Professional Documents
Culture Documents
A Primer
by
Helmut Es hrig
Dire tor of the Institute for Solid State and Materials Resear h and
to Musik Kaganov
2
Contents
1 INTRODUCTION 4
2 PHENOMENA, LONDON THEORY 5
2.1 Phenomena . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2 London theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3 Gauge symmetry, London gauge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3 THE THERMODYNAMICS OF THE PHASE TRANSITION 13
3.1 The Free Energy . . . . . . . . . .. .. . .. .. .. . .. .. .. . .. .. .. . .. .. 13
3.2 The Free Enthalpy . . . . . . . . .. .. . .. .. .. . .. .. .. . .. .. .. . .. .. 14
3.3 The thermodynami
riti
al eld .. .. . .. .. .. . .. .. .. . .. .. .. . .. .. 15
3.4 Heat
apa
ity jump . . . . . . . . .. .. . .. .. .. . .. .. .. . .. .. .. . .. .. 16
4 THE GINSBURG-LANDAU THEORY; TYPES OF SUPERCONDUCTORS 17
4.1 The Landau theory . . . . . . . . . . .. . .. .. .. . .. .. .. . .. .. .. . .. .. 17
4.2 The Ginsburg-Landau equations . . .. . .. .. .. . .. .. .. . .. .. .. . .. .. 18
4.3 The Ginsburg-Landau parameter . . .. . .. .. .. . .. .. .. . .. .. .. . .. .. 20
4.4 The phase boundary . . . . . . . . . .. . .. .. .. . .. .. .. . .. .. .. . .. .. 21
4.5 The energy of the phase boundary . .. . .. .. .. . .. .. .. . .. .. .. . .. .. 23
5 INTERMEDIATE STATE, MIXED STATE 25
5.1 The intermediate state of a type I super
ondu
tor . . . . . . . . . . . . . . . . . . . . 25
5.2 Mixed state of a type II super
ondu
tor . . . . . . . . . . . . . . . . . . . . . . . . . . 27
5.3 The
ux line in a type II super
ondu
tor . . . . . . . . . . . . . . . . . . . . . . . . . . 28
6 JOSEPHSON EFFECTS 32
6.1 The d.
. Josephson ee
t, quantum interferen
e . . . . . . . . . . . . . . . . . . . . . . 33
6.2 The a.
. Josephson ee
t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
7 MICROSCOPIC THEORY: THE FOCK SPACE 37
7.1 Slater determinants . . . . . . . . . . .. . .. .. .. . .. .. .. . .. .. .. . .. .. 37
7.2 The Fo
k spa
e . . . . . . . . . . . . .. . .. .. .. . .. .. .. . .. .. .. . .. .. 38
7.3 O
upation number representation . .. . .. .. .. . .. .. .. . .. .. .. . .. .. 39
7.4 Field operators . . . . . . . . . . . . .. . .. .. .. . .. .. .. . .. .. .. . .. .. 40
8 MICROSCOPIC THEORY: THE BCS MODEL 42
8.1 The normal Fermi liquid as a quasi-parti
le gas . .. . .. .. .. . .. .. .. . .. .. 42
8.2 The Cooper problem . . . . . . . . . . . . . . . . .. . .. .. .. . .. .. .. . .. .. 44
8.3 The BCS Hamiltonian . . . . . . . . . . . . . . . .. . .. .. .. . .. .. .. . .. .. 46
8.4 The Bogoliubov-Valatin transformation . . . . . .. . .. .. .. . .. .. .. . .. .. 47
9 MICROSCOPIC THEORY: PAIR STATES 51
9.1 The BCS ground state . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
9.2 The pair fun
tion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
9.3 Non-zero temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
10 MICROSCOPIC THEORY: COHERENCE FACTORS 55
10.1 The thermodynami
state . . . . . . . . . .. .. .. . .. .. .. . .. .. .. . .. .. 55
10.2 The
harge and spin moment densities . . .. .. .. . .. .. .. . .. .. .. . .. .. 55
10.3 Ultrasoni
attenuation . . . . . . . . . . . .. .. .. . .. .. .. . .. .. .. . .. .. 56
10.4 The spin sus
eptibility . . . . . . . . . . . .. .. .. . .. .. .. . .. .. .. . .. .. 57
3
1 INTRODUCTION
These le
ture notes introdu
e into the phenomenologi
al and qualitative theory of super
ondu
tiv-
ity. Nowhere any spe
i
assumption on the mi
ros
opi
me
hanism of super
ondu
tivity is made
although on a few o
asions ele
tron-phonon intera
tion is mentioned as an example. The theoreti
al
presuppositions are ex
lusively guided by phenomena and kept to a minimum in order to arrive at
results in a reasonably simple manner.
At present there are indi
ations of non-phonon me
hanisms of super
ondu
tivity, yet there is no
hard proof up to now. The whole of this treatise would apply to any me
hanism, possibly with
indi
ated modi
ations, for instan
e a symmetry of the order parameter dierent from isotropy whi
h
has been
hosen for the sake of simpli
ity.
This is a primer. For ea
h
onsidered phenomenon, only the simplest
ase is treated. Referen
es
are given basi
ally to the most important seminal original papers. Despite the above mentioned
stri
t phenomenologi
al approa
h the te
hni
al presentation is standard throughout, so that it readily
ompares to the existing literature.1
More advan
ed theoreti
al tools as eld quantization and the quasi-parti
le
on
ept are introdu
ed
to the needed level before they are used. Basi
notions of Quantum Theory and of Thermodynami
s
(as well as of Statisti
al Physi
s in a few o
asions) are presupposed as known.
In Chapter 2, after a short enumeration of the essential phenomena of super
ondu
tivity, the
London theory is derived from the sole assumption that the super
urrent as an ele
tri
al
urrent is
a property of the quantum ground state. Thermoele
tri
s, ele
trodynami
s and gauge properties are
dis
ussed.
With the help of simple thermodynami
relations, the
ondensation energy, the thermodynami
riti
al eld and the spe
i
heat are
onsidered in Chapter 3.
In Chapter 4, the Ginsburg-Landau theory is introdu
ed for spatially inhomogeneous situations,
leading to the
lassi
ation of all super
ondu
tors into types I and II. The simplest phase diagram of
an isotropi
type II super
ondu
tor is obtained in Chapter 5.
The Josephson ee
ts are qualitatively
onsidered on the basis of the Ginsburg-Landau theory in
Chapter 6. Both, d.
. and a.
. ee
ts are treated.
The remaining four
hapters are devoted to the simplest phenomenologi
al weak
oupling theory
of super
ondu
tivity on a mi
ros
opi
level, the BCS theory whi
h gave the rst quantum theoreti
al
understanding of super
ondu
tivity, 46 years after the experimental dis
overy of the phenomenon. For
this purpose, in Chapter 7 the Fo
k spa
e and the
on
ept of eld quantization is introdu
ed. Then,
in Chapter 8, the Cooper theorem and the BCS model are treated with o
upation number operators
of quasi-parti
le states whi
h latter are introdu
ed as a working approximation in Solid State Physi
s.
The nature of the
harged bosoni
ondensate, phenomenologi
ally introdu
ed in Chapter 2, is derived
in Chapter 9 as the
ondensate of Cooper pairs. The ex
itation gap as a fun
tion of temperature is
here the essential result. The treatise is
losed with a
onsideration of basi
examples of the important
notion of
oheren
e fa
tors.
By spe
ifying more details as lower point symmetry, real stru
ture features of the solid (for instan
e
ausing pinning of vortex lines) and many more, a lot of additional theoreti
al
onsiderations would
be possible without spe
ifying the mi
ros
opi
me
hanism of the attra
tive intera
tion leading to
super
ondu
tivity. However, these are just the notes of a one-term two-hours le
ture to introdu
e
into the spirit of this kind of theoreti
al approa
h, not only addressing theorists. In our days of lively
spe
ulations on possible
auses of super
ondu
tivity it should provide the new
ommer to the eld
(again not just the theorist) with a safe ground to start out.
1 Two
lassi
s are re
ommended for more details: J. R. S
hrieer, Theory of Super
ondu
tivity, Benjamin, New York,
1964, and R. D. Parks (ed.), Super
ondu
tivity, vol. I and II, Dekker, New York, 1969.
4
Figure 1: Resistan
e in ohms of a spe
imen of mer
ury versus absolute temperature. This plot by
Kamerlingh Onnes marked the dis
overy of super
ondu
tivity. (Taken from: Ch. Kittel, Introdu
tion
to Solid State Physi
s, Wiley, New York, 1986, Chap. 12.)
Helium was rst liqueed by Kammerling Onnes at Leiden in 1908. By exhausting the helium vapor
above the liquid the temperature
ould soon be lowered down to 1.5K.
Shortly afterwards, in the year 1911, it was found in the same laboratory1 that in pure mer
ury
the ele
tri
al resistan
e disappeared abruptly below a
riti
al temperature, T
= 4.2K.
Deliberately in
reasing ele
tron s
attering by making the mer
ury impure did not ae
t the phe-
nomenon. Shortly thereafter, the same ee
t was found in indium (3.4K), tin (3.72K) and in lead
(7.19K). In 1930, super
ondu
tivity was found in niobium (T
= 9.2K) and in 1940 in the metalli
ompound NbN (T
= 17.3K), and this remained the highest T
until the 50's, when super
ondu
tivity
in the A15
ompounds was found and higher T
-values appeared up to T
= 23.2K in Nb3Ge, in 1973.
These materials were all normal metals and more or less good
ondu
tors.
In 1964, Marvin L. Cohen made theoreti
al predi
tions of T
-values as high as 0.1K for
ertain
doped semi
ondu
tors, and in the same year and the following years, super
ondu
tivity was found in
GeTe, SnTe (T
0.1K, ne 1021
m 3 ) and in SrTiO3 (T
= 0.38K at ne 1021
m 3, T
0.1K at ne 1018
m 3).
In the early 80's, super
ondu
tivity was found in several
ondu
ting polymers as well as in \heavy
fermion systems" like UBe13 (T
1K in both
ases). The year 2000 Nobel pri
e in Chemistry was
dedi
ated tho the predi
tion and realization of
ondu
ting polymers (syntheti
metals) in the late
70's.
1 H. K. Onnes, Commun. Phys. Lab. Univ. Leiden, No124
(1911); H. K. Onnes, Akad. van Wetens
happen
(Amsterdam) 14, 818 (1911).
5
September, 1993
(under pressure)
160
140
Liquid
CF4 Hg−1223
[ August, 1993
(under pressure)
April, 1993
September, 1992
Tl−2223 [ February, 1988
120
Bi−2223 January, 1988
Tc(K)
100
Shuttle Y−123 January, 1987
Liquid
80 N2
December, 1986
60 (under pressure)
40
(La−Ba)−214 [ December, 1986
(under pressure)
April, 1986
Liquid Nb3Ge
Nb3Sn
H2 NbN
20 NbC
Liquid Pb Nb−Al−Ge
V3Si
He
Hg Nb
Figure 2: The evolution of T
with time (from C. W. Chu, Super
ondu
tivity Above 90K and Beyond
in: B. Batlogg, C. W. Chu, W. K. Chu D. U. Gubser and K. A. Muller (eds.) Pro
. HTS Workshop
on Physi
s, Materials and Appli
ations, World S
ienti
, Singapore, 1996.).
In 1986, Georg Bednorz and Alex Muller found super
ondu
tivity in (La,Sr)2CuO4 with T
= 36K,
an in
redible new re
ord.1
Within months, T
-values in
uprates were shooting up, and the re
ord is now at T
135K.
2.1 Phenomena
Zero resistan
e2 No resistan
e is dete
table even for high s
attering rates of
ondu
tion ele
trons.
Persistent
urrents magneti
ally indu
ed in a
oil of Nb0:75 Zr0:25 and wat
hed with NMR yielded an
estimate of the de
ay time greater than 105 years! (From theoreti
al estimates the de
ay time may
10
be as large as 10 years!)
10
Absen
e of thermoele
tri
ee
ts3 No Seebe
k voltage, no Peltier heat, no Thomson heat is
dete
table (see next se
tion).
Ideal diamagnetism m = 1. Weak magneti
elds are
ompletely s
reened away from the bulk
of a super
ondu
tor.
Meissner ee
t4 If a super
ondu
tor is
ooled down in the presen
e of a weak magneti
eld, below
T
the eld is
ompletely expelled from the bulk of the super
ondu
tor.
Flux quantization5 The magneti
ux through a super
ondu
ting ring is quantized and
onstant
in time. This phenomenon was theoreti
ally predi
ted by F. London in 1950 and experimentally
veried 1961.
1 J. G. Bednorz and K. A. M uller, Z. Phys. B64, 189 (1986).
2 J. File and R. G. Mills, Phys. Rev. Lett. 10, 93 (1963).
3 W. Meissner, Z. Ges. K alteindustrie 34, 197 (1927).
4 W. Meissner and R. O
hsenfeld, Naturwiss. 21, 787 (1933).
5 B. S. Deaver and W. M. Fairbank, Phys. Rev. Lett. 7, 43 (1961); R. Doll and M. Nabauer, Phys. Rev. Lett. 7,
51 (1961).
6
2.2 London theory1
Phenomena (a) and (b)
learly indi
ate that the super
urrent (at T = 0) is a property of the quantum
ground state:
There must be an ele
tri
ally
harged (
harge quantum q), hen
e
omplex bosoni
eld
whi
h
ondenses in the ground state into a ma
ros
opi
amplitude:
nB = j j2 ; (1)
where nB means the bosoni
density, and is the
orresponding eld amplitude.
Sin
e the eld is ele
tri
ally
harged, it is subje
t to ele
tromagneti
elds (E ; B) whi
h are usually
des
ribed by potentials (U; A):
A U
E= ; (2a)
t r
B= A: (2b)
r
The eld amplitude should obey a S
hrodinger equation
1 ~ qA2 + qU = E B ; (3)
2mB i r
where the energy is reasonably measured from the
hemi
al potential B of the boson eld, sin
e what
is measured in a voltmeter is rather the ele
tro
hemi
al potential
= B + qU (4)
than the external potential U , or the ee
tive ele
tri
eld
A 1
Ee = : (5)
t q r
As usual in Quantum Me
hani
s, i~= r is the
anoni
al momentum and ( i~= r qA) = p^m is
the me
hani
al momentum.
The super
urrent density is then
q 2
js = q
pm
n
mB B
= mB
< p^m = 2iqm~ r r mq A: (6)
B B
It
onsists as usual of a `paramagneti
urrent' (rst term) and a `diamagneti
urrent' (se
ond term).2
In a homogeneous super
ondu
tor, where nB =
onst., we may write
(r; t) = pnB ei(r;t); (7)
and have
js = ~q r A; = nmBq2 : (8)
B
Sin
e in the ground state E = , and E = i~ =t, we also have
= :
~
t
(9)
The London theory derives from (8) and (9). It is valid in the London limit, where nB =
onst. in
spa
e
an be assumed.
1 F. London and H. London, Pro
. Roy. So
. A149, 71 (1935); F. London, Pro
. Roy. So
. A152, 24 (1935); F.
London, Super
uids, Wiley, London, 1950.
2 These are formal names: sin
e the splitting into the two
urrent
ontributions depends on the gauge, it has no
deeper physi
al meaning. Physi
ally, paramagneti
means a positive response on an external magneti
eld (enhan
ing
the eld inside the material) and diamagneti
means a negative response.
7
The time derivative of (8) yields with (9)
js A 1 ;
t
= t q r
or
js
t
= Ee (10)
This is the rst London equation:
A super
urrent is freely a
elerated by an applied voltage, or, in a bulk super
ondu
tor
with no super
urrent or with a stationary super
urrent there is no ee
tive ele
tri
eld
(
onstant ele
tro
hemi
al potential).
The rst London equation yields the absen
e of thermoele
tri
ee
ts, if the ele
tro
hemi
al poten-
tials of
ondu
tion ele
trons, el, and of the super
urrent, , are
oupled. The thermoele
tri
ee
ts
are sket
hy illustrated in Fig. 3. The rst London equation
auses the ele
tro
hemi
al potential of
the super
urrent
arrying eld to be
onstant in every stationary situation. If the super
urrent
ar-
rying eld rea
ts with the
ondu
tion ele
tron eld with n ele
trons forming a eld quantum with
harge q, then the ele
tro
hemi
al potentials must be related as nel = . Hen
e the ele
tro
hemi
al
potential of the
ondu
tion ele
trons must also be
onstant: no thermopower (Seebe
k voltage) may
develop in a super
ondu
tor. The thermoele
tri
urrent
owing due to the temperature dieren
e
is
an
eled by a ba
k
owing super
urrent, with a
ontinuous transformation of
ondu
tion ele
trons
into super
urrent density at the one end of the sample and a ba
k transformation at the other end.
If a loop of two dierent normal
ondu
tors is formed with the jun
tions kept at dierent tem-
peratures, then a thermoele
tri
urrent develops together with a dieren
e of the ele
tro
hemi
al
potentials of the two jun
tions, and several forms of heat are produ
ed, everything depending on the
ombination of the two metals. If there is no temperature dieren
e at the beginning, but a
urrent is
maintained in the ring (by inserting a power supply into one of the metal halfs), then a temperature
dieren
e between the jun
tions will develop. This is how a Peltier
ooler works. In a loop of two
super
ondu
tors non of those phenomena
an appear sin
e a dieren
e of ele
tro
hemi
al potentials
annot be maintained. Every normal
urrent is lo
ally short-
ir
uited by super
urrents.
If, however, a normal metal A is
ombined with a super
ondu
tor B in a loop, a thermoele
tri
urrent will
ow in the normal half without developing an ele
tro
hemi
al potential dieren
e of the
jun
tions be
ause of the presen
e of the super
ondu
tor on the other side. This yields a dire
t absolute
measurement of the thermoele
tri
oeÆ
ients of a single material A.
The
url of Eq. (8) yields (with r r = 0)
r
js = B : (11)
This is the se
ond London equation. It yields the ideal diamagnetism, the Meissner ee
t, and the
ux quantization.
Take the
url of Maxwell's equation (Ampere's law) and
onsider r r B = r Br r22 B:
B
r
B = 0 js + j ; r
= 0; (12)
r
r
B = 0
r
js + j ;
2
r2
B = 0 js + j ;
r
2 0
r2
B = B 0 r j :
8
normal
ondu
tor super
ondu
tor
T1 < T2
T1 < T2 super
ondu
tor
thermoele
tri
urrent
normal
ondu
tor 2;el n1;el = j T n2;el =
1;el
ne ! q super
urrent j q ! ne
Seebe
k voltage
2 1 T 2 T 1
Sin
e j (t) =
onst.
) js (t) =
onst.
fT1 ( 1;el ) fT2 ( 2;el ) ) Ee = =q = 0; = 0
The dieren
e of the Fermi distribution due to the rst London equation.
fun
tions fT in
onne
tion with a
non-
onstant density of states results in
a dieren
e of ele
tro
hemi
al potentials
due to the detailed balan
e of
urrents.
T1 < T2 T1 < T2
Peltier heat Peltier heat
BA(T1; J ) AB (T2 ; J )
B B
Ohm's heat +
Thomson heat
B ( T; J ) A J
normal
ondu
tor
T1 < T2
super
ondu
tor
Js
B
Figure 3: Thermoele
tri
phenomena in normal
ondu
tors and super
ondu
tors.
9
external eld B0
z
js B L
1111
0000 x
y
e z=L
z super
ondu
tor
Figure 4: Penetration of an external magneti
eld into a super
ondu
tor.
If j = 0 or r j = 0 for the normal
urrent inside the super
ondu
tor, then
s
2 B = r mB
2 B = 2 ; L = (13)
r L 0n q2 B 0
with solutions
B = B0 e nr=L ; n2 = 1; n B0 = 0 (14)
several of whi
h with appropriate unit ve
tors n may be superimposed to fulll boundary
onditions.
L is London's penetration depth.
Any external eld B is s
reened to zero inside a bulk super
ondu
ting state within a surfa
e layer
of thi
kness L. It is important that (11) does not
ontain time derivatives of the eld but the eld
B itself: If a metal in an applied eld B0 is
ooled down below T
, the eld is expelled.
Consider a super
ondu
ting
ring with magneti
ux pass-
ing through it (Fig. 5). Be
ause of B
(14) and (12), js = 0 deep inside
the ring on the
ontour C . Hen
e,
from (10), Ee = E = 0 there.
From Faraday's law, (= r)E =
B,
d d
Z I
=
dt dt A
B dS = E dl = 0;
C
(15) C
where A is a surfa
e with bound- d
ary C , and is the magneti
ux
through A. Figure 5: Flux through a super
ondu
ting ring.
Even if the super
urrent in a
surfa
e layer of the ring is
hang-
ing with time (for instan
e, if an
applied magneti
eld is
hanging with time), the
ux is not:
The
ux through a super
ondu
ting ring is trapped.
Integrate Eq. (8) along the
ontour C :
I I
~
A + js dl = dl:
C q C r
10
The integral on the right hand side is the total
hange of the phase of the wavefun
tion (7) around
the
ontour, whi
h must be an integer multiple of 2 sin
e the wavefun
tion itself must be unique.
Hen
e, I
~
A + js dl = 2n: (16)
C q
The left hand integral has been named the
uxoid by F. London. In the situation of our ring we nd
= ~q 2n: (17)
By dire
tly measuring the
ux quantum 0 the absolute value of the super
ondu
ting
harge was
measured:
jqj = 2e; 0 = 2he : (18)
(The sign of the
ux quantum may be dened arbitrarily; e is the proton
harge.)
If the super
urrent js along the
ontour C is non-zero, then the
ux is not quantized any more,
the
uxoid (16), however, is always quantized.
In order to determine the sign of q,
onsider a super
ondu
ting sample whi
h rotates with the
angular velo
ity !. Sin
e the sample is neutral, its super
ondu
ting
harge density qnB is neutralized
by the
harge density qnB of the remainder of the material. Ampere's law (in the absen
e of a
normal
urrent density j inside the sample) yields now
B = 0 js qnB v ;
r
where v = ! r is the lo
al velo
ity of the sample, and js is the super
urrent with respe
t to the
rest
oordinates. Taking again the
url and
onsidering
v = r ! r = ! rr
! r = 3! ! = 2!
r r
leads to
2
B = 0 js 20qnB !:
r2 r
We dene the London eld
BL 22L0qnB ! = 2mq B ! (19)
and
onsider the se
ond London equation (11) to obtain
2 B BL
B= : (20)
r2 2L
Deep inside a rotating super
ondu
tor the magneti
eld is not zero but equal to the homogeneous
London eld.
Independent measurements of the
ux quantum and the London eld result in
q = 2e; mB = 2me : (21)
The bosoni
eld is
omposed of pairs of ele
trons.
11
2.3 Gauge symmetry, London gauge
If (r; t) is an arbitrary dierentiable single-valued fun
tion, then the ele
tromagneti
eld (2) is
invariant under the gauge transformation
A ! A+ ;
r
U ! U t
: (22a)
Sin
e potentials in ele
trodynami
s
an only indire
tly be measured through elds, ele
trodynami
s
is symmetri
with respe
t to gauge transformations (22a).
Eqs. (8, 9), and hen
e the London theory are
ovariant under lo
al gauge transformations, if (22a)
is supplemented by
!
2e ;
~
! t
: (22b)
From (8), the super
urrent js is still gauge invariant, and so are the ele
tromagneti
properties of a
super
ondu
tor. However, the ele
tro
hemi
al potential is dire
tly observable in thermodynami
s
by making
onta
t to a bath. The thermodynami
super
ondu
ting state breaks gauge symmetry.
For theoreti
al
onsiderations a spe
ial gauge is often advantageous. The London gauge
hooses
in (22b) su
h that the phase 0: Then, from (8),
js = A; (23)
whi
h is
onvenient for
omputing patterns of super
urrents and elds.
12
3 THE THERMODYNAMICS OF THE
PHASE TRANSITION
1
Up to here we
onsidered super
ondu
tivity as a property of a bosoni
ondensate. From experiment
we know, that the
onsidered phenomena are present up to the
riti
al temperature, T
, of the tran-
sition from the super
ondu
ting state, indexed by s, into the normal
ondu
ting state, indexed by
n, as temperature rises. The parameters of the theory, nB and L , are to be expe
ted temperature
dependent: nB must vanish at T
.
In this and the next
hapters we
onsider the vi
inity of the phase transition, T T
T
.
13
3.2 The Free Enthalpy
The relations between the Free Energy F and the Free Enthalpy (Gibbs Free Energy) G at B = 0
and B 6= 0 read
Fs (T; V; 0) = Gs (T; p(T; V; 0); 0) p(T; V; 0)V
and
V B2
Fs (T; V; 0) + = Fs (T; V; B ) =
20
= Gs (T; p(T; V; B ); B ) p(T; V; B )V =
B2 2
= Gs (T; p(T; V; 0)
20 ; B) p(T; V; 0)V + V2B :
0
These relations
ombine to
B2
Gs (T; p(T; V; 0); 0) = Gs (T; p(T; V; 0)
20 ; B);
or
B2
Gs (T; p; B ) = Gs (T; p +
20 ; 0): (32)
In a
ord with (31), the ee
t of an external magneti
eld B on the Free Enthalpy is a redu
tion of
the pressure exerted on the surroundings, by B2=20. In the normal state, from (29),
Gn (T; p; B ) = Gn (T; p; 0): (33)
The
riti
al temperature T
(p; B) is given by
B2
Gs (T
; p +
20 ; 0) = Gn(T
; p; 0): (34a)
Likewise B
(T; p) from
B
2
Gs (T; p +
20 ; 0) = Gn(T; p; 0): (34b)
B (T )
T
(B )
B
T
T
Figure 6: The
riti
al temperature as a fun
tion of the applied magneti
eld and the thermodynami
riti
al eld as a fun
tion of temperature.
14
3.3 The thermodynami
riti
al eld
The Free Enthalpy dieren
e between the normal and super
ondu
ting states is usually small, so that
at T < T
(B = 0) the thermodynami
riti
al eld B
(T ) for whi
h (34) holds is also small. Taylor
expansion of the left hand side of (34) yields
B
2 Gs B
2
Gn (T; p) = Gs (T; p) + = Gs (T; p) +
20 p 20 V (T; p; B = 0): (35)
Experiment shows that at B = 0 the phase transition is se
ond order,
Gn (T; p) Gs (T; p) = a T
(p) T 2 :
(36)
Hen
e,
B
(T; p) = b T
(p) T ; (37)
p
where a is a
onstant, and b = 20a=V . T
(p) is meant for B = 0.
0 M = V Bm B (T ) B
B
=0 0 G normal state
B T
Meissner
b(T
(p) T ) ee
t
B
(T )
P m = 1
T (p) T
FIG. 7: The thermodynami riti al eld. FIG. 8: The magnetization urve of a super ondu tor.
We
onsider all thermodynami
parameters T; p; B at the phase transition point P of Fig. 7. From
(32),
Gs B2
Ss (T; p; B ) = = Ss (T; p +
T 20 ; 0);
Gs B2
Vs (T; p; B ) = = Vs (T; p +
p 20 ; 0): (38)
Dierentiating (34b) with respe
t to T yields, with (38),
B 2 (T; p)
Gs (T; p +
; 0) = Gn (T; p; 0 or B
);
T 20 T
V (T; p; B
) 2
Ss (T; p; B
) + s
20 T B
(T; p) = Sn(T; p; B
);
0
(T; p) :
S (T; p; B
) = Ss(T; p; B
) Sn(T; p; B
) = Vs (T;p; B
) B
(T; p) B
T (39)
15
A
ording to (37) this dieren
e is non-zero for B
6= 0 (T < T
(p)): For B 6= 0 the phase transition
is rst order with a latent heat
Q = T S (T; p; B
): (40)
For T ! 0, Nernst's theorem demands Ss = Sn = 0, and hen
e
lim B
(T; p) = 0:
T !0 T
(41)
3.4 Heat
apa
ity jump
For B 0, T T
(p) we
an use (35). Applying T 2=T 2 yields
2 T V (T; p) 2 2
Cp = Cp;s Cp;n = T
T 2
Gs (T; p) G n (T; p) = 20 T 2 B
(T; p): (42)
The thermal expansion V=T gives a small
ontribution whi
h has been negle
ted. With
!2
2 2 B B
2
T 2 B
= 2B
T T
=2 T
+ 2B
TB2
we nd " !2 #
TV B
2
Cp = 0 T
+ B
TB2
(43)
For T ! T
(p), B
! 0 the jump in the spe
i
heat is
!2
T
V B
Cp = 0 T
= T
V b2: (44)
0
It is given by the slope of B
(T ) at T
(p).
Cp
S
B=0 B=0
Cp
Sn
Ss
T (p) T T (p) T
FIG. 9: The entropy of a super ondu tor. FIG. 10: The heat apa ity of a super ondu tor.
16
4 THE GINSBURG-LANDAU THEORY;
1
TYPES OF SUPERCONDUCTORS
A
ording to the Landau theory of se
ond order phase transitions with symmetry redu
tion2 there is a
thermodynami
quantity,
alled an order parameter, whi
h is zero in the symmetri
(high temperature)
phase, and be
omes
ontinuously non-zero in the less symmetri
phase.
4.1 The Landau theory
The quantity whi
h be
omes non-zero in the super
ondu
ting state is
nB = j j2 : (45)
For nB > 0, the ele
tro
hemi
al potential has a
ertain value whi
h breaks the global gauge sym-
metry by xing the time-derivative of the phase of (
f. (22b)). A
ording to the Landau theory,
the Free Energy is the minimum of a \Free Energy fun
tion" of the order parameter with respe
t to
variations of the latter:
F (T; V ) = min F (T; V; j j2): (46)
F T > T
(t > 0)
T = T
(t = 0)
T < T
(t < 0)
0110 j j2
n1B;
0min
Figure 11: The Free Energy fun
tion.
17
Sin
e jtj 1, we put
A(t) tV; B (t) V: (49)
Then we have
Fn (t) = Fn (t) for t 0; (50)
and
1 F = t + j j2 = 0; that is,
V j j2
2 t2
j j2 = t
; Fs (t) = Fn (t)
2 V for t < 0: (51)
Re
alling that small
hanges in the Free Energy and Free Enthalpy are equal and
omparing to
(35) yields
2 t2 B
2
r
0
2 = 20 =) B
(t) = jtj : (52)
From (43), !2
T
V B
2
Cp = 0 T
t
= TV (53)
follows. While Cp
an be measured, this is not always the
ase for the thermodynami
riti
al eld,
B
, as we will later see.
Eqs. (51) and (52) may be rewritten as
2 t2 0
nB (t) = jtj; B
2 (t) = ;
hen
e,
B
2 (t) B
2 (t)
= 2 ; = : (54)
0 nB (t) 0 jtjnB (t)
Sin
e a
ording to (37) B
t, it follows
nB t: (55)
The bosoni
density tends to zero linearly in T
T.
where also (21) was
onsidered. The rst
orre
tion term is the eld energy of the eld Bm
reated
by , in
luding the stray eld outside of the volume V while 6= 0 inside V only. A is the ve
tor
potential of the total eld a
ting on :
r
A = B + Bm : (56b)
18
The Free Energy is obtained by minimizing (56a) with respe
t to (r) and (r). To prepare for a
variation of , the se
ond integral in (56a) is integrated by parts:
" #" #
Z
d3 r
2ie
+ A 2 ie
A =
V r ~ r ~
2
2ie 2ie
Z Z
= d3 r + A + d2 n + A : (56
)
V r ~ V r ~
From the rst integral on the right we see that (56a) indeed
orresponds to (3). The preferen
e of the
writing in (56a) derives from that kineti
energy expression being manifestly positive denite in any
partial volume.
Now, the variation ! + Æ yields
( )
~2 ie 2
!
Z
3 2
0 = ÆF = d rÆ 4m r + ~ A + t + j j + 2
V
~2
Z
2
+ d nÆ 4m r + ~ A : 2ie
V
F is stationary for any variation Æ (r), if
1 ~ + 2eA2 jtj + j j2 = 0 (57)
4m i r
and
~
+ 2eA = 0:
n
i r
(58)
The
onne
tion of with Bm must be that of Ampere's law: (= r) Bm = 0js with js given
by (6). Sin
e in thermodynami
equilibrium there are no
urrents besides js in the super
ondu
tor,
(= r) B = 0 there. Hen
e, we also have
r
Btot = 0 js ; Btot = B + Bm =
r
A;
ie~
2e2 (59)
js =
2m r r m
A :
It is interesting to see that (59) is also obtained from (56a), if , and A are varied independently:
The variation of A on the left hand side of (56
) yields
2ie Z d3rÆA 2ie A + 2ie A :
~ V r ~ r ~
With ÆBm = (= r) ÆA the variation of the rst integral of (56a) yields
Z 1 Z 1 Z 1
Æ d3 rBm
2 =2 d3 rÆBm Bm = 2 d3 r Æ A Bm =
r
z }| {
Z 1 1 Z
= 2 d3r r ÆA Bm = 2 d3 rÆA r Bm =
Z
= 2 d3 rÆA r Btot + : (60)
V
In the fourth equality an integration per parts was performed, and a (b
) = b (a
) was used.
(The over bra
e indi
ates the range of the dierential operator.) Finally, the integral over the innite
19
spa
e is split into an integral over the super
ondu
tor (volume V ), where (= r)Bm = (= r)Btot,
and the integral over the volume outside of the super
ondu
tor, indi
ated by dots, sin
e we do not
need it. Now, after adding the prefa
tors from (56a) we see that stationarity of (56a) with respe
t to
a variation ÆA inside the volume V again leeds to (59).
This situation is no a
ident. From a more general point of view the Ginsburg-Landau fun
tional
(56a) may be
onsidered as an ee
tive Hamiltonian for the
u
tuations of the elds and A near
the phase transition.1 This is pre
isely the meaning of relating (56a) to (3).
Eqs. (57) and (59) form the
omplete system of the Ginsburg-Landau equations.
The boundary
ondition (58)
omes about by the spe
ial writing of (56a) without additional surfa
e
terms. This is
orre
t for a boundary super
ondu
tor/va
uum or super
ondu
tor/semi
ondu
tor. A
areful analysis on a mi
ros
opi
theory level yields the more general boundary
ondition
~ i
n
i r
+ 2eA = ;
b
(61)
where b depends on the outside material: b = 1 for va
uum or a non-metal, b = 0 for a ferromagnet,
b nite and non-zero for a normal metal.2
In all
ases, multiplying (61) by and taking the real part yields
n js = 0 (62)
as it must: there is no super
urrent passing through the surfa
e of a super
ondu
tor into the non-
super
ondu
ting volume.
Btot must be
ontinuous on the boundary be
ause, a
ording to Btot = r = 0 and (59), its
derivatives are all nite.
4.3 The Ginsburg-Landau parameter
Taking the
url of (59) yields, like in (13),
2 Btot
r2
= Btot2 ; m
2 =
m
20e2j j2 = 20e2jtj ; (63)
where (51) was taken into a
ount in the last expression. is the Ginsburg-Landau penetration depth;
it diverges at T
like jtj 1: if T
is approa
hed from below, the external eld penetrates more and
more, and eventually, at T
, the diamagnetism vanishes.
Eq. (57)
ontains a se
ond length parameter: In the absen
e of an external eld, A = 0, and for
small , j j2 jtj=, one is left with
2 2 = ~2 :
=
r2 2
;
4mjtj (64)
This equation des
ribes spatial modulations of the order parameter j j2
lose to T
. is the Ginsburg-
Landau
oheren
e length of su
h order parameter
u
tuations. It has the same temperature dependen
e
as , and their ratio, s
= =
2m2 ; (65)
~2 e2 0
is the
elebrated Ginsburg-Landau parameter.
1 L. D. Landau and E. M. Lifshits, Statisti
al Physi
s, Part I, x147, Pergamon, London, 1980.
2 P. G. De Gennes, Super
ondu
tivity in metals and alloys, New York 1966, p. 225 .
20
Introdu
tion of dimensionless quantities
x = r=;
,s
jtj
=
;
p r
2B
(t) = B jtj 20 ; (66)
.
b=B
.p
is = js 0 2B
(t) ;
.p
a=A 2B
(t)
yields the dimensionless Ginsburg-Landau equations
1
2
i x
+ a + j j2 = 0;
i
(67)
b = is ; is = 2 x a
x x
whi
h
ontain the only parameter .
4.4 The phase boundary
We
onsider a homogeneous super
ondu
tor at T . T
in an homogeneous external eld B B
(T ) in
z -dire
tion. We assume a plane phase boundary in the y z -plane so that for x ! 1 the material is
still super
ondu
ting, and the magneti
eld is expelled, but for x ! 1 the material is in the normal
state with the eld penetrating.
z p
b=0 b = 1= 2
=1 b =0
is phase
boundary
We put
= (x); bz = b(x); bx = by = 0;
21
ay = a(x); ax = az = 0; b(x) = a0 (x):
Then, the super
urrent is
ows in the y-dire
tion, and hen
e the phase of depends on y. We
onsider y = 0 and may then
hoose real. Further, by xing another gauge
onstant, we may
hoose
a(0) = 0:
Then, Eqs. (67) redu
e to
1 00 + a2 + 3 = 0; a00 = a 2 : (68)
2
Let us rst
onsider 1: For large enough negative x we have a 0 and 1. We put = 1 (x),
and get from the rst equation (68)
p
00 2 1 1 + 3 = 22; e 2x ; x . 1 :
On the other hand, for large enough positive x we have b = 1=p2; a = xp2; 1; hen
e, again from
the rst equation (68),
2 2
00 x ; x2 =2
p2
e ; x2 1:
2
1 1
p
1
e 2x
p1 b
2
p
0 x2 =2
p2
e
e 0x
1=p > 1
1= =
In the opposite
ase 1; falls o for x & 1; where b 1=p2; a x=p2; and for x 1;
00 2 x2 =2 :
22
1 1
p1 b
2
x2 =2
p2
e
1=p < 1
1=
24
5 INTERMEDIATE STATE, MIXED STATE
In Chapter 2 we
onsidered a super
ondu
tor in a suÆ
iently weak magneti
eld, B < B
; where we
found the ideal diamagnetism, the Meissner ee
t, and the
ux quantization.
In Chapter 3 we found that the dieren
e between the thermodynami
potentials in the normal
and the super
ondu
ting homogeneous phases per volume without magneti
elds may be expressed
as (
f. (35))
1 hGn(p; T ) Gs (p; T )i = 1 hFn (V; T ) Fs (V; T )i = B
2(T ) (73)
V V 20
by a thermodynami
riti
al eld B
(T ): (We negle
t here again the ee
ts of pressure or of
orre-
sponding volume
hanges on B
:)
If a magneti
eld B is applied to some volume part of a super
ondu
tor, it may be expelled
(Meissner ee
t) byR
reating an internal eld Bm = B through super
urrents, on the
ost of an
additional energy d3rBm2 =20 for the super
ondu
ting phase (
f. (56a)) and of a kineti
energy
density (~2 =4m)j(= r + 2ieA=~) j2 in the surfa
e where the super
urrents
ow. If Bm > B
;
the Free Energy of the super
ondu
ting state be
omes larger than that of the normal state in a
homogeneous situation. However, B itself may
ontain a part
reated by
urrents in another volume
of the super
ondu
tor, and phase boundary energies must also be
onsidered. There are therefore
long range intera
tions like in ferroele
tri
s and in ferromagnets, and
orresponding domain patterns
orrespond to thermodynami
stable states. The external eld B at whi
h the phase transition appears
depends on the geometry and on the phase boundary energy.
Bext
Bmax
Bext + Bm
SC
SC
n
Bmax
25
This, however,
annot be stable either: the point of Bmax = B
has now moved into the super-
ondu
tor to a point of the phase boundary between the normal and super
ondu
ting phases, whi
h
means that in the shaded normal area B < B
; this area must be
ome super
ondu
ting again (Fig. 16).
Forming of a
onvex island would
ause the same problem (Fig. 17).
Bmax
n SC
SC
26
5.2 Mixed state of a type II super
ondu
tor1
If the phase boundary energy is negative, germs of normal phase may form well below B
; at the lower
riti
al eld B
1 ; and germs of super
ondu
ting phase may form well above B
; at the upper
riti
al
eld B
2 ; in both
ases by gaining phase boundary energy.
At B / B
2; in small germs 1 and b
onst. The dimensionless Ginsburg-Landau equations
(67) may be linearized:
1 + a2 = : (74)
i x
We apply the external B-eld in z-dire
tion (Fig. 20),
b bx = by = 0; bz = b; ax = by; ay = az = 0;
and assume germ laments along the eld lines:
a = (x; y):
Then, (74) is
ast into
. " #
y 1 by2 1 2 = :
i x 2 y2
With (x;2y) =2 eipx2 (y); p=(b)2 = y0 this equation simplies
x to [ (1= )(d =dy )+ b(y y0 ) ℄ = ; or, after multiplying
with =2 and by dening u2 (y y0)2 :
" #
Figure 20: 1 d2 + 1 b2u2 = :
2 du2 2 2 (75)
This is the S
hrodinger equation for the ground state of a harmoni
os
illator with
p
! = b = =) B = 2B
:
Hen
e, p
B
2 (T ) = 2B
(T ): (76)
If a germ
lose to the surfa
e of a super
ondu
tor at y=
onst. is
onsidered, then the u-
oordinate
must be
ut at some nite value. There, the boundary
ondition (58) yields d=du = 0 (sin
e na = 0).
Therefore, instead of the boundary problem, the symmetri
ground state in a double os
illator with
a mirror plane may be
onsidered (Fig. 21).
0 (0) = 0
~!
min
u0 0
E = 0:59
2
u0 u
27
Hen
e, at
B
2
= 1:7B
2 = 2:4B
B
3 = (77)
0:59
super
ondu
tivity may set in in a surfa
e layer of thi
kness .
In a type I super
ondu
tor, B
2 < B
. However, only below B
2 germs with arbitrarily small
-values
ould form where the super
ondu
ting phase is already absolutely stable in a type I super-
ondu
tor. Here, for B
2 < B < B
germs
an only form with a non-zero minimal -value, whi
h is
inhibited pby positive surfa
e energy. In this region the super
ooled normal phase is metastable. For
> 0:59= 2; B
3 > B
; and surfa
e super
ondu
tivity may exist above B
:
Bz
j j
= j jei ; B = ez B(); 2 = x2 + y2
One
ould try to solve the Ginsburg-Landau equations for that
ase. However, there is no general
analyti
solution, and the equations are valid
lose to T
only, where j j is small. Instead we assume
1; that is, ; and
onsider only the region ; where j j =
onst. Then, from Ampere's
law and (6),
2
r
B = 0 js = 0 e~ j j2 20 e j j2 A =
m r m
0e2 j j2 ~
= 2 A = 2 20 r
1 A ;
m 2e r
hen
e,
A + 2
r
B = 20 r :
We integrate this equation along a
ir
le around the
ux line with radius and use Stokes'
theorem, H dsA = R d2n ( r A) :
Z I
d2 n B + 2 ds B = 0 : (79)
r
28
The phase must in
rease by 2 on a
ir
le around one
uxoid.
Now, we take : Then, the rst term in (79) may be negle
ted. We nd
22 dB
d
= 0 (80)
or
B () =
0 0
22 ln ; 0 : (81)
The integration
onstant 0 was
hosen su
h that (81) vanishes for & ; where a more a
urate
analysis of (79) is ne
essary to get the
orre
t asymptoti
s.
By applying Stokes' theorem also to the se
ond integral of (79), we have for all
2
Z Z
d2 n B + 2 B = d2 n B 2 B = 0 :
r r r2
Sin
e the right hand side does not
hange if we vary the area of integration,
2
B 2 B=0
r2
must hold. In
ylindri
oordinates,
2 1 2 2
r2
= 1
+ 2 2 + 2;
z
hen
e,
B + 2
1 B=0
or 1 1
B 00 + B 0 B = 0: (82)
2
This equation is of the Bessel type. Its for large de
aying solution is
0 K0 ! p 0 e = ;
B () = (83)
22 !1 82
where K0 is M
Donald's fun
tion (Hankel's fun
tion with imaginary argument), and the
oeÆ
ient
has been
hosen to meet (80) for :
The energy per length of the
ux line
onsists of eld energy and kineti
energy of the super
ur-
rent:
B 2 mB
Z
= d2 r + 2 nB v2 ; js = 2enB v; nB = j j2;
Z
2
2
0
b m 2
= d2 r
20 4e2nB s =
+ j
Z 2 2
2 B m
= dr
20 + 40e2j j2 rB =
2
2
Z
B
= d2 r
2 +
0 20 r
B r B =
= 1 Z d2 rB B + 2 B
20 r
!
r
2
I I
ds B B :
20 !1 r
29
The rst integrand was shown to vanish for ; and the last
ontour integral vanishes for ! 1.
We negle
t the
ontribution from . ; and nd with (80))
2 dB B ( ) 0 B (0)
2B 0 (84)
20 d 20 20 :
With (81), the nal result with logarithmi
a
ura
y (ln(=) 1) is
20
20 ln :
ln
402 402 = (85)
This result also proves that the total energy is minimum for
ux lines
ontaining one
uxoid ea
h:
For a
ux line
ontaining n
uxoids the energy would be n2 while for n
ux lines it would only be
n ( > 0).
In this analysis, B was the eld
reated by the super
urrent around the vertex line. Its intera
tion
energy per length with a homogeneous external eld in the same dire
tion is
Z
BB B
d2 r ext = 0 ext : (86)
0 0
(85) and (86) are equal at the lower
riti
al eld Bext = B
1:
ln
B
1 = 0 2 ln = B
p ; 1: (87)
4 2
The phase diagram of a type II super
ondu
tor is shown in Fig. 23. (There might be another phase with
spatially modulated order parameter under
ertain
onditions, theoreti
ally predi
ted independently
by Fulde and Ferell and by Larkin and Ovshinnikov; this FFLO phase has not yet been
learly observed
experimentally.)
B
surfa
e B
3 n
s
FFLO? B
2
mixed
phase
B
B
1
s
T
T
Figure 23: The phase diagram of a type II super
ondu
tor.
A more detailed numeri
al
al
ulation shows that in an isotropi
material the energy is minimum
for a regular triangular latti
e of the
ux lines2in the plane perpendi
ular to them. From (87), at
B p= B
1 the density of
ux lines is ln =4 ; that is, the latti
e
onstant a1 is obtained from
a21 3=2 = 42 = ln : s
a1 = p
8 (88)
3ln & :
30
The lines (of thi
kness ) indeed form nearly individually (Fig. 24). Sin
e B
2 = B
122= ln ; the
latti
e
onstant a2 at B
2 is
r s s
a2 = a 1
ln 4 4
22 = p3 = p3 & : (89)
The
ores of the
ux lines (of thi
kness ) tou
h ea
h other while the eld is already quite homoge-
neous (Fig. 25). Sin
e j j 1 in the
ore, the Ginsburg-Landau equations apply, and j j may rise
ontinuously from zero: the phase transition at B
2 is se
ond order.
B
j j
solution
B j j of (4.3)
a1 & a2 &
FIG. 24: Mixed phase for Bext B 1 . FIG. 25: Mixed phase for Bext B 2 .
For a long
ylindri
rod the stray eld
reated by super
urrents outside of the rod may be negle
ted,
and one may express the eld inside the rod as
B = Bext 0 M
by a magnetization density M: The
hange in Free Energy at xed T and V by tuning up the external
magneti
eld is
dF = MdBext
B2
Z B
2
Fn F s = dBextM =
:
0 20
0 M
B
1 B
B
2 Bext
111111111
000000000
000
11100000000
11111111
000000000
111111111
000
111
000000000
11111111100000000
11111111
000
111
000000000
111111111
000
111
000000000
111111111
000
111
000000000
111111111
000
111
jump or
innite slope
000
111
Figure 26: Magnetization
urve of a type II super
ondu
tor.
The dierently dashed areas in Fig. 26 are equal. For type II super
ondu
tors, B
is only a theoreti
al
quantity.
31
6 JOSEPHSON EFFECTS
1
The quantitative des
ription of Josephson ee
ts at T T
(the usual
ase in appli
ations) needs a
mi
ros
opi
treatment. However, qualitatively they are the same at all temperatures T < T
, hen
e
qualitatively they may be treated within the Ginsburg-Landau theory.
Consider a very thin weak link between two halfs of a super
ondu
tor (Fig. 27).
2
S1 S2
x1 x2
2 = 1
1 (x1) 1 = j jei1
2
(x2) 2 = j jei2
2 = 1 +
FIG. 27: A weak link between two halfs of a super
ondu
tor.
The order parameter has its thermodynami
value on both sides x < x1 ; x > x2; but is expo-
nentially small at x = 0: Hen
e, any super
urrent through the weak link is small, and may be
onsidered
onstant in both bulks of super
ondu
tor. In the weak link, not only j j is small, also its
phase may
hange rapidly (e.g. from 2 = 1 to 2 = 1 + by a very small perturbation).
Without the right half, the boundary
ondition (58) would hold at x1 :
2ie
+ A = 0:
x ~ x x1
In the presen
e of the right half, this
ondition must be modied to slightly depending on the value
2 :
2ie
+ Ax =
2;
x ~ (90)
x1
1 B. D. Josephson, Phys. Lett. 1, 251 (1962).
32
where
is a small number depending on the properties of the weak link. Time inversion symmetry
demands that (90) remains valid for ! ; A ! A; hen
e
must be real as long as the phase of
does not depend on A. For the moment we
hoose a gauge in whi
h Ax = 0: Then, the super
urrent
density at x1 is
" #
ie~
js;x (x1 ) =
2m 1 x x1 1 x x1 =
h i
= 2iem~
1 2 1 2 =
jm
= 2 = jm sin 2 1:
ei(2 1 ) ei(1 2 ) (91)
We generalize the argument of the sine fun
tion by a general gauge transformation (22):
!
2e ;
~
A ! A+ ; Ax = 0 ! A x = ;
r x
! t
2e 2e 2
Z
2 1 !
= 2 1 ~
2 1 = 2 1 +
~
dxAx ; (92)
1
d
dt
= 2~e 2 1: (93)
Re
all that is the ele
tro
hemi
al potential measured by a voltmeter.
Now, the general Josephson equation reads
js = jm sin
: (94)
6.1 The d.
. Josephson ee
t, quantum interferen
e
A
ording to (94), a d.
. super
urrent of any value between jm and jm may
ow through the jun
tion,
and a
ording to (93) the potential dieren
e 2 1 is zero in that
ase.
Now,
onsider a jun
tion in the y z-plane with a magneti
eld applied in z-dire
tion. There are
super
urrents s
reening the eld away from the bulks of the super
ondu
ting halfs.
y
11
00
00
11 B
(y)11
00
b S1 4 11 00 3 S2
1
0 11
00
00 2
11 x
d 2
33
Let us
onsider
(y); and let
(0) =
0 at the edge y = 0: We have
Z Z 2 Z 3 Z 4 Z 1
Byd = ds A = dxAx + dyAy + dxAx + dyAy :
1234 1 2 3 4
In the jun
tion we
hoose the gauge Ax = By; Ay = 0: Then, the y-integrals vanish, and from (92),
(0) = 2 1 =
0 ;
(y) =
0 +
2 Z 3 dxBy =
0 + 2Bd y:
0 4 0
The d.
. Josephson
urrent density through the jun
tion os
illates with y a
ording to
!
js (y) = jm sin
0 +
2Bd y :
0
In experiment, at B = 0 one always starts from a biased situation with js = jm; hen
e
0 = =2; and
js (y) = jm
os
2Bd y: (95)
0
The total
urrent through the jun
tion is
Z b
Is =
jm dy
os
2Bd y;
0 0
where
is the thi
kness in z-dire
tion. F = b
is the area of the jun
tion. With
eib 1 sin(b)
Z b Z b
dy
os(y) = < dyeiy = < =
0 0 i
we nd that, depending on the phase
0; the maximal
urrent at a given eld B is
Is;max = F jm
j sin(2Bbd=0 )j : (96)
2Bbd=0
An even simpler situation appears, if one splits the jun
tion into a double jun
tion: Now, H ds A =
is the magneti
ux through the
ut-out, and
a
b =
2
I
ds A = 2 :
0 0
11
00
00
11
b
00
11
S1 00
11
00B
11
S2
00
11
00
11
s
00
11
00
11
a
34
Hen
e, the maximal d.
. Josephson
urrent is
" #
Is;max = F jm max
2
sin
a + sin
a +
:
=2F jm 2=0
a 0
a x
I s;m
a
Is;max = 2
F jm
os : (97)
0
Figure 30: Phase relations in Eq.(97).
This is the basis to experimentally
ount
ux
quanta with a devi
e
alled super
ondu
ting
quantum interferometer (SQUID).
35
I
~!r =2e V=R
d.
. Josephson spike, Im
If the impedan
e of the radio sour
e is large
ompared to the impedan
e of the Josephson jun
tion,
as is usually the
ase, we have a
urrent-sour
e situation, where the fed-in total
urrent determines
the voltage a
ross the jun
tion:
V = 2~e d
dt
= R(I Is ) = R(I Im sin
): (102)
I
Shapiro steps
V =R
Im
V
Figure 32: Josephson
urrent vs. voltage in the
urrent-sour
e situation.
36
7 MICROSCOPIC THEORY: THE FOCK SPACE
The S
hrodinger wavefun
tion of an (isolated) ele
tron is a fun
tion of its position, r; and of the
dis
rete spin variable, s : = (r; s): Sin
e there are only two independent spin states for an ele
tron
| the spin
omponent with respe
t to any (single)
hosen axis may be either up (") or down (#) |,
s takes on only two values, + and , hen
e may be thought as
onsisting of two fun
tions
(r; s) = ((rr;; +)) (103)
forming a spinor fun
tion of r: The expe
tation value of any (usually lo
al) one-parti
le operator
A(r; s; r0 ; s0 ) = Æ(r r0 )A^(r; s; s0) is
XZ
hAi = d3 r (r; s)A^(r; s; s0 )(r; s0 ): (104)
s;s0
They are
ommonly
alled (spinor-)orbitals. The quantum number l refers to both the spatial and
the spin state and is usually already a multi-index (for instan
e (nlm) for an atomi
orbital or (k)
for a plane wave), and we agree upon a
ertain on
e and forever given linear order of those l-indi
es.
Choose N of those orbitals, l1 ; l2 ; : : : ; lN ; in as
ending order of the li and form the determinant
L(x1 : : : xN ) = p1N ! det kli (xk )k: (107)
L = (l1 : : : lN ) is a new (hyper-)multi-index whi
h labels an orbital
onguration. This determinant of
a matrix aik = li (xk ) for every point (x1 : : : xN ) in the spin-position spa
e has the proper symmetry
property (105). In view of (106) it is normalized, if all li are dierent, and it would be identi
ally
37
zero, if at least two of the li would be equal (determinant with two equal raws): Two fermions
annot
be in the same spinor-orbital. This is the
ompared to (105) very spe
ial
ase of the Pauli prin
iple
(whi
h is the
ommonly known
ase).
Now, given a
omplete set of orbitals (106), we mention without proof that all possible orbital
ongurations of N orbitals (107) form a
omplete set of N -fermion wavefun
tions (105), that is, any
wavefun
tion (105) may be represented as
X
(x1 : : : xN ) = CLL(x1 : : : xN ) (108)
L
+ 21
XX Y
(lP ilP j jwjlj0 li0 ) ( 1)jPj ÆlPk l0k ; (110)
i6=j P k(6=i;j )
where P is any permutation of the subs
ripts i; j; k; and jPj is its order. The matrix elements are
dened as XZ
0
(li jhjli) = d3 rli (r; s)h^ ss0 (r)l0i (r; s0 ); (111)
ss0
and
X Z
(li lj jwjlj0 li0 ) = d3 ri d3 rj li (ri ; si )lj (rj ; sj ) wsi s0i ;sj s0j (ri ; rj ) l0j (rj ; s0j )l0i (ri ; s0i ): (112)
si s0i ;sj s0j
(Note our
onvention on the order of indi
es whi
h may dier from that in other textbooks but leads
to a
ertain
anoni
al way of writing of formulas later on.) The rst line on the r.h.s. of (110) is
non-zero only if the two
ongurations L and L0 dier at most in one orbital, and the sum over all
permutations P has only one non-zero term in this
ase, determining the sign fa
tor for that matrix
element. The se
ond line is non-zero only if the two
ongurations dier at most in two orbitals, and
the sum over all permutations has two non-zero terms in that
ase: if P is a perturbation with lP k = lk0
for all k 6= i; j , then the
orresponding
ontribution is (1=2)[(lP ilP j jwjlj0 li0 ) ((lP j lP ijwjlj0 li0 )℄( 1)jPj.
For L = L0 and P = identity (and sometimes also in the general
ase) the rst matrix element is
alled dire
t intera
tion and the se
ond one ex
hange intera
tion.
7.2 The Fo
k spa
e
Up to here we
onsidered representations of quantum me
hani
s by wavefun
tions with the parti
le
number N of the system xed. If this number is ma
ros
opi
ally large, it
annot be xed at a single
denite value in experiment. Zero mass bosons as e.g. photons may be emitted or absorbed in systems
of any s
ale. (In a relativisti
des
ription any parti
le may be
reated or annihilated, possibly together
with its antiparti
le, in a va
uum region just by applying energy.) From a mere te
hni
al point of
view, quantum statisti
s of identi
al parti
les is mu
h simpler to formulate with the grand
anoni
al
ensemble with varying parti
le number, than with the
anoni
al one. Hen
e there are many good
reasons to
onsider quantum dynami
s with
hanges in parti
le number.
38
In order to do so, we start with building the Hilbert spa
e of quantum states of this wider frame:
the Fo
k spa
e. The
onsidered up to now Hilbert spa
e of all N -parti
le states having the appro-
priate symmetry with respe
t to parti
le ex
hange will be denoted by HN . In the last subse
tion we
introdu
ed a basis fLg in HN . Instead of spe
ifying the multi-index L as a raw of N indi
es li we
may denote a basis state by spe
ifying the o
upation numbers ni (being either 0 or 1) of all orbitals
i: X
jn1 : : : ni : : :i; ni = N: (113)
i
Our previous determinantal state (107) is now represented as
jL i = j0 : : : 01l1 0 : : : 01l2 0 : : : 01lN 0 : : :i:
Two states (113) not
oin
iding in all o
upation numbers ni are orthogonal. HN is the
omplete
linear
P spa
e spanned by the basis ve
tors (113), i.e. the states of HN are either linear
ombinations
jLiCL of states (113) (with the sum of the squared absolute values of the
oeÆ
ients CL equal to
unity) or limits of Cau
hy sequen
es of su
h linear
ombinations. (A Cau
hy sequen
e is a sequen
e
fj n ig with limm;n!1 h m n j m n i = 0. The in
lusion of all limits of su
h sequen
es
into HN means realizing the topologi
al
ompleteness property of the Hilbert spa
e, being extremely
important in all
onsiderations of limits. This
ompleteness of the spa
e is not to be
onfused with
the
ompleteness of a basis set fi g.
The extended Hilbert spa
e F (Fo
k spa
e) of all states with the parti
le number N not xed is
now dened as the
ompleted dire
t sum of all HN . It is spanned by all state ve
tors (113) for all N
with the above given denition of orthogonality retained, and is
ompleted by
orresponding Cau
hy
sequen
es, just as the real line is obtained from the rational line by
ompleting it with the help of
Cau
hy sequen
es of rational numbers.
Note that F now
ontains not only quantum states whi
h are linear
ombinations with varying
ni so that ni does not have a denite value in the quantum state (o
upation number
u
tuations),
but also linear
ombinations with varying N so that now quantum
u
tuations of the total parti
le
number are allowed too. (For bosoni
elds as e.g. laser light those quantum
u
tuations
an be
ome
important experimentally even for ma
ros
opi
N .)
7.3 O
upation number representation
We now
ompletely abandon the awful wavefun
tions (105) and will ex
lusively work with the o
-
upation number eigenstates (113) and matrix elements between them. The simplest operators are
those whi
h provide just a transition between basis states (113) whi
h are as
lose to ea
h other as
possible: those whi
h dier in one o
upation number only.
The denition of these
reation and annihilation operators for fermions must have regard to the an-
tisymmetry of the quantum states and to Pauli's ex
lusion prin
iple following from this antisymmetry.
They are dened as P
^i j : : : ni : : :i = j : : : ni 1 : : :i ni ( 1) j<i nj ; (114)
P
^yi j : : : ni : : :i = j : : : ni + 1 : : :i (1 ni ) ( 1) j<i nj : (115)
The usefulness of the sign fa
tors will be
ome
lear below. By
onsidering the matrix elements with
all possible o
upation number eigenstates (113), it is easily seen that these operators have all the
needed properties, do parti
ularly not
reate non-fermioni
states (that is, states with o
upation
numbers ni dierenty from 0 or 1 do not appear: appli
ation of
^i to a statey with ni = 0 gives zero,
and appli
ation of
^i to a state with ni = 1 gives zero as well). The
^i and
^i are mutually Hermitian
onjugate, obey the key relations
n^ i j : : : ni : : :i
^yi
^i j : : : ni : : :i = j : : : ni : : :i ni (116)
and
[^
i ;
^yj ℄+ = Æij ; [^
i;
^j ℄+ = 0 = [^
yi ;
^yj ℄+ (117)
39
with the anti
ommutator [^
i ;
^yj ℄+ =
^i
^yj +
^yj
^i dened in standard way. Reversely, the
anoni
al
anti
ommutation relations (117) dene all the algebrai
properties of the
^-operators and moreover
dene up to unitary equivalen
e the Fo
k-spa
e representation (114, 115). (There are, however, vast
lasses of further representations of those algebrai
relations with a dierent stru
ture and not unitary
equivalent to the Fo
k-spa
e representation.)
The basis (113) of the Fo
k spa
e is systemati
ally generated out of a single basis ve
tor, the
va
uum state ji j0 : : : 0i (with N =0) by applying
^y -operators:
jn1 : : : ni : : :i = : : :
^yi : : :
^y1 ji: (118)
Observe again the order of operators dening a sign fa
tor in view of (117) in agreement with the
sign fa
tors of (114, 115). Sin
e produ
ts of a given set of N
^y-operators written in any order agree
with ea
h other up to possibly a sign, all possible expressions (118) do not generate more dierent
basis ve
tors than those of (107) with the
onvention on the order of the li as agreed upon there.
Hen
eforth, by using (118) we need not bother any more about the given linear order of the orbital
indi
es.
With the help of the
^-operators, any linear operator in the Fo
k spa
e may be expressed. It is
not diÆ
ult to demonstrate that the Hamiltonian
H^ =
X
^yi (ijhjj )^
j +
1 X
^y
^y(ij jwjkl)^
k
^l (119)
ij ijkl
2 i j
has the same matrix elements with o
upation number eigenstates (113) as the Hamiltonian (109))
has with determinantal states in (110). Be
ause of the one-to-one
orresponden
e between the deter-
minantal states (107) and the o
upation number eigenstates and be
ause both span the Fo
k spa
e,
by linearity the Hamiltonians (109) and (119) are equivalent. The building prin
iple of the equivalent
of any linear operator given in the S
hrodinger representation is evident from (119).
The S
hrodinger wavefun
tion of a bosoni
many-parti
le quantum state must be totally symmetri
with respe
t to parti
le ex
hange (omission of the minus sign in (105)). The determinants are then
to be repla
ed by symmetrized produ
ts (permanents), with a slightly more involved normalization
fa
tor. The orbitals may now be o
upied with arbitrary many parti
les: ni = 0; 1; 2; : : : : This
ase
may be realized with bosoni
reation and annihilation operators
^bi j : : : ni : : :i = j : : : ni 1 : : :ipni; (120)
p
^byi j : : : ni : : :i = j : : : ni + 1 : : :i ni + 1; (121)
n^ i j : : : ni : : :i ^bi ^bi j : : : ni : : :i = j : : : ni : : :ini :
y (122)
with the
anoni
al
ommutation relations
[^bi; ^byj ℄ = Æij ; [^bi; ^bj ℄ = 0 = [^byi ; ^byj ℄ : (123)
The basis states of the Fo
k spa
e are
reated out of the va
uum a
ording to
^by1n1 ^byi ni
jn1 : : : ni : : :i = pn ! pn ! ji: (124)
1 i
The order of these operators in the produ
t does not make any dieren
e. The
hoi
e of fa
tors
on the r.h.s. of (120, 121) not only ensures that (122) holds but also ensure the mutual Hermitian
onjugation of ^bi and ^byi .
7.4 Field operators
A spatial representation may be introdu
ed in the Fo
k spa
e by dening eld operators
^(x) = X i (x)^ai ; ^y (x) = X i (x)^ayi ; (125)
i i
40
where the a^i mean either fermioni
operators
^i or bosoni
operators ^bi . The eld operators ^(x) and
^y (x) obey the
anoni
al (anti-)
ommutation relations
[ ^(x); ^y (x0 )℄ = Æ(x x0 ); [ ^(x); ^(x0 )℄ = 0 = [ ^y (x); ^y (x0 )℄: (126)
They provide a spatial parti
le density operator
n^ (r) =
X
^y (r; s) ^(r; s) (127)
s
These relations are readily obtained from those of the
reation and annihilation operators, and by
taking into a
ount the
ompleteness and orthonormality (106) of the orbitals i .
In terms of eld operators, the Hamiltonian (109) or (119) reads
XZ
H^ = d3 r ^y (r; s) h^ ss0 (r) ^(r; s0 ) +
ss0
+ 21
Z
d3 r1 d3 r2 ^y (r1 ; s1 ) ^y (r2 ; s2 ) ws1 s01 ;s2 s02 (r1 ; r2 ) ^(r2; s02) ^(r1; s01): (129)
X
s1 s01 s2 s02
41
8 MICROSCOPIC THEORY: THE BCS MODEL
The great advantage of the use of
reation and annihilation or eld operators lies in the fa
t that we
an use them to manipulate quantum states in a physi
ally
omprehensible way without expli
itly
knowing the wavefun
tion. We even
an think of modied operators of whi
h we know little more than
their algebrai
properties. The point is that the o
upation number formalism applies for every orbital
set (106). The transition from one set of operators obeying
anoni
al (anti-)
ommutation relations to
another su
h set is
alled a
anoni
al transformation in quantum theory.
8.1 The normal Fermi liquid as a quasi-parti
le gas
A normal
ondu
ting Fermi liquid has a fermioni
quasi-parti
le ex
itation spe
trum whi
h behaves
very mu
h like a gas of independent parti
les with energies k (for the sake of simpli
ity we assume it
isotropi
in k-spa
e although this assumption is not essential here). Non-intera
ting fermions would
have a ground state with all orbitals with < o
upied and all orbitals with > empty; is
the
hemi
al potential. By adding or removing a fermion with = new ground states with N 1
fermions are obtained. By adding a fermion with > an ex
ited state is obtained with ex
itation
energy . By removing a fermion with < | that is,
reating a hole in the original ground
state | an ex
ited state is obtained with ex
itation energy j j: rst lift the fermion to the level
and then remove it without
hanging the
hara
ter of the state any more (Figs. 33 and 34).
k
quasi-ele tron
0 jk j
kF k
quasi-hole
kF k
FIG. 33: Creation of an ex
ited ele
tron and of a hole, resp. FIG. 34: Ex
itation spe
trum of a Fermi gas.
The ground state j0i of a normal metal has mu
h the same properties:
ondu
tion ele
trons with
> and holes with < may be ex
ited with ex
itation energies as above. These are not the
original ele
trons making up the metal together with the atomi
nu
lei. Rather they are ele
trons or
missing ele
trons surrounded by polarization
louds of other ele
trons and nu
lei in whi
h nearly all
the Coulomb intera
tion is absorbed. We do not pre
isely know these ex
itations nor do we know the
ground state j0i (although a quite elaborate theory exists for them whi
h we ignore here). We just
assume that they may be represented by fermioni
operators with properties like those in the gas:
k < :
^yk j0i = 0;
^yk
^k j0i = j0i; (130)
k > :
^k j0i = 0;
^k
^yk j0i = j0i; (131)
42
k is the waveve
tor and the spin state of the quasi-parti
le.
Sin
e all intera
tions present in the ground state j0i are already absorbed in the quasi-parti
le
energies k , only ex
ited
ondu
tion ele
trons or holes exert a remainder intera
tion. Hen
e, we may
write down an ee
tive Hamiltonian
^H~ = X
^yk (k )^
k + 1 k Xk0
^yk+q
^yk0 q0 wkk0 q
^k0 0
^k +
>; >
k
2 k; k0 0 ; q
k >;
X k0 <
+
^yk+q
^k0 q0 wkk0 q
^yk0 0
^k +
k; k0 0 ; q
k <; k0 <
+ 21
X
^k+q
^k0 q0 wkk0 q
^yk0 0
^yk : (132)
k; k0 0 ; q
The matrix elements in the three lines are qualitatively dierent: they are predominantly repulsive
in the rst and last line and attra
tive in the se
ond line; ele
trons and holes have opposite
harges.
With the relations (130, 131) one nds easily
k <
X
h0jH^~ j0i = (k ) = E~0 H^~ j0i = j0iE~0 : (133)
k
43
The intera
tion term is obtained with the rule
kii
ykii j0i = j0i. One
ontribution appears from
k = k1 1 ; k0 0 = k2 2 ; and another
ontribution
^yk2 +q2
^yk1 q1 j0i wk2 k1 q from k0 0 = k1 1 ;
k = k2 2 : The minus sign in this
ontribution is removed by anti
ommuting the two
^y -operators,
and then, by repla
ing q with q under the q-sum and observing wk2 k1 q = wk1 k2 q whi
h derives
from w(r1 ; r2) = w(r2 ; r1), this se
ond
ontribution is equal to the rst one, when
e omitting the
fa
tor (1/2) in front. (This is how ex
hange terms appear automati
ally with
^-operators sin
e their
anti
ommutation rules automati
ally retain the antisymmetry of states.) For the simpli
ity of writing
we omitted here and in (132) the spin dependen
e of the intera
tion matrix element. It is always
present in the ee
tive quasi-parti
le intera
tion, and it may always be added afterwards without
onfusion. (We might introdu
e a short-hand notation k for k.) The ee
tive intera
tion of two
quasi-parti
les with equal spin diers from that of two quasi-parti
les with opposite spin. The spin-
ip s
attering of quasi-parti
les | an intera
tion with
hanging 1 and 2 into 10 and 20 | may
often be negle
ted. Then, the q-sum of (139) need not be
ompleted by additional spin sums.
8.2 The Cooper problem1
From (139) it
an be seen that the state (138) is not any more an eigenstate of the ee
tive Hamiltonian,
not even within the approximations made in the previous subse
tion. The two ex
ited quasi-parti
les
intera
t and thus form a
orrelated pair state. We try to nd this pair state of lowest energy for
two ele
trons ( > 0) within our approximate approa
h. Sin
e we expe
t that this state is formed
out of quasi-parti
le ex
itations with energies 0 our approximations
annot be
riti
al. (The
quasi-parti
le lifetime be
omes innite for jj ! 0:) We expe
t that the state lowest in energy has
zero total momentum, hen
e we build it out of quasi-parti
le pairs with k2 = k1:
X
j i = ak jk k0 i: (140)
k
where we assume a xed
ombination of and 0 and the yet unknown expansion
oeÆ
ients to
depend on k only, be
ause the sought state is to be expe
ted to have a denite total spin. (Re
all
that we are
onsidering an isotropi
metal in this
hapter.) We want that this pair state j i is an
eigenstate of H^ :
H^ j i = j iE; H^ j i =
X X
^yk
^y k0 j0i 2k ak +
^yk+q
^y k q0 j0i wk kq ak : (141)
k kq
Multiply the last relation with h0j
^ k0 0
^k0 and observe h0j
^ k0 0
^k0
^yk
^y k0 j0i = h0j
^ k00 (Ækk0
^yk
^k0 )^
y k0 j0i = Ækk0 h0j(Ækk0
^y k0
^ k0 0 )j0i h0j
^ k0 0
^yk (Æ kk0 Æ0
^y k0
^k0 )j0i = Ækk0
h0j(Æ kk0 Æ0
^yk
^ k0 0 )Æ kk0 Æ0 j0i = Ækk0 Æ kk0 Æ0 to obtain
X
E ak0 a k0 Æ0 = 2k0 ak0 a k0 Æ0 + wk0 q; k0 +q;q a k0 q a k0 +q Æ0 : (142)
q
In the last term, we also used again w k0 q;k0+q;q = wk0 +q; k0 q; q and then repla
ed the sum over
q by a sum over q.
Due to the isotropy of our problem we expe
t the solution to be an angular momentum eigenstate,
hen
e ak should have a denite parity. It is immediately seen that a non-trivial solution with even
parity a k = ak (even angular momentum) is only possible, if Æ0 = 0, that is for a singlet 0 = .
For a spin triplet 0 = only a non-trivial solution with odd parity (odd angular momentum) is
possible. To be spe
i
,
onsider the singlet
ase. (The triplet
ase is analogous.) Assume
ak = ak Ylm (k=k) (143)
1 L. N. Cooper, Phys. Rev. 104, 1189 (1956).
44
with even l: In (142), rename k0 ! k, k0 q ! k0. The matrix element wk0 ; k0 ;k k0 determines the
s
attering amplitude from states k; k into states k0; k0. we use an expansion
X
wk0 ; k0 ;k k0 = l wkl wkl0 Ylm (k)Ylm
(k0 ): (144)
lm
The sought lowest pair energy
orresponds to the lowest solution of F (Elm ) = 1=l.
The fun
tion F (Elm )
has poles for Elm = 2k
F
where it jumps from 1
to +1. Re
all that the
k -values are all positive
and start from zero. For
Elm ! 1, F (Elm ) ap-
proa
hes zero from nega-
tive values. Hen
e, if l >
0, then the lowest solu-
tion Elm of (146) is posi-
tive and the ground state
1 j0i of the normal metal
l > 0 is stable. If at least one
l -value is negative (at-
1
0 E tra
tive intera
tion), then
l < 0 there is unavoidably a neg-
ative solution Elm of (146):
Elm the `ex
ited pair' has nega-
2k tive energy and the normal
ground state j0i is unsta-
ble against forming of pairs
of bound quasi-parti
les,
no matter how small jl j
is (how weak the attra
-
tive intera
tion is). Pairs
are spontaneously formed
Figure 35: The fun
tion F (E ) from Eq. (146). and the ground state re
on-
stru
ts. This is the
ontent
of Cooper's theorem.
If the intera
tion is
ut of at some energy !
,
wkl = 10 for 0 < k < !
;
elsewhere (147)
and the density of states for k is nearly
onstant in this interval, N () = N (0); then, with negative
Elm ,
45
" #
X
j j2 1 Z !
1 = N (0) jElm + 2!
;
k
wkl
Elm 2k = N (0)
0
d
jElm j + 2 2 ln jElm j
hen
e,
jElm j = h
2!
i : (148)
exp 2 1
N (0)jl j
This yields 8 " #
> 2
jElm j > 2!
exp N (0)jl j for N (0)jl j
<
1 (149)
:
N (0)jl j!
in the weak and strong
oupling limits. In this
hapter we only
onsider the weak
oupling limit where
jElm j is exponentially small.
The whole analysis may be repeated for the
ase where the pair has a non-zero total momentum
q. In that
ase the denominator of (146) is to be repla
ed with Elm (q) k+q=2 k+q=2 where now
jk q=2j must be larger than kF . For small q; this
ondition redu
es the density of states in ee
t in
an interval of thi
kness j= kjq=2 = vF q=2 at the lower -integration limit; vF is the Fermi velo
ity.
The result is
Elm (q) Elm + vF q=2: (150)
In the weak
oupling limit, Elm(q)
an only be negative for exponentially small q.
We performed the analysis with a pair of parti
les. It
an likewise be done with a pair of holes
with an analogous result.
8.3 The BCS Hamiltonian
Frohli
h1 was the rst to point out that the ele
tron-phonon intera
tion is
apable of providing an
ee
tive attra
tion between
ondu
tion ele
trons in the energy range of phonon energies.
From Cooper's analysis it follows that, if there is a weak attra
tion, it
an only be ee
tive for
pairs with zero total momentum, that is, between k and k. With the assumption that the attra
tion
is in the l = 0 spin singlet
hannel, this led Bardeen, Cooper and S
hrieer2 to the simple model
Hamiltonian
k ;k0 <+!
!
<X
g
H^ BCS =
X
^yk (k )^
k
^yk0 "
^y k0 #
^ k#
^k" : (151)
k
V kk0
Here, g > 0 is the BCS
oupling
onstant, and V is the normalization volume. Sin
e the density of
plane-wave states in k-spa
e is V=(2)3: Pk = V=(2)3 R d3 k, the matrix element of an n-parti
le
intera
tion (appearing in an n-fold k-sum) must be proportional to V (n 1) in order that the Hamil-
tonian is extensive ( V ). The modeled attra
tive intera
tion is assumed in an energy range of width
2!
around the
hemi
al potential (Fermi level in the
ase T = 0), where !
is a
hara
teristi
phonon
energy for whi
h the Debye energy of the latti
e
an be taken.
The state j0i of (130, 131)
annot any more be the ground state of this Hamiltonian sin
e Cooper's
theorem tells us that this state is unstable against spontaneous formation of bound pairs with the gain
of their binding energy. The problem to solve is now to nd the ground state and the quasi-parti
le
spe
trum of the BCS-Hamiltonian. This problem was solved by Bardeen, Cooper and S
hrieer, and,
shortly thereafter and independently by means of a
anoni
al transformation, by Bogoliubov and
Valatin. Bardeen, Cooper and S
hrieer thus provided the rst mi
ros
opi
theory of super
ondu
-
tivity, 46 years after the dis
overy of the phenomenon.
1 H. Fr
ohli
h, Phys. Rev. 79, 845 (1950).
2 J. Bardeen, L. N. Cooper, and J. R. S
hrieer, Phys. Rev. 108, 1175 (1957).
46
8.4 The Bogoliubov-Valatin transformation1
Suppose the ground state
ontains a bound pair. Ex
iting one parti
le of that pair leaves its partner
behind, and hen
e also in an ex
ited state. If one wants to ex
ite only one parti
le, one must annihilate
simultaneously its partner. Led by this
onsideration, for the quasi-parti
le operators in the ground
state of (151) an ansatz
^bk" = uk
^k" vk
^y k#; ^bk# = uk
^k# + vk
^y k"
is made. uk and vk are variational parameters. Again we
onsider the isotropi
problem and hen
e
their dependen
e on k = jkj only. The reason of the dierent signs in the two relations be
omes
lear in a minute. Sin
e for ea
h orbital annihilated by
^k" ;
^ k#; ^bk"; ^b k# an r-independent phase
fa
tor may be arbitrarily
hosen, uk and vk may be assumed real without loss of generality. These
Bogoliubov-Valatin transformations together with their Hermitian
onjugate may be summarized as
^bk = uk
^k vk
^y k ; ^byk = uk
^yk vk
^ k : (152)
We want these transformations to be
anoni
al, that is, we want the new operators ^bk ; ^byk again to
be fermioni
operators. One easily
al
ulates
[^bk ; ^bk0 0 ℄+ = uk vk0 0[^
k ;
^y k0 0 ℄+ + [^
y k ;
^k0 0 ℄+ =
= uk vk0 0Æk k0 Æ 0 + Æ kk0 Æ 0 = uk vk Æ kk0 ( + ) = 0:
In the rst equality it was already
onsidered that annihilation and
reation operators
^ and
^y ,
respe
tively, anti
ommute among themselves. The analogous result for the ^by-operators is obtained in
the same way. The sign fa
tor in the transformation ensures that the anti
ommutation is retained
for the ^b- and ^by-operators, respe
tively. Analogously,
[^bk ; ^byk0 0 ℄+ = u2k [^
k ;
^yk0 0 ℄+ + 0 vk2 [^
y k ;
^ k0 0 ℄+ = u2k + vk2 Ækk0 Æ0 ;
hen
e the
ondition
u2k + vk2 = 1 (153)
ensures that the transformation is
anoni
al and the new operators are again fermioni
operators.
Multiplying the rst relation (152) by uk , repla
ing in the se
ond one k with k , multiplying
it with vk , and then adding both results yields with (153) the inverse transformation
^k = uk^bk + vk ^by k ;
^yk = uk^byk + vk^b k : (154)
Observe the reversed sign fa
tor.
The next step is to transform the Hamiltonian (151). With
^yk
^k = uk^byk + vk^b k uk^bk + vk^by k =
= u2k^byk^bk + vk2^b k ^by k + uk vk ^byk^by k + ^b k ^bk
and the anti
ommutation rules it is easily seen that the single-parti
le part of the BCS-Hamiltonian
transforms into
2 (k )vk2 + (k )(u2k vk2 ) ^byk^bk + 2 (k )uk vk ^byk"^by k# + ^b k#^bk" :
X X X X
k k k
It has also been used that under the k-sum k may be repla
ed by k: Further, with
B^k =
^ k#
^k" = uk^b k# vk^byk" uk^bk" + vk^by k# =
= u2k^b k#^bk" vk2^byk"^by k# + uk vk ^b k#^by k# ^byk"^bk" ; (155)
1 N. N. Bogoliubov, Nuovo Cimento 7, 794 (1958); J. G. Valatin, Nuovo Cimento 7, 843 (1958).
47
the full transformed BCS-Hamiltonian reads
H^ BCS = 2 (k )vk2 + (k
X X
)(u2k vk2 )
X
^byk^bk +
k k
g
+2 (k
X
^ ^
uk vk byk" by k#
) + ^b k#^bk"
X
B^ky 0 B^k ; (156)
k
V kk0
The nk are the eigenvalues (0 or 1) of the o
upation number operators (157) in the ground state
j 0 i.
This energy expression still
ontains the variational parameters uk and vk whi
h are
onne
ted by
(153), when
e vk =uk = uk =vk . For given o
upation numbers, (158) has its minimum for
" #
E u2 v2
uk
= 4(k )uk + 2 k k
vk
1 nk" nk# = 0;
= Vg
X
uk vk 1 nk" nk# : (159)
k
2
gN2(0) ln 4!2
0
resulting in
0 = 2!
exp gN (0) 1 (163)
for the value of in the ground state (atPzero temperature).
If one repla
es the last termP (g=V ) B^ky 0 B^k of the transformed BCS-Hamiltonian (156) by the
mean-eld approximation k B^k +B^ky (re
all that was introdu
ed as = (g=V )h 0j Pk B^k j 0i,
f. (155, 159)), than it is readily seen that the relation (160) makes the anomalous terms (terms ^by^by
or ^b^b) of this Hamiltonian vanish: In mean-eld approximation the BCS-Hamiltonian is diagonalized
by the Bogoliubov-Valatin transformation, resulting in
H^ m-f vk2 )^byk ^bk ^byk^bk =
X X X X
= 2 (k )vk2 + (k )(u2k 2 uk vk 1
k k k
Xh i
=
onst. + (k )(u2k vk2 ) + 2uk vk ^byk^bk =
k
k ^byk^bk
X
=
onst. + (164)
k
50
9 MICROSCOPIC THEORY: PAIR STATES
In mean-eld BCS theory, the ground state is determined by the
omplete absen
e of quasi-parti
les.
With the properties of the Bogoliubov-Valatin transformation, this ground state is found to be the
ondensate of Cooper pairs in plane-wave K = 0 states of their
enters of gravity. By o
upying
quasi-parti
le states a
ording to Fermi statisti
s, thermodynami
states of the BCS super
ondu
tor
are obtained.
is the properly normalized state with properties (167). The normal metal ground state j0i of (130,
131) is
k <
Y k <
Y
j0i =
^yk ji =
^yk"
^y k# ji (169)
k k
and hen
e has the form (168) too, with uk = 0 for k < and uk = 1 for k > and the opposite
behavior for vk .
We now demonstrate the properties of (168). Sin
e
hj uk + vk
^ k#
^k" uk + vk
^yk"
^y k# ji = u2k + vk2 = 1;
j 0 i of (168) is properly normalized. Moreover,
" #
^bk0 "j 0i = Y
uk + vk
^yk"
^y k# uk0
^k0 "
vk0
^y k0 # uk0 + vk0
^yk0 "
^y k0 # ji =
k(6=k0 )
" #
Y
= uk + vk
^yk"
^y k# uk0 vk0
^k0 "
^yk0 "
^y k0 #
^y k0 # ji =
k(6=k0 )
" #
Y
= uk + vk
^yk"
^y k# uk0 vk0 1
^yk0 "
^k0 "
^y k0#
^y k0 # ji = 0 (170)
k(6=k0 )
and analogously ^bk0 #j 0i = 0. This
ompletes our proof. Histori
ally, Bardeen, Cooper and S
hrieer
solved the BCS model with the ansatz (168), before the work of Bogoliubov and Valatin.
51
9.2 The pair fun
tion
Next we nd the wavefun
tions
ontained in j 0i: Re
all, that
^yk
reates a
ondu
tion ele
tron in
the plane-wave state exp(ik r) (s) so that the eld operator ^y (rs) is
(That is, rs denotes the
enter of gravity and the spin of the ele
tron with its polarization
loud whi
h
together make up the `
ondu
tion ele
tron'.)
The N -parti
le wavefun
tion
ontained in j 0i and depending on these variables is
0(x1 ; : : : ; xN ) = hx1 : : : xN j 0i = hj ^(xN ) ^(x1 ) uk + vk
^yk"
^y k#ji:
Y
(172)
k
k1 :::kN k k
=0; if not ki i 6=kj j for i6=j
p p Y2
N= q q !
X 0 i(k1 r1 ++kN rN ) Y
= e hj
^kN N
^k1 1 gk
^yk"
^y k# ji uk
k1 :::kN k k
| {z }
sum over all possible
ontra
tions
!
X0
eik2 (r1 r2 ) g
k2 Æ1 2 eikN (rN
1 rN ) g Æ
kN N 1 N =
fk2i g
A(r1 r2 )singlet (r3 r4 )singlet (rN 1 rN )singlet (173)
with X vk
() gk eik ; gk = : (174)
k
uk
In the se
ond line of (173), (171) was inserted for the ^(xi ) of (172), and the uk s were fa
tored out of
the produ
t of (172) (leaving gk = vk =uk behind in the se
ond item of the fa
tors). The k-produ
ts
run over all grid points of the (innite) k-mesh, e.g. determined by periodi
boundary
onditions for
the sample volume V , while the sum runs over all possible produ
ts for sets of N disjun
t k-values out
of that mesh. This disjun
t nature of the k-sums is indi
ated by a dash at the sum in the following
lines. Expansion of the rst k-produ
t yields terms with 0, 2, 4, . . .
^y -operators of whi
h only the
terms with exa
tly those N
^y -operators that
orrespond to the N
^-operators of an item of the k-sum
left to the produ
t produ
e a non-zero result between the
^-va
uum states hj ji. These results are
most easily obtained by anti
ommuting all annihilation operators to the right of all
reation operators
and are usually
alled
ontra
tions; depending on the original order of the operators, ea
h result is
1. The produ
t over the uk , whi
h multiplies ea
h
ontribution and whi
h as previously runs over
all innitely many k-values of the full mesh, yields a normalizing fa
tor whi
h is independent of the
values of k1; : : : ; kN of the sums. Like ea
h individual fa
tor uk , it depends on the
hemi
al potential
and on the gap 0 . At this point one must realize that the gk are essentially non-zero inside of
the Fermi surfa
e (
f. (161)). Hen
e, the
ontribution to 0(x1 ; : : : ; xN ) has a non-negligible value
only for all ki-ve
tors inside the Fermi surfa
e, and this value in
reases with an in
reasing number of
su
h ki -ve
tors and de
reases again, if an appre
iable number of ki -ve
tors falls outside of the Fermi
surfa
e: the norm of (172) is maximal for N -values su
h that the k2i o
upy essentially all mesh points
inside the Fermi surfa
e. That is, this norm is non-negligible only for those N -values
orresponding to
52
the ele
tron number in the original normal Fermi liquid: j 0i is a grand-
anoni
al state with a sharp
parti
le-number maximum at the
anoni
al N -value. We did not tra
e normalizing fa
tors in (173)
and use a sign of approximation in (174) again, assumig () to be normalized.
The result is a state j 0i
onsisting of pairs of ele
trons in the geminal (pair-orbital) ()singlet .
In order to analyze what this pair-orbital () looks like, re
all that uk and vk may be written as
fun
tions of 0 and (k )=0 ~vF k=0 (
f. (161)). Hen
e, gk g~(~vF k=0), and
Z Z 1 Z 1 Z 1
eik e ik
() d3 kgk eik dkk2 gk deik = dkk2 gk
0 1 0 ik
1 Z 1 dkkg~ ~vF k sin k 1 Z 1 dxxg~(x) sin 0x f (=0);
0 0 0 ~vF
0
1 = :~vF
(175)
Æk 0
By
omparison with (173), is the distan
e ve
tor of the two ele
trons in the pair, their distan
e
on average being of the order of 0 , while the pair-orbital does not depend on the position R of
the
enter of gravity of the pair: with respe
t to the
enter of gravity the pair is delo
alized, it
is a plane wave with wave ve
tor K = 0. Moreover, again due to (173), all N=2 1023 ele
tron
pairs o
upy the same delo
alized pair-orbital in the BCS ground state j 0i : this ma
ros
opi
ally
o
upied delo
alized (that is,
onstant in R-spa
e) pair-orbital is the
ondensate wavefun
tion of the
super
ondu
ting state, and hen
e the stru
ture (173) of j 0i ensures that the solution of the BCS
model is a super
ondu
tor.
For a real super
ondu
tor, the gap 0
an be measured (for instan
e by measuring thermodynami
quantities whi
h depend on the ex
itation spe
trum or dire
tly by tunneling spe
tros
opy. With the
independently determined Fermi velo
ity vF of ele
trons in the normal state, this measurement yields
dire
tly the average distan
e 0 of the ele
trons in a pair whi
h
an be
ompared to the average
distan
e rs of to arbitrary
ondu
tion ele
trons in the solid given by the ele
tron density. For a
weakly
oupled type I super
ondu
tor this ratio is typi
ally
0
rs
103 : : : 104: (176)
In this
ase, there are 109 : : : 1012 ele
trons of other pairs in the volume between a given pair: There
is a pair
orrelation resulting in a
ondensation of all ele
trons into one and the same delo
alized pair
orbital in the super
ondu
ting state, however, the pi
ture of ele
trons grouped into individual pairs
would be by far misleading.
In order to
reate a super
urrent, the
ondensate wavefun
tion, that is, the pair orbital must be
provided with a phase fa
tor
(; R) eiK R
(177)
by repla
ing the
reation operators in (168) with
^yk+K=2"
^y k+K=2#: Obviously, it must be K0 1
in order not to deform (and thus destroy) the pair orbital itself. Hen
e, 0 has the meaning of the
oheren
e length of the super
ondu
tor at zero temperature.
54
10 MICROSCOPIC THEORY: COHERENCE FACTORS
As already done in the last se
tion, thermodynami
states of a super
ondu
tor are obtained by o
-
upying quasi-parti
le states with Fermi o
upation numbers. External elds, however, in most
ases
ouple to the
^-operator elds. Due to the
oupling of
^-ex
itations in a super
ondu
tor interferen
e
terms appear in the response to su
h elds.
10.1 The thermodynami
state
Let fkg be any given disjun
t set of quasi-parti
le quantum numbers. Then,
Y
j fkg i = ky j 0i (186)
k2fkg
is a state with those quasi-parti
les ex
ited above the super
ondu
ting ground state j 0i: If there are
many quasi-parti
le ex
itations present, they intera
t with ea
h other and with the
ondensate in the
ground state (the latter intera
tion is in terms of the operators P^
^yk and
^kpP^y), and this intera
tion
leads to the temperature dependen
e of their energy dispersion law k = (k )2 + 2 via the
temperature dependen
e = (T ): The thermodynami
state is
Q
k2fkg fk
P =
X
j fkg i Z
h fkg j; (187)
fkg
where the summation is over all possible sets of quasi-parti
le quantum numbers,
fk =
1 (188)
exp(k =kT ) + 1
is the Fermi o
upation number, and Z = Z (T; ) is the partition fun
tion determined by
Q
k2fkg fk
trP =
X
Z
= 1: (189)
fkg
As usual, the thermodynami
expe
tation value of any operator A^ is obtained as trA^P: For instan
e
the average o
upation number itself of a quasi-parti
le in the state k00 is
Q
nk0 0 = tr ^ky 0 0 ^k0 0 j fkg i fk h fkg j =
X
fkg Z
Q
fk
j^y 0 0 ^k0 0 j fk0 trP = fk0 :
X
= h fkg k fkg i Z = (190)
fkg
The state ^k00 j fkgi is obtained from the state j fkg i by removing the quasi-parti
le k00 : Hen
e,
if one fa
tors out fk0 , the remaining sum is again P.
10.2 The
harge and spin moment densities
The operator of the q Fourier
omponent of the
harge density of ele
trons in a solid is
X X
n^ (q) = e
^yk+q
^k = e
^yk+q"
^k" +
^y k#
^ k q# : (191)
k k
That of the spin moment density (in z-dire
tion) is
X X
^ (q) = B
^yk+q
^k = B
^yk+q"
^k"
^y k#
^
m k q# : (192)
k k
55
The statisti
al operator P of a normal metalli
state is
omposed in analogy to (187) from eigenstates
j fkg i of
^-operators. Then, in
al
ulating thermodynami
averages, ea
h item of the k-sum of
(191) and (192) is averaged independently. In the super
ondu
ting state, the items in parentheses of
the last of those expressions are
oupled and hen
e they are not any more averaged independently:
there appear
ontributions due to their
oherent interferen
e in the super
ondu
ting states j fkg i:
Those
ontributions appear in the response of the super
ondu
ting state to external elds whi
h
ouple
to
harge and spin densities.
Performing the Bogoliubov-Valatin transformation for the
harge density operator yields
X
^yk+q"
^k" +
^y k#
^ k q# =
k
X
= ujk+qj^byk+q" + vjk+qj^b u ^
b + v ^
by +
k q# k k" k k#
k
+ uk^by k# vk^bk" ujk+qj^b k q# vjk+qj^bk+q" =
X
= ujk+qjuk^byk+q"^bk" + vjk+qj vk^b ^y + ujk+qjvk^byk+q"^by k# + vjk+qjuk^b
k q# b k# ^ +
k q# bk"
k
+uk ujk+qj^by k#^b k bk"^byk+q"
q# + vk vjk+qj^ uk vjk+qj^by k#^byk+q" vk ujk+qj^bk"^b k q# =
X
= ujk+qjuk vjk+qj vk ^byk+q"^bk" + ^by k#^b k q# +
k
+ ujk+qjvk + uk vjk+qj ^byk+q"^by k# ^bk"^b k q#
In obtaining the last equality some operator pairs were anti
ommuted whi
h leads to the nal result
X
n^ (q) = e ujk+qjuk vjk+qjvk ^byk+q"^bk" + ^by k#^b k q# +
k
+ ujk+qj vk + uk vjk+qj ^byk+q"^by k# ^bk"^b k q# : (193)
An analogous
al
ulation yields
X
^ (q)
m = B ujk+qj uk + vjk+qj vk ^byk+q"^bk" ^by k#^b k q# +
k
+ ujk+qjvk uk vjk+qj ^byk+q"^by k# + ^bk"^b k q# : (194)
The rst line of these relations re
e
ts the above mentioned
oupling between
^-states, and the se
ond
re
e
ts the
oupling to the
ondensate. Both lines
ontain
oheren
e fa
tors
omposed of u and v.
10.3 Ultrasoni
attenuation
As an example of a eld (external to the ele
tron system)
oupling to the
harge density we
onsider
the ele
tri
eld
aused by a latti
e phonon. The
orresponding intera
tion term of the Hamiltonian
is g Xp
H^ I = p
!q a^y q + a^q n^(q); (195)
V q
where g is a
oupling
onstant relating the ele
tri
eld of the phonon to its amplitude, V is the
volume, ! is the phonon frequen
y and a^y its
reation operator.
56
We
onsider the attenuation of ultrasound with ~!q < ; then, in lowest order pair pro
esses do
not
ontribute. A
ording to Fermi's golden rule the phonon absorption rate may be written as
Ra (q) =
2 tr H^ I Æ(Ef Ei )H^ I P =
~
2
= 4~gV !q nq ujk+qjuk
X 2
vjk+qj vk fk (1 fjk+qj)Æ(jk+qj k ~!q ): (196)
k
Here, Ef and Ei are the total energies of the states forming the H^ I -matrix elements and nq is the
phonon o
upation number of the thermodynami
state whi
h in this
ase in extension of (187) also
ontains phononi
ex
itations in thermi
equilibrium. Half of the result0 of the last line is obtained
from the rst term in the rst line of (193). After renaming k q ! k ; the se
ond term yields the
same result. The phonon emission rate is analogously
R (q) =
2 tr H^ I Æ(Ef Ei )H^ I P =
e
~
2
= 4~gV !q nq ujk+qjuk
X 2
vjk+qj vk fjk+qj(1 fk )Æ(jk+qj k ~!q ): (197)
k
58