all electrons in a narrow shell straddling the Fermi surface of thickness, in units of energy, less than the average energy of the lattice and that V,, vanishes elsewhere. Measuring electron energy from the Fermi surface and calling tkthe energy of an electron in state k, one can state this formally by the equations

is the Debye frequency of the material and Vis positive so that superconductivity is possible. T h e hamiltonian of our system, written for convenience in the direct space, is
U,,

124

M Cyrot

where #,(r) is the creation operator for an electron of spin B at point Y. T h e first term is the normal state energy and the second one must be interpreted with equation (24). U ( r ) describes the potential due to the impurities. 4.2. The general free energy functional for a superconductor The usual derivation of the Ginzburg-Landau free energy leads to the GinzburgLandau equations. From these equations it is possible to construct a free energy functional whose stationary variations with respect to A, A* and A reproduce the equations. We will proceed here in a different way. We first derive a free energy functional for a superconductor (Cyrot 1970) from which we derive as a special case first the Ginzburg-Landau free energy and later its generalizations. We calculate the grand partition function (26) where H is given by equation (25). 2 cannot be easily calculated because of the interaction term, so we employ a mere trick using the identity valid for any bound operator exp 1 a l2
= --/d2r

2 = T r exp -,8(H- p V ) )

exp { - (x*+ax* +a%)}.

The partition function can be rewritten


2 = cJSA(r)exp (-~[d3rlA(r)12) Trexp(-PH,,)

with

T h e main interest of this transformation is to eliminate the two-body potential. We have to calculate a fairly simple partition function for noninteracting particles moving in an arbitrary pairing potential A(r). The integral is a functional integral which means that we have to integrate over all functions A(r). c is a normalization constant. I n this derivation we have overlooked some difficulties which are mainly due to the fact that the two terms of the hamiltonian do not commute so equation (27) is not an exact result. T h e approximation done is of the same nature as a Hartree-Fock approximation (Langer 1964). We have now to calculate the trace of the effective hamiltonian (equation 28). This hamiltonian is quadratic so we diagonalize it by using a unitary transformation
# n ( ~= )

Z ~n 7 Un(Y) n

ti ~ ; E ( Y ) .

(29)

This diagonalization is discussed by de Gennes (1966) so we will only quote the different steps of the calculation. The transformation (29) must diagonalize the hamiltonian, that is -IHe, = E g + E en nu Ynu (30)
nu

Ginzburg-Landau theory for superconductors

125

where Eg is the ground state energy of He,, and E , is the energy of the excitation n created by the operator y:. Thus we must have

nul = - E n Ynu + [ H e r , YLI = E n Ynv.


[ffeff,

These conditions fix U, and v,. We obtain the following set of equations for and v, known as the Bogolubov equations
EU(Y) =

U,

(- ikV + $1

+ U ( Y-E , )
U ( r )- E , v ( r ) + AL(r) u(r).

(ihV + \2m It is then easy to show that


E ~ ( Y ) =

$1 +

d 3 r I A ( r ) 1 2n ~ l n { 1 + e x p ( - P + en)))

(32)

where E , are the eigenvalues of the Bogolubov equations. T h e argument of the exponential is an effective free energy for the pair potential A(r)

F ( A ( r ) )= 1 -I A ( r ) 1 2 d 3 r1 p ~ l n { 1 + e x p ( - P ~ n ) } . -

vj

(33)

This functional for the free energy is valid at all temperatures and for any function A ( r ) without restriction. We would mention that it is equivalent to the free energy functional derived by Eilenberger (1965). H e shows that Gorkovs original Greens function equations for superconductors follow from it. I n our derivation Bogolubov equations play the same role as the Gorkov one in the Greens function formalism. Let us now minimize the functional with respect to A*(r)

Differentiating the equation for U, and v, and taking into account that if E , is an eigenvalue and U, and v, are the corresponding eigenfunctions, - E , is also an eigenvalue corresponding to vg and -U;, we obtain

A(r)

v~~n(~)~~(~){1-22f(~n)>. n

(34a)

This equation corresponds to a self-consistent equation for the order parameter. Using equation (29) we have (34b) This is just the definition we gave in the first part for the order parameter. T h e results of this section are quite general. If A is assumed to be constant we obtain the usual BCS theory. We quote the BCS result which is very easy to obtain and which will be useful for the following. Let w,(Y) be the wavefunction in the normal metal

4 ~ ) v(4~) +n(Y>>* =1( Y

126

M Cyrot

The solution of the Bogolubov equations are

with

where

E,

is the energy of the excitation n and is given by


Em = ( f n 2

(37) T h e self-consistent equation for A which represents a gap in the density of states is given by the BCS equation

+ A*)li?

where N ( 0 ) is the density of states at the Fermi energy. Figure 6 gives the energy gap A as a function of temperature.

TIT,

Figure 6. The energy gap as a function of temperature for a BCS superconductor.

4.3. Derivation of the Ginxburg-Landau equations I n $ 3 we gave the Ginzburg-Landau equations from a postulated form of the free energy. They are based on an expansion in powers of the order parameter and on the assumption of slow variation. We want now to derive them from the microscopic theory that we have just stated. That is, we solve the Bogolubov equations and apply equations (33) and (34). The order parameter is treated as a perturbation. We follow closely the simplified version of this calculation given by de Gennes (1966) and we will calculate directly A and not the free energy. 4.3.1, The linear Ginxburg-Landau equation. We expand in powers of A the solution of the Bogolubov equations. T h e zero-order solutions are

where

4%are the eigenfunctions

of the hamiltonian in the normal state

Ginnburg-Landau theory for superconductors


I n order to determine the first-order corrections U, and v, we set
U ,= Eenm+m m
U,

127

= Ednm+m
m

T h e Bogolubov equations are then written

I I - em) enm = J+g(r) A(Y> vi(^) d3r


en

( I t n I + t m ) dnm

= J+m(r)

A*(Y> i ( ~d3r. u )

Inserting these results into equations (34) we obtain to first order in

(41)
where

This can be written using equation (39)

nm

It is sometimes useful to transform this result using the well known formula
(431
WV

where Rw, = 27rkB T(v+ 4. T h e sum is over all positive and negative integers v )

wy

mn

We have to calculate the matrix element between states $2 and 4;. These are not the usual ones, so we introduce, following de Gennes (1966), an operator K which, by definition, changes a function into its complex conjugate. So we have

4 3 4 +t(r>= ( n I a(r - r ) KI m>


where ( n ... m ) is a matrix element between wavefunctions of one electron in the normal state. We introduce the function

I I

and its Fourier transform

S(r, s, t ) =
=

dQ S(r, s, R)exp ( - iRt) 27r

(8(r(O) - r ) K ( 0 )K+(t) 8(r(t)- s ) ) ~ * .

(45)

128

M Cyrot

We have assumed that S(Y, C l ) is only slightly dependent on g so the bracket s, ' means an average over all the one-electron states at the Fermi energy. The function S(Y,t) has a simple physical meaning. It is, apart from a phase factor, the probas, bility in the normal state that an electron at the Fermi level, being at point r at time zero, arrives at time t at point s. This is a function which depends only on the properties of the normal metal. T h e kernel of equation (41) is simply related to this function by

K ( s , r ) = 2.rrkBTN(0) V c
WY

dtexp(-21wVlt)S(r,s,t).

(46)

T o calculate S we remark first that in a superconductor all the characteristic lengths are greater than the Fermi wavelength, so we can treat K as a classical operator. dK _ - i[H, K] = i{A . p + p .A) dt we neglect the noncommutation and obtain

T h e integral is along a trajectory of an electron in the normal metal between time zero and t. U p to now, we have only made use of one of the two assumptions of Ginzburg and Landau. We just assumed A to be small. We now assume the slow variation of A and the vector potential. I n fact this means that the characteristic lengths over which they vary are small compared to the range of the kernel K. T h e spatial variation of A can be neglected in equation (47) and we have

Using the two identities A(Y)= exp {(r - s) , V} A($) and

we obtain

T h e slow variation assumption permits us to develop the exponential in powers of ( r -s). (V - 2ieA/cK). So we obtain to second order

(48)
T h e zero-order term Q cannot be obtained directly as it diverges. We have to go back to equation (42), assume constant A and no magnetic field. We obtain

Ginxburg-Landau theory for superconductors

129

AwD

Recalling that the interaction V is different from zero only for energies within of the Fermi level we have 1.14kwD
kBT
*

(49)

As the slow-variation approximation is only valid near T, which is defined by

1 = N ( 0 )Vlnwe have

1.146~~
kB

T,

Q
T h e coefficient L is given by

T-T l+N(O)V+.
JC

nkB T N ( 0 )V

soa

dt
wy

exp ( - 21 wyl t) (R2(t)),,.

One can explicitely compute S(r,s,t) and then the average of R2. Here we will proceed directly. T h e simplest way to describe a normal metal is to introduce a relaxation time T so that

Therefore we obtain

L is now given by

Following Gor'kov, we introduce the function

which brings in the effect of impurity scattering. For example, in .the limit of infinite mean free path x(p) = x(0) = 1 while in the dirty limit where the electronic mean free path is much shorter than the coherence length [ ( O ) , we have

4.3.2. Addition of the nonlinear term. If we expand to third order the self-consistent equation (34), we get
A(s) = / K ( s , Y) A(r) d3r+ R(s,r, I, m) A*(r) A(2) A(m)d3r d31d3m.

T h e kernel R can be calculated exactly but thanks to the slow variation approximation of A near T, we can replace the last term by

R I4

s )

l2 A($>

130

M Cyrot

with R = JR(s, I, m) d3r d31d3m. T o calculate R we can take A independently and Y, we have just to expand the BCS result to third order in A. We obtain

R=

16x2kB2 T 2

This coefficient is independent of the mean free path for the same reason as holds for Q. The first Ginzburg-Landau equation valid for all mean free paths near T, is then

1--

I )

1 7iv 27((3) R A(Y)-- - 6(77k<c) 8 '(P,k,c7)

( - iRV - =)'A(r)
C

T h e second Ginzburg-Landau equation for the current can be obtained in a similar way using the expression for the current

j ( r ) = Re
T h e result is

(( ( - iRV -g , $+(Y)
C

As is seen from equation (48) the spatial variation of the order parameter takes place on distance given by

which diverges as T-tG while the range of the kernel K(r,s) which is given by Ll'z remains finite as T tends to T,. Our assumption about the slow variation is then verified near From equation (58) we can calculate the penetration depth

e.

where hL(0) is the London's penetration depth. X(T) also diverges at T,. Before ending this microscopic derivation, we point out that the correspondence with the phenomenological Ginzburg-Landau equation becomes exact upon making the identification

4.4. Applications We wish to discuss the applications which follow from the mixture of the phenomenological approach and the microscopic theory. They are in fact limited as we will see in the following. In a direct microscopic approach we have to solve the two coupled equations (3 1) with an unknown pair potential A(r) and determine it using equation (34). It is obvious that this is impossible in the general case. I n the BCS case we were able to solve these equations because we assumed a

Ginzburg-Landau theory f o r superconductors

13I

constant pair potential and overlooked some difficulties in the case where U ( r )# 0, that is, when there are impurities (Caroli et a 1963). Thus, in fact, there Z are two possibilities for solving equation (31); either we guess A ( r ) and after solving these equations justify our guess by the self-consistent equation (34), or we know A(r) a priori by other methods and have only to solve equation (31). How is it possible to know A(Y)a priori ? Just by using the Ginzburg-Landau equation when possible. This permits a large number of applications because when equation (31) is solved we know the complete excitation spectrum of the superconductors and can then calculate transport properties. As a first example we can give the direct calculation of Caroli et aZ(l964) using equation (31) of the excitation spectrum in the core of a vortex. They restricted their calculation to a pure superconductor and take for A the Abrikosov (1957) solution. We can then quote the calculation of the density of states of a dirty type I1 superconductor in the mixed state (de Gennes 1964) or the same calculation for the clean Z case (Brandt et a 1967). More generally one can say that any calculation of properties of the mixed states relies on the assumed Abrikosov solution for the mixed state (equation 23).

5. Validity and breakdown of the Landau theory


The original formulation of the BCS theory seemed to make superconductivity a special class of phenomena, with pairing introduced in a somewhat ad hoc manner. T h e Gorkov reformulation (or the Bogolubov one) of the BCS theory made it possible to link the superconductive phase transition with other phase transitions. This theory introduces a new collective degree of freedom into the problem, t h e pair potential which is treated in a self-consistent manner. This potential influences the individual electrons, which in turn determine the strength of the collective pair potential. This is similar to a Hartree-Fock method for determining atomic structure and to the Weiss theory of the molecular field for ferromagnets. Interest is turned from the individual particles to the collective field. All of these methods are mathematically similar and called self-consistent field theories. However, these theories are known to be incorrect in many cases because they do not treat thermodynamic fluctuations. I n spite of the absence of magnetization in a ferromagnet above the Curie temperature, fluctuations away from zero magnetization are found. The occurrence of fluctuations is a universal phenomenon regardless of any phase transition, Thus we want to comment on the fact that the Ginzburg-Landau theory has been so successful, although it is an example of a mean field theory. T h e basic reason why fluctuations are more difficult to detect in superconductors is the fact , that the coherence length is large compared to k. This tends to average the selfconsistent field and suppress the fluctuations. However, one can calculate the criterion for the validity of the Landau approach. We first do this and show that, nevertheless, in some particular cases one can observe the breakdown of this approach. We finally describe some examples. 5.1. The Ginxburg criterion Let us go back to our microscopic derivation of a free energy functional. We have in fact to calculate the partition function (equation 32) which is given as a functional integral. However, instead of computing this integral we have taken the

132

M Cyrot

value of $ which minimizes the argument of the exponential under given external conditions. Fluctuations from this value were assumed to be negligible and we did not calculate the total partition function by integrating over all the function $(r). We now calculate the mean square of these fluctuations and see when they are in fact negligible (Ginzburg 1960). We have
Z = exp(-PF)
= jd$(r)exp(-p[F($)d3r)

with

where we have assumed $ to be small as we have restricted ourselves to a temperature close to T,. For a spatially uniform situation the average order parameter is given by

Retaining terms quadratic in! , I J

- $o, one obtains

where k is the wavenumber of the fluctuation. Thus we have

T h e critical temperature interval at which fluctuations become important may be obtained by setting the left hand side equal to unity. I n the calculation of the right hand side there is a difficulty due to the divergence of the integral for large k vectors so we have to introduce a cut-off. Mathematically this stems from the slow variation approximation: we cannot, using the Landau free energy, calculate fluctuations of the wavevector larger than R-l, the inverse of the range of the kernel K(Y- s) (equation 41). T h e criterion calculated with this cut-off (R= to a clean material and (fOZ)'/% for a dirty material) is appropriate for for the breakdown of the microscopic mean field theory (Hohenberg 1968). However, this breakdown just involves a shift of T, and other parameters in the theory. Since these quantities are left arbitrary in the phenomenological approach, this criterion is not appropriate for the breakdown of the Landau theory. Ginzburg used as an effective cut-off &,-I. We see that fluctuations over the range tT-I< k < R-I are incorporated into a change of the coefficients of the theory and are therefore not considered 'critical' in the Landau theory. In any experiment the transition temperature is usually taken to be the point at which $vanishes, so that the Ginzburg criterion is the appropriate one. We would have, for instance, deviation of the critical exponents from their classical values. Using the Ginzburg cut-off we obtain

Ginzburg-Landau theory f o r superconductors


for a clean superconductor and

133

for a dirty superconductor. T h e breakdown of the microscopic mean field theory would have been obtained for I T - T,l/T, = ( k , and ( k , T/EF)'12 k , ( for clean and dirty superconductors respectively. I n practice this breakdown is not observable for the reason given above. It is clear that due to the large factor E,/k, T, lo4, it is not possible to observe fluctuations. Short-range order effects are very small: this is the fundamental reason why the simple Landau assumption works for most superconductors while it does not in other cooperative systems. It must, however, be pointed out that in samples of lower dimensionality it is possible to observe the breakdown of the Ginzburg-Landau theory. If we consider a thin film of thickness d small compared to f so that the order parameter cannot vary in the thickness we obtain in clean materials
N

--

T-T, T,

E, 1 kgT,k~~Pd

and in dirty materials

Fluctuations are thus observable in one or two dimensions.

5.2. Fluctuations above T, As we concluded that it must be possible to observe fluctuations, that is, the breakdown of the Ginzburg-Landau theory in a two-dimensional system, we shall calculate now some effects linked with these fluctuations. Just above T, superconducting regions grow and decay and they must have an effect on the conductivity above T,. This phenomenon was called paraconductivity by Ferrell (1969) in a loose analogy with the behaviour of a ferromagnet in the paramagnetic range above the Curie temperature. As the superconducting region is approached from above the resistivity drops continuously to zero, that is, upon cooling one observes an increase in the conductivity from the normal value before the critical temperature T,. This was first observed by Glover (1967) and studied theoretically by Schmidt (1968) and Aslamazov and Larkin (1968). We shall give here a brief derivation of the result following Schmidt without entering into some difficulties which appear in the clean case (see, for instance, Takayama and Maki 1971). If #q is the Fourier transform of waveform of wavevector q, the free energy associated is

where we have neglected the nonlinear term and the current is

134
T h e conductivity
U

M Cyrot
is given by the Kubo formula

where the brackets mean a statistical average. a(q) is the product of four # but as the fluctuations of wavevector q are statistically independent, we can separate it into the product of two averages of two #. One has

T h e first factor is the Ornstein-Zernicke factor where (-2 = 2m 101 17 and r k /2 is the relaxation time of fluctuations of wavevector k. T h e usual assumption is that 4approaches the equilibrium value corresponding to a minimum of the free energy F at a rate governed by the slope of F in the functional space

so we obtain
r k

m T y ZY = 71, k + f - 2 '

T h e final result for ~ ( 0is)

~ ( 0= 4k Tye2)
for a three-dimensional system. One can now introduce the effect of a finite sample by restricting the allowed values of k to those allowed by the boundary conditions o n the order parameter at the surface, namely that the normal derivative should be zero. For those directions in which the dimensions are less than ( ( T ) , one thus only obtains the contribution from k = 0. T h e sum over k thus gives

From the microscopic study of the time-dependent Ginzburg-Landau equation we will find (equation 79)

thus
U =-

e2 T, 1675d(T-T,)'

This result has been obtained microscopically by Aslamozov and Larkin (1968) neglecting the scattering of the electrons by the fluctuations. This scattering can be important in clean materials. One should now mention that fluctuations have also an effect on the other properties like diamagnetic susceptibility; an effect which has been observed (Gollub et al 1969). So our conclusion will be that, in some special cases, one can predict and observe the breakdown of the Ginzburg-Landau approach but, in practice, it is of very limited consequence.

Ginxburg-Landau theory f o r superconductors

135

6. Extension of the Ginzburg-Landau equations 6.1. Preliminary remarks


We have seen the considerable interest in having a closed equation for A. I n addition, the simplicity of the Ginzburg-Landau equations makes the problem of generalizing the theory of primary importance. T h e Gorkov derivation relies heavily, of course, on the assumption that T is just below T,. This permits two different kinds of expansion: the expansion in powers of A / k B T , since A is small and in powers of (V-2ieAlcK) since A and H are slow~lyvarying. We shall see whether the two approximations have a certain range of validity at lower temperature and if one of the two approximations can be relaxed without losing all simplicity. We shall discuss, successively, the two approaches :expansion in powers of the order parameter and expansion in powers of gradients without restriction on the magnitude of A. However, we want first to point out that there is a third assumption which is not apparent but which is much more basic and which cannot be relaxed in any generalization without losing all the simplicity of a Ginzburg-Landau approach: the superconductor must obey a local electrodynamic. This is clearly seen from the second Ginzburg-Landau equation. T h e current at one point is related to the vector potential at the same point. I n the general case the current j ( r ) at one point depends on the vector potential A ( r ) at all neighbouring points r f such that 1 Y - r f I < Eo in the clean case. Thus when the electrodynamic is nonlocal (Pippard electrodynamic), that is, when the distance over which A varies, A( T ) , is smaller than Eo, the general equations for a superconductor cannot be reduced to a simple Ginzburg-Landau form. This prevents, for instance, any generalization to type I superconductors (A < () at low temperature. T h e two other assumptions have, however, not the same physical significance and we will see in what limits one can relax one of them. It turns out that one attempt has been far more successful than the other. This is the assumption that A is small with no assumption on the variation of A. This situation is realized in type I1 superconductors just below H,,. At the critical field H,, the transition is second order, so below H,, A is small. I n the dirty limit, 1 < to,the linearized Ginzburg-Landau equation holds at all temperatures. Thus many results derived near T,are valid down to zero temperature in fields near Hca. T h e second attempt which keeps only the slow variation approximation is unfortunately restricted to small deviations from the BCS state. For instance, it is not able to describe the mixed state near H,, except near T, where A is small. 6.2. Small order parameter T h e more important and successful attempt to generalize the Landau theory follows this line. T h e order parameter is small near a second-order transition to the normal state or near the transition at H,, for type I1 superconductors. However, this alone cannot give the necessary simplification and we must assume a local electrodynamic. An interesting situation was pointed out by de Gennes (1964) and Maki (1964): the transition at H,, of a dirty type I1 superconductor. They studied the region just below the transition at H,, but at all temperatures. They both restricted themselves to the case of a very short mean free path 1< to.At first sight this could be generalized to an arbitrary mean free path just below the transition if the electrodynamic is local. However, one should mention that in the pure limit, where the electronic mean free path is large (ZBEo), there exists a difficulty if one
6

136

M Cyrot

makes an expansion in powers of the order parameter. It was first noted by Maki and Tsuzuki (1965) that the expansion for the free energy in powers of A breaks down at low temperatures in the gapless region. Cyrot and Maki (1967)have shown that in general it is not possible to expand in powers of A the correlation functions and the Orsay Group (1966) found a logarithmic divergence in the density of states for low excitation energies. Thus, due to this complication, we will only consider in the dirty case I< to the following.

6.2.1. The linear term. We go back to the linear integral equation (equation 41) where the kernel is given in full generality if one carries out an expansion in powers of A, by equations (45) and (46). T h e procedure used by de Gennes is to find out what differential equation is satisfied by S(r,s, t ) . This function represents the probability, calculated in the normal state, that an electron being at Y at time zero will be at s at time t. There is also a phase factor given by equation (47). For this purpose one considers two successive times t and t + E ( E being small). T h e probability that the particle is at point r1 at time t + E is given by the probability that being at r' at time t it arrives at r1 by diffusion during the time E. The average ' distance between Y and r1 is small and thus the corresponding increment in phase A+ of K is 2e A$ = - A ( r l ) . ( r l - r ' ) .
C7i

Thus

S(r1,y2, t

E)

= /d3r'

S(Y', t ) G(rl,Y ' , yZ,

E)

exp (iA4)

where G is the Green's function describing the diffusion from a point source and is given by

D is the diffusion coefficient, D = &vF1. Expanding S(r', y 2 , t ) in a Taylor series around r' = Y and retaining all terms up to first order in E , we obtain the required equation

We can now explicitly write S in terms of the eigenfunctions g,(y) defined by

Taking into account that for t+O,

S reduces to a delta function, we have

S(r,s,t ) =

Ginxburg-Landau theory f o r superconductors

137

We insert this explicit form into equation (46) and perform the integration over t. T h e linear integral equation for A becomes

A(s)

N(0) V2nkB T

2
WV

1 4s). 2h I U , I - %D(V- 2 i e A / ~ h ) ~

(64)

T h e eigenfunctions of this equation for A are simply theg. If we choose A(Y)= g,(r), this equation is satisfied provided that

T h e sum over U , diverges. This is because we did not take into account the frequency cut-off of the interaction. So we subtract equation (49) which we rewrite

WV

Transforming l/ni7(0)V we obtain

where is the digamma function. For a fixed magnetic field we are interested in the highest temperature T at which we have a nonzero solution. Thus the order parameter obeys the linear , differential equation, equation (62), with the lowest eigenvalue q = 2eDH/c. We wish to point out that we have not assumed slow variations even if we obtain the simple Ginzburg-Landau form, equation (62). Equation (64) shows that the equation for A is an infinite order differential equation which can be solved simply if we are able to solve equation (62). T h e eigenvalues are given by equation (65). We also mention that a similar calculation near the boundary of the specimen leads to the same boundary conditions as near T,.

6.2.2. The nonlinear term. Although the linearized version of the Maki-de Gennes theory has been often used in a variety of situations the nonlinear portion is more complicated and has been used only in the mixed state. It is, however, important for the normalization of A. It was first derived incorrectly by h'laki (1964) and was later corrected by Caroli et aZ(l966). A generalized version is given by Werthamer (1969) in his review article and we will not give any details here. However, we wish to point out that there is really no differential equation of finite order for the order parameter. It is strictly speaking a differential equation of infinite order and one cannot speak of a local equation for A. There is, however, a considerable simplification compared to the starting equation. T h e coefficients of the expansion in powers of A are functions only of ( V - 2 i e A I ~ h ) ~So if A is an . eigenfunction of this operator it is an eigenfunction of the total equation for A. This simplification, which gives coefficients depending only on this combination, comes from the locality approximation: that is, the components of V - 2ieA/Ac

138

M Cyrot

commute with each other. The final result is

x {( I 1 - IT3)2 I

+ (IT2 - I I

4)

'} A:

A2 A$
1=2=3=~

= 0.

AT and

The operator IT, = (-io, f 2eAl6c) operates on A, and A; with the - sign for A,.

sign for

6.3. The slow variation approximation


The assumption of a small order parameter is a very convenient mathematical approximation. We will now relax this assumption but, of course, keep the local electrodynamic condition. I n the following, we will make an expansion in powers of gradients which means an expansion in powers of VI AI and in powers of the superfluid velocity vs = (1/2m) (V (arg A) ZieA/c). We hope (and we will verify later) that such a generalization can describe the BCS state slightly perturbed. This calculation was carried out mainly by Werthamer (1963,1964) and Tewordt (1963, 1965a,b)and confirmed by Eilenberger (1965).

6.3.1. Expansion in powers of the superjluid velocity. For pedagogical purposes, we first consider a clean material in which we assume IAJ to be constant. We allow a space-varying phase and a vector potential (Cyrot 1965). I n the Bogolubov equations (equation (31)), we can always eliminate this phase by a gauge transformation and rewrite them
EU(Y) =

EV(Y) =

(A (z"

( i6V + mz#

+ U ( r )- E ,

(i6V + mqJ2+ U ( Y - E , )

u(r)

+ I A Iv(y)

v(r)+ I A u(r).

(66)

We make the slow variation assumption, that is, we assume that the superfluid velocity varies over large distance compared to the extension of the excitations which is of order f . Thus we can divide the space into cells, large compared to E, but small compared to the distance over which vs varies. I n each cell the superfluid is practically in uniform translation. The excitations are the same as in an infinite medium because the dimensions of the cell are large compared to E. We calculate the excitation spectrum and the contribution to the free energy of each cell and then sum over all the cells. Equation (66) can be solved in the clean limit ( U ( r )= 0 ) and we obtain for the excitations
'1%

(67)

Ginxburg-Landau theory f o r superconductors

139

T o calculate the free energy we just apply the general functional given by equation (33). We expand the result in powers of cs and restrict ourselves to second order. We obtain

F(A(y))= FBcs( A I ) +

2
k

ggs2PF2 +mvs2N .

T h e first term of the right hand side is the BCS free energy. The last term takes into account the change in chemical potential introduced by the current (equation (67) ). N is the total number of electrons, We can rewrite the free energy as

F(A(r))= FBcs( I ) + *mvs2 1+ S a ft d f N. IA a

(68)

N(l+J(af/a()d() is the number of superfluid electrons so equation (68) is the direct generalization of the Ginzburg-Landau free energy. We have added to the BCS free energy the kinetic energy of the superfluid current. We will show that this result holds in the dirty limit and that we have also to add a term corresponding to the stiffness of lAl in the general case. 6.3.2. General expansion. We now allow the modulus of the order parameter to vary slowly in space and do not restrict ourselves anymore to the clean case. We always start from the general free energy functional and calculate the eigenenergy to second order in the generalized gradient (V - 2ieA/c), which means to second order in currents and in V I A / . For this purpose we use the same arguments as above. We divide space into cells large compared to the extension of the excitations but small compared to the range of variation of A. I n each cell we compute the eigenenergy to second order in vs and V I A I and sum the contribution to the free energy due to each cell. I n the cell centred at point r there is a small deviation A, from the value I A(r)I. We expand to second order in A, the Bogolubov equations and obtain for the second order correction to the unperturbed energy

the usual expression

m#n

Now if we use equation (36) which gives the value of constant order parameter A(Y)we obtain

U,

and v, for a medium with a

where the function S(s,s) has the same definition as in equation (45). Expanding the general free energy functional we get for the correction to the local BCS free energy FBCS( A(r>I (1- 2 f n ) E L 2 )

140

M Cyrot

which using the same transformation for 1-2f, as in equation (43) gives

d3s d3sA,(s) A,(s) S(s,s, t).

Carrying out the Taylor expansion

and recalling from equation (52) that


[S(s, s, t ) (s - s)2 d3s d3s
=

we obtain a calculation similar to the one we carried out near T,. Making first the , integration over 5 and t,, then over t, we obtain the following result

Fs

= F,

+ /d3r
L

A(r) l 2

+ N ( 0 )2rkB T

2
CO

- 2{(

I l 2 + I A(r) I2)/a U,

U}

U= 0

(69) This final result has been written on the form given by Werthamer (1969), that is, we have expanded the BCS free energy as a sum over w,.
FBcs
= T - N ( 0 ) 2 ~ k sT
A2

E -2{(~,2+A~)~2-~}.
v=o

(70)

One can easily check that aFB,,/aA = 0 gives the BCS self-consistent equation (equation (38)). T h e last two terms give the kinetic energy of the superfluid current and the energy linked to the variation of IAl. It is interesting to note that the kinetic energy of the current

can be rewritten
Ekin =

3QA

with

J = QA.

Q is given by the linear response of a BCS superconductor to a varying vector potential A when the electrodynamic is local (London limit) (cf Abrikosov et al 1962). There is, however, no simple way to predict the coefficient of the gradient of the modulus of the order parameter.

Ginxburg-Landau theory for superconductors

141

Let us now consider the validity of equation (69). We have assumed that the order parameter is slowly varying over a distance of the order of the range of the kernel S , that is, over f. Using the BCS formula for the order parameter, that is, aF,,,/aA = 0, we can rewrite

- = N ( 0 )2nkB T V

2
a 3

(: U

1-0

+ AT2)l1a

where AT is the BCS value of the order parameter at temperature T. T h e term which does not contain a derivative in the equation for A obtained from the condition a F p A = 0 can be rewritten

We see that AT - I A ( r )1 plays the role of T,- T in the Ginzburg-Landau equation and permits a slowly varying order parameter if

This means that only small deviations in [ A I from its BCS value can be maintained over large distance and permits the slow variation approximation. This condition is equivalent to our

required for the validity of the Ginzburg-Landau free energy. However, we must add explicitly that the electrodynamic is local. This is not implied by condition (71), that is, we must add A( T ) .

e<

6.4. Applications T h e linearized version of the de Gennes-Maki theory has been used in a variety of situations. It allows one to extend all results obtained near T, to all temperatures T h e nondown to zero close to the critical field in dirty superconductors (Z< linear term has permitted the generalization of the Abrikosov solution down to T = 0 near H , , (Maki 1964). I n addition to being itself an important result, this is very important for the microscopic theory. As we mentioned earlier, it allows one to calculate many transport properties because the order parameter is known independently of the Bogolubov equation. T h e Werthamer-Tewordt expansion is of more restricted application since condition (71) prevents its application to the mixed region of type I1 superconductors at low field and low temperature. It has been used to calculate H,, but the result is only valid not too far from T,. Tewordt (1965a,b) added all terms of fourth order and worked only in the temperature region (T,- T ) / T < 1. Thus no , restrictions on A and K are required, This is probably the most important application of this generalized Ginzburg-Landau equation (Neumann and Tewordt 1966a,b).

eo).

142

M Cyrot

6.5. Other extensions A number of other extensions have been investigated. We will only mention these extensions here in passing because they are restricted to certain specialized applications and should, more properly, be left to the discussion of the special cases.
6.5.1. Spin paramagnetism. If a material remains superconducting in very high fields, the Pauli paramagnetic energy of the electron spins becomes a significant fraction of the total energy of superconductivity and must be taken into account. This is important for the high critical field materials that now exist. This effect has been considered by Maki (1966) and Werthamer et aZ(1966). 6.5.2. Anisotropy o the Fermi surface. A basic assumption of the theory is that the f normal Fermi surface is isotropic. This assumption was removed from the phenomenological theory by Ginzburg (1952) and confirmed microscopically by Caroli et al (1963) and by Gorkov and Melik-Barkhudarov (1964). 6.5.3. Inhomogeneous systems. On a phenomenological basis, the theory has been generalized by Larkin (1970). One can also mention the generalization of Barisic and de Gennes (1968) for coupled linear chains and Lawrence and Doniach (1971) for superposition of thin films.

7. Time-dependent Ginzburg-Landau equations


7.1. Position of the problem T h e original Ginzburg-Landau theory and all extensions discussed so far have concentrated on thermal equilibrium situations. However, considerable attention has been devoted to the time-dependent nonequilibrium situation. T h e most direct application is, for instance, transport of current in type I1 superconductors. Here the superconductor develops a resistance which is associated with the motion of the order parameter. Another important application is the normal-superconductor phase-boundary propagation. Thus the question of interest is whether a generalization of the GinzburgLandau theory exists which can describe variations of the order parameter with time: that is, is it possible to establish a simple differential equation in time and space variables for the order parameter containing only the electromagnetic fields and the parameter A ? As, in the case of static problems, there is considerable simplification in the range of temperatures close to the critical temperature, one may hope that the general scheme of superconductivity can also be simplified for nonstationary problems. Unfortunately this happens to be in general untrue. This is connected to a considerable degree with the fact that the energy spectrum has an energy gap. There is no simple scheme, such as the Ginzburg-Landau theory in the static case, capable of describing the dynamic behaviour of the order parameter. One cannot proceed further than the linear term in the general case. However, in gapless superconductors the situation turns out to be simpler and one can derive a nonlinear term. This term is identical to the static one for the gapless region of a superconductor with magnetic impurities. I n the gapless region of a dirty type I1 superconductor there exists an anomalous term which has no equivalent in the static case. However, there still exists controversy because some

Ginxburg-Landau theory for superconductors

143

authors introduced it in the equation for A and others in the equation for the current, Attempts to derive time-dependent equations have been numerous. However, in the light of the work of Gor'kov and Eliashberg (1968a,b, 1969) many of them are invalidated or restricted to the linear term or to much more serious conditions than claimed by the authors. Theoretical investigations have pursued both the lines that we have discussed for the static case: slow variations of small amplitude from the equilibrium state and a small order parameter just below the second-order transition to the normal state. T h e first calculation along the former of these lines was reported by Stephen and Suhl (1964) and confirmed by Anderson et aZ(l965). O n the same lines, detailed investigations have been carried out by Stephen (1965), Abrahams and Tsuneto (1966) and Kemoklidze and Pitaevskii (1966, 1967). T h e latter line has been investigated by Schmid (1966), Caroli and Maki (1967a,b) and Gor'kov and Eliashberg (1968a,b, 1969, 1970). It turns out that the main results have been obtained following this second line. So we will mainly restrict ourselves to it and will discuss only briefly the other line. As in the static case, we first give a phenomenological approach to nonequilibrium thermodynamics of superconductors before the microscopic theory. This will show that there exists a dissipative process in the superconductor connected with the time variation of the order parameter. Then we discuss the possibility of a linear equation and its derivation in the case of small A. T h e problems which arise from the nonlinear term are then faced and some applications of the nonlinear electrodynamic are discussed.

7.2. Nonequilibrium thermodynamics of superconductors I n this section we want to apply the usual scheme of the phenomenological nonequilibrium thermodynamics to superconductors (Schmid 1966, Weller 1968). It is assumed that the superconductor is always in a state of local equilibrium described by the variables: charge density of the electrons p, internal energy density of the superconductors Eint, vector potential A and the order parameter A. T h e elementary excitations are assumed to be always in equilibrium with the prescribed values of the above mentioned variables. We have the conservation laws for charge, entropy and total energy E,
-+divJ aP at
=

0 R

as d i v J S 7 tf d

=T

2 + d i v J E = 0.
p is the density of entropy and J, Js, JEthe three currents associated with charge, entropy and energy. R is the dissipative function. There are losses due to joule heating of the normal electrons but according to Tinkham (1964) an additional dissipative mechanism due to the finite relaxation time of the superfluid electrons should be taken into account. We will show that this is in fact the case. Together with these equations, we need the Maxwell equations and an equation for the motion of the order parameter. If F(A) is the free energy functional of the

144

M Cyrot

superconductor, the general assumption usually made in a phenomenological approach is that the functional derivative of the free energy with respect to A is a generalized force, in the sense of irreversible thermodynamics, acting on the order parameter. Thus the order parameter relaxes to its equilibrium value proportionally to the derivative of the free energy in the functional space

We will now calculate the dissipative function. T h e density of total energy is

I n the framework of the phenomenological thermodynamics, the internal energy in a nonequilibrium state is assumed to be the same functional of the thermodynamical variables as in equilibrium
Eint

Eint(S,P, A, A)*

For the functional derivatives we have the thermodynamic reIations

,L L

is the chemical potential of the electrons and Js the density of supercurrent. Using these relations and the identity 2ie A*--A-

we obtain

: ;

)1 :

divJs

aE - _ - R+JS.VT-(J-Js)
at

[g +

2ie( 4+

A*/

+ divergences.

Comparing with the energy conservation law and introducing the normal current J = J- Js, we have ,

Ginxburg-Landau theory f o r superconductors

145

Our equation for the time derivative of the order parameter is not gauge invariant. From the non-negativity of the dissipation function we see that we must replace equation (72) by
y ---le

{E

so we have

-JS.VTi-J, E--Vp

I! ) i :I
e

(,++-

8F A = -aA*

(73)

+2y

--2ie K t

i++-I! It 1
A

(74)

and as predicted by Tinkham (1964) there is a contribution to R which is connected with the time variation of the order parameter. However, we emphasize that this approach is valid only if the time variations are slow. Rapid processes, such as the conversion of the thermally excited excitation to superfluids, must be, in some sense, irrelevant for such a description to apply. We discuss this point in the following section.

7.3. Possibility of a linear time-dependent Ginxburg-Landau equation Because of the complexity of the general theory we want first to discuss the following point : when is it possible to derive a linear time-dependent GinzburgLandau equation ? A priori, one can argue that we can generalize, in principle, the derivation we gave in the static problem by treating the time variation in the same manner as the spatial variation. We assume that the variation of A is slow compared to the time of importance in the kernel and expand in a space-time Taylor series about the local value. There is no mathematical difficulty in carrying out this programme and obtaining a time-dependent generalization assuming small A or slow variation. Unfortunately this time-dependent Ginzburg-Landau equation is not valid in general. Physically the reason is that there is a possibility of local conversion of the thermally excited normal excitations to superfluids. Mathematically the difficulty is due to the existence of singularities in the kernel of the integral equation. Let k and w o be the characteristic wavenumber and frequency of the space-time variations of the order parameter. I n the Taylor series we calculate, for instance, the time variation by setting k = 0 and taking the limit as wo goes to zero and the space variation by setting w o = 0 and taking the limit as k goes to zero. I n the actual situation the relevant ratio is U .k / w o where v is the group velocity of the thermally excited quasiparticle. Then if k < w o we are calculating the time derivative correctly and the space derivative incorrectly and the reverse in the other case. So the existence of terms a k / w o prevents simultaneous expansion in k and w o and makes the existence of even a linear time-dependent Ginzburg-Landau equation impossible, T h e only possibility is when this ratio is physically irrelevant. A first example is at zero temperature where there is no thermally excited quasiparticle, The method we describe leads to a time-dependent Ginzburg-Landau equation. A second example is near a second-order transition. We expand if possible in powers of A so, in some sense, we force the limit w o> A and the ratio vk/wo becomes once more irrelevant. This is the physically interesting situation. We will restrict ourselves to this situation. Near T, a time-dependent equation was first derived by Schmid (1966). I n fact he added the static nonlinear term to this equation but we will show in the following that this is not correct. Caroli and Maki

146

M Cyrot

(1967) succeeded in the gapless region of dirty type I1 superconductors. They computed the linear response due to an external field and a time-varying order parameter. We will consider now only their linear term. A time-dependent equation was also derived by Gorkov and Eliashberg (1968a) for the gapless region of a superconductor with magnetic impurities, starting from the self-consistent equation for the order parameter.
7.4. The linear time-dependent Ginxburg-Landau equations We introduce in our basic hamiltonian (equation (25)) time-dependent external fields. T h e linearization that we made to obtain equation (28) and the self-consistent equation (34) can be generalized by introducing a time-dependent order parameter whose definition is just the generalization of equation (34)

A(r, t ) = <+L(Y,

r(Y,

t)>.

(75)

T o define the nonequilibrium average we proceed as usual. We imagine that in the distant past ( t = --CO)the system is in thermodynamic equilibrium without the time-dependent interactions. T h e time-dependent interactions include external time-dependent fields and the time-dependent order parameter through the selfconsistent equation. is now a function of time. We perform the same transformation on the $ I (equation (29)):

t(r, t ) =

T h e equation of motion for Bogolubov equations


*

+ leads to the time-dependent generalization of the


u(y, t )

Z Yn n

zin(y,

t ) - Y$ :

v n ( y , t).

(76)

au -=
at
=

1 av -

at

(&( +$1 + u(r>E , 1 - (&(itiv + $I2 + U ( r )- E ,


- ihiv

+ ~ ( rt>,v(Y, t )

U(Y,

t ) + A*(r, t ) u(r,t).

We can write A in the same form as equation (34)

u,(Y, t ) and vn(r,t ) being given by equation (77) and time-dependent term occurs, that is, at t = --CO.

E,

is the eigenvalue when no

7.4.1. Temperature close to the critical temperature T,. This is the region of validity of the original Ginzburg-Landau equation. We will proceed in exactly the same way as we used to derive the static equation. T h e order parameter is small and we calculate by time-dependent perturbation un(r,t ) and vn(r,t ) to first order in A(Y,t). We begin by assuming only a time-dependent energy gap and obtain

x exp (iCp t) A(#, t)V;(Y, t )

and a similar equation for vn(r,t). Inserting both equations in the self-consistent

Ginxburg-Landau theory f o r superconductors equation we obtain

147

We now assume that A varies slowly over the range of the kernel of the integral equation and in addition we assume a slow variation in time, thus we can write

[( r ,- r,) .v + i r2). v 4(r14nTVN(O)

T h e spatial part has already been calculated in the static case. T h e coefficient of the time derivative is given by

+ (t1-

t 2 ) 4 A(y1, tl).

C
WY>O

Sdr2
-m

dt2 exp {- 2wp(t1- t2)) ( t - tz) S(r1, ~ 2 t,,t2>. ~ ,

Going back to the definition of S and introducing a switching parameter exp (26t) (8 small and positive) so that in the distant past the system was in thermodynamic equilibrium, we obtain (79) If we calculate the coefficient of the next-order term a2A/at2, we are led to the assumption that the expansion parameter of this series is of the order KIAB T,. I n the case of an external field we include the effect of a vector potential A(r, t) replacing V by V - 2ieA/cK. I n order to have a gauge-invariant equation we have to substitute for ajat a term a/at-2ie+, where is the scalar potential. A gauge transformation consists of the simultaneous substitutions

AaAexp@)X When calculating the current

A+A+Vx

+++---%. 1 ax

$ is given as a function of
case.

and

U, we

obtain the same formal result as in the static

7.4.2. Dirty superconductor near Hc2. As we can expand in powers of A the above derivation is still valid and we start from the general formula

where we have written t = t, - t2 and r = r1- r2.

148

1WI Cyrot

We have obtained the same equation as in the static case except that 2w, is replaced by 2w,+a/at. Consider first the field-free case. S is given by equation (63)) thus we have
1

A can be taken as the solution of


-+D a h
at where
eo

(V-- f';)A '

E,,A

is given by

I n the case of external fields we have to include a vector potential or a scalar potential. I n order to have a gauge-invariant equation we substitute alat-2ie4 for a/i?t. Thus we obtain the equation first derived by Caroli and Maki:

However, there still exists controversy about the order of the operator. Takayama and Ebisawa (1970) claimed that the proper order is

1 ZD(V - 2ieA/cZ)2 $4nkT ~

) -+(;)

+In-- T

T,

T h e first-order term with respect to &Y), exactly the same because

in the Caroli and Maki expression, is not

The two equations reduce to the same time-dependent Ginzburg-Landau equation near T,

The expansion parameter in deriving the static Ginzburg-Landau equation is A/max (&, eo), that is, Alk, T, near the transition temperature T, and Aleo near T = 0. I n the dynamical case the expansion parameter is A/iw+ eo as ~7ellas A / k BT,.

7.5. Applications of the linear time-dependent Ginxburg-Landau equation This equation has been extensively used by Maki (1969) to calculate properties of the mixed state in the presence of a transport current. He considered the ideal case in which the Abrikosov state moves uniformly carrying a finite transport

Ginxburg-Landau theory f o r superconductors

149

current. We can look at this problem by considering the solution of the timedependent equation in the presence of an electric field in the x direction. We assume the static magnetic field along the x axis. The scalar potential is given by

#J(4 E,* =
T h e equation ( 8 3 ) was first solved by Schmid. T h e general solution is
&Y (,

t ) = exp {ik(y+ ut))exp

- eH

(x

- -- -

k 2eH 4eHD

where

U =

E/H,, and H is given by

T h e moving Abrikosov state is given by linear superposition of

& (cf equation (23))

It is then easy to show that

-A(Y, t ) = u;;-A(y, t ) Bt O Y which expresses the fact that the Abrikosov states moves in the y direction with velocity - U . This relationship in fact follows directly from the gauge invariance of the time-dependent Ginzburg-Landau theory. 7.6. The nonlinear term 7.6.1. Introduction. T h e nonlinear problem is far more difficult and has just begun to be handled in a serious way. It is, however, very important because it gives the amplitude of the order parameter. For instance, the microscopic theory calculates only the interaction of weak electromagnetic radiation with superconductors, that is, calculates the response function to first order in the vector potential. T h e value of the order parameter is not changed from the value in zero field. I n the nonlinear problem we want to take into account the electromagnetic field to any order and to recalculate self-consistently the order parameter. T h e kinds of problem which can be solved are the following. The first is that of determining the frequencies characterizing the behaviour of the parameter A. It is clear beforehand that A(t) will follow the field adiabatically at external field frequencies below a certain characteristic frequency R,. Therefore to determine A ( t ) when w < R, it is possible in the first approximation to use the equation of the Ginzburg-Landau theory in which the field depends on t as a parameter. On the other hand, when U > Q we can expect A to respond, in the main, to the time-averaged value of the , square of the field whereas the high frequency oscillations of A will have a small amplitude. I n this case, the time-averaged value of A will be determined by the Ginzburg-Landau equations which contain the mean square of the field. Another important problem is that of the behaviour of the field A ( r , t ) itself. Here, too, there should exist a characteristic frequency RI that separates the region of low frequencies, at which the penetration of the field into the superconductor is determined by the Meisner effect, from the high frequency region where the main role is played by the skin effect.

150

M Cyrot

It was first suggested that this type of problem can be handled just by adding to the linear time-dependent Ginzburg-Landau equation the nonlinear term of the static equation. It was first pointed out by Gorkov and Eliashberg that this is, in general, untrue and that the situation is far more complicated. Near T,,it has been shown by Eliashberg (1969) that in the general case there is no nonlinear timedependent equation. T h e first success at deriving such an equation was in the gapless regime of a superconductor with magnetic impurities. Gorkov and Eliashberg (1968a) were able to show that the nonlinear term is, in that case, identical to the corresponding one in the static equation. Caroli and Maki claimed that the same result holds in the gapless regime of a type I1 superconductor. However, this contradicts a result of Gorkov and Eliashberg (1970) which shows that, in addition to the static nonlinear term, there exists an anomalous term which has no counterpart in the static equation. I n the following, we want to explain the reason why there is no nonlinear timedependent equation except perhaps in a gapless superconductor. Then we first examine the gapless regime of a superconductor with magnetic impurities and afterwards turn to the controversial case of the gapless regime of a type I1 superconductor. We end with an example which shows how the order parameter follows an AC magnetic field.

7.6.2. Existence of a nonlinear time-dependent equation. T o calculate the nonlinear term we have to solve the time-dependent Bogolubov equations to second order in U and U . Letting

we have

where

T o second order in A, we have +(r, t) = 1 with Hint(t) = exp ( - iHot) HI exp (iHot). When introducing this result into the self-consistent equation (75) the term of third order in A will be

Hint(t1) dt,+

jl dt1
--CO

-m

dtz Hint(t1) Hint(tz)

T h e function H will be given in terms of the normal state wavefunctions and can be handled using the same method as in the static case. I n the static case near T,,we have shown that A varies over distance 3) and that the range of the kernel is ( so we can take y1 = rz = y3 = Y. In the nonstatic case we have to investigate the range of the kernel for the time behaviour. We have seen that the important

((c-

Ginxburg-Landau theory foy superconductors


correlation function which occurs in this kernel is

151

(K+(O) t ) ) W assuming from our reasoning that A is not space dependent. T h e time behaviour
'

of this function determines the time behaviour of the kernel. First we notice that for all cases where the one electron hamiltonian is invariant by time reversal

[ K , H ]= 0
we have (K+(O)K(t)) 1 and the range of the kernel is infinite. There is no hope = of transferring the simple procedure for the space to the time and we cannot obtain a time-dependent Ginzburg-Landau equation. This case corresponds to the usual BCS superconductor even near T,, On the other hand if

[ K ,HI z 0
and more particularly if we have

where T , is a function of the strength of the perturbation, for instance, magnetic field or magnetic impurity concentration. I n that case the range of the kernel will be of the order of T , and if the time variation of A is larger than r Kwe will be able to put t, = t, = t, = t. T h e nonlinear term of the time-dependent equation will be exactly the same as in the static equation. This case corresponds to gapless superconductivity as shown by de Gennes (1966). After these considerations we must explain why we were able to derive a linear equation near T, when ( K ( O ) K ( t ) ) 1. I n fact the range of the kernel for the = linear term remains finite and of the order of ElkB T, because we have to take into account the sum over U,. We have zexp(-2wvt) =
WV

exp ( - rRkB T t ) 1 - exp ( - 2 r ~ k T t ) ,

Thus, near T, the finite range of the kernel comes from this term and as A varies over time of order E/kB(T,- T ) , we were able to obtain a time-dependent equation. If we calculate the kernel for the nonlinear term we have a succession of four times and the equivalent temperature-dependent factor is not sufficient to maintain all the four times within ti/kBT,. Thus the procedure used cannot be applied to the nonlinear term. 7.6.3. Magnetic impurities. T h e first case studied without ambiguity was a superconductor with magnetic impurities. Abrikosov and Gor'kov (1961) had shown that magnetic impurities drastically suppress superconductivity because their spins acting differently on the two electrons of a Cooper pair tend to destroy it. However, for concentrations just below the concentration needed to suppress superconductivity, there is no gap in the energy spectrum. This is related to the existence of l/r,. If the perturbation due to the impurities is of the form

where Si is the spin of the impurity at site i and Seis the electronic spin, l / T K is
7

152

M Cyrot

directly related to the power spectrum of K and is given by (de Gennes 1966)

n is the concentration of impurities. When r K h < 1, that is, the energy of pair is breaking, %-/TK, greater than the order parameter, there is no gap in the spectrum of the excitations. I n that limit, by direct expansion in powers of AT, and W T ~ Gorkov and Eliashberg (1968a) obtained the time-dependent Ginzburg-Landau equation which is just the generalization guessed previously when one adds the nonlinear static term, T h e characteristic time of variation of the order parameter is 1/7KA2 and is always greater than T~ in the gapless regime.
7.6.4. Gapless superconductivity of dirty materials. The second case turns out to be more complicated. The reason is that TK is linked with the spatial variation of the order parameter and is a function of temperature. We have found

1 _ - Eo TK

where

eo

is given by equation (65)

thus 1 / T ~ N ( 7 T / 8 ) k g ( TT ) tends to zero as T tends to T,. Thus some difficulties can occur near T,. From their general equations determining the nonstationary properties of superconductors, Gorkov and Eliashberg have shown that the expression for A(r, t ) consists of two terms, T h e first part is regular and has the properties that near the transition it can be expanded in powers of A/T, and when the field frequency is decreased it goes over into the corresponding static expression. This only partly gives the time-dependent Ginzburg-Landau equation with the static nonlinear term. However, they found a second part which is very sensitive to the detail of the spectrum and which they called the anomalous part AU. One must add it to the equation for A and this term, in general, prevents a simple GinzburgLandau theory. I n the case of magnetic impurities this term reduces to the potential 4. However, in the case of dirty type I1 superconductors near He,, they claimed that it does not. So one would not obtain in that case the time-dependent equation derived by Caroli and Maki where the nonlinear term is just the static one. As w,, and k, the frequency and the wavevector of the varying field, are small, Gorkov and Eliashberg only calculated the anomalous part with quantities Id l2 averaged over large volumes compared with the parameter of the local structure. I n that case U is the sum of two terms U, and U,. U , obeys the following equation

with

(( 12x4)-1

for x Q 1

Ginxburg-Landau theory for superconductors The quantity U , can be expressed in terms of A and

153

+.

at absolute zero U, reduces the potential

U2 -- _ - ie+
EO

(93)

We notice that U , = 0 at zero temperature. Thus we obtain the Caroli-Maki Ginzburg-Landau equation. I n the same manner, if for magnetic impurities we change eo by 1 / T K , as T K T < 1 there is no contribution to U,. U,, in both cases, gives the potential needed to preserve gauge invariance. However, for finite temperatures the result disagrees with the Caroli and Maki result. As predicted from general considerations the discrepancy is larger near T,. Weller (1969b) has already predicted that the Maki and Caroli results could be obtained from Gorkov and Eliashberg formalism only at low temperature. However, Takayama and Ebisawa (1970) claim that they can justify the Caroli and Maki nonlinear term by showing that the pair breaking parameter which is introduced by the spatial dependence of the order parameter plays exactly the same role as in the magnetic impurities case. This is in direct conflict with the Gorkov and Eliashberg result. Thus no final conclusion can be drawn except that this anomalous term is likely to exist. According to Gorkov and Eliashberg the expression for the electrical current is the same as in the static case except for the contribution due to the electric field

kD

(94)

There arises another unsolved problem. In the presence of a transport current as we have shown previously on the linear equation, the whole vortex structure moves. This causes dissipation, so there exists a finite conductivity in the superconductors. It was first calculated by Caroli and Maki (1967a) wrongly and corrected by Thompson (1970). However, application of the time-dependent equation with the nonlinear static term and equation (94) does not give the right answer. Houghton and Maki (1971) argued that in the calculation of the current there is also an anomalous term (Maki term or Thompson term) which would have not been taken into account by Gorkov and Eliashberg. In order to obtain the right answer they showed that in the conductivity equation one must replace aE by asE with

It appears rather surprising that some authors do not get an anomalous term in the equation for A but do get one in the equation for the current when Gorkov and Eliashberg obtained the reverse. It seems, however, that the anomalous term gives, in both cases, exactly the same result for the dissipation as the two equations have to be used together (cf equation (95) and equations (90) and (91)). We add in

154

M Cyrot

equation (94) the anomalous term U,A* given by equation (90) to

where we neglect the spatial dependence of because we have averaged over the local structure. We also assumed that U , is given by equation (93)' then we obtain the same result as Thompson for the conductivity

Thus everything would behave as if the anomalous term would appear in the current when using a linear (or nonlinear) response theory as used by Maki and would appear in the equation for the order parameter when expanding the selfconsistent equation. From a fundamental point of view the second procedure is more justifiable and gives more generality to the result. Let us end this part with a last comment on the Gor'kov and Eliashberg result. It seems strange that the equation for U , gives a solution of equation (96) so sensitive to the initial condition. Gor'kov and Eliashberg argued that the reason is the absence of any mechanism of homogeneous relaxation in the model. I n order that relaxation may take place, one must take into account the inelastic processes such as electron-electron and electron-phonon scattering. The pair breaking parameter associated with these processes is of order T 2 / E , or T3/BD2 respectively (8, Debye temperature). If one allows for this damping instead of equation (90) we get

Thus the anomalous term vanishes in the static case.

7.7. Applications
We wish now to give some examples of the kind of problems that we are able to handle with a time-dependent Ginzburg-Landau equation. We mentioned at the beginning that it is important to know the frequency characterizing the behaviour of the order parameter and its response to frequencies higher than this characteristic frequency. We will now give an example. We consider a thin film of a paramagnetic alloy with dimension 2d (see figure 7 ) less than the characteristic dimension of the vortices so we will be able to assume [ A I = Cte. I t is placed in a varying magnetic field which is the same on both sides and parallel to it. We want to calculate the order parameter as a function of the amplitude and frequency of the external field. T h e condition that the magnetic field distribution is symmetric imposes that the total current through the film vanishes. Neglecting the induced electric field and letting A = I A I exp (2ieic) x we obtain

So we can take the gauge

( A- vx)x = - H, tz.

Ginxburg-Landau theory for superconductors


T h e equation for the modulus of the order parameter is

155

We average over the thickness of the film with the boundary condition

and obtain

~ + ~ ( - v ~ ( ~ ~ - -

0.

(97)

Figure 7. Schematic diagram of a thin film of paramagnetic alloy with dimension 2d.

T h e static critical field is then given by

If we introduce a reduced unit for the field

we have to solve

u at 7 1 4 ] 3 - / 3 ( 1 - h 2 ( t ) ) A = 0 +
with

Gorkov and Eliashberg showed that the equation can be solved by introducing a new function q(t)
1

4 = q ( t ) e x p f l ]0 1 - k 2 ( t ) ) d t . (

(101)

156

M Cyrot

T h e equation for 17 can be simply solved to give A(t)


=

exp p Jb dt'( 1- h2(t ' ) ) [C++ ~ & d t , e x p {ZP j$dt2(l -h2(t2)))]1~2' H(t) = H,sinwt

If the external field has the frequency


we have

A(t) =

[C + & & ,

exp (p(1-72) t + ( P / Z w ) Fsin Zwt} dt, exp 2@( 1- 7?') t, + (Pi2w)hz sin 2~tJ1~1'2

(103)

where the bar indicates the average value over a period. We first assume wt > 1 and 1- > 0 and consider the two limiting cases P/Zw < 1 and P/w 1, I n the first one we get

A, is the thermodynamic equilibrium value of A. I n the opposite limit we have


A ( t ) = A,(l -h2(t))1'2.

A(t) follows the field adiabatically. Kow if 1 - h2(t)< 0 from equation (102) we see that A(t) is exponentially small. We notice that the frequency region in which the transition from the adiabatic variation of A to the high frequency variation takes place is of the order of /3 = T~ Ao2. I n fact the same frequency separates the region in which the penetration of the field into the superconductor is determined by the skin effect w B T ~ A ~ ~ from the region of the Meisner effect w < r g A o 2 . As noticed by Gor'kov and Eliashberg (1970) this is not always the case. If the anomalous term contributes to the equation for A, the frequency characteristic of the variation of A turns out to be much smaller than the frequency at which the skin effect begins to play a role.

Acknowledgments I wish to thank Professor P Martin and D r F Cyrot-Lackmann for reading through the manuscript and Professors P G de Gennes and K Maki for numerous
discussions.

References
ABRAHAMS E and TSCNETO T 1966 Phys. Rev. 152 416 ABRIKOSOVA 1957 Sov. Phys.-JETP 5 1174 A ABRIKOSOV and GOR'KOV P 1961 Sov. Phys.-JETP 12 1243 AA L ABRIKOSOV A, GOR'KOV P and DZYALOSHINSKII 1965 Quantum Field Theoretical A L T Y Methods in Statistical Physics (Oxford : Pergamon Press) N JM ANDERSONW, WERTHAMERR and LUTTIKGER 1965 Phys. Rea. 138 A1157 P ASLAMAZOV and LARKIN I 1968 Phys. Lett. 26A 238 LG A BARDEEN COOPER N and SCHRIEFFER 1957 Phys. Rea. 108 1175 J, L J R BARISIC and DE GEKNES G 1968 Solid S t . Commun. 6 281 S P BLATT M 1964 Theory of Superconductivity (New York: Academic Press) J U, W L 1967 BRANDT PESCH and TEWORDT 2. Phys. 210 209 CAROL1 c, CYROT and DE GENNES G 1966 Solid St. Commun. 4 17 M P

Ginxburg-Landau theory for superconductors

157

CAROLI DE GENNES G and MATRICON 1963 Phys. Mat. Cond. 1 176 C, P J CAROLI and MAKIK 1967a Phys. Rev. 159 306 C -196713 Phys. Rev. 159 316 L COOPER N 1956 Phys. Rev. 104 1189 CYROT 1965 Phys. Mat. Cond. 3 374 M -1970J. Low Temp. Phys. 2 197 CYROT and MAKIK 1967 Phys. Rev. 156 433 M G EILENBERGER1965 2. Phys. 182 427 __ 1966 2. P h y ~ 190 142 . ELIASHBERGVI 1969 Sov. Phys.-JETP 28 1298 G FERRELL 1969J. Low Temp. Phys. 1 241 RA DE GENNES G 1964a Phys. Mat. Cond. 3 79 P -1964b Rev. Mod. Phys. 36 225 -1966 Superconductivity of Metals and Alloys (New York: Benjamin) GIKZBURG 1952 Zh. Eksp. Teor. Fiz. 23, 216 VL -1960 Sov. Phys.-Solid St. 2 1824 V L GINZBURG L and LANDAUD 1950 Zh. Eksp. Teor. Fiz. 20 1064 GLOVER E 1967 Phys. Lett. 25A 542 R M R GOLLUB P, BEASLEY R, NEWBOWER S and TINKHAM RiI 1969 Phys. Rev. Lett. 22 1288 J L GORKOV P 1959 Sov. Phys.-JETP 9 1364 -1960 SOV. PhyS.-JETP 10 593 -1958 Sov. Phys.-JETP 8 505 L G GORKOV P and ELIASHBERG hI 1968a Sov. Phys.-JETP 27 328 -1968b Sov. Phys.-JETP Lett. 8 202 -1969 SOV. Phys.-JETP 28 1991 __ 1970J. Low Temp. Phys. 2 161 L TK GORKOV P and MELIK-BARKHUDAROV 1964 Sov. Phys.-JETP 18 1031 HOHENBERG1968 Conf. Fluctuations in Superconductors, Asilomar, California unpublished PC HOUGHTON A and MAKIK 1971 Phys. Rev. 3 1625 R J A J E JAKLEVIC C, LAMBEJ , SILVER H and MERCEREAU 1964 Phys. Rev. Lett. 12 274 JOSEPHSON D 1962 Phys. Lett. 1 251 B -1965 Adv. Phys. 14 419 L KELDYSH V 1964 Sov. Phys.-JETP 20 1018 KEMOKLIDZEP and PITAEVSKIIP 1966 Sou. Phys.-JETP 23 160 lLl L -1967 SOV. PhyS.-JETP 25 1036 LANDAUD and LIFSHITZ &I 1950 Statistical Mechanics (Oxford: Pergamon Press) L E LANGER J 1964 Phys. Rev. 134 A553 LARKIKA 1970 Sov. Phys.-JETP 31 784 I LAWRENCE and DOKIACH 1971 Proc. 12th Int. Conf. Low Temperature Physics ed W E S E Kanda (Academic Press of Japan) p361 LONDOK 1950 Superfluids vol I (New York: Dover) F MAKIK 1964 Physics 1 21 __ 1966 Phys. Rev. 148 363 -1969J. Low Temp. Phys. 1 45 MAKIK and TSUZUKI T 1965 Phys. Rev. 139 A868 L L 1966a 2. Phys. 189 55 NEUMANN and TEWORDT -1966b 2. Phys. 191 7 ORSAY GROUP 1966 Phys. Mat. Cond. 5 141 PARKS D 1969 Superconductivity (New York: Marcel Dekker) R SAINT-JAMES and DE GENNES G 1963 Phys. Lett. 7 306 D P SAINTJAMES SARMAand THOMAS1968 Type I 1 Superconductivity (Oxford : Pergamon D, G EJ Press) SCHMID 1966 Phys. Mat. Cond. 5 302 A SCHMIDT 1968 2. Phys. 216 336 H STEPHEN J 1965 Phys. Rev. 139 A197 NI STEPHEN J and SUHL 1964 Phys. Rev. Lett. 13 797 M H TAKAYAMAEBIsAw.4 H 1970 Prog. Theor. Phys. 44 1450 H and TAKAYAMA MAKIK 1971 Prog. Theor. Phys. 46 42 H and

158
TEWORDT Phys. Rev. 132 595 L 1963
__ 1965a Z. Phys. 180 385

M Cyrot

1965b Phys. Rev. 137 1745 -1 9 6 5 Z. Phys. 184 319 ~


~

THOMPSON1970 Phys. Rev. B 1 327 RS TINKHAM M 1964 Phys. Rev. Lett. 13 804 WELLER 1968 Phys. Stat. Solidi 30 373 W -1969a Phys. Stat. Solidi 35 567 -1969b Phys. Stat. Solidi 35 583 WERTHAMERR 1963 Phys. Rev. 132 663 N -1964 Rev. Mod. Phys. 26 292 __ 1969 The Ginzburg-Landau Equations and their Extensions in Superconductivity ed R D Parks p321 WERTHAMERR, HELFANDand HOHENBERG 1966 Phys. Rev. 147 295 N E PC R ZAITSEV 0 1965 Sou. Phys.-JETP 21 1178

Вам также может понравиться