Lecture 7: Jellium model for electrons in a solid: Part I

Christopher Mudry
Paul Scherrer Institut, CH-5232 Villigen PSI, Switzerland.
(Dated: April 12, 2010)

Abstract
The jellium model for a three-dimensional Coulomb gas is defined. Second quantization of
fermions is presented, followed by the path-integral representation of the grand canonical partition
function. Collective degrees of freedom are introduced through a Hubbard-Stratonovich transformation. The low energy and long wave-length limit of the effective theory for the collective degrees
of freedom that results from integration over the fermions is derived within the random phase
approximation (RPA).

Electronic address: Christopher.Mudry@psi.ch; URL: http://people.web.psi.ch/mudry

Contents

I. Introduction
II. Definition of the Coulomb gas in the Schr
odinger picture

A. Three-dimensional classical Coulomb gas

B. Three-dimensional quantum Coulomb gas

III. Path integral representation of the Coulomb gas

IV. The random phase approximation

11

A. Hubbard-Stratonovich transformation

11

B. Integration of the electrons

12

C. Gaussian expansion of the fermionic determinant

13

A. Grassmann coherent states

16

B. Path integral representation for fermions

20
22

References

I.

INTRODUCTION

The so-called jellium model is a very naive model for interacting electrons hosted in a
three-dimensional solid. Electronic interactions are taken to be of the Coulomb type. The
ions making up the solid are treated in the simplest possible way, namely as an inert positive
background of charges that insures overall charge neutrality. In spite of its simplicity the
jellium model is very instructive as it displays interesting many-body effects. We shall see
that:
Screening of the Coulomb interaction takes place.
There are Friedel oscillations.
There are collective oscillations called plasmons.
The method that I will employ to derive these phenomena is very general and powerful
although it is not elementary. The idea is to introduce a collective degree of freedom for
which a low-energy and long wave-length effective theory is derived. The effective theory
cannot be obtained exactly. Today, the effective theory will be derived within the random
phase approximation (RPA). Its analysis will be left for next weeks lecture.

II.

DEFINITION OF THE COULOMB GAS IN THE SCHRODINGER

PICTURE
A.

Three-dimensional classical Coulomb gas

The definition of the three-dimensional classical jellium model is given by the Hamiltonian
N
X
X
p2i
H :=
+
e2 Vcb (ri rj )
2m
i=1
1i<jN

(2.1a)

that describes N identical electrons of charge e and mass m. The electronic interaction is
mediated by the electrostatic potential Vcb (r) that satisfies Poisson equation (~ = 1 and
is Laplace operator in d = 3 space dimensions)


1
Vcb (r) = 4 (r)
.
V

(2.1b)

All N electrons are confined to a box of linear size L and volume V = L3 . The condition of
total (electronic and ionic) charge neutrality


Z
N
3
0 = d r (r)
,
V

(2.2)

where (r) is the local electronic density and N/V is the local ionic density, is the only
remnant of the underlying ions in the solid. If periodic boundary conditions are imposed on
any solution to Poisson equation, one verifies that
1
V

q,

L
qZ3
2

1 X 4 iqr
4 iqr
e

e
q2
V q q2

(2.3)

is a solution to Poisson equation

(r) = 4(r)

(2.4)

and that
1 X 4 iqr
e
V
q2
q6=0


1 X 4
1 q,0 eiqr
=
2
V q q
Z
 iqr
1
d3 q 4 
3
3
1

(2/L)
(q)
e ,
L
(2/L)3 q 2

Vcb (r) =

as L ,

(2.5)

is a solution to Poisson equation (2.1b). Because of translational invariance, it is more

advantageous to use the representation
X

1i<jN

1 X 4e2 iq(ri rj )
e
2
V
q
1i<jN
q6=0
" N
! N
!
#
X
X
1 X 4e2 1
eiqrj
e+iqri N
=
V
q2 2
j=1
i=1

e2 Vcb (ri rj ) =

q6=0

1 X 2e2

N
,
+q
q
V
q2
q6=0

(2.6)

whereby I have introduced the Fourier transform in reciprocal space of the local electronic
density
q :=

N
X

eiqrj

j=1

d re

d3 r eiqr (r).

iqr

j=1

=:

N
X

(r rj )
(2.7)

Hence,
N
X


p2i
1 X 2e2
H=
+

N
.
+q q
2m V q6=0 q 2
i=1

B.

(2.8)

Three-dimensional quantum Coulomb gas

The quantum jellium model is defined by Schrodinger equation

,
i~t = H
N
N
X p2j


1 X 2e2
=
H
+

N
,
N
+q q
2m V q6=0 q 2
j=1
q :=

N
X
j=1

exp iq rj ,


ria , pjb = i~ ij ab ,

q=

2
l,
L

(2.9a)
(2.9b)

l Z3 ,

i, j = 1, , N,

a, b = 1, 2, 3.

(2.9c)
(2.9d)

The N-electrons wavefunctions spanning the Hilbert space H(N ) are antisymmetric under
exchange of any two electrons and obey periodic boundary conditions in the box of volume
V = L3 that defines the solid. Equilibrium properties at inverse temperature are obtained
from the canonical partition function
Z,N := TrH(N)

h

i

exp HN .

(2.10)

What makes the task of solving for the eigenvalues of the Hamiltonian (2.9), that defines
the jellium model, difficult is the competition between the kinetic energy and the potential
energy. The kinetic energy favors extended or delocalized states in real space. The potential
energy favors localized states in real space. Accordingly, the kinetic energy is local in
momentum space whereas the potential energy is nonlocal in momentum space.
5

The choice to work in the momentum space representation in Eq. (2.9) is motivated
by the strategy to solve the noninteracting problem first and then to treat the Coulomb
interaction by a perturbative expansion. If so, what could be the small expansion parameter
that would, a-priory, justify treating the Coulomb interaction perturbatively? We can try
the following estimates. Identify the microscopic length scale
a

N
V

1/3

the classical interparticle separation.

(2.11)

With this characteristic microscopic length scale in hand, the following estimates of the
characteristic kinetic and Coulomb energies
Ekin

~2
,
2ma2

Ecb

e2
,
a

(2.12)

respectively, can be made. The ratio of the characteristic Coulomb and kinetic energies
defines a dimensionless number rs ,
Ecb
me2
a

Ekin
~2
a
aB :=
=
aB
=: rs .

~2
me2

being the Bohr radius

(2.13)

When rs 1, the kinetic energy is the largest energy scale. When rs 1, the potential
energy is the largest energy scale. I will only treat the jellium model in the limit rs 1.
The two limits rs 1 and rs 1 are not smoothly connected. In the limit rs 1,
the electronic ground state is known to break spontaneously translational invariance and is
called a Wigner crystal. Although it is believed that Wigner crystals could be realized in
some regimes of the quantum Hall effect, the limit rs 1 seems to be the relevant one to
most solids. Discussion of the transition to and of the Wigner crystal is beyond the scope of
this class. The physics of screening, Friedel oscillations, and plasmons that emerge from the
so-called RPA in the regime rs 1 will be seen to depend in an nonanalytic way on rs . By
this measure, the RPA is a highly sophisticated approximation. In fact, the first calculation
of the ground state energy within the RPA relied on perturbation theory and was a tour de
force in diagramatics. Fortunately, it is now days possible to reproduce the RPA in a more
economical, although perhaps less elementary, way. The modern methods that I will
follow here present the advantage of being of more general use than the pioneering methods
6

of the 50s (see chapters 5 and 6 of Refs. [1] and [2], respectively, for a historical perspective.)
The price to be paid is that some machinery to reformulate quantum mechanics as a path
integral has to be introduced.
Application of the rules for second quantization to the jellium model yields the identifications (~ = 1)
(r, t)

(r, t) (r, t),

(2.14a)

=,

1 X +iqr
e
q (r),
V q
Z
XX
q d3 r eiqr (r) =
ck, ck+q, ,

(r, 0)

=,

(2.14b)
(2.14c)

and
in Eq. (2.9)
H
N
=

X k2


1 X 2e2
ck ck +
+q q N
2
2m
V q6=0 q
k,

X k2
k,

2m

ck ck +

1 X 2e2 X X
ck+q, ck q, ck, ck, ,
2
V q6=0 q

k,k ,

(2.14d)

with the constraint

q=0 =

d3 r (r, t) = N

(2.14e)

on the Fock space F , here defined by application of the creation operators (r, t) on the
vacuum |0i that is annihilated by the operators (r, t). The equal-time fermionic algebra
n

o
(r, t), (r , t) = , (r r ),

n
o n
o
(r, t), (r , t) = (r, t), (r , t) = 0

(2.14f)

is obeyed. The Fourier transform of this equal-time algebra is

{
ck, , ck, } = k,k , ,

{
ck, , ck , } = {
ck, , ck , } = 0,

(2.15a)

where
1 X ikr+ik t
ck, e
,
(r, t) =
V k

1 X
(r, t) =
ck, e+ikrik t
V k
7

(2.15b)

with the single-particle momenta and energies (V = L3 )

k=

2
n,
L

n Z,

k =

k2
.
2m

(2.15c)

Observe that the constraint

= q=0 =
Q

d3 r (r, t) = N

(2.16)

on the total number of electrons means projection onto the subspace H(N ) of the fermion
can be replaced
Fock space F , for it is only in H(N ) that the total particle number operator Q
by the C-number N.
The second-quantized Hamiltonian for the jellium model in the grand canonical ensemble
according to
will be defined from the normal ordered representation of H
N

:=
H


X  k2
1 X 2e2 X X
ck ck +
ck+q, ck q, ck , ck, ,
2
2m
V q6=0 q

k,
k,k ,

(2.17)

on the one hand, and with the constraint that the chemical potential R be adjusted so
that the quantum statistical average of the local electron density be N/V , on the other hand.
, the expectation values of H
and H
in the state |0i differ by
With this definition of H



N X 2e2

h0| HN H |0i =
.
V q6=0 q 2

(2.18)

This C-number does not enter observables obtained from taking logarithmic derivatives of
the partition function, but it has to be accounted for when evaluating the logarithm of the
partition function, say, as is the case for the ground-state energy.

III.

PATH INTEGRAL REPRESENTATION OF THE COULOMB GAS

The exact number of electrons is not accessible experimentally in a macroscopic piece of

metal. Hence, we can choose the grand canonical partition function instead of the canonical
partition function to describe the jellium model. In the grand canonical ensemble, the
uniform density of electrons is allowed to fluctuate around its average, N/V , held fixed by
the choice of the chemical potential as the temperature is varied or as the thermodynamic

limit N, V is taken. In mathematical terms, the transition from the canonical to the
grand canonical ensemble is encoded by,
, H(N ) , Z

H
N
,N H , F , Z, ,
(3.1a)

X X  k2
1 X 2e2 X X

H :=
ck, ck, +
ck+q, ck q, ck , ck, ,
2
2m
V q6=0 q

k =,
k,k ,

{
ck, , ck, } = {
ck, , ck, } = 0,

{
ck, , ck , } = k,k , ,

F :=

span

1
V

XX
k

ckj ,j

Z, := TrF
*

Y

mj,



|0i mj,j = 0, 1,

h

i

exp H ,
ck, ck,

Z,

(3.1b)

:=

ck, |0i = 0,

L
k Z3 ,
2

(3.1c)
)

=, ,

(3.1d)

(3.1e)

1 ln Z,
0 held fixed fixes as a function of .
V

(3.1f)

Observe that the Coulomb interaction is normal ordered, i.e., with all creation to the left of
annihilation operators.
Instead of calculating the grand canonical partition function or its logarithmic derivatives
by performing the trace over the fermion Fock space F , I will choose to represent the grand
canonical partition function as a path integral over Grassmann coherent states. As is shown
in appendices A and B, the creation ck, (t = i ) and annihilation ck, (t = i ) operators in
the Heisenberg picture are replaced, at finite temperature, by the imaginary-time dependent

Grassmann fields k,
( ) and k, ( ), respectively. Whereas ck, (t = i ) is the adjoint of

ck, (t = i ), the two Grassmann fields k,

( ) and k, ( ), are independent of each other.

The

should not be construed as implying Grassmann conjugation. With the help of the

Grassmann integration rules defined in appendices A and B,

Z


Z, = D[ ]D[] exp S, ,
S, =

"
X

k,
( )

k,

+q ( ) :=

(3.2a)
#

k2
1 X 2e2
+
k, ( ) +
( )q ( ) ,

2m
V
q 2 +q
q6=0

(3.2b)

k,
( )k+q, ( ),

k,

(3.2c)

k,
( + ) = k,
( ),

k, ( + ) = k, ( ),
(3.2d)


1 ln Z,
1

q=0 ( ) Z :=
0 held fixed fixes as a function of .
,
V
V

(3.2e)

Here, the Grassmann measure D[ ]D[] for the two independent Grassmann fields (r, t)
Q

and (r, t) is to be understood as the product measure k,n, dk,

dk,n , . The discrete
n ,

Matsubara frequencies

n =

(2n + 1),

n Z,

(3.3)

that enter the Fourier series

1 X 1 X

exp [i (k r n )] ,
(r, ) =
nZ V k k,n ,
1 X 1 X

(r, ) =

exp [+i (k r n )] ,
nZ V k k,n ,

(3.4a)
(3.4b)

are a consequence of the antiperiodic boundary condition in the imaginary-time ( ) direction,

(r, + ) = (r, ),

(r, + ) = (r, ),

(3.5)

that must be imposed in order to construct the Grassmann path integral. Inversion of
the Fourier expansion (3.4) yields the real space and imaginary-time representation of the

10

Euclidean action S, ,
S, =

1
+
2

d3 r

"
X

IV.

dr

#



(r, )
(r, )
2m

d3 r [(r, ) 0 ]

e2
[(r , ) 0 ] .
|r r |

(3.6)

THE RANDOM PHASE APPROXIMATION

A.

Hubbard-Stratonovich transformation

I begin with the extension to infinite product measures of the Gaussian identity
r

2 + 1 z 2
e 2a =
a

Z+
1
2
dx e 2 ax +xz ,

z C,

(4.1)

valid for any a > 0. Second, I introduce the real-valued field

(r, ) =

1 XX
q,l ei(qrl ) ,
V lZ q

L
q Z3 ,
2

q=0,l = 0,

Z, (4.2)
2 l

that
obeys periodic boundary condition in the imaginary-time direction and in space
and implements the charge neutrality condition.
This field is the field conjugate to the local electronic density
(r, ) =

( )(r, )

(4.3)

through the Hubbard-Stratonovich identity

R
0

d V1

q6=0

2e2
+q ( )q ( )
q2

1
8

D[] e

e
by charge neutrality +q=0 ( ) = 0

1
8

D[] e

11

q6=0

R
0

ie

2 V

q2 +q ( )q ( )

q6=0

[+q ( )q ( )++q ( )q ( )]

q2 +q ( )q ( )

ie

2 V

[+q ( )q ( )++q ( )q ( )]

(4.4)

The physical interpretation of is that of the scalar potential associated to charge fluctuations. Charge neutrality is here implemented by the condition
Z
1
d3 r(r, ) = 0,
0 = +q=0 ( ) =
V

(4.5)

In this way, the grand canonical partition function becomes

Z, =

S , =


+1/2 Z
Z



Det
D[] D[ ]D[] exp S , ,
4

(4.6a)
#

"


X 

() +

+ ie (r, )
d r
8
2m

X
q2
+q,+l
=
q,l
8
q,l
)



2
X
ie
k

q,0 l ,0 + +q,+l k,n , .

+
k+q,
in +
n +l ,
2m
V
k,n,

(4.6b)

By analogy to the Gaussian integration over Grassmann numbers, the product measure


R
1
for any symmetric bilinear form
D[] is normalized so that D[] exp 12 A = Det
A
(symmetric kernel) A.

B.

Integration of the electrons

With the extension of the Grassmann Gaussian identity

Z

d d e a = a,
a C,

(4.7)

to infinite product Grassmann measures, integration over the fermions in the background of
the field yields
+1/2 Z



D[] exp S , ,
Z, = Det
4


Z Z

1
3

+ ie .
S , = d d r ()(r, ) 2Tr ln
8
2m
0

12

(4.8a)
(4.8b)

The functional trace Tr appears when exponentiating the fermionic determinant,



2





Det
+ ie
+ ie
= exp 2 ln Det
2m
2m




+ ie .
= exp 2Tr ln
2m

(4.9)

The prefactor of 2 multiplying the trace is due to the spin degeneracy.

C.

Gaussian expansion of the fermionic determinant

We need to approximate
Tr ln M := Tr ln(M0 + M1 )



= Tr ln M0 1 + M01 M1

= Tr ln M0 + Tr ln 1 + M01 M1 ,

where

+ ie,
2m

M0 :=
,
2m
M1 := ie.

(4.10a)

(4.10b)

M :=

(4.10c)
(4.10d)

Define the unperturbed Green function

G0 := M01 =

2m

(4.11)

The sign is convention. Perform the expansion

Tr ln M = Tr ln(G1
0 + M1 )


= Tr ln G1
+ Tr ln (1 G0 M1 )
0

X
1
1
= Tr ln(G0 )
Tr (G0 M1 )n
n
n=1

(4.12)

to the desired order. The RPA truncates this expansion to second order.
Use the compact notation
k := (k, n )
13

(4.13)

for the fermionic momentum and Matsubara indices. To first order,

Tr (G0 M1 ) =
=
=

(G0 )kk (M1 )k k

k,k

G0k k,k (M1 )k k

k,k

G0k (M1 )kk .

(4.14)

To second order,
1
1X
Tr (G0 M1 )2 =
G (M1 )kk G0k (M1 )k k
2
2 0k
k,k
1X
=
G (M1 )k(k+q) G0(k+q) (M1 )(k+q)k .
2 k,q 0k

(4.15)

Here,
(M0 )kk
(G0 )kk
(M1 )kk



k2
= in +
k,k ,
2m
1
1

=
G0k k,k ,

2
k,k
k
in k k,k
+
in 2m
ie
= kk .
V

(4.16a)
(4.16b)
(4.16c)

We see that charge neutrality insures that the first-order contribution vanishes. The first

nonvanishing contribution to the expansion of the fermion determinant in powers of e/ V

is Gaussian and given by
X e2
1
2
21
Tr (G0 M1 ) = ()
2
2 q V

X
k

G0k G0(k+q) +q q .

(4.17)

In summary, expansion of the fermionic logarithm in S , in powers of ie/[ (/2m)

] yields the first nonvanishing contribution to second order only, owing to the charge neu-

14

trality condition q=0 ( ) = 0,

RPA
Z, Z,
,

(4.18a)


+1/2 
2 Z


RPA
RPA
Z,
= Det

Det
D[] exp S,
,
4
2m
(4.18b)


2
X 1
L

q
RPA
S,
=
q ,
+q
e2 RPA
q Z3 ,
= l Z,
q
2
4
2
2 l
q=(q,l )

(4.18c)

RPA
q

2 X
G G
,
=+
V k 0k 0(k+q)

G0k =

1
in

k2
2m

1
,
in k

(4.18d)
L
k Z3 ,
2

n =

(2n + 1), n Z.

(4.18e)

Comments:
The kernel RPA
that results from the integration over the fermions within the RPA
q
approximation is called the polarization function. It endows the field with a nontrivial dynamics, i.e., is no longer instantaneous. This kernel is a property of the
occupied states of a Fermi gas at temperature T = 1 , i.e., of all states within a small
window of energy T /F 1 around the Fermi energy F = kF2 /2m, kF = kF3 N/(3 2 V )
at sufficiently low temperatures relative to the Fermi energy.
It is important to stress that the exact effective action S , in Eq. (4.8) for the
order parameter has been expanded up to quadratic order in within the RPA
approximation. However, since this expansion takes place in the argument of an
exponential, this is clearly not second-order perturbation theory in the electric charge
e.
I am using the terminology of an order parameter for the field to stress that our
treatment of the Coulomb interaction is nothing but a mean-field theory with the
inclusion of Gaussian fluctuations around the mean-field value of the order parameter.
Indeed, assume that S , can be Taylor expanded around some field configuration
15

0 ,


X
1 m S , 
S , [0 + ] =
()m ,
m 
m!

=
m=0

(4.19)

and impose the selfconsistency condition that 0 be a local extrema of S , . Then,

the mean-field value of the order parameter is the one solving


+
*
 +ie(r r )( ) 
1


0 = + ()(0 )(r, ) 2 r 
 r . (4.20)
r


8
+ ie (r , ) 

2m

The Ansatz that we made to solve this condition is 0 (r, ) = 0 for all r and , i.e., we

extended the charge neutrality condition q=0 = 0 to all qs. Since mean-field theory
is here nothing but a saddle-point approximation, it should be exact in the
limit, provided the saddle-point is an absolute minimum.
One a posteriori justification for the truncation of S , in Eq. (4.8) up to second order
RPA
in is that the functional S,
is always strictly positive as soon as 6= 0. Hence,

the mean-field configuration = 0 is, at the very least, a local minimum of the exact
effective action S , in Eq. (4.8). This is not to say that = 0 is a global minimum
of S , . Unfortunately, short of an exact solution of the problem or the identification
of an instability, for example another mean-field Ansatz with lower energy than the
energy of the Fermi gas (the ground state when = 0), it is impossible to prove that
= 0 is an absolute minimum of S , .
Mean-field theory with eventual inclusion of fluctuations within the RPA approximation or to more than Gaussian order should not be thought of as a systematic method
to solve an interacting many body problem. Rather, it is a practical method based on
physical intuition or prejudice that can only be justified a posteriori by comparison
with experiments.

APPENDIX A: GRASSMANN COHERENT STATES

To simplify notation I will consider the two-dimensional fermion Fock space F spanned
by the fermion algebra of operators
{
c, c } = 1,

{
c, c} = {
c , c } = 0.
16

(A1)

In other words F is spanned by the two normalized and orthogonal vectors

|1i := c |0i,

|0i,

(A2)

where c|0i = 0 defines the vacuum |0i. To be concrete, define the fermion harmonic oscillator
by the Hamiltonian
:= c c.
H

(A3)

There are two eigenvalues, 0 and 1, and the partition function at inverse temperature is

Z := tr|F e H = 1 + e .

(A4)

Is it possible to construct a path integral representation of this partition function as was

done for the boson harmonic oscillator in lecture 1? If the boson harmonic oscillator is
to be a guide, we need to construct coherent states defined by the conditions that they
are eigenstates of the annihilation operator c, overcomplete, and provide a resolution of the
identity. The route to fermion coherent states goes through the introduction of a Grassmann
algebra.
A four dimensional Grassmann algebra is defined by considering all possible polynomials
with complex valued coefficients that can be built from monomials in the two independent
Grassmann numbers and that obey the following multiplication rules
{, } = {, } = { , } = 0.

(A5)

In other words, and are anticommuting numbers and a generic element of the Grassmann
algebra G is written
a a1 + a2 + a3 + a4 = a1 + a2 + a3 a4 ,

a1,2,3,4 C.

(A6)

Observe that the Fock space F is a vector space spanned by |0i and |1i over the complex
numbers,

n

F := |a1 , a2 i  |a1 , a2 i = a1 |0i + a2 |1i,

a1,2 C .

(A7)

By analogy, define the Grassmann Fock space FG to be the vector space spanned by |0i and
|1i over the Grassmann algebra (A5-A6),

n




o

FG := |a, bi|a, bi = a1 + a2 + a3 + a4 |0i + b1 + b2 + b3 + b4 |1i, ai , bi C ,

(A8)

17

whereby the consistency rule that Grassmann numbers and anticommute with fermion
annihilation c and creation c operators,
{, c} = {, c} = { , c} = { , c } = 0,

(A9)

must be imposed.
Fermion coherent states |a2 i and |a3 i of the Grassmann Fock space FG are defined by

|a2 i := ea2

|a3 i := ea3

X
a2 c
|0i
n!
n=0

|0i

X
n=0

n

n
a3 c
n!

|0i = 1 a2 c |0i,

a2 C,

|0i = 1 a3 c |0i,

a3 C.

(A10a)

(A10b)

The corresponding adjoint coherent states ha2 | and ha3 | of the Grassmann Fock space FG
are defined by

c a2

ha2 | := h0|e

ha3 | := h0|ec

a3

h0|

X
(
c a )n

h0|

n!

n=0

n=0

= h0| (1 c a2 ) ,

a2 C,
(A11a)

(
c a3 )
= h0| (1 c a3 ) ,
n!

a3 C.
(A11b)

With these two definitions in hand, one verifies that |a2 i (|a3 i) is a right eigenstate of c
with Grassmann eigenvalue a2 (a3 ) and that ha2 | (ha3 |) is a left eigenstate of c with
Grassmann eigenvalue a2 (a3 ),

c|a2 i = ()2 a2 |0i = a2 |a2 i,

ha2 |
c = h0|a2 ()2 = ha2 |a2 ,
(A12a)

c|a3 i = ()2 a3 |0i = a3 |a3 i,

ha3 |
c = h0|a3 ()2 = ha3 |a3 .
(A12b)

18

Fermion coherent states are neither normalized nor orthogonal,

h | i = h0| (1 c ) 1 c |0i = 1 + ()4


h | i = h0| (1 c ) 1 c |0i = 1 + ()4


h | i = h0| (1 c ) 1 c |0i = 1 + ()4


h | i = h0| (1 c ) 1 c |0i = 1 + ()4

= e+ ,

(A13a)

= e ,

(A13b)

= 1,

(A13c)

= 1.

(A13d)

c , c) in the fermion coherent state

The expectation value of any normal ordered operator C(
|i is
c , c)|i = e+ C( , ).
h|C(

(A14)

c , c) by
Here, the Grassmann valued function C( , ) is obtained from the operator C(
replacing c by and c by .
Fermion coherent states are merely a mathematical trick that allows a path integral
representation of partition functions for fermions.1 To this end, we still need a resolution
of the identity, which, in turn, demands the notion of Grassmann integration. Grassmann
R
R
integrations d and d are multilinear mappings from the Grassmann algebra (A5-A6)

to the complex numbers which are defined by linear extension of the rules
Z
Z
Z
Z

d 1,
1 = d ,
0 = d 1,
0 = d =
Z
Z
Z
Z

0 = d 1,
0 = d =
d 1,
1 = d ,
Z
Z
Z
Z
Z

d
= d
d
d d .
d

(A15a)
(A15b)
(A15c)

Thus,

d (a1 + a2 + a3 + a4 ) = a2 a4 ,

d (a1 + a2 + a3 + a4 ) = a3 + a4 ,
Z
Z

d (a1 + a2 + a3 + a4 ) = a4 ,
d

a1,2,3,4 C,

(A16a)

a1,2,3,4 C,

(A16b)

a1,2,3,4 C.

(A16c)

Grassmann integration over the Grassmann Fock space FG in Eq. (A9-A9) is the same as

Grassmann integration over the Grassmann algebra with the caveat that c and c anticomR
R
mute with d and d .
1

A word of caution here. Gauge potentials in classical electrodynamics were also thought to be mathematical curiosities before the advent of quantum mechanics.

19

We are now ready to establish a resolution of identity for fermion coherent states. Indeed,
with the help of the resolution of the identity
11F = |0ih0| + |1ih1|,

(A17)

in F , we have
Z
Z
Z
Z

c |0ih0| (1 c )
d (1 ) 1
d e
|ih| =
d
d
Z 
Z


= d d |0ih0| |0ih1| |1ih0| + |1ih1| |0ih0|

= |1ih1| + |0ih0|
= 11F .

(A18)

Moreover, we have the trace formula

Z
Z
Z
Z

 


+ i =
d (1 ) h0| h1| C |0i + |1i
d e
h|C|
d
d
Z
Z


+ h0|C|1i

h1|C|1i

d h0|C|0i
h1|C|0i
=
d


h0|C|0i
+ h0|C|0i

= h1|C|1i

= trF C,

(A19)

for any linear operator C : F F . The asymmetry in the sign of in the fermion coherent
state trace formula should be contrasted with the bosonic coherent state trace formula. This
asymmetry will lead to the Fermi-Dirac distribution function.

APPENDIX B: PATH INTEGRAL REPRESENTATION FOR FERMIONS

With the help of the overlap, resolution of the identity, and trace formula in Eqs. (A13),
(A18), and (A19), respectively, it is possible to represent the partition function (A4) as the
Grassmann path integral,
M
1 Z
Y

M 
X

j j j1 + H(j , j1)
dj dj exp
Z = lim
M
M
j=1
j=0

Z
Z


D[ , ] exp d j ( ) ( ) + H[ ( ), ( )] ,
Z

20

)
(B1a)

where
M 0 = ( + ) = ( ),

(B1b)

M
0 = ( + ) = ( ).

(B1c)

The manipulations that lead to Eq. (B1a) are identical to those made for the boson harmonic
oscillator in lecture 1. The only change comes about due to the asymmetry of the trace
formula (A19). It leads to the integration variables in the path integral representation
in
obeying antiperiodic boundary conditions in time. Equation (B1a) holds not only for H
Eq. (A4) but for any normal ordered operator. However, since Eq. (A4) is quadratic, direct
evaluation of the first line in Eq. (B1a) can be performed,

1
0
0 0
0
0
1 M

1
0

0
0
0
0
M

1
1

0
0
0
0

Z = lim det .

.
.
.
.
.
.
.
M
..
..
..
..
.. .. ..
..

0
0

1
1
0
M

1
1
0
0
0 0
0
M
#
"
 M 1




Y
M 1
() 1
= lim 1 + ()
1
M
M j=1
M
#
"
M


= lim 1 + 1
M
M
= 1 + e ,

(B2)

as it should be. It follows that the average energy level occupation number
f := ln Z

(B3)

is nothing but the Fermi-Dirac distribution function,

e
1
f := ln Z =
=
.
1 + e
e+ + 1

(B4)

Had we imposed a boson algebra instead of the fermion algebra in Eq. (A1), the path integral
representation (B1a) would have to be modified in two ways. First, the integration variables
21

would be conventional complex numbers and the path integral would be an infinite product
of one dimensional Riemann integrals. Second, the integration variables would obey periodic
boundary conditions. These two changes would turn Eq. (B2) into
Z = 1 e

1

(B5)

from which follows that the average energy level occupation number
f := ln Z

(B6)

obeys the Bose-Einstein distribution

f := ln Z =

e
1
= +
.

1e
e 1

[1] D. Pines and P. Nozi`eres, The theory of quantum liquids, (Benjamin, New York, 1966).
[2] C. Kittel, Quatum theory of solids, (John Wiley, New York, 1963).

22

(B7)