You are on page 1of 26

LECTURES ON STATISTICAL

MECHANICS

E. MANOUSAKIS
February 18, 2011
2
Contents

1 Need for Statistical Mechanical Description 9

2 Classical Statistical Mechanics 13


2.1 Phase Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2 Statistical Distribution and Equilibrium . . . . . . . . . . . . . . 16
2.3 Ensemble and Macroscopic State . . . . . . . . . . . . . . . . . . 18
2.4 Liouville’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.5 Measurement and Averages . . . . . . . . . . . . . . . . . . . . . 23
2.6 Micro-canonical Ensemble . . . . . . . . . . . . . . . . . . . . . . 25
2.7 Statistical Independence-Fluctuations . . . . . . . . . . . . . . . 28
2.8 Entropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.9 The Law of Entropy Increase . . . . . . . . . . . . . . . . . . . . 32
2.10 Temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.11 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

3 Derivation of Thermodynamics 37
3.1 Reversibility and Adiabatic Process . . . . . . . . . . . . . . . . . 37
3.2 First Law of Thermodynamics . . . . . . . . . . . . . . . . . . . . 39
3.3 Second law of thermodynamics . . . . . . . . . . . . . . . . . . . 41
3.4 Thermodynamic Functions . . . . . . . . . . . . . . . . . . . . . . 41

4 Quantum Statistical Mechanics 47


4.1 The state of a macroscopic system . . . . . . . . . . . . . . . . . 47
4.2 Density Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.3 Statistical Averages . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.4 Micro-canonical Ensemble . . . . . . . . . . . . . . . . . . . . . . 53
4.5 Entropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.6 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

5 Other Ensembles 57
5.1 Canonical Ensemble . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.2 Thermodynamic Quantities . . . . . . . . . . . . . . . . . . . . . 59
5.3 Examples from the Canonical Ensemble . . . . . . . . . . . . . . 59
5.3.1 Distinguishable Free-Particles . . . . . . . . . . . . . . . . 59

3
4 CONTENTS

5.3.2 Classical Partition Function . . . . . . . . . . . . . . . . . 62


5.3.3 Magnetic Dipoles in a Magnetic Field . . . . . . . . . . . 62
5.4 Grand Canonical Ensemble . . . . . . . . . . . . . . . . . . . . . 66
5.5 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

6 Simulations 69
6.1 The general idea . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
6.2 Monte Carlo Simulation . . . . . . . . . . . . . . . . . . . . . . . 70
6.3 Solving the equations of motion . . . . . . . . . . . . . . . . . . . 72
6.4 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
6.4.1 The Classical gas and liquid . . . . . . . . . . . . . . . . . 73
6.4.2 The Heisenberg model . . . . . . . . . . . . . . . . . . . . 74
6.5 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

7 Quantum Gases 81
7.1 Identical Particles . . . . . . . . . . . . . . . . . . . . . . . . . . 81
7.2 Density Matrix of Free Particles . . . . . . . . . . . . . . . . . . 83
7.3 Fermi Gas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
7.4 Bose Gas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
7.5 Bose-Einstein Condensation . . . . . . . . . . . . . . . . . . . . . 87
7.6 Blackbody Radiation . . . . . . . . . . . . . . . . . . . . . . . . . 90
7.7 Cosmic Background Radiation . . . . . . . . . . . . . . . . . . . 93
7.8 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94

8 Thermodynamics of a Solid 97
8.1 Atomic Vibrations . . . . . . . . . . . . . . . . . . . . . . . . . . 97
8.1.1 A Single Harmonic Oscillator . . . . . . . . . . . . . . . . 97
8.1.2 Thermodynamics of a Single Harmonic Oscillator . . . . . 98
8.2 Mono-atomic Chain . . . . . . . . . . . . . . . . . . . . . . . . . 99
8.2.1 Classical Treatment . . . . . . . . . . . . . . . . . . . . . 100
8.2.2 Quantization of the phonon field . . . . . . . . . . . . . . 101
8.3 Thermodynamic Behavior of Lattice Vibrations . . . . . . . . . . 103

9 The Fermi Gas at Low Temperature 109


9.1 Low temperature expansion in a Fermi gas . . . . . . . . . . . . . 109
9.2 Magnetism of a Fermi-Gas . . . . . . . . . . . . . . . . . . . . . . 112
9.3 Relativistic electron gas–White Dwarfs . . . . . . . . . . . . . . . 113
9.4 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116

10 The Ising Model 117


10.1 The Ising Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
10.2 The One-dimensional Ising Model . . . . . . . . . . . . . . . . . . 118
10.3 Mean-field approximation . . . . . . . . . . . . . . . . . . . . . . 119
10.4 Bragg-Williams Approximation . . . . . . . . . . . . . . . . . . . 123
10.5 Bethe-Peierls Approximation . . . . . . . . . . . . . . . . . . . . 126
10.6 Exact solution for the square lattice . . . . . . . . . . . . . . . . 129
CONTENTS 5

10.7 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130

11 Critical Phenomena 133


11.1 The order parameter . . . . . . . . . . . . . . . . . . . . . . . . . 135
11.2 Universality-Critical exponents . . . . . . . . . . . . . . . . . . . 136
11.3 Landau Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
11.4 Effective Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
11.5 General Approach and Conclusions . . . . . . . . . . . . . . . . . 142
11.6 Role of Long Wave-Length Fluctuations . . . . . . . . . . . . . . 143
11.7 Finite-Size Scaling . . . . . . . . . . . . . . . . . . . . . . . . . . 145
11.8 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148

12 Path Integrals 151


12.1 Path Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
12.2 Trotter Approximation . . . . . . . . . . . . . . . . . . . . . . . . 152
12.3 Mapping the quantum to a classical system . . . . . . . . . . . . 154
12.4 Path integral for identical particles . . . . . . . . . . . . . . . . . 156
12.5 Quantum spin models . . . . . . . . . . . . . . . . . . . . . . . . 160
12.6 Path integral-Boson coherent states . . . . . . . . . . . . . . . . . 164
12.7 Path integral-Fermion coherent states . . . . . . . . . . . . . . . 165

13 Perturbation Theory 167


13.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
13.2 Green’s functions . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
13.3 Generating function . . . . . . . . . . . . . . . . . . . . . . . . . 170
13.4 Free particles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
13.5 Higher order Green’s functions . . . . . . . . . . . . . . . . . . . 175
13.6 Expansion of the partition function . . . . . . . . . . . . . . . . . 175
13.7 Hugenholtz Diagrams . . . . . . . . . . . . . . . . . . . . . . . . 182
13.8 Summary of diagram rules . . . . . . . . . . . . . . . . . . . . . . 183
13.9 Linked Cluster Expansion . . . . . . . . . . . . . . . . . . . . . . 186
13.10Averages of operators . . . . . . . . . . . . . . . . . . . . . . . . 188
13.11Zero temperature expansion . . . . . . . . . . . . . . . . . . . . . 191
13.12Finite-density of fermions . . . . . . . . . . . . . . . . . . . . . . 196
13.13Frequency Fourier transform of G . . . . . . . . . . . . . . . . . . 199
13.14Perturbation theory for bosons . . . . . . . . . . . . . . . . . . . 201
13.15Analytic properties of the Green’s function . . . . . . . . . . . . 206

14 Response Theory 211


14.1 Linear Response . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
14.2 Density-density response function . . . . . . . . . . . . . . . . . . 212

15 Normal Fermi Fluid 215


15.1 Charged Fermi system . . . . . . . . . . . . . . . . . . . . . . . . 215
15.2 Fermi-gas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
15.3 Plasmon mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
6 CONTENTS

15.4 Quasiparticles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223

16 Bose Fluid 231


16.1 Green’s functions . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
16.2 Dyson’s Equations . . . . . . . . . . . . . . . . . . . . . . . . . . 233
16.3 Bogoliubov’s Trasformation . . . . . . . . . . . . . . . . . . . . . 239

17 Instabilities of the Fermi fluid 241


17.1 Pairing Instability . . . . . . . . . . . . . . . . . . . . . . . . . . 241
17.2 Bogoliubov Transformation . . . . . . . . . . . . . . . . . . . . . 242
17.3 Superconductivity . . . . . . . . . . . . . . . . . . . . . . . . . . 243
17.4 Spin Density Waves . . . . . . . . . . . . . . . . . . . . . . . . . 244
17.5 Charge Density Waves . . . . . . . . . . . . . . . . . . . . . . . . 245
17.6 Other Possible Instabilities . . . . . . . . . . . . . . . . . . . . . 246
17.7 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247

18 Renormalization Group Approach 249


18.1 Scale Transformation . . . . . . . . . . . . . . . . . . . . . . . . . 250
18.2 RG Flow-Fixed Points . . . . . . . . . . . . . . . . . . . . . . . . 252
18.3 Real Space RG Transformation . . . . . . . . . . . . . . . . . . . 253
18.4 Momentum Space RG . . . . . . . . . . . . . . . . . . . . . . . . 254
18.5 The Gaussian Model . . . . . . . . . . . . . . . . . . . . . . . . . 255
18.6 One-Component φ4 Theory-Ising Model . . . . . . . . . . . . . . 256
18.7 The Heisenberg and the non-linear σ Model . . . . . . . . . . . . 257
18.8 Renormalization of the σ Model . . . . . . . . . . . . . . . . . . . 258
18.9 Loop and  Expansion . . . . . . . . . . . . . . . . . . . . . . . . 261
18.10RG Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
18.11n-Component φ4 Theory . . . . . . . . . . . . . . . . . . . . . . . 263
18.12Fluctuations in 2D . . . . . . . . . . . . . . . . . . . . . . . . . . 266
18.13Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267

19 The 2D XY Model-Vortices 269


19.1 Role of Spin Waves in 2D . . . . . . . . . . . . . . . . . . . . . . 273
19.2 KT Phase Transition . . . . . . . . . . . . . . . . . . . . . . . . . 275
19.3 Renormalization Group approach . . . . . . . . . . . . . . . . . . 276
19.4 Results from Simulations . . . . . . . . . . . . . . . . . . . . . . 277

A Second Quantization 311


A.1 Hilbert space for identical particles . . . . . . . . . . . . . . . . . 312
A.2 Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 314
A.3 Creation and annihilation operators . . . . . . . . . . . . . . . . 315

B Coherent States 319


B.1 Coherent States for bosons . . . . . . . . . . . . . . . . . . . . . 319
B.2 Grassmann algebra . . . . . . . . . . . . . . . . . . . . . . . . . . 322
B.3 Fermion coherent states . . . . . . . . . . . . . . . . . . . . . . . 324
CONTENTS 7

B.4 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 328


B.5 Gaussian Integrals-Bosons . . . . . . . . . . . . . . . . . . . . . . 329
B.6 Gaussian Integrals-Fermions . . . . . . . . . . . . . . . . . . . . . 329
310 CONTENTS
Appendix A

Second Quantization

Here we shall present the formalism of second quantization. There is no addi-


tional quantization, the quantization procedure is only one. The name “second
quantization” originates in the history of the problem and the name itself gives
no information on what it is it might even be confusing. A more appropriate
name is “formulation of operators in terms of creation and annihilation opera-
tors”. This is exactly what we plan on doing.

Consider that the particle number is not definite, not because that we nec-
essarily plan to introduce interactions in which particles are destroyed or cre-
ated. The deeper reason is that in many problems the physical eigenstates of
the system, might behave as if they do not respect the particle number defi-
niteness. Sometimes they behave as the definiteness of the particle number is
“transcended”. Using a more technically precise vocabulary, even though the
particle conservation is an exact symmetry of the Hamiltonian of the system,
the state of the system might show little respect for that symmetry in the limit
of N → ∞. The symmetry of the Hamiltonian is restored in any finite size
system by mixing many states and each one of them does not respect the sym-
metry but the entire state does. However, there is an energy associated with
precession of one such state into another (the mixing of these states) which as
N becomes larger and larger this frequency becomes smaller and smaller. In
the thermodynamics limit the system might find itself stuck in one such sector
where the symmetry is not respected. The way one detects such a tendency in
a given system is the development of correlations in the particle creation and
in its creation at some very far distance from the creation place. The particle
actually has not vanished, it has simply disappeared from the part of space we
are looking, thus, creating a nice illusion for us, which has marvelous conse-
quences since this distance gets larger and larger and in the thermodynamic
limit it becomes infinite. When one measure disappears by becoming infinite,
other quantities also leap into the immeasurable.

311
312 APPENDIX A. SECOND QUANTIZATION

A.1 Hilbert space for identical particles


Let us start by considering the Hilbert space H which has no definite number
of particles, symbolically:

H = H0 ⊕ H1 ⊕ H2 ⊕ ...HN ⊕ ... (A.1)

Namely, it is thought as the direct sum of the Hilbert subspaces Hn with well-
defined number of particles n.
Now a complete basis of the HN for distinguishable particles can be con-
structed out of the basis elements of H1 as follows:

|γ1 , γ2 , ..., γN i = |γ1 i1 ⊗ |γ2 i2 ⊗ ...|γN iN (A.2)

where |γii for fixed i (here γ labels the different basis states) is a complete basis
of H1 and i is the particular particle identity index. This is the same as say
that
HN = H1 ⊗N (A.3)
where the direct product of H1 with itself is taken N times.
For indistinguishable particles the situation is different. In such cases one
needs to project out only the symmetric or antisymmetric part of the Hilbert
space as discussed in Chapter 7. In the case of bosons or fermions, a basis in
the HN is constructed as follows:

|γ1 , γ2 , ..., γN )± = P̂± |γ1 i1 |γ2 i2 ...|γN iN (A.4)

where P̂± is a symmetrization (+) or antisymmetrization (-) operator defined


as
1 X
(±1)[P ] |γ1 iP 1 |γ2 iP 2 ...|γN iP N

P̂± |γ1 i1 |γ2 ...|γN iN = √ (A.5)
N! P

and the notation regarding the summation is the same with that used in Chapter
7. Here we have chosen to permute particle indices, however, the same can
be done by choosing to permute the indices of the single-particle states. The
larger Hilbert space H can be projected to two (dimensionally smaller) spaces
each one separately transforming according to the representations of the pair
permutations. Symbolically,
H± = P̂± H. (A.6)
Let us say that we have ordered the single-particle basis states according to our
choice so that they are in one to one correspondence with the natural numbers
1,2,3,...,k,.... Then, let the set ni , i = 1, 2, ..., k, ..., ∞ give the number of par-
ticles occupying each one of these single-particle states using the same order
ordering. The numbers ni are called, occupation numbers. Therefore given a
state characterized by the set of single particle states γ1 , γ2 , ..., γN there is a
corresponding set of occupation numbers ni with at most N non-zero elements.
Now it is clear by examining Eq. (A.4) that for fermions each ni can only take
A.1. HILBERT SPACE FOR IDENTICAL PARTICLES 313

two possible values ni = 0, 1, otherwise the state collapses to zero owing to


the cancellation of two identical terms (with opposite signs) the following: any
term and the corresponding one produced by a permutation of the two particles
which are placed in the same state. With the convention that 0! = 1, there is
no ambiguity in the case of fermion states. However, in the case of bosons in
order guarantee normalization of the state when ni 6= 0, 1, we need to divide by
a normalization factor, which the square root of the number of ways by which
one is able to re-arrange n1 states, n2 states and ,....
Next we shall examine the normalization of the states. Let us consider the
overlap of two states. It is immediately clear that the states are orthogonal if
they are states corresponding to different particle numbers. Thus, we consider
states having the same particle number

1 X 0
(γ1 , γ2 , ...γN |γ10 , γ20 , ..., γN
0
)± = (±1)([P ]+[P ]) hγP 1 |γP0 0 1 i1
N! 0P,P

hγP 2 |γP0 0 2 i2 ...hγP N |γP0 0 N i


1 XX 0 0 0
= (±1)[Q] hγP 1 |γQP 1 i1 hγP 2 |γQP 2 i2 ...hγP N |γQP N i. (A.7)
N!
P Q

Here, for convenience we have permuted the state indices. The second part
of the equation is obtained by realizing that we can write P 0 = QP where Q
another permutation and change the summation variables from P and P 0 to P
and Q. The overall sign is determined by the sign of the relative permutation
Q. Then the sum over Q is independent of P which can be regarded as the
reference permutation. Thus, one find that
X
(γ1 , γ2 , ...γN |γ10 , γ20 , ..., γN
0
)± = 0
(±1)[Q] hγ1 |γQ1 0
i1 hγ2 |γQ2 i2
Q
0
...hγN |γQN i. (A.8)

Now, it is clear that, unless the set of states {γ1 , γ2 , ...γN } is the same as the
set {γ10 , γ20 , ...γN
0
} the overlap is zero. Therefore they can be different by only a
permutation P0 in which case the overlap is the same as the one with itself mul-
tiplied by the factor (±1)[P0 ] which is required to create the same permutation.
Therefore, let us consider
X
(γ1 , γ2 , ...γN |γ1 , γ2 , ..., γN )± = (±1)[Q] hγ1 |γQ1 i1 hγ2 |γQ2 i2
Q
...hγN |γQN i. (A.9)

For Fermions all single-particle states in both N -particle states are different thus,
only the identity permutation gives non-zero contribution. Thus for fermions,
the states defined by Eq. (A.4) are normalized.
Q∞ For the case of bosons, however,
there can be ni > 1, then, there are i=1 ni ! permutations corresponding to the
same state. Thus, for both fermions (for fermions always ni ! = 1) and bosons
314 APPENDIX A. SECOND QUANTIZATION

one can write:



Y
(γ1 , γ2 , ...γN |γ1 , γ2 , ..., γN )± = ni !. (A.10)
i=1

Thus an orthonormal set of states can be defined as


1
|γ1 , γ2 , ...γN i± = pQ∞ |γ1 , γ2 , ..., γN )± . (A.11)
i=1 ni !

Once such a complete set of orthonormal states is constructed, then the com-
pleteness relationship in each of the spaces H± is automatic
∞ X
X
|0ih0| + |γ1 , γ2 , ..., γN ihγ1 , γ2 , ..., γN | = 1. (A.12)
N =1 {γi }

For simplicity in the notation, we have omitted the label ± of the state which
stands for its symmetrization or antisymmetrization. Here |0i stands for the no-
particle state. Here, we need to worry about the meaning of the summation. We
should sum over all different states. Any permutation of the states γ1 , γ2 , ..., γN
does not produce any different state because of the symmetrization (or anti-
symmetrization). Thus, the summation is over all different sets {γi }. Two sets
with the same elements but differing by a permutation are still the same sets.

A.2 Operators
An operator Ô which does not change the particle number can be expressed in
this orthonormal basis as
X X
Ô = |γ1 , ..., γN ihγ1 , ..., γN |Ô|γ10 , ..., γN
0
ihγ10 , ..., γN
0
|. (A.13)
N {γ},{γ 0 }

Here, again, we have omitted the anti-symmetrization/symmetrization label for


simplicity.
Let us, first, consider a one-body operator such as an external field acting
on each one of the particles or the kinetic energy operator. In the standard
notation:
N
(1)
X
Ô(1) = ôi . (A.14)
i=1

In this case, it is straightforward for one to carry out the evaluation of the
matrix elements which enter in (A.13) and show that

N
XX X
0 0
Ô(1) = |γ1 , ..., γm , ..., γN ihγm |ô(1) |γm ihγ1 , ..., γm , ..., γN |. (A.15)
N m=1 0
{γ},γm
A.3. CREATION AND ANNIHILATION OPERATORS 315

In the case of two-body operators such as the usual pair interaction


N
1 X (2)
Ô(2) = ô . (A.16)
2 i=1,j=1 ij

we find
N
1X X X
Ô(2) = |γ1 , ..., γm , ..., γn , ..., γN i
2 m=1,n=1 0 ,γ 0
N {γ},γm n

0
hγm , γn |ô(2) |γm , γn0 ihγ1 , ..., γm
0
, ..., γn0 , ..., γN |. (A.17)

In the same way one can define other many-particle operators. We are only going
to be concerned with problems involving at the most two-particle operators.

A.3 Creation and annihilation operators


We, now, wish to define the following operators

â†γ |γ1 , γ2 , ..., γN )± ≡ |γ, γ1 , γ2 , ..., γN )± . (A.18)

and its application on the orthonormal set of states (A.11) gives



â†γ |γ1 , γ2 , ..., γN i± = ni + 1|γ, γ1 , γ2 , ..., γN i± . (A.19)

The operator itself increases the occupation number of that particular state
by one. Thus the state on the left-hand-side is an N particle state while the
state produced on the right-hand-side is an N +1 state with the same symmetry
(symmetric
√ or antisymmetric). It is clear that for the case of fermions the factor
ni + 1 will never play any role because if ni = 0 then this factor is 1, while if
ni ≥ 1 we cannot produce a fermion state with ni + 1 ≥ 2 fermions in the same
single-particle state, thus, the state in the right-hand-side will be zero.
The adjoint operator of â†i , namely âi ≡ (â†i )† . We wish to find what is the
result of the action of âi on any state |γ1 , γ2 , ..., γN i± as defined by Eq. (A.4).
We obtain:

âi |γ1 , γ2 , ..., γN )± = h0|âi |γ1 , γ2 , ..., γN i± |0i +



X X
hγ10 , γ20 , ..., γn0 |âγ |γ1 , γ2 , ..., γN )± |γ10 , γ20 , ..., γn0 )±
n=1 {γi0 ,i=1,...,n}

N
X
= (±1)m−1 δγ,γm |γ1 , γ2 , ..., γ̌m , γN )± . (A.20)
m=1

where the notation γ̌P 1 means that the state γP 1 is missing, thus, the state
|γ̌P 1 , γP 2 , ..., γP N i± is an N − 1 particle state. The last equation is obtained by
316 APPENDIX A. SECOND QUANTIZATION

noticing that

hγ10 , γ20 , ..., γn0 |âγ |γ1 , γ2 , ..., γN )± = hγ1 , γ2 , ..., γN |â†γ |γ10 , γ20 , ..., γn0 )∗±
= hγ1 , γ2 , ..., γN |γ, γ10 , γ20 , ..., γn0 )∗± (A.21)

The last part is non-zero only when n = N − 1 and the set of single-particle
states {γ, γ10 , γ20 , ..., γN
0
−1 } is the same as the set {γ1 , γ2 , ..., γN }. Therefore,
there are N different possibilities so that this can happen which are: γ = γm ,
and the set γ10 , γ20 , ..., γN0
−1 is the same as the set γ1 , ..., γˇ
m , ..., γN .
Therefore we have been able to show the following:
√ √
âγ |γ1 , γ2 , ..., γN i± = ni |γ, γ1 , γ2 , ..., γN i± , âi |{ni }i = ni |{ni }i. (A.22)

One can easily show that the following commutation or anticommutation


relations are valid respectively for the boson or the fermion creation and anni-
hilation operators.
[aγ , aγ 0 ]∓ = 0 (A.23)

[a†γ , a†γ 0 ]∓ = 0 (A.24)

[aγ , a†γ 0 ]∓ = δγ,γ 0 (A.25)


where the symbol [a, b]∓ = ab ∓ ba and the commutator [a, b]− is used for the
case of bosons and the anticommutator [a, b]+ for fermions. These relations can
be easily shown by straightforward application to an N -particle state using Eq.
(A.18) and Eq. (A.20).
Since for identical particles there is no meaning to ask which particle occu-
pies which state, the many-particle state is specified by giving the occupation
numbers of all states, namely in the state |n1 , n2 , ..., nk , ..., i± the integers ni
with i = 1, 2, ... give how many particles occupy the ith state.
It can be easily shown that the one-body operator which has been expressed
as (A.15) can be written in terms of creation and annihilation operators as:
X
Ô(1) = hγ|ô(1) |γ 0 ia†γ aγ 0 . (A.26)
γ,γ 0

Also using the expression (A.17) for the two-body operator we can show that

1 X
Ô(2) = hγ1 , γ2 |ô(2) |γ10 , γ20 ia†γ1 a†γ2 aγ20 aγ10 . (A.27)
2
γ1 ,γ2 ,γ10 ,γ20

As an example of this we consider the case of a uniform system of interacting


non-relativistic particles such as the Hamiltonian
N
h̄2 X 2 1 X
H=− ∇ + u(rij ), (A.28)
2m i=1 i 2
i6=j
A.3. CREATION AND ANNIHILATION OPERATORS 317

for a uniform fluid or electron gas. In the non-interacting plane wave basis this
Hamiltonian takes the form
1
e(p)a†p~σ ap~σ + p1 , p~2 ia†p~0 σ1 a†p~0 σ2 ap~2 σ2 ap~1 σ1 .
X X
H= p01 , p~02 |u|~
h~
2 0 0
1 2
p
~σ p
~1 ,~
p2 ,~ p2 ,σ1 ,σ2
p1 ,~
(A.29)
where the creation operator a†p~σ adds a particle in an eigenstate of the momen-
tum and spin:
a†p~σ |0i = |~
pi|σi (A.30)
where
1 i
pi = √ e h̄ p~·~r
h~r|~ (A.31)
V
and |σi is the spin-state with spin projection along the z-axis σ. The matrix
element can be easily evaluated
1
p01 , p~02 |u|~
h~ p1 , p~2 i = ũ(~q)δp~1 +~p2 ,~p01 +~p02 (A.32)
V
where the Kronecker δ expresses the momentum conservation and ũ(~q) is the
Fourier transform of u(r)
Z
i
ũ(~q) = d3 ru(r)e h̄ q~·~r (A.33)

and ~q = p~1 − p~01 = p~2 − p~02 is the momentum transfer.


The same Hamiltonian can be expressed in terms of creation and annihilation
operators ψ̂ † (~r, σ) which add particles in eigenstates of the position operator:

h̄2 X
Z
H=− d3 xψ̂ † (~r, σ)∇2 ψ̂(~r, σ) +
2m σ
X 1Z
d3 xd3 yu(|~x − ~y |)ψ̂ † (~x, σ1 )ψ̂ † (~y , σ2 )ψ̂(~y , σ2 )ψ̂(~x, σ1 ). (A.34)
σ ,σ
2
1 2
318 APPENDIX A. SECOND QUANTIZATION
Appendix B

Coherent States

B.1 Coherent States for bosons


Now, we wish to try to construct eigenstates of the annihilation operator, namely

aλ |{φµ }i = φλ |{φµ }i. (B.1)

The states |{φµ }, µ = 1, 2, ...i are the so-called coherent states. However, we
wish to construct them explicitly. Let us expand any such state, in the complete
basis |{nµ }i as X
|{φµ }i = C({nµ })|{nµ }i (B.2)
{nµ }

This equation implies:



h{mν }|aλ |{φµ }i = C({mν6=λ }, mλ + 1) mλ + 1. (B.3)

Using Eq. (B.1) we find that

h{mν }|aλ |{φµ }i = φλ C({mν }). (B.4)

Thus, from the last two Eqs. we find


φλ
C({nµ6=λ }, nλ + 1) = √ C({nν6=λ }, nλ ). (B.5)
nλ + 1
This equation can be used recursively to yield:
Y φ nλ
C({nλ }) = √ λ C(0, 0, ...) (B.6)
λ
nλ !

and normalizing the state such that for the vacuum c(0, 0, ...) = 1 we obtain:
X Y (φλ a† )nλ
λ
|{φλ }i = |0i. (B.7)
nλ !
{nλ } λ

319
320 APPENDIX B. COHERENT STATES

which implies that P


φλ0 a†λ0
|{φµ }i = e λ0 |0i. (B.8)
The state |{φµ }i is not normalized and it is straightforward to show that its
norm is given by: P
φ∗
µ φµ
h{φµ }|{φµ }i = e µ . (B.9)
In addition, one can show that
P
φ∗ 0
µ φµ
h{φµ }|{φ0µ }i = e µ , (B.10)

which implies that the set of coherent states is an overcomplete set of states. It
can be easily shown that the completeness relation is the following:
 dφµ dφ∗µ  − P φ∗ φλ
Z Y
e λ λ |{φν }ih{φν }| = 1. (B.11)
µ
2πi

A simple way to show this is by computing the matrix elements of both sides
using states in the occupation number representation.
If we differentiate both sides of the Eq. (B.8) with respect to φλ we obtain:


a†λ |{φµ }i = |{φµ }i. (B.12)
∂φλ

From Eq.(B.8) we also find that

hφ|a†λ = hφ|φ∗λ (B.13)

and

hφ|aλ = hφ|. (B.14)
∂φ∗λ
Therefore for any state |Ψi we obtain:


hφ|aλ |Ψi = Ψ(φ∗ ) (B.15)
∂φ∗λ

and
hφ|a†λ |Ψi = φ∗λ Ψ(φ∗ ) (B.16)

where Ψ(φ ) = hφ|Ψi.
Using the completeness relation (B.11) we can express the overlap between
two states |Ψi and Ψ0 i as

 dφµ dφ∗µ  − P φ∗ φλ
Z Y
hΨ|Ψ0 i = e λ λ hΨ|{φν }ih{φν }|Ψ0 i
µ
2πi
 dφµ dφ∗µ  − P φ∗ φλ ∗
Z Y
= e λ λ Ψ ({φν })Ψ0 ({φ∗ν }). (B.17)
µ
2πi
B.1. COHERENT STATES FOR BOSONS 321

Therefore we we that the operators a†λ and P̂λ = ih̄aλ behave as position
and momentum operators respectively in the coherent state representation since

a†λ → φ∗λ (B.18)



P̂λ = ih̄aλ → ih̄ ∗ (B.19)
∂φλ

In addition, the standard momentum-position commutation relations hold for


these operators also.
It is useful to find the analogous relation to the relation
1 i
hx|pi = √ e− h̄ px (B.20)
V
which gives the overlap between momentum and position eigenstates. Let |P i
be the eigenstates of the “momentum” operator P̂λ :

P̂λ |{Pµ }i = Pλ |{Pµ }i. (B.21)

Where the eigenvalues Pλ are ih̄φλ . This equation implies:

hφ|P̂λ |P i = Pλ hφ|P i. (B.22)

which can be written as follows



ih̄ hφ|P i = Pλ hφ|P i (B.23)
∂φ∗λ

and by integrating with respect to φλ we find:


i
P ∗
hφ|P i = e− h̄ λ Pλ φλ (B.24)

which is the relation we were looking for and it can be identified that it is the
same as Eq. (B.10).
Using the completeness relation (B.11) we can expand the state |Ψi in the
basis of coherent states as
 dφµ dφ∗µ  − P φ∗ φλ
Z Y
|Ψi = e λ λ hφ|Ψi|φi (B.25)
µ
2πi

and, thus, we obtain


 dφµ dφ∗µ  − P (φ∗ −φ0 ∗ )φλ
Z Y
0

Ψ(φ ) = e λ λ λ
Ψ(φ∗ ) (B.26)
µ
2πi

which can be rewritten in terms of the momentum operator eigenvalues as


 dφ∗µ dPµ  i P (φ∗ −φ∗0 )Pλ
Z Y
Ψ(φ0∗ ) = e h̄ λ λ λ Ψ(φ∗ ). (B.27)
µ
2πh̄
322 APPENDIX B. COHERENT STATES

Eq. (B.27) is a consequence of the fact that


Z
dPλ i Pλ (φ∗λ −φ0∗
λ ) = δ(φ∗ − φ0∗ ).
e h̄ λ λ (B.28)
2πh̄
The next question is how one evaluates the matrix elements, using the coherent
states as a basis set, of an an operator Ô({â†λ , âλ }) which is a function of creation
and annihilation operators. Let us further assume that the operator is in the
so-called normal ordered form, which is a polynomial of products of creation
and annihilation operators in which all the creation operators are on the left of
the product of the annihilation operators. The matrix element of an operator
which is in a normal ordered form can be calculated using Eqs. (B.1), (B.13)
and (B.10)
P ∗ 0
h{φλ }|Ô({â†λ , âλ })|{φ0λ }i = O({φ̂∗λ , φ0λ })e λ φλ φλ (B.29)
where the exponential factor is due to the overlap of the coherent states Eq.
(B.10).

B.2 Grassmann algebra


In order to introduce fermion states which are analogous to the boson coherent
states it is necessary to introduce the so-called Grassmann algebra.
We introduce a set of generators ψ1 , ψ2 , ..., ψn of the algebra which are an-
ticommuting numbers:
ψλ ψλ0 = −ψλ0 ψλ (B.30)
thus ψλ2 = 0. We also need to define the operation of conjugation on one such
Grassmann variable:

ψλ → ψλ∗ (B.31)
In other words, every Grassmann number has a counterpart in the algebra to
which is related by conjugation operation. This can be written more simply as
(ψλ )∗ = ψλ∗ (B.32)
(ψλ∗ )∗ = ψλ . (B.33)
The properties of the conjugation operation are the following:
(ψλ1 ψλ2 ...ψλn )∗ = ψλ∗n ψλ∗n−1 ...ψλ∗1 (B.34)
∗ ∗
(cψλ ) =c ψλ∗ , (B.35)
where c is an arbitrary complex number, not a Grassmann number. We shall
consider the case of an algebra with two generators ψ and ψ ∗ to illustrate
the properties of the algebra. The generalization to an arbitrary number of
generators is straightforward. All possible powers of them are 1, ψ, ψ ∗ and
ψψ ∗ . The most general function f (ψ) of ψ is a linear function:
f (ψ) = f0 + f1 ψ (B.36)
B.2. GRASSMANN ALGEBRA 323

because higher powers of ψ vanish. The most general function O(ψ, ψ ∗ ) of both
ψ and ψ ∗ is the following:

O(ψ, ψ ∗ ) = O0 + O1 ψ + Ō1 ψ ∗ + O12 ψ ∗ ψ. (B.37)

Having defined the most general functions of Grassmann variables we can de-
fine the operations of the “derivative” and of the definite “integral” on these
functions. We introduce the derivative as follows:
∂ψ
=1 (B.38)
∂ψ
∂(ψψ ∗ )
= ψ∗ (B.39)
∂ψ
∂(ψ ∗ ψ)
= −ψ ∗ (B.40)
∂ψ

where the last equation follows from the fact that these variables anticommute.
Therefore using these definitions we can “compute” the derivative of any func-
tion f (ψ) or of O(ψ, ψ ∗ ) as follows:

∂f (ψ)
= f1 (B.41)
∂ψ
∂O(ψ, ψ ∗ )
= O1 − O12 ψ ∗ (B.42)
∂ψ
∂O(ψ, ψ ∗ )
= Ō1 + O12 ψ (B.43)
∂ψ ∗
∂ ∂
O(ψ, ψ ∗ ) = −O12 (B.44)
∂ψ ∗ ∂ψ
∂ ∂
O(ψ, ψ ∗ ) = O12 (B.45)
∂ψ ∂ψ ∗

which imply that


∂ ∂ ∂ ∂

=− (B.46)
∂ψ ∂ψ ∂ψ ∂ψ ∗
which implies that the derivatives with respect to Grassmann variables anticom-
mute. Notice that the “derivative” has no similar meaning with the definition
of the derivative of an analytic function. In this case we use the name, namely
derivative of a function of Grassmann variables, to define the class of operations
on such functions which have been defined before.
We also need to define the operation of definite “integration” of a function
of Grassmann variables. We define:
Z
dψ1 = 0 (B.47)
Z
dψψ = 1, (B.48)
324 APPENDIX B. COHERENT STATES

and based on these we also have


Z
dψ ∗ = 0 (B.49)
Z
dψ ∗ ψ ∗ = 1. (B.50)

Using these definitions we can integrate any function of Grassmann variables:


Z
dψf (ψ) = f1 (B.51)
Z
dψO(ψ, ψ ∗ ) = O1 − O12 ψ ∗ (B.52)
Z
dψ ∗ O(ψ, ψ ∗ ) = Ō1 + O12 ψ (B.53)
Z
dψ ∗ dψO(ψ, ψ ∗ ) = −O12 (B.54)
Z
dψdψ ∗ O(ψ, ψ ∗ ) = O12 (B.55)

(B.56)

which imply that


Z Z
dψ ∗ dψO(ψ, ψ ∗ ) = − dψdψ ∗ O(ψ, ψ ∗ ) (B.57)

The overlap integral of two functions f (ψ) and g(ψ) of Grassmann variables is
defined as follows:
Z

hf |gi ≡ dψ ∗ dψe−ψ ψ f ∗ (ψ)g(ψ ∗ ) (B.58)

in analogy with Eq. (B.10) for bosons. Using the definition of the integration
and the most general linear functions such as (B.36) we can easily show that

hf |gi = f0∗ g0 + f1∗ g1 (B.59)

which coincides with the definition of the inner product in a vector space.

B.3 Fermion coherent states


We shall use these Grassmann variables to define fermion coherent states. These
states are not physical states but when we shall express the observables, such
as the trace of e−β(Ĥ−µN̂ ) in terms of coherent states they should give the same
result as that computed in the physical Fock space.
In order to construct fermion coherent states as a linear combination of
fermion antisymmetric states using Grassmann variables as coefficients we first
B.3. FERMION COHERENT STATES 325

need to associate Grassmann variables ψλ and ψ ∗ λ to each fermion annihilation


and creation operator aλ and a∗λ respectively. Having done that one needs to
assume that the Grassmann variables anticommute with the fermion creation
or annihilation operators. Namely,
{ψλ , a†λ0 } = {ψλ , aλ } = 0 (B.60)
{ψλ∗ , a†λ0 } = {ψλ∗ , aλ } = 0. (B.61)
In addition the adjoint operation is related to the conjugation of the Grassmann
variables as follows:
(ψλ aλ0 )† = a†λ0 ψλ∗ (B.62)
(ψλ a†λ0 )† = aλ0 ψλ∗ (B.63)
(ψλ∗ aλ0 )† = a†λ0 ψλ (B.64)
(ψλ∗ a†λ0 )† = aλ0 ψλ . (B.65)
It is straightforward to show that the state defined as follows
P †
|ψi = e− λ ψλ aλ |0i (B.66)
which has the same form as the boson coherent state. In fact the introduction
of the Grassmann algebra was made to achieve this goal. Since ψλ a†λ commutes
with ψλ0 a†λ0 we can write
Y
|ψi = (1 − ψµ a†µ )|0i (B.67)
µ

and we can prove that (B.66) is an eigenstate of the fermion annihilation oper-
ator by considering the action of aλ on such state
(1 − ψµ a†µ )aλ (1 − ψλ a†λ )|0i.
Y
aλ |ψi = (B.68)
µ6=λ

And since the following statements are true


aλ (1 − ψλ a†λ )|0i = aλ |0i + ψλ aλ a†λ |0i = ψλ |0i = ψλ (1 − ψλ a†λ )|0i,
we find that
aλ |ψi = ψλ |ψi. (B.69)
One can show in a similar way that
P ∗
hψ| = h0|e λ ψλ aλ . (B.70)
In addition, with straightforward application of the properties of Grassmann
variables and Grassmann derivatives one can show that
hψ|a†λ = hψ|ψλ∗

a† |ψi = − |ψi
∂ψλ

hψ|aλ = hψ|. (B.71)
∂ψ ∗
326 APPENDIX B. COHERENT STATES

The overlap of two fermion coherent states has the same form as the one for
boson coherent states, P ∗
hψ|ψ 0 i = e λ ψλ ψλ (B.72)
This can be shown in a straightforward way by writing the left and right coherent
states as a product (B.67) and by evaluating the vacuum to vacuum expectation
values of the terms for each set of single-fermion quantum numbers λ. Finally

one needs to use that (1 + ψλ∗ ψλ0 ) = eψλ ψλ0 .
Again as in the case of boson coherent states the completeness relation is
given as: Z Y P ∗
dψλ∗ dψλ e− λ ψλ ψλ |ψihψ| = 1. (B.73)
λ

In order to show that, we shall consider two fermion determinants

|γ1 , γ2 , ..., γn i = a†γ1 a†γ2 ...a†γn |0i (B.74)


|α1 , α2 , ..., αm i = a†α1 a†α2 ...a†αm |0i, (B.75)

and we shall compute the matrix elements with these two states of both sides
of Eq. (B.73). Thus, we need to evaluate

(1 + ψλ a†λ )|0i
Y
hγ1 , γ2 , ..., γn |ψi = h0|aγn aγn−1 ...aγ1
λ
= † †
h0|aγn aγn−1 ...aγ1 (ψγ1 aγ1 )(ψγ2 aγ2 )...(ψγn a†γn )|0i
= (−1)n ψγn ψγn−1 ...ψγ1 . (B.76)

Similarly
hψ|α1 , α2 , ..., αm i = (−1)m ψα∗ 1 ψα∗ 2 ...ψα∗ m . (B.77)
Therefore we find that the left-hand-side of (B.73) when evaluated in between
the states (B.74,B.75) is given by
Z Y
 ∗
m+n
dψλ dψλ (1 − ψλ∗ ψλ ) ψγn ψγn−1 ...ψγ1 ψα∗ 1 ψα∗ 2 ...ψα∗ m .

(−1) (B.78)
λ

Let us ask when this is not zero. For this lets not worry at first for the signs
that we need to keep track when permuting Grassmann variables. There are
the following two possibilities for each different single-fermion set of quantum
numbers λ.
Z
dψλ∗ dψλ (1 − ψλ∗ ψλ ) = 1 (B.79)
Z
dψλ∗ dψλ (1 − ψλ∗ ψλ )ψλ∗ ψλ = 1 (B.80)

If there is only ψλ or only ψλ∗ the integral is zero and if there is ψ n or (ψ ∗ )n


with n > 1 is also zero. This implies that if a ψλ∗ exists in the product, then ψλ
must also exist. This means that the states |γ1 , γ2 , ..., γn i and |α1 , α2 , ...αm i are
B.3. FERMION COHERENT STATES 327

identical, thus, m = n and αi = γi , i = 1, 2, ..., n. Thus, in order to separate the


integrals one needs to simultaneous move pairs ψλ∗ ψλ . Namely, one moves first
ψγ∗1 ψγ1 (α1 = γ1 ), next one moves ψγ∗2 ψγ2 (which now become consecutive), etc.
Moving such pairs of Grassmann variables produces no minus signs which we
would need to keep track. Thus, when the two states (B.74,B.75) are identical we
find 1 otherwise zero, which means that we have evaluated the matrix elements
of the unit operator.
This property of the fermion coherent states demonstrates the utility of the
Grassmann variables. Namely, even though the coherent states are not states in
the physical Fock space, their algebra allows construction of the unit operator
in the physical Fock space. This enables us to calculate the trace of an operator
Ô as follows. We start from the standard definition of trace, using a basis |ni
of the physical Fock space:
P RQ Q  ∗  − P ψ∗ ψλ P
trÔ = n hn| Ô|ni = λ λ dψ λ dψ λ e λ λ
n hn|ψihψ|Ô|ni
 − P ∗
RQ  ∗ ψλ ψλ P
= λ dψλ dψλ e n h−ψ|Ô|nihn|ψi
λ

RQ  ∗  − P ψ∗ ψλ
= λ dψλ dψλ e
λ λ h−ψ|Ôψi. (B.81)

Notice that when we permuted the hn|ψi and hψ|Ô|ni we picked up a minus
sign in the coherent state. Namely, the following is true:

hf |ψihψ|gi = h−ψ|gihf |ψi. (B.82)

Thus, the trace of any operator in the fermion Fock space using coherent states
is given by Eq. (B.81), thus, the Grassmann variables as defined in the previous
section are useful because one can cast the trace and as we shall see the path
integral over fermion coherent states in the same form as for bosons. Of course
the price that we have to pay is that we are dealing not with complex numbers
but with Grassmann variables.
In the fermion coherent basis, any state |Ψi is expressed as
Z Y
 P ∗
dψλ dψλ e− λ ψλ ψλ Ψ(ψ ∗ )|ψi.
 ∗
|Ψi = (B.83)
λ

where Ψ(ψ ∗ ) = hψ|Ψi. In addition, we find that the representation of the


operators aλ and a†λ is the following:


hψ|aλ |Ψi = Ψ(ψ ∗ ) (B.84)
∂ψ ∗
hψ|a†λ |Ψi = ψλ∗ Ψ(ψ ∗ ). (B.85)

On the other hand matrix elements of an operator Ô(a†λ , aλ ), which is a normal


ordered form with respect to the operators a† and a, are given as follows:
P ∗
hψ|Ô(a†λ , aλ )|ψ 0 i = e λ ψλ ψλ O(ψλ∗ , ψλ0 ). (B.86)
328 APPENDIX B. COHERENT STATES

B.4 Problems
1. As we have discussed the coherent states are not physicalPstates, for example
the expectation value of the particle number operator N̂ = λ a†λ aλ is given by

hψ|N̂ |ψi X ∗
= ψλ ψλ (B.87)
hψ|ψi
λ

which has no physical interpretation. However, this is because the coherent


states are not forming basis in a physical space. We saw how to use them to
express physical quantities. For example, take a state |ni from the physical
Fock space, such as (B.74) and using coherent states express the expectation
value hn|N̂ |ni in terms of integrals over coherent states. Show that one finds
the same particle number as in the state |ni.
2. Show that the following behaves as a δ function among functions of
Grassmann variables: Z
0
δ(ψ, ψ 0 ) = dξe−ξ(ψ−ψ ) , (B.88)

Namely, show that given any function of Grassmann variables: f (ψ)


Z
dψ 0 δ(ψ, ψ 0 )f (ψ 0 ) = f (ψ). (B.89)

3. For Grassmann variables show that:

hf |ψihψ|gi = h−ψ|gihf |ψi. (B.90)

You might also like