Superconductivity 763645S: Erkki Thuneberg

SUPERCONDUCTIVITY
763645S
Erkki Thuneberg
Department of Physics
University of Oulu
2011
Practicalities 1. Introduction
Write your name, department/group, year class, and By superconductivity we mean a phenomenon where the
email address on the list. electrical resistivity of a material (e.g. metal) disappears
The lectures are given in English if there is sufficient below some temperature.
demand for it, otherwise in Finnish. electric resistance
The web page of the course is

https://wiki.oulu.fi/display/763645S/Etusivu
The web page contains the lecture material (these notes),
the exercises and later also the solutions to the exercises.
See the web page also for possible changes in lecture and 0
exercise times. 0 Tc T
We are interested in superconductivity because of the

Time table 2011 following reasons:
Lectures: Janne Viljas, Thu 12-14, TE320
Exercises: Wed 14-16, TE320 • The theory of superconductivity is very interesting.
Examination: • Superconductivity has several technical applications.

Teaching assistant:
• Superconductivity is under active study around the
Doing exercises is essential for learning. In addition, world (e.g. in connection with quantum computing).
showing completed exercised will affect your final evalua-
tion. (You can improve by one, for example, from 3 to • We do theoretical research related to superconducti-
4.) Start calculating (at home) before the exercise time, 2 vity in Oulu, and can suggest some topics for master
hours is too short time to start from scratch. thesis.
Short content of the course
• short revisal/introduction to some fundamental re-

sults of statistical and condensed matter physics
• Thermodynamics in magnetic field
• BCS theory
• Ginzburg-Landau theory
– Type II superconductivity
• Josephson effect
Books
• M. Tinkham, Introduction to Superconductivity

(1975, 1996). Very broadly used book. More experi-
mental view, and therefore not ideal for this theory
course.
• A.L. Fetter and J.D. Walecka, Quantum theory of

many-particle systems (1971). Superconductivity stu-
died in chapters 10 and 13. The problem is that most of
the microscopic theory is treated using Green’s func-
tions, which are avoided in this course.
• J.B. Ketterson and S.N. Song, Superconductivity

(1999). Just another book.
1
• A.A. Abrikosov, Fundamentals of the Theory of Me- Meissner effect
tals (1988). Very extensive book. Half of the book
A more fundamental phenomenon than infinite conduc-
discusses normal state metals. Derivation of BCS theo- tivity is seen when a normal state metal is first placed in
ry not good. magnetic field, and is then cooled into the superconduc-
• P.G. de Gennes, Superconductivity of Metals and Al- ting state. It is observed that the magnetic field is expelled
loys (1961). Old but still useful. from the sample. This is called Meissner effect.
1.1 Properties of superconductors
Occurrence of superconductivity
normal state then cooled

• several metallic elements : Al, Nb, Sn, (but not in mag-
metal in field below Tc
netic metals and in noble metals: Cu, Au, Ag)
• many alloys, e.g. Nb-Ti Thus the fundamental property is that the magnetic field
is zero inside a superconductor. (Not only constant, as
• some compounds: Nb3 Ge, MgB2 , Yt-Ba-Cu-O etc. would follow from infinite conductivity.)
The temperature below which superconductivity occurs is

called critical temperature, Tc . The list gives some critical
temperatures.
material Tc (K) µ0 Hc (T = 0) (mT)
Al 1.196 9.9 Critical field
Hg 4.15 41 The Meissner effect is observed only if the field is not too
In 3.40 29.3 large. Let us for simplicity consider a thin bar sample that
Pb 7.19 80.3 is parallel to the field. (In this case the magnetizing field
Nb 9.25 H is constant.) It is observed that a transition between su-
Nb3 Ge 23 perconducting state and normal state takes place in critical
MgB2 39 field Hc , whose dependence on temperature is approxima-
YBa2 Cu3 06+x 98 tely
Tl2 Ca2 Ba2 Cu3 010 125
Infinte conductivity " 2 #

T
In normal state metals the electric current j is propor- Hc (T ) = Hc (0) 1 − . (3)
tional to the electric field E: Tc
j = σE. (1)
If σ → ∞ then E → 0. Maxwell’s equation
∂B H
∇×E = − (2)
∂t Hc(0) normal state
then gives that the magnetic field intensity B is constant.
Let us apply this to the case that a superconductor is coo-
led below Tc while B = 0. When the field is switched on,
it will not penetrate into the superconductor. supercon-
ducting state T
Tc
Material showing this behavior are called type I su-
perconductors.
Some superconducting materials have a mixed state
cooling then the field between Meissner and normal states. These are called type
in zero field is switched on II superconductors.
2
H This gap is somewhat less than 2kB Tc in most supercon-
normal state ductors.
Hc2 Isotope effect

mixed state Hc Different isotopes of the same element have difference in
Hc1 Tc , which depends on the ion mass M approximately as
T Tc ∝ M −1/2 . (6)
Meissner state
Tc This can be used to deduce that the motion of the ions is
important for superconductivity.
Persistent currents and flux quantization

1.2 Superfluidity
Let us place a normal state ring in perpendicular mag-
netic field. When it is cooled below Tc , the magnetic field An analog phenomenon to superconductivity is su-
is expelled form the inside of the superconductor, but a perfluidity. It means the a fluid can flow persistently (wit-
magnetic flux through the ring remains. When the exter- hout viscosity), for example, in a ring shaped tube. There
nal field is removed, this flux remains unchanged. Thus a are two well know superfluids: 4 He
persistent current I is induced in the superconducting ring 6
that generates to the

R magnetic field B. In addition, the
5
Solid
magnetic flux Φ = da · B through the ring is quantized:
Pressure (MPa)
4 (hcp)
it is an integer multiple of the flux quantum 3 (bcc)

Normal Liquid
h 2
Φ0 = = 2.07 × 10−15 Wb. (4)
2|e| 1 Superfluid
Gas
0
0 1 2 3 4 5 6
B Temperature (K)
and 3 He
4
Solid (bcc)
Pressure (MPa)
3 Superfluid
A phase
2 Superfluid
B phase
Normal liquid
I 1
Gas
0
Specific heat 0.0001 0.001 0.01 0.1 1 10 100
Temperature (K)
C The superfluid state is achieved in some gases of alkali

Cs metals 87 Rb, 7 Li, 23 Na, 1 H,. . .
Cn
1.3 History
T
• 1911 H. Kamerlingh Onnes finds superconductivity in
0 Tc mercury
The transition between normal and superconducting sta- • 1933 Meissner effect
tes is also seen in thermodynamic properties. The specific
heat has discontinuity but no latent head (in zero field). • 1935 London theory
This kind of change of state is called second order phase • 1950 Ginzburg-Landau theory
transition. In the normal state the specific heat is linear
at low temperatures. In superconducting state the specific • 1957 Bardeen-Cooper-Schrieffer theory
heat is exponential when T → 0:
• 1957 theory of type II superconductivity
∆0
Cs ∝ exp − . (5) • 1962 Josephson effect
kB T
This can be understood so that there is an energy gap ∆0 • 1986 Bednorz and Müller find “high temperature su-
between the ground state and the lowest excited states. perconductors”
3
Superfluidity 2. Fundamental results of statistical
physics
• 1938 discovery of superfluid state of 4 He In order to understand superconductivity one has to
know quantum mechanics and statistical physics. This is
• 1972 discovery of superfluid states of 3 He
a short collection of some central results of statistical phy-
• 1995 Bose-Einstein condensation in alkali atom gases sics.
Let us consider a system consisting of a large number of
particles. It is described by a Hamiltonian operator Ĥ. It
has eigenstates Ψi :
ĤΨi = Ei Ψi . (7)
It should be emphasized that the system we study has on

the order of 1023 particles, so the function Ψi (r1 , r2 , ...)
has equally many arguments.
Equilibrium
An important basic result is Gibbs distribution: the pro-
bability ρi that the state Ψi occurs is
ρi = eβ(F −Ei ) . (8)
Here the constant β can be interpreted as inverse tem-

perature: β = 1/(kB T ). Here is Boltzmann’s constant
kB = 1.38 × 10−23 J/K, which is needed to express the
temperature T inPKelvins. The constant F is determined
by the condition i ρi = 1.
The Gibbs distribution can be derived under the fol-
lowing assumptions: 1) the system we study interacts with
a much larger system called surroundings. 2) all states of
the whole system (system under study + surroundings)
occur with equal probability within some energy interval.
The most straight forward derivation of this I have seen in
the book R.P. Feynman, Statistical physics.
surroundings
system
energy
The expectation value hÂi of an arbitrary operator Â is

given by
X X
hÂi = ρi hΨi |Â|Ψi i = hΨi |Âeβ(F −Ĥ) |Ψi i = Tr(Âρ̂),
i i
(9)
where we have defined a (probability) density operator
ρ̂ = eβ(F −Ĥ) . (10)
This is the Gibbs distribution represented in operator form.

The normalization condition Trρ̂ = 1 gives
1 1 X
F = − ln(Tre−β Ĥ ) = − ln( e−βEi ). (11)
β β i
We additionally define the entropy
S = −kB hln ρ̂i (12)
4
and internal energy E = hĤi. Show as an exercise that internally in equilibrium, but the parts are not in equili-
brium with each other. For simplicity we assume that the
F = E − ST. (13) parts can only exchange heat with each other (the volumes
are constants). For both subsystems (i = 1, 2) the first law
Let us suppose that the Hamiltonian depends on a para- (17) gives
meter λ: Ĥ(λ). Differentiating the normalization condition dEi = Ti dSi . (20)
Trρ̂ = 1 show as an exercise that We define that total entropy as sum of the entropies of the
subsystems,
dĤ S = S1 + S2 . (21)
dF = −SdT + h idλ. (14)
dλ The change of entropy is given by
Supposing that λ is the volume V of the system and defi- dE1 dE2

1 1

ning the pressure dS = dS1 + dS2 = + = − dE1 , (22)
T1 T2 T1 T2
dĤ
p = −h i, (15) because the total energy is conserved. We suppose that
dV
T1 > T2 . Because T is monotonically increasing function
we get the equation (14) into the form of energy, the time derivative of E1 has to be negative,
dE1 /dt < 0, in order to the equilibrium state to be stable.
dF = −SdT − pdV. (16) This case (and also the opposite case T < T ) implies that
1 2
dS/dt > 0. Thus the entropy always grows in transition
We recognize that equations (13) and (16) are familiar from
from nonequilibrium to the equilibrium state.
thermodynamics, and they could be used to derive the first
law of thermodynamics Let us return to study a system connected to a surroun-
dings. The surroundings can be interpreted as ideal heat
dE = T dS − pdV. (17) bath, whose energy change satisfies dEb = T dSb . Here T
is constant because the surroundings is much bigger than
Thus the quantities T , S etc. can be identified as the sa- the system. Because the total system is closed, we have
me quantities as defined in thermodynamics. Especially we dEtot = dE + dEb = 0, and
identify T dS = dQ as the heat absorbed by the system, and
dStot dS dSb
we can define the specific heat in constant volume = + ≥ 0. (23)
dt dt dt
2
So we get dE − T dS ≤ 0. We define the nonequilibrium F

dQ dS d F
CV = =T = −T . (18)
dT V dT V 2
dT V by the expression F = E − T S. It follows
dF
≤ 0 (T and V constants). (24)
Next we show that the temperature T as defined above dt
has properties that we expect it to have. Thus in equilibrium the free energy F has its minimum
We study a system consisting of two parts: Ĥ = Ĥ1 + Ĥ2 . value.
According to Gibbs distribution the equilibrium has Let us restate the previous result more precisely. In no-
nequilibrium state F (T, V, λ1 , λ2 , . . .) depends on several
ρ̂ = eβ(F −Ĥ) = eβ(F1 −Ĥ1 ) eβ(F2 −Ĥ2 ) . (19) variables λi . In equilibrium it is
So we see that the temperatures T = 1/(kB β) of both F (T, V ) = minλ1 ,λ2 ,... F (T, V, λ1 , λ2 , . . .). (25)
subsystems are the same. This is the crucial property that
we require the temperature to satisfy. The condition F = E − T S = minimum generalizes the
zero-temperature condition E = minimum to finite tempe-
Usually there is no upper bound for the energy eigenva- rature.
lues Ei . In order to the Gibbs distribution (8) to be reaso-
nable, we must have β > 0, which implies T ≥ 0. Example Phase equilibrium
We see directly from the Gibbs distribution (8) that hig- Often the system can appear in two different phases, for
her temperature (smaller β) means that states with high example liquid and gas. Suppose that we have calculated
energy have larger probability. Thus the internal energy E the free energies F1 (T, V ) and F2 (T, V ) for the two phases
is a monotonically increasing function of T (assuming V is as shown in the figure.
constant). If T = 0, only the ground state (which has the
minimum Ei ) is possible. F
F1
Nonequilibrium F2
The previous analysis can be extended to nonequilibrium
systems as follows. We suppose that the system consists T
of two macroscopic parts. (The generalization to arbitra-
Tc
ry number of parts is trivial.) We suppose both parts are
5
According to the previous result, the phase that is rea- where the wave vector k appears as a parameter. The ener-
lized in equilibrium is the one having lower free energy. gy of these states is k = h̄2 k 2 /2m. In order to count the
We conclude that there is a phase transition between the states, it is most simple to require that the wave functions
phases at temperature where are periodic in a cube of volume V = L3 , which allows the
wave vectors k (nx , ny and nz integers)
F1 = F2 . (26)
2πnx 2πny 2πnz
kx =
, ky = , kz = . (35)
L L L
Variable particle number We suppose that the volume V is very large. Then we can
take the limit V → ∞ in quantities that do not essentially
Often it is mathematically easier to study case, where depend on V .
the particle number is not fixed. This can be achieved by
thinking the system as connected to a “particle bath”, an In addition to the location r, the wave function of a fer-
ideal reservoir of particles at constant energy µ, which is mion depends on spin index σ, which describes the com-
called the chemical potential. In Gibbs distribution (10) we ponent of spin angular momentum on some chosen z axis.
can generalize For spin- 12 fermions this can have two values. These can be
denoted by σ = ± 21 , or alternatively by ↑ and ↓. For free
Ĥ → Ĥ + µN̂b = Ĥ − µN̂ + constant (27) particles we can thus choose “spin-up states”
(
(because Ntot = N + Nb = constant). (Here N̂ is the √1 eik·r if σ = 21
φk↑ (r, σ) = V . (36)
particle-number operator of the system: N̂ Ψi = Ni Ψi , whe- 0 if σ = − 21
re Ni is the number of particles in state Ψi .) In making the
substitution (27) one also replaces the constant F (Helm- and “spin-down states”
holtz free energy) with another constant Ω (grand poten-
(
0 if σ = 12
tial). Therefore φ (r, σ) = . (37)
√1 eik·r
k↓
β(Ω−Ĥ+µN̂ ) V
if σ = − 21
ρ̂ = e . (28)
There are several ways to present many-body states Ψ,
In the same way as for F , one can derive for Ω the defi-
as will be discussed later. One useful way to think of these
nition (also nonequilibrium, N = hN̂ i)
states is first to list all single particle states (36)-(37) in
Ω = E − µN − ST, (29) some arbitrary order, for example
φ0↑ , φ0↓ , φk1 ↑ , φk1 ↓ , φk2 ↑ , φk2 ↓ , . . . (38)
the equilibrium expression
" # Then the basis states of the many-body space can be
1 h −β(Ĥ−µN̂ ) i 1 X
−β(Ei −µNi ) expressed by telling how many particles is in any of the
Ω = − ln Tre = − ln e ,
β β single-particle states,
i
(30) |Ψi i = |n1 , n2 , n3 , . . . , n∞ i. (39)
the differential for equilibrium states
Here nα is the number of particles in the α’th single-
dΩ = −SdT − pdV − N dµ (31) particle state (38). [In practice the writing of the state (39)
is difficult because there is an infinite number of single-
and the time development particle states and thus the great majority of the numbers
dΩ nα areP zeros.] The energy of the many body state (39) is
≤ 0 (T , V and µ constants). (32) E = α nα α . Fermions obey the Pauli exclusion principle
dt
and thus all occupations nα are either 0 or 1.
In addition we deduce from equation (31)
In order to determine the equilibrium state, it is easiest
to use the formulas for variable particle number. Starting

∂Ω ∂Ω ∂Ω
S=− , p=− , N =− . from equation (30) we get
∂T V,µ ∂V T,µ ∂µ T,V
(33) e−βΩ = Tre−β(Ĥ−µN̂ )
XX
= . . . hn1 , n2 , . . . |e−β(Ĥ−µN̂ ) |n1 , n2 , . . .i
Ideal Fermi gas n1 n2
XX
Above we formally discussed the many-body wave func- = . . . e−β(1 −µ)n1 e−β(2 −µ)n2 . . .
tions Ψi . The calculation of these is possible only in very n1 n2
special cases. One case is ideal gas, where we assume that X X
= e−β(1 −µ)n1 e−β(2 −µ)n2 . . .
there are no interactions between the particles.
n1 n2
The natural choice for wave functions of a single free =
YX
e−β(α −µ)nα
particle are plane wave states α nα
Yh i
1 ik·r = 1 + e −β(α −µ)
. (40)
φk (r) = √ e , (34)
V α
6
Thus where the factor 2 comes from spin. From this we get a
relation between the Fermi wave vector and the particle
1 Y h i 1X h i
Ω = − ln 1 + e−β(α −µ) = − ln 1 + e−β(α −µ) . density,
β α
β α N k3
(41) = F2 . (46)
V 3π
From this we can calculate all thermodynamic quantities.
Especially the particle number (31) is When T > 0, the occupation f () gets rounded so that
the change from f ≈ 1 to f ≈ 0 takes place in the energy
1 X e−β(α −µ) β interval ≈ kB T .

dΩ
N =− =
dµ T,V β α 1 + e−β(α −µ) kBT
f T=0
X 1
= β( −µ)
. (42)
α
e α +1
T>0
Here we see that the average occupation probability of each
state depends on the energy and is ε
0 εF
1
f () = β(−µ) . (43)
e +1
This is the familiar Fermi-Dirac distribution.
In a similar fashion we can derive for an ideal Bose gas
(possible occupation numbers nα = 0, 1, 2, . . . , ∞) the
Bose-Einstein distribution
1
f () = . (44)
eβ(−µ) −1
The purpose above was to show that ideal Bose and

Fermi distributions can be derived from the more gene-
ral Gibbs distribution, which can be applied to arbitrary
interacting systems as well.
Let us remind about the main features of Fermi distribu-
tion. When the temperature T → 0, the occupation beco-
mes a step function

1 for < µ
f () = (45)
0 for > µ.
Here the maximal kinetic energy of an occupied state is

called Fermi energy F = kB TF , and expressed in tempe-
rature units it is called Fermi temperature TF . We also
define the Fermi wave vector kF so that h̄2 kF2 /2m = F .
In momentum space all states inside (k < kF ) of the Fermi
surface (k = kF ) are occupied, and the ones outside are
empty.
ky
× × × × × × × × × × × × × × ×
× × × × × × × × × × × × × × ×
× × × × × × × × × × × × × × ×
× × × × × × × × × × × × × × ×
kF
× × × × × × × × × × × × × × ×
× × × × × × × × × × × × × × ×
× × × × × × × × × × × × × × × kx
× × × × × × × × × × × × × × ×
× × × × × × × × × × × × × × ×
× × × × × × × × × × × × × × ×
× × × × × × × × × × × × × × ×
× × × × × × × × × × × × × × × 2π
× × × × × × × × × × × × × × ×
× × × × × × × × × × × × × × × L
× × × × × × × × × × × × × × ×
Equation (42) gives at zero temperature

4 3
3 πkF
X
N =2 1=2 ,
(2π/L)3
k<kF
7
3. Thermodynamics of supercon- of the current source just by the change of the additional
ductors term. (more details as an exercise)
When we study the thermodynamics of a magnetic mate- Let us apply this to a long cylindrical superconducting
rial, we should consider the field as an additional variable. sample in parallel magnetic field When H increases from
The equation (14) can be written (supposing the volume zero, equation (56) gives
V constant) Z H
dF = −SdT + V H · dB. (47) G(T, H) − G(T, 0) = −V dH 0 B(H 0 ). (57)
0
In the normal way one gets from this
In the normal state of a superconductor, the magnetization
∂F 1 ∂F is very weak. Thus to a good approximation B = µ0 H
S=− , H= , (48)
∂T B V ∂B T (Appendix A). We get
where the vector relation should be understood as Hx = Gn (T, H) − Gn (T, 0) = − 21 V µ0 H 2 . (58)

∂F/V ∂Bx etc.
In the superconducting state B = 0, so we get
Here is more detailed justification of equation (47). The
Maxwell equations in presence of material are (Appendix Gs (T, H) = Gs (T, 0). (59)
A)
Because these two states are in equilibrium in the critical
∇·D = ρf , (49) field Hc , their potentials must be the same (26):
∂B Gn (T, Hc ) = Gs (T, Hc ). (60)
∇×E = − , (50)
∂t
∇ · B = 0, (51) From above we deduce
∂D Gs (T, 0) = Gn (T, 0) − 21 V µ0 Hc2 . (61)
∇×H = + jf . (52)
∂t
So the energy of the superconducting state is lower than
Together with material relations, e.g.
that of the normal state by the condensation energy
1 2
D(E) ja H(B), (53) 2 V µ0 Hc .
Rearranging we get from formulas (58), (59) and (61)
they form a complete set of equations. From these one can
derive the conservation law Gs (T, H) − Gn (T, H) = 21 V µ0 (H 2 − Hc2 ). (62)
Z
dV E · dD + H · dB + E · jf dt
According to equation (56) we see that the corresponding
I difference in entropy is obtained by derivation:
= −dt da · (E × H). (54) dHc (T )
Ss (T, H) − Sn (T, H) = V µ0 Hc (T ) . (63)
dT
The left hand side is energy differential in the system and Because H (T ) decreases as the temperature increases
c
the right hand side is the flux of electromagnetic energy (equation 3), we see that the entropy in the superconduc-
through the surface of the system. When D = 0 and jf = ting state is lower than in the normal state. Note that the
0, we get the magnetic energy change of a homogeneous entropy difference (63) is independent of the field H. From
system as given by equation (47). equations (63) we can also deduce that the latent heat in
In order to study the sample in a given external field H, the transition between normal and superconducting states
we have to add to F an energy term (“bath”) that gives rise T (Ss − Sn ) vanishes at T = 0 and at T = Tc , but not at
to this field. The needed term can be justified as Legendre intermediate temperatures.
transformation from variable B to H. So we get as the The specific heat in constant field can be calculated using
quantity to be minimized
∂S
G = F − V H · B. (55) CH = T (64)
∂T H
and its differential in equilibrium state From this we get the discontinuity of the specific heat
dG = −SdT − V B · dH. (56) Cs (T, H) − Cn (T, H)
" 2 #
dHc d2 Hc
More physically, the additional term in equation (55) can = T V µ0 + Hc . (65)
dT dT 2
be understood as the energy of the current source. The cur-
rent source drives current to a coil that generates the mag- In the special case T = Tc this reduces to
netizing field H. (Both the coil and the current source are 2
idealized as dissipationless.) When B in the sample chan- dHc
Cs (T, H) − Cn (T, H) = µ0 V Tc . (66)
ges, it induces a voltage in the coil and changes the energy dT Tc
8
In summary, the thermodynamics sets some conditions where D = Dn + Ds . The minimization of this would lead
between different measurable quantities. to minimizing the interface by D → ∞. This tendency is
opposed by the fact that then the field outside the slab
3.1 Intermediate state would be very inhomogeneous. In vacuum H = B/µ0 and
the energy density is 2µ1 0 B 2 . We estimate the order of mag-
The discussion above is valid for a thin sample that in nitude of the additional energy caused by the inhomoge-
parallel field, where the magnetizing field H ≈ constant. neity of the field by
[Show this starting from equation ∇×H = 0 (52).] In other
cases H(r) 6= constant. For example, for a sphere we get µ0 2 Dn 1 2 2Ds Dn
finhomog = H − B (70)
that the field at polar angle θ = π/2 is H = 32 H(r = ∞). 2 c D 2µ0 ∞ D
The expression in parenthesis gives the field energy near

the slab (weighted by the area factor Dn /D), from which
the energy of homogeneous field is subtracted. The factor
2Ds Dn /D ≈ 2 min(Ds , Dn ) estimated the thickness where
the inhomogeneity is important. The total energy is
" 2 2 #
2 dδ B∞ B∞
f = µ0 Hc +D 1− (71)
H = constant = H∞ H = (3/2)H∞ D µ0 Hc µ0 Hc
Minimizing with respect to D we get

If H(r = ∞) > 23 Hc , the critical field is exceeded on the √
equator. On the other hand, the whole sphere cannot go dδ
into the normal state below H(r = ∞) = Hc . Between the- D= (72)
B∞
1 − µB0 H
∞
se fields one gets so called intermediate state intermediate µ0 Hc c
state (of type I superconductor).

The essential thing is the numerator, which gives the geo-
Let us consider superconducting slab in perpendicular metric mean of a microscopic and macroscopic length, δ
magnetic field. We estimate orders of magnitude assuming and d.
the intermediate state consists of alternating layers of nor-
mal and superconducting states (thicknesses Ds and Dn ). Experiments show alternating normal and superconduc-
ting layers that can for quite complicated structures.
Ds d
Dn ≈DsDn/D
D = Dn+Ds B∞
B = µ0Hc
B=0
It follows from Maxwell equations (51) that the average
flux through the slab is the same as the field B∞ far from
the slab. This goes through the normal state part only.
In order the normal and superconducting states to be in
equilibrium, the normal state must have B = µ0 Hc . This
gives
B∞ (Ds + Dn ) = µ0 Hc Dn . (67)
It is obvious that there is some extra energy associated

with the interface between normal and superconducting
Intermediate state of indium, in which superconducting
states, which is proportional to the area A of the interface.
regions (black) are decorated with niobium. The applied
We write this energy in the form
field is close to the critical field (H/Hc = 0.931). F. Haenss-
µ0 2
Fsurface = σA, σ = H δ, (68) ler and L. Rinderer, Helv. Phys. Acta 40, 659 (1967).
2 c
The intermediate state described above occurs in type I
where δ is some microscopic quantity of dimension length. superconductors. Type II superconductors show essentially
The surface energy per unit area of the slab is different behavior in magnetic field, as will be seen later.
2dσ µ0 2 2dδ
fsurface = = H , (69)
D 2 c D
9
4. Microscopic theory where
mkF
4.1 Normal state metal N (0) = (77)
2π 2 h̄2
The properties of metals are more thoroughly studied
in the course of condensed matter physics. Here we only
consider a very simplified model of a metal, which still is is the density of states at the Fermi surface. [The notation
sufficient as a starting point for understanding basic pro- N (0) comes from the fact that often it is most convenient to
perties of superconductivity. count the energy starting from the Fermi level. In conden-
We assume that there are conduction electrons, which sed matter physics the zero value of the energy can usually
can move freely like ideal gas through the metal. We assume be chosen freely.]
that the rest of the electrons called valence electrons are In the free electron model it is easy to calculate the speci-
bound to the atomic nuclei so that they do not contribute fic heat. The result is that C is linearly proportional to the
to the electric conductivity. The kinetic energy k of the temperature when T TF . This explains the specific heat
conduction electrons depends on the wave vector k. In the of the normal metal mentioned in the introduction. This
simplest case this dependence is of the form will be calculated in detail later on in this course.
h̄2 k 2 How can we understand the electrical resistivity of a nor-

k = . (73) mal metal? At low temperatures the main reason for elect-
2m
rical resistance is that the conduction electrons collide with
Here m is the effective mass, which can be different from impurities that occur in the metal. If the ideal gas has net
the free electron mass. We can estimate that the Fermi momentum in the beginning, it will decay gradually as in-
energy F is on the order of a few eV corresponding to dividual electrons collide with impurities and are scattered
TF ≈ 105 K. This should be compared with the typical in random directions.
temperatures T ∼ Tc ∼ 10 K that occur in superconduc-
tivity. The theory thus has two very different energy sca-
les, Tc /TF ∼ 10−4 . We will often make use of the relation
T TF .
ky
kF
kx
Excitation picture
g≠0
Previously we discussed that the ground state (= lowest
energy state) of a Fermi gas is such that all single-particle
states inside the Fermi surface are filled and the states
Example One often encounters integration over k space outside are empty. The most simple excited states are of
of a function g(k). Show as an exercise that two different types.
Z 2 Z
1 X d Ω
3
g(k) = d N ()g(k). (74)
L 4π • one extra electron is in state with k whose k > µ.
k
This excitation has energy ξk ≡ k −µ and momentum
This contains integration over the solid angle p = h̄k.
Z Z 2π Z π
2
d Ω= dφ dθ sin θ, (75)
0 0
• one electron is missing from state with k whose k < µ.
where θ and φ are the polar and azimuthal
R angles of k). This is called a hole. Its excitation energy is µ − k > 0
The second integration is over energy, d. The factor N () and momentum p = −h̄k.
is the density of states per unit energy (and per volume).
If g(k) is different from zero only near the Fermi surface
(figure), we can further approximate In both cases the excitation energies are positive. The
figure illustrates the same hole type excitation seen both
d2 Ω
Z Z
1 X
g(k) = N (0) d g(k), (76) as a missing particle and as an excitation. (The momentum
L3 4π space is simplified as one-dimensional in the figure.)
k
10
particle view excitation view Let us study the wave functions of spin- 21 fermions. The
single-particle wave function φ(rσ) [e.q. (36) or (37)] de-
|εk-µ| pends on the location r and on the spin index σ = ± 12 .
In general form the Pauli exclusion principle says that the
εk wave function has to be antisymmetric in any exchange of
p two electron coordinates. For a two-electron wave function
µ 0 Ψ(r1 σ1 , r2 σ2 ) this requirement is
Ψ(r1 σ1 , r2 σ2 ) = −Ψ(r2 σ2 , r1 σ1 ). (79)

p
Starting from an arbitrary function ψ0 (r1 σ1 , r2 σ2 ), one can
If the particle number is fixed, the particle and hole type always by antisymmetrizing construct a function that sa-
excitations must always appear as pairs. tisfies this condition:
Ψ(r1 σ1 , r2 σ2 ) = ψ0 (r1 σ1 , r2 σ2 ) − ψ0 (r2 σ2 , r1 σ1 ). (80)

4.2 Ideas about superconductivity
Materials can be studied on two very different scales: Thus the pair state formed from two single-particle states
φ1 and φ2 has the wave function
• atomic scale. Quantum mechanics is essential. There
is no friction. Ψ(r1 σ1 , r2 σ1 ) = φ1 (r1 σ1 )φ2 (r2 σ2 ) − φ1 (r2 σ2 )φ2 (r1 σ1 ),
(81)
• macroscopic scale. Material is described used laws of so called Slater determinant. We see immediately that this
classical physics (e.g. theory of elasticity). Friction is vanishes, Ψ ≡ 0, if the single-particle states are the same:
essential. φ1 = φ2 . Thus two fermions cannot be placed into the same
single-particle state.
The phenomena on atomic scale are not usually visible
on the macroscopic scale. The reason for this is that mac- In case of many particles, the antisymmetry is required
roscopic bodies consists of enormously large number of par- in any pairwise exchange of the coordinates of two particles
23
ticles (∼ 10 ). The different particles are in general in dif-
Ψ(. . . , ri σi , ri+1 σi+1 , . . . , rk σk , . . .)
ferent quantum states, and only the average of them is vi-
sible. Friction is needed because it is not possible to take = −Ψ(. . . , rk σk , ri+1 σi+1 , . . . , ri σi , . . .) (82)
into account the motion of individual particles.
Idea: Superconductivity is an exception to the rule abo- Idea: fermions form pairs.
ve: it is a quantum phenomenon that is still preserved on
a macroscopic scale. ψ0 (r1 σ1 , r2 σ2 , r3 σ3 , . . .) = φ(r1 σ1 , r2 σ2 ) ×
φ(r3 σ3 , r4 σ4 )φ(r5 σ5 , r6 σ6 ) . . . , (83)
How is this possible? As a simple example, let us con-
sider ideal Bose gas. At zero temperature its distribution
All pair states are the same!. This function does not va-
function (44) reduces to the form
nish in antisymmetrization as long as the pair function is
antisymmetric,

N lowest single-particle state (i = 0 )
f () = (78)
0 other single-particle states (i > 0 ).
φ(r1 σ1 , r2 σ2 ) = −φ(r2 σ2 , r1 σ1 ). (84)
Also at finite temperatures below so-called Bose conden-
sation temperature, the occupation of the lowest state In exchanging pairs one gets the factor (−1)2 = 1, similarly
N0 /N > 0, whereas for all other single-particle states to bosons.
fi /N → 0 when N, V → ∞. It is expressed by saying that
We see that superconductivity could arise from mac-
the wave function of the lowest state becomes macroscopic
roscopic occupation of a pair state. Before we can accept
wave function because a macroscopic number of particle is
this claim, we must be able to answer the following ques-
in the same single-particle state.
tions.
It can be said that the superfluid phases of alkali atom
gases and 4 He liquid are based on Bose condensation. • Is there a force that binds the pairs?
However, the ideal Bose gas is a too simplified model for
them because more detailed study shows that the interac- • Is it sufficiently strong?
tions between particles have an essential role in the su-
perfluid state. In this course we will not study this interes- • Does the pair state have a lower energy than the nor-
ting problem more. mal state?
Electrons are fermions. One can put only one fermion
into a one single-particle state. Thus the discussion above
as such cannot explain the superconductivity of metals. 4.3 Cooper problem
11
Let us study two fermions that interact with each other Substituting back on the right hand side and cancelling
but not with other fermions at temperature T = 0. We the common factor I, we get
ignore the spin for a while. The Schrödinger equation is
1 1 X 1

h̄2 2 h̄2 2
= 3 . (92)
− ∇1 − ∇2 + V (r1 , r2 ) φ(r1 , r2 ) = Eφ(r1 , r2 ). g L F <k <F +c
2 k −E
2m 2m
(85) Applying the integration formula (76) we get
It is likely that in the lowest energy state, the center of the
mass of the pair is at rest. Therefore we write the wave Z F +c
1 1 1 2F − E + 2c
function as dependent only on the difference r1 − r2 , = d = ln . (93)
N (0)g F 2 − E 2 2F − E
1 X
φ(r1 , r2 ) = 3 χ(k)eik·(r1 −r2 ) . (86) We make weak coupling approximation gN (0) 1, or equi-
L
k
valently 2EF − E 2c , in order to get the formula into
We substitute this into the Schrödinger equation (85). We the form
0
multiply by e−ik ·(r1 −r2 ) , integrate over r1 :n ja r2 and get E = 2F − 2c e−2/gN (0) . (94)
X
0 0 0
(2k0 − E) χ(k ) = − hk , −k |V |k, −kiχ(k). (87) We see that the energy is lower than the energy 2F of a
k noninteracting pair when g > 0 (90).
Here we have used the notation In is interesting to note that this bound state is formed
Z Z
1 even for arbitrarily weak attractive interaction. [When the-
hk10 , k20 |V |k1 , k2 i = 6 d3 r1 d3 r2
L re are no other particles, the attractive interaction for for-
0 0 ming a bound state must exceed some threshold value. This
×e−ik1 ·r1 e−ik2 ·r2 V (r1 , r2 )eik1 ·r1 eik2 ·r2 , (88)
corresponds to the case kF = N (0) = 0 (77).]
which looks slightly clumsy here but is useful later.
The wave function (91) of the pair is independent of the
The presence of other electrons is taken into account direction k of the wave vector. [Then, according to equation
only through the Pauli exclusion principle. They restrict (86), the wave function in the r space is independent of the
the pair function so that direction of r − r 0 .] Thus the pair is formed in s-wave state
χ(k) = 0 for k < kF . (89) (not p, d, f . . . ). The s0 state is symmetric in the exchange
of coordinates r and r . In order the total wave function to
We proceed from equation (87) by assuming that V is a be antisymmetric, the spin state must be antisymmetric,
constant in a thin layer on both sides of the Fermi surface, and thus singlet:
and zero elsewhere, 1
φtot (rσ, r 0 σ 0 ) = φ(r − r 0 ) √ [δσ, 21 δσ0 ,− 12 − δσ,− 21 δσ0 , 12 ].
hk0 , −k0 |V |k, −ki (90) 2

3
−g/L for |k0 − F | < c ja |k − F | < c , (95)
=
0 otherwise, As a function of the magnitude of the wave vector k, the
where we assume that g > 0 and c F . [The cutting of wave function has its maximum at the Fermi surface, and
the interaction (90) is slightly problematic since according decreases when k moves away from it. It is thus likely that
0 0
to (88) the interaction hk , −k |V |k, −ki depends only on the cut off of the interaction (90) is not essential.
k0 − k. Later we will see that indeed the interaction has a We see that an attractive interaction makes a Fermi gas
more complicated form. The cutting also will not have a unstable against formation of pairs at T = 0.
serious effect since χ(k) will be small when k ≈ F + c .]
We see that inside of the layer the right hand side of the
equation (87) is independent of k0 , and we mark it by I.
4.4 Attractive interaction
From the left hand side we get easily Between electrons there is the repulsive Coulomb inte-
raction
I 1 e2
χ(k) = for k > kF . (91) V (r) = . (96)
2k − E 4π0 r
What could cause an attraction so that pairs could be for-
med?
Warning: in the following we only attempt to make it
plausible that an attractive interaction might exist.
The first observation is that a metal has a great number
χ
of conduction electrons and ions. (By an ion we mean the
nucleus and the valence electrons of the atom.) The whole
ε
system is charge neutral because of the ions. When we look
εF+εc at the interaction between electrons, we also should take
E/2 εF
into account also other electrons. These tend to go away
12
from a negative charge, and thus the electron is surroun- (En(0) = Em
(0)
) is
ded by a net positive charge from the ions. This is called
screening of the electron charge. In the simplest case this hψm |V eff |ψn i = hψm (0)
|H1 |ψn(0) i
leads to a potential (µ is constant) X hψm (0) (0) (0)
|H1 |ψα ihψα |H1 |ψn i
(0)
+ (0) (0)
1 e2 exp(−µr) α Em − Eα
V (r) = , (97)
4π0 r +O(H13 ), (100)
where the potential decays exponentially at large distances.
where α goes through all nondegenerate states.
A more accurate calculation gives also an oscillating com-
ponent to the potential. However, the attraction caused by We estimate the effect of lattice vibrations on the inte-
this overscreening is too weak to cause formation of pairs raction between two electrons by calculating the effective
in usual superconductors. matrix element (100) in the case where the initial state ψn
has two electrons at wave vectors k1 and k2 and the final
It is found experimentally that the superfluid state de-
state ψm at wave vectors k10 and k20 . The first order term
pends on the mass of the ion (isotope effect). The ion mass (0)
can appear because the ion lattice is not at rest but oscil- vanishes. The intermediate state in second order ψα has
lates. In the following we study what effect the ion motion two alternatives
has on the interaction between electrons. k1'=k1-q k2'=k2+q k1'=k1-q k2'=k2+q
Lattice vibrations are studied in more detail in the course q -q
intermediate state
of condensed matter physics. Here we only state that the
eigenstates of the lattice are vibrations that have a well k1 k2 k1 k2
defined wave vector k and a frequency ω(k), which de-
pends on the wave vector. According to quantum mecha-
nics, the vibrations are quantized, i.e. they consists of pho- Assuming the matrix elements are constants, we get the
nons that have energy h̄ω(k) ja quasimomentum h̄k. This corresponding terms
is analogous to the photos, which are oscillation quanta of
the electromagnetic field. hk1 − q, k2 + q|V eff |k1 , k2 i
The coupling of lattice vibrations with the electrons is
X |Vq |2
=
described as follows. An electron with momentum h̄k emits q
(k1 ) − (k1 − q) − h̄ω(q)
a phonon with momentum h̄q, and the momentum remai- X |Vq |2
ning for the electron is h̄k0 . This process has the matrix + + O(H13 ).
(k 2 ) − (k 2 + q) − h̄ω(−q)
element q
Z
hψm(0) 0
|H1 |ψn(0) i ∝ dV eik·r e−iq·r e−ik ·r . (98) Now we use energy conservation (k1 ) + (k2 ) = (k10 ) +
(k20 ) and because ω(−q) = ω(q), we get
It follows that the momentum is conserved, k0 = k−q. The hk1 − q, k2 + q|V eff |k1 , k2 i
process is often depicted by the graphs (a) in the figure. X 2h̄ω(q)|Vq |2
(a) (b) =−
q
[h̄ω(q)] − [(k1 ) − (k10 )]2
2
k+q
q k-q final state +O(H13 ) (101)
k q In the Cooper problem the total momentum of the
k initial state
pair was assumed zero and therefore the used potential
hk0 , −k0 |V |k, −ki is obtained as the special case k2 = −k1 .
Correspondingly, there must exist a process (b), where
the electron absorbs a phonon. We see that the interaction (101) is attractive when
h̄ω(q) > |(k1 ) − (k10 )|. The former quantity is typical-
We study the effect of lattice vibrations on electrons
ly on the order of the Debye temperature ∼ 100 K, which
using perturbation theory. We recall time independent per-
means that this inequality possibly is satisfied. Ultimate-
turbation theory (Quantum mechanics I, nondegenerate
ly, the formation of pairs depends if this attractive force is
case), which gives
sufficiently strong that it wins the repulsive Coulomb force.
(0) (0)
En = En + hψn |H1 |ψn i (0) With some more work one could show that these two forces
have the same order of magnitude. Therefore, the forma-
X |hψα(0) |H1 |ψn(0) i|2
+ (0) (0)
3
+ O(H1 ). (99) tion of Cooper pairs and thus superconductivity depends
α6=n E n − E α on the detailed structure of each metal.
In a similar way but with slightly more work (Landau-

Lifshitz, Quantum mechanics) one can show that the effec- 4.5 Creation and annihilation operators
tive interaction between two degenerate states ψm and ψn This part is well presented in FW pages 3-19.
13
Above we already studied many-body states a bit. Be- Although the (anti)symmetrization requirement (103)
fore we continue, it is good to introduce a new notation, was automatically satisfied with the notation (104), dea-
which is commonly used to describe many-body systems. ling with these states is still difficult due to the long list of
This involves the introduction of the so-called creation and numbers. To get rid of this, let us first study purely mat-
annihilation operators. These are also referred to as the hematically the operators b̌k and b̌†k , where the latter is
operators of second quantization (we talk about the names the Hermitian conjugate of the former with respect to so-
later). me inner-product space. Let us require these operators to
The basic idea, which appears also more generally, is the satisfy the following commutators ([Ǎ, B̌] ≡ ǍB̌ − B̌ Ǎ)
following. Originally we have some notation, where the- h
†
i
re are extra (unphysical) degrees of freedom. Then it ma- b̌ k , b̌k 0 = δk,k0 ,
kes sense to take into use a notation which automatically [b̌k , b̌k0 ] = [b̌†k , b̌†k0 ] = 0 (106)
excludes the unphysical degrees of freedom.
Examples: We claim that these conditions (alone) are yield the fol-
lowing relations
• moving from coordinates x, y and z to a vector r, †
b̌ b̌k | . . . , nk , . . .i = nk | . . . , nk , . . .i, nk = 0, 1, 2, . . . , ∞
which is not dependent on the choice of the coordinate k √
system b̌k | . . . , nk , . . .i = nk | . . . , nk − 1, . . .i
† √
b̌k | . . . , nk , . . .i = nk + 1 | . . . , nk + 1, . . .i, (107)
• going over from time t and position r to the four-space
xα in relativity theory where the states are of the form (104). Interpretation: b̌† b̌ k k
is a particle-number operator, b̌k is an annihilation opera-
The case under study here is a many-body system. As
tor, and b̌†k a creation operator for bosons.
a starting point we have some single-particle basis states
φi (r, σ) [e.g. (34)], i = 1, . . . , ∞. From these we form simple In the literature no hat seems to be written on the b̌k
n-particle states as a the product operators. Here it has been added so that all operators of
second quantization would be denoted uniformly.
Ψ(r1 σ1 , . . . , rn σn ) = φi1 (r1 σ1 )φi2 (r2 σ2 ) . . . φin (rn σn )
(102) Proof. (read at home) We study the operator
Based on experimental evidence it has been judged that
ň = b̌† b̌. (108)
all such states are not found in nature, but that physical
states must additionally satisfy a symmetry with regard to It is Hermitian (verify). The eigenvalues of a Hermitian
exchange of indistinguishable particles: operator are real valued. Let us label the eigenstates of the
operator ň by using the eigenvalue n (a real number)
Ψ(. . . , ri σi , ri+1 σi+1 , . . . , rk σk , . . .)
= ±Ψ(. . . , rk σk , ri+1 σi+1 , . . . , ri σi , . . .), (103) ň|ni = n|ni. (109)
where + is for bosons and − is for fermions. We can make We assume the eigenstates are normalized, hn|ni = 1. We
states of the type (102) physical by symmetrizing (bosons) see that the eigenvalue n cannot be negative,
of antisymmetrizing (fermions) them. A general many- X
body state is a linear combination of (anti)symmetrized n = hn|ň|ni = hn|b̌† b̌|ni = hn|b̌† |mihm|b̌|ni
states. m
X
2
Writing the wave function in terms of coordinates in = |hm|b̌|ni| ≥ 0. (110)
unsatisfying because it does not directly forbid writing m
down un-(anti)symmetrized functions. We can deal with We easily calculate
this as already discussed earlier (39). Let us write all single-
particle wave functions (34) in some arbitrary order φ1 , φ2 , [ň, b̌] = −b̌. (111)
. . . , φ∞ . Then we express the (noninteracting) many-body
state by expressing how many times each single-particle This implies
state is included:
ň(b̌|ni) = b̌ň|ni − b̌|ni = (n − 1)(b̌|ni), (112)
|n1 , n2 , . . . , n∞ i, (104)
and we see that b̌|ni either is a state corresponding to the
(ni = 0 or 1 for fermions; ni = 0, 1, . . . , ∞ for bosons). The eigenvalue n − 1, i.e. b̌|ni = c|n − 1i, or else b̌|ni = 0. The
basis states obtained in this way can be used to express an latter alternative implies ň|ni = 0, so it is possible only if
arbitrary (interacting) many-body state in the form n = 0. In the former case normalization gives
X
|Ψ(t)i = cn1 ,n2 ,...,n∞ (t)|n1 , n2 , . . . , n∞ i. (105) |c|2 = hn|b̌† b̌|ni = n, (113)
n1 ,n2 ,...,n∞
and therefore we fix
√
Bosons b̌|ni = n |n − 1i. (114)
14
If one operates sufficiently many times with b̌, one should and a term where the particles have been exchanged with
arrive at negative eigenvalues, which is in contradiction each other. More in the exercises.
with equation (110). The way out of this is that n is an
integer, so that b̌|0i = 0, and the process (114) ends. (Note Fermions
the essential difference between the n = 0 eigenstate |0i
and the zero of the linear space 0.) Correspondingly one The development in the case of fermions differs in that
can deduce results for b̌† . Inserting the indices k we get all we require the anticommutators {Ǎ, B̌} ≡ ǍB̌ + B̌ Ǎ and
formulas (107). change the letter just to be sure
n o
The many-body Schrödinger equation in coordinate ǎk , ǎ†k0 = δk,k0 ,
representation is (we forget spin)
{ǎk , ǎk0 } = {ǎ†k , ǎ†k0 } = 0. (122)
∂Ψ
ih̄ (r1 , . . . , rn , t) = HΨ(r1 , . . . , rn , t), (115) It is now claimed that these conditions alone yield the fol-
∂t
lowing relations (exercise)
where
ǎ†k ǎk | . . . , nk , . . .i = nk | . . . , nk , . . .i, nk = 0, 1
X 1 XX
H= T (ri ) + V (ri , rj ) ǎk | . . . , 0k , . . .i = 0
i
2 i j
j6=i
ǎk | . . . , 1k , . . .i = | . . . , 0k , . . .i
X h̄ 2
1 XX ǎ†k | . . . , 0k , . . .i = | . . . , 1k , . . .i
=− ∇2i + V (ri , rj ). (116)
i
2m 2 i j ǎ†k | . . . , 1k , . . .i = 0, (123)
j6=i
where 0k and 1k mean numbers 0 and 1 in the argument
How to write this for states (105)? That is, when k of state (104). Interpretation: ǎ†k ǎk is a particle-number
operator, ǎk an annihilation operator, and ǎ†k a creation
∂ operator for fermions.
ih̄ |Ψ(t)i = Ȟ|Ψ(t)i, (117)
∂t
The Hamilton operator in second quantization
what is Ȟ? X † 1X † †
Ȟ = ǎi hi|T |jiǎj + ǎi ǎj hi, j|V |k, liǎl ǎk . (124)
The answer: 2
i,j ijkl
X † 1X † †
Ȟ = b̌i hi|T |jib̌j + b̌i b̌j hi, j|V |k, lib̌l b̌k . (118) is exactly of the same form as for bosons. Now in particu-
2
i,j ijkl lar the order of the operators ǎl ǎk is essential, because an
opposite order would change the sign. Exercises.
Here hi|T |ji and hi, j|U |k, li are familiar matrix elements
calculated for single-particle states The interaction potential is assumed to depend only on
the distances between particles: V (r, r 0 ) = V (r − r 0 ). In
h̄2 2 general the interaction potential could also depend on the
Z
3 ∗
hi|T |ji = d rφi (r) − ∇ φj (r) (119) spins of the particles. However, let us study only spin-
2m
hi, j|V |k, li (120) independent interactions, i.e. the interaction cannot chan-
Z Z ge the spins of the particles, and the value of the potential
= d3 r1 d3 r2 φ∗i (r1 )φ∗j (r2 )V (r1 , r2 )φk (r1 )φl (r2 ). does not depend on the spins.
Exercise: calculate the matrix elements (119) and (120)
Note in particular that hi, j|V |k, li is calculated so that using the plane wave states (36)-(37) and then derive the
one electron scatters from state k to state i and the other second-quantized Hamilton operator
from state l to state j. Also note that (88) used above
1 X X XX
k ǎ†kσ ǎkσ +
X
is a special case of the formula (120), where plane waves Ȟ = V (k3 − k1 )
φk (r) = (1/L3/2 )eik·r (34) have been used. Using these 2L3
k,σ k1 ,σ k2 ,λ k3 k4
states the kinetic energy matrix element is ×δk1 +k2 ,k3 +k4 ǎ†k3 σ ǎ†k4 λ ǎk2 λ ǎk1 σ , (125)
h̄2 2 ik·r
Z
0 1 3 −ik0 ·r where spin indices σ and λ can have the values ↑ and ↓.
hk |T |ki = 3 d re − ∇ e
L 2m Note that due to the Kronecker delta the momentum is
h̄2 k 2 conserved also in interactions. Let us also remind that the
= δk0 ,k . (121) Fourier transformation is defined as
2m Z
V (k) = d3 rV (r)e−ik·r , (126)
Instead of deriving (118) generally (see e.g. FW) we only
P certain cases. 1) Normally hi|T |ji so that its inverse transformation is
note that is sensible in
diagonal ⇒ Ȟkin = i ňi hi|T |ii. 2) Interaction term V̌
does not count interaction of a particle with itself. 3) In 1 X
V (r) = 3 V (k)eik·r . (127)
a two-particle state hV̌ i gives both the direct interaction, L
k
15
†
As a special case let us inspect an interaction that can (133) is called canonical. The operators γ̌kσ ja γ̌kσ can
be described with a delta function be interpreted as annihilation and creation operators of an
excitation. Inside the Fermi sphere we use
V (r, r 0 ) = −gδ(r − r 0 ). (128)
ǎ†kσ ǎkσ = 1 − ǎkσ ǎ†kσ (134)
For this we get (125) in the form
and using the notation (133) we get
g X X XX
k ǎ†kα ǎkσ −
X
Ȟ = X †
2L3 Ǩ = |ξk |γ̌kσ γ̌kσ + Ω0 , (135)
k,σ k ,σ k ,λ k
1 k 2 3 4
k,σ
×δk1 +k2 ,k3 +k4 ǎ†k3 σ ǎ†k4 λ ǎk2 λ ǎk1 σ . (129)
where the last term is a constant (the grand potential of
With a change of variables k1 ↔ k2 and using (122), we the ground state). The ground state |0i satisfies
can see that a nonzero interaction is obtained only in the
case that the spins are opposite: γ̌kσ |0i = 0 ∀k, σ. (136)
g XXXX †
The number operator of the excitation kσ is γ̌kσ γ̌kσ , and
k ǎ†kα ǎkσ − 3
X
Ȟ =
L all excitation energies are positive.
k,σ k1 k2 k3 k4
×δk1 +k2 ,k3 +k4 ǎ†k3 ↑ ǎ†k4 ↓ ǎk2 ↓ ǎk1 ↑ . (130)

4.7 Many body problem
This can be understood so that two fermions whose spins We study an interacting system. We write the Hamilto-
are the same cannot coexist in the same place, so they nian presented above (124) using a shorter notation
cannot feel a delta-function interaction.
1X
ξij ǎ†i ǎj + Vijkl ǎ†i ǎ†j ǎl ǎk . (137)
X
This concludes the introduction to second quantization. Ǩ ≡ Ȟ − µŇ =
i,j
2
Here we still give a short account of the terminology. When ijkl
we take as a starting point the many-body wave equation where ξij = hi|T |ji − µδij and Vijkl = hi, j|V |k, li.
(103), (115), and (116), then the second quantization is
just new way of denoting it. An alternative approach, used The essential problem is the interaction term in (137),
in quantum field theory and from which the name second which is fourth order in ǎ. This is the many body problem.
quantization comes from, is that one makes the switch from The noninteracting case (=ideal gas) can be calculated
single-particle quantum mechanics to many-body theory exactly, but there is no general method to solve exactly the
by postulating the commutators (106) [or for fermions the problem of many interacting particles. Instead, there exists
anticommutators (122)] and only after that deduce (if need numerous approximation methods that can be applied in
be) the formulas (103), (115), and (116). different cases.
It turns out that in case of superconductivity, there
4.6 Noninteracting system exist a quite good approximation method (known as qua-
siclassical theory of Fermi liquids). There the many body
We study fermions that do not interact with each other. problem is solved using the fact that the ratio Tc /TF is
It is easiest to use grand canonical ensemble. The effective small. In general form this theory is very complicated, and
Hamiltonian operator (27) is therefore it is not presented in this course. Although we
have to refrain from firm justification, we attempt to make
ξk ǎ†kσ ǎkσ ,
X
Ǩ ≡ Ȟ − µŇ = (131) the main results understandable in the following.
k,σ
A quite general approximation method in many-body
where ξk = k −µ. Its ground state (45) and the elementary systems is Hartree-Fock approximation. The main idea is
excitations are described above. We see that in the notation the following. Because the difficulties in the Hamiltonian
of second quantization the ground state |0i can be written (137) arise in the fourth order term, one approximated it
in the form Y † † with a second order term
|0i = ǎk↑ ǎk↓ |vaci, (132) 1X
Vijkl ǎ†i ǎ†j ǎl ǎk ≈ Aij ǎ†i ǎj
X
|k|<kF V̌ = (138)
2 ij
Q ijkl
where |k|<kF is a product over all wave vectors having
k < kF and |vaci is the vacuum state |0, 0, 0, . . .i. Because this term has the same form as the noninterac-
ting system, its solution is easy (at least relatively). Now
In order to get clear connection to the excitation picture,
one must determine the coefficients Aij . In Hartree-Fock
we define new operators
approximation this is done by replacing the removed ope-
rators by their expectation values. Thus
†
ǎkσ = γ̌−kσ when ξk < 0 . (133)
γ̌kσ when ξk > 0. 1X
V̌ ≈ V̌HF = Vijkl (hǎ†i ǎk iǎ†j ǎl + ǎ†i ǎk hǎ†j ǎl i
† 2
The operators γ̌kσ ja γ̌kσ satisfy the same commutation ijkl
†
relations (122) as ǎkσ ja ǎkσ . Therefore the transformation −hǎi ǎl iǎj ǎk − ǎ†i ǎl hǎ†j ǎk i) + constant.
† †
(139)
16
The first term can be understood so that a particle that • we assume that momentum is conserved in expectation
scatters from state l to state j feels an interaction that values (although the particle number is not). One can
is averaged over all states of other particles. The second show that this limits the study to the case of stationary
term is of the same type but has different indices. The (not flowing) state of the superconductor.
third and fourth terms are caused by the fact that it is
impossible to distinguish the particles, but one must allow With these assumptions we get the Hamiltonian
them to interchange (so called exchange interaction). [We g XX
ξk ǎ†kσ ǎkσ − 3
X
have neglected in equation (139) correction terms that arise Ǩeff =
L
from the fact that a particle cannot interact with itself.] k,σ 0 k k
The expectation value is calculated as above h. . .i = ×(ǎ†k↑ ǎ†−k↓ hǎ−k0 ↓ ǎk0 ↑ i + hǎ†k↑ ǎ†−k↓ iǎ−k0 ↓ ǎk0 ↑
Tr(. . . ρ̌) (9), but in order to get a closed theory, one must − hǎ†k↑ ǎ†−k↓ ihǎ−k0 ↓ ǎk0 ↑ i). (143)
use the same approximation for Ǩ in the density matrix:
Tr(. . . e−β ǨHF ) We write this into the form

h. . .i = . (140)
ξk ǎ†kσ ǎkσ − (∆ǎ†k↑ ǎ†−k↓
X X
Tre−β ǨHF Ǩeff =
k,σ k
This is known as the self-consistency equation. It must be
solved together with (139) because they depend on each + ∆∗ ǎ−k↓ ǎk↑ ) + C. (144)
other.
Here we have defined
The Hartree-Fock approximation is widely used in calcu- g X
lation of the electronic states of atoms and molecules. For ∆= 3 hǎ−k↓ ǎk↑ i. (145)
L
conduction electrons is clearly insufficient already in the k
normal state (because it does not take into account the sc- For its complex conjugate we get
reening of the Coulomb potential). Also, it does not lead
to superconductivity. g X † †
∆∗ = 3 hǎk↑ ǎ−k↓ i. (146)
L
k
4.8 Superconducting state The Hamiltonian (144) can also be written in the form
We make so called anomalous Hartree-Fock approxima- ξ ǎk↑

tion, where we also take terms of the type ǎ†i ǎ†j hǎl ǎk i. This
X † k −∆
Ǩeff = ǎk↑ ǎ−k↓ †
is not included in the standard Hartree-Fock approxima- −∆∗ −ξk ǎ−k↓
k
tion because the expectation values of the operators, which +C2 . (147)
change the particle number, vanish. It turns out, however,
that just these terms are essential for superconductivity.
The Hamiltonian (144) is second order in ǎ, as was de-
The anomalous HF approximation can also be justified sired. However, it is not yet of the same form as for ideal
by first writing exactly gas (131) because in addition to diagonal terms ∝ ǎ†kσ ǎkσ
it contains nondiagonal terms. In order to achieve a diago-
ǎi ǎj = hǎi ǎj i + (ǎi ǎj − hǎi ǎj i) (141) nal form we make Bogoliubov transformation. We introduce
the operators γ̌k↑ ja γ̌k↓ by defining
and correspondingly for the Hermitian conjugate operator.
†
Now we assume that the expectation value is a good ap- ǎk↑ = u∗k γ̌k↑ + vk γ̌−k↓
proximation, i.e. the term in the parenthesis on the right †
hand side is small. Its order of magnitude is denoted by ǎk↓ = u∗k γ̌k↓ − vk γ̌−k↑ . (148)
. We substitute in the Hamiltonian (137) and drop terms Here u and v are generally complex valued coefficients.
k k
that are proportional to 2 , and get For the Hermitian conjugates we get
1X
V̌anom = Vijkl (hǎ†i ǎ†j iǎl ǎk +ǎ†i ǎ†j hǎl ǎk i−hǎ†i ǎ†j ihǎl ǎk i). ǎ†k↑ = uk γ̌k↑
†
+ vk∗ γ̌−k↓
2
ijkl
ǎ†k↓ = uk γ̌k↓
†
− vk∗ γ̌−k↑ . (149)
(142)
This was the most essential approximation. In addition, We wish to make the transformation canonical, which
we make the following simplifying approximations. means that the new operators should satisfy the same an-
ticommutation relations as the original operators (122). We
• the interaction is approximated by a contact interac- calculate
tion (128).
{ǎk↑ , ǎ†k↑ } =
• the normal Hartree-Fock terms (139) are dropped as- † †
|uk |2 {γ̌k↑ , γ̌k↑ } + |vk |2 {γ̌−k↓ , γ̌−k↓ }
suming that they shift equally the energies of the nor-
† †
mal and superconducting states. + u∗k vk∗ {γ̌k↑ , γ̌−k↓ } + uk vk {γ̌−k↓ , γ̌k↑ } = 1.
17
This succeeds if Gap equation
|uk |2 + |vk |2 = 1. (150)
In order to fix the theory, we should determine ∆ (145).
Similarly as in the Hartree-Fock approximation we have to
Exercise: Show that the inverse of the transformation
calculate the expectation value using the effective Hamil-
(148) is
tonian (156). We get
γ̌k↑ = uk ǎk↑ − vk ǎ†−k↓ g X g X Tr(ǎ−k↓ ǎk↑ e−β Ǩeff )
hǎ−k↓ ǎ∆=
k↑ i = .
γ̌k↓ = uk ǎk↓ + vk ǎ†−k↑ . (151) L3 L3 Tre−β Ǩeff
k k
(157)
Show that the γ̌ operators satisfy all the same anticommu- Using the transformation formulas (148) and expressing
tation rules (122) as the ǎ operators. the trace (Tr) in eigenstates of the excitation number ope-
Now we substitute the Bogoliubov transformation to the rators we get
Hamiltonian (144). We see that the inconvenient terms g X
drop out if one chooses ∆= 3 uk vk [1 − 2n(Ek )], (158)
L
k
2ξk uk vk − ∆uk 2 + ∆∗ vk2 = 0, (152) where the Fermi distribution
and the Hamiltonian gets the form 1
n() = . (159)
eβ + 1
† †
X
Ǩeff = {ξk [|uk |2 (γ̌k↑ γ̌k↑ + γ̌k↓ γ̌k↓ )
k Using still the expressions for uk and vk (154) and rear-
+ †
|vk |2 (γ̌k↑ γ̌k↑ †
+ γ̌k↓ γ̌k↓ )] + (∆∗ u∗k vk + ∆uk vk∗ ) ranging one gets the consistency equation in the form
† † g X ∆ Ek
× (γ̌k↑ γ̌k↑ − γ̌k↓ γ̌k↓ )} + C. (153) ∆= tanh . (160)
2L3 Ek 2kB T
k
Now we suppose, for simplicity, that ∆ is real valued. This has at least the trivial solution ∆ = 0, which cor-
(We shall return to the general case later.) The solution of responds to the normal state. Any other solutions should
equations (150) and (152) is satisfy
g X 1 Ek
1= tanh . (161)
s s
1 ξk 1 ξk 2L3 Ek 2kB T
uk = 1+ , vk = 1− , (154) k
2 Ek 2 Ek
We see that because of the factor 1/Ek the summand is
where q largest at the Fermi surface. Indeed the contact interaction
Ek = ξk2 + ∆2 . (155) (128), whose Fourier transform is a constant, is too idea-
lized, and at large momenta the Fourier transform should
Substituting these into (153) gives the Hamiltonian in the drop to zero. Similarly as in the Cooper problem, we cut off
desired form the summation at energy c . Using the summation formula
(76) we get
†
X
Ǩeff = Ek γ̌kσ γ̌kσ + Ω0 . (156)
gN (0) c
p
ξ 2 + ∆2
Z
k,σ 1
1= dξ p tanh . (162)
2 −c
2
ξ +∆ 2 2kB T
(We return to the constant Ω0 later.) Because Ek > 0, it
can be interpreted, similarly as in the normal state (135- Noticing the symmetry of the integral we get it to the form
†
136), as the excitation energy and γ̌kσ γ̌kσ as the number Z c p
1 1 ξ 2 + ∆2
operator of excitations. The excitation energy (155) in the = dξ p tanh . (163)
neighborhood of the Fermi surface is depicted in the figure. gN (0) 0 ξ 2 + ∆2 2kB T
In the superconducting state the particle and hole type This gap equation determines the energy gap as a func-
excitations change to each other smoothly as k changes. tion of temperature, ∆(T ). Generally it should be solved
The excitations have a minimum energy ∆, which is known numerically. We study two limiting cases.
as energy gap. The excitation spectrum reduces to the one
in normal state when ∆ = 0. 1) T = 0. One gets
Z c
1 1
Ek gN (0)
= dξ p
0 ξ + ∆2
2
2c
= ln , (164)
∆
∆
k where the integral can be calculated with Mathematica,
0 and the result is valid in the limit c ∆. We get
kF 1
∆(T = 0) = 2c e− gN (0) . (165)
18
2) T = Tc . Here the superconducting state vanishes, Let us study how well the previous assumptions are sa-
∆ → 0. When c kB Tc we get tisfied. The table shows experimental values except that
Z c gN (0) is calculated from equation (167). We see that for
1 1 ξ
= dξ tanh elemental metals vc kB Tc holds reasonably. Equation
gN (0) 0 ξ 2kB Tc (168) is rather well satisfied. The greatest deviation ap-
2c eγ
= ln , (166) pears in Lead. The deviations can be more or less unders-
πkB Tc tood using strong coupling theory, which takes more accu-
γ
where the Euler constant γ = 0.5772, e = 1.781. We get rately into account e.g. the energy dependence of the pho-
non mediated interaction.
2eγ − gN1(0) ∆(0)
kB Tc = c e . (167) Tc (K) TD (K) gN (0)
π kB Tc
BCS 1.764
The result (167) in principle predicts the transition tem- Cd 0.56 164 0.18 1.6
perature of the superconductor. It contains the parameters Al 1.2 375 0.18 1.3-2.1
N (0), c and g. Out of these N (0) (77) can be measured Sn 3.75 195 0.25 1.6
independently and c is estimated to be on same order of Pb 7.22 96 0.39 2.2
magnitude as the Debye temperature (∼ 100 K). The grea- The quasiclassical theory of Fermi liquids is based on the
test uncertainty appears in the constant g. Because Tc de- following idea. There are a few microscopic parameters like
pends exponentially on g, it makes the calculation of Tc Tc , effective mass etc., whose values cannot be calculated
very uncertain. Therefore the equation (167) as such is not in this theory, but their values can be determined experi-
as remarkable as one could initially think. mentally. Once the parameters are known, the quasiclas-
Considerably more reliable result is obtained if we elimi- sical theory can be used to calculate several properties of
nate c e−1/gN (0) from equations (165) and (167): the superconducting state, e.g. the gap function.
∆(T = 0) = πe−γ kB Tc = 1.764 kB Tc . (168)

BCS ground state
We see that the energy gap and the transition temperature In their original work Bardeen, Cooper and Schrieffer
are of the same order of magnitude, and also get a precise presented the ground state of the superconductor
factor of proportionality between the two.
(uk + vk ǎ†k↑ ǎ†−k↓ )|vaci,
Y
More generally, g and c can be eliminated from the gap |ψ0 i = (171)
equation in the weak coupling limit c kB Tc . This can be k
accomplished by subtracting from the gap equation (163) where Q is the product over all wave vectors and |vaci
k
its Tc condition (166). Utilizing the limit c kB Tc the is the vacuum state. Show as an exercise that this is con-
resulting equation can be written in the form (exercise) sistent with the theory above (156) by showing that γkσ
Z ∞" (151) annihilates this state,
p #
tanh(ξ/2kB T ) tanh( ξ 2 + ∆2 /2kB T )
− p dξ
0 ξ ξ 2 + ∆2 γ̌kσ |ψ0 i = 0. (172)
Tc
= ln . (169) The functions uk and vk (154) near the Fermi surface are
T
shown in the figure.
We see that ∆(T )/kB Tc is a universal (= independent of 1.0
the material) function of T /Tc . The only material depen- vk uk
dent parameter is thus Tc . 0.8
2.0
0.6
∆
kBTc
0.4
1.5
0.2
1.0 ξk
0.0
-5 -4 -3 -2 -1 0 1 2 3 4 5 ∆
0.5 The state (171) shows that the occupations of the single
particle states k ↑ and −k ↓ are correlated. If a particle is
0.0 T present in state k ↑, it is also present in state −k ↓, and if
0.0 0.2 0.4 0.6 0.8 1.0 Tc not, also not in the other state. For ideal gas (132) this is
satisfied trivially because the occupation number is always
In the neighborhood of Tc one can derive (ζ(3) = 1.202) either 0 or 1, depending on k. The BCS state differs from
s r the normal state at those k for which uk vk 6= 0. This takes
8 T place around the Fermi surface in the energy range of a few
∆(T ) ≈ kB Tc π 1− . (170)
7ζ(3) Tc ∆ ∼, or, equivalently, of a few kB Tc .
19
One can also show that the state (171) is the one consis- We attempt to understand that the number of particles
ting of pairs, as was expected above using the wave function in the BCS state is not fixed.
presentation (83) (see de Gennes). Exercise: suppose that ∆ defined in equation (145) is not
real valued but
The size of a pair ∆ = eiφ |∆|, (176)
In order to see the general picture, it is important to where φ is a real phase. Show that all the preceding results
study the size of one pair. This can be justified in more de- are valid also in this case when equations (154) and (155)
tail, for example by studying the wave function obtained are replaced by
from the Cooper problem (see Ketterson-Song), but the fol-
lowing gives the same result more directly. The supercon- 2 1 ξk 2 1 2iφ ξk
uk = 1+ , vk = e 1− , (177)
ducting state differs from the normal one in the energy 2 Ek 2 Ek
range δξ ∼ kB Tc . By calculating q
2 Ek = ξk2 + |∆|2 . (178)
p p
δξ = δ = δ = δp ≈ vF δp (173) and in equations (162)-(170) one replaces ∆ → |∆|.
2m m
We conclude that the BCS state (171) is degenerate, i.e.
we get that in momentum this corresponds to a shell of corresponding to each value of φ there is a state with the
thickness δp = h̄δk ∼ kB Tc /vF around the Fermi surface. same energy, where the original v (154) is multiplied by a
k
It is a general property of wave motion that the minimum phase factor exp(iφ):
size δx for a localized wave packet satisfies δx δk ∼ 1 when
(uk + vk eiφ ǎ†k↑ ǎ†−k↓ )|vaci.
Y
the available wave vectors are in range δk. Applying this |ψφ i = (179)
to the superconducting state we get that the minimum size k
of a Cooper pair has the magnitude
Let us consider more generally the relationship between
h̄vF
ξ0 = , (174) phase and pair number. We define the pair number opera-
2πkB Tc tor ň. Its eigenstates are |ni,
which is called coherence length. (Because it is an order of ň|ni = n|ni. (180)
magnitude estimate, the factor 2π can be added without
justification.) The table shows calculated values for a few [For mathematical completeness we consider the integer n
metals. in the range (−∞, +∞) but in practice a much smaller
ξ0 (nm) range is sufficient.] In addition we define the operator
Al 1500 X
Sn 480 Ť = |n − 1ihn|. (181)
n
Pb 160
Nb 14 Operated on an arbitrary pair state it removes one pair out
The essential result is that ξ0 is much larger than the of it. We look for the eigenstates of Ť :
atomic scale ∼ 0.1 nm. The pairs are so huge that within Ť |ti = t|ti. (182)
each pair there are ∼ 1010 conduction electrons! The situa-
tion is quite opposite, for example, to O atoms forming O2 In order to achieve this we write
molecules in oxygen gas. X
|ti = fn |ni. (183)
Because the pairs are strongly overlapping, the identi- n
ty of electrons and the Pauli exclusion principle are quite
essential in the theory of superconductivity. For example, Now the condition (182) gives fn+1 = tfn , whose solution
one cannot answer the question which two electrons form is fn = C exp(inφ) and the eigenvalue t = exp(iφ). Thus
the pair. Therefore the BCS ground state (171), which on- corresponding to each eigenstate of the operator Ť there is
ly indicates pair correlations, is much more useful way of a corresponding value of φ. We use this value to label the
presenting than the wave function presentation (83) where eigenstates and therefore write instead of (182)
the effect of antisymmetrization is essential. Ť |φi = eiφ |φi. (184)
The size of a pair compared to the atomic scale a ∼
kF−1 is equivalent with the previously mentioned fact that This eigenstate of the phase can be presented using pair
kB Tc F : number eigenstates
ξ0 F 1 X inφ
∼ 1. (175) |φi = √ e |ni. (185)
a kB Tc 2π n
Superconductivity is a relatively low energy phenomenon
and therefore the associated length scale is large. A calculation gives the inverse relation
Z π
1
Particle number and phase |ni = √ dφ e−inφ |φi. (186)
2π −π
20
We see that the BCS state (179) has the form we get
2X
ln 1 + e−βEk .

X Ω = Ω0 − (192)
|ψφ i = einφ fn |ni, (187) β
k
n
We see that at zero temperature Ω(T = 0) = Ω0 .
where fn is real and differs from zero in an interval of ∆n
1. (∆n denotes here the difference in n.) Here the phase is The energy functional (192) has a couple of interesting
rather well defined but the pair number is uncertain. If oneproperties.
wants a state with fixed number of pairs, it can be formed 1) If one minimizes Ω0 with respect to vk [taking into
as linear combination of phase eigenstates, as in equation account (150), ∆ = constant] one gets the same condi-
(186). Then the phase is fully uncertain. More generally tion (152) that was derived above in another way.
one can construct an uncertainty relation 2) In the reminder of statistical physics we stated that Ω
has a minimum with respect to all internal degrees of free-
1 dom. In particular, this should apply to ∆. Verify as an
∆n ∆φ ≥ . (188)
2 exercise that the condition
∂Ω(T, V, µ, ∆)
Justification: We consider an arbitrary state |ψi. Its =0 (193)
∂∆
representation in the pair number eigenbasis is ψn = hn|ψi
and in phase eigenbasis ψ(φ) = hφ|ψi. From equation (186) is equivalent with the gap equation (158).
we get Z π Based on the energy functional (192) one can calculate
1
ψn = √ dφ einφ ψ(φ). (189) all thermodynamic quantities. Here we calculate the diffe-
2π −π rence in energy between the superconducting and the nor-
This shows that these representations are constructed from mal state at zero temperature:
each other by Fourier transformation. X L3 2
Ω0 − Ω0 (∆ = 0) = 2 (ξk vk2 − ∆uk vk ) + ∆
The relation (189) is similar as between the momentum g
k
and coordinate representations X 1 ξk
Z L/2 −2 ξk (1 − ). (194)
1 −ipx/h̄
2 |ξ k|
ψp = √ dx e ψ(x) (190) k
L −L/2 In calculating the integrals one can proceed as follows.
for a particle in a one-dimensional box of width L. The Because of the gap equation
P (158) the term L3 ∆2 /g cancels
uncertainty relation (188) can be derived from relation half of the term −2 k ∆uk vk . Using expressions (154) and
(189) in the same way as the Heisenberg uncertainty re- (155) we get the summation in the form
lation ∆p∆x ≥ h̄/2 for the particle’s momentum and po- 1X ∆4
sition can be derived from equation (190). Here we do it Ω0 − Ω0 (∆ = 0) = − . (195)
2 Ek (Ek + |ξk |)2
only up to an order of magnitude. Let ψ(φ) be different k
from zero when |φ| < ∆φ/2. In order to ψn to change from We see that this converges well for large ξk . Therefore we
its maximum value, n has to change so that the exponent can use formula (76). We get
in equation (189) changes essentially, and thus ∆n∆φ ∼ 1.
1
Ω0 − Ω0 (∆ = 0) = − L3 N (0)∆2 . (196)
2
As a conclusion we realize that the anomalous Hartree-
This is the condensation energy of the superconducting sta-
Fock approximation leads to BCS states where the phase is
te, whose existence we deduced above by studying thermo-
well defined but the pair number is uncertain. As a linear
dynamics in magnetic field (61). We see that the energy of
combination of such states it is possible to form states of
the superconducting state is lower than that of the normal
fixed number of pairs.
state. The energy difference (196) can be roughly unders-
tood that the energy of those single particle states, which
4.9 Thermodynamics are around the Fermi surface in a shell of energy ∆, is re-
We follow the principle in the beginning of the course duced by ∆.
that we first calculate the thermodynamic potential, and The entropy can be calculated from the formula (33)
from it we get all thermodynamic quantities as derivatives. and the specific hear from formula (18). In calculating the
To simplify the notation, we assume ∆ real. We first entropy one should in principle derivate with respect to all
calculate the constant appearing in the diagonalized Ǩ temperature dependent parameters, but because of relation
(156) (exercise), and get (193) the temperature dependence of ∆ does not contribute
to the final result. Show as exercise that the specific heat
3
X L is given by
Ω0 = 2 (ξk vk2 − ∆uk vk ) + ∆2 . (191)
g L3 N (0) ∞
Z
1

d∆

k 2 2
C= dξ √ ξ + ∆ − T ∆ .
2kB T 2 −∞ 2 ξ 2 +∆2 dT
Because Ǩ (156) is diagonal, the grand potential (30) can cosh 2kB T
be calculated in the same way as for ideal Fermi gas and (197)
21
Using numerical calculation one could show that this gives state can to a large extent be understood similarly as su-
similar curve a plotted on page 3. perconductivity above.
Exercise. Show from formula (197) that the specific heat An essential difference is that in 3 He the pairs form in
of the normal state is given by a p wave state instead of the s wave state found in most
superconductors. This means that vk has to be replaced
2π 2 3
C= L N (0)kB T, 2
(198) by vk , which also depends on the direction of k according
3 to a combination of spherical harmonic functions Y1m (θ, φ)
which is linear in T . From this one can determine N (0) [thus ` = 1 ja m = 0, ±1]. Simultaneously the spin state
experimentally. has to be written more generally. The BCS state (171) thus
has the form
Note. Because we used grand canonical ensemble (28)  
the specific hear (197) is calculated at constant µ. In the
vkσσ0 ǎ†kσ ǎ†−kσ0  |vaci.
Y X
courses of condensed matter physics and statistical physics |ψ0 i = uk + (202)
it is shown that in the case studied (T TF ) the specific k σ,σ 0
heat at constant volume is the same.

Because v−kσσ0 = −vkσσ0 we see that only the spin sym-
metric part vkσσ0 + vkσ0 σ gives something nonzero (exerci- √
4.10 Inhomogeneous superconductor se). Thus the spin state is triplet [↑↑, ↓↓ or (↑↓ + ↓↑)/ 2]
Previously in equation (144) we assumed that only those instead of the singlet in superconductors (95). It follows
3
expectation values conserving the momentum were nonze- that the order parameter of He is a 3 × 3 matrix, where
ro. We start to generalize the calculation to the case that the indices refer to three p wave states and to three spin
this assumption is not made. The effective Hamiltonian can triplet states. j
be written in the form Also some superconductors show properties from which
1 one can infer an order parameter consisting or more than
ξk ǎ†kσ ǎkσ − 3
X X X
Ǩeff = one component (e.g. UPt3 ). In high temperature supercon-
L
k,σ k0 q
ductors the pairs are found to form in the d wave state
×(ǎ†1 q+k0 ↑ ǎ†1 q−k0 ↓ ∆(q) + ∆∗ (q)ǎ 21 q−k0 ↓ ǎ 21 q+k0 ↑ ) vk ∝ kx2 − ky2 . The superfluid states of 3 He are studied a
2 2
lot because there one can avoid such complications occur-
+ C, (199)
ring in metals such as the ion lattice (which causes non
where spherically symmetric Fermi surface) and impurities.
X
∆(q) = g hǎ 12 q−k↓ ǎ 21 q+k↑ i. (200)
k
We define the inverse Fourier transform of ∆(q) in the

usual way
1 X iq·r
∆(r) = e ∆(q)
L3 q
g X X iq·r
= 3 e hǎ 21 q−k↓ ǎ 12 q+k↑ i. (201)
L qk
We realize that h̄q the total momentum of a pair. Thus we

can deduce that ∆(r) can be interpreted as the wave func-
tion describing the center of mass of the pair. The function
∆(r) is called order parameter. If the momentum in the
expectation value (201) is conserved, ∆(r) reduces to the
same constant as above (145).
We realize that this generalization is needed in the case
where the superfluid is in motion or in other inhomoge-
neous state. We will not continue the analysis in general
form any further, but we will later use another approach
to study this very important case.
4.11 Superfluid 3 He
As an exercise we found out that 3 He atom is a fermion
and its Fermi temperature TF ∼ 1 K. It is found expe-
rimentally that liquid 3 He has a transition to superfluid
state at temperatures Tc = 1 . . . 2.5 mK. This superfluid
22
5. Ginzburg-Landau theory Often one wants to study a system in a given external
magnetic field. Then, instead of F , on should minimize G
(55), in this case
5.1 Introduction Z
This part is well presented in FW pages 430-439 G = F − d3 r H · B. (208)
Similarly as Ginzburg and Landau (GL), we derive the
GL theory phenomenologically. We will discuss later, how Let us still write the G in GL theory in its full form
it can be derived from microscopic theory. Z
G = F0 + d3 r g,
Ginzburg and Landau presumed that the superconduc-
2
ting state is described by a complex-valued order parame-

2 1 4
h̄
ter Ψ. This parameter is assumed to be different from zero g = α|Ψ| + β|Ψ| + γ ∇ − qA Ψ
2 i
only in the superconducting state. Further it is assumed 1 2
that Ψ is small near the transition temperature. We sup- + B − B · H. (209)
2µ0
pose that near the transition temperature the free energy
density f = F/L3 can be written as Taylor series in Ψ and
Ψ∗ , GL differential equations
1
f = f0 + α|Ψ|2 + β|Ψ|4 + . . . . (203) Earlier we derived the result that in equilibrium the free
2
energy must be minimized. In a given external field one
The terms appearing here are restricted by the fact that
must thus minimize G (209) both with respect to Ψ and
Fs has to be real valued for arbitrary complex valued Ψ.
to A. {In minimizing with respect to Ψ the independent
Therefore, the term cΨ cannot appear. Instead, Ψ has to
variables [e.g. (Re Ψ, Im Ψ) or (|Ψ|, arg Ψ)] can be chosen
appear in product with Ψ∗ : Ψ∗ Ψ = |Ψ|2 . Also the term
arbitrarily. The shortest calculation follows by treating Ψ
c Re Ψ is not accepted. The reason is that we require Fs to
and Ψ∗ as independent variables.} Let us leave the mi-
remain unchanged in the transformation Ψ → eiφ Ψ, where
nimization as an exercise. As a result we obtain the GL
φ is a real-valued constant.
differential equations
The expansion (203) is incomplete because nothing in it 2
prevents a spatial dependence Ψ(r). Such a spatial depen- h̄
γ ∇ − qA Ψ + αΨ + β|Ψ|2 Ψ = 0, (210)
dence can be limited by adding a term |∇Ψ|2 that increases i
the energy of inhomogeneous states. However, also this is 1 qh̄γ
unsatisfactory in the case of a nonzero magnetic field. The ∇×B= (Ψ∗ ∇Ψ − Ψ∇Ψ∗ )
µ0 i
magnetic field B can be described with a vector potential
−2q 2 γ|Ψ|2 A. (211)
A:
B = ∇ × A. (204) The surface terms arising from integration by parts must
also vanish. From this we get the boundary conditions at
From the course of analytical mechanics we know that the the surface of a superconductor
real momentum mv = p − qA. Here p is a canonical mo-
mentum, which in quantum mechanics is replaced by the h̄
n̂ · ∇ − qA Ψ = 0, (212)
operator h̄i ∇. Analogously to this, GL chose the additional i
energy term to be n̂ × (B − µ0 H) = 0. (213)
2
h̄ [It is noted in passing that the transformation (208) is es-
γ ∇ − qA Ψ , (205) sential only for the surface terms.]
i
It is noted that based on the Maxwell equation
where q is some charge. Let us take into account also the
energy density of the magnetic field in the sample ∂E
∇ × B = 0 µ0 + µ0 j (214)
∂t
1 2
B . (206) we identify the quantity appearing in (211) as an electric
2µ0
current density
In this way we obtain the total energy in Ginzburg-Landau
qh̄γ
theory as j= (Ψ∗ ∇Ψ − Ψ∇Ψ∗ ) − 2q 2 γ|Ψ|2 A. (215)
i
Z
F = F0 + d3 r f, Exercise: show that the equations guarantee current con-
2 servation
1 h̄ ∇ · j = 0, n̂ · j = 0. (216)
f = α|Ψ|2 + β|Ψ|4 + γ

∇ − qA Ψ
2 i
1 2
+ B . (207) 5.2 Special cases
2µ0
23
The GL equations (210) and (211) constitute a coupled where we have defined the GL coherence length
set of differential equations, whose solution gives Ψ(r) and s
A(r). In the general case this is very complicated. Let us h̄2 γ
start by considering simple special cases. ξ GL = . (224)
|α|
1) Homogeneous superconductor, H = A = 0. Equation
(210) gives as possible solutions We see that ξGL determines that length scale on which Ψ
can vary essentially. As an example we give the solution
Ψ = 0, (217) of equation (223) in the case of a one-dimensional depen-
α
|Ψ|2 = − . (218) dence: x
β f (x) = tanh √ . (225)
2ξGL
The former solution describes normal state. The latter, su-
perconducting state, is possible only if α/β < 0. In order 3) Let us investigate the case
for F (207) to be sensible (minimum energy must be ac-
hieved with a finite Ψ) we must always have β > 0. The Ψ(r) = eiφ(r) |Ψ| (226)
condition for the latter state is therefore α < 0. The ener-
gies corresponding to the states (217) and (218) are found where |Ψ|2 ≈ |α|/β is constant. By inserting into the
by inserting into the functional (207): expression of current (215) we find
F = F0 , (219)
j = 2qγ|Ψ|2 (h̄∇φ − qA) . (227)
α2
F = F0 − V . (220)
2β By taking the rotor of this we have the London equation
We thus see that if α > 0, only the normal state is pos-
∇ × j = −2q 2 γ|Ψ|2 B. (228)
sible, while in the case α < 0 the superconducting state
has the lowest energy. The transition temperature Tc thus By using the Maxwell equations (214) and (51) this yields
corresponds to the point where α = 0. In GL theory it is as-
sumed that the temperature dependence of the coefficient ∇×j ∇ × (∇ × B)
α is linear B=− 2 =−
2q γ|Ψ|2 2µ0 q 2 γ|Ψ|2
0 T
α(T ) = α −1 , (221) ∇ B2
Tc = . (229)
2µ0 q 2 γ|Ψ|2
and the other coefficients (β, γ, q) are temperature-
independent. Thus we have
The dependence of the free energy on the order parame- B = λ2 ∇2 B, (230)
ter can be illustrated with the following pictures. where s
T<Tc F T>Tc F β
λ= . (231)
2µ0 q 2 γ|α|
Bz
B0
Re Ψ Im Ψ Re Ψ Im Ψ
It is observed that in the normal state Ψ = 0 is comple-

0 λ x
tely determined, but in the superconducting state only the
absolute value |Ψ| of the order parameter is fixed while the Now we study a superconducting half-space x > 0. Let
phase arg Ψ is arbitrary. us assume that outside the superconductor (x < 0) there
2) Changing |Ψ|, A = 0. From the boundary condi- is a field B0 = B0 ẑ parallel to the surface of the supercon-
tion (212) it follows that n̂ · ∇Ψ = 0. Thus a position- ductor. The solution of equation (230) inside the supercon-
independent |Ψ| (218) is a valid solution everywhere in the ductor is
superconductor, also close to boundaries. Despite this we Bz (x) = B0 e−x/λ . (232)
consider a case where Ψ deviates from its equilibrium va-
Thus we explain the Meissner effect: the magnetic field does
lue. From equation (210) we find
not penetrate into the superconductor, apart from a layer
h̄2 γ∇2 Ψ − αΨ − β|Ψ|2 Ψ = 0. (222) with thickness on the order of λ. In this layer a current is
p flowing (calculate it), which cancels the external field in-
Assuming Ψ to be real and writing Ψ = |α|/βf we put
side the superconductor. The result thus also implies the
this in the form
existence of a dissipationless current. Below some experi-
2
ξGL ∇2 f + f − f 3 = 0, (223) mental values for the penetration dept are given.
24
λ(T Tc ) (nm) By minimizing G with respect to B we find
Al 49
Sn 51 1
G = F0 − V µ0 H 2 , B = µ0 H. (237)
Pb 39 2
>>λ In order for the energies to be equal when H = Hc we have
1 α2
Γ µ0 Hc2 = . (238)
2 2β
Φ
From this and the linearity (221) of α(T ) it is concluded
that Hc (T ) is linear close to Tc . This is consistent with the
experimental observation (3).
7) Next we investigate the interface between normal and
4) We look at the superconducting state in a ring (loop, superconducting states. This requires the field H to be of
torus), whose cross-section is considerably larger than the the critical magnitude Hc , because the interface can only be
penetration depth. Then inside the ring j = 0. From (227) stable if the two phases are in equilibrium. The structure of
we find that for a path Γ going around inside the ring we the interface can be solved exactly from the GL equations,
have be here we are satisfied by a qualitative analysis. In the
I Z previous point we identified the essential terms of normal
0 = dl · (h̄∇φ − qA) = h̄2πN − q da · ∇ × A. (233) and superconducting states in the functional (209). Let us
see how these are involved in the interface.
Here N is an integer, which follows from that the fact that λ λ
a unique single-valued Ψ (226) only allows for φ to change
by a multiple of 2π when going around Γ. Thus for the B B
Ψ Ψ
magnetic flux threading the loop we find
Z
2πh̄
Φ = da · B = N . (234)
q x x
ξGL ξGL
It has been experimentally observed that the flux is quan-
tized according to this formula. From the magnitude from In the figure a light shading roughly describes the regions
h
the observed flux quantum, Φ0 = 2|e| (4), we deduce that where the energy is lowered from F0 by the amount (238).
q is twice the charge e of an electron (the sign of the charge The situation in an S-N interface depends essentially on
cannot be deduced from this). the ratio of the penetration depth and the GL coherence
5) Above we have defined two length: the GL coherence length. If λ ξGL a region of thickness ≈ ξGL is for-
length ξGL (224) an the penetration length λ (231). Both med, where neither of the negative contributions 1
is reac-
2
have the temperature dependence hed. This means an interface energy σ ≈ 2 ξ GL µ0 H c [com-
pare to equation (68)].In the opposite case λ ξGL , both
1 1 negative contributions are present within a thickness ≈ λ
λ(T ), ξGL (T ) ∝ p ∝p , (235)
|α| 1 − T /Tc and we find a negative interface energy σ ≈ − 12 λµ0 Hc2 .
This latter case leads to completely new types of proper-
so that they diverge when T → Tc . The ratio of the lengths ties. The description of the intermediate state given in the
is called the GL parameter beginning if the course is clearly not valid in this case.
s A superconductor where the interface energy in negative
λ(T ) β is called a type II superconductor, as opposed to the type
κ= = . (236)
ξGL (T ) 2µ0 q 2 h̄2 γ 2 I superconductor that has a positive interface energy. By
solving the GL equations we find that√ the limit between the
It is a temperature-independent constant. By writing the two cases goes at the value κ = 1/ 2 of the GL parameter.
√
GL equations in a dimensionless form we observe that this Thus for a type I superconductor κ < 1/ 2 and for a type
√
is the only dimensionless parameter in the theory. II superconductor κ > 1/ 2.
6) The equilibrium between normal and superconducting
states in an external field was studied already in the be-
ginning of the course, but it is instructive to see the same 5.3 Derivation from microscopic theory
by starting from the GL functional (209). In the supercon- The Ginzburg-Landau theory can be derived starting
ducting state we obtain from the terms α|Ψ|2 + 12 β|Ψ|4 from microscopic theory. The correspondence to microsco-
a negative contribution that was calculated above (220). pic theory is achieved when one identifies
This is independent of the field H, because B ≡ 0. In the
normal state only the terms 2µ1 0 B 2 − B · H are nonzero. Ψ(r) = ∆(r), (239)
25
where ∆ is defined by equation (201). [Note that relation Metals always have impurities. It was discussed above
(239) can contain an arbitrary constant factor of propor- that these scatter the conduction electrons and thus cause
tionality, and it is often also used.] the electrical resistance in the normal state. What happens
We state the conditions under which the general theory to the superconducting state when impurities are present?
reduces to the GL theory: Are the pairs broken?
The effect of impurities can be studied theoretically by
1) The temperature is near the transition temperature,
Tc − T Tc ⇔ ∆ kB Tc . adding an external potential U (r) besides the kinetic ener-
2) The order parameter is not changing too steeply, ∇∆ gy (119),
∆/ξ0 . h̄2 2 h̄2 2
− ∇ →− ∇ + U (r). (246)
2m 2m
The Ginzburg-Landau theory can be understood as Tay-
In the simplest case one can study a delta-function poten-
lor expansion in both ∆ and ∇∆ where one keeps only the
tial U (r) = uδ(r−r0 ), where r0 is the location of the impu-
lowest order terms.
rity. [This scatters electrons from one plane wave state (34)
From microscopic theory one can derive the following to another but takes up no volume.] The interesting result
expressions for the parameters of the GL theory of this calculation is that out of the GL coefficients (240)-
T − Tc (243) only γ (242) changes. The fact that Tc is not chan-
α = N (0) (240) ged could be understood by saying that although electrons
Tc
scatter from one state to another, they always can find a
7ζ(3)N (0) new partner to form a pair.
β= (241)
8(πkB Tc )2
The previous does not hold in “unusual” superfluid sta-
7ζ(3)N (0) 2
γ= ξ 0 (242) tes where vk depends on direction, i.e. in high temperature
12h̄2 superconductors or in 3 He, and pairs are broken there.
q = 2e, (243)
What happens to γ? Let us study the case where the
where ζ(3) = 1.202, ξ0 is defined in equation (174) and density of impurities is so large that the mean free path
e is the electron charge (e < 0). The last relation can be ` of a particle between scattering events is much smaller
easily understood: because Ψ the wave function of a pair, than ξ0 . This case ` ξ0 is called the “dirty limit”. The
its kinetic energy contains the pair charge q = 2e. electron propagates randomly as its direction changes after
The two first equations [(240) and (241)] can be derived an average flight by `. A simple calculation gives that if the
directly from the energy functional (192). The calculation total distance √ travelled by the the particle is ξ0 , it is at the
is though somewhat complicated and therefore is not done distance R ≈ `ξ0 from its starting point:
here. The two other equations [(242) and (243)] can be !2
derived from the inhomogeneous state theory mentioned
X X XX
2
R = ∆xi = (∆xi )2 + ∆xi · ∆xj
above (199) when also the vector potential is included in i i i j
the kinetic energy (119): j6=i
X ξ 0
h̄2 2
2
≈ (∆xi )2 ≈ `2 = ξ0 `. (247)

1 h̄
− ∇ → ∇ − eA . (244) i
`
2m 2m i
This is because on the average ∆xi · ∆xj = 0 when j 6= i,
From coefficients (240) and (242) we get for the GL co- and ξ0 /` is the number of terms in the last summation
herence length (224)
s Supposing now that the estimate about the pair size
(174) concerns the total path length, we arrive at the
r
h̄2 γ 7ζ(3) 1
ξGL (T ) = = ξ0 result
|α| √ that in the dirty limit the pair size is reduced to
p
12 1 − T /Tc
∼ `ξ0 ξ0 . This could be described by saying that in
1 dirty case the members of the pair loose each other slower
= 0.837ξ0 p . (245)
1 − T /Tc than in the pure case.
The coherence lengths are thus of same order of magnitude For the GL parameter γ instead of (242) we get
except the case T → Tc , where ξGL (T ) → ∞.
N (0)
It turns out that when one goes very near Tc , the GL γ∼ `ξ0 , (248)
h̄2
theory is not valid any more. This is caused by critical
fluctuations, that are common to all second order phase i.e. a gradient of the order parameter does not increase the
transitions. These are more discussed in the course of sta- energy as much as in the case of a pure superconductor. It
tistical physics. In ordinary superconductors the tempe- follows from equation (236) that κ increases with increasing
rature region where critical fluctuations are important is impurity. Pure elemental metals are almost exclusively of
vanishingly small. type I, but they change to type II with increasing impurity.
Effect of impurities 5.4 Type II superconductivity
26
Previously we studied the intermediate state of a type For a type II superconductor one gets the following phase
I superconductor. In type II superconductor the interface diagram.
energy is negative. It follows that the magnetic field penet- H
rates into the sample is as small units as possible in order
normal state
to maximize the amount of the interface. Because of flux B/µ 0
quantization (234) we deduce that the smallest unit is one Hc2
flux quantum Φ0 . We sketch the corresponding solution of
the GL equations. In cylindrical coordinates mixed state Hc
Hc1
Ψ(r, ϕ, z) = C(r)eiϕ . (249)
Meissner state T
C(r) Tc H c1 H c2 H
ϕ
r Between the critical fields Hc1 and Hc2 the magnetic
r field partly penetrates to the sample. The density of vor-
ξ GL tices is n = B/Φ0 . The solution for one vortex describes
the situation at fields near Hc1 , where the vortices are far
Here the phase φ of the order parameter is the sa- apart from each other. In increasing field also the density of
me as the azimuthal angle ϕ of the cylindrical coordi- vortices increases until at Hc2 they are so dense that there
nates. Because the order parameter is independent of z, is no space for superconductivity in between. Because the
we consider it in the x-y plane. Ψ has to be continuo- vortex core size is approximately ξGL , we estimate from
2
us everywhere. It has a zero at r = 0, where it is ana- this that Bc2 ∼ Φ0 /ξGL . An accurate calculation with GL
lytic in spite of the singularity of the coordinate system, theory gives
Ψ(x, y, z) = a(x + iy) + O(r2 ). h̄ Φ0
Bc2 = µ0 Hc2 = = 2 . (253)
The dependence eiϕ (249) on the phase causes a cur- 2
2|e|ξGL 2πξGL
rent (227) that circulates around the z axis. We suppose
Near Hc2 the order parameter goes continuously to zero.
that also the vector potential A is in the direction of the
In this case the third order term Ψ|Ψ|2 in the GL equation
azimuthal angle, A = A(r)ϕ̂. The current
(210) can be dropped, and the remaining equation is the
same as the Schrödinger equation for a charged particle in
h̄
j = 2qγC 2 (r) − qA(r) ϕ̂. (250) constant magnetic field B. We leave the mathematics of
r this problem to the condensed matter course.
At large r the order parameter approaches its equilibrium In the equilibrium state the vortices fill the sample as
value (218). There the current (250) must vanish (expo- uniformly as possible. This leads to a lattice that is hexa-
nentially. Thus gonal.
h̄ ϕ̂
A(r) = (r → ∞). (251)
q r
Requiring that A(r) is regular at origin, we can guess its

shape. Finally we calculate
ẑ d(rA)
B=∇×A= . (252)
r dr
A(r) B(r)
Spectroscopic image of the vortex lattice in NbSe2 at 4.2
Kelvin and 1 Tesla. Dark corresponds to the normal vor-
tex cores, and bright to the superconducting regions. The
vortex lattice imaging by scanning tunneling spectrosco-
λ r λ r py relies on spatial variations of the density of states in
The accurate forms of the functions are obtained by sol- the mixed state. Indeed, the local quasiparticle density of
ving the GL equations, which generally is possible only states is different at the center of vortex cores compared
numerically. to the surrounding superconducting regions. Plotting these
differences as a function of position yields a spectroscopic
The solution of the type (249) is called a quantized vortex
real space image of the Abrikosov vortex lattice. (lähde:
or vortex or flux line. We see from equation (251) that the
magnetic flux associated to a vortex is precisely one flux http://dpmc.unige.ch/gr fischer/)
quantum Φ0 (4). Check as an exercise that the dimensions of the figure
27
and the given field are consistent. Here m = m4 for 4 He and m = 2m3 for 3 He, where m4 and
m3 are the corresponding masses of the atoms (why so?).
Force on a flux line When |Ψ|2 is constant one can define superfluid velocity vs :
because h̄∇φ is momentum,
A Lorentz force
h̄
F = q(E + v × B). (254) vs = ∇φ. (258)
m
acts on a charged particle. This can be generalized to con-
tinuously distributed matter as It follows that
Z ∇ × vs = 0. (259)
F = d3 r(ρE + j × B), (255) We compare this to uniform rotation where
where ρ is the charge density and j the electric current v = Ω × r ⇒ ∇ × v = 2Ω. (260)
density. We apply this to a single flux line in an applied flow
field jext (r). Because of charge neutrality ρ = 0. Supposing We thus get the interesting result that superfluid cannot
that jext (r) is approximately constant in the cross section rotate in such a way that |Ψ|2 is constant. This is called
of the vortex (area ≈ λ2 ), we can calculate the integral in rotation paradox.
transversal plane and get the force action on the flux line The solution of the rotation paradox is that quantized
vortices are formed when superfluid is rotated. Because
Z
F = Φ0 jext × dl. (256) q = 0, we get for a single vortex in the simplest case
This force drives the vortex in direction that is perpendicu- Ψ(r, ϕ, z) = C(r)eiϕ , (261)
lar to the applied current.
h̄
j int jmass = 2mγC 2 (r) ϕ̂. (262)
jext r
C(r)
ϕ
r r
ξ GL
F
Notice that the direction of the force is such that it tends to
decrease the region where the total flow velocity is largest The circulation of the superfluid velocity around a single
(compare to the Magnus force). vortex is
One application of superconductivity is to built strong I I
h
h̄
magnets. Because Hc is relatively small (see the table on dl · v s = m dl · ∇φ = .
m
page 2), superconductors of type II are used. A large part
os applications use Nb-Ti alloy, where 45 wight per cent Vortices are formed in a superfluid when the container
is titanium. Because the alloy is disordered, the mean free is rotated.
path is very short and therefore Hc2 is high, Bc2 ≈ 10 T
at T = 4 K. Notice that if the alloy would be ordered, the
mean free path could in principle be as long as in a pure
element (see condensed matter course for justification).
The motion of flux lines leads to dissipation, which
means that the “superconductor” is not conducting wit-
hout resistance. This can be prevented by grain bounda-
ries, precipitates or other impurities that trap flux lines.
Especially Nb-Ti has titanium precipitates which are not
superconducting. In these regions the flux lines have lower
energy and thus are trapped there.
Rotation of a superfluid
Many of the results described above are also valid in 4 He
and 3 He superfluids. An essential difference is that there
are uncharged (effectively q = 0) so that instead of electric
current (227) one gets mass current
jmass = 2mγ|Ψ|2 h̄∇φ. (257)
28
6. Josephson effect where we also assumed that τ depends only weakly on ener-
gy The result is linear in voltage V . Thus the factor mul-
tiplying it can be identified as conductance (inverse of re-
6.1 Tunneling sistance) G = 1/R, and
Let us consider two metals that separated from each ot-
her by a thin insulating layer. Although the potential in J = V /R. (266)
the insulator is higher that the energy of electrons, accor-
A more complicated result for the current is obtained
ding to quantum mechanics the electrons can tunnel th-
between a superconducting and normal state metal. The
rough the potential barrier. In equilibrium, the chemical
density of states in the superconducting state is
potentials of the metal are equal (why?) We study the case N (0)E
when a current source is connected between the metal, and √ kun E > ∆
Ns (E) = E 2 −∆2 (267)
it causes a potential V across the junction: µ1 − µ2 = eV . 0 kun E < ∆,
U(x) E which is easily obtained from the dispersion relation (155)

taking into account that the states are uniformly distribu-
ted in ξk :ssa (exercise).
µ1
eV µ2 E
µ1
eV ∆ µ2
Metal 1 Metal 2 ∆
(Note. The figures are often drawn as if eV > 0. If one
wants to take into account that e < 0, it is easiest to think N1(E) N2(E)
that also V < 0.)
Metal 1 Metal 2
For the tunneling current we get the expression
Z At zero temperature the current is different from zero
only if |eV | > ∆. One can deduce that the current-voltage
J = c dE τ (E)
relationship is qualitatively of the type shown in the fi-
×{N1 (E)f (E)N2 (E + eV )[1 − f (E + eV )] gure. (More accurately one should take into account that
−N1 (E)[1 − f (E)]N2 (E + eV )f (E + eV )}. the tunneling objects (electrons) are not the same as the
excitations on the superconductor side, but this does not
(263) change the the result, see Tinkham).
Here c is a constant, τ (E) is the tunneling probability and J
N1 and N2 are the densities of states in the metals. The
first term in equation (263) describes tunneling from left to
right and the second from right to left. The Fermi functions
/R
f (43) in both terms take into account that only those cases
V
J=
are counted where the state on the starting side is initially T>0
occupied, and that the electron arrives on the other side to T=0
a state that initially was empty. From equation (263) we
get by direct calculation
∆ eV
Z This kind of dependence of the current on the energy gap
J = c dE τ (E)N1 (E)N2 (E + eV ) is applied, among other things, in measuring the figure on
page 27.
×[f (E) − f (E + eV )]. (264)
Between two superconductors one gets even more
Notice that the factor containing the Fermi functions is complicated current-voltage relationships depending on the
different from zero only in an energy interval of ∼ eV + magnitudes of the energy gaps.
kB T .
J
We study tunneling between two normal state metals.
There the density of states is approximately constant and
we get
/R
Jc
V
J=
Z
J ≈ cN1 (0)N2 (0)τ (0) dE T>0
T=0 eV
×[f (E) − f (E + eV )]
= cN1 (0)N2 (0)τ (0)eV, (265) |∆ 1 -∆ 2 | ∆ 1 +∆ 2
29
The most interesting feature is the current that is obtai- from the factor exp(−iEt/h̄), which for the order parame-
ned at precisely zero voltage. ter Ψ = eiφ |Ψ| is exp(−i2µt/h̄), and thus one gets (273).
6.2 Josephson effect We apply (273) to a Josephson junction. For the phase
We consider two superconductors that are weakly difference (271) we get
coupled to each other. We assume in the beginning that d∆φ 2eV
B = A = 0. = , (274)
dt h̄
since the difference in the chemical potentials is related to
the voltage V by ∆µ = µ2 − µ1 = −eV .
Ψ1 Ψ2 The equations (272) and (274) are known as Josephson
equations. The first gives that in equilibrium (V = 0) a
constant current flows trough that depends sinusoidally on
the phase difference ∆φ ≡ φ2 − φ1 . This is known as dc
Analogously to the phenomenological derivation of the Josephson effect.
GL theory, we form expression for the energy associated If the voltage V is constant, one gets from equation (274)
with such a junction. Let Ψ1 and Ψ2 denote the order pa- that the phase grows linearly in time,
rameters on the left and right hand sides. We require 1) rea-
2eV
lity, 2) independence of a constant phase factor exp(iφ). 3) ∆φ = t. (275)
independence on exchange of the two sides, and 4) take on- h̄
ly the leading order terms. This way we get the Josephson Substituting this in to equation (272) one gets alternating
energy current at angular frequency
2e
FJ = −a(Ψ∗1 Ψ2 + Ψ1 Ψ∗2 ) = −2a Re(Ψ∗1 Ψ2 ). (268) ω= V. (276)
h̄
We substitute This is known as ac Josephson effect.
Ψ1 = Ceiφ1 , Ψ2 = Ceiφ2 , (269) At voltage 0.1 mV (which is typical in the figure on page
29) the equation (276) gives the frequency ν = ω/2π = 48
and get GHz.
FJ = −EJ cos ∆φ. (270)
Using Josephson junctions it is possible to make sensiti-
We have defined the phase difference ve measuring devices. For example, equation (276) makes
possible a voltage standard, as the frequency can accura-
∆φ = φ2 − φ1 (271)
tely be measured.
and EJ = 2aC 2 . The current source driving the junction makes in time dt
From the Josephson energy (270) we get the electric cur- the work V Jdt. According to energy conservation we must
rent through the junction have
dFJ
= V J. (277)
J = Jc sin ∆φ, (272) dt
We see that using this relation we can derive the third
where Jc = (q/h̄)EJ = (2e/h̄)EJ . relation based on any pair of the relations (270), (272) and
Justification of (272): We substitute (226) in to the GL (274).
energy (207). We make variation of it with respect φ on If the vector potential A is different from zero, the phase
both sides (1-dimensional model is sufficient). The varia- difference (271) used above should be redefined as follows
tion gives surface terms, and these counted together with Analogously to equation (227) one defines an gauge inva-
the Josephson energy (270) should vanish, which gives re- riant phase difference
lation (272).
2e 2
Z
Another important relation is the following, which gives ∆φ = φ2 − φ1 − dl · A. (278)
h̄ 1
the time derivative of the phase,
dφ 2µ
=− . (273)
φ2
dt h̄ φ1
Justification of (273): the order parameter Ψ of the GL Γ
theory was interpreted as the wave function describing a
Cooper pair. In equilibrium the Cooper pairs are in equili-
Φ
brium with electrons so that the energy of a Cooper pair is
2µ, twice the electron chemical potential. The time depen-
dence of an energy eigenstate in quantum mechanics comes
30
3.5
Let us consider a ring that contains one Josephson junc- E
EJ 3.0
tion. The total flux is
2.5
Z I Z 2 Z 1
Φ = da · B = dl · A = dl · A + dl · A (279) 2.0
1 2 1.5
1.0
where the last form has two terms, the former across the
junction and the latter over the rest of the ring. Using (227) 0.5
and j = 0 the latter contribution is 0.0

0.0 0.2 0.4 0.6 0.8 1.0
Φ/Φ0
1 1
h̄ h̄
Z Z
dl · A = dl · ∇φ = (φ1 − φ2 + 2πN ). (280)
2e 2e
2 2
6.3 Alternative treatment of the Joseph-
Substituting in (279) we get the phase difference across the son effect
junction We showed above that the BCS state (171) can be repre-
2πΦ sented as a superposition of states |ni, where the pair num-
∆φ = + 2πN. (281)
Φ0 ber n is fixed, X
If there are more junctions, this generalizes to |ψi = einφ fn |ni. (285)
n
X 2πΦ Here we have written the phase φ (176) explicitly and the-
∆φj = + 2πN. (282)
j
Φ0 refore fn is real valued.
We consider two superconductors L and R, which are
J weakly coupled to each other. The coupling can be simplest
be described by adding to the Hamiltonian the term
J2 J1
∆φ2 ∆φ1 EJ X X
Φ ȞT = − |m − 1, n + 1ihm, n|
2 m n
!
+|m + 1, n − 1ihm, n| . (286)
Consider the circuit that has thick ( λ) superconduc-
ting wires and two Josephson junctions. This is described Here |m, ni is a state where the superconductor L has m
by equations pairs and superconductor R has n pairs. The first term
describes process where one pair jumps from L to R, and
2πΦ the second term describes the opposite process. EJ is a
∆φ1 + ∆φ2 = + 2πN
Φ0 constant describing the strength of the coupling.
J = Jc1 sin(∆φ1 ) − Jc2 sin(∆φ2 ). (283) The states of uncoupled superconductors are described
by the product of two states of the type (285):
For simplicity assume Jc1 = Jc2 . By calculation one sees
that the two junctions behave as if a single junction (272),
XX
|ψi = eikφL eilφR fk fl |k, li. (287)
whose critical current Jc depends on the flux k l
It the coupling is weak, it can be treated as a small pertur-

πΦ
Jc = 2Jc1 cos
. (284) bation. The first order correction to the energy is obtained
Φ0
from the expectation value
Changing the field thus changes the critical current. This is
E1 = hψ|ȞT |ψi. (288)
used in very sensitive measuring devices of magnetic field.
The device based on the circuit is called dc SQUID (Su- Substituting the expressions (286) and (287) we get
perconducting quantum interference device).
EJ X X i(φL −φR )
∆φ2 E1 = − (e fm−1 fn+1
2 m n
+e−i(φL −φR ) fm+1 fn−1 )fm fn
Φ ∆φ1 ≈
−EJ cos(φR − φL ) (289)
2
P P
supposing m fm−1 fm ≈ m fm = 1. The result is the
∆φ3 same as obtained above (270) in a different way.
We can study a ring with three Josephson junctions. At Based on the same assumptions, we can also calculate
certain fields Φ ≈ 21 Φ0 this has two possible energy states the current
d
that correspond to currents flowing in opposite directions. J = 2e hψ|ňR |ψi, (290)
dt
31
where XX of S is marked by φ.) The tunneling of one pair changes
ňR = n|m, nihm, n| (291) the electric charge Q of the island by 2e. This leads to a
m n change of the electrostatic potential Q2 /2C, where C is the
is the pair number operator on the side R. capacitance of the island. The capacitance can be estima-
ted using the plate-capacitor formula C = A/d, where A
The time derivative in equation (290) can easily be calcu- is the area of the Josephson junction, the permeability
lated using time dependent Schrödinger equation. More di- of the insulator and d the thickness of the insulating layer.
rectly the same result is obtained in the Heisenberg pictu- The order of magnitude of the charging energy is given by
re, where the states are time independent but the operators E ≡ e2 /2C.
c
obey the equation of motion
The phase is well defined if the charging energy is small
dňR i i compared to the Josephson energy, Ec EJ . In the other
= Ȟ, ňR = ȞT , ňR . (292)
dt h̄ h̄ case, when Ec has same order of magnitude as EJ (or is lar-
ger), the phase is a quantum mechanical variable in a sense
Here the latter equality follows because the only term in
that it should be described by a wave function ψ(φ). This
the Hamiltonian that does not commute with ňR is the
is known as macroscopic quantum mechanics, as φ descri-
tunneling term (286). Substituting (286) we get
bes a large number of electrons but still behaves according
to quantum laws.
dňR iEJ X X
= − − |m − 1, n + 1ihm, n| Other conditions for the observation of quantum beha-
dt 2h̄ m n
! vior of phase is low temperature, and that the charge can-
not escape too fast by other means. In practice this means
+|m + 1, n − 1ihm, n| . (293)
very small junctions of area A < 10−12 m2 .
In order to get a quantitative theory, we construct a
The expectation value of this can be calculated similarly Hamiltonian. The charging energy is 4E (n − Q /2e)2 .
c 0
as above, and we get This is the capacitive energy Q2 /2C caused by charge
2e Q = 2en + Q0 , where 2en describes the charge depen-
J = EJ sin(φR − φL ). (294) dent on the number of Cooper pairs and Q0 is an inde-
h̄
pendent contribution that is called the background charge.
Also this result is the same as above (272). Although n gets only integer values, Q0 can be an arbit-
We apply the time dependent Schrödinger equation rary real number, since it is changed, for example, by the
motion of an impurity ion in the insulator between the ca-
∂ pacitor plates. The charging energy is diagonal when using
ih̄ |ψi = Ȟ|ψi (295)
∂t the pair number eigenstates |ni. The tunneling part was
postulated above (286), but here we concentrate only on
to the state (185). By definition of the chemical potential the island. (We suppose the other superconductor so large
µ, the energies of states with different number of particles that the changes in its particle number are unimportant.)
differ by µ, i.e., Ȟ = 2µŇ + constant. We get Thus we get the Hamiltonian
dφ 2µ
" 2
=− . (296)
X Q0
dt h̄ Ȟ = 4Ec |ni n + hn|
n
2e
which also was derived above (273). #
EJ
− (|n + 1ihn| + |n − 1ihn|) . (297)
2
6.4 Macroscopic quantum mechanics
Previously we considered the order parameter and its It is interesting to express this Hamiltonian using the
phase as classical quantities, which have definite values. phase eigenstates. Using the transformation formulas (185)
The precision of the phase, however, is limited by the uncer- and (186) we get the form
tainty relation (188). Let us study when this restriction Z π
∂ Q0
2
becomes essential. Ȟ = dφ|φi 4Ec i +
−π ∂φ 2e

S1 2e S
−EJ cos φ hφ|. (298)
φ1=0 φ
Q=2en-Q0
Verify this as an exercise. We see that in φ representation
C
the operator giving n is
∂
We consider a small superconducting body S, which we nop = i . (299)
call island. It is connected via a Josephson junction to anot- ∂φ
her larger superconductor S1 . (Without loosing generality This and the phase have commutation relation [φ, nop ] =
we can choose the phase of S1 to vanish, and the phase −i, as generally applies to quantities that are obtained by
32
Fourier transform from each other. [Depending on defini- We note that the classical problem considered here (302)
tions the sign of the commutator is ±i. Here the sign of is the same as for a simple pendulum, and the quantum
the exponent in equation (185) is chosen such that it leads mechanical problem (300) is the same as for an electron in
to the same definition of φ as usually used in the literature a 1-dimensional crystal with sinusoidal periodic potential.
of superconductivity.]
Because the Hamiltonian (298) is diagonal, it is more
simple to write it as
2
∂ Q0
H = 4Ec i + − EJ cos φ. (300)
∂φ 2e
This Hamiltonian determines the form of the phase wave
function ψ(φ). If Ec EJ , can ψ(φ) be almost a delta
function, and the phase is well defined. In the opposite
case ψ(φ) is distributed and the phase uncertain.
Finally we mention that the macroscopic quantum mec-
hanics is actively studied at present. One of the motivations
is to make a qubit, the bit of a quantum computer. For
example, the quantum tunneling between two macroscopic
states has been observed experimentally in the ring with
three Josephson junctions described above [van der Wal et
al, Science 290, 773 (2000).].
Alternative derivation
Here we give alternative derivation for the quantum pro-
perties of phase If you are satisfied with the above, you can
skip this.
Tunnel junctions always have also electric capacitance
C. Its energy is
2
1 1 h̄
EQ = CV 2 = C φ̇2 (301)
2 2 2e
where we have used formula (275). This can be thought as
some kinetic energy and the corresponding potential energy
is then the Josephson energy (270). Out of these we can
form the Lagrange function
2
1 h̄
L= C φ̇2 + EJ cos φ. (302)
2 2e
We calculate the canonical momentum

2
∂L h̄ h̄ h̄
p= =C φ̇ = CV = Q. (303)
∂ φ̇ 2e 2e 2e
The Hamiltonian is
Q2
H= − EJ cos φ. (304)
2C
The commutation rule for canonical variables x ja p gives
[p̌, x̌] = −ih̄, (305)
and based on analogy we have
[Q̌, φ̌] = −2ei. (306)
Using Q̌ = 2eň − Q0 we see that we get the same formulas
as before (299) and (300) except of irrelevant sign difference
in (299) and the Q0 contribution in (300).
33
7. Conclusion Appendix
The course had two main topics: the microscopic theo- A. Maxwell’s equations
ry (BCS) and the macroscopic theories (thermodynamics,
Maxwell’s equations are
GL, Josephson effect). The purpose was to give basic un-
derstanding of both. In addition we considered several mo- ρ
∇·E = , (307)
re detailed questions. In conclusion one should read again 0
the introduction to see if the topics mentioned there were ∂B
sufficiently understood. ∇×E = − , (308)
∂t
∇ · B = 0, (309)
∂E
∇ × B = 0 µ0 + µ0 j. (310)
∂t
Here E is the electric field and B the magnetic field. In
many cases it is convenient to represent them using poten-
tials
∂A
E = −∇ϕ − , (311)
∂t
B = ∇ × A, (312)
where ϕ is the scalar potential and A the vector potential.

The Maxwell’s equations contain the charge density ρ
and electric current density j. In studying electromagne-
tic phenomena in materials, it is useful to separate the
charges arising from the polarization and other charges.
Correspondingly, the electric current can be divided to the
current arising from magnetization and polarization of the
medium, and to other currents. Without proper justifica-
tion we claim that this can be written as the following
equations
ρ = ρf − ∇ · P, (313)
∂P
j = jf + ∇ × M + . (314)
∂t
Here P is the electric polarization and M the magnetiza-
tion of the medium. Other charges and currents are denoted
by index f meaning free. We substitute these relations to
Maxwell’s equations (307)-(310) and define two new fields
D ≡ 0 E + P, (315)
1
H≡ B − M. (316)
µ0
We get Maxwell’s equations in a medium
∇·D = ρf ,
∂B
∇×E = − ,
∂t
∇ · B = 0,
∂D
∇×H = + jf . (317)
∂t
34

Superconductivity 763645S: Erkki Thuneberg

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Superconductivity 763645S: Erkki Thuneberg

Uploaded by

Copyright:

Available Formats

SUPERCONDUCTIVITY

The web page of the course is

We are interested in superconductivity because of the

Examination: • Superconductivity has several technical applications.

Short content of the course

• short revisal/introduction to some fundamental re-

• Thermodynamics in magnetic field

• M. Tinkham, Introduction to Superconductivity

• A.L. Fetter and J.D. Walecka, Quantum theory of

• J.B. Ketterson and S.N. Song, Superconductivity

1.1 Properties of superconductors

normal state then cooled

The temperature below which superconductivity occurs is

Infinte conductivity "  2 #

If σ → ∞ then E → 0. Maxwell’s equation

Hc2 Isotope effect

Persistent currents and flux quantization

that generates to the

it is an integer multiple of the flux quantum 3 (bcc)

C The superfluid state is achieved in some gases of alkali

It should be emphasized that the system we study has on

ρi = eβ(F −Ei ) . (8)

Here the constant β can be interpreted as inverse tem-

The expectation value hÂi of an arbitrary operator Â is

ρ̂ = eβ(F −Ĥ) . (10)

This is the Gibbs distribution represented in operator form.

We additionally define the entropy

S = −kB hln ρ̂i (12)

The purpose above was to show that ideal Bose and

Here the maximal kinetic energy of an occupied state is

Equation (42) gives at zero temperature

where the vector relation should be understood as Hx = Gn (T, H) − Gn (T, 0) = − 21 V µ0 H 2 . (58)

The expression in parenthesis gives the field energy near

Minimizing with respect to D we get

state (of type I superconductor).

It is obvious that there is some extra energy associated

h̄2 k 2 How can we understand the electrical resistivity of a nor-

Ψ(r1 σ1 , r2 σ2 ) = −Ψ(r2 σ2 , r1 σ1 ). (79)

Ψ(r1 σ1 , r2 σ2 ) = ψ0 (r1 σ1 , r2 σ2 ) − ψ0 (r2 σ2 , r1 σ1 ). (80)

In a similar way but with slightly more work (Landau-

×δk1 +k2 ,k3 +k4 ǎ†k3 ↑ ǎ†k4 ↓ ǎk2 ↓ ǎk1 ↑ . (130)

Tr(. . . e−β ǨHF ) We write this into the form

∆(T = 0) = πe−γ kB Tc = 1.764 kB Tc . (168)

heat at constant volume is the same.

We define the inverse Fourier transform of ∆(q) in the

We realize that h̄q the total momentum of a pair. Thus we

It is observed that in the normal state Ψ = 0 is comple-

Effect of impurities 5.4 Type II superconductivity

Requiring that A(r) is regular at origin, we can guess its

U(x) E which is easily obtained from the dispersion relation (155)

and j = 0 the latter contribution is 0.0

It the coupling is weak, it can be treated as a small pertur-

We calculate the canonical momentum

where ϕ is the scalar potential and A the vector potential.

You might also like

Infinte conductivity " 2 #