Theory of Many-Particle Systems

Lecture notes for P654, Cornell University, spring 2005.

c
Piet Brouwer, 2005. Permission is granted to print and copy these notes, if kept together
with the title page and this copyright notice.
Chapter 1

Preliminaries

1.1 Thermal average

In the literature, two types of theories are constructed to describe a system of many particles:
Theories of the many-particle ground state, and theories that address a thermal average at
temperature T . In this course, well limit ourselves to finite-temperature thermal equilibria
and to nonequilibrium situations that arise from a finite-temperature equilibrium by switch-
ing on a time-dependent perturbation in the Hamiltonian. In all cases of interest to us, results
at zero temperature can be obtained as the zero temperature limit of the finite-temperature
theory.
Below, we briefly recall the basic results of equilibrium statistical mechanics. The thermal
average of an observable A is defined as
,
hAi = tr A (1.1)
where is the density matrix and A is the operator corresponding to the observable A.
The density matrix describes the thermal distribution over the different eigenstates of
the system. The symbol tr denotes the trace of an operator, the summation over the
expectation values of the operator over an orthonormal basis set,
X
tr A = hn|A|ni. (1.2)
n

The basis set {|ni} can be the collection of many-particle eigenstates of the Hamiltonian H,
or any other orthonormal basis set.
In the canonical ensemble of statistical mechanics, the trace is taken over all states with
N particles and one has
1
= eH/T , Z = tr eH/T , (1.3)
Z
1
2 CHAPTER 1. PRELIMINARIES

is the Hamiltonian. Using the basis of N -particle eigenstates |ni of the Hamiltonian
where H

H, with eigenvalues En , the thermal average hAi can then be written as

P En /T
e hn|A|ni
hAi = n P En /T . (1.4)
ne

In the grand-canonical ensemble, the trace is taken over all states, irrespective of particle
number, and one has
1
= e(HN )/T , Z = tr e(HN )/T . (1.5)
Z
Here is the chemical potential and N is the particle number operator. Usually we will
include the term N into the definition of the Hamiltonian H, so that expressions for the
thermal average in the canonical and grand canonical ensembles are formally identical.

1.2 Schr
odinger, Heisenberg, and interaction picture
There exist formally different but mathematically equivalent ways to formulate the quantum
mechanical dynamics. These different formulations are called pictures. For an observable
the expectation value in the quantum state |i is
A,

A = h|A|i. (1.6)
The time-dependence of the expectation value of A can be encoded through the time-
dependence of the quantum state , or through the time-dependence of the operator A, or
through both. You can find a detailed discussion of the pictures in a textbook on quantum
mechanics, e.g., chapter 8 of Quantum Mechanics, by A. Messiah, North-Holland (1961).

1.2.1 Schr
odinger picture
In the Schr odinger picture, all time-dependence is encoded in the quantum state |(t)i;
operators are time independent, except for a possible explicit time dependence.1 The time-
evolution of the quantum state |(t)i is governed by a unitary evolution operator U (t, t0 )
that relates the quantum state at time t to the quantum state at a reference time t0 ,
|(t)i = U (t, t0 )|(t0 )i. (1.7)
1
An example of an operator with an explicit time dependence is the velocity operator in the presence of
a time-dependent vector potential,
1  e 
=
v i~ A(r, t) .
m c

1.2. SCHRODINGER, HEISENBERG, AND INTERACTION PICTURE 3

Quantum states at two arbitrary times t and t0 are related by the evolution operator U (t, t0 ) =
U (t, t0 )U (t0 , t0 ). The evolution operator U satisfies the group properties U (t, t) = 1 and
U (t, t0 ) = U (t, t00 )U (t00 , t0 ) for any three times t, t0 , and t00 , and the unitarity condition

U (t, t0 ) = U (t0 , t). (1.8)

The time-dependence of the evolution operator U (t, t0 ) is given by the Schrodinger equation

U (t, t0 ) U (t, t0 ).
i~ =H (1.9)
t

For a time-independent Hamiltonian H, the evolution operator U (t, t0 ) is an exponential

function of H,
U (t, t0 ) = exp[iH(t
t0 )/~]. (1.10)
If the Hamiltonian H is time dependent, there is no such simple result as Eq. (1.10) above,
and one has to solve the Schrodinger equation (1.9).

1.2.2 Heisenberg picture

In the Heisenberg picture, the time-dependence is encoded in operators A(t), whereas the
is related to
quantum state |i is time independent. The Heisenberg picture operator A(t)
the Schrodinger picture operator A as
= U (t0 , t)AU
A(t) (t, t0 ), (1.11)

where t0 is a reference time. The Heisenberg picture quantum state |i has no dynamics
and is equal to the Schrodinger picture quantum state |(t0 )i at the reference time t0 . The
then follows from Eq. (1.9) above,
time evolution of A(t)

dA(t) i
= [H(t), + A(t) ,
A(t)] (1.12)
dt ~ t

where H(t) = U (t0 , t)H(t
0 )U (t, t0 ) is the Heisenberg representation of the Hamiltonian and
the last term is the Heisenberg picture representation of an eventual explicit time dependence
of the observable A. The square brackets [, ] denote a commutator. The Heisenberg picture

Hamiltonian H(t) also satisfies the evolution equation (1.12),

dH(t) i
H
H(t)
= [H(t),
H(t)] + = .
dt ~ t t
4 CHAPTER 1. PRELIMINARIES

For a time-independent Hamiltonian, the solution of Eq. (1.12) takes the familiar form

= eiH(tt
0 )/~ iH(tt0 )/~
A(t) Ae . (1.13)

You can easily verify that the Heisenberg and Schrodinger pictures are equivalent. Indeed,
in both pictures, the expectation values (A)t of the observable A at time t is related to the
0 ) and quantum state |(t0 )i at the reference time t0 by the same equation,
operators A(t

(A)t = h(t)|A|(t)i
= h(t0 )|U (t0 , t)AU
(t, t0 )|(t0 )i

= h|A(t)|i. (1.14)

1.2.3 Interaction picture

The interaction picture is a mixture of the Heisenberg and Schrodinger pictures: both the
are time dependent. The interaction picture is
quantum state |(t)i and the operator A(t)
usually used if the Hamiltonian is separated into a time-independent unperturbed part H 0
and a possibly time-dependent perturbation H 1,

H(t) 0 + H
=H 1. (1.15)

In the interaction picture, the operators evolve according to the unperturbed Hamiltonian
H 0,
= eiH 0 (tt0 )/~ A(t
A(t) 0 )eiH 0 (tt0 )/~ , (1.16)
whereas the quantum state |i evolves according to a modified evolution operator UI (t, t0 ),
I (t, t0 ) = eiH 0 (tt0 )/~ U (t, t0 ).
U (1.17)

You easily verify that this assignment leads to the same time-dependent expectation value
(1.14) as the Schrodinger and Heisenberg pictures.
The evolution operator that relates interaction picture quantum states at two arbitrary
times t and t0 is
I (t, t0 ) = eiH 0 (tt0 )/~ U (t, t0 )eiH 0 (t0 t0 )/~ .
U (1.18)
The interaction picture evolution operator also satisfies the group and unitarity properties.
is readily obtained
A differential equation for the time dependence of the operator A(t)
from the definition (1.16),

dA(t) i
A(t)
= [H 0 , A(t)] + . (1.19)
dt ~ t

1.2. SCHRODINGER, HEISENBERG, AND INTERACTION PICTURE 5

The time dependence of the interaction picture evolution operator is governed by the interaction-
picture representation of the perturbation H 1 only,

UI (t, t0 ) 1 (t)U
I (t, t0 ),
i~ =H (1.20)
t
where
1 (t) = eiH 0 (tt0 )/~ H
H 1 eiH 0 (tt0 ) . (1.21)
The solution of Eq. (1.20) is not a simple exponential function of H 1 , as it was in the case

of the Schrodinger picture. The reason is that the perturbation H1 is always time dependent
in the interaction picture, see Eq. (1.21).2 (An exception is when H 1 and H 0 commute, but
in that case H 1 is not a real perturbation.) However, we can write down a formal solution
of Eq. (1.20) in the form of a time-ordered exponential. For t > t0 , this time-ordered
exponential reads Rt
I (t, t0 ) = Tt e(i/~) t0 dt0 H 1 (t0 ) .
U (1.22)
In this formal solution, the exponent has to be interpreted as its power series, and in each
term the time-ordering operator Tt arranges the factors H with descending time arguments.
For example, for a product of two operators H 1 taken at different times t and t0 one has
1 (t)H 1 (t0 ) if t > t0 ,

0 H
Tt H1 (t)H1 (t ) = 1 (t0 )H
1 (t) if t0 > t. (1.23)
H
Taking the derivative to t is straightforward now: since t is the largest time in the time
integration in Eq. (1.22), the time-ordering operator ensures that upon differentiation to t
the factor H 1 (t) always appears on the left, so that one immediately recovers Eq. (1.20).
Similarly, the conjugate evolution operator U (t0 , t) is given by an anti-time-ordered expo-
nential,
(i/~) tt dt0 H
1 (t0 )
R
UI (t0 , t) = Tt e 0 , (1.24)
where the operator Tt arranges the factors H 1 with ascending time arguments.
Operators in the Heisenberg picture can be expressed in terms of the corresponding
operators in the interaction picture using the evolution operator UI . Denoting the Heisenberg
picture operators with a subscript H, one has
AH (t) = UI (t0 , t)A(t)
U I (t, t0 ), (1.25)
2
Indeed, one would be tempted to write
1 (t0 )
Rt
I (t, t0 ) = e(i/~) dt0 H
U t0
.
However, since H 1 does not commute with itself if evaluated at different times, taking a derivative to t does
not simply correspond to left-multiplication with H 1 (t).
6 CHAPTER 1. PRELIMINARIES

t0 t
Figure 1.1: Contour used to the operator AH (t) in the Heisenberg picture from the corresponding
in the interaction picture.
operator A(t)

t0 t
Figure 1.2: Keldysh contour. The arguments t and t 0 can be taken on each branch of the contour.

where A(t) is the interaction picture operator, see Eq. (1.16). Using the formal solution
(1.22) for the evolution operator, this expression can be written as
(i/~) tt dt0 H
1 (t0 ) (i/~) tt dt0 H
1 (t0 )
h R i h R i
AH (t) = Tt e 0
A(t) Tt e 0

(i/~) tt dt0 H 1 (t0 )

h R t0 0 i h i
0
R
= Tt e(i/~) t dt H1 (t ) A(t) Tt e 0 . (1.26)

It is tempting to combine the two integrals in the exponent into a single integral starting out
at time t0 , going to t, and then returning to t0 . In order to do so, one has to be careful that
the order of the operators is preserved: the time-ordered integral from t0 to t should be kept

at the right of the operator A(t), whereas the anti-time-ordered integral from t back to t0

should remain at the left of A(t). The correct order is preserved if we shift to an integration
over a contour c that starts at time t0 , goes to time t, and returns to time t0 , and order all
products according to their position along the contour: operators that have time arguments
that appear late in the contour appear on the left of operators with time arguments that
appear early in the contour. The contour c is shown in Fig. 1.1. Note that early and
late according to the contour-ordering does not have to be earlier or later according to
the true physical time. Writing such a contour-ordering operator as Tc , we then find the
formal expression
0 0
AH (t) = Tc e(i/~) c dt H1 (t ) A(t).
R
(1.27)
In fact, instead of the contour shown in Fig. 1.1, one can use the Keldysh contour, which
is shown in Fig. 1.2. The Keldysh contour starts from t0 , extends past time t up to infinite
time, and then returns to t0 . The integration for times larger than t is redundant and cancels
from the exponent.
In these notes well use all three pictures. Mostly, the context provides sufficient informa-
tion to find out which picture is used. As a general rule, operators without time argument
1.3. IMAGINARY TIME 7

are operators in the Schrodinger picture, whereas operators with time argument are in the
Heisenberg and interaction pictures. Well use this convention even if the Schrodinger picture
operators have an explicit time dependence, cf. Eq. (1.15) above.

1.3 imaginary time

When performing a calculation that involves a thermal average, it is often convenient to
introduce operators that are a function of the imaginary time .
Unlike real time arguments, imaginary times have no direct physical meaning. Imagi-
nary time is used for the theorists convenience, because Green functions, the mathematical
machinery used to approach the many-particle problem, have very useful mathematical prop-
erties if regarded as a function of a complex time and frequency, instead of just real times and
frequencies. The imaginary time formalism is usually not used for time-dependent Hamilto-
nians: it would be awkward to specify how a certain time dependence H(t) translates into
imaginary time!
You can see why imaginary times can be useful from the following observation: A thermal

average brings about a negative exponents eH/T , whereas the real time dependence brings

about imaginary exponents eiHt . For calculations, it would have been much easier if all
exponents were either real or imaginary. To treat both real and imaginary exponents at
the same time is rather awkward. Therefore, well opt to consider operators that depend

on an imaginary time argument, so that the imaginary exponent eiHt is replaced by eH .
If only imaginary times are used, only real exponents occur. This leads to much simpler
calculations, as well see soon. The drawback of the change from real times to imaginary
times is that, at the end of the calculation, one has to perform an analytical continuation
from imaginary times to real times, which requires considerable mathematical care.
In the Schrodinger picture, the imaginary time variable appears in the evolution operator
U (, 0 ), which now reads
0 )/~
(, 0 ) = eH(
U , ~ U (, 0 ) = H
U (, 0 ). (1.28)

Since a thermal average necessarily involves a time-independent Hamiltonian, we could give
an explicit expression for the evolution operator. Note that the evolution operator (1.28)
satisfies the group properties U (, ) = 1 and U
(, 0 )U
( 0 , 00 ) = U (, 00 ), but that it is no
longer unitary.
In the Heisenberg picture, the imaginary time variable appears through the relation

/~ H
/~
)
A( 1
) = eH
A( Ae , = [H, A] , (1.29)
~
8 CHAPTER 1. PRELIMINARIES

0
t

1/T

Figure 1.3: Relation between real time t and imaginary time .

where A is the corresponding real-time operator. Again, H is assumed to be time indepen-

dent.
In the interaction picture, the time dependence of the operators is given by

A( H 0 /~ , A( ) = 1 [H
) = eH 0 /~ Ae 0 , A]
, (1.30)
~
whereas the time dependence of the quantum state |( )i is given by the interaction picture
evolution operator

I (, 0 ) = eH 0 /~ eH(
0 )/~ H
0
U e 0 /~ , ~ UI (, 0 ) = H
1 ( )U (, 0 ). (1.31)


1 ( ) = eH0 /~ H
1 eH0 /~ is the perturbation in the interaction picture.
Here H
The relation between the real time t and the imaginary time is illustrated in Fig. 1.3.
Since the interaction Hamiltonian H 1 ( ) depends on the imaginary time , the solution
of Eq. (1.31) is not a simple exponential. However, as in the previous section, we can write
down a formal solution using the concept of a time-ordering operator acting on imaginary
times, R 00 00
UI (, 0 ) = T e(1/~) 0 d H1 ( ) . (1.32)
The imaginary-time-ordering operator T arranges factors H 1 ( 00 ) from left to right with
0
descending time arguments (ascending time arguments if > ). Similarly, the Heisenberg
picture operator AH ( ) can be expressed in terms of a contour-ordered exponential,
1 ( 0 )
d 0 H
AH ( ) = Tc e ),
R
c A( (1.33)

where the contour c goes from 0 = 0 to 0 = and back, see Fig. 1.4.
1.4. EXERCISES 9

0
Figure 1.4: Contour used to the operator AH ( ) in the Heisenberg picture from the corresponding
) in the interaction picture. The horizontal axis denotes imaginary time, not real time.
operator A(

1.4 Exercises
Exercise 1.1: Evolution equation for expectation value (Ehrenfest Theorem)

Show that the time dependence of the expectation value of an observable A satisfies the
differential equation
d
i 
A t= [H, A] ,
dt ~ t

if the operator A has no explicit time dependence.

Exercise 1.2: Free particle, Schr

odinger picture.

For a free particle with mass m, express the expectation values (x )t and (p )t of the com-
ponents of the position and momentum at time t in terms of the corresponding expectation
values at time 0 ( = x, y, z). Use the Schrodinger picture.

Exercise 1.3: Free particle, Heisenberg picture

Derive the Heisenberg picture evolution equations for the position and momentum operators
of a free particle. Use your answer to express the expectation values (x )t and (p )t of
the components of the position and momentum at time t in terms of the corresponding
expectation values at time 0 ( = x, y, z).

Exercise 1.4: Harmonic oscillator

Derive and solve the Heisenberg picture evolution equations for the position and momentum
operators x and p of a one-dimensional harmonic oscillator with mass m and frequency 0 .
10 CHAPTER 1. PRELIMINARIES

Do the same for the corresponding creation and annihilation operators

a = (0 m/2~)1/2 x i(2~0 m)1/2 p

a = (0 m/2~)1/2 x + i(2~0 m)1/2 p. (1.34)

Verify that the energy is time-independent.

Exercise 1.5: Imaginary time

Derive and solve the imaginary-time Heisenberg picture evolution equations for the position
and momentum operators of a one-dimensional harmonic oscillator with mass m and fre-
quency 0 . Do the same for the corresponding creation and annihilation operators, see Eq.
(1.34).
Chapter 2

Green functions

Throughout this course well make use of Green functions. These are nothing but the
expectation value of a product of operators evaluated at different times. There is a number
of important ways in which Green functions are defined and there are important relations
between the different definitions. Quite often, a certain application calls for one Green
function, whereas it is easier to calculate a different Green function for the same system.
We now present the various definitions of Green functions in the general case and the
general relations between the Green functions. In the later chapters we discuss various
physical applications that call for the use of Green functions.
For now, well restrict our attention to Green functions (or correlation functions) de-
fined for two operators A and B,
which do not need to be hermitian. Green functions defined
for more than two operators can be defined in a similar way. In the applications, well en-
counter cases where the operators A and B are fermion or boson creation or annihilation
operators, or displacements of atoms in a lattice, or current or charge densities.
We say that the operators A and B satisfy commutation relations if they describe bosons
or if they describe fermions and they are even functions of fermion creation and annihilation
operators. Well refer to this case as the boson case, and use a sign in the formulas
below. We say that the operators A and B satisfy anticommutation relations if they describe
fermions and they are odd functions of in fermion creation and annihilation operators. Well
call this case the fermion case, and use a + sign in the formulas of the next sections.

2.1 Green functions with real time arguments

The so-called greater and lesser Green functions are defined as

G> 0 0
A;B (t, t ) ihA(t)B(t )i, (2.1)

11
12 CHAPTER 2. GREEN FUNCTIONS

G< 0 0
A;B (t, t ) ihB(t )A(t)i, (2.2)

where the + sign is for fermions and the sign for bosons. Here the brackets h. . .i denote an
appropriately defined equilibrium or non-equilibrium average more on that below. One
further defines retarded and advanced Green functions, which are expectation values of
and B(t
the commutator of A(t) 0 ),

GR 0 0 0
A;B (t, t ) i(t t )h[A(t), B(t )] i,
= (t t0 )[G> 0 < 0
A;B (t, t ) GA;B (t, t )] (2.3)

GA (t, t0 ) i(t0 t)h[A(t), 0 )] i
B(t
A;B
= (t0 t)[G< 0 > 0
A;B (t, t ) GA;B (t, t )]. (2.4)

respectively, where (x) = 1 if x > 0, (x) = 0 if x < 0 and (0) = 1/2. Here, [, ]+ is the
anticommutator, [A, B]
+ = AB +B A.
Note that, by definition, GR is zero for t t0 < 0 and
A 0
G is zero for t t > 0.
One verifies that greater and lesser Green functions satisfy the relation

G> 0 > 0 < 0 < 0

A;B (t, t ) = (GB ,A (t , t)) , GA;B (t, t ) = (GB ,A (t , t)) , (2.5)

from which it follows that

GA 0 R 0
B ;A (t ; t) = GA;B (t; t ) . (2.6)
Also, note that by virtue of their definitions, one has the relation

G> 0 < 0
A;B (t, t ) = (1)GB;A (t , t), (2.7)

from which it follows that

GR 0 A 0
A;B (t; t ) = (1)GB;A (t ; t), (2.8)
where the + sign applies to fermions and the sign to bosons.
In the previous chapter we have seen that formal expressions for Heisenberg picture
operators can be obtained using time-ordered and contour-ordered expressions. For that
reason it is useful to define time-ordered and contour-ordered Green functions. Another
reason why it is useful to consider time-ordered and contour-ordered Green functions is that
there exists more mathematical machinery to calculate them than to calculate the Green
functions defined sofar. A typical calculation thus starts at calculating time-ordered or
contour-ordered Green functions and then uses these results to find physical quantities of
interest.
2.1. GREEN FUNCTIONS WITH REAL TIME ARGUMENTS 13

Time ordered Green functions are defined as

B(t
GA;B (t, t0 ) ihTt A(t) 0 )i
B(t
= i(t t0 )hA(t) 0 )i i(t0 t)hB(t
0 )A(t)i

= (t t0 )G> 0 0 < 0
A;B (t, t ) + (t t)GA;B (t, t ). (2.9)

As before, the symbol Tt represents the time-ordering operator. For fermions, time
ordering is defined with an additional factor 1 for every exchange of operators, hence

B(t 0 ) if t > t0 ,

B(t 0) A(t)
Tt A(t) 0 )A(t)
(2.10)
()B(t if t < t0 .

The anti time-ordered Green function and contour-ordered Green functions are defined
similarly. Recall that for contour-ordered Green functions it is not the physical time argu-
ment, but the location on a contour along the time axis that determines the order of the
operators.
If the averages h. . .i are taken in thermal equilibrium (or, more generally, if they are
taken in a stationary state), the Green functions depend on the time difference t t0 only.
Then, Fourier transforms of these Green functions are defined as
Z
>
GA;B () = dteit G>
A;B (t), (2.11)

Z
<
GA;B () = dteit G<
A;B (t), (2.12)

with similar definitions for the retarded, advanced, and time-ordered Green functions. Since
GR (t t0 ) is zero for t < t0 , its Fourier transform GR () is analytic for Im > 0. Similarly,
the Fourier transform of the advanced Green function, GA (), is analytic for Im < 0. The
inverse Fourier transforms are
Z
> 1
GA;B (t) = deit G>
A;B (), (2.13)
2
Z
< 1
GA;B (t) = deit G<
A;B (). (2.14)
2

Note that the fact that GR () is analytic for Im > 0 implies that GR (t t0 ) = 0 for t < t0 ,
and, similarly, that the fact that GA () is analytic for Im < 0 implies that GR (t t0 ) = 0
for t > t0 .
14 CHAPTER 2. GREEN FUNCTIONS

2.2 Green functions with imaginary time arguments

Green functions for operators with imaginary time arguments are referred to as temperature
Green functions. The temperature Green function GA;B (1 ; 2 ) for the operators A and B
is defined as
1 )B(
GA;B (1 ; 2 ) hT A( 2 )i. (2.15)
Here, the brackets denote a thermal average and the symbol T denotes time-ordering,
1 )B(
T A( 2 ) = (1 2 )A(
1 )B(
2 ) (1)(2 1 )B(
2 )A(
1 ). (2.16)

As discussed previously, the + sign applies if the operators A and B satisfy fermion anticom-
mutation relations, i.e., if they are of odd degree in fermion creation/annihilation operators.
The sign applies if A and B are boson operators or if they are of even degree in fermion
creation/annihilation operators. Note that, again, a factor 1 is added for every exchange
of fermion operators. The definition (2.15) of the temperature Green function is used for the
interval ~/T < 1 2 < ~/T only.
Temperature Green functions are used only for calculations that involve a thermal equi-
librium at temperature T . Hence, making use of the cyclic property of the trace, we find
1
1 H/~
2 H/~
GA;B (1 , 2 ) = (1 2 ) tr e1 H/~ Ae e2 H/~ Be eH/T
Z
1
2 H/~
1 H/~
(2 1 ) tr e2 H/~ Be e1 H/~ Ae eH/T
Z
1
(1 2 )H/~
H/T
= (1 2 ) tr e(1 2 )H/~ Ae Be
Z
1 (1 2 )H/~

(1 2 )H/~
(2 1 ) tr Be Ae eH/T
Z
= GA;B (1 2 , 0), (2.17)

so that G(1 , 2 ) depends on the imaginary time difference 1 2 only. Hence, we can write
G(1 2 ) instead of G(1 , 2 ). For GA;B ( ) with 0 < ~/T we further have

1
H/~
H/T
GA;B ( ) = tr e H/~ Ae Be
Z
1
H/~
= tr eH/T e H/~ Ae B
Z
1
( ~/T )H/~
= tr e( ~/T )H /~ Ae eH/T B
Z
= (1)GA;B ( ~/T ). (2.18)
2.2. GREEN FUNCTIONS WITH IMAGINARY TIME ARGUMENTS 15

Hence, on the interval ~/T < < ~/T , G( ) is periodic with period ~/T and antiperiodic
with the same period for fermions. This property is used to extend the definition of the
temperature Green function G( ) to the entire imaginary time axis
If the operators A and B and the Hamiltonian H are all symmetric, the temperature
Green function is symmetric or antisymmetric in the time argument, for the boson and
fermion cases, respectively,
B,
GA;B ( ) = (1)GA;B ( ) if A, H
symmetric. (2.19)
Using the periodicity of G, one may write G as a Fourier series,
T X in
GA;B ( ) = e GA;B (in ), (2.20)
~ n

with frequencies n = 2nT /~, n integer, for bosons and n = (2n + 1)T /~ for fermions.
The inverse relation is Z ~/T
GA,B (in ) = d ein GA,B ( ). (2.21)
0
The frequencies n are referred to as Matsubara frequencies.
There exists a very useful expression for the temperature Green function in the interaction
picture. Let us recall the definition of the temperature Green function GA;B (1 , 2 ) for the
case 0 < 2 < 1 < ~/T ,
GA;B (1 , 2 ) = hAH (1 )B
H (2 )i

tr eH/T AH (1 )BH (2 )
=
. (2.22)
tr eH/T
We wrote the index H denoting Heisenberg picture operators explicitly, to avoid confusion
with the interaction picture operators to be used shortly.
In the previous chapter, we have seen that the imaginary-time Heisenberg picture op-
erators AH (1 ) and B
H (2 ) can be written in terms of the contour-ordered product of the
corresponding interaction picture operators and an exponential of the perturbation to the

Hamiltonian, see Eq. (1.27). Similarly, nothing that the exponential eH/T is nothing but
the imaginary time evolution operator, Eq. (1.22) implies that it can be written as
R ~/T
1 ( 0 )/T
d 0 H
eH/T = eH0 /T T e 0 . (2.23)
Substituting these results into Eq. (2.22), we find
1 ( 0 )/T
d 0 H 1 )B(
2)
R
tr Tc eH0 /T e c A(
GA;B (1 , 2 ) = R ~/T
1 ( 0 )/T
, (2.24)
tr T eH 0 /T e 0
d 0 H
16 CHAPTER 2. GREEN FUNCTIONS

0 1 1/T 0 1/T
(a) (b)

Figure 2.1: Integration contour for the evaluation of the imaginary time Green function
GA;B (1 , 2 ) using the interaction picture. The contour shown in (a) can be deformed to a
simple line connecting the points = 0 and = ~/T (b).

where the contour c is shown in Fig. 2.1. The contour can be deformed into a straight
integration from 0 to ~/T , see Fig. 2.1, so that we obtain the remarkably simple result
R ~/T
1 ( 0 )/T
d 0 H 1 )B(
2)
tr T eH0 /T e 0 A(
GA;B (1 , 2 ) = R ~/T
1 ( 0 )/T
tr T eH 0 /T e 0 d 0 H
R ~/T
1 ( 0 )/T
d 0 H 1 )B(
2 )i0
hT e 0 A(
= R ~/T
1 ( 0 )/T
, (2.25)
d 0 H
hT e 0 i0

where the brackets h. . .i0 denote an average with respect to the unperturbed Hamiltonian
0 . One easily verifies that this final result does not depend on the assumption 2 > 1 .
H

2.3 Green functions in thermal equilibrium

In thermal equilibrium, all Green functions defined in the previous two sections are related.
This is an enormous help in actual calculations.
In many applications is it difficult, if not impossible, to calculate Green functions ex-
actly. Instead, we have to rely on perturbation theory or other approximation methods.
Whereas physical observables are often expressed in terms of greater and lesser Green func-
tions (for correlation functions) or advandced and retarded Green functions (for response
functions), the theoretical machinery is optimized for the calculation of the time-ordered,
contour-ordered, and imaginary time Green functions. The relations we explored in this
chapter allow one to relate retarded, advanced, and temperature Green functions to the
temperature, time-ordered, and contour-ordered Green functions. Hence, these relations are
a crucial link between what can be calculated easily and what is desired to be calculated.
Before we explain those relations, it is helpful to define a real part and imaginary
part of a Green function. In the time representation, we define the real part <G of the
2.3. GREEN FUNCTIONS IN THERMAL EQUILIBRIUM 17

retarded, advanced, and time-ordered Green functions as

1 R
<GR 0
GA;B (t; t0 ) + GR 0


A;B (t; t ) = B ;A (t ; t) , (2.26)
2
1 A
<GA 0
GA;B (t; t0 ) + GA 0


A;B (t; t ) = B ;A (t ; t) , (2.27)
2
1
<GA;B (t; t0 ) = GA;B (t; t0 ) + GB ;A (t0 ; t) .


(2.28)
2
Similarly, we define the imaginary part =G as

1
=GR 0
GR 0 R 0


A;B (t; t ) = A;B (t; t ) GB ;A (t ; t) , (2.29)
2i
1
=GA 0
GA 0 A 0


A;B (t; t ) = A;B (t; t ) GB ;A (t ; t) , (2.30)
2i
1
=GA;B (t; t0 ) = GA;B (t; t0 ) GB ;A (t0 ; t) .


(2.31)
2i
These real and imaginary parts of Green functions are different from the standard real
and imaginary parts Re G and Im G. In particular, <G and =G are not necessarily real
numbers. The quantities <G and =G are defined with respect to a hermitian conjugation
that consists of ordinary complex conjugation, interchange and hermitian conjugation of the
operators A and B,
and of the time arguments. The assignments < and = are invariant under
this hermitian conjugation and are preserved under Fourier transform of the time variable t,

1
GA;B () + GB ;A () ,


<GA;B () = (2.32)
2
1
GA;B () GB ;A () .


=GA;B () = (2.33)
2i
with similar expressions for the retarded and advanced Green functions.
By virtue of Eq. (2.6), the retarded and advanced Green functions are hermitian conju-
gates, hence

<GR () = <GA (), (2.34)

=GR () = =GA (). (2.35)

Similarly, by employing their definitions and with repeated use of the relation

B(t
hA(t) 0 )i = hB
(t0 )A (t)i, (2.36)
18 CHAPTER 2. GREEN FUNCTIONS

one finds a relation between the real parts of the time-ordered and the retarded or advanced
Green functions,

<GR () = <G(), (2.37)

<GA () = <G(). (2.38)

Shifting the t-integration from the real axis to the line t i~/T , one derives the general
relation Z Z
it A(t)i,

dte hA(t)B(0)i = e ~/T
dteit hB(0) (2.39)

valid in thermal equilibrium. Using Eq. (2.39), together with Eq. (2.36), one finds a relation
for the imaginary parts of the time-ordered and retarded or advanced Green functions. This
relation follows from the fact that the imaginary parts of the retarded, advanced, and time-
ordered Green functions do not involve theta functions. Indeed, for the imaginary part of
the retarded and advanced functions one finds

=GR A
A;B () = =GA;B ()
1
Z
= B(0)
dteit hA(t) A(t)i
B(0)
2
1

Z
= 1e ~/T B(0)i
dteit hA(t)
2
i ~/T

>
= 1e GA;B (), (2.40)
2
whereas for the time-ordered Green function one has
1
Z
=GA;B () = B(0)
dteit hA(t) (1)B(0) A(t)i

2
1

Z
= 1 (1)e ~/T B(0)i
dteit hA(t)
2
i
= 1 ()e~/T G>


A;B (). (2.41)
2
Hence, we find

1 + (1)e~/T
=GR () = =G(), (2.42)
1 (1)e~/T
1 + (1)e~/T
=GA () = =G(), (2.43)
1 (1)e~/T
2.3. GREEN FUNCTIONS IN THERMAL EQUILIBRIUM 19

where the + sign is for fermion operators and the sign for boson operators.
With the help of the Fourier representation of the step function,
Z i 0 t
1 0 e
(t) = d 0 , (2.44)
2i i

where is a positive infinitesimal, and of the second line of Eq. (2.40), one shows that
knowledge of the imaginary part of the retarded Green function (or of the advanced Green
function) is sufficient to calculate the full retarded Green function,
Z
1 1
Z
0
R
GAB () = 0
d 0 B(0)
dtei t hA(t) A(t)i
B(0)
2 i
1 0 =GR 0
A;B ( )
Z
= d 0 . (2.45)
i
Similarly, one finds

1 =GA 0
A;B ( )
Z
GA
A;B () = d 0 . (2.46)
0 + i

The numerator of the fractions in Eqs. (2.45) and (2.46) is known as the spectral den-
sity,
AA;B () = 2=GR () = 2=GA (). (2.47)
The spectral density satisfies the normalization condition
Z
1 B]
i.
dAA;B () = h[A, (2.48)
2

Similarly, we can express the Fourier transforms G> () and G< () of the greater and
lesser Green functions in terms of the spectral density,
iAA;B ()
G>
A;B () = , (2.49)
1 e~/T
iAA;B ()
G<
A;B () = . (2.50)
1 e~/T
Since the spectral density is real (in the sense that A = <A), we conclude that the greater
and lesser Green functions are purely imaginary (G> = i=G> , G< = i=G< ).
The greater and lesser Green functions describe time-dependent correlations of the ob-
servables A and B. In the next section, well see that the retarded and advanced Green
20 CHAPTER 2. GREEN FUNCTIONS

0 1/T

Figure 2.2: Integration contours for derivation of Eq. (2.51).

functions describe the response of the observable A to a perturbation B. The imaginary

part of the response represents the dissipation. For that reason, Eqs. (2.49) and (2.50) are
referred to as the fluctuation-dissipation theorem.
Finally, by shifting integration contours as shown in Fig. 2.2, one can write the Fourier
transform of the temperature Green function for n > 0 as
Z ~/T
GA;B (in ) = )B(0)i
d ein hA(
0
Z
= i B(0)ie
dthA(t) n t

Z0
+i i~/T )B(0)ie
dthA(t n t+i~n /T
0
Z
= i B(0)ie
dthA(t) n t

Z0
+i A(t)ie
dthB(0) n t+i~n /T
0
= GR
A;B (in ). (2.51)

Here, we made use of the fact that exp(in /T ) = (1) and of the fact that the Fourier
transform of the retarded Green function is analytic in the upper half of the complex plane.
In order to find the temperature Green function for negative Matsubara frequencies, one can
make use of the relation
GA;B (in ) = GB ;A (in ) . (2.52)
2.4. GREEN FUNCTIONS OUTSIDE THERMAL EQUILIBRIUM 21

Writing the retarded Green function in terms of the spectral density using Eqs. (2.45) and
(2.47), one thus obtains the general expression

1 AA;B ( 0 )
Z
GA;B (in ) = d 0 0 , (2.53)
2 in
which is valid for positive and for negative Matsubara frequencies.

2.4 Green functions outside thermal equilibrium

While there exists an elaborate and well-established framework to deal with systems in
thermal equilibrium, a description of systems out of equilibrium seems to be a rather hopeless
task. Here we describe a method to deal with systems out of equilibrium if it is known that
the system was in equilibrium at some reference time t0 . After time t0 , a known perturbation
was added to the Hamiltonian that drove the system out of equilibrium.
In order to describe the effect of the perturbation, we write the Hamiltonian as
=H
H 0 + H
1, (2.54)

where the time dependence of the perturbation H 1 is such that it is switched on slowly after
a time t0 long before the times we are interested in. Since we know the Hamiltonian for
all times, we can we can express any operators at times t in terms of the corresponding
operators at time t0 . In the interaction picture, this relation is given by Eq. (1.27), which
we repeat here
0 0
AH (t) = Tc e(i/~) c dt H1 (t ) A(t).

R

Here we choose c to be the Keldysh contour, see Fig. 1.2, which starts at the reference time
t0 runs to infinite time, and then returns to time t0 .
What does this imply for Green functions? Expressing both A(t) and B(t
0 ) in terms of
the corresponding operators without the perturbation, one has for, e.g., the greater Green
function
i c dt1 H1 (t1 ) 0 )i0 ,
R
G> 0
A;B (t, t ) = ihTc e A(t)B(t (2.55)
where the brackets h. . .i0 indicate a thermal average at time t0 . The contour c now consists
of two trips from t0 to infinity and back: one from the time evolution of A and one from
the time evolution of B, see Fig. 2.3. The time t corresponding to A will be assigned to
the second part of the contour, whereas the time t0 corresponding to B will be assigned
to the first part of the contour. However, matters simplify considerably if we look at the
contour-ordered Green function. In that case, the times t and t0 represent positions on the
22 CHAPTER 2. GREEN FUNCTIONS

t0 t

t
Figure 2.3: Contours used to calculate the greater Green functions G > .

Keldysh contour. For a contour-ordered Green function, the contour time that appears on
the left in the Green function is always later (in the contour sense). One can deform the
contour of Fig. 2.3 such that the contour goes directly from t0 to t, without the intermediate
excursion to the reference time t0 without one of the excursions to infinity. What remains is
the standard Keldysh contour of Fig. 1.2, so that one has the simple result
B(t 0 )i0 ,
R
GA;B (t, t0 ) = ihTc e(i/~) c dt1 H1 (t1 )
A(t) (2.56)

where c is the Keldysh contour.

Notice that if both both arguments are on the upper branch of the Keldysh contour, the
contour-ordered Green function is nothing but the time-ordered Green function,

G(t, t0 ) = ihTt AH (t)B

H (t0 )i, t, t0 upper branch. (2.57)

(We dropped the index A; B 00 of the Green functions as no confusion is possible here.)
Similarly, if both arguments are on the lower branch, G(t, t0 ) is equal to the anti-time-
ordered Green function,

G(t, t0 ) = i(t0 t)hAH (t)B H (t0 )i i(t t0 )hB

H (t0 )AH (t)i
ihTt AH (t)B
H (t0 )i, t, t0 lower branch, (2.58)

whereas, if t is on the upper branch and t0 is on the lower branch, or vice versa, one has

G(t, t0 ) = G< (t, t0 ) if t upper branch, t0 lower branch,

G(t, t0 ) = G> (t, t0 ) if t lower branch, t0 upper branch. (2.59)

We use this property to represent the Green function G(t, t0 ) as a 2 2 matrix, where the
matrix index indicates what branch of the Keldysh contour is referred to (1 for upper branch
and 2 for lower branch)

G11 (t, t0 ) G12 (t, t0 ) ihTt AH (t)B

H (t0 )i G< (t, t0 )
   
= H (t0 )i . (2.60)
G21 (t, t0 ) G22 (t, t0 ) G> (t, t0 ) ihTt AH (t)B
2.5. LINEAR RESPONSE 23

In the matrix notation, the arguments t and t0 refer to physical times, not contour positions.
In the literature, one usually uses a different representation of the matrix Green function
(2.60),
G11 (t, t0 ) G12 (t, t0 )
   
0 1 1 1 1 1
G(t, t ) = . (2.61)
2 1 1 G21 (t, t0 ) G22 (t, t0 ) 1 1
Note that the transformation in Eq. (2.61) is invertible, although it is not merely a shift of
basis. Using Eq. (2.60) you quickly verify that
 R
G (t, t0 ) GK (t, t0 )

0
G(t, t ) = , (2.62)
0 GA (t, t0 )

where GR (t, t0 ) and GA (t, t0 ) are the standard retarded and advanced Green functions (but
now calculated outside equilibrium) and

GK (t, t0 ) = G> (t, t0 ) + G< (t, t0 ) (2.63)

is the so-called Keldysh Green function. Note that the special structure of the matrix (2.62)
is preserved under matrix multiplication.
In equilibrium, or in a steady state situation, the matrix Green function G(t, t0 ) depends
on the time difference t t0 only. In that case, one can look at the Fourier transform
G(). The Fourier transforms of the retarded and advanced Green functions were discussed
previously. In thermal equilibrium, the Fourier transform of the Keldysh Green function
satisfies the relation
e~/T (1) e~/T (1)
GK R A
A;B () = (GA;B () GA;B ()) = iA A;B () . (2.64)
e~/T + (1) e~/T + (1)

2.5 Linear Response

A special application of a non-equilibrium calculation is if we are interested in the response
1 only. This situation is called linear response. Expanding
to first order in the perturbation H
Eq. (1.27) to first order in the perturbation H 1 , one has

t
i
Z
AH (t) = H
dt0 (A(t) 1 (t0 ) H
1 (t0 )A(t)),
(2.65)
~ t0

and H
where A(t) 1 (t) are the operators in the interaction picture: their time dependence is
given by the unperturbed Hamiltonian H 0.
24 CHAPTER 2. GREEN FUNCTIONS

The thermal average is performed for the observables at time t0 . Since the Hamiltonian
0 , this simply corresponds to a thermal average with respect
for all earlier times is given by H
to the Hamiltonian H 0,

t
0 = i
Z

hA(t)i hAi
h[A(t), 1 (t0 )] i0 .
H (2.66)
~ t0

1 (t) = 0 for t < t0 , this can be rewritten as

Since H

i t
Z

hA(t)i hAi0 =
h[A(t), H 1 (t0 )] i0
~
1 0 R
Z
= dt GA;H 1 (t, t0 ), (2.67)
~

Here GR is the retarded Green function, and we used the fact that GR (t, t0 ) = 0 for t0 > t.
Quite often, we are interested in the response to a perturbation that is known in the
frequency domain,
1
Z

H1 = 1 ().
dei(+i)t H (2.68)
2

Here is a positive infinitesimal and the factor exp(t) has been added to ensure that H 1 0
if t . (In practice one does not need to require that H 1 = 0 for times smaller than a

reference time t0 ; in most cases it is sufficient if H1 0 fast enough if t .) Then Eq.
(2.67) gives, after Fourier transform,
Z

hA()i
hA()i 0 =
hAi
dteit (hAi 0)

= GR
A;H1 ( + i). (2.69)

Equation (2.69) is known as the Kubo formula. It shows that the linear response to
a perturbation H 1 , which is a nonequilibrium quantity, can be calculated from a retarded
equilibrium Green function. For this reason, retarded Green functions (and advanced Green
functions) are often referred to as response functions.
Although the Kubo formula involves an integration over the real time t0 , for actual
calculations, we can still use the imaginary time formalism and calculate the temperature

Green function GA;H1 ( ). The response hA()i is then found by Fourier transform of G to
the Matsubara frequency domain, followed by analytical continuation in .
2.6. HARMONIC OSCILLATOR 25

2.6 Harmonic oscillator

Well now illustrate the various definitions and the relations between the Green functions by
calculating all Green functions for the one-dimensional harmonic oscillator.
The Hamiltonian for a one-dimensional quantum-mechanical harmonic oscillator with
mass m and frequency 0 is
H = 1 p2 + 1 m 2 x2 . (2.70)
0
2m 2
The momentum and position operators satisfy canonical commutation relations,

[
p, p] = [ p, x] = i~.
x, x] = 0, [ (2.71)

We now demonstrate two methods to calculate various harmonic oscillator Green func-
tions if the harmonic oscillator is in thermal equilibrium at temperature T . First, we calculate
all Green functions explicitly using the fact that we can diagonalize the harmonic oscillator
Hamiltonian. Of course, our explicit solution will be found to obey the general relations be-
tween the different Green functions derived in the previous chapter. Then we use a different
method, the so-called equation of motion method to find the temperature Green function
Gx;x .
The harmonic oscillator can be diagonalized by switching to creation and annihilation
operators, r r r r
m0 1 m0 1
a
=x + i p , a
=x i
p , (2.72)
2~ 2m0 ~ 2~ 2m0 ~
so that  
1
H = ~0 aa + . (2.73)
2
The creation and annihilation operators satisfy the commutation relation

[
a, a a , a
] = [ ] = 0, [ ] = 1.
a, a (2.74)

The explicit calculation of the Green functions uses the known Heisenberg time-evolution of
the creation and annihilation operators,

a(t) = ei0 t a (t) = ei0 t a

(0), a (0), (2.75)

the commutation relations (2.74), and the equal-time averages

1
h a (0)
a(0)i = h
a(0) a (0)i = 0, h
a (0)
a(0)i = . (2.76)
e~0 /T 1
26 CHAPTER 2. GREEN FUNCTIONS

With the help of these results, one easily constructs averages involving the creation and
annihilation operators at unequal times,

h
a(t)
a(0)i = 0
ei0 t
h a (0)i =
a(t) ,
1 e~0 /T
ei0 t
h
a (t)a(0)i = ~0 /T ,
e 1
a (t)
h a (0)i = 0. (2.77)

Knowing these expectation values, it is straightforward to calculate all different Green func-
tions involving the creation and annihilation operators,

iei0 t
G>
a,a (t) = ,
1 e~0 /T
iei0 t
G<
a,a (t) =
e~0 /T 1
Ga,a (t) = i(t)ei0 t ,
R

GA
a,a (t) = i(t)e
i0 t
,
iei0 t sign (t)
Ga,a (t) = . (2.78)
1 e~0 sign(t)/T
Their Fourier transforms are
e~0 /T
G>
a,a () = 2i( 0 ) ,
e~0 /T 1
1
G<
a,a () = 2i( 0 ) ~0 /T ,
e 1
1
GR
a,a () = ,
0 i
1
GA
a,a () = ,
0 + i
1 1
Ga,a () = , (2.79)
(0 i)(1 e 0 ) (0 + i)(1 e~0 /T )
~ /T

where is a positive infinitesimal. From this we conclude that the spectral density is

Aa,a = 2( 0 ). (2.80)
2.6. HARMONIC OSCILLATOR 27

For the calculation of the imaginary time Green functions we need the Heisenberg time
evolution for imaginary time, which is found from Eq. (2.75) by substituting t i ,
( ) = e0 a
a ( ) = e0 a
(0), a (0). (2.81)
From this, one finds
e0 e 0
h a (0)i =
a( ) , h
a
( )
a (0)i = , (2.82)
1 e0 ~/T e0 ~/T 1
hence, for ~/T < < ~/T ,
e0 e0 0 ~/T
Ga;a ( ) = ( ) ( ) . (2.83)
1 e0 ~/T 1 e0 ~/T
You verify that the temperature Green function is periodic in , G( + ~/T ) = G( ). The
Fourier transform is
1
Ga;a (in ) = , (2.84)
0 in
where n = 2T n/~, n integer, is a bosonic Matsubara frequency.
We can also calculate Green functions involving the position x at different times. Upon
inverting Eq. (2.72) one finds
G>
x;x (t) = ih
x(t)
x(0)i
i~
= h a (0) + a
a(t) (t)
a(0)i, (2.85)
2m0
plus terms that give zero after thermal averaging. Expressions for the other Green functions
are similar. Using the averages calculated above, we then find
i~ ei0 t ei0 t
 
>
Gx;x (t) = + ,
2m0 e0 ~/T 1 1 e0 ~/T
i~ ei0 t ei0 t
 
<
Gx;x (t) = + ,
2m0 e0 ~/T 1 1 e0 ~/T
~
GRx;x (t) = (t) sin(0 t),
m0
~
GAx;x (t) = (t) sin(0 t),
m0
i~ ei0 |t| ei0 |t|
 
Gx;x (t) = + ,
2m0 e0 ~/T 1 1 e0 ~/T
e 0 e0
 
~
Gx;x ( ) = + . (2.86)
m0 e0 ~/T 1 1 e0 ~/T
28 CHAPTER 2. GREEN FUNCTIONS

The Fourier transforms are

i~ ( + 0 ) ( 0 )
 
>
Gx;x () = +
m0 e0 ~/T 1 1 e0 ~/T
i~ ( 0 )
 
< ( + 0 )
Gx;x () = +
m0 e0 ~/T 1 1 e0 ~/T
 
R ~ 1 1
Gx;x () = + ,
2m0 0 i 0 + + i
 
A ~ 1 1
Gx;x () = + ,
2m0 0 + i 0 + i
 
~ 1 1
Gx;x () =
m ( 2 (0 i)2 )(1 e0 ~/T ) ( 2 (0 + i)2 )(e0 ~/T 1)
~
Gx;x (in ) = 2
(2.87)
m(0 + n2 )
In this case, the spectral density is
~
Ax;x () = (( 0 ) ( + 0)). (2.88)
m0
Note that both the Green functions Ga;a and the Green functions Gx;x we just calculated
satisfy all the relations derived in the previous chapter.
We now illustrate a different method to calculate the Green functions. Well look at
the example of the temperature Green function Gx;x ( ), although the same method can also
be applied to time-ordered real-time Green functions and Green functions involving other
operators. The starting point in this method are the equations of motion for the Heisenberg
operators x and p,
1 i
x( ) = [H, x( )] = p( ), (2.89)
~ m
1
p( ) = [H, p( )] = mi02 x( ). (2.90)
~
Then, using the definition of the temperature Green function, we find
i
Gx;x ( ) = ( )h x(0) x(0)
x( ) x( )i + [( )hp( )
x(0)i + ( )h
x(0)
p( )i](2.91)
.
m
The first term arises from the derivative of the theta-function and is zero upon setting = 0
in the thermal average. In the second term we recognize the temperature Green function
Gp;x ( ), hence
i
Gx;x ( ) = Gp;x ( ). (2.92)
m
2.6. HARMONIC OSCILLATOR 29

Taking one more derivative to , we find

i
2 Gx;x ( ) = ( )h
p( ) p( )i 02 [( )h
x(0) x(0) x( )
x(0)i + ( )h
x(0)
x( )i]
m
~
= ( ) + 02 Gx;x ( ). (2.93)
m
This equation is best solved by Fourier transform,
~
n2 Gx;x (in ) = + 02 Gx;x (in ), (2.94)
m
which reproduces Eq. (2.87) above. The advantage of the equation of motion approach is
that one does not have to diagonalize the Hamiltonian in order to use it. However, more
often than not the set of equations generated by this approach does not close, and one has
to use a truncation scheme of some sort.
30 CHAPTER 2. GREEN FUNCTIONS

2.7 Exercises
Exercise 2.1: Lehmann representation

Explicit representations for Green functions can be obtained using the set of many-particle
eigenstates {|ni} of the Hamiltonian H, as a basis set. For example, the greater Green
function in the canonical ensemble can be written as
i 0 H/T
G> 0
A;B (t, t ) = tr A(t)B(t )e
Z
i X En /T B(t 0 )|ni,
= e hn|A(t) (2.95)
Z n

where Z = n eEn /T is the canonical partition function. (In the grand canonical ensemble,
P
similar expressions are obtained.) Inserting a complete basis set and using
X
|n0 ihn0 | = 1,
n0

we write Eq. (2.95) as

i X En /T +i(tt0 )(En En0 ) 0 ihn0 |B|ni.

G> 0
A;B (t, t ) = e hn|A|n (2.96)
Z n,n0

Performing the Fourier transform to time, one finds

2i X En /T 0 ihn0 |B|ni.

G>A;B () = e (En En0 + )hn|A|n (2.97)
Z n,n0

With the Lehmann representation, the general relations we derived in Sec. 2.3 can be verified
explicitly.

(a) Verify that G>A;B is purely imaginary (with respect to hermitian conjugation, as defined
in Sec. 2.3).
(b) Derive Lehmann representations of the retarded, advanced, lesser, time-ordered, and
temperature Green functions. Use the frequency representation.
(c) Derive the Lehmann representation of the spectral density AA;B ().
(d) Verify the relations between the different types of Green functions that were derived
in Sec. 2.3 using the Lehmann representations derived in (b) and (c).
2.7. EXERCISES 31

Exercise 2.2: Spectral density

In this exercise we consider the spectral density AA;B () for the case A = c, B = c that the
operators A and B are fermion or boson annihilation and creation operators, respectively.
In this case, the spectral density Ac;c () can be viewed as the energy resolution of a particle
created by the creation operator c . To show that this is a plausible idea, you are asked to
prove some general relations for the spectral density Ac;c (). Consider the cases of fermions
and bosons separately.
(a) The spectral density Ac;c () is normalized,
Z
1
dAc;c () = 1. (2.98)
2

(b) For fermions, the spectral density Ac;c () 0. For bosons, Ac;c () 0 if > 0 and
Ac;c () 0 if < 0. (Hint: use the Lehmann representation of the spectral density.)
= hc ci is
(c) The average occupation n
Z
1 1
n
= dAc;c () ~/T . (2.99)
2 e 1

(d) Now consider a Hamiltonian that is quadratic in creation and annihilation operators,
X
H= n cn cn .
n

For this Hamiltonian, calculate the spectral density A(n, ) Acn ;cn ().

Exercise 2.3: Harmonic oscillator

For the one-dimensional harmonic oscillator with mass m and frequency 0 , calculate the
retarded, advanced, greater, lesser, time-ordered, and temperature Green functions Gx;p and
Gp;p . List your results both in frequency and time representations.
32 CHAPTER 2. GREEN FUNCTIONS
Chapter 3

The Fermi Gas

3.1 Fermi Gas

Let us now consider a gas of non-interacting fermions. Since we deal with a system of
many fermions, we use second quantization language. In second quantization notation, the
Hamiltonian is written in terms of creation and annihilation operators (r) and (r) of
an electron at position r and with spin ,
Z
0 (r)H0 (r),
X
H = dr (3.1)
, 0

where H is the first-quantization Hamiltonian. For free fermions, one has H = p2 /2m with
p = i~r , but many of the results we derive below also apply in the more general case when
H contains a scalar potential, a magnetic field, or spin-orbit coupling. The operators (r)
and (r) satisfy anticommutation relations,

[ (r), 0 (r0 )]+ = [ (r), 0 (r0 )]+ = 0, [ (r), 0 (r0 )]+ = 0 (r r0 ). (3.2)

In the Heisenberg picture, the operators and are time dependent, and their time-
dependence is given by
i iX
t (r, t) = [H, (r, t)] = H0 0 (r, t). (3.3)
~ ~ 0

Note that the time-evolution of the Heisenberg picture annihilation operator is formally
identical to that of a single-particle wavefunction in the Schrodinger picture.

33
34 CHAPTER 3. THE FERMI GAS

Below we calculate the Green functions for the pair of operators (r) and 0 (r0 ). We
will denote these Green functions as G,0 (r, r0 ; t). Using the equation of motion for the
t), one can derive an equation of motion for the retarded Green function,
operator (r,

t GR 0 0 0
, 0 (r, r ; t) = i(t)h (r, t) 0 (r , 0) + 0 (r , 0) (r, t)i

i(t)t h (r, t) 0 (r0 , 0) + 0 (r0 , 0) (r, t)i

iX
= i(t)(r r0 )0 H00 GR 0
00 , 0 (r, r ; t). (3.4)
~ 00

where H0 is the first-quantization form of the single-particle Hamiltonian operating on the

first argument of the Green function, see Eq. (3.3) above. The differential equation (3.6)
is solved with the boundary condition GR (t) = 0 for t < 0. A similar calculation shows
that the advanced Green function GA satisfies the same equation, but with the boundary
condition GA (t) = 0 for t > 0. The Keldysh Green function GK satisfies the equation

iX
t GK 0
, 0 (r, r ; t) = H00 GK 0
00 , 0 (r, r ; t) (3.5)
~ 00

There are no boundary conditions for this equation. However, in equilibrium, GK can be
expressed in terms of GR and GA , see Eq. (2.64). This information is sufficient to make the
solution of Eq. (3.5) unique, even in non-equilibrium situtations.
Performing a Fourier transform, one finds
1X
(~00 H00 )GR 0 0
00 , 0 (r, r ; ) = 0 (r r ), (3.6)
~ 00

with similar equations for GA and GK .

A formal solution of Eq. (3.6) can be obtained in in terms of the eigenvalues and
eigenfunctions , (r) of H,
X , (r),0 (r0 )
GR 0
, 0 (r, r ; ) = ~ , (3.7)

~ + i
X , (r),0 (r0 )
GA 0
, 0 (r, r ; ) = ~ , (3.8)

~ i

where is a positive infinitesimal. One verifies that both the retarded and the advanced
Green functions are analytical in the upper and lower half planes of the complex plane,
3.2. IDEAL FERMI GAS 35

respectively. The spectral density A0 (r, r0 ; ) is easily calculated from Eqs. (3.7) or (3.8),
X
A0 (r, r0 ; ) = 2 , (r),0 (r0 )( /~). (3.9)

Notice that our calculation of the retarded and advanced Green functions did not use the
requirement of thermal equilibrium. In thermal equilibrium, the other Green functions
(greater, lesser, temperature, time-ordered) follow directly from the spectral density cal-
culated here. Outside thermal equilibrium, calculation of the other Green functions requires
knowledge of how the non-equilibrium state has been obtained.

3.2 Ideal Fermi Gas

In the absence of an impurity potential, the Hamiltonian is

~2 2
H0 = 0 (3.10)
2m r
For fermions confined to a volume V with periodic boundary conditions, the eigenfunctions
of the Hamiltonian (3.10) are plane waves,

1
k (r) = eikr . (3.11)
V

(We dropped the spin index here.) Hence the free fermion Green function reads
0 0
R 0 1 X eik(rr ) A 0 1 X eik(rr )
G (r, r ; ) = , G (r, r ; ) = , (3.12)
V k + i k /~ V k i k /~

with k = ~2 k 2 /2m. Replacing the k-summation by an integration, one finds

0
R 0 m eik|rr |
G (r, r ; ) = , (3.13)
2~ |r r0 |
0
A 0 m eik|rr |
G (r, r ; ) = . (3.14)
2~ |r r0 |

For the retarded Green function, k is the solution of + i k = 0 such that Im k > 0,
whereas for the advanced Green function, k is the solution of i k = 0 with Im k < 0.
36 CHAPTER 3. THE FERMI GAS

It is interesting to note that the same result can be obtained without using the quadratic
dispersion relation k = ~2 k 2 /2m. Hereto one defines k as above and linearizes the spectrum
around ~
k0 = ~ + ~v(|k0 | k), (3.15)
where v is the velocity. Then, performing the Fourier transform, one finds
Z 2 Z 1 Z ik 0 |rr0 | cos
R 0 1 02 0 e
G (r, r ; ) = d d cos k dk
(2)3 0 1 0 i v(k 0 k)
Z ik 0 |rr0 | 0 0
1 0 0e eik |rr |
= k dk
(2)2 i|r r0 | 0 i v(k 0 k)
Z 0 0
k ei(k+x/v)|rr | ei(k+x/v)|rr |
dx
4 2 i|r r0 |v i x
ik|rr0 |
m e
= , (3.16)
2~ |r r0 |

and a similar result for the advanced Green function. For non-infinitesimal , the final result
0
needs to be multiplied by e|rr |/v .
Because the ideal Fermi gas is translationally invariant, the Green functions depend on
the position difference r r0 only. Fourier transforming with respect to r r0 , one has
Z
R 0
Gk () = drGR (r r0 ; )eik(rr )
1
= , (3.17)
+ i k /~
1
GA
k () = . (3.18)
i k /~

Alternatively, one may consider the Green function corresponding to the operators k and k 0
that annihilate and create a fermion in an eigenstate with wavevector k and k0 , respectively.

In terms of the operators (r) and (r), one has

1 1
Z Z

k =
dr(r)e ikr
, k = dr (r)eikr .
V V
Denoting this Green function with Gk,k0 , one has

1
Z
R 0 0
Gk,k0 () = drdr0 GR (r, r0 ; )eikr+ik r
V
3.3. BOLTZMANN EQUATION 37

k,k0
= , (3.19)
+ i k /~
k,k0
GA
k,k0 () = . (3.20)
i k /~

3.3 Boltzmann equation

If the potential U (r) varies slowly on the scale of the Fermi wavelength, one expects that
the electrons will follow trajectories that are described by classical mechanics. In such a
situation, one would expect that the Boltzmann equation will be a valid description of the
many-electron system.
The Boltzmann equation describes the time-evolution of the distribution function fk (r, t)
that is the occupation of electronic states with wavevector k and position r. Without col-
lisions between electrons and without scattering from quantum impurities or phonons, the
Boltzmann equation reads

t + vk r ~1 R U k fk (r, t) = 0,


(3.21)

where vk = ~1 k k is the electrons velocity. The use of a distribution that depends on an

electrons position as well as its momentum is allowed only if the spatial variation of f is
slow on the inverse of the scale for the k-dependence of f . A faster spatial variation would
violate the Heisenberg uncertainty relations. Hence, we expect that the Boltzmann equation
works for slowly varying potentials.
Well now show how the distribution function f and the Boltzmann equation can be ob-
tained in the Green function language. For this, well use the real-time Keldysh formulation.
So far we have dealt with Green functions that depend on two spatial coordinates r and
r or on two wavenumbers k and k0 . For a semiclassical description, it is useful to use a
0

mixed representation in which the Green function G is chosen to depend on sum and
difference coordinates, followed by a Fourier transform to the difference coordinate,
Z Z
G(R, T ; k, ) = dr dteikr+it G(R + r/2, T + t/2; R r/2, T t/2).
(3.22)

Well refer to R and T as center coordinate and center time, and to k and as mo-
mentum and frequency. The reason why one chooses to Fourier transform to the differences
of spatial and temporal coordinates is the G(r, t; r0 , t0 ) is a fast oscillating function of r r0
and t t0 , whereas it is a slowly varying function of the center coordinates R = (r + r0 )/2
and T = (t + t0 )/2. In fact, for the ideal Fermi gas, G(r, t; r0 , t0 ) does not depend on R and
38 CHAPTER 3. THE FERMI GAS

T at all. After Fourier transform, the oscillating dependence on the differences r r0 and
t t0 results in a much less singular dependence on k and .
In order to write the first-quantization operators H and t in the mixed representation,
we first these operators in a form in which they depend on two spatial coordinates r and r0
and two time coordinates,
~2 2
 
0 0
H(r, t; r , t ) + U (r) (r r0 )(t t0 ), (t )(r, t; r0 , t0 ) t (r r0 )(t t0 ).
2m r
(3.23)
Here U is the potential and we included the chemical potential in the definition of the
Hamiltonian. In this notation, the operator action corresponds to a convolution with respect
to the primed variables. Transforming to the mixed representation, we then find that the
first-quantization Hamiltonian H and the time derivative t take a particularly simple form,
H = k + U (R, T ), t = i. (3.24)
The matrix Green function G satisfies the differential equation
 
i
t + H G(r, t; r0 , t0 ) = i(t t0 )(r r0 )1, (3.25)
~
where 1 is the 22 unit matrix, cf. Sec. 2.4. The equation has to be solved with the boundary
condition GR = 0 for t < t0 , GA = 0 for t > t0 . In equilibrium, GK is given by Eq. (2.64). In
Eq. (3.25), the Green function G can be viewed a first-quantization operators. In operator
language, Eq. (3.25) reads  
i
t + H G = iI, (3.26)
~
where I is the identity operator and I = I1. Implicitly, we already used the operator picture
in Sec. 3.1, when we calculated the Green functions of the Fermi gas using the equation of
motion method and in Sec. 4.1. Considering the derivative to t0 , we find the related operator
identity  
i
G t + H = iI. (3.27)
~
Instead of working with Eqs. (3.26) and (3.27), one prefers to work with the sum and differ-
ence equations,
  
i
t + H , G = 0 (3.28)
~
  
i
t + H , G = 2iI, (3.29)
~ +
3.3. BOLTZMANN EQUATION 39

where [, ] and [, ]+ denote commutator and anticommutator, respectively.

Mathematically, the way Green functions act as first-quantization operators, as well as the
way first-quantization operators act on single-particle Green functions, is a convolution.
The main disadvantage of the mixed representation (3.22) is that convolutions become rather
awkward. Using Eq. (3.22) and its inverse to express the Wigner representation of the
convolution (or operator product) A1 A2 in terms of the Wigner representations of the two
factors A1 and A2 , one finds

i
[A1 A2 ](R, T ; k, ) = e 2 D A1 (R, T ; k, )A2 (R, T ; k, ),
1 2 1 2 2 1 2 1
D = + , (3.30)
1 R 2 k 1 T 2 2 R 1 k 2 T 1

where 1 refers to a derivative with respect to an argument of A1 and 2 refers to a derivative

40 CHAPTER 3. THE FERMI GAS

with respect to an argument of A2 .1

We now make use of the fact that the potential U is a slowly varying function of its
arguments R and T . Then we expect that the Green function GK will also be a slowly
varying function of R and T . Calculating the operator products according to the rule (3.30)
we can expand the exponential and truncate the expansion after first order. One then finds
that a commutator becomes equal to a Poisson bracket,

i[A1 , A2 ] = [A1 , A2 ]Poisson (3.31)

 
1 2 1 2 2 1 2 1
= + A1 A2 .
1 R 2 k 1 T 2 2 R 1 k 2 T 1

Similarly, truncating the expansion of the exponential after the first order, one finds that
1
One may prove Eq. (3.30) as follows. Writing A = A1 A2 , we have
Z Z Z Z
ikr+it 0
A(R, T ; k, ) = dr dte dr dt0 A1 (R + r/2, T + t/2; r0 , t0 )A2 (r0 , t0 ; R r/2, T t/2).

Now shift variables to r = r1 + r2 , r0 = R + (r2 r1 )/2, t = t1 + t2 , t0 = T + (t2 t1 )/2. This variable

transformation has unit jacobian, hence
Z Z
A(R, T ; k, ) = dr1 dr2 dt1 dt2 eikr1 +it1 ikr2 +it2

A1 (R + (r1 + r2 )/2, T + (t1 + t2 )/2; R + (r2 r1 )/2, T + (t2 t1 )/2)

A2 (R + (r2 r1 )/2, T + (t2 t1 )/2; R (r1 + r2 )/2, T (t1 + t2 )/2).

If it werent for the fact that the argument of A2 contained r1 and t1 , the integrations over r1 and t1 could
be done and would give us the mixed representation Green function A1 (R + r2 /2, T + t2 /2; k, ). Similarly,
without the appearance of r2 and t2 in the arguments of A1 , the integrations over r2 and t2 would simply
give A2 (R r1 /2, T t1 /2; k, ). We can formally remedy this situation by writing

A1 (R + (r1 + r2 )/2, T + (t1 + t2 )/2; R + (r2 r1 )/2, T + (t2 t1 )/2)

= e(1/2)r2 R +(1/2)t2 T A1 (R + r1 /2, T + t1 /2; R r1 /2, T t1 /2),
A2 (R + (r2 r1 )/2, T + (t2 t1 )/2; R (r1 + r2 )/2, T (t1 + t2 )/2)
= e(1/2)r1 R (1/2)t1 T A2 (R + r2 /2, T + t2 /2; R r2 /2, T t2 /2).

Now the integrations can be done, with the formal result

A(R, T ; k, ) = A1 (R, T ; k (i/2)2,R , + (i/2)2,T )A2 (R, T ; k + (i/2)1,R , (i/2)1,T )

= e(i/2)D A1 (R, T ; k, )A2 (R, T ; k, ).

Here the derivatives 2,R and 2,T act on the factor A2 , whereas the derivatives 1,R and 1,T act on the
factor A1 .
3.3. BOLTZMANN EQUATION 41

the derivatives cancel when calculating an anticommutator,

[A1 , A2 ]+ = 2A1 A2 . (3.32)

This approximation is known as the gradient expansion.

In the gradient approximation, one quickly finds the retarded and advanced Green func-
tions,

~
GR (R, T ; k, ) = , (3.33)
~ + i k + U (R, T )
~
GA (R, T ; k, ) = , (3.34)
~ i k + U (R, T )

where is a positive infinitesimal. (Addition of is necessary to enforce the appropriate

boundary conditions for advanced and retarded Green functions.) From this result, we
conclude that the spectral density A is a delta function,

A(R, T ; k, ) = 2~(~ k + U (R, T )). (3.35)

Calculation of the Keldysh Green function from Eq. (3.29) shows that GK (R, T ; k, )
is proportional to 2~(~ k + U (R, T )), but fails to determine the prefactor. In
order to determine the prefactor, which determines the distribution function, we consider the
Keldysh component of Eq. (3.28. In the gradient approximation, the differential equation
(3.28) for the Keldysh component reads

GK GK H GK H GK
+ vk + = 0, (3.36)
T R R k T
where vk = k H. Integrating this equation over , and using Eq. (3.24) for H, we find

(T + vk R R Uext k ) fk (R, T ) = 0 (3.37)

where
1 1
Z
fk (R, T ) = dGK (R, T ; , k). (3.38)
2 4i
Equation (3.37) is nothing but the Boltzmann equation (3.21). The function f is the dis-
tribution function: it specifies the density of electrons with wavevector k. Indeed, since the
definition of f contains an integration over frequency , fk is related to an equal-time Green
function.
42 CHAPTER 3. THE FERMI GAS

Using the relationship G< = (GK + iA)/2, together with Eq. (3.35), one quickly finds
expressions for the charge density and current density in terms of the distribution function
f . For the charge density, one has

h (r, t) (r, t)i

X
e (r, t) = e

X
= ie G<
, (r, t; r, t)

ie X
Z
= dG<, (r, t; k, )
2V
k,
e X
= fk, (r, t), (3.39)
V
k,

where we reinserted the summation over spin. Similarly, for the current density, one finds
e X
je (r, t) = vk fk, (r, t). (3.40)
V
k,

Also note that, by Eqs. (2.64) and (3.35), one has fk = [1+exp((k )/T )]1 in equilibrium,
independent of R and T .
Our derivation of the Boltzmann equation was for a particle moving in a smooth potential.
It does not hold for impurities with a short-range potential. Such impurities scatter particles
into different quantum states. Such quantum scattering events give rise to a collision term
in the Boltzmann equation. We return to this issue in the next chapters, when we consider
the effect of impurities in more detail.

3.4 Exercises
Exercise 3.1: Fermi gas in one or two dimensions

Calculate the retarded Green function for the ideal Fermi gas in one and in two dimensions
in the coordinate representation.

Exercise 3.2: Non-interacting bosons

Repeat the analysis of Sec. 3.1 for a gas of non-interacting bosons. In particular, show that
the retarded and advanced Green functions for the Bose gas are identical to the retarded and
3.4. EXERCISES 43

advanced Green functions of the Fermi gas. (Of course, Bosons will have integer spin, whereas
Fermions will have half-odd-integer spin.) Will the other Green functions be identical as well?

Exercise 3.3: Vector potential and Gauge invariance

In Sec. 3.3 we derived the Boltzmann equation in the presence of a scalar potential. In this
exercise you are asked to see what happens if one has both a scalar potential and a vector
potential that both vary slowly in time and space.
The first-quantization Hamiltonian now has the form
1  e 2
H= i~r A(r, t) + e(r, t), (3.41)
2m c
where we wrote the scalar electromagnetic potential instead of the potential U . The vector
potential A and the scalar potential are related to the magnetic and electric fields as
1
E = t A r , B = r A. (3.42)
c
The vector potential A and the scalar potential are not uniquely determined by the electric
and magnetic fields. Gauge invariance requires that all physical observables be invariant
under the transformation

A(r, t) A(r, t) + r (r, t), (r, t) (r, t) t (r, t), (3.43)

where is an arbitrary function of r and t.

(a) Show that in the mixed representation and in the gradient approximation, one has
1  e 2
H= ~k A(R, T ) + e(R, T ). (3.44)
2m c

(b) Argue that the wavevector k has to change according to the rule
e
kk+ R (R, T ) (3.45)
~c
if the gauge transformation (3.43) is applied.

The canonical momentum ~k is an argument of the distribution function f in the Boltzmann

Equation (3.21). However, as you have just showed, ~k is not a gauge invariant quantity.
44 CHAPTER 3. THE FERMI GAS

One may make the momentum argument of f gauge invariant by switching to the kinetic
momentum ~kkin ,
k = kkin + (e/~c)A(R, T )). (3.46)
A differential equation for the Keldysh Green function GK (R, T ; kkin , ) in the gradient
approximation can be obtained from Eq. (3.36). However, one must take care in dealing
with the partial derivatives in that equation: The partial derivatives to R and T are meant
to be taken at constant canonical momentum ~k, not at constant kinetic momentum ~kkin .

(c) Show that GK (R, T ; kkin , ) satisfies the differential equation

GK GK e GK A GK
   
e A
~ +v v + v
T R c kkin R c R kkin
K
A GK
   
H e A G H e
+ + v = 0, (3.47)
R c T kkin T c T

where now the partial derivatives to R, T , and kkin have their standard meaning.

(d) Integrate over frequency to find a kinetic equation for the distribution function fkkin (R, T ).
Express your final answer in terms of the electric field E and the magnetic field B.

(e) One may want to further adapt Eq. (3.47) by replacing the frequency ~ by ~
e(R, T ). Show that this substitution gives a manifestly gauge invariant evolution
equation for the Green function.

Exercise 3.4: Quasiclassical approximation

If the Green function G has an important -dependence, as is, e.g., the case for electron-
phonon scattering, one does not want to define a distribution function by integrating the
Green function over the frequency . In such circumstances, G may still be a sharply peaked
function of momentum. When that is the case, an equation can be derived for the so-called
quasiclassical Green function

i
Z
0
g(R, n, t, t ) = dG(R, k, t, t0 ), (3.48)

where = k and n indicates the direction of k.
3.4. EXERCISES 45

We now derive an equation for the quasiclassical Green function using the gradient ex-
pansion for the spatial coordinates, but not for the temporal coordinates. As in Ex. 3.3
we want to use the kinetic momentum, not the canonical momentum, as the argument of
the Green function in the mixed representation. If we use the kinetic momentum, n repre-
sents the direction of the electrons kinetic momentum, not its canonical momentum. Thus,
instead of Eq. (3.22), we define
Z
0 0
G(R, kkin ; t, t ) = drei[kkin +(e/2~c)(A(t,R)+A(t ,R))]r G(R + r/2, t; R r/2, t0 ).
(3.49)

(a) Show that the Green function G of Eq. (3.49) satisfies the differential equation

0 = (~t1 + ei(R, t1 ) + ~t2 ei(R, t2 )) G

 
ei
+ v R (A(R, t1 ) A(R, t2 )) G (3.50)
c
e G 1 G
+ (v (B(R, t1 ) + B(R, t2 ))) (E(R, t1 ) + E(R, t2 )) .
2c kkin 2 kkin

(b) Integrate this equation over = kkin to find an equation for the quasiclassical Green
function g.

(c) Argue that the advanced and retarded components of the quasiclassical Green function
are particularly simple:
g R = 1, g A = 1. (3.51)

(d) If the potentials and A are a slow function of time, one may Fourier transform
to t1 t2 and use the gradient expansion for the temporal coordinates. What is
the corresponding kinetic equation? Comment on the subtle differences between the
equation you just obtained and the Boltzmann equation.
46 CHAPTER 3. THE FERMI GAS
Chapter 4

Scattering from a single impurity

In this chapter, we consider non-interacting electrons in the presence of a external potential.

This is another problem for which an exact solution is possible. In the first section of this
chapter, we recall the standard quantum-mechanical solution of the problem. In the second
section, we show how the same results can be obtained using the many-electron Green
function formalism and diagrammatic perturbation theory.
In first quantization notation, the Hamiltonian we consider is

~2 2
H= + U (r), (4.1)
2m r

where U (r) is the impurity potential. Here and in the remainder of this chapter we drop
reference to the spin index . In second quantization notation, the Hamiltonian is written
in terms of creation and annihilation operators (r) and (r)
of an electron at position r

2 Z Z
=~
H dr (r)r2 (r)
+ dr (r)U (r)(r).
(4.2)
2m

4.1 Scattering theory and Friedel Sum Rule

In the presence of an impurity, which is described by the potential U (r), eigenfunctions of
the Hamiltonian (4.1) are no longer plane waves. In this chapter, we consider the case of
a single impurity, and assume that the potential U has a finite range. Then we can try to
find an eigenfunction |k i of H with incoming wave boundary condition corresponding to

47
48 CHAPTER 4. SCATTERING FROM A SINGLE IMPURITY

the unperturbed state |ki far away from the impurity. This solution can be written as1
1
|k i = |ki + G0R (k /~)U |k i (4.3)
~
1 1
= |ki + G0R (k /~)U |ki + 2 G0R (k /~)U G0R (k /~)U |ki + . . . , (4.4)
~ ~
where U is the first-quantization operator corresponding to the potential U (r) and G0 is
the free particle Green function, seen as an operator on the wavefunction |k i. The easiest
way to see that Eq. (4.4) provides the correct wavefunction is to verify that | k i solves the
Schrodinger equation and has the correct boundary conditions. The series (4.4) is known
as the Born series. Truncating the series after the nth order in U yields the nth Born
approximation. Defining the T -matrix as
1 1
T ( + ) = U + 2 U G0R ()U + . . .
~ ~
= U [~ G0R ()U ]1
1 1
= U + U G0R ()T ( + ), (4.5)
~ ~
where + = + i, being a positive infinitesimal, we can rewrite this as
X Tk0 k (k + i)
|k i = |ki + |k 0 i, (4.6)
(k k0 )/~ + i
k0

with Tk0 k () = hk0 |T ()|ki. The T -matrix is related to the retarded Green function GR of
the total Hamiltonian H as

GR () = G0R () + G0R ()T ( + )G0R (). (4.7)

Note that in this notation, where we have a summation over discrete set of wavevectors, the
T -matrix is of order 1/V . Using Eq. (3.13) one then finds that the asymptotic form of the
wavefunction of the scattered wave (4.6) is
0
ikr 0 eik r mV
k (r) = e + f (k , k) + O(1/r 2 ), f (k, k0 ) = Tkk0 (k /~ + i), (4.8)
r 2~
where the wavevector k0 is defined as k
r , r being the unit vector in the direction of r.
1
You can find a good discussion of stationary scattering theory in section 37 of Quantum Mechanics, by
L. I. Schiff, McGraw-Hill (1968).
4.1. SCATTERING THEORY AND FRIEDEL SUM RULE 49

With this result we can now write expressions for two useful quantities from collision
theory: the differential cross section d/d, where d denotes an element of solid angle,
which is
d m2 V 2
= |f (k0 , k)|2 = 2 2 |Tk0 ,k (k /~ + i)|2 , (4.9)
d 4 ~
and the transition rate

Wk0 k = 2~|Tk0 k (k + i)|2 (k0 k ). (4.10)

The transition rate gives the probability per unit time that a particle is scattered out of
plane wave state k into plane wave state k0 . To lowest order in the impurity potential U ,
Eq. (4.10) is nothing but the Fermi golden rule. The total rate for scattering out of the state
|ki is found by summation of Eq. (4.10) over all outgoing momenta,

1 X
= 2~ |Tk0 k (k + i)|2 (k0 k ). (4.11)
k k0

Taking the hermitian conjugate of Eq. (4.5) and substituting U = U = ~T T G0A U

in the second term of Eq. (4.5), one finds
1 1
T = U + T G0R T T G0A U G0R T . (4.12)
~ ~
The first and third term on the right hand side are hermitian, so that

=T = T =G0R T . (4.13)

Using Eq. (3.20), this result can be written as

1 X
Tkk0 ( + ) Tk0 k ( + ) = Tk1 k ( + )Tk1 k0 ( + )( k1 /~).


(4.14)
2i k 1

This result is known as the generalized optical theorem. The standard optical theorem
is obtained setting k0 = k,
X
Im Tkk ( + ) = |Tk1 k ( + )|2 ( k1 /~). (4.15)
k1

The optical theorem relates the forward scattering amplitude to the intensity of the total
scattered wave.
50 CHAPTER 4. SCATTERING FROM A SINGLE IMPURITY

The density of levels dN ()/d~ is expressed in terms of the Green function as

dN () X 1
= (~ ) = Im tr GR (). (4.16)
d~
~

The impurity density of states (dN/d~)imp is defined as the difference between the density
of states with and without impurity. Using the general relation

tr G() = ln det G(), (4.17)

one can show that  
dN 1 ()
= , (4.18)
d~ imp ~
where () = arg det T () is the so-called total scattering phase shift. Zero eigenvalues
of T , which correspond to states that do not scatter at all, should be left out from the
calculation of the total scattering phase shift. Equation (4.18) is known as the Friedel sum
rule. At zero temperature, as a result of the introduction of the impurity, the number of
electrons in a many-electron system is changed by the amount2
Z F  
dN (F )
Nimp = ~ d = , (4.19)
d~ imp

where we took () = 0 and F is the Fermi energy.

In order to make things more explicit, let us now study the case U (r) = u(r) of a delta-
function potential at the origin in detail. Fourier transforming to the basis of plane-wave
states and using second-quantization notation, the potential U is written as
u X
U= k . (4.20)
V k,q k+q

Then the equation for the T -matrix becomes

u 1
Tk,k0 ( + i) = , (4.21)
~V 1 (u/~)G0R ()
2
Strictly speaking, the derivation of Eq. (4.19) applies to potential scattering only. One may, however,
extend the Friedel sum rule to the case of scattering from a dynamical impurity. See, e.g., chapter 5 of The
Kondo problem to heavy fermions, by A. C. Hewson.
4.1. SCATTERING THEORY AND FRIEDEL SUM RULE 51

where we abbreviated
1 X 0R 1 X 1
G0R () = Gkk () = . (4.22)
V k V k + i k0 /~

(A similar definition can be made for G0A ().) Note that the T -matrix is isotropic: For a
delta-function scatterer, scattering is equally probable in all directions. Since T is isotropic,
only one of the eigenvalues of T is nonzero. The simplest way to calculate the corresponding
phase shift is to note that T ( + i) = T ( i) , so that

T ( + i) 1 (u/~)G0 ( i)
e2i() = = . (4.23)
T ( i) 1 (u/~)G0 ( + i)

The real part of G0 ( + i) depends weakly on . Formally, for a quadratic dispersion

relation k = ~2 k 2 /2m and a delta-function scatterer, Re G0 ( + i) is ultraviolet divergent.
The high-energy part of the energy integration can be cut off by introducing a finite width
of the conduction band, or by noting that for high energies the approximation of a delta
function scatterer breaks down: high energies correspond to small length scales, and on
sufficiently short length scales the spatial structure of the potential must be resolved. The
imaginary part of G0 ( + i) is related to the density of states 0 (per unit volume) at the
energy ~ in the absence of the impurity,
k2
Im G0 ( + i) = ~0 () = , (4.24)
2v
Here v = ~1 /k is the velocity at wavevector k.
With the help of Eq. (4.24), we can express T in terms of and 0 ,
1
Tkk0 ( i) = ei() sin ()
~0 V
2v
= 2 ei() sin (). (4.25)
k V
Hence, in this case we find for the differential cross section
d m2 v 2
= 4 2 sin2 (), (4.26)
d k ~
and for the total scattering cross section (the integral of d/d over the full solid angle)
4m2 v 2
= sin2 (). (4.27)
k 4 ~2
52 CHAPTER 4. SCATTERING FROM A SINGLE IMPURITY

4.2 Diagrammatic perturbation theory

In this section, well use the Green function formalism to describe the electron gas in the
presence of an impurity. The main goal of this section is to introduce the formal machinery
of diagrammatic perturbation theory. We first describe the calculation using the imaginary-
time formalism. Then the same calculation is described using the real-time formalism.

4.2.1 imaginary time formalism

We are interested in the electron density n(r) of a gas of noninteracting electrons in the
presence of a potential U (r). The electron density can be expressed in terms of the many-
electron temperature Green function, using
n(r) = h (r)(r)i
= lim G(r, r; ). (4.28)
0

Here G is the temperature Green function in the presence of the impurity potential. We now
calculate G using diagrammatic perturbation theory. It is advantageous to change to the
momentum representation,
1 = 1
Z
k eikr ,
X
ikr
k = dr (r)e , (r)
V 1/2 V 1/2 k

so that the temperature Green function in momentum representation reads

Gk;k0 ( ) = hT [k ( )k 0 (0)]i
1
Z
0 0
= drdr0 G(r, r0 ; )ei(krk r ) . (4.29)
V
The Hamiltonian is separated into an unperturbed part H0 and the perturbation H1 ,
H = H0 + H1 , (4.30)
with
(k )k k ,
X
H0 = (4.31)
k

Uq k+q k ,
XX
H1 = (4.32)
k q

where Uq is the Fourier transform of the scattering potential,

1
Z
Uq = drU (r)eiqr . (4.33)
V
4.2. DIAGRAMMATIC PERTURBATION THEORY 53

Here, we have included the chemical potential into the unperturbed Hamiltonian H0 .
It is not clear a priori that the perturbation H1 is small. Still, in the method of dia-
grammatic perturbation theory, it is assumed that an expansion in H1 is possible, and that
resummation of that expansion gives the correct results. Since we are going to expand in
H1 , we need the temperature Green function for the unperturbed Hamiltonian H0 ,

0 e(k ) /~ e(k ) /~
Gk;k 0 ( ) = k,k0 ( ) + k;k 0 ( ) . (4.34)
1 + e(k )/T 1 + e(k )/T
The Fourier transform of the temperature Green function is
0 k,k0
Gk;k 0 (in ) = . (4.35)
in (k )/~
0 0
Since Gk;k 0 is nonzero only if k = k , we drop the second index if no confusion is possible.

In order to prepare ourselves for a perturbative calculation in H 1 , we switch to the

interaction picture. Using Eq. (2.25), we then write the temperature Green function as
R ~/T
1 ( 0 )/~
hT e 0
d 0 H
k ( )k 0 (0)i0
Gk;k0 ( ) = R ~/T
1 ( 0 )/~
, (4.36)
d 0 H
hT e 0 i0
where the brackets h. . .i0 indicate an thermal average with respect to the unperturbed Hamil-
tonian H0 only. In order to avoid misunderstandings about the order of the creation and
annihilation operators in the perturbation H 1 , we increase the imaginary time argument of
the creation operators by a positive infinitesimal ,
1 ( ) = Uq k+q ( + )k ( ).
XX
H (4.37)
k q

The next step is to expand the evolution operator in powers of H1 and calculate the
thermal average with respect to the Hamiltonian H 0 . The calculation of the thermal average
is done with the help of Wicks theorem. In the language of the general formulation of
Sec. 2, Wicks theorem allows one to calculate the thermal average of an imaginary time-
ordered product of the form hT A1 (1 ) . . . A2n (2n )i0 , where all operators Ai , i = 1, . . . , 2n
are linear in fermion creation and annihilation operators and the average is taken with
respect to a Hamiltonian H0 that is quadratic in fermion creation and annihilation operators.
Clearly, every order in the perturbation expansion of Eq. (4.36) satisfies these criteria. Wicks
theorem states that
 n X
1
hT A1 (1 ) . . . A2n (2n )i0 = (1)P GA0 P (1) ,AP (2) (P (1) , P (2) ) . . .
2 P
GA0 P (2n1) ,AP (2n) (P (2n1) , P (2n) ), (4.38)
54 CHAPTER 4. SCATTERING FROM A SINGLE IMPURITY

where the sum is over all permutations P of the numbers 1, 2, . . . , 2n and (1)P is the sign
of the permutation. In our case, only Green functions with one creation operator and one
annihilation operator are nonzero. That simplifies the set of allowed permutations, and we
can write Wicks theorem as

hT k1 (1 ) . . . kn (n )k 0n (n0 ) . . . k 0 (10 )i0

1
X
n
= (1) (1) Gk1 ,k0 (1 , P (1) ) . . . Gk0n ,k0
P 0 0
(n , P0 (n) ), (4.39)
P (1) P (n)
P

where now P is a permutation of the numbers 1, 2, . . . , n only. For a proof of Wicks

theorem, see, e.g., the book Methods of quantum field theory in statistical physics, by A. A.
Abrikosov, L. P. Gorkov, and I. E. Dzyaloshinski, Dover (1963). Wicks theorem applies to
time-ordered products only.
We have to average numerator and denominator in Eq. (4.36) separately. To zeroth order
in the perturbation H 1 , the numerator gives G 0 0 ( ), whereas the denominator gives 1, hence
k,k
we find
Gk,k0 ( ) = Gk0 ( )k,k0 + higher order terms (4.40)
Let us now calculate the first order corrections. The first-order correction to the numerator
of Eq. (4.36) is
~/T
1
Z X
d 0
 0 0 0 0 0 0

Uq Gk,k 00 +q ( )Gk00 ,k0 ( 0) Gk00 ,k00 +q ()Gk,k0 ( )
~ 0 q,k00
~/T ~/T
1 1
Z Z X
0
= d Ukk0 Gk0 ( 0
)Gk00 ( 0 ) Gk0 ( )k,k0 d 0 U0 Gk000 (),
~ 0 ~ 0 k00

whereas the first-order correction to the denominator is

1 ~/T 0 X 1 ~/T 0 X 0
Z Z
0
d Uq Gk00 ,k00 +q () = d U0 Gk00 ()
~ 0 00
~ 0 00
q,k k

Dividing numerator and denominator and keeping terms to first order in U only, we find that
the first-order correction to the numerator cancels against the second term of the first-order
0
correction to the denominator. Using this, together with the fact that Gk,k 0 is nonzero only

if k = k0 , we find the simple result

~/T
1
Z
Gk;k0 ( ) = Gk0 ( )k,k + d1 Ukk0 Gk0 ( 1 )Gk00 (1 ) + . . . , (4.41)
~ 0
4.2. DIAGRAMMATIC PERTURBATION THEORY 55

numerator
0st order k=k
1 k=k
numerator
k k
1st order 1
U kk k U0

denominator
1
1st order
k U0
Figure 4.1: Diagrammatic representation of the numerator and denominator of Eq. (4.36) up to
0 (, 0 ), whereas
first order in U . Single arrows correspond to the unperturbed Green function G k,k 0

the dashed lines with the stars correspond to the impurity potential U q .

where the dots indicate terms of order two and higher.

The number of terms contributing to the Green function in higher orders in H 1 increases
very rapidly. In order to handle the higher-order contributions it is useful to use diagrams
to represent the terms. The diagrams that represent the first order calculation are shown in
Fig. 4.1. The solid arrows represent the Green functions G 0 of the unperturbed Hamiltonian
H 0 , whereas the dashed lines with the star represent the impurity potential U .
For the terms contributing to the numerator in Eq. (4.36), we discriminate two types of
contributions: disconnected diagrams (the right diagram) and connected diagrams (the left
diagram in Fig. 4.1). It was the disconnected diagram that canceled against the first-order
correction to the denominator in Eq. (4.36). In fact, one can easily convince oneself all
disconnected diagrams in the perturbative expansion of the numerator cancel against the
diagrammatic expansion of the numerator. Thus, for the calculation of G we can ignore the
denominator if we keep connected diagrams only.
Now we can write down the higher-order contributions to the Green function,
~/T
1
Z
Gk;k0 ( ) = Gk0 ( )k,k + d1 Gk0 ( 1 )Ukk0 Gk00 (1 ) (4.42)
~ 0
~/T ~/T
1
Z Z X
+ 2 d1 d2 Gk0 ( 1 )Ukk1 Gk01 (1 2 )Uk1 k0 Gk00 (2 ) + . . . .
~ 0 0 k1
56 CHAPTER 4. SCATTERING FROM A SINGLE IMPURITY

1 1 2
= + k k + k k1 k + ...
k k k=k

Figure 4.2: Diagrammatic representation of Eq. (4.42). The double arrow corresponds to the full
Green function Gk,k0 (, 0 ). The momentum argument of the impurity potential is the difference in
electron momenta at the scattering vertex.

1
= + k k1 k
k k k=k

Figure 4.3: Diagrammatic representation of the Dyson equation for impurity scattering, Eq. (4.43).

Diagrammatically this result may be represented as in Fig. 4.2. The double arrow represents
the full Green function, while the single arrows correspond to the Green functions of the free
electron Hamiltonian H0 . As before, the dashed lines with the stars correspond to scattering
from the impurity potential U . The series can be written as a self-consistent equation for
Gk,k0 ( ),
1 X 1/T
Z
0
Gk;k0 ( ) = Gk ( )k,k0 + d1 Gk0 ( 1 )Ukk1 Gk1 ;k0 (1 ). (4.43)
~ k 0
1

This equation is called the Dyson equation. It is represented diagrammatically in Fig. 4.3.
Equation (4.43) is an integral equation for the Green function Gk;k0 ( ): it contains both
a convolution with respect to the imaginary time variable 1 and a summation over the
intermediate momentum k1 . The convolution can be handled by Fourier transformation,
1X 0
Gk;k0 (in ) = Gk0 (in )k,k0 + G (in )Ukk1 Gk1 ;k0 (in ), (4.44)
~ k k
1

but the summation over the momentum k1 is more complicated. It can not be carried out
exactly for an arbitrary impurity potential. As a simple example, we return to the case of
a point scatterer, U (r) = u(r). In this case, Ukk1 = u/V does not depend on k and k1 ,
see Eq. (4.20). Summing Eq. (4.44) over k, we find
! !
X
0 u X 0 X
Gk;k0 (in ) = Gk0 ;k0 (in ) + Gk (in ) Gk;k0 (in ) , (4.45)
k
~V k k
4.2. DIAGRAMMATIC PERTURBATION THEORY 57

n n
n = n + k k1 k
k k k=k

Figure 4.4: Diagrammatic representation of the Matsubara frequency representation of the Dyson
equation for impurity scattering, Eq. (4.44). Notice that the Matsubara frequency n is conserved
at every scattering event.

from which we derive

X Gk00 (in )
Gk;k0 (in ) = , (4.46)
1 (u/~V ) k Gk0 (in )
P
k

and, hence,
u 0 1
Gk;k0 (in ) = Gk0 (in )k,k0 + Gk (in ) Gk00 (in ), (4.47)
~V 1 (u/~) G 0 (in )

where we abbreviated
1 X 0
G 0 (in ) = G (in ). (4.48)
V k k;k

What remains is to perform the inverse Fourier transform,

T X in
Gk;k0 ( ) = e Gk;k0 (in ). (4.49)
~ n

This is done with the help of the following trick. The summation in Eq. (4.49) is written as
an integration over the contour C1 , see Fig. 4.5,

1
Z
Gk;k0 ( ) = dzez [tanh(~z/2T ) 1] Gk;k0 (in z). (4.50)
4i C1

We are interested in the case of negative .3 Then we make use of the fact that G is analytic
for all frequencies away from the real and imaginary axes, so that we may deform the contour
C1 into the contour C2 , see Fig. 4.5. Note that there is no contribution from the segments
3
If we would have wanted to calculate the temperature Green function for positive , we would have used
Eq. (4.50) with a factor [tanh(~z/2T ) + 1] instead of [tanh(~z/2T ) 1].
58 CHAPTER 4. SCATTERING FROM A SINGLE IMPURITY

C1 C1

C2

C2

Figure 4.5: Integration contours for the derivation of Eq. (4.50) and (4.51).

near infinity by virtue of the factor [tanh(~z/2T ) 1] for Re z > 0 and by virtue of the factor
ez for Re z < 0. We thus obtain
Z
1 X
Gk;k0 ( ) = () de [tanh(~/2T ) 1]Gk;k0 (in i),
4i
Z
e(k ) /~ u X 1 e

1
= ( )/T
d
1+e k ~V 2i 1+e ~/T (k )/~
1 1
, (4.51)
1 (u/~) G 0 ( ) ~ (k0 )/~

where = i and is a positive infinitesimal. The first term in Eq. (4.51) is the
temperature Green function in the absence of the potential, cf. Eq. (4.34). The second term
is the correction that arises as a result of scattering from the impurity at r = 0.
We first use this result to calculate the change in the total number of electrons due to
the presence of the impurity,
X
0


Nimp = lim Gk;k ( ) Gk;k ( )
0
k
u X 1 X
 
1 1 1
Z
= d 2
~V 2i k 1+e ~/T ( (k )/~) 1 (u/~)G 0 ( )
X 1 Z
 
1 h i
= d 0
log 1 (u/~)G ( ) . (4.52)

2i 1 + e~/T
4.2. DIAGRAMMATIC PERTURBATION THEORY 59

This result can be rewritten in terms of the T matrix and the scattering phase shifts. Using
the relation (4.23) between G, the T -matrix, and the scattering phase shift, we recover the
Friedel sum rule,  
1 1
Z
Nimp = d 0 (~ + ). (4.53)
1 + e~/T
At zero temperature, Eq. (4.53) simplifies to Eq. (4.19) above.
In order to find how the density change depends on the distance r to the impurity site,
we have to change variables from the momenta k and k0 to the positions r and r0 . Hereto, we
start again from the second term in Eq. (4.51) and note that the k-dependence is through the
factors 1/( + (k )/~) only. The corresponding Fourier transform has been considered
in Sec. 4.1 and the result for the asymptotic behavior as r is
X  1  Z m2 v sin 0 (~ + )e i[0 (~+)+2(kF +/vF )r]
nimp (r) = lim d e .
0

2i 2r 2 k 2 ~2 1 + e~/T

In order to perform the -integration, we expand the phase shift 0 (~ + ) around = 0

and keep terms that contribute to leading order in 1/r only. The result is
 
sin 0 () 1
Z
nimp (r) = d cos[0 () + 2(kF + /vF )r]. (4.54)
4 2 r 3 1 + e~/T
At zero temperature, this simplifies to Eq. (4.79). These density oscillations are known as
Friedel oscillations. At zero temperature, Friedel oscillations fall off algebraically, propor-
tional to 1/r 3 . At finite temperatures, the main cause of decay are the thermal fluctuations
of the phase shift 2r/vF , which are important for T & ~vF /r.
The final results (4.53) and (4.54) are to be multiplied by two if spin degeneracy is taken
into account.

4.2.2 real-time formalism

We now present the same calculation using Green functions with real time arguments. The
real-time formalism makes use of contour-ordered Green functions, where the integration
contour is the Keldysh contour. The electron density can be calculated as
n(r, t) = h (r)(r)i
= iG< (r, t; r, t). (4.55)
Alternatively, we can use the anticommutation relations of the fermion creation and annihila-
tion operators, to write G< (r, t; r0 , t) = G> (r, t; r0 , t)+i(rr0 ). Hence, using GK = G> +G< ,
we find
i 1
n(r, t) = GK (r, t; r, t) + lim (r r0 ). (4.56)
2 2 r0 r
60 CHAPTER 4. SCATTERING FROM A SINGLE IMPURITY

If we are interested in the change of the density due to the presence of the impurity, we
dont care about the delta function contribution. Hence, it will be our goal to calculate the
Keldysh Green function GK (r, t; r, t).
As before, we like to calculate n(r) in perturbation theory using Wicks theorem. How-
ever, Wicks theorem applies to time-ordered or contour-ordered Green functions only. Hence,
although we are interested in the Keldysh Green function only, well have to calculate the
full contour-ordered Green function G(r, t; r0 , t0 ) where t and t0 are located on the Keldysh
contour.
We now arrange the perturbation theory in such a way that the Hamiltonian H 1 corre-
sponding to the impurity potential is switched on slowly after the reference time t0 . For a
time t on the Keldysh contour, we write
Z
H1 (t) = dr (r, t + )U (r, t)(r,
t), (4.57)

where the time dependence of U ensures the switching on of the impurity potential. The
infinitesimal is added to the argument of the creation operator to guarantee that the
creation operator appears left of the annihilation operator in a contour ordered product: the
contour time t1 + is an infinitesimal amount later (in the contour sense) than t1 . With
this convention, the thermal average at time t0 will be done with respect to the unperturbed
Hamiltonian H 0 only. Of course, at the end of the calculation we will take the limit t0 .
Note that in writing Eq. (4.57) we have kept the coordinate representation. The calcu-
lation outlined below can be done using the momentum representation as well.
We start from Eq. (2.56),
1 (t1 )/~ t) (r0 , t0 )i0 ,
R
G(r, t; r0 , t0 ) = ihTc ei c
dt1 H
(r, (4.58)
and expand the exponential. For real-time Green functions, Wicks theorem reads
1 , t1 ) . . . (r
hTc (r n , tn ) (r0 , t0 ) . . . (r0 , t0 )i0
n n 1 1
X
n
= i P 0 0 0
(1) G (r1 , t1 ; rP (1) , tP (1) ) . . . G0 (rn , tn ; r0P (n) , t0P (n) ). (4.59)
P

1 one finds
Up to first order in H
1
Z Z
0 0 0 0 0
G(r, t; r , t ) = G (r, t; r , t ) + dt1 dr1 U (r1 , t1 )[G0 (r, t; r1 , t1 )G0 (r1 , t1 ; r0 , t0 )
~ c
G0 (r, t; r0 , t0 )G0 (r1 , t1 ; r1 , t1 + )] + . . . , (4.60)
where the dots represent terms of order two and higher. Neither the potential U (r1 , t1 ) nor
the Green function G0 (r1 , t1 ; r1 , t1 +) depends depend on whether t1 is on the upper or lower
4.2. DIAGRAMMATIC PERTURBATION THEORY 61

tr tr
0th order

tr t1r1 tr tr tr
1st order
t1r1

Figure 4.6: Diagrammatic representation of Eq. (4.61). The solid arrows represent the
unperturbed contour-ordered Green function G0 . The dashed lines and the stars represent
the impurity potential U .

branch of the Keldysh contour. Hence, upon integration over t1 the second term between
brackets vanishes and one finds
1
Z Z
0 0 0 0 0
G(r, t; r , t ) = G (r, t; r , t ) + dt1 dr1 U (r1 , t1 )G0 (r, t; r1 , t1 )G0 (r1 , t1 ; r0 , t0 ) + . . .
~ c
(4.61)

Diagrammatically, the two first order terms can be represented as in Fig. 4.6. Upon
integration over t1 , the disconnected rightmost diagram vanishes. This property is preserved
in higher orders of perturbation theory: all disconnected diagrams vanish upon performing
the intermediate integration over the Keldysh contour. Hence, only connected diagrams con-
tribute to the perturbation expansion of G, see Fig. 4.7. Recognizing G in the perturbative
series, one arrives at a self-consistent equation for G,
1
Z Z
0 0 0 0 0
G(r, t; r , t ) = G (r, t; r , t ) + dt1 dr1 U (r1 , t1 )G0 (r, t; r1 , t1 )G(r1 , t1 ; r0 , t0 ).(4.62)
~ c
The self-consistent equation is shown in the second line of Fig. 4.7.
Solving the integral equation (4.62) is not entirely straightforward because the interme-
diate integration over t1 is over the Keldysh contour, not over the real time axis. We write
the contour integration as
Z Z Z
dt . . . dt[ upper branch ] dt[ lower branch].
c t0 t0

If we write the contour-ordered Green function G(t, t0 ) as Gij (t, t0 ), with i = 1 or 2 (j = 1 or

2), if t (t0 ) is on the upper or lower branch of the Keldysh contour, respectively, the contour
62 CHAPTER 4. SCATTERING FROM A SINGLE IMPURITY

rt rt rt rt rt r1 t1 rt rt r1 t1 r2 t2 rt
= + +

rt rt rt r1 t1 rt
= +

Figure 4.7: Diagrammatic representation of the contour-ordered Green function G. The

double arrow represents the contour-ordered Green function. The single arrows represent
the unperturbed contour-ordered Green function G0 .

integration in Eq. (4.62) amounts to matrix multiplication at each impurity vertex, combined
with insertion of a minus sign at the lower branch argument. Hence, we have
Gij (r, t; r0 , t0 ) = G0ij (r, t; r0 , t0 ) (4.63)
1
Z Z X
+ dr1 dt1 G0ik (r, t; r1 , t1 )U (r1 , t1 )z,k Gkj (r1 , t1 , r0 , t0 ),
~ k

where z,1 = 1 and z,2 = 1. Hence, in this matrix notation, every impurity vertex carries
a Pauli matrix z . Upon transformation to the matrix representation of Eq. (2.61), z maps
to the unit matrix, so that one has the simple result
1
Z Z
0
0 0 0 0
G(r, t; r , t ) = G (r, t; r , t ) + dr1 dt1 G0 (r, t; r1 , t1 )U (r1 , t1 )G(r1 , t1 , r0 , t0 ). (4.64)
~
The solution of this equation is found by Fourier transform. As before, we choose the
impurity potential to be of delta-function form. We switch on the potential by setting
U (r, t) = u(r)et , (4.65)
where is a positive infinitesimal. Then, upon Fourier transforming Eq. (4.64) to time and
space coordinates, one finds
!
u X
Gk,k0 () = G0k,k ()k,k0 + G0 () Gk1 ,k0 () . (4.66)
~V k,k k 1

Summing over k, we find

" #1
X u X 0
Gk,k0 () = 1 G () G0k0 (), (4.67)
k
~V k k,k
4.2. DIAGRAMMATIC PERTURBATION THEORY 63

where the inverse means matrix inversion. Hence, upon resubstitution into Eq. (4.66),
" #1
u 0 u X 0
Gk,k0 () = G0k ()k,k0 + G () 1 Gk () G0k0 (), (4.68)
~V k ~V
k

The matrix Green function G has a special matrix structure: the lower left element of the
matrix is zero, see Eq. (2.61). This special matrix structure is preserved under multiplication
and matrix inversion, which makes the matrix inversion in Eq. (4.68) rather straightforward.
Abbreviating
1 X 0
G0 () = Gk (), (4.69)
V
k

we then find that the Keldysh component of G reads

u G0K 0A
k ()Gk0 () u2 G0R 0K 0A
k ()G ()Gk0 ()
GK 0K
k,k0 () = Gk0 () + +
V ~ uG0A () V (~ uG0R ())(~ uG0A ())
u G0R 0K
k ()Gk0 ()
+
V (~ uG0R ())
" #
0R 0R 0A 0A
u Gk ()Gk0 () Gk ()Gk0 ()
= G0K
k0 () + tanh(~/2T ) , (4.70)
V ~ uG0R ()) ~ uG0A ())

where, in the last equality, we used Eq. (2.64) for the Keldysh Green function G0K .4 Using
Eqs. (3.17) and (3.18) for G0R and G0A it is easy to verify that one recovers the same result
as in the imaginary time formalism upon Fourier transform to time and space.
Both the imaginary-time and Keldysh formalism can be used for equilibrium calculations.
In spite of their simplicity, the calculations in this section have shown the advantages and
disadvantages of the imaginary-time and Keldysh formalisms: The imaginary-time formalism
requires a tedious analytical continuation in order to find physically meaningful results,
whereas the Keldysh formalism needs matrix operations. What formalism you prefer to use
is a matter of taste. For nonequilibrium calculations, the situation is different, however.
Nonequilibrium calculations can be done using the Keldysh formalism only.

4
The same result can be obtained in a simpler way if one solves Eq. (4.68) for the retarded and advanced
components of G and then applies Eq. (2.64) to the Keldysh Green function GK .
64 CHAPTER 4. SCATTERING FROM A SINGLE IMPURITY

4.3 Exercises
Exercise 4.1: Friedel oscillations

For an impurity with a general potential U (r), calculate the shift in the electronic density
to first order in U . Compare your result to Eq. (4.54).

Exercise 4.2: Scattering from a single impurity in one dimension

In this exercise, we consider scattering from a delta-function impurity in one dimension. The
impurity has potential U (x) = u(x) and is located at the origin.
It is our aim to calculate the full Green function G(x, x0 ; in ) in coordinate representation.
In order to calculate G(x, x0 ; in ), one may either use the Green function in momentum space
and Fourier transform back to coordinate space, as we did in Sec. 4.2, or derive the Dyson
equation in coordinate representation. Following the latter method, you can verify that one
finds
u
G(x, x0 ; in ) = G 0 (x, x0 ; in ) + G 0 (x, 0; in ) 0
G 0 (0, x0 ; in ).
~ G (0, 0; in)u
(a) Calculate the temperature Green function in coordinate representation.
(b) Use your answer to (a) to calculate the retarded Green function. If x0 < 0, show that
it can be written as
0
GR (x, x0 ; ) = tG0R (x, x0 ; )(x) + [1 + rei(x,x ) ]G0R (x, x0 , )(x). (4.71)

The complex numbers t and r are called transmission and reflection amplitude.
(c) Explain the origin of the phase shift (x, x0 ) in Eq. (4.71).

Exercise 4.3: Alternative derivation of the Friedel Sum rule

The Friedel sum rule(4.19) relates the change N in the total number of particles as the
result of the presence of an impurity to the phase shifts that particles experience while
scattering off the impurity. You have seen a formal derivation of the Friedel sum rule in Sec.
4.1. Here, we will consider a more elementary derivation of the Friedel sum rule for the case
of a spherically symmetric scatterer.
4.3. EXERCISES 65

If a flux of noninteracting particles is elastically scattered on a spherically symmetric

potential at r = 0, the wave function at large values of r will be of the form

ikz eikr
(r) e + f () , (4.72)
r
where z is the direction of the incoming flux, see Eq. (4.8) above. Since the scattered wave in
Eq. (4.72) has axial symmetry about the z-axis, it may be expanded in Legendre polynomials,

X
(r) = a` P` (cos )Rk` (r) (4.73)
`=0

The constants a` are expansion coefficients.

(a) Show that the radial functions Rk` satisfy the radial Schrodinger equation
 
2 2 Rk` 2 `(` + 1) 2m
r r + k 2 U (r) Rk` = 0, (4.74)
r r r2 ~
where the energy eigenvalues are written as ~2 k 2 /2m, the energy of a free particle, and
U (r) is the scattering potential.
(b) Show that, at distances sufficiently large that U (r) 0, the radial functions Rk` will
approach the asymptotic expansion
2
Rk` sin(kr `/2 + ` ) (4.75)
r
Both the factor of 2 and the factor `/2 are introduced for convenience, so that the
asymptotic behavior of Rk` reduces to the solutions for Rk` when U (r) vanishes iden-
tically, by putting ` = 0. The ` are the so-called scattering phases. They depend on
the scattering potential U (r) and approach zero when U (r) vanishes.
(c) The change in the number of states near the Fermi surface as a result of the scattering
potential generally depends on the ` s. To see this, we first have to connect the
general ` s in the expression (4.75) for the radial functions in the expansion of the
total wavefunction (4.73) to the function f () defined in (4.72). Apart from an overall
normalization, the coefficients a` in (4.73) are fixed by the requirement that the function
eikz is of the form f ()eikr /r, i.e. has only terms of the form of an outgoing wave,
eikr /r. Argue that this requirement leads to the identification

1 X
(2` + 1) e2i` 1 P` (cos ).
 
f () = (4.76)
2ik `=0
66 CHAPTER 4. SCATTERING FROM A SINGLE IMPURITY

Hint: use the fact that for large r the plane wave eikz can be expanded in terms of
Legendre polynomials as

ikz ikr cos 1 X
(2` + 1)P` (cos ) (1)`+1 eikr + eikr .
 
e =e =
2ikr `=0

Let us now consider the change in the number of states in the presence of the impurity. Hereto
we imagine a large sphere of radius L about the origin, and impose that the wavefunction
vanishes on this sphere. In the absence of the potential (` = 0), Eq. (4.75) then yields

` ` n
kL = n = kn` = + (4.77)
2 2L L
0 0
In the presence of the scattering potential, the wavenumbers kn` satisfy kn` = `/(2L) +
n/L ` /L, so we get for the change in wavenumber

0 `
kn` kn` = (4.78)
L
Let us first consider the special case in which only 0 is nonzero. The spacing of successive
values of kn0 is /L [see Eq. (4.77)]; if we now consider a given value of k, say k = kF , and
imagine turning on the potential, then each allowed k value shifts, according to Eq. (4.78),
by an amount 0 (kF )/L. Then the number of states k with k < kF increases by an amount

0 L 0
Nimp = = .
L
(d) Extend these arguments to prove the general Friedel sum rule (4.19), which reads, in
terms of the phase shifts ` ,
X
Nimp = 1 (2` + 1)` (kF ).
`

(e) Using similar arguments, argue that, at zero temperature, the density of electrons far
from a small spherically symmetric impurity shows oscillations
1
nimp (r) cos[2kF r + 0 (kF )] sin[0 (kF )], (4.79)
4 2 r 3
where 0 is a scattering phase shift evaluated at the Fermi energy.
4.3. EXERCISES 67

Exercise 4.4: Anderson model for impurity scattering

In some applications, a simple scattering potential is not a good description for an impurity.
This is the case, for example, if the impurity introduces an extra localized state for electrons
in an otherwise perfect lattice. In this case, one can use the following Hamiltonian to describe
the impurity,

k k

i i + (Vk k

i + Vk i

X X X
H = k + imp k ). (4.80)
k, k

Here imp is the energy of the extra localized state. The second line of Eq. (4.80) represents
hybridization of the impurity state with the Bloch states |ki of the lattice. The matrix
element Vk describes the overlap of the impurity state with the state |ki. For a strongly
localized impurity orbital at the origin, one can neglect the k-dependence of the matrix
elements Vk .
For this system, we define Green functions with respect to the wavevectors of the Bloch
states and with respect to the impurity site. Hence, the temperature Green function is
denoted as Gk0 ,k (in ), Gi,k (in ), Gk0 ,i (in ), and Gi,i (in ), etc.

(a) Show that the temperature Green function satisfies the equations
X
1 = (in imp )Gi,i (in ) Vk0 Gk0 ,i (in ),
k0
0 = (in k0 )Gk0 ,i (in ) Vk0 Gi,i (in ),
X
0 = (in imp )Gi,k (in ) Vk0 Gk0 ,k (in ),
k0
k0 k = (in k0 )Gk0 ,k (in ) Vk0 Gi,k (in ).

Solve these equations and analytically continue your answer to find the retarded Green
function GR .
(b) Show that the T -matrix is given by

Tk0 k () = Vk0 Gi,i ()Vk .

Calculate the total scattering phase shift (), and the impurity density of states
imp ().
68 CHAPTER 4. SCATTERING FROM A SINGLE IMPURITY

(c) Evaluate your expressions for the special case of a flat conduction band: the conduc-
tion electron states have a constant density of states 0 for energies D < < D, where
D is the bandwidth. For the hybridization matrix elements, you may set |Vk |2 = V 2 /Ns ,
where Ns is the total number of lattice sites in the sample. (For a lattice, the number
of available wavevectors also equals Ns .) The conduction electron density of states 0
is defined as density of states per site.

Exercise 4.5: Resonant tunneling

In this exercise we consider a quantum well, which is formed between two tunnel barriers in
a three-dimensional metal or semiconductor structure. The tunnel barriers have a potential
U (r) = u((x a) + (x + a)).
Since the system is translation invariant in the y and z directions, but not in the x
direction, we use a mixed notation, in which Green functions are Fourier transformed with
respect to y and z, but not with respect to x. The Green function is diagonal in the momenta
ky and kz .

(a) Following the same method as in exercise 4.2, calculate the retarded Green function
GR 0
ky ,kz ;ky ,kz (x, x ; ). Express your answer in the same form as Eq. (4.71) and calculate
the transmission and reflection amplitudes tky ,kz and rky ,kz .

(b) When is the transmission is unity? Find a simple physical argument for this relation.
(Hint: use your answer to Ex. 4.2.)
Chapter 5

Many impurities

We have seen that it was a difficult (but still doable!) task to calculate the Green function for
a single impurity. In the presence of many impurities, the calculation of the Green function
is as good as impossible.
One might ask whether that is a big problem. Do we really want to calculate the Green
function in the presence of many impurities? The answer is no, and the reason is very simple:
Nobody knows the exact locations and potentials of all impurities in a realistic sample! Of
course, if we dont know the exact potential U (r), it does not make sense to try to calculate
the exact Green function in the presence of that potential.
The information one usually does have is the concentration of impurities, i.e., the number
of impurities per unit volume and, if the chemical species of the impurity atoms and their
preferred location inside the lattice unit cell are known, the form of the potential of a single
impurity. Because that is all information one has, one considers an ensemble of different
samples, all with the same impurity concentration, but with different locations of the impu-
rities. Such samples are called macroscopically equivalent but microscopically different.
We then calculate quantities that are averaged over this ensemble. Such an average is called
a disorder average, to distinguish it from the thermal average we have been considering so
far.
There is an important difference between disorder averages and thermal averages. In a
thermal average, one averages over time-dependent fluctuations. According to the ergodic
hypothesis, if one waits long enough, a sample will access all accessible states. Hence, the
thermal average equals the time average. In a disorder average, one averages over impurity
configurations. For a given sample, the impurities typically do not move. Hence, in order to
obtain a disorder average experimentally one has to, somehow, involve many samples that

69
70 CHAPTER 5. MANY IMPURITIES

Figure 5.1: Reproducible magnetoconductance curves measured at T = 45 mK. Different curves

are measured for different cool downs of the same GaAs wire. Figure is taken from D. Mailly and
M. Sanquer, J. Phys. (France) I 2, 357 (1992).

have the same impurity concentration but different impurity locations.1

One way of obtaining a disorder average experimentally is thermal cycling: to repeat-
edly heat up and cool the same sample. The reason why this works is that, at sufficiently
high temperatures, impurities are somewhat mobile. Hence, after heating up and cooling
down the sample once, a different static impurity configuration is obtained. An example
of this procedure is shown in Fig. 5.1. Here, the conductance of a metal wire is measured
as a function of the magnetic field. The different curves correspond to different cool downs
of the same wire. Within one cool down, the conductance curves are fully reproducible, as
one would expect for a static impurity configuration. There are small differences between
cool downs. Indeed, different cool downs should correspond to slightly different impurity
configurations, for which one expect slightly different Green functions, and, hence, a slightly
different conductance. From the experimental data of Fig. 5.1, one can calculate the average
conductance for this ensemble of wires, the variance, and so on. The average and variance
can then be compared to a theory. You can see plots of the average conductance and of the
variance in Fig. 5.2.
In this chapter we discuss how a disorder average is taken in the Green function formalism.

1
Some observables may be self-averaging: all samples in the ensemble will behave in the same way,
irrespective of the microscopic details of impurity configuration. In the case of a self-averaging observable,
it is sufficient to measure one sample only.
5.1. DISORDER AVERAGE 71

Figure 5.2: Average (a) and variance (b) of the magnetoconductance curves of Fig. 5.1, together
with theoretical predictions. The average and variance were taken over 50 different disorder con-
figurations. The figure is taken from D. Mailly and M. Sanquer, J. Phys. (France) I 2, 357 (1992).

5.1 Disorder average

The joint effect of all impurities is described by the potential
Nimp
X
U (r) = u(r rj ), (5.1)
j=1

where rj is the location of the jth impurity. The Fourier transform of the impurity potential
is
1
Z
Uq = drU (r)eiqr
V
Nimp
X
= uq eiqrj , (5.2)
j=1

where uq is the Fourier transform of the scattering potential for a single impurity. For
delta-function impurities with u(r) = u(r) one has uq = u/V , see Eq. (4.20).
In performing the disorder average for a Green function, we have to average products of
the disorder potential Uq over the positions of the impurities. The corresponding averages
are

hUq i = uq Nimp q,0 ,

hUq1 Uq2 i = uq1 uq2 [Nimp (Nimp 1)q1 ,0 q2 ,0 + Nimp q1 +q2 ,0 ],
hUq1 Uq2 Uq3 i = uq1 uq2 uq3 [Nimp (Nimp 1)(Nimp 2)q1 ,0 q2 ,0 q3 ,0
72 CHAPTER 5. MANY IMPURITIES

+ Nimp (Nimp 1)(q1 ,0 q2 +q3 ,0 + q2 ,0 q1 +q3 ,0 + q3 ,0 q1 +q2 ,0 )

+ Nimp q1 +q2 +q3 ,0 ], (5.3)

and so on. Well consider the case when the sample is large and there are many impurities.
In that case, we can neglect the difference between Nimp 1 or Nimp 2 and Nimp in the above
formulas. Studying the case of many impurities does not mean that one can neglect, e.g.,
2
the term Nimp q1 +q2 ,0 in the second line of Eq. (5.3) with respect to the term Nimp q1 ,0 q2 ,0 .
The reason is that the former term has one more summation over the momenta, which may
bring in extra factors volume. (Note that, for a fixed concentration of impurities, Nimp is
proportional to the sample volume V .)
The disorder average is taken after the thermal average. Based on Eq. (5.3), one can
formulate a set of diagrammatic rules for the impurity average that should be implemented
on top of the diagrammatic rules for the thermal average. These rules are

Momentum has to be conserved at an impurity vertex (a star in the diagram),

Every impurity vertex (represented by a star in the diagram) contributed a factor N imp .

An arbitrary number of dashed lines can meet at a single impurity vertex. Every
dashed impurity line carrying a momentum q contributes a factor uq .

5.2 Disorder average of single-particle Green function

We now use the diagrammatic rules derived in the previous section to calculate the disorder
average of the single-particle Green function. In this section, well use the imaginary-time
formalism. The derivation in the real-time Keldysh formalism proceeds along the same lines.
Our first observation is that the disorder average restores translation invariance. Hence,
hGk,k0 ( )i is nonzero only if k = k0 ,

hGk,k0 ( )i = hT k ( )k 0 (0)i hGk ( )ik,k0 . (5.4)

A diagrammatic expansion for hGk (in )i is shown in Fig. 5.3. In this figure, all labels for
the imaginary frequency and wavevector have been suppressed. You can include them in
accordance with the diagrammatic rules of the previous section.
The diagrammatic expansion of Fig. 5.3 looks rather wild. One can introduce some order
in this expansion with the concept of irreducible diagrams. A diagram is called irre-
ducible if it cannot be cut into two pieces by cutting a single internal fermion line. For
example, the first, second, third, fifth, and eighth diagram in the expansion of Fig. 5.3 are
5.2. DISORDER AVERAGE OF SINGLE-PARTICLE GREEN FUNCTION 73

n = + + +
k k

+ + +

+ + + ...

Figure 5.3: Diagrammatic expansion of the disorder averaged single-particle Green function
hGk,k0 (in )i.

n









































 + + + + ...















=

k
Figure 5.4: Diagrammatic expansion of the self energy k for the disorder averaged single-particle
Green function hGk,k0 (in )i. The self energy is denoted by the shadded bubble at the left hand side
of the figure.

irreducible. The other diagrams in Fig. 5.3 are reducible. The sum of all irreducible dia-
grams, without the external fermion lines, is called the self energy k . Diagrammatically,
the self energy is given by the expansion of Fig. 5.4. In terms of the self energy, the disorder
averaged Green function is seen to obey a Dyson equation, see Fig. 5.5,
0 0
hGk,k (in )i = Gk,k (in ) + Gk,k (in )k (in )hGk,k (in )i. (5.5)

The solution of the Dyson equation is

0
Gk,k (in ) 1
hGk,k (in )i = 0
= . (5.6)
1 Gk,k (in )k (in ) in k + k (in )

With this result, we have reduced the calculation of the disorder averaged single-particle
Green function to that of the self energy k (in ).
74 CHAPTER 5. MANY IMPURITIES

n




n



n n
  n  n
  n  n
n = n + 







+       + ...
k k k k k k k k
k k k

n
       
n
n
= n +        
k k k
k

Figure 5.5: Diagrammatic representation of the Dyson equation for the disorder averaged Green
function hGk,k0 (in )i.

The reason why the quantity k is called self energy is clear from Eq. (5.6): the self
energy provides a shift of the energy k that appears in the denominator of the single-particle
Green function. The shift is caused by the average effect of scattering from all impurities.
Still, the diagrammatic expansion for the self energy is too complicated to be solved
exactly. In the limit of a low impurity concentration, one can restrict the diagrammatic
expansion to those diagrams that have one impurity vertex. Indeed, according to the dia-
grammatic rules of the previous section, every impurity vertex contributes a factor N imp , so
that diagrams with two or more impurity vertices contribute to higher order in the impurity
concentration. (See also exercise 5.1).) The approximation in which one considers diagrams
with one impurity vertex is referred to as the Born approximation. One refers to the nth
order Born approximation as the Born approximation where only self energy diagrams with
n + 1 or less impurity lines are kept.
Diagrammatically, the Born approximation corresponds to the expansion shown in Fig.
5.6. For the self energy this implies
"
X
0
k (in ) = Nimp u0 + uq Gk+q (in )uq
q
#
X
0 0
+ uq1 Gk+q 1
(in )uq1 q2 Gk+q 2
(in )uq2 + . . . . (5.7)
q1 ,q2

If we compare this with the expression for the T -matrix of a single impurity, Eq. (4.5), we
arrive at the important conclusion
k (in ) = Nimp Tk,k (in ). (5.8)
5.2. DISORDER AVERAGE OF SINGLE-PARTICLE GREEN FUNCTION 75



  

n 


 

     
 
 
    
  = + + + ...
       
k
Figure 5.6: Diagrammatic expansion of the Born approximation for the self energy k for the
disorder averaged single-particle Green function hG k,k0 (in )i.

In calculations, well be interested in the analytical continuation of the self energy to real
frequencies i. After analytical continuation, the real part of the self energy is nothing
but a shift of the pole of the single-particle Green function. It will usually be neglected
since it is a slow function of k and that can be absorbed in a redefinition of the chemical
potential or in a redefinition of the energy k . The imaginary part of the self energy is more
important. It can be expressed in terms of the scattering time k using the optical theorem,
Eq. (4.15),

Im k ( + i) = Im k ( i)
X
= Nimp |Tk0 k ()|2 ( k0 )
k0
1
, (5.9)
2k

where 1/k is the rate for scattering out of the momentum eigenstate |ki. The scattering
rate is Nimp times the scattering rate due to a single impurity only, see Eq. (4.11). (It is
only in the limit of a low impurity concentration that the scattering rate is proportional to
the number of impurities.)
One can also understand the first equality in Eq. (4.11) without making use of the optical
theorem. To see this, we start from the second line in Eq. (4.11), write the T -matrix in terms
of the retarded and advanced Green functions for the unperturbed Hamiltonian, cf. Eq. (4.5),
and use the equality 2( k ) = iG0R 0A
k () iGk (),
X X
2 |Tk0 k ()|2 ( k0 ) = i (G0R 0A
k0 () Gk0 ())
k0 k0
" #
X
ukk0 + ukk1 G0R
k1 ()uk1 k0 + . . .
k1
76 CHAPTER 5. MANY IMPURITIES
" #
X
uk0 k + uk0 k1 G0A
k1 ()uk1 k +... . (5.10)
k1

In this expansion, all terms that contain at least one retarded and one advanced Green
function cancel. What remains is a difference of two terms, one involving retarded Green
functions only and one involving advanced Green functions only,
" #
X i X X
|Tk0 k ()|2 ( k0 ) = ukk0 + ukk1 G0R 0R
k1 ()uk1 k0 + . . . Gk0 ()uk0 k
2
k0 k0 k1
" #
iX X
ukk0 G0A
k0 uk0 k + uk0 k1 G0A
k1 ()uk1 k + . . .
2 k0 k 1

= 2Im Tk0 k (). (5.11)

This is precisely the content of the optical theorem.
In the so-called self-consistent Born approximation one replaces the unperturbed Green
function G0 by the disorder averaged Green function hGi in the calculation of the T -matrix.
In this way, one captures some effects of higher impurity concentrations. See exercise 5.3 for
more details on the self-consistent Born approximation.
The imaginary part of the self energy gives the particle in a momentum eigenstate |ki
a finite lifetime, as you would expect for impurity scattering. You can see this explicitly if
you use the result (5.6) for the disorder averaged temperature Green function to calculate
the retarded Green function,
hGR 0R
k,k (t)i = Gkk (t)e
t/2k
. (5.12)
In frequency representation, the finite lifetime shows up as a finite width of the delta-function
peak of the spectral density,
1/k
hAk ()i = . (5.13)
( k )2 + (1/2k )2
We call the disorder weak if the broadening of the peak in the spectral density is much
smaller than the energy k or, equivalently, if the scattering life time is much larger than the
period 1/k .

5.3 Boltzmann Equation with impurity scattering

Short-range impurities scatter electrons into states with a different momentum. In this
section we discuss how impurity scattering can be incorporated into the Boltzmann equation.
5.3. BOLTZMANN EQUATION WITH IMPURITY SCATTERING 77

Our derivation of the Boltzmann equation from the Green function formalism uses the
real-time Keldysh formalism. In the real-time formalism, the disorder average is done in the
same way as in the imaginary-time formalism. In the Born approximation, one derives the
Dyson equation, for the disorder averaged Green function,
1
hGi = (it H0 )1 1

, (5.14)

where we have used operator language to write down the action of the single-particle Green
functions. Here H0 is the Hamiltonian without the short-range impurity potential (but H0
may include a slowly varying external potential) and is the self energy, the sum of all
irreducible diagrams (excluding external fermion lines). The self energy has same matrix
structure as the Green function,

R (r, t; r0 , t0 ) K (r, t; r0 , t0 )
 
= . (5.15)
0 A (r, t; r0 , t0 )

Repeating the calculation of the previous section for the self-energy , we find, in momentum
representation,

R
k = Nimp Tkk ( + i)
A
k = Nimp Tkk ( i) (5.16)
X
K
k = N imp Tk,k0 ( + i)G0K
k0 Tk0 ,k ( i)
k0
X
= Nimp |Tk,k0 ()|2 G0K
k0 .
k0

In the self-consistent Born approximation, G0K is replaced by hGK i.

The derivation of the Boltzmann equation in the presence of impurity scattering closely
follows the derivation for the Boltzmann equation without impurities. We start by rewriting
the Dyson equation as

(it H0 )hGi = 1 = hGi(it H0 ) (5.17)

and subtract the far left and right hand sides. Taking the Keldysh component of the re-
maining equation, we find

i i
[(it H)1 <, hGK i] [K , h<Gi] = [K , hAi]+ [, hGK i]+ , (5.18)
2 2
78 CHAPTER 5. MANY IMPURITIES

where A = i(GR GA ) is the spectral weight of the full Green function and

= i(R A ),
1 R
< = ( + A ),
2
1 R
<G = (G + GA ). (5.19)
2
The definition of real and imaginary parts is the same one as used in Chapter 2. Note
that in equilibrium one has GK = i tanh(/2T )A and K = i tanh(/2T ), so that the
right hand side of Eq. (5.18) vanishes.
The self energy represents the effect of impurity scattering. We neglect spatial and
temporal derivates of the self energy, as well as its dependence on momentum and energy.
(This is consistent with the calculation of the self energy in the previous section.) This
allows us to disregard the self energy in the left hand side of Eq. (5.18). For the right-hand
side, we use the fact that the spectral density A is close to a delta function if k  1, cf.
Eq. (5.13)
hAi = 2[ (k ) U ]. (5.20)
Similarly, the disorder-averaged Keldysh Green function hGK i is also strongly peaked as a
function of ,
hGK i = 2i[ (k ) U ](1 2fk ), (5.21)
where the distribution function f is defined as in Eq. (3.38). Then, using the gradient
expansion in Eq. (5.18), integrating over , and performing one partial integration, we find

Nimp X
(T + vk R R U k ) fk (R, T ) = 2 |Tk,k0 |2 (k k0 )(fk fk0 ), (5.22)
V 0k

where vk = k k is the velocity. The right-hand side of the Boltzmann equation is the
collision integral: it contains the effect of the collisions between electrons and impurities
with a short-range potential.

5.4 Electrical conductivity

Impurities in a conductor form an obstacle for the flow of electrical current. In normal metals,
application of an external electric field E gives a finite electrical current density j = E. The
Boltzmann equation can be used to find a microscopic expression for the conductivity .
5.4. ELECTRICAL CONDUCTIVITY 79

In equilibrium, the distribution function fk is the Fermi function f 0 ,

1
fk0 = . (5.23)
1+ e(k )/T
The presence of an electric field changes the distribution f . In this section, we are interested
in the response to small electric fields. Hence, we only consider deviations from f 0 to linear
order in the electric field,
fk = fk0 + fk1 . (5.24)
After some algebra, one can rewrite Eq. (5.18) into a simple form,
e f1
E k fk = k , (5.25)
~ ~k,tr
In general, this simple form is known as the relaxation time approximation, although it is
exact for the problem we consider here. The time k,tr is known as the transport mean free
time. It is expressed in terms of the T -matrix of a single impurity as
1 2Nimp X
= |Tkk0 (k )|2 (1 cos 0 )(k k0 ), (5.26)
k,tr ~ k0

where 0 is the angle between k and k0 .

Linearizing in the electric field E, we find
fk1 = ek,tr E k f 0 (k). (5.27)
Using Eq. (3.40) for the current density j, we have
2e2 X fk0
 
j = vk (E vk ) k,tr , (5.28)
V k
k

where we inserted a factor two to account for spin degeneracy. We then perform the inte-
gration over k by first averaging over the angles and find
2e2 X 2 f
= vk k,tr ,
3V k k

which we can rewrite in terms of the density n of electrons and for low temperatures T as
ne2 tr (kF )
= . (5.29)
m
This is nothing but the well-known Drude formula for the conductivity.
80 CHAPTER 5. MANY IMPURITIES

5.5 Gaussian white noise

Dealing with the precise impurity potential is often cumbersome: whereas calculations start
from the impurity potential, final results will have to be formulated in terms of the T -matrices
and, subsequently, in an appropriate mean free time.
In the remainder of this course, well use a simpler model for the impurity potential than
Eq. (5.1) above. In this model, which has been chosen to minimize calculational difficulties,
both the positions and the potentials of the impurities are random: the entire impurity
potential U (r) is considered a random function, with a Gaussian distribution of zero average,

hU (r)i = 0, (5.30)

and with a delta-function correlated second moment,

1
hU (r)U (r0 )i = (r r0 ), (5.31)
2
where is the mean free scattering time (see Sec. 5.2) and is the density of states at the
Fermi level per unit volume,
k2
= 2 . (5.32)
2 ~vF
Higher moments of the impurity potential can be constructed using the rules for Gaussian
averages. This random potential U (r) is called white noise because it is delta-function
correlated in space.
The model (5.31) is not faithful to the microscopic potentials. Yet, as many observables
depend in essence only on a suitably defined mean free time (such as the transport mean free
time for the conductivity), one can imagine doing a calculation in two steps: use a correct
microscopic theory to calculate the relevant scattering times, and use the model (5.31) to
find physical observables that are hard to get at using the full microscopic theory.
Performing the Fourier transform,
1
Z
Uq = drU (r)eiqr ,
V
we find that the potential Uq is also random and Gaussian. Its first two moments are given
by

hUq i = 0
1
hUq Uq0 i = q+q0 ,0 . (5.33)
2 V
5.6. ELECTRICAL CONDUCTIVITY FROM KUBO FORMULA 81

n


 
 
n
 
  
  
  kq

     

 
 

  
=

k
Figure 5.7: Self energy diagram for Gaussian white noise potential.

All higher moments can be factorized into the first and second moments. For the diagram-
matic rules for impurity average this means that only those impurity vertices with two dashed
lines contribute to the average.
To leading order in the impurity concentration, we then find only one diagram for the
self energy, see Fig. 5.7. Hence, the self energy is given by

X 1
k = Gkq,kq (in )
q
2 V
1 dq 1
Z
= 3
2 (2) in kq +
i
= sign (n ). (5.34)
2

This is the same answer as we obtained previously in Sec. 5.2. In fact, the normalization
(5.31) was chosen precisely to get this answer.2

5.6 Electrical conductivity from Kubo formula

We now re-calculate the electrical conductivity of a disordered metal in the Green function
formalism. The calculation is considerably more cumbersome than the Boltzmann equation
calculation of Sec. 5.4 (which, of course, is fully rigorous given our microscopic derivation
of the Boltzmann equation). In order to keep the calculation simple, we will consider use
Gaussian white noise as our model for the disorder potential.

2
The most important difference between the Gaussian white noise potential and realistic impurity poten-
tials is that, for Gaussian white noise, the impurity scattering lifetime and the transport mean free time are
equal, whereas they are not for generic impurity scattering.
82 CHAPTER 5. MANY IMPURITIES

For a normal metal, the general response to an electrical field E takes the form
Z Z X
0
je (r, t) = dt dr0 (r, t; r0 , t0 )E (r0 , t0 ), (5.35)

where is the (nonlocal) conductivity tensor. The electric field is related to the electric
potential and the vector potential A,
1
E = r t A. (5.36)
c
This relation holds both for the external electric field and for the total electric field. If the
current is carried by electrons, the electrical current operator is a sum of a paramagnetic
contribution jp and a diamagnetic contribution jd ,
je (r) = jpe (r) + jde (r), (5.37)
where
~e X h
i
jpe (r) =
(r)(r (r)) (r (r)) (r) , (5.38)
2mi
e2
(r) (r).
X
jde (r) = A(r) (5.39)
mc

In Sec. 2.5 we derived a general formalism to calculate the linear response to a perturba-
tion H 1 . In our case, we want to find the current density as a response to an applied electric
field. To linear order in the electric field, the electrons Hamiltonian is H0 + H 1 , where H0
is the Hamiltonian in the absence of the electric field and
Z Z

H1 = dre (r)(r, t) drje (r)Aext (r, t), (5.40)

where
(r) (r)
X
e (r) = e (5.41)

is the operator for the electronic charge density. Since we are interested in linear response,
we only need to know H 1 to linear order in A, so that we can replace the electrical current
density je in Eq. (5.40) by the paramagnetic current density jpe .3 We choose a gauge in which
= 0. This would be the natural choice if, e.g., the sample is a cylinder and the electric
field is generated by a time-dependent flux through the sample, see Fig. 5.8.
3 0 contains a vector potential A0 due to a magnetic field, we should replace jpe by
If the Hamiltonian H
jpe + (e/m)A0 e .
5.6. ELECTRICAL CONDUCTIVITY FROM KUBO FORMULA 83

E(t)

(t)

Figure 5.8: The gauge with zero scalar potential and a time-dependent vector potential A is the
natural choice if the electric field is generated by a time-dependent flux through the sample.

Fourier transforming Eq. (5.35), we find

Z X
je (r, ) = dr0 (r, r0 ; )E (r0 , ), = 1, 2, 3, (5.42)

where, in frequency representation, the electric field is related to the vector potential as
1
Z
Aext (r, ) = Eext (r, ) = dteit Aext (r, t) (5.43)
i
and Z
0
(r, r ; ) = dteit (r, t; r0 , 0). (5.44)

According to the Kubo formula (2.69), the paramagnetic contribution to the conductivity is
expressed in terms of the retarded current-density autocorrelation function, R (r, r0 ; t),
i R ee (r)
(r, r0 ; ) = (r, r0 ; ) (r r0 ) . (5.45)
im
where
R 0 R 0 0
(r, r ; t) Gje ,je (r, r ; t) = i(t)h[je (r, t); je (r , 0)] i. (5.46)
84 CHAPTER 5. MANY IMPURITIES

The second term in Eq. (5.45) represents the diamagnetic contribution to the current density.
For a translationally invariant system, the conductivity depends on the difference r r0
only. Performing Fourier transforms for the conductivity and for the current density,
1
Z
0
(q, ) = drdr0 eq(rr ) (r, r0 ; ), (5.47)
V
Z
je,q = dreiqr je (r), (5.48)

we can write the Kubo formula for the conductivity as

i R ee
(q, ) = (q, ) . (5.49)
im
Here e is the electron charge density averaged over the entire sample (i.e., the q = 0
component of the charge density divided by the volume), and
1
R
(q, t) = i(t) h[je,q (t), je,q (0)] i. (5.50)
V
In general, the conductivity is defined as the response to the actual electric field in the
sample, not as the response to the external field. If the electric fields are time dependent,
the currents are time dependent and, thus, generate their own magnetic and electric fields.
In addition an applied electric field may lead to a build-up of charge, which can screen the
field. What we calculated above is the response to the external electric field. This procedure
gives the correct result only if we can neglect the induced electric field. This is the case if,
on the one hand, the frequency of the signal is sufficiently low that no magnetic fields are
induced, whereas, on the other hand, the frequency is large enough compared to the size of
the sample or the wavelength of the electric field that no screening charges are built up. In
practice, both conditions are met if the electric field is perpendicular to the wavevector q or
if the wavevector q is taken to zero before the frequency is made small.
Let us now calculate the conductivity of an electron gas. If there are no impurities, the
conductivity is infinite (see exercise 5.6). So, in order to do a meaningful calculation, we
consider the conductivity of an electron gas with impurities.
The conductivity follows from the current-current correlation function (5.50). For our
calculation, we consider the imaginary-time version of that correlation function,
1
, (q, 0 ) = hT jq, ( )jq, ( 0 )i, (5.51)
V
and perform an analytical continuation to real frequencies at the end of the calculation. The
current density operator jq is expressed in terms of the creation and annihilation operators
5.6. ELECTRICAL CONDUCTIVITY FROM KUBO FORMULA 85

for electrons in momentum eigenstates |ki as

e X
jq = (2k + q)k,

k+q, . (5.52)
2m k

Since we are dealing with a non-interacting electron gas, we can perform the thermal average
in Eq. (5.51) using Wicks theorem, with the result

e2 X X
; (, 0 ) = Gk+q,;k0 ,0 ( 0 )Gk0 q,0 ;k, ( 0 )
4m2 V k 0 0
k
(2k + q )(2k0 q ), (5.53)

plus a term that does not depend on 0 and can be discarded because it vanishes after
Fourier transform to the Matsubara frequency in and subsequent analytical continuation
in + i. It is important to note that the Green function in Eq. (5.53) is the full
temperature Green function, calculated in the presence of the exact impurity potential.
Fourier transforming Eq. (5.53) and changing variables k0 k0 + q, we find

e2 T X
; (in ) = Gk+q,k0 +q (im + in )Gk0 ;k (im )
2m2 V k,k0 ,m
(2k + q )(2k0 + q ), (5.54)

where we summed over the spin index. The Matsubara frequency in is a bosonic Matsubara
frequency, because the current-current correlation function involves operators that are even
in the fermion creation and annihilation operators. The arguments of the single-particle
Green functions, however, are fermionic Matsubara frequencies.
In order to calculate the full temperature Green function, we use the perturbative expan-
sion of G in powers of the impurity potential, which is shown in Fig. 4.2, and then perform
the disorder average. It is important to realize that, here, we calculate the impurity average
of the product of two Green functions. That need not be the same as the product of the
averages!
In order to keep the calculations simple, well restrict our attention to the case of a
Gaussian white noise random potential. Then, diagrammatically, the impurity average means
that we connect all pairs of impurity lines, as in Fig. 5.7. Still, it is an impossible task to
sum all diagrams. We can further simplify the calculation by limiting ourselves to those
diagrams in which no impurity lines cross. In exercise 5.1 youll verify that this is a sensible
approximation: diagrams in which impurity lines cross are a factor F smaller than diagrams
without crossed impurity lines.
86 CHAPTER 5. MANY IMPURITIES

m m m m m
k k k k k" k
+ + + ...
k+q k+q k+q k+q k"+q k+q
m+ n m+ n m+ n m+ n m+ n

Figure 5.9: Diagrammatic representation of all diagrams without crossed lines that contribute to
the conductivity. The double lines represent the disorder averaged Green function.

The diagrams without crossed lines can be summed up as in Fig. 5.9. The building
blocks are the impurity averaged Green functions, which are represented by double arrows.
The open circles at the two ends are called current vertices. They represent the factors
e(2k + q)/2m and e(2k0 + q)/2m that appear in the calculation of the current-current
correlation function. These factors are referred to as the bare current vertex, and they are
denoted by the bare vertex function 0 (k, im ; k + q, im + in ),
e(2k + q )
0 (k, im ; k + q, im + in ) = . (5.55)
2m
The middle part of the diagram is a ladder which has the impurity averaged Green func-
tions and impurity lines as building blocks.
All but the rightmost part of the ladder diagram can be written as a solid dot, see Fig.
5.10. This part of the diagram is known as a vertex correction. It is described by the vertex
function (k, im ; k + q, im + in ). A self-consistent equation for the vertex function can
be found from the second line of Fig. 5.10,
(k, im ; k + q, im + in ) = 0 (k, im ; k + q, im + in )
1 X
+ hGk0 (im )ihGk0 +q (im + in )i (k0 , im ; k0 + q, im + in ), (5.56)
2 V 0
k

where 0 is the bare vertex function, see Eq. (5.55) above. Recall that the impurity averaged
Green functions have only one momentum index because translation invariance is restored
upon taking the disorder average. A self-consistent equation for the vertex function is found
upon multiplication of Eq. (5.56) with hGk (im )ihGk+q (im + in )i and summation over k.
Solving for , we find
(k, im ; k + q, im + in )
0
 P 
e k0 (2k +Pq )hGk0 (im )ihGk0 +q (im + in )i
= 2k + q + . (5.57)
2m 2 V k0 hGk0 (im )ihGk0 +q (im + in )i
We are interested in the current-current correlator in the limit q 0. The numerator of
the last term of Eq. (5.57) goes to zero if q 0. Hence, we conclude that, for our choice of
5.6. ELECTRICAL CONDUCTIVITY FROM KUBO FORMULA 87

m m
m k m k k k k k" k
= + + + ...
k+q k+q
m+ n m+ n k+q
k+q +
k+q k"+q
k+q
m+ n m n

m
m k k k
= +
k+q
m+ n k+q
k+q
m+ n

Figure 5.10: Diagrammatic representation of vertex correction necessary for the calculation of
the current-current correlator. The double lines represent the disorder averaged Green function.

m
k

k+q
m+ n

Figure 5.11: Diagrammatic representation of the current-current correlator in terms of the vertex
function (solid dot), the bare vertex function 0 (open dot) and the disorder averaged Green
function (double lines).

the random potential, there is no vertex correction.4 This simplification one of the reasons
why we have chosen the Gaussian white noise model for the random potential.
The current-current correlation function is easily expressed in terms of the vertex function
and the bare vertex function 0 , see Fig. 5.11,

2T X
, (in ) = (k, im ; k + q, im + in )0 (k + q, im + in ; k, im )
V k,m
Gk (im )Gk+q (im + in ). (5.58)

4
Unlike for the Gaussian white noise used here, and 0 are different for a realistic impurity potential.
This vertex correction amounts to the replacement of the impurity scattering lifetime with the transport
mean free time tr , see Sec. 5.3.
88 CHAPTER 5. MANY IMPURITIES

It remains to perform the summations over k and m and do the analytical continuation
in + i. We first do the summation over m. As before, this is done by representing
the Matsubara summation as an integral in the complex plane,
1
Z X
, (in ) = dz (k, z; k + q, z + in )0 (k + q, z + in ; k, z)
2iV C1
k
Gk,k (z)Gk+q,k+q (z + in ) tanh(z/2T ). (5.59)
The integrand has singularities at the real axis and at the axis z = x in , x real. Hence,
we deform the contour as shown in Fig. 5.12.
1
Z XX
, (in ) = d tanh(/2T ) (1)
2iV k

(k, ; k + q, + in )0 (k + q, + in , k, ) Gk ( )Gk+q ( + in )



+ (k, in ; k + q, )0 (k + q, , k, in ) Gk ( in )Gk+q ( ) ,


where we used tanh(z/2T ) = tanh[(z + in )/2T ] and = i. It is not necessary to

add an infinitesimal i to in , since in is imaginary itself. Now we can perform the
analytical continuation in + i,
1
Z X
, ( + i) = d tanh(/2T )
2iV k
(k, ; k + q, + ) (k + q, + ; k, )GR
 RR 0 RR R
k ()Gk+q ( + )

AR 0 RA
(k, ; k + q, + ) (k + q, + ; k, )GA R
k ()Gk+q ( + )
+ AR 0 RA
(k, ; k + q, ) (k + q, ; k, )GA R
k ( )Gk+q ()

AA 0 AA
(k + q, ; k, )GA A

(k, ; k + q, ) k ( )Gk+q () . (5.60)

Sofar our result has been quite general. Equation (5.60) is valid for any form of the
vertex correction and even holds in the presence of electron-electron interactions (which will
be discussed in a later chapter). We now specialize to the limit q 0, which is needed to
find the dc conductivity. As explained above, we should take the limit q 0 before we take
the limit 0. For zero wavevector q there is no vertex renormalization, and hence
e2 X k k
Z
, ( + i) = d tanh[( )/2T ] (5.61)
2iV k m2

1 1

( k + i/2 )( + k + i/2 ) ( k i/2 )( + k + i/2 )
5.6. ELECTRICAL CONDUCTIVITY FROM KUBO FORMULA 89

C1 C1
C2

C2
C2

C2

Figure 5.12: Integration contours for the calculation of the current-current correlator.


1 1
+ ,
( k i/2 )( k + i/2 ) ( k i/2 )( k i/2 )

where we shifted the -integration by the chemical potential . We first perform a partial
integration to ,

e2 1
Z
, ( + i) = d 2
4iV T cosh [( )/2T ]
X k k  1 ( k )2 ( + i/2 )2

1
ln .
k
m2 + i/ ( k )2 + 1/4 2

With this manipulation the summation over the wavevector k is convergent. The summation
over k is separated into an angular average over the direction of k and an integration over
the magnitude. The latter integration can be rewritten as an integral over the energy k and
yields

2e2 kF2 ,
, ( + i) = . (5.62)
3(1 i )m2

We use the relation = mkF /2 2 ~2 for the density of states per spin direction and per unit
90 CHAPTER 5. MANY IMPURITIES

volume, and the relation n = kF3 /3 2 = e /e for the total particle density to rewrite this as5
e e,
R
, () = . (5.64)
m(1 i )
Substituting this in Eq. (5.49), we find that the conductivity at wavevector q = 0 and
frequency is given by
ee
(0, ) = . (5.65)
m(1 i )
In this limit 0 this reproduces the well-known Drude formula.
It may be somewhat disappointing to recover the Drude formula after so much work.
However, this whole calculation was not in vain. First of all, we have found a firm microscopic
derivation of the Drude formula. But more importantly, with this formalism we are ready
to study corrections to the Drude formula. One of these corrections, the weak localization
correction, will be studied at length in a later chapter.
5
You can verify this result in the limit 0 without expanding around the Fermi surface. Starting
point is Eq. (5.61), which we rewrite as

e2 X k k
Z
R
, (0) = d tanh[( )/2T ]Im hGR 2
k,k ()i .
V m2
k

We make use of the fact that the integral of hGR ()i2 vanishes and perform the angular average over the
wavevector k,

4e2 X k 2 1
Z
R, (0) = d ()/T
Im hGR 2
k,k ()i .
3V m 2m 1 + e
k

Using the explicit form of Im hGR

k,k ()i one verifies that

1
Im hGR 2
k,k ()i = hAk,k ()i. (5.63)
2 k
Performing a partial integration with respect to k one thus finds

2e2 X 1
Z
R
, (0) = k d ()/T
hAk,k ()i
3V m 1 + e k
k
2e2 X d 1
Z
= hAk,k ()i
Vm 2 1 + e()/T
k
e e
= .
m
1/2
Note that the partial integration gives a factor 3/2 since the density of states is proportional to k . This
is the desired result.
5.7. EXERCISES 91

n n n n n n

kqq
kqq

kq
kq

kq

kq
Figure 5.13: Diagrams that were omitted in the calculation of the self energy.

5.7 Exercises
Exercise 5.1: Corrections to the self energy

For the Gaussian white noise potential, we neglected some self-energy diagrams. The simplest
diagrams we left out are shown in Fig. 5.13. Calculate these contributions to the self energy
and show that they can be omitted if disorder if the impurity concentration is small.

Exercise 5.2: Boltzmann Equation

Derive Eq. (5.25).

Exercise 5.3: Self-consistent Born approximation

In the self-consistent Born approximation, one substitutes the unperturbed Green function
G 0 with the disorder averaged full Green function G in the calculation of the self energy.
Since G depends on , the result of this substitution is a self-consistent equation for the self
energy .

(a) Draw diagrams representing the equation for the disorder averaged Green function G
and the self energy in the self-consistent Born approximation.

(b) Write down a self-consistency equation for the self energy in terms of the T -matrix.
Hint: since the real part of can be absorbed into the chemical potential, it is the
imaginary part only that plays a role.
92 CHAPTER 5. MANY IMPURITIES

(c) Discuss the solution if the T -matrix has only a weak energy dependence for energies
within a distance  F from the Fermi level.
(d) Discuss the self-consistent Born approximation for the Gaussian white noise potential.

Exercise 5.4: One-dimensional conductor

It is possible to fabricate samples that behave as defect-free one-dimensional conductors.

Examples are quantum wires in semiconductor heterostructures, or carbon nanotubes. At
very low temperatures, interactions lead to a strongly correlated electronic ground state (the
so-called Luttinger Liquid). However, at not too low temperatures, a description in terms
of the Boltzmann equation may still be appropriate.
In a one-dimensional conductor, the relevant momenta p are in the vicinity of either pF or
pF , see figure 5.14. One usually refers to the two types of excitations as right movers and
left movers. As a result, in one dimension the distribution function f can be represented
by two functions fR and fL that each depend on the kinetic energy kin of the electrons only.
In the absence of collisions, the Boltzmann equation then reads
fR fR fR
+ vF + eEvF = 0, (5.66)
t x
fL fL fL
vF eEvF = 0, (5.67)
t x
where vF is the Fermi velocity.

(a) Show that in one dimension, the Fermi velocity is related to the density of states at
the Fermi level (F ).

Consider a one dimensional conductor in which the electric field exists only within a
segment of length L. Assume that all electrons entering that segment are in thermal equi-
librium, i.e., their distribution function f is the Fermi function. Also assume that the one
dimensional conductor is perfect, i.e., that there are no collisions between the electrons or
between electrons and impurities or defects.
(b) Solve the Boltzmann equation to find the electrical current in the conductor.
(c) What does your answer to (b) imply if expressed in terms of a conductivity? Does
your answer make sense?
5.7. EXERCISES 93

p p p
F F

Figure 5.14: Dispersion relation of a one dimensional metal. All relevant momenta are close to p F
and pF .

(d) One can also express the answer to (b) in terms of the conductance G, defined as
the quotient of the current I and the voltage V = EL over the conductor. What
conductance do you find? What physical properties does it depend on?

Exercise 5.5: Electrical conduction

In this exercise we return to the problem of electrical resistance of a perfect one-dimensional

conductor. A possible measurement setup is shown in Fig. 5.15: The wire is connected to
source and drain reservoirs. There exists a voltage difference V between the electrostatic
potentials in the two reservoirs, and the current through the wire is measured. Technically,
such a measurement of the resistance is referred to as a two-terminal measurement, because
the voltages and currents are measured with respect to the same reservoirs. The Boltzmann
equation you solved in Ex. 5.4 is for this measurement setup.
In Ex. 5.4 you saw that, in a one-dimensional conductor, one can use the kinetic energy
F to label an electronic state, instead of the momentum p, together with a label that
specifies whether the electron moves to the right or to the left. The Boltzmann equation
[Eqs. (5.66 and (5.67) of Ex. 5.4] expresses the fact that the kinetic energy of a moving
electron changes because of the presence of the electric field. However, the electrons total
energy tot = + e, which is the sum of potential energy e and kinetic energy , does not.
We can use the fact that the total energy is conserved to write down a simpler version of
the Boltzmann equation.6

(a) Argue that, if we describe the state of an electron in a one-dimensional conductor by

6
This version is nothing but the quasiclassical kinetic equation you derived in Ex. 3.4.
94 CHAPTER 5. MANY IMPURITIES
          
drain reservoir
 
         source  reservoir         
 
           
               I             
           
         
 conductor 
 
                  
       
V

Figure 5.15: Setup for the two-terminal measurement of the electrical resistance of a metal.

specifying its total energy instead of its kinetic energy kin , the Boltzmann equation
reads simply
fR fR
+ vF = 0, (5.68)
t x
fL fL
vF = 0. (5.69)
t x
Here 0 < x < L is the coordinate along the conductor.
(b) What are the appropriate boundary conditions for the distribution functions fL and
fR at x = 0 and x = L?
Hint: both source and drain reservoirs have the same electron density.
(c) Solve the Boltzmann equation, calculate the current I as a function of the applied
voltage V , and compare your answer with Ex. 5.4.

Exercise 5.6: Conductivity without impurities

Calculate the retarded current density correlation function R ; () for a non-interacting

electron gas without impurities. Is the result the same as the limit of Eq. (5.64)?
What does your result imply for the relation between current density and electric field?
Chapter 6

Diffusion modes

Consider changing the potential at a certain time t. The electron density will adjust to
the change in the potential. In this chapter well investigate how this adjustment takes place
and how it depends on the presence of impurities.

6.1 Dielectric function

Consider a metal or other dielectric with an added charge. One example of such a charge is
an ion with a different valence in a metal. Well call this charge the external charge and
denote the corresponding charge density by ext .
The external charge determines the electrical displacement D, which is the solution of
the Maxwell equation
0 r D = 4ext . (6.1)

Here, 0 is the permittivity of vacuum. In a dielectric, the external charge will induce a
screening charge ind . The total charge, screening charge plus external charge, determines
the electric field E,
0 r E = 4(ext + ind ). (6.2)

In the language of external and induced quantities, you may consider E as the total
electric field, and D as the external electric field. The difference E D then corresponds
to the induced electric field.
In general, the electric field, the electric displacement, and their difference (or, more pre-
cisely, the longitudinal parts of these fields) can be written as gradients of electric potentials
tot , ext , and ind , respectively. These potentials are expressed in terms of the corresponding

95
96 CHAPTER 6. DIFFUSION MODES

charges as Z
(r) = dr0 VC (r r0 )(r0 ), (6.3)

where VC (r r0 ) = 1/0 |r r0 | is the Coulomb interaction.

For a sufficiently small external charge, the induced and total potentials are proportional
to the external potential. This is the regime of linear screening. The proportionality
constant, which may be a non-local function in space and time, is called the inverse relative
permittivity of the dielectric, or the dielectric response function,
Z Z
0
tot (r, t) = dr dt0 1 (r, t; r0 , t0 )ext (r0 , t0 ). (6.4)

Instead of the dielectric response function, one often uses the so-called polarizability func-
tion e (r, t; r0 , t0 ), which gives the induced charge as a function of the external potential,
Z Z
0
ind (r, t) = dr dt0 R 0 0 0 0
e (r, t; r , t )ext (r , t ). (6.5)

From Eq. (6.5) we then conclude that

Z
1 0 0 0 0
 (r, t; r , t ) = (r r ) (t t ) + dr00 VC (r r00 )R 00 0 0
e (r , t; r , t ). (6.6)

In a translationally invariant medium, the dielectric response function 1 (r, t; r0 , t0 ) and

the polarizability function e (r, t; r0 , t0 ) depend on the differences r r0 and t t0 only. Then
it is advantageous to perform a Fourier transform to position and time. Defining the Fourier
transforms according to
1
Z Z
1 0 0
 (q, ) = drdr dteiq(rr )+it 1 (r, t; r0 , 0), (6.7)
V
Z Z
(q, ) = dr dteitiqr (r, t), (6.8)

with similar definitions for e (q, ) and (q, ), we find

tot (q, ) = 1 (q, )ext (q, ), (6.9)

ind (q, ) = e (q, )ext (q, ). (6.10)

Fourier transforming Eq. (6.6), we have

1 (q, ) = 1 + VC (q)R
e (q, ), (6.11)
6.1. DIELECTRIC FUNCTION 97

where VC (q) = 4/0 q 2 is the Fourier transform of the Coulomb potential.

Let us now calculate the polarizability function e for an electron gas without impurities.
The simplest approach is to use the Boltzmann equation. We write f = f 0 + f 1 , where
f 0 is the equilibrium distribution function and f 1 is the first-order response to the external
potential ext . Linearizing and Fourier transforming the Boltzmann equation (3.21), we find

(i + ivk q)fk1 (q, ) i(e/~)ext (q, )q k fk0 (q, ) = 0. (6.12)

Solving for f 1 , we find

e2 X q vk
R f 0 (k ) ,


e (q, ) = 2 (6.13)
V k q vk i

where a factor two has been inserted to account for spin degeneracy.
A calculation of e that is totally inside the Green function formalism makes use of the
Kubo formula, see Sec. 2.5. We make use of the fact that the external charge causes a change
of the Hamiltonian given by
Z

H1 = dre (r)ext (r, t). (6.14)

Then, according to the Kubo formula (2.69), we find that the induced charge density is given
by Eq. (6.5), with
e (r, t; r0 , t0 ) = GR 0 0
e ,e (r, t; r , t ). (6.15)
Using the Fourier transform of the charge density,
XZ
dreiqr (r) (r) = e k, k+q, ,
X
e (q) = e (6.16)
k,

tother with the proper time dependence of the operators k+q, and k,
in the ideal electron
gas, we find
1 X X
e (q; t) = ie2 (t) h[k, k+q, , k 0 ,0 k0 q,0 ] iei(k k+q )t/~ . (6.17)
V k 0 0
k

The thermal average of the commutator is easily found to give [f 0 (k ) f 0 (k+q )]k,k0 q 0 ,
where f 0 is the Fermi distribution function. Finally, performing a Fourier transform to time,
we find
2e2 X f 0 (k ) f 0 (k+q )
e (q, ) = , (6.18)
V k k k+q + + i
98 CHAPTER 6. DIFFUSION MODES

C1 C1

C2

k
k+q i n
C2

Figure 6.1: Integration contours for the calculation of the polarizability of the noninteracting
electron gas.

where is a positive infinitesimal. Note that e , being a retarded correlation function, sat-
isfies the general integral relation (2.45) that expresses the full retarded correlation function
in terms of its imaginary part only.
An alternative method to calculate e is to use the imaginary time formalism. One first
calculates
e2 X X
e (q, ) = T hk,

( + )k+q, ( )k 0 ,0 ()k0 q,0 (0)i (6.19)
V k 0 0
k
e2 X X
= [Gk+q,;k0 ,0 ( )Gk0 q,0 ;k, ( ) Gk+q,;k, ()Gk0 q,0 ;k0 ,0 ()] .
V k k0 0

The positive infinitesimal has been added to ensure the placement of creation operators in
front of the annihilation operators at the same imaginary time in the time-ordered product.
The second term does not depend on . After Fourier transform with respect to it will give
a contribution proportional to n ,0 , which can be discarded for the analytical continuation
in + i we need to do at the end of the calculation. Fourier transform of the first term
and substitution of the exact results for the electron Green functions gives

2T e2 X 1
e (q, in ) = .
V k,m [im (k )][i(m + n ) (k+q )]
(6.20)

Here n is a bosonic Matsubara frequency (recall that q is quadratic in fermion creation and
annihilation operators) and m is a fermionic Matsubara frequency. In order to perform the
6.1. DIELECTRIC FUNCTION 99

summation over m , the summation is written as an integral over a suitably chosen contour
in the complex plane, see Fig. 6.1. The integrand is a function of the complex variable
z and has poles at the fermionic Matsubara frequencies z = im , at z = k , and at
z = k+q in , hence

e2 X tanh(z/2T )
Z
e (q, in ) = dz
2iV k C1 [z (k )][z + in (k+q )]
e2 X tanh[(k+q )/2T ] tanh[(k )/2T ]
= . (6.21)
V k k k+q + in

Now the analytical continuation n + i is easily done and one recovers the result (6.18)
upon substituting tanh(( )/2T ) = 1 2f 0 ().
The Boltzmann equation result (6.13) agrees with what is found from the microscopic
calculation in the limit q 0. This is no surprise, once we recall that the Boltzmann
equation was derived in the gradient approximation, i.e., under the assumption that all
external potential are slowly varying functions of spatial and temporal coordinates.
Let us now consider the dielectric response at finite frequency in more detail. The polar-
izability describes how the induced charge varies with the time-dependent external charge.
A time-dependent induced charge requires a time-dependent current density j, which, by the
continuity relation, is proportional to the time derivative of the induced charge. The real
part of the polarizability describes that part of the induced current that is out of phase with
respect to the external electric field. For such currents, the time-average of j E is zero, so
that no energy is dissipated in one cycle. Hence, Re e describes the dissipationless response
to an external potential. Similarly, the imaginary part of the polarizability describes dissi-
pation: current and field are in phase, so that the time average of j E is nonzero. Hence,
the imaginary part of the polarizability gives information about the possible excitations of
the electron gas.
For the noninteracting electron gas, the dissipative response is given by the imaginary
part of Eq. (6.18) above,

2e2 X 0
Im e (q, ) = [f (k ) f 0 (k+q )](k k+q + ~). (6.22)
V k

Hence, electromagnetic energy can be dissipated by the excitation of particle-hole pairs of

energy ~ and momentum q. Let us investigate Eq. (6.22) in more detail for zero temperature
and > 0.1 Then the difference f 0 (k ) f 0 (k+q ) is one if k is inside the Fermi sphere
1
The case < 0 follows from the relation R R
e (q, ) = e (q, ).
100 CHAPTER 6. DIFFUSION MODES

                    
/ F 4                    
                   
                   
                    
           
2
                    
0
0 1 2 3 4
q/k F

Figure 6.2: Support of the imaginary part of the polarizability e (q, ) for a noninteracting electron
gas.

and k + q is outside the Fermi sphere, and zero otherwise. Taking a parabolic dispersion
relation, k = ~2 k 2 /2m, the delta function then implies = q 2 /2m~ + k q/m~. In general,
one has |k q| mvq/~, where v = ~k/m is the velocity for an electron with wavenumber
k. Since k is inside the Fermi sphere, one has v vF , so that the momentum and frequency
range where Im e is nonzero is given by

q2 q2
vF q < < + vF q. (6.23)
2m~ 2m~

The support of Im e is shown in Fig. 6.2.

6.2 polarizability function for a dirty metal

How do electrons in a disordered metal respond to a change in the potential?
At high frequencies or small wavelengths, we do not expect that the disorder will play
an important role. Hence, for high frequencies or small wavelength we expect that the
polarizability function e is given by the results obtained in the previous section. However,
at low frequencies or large wavelengths, impurity scattering becomes important and should
affect the way the electron density responds to a change of the potential.
As before, let us first examine the Boltzmann equation. Looking at linear response only,
we write f = f 0 + f 1 . For Gaussian white noise, the Boltzmann equation reads

1

(t + vk r (e/~)r k ) fk (r, t) = fk (r, t) f k (r, t) (6.24)

6.2. POLARIZABILITY FUNCTION FOR A DIRTY METAL 101

where f is the average of fk over all directions of k, see Eq. (5.22). Linearizing the Boltzmann
equation and Fourier transforming with respect to r and t, we find

fk (q, ) + i (e/~)vk q( fk0 )(q, )

fk1 (q, ) = . (6.25)
1 + i vk q i

Upon performing an angular average one finds a self-consistent equation for f . Since the
effect of impurities is important only if  1 and ql  1, where l = vF is the mean
free path, we can expand the denominator in Eq. (6.25) for small q and , which greatly
simplifies the calculation. Upon performing the angular average, the leading terms in Eq.
(6.25) cancel. Solving for f from the remaining sub-leading terms, we find

Dq 2
f k (q, ) = e 2 ( fk0 )(q, ), (6.26)
Dq i

where D vF2 /3 is the so-called diffusion constant. Hence, we conclude

2e2 X Dq 2 0 2e2 Dq 2
e (q, ) = ( f k ). = , (6.27)
V k Dq 2 i Dq 2 i

where is the density of states at the Fermi surface.

To see what this result means, it is useful to perform a Fourier transform to time t
and coordinate r. For the density response to a sudden increase of the potential (r, t) =
(r r0 )(t) we find from Eq. (6.27)
  3/2 
2 0 |rr0 |2 /4Dt
e (r, t) = 2e (r r ) e . (6.28)
Dt

This result has a very simple interpretation: the change in the potential at r = r0 creates a
density profile that spreads diffusively.2
The behavior of the polarizability in the limit that both the frequency and the wavevec-
tor q are taken to zero is singular. It depends on which of those is taken to zero first. If
the frequency is taken to zero first, we essentially look at the response to a static potential.
In that case, the fact that the electron motion is slowed down by the impurities plays no
2
Note that the Drude conductivity is a function of the diffusion coefficient as well,

() = 2e2 D .
102 CHAPTER 6. DIFFUSION MODES

(r,t)

electron motion
to screen potential

Figure 6.3: The electrons move periodically in order to screen an external time-dependent poten-
tial. It depends on the frequency of the potential whether the electron gas can sustain this periodic
motion in the presence of disorder.

role. Sooner or later theyll adjust their density profile to accomodate the external poten-
tial, e = 2e2 . This response to an external potential is referred to as Thomas-Fermi
screening. On the other hand, if the wavenumber q is taken to zero first, the presence of
the impurities does not allow the electrons to travel between the minima and maxima of the
external potential within one period 1/, see Fig. 6.3. Hence, the charge density does not
adjust itself, e = 0.
Can the same result be obtained using the diagrammatic method? Repeating the calcu-
lations of Sec. 6.1, we find that the polarizability function e is equal to
2T e2 X
e (q, in ) = Gk+q,k0 +q (im + in )Gk0 ,k (im ), (6.29)
V 0
k,k ,m

where the Green function G is calculated in the presence of the exact impurity potential.
We will be interested in the impurity averaged polarizability function he i. For the disorder,
we use the model of Gaussian white noise, see Sec. 5.5. Evaluating the average, we limit
ourselves to the ladder diagrams shown in Fig. 6.4. The middle part of these diagrams is
a geometric series, and can be summed up separately. The result is called the diffusion
propagator or diffusion D shown in Fig. 6.5,
" #1
1 X
D(q; im , im + in ) = 1 hGk0 +q (im + in )ihGk0 (im )i . (6.30)
2 V k0

We can replace the summation over k0 by an integration over the energy = k0 and the
angle between k0 and q. We are interested in wavevectors q  kF , so that we can expand

k+q = k + vF q cos . (6.31)

Since the main contribution of the -integration comes from a window of width ~/ around
6.2. POLARIZABILITY FUNCTION FOR A DIRTY METAL 103
 
k   
m m m m m m
k k k k k" k

k+q
   = + + + ...

+
m n
   m+ n
k+q
m+ n
k+q k+q
m+ n m+ n
k+q k"+q k+q
m+ n

Figure 6.4: Diagrammatic representation of ladder series for the polarizability function is a disor-
dered non-interacting electron gas.

the Fermi energy, we can assume a constant density of states for the entire integration
range. The result of the -integration depends on the position of the Matsubara frequencies
im + in and im with respect to the real axis. If they are on the same side of the real axis,
the integration over the energy vanishes, and one finds D = 1. If m + n > 0 whereas
m < 0, the -integration is finite, and one has

1 X 1 1 1
Z
hGk0 +q (im + in )ihGk0 (im )i = d cos
2 V 0 2 1 1 + n + ivF q cos
k
= 1 n Dq 2 , (6.32)

where, in the last equality, we expanded for low frequencies, in  1/ and for long
wavelengths Similarly, for m + n < 0 and m > 0, one finds
1 X 1
hGk0 +q (im + in )ihGk0 (im )i = 1 + n vF2 q 2 2 . (6.33)
2 V 0 3
k

We thus conclude that for |n |  1, qvF  1, one has

1 if m + n > 0, m > 0,
1/[ (Dq 2 + n )] if m + n > 0, m < 0,

D(q; im , im + in ) = . (6.34)

1/[ (Dq 2 n )] if m + n < 0, m > 0,
1 if m + n < 0, m < 0,

Performing the analytical continuation im i, im + in + i, we refer to the

four cases listed in Eq. (6.34) as retarded-retarded, retarded-advanced, advanced-retarded,
and advanced-advanced,

D RR (q, ) = 1,
D AR (q, ) = 1/[ (Dq 2 i)],
D RA (q, ) = 1/[ (Dq 2 + i)], (6.35)
D AA (q, ) = 1.
104 CHAPTER 6. DIFFUSION MODES

#"!#"!#"! #"!#"!#"! #!#!#!

#!"#!"!"# #"!#"!!"# #!#!!#
#!"#"!#"! #"!#"!#"! #!#!#!
#!"#!"#!"#"!#"!#"! #!#!#! = 1 + + + ...
#!"#"! #"!#"! #!#!
Figure 6.5: Diagrammatic representation of the diffusion propagator.

The propagators D RA and D AR contain the denominator Dq 2 i , which is the kernel

of the diffusion equation. For that reason, they are referred to as the diffusion propagator:
D RA and D AR are Green functions of the diffusion equation. If no confusion is possible
well refer to the retarded-advanced propagator D RA simply by D. The advanced-retarded
propagator is found by complex conjugation.
In terms of the diffusion propagator, the polarizability reads

2T e2 X
e (q, in ) = hGk+q (im + in )ihGk (im )iD(q; im , im + in ). (6.36)
V k,m

The calculation of e differs in an essential way from the calculation of the diffusion prop-
agator, since Eq. (6.36) involves both a summation over Matsubara frequencies m and a
summation over wavevectors k. The summation over m , which needs to be performed first,
can be done as in Sec. 5.6. Replacing the summation over Matsubara frequencies by an
integration in the complex plane and deforming the contours as in Fig. 5.12, and performing
the analytical continuation in + i, we find

e2
Z X
R
e (q, ) = d tanh(/2T )
2iV k
hGk ()ihGk+q ( + )iD RR (q, ) hGA
 R R R AR
k ()ihGk+q ( + )iD (q, )
A R AR A A AA

+ hGk ( )ihGk+q()iD (q, ) hGk ( )ihGk+q()iD (q, ) .

For the first and fourth terms between the brackets, the diffuson propagator D RR = D AA = 1.
For the second and third term, we write D AR = 1 + (D AR 1). Then there is a contribution
from all four terms between brackets, which is identical to that calculated in Eq. (5.62),
and a contribution from the remaining part of the second and third term between brackets.
Adding both terms and making use of the limits qvF  1,  1 and F  1, we find

e2
 
+
Z
R 2
e (q, ) = 2e d tanh tanh
2iV (Dq 2 i) 2T 2T
6.3. WEAK LOCALIZATION 105
X
hGA R
k ()ihGk+q ( + )i
k
2e2 i
= 2e2
Dq 2 i
2e2 Dq 2
= 2 , (6.37)
Dq i
in agreement with Eq. (6.27).

6.3 Weak localization

Our calculation of the conductivity in Ch. 5 was done to leading order in the impurity
concentration. The leading contribution consisted of diagrams without crossed impurity
lines. Diagrams that have impurity lines are a factor 1/F smaller. However, as well show
here, such diagrams can have a singular temperature dependence in one and two dimensions,
actually diverging in the limit of zero temperature. The singularity arises from the diffusive
electron motion we considered in the previous section.
The leading correction in powers of 1/F is given by the so-called maximally crossed
diagrams.3 The maximally crossed diagrams are shown in Fig. 6.6. They can be represented
in terms of the renormalized current vertex (see Sec. 5.6) and the so-called cooperon prop-
agator. The cooperon is a series of ladder diagrams, similar to the diffusion propagator we
encountered in the previous section. However, unlike for the diffuson, where the difference of
the momenta is constant throughout the ladder, for the cooperon the sum of the momenta
is constant throughout the ladder, see Fig. 6.6. The cooperon propagator derives its name
from its close resemblance to the electron-electron correlation function that describes the
superconducting instability.
The cooperon propagator reads
P
k hG qk (im + in )ihGk (im )i
C(q; im , im + in ) = P .
2V k hGqk (im + in )ihGk (im )i
(6.38)
Performing the momentum summations, we find for qvF  1 and |n |  1
1
C(q; im , im + in ) = [(m )(m + n )].
(Dq 2 + |n |)
(6.39)
3
Dont be confused by the word maximally crossed. These diagrams, in fact, have the smallest number
of crossed impurity lines of all the diagrams that remain after calculation of the Drude conductivity.
106 CHAPTER 6. DIFFUSION MODES

&%$%& &%$%& &%$%& &%$%& &$&

m
k
m
k k
m
$%&%$&%%$&%$&%%$&%$&%%$&%$&%$&$&
m
+ + ... =
1
$%&%&$ $%&%&$ $%&%&$ $%&%&$ $&&$
2 V %&$$ %&$%$ %&$%$ %&$%$ &$$
m+ n %
k+q k+q k+q
m+ n m+ n m+ n %%%%

(%'(%' (%'(%' (%'(%' ('(' k k k

('%('%(%'(%' (%'(%' ('('

(%'('%(%'(%' (%'(%' ('(' = 2 V + + ...
('%(%' (%' ('
qk qk qk

Figure 6.6: Diagrammatic representation of maximally crossed diagrams that contribute to the
conductivity (top). The double lines represent the disorder averaged Green function; the solid dots
refer to the renormalized current vertex, see Fig. 5.10. The maximally crossed diagrams can be
represented in terms of the cooperon propagator (right and below).

The expression is formally identical to the corresponding expression for the diffusion prop-
agator. However, you have to keep in mind that for the Cooperon q is the sum of the
momenta, whereas for the diffuson q is the difference of the momenta.
For the current-current correlator, the diagrams of Fig. 6.6 imply a correction
2e2 T X 0
X
(q 0, in ) = k k
hGk (im )ihGk (im + in )
m2 V 2 0 m k,k
~
C(k + k0 ; im , im + in )hGk0 (im )ihGk0 (im + in )i,
(6.40)
2
The summation over m and the integrations over k and k0 can be done as in the previous
sections. Alternatively, we note that the Green functions have a very weak dependence on m
for |m | < 1/ , so that we may set in i in the arguments of the Green functions. Then
summation over m gives nothing but a factor n = n /2T . Now the analytical continuation
in + i is possible, followed by division by and the limit 0. Also, if we write
k0 = k + Q and note that the mean contribution of the summation over Q comes from
Q close to 0, we can ignore the Q dependence of the single-electron Green functions in Eq.
(6.40). The remaining integration over k then gives
1 X
k (k )hGR 2 A 2 3 2
k,k (0)i hGk,k (0)i = 4 kF /d, (6.41)
V k
6.3. WEAK LOCALIZATION 107

where d is the dimensionality of the conductor. Combining everything, we find

2 e2 kF2 X i
R
(q 0, ) = . (6.42)
dm2 V Q DQ2

Hence, using the general relation D = vF2 /d, we find that the correction to the conductivity
reads

1 R

= (q 0, )
i 0
2 e2 kF2 X 1
=
dm2 V Q DQ2
2e2 X 1
= . (6.43)
V Q Q2

In three dimensions, the summation over Q is divergent for large Q. The reason for the
divergence is that the approximations we made cease to be valid for Q larger than the inverse
mean free path l = vF . Truncating the summation for Ql 1, we find

e2
. (6.44)
l
This correction to the conductivity is a factor kF l smaller than the Drude conductivity.
Hence, for three dimensions, we conclude that the strategy of Sec. 5.6, where we kept dia-
grams without crossed lines only, is valid.
For two dimensions, the summation over Q is divergent both for large Q and for small
Q. The divergence at large Q is removed by truncating the integration at Q 1/l. The
divergence at small Q, however, has an entirely different origin. Small Q correspond to
large length scales, and on large length scales our description in terms of non-interacting
electrons in a static and unbounded environment breaks down. In a real sample, electrons
exit the sample through the contacts, or they loose their phase memory because of scattering
from photons or phonons, or because of electron-electron interactions. That means that
our description, in which such processes are not taken into account, breaks down at long
time scales or long length scales. If we do not want to formulate a full microscopic theory
that includes the contacts and decohering processes, we can simply cut off the momentum
summation from below at Q max(1/L, 1/L ), where L is the system size and L is the
length scale at which the electrons loose their phase memory. Alternative, we keep the
full momentum summation but replace the frequency i in the Cooperon propagator by
108 CHAPTER 6. DIFFUSION MODES

max(1/esc , 1/ ), where esc = L2 /D is the time it takes to exit the sample through the
contacts and = L2 /D is the dephasing time. In both cases we find45

(e2 / 2 ) ln[min(L, L )/l], if d = 2,


= (6.45)
(e2 /)[min(L, L ) l], if d = 1.

These results are very important for our understanding of electron transport through
disordered conductors. On the one hand, you verify that is, at first sight, a factor kF l
smaller than the Drude conductivity. However, has an additional dependence on system
size or temperature that leads to a divergence at large system sizes and low temperatures.
(The dephasing length L if T 0.) Hence, we see that these small corrections become
large in large samples at low temperatures, and that the conductivity becomes significantly
smaller than the Drude conductivity because of the quantum correction . In one and
two dimensions, the correction we calculated is the first signature of the phenomenon of
Anderson localization, the statement that all metals in one and two dimensions become
insulators at sufficiently low temperatures. For that reason, the correction is called weak
localization correction. Moreover, is very sensitive to the size of the conductor or the
temperature (through the low-momentum cut-off) and, as well see soon, to the presence of
a magnetic field, unlike the Drude conductivity , which has at most a weak dependence on
these quantities.
At this point, it is important to stress what is meant by two dimensions and one
dimension in the context of disordered conductors. The meaning is not that the electron
gas has to be strictly two-dimensional or one-dimensional, as is the case, e.g., in the two-
dimensional electron gas formed at the interface in a semiconductor heterostructure, or in
carbon nanotubes. What is meant is that the transverse sample direction (thickness of a
film for a two-dimensional sample or thickness of a wire for a one-dimensional sample) is
much smaller than the relevant large-length cut-off scales, such as L or L .6 In samples of
4
In obtaining the numerical factor for d = 1 and for d = 2, we replaced by i/ in the diffuson
propagator and used = L2 /D to express in terms of L .
5
In three dimensions, the fact that we need to add a lower cut-off to the momentum integration has a
minor effect. It simply amounts to the replacement of Eq. (6.44) by

e2
 
1 1
= 2 .
2 l L

Yet, the temperature and magnetic-field dependence of the conductance correction is dominated by the
low-momentum cut-off.
6
Alternatively, and more appropriately, a sample is referred to as being in a reduced dimension if the
time scale for transverse diffusion is small in comparison to the relevant large-time cut-off scales, such as the
dwell time, the time it takes to exit from the sample, or the dephasing time . Whereas this definition
6.3. WEAK LOCALIZATION 109

reduced dimension, the smallest wavevector in the transverse direction is Q = 0. The next
wavevector is already so large that the corresponding length scale 2/Q is larger than the
low-wavevector cut-off. Although this does not mean that these wavevectors are excluded
from the summation (6.43) after all, they are likely to be smaller than the high-wavevector
cut-off 1/l , their contribution to is non-singular and independent of the sample size
L, the dephasing length L , or any other relevant experimental parameter.
The weak localization correction has a simple physical explanation. In a semiclassical
picture, the electrons are described as wave packets that move along well defined trajectories,
acquiring a phase as they move along. Hence, the electron propagation along a certain path
is described by an amplitude, i.e., by an intensity and a phase. The probability for an
electron to go from one location in the sample to another location is proportional to the
square of the amplitudes for all trajectories linking start and finish. Without quantum
phase coherence, one would add intensities of all trajectories, not amplitudes. In other
words, without quantum phase coherence, one squares the amplitudes first and then adds
them. The difference between the two descriptions add squared amplitudes or square
added amplitudes is interference. Interference can be constructive or destructive. In
most cases, the interference correction contains a sum over many terms with different phases,
and averages to zero. However there is one exception: interference of trajectories that return
to the point of departure. In this case, time-reversed trajectories acquire the same phase,
and, hence, interfere constructively, see Fig. 6.7. Thus, the probability to return to the point
of departure is enhanced by quantum interference. This also implies that the rate at which
particles diffuse through the metal is decreased, which, in turn, leads to a reduction of the
conductivity. This reduction is precisely what we calculated above!
The interference between time-reversed trajectories is described by the Cooperon prop-
agator. Recall that the Cooperon propagator was most important precisely for the case of
opposite momenta k and k0 .
In this picture, the cut-offs needed for the calculation of the weak localization correction
are more transparent. The large momentum cut-off corresponds to the break down of dif-
fusion theory at length scales below the mean free path. Indeed, since one needs impurity
scattering in order to return to the point of departure, there is a minimal length for such
a trajectory. Also, the interference correction exists only if the electrons keep their phase
memory along the trajectory and if the electrons return to the point of departure at all. This
explains the cut-off at low momenta at 1/L or 1/L, respectively.
is the same as the length-based definition given in the text for samples with a simple geometry and extended
contacts, it can more readily be generalized to samples of arbitrary shape and to samples with point contacts
or tunneling contacts. In particular, it allows for the definition of a zero dimensional sample: a sample in
which the time it takes to diffuse through the entire sample is small in comparison to all other relevant time
scales.
110 CHAPTER 6. DIFFUSION MODES

Figure 6.7: Constructive interference of two time-reversed trajectories returning to the point of
departure.

In a magnetic field B, time-reversed trajectories acquire a phase difference equal to 2

times the enclosed flux divided by the flux quantum 0 = hc/e. If the typical flux
accumulated by a trajectory that ventures a distance L away from the point of departure
is denotes (L), trajectories of length L such that (L) > 0 do not contribute to the
interference. Hence, in a magnetic field, we should impose a lower momentum cutoff at
Q 1/lm , where lm is the so-called magnetic length, defined using the criterion (lm ) = 0 .
If lm is smaller than the dephasing length L and the system size, the dephasing length
L should be replaced by LB in Eq. (6.45). Hence, for large magnetic fields, the weak
localization correction to the conductivity is suppressed. In a three-dimensional sample
or in a two-dimensional sample with the magnetic field perpendicular to the plane of the
sample, the typical area swept out for a trajectory that ventures a distance L away from the
point of departure is of order L2 , hence lm (0 /B)1/2 . In a film of thickness a with the
magnetic field in the plane of the film or in a wire of thickness a, this typical area is La,
hence lm 0 /Ba.

6.4 Formulation in real space

In Sec. 6.2 we saw that the diffuson is equal to the pole of the diffusion equation. We arrived
at this conclusion from an analysis in momentum representation, which, in principle, is valid
6.4. FORMULATION IN REAL SPACE 111

,*,*-*-*,, 1-*-*,, R--,, 1

)*)*)*)*++1 )*)*++ ))++
,*,*-*-*,, -*-*,, --,, )*)*)*)*++ )*)*++ ))++
,*,*-*-*,, -*-*,, --,, = 2 + + ... = 2 + )*)*)*)*++ )*)*++ ))++
,*-*, -*, -, h h )*)*+ )*+ )+
2A 2 2

Figure 6.8: Defining equation for the diffusion propagator.

for a bulk system only. For finite-sized systems or for inhomogeneous systems, a description
in real space is more useful. In this section, we rederive the results of Sec. 6.2 using the
coordinate and time representation.
Before embarking on a diagrammatic calculation in coordinate representation, let us recall
the expressions for the impurity averaged Green function,
m i(kF +/vF )|rr0 ||rr0 |/2l
hGR (r, r0 ; )i = e ,
2~|r r0 |
m 0 0
hGA (r, r0 ; )i = 0
ei(kF +/vF )|rr ||rr |/2l . (6.46)
2~|r r |

Fourier transforming to time, one has

m 0 0
hGR (r, r0 , t, t0 )i = 0
eikF |rr ||rr |/2l (t t0 |r r0 |/vF ),
2~|r r |
m ikF |rr0 ||rr0 |/2l
hGA (r, r0 ; t, t0 )i = e (t0 t |r r0 |/vF ). (6.47)
2~|r r0 |

We also recall the averaging rule for the Gaussian white noise potential,

~
hU (r)U (r0 )i = (r r0 ), (6.48)
2
where is the elastic mean free time and is the density of states per spin direction.
Equations (6.46) and (6.47) are valid in three dimensions; similar results can be obtained
in one and two dimensions. Throughout this discussion we assume weak disorder, kF l  1,
where l = vF / is the elastic mean free path.
Starting point of our discussion is the diagrammatic expression for the diffusion propa-
gator, which we repeat in Fig. 6.8. In principle, the diffusion propagator has four coordinate
and four time arguments. However, because of the special structure of the Gaussian white
noise average, the spatial arguments are pairwise equal. Hence, we may write the arguments
112 CHAPTER 6. DIFFUSION MODES

of the diffusion propgator as D(r, r0 ; t1 , t01 ; t2 , t02 ). With this notation, using the coordinate
and time representation, the defining expression in Fig. 6.8 reads
1
Z
0 0 0 0 0 0
D(r, r ; t1 , t1 ; t2 , t2 ) = (r r )(t1 t1 )(t2 t2 ) + dr00 dt001 dt002 hGR (r, t1 ; r00 , t001 )i
2 ~
D(r00 , r0 ; t001 , t01 ; t002 , t02 )hGA (r00 , t002 ; r, t2 )i (6.49)
Because of the delta-function time dependence of the impurity-averaged Green functions,
the solution D(r, r0 ; t1 , t01 ; t2 , t02 ) of Eq. (6.49) will be proportional to (t1 t01 + t02 t2 ). In
coordinate representation, the disorder-averaged Green functions are short ranged, see Eq.
(6.46). Anticipating that D will be a slowly varying function of it arguments (apart from
the delta-function dependence noted above), one might be tempted to replace the product
of single-electron Green functions by a delta functions in space and in time. In that case,
one finds
1
hGR (r, t1 ; r00 ; t001 )ihGA (r00 , t002 ; r, t2 )i = (r r00 )(t1 t001 )(t2 t002 ). (6.50)
2 ~
Using Eq. (6.50) for the product of single-electron Green functions, the diffusion propagator
cancels from Eq. (6.49). Hence, we need to go beyond the simple delta-function approxima-
tion used in Eq. (6.50). Taking the spatial and temporal dependence of the Green functions
into account to leading order, one has
1
hGR (r, t1 ; r00 ; t001 )ihGA (r00 , t002 ; r, t2 )i = (1 + D r2 (t1 + t2 ))
2 ~
(r r00 )(t1 t001 )(t2 t002 ), (6.51)
where D = (1/3)vF2 is the diffusion constant.7 Note that the temporal derivatives appear
as a sum, so that there is not contribution from the delta function (t1 t01 + t02 t2 ) in the
diffusion propagator. In this notation, the diffusion propagator satisfies the equation
1
Dr2 + t1 + t2 D(r, r0 ; t1 , t01 ; t2 , t02 ) = (r r0 )(t1 t01 )(t2 t02 )


(6.52)

This is precisely the diffusion equation. Note that Eq. (6.52) contains a derivative to the
sum t1 + t2 only. Hence, in solving Eq. (6.52) one can consider all time differences and
the sum t01 + t02 as parameters. For a bulk system, Fourier transform of Eq. (6.52) gives
D(q; ) = [ (Dq 2 + i)]1 , in agreement with the results of Sec. 6.2.
7
Equation (6.51) is derived under the integral sign, considering the integral of
hG(r, t1 ; r00 ; t001 )ihG(r00 , t002 ; r, t2 )i times an arbitrary function f (r00 , t001 , t002 ) over r00 , t001 , and t002 . Expand-
ing the function f in a Taylor series around r00 = r, t001 = t1 , and t002 = t2 then gives the desired
result.
6.4. FORMULATION IN REAL SPACE 113
. 0/
0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0.
0./0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0.
0./0/ . 0/ . 0/ . insulator . 0/
0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0.
0./0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0.
0./0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0.
./
0 ./ 0 ./ 0 ./ 0 ./ 0 ./ 0 ./ 0 ./ 0 ./ 0 ./ 0 ./ 0 ./ 0 ./ 0 ./ 0 ./ 0 ./ 0 .0 r
0./0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . r0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0.
0./0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . r0/ . 0/ . 0/ . 0/ . 0/ . 0.
0./0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0.
0./0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0.
0./0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0. conducting
0./0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0.
0./0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0. sample
0./0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0.
0./0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0/ . 0.
Figure 6.9: Relative positions of a point r close to the sample boundary and its mirror image r r .
For the derivation of the boundary conditions for the diffusion propagator, the point r is chosen
well within a mean free path from the boundary, but much further away than a Fermi wavelength.

The boundary conditions for the diffuson propagator can be obtained in a similar way. If
we fix the point r0 to be inside the sample, the boundary conditions are found by taking the
correct form of the disorder averaged single-electron Green functions at the sample boundary.
Since the wavefunctions have to vanish at the sample boundary, you verify that one has

hG(r, r00 )i = hG(r, r00 )ibulk hG(rr , r00 )ibulk , (6.53)

where rr is the mirror image of r in the sample boundary, see Fig. 6.9, and the subscript
bulk indicates the expression for the average Green function deep inside the sample, see
Eqs. (6.46) and (6.47) above. Since we are interested in spatial variations of the diffusion
propagator on length scales of the mean free path l and beyond, we take the point r such
that its distance from the boundary is much smaller than the mean free path l but much
larger than the Fermi wavelength F = 2/kF . Then, using Eq. (6.47), we find

1
hGR (r, t1 ; r00 , t01 )ihGA (r00 , t002 ; r, t2 )i = (1 + DvF1 n
r )(r r00 )(t1 t001 )(t2 t002 ),
2 ~
(6.54)

where n is the outward oriented unit vector normal to the sample boundary. Hence, using
Eq. (6.49), we conclude that, at the sample boundary, the diffusion propagator must satisfy
the boundary condition

r D(r, r0 ; t1 , t01 ; t2 , t02 ) = 0.

n (6.55)
114 CHAPTER 6. DIFFUSION MODES

Similarly, at points where the sample is connected to a perfect conductor, one has the
boundary condition
D(r, r0 ; t1 , t01 ; t2 , t02 ) = 0. (6.56)

The boundary condition (6.55) describes the fact that, at a boundary with an insulator, the
current density is zero, whereas the boundary condition (6.56) describes the fact that, at a
boundary with a perfect conductor, the particle accumulation must vanish.
In the presence of a vector potential A and a scalar potential that vary on a spatial and
temporal scales slow compared to the mean free path and the mean free time, the retarded
and advanced Green functions (6.46) acquire an extra factor exp[ie|r r0 | + i(eA/c~)
(r r0 )] and exp[ie|r r0 | + i(eA/c~) (r r0 )], respectively. The effect of these factors is
to replace the derivatives in Eq. (6.52) and (6.55) by covariant derivatives. Hence, we find
"  2 #
ie
D r (A(r, t1 ) A(r, t2 )) + t1 + t2 + ie(r, t1 ) ie(r, t2 )
~c
1
D(r, r0 ; t1 , t01 ; t2 , t02 ) = (r r0 )(t1 t01 )(t2 t02 ), (6.57)

whereas the derivative r in the boundary condition (6.55) is replaced by the covariant
derivative r (ie/~c)(A(r, t1 ) A(r, t2 )).
The same considerations apply to the Cooperon. Using the space and time representation,
the defining equation for the Cooperon reads
1
Z
0 0 0 0 0 0
C(r, r ; t1 , t1 ; t2 , t2 ) = (r r )(t1 t1 )(t2 t2 ) + dr00 dt001 dt002 hGR (r, t1 ; r00 , t001 )i
2 ~
A 00 00 00 0 00 0 00 0
hG (r, t2 ; r , t2 )iC(r , r ; t1 , t1 ; t2 , t2 ) (6.58)
From this equation, one concludes that the Cooperon propagator C(r, r0 ; t1 , t01 ; t2 , t02 ) is pro-
portional to (t1 t01 + t2 t02 ). Replacing the product of retarded and advanced Green
functions by the appropriate sum of a delta functions and derivatives of delta functions, one
finds that the Cooperon then satisfies the equation
"  2 #
ie
D r (A(r, t1 ) + A(r, t2 )) + t1 t2 + ie(r, t1 ) ie(r, t2 )
~c
1
C(r, r0 ; t1 , t01 ; t2 , t02 ) = (r r0 )(t1 t01 )(t2 t02 ), (6.59)

The boundary conditions are the same as those for the diffuson propagator. In the presence
of a magnetic field, the differential operator r is replaced by the covariant derivative r
6.5. RANDOM MATRIX THEORY 115

(ie/~c)(A(r, t1 ) + A(r, t2 )). Note that Eq. (6.59) contains a derivative to the time difference
t1 t2 . In solving Eq. (6.52) one can consider all time sums and the difference t01 t02 as
parameters.
A time-independent magnetic field, which is represented by a time-independent vector
potential, has no effect on the diffuson, see Eq. (6.57). This is understandable, since the
diffusion propagator, which governs the diffusive spreading of charge in a disordered metal,
should not be affected by the presence of a magnetic field. On the other hand, the Cooperon,
which contains information about the interference of time-reversed trajectories, depends
strongly on a magnetic field. How strong this dependence is one can see by realizing that Eq.
(6.59) is formally equivalent to the defining equation for the imaginary-time Green function
of a particle of mass ~/D and charge 2e. This mass is much smaller than the electron mass
if the disorder is weak, kF l  1, so that the effect of a magnetic field is unusually strong. A
very small magnetic field will significantly affect the Cooperon propagator and destroy the
interference leading to weak localization.

6.5 Random matrix theory

Now let us look at an isolated piece of a disordered conductor of finite size. We are interested
in the density of states in the conductor. Since the sample has a finite size, the density of
states will be a sum of Dirac delta functions,
X
N () = 2 ( ), (6.60)

where labels the eigenvalues of the single-electron Hamiltonian (without spin) and the
factor two accounts for spin degeneracy.
The positions of the energy levels and, hence, the density of states, will depend on the
precise impurity configuration. As before, well take a statistical approach, and calculate the
probability distribution of the density of states.
Calculation of the average is straightforward, hN i = 2V , where V is the volume of the
sample. The more interesting question is that of the fluctuations of the density of states,
which is described by the function

R(1 2 ) = hN (1 )N (2 )i hN (1 )ihN (2 )i. (6.61)

In order to calculate R, we first express the density of states in terms of the electron Green
functions,
1
Z
N () = dr[GA (r, r; ) GR (r, r; )]. (6.62)
i
116 CHAPTER 6. DIFFUSION MODES

213321 i321321 3131 2 267726 i726726 7676 2

r 31232
3123121321321 313131 r 76272
7627626726726 767676
r 32
3211 321321 3131 r 72
7266 726726 7676
i1 i1

r r r r
1 2 1 2
5244252442544
i1
9288292882988
i1
r52
5244 525244 5544 r92
9288 929288 9988
542542524524 5454 982982928928 9898
r54252
524 5244 5454 r98292
928 9288 9898
i2 i2

Figure 6.10: Diagrams contributing to the fluctuations of the density of states in a disordered
metal grain.

With this expression, R(1 2 ) can be calculated with the help of diagrammatic pertur-
bation theory. The relevant diagrams for R(1 2 ) are shown in Fig. 6.10. (Note that the
diagrams without any lines connecting the inner and outer electron lines contribute to the
average of the density of states and not to the fluctuations.)
After substitution of Eq. (6.62) into Eq. (6.61), we distinguish four contributions to
the density of states fluctuations. Performing the same calculation as the one leading to
Eq. (6.50), one finds that the contributions involving two retarded Green functions or two
advanced Green functions are smaller than the contributions involving one retarded and one
advanced Green function by a factor kF l  1. Hence, we conclude
2
Z
R(1 2 ) = 2 Re dr1 dr2 hGR (r1 , r1 , 1 )GA (r2 , r2 , 2 )i. (6.63)

Making use of the auxiliary results
GR (r1 , r0 , 1 )GR (r00 , r1 , 1 )GA (r0 , r00 , 2 ) = 2i 2 (r0 r1 )(r00 r1 )
GR (r1 , r0 , 1 )GR (r00 , r1 , 2 )GA (r00 , r0 , 2 ) = 2i 2 (r0 r1 )(r00 r1 ),
(6.64)
we find that, according to the diagrams of Fig. 6.10,
2 2
Z
R(1 2 ) = Re drdr0 [D RA (r, r0 ; 1 , 2 )D RA (r0 , r; 1 , 2 )
2
6.5. RANDOM MATRIX THEORY 117

+ C RA (r, r0 ; 1 , 2 )C RA (r0 , r; 1 , 2 )]. (6.65)

We now calculate the density of states correlator at energy 1 and vector potential A1
and at energy 2 and vector potential A2 . Hereto, we write the diffuson and cooperon
propagators as a sum over eigenmodes of the diffusion equation,
 2
ie
D r + (A1 ()A2 ) m (r) = m m (r), (6.66)
~c

where the m are properly normalized eigenfunctions and the m are the corresponding
eigenvalues,

1 X m+ (r)m+ (r0 )
D RA (r, r0 ; 1 , 2 ) = , (6.67)
m m+ i(1 2 )
1 X m (r)m (r0 )
C RA (r, r0 ; 1 , 2 ) = . (6.68)
m m i(1 2 )

Substituting this into Eq. (6.65) we find

2 X 1 1

R(1 2 ) = 2 Re 2
+ 2
. (6.69)
m
( m+ i( 1 2 )) ( m i( 1 2 ))

This result, which was first obtained by Altshuler and Shklovskii [Sov. Phys. JETP 64,
127 (1986)], is a key result in the study of spectral statistics in small disordered conductors.
Let us look at it in more detail. First, we consider the case in which there is no magnetic
field. Then, the cooperon and diffuson contributions are equal. Finding an exact expression
for R(1 2 ) requires solving the diffusion equation, which may be difficult in a sample
with an irregular shape. However, we can make three general statements: First, all m are
non-negative. Second the smallest eigenvalue of the diffusion equation is 0 = 0. Third, the
second smallest eigenvalue 1 is of order D/L2 , where L is the sample size. (The length L is
the longest dimension of the sample for a sample with an anisotropic shape.) The smallest
eigenvalue 1 in the absence of a magnetic field is known as the Thouless energy, ETh .
Hence, if |1 2 |  ETh the sum in Eq. (6.69) is dominated by the m = 0 term, and one
has
4
R(1 2 ) = 2 if no magnetic field. (6.70)
(1 2 )2
Notice that this result is completely universal: it does not depend on sample size, sample
shape, or disorder concentration.
118 CHAPTER 6. DIFFUSION MODES

If, on the other hand, there is a finite magnetic field, the lowest eigenvalue 0 for
the cooperon propagator is nonzero. For small magnetic fields, one can find 0 using
perturbation theory,  2
4De2 e ETh
Z
2
0 = 2 2 dr|A| , (6.71)
~c hc
where = 1/V is the mean spacing between energy levels and the vector potential A is
taken in the London gauge (r A = 0 inside the sample and n r A = 0 on the sample
boundary, where n is the unit vector perpendicular to the sample boundary). Equation
(6.71) is valid as long as 0  ETh . Hence, for magnetic fields for which 0  |1 2 |,
the cooperon contribution to the density of states fluctuations is suppressed, and one finds
2
R(1 2 ) = with magnetic field. (6.72)
2 ( 1 2 )
2

The results (6.70) and (6.72) are divergent if 1 2 0. This divergence appears
because of a levels self correlation: Looking at the definitions (6.60) and (6.61), one sees
that the correlator R(1 2 ) contains the singular term
X
Rself (1 2 ) = 4 [h(1 )(2 )i h(1 )ih(2 )i]

4 4
= (1 2 ) 2 . (6.73)

The origin of the divergence in our diagrammatic calculation is that the diagrammatic per-
turbation theory breaks down when the energy difference 1 2 becomes comparable to
the spacing = V / between energy levels in the sample and thus fails to resolve the delta-
function correlations of Eq. (6.73). One way to cure this problem is to give all levels a finite
width , which regularizes the divergence arising from the self correlations. This amounts
to replacing Eqs. (6.70) and (6.72) by
4 1
R() = 2
Re
( i)2
4 2 2
= , (6.74)
2 ( 2 + 2 )2
where = 2 with a magnetic field and = 1 without a field. Using field-theoretic methods,
Efetov has been able to calculate the exact correlation function R() down to = 0 [Adv.
Phys. 32, 53 (1983)]. Without magnetic field he found
Z
4 sin2 (/)
 
4 sin(/) sin(x)
R() = 2 2 dx for 6= 0. (6.75)
2 / x
6.5. RANDOM MATRIX THEORY 119

With a magnetic field, only the first term on the r.h.s. of Eq. (6.75) is kept. Note that Eq.
(6.75) simplifies to our results (6.70) and (6.72) if ||  .
One striking consequence of the correlator (6.70) and (6.72) is that the spectrum is rigid.
This means that the fluctuations of the number of levels N (2 , 1 ) in a certain energy interval
1 < < 2 increases slower with |1 2 | than |1 2 |1/2 . (Fluctuations proportional
to |1 2 |1/2 are expected if the levels were distributed as uncorrelated random numbers.)
Indeed, one finds

Z 2
var N (2 , 1 ) = dd 0 R( 0 )
1
4 (1 2 )2
ln , (6.76)
2 2

where we used the regularized result (6.74) with to cut off the divergence of R() at
small .
The universality of the spectral fluctuations for energies below the Thouless energy is
more general than the case of a disordered conductor of arbitrary shape we considered here.
In fact, one finds the same spectral fluctuations for the energy levels in a ballistic conductor
with an irregular shape, for the resonances in a heavy nucleus, and for the eigenvalues
of a hermitian matrix with randomly chosen elements.8 In these cases, the only relevant
input is the average spacing between levels and the symmetry of the system: without
a magnetic field, time-reversal symmetry is present, whereas with a magnetic field, time-
reversal symmetry is broken. For hermitian matrices, the presence or absence of time-
reversal symmetry corresponds to real or complex matrix elements. Mathematically, random
matrices are much easier to deal with than disordered conductors. The study of eigenvalues
and eigenvectors of random matrices has grown into a field of its own, called random matrix
theory. In fact, the label random matrix theory is also used to describe the universal
aspect of spectral statistics of disordered metals and heavy nuclei within an energy window
of width  ETh . You can find more about the mathematical aspects of the theory of random
matrices in the book Random matrices by M.L. Mehta (Academic, New York, 1991).

8
For a ballistic conductor, the Thouless energy is ETh vF /L, whereas for a random matrix of size N
one has ETh N . For a ballistic conductor, R(1 2 ) for |1 2 |  vF /L does not depend on the
precise sample shape as long as the shape is irregular, whereas for a random matrix R( 1 2 ) does not
depend on the distribution of the matrix elements if |1 2 |  N . For a ballistic sample, the universality
of spectral statistics has the status of a conjecture, although there is extensive numerical evidence.
120 CHAPTER 6. DIFFUSION MODES

6.6 Exercises
Exercise 6.1: Tunneling density of states

In this exercise we consider two pieces of metal that are weakly coupled by a tunnel barrier,
see Fig. 6.11. The two pieces of metal are labeled A and B. In the tunneling Hamiltonian
formalism, the total Hamiltonian H is written as
A + H
H=H B + H
AB , (6.77)

where H A and H B operate on states of each system separately, with creation and annihilation
operators cA , cA , and cB , cB for the single-electron states in A and B, and H
AB is a
tunneling Hamiltonian H AB that describes the coupling between the two systems,

AB =
X
H (T cA cB + T

cB cA ), (6.78)

where, in principle, the indices and refer to any choice of an orthogonal basis for the
single-electron states in the two systems A and B. The tunnel matrix elements T are
complex number numbers that depend on the extension of the single-electron wavefunctions
in A and B into the insulating region between them. In general, the tunnel matrix elements
are exponentially small in the separation between A and B.
The current through the tunnel barrier follows from the rate of change of the charge QA
in metal A (or metal B) as
QA X
I= = i[H, QA ] , QA = e cA, cA, . (6.79)
t

Substituting Eqs. (6.77) and (6.78) for H, one finds that only the tunneling Hamiltonian
contributes to the commutator, and that
X
I = ie (T cA cB T

cB cA ). (6.80)

The current through the tunnel barrier is linear in the tunneling matrix elements T . In
order to calculate I to lowest (second) order in the tunneling matrix elements, it is sufficient
to calculate the (change of the) current to linear response in the tunneling matrix elements.
The Kubo formula then gives
Z
I(t) = dt0 CI,H
R
AB
(t, t0 ), (6.81)

6.6. EXERCISES 121
=>;
>;
=>=;>; =>;
=>; =>;
>;
=>; =>;
>;
=>; =>;
>;
=>; =>;
>;
=>; =>;
>;
=>; =>;
>;
=>; >=>= @; ?@;
?@; ?@;
@;
?@; ?@;
@;
?@; ?@;
@;
?@; ?@;
@;
?@; ?@;
@;
?@; ?@;
@;
?@; ?@;
@;
?@; ?@;
@;
?@; @?@?
>=; = =>;
>; = =>;
>; = =>;
>; = =>;
>; = =>;
>; = =>;
>; = =>;
>; = > >=>= @; ? ?@; ? ?@;
@; ? ?@;
@; ? ?@;
@; ? ?@;
@; ? ?@;
@; ? ?@;
@; ? ?@;
@; ? ?@;
@; @?@?
>=;
>=; = =>; = =>; = =>; = =>; = =>; = =>; = =>; >=>= @; ? ?@; ? ?@; ? ?@; ? ?@; ? ?@; ? ?@; ? ?@; ? ?@; ? ?@; @?@?
=>=;>;
>; =>;
=>;
>;
= =>;
>;
>;
= =>;
>;
>;
= =>;
>;
>;
= =>;
>;
>;
= =>;
>;
>;
= =>;
>; >=>= @; ? ?@;
@;
? ?@;
@;
@;
? ?@;
@;
@;
? ?@;
@;
@;
? ?@;
@;
@;
? ?@;
@;
@;
? ?@;
@;
@;
? ?@;
@;
@;
? ?@;
@; @?@?
>=;
>=; = =>;
>; = =>;
>; = =>;
>; = =>;
>; = =>;
>; = =>;
>; = =>;
>; >=>= @; ? ?@; ? ?@;
@; ? ?@;
@; ? ?@;
@; ? ?@;
@; ? ?@;
@; ? ?@;
@; ? ?@;
@; ? ?@;
@; @?@?
>=;
==;
>; = ==; = ==; = ==; = ==; = ==; = ==; = ==; >==> @; ? ??; ? ??; ? ??; ? ??; ? ??; ? ??; ? ??; ? ??; ? ??; @?@?
>>=;>; >>;
=>;
>;
> =>;
>;
>;
> =>;
>;
>;
> =>;
>;
>;
> =>;
>;
>;
> =>;
>;
>;
> =>;
>; ;
: <:;
>: =>= @; @ ?@; ;
: <:;
@: ;
@;
@ ?@; : <:<: @;
@;
@ ?@;
@;
@ ?@;
@;
@;
@ ?@;
@;
@;
@ ?@;
@;
@;
@ ?@;
@;
@;
@ ?@;
@;
@;
@ ?@;
@; ?@@?
>=;
=; = = = = = = = < <; ? < <; ? <:<: ?; ? ? ? ? ? ? ? @??@
=>=;=;
>>; >>;
=>;
=>;
=;
>>;
= =>;
>;
=;
>>;
= =>;
>;
=;
>>;
= =>;
>;
=;
>>;
= =>;
>;
=;
>>;
= =>;
>;
I =; >>;
= =>;
>; : ;
< : <;
:
=>>= ?;
>=>= @;
@@;
? ?@; : ;
< : <;
:
?;
@@;
? ?@;
@; <<:: @;
@@;
? ?@;
?;
@@;
? ?@;
@;
?;
@@;
? ?@;
@;
?;
@@;
? ?@;
@;
?;
@@;
? ?@;
@;
?;
@@;
? ?@;
@;
?;
@@;
? ?@;
@; @?@?
>=;
>=; = =>;
>; = =>;
>; = =>;
>; = =>;
>; = =>;
>; = =>;
>; = =>;
>; ;
< : <;
: >=>= @; ? ?@; ;
< : <;
: ? ?@;
@; <:< @; ? ?@; ? ?@;
@; ? ?@;
@; ? ?@;
@; ? ?@;
@; ? ?@;
@; ? ?@;
@; @?@?
>=;
=>=;>;
>; = =>; = =>;
>; = =>;
>; = =>;
>; = =>;
>; = =>;
>; = =>;
>; ;
< >=>= @; ? ?@; ;
< ? ?@;
@; ? ?@;
@; ? ?@;
@; ? ?@;
@; ? ?@;
@; ? ?@;
@; ? ?@;
@; ? ?@;
@; @?@?
>=; = =>;
>; = =>;
>; = =>;
>; = =>;
>; = =>;
>; = =>;
>; = =>;
>; >=>= @; ? ?@; ? ?@;
@; ? ?@;
@; ? ?@;
@; ? ?@;
@; ? ?@;
@; ? ?@;
@; ? ?@;
@; ? ?@;
@; @?@?
>=;
>=; = =>; = =>; = =>; = =>; = =>; = =>; = =>; >=>= @; ? ?@; ? ?@; ? ?@; ? ?@; ? ?@; ? ?@; ? ?@; ? ?@; ? ?@; @?@?
=>=;>;
>; =>;
=>; >A;
>;
= =>;
>;
= =>;
>;
>;
= =>;
>;
>;
= =>;
>;
>;
= =>;
>;
>;
= =>;
>; >=>= @; ? ?@;
@;
? ?@;
@;
@;
? ?@;
@;
@;
? ?@;
@;
@;
? ?@;
@;
@;
? ?@;
@;
@;
? ?@;
@;
@;
? ?@;
@; @B;
@;
? ?@; @?@?
>=;
>=; = =>;
>; = =>;
>; = =>;
>; = =>;
>; = =>;
>; = =>;
>; = =>;
>; >=>= @; ? ?@; ? ?@;
@; ? ?@;
@; ? ?@;
@; ? ?@;
@; ? ?@;
@; ? ?@;
@; ? ?@;
@; ? ?@;
@; @?@?
>=;
=>=;>; = =>; = =>; = =>; = =>; = =>; = =>; = =>; ? ? ? ? ? ? ? ? ? @?
>; = >; = >; = >; = >; = >; = >; = >=>= @; ?@;
? @; ?@;
? @; ?@;
? @; ?@;
? @; ?@;
? @; ?@;
? @; ?@;
? @; ?@;
? @; ?@;
? @?@?

V
Figure 6.11: Setup for tunnel spectroscopy. Two metals, labeled A and B, are separated by an
insulating material, e.g., an oxide or simply vacuum. A battery maintains a finite difference between
the chemical potentials of A and B.

R
where the correlation function CI,H AB
is defined as

R
CI,H AB (t0 )] i.
(t, t0 ) = i(t t0 )h[I(t), H (6.82)
AB

Here, the thermal equilibrium average h. . .i and the time-dependence of the tunneling Hamil-
tonian H AB are taken with respect to the Hamiltonian H A + H
B only, i.e., without inclusion
of the tunneling Hamiltonian H AB .
In the average (6.82), the two systems are completely separated. Hence, in thermal
equilibrium, their chemical potentials A and B need not be equal. In fact, it is only if
B A = eV 6= 0 that a nonzero current flows between the reservoirs. Even in the presence
of the tunneling current, a quasi equilibrium in which B A = eV 6= 0 can be maintained
with the help of a battery, as long as the contacts between the battery and the metals A and
B have a much smaller resistance than the tunnel barrier between A and B.

(a) Calculate the correlator CI,HAB and show that the tunneling current can be written

d X
Z
I = e |T |2 AA, ()AB, ( + eV ) (nF ( + eV ) nF ()) , (6.83)
2
122 CHAPTER 6. DIFFUSION MODES

where nF () = 1/(1 + exp(( )/T )) is the Fermi distribution function. Be sure that
your derivation makes no specific assumptions about the Hamiltonians H A and HB !
Equation (6.83) is valid both with and without electron-electron interactions.
(b) If metal B has a spectral density that is more or less constant,
X
|T |2 AB, () const., (6.84)

show that the tunneling current can be used as a direct measurement of the spectral
density of metal A. This method of measuring the spectral density is known as tunnel
spectroscopy.

Exercise 6.2: Polarizability function

In this exercise, we consider the polarizability function for the non-interacting electron gas
at zero temperature
0 2e2 X ( k+q ) ( k )
e (q, ) = (6.85)
V k k+q + k + i
2e2 X
 
1 1
= .
V k<k k+q + k + i k + kq + i
F

(a) Perform the summation over k to show that the real part of e is
Re e (q, ) = 2e2 (f (x, x0 ) + f (x, x0 )) , (6.86)
where is the density of states at the Fermi level per spin direction and per unit
volume, x = q/2kF , x0 = /4F , and
1 (1 (x0 /x x)2 )  x + x2 x0 
 
f (x, x0 ) = + ln  .
4 8x x x2 + x0 
(b) The expressions for Im e are somewhat more complicated. Show that if q < 2kF and
0 one finds

2 4x0 /x if 0 < x0 < x x2 ,
e
Im e (q, ) = (1 (x0 /x x)2 ) if x x2 < x0 < x + x2 ,
8x
0 if x0 > x + x2 .
(6.87)
6.6. EXERCISES 123

The dimensionless quantities x and x0 were defined above. On the other hand, if
q > 2kF , one finds
2 2 2
2 (1 (x0 /x x) ) if x x < x0 < x + x,
e
Im e (q, ) = 0 if 0 < x0 < x2 x
8x
or if x0 > x + x2 .
(6.88)

Exercise 6.3: Conductivity from polarizability function

Whereas the polarizability function e (q, ) measures the charge response to a time-dependent
shift of the scalar potential, the conductivity (q, ) measures the current response to a
time-dependent electric field. However, charge and current are related, as well as electric
field and scalar potential. Hence, one should be able to calculate the conductivity from the
polarizability function.
Charge and current are related via the continuity relation, t e + r j = 0. Upon Fourier
transformation, the continuity equation reads

e (q, ) = q j(q, ).

The electric field and the scalar potential are related as

E(q, ) = iq(q, ).

(a) Express the longitudinal current component q 2 q(q j) in terms of the polarizability
function and the electric field E(q, ).

(b) For an isotropic system, we expect that the current density j is parallel to the electric
field E. For electric fields generated by a scalar potential, E is always parallel to q.
Use this, and your answer to (a) to calculate the dc conductivity . Take care to take
the limits 0 and q 0 in the correct order!

(c) Derive a general relation between the polarizability e (q, ) and the conductivity
(q, ).

Note that the restriction that E and q are parallel did not appear when the conductivity
was calculated using the time-dependent vector potential.
124 CHAPTER 6. DIFFUSION MODES

Exercise 6.4: Magnetic-field dependence of weak localization

In the presence of a magnetic field, the expression we found for the weak-localization cor-
rection to the conductivity, Eq. (6.43), should be supplemented with a cut-off for small
wavevectors Q. In the text, we argued that this cut-off should be Q 1/lm , where lm is the
so-called magnetic length. In this exercise, youre asked to give a more rigorous calculation.
Starting point of our analysis is the expression of the conductivity correction in terms of
the Cooperon propagator C(Q; im , im + in ), which we write as

e2 D X
= 2 C(Q; 0, 0). (6.89)
V Q

Shifting to coordinate representation and replacing the summation over the wavevector Q
to an integration over space, we find
e2 D
Z
= 2 drC(r, r; 0, 0). (6.90)
V
In the presence of a magnetic field B, the Cooperon propagator satisfies the differen-
tial equation (6.59). With time-independent vector and scalar potentials and after Fourier
transforming to time, one finds
2ie 1
D(r A(r))2 C(r, r0 ; 0, 0) = (r r0 ). (6.91)
~c
This equation is formally equivalent to the defining equation for the Green function of a
single particle of charge 2e and mass ~2 /2D in a magnetic field. The eigenfunctions of that
problem are known: they are the Landau level wavefunctions n,,qk . These wavefunctions
are labeled by the Landau level index n = 0, 1, 2, . . ., the wavenumber qk parallel to the
direction of the magnetic field, and the degeneracy index . At a fixed value of qk , the
index can take 2eBA/hc values, where A is the sample cross section perpendicular to the
magnetic field.

(a) Show that the solution of Eq. (6.91) can be written as

0
0 1 X n,,qk (r)n,,qk (r )
C(r, r ; 0, 0) = , (6.92)
n,,q Dqk2 + B (n + 1/2)
k

where B = 4DeB/c~ is the cyclotron frequency for a particle of mass ~2 /2D. Note
that no regularization of the denominator is needed.
6.6. EXERCISES 125

(b) Derive an expression for the weak localization correction to the conductivity.

(c) Consider a metal film of thickness a. You may assume D/a2  B , so that the film
is effectively two-dimensional. Discuss the magnetic-field dependence of the weak-
localization for the case that the magnetic field is perpendicular to the plane of the
film and for the case that the magnetic field is parallel to the plane of the film.

Exercise 6.5: Kubo formula for conductance

The response of a bulk sample to an applied electric field is characterized by the conduc-
tivity. However, for a sample of finite size, one measures the conductance, the coefficient
of proportionality between the total current I flowing through a cross section of the sample
and the total voltage drop V over the sample.
In order to derive a relation between the total current I and the voltage drop V , we
introduce a coordinate system (, ), such that is parallel with the electric field lines and
is parallel with the equipotential lines, see Fig. 6.12. The current can then be written as
an integral of the current density,
Z
I() = d j(, ), (6.93)

where is the unit vector in the -direction. We are interested in the dc conductance only.
Then, current conservation implies that I() does not depend on the choice of the cross
section, i.e., I() does not depend on .

(a) Use the Kubo formula for the dc conductivity to express I() in terms of an integral
over the electric field E( 0 ,
0
).

(b) Use current conservation to show that the conductance G can be written as
i R
G = lim Re C (), (6.94)
0 II
R
where CII is the retarded current-current correlation function,
R
CII (t) = i(t)h[I(t), I(0)] i. (6.95)

The current can be calculated among an arbitrary cross section along the sample.
126 CHAPTER 6. DIFFUSION MODES
CBACBA CBACBA CACA EBDEBD EBDEBD EBDEBD EDED
CABCABCBACBA CACA EDBEDBEBDEBD EBDEBD EDED
CBACABCBACBA CACA EBDEDBEBDEBD EBDEBD EDED
CABCBA CBACBA CACA EDBEBD EBDEBD EBDEBD EDED
CABCABCBACBA CACA EDBEDBEBDEBD EBDEBD EDED
CBACABCBACBA CACA EBDEDBEBDEBD EBDEBD EDED
CABCBA CBACBA CACA
EDBEBD EBDEBD EBDEBD EDED
CABCABCBACBA CACA EDBEDBEBDEBD EBDEBD EDED
CBACABCBACBA CACA EBDEDBEBDEBD EBDEBD EDED
CABCBA CBACBA CACA EDBD EBDEBD EBDEBD EDED
CABCABCBACBA CACA equipotential linesEB EDBEBDD EBDEBD EDED
electric field lines EDBEB
CBA CBA CA EBD EBD EBD ED
I

Figure 6.12: Schematic of a conductance measurement. In the derivation of the Kubo formula for
the conductance one uses a coordinate system where the coordinate points along the electric field
and the coordinate points along equipotential lines.

(c) Discuss when the relation G = A/L is true for a sample of cross section A and length
L.

(d) Use the Kubo formula (6.94) to calculate the conductance of a one-dimensional wire
without impurities.

Exercise 6.6: Weak localization and kinetic equation

One can study the effects of weak localization using the kinetic equation. In that approach,
the effect of impurity scattering is described through a correction to the self energy , which
is then used to calculate the collision integral. You may find it helpful to re-read Sec.
5.3 before proceeding with this exercise. This exercise is based on Sec. IV of the review
Quantum field-theoretical methods in transport theory of metals by Rammer and Smith [Rev.
Mod. Phys. 58, 323 (1986)].
6.6. EXERCISES 127

= + + + ...

Born =

R R R
K
WL = + + ...
K K
A A A

Figure 6.13: Top: Diagrammatic representation of the self energy for impurity scattering.
Middle: Leading contribution to the disorder-averaged self energy. Bottom: Maximally crossed
diagrams contributing to the Keldysh component of hi.

(a) With impurity scattering, the self-energy is given by the Born series shown in Fig.
6.13. Show that the Born series shown in the top panel of Fig. 6.13 implies the following
matrix structure for the nth order contribution to the self energy
Pn1 0R p1 0K
(G0R U )n1 0A np1
 
(n) p=1 (G U ) G U (U G U )
=U . (6.96)
0 (G0A U )n1

In Sec. 3.3 you showed how the leading contribution hBorn i to the disorder-averaged self
energy, shown in the middle panel of Fig. 6.13, gave rise to the collision term in the Boltzmann
equation and, hence, to the Drude conductivity of a disordered metal. Here we consider a
contribution to the Keldysh component hK WL i that involves a maximally crossed diagram,
see the bottom panel of Fig. 6.13. Clearly, the maximally crossed part of the diagram is
nothing but the Cooperon propagator.

(b) Show that the weak-localization correction to the self energy hK WL i can be expressed in
0 0 0
terms of the Cooperon propagator C(r, r ; t1 , t1 ; t2 , t2 ) and the Keldysh Green function
G0K in the absence of impurity scattering as
Z
K 0 0 ~
hWL (r, t; r , t )i = dt2 dt01 C(r, r0 ; t, t01 ; t2 , t0 )GK (r0 , t01 ; r, t2 ). (6.97)
2
Here we made the expression self-consistent by including the effect of impurity scat-
tering on GK .
128 CHAPTER 6. DIFFUSION MODES

(c) Why does one consider maximally crossed diagrams for the Keldysh component of the
self energy only?

(d) In order to study the effect of the self energy contribution hWL i on the kinetic equa-
tion, one should shift to a mixed representation with respect to time and space co-
ordinates. Write down the expression you derived under (b) in terms of the mixed
representation.

(e) Now consider the kinetic equation (5.18) in the gradient approximation. Consider the
case of a uniform and time-independent electric field. As before, we drop the real parts
of the self energy and the Green function from the equation, so that one obtains,

eE k GK = ihK (GR GA ) i(R A )GK . (6.98)

Note that the self-energy only affects the right-hand side of the equation, which is
the collision term. Instead of calculating the collision term explicitly, one can find
the effect of weak localization on the linear conductivity from the solution of the full
equation (6.98). In linear response in E, the Keldysh Green function on the l.h.s. can be
replaced by the equilibrium Green function, which is unaffected by impurity scattering.
[In any case, there is no weak-localization correction to GK in equilibrium, because in
equilbrium GK is proportional to the spectral density.] Hence, the net effect of h WL i
on the right-hand side of Eq. (6.98) must be zero. If the effect of weak localization is
small in comparison to that of classical impurity scattering, show that one then finds
that weak localization causes the following small correction to hGK i,

GR G A K
hGK
WL i = h i = 2 hK
WL i/~. (6.99)
R A WL

(f) Now calculate the effect of weak localization on the current density.
Chapter 7

The interacting electron gas

Sofar we have discussed electrons mainly as if they were non-interacting particles. This is
a huge simplification of reality that turns out to work remarkably well. In this chapter we
include electron-electron interactions into our theory and uncover part of the picture why it
is that the approximation of noninteracting electrons is as good as it is.
In our theoretical description, well consider a general electron-electron interaction Hamil-
tonian of the form
H =H 0 + H1, (7.1)
where H 0 is the unperturbed Hamiltonian, which we assume to be quadratic in fermion or
boson creation and annihilation operators, whereas H 1 is the perturbation. The Hamiltonian
H 1 will include the effect of electron-electron interactions, which, in the most general case,
read

1 = 1 V, 0 0 0 0 .
X
H (7.2)
2 0 ,0

Here the indices , , 0 , and 0 label a complete set of single-electron states and V, 0 0 is
the corresponding matrix element. For electrons, the fermion statistics is taken care by the
anticommutation relations of the creation and annihilation operators and . Depending
on the problem of interest, the effects of impurity scattering may be included in H 0 or as an
additional term in H1.
Of course, the true microscopic electron-electron interaction is the Coulomb interaction.
Using a representation in real space, the Coulomb interaction Hamiltonian reads

1 XZ e2
Z
1 =
H dr1 dr2 (r1 ) 2 (r2 )2 (r2 )1 (r1 ). (7.3)
2 , 40 |r1 r2 | 1
1 2

129
130 CHAPTER 7. THE INTERACTING ELECTRON GAS

In momentum space, the Coulomb interaction reads

1 = 1 Vq k 1 +q,1 k 2 q,2 k2 ,2 k1 ,1 ,
X XX
H (7.4)
2V , k ,k q
1 2 1 2

where
e2 iqr e2
Z
Vq = dr e = . (7.5)
40 r 0 q 2
One important aspect of the Coulomb interaction is its long range. As a matter of fact,
the energy required to charge a system is infinite! That is the reason why all objects are
electrically neutral. In most cases, charge neutrality is maintained by the positive charge
density of the ion lattice, but charge neutrality can also be maintained by the charge on
objects that are near the sample of interest, but that are not a part of it, such as nearby
pieces of metal or condensator plates.

7.1 Perturbation theory

The effect of electron-electron interactions can be accounted for by means of a perturbative
expansion in the interaction Hamiltonian H 1 . This perturbative expansion can be organized
using diagrammatic rules, in exactly the same way as the perturbation expansion for impurity
scattering could be organized using the diagrammatic technique, see chapter 4.
In this section, we present the rules for the specific case of the calculation of a single-
particle imaginary time Green function, and then discuss the same calculation in the frame-
work of the real-time formalism. The extension to more complicated correlation functions is
straightforward.

7.1.1 Imaginary-time formalism

Recall that the single particle imaginary time Green function can be expressed as a thermal
average over the non-interacting Hamiltonian H 0,
R 1/T
1 ( 0 )d 0
hT e 0 H
1 (r1 , 1 ) 2 (r2 , 2 )i0
G1 ,2 (r1 , 1 ; r2 , 2 ) = R 1/T
1 ( 0 )d 0
. (7.6)
hT e 0 H
i0

Here operators are represented in the interaction picture. As before, the Green function
1 . Each term in the series can be represented by a Feynman
is calculated in a series in H
diagram. The diagrammatic representation of the interaction (7.2) is as in figure 7.1: it
7.1. PERTURBATION THEORY 131

Figure 7.1: Diagrammatic representation of an interaction vertex

consists of four fermion lines connected by a dotted line.1 Each interaction line carries the
weight V;0 0 appropriate to the representation that is used, the minus sign arising from
the minus sign in the exponent of the evolution operator in Eq. (7.6). For the Coulomb
interaction in coordinate representation, the interaction involves one particle at coordinate
r and one particle at coordinate r0 and has weight is e2 /40 |r r0 |. For the Coulomb
interaction in momentum representation, the Coulomb interaction scatters electrons in mo-
mentum states k and k0 to momentum states k + q and k0 q, respectively, and has weight
Vq /V . The factor 1/V follows from the normalization of the Fourier transform of the
Coulomb interaction, see Eq. (7.4).
The diagrammatic rules for a perturbation H 1 that includes interactions are the same as
those for the case of impurity scattering we discussed in chapter 4. There are two additional
rules, specific for the case of interactions,

1. The interaction Hamiltonian H 1 involves four creation and annihilations operators at

the same time. In order to ensure that creation operators are always kept in front of
annihilation operators, one adds a positive infinitesimal to their imaginary times,

1 ( ) = 1 V, 0 0 ( + ) ( + )0 ( ) 0 ( ).
X
H (7.7)
2 0 0
,

With this convention, the time ordering operator T guarantees that ( ) always
).
appears in front of (

2. With interactions, connected diagrams may include closed particle loops, see, e.g., Fig.
7.2. For fermions, each loop corresponds to an odd permutation of the creation and
annihilation operators in the perturbation expansion, and hence carries a weight 1.
1
In the literature, the dotted line is often replaced by a wiggly line.
132 CHAPTER 7. THE INTERACTING ELECTRON GAS

As before, all internal times are integrated over, all internal spin indices are summed over,
and all internal coordinates or momenta are integrated over or summed over, depending
on the representation that is used in the calculation. You verify that combinatorial factors
conspire in such a way that (i) all disconnected diagrams cancel between numerator and
denominator and (ii) one has to sum over topologically different diagrams only, with the
same weight for each diagram, up to a factor 1 for each fermion loop. You can find a
detailed proof of these rules in the book by Abrikosov, Gorkov, and Dzyaloshinski.
One often studies a generalization of the interaction (7.7), in which the instantaneous
interaction is replaced by a time-dependent interaction (although the time dependence is in
imaginary time),
1
Z Z

d H1 ( ) d dV, 0 0 ( ) ( + ) ( + )0 () 0 ( ). (7.8)
2
In this notation, the original interaction corresponds to V, 0 0 ( ) = V, 0 0 ( ).
We refer to V, 0 0 ( ) as the interaction propagator. The usefulness of an interaction
propagator that depends on an imaginary time difference will be shown later, when we
consider screening. With the time dependence included, the interaction progagator has the
same arguments as an imaginary time Green function. Since the interaction propagator
relates the interaction of densities, we take it to be a periodic function of the imaginary time
difference , with period ~/T .
All diagrams for the single particle Green function, up to second order in the Coulomb
interaction, are shown in Fig. 7.2. The second, fourth, fifth, eleventh, twelfth, and thirteenth
diagram in the figure have an odd number of fermion loops and, hence, receive an extra factor
1. The first three terms in the expansion of the single-particle Green function read

G,0 (r, ; r0 , 0 ) = G0 (r, ; r0 , 0 ),0

Z 1/T XZ Z
d1 d2 dr1 dr2 G0 (r, ; r1 , 1 )(V (r1 , r2 ; 1 2 ))
0 1
0
G1 (r2 , 2 ; r2 , 2 + )G0 (r1 , 1 ; r0 , 0 ),0
Z 1/T Z Z
+ d1 d2 dr1 dr2 G0 (r, ; r1 , 1 )(V (r1 , r2 ; 1 2 ))
0
G0 (r1 , 1 ; r2 , 2 + )G0 (r2 , 2 ; r0 , 0 ),0 , (7.9)

if the coordinate representation is used, whereas in momentum representation the first three
diagrams read

Gk;k0 0 ( ; 0 ) = Gk
0
( ; 0 ),0 k,k0
7.1. PERTURBATION THEORY 133

k k
k1
= + + +
k1 k k
k k k

+ + + +

+ + +

+ + + ...

Figure 7.2: Diagrammatic representation of all diagrams for the single particle Green function, up
to second order in the Coulomb interaction. The Coulomb interaction lines are drawn dotted.

1/T
1
Z XX
0
d1 d2 Gk ( ; 1 )(V0 (1 2 ))
V 0 1 k 1

Gk01 1 (2 ; 2 + 0
)Gk (1 ; 0 ),0 k,k0
Z 1/T
1 X
0
+ d1 d2 Gk ( ; 1 )(Vkk1 (1 2 ))
V 0 k1

Gk01 (1 ; 2 + 0
)Gk (2 ; 0 ),0 k,k0 . (7.10)
Note that the interaction matrix element that appears in the integrations depends on the
distance r1 r2 only (if the coordinate representation is used) or on the transferred mo-
mentum k k1 (if the momentum representation is used). Also note that the Coulomb
interaction does not flip spins, so that spin needs to be conserved separately at each end of
an interaction line. We used the generalized version of the interaction, see Eq. (7.8) above.
For a standard interaction Hamiltonian, the time-dependence of the interaction potential
V (1 2 ) is proportional to (1 2 ).
In order to deal with the integrations over imaginary time, it is advantageous to make a
Fourier transform to imaginary time, i.e., to write
0
X
G,0 (r, ; r0 , 0 ) = T G, (r, r0 ; in )ein ( ) ,
n
134 CHAPTER 7. THE INTERACTING ELECTRON GAS

0
X
Gk;k0 0 ( ; 0 ) = T Gk;k0 0 (in )ein ( ) , (7.11)
n

where the n are the fermionic of bosonic Matsubara frequencies. Each internal integration
over imaginary time for each interaction vertex gives 1/T times a Kronecker delta for the
Matsubara frequencies. Since an interaction vertex carries only one internal time, this im-
plies that only the sum of the four Matsubara frequencies is conserved at each interaction
line, not the Matsubara frequencies of the individual fermions or bosons participating in the
interaction.2 In other words, at interaction vertices a Matsubara frequency difference can be
transferred between particles. Note that a transferred Matsubara frequency difference
is of the boson type if it is transferred between bosons and bosons or fermions and fermions,
and of fermion type if it is transferred between a boson and a fermion. Since the Coulomb
interaction is instantaneous, the weight of the Coulomb interaction lines do not depend on
the Matsubara frequency that is transferred at the interaction line. All remaining interme-
diate Matsubara frequencies that are not fixed by the conservation rules are summed over;
normalization requires an extra factor T for each summation over Matsubara frequencies, in
accordance with Eq. (7.11).
To illustrate the use of Matsubara frequencies, we rewrite Eq. (7.10) using Matsubara
frequencies,

Gk;k0 0 (in ) = G0k (in ),0 k,k0

T XXX 0
Gk (in )(V0 (i0))G0k1 1 (im )eim G0k (in ),0 k,k0
V m k
1 1

T X X
+ G0k (in )(Vkk1 (in im ))G0k1 (im )eim G0k (in ),0 k,k0 .
V m
k1
(7.12)

The factors exp(im ) were added to account for the infinitesimal time difference in the in-
termediate Green functions, cf. Eqs. (7.10) and (7.11). Note that in the case of an interaction
potential, the Fourier transform V (im ) does not depend on the frequency m .
Looking at all the indices and summations that appear in Eq. (7.12), it becomes useful
to group all these indices together into one four-vector that includes the momentum, the
spin, and the Matsubara frequency,

k = (k, , in ).
2
The statement that the sum of the four Matsubara frequencies is conserved means that the sum of the
Matsubara frequencies of the two outgoing particles equals the sum of the Matsubara frequencies of the
incoming particles at the interaction vertex.
7.1. PERTURBATION THEORY 135

Then the symbol

X T XXX
=
k
V k n
will denote summation over the momentum k, the spin , and the Matsubara frequency n
and includes the appropriate normalization factors.

7.1.2 Real-time formalism

For the real-time formalism, the expression for the Green function is given by Eq. (2.56),
where H 1 is the interaction Hamiltonian. If the contour language is used, the diagrammatic
rules are precisely the same as in the case of the imaginary time formalism, except for two
minor differences:
1. The meaning of the infinesimal is such that t + is further along the contour than
t. This preserves the correct ordering of operators evaluated at equal times.

2. The interaction propagator gets a factor i, instead of a minus sign, which is a con-
sequence of the fact that the real-time Green functions are defined with a factor i,
whereas the imaginary-time Green function carry a factor 1.
With these rules, the diagrammatic expansion of the single-particle Green function is the
same as shown in Fig. 7.2. Calculating the Green function up to first order in the interaction,
we find

G,0 (r, t; r0 , t0 ) = G0 (r, t; r0 , t0 ),0

1X
Z Z
dr1 dr2 dt1 dt2 G0 (r, t; r1 , t1 )(iV (r1 r2 ; t1 , t2 ))
~ c
1

G01 (r2 , t2 ; r2 , t2 + )G0 (r1 , t1 ; r0 , t0 ),0

1
Z Z
+ dr1 dr2 dt1 dt2 G0 (r, t; r1 , t1 )(iV (r1 r2 ; t1 , t2 ))
~ c
G (r1 , t1 ; r2 , t2 + )G0 (r2 , t2 ; r0 , t0 ),0 .
0
(7.13)

The Green function G01 (r2 , t2 ; r2 , t2 + ) from the third line of Eq. (7.13) is always a lesser
Green function, because of the presence of the infinitesimal .
In the real-time formalism, the interaction propagator V depends on two contour points
t1 and t2 , and is assumed to have the same causal structure as a contour-ordered Green
function. This condition makes the replacement of an instantaneous interaction potential by
an interaction propagator is not entirely straightforward. At first sight, one is tempted to
136 CHAPTER 7. THE INTERACTING ELECTRON GAS

set V (t1 t2 ) (t1 t2 ) if t1 and t2 are both on the upper (+) or lower () branch on
the Keldysh contour, and zero otherwise. Indeed, with this choice the integration over t2 in
Eq. (7.13) gives one, and the Green function G0 (r1 , t1 ; r2 , t2 + ) from the fifth line of Eq.
(7.13) becomes a lesser Green function. That is precisely what it should be, since a lesser
Green function represents an electron density, as is appropriate for an interaction correction.
However, this simple choice for the interaction propagator does not have the causal structure
of a true real-time propagator. To see this, one shifts to the matrix notation, in which V
has the form
 R
V (t1 t2 ) V K (t1 t2 )
  
(t1 t2 ) 0
V (t1 t2 ) = =V .
V Z (t1 t2 ) V A (t1 t2 ) 0 (t1 t2 )
Without regularization, it is not clear that the advanced and retarded components are zero
for t1 < t2 and t1 > t2 , respectively. You verify that the correct regularization, which gives
a lesser Green function in the fifth line of Eq. (7.13) and which obeys the conditions that
V R (t1 t2 ) = 0 for t1 < t2 , V A (t1 t2 ) = 0 for t1 > t2 , and V Z (t1 t2 ) = 0 for all t1 , t2 is3
 R
V (t1 t2 ) V K (t1 t2 )

V (t1 t2 ) =
V Z (t1 t2 ) V A (t1 t2 )
 
(t1 t2 ) (t1 t2 + ) (t1 t2 )
= V . (7.14)
0 (t1 t2 + )
With this choice of the interaction propagator, the infinitesimal can be dropped from the
Green function G0, (r1 , t1 ; r2 , t2 + ) in the fifth line of Eq. (7.13). (In the general case that
V (t1 t2 ) has a true time dependence, the times t1 and t2 are different and addition or
subtraction of the infinitesimal is irrelevant.)
Equation (7.13) is written in the contour language: the arguments t and t0 of the Green
function, as well as the integration variables t1 and t2 represent points on the Keldysh
contour. It is customary to switch to the matrix representation, in which the contour-ordered
Green function is represented by a 2 2 matrix as in Eq. (2.61),
 R
G GK

G= . (7.15)
0 GA
3
In contour language, Eq. (7.14) corresponds to
V11 (t1 , t2 ) = V (t1 t2 + ),
V12 (t1 , t2 ) = V (t1 t2 + ) V (t1 t2 ),
V21 (t1 , t2 ) = 0,
V22 (t1 , t2 ) = V (t1 t2 ).
Here the subscripts 1 and 2 refer to contour points on the upper and lower branches of the Keldysh contour,
respectively.
7.1. PERTURBATION THEORY 137

i i
~
k ijk
k ijk
absorption j emission j

Figure 7.3: Diagrammatic rules for interaction Coulomb vertices.

In the matrix notation, the arguments t, t0 , as well as the integration variables t1 and t2 , are
simple times, not contour points. Switching to the notation (2.61), one then finds
G,0 (r, t; r0 , t0 ) = G0 (r, t; r0 , t0 ),0
1X
Z Z
dr1 dr2 dt1 dt2 G0 (r, t; r1 , t1 )G0 (r1 , t1 ; r0 , t0 ),0
~
1

(iV (r1 r2 ; t1 , t2 )R ) G0<

1 (r2 , t2 ; r2 , t2 )
1 X Z Z
+ dr1 dr2 dt1 dt2 G0 (r, t; r1 , t1 )m G0 (r1 , t1 ; r2 , t2 )
~ mn
n G0 (r2 , t2 ; r0 , t0 )(iV (r1 r2 ; t1 , t2 )mn ),0 , (7.16)
where the indices m and n now refer to the matrix structure (7.15) and
   
1 1 0 1 0 1
1 = 2 = , 2 = 1 = . (7.17)
2 0 1 2 1 0
The matrix manipulations that lead to Eq. (7.16) can be summarized in terms of a general
diagrammatic rules: the vertices with the interaction propagator carry a matrix or , see
Fig. 7.3. Unlike in the imaginary-time formalism or in the contour language, the interaction
lines need to have a direction if the representation (7.15) is used. However, the final results
do not depend on the direction of the assigned direction and the direction of the interaction
propagator has no physical meaning at this point.
Once the matrix structure has been laid out, Fourier transform to time and space coor-
dinates is straightforward.
Gk;k0 0 () = G0k (),0 k,k0
1 X
Z X
d 0 G0k ()G0k (),0 k,k0 (iV0 (0)R ) G0< 0
k1 1 ( )
2V k1
Z1
1 X X
+ d 0 G0k ()m G0k1 ( 0 )
n G0k (),0 k,k0 (iVkk1 ( 0 )mn ).
2V m,n
k1
(7.18)
138 CHAPTER 7. THE INTERACTING ELECTRON GAS

In this case, we can simplify the notation by writing symbolically k = (k, , ), and
1 XX
X Z
= d.
k
2V k

7.2 Self energy and Hartree Fock approximation

An important approximate method to treat the effect of interactions is the Hartree Fock
approximation. In the Hartree Fock approximation, the electron-electron is replaced by
an effective potential, which is then determined self-consistently. Replacing the interaction
by a potential is a very useful approximation, because potentials can be dealt with on the
single-particle level.
If the interaction is replaced by an effective potential, the Hamiltonian is quadratic in
electron creation and annihilation operators. This means that Wicks theorem holds for this
Hamiltonian. The principle of the Hartree Fock approximation is that the interaction Hamil-
tonian is replaced with a Hamiltonian that is quadratic in electron creation and annihilation
operators in a way that respects Wicks theorem,4

HF = 1
 

X
H 1 V 0
, 0 0 h 0 i + 0 h 0 i h 0 ih 0 i
2
1  
V, 0 0 0 h 0 i + 0 h 0 i h 0 ih 0 i .
X

2
(7.19)
Here, the first line corresponds to what is known as the Hartree Hamiltonian, whereas
the second line is the Fock Hamiltonian. If the labels and refer to different particles,
the Fock contribution is zero; for indistinguishable particles, however, the Fock contribution
is important. The averages h 0 i are to be regarded as parameters in the Hartree-Fock
Hamiltonian that should be calculated self-consistently at the end of the calculation.
In a non-magnetic translationally invariant system, any average of the form hk
k0 0 i
0 0
should be nonzero only if k = k and = . Denoting
nk = hk

k i, (7.20)
the Hartree-Fock Hamiltonian in a translationally invariant system can be written
0 + H
HF = HF k ,
X
H HF = H 1 k k (7.21)
k

4
You verify that any average of the Hamiltonian H 1HF is the same as that of the original interaction

Hamiltonian H1 if the Wick theorem were to be used.
7.2. SELF ENERGY AND HARTREE FOCK APPROXIMATION 139
FG
G
FGFG FG
FG FG
G
FG GFGF
GF F FG
G F FG
G GFGF
GF
GF F FG F FG GFGF
FGFG
G FG
F G
G
F F GFGF
= + + +

+ + + + + ...

Figure 7.4: Diagrammatic expansion of the self energy, up to second order in the Coulomb inter-
action.

where the Hartree Fock energy is given by

X
HF
k = k + (V0 0 Vkk0 )nk0 0 . (7.22)
k0

Note that the Hartree-Fock energies HF k depend on the occupation numbers nk0 0 which, in
turn, depend self-consistently on the Hartree-Fock energies HF k .
The Hartree Fock approximation can be given a diagrammatic interpretation by consider
the concept of a self energy. In previous chapters, we saw that this concept was very useful
in order to organize the summation of a diagrammatic perturbation series for a disorder-
averaged Green function. The same is true for interacting electrons.
As in the case of impurity diagrams, we call a diagram irreducible if it cannot be cut
in two by removal of a single fermion line. In Fig. 7.2, the seventh, tenth, twelfth, and
thirteenth diagrams are not irreducible. The self-energy is defined as the sum over all all
irreducible diagrams without the two external fermion lines, see Fig. 7.4. With this definition
of the self energy, the single-particle Green function obeys the Dyson equation
X
Gk,k0 = G0k,k0 + G0k,k1 k1 ,k2 Gk2 ,k0 . (7.23)
k1 ,k2

In the real-time formalism, is a 2 2 matrix and matrix multiplication is implied. Note

that, in contrast to the case of the disorder average we studied before, the self energy k,k0
does not need to be diagonal in the four-vectors k, k 0 .
A perturbative expansion for the self energy is shown in Fig. 7.4. The first two diagrams of
Fig. 7.4 have our special attention: The interaction line can be seen as an effective impurity
140 CHAPTER 7. THE INTERACTING ELECTRON GAS

Figure 7.5: Interaction vertex and exchanged interaction vertex.

line. In other words, it can be represented by an effective potential. This is particularly

clear for the first diagram, which is known as the Hartree diagram. For the second diagram,
which is known as the Fock diagram, you should consider the fermion line as being part
of the effective impurity.
At first sight, the Hartree and Fock diagrams appear to be rather different. However, they
are related by simple exchange of particles participating in the interaction! To understand
this, look at the effect of exchange on the basic interaction vertex, see Fig. 7.5. It is precisely
this exchange operation that maps the Hartree diagram into the Fock diagram and vice
versa.
Let us calculate what these two contributions to the self energy are. We use the imaginary-
time formalism. Calculation of the Hartree diagram gives, cf. Eq. (7.12),
T XXX
Hk (i n ) = (V0 )Gk00 0 (im )eim . (7.24)
V k0 0 m
The factor exp(im ) arose because we take the creation operator in the interaction Hamilto-
nian at an infinitesimally larger imaginary time than the annihilation operator. Performing
the summation over the Matsubara frequency m and the momentum k, we find
X
H = V 0 n0k = V0 n, (7.25)
k

where n is the particle density. Similarly, calculation of the Fock diagram gives, cf. Eq.
(7.12),
T XXX
Fk (in ) = (Vkk0 0 )Gk0 0 (im )eim
V k0 0 m
1 X
= Vkk0 n0k0 . (7.26)
V k0
Here, there is no summation over spin, because spin is conserved throughout the diagram.
In the real-time formalism, the self-energy is given by Eqs. (7.25) and (7.26), multiplied by
the 2 2 unit matrix.
7.3. DIELECTRIC RESPONSE 141
HJI
JI
HJHIJI HJI
HJI HJI
JI
HJI JHJH
JHI H H I
J H H I
J JHJH
JHI
HJHIJI
JI HJI
HJI HJI
JI
H JHJH =
JHIHJI H JI HJI
H JHJH +

Figure 7.6: Diagrammatic representation of the Hartree-Fock approximation. The self energy is
expressed in terms of the full single-particle Green function, which, in turn, depends on the self
energy as in Fig. 5.7.

Looking at these results, we first note that the Hartree and Fock corrections to the self
energy are real: they correspond to a shift of the particles energy, but the particles lifetime
remains infinite. The results (7.25) and (7.26) are almost identical to the correction to the
single particle-energy in the Hartree-Fock approximation. The only difference is the absence
of self-consistency in Eqs. (7.25) and (7.26): in those equations, the occupation numbers n k
are calculated in the non-interacting ground state. Self-consistency can be achieved if the
single-particle Green function G 0 in the Hartree and Fock diagrams is replaced by the full
Green function G, see Fig. 7.6. In that case one recovers precisely the Hartree-Fock equation
(7.22). You easily verify that, because of the self-consistency condition, the Hartree-Fock
approximation includes all diagrams of Fig. 7.4 with the exception of the fourth and the
eighth diagrams of that figure.
Note that, formally, the Hartree self energy is divergent. In practice, the negative charge
of the electrons is balanced by the positive charge of the lattice ions, and there is no contri-
bution from the q = 0 mode of the interaction. Hence, for a translationally invariant system,
the Hartree contribution to the self energy can be neglected.

7.3 Dielectric response

Interactions profoundly modify the electronic response to an external potential. This phe-
nomenon is known as screening. Screening is described by the polarizability function e ,
which was defined in Sec. 6.1. The polarizability function expresses the induced charge
density as a function of the external potential,

ind (q, ) = R
e (q, )ext (q, ). (7.27)

Below, well calculate the polarizability function using the Green function formalism.
However, for long-wavelength and low frequency perturbations, one can use the Boltzmann
142 CHAPTER 7. THE INTERACTING ELECTRON GAS

equation to find e . Starting from the Fourier transformed linearized Boltzmann equation,

(i + ivk q)fk1 (q, ) i(e/~)(q, )q k fk0 = 0, (7.28)

with (q, ) = ext (q, ) + ind (q, ) equal to the total electrostatic potential, we find that
the presence of an external potential perturbation ext causes the response

eq vk (ext (q, ) + ind (q, )

fk1 (q, ) = (k f 0 (k )). (7.29)
~(q vk )

In order to eleminate ind from this equation, we multiply by 2e/V and sum over k, with
the result
2e X 1
ind (q, ) = f (q, )
V k k
2e2 X q vk
= (k f 0 (k ))(ext (q, ) + ind (q, ))
V ~ k q vk
= 0R
e (q, )(ext (q, ) + ind (q, )), (7.30)

where 0R is the polarizability function of the non-interacting electron gas. Of course,

this equation could have been derived without use of the Boltzmann equation. It simply
expresses that the response of the interacting electron gas to the potential ext is the same
as the response of the non-interacting electron gas to the total potential ext + ind . Using
1
ind (q, ) = ind (q, ), (7.31)
0 q 2
we find finally
0R
e (q, )
R
e (q, ) = . (7.32)
0R
1 e (q, )/0 q 2
Before we study the properties of e in more detail, we show how the polarizability
function can be derived in the Green function formalism. In the Green function formalism,
e is calculated as the retarded charge density autocorrelation function, which is obtained
as the analytical continuation of the temperature charge density autocorrelation function
1
e (q, ) = hT e,q ( )e,q (0)i (7.33)
V
e2 X X
= hT k,

( + )k+q, ( )k 0 +q,0 ()k0 ,0 (0)i.
V k k0 0
7.3. DIELECTRIC RESPONSE 143
MLKMLK MLKMLK MKMK
e =
k
MKLKLM MLKKLM MKKM k

k+q
MKLKLMLKKLMKK k+q
= + + +

+ + + +

+ + + +

+ + ...

Figure 7.7: Diagrammatic expansion of the polarizability function e . The figure contains all
diagrams to first order in the interaction and a subset of the diagrams to second order in the
interaction.

Diagrammatically, this function is obtained as in Fig. 7.7. We can organize the diagrammatic
expansion as we did previously when we looked at the single-particle Green function. A
polarization diagram is called reducible if the two end points can be separated by cutting
a single interaction line. For example, of the diagrams in Fig. 7.7, the seventh, ninth, tenth,
eleventh, and thirteenth diagrams are reducible. All other diagrams are irreducible in this
sense. Defining the irreducible polarizability function as (minus) the sum over all irreducible
diagrams, we can write the polarizability function as
irr
e (q, in )
e (q, in ) = . (7.34)
1 Vq irr
e (q, in )/e
2

The diagrammatic series for irre is as in Fig. 7.8. In the so-called random phase approxi-
mation (RPA), one only keeps the zeroth order diagram for irr e , which is the polarizability
function for the noninteracting electron gas. After analytical continuation to real frequencies,
one finds
RPA 0R
e (q, )
e (q, ) = , (7.35)
1 Vq 0e (q, )/e2
which is the same as we obtained from the Boltzmann equation previously.
We can now use our result for the polarizability function to discuss the dielectric response
function
1 (q, ) = 1 + Vq R 2
e (q, )/e . (7.36)
144 CHAPTER 7. THE INTERACTING ELECTRON GAS
PONPON PONPON PNPN
eirr =
k
PNOPNOPONPON PNPN k

k+q
PNOPON PN k+q
= + + +

+ + + + + ...

Figure 7.8: Diagrammatic expansion of the irreducible part irr

e of the polarizability function. The
figure contains all diagrams to first order in the interaction and a subset of the diagrams to second
order in the interaction.

Substituting the RPA result (7.35), we find

2 1
1 (q, ) = 1 0R
 
e (q, )Vq /e . (7.37)

The polarizability function of the non-interacting electron gas was calculated in Ex. 6.2. In
the static limit, 0, the result is
    
0 kF q  2kF + q 
e (q, 0) = 1 + ln   , (7.38)
q 4kF 2kF q 
where = mkF /2 2 is the density of states at the Fermi level. Hence,

e2 k F m  2kF + q  1
     
1 kF q
 (q, 0) = 1 + 1+ ln 
  . (7.39)
20 q 2 2 q 4kF 2kF q 
The long wavelengths limit of Eq. (7.39) is the Thomas-Fermi dielectric function,
ks2
(q, 0) = 1 + 2 , (7.40)
q
where ks2 = me2 kF / 2 0 is the inverse Thomas-Fermi screening length.
An important feature of the RPA result for the static dielectric response function is that
it is singular at q = 2kF . This has a number of physical consequences, akin to the Friedel
oscillations we discussed in an earlier chapter.
For large frequencies  qvF , the polarizability function 0 can be expanded as
kF3 q 2
 
R 3  qvF 2
0 (q, ) = 1+
3 2 m 2 5
2
 
nq 3 qvF 2

= 1+ , (7.41)
m 2 5
7.3. DIELECTRIC RESPONSE 145

= + + + ...

= +

Figure 7.9: Diagrammatic representation of the effective interaction in the random phase approx-
imation.

where n = kF3 /3 is the density of electrons. Substituting this into the dielectric response
function gives
1
ne2
 
1 3  qvF 2
(q, ) = 1 1+ . (7.42)
m0 2 5
Just as the dielectric response function gives the interaction the electron gas to an external
charge, it also gives the response of the electron gas to itself. In other words, it describes how
the electron-electron interaction is modified by the presence of other electrons. The effective
electron-electron interaction is then simply Vq 1 (q, in ). In the RPA approximation, this
corresponds to

Vq e2
VqRPA (in ) = = . (7.43)
1 0e (q, in )Vq 0 q 2 0R
eq (in )

Diagrammatically, the effective interaction is calculated as in Fig. (7.9). Again, the ran-
dom phase approximation consists of taking the non-interacting polarizability function for
polarization bubbles in the diagrammatic expansion.
Note that the effective interaction has a nontrivial time dependence. Because of that,
we call the effective interaction retarded. Now you understand why we went through
the trouble of defining the diagrammatic rules for the general case of a time-dependent
interaction instead of simply restricting ourselves to an instantaneous interaction potential:
if the effective interaction is used as the building block for a diagrammatic expansion, the
interaction lines carry a true time dependence.
For small momentum and energy exchange, we can replace the polarizability 0 by its
value at q = 0,
0R 2 2
e,q0 (in 0) = 2e , = 0 ks , (7.44)
where is the density of states at the Fermi level, per spin direction and per unit volume, and
ks is the inverse Thomas-Fermi screening length introduced previously. Then the effective
146 CHAPTER 7. THE INTERACTING ELECTRON GAS

RPA
Uq = = + + + ...

RPA
Vq = = + + + ...

Figure 7.10: Diagrammatic representations of impurity potential and screened Coulomb interaction
in the RPA approximation.

Coulomb interaction in the limit of small q and reads

e2
VqRPA (in ) = if q, n 0. (7.45)
0 (q 2 + ks2 )

The quantity 1/ks is known as the Thomas-Fermi screening length. The singularity of the
Coulomb interaction at large wavelengths is cut off by the screening of the interaction by
the electron gas. The Thomas-Fermi screening potential is found when we use the limiting
form of Eq. (7.45) for all momenta and frequencies. In that case, the screened Coulomb
interaction is instantaneous and has the spatial dependence of the Yukawa potential,

e2 ks r
V TF (r) = e . (7.46)
40 r

Note, however, that true RPA screened interaction we derived above, however, has a different
from on short length scales and, more importantly, is retarded (i.e., it depends on frequency).
At this point, it is important to make a comment on the effect of impurities in the
metal. The impurities are charged objects otherwise they would hardly interact with the
electrons. However, in our previous treatment, we have modeled impurities with a short-
range potential. Now we see why that was correct: the impurity potential is screened by the
electron gas, so that
U q U q + U q R
e (q, 0)Vq . (7.47)
Diagrammatically, the screening of the impurity potential and of the Coulomb interaction
in the RPA approximation can be represented in striking similarity, see Fig. 7.10.
The random phase approximation may seem rather ad-hoc: one decides to sum a sub-
set of diagrams, and to ignore all others. In fact, this is not true. One can show that the
RPA becomes exact in the limit of a high density of electrons. This is what we now show.
7.3. DIELECTRIC RESPONSE 147

In defining the relative importance of interactions, one introduces the so-called gas
parameter rs , which is a dimensionless parameter measuring the density of the electron
gas. The parameter rs is defined in terms of the electron number density n as

a0 rs = (4n/3)1/3 . (7.48)

A sphere of radius a0 rs contains precisely one electron on average. Using the relation n =
kF3 /3 2 , we can rewrite this as

rs = (9/4)1/3 (a0 kF )1 . (7.49)

A low gas parameter corresponds to a high density of electrons. The gas parameter also
relates the Fermi wavenumber kF and the inverse Thomas-Fermi screening length ks ,
 2/3
4 kF 16
ks2 = = kF2 rs . (7.50)
a0 3 2

The importance of the electron density for a perturbation expansion in the Coulomb
interaction follows from the observation that the gas parameter is proportional to the ratio
of typical Coulomb interaction for neighboring electrons, which is of order e2 /0 kF , and the
electronic kinetic energy, which is ~2 kF2 /2m. Hence, at high electron densities the main
contribution to the electrons energy is kinetic.
For realistic metals, rs is between 1.5 and 6. That is not really small. Hence, our small-
rs expansion is an idealized theory, and we should expect that some of our quantitative
conclusions cannot be trusted.
In order to see which diagrams are important in the limit of high density, we need two
observations.

First, we consider a diagram contribution to the self energy to nth order in the Coulomb
interaction. In order to estimate its dependence on the gas parameter rs , we measure
energy and temperature in units of the Fermi energy, which is kF2 . Then, noting that
every nth order diagram has n integrations over internal momenta (there are 2n 1
fermion lines and n 1 constraints from the interaction lines) and that it contains
2n 1 fermion Green functions, we find that

(n) /F rsn . (7.51)

(The summations over internal Matsubara frequencies do not contribute to this esti-
mate, since, for each internal Matsubara frequency, the diagram carries a factor T .)
148 CHAPTER 7. THE INTERACTING ELECTRON GAS
QSR
SR
QSQRSR QSR
QSR QSR
SR
QSR SQSQ
QR
SSQRSSR Q R
Q Q R
Q SSR SQQS
SQR Q QQR Q QQR SQSQ
SSQRSR
QR SSR
Q SR
SR
S Q QSSQ
= + + + ...

Figure 7.11: Diagrammatic representation of the random phase approximation for the self energy
in a translationally invariant system.

Second, we note that the Coulomb potential is singular: for small transferred mo-
mentum q, the Coulomb potential diverges Vq q 2 . The role of these divergences
is stronger if the same momentum q is transferred more than once, i.e., if there is
more than one interaction vertex with precisely the same momentum transfer. Hence,
we conclude that those diagrams with the maximal number of identical transferred
momenta have the largest contribution.
With these two observations, we can identify the leading diagrams in the limit of high
electron density. Namely, for each set of diagrams with the same largest number of identical
transferred momenta, the diagram of minimal order in the interaction potential dominates for
small rs . These are precisely the diagrams one retains in the random phase approximation.
Note that the random phase approximation does not treat the direct interaction and the
exchange interaction on the same footing: performing the same transformation as in Fig. 7.5
on the diagrams of Fig. 7.6 leads to different contributions to the self energy. The reason that
these contributions are left out in the present calculation is that they are less singular. In
the next chapter we consider the effects of a short-range interaction, for which it is essential
that exchange and direct interactions are treated equally.
We close this section with a calculation of the self energy in the random phase approxi-
mation. For the self energy, the random phase approximation corresponds to retaining the
diagrams shown in Fig. 7.11. If we sum the entire RPA series for the self energy, we find
RPA im
T X X Vkk 0 (in in )e
RPA
k (i n ) = . (7.52)
V 0 m (im k0 + )
k

Unlike the Hartree-Fock self energy, the self energy in the random phase approximation
depends on frequency.

7.4 Plasmon modes

For some frequency/wavelength combinations, the dielectric response function can vanish.
Then, there is an induced charge density without an external one! Such a situation is referred
7.4. PLASMON MODES 149

UTUT UTUT UTUT UTUT UTUT UTUT UTUT UTUT UTUT UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV
/ F 4 UTUTUTUT UTUT UTUT UTUT UTUT UTUT UTUT UTUT UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV
UTUTUTUT UTUT UTUT UTUT UTUT UTUT UTUT UTUT UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV
UTUTUTUT UTUT UTUT UTUT UTUT UTUT UTUT UTUT UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV
UTTU UTTU UTTU UTTU UTTU UTTU UTTU UTTU UTTU UTTUWVVW UTTUWVVW UTTUWVVW UTTUWVVW UTTUWVVW UTTUWVVW UTTUWVVW UTTUWVVW UTTUWVVW UTTUWVVW UTTUWVVW UTTUWVVW
UTUTUTUTUTUTUTUTUTUTWV UTWV UTWV UTWV UTWV UTWV UTWV UTWV UTWV UTWV UTWV UTWV
2
UTUT UTUT UTUT UTUT UTUT UTUT UTUT UTUT UTUT UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV UTUTWVWV
0
0 1 2 3 4
q/k F

Figure 7.12: Plasmon dispersion relation (thick curve) together with the support of Im . Plasmon
modes are strongly damped (Landau damping) if Im  6= 0.

to as a collective mode of the electron gas. It is a motion that can sustain itself without
external input.
In principle, the dielectric response function is complex. In those cases, the occurrence
of zeroes is unlikely. However, there are regions of (q, ) space where  is real. In such
regions, long-lived collective excitations can be expected. The support of the imaginary part
of the polarizability function of the noninteracting electron gas (which equals the support of
Im 1 ) was shown in Fig. 6.2. Based on that figure, we expect collective excitations in the
region of finite frequencies and long wavelengths. In this parameter regime, the dielectric
response function is given by Eq. (7.42) above. We see that, indeed, there are zeroes, given
by the dispersion relation

3q 2 vF2 ne2
(q) = p + , p2 = . (7.53)
10p m0

The frequency p is known as the plasma frequency and the corresponding collective
modes are known as plasma oscillations or plasmons. The plasmon modes are long-lived
as long as the dielectric response function remains real. At the point when the dispersion
curve (7.53) enters that region of (q, ) space where 1 is complex, the zeroes of  at real
frequency and wavevector q cease to exist. In this case, one can still find zeroes of 
at complex frequencies. These correspond to damped collective modes. The damping that
arises from the fact that  is complex is called Landau damping.
The plasmon dispersion relation is shown schematically in Fig. 7.12, together with the
support of Im (q, ).
150 CHAPTER 7. THE INTERACTING ELECTRON GAS

7.5 Free energy

What do the interactions imply for the free energy of the electron gas? At first sight, the
free energy cannot be calculated using the Green function method after all, it is not a
Green function. However, there is a nice technical trick that allows us to apply the Green
function formalism to the calculation of the free energy.
1 is multiplied by a coupling constant ,
In this trick, the interaction Hamiltonian H

H() 0 + H
=H 1. (7.54)

Hence, H(0) corresponds to the unperturbed Hamiltonian H 0 whereas H(1) corresponds to
0 + H
the full Hamiltonian H 1 . For this Hamiltonian, the Free energy becomes a function of

F () = T ln Z(), Z() = tr e(H0 +H1 )/T . (7.55)
Taking a derivative to , we find

F () tr H 1 e(H 0 +H 1 )/T 1
=
= hH 1 i (7.56)
tr e(H0 +H1 )/T

where the brackets h. . .i denote a thermal average with respect to the Hamiltonian H().
Integrating Eq. (7.56) from = 0 to = 1 we find the free energy difference F between
the full Hamiltonian H0 + H 1 and the unperturbed Hamiltonian H 0,
Z 1
1
F = d hH 1 i . (7.57)
0

Hence, if we know the average hH 1 i for all , we can find the change in the free energy.
This average is nothing but a Green function!
Rewriting H 1 as

1 X X 1 X
1 =
H Vq k 1 +q,1 k1 ,1 k 2 q,2 k2 ,2 n Vq , (7.58)
2V , k ,k ,q 2 q
1 2 1 2

1 i as
we can express the average hH

1i = 1 X 1 X
hH lim V
q e (q, ) n Vq ,
2e2 0 q 2 q
T X X in 1 X
= 2 e Vq e (q, in ) n Vq , (7.59)
2e q n 2 q
7.5. FREE ENERGY 151

where e is the polarizability function. Within the random phase approximation we can use
Eq. (7.35) for e , replace Vq by Vq and perform the integration to . The result is
T X X in
1 X
F F0 = e ln 1 Vq 0e (q, in )/e2 n Vq . (7.60)
2 q n 2 q

We can perform the summation over n writing F F0 as a contour integral in the complex
plane and shifting the contours to the real axis,
X Z d 1 X
F F0 = coth(/2T )Im ln ((q, )) n Vq . (7.61)
q 4 2 q

Here we used Eq. (7.37) to express the argument of the logarithm in terms of the RPA
dielectric response function.
Evaluation of this expression for the Free energy is not entirely straightforward. Some
insight into the physics of Eq. (7.61) is obtained if we look at the contribution from the
plasmons. Using Eq. (7.42) for the dielectric response function (q, ) near the plasmon
branch p (q)  qvF at low wavenumber, we find

0 if || > p (q),
Im ln (q, ) = if qvF  < p (q), (7.62)
if p (q) <  qvF .

We then see that each plasmon mode with vF q  p (q) has a contribution
1 p (q)
Z
Fq = d coth(/2T )
2
= T ln(sinh(p /2T )) const, (7.63)
where the additional constant reflects the contribution from the integration close to = 0
where the approximation (7.62) does not hold. The first term on the r.h.s. of Eq. (7.63)
represents the free energy of a plasmon mode at wavevector q, while the additional constant
represents the contribution from low energy particle-hole excitations to the free energy.
In the literature (e.g., the book by Bruus and Flensberg or Mahans book), one can find
a result for the ground state energy as an expansion for small gas parameter rs ,
 
0.916
E E0 = + 0.0622 ln rs 0.094 Ry. (7.64)
rs
In the same units, the ground state energy of the non-interacting electron gas is (2.21/r s2 )
Ry. The unit Ry, for Rydberg, corresponds to e2 /2a0 13.6 eV. Higher-order terms in the
small-rs expansion of the ground state have been calculated. However, for those terms, one
needs to go beyond the RPA approximations.
152 CHAPTER 7. THE INTERACTING ELECTRON GAS

7.6 Interactions and disorder

The fact that dynamics in a dirty metal is diffusive, rather than ballistic, for time scales
longer than the mean free time and length scales longer than the mean free path l = vF ,
has important consequences for the screening of the Coulomb interaction: since it is more
difficult to move charge around, the electron gas has a reduced capacity to screen time-
dependent potentials. The relevant frequency for the time dependence is not the ballistic
frequency vF q (as it is in the response of a clean electron gas), but the diffusive one, Dq 2 ,
where 1/q  l is the length scale that the system is probed at. Qualitatively, the effect of
disorder on the screened Coulomb interaction can be seen from Fig. 6.3. In order to screen
a time-dependent potential with frequency and wavevector q, the electrons need to move
a distance 1/q in a time 1/. As long as  Dq 2 the electron gas can accommodate
such a displacement, and Thomas-Fermi screening theory applies. If, however,  Dq 2 the
electron gas cannot follow the potential, and screening is absent.
The reduced screening capability is seen quantitatively in the RPA calculation of the
dielectric function of the interacting electron gas, which now becomes
ks2 Dq 2
RPA (q, in ) = 1 + , (7.65)
q 2 |n | + Dq 2
where ks2 = 2e2 /0 is the Thomas-Fermi wavenumber. Similarly, the effective electron-
electron interaction becomes
e2
VqRPA (in ) =
0 q 2 RPA (q, in )
1
e2 Dq 2

2 2
= q + ks . (7.66)
0 |n | + Dq 2
For ||  Dq 2 the effect of the disorder is negligible, whereas for  Dq 2 the Coulomb
interaction is essentially unscreened. This is different from the case of a clean metal, where
Thomas-Fermi screening breaks down at the much higher frequency vF q.
One example where the modified screening capability of electrons plays a role is the
so-called zero bias anomaly in the tunneling density of states. The density of states can
be measured by a tunneling experiment, in which electrons from a reference conductor with
known (and constant) spectral density tunnel into the sample, see Fig. 7.13. Indeed, as shown
in Ex. 6.1, the current I through a tunneling barrier contacting the sample at position r is
proportional to the spectral density,
Z
I(r, eV ) 1 X 1
dA(r, ) 2 . (7.67)
V 4T
cosh [( eV )/2T ]
7.6. INTERACTIONS AND DISORDER 153
\Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \[\[\[ ]Y]Y]Y^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^]^]^]
\[Y\[Y\[Y\Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \[\[\[ ]Y]Y]Y^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^]^]^]
\[Y\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \[\[\[ ]Y]Y]Y^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^]^]^]
\[Y\[Y\[Y\Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \[\[\[ ]Y]Y]Y^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^]^]^]
\[Y\Y[[Y\ \Y[\Y[[Y\ \Y[\Y[[Y\ \Y[\Y[[Y\ \Y[\Y[[Y\ \Y[\Y[[Y\ \Y[\Y[[Y\ \Y[\Y[[Y\ ZYX \[\[[\ ]Y]Y]YZYX ^Y]^Y]]Y^ ZX ^Y]^Y]]Y^ ^Y]^Y]]Y^ ^Y]^Y]]Y^ ^Y]^Y]]Y^ ^Y]^Y]]Y^ ^Y]^Y]]Y^ ^Y]^Y]]Y^ ^]^]]^
\[Y\[Y\[Y\Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ I \Y[\Y[\Y[ XZYXZYXZY\[\[\[ ]Y]Y]YXZYXZYXZY^Y]^Y]^Y] XZXZXZ ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^]^]^]
disordered \[Y\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ XZYZYXZYX \[\[\[ ]Y]Y]YXZYZYXZYX ^Y]^Y]^Y] XZZXZX ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^]^]^] reference
\[Y\[Y\[Y\Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ ZYX \[\[\[ ]Y]Y]YZYX ^Y]^Y]^Y] ZX ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^]^]^]
conductor \[Y\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \[\[\[ ]Y]Y]Y^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^]^]^] electrode
\[Y\[Y\[Y\Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \[\[\[ ]Y]Y]Y^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^]^]^]
\[Y\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \Y[\Y[\Y[ \[\[\[ ]Y]Y]Y^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^Y]^Y]^Y] ^]^]^]
\[Y\Y[ \Y[ \Y[ \Y[ \Y[ \Y[ \Y[ \[ ]Y^Y] ^Y] ^Y] ^Y] ^Y] ^Y] ^Y] ^Y] ^]

V
Figure 7.13: Tunnel spectroscopy setup for a measurement of the spectral density of a disordered
conductor.

(At zero temperature, this simplifies to I/V A(r, eV ).)

For a non-interacting electron gas, the spectral density is essentially energy independent,
hence the tunnel current is proportional to the applied bias voltage V and independent
of the temperature T . For the disordered interacting electron gas, the situation is more
complicated. When an electron tunnels into the sample, it charges the sample locally. For
a cloud that is bigger than the mean free path l, dispersion of the charge cloud is a slow
process, which is determined by the diffusive motion of the electrons. Until the energy of the
charge cloud, which is a sum of kinetic and the Coulomb energy contributions, has dropped
below the bias voltage eV or the temperature T , the electron that tunnels into the metal has
to exist in a virtual state. The prolonged existence in a virtual state suppresses the tunneling
current or, in other words, the density of states. At low voltages and low temperatures, the
suppression should be maximal, the tunneling density of states being an increasing function
of V and T for small voltages and low temperatures.
A particularly illuminating qualitative picture of how interactions affect the density of
states and the conductivity in disordered metals is given by Aleiner, Altshuler, and Ger-
shenson [Waves in Random Media 9, 201 (1999)]. As we have seen in Ch. 4, the electron
density near an impurity will show oscillations with wavevector 2kF . Because electrons are
interacting, they see an oscillating potential (r) with wavevector 2kF . These oscillations
strongly modify the scattering properties of the impurity. The modification is strongest near
the Fermi energy because of the resonant scattering off the oscillating potential. Modified
154 CHAPTER 7. THE INTERACTING ELECTRON GAS

_`_`_`_`_`_`aaa _`_`_`aaa ___aaa c`bc`bb`c c`bc`bb`c c`bc`bb`c cbcbbc e`de`dd`e e`de`dd`e e`de`dd`e ededde f`f`f`f`f`f`ggg f`f`f`ggg fffggg
_`_`_`_`_`_`aaa _`_`_`aaa ___aaa cb`cb`cb`c`bc`bc`b c`bc`bc`b cbcbcb ed`ed`ed`e`de`de`d e`de`de`d ededed f`f`f`f`f`f`ggg f`f`f`ggg fffggg
_`_`_`_`_`_`aaa _`_`_`aaa ___aaa cb`c`bc`b c`bc`bc`b c`bc`bc`b cbcbcb ed`e`de`d e`de`de`d e`de`de`d ededed f`f`f`f`f`f`ggg f`f`f`ggg fffggg
_`_`_`_`_`_`aaa _`_`_`aaa ___aaa cb`cb`cb`c`bc`bc`b c`bc`bc`b cbcbcb ed`ed`ed`e`de`de`d e`de`de`d ededed f`f`f`f`f`f`ggg f`f`f`ggg fffggg
_`_`_`_`_`_`aaa _`_`_`aaa ___aaa b`cc`bc`b b`cc`bc`b b`cc`bc`b bccbcb d`ee`de`d d`ee`de`d d`ee`de`d deeded f`f`f`f`f`f`ggg f`f`f`ggg fffggg

Figure 7.14: Lowest order corrections to the single-particle Green function in a disordered inter-
acting electron gas. Left: Hartree diagram, right: Fock diagram. The shaded blocks represent the
diffusion propagator, see Fig. 6.5.

scattering properties include a modified scattering phase shift, and, hence, a modified impu-
rity contribution to the density of states and a modified scattering rate. Below, we focus on
the effect of interactions on the density of states, using diagrammatic perturbation theory. 5
A simple calculation illustrating interaction effects on transport is given in Ex. 7.5.
Our calculation of the density of states follows the original work of Altshuler and Aronov
[Sov. Phys. JETP 50, 968 (1979)] and Altshuler, Aronov, and Lee [Phys. Rev. Lett. 44,
1288 (1980)]. To lowest order in the interaction, this correction is given by Hartree and Fock
diagrams. The diffusive electron motion is reflected in the presence of the diffusion ladders
around the interaction line, which, in turn, represents the effective interaction of Eq. (7.66)
above. The two relevant diagrams for the correction to the single-particle Green function
are shown in Fig. 7.14.6
We consider the Fock diagram first. The correction to the single-electron Green function
shown in Fig. 7.14 reads
T XX 0
hGk,k (in )i = hG (in )i2 D(k, in ; k + q, im )2
V q m k,k
0
(Vq (in im ))hGk+q,k+q (im )i. (7.68)
5
A non-perturbative semiclassical theory of the density of states has been formulated by Shytov and
Levitov [JETP Lett. 66, 214 (1997)].
6
There are two additional diagrams containing cooperons instead of diffusons. These diagrams involve
interactions at high momentum transfers, unlike the Fock diagram shown in the figure, which is dominated
by interactions at low momenta. For that reason, the appropriate effective interaction in these diagrams is
more complicated than the RPA interaction V RPA that is used in the calculation below. For details, see
the article Electron-electron interaction in disordered conductors, by B.L. Altshuler and A.G. Aronov, in
the book Electron-electron interactions in disordered systems, edited by A.L. Efros and M. Pollak (North-
Holland, 1985).
7.6. INTERACTIONS AND DISORDER 155

C2

C1

C1

Figure 7.15: Integration contour for the calculation of the zero-bias anomaly to the tunneling
density of states.

The correction to the spectral density is found as

A(r, ) = 4Im GR (r, r; )

4T XX
0
= 2 Im hGk,k (in )i2 D(k, in ; k + q, im )2
V k,q m
0

(Vq (in im ))hGk+q,k+q (im )ii +i . (7.69)
n

where the extra factor two accounts for spin degeneracy.

Since the diffusion propagator is divergent for small momentum transfers, the main con-
tribution to the integration in Eq. (7.68) comes from small q. For these momenta, we can
use the screened interaction of Eq. (7.66).
A singular contribution to the density of states can come from the pole of the diffusion
propagator at small momentum and frequency. For the case n > 0, this pole exists for
m < 0 only. Replacing the summation over m by an integration in the complex plane,
deforming the integration contour as in Fig. 7.15, and performing the analytical continuation
in + i we find

1 X 1
Z
0R
A(r, ) = Im 2 d tanh(/2T )hGk,k ()i2
V i k,q (Dq 2 + i i)2 2
0A
(Vq ( + i))hGk+q,k+q ()i. (7.70)

Since the main dependence on and q comes from the pole of the diffusion operator, we
156 CHAPTER 7. THE INTERACTING ELECTRON GAS

set q = 0 and = in the arguments of the Green functions in Eq. (7.70). Performing the
summation over k and shifting integration variables we then find

2 X Vq ( + i)
Z  
A(r, ) = Im d tanh . (7.71)
V q 2T (Dq 2 i)2

In three dimensions, the summation over q and the integration over are, in fact, divergent
for large q and large . The divergence is not physical, since our approximations break down
when vF q and become of order unity. In order to arrive at a meaningful result that
describes the temperature and energy dependence of the density of states, we subtract A
in the limit 0, T 0 (limits taken in this order). In Eq. (7.71) this amounts to the
replacement of tanh[( )/2T ] by tanh[( )/2T ] sign (). For the interaction, we
substitute the screened Coulomb interaction in the presence of disorder, Eq. (7.66) above.
First performing the remaining summation over q, one finds
Z 


1
A(r, ) = Im d tanh sign ()
4D 3/2 2T
" #
1 1
p . (7.72)
i Dks2 i

We omit the second term between the square brackets, which gives a vanishing contribution
to the spectral density if , T  Dks2 . For the remaining integration, we find

T 1/2
A(r, ) = f (/T ), (7.73)
2(~D)3/2

where  
dy sinh y
Z
f (x) = 1 . (7.74)
0 2y 1/2 cosh y + cosh x
Asymptotically, the function f (x) behaves as

2
f (x) x x3/2 if x  1,
24
f (x) 1.072 if x  1. (7.75)

You verify that, indeed, A 0 if we take the limit 0, followed by T 0.

For two dimensional samples and one dimensional samples, the calculation of the density
of states is very similar. The first difference is that the Coulomb interaction has a different
7.6. INTERACTIONS AND DISORDER 157

form in one and two dimensions,

h i1
(e2 /20 ) q + (e2 /0 ) Dq2 2 if d = 2,
|n |+Dq
Vq (in ) = h i 1 (7.76)
(e2 / ) 4 + 2(e2 / ) Dq2 if d = 1.
0 ln(q 2 a2 ) 0 |n |+Dq 2

Here is the density of states per spin direction and per unit area (in two dimensions) or
unit length (in one dimension), whereas a is the thickness or radius of the sample. In quasi
two-dimensional samples, = 3 a, where 3 is the three-dimensional density of states. With
this modification, the calculation proceeds as in the three dimensional case.
In two dimensions, the final integration over and summation over q is logarithmically
divergent. Performing the summation over q first and truncating the -integration at
1/ , we find
Z

 
1 d +
A(r, ) = ln tanh + tanh
4D 0 De2 /0 2T 2T
 
1 max(, T )
ln 2 2 ln [ max(, T )] . (7.77)
4D D (e /0 ) 2

Similarly, in the one-dimensional case one has

1/2 Z
Dks2

aks dy sinh y
A(r, ) = ln , (7.78)
2 2T D max(, T ) 0 y 3/2 cosh y + cosh(/T )

where ks is the three-dimensional Thomas-Fermi wavenumber.

The contribution from the Hartree diagram is similar. The transferred momentum q in
interaction factor of the Hartree diagram does not have to be small for the momentum in
the diffusion propagators to be small. Since the Hartree diagram involves large momentum
exchanges in the interaction matrix element, the precise magnitude depends on the details
of the short-range part of the Coulomb interaction. In that case we cannot use the sim-
ple screened form (7.66) above, but need information about the microscopic details of the
sample. For the discussion here, we limit ourselves to the statement that the energy and
temperature dependence of A is still the same as that of Eq. (7.73), whereas the prefactor
depends on sample details. For details of this calculation we refer to the article Electron-
electron interaction in disordered conductors, by B.L. Altshuler and A.G. Aronov, in the
book Electron-electron interactions in disordered systems, edited by A.L. Efros and M. Pol-
lak (North-Holland, 1985).
158 CHAPTER 7. THE INTERACTING ELECTRON GAS

7.7 Exercises
Exercise 7.1: Coulomb interactions in two dimensions

In certain materials, such as a quantum well at the interface of the semiconductors GaAs
and GaAlAs, the electrons are confined to two spatial dimensions only.
For such a system, the non-interacting part of the Hamiltonian can be written
0 = k k

X
H k , (7.79)
k

where the summation over wavevectors k is restricted to wavevectors in the plane of the
two-dimensional electron gas. The creation operator k
creates an electron with spin and
orbital wavefunction
1
k (r, z) = 1/2 eikr (z), (7.80)
A
where r is the coordinate in the plane of the two-dimensional electron gas, and z is the
transverse coordinate, and A is the area of the two-dimensional electron gas. For most
practical purposes, the function (z) can be approximated by the delta function (z).
The interaction between the electrons in the two-dimensional electron gas is the three-
dimensional Coulomb interactions,
1
V (r) = . (7.81)
40 r |r|
The relative permittivity r appears because the two-dimensional electron gas is not sur-
rounded by vacuum, but by a semiconducting material. For this system, write down the
interaction Hamiltonian for the Coulomb interaction using the momentum representation.
Find an explicit expression for the Coulomb-interaction matrix element Vq in two dimensions.

Exercise 7.2: Four-vertex formalism

Abrikosov, Gorkov, and Dzyaloshinski discuss a different diagrammatic formalism to describe

electron-electron interactions. The diagrams for the direct and the exchange interaction are
combined into one diagram, as shown in Fig. 7.16.
Discuss the diagrams contributing to the self-energy to second order in the interaction
using this formalism. Also discuss the Hartree-Fock approximation. Does the simplest four-
vertex formulation of an effective interaction agree with the RPA approximation? Why
(not)?
7.7. EXERCISES 159

Figure 7.16: Diagrammatic representation of both the direct and exchange parts of the electron-
electron interaction, following the notation of the book by Abrikosov, Gorkov, and Dzyaloshinski.

Exercise 7.3: Plasma oscillations

It is possible to find the plasma frequency p from simple classical considerations. Consider
an electron gas in a rectangular box of length Lx and cross section A. Since the mass of
the ions is much larger than the mass of the electrons, the ions can be treated as an inert
positive background charge that compensates the charge of the electrons. The electron gas
can be set in oscillating motion by translating it a distance x in the x-direction, leaving
the ion background fixed. Show that the frequency of this oscillation is precisely the plasma
frequency p .?

Exercise 7.4: Plasma oscillations in two dimensions

In this exercise we return to the case of a two-dimensional electron gas that exists at a
quantum well in a semiconductor heterostructure.

(a) Calculate the dielectric response function (q, ) for a two-dimensional electron gas in
the RPA approximation.

(b) From the zeroes of the dielectric response function, calculate the dispersion relation of
plasma oscillations in the two-dimensional electron gas. Discuss for what wavenumbers
plasma oscillations are damped by Landau damping (excitation of particle/hole pairs).

(c) Calculate the screened Coulomb interaction in a two-dimensional electron gas in the
RPA approximation.

(d) Show that, at small wavevectors q  kF and low frequencies ~  F the screened
interaction is given by
e2
VqRPA = , (7.82)
20 r (q + ks )
160 CHAPTER 7. THE INTERACTING ELECTRON GAS

where ks is the two-dimensional Thomas-Fermi screening wavenumber. Calculate ks

for the case of a GaAs/GaAlAs heterostructure, for which the effective electron mass
is m = 0.067m, relative permittivity r = 13, and electron density n = 1015 m2 . How
does it compare to the Fermi wavelength for this system?

Exercise 7.5: Interacting wire with impurity

In this exercise youll consider a one-dimensional wire with an impurity at x = 0. The

impurity is characterized by its transmission amplitude t. Here youll calcualte how t is
modified by the presence of interactions. This exercise is based on the article Tunneling in
one-dimensional non-Luttinger electron liquid by Matveev, Yue, and Glazman [Phys. Rev.
Lett. 71, 3351 (1993)].
Without interactions, the retarded Green function for the wire reads
1 h ik(xx0 ) 0 0
i
G0R (x, x0 ; ) = te (x) + (eik|xx | + reik(x+x ) )(x) if x0 < 0,
i~vk
1 h ik(xx0 ) 0 0
i
G0R (x, x0 ; ) = te (x) + (eik|xx | + reik(x+x ) )(x) if x0 > 0, (7.83)
i~vk
see Ex. 4.5. Here r and t are the reflection amplitude and transmission amplitude of the
impurity, and k is the positive solution of k = . The corresponding wavefunctions are

0k (x) = (x)(eikx + reikx ) + (x)teikx ,

0k (x) = (x)teikx + (x)(eikx + reikx ). (7.84)

The transmission amplitude t gives information about the conductance of the wire. The goal
of this exercise is to find how the interactions change the transmission amplitude t. Well
restrict ourselves to the case of spinless electrons.

(a) The lowest order in perturbation theory corresponds to the Hartree-Fock approxima-
tion, but without the self-consistency condition. Because the Hartree-Fock approxi-
mation interactions are described by means of an effective potential U , it will be
sufficient to find the wavefunctions k and k in the presence of this potential. Show
that the wavefunction k satisfies the equation
Z Z
0 0R
k (x) = k (x) + dyGk (x, y) dzU (y, z)k (z), (7.85)
7.7. EXERCISES 161

where the potential U (x, y) is the sum of Hartree and Fock contributions,

U (x, y) = UH (x)(x y) UF (x, y), (7.86)

which read
Z
UH (x) = dyV (x y)h (y)(y)i

1
Z Z
= dk f (k0 ) dyV (x y)(|k0 (y)|2 + |k0 (y)|2 ),
0
(7.87)
2
UF (x, y) = V (x y)h (y)(x)i

1
Z
= dk 0 f (k0 )V (x y)(k0 (y)k0 (x) + k0 (y)k0 (x)). (7.88)
2
Here V is the interaction potential.

If we are interested in the lowest order perturbation theory in V only, we can replace the
wavefunctions by 0 in the r.h.s. of Eq. (7.85) and in Eqs. (7.87) and (7.88).

(b) Show that the interactions modify the transmission amplitude as t t + t, where

1
Z
t = dy t(1 + re2ik|y| )UH (y) + ((y)(eiky + reiky ) + (y)teiky )
i~vk
Z 
ikz ikz ikz
dz UF (y, z)((z)(e + re ) + (z)te ) . (7.89)

(c) The Hartree and Fock potentials contain an integration over all wavenumbers k 0 for
which states are occupied. We first calculate the shift t(k 0 ) from the states with
wavevector k 0 and then integrate over k 0 . Calculating t(k 0 ) youll find a non-singular
contribution, together with a contribution that is singular as k k 0 . Were inter-
ested in the singular contribution only, because that contribution will determine the
temperature dependence of t. Show that the singular contribution reads
1
t(k 0 ) = |r|2 t(V2k + V2k 2V0 )
2~vk (k k 0 )
1
= (1 |t|2 )t(V2k + V2k 2V0 ), (7.90)
2~vk (k k 0 )

where Vq is the Fourier transform of the interaction potential.

162 CHAPTER 7. THE INTERACTING ELECTRON GAS

(d) The total change t is now found upon multiplication of t(k 0 ) by f (k0 )/2, followed
by integration over k 0 . You may cut off the logarithmic divergence at small k 0 at
k k 0 ' , where 1/ is of the order of the Fermi wavelength or the spatial range of
the potential V , whichever is larger. Calculate t and discuss its asymptotic properties
for low temperatures and electrons close to the Fermi surface.

Transport at low temperatures corresponds to the limit k kF , T 0. In this limit, the

first-order perturbation theory shift to the transmission amplitude is logarithmically diver-
gent. This situation is reminiscent of the Kondo problem, where one finds that the rate for
scattering off a magnetic impurity diverges logarithmically upon lowering the temperature.
For the Kondo problem, good insight could be obtained by Andersons poor mans scaling
method. In this approach, the cut-off is gradually decreased, leading to an effective theory
with a smaller cut-off . This procedure can be repeated until max(|k kF |, T /vF ). In
the language of perturbation theory, poor mans scaling amounts to summing up the most
divergent contributions to t in each order of perturbation theory.

(e) Calculate the shift to the transmission amplitude t that arises from taking into account
only states with kF < k 0 < kF ( + d), with d negative, and write your result
in the form of a differential equation that expresses dt/d in terms of t and .

(f) Solve the differential equation you find under (e) to find how the effective zero tem-
perature transmission amplitude depends on |k kF | for k kF | 0. Express your
answer in terms of the transmission amplitude t0 at a reference cut-off 0 .

The result you find in (f) is very interesting: For a repulsive interaction, you find that t 0
if k kF at zero temperature, irrespective of the bare value of t for k far away from
kF . Hence, even the weakest impurity will turn a one-dimensional wire into an insulator at
low temperatures. Moreover, you find that t approaches zero as a power law of |k k F |,
with an exponent that depends on the interaction. Such power laws are typical of one-
dimensional interacting metals. We return to this issue when we discuss the Luttinger
Liquid description of one-dimensional metals.
Chapter 8

Magnetism

In this chapter we study the magnetic excitations in a solid. We start with a calculation
of the magnetic susceptibility , which is the coefficient of proportionality between the spin
polarization induced by a magnetic field and the magnetic field h. The susceptibility not
only tells us how easy it is to magnetize a solid, it also points to the onset of a magnetic
phase. This happens at the point where the susceptibility diverges. A divergent susceptibility
signals an instability: a spin polarization is formed for arbitrarily weak magnetic field.
The magnetic instability is governed by a competition between two effects. On the one
hand, the kinetic energy of the electrons is minimal if there is no net spin polarization,
because then all energy levels k can be doubly occupied. On the other hand, the short-
range part of the electron-electron interaction favors a magnetic state. This is because
electrons with the same spin have a strongly reduced repulsion since the Pauli principle
already guarantees that they are spatially separated.
We first consider the susceptibility for a gas of interacting electrons with a parabolic
dispersion (free electrons). This is the example that we studied in detail in the previous
chapters. A calculation within the RPA approximation shows that electron-electron inter-
actions enhance , but also that they do not lead to a divergence. In other words, for free
electrons, the competition between kinetic energy and interaction energy always favors the
kinetic energy, and, hence, forbids a spontaneous spin polarization.
The picture is different once the ionic lattice is taken into account. Band structure effects
turn out to be crucial for the occurrence of a ferromagnetic instability. In a narrow band, the
kinetic energy cost for polarizing electron spins is strongly reduced, and a magnetic instability
becomes possible. Depending on the material one looks at, one may have a ferromagnetic
instability (uniform magnetization) or a spin-wave material (nonuniform magnetization).
One example of a spin-wave material is an antiferromagnet, where the spin polarization has
opposite direction on neighboring lattice sites.

163
164 CHAPTER 8. MAGNETISM

8.1 Magnetic susceptibility

In order to make the above picture more quantitative, we consider an electron gas in the
presence of a magnetic field h. We are interested in the spin density s(r),
1 X
s (r) = (r)( ),0 0 (r), = x, y, z, (8.1)
2 ,0

where is the Pauli matrix. The response of a magnetic material to an applied magnetic
field is characterized by the magnetic susceptibility (r, t; r0 , t0 ), which relates the change in
spin density at position r and time t to the magnetic field at position r0 and time t0 ,
XZ Z
0
s (r, t) = dr dt0 (r, t; r0 , t0 )h (r0 , t0 ). (8.2)

Note that is a tensor: in principle, a magnetic field h can cause a spin density in a
different direction than the direction of h. The magnetic field h enters into the Hamiltonian
through the Zeeman coupling,
Z
~
H1 = B g drh(r, t) s(r), (8.3)
2
where B = e~/2mc is the Bohr magneton and g = 2 the electron g factor. For electrons, it
also enters the Hamiltonian through the kinetic energy (1/2m)(p eA/c)2 . We neglect the
role of the vector potential in the discussion below.
According to the Kubo formula, the spin susceptibility is equal to the retarded spin-spin
correlation function,
B g
(r, t; r0 , t0 ) = i (t t0 )h[s (r, t), s (r0 , t0 )] i. (8.4)
2
Note the absence of a minus sign on the r.h.s. of Eq. (8.4) because of the minus sign in
front of the magnetic field in the expression (8.3) for the Zeeman energy. We may perform
a Fourier transform to find the susceptibility in frequency representation,
XZ
s (r, ) = dr0 (r, r0 ; )h (r0 , ), (8.5)

where
B g
Z
0
(r, r ; ) = i dteit h[s (r, t), s (r0 , 0)] i. (8.6)
2 0
8.1. MAGNETIC SUSCEPTIBILITY 165

For a spatially homogeneous system, we may perform one further Fourier transform to the
coordinate r, which gives a susceptibility (q, ),

1
Z
0
(q, ) = drdr0 eiq(rr ) (r, r0 ; ). (8.7)
V
For the calculation of (q, ) it is useful to write the spin density in terms of the plane
wave basis,
B g
(q; t, t0 ) = i(t t0 ) h[sq, (t), sq, (t0 )] i, (8.8)
2V
where
1 X
sq, = k, ( ),0 k+q,0 . (8.9)
2 0 k,,

In practice, the x and y labels of the tensor are often replaced by labels + and
, referring to the linear combinations

s = s1 is2 . (8.10)

These combinations are expressed in terms of raising and lowering operators for the spin,
which are much easier to calculate. In particular, we have the susceptibility
B g
+ (r, t; r0 , t0 ) = i (t t0 )h[s (r, t), s+ (r0 , t0 )] i. (8.11)
2
The susceptibility + is known as the transverse susceptibility, while zz is known as the
longitudinal susceptibility. For an isotropic system, the tensor is proportional to the
unit tensor. In that case one has
1
= + , , = x, y, z. (8.12)
2
A calculation of the susceptibility for non-interacting electrons is very similar to the
calculation of the polarizability that we discussed in Sec. 6.1. One finds
B g
h[k, (t)k+q, (t), k 0 , (0)k0 q, (0)] i.
X
+ (q, t) = i (t) (8.13)
2V 0 k,k

Diagrammatically, the susceptibility for non-interacting electrons can be represented by the

diagram shown in Fig. 8.1. The white dots are spin vertices. For the transverse susceptibility
+ , they enforce that the ingoing and outgoing spins have different sign, see Fig. 8.1.
166 CHAPTER 8. MAGNETISM

k k
+=
k+q k+q

Figure 8.1: Diagrammatic representation of the transverse spin susceptibility + for non-
interacting electrons.

As a matter of fact, we can find + without doing a calculation if we compare Eqs. (8.13)
and the expression for the polarizability of the non-interacting electron gas, Eq. (6.18). We
then immediately conclude

B g X nF (k ) nF (k+q )
0+ (q, ) = , (8.14)
2V k k k+q + + i

where nF is the Fermi function and is a positive infinitesimal.

In the limit q 0, 0 of a static and uniform magnetic field, one has + B g/2,
where is the density of states per unit volume and per spin direction. This result is known
as the Pauli susceptibility. The Pauli susceptibility accounts for the fact that it costs a
finite kinetic energy to have a finite spin density. The value of the Pauli susceptibility then
follows from the competition of the magnetic energy gain (proportional to the spin density
and the magnetic field) and the kinetic energy loss (proportional to the square of the spin
density).
The imaginary part of the susceptibility 0 (q, ) characterizes energy dissipation through
magnetic excitations. The support of the imaginary part of 0 is the same as that of the
imaginary part of the polarizability of the non-interacting electron gas, see Fig. 6.2.
For the interacting electron gas, we can use the diagrammatic theory to compute the
susceptibility. Some diagrams are shown in Fig. 8.2. They can be rearranged in three
steps. First, we incorporate all interaction lines that connect the upper or lower fermion
line to itself into single-particle Green functions. Second, we replace the interaction by the
screened interaction by inserting any number of bubbles into the interaction. Below, the
single-electron Green function will be calculated using the Hartree-Fock approximation, the
effective interaction in the RPA. Third, we write the diagram in terms of a vertex correction
as in Sec. 5.6, see Fig. 8.2. The renormalized vertex, which is indicated by a black dot, then
contains all contributions from interaction lines that connect the upper and lower fermion
lines.
In the spirit of the RPA approximation, we keep only ladder diagrams for the vertex
correction, see Fig. 8.3. In this approximation, the calculation has become very similar to
8.1. MAGNETIC SUSCEPTIBILITY 167
iOhiOh iOhiOh ihih
+=
k
ihOihOiOhiOh ihih k

k+q
ihOiOh iOhiOh ihih k+q
= + + +

+ + + + + ...

m
k
=
k+q
m+ n

Figure 8.2: Diagrammatic representation of the contributions to the transverse spin susceptibility
+ for interacting electrons, up to second order in the interaction. The arrows indicate the
spin direction, while denotes an additional summation over spin. Note that spin is conserved
throughout the upper and lower fermion lines.

that of the current-current correlation function of Sec. 5.6. Denoting the renormalized vertex
by (k, im ; k + q, im + in ) and the screened interaction by Vqeff (in ), we have

T XX
(k, im ; k + q, im + in ) = 1 Gk0 ,k0 (ip )Gk0 +q,k0 +q (ip + in )
V k0 p
eff 0 0
Vkk 0 (im ip )(k , ip ; k + q, ip + in ). (8.15)

Note that the effective interaction depends on the Matsubara frequency difference im ip ,
in contrast to the bare interaction which is frequency independent. Then, in terms of the
renormalized vertex, the spin susceptibility is calculated as
T B g X X
RPA
+ (q, in ) = (k, im ; k + q, im + in )
2V k m
Gk,k (im )Gk+q,k+q (im + in ). (8.16)

The analytical continuation in + i is done as in Sec. 5.6. The result is

B g
Z X
+ (q, + i) = d tanh(/2T )
8iV k
(k, ; k + q, + )GR
 RR R
k,k ()Gk+q,k+q ( + )
AR (k, ; k + q, + )GA R
k,k ()Gk+q,k+q ( + )
168 CHAPTER 8. MAGNETISM

p p
m k m k k k k k" k
= + + + ...
k+q k+q
m+ n m+ n k+q
k+q +
k+q k"+q k+q
p +n m n

p
m k k k
= +
k+q
m+ n k+q
k+q
p +n

Figure 8.3: Diagrammatic representation of the RPA approximation for the transverse spin sus-
ceptibility + .

+ AR (k, ; k + q, )GA R
k,k ( )Gk+q,k+q ()

AA (k, ; k + q, )GA A

k,k ( )Gk+q,k+q () .
(8.17)

For the full frequency and momentum dependent screened interaction, the RPA equation
for the renormalized vertex cannot be solved in closed form. Here, well adopt the Thomas-
Fermi approximation and replace the screened interaction by the interaction at q = 0, = 0,

e2
VqTF (in ) , (8.18)
0 ks2

where ks2 = 2e2 /0 is the Thomas-Fermi wavenumber. This approximation corresponds to
a delta-function interaction in real space, V TF (r) (2)1 (r). With this approximation,
calculation of the renormalized vertex is straightforward, and we find the simple result

(k, im ; k + q, im + in )
" #1
T X X
= 1+ Gk0 ,k0 (ip )Gk0 +q,k0 +q (ip + in ) . (8.19)
2V 0 p
k

The summation in the denominator has the same expression as the susceptibility for non-
interacting electrons, the only difference being that the single-electron Green functions have
been replaced by their values in the interacting electron liquid. In other words, this is the
susceptibility for the interacting electron liquid without vertex renormalization. Using the
8.1. MAGNETIC SUSCEPTIBILITY 169

Hartree-Fock approximation for the single-particle Green function, we write

B gT X X
0,HF
+ (q, in ) = Gk,k (im )Gk+q,k+q (im + in )
2V k m

B g X nF (HF HF
k, ) nF (k+q, )
= , (8.20)
2V k HF HF
k, k+q, + in

so that we can express the full transverse susceptibility in the Thomas-Fermi approximation
as
0,HF
+ (q, )
TF
+ (q, ) = . (8.21)
1 1B g 0,HF
+ (q, )

For a uniform static magnetic field, one has 0,HF

+ (0, 0) = B g/2, where the density of
states should be interpreted as the density of poles of the full Green function, i.e., the
density of Hartree-Fock energy levels. Since this is the same density of states as the one that
occurs in the screened interaction, one has

TF
+ (0, 0) = B g. (8.22)

We conclude that, according to the Thomas-Fermi approximation, in an interacting electron

gas the electron-electron interactions increase the spin susceptibility by a factor two compared
to the case of the non-interacting electron gas.1
The enhancement of the susceptibility by the interactions follows from the fact that the
repulsive interaction between the electrons is lowered if a finite spin density is created. After
all, the Pauli principle forbids that electrons of the same spin are at the same location, so
that they have a smaller electrostatic repulsion. The interaction energy gain is proportional
to the square of the spin density, so that the net energy cost for polarizing the system is
reduced with respect to the case of non-interacting electrons.
If the interaction energy gain is larger than the kinetic energy cost, a spin polarization
will form spontaneously. This instability is known as the Stoner instability. We see that
for the standard case of a parabolic band the Stoner instability does not occur. The situation
becomes different once the precise band structure of the material is taken into account. The
next section discusses a simple model for this.
1
The Thomas-Fermi approximation is not quantitatively correct. A more correct treatment of the RPA
approximation which takes into account the momentum dependence of the effective interaction gives
B g
RPA
+ = ,
2 (2/3 2 )2/3 rs ln(1 + (3 2 /2)2/3 /rs )
see Ex. 8.1.
170 CHAPTER 8. MAGNETISM

8.2 Hubbard model

The approximations of the previous subsection are appropriate for the standard parabolic
band of free electrons. In real materials, the electrons occupy bands, and the band structure
needs to be taken into account for a calculation of the magnetic susceptibility. A simple
model to take the band structure into account is the Hubbard model.
For the construction of the Hubbard model, we consider electrons in an energy band with
Bloch energies k described by the Hamiltonian

0 = k k k .
X
H (8.23)
k

For such a band, wavefunctions can be represented in the basis of Wannier wavefunctions at
lattice site ri ,
eikri k (r),
X
i (r) = N 1/2 (8.24)
k

where the summation is restricted to wavevectors in the first Brillouin zone and N is the
number of lattice sites. The Wannier wavefunctions at different lattice sites are orthogonal.
The Coulomb interaction can be written in terms of the basis of Wannier wavefunctions.
The corresponding matrix elements are
Z
Vij,i0 j 0 = dr1 dr2 i (r1 )j (r2 )V eff (r1 r2 )j 0 (r2 )i0 (r1 ), (8.25)

where V eff is the screened Coulomb interaction. The approximation of the Hubbard model
consists in neglecting all interaction matrix elements except for the matrix element U = Viiii .
The justification for this approximation is that the screened interaction is short range, while
the Wannier wavefunctions are mainly localized at one lattice site. This approximation leads
to the interaction Hamiltonian

1 = U/2
X
H ni ni0 , (8.26)
i 0

where ni = i

i is the Wannier number operator and

i = N 1/2 k
X
eikri (8.27)
k

is the Wannier annihilation operator.

8.2. HUBBARD MODEL 171

The interaction Hamiltonian can be further simplified if we use n2i = ni , so that

1 = U
X X
H ni ni + (U/2) (ni + ni ). (8.28)
i i

The second term is nothing but a shift of the chemical potential and will be neglected.
Fourier transforming the first term in Eq. (8.28) we then arrive at

1 = U k+q, k 0 , k0 q, .
X
H (8.29)
N k,k0 ,q k,

Let us now calculate the spin susceptibility for the Hubbard model. The calculation
proceeds along the same lines as that of the previous section. The only differences are that,
for the Hubbard model, the wavevector summation is restricted to the first Brillouin zone,
and that the strength of the interaction is set independently by the interaction constant U .
The result for the transverse susceptibility is
0,HF
+ (q, )
+ (q, ) = 0,HF
, (8.30)
1 2U
Bg
+ (q, )

where
B g X nF (HF HF
k, ) nF (k+q, )
0,HF
+ (q, ) = . (8.31)
2N k HF HF
k+q, k, + + i

For the Hubbard model, the Hartree-Fock energy levels have a very simple connection to
the bare energy levels k ,
1 X
HF
k = k + U n
, n
= nF (HF
k )i. (8.32)
N k

In the absence of a spontaneous magnetization, the levels of electrons with spin up and
spin down experience the same uniform shift, which can be incorporated into the chemical
potential. In particular, this implies that the support of the imaginary part of 0,HF + and,
hence, of the imaginary part of TF
+ , is the same as that of 0
+, which is shown in Fig. 6.2.
In the presence of a spontaneous magnetization, the Hartree-Fock energy levels for electrons
with spin up and spin down experience different shifts. We return to this case in Sec. 8.5.
As before, the limit of large wavelength and low frequency has 0,HF + (0, 0) = B g/2,
where is the density of states per spin direction and per lattice site. Hence
B g/2
TF
+ (0, 0) = . (8.33)
1 U
172 CHAPTER 8. MAGNETISM

We can use these results to find the interaction correction to the Free energy for the
Hubbard model. Repeating the analysis of Sec. 7.5, we find the result
Z
d X
F F0 = coth(/2T )Im ln(1 2U 0,HF (q, )/B g). (8.34)
4 q

This result may be analyzed using the detailed expressions for 0,HF of Ex. 6.2. For low
temperatures, the main contribution comes from frequencies around = 0, and one finds
that the specific heat is enhanced close to the Stoner instability,
CV T ln(1 U ). (8.35)
This is to be compared to the usual specific heat of the noninteracting electron gas, which
is proportional to T .

8.3 The Stoner instability

As long as the interaction constant U is small enough, the susceptibility remains finite. A
finite susceptibility implies that it requires a finite value of the magnetic field h to create
a spin polarization. However, if the interaction strength increases, the susceptibility in-
creases. If the static susceptibility (q, 0) diverges, an instability occurs, and a spontaneous
magnetization will form.
From our general result (8.30) for the Hubbard model we find that the instability criterion
2
is
U = Uc (q) = B g/20,HF (q, 0). (8.36)
This is the Stoner criterion.
For zero wavevector q = 0, the Stoner criterion describes the ferromagnetic instability,
which occurs at critical interaction strength U = 1. For a parabolic band, the instability
occurs first at q = 0, since 0,HF (q, 0) is a monotonously decreasing function of |q|, cf. Ex.
6.2. However, in the presence of a more complicated band structure, instabilities at finite
wavenumbers are also possible. An instability at finite q leads to the formation of a so-
called spin density wave state, as was first pointed out by Overhauser [Phys. Rev. 128,
1437 (1962)]. As an example, we consider the band
X
k = 2t cos k a, (8.37)
=x,y,z

2
For finite frequencies, the denominator of Eq. (8.30) is generally complex, so that one can rule out the
existence of zeroes. Physically, one expects the instability to happen at zero frequency, since an instability
at nonzero frequency corresponds to a time-dependent polarization, which is damped. Response at nonzero
frequencies is studied in a later section.
8.4. NEUTRON SCATTERING 173

which corresponds to electrons on a simple cubic lattice with lattice constant a and hopping
amplitude t, see Ex. 8.7. For this band, there is a competition between the ferromagnetic
instability at wavevector q = 0 and an antiferromagnetic instability at wavevector q = q A =
(/a)(1, 1, 1) at the far corner of the first Brillouin zone. You verify that one has

B g ()
Z
0,HF
(qA , 0) = d . (8.38)
2 6t

For chemical potentials close to the band edges 6t, the instability occurs at wavevector
q = 0, while for a chemical potential close to the band center at = 0, the instability
occurs at q = qA . For this simple cubic band structure, one has 0,HF (qA , 0) as
0, so that the antiferromagnetic instability occurs already for arbitrarily small repulsive
interaction!
Note that for interaction strength U > Uc , our expression (8.30) for the spin susceptibility
in the Thomas Fermi approximation is, again, finite. However, this result is unphysical, since,
for U > Uc the ground state has a spontaneous spin polarization (or a spin-density wave
state). Hence, we conclude that Eq. (8.30) is valid for all U < min q Uc (q), whereas a state
with spontaneous magnetization is formed for U > Uc . A description of the ferromagnetic
state is given in Sec. 8.5 below.

8.4 Neutron scattering

Neutrons have spin 1/2. The neutron spins interact with the spins of electrons in a solid
through the magnetic dipole interaction. Hence, the neutron scattering cross section is
sensitive to the magnetic structure of a solid.
The inelastic scattering cross section d2 /dd tells us the amount of scattered neutrons
that emerge from the sample at a given energy ~ and solid angle . In order to describe
the scattering of slow neutrons, we start from the first Born approximation for the neutron
scattering cross section,3
 2
d2 1 X k0 M
0
|hk f , f |H 0 |ki , ii|2 ( + Ei Ef ),

= (8.39)
dd 2 f, , k 2
f i

where f and i denote the outgoing and incoming spin states of the neutron, k is the
wavevector of the incoming neutron, k0 is the wavevector of the outgoing neutron, M is the
3
This result is derived in standard texts on quantum mechanics, e.g., Quantum Mechanics, by A. Messiah,
North-Holland (1961).
174 CHAPTER 8. MAGNETISM

reduced mass of the neutron, i and f refer to initial state and final state, respectively, H 0 is
the interaction between the neutron and the target, and the outer brackets h. . .i indicate a
thermal average over the initial states i of the target. The prefactor 1/2 is added since we
average over the possible incoming spin states. For the interaction between an electron and
a neutron we take the first quantization form4
H 0 = V0 (rn r) + V1 (rn r)sn s, (8.40)
where rn and sn are the position and spin of the neutron, while r and s are the position and
spin of the electron. Using the second quantization language for the electrons, H 0 can be
rewritten as
Z Z
0
H = drV0 (rn r)n(r) + sn drV1 (rn r)s(r), (8.41)

where n(r) and s(r) are the second quantization operators for the electron number density
and spin density. Writing q = k k0 , we then find
Z
0 0
hk f , f |H |ki , ii = drn dreiqrn V0 (rn r)f ,i hf |n(r)|ii
1X
Z
+ drn dreiqrn V1 (rn r)( )f i hf |s(r)|ii
2
V1q X
= V0q hf |q |iif ,i + ( )f i hf |sq |ii, (8.42)
2

where V0q and V1q are the Fourier transforms of V0 and V1 , respectively. Next, we write the
delta function in Eq. (8.39) as an integration over time, absorb the factors exp(iEi t) and
exp(iEf t) into a time-shift of the operators n and s, and perform the summation over final
states. The result is5
 2 0
d2 |V1q |2
 
M k
Z
it 2
= dte |V0q | hnq (0)nq (t)i + hsq (0) sq (t)i .
dd 2 2k 4
The spin-spin correlation function appearing in the spin-dependent part of the inelastic
neutron scattering cross section is nothing but a greater Green function. For an isotropic
system, it can be written in terms of the spin raising and lowering operators s+ and s ,
3
hsq (0) sq (t)i = hs,q (0)s+,q (t)i. (8.43)
2
4
For more details: see L.D. Landau and E.M. Lifschitz, Quantum mechanics, non-relativistic theory,
Addison and Wesley (1958).
5
See, e.g., the book by Doniach and Sondheimer for details.
8.5. THE FERROMAGNETIC STATE 175

Using the general relation (2.49) between Green functions, the Fourier transform with respect
to time can be written in terms of the imaginary part of the retarded spin density correlation
function,
3 Im R+ (q, )
Z
dteit hsq (0) sq (t)i = . (8.44)
2B g 1 e~/T
The minus-sign in the denominator appears because is a correlation function of spin den-
sities, which are even in fermion creation and annihilation operators.
The approach to the instability point leads to a strong enhancement of the neutron
scattering cross section. For small q and and a parabolic band, one has (cf. Ex. 6.2)

B g/8qvF
Im R
+ (q, ) = . (8.45)
(1 U )2 + (U /4qvF )2

Thus, as the ferromagnetic instability is approached, inelastic neutron scattering is strongly

enhanced for low frequencies. The physical origin of the enhanced neutron scattering rate
at low temperatures is the slowing down of spin relaxation at the approach of the Stoner
instability.

8.5 The ferromagnetic state

Beyond the instability point a spontaneous magnetization forms. In principle, the magne-
tization can point in any direction, although, for each sample, a specific direction will be
selected. Here, we assume that the magnetization points in the positive z-direction.
The main difference between the ferromagnetic state and the paramagnetic state is that,
in a ferromagnet, the Hartree-Fock energy levels are different for up spins and down spins,
see Eq. (8.32). For the Hubbard model, the difference between levels for spin up and spin
down does not depend on the wavevector,

= HF HF
k k = U (
n n
). (8.46)

The existence of the gap implies that electrons with spin up have a larger Fermi momentum
than electrons with spin down, see Fig. 8.4, which, in turn reflects the fact that there is a net
spin polarization. The electrons with spin parallel to the magnetization direction are referred
to as majority electrons, whereas those with opposite to the magnetization direction are
called minority electrons.
One can calculate self-consistently from the requirement that the total number of
particles does not change upon the formation of the magnetic moment. Together with Eq.
(8.32), one can then solve for the energy gap . Nontrivial solutions exist for U > 1 only.
176 CHAPTER 8. MAGNETISM

kF
kF

Figure 8.4: Support of the imaginary part of + for a ferromagnet and spin wave dispersion.

The difference between Hartree-Fock energy levels for different spin directions leads to a
modification of the expression for the transverse spin susceptibility,

B g X nF (HF HF
k, ) nF (k+q, )
0,HF
+ (q, ) = . (8.47)
2N k k k+q + + i

The support of the imaginary part of 0,HF and, hence, of , is affected by the appearance of
the gap : The imaginary part of + is nonzero if there are excitations of pairs of a particle
and a hole of opposite spin. Such excitations are called Stoner excitations. The existence
of the gap implies that Stoner excitations at q = 0 require an energy . Zero frequency
Stoner excitations require q to be at least equal to the difference between the Fermi momenta
of electrons with spin up and spin down. Repeating the arguments of Sec. 6.1, we can find
an expression for the support of Im + in the, for a ferromagnet, somewhat unrealistic case
of a parabolic band. For positive one finds that, at zero temperature, Im + is nonzero
for
q2 q2
vF q + + < < vF q + + . (8.48)
~ 2m~ ~ 2m~
This support is shown in Fig. 8.5.
Substitution into Eq. (8.30) then yields for q 0
 g 
B
+ (q 0, ) = . (8.49)
2 U
8.6. THE TRANSITION TO A MAGNETIC INSULATOR 177
lYlYl lYlYl lYlYl lYlYl lYlYl lYlYl lYlYl lYlYl lYlYl lYlYl lYlYl lYlYl lYlYl lYlYl lYlYl lYlYl lYlYl lYlYl lYlYl lll
mkjYmlkjYmlYmYkjmYlkjmYl mYkjmYlkjmYl mYkjmYlkjmYl mYkjmYlkjmYl mYkjmYlkjmYl mYkjmYlkjmYl mYkjmYlkjmYl mYkjmYlkjmYl mYkjmYlkjmYl mYkjmYlkjmYl mYkjmYlkjmYl mYkjmYlkjmYl mYkjmYlkjmYl mYkjmYlkjmYl mYkjmYlkjmYl mYkjmYlkjmYl mYkjmYlkjmYl mYkjmYlkjmYl mkjmlkjml
jmlkYlkjmYkj jmYlklkjmYkj jmYlklkjmYkj jmYlklkjmYkj jmYlklkjmYkj jmYlklkjmYkj jmYlklkjmYkj jmYlklkjmYkj jmYlklkjmYkj jmYlklkjmYkj jmYlklkjmYkj jmYlklkjmYkj jmYlklkjmYkj jmYlklkjmYkj jmYlklkjmYkj jmYlklkjmYkj jmYlklkjmYkj jmYlklkjmYkj jmYlklkjmYkj jmlklkjmkj
llYmkjmkjl llYmkjmkjl llYmkjmkjl llYmkjmkjl llYmkjmkjl llYmkjmkjl llYmkjmkjl llYmkjmkjl llYmkjmkjl llYmkjmkjl llYmkjmkjl llYmkjmkjl llYmkjmkjl llYmkjmkjl llYmkjmkjl llYmkjmkjl llYmkjmkjl llYmkjmkjl llYmkjmkjl llmkjmkjl
/ F 4 mkjYmlkjYmlYmYkjmYlkjmYl mYkjmYlkjmYl mYkjmYlkjmYl mYkjmYlkjmYl mYkjmYlkjmYl mYkjmYlkjmYl mYkjmYlkjmYl mYkjmYlkjmYl mYkjmYlkjmYl mYkjmYlkjmYl mYkjmYlkjmYl mYkjmYlkjmYl mYkjmYlkjmYl mYkjmYlkjmYl mYkjmYlkjmYl mYkjmYlkjmYl mYkjmYlkjmYl mYkjmYlkjmYl mkjmlkjml
lkjmYlkjlYmkj lkjmYlkjlYmkj lkjmYlkjlYmkj lkjmYlkjlYmkj lkjmYlkjlYmkj lkjmYlkjlYmkj lkjmYlkjlYmkj lkjmYlkjlYmkj lkjmYlkjlYmkj lkjmYlkjlYmkj lkjmYlkjlYmkj lkjmYlkjlYmkj lkjmYlkjlYmkj lkjmYlkjlYmkj lkjmYlkjlYmkj lkjmYlkjlYmkj lkjmYlkjlYmkj lkjmYlkjlYmkj lkjmYlkjlYmkj lkjmlkjlmkj
mlkjYmkjmlYmlYkjmkjmYl mlYkjmkjmYl mlYkjmkjmYl mlYkjmkjmYl mlYkjmkjmYl mlYkjmkjmYl mlYkjmkjmYl mlYkjmkjmYl mlYkjmkjmYl Stoner m m m m m m m m m m
kjmlkjYmYlkj kjmYlkjmYlkj kjmYlkjmYlkj kjmYlkjmYlkj kjmYlkjmYlkj kjmYlkjmYlkj kjmYlkjmYlkj kjmYlkjmYlkj excitations kjmYlkjmYlkj kjmYlkjmYlkj lYkjmkjmYlkjmYlkjmYlkj lYkjmkjmYlkjmYlkjmYlkj lYkjmkjmYlkjmYlkjmYlkj lYkjmkjmYlkjmYlkjmYlkj lYkjmkjmYlkjmYlkjmYlkj lYkjmkjmYlkjmYlkjmYlkj lYkjmkjmYlkjmYlkjmYlkj lYkjmkjmYlkjmYlkjmYlkj lYkjmkjmYlkjmYlkjmYlkj lkjmkjmlkjmlkjmlkj
llYmkjmlkjYllYmkjmlYkj llYmkjmlYkj llYmkjmlYkj llYmkjmlYkj llYmkjmlYkj llYmkjmlYkj llYmkjmlYkj llYmkjmlYkj llYmkjmlYkj llYmkjmlYkj llYmkjmlYkj llYmkjmlYkj llYmkjmlYkj llYmkjmlYkj llYmkjmlYkj llYmkjmlYkj llYmkjmlYkj llYmkjmlYkj llmkjmlkj
mkjmlkjYmlYmkjmYlkjmYl mkjmYlkjmYl mkjmYlkjmYl mkjmYlkjmYl mkjmYlkjmYl mkjmYlkjmYl mkjmYlkjmYl mkjmYlkjmYl mkjmYlkjmYl mkjmYlkjmYl mkjmYlkjmYl mkjmYlkjmYl mkjmYlkjmYl mkjmYlkjmYl mkjmYlkjmYl mkjmYlkjmYl mkjmYlkjmYl mkjmYlkjmYl mkjmlkjml
2
kjmYljmYlkkj kjmYljmYlkkj kjmYljmYlkkj kjmYljmYlkkj kjmYljmYlkkj kjmYljmYlkkj kjmYljmYlkkj kjmYljmYlkkj kjmYljmYlkkj kjmYljmYlkkj kjmYljmYlkkj spin kj kj kj waves kj kj kj kj kj kj
lmlkjYmlkjYlmYlkjmYlkj lmYlkjmYlkj lmYlkjmYlkj lmYlkjmYlkj lmYlkjmYlkj lmYlkjmYlkj lmYlkjmYlkj lmYlkjmYlkj lmYlkjmYlkj lmYlkjmYlkj mYljmYlklkjmYlkjmYlkj mYljmYlklkjmYlkjmYlkj mYljmYlklkjmYlkjmYlkj mYljmYlklkjmYlkjmYlkj mYljmYlklkjmYlkjmYlkj mYljmYlklkjmYlkjmYlkj mYljmYlklkjmYlkjmYlkj mYljmYlklkjmYlkjmYlkj mljmlklkjmlkjmlkj
mkjYmkjYmYmYkjmYkjmYmYkjmYkjmYmYkjmYkjmYmYkjmYkjmYmYkjmYkjmYmYkjmYkjmYmYkjmYkjmYmYkjmYkjmYmYkjmYkjmYmYkjmYkjmYmYkjmYkjmYmYkjmYkjmYmYkjmYkjmYmYkjmYkjmYmYkjmYkjmYmYkjmYkjmYmYkjmYkjmYmYkjmYkjmYmkjmkjm
0
0 1 2 3 4
q/k F

Figure 8.5: Support of the imaginary part of + for a ferromagnet and spin wave dispersion.

The pole of as 0 corresponds to a uniform rotation of the magnetization. This

collective excitation has zero frequency because no energy is required for such a uniform ro-
tation. At finite wavevector q, the transverse spin susceptibility + has a pole at frequencies
= Dq 2 , where D is a proportionality constant. Such an excitation is known as a spin
wave. Spin waves are virtually undamped for small wavenumbers. At larger wavenumbers,
the spin wave branch enters into the region of the (q, ) plane where + is complex. Then,
spin waves can decay by excitation of Stoner excitations, a pair of a particle and a hole of
opposite spin, see Fig. 8.5.
The fact that spin waves exist at low frequencies if q 0 is guaranteed by the rotation
symmetry of the ferromagnetic state. A uniform rotation of all spins does not cost energy,
and, hence, a long-wavelength spin wave has energy proportional to q 2 . Quite generally,
low-lying collective modes whose existence is guaranteed by a continuous symmetry of the
problem are referred to as Goldstone modes.

8.6 The transition to a magnetic insulator

In Sec. 8.2 we considered the interaction term in the Hubbard model as a perturbation.
Without interactions, the electrons are delocalized. In our treatment, the interaction does
not change that; the main effect of interactions is to change the spin susceptibility, and,
eventually, beyond the Stoner instability, to create a gap between the self-consistent
energy levels of electrons with spin up and spin down.
178 CHAPTER 8. MAGNETISM

This perturbative picture should remain correct as long as the interaction energy is small
compared to the band width. In this section, the opposite case is investigated, when the
band width is small compared to the interaction strength U . This is called the atomic
limit of the Hubbard model.
In the atomic limit, the coordinate representation is preferred over the momentum rep-
resentation. Fourier transforming the non-interacting part of the Hamiltonian, Eq. (8.23)
above, we find
0 = t(ri rj )i

X
H j , (8.50)
ij,

where i is the annihilation operator for an electron with spin in Wannier wavefunction
i (r). The interaction Hamiltonian is

1 = U
X
H ni ni , (8.51)
i

which is the first term on the r.h.s. of Eq. (8.28) above. The indices 0 and 1 refer to
the non-interacting and interacting parts of the Hamiltonian, respectively. As we discussed
above, in the atomic limit the Hamiltonian H 1 is, in fact, dominant.
The problem of studying the full Hamiltonian H 0 + H 1 has proven extremely difficult.
Whereas an exact solution has been obtained in one dimension, in two and three dimensions
no more than a few limiting are fully understood. In order to make progress, we now make
use of the fact that the interaction U is much larger than the band width. Hereto, we write
0 = H
H 00 + H
01 , (8.52)

where
00 = t(0)i
01 = t(ri rj )i

X X
H i , H j , (8.53)
i, i6=j,

and consider H 01 as a perturbation. (Note that H 00 cannot be considered a perturbation,

since t(0) determines the position of the band, not the band width.)
Below, we are interested in the retarded single-particle Green function,

Gij, (t) = i(t)h[i (t), j

(0)]+ i. (8.54)

Recall that the poles of the single-particle Green function contain information about the
particle-like excitations of the system. Since the unperturbed Hamiltonian H 00 + H
1 is
not quadratic in the fermion creation/annihilation operators, the diagrammatic perturbation
theory we used in the previous chapters cannot be employed to calculate Gij, . Instead, we
8.6. THE TRANSITION TO A MAGNETIC INSULATOR 179

have to calculate Green functions using the equation of motion approach. An illustration of
the equation of motion approach for the calculation of phonon Green functions was given in
Sec. 2.6 above.
Let us first calculate the single-particle Green function in the absence of the perturbation

H01 . In that case, one has

t Gij, (t) = i(t)h[i (t), j

(0)]+ i i(t)h[t i (t), j

(0)]+ i. (8.55)

Using the equation of motion for the annihilation operator i (t),

t i (t) = it(0)i (t) iU ni, i (t), (8.56)

we find

(it + t(0)) Gij, (t) = (t)ij (i)U (t)h[ni, (t)i (t), j

(0)]+ i. (8.57)

Repeating the same procedure for the Green function that appeared at the r.h.s. of Eq.
(8.57), we find
h i
t i(t)h[ni, (t)i (t), j

(0)]+ i =
= i(t)h[ni, (t)i (t), j

(0)]+ i t(0)(t)h[ni, (t)i (t), j

(0)]+ i
U (t)h[ni, (t)2 i (t), j

(0)]+ i
h i

= i(t)ij hni, i i(t(0) + U ) i(t)h[ni, (t)i (t), j (0)]+ i .

After Fourier transforming and substitution into Eq. (8.57), we find

(1 hni, i)ij hni, iij

Gij, () = + . (8.58)
t(0) + i t(0) U + i

In a paramagnetic state, the expectation value hni, i = n/2, where n is the electron number
density.
The result (8.58) has a very simple interpretation: If an electron of spin is present at
a site, an added electron with spin has energy t(0) + U , so that its retarded Green function
is Gii, () = 1/( t(0) U + i). If no electron of spin is present, the Green function is
Gii, () = 1/( t(0) + i). The probability that an electron of spin is present is hni, i,
hence Eq. (8.58).
Once the hopping term H 01 is included, the equation of motion for the Green function can
no longer be truncated. However, following Hubbard, one can make the following guess
180 CHAPTER 8. MAGNETISM

for the result. It should be pointed out, however, that this guess has no formal justification;
nevertheless, as youll see below, it conveys a compelling picture of how the system behaves.
The single-electron Green function (8.58) can be interpreted as the inverse of i , where
i is a measure of the energy of an electron at site i, including the interaction modifications.
Without interactions, the Green function of the electron would be Gk, = 1/( k + i),
where X
k = t(0) + eik(ri rj ) t(ri rj ). (8.59)
i6=j

Note that we switched to the momentum representation. With interactions, we replace

t(0) by (Gii, ())1 . Hence, we find
1
GH
k = . (8.60)
(Gii ())1 (k t(0)) + i
Let us now consider the positions of the poles of the Green function GH
k . They are at
the solutions of the equation
( t(0))( t(0) U ) = (k t(0))( t(0) U (1 n/2)). (8.61)
You verify that for n 0 or U 0 one recovers the Bloch band = k . You also verify
that for the limit of very small bandwidth max |k t(0)|  U the solutions approach the
on-site energies t(0) and t(0) + U of the Hubbard model in the atomic limit. The effect of
the inter-site hopping is to change the N -fold degenerate solutions = t(0) and = t(0) + U
into narrow bands around t(0) and t(0) + U . For max |k t(0)|  U these bands are given
by
= t(0)(n/2) + k (1 n/2), = t(0)(1 n/2) + U + k (n/2). (8.62)
A schematic picture of these bands is shown in Fig. 8.6.
In general, the two bands may overlap. An exception is the case that n is close to 1.
The case n = 1 is known as half filling, since it amounts to half of the maximum possible
number of electrons in the system. At n = 1, the two bands are always separated, although
one may question whether this result is realistic for small U . Since the lowest band is fully
filled at n = 1, there is an excitation gap, and the system becomes an insulator. In this case,
the excitation gap is referred to as Coulomb gap and the insulating state is known as a
Mott insulator.

8.7 Heisenberg model

The Heisenberg model gives a description for an insulating magnet. The magnetization
arises from the alignment of local magnetic moments, not from the shift of Fermi levels for
8.7. HEISENBERG MODEL 181

t 0+U

k
t0

Figure 8.6: Schematic picture of the two energy bands for the Hubbard model for the case that
the interaction U is much larger than the bandwidth.

majority and minority electrons.6

The Heisenberg model is described by the Hamiltonian
X
HHeis = Jsi sj
ij nn
X
= J(sxi sxj + syi syj + szi szj ) (8.63)
ij nn
X 1 1 +

+ z z
= J s s + si sj + s i sj . (8.64)
ij nn
2 i j 2

In writing (8.64), we consider a lattice whose sites are labeled with the index i, the summation
is over nearest neighbor pairs i and j on this lattice only, and si denotes the spin 1/2
operator at site i. In principle, one can include next-nearest neighbor interactions, etc., in
the Hamiltonian (8.64). The spin operators commute for different sites, and obey the usual
spin commutation relations on the same sites. Thus, in terms of the spin raising and lowering
operators,
[s z + z +
i , sj ] = ij si , [si , sj ] = ij si [s+ z
i , sj ] = 2ij si (8.65)
[sxi , syj ] = iij szi , and cyclic. (8.66)
We will take the quantum Heisenberg model as a starting point here; the Heisenberg
model with antiferromagnetic coupling (J > 0) can be derived from the Hubbard model at
half filling, with nearest-neighbor hopping and large positive U , see exercise 8.7.
6
Parts of this section are based on lecture notes by W. van Saarloos, Leiden, 1996.
182 CHAPTER 8. MAGNETISM

In the literature, sometimes an anisotropic version of the Heisenberg model (8.64) is

studied,
X
anis
J (sxi sxj + syi syj ) + Jz szi szj

HHeis =
ij nn
X  J 
= (s+
i sj + s +
i sj ) + Jz szi szj . (8.67)
ij nn
2

Clearly, in this case the model is not invariant under arbitrary rotations of the spins, but is
invariant for rotations of the spins about the z-axis.7
Despite its simplicity, the Heisenberg model is difficult to study analytically. One reason
that is of interest to us, is that Wicks theorem does not hold for the Heisenberg model.
Hence, we cannot rely on the diagrammatic perturbation theory.
An important question of interest for the Heisenberg model is when and how a mag-
netic phase is formed. This question is in the realm of the theory of phase transitions and
critical phenomena, and we will not discuss it here. Instead, we will assume that a low-
temperature magnetic phase exists, and aim at a description of the possible excitations of
the system in the low-temperature magnetic phase and the high-temperature paramagnetic
phase. We will carry out this program in detail for the ferromagnetic Heisenberg model, for
which the low-temperature magnetic phase is the fully polarized phase. For the antiferro-
magnetic Heisenberg model, this question is much more difficult, as the precise nature of the
antiferromagnetic phase is not known.
Let us now turn to the isotropic ferromagnetic (J < 0) Heisenberg model (8.64) in absence
of a field to introduce the concept of spin-waves in more detail. The case of the ferromagnetic
Heisenberg model is particularly simple but somewhat a-typical in that the order
parameter associated with the broken symmetry, the total magnetization, is also a conserved
quantity. (This is sometimes referred to as an exact spontaneously broken symmetry.)
Since the total spin S commutes with HHeis , we can diagonalize the Hamiltonian in each
subspace of eigenvalues of the operator S z . Let |+i denote the state with all the spins
pointing up, so that S z |+i = (N/2)|+i, where N is the total number of spins. The state |+i
is the ground state of HHeis if J < 0, although it is not the only possible ground state. In
7 anis
In passing, we note that for J = 0, HHeis reduces to the celebrated Ising model, which plays a central
role in the theory of critical phenomena. Unlike the Heisenberg model, the Ising model does not have a
continuous symmetry, and, as a result, the Ising model has no Goldstone modes. On a bipartite lattice, the
antiferromagnetic Ising model at zero magnetic field is equivalent to the ferromagnetic Ising model. This
result does not extend to the Heisenberg model; the ferromagnetic and antiferromagnetic Heisenberg models
are fundamentally different.
8.7. HEISENBERG MODEL 183

fact, all states obtained from |+i by multiple operation of the total spin lowering operator,
X
S = s
i (8.68)
i

are ground states of the Heisenberg Hamiltonian as well. For a hypercubic lattice in d
dimensions, the ground state energy is E0 = dN J/4.
Since the ground state of the ferromagnetic Heisenberg model is so simple, some low-lying
excitations, the spin waves can be determined exactly. A spin-wave state with wavevector
q is created by the operator Sq operating on the fully polarized state |+i, with
X
Sq = N 1/2 eiqri s
i . (8.69)
i

You verify that a state with a single spin wave is an exact eigenstate of the Heisenberg
Hamiltonian HHeis , at energy E0 + ~q , where E0 is the ground state energy and

|J| X
~q = (1 eiq ), (8.70)
2

where the vector is summed over all nearest-neighbor directions in the lattice. For small
q, the energy q increases proportional to q 2 . Note that, for q 0, the state Sq |+i = |qi
approaches the uniformly rotated state S |+i continuously, and therefore we expect the
excitation energy ~q of this mode to vanish continuously as q 0. This again illustrates
that the appearance of such low lying modes or Goldstone modes for q small is a general
phenomenon when a continuous symmetry is broken.
The spin-wave state |qi is a collective mode; it describes a delocalized state with many
spin excitations involved. For small q, it describes a slow rotation of the spins about the z-
axis. To make this more explicitly, one can calculate the expectation value of the transverse
spin correlation in the state |qi,

1
hq|sxi sxj + syi syj |qi = cos[q (i j)]. (8.71)
N
Spin waves do interact: the two-spin wave state Sq1 Sq2 |+i is not an exact eigenstate of
the Hamiltonian (8.64). On the other hand, spin-wave interactions are small if the number of
excited spin wave modes is small. Then, the spin waves can be considered independent boson
modes, their distribution being given by the Bose-Einstein distribution function. The picture
of independent spin wave modes leads to the Bloch T 3/2 law for the temperature dependence
184 CHAPTER 8. MAGNETISM

u e u e u
u: A
e u e u e
e: B
u e u e u

Figure 8.7: An example of a bipartite lattice. Every neighbor of the A sublattice is a site on the
B sublattice, and vice versa.

of the low-temperature magnetization in three dimensions, hSz (T )i hSz (T = 0)i T 3/2 .8

For one and two dimensions, the density of states of low-energy spin waves is higher than in
three dimensions; there is no regime of a small number of excited spin waves, leading to a
breakdown of the ferromagnetic state in one and two dimensions.
Within the Green function approach, these conclusions can be arrived at if we look at
the retarded transverse spin-spin correlation function,
R
Rij (t) = i(t)h[s +
i (t), sj (0)] i, (8.72)

which is the Heisenberg-model equivalent of the transverse spin susceptibility + we stud-

ied previously. (Note that the correlation function contains a commutator of spin operators,
R
since spin operators satisfy commutation relations.) The function Rij (t) describes the re-
sponse of the spin polarization to a magnetic field, where both the response and the field are
perpendicular to the z axis, which is presumed to be the direction of spontaneous magnetiza-
tion. The poles of the correlation function corresponds to the collective modes. Further, the
spin-spin correlation function is related to the average magnetization through the relation
R
lim Rij (t) = 2ihsz iij . (8.73)
t0

In order to calculate the transverse spin-spin correlation function, we consider its equation
of motion
R X
Rij (t) = 2i(t)hszi i (t) hJli [s z z +
l (t)si (t) sl (t)si (t), sj (0)] i, (8.74)
t l

where we defined Jli = J if l and i are nearest neighbor sites and Jli = 0 otherwise. In order
to close this equation of motion, we replace the operator szi (t) by the average hsz i hszi i in
8
See, e.g., Solid State Physics, by N. W. Ashcroft and N. D. Mermin, Saunders (1976), and the discussion
below.
8.7. HEISENBERG MODEL 185

the ferromagnetic state,9

R X
Rij (t) = 2i(t)hszi i ihsz i R
Jli (Rlj R
(t) Rij (t)). (8.75)
t l

The justification for this approximation is that, in the ferromagnetic state, fluctuations of
szi around its average are small. Moreover, since Eq. (8.75) contains a sum over nearest
neighbors, only the sum of szi over nearest neighbors is important. In high dimensions, the
summation over nearest neighbors leads to an even further reduction of fluctuations, and one
expects the approximation szi hsz i to continue to hold outside the almost fully polarized
ferromagnetic state. Solving Eq. (8.75) by means of a Fourier transform, we find
2hsz i
Rk () = , (8.76)
+ k + i
where X
k = hsz iJ (1 eik ), (8.77)

the summation being over all nearest-neighbor vectors . Note that this solution satisfies
the relation (8.73).
For zero temperature, hsz i = s, and we recover the spin-wave dispersion relation (8.70).
For finite temperatures, the spin-wave frequencies depend on the average spin polarization
hsz i, which, in turn, is a function of the temperature T . In order to determine hsz i, we use
the greater correlation function,
>
Rij (t) = ihs +
i (t)sj (0)i, (8.78)
for which one has
Rii> (0) = is(s + 1) + ihsz (sz + 1)i. (8.79)
In keeping with the approximations we made previously, we neglect fluctuations of sz and
set
Rii> (0) is(s + 1) + ihsz i(hsz i + 1). (8.80)
>
We calculate Rij (t) from the retarded correlation function (8.76) using the general relation
(2.49),
Im RkR ()
Rk> () = 2i
1 e~/T
hsz i( + k )
= 4i . (8.81)
1 e~T
9
This approximation is the equivalent of the Weiss molecular field approximation, see, e.g, Statistical and
Thermal Physics, F. Reif, Mc Graw Hill, 1965.
186 CHAPTER 8. MAGNETISM

Performing the inverse Fourier transform, we then find

2i X 1
Rii> (0) = /T
, (8.82)
N k e k 1
~

from which we derive the self-consistency equation

2 X 1
s(s + 1) hsz i(hsz i + 1) = hsz i /T
. (8.83)
N k e ~ k 1

In the limit T 0 the r.h.s. vanishes in three dimensions and above, recovering the solution
hsz i = s at T = 0. For one and two spatial dimensions, the r.h.s. does not vanish as T 0,
implying the breakdown of the ferromagnetic state by spin wave generation, as we discussed
previously. For low temperatures and three dimensions, one find hsz (T = 0)i hsz (T )i
T 3/2 , which is the Bloch T 3/2 law mentioned previously.
The self-consistent relation could be used to find the Curie temperature, the highest
temperature at which the spontaneous magnetization exists. However, close to the Curie
temperature, spin fluctuations are abundant, and the approximation szi hsz i we made in
the derivation of Eq. (8.83) is not justified in systems with low dimensionality.
We now turn to the antiferromagnetic Heisenberg model J > 0. Classically, an antifer-
romagnetic state is possible only if the lattice is bipartite: it can be separated in sublattices
A and B such that all nearest neighbors of a site in lattice A belong to lattice B and vice
versa, see Fig. 8.7. If the lattice is not bipartite, the antiferromagnet is frustrated, and a
completely different physical picture arises, which we do not discuss here. The classical an-
tiferromagnetic ground state is a state in which spins on one sublattice point up, whereas
spins on the other sublattice point down. This state is known as the Neel state. Intu-
itively, we expect that the Neel state is the ground state of the quantum antiferromagnetic
Heisenberg model as well. However, the Neel state is not an exact eigenstate of H Heis !10
Nevertheless the true ground state has strong antiferromagnetic correlations reminiscent of
this state.
Even though the classical Neel state is not the exact ground state, it does suggest a defi-
nition of the antiferromagnetic order parameter, which takes the role of the magnetization
for the ferromagnetic Heisenberg model. Let us consider a hypercubic lattice with lattice
spacing a in d dimensions and let a reference spin on sublattice A point up; then the so
called staggered magnetization is defined as
 z
AF hsi i if ri in sublattice A,
Mi = . (8.84)
hszi i if ri in sublattice B
10 anis
The classical Neel state is the ground state of HHeis with Jz > 0 and J = 0, but not of the model with
J 6= 0.
8.7. HEISENBERG MODEL 187

In the classical Neel state, this order parameter is 1/2 for all sites.
This choice of the order parameter may be justified for a Heisenberg model with spin
s  1/2. The larger the spin s is, the more classical the behavior of the model becomes, as
the commutation relations (and hence fluctuations) become less important. In the classical
limit, we can represent the spins by vectors on a sphere of radius s. In this s classical
limit, the ground state of the antiferromagnetic Heisenberg model consists of vectors pointing
in opposite directions on the two sublattices. In this limit, (8.84) is a good order parameter,
whose value approaches s for s large.
In recent years, there has been a lot of research on the properties of quantum anti-
ferromagnets, in particular in d = 2. Research on two-dimensional antiferromagnets was
motivated to a large extent by the discovery of high temperature superconductors, which
are obtained by doping compounds which are layered insulating quantum antiferromagnets.
The consensus, based on many large scale numerical simulations, is that the ground state of
the two dimensional Heisenberg model with spin 1/2 does have antiferromagnetic order, but
that the order parameter (8.84) is reduced from the classical Neel value 1/2 to about 0.34.
In the ferromagnetic case, the order parameter was a conserved quantity and as a result
a single spin wave state |qi was an exact eigenstate of the Hamiltonian. In the present case,
the staggered order parameter (8.84) is not a conserved quantity, and the spin wave analysis
can not be performed exactly (after all, the ground state is not known exactly either!), so
we have to resort to an approximate treatment. In order to understand the nature of the
approximation better, it will turn out to be useful to look at the Heisenberg model for the
case of large spin s. Since we expect the ground state to get closer to the classical Neel state
for large s, increasing s gives us a way to get a controlled approximation.
As in the case of a ferromagnet, to analyze spin waves we need to consider states where
one spin is flipped. Since the ground state is not explicitly known, it is more convenient to
use (in the Heisenberg picture) the equation of motion for the spin flip operator. Using the
commutation relations (8.65) (which are also valid for arbitrary spin s), we get

s 1 J X z
i
= [s si si+ s z


i , H Heis ] = i+ si . (8.85)
t i~ i~

As before, the sum over is a sum over the nearest neighbor vectors on the quadratic (d = 2)
or cubic (d = 3) lattice.
For large s, spin waves consist of minor distortions of the Neel state, with the z-component
of the spin slightly reduced from its maximal value +s on the A sublattice, and slightly
increased from its minimal value s on the B sublattice. To a good approximation, we can
then write in (8.85) szi |ground statei s|ground statei depending on whether ri is on the
188 CHAPTER 8. MAGNETISM

A or B sublattice. Using the notation

|iiA = s
i |ground statei for ri in sublattice A, (8.86)
|iiB = s+
i |ground statei for ri in sublattice B, (8.87)

we then get from (8.85)

!
J X
|iiA = 2ds|iiA s |i + iB ,
t i~
!
J X
|iiB = 2ds|iiB + s |i + iA . (8.88)
t i~

Since the two sublattices are distinct, we now have two coupled equations, instead of one in
the case of a ferromagnet. The equations for the spin waves can be solved by introducing
Fourier transforms,
1/2 1/2
X X
|qiA = NA eiqi |iiA , |qiB = NB eiqi |iiB . (8.89)
iA iB

Writing the temporal evolution of these modes as eiq t , we then get from (8.88) the disper-
sion relation  
 = 0, q = 1
 ~q + 2dsJ 2dsJq X
eiq .


 2dsJq ~q 2dsJ (8.90)
 2d
Solution of Eq. (8.90) yields q
~q = 2dsJ 1 q2 . (8.91)
By expanding q for small q, we find that long wavelength antiferromagnetic spin waves have
a linear dispersion
q 2sJq d, q  1. (8.92)
This linear dispersion is reminiscent of that of acoustic phonons in a crystal. The difference
between the quadratic dispersion of ferromagnetic spin waves and the linear dispersion of
the antiferromagnetic spin waves the Goldstone modes of the antiferromagnet can be
traced to the fact that in the first case, the order parameter itself is conserved. This is rather
exceptional, and indeed Goldstone modes usually have a linear dispersion.
Our earlier discussion implies that (8.92) is a good approximation for large s. Although
the estimate of the spin wave velocity c = sJ 2d/~ may not be that accurate for spin 1/2,
the linearity of the antiferromagnetic spin wave dispersion is a robust feature that is not
affected by the approximations made.
8.8. EXERCISES 189

In closing this section, we note that the present analysis may be extended to a systematic
1/s expansion for quantities like the spin wave velocity and the order parameter. Quantum
mechanical fluctuations are more and more suppressed as s increases. Intuitively, this be-
comes clear from the fact that for s = 1/2, the operators s+
i or si lead to a local reversion of
the spin, and hence tends to locally make the order parameter of the wrong sign. For large
s, these operators just reduce the z-component of the spin slightly they do not lead to a
complete spin reversal.

8.8 Exercises
Exercise 8.1: Spin susceptibility

The Thomas-Fermi approximation (8.22) for the spin susceptibility in an interacting electron
gas is not correct. It neglects the structure of the interaction on the scale of the Fermi wave-
length. A better approximation is made when the full Thomas-Fermi screened interaction is
used,
e2
VqRPA (in ) = if q, n 0,
0 (q 2 + ks2 )
see Eq. (7.45). Hence, the momentum-dependence of the screened interaction is taken into
account, but retardation effects are neglected.

(a) Argue that, in the limit q 0, the renormalized vertex (k, im ; k + q, im + in ) of

Eq. (8.15) does not depend on the direction of the wavevector k or on the energy m .

(b) Since (k, im ; k + q, im + in ) of Eq. (8.15) does not depend on the direction of
eff
the wavevector k, the effective interaction Vkvk 0 in the vertex equation (8.15) may be

integrated over the directions of k0 . Show that the integrated interaction is

eff e2 ks2 + (k + k 0 )2
Vk,k 0 = ln . (8.93)
40 kk 0 ks2 + (k k 0 )2

(c) Argue that the renormalized vertex is determined by the effective interaction at the
Fermi level. From here, derive the result
B g
RPA
+ = . (8.94)
2 (2/3 2 )2/3 r 2
s ln(1 + (3 /2)
2/3 /r )
s

(d) Can you explain this result qualitatively in the limits rs  1 and rs  1?
190 CHAPTER 8. MAGNETISM

Exercise 8.2: Specific heat

Verify Eq. (8.35) for the low-temperature specific heat of the Hubbard model close to the
Stoner instability. For your verification is is sufficient if you find the order of magnitude
of the proportionality constant in Eq. (8.35). You do not need to find precise numerical
coefficients.

Exercise 8.3: Ferromagnet with parabolic dispersion relation

In this exercise we consider the Hubbard model, with a parabolic dispersion of the band
energies k . This model is somewhat unrealistic, but it allows us to calculate many quantities
in closed form.

(a) Using the Hartree-Fock energy levels, show that the spin polarization
n
n
= (8.95)
n
+ n

obeys the equation

 n 1/3 2m
U = (1 + )2/3 (1 )2/3 . (8.96)
9 4 ~2
Here n = n
+ n
is the total electron density. Find a relation between and the energy
gap .

(b) Analyze Eq. (8.95) and find for what values of the Hubbard interaction strength U a
spontaneous spin polarization is formed.

(c) Show that, if the Hubbard interaction strength U is further increased, a full polarization
is achieved, = 1. This situation is known as a half metal.

(d) Calculate the transverse spin susceptibility + for low frequencies and long wave-
lengths. What is the spin-wave dispersion that you find from your answer?
8.8. EXERCISES 191

P= + + + ...

Figure 8.8: Singular contribution to the electron self energy at the Stoner instability.

Exercise 8.4: Self-energy correction

Close to the Stoner instability point there are important corrections to the self energy that
are not included in the Hartree-Fock approximation. One of the leading corrections is shown
in Fig. 8.8.

(a) Calculate the contribution to the self energy shown in Fig. 8.8 for the Hubbard model.

(b) Discuss how the self-energy correction you calculated under (a) affects the electron
effective mass m and the density of states at the Fermi level.

Exercise 8.5: Coulomb gap

In this exercise, we consider the ansatz (8.58) for the single particle Green function of the
Hubbard model. The Green function not only contains information about the possible single-
particle excitations through its poles, but also through the spectral weight of each pole.

(a) Find the spectral weights for the case of large U , U  max |k t(0)|, for which you
can use Eqs. (8.62) for the poles of the Green function.

(b) Do the same for the case of small U , U  max |k t(0)|. Formulate your answer for
the case n  1 and n = 1 separately.

Exercise 8.6: Bloch T 3/2 law

Derive the Bloch T 3/2 law for the magnetization of a three-dimensional ferromagnet from
Eq. (8.83).
192 CHAPTER 8. MAGNETISM

Exercise 8.7: Antiferromagnetic Heisenberg model from large-U Hubbard model at half filling

In this exercise we consider the Hubbard model with nearest neighbor hopping,
X  
i j + j

X
HHub = t i + U ni ni . (8.97)
ij nn, i

(a) When U = 0, the Hubbard model describes non-interacting electrons, and we can
explicitly determine the single electron energies k . Show that these are given by Eq.
(8.37) above.
Because of the Pauli principle, there can not be more than two electrons per site. Let us
from now on consider half filling, i.e. the case where there is one electron per site. Moreover,
we consider the limit of large positive U (U  t). Then it is energetically very unfavorable
to have two electrons on the same site in other words the low energy sector of the Hilbert
space is the one where all sites are occupied by one electron. Our strategy now will be to
project onto this part of Hilbert space of singly occupied sites.
What will the interactions look like? Well, the operator ni ni only counts whether a site
is doubly occupied, therefore it does not depend on the absolute spin direction. Therefore,
the effective spin interaction will only depend on the relative spin orientation, in other words,
it will be of the Heisenberg type si sj .
Furthermore the interaction will be antiferromagnetic. Physically, this can be seen as
follows. Virtual excitations to states in which one site is doubly occupied involve the hop of
an electron to a neighboring site and back (see the figure); in the intermediate state, there
is an extra energy U associated with the double occupancy.

s6 s 6
s s 6
s s 6
s 6
s
? ? ?

(a) (b) (c)

initial configuration virtual state final state hopping blocked
hop t hop t

(b) Argue that, with these considerations, the large-U Hubbard model at half filling reduces
to an antiferromagnetic Heisenberg model with interaction J = 2t2 /U .
The range over which the antiferromagnetic phase exists as a function of the ratio U/t
and doping of the Hubbard model, is still under active investigation.
Chapter 9

Phonons and Electron-phonon

interaction

In the previous chapters we have looked at the dynamics of electrons in a metal. In this
chapter, well consider the lattice ions, their mutual interaction, and the interaction between
lattice ions and electrons. Just like the presence of the electron gas modifies the electron-
electron interaction, and leads to an effective, or screened interaction, the presence of the
ionic lattice and the interaction between electrons and phonons causes a modification of the
electron-electron interaction. This phonon-modified interaction is attractive, and is the root
cause for superconductivity.

9.1 Normal modes

We consider the dynamics of the atoms of a lattice. The atoms interact via a potential
V({xi }) that depends on the location xi of all atoms i = 1, . . . , N . Each atom has an equi-
librium position ri . For small displacements ui = xi ri from equilibrium, the Hamiltonian
for the lattice atoms can be expanded in a series in the displacements ui . The first order term
in the expansion vanishes, since the potential V is a minimum at the equilibrium positions
of the atoms. If we truncate the expansion at the second order, we have
N
X p2i 1 XX
H= + ui uj V; (ri , rj ), (9.1)
i=1
2M 2 i,j

where the numbers V; (ri , rj ) are derivatives of the potential V at the equilibrium positions
ri , rj ,
V; (ri , rj ) = xi xj V({ri }). (9.2)

193
194 CHAPTER 9. PHONONS AND ELECTRON-PHONON INTERACTION

The displacement and momentum operators ui and pj have canonical commutation relations,

[pi , pj ] = [ui , uj ] = 0 (9.3)

[pi , uj ] = i~ ij . (9.4)

For an infinite lattice or for a lattice with periodic boundary conditions, the real numbers
V; (ri , rj )) depend on the difference ri rj only. Inversion symmetry implies V; (ri rj ) =
V; (rj ri ), and, since derivatives can be taken in any order, V; (ri rj ) = V; (ri rj ).
The approximation (9.1) in which the Hamiltonian of the lattice atoms is truncated after
the quadratic term in the displacements is known as the harmonic approximation.
The Hamiltonian (9.1) can be brought to diagonal form by a suitable basis transformation.
We construct this basis transformation in two steps. First, we perform a Fourier transform
with respect to the lattice coordinate ri and define
1 X ikrj
pk = e pj (9.5)
N j
1 X ikrj
uk = e uj . (9.6)
N j

Here and below, the wavevector k is in the first Brillouin zone of the reciprocal lattice. After
Fourier transform, the operators pk and uk still obey canonical commutation relations,

[pk , pk0 ] = [uk , uk0 ] = 0 (9.7)

[pk , uk0 ] = i~ kk0 . (9.8)

The momenta and displacements corresponding to opposite wavevectors k and k are related
by hermitian conjugation, pk = pk and uk = uk . Note that the total number of degrees
of freedom has not changed upon Fourier transform.
In terms of these new variables, the Hamiltonian reads
XX 1 1

H= pk p,k + V,k uk u,k , (9.9)
k ,
2M 2

where
1 X ik(ri rj )
V,k = e V; (ri rj ). (9.10)
N i,j

The 3 3 matrix Vk is symmetric because V; (r) is symmetric and real because V; (r) =
V; (r). Moreover, V,k = V,k . Since the matrix Vk is real and symmetric, it is
9.1. NORMAL MODES 195

diagonalized by three real and orthogonal eigenvectors. We denote the eigenvectors as v k ,
2
= 1, 2, 3, and the corresponding eigenvalues as M k . Then we can introduce normal
modes
X
pk =
vk pk , (9.11)

X
uk =
vk uk , (9.12)

that bring H into diagonal form,

X 1 1

2
H= p p + M k uk uk . (9.13)
k,
2M k k 2

One verifies that the normal mode operators pk and uk still obey canonical commutation
relations.
The Hamiltonian (9.13) represents 3N harmonic oscillator modes. One can write the
normal mode Hamiltonian in terms of phonon creation and annihilation operators,
r r
M k 1
bk = u k + ipk , (9.14)
2~ 2M k ~
r r
M k 1
b
k = uk ipk , (9.15)
2~ 2M k ~
that obey the commutation rules
h i h i
0 0
bk , bk0 = b
k , b k0

= 0,
h 0
i
bk , bk0 = kk0 0 . (9.16)

In terms of the new operators b

k and bk , the Hamiltonian reads
 
X 1
H= ~k bk bk + . (9.17)
2
k

Since the operators b

k and bk satisfy creation and annihilation operators for bosons, the
Hamiltonian (9.17) is interpreted as describing phonons, quantized lattice vibrations with
boson statistics. With or without such interpretation, what is important for us is that the
196 CHAPTER 9. PHONONS AND ELECTRON-PHONON INTERACTION

Hamiltonian (9.17) is diagonal, so that any average of displacements and momenta of lattice
atoms can be calculated straightforwardly once it is expressed in terms of the operators b
k
and bk . For this, it is important to have an expression for the inverse of the normal-mode
transformation we just derived,
r
1 X ikrj ~ 


uj = e vk b + bk , (9.18)
N k 2M k k
r
1 X ikrj M ~k  
pj = e vk ibk ibk . (9.19)
N k 2

Since the phonon Hamiltonian is nothing but the Hamiltonian of 3N harmonic harmonic
oscillators, phonon Green functions can easily be derived from the harmonic oscillator Green
functions, see Sec. 2.6.
For phonons one defines Green functions with respect to the displacement operators ui .
The phonon temperature Green function then reads
Di;j (1 ; 2 ) = hT [ui (1 )uj (2 )]i. (9.20)
For phonons, D is real, since u and H are real. According to Eq. (2.19), this implies
Di;j ( ) = Di;j ( ). (9.21)
Similarly, one defines phonon retarded, advanced, time-ordered, greater, and lesser Green
functions.
Calculation of the temperature Green function then yields, for 0 ~/T ,
X ek ek
  
~ ik(ri rj )
Di;j ( ) = v v e + . (9.22)
k
2M N k e~k /T 1 1 e~k /T

It is easily verified that Di;j (0) = Di;j (~/T ), so that this result can be extended pe-
riodically to the entire imaginary time axis. Fourier transform to the imaginary time
yields
X ~  1
Di;j (in ) = v v eik(ri rj ) 2 2
. (9.23)
k
M N k + n

Similarly, for the retarded and advanced Green functions, one finds
X ~

R
Di;j (t) = (t) v v eik(ri rj ) sin(k t), (9.24)
k
M N k
 
A
X ~
Di;j (t) = (t) v v eik(ri rj ) sin(k t). (9.25)
k
M N k
9.1. NORMAL MODES 197

Performing a Fourier transform to time, we have

X ~
 
1 1

R ik(ri rj )
Di;j () = v v e . (9.26)
k
2M N k k + i + k + i

Finally, the phonon time-ordered Green function is

X ~

v v eik(ri rj ) hnk ieik |t| + hnk + 1ieik |t| (9.27)


Di;j (t) = i ,
k
2M N k

where hnk i = 1/(exp(~k /T ) 1) is the occupation number.

The spectral density can now be found by taking the imaginary part of D R (),
X ~

Ai;j () = 2 v v eik(ri rj ) (( k ) ( + k )) . (9.28)
k
2M N k

As an application of the use of Green functions for phonons, olne may again consider
inelastic neutron scattering. The relevant quantity to calculate in such a scattering experi-
ment is the differential cross section, the amount of scattered neutrons that emerge from the
sample at a given energy ~ and solid angle .
Starting point of the analysis is, as in Sec. 8.4, the expression for the differential scattering
cross section,

d2 X k 0  M 2

|hk0 , f |H 0 |k, ii|2 ( + Ei Ef ),


= (9.29)
dd f
k 2

where k is the wavevector of the incoming neutron, k0 is the wavevector of the outgoing
neutron, M is the reduced mass of the neutron, i and f refer to initial state and final state,
respectively, H 0 is the interaction between the neutron and the target, and the outer brackets
h. . .i indicate a thermal average over the initial states i of the target.
Equation (9.29) can be rewritten as
2
d2 k0

M
Z
2
= |Vq | dteit F (q, t), (9.30)
dd 2k 2

where q = k k0 and the correlation function F (q, t) is defined as

X
eiq(rl rj ) eiquj (t) eiqul (0) .


F (q, t) = (9.31)
j,l
198 CHAPTER 9. PHONONS AND ELECTRON-PHONON INTERACTION

In Eq. (9.31), the summation is over the lattice sites j and l and the operators uj (t), ul (0)
are the corresponding displacements of the lattice atoms in the Heisenberg picture. The
equilibrium positions of the lattice atoms are denoted rj and rl . In Eq. (9.30), Vq is the
Fourier transform of the potential that describes how a neutron scatters off a single atom.
The combined effect of many atoms in a vibrating lattice is described by the correlation
function F (q, t).
For small q, we now write
D E
iq(uj (t)ul (0))+ 12 [quj (t),qul (0)]
X
iq(rl rj )
F (q, t) = e e , (9.32)
j,l

where the expression in the exponent is correct up to corrections of order q 3 . In order to

perform the thermal average, we use the cumulant expansion,1 and obtain
X 1 2 1 2
F (q, t) = eiq(rl rj ) 2 h(quj (t)) i 2 h(qul (0)) i+h(quj (t))(qul (0))i , (9.33)
j,l

again up to corrections of order q 3 in the exponent.2

The second and third terms in the exponent do not depend on j, l, or t; they form an
overall prefactor known as the Debije-Waller factor. Writing the Debije-Waller factor as
e2W , we may calculate W from the time-ordered phonon Green function
iX
W = q q lim Dj;j (t). (9.34)
2 t0

Substituting the exact result (9.27) for the time-ordered Green function, we obtain
XX ~

W = q q v v coth(~k /2T ). (9.35)
k
4M N k

We further expand F in the fourth term of the exponent, and separate F into an elastic
and inelastic contribution,
F (q, t) = F el (q, t) + F inel (q, t) (9.36)
1
According to the cumulant expansion, any average of the form hexp(A)i can be calculated as
heA i = exp hAi + 21 var A + . . . ,

where var A = hA2 ihAi2 is the variance (i.e., the second cumulant) and the dots indicate higher cumulants.
If the probability distribution of A is Gaussian, the first two terms in the cumulant expansion are sufficient.
2
In fact, Eqs. (9.32) and (9.33) are exact because the Hamiltonian is quadratic in the displacements u i ,
i = 1, . . . , N .
9.2. ELECTRON-PHONON INTERACTION 199

with
X
F el (q, t) = e2W eiq(rl rj ) (9.37)
j,l
X X
F inel (q, t) = e2W eiq(rl rj ) q q huj (t)ul (0)i. (9.38)
j,l

Fourier transform of F el gives a term proportional to () in the scattering cross section,

corresponding to elastic scattering from the lattice without the excitation of phonons. Fourier
transform of F inel gives
Z X X
dteit F inel (q, t) = ie2W eiq(rl rj ) >
q q Dj;l (), (9.39)
j,l

>
where Dj;l () is the phonon greater Green function. Using Eqs. (2.49) together with the
>
exact result (9.28) to calculate Dj;l (), we find

~e/2T
Z XX
dteit F inel (q, t) = N e2W q q v v
k
2M sinh(/2T )
(( k ) ( + k )) . (9.40)

9.2 Electron-phonon interaction

For a discussion of the phonon-mediated electron-electron interaction we first need to de-
scribe the interaction between electrons and phonons. We first describe the electron-phonon
interaction in the normal-mode picture, and then switch to the jellium model.
Starting point is the potential the electrons feel from the lattice ions,
Z N
lattice =
X
H dr
n(r) Vion (r xj ), (9.41)
j=1

where the vector xj runs over the positions of all ions in the lattice and n
is the electron

number density operator. Expanding Hlattice around the equilibrium positions xj = rj ,
j = 1, . . . , N , we have
Z N Z N
lattice =
X X
H drn(r) Vion (r rj ) drn(r) uj r Vion (r rj ). (9.42)
j=1 j=1
200 CHAPTER 9. PHONONS AND ELECTRON-PHONON INTERACTION

Here uj = xj rj is the displacement of the jth lattice ion. The first term represents a static
potential and gives rise to the electronic band structure. The second term depends on the
actual ionic positions. It is the second term that leads to the electron-phonon interaction.
Fourier transforming the second term of Eq. (9.42), we find
s !
i XX N~
ep
H = v qVqie k+q,

k, (bq + b
q ), (9.43)
V k q 2M q q

where k is the annihilation operator for an electron with wavevector k and spin , bq the
annihilation operator for a phonon with momentum q and polarization vq . In the derivation
of Eq. (9.43) we used Eq. (9.18) to express the displacement uj in terms of the phonon
creation and annihilation operators b
q and bq .
At low temperatures, the phase space available for umklapp processes, corresponding
to a momentum transfer q outside the first Brillouin zone, is small. Moreover, the weight of
umklapp processes is suppressed because of the q-dependence of the potential Vqie for large
q. Hence, we will restrict the summation over q to the first Brillouin zone.
In an isotropic material, the polarization vector vq is either perpendicular or parallel to q.
The corresponding phonon states are labeled transverse and longitudinal, respectively.
From Eq. (9.43), we conclude that only longitudinal phonons interact with electrons. This
is because the longitudinal phonons carry a charge density, whereas transverse phonons do
not. If we want to simplify Eq. (9.43) to the case that we keep longitudinal phonons only,
we must remind ourselves that in our original formulation of the phonon modes, we have
chosen the same polarization vectors for opposite wavevectors, vq = vq
. Hence, if the
longitudinal polarization vq points along q, vq points opposite to q. We resolve this
problem by defining a function sign (q), which can take the values 1 or 1 and which is
such that sign (q) = sign (q). Keeping longitudinal phonons only, the expression for the
electron-phonon interaction then simplifies to
s !
i XX N~
ep
H = Vqie q sign (q) k+q,

k, (bq + bq ). (9.44)
V k q 2M q

where the creation and annihilation operators b and b refer to longitudinal phonons only.
The calculation of Green functions for a system of electrons and phonons proceeds in a
way that is very similar to what we have seen in the previous chapters. Since our discus-
sion of phonon Green functions there focused on the Green functions for the displacements
rather than the phonon creation and annihilation operators, we rewrite the electron-phonon
9.3. JELLIUM MODEL FOR PHONONS 201

Hamiltonian in the form

1 X X
ep =
H gq k+q, k, uq , (9.45)
V k q

where uq is a longitudinal displacement,

s
q ~
uq = (bq + bq ), (9.46)
q 2M q

and gq describes the strength of the electron-phonon coupling,

gq = iN 1/2 qVqie sign (q). (9.47)

9.3 Jellium model for phonons

In Sec. 9.1 we saw that lattice ions perform vibrations around their equilibrium positions,
the frequency of the vibrations being determined by the second derivative of the interac-
tion between ions. However, this interaction is an effective interaction, mediated by the
electrons. Without electrons, the potential between lattice ions would be the Coulomb po-
tential. As we have seen in the previous section, the long-range Coulomb potential does not
allow for phonon-like low-frequency collective modes. Instead, the collective modes allowed
by the Coulomb interaction are plasma modes, which have a finite frequency even for zero
wavevector.
A description of the interaction between electrons and the lattice and the lattice dynamics
itself is not straightforward. After all, in such a theory the electrons interact with lattice
vibrations, that themselves exist by virtue of the existence of the electrons. A theory of
phonon-mediated electron-electron interactions in which the phonon dispersion is, in turn,
governed by the electron-mediated interaction between the lattice ions is a risky enterprise.
In order to circumvent such a chicken and egg problem, it is preferable to start from first
principles, and use a description in which the only interaction is the Coulomb interaction,
just like we did for the case of the electron gas. As a complete ab-initio treatment is
impossible here for that wed have to find the lattice structure from first principles , we
use a simplified model for the ionic lattice. That simplified model is referred to as jellium
model. In the jellium model, the lattice is replaced by a positively charged ion fluid (or
jellium). In such a model, the only difference between the ions and the electrons is the
different ratios of mass density to charge density: the ion jellium has the same average charge
density as the electron gas, but its mass density is a factor 105 higher. A simplified jellium
202 CHAPTER 9. PHONONS AND ELECTRON-PHONON INTERACTION

description may be used, e.g., if one is interested in long-wavelength phenomena, or if one

only needs an order-of-magnitude estimate of the phonon contribution to a certain physical
observable.
In the jellium model, we describe the lattice ions by means of their charge density ion (r),
which is taken to be a continuous function of the coordinate r. Let us first establish what
the normal modes in the jellium model. Hereto, we first treat the electron gas as a negatively
charged static background that neutralizes the positive charge of the ions. Of course, this is
not a correct description of the electrons they are much lighter than the ions and, hence,
much more mobile, but it serves to make an important point needed later on.
If the electrons are treated as a static negative background charge, any deviation from
equilibrium ion = ion 0ion is associated with an electric field
1
E = ion , (9.48)
0
where 0 is the electric permittivity of vacuum. The electric field, in turn exerts a force field
on the ions,
f = 0ion E. (9.49)
The ion charge density then satisfies the equation
2 Ze0ion
M ion + ion = 0, (9.50)
t2 0
where M is the ionic mass and Ze is the charge on one ion. The excitations described by
Eq. (9.50) all have the same energy energy does not depend on wavevector! For sure,
these excitations are not the phonons we studied in the previous section. The physical cause
that the jellium excitations we found are not phonons is the long range of the Coulomb
interaction. In fact, those excitations are nothing but the ion lattice equivalent of the
plasmon excitations. They are an artefact of the approximations we used in order to
derive Eq. (9.50); They do not exist for a real ion lattice.
In a real material, the long range Coulomb interaction between the ions is screened by
the electrons. Since electrons are much lighter than the phonons, they react quickly to any
change in the ion density. In fact, to a good approximation we can assume that the electrons
follow the ion density instantaneously, maintaining charge neutrality throughout the material
at all times. As a consequence, the energy price for a perturbation ion of the ion density is
not the Coulomb energy, but the kinetic energy cost of increasing or decreasing the electron
density: If the ionic charge density is changed by an amount ion , the electronic charge
density needs to be changed by ion in order to maintain charge neutrality. The energy
cost associated with the change of electron density leads to a restoring force for the ions.
9.3. JELLIUM MODEL FOR PHONONS 203

The easiest way to estimate that restoring force, is to view it as an electron pressure.
One easily shows that, for a free electron gas, the ground state energy for N electrons confined
to a volume V is
5/3
~2 (3 2 N )
Ee = . (9.51)
10 2 mV 2/3
Hence, one argues that the electronic pressure equals
~2
5/3
pe = 2
3 2 e . (9.52)
15 m
The ionic motion is driven by the gradient of the electronic pressure pe . This leads to a wave
equation for ion ,
2 2
M 2 ion ZF 2 ion = 0. (9.53)
t 3
Solutions of Eq. (9.53) have a valid dispersion relation
r
Zm
q = v F q . (9.54)
3M
Note that, in the jellium model, there is only one phonon mode per wavevector. For a
real lattice, there are three phonon modes per wavevector: one longitudinal mode and two
transverse modes. The two transverse modes are not present in the jellium model, because
they do not correspond to a change of the charge density. Since the electrons interact with
longitudinal phonons only, the absence of transverse normal modes in the jellium model is
not a real problem.
We now present a more microscopic picture of how the conduction electrons determine
the phonon dispersion relation. In our theory well consider the electron-phonon interac-
tion as a perturbation. Without electron-phonon interaction, the phonons are the quantized
oscillations of the ion jellium without electron screening, see Eq. (9.50) above. Such oscil-
lations, which are the equivalent of plasma oscillations in the interacting electron gas, have
frequency E
Z 2 e2 N Ze2 n
2E = = , (9.55)
0 M V 0 M
where n is the electron density. The frequency does not depend on the wavelength of the
phonons, as in the Einstein model of lattice vibrations. Also, in the jellium one has longitu-
dinal phonons only; transverse phonons, which do not correspond to a change in ion density,
can not be described in a jellium model. Thus, the phonon Hamiltonian in the jellium model
is
phonon =
X
H ~E (bq bq + 1/2). (9.56)
q
204 CHAPTER 9. PHONONS AND ELECTRON-PHONON INTERACTION

k q

k+q
Figure 9.1: Diagrammatic representation of phonon line and of electron-phonon interaction vertex.

The relevant phonon Green function is defined with respect to the displacement operators
uq . It is a harmonic oscillator Green function,
1
Dq ( ) = hT uq ( )uq (0)i, Dq0 (in ) = . (9.57)
M (2n + 2E )
The superscript 0 indicates that this phonon Green function is calculated without taking
the electron gas into account.
In this first-principle jellium model, the interaction potential between electrons and ions
is the Coulomb interaction. We thus set Vqie = Ze2 /0 q 2 in Eq. (9.47),
s
2
Ze i 1/2 E M e 2
gq0 = N =i . (9.58)
0 q 0 q 2
Diagrammatic rules for a combined electron-phonon system are quite similar to those of
the previous chapters for electronic systems. We denote a correlation function of displace-
ments with a springy line, see Fig. 9.1. The electron-phonon interaction is a vertex where
a phonon line meets an incoming and outgoing electron line. Energy and momentum are
conserved at the electron-phonon vertex. In the imaginary time formalism, each phonon line
is represented by minus a phonon Green function, D 0 ; each electron-phonon vertex has
weight gq0 . Note that gq0
= gq0 . If the real-time Keldysh formalism is used, the phonon line
represents iD. The phonon lines are directed (although the final answer does not depend on
the direction chosen). The vertices carry weight gq , multiplied by a 2 2 matrix in Keldysh
space, see Sec. 7.1.
The first effect of the electron-phonon interaction is to modify the phonon Green function.
Diagrams for the phonon Green function in the random phase approximation are shown in
Fig. 9.2. In this diagram you recognize the electron polarizability function , which we
calculate in the RPA approximation as well, see Eq. (7.35). We find
Dq0 (in )
DqRPA (in ) =
1 RPA
q (in )Dq0 (in )|gq0 |2 /V e2
9.3. JELLIUM MODEL FOR PHONONS 205

npo
po
npnopo npo
npo pnpn
pno n nno pnpn
ppnopo
no ppo nppn
= + npnonpo
po npo
npo pnpn
pno
no
p no n p pnnp
D = D 0 (D0 ) ( e) (D)
Figure 9.2: Diagrams for the effective phonon Green function D.
qsr
sr qsr qsr sqsqq
qsqrsr
qsr qsr
qsr
qsr
sr
qsr
q qsr
r
s ssqq
sqr
sqr
qsqrsr
q qsr
qsr
q qsr
sr
qsr ssqq
= + sr q q ssqq
sqr
sqr q r
s qsr q r
s q ssqsq
qsr
sqrsr q sr
g = g0 + V ( e) g0
Figure 9.3: Diagrams for the effective electron-phonon interaction g q .

1
= , (9.59)
M (2n + q2 )

where
2E Ze2 n
q2 = = . (9.60)
RPA
q (in ) RPA 0 M
Here we wrote the polarizability function in terms of the dielectric response function RPA =
(1+Vq RPA /e2 )1 , where Vq = e2 /0 q 2 denotes the electron-electron interaction. Substituting
the Thomas-Fermi approximation (7.40) of the dielectric response function, we find, for small
q, r
Zm
q = v F q , (9.61)
3M
which is the Bohm-Staver expression for the phonon dispersion, see Eq. (9.54). We thus
arrive at the satisfactory conclusion that, once the electron-phonon interaction is taken into
account, the phonon frequencies are renormalized from to q , where q is proportional to
q for large wavelengths. Of course, we already reached the same conclusion in Eq. (9.54),
based on macroscopic arguments.
Next, let us consider the electron-phonon interaction. This interaction is also modified
by the presence of the electron gas. Instead of exciting a phonon directly, as is described by
206 CHAPTER 9. PHONONS AND ELECTRON-PHONON INTERACTION

= +

V eff = V RPA + g D g
Figure 9.4: Diagrams for the effective electron-electron interaction V qeff .

the electron-phonon Hamiltonian (9.45), an electron may excite an electron-hole pair, which,
in turn excites a phonon. Again using the random phase approximation, such processes can
be described by the diagrams listed in Fig. 9.3. We thus find

gqRPA (in ) = gq0 (1 + Vq RPA

q /e2 )
gq0
= RPA . (9.62)
 (q, in )
Let us now turn to the electron-electron interactions. A diagrammatic expression for
the effective electron-electron interaction is shown in Fig. 9.4. The left diagram in Fig. 9.4
corresponds to interactions that do not involve the excitations of phonons. This is the RPA
effective interaction we studied in Sec. 7.3. The right diagram contains all processes where
a phonon is excited. Notice that the electron-phonon vertices are renormalized vertices of
Eq. (9.62) and that the phonon propagator is the full renormalized propagator of Eq. (9.59).
Combining everything, we find that the effective electron-electron interaction is
Vq
Vqeff (in ) = RPA
+ gqRPA (in )gq
RPA
(in )DqRPA (in )
q (in )
q2
 
Vq
= RPA 1 2 , (9.63)
q (in ) n + q2

where we have made use of the identity

gq0 gq
0
= M q2 RPA
q (in )Vq . (9.64)

Transforming to real frequencies, we finally obtain the phonon-mediated effective electron-

electron interaction
Vq 2
Vqeff () = RPA . (9.65)
q () 2 q2
9.3. JELLIUM MODEL FOR PHONONS 207

Vqeff()

VqRPA

Figure 9.5: Sketch of the frequency dependence of the effective electron-electron interaction.

Using the Thomas-Fermi approximation for the dielectric response function, this simplifies
to
e2 2
Vqeff, TF () = . (9.66)
0 (q 2 + ks2 )( 2 q2 )
In contrast to the RPA screened interaction, which hardly depended on frequency as long as
 F , the inclusion of electron-phonon interactions gives rise to a frequency dependence
on a much smaller energy scale: V eff depends on frequency on the scale q , as sketched in
Fig. 9.5. For large frequencies q  the effect of the lattice is very small; V eff is very close
to V RPA . The effect of the lattice is most dramatic, however, for small frequencies,  q ,
where the effective electron-electron interaction becomes attractive, rather than repulsive.
What is the effect of the attractive interaction? At first sight, one is inclined to believe
that the effect is small. After all, the interaction is attractive only in a very small frequency
range, . D , where the Debije frequency D is the typical phonon frequency. Nevertheless,
at low temperatures only electrons close to the Fermi level play a role. For these electrons
interactions with small transferred energy are most important. It is precisely for those
processes that the interaction is attractive.
208 CHAPTER 9. PHONONS AND ELECTRON-PHONON INTERACTION

Figure 9.6: Diagram contribution to the renormalization of electron-phonon vertex by phonon

scattering.

9.4 Exercises
Exercise 9.1: Phonon velocity in jellium model

Calculate the dispersion relation for phonons in the jellium model. What is the phonon
velocity? This expression for the phonon disperson is known as the Bohm-Staver law.

Exercise 9.2: Migdals theorem

In principle, the electron-phonon vertex can also be renormalized by the repeated exchange
of phonons, as, e.g., in the diagram of Fig. 9.6. (Note that the phonon line exchanged
between the electrons is a renormalized phonon line.) Show that such diagrams
p are smaller
than the diagrams without the exchange of a phonon by a large factor M/m, where M
and m are ion and electron mass, respectively.
Chapter 10

Superconductivity

The electron-phonon interaction leads to an attractive effective electron-electron interaction.

In this chapter we will review the consequence of the attractive interaction: an instability
that leads to the superconducting state. We also dicuss the fundamentals of a Green function
description of superconductors.

10.1 Cooper instability

What is the effect of an attractive interaction? In order to investigate the effect of a weak
attractive interaction, we consider a cartoon version of the effective interaction we derived
in the previous section.
The cartoon version, used since the original formulation of the theory of superconductivity
by Bardeen, Cooper, and Schrieffer, reflects the fact that the attractive interaction exists for
small energy transfers only, and combines it with our knowledge that only electrons near the
Fermi surface play a role. With this motivation, we look at the interaction Hamiltonian

ee = k 1 1 k 2 2 k3 2 k4 1 k1 +k2 ,k3 +k4 (k1 )(k2 )(k3 )(k4 ),(10.1)

X X
H
2V k1 ,k2 ,k3 ,k4 1 ,2

where the function (k) is nonzero only for |k | < D and is a positive constant.
Note that the interaction (10.1) is not the result of a formal manipulation starting from the
effective phonon-mediated electron-electron interaction. Instead, it is merely a cartoon that is
used because of its technical simplicity. In some respects, it is quite different than the effective
phonon-mediated electron-electron interaction. The interaction (10.1) is instantaneous but
velocity dependent (the interaction depends on k ), whereas the true effective electron-

209
210 CHAPTER 10. SUPERCONDUCTIVITY

electron interaction is retarded. For the application we consider here, these differences are
believed not to play an important role.
In some applications, it is useful to rewrite the interaction (10.1) in coordinate space.
The interaction is local on length scales large compared to vF /D ,
X
Z

Hee = dr 1 (r) 2 (r)2 (r)1 (r). (10.2)
2 ,
1 2

If the coordinate representation (10.2) is used, care must be taken that it is, in fact, the
Fourier transformation of Eq. (10.1), and, thus, requires a momentum cut-off at momenta
of order D /vF .1
ee leads to a new phase, the superconducting phase. In order
The attractive interaction H
to find the temperature or interaction strength at which this new phase starts to appear,
we follow the road taken in our dicussion of magnetism, and search for a zero-frequency
divergence of a response function. A diverging response function signals an instability: A
finite response for no applied perturbation. In the case of a ferromagnet, the respone is
the spin density, the perturbation is a magnetic field, and we found that for sufficiently
strong interactions, the spin density in response to an applied magnetic field diverges. This
divergence signalled the Stoner instability and the appearance of a ferromagnetic state. For
the case of a superconductor, both the perturbation looks rather unphysical: We take
Z
H1 = dr((r) (r) (r) + (r) (r) (r))
(10.3)

as the Hamiltonian of the perturbing field. It is important that we use a grand canonical
formulation, since the perturbation H 1 does not conserve particle number. Indeed, the
perturbation H 1 is unphysical. It can not be a realistic perturbation in an experiment. But
that is not the point! If we were to live in a world one did not know how to apply a magnetic
field, one would consider the perturbation we looked at in Ch. 8 as unphysical. Yet, in such
a world a theoretical physicist could stil ask the question what would happen if one were
1
The exact Fourier transform of Eq. (10.1) has the form

ee =
X Z Z
H dr dr1 dr2 dr3 dr4 (r r1 )(r r2 )(r r3 )(r r4 )
2 ,
1 2

1 (r1 ) 2 (r2 )2 (r3 )1 (r4 ),

where
1 X ikr
(r) = e (k).
V
k
10.1. COOPER INSTABILITY 211

m m m p m p+ n
k k k k k k k k
R= +
k+q k+q k+q k+q k+q k+q k+q k+q
m+ n m+ n m+ n p+ n m+ n p

m p m p
k k k k k k
+ + ...
k+q k+q k+q k+q k+q k+q
m+ n p+ n m+ n p+ n

Figure 10.1: Ladder diagrams contributing to the Cooper correlator R of Eq. (10.8).

to apply a fictitous magnetic field. Inevitably she/he would find that this field creates a
finite spin density, that the spin density response diverges at the Stoner instability, and that
there is a ferromagnetic state for interaction strengths beyond the critical value. Here we
are equally justified in discussing the nonphysical perturbation H 1 . As soon as we find that
it creates a diverging response, we know that we have discovered a new phase.
The response we look for is of the quantity

F (r) = h (r) (r)i. (10.4)

The response of F to the perturbing field (r) is easily found using the Kubo formula,
Z Z
0
F (r, t) = dt dr0 RR (r r0 ; t t0 ) (r0 , t0 ), (10.5)

where
RR (r r0 , t t0 ) = i(t t0 )h[ (r, t) (r, t), (r0 , t0 ) (r0 , t0 )] . (10.6)
Here we omitted a contribution to the response that involves a commutator of four creation
operators, which is nonzero in the non-interacting ground state. Fourier transforming and
switching from the retarded correlation function to the temperature correlation function, we
find
X
Rq ( ) = Rkk0 q ( ), (10.7)
k,k0
1 2 ,3 4
Rk,k 0 ,q ( ) = hT k+q,1 ( )k,2 ( )k

0 +q, (0)k0 , (0)i.
3 4
(10.8)

We calculate the response function R using diagrammatic perturbation theory, taking

the effective electron-electron interaction (10.1). The diagrams contributing to R are very
212 CHAPTER 10. SUPERCONDUCTIVITY

similar to the diagrams that contributed to the spin susceptibility in chapter 8. They are
shown in Fig. 10.1. Note that there are contributions from the direct interaction and from
the exchange interaction, which come with an extra minus sign. Summing the geometric
series in Fig. 10.1, we find
!
1 2 ,3 4
X
Rk,k 0 ,q (in ) = (k,k0 1 3 2 4 k,qk0 1 4 2 3 ) T Gk+q (im + in )Gk (im )
m
!
1 X
+ (1 3 2 4 1 4 2 3 ) T Gk+q (im + in )Gk (im )
V m
!
X
T Gk0 +q (ip + in )Gk0 (ip ) . (10.9)
p

is a renormalized interaction,
Here
" #1
= 1+ T X X
0
Gk00 (im )Gk00 +q (im + in ) , (10.10)
V 00 mk
P0 P
where the symbol k denotes k (k)(k + q). Performing the Matsubara summations,
we have
X
T Gk+q (im + in )Gk (im )
m
X 1
= T
m
[im + in (k+q ][im (k )]
1 tanh[(k+q )/2T ] + tanh[(k )/2T ]
= . (10.11)
2 k + k+q 2 in
We perform the analytical continuation in + i and consider the limit q 0,
0. Substituting Eq. (10.11) into the expression (10.10) for the effective interaction
strength, replacing the momentum summation by an integration over energy, and integrating
by parts, we find
Z ~D
1 X tanh[(k )/2T ] tanh(/2T )
(k) = d (10.12)
2V k k ~D 2
Z ~D /2T
~D ~D ln x
= tanh ln dx .
2T 2T 0 cosh2 (x)
10.2. GREENS FUNCTION DESCRIPTION OF A SUPERCONDUCTOR 213

The second term is convergent and approaches the value 0.81 as T 0, whereas the first
term diverges logarithmically. We thus see that the denominator in Eq. (10.10) diverges if

1 = (ln(~D /2T ) + 0.81), or T = Tc = 1.14~D e1/ . (10.13)

It is important to realize that this divergence happens for an arbitrarily weak attractive
interaction, provided the temperature is sufficiently low.
corresponds to a divergence of the correlation function R. The
Of course, a divergence of
reason why we studied the correlation function R was precisely because of this divergence:
a divergent correlation function is a signal of an instability. At high temperatures T  Tc ,
the response function Rq is finite, indicating a finite response to the perturbation (r).
Lowering the temperature, we see that the response to the fictitious field (r) diverges upon
approaching the critical temperature Tc . The divergence signals a phase transition to a
state in which F (r) acquires a nonzero value spontaneously.2 This is the superconducting
state. This important observation is the basis for the Green function description of the
superconducting state of the next section.
The fact that the quantity F becomes nonzero in the superconducting phase has two
consequences: First, the absolute value |F | becomes nonzero, and, second, F acquires a
phase. Upon rethinking this, the second statement may be more striking than the first one.
After all, what determines the phase of F ? For this, it is useful to again compare the
normal-metal-superconductor phase transition to the normal-metalferromagnet transition
we studied in the previous chapter. Upon passing the Stoner instability, the spin polarization
acquired a finite magnitude, and a certain direction, although the Hamiltonian did not
contain any preferred direction. The spin-rotational symmetry is broken spontaneously upon
entering the ferromagnetic phase. In principle, the direction of the magnetization could be
fixed by the addition of an infinitesimal magnetic field to the Hamiltonian. In the same
way, the superconducting phase is a broken symmetry phase. The quantity F acquires a
phase, whereas the Hamiltonian does not contain a term that determines the phase of F .
As in the case of the ferromagnet, the symmetry may be broken by a small perturbation,
which has the form of the perturbation of Eq. (10.3).

10.2 Greens function description of a superconductor

In this section, we will derive an equation of motion for the superconductor Green functions.
Hereto, we need the equation of motion for the creation and annihilation operators. Using
2
If one would insist on a canonical formulation, the quantity that turns nonzero at the normal-metal
superconductor transition is hN + 2|k

k |N i.
214 CHAPTER 10. SUPERCONDUCTIVITY

the coordinate representation (10.2) of the interaction Hamiltonian, we find

" #
0 ) (r) + 0 (r)0 (r) (r)
X
(r) = (H

0
" #
) (r) (r) 0 (r)0 (r) .
X
(r) = (H (10.14)
0
0

Here H is the complex conjugate of the Hamiltonian H 0 . Eventually, the repulsive part of
0
the interaction at higher frequencies can be included into H 0 as a self-consistent Hartree-Fock
potential.
Slightly generalizing the arguments of the preceding section, we characterize the super-
conducting state by means of the standard Green function G0 (r, ; r0 0 ) and by the so-called
anomalous Green function

F0 (r, ; r0 , 0 ) = hT (r, )0 (r0 , 0 )i,

+
F 0 0
0 (r, ; r , ) = hT (r, ) 0 (r0 , 0 )i, (10.15)

where H 0 is the electron Hamiltonian without interactions. Note that the absence of a
minus sign in the definition of the anomalous Green functions F and F + . Writing down the
equation of motion for G, we the find
 
0 G0 (r, r0 ; )
G0 (r, r0 ; ) = (r r0 )( )0 + H (10.16)
D E
T 00 (r, )00 (r, ) (r, ) 0 (r0 , 0) .
X
+
00

The imaginary-time arguments of the creation and annihilation operators in the last term
on the r.h.s. of Eq. (10.17) should be increased by an infinitesimal amount in order to ensure
the correct order under the time-ordering operation T . Anticipating a description in terms
of a self-consistent field (as in the Hartree-Fock approximation we discussed in Ch. 7), the
average of a product of four creation and annihilation operators on the right hand side of Eq.
(10.16) can be written as a product of two pair averages using Wicks theorem. Application
of Wicks theorem results in three terms. Two of those are the Hartree and Fock terms we
encountered before in Chapter 7. These can be absorbed into the Hartree and Fock self
consistent potentials, which is included in the Hamiltonian H 0 . The third term involves the
anomalous Green function F ,
D E

T 00 (r, ) (r, ) (r, )0 (r , 0) F00 (r, + ; r, )F+00 0 (r, ; r0 , 0),
00
0
10.2. GREENS FUNCTION DESCRIPTION OF A SUPERCONDUCTOR 215

where is a positive infinitesimal. Defining the field 0 (r) as3

0 (r) = 0 (r) = F0 (r, + ; r, ), (10.17)
we arrive at the equation of motion
 

(H0 ) G0 (r, ; r0 , 0 )

X
+ 00 (r)F+00 0 (r, ; r0 , 0 ) = (r r0 )( 0 ). (10.18)
00

Note that (r) is antisymmetric in the spin indices. This reflects the fact that the super-
conducting instability exists for pairs of electrons with zero spin only. Similarly, one derives
an equation of motion for the anomalous Green function,
 
) F + 0 (r, ; r0 , 0 )

X
(H 0 00 (r)G00 0 (r, ; r0 , 0 ) = 0, (10.19)

00
 
0 ) F0 (r, ; r0 , 0 )
X
(H 00 (r)G0 00 (r0 , 0 ; r, ) = 0, (10.20)
00

where
0 (r) = F
+
0 (r, + ; r, ). (10.21)
These equations are known as Gorkovs equations. Note that Eq. (10.19) contains the com-
plex conjugate of the Hamiltonian H 0.
It is the presence of the field (r) that makes the Gorkov equations different from the
equations for the Green function in a normal metal. The field is related to the energy
gap for quasiparticle excitations in a superconductor (see below). It also serves as an order
parameter that distinguishes the superconducting state from the normal state. In that sense,
it plays the same role as the magnetization in the normal-metalferromagnet transition.
Solving these equations by means of a Fourier transform, we find
(in + (k ))kk0 0
Gk,k0 0 (in ) =
n2 + (k )2 + | |2
+ 0 kk0
Fk,k 0 0 (in ) = . (10.22)
n2 + (k )2 + | |2
3
Strictly speaking, one would need to take the finite range of the attractive interaction into account. In
accordance with the momentum cut off at |k | ~D , the interaction (10.2) has a range of order vF /D .
However, as well see below, the anomalous Green function F(r, r0 ) is smooth on the scale vF /D , its spatial
variations occurring on the much larger scale vF /Tc only. This implies that the approximation made in
setting both coordinates equal in the argument of F is justified.
216 CHAPTER 10. SUPERCONDUCTIVITY

Self-consistency is achieved upon substitution of the solution for F into the definition of ,
T XX 1
1= . (10.23)
V n n + (k )2 + | |2
2
k

Performing the summation over Matsubara frequencies, replacing the summation over mo-
menta k by an integration over energies k and remembering that only momenta with
|k | < ~D play a role, we find the self-consistency condition
Z ~D p
tanh[ 2 + | |2 /2T ]
1 = d p . (10.24)
0 2 + | |2
This is the same equation as the gap equation from standard BCS theory.
The highest temperature with a nontrivial solution of the self-consistency equation is the
point where the normal-metalsuperconductor phase transition occurs. Since = 0 at
T = Tc , we find that the critical temperature satisfies the equation
Z ~D
tanh(/2Tc)
1 = d . (10.25)
0
This is precisely the same equation as we found for the divergence of the correlator R in the
previous section, see Eq. (10.12) above. Hence
Tc = 1.14~D e1/ . (10.26)
Quasiparticle excitations of the superconductor correspond to poles of the retarded Green
functions G and F . Analytical continuation of Eq. (10.22) gives
( + k )kk0 0
GR
k,k0 0 () =
(k )2 + ||2 ( + i)2
+R 0 kk0
Fk,k 0 0 () = . (10.27)
(k )2 + ||2 ( + i)2
We thus conclude that the quasiparticle energies are
p
k = (k )2 + ||2 . (10.28)
From here we conclude that the excitation spectrum has a gap of size ||.
We close this section with a few remarks on the role of a vector potential A(r) and gauge
invariance. A vector potential can be included in the Hamiltonian H 0 by the replacement
ie
r r A. (10.29)
~c
10.3. A SUPERCONDUCTOR IN A WEAK ELECTROMAGNETIC FIELD 217

In the complex conjugate Hamiltonian H this corresponds to the replacement r r +

0
ieA/~c. Gauge invariance requires that the theory is invariant under transformations

A A + r , (10.30)

where is an arbitrary function of the position r. Green functions are not gauge-invariant
objects. Under the gauge transformation (10.30), the Green functions transform as
0
G(r, r0 ) G(r, r0 )eie((r)(r ))/~c ,
0
F (r, r0 ) F (r, r0 )eie((r)+(r ))/~c , (10.31)
0
F + (r, r0 ) F + (r, r0 )eie((r)+(r ))/~c .

In particular, this implies that the field (r) transforms as

(r) (r)e2ie(r)/~c . (10.32)

The Gorkov equations serve as the basis for further investigations of the superconducting
state. Results obtained from the Gorkov equations agree with those obtained from the
mean-field BCS theory. For reviews, we refer to the book by Tinkham.4 Here we forego a
discussion of thermodynamic properties and the Josephson effect, and limit ourselves to a
discussion of the response of superconductors to electromagnetic radiation.

10.3 A superconductor in a weak electromagnetic field

In this section, we calculate the current density in a superconductor in response to an external
vector potential A.
Starting point of the calculation is the Kubo formula. Repeating results of Sec. 5.6, we
have
X ee
je,q () = R (q, )A,q () A,q (), (10.33)

m

where R is the retarded current density autocorrelation function,

1
(q, ) = hT je,q ( )je,q (0)i, (10.34)
V
4
See Introduction to Superconductivity, M. Tinkham (Mc Graw Hill College Div., 1995). Other useful
references are Superconductivity of Metals and Alloys, P.G. de Gennes (Perseus, 1999) and Methods of
Quantum Field Theory in Statistical Physics, A.A. Abrikosov, L.P. Gorkov, and I.E. Dzyaloshinski (Dover
1977).
218 CHAPTER 10. SUPERCONDUCTIVITY

and
e X
je,q = (2k + q)k,

k+q, . (10.35)
2m k
Equation (10.33) describes the linear response of the current density to a vector potential
A. We used this relation in Sec. 5.6 to find the conductivity of a disordered normal metal,
the coefficient of proportionality between the current density and the electric field E() =
iA(). In the normal metal, the choice of the gauge for the vector potential A was not
relevant. This is different for the case of a superconductor. The reason is that the vector
potential A enters in the equation for the single-particle Green functions G and F , and,
hence, affects the value of the order parameter . To linear order in A, can only
depend on r A. Hence, if we choose the gauge such that r A = 0, is unchanged
to linear order in the vector potential. This choice of the gauge is known as the London
gauge.
Let us now calculate the current density autocorrelation function for the case of a clean
superconductor. Using the Matsubara frequency language, we find

e2 T X
(q, in ) = (2k + q )(2k + q ) (10.36)
2m2 V k,m
+
 
Gk+q,k+q (im + in )Gk;k (im ) + Fk+q,k+q (im + in )Fk,k (im ) .

Changing variables to k = k (q/2) and m = m (n /2), we rewrite this as

2e2 T X h
+
i
(q, in ) = k k
G k+ ,k+ (i m+ )G k ,k (i m ) + F k+ ,k+ (i m+ )F k ,k (i m )
m2 V
k,m

2e T X2
(im+ + k+ )(im + k ) + ||2
= k k 2 2
,
m2 V k,m (m+ + (k+ )2 + ||2 )(m + (k )2 + ||2 )

where we substituted the solution (10.22) for the Green functions G and F . Next, we replace
the summation over k by an integration over = k and over the angle between q and
k.
Next we introduce polar coordinates for the vector k. We choose the direction of q as the
polar axis. The average over the azimuthal angle can be performed directly. Noting that
q and A are orthogonal in the London gauge, upon performing the average over we find
that is diagonal, and proportional to (1/2) sin2 , where is the angle between q and k.
Further, since the main contribution comes from momenta close to the Fermi momentum,
we set the magnitude of the factors k and k equal to kF . Finally, using the equality
kF2 /m = 3n/2, where n is the electron density and the normal-metal density of states at
10.3. A SUPERCONDUCTOR IN A WEAK ELECTROMAGNETIC FIELD 219

the Fermi level (per spin direction and per unit volume), and replacing the summation over
k by an integration over = k and an integration over , we find
Z
3T e2 n XZ (im+ + + )(im + ) + ||2
(q, in ) = d d sin3 2 2 2 )( 2 2 2)
,
4m m 0 ( m+ + + + || m + + ||

where = (vF q/2) cos .

Because of the approximations involved, the summation over Matsubara frequencies
should be performed before the integration over . Performing the integration over first
gives a different answer. The answer depends on the order of the integrations because the in-
tegral (10.37) is formally divergent. Our formulas for the -dependence of the single-electron
Green functions are valid close to the Fermi level only. After the summation over the Mat-
subara frequencies, the integration is, in fact, convergent, the main contribution coming
from momenta near the Fermi level.
For the case of the superconductor, there is a trick that allows us to do the integration
first, nevertheless. The trick is to look at the difference between current density correlation
functions for the normal metal and for the superconductor. For this difference, the integra-
tions and summations over and m are convergent, so that the order of integrations is no
longer important. Hence, if we calculate the difference between current-current correlation
functions for the normal metal and superconducting states, we can perform the integra-
tion before the summation over Matsubara frequencies. The current density autocorrelation
function for the normal metal is
Z
N 3T e2 n XZ 1
(q, in ) = d d sin3
4m m 0 (im+ + )(im )
ee
= , (10.37)
m
see Ex. 5.6. Integrating with respect to we then obtain

ee 3T ee XZ
(q, in ) = + d sin3
m 4m m 0
" p p
(im+ + i m+ 2
+ ||2 )(im qvF cos + i m+ 2
+ ||2 ) + ||2
2
p
2
p
2
(m + ||2 + (qvF cos i m+ + ||2 )2 ) m+ + ||2
p p #
(im + i m 2
+ ||2 )(im+ qvF cos + i m 2
+ ||2 ) + ||2
+ 2
p
2
p
2
.
(m+ + ||2 + (qvF cos i m + ||2 )2 ) m + ||2
(10.38)
220 CHAPTER 10. SUPERCONDUCTIVITY

C1 C1
C2

C2
C2

i n i n C2

Figure 10.2: Integration contour for the calculation of Eq. (10.40).

Next, we perform the summation over the Matsubara frequency m . We undo the shift
of variables for the Matsubara frequencies, i.e., we replace m+ m + n and m m .
We then represent the summation over Matsubara frequencies as an integral in the complex
plane and deform the integration contour as shown in Fig. 10.2. The integrand has a square
root branch cut. In identifying the correct sign of the square root, one needs the relations
 p
p
2 2 ||2 p 2 if || < ||,
|| ( i) = 2 2
(10.39)
i sign() || if || ||.
After the integration contours have been deformed as in Fig. 10.2, we can perform the
analytical continuation in + i.
Obtaining the final equations for (q, ), which contain one integration over a real
energy variable and one integration over the angle is straightforward, but tedious. Below,
well restrict ourselves to static response 0. In this case, the integration above the lower
branch cut cancels the integration below the upper branch cut in Fig. 10.2. What remains is
the integration above the upper branch cut and the integration below the lower branch cut,
Z Z
R ee 3ee 3
(q, 0) = + d sin d tanh(/2T )
m 8m 0
X ||2
p .
2 ||2 (||2 ( i)2 + (vF2 q 2 /4) cos2 )
(10.40)
The easiest way to evaluate this integral is to rewrite it as a sum over Matsubara frequencies,
ee 3ee T XZ
R
(q, 0) = + d sin3
m 4m m 0
10.3. A SUPERCONDUCTOR IN A WEAK ELECTROMAGNETIC FIELD 221

||2
p . (10.41)
2 + ||2 + (v 2 q 2 /4) cos2 )
(m 2 + ||2
m
F

Evaluation of this expression is further simplified in the limits qvF  max(||, T ) and
qvF  max(||, T ).
If qvF  max(||, T ), we can neglect the q-dependence of the summand and find
 
R ee ns (T )
(q, 0) = 1 , (10.42)
m n
where ns is the so-called superconducting density,

nT ||2 X 1
ns = . (10.43)
2 m
(m + ||2 )3/2
2

For zero temperature the superconducting density is equal to the electron density n. The
superconducting density approaches zero as T approaches Tc .
If qvF  max(||, T ), the main contribution to the integral over the angle comes from
angles close to = /2. For those angles we may set sin = 1. Replacing the -integration
by an integration over cos and extending the integration interval to the entire real axis, we
find
ee 3 2 ee T X ||2
R
(q, 0) = + 2 + ||2
m 2mqvF m
m
ee 3 2 ee || ||
= + tanh . (10.44)
m 4mqvF 2T
Unlike in the case of a normal metal, we find that for a superconductor the current density
j(q, ) is proportional to the vector potential A itself, and not to its time-derivative. This
dependence is the cause of the Meissner effect, the phenomenon that no magnetic field can
exist in a superconductor, except for a thin layer of thickness close to the superconductor
surface. For a superconductor for which vF /  max(T, ), the Meissner effect can be
described with the help of Eq. (10.42) above,

ns e 2
j= A. (10.45)
m
2
p (10.45) with the Maxwell equation r A = j for the London gauge r A = 0,
Combining Eq.
we find = m/ns e2 . Equation (10.45) is known as the London equation, and is known as
the London penetration depth. A superconductor for which  vF / max(T, ) is said to be
222 CHAPTER 10. SUPERCONDUCTIVITY

of London type. For a superconductor for which  vF / max(T, ), Eq. (10.44) has to be
used instead of Eq. (10.42). In this case, the mathematical solution is more complicated, but
the qualitative conclusion, no magnetic field inside the superconductor, except for a surface
layer of thickness , remains true. Superconductors for which  vF / max(T, ) are said
to be of Pippard type. Equation (10.44) is known as the Pippard equation.
Most pure superconductors at T  Tc are of Pippard type. Close to Tc , increases, and
a crossover to the London type of behavior is seen.
These calculations can be repeated for the case of a disordered superconductor with
elastic mean free time . The calculation proceeds along the lines of Sec. 5.6 and has the
following results: First, the addition of impurities does not affect the magnitude of the
superconducting order parameter . Second, the effect of impurities is small if qvF  1,
whereas, in the opposite limit, qvF  1, one recovers the London equation (10.45), with a
reduced superconducting density
nT ||2 X 1
ns = p . (10.46)
2 m ( 2 + ||2 )(
m 2 + ||2 + 1/2 )
m

As a result of the reduced superconducting density, the London penetration depth is

strongly enhanced. As a result, for sufficiently strong disorder, disordered superconductors
are always in the London limit.

10.4 Quasiclassical theory

The Green functions for a superconductor satisfy the Gorkov equation (10.19). Apart from
the fact that they involve three different Green functions, G, F , and F + , and apart from the
self-consistency condition (10.21), the Gorkov equations are nothing but the defining equa-
tion for the single-particle Green function of non-interacting particles. For non-interacting
particles, we saw that a full quantum mechanical treatment was not necessary in most cases.
(The exception is if one is interested in quantum interference effects, such as the weak local-
ization correction to the conductivity.) Instead of a full quantum solution, most problems
are amendable to a semiclassical treatment. For normal metals, the resulting equation is the
Boltzmann equation. A semiclassical theory is particularly useful if a system is not homoge-
neous, as is the case, e.g., in the presence of weak electromagnetic fields or near the interface
of a normal metal and a superconductor. We now discuss a semiclassical treatment of the
Gorkov equations.
As before, we consider the single-particle Green functions as operators on the Hilbert
space. The operator action is then nothing but the convolution. In order to be able to deal
with non-equilibrium situations, we use the real time formalism. We will be considering
10.4. QUASICLASSICAL THEORY 223

the effect of spatial dependences of the potentials, but we will not consider time-dependent
potentials or fields.
In order to simplify our notation, we combine the standard and the anomalous Green
functions into a 2 2 matrix,
!
ihT
c (r, t)
0 0
(r , t )i ihT
c (r, t)
(r 0 0
, t )i
t; r0 , t0 ) =
G(r,
. (10.47)
ihTc (r, t) (r0 , t0 )i ihTc (r, t) (r0 , t0 )i

The matrix notion of Eq. (10.47) was first proposed by Nambu. Using Nambus notation,
and using the operator notation for the Green function, the Gorkov equations read
 
0 +
3 t H
i~ G = (r r0 )c (t t0 ) (10.48)

where 3 is the Pauli matrix with respect to the Nambu grading of Eq. (10.47), H 0 is the
Hamiltonian without superconducting order parameter,
" 2 #
2

~ ie
H0 (r, t; r0 , t0 ) = (r r0 )(t t0 ) r 3 A(r) + e(r) , (10.49)
2m ~c

is the matrix notation for the superconducting order parameter (r) = h (r, t) (r, t)i,

 
t; r0 , t0 ) = (r r0 )(t t0 ) 0 (r)
(r, , (10.50)
(r) 0

and the delta function c (t t0 ) is defined with respect to the contour points t and t0 , not
with respect to times.5
If we consider a superconductor with impurities with a short-range potential, the ensemble
average over the impurities may be performed. This gives an additional self-energy term that
must be included in the in the Gorkov equation. With inclusion of the self energy, the Gorkov
equations (10.48) read
 
0
3 t H
i~ + G = (r r0 )c (t t0 ). (10.51)

Following the procedure we used previously for normal metals, we introduce the matrix
notation for the Keldysh Green functions, cf. Eq. (2.61). Together with the Nambu matrix
5
If a notation in terms of times is used, c (t t0 ) = (t t0 ) if t and t0 are both on the upper branch of
the Keldysh contour, c (t t0 ) = (t t0 ) if t and t0 are both on the lower branch of the Keldysh contour,
and c (t t0 ) = 0 otherwise.
224 CHAPTER 10. SUPERCONDUCTIVITY

notation, this means that the Green function is now represented by a 44 matrix. Rewriting
the Gorkov equations in this way, one finds
 
0
3 t H
i~ +
G = I, (10.52)

are 4 4 matrices in
where I is the identity operator. In Eq. (10.52), the matrices 3 and
the combined Keldysh/Nambu grading. They are obtained from the corresponding matrices
with the Nambu grading only by taking the tensor product with the 2 2 unit matrix in the
Keldysh grading.
The Gorkov equation may also be written as
 
i~
G 0
3 t H +
= I. (10.53)

Taking the difference of Eqs. (10.52) and (10.53), one finds

h i
0
i~t 3 H + ,
G = 0. (10.54)

This is the equation well study in more detail below.

In order to arrive at a quasiclassical description, we perform a Fourier transform with
respect to the coordinate difference r r0 . We also perform a Fourier transform to the time
difference t t0 . Since the Green functions and the Hamiltonian do not depend on time, the
Fourier transform to time can be taken exactly. The Fourier transform to the coordinate
R; ) satisfies
difference is treated in the gradient approximation. One then finds that G(k,
the differential equation
h i h i h i
3 + i~v R +
~ ,
G i R ( + + e), k G
+ i k ,
R G = 0.
+ +
(10.55)

where v = (~/m)(k 3 eA/~c) is the electron velocity and

R = R 3 (ie/~c)A(R) (10.56)

the covariant derivative. In writing down Eq. (10.55) we neglected the effect of a magnetic
field on the motion of the electrons.
If impurities are modeled by the Gaussian white noise potential and impurity scattering
reads
is treated in the lowest order Born approximation, the self-energy
1
Z

= R, ),
dk G(k, (10.57)
2
10.4. QUASICLASSICAL THEORY 225

cf. Eq. (5.34).

We now define the quasiclassical Green function by parameterizing the wavevector k
through the energy and the direction n, and integrating over ,
i
Z
g(R, n, ) = dG(R, k, ). (10.58)

Neglecting the weak dependence of the self energy, as well as the direction of G on n if
multiplied with a spatial derivative, we arrive at an equation for the quasiclassical Green
function only,
h i
i~vF n R + ~
3 , g = 0. (10.59)

Equation (10.59) is known as the Eilenberger equation. The quasiclassical g has a matrix
structure in Nambu space,  
g f
g = . (10.60)
f + g
You can verify this either from the definition of the quasiclassical Green function itself, or
by noting that the Eilenberger equation contains commutators only, so that one can search
for solutions that are linear combinations of the three Pauli matrices 1 , 2 , and 3 . Also,
f + is the time reversed of f , which need not be the complex conjugate of f . The self-energy
now reads
i
= h
g in , (10.61)
2
where the brackets h. . .in indicate an average over all directions n.
Note that the Eilenberger equation is homogeneous in g, so that it does not fix the
magnitude of the quasiclassical Green function. One can fix the magnitude of g as follows:
Because of the special matrix structure of g, cf. Eq. (10.60), the square gg is proportional
to the unit matrix. Upon substition into Eq. (10.59) one finds that the proportionality
constant is uniform throughout space. Both deep inside a normal metal and deep inside a
superconductor the solution for the quasiclassical Green function is known: it obeys

gg = 1, (10.62)

so that the proportionality constant must be unity throughout space.

It is instructive to write down the solutions for the quasiclassical Green functions in
equilibrium. In that case, it is sufficient to calculate the retarded Green function, which
obeys the equations
[~ gR ] = 0, gRgR = 1.
3 + , (10.63)
226 CHAPTER 10. SUPERCONDUCTIVITY

Together with the condition that gR has the matrix structure (10.60), one then finds
 
R 1 ~
g = p . (10.64)
2 ||2 ~
For ~ < || one determines the phase of the square roots by setting + i, where
is a positive infinitesimal. The advanced Green function then reads gA = 3 gR 3 , where
the superscript means hermitian conjugation. In a normal metal ( = 0), the solutions
for the quasiclassical Green functions are gR = 3 , gA =
3 , in agreement with Ex. 3.4.
If scattering from impurities is very strong, the electron motion is diffusive, not ballistic,
and the equations for the quasiclassical Green functions can be further simplified. As well
see below, strong means that the scattering rate ~/ is large in comparison to both T and
. The role of impurity scattering is to randomize the direction of the electrons velocity.
Hence, one expects that a description in terms of a quasiclassical Green function that is
averaged over all directions of the electrons velocity is possible. In order to arrive at such
a description, we first use an intermediate formulation in which we keep information on the
current density: we take g(n) to be of the form
g(n) = g0 + n g
1 . (10.65)
Similarly, the self energy is written
0 + n
(n) = 1. (10.66)
For a white noise potential, one then finds from Eq. (10.61) that
i
0 = 1 = 0.
g0 , (10.67)
2
For strong impurity scattering, one expects that g 1 is small in comparison to g
0 . Hence,
we expand the Eilenberger equation and the normalization condition up to terms of first
1 . The normalization condition for the quasiclassical Green function thus reads
order in g
g0 g0 = 1, [ 1 ]+ = 0.
g0 , g (10.68)
An equation for the n-averaged Green functions g0 and g 1 is obtained upon substitution of
the angular dependencies (10.65) and (10.66) into the Eilenberger equation (10.59). One
then finds
h i h i
0 = ~ 1 + ivF n R , g0 + n g
g0 + n g
3 , 1 [ 0 , g0 + n g
1 ]

h i h
g0 + ivF n R , g0 + n g
i i
~ 3 , 1 g0 , n g
[ 1 ] . (10.69)
2
10.4. QUASICLASSICAL THEORY 227

superconductor normal metal

x
x= 0
Figure 10.3: Schematic drawing of the interface of a normal-metal and a superconductor.
The Usadel equation is used to calculate the density of states in the normal metal.

Since the scattering rate ~/ is large in comparison to ~ and , we have neglect the term
proportional to g 1 in the first commutator in comparison to the commutator of g0 and g1 .
Separate equations for g0 and g 1 are found upon averaging Eq. (10.69) over n and upon
multiplication of Eq. (10.69) by n, followed by averaging over n,
h i vF h i h i 1
~ g0
3 , +i 1 = 0, vF R , g0 +
R , g [ 1 ] = 0.
g0 , g (10.70)
3 2
Using the normalization condition for the quasiclassical Green function, the explicit solution
1 can be found from the second equation,
for g
h i
1 = vF g0 R , g0 .
g (10.71)

Upon substitution of this equation into the first equation (10.70), one finds
h i  h i 
~ g0 iD R , g0 R , g0
3 , = 0, (10.72)

where D = (1/3)vF 2 is the diffusion constant. Equation (10.72) is known as the Usadel
equation.
As an application of the quasiclassical formalism, we now calculate the density of states
in a diffusive normal metal wire that is attached to a superconductor, see Fig. 10.3. This
is a calculation that is described in detail in, e.g., W. Belzig, C. Bruder, and G. Schon,
Phys. Rev. B 54, 9443 (1996). The density of states is an equilibrium property, so that it
is sufficient to calculate the retarded Green function alone. Taking the retarded component
228 CHAPTER 10. SUPERCONDUCTIVITY

of the Usadel equation (10.72), one finds

h i  h i 
R
3 , g0
~ R R
iD R , g0 R , g0 = 0. (10.73)

In matrix notation, the commutator [R , g] reads

 
g ( (2ie/~c)A)f
[R , g] = R R
. (10.74)
(R + (2ie/~c)A)f + R g
Applying this relation twice, one finds that Eq. (10.73) gives the following equation for the
anomalous Green function
2~f R + g R iDg R (R (2ie/~c)A)2 f R + iDf R R
2 R
g = 0. (10.75)
We specialize to the case that there is no magnetic field, A = 0, in which one may set
f = f + . Then Eq. (10.75), together with the normalization condition (g R )2 + (f R )2 = 1
is sufficient to determine the quasiclassical Green functions. We take the order parameter
to be purely imaginary, and i instead of , and parameterize the quasiclassical Green
function as  
cos sin
g = . (10.76)
sin cos
We consider the case that the region x < 0 is a superconductor and that the region
x > 0 is a normal metal. We do not calculate the superconductor order parameter (r)
self-consistently. Instead, we set (r) = , independent of r, for x < 0, and (r) = 0 for
x > 0. Not enforcing self-consistency will result in a small quantitative error, but it will not
affect our qualitative conclusions. For simplicity, we assume that the superconductor and the
normal metal have the same densities of states and the same diffusion coefficients. In that
case, is continuous at the interface between the normal metal and the superconductor.
Rewriting Eq. (10.75) in terms of the angle and solving it with the boudnary conditions
that is continuous at the normal-metalsuperconductor interface, that cos() 1 deep
inside the normal metal, and that cos() ~/( 2 2 )1/2 deep inside the superconductor,
one finds
h i
4 arctan ex 2i/D tan 0 if x > 0,
4
(x, ) = h i (10.77)
S + 4 arctan ex 2(2 2 )1/2 /D tan 0 S if x < 0,
4

where the angles S and 0 are defined as

i i~ 0
S = arctan , sin(0 S ) = 2 2 1/4
sin . (10.78)
~ ( ) 2
10.5. EXERCISES 229

From this solution one can calculate the spectral density in the normal metal. Simple
expressions are obtained in the limit ||  , for which one finds S = 0 = /2, and

(x, ) 4 tan(/8)ex 2i/D for x  D/||.
p
(10.79)
Hence, we find that the density of states is reduced below the normal-state value 0 , even
in the normal metal,
() = 0 Re cos (x, )
h p i
0 1 8 tan2 (/8)e2x ||/D cos(2x ||/D) . (10.80)
p
This result is valid for x  D/|| only. One can find a more accurate answer valid for all
x by using Eq. (10.77) for (x, ).
Of more interest than the actual numbers one gets is the observation that, in the normal
metal, is different from zero. This not only means that the density of states is different
from the normal-state density of states. It also means the anomalous Green function is
nonzero in the normal metal! Clearly, it is the proximity to the superconductor that is
responsible for this fact. The transfer of superconducting properties from a superconductor
to a nearby normal metal is referred to as the superconductor proximity effect.
The proximity effect in normal metals has been very well studied, both experimentally and
theoretically. You can find more information in the quasiclassical Green function approach
in, e.g., W. Belzig et al., Superlatt. Microstruct. 25 1251 (1999) or in V. Chandrasekhar,
in The Physics of Superconductors: Vol. II, Ed. K. H. Bennemann and J. B. Ketterson
(Springer, Berlin, 2004), cond-mat/0312507.

10.5 Exercises
Exercise 10.1: Thermodynamic properties

One can calculate the effect of the superconductivity on the free energy F using the general
relation between F and the interaction Hamiltonian H 1 of Eq. (10.1).
Z
1 0
F F0 = d0 0 hH 1 ( )i. (10.81)
0
(a) Keeping contributions that are specific for a superconductor only, show that the su-
perconductivity contribution F to the free energy reads
Z 0
d
F = 02
|(0 )|2 , (10.82)
0
230 CHAPTER 10. SUPERCONDUCTIVITY

where (0 ) is the superconducting order parameter at interaction strength 0 .

(b) One can rewrite the gap equation (10.24) as

" #
2D X
1 = ln 2 (1)n+1 K0 (n||/T ) . (10.83)
|| n=1

Using this result, together with Eq. (10.82), show that

"
#
2 Z n||/T
||2 X T
F = 2 (1)n+1 2 K1 (x)x2 dx , (10.84)
2 n=1
n 0

where K1 is a Bessel function.

(c) Show that, at temperatures T  Tc , the heat capacity of a superconductor is

r
250 0 /T
C = 2 e , (10.85)
T3
where 0 is the magnitude of the order parameter at zero temperature.
Chapter 11

Electrons in one dimension

11.1 Why is one dimension different?

With the exception of a short discussion in chapter 6 we have studied electrons in three
dimensions. We encountered two possible excitations of an electron liquid in three dimen-
sions: particle-hole pairs and plasma modes. These excitations appeared as the support of
the imaginary part of the polarizability and susceptibility at finite frequency and finite
wavevector q. The plasma modes have a high threshold frequency for excitation, leaving the
particle-hole pairs as the main type of excitations at long wavelengths and low frequencies.
This observation, together with the observation from Fermi liquid theory that particle-hole
pairs are long-lived excitations if their energy is low, is the main justification of the use of
the non-interacting electron picture as the starting point for our description of the three-
dimensional electron liquid.
In one dimension, the situation is completely different. Here, with one dimension we
mean a truly one dimensional system, i.e., a quantum wire with a width a so small that
only one transverse mode is populated at the Fermi energy. In practice, this means a F ,
where F is the Fermi wavelength.1 In order to see why the one-dimensional electron liquid
is so different, we reexamine the polarizability and spin susceptibility in a one-dimensional
wire.
Before we look at the polarizability, we need to know the Coulomb interaction in one

1
This use of the word one dimensional is different from that of Chapter 6, where one dimensional
referred to a system that is much longer than it is wide, but that does not need to be as narrow as F . In
order to distinguish the two cases, the latter case is sometimes referred to as quasi one dimensional.

231
232 CHAPTER 11. ELECTRONS IN ONE DIMENSION

dimension. For a wire of thickness a in vacuum, the Coulomb interaction is

e2 eiqx
Z
Vq = dx
40 |x|
2
e 1
ln , (11.1)
20 qa
which is valid for qa  1. We have cut off the integral at x a, since for x . a a three-
dimensional form of the Coulomb interaction should be used. If interactions in the one-
dimensional wire are screened by a nearby metal at distance b, the interaction Vq saturates
at q 1/b.
In the random phase approximation, the polarizability reads
0R
e (q, )
e (q, )RPA = , (11.2)
1 Vq 0R
e (q, )/e
2

where Vq is given by Eq. (11.1) above and

e2 X tanh[(k+q )/2T ] tanh[(k )/2T ]
0R
e (q, ) = (11.3)
V k k k+q + + i

is the polarizability of the one-dimensional non-interacting electron gas. In one dimension,

the k summation in Eq. (11.3) is easily done: the momentum k is represented by a real
number k, and one has k+q = k + qvF sign k. Then, replacing the summation over k by an
integration over k , one has
e2
 
0R 2qvF 2qvF
e (q, ) =
2 + i qvF + i + qvF
2 2
2e q vF
= , (11.4)
[( + i)2 q 2 vF2 ]
where we used the fact that, in one dimension, the density of states is related to the Fermi
velocity vF as = 1/vF . Substituting this into Eq. (11.2), one finds
2e2 q 2 vF
e (q, )RPA = . (11.5)
( + i)2 q 2 vF2 (e2 /0 ) ln(1/qa)q 2 vF
The one-dimensional electron liquid can dissipate energy only if the imaginary part of
e (q, ) is nonzero. Inspection of Eq. (11.5) shows that that is the case for

= qvc , (11.6)
11.1. WHY IS ONE DIMENSION DIFFERENT? 233

where the velocity vc is

s
e2 1
vc = v F 1+ 2
ln . (11.7)
vF 0 qa
In the presence of a metal gate at distance b from the wire that screens the Coulomb inter-
action inside the wire, the momentum q in Eq. (11.7) should be replaced by 1/b.
The result (11.6) describes the one-dimensional version of plasma oscillations. Unlike in
three dimensions, where the plasma frequency remained finite (and large) if the wavevector
q 0, in one dimension 0 if q 0. However, the plasmon speed is larger than the
Fermi velocity, since the plasmon propagation speed is enhanced by the repulsive Coulomb
interactions. Hence, in one dimension, plasmons are relevant perturbations in the low fre-
quency, long wavelength limit.
However, this is not the entire story. The result (11.5) differs in one more aspect from its
three dimensional counterpart: There is no particle-hole continuum for excitations, not even
a single particle-hole branch! Note that at = qvF the polarizability of Eq. (11.5) is well
behaved and has no imaginary part. Hence, we conclude that the only charged excitations
in a one dimensional electron gas are the plasmon collective modes.
The situation is different if the wavevector q is of order kF , or if the temperature or
frequency are of order F . In that case, one cannot approximate the spectrum by a linear
dispersion. Without interactions, the polarizability function 0R e (q, ) becomes a truly com-
2 2
plex function of q and for qvF q /2m < < qvF + q /2m, just like in three dimensions,
and, hence, the full polarizability RPA
e becomes complex as well for that wavevector and
frequency range. The support of Im e for the entire frequency range is shown in Fig. 11.1.
If we want to investigate spin excitations, we should look at the spin susceptibility. Taking
the Hubbard-model result for the transverse susceptibility + (q, ),
0R
+ (q, )
R
+ (q, ) = , (11.8)
1 (2U/B g)0R+ (q, )

where
G g X tanh[(k+q )/2T ] tanh[(k )/2T ]
0R
+ (q, ) =
4V k k k+q + + i
B gq 2 vF
= (11.9)
2[( + i)2 q 2 vF2 ]
is the transverse susceptibility of the non-interacting electron gas, we find that the imaginary
part of + (q, ) is nonzero if and only if
= vs q, (11.10)
234 CHAPTER 11. ELECTRONS IN ONE DIMENSION

tvu
vu
tvtuvu tvu
tvu tvu
vu
tvu tvu
vu
tvu tvu
vu
tvu tvu
vu
tvu tvu
vu
tvu tvu
vu
tvu tvu
vu
tvu tvu
vu
tvu tvu
vu
tvu xwxw vu tvu
tvu xwxw vu tvu
tvu xwxw vu tvu
tvu xwxw vu tvu
tvu xwxw vu tvu
tvu xwxw vu tvu
tvu xwxw vu tvu
tvu xwxw vu tvu
tvu xwxw vu tvu
tvu xwxw vtvtxwxw
vtu t tvu
vu t tvu
vu t tvu
vu t tvu
vu t tvu
vu t tvu
vu t tvu
vu t tvu
vu t tvu
vu t tvu
vu w x xwxw particlehole t tvu
vu w x xwxw vu t tvu w x xwxw vu t tvu w x xwxw vu t tvu w x xwxw vu t tvu w x xwxw vu t tvu w x xwxw vu t tvu w x xwxw vu t tvu w x xwxw vu t tvu xwxw vtvtxwxw
4 vtu
t u
v t tvu
vu t tvu
vu t tvu
vu t tvu
vu t tvu
vu t tvu
vu t tvu
vu t tvu
vu t tvu
vu t tvu
vu xwxw vu t tvu xwxw vu t tvu xwxw vu t tvu xwxw vu t tvu xwxw vu t tvu xwxw vu t tvu xwxw vu t tvu xwxw vu t tvu xwxw vu t tvu xwxw vtvtxwxw
ttu
vu
vvtutu t vvu vplasmon t tu t tu
vvu t tu
vvu t tu
vvu t tu
vvu t tu
vvu t tu
vvu t tu
vvu t tu
vvu wxxw excitations t tu
vvu wxxw tu t vvu wxxw tu t vvu wxxw tu t vvu wxxw tu t vvu wxxw tu t vvu wxxw tu t vvu wxxw tu t vvu wxxw tu t vvu xwwx vttvxwwx
t t u
v t t t t t t t t t t t t t t t t t xwxw vtvtxwxw
/ F vtu vtu
tvtuvu
vu
tvu
vu
t tvu
tvu
vu
t tvu
vu
tvu
vu
t tvu
vu
tvu
vu
t tvu
vu
tvu
vu
t tvu
vu
tvu
vu
t tvu
vu
tvu
vu
t tvu
vu
tvu
vu
t tvu
vu
tvu
vu
t tvu
vu
tvu
vu
t tvu
vu
xwxw vu
xwxw vu
tvu
t tvu
xwxw vu
xwxw vu
tvu
t tvu
xwxw vu
xwxw vu
tvu
t tvu
xwxw vu
xwxw vu
tvu
t tvu
xwxw vu
xwxw vu
tvu
t tvu
xwxw vu
xwxw vu
tvu
t tvu
xwxw vu
xwxw vu
tvu
t tvu
xwxw vu
xwxw vu
tvu
t tvu
xwxw vu
xwxw vu
tvu
t tvu xwxw vtvtxwxw
vtu t tvu
vu modes t tvu
vu t tvu
vu t tvu
vu t tvu
vu t tvu
vu t tvu
vu t tvu
vu t tvu
vu t tvu
vu xwxw vu t tvu xwxw vu t tvu xwxw vu t tvu xwxw vu t tvu xwxw vu t tvu xwxw vu t tvu xwxw vu t tvu xwxw vu t tvu xwxw vu t tvu xwxw vtvtxwxw
vtu t t t t t t t t t t t t t t t t t t t xwwx vttvxwwx
2 tu tvtutu
vvu vvu
tvu
tvu
tu
vvu
t tvu
vu
tu
vvu
t tvu
vu
tu
vvu
t tvu
vu
tu
vvu
t tvu
vu
tu
vvu
t tvu
vu
tu
vvu
t tvu
vu
tu
vvu
t tvu
vu
tu
vvu
t tvu
vu
tu
vvu
t tvu
vu
wxxw tu
xwxw vu
vvu
t tvu
wxxw tu
xwxw modes
vvu
t tvu
vu
wxxw tu
xwxw vu
vvu
t tvu
wxxw tu
xwxw vu
vvu
t tvu
wxxw tu
xwxw vu
vvu
t tvu
wxxw tu
xwxw vu
vvu
t tvu
wxxw tu
xwxw vu
vvu
t tvu
wxxw tu
xwxw vu
vvu
t tvu
wxxw tu
xwxw vu
vvu
t tvu xwxw vtvtxwxw
vtu
vtu t tvu
vu t tvu
vu t tvu
vu t tvu
vu t tvu
vu t tvu
vu t tvu
vu t tvu
vu spin t tvu
vu t tvu
vu xwxw vu t tvu xwxw vu t tvu xwxw vu t tvu xwxw vu t tvu xwxw vu t tvu xwxw vu t tvu xwxw vu t tvu xwxw vu t tvu xwxw vu t tvu xwxw vtvtxwxw
vtu
tvtuvu
vu t tvu t tvu
vu t tvu
vu t tvu
vu t tvu
vu t tvu
vu t tvu
vu t tvu
vu t tvu
vu t tvu
vu xwxw vu t tvu xwxw vu t tvu xwxw vu t tvu xwxw vu t tvu xwxw vu t tvu xwxw vu t tvu xwxw vu t tvu xwxw vu t tvu xwxw vu t tvu xwxw vtvtxwxw
tu
vvtutu t vvu t tu
vvu t tu
vvu t tu
vvu t tu
vvu t tu
vvu t tu
vvu t tu
vvu t tu
vvu t tu
vvu wxxw tu t vvu wxxw tu t vvu wxxw tu t vvu wxxw tu t vvu wxxw tu t vvu wxxw tu t vvu wxxw tu t vvu wxxw tu t vvu wxxw tu t vvu xwwx vttvxwwx
0 vtuvu
t t vu t t vu t t vu t t vu t t vu t t vu t t vu t t vu t t vu t txw vu t txw vu t txw vu t txw vu t txw vu t txw vu t txw vu t txw vu t txw vu t txwxw vtvtxwxw
0 1 2 3 4
q/k F
Figure 11.1: Support of the polarizability e and the transverse spin susceptibility + for a
one-dimensional interacting electron liquid in the RPA approximation.

where p
vs = v F 1 U/vF . (11.11)
The spin excitations are collective excitations as well, moving at a speed that is below the
Fermi velocity. The reason why vs is smaller than vF is that the propagating of spin excita-
tions is slowed down by the ferromagnetic exchange interaction.2 At the Stoner instability,
U = 1, the propagation speed has come to zero, and large-scale ferromagnetic fluctuations
become possible. Note that, again, + is purely real at = vF q.
The phenomenon that, in general, vs is different from vc is known as spin-charge sepa-
ration. In a one-dimensional electron liquid, excitations with spin move at a different speed
than excitations with charge. In view of what we just discussed, this is not a big surprise.
Since, in one dimension, the only possible excitations are collective modes, there is no rea-
son to expect that spin modes and charge modes have the same velocity. You may recall
the calculations of the zero sound velocity in a Fermi liquid, where spin and density modes
depend on different Fermi-liquid constants and, hence, have different propagation velocities.
Quantitatively, our observations have been based on the random phase approximation.
Going beyond the random phase approximation will most certainly change our estimates
2
In three dimensions, the fact that vs < vF implies that spin excitations are strongly damped because
they can decay into particle-hole pairs (Landau damping). In one dimension, there are no particle-hole
excitations, so that spin excitations are long-lived in spite of the fact that their propagation speed is below
the Fermi velocity.
11.2. EFFECTIVE HAMILTONIAN 235

for the velocities vc and vs of collective charge and spin excitations in the one-dimensional
electron system, but not the qualitative conclusion that the collective modes are the only
long-wavelength excitations of the one-dimensional electron system. Well see this in Sec.
11.3, where we start from a microscopic model for electrons in one dimension and show
rigorously that all dynamics is collective. Another way to see this is to note the similarity
between electrons in one dimension and ions in a lattice. The similarity arises because, in
one dimension, electrons cannot pass each other. In that sense, a description in terms of a
solid would be more appropriate than a description in terms of a liquid.3 Indeed, all
long-wavelength dynamics of lattice ions is collective; it is only at short wavelengths that
the properties of individual ions become important.
We conclude that, for long wavelengths and low frequencies, all excitations of the one-
dimensional metal are collective modes; there are no dissipative modes at |/q| = vF , which
is the frequency-wavevector relation one expects for particle-hole excitations in a degenerate
one-dimensional Fermi gas. This observation makes us wonder whether a better description
than a description in terms of particles is possible. Can one make a theory that uses the
collective modes as the building blocks, rather than the individual fermions?
Another reason to look for a theory that does not start from a system of non-interacting
fermions is that, in one dimension, the quasiparticle lifetime scales inversely proportional to
the excitation energy of the quasiparticle. This is different from three dimensions, where the
life time is inversely proportional to the square of the excitation energy. In three dimensions,
this dependence was the basis for Fermi Liquid theory, the statement that a picture based
on non-interacting electrons is a good starting point. In one dimension, quasiparticle decay
is much faster, and Fermi Liquid theory is not valid. What description one uses instead
follows from the above considerations, as well see in the next sections.

11.2 Effective Hamiltonian

Based on the considerations of the previous section, we now write down an effective Hamil-
tonian that has the same (collective) excitations as the one-dimensional electron liquid. For
simplicity, we consider the case of spinless fermions, so that the only collective excitations are
the plasmon modes. Also, we assume that the long-range part of the Coulomb interaction
in the wire is screened by a nearby piece of metal, and, hence, take the plasmon velocity vc
3
The difference between a solid and a liquid has to do with the presence of transverse rigidity (shear).
In one dimension, there is no transverse direction, so that this formal difference between solid and liquid
disappears. We usually denote the one-dimensional electron system as a liquid, because of the lack of
long-range order. However, as we see from the present discussion, one-dimensional electron systems have
characteristics of both solids and liquids.
236 CHAPTER 11. ELECTRONS IN ONE DIMENSION

to be independent of q.
With these observations, we are tempted to write down the following Hamiltonian as an
effective Hamiltonian for the one-dimensional electron liquid,
X
H= q (aq aq + 1/2), (11.12)
q

where the summation over the wavenumber q extends over both positive and negative q,
with q = vc |q|, and where aq and aq are boson creation and annihilation operators that
obey commutation relations,
h i h i

aq , aq 0 = aq , aq 0 = 0
h i

aq , aq 0 = q,q0 .

As we discussed above, the Hamiltonian (11.12) is valid for long wavelengths and low fre-
quencies only. At high wavenumbers |q| kF , particle-hole excitations become important,
and our description in terms of collective modes only ceases to be valid.
Although there is nothing wrong with the Hamiltonian (11.12) as an effective Hamiltonian
for the spinless one-dimensional electron liquid, we can gain considerably more insight if we
replace the creation and annihilation operators a and a by momentum and displacement
operators and , where
s
~
q = (a + aq ),
2Kq q
r
K~q  
q = iaq iaq . (11.13)
2
We will use the freedom of the arbitrary constant K later in order to give a physical inter-
pretation of the operators and . You verify that these momentum and displacement
operators satisfy the usual commutation rules for canonically conjugate variables,

[q , q0 ] = [q , q0 ] = 0
[q , q0 ] = i~qq0 . (11.14)

In terms of the new variables and we write the Hamiltonian as

X 1 1

2
H= q q + Kq q q . (11.15)
q
2K 2
11.2. EFFECTIVE HAMILTONIAN 237

Next, we perform a Fourier transform to real space,

r
1X
(x) = q eiqx ,
L q
r
1X
(x) = q eiqx . (11.16)
L q

Since the effective Hamiltonian (11.12) is valid for low q only, the summation over q should
be restricted to q  kF only. This means that delta functions and divergences appearing
in the real-space formulation should be cut off at distances x 1/kF . In the real-space
formulation, the effective Hamiltonian becomes
 
1 1
Z
2 2 2
H = dx (x) + Kvc (x (x)) . (11.17)
2K 2

The commutation relations of the fields (x) and (x) are

[(x), (x0 )] = [(x), (x0 )] = 0

[(x), (x0 )] = i~(x x0 ). (11.18)

Let us now calculate time-derivative of (x),

i 1
(x) = [H, (x)] = (x). (11.19)
t ~ K
We want the operators and to be related to the electron current density and particle
density, respectively. If such an identification is to hold, Eq. (11.19) should represent the
continuity equation t n = x j. Taking a derivative to x on both sides of the equation, we
see that such an identification holds, if we identify the (excess) particle density n(x) with
the x-derivative of (x) and the current with (x). In fact, the following identifications are
made,

1
n(x) = x (x), (11.20)

j(x) = vF (x), (11.21)

which implies K = 1/vF . However, different identifications are possible, depending on the
microscopic details of the one-dimensional electron liquid.
238 CHAPTER 11. ELECTRONS IN ONE DIMENSION

Combining everything, we see that we have arrived at the effective Hamiltonian

" 2 #
2

(x)
Z
~vF 2 2
H= dx 2 (x) + g , (11.22)
2 ~ x

where we introduced the dimensionless parameter

vc
g= . (11.23)
vF
You may recognize Eq. (11.22) as the Hamiltonian of an elastic string, where is the displace-
ment and is the momentum density. The Hamiltonian (11.22) is quite different from the
Hamiltonian you would write down for non-interacting fermions in one dimension. It reflects
the fact that all excitations are collective and, hence, bosonic, rather that quasi-particle like.
In order to stress the difference of the collective dynamics of the one-dimensional electron
liquid and the quasi-particle dynamics of the Fermi liquid in higher dimensions, the former
case is referred to as Luttinger Liquid.
The Hamiltonian (11.22) is a great starting point if one wants to study the excitation
spectrum and thermodynamic quantities of a one-dimensional spinless electron liquid. One
example is the calculation of the compressibility , which is the derivative of the particle
density to the chemical potential = E/n, where E is the energy of the electron liquid,
 2 1
n E
= = . (11.24)
n2
According to the effective Hamiltonian (11.22), one has = vF /vc2 . Without interactions,
when vc = vF , this simplifies to = 1/vF = , being the density of states. With
interactions, is decreased by a factor g 2 = (vc /vF )2 , reflecting the increased energy cost for
addition of charge. Similarly, you verify that the specific heat is decreased by a factor 1/g 2
with respect to the non-interacting electron case.
For some applications it is important to be able to make reference to individual electrons.
An example is a tunneling experiment, where electrons tunnel into a one-dimensional electron
liquid from a weakly coupled electrode. Since we employ a long-wavelength description, we
will not attempt to describe a process in which an electron that is created or annihilated
is localized within a wire segment of length comparable to the Fermi wavelength. Instead,
well want to write down an operator that creates an electron that is delocalized over a piece
of wire of length F . This spatial smearing amounts to a momentum cut off factor
exp(|q|/2) in Eq. (11.16) or, equivalently, replacing the fields and by
2
Z
(x) dx0 2 0 2
(x0 ),
[ + 4(x x ) ]
11.2. EFFECTIVE HAMILTONIAN 239

2
Z
(x) dx0 (x0 ). (11.25)
[2 0 2
+ 4(x x ) ]
On the other hand, since we give up spatial resolution on length scales below F , we can
specify the momentum of the electron to within kF , i.e., we can specify whether the electron
moves left (L) or right (R). The creation operator of a right moving electron at position x
then becomes4
1
R (x) = UR eikF x+i(x)i(x) . (11.26)
2
Similarly, for a left-moving particle one has
1
L (x) = UL eikF xi(x)i(x) . (11.27)
2
Here the auxiliary field (x) is defined as

Z
(x) = dx0 (x0 )sign (x x0 ). (11.28)
2~
You verify that commutes with itself, whereas the commutator with is given by

[(x), (x0 )] = [(x), (x0 )] = i sign (x x0 ). (11.29)
2
In these equations, the fast exponential factor exp(ikF x) corresponds to the orbital phase
of left and right moving electrons. The operator U is known as the Klein factor. It is
a formal operator that increases the total number of left or right moving particles by one.
It is necessary, because the operators (x) and (x) conserve the total number of particles.
The operator exp[i(x)] produces a shift of the field (x0 ) by /2 at x0 < x and by /2
at x0 > x. Hence, the derivative x is peaked near x, corresponding to an excess particle
density at that point, see Fig. 11.2.5 The operator ei(x) changes sign each time is increased
by , i.e., each time a particle passes through the point x. This property enforces the fermion

anticommutation rules for the -operators. Furthermore, ei(x) displaces the field (x) by
one, corresponding to a current density vF (x). Finally, the prefactor (2)1/2 is chosen
in conjunction with the regularization (smearing) procedure for the fields and . It is
chosen such that a calculation of the fermion Green function for non-interacting electrons
gives the same result in boson and in fermion language.
4
In the next section, where we give a more formal derivation of the Hamiltonian (11.22), you can find
more details of the formal relationship between the fields (x) and (x) and the underlying electron creation
and annihilation operators and .
5
This statement does not conflict with the fact that the operator conserves particle number. At the
upper and lower ends of the wire there is an antikink in , corresponding to a decrease of the particle
number by one (in total). It is the operator U that keeps track of the total particle number.
240 CHAPTER 11. ELECTRONS IN ONE DIMENSION

x x
Figure 11.2: Upon creation of an electron at position x, the field acquires a kink, corresponding
to a peak in the excess density .

The information of the boson fields and is contained in their Green functions. The
calculation of these Green functions is quite similar to that of the phonon Green functions.
For technical reasons, we calculate Green functions for the fields and . In terms of the
fields and , the Hamiltonian reads
" 2  2 #
(x) (x)
Z
~vF
H= dx + g2 . (11.30)
2 x x
In order to calculate the Green functions of the fields and , we use the equation of motion
approach. Hereto we need the imaginary time evolution,
(x, )
(x, ) = ivF g 2 ,
x
(x, )
(x, ) = ivF . (11.31)
x
We define the temperature Green functions
D (x, ) = hT (x, )(0, 0)i,
D (x, ) = hT (x, )(0, 0)i,
D (x, ) = hT (x, )(0, 0)i,
D (x, ) = hT (x, )(0, 0)i. (11.32)
Using Eq. (11.31) we find that these Green functions satisfy the equations of motion
D (x, )
D (x, ) = ivF g 2 ,
x
11.2. EFFECTIVE HAMILTONIAN 241

D (x, )
D (x, ) = i ( ) sign (x) ivF g 2 ,
2 x
D (x, )
D (x, ) = i ( ) sign (x) ivF ,
2 x
D (x, )
D (x, ) = ivF . (11.33)
x
The solution to these equations is easily found by inspection,

D (x, ) = D (x, )
1 1
= ln sin[T ( + ix/vc )sign ( )] ln sin[T ( ix/vc )sign ( )],
4 4
vF vc
D (x, ) = D (x, )
vc vF
1 1
= ln sin[T ( + ix/vc )sign ( )] + ln sin[T ( ix/vc )sign ( )].
(11.34)
4 4
Regularization amounts to smearing of the boson fields by a Lorentzian factor 2/[( 2 +
4(x x0 )2 ], see Eq. (11.25) above. In the Green functions such a smearing amounts to the
substitution x xisign ( ). This corresponds to the addition of T /vF to the argument
of the sine function, so that the divergence of the Green functions at x 0 and 0 is
cut off at distances of order and times of order /vF .
As an example, let us now calculate a fermion Green function using the boson language.
We consider the temperature Green function for right-moving electrons,

GRR (x, ) = hT R (x, )R (0, 0)i. (11.35)

Using Eq. (11.26), we write this as

eikF x
hT UR ( )UR (0)i T ei(x, )+i(x, )+iR (0,0)i(0,0) .


GRR (x, ) = (11.36)
2
Since the total Green function needs to be antiperiodic in , whereas the boson part of the
Green function is manifestly periodic in , we require that the time-ordering for the Klein
factors UR and UR is that of fermions. Since the operators UR and UR are simply ladder
operators, we have
hT UR ( )UR (0)i = sign ( ). (11.37)
Calculating the boson part of the Green function is easier than it seems at first sight.
Since the Hamiltonian of the boson fields is quadratic, we can use the cumulant expansion,
242 CHAPTER 11. ELECTRONS IN ONE DIMENSION

which, for h(x, )i = h(x, )i = 0, reads

i(x, )+i(x, )+i(0,0)i(0,0)  1 2
T e = e 2 h(i(0,0)i(0,0)i(x, )+i(x, )) i
= eD (0,0)D (0,0)D (0,0)+D (0,0)
eD (x, )+D (x, )+D (x, )D (x, ) . (11.38)

Without regularization, the equal-time and equal-position Green functions D (0, 0)+D (0, 0)
are divergent (and negative). With regularization, this divergence is cut off at distances
(see above). Substituting Eq. (11.34) and putting everything together, we find
 /21/2
1 T /2vc
GRR (x, ) = sign ( )
(2)1 sin[T ( + ix/vc )sign ( )]
 /2+1/2
T /2vc
, (11.39)
sin[T ( ix/vc )sign ( )]
where we abbreviated    
1 1 1 vc vF
= g+ = + . (11.40)
2 g 2 vF vc
It is instructive to compare this result to the electron Green function calculated in the
fermion language,
T /2vF
GRR (x, ) = . (11.41)
sin[T ( ix/vF )]
Taking the result obtained in the boson language, and setting g = 1, we recover the fermion
Green function. This argument, a posteriori, fixes the cut-off dependent prefactor (2)1/2
in the relations (11.26) and (11.27) between the fermion creation/annihilation operators and
the boson fields and .
There is an important difference between the single-electron Green function for g = 1 and
for g 6= 1. In the absence of interactions, the single-electron Green function has a simple pole
at i = x/vF . With interactions, the location of the singularity shifts, and the singularity
acquires a different analytical structure. This may be brought to light by a calculation of the
retarded Green function, i.e., by performing a Fourier transform to Matsubara frequencies
followed by analytical continuation in +i. For simplicity we look at the equal-position
Green function,
Z 1/T
GRR (0, in ) = GRR (0, )ein , (11.42)
0

which is related to the spectral density ARR for right-moving electrons. We deform the -
11.2. EFFECTIVE HAMILTONIAN 243

0 1/T

Figure 11.3: Integration contour for the calculation of the density of states in a Luttinger liquid.

integration as shown in Fig. 11.3, avoiding the branch cuts for Re = 0 and Re = 1/T .
After the deformation of the contours, the integral reads
Z  
2i n t T /2vF
GRR (0, in ) = dte Re
(2)1 0 sin[T (it + /vF )]
Z  
2i n t i/2 T /2vF
= dte Re e . (11.43)
(2)1 0 sinh[T (t i/vF )]

Now we can take the analytical continuation in + i and calculate the retarded Green
function,
Z  
R 2i it i/2 T /2vF
GRR (0, ) = dte Re e .
(2)1 0 sinh[T (t i/vF )]
(11.44)

Evaluation of the remaining integral depends on whether the energy is small or large in
comparison to T . If  T , we find
 1
2i T
GR
RR (0, ) . (11.45)
2vF vF
244 CHAPTER 11. ELECTRONS IN ONE DIMENSION

(The numerical prefactor is valid for close to unity.) You verify that one obtains GR
RR (0, ) =
i/2vF for the non-interacting case = 1, which implies ARR = 1/vF , the result that we
expected for the spectral density of right-moving electrons. If, on the other hand, T  ,
one finds a result similar to Eq. (11.45) with T replaced by .
We conclude that for an interacting one-dimensional electron liquid the density of states
at the Fermi level has a power-law singularity, and vanishes proportional to max(, T )1 .
This behavior is quite different from the case of a Fermi liquid, where the density of states is
non-singular at the Fermi level. Also note that the cut-off length enters into the expression
for the density of states if 6= 1. For a quantitative estimate, one should replace by F .

11.3 Luttingers model

11.3.1 Formulation of the model
Although the discussion of the previous section provided the physical motivation of the ef-
fective Hamiltonian (11.22), it might leave some uneasiness with those of you who want a
constructive description of a field theory of interacting electrons in one dimension. There-
fore, we now return to a description of a one-dimensional electron liquid in fermion language,
and derive the effective Hamiltonian (11.22) using formal manipulations. The discussion of
this section follows that of F. D. M. Haldane, J. Phys. C 14, 2585 (1981), although it is
less rigorous at points. Those readers who want a fully rigorous treatment are referred to
Haldanes article, or to the tutorial review by J. von Delft and H. Schoeller, Annalen Phys.
7, 225 (1998).
As a starting point, we use a description of the one-dimensional electron system in which
we have linearized the electron spectrum, see Fig. 11.4. We shift momenta by an amount
kF (kF ) for right (left) moving electrons, so that the Fermi points are shifted to k = 0,
and denote the creation/annihilation operators for right moving (left moving) electrons with
shifted momentum k by ckR and ckR (ckL and ckL ), respectively. We then write the kinetic
energy as
kin = vF
X
H (k : nkR : k : nkL :), (11.46)
k

where nkR = ckR ckR is the density of right-moving electrons and nkL = ckL ckL is the density
of left-moving electrons. The symbol : . . . : refers to normal ordering, i.e., to subtraction
of the densities in the (non-interacting) ground state. The summations over k should be
truncated at |k| . kF because the linear dispersion is valid for a window of wavevectors
of size . kF around k = 0 only and because the distinction between left movers and right
movers implies that the (translated) wavevector of a right mover cannot be smaller than
11.3. LUTTINGERS MODEL 245

left right
movers movers

kF kF k k

= F vF (kkF ) = F+ vF (kkF ) k=0 for k=0 for

left movers right movers

Figure 11.4: Linearized spectrum for electrons in one dimension. In our description, we separate
right and left moving electrons and shift momenta such that the Fermi points correspond to k = 0.

kF , whereas the (translated) wavevector of a left mover cannot be larger than k F , see Fig.
11.4. In terms of the fields
1 X ikx 1 X ikx
R (x) = e ckR , L (x) = e ckL , (11.47)
L k L k

the Hamiltonian (11.46) reads6

Z h i
kin = vF dx : (x)(ix )L (x) : : (x)(ix )R (x) : .
H (11.48)
L R

11.3.2 Bosonization
Well now show that H kin can be written in terms of the densities qR and qL of right and
left moving electrons,
X X
qR = ck,R ck+q,R , qL = ck,L ck+q,L . (11.49)
k k

The normal ordering in Eq. (11.48) means that has to be replaced by for states with momen-
6

tum smaller than kF . This prescription is not transparent in a real space formulation.
246 CHAPTER 11. ELECTRONS IN ONE DIMENSION

We consider the case q = 0 separately; note that we do not need to use normal ordering for
q 6= 0. The commutation relations of the density operators are
qL
[q,R , q0 ,R ] = qq0 ,
2
qL
[q,L , q0 ,L ] = qq0 ,
2
[q,R , q0 ,L ] = 0. (11.50)

To see how this result is obtained, let us look at the first line of Eq. (11.50) in detail,
Xh i
[q,R , q0 ,R ] = ck,R ck+q,R , ck0 ,R ck0 q0 ,R

k,k 0
X 
= ck,R ck+qq0 ,R ck+q0 ,R ck+q,R . (11.51)
k

If q 6= q 0 , this is zero, as one can see from relabeling the summation index k. Otherwise, if
q = q 0 we find X
[q,R , q0 ,R ] = (nk,R nk+q,R ) . (11.52)
k

The summation on the right hand side of Eq. (11.52) might appear ambiguous. It can
be computed unambiguously by normal ordering of the summand. Normal ordering of the
operators between brackets gives
qL X
[q,R , q,R ] = : nk,R nk+q,R :
2 k
qL
= . (11.53)
2
The summation of normal ordered terms gives zero after relabeling of the summation index
k. We had to normal order first, since relabeling the summation index is allowed for normal-
ordered summands only.
With these commutation relations, the density operators can serve as boson creation and
annihilation operators. Operators qR with q < 0 play the role of creation operators, whereas
operators qR with q > 0 are annihilation operators. Similarly, for left-moving particles, the
creation operators are the qL with q > 0 and annihilation operators are qL with q < 0.7
When regarded as creation and annihilation operators, the density operators span the entire
7
Formally, this identification requires that we multiply the density operators by (L|q|/2) 1/2 .
11.3. LUTTINGERS MODEL 247

Hilbert space with fixed numbers NR and NL of left moving and right moving electrons.
You can verify this statement by explicit construction of the states, or by comparing the
partition functions for the electron liquid in fermion and boson representations, see footnote
8 below. Since the density operators span the entire Hilbert space, the kinetic energy can be
represented in terms of the density operators, rather than the fermion operators. In order
to achieve this, we look at the time derivative of q,R and q,L , for which we find
i X h
i
q,R = vF k ck,R ck,R , ck0 ,R ck0 +q,R
t ~ k,k0

i X 


= vF k ck,R ck+q,R ckq,R ck,R
~ k
iqvF
= q,R . (11.54)
~
Similarly, for the left-moving particles we get
iqvF
q,L = q,L . (11.55)
t ~
Utilizing the commutation relations of the density operators, we find that we can write the
kinetic energy in terms of the boson operators q,R and q,L ,

kin = 2vF vF
X
H (q,R q,R + q,L q,L ) + (NR2 + NL2 ), (11.56)
L q>0 L

where the additive constant reflects the additional cost of adding electrons to the system. 8
8
Now we are in a position to prove that the set of states that is spanned by the density operators is
complete. Calculating the grand canonical partition function for the fermions, we find
" #2
Y
Z= (1 + w2n1 )2 ,
n=1

where we abbreviated w = exp(vF /T L) and put the chemical potential precisely between two successive
energy levels. The square within the square brackets corresponds to states with different signs of k, i.e.,
to particle and hole-like excitations, whereas the overall square reflects the identical contributions from left
movers and right movers. For the boson basis, we find

!2
m2
Y X
2n 2
Z= (1 w ) w .
n=1 m=

Here, the first factor is the partition function corresponding to the boson degrees of freedom, whereas the
second factor gives the contribution of the total number of particles to the energy. You verify that these two
partition functions are, indeed, equal.
248 CHAPTER 11. ELECTRONS IN ONE DIMENSION

Another way to obtain Eq. (11.56) is to note that, by Eq. (11.49), the operator q creates
a multitude of electron-hole pairs, all with the same energy vF q. Hence, the energy of the
boson created by q is vF q. Taking into account the normalization factor (2/qL)1/2 from
the commutation relations (11.50), one arrives at Eq. (11.56).
The interaction Hamiltonian is, in fact, easier to deal with. Since it contains four fermion
creation or annihilation operators, it is quadratic in the boson operators q,R and q,L . Using
the observation that the interaction couples to the total density only, one arrives at the
Hamiltonian
int = 1 (V0 V2k )
X
H F
(q,R + q,L )(q,R + q,L ). (11.57)
2L q6=0

The prefactor (V0 V2kF ) follows from the difference of Hartree and Fock type interactions,
assuming that the interaction Vq depends only weakly on q on scales  kF . For a point-like
interaction, one has V2kF = V0 , and H int vanishes, as is required by the Pauli principle.
The above manipulations, in which the fermion operators in the Hamiltonian are replaced
by boson operators, are known as bosonization.
The total Hamiltonian H = H kin + H int can be rewritten in terms of boson fields (x) and
(x) that depend on the coordinate x only. These fields are related to the particle density
and current density as in Eqs. (11.20) and (11.21). In terms of the density operators, the
field (x) reads
Z x
(x) = dx0 : (x) :, (11.58)

where (x) is the total electron density,

1 X iqx
(x) = e (q,R + q,L )e|q|/2 , (11.59)
L q6=0

whereas the field (x) is given by

X
(x) = eiqx (q,R q,L) e|q|/2 . (11.60)
q6=0

(The exponential high-momentum cutoff is the same as in the previous section.) Using
the commutation relations between the density operators q,R and q,L , you verify that the
fields (x) and (x) obey canonical boson commutation relations, cf. Eq. (11.18). With this
change of variables, the Hamiltonian H = H kin + H int acquires the form (11.22), with
r
V0 V2kF
vc = v F 1 + . (11.61)
vF
11.3. LUTTINGERS MODEL 249

Instead of a formulation in terms of the fields (x) and (x), which refer to total electron
density and current density (after normal ordering), one often uses fields R and L that are
related to the densities of right and left moving electrons,
1 X iqx 1 X iqx
R (x) = e q,R e|q|/2 , L (x) = e q,L e|q|/2 , (11.62)
L q6=0 L q6=0

as
Z x Z x
0 0
R (x) = 2 dx : R (x ) :, L (x) = 2 dx0 : L (x0 ) : . (11.63)

These fields have commutation rules

[R (x), R (x0 )] = isign (x x0 ),

[L (x), L (x0 )] = isign (x x0 ), (11.64)
[R (x), L (x0 )] = 0.

They are related to the original fields (x) and (x) as

Z
R (x) = (x) dx0 (x0 )sign (x x0 ), (11.65)
2~

Z
L (x) = (x) + dx0 (x0 )sign (x x0 ). (11.66)
2~

In terms of the fields R (x) and L , the Hamiltonian reads

" 2  2 #
vF R L
Z
H = dx +
4 x x
2
V0 V2kF

R L
Z
+ dx + . (11.67)
2 x x

11.3.3 Fermion creation and annihilation operators

Having formally rewritten the electron Hamiltonian in terms of boson fields (x) and (x),
we still want to be able to express the creation and annihilation operators of a single electron
in terms of the boson fields. Before we do this, let us recall the form of the Hilbert space in
the boson representation. In the boson representation, any electronic state can be written
as density operators q,R and q,L acting on the ground state with NR right moving electrons
250 CHAPTER 11. ELECTRONS IN ONE DIMENSION

and NL left moving electrons. Hence, every electronic state is represented by occupation
numbers nq,L with q > 0 and nq,R with q < 0 for the boson modes, and the numbers NR and
NL ,
|statei = |{nq }, NR , NL i. (11.68)
What is the action of a single-fermion creation operator L (x) or R (x) in this Hilbert space?
First of all, R (x) (L (x)) increases NR (NL ) by one. Inside the boson Hilbert space, this is
achieved by ladder operators UR and UL ,

UR |{nq }, NR , NL i = |{nq }, NR + 1, NL i,
UR |{nq }, NR , NL i = |{nq }, NR 1, NL i,
UL |{nq }, NR , NL i = |{nq }, NR , NL + 1i,
UL |{nq }, NR , NL i = |{nq }, NR , NL 1i. (11.69)

In addition to this, the operator R (x) creates a multitude of electron-hole pairs, so that,
in the end, the change in particle and current densities is sharply peaked around x. In order
to capture these particle hole pairs in terms of the boson fields, the notation that uses the
separate fields R (x) and L (x) for right and left moving electrons is particularly useful.
Returning to the qualitative arguments of the previous section the addition of an electron
corresponds to a kink in the fields R (x) and L (x) , we now make an educated guess
for R (x) and L (x),

1 1
R (x) = UR eiR (x) , L (x) = UL eiL (x) , (11.70)
2 2

where the cut-off dependent prefactor (2)1/2 has already been inserted. We verify that
this gives, indeed, the correct commutator with the density fields R (x) and L (x),
 
h
0
i 1 iR (x) 0
R (x), R (x ) = UR e , 0 R (x )
2 2 x
1
= UR eiR (x) 0 sign (x x0 )
2 2 x

= R (x)(x x ), 0
h i
L (x), L (x0 ) = L (x)(x x0 ). (11.71)

In deriving this result, we used the fact that the commutator [eX , Y ] = eX [X, Y ] if [X, Y ]
is proportional to the identity operator. Similarly, you verify that the R (x) and R (x0 )
11.3. LUTTINGERS MODEL 251

anticommute, as well as L (x) and L (x0 ). There is a problem, however, with the calculation
of the anticommutator of R (x) and R (x0 ), which becomes ambiguous when x x0 . The
ambiguity is lifted once the regularizing momentum cut off in Eqs. (11.59) and (11.60) or Eq.
(11.62) is taken into account. In coordinate representation, this momentum cut-off amounts
to the replacement of the boson fields R (x) and L (x) in the exponents in Eq. (11.70) by
the smeared fields

2R (x0 )
Z
R (x) dx0 ,
[2 + 4(x x0 )2 ]
2L (x0 )
Z
L (x) dx0 , (11.72)
[2 + 4(x x0 )2 ]

where F is the length scale over which the electrons are delocalized. This way, the
kink in the fields R (x) and L (x) that is created by the operators R (x) and L (x) is
smeared out over a length , and, hence, the added electron is delocalized over a segment of
length .9 For the regularized fields, the sign-function in the commutator is smeared over
a distance ,

[R (x), R (x0 )] = 2i arctan[(x x0 )/],

[L (x), L (x0 )] = 2i arctan[(x x0 )/]. (11.73)

With this smearing of the boson fields, it can be shown that the ansatz (11.70) provides the
correct anticommutation relation for the fermion creation and annihilation operator. For
details, see the original paper F. D. M. Haldane, J. Phys. C 14, 2585 (1981) or the tutorial
review J. von Delft and H. Schoeller, Annalen Phys. 7, 225 (1998). Writing the fields L (x)
nd R (x) in terms of the original fields (x) and (x) and undoing the momentum shift
q q kF we performed at the beginning of this section, you recover Eqs. (11.26) and
(11.27) of the previous section.
Finally, we have to ensure that operators for left moving and right moving fermions
anticommute. There are various ways to achieve this. The simplest method is to postulate
that the ladder operators UR and UL anticommute. An alternative way is to postulate that
the commutator between the left-moving and right-moving boson fields L and R is not
zero, but i instead. You are referred to the specialized literature for details.

9
We have chosen a regularization integral in the exponent of Eq. (11.70), instead of a regularization
integral before the exponent. The latter choice would amount a superposition of sharply localized electrons,
which does not need to be the same as a delocalized electron.
252 CHAPTER 11. ELECTRONS IN ONE DIMENSION

11.4 Electrons with spin

For electrons with spin, one can take the same route as we took in Sec. 11.2, and write down
an effective Hamiltonian based on the spin and charge velocities in the one-dimensional
electron liquid. Introducing fields c and c that are related to the particle current density
and particle density as
1
nc (x) = n (x) + n (x) = x c (x) 2 (11.74)

jc (x) = j (x) + j (x) = vF c (x) 2, (11.75)

together with fields s and s that are related to the spin current density and spin density,
1
ns (x) = n (x) n (x) = x s (x) 2 (11.76)

js (x) = j (x) j (x) = vF s (x) 2, (11.77)

we can write the Hamiltonian as

 2

Z
~vF
dx 2 c (x)2 + s (x)2


H =
2 ~
+ gc2 (x c (x))2 + gs2 (x s (x))2 ,

(11.78)

where gc = vc /vF and gs = vs /vF . The boson fields c,s and c,s are normalized such that
they satisfy canonical commutation relations,

[c,s (x), c,s (x0 )] = [c,s (x), c,s (x0 )] = 0

[c (x), c (x0 )] = [s (x), s (x0 )] = i~(x x0 ), (11.79)
[c (x), s (x0 )] = [s (x), c (x0 )] = 0.

The form of the fermion operators is found in the same way as for the spinless case,
1
R

(x) =
UR eikF xiR (x) ,
2
1
L

(x) = ikF x+iL (x)
UL e , (11.80)
2
where =, and where the fields L and R are defined as

1
R (x) = c (x) + s (x)
2
11.5. DOES A ONE-DIMENSIONAL ELECTRON LIQUID EXIST? 253


Z
0 0 0 0
dx [c (x ) + s (x )]sign (x x ) ,
2~

1
L (x) = c (x) + s (x)
2


Z
0 0 0 0
+ dx [c (x ) + s (x )]sign (x x ) .
2~
(11.81)

Alternatively, one can find a description starting from a microscopic picture, as in the
previous section. In this case, bosonization of the kinetic energy is done as before. For this,
we refer to the references cited above for more details.
One important aspect of the one-dimensional electron liquid with spin, is that spin exci-
tations and charge excitations travel at different speeds, see our discussion in Sec. 11.1. We
again emphasize that this phenomenon, which is known as spin-charge separation, also ex-
ists for collective modes in higher dimensions. The difference between the higher-dimensional
case and the one-dimensional case is that in higher dimensions there are quasiparticle exci-
tations that carry both spin and charge, in addition to the spin-charge separated collective
excitations. These quasiparticles dominate the low-energy physics of higher dimensional
electron liquids, so that, in the end, spin and charge always travel together. In one dimen-
sion, there are no quasiparticle excitations, and the spin-charge separated collective modes
dominate the low energy physics.
To illustrate the phenomenon of spin-charge separation, let us create a right-moving
electron with spin up at position x = 0 at time t = 0. As we discussed above, creating
an electron corresponds to making a kink of size in the field R around x = 0. The
time-dependence of the charge density and the spin density then follows from decomposing
R in terms of the spin and charge fields, and solving for their time-evolution using the
Hamiltonian (11.78). A kink in R corresponds to kinks in both c and s , but these two
kinks travel at different velocities. Hence, although initially charge and spin were located at
the same point, they are separated after a finite time. Only for a non-interacting electron
gas, for which vc = vs = vF , spin and charge are not separated.

11.5 Does a one-dimensional electron liquid exist?

Now that we have seen that electrons in one dimension behave quite differently from electrons
in three dimensions, it is natural to ask whether all of this was an academic exercise, or
whether one-dimensional electron systems exist in nature. Fortunately, the answer is positive.
254 CHAPTER 11. ELECTRONS IN ONE DIMENSION

One example of a one-dimensional electron liquid is realized in extremely thin and

clean wires formed in the two-dimensional electron gas that exists at the interface in a
GaAs/GaAlAs heterostructure. This is the example that comes closest to the physics we
have discussed in this chapter. Spin-charge separation in these thin wires has been observed
directly, in the form of interference between charge and spin excitations in a wire of finite
length. You can find more about the experiment and the theory in Y. Tserkovnyak, B. I.
Halperin, O. M. Auslaender, and A. Yacoby, cond-mat/0302274 and references therein.
Another example is that of a carbon nanotube. Nanotubes are very good candidates to
observe Luttinger Liquid behavior, since they are thin and clean by virtue of their chemistry.
(The wires, on the other hand, have to be carefully engineered.) However, carbon nanotubes
have two propagating modes at the Fermi level (per spin direction), which slightly modi-
fies some of the properties. The anomalous density of states has been observed in carbon
nanotubes, see, e.g., M. Bockrath et al., Nature 397, 598 (1999).
The third example is that of edge states in the quantum Hall effect. The main difference
between edge states and the one-dimensional electron liquid we considered here is that edge
states are chiral: all electrons move in one direction. For more details, you are referred to
the original papers by Wen [Phys. Rev. B 41, 12838 (1990); 43, 11025 (1991)].
In addition to the question of real existence, one-dimensional electron systems have been
used as a model for quite a variety of problems in condensed matter physics. One example
is the Kondo problem with a point-like magnetic impurity, where it can be argued that the
impurity interacts only with s-wave states in the metal, which, in turn, can be modeled as
a one-dimensional electron gas. Another example is that of a Coulomb blockaded quantum
dot, where the electron gas in the point contacts between the dot and the bulk electrodes
is modeled as a one-dimensional electron gas. In these applications, most of the electron
liquid is non-interacting, the role of the impurity or of the quantum dot being to introduce
a localized interaction in the Luttinger Liquid.