BE VIEWS OF MODE RN PHY'SIC S V 0 LU iVI E 3 4, N UM BE R 4 OC TOBE R 1962

Critica. . . ~ie. .c s anc Currents in Superconi uctors

JOHN BXRDEKX
University of Illinois, Urbana, Illinois

ing the value which makes the free energy a mini-

I. INTRODUCTION mum for given fields or currents.
&0 determine critical fields and currents in small Several calculations of critical currents and fields
superconducting specimens, it has long been have been given, particularly by Gin z burg and
recognized that it is necessary to take into account Abrikosov, 4 on the basis of the original version of
the effect of a change in the order parameter with the G-L theory. This work can be taken over with
Geld or current. Pippard' discussed the problem in a little or no Inodification for application to tempera-
qualitative way and showed that under some con- tures near T, . Douglass' has used the G-L equations
ditions the order parameter goes gradually to zero to estimate the magnetic-field dependence of the
with increasing field, giving a second-order transi- energy gap, and finds good agreement between
tion, while in others the transition is of first order. theory and experiments of Giaevar and Megerle. '
He suggested that the concentration of supercon- By using the gap as a parameter, and by determining
ducting electrons n, of the two-Quid model might the dependence of free energy on gap at lower tem-
be taken as the order parameter, and that an extra peratures from a phenomenological two-Quid model,
energy is involved if n, changes appreciably over Tinkham' was able to get a significant improvement
distances comparable to the coherence distance. over the G-L theory in fitting his data on the de-
A phenomenological theory for treating such pendence of thermal conductivity on magnetic Geld.
problems quantitatively was given by Ginzburg and Rogers' used the gap as a free parameter and deter-
Landau (G-L), ' in which they introduced an effec- Inined how it changed with field or current. The
tive wave function + whose square is proportional present paper is an attempt to review this work, to
to n, . Their equations reduce to those of the London put the theory on a more systematic basis, and to
theory when space variations of n, are ignored, and apply it to several problems.
thus are not valid when the Pippard nonlocal theory In a few cases, more basic calculations have been
must be used. Gor'kov' has shown that the present made from the microscopic theory in which a modi-
microscopic theory leads to the G-L theory near 7', Ged equation for the gap has been obtained for a
where the London equations do apply. The only superconductor in the presence of fields or currents.
change is that the effective charge is — 2e rather Rogers' treated the case of a uniform current Qow
than — e, a consequence of the pairing in the super- in a bulk specimen. Nambu and Tuan' have calcu-
conducting ground state. Gor'kov takes +(r) propor- lated the reduction in gap to the second order in
tional to the energy gap parameter A(r). As we shall Geld strength for a bulk superconductor with Gelds
see, there is a close connection between this choice of arbitrary wavelength. Results obtained are equiva-
and the original concept of n, as the order parameter. lent to those obtained. by the variational method in
In the present paper we shall be primarily con- cases where they can be compared.
cerned with small specimens of dimensions less than Section II is concerned with a general formulation
the coherence distance, so that changes in the energy of the thermodynamic relations when the gap is
gap with position may be neglected. One may then
extend the theory to arbitrary temperatures by 4 V. L. Ginzburg, Doklady Akad. Nauk. S.S.S.R. 83, 385
(1952); J. Expt. Theoret. Phys. (U. S.S.R.) 34, 113 (1958);
treating the gap as a variational parameter and find- —
Soviet Phys. JETP 7, 78 (1958); A. A. Abrikosov, Doklady
Akad. Nauk. S.S.S.R. 86, 489 (1952).
D. H. Douglass, Jr. , Phys. Rev. Letters 6, 346 (1961);
&

*Supported in part by the U. S. Army Research Office IBM J. Research Develop. 6, 44 (1962).
(Durham). s I. Giaever and K. Megerle, Phys. Rev. 122, 1101 (1961).
&
A. B. Pippard, Phil. Mag. 43, 278 (1952). 7 M. Tinkharn, IBM J. Research Develop. 6, 49 (1962);
2V. L. Ginzburg and L. D. Landau, J. Exptl. Theoret. D. E. Morris, Ph. D. thesis, University of California(Berkeley),
Phys. (U. S.S.R. ) 20, 1064 (1950). 1962 (unpublished).
s L. P. Gor'kov, J. Exptl. Theoret. Phys. (U. S.S.R. ) 36,
— E. T. Rogers, Ph. D. thesis, University of Illinois, 1960
1918 (1959); 37, 883 (1959); Soviet Phys. JETP 9, 1364 (unpublished).
(1959); 10, 593 (1960). s Y. Nambu and S. F. Tuan, Phys. Rev. (to be published).

667
 JOHN BAH, DEEN

treated as a parameter, Sec. III with the Ginzburg- More generally, one may define p, by the equation
Landau equations, Sec. IV and Appendix C with
explicit calculations of the free-energy difference p, (v, ) = dJ, (v. )/dv, , (2.3)
between superconducting and normal states for so that
general values of the gap parameter, and Sec. V
deals with applications ef the methods to the calcu- dJ, dJ. dv. ep, G
(2.4)
lation of changes of gap with field or current in small dt dv, dt m
specimens. Expressions for the magnetic moment of
small specimens are listed in Appendix A and Ap-
It is assumed that the variation with time is suf-
ficiently slow, so that the quasi-particle distribution
pendix B gives Roger's calculation for uniform cur-
attains a steady-state distribution appropriate to
rent Aow.
the velocity v, of the ground pairs.
II. THERMODYNAMIC RELATIONS One may use (2.2) and (2.4) to define v, and p.
when impurity scattering is present so that the wave
To discuss the thermodynamics of a supercon-
vector k is not a good quantum number. It is neces-
ductor in a magnetic field or with current How, it is
sary to include eA'ects of impurity scattering in thin
most convenient to take the external field II and the
films or other small specimens where electrons can
superQuid velocity v, as independent variables. The
be randomly scattered from the surface.
latter is a significant variable only for thin films or
The displacement of the pairs causes an increase
wires of dimensions small compared with the coher-
in free energy of the system which may be expressed
ence distance. In a bulk specimen, the current How
is confined to the penetration region near the surface
J,
simply in terms of . The rate at which work is done
per unit volume on the supercurrent by the electric
and is determined by the magnetic field at the
field is
surface. The critical current is determined by the
bulk critical field (Silsbee s hypothesis). In the dis- dW/dt = —eG g v; = —ee J;()/m = V J.(dvv/dt).
cussion of thin films or wires, we shall assume that v,
is constant across the cross section. (2.5)
We may take v, to be the common velocity of the The net increase in free energy obtained by inte-
pairs in the ground state. More precisely, if the grating with respect to t is
ground-state pairing" is (k+ q $, — k+ q J, ), one
may define v, = h, q/m, where, in a periodic potential,
m is an effective mass. In an electric field 8, , the J(v,')dv! . (2.6)
acceleration is given by
In addition to the work associated with the super-
mdv, /dt = —eG . (2. 1) fluid component, there also will be a dissipation of
energy from scattering of quasi-particles. 'By ac-
The entire distribution of electrons, including the
celerating the electrons in a suKciently small electric
pairs, is displaced in momentum space by the electric
field, this ohmic energy dissipation can be made
field. Scattering of quasi-particles tends to reduce the
negligible with respect to E&.
current, but does not change the value of v, . In a
The increase in free energy as a result of an external
normal metal such scattering reduces the current to
magnetic field H is
zero, but in a superconductor a net Row remains.
It is convenient to use rather than the electric
current density, the density of mass How defined by
M(H') dH', (2.7)

(2.2) where M is the total magnetic moment. This expres-

sion applies to a body of any shape. For a body of a
where 'U is the volume and v; the velocity of the ith simple shape such that there is no demagnetizing
electron. The electric current density is then —e J,/m. field,
When v, is small, the equilibrium density J,is
proportional to v„and the ratio is the density p, of (2 8)
the superAuid component of the two-Quid model. " where x is the susceptibility, equal to —1/4s. for a
bulk superconductor.
0 J. Bardeen, L. N. Cooper, and J. R. Sehrieffer, Phys.
Rev. 108, 1175 (1957).
For a specimen with dimensions small compared
n J. Bardeen, Phys. Rev. Letters 1, 899 (1959). with the coherence distance, the magnetic moment
 CRITICAL FIELDS IN SUPERCONDUCTORS 669

for small fields is given by an expression of the form Integration of (2.12) gives the usual expression

M = 'U—
aHN(0) 606 tanh (-', Ph), (2.9)
F, (H, v, ) = F, (0,v, ) — M(h(H', v, ),H')dH' . (2.18)
where a is a constant dependent on the shape of the
body but independent of H and of the energy gap In Eq. (2. 10), the dependence of F„, M, and J, on
parameter A. Here N(0) is the density of states of the gap has been indicated explicitly. Methods for
one spin in energy at the Fermi surface and Ap the determining these functions are discussed in the
gap at T = 0 in the absence of fields. The method of following sections.
derivation of (2.9) together with a list of values of If we assume that x(A) and p, (A) are independent
the parameter c for different specimen shapes is of H and v„we have
given in Appendix A.
In large fields, the gap may depend on H and v, .
Further, for very large fields there may be strictly
.
F. = F, (~) — ~x(a)H'+ , ~p. (Z)v'. .
-', —'
(2.14)

nonlinear effects not included in the change of A. In this approximation, the gap is treated as an order
The fact that there is only a very small change of parameter, and the dependence of x and p, on 0 is
included, but other nonlinear effects are ignored.
penetration depth up to the critical field, indicates
While this procedure can be justified for tempera-
that such effects are unimportant for fields up to the
tures near T„and, in fact, gives the same results as
usual bulk critical fields of a few hundred oersteds,
the Gor'kov version of the Ginzburg-Landau theory,
but they may be important for specimens of small
dimensions with critical fields of 104 or even 10' Oe.
it may be in error for small specimens at low tem-
peratures.
We shall ignore such effects here, and include only
An equation similar to (2.14), but which ignores
the dependence of 6 on H and v. .
The procedure is to take 6 as an independent changes in x and p, with the gap or equivalent order
parameter has sometimes been used to estimate
parameter along with H and v, . This implies that
critical fields or currents. The transition to the normal
calculations of the free energy are made with a set
of quasi-particle states appropriate to a gap state would then occur when the added terms are
giving
equal to the free energy difference, F„— .
F. . be-
tween normal and superconducting states in the
absence of fields and currents. The critical field for
F, (A,H, v. ) = F,o(h) — M(H', 6)dH' v. = 0 would then be

+ '0 J,(v'„A)dv!, (2. 10) H. = H. g( m4i xi)

' ', (2.15)
where H, t, is the bulk critical field, for which y
where F„(A) is the free energy in the absence of = 1/4~. The critical current density for H = 0
fields and currents, but with a gap which may differ would be
from the usual value. One then finds the value of
D(H, v, ) which makes the over-all free energy a mini-
J, = p, v. = (p./4m)H. g. .(2.16)
mum for given H and v, by setting At T = O'K, and in the absence of impurity scatter-
ing, p, = p = nm and H, f, ——Hp so that
(BF,/Bh)~, .„r = 0. (2. 11)
—[nmN(0)ho]
When this value of 6(H, v, ) is substituted in (2. 10),
J, = (p/4v)' 'Ho

one obtains an upper bound on the true free energy. = (2/3)' 'N(0)hmv~, (2. 17)
It should be noted that the moment M(H, v, ) is where in the last form v& is the velocity at the Fermi
given by the usual expression,
surface for a simple free electron model. We have
used the relation H20/8~ = —, 'N(0) 620, where Ho and
Ap are the bulk critical field and gap parameter at
T = O'K, respectively. These expressions for H, and
J, are equivalent to those which have been derived
(2. 12) some years ago from the original London theory.
When changes of 6 with field or current are taken into
Since the second term on the right vanishes by virtue account, larger values are obtained for critical fields
of (2. 11), one finds that M(H, v, ) = M(d, (H, v, ),H). and somewhat smaller values for critical currents.
670 JOHN BARDEEN
III. GINZBURG-LANDAU EQUATIONS FOR T —To One may express p, /p in terms of the penetration
'
Gor'kov, by use of a Green's-function method, depth X(T) for the given temperature and Xr.
has shown that the Ginzburg-Landau equations may (pc'/4s n'e') "~', the London penetration depth
be derived from the microscopic theory if the tem- for density p'.
perature is near T, so that the local London theory = =
p, /p Xi/X' n/p . (8 6)
applies. At these temperatures, p, is proportional to
the square of the gap, so that if one defines an effec-
tive wave function 0', such that = H', gX'/2~ = H', tX'/2xn), 1, . (8.7)
n n)ir, p
Ps/P ) (8.1) For temperaturesnear T. such that (T. —T)/T
+ will be proportional to the gap. Gor'kov used a 1 —t «1, the
model of Cooper, Schrieffer, and
somewhat different normalization for q, one which the author (BCS)" gives
is not as close to that of G-L. With our definition,
the G--L equation may be written
p, /p 2(l —t) . (8.8)
One may also write for temperatures near Te)

2m*
—
Br
—A(r)
8 —ie
Ac
4 —nq + P~%~ q = 0, H'„= (dH. ,/dt)'. (1 —t)' (8.9) .
(8.2) The slope of the critical Geld curve near T, may be
where m* = 2m and e* = — 2e, representing the expressed in terms of the jump in speciGc heat at T.
mass and charge of a pair. With Gor'kov's deGnition by use of Rutgers relation:
of 0', m rather than 2m appears in the kinetic energy
term. Equation (8.2) may be regarded as representing
(1/Ss-)(dH, b/dt). = -', T, (C. —C„) . (8.10)
the center-of-mass motion of a bound pair; the wave For the BCS model, T.(C, —C.) = 6.2(H,'/Ss. ) and
vector for%'is q„, that of a pair. = T, (C, —C.)(l —t) .
nn (8. 11a)
The free-energy difference per unit volume is
obtained by integrating over space and multiplying np = ' T, (C, —C„) .
—, (8. lib)
by n/2, the number of pairs per unit volume at Near T„ the equilibrium gap is given by

(A(T)/A) = 81(1 —t) (8.12)

p', —p'„=—
n 8 ——
— ie*
A (r) The expressions (8.7) are valid when impurity
2 2m* Br hc
scattering is present, if the appropriate value of X
I&l'+ l el+I') &' (8.8) is used. Although H, ~ and the equilibrium gap are
not changed very much by scattering, P is increased.
The density of mass Bow is given by According to Miller, "
in the limit for which the
mean free path / is much less than the coherence
84
+—
n ih 8%'* e*
i%'i'A . (8.4) distance, $o = Avs/vrAp,
2 2 Br Br e
p. ) '. ~tA tanh (-,'PA)
The kinetic-energy density for a wave vector q„ (8.18}
= 2q for a pair is -,'n(fPq„'/2m*) ~N~' and the density p )' hv~

of flow J = -,'nhq„~%'~'. Note that with v, = Aq„/m* where P = 1/kT Equation (8.18.) leads to values of
= Aq/m and p, = p(q ~' = nmtq t', these become n and p equivalent to those derived by Gor'kov for
p, v', and p, v„as they should.
—', the limit T —+ T, . Note that (8.8) is no longer valid
The coefficients n and p are temperature-dependent when impurity scattering is present. I:n the limit
parameters which were evaluated by Gor'kov from T=O) 6 —+Ap)and
the microscopic theory. They also can be obtained
directly from the equilibrium values of p, /p and of p, = lp/fs ——so.m'Ao/(he'), (8.14)
F. —F, = H', t/Svr in the absence of applied fields. where X = ne'r/m is the conductivity in the normal
The equilibrium value of ~+ ' = n/p, which is that state.
which makes — n~+t'+ —,'p%'~' a minimum. The
~2 P. B. Miller,
minimum free energy per unit volume for this value Phys. Rev. 113, 1209 (1959). Equivalent
results werc obtained by a different method by A. A. Abriko-
of [ef' is sov and L. P. Gor'kov, J. Kxptl. Theoret. Phys. (U.S.S.H, .)
35, 1558 (1958); 36, 819 (1959); Soviet Phys. JETP 8, 1090: —
F, —F„= —nn'/4p = H, b/87r . (8.5) —
(1959); 9, 220 (1959).
 yeA. I »F- LD8 Ig S UPKH, CO»U eT«S 67 &

gh nges in e with
aPP lied«ldso
an es in the energy g
corresponding chang
1
js
" simpy,
~

E = (' +
Fermj. suI'face,
a nd »f'
the Fermi (F D) funct&on
T e 1a t term is the in
'
t l temperatures, it is tion ene gy f ~ tisfies the in teg»ie@u
eg
ost convenient to "
parameter "d ation which g~ve
. &(0
fiu

[1 —2f(E)1
c4
E ' (4.2)
whei1 a «rr nt "
present is g iven in Appen ix h;,
is term besom~~ +2/ p and onee recovers the
determine + d the critical currennt in the S expression
absence of magne ic h ve for minimum In general, onee must evaluatet (4 ]) by numer~
eri. cal
free energy When the gap &s sm ll compared wit
' —n04 g)2 where
pv', lel n~l+I'+ n&l+l' = o
OI'

temperatures, one m
/ = (nn —pv, )/nP . powers of e ~E as discusse in Appendix C. e
The mass How ls consi er g - p
—pv')inPjpv. The integral nee e in
p v. = l:(n~ , may be written
'
which i a maximumm for
w ic is or v, = (n42/8p)'": ko
—2f(E)1
(p.v, )~. . —= -'
3 (~p/P)(n~/8p)
p (8.17)
l:1
G6
E
—2f(~)
The value of p„p at the maximum is 1 —2f (E) 1
= E
p., p 3 (~i 4) (8.18) Ace
1 2f( )
so that a aat the current How
a near T, the gap w m
maximum (4.8)
is reduced to (2, 8)'~2 = 0.81 o e
ral on the right-hand si e
T e of the mass Aow is
K"'"tly r"idl
-(-"")"= (-')" (",'
4.8) converges sufficien for 1
' — =-.' h t the upper limit or
out appreciable error. the variable o
(8.19) integration from e
x + xo . To expan
which variess aas (1 —t)' '. A further
'
discussion o ' we need to ta e
in x„w
critical curre
rent s is g ive11 in e
~xp 8 a
CE OF FREE ENERG
IV. DEPENDENCE GY
G ON GAP
Bxp Bxp 8X
we shall derive expressions nd
e a taken as a r
ee 'y a 1 —2f(e) 1 —2f(E))
g& (xo 4
the author" in w ic = 02ZO + 04$P +
interaction V e (h~ of the
where
gra equ . (8.86) of f
"1 8 1 —2f( ) 7t (8)
for thee free-energy difference
i b et w 2x Bx 8m
45)
as becomes
ducting and normaal p hases The second integra
e ral on the rig
P —F„o ——2N(0)
sp np
(4.8) becomes
—2f(x)
„
Ace

—2' (e) —e —«/E) Id»e 1

X I 2Ef (E) 1
x 7

(4 6)
—X(0)'V
. E1 —2f(E) )de, 1 (4. 1
T and it has been assumeded that
where X(0) is the densityt oof sstates of one sp'
s in at the f(/Aced) is vanishingly small. a i
672 JOHN BARDEEN

gral equation (4.2) for the gap at T = T„we find where

that the first integral on the right-hand side of (4.6) —g, (xo) —
is equal to 1/(N(0) V) . Thus, we have 1 = 2xo' {X[1—2f(X)] —k (x:/x)
Ace X [1 —2f(x)] —x[1 —2f(x)] }dx .
[I —2f(E)]E = g—
(p~) +N 0 V+»— (4.11)
Again, g, may be expanded in a power series in
x02, the first term of which is
In the limit T ~ 0, we may neglect f and find =
go(xo) —aoxo2
, +
g, (pt), ) ~ ln (pa/p, ao) .
with as given by (4.5). In the low-temperature limit,
The integral (4.7) set equal to 1/[N(0) V] is just the x0 large, we have by direct integration
integral equation (4.2) for the gap. From previous
calculations of p 6 as a function of p/p. = T,/T, one
g, (xo) = ln (xo/P. ao) y ' —2(. (2)/xo
—, .
can determine g, (ph). A direct evaluation is given Plots of g&(x) and go(x) as calculated by McMillan
in Appendix C. are given in Figs. 1 and 2.
The first line on the right-hand side of (4.1) may
be written, after an integration by parts,
jo-

L, = 2N(0) {(Aa)' —h(u[(her)' + 6']' '}

gi(X)

+ 2N(0) {E[1—2f(E)] —e[1 —2f(o)] }ds .

0
(4 8) 12

To get rapid convergence of the integrand of the

second term, we may write it in the form X~
Fzo. 1. Plot of the function g&(X), as defined by Eq. (4.4).
2N(0) {E[1—2f(E)] — (6 /s) [1 —2f(e)]
is

—e[1 —2f(o)]}de+ 6 N(0)

"
1 —2f(o)d

(4.9)
gq(X)
.8-
The upper limit of the first integral may now be
replaced by infinity without appreciable error. In
this way we find that LI may be expressed to a
good approximation in the form
12

L = N(0)~'{1/N(o)V+1 (P/P ) —g (P~)} 0

I

1
I

2
I

3 4 5
X

(4. 10) FIG. 2. Plot of the function go(X), as defined by Eq. (4. 11).

The second line of (4.1) may be evaluated with use of (4.7). We thus find

F„—F„= +N(0) 6'{1/N(0) V + ln (p/p. ) —g (ph) —N(0) V[ —g (ph) + 1/N (0) V + ln (p/p. )]'}
= —N(0)~ {[—2gi +» (P/P. )][1+N(o) V» (P/P. )] + N(o) Vg~ + go} (4.13)

Expanding to order 6', we find where t = p, /p = T/T, .

T —0 is
.
F, —F„= N (0) a' {ln —N(0) V ln
t[1 t]
The limiting expression as &

+ ao(PA)'[1 —4N(0) V ln t] + }, F.o —F„= —N (0) 6'

(4.14) X {-', —ln (6/6o) + N(0) V[ln (i1/6o)]'} . (4. 15)
 —.2

CI4 O
CI
Lc
I Z
N

=4

".8

i'=0
-s.o
.2
l

0 .8 I.O l.2

Fze. 8. Free-energy difference (F„(a) —F„) inunits of sN(0)as =

Hs /87r as a
function of 6/A0 for several reduced temperatures. The minima of these curves corre-
spond to the usual equilibrium gap for the corresponding temperatures. The curves
apply to a coupling strength N(0) V = 0.8.

A plot of (P,. —P„)/[-,'X(0) 60] asa function of large fields or currents with random scattering at the
6/Dc, for several different reduced temperatures, surface or that corresponding to l &( $0.
based on calculations of McMillan, is given in Fig. 3. We shall use (2.9) for the magnetic moment and
(3.13) for p, . Since the dependence on 6 is the same
V. APPLICATION TO SMALL SPECIMENS
for both, we have for unit volume
In this section we consider the application of the
theory to a thin film or filament with thickness small Z. = r', y ' (aH'+ l.', )X(O)~.~ tanh (-', PS),
—,

compared with the penetration depth. We suppose (5.1)

that the magnetic field is either transverse or parallel
where a is the symbol in (2.9) and, from (3.13)
to the filament or is in the plane of the film. Scatter-
ing at the boundaries may be taken into account by lp/[&sX(0) 6s]
use of an effective mean free path /, which by our
assumptions, is such that t/(c &(1. We shall use The equilibrium value of 6 is that which makes
(2.14) for the free energy, which takes into account the free energy a minimum. Analytic solutions can
the dependence of x and p, on 6 but neglects other be obtained only for the limits P 6 &( 1 and P 6 )& 1.
nonlinear effects. This approximation is valid when For the former, we use (4. 14) and for the latter
PA « 1, and probably also at lower temperatures (4 15)
In the limit P 6 ((
when l « Ps, since critical currents and fields may
1, we have
then not be so large as to get into the true nonlinear P, —F„= X(0)6' I ln t[1 —X(0) V ln t]
range. It is certainly not valid when PA
t) $&&. As shown
)
1 and
in Appendix B, for the latter case + as(PD) [1 —4X(0)Vln t])
one may still use the concept of a gap varying with + AX(0)hc(-', PD ——,', P 6 ), (5.2)
v. , but the current density is not given by p, (A)u,
where A = -', (aII' + bv', ). The equilibrium value
as assumed in (2. 14). Whereas a calcula, tion similar
to that of Appendix 8 could be made for small speci-
of 6 is given by
mens for a boundary condition corresponding to —{ln t[l —X(0) V ln t] — PA 60 -',

specular reAection, it is diFicult to treat the case of 2a, [l —4X(0) V ln t] —' PALEO —,
674 JOHN BARD KEN

Kith increasing magnetic field, there is a second- The value of y which makes F, a minimum is given
order transition if 6 gradually goes to zero. This by the solution of
means that there must exist a solution of (5.3) as
—-,'4ylnyL —X(0)V —X(0)Vlny] = A. (5.10)
6 —+ 0, which would occur for
If A is larger than a critical value A., the free energy
pAoA = —2 ln t[l —X(0) V ln t] . (5.4) in the normal state will be lower than that of the
It is necessary that the denominator remain positive superconducting state. To determine the critical
for this value of A, or that t be such that value of y = y„we substitute the value of A given

2a2 [1 —4iV (0) V ln t] )— ~~ ln t[1 —X (0) V ln t] .

by (5.10) into (5.9), set F, = F„„and solve for ln
yc:
(5.5)
lny,
The reduced temperature tl, above which the transi-
1 —2X(0) V —I 1 —2X(0) V+ 4[%(0) V]'}' '
tion is second order, is given by the solution of the
quadratic equation 2X(0) U
(5.11)
+ ( ) "] y~ ( y ' ( ' ) For x(0)U = 0.3, lil y. ——0.80 aild
where y = — ln t& A. plot . of t, as a function of X(0) V
is given in Fig. 4. A, = 0.68.
At reduced temperatures above t&, such that (5.4) For a value of A less than A„ the minimum in I',
applies, the critical field II, is given by will be less than F„p and for larger values of A greater
—2 ln —X (0) U ln than F p. At this critical value, a first-order transi-
t[1 t] tion to the normal state will occur. Such a transition
C
pro has not been observed in small specimens of dimen-
Near T„ this expression reduces to: sions of the order of the penetration depth or less.
II'. = 4(l —t)/P. d, a. (5.8)
Numerical calculations are required to determine
the variation of the gap with field for the complete
Values of a for various specimen shapes are given in range of temperatures. Figure 5 is a plot of (6/t4)
Appendix A. vs the parameter A for several reduced temperatures
In the opposite limit of low temperatures, p, and for X(0) V = 0.3, based on the free energy plots of
x are proportional to A. With y = 6/60, as in Fig. 3. For this case, the gap gradually goes to zero,
(4.15),
F, = F„o— X(0)day'
—', I.O

X [1 —21ny[1 —X(0)V»y]} + ~X(0)~,t), .

(5.9) .St

.2

.2 0
0 .i .2 .3 4 .5 .6 .7
I
A poH
N(0) V FIG. 5. Change in gap parameter with A = z aH~ for cou-
Fxo. 4. Reduced temperatures tl, which mark the change pling strength X(0)V = 0.3 at several reduced temperatures.
from a 6rst- to a second-order transition of small specimens the plotted points are deduced from data of Tinkham and
in a magnetic Geld as a function of the coupling strength, Morris (reference 7) on thermal conductivity of a thin
X(0)V. ( 650A) film of indium for t = 0.63.
 CRITICAL FIELDS IN SUPER COND UCTORS

giving a second-order transition, if t & 0.325, and H, varies as (1 —t) while the theory [Eq. (5.8)] for
the transition is of Grst order for t & 0.325. At the small specimens of uniform size predicts H', rather
higher reduced temperatures, t = 0.6 and 0.8, than H. is linear in t near T,. The reason for this
(6/dc)' is very close to a linear function of A or of discrepancy is not known. From the plot of Fig. 6,
H'. Even at t = 0.4, the departures are not very one might expect that at very high fields, such that
large. An asymptotic behavior of this sort is to be t & 0.3, the transition would change to one of first
expected from (5.3) near t = 1, when the first term order. To date, no such first order transition has
of the numerator as well as the ln t term in the been observed, although measurements on lead' have
denominator can be neglected. It also follows from been made at temperatures as low as t = 0.2.
the Ginzburg-Landau theory, as shown by Douglass. ' Values of A(H)/A(0) deduced by Tinkham and
However, it is surprising that this limiting behavior Morris' from measurements of thermal conductivity
is approximately valid at reduced teInperatures as in thin ( 650A) films of indium in high magnetic
low as t = 0.6, or even 0.4. Gelds are compared with theory in Fig. 5. The re-
In Fig. 6 is plotted, again for X(0) V = 0.3, the duced temperature for the experimental run is
t = 0.63. Agreement between theory and experiment
C R ITICAL FIELD
is excellent. Data taken at t = 0.36 (not shown) are
j in much poorer agreement with theory.
.6 Elukhara" has measured critical fields of de-
posited mercury films with thickness d, varying
from about 1 to 10 X 10 ' cm. The fields were
parallel to the plane of the Glms. Measurements made
near T, were found to be in good quantitative agree-
0.3
2.—-—
N(o)V=
ment with predictions of the Gor'kov version of the
Ginzburg-Landau theory as modified to take im-
.I
purity scattering into account (see Sec. III). Sus-
ceptibilities were calculated by use of the local
0 ,I .2 3 .4 .5 .6 .7 ,8 .9 I 0
London theory for a mean free path l (
d (Appendix
A). The theory predicts, as observed, that H. varies
FIG. 6. Critical values of A, = ~ OII,2 at which the transi-
tion from superconducting to normal state occurs for N(0)V inversely with d. In the Pippard nonlocal limit with
= 0.3. Below about t = 0.325, the transition is of Grst order, t & d, one would expect H, to vary as d '/' when d
above it is second order.
is very small. While measurements were made on
annealed films with longer mean free path, the film
critical value of A, denoted by A„at which the thicknesses were too large to expect this limit to
transition from superconducting to normal behavior apply. It was found that critical fields of the an-
occurs. Below the inAection point near t = 0.325, nealed films as well as of the unannealed Glms varied
the transition is of Grst order, above of second order. as 4
Experimental evidence for a change in gap with %e next consider the application of the equations
field and a second order transition comes from (1) to a calculation of critical currents. As noted in Sec.
tunneling experiments of Giaevar and Megerle, ' III, the current density p, v, increases to a maximum
(2) changes of thermal conductivity in thin films and then decreases with increasing v„as the decrease
with field as measured by Tinkham, ' and (3) changes in p, with decreasing gap more than compensates
in the specific heat curves of high Geld supercon- for the increase in v. . Only the values up to the
ductors, which presumably have superconducting maximum can be realized in practice. Since at the
domains of very small size. Morin and his co- maximum there is only a moderate decrease in the
workers" have measured the specific heat of V3 Ga gap, large changes in gap cannot be obtained simply
in fields of 40 kG and of 70 kG, and find that T, de- by increasing the current. A magnetic Geld is neces-
creases with increasing Geld, but the transition ap- sary. Vfe shall first derive the expression for the
parently remains one of second order, contrary to critical current when a magnetic field is present for
the behavior expected in bulk specimens which the limit PD « 1, which may apply either near T,
exhibit a Meissner effect. Their data indicate that or in high magnetic fields near H, .
~3 F. J. Morin, J. P. Maita, H. J. Williams, R. C. Sherwood,
J. H. Wernick, and J. E. Eunsler, Phys. Rev. Letters (to be r4 I. S. Ehukhareva, J. Exptl. Theoret. Phys. (U.S.S.R.)
—
published). 41, 728 (1961); Soviet Phys. JETP 14, 626 (1962).
676 JOHN BARDEEN

form

where
Using (5.8) and (8.18), we may express p, in the

ps
8bX(0)62 c,
2p C2
—v',
—V, (5.18)
by e/m:

where XL,
H.
I2.E. = 4m ',
electrical current density is obtained by multiplying

b 4x pse
m
'
1/2
= H,

is the London penetration depth. Equation

47I
',
b
{emu. ) (5.21)

c, = —(4/62pb) t[1 —X (0) V ln —aII'/b (5.21) may be used to estimate critical current
I ln t] }
densities at temperatures different from zero, if
(5.14) H, &(T) and p, (T) are the temperature dependent
—4X (0) V ln —aH'/b quantities. The dependence of p, on scattering mean
c, = (12a2/3, 2Pb) [1 t) .
free paths should be taken into account, with use of
(5. 15) (8.18) if t &( $2. The values obtained from (5.21) are
somewhat too large because this expression does not
The value of v. = v„which makes p, v, a maximum
take into account the dependence of the gap on v, .
is given by setting
Near T„when P 6 &( 1, we have from (8.19)
d v, (c, —v', ) —(2/. 8.) ' = 0.545.
dvs c2 —v, (5.16) I./I2 F, (5.22)
Near T = 0, with t «f2, we have from (4.15) and
which gives
(5 1)
v', .= ' —, I8c2 —cl —[(Qc2 —c, ) (c, —cl)]'" } . (5.17) I,/I2. , —' I-', [1 —X(0)V]}' '
—, '. (5.28)
-,

The value of (P 6) for this critical velocity is

2
Near T = 0, with no scattering, we have from
1/2 —8 (c2 1/2 (B12)
(9c2 cl) cl)
(p~)2 8 (5 18) '
(9C2 Cl) (C2 Cl) I./I , 2.
—
E— 1.017 (2/8) 0.82 . (5.24)
The solution is valid only when (ph), &( 1, which, An approximate formula valid for all temperatures
in turn, implies cl «c2. In this limit, v, cl/8 and ~ may be obtained by using the relations H. &(t)
(P 6)', 2cl/c2. The critical current is = H2(1 —t'), XL, (t) = X(0)/(1 —t4)'~2, so that,
J, = p, v„[bc,X(0)h,/Pc2](c, /8)' '. (5.19) H. (t)/X(t) = [H2/Xl(0)](1 —t )
l, (1+ t ) . (5.25)

As indicated by (8.18), p, is proportional to To obtain a useful approximation valid for all

near T„and near T = 0 is independent of 6 for temperatures one may express p in terms of the
o = ne2t/mvz,
t ))
g2 and proportional to 6 for t p2. In Fig. 5 is (( conductivity, of the normal state,
and find for t(& $2 (the usual situation for thin
shown how p, varies with A. Note that in the ab-
sence of a magnetic field, A is proportional to v', . films):
Thus, near T„p, decreases linearly with v', Near
((
T = 0, for t $2, 6 and thus p, drop roughly linearly
I. = —; Ho(~/&2)'"(1 —t')'"
with A or v', until a critical field is reached at which 500.
0 i/2

an abrupt first-order change to the normal state

occurs . The slope in the limit A T &0, is ~0, — Since H2 is proportional to D2 and p, to 6&, varies I.
as T', '/'. A rough formula based on a free-electron
dA 1 dp,
dA p, dA 2[1 —X(0) V] ' (5.20)
The current density, proportional to p, v. , increases
I. = (8t/)t'" T{l —t')" X 10'A/em'. (5.27)
to a maximum and then decreases with increasing The best experimental observations of critcial
v, ; the critical current density is given by the maxi- currents in thin film are in fair agreement with
mum of the curve. Several limiting forms for the (5.26), both as regards temperature dependence and
critical current density, J.
= (p, v, ) „., are of interest. . magnitude. Care must be taken to avoid eGects of
It is convenient to normalize in terms of the critical magnetic fields produced by the current on the cur-
density JF.~. given by the free-energy criterion for rent distribution by use of a compensated geometry,
a bulk specimen as given in (2. 18). The corresponding and also as pointed out particularly by Bremer and
 CRITICAL FIELDS IN SUP ERCONDUCTORS 677

Newhouse" to avoid heating effects by use of short basic studies because the current distribution may
pulses. As far as the writer is aware, no experiments be altered by the magnetic field and not be uniform
have been done in which both precautions have been across the film. Nevertheless, results have generally
taken. Ginzburg and Shalnikov" and Alekseevskii been in qualitative accord with theory.
and Mikheeva" have used a compensated geometry
VI. CONCLUDING REMARXS
with a specimen in the form of a thin film deposited
on the outside of a circular cylinder. They found The discussion has been confined to specimens
that I, varies as (1 —t)'~' near T., as predicted by with at least one dimension small in comparison
the 6-L theory. However, they used direct currents, with the coherence distance, so that the variation
and so may have had heating effects and observed of the gap parameter 5 with position may be
the critical current for propagation of a normal- neglected. To determine changes in gap with applied
superconducting boundary rather than a true critical fields and currents, we use a variational method and
current. "
Mercereau and Hunt" have observed the treat 6 as a free parameter, chosen to make the
Aux trapped in thin film rings of tin of very small over-all free energy a minimum. As discussed in Sec.
dimensions. With films less than 700A in t,hickness, II, there are three contributions to the free energy:
they find current densities greater than 10' A/cm' (1) that from the pairing interaction, which is a
and near T, the predicted temperature dependence minimum for the value of 6 determined from the
of critical current. A plot of their data is given in usual gap equation; (2) that from the magnetization,
Fig. 7. Observations with pulses so as to avoid heat which decreases as 6 decreases, allowing further
effects have been made by a number of workers"" penetration of the field; and (3) that from the
on planar film strips, mainly in connection with kinetic energy of the supercurrents. Section IV and
cryotron studies. This geometry is unfavorable for Appendix C are concerned with the calculation of
the first of these, the free energy difference F,o —I'.
between superconducting and normal states for
Ioo
general values of A.
To determine the magnetization, 3II(II, A), and
the supercurrent density, J, (v„A), for a velocity,
v„of the ground-state pairs, we have generally taken
~a
4 the first nonvanishing terms, those proportional to
Ch
ias
4
H' and v'„respectively. Thus, we take into account
IO— eAects of changes of gap, but neglect other specifi-
cally nonlinear effects. This procedure is justified
for temperatures near T„and, in fact, gives results
equivalent to those derived from the Gor'kov version
of the Ginzburg-Landau theory. This latter theory
is reviewed in Sec. III. It is likely that the method
gives reasonably satisfactory results at all tempera-
I If I
tures for fields and currents up to the critical values
O.OI O.I
(Tc- T) if the mean free path of the electrons is much less
Fig. 7. Critical currents of thin ( ~700A) Glms of tin as a than the coherence distance, the usual situation in
function of T, —T, from data of Mercereau and Hunt small specimens. Applications of the theory to
(reference 18). Theory predicts a slope of 1.50.
various problems and comparisons of theory and
experiment are discussed in Sec. V.
5 J. W. Bremer and V. L. Newhouse, Phys. Rev. 116, 309 While agreement between theory and experiment
(1959).
6 N. I. Ginzburg and A. I. Shalnikov, J. Exptl. Theoret. is good in general, there are some discrepancies.
—
Phys. (U. S.S.R.) 3'7, 899 (1959); Soviet Phys. JETP 10, Further experiments on well-defined specimens are
28' (i.960).
7 N. E. Alekseevskii and M. N. Mikheeva, J. Exptl. desirable. In particle, it would be of interest to study
Theoret. Phys. (U. S.S.R.) 38, 292 (1960); Soviet Phys. — combined effects of field and current on specimens
JETP 11, 211 (1960).
8 J. E. Mercereau and T. K. Hunt, Phys. Rev. Letters 8, of simple geometry. Measurements of the change of
243 (1962).
gap with field at very low reduced temperatures
9 The problems encountered are discussed in A. M. Kolchin,
Yu. Q. Mikhailov, N. M. Reinov, A. V. Runyantreva, A. P. (t
Smirnov, and V. N. Totubaliv, J. Exptl. Theoret. Phys. 40,
( 0.3) are required to see whether or not the
—
1548 (1961);Soviet Phys. JETP 13, 1088 (1961), where other
predicted first-order transition occurs.
references to the literature may be found. Rather than use the variational method, one could,
 JOHN BARDEEN

from a more basic point of view, rederive an integral APPENDIX A. MAGNETIC MOMENTS
equation for the gap in the presence of fields and FOR VARIOUS SPECIMEN SHAPES
currents, and thus determine changes in 6 with II Here we shall tabulate results of calculations
and v, directly from an integral equation. This latter which have been made of the magnetic moment of
method is applied in Appendix 8 to determine specimens whose smallest dimension d is less than
changes in gap with a uniform current How. In more the coherence distance $0 (typically of the order of
general cases, the integral equation may be derived '
10 ' to 10 cm). Different methods have been used
from the general Gor'kov equations, valid for all to make such calculations. The London theory, with
temperatures, or directly from microscopic theory. a penetration depth X given by (3.13), may be used
In the few cases where comparisons can be made,
results obtained by the more general methods are
when the mean free path When / t((d.
d, a non-
local theory must be used to determine the current
)
equivalent to those obtained by the variational density from the vector potential. In this case the
method. It would be desirable to make further calcu- results depend to some extent on the boundary
lations by the more general methods, particularly in condition assumed for scattering of electrons from
cases where nonlinear effects other than those repre- the surface. May and Schafroth" have given a
sented by changes in gap are important. general method by which the magnetic moment of a
We have not discussed the closely related problem specimen of any shape can be determined for the
of changes of gap with rotation in nuclei. To first case of specular reHection. Random scattering is
order, the Coriolis force in the rotating frame is more dificult 4o treat; a method suggested by Pip-
equivalent to a magnetic field. As pointed out first
by Mottelson and Valaten, "
one expects, as for the
pard can be used for specimens of simple shape.
Recently, Hauser and Helfand" have given calcu-
corresponding magnetic case, that the gap will de- lations for several specimen shapes according to the
crease with increasing rotational velocity, and even- London theory and also for the nonlocal theory with
tually go to zero. These authors derived a modifmd use of the method of May and Schafroth. Included
gap equation valid to second order for the rotational were a sphere, a cylinder in a transverse field, a
velocity. Recently Grin" has treated the problem cylinder in a parallel field, and a thin plate in a
from the general Gor'kov equations. parallel field. They took into account effects of a
mean free path in the expression for the current
ACKNOWLEDGMENTS density. Earlier calculations have been made with
The author is greatly indebted to Dr. D. H. use of the random scattering boundary condition
Douglass, Jr. , Dr. J. J. Hauser, Dr. E. Helfund, Dr.
for a sphere by Whitehead and for a thin plate by
Y. Nambu, and Dr. M. Tinkham for illuminating Schrieffer. In Table I are given the results of these
discussions of some of the problems treated herein,
and for sending manuscripts of their own contribu- Tash, x I. CoefFicients for calculation of magnetization of
tions in advance of publication. Part of the original small specimens from Eqs. (A6) and (A7).

work for this article was done at the Summer Insti- Random London
tute for Theoretical Physics at the University of Specimen scattering Specular limit
Wisconsin, supported by the National Science Foun- geometry g reflection g gL,

dation, in the summer of 1961. While there, the I Thin Glm parallel
author had the benefit of the critical mind and clear Geld 0.876'b 0.500d 0.333'
II Cylinder parallel
insight of Professor Wigner in frequent discussions Geld 0 100g 0.122d 0.125"
of this and other problems of superconductivity. W. IIICylinder trans-
verse Geld 0.200g 0.316d 0.250"
L. McMillan has been of great help in preparing IV Sphere 0 0625c 0.0845 ' 0.100d'
this review; he is largely responsible for Appendices
A and C, has made most of the numerical calcula- a J. R. Schrieffer, Phys. Rev. 105, 47 (1957).
E. T. Rodgers, Ph. D. Thesis, University of Illinois (1960) (unpublished).
tions reported, and has checked many of the equa- 0 t . S. Whitehead, Proc. Roy. Soc. (London) A238, 175 (1956).
& J. J. Hauser and E. Helfand, reference 23.
tions. e R. M. May and M. R. Schafroth, Proc. Phys. Soc. (London) 74, 153
(1959).
f F. London, Suyerjluids (John Wiley k Sons, Inc. , New York, 1950).
& Hitherto unpublished calculations of %. L. McMillan.

B. R. Mottelson and J. G. Valatin, Phys. Rev. Letters 5,

511 (1960). ~2 R. M.
May and M. R. Schafroth, Proc. Phys'. Soc.
2r Yu T. Grin', J.— Theor. Phys. (U. S.S.R. )
Exptl. 41& 410 (London) 'l4, 155 (1959).
(1961); Soviet Phys. JETP 14, 320 (1962), s3 J. J. Hauser and E. Helfand, Phys. Rev. 12'7, 386
(1962).
 CRITICAL FIELDS IN SUPERCONDUCTORS
calculations for the limit, d (& to, together with field, (III) a small cylinder of radius r in a transverse
hitherto unpublished calculations of McMillan for field, and (IV) a small sphere of radius r. For the
a cylinder in a transverse field and in a parallel Q.eld. nonlocal theories, the coeKcient a may be expressed
The general expression for the current density, in the form
including the effect of a scattering mean free path a = gr /Xr. fpHp, (A6)
is10, 12 .
where g is a dimensionless geometrical factor. Values
p(r) =—4~c&r)o
3 R[R A(r')] J(R, T)e
g4 T for g are listed in Table I for both random scattering
and specular reQection.
where R = r —r'. For specimens with d (& $p, one In the London limit, a is proportional to r' rather
may replace J(B,T) by J(O, T) without appreciable than r' and may be expressed in the form
error, where = 2 2 2
a gJr jX Hp. (A.7)
J (0, T) = (A,/A)
(6/Ao) tanh (PA/2) (A2) . Values of the dimensionless coeKcient g& are also
When in addition / (& d, so that A(r) does not vary listed in Table I, prepared by McMillan.
much over a mean free path, one may replace A(r')
by A(r), carry out the integration, and find the APPENDIX B. DIRECT CALCULATION
London relation OF CHANGE OF GAP WITH CURRENT24

j (r) = —(c/4n-X')A (r), (A3) We discuss in this section a direct calculation of

the change in gap with current Row which applies
with the penetration depth X given by (3.13).
for all T. The problem is an idealized one because
For the case of random scattering with siInple
we disregard effects of magnetic fieMs and assume,
specimen shapes such as those considered here, one
contrary to fact, that one can have a uniform current
simply carries out the integration over the specimen.
fI.ow in a bulk specimen. It is of interest because the
In the limit d/X ~
0, the field in the specimen is calculation goes back to first principles and because
little changed from the applied field, so that one
the change in gap is determined directly from an
takes for A(r) the vector potential of the external
integral equation. Results near T, are equivalent to
field H. The magnetic moment is
those obtained from the 0-L theory.
= 1 The calculation is similar to that required for the
M r X j(r)dr
2c
36 tanh (-', Pd), (r X R)[R.A(r')],
two-fIuid model, but is extended to arbitrarily large
currents. We consider first a situation in which the
ground-state pairs have zero net momentum, but
(A.4) for which there is a distribution of quasi-particle
in which we have assumed d, and replaced l)) excitations giving a net current J.
Let fop be the
e ~~' by unity. Comparing (AB) and (2.9), and using
probability of an excitation in the state kf and
the relations Hp/8m = -', X(0) LQ, $o = hvar/pr&o, f o, that in —lr $. Then the mass ffow is

A = 4~) 'I/c', we find for the coefficient a of (2.9): J. = Q hk. (fo, —f oo) . (Bl)
a= &r I
o
0
p
0
dr dr', (r &4,
X R)[R A(r')]
(A5)
We want to determine the fo to give a minimum
free energy subject to a given J.
We use the BCS
model for which the pair interaction is — V in an
which is the expression used by McMillan for his
energy zone o&~ & h~ about the Fermi surface,
calculations.
where oo —A, 'k'/2m —p. If v is a Lagrange multiplier
~

The four specimen geometries considered are (I)

for J„we have [cf. Fq. (3.16) of reference 9].
a thin film of thickness 2r in a parallel magnetic
field, (II) a small cylinder of radius r in a parallel 24 Based on the thesis of K. T. Rogers, reference 8.

~+ v. ~ = Q oi(f~. + f - '. ) + 2 Q irk((A

oA. (1 —fo. —f op)-
" —f o&) —V Q —ho)ho(l —hg, )]'"(1 —fo„—f og)
+ 5 'v Q BA;
(fo~ [h~(l
Ilk )(AQl

X (1 —fo'o —f o'o) + kP -'I g foo ln f~o + (1 —foo) ln (1 —fo, )

+f go ln f oo + (1 —f oo) ln (1. —f ~~-)],
680 JOHN HARDEEN

where p = (kET) ' and 1't& is the probability of a to transfer of an electron from one side of the Fermi
pair in k. The values of f& and hp are determined so sea to the other. This criterion is (depairing condi-
as to make I/ +
v J a minimum. The solution is tion)
similar to that for v = 0:

ttp = (1 —es/E. )
-', (B3)
, m (hkr/m + v, ) —
—' -', m()'tkE/m —v, ) ) 2h, (B10)
Ep=
2 2
ep+& 2
(B4)
hkgv, & A.
1
The gap decreases very rapidly when v, exceeds the
1 + exp Ltt (E, + v k)] ' A,
critical value. The maximum in the Row occurs for
1
1 + exp [P(E, —v hk)] (B5) v„= 1.03 ep/AkE, (Bll)
To get a state corresponding a velocity v, of the with the maximum being only slightly greater than
ground pairs and a normal component of current that corresponding to the depairing condition:
equal to zero, one may displace the above configura-
tions in momentum space by mv, and then take
J = 1.017 Nep (0)/vE ~ (B12)
v = — v, . An equivalent solution can be obtained This is roughly (2/3)'/2 or about 80% of the value
from the general Gor'kov equations for the thermal corresponding to the free-energy criterion.
Green's function, valid for all temperatures. "
%e may write to a close approximation APPENDIX C

v'V = ~r&J cos 0 = ~I('&x (86) In order to find the free energy as a function of
6, we must evaluate the functions gr(xp) and gs(xp)
where k& is the magnitude of the wave vector of the numerically. It is convenient to expand the F-D
Fermi surface, 0 is the angle between v and k and x function f(X) in powers of e E. From (4.4)
= cos 8. The equation for the energy gap then be-
comes 1 —2f(x) 1 —2f(X)
g(*) —= S x
1 1 "dc
E —2f(x) "f(X) dx
N(0)V 2

X
p

1,
1
~, 1
(+ —
1
g I dx +2 X
P (@+gx)
1 ~e '
gz)
e P
=ln
,Ap
+2+(—1)"" n=l
(B7)
= AA:J V, . The integration
" exp [—n(x' + x', )'"]
where a over angles may 2 dx (C1)
2)1/2
be carried out explicitly to give ( +
1
t~
1
t/Es
+ s Here X = (x'+
xp)'/2, a, nd we make weak cou-
pling approximations, [(Ap/hcc)2 1; f(P, hcc) &(1], «
throughout.
In the limit v = a = 0, this equation reduces to Substituting x = xo sinh y, we find
the usual equation for the gap. At temperatures
near T, such that Pct «1
and Pep(T) 1, one may (( " exp [—n(x' g x', ) ' '] -nzp sinh y
expand in a series in (Ptt)2 and (Pep)2. To terms of e Qg
( +
2)1/2
2

order v2, one gets

= Kp(nxp) .
[0 ~(&)]' = [8 '/7t (3)] (1 —t) —l (P.&k )'v', (B9)
which is in agreement with (3.15)
In the low temperature limit, P — + ~, there are
no excitations formed and thus no change in
until the velocity v, reaches the value for which it is gg (xp) = ln
P.kp
+2 g ( —1)"+Kp(nxp)
n=1
. (C2)
favorable to form pairs of excita, tions, corresponding
W. Magnus and F. Oberhetinger, Formulas and Theorems
26
22 K. Maki and T. Tsuueto, Progr. Theoret. Phys. (Kyoto) for of Mathematicat Physics (Chelsea Publish-
the Functions
2'7, 228 (1962). ing Company, New York, 1954), p. 27.
 CRITICAL FIELDS IN SUPE RCON D UCTORS 681
We may evaluate g&(xp) in a similar way. From remaining integrals of (C8) are easily evaluated.
(4.11) We find

— = ' + ln —2i (x)/x;

gp(xp) = 1—2 [1 —2f(X)] g, (x, ) —, (x,/P. t),p)
Xp
= 4 Q ( —1)""[Kp(nxp) + ' '-],
2 (C4)
2x
[1 —2f(x)] —x[1 —2f(x)] dx
"
where f(x) is the Riemann zeta function. The
~ cAoo —2f(x) Hankel functions E.
(x) are related to the Bessel
=1— p Xp
1 2x
dX functions of the third kind" H„(x) by

K„(x) = (s-i/2)e"" 'H„"'(ix) .

+ p
Xp 0
[Xf(X) —xf (x)]dx . The Hankel functions decrease exponentially for
large values of the argument allowing the summa-
We again expand f(X) in powers of e ~ tions to be truncated at n 5/xp. The functions
(C2) and (C4) are
Xf (X)dx =, Q ( —1)
"+'
gr(xp) and gp(xp), calculated from
given in Figs. 1 and 2. Using these functions and
(4. 18), we can calculate the free energy as a function
—n[(x'+ x:)]' 'I[(x'+ x', )]"dx. of 6 in the absence of external fields or currents.
exp I The free energy as a function of 6 for X(0) V = 0.8
is given in Fig. 3. Finally, we can And the free energy
Substituting x = xp sinh y as a function of 3, when both a magnetic Geld and
current Row are present, and minimize this free
X Xdx=4 —1"+' energy with respect to 6 to determine the gap 6 in
the presence of the field and current. The energy
gap as a function of the field has been determined, for
exp ( —nx, sinh y)(1 + sinh'y)dy X(0) V = 0.8, with the free energy given by (4.18)
and (5.1), and is given by the solid lines in Fig. 5.
=4 g ( —1)"" Z. (n*.) y »K. Zahnke and F. Emde, Tables of Functions (Dover
Publications, Inc. , New York, 1945}, p. 236.
28 These functions are tabulated in reference 27, p. 236 and
where Ei(x) is the Hankel function of order one. The elsewhere.

RE VIE W S OF MODE RN PH Y SIC S VOLUME 34, NUMBER 4 OCTOBER 1962

Xo1:e on tie . .nterac1:ions ~e1:ween 1.. ie S sins oI:

:Y.:agnetic ..ons or :5'uc. .ei in .Y. :eta. .s
J. H. VxN VLEcK
Lyman Laboratory of Physics, Harvard University, Cambridge, JVassachusetts

&HE coupling between spins of magnetic ions, or by Frohlich and Nabarro. 2 The resulting nuclear
of nuclei, which results indirectly from the inter- coupling is, of course, very weak, but capable of de-
action of such spins with those of conduction elec- tection in some metals by nuclear resonance. Calcu-
trons in metals has been the subject of a number of lations of the Zener-Frohlich-Nabarro (ZFN) type
papers. Zener' proposed that this indirect mecha- are incomplete because they neglect the effect,
nism is the cause of ferromagnetism. The correspond- essentially a second-order or polarization one, of
ing calculation for nuclear spins was made still earlier the matrix elements which are nondiagonal in the
s H. Frohlich and F.R.N. Nabarro, Proc. Roy. Soc. (London)
i C. Zener, Phys. Rev. 81,440 (1951). Al'75, 882 (1940).