NON-LINEAR CONDUCTION IN SUPERCONDUCTORS

B. D. Josephson

Dissertation submitted in application for the degree of

Doctor of Philosophy
at the University ' of Cambridge

THE BOt\RD OF RESEARCH STUDI.ES

APPR(l'f~ :· , ...is l)l(S'"'""" rioN

FOR THE ffl. • UEGh.. ,i'.'."7 DEC 1964

Trinity College,
Cambridge.

August 1964.
 i

PREFACE
The work described in this dissertation was carried out at the

Roy al Society Mend Laboratory, Cambridg e. The dissertation is not

sub stantially the same as any which I mi g ht have submitted for a

degree or di p loma or other qualification at any University other

than that of Cambridge, and i ndeed no part o f it has already, or

is bei ng concurrently , su bmitted for any such de g ree, diploma

or other qu alification.

Througho ut the text, ackn owledg e ment of t h e work of ot h ers

has been made explicitly or by references. In part I, c h apters

1 and 4 summarize the work carried out by others in the field and

are essentially not original. The experiments de scribed in

chap ters 2 and 3 are believed to be original as is also the

applicati on of the calculation of t h e a p pendix to chapter 3.
Part II is an account of a field of research ori ginated by the

author and is nearly all ori ginal, exce p t where results obtained

by oth ers wo rkin g in t h e field h ave b een quoted for completeness

or for illu strative pur p oses. In such cas e s t h e fact t h at the

r esult is not due to t h e p resent author is made clear in the te x t.

It is, however, probably inevitable that in t h e e xp ository p arts

of the work ideas picked up in conversation with o thers have been

incorpo rated.

I am g rateful to my supervisor, Professor A. B. Pippard , for

sug g esting t h e experiment described in Part I of t h e dissertation

and for his e v e r rea dy advice on how to carry it out and ove rcome
 ii

the diffi culties that arose; to Dr. P. W. Anderson for many

stimulating discussions during his stay at Cambridge during t h e

year 196 1-2; to t h e wo rkshop staff for the c o nstruction o f the mai n
parts of the appar a tus and t h e supply of liquid helium, and for

t h eir he l p ful advice on matters of design and technique; and to

t he many o ther me mbers of the staff and research students with

whom I have had useful discussi ons on experimen tal and theoretical

top ics. I should also like to thank Trinity College for their

g enerous financial support throughout t he period of the research

project , and the DepaEtment of Scientific and Industrial Research

for providing par tial financial support for the first two years.

th
August 10 , 1964.
 CONTENT S

Preface i

Introduction 0

Part I Field dependence of the surface i mpedance of supercon ductors

Chapter 1. General background

(i) The pene tration depth 1

(ii) Temperature dependence of the penetration depth 2

(iii) Field dependence of the p enetration depth 3

(iv) The Ginzburg-Landau theory 4
(v) The surface i mpedance 5
(vi) Effects of non-linearity: the quasi-static limit 6

(vii) Frequency-dependent effects 9

(viii) Previous experimental res ul ts 9

(ix) The present experiment 10

Chapt e r 2. The exp eriment

(i) The resonance meth od 11

(ii) The tran s mission method for determining resonant

frequency 12

(iii) The apparatus 1: the cryostat 13

(iv) Th e a pparatus 2: e x ternal circuitry 15
(v) Specimen preparation 17

(vi) Th e preliminary experiments and t h e consequent

modifications 25
(vii) Miscellaneous ob servations a nd calibrations 30
 iv

Chapter 3. Results and discussion

(i) Specimen characteristics 33

(ii) The field dependence of the penetration depth 34

(iii) Err ors 34

(iv) Discussion of t h e results 40

(v) Summary of the discussion of results 43
Appendix The unconvolution of the inte grals occurring in
the transverse field confi guration 44

Chapter 4. Theory 48

Part II Supercurrents through barriers

(i) Introduction 52
(ii) Basic phenomenolo gical formulae 57
(iii) Thermodynamics 62
(iv) Equilibrium properties 64
(v) Non-equilibrium properties 68
(vi) Mi crosc opic theory 78

Appendix Calculation of Hc
1
92

References 94
 v

Location of tables, diagrams and photographs

N.B. In t his list * denotes 'opposite' and - denotes 'beneath 1 •

Fig. ~ Fig. ~
1 11* 14 34*-
2 12* 15 34*
3 13* 16 L~6*
4 15* 17 69*
5 20* 18 87*
6 21* 19 88*
7 23*
8 27*
9 30* Table
10 31*
11 33*
34* ___ I 33
12 II 35
13 34*-- III 36
 0

INTRODUCTION

Part I of this dissert ation is concerned with the problem of

the magnetic field dependence of t h e surface impedance of super-

conductors, with particular reference to tin. In c hap ter 1 the

predictions of the simple theor y for the behaviour of the surface

i mpe dance for different frequencies and fieid confi gurations are

described and compared with the results of pr evious exp eriments.

Chapter 2 deals wi th the experimental side of the p resent work,

where a . frequency of 170 Mc/s was used, and in chapter 3 t h e

results obtained are discussed. In chapter 4 the attemp ts which

have been made to account theoretically for the e xp erimental

obs erva tions are reviewed.

Part II of t h e dissertation is con cerned with another problem

in superconductivity, namely the behaviour of superconducting

systems p artitio ned by thin barriers of substances which in bulk

are not superconducting. The t h eory of such systems is developed,

the c onsequences investiga t ed in soDe detail and the present

e xpe rimental situation reviewed.

 Part I

FIELD DEPENDENCE OF THE SURFACE I MPEDANCE OF SUPERC ONDUCTORS

Chapter 1. General backg round.

(i) The penetration depth

One of the most basic properties of superconductors is the

Meissner effect, discovered in 1933 by Meissner and Ochsenfeld,

who showed that pure superconductors in a magnetic field behave

as if the magnetic induction~ in their interior is zero. This

observation suggested the question of how fast a magnetic field

applied to the surface of a superconductor falls off with distance

from the surface. An answer to this question was provided by

the theory of F. and H. London (1935). This theory assumed the

existence in a superconductor of 'superconducting electrons'

which under the influence of an electric field accelerate freely

without encountering resistive forces. The basic equation of

the Londons' theory for superconductors in equilibrium is the

follo wing :

cAcurl !!_ +] =0 ( 1)

where /\ = m/n s e 2 , m, n
s
and e being the mass, number density and

charge respectively of the supercobducting electrons. From (1)

and Maxwell's equations can be deduced the penetration equation

(2)

which shows that B falls off e xponentially with distance from the

surface in a characteristic distance A, the penetration depth, given by

2 2 1
).. = (me /4nn e ) 2
s
(3)
 2

In the general case when the Londons' equations may not be obeyed,

it is convenient to define the penetration depth as

00
JH(z)dz
0
). = (4)
H(O)

where H(z) is the field at a depth z below the surface. Wh en

the Londons' equations are valid, H(z) = e-~I.U H(O), so this

definition reduces to the previous one.

(ii) Temperature dependence of the penetration depth

Ori~inally the Londons identified n with n, the number density

s
o f conduction electrons, which occurs in the discussio~ of such

normal state properties as the conductivity and the Hall effect.

This predicted a penetration depth of the order of 10 -6 cm. When

measurements of the penetration depth were carried out it was found

to be considerably greater, indicating that n is much smaller than

s
n. Furthermore, it was found that as one approaches the transition

temperature T the penetration depth tends to infinity, so that n

c s
tends to zero at T.
c
The Londons' theory may be combined with the two-fluid model

of Gorter and Casimir (1934), according to which a superconductor

consists of an equilibrium mixture of two fluids, composed of

'superconducting electrons' and 'normal electrons' r e spedtively.

n is then the number density of the superconducting component

s
(alternatively it may be thought of as an order parameter, in the sense

of the Landau theory of phase transitions (Landau and Lifshitz 1959)),

and T is t h e point at which n becomes zero and the characteristic

c s
 supercondu cting properties disappear (through/\ becoming infinite).

(iii) Field dependence of the penetratio n depth.

It is to be expected that the proportion of supercondu cting

electrons will be a function n ot only of temperatur e, as in

section (ii), but also of magnetic field. We know in p articular

that application of fields greater than the critical field causes

a supercondu ctor to chang e into the normal state (n = O). One

s
would therefore e xpect that in fields less tha.h the critical field

n might be reduced in t ::ne penetration region, with a correspond ing

s
increase in the penetration depth.

A more sophistica ted ar g ument giving rise to the same conclusion

is due to H. London (unpublish ed, see Pippard 195 0), and is as

follows. If we i g nore the penetration region, and suppose that

the specimen has such a shape that demagnetizi ng effects can be

ignored, then the mag netization per unit volume of a super conductor,

M, is H/4n, where H is the applied field. The effect of the finite

distance of penetration of the field is to reduce M sli g htly. Since

the penetratio n depth increases with temperatur e, it follows that

(oM/oT)H is ne gative. By a Maxwell relation it fo l lows that (oS/oH)T

is positive, so that the entropy of the supercondu ctor increases

with H. This entropy increase is :presumably associated with a

decrease inns, since the entropy of the normal state (ns=O) is

known from specific he at. measuremen ts to b e gr eater that that of the

supercondu cting state (n

s
> 0).
(iv) The Ginzburg-Landau theory

A theory which makes det a iled predictions of the field dependence

of the penetration depth is the Ginzburg-Landau theory (Ginzburg

and Landau 1950). This differs fro m the two-fluid model in using
2
a complex or der parameter '{J• l tl is proportional ton
s
of the

two-fluid model, and the appearance of an order parameter with phase

as well as amplitude reflects the wave nature of the electron in

quantum theory, which manifests i tself on a macroscopic scale in

superconductivit y .

The Ginzburg-Landau theory (hereafter referred to a s the G-L

t h eory) p rocee d s by applying the variational principle to a suitable

thermodynamic potential (Gi bbs free energ y). The main advance of
1
the G-L theory is the inclusion of a term
2ni
containing the magnetic vector p otential!, where according to

Gor'kov (1959) e* is twice the electronic charge. Th e equili brium ·

state of the s y stem in the G-L t h eory is obtained by minimizing the

Gibbs free ener gy with r e spect to variati ons of both f'and A.

Equations are t h en obtained fro m whi ch the properties of a

superconductor in a magnetic field can be deduced. In weak fields

these reduce to the Londons' equations.

The G-L equations at a given tempera ture involve only two

adjustable parameters, which can be determined from two experimental

parameters such as the critical field H and the s mall field

c
penetration depth A (0). From these may be derived the important
 5

dimensionl ess quantity

Experiment shows that for pure tin K = 0.16 (Lynton 1962). The
G-L equations make the followin g prediction s for the field dependence

of ). (Ginzbur g and Landau 1950, Sharvin a nd Gantmakher 1961):

). (H) -- 1 (o){ 1 + a.(H/H ) 2 + l3(H/H) 4 + ••••• ) (6)

c c
where a. = ~ ( K + 2 ...J2) (?)
()(. + .v2 )2.

and 13 = 0.047 l<.

2 2
(1 - 5.1)(. + 7\.(.. ) (8 )

for K-l..<. 1. For Sn, 13 is very small and the quadratic approximat ion

for )(H) should be quite good for all fields up to the critical field.

It should be noted in particular that since a.;>O the increase in )

with increasing H expected from simplified ar guments is confirmed

by the G-L theory.

(v) The surface i mpedance

There exist no methods using only static fields capable of

measuring the penetrati on depth of a supercondu ctor with sufficient

accuracy to give useful informatio n about its field dependence ,

and generally methods using hi gh frequency fields are employed.

These measure a quantity known as the surface i mp edance, defined

a s the electric field at the surface divided by the total current

crossing unit length of surface in a direction parallel to the

electric field. These two quantities will in general not be in

pl;lase, so that the surface i mpedance Z is in g eneral complex. Its

 real and imaginary parts are known as the surface resistance Rand

surface reactance X respective ly. The surface i n ductance L may

be defined by the familiar relation X = wL, where w is the angular

frequency.

These quantities may be related to the penetration depth in

the followi ng way. Suppose that the surface of the specimen is in

the xy plane, with the electric field and currents in the x

direction and the magnetic field in the y direc tion. We have fro m
1 oB
the Maxwell relation curl E =-c 5t:
E (o) =-iw
X C
Jao Hy (z)
O
dz (9)

and fro m the relation curl H = 4nj/c (ignoring displaceme nt currents):

R (0,) =- 4 n foo .i (z) dz ( 10)

Y c Jo --x
It follows from the definition of Z that
E (o) -~
. JeoH (z) dz
x
Z = --,.cic,~~~~ __ · c · 9 y = 4niw;\./c 2
0
J jx< z) dz - :n Hy ( O)
( 11)

SaoG Hy (z) dz
where as in (4), but the H may now be
H (O) y
y
complex. In general A will also be complex, though it is clear
that in the limit w -4, 0 it will simply reduce to t h e static penetration

depth defined earlier. Clearly, then,measu rements of t h e surface

i mpedance at sufficient ly low frequencie s can be used to give

informa tion a b out the penetratio n depth.

(vi) Effects of non-l i nearity; the quasi-stat ic limit

Experbient ally a situati::m in which a s mall a ltermating magnetic

field is superimpose d on a ste ady field is often used. If non-linearity

is present the steady and alternating fields cannot be considered

in isolation and the results of the pre~ing section must be

modified. It should be noted in passing that from definition (4),

a field-dependent penetration depth implies the existence of

non-linearity. We shall, however, concern ourselves only with the

case in which the alternating fields are so small that we need

consider only quantities which are of first order in them (though

quantities of higher order in the statie fields will have to be

taken into account), and ac surface impedance can be defined as usual.

In this section we shall consider only the quasi-static limit,

in which the freq u.ency of the alternating fields is assumed to be

so low that the superconductor is at each instant in the e~uilibrium

state corresponding to the p articular magnetic field pre s ent at

its surface at t h at instant.

Following Ginzburg and Landau (1950), we ne g lect anisotropy

effects, so that 1. may be written as a function of H(O):H , so that

0
):!!(z) dz = lio (12)

for quasi-static fields, ').. being the - field-dependent static penetr-

ation depth. For small variations f£!o

f~g !f(z)dz = ~ (H 0 ) S lio + lio ~ (H0 ) ~H 0 (13)

(note the distinction between &H0 and £ lio, b H0 bein g the chang e

in the amplitude of H ).
0
 In the present case we may regard the oscillati ng componen t of the

field as a small variation f/g. The surface i mpedance is then given

by the following modifica tion of (11):

,v 2
Z = 4n:iwl./c ( 11 I )

oO

where now
S0 ~Blz)dz.
( 14)
~!{o
Let us now consider two spectial cases: (i) ~!!o 1 E.o and (ii) S'!!,0 11E.Q,
which for e x periment al reasons are known as the longitud inal and

transver se cases respectiv ely. In the longitud inal case we have

&H
0
= O, and we obtain from (13) and (14) simply

'J. (Ho)::: A(He) (long itudinal ) ( 15)

In the transvers e case all magnetic fields lie in a fixed plane, so

that (13) redttces to

r:'oH(z) dz = (A1(Ho) + HoX (Ho)) ifH()

from which we deduce

(transve rse) ( 16)

This formula is valid for any penetrat ion law and assumes only

isotropy and the quasi-st atic approxim ation. A more general relation
can be derived if anisotrop y is present.

A case of particul ar interest is when A, depends paraboli cally

on H (as will always be the case for small fields). Assuming

0
~ ( H) = A ( 0) { 1 + a. ( H/H c )2 }
f- (0) f >2 l
1
we have A (H) + H ').. ( H) = 1 + 3a. ( H/Hc
so that as pointed out by Ginzburg and Landau the field dependen ce of

the surface impedanc e in the transver se case should be three times

 as great as in the longitudin al case in the quasi-stat ic limit.

It should be noted that this derivation makes no assumption s as to

the detailed form of the penetration law.

(vii) Frequency- dependent effects

This difference between the longitudin al and transverse cases

in the quasi-stat ic limit results from the fact that in the

transverse case the fields are changing in magnitude and consequent ly

the order parameter also changes. Naively one would expect that

the change in the order parameter might be some kind of relaxation

process, and that with increasing frequency the order parameter

would become less able to follow the changes in magnitude of the

field. At high frequencie s the order parameter would not change at

all and the surface impedance would be the same in a given static

field whatever was the direction of the oscillating field. Thus

SZT/SZL' the ratio between the field dependence s in the transverse

and longitudin al cases would change smoothly fr9m 3 in the quasi-stat ic

limit to 1 at high frequencie s.

(viii) Previous experiment al results

The experiment ally observed results on the field dependence

.. · · of the su~face reactance in supercondu cting

tin are not at all in accordance with the simple picture derived

in the last section. Observatio ns have been made at various

frequencie s, 9.4 Gc/s by Pippa.rd (1950), 3 Gc/s by Richards (1962),

1 Gc/s by Spiewak (1959 ) and 2 Mc/s by Sharvin and Gantmakher (1961).

Sharvin and Gantmakher observed values of tz in fair agreement with

the G-L theory, and also a value of SzT/8z1 consistent with the

quasi-static value of 3. The observations at higher frequencies

have been summarized by Pippard (1961), essentially as follows:

(i) In the longitudinal case &x is normally positive at all

temperatures, as in the quasi-static limit.

(ii) In the transverse case 6x is positive for low temperatures

(say below 0,7 T ) •

c
Close to T
c
Sx is ne gative at the lower frequen-

cies and becomes positive as the frequency is increased.

It follows that at moderately high frequencies and close to T ,

c
~T/&x may actually be negative, showing that our picture of the
1
last section is considerably oversimplified. It is clear that a

significant change in the behaviour of &cT occurs between the

frequencies 2 Mc/sand 1 Gc/s, and even for frequencies as low

(compared, say, to the g ap frequency) as 1 Gc/s the quasi-static

approximation has broken down comp letely.

(ix) The present experiment

The object of the present e x periment, suggested by Professor

Pippard, was to start to fill in the previously mentioned gap in

frequency, a factor of 500, between 2 Mc/sand 1 Gc/s. The frequency

chosen was 170 Mc/s, and measurements were made of the field

dependence of the surface react~nce of superconducting tin over a

wide range of temp eratures in both the transverse and the longitudinal

configurations.
11* 11

Chapter 2. The experiment

(i) The resonance method

The essentials of the method employed to measure changes of

surface impedance in the present experiment were suggested by

Professor Pippard and are described in a paper by him (Pippard 1947).

The basis of the method is to make a specimen of the substance under

investigation in such a way that it will resonate electromagnetic-

ally at the frequency of interest (typically the specimen might

be in the form of a resonant transmission line). Changes in the

surface impedance of the specimen affect the boundary conditions

satisfied by the electromagnetic field at its surface and hence its

resonant frequency. More precisely, an increase in the surface

reactance (assumed inductive) decreases the resonant frequency

and an increase in the surface resist a nce decreases the Q of the

resonance. Assuming that there are no other imp ortant resonances

nearby and that the fractional frequency c hange is small, the

change in frequency can be taken as proportional to the change in

surface re a ctance.

In order to make the sensitivity of the method as great as

possible the specimen must be in the form of a wire. In fact the

fractional change in frequency is then of the order of the change in

penetration depth divided b y the diameter of the wire, so the

diameter of the wire must be as s mall as possible. Its length is

g overned by the fact that the lowest frequency resonance of a

conductor has a free space wavelength of the order of twice its

 12

greatest linear dimension.

(ii) The transmissio n method for determinin g resonant frequency

The resonant frequency was determined by the tnan"Bni:i.::B.si.:on method,

the details of which can be seen from the photo g raph of fig. 2.

The specimen is held inside a metal can into which two transmissio n

lines go (in the photograph the can is removed to show details of

the interior). Th e outer conductors of the transmissio n lines are

connected to each other directly via the top of the can, a nd the

inner conductors are joined by a loop of wire inside the can.

An oscillator is connected to one transmissio n l ine and a matched

detector to the other, the matching being adjusted so that in the

absence of the specimen all the power is transmitted to the detector

and none reflected back to the oscillator . When the s p ecimen is in

position and the oscillator is near the resonant frequenc~ the

specimen is set into oscillation , and some power is absorbed and

some reflected, so that the power transmitted to the detector drops.

An .attenuator is placed between the oscillator and the specimen

to prevent the oscillator being pulled by the reflected wave.

In the ideal case a s :rmmetrical resonance curve of power transmitted

a gainst frequency is obtained, but if the matching is not perfect,

Fi g . 2 an unsymmetri cal resonance curve ~s obtained in general. Changes in

the surface reactance will to a good a p proximatio n merely displace

the whole resonance curve in frequency, if as in the present instance

the frequency is so low that most of the losses are due to causes
 13

other than the specimen. The eIIects of s mall changes in Q can

be partly eliminated by the observational technique described later

(section vi).

(iii) The apparatus 1: the cry o stat

The details of the cryostat desi g n were worked out in conjunction

with Professor Pippa rd and Mr. F. Sadler, and the construction

was carried out by 1Ir. W. Undrill of the Mond Laboratory workshop.

The essenti~l parts are the two transmission lines for carrying the

rf signal, the specimen holder, and a screening can to prevent loss

Tv-ov9 l'l for
holdi119 c.ooli ng\ of pb wer from the specimen by radiation and to allow the specimen
W'ctte.v- clvr-i ng -------
so\cleriti 1 to be at a different pressure fro m the helium bath. In addition
opera.t ion
a tube was provided to allow specimens to be inserted from the top

of the cryostat, which also allowed the separation between the

specimen and the coupling loop to be varied during an experiment.

The design of the lower part of the cryostat can be seen from the

photograph (fig 2), and the diagram of fig. 3. The coaxial lines

each consisted of a 3mm. O.D. copper.- nickel tube outer conductor

and an inner conductor of 26 S. W.G. Eureka wire. The wire was

supported in the tube by some of the polythene insulator obtained

Fig. 3
from a commercial coaxial cable. The polythene was cut up into

short lengths to avoid diffj_cul ties due to contraction and was

threaded on to the wire before being inserted into the tube. The

polythene was prevented from falling out at the bottom by a piece

of polystyrene, which also served to support the inner conductor.

The two lines were made separately and the inner wires soldered
 I

together afterward s to make the coupling loop. At the top of

each transmis sion line was a socket which would accept a standard

G.E.C. plug; this was a design of Mr. Sadler's using a tapered PTFE

washer as a vacuum seal. The tube for insertin g the specimen was

an 18 mm. O.D. copper-n ickel tube. The specimen was held on the

end of a long 5 mm. tube which at the top end passed through an

0 -ring seal and at the bottom end had attached two sets of brass

spring fingers which bore a g ainst the inside of the 18 mm. tube.

At the bottom of the tube holding the specimen was inserted a

short length of copper wire, which could be bent in order to align

the specimen correctly for insertion in the tube (a system invented

by Chambers (unpubli shed)). The screenin g can was made from 30 mm.

O.D. copper-n ickel tube electrop lated on the inside with copper to

reduce resistive losses. The coaxial lines and the specimen

insertion tube were soldered into a brass top plate, which rested

via an 0-ring seal on top of an assembly wh ich could be connected

to the laborato ry vacuum line, a booster pump, a small rotary

pump, or the helium return line (fig. 1). Provision was made for

pumping the specimen space and for filling the specimen space

with helium from the helium bath. The pressure in the specimen

space could be read with a mercury manomete r and the bath pressure

by either a mercury manomete r, a differen tial oil gau g e or a McLeod

g auge. Specimen temperat ures were obtained fro m the bath pressure ,

an estimated hydrost a tic correctio n of 0.45 mm. of mercury being

added to those obtained above the l ambda- point.

 15* 15

All soldered joints near the specimen had to b e made of

-- non-sup erconducting solder to avoid interference with the magnetic

field. Hard solder was used for all joints except that holding the
s: ~ - ~
Q) 7) ti d 0 screening can, which it was forseen might have to be removed later.
f +J O
-~ ·; () '-r -{..J
V) ~
QJ
d u ~ For this joint a zinc-cadmium alloy was used as solder. To avoid
Cl)
a... v
V'I , s:
~
IJ
t
Q
.;:,:::; damag e to the coaxial lines while this joint was being made, a
water
trough which could be filled with coolingAwas incorporated i nto the
'\ I'
'1' design and the heating was done electrically.
~
Thermal contact between the specimen and the helium bath was
~ ~
c:v !...
0 QJ
~
~
C) ·-
.<±:
obt a ined by admitting into the specimen volume 1 c m. pressure of
::1
s: a -a.
a,
- ::::. i
helium before cooli ng. The helium bath was contained in a glass

~
+' ,<t
. ./
<
,+J l Dewar surrounded by a liquid nitro g en shield. A temperature of
o. 8 ° K could be reached using the booster pump.
,1,
'/'

(iv) The apparatus 2: external cirauitry.

~
Q.I I. ~
-
~
f:i
a,
-,.J
0
~
~
s:. Q)
s: QI
-d Q.I-+,)
~!-.
>- -e
Q) The experimental arran g ement in its final form is shown in the

4.
s:
:s
..___
--·-d

v
\.
r
O
~
.µ s;.
a> 0
OJ~~~
>d
()I O
Lu
:> I
x.
~
\.)
Q)
block diag ram, fig.

will be descri bed •

In this section the arran g ement used initially

VI :c u lL ~
0
Much of the equi pment wa s that used previously for penetration

... , dept h measurements by McLean. We refer to McLean (1960) for fuller

details not given here. The rf signal was g ener a ted by a triode
I C1'l -1-Jcl) oscillator constnucted by McLean using a CV 273 triode with coaxial
QJ
,....., ~ ~
..µ
..£l
-0.... \. QI \,.
::> s: ,
\.. s:
i::J c.) r i.. (,, en lines shorted by movable spring fingers as the frequency determining
-d CL ~
:::>
~ d ~ d
d :;) ::> QI
> V"I v ~
V'\
z resonant elements. Coarse control of the frequency could b e obtained

by movi n g the s p rin g fingers, while fine control could b e obtained

 16

by inserting metal or silica rods into the resonant line. The

output was taken from one of the lines by variable capacitative

coupli ng. The filament voltag e was obtained from an accumulator

and the HT supply from a stabilized power pack. Two-stage RC and LC

filters of standard design in a screening can were used to prevent

trouble due to power getting back along the supply lines. By

careful setting of all the controls (grid voltage, feedback, length

of coaxial lines) it was possible to get the oscillator to oscillate

stably at t h e required frequency, with a drift which settled down

2
eventually to a value of the order o f 5 cycles/sec •

The fre q uency of the output was measured by a Beckmann/Berkeley

pulse counter model 7370 together with a model 7572 hetrodyne

frequency converter, which to g ether displayed the output in digital

form. Both models were controlled by the same thermostatically

controlled crystal oscillator with a stability of± 3x10- 7 per week.

The oscillator was connected to the input line of the cryostat

by attenuating cable, which was also used between the oscillator and

the frequency meter, t h e object t in both c a ses being to prevent

reflected signals.

The output from the cryostat was taken to an adjustable matching

I

device and then to a rectifier consisting of a point-contact diode

I

inserted in the inner conductor of the transmission line. The

I
output from the rectifier was then taken, in the preliminary

experiments, to a galvanometer.

A pair of Helmho l tz coils was mounted on a Dexion framework

 17

to provide horizontal (transverse) magnetic fields and a solenoid

was used to provide vertical (longitudinal) fields. The field

distributions were measured and found to vary by less than 5%

over the specimen region. Each magnet had a resistance of the

order of an ohra . The magnets were powered by a 25 volt battery

supply and the current was controlled by series rheostats. In

the preliminary experiments, the current was measured directly by

a sub-standard ammeter. The earth's field was not conpensated.

(v) Specimen preparation

As mentione d in sedtion (i), the specimen must be in the form

of a thin wire about half a wavelength long, or in the present case

about 90 cm. The only suitable method for making and supporting

wires of the necessary thinness is to cast them into fused silica

capillary tubes and to leave them in the tubes during the experiment.

Silica is suitable for this purpose because of its low electromag-

netic losses and b ecause tin con tracts more than silica on cooling

and does not stick to it, so that specimens prepared in this way

have surface properties similar to those of unsupported specimens, as

shown by Pippard (1950).

The first problem was to make capillaries of sufficient length.

Capillaries are normally made by heating ordinary tubing in a flame

while holding b oth ends till the viscosity of the quartz becomes

very low, re moving it from the flame and pulling sharply. The

length of capillary · required in the present case prevented this

method on account of two factors, the length of the exp erimenter's

arms and the leng th to which the tube could b e pulled before it

got too cool to be stretched any further. The following met h od of

making capillaries was accordingly devised. The top of the tube

was held vertically in a clamp at a hei gh t of about six feet, and

a weight of about i kg. was clamped to the bottom of the tube,

leavin g a gap of a few centimetres in betwe en. To heat up the

necessary volume of silica an oxy-acetylene torch fitted with a

jet of size no~ 10 or larger was used. This was directed at the

region between the c.lamps while the weight at the bottom was held

in the free hand. When the silica was almost molten the flame was

removed and the weight immediately released and allowed to fall

into a bucket of sand, drawing out the tu b e into a capillary in the

process. Some practice was neaessary to get suitable results; if

an insufficiently hot flame was used the weight would not fall far

enoug h, and if an insufficiently great region was heated up the

resulting capillary would have an e x tremely fine bore, though

there seemed to be no danger of the tube closing up comp letely.

By this method capillaries of length about 130 cm. and bore of the

order of 0.1 mm. could be made.

The long capillaries then had to be bent to a suitable compact

shape which would g o into the cryostat. An important factor in

determioing the shape was the necessity of being able to se parate

out the longitudinal and transverse field cases. Since the currents

flow along the axis of the wire the rf field is transverse to the

wire, so that the longitudinal and transverse field cases defined

 in chap ter 1 actually c orresp ond to steady fields longitudinal and

tran sverse to the wire in this case. Professor Pippard drew

attention to a paper by Tapp (1932) describing a method for makin g

helices fro m silica fibres. With a helical wire of s mall pitch

an g le the more interesting transverse field case can b e realized

approximately when the steady field is p arallel to the axis of the

helix , while with the field at ri g ht ang les to the axis the field

will be transve r se in some places, lon gitudinal in others and in

between the field will h a ve comp onents in both directions.

Atte mpts were made to apply the Tapp method to the production

of helical capillaries. Essenti &lly the princip le was to win d the

capillaries on to a grap hite rod with a s mall flame p laying on the

point where the cap illary first made contact with the rod, in order

to bend the capillary. A minature oxy-coal g as burner wa s made for

this purpose. The rod was caused to move along its a x is as it

rotated by attaching it to a threaded rod rotati ng in a fi x ed mount.

The capillary was g uided in its approach to t h e rod by two slots

and tensioned by friction o :r g ravity. Th e main difficulty with

this method was that the capi llary tended to bend in polyg ons rather

than s mooth curves, and to become very fra gile at the corners of the

p olyg ons. The ma x imum number of turns obt ained by this met h od

before the capillary stretc h ed and snapped or met some other disaster

was about ten, and,. was g eneral l y less. This was less than half of

t h e leng th required, and various mo difications tried met with no

g reater success.
 20

In view of the diffic ulties inv6 lved in this method it was

decide d to leave the helica l geomet ry for the time being and try

a differ ent one. This was essent ial ly a zi g-zag shape bent into
a cylind er, and one such specim en can be seen in fig. 5. The
produc tion of such a shape involv ed only making a number of 180°

bends in the capill ary. One advant a ge of this shape is that apart
fro m the bends all the wire is in one direct ion only, so that the

transv erse and lon gitudi nal field cases can be invest igated separa
tely.
Howeve r, there is the disadv antage that the length of the zig-za g

needs to be consid erably greate r than h a lf the free space wavele ngth

at the resona nt freque ncy. The re a son for this is that adjace nt
sectio ns of the zi g-zag carry curren ts in opposi te direct ions, which

tends to decrea se the ma gnetic energy for given curren ts and hence

the self-in ductan ce, while they carry charge s of the same sign, which

increa ses the electr ostati c energy for given char ge distrib ution

and hence decrea ses the self-ca pacita nce. The net effect is that the
resona nt freque ncy for a specim en of given leng th is increa sed.

Estima tes were made of the magnitu de of this effect by using the

variat ional princi ple to determ ine the freque ncy of resona nce.

If we assume the curren t distrib ution along the wire in a normal

mode to be j(x) thehth e charge distrib ution (out of ph a se by n/2)

Fig . 5 Specimens b efore and after being )fille d with tin \ is given by the contin uity equati on to be w-1 ~,
d"
w being the freque ncy
of the mode. At one point in the cycle the energy is entire ly
magne tic, the curren t being at its peak, and a quarte r of a cycle

later the energy is entire ly electr ostati c, the charge densit y being
 21

being at its maximum. Equating the two energies gives an expression

2
for 00 in terms of j(x), and the variational principle tells us

that the frequency and current distribution of the normal mode of

lowest frequency can be found by choosing j(x) so as to minimize 2

Ol •

In the present case the energies were calculated by treating the

sections of the zi g -zag as infinitely long, each one carrying a

uniform charge density or current. For j(x) was assumed a Fourier

series containing terms up to the f o ur t h order, all except two

vanishing by symmetry. The effect of the screening can was taken

into account by the method of images.

The calculation showed that increases in: length of the order of

50% would be required to make a specimen of the same resonant

frequency as one which was half a free space wavelength long. The

length would of course depend to a small extent on the diameter of

the specimen and the number of sections of the zig-zag. To check

t his prediction a test specimen (fig. 6) was made up from some

o.005r1 niobium wire kindly provided by Dr. B. Abrahams. The wire
was threaded throug h holes in two PTFE discs on a silica rod and

held at the ends with araldite. The resonant frequency, measured

Fig. 6 Test spec i men made from Nb wire
in a liquid helium run, was about 10% less than the predicted value.
In decidin g the length of capillary to be used for the real

s p ecimen to give the required resonant frequency, allowance was made

for the dielectric nature of the silica, and on the strength of

the result obtained with t h e test specimen a length of about 105 cm,
was decided on. It was also decided to make the speiimen with 12
 22

sections of zig-zag, this being a compromise between the necessity

of keeping the specimen short so that it should be in a f a irly

h omogeneous field, and the difficulty of making a specimen with

a larger number of sections.

The silica capillary was supported in trian g ular slots on the

ed ge of three mica discs of diameter 15 mm, spaced 3 cm. apart on

a 2 mm. silica rod, allowing just over 1 mm. clearance in the

specimen insertion tube. Before the disc s were attached to the

rod one end of the rod was enlarged by heating in a flame and the

other end in serted into a brass sleeve just larg e enough to take

t he main part of the rod. A 6 BA threaded rod was then screwed

into the sleeve and held by a locknut. When the specimen had been

made the silica rod was held firmly in the sleeve with black wax

and the threaded rod used to attach the specimen to the s p ecimen

holder.

The adhesive used for attaching the mica discs to the central

rod and the capillary to the mica discs was araldite I. This had

two advantag es compared to the usual two-tube form. In the first

place it is better able to withstand the temperatures involved in

filling the capillaries with molten tin. Secondly, as will be seen,

its method of application made it possible to cure all the 36 joints

together instead of separately a s would have been necessary using

two-tube araldite.

In the first attempts to make zig-zag specimens flat zig-zags

were first made, using the minature flame to make the bends. This

-
 23

operation proved to be fairly stra i ghtforward, but difficulty

was experienced making the zig-zag lie on the surface of a cylinder

w:iile sticking it into the slots on the mica discs. It was then

realized that it would be much simpler to do the bending and

sticking operations in a single operation. To facilitate the

operation a jig was constructed (fig. 7) which allowed the specimen

to be rotated about its axis and about a line at right angles to the

axis. It should be explained at this point that araldite I is

a wax-like solid with a low melting point, which is cured by

heating. Since the adhesive is solid at room temperature the joints

have a certain amount of strength even before the adhesive is

cured, although of course the adhesive becomes liquid again when

it is heated for the curing process.

The specimens were made in the following way. First the mica

discs were stuck to the silica rod and the joints cured. The

specimen was then placed in the jig and the capillary rested in

the slots in the mica discs, and fixed temporarily with araldite I

as described above, using a minature soldering iron run at

:t reduced voltage to apply the adhesive. The capillary was then bent
Fig. 7 Ji g f o r making s p ecimens ;
in the minature flame, the specimen being rotated so that gravity

would cause the capillary to fall into the next set of slots. The

capillary wa s then stuck to the next set of slots and the process

continued a s b efore.

When this process wa s finished the specimen wa s p lac e d in a n

oven to cure the joints. The capillaries were held in the slots
 24

the wires
during this proce ss by wrapp ing fine wires round them,

being kept in tensio n by attach ing weigh ts to their ends.

the
The final step in prepa ring the specim ens was to fill

capil laries with tin. In princ iple, the metho d was the follow in g .
was evacu ated
One end of the capil lary was sealed up. The capil lary

and the open end immers ed in molte n tin. Nitrog en at atmos pheric

the capil lary.

press ure was then used to force the molte n tin into

The exper iment al detai ls were as follow s. The tin used (Johns on

Matth ey spec. pure) was heat~ d to red heat in a rotary pump

break s due
vacuum to out g as it and reduce the dan ger of gettin g

to bubbl es in the colum n of tin in the capil lary. The tin was bent

pyrex tube
into a ring and placed in the bottom of a metre long
meltin g.
which had a small pip on the end to conta in the tin after
a rubbe r bung
The specim en was attach ed to a long rod going throug h

at the top of the tube. The b ung was also fitted with a tube for

evacu ating the s~stem . A few inche s of the open end of the
as sugge sted
capil lary were left protru ding beyon d the specim en and,
as would
by Profe ssor Pippa rd, the end was bent into such a shape
tube by
preve nt it being pulled again st the side of the pyr ex

the f orce due to the incom i ng tin.

of the
The specim en having been positi oned with the open end
d off. It
capil lary in the pip, the tube was evacu ated and clippe
ce, which
was then placed vertic ally with its lower end in a furna
0
the
had been pr~vi ously se t to a tempe rature about 20 C above

meltin g point of tin. When the tin had melte d and gone into. the

pip nitrog en was admit ted into the tube. After allow ing a few
 25

minutes for the capillary to fill, the specimen was withdrawn

sli g htly so that its end was clear of the tin, and the tube was

removed from the furnace. At the first attempt a good specimen

was obtained, all but about 1 cm. of the capillary being filled

with a continuous length of tin. The protruding end of the

specimen was broken off before the specimen was mounted.

(vi) The preliminary experiment s and the consequent modificatio ns

The first task was to find the resonant frequency of the

specimen. This was originally done by using the fi n e tuning control

of the oscillator and systematic ally covering the whole frequenc y

ran g e by setting the coarse controls suitably. The resonance

showed itself as a sharp drop in output on the galvanomet er.

This process took a few hours, but was later s p eeded up by using

a Marconi signal generator no. TF 801D/1 lent by Dr. Dobbs as the

oscillator . This could be tuned through the frequenc y range in a

few minutes without the output power fluctuating violently as

would have happened using the coarse control on the McLean

oscillator . To make the det ~c tor fast enough to respond to the

sudden chang e in output on g oing through the resonance, the

attenuating cable was dispensed with, thus making the output large

enough to be deteoted on an oscilloscop e.

Trouble was exp erienced at the be ginning due to short circuits

in the transmissio n lines down to the cryostat. This was traced

to damage caused to the polythene and polystyren e during the

operation of soldering on t h e screen ing can. This difficulty was I

overcome by replacing the polythene in the relevant parts by pyrex

'
 26

and the polystyrene by araldite.

The method of measuring changes in resonant frequency of the

sample was the following. The changes of resonant frequency on

applying a ma gnetic field were so small (usually less than 1 kc/s

or about 1% of the width of the resonance) as to make it unsa tisfactory

to proceed by plotting out complete resonance curves. Instead, the

oscillator was set to a frequency on one side of the resonance,

and the magnetic field switched on and off several times while

noting down the galvanometer deflection, if necessary adjusting

the oscillator tuning c ontrol to compensate for drift. These

amplitude changes were then rela ted to changes in resonant frequency

by plotting a resonance curve at fixed magnetic field. To allow

for possible changes in surface resistance the measurements were

repeated on the other side of the resonance curve. Ch anges in

surface resistance produce contributions of opposite sign in the

two cases, and so averaging will lar g ely eliminate the effects,

in any case small, of surface resistance chang es.

One disadvantag e of the cryostat desi g n manifested itself

when the resonance had been found; the galvanometer spot s h o wed

violent jerky, irregula r movements a b out the scale, making it

difficult to estimate what to take as its mean p osition. This

effect only occurred on resonance, indicating that it was due to

fluctuations in resonant frequency, presuma bly due to motion of

t h e s p ecimen relative to the screening can. This effect was

enhanced b y the eccentric position of the s p ecimen in the screening

27*
27
I,
can and by the lack of symm etry of the speci men
about its axis
(sinc e the ends of the capil lary defin e a prefe
rre d direc tion) .
These assym metri es allow chang es of frequ ency
which are first order
in the displ acem ent. The lar g est burst s of activ ity, corre spond ing

II
C'>
""' to frequ ency chang es of some kiloc ycles per secon
d, could be
clear ly iden tified with event s such as slamm ing
doors in the
Q)
. 0
,.....
.r;- vicin ity, peopl e walki ng in the corri dor or noisy
mach ines being
t::l rn
0 t: used in the cons truct ion of the new prac tical
class es nearb y. The
Q) rd
0 (]) Q)
s:: ...c: +>0 other troub lesom e distu rbanc es were proba bly due
ell+> . to simil ar cause s
§ •

"' -
§Ol ~O ~ of inten sity too s mall to be notic ed by other
Q) 0 (]) mean s, rathe r than
H ...C: o
0
0
cd
t- Z'
Cll qQ ~ ti:, the boili ng of the nitro gen or the vibra tions
Q) H of the pump, which
-~
~ 0
(]) -~
Q)
+>
a CQ
0

would have cause d effec ts simil ar to rando m noise

+> .0 0
CU .
Q)
+> s:: ~
H
,.
:i.

-
Ol
Q) ~ ·rl ::, The exact way in which vibra tions affec ted the
rd >< rd outpu t power
+> cd (]) ..:J. u
0
+> -~ ><; -~
cd
I""
~
did not becom e clear until the outpu t was plott
ed on a pen recor der
rrj Q) +>
(]) +> ...c:
~
.0
Ol Q) +> 0
::l a:) (figs . 8,9). It will be seen from these t~ace s that the distu
rbanc es
Q) Ol rrj"
Q)

t 6
0 cd
cd
S... f;:
+>
Q)

1! . ::l
0
'°t'- GIs:~
tend to be in one direc tion only, and that their
magn itude and sign

+>
S...
8~ (])
!1l
cd 0
Q)
vary in a simil ar mann er to the secon d deriv ative
Qlcd!ll :, of the ampl itude
~
.O
rrjrl Q) r:,A
H rl ·r! a> s:: OJ with respe dt to frequ ency. This indic ates that the meas uring
0 ·r! 0 ~ 0 .\..
0 0 S:: • Ill
Ls...
Q) Ol Q) +> (1)
p::: 0 ::l p:::
o< cd devic e is not respo nding to the fast vibra tions
(])
• ...C: H
Q)
o . Oo of the s p ecime n,
~ 8 ~ +> P'.:l
" but is actin g as a squar e law detec tor, owing
to the non- linea rity
of the ampl itude -freq uency curve . There fore the noise can be

~
0
Oo reduc ed by worki ng near the point of infle xion
r:-- of the re,son ance
m
t-- curve , and by broad ening the reson ance curve by
coup ling the
speci men more stron gly.

~
~ - - - - - - - - - - ~~ ()~+''G
~.; 7'
0 .!..
"' .
 28

The effect s of the extern al vibrat ions could probab ly have

been made ne g ligibl e if the specim en could have been fixed

rigidl y
agains t the screen ing can. Howev er, this was not very feasib le

with the presen t design , adopte d to avoid the necess ity of

freque nt
solder ing and unsold ering of the sc~een ing can, in which the
specim en
was insert ed from the top of the cryost at. Failin g this, variou s
other device s were tried to reduce the . vib~at ion of the specim
en.
The screen ing can was suppor ted in the helium dewar by spring

finger s. The copper wire link to the s p ecimen mentio ned in sectio n

(iii) was stiffen ed with araldi te once it had been bent into
the
cor. rect positi on, and the effect of replac ing the spring finger
s
suppor ting the specim en in the specim en insert ion tube by a

kinem atic slide was noted but found to be ne g ligibl e. A device

invent ed by Preston -Thom as (1949) to preven t the nitrog en bubbli ng
was made, but this wa s not very succes sful since it tended
to
block up when the nitrog en evapor ated after a run. Finall y, the
only measur e which had any signif icant effect was to work at
niwt.
Apart from the period when the porter made his nightl y tour
of the
labora tory, there were few larg e disturb ances then. I I

Improv ements were also made to the detect ion side of the

experi ment. A twin-c hannel record ing milli _a mmeter was used to

record the inform ation obtain ed, and allow a less subjec tive

separa tion of the signal from the noise. Into one channe l was fed
the output si g nal amplif ied by a galvan ometer amplif ier. A filter
was p laced betwee n the cry stal detect or and the a mplifi er,
just
as was used in the power supply to the oscillator, to prevent

variable amounts of rf power being reflected back from the amplifier

to the detector. To the other channel was connected a si g nal

derived from the magnet curr ent. The recorder was oil damped with

a time constant of a few seconds. Readin g s were taken as before

by switching the magnet on and of f several times. Some trouble

was had near the transition tempenature owing to drifts in temperature

of the bath, so an Adkins temperature stab ilizer (Adkins 1961) was

fitted. This type of stabilizer detects chang es in the bath

tempera ture by means of a differental oil manometer, and controls

the tempeeature by altering the current throug h a heater at the

bottom of the cryostat.

A few observation s were made by this method, but it was soon

decide d to replace t h e recordin g milliammeter by a Moseley servo-

controlled X-Y recorder. One advantage of doing this was that the

whale behaviour of penetrati o n depth as a function of field could

be displayed by varying the field wi th the rheostats while plotting

output powe r directly a gainst ma gnet current. The si g nal corres-

ponding to the ma g net current was in this case o btained from the

voltage develo r ed across a shunt i n series with the magnet. Further,

the X-Y recorder had such a f a st response time that it was p o ssi b le

to dr aw comp lete plots in between ma j or disturbances affecting the

specimen. It was also p ossi b le to comp e n sate for t h e component of

the earth's field in the direction of the magnet's field by plotting

the behaviour with both directions of field and making use of the

symmetr y of the curve to fin d th e p oi nt corresponding to zero total

 +>
.
e.-.
(I) 30
30* r-1
(I)
..c:
+> field. A typical X-Y reconder p lot, including traces giving the
g
0)
relation between output and oscillator frequency, is shown in fig. 9.
(I)
0
c1l ! I I
~
+> (vii) Miscellaneous observations and calibrations
(I)
..c: Critical field curve
+>
~
0
e.-. The critical field of the specimen was measured at various te mp er-
11)
(I)
l:lO atures by setting the oscillator so that the specimen was in the
s:1
c1l
~ ~ ..c:0 centre of the resonance, and measuring output as a function of
IJ v
~ ~ ...Y. ..:.!
:>,
I,) tJ I.I'\
~ ..:..! V\ 0
6 0 s:1 applied longitudinal field, the coupling to the loop being made
(I)

& fairly stron g .

(I)
~
Wh en the specimen became normal the Q of the
e.-.
+> resonance dropped, c a using the output to increase. There was a
§
§ tendency for supercooling to occur, and for the transition to
0)
(I)
~
occur in a series of step s, but there was a fairly well defined
0
+> field
.p maximum transitionAin incre a sing fields, and this was taken as
s:1
(I)
s(I) defining the critical field~
0
c1l
r-1
P..
0)
Critical temperature and transition width
~ ·
The critical temperature of the specimen was obtained from the
(I)
+> critical field curve as the temperature where the critical field
c1l
r-1
(I)
~ extrapolated to zero. To find the wid th of the transition, the

range of fields over which the transition took place at constant

temp erature was esti~ated from the curves used to find the critical
11)
(I)
fields close to the critical tempera ture, and this range was then
t::s
0
converted to a walue of the broadening of the critical temper a ture

in zero field, using the slope of the critical field curve near the

transition point.
 31
31 *
Constant s relatin g ma gnetic fields and currents through the ma g nets

The fields corresp onding to a given current in the ma g nets were

found by compariso n with a standard solenoid made by Dr. Faber. A

search coil connected to a ballistic g alva nometer was used to

173 ·B00 compare the magnetic fields.

Constant relating resonan t frequency c h an g es to chan g es

in penetrat ion depth

The resonant frequency chang es observed in the e xp eriment are

Fig . 10 proportio nal to the chang es in penetrat ion dep t h, the constant of

pro p ortional ity dependin g on t h e detailed g eometry of the system.

ao
In the past, methods of varying de g rees of sop histicati on have b een

used to d et ermine t his constant , b ut here we h ave adop ted a fairly

Resona.nt
freq,uency simple approac h, as follo ws. Resonanc e curves were plotted in
(Mds) zero fiel d at differen t te mperatur es ( a t y pical p lot is shown in

&O fi g . 8 ). At such low frequenc ies as those used in t h e present

e x periment the penetra ti on depth in zero field is p rop ortional to

4 _..1.
This
z :::: (1-(T/T) ) 2 over a fairly wide ran g e of temper a tures.
- c
is confirme d by the e x periment al p i ot of r eso n ant fnequ e ncy a gainst

z (fig 10). By extrapol atin g t h e linear p ortion of the curve to

j '
40 z=O the frequency corres p ondin g to zero penetrat i on de p th was found.
11
The e x periment al curve was t h en used to find the frequency c h an g e

corres p ondi n g to the penetrat ion de p th a t any temp era ture. The

met h od described was used to pre s ent all t h e e xperimen tal r e sul ts

as fract i onal chang es in penetrat ion de p th a t the temp erature co n cerned.

173-7100 3
Residua l resi s tivity ratio

The re s idua l resi s tivit y ratio o f t h e sp e c i men wa s me a sur ed by

 32
break ing the c a pilla ry at the ends of the s p ecime
n and solde ring
leads to it. The specimen was conne cted in serie s with a stand
ard
resis tor and t h e volta ges appea ring a c ross each
were compa red at
room tempe ra ture and liqui d helium tempe ratur e
by means of a
p o tenti omete r, the curre nts being rever sed to
allow for t h ermal e.m. f's.
 33
33*
Ch apter 3. Results and discussion

(i) Specimen charact e ristics

The characteristic s of the specimen are giv e n in the fo l lowing table.

TABLE I. Specimen characteristic s

Material Johnson Matthey 'spec. pure' tin

m Total leng th 105 cm.

Room temperature resistance (291°K) 9.10 ohms

Residual resistance (4.2°K) 1.232 milliohms

Residual resistivity ratio 7390

-
~
...__,
0
Mean diameter 0.0144 cm.

1.55 x 10-
10
sec.
Q)
Relaxation time (4.2°K)
~
rl :::> Mean free path (4.2°K) 0.008 4 cm.
..µ
d Transition temperature 3.715°K
(l) l...
I>
s
t)
Q)
Q._
Transition width (field) 2.2 gauss
rd E (temperature) 0.015°K =
0.004 T
c
r-l (U
(l)
·rl
l°H f- Critical field curve see fig. 11
r-l
c1l
t) Resonant frequency 174 Mc/s
·rl
+>
·rl
~
0 The determination of most of the quantities in table I has been

~
described alre a dy . The mean diameter was determined from the ro om

t~ mperature resistance, assumin g a resistivity of 1.4 x 10-5 ohm cm.

(from the Chemical Rubber Publishing Company Handbook of Chemistry

and Ph y sics 1 962-3). The relaxation time and mean free p ath at

helium temperature were obtained from the observed r e sistivity and

data for tin give n by Fawcett (1 96 0).

 ,........_
() rd
~ ~
::i:: ·r-1
~
.µ
(/) (I)
~ tQ
·r-1 H
(\l. (I)
0 Q.Q p
ell tQ

r< §
> ..._,,,
<I
.µ
H

N
rl
.
bi)
·r-1
Iii

--+ 0 LI\
...,.0
0
,
0 .
0
I
~ ~I
+
 r-1

~
o.l
.
..d-
0
, ::i
~
•M v
llO
~
0
:I:
r-1 ..........
.._,, :I:
()
~
·~
'C)
::r::
~
!I.I

-~rotu)
ro
t<
> <]

~
·M
~
 0( 'Y' ~
O C) 0
II v U II
:I' v v
. . ..............:r r
..................
:r: :r I

0 <l CJ

U\
~

~ ()~
,...;
~
E-i
Q)
·r-1
(j...j
+>
a) Q)

r:- -~cu
a)
r-t
:c ~
4)
>
EJ cu a)
~

~
,<
"!> . ............
,< +>
<J ~

El \"

~
..:I-

'!"

0
'i"
 34

The crystal orientation of the specimen was not me~sured, but

owing to its shape it is unlikely that solidification would have

proceeded gradually from one end to the other, and so the specimen

.
l1)
C)
..j,- rt'\
~
probably consisted of a large number of differently oriented
II
... II
::c
... crystals of length of the order of 1 cm.
:::c
--... -......
:r
-.
0
.:c
<:)
(ii) The field dependence of the penetration depth
El
The experimental results are tabulated in tables II and III (next

page). The magnetic fields are given in gauss and the frequency

changes (ly in kc/s. The results are disnlayed graphically in

figs. 12-15. In figs 12 and 13 the fractional penetration depth

C.)

~ change A)./). is plotted against the reduced magnetic field h = H/H

8
reduced c
+:>
rll
s:: for variousAtemperatures in the transverse and longitudinal cases,
•rl
C1l
Qj)
C1l
and in figs. 14 and 15 ~)./)... is plotted as a function of reduced
..<
>
<]
temperature for various values of h, using interpolated values of

AX/A obtained from the curves of figs. 12 and 13. For clarity

some curves have been omitted from fig. 12, but fig 14 contains the

points derived fr om the omitted curves as well. In fig. 14 temp-

r:I
0
erature has been plotted on a lo garithmic scale to expand the region

near T.
c

0 (iii) Errors
'!" -'!- ~
0 0 '+>' 0
Almost certainly the main sour~e of error in the experiment
~£
r< \.)
I,,. was that caused by the irregular mechanical disturbances to the
<3
-
a,
a...
specimen mentioned in chapter 2, section vi. The pen recorder trace

reproduced in fi g . 9 was taken at a te mperature where the chang e in

(continued on page 37)
 35

EXPERIMENTAL DATA ON THE FIELD DEPENDENCE OF THE PENETRATION DEPTH

Note In tables II and III the ieft hand column is the temperature

in degrees Kelvin. The pairs of numbers to the ri gh t are the

experimental observations, the upper member of the pair being the

field in g auss and the lower member the observed resonant frequency

chang e in kc/s.

Table II Transverse field

3.686 0.14 0.28 o.43 0.57 0.71 0.85

0.18 o.44 o.86 1.27 1.70 2.00
0.14 0.28 0.43 0.57 0.71 0.85 0.99 1o14
0.06 0.26 0.70 1.02 1.32 1.64 1.88 2.06

3.664 0.28 0.57 0.85 1.14 1.42 1. 70 1. 99

0.09 0.34 0.74 1.15 1.63 1.97 2.28

3.647 0.57 1.14 1.70 2.27 2.84

0.24 0.65 1.07 1.48 1.79
0.57 1.14 1.70 2.27 2.84 3.41
0.12 o.46 o.88 1.15 1.50 1.72

3.610 0.57 1.14 1.70 2.27 2.84 3.41 3.98 4.54 5.11 5.68
0.04 0.14 0.27 0.42 0.56 0.71 o.86 0.98 1.08 1.14
0.71 1.42 2.13 2.84 3.55 4.26 4.97 5.68
0.04 0.10 0.18 0.27 0.35 0.42 0.50 0.57
1.4 2.8 4.3 5.7 7.1
0.07 0.18 0.30 0.41 0.48

3.520 1.4 2.8 4.3 5.7 7.1 8.5 9.9 11.4

0.03 0.09 0.13 0.15 0.18 0.21 0.22 0.21

3.479 2.8 5.7 8.5 11.4 14.2 17.0

0.03 0.08 0.08 0.03 -0.07 -0.23
2.8 5.7 8.5 11.4 14.2 17.0 19.9
o.oo o.oo o.oo -0.01 -0.05 -0.11 -0.20
5.7 11.4 17.0 22.7 28. L,. 34.1 39.8 45.4
o.oo -0.01 -0.05 -0.10 -0.18 -0.27 -0.38 -0.53
 Table II (continued)

2.47 8 .5 17.0 25.6 34.1 42.6 51.1 59.6 68.2 76.7

o.oo -0.01 -0.04 -0.08 -0.13 -0.20 -0.29 -0.42 -0.64
1,97 17 34 51 68 85 102
o.oo -0.01 -0.03 -0.11 -0.24 -0.51
1.65 17 34 51 68 85 102
o.oo +0.02 +0.03 +0.02 -0.04 -0.21
1.26 17 34 51 68 85 102 119
+0.03 +0.08 +0.16 +0.21 +0.21 +0.15 -0.02

1.19 17 34 51 68 85 102 119

+0.02+0.05 +0.09 +0.12 +0.11 +0.06 -0.08

Table III Langi tudinal field

3.662 0.85 1.69 2.54 3.38 4.23

o.oo -0.02 -0.11 -0.35 -0.79

3.473 4.2 8 .4 12.7 16.9 21.1

-0.01 -0.03 -0.08 -0.16 -0.34

3.041 7.6 15.2 22. 8 30.4 38.0 45.6 53.2

o.oo -0.01 -0.04 -0.06 -0.11 -0.17 -0.27
2.076 15 30 46 61 76 91 106
o.oo -0.01 -0.03 -0.06 -0.08 -0.13 -0.19
1.0 19 38 57 76 95 114 133 152
-0.01 -0.02 -0.04 -0.06 -0.09 -0.12 -0.15 -0.19

I
I I
 37

penettation depth was large, and most traces had considerably

worse signal to noise ratio than fig. 9. The values of A~ given

in Table I were derived not from the recorder traces themselves

but from hand-drawn smooth curves drawn on top of the traces

following the general trend of the traces but with the noise removed.

Traces where the noise was so great that the location of such a

curve was doubtful wer e discarded. The procedure described is

somewhat subjective, but in view of the non-random nature of the

noise probably gave better results than an exact integration

procedure would have done. However, since the points of the curves

of figs. 12 and 13 have themselves been deriveq via hand-drawn

curves, no information can be deduced from the accuracy with which

they lie on the curves drawn. This comment is not applicable to

figs. 14 and 15, since there the points on a gi ven curve correspond

to different temperatures and involve independent calculations

from a set of different recorder traces. Therefore an estimate of

the random errors can be obtained from the scatter of the points

in figs. 14 and 15. This indicates that the absolute error in AA/)..
should be generally less than 0.2% in the transverse case and 0.1%

in the longi tudinal case, the corresponding relative errors being

of the order of 15%.

The errors in temperature determination a b ove 1.5°K, where the

pressure was determined from the mercury manometer, would be important

only near T. Any systematic errers would affect the determination

c
of T in the same way, and h enc e n ot affect T/T , while random errors
c c
 38

are small compar ed with the width of the transi tion in zero field.

Tempe rature determ ination s below 1.5°K using the McLeod gauge are

liable to be too great by an amount of the order of 0.1°K, owing

to the presen ce of a leak which was not notice d at the time in the

connec tion to the McLeod gauge.

The broade ning of the transi tion (l\. T /T

c c
= 0.004) should not
be import ant except for the two highes t temper atures at which obs-

ervatio ns were carrie d out. These observ ations appear only in

fig. 14, and it can be seen that the points for the temper atures

concer ned do indeed deviat e apprec iably from the curve.

The earth' s magnet ic field is approx ima tely a quarte r of that

corres pondin g to the width of the transi tion, and should theref ore

not have had an import ant effect on the result s. The errors in
field determinatio n and the inhomo geneit ies of the field over the

specimen were both s mall compar ed with the errors in the determ ination

of lll.

Variou s other possib le source s of error were invest i gated. A

trace was taken in t h e usual way with t h e specimen off resona nce,

usin g somewh at greate r magnet ic fields and sensit ivitie s than

those used during the penetr ation depth measur ements , and it was

found that the magne tic field had no detect able effedi'.t on the

amplit ude of the output of the oscill ator. It also had no signif -
icant ef f ect on its freque nc y .

Anot her possib le source of error would occur if the magne tic
field could exert a mechan ical force on the superco nductin g specime n

and displac e it relativ e to the screeni ng can, thus alterin g the

resonan t frequen cy. To investi g ate this possib ility, the Helmho ltz

coils were moved from their symmetr ical positio n so as to create a

strong ly inhomog eneous field near the specime n, and the effect on

the recorde r trace noted. There was no appreci able effect apart

from a small change in the scale of the X axis due to the change in

the field at the specime n, indicat ing that magneti c fories on the

specime n were not a signifi cant source of error.

Another source of error which must be conside red is that at the

ends of the zig-zag the current is flowing in a differe nt directio n

to that elsewhe re, causing a mixing of the longitu dinal and transve rse

field configu rations . One observe d effect of this was that the Q

of the resonan ce decreas ed sharply at fields of the order of fH c in

both the longitu dinal and transve rse field configu rations . Howeve r,

from the short length of the end section s one would e~timat e that

the error introdu ced by them in fields smaller than those require d

to produce the sharp decreas e in Q would be less than five per cent,

except possibl y in the longitu dinal case near T c , where an appreci able

reducti on in the observe d value of a.~/).m ight result.

Finally , there is the questio n of whether the surface of the

specime n was suitabl e for this type of exp eriment . Not very much

can be said on this point, but as will be seen later the observa tions

in the longitu dinal field case are in fair quantit ative agreeme nt

with t h e predict ions of the G- L theory, which increas es confide nce

 40

in the validity of the observations in the transverse case (where

a g reement with the G-L theory is not expected). Sharvin and

Gantmakher (1961) found that straining their specimens considerably

increased the value of a (eqn. 6, chapter 1), but this may not

be a relevant point here as they were using a considerably different

frequency.

In summary, it would appear that the results obtained should

show the correct general behaviour of the penetration depth in a

magnetic field, but that the vibratory disturbances to the specimen

have prevented results more accurate than about 15% from being

obtained.

(iv) Discussion of the results

Longitudinal field

It will be seen from fig. 15 that the value of ~l/).. for given

longitudinal his not strongly temperature dependent, changing by

only a factor of 2 over the temperature range couered. The effect

tends to become less as the transition temperature is approached.

Assuming a value f or )<.of 0.16, the prediction of the G-1 theory

has been plotted in fig.13, and it is seen that it is in fair

agreement with the experimental results. These conclusions are in

agreement with those found at 2 Mc/s for their good specimens by

Sharvin and Gantmakher (1 96 1). However , it should be pointed out

that the region of validity of the G-L theory for tin is quite

small (i.e. T/T :> 2

1-X. See , for examp le Gor ' kov 1960), so that
CN

agreement would be expected only for the curve T/T ; 0.9~6 in any
c
 41

case. The regio n where the G-1 theor y is not valid has,
it is
belie ved, not been inves tigat ed theo retic ally.

Trans verse field

In the trans verse field case there is no resem

blanc e at all
betwe en the predi ction s of the G-1 theor y and
the exper imen tal
resu lts. For exam ple, for h=0.4 , the G-1 theor y pred icts
a (\l/i\ .
of 2.3% (allow ing for the avera ging effec t discu
ssed in the appen dix,
which incre ases 6,.)../~ by a facto r 2), while the maxim um obser ved
at any temp eratu re is 1.1%. More si g nific antly , the maxim um value
of the ratio in the trans verse and long itudi nal
cases is only o.6
times that expec ted in the quas i-sta tic limit
(6, allow ing for the
avera ging effec t). Furth ermo re, for temp eratu res above about 0.93
T
c
and below about 0.45 T ll~/A even has the wrong sign.
c These
obser vatio ns are in disag reem ent with those at
2 Mc/s, where the
G-1 theor y gives good agree ment with exper iment
as in the longi tudin al
case, but are more simil ar to those obtai ned at
highe r frequ encie s
(1 and 3 Gc/s ). Howe ver, at the highe r frequ encie s the low

temp eratu re sign rever sal of AA/A does not occur

and the high
temp eratu re sign rever sal occur s at a lower temp
eratu re (of the
order of 0.75 T as oppos ed to 0.93 T ). These diffe rence s could
c c
possi bly be due to aniso tropy effec ts.

One obser vatio n which may be impo rtant in testi

ng any theor y
which may be propo sed to expla in the exper imen
ts is the detai led
shape of the LU/~ a gains t H curve s (figs . 12 and 13). In the long -
itudi nal case these are roug hly p arabo lic (as
predi cted by the G-L
 42

theory), as also in the transvers e case in the re gion defined

approxim ately by 0.5 T

c
< T..:::: 0. 8 5 T •
c
At low temperat ures in the

transver se case there is an H4 term in /::;. >../)... sufficie nt to cause

its si g n to chang e fr om ne gative to positive a s His increase d.

More interesti ng, however, is the behaviou r near T. At temperat ures

c
near 0.97 Tc the behaviou r is quite well approxim ated by the law

AJ../A.>oG IHI • This relation i mplies that the quadratic approxim ation,

l(H) = ;\..(o) 2
{ 1 + a. (H/Hc) }, which must be a good one in the limit H"?O,

must break down in very small fields. Indeed, the breakdow n occurs

at such a low field that it is best seen not in the plotted curves

but in the ori g inal recorder traces. Fig. 9, taken at 0 .• 972 T , is

c
a g ood example. Owing to the noise on the recorder traces it is

difficul t to te l l exactly where the quadratic approxim ation breaks

down, but the followin g observat ions can be made:

(i) As T is reduced the ran ge of hover which the quadratic approx-

imation is valid becomes less, t h e correspon ding field H being

roug hly constant and of the order of 0.5 gauss.

(ii) At the temperatu re where the effect is most pronounc ed (.-0.97 T)

c
the quadratic app roximatio n is valid only over a ran g e lhl<0.1 or

less.

(iii) The effect is visible in traces taken down to 0.937 T , but

c
at lower temperat ures becomes lost in the noise.

The effect observed cannot be e x plained as due to t h e earth's

mag netic field, whose only effect would be to tend to smear out such

an effect if it e x i s ted. A similar re mark would a pply to e xp l a nations

 43

based on inhomogeneities of the specimen or the field, and it does

not appear possible either to explain the effect as due to the

presence of normal regions. It is, of course, possible that

smearing out of the effect has occurred in the present case, and

that the field concerned is even smaller than 0.5 gauss.

The same effect has also been observed at higher frequencies,

though not so clearly, since observations were not made in very

small fields, as they were in the present experiment. In particular,

the curves plotted by Spiewak (1958) at 3.320°K and by Richards (1962)

at 0.894 Tc are similar to the curve of fig. 9.
It would clearly be desirable for the effect in small fields to

be investigated further, using compensating coils the earth's

field and a detection system which measured dH or

dA directly.

(v) Summary of the discussion of results

(i) In a longitudinal field 4.A/A is always positive and is

roughly in accordance with the G-L theory.

(ii) In a transverse field A.l/).. is negative at low temperatures

and near T , and positive otherwise. The observations bear no

c
relation to the predictions of the G-L theory, but are in many ways

similar to those at higher frequencies.

(iii) There is evidence that in the transverse case near T the

c
quadratic approximation for the dependence of~ on H breaks down

in very small fields, of the order of 0.5 gauss.

 44

Appendix The unconvolut ion of the integrals occurrin ~

the transverse field configurat ion

When a transverse field His applied to a cylindrica l supercon-

ducting wire the field at a point on its surface is 2Hsin~, where

~ is the angle between the field direction an d the radius vector

to the point. This local field h (we abandon previous notation for

this appendix) is position-de pendent and in general different from

H, so that when one measures a field-depen dent quantity such as

the penetration depth, which in a local field h has a value f(h),

what one obtains is not f(H) (H=applied field) but an average over

the surface of the cylinder

'11'/4
f ( H) = ~lt SO f ( 2H sin ~) df1 ( 1)

where it is assumed that f depends on the magnitude of h only and

not on its direction.

It is easily verified that if f(h) has the form a+bh 2 then

-f(H) is a+2bH,
2 a fact
we have used previously in this chapter. In
general, however, f will be a more compticate d function and the

question arises how f can be found if f is known (the unconvolut ion

problem), and whether the solution is unique. This problem has

not in general been attempted in the past, in the belief that its

solution would involve a tedious trial and error procedure. The

object of this appendix is to show that such a procedure is not

necessary. A sim p le analytic solution will be derived and applied

to an experiment al curve.

Consider the expression

 45

I ( H) = {" f (1) ) d 1J
rlt
= H ): f(H cos e) sine d-8 (putting"?= H cos e)

=
2Hs~ d~Ji f(Hcose) sin0 d8
1t O O
2H ) ,r;i, \ Tr /"J..
= - J sin e d'9 d~ • f ( H cos e)
1t rp-:0 4P=O

2
= 1tH ~~ f ( H cos 0 ) dS

where the last integral is over an octant of a unit sphere and

e, 0 are polar coordinates~*

Going over now to Cartesian coordinates

x = r sin 0 cos 0
y = r sin e sin 0
z = r cos e

we obtain
2
I = 1tH ) f f ( Hz ) dS
unit
sphere
x>o, y>O, z)O
whence by symmetry we deduce

I = ~H )
unit
s f (Hy) dS

sphere
x '>O, y>O, z>o

2H J. 4tT /;].(
)
"'tr/2. si· n
& d9 d0 f(H sine sin 0)
= 1t j-=O Q:::0
'ri/2. ~ "rr/2..
= ~5. sine d0 f(H sin@ sin 0) d~
1t <> 0

* Writing as an ex- mathematican to physicists, I suppose it is

necessary here for me to interject the remark that the sphere has
no physical significance whatever. This appendix may serve as
a good example of the fact that out of mathematical abstr a ction
may come results which are useful to the physicist (as it is hoped
fig . 16 will demonstrate). ·
 46
46*
rirtJ..
= HJ. f(iH sine) sin 8 de from (1)
0

Combining this result with the definit~on of I(H), we obtain

dI
f (H) = dH
= ddH i. H
,;ra.. f(iH sine) sin 8 d0 }
J0 (2)
Figure 16

which provides an explicit solution to the unconvolution problem.

{i) E)(pev-i wieL1ta l curv e
If f is known for fields up to H, e quation (2) gives f uniquely
c
(ii) ~kvlat ed. Vr\COVlVO lvte.d C.IJt"Ve - 0

(iii) Cvrve ob·bo.it'\ ed f.rowi (ii) by a.vero.giV\g for fields up to 2H.

0
In particular, although one can carry out

measurements in a tr a nsverse field only up to about i H , information

c
can be obtained ab out the local behaviour in all fields up to H.
c
Equation 2 was tried out on the values of /)..)./}.. fro m the

observations i n a transverse field at 0.341 T (curve (i) of fig. 16).

c
Curve (ii) is the lJ..l/A curve obtained by substituting in eqn. (2).
Finally, the val ues of D.).,/ ).. obtained were substituted back in

eqn. (1), giving curve (iii). It will be observed that the agreement

between (i) and (iii) is quite g ood, consi dering the rather crude

methods which were used to perform t h e inte g rations .

I am ' indebted to Mr. N. Th. · Varopoulos for drawing my attention

to the fact that a problem to which the present one can be trans-

formed was solved by the mathematician Abel in the nineteenth century.

Abel's work ap p ears not to have be e n used previously in connection

with the present p roblem.

In view of the fact that the unconvolution process in this and

other cases investigated produced no important changes in the g eneral

shape of the curves, it wa s decided that it was not worth doing a

systematic u n convolution of all the results. In any case, any

 47

theo retic al formu la can be convo luted to compa

re with the expe rimen tal
resu lts more quick ly than the exper imen tal resu
lts can be uncon volut ed
to comp are with the theor y, as the forme r opera
tion need be done
only once to deal with all the resu lts.
 48

Chapter 4 Theory

In this chapter the theorie s which have been put forward to

attempt to explain the anomalo us (negati ve) si gn of A1/Aw ill be

describ ed.

Firstly there is the theory of Dressel haus and Dressel haus (1960).

This theory assumes a simple two-flu id model, with the superfl uid

obeying the London s'equati on and the normal fluid being describ ed

by the Boltzma nn equatio n, allowin g for the effect of the magneti c

field in bending the orbits of the normal electro ns. Some feature s
of the observa tions can be explain ed qualita tively, but the theory

has the followi n g defects :

(i) A detaile d fit of the theory to the observa tions require s

the effecti ve mass of the normal electro ns to be less by a factor of

nearly 10 3 in some cases than it is in the normal state (Richar ds 1962).

(ii) The theory does not account for the fact that the large

deviati ons from the G-L theory are confine d to the transve rse case.

(iii) The theory does not account for the small-f ield

breakdo wn of the quadrat ic approxi mation for)...

(iv) The anomalo us effects at low tempera tures observe d in the

present experim ent cannot be explain ed on the theory since at such

low tempera tures the number of normal electro ns is negligi ble.

(v) The theory does not explain the observa tions of Richard s on

the effects of i mpuriti es (Richar ds 1962).

We turn next to the theory of Bardeen (1958). Bardeen assumed

that a wave disturb ance analo g ous to that of second sound in

superfl uid helium mi ght be p ossible in supere o nductor s. In such a

 49

wave the osci llati ng quan tity woul d be not

the tota l dens ity of
elec tron s as in plas ma osci llati ons but inste
ad the temp eratu re
and the orde r param eter ns. Sinc e in the surfa ce i mp edan ce expe r-
imen ts a trans vers e magn etic field but not
a lon gitud inal one
caus es osci llati ons inn (cha pter 1), secon d s ound can be exci ted
s
in the for mer case but not the latte r. By some mech anism not
spec ified in deta il by Bard een, this migh t
cause the anom alous
sign of A)./}.. in the trans vers e case . In addi tion t h e effe cts
migh t be larg er near T owin g to the larg er
c numb er of norm al
elec tron s pres ent ther e, whil e the effe ct
of addin g impu ritie s
migh t be to damp out t h e secon d soun d wave
s, tlius ·expl ainin g the
obse rvati ons of Rich ards.

Altho ugh some type of colle ctive exci tatio

n simi lar to secon d
soun d may be resp onsi ble for the effe cts obse
rved , it does not
seem lik ely that the typ:e occu rring in heliu
m can occu r unde r the
cond ition s of the expe rime nts. In secon d soun d in heliu m the
norm al fluid unde r g oes 'hydr odyn amic flow
'. By t h is is mean t that
there is loca l equi libri um,w ith non- zero mean
velo cit~ of the
exci tatio ns of the norm al fluid , cause d by
exci ta tion- exc.i tatio n
coll ision s, whic h do not alte r the tota l mome
ntum of the norm al
fluid . Unde r the cond ition s of t h e expe rime nts on
surfa ce impe danc e,
the elec tron -elec tron coll ision relax ation
time is long comp ared
with the relax ation time for the decay of
a normal curre nt throu gh
colli sion s of the normal elec tron s with the
boun darie s of the
s p ecimen or wi t h i mpur ities . Henc e hydro dyna mic flow of the norm al
elec tron s is not poss ible. Furth ermo re, in the Bard een t h eory , the
.
I
j
wave leng th of the seco nd soun d is supp
osed to be of the orde r of
the pene trati on dept h. This is cons ider
ably shor t er than the mean
free path in the expe rime nts on pure tin,
so that agai n the
cond ition s for trea ting the moti on of
the normal elec tron s as
hydr odyn amic flow are viol ated .

The ques tion aris es as to whet her a diff

eren t type of seco nd
soun d can exis ~ in whic h the extre me oppo
site cond ition , mean free
path >-) wav elen gth is sati sfie d. This
type of seco nd soun d woul d
bear the same rela tion to the othe r kind
as that of zero soun d to
ordi nary soun d in flui ds. The pos sibi lity has been inve stig ated
inde pend ently by Thou less (196 0) and by
Jose phso n (unp ubli shed ), 11

who arriv ed by diff eren t meth ods at esse

ntia lly the same disp ersio n
rela tion . This is very com plica t ed and
has not so far been
anal ysed in deta il. How ever, the follo wing conc lusi o ns can
be stat ed:
(i) The velo city of such a distu rban ce if
it exis ted woul d be of
the orde r of the Ferm i velo city , i.e.
of the orde r of 10 4 time s as
grea t as that of Bard een' s type of seco
nd soun d.
(ii) The disp ersio n equa tion has no ph onoR
-like · solu tion s -at any tern-
perc1: ture , with the poss ible exce ptio n of
a regi on clos e to T (a
c
publ ishe d stat emen t to the cont r ary by
Thou less and Till ey has
sinc e been withd rawn follo wing the disc
over y of a num erica l erro r).
The sit u atio n c lose to T has not hith erto been inve stig ated .
c
It can be seen from the ab o ve that so
far t h ere is no sati sfac tory
theo ry to acco unt for all the obse rvat
ions and thei r comR lica ted
depe nden ce on all the para mete rs invo lved
. Poss ibly some comb inat ion
of the idea s of the Bard een and Dres selh aus
theo ries may solv e the
prob lem, but it seem s like ly that a new idea
may be nece ssar y
befo re a full expl anat ion is obta ined .
 52

Part II

SUPERCURRENTS THROUGH BARRIERS

(i) Intro ducti on

The char acter istic featu res of super cond uctin g

syste ms, the
Meiss ner effec t, zero resis tivit y, quan tized persi
stent curre nts
in macro scopi c rings and quan tized flux lines ,
are the conse quenc es
of a long- range order ing proce ss of an essen tially
quant um natur e.
Part II of this disse rtatio n deals with t h e prop
ertie s of super con-
ducti ng syste ms parti tione d by thin barri ers of
subst ances which
in bulk are not super cond uctin g. Such s y stems have · becom e the
subje ct of much recen t exper imen tal work, in con
necti on wi th both
diele ctric barri ers (tunn ellin g ) and barri ers
of norm al meta ls
(prox imity effec ts). We shall d eal with such quest ions as the

fo l lowin g: What effec t does a barri er have on the super cond

uctin g
long -rang e order ? What happe ns to the char acter istic featu res of

a super condu ctor as the barri er is made thick er?

Are new type s
of behav iour possi ble?

\ 1hat distin guish es super cond uctin g syste ms from all

other
syste ms is a spec ial prope rty of t h e param eter
which descr ibes the
order i n g p roc es s. In norma l syste ms t h e order param eter (e. g . spin

d ensi t y in a ferrom ag netic syste m) can be uniqu

ely defin ed irres -
pec t ive of the g aug e chose n for the elect romagneti
c pote ntial s. This
is not so for t h e order param eter "1 of a super cond uctor , which
chang es under a g aug e trans form ation in the same
way as do wave
funct ions in eleme ntary quant um mecha nics;
 53

A~A + grad X
~~~ _ 1
cat
ox (1 )
"1~ '/' exp l (ie* /;le) X.}
wher e~ and A are the scala r and vecto r elect roma
gneti c pote ntial s
respe ctive ly, X. is the scala r funct ion of posit ion defin ing the
gauge trans form ation and e* is a const ant havin
g the dimen sions
of elect ric charg e. Since there exist s a parti cular gauge trans -

forma tion (with /v indep enden t of posit ion and time) which mult iplie
s
~ by a const ant phase facto r and leave s the elect
roma gneti c
pote ntial s uncha nged, it follow s that the phase
of 'f is not a
phys ically mean ingfu l quan tity; it is only the
phase diffe rence s
betwe en its value s at diffe rent point s of space
that are impo rtant . I I

The theor y of Ginzb urg and Landa u (1950 ) assum

ed an order param eter
of this type. It can be shown that the theor y leads to a Meissn
er
effec t, and to quan tized persi stent curre nts and
flux lines (Ginz burg
and Landa u (1950 ), Abrik osov (1957 ), Kelle r and
Zumin o (1961 )). It
was shown by Gor'k ov (1959 ) that the exist ence
of an order param eter
with the prop ertie s postu lat~d follow s from the
micro scopi c theor y
of super cond uctiv ity, with e* equal to 2e, the
charg e of the Coop er
pairs prese nt in the super condu cting state . It will not be neces sary
in setti ng up the pheno meno logic al theor y of barri
ers to consi der
the exact mean ing of the order param eter, and
so we shall not
deal with this quest ion here.

After these preli mina ries le't us retur n to consi

derat ion of the
prop ertie s of barri ers. We have seen that the order ing proce ss
 cons ists of a long- rang e corre latio n of the phase
of yJ • If now a
barri er divid es a super condu ctor into two parts
, it is clear that
if it is thick enoug h it will preve nt the phase
corre latio ns
getti ng acros s it. It shoul d there fore be possi ble in that case

to vary the p h ases of 'fin the two parts indep

enden tly of each other .
Cons ider now the oppo site extrem e of a very thin
barri er, which we
can think of as being a two-d imen siona l array
of impu rity atoms .
As its thick ness tends to zero we appro ach a situa
tion in which
there is no barri er at all. In the absen ce of a barri er -f must
vary conti nuou sly with posit ion; it is not possi
ble to draw a
plane divid ing the super condu ctor into two parts
and to alter the
phase s of ~in the two parts indep ende ntly, as
this would intro duce
a disco ntinu ity in r in cross ing the plane .

We see that as the barri er is chan g ed from a very

thick one to
a vanis hingl y thin one the syste m loses the degre
e of freedo m
corre spond ing to the abili ty to alter the phase
s indep enden tly as
just descr ibed. By what mecha nism can this loss take place ? The
answe r is the follo wing very sti:lpl e one. The free energ y of the
syste m conta ins a contr ibuti on from the barri er
regio n which depen ds
on the relat ive phase s of t h e value s of 1 on the two sides of the
barri er, and whose magn itude becom es great er as
the barri er is
made thinn er. With very thick b arrie rs the free energ y contr
ibuti on
from t h e barri er is negli gible and the phase s
are able to vary
arbit raril y with respe ct to each other . As the barri er is made
thinn er, howe ver, the barri er ener gy be gins to
exert an influ ence
 55

and the pha ses beco me effe ctiv ely

lock ed tog eth er.
A chan g e in beh avio ur as a fun
ctio n of bar rier thic kne ss whi ch
is mor e dire ctly rela ted to exp erim
ent is t h e que stio n of whe ther
a bar rier can pass sup ercu rren ts,
i.e . cur ren ts at zero app lied
vol tage . In the l imit of a van ishi ng ly thin
bar rier sup ercu rren ts
mus t be able to flow jus t as in
the bulk met al, whi le thic k bar rier
s
do not tran smi t sup ercu rren ts.
In pra ctic e it is obs erve d tha t
tun nell ing spec ime ns of resi stan ce
less than abo ut an ohm alm ost
inv aria bly show cri tica l sup ercu
rren ts of the o~tl er of mil liam ps,
whi le spec ime ns of resi stan ce gre
ater than abo ut 10 ohm s show no
dete ctab le cur ren ts at zero vol tage
. Bef ore disc uss ing the
rela tion ship betw een this obs erva
tion and the pha se lock ing , it
wil l be con ven ient to con side r the
rela tion ship betw een long -ran ge
ord er and zero res ist i Yity in sup
erco ndu ctor s in g ene ral.
Let us con side r in the firs t plac
e a norm al met al in a stat e not
too far rem ove d from equ ilib rium
(e.g . a wire fed from a cur ren t
gen erat or). It is then pos sibl e to defi ne app
roxi mate ly a loc al
che mic al pot ent ial for elec tron s
(the Ferm i lev el). Qua si-p arti cles
con tinu ous ly mov e abo ut· from one
par t of the met al to ano ther ,
tend in g to prod uce an equ ilib rium
stat e with the sam e che mic al
pot ent ial eve rywh ere. How ever ,
sinc e the qua si-p arti cles mov e
only a sho rt dist anc e betw een col
lisi ons , it is qui te pos sibl e
to have a situ atio n whe re ther e
is loc al equ ilib rium eve ryw here
tri thou t gen eral equ ilib rium , i .,e.
the che mic al pot ent ial can chan ge
slow ly with pos itio n. Fur ther mor
e, it is only in non -equ ilib rium
 stat es that non- zero curr ents can occu
r, sinc e only then is ther e
anyt hin g to dete rmin e in whic h dire ctio
n the curr ents shou ld flow .
This sit u ati o n can be avoi ded in the case
of the supe rcon duct or,
as we shal l see, by assu ming that t h e
time depe nden ce of the orde r
p aram eter is rela ted to the loca l chem
ical pote ntia l. The e x act
form of the rela tion can be obta ined by
exte ndin g a basi c resu lt of
the Gor' kov (195 8 ) theo ry, acco rdin g to
whic h the orde r para mete r
of a supe rcon duct or in equi libri um at
a chem ic al p o tent ial t<-' is
ever ywh ere prop ortio nal to e - 2 if- t/~ The
natu ral exte nsio n of this
resu lt to non- equi li brium situ atio ns is
that the rate of chan ge of
the phas e of the orde r para mete r at any
poin t is equa l to -2 time s
the loca l chemical pote ntia l. This assu ~
mpti on is cons isten t with
the gaug e tran sfor mati on (1). From it
one can dedu ce that in a
stea dy stat e ther e can be no pote ntia l
diff eren ces in a supe rcon -
duct or if long rang e phas e corr elat ions
are pres ent in ,..jl, sinc e a
diff eren ce in t h e pote ntia l at t wo p oint
s woul d resu lt in the phas e
rela tion ship betw een the valu es of f a
t the two poin ts chan ging
with time .

The p ossi bili ty of deri ving zero resi stan

ce from long rang e
orde r in this way is conf ined to supe rflu
id s y stem s, sinc e only
in thes e does the orde r para mete r tran
sfor m unde r g auge tran s form -
atio ns in suc h a way as to allow a suit
able expr essio n for the
chem ical pote ntia l to b e obta ined . It shou ld be note d also that
the e x isten ce of a posi tion -dep ende nt
orde r para mete r perm its
curr ents to o ccur even when the ch emic
al p oten tial is inde pend ent
of p osit ion, sinc e thei r mag nitud e and
di rect ion can be d eterm ined
by the deta ils of the s pati al vari atio
n b f the orde r p aram eter.
 57

Retur ning now to the effec ts of phase locki ng in

barri ers, by
a simil ar argum ent we see that if there is phase
locki ng zero
volta g e super curre nts of magn itude deter mine d by
the phase relat ion-
ships betwe en the value s of ,f on the two sides of the barri er
may occur , but if the phase s are not locke d then
the usual mech anism
for maint ainin g the pote ntial of the super c onduc
tor the same
every where is absen t, and only resis tive curre nts
can flow acros s
the barri er.

(ii) Basic . Pheno menol og~c al formu lae

In this secti on we shall deriv e a set of pheno menol
o gical
formu lae with which to inve sti gate the macro scopi
c pro p ertie s of
barri ers. While in t h e deriv ation of these formu lae fro m
micro scopi c
t n eory it may well be neces sary to consi der t h e
full detai ls of
how quan tities such as 'f v a ry as one g oes t h rough the ba rrier ,
it i s c l ear that as far as macro sc opic p henomena
are conce rned the
probl em is esse ntial ly a t wo-di mensi onal one, and
we must first
consi der wh a t set of physi cal quan tities is neces
sary to speci fy
the pr oblem. In the first p l ace let us consi der t h e follo wi ng
set:
(i ) the or der p arame ters on each side of t h e b arrie
r.
(ii) t h e curre nts and ma g netic fiel d s in the nei Eh b ourho
od
of the barri er.

(iii) the pote ntial diffe rence ac r oss the barri er.

A di f f ic u lty aris e s in conn ect ion wit h the quan

t i t i es (ii ),
s in ce owi n g to t h e s malln ess of t h e pene tratio n
de p t h these quan titi e s
 shou ld be re garde d as vary ing in three dime
nsion s. To avoid
exce ssive comp licat ions we shal l here deal
only with situa tion s
wher e the re gion s on each side of the barr
ier exten d away from the
barr ier for dista nces lar g e comp ared with
the pene trati on dept h, and
that magn etic field s if they are pres ent are
appr eciab le only in
the neigh bour hood of the barr ier. In this case we can assum e that
the magn etic field s are para llel to the barr
ier and are deter mine d
uniq uely , acco rding to some pene trati on law,
by the field in the
barr ier itse lf. This devic e redu ces the probl em back to two-

dime nsion al form , since only the field in

the barr ier need appe ar
in the equa tions . Gene raliz ation to the case of thin films

sepa rated by a barr ier could be obta ined by

inclu ding also the
field s at the outs ide edge s of the film s,
but we shal l not attem pt
this here .

For our purp oses the set of quan titie s (i)

is more than we
requ ire. We shal l be inter este d here only in s mall
magn etic field s
and curr ents , whic h do not appr eciab ly affe
ct the magn itude of the
orde r param eter but only its pha se. Furth
ermo re, only diffe renc es
in the phas e of the orde r param eter are phys
icall y mean ingfu l. It
follo ws that the effe cts of the orde r param
eter in the neigh bour hood
of a poin t of the barr ier can ente r only throu
gh the diffe renc e
between the valu es of its phas e on the two
side s.
For simp licit y we shal l assum e that the barr
ier is flat and
occu pies the plane z=O. The quan titie s requ ired to spec ify the

probl em are now

 59
(i) 0(x ,y,t ); ar g 'j(x,y ,O+, t) - ar g '1J(x ,y,O -,t),
the chan g e
in the phas e off as one cros ses the barr ier
from ne g ative z to
p ositi ve z,

(ii) Hx ( x ,y,t) and H (x,y ,t), the comp onen ts of the tang
y enti a l
ma gnet ic field in the barr ier,

(iii) V(x ,y,t) ~·~(x ,y,O -,t) - -p.(x ,y,O+ ,t)J, the
pote ntial
diffe renc e acro ss t h e barr ier,

and (iv) j
z (x,y ,t), the curr ent throu gh unit area of barr ier.
We must now deriv e equa tions rela ting these
quan titi es. Firs tly,
fro m Maxwe ll's equa tions we have

OH OH 4rcj 1 OD
ol - a; = c
z z
+ - crr-
c ut (2)

The disp lace ment curre nt term must be inclu

ded since there may be
larg e elec tric field s in the barr ier. An expr essio n for it may
be found by rega rding the barr ier as a capa
citor . Sinc e the normal
comp onent of Din the re gion betw een the plate
s of a capa citor is
equa l to 4rc time s the char g e per unit area
on its plat es, the
displ acem ent curre nt may be expr essed in term
s of C, the self -
capa citan ce of t h e barr ier per unit area ,
and the p oten tial
diffe renc e V acro ss it, givin g the resu lt

OH OH 4rcC OV 4rcj
_JL x = z
ox - oy - ca t c (3)

Next we must find equa tions gi vin g the time

and spat ial vari ation s
of 0. Firs tly, from the relat ion assum ed (sec tion
(i)) betw een the
chem ical p oten tial and the rate of chan g e
of the phas e of 'I, whic h
11 '

may be writ ten as

 (4)
we obtain the equatio n

~ _ 2e V
ot - V (5)

To derive a formula for the spatial variatio n of~ is rather

more difficu lt. The problem was first solved by Anderso n (pri vate

commun ication) , who p ointed out that the situatio n was similar to

those occurri n g in the phenome na of interfe rence between electro n

beams in the presenc e of a mag netic field ( Ehrenbe rg and Siday 1949)

and of flu x quantiz ation in superco nductor s. By using gauge

invaria nce he was able to prove the follo wing result (Anders on and

Rowell 1963, Anders on 196 3): conside r t wo points P and Q on opp osite

sides of the barrier and outside the re gion of the ma g netic field,

joined by two curves crossin g the barrier at differe nt points A and B.

The differe nce bet ween the phase s of~ at A and Bis proport ional

to the flux enclose d bet ween the two curves, one flux quantum hc/2e

corresp onding t o a phase differe nce of 2n. This result may also
be obtaine d directl y fro m t h e Ginzbur g- Landau equatio n s, by similar

calcula tions to those used by Keller and Zumino (1961) in their

treat ment of flu x quantiz ation. In differe ntial for m the result

may be written :

~= 2ed H
ox Tc y
(6)
M __ 2ed H
oy - Tc x
wh ere dis the effect i ve thick ness of the re gi on of ma gnetic fiel d
 near the barri er, i.e.{f _':, H(z)d z1 / H( O). dis equal to~ +~+t ,
1
where 1.1 , ~ are the penet ration de p ths on each side of the barri
er
and t is the barri er thickn ess.

One more equat ion is neede d to comp lete the descr iption
of the
system , namel y an expre ssion for the curre nt throu gh
the barrie r. j •
z
This can only be obtain ed fro m basic theory , which we
defer to
sectio n (vi), b u t the most impor tant featu res can be
deduc ed from
g enera l consi derati ons. First ly we recal l that in the situa tion

where phase lockin g occur s there is a zero v oltag e super

curre nt
throu gh t h e barri er which is determined by the phase
differ ence.
We may there fore write in this case:

(7)
jz must of c ourse be a perio dic functi on owing to the
nature of~-
In the more ge neral case (V non-z ero and~ time-d ep e
ndent ) we must
expec t that extra terms will have to b e added to the
ri ght-ha nd
side of (7). One such term will be the term corre s p ondin g to the

ohmic curre nts acros s the barri er, which will be a functi
on of V.
In the case in which Vis const ant and the transmission
coeff icien t
throu gh the barri er for quasi -parti cles is assum e d to
be s mall
compa red to unity , j is given (Josep hson 1962) by an e xpress ion o f
z
the for m

In deriv in g the macro sco pi c prope rties of barri ers we

shall ke ep
only t h e t erm in (8 ) corres pondi ng to zero- volta ge super
curr ents, 6n
the groun ds that t h is term should b e the chief factor
in determ ining
 62

the basic proper ties, thoug h i t must be borne in mind that we are

thereby leaving out of conside ration a ll damping effects , which

are due to the terms omitted . We shall therefo re replace (8) by

the relatio n

1 sin ,;!
jz = j (9)
with a constan t j • Equatio ns (3), (5), (6 ) and (9) are the
1
require d set of equatio ns for providi ng a complet e phenome nologic al

descrip tion of the behavio ur of barrier s. From them can be

derive d an e quation involvi ng only~:

2 1 ;:\217( -2
~ ~ - - :!J:.. =
2 2 A sin ~ ( 10)
v ot
where v = c/(41tdC )°~- ( 11)

( 12)
2
and 'fi;/ is the two-dimensiona l operato r Since (10) is

a non-lin ear p artial differe ntial equatio n, it will not be possibl e

to obtain the general solutio n, and we shall confine ourselv es to

conside ring i mportan t limitin g cases.

(iii) Thermod ynamics

In t h is section we shall derive a relation ship b etween t h e

supercu r r e n t throu gh a b arrier under equilib rium conditi ons and the

free energy associ a ted with t h e barrier . This can be used to

investi g ate the stabili ty of time-i ndepend ent s olution s of t h e

equatio ns derived in the last secti o n. The p roof given h ere is

based on a g e dankene x perimen t, in the s pirit of classic al t h e rm o-

dynamic s . A more mathema t i cal type o f p r o of has b een given by

 Anders on (1963) .

Since we expect both the free ener gy associ ated w~th the barrie
r
and the curren t densit y throu gh it to be determ ined by the
local
value of W, it will be suffic ient in derivi ng the relatio nship

betwee n them to consid er a barrie r so small that the spatia

l
variat ions of Wover it may be neglec ted. Let us consid er two
system s A and B, which are identi cal except for the fact that
A
contai ns a barrie r and B does not, i.e. Bis a single superc
onduct in g
region . Suppos e that A and Bare connec ted to gener a tors supp lyin g

an identi cal curren t I to each. The free energi es of A and B will

be functio ns FA' FB respec tively of I. If I is chang ing voltag es
VA' VB will ap p ear across A and B. Equati ng the free energy
chang es to the work d one by the g enerat ors, we obt a in

dF. = V.Idt (i=A or B)

J. J.

so that

Now since A and Bare identi cal exce p t for the barrie r, FA
- FB is
simply F, the free ener gy associ ated with the barrie r, and
VA - VB
is equal to V, the voltag e across the barrie r.

Hence dF = VI dt = Ji. o~
2e ot
I dt = JL
2
e Id~

so that I= 2e dF
v d.0' ( 13)

Ex pressi ng (13) in terms of quanti ties p er unit area, we obtain

the more g eneral ly valid result

2e of
jz = 1f a}3 (14)

where f is the barrie r free ener g y p er unit a rea. In the specia l

 cas e of eqn. (9), therefor e, we obtain for the barrier free ener gy
per unit area:

f =- fe j
1
cos fo + const. (15)

(iv) Equilibri um properti es

If fo is independ ent of time, (10) reduces to

V 2~y.; = '\-2
"'- sin fo ( 16)
This equation was first derived by Anderson (private communic ation)

and by Ferrell and Prange (1963).

If the transvers e dimensio ns of the barrier are small comp ared

with A., one solution of (16) has fo approxim ately constant . It

then follows from (9) that jz is also constant over the barrier, and

takes its maximum value j when fo is equal to n/2. In this case j

1 1
therefore has the interpre tation of the critical supercur rent per

unit area of barrier.

Interfere nce of supercur rents

A more general solution of (16), obtained by ne g lecting the

right-han d side, and valid when j and conseque ntly ).. - 2 is s mall,
1
is

where H, H and a. are constant s. From (6) we see that this is a

x y
solution with a constant field (H, H, 0) in the barrier. This is
x y
the solution which will be set up when an external field (H, H, 0)
x y
whose direction lies in the plane of the barrier, is applied to the

barrier. The total current through the barrier is t he n given by

 where Im denote s the imagin ary part and the integr als are taken

over the barrie r. The critic al superc urrent throug h the barrie r is

given by the maximu m value of this expres sion with respec t

to
change s in a, and is

The integr al occurr ing in (17), regard ed as a functio n of a

two dimens ional vector (H , H ), is simply the Fourie r transfo

x y rm
of j , with suitab le scalin g factor s and rotate d throug h a right
1
angle. This fact allows us to make the follow ing observ ation: the

square of the critic al curren t throug h the barrie r, regard ed

as a
functio n of the magne tic field applie d to t h e barrie r, is simila
r
in functio naL:fro rm to the Fraunh ofer diffra ction patter n of
an
apertu re having the same shape as the barrie r and transm ission

coeffi cient everyw here propor tional to j • The field depend ence
1
of the critic al curren t has been studie d experi mental ly in
two
cases. For a rectan gular barrie r (17) predic ts a critic al curren t

for fields parall el to an edge of the barrie r of the form I

(sin x)/ xl ,
with minima occurr ing when the flux enclos ed by the barrie r
is a
non-ze ro multip le of the flux quantum . Behavi our of this type has
been observ ed by Rowell (1963) . The second case is the analog ue of

the interfe rence patter n produc ed by two cohere nt point source

s, and
has been invest igated by Jaklev ic et. al. (1964) , whose experi mental
 arrang ement consisted of two tunnelling barriers connected in

parallel. The critical current of the combinatio n was found to

be ap proximately a periodic functi on of the applied field. It is

interestin g to note that according to (17), the critical current

of the barriers connected in parallel is in g eneral less than the

sum of the critical currents which each barrier would have if

measured by itself. As p ointed out by Jaklevic et. al., their

experiment shows that the rang e of the supercondu cting long -ran g e

order is at least as great as the separation of the two barriers,

which in one cas~ was as gre a t as 3.5 mm.

The Meissner effect and the mixed state

Anot h er case o f interest occurs when% is everywhere small, so

that in (16) sin% can be replaced b y %. (16) then reduces to

'12% = )._-2% ( 18)

This is simply the Londons' equation in two dimensions , and i mplies

that in small magnetic fields there will be a Meissner effect, i.e.

curre n ts and magnetic fields will be confined to a re gion near the

edg e of t h e barrier and will fall off with distance from the edge

as e-r/~ Anderson ( p rivate communicat ion) and Ferrell and Prange

(1963) haue pointed out that this effect will reduce the critical

current even in zero mag netic field below the value corres p onding

to a current densit y of j per unit area.

1
For typical barriers used in tunnel l ing experiment s ,Ai is of the

order of 1 mm.

Let us now consider qualitativ ely wh a t will happ en in the g eneral

 case, given by equation (16). It is clear that the Meis sner effect

will be destroyed in sufficie ntly large fields, just as in bulk

supercon ductors, because the exclusion of the ma gne tic field

increase s the Gibbs free energy of the system, and in sufficie ntly

lar g e fields t h is increase cannot be compensa ted by the lowering

of o arrier energy obtained by having~ everywhe re small in the

interior of the barrier. The situation when the Meissner effect

has been destroyed is analogou s to that o ccurring in the mixed state

of type II supercon ductors, and t he transitio n to the mixed state

wi ll occur wh en it becomes energeti cally favourab le for a single

quantized flux line to enter the barrier. As shown by Abrikoso v (1957)

in the case of type II supercon ductors, the critical field is

simply related to t he Helmholt z free energy per unit l ength of a

flux line. This ener gy is calculate d in the appendix , and the

correspo nding critical field shown to be

( 19)
This is typically of the order of a gauss.

Ferrell and Prange (1963) have investig ated the one-dime nsional
form of equation (16) and in particul ar have found the maximum field
at the edg e of the barrier for which a solution with a Meissner

effect (field decaying exponen tially at large distance s from the

edge) is possibl e. This field they find to be (n/2)Hc ' i.e.

1
g reater than the thermodyn amic transitio n field. The solution s
with H;::, Hc must be metastab le. The explanat ion for their existenc e,
1
as in the case of type II supercon ductors (Bean and Ll.vingsto n 1964),
 68

is that there is an attract ive force between a flux line and the

edge of the barrier , so that a flux line has to overcom e a potenti al

barrier before it can get into the interio r of the barrier .

Similar ly, metasta ble solutio ns corresp onding to the mixed state

can occur in all applied fields down to zero field.

We have so far assumed the barrier to be homogen eous. Let

us now conside r briefly the effects of inhomo geneitie s. As in the
case of type II superco nductor s inhomog eneitie s will tend to pin the

flux lines (Anders on 1962), and this will lead to the possib ility

of supercu rrents being carried by the interio r of the barrier as

well as by its edge, the Lorentz force on the flux line being

balance d by the pinning force. A particu lar exrunple of this is

the supercu rrent carried by a superco nductin g bridge across the

barrier .

(v) Non-eq uilibriu m propert ies

In this section we shall confine ourselv es to the case of

dielect ric barrier s, since in the case of normal metal barrier s

the damping effects due to shuntin g of the supercu rrents by normal

current s will be very importa nt and cannot be neglect ed.

Non-eq uilibriu m propert ies of barrier s can be classif ied into

two t y pes, accordi n g to whether there is a de voltag e across the

barrier or not. Accorei ng to (5), in the former case~ will

increas e progres sively with tiille, so that oscilla tory effects will

occur through oscilla tions in sin~ and hence in j , even in the

; z
absence of applied oscilla ting fields. In the latter case~ will
 oscillate abo ut a mean value, and these oscillations will be damped

out by resistive effects unless they are sustained by applied

oscillating fields.

The non-equilibrium properties of barriers are described by

equation ( 10):
2
V
2
5i:1 - -;. q = )..-
2
sin ~
I . \ I v ot
\
I • \ In the limit of zero supercurrents, i.e. ).-2 ~ O, this reduces to
ki
z inductiv
gin ry '. '
I • \
k real
z resistive the two-dimensional wave equation . This corresponds to the fact
I .
I ' that in this limit the barrier is behaving as a transmission plane

I (two -dimensional transmission line), the velocity of propagation

I being v. We shall now consider the properties which depend
I - - -IZI specifically on the existence of supercurrents.
I
A. No de voltage

We shall deal only with the situation in which~ undergoes small

oscillati ons about the solution yj =0, corresponding to zero

/ a nd the barrier is uniform.
/ magne tic field and current.A In this case we may replace sin~ by~,
/
/ obtaining the equation
/
/

Frequency dependence of the wave number k and the . On substituting into (20) the plane wave solution 0 = exp i ( wt - k. r),
Fig. 17
edge impedance Z for small amplitude electromagnetic
waves propagated a long a ba rrier in the absence of we obtain the dispersion relation, illustr~ted in fig. 17,
steady fields and currents.
= (k).)2 + 1 (21)

where w0 = v/').. (22)

w0 is typically a few Gc/s. It follows from (21) that waves cannot

be propagated for w<.w0 • In the static limit k tends to i/)... This

 70

is merely a restatement of the Meissner effect. For w>>cu the

0
pr opaga tion velocity tends to v, its v a lue in the absence of
supercurrents.

A case of some interest is that for which w =w0 and k =O. In

this case the phase is the same over the whole barrier, and so by

(6) there are no ma g netic fields. This type of disturbance is in

effect a plasma oscillation, reduced in frequency to the microwave

region owing to the small density of charge carriers in the barrier,

and localized in the barrier owing to its fre quency b eing b elow the

plasma cut-off frequen c y in the bulk superconduc tor.

In the k=O case, in which~ is inde p endent o f position, we

may drop the assumption of 95 being small, and obtain the p endulum

equation (Anderson 1963)

2
d 95 2 (23)
- - + w sin ~ = 0
dt2 0

As mentioned previously, in the absence of de voltag es oscillatory

disturbances can be set up only if t hey are e xci te d ex t ernally, for

example by applying ac voltages across the edg e of the barrier. It

is therefore of interest to calculate t he i mpedance p resented to a

source connected across the edg e of the barrier, i.e. the voltag e

across t h e bar r ier divided by the current flowing across the edg e.

The latter is related to the magnetic field in the barrier by

Maxwell's equations, so that using equations (5 ), (6) and assuming

for 95 the p lane wave solution, valid in the case of a s emi-infinite

b arrier, we obtain for the edge imp edance per unit len g th
 71

2
Z = 4rcdw/kc (2 4 )

The variation of Z vTith fre quency is s h own i n fi g . 17. Below w

0

Z i s p ure ly i nductive , while a b ove it Z is p urely resistive. This

is due to the fact that onl y ab ove w c a n travel l ing waves carry
0

ener gy a wa y from t h e barrier.

Owi n e; t o t h e s mallness of d, Z is n or mally very s mall (< 10 - 3 ohm cm.),

b ut it is l a r ge r near w , by an a mount dependin g on the d amp ing of

0

the p lasma t y pe oscilla tion.

The phenomena discuss e d here could p rob ab l y be ob serve d in a

barrier in the for m of a thin-film transmission line, it being

necessary of c our s e to ensure t h at t h e power a p plie d was not so

g reat as to cause t h e breakdown of the linear approx imation (20).

Th e s e p aration into n or ma l modes of t h e typ e discussed here of

th e t h ermal and zero-point fluctuat i ons of~ could b e used as the

starti ng p oint of~ calculation of t h eir ma g nitude, a quantity

which would det e r mi n e whether t he p has e l o ck ing discussed in

section (i) wo u ld oc cur. Detailed calculations have hith erto not

been performe d (see, h owever, Anderson and Rowell (1963), Anderso n

(19 63) and Josephson (1964) for a qualitative disc u ssion).

B. Finite de voltag ~

In this cas e~ increases prog ressively with time, so that the

app r oxi mation t h at~ is s mall can no lon g er be made. In g eneral (10)

can only b e solved n ume rically, and the solut i ons may be e xp ected

to have v e ry comp licated behaviour, as is confirmed by e xp erime n tal

results (Fisk e 1964, Eck et. al. 1964, and v a rious un p u b lished
 72

o bservati ons). For s mall j , however, (10) ma y be solved ap p rox -

1
i mately by an itera tive procedur e, t h us allowing the main fe a tures

of t h e solut ions to b e see n . To do this we wr i te ( 10) as t wo equation s:

v20 - ..1.
2
_a:2 = . /( . J._2)
Jz J1 (25)
v at

jz = j1 sin la (26)

As a zero order appro xi mation we may solve (25 ) with the a ppropria te

boundary c ondition s, putting t h e ri gh t-hand side equal to zer o.

These boundary c ondition s will depend on the de voltag e and on vrhat

oscillato ry fiel d s if any are b eing app lied to t h e barrier. The

value of 0 obtained can then b e substitu ted into (26) to obtain a

value of j z which is use d as the ri gh t-h a n d sid e of (25) in the

next app roxi mation, a nd so on. :·Je shall n ow c o nsider s eparat e l y

the cases with an a pp lied oscillat i ng field present or ab s ent.

(a) No oscillat ing field applied

In this case an a p propriate zer o order solution of (25 )

(Eck et.al. 196 4) is

+ a. (27)

. . . . {2 e d ( H H
giving Jz = J1 sin Tc yx - xy
) 2e Vt
+ "j" + a.
} (28 )

where Vis the de p otential differenc e a n d (H, H, 0) is the steady

x y
magnetic field. We see that the supercur rent oscilla tes with ang ular

frequenc y w = 2eV/ ~ ( the ac s u percurre nt), and that if t h e ma g netic

fiel d is non- z ero the current di s tri buti on varies as in a travelli ng

wave , with v eloc i ty Vc/ lH ld. Th e value (28 ) o f j must n ow be

z
substit u ted i n to (25) to give a f i rst or d er c o rrection term ~ to~-
1
 73

~ will oscillate with the frequency w of the driving term jz. We

1
shall not here attempt any detailed calculations, but only note

two qualitative features. Firstly, ~ may be expected to be large

1
when w is near a resonant frequency of the barrier, i.e. a

frequency at which the h omogeneous e quation

2
V 2
~ + 5;z ~ = 0 (29)

has a non-zero solution satisfying appropriate boundary conditions.

Secondly, _0 may also be expected to be large when the ratio between

1 I I

the p otential difference and the magnetic field is such that the

velocity Vc/lHld of the supercurrent distribution is equal to v,

the velocity of propagation of electromagneti c waves in the barrier.

The e x istence of the _0 term will show itself experimentally as

1
a modification of the de curr ent-volta ge c h aracteristics of the

barrier. However, t h e equations set up in s ec tion (ii) are not

sufficient in t h emselves to treat t h ese modifications . This is

because of our ne g lect of dissipative p rocess.as. If a barrier h as

a de p otential across it and a de current t h roug h it, energy is

being continuously supplied to the barrier, and a steady state

sit u ation can occur only if there is some mechani s m p resent for

dissipating the energy . He may turn this ar g ument round and

o b tain a qualitative picture of the behaviour of t h e de c h aracter-

i stics by noting that any dissi p ative process must lead to a de

current throu gh t h e barrier in order t h at energy c a n be co n served.

The rate of dissi p ati on will be greater whe n larg e oscillating

electroma g netic f i el d s are present, i.e. wh en ~ is lar g e for either

1
 74

of the reasons c o nsidered previ ously (the ~O term gives rise only

to de fields when substituted into (5) and (6)). In both these cases

the condition for ~ to be larg e is one which involves the voltag e

1
across the barrier, so that the current-voltage characteristics

should show a resonant peak at a particular voltage or voltages.

This has been confirmed by detailed calculations in one particular

case by Eck et. al. (1964), and from these calculations it can be
seen that t h e e x c e ss de current arises from a distortion of t h e

waveform of the ac sup ercurrent (cf. also sub-section (b)).

Experimental evidence for the occurrence of both types of

resonant process has been obtained. Fiske (1964) has observed

step-lik e ch a racteristics with ra p id rises of current at voltag es

which are inde p endent of magnetic field, thoug h oth er features of

the characteristics are strongly field-dependent just as is the

value of the critical zero-voltag e supercurrent. The excess currents

observed by Fiske are probably due to the e x citation of standing

waves in the barrier at characteristic r esonant frequencies g iven

by (29). The negative resist a nce portions of t h e reso nant p eaks

were presumably unstable under t h e conditions of the exp eriment

and were not observed.

Eck, Scalapino and Taylor (1964) have obtained evidence for the
occurrence of the other type of resonant process. They observed

characteristics with a reson ant peak at a field-de p endent volt age

given , to within e xperimental error, by the velocity matching

condition Ve/( I HI d) = v.
One further c as e of interest is t ha t of zero magnetic field and
a uniform barrier, which reduces as in the z e ro-voltage case to

the p endulum equa tion (23). It is cle a r that in this case there

can be no resonant type of behaviour (remember that we are now

deali ng with solutions in which~ i n creases progressively with time ).

This conclusion depends spe cifically on our assumptions of ze r o

ma g netic field and barrier uniformity, which allow e quation (10)

to be reduced to an ordinary differential equation. S ome exper-

imental support for it is provided by the observation that when

barriers are made by a method which gives g ood uniformity of

barrier thickness the step-like structure often becomes more

pronounced in a magnetic field.

For c ompl eteness it s h ould be noted that a more complicated

typ e of behaviour is also possible (Shapiro 1963), in which the

barrier voltag e jumps back and f or ·e b etween two different values at

a r ate determined by the external circuit parameters. This

behaviour may be due to a negative resistance re gion of the

characteristic, as sug geste d by Shapiro, or to t h e fact that t h e

load line of the barrier may not intersect its de chara cteristic

(as happens in the neon-tube time base).

( b) Oscillating fiel d s applied

When both de voltag es and oscillating fields are app lied

to g ether to a barrier a ne w p ossibility arises; the ac supercurrent

resulting from t h e d e volt age may b ecome locked to the oscillations

of t h e a pplied field. This can happ en only if the ratio between

the t wo frequencies is a simple rati ona l numbe r. If l o ck ing of

 76

the ac supercurrent to an applied field of fixed fre quency occurs

then the de component of the current is free to v a ry, since the

system has available one degree of freedom conresponding t o the

possibility of a ltering the phase relationship between the two

oscillations (a specific example will be c onsidered later), while

on the other hand the de comp onent of the voltag e is fixed by the

voltag e-frequency relation (5), since the mean value o f d~/dt is

fixed by t h e frequency of t h e field to which t h e ac supercurrent is

locked. Th is situa tion is an a l ogous to that occurring in the

discussi o n of the de s u percurrent (section (i)), where the voltag e

i1as fixed a t zero while the current could v a ry over a finite ran g e.

In the p resent case the de characteristics will have a number of

c o nstant-voltag e re gions at finite voltag es as well as at zero

voltag e. The volt ag e of these regions sho uld be exactly constant

as far as the p resent t h eory g oes, being determined only by the

fr e quency of the applied oscillating field, while in the case of

Fiske's exp eriments t h e voltage is determined by the fre quency of a

res o na nce, which must have a finite wi d th. Fr om a more practical

viewpoint we may say t h at a barrier whose phase oscillations are

locked to t h o s e o f an external oscillator of fi x ed frequency should

behave as a p erfect volt ag e st a bilizer, i.e. within limits c h an ging

I'1'!
the current throug h it has no effect at all on the voltag e. Pippard

( p rivate communication) has pointed out t h at this pro p erty mi ght b e I

used for a n a ccurate detef.mination of the c onstant h/e on which the

voltag e-frequency relationship depends.

 77

We shall now make a quantitative analysis in the case where

s patial variations may be ne glected and j is s mall. Since j is

1 1
s mall, we ma y asoume t hat the supercurrent has no influence on the

voltage across t h e barrier, which is therefore simp ly t h e sum of t he

app lied de and oscillating potentials:

v = v0 + v1 cos wt (30)

From (5) we therefore obtain

!i1 = (2eV /}{)t + (2eV /p{w) sin wt+

0 1
a. (31)

so that jz = j 1 sin{.(2eV0 /}{)t + (2eV /}{w) sin wt +

1
a.}
=j1~ [..rn (2eV /}i{w) sin {(nw + 2eV0 /}i{) t +
1
a.}] (32)
l'l=-oo

,·,here J is t h e Bessel function of order n. ;fo see t hat if 1.eY0 /l{w

n
is an inte ger n t hen j will have a n on-zero mean va lue which
z
depends on t he c onstant of inte gr a tion a.:
J= (-1 )nJ J (2eV /}(w) sin ex. = (-1 )nJ J (nV /v ) sin ex. (33)
z 1 n 1 1 n 1 0

jz can therefore b e either in t he direction of t h e a pp lied de

potential or a gainst it, and to c onserve ener gy power mu s t

accor ding ly be supplied to or absorbed from the applied oscillating

field. jz can vary between the limi ts+ j Jn(2eV /}i{w), so that the
1 1
size of the step a t de volt age np{w/2e is e qual to 2j Jn(2eV /}{w)
1 1
-fore
and thereAva ries in an oscillatory manner with the amplitude of the

oscilla ting c omponent of the fiel d>V (cf. the situation in p4 ot on

1
assisted single- particle tunnelling ( Tie n a nd Gordon 1963) ).

The predicted behaviour of t he de c har a ct eristics of barriers

i n a microwave field, includi ng t h e possibility e x~ected from ( 33 )

 78

of de curre n ts flo wing in t h e r everse dir e ction to t h e app lied

voltag e, h as been observe d by Sh a~iro (1963) and Shapiro et. al.

(1964). Howeve r, in real barrier s a d ditio n al co n tributi o ns to t h e 'I I

I
curre n t not conside red here s h ift t h e co n stant-v olt ag e steps so

t h at t h ey are n ot symmetr ical abo u t zero current as p redicte d by (33).

(vi) Microsco ~ic t h eory

The tunnell ing Ha miltoni an method

Th e basic p roblem of t h e microsc o pic theory is the d eriva tion o f the

relatio n betwe en j
z
a nd 5D. One meth od , p_art'i c u1:a:rl_y usefu l for

dielect ric barrier s, is the tunnell ing Hamilto nian met h od of Cohen,

Falicov and Phillip s (1962). In t his me th od it is ass umed t h at t h e

Hamilto nian o f t h e s y stem can be wr i tten in the f or m

(34)
where H and H are the Hamilto nians for the re gi ons on the t wo
1 2
sides of t h e barrier and RT is a term describ in g tunnell ing across

t h e barrier , assumed to be of the form '!_ T1 r ai a r, where a and ar

1
are electro n annihil ation opera tors f or states on opp osite sides of

the barrier an d HT is t h e tunnell ing matrix element . The tunnell ing

Hamilto ni an a p pro xi mation is equival ent to t h e assu.~p tion that

tunnell ing is an instanta neous process during which t h e effects of

interac tions may b e .i gnored.

Since the effects of tunnell i ng are i n clude d e n tirely in the

t e rm HT' it is reas onable when calcul atin g the tunnell ing c u rren t to

r e gard t h is term as a pe rtur bati on . F or calcula ting the zero-

voltag e curre n t t h e s i mp lest meth od is to calcula te t h e free e n er gy

I
I,
 of the barrier by perturb ati on theory and use (14) to de rive the

supercu rrent (Anders on 1963). F or calcula tin g the current at

non-zer o voltag e one must use ti me-de p endent p erturba tion t h eory.

At t = - CIO the s ys tem is assumed to consist of t wo indepen dent

subs y stems each in equilib rium at differe nt potenti als with a

definit e (time -depend ent) value of~. The perturb ing term HT is

suppose d to be s wi tche d on adiab a tically and t h e barrier current

resulti n g is calcula ted. A calcula tion in which it uas ass um ed

that t h e Hamilto ni a ns H and H _could' he approx i ma ted by t ho se of

1 2
non-int eractin g quasi-p articles h as been given by Josephs on (1962 ).
In this calcula tion there are cert a in difficu lties in t h e formali sm

d ue to the fact that t h e usual simplif ication of me asurin g all

energie s relativ e to t h e Fermi ener gy cannot be made since the Fer~i

level is n ot t he same on both sides of the barrier . This difficu lty

is related t o t he fact that~ is time depende nt. One way of avoidin g
it,whic h has t h e adva nt a s e of not being limited to the weak couplin g

limit in which indepen dent quasi-p articles exist, is to express the

correla tion functi ons which enter into the formula for the tunnell ing

curre n t in ter ms of Green's functio ns (Ambega okar and Baratof f 1963).

Two terms are found, one which involve s t he n or mal sin g le-elec tron

Green' s functio ns ( G functio ns) and is the s am e as the express ion

found for quasi-p article tunnell ing by Schrief fer et. al. (1963), I 1111

and one which involve s the Gor'kov F functio ns and is a periodi c

functio n of~. This latter term ave ra g es out to zero f or non-zer o

p ote n tial differe nces, so that in an experim ent me asurin g t he de

 80

current-vo ltag e characteri stics the f or mula of Schrieffer et. al.

e ives the correct result in the approximat ion used by .A.mbegaokar

and Baratoff. The term involving F functions gives rise to the

ac and de supercurre nts, and may be regarded as the contributio n

from the tunnelling of 'Cooper pairs'. The Cooper pair sta te

(Cooper 1956, Gor'kov 1958 ) is a boson state of charge 2e wh ose

occupation number is macroscopi c. Its existence in supercondu cting

systems is imp lied by the long-rang e order of the phase of the

order parameter. The energy of the Cooper pair state is twice the

Fermi energy, and there is a single state only when the order

parameter has phase coherence over the whole of the region which

is supercondu cting. * Wh'. e n a finite voltag e is app lied across a

barrier, t h e time-de p endence of~ implies that phase coherence is

present only in each of the sub-re gions on either side of the

barrier, and instead of a single Cooper p air state there is one

state for each side of the b a rrier. These have different energies

owing to t h e potential difference across the barri~r, so that

the processes involving tunnelling of Cooper pairs across the

barrier are virtual processes. These virtual processes constitute

the ac supercurre nt. Real processes involving the tunnelli ng of

Cooper pairs, in which the ex istence of a dissi p ative process, or

'
the pos s ibility of interaction with photons from a microwave field, II
allows the ener gy balance to b e k ept, are also possible, and have
*An
I'
instructiv e analo gy can be found in the behaviour of lasers. When
a laser oscillates in a single mode, i.e. a sing le photon st a te has
macrosco pic occupation number, the electroma g netic field has p h a se
coherence over the who le volume of the laser. In t h is a nalog y the
photons and electromag netic f i eld of the l a ser corres p ond to the
Cooper pairs and the order parameter of the supercon d uctor.
 been discuss ed from a differe nt viewp oint in section (v) B.

The mag nitude of t h e zero-vo ltag e supercu rrent has b een

calcula ted by Anderso n (1963) at absolut e zero an d by Ambe g aok ar

and Ba ratoff (1963) for fi nite tempera tures. An i mp ortant result

of the calcula tion is t h at the supercu rrent terms are of t h e same

order of ma gnitude as t h e quasi-p article current s o cc u rring at

vol t a g es of t h e order of t h e ener gy gap.

The Gor'kov method

It is possibl e to calcula te t h e zero volt age supercu rrent in a

way that avoids complet ely the tunnell ing Ha miltoni an met h od and

d escribe s propa ~ation through t h e barrier in terms of Gr e en's

functio ns. This met h od h as t h e advanta GeS t hat it shows p recisel y

the factors determi ni ng t h e ma gnitud e of t h e sup ercurre nt, it

indicat es t h e rel a tion between sup ercurre nts through barrier s and

the usua l kind of sup e r current , and it is not restric te d to

dielect ric barrier s. Th e method is based on t ha t used by Gor'kov

(1 95~ Jto derive the Ginzbur g - Landau equati ons fro m microsc opic

theory.

We shall u s e in our proof the st a tistica l Green's functio ns

G and F + , defined by
I'
G(r·r, r '1'') = ( T ( '1-(r, Thyl(r' , i')i (35) I I

F + (r1", r '1' ' ) = ( T('1l (r ,'T)·J/ (r •, '1 1 ) )) (36)

where T is t h e time-or dering operato r and for any operato r

A('Y) = e Hi, Ae -H""r

For simplic ity we ahall use units in which~ = 1, Boltzma nn's constan t=1
 and energie s are measure d from the Fermi level (so that
fL = O).
+
G and F may be express ed in the usual way as Fourier series:

G(ry,r 'f') = Tl e-iw(y- r•) G Cr r•)

(37)
w (I) '

the sum being taken over the values w=w: (2n+1)n T for integra l n,
n
T being the tempera ture and G being given by the relatio n
w
1/T
G (r,r')
w
= 5 O
(38)

with similar relatio ns for F+. Taking as the single- particl e

Hamilto nian (includ ing self-en ergies but i gnoring interac tions)

Hr = 21m (i\7 + ~ A(r)) 2 + V(r)

r c (39)

we may write the Gor'kov equatio ns in the form

(iw-H )G (r,r') +~ (r) F+(r,r' )

r w w = ~(r-r•) (40)

( i w-H * ) F + ( r , r
r' w
1 ) - J.. * ( r
l>
') G
-w ( r ' , r ) =0 (41)

H* being the operato r obtaine d fro m H by t h e replace ment i~-i •

.6.(r )=gT ?- F. ·er ,r') is the energy gap paramet er, g being the
(I)
(I)

interac tion constan t.

Using the Green's functio n for non-int eractin g pa rticles

,,...,
G (r,r'), which s a tisfies t h e equati ons
w
.IV
(iw-H )G (r,r')
r w =t (r-r') (42)
* ,-,
(iw-H , )G (r ,r') = o(" (r-r •) (43)
r w
(40) and (41) can be convert ed into the inte gral equatio ns
G (r,r')
w
:'Gw(r,r') - !'Gw(r,s)~( s)F+(s
w
,r') ds (4l1-)

F+(r,r' )
(I)
= l(G (I)
(s,r')6 . *(s)G
-w
(s,r) ds (45)
 Equation (45) is sli gh tly differen t fro m Gor'kov' s e quation, being

obtaine d fro m it by interc h anging G and ""'

G. This for m will b e I:J. ore

convenie nt for our pur p oses. Ob tainin g a differen t equation was the

result of usin g an e quation for F+ derived fro m the equat ion of

motion for \f't(~) r a ther than t h e one for \f't(r).

F+ ma y now be eliminate d from (41+) and (45), giving the result

G (r,r')
w
=Gw(r,r') - (('cf (r,s)a(s )
JJ w G '
-w
(s' ,s)~*(s 1 )G (s' ,r') ds ds'
w
I I

(46)
The expectat ion value of t h e current density,
2
j(r)= ({ie(V ,-'2) , - 2e
m r A(r)}·"t'C r 1 ,o) cl,(r,o)) (47)
r r =r me l 1'
may now be found by expressin g it in terms of G (equation 35), as
follows:
2
j(r) = _s1. iecv
m r
,-V) , - 2 e A(r)}G(r
,O; r' ,O+) (48)
r r =r me

Note t ha t in (47) and (48) is included a factor 2 due to summation

11
over spins. If in the right..:.han ct side of (48 ) we replace G by rv
G,
we obtain the current density for a non-inte ra cting system, which

is zero in equ ilibrium . Hence (48) remains true if in it we replace

"'
G by G-G, and using (37) and (46) we obtain

j (r) = T ~L.w { -(V

ie
m r
,-\7 ) ,
r r =r
2e
2 (r)
- -A
me
}ill w -w
IV*
(r ,s)A(sY' G (s!ts )l}(s' )G(s ; r ') ds ds'
w
(49)
I i

=SS K ( r , s, s' ) D.( s) .t l * ( s' ) ds ds' (50)

where
ie
X [- ( G (r
m 1 w
f'J
K(r,s,s') = T

'r"'w
!.
w
G
-w
,..,
, s ) '('7<} ( s ' , r ) -G
w
(s' ,s)
rv
( s ' , r ) V. G ( r , s
r w
?-
'J -me
2e
2
-
w
""'
A ( r) G ( r , s ) G ( s
w
1
,
u
r)

(51)
 84

(49) is valid at all tempera tur es and assumes only the validity of

the Gor'kov equations. As a first approximat ion, valid n ear T

c
,v
where bis small, we riay replace G by G. K then becomes independen t
11
of /1 and (49) reduces to Gor'kov's result. Gor'kov derived the

Ginzburg - Landau equa tions from it by assumin g .Ato be slowly varying

with position and ex uanding Aina Taylor series about the point r.

If we are calculatin g the current at a point in a barrier t h is

app roximation is clearly n ot a g ood one. Howeuer, since K has a

ran ge of t he order of a coherence length and t h e important variations

in A. take place over a much smaller ran ge, i.e. t h e barrier thickness,

we may make a different approxi mation, namely that in the neighbour-

hood of r b. take s different constant values l\.1 and A2 on the two

sides of the barrier. ( 50 ) then reduces simply to t he form

j (r) = L
. .
K ..
J.J
* J
!J.D.
J_
(i,j=1,2) (52)
J
1.,

Further, no supercurre nt can flow acros s the barri er unless both I

sides are supercondu cting, so that the ri gh t- hand side of ( 52) II
must vanish if either A. or
1
62 is zero. Hence K :K :0,
11 22
If I
time-rever sal symmetry is present j must change si gn when /::;. is
I
replaced by uA::f.• In addition j must be real. j therefore reduces

to the form
(53)

where K is real. Putting 0 = arg A2 - arg A, and j

1
=-a-K1~1 I I~ I ,
this reduces to (9).

More li gh t is thrown on the factors determinin g the magnitude of

the supercurre nts if we calcula te inste a d of the current density the

 total current through the barrier. Let S be a surface whose

intersection vii th the superconductor lies inside the barrier, and

which divides space into two parts v1 and v2 • If n is a unit

vector normal to S p ointing into v 2 , the total current through the

barrier is

(the suffix r denotes that the surface integral is taken with respect

to t h e point r). The quantity r{K(r,s,s').~d Sr can be calculated . I

by usin g the identity I
tlr [~ {a w(r,s)VG (s' ,r) -Gw(s' ,r) (l Gw(r,s)} 2 e_A(r)G
__ 2 ,.,, (s',r)~
"" (r,s)G
m r w r me w w

= 2ie{Gw(r,s)b(r-s') - Gw(s' ,r) ~ (r-s)1.s (55)

This identity can be proved by direct computation using equations

(39), (42) and (43), but it is more instructive to derive it fro m

the law of conservation of charge , as follo ws. Consider two single-

-
electron states Is) and Is'), with wave functions

=G (r,s)
w
={G/s',r)\
* (56)

respectively. Note t h at 'l.s and \JI s , are ordinary functions of t h e

p osition coordinate r and not field operators. From (L~2) and (43)

they satisfy t h e equations

H -.}' (r) = iw
r s
"iJ.s (r) - ~(r-s)
(57)
H ~ ,(r) = -iwcl) (r) - ~(r-s')
r s Ts•

Now if Hr is a charg e-conserving Hamiltonian and p(r) is the charg e

density operator then

 86

vr j(r) = -i (Hr,P (r)] (58)

so that ~ (s 1 lj(r)ls) = -i(s 1 jHrf(r)-f(r)Hrfs)

= w(s•Jp(r)ls) + i(hs 1 lp(r)(s)- w <s•fpCr)fs)- i<s•lpCr)[s)

from (57),

(59)

where 6,
s ~,
s are the states with wave functions b(r-s), ~(r-s•)
respectively. If we now substitute the expressions
.
=-~: ~ )( *
<' ') = C,/'i
2
i j(r) k (r)Vr'4_(r)-t_k:.(r)V rt (r))-:cA(rhpi (r)'fk(r)

"( il,O(r)lk) e~(r)'fi/r)

and use the definitions (56), then (59) reduces to (55).

and the definition (5 1)
Using the identity (55)Awe obtain

5{ K( r , s , s ' ) •n1 dS r = J'Or K( r , s , s ' ) dr

v1

= 2ieTl_ G (s' ,s)r (G (r,s)&'(r-s•) - G(s',r)S(r-s))dr I

w -w Jv 1 w w 111

= 2i e T r
(I)
G
-w
(s ' , s )Gw( s ' , s) {n ( s • ) - n ( s ) 1
I

where n is a function which is equal to 1 in v and O in v • Noting

2 1
t hat the funct i on n(s') - n(s) is zero unless s and s' a re on

opposite sides of S, we finally obtain

I = 2ieT ( ~)
s'ErV 1 , s~V
- Jfsev ,s•~v
) ds ds'
2 1 2
,..,,,
XIw G
-w
(s',s) G (s',s) A(s)
w ~ *Cs•) (60)

It will be observed that (60) has the desirable property that the

I
 87

Gree n's funct ions

expre ssion for the curre nt t h r ou gh S conta ins only
s' on oppo site
repre senti ng propag ation throu gh S (i.e. with sand

sides of S). In addit ion it is gaug e-inv arian t.

fican ce
The diagr am repre senta tion and its phys ical signi
e by repre senti ng
Equa tion (60) may be given phys ical si gnifi canc

it by Feynm an diagr ams (fig . 18). In the diagr ams the direc ted

ane line
lines repre sent singl e-p artic le Gree n's funct ions,
"' ion, and the y are
+ repre senti ng a G funct ion and the other a G funct
x to the Gree n's
label led by the frequ ency which occur s as a suffi

funct ion. A dot at a verte :~ repre sents a fa c t or t>* or Cl according

into the ve rtex, and
to wheth er two direc ted lines e mer g e from or go
at a verte x r:ms t
the sum of the fre quenc ies label ling the two lines
into
be zero. The dashe d line rep~e sents the barri er, and takes
+ accou nt the facto r n(s') -n(s ). More prec isely , one must multi ply

r of elect rons
each diagr am by a fa ctor p ropo rtion al to the numbe I

O, 2, O, -2
which t h e diagr am shows cross ing the barri er (i.e. 11

3 respe ctive ly for the diagr ams of fi g . 18). The quan tity which a
I
ect to the coord inate s
diagr am re p resen ts must be integ rated with res p

+ of the ve r tices and summe d Hith res p ect to frequ

ency in the usual

prop ortio nal to the

way, and the sum ov e r all diagr ams is a quan tity

barri er curre nt. Phys ically , we may say t h at the diagr ams repre sent
s', propa gates to
3 the follo win g proce ss: a Coope r pair scatt ers at
I
sand then scatt ers a g ain. If in going from s' to s i t cross es t h e

supe rcurr ent. From

barri er, this ~rece ss c ontri b u tes to the barri er
n in the barri er
this we see that the exist ence of an inter actio
 88

itself is not a necessary condition for supercurrents to be able

to flow throu gh the b arrier.

Processes involving more t h an two Cooper p airs can also occur,

and it is these processes which require on e of the Green's functions

,oJ
to be a G instead of a G. In fact, if an expression for Gin

terms of 'a,~ and A* is obtained by iter ating (46) and substituted

into (60) the resulting expression is represented diagrammatically

by an infinite set of connected diagrams, a typical one being that

of fi g . 19, which represents a p rocess involving four electrons

crossing the barrier and four scattering processes inv olvin g the

supercon ductin g interaction. In these diagrams all directed lines

,.,
are G functions.

If one examines the fore going explanation of t h e processes

occurring mor e closely some rath er curious f ea tures e merge. For

example, Feynman diagrams normally represent probability amplitudes

for p rocesses and one would therefore have expected the current I I
Fig. 19 I

throug h the barrier to have been the square of the modulus of the

sum of the Feynman diag rams ins t ead of just the sum. A similar

feature has also b-een noted in p revious analyses (Josephson 196 2,

Ferrell and Prange 1963), and it is res p onsible for the fact that
supercurre nt s t h rough barriers, whic h are due to second order

process es, are similar in ma g nitude to the first order quasi-particle

processes. Another curious feature is that the c ontribution to the

current fro m each diag ram is in ge neral comp lex, t h ough the sum is

real, as it must be, each diag r am havin g a complex c onju g ate

diagram obtained from it by reversing all its arrows. It is clear

t hat i n tuition is of no g reat help in un derstanding the s upe rcurrent

as a flow of Cooper pairs. Eowever, as will now be shown, there

are some ver y familiar situations in which simi lar difficulties

arise if one tries to interpret t h em in terms of a · flowof p articles

(for a different example, see Ferrell and Prang e 1963).

Consider a charg ed particle in a harmonic p otential well,

interacting with an external oscillatitig .~lec t ric field. The

i nteraction between the particle and the field may be p ictured as

follo ws: the quantum number of t h e state of t h e electron may cha n g e

by ±1 and a photon be simultaneously absorbed or e mitted. The

matrix element for these p rocesses is proportional to the electronic

charg e e, and as shown by standard t h eory the probability of

emission of a photon (in t h e absence of an applied f i eld) or

absorption or stimulated e mission (in the presence of a continuous

spectrum of electromagn etic radiation) is proportional to the

2
squared matrix element, i.e. to e . Now consider instead a coherent

applied electroma gnetic field oscillating in phase with the electron.

If t h e quantum numbers of the electron an d the field are larg e we

may calculate classically t h e mean rate at which p ower is transferred

from the field to t h e electron:

P = eEv sin $0 (61)

where E and v are t h e root mean s quare values of t h e electric field

and t h e veloc i ty of the el e ctron r e spectively and $0 is the phase

difference between the t wo oscilla tio n s. Since the photon ener gy

is fixed, n, t h e mean rate at 1vhich quanta are transferred from

 90

the field to the electron, is proportional to P and hence to e sin~,

i.e. proportional to the ma tri x element and no t to its square. It

should be noted that the usual perturbation the ory tre a t ment for

the rate at which a process occurs is not applicable here, since

the spectrum of final states is not quasi-continuo us as is usually

assumed. Note also the ap p earance in (61) of the factor sin <J,
just as in the case of barriers (equation 9).
At this p oint we may take a lead fro ~ the treatment of barriers

and say that there ate two contributions ton, one proportional to

ie exp(-ilJ) from the abso rption of photons and one proportional to

-ie exp(ilJ) from the emission of photons. As with other interfe rence

processes, any attempt to observe which way any individua l process

h as taken place alters t h e nature of t h e effect. To be able to

tell wh ether between two :·. instants o f time a photon has been

a bs orb ed or emitted it is necessary to know the quantum n umb er of

eit h er the electron or the field at b oth times to better than unity,

which destroy s knowled ge of t h e phases. Equation (61) h as meaning

only when averaged over a lar g e number of quantum processes.

Interference effects

After this di gression, let us return to the discussion of

barriers. In particular we shall consider the question of whether

the summations and inte grations i mp lied by equation ( 6 0) involve

quantities whi ch tend to h ave t h e same sign ( more precisely, whose

real p arts tend to have the same si g n), or whet he r the si gns tend

to be random, whic h will tend to reduce the ma g nitude of the

supe rcurrents.
In the first place, phase chang es may occur in the p roduct
 91

ll(s)A*( s') as s and s' mov e in a direction p arallel to t h e p lane

of t h e barrier. Th e effect of t h is on the barrier s u percurr ent is

just t h e interfere nce effect discus s ed in sect i on (iv), and the

phase chan g es occur on l y i n the presence of a ma gn etic field. Then

t h ere is t h e p os s i bi l i ty of in terferenc e b etwee n t h e t wo inte g ral s

in equation ( 6 0). These chang e relati v e pb,ase as the phases of 6.

on the t wo sides of t h e b arrier cha n g e relative to each ot h er, but

t his effect merely gives rise to t h e sin~ dependen ce of j • Ne x t

z
there is t h e p ossi bi lity t h at A ma y c h an g e si g n as on e i aves n o r mal
to t h e b a rrier, owi n g to the Coulomb r epulsion dominatin g in t h e

barrier. Th i s eff e ct is probab l y s mall for t ypical t u nnelling

barri ers s i nce t h eir t h ickn ess is so s mall, but it ma y play an

i mportant r ole in barr iers of normal metal. Finally, th ere is t h e

p ossibili ty o f fl u ct u a t i ons in si g n of G
-w ""' w
(s•,s)G (s',s). Since
Green's f unctions oscillate in si gn wit h a peFiod of the order of
I I

an inverse Fermi wave- number, one mi g h t e xp ect t his effect t o be

ver y i mp ortant, but we shall see t h at in t h e ab se n ce of ma g ne tic

i mpurit i es this is not so. If ma gnetic fields and i mpuritie s are

absent t h en H* = H , so that by (42)

r r
(',,
G (s•,s) 5:-""' (s',s) }*
= {G
w -w
,.,
In additi on G a nd G are identica l at the tran s iti o n temperatu re a n d

g enera lly tend to have t h e s ame p hase ( G d ecreas e s fas t er wit h

distance s e p arat ing t h e t wo ar guments than N.

,...,
G). Hence G (s 1 ,s)G(s 1 ,s)
-w w
te nd s to h ave t h e s am e sig n as IGw(s•,s)f, which is alwa y s p ositive.

Thus non-ma gne tic i mpuritie s s hou l d no t h ave a drastic effect on

 92

barrier supercurre nts, but with magnetic impurities destructive

interferenc e is possible. This distinction between the effects of

magnetic and non-magnet ic impurities is similar to that occurring

in bulk supercondu ctors (Anderson 1959).

To conclude this section, it should be pointed out that the

method used here may be g eneralized in several directions . For

example, spin-chang ing interaction s can be dealt with by using a

G, 1 and F+ which are matrices in the spi·n indices. Non-l o cal

interaction s can be dealt with by using a 11 which is a function

of two position coordinate s instead of one, and retarded interaction s

by using a l1 which depends on w. Once t h e ap p ropriate g eneraliz-

ation of the Gor'kov equation h as been written down, the subsequent

manipulatio ns are essentially identical to those carried out here.

Appendix Calculatio n of Hc
1
As pointed out by Abrikosov (1957), the field at which the I I

transition to the mi x ed state occurs is e qual to 41tF/~, where

0

y1
0
= hc/2e is the flux quantum and F is the free energy of an

isolated flux line per unit length. This expression comes fro m usin g

t h e equilibrium condition dF = H dM (M = ma g netization ) and the fact

that the magnetic moment contrib uted by unit leng th of a flux line

is y1
0
/4rc.
The free ener gy of a barrier per unit area is

(62)

The first term is that due to t h e barrier itself, and was derived in

sec t ion (iii). The second term is t he ener gy of the ma gn etic

field plus t he k inetic energy of the supercurre nts in t h e penetration

re g ion. This contribution is clearly proporti onal to the s qu a re

of the field in the barrier and so to (V0) 2 , and the c o efficient

can most simply be found by using the fact that a p plying the

v a riational principle to the free energ y expression obtained from

(62) must lead to (16). For the barrier containing no flux line the

appro p riate soluti on of ( 16) is 0 = 0, and the solution corresponding

to an isolated flux line in the barrier is the one derived by

Ferrell and Prang e (1963) in a different con n ection:

1
0f =2 sin - sech (x/l.) (63)

which clearly corresponds to a disturbance localized near the line

x = O. Substituting into t h e free energy for mula (62), we obtain

for the additional energy per unit length of a flux line

= (°" 2
(2p{j /e)j~ sech (x/l) dx
1 = s.ao
(2p{j )./e) -c4 sech
1
2
x dx

= 4.tfj 1 J../e (64)

= (2~ 3 j 1 c 2 /ne 3d) .12 (65)

Therefore Hc 1 = 4nF/(~:)
(66)
 REFERENCES

Abrikosov, A, A. (1957), Soviet Phys. JETP, 2, 1174,

Adkins, C. J. ( 1961), J. Sci. Instr. 38, 305.

Ambe gaokar, V. and Baratoff, A. (1963), Phys. Rev. Letters, .:!.Q., 486

and .:11, 104 (erratum).

I
Anderson, P. W. (1959), J. Phys. Chem. Solids, .:11, 26.
Anderson, P . W. (1962), Phys. Rev. Letters 2, 309.
Anderson, P. W. (1963), Lecture at the Ravello Spring School.

Anderson, P. w. and Rowell, J. M. (1963), Phys. Rev. Letters

19.' 230.
Bardeen, J. (1 958), Phys. Rev. Let ters .:!., 399.
Bean, c. P. and Livingston, J. D. ( 1964) , Phys. Hev. Letters _g, 14.
Cohen, i~: . H. ' Falicov, L. M. and Phillips, J.C. ( 1962) ,

Phys. Rev. Letters f, 316.

Cooper, L. N. (1956), Phys. Rev. 104, 11 89.

Dresselhaus, G. and Dresselhaus, M. S. (1960), Phys. Rev. 118, 77

and Phys. Rev. Letters ~, L,.01.

Eck, R. E., Scalapino, D. J. and Taylor, B. N. (1964),

Phys. Rev. Letters .12., 15.

Ehrenberg, W. and Siday, R . E. (1949), Proc. Phys. Soc. (London), B62, 8.

Fawcett, E. (1960), The Fermi Surface, ed. Harrison and Webb, Wiley,

N.Y. and London, p. 197.

Ferrell, R. A. and Pran ge, R . E. (1963), Phys. Re v. Letters 10, 479. *

* Assiduous readers detecting an inconsistency between the declaration
mad e in the first paragraph of the preface and reference 4 of this paper
!=lr.~ .as&ured that' I h:1.ve ne:ver even set foot in the state of Massachuset t s,
let alone submitted a Ph. D. t hesis to any university there. Also,
(cont.)
Fiske, M.D. (1964), Revs. Mo d. Phys. 36, 221.

Ginzburg, V.L. and Landau, L.D. (1950), J. Exptl. Theoret. Phys.

( USSR ), 20, 1064.

Gor'kov, L.P. (1958 ), Soviet Phys. JETP z, 505.

Gor 'kov, L.P. (1959), Soviet Phys. JETP 2., 1364.
Gor'kov, L.P. ( 1960) , Soviet Phys, JETP .lQ., 593.

Gorter, C.J. and Casimir, H.B.G. (1934) Phys. z. 35, 963 and
Z. Techn. Phys. 12., 539.

Jaklevic, R.C., Lamb e, J., Silver, A.H. and Mercereau, J.E. (1964 ),

Phys. Rev. Letters 12, 159.

Josephson, B.D. (1962), Phys. Letters ..:!_, 251 (submitted with this
dissertation).

Josephson, B.D. (1964), Revs. Mod. Phys. 36, 216 (submitted with this
dissertation).

Keller, J.B. and Zumino, B. (1961), Phys. Rev. Letters z, 164.

Landau, L.D. and Lifshitz, E. N. (19 59 ), Statistical Physics,

Pergamon Press, London, (chapter 14).

London, F. and H. (1935), Proc. Roy. Soc. (London) A149, 71 and

Physic a _s, 341.

Lynton, E.A. (1 962), Superconductivity, Methuen, London, (chaps. 5 and6).

Mc Lean, W.L. (1960), Thesis, Cambridge University, England.

Pippard, A. B. ( 1947), Proc. Roy. Soc. ( London) A191, 370. 11

Pippard, A. B. (1950), Proc. Roy. Soc. (London) A203, 98 and 210.

Pippard, A. B. ( 196 1) , Proc. VIIth Int. Conf. Low Temp. Phys. , North

Holland Publishing Company, Amsterdam, p. 320.

this dissertation has not taken quite so long to prepare that the
authors could have had access in June 1963 to a pre limina ry draft
of it.
Preston-Tho mas, H. (1949), J. Sci. Instr. 26, 381.

Richards, P.L. (1962), Phys. Rev. 126, 912.

Rowell, J. M. (1963), Phys. Rev. Letters ,11, 200.

Schrieffer , J.R., Scalapino, D.J. and Wilkins, J.W. (1963),

Phys. Rev. Letters ..1Q., 336.

Shapiro, S. (1963), Phys. Re v. Letters ,11, 80.

Shapiro, S., Janus, A.R. and Holly, S. (1964), Revs. Mod. Phys. 36, 223· 1,

Sharvin, Yu.V. and Gantmakher , V.F. (1961), Soviet Phys. JETP 12, 86 6.

Spiewak, M. (1958), Phys. Rev. Letters 1, 136.

Spiewak, M. (1959), Phys. Rev. 113, 1479.

Tapp, J.S. (1932), Can. J. Res.~, 584,

Thouless, D.J. (1960), Ann. Phys. (N. Y.) 10, 553,

Tien, P. K. and Gordon, J.P. (1963), Phys. Rev. 129, 647,