T h is d isse r ta tio n h a s b e e n 65—10,556

m ic r o film e d e x a c tly as r e c e iv e d

PE A R L , Judea, 1 9 3 5 -
VORTEX THEORY OF SUPERCONDUCTIVE
MEMORIES.

P o ly tech n ic Institute of B rooklyn, P h .D ., 1965

P h y s ic s , so lid sta te

University Microfilms, Inc., Ann Arbor, Michigan

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 VORTEX THEORY OF SUPERCONDUCTIVE MEMORIES

DISSERTATION

Submitted in Partial Fulfillment

of the requirements for the

degree of

DOCTOR OF PHILOSOPHY (Electrical Engineering)

at the

POLYTECHNIC INSTITUTE OF BROOKLYN

by

Judea Pearl

JUNE 1965

Approved:

Head of Department

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Approved by the Guidance Committee;

Major; Electrical Engineering

Leonard Strauss
Assoc. Prof. of Electrical Engineering
Chairman of Guidance Committee

Leonard Bergstein
Assoc. Prof. .of Electrical Engineering

Minor: Physics

I | M e i r Menes
Associate Professor Physics

Minor; Mathematics

-1^
[J
Harry Hochstadt
Professor of Mathematics
Head of Department of Mathematics

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Microfilm or other copies of this dissertation are
obtainable from the firm of

University Microfilms
313 N. First Street
Ann Arbor, Michigan

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Judea Pearl was born on September 4, 1936 in Tel-Aviv, Israel. He

received his B.S. degree in Electrical Engineering in 1960 from the

Technion, Haifa, Israel, and his M.S. degree in Electrical Engineering

in 1961 from Newark College of Engineering, Newark, New Jersey. While

at Newark College of Engineering, he instructed mathematics, and was

engaged in medical electronic research at New York University, New York

New York.

In June, 1961, Mr. Pearl joined the staff of RCA Laboratories,

Princeton, New Jersey, where he is presently employed. While on the

Research Training Program, he has done research on computer micromag-

netic memories, microminiature solid state circuits, and low noise

electron guns. In 1962 he joined the Cryoelectric Computer Research

group of RCA Laboratories. He was engaged in the development of super­

conductive parametric devices, for which he was a corecipient of the

RCA Award for Outstanding Achievement in 1963.

In February, 1963, Mr. Pearl received a Graduate Study Award from

RCA Laboratories enabling him to pursue full time studies in physics at

Rutgers University, New Brunswick, New Jersey. In February, 1964 he re

turned to RCA Laboratories, and continued his work towards a doctorate

at Polytechnic Institute of Brooklyn. Mr. Pearl has worked on the re­

search for his dissertation from May, 1964 to June, 1965. The research

work was conducted in R C A Laboratories under the R CA Graduate Study

Program.

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 ACKNOWLEDGEMENT

I thank the management of R CA Laboratories and especially my

supervisors Dr. J. A. Rajchman and L. L. Burns for fostering and en ­

couraging the research leading to this dissertation. I thank L. S.

Cosentino, R. A. Gange, R. S. Mezrich, Dr. J. C. Miller, and R. W.

Nosker of the Cryoelectric Computer Research group for helpful dis­

cussions in connection with the operation of the memory, and for the

use of their experimental results. I thank Drs. W. H. Cherry, G. D.

Cody, J. I. Gittleman, and R. H. Parmenter of the Materials Research

Laboratory for stimulating discussions on the physical aspects of

superconductivity.

I acknowledge the help of Mr. E. K. Annavedder and Dr. R. W.

Klopfenstein in programming the numerical computations on the R C A 601

computer, and gratefully appreciate the typing and aid of Elaine L.

Harrison.

Finally, I express my sincere appreciation to Professors C. A.

Hachemeister, L. Strauss, and L. Bergstein of Polytechnic Institute of

Brooklyn for their helpful suggestions and the many hours of discussions

on the various aspects of this dissertation.

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 AN ABSTRACT

VORTEX THEORY OF SUPERCONDUCTIVE MEMORIES

by

Judea Fearl

Advisers: L. Strauss
L. Bergstein

Submitted in partial fulfillment of the requirements for

the degree of Doctor of Philosophy (Electrical Engineering)

It is the purpose of this study to establish a physical model which

describes storage phenomena in continuous superconducting films. The

model is based on the Ginsburg-Landau theory and emphasizes the concepts

of flux quantization and vortex excitation. With the concepts developed,

the operation of superconductive memories which employ such films can be

formulated in terms of simple rules governed by the principles of vortex

mechanics.

A special Green's function is developed to handle electromagnetic

problems which involve thin films. With its aid, the magnetic properties

of thin superconductive films versus those of bulk superconductors are

discussed and an effective penetration depth for thin films is derived.

Th^ function is applied to the solution of a practical problem, namely

the current distribution in a superconducting film, under a strip line,

which carries a uniform current.

Abrikosov's flux tubes model and the stability of vortex structures

in thin films made of type I superconductors, are discussed. The mechanic

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 of quantized vortices in superconducting films is developed on the basis

of the narrow core approximation. The interaction of a superconducting

vortex with external magnetic fields and with neighboring vortices is

derived.

The basic properties of vortices in thin films are derived and com­

pared to those of vortices in bulk superconductors. It is shown that in

films containing a vortex, the current density in areas of the film far
2
from the vortex center is proportional to 1/r , and in areas of the film

close to the vortex center is proportional to 1/r. It is shown that by

virtue of this behavior the radius of the core, the magnetic flux and

the self energy of vortices in thin films are almost the same as those

of vortices in bulk superconductors, but the pair interaction energy is

very long-ranged and the magnetic moment is infinite. These distinctive

properties of thin film vortices have far reaching consequences on the

transverse magnetization of thin films which is shown to have a second

order transition at H ..
cl
Vortex pinning, which is responsible for stable storage, is discussed

Expressions are derived for the interactions of a vortex with material de­

fects and with diamagnetic bodies close to the storage film.

The transition processes, between states of different vortex con­

figurations, are studied. The applicability of Ginsburg-Douglass formula

for the critical field of superconductive films is examined under the geo­

metrical conditions of the memory cell. The role of film defects in trig­

gering instabilities is studied by linearizing the Ginsburg-Landau equatio

It is shown that many flux switching mechanisms can be regarded as special

cases of a single vortex production process which can be triggered with

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 various degrees of stimulation. A study of the Disturb characteristics

of the memory cell serves to portray the power and simplicity of vortex

me c h a n i c s .

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 ix

TABLE OF CONTENTS

Page

Introduction to Superconductivity and Superconductive Memories . . 1

1. Magnetic Response of Thin Films in the Meissner State . . . . 16

2. Film Response to Drive Line Current . . . . . . ......... . . 25

3. The Excited State of Superconducting Films ............... . 33

4. Relation of Fields Connected with the Fluxoid Concept . . . . 37

5. The Vortex Force, Meaning and Definitions . .............. . . 43

6. Forces in Isolated Vortex Systems . . . . . . . . . . . . . . 46

7. Interaction of Vortices with External Fields . . . . . . . . 52

8. Spatial Distribution of Vortex Current .................... . 59

9. Structure of Vortex Core . . . . . . ............... ... . 70

10. Self-Energy, Interaction Energy and Magnetic Moment of

Superconducting Vortices . ........... . ...............

11. Reversible Magnetization of Superconducting Films in

Perpendicular Fields . . . . . . . . . . . . ......... . . 88

12. Interaction of Vortices with Material Defects . . ......... . 100

13. Vortex Interaction with Diamagnetic Films . . . . . . . . .

14. Critical Current for Uniform Switching of Superconducting

F i l m s ............. ....................................... . . 122

15. Effect of Film Defects on the Superconductive Transition . 140

16. Vortex Creation, Flux Switching and Disturbs in Memory

Cells ................................. . . . . . . . . . . . 155

Appendix I ................................... 186

Appendix I I ...................... 187

Appendix I I I ................................. 188

List of R e f e r e n c e s ............... 190

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 X

TABLE OF FIGURES AND GRAPHS

Page

Fig. 1 Continuous Sheet Memory Cell 10

Fig. 2 Operation of a Continuous Sheet Memory Cell
[After Burns et al (26)] 11
Fig. 3 Continuous Sheet Memory Plane 14
Fig. 4 Current Applied Above Superconducting Film 17
Fig. 5 Wave Number Dependence of Magnetic Permeabilities
for Thin Film and Bulk Superconductors 23

Fig. 6 Strip Carrying Uniform Current Above a Super­

conducting Film 26
Fig. 7 Numerical Solution of Eq. (2.5) 28
Fig. 8 Graphical Representation of Eq. (2.8) 30
Fig. 9 Current Distribution in Superconducting Film Below
Drive Line 32
Fig. 10 Structure of Vortex Line in Bulk Type II
Superconductors 34

Fig. 11 Vortex System Under the Effect of External Magnetic

Field 53
Fig. 12 Schematic Representation of Interactions Between
Vortices and External Field , 56
Fig. 13 Current Distribution of Thin Film Vortex 63
Fig. 14 Structure of Vortex Core for n = 1 73
Fig. 15 Structure of Vortex Core for n = 2 74

Fig. 16 Structure of Vortex Core for n = 3 75

Fig. 17 Structure of Vortex Core for n = 10 76
Fig. 18 k ' Dependence of P, Eq. (9.14) 78
Fig. 19 A Vortex Near an Interface Between Two Superconducting
Regions of Different Characteristics 104
Fig. 20 Vortex Force vs. Distance From Boundary (a/| = 2) 106

Fig. 21 Vortex Cpvered by Perfect Diamagnetic Film 111

Fig. 22 Free Energies of Covered and Uncovered Vortex Pairs
vs. Separation Distance a l 2d^ = 1 119
Eig. 23 Geometry of Continuous Film Memory Cell - 123
Fig. 24 Current Dependence of Order Parameter 129
Fig. 25 I vs. I ['Data by Cosentino and Miller (50)] 138
x y
Fig. 26 Stability of Order Parameter Against Defects
Perturbation 149
Fig. 27 Root Locus of A(q) for Various Values of f and
K 1 (=2A /nT2 |d) ° 153
Fig. 28 Cell Switching by Normal Channels 156
Fig. 29 Successive Steps in the Process of Vortex Pair
Production (Representing a Small Region in the
Bridge Area) 159
Fig. 30 Effect of Drive Pulse Rise Time on Threshold Current
[Data by Cosentino and Miller (50)] 167

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 xi

Page

Fig. 31 Electron Micrographs of Typical Tin Films Used for

Memory Planes (After Mezrich) 169
Fig. 32 Threshold Curves for Coincident Disturb 175
Fig. 33 Threshold Curves for Non-coincident Disturbs 179
Fig. 34 Illustrative Comparison of Vortex Motions Under
Interlaced and Single-Line Disturbs 185

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 1

INTRODUCTION TO SUPERCONDUCTIVITY AND SUPERCONDUCTIVE MEMORIES

Superconductive materials exhibit the extraordinary phenomenon of

losing all trace of electrical resistivity when cooled down below a

critical temperature, T . This phenomenon was discovered in 1911 by

1 2
Kamerlingh Onnes, and was confirmed by more recent experiments, in

which a superconducting ring carrying current was kept below Tc for

about two and a half years without any detectable decay of the current.
3
In 1916 Silsbee found that below T , the superconducting .behavior can

be quenched and normal conductivity restored by the application of an

external magnetic field. This field, H c , is called the critical m a g ­

netic field, and was found to vary approximately as

H c (T) = H c (0) [1 - (T/Tc)2 ] (0-1)

The second basic property of superconductors is the Meissner effect.

4
In 1933, Meissner and Ochsenfeld discovered that a superconductor is a

perfect diamagnet. That is, below T , the magnetic field is completely

excluded from the interior of a massive superconductor. Perfect con­

ductivity alone would prevent a field from penetrating the specimen by

setting up induced surface currents which would just cancel the applied

field. But if a normal conductor were in an external field before it b e ­

came perfectly conducting, the internal flux would be locked in by induced

persistent currents even if the external field were removed. Meissner's

experiment showed that, even if the field were already inside the metal

before cooling through T£ , the flux was actually expelled below T . Thus,

the Meissner effect shows that the superconducting transition is reversible

and that the superconductive state is a thermodynamic f unction’of the vari­

ables T and H.

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 2

The discovery of the Meissner effect enabled Gorter and Casimir^

to develop a thermodynamic treatment of the superconducting phase trans-

2
ition. The exclusion of flux introduces an additional energy 1/2 p H

per unit volume into the Gibbs free energy. Since at H = H it takes
2
an amount 1/2 of magnetic energy for a normal material to become

superconductive, one concludes that the Gibbs free energy (at zero field)

of the superconductive state is lower than that of the normal state, by

that same amount. So, the basic equation of the thermodynamic treatment

is

G (H = 0) - G (H = 0) = 1/2 ii H 2 (T) (0-2)

n. s o c

where G and G are the Gibbs free energies (per unit volume) of the
n s

normal and superconductive states, respectively. Knowing the temperature

dependence of H c , one can use Eq. (0-2) to derive all thermodynamic prop­

erties of superconductors at zero field.

A useful picture of the superconducting state, which also infers on

its electromagnetic properties, was provided by the "two-fluid" model of

Gorter and Casimir. They assumed that below T the metallic electrons
c
could be divided into two distinct groups. A fraction n was assumed
n

to remain "normal", while the remainder n = n - n "condensed" into a

s n

superconducting aggregate. From the parabolic nature of the critical

field (Eq. (0-1)) the temperature dependence of n becomes

s
n = n [1 - (T/T )4 ] (0-3)
s c

A simple electrodynamical description of superconductors was obtained

by F. and H. London,^ treating the condensed electrons as free particles

obeying Newton's acceleration law

5v —
m ^ = e E (0-4)

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 3

In terms of the current density J = e n v, E q . (0-4) can be written

s s

in the form

1 = ^o\2 <°-5)

where

\ Z = 12 2 (0“6>
n s ^o 6

To obtain the Meissner effect, the Londons added a second equation

to (0-5) writing

V * = ' l,\ 2 ® ^°"7 ^

Combined with a Maxwell equation, (0-7) leads to

2 — 2 —
V H - 1/A H =0 (0-8)

whose solution shows that H decays exponentially upon penetrating into

a superconducting specimen. The characteristic length A , over which

Li
magnetic fields can penetrate a superconductor, is called the London pene­

tration depth, and is given by Eq. (0-6) in terms of fundamental constants

of the superconducting change carrier. Empirical values (=s500 X) are

consistently higher than those predicted by Eq. (0-6). The temperature

dependence of A is obtained from E q . (0-3)

Li

Al 2 ( t ) ’= Al 2 (°) - (T/Tc)4 ] (0-9)

7
This temperature behavior of A is in good agreement with empirical data.

In terms of the magnetic vector potential A,(B = y x A) Eq. (0-7)

becomes

Js = ^~2 1 (°'10)
o

where the gauge V ‘ A = 0, A = 0 is chosen and a singly connected

n
surf

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 4

body is assumed. This relation between J g and A is local in the sense

that the supercurrent density at any point depends on the vector poten-
g
tial at the same point. Several experimental facts led Pippard to p r o ­

pose a non-local modification of the London equations with a characteristic

-4
"coherence length" £ ~ 10 cm, considerably larger than the penetration

depth A. The motivation for this modification was the mean-free-path

dependence of A which was found to be similar to that of the skin depth

9
of normal metals. Pippard drew an analogy to Reuter and Sondheimer

treatment of the anomalous skin effect and obtained the relation

J (r) - - --- ----- f

4 5o Ho X L vol
J ,
f ^^ R
6— — d3r' (0-11)

where R = r - r 1, and where

1/1 = l / t 0 + S » ( 0 - 12 )

a is an empirical constant near unity, Z is the mean free path, and g

is a parameter characteristic of the metal empirically found to be

-ft v/f
| = 0 . 1 5 ^ (0-13)

The form of (0-11) shows that the supercurrent is not a local function

of the vector potential, but depends on an average over a volume of

radius ~ |. The solution of this equation is difficult, however, an

effective A can be derived from the non-local theory which can be used

in (0-10) in two limiting cases. The result is

A = Al ^°r ^ ^ ^ (London limit) (0-14)

1/3
A = for £ » A (Pippard limit) (0-15)
0
27t £o A L

Equation (0-14) is applicable when the mean-free-path is shorter than

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 5

the penetration depth, which is the case for thin films, where the mean

free path is limited by the film thickness. Thus, the effective penetra­

tion depth for thin films is larger than A by the factor ^o/

In 1950Giosburg and Landau

10 ^'
(G-L) introduced a phenomenological approach

to superconductivity which is the basis for the present thesis. G-L

introduce an order parameter (or effective wave function) \jr which they
2
normalized so as to make |\[f| = ns • -*-n t*:ie absence of an external field,

the free energy of superconductors can be expanded in power series of

2
|ij;| for temperatures near T :

G g (T) = G n (T) + a(T) |i|2 + ^ - |i|4 (0-16)

The coefficients a and B are found from the conditions that G (T) be mini-
s
2 2
mum when |\|r| = |\{rq | , the zero field order parameter, and that (0-2) be

satisfied. Thus,

-a/p = l i j 2 (0-17)

a 2/p = h o h c2 (0-18)

In the presence of magnetic field G-L added to (0-16) magnetic and kinetic

energy terms, writing

G g (T,H) = G n (T,0) + a |i|2 + p/2 |ij,|4 + h q/2 H 2 + — |-M% - e*A±\2

2m

(0-19)

Where e and m are the electric charge and mass of the superconducting

carriers. Minimizing (0-19) with respect to ± and A leads to the two

G-L equations:
1 -*■ * 2 2
— x (-i'fiv -e A) ± +a ± + P \±\ £ = 0 (0-20)
2m

*2 r •
- V 2 A/m-o = Jg = j -A |jf|2 + ( f v l - l v ±) f (0-21)
m 2e J

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 6

The second equation reduces to London equation (0-10) if the phase of

\[r remains uniform throughout the superconductor. Thus, the Meissner

effect is shown to be related to a long ranged phase coherence. The

term (\jr V & “ it V i ) will be shown in Section 5 to describe the ex­

citation of superconducting vortices which break the superconducting

phase coherence, and so reduce the Meissner effect.

The outstanding contribution of the G-L model arises from its

ability to treat the position and magnetic field dependence of the

order parameter, and so, lends itself to the analysis of critical phe­

nomena. This model, however, is restricted to temperatures near T

c
for two reasons: in the first place because of the power series expan­

sion of (0-16) and secondly because only near T£ is A » (■ , and can

the non-local electromagnetic character of superconductivity be ignored.

In 1957 Bardeen-Cooper and S c h r i e f f e r ^ (BCS) presented a m i cro­

scopic theory of superconductivity which successfully predicted a large

variety of superconductive properties on the basis of only few material

parameters. The theory is based on phonon-induced pairing of electrons

with opposite momentum and opposite spin. The paired electron state is

separated from the normal state by an energy gap 2A. Given the critical

temperature, the density of states at the Fermi level, and the Fermi

velocity, the BCS theory can successfully predict the magnitude and tem­

perature behavior of H c (T), 2A(T), A(T), the specific heat, the thermal

conductivity, and a variety of other phenomena.

12
Gorkov was able to show that the G-L equations can be derived

from his formulation of the BCS theory at T « T^ if one identified the

order parameter, \|r, with a position dependent energy gap, A. The effec­

tive charge, e , in (0-21) should be replaced by the charge of a

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 7

superconducting pair, 2e. This replacement received its experimental

13 14
confirmation by the discovery of flux quantization ’ in units of

0 = h_ = h_ .

o * 2e
e

Recent developments in the field of type II superconductivity fur­

ther verified the ability of the Ginsburg-Landau-Gorkov theory to give

an accurate description of the superconducting phase transition.

15
Abrikosov was able to show that for materials with A > | a stable

state exists (called the mixed state) consisting of a periodic varia­

tion of the order parameter, which can successfully describe the rever-

16
sible and broad magnetic transition of type II superconductors. Tinkham

has noticed t. <_t the microstructure of the Abrikosov's mixed state con­

sisted of a periodic array of singly quantized flux vortices. Saint

17
James and deGennes have shown that the G-L equations predict the ex­

istence of surface superconductivity, namely a surface layer can remain

superconducting up to magnetic fields appreciably higher than those caus­

ing the bulk material to go normal. The width of this surface sheath is

of the order of the coherence distance |, and the magnetic field necessary

to quench its superconductivity is = 1-69 A/| The existence of

such surface superconductivity was confirmed by a large variety of measure­

ments, ^ on both type I and type II superconductors.

Superconductivity with its fast phase transition and inherent low

20
power consumption lends itself to computing devices. In 1956 D. Buck

described a switching element called a "cryotron" which employed niobium

coil to control the resistive state of a tantalum wire. The wire wound

cryotron has a gain which is proportional to the number of turns of the

coil, but its relatively relatively high L/R time constant and large size

make it unattractive as a computer element. The thin film cryotron was

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 later introduced to overcome these two disadvantages, and gain values
22
close to 7 have been reported. A great variety of logical circuits
23
can be constructed employing interconnected cryotrons. If the cryotron

gain is kept at a value greater than unity the change in current of one

cryotron can be used to control another, and so, the design of logic cir­

cuits is similar to that which employs vacuum tubes.

The small size of evaporated cryogenic circuits, and the character­

istic of diamagnetic shielding make the application of superconductivity

especially attractive to large-capacity memories. There are two types of

cryoelectric memories. One consists of the interconnection of two cryo­

trons to form a flip-flop type of storage element, and in the second type

information is stored in the form of a persistent current loop having one

24
polarity or another. In 1958 Crowe described a memory cell which con­

sists of a thin superconducting film with two small holes separated by a

narrow bridge. A drive "wire" in the form of a narrow strip lies_just

above the bridge, separated only by a thin insulating layer. A current

flowing in the drive strip may momentarily quench the superconductivity

of the bridge, and allow magnetic flux to link through the holes. When

the drive current is removed the magnetic flux remains stored in the cell

being supported by two persistent current loops flowing around the holes.

For reading, current is again applied to the drive line and if the sum of

the fields from the drive line and the field of the stored persistent cur

rent is large enough the bridge switches, causing a voltage to be induced

in the sense line located under the bridge. Conversely, if a persistent

current had been stored in the opposite direction then the sum of the

fields would be too small to cause the bridge to switch and no voltage

would be induced in the sense line.

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 9

Memories employing Crowe cells have the difficulty of reproducing

the same cell configuration over large arrays. The problems arise from

the difficulty of aligning the bridge to the drive lines. Also, rela­

tively large sizes are required in order to fabricate the two-hole struc­

ture in a reproducible fashion.

To overcome those difficulties, the continuous film memory (CFM)

25 26
was developed by Burns et al. ’ The structure of this memory cell

is shown in Fig. 1; it resembles that of the Crowe cell except that the

storage medium is a continuous superconducting film rather than an array

of actual holes. This memory is extremely attractive because of the

relative ease by which it can be produced in large numbers, since the

drive lines which cross need not be aligned with any particular loca­

tion at the film surface.

26
Burns et al explained memory operation as follows: The drive

current in Fig. 2a creates the field pattern shown. A current equal

to the drive current is induced in the memory film. As this current

exceeds the critical value, its field breaks through the superconducting

plane as in b and two normal regions separated by a superconducting bridge

are formed as shown in c. The induced current in the memory film must re­

duce to I in the steady state. Wh e n the drive current is removed, there

remains a stored current equal the difference between the amplitude of

the drive current pulse and the critical current, Ic * The field pertain­

ing to this current is shown in d. This operation is analogous to the

cell described by Crowe where physical holes and a bridge actually exist.

In order to explain the formation of the bridge and thereby to

27
analyze the transient response of the cell, Ahrons proposed a different

switching mechanism whereby the image current in the film is forced to

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 STRIPS
LEAD

FILM
T IN

MAGNETIC FIELD LINES

FIG. I. CONTINUOUS S H E E T MEMORY CELL

DR)VE
SENSE

DRIVE
Reproduced with permission of the copyright ow ner. Further reproduction prohibited without p erm is sio n .
 11

A P P L IE D PULSE

DRIVE

v ///////////// a // a

SENSE SUPERCONDUCTOR

DRIVE Ic !

t\_ INDUCED \
s\ ~ ~ CURRENT _ \
-i_ -V-,
C v/
SEN SE _
VOLTAGE V
'(a) (b)

v/////\ y/j /’/ / / / / A X/////A [177-Xfny//A/Jl

(c) (d)

FIG. 2 . OPERATION OF A CONTINUOUS S H E E T MEMORY

C E L L (A F T E R BURNS E T AL (2 6 ))

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 12

flow outside the intersection area and so doing; two circulating current

loops are established in the diagonal corners of the cell. In Ahrons'

model "there is no fixed normal region in the CFM" and it is necessary

to speculate that "the currents circulate about imperfections such as

grain boundaries and impurities."

The main difficulty in constructing a switching model for the CFM

lies in explaining the stability of the flux storage, namely, one should

explain why the two normal regions in Fig. 2d do not grow toward each

other until eventually destroying the superconducting bridge which sup­

ports the current circulation. Another question that should be answered

is whether the entire area under the drive lines intersection becomes

normal during the switching (Fig. 2b), and if so, what causes the forma­

tion of the superconducting bridge in c. The answers to these questions

are essential for the calculation of the physical size of the memory cell,

the magnitude of the sense voltage, the amount of hysteretic losses, the

degree of flux degradation by disturb currents and the amount of inter­

ference between neighboring cells.

The most obvious contribution of the present study is that of trans­

lating various pictures of the CFM operation into a standard language which

uses a single type of excitation of the superconducting state. These ele­

mentary excitations, called vortices, obey very simple mechanical rules

and can represent any pattern of flux and currents in superconductive

films by an appropriate vortex distribution. The switching of magnetic

flux is said to be due to creation or destruction of vortices, and the

expansion of magnetic field permeable areas is described in terms of v or­

tex motion. The electromagnetic problems associated with the geometry of

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 13

CFM cell are complex, and require the solution of three dimensional

boundary value problems. With the aid of vortex mechanics concepts,

those problems are reduced to the simple language of forces and motion.

To complete this introductory chapter, a short description of the

organization of the CFM will be given. Figure 3 describes a single

memory plane containing 4 x 4 cells in an x-y array. The selection of

a single element in this array is done by sending current pulses in one

of the X drive lines and one of the Y drive lines simultaneously. The

level of the drive line current alone should not be sufficient to cause

flux switching, only the combined field of the X and Y pulses is higher

than the critical field and causes switching at the appropriate X-Y

intersection. A zigzag sense line links all the memory cells and the

appearance of a sense signal at its terminals indicates that switching

has occurred at the addressed location. The selection of the appropriate

X line is performed by a network of cryotrons in a "tree" arrangement.

The currents flowing in the "address" lines control the superconductive

state of the cryotrons and so determine the path of current flow. Every

binary combination of the address line currents corresponds to only one

superconductive path which is chosen to carry the drive current to the

selected memory location.

Every memory word corresponds to a specified location of the X-Y

array, and each of its digits is located in a different plane. For e x ­

ample, the single plane of Fig. 3 constitutes a single digit of a 16 word

memory. Fifty digit words.will require a stack of fifty such planes. To

select a single word, the same array location is simultaneously energized

in all the planes of the stack. The voltages appearing on the sense lines

of the various planes represent the various digits of that selected word.

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 FIG. 3 CONTINUOUS SHEET MEMORY PLANE
STORAGE
CELL
ADDRESS
LI NES
X" .DRIVE
Reproduced with permission of the copyright ow ner. Further reproduction prohibited without p erm is sio n .
 15

The analysis of more elaborate schemes of memory organization is

27
given by Ahrons. The present study is concerned with the physical oper­

ation of a single memory cell.

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 16

1. MAGNETIC RESPONSE OF THIN FILMS IN THE MEISSNER STATE

The method of analysis which is repeatedly used throughout this

thesis is based on the Green's function appropriate to thin supercon­

ducting films. Since the basic equations describing the Meissner effect

are linear, the complete magnetic behavior of a film in the Meissner

state is determined once the response to a singular excitation is known.

Consider the situation in which an arbitrary current distribution

J is applied parallel to and above a continuous superconducting film of

thickness d and of infinite extent. The film is assumed to lie in the

x-y plane of a Cartesian coordinates system, as shown in Fig. 4, and has

a penetration depth A. Being infinitely wide (much wider than the dimen­

sions of the energizing circuit) the film not only expels the magnetic

field from its interior, but allows supercurrents to flow on its surface

so as to screen the whole region below. It is the relation between the

applied current J and the induced currents J that determines the mag-
3 S

netic properties of the film. The exact analysis of the geometry in

28
Fig. 4, for any arbitrary d, results in highly complicated expressions,

the use of which is rather limited. Instead, a simpler approximate ap­

proach will be used, which takes advantage of the small thickness of the

film, and which clearly exhibits the distinctive properties of thin super­

conductive films.

For small ratios of d/A, the current density in the film is essen­

tially uniform across the film thickness. This is due to the fact that

London's equations (see Eq. 0-8) forbid rapid variations of currents over

distances shorter than A. Since we shall be mostly interested in the be­

havior of the current over distances larger than d, any variation of J g

in the z direction will not cause a significant contribution. We can

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 17

CURRENT SOURCE

APPLIED CURRENT
SUPERCONDUCTING
\ F IL M

SCREENING
CURRENTS

F IG .4. CURRENT APPLIED ABOVE SUPERCONDUCTING

FILM

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 18

replace the true current distribution in the film by an infinitesimally

thin current sheet

J s = K s ( p ) S ( z) (1 .1 )

where the strength of the surface current K g (Amperes per meter) is the

integral value of the true current density over the film thickness

+d/2

K s (P) = J J s (P,z) dz = d J s (p,0 ) (1.2)

-d / 2

Using this representation of the film current, and assuming that London's

equation for Meissner superconductors holds for the superconducting film,

_ — 2
i.e., that J = - A/A ll , Maxwell-London's equations read
’ s o

V 2 A = -p. J = p (J + K S(z) ) =
o total o a s 1

= -p Ja + d/A 2 A 5(z) (1.3)

where pQ H = curl A. The three dimensional Fourier transform of (3) is

given by

A(q) = P o/q 2 [Ja (q) + K g (q,.) ] (1-4)

where

K s (qt) = " — 1/-2TT j A (q) dq^ (1.5)

A [i J
o
✓
"S /S /S /
N

q = X qx + Y qy + Z qz = qt+ Zqz (1.6)

A (q)= / A(r) e " 1^ ' r dr (1.7)

A A A

X, Y and Z are unit vectors along the Cartesian axes, and cj is the trans­

verse projection of the wave vector q".

The integral equation (1.4) can be ^solved by integrating both sides over

all q •

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 19

A- (q) dqz = M-o \ + K s (qt)] dq; ( 1 . 8)

q

AV
By definition (Eq. (1.5)), the left hand side of (1.8) gives - K-s (q^)

The second term on the fight gives

uu —»» * .

r W - -
dq„ = u K (qj (1.9)
/ " 1 ---- d q z = ^ o K S (qt } 2 “X ° * qfc
-oo
q. + q
^Z

The first term on the right hand side of Eq. (1.8) can be integrated

in the complex plane, taking advantage of the fact that due to the ab­

sence of applied currents in the region z < 0, J (q) is analytic in the

3.

lower half of the complex q z plane. Thus,

J ( qt; q ) Ja(V V
I — 2 ----- 2 ~ dqz = -27ri R e s ‘ 2 2
-OO q„ + qZ z + 9, _ q = - xq
nz t

= —-- J (q ; - iq ) ( 1 . 10)
q t a KHt ’ 4 t'

Substituting Eqs. (1.9) and (1.10) into (1.8), one gets an algebraic

equation for K (q ):
s t

-A2u
^o7r
o 1 ^ (^f.) = --- d (q : - iq. ) d---- K (q.) ( 1 . 11 )
2 7T d s nt q^ a ^t t- q s ^t
>t

Solving for K (q ), one obtains the desired relation between the applied

currents and the film screening currents

da(V ' iqt)

(1 .12)
V qt>
1 + 2 q A /d

The appropriate Green's function is found by computing the film re­

sponse to a unit current element situated at p = 0 z = z',

✓\
J (r) = u 6 (p) 6 (z-z') (1*13)
3

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 20

where u is a unit vector along the direction of the applied current.

(Such a current element cannot be realized since it does not obey the

continuity rule s? • J =0. This should not in any way interfere with
Si

the formal analysis, since it is always understood that in practical

computations all quantities are computed by integrating J over a com-

cl

plete closed loop.) The Fourier transform of (1.13) is given by

J„(q) = u e 1 qz z (1.14)
a

and the film's current is obtained from (1 .1 2 ):

„ -qt Z '
K (q ) = - u — 2 --------t d-15)
S C 1 + 2 q A /d

Thus, the Fourier transform of the diadic Green's function is given by

-qt
G(qJ = - 1 --- 1 - p (1.16)
1 + 2 q A /d

where 1 is the unit diadic.

To get the real space transformation from J ('r) to K (p), we take

3 s

the inverse Fourier transform of both sides of Eq. (1.12) and obtain

-qtzl iqt ’ (P - P 1)
K (P,0) =
(2tt) J J
[f ------ --------5
1 + 2 q A /d
J ("r') dq dr'
a t
(1.17)

The transformation portrayed by Eqs. (1.12), (1.16) and (1.17) repre­

sents a dispersive relation between the applied current and the film

screening currents, namely, the screening currents are not merely the

negative image reflection of the applied currents, but are related to the

latter by some average process. Current components of high wave numbers

induce less screening than current components of low wave-number. A com­

plete screening, i.e. G = -1, is obtained only for zero wave-number, that

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 21

is, for uniform current.

The dispersion between J and K has two sources; a geometrical

d. S

dispersion, and an inherent superconductive dispersion. The geometrical

dispersion arises from the geometrical distance between the applied cur­

rent and the film, and represents the loss of resolution in the current

pattern that occurs due to the spreading of magnetic fields in free

-qtz'
space. This part of the dispersion is represented by the factor e

in Eq. (1.16)j the characteristic wave length being z', the distance be-
-qtz'
tween source and film. The effect of the factor e on the distribu­

tion of magnetic fields corresponds to the ordinary image sources placed

on the other side of the film, and represent therefore a perfect diamag­

netic film, allowing no normal component.of magnetic field to exist at

the symmetry plane z = 0 .

Even when the current sources are placed infinitesimally close to

the film surface, the screening current does not follow the same pattern

as the source currents. This part of the dispersion portrays the inability

of superconducting currents to screen away a rapidly varying excitation.

It is represented by the factor ----------- x— in Eq. (1.16). Its characr

1 + 2 q A /d
2
teristic length is 2A /d, which is larger than A (d/A < 1) - the char­

acteristic length for dispersion in bulk superconductors.

In order to compare the screening characteristics of thin films to

those of bulk superconductors, it is convenient to define the quantity

|i(q), a wave-vector dependent permeability of a superconductor. In ordinary

magnetism, if a test current J is applied within a magnetic material of

relative permeability p, the relation between the induced molecular currents

J and J is given by
S SL

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 22

. i
total J a + sJ ......
V = ~~j = ~~j---- <L *18)
a a

Adapting the same definition of p. to superconducting films, by letting

the test current J get very close to the film surface (z'-~0) and allow-
3.

ing |i to become wave-number dependent, we obtain (using 1 .1 2 )

W + 3 s^t> q t 2}-2/d
t j (q.) , , 9,2 . C1 -19;
a t 1 4- q 2A /d

In bulk superconductor the magnetic vector potential satisfies a

differential equation similar to (1.3) except for the absence of the

factor d S ( z ) . Thus,

V 2 A = -u1 T r i = - i
u- ~
J + l/A2 A (1.20)
o J*. ^
total o a

giving

-1 - - ®
Js(q) J A (q) = " 2 2 -- (1-21>
p A A q + 1
o n

and for the bulk permeability

- ^aC q ) + J s^ A 2 a2

"b(q) = Ja ^ ' = 1 + A 2 q2 (1’22)

Both bulk and thin film permeabilities are plotted in Fig. 5 against

the wave number q. Thin films are shown to be more permeable than bulk

to low q components of fields. The characteristic wave length, below

which the superconductor losses its diamagnetic properties, is given by

2
A for bulk and 2A /d for thin films. The reason for this difference in

screening behavior lies, in the fact that while in bulk the test current

is heavily screened in all directions by the free electrons, in thin film,

distant regions are inductively coupled through free space, and screening

is less efficient. The length

d = 2 A 2/d (1.23)

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 23

.5

FIG.5. WAVE NUMBER DEPENDENCE OF MAGNETIC

PER M E A B ILITIES FOR THIN FILM AND BULK
SUPERCONDUCTORS

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 24

can be thought of as an effective penetration depth of thin films since

it represents the distance over which the film current decays along a

direction transverse to the film surface. Many properties of super­

conductors, which in bulk samples depend on A, will be shown to depend

2
on the quantity a = 2A /d when thin films are involved.

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 25

2. FILM RESPONSE TO DRIVE LINE CURRENT

As a special case of the problem discussed in the previous section,

consider a very thin strip of width 2 b, situated a distance d£ above the

film, and carrying a uniform current, K q , throughout its cross section

(see Fig. 6 ). This geometry approximates the situation which prevails

when a drive current is applied to the lead drive lines before the memory

cell has-'&een switched. The distribution of drive current within the

lead strip is such that all the drive line current flows within a dis­

tance A from the lower surface of the strip. Since the penetration depth

of lead at the temperatures involved is much smaller than the thickness

of the lead strip as well as the thickness of the insulation, it is, to a

good approximation, possible to replace the actual driving current by the

current sheet of Fig. 6 . Due to the curvature of magnetic field lines

near the lead strip edges, the current density at the edges is expected

to be higher than that in the center of the strip. Thus, the uniform cur­

rent model of Fig. 6 is only a first approximation, and strictly speaking

it represents a memory film driven by normal drive lines. This model,

however, can clearly demonstrate the effect of the two sorts of dispersion

previously discussed.

Writing the expression for the applied current

A

J (r) = Y K 6 (z-d.) Ixl < b

a o 1 11
( 2 . 1)
= 0 |x | > b

The Fourier transform of J (7) reads

a .
/n 2 sin q b -iq d

= Y Ko — e 2 (2-2)
x

Substituting expression (2.2) in (1.12), we obtain for the Fourier

transform of the screening current

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 STRIP CARRYING UNIFORM CURRENT

INSULATION

SCREENING CURRENTS SUPERCONDUCTING FILM

FIG. 6. STRIP CARRYING UNIFORM CURRENT ABOVE

A SUPERCONDUCTING FILM

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 The distribution, of the film current in real space is obtained by

the inverse transform of (2.3),

00 . Co

= k f W e x 6\ ~ w I t sK'> cos < v > dq* -

3 -q d
Y K X 1
[sin (x+b)q - sin (x-b)q ] y — ■------ 7 dq
tr x ' q (1 + a q ) x
x x

= - Y K (2.4)

T P • / \ -V X
1 / sin(ux) e
where I(u,v) = — / --- 7 T 7 \ ------dx: (2.5)
’ 7T J x (1 +x)
o

I(u;v) = — i tan ^ u/v + -It ,(v ^u ^E^(-v+iu)-e^v + ^u ^E^(-v-iu)

IT (_ 2i

(2 . 6)

The form (2.6) is not particularly suitable for computation, because of

the lack of suitable tables and subroutines for the exponential integral

function E^(z) of complex arguments. Instead, a differential equation

for I(u,v) in terms of u can be formed, which may be integrated numeri­

cally. The curves in Fig. 7 show the computer solution of (2.5).

The difference between the curves in Fig. 7 and the line I = .5

illustrates the drop in the film's current, from its value right below

the drive line edge, down to zero, at far distances from the drive line.

The spread in the current distribution is caused by the two aforementioned

contributions to the dispersion. The larger the distance between the

drive line and the film the larger the current spread (note curve for

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 28

in

O F E Q .(2 -5 )
o>

oo

SOLUTION
FIG.7. NUMERICAL
in

ro

OJ

m fO c\j

(x+| )x ji
= (A‘n ) i
xp ;
XA-9(xn)uis
' ' 00

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission
 29

dX
v = — = 10, Fig. 7). For drive lines situated in a very close prox­

imity to the film (v — 0 ), the current spread approaches its minimal

value - OL, (note curve for v = .001, Fig. 7) .

It is of practical importance to estimate the current behavior at

distances far from the drive line edge. The rate at which these cur­

rents decay to zero determines the maximum packing density of a continu­

ous sheet memory. The asymptotic behavior of the film currents for large

distances from the edges (u » 1) can be found directly from Eq. (2.5).

For u » 1 the main contribution to the integral comes from small values

of x. The contribution from high values of x cancels due to the rapid

oscillations of the sin(ux) term. The cancellation is further enhanced

_
if d^ is greater than CL in which case the exponential term e will cut

off the integral at relatively low values of x. We can therefore approxi­

mate the integral in Eq. (2.5) by replacing the factor with an exponen-
I't X
—

tial factor e , both having the same behavior at small values of x. The

approximated integral assumes the form

I(u,v) :
00

J o
e-x(v+1>dx (2.7)

which can be evaluated in terms ofelementary functions, giving

I(u,v) = — tan 1 — - (2 .8 )
1 TT V+l '
u » 1

Expression (2.8) is represented by the curves in Fig. 8 , for comparison

with the exact numerical result. Thus, except for distances as small
2 A 2/
as /d from the drive line edges, and for very thin insulations, the

current distribution in the film is given by (see Eq. (2.4))

^ i —
- 1I vj-r.
x+b .11
- v —k
x-b
tan -— -t - tan (2.9)
d ^+d d ^4a

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 30

— in oj

> >

OF EQ. (2 -8 )
oo

REPRESENTATION
in

FIG. 8. GRAPHICAL

O
in ro oo O

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 31

This is also the expression for the surface current that would flow on

the surface of a perfectly diamagnetic film had the drive line been lo­

cated a distance d^ + a away from the film. Expression (2.9) can be

rewritten:

2 2 2
/s K . x + (QH-d..) -b

V * > * - Y F ct“" - T b'-f e r t p > ^ . <2 -10>

s

but since in practice the width of the drive line is much greater than
2
a. + d^, the term (Q! + d^) can be neglected (even at x = b), and we obtain

K 2 2
K_(x) = -Y f c t n 1 2-* ^ (2.11)

QH-d
Expression (2.10) is plotted in Fig. 9, for various values of — -— .

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 32

o>

FIG. 9. CURRENT DISTRIBUTION IN SUPERCONDUCTING

00

FILM BELOW DRIVE LIN E

(0

ID

Xl-Q

ro

OJ

J22L

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 33

3. THE EXCITED STATE OF SUPERCONDUCTING FILMS

In the previous sections we have been dealing with the response of

a superconducting film which is in the Meissner state, that is, a film

— 2 —
which obeys London's Equation, curl J = - 1/A H, everywhere throughout

its surface. If the film consists of normal regions, this equation can­

not be satisfied everywhere since in normal areas the density of super­

conducting electrons is zero and A becomes infinite. When the drive

current exceeds a certain threshold value, the film loses its supercon­

ductivity in some regions, letting magnetic flux penetrate its surface.

Such a film is no longer said to be in the Meissner state, but in some

excited state with non-zero "Fluxoid". In the sections to follow we shall

be concerned with the properties of this excitation which constitutes the

basis for memory action.

It was shown by A b r i k o s o v , ^ that when a bulk type II superconductor

is excited by flux penetration, the magnetic field penetrates the super­

conductor in the form of periodic array of flux lines, called vortices,

each carrying one quantum of flux. The structure of flux lines in bulk

material consists of two regions: a core of a radius approximately equal

to the coherence length £, and an electromagnetic region of radius equal

to the London penetration depth A (Fig. 10). In the core region the num­

ber of superconducting electrons is severely modified, it reaches zero

J

at the very center of the core, and approaches its zero field value out­

side the core region. The electromagnetic region is the space where cir­

culating currents and axial magnetic fields exist. To understand the

function of these two regions, and the role they play in determining the

condition for the vortex texture to exist, let us review the process of

flux penetration into an ideal bulk superconductor of the second type.

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 34

CORE
REGION

ELECTROMAGNETIC
REGION

F IG .10. STRUCTURE OF V O R T E X L I N E IN
BULK T YPE H SUPERCONDUCTOR

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 35

In an increasing magnetic field the sample remains a perfect dia-

magnet until a field H is reached. At this value of H the energy

required to excite a single vortex is just equal to the decrease (-p^M-H)

in the free energy of the system caused by the penetration of one flux

unit, and so, from an energy viewpoint it becomes more favorable for the

supe-rconductor to allow many flux lines to penetrate its interior, rather

than stay diamagnetic. The number of lines is then increasing until

limited by the repulsive inter-lines force. In other words, the density

of lines increases until the distance between adjacent lines become com­

parable to the radius of the electromagnetic region; in which case electro­

magnetic repulsion forces tend to keep them apart. As the field further

increases, more and more flux lines enter the material, until at some

upper field cores begin to overlap and superconductivity is

destroyed. It follows directly from this picture that in order to get

a second type behavior (namely H £ > t^ie electromagnetic region

should be greater than the core, or K = > l/^. (K, defined here as

the ratio of the electromagnetic radius to the radius of the core, is an

essential parameter which determines whether the mixed state is stable

or not.) If this condition is not satisfied, then as the field exceeds

H ^ ji, the density of flux lines will rise to the point of cores overlap,

before the flux lines have a chance to repel each other via the inter­

action of their electromagnet!:-, regions. The result is a sharp phase

transition, typical to superconductors of the first kind.

The difficulty in applying such a model to superconducting films

lies in solving the electromagnetic equations with the proper boundary

conditions imposed by the metal-air interfaces. In a previous section

a simple technique was developed for handling magnetic problems involving

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 36

superconducting films in the very thin limit (d < 2 A ) . Using this tech­

nique it is possible to find the distribution of currents induced by ex­

ternal sources, as well as the distribution of currents circulating

around a thin film vortex. The main difference between the structure

of a vortex in bulk material and the one in thin films lies in the extent

of the electromagnetic region. While in bulk this region is of radius A

and decays exponentially away from the center, the one in films will be
2 1\ f
shown to be of radius 2A /d and to decay very slowly 1 — ) away from the

center. The size of the core, on the other hand, is almost unchanged.

Although present superconducting memories employ type I superconductor

as a storage medium, the quantized vortices model, taken from the theory

of type II superconductors, is believed to remain valid for the two main

reasons. The first relies on the reduction of the electronic mean free

path imposed by the film boundaries, which results in a shorter coherence

length (see Eq. 0-12) or higher K material. The second argument is based

on the peculiar structure that a superconducting vortex acquires when im­

bedded in a thin film.

The long range interaction between vortices in thin films makes it

possible to maintain a dilute system of vortices in equilibrium with the

external fields, long before the cores begin to overlap. This suggests

that due to the large ratio of the electromagnetic radius to the core

radius, thin films made of type I superconductors will exhibit a second

type behavior and will prefer the mixed state over the macroscopic inter-

16 29 30
mediate state. Indeed, many experimental observations ’ ’ support

the existence of a vortex structure in films made of type I superconductors.

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 37

4. RELATIONS OF FIELDS CONNECTED WITH THE FLUXOID CONCEPT

31
The term "Fluxoid" was first used by London to name the quantity:

F = Fluxoid = (A + n A J ) d"J (.4.1)

o s

A is the magnetic vector potential (B = curl A), p. is the magnetic per­

meability of free space, A is the London penetration depth, and J g is the

supercurrent. London has shown that the quantity F is zero, unless the

path of integration includes a normal region (a singularity of A), in

which case F is a constant, independent on time and on the contour of

integration. This led London to suggest that F is an adiabatic invariant

variable of a superconductor and should be quantized (using Bohr Sommer-

feld quantitation condition) in units of a universal constant $ = — ,

e<
e being effective charge of the superconducting carriers.

The term "Fluxoid" has since been used to designate other entities,

like the integrand of E q . (4.1), .^nd the excitation of the Abrikosov type.

For the sake of definiteness we shall call the integrand of Eq. (4.1) the

"Fluxoid Field", 0.

$ = A + po A 2 J s , (4.2)

and the localized excitation connected with the fluxoid field will be r e ­

ferred to as "Vortex".

Instead of following London's treatment of the fluxoid, we shall

focus our attention on the Ginsburg-Landau theory since the latter con­

tains the properties of superconducting vortices in a more direct and

general way.

The Ginsburg-Landau expression for the supercurrent is given by

*2 2
2 — e" l l I f _ 2 i(t) ,'c* * 1
-V A / p o = 3g = -----^ — 1 -A [ij/| +~ - (i|r7 \|r - if v i|T) f =

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 38

2 r 3 + 2? ^ Sr <A -3>

2 ^ ^ g

where A is the zero field penetration depth A = ~*T,--- 72— > ^ = M e

e U p
1^-o 1 o

is the normalized Ginsburg-Landau wave function (|i|/| = 1 in zero field),

and m is the effective mass of the superconducting charge carriers.

Thus, one concludes that the fluxoid field is proportional to the gradient

of the wave-function phase:

, 2
p A <t>
t = A + ° , J (4.4)
PI w

The phase S may in general be a multi-valued function of position, but

the requirement on the single-valueness of i|r, restricts S to increase by

an integer multiple of 2w, each time one retraces its path around any

branch of S. Combining this requirement with Eq. (4.4), and assuming

only straight branch lines along the z direction, one gets

curl 0 = < t > E . n . S ( p - ' p . ) Z (4.5)

O 1 1 1

tbi
where n^ is the multiplicity of the i branch, (or its flux quantum num-

til
ber), p is the 2 dim. radius vector, is the location of the i branch
/N

(or vortex), and Z is a unit vector in the z direction. In the course of

calculations one is tempted to set y x = 0 but this is valid only

when S is a single valued potential. Equation (4.5), on the other hand,

contains the multiplicity of S, and is therefore more convenient to work

with than Eq. (4.4). Note that Eq. (4.5) is also valid for position de-
2
pendent |t|r| , while the London equations assume a uniform carrier

concentration.

The general behavior of \|/ in the immediate vicinity of the vortex

center can be inferred from the first Ginsburg-Landau equation

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 As one approaches the vortex axis, the effect of neighboring vortices

on the fluxoid field is negligible and from Eq. (4.5) one obtains

<i> n 1 ^
® n =n
p — 0 T2 ~TT o
P 0 '(4.7)
v '
A

6 being a unit vector in the azimuthal direction around the vortex con­

sidered. To insure finite magnetic fields, A must be finite, and can

be neglected compared with VS which diverges like 1/p. Therefore,

Eq. (4.6) assumes the approximate form

-v2 kl +(£)2 kl =-^2 *2 <4-8>

and the asymptotic behavior of \|/ in the vicinity of the vortex axis

becomes

~ n in 0
^ p =0 C p e (4.9)

We conclude that the order parameter must go to zero at the center of

the vortex, and in this sense the superconductor can be regarded as

multiply connected. The higher the flux quantum number, the slower the
2
rise of |i|r| away from the axis. A detailed analysis of the core struc­

ture is given in a separate section devoted to this subject. The be ­

havior of currents near the axis can be obtained from (4.4):

2 <t>
- ~ C o 2m-1^ ,, .
Jp S 0 ~ 2 2? “P 9 <4'10)

Note that in the vicinity of theaxis, allproperties ofvortices are

common to both bulk and thin films,since none of Eqs.(4.4), (4.5) and

(4.6) is thickness dependent. It is only the distant field behavior

which is significantly different in bulk and thin film vortices.

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 40

In order to find the force acting on the vortices, it is necessary

(and sufficient) to calculate the current distribution in the supercon­

ductor. Writing Maxwell equation for a superconductor in the presence

of an externally applied current J

V x V x A = (J + J ) u = (-A + 0) + u J (4.11)
o a U O a
A
2 —.
we notice two main difficulties. First, |i|/| is a function of J (Eq.

(4.6)) so that Eq. (4.11) is nonlinear. In addition, the dependent

variable A is multiplied by a position dependent (generally unknown)

function, |\|/| , which makes the exact solution of (4.11) formidable.

Both difficulties disappear in the "narrow core approximation". In this

approximation one assumes that the mean distance between neighboring

vortices is large compared with the radius of the cores, |i|r| is assumed

to be unity everywhere except at the vortex axis where it abruptly drops

to zero. This approximation does not take into account the change in the

structure of one core due to currents of neighboring vortices. It also

does not include the disturbance that the cores present to the free flow

of currents. The latter does not introduce a severe error, since from

Eq. (4.9) it follows that ~ 100 percent of the metal is still in a state

of non-zero order-parameter, and is capable of transporting currents.

Within the limits of the narrow core approximation, Eq. (4.11) is

linear and therefore superposition is valid. The total current in the

superconductor can be regarded as a superposition of currents induced by

external sources, plus currents associated with the presence of vortices.

Note that the fluxoid field $ plays the role of an externally applied

current source situated in the interior of the superconductor.

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 41

In an infinite superconductor we get the current distribution sim-

2
ply by setting J\j/1 = 1 in (4.11), solving for A, and computing J =

— t (-A + <t>) . In superconductors of finite dimensions |\j/| should be

,2
|i A
replaced by a function which is unity in the superconductor and zero

outside, to confine the supercurrents to superconducting regions only.

For an infinitely wide film of thickness d, in the very thin limit, E q .

(4.11) becomes

V x V x A = (-A + $) + p. J (4.12)
A

where, like in Eq. (1.1) a current sheet J g = &(z) K g is assumed for the

film current. In Section 1 (Eq. (1.12)) it was shown that the Fourier

solution of the quation

C y 2 u(r) = 6 (2 ) u(r) - v(r) (4.13)

in the plane z = 0 , is given by

v (q ; -iq )
“ <’t> - T T T c i f - <4 '14>

Here qfc is the transverse projection of the wave vector q, C is a constant,

2
and y is the three dimensional Laplacian operator. Applying this result

to Eq. (4.12), we get for the film current

K = K + S K (4.15)
s s . s. v
O i l

(q ; - iq)
Ks ® - — — ac ' A;,16>
o .* *.

% \ ® - h ® <4 - 17>

a = 2A2/d (4.18)

For simplicity the subscript twas dropped from . The fluxoid field

$^(q) is obtained by taking theFourier transform of Eq. (4.5). We obtain,

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 42

/N ^ ^

iq" x <j>^(q) = <t>o n_^ Z e 1C^ (4.19)

Cross multiplying both sides of (4.19) by iq and using

y . $ = y . j__ = y •A = 0 (4.20)

one obtains

. — 2 -i q * P.
$. (q) = n. bo 1 q ^ e 1 (4.21)
q
Equations (4.17) through (4.21) summarize the relations necessary to

describe superconducting films supporting quantized vortices. For the

calculation of interaction forces we also need an expression for the

magnetic vector potential that would exist in the plane z = o when the

film is removed. The result is

l_i

*o<3> = 4 ?a “iq) (4'22)

This can be obtained directly from Maxwell equation

-y 2 A = p. J (4.23)
v o o a j— '

or from (4.15) and (4.18), letting A — °°.

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 43

5. THE VORTEX FORCE, MEANING AND DEFINITIONS

The concept of "force", when applied to objects such as supercon­

ducting vortices, requires a special consideration. In ordinary mechan­

ics, forces can only be measured via their effect on mass particles. The

object responding to the action of the forces (like electrical charges

on currents) must be attached to real matter before a force can be

measured, both statically (balancing it against a known mechanical force

acting on the attached matter) or dynamically (observing the acceleration

of the object and multiplying by the mass of the attached matt e r ) . Dynam­

ical measurements on vortices moving in a superconductor can hardly give

any information on the driving force, since the inertia and viscosity

associated with the moving core are still unknown.

In classical hydrodynamics vortex-lines are material lines, which

means that a set of particles which composes a vortex-line at one instant

will continue to form a vortex line at later instants. Whether this is

also the case for vortices in superconductors is quite unclear. As the

vortex moves slowly through the metal the pattern of currents and super­

electrons density moves along with it,but whether it is the same group of

electrons that constitute the vortex or whether different groups take

turns in the circulating motion at different positions along the vortex

path, remains a question. In fact, this question looses its meaning

when we recall that the vortex is a collective excitation in a dense sea

of indistinguishable non-localized particles. The Ginsburg-Landau \|r-

function, which seems to move along with the vortex, is in no way a true

wave-function of the electrons in the metal, but is related to the latter

by some averaging process. Therefore, there is no reason to assume that

the forces acting on a vortex are the same forces acting on the electrons

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 44

composing that vortex at any given instant.

However, it is possible to define the force acting on a vortex in

a unique way, if we assume the vortex to be pinned to an impurity center.

Being attached in such a way to the crystal lattice one can measure the

force experienced by the vortex, by observing the stress developed in

the metal in the vicinity of the impurity center. This way of introduc­

ing the concept of force, completely avoids the aforementioned difficulty

of relating vortex motion to the motion of real matter.

To include moving vortices in this picture, imagine the vortex core

to be attached to a small probe which is free to move throughout the metal;

say a small magnet of sub-atomic dimensions. To measure the force on the

vortex core connect the imaginary probe to a mechanical spring. The v o r ­

tex can be moved by pulling the probe and dragging the vortex core behind

it. The calibrations on the spring measure the force on the moving vortex.

Up to this point we have specified what we mean by a force acting on

a vortex. We are now in a position to develop the necessary methods for

calculating that quantity. Consider a system of superconductors support­

ing a specified distribution of vortices. The superconductors are under

the influence of external magnetic fields produced by constant current

sources connected to normal conductors. We now perform an infinitesimal

t th
displacement 5 1* in the location of the i vortex, keeping all the others

fixed in their position. In doing so we supply an amount -Tf^ • Sr of

mechanical work, where (by definition) is the force acting on the i t '1

vortex. In keeping the source current constant during the displacement

process, the energy 6^ supplied by the sources can be found by calculating

changes in the potential at the source terminals. By energy conservation,

the sum of mechanical plus electrical energies supplied, is converted

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 45

into an increase 5 F g in the free energy of the system. Thus,

5 W„ - 7. ' 8 r. = 6 F (5.1)
B l l s v '

or
8 (Fs - W B) b
7. - - s r - (5.2)
x 5 r. 5 r.

til
Thus, the force on the i vortex is given by minus the derivative of

the thermodynamic potential G = F - W with respect to the position of

S D

that vortex. A system of vortices will be in equilibrium when the po­

tential G is stationary with respect to a virtual displacement of the

vortices location. Note that F is not merely the internal energy of

s

the system but the Helmholtz free energy U - TS, where S is the entropy,

in this way we include the thermal energy supplied to the system in keep­

ing the temperature constant. The function W„ depends on the form of

j}
coupling that exists between the source circuit and the superconductors.

The coupling is magnetic if there is no direct contact between the super­

conductors and the current source; and it is electromagnetic if galvanic

connections exist between the two. These two cases should be analyzed

separately.

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 46

6 . FORCES IN ISOLATED VORTEX SYSTEMS

If the vortex system is electromagnetically isolated from energy

sources, the force is given by the negative of the gradient of the free

energy with respect to the position of thevortex considered. For the

free energy Fg we take the Ginsburg-Landau expression:

Fs = Fn + J^alil2 + f lil4 + ~ H2 + I-i -HTvk - e* A i|2j dv (6.1)

j '— 2
where \j/ is the non-normalized Ginsburg-Landau wave function, |ijr| mea­

sures the concentration of the superconducting carriers. Equation (6 .1 )

represents the condensation, magnetic, and kinetic energies respectively.

The last term can be written in terms of the magnitude and phase of jr;

Fkinetic = f [* 2 (^ ^ l )2 + l^l2 V S ' e* A ) 2 ] dv (6.2)

^ 2m

Using (4.3) this gives,

( V W ) ^ V +/ (6.3,

The first term describes the increase in free energy due to spatial

variations in the order parameter and gives rise to a surface energy. The
* 2
second term is the classical kinetic energy 1/2 m v . Let us now combine

the magnetic energy term with the kinetic energy term;

^o r*-2 , m --2
fmk -j ( - h + 175^2 j s ; . dv <6-4>
Integrating the first term by parts, gives

FM K = 1/2f (4 + 7*^|2 ) J s dv = 1/2 ' 7 s dv (6'5)

If by displacing the i ^ vortex the shape of all cores remain u n ­

altered then the condensation and surface energy terms remain constant

giving no contribution to the force. The only part of the free energy

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 47

which is severely affected by such a displacement is the magneto-

kinetic part,, given by Eq. (6.5) as the scalar product of the fluxoid

field $ with the supercurrent field J g .

32
The result stated in Eq. (6.5) was obtained by London as a direct

consequence of his equations, but is also valid for superconductors with

position and field dependent order-parameter. The evaluation of the in­

tegral in (6.5) was carried out by London using a single valued but dis­

continuous scalar function to replace the multi-valued continuous function

S in Eq. (4.4). The result is

$
F._. = j 2- 2 n. 1 , (6 .6 )
MK 2 . i k
l
fch
where 1^ is the total current crossing the area bounded by the i vortex

line and the metal boundary. A similar proof, using relation (4.5), is

given in Appendix I. The important result of Eq. (6 .6 ) is the (often

ignored) origin of the familiar "Lorentz Force" equation, which is com­

monly used to explain the magnetic behavior and resistive state of bulk

33 34
type II superconductors. ’

To derive the "Lorentz Force" equation let us now restrict the dis­

cussion to a system of vortices with very narrow cores, in the midst of

a thin superconducting film. Since for such a system superposition is

valid, we can use Eq. (4.15) andrewrite the integral in (6.5) in the

following way:

F__. = 1/2 / $ • J dv = 1/2 Z / <D. • J .dv =

MK J J i sj
J

= 1/2 Z j 0, (q) • K (q) dq =

i,j (27T) J j_______________ _
1 r -o * - - o - iq-(Pr p) _
= 1/2 Z n. n. — — / $ (q) • K (q) e 1 dq (6.7)
i,j 1 J (27rr J

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 48

where the superscript o refers to a unit vortex located at the origin

p = o. The force on vortex k is obtained by differentiating (6.9):

-* „ i f >, 7 o - ic3'(p i p ^
1/2 z . Vj ~77iJ 4 <D-Ks W iq e (5 i k - V dq
k i,j (27r) u

i q - ( p k -p.) iq-(p.-pw )
- 1/2 -- ^ - 7 n I $°' (q)-K "(q) iq Z n. e J -Z n. e dq
(277-r J .1 J i 1

- 1/2 r n / [0 "(q)-K (q)] iq Z n 2i sin q • (p -p ) dq

(2 tt) kJ S j J k J

1C3Pk _ „ -iqP:

(2tt)'
\ $° '(q) e ' z nj V (q)6 J iq dq

A," (q) • K (q)] iq dq ( 6 . 8)

(2tt)
From Eq. (4.21) we get

0 n. - i q 'Pk -
o k (-iq X Z) • K (q) iq e dq
2 a
(27r) 2 q J
0 n, O iqp.
— — x / K (q) X ["q X (q X z) ] e dq-
(2?r) J
0 n, _ _ ^ ic& k -
o k
[Kg (q) X Z] e K dq , (6.9)
(2tr)2

giving

V *0 “k ^ s (pk> X * (6 ’10)

Thus, in an isolated system, the force per unit length acting on a vortex

is proportional and at right angle to the current density that would exist

at the vortex axis if the vortex was absent. This is the famous "Lorentz

Force" result. The name Lorentz was associated with the present result

due to the striking similarity between the form of Eq. (6.10) and the

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 49

Lorentz force

7 = J x B (6 .11)
If we assume the vortex magnetic field is localized at its axis, and has

the strength n<J>Q , then (6.10) follows from (6.11). The similarity between

the two, however, is only incidental. The fact is that the vortex m a g ­

netic field is rather non-localized, in thin films it occupies a radius

of 2A /d. Indeed, if f^ was originated by volume forces of the Lorentz

type, a correction term should have appeared in (6 .1 0 ) to account for pos­

sible variations of K within the active region of the vortex. This is not
s

the case, the force in Eq. (6.10) is solely related to the current at the

very center of the vortex.

A more obvious dissimilarity in the nature of the forces in (6.10)

and (6.11) is manifested in the forces' directions. If by f ^ we desig­

nate the force on object 1 due to object 2, then the ordinary Lorentz force

reads

(6. 12)

but the force described by Eq. (6.10) has exactly the opposite polarity;

if (<t> n^ Z) is the magnetic field of vortex 1, then Eq. (6.10) reads

(6.13)

This disparity demonstrates again the fact that forces on superconducting

vortices do not arise from forces acting on their circulating electrons

but from direct interaction with the center of excitation. Because of

the common use in the literature, we shall continue to call the force in

(6.10) the "Lorentz Force", the similarity between the two being formal

rather than essential.

Another peculiar feature of the force in (6.10) is the absence of

I
.

electromagnetic constants in that expression; writing = — and

e

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 50

K g = N'(-e')V, (6.10) becomes

(6.14)

containing only mechanical quantities. Expression (6.14) reminds us of

the hydrodynamical force that a classical fluid exerts on a vortex line.

The latter is given by

i ^
f = n “ K (Vt - V) x Z, (6.15)
Li

where N is the particles concentration, K is the circulation strength,

V is the velocity of the fluid with respect to some fixed frame of refer­

ence, and V T is the velocity of the vortex line with respect to the same
Li

frame. If the circulation K is quantized in units of h, the force on a

stationary vortex is identical for both superconductors and uncharged

classical fluids. The origin of this force lies in the Bernulli effect;

as the fluid velocity is superimposed on the circular motion of the vortex,

the particles velocity increases on one side of the vortex and decreases

on the opposite side. This difference in speed results in a Bernulli type

pressure-gradient which tends to pull the vortex toward the region of

higher speed. It is tempting to suppose that the motion of superconducting

35 36
vortices can be derived from the same basic principle. ’ If this war.

indeed the case, then the first term in Eq. (6.15), usually referred to
/\
as the Magnus force = N“ K x Z, implies that a superconducting vortex

moving in a (otherwise currentless) superconductor with a velocity V T

Li
should be subject to a force Such a force would give rise to vibra­

tional modes, and a large Hall angle in type II superconductors. There

37 38
is some experimental evidence that no such force exists. Bardeen has

argued that the Magnus force is cancelled by the reaction from the corre­

sponding force on the positive lattice background, so that the total force

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 51

due to motion of the line vanishes. If we accept this argument, we re­

attribute electromagnetic nature to the superconducting vortex, allowing

it to interact differently with the electron fluid than with the positive

background, both moving in the same direction, differing only in the sign

of their electrical charge.

Expression (6.10) for the vortex force is usually taken to be an

absolute truth, regardless of whether the system is isolated or not.

The Ampere's force for example can be computed as a sum of interactions

between current elements regardless of how the system is coupled to energy

sources. Such an absolute standard would be justified if (6.10) could be

derived from direct interaction between an arbitrary current and the vor­

tex core. The derivation of the vortex force, however, is inseparable

from energy considerations, (the current K in (6.10) was specifically

assumed to be produced by other vortices) it should therefore be separately

checked for every energy configuration of the system.

Note that our derivation of the Lorentz force is also valid for films

of finite size, since the only assumption made is that of superposition.

Relation (4.17) connecting the vortex current and the fluxoid field,

which is only valid for films of infinite extent, was not used in arriv­

ing at (6.10). In finite films the relation between K and 0 will depend

on the film geometry, and should be obtained from the solution of (4 .1 1 )

2
with |\|r| = 0 outside the superconductor.

One can generalize the result (6.10) to include films of finite

thickness by regarding" the latter as a superposition of many thin layers,

each carrying an infinitessimally thin current sheet of strength J(p,z)dz.

fch
The total force on the k vortex is then given by
d

J (Pk >z) X Z dz (6.16)

o

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 52

7. INTERACTION OF VORTICES WI T H EXTERNAL FIELDS

In order to find a potential function, whose derivative gives the

force, we need to calculate the amount of energy 5 W supplied by the

current sources, during a virtual displacement of one of the vortices.

Consider (see Fig. 11) an energizing circuit, consisting of a current

source connected to a normal conductor of conductivity cr. The source

maintains a constant current I throughout the circuit, and the current

density distribution in the normal wire, J^, is kept constant in time.

A superconductor supporting vortices is exposed to the magnetic field

of the energizing circuit, but has no galvanic connection to it.

To find the instantaneous voltage of the source we write Maxwell's

equation:

VT ■*
e = - v <p - ^ ; (7-i)

E is the electric field, cp is the electrostatic scalar potential (single

valued), and X is the magnetic vector potential. Multiplying both sides

of (7.1) by J q and integrating over all space excluding the source, we

obtain
a
J E dv = - / J0 • V 9 dv - / J q • A dv (7.2)

v-vB v-vB v-vB

The first term on the right can be integrated by parts, giving

j)
J n ' E dv = - / cp J^ • 3s - ^ j J n • A dv (7.3)
o j ' o dt j o
w -\ s+sB v-vB

The surface integral over S vanishes, since J g vanishes on S. The sur­

face integral over the surface S , containing the source, gives 1(9 -cp)
B B A
where 9 and 9 are the potentials at the source terminals. Writing
D A

E = ~3 /cr for the field in the normal conductor, (7.3) becomes

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 CURRENT
SOURCE

NORMAL
CONDUCTOR

SUPERCONDUCTOR
SUPPORTING VORTICES

FIG.II. VORTEX SYSTEM UNDER THE EFFECT

OF E X T E R N A L MAGNETIC FIELD

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 54

i (<pA - v = f 3o h d v + s! / 3? iv <? -4 >

v-vB v-vB

The first term is unaffected by vortex motion; it only adds a con­

stant to G and can be dropped out. The second integral can be extended

to include all space; the difference involved is of the order of the

source volume, and since J q and A are finite, by shrinking that volume

to zero the contribution from the excluded region vanishes. Thus, the

instantaneous power delivered by the source is given by

d
PB (t) = I V(t) = dv (7.5)

~V

and the total energy supplied by the battery becomes

5 WB = J
C

o
O

P B (t)dt = J J q [A(oo) - A(o) ] dv = 5J Jq • a dv

(7.6)

Integrating the last result by parts, gives

5 w_ = 5 / |i H • H dv (7.7)
B J o o v '

where H q is the field produced by the energizing circuit in the absence

of the superconductor, and H is the existing field.

We have arrived at this familiar result in a rather lengthy but

thorough way, in order to show that the coupling between the vortex sys­

tem and the constant current source is the ordinary magnetic interaction

between current carrying circuits, and all changes which occur in the

superconductor can only affect the source circuit via their magnetic

manifestation.
39
In ordinary magnetism, however, it is well known that when a system

of electric currents undergoes an infinitesimal displacement, then the

amount of energy supplied by the current sources is exactly equal to

twice the increase in the stored energy. It is easy to see that this

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 55

simple rule does not apply to vortex motion. If one assumes 8W = 2 6E

B s
in Eq. (5.2), an erroneous result is obtained.

From (6.4) and (7.7) we can write the required potential function

G - rH K - W B = 1,2f ^
(B - 5 o)2 + ...2m , , .2
e“ |i|f|
Js2] <7 '8>

1 2 ji *
where a constant -z U H was addea for convenience. To express G in
z o o r

terms of integrals over the superconductor only, we integrate (7.8) by

parts, and obtain

G = 1/2 A ( a - A q) • J s + - f — j J s2] dv (? -9>

J e" |jr|

G = I/ 2 0 • J s dv - 1/2 J . A o J g dv (7.10)

Let us now decompose the supercurrent into two parts,

J = J + J (7.11)
s so sv

where J is the current induced in the superconductor in the absence of

so
the vortices, and J is the current due to the vortices alone, in the
’ sv ’
absence of external fields.

G = 1/2 I dv + 1/2 / 0- J dv - 1/2 •J dv - 1/2 / •J dv

J so J sv J o so J o sv

(7.12)

The four terms in (7.12) are illustrated by the schematic in Fig. 12. The

first term represents the magneto-kinetic interaction between the vor­

tices and the induced current. The second term represents the internal

interaction between the vortices, and is identical with the free energy

of an isolated system (Eq. (6.5)). The third term arises from magnetic

interaction between the source and the superconductor diamagnetic cur­

rents, it stays constant and can be dropped out from the calculation.

The last term describes the magnetic interaction between the source and

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 56

E N E R G IZIN G
C IR C U IT

Jo

M A G NETIC M A G N ETIC
IN TE R A C TIO N . IN TER A C TIO N

M A G N ETIC
a K IN E T IC

'Jso IN TE R A C TIO N S JV

M E IS S N E R VORTEX
SUPERCONDUCTOR SYSTEM

FIG.12. SCHEMATIC REPRESENTATION OF

IN T E R A C T IO N S BETWEEN V O R T I­
CES AND E X T E R N A L FIELD

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 57

the vortex currents.

If the superconductor under consideration is a thin film of infi­

nite extent, we can use Eqs. (4.10) through (4.23) to prove that the

first and last terms in (7.12) are equal. We have,

(q) • E K g . (q) dq
i

(27T) 2 /
1

(27T) 2 / d
(q; - iq) ' 2
.
i
0
1
“ (q) dq

so that

(7.13)

This somewhat surprising result states that the direct magnetic in­

teraction between a vortex and a current source is equal to the negative

of the magneto-kinetic interaction between a vortex and the currents in­

duced in the superconductor by that same current source. This is valid

for superconductors of any shape; the general proof is given in Appendix

II.

Using (7.13), the active part of G becomes

(7.14)

and going through the same operations indicated by (6 .1 0 ), we find

The "Lorentz force" expression for the vortex force is also valid for

systems which are magnetically coupled to external sources. To calcu­

late the force between an externally applied field and a thin film vortex,

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 58

we need not go into elaborate calculations of mutual inductances, but

simply calculate the screening currents induced in a vortex-free film,

at the position of the vortex considered.

The reappearance of the "Lorentz force" formula in non-isolated

systems further supports the accepted postulate that (56) is a univer­

sal expression for the vortex force and will also describe the interaction

between a vortex and transport currents. The fact is, however, that when­

ever galvanic connections exist between the superconductor and the current

source, the exact relation, between E(t) and J(t) in the superconductor is

required before one can construct the potential W,.. This situation is

the subject of a separate study and will not be discussed further.

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 59

8 . SPATIAL DISTRIBUTION OF VORTEX CURRENT

The general laws governing the interactions of superconducting vor­

tices were derived by analysis which was confined to the Fourier space.

The details of this interaction as well as other distinctive properties

of vortices in superconducting films are best portrayed by real space

analysis. In this section we shall be concerned with the spatial dis­

tribution of currents which accompany superconducting vortices in thin

films. If the density of vortices is not too high (no cores overlap)

the narrow core approximation can be applied and superposition is valid.

One can first find the current distribution due to a single vortex, and

then apply superposition using (4.15).

Consider a single superconducting vortex, situated in the midst of

infinitely wide thin film of thickness d. Let the origin of a cylindri­

cal coordinate system r,0,z, coincide with the vortex center. The vor­

tex is assumed to carry one quantum of flux and has a vanishingly narrow

core. The finite size of the core will later be taken into account as

the proper radius for cutting off the current singularity. From Eq. (4.5)

the fluxoid field of such a vortex is given by

and the differential equation for the vector potential, Eq. (4.12),

becomes

(8 . 2)
2
OL being 2A /d.

Using the angular symmetry of our geometry, we can write

A =6 f(r,z) (8.3)

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 60

and the equation for f(r,z) becomes

d2f . d 1 3 , „ 2 „ c , N l 0O 5Cz) ..
— T + 3 (r f)= — f 5(2) ^ (8.4)
>2 or r or O' a ir r v '
oz

The Fourier transform of f(r,z) along the z direction must satisfy the

integro-differential equation

- P2 + si r 57 r) 1(r’w = 3F 8(r)■ m r (8-5)

CO

where f (r,z) = ~ J f (r,p) e 1 ^ 2 dp (8 .6 )

and g(r) = / f (r,p) dp (8 .7 )

We now expand f (r,p) in terms of a complete orthonormal set of functions

u (r)
7

= f
f (r,p) = I d
’ 7 F( 7 ,p) u (r) (8 .8 )
d 7
7
where Uy(r) is the eigenfunction set of the operator

Lr = ^ r ^ r <8>9>
2
with eigenvalue -7 . Thus u^(r) satisfies

\ « 7 (r) = - 72 u^(r) (8 .1 0 )

and meets the orthonomality condition

J u7(r)u7(r)
o
00

r dr = 5 (7 -7 ') (8 .1 1 )

Expanding the functions ^ and g(r) in terms of u^(r),

“ = J « (r) h (7 ) d y (8 .12)

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 61

g(r) o j »r (r) C (r) d 7 = J dp J d?| ^ (r) F (7;p ) (8 n )

7 P 7

C(7) = J dP F(7,P) (8.14)

(3

and inserting in (8.5), we get

J
f (-P2 -?2) u (r)
7
F (7 ,P) d 7 =
ira J
/ [C (7 ) - ^ b (L) ] u (r)dy
7
(8.15)

Multiplying both sides of (8.15) by ru^,(r) and integrating over r

(using (8 .1 1 )), gives

_L
F(7,p) = oar ~ 2 ~^— 2 [0O b ^ " ] (8.16)
P + 7

Integrating (8.16) over (3, and using (8.14), we get an algebraic equation

for C (7 )

c (7 ) = £ ~ [0o b (7 ) - C(7)] (8.17)

resulting in

c(7) = b(7) T T ^ c (8>18)

From (8.2), (8.3), (8 .6 ) and (8.7), the vortex current, K g (r), is related

to b(y) by

K„(r) F [«fe b(7) - C (7 ) ] u (r) d 7 (8.19)

o mi j o 7
7

and using the result in (8.18), (8.19) becomes

K (r) = — — f 7- u (r) dy (8.20)

s ^ IT J 1 + 7<i 7
7

Getting back to the definition of u^(r), it is seen that Eqs. (8.10)and

(8 .1 1 ) are satisfied by the function

u (r) = 7^ J x (7 ^) (8 .2 1 )

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 62

J being the Bessel function of order one. b(y) is thus given by the

Hankel transform of —
r

b(7 ) = J 7 2 J^yr) rdf = 7 2 (8 .2 2 )

resulting in the current integral

00

- ir F / r h a ^ Crr) d7 (8.23)
o

The solution of the integral (8.23) can be found in Ref. 40 giving

the final result

K g (r) “ S 2 u" - 7 [S]_ (r/a) - ^ (r/a) - 2/tt] (8.24)

o a
is the Struve*s function of order one, is the Neumann function of

41
order one, both functions are tabulated inJahpke and Emde. Figure 13

shows how the vortex current varies in distance from the vortex center.

The asymptotic behavior of the vortex current for large and small radii

is given by

~ ~ ^ 1
K (r) == 6 — - — for r « a (8.25a)
s |i TT ar
o

- ~ ~ 0 1
K g (r) = 6 — ^ — for r » a (8.25b)
o r

As a check to the result obtained in (8.23) we can take the inverse

transform of Eq. (4.17) and (4.21);

* ■t ? f t ^ s r
( 8 . 26)

Due to angular symmetry we have

K Q (r) = K (x = r, y = o) =
sS sy

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 -3-CS,(x)-N,(xn-

.0 5
.0 4

.03

.02

.01
10 X = r / a
FIG. 13. CURRENT D IS T R IB U T IO N OF T H IN F IL M VORTEX

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 64

lqxr
0 n n - iq e
"2+^ j 2 2' d\ dc^y (8-27)
<2i r)2 *0 J J ( l + a J q x z + <,y z j jq / + V
Transforming (8.27) to cylindrical coordinates, gives
o CO 2 f
. 20 p p . irp cos 6
„
se " (27r)2 f, J J
/ . 1 o

O O
(1
- ip cos 8 e
+ap> p pdpde
, ,„

4> “

= — / •? ~ / cos 6 sin (rp cos G) d6

TT J 1 + Ctp W .I
o
O 0

00

= s r I J i (rp) (8-28)
o

identically with Eq. (8.23).

Expression (8.24) is not too convenient to work with if closed form

solutions are required. The actual current distribution can be approxi­

mated by a simpler expression having the same asymptotic behavior indi­

cated in (8.25) and (8.26). The simplest function to satisfy these

asymptotic condition is

K s (r) = iTV
o
7 (5 V rr <8-2!»

which is plotted in Fig. 13 for comparison with the actual current dis­

tribution (8.23). The deviation between the two is seen to be very small

and assumes its maximum value of about 10 percent near the transition

region at r « a.

The distribution of vortex currents as given by Eqs. (8.24) and

(8.29) portrays the basic differences between bulk vortices and vortices

in thin films. The ^ behavior of currents at short distances from the

vortex axis (r < Q!) agrees with. A b r i k o s o v ' s ^ solution for bulk vortices.
42 43
According to the theory of Onsager and Feynman, the superfluid velocity

V in helium XI diminishes in the same way, with increasing distance from

s

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 65

the center of the vortex filament. In superconducting vortices, how­

ever, the ^ behavior of currents is only true in the region | < r < a.

At shorter distances from the center, the density of superconducting

electrons is reduced and the current is correspondingly modified. F o l ­

lowing the analysis of Eqs. (4.9) and (7.10), the current in the core

of a unit quantum vortex is linear with increasing distance from the

center. The particle velocity, on the other hand, still obeys the —
r
law. Thus the relation

vs = mh 2^ <8-30>

is common to both superfluid and superconducting vortices.

The main differences between vortices in bulk and thin film super­

conductors lies in the current behavior at large distances from the v o r ­

tex center. Instead of exponentially decreasing currents which is typical

15
to vortex lines in bulk superconductors, we now find a slow decay of

currents following a law, So, those properties of superconducting

r
vortices which depend on the nature of the core are expected to remain

the same for bulk and thin films, but large differences are anticipated

in those properties which depend on the far field behavior.

It was already noted in Section 1 that the basic differences in the

magnetic behavior of thin films and bulk superconductors stem from sur­

face phenomena; the presence of free space in both sides of the film

reduces the effectiveness of the film as a shield and enhances mutual

magnetic coupling between different parts of the film area. The differ­

ences in the current distributions of thin film and bulk vortices are of

the same nature. One would expect therefore to find a change in vortex

properties along a vortex line which is supported by a thick film. Those

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 66

parts of the vortex which are close to the metal-air interfaces will

exhibit thin film properties, while the parts embedded inside the metal

will perform like vortices in bulk. To demonstrate this continuous

change of behavior, we shall now calculate^the fields of a vortex line

emerging from a semi-infinite superconductor.

Let the superconductor fill the half space z > o, and a single

quantum vortex be located along the axis r = o. The fluxoid field of

such a vortex is given by

<t> /s 1
3 = ^ e 7 (8.31)

and the vector potential should satisfy

<t> ~
V i V j J + ^ I ' r 2- ^ for z > o (8.32)
A2 2 w o A2 r

V x ^7x 1 = 0 for z < o

Using the angular symmetry of the problem, we can write

1 = 6 f(r,z) (8.34)

and denoting the regionsz < o and z > o b y the 1 and 2 respectively, we have

— i + L = 0 (8.35)
o z

^ 2f2 1 <t> 1
7 2 + Lr f2 " ~2 f2 = " 2 7 (8.36)
o z A 27tA

where Lr is defined by (8.9). The function f^ is composed of a homo­

geneous part f °, and an inhomogeneous part , which is any particular

solution of (8.36) and can be a funttion of r only. Expanding all f's

in terms of their Hankel components (Eqs. (8.10) and (8.21)), we have

00

fl = / 7* d-L^r) e7Z a 1 (7) dy (8.37)

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 67

I 2 . ., 2 '
-I y + 1 / z
f2 = / 7 J L (7r) e a 2 (7 ) dy (8.38)

f2 1 = / 7^ J 1 (7r) a 3 (7) ^7 (8.39)

Using relations (8.12), (8.21) and (8.22) we have the expansion

00

“ = / Jj_ (7^) d 7 (8.40)

from which we find that in order for to satisfy (8.36), a^( 7 ) must

be
<t> -%
ao (7) = 2 7 ^ o (8.41)
2-n-A 7 + 1/A

Combining its two parts, we find for f2

f2 = f2 X + f2°

r *0 J i (7r) F h , - J 72 + i/a2’ 2
/ —::-----------2 dd-v
1■■'2 ~2------ -4- /
7 + /
"V
7
T f'vr'4 p V
J i^7 r ^ e a 2 (7) 87 (8.42)
^ 2ttA 7 + 1/A'
o o

The boundary conditions for f (continuity and continuity of its normal

derivative at the interface z = o) result in two algebraic equations for

a2 (7 ) and (7)

+ 7 2 ao (7 ) = 7 2 a i (7) (8.43)
9 2
72 j. 1 /-x2 2 1
2tt + 1 /A

- \|a 2 + 1/A2 a2 (7) = a1 (7) (8.44)

yielding

7 2 I ? "" 2* (8*45)
2ttA 7 + 1/A 7 + p ~ + 1/A

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 68

and

<t> J^r)
- \jy2 + 1/A2’

f2 (r >z) = ,2 ./ , 2 . W .2X I 2
-^r~
2
J dy (8.46)
7 + p + 1 /A

The current distribution in the superconductor is found from (4.2),

(8.31) and (8.34)

<t>
(8.47)
“ n 2 |_2 tt r ' f2 (r’z)
o A L

or.

8 0
\ly2 + 1/A2
J(r,z) =
r Ji (?r) 7 2 A 2 + 2-2- d7 (8.48)
2 2 2
p. A s i r 1 + 7 A
o o
7 + ~Jy2 + 1/A2 '

Although the evaluation of the integral in (8.48) for arbitrary z

is hard, it is manageable in the two regions of interest; at the surface

z = o, and deep inside the metal, z — oo. For the current distribution

on the surface we obtain

0 C/r)
o 1 1 _ 1 yb 2 + i/a2
J(r,o) =
r 2 I 2 2 T 2 ---- —91
2 w o A2 A o (7 + 1/A ) (7 + V7 + 1/A )

-2 r r- - A
-i K (8.49)
2T7T-1- 1 1V2A
o A

I. is the modified Bessel function of the first kind and of order one,
j.

K is the zero order modified Bessel function of the second kind. At

o
the other extreme limit, z — 00, we have

" 2 d>_ .
J(r,oo) =
_2L 3 K x (r/A) (8.50)
2ir\i Y~2 J i(Ar) dA = 27711
o A

To study the difference in vortex behavior in the regions of inter­

est we compare the asymptotic behavior of the vortex current at very

large distances from the core. From (8.49) and (8.50) we obtain

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 69

J(r,o) - 27J.A 2 (8.51)

r » A o r

"r/A
1
J(r,») = ---- 2 j. -o — (8-52)
r » A 2(io (2TrA)2 r r2

Thus, as the vortex line emerges from the bulk toward free space, the

electromagnetic region spreads like a mushroom; the current density at

2
the metal-air interface follows the 1 /r law (characteristic to thin film

vortices) while deep inside the metal it falls off exponentially. The

range of this transition is of the order A away from the surface. Since

the range of the inter-vortices forces depend on the current distribution

range (see Eq. (6.10) ) we conclude that Abrikosov's flux lines are

loosely coupled inside the metal but strongly repel each other at their

ends, forming a sort of "surface compression layer." In thin films of

thickness larger than the penetration depth we expect the vortex lines

to exhibit thin film characteristics a distance A away from both surfaces,

and bulk properties at the center of the film. The effect of these proper­

ties on the macroscopic behavior of the film will be discussed in Section

11.

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 70

9. STRUCTURE OF VORTEX CORE

For the derivation of the vortex current we have assumed that the

order parameter is constant throughout the film. The solution we have

obtained indicates a — increase of current for decreasing distance from

the vortex center. Such a pattern of current cannot be supported by a

superconductor because at some distance the current density will exceed

the critical current density of the superconductor and will cause a cer­

tain region of the film to turn normal. The transition does not occur

abruptly but involves a continuous adjustment of the electron density;

the increase of currents toward the vortex center causes a reduction in

the density of the superconducting electrons, this results in a corre­

sponding current decrease, and equilibrium is reached. The relation

governing the effect of currents on the superelectrons density is given

by the first Ginsburg-Landau equation (4.6). Since solution requires

the knowledge of the current density, as a first iteration, we substitute

expression (8.24), the current density for a vortex in films of uniform

order parameter. Using (8.2) and (8.24) we have for•a single vortex of

multiplicity n

(9-1)
o o

and the differential equation for the radial distribution of the order

parameter becomes

(9.2)

«i»nil0
where f(r) is the absolute value of t|t (\|r = f(r) e ). It is convenient

to introduce a dimensionless variable

r
x (9.3)
a

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 71

in terras of which (9 .2 ) assumes the form

- 4 - x T - “ n 2 K 2 (x) f + k '2 (f f3 ) = 0 (9.4)

x dx dx 1

where

tr
Kj(x) - | S 1 (x) - N x (x) 2 (9.5)

and

(9.6)

15
Equation (9.4) is similar to the one appropriate for vortices in bulk,
2A
except for the factor K 1 = replacing the ordinary K = A/-/?! per-
~ 7 m
2A 2
tinent to bulk. Thus, we can regard as the effective K of thin

film vortices. The thinner the film the higher its effective K , and the

higher the stability of its mixed state texture.

The boundary conditions imposed on f are

f (oo) = 1 (9.7)

f (o) = 0 (9.8)

These conditions uniquely determine the solution of (9.4). In general,

however, the solution can be obtained only by numerical integration, and

this must be performed separately for each value of n and /<’ . The diffi­

culty faced by the numerical analysis arises from the fact that the con­

ditions (9.7) and (9.8) are given at two different points of x. Instead

of a direct integration of (9.4), a trial and error method should be used,.

The value of the first non-vanishing derivative of f at x = o is assumed,

and (9.4) is then integrated from x = o up. If the assumed value is dif­

ferent from its true value, f will diverge for increasing x. Only for

one particular value of the first non-vanishing derivative will the inte­

grated solution approach unity as x goes to infinity. A systematic search

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 72

for the optimum starting at x = o was programmed on an RCA 601 computer.

The results of this computation are shown in Figs. 14, 15,16 and17,

representing n = 1, 2, 3, 10 respectively. As is expected, the radius

of the core increases with increasing n and decreasing k '.

The behavior of superconductors at very high fields is known to de­

pend very critically on whether K is greater or smaller th ^ There

is no such critical change in the shape of the core with decreasing k '.

On the contrary, the vortex excitation is a proper solution of the

Ginsburg-Landau equation for all value of K 1. However, in order to

make sure that it is a stable solution, one has to compare the energy

associated with vortex excitation to the energy involved with other

mechanisms causing destruction of superconductivity like current quench­

ing. This is done in Section 11 after the vortex self-energy is computed.

The most significant feature which is manifested in Figs. 14 - 17

is the fact that unlike the size of the electromagnetic region which for

thin film vortices is inversely proportional to film thickness, the size

of the core remains close to its value in bulk superconductors. To demon­

strate this feature more clearly it is desired to obtain an approximated

close form relation between the core size and k ' . Such a relation can be

obtained by regarding E q . (9.4) as an Euler-Lagrange variational equation,

minimizing some functional of f. The pertinent functional is readily

found to be
00

(1 - f2) dx (9.9)
o

subject to the restrictions that f satisfies che boundary conditions of

(9.7) and (9.8). W e now restrict f to a single-parameter class of func­

tions which beside satisfying (9.7) and (9.8) also conforms with the

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 73

ID

FOR n =
CORE
FI 6.14. STRUCTURE OF VORTEX

V N*
V CM \
/c'slOO

<+—

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 74

in

ro

<\i
OJ
it
c

FOR
CORE
o>

OF VORTEX
oo

STRUCTURE
<0

in

FIG. 15.
ro
CM

<vl

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 75

FOR
CORE
OF VORTEX
STRUCTURE
in
FIG. 16.

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 n = IO
FOR
CORE
OF VORTEX
r /a
ID

STRUCTURE
FIG. 17.

in

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 77

asymptotic behavior of f. For a single quantum vortex f was shown

(Eq. 4.9) to increase linearly with x near the origin. We therefore

choose the class of functions

f = x/P for 0 < X < P

(9.10)
=1 for P < x

P being the variable parameter to be determined by the minimization

condition

§ = 0 (9.11)

Inserting (9.10) and the approximate form of

K 1 (x) = — r
1 x(l + x)

into (9.9), we obtain

I(P ) = f ^ P 2 + l { ( ! + h ) l o8e (1 + P) - loSe P - f } + 4 <9' 13)

By (9.11) P must satisfy

.2 12
loge (1 + P) - 2 (9.14)

and this is plotted in Fig. 18. The parameter P, representing the extent

of the core region, can be expressed in terms of an effective core radius

r
c

P = rja (9.15)

In order for the calculation of the vortex current to be valid in the

region r < QL, P should be less than unity. Figure 18 shows that as «'

exceeds 2.32 P becomes less than unity, namely the effective radius of

the core becomes smaller than the effective radius of the electromagnetic

region. For that value of K *, Fig. 14 indicates that at r = a f exceeds

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 78

OF P EQ. ( 9 1 4 )
FIG. 18. k * DEPENDENCE

O
CL

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 79

already 90 percent of its asymptotic value at infinity.

The dependence of the effective core radius on the film thickness

can be extracted from (9.14) which for P « 1gives

rc = (6 « ) 1 / 3 ) 7 = (12a ) 1 / 3 S 2 / 3 (9.16)

thus, the core radius has a 2/3 power dependence on £ and only 1/3 power

-1/3
dependence on a. It appears as though the core radius depends on d ,

but if we also include the reduction of £ with decreasing film thickness

(which occurs whenever d is smaller than the intrinsic coherence length)

taking for simplicity

i = d (9.17)

Eq. (9.16) then reads

rc = (24A2d ) 1/3 (9.18)

Thus, the thinner the film, the smaller the core radius, and the better

the narrow core approximation

It is of interest to examine the maximum value of the vortex current

density. In the region r 'a the superfluid velocity is given by (8.30),

the supercurrent will be determined by the product

i, i2 * c2\, j2 e"h 1 " c l . ,2

J = m e V = f h M — o— = --- 7 — f (9.19)
s 1 1 s 1^ o 1 “ 27rr u Tfd ra
m 'o

Taking the approximated form of Eq. (9.10) to represent f, J g becomes

<t>
J = — (— J for r < r < a (9.20)
s u. Trd ra V r J c y J
o N c

J = j — for r < r < a (9.21)

s [iird ra c v '
o

Thus, the current reaches its maximum value at the edge of the core region,

this value being

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 80

J (9.22)
s max (i 7Td r Q!
o c

Inserting expression (9.16) for r^, and using the relation

<t> = % | A H p. (9.23)
O C O

Eq. (9.22) becomes

1/3
H
c
J (9.24)
s max A

Recalling that for bulk superconductors the critical current density is

given by

H
(9.25)
J cb = IT

We conclude that the vortex adjusts the size of its core in such a way

that the maximum current density remains of the order of the critical

current of the superconductor. For practical values of K *, the co­

efficient of H c/A in Eq. (9.24) is of order u n ity , so

J = 0(1) J , (9.26
s max cb

where 0 (1 ) is a factor of order unity.

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 81

10. SELF-ENERGY, INTERACTION ENERGY AND MAGNETIC MOMENT OF SUPERCON­

DUCTING VORTICES

Knowing the current distribution characterizing a vortex in thin

superconducting films, we are now in a position to calculate the other

basic properties of the vortex. The most important properties for under­

standing the macroscopic behavior of an assembly of many vortices are

the self-energy of a vortex, the pair interaction energy and the vortex

magnetic moment.

The vortex self-energy (or excitation energy) is the energy required

to excite a single vortex in a current-less superconductor. It is given

by the difference in the free energy of a superconductor supporting a

single vortex and a superconductor in the Meissner state. The Ginsburg-

Landau expression for the free energy of a superconductor, Eq. (6.1),

can be rewritten in terms of the normalized order parameter

oo.i)
It'2 -“ /P

in the form
1 2 i 2 p a /
-
*

F = F - t ^ H V + ^ k i H / (f^-l)Z dv +
s n 2 "o c s 2 o c ,/ ' '

+ ? h, Hc 2 / 4?2 ( V £ >2 dv + PM k ’ (1°-2)

F n is the free energy of a normal material in zero field, | is the co­

herence length defined by

5 u r n . <10-3>
e po c J

F^ is the magneto-kinetic part of the free energy as defined in (6 .5 ),

and H c is the thermodynamic critical field. Beside the magneto-kinetic

energy, the other contributions to the vortex self-energy are the con-
2
densation energy, arising from the fraction 1 -f of electrons which are

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 82

not paired, and the surface energy coming from spatial fluctuations

in f.

Comparing Eq. (10.2) and (9.4) and assuming that form (9.10)holds

for the distribution of the order parameter, we find that the vortex

self-energy U is given by

4> 2d 2 2
u =— ^ h I(P) = 8mi H df I(P) (10.4)
H0AZ (2tr)
r
where I(P) is given by Eq. (9.13). For small ratios of P = ~ , I(P)

assumes the form

2 1 1 f 1 1 \
I(P>= fj- P + 4 + 2 ( l0S p - 2 J <10'5>

representing the contributions from condensation, surface, and magneto-

kinetic energies respectively. Thus the surface energy contribution is

cancelled by the constant part of the magneto-kinetic term, and the ex­

pression for the vortex self-energy becomes

2
2 2
U = 2tt M- H d 1/12 -y- + 2 In a / r c ^ (1 0 .6 )
o c
3

As the radius of the vortex core decreases, the contribution from

the magneto-kinetic energy dominates, and the self-energy diverges

logarithmically.

This energy divergence in the narrow core limit stems from the fact

that the — behavior of current density near the core results in a loga­

rithmically diverging total current. Only the finite size of the core

which cuts off the current density results in a finite total current, and

a finite self-energy.

Due to the fact that r is not proportional to | , the logarithmic

term in (10.6) does not become the leading one even for very high K

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 83

materials. This becomes apparent as we substitute expression (9.10)

into (1 0 .6 ), obtaining

U = 27T H H 2 d|2 (10.7)

o c

or,

U = 2t M.q H ^ 2 df;2 ^a/rc + 2 logg a /r^ (10.8)

The presence of the linear term in (10.8) is in slight contrast to

the expression of the self-energy of vortex in bulk superconductors where

15
the logarithmic term dominates for large value of A/|. For the films

presently employed in superconducting memories, the value of Ql/f; varies

from 3 to 10 and correspondingly, a,/r£ varies from 1.14 to 2.52. Within

this range, the contribution from the linear term in (1 0 .8 ) is always

higher than the one from the logarithm term, the ratio between the two,

however, never exceeds the value of 4.5.

The pair interaction energy is defined as the amount of energy neces­

sary to bring a vortex from infinity to the vicinity of a second vortex

of the same polarity. If the vortices are not too close together the

order parameter of one vortex is not affected by the field of the other;

one can therefore neglect the contribution from the condensation and sur­

face energies, and consider only the magneto-kinetic interaction in the

narrow core limit. The force ? ^ that vortex 2 exerts on vortex 1 is

given from Eq. (6.10)

^12 = ^o11! Ks 2^1^ X Z (1°-9)

where n^ is the multiplicity of vortex 1 and K ^ C p p is the current pro­

duced by vortex 2 at the location of vortex 1. The amount of work necessary

to bring vortex 1 from infinity to a distance r ^ from vortex 2 is obtained

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 84

by integrating the force in (10.9), giving

U 12 = <!>on lI21 (10.10)

where is that fraction of the current produced by vortex 2 which em­

braces vortex 1. The sign of is positive if both vortices have the

same polarity, and negative when the polarities are opposite. The symme­

try property of the interaction energy requires that

n l ^'21 — n2 ^"12 (10.11)

Using the expression (8.24) we have derived for the vortex current

distribution, the interaction energy for a pair of single quantum thin

film vortices situated a distance r.^ apart, assumes the form

^ 2 00

U 12 = 2 jJ- ^2 J [S^r/a) - Nj (r/a) - 2/tt] dr

r 12

= 5 lSo (rl2/a) - N o <r-12/o)l < <1 0 '12>

o

where S and N are the zero order Struve’s and Neumann's functions re-
o o

spectively. The asymptotic behavior of (10.12) for large and small values

of Ty^GL is given by

* 2 i
U 12 = r— for r l2 >> a (10.13)
O 12

<t> 2

= 2ST5 £or r1 2 < < a <10-14>

where y - 0.5772157 is Euler's constant. Thus, at large distances apart,

thin film vortices interact like coulomb particles having a very long

range interaction force. As discussed in Section 1 this interaction oc­

curs mostly in free space. The superconducting film, on the other hand,

rather than screening the interacting fields, acts as a guide enhancing

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 85

the long range propagation of the magneto-kinetic interaction.

A third distinctive feature of vortices in thin films is the high

magnetic moment associated with a single vortex. Following the usual

definition of the magnetic moment

®= \J * X ^ dv (10.15)

we encounter the difficulty of linear divergence of expression (10.15)

at very large distances from the vortex. This becomes apparent as we

examine the behavior of the integrand at the limit r — °°. From (8.26)

^o 1
the vortex current decays like --- — , the integrand of (10.15) be-

♦ , 2» ^ r
o Zjrr o , . j-. .
comes a constant r ------ t: = --- . and the magnetic moment is infinite
V r
if the integral is to be carried from zero to infinity. For films of

finite extent the integral should only be carried out to the film edges,

and the magnetic moment depends linearly on the film dimension.

To estimate the magnitude of the magnetic moment of a single vortex,

assume a circular film of radius R, the vortex being situated at the cir­

cle center. In evaluating expression (10.15) the contribution from current

filaments very near the center is negligible, since the latter are weighted
2
by a factor r . One can approximate J(r) by its asymptotic behavior for

r » a,, Eq. (8.26), obtaining

M = — R (10.16)
^o

The current distribution expressed by Eq. (8.26) is only valid for a film

of infinite extent. For film of finite size an account should be taken

of the effect of the flux return path, causing additional currents to cir­

culate at the film edge. An estimate of the magnitude of this effect

shows that an additional term of the same order of magnitude as (10.16)

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 86

and linearly dependent on R, should be added to M. Equation (10.16) is

to be only regarded as an estimate of M, the exact value may only differ

by a constant factor which is of order unity.

This extremely high magnetic moment accompanying a single vortex

in thin films is not surprising. When a superconducting film in its

Meissner state is placed perpendicular to a uniform magnetic field, it

3
expels flux from a volume comparable to R which is a factor of R/d higher

than the films own volume. A similar situation prevails when a vortex

is stored in the center of such film; the diamagnetic properties of the

film force the flux return path to extend all the way around the film
3
edges. As a result, the entire volume comparable to R is exposed to

the vortex magnetic field, which explains the linear dependence of M on

R instead on d. (The pertinent relation for vortex in bulk superconductor

$
of length d is (M = — ■- d ) .
o

To summarize the results obtained in this section, a comparison is

made in Table 1 between thin film vortices to vortices in bulk. While

the total flux carried by a vortex, its core radius and its excitation

energy are of the same order of magnitude for bulk and thin film vortices,

the pair interaction energies and the vortex magnetic moment are appreci­

ably different for the two. For vortices at large distances we find a

cutoff interaction in bulk and a long range coulomb interaction ( l / r „ )

in thin film. The magnetic moment associated with a single vortex in thin

films is larger than the one in bulk by the factor R/d.

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 87

BULK FILM

^ °d 1 ^ °d 1 . _ v2
CURRENT 77 OTrvg -r - r « 2 X /d
DISTRIBUTION M 02TTX2 r r M o 2 T T \2 r
Ks
*°d e7 /X r -X ' r ''''2 X 2d
2 X 2 At o(2TTX)'/2 r'/z M oTT r* r A

TOTAL FLU X
* 4>Q * < P o

<t>

RADIUS
OF CORE
rc

EXC ITA TIO N <£o2 d p / X2

ENERGY
BTTMo AZ^VSdC '
U

PAIR
4 > Z0 6 , x
INTERACTION
U ij ^ 0i r r jj ri j > > 2 X /d
2 X 2M o (2 F )l/2 ^ r i j 6 r' i> > X

MAGNETIC
MOMENT « d r
Mo Mo
M

TA B LE I BASIC PROPERTIES OFA SINGLE QUANTUM

VORTEX IN BULK AND T H IN FILM SUPER -
C O N D U C T O R S (X > £ )

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 88

11. REVERSIBLE MAGNETIZATION OF SUPERCONDUCTING FILMS IN PERPENDICULAR

FIELDS

Having derived the basic properties of vortices in thin films, we

can now consider their effect on the macroscopic behavior of an assembly

of loosely coupled vortices (no overlapping cores). A simple experiment

for studying such a system is provided by the transverse magnetization

of superconducting films. Consider a circular superconducting film of

radius R placed in a uniform magnetic field H , normal to the film sur­

face. As the field increases from zero, the film stays in the Meissner

state, shielding the magnetic field from a large volume of space. At

some critical value H ^ it becomes energetically favorable for vortices

to be formed at the film edge and penetrate the film. The magnetic

moments of these vortices are in the direction of the applied field,

and thus, the vortex penetration reduces the diamagnetic moment of the

film. In an ideally uniform film the vortices will be free to move along

the film area, responding to the driving forces described in Sections 5,

6, and 7. In such a film the vortices are said to be in equilibrium with

the external field; for every value of H q the total number of excited

vortices as well as their density distribution is so adjusted as to m i n i ­

mize the Gibbs free energy of the system. Being at every instant of

time in equilibrium with the applied field, such a film will yield a re ­

versible magnetization curve; a single curve for both increasing and d e ­

creasing fields, with no trapped flux.

In practice, the preparation of such defect-free films is almost

impossible. Practical films will not exhibit a sharp reduction of m a gne­

tization at H ^ because the excited vortices remain pinned down close to

the film edge, preventing the excitation of new vortices. If the applied

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 89

field exceeds H , and then reduced to zero, the vortices are forbidden
cl

from escaping the film, and trapped flux is observed. A thermodynamic

analysis of the reversible magnetization of thin films is incomplete

without an additional knowledge of the nature of the pinning forces.

It does not represent therefore a real film, except perhaps at very high

temperatures where the pinning forces become vanishingly small. However,

since the forces which drive the vortices to overcome the pinning resis­

tance are determined by the tendency of the system to reach equilibrium,

all theories of irreversible magnetization must start with the analysis

of the equilibrium state.

The simplicity of the geometry used in transverse magnetization ex­

periments is particularly suitable for illustrating the distinctive proper­

ties of thin film vortices and the way these are manifested in the collec­

tive behavior of vortex assemblies. Special attention will be focused on

the difference in behavior of thin films and bulk samples under similar

conditions. In spite of differences in geometries, some of the features

portrayed by the transverse magnetization curve are also common to the

vortex groups which are involved in the memory cell.

We return now to the circular film in a magnetic field perpendicular

to its plane, the Gibbs free energy of the system is given by

G = F - W„ (11 .1 )
s B

where F is the Helmholtz free energy of the superconductor, and W is

s B

the energy supplied by the current source maintaining a constant field H q .

Substituting the appropriate expressions for F (Eq. 10.1) and W (Eq. 7.7),
S ■ o

yields

( 11 . 2 )

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 90

where F go is the free energy of the superconductor at zero temperature

and zero field, and the constant P Q/2 J ' Hq2 dv was added for convenience.

Integrating the last term by parts gives

G = F so + \ % H c2 / [f2 " 1)2 + (21 V f)2 l dv + 2 / ^ ' dv" l / V s dv

(11.3)
For a dilute system of vortices, assuming the wave-functions asso­

ciated with separated vortices do not overlap, the first integral can

be computed by a superposition of the condensation and surface energies

of individual vortices. Assuming further that superposition of currents

is valid, we can write

J = J + E jf . $ = Z $. (11.4)
s so . si 1 v '
1

Where J is the current induced in the film in the absence of vortex

so
t th
excitation, J . is the current accompanying the i vortex, and i runs
SX

over all vortices present in the film. In terms of fields of individual

vortices G assumes the form

G = F + - V H 2 2 / [f.2-l)2 +(2£Vf.)2] dv +
so2 o c . ./ 1 b 1
1 u

+ ~r L / * J . dv + Z 1$. ' J . dv +
2 •J 1 si . J 1 sj
1 u i<j a J

+ Z \ • J dv - ~ / A •J dv- Z / A •J . dv (11.5)
^ IJ 1 so 2J o so ^ 2J o si v '

Using E q . (7.13)

/ $. • J dv = - / A • J . dv (H.6)
J 1 so J o si v '

and combining the appropriate terms, we obtain

G = F + Z U . + Z U.. - Z 3 j . d v - J A ' J dv (11.7)

S° i 1 i<j 1J J 0 si 2J o so

where U. is the self-energy of the i ^ vortex, and U.. is the interaction

1 0 ij

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 91

energy for the i,j pair. The integrals in (11.7) can be simplified by

the choice of gauge

A
o
= (I /2 H
o o
X r ( 11 . 8)

yielding

dv = |i H ■ M. (11.9)
o o 1

so that

( 1 1 . 10)

At any particular value of H q the last term is a constant, inde­

pendent on the vortex configuration. The number of vortices and their

distribution in space are determined by the equilibrium between the three

other terms; the total self-energy of individual vortices, the interaction

energy of all vortex pairs and the magnetic interaction with the applied

field. The fact that only pair interactions enter into the total inter­

action energy is a direct consequence of the superposition principle used

in the derivation of (11.10). It states that the interaction energy for

a single vortex-pair is independent of the presence of other vortices in

the neighborhood. This picture is no longer valid when the size of the

vortex cores is comparable to the distance between the pair, in which

case the presence of a third vortex might distort the current pattern of

the pair members, resulting in a different pair-interaction. For very

high vortex densities (11.10) is no longer valid and summation over higher

order vortex-groups is required to properly represent the interaction

energy.

In arriving at (11.10) no assumptions were made as to the size and

shape of the film; it is valid therefore for all film geometries if we

properly interpret the components U., U . . and M . . Strictly speaking,

1 1 1 1

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 92

all three properties depend upon the exact location of the vortices

with respect to the film boundaries. However, to get useful re­

sults from (11.10) we must replace the summation over individual vortices

by their average properties, being a representative of a large fraction

of the vortex population. The self-energy of a vortex is only slightly

modified by the film boundaries because the main contribution to the in­

tegral in (9.9) comes from areas close to the vortex axis (r < a ) . We

can therefore replace IL by an average energy U, being given by Eq. (10.8).

As for the magnetic moment M^, we have shown previously that its

magnitude is mainly determined by the current at very large distances

from the vortex center. We anticipate therefore, wide differences in

the magnetic moments of different vortices depending on their location

with respect to the film boundaries. Vortices located close to the film

edge will carry a lower magnetic moment than those located deep inside

the film. A rough estimate of for an arbitrary vortex position shows

that Eq. (10.16) still holds if R is replaced by the shortest distance

between the vortex center to the film edge. We can now define an average

magnetic moment M by
4>
M=-^R (11.11)
Ko

where by R we mean the average distance to the edge. For more or less

uniform distribution of vortices, R is of the order of the sample dimension.

Replacing IL and by their average values U and M, Eq. (11.10)

becomes
F —
G = F - ^ H M + N U + f (N) - p N H • M (11.12)
s o ^ o o o o

where N is the total number of vortices excited in the film, and f(N) is

the total interaction energy,

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 93

f(N) = L U.. i,j = 1 . . . N (11.13)

i < j 1J

Strictly speaking f(N) also depends on .the vortex density distribution,

but since we are only interested in the order of magnitude of the quanti­

ties involved, we can use the uniform density as a representative, and

let N be the only dynamical variable affecting G.

Knowing f(N) provides all the information necessary to determine

the magnetization curve; the equilibrium condition

3g
= 0 (11.14)
c>N

uniquely defines N as a function of Hq , and the total magnetization of

the film is given by

M. = M + N M (11.15)
t o

The first critical field, H is the field just sufficient to cause

cl
the penetration of a single vortex. Such a penetration is only possible

when the energy U, necessary to excite a single vortex is compensated by

a decrease, -(i H -M, in the magnetic interaction. This condition is

satisfied when H reaches the value

° 2/3 / o v 2/3-i
U ~ d E
+ 2 loge (11.16)
cl n M c 2R A
o

The rate at which the magnetization increases after vortex penetra­

tion has started can be obtained from (11.14) and (11.12)

U - n H M + = 0 (11-17)
o o dN

or,

H
o
= H
c l u M d N
4
+ - ;= - --p - (11.18)
o

Differentiating both sides of (11.18) with respect to N, gives

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 94

-1
'dH
dN __ c d2 f(N )
= (i M (1 1 .1 9 )
dH dN o , 2
dN

and using (11.15), we get

-1
dM
d f (N) __ c
dMt
d T ' ^ o *
M2
, 2 dH
( 1 1 . 20)
O dN

It is understood that the first term in (11.20) vanishes identically

for H q < H since at this range N = 0 and = Mq . From (11.20) we

learn that the rate of magnetization increase is inversely proportional

to the second derivative of f(N). If f(N) is only a slowly varying

function of N, M. will increase very fast as soon as H exceeds H ,. This

t o cl
is the situation in bulk superconductors; because the pair-interaction is

cut off exponentially, a large number of vortices can penetrate the super­

conductor before they .see each other. f(N) is essentially zero, until

the vortex density increases to the point where the mean distance between

adjacent vortices is A. This plateau in f(N) makes the first term in

(11.20) infinite, which explains the well known magnetization jump at

H = H ..
o cl

The situation in thin film is entirely different; as was shown in

Section 10, the interaction energy has a very long range, each vortex

interacts with all the others, making f(N) depend very strongly on N even

at low densities. The strength of this dependence can be found by re­

placing the summation in (11.13) by an equivalent integral over a con­

tinuous variable. Taking the uniform density as a convenient represen-

N
tative, the vortex density is given by ~, S being the film area. Making
b
the substitution

Z N a (11 .21)
i
s dsi

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 95

f(N) becomes

f(N) = • ds . U(7. - r\ ) ( 11. 22)

J J i

(11.23)

(11.24)

(11.25)

The approximation leading from (11.22) to (11.23) replaces the inter­

action of a vortex j with all the others by its interaction with single

vortex of equivalent strength situated at the center of the film. Equa­

tion (11.24) was obtained using (10.13) for the interaction energy,

assuming that the large majority of vortex pairs still interact like in
2
infinite films. Note that the N dependence of f(N) is always a direct

consequence of replacing the summation by an integral, without any further

assumptions (see (11.22)). Such a replacement is only valid when the

interaction range is large compared with the mean distance between adja­

cent vortices. In thin films, where the range of interaction is infinite,

the integral substitution is permissible already for low values of N. In

bulk superconductors, on the other hand, the interaction range is A, and

2 2
f(N) will attain its N dependence only for N > S/A .
2
From (11.20) the N dependence of f(N) results in a constant magneti­

zation slope, or linear magnetization. Thus, using (11.11), (11.15) and

(11.25), it is seen that the magnetization curve of thin films in the range

Hq « H c 2 is bilinear,

for H < H 1 (11.26)

o cl

= -V H + 0 (R3) H for H > H 1 (11.27)

o o o o cl

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 96

Where V q is the effective volume from which flux is being excluded when

the film is in its Meissner state

dM

'o dH (11-28)
o

For a perfectly diamagnetic film of radius R, V q is known to be

2/7T times a spherical volume of the same radius. In its Meissner state

the film behaves almost like a perfect diamagnet except for a small field

penetration at the film edge, and we can write

V q = 8/3 R 3 (11.29)

3
The second term in (11.27) also involves a volume comparable to R, but

is of opposite sign. For the magnetization curve to rise after H > H -

cl
3
it is necessary that the factor 0(R ) be larger than V . The bilinear

character of the magnetization curve for thin films indicates a second

order transition at the first critical field, accompanied by a jump in

the specific heat but with no latent heat.

To summarize the results of this section, Table 2 compares the r e ­

versible magnetization curves for thin film and bulk superconductors. To

emphasize the differences between the two it is convenient to consider a

bulk sample with length d, that is, instead of comparing two samples of

different volumes and shapes, we consider a single thin film of thickness

d and represent bulk behavior by letting the vortices acquire bulk proper­

ties. Similar results are obtained by comparing thin film magnetization

in perpendicular and parallel field; w hen the field is parallel to the

plane of the film, the vortices have bulk properties which allows a com­

parison of thin film and bulk both having the same volume.

Table 2 indicates a significant difference in the first critical

field. In thin films the first critical field is lower than that of bulk

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 97

BULK F IL M

Hci U d * [/ X2 \ 2/3 . o i 0 q / X2 .2/1

^ Tlon X -t-
Hc MoMHc x L 9V l £ + 08J 2R

s o r^ X
f ( N)= S u ij = ^ 0 (1 )
i < j ^ 0R
° ( - I - ) « < r ‘i<x

d Mf |
GO F IN IT E 0 ( R 3)
dH° I
HCI

d Mt
0 ( R 2d ) 0 ( R3)
dH0
HC2» H o>HC(

-M f -M t
v ^

l \ /A]

t\ I ^
HCIHC HC2 *Ro He. HC HCZ H0

TABLE 2 REVERSIBLE MAGNETIZATION CURVES FOR

BULK AND THIN FILM SUPERCONDUCTORS

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 98

by a factor of the order of d/R. Such a difference in H . can also be

cl

understood on the basis of current quenching at the film edges. However,

the maximum current density at the edge of the film (in the Meissner .state)

44
has been calculated by Marcus and is approximately given by

H A-
J = ^ (R/d)2
max A

The critical field calculated on this basis is proportional to

(d/R)2, and for practical ratios of R/d it is much larger than the field

necessary for vortex excitation. This implies that vortex excitation is

a more favorable destruction of superconductivity than current quenching.

It should be mentioned that the magnetization parameters derived in

this section are only valid for fields below W h e n H q approached

the vortex density increases to the point where the wave functions begin

to overlap and the narrow core approximation is no longer valid. The ex­

tension of this model which allows for the variation of the order parame­

ter in high magnetic fields is given in Abrikosov's theory. In that limit,

where the superconductor has almost lost all diamagnetism, thin films ex-
s
hibit the same behavior as bulk and Abrikosov theory is applicable.

An interesting consequence of the long range vortex interaction is

that the free energy is extremely insensitive to the exact configuration

of the vortex texture. In bulk superconductors, where the inter-vortex

forces are short ranged, the vortex lines form a two dimensional "crystal"

and some properties of the superconductor are sensitive to the exact nature

45
of this crystal. The calculations of Matricon shows that the free energy

is lower for a triangular than for a square lattice; the latter possesses

a negative shear modulus which implies instability.

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 99

The necessary condition for the existence of an ordered lattice

with positive shear modulus is for the interaction energy to come mainly

from near neighbors interaction, since only the latter is sensitive to

the exact configuration of the elements involved. In thin films, where

the range of interaction is infinite, the near neighbors contribute

only a small fraction of the vortex interaction with the overall vortex

system. As a result, the shear modulus vanishes, and the vortices do

not assume a particular lattice order.

Examination of Eq. (11.25) shows that the interaction energy per

unit area of the film is proportional to the sample's linear dimension,

therefore, the total free energy is very sensitive to deviations from

the equilibrium density distribution, which results in a high Young

modulus. A simple way to look at the vortex texture and to determine

its preferred distribution is to regard it as a collection of charged

particles on the surface of very thin conductor. Both systems have

zero shear modulus and high sensitivity to density fluctuations.

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 100

12. INTERACTION OF VORTICES WITH MATERIAL DEFECTS

In this chapter we wish to find the nature of the forces that pin

down the vortices to restricted areas of the metal, opposing the ten­

dency of the Lorentz force to move them along. To represent defects,

imagine a superconductor with characteristics which vary in position.

The critical temperature for example, is known to depend on the orien­

tation of the crystal with respect to its boundary, and varies therefore

from one crystallite to another.

There are two main contributions to the pinning energy; condensa­

tion energy and magneto-kinetic energy. It takes an amount of 1/2 u,

o
2
Hc per unit volume to make a portion of a superconductor normal, there-
2 2
fore, it costs an amount 1/2 (H V -H . V , ) to move a vortex from
ca ca cb cb
region b to region a, V c is the volume of the core. The magneto-kinetic

contribution arises from the redistribution of currents and magnetic

fields caused by the defects.

Consider the magneto-kinetic part of the thermodynamic potential G,

in the presence of external field H q [Eq. (7.8)],

G = y ~ f [ ( H - H q)2 + A2 J o2 ] dv (12.1)
s
2
The only material property that enters into G is A . As a first approxi-
2
mation, we let A vary slightly with position

A2 = \ Z + e(r) , e(r) « A q2 , (12.2)

2 . 2
Aq is the average value of A over the whole sample. The space varia-
2
tion of A will perturb the distribution of currents and fields:

H = H + h J = J + 7 A = A + a (12.3)
u ’ s su s 3 u

the subscript u refers to the value of the fields in a uniform material

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 10 1

2 2
of A = A . The first order change in G is given by
o

AG =
J
f (a + M-0 Aq2 T s) ‘J s dv + y - f
u ^
e(r) J g 2 dv
u
(12.4)

The variations in the fields quantities are related to e(r) by the con­

servation of fluxoids. W e assume here that the fluxoid,'as defined by

Eq. (4.1), is an adiabatic invariance of the system; it stays unchanged,

as the system undergoes a small and slow change of parameters. If we

2
regard the position dependent part of A to arise from such an adiabatic

change of parameters, we conclude that the fluxoid computed in the un-

2 2
perturbed'system (A ) is equal to that of the perturbed system (A*-),

or:

(A + n A2 J )•
' o s
dl = fif/ (A
u
+ u A 2 J )
o o s
• dT (12.5)
° u
C C

for every contour C. Extracting the first order difference between both

sides of (12.5) and equating it to zero, gives

(a + n A 2 j + p. e J ) •dT = 0 (12.6)
o o Js s
u
C

for every C. The integrand of (12.6) must therefore be a.gradient of' a

scalar (single valued) functionX :

a + p. A 2 T + p e J = V X (12.7)
o o Js o s
u

Substituting back in Eq. (12.4), gives for the free energy difference:

G = f
J
(V X " e d s )-d s
u u
dv + ^
^
f e J s 2 dv
u
(12.8)

Using V ‘ J = 0 and the fact that thenormal component of J vanishes

s s
u u
on the superconductor surface, the contribution from the term involving

V X vanishes, and the final result for the pinning energy becomes:

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 102

/ e (r) J
s
u
dv (1 2 .-9 )

The range and magnitude of the pinning forces depend on the vortex cur­

rent distribution in a uniform material. Because of the minus sign

associated with Eq. (12.9), in order to minimize the system energy, a

vortex will adjust its position in such a way that the integral in (12.9)

is maximized. This is achieved when that part of the vortex which carries

the highest current (near the vortex core) is placed in areas of highest

penetration depth. A vortex will tend therefore to move toward regions

of lowest order parameter. The magneto-kinetic pinning forces act in the

same direction as the condensation pinning forces, but have a larger

range of interaction.

Knowing the vortex current distribution in a uniform superconductor,

the problem of finding the pinning force reduces to that of evaluating

the integral of Eq. (12.9). As an example we shall study the pinning

r-
forces which act on a superconducting vortex in a polycrystalline thin

film. We consider a film which is composed of many adjacent regions,

each characterized by a penetration depth A^ which is uniform over the

area of any single region. The spatial variation of e has a step like

2 2 2 2
shape given by = A_^ - A q , where A q is the average value of A over

the whole sample. In order for the linearized theory to be valid, the
2
condition e. « A must be satisfied for all i's.
i o

Consider a superconducting vortex situated in region 2 close to

its boundary with region 1. If the size of crystallites is much greater

than the effective size of the electromagnetic region of the vortex, we

need only consider that part of the interaction which arises from the

boundary between regions 1 and 2. The interaction with other boundaries

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 103

is much weaker because of the larger distance from the vortex center.

Also, if the vortex is near a boundary of low curvature, the latter can

be approximated by a straight line separating the whole film into two

2
regions as in Fig. 19. We can arbitrarily choose A 2 to coincide with
2 —
the average value A , and so e(r) is zero in region 2 and is equal to
o
2 2
Aq - A^ in region 1. With that shape of e(r) we need to evaluate ex­

pression (12.9) as a function of x, which is the distance between the

vortex and the straight boundary. For the current distribution of a

thin film vortex we take expression (8.29)

K (r) = -9 - 2 - ■ - j - ■ - (12.10)
s p o7T r (a + r) '

which differs by only a few percent from the true current function, but

contains allits important features. A further modification of Kg(r ) is

required to take into account thefinite size of the core. A simple way

to cut off the current near the origin is to set K g (r) equal zero for

r < |. Although the Ginsburg-Landau theory predicts a linear increase

in currents at the center of a single quantum vortex (Eq. 4.10), this

result is highly doubtful, since in the core region non-local effects

become most important, and are not included in the theory. A complete

current depletion from the region r < £ gives a better representation

, , 46
of the core.

Substituting (12.10) in (12.9), gives (for |x| < |)

00 00

^ ° - 2d - * 2 / W 7 7 2 - d' ~ ,2,"'.— 1 ,2 ,2:2 (12-u >

^o77" - 0= x ( + y ) (Q + y )

V?
In this section | is used as the effective radius of the core. A better
approximation is given in Eq. (9.16), but since the dimensions of the
core are not well defined, only qualitative results are expected, and
r^ = | is as good an assumption as any.

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 REGION 2

VORTEX
CORE

FIG. 19. A VORTEX NEAR AN INTERFACE BETWEEN

TWO SUPERCONDUCTING REGIONS OF D IF­
F E R E N T CHARACTERISTICS.

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 105

The force acting on the vortex along the x direction is

2 00

f =
x ox
3x " d’
2 2 1
J. 7T
[
~n -oo
(x 2 + y'2) (a + \/x2 + y '2 ' ) 2

-1
(x /a ) -2 cos (a /x ) . ,
(x /a ) ' + 1
i n x ^ (x /a ) - l
o
(12.12)
For Jx| < | the force is given by
2 00

jzl (12.13)
"x d
IXW 2 21 (x + y * ) (a + v x + y )

The force reaches its maximum value at x = and at the center of

the core one has

2
o 1
(12.14)
fx (x - 0) = a/i gf f - H “ 2 log (1 + a n )
dLL 7T2 a 3
'o

In Figure 20 the force f is plotted against x, for

Oi/i = 2. The boundary is the least stable location for the vortex, as

indicated by the peaking o f f at x = (■ . The area under this curve gives

the total energy required to move the vortex from region 1 deep into

region 2. Thus,

h2 00
o r 2irr dr
2 / 2 2
^oT r

£ 0 1 a/i
log (l + a/|) - (12.15)
l + a/g

To estimate e we assume that the film crystallites only differ from

each other by their critical temperature. Hence, if the critical tempera­

tures of regions 1 and 2 are different by an amount A T , e is given by

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 106

1.0 2.0 2 .5 3 .0 3 .5
X/a — ►

FIG.20. VORTEX FORCE vs DISTANCE FROM BOUNDARY

a /£=2

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 107

2
= A (A 2 ) = A f (1 2 .1 6 )
T
c

and using the r e l a t i o n

-1
2 2
A = A
o
(12.17)

we obtain

? , -2 , A T
e = -4A q ( 1 - t ) t - j - 2- (12.18)
c

where t is the reduced temperature t = T/T .

It is interesting to compare the magneto-kinetic energy change of

Eq. (12.15) with the change in condensation energy A p . Q A ( H c2|2 ) .

Relating A, H c and | to the reduced temperature t by the usual expressions

2 2 2 2 2 2 4-1
H = h (1-t ) A = A (1-t ) * = 4? E A H |i (12.19)
C C O O C O
o

and assuming the difference between the two regions to arise from an

equivalent difference in reduced temperature A t , we obtain

££ = 4 log (1 + a/i) - T f i 7r ( 12. 20)

AG
c

The ratio b e t w e e n the m a g n e t o - k i n e t i c pinning energy and the c o n d e n s a ­

tional pinning energ y is of order unity, and is only a slowly varying

function of temperature. In h i g h K m a t e ria ls the m a g n e t o - k i n e t i c part

becomes mor e significa nt than in low K materials.

The temperature b e h a v i o r of Ag is given by

3
AG (t) ~ A t t (12.21)

where A t is the d i f f e r e n c e in reduce d temperatures o n b o t h sides of the

boundary. The increase of the pinni ng energy w i t h temperature is due to

the increase in the e x t e n t of the v o r t e x e l e c t r o m a g n e t i c region. It does

not mean, however, that the resistance to v o r t e x m o t i o n increases wit h

temperature. The latter depends on the m a x i m u m value of the pinning force,

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 1 08

since no motion is possible unless that value is exceeded. Taking the

force at the vortex center for a representative of the maximum force,

and neglecting the logarithmic term (small for high K material) in

(12.14), we get

3 2 ^
T (1-T )
£x (0) • % " (1 2 .2 2 )
(1+t 2)
f (0 ) vanishes at the critical temperature and has a maximum around
X
T
f = -73.
c
The reason for such a maximum lies in our simplified model of two

regions differing only in their critical temperatures. In such a case,

the difference between the penetration depths vanishes as the temperature

approaches zero, and no magneto-kinetic pinning force exists. In real

materials, two crystallites may also differ in their mean free paths,

so that A^(Q)^ A 2 (0 ), and the pinning force may continue to increase as

the temperature goes down.

It is worth noting that the tendency of a vortex to migrate toward

regions of lower order parameter may give rise to an attraction force

between vortices that may exceed their magnetic repulsion. Since the

vortex core has a lower order than the rest of the superconductor, it

33
tends to attract other vortices into it neighborhood, and a "flux bundle"

may be formed. It is possible that a certain minimum number of vortices

is required to trigger such a process, and this determines the minimum

number of vortices in a bundle.

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 109

13. VORTEX INTERACTION WITH DIAMAGNETIC FILMS

In the previous section we have seen how the presence of film

nonuniformities can confine a vortex to stay in restricted areas. This

pinning mechanism is of major importance for the stability of flux

storage in the continuous sheet superconducting memory. In the absence

of any pinning forces, the superconducting vortices which are excited

in the switching process will be free to move along the film surface

and respond to the inter-vortices interaction forces. As was established

in Section 6 , the forces acting between vortices of opposite polarities

are attractive and therefore, the tendency of the two vortex-groups which

carry the flux in the memory cell (see Fig. 1) is to attract each other.

In the absence of any forces opposing this tendency, the two vortex-

groups will migrate toward and eventually anihilate each other. In such

a case the vortex stored energy is converted into heat which will be

generated in the vortex cores by viscous drag.

In addition to the intrinsic pinning forces due to film defects

there exist other forces in the memory cell which tend to localize the

stored flux and prevent the two vortex-groups from getting together. The

latter is provided by the presence of the lead drive lines which have a

higher critical field than the memory plane and stay diamagnetic at all

times. The two vortex-groups, which are found on both sides of such a

diamagnetic strip, have to enter the region below the lead line before

they can approach each other. If the energy of a vortex is increased on

entering the region ccvered by. the lead strip the vortex will tend to

avoid that region and thus, the lead strip edge acts as an energy barrier

which provides storage stability and enables one to operate a memory with

defect-free films.

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 110

The fact that the force between a vortex and the lead line is re­

pulsive can be demonstrated by the following simple argument. We know

from ordinary magnetism that when a perfect diamagnetic material is

placed in a region of a non-uniform magnetic field, surface currents are

induced which result in a force tending to move the diamagnetic material

toward regions of low field intensity. Therefore, as a vortex approaches

the lead line, a force is exerted on the line tending to move it toward

regions of low vortex magnetic fields, namely, as far as possible from

the vortex center. Using Newton's principle of action and reaction we

conclude that the force between a vortex and the lead line is repulsive,

and that a vortex undergoes an energy increase on entering a region

covered by a diamagnetic film. It is our purpose in this section to esti­

mate the magnitude of this interaction.

We first consider the situation illustrated in Fig. 21 in which a

vortex is supported by a superconducting film separated a distance

apart from a perfect diamagnetic film, both films being of infinite ex ­

tent. One should not be disturbed by the fact that the infinite size

of the films prevents the vortex magnetic fields to form closed lines,

since vortices are always created in pairs of equal and opposite fluxoids.

The magnetic flux of one vortex finds a return path through the electro­

magnetic region of the other pair member, and the combined fields of the

vortex pair cancel at large distances from the vortex centers. In

analyzing a single vortex in infinite films we keep in mind that the re­

sults are to be combined with those of the other pair member and the

physical difficulty of treating an object.with infinite magnetic moment

disappears.

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 Ill

SCREENING CURRENTS

VORTEX
CORE CIRCULATING
CURRENTS

FIG. 21. VORTEX COVERED BY PERFECT DIAMAGNETIC

FILM

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 112

W e assume the storing film to have a thickness d and a penetration

depth A satisfying d < 2A. The diamagnetic film, which represents the

lead strip, is much thicker than its penetration depth A^. . Since the

penetration depth of lead at the pertinent temperatures is much smaller

than the distance between the films, we can regard the lead film as

being a perfect diamagnet, namely A^ — 0 , without introducing a signifi­

cant error. In order to apply the previously developed technique to such

diamagnetic films, we let the diamagnetic currents be confined to a layer

of thickness d^ at the lower surface of the lead film, and later set d^

equal to A^ and let A^ approach zero.

Since the diamagnetic film is of zero fluxcid, its current density,

J , obeys London's equation

S1

J ------------------------------------------------------------ (13.1)
S1 n A/
'o 1

Replacing J by a surface current sheet of the appropriate strength, we

S1
obtain

f d]L -
Js ------ 2 A 8 (z " d 9 > (13.2 )
1 (j. A. z
'o 1

The storing film, on the other hand, carries a finite fluxoid, and

is governed by the relation

J s = - ~^~~2 (A - 0) = - (A - I) S(z) (13.3)

p. A u. A
'o "o

The vector potential, at any point in space, must satisfy

j d
curl curl A = pQ ($ - X) 6 (z) - A 5 (z - d ^ (13.4)
A A^

For a single quantum vortex situated at the origin of the cylindri­

cal coordinates, r, 6, z, we have

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 the equation for f becomes

p d f *o 1 \ r/ \ , d l r. C/ , .
aT + s ; 7 s7 y f = ^ v ' 5 y ^

(13.7)

Taking the Fourier transform of (13.7) along the z direction, we get

2 S 1 3 \ . d . ' % d 1 dl " ipd2

-P + 57 7 3 7 ^ F(r>p) = ^2 si(r) " 27~2 7 + ^~2 e §2(r)

(13.8)

where
p -i?z
F(r,p) = J f(r,z) e dp (13.9)

g x (r) = 27 / F (r ^ ) dP (13.10)

P
i r d^ d 2
g2 (r) = 27 J F <r 'P> e dP (13.11)

P
Following the procedure of Section 8 we now expand all functions in

Si S
terms of the eigenfunctions of the operator ^ ^ r, writing

F ( r ,(3) = J dy u^(r) a( 7 ,p) (13.12)

^=J dy u^(r) b (7 ) (13.13)

g t (r) = J d 7 u^(r) (^( 7 ) (13.14)

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 114

g 2 (r) = J d7 u^(r) C 2 (7) (1 3 .1 5 )

Inserting into (13.8), and using the orthonormal properties of u^, we

obtain

d, -ipd
a( 7 >P) = " "o 2 o b(7) + . e C 2 (7) + C 1 (7) (13.16)
P + 7 A A, A

From the definitions (13.10) and (13.11), C^( 7 ) and C 2 (7 ) are related to

a (7 ,p) by

C x (7) = ^ f a(7,P) dP (13.17)

1 P dpd 2
^ 2 (7 ) = jp J a( 7 ,P) e dp (13.18)

Integrating Eq. (13.16) over (3 and equating to 2ir C. (7 ) gives

d, -7 d 2
7T
^7TC1 (7) = - - (13.19)
2 Cl ^ ' 27T . 2 + 2 6 C 2 ^7)
A A A-
ipd 2
Multiplying Eq. (13.16) by e , integrating over (3 and equating

to 2 ttC2 (7 ) gives
-7 d d
' 2 o d '7 d 2 . dl
2ttC2 (7) = “ ^ \ C i (A) e C 9 (7) (13.20)
2ir 2 b(7) 6 + 2 ^2
A 1 A A

Equations (13.19) and (13.20) constitute two algebraic equations for

C 1 (7 ) and (^( 7 ), yielding

2 "27d2
$ 1 + 2 7 A /d - e
c l(7> = £ b(r) ---------- — ------------- (13.21)
-2 7 d r
(1 + 2 7 A l / d 1) ( l + 2 7 A/d) - e

We can now impose the condition of perfect diamagnetism by first

setting d^ equal A^ and then letting A^ approach zero, which results in

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 115

♦ , -lyi2
Cl(r) = H r ) ■1— h % r (1 3 .2 2 )
2 2
1 - e + 2 ) A /d

From E q s . (13.3) and (13.22) we can write for the current distribution

in the lower film

d o 1 / s
K s (r) =
2ir r 81 r
^oA

e d /
2^ b (7) " C x (7) u (r) dy
n oA2^7

?b(y)
a e u (r) d 7 (13.23)
(i irj ■2 7dr 2 7
0 7 1 - e + 2 7 d

Using (8.21) and (8.22) for the explicit form of b( 7 ) and u (r), we finally

have

J 1 (7 r) d 7 (13.24)
- 2 7 dr
o 1 - e + 27 A /d

Examining the asymptotic behavior of It (r) at r — o and r -* 00, we find

s

~c|>
K (r) = — (13.25)
s LI T ar
r — o o

0 0
K m (13.26)
s r - 00 ^oT (a + 2 d 2 )r

The 1/r behavior of K g (r) for small r is a direct consequence of

Eq. (13.3) and is a general property of superfluid vortices under all

conditions (see Eq. (8.30)). The asymptotic behavior of K g (r) for very

large x, Eq. (13.26), is obtained by expanding the integrand of (13.24)

in power series of J/ r -'ini keeping the first term only.

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 116

Since the structure of the vortex core is mostly sensitive to cur­

rents at short distances from the vortex axis, we conclude that the core

radius of a covered vortex is almost equal to that of an uncovered vor­

tex. However, due to the slow decay of current at large distances from

the vortex axis, the vertex self energy undergoes a significant change

upon entering a region covered by a diamagnetic film. This becomes ap­

parent when we examine the expression for the magneto-kinetic part of

the vortex self energy, using Eq. (6 .6 )

00

FM K = 1/2 *0 f K s (r) dr (13-27)

o

Due to the 1/r behavior of K g (r) for r — <» and r — o, E q . (13.27) diverges

logarithmically at both limits of the integration. The divergence for short

radii is common to all superconducting vortices and is remedied by the

finite size of the vortex core. The divergence at large distances, how­

ever, is a property which is unique to vortices facing diamagnetic films,

and indicates an infinite self-energy for the vortex described in Fig. 21.

In practice, however, the diamagnetic film has a finite width, so

that the integral in (13.27) should only be carried out to a distance of

the order of the lead line width. Hence, the self energy of a single vor­

tex in the region covered by the lead drive line has a logarithmic depen­

dence on the drive line width.

Another way of avoiding the energy divergence is to consider a pair

of equal and opposite vortices under an infinite diamagnetic film. From

<t>
Eq. (6 .6 ) the total magneto-kinetic energy of such a pair is given by —

times the total current that flows between the vortex centers. Hence, if

the distance between the pair members is we can w r ite

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 117

r 12

F = <D / K g (r) dr (13.28)

where K (r) is the current produced by a single vortex, and E is taken

s '
to represent the size of the vortex core. Since we are mainly concerned

with the increase in the pair energy upon entering the covered area, it

is convenient to calculate the energy difference between covered and u n ­

covered vortex pairs, both pairs having the same distance of separation.

If by K (r) we denote the current distribution of an uncovered vortex

su
and by K (r) that of a covered one, the free energy difference can be
sc
written
12

AF = <t>o J [K s c (r) " K s u ^r ^ dr (13.29)

For the current distribution of uncovered vortex we have found Eq. (8.23))
00

Ks u « - r 0 f I T + *ih ^ (13-30)
O

which combined with (13.24) gives

* 2 r 12
$ r\

AF = - 2- / dr I JL J 1 (7 r) d 7 =
-2 7 d2 1 + ya
W J J
& o 1 - e + 7

2 r l2 « - 2 7 dr

* - ! dr j ;-------------- 3 J,( 7 r) d 7 (13.31)

o
o (l + ya) ^ + ya. - e

From the fact that the coefficient of J^( 7 r) in the integrand of (13.31)

is a positive monotonically decreasing function of j we conclude that A F

is always positive. Thus, a covered vortex pair is always in a state of

higher energy compared with an uncovered pair of the same separation

distance.

R eproduced with permission o f the copyright owner. Further reproduction prohibited without permission.
 118

Reversing the order of integration in (13.31) and performing the

integral over r, gives

$ 2 00 “ 2 7d 2
f [Jo (« > - J 0 ( V 12)1 -j - J - ----------------- (13.32)
o (1 + yet) (l + ya - e
or, \ /
? - xd' /a
<P [J (x |/a) - J (xr „/a)]e
AF = — / — 2 --------- --------=--- 9 , , . dx (13.33)
|-i 77X7 J / -2xd /o;\ v '
° o ( l + x ) ( l + x- e )

-2xd^/a
Since the factor e forces the integrand to decay rapidly for values

of x which are greater than a/2d^, the main contribution from the denomi­

nator comes from small values of x. Therefore, we can approximate (13.33)

by taking only the first term of the power series expansion of the denomi­

nator . We have

^ 2 00 -2xd^loi

** - 57 z n h r j [V X ^/a> - J0(x ri2^)' ^ < 13-34)

o

or, ,
2
2
<t> , 1 + 1 + ( r l2 ./2 d )
o 1

** “ ^ l o g e T 7 / (13>35)

1 + v 1 + (l/2d2)

For the geometry involved in a practical memory cell we are mainly

IT
concerned with vortex pairs with very high ratios of 1 2 / 2 d2 (a typical

3T
value for a practical cell is 1 2 /2 d2 = 100) which allows a further simpli­

fication of (13.35):

~ * 2 1
^ “ (io7T a + 2d 2 loge ^r i2/ 2 d 2) (13.36)

Knowing the free energy differences between covered and uncovered

pairs, we can now plot the free energies of these two cases as a function

of the separation distance. The lower curve in Fig. 22 represents (except

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Reproduced
with
permission
of the copyright ow ner.

COVERED PAIR

TRANSITIO N
CURVE UNCOVERED PAIR
Further reproduction
prohibited

0 10 20 50 100 150
without p erm is sio n .

FIG .22. FREE ENERGIES OF COVERED AND UNCOVERED VORTEX

PAIRS VS SEPARATION DISTANCE
a / 2cf i 2 =I

119
 120

for an additive constant equal to the self energy of the two vortices

involved) the free energy of a vortex pair in an uncovered film, as

given by Eq. (10.13). The upper curve in Fig. 22 represents the free

energy of a vortex pair which is covered by an infinitely wide diamag­

netic film.

Returning now to the memory cell, we may examine the free energy of

a typical vortex pair under a lead strip of width w. For very large

separations > w ) the effect of the lead strip is negligible and the

pair energy is essentially that of an uncovered vortex pair. As the

separation distance becomes very small and the vortices enter deep into

the region covered by the lead line, the vortices behave as though they

are completely covered and the pair energy corresponds to the upper curve

of Fig. 22. In between these limits the pair energy undergoes a smooth

transition from the lower to the upper curve, indicated by the dashed

line in Fig. 22. The slope of this transition curve depends on how far

should the vortices penetrate under the lead line before all magnetic

fields are confined to the region below the lead, at which moment the

lead strip is equivalent to an infinite plane and Eq. (13.36) is appli­

cable. Regardless, however, of the exact value of this transition range,

Fig. 22 indicates that for all practical values of w and 62 the slope of

the energy versus distance curve is negative, representing a force which

tends to keep the vortex pair apart.

If a zigzag lead line is used to sense the Read Out signal of the

cell, then the memory plane is sandwiched between two diamagnetic films

(see Figs. 1 and 3). An analysis of a vortex covered on both sides by

diamagnetic films (Appendix III) shows that such a vortex is not essentially

different from the one illustrated by Fig. 22, having only one side covered

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 121

by diamagnetic film. The current distribution and the pair energy in

the sandwich case are similar to those expressed by (13.26) and (13.36)

if we replaced a + 2d^ by a, + d^- Thus, the expression for the free

energy difference of a sandwiched pair assumes the form

* 2 i
^ “ iT ir c T + X lo «e < ri2/2d2) <13' 37
'o 2

indicating a significant increase in the height of the energy barrier

which keeps the vortex pair apart.

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 122

14. CRITICAL CURRENT FOR UNIFORM SWITCHING OF SUPERCONDUCTING FILMS

In the remaining three sections we shall be concerned with critical

processes underlying the operation of the continuous sheet superconduc­

tive memory. Figure 23 serves to define the terms which are later used

in connection with the geometry of the memory cell. Figure 23a shows

a top view of the memory plane with the projection of the drive lines

(X and Y) and the zigzag sense line (ZZ). That area of the memory plane

which lies beneath the intersection of the X-Y drive lines is called

the "bridge". The circled areas C-C represent the regions along the

diagonal A-A which, under storage condition, accommodate the two v o r ­

tex groups. Figure 23b shows a section view along the diagonal A-A.

The geometry of this view is similar to the one described in Fig. 6 ,

except for the presence of two superposed drive lines.

When electric currents are applied to the drive lines, image cur­

rents, of the nature shown in Fig. 9, are flowing in the memory plane,

in areas beneath the drive lines. In the bridge area, these image cur­

rents produced by the X and Y driving currents are combined vectorially.

It is the bridge area, therefore, which carries the highest current

density, and where switching occurs first. In analyzing the switching

of the bridge area we need to consider the image currents due to the X

and Y drive lines as well as the stored currents which circulate around

the vortex regions C-C. The combination of these two current components

will be referred to as "film current".

In a practical memory cell the width of the drive lines is about

75 K& and their thickness is 5 K&. The thickness of the memory plane

varies between 1 to 3 k£, and the insulation layers (not shown in Fig. 23)

are 4 - 6 K & thick.

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 123

/
/ zz

/
/
/
.z.
/ /
/
J / . /
SHADED AREA "BRIDGE"
/ f x I
/ CC = REGIONS CONTAIN­
A ING VORTEX GROUPS
X /
V /
MEMORY
PLANE

I T -C
MEMORY
m
PLANE
c ZZ
SECTION A-A

(b)

FIG. 23- GEOMETRY OF CONTINUOUS FILM

MEMORY CELL.
(o) TOP VIEW
(b) SECTION VIEW (INSULATION LAYERS
OMITTED FOR CLARITY)

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 124

A typical value of the threshold current I , necessary for switching;

is about 150-250 ma, and the operating temperature is 3.4 - 3.6 K. 1^ is

defined as the minimum drive current (for an equal amplitude; coincident;

Read-Write pulse pair) required to cause the abrupt appearance of a sense

47
signal. This type of sense signal, sometimes called "snap action", is

characterized by a very fast rise time (~ 20 nsec) and a magnitude of

1 - 2 mv. The magnitude and width of such a sense signal are almost inde­

pendent on the rise time of the interrogating drive current. The nature

of the instability associated with the snap action will be discussed in

Sections 15 and 16.

Consider a uniform superconducting film in the Meissner state under

the influence of a drive line current I, as in Fig. 6 . The current I

represents the combined effect of the X and Y driving currents, and for

the moment is taken to have a uniform distribution across the drive line

strip

K = (14.1)
o 2b

The state of the superconducting film under such a current is deter­

mined by an effective wave function which is a solution to the Ginsburg-

Landau equations
- 2 K2
(i V + 2jt A/<t> ) \|r = — j (i|r - \|/ (14.2)
° A

— V X V A = J + J = r + j
u s a -.2
'o p. A
(14.3)

As in Eq. (4.3) \|r is normalized to unity at zero field, A is the zero

field penetration depth, and J is given by (2.1).

Denoting the magnitude and phase of if by f and S respectively

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 125

\|r = f e iS , (14.4)

equations (14.2) and (14.3) assume the form

-V 2 f + ( | 2: A - V s) 2 f = % (£ ' f3) (1 4 '5>

' o A

v 2 S = ~ (f : A - v s j (14.6)

f2 f * \
— V X V X A = Js + Ja =— 2 ^ - A + ^ v s j + J a(14.7)

In general f may be a function of all space variable. However, due

to the surface energy associated with fluctuations in the order parameter,

f can only vary significantly over distances comparable to A /k . In view

of the small thicknesses of the films involved we can regard f as being

uniform across the film. This approximation is valid for film thickness

up to d = which amounts to approximately 10,000 X for pure tin.

A further simplification of the Ginsburg-Landau equations is affected

by eliminating the phase variable S. Since all the currents in the configu­

ration of Fig. 6 are flowing in the y direction, we can choose the gauge
✓N
A = Y A (x,z) (14.8)

and assume no variation in f along the y direction:

f = f(x) (14.9)

Combining (14.6) and (14.7), we obtain

2 — —
2 4tt u A J • V f
V S= j---------- ^--- (14.10)
o f

which is'in the form of Poison's equation for S, with equivalent source

distribution proportional to J ’V f- In view of (14.8) and (14.9) J

s s

is perpendicular to Vf> the right hand side of (14.10) vanishes, and we

can choose

V S =0 (14.11)

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 126

as a proper solution. If in addition, S is a single valued function

then this solution is unique in view of the boundary condition

(
\VV S )/ n = o (14.12)

on the surface of the superconductor. The single valueness of S is

guaranteed by the requirement that the film be of zero fluxoid (in the

Meissner state). This is not the case, however, if vortex excitation

is allowed, in which case (14.10) is satisfied by the "streamline" type

fluxoid field (4.5).

Inserting (14.10) in (14.5) and (14.7), and using the current

sheet representation for the thin film current, yields

(14.13)

- 1/M-0 V A(x,z) = J g + J a = - ^ 2 A(x,o) 5(z) + J g (14.14)

Equations (14.13) and (14.14) constitute a pair of coupled nonlinear

differential equations which determine f and A for every value of the

applied current J . The solution of (14.13) - (14.14) is not unique as

3

can be seen for example from the fact that for all values of J Eqs (14.13) -
3

(14.14) are satisfied by the trivial solution f = 0, representing a normal

film. Such spurious solutions do in fact satisfy the stationary conditions

from which the Ginsburg-Landau equation stem, but are not stable in the

sense that they do not represent true minima (of the free energy) in phase

space; every finite perturbation will cause an abrupt transition to other

solutions which are stable. The selection of the right stable solutions

can be done by examining the behavior of the free energy, or more conven­

iently by physical reasoning. In the limit J =0, for example, we know

3

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 127

that the proper stable solution is f = 1. As I increases from zero, we

anticipate finding a continuous curve in phase space, starting at f = 1

and representing the sequence of stable solutions for all values of I.

If we continue to follow this curve in a direction corresponding to in­

creasing I, we shall eventually come to a point where an infinitessimal

increase in I corresponds to a finite increase in arc length. Such a

point should be identified with the critical current of a first order

phase transition, since it represents a finite change of state with

only virtual perturbation in external conditions.

The exact solution of (14.13) and (14.14) is very difficult. We

can however make use of the similarity between our geometry and the one
48 49
treated by Ginsburg and Douglass of a uniform magnetic field applied

to one side of a superconducting film. Such a situation exists for ex­

ample when a thin cylindrical shell is placed in a uniform magnetic field

parallel to the cylinder axis, or when a transport current is applied to

a superconducting strip in close proximity and above a ground plane. The

situation in the memory cell is that the magnetic field in the region

under the drive lines is almost uniform but drops to zero very fast out­

side this region, following Fig. 9. Therefore, as long as we restrict

our discussion to switching in the center of the bridge area, the Ginsburg-

Douglass treatment is applicable. The limitations on the applicability of

their results will be discussed later in this section.

In the uniform field case all quantities become position dependent

and the solution of (14.13) and (14.14) reduces to that of a simple alge­

braic equation for f; J being uniform in x, reduces (14.14) to

[1 A
A = K , (14.15)
d f

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 128

and inserted in (14.13) it results in

2
\2
2ir (j. A K \ 2 o /
o o \ K ,, .2. A
^--- 1 = ^ (1 ' f ) f (14.16)
o / A

Using (10.3), Eq. (14.16) becomes

2

= 2 (1 - f2) f4 (14.17)
d H cb

where H ^ is the critical field of bulk superconductors. Equation (14.17)

2
is plotted in Fig. 24, illustrating the dependence of f on the applied
2
current K . As K increases from zero, f starts from unity, decreases
o o

slowly, and then reaches the point where the slope of the curve becomes

infinite. This point corresponds to the maximum value of current that

can maintain a non-zero order parameter. Any further increase in K q

2
causes an abrupt drop of f to the trivial solution f = 0 .

The critical current is thus given by the maximum of expression

(14.17)

K = \l8/27' y H . (14.18)
oc A cb

and the value of the order parameter at the critical point is

f £ 2 = 2/3 (14.19)

In order to examine the validity of these results to the geometry

of Fig. 6 we can attack Eqs. (14.13) and (14.14) by perturbation method

and find out how uniform A and fremain asthe applied current starts to

increase from zero. The zero order solution is given by

f (o) = 1 A (o) = 0 (14.20)

The first order perturbation only affects A, giving

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 FIG. 24. CURRENT DEPENDENCE OF ORDER-
PARAMETER

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 130

which is exactly the differential equation treated in Section 2, the

solution of which is given by Eq. (2.4) and plotted in Fig. 9. Using

the approximation (2 .1 0 ) we can write for A ^ ) ( x , o )

P qA 2 |i o A2 ^ x2 + (a + d^)2 - b2

V x’o) “ -J- ' " V IT c“ ' — 2b ( a + dp 04.22)

For the second order perturbation we insert (14.22) in (14.13) and obtain

^2 _ f | 2 1 . f (2) = |j/2 A/d K (l) (-x)/ \ t) 2 (14.23

having the solution

a, -|x-x'|/{-

f (2 ) (x) = - ( in A/d H c ^ ) 2 J -e 2 t K (1 )2 (x ’> d x ’ U^.24)

-0 0

-|x-x'|/|
e 2
The kernel ---- — -— — indicates a localized averaging of (x)

over a distance £. Since in most of the region below the lead drive line
2
(x) (as illustrated by Fig. 9) is a slowly varying function of x, the
2
result of the averaging operation gives back the function (x). Right
2
under the drive line edge, where the function (x) varies rapidly in
2
position, the integration of (14.24) results in smearing out the (x)

function over a range equal to |, however, the general shape of the curve

remains unchanged. W e conclude that the reduction in the order parameter

is strongest under the center of the drive line and is almost uniform in

the range |x| < b. A further pursuance of the perturbation method shows

that also in higher orders of perturbation the general shape of A and f

follows the pulse type curves of Fig. 9, indicating that the highest value

of A and the lowest value of the order parameter occur in the center of the

cell, with no significant variations of these quantities throughout the

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 131

region |x|< b.

The fact that the order parameter stays almost uniform in the region

under the drive line is partly a consequence of assuming the distribution

of applied current to be given by (2.1) for all times. The fact is how­

ever that the current distribution in the lead drive lines depends on the

state of the memory film. When the film is of high order parameter it

acts as a good magnetic shield and the current distribution in the lead

line is almost uniform according to (2.1). W h e n the order parameter of

the memory film is reduced, the film loses its shielding properties and

the current distribution in the lead line approaches its free-space dis­

tribution, namely one where the current is highly peaked at the strip's

edges. This positive feedback mechanism is partially responsible for the "snap

action" sense signal; the drive currents, piling up at the edges of the

lead strip, cause a further reduction of the order parameter in the

memory plane, which in turn enhances the redistribution of the drive line

currents.

It should be noted, however, that the current redistribution before

switching is not so intense as it sounds. According to Eq. (14.19) the

memory film still contains two thirds of its superconducting carriers

before it switches. A reduction of one third in the electronic population

of the film has only a little effect on its magnetic shielding properties,

the reason being that the remaining superconducting carriers are still free

to short out any electric fields within the film and thereby prevent all

magnetic fields from penetrating its surface. As can be seen from Fig. 9,

a 3/2 increase in CC has only a slight effect on the current distribution

in the range |x| < b. We anticipate therefore, that before the switching

takes place, the film current distribution below the drive line remains

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 132

essentially that of Fig. 9. In order to maintain this current, the re­

duction of carrier concentration must be accompanied by a corresponding

increase of carriers' velocity. Such is also the case in the geometries

treated by Ginsburg and Douglass where the application of Magnetic field

on one side of the film fixes the current in the film and allows only

variations in carriers' velocity. We conclude that the first order phase

transition at high values of order parameter, which is predicted by the

uniform field analysis, is consistent with the assumption of uniform

fields, and therefore, (14.18) and (14.19) are self-consistent.

d2 f
In the uniform field analysis, the elimination of the term — x from
dx
Eq. (14.13) amounts to neglecting the surface energy from the Ginsburg-

Landau free energy expression. Since the magnetic field is not entirely

uniform along the switching area, the order parameter will vary corre­

spondingly giving rise to surface energy. One might expect to find a

lower critical current than the one obtained on the basis of uniform

order parameter.

To estimate the magnitude of this effect we let the order parameter

and the vector potential acquire a small curvature near the origin:

f = f + m, x 2 (14.25)
o 1

A2 = A 2 + m 0 x2 (14.26)
o z
2
Where m^ and m 2 are the measures of the curvatures of' f and A near the

origin, and f and A are the curvature-free values of f and A. Inserting

7 o o

expressions (14.25) and (14.26) into Eq. (14.13), and keeping only zero

order terms in x, we obtain

-2 e V f w j r ) f» = £ (£o ■ fo3> <1 4 -27>

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 133

which is independent on im,. Defining the positive dimensionlcss quantity

u2
= _ U H i = £.(q)_:_L(Sl (14.28)
f f (°)

and using (14.15) for A , we obtain the relation

d" % ) = 2 a ‘ £°2) f °4 ' 8 6 f °4 <14'29>

In comparison with Eq. (14.17), (14.24) describes a curve similar to that

4
of Fig. 24, but shifted to the left by an amount 8 e f . The critical

current is found by the maximum of expression (14.29) which occurs at

f 2 = 2/3 (1 - 4e), (14.30)

c

and has the value

K = d/A H cb J 8 / T F (1 - 4e ) 3 / 2 (14.31)
c

Compared with Eq. (14.18) we notice a decrease in the critical current by

3/2
a factor of (1 - 4e) which is due to spatial curvature of the order

parameter.

In order to estimate e we can regard Eq. (14.17) as a point relation,

connecting the order parameter and the film current density at every point
2
along the film. For f not too far from unity Eq. (14.17) reads

A K (x) ' 2

f = 1 " ( Sd-g ) <1 4 -32)

where K (x) is given by (2.11). For values of (a + d n)/b much smaller

S J_

than unity the behavior of f near the origin is

A K \ 2 f 4(a + d )
£ = 1 - 1 2 d-g— / ( 1+ - / *I <W -33>

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 and hence, e is given by

4 (a + d p i 2
1
2 (1 4 .3 4 )
2d H
1
A K
C

Near the critical point e assumes the approximate value

€ = 8/27 (a + d ) i2lb3 (14.35)

which is less than one thousandth for all practical geometries. We con­

clude, therefore, that the effect of the order parameter curvature on the

critical current is negligible and expression (14.18) is valid for the

critical current necessary to cause a uniform switching of the area under­

neath the drive line.

It should be noted that A which appears in (14.18) is the temperature

and mean-free-path dependent penetration depth. Using relations (12.16)

the temperature dependence of the critical current is given by

Ic (T) = I c (o) (1 + T 2 ) V 2 (1 - T2) (14.36)

Up to now we were only treating the switching of a virgin cell, namely

a cell with no stored flux. In an ordinary operation of the superconducting

memory there is always flux stored in the cells so that the critical driving

current under storage condition is the important parameter which determines

the operation of the memory. The presence of two vortex-groups in the

regions C-C (see Fig. 23) suggests the possibility of a different mode of

switching, that of vortex migration. If the drive current used to estab­

lish a certain flux storage is positive, then a negative drive current

pulse is used to interrogate that information and it exerts a force on the

vortex groups which tends to bring them closer together. If the interroga­

tion current exceeds a certain value this current-induced force may overcome

R eproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission.
 135

the pinning forces and vortex motion toward the center of the cell may

start. Such a motion is accompanied by a heat producing viscous drag

which raises the local temperature and further enhances the vortex motion.

If this thermal feedback has a loop gain greater than unity it will ac­

count for the abrupt switching associated with the "snap action".

There is some evidence however, that such a process, although impor­

tant to explain the disturb characteristics of the cell, is not the one

responsible for switching. If indeed the critical current for observed

switching is the current necessary to produce the critical Lorentz force

on a vortex group, one would expect to find a vast difference in the

critical current of a virgin cell and that o' storing cell. As was

shown in Section 9, a superconducting vortex adjusts the size of its

core in such a way that its maximum current density is of the order of

the critical current density of the superconductor. Therefore, the

presence of a vortex under the drive line amounts to having some area

which is on the verge of switching and a small drive current will ini­

tiate the switching of the cell.

The fact is, however, that experiments'*^ show at most a 40 percent

difference between the threshold currents of virgin and storing cells.

This leads us to believe that vortex motion is not the mechanism respon­

sible for cell switching. It seems more plausible that the stored vortex

groups are tightly pinned at the corners outside the bridge area, and

switching initiates at new nucleation sites.

Once information is stored, there is circulating current in existence

in the bridge region. Since the switching point depends on the strength

of the total film current, under this condition, the necessary drive current

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 136

to cause switching will not be the same as when there is no storage. The

amount of stored current in the center of the cell is determined by the

far field behavior of the vortex currents. To estimate this amount we

can regard each vortex group carrying a total flux 0 as a single vortex
0
of quantum number n, n being given by — . From Eq. (8.25b), the stored
o
current at the center of the cell is approximately given by

K = (14.37)
st p q 7T

where the two vortex groups are assumed to be a distance rQ away from the

center. The presence of the lead drive line should not distort the cur­

rent pattern by any significant amount as long as the vortices are situated

outside the lead strip. The reason for this is that most of the flux em ­

braces the lead line without being distored by its diamagnetic properties.

Since K is independent of the order parameter, the total film cur­

rent flowing in the film can be written as a superposition of K and the

current induced by the drive, K . In view of the shilding properties of

the film before switching, K gd is essentially equal to the drive current

image and is uniform at the center region of the cell. From Eq. (4.3) we
lir - 2 jr ^ A2 - -
can replace the term — A - yS of Eq. (14.5) by ------- "j- (Kgd + K ),
o b f s
o

and assuming again a uniform order parameter in the center region, the

equation for f becomes

f -X--- -) <X , + K J 2 - 2 (1 - £2) f4 (14.38)

V d H cb/ Sd St

Comparing (14.38) to E q . ( 1 4 . 1 7 ) , we find that the two are identical if

K q is replaced by the vector addition of the image current and the stored

current. Hence, the critical condition becomes (from (14.18))

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 137

<K s d + K st ) 2 = 8/27 (d/A H c b ) 2 (14.39)

cr

Since along the diagonal A-A the direction of K is at 45° to

the X and Y drive lines (see Fig. 23) we can write for the Cartesian com­

ponents of the critical current

(K + K / nT2)* + (K + K J s/~2) 2 = 8/27 (d/A H J 2 (14.40)

cx st cy st cb

Hence, the phase diagram describing the minimum combination of X and Y

currents necessary for switching lies on a c-Trcle in the K - K plane.

cx cy
The center of the circle is located at K = K = - K / ^ 2 and its
cx cy st

radius is given by s/ 8/271 (d/A H ^ ) ' ^ typi-0 3 ! 1 diagram is shown

in Fig. 25, according to data taken by Cosentino and M i l l e r . T h e Write

current amplitudes were plotted while Read current amplitudes were held

constant throughout the experiment. The resemblance of this plot to a

circle of the type described by Eq. (14.40) is only good in the range

where the difference between I and I does not exceed 150 ma. A large
x y
departure from a circular locus is noted when one of the currents becomes

much larger than the other. In this range the line carrying the large cur­

rent may switch the entire plane beneath it or may cause switching at dif­

ferent weak spots along this area before affecting the cell being tested.

Thus the data at this range is masked by side effects and is subject to

different interpretations.

Another reason for the departure of the I - I curve from the pre-
x y

dieted circular shape is due to the interaction between the two drive lines.

Being made of lead, which is superconducting at the temperatures involved,

the drive lines exhibit diamagnetic properties. Current applied to the X

drive line induces diamagnetic circulating currents on the surface of the

Y drive line and vice versa. This effect causes distortions in the image

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 138

300 ,
275

250
250 ma

225

250 ma
200 \

| 175 + + s

| 150
+> \

£125
\
\+
100 \
\ +
75 \
\ +
50
\ -
25 \ +
\ +
0 J L ± J I I I I fl I L+.J
0 25 50 75 100 125 150 175 200 225 250 275 300
I x (m a)-^

FIG .25. I x v s . I Y (DATA BY C0N S E N TIN 0 AND

M IL L E R ( 5 0 ) )

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 139

current induced in the memory plane and give rise to additional current

components which were not taken into account in the derivation of (14.40).

Indeed, drive lines made of normal material, where the diamagnetic inter­

action does not take place, yield 1^ - I curves with a perfect circular

shape.

It should be noted that E q . (14.40) only represents the conditions

under which the superconducting state in the center of the cell becomes

unstable. It is possible that this still is not a sufficient criterion

for overall cell switching and that an additional triggering mechanism is

necessary to force the fast "snap action". It is also possible that before

the critical state is achieved another process, presently forbidden by our

restriction on f to be uniform, can take place, and result in a lower criti­

cal current than the one predicted by (14.40). Those possibilities are the

subject of the following sections.

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 140

15. EFFECT OF FILM DEFECTS ON THE SUPERCONDUCTIVE TRANSITION

In the previous section we have seen how the state of a superconduct­

ing film varies upon the application of a uniform magnetic field on one

side of the film. The film was assumed to be idealy uniform and for that

reason it was reasonable to restrict the solution of the Ginsburg-Landau

equations to one narrow class of solutions namely that which involves a

uniform order parameter throughout the film area. By applying this r e ­

striction on f, we have left a large portion of phase space, that which

includes all position dependent solutions, unexplored. One might argue

that such solutions, even if they exist, are not excited since all con­

ditions are translation invariant along the film surface. This argument

breaks down if the film contains material defects which perturb the trans­

lational invariance of the system and may trigger position dependent

solutions.

It is the purpose of this section to examine the system behavior

when the film characteristics are allowed to vary in position. We shall

first study under what conditions the Ginsburg-Landau equations remain

stable against defect stimulation, and then examine what type of defects

is most likely to trigger instabilities.

Dealing with film characteristics which vary in position, the order

parameter in the absence of magnetic fields is itself position dependent,

and hence, a renormalization of the Ginsburg-Landau equations is required.

We start with the original Ginsburg-Landau free energy expression (Eq.

(6 .1))

(1 5 .1 )

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 141

where describes the normal state free energy density, a and p / 2 are

the first two coefficients in the expansion of the condensation energy

2
in powers of |i|r| , and ijr is the Ginsburg-Landau "effective" wave function.
2
\jr is normalized in such a way that |\j/| equals the concentration ng of

the superconducting carriers. To represent defects we let a and p be

position dependent. Writing \j/ in terms of its magnitude and phase

± = R e lS (15.2)

reduces (15.1) to
2

F = I “If + OR + p/2 + ' (vR) + M- H + ~ R V S -eA) rdv

b J ^ n 2 tn" 2m ” J

(15.3)

The equations for R,S and A may be found by the requirement that F

be stationary with respect to these three variable, that is, by the Euler-

Langrange equations

2
- V 2R + OR + pR 3 + ~ (h V S - e A) 2 R = 0 (15.4)
2m 2m

V 2 s = - ^ (VS -f - A) (15.5)

v x V X A/m- = 7 = R 2 ^ («vS - e ' A) (15.6)

m

We see that letting a and p vary with position does not change the form

of the Ginsbrug-Landau equations.

Since we are mainly interested in the point where the effect of de­

fects starts to become significant, we can investigate the system behavior

in the linear approximation. We let Ct and p have uniform parts and p^,

plus small position dependent parts and p^. Thus,

a = «o + a (?) (15.7)

p = Po + p x (?) (15.8)

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 142

Due to these fluctuations in film characteristics the dependent vari­

able R, S and A will correspondingly acquire small flucuations;

R = R q + R (?) (15.9)

S = S q + S 1 (7) (15.10)

A = ^ + ^ ( 7 ) (15.11)

The fluctuations designated by the subscript 1 are assumed to be much

smaller (for all r) than the corresponding uniform components.

As was discussed in the previous section, and can be seen directly

from (15.5), the position dependent of part of S must be taken into con­

sideration, if we allow the order parameter to vary along the current

direction. These phase fluctuations represent variations in carriers’

velocity which must occur in regions of different carriers' concentration

if the current is to remain divergence-less.

Equation (15.6) only describes the vector potential within the super­

conducting volume. The applied current, J , should be added to the right

side of (15.6), to make it applicable to every point in space. We again

treat the case of a uniform current sheet applied on one side of the film.

Inserting (15.7) - (15.11) into (15.4) - (15.6) and separating terms of

different orders of magnitude gives:

.,.2
o #v a
a R + P R + ^-7 A R = 0 (15.12)
o o ro o 0 « o o
2m

2 1 (Z) e*2 2
-V "J = ~ V Rq (z) A o (z) + J a (z) (15.13)
o m

and for the first order position dependent parts:

0
^ 0 *'v
D_ •.
’*Z
‘ ? E1 + “o R1 +3P0R02 Rj ■ < e % - h V Sl)+ =

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 143

= -a,R - p,R 3 (15,14)

1 o 1 o

0 0 * 7 R-, ‘ A
v si = ^ — V - 2 <15-l5)
o

9 9 /%
J-
/v
-V A, , =R - ( ^ V S , - e" A.) - 2 R R , — A (15.16)
1/Li O 1 1 O 1 O
o m m
2
Note that in Eq. (15.13) R q (z ) is uniform (for thin films) across the

film, but is zero outside. Making use of the current sheet representa­

tion of the film, Eq. (15.13) gives

A0 = " ^ 2 ^ 2 <15 -17>

e o

where J q is the uniform current density in the film which is related to

the applied field by

H
Jo = T
a <15-18>

Thus, Eq. (15.12) becomes

2
1 1 * J
a R +p R + 4 S-s— ^ = 0 (15.19)
o o o o 2 j.2 3
e“ R
o

It is now convenient to normalize R to the value it will acquire in

a uniform film and in zero fields;

|P 1
f = R — = f + f. (r) (15.20)
\ -a o lv v
\l o

Also, defining the reduced variables

a = — ----- = 1 + a . (r) (15.21)

H A 0 1
o ,c

1 = 7 ^ - =j + j;(r) (15.22)
■R H ° 1
c

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 144

a 2
and using the relation -— = p. H , (15.23)
° (3 o c

we obtain the set of equations

j 2 = f 4 - f 6 (15.24)
o o o

ao = " Jo /f (15.25)

(s/2 |V V a ^ f J fQ

(15.26)

(15.27)
f °

2 2 -
-A V ax + f0 (aL - 2 | VS]L) - 2j
o f
d 5 (z) = 0 (15.28)

In these equations A is the penetration depth of a uniform film in zero

magnetic field, | is the coherence length given by

(15.29)
2 e )_i H A
o c

and H is the thermodynamical critical field'of a uniform film. Note that

c

the Laplacian operator in (15.26) is two dimensional while that appearing

„J
in (15.28) is three dimensional.

Equation (15.24) is identical to the one obtained in Eq. (14.17) where

f was restricted to be position independent. From Fig. 24 we gather that

2
as long as j < 4/27 the linearized Ginsburg-Landau equations have a solu-
2
tion with a uniform component of f given by (15.24). When j exceeds this

critical value of 4/27 the only solution to the linearized Ginsburg-Landau

2 2
equations is the trivial one f = f^ =0. It does not mean, however,

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 145

that the system undergoes an abrupt transition to the normal state, since

such a transition, under the constraint of uniform currents, amounts to

infinite kinetic energy of the superelectrons and therefore is not plausi­

ble. The meaning of the break in the f - j curve of Fig. 24 is that

the system, undergoing a phase transition, is no longer in thermodynamical

equilibrium and can no longer be described by the Ginsburg-Landau equations

which stem from the requirement that F g be stationary. In terms of the

2
system's behavior in phase-space one can say that as j reaches the value

4/27 the point given by Eq. (15.24) is no longer a true minimum (of F )

but becomes an inflection point. Thus, the representative point "slides"

toward another minimum along a curve which is no longer a solution of the

time independent Ginsburg-Landau equations. This "sliding path" is not

arbitrary but is chosen by the system in accordance with some time depen­

dent variational principle like, for instance, the Hamilton principle

which determines the time development_of classical systems.

In view of this discussion, Eqs. (15.24) - (15.28) should be sepa-

2 2
rately considered in the regions j < 4/27 and j > 4/27. In the first

region f is given by (15.24) while the first order corrections to f and

to A are given by (15.26) - (15.28) with j as a parameter. It still re­

mains to be checked whether in this range the solution for f^ and a^ is

stable against all Fourier components of the excitation function

In affecting the solution of (15.26) - (15.28) we make use of the

Green's function technique which was developed in Section 1. We notice

that Eq. (15.28) is of the form

C V 2 u(r) = 5 ( z) u (r) - v(r) (1 5 .3 0 )

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 1 46

which was treated in Section 1 (Eq. 1.3)). The transverse Fourier com­

ponents of u and v in the plane 2 = 0 were shown (Eq. (1.12)) to be

related by

v (q ; - iq )
(15.31)
= ■i + r s r —

Taking the transverse Fourier transforms of all the functions involved in

(15.26) - (15.28) and applying (15.31) to (15.28) results in:

12 n/T ^ j 2j
* J<
(-2| 2 q 2 -D) fx (q) - £ S l (q) + J -2 ai (q) = p(q) (15.32)
j
.
o

i2q 2
f (q) - q S (q) = 0 (15.33)
s/2 e

i s/7 |
~— ;---- S n (q) + a. (q) = 0 (15.34)
1 + a q 1 lx

•2 j fi(q) 1 7? | q
- - i S l(,) + a (,) (15.35)
3 1 + aq 1 + a q

In Eqs. (15.32) - (15.35) D is a function of j and f given by

2 3
D = 3f H - 1 , (15.36)
o f 4
o

j is assumed to flow along the y direction, p(q) is the Fourier transform

of the defect induced driving function

g (?) 2 a ( r)
P(?) = £ (15.37)
a

a,^ and a ^ are the x and y component of the reduced vector potential, q

is the magnitude of the transverse wave vector

2 2 2
q = q + q (1 5 .3 8 )
x ^y

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 147

2
and a is equal to 2A /d (not to be confused with a which appears in (15.3))

Equations (15.32) - (15.35) can be written in a matrix form

P(q) = M (q) X (q) (15.39)

where the matrix M (q), connecting the driving vector

P (q) = (p (q); 0, 0, 0) (15.40)

to the system response vector

X (q) = (fj^Cq); Sx(q); alx (q), ai y^>) ' (15.41)

ly

is given by
-i 2 £ j -2 j.
2 2
(-2 q - D)

i 2 J,
0 0

M (q) = (15.42)

- i \/~2 £
1 + a q x

"2 J, -i \/~2 £
1 + a q
f (1 + a q )
o

The determinant of M is given by

2 2 2

A(q) = q {2 | q + D - ~ ~ + (15.43)

For X to be stable A should not vanish for all real q's. It can

be readily verified that a sufficient condition for A to have no zero in

the real q ^ ; q^ plane (except for the double zero at the origin) is

4j
+ = 3£ 2 - 1 < 0 (1 5 .4 4 )
4 o
f 4
o

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 148

This stability criterion is shown in Fig. 26 by the curve which separates

the f , j plane into stable and unstable regions. This curve passes
2 2
through the critical point fQ = 2/3 j = 4/27, leaving the upper branch

of the f (jQ) curve in the stable region. Hence, we conclude that in the
2
range j < 4/27 the linearized Ginsburg-Landau equations have a solution

which is stable against defects perturbations. In this region the system's

response is proportional to the magnitude of the defects - induced driving

functions, and has different sensitivity to different Fourier components

of the latter.

Using the inverse of M we find for the components of X

2
fi (q) = -9=*- p(q) (15.45)
*(q)

Sl<*> = - 2i - 4 ^ 7 A^fj-P(9) (15-46)

o

aix ® ’ “ t (1 +7^ ) A ® p® <1 5 '47>

O
2j q 2

V ’>= ■ 7 r +^ i )- A(i T p® (15-48)

o

Let us examine now the behavior of the order parameter in the range

j < 4/27. Using (15.24) and (15.43) Eq. (15.45) gives

f 1 ( q ) = - £^9)----------------------------- ------- ----- (15>49

2 2 , 2 2 a qx
2 | q + 6 f - 4 + 4(1 - f )
o v o q(l + aq)

It is seen that for the same value of q, f^(cj) increases with increas­

ing ratios of q ^ / q ^ We conclude that material fluctuations which vary

along the direction of current flow have a larger effect on the order

parameter than those occurring in parallel with the current. This is in

accordance with our intuitive feeling that obstacles blocking the current

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 oo uJ
cn >
O lj
Q_ QC
tE

C/ ) 0

M O

FIG. 2 6 . STABILITY OF ORDER- PARAMETER

AGAINST DEFECTS PERTURBATION

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 150

flow have a higher effect than those laid in laminar order parallel to

the flow.

The highest p e r turb ation of f^ is obtained by putting = q in

(15.49), giving

f.(5) = 2 P ( ^ / '2 , 2---- = - T - (15.50)

5 q ( £„ - £0C2)

The order parameter is more sensitive to long range defects (low q's)

than to short range ones, the latter are averaged out over a distance

comparable to £. The effective band in the s pect r u m of p(q) is determined

2
by how close f is to its critical value of 2/3. In real space, the re ­

lation between f^(y) and p(y) is given by

J
CO

f i(y ) = 2jr p ( y ' ) e ^y "y ^ 7 ^ d y' (15.51)

-0 0

2 2 2
where 7 is equal to 3(fQ - f ). The integral of (15.51) describes a

localized average of p(y) over a distance £/ 7 . As the u n i f o r m component

of the order parameter approaches its critical value, the contribution

from low q components of p becomes higher and higher. Finally, when the

critical current is reached, the relation between f(y) and p(y) becomes

2 d2f i(y) = i /2 p(y)

-I ------------------------------------------------------------- (15.52)
dx

If the fluctuations in the material properties vary periodically over a

distance h, then
.2
f, (y) = P(y) — 7 (15.53)
2r

large fluctuations in f^ can be affected if h is much larger than (• .

2
The behavior of the film in the range j > 4/27 requires special

consideration. As was discussed earlier in this section^ in this range

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 151

the system is no longer in equilibrium and so, the relation between f

and i is no longer given by (15.24) but rather follows one of the ire-
Jc
reversible paths indicated in Fig. 26- If j increases very fast to

above its critical value, the horizontal superheating path is likely to

take place. If j reaches its critical value and then remains constant,

f follows the vertical path. The behavior of f^ in this range is d e ­

termined by some time dependent principle which is not implied by the

stationary Ginsburg-Landau equations.

In the absence of time-dependent Ginsburg-Landau equations one is

forced to use plausible speculations as to the actual shape of the se­

lected irreversible path. In this respect the analogy with other non­

equilibrium processes might serve as a guide. When a ball is sliding

on a surface in a gravitational field, the selected path is not the one

along the highest potential gradient, but deviates from the latter be­

cause of the finite inertial mass which tends to keep the ball moving
52
in a straight line. According to Suhl's analysis, the kinetic energy

associated with a changing order parameter is very small compared with

the other quantities involved in the Ginsburg-Landau free energy. This

is valid as long as the characteristic frequency of the changes involved

is less than the superconductor's gap frequency. It is reasonable there­

fore to assume that the transition path selected by the system lies close

to the path with the highest free energy slope. Namely, for every value

of j and fQ the position dependent part, of the order parameter will still

satisfy the Ginsburg-Landau equation.

As is indicated by the stability criterion in Fig. 26, the range

2 —
j > 4/27 is also characterized by possible instabilities, namely A(q)

might have zeros for finite values of q\ In such a case the order parameter

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 152

will resonate to a p ar ticula r band in the p(q) spe ctrum depending on the

location of the re pres e n t a t i v e point along the irreversible path chosen.

2
Kee p i n g jQ at its critical value and allowing f to decline along

the vertical line of F i g . 26, the zeros of the determin ant in (15.43) are

satisfied by the equation

^ 2 2 2 12 16
q (1 + osq) ( 21 q + 3fo - ~ - 1J + fj - f - = 0 (15.54) ’

^ ^ ' fo

Eq. (15.54) describes a locus in the q^ - q^ plane w h i c h determines what.

combination of q and q is required to make the d e t e r m i n a n t vanish. A

^x y

family of such root loci is plotted in F i g . 27, for diff e r e n t values of

O'
f and K 1 = ---- . For low values of K 1 the loci are al most symmetric
0 n/ 2 £
in q and q , w h i l e for h ig h values of K 1 they curv e inward on the q
nx y x

axis.
2
As f decreases from its critical value of 2/3, higher wave-vector

components of the p (q) s p e c t r u m are allowed to trigger instabilities. From

the asymmetry of the root loci in Fig. 27 one conclu des that p(q) spectrum

components of large ratios of q / q^ are scanned faster than those wit h low

2
q^./- As a result, when fQ drops below 2/3, the instability in the

order parameter w h i c h is m o s t likely to occur first, is in the form of

very sharp fluctuations along the current flow, and only smooth variations

transversely to the current. This is in accordance w i t h the film response

2
in the range j < 4/27.

The m a i n part of the p(q) spectrum is con cen t r a t e d around a wave

vector which is the reciprocal of the mean grain size, and is symmetrical

in q and q . The size of the film's grains is u s u a l l y of the order of I*,

x y
2
We see that only a slight decrease in f is n ecess ary to fully scan the

who le p(q) spectrum. While.the average order paramet er along the film can

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 FIG. 27. ROOT LOCUS OF A ( q ) FOR DIFFERENT
VALUES O F f 0 AND k '(=2>?/Vz£d)

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 154

be as high as 504> of its zero field value, unlimited fluctuation in its

position dependent part can occur and allow flux penetration to change

the film's fluxoid state.

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 155

16. VORTEX CREATION, FLUX SWITCHING AND DISTURBS IN MEMORY CELLS

As was discussed in the preceeding section, the switching of the

memory film does not occur uniformly over the bridge area. Instead, it

was shown that the transition involves a fast increase in the nonuniform

part of the ordered parameter, triggered by long ranged defects, which

lie perpendicular to the direction of film current flow. Such defects

will cause the order parameter to vary along the direction of the cur­

rent in a form of a wave (representing a varying degree of superconductivty)

whose amplitude is a rapidly increasing function of the applied current.

Carrying this picture back to the structure of the memory cell, and

extending it beyond the limits of the linear theory, we can envision the

generation of normal channels, perpendicular to the film current, which

allows the magnetic flux to penetrate the superconducting film, as illus­

trated in Fig. 28. After a sufficient amount of flux has penetrated

through the normal channels, the current in the bridge area is reduced

and the channels are healed, leaving two vortex groups at the regions C-C.

This model of switching is physically more appealing than the one describ­

ing an overall switching of the bridge area (see Fig. 2), since it demon­

strates a process which, leading to the same final state, involves a

lower rate of entropy production.

The principle of minimum entropy production, which is used to de ­

scribe many irreversible processes in physics, can be very helpful in

trying to construct a switching model for the memory cell. Basically,

the superconductive analog of this principle can be described in the

following simple terms. Since the superconducting state is of a higher

order than the normal state at the same temperature, when superconductivity

is destroyed by magnetic fields the entropy of the metal increases abruptly

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 BRID G E NORMAL VORTEX
AREA CHANNELS GROUPS

oa

CU R R EN T LIN ES -
IN M E M O R Y F IL M

B E F O R E SW ITCHING D U R IN G SW ITCHING A F T E R SWITCHING

F IG .2 8 C ELL SWITCHING BY N O R M A L C H A N N EL S
 157

(analogous to the latent heat associated with liquid-gas phase transition),

that is, the system requires an external supply of heat in order to m a i n ­

tain the same temperature. In vi of the relatively long time associated

with heat transfer processes, rat 1 ;r than absorbing that heat, the tempera­

ture of the metal drops and tends to slow down the transition to the normal

state. It is clear now that if two processes can lead the system toward

the same final state, the process with the lower rate of entropy production

has a higher probability of occurance; it will take place at the highest

temperature and so will face the least opposition to its development. Since

the entropy of a superconductor depends on the fraction of unpaired elec­

trons, that process which switches the minimum amount of superconducting

volume into the normal state will produce the minimum entropy and will

occur first. During the switching process described by the channel model

of Fig. 28, only the small volume of the channel lines becomes normal while

the rest of the bridge area remains superconductive. On the other hand,

the model describing a uniform overall switching of the bridge area in­

volves an entropy production which is proportional to the whole bridge

volume.

In view of the fast switching time of the memory cell, it would be

difficult to experimentally decide which of the two switching models ac­

tually takes place. There exists however some experimental evidence con­

nected with the possibility of orthogonal storage, which supports the

channel model. By the proper combination of X and Y pulses it is possible

to store information along the two diagonals of the bridge area. It was

found that these two storage modes are independent of each other, that is,

one can store information along the first diagonal and then come back and

read it correctly, while in between, information is stored along the second

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 158

diagonal. If the total area of the bridge turns normal during the switch­

ing then the flux stored along the first diagonal would decay during the

application of a Read-Write cycle to the second diagonal. Such a flux

decay is not observed; confirming the channel model in which the normal

channels formed only connect the corners C-C of a single diagonal while

the corners along the other diagonal remain entirely unaffected.

It is also possible to explain the development of such switching

mechanism using "vortex language". Assume that at some "weak" spot along

the film the order parameter has dropped to zero and a normal nucleous

has been formed as in Fig. 29a. It is reasonable to assume that such a

nucleous is formed in the vicinity of defects such as grain boundaries;

where the variations in the film characteristics reduce the amount of

surface energy associated with any normal superconducting interface. The

rest of the film still has a finite order parameter and carries a current

close to critical. If the intensity of the nucleating defect is not too

high we expect, frm the discussion of the preceeding section; that the
2 2
rest of the film is close to the point f = 2/3, j = 4/27.
o 3 Jo

The views in Fig. 29 represent a small region which is part of the

bridge area. The current lines shown represent the total current flowing

in film and is almost uniform throughout the region considered.

After the formation of the normal nucleous, the current will not

flow through the latter, but instead will seek out new paths in the

manner shown by Fig. 29b. The change in the flow pattern results in a

more intense current density inthe neighborhood of the normal nucleous

which causes the normal area toelongate in the direction perpendicu­

lar to the current (see Fig. 29c). This propagation of the normal

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 159

M A G N E TIC
F IE L D L IN E S

SIDE =====
■ X
VIE W S =E Z E

CURRENT
^ L IN E S

TOP m
VIE W S

(a ) (b ) (c )

SIDE
V IEW S i7i . ty£ ~

TO P
V IE W S

W •)

SUPERCO NDUCTING
B R ID G E

(d ) (e )

FIG. 29 SUCCESSIVE STEPS IN THE PROCESS OF VORTEX

PAIR PRODUCTION (REPRESENTING A SMALL
REGION IN THE BRIDGE AREA)

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 160

phase is particularly easy in view of the fact that the entire film is

already on the verge of switching. Also, since a normal superconducting

interface is already in existence, the additional surface energy required

for expansion is small.

The normal area continues to grow as indicated by Fig. 29c, with a

further increase in the current density at the edges, resulting in a

further change of shape. We now see how the aforementioned normal channels

can regeneratively grow via the interaction between the film currents and

a normal nucleous. Such a growth will indeed continue to the edges of

the cell region if the temperature rise due to Joule losses in the normal

region prevents the recovery of the central portion of the channel. If,

however, the growth is slow, or if the film is in good thermal contact

with the bath, the heat generated by Joule loss is conducted away and a

different process takes place.

We notice (Fig. 29b) a reduction in the current density at these

portions of the normal area boundary which are facing the current flow,

and that the reduction in current is even more pronounced after the shape

of the normal area has stretched out as indicated by Fig. 29c. Thus, the

central portion of the normal area is exposed to.almost no current; it

can therefore recover back to the superconductive state, forming a super­

conducting bridge between two normal regions (see Fig. 29d). Once a super­

conducting bridge is formed, a pair of equal and opposite vortices is

created which are drivenaway in the manner shown by Fig. 29e.

The formation of the superconducting bridge is not possible at any

arbitrary stage of the process, since the fluxoid quantization rules

should be obeyed. Thus, the necessary condition for the formation of the

bridge is that the total fluxoid carried by each of the normal regions be

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 161

a multiple integer of <l> .

From the analysis of Section 9 we learn that in order to accommodate

one unit of fluxoid; a superconducting vortex should contain a normal

core whose radius is of the order of the coherence length £. Therefore,

it is reasonable to assume that the formation of the superconducting

bridge is only possible if the two vortices thus created are at least 2|

apart. This condition is equivalent to the quantization requirement since,

as shown by (9.26), the proper strength of the fluxoid is guaranteed by

having the critical current density circulating around a radius | .

Another way of arriving at this condition is to calculate the cur­

rent strength necessary to separate the vortex pair. We consider a

system of two equal and opposite vortices, situated a distance 2| apart

and immersed in the field of a uniform current density J q .According to

Eq. (7.15) the pair is subject to two opposing forces; the attractive vor-

tex-vortex interaction, and the interaction with the uniform current which

tends to separate the pair apart. A necessary condition for the pair to

persist is that the dispersive force be greater than the attractive one.

In case this is not satisfied the pair is annihilated. If, however, the

attractive force is smaller than the repulsive force the vortex members

will separate and be driven to the corners C-C of the memory cell where,

in the absence of any driving force, they will be stopped by some pinning

defect. Since the attraction force decreases with pair separation dis­

tance, it assumes its highest value upon creation, and once overcome by

the constant dispersive force can never restore the force balance.

The attraction force is determined by the close field behavior of

the vortex current, giving (using (6 .1 ) and (8.25a))

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 162

"c 2 1
fatt ~ (J.o7T a ( 2 |) (16.1)

while the dispersive force is given by

f,. = <t> K (16.2)

dis o o v '

Equating the two forces and using Eq. (15.29) gives the condition for. K :
o'
• <t> d
K = S = H c d/A > d 6 -3)
P oTT 2A 2|

and for the current density:

K
I
J'o d = “c U 6 .4)

Thus, the current density necessary to break up a vortex pair is

just equal to the critical current of the superconductor. This surpris­

ing result makes one wonder whether what is usually called current induced

switching of superconductors is any different from current induced pair

creation.

Since the mechanics of Sections 6 and 7 are based on the narrow

core approximation, and in view of the ambiquity in the exact definition

of the core boundary, we only expect Eq. (16.4) to give the order of m a g ­

nitude of the critical J q . Keeping this in mind, and comparing (16.4) to

(14.18) we learn that the current level at which the Meissner state b e ­

comes unstable is within the range of currents which can excite vortex
\
pairs. This supports the thesis that vortex creation is the mechanism

responsible for flux switching in the regular operation of the memory

cell.

W e recall that one of the arguments used to support the channel

model was the principle of minimum rate of entropy production. We might

note that the creation of a single vortex pair actually involves a much

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 163

lower entropy production than the formation of a complete normal channel.

The production of vortex pairs by an imperfection in the film is not

limited to a single pair but can continue as long as the total current in

the vicinity of the imperfection is sufficient to trigger that excitation.

»
As shown in Fig. 29, the current in the creation center is reduced by the

vortex creation, because the total current is now made of the superposition

of the uniform current J q and the vortex current, given by Eq. (8.25).

Since the vortex current is inversely proportional to the square of the

distance, the total current at the creation center is small as long as

the vortices are close together, but as they travel far enough, the total

current approaches J q and the creation process can start all over again

(Fig. 2 9 e ) . Thus, we see that a single imperfection can act as a source

of vortex pairs. The spacing between the moving vortices as well as the

total number of pairs created depend on the amount of current overdrive,

that is, the amount by which the initial current J q is higher than the

one required to trigger the creation process. A simple calculation shows

that if K, is the driving current and K is the triggering current then

d o

the spacing between the first adjacent vortices is given by

'
rs = V <K°-K0) < 1 6 -5 >

The pair creation will eventually stop when a sufficient number of v o r ­

tices are created and the combined current reduces below K • If the two
o

vortex groups find rest at a distance w away from the creation center

then the maximum amount of flux that can be generated is given by

(K, - K ) _
$ = --- 2 --- T w (16.6)

This is the total flux linking a toroid of radius w carrying a current

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 164

(K,-K ) 1/2 per unit length. The speed at which the vortex pairs move
d o

toward the corners C-C of the cell is determined bv the driving Lorentz
53
force and by the viscous drag. Stephent and Bardeen have calculated

the viscous drag associated with a moving vortex to be

* 2
T1 = - 2-^o (16.7)
2tt a

where a is the effective radius of the core and cr is the normal conduc­

tivity of the metal. Taking a = £, and equating the driving force to

the viscous force we obtain for the velocity of a moving vortex

V i 1 (16.g)
L J , A Li W
cb o

For driving currents close to critical and in pure tin, expression (16.8)
2 3
gives velocities in the order of 10 + 10 meters per second.The time

required for a vortex pair to traverse a distance comparable to that of

the memory drive lines is in the order of . 1 + 1 . 0 psec. This time is

short enough to develop a detectable sense voltage if sufficient number

4
of vortex pairs (« 10 ) could be excited simultaneously. If however

only few nucleation- centers take part in the pair production process, the

rate of total flux change is not sufficient to be detected by the sense

amplifier. Some experimental observations, which will be discussed later

in this section, indeed support the fact that a certain amount of flux can

be switched without the detection of a sense signal.

The density of nucleation centers is primarily dependent on the

amount of current overdrive. As the film current increases above its

thermodynamic critical value (Eq. 14.18) more and more nucleation cites

are established and start to act as pair creation centers. If the density

of production centers is high, one can envision a process in which vortex

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 165

pairs are simultaneously delivered by many production centers. Vortices

moving in the opposite directions annihilate one another, and only the

ones at the two edges of the cell survive. Since every vortex only

travels half the mean distance between nucleation centers before being

annihilated, the net result of such a process is the almost instantaneous

transfer of flux to regions C-C of the memory cell. Such a rapid flux

transfer is very similar to-the one portrayed by the channel model. In

fact as the mean distance between nucleation centers approaches 2| the

pair production mode is in all respects equivalent to the channel mode.

Thus, the pair production, the channels and the uniform switching models

all describe switching modes which take place at different degrees of

nucleation density.

From the nature of the "snap action" sense signal it is clear that

the switching process must contain an avalanche mechanism, which is

either of thermal or electromagnetic nature. The avalanche picture is

analogous to ionization breakdown in gases where the acceleration of few

ions may result in a total ionization of the gas. As a vortex moves along

the film it can generate more nucleation centers along its trajectory path,

which turn into new pair production centers.

The thermal aspect of this process is a consequence of the heat

generated by the viscous motion of the vortex which raises the film

temperature along the vortex path and thus reduces the current threshold

for new nucleation centers to form. The onset of the avalanche process

requires that the heat generated by vortex motion exceeds the amount of

heat which is conducted away from the cell, and so it depends both on the

amount of current overdrive and on the rise time of the driving currents.

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 166

The electromagnetic aspect of the avalanche can be un derstood by-

noticing that the m o v i n g vort ex carries w i t h it circulating currents of

high intensity which, when combined wi t h the u n i f o r m applied current,

can e sta b l i s h more nucleation centers along the v ort e x path. For the

ele ctromagnetic avalanche to occur the formation of new p roduction

centers should be faster than the reduction of nu cle a t i o n threshold

caused by the v ort e x groups settling at the cell edges. This also d e ­

pends on the rise time of the driving currents.

Pre vi o u s l y in this section we have obt ained an expressio n (Eq.

(16.6)) for the amount of switching flux as a function of the current

overdrive. The derivation of this expression assumed an isothermal

switching process in which the nucleation threshold stays constant in

time. The heat generated during the thermal avalanche clearly reduces

the nucleation threshold and results in a total flux which is greater

than the one given by (Eq. 16.6). Also, in an isothermal process the

motion of the flux carrying vortices stops as they arrive at the

regions C-C where the intensity of the driving force is reduced, while

in a non-isothermal process the temperature rise reduces the pinning

forces and allows the vortex motion to extend beyond the cell boundaries.

In this way more vortex pairs can be created and more flux remains trapped

after the switching has elapsed. Hence, both increasing amount of over­

drive and faster current rise times (which reflect an increasing degree

of heating) will increase the amount of stored flux.

The dependence of stored flux on the level and rise time of Write

currents is shown in Fig. 30 and is reflected by the corresp onding d e ­

pendence of the threshold current for Read. The following was done"^ for

pulses of three different rise times, 50 nsec (fast), 300 nsec (normal),

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 167

I WRITE
I READ

_ n i
I X=L
IT
3001 -
250

£ 200

150
RISE TIME
100 □— 50 nsec
+— 300nsec
a— 1500 nsec
50

o 50 100 150 250 300

I READ (ma)

FIG. 3 0 . EFFECT OF DRIVE PULSE RISE TIM E

ON THRESHOLD C U R R E N T (D A TA BY
CONSENTINO AND M IL L E R (5 0 ))

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 168

and 1500 nsec (slow). The switching threshold was first observed for a

virgin cell when subjected to a Read-Write sequence (equal amplitude

pulses). The Write current was then increased in steps, noting the

minimum Read current required for sense. Figure 30 shows that if we

keep the level of the Write current the same but decrease its rise time,

more flux is stored in the film, and the Read threshold is reduced.

It should be mentioned that although the temperature rise during

the "snap action" switching can severely affect the current thresholds

and the duty cycle of a memory, it is not sufficient to cause the whole

bridge area to go normal during switching. This is clear from the fact

that independent orthogonal storage has been observed experimentally

even under the extreme conditions of fast rise times and high current

levels.

Up to this point we have considered pair production centers, which

originate from normal nuclei. The formation of such nuclei was assumed

to take place when the film’s current exceeds its thermodynamic critical

value. This picture is probably valid when the film contains only weak

defects. If, however, very intense irregularities exist along the cur­

rent carrying areas of the film, the production of vortex pairs can be

initiated at much lower current levels. Figure 31 shows an electron

microscope pictures of typical tin films used in the memory, as taken

54
by Mezrich. The surface of the films is shown to be very irregular

and even contains actual holes of worm type shape. The worm holes, when

present in a current carrying area of the film, introduce a severe dis­

continuity in the current path, especially if they are oriented at right

angle to the current flow.

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 169

10 A

RATE RATE

6 0 A/sec
m
6 0 A/sec

^■j
9

4 0 A /sec I 4 0 A/sec

2 0 A /sec 2 0 A/sec

PRES SURE - I x I 0 " 6

T E M P - 5 0 °C c
THICKNESS - 1 0 0 0 A

F16. 31 ELECTRON M IC R O G R A P H S OF T Y P I C A L TIN

F I L M S U S E D FOR M E M O R Y PLANES
(AFTER M EZRITCH)

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 170

From classical hydrodynamics we know that if an obstacle with sharp

edges is placed in the path of flowing fluid, the velocity field at the

edges is many times greater than the original fluid velocity. In fact,

the velocity of an irrotational fluid has a logarithmic singularity near

a sharp edge. In real fluids the singularity is smeared out by the ap­

pearance of a rotational component which is supported by the finite vis­

cosity of the fluid. The current field on the surface of a perfect

diamagnetic material is also irrotational and has therefore a singularity

near the sharp edge of an electrical discontinuity. In view of the

finite penetration depth of superconducting films, there exists a non­

vanishing component of magnetic field which is perpendicular to the film

surface and gives rise to a small rotational component in the super­

currents field. In Section 1 we have shown that the transverse pene-

2
tration depth of thin films is 2A /d and so we conclude that the current

singularity near a sharp edge is smeared out over a distance comparable

to 2A^/d.

Irrespectively however of the exact behavior of the current field,

it is apparent that the current intensity near the sharp edge of a worm

hole, can be many times higher than the one in a uniform film. Hence, the

nucleation of a normal region and the formation of a vortex at the corner

of the worm hole can take place at current levels lower than the thermo­

dynamic critical value. The vortex formation follows the steps indicated

in Fig. 29 except that the defect which now triggers the process is not

the spot-like normal region of Fig. 29a, but an elongated region of material

deficiency. Once a vortex is formed, the current-induced Lorentz force

drives it away from the hole toward the corner of the cell. If the shape

of the worm hole is not symmetric, for example if one edge is sharper than

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 171

the other, it is possible to excite only a single vortex leaving the hole

with a net fluxoid of the opposite polarity. In order to separate the

vortex the Lorentz driving force should overcome the attraction force b e ­

tween the vortex and the hole. This force can be thought of as arising

from the interaction between the excited vortex and its negative imagei

Because of the high degree of current accumulation near the hole edge,

the driving force is sufficient to overcome the image attraction and the

vortex is driven away from the hole.

The most important difference between the vortex production process

induced by material deficiencies and the one described in Fig. 29 lies in

the fact that in the first case the excited vortex travels in a super­

conducting film whose average order parameter is close to unity, while

in the second case the film is on the verge of thermodymamic instability.

As a result, in films of weak defects the creation of vortex pairs invari­

ably brings about thermal avalanche. In films with very intense irregu­

larities the process of vortex creation can proceed isothermally until

the cell is "charged" with a sufficient number of vortices such that the

current near the creation site decreases, and the production process stops.

In other words, one would expect to find two critical currents in a non-

uniform film; the driving current necessary to initiate vortex pro­

duction, and Ic 2 ^ t^e driving current necessary to cause a thermal ava­

lanche. Since the vortex production in a nonuniform film only takes place

at few isolated spots and at a relatively low current level, the rate of

flux change is low and is not detected by the sense line. Thus, the criti­

cal current as measured by the appearance of a "snap action" sense signal

is Ic 2 ‘ The flux stored in the memory cell can be changed at lower current

levels without the appearance of sense signals.

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 172

From the nature of the two critical currents it is apparent that

I , is a constant depending on the film characteristics, while I „ can

cl c2

vary depending on the rise time of the driving currents and on the

thermal resistance of the cell. The upper limit on I ^ is the thermo­

dynamic critical value I q , and the lower limit is given by I If the

density of film vacancies is high, and the cell is thermally isolated

from the bath, then the creation of only a few vortices may result in

a chain reaction and a thermal avalanche is affected. In such a case

I „ is close to I ,. If, however, the film contains only a few vacan-

c2 cl

cies per cell and the cells have a good thermal contact to the helium

bath, then isothermal vortex creation is possible up to current levels

close to the thermodynamic critical value. In a uniform film I , and

cl
I 2 become close together because the vortex production occurs simul­

taneously at many nucleation points; the film then has a single critical

value close to I .
o

From a standpoint of cells1 uniformity, it is clear that since I

depends strongly on the size, shape and orientation of the film vacan­

cies, it cannot be controlled and unless the film is made entirely free

of vacancies will vary vastly from cell to cell. The effect of a nonuni­

form I ^ on the overall performance of the memory is twofold. First,

variations in I ^ are usually accompanied by corresponding variations in

I 2 (the faster the vortex creation the faster the thermal latch) which

is the critical current for observed switching. Secondly, large differ­

ences between I ^ and I make it possible for the X (or Y) current alone

to diminish the flux stored in the cells along the X line and thus result

in a "Disturb" sensitive memory.

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 173

The name "Disturb" designates all measurable perturbations of the

cell's stored flux which are caused by currents addressed to other lo­

cations in the memory. An ideal memory should be disturb insensitive,

that is, the only allowed perturbance in the cell's stored flux should

be caused by switching due to coincident X and Y current pulses. If

the stored flux is modified in response to the X or Y current alone,

then the sense output as well as the critical current will vary from

cell to cell, according to the sequence of Disturb pulses they are ex­

posed to, and this will result in a malfunctioning memory. Therefore,

the cell's Disturb characteristic is an essential parameter tc determine

memory performance. In addition, the behavior of the memory cell under

the influence of different Disturb patterns furnishes essential informa- •

tion on the switching mechanism and on the dynamics of the stored flux.

Disturb measurements on a large number of cells and under a variety

of conditions show a full agreement with our physical model based on

vortex dynamics. The standard Disturb test consists of measuring two

currents: the threshold current IT and the breakthrough current 1^.

Equal amplitude Read and Write pulses are applied to both X and Y lines,

the current amplitudes are increased until at I = I the first sign of

(snap action) sense voltage appears. The X and Y Write pulses are then

taken out of coincidence making the sense voltage disappear. All current

amplitudes are further increased until, at the value 1^, the sense voltage

reappears under the Read pulse. The difference between 1^ and I defines

the current tolerance for the memory cell, that is, theapplied current

level must be greater than 1^, but should not exceed I . The larger the

difference 1^ - 1^ the better the chance of finding all the memory cells

operating at a single current level.

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 174

The reappearance of the sense voltage, as the current level is

raised up to 1^, indicates that the separated X and Y current pulses

have succeeded in modifying the stored flux in such a manner that the

following Read pulse (the same pulse which is responsible for the present

stored flux) causes switching of the cell. Since the test is performed

in a steady state mode, the total flux that is switched under the Read

pulse must be switched back under the disturbing X and Y pulses. However,

no sense signal is observed under these pulses. This experimental fact

leads to the inevitable conclusion that flux can be reversed in two

different modes; a fast mode, resulting in a snap-action voltage, and

a slow mode, which does not yield any detectable sense signal. It is

logical to associate the fast mode with the thermomagnetic avalanche and

the slow mode with isothermal processes like vortex motion and vortex

production. The explanation of the cell's response under Disturb condi­

tions is that the X and Y current when applied separately are not suffi­

cient to cause a thermal avalanche but are above the value required to

start vortex production. In other words, the film's current during the

Disturb pulses is above 1 ^ but below I ^ , and as a result, the total

flux stored is slowly reduced. During the Read pulse, the total current

flowing in the film is larger than that which would exist without the

flux degradation, hence, the current overdrive is larger, I 2 exceeded

and a thermal runaway is achieved.

The results of a more extensive test which

I demonstrate the existence

of these two switching modes are shown in Fig. 32. Read - Write current

pulses of equal amplitudes in X and Y lines were applied to a single memory

■cell. Starting with equal amplitudes of Read and Write, the Write pulse

was continuously increased and the threshold values of the Read pulses

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 175

250

iyy (md)

W R ITE READ

100 150 250

Ip(ma)

FIG. 32. THRESHOLD CURVES FOR COINCIDENT

DISTURB.

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 1.76

were recorded. Two distinct threshold values of Read currents were

observed; I „ T> the minimum value of the Read pulse which is sufficient
(jW

to cause a sense signal to appear in coincidence with the Write pulse,

and I-,.,* the minimum value of the Read pulse which is required to cause —
CR
the appearance of a sense signal in coincidence with the Read pulse. As

may be seen from Fig. 32, I and I _ follow two separate curves. Start-
cw CR
ing at a common point I = I = 115 ma, I,_T is strongly depressed with
CW O K CW

increasing Write currents and drops to zero as 1^ reaches 190 ma.

on the other hand, is only slightly sensitive to increasing Write currents

and is seen to remain finite at high values of T .

w
The distance T _ - 1^.. is a measure of the difference between the
CR CW
threshold currents for the fast and slow switching modes. At any par­

ticular value of 1^, which represents a certain amount of stored flux,

^CW a ^way s higher than the current required to trigger the slow

mode of flux switching. This is appreciated from the fact that the sense

signal under the Write pulse disappears when 1^ is made lower than I

and reappears again when I exceeds I . For the same value of I , I

R CW vv CR

is equal to Ic 2 > t*16 current required to trigger the fast mode of flux

switching and which gives rise to the sense signal under the Read pulse.

Figure 32 shows that as the amount of stored flux increases there comes

a point where only a slight enhancement of the stored current is sufficient

to cause flux degradation. On the other hand, a finite amount of overdrive

is always required in order to trigger a thermomagnetic avalanche.

The point 1^ = 190 ma, 1 ^ = 0 requires special attention. This

point represents the situation in which repetitive Write pulses (not

followed by Read) are applied to the memory element and yet a sense signal

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 177

of the same polarity appears in coincidence, with the leading edge of

each pulse. Since the flux stored in the memory cell cannot possibly

build up indefinitely, there must be some mechanism by which the flux

relaxes back to its original value after the snap action of the rising

current is over. Indeed, a small sense signal of the opposite polarity

was observed during the tail edge of the current pulse, but of a much

lower magnitude than that of the snap action signal. Thus the relaxa­

tion process is slow and does not involve a thermal avalanche. This

is expected since the forces which act on the vortex groups after the

current pulse has terminated are only the interaction forces between

vortices. These forces are weak and consequently the vortices do not

move fast enough to develop a thermal catastrophy.

Another explanation for the slow speed of the flux relaxation

process is that it involves an inward motion of vortices. Such a m o ­

tion is opposed by the diamagnetic properties of t;he lead strips that

sandwich the memory film. It can be seen from the negative slope of

the transition curve in Fig. 22 that regardless of the polarity of the

vortices involved, the nature of the diamagnetic interaction is such

that it assists outward vortex motion and impedes inward motions. Thus,

when vortices are created and driven to the edges of the cell by the

current induced Lorentz forces, the diamagnetic interaction assists

their motion. This eventually results in a violent injection of flux

into the film. However, when the flux decays a diamagnetic energy bar­

rier blocks the return path resulting in a slow decay, too slew to

trigger thermal avalanche.

It should be mentioned that as one further increases the drive cur­

rent amplitude, a point is reached when a sense signal appears on the

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 178

falling edge as well as on the leading edge of the current p u l s e . T h e

amount of flux initially injected into the film is so great that the re­

storing forces are sufficient to accelerate vortices to higher speeds,

and this results in a second thermomagnetic avalanche that is observed

as a snap action signal in a direction opposite to that produced by the

leading edge.

The same memory cell was also tested for its response to non-coinci­

dence Disturb pulses. The pattern of pulses used in described in Fig. 33.

The test procedure was the same as that of the previous experiment except

that this time the Write pulse is followed by non-coincident X and Y Dis­

turb pulses of equal amplitude. Two threshold currents were recorded:

I^, the minimum amplitude of Disturb pulses necessary to develop a

voltage in coincidence with the Write pulses, and Ip n > the minimum Disturb

current necessary to cause a signal during one. of the Disturb pulses. In

all cases the sense signal first appears under the Y pulse, irrespecively

of the chronological order at which X and Y Disturb pulses are applied.

The reason for that is that the Y drive line is closer to the memory film

than the X line, and so the Y-current induces image currents of higher in­

tensity than those induced by the X current.

The results obtained for non-coincident Disturbs are plotted in

Fig. 33. Again, two distinct curves are observed for I and I , having
L»W CiD

similar shapes to those plotted in Fig. 32 for the case of coincident Dis­

turbs. The irregularities in the I curve is probably associated with

(jD

the severe conditions of forcing the memory film to switch under the whole

length of the Y line. Another difficulty associated with the I curve is

VjU

defining the exact threshold value at which sense signals start to appear.

It so happened that there is a wide range of current where the sense signal

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 179

250

200 -

CD
lew
50

Iw^m Q )

100

y— TL___ 2 -
50 WRITE DISTURBS

0 50 100 150 200 250

Ip (ma)

FIG. 33. THRESHOLD CURVES FOR NON-COINCIDENT

DISTURBS.

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 180

may appear under only few Disturb pulses but not under the rest, and so,

it is difficult to determine exactly at what current level switching

actually took place.

The general similarity in the behavior of the curves in Fig, 32 and

Fig. 33 shows that the mechanism of flux degradation is the same for non­

coincident and coincident Disturb. However, non-coincident current levels

necessary to affect a measurable disturbance of the stored flux are more

than twice as high as those necessary to affect the same disturbance in

the coincidence case. It looks as though the disturbing effect of the co­

incident X and Y pulses is as strong as though theywere combined linearly

rather than vectorially (a proportionality factor of l/\/~2 would characterize

vectorial addition). This apparent linear addition of X and Y currents in­

dicates the existence of a strong stored current at the position where

the disturbance takes place. Since the direction of the stored current

is at 45° to X and Y, only those components of X and Y image currents

which are normal to the diagonal A-A (see Fig. 23) combine with the stored

current and have a sufficient intensity to affect the disturbance. The

image current components which are along the diagonal A-A are perpendicular

to the stored current, and have only a second order effect.

Two possible mechanisms were mentioned which can be responsible for

the slow changes that take place in the stored flux; vortex production,

and vortex migration. In the former, new vortices are created and are

driven to the regions C-C of the memory cell where they annihilate some

of the previously stored flux. In the latter, already established vortices

are forced to migrate toward the direction where the opposite vortex group

is located and eventually get annihilated. For both processes to occur,

it is necessary that the driving force be of sufficient strength to push'

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 181

the vortices over pinning barriers, and that the moving vortices find

an "easy" path with the least resistance to vortex motion. The varia­

tions in vortex energy with position of its core can be represented by

a topological map containing a random pattern of hills and valleys. The

hills represents regions of high energy, which may be a result of high

film thickness or high order parameter, and the valleys represent regions

of relatively lower energy. A vortex subject to a driving force can

only travel as long as it is not confronted by an. energy slope higher

than the driving force. Thus, the distance traversed by the vortex d e ­

pends not only on the magnitude of the driving forces but on their direc­

tion as well.

This model of vortex motion is in excellent agreement with the data

taken by Nosker"^ showing the dependence of flux degradation on the

pattern of Disturb pulses.

A standard Write-Read cycle, of equal magnitude in X and Y currents,

was applied, followed by many various patterns of Disturb sequences. All

the currents used were kept at about 20 ma above, the threshold value, and

the magnitude of the sense signal under the Write pulse (the pulse follow­

ing the Disturb pattern) was measured and compared to its value under con­

ditions of no Disturbs. The degradation in the magnitude of this sense

signal is a measure of (though not necessarily proportional to) the amount

of flux degradation upon Disturbs, This way, different sequences of Dis­

turb pulses could be compared as to their effect on the stored flux under

standard initial and final conditions.

The data of Table 3 represents a typical set of characteristics for

a memory cell at 3.5°K. The notation used in Table 3 is the following:

+
a positive current pulse on the X line is denoted by X , coincident positive

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 18 2

Percentage of Percentage of
Voltage With Voltage With
No. Disturbs No Disturbs No. Disturbs No Disturbs

1 None 1.0 25 K y V ) .93

2 lx+ .93 26 2 ( y V ) .93

3 5x+ .93 27 10(y+-y") .93

4 103x+ .93 28 103 ( y % ‘) .93

5 iy+ .93 29 lx+ y' .93

c + -
6 5y+ .93 30 5x y .91

7 103y+ .93 31 20x+ y" .89

'+ +
8 l(x+-y ) .89 32 150x+y' .85

9 2(x+-y+ ) .85 33 1500x+y" .78

10 10(x+— y+ ) .85 34 lx t"*" .93

11 103 (x+~ y + ) .82 35 5x y ' •91

n r\
12 106 ( x V ) .82 36 20x'y+ . oy
- +
13 108(x+-y+) .82 37 150x y .85
4- +
14 2x — 2y ) .89 38 1500x'y+ .78
<

+
1
r-f

.89 39 .93
f

15 10x+~10y+
3 + 3 +
16 10 x -.10 y .87 40 2(x+y - x y+ ) .91

17 lx' 1.0 41 10(x+y"-x“y+ ) .90

3 -
18 10 x 1.0 42 150(x+ y — x y+ ) .78
3 + _ _ +
19 ly" 1.0 43 10 (x y — x y ) .74
3 -
20 10 v 1.0 44 106 (x+y"-x~y+ ) .70
in8. + - - +.
21 l(x+-x') .93 45 10 (x y — x y ) .70
00

*+1
— 1

+
4:

22 .93 46
1
o

2(x+-x')
+ - - +.
23 10(x+-x") .93 47 -y - x y - x y ) .70
3 + -
24 10 (x - x ) .93

Table 3. DEGRADATION OF SENSE VOLTAGE WITH VARIOUS PATTERNS.

OF DISTURB PULSES [DATA BY NOSKER (55)]

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 183

current pulses on. X and Y lines are called X Y , a positive Y pulse

+
followed by a negative X pulse is denoted by Y — X , the number appear­

ing before the pulse sequence indicates how many times the sequence was
“f" -|>
repeated. For example, the description 2 (X ~Y ) is short for X — Y — X — Y .
+ + - -
The data obtained is for a Write-Read pattern (X Y - X Y ) followed by a

sequency of Disturbs, and the voltage measured is the one appearing in

coincidence with the X Y Write pulse.

Several interesting observations can be made from the Disturb degra­

dation data of Table 3. It appears that, with the current levels used, the

sense voltage can be degraded by at most 30$, regardless of the number of

Disturb pulses. The first several Disturbs appear to do most of the

damage after which many more Disturbs produce only negligible further

degradation. The number of Disturbs after which the sense signal levels

off depends strongly on the pattern of Disturb pulses used. Under a single

line Disturbs (no. 2 -5- 7) the stored flux seems to remain fixed, at 7$

less than its original value, already after the first pulse. Interlaced

X +-.v+,X+-.Y+ . . . Disturbs (no. 8 13) degrade the sense voltage to 18$

below its original ralue and saturation is reached only after several hun-

“h “h “I* ■f*’
dred pulses. Furthermore, interlaced X'— Y — X — Y Disturbs degrade the

sense voltage more than X — Y . . . Y —Y ... Disturbs. For example,

•j* -}■
the sequence 10X — 10Y has the same effect on the stored flux as the se-
-|* -J- "j-
quence 1(X — Y ) while the. sequence 2(X -Y ) degrades the voltage by an

additional 4$. Higher current pulses, if used in an interlaced sequence,

may cause the sense voltage to gradually decrease to zero, while the single

line Disturbs of the same amplitude only cause a reduction of 25$.

The explanation of these phenomena, in terms of vortex mechanics, is

as follows: when the first Disturb pulse is applied, a certain fraction

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 184

(in our example 7$) of the vortices present succeeds in finding a permea­

ble path along the direction of the driving force and get annihilated,

while the other vortices are stopped along their way by high energy

barriers. The reapplication of the same Disturb pulse will produce no

further effect since the force tends to push the vortices over the same

barriers which were proven to be impermeable. If however, a successive

pulse is applied to the other drive line, the direction of the force has

changed by 90°, and along this new direction some vortices may find a

permeable path. This explains why a single sequence of X — Y has more

effect than thousands of pulses on the X (or Y) line alone. But even

among those vortices that did not make it the second time, the change

in force direction causes a redistribution, so that when the first Dis­

turb pulse is applied again, new paths are opened and more vortices can

travel toward annihilation. In the interlaced sequence every pulse may

cause a redistribution of vortices and so increases the probability that

the pulse to follow will succeed in further degrading the stored flux.

This explains the large number of pulses necessary to reach a steady state.

An example of a possible vortex path under two types of Disturb se­

quences is illustrated in Fig. 34. The pictures represent a small region

in the bridge area, and the moving vortex is reprsented by the black cir­

cle. This vortex can be part of the stored flux which has penetrated the

bridge area or, a newly created vortex produced by the disturbing currents.

The black lines represent impermeable energy barriers while the white

regions among them are accessible. It is shown how the interlaced Disturbs

are able to drive a vortex a far greater distance than single line Disturbs.

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 DIRECTION
OF FORCE
t t

)
I A
VORTEX
POSITION

t
*
j ) •)
I / i
b.) INTERLACED DISTURBS

FIG. 3 4 . IL L U S T R A T IV E COMPARISON OF VORTEX

MOTIONS UNDER IN TE R LA C E D AND SINGLE-LINE
DISTURBS.
 186

APPENDIX I

In order to evaluate the integral 1/2 J ' ® we extend the

definition of $ to include all space. The nature of the extension

should not affect the result, since the integral to be evaluated depends

only on 0 in the superconductor. It is convenient to extend 5 in a con­

tinuous fashion and for this purpose Eq. (4.5) is taken as the definition

of $ everywhere in space

F'mk = 1 / 2 J' $ * J g dv = 1/2 J ' • curl H dv =

= 1/2 J ' H •curl 0 dv + 1/2 J (H x "$) ‘ ds-

The surface integral vanishes if we extend the surface to infinity, since

H falls off like 1/r^.Using (4.5), we get:

F =1/2 / H Z n. O 5(p - p.) dv

MK ./ z . lo
u l

+ oo
n. 0 n

1
“P J
-00
Hz<Pi> dz
But the integral of H z is just equal to the total current 1^, that em-

braces the i “ vortex, so

0 r~i
MK = 2 o2- / ),n - i 1 i-

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 187

APPENDIX II

We wish to express the magnetic interaction energy between a vor­

tex and a current source in terms of the fluxoid field and the current

induced in the superconductor. We decompose the vector potential and

the current field into their components:

A = A + A + A^r J = J + J„
o . so V s so V

A q is the vector potential due to the applied currents when the film is

removed, A gQ is the vector potential produced by alone, A^ is the

vector potential produced by the vortex currents J^. alone, and J is

the current induced on the superconductor in the absence of the vortices.

The vortex-source magnetic interaction energy is given by U = J dv

thus,

U = / (A-A - A ) • J dv = / [-J •A + (A - A )■J ] dv =

J so v v J so v v v

= /-(J *A + A^p. J ■ J ) d v = - / j • $ dv
J so v o so v . J so v

where we have used the relations

/ J -A dv = / J -A dv
J v so J so v

- - 2 ■—
and A - A ^ = - A P Q J go in the superconductor. This completes the general

proof of the relation

/ a ’J dv = - / J dv
. o v . so v

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 188

APPENDIX III

We wish to find an expression for the superconducting current in

a thin film which supports a single quantum vortex and is sandwiched

between two perfect diamagnetic films. Let the film carrying the vor­

tex have a thickness d, a penetration depth A, and be located a distance

away from each of the diamagnetic films. As was shown in Section 4

(Eq.4.12) the fluxoid field plays the role of an externally applied cur­

rent source situated in the interior of the film that carries the vortex.

Therefore, we can regard the current distribution in the middle film as

arising from three different current sources, these are the fluxoid field

and the surface currents induced on the other two films. Owing to the

linearity of the system we can use superposition and independently com­

bine these three contributions.

From (4.17), that part of the film's current which is due to the

fluxoid field 0 is related to the latter by

(III.l)

A' 'o, from (4.16), the part of K g which is due to external surface cur­

rent K , situated a distance d„ from the film, is given by

K (q) = - K (q) f — --- (III.2)

s a 1 + aq

Denoting the surface currents of the diamagnetic films by and K^, we

get for the Fourier transform of K :

s

(III.3)

From symmetry, the two diamagnetic films carry the same current distribu­

tion, and so

K x (q) = K 2 (q) (III.4)

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 189

The surface current is produced by two sources; located a distance

away, and K 2 , located a distance 2d2 away. Combining these two contri­

butions and letting be zero to represent perfect diamagnetism, we obtain

(using III.2)

-d„q "2d„q
K L (q) = - K s (q) e - K 2 (q) e (III.5)

Equations (III.3), (III.4), and (III.5) constitute three algebraic equa­

tions for K g , and K 2 , giving

\ ® - ir *® ---- /---- 23^

l + a q ^ 2 / l + e

Substituting expression (4.21) for $(q) a^d going through the same

procedure as in (8.27) and (8.28), we get

* 00

KsU(r) ■ s r j ----7 2d M Ji(r<1> (IH-D

°° l + a q - 2 / l + e

This integral is similar to the one in (13.24) except for the term
' 2qd2 f I
e [ 2/ 1 + e ) , which resylts in a slightly different behavior

of the integrand at low q's. Correspondingly, the asymptotic behavior of

the current at large distances becomes

K (r) = 9 — - y (III. 8)
s p. 7T (a + d„)r '
r-*» 'o 2

which is similar to (13.26) but contains d2 instead of 2d2>

The derivation of the free energy difference between sandwiched and

uncovered vortex pairs follows the same lines as in Eqs. (13.31) - (13.36),

giving

•„2 2
AF — loge (r12/2d2) (III.9)
p ott a + ~2

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 190

REFERENCES

1. Kammerlingh Onnes, H. Leiden Comm. Suppl. 34, (1913).

2. Collins, S. C. (1956), unpublished, E. A. Lynton "Superconductivity",

p. 3, .John Wiley and Sons Inc.

3. Silsbee, F. B., J. Wash. Acad. Sci. 6_, 597, (1916).

4. Meissner, W. and Ochsenfeld, R., Naturwiss, 2JL, 787, (1933).

5. Gorter, C. J. and Casimir, H. B. G., Physica 1, 306, (1934).

6. London, F. and London, H.,-Proc. Roy. Soc. A149, 7JL, (1935).

7. Daunt, J. G. Miller, A. R., Pippard, A. B. and Shoenberg, D., Phys.

Rev. 74, 842, (1948).

8. Pippard, A. B., Proc. Camb. Phil. Soc. 47, 617, (1950).

9. Reuter, G. E., and Sondheimer, E. H., Proc. Roy. Soc. A195, 336,
(1948).

10. Ginsburg, V. L. and Landau, L. D., J.E.T.P. USSR _20, 1064, (1950).

11. Bardeen, J., Cooper, L. N., and Schrieffer, J. R., Phys. Rev., 108,
1175, (1957).

12. Gorkov, L. P. Translation, Soviety Phys. J.E.T.P., 9, 1364, (1959).

13. Doll, R. and Nabauer, M., Phys. Rev. Lett. 1_, 51 (1961).

14. Deaver, Jr., B. S. and Fairbank, W. M., Phys. Rev. Lett. ]_, 43 (1961).

15. Abrikosov, A. A. Ah. Eksperin i Teor. Fiz. 32^, 1442 (1957) [Translation:
Soviet Phys. - J.E.T.P. _5, 1174 (1957)].

16. Tinkham, M. Phys. Rev., 1 2 9 , 2413 (1963).

17. Saint James, D. and de Gennes, P. G., Phys. Lett. 306, (1963).

18. Cardona, M. and Rosenblum, B., Phys. Lett. j3, 308 (1964).

19. Tomasch, W. J., Ehys. Lett. 9_ 104 (1964).

20. Buck, D. A., Proc. IRE, 44, 482 (1956).

21. Newhouse, V. L. and Brener, J. W., J. App. Phys., 3(), 1458 (1954).

22. Brenneman, A. E., Proc. IEEE, 5_1, No. 3, 442 (1963).

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 191

23. Young, D. R., Progr. Cryogenics, Vol. I, p. 1, K. Mendelssohn, ed.;

London, H e y w o o d and Co.

24. Crowe, J. W., IB M Journal _1, 295 (1957).

25. Burns, Jr., L. L., Leek, G. W., Alphonse, G. A. and Katz, R. W., ONR
S y m p . , 167 (1960).

26. Burns, Jr., L. L., Alphonse, G. A., and Leek, G. W., Trans. IRE on EC,
EC-10, 438 (1961).

27. Ahrons, R. W., Ph.D. Thesis, Electrical Engineering Dept., Polytechnic

Insti. of Brooklyn, (1963).

28. Sass, A. R. , Jour nal of App. Phys. 3^5, 516 (1964).

29. Parks, R. D., and Mochel, J. M., Phys. Rev. Letters, 1_1, 354 (1963).

30. Anderson, P. M. and Dayem, A. H., Phys. Rev. Letters, ^13, 195 (1964).

31. London, F., Superfluids, I, 47, J o h n W i l e y and Sons, N e w Yo r k (1950).

32. Ibid p. 73-76.

33. Anderson, P. W., Phys. Rev. Letters, _9, 309 (1962).

34. Anderson, P. W., and Kim, Y. B., Rev. Mod. Phys. 36_, 39 (1964).

35. Friedel, J., deGennes, P. G., and Matricon, J., Appl.Phys. Letters,
2, 119 (1963).

36. deGennes, P. G., and Matricon, J. Rev. Mod. Phys., J36, 45 (1964).

37. Borcherds, P. W., Gough, C. E., Vinen, W. F., and Warren, A. C., Phil.
Mag. 10, 349 (1964).

38. Bardeen, J., Phys. Rev. Letters, _13, 747 (1964).

39. Ferraro, V. C. A., "Electromagnetic Theory", p. 447, The Athlone

Press, London (1954).

40. Bateman, H., "Tables of Integral Transforms", Vol. II, p. 22, New York,
McGraw-Hill, (1954).

41. Jahnke, E. and Ende, F., "Tables of Functions", p. 191-219, New York,
Dover, (1945).

42. Onsager, L., Nuovo Cimeno , <6, Supp. 2, 249 (1949).

43. Feynman, R. P., "Progress in Low Temperature Physics" I (North Holland

Publish. Co., Amsterdam, 1955).

44. Miller, P. B., Kington, B; W., and Quinn, D. J., Rev. Mod. Phys., 3 6 ,
70 (1964).

R eproduced with permission of the copyright owner. Further reproduction prohibited without permission.
 192

45. Matricon, J. Proc. of the IX Inter. Conf. on Low Temp. Phys., Columbus,
Ohio (1964).

46. Caroli, C., deGennes, P. G., and Matricon, J. Phys. Letters, _9, 307,
(1964).

47. Ro me A i r Developme nt Center Technical Documentar y R e p o r t No. RADC-TDR-

63-351., p. 56, O c t o b e r (1963).

48. Ginsburg, V. L., Doklady 3, 102 (1958).

49. Douglass, D. H- Jr., Phys. Rev., 1 2 4 , 735 (1961).

50. Miller, J. C. and Cosentino, L. S., RCA Laboratories, personal

communication.

51. Mailer, V. A. J., ICT, Stevenage, Hertfordshire, England, personal

communication.

52. Suhl, H., Phys. Rev. Letters, _14, 226 (1965).

53. Stephen, M. J.'and Bardeen, J. Phys. Rev. Letters, _14, 112 (1965).

54. Mezrich, R. S. R C A Laboratories, personal communication.

55. Nosker, R. W., R C A Laboratories, personal communication.

R eproduced with permission of the copyright owner. Further reproduction prohibited w ithout permission.