PH 87

THE S-MATRIX FORMULATION OF QUANTUM
STATISTICAL MECHANICS, WITH APPLICATION TO

COLD QUANTUM GAS
A Dissertation
Presented to the Faculty of the Graduate School
of Cornell University
in Partial Fulfillment of the Requirements for the Degree of

Doctor of Philosophy
by
Pye Ton How
August 2011
c 2011 Pye Ton How

ALL RIGHTS RESERVED
THE S-MATRIX FORMULATION OF QUANTUM STATISTICAL
MECHANICS, WITH APPLICATION TO COLD QUANTUM GAS
Pye Ton How, Ph.D.
Cornell University 2011
A novel formalism of quantum statistical mechanics, based on the zero-temperature

S-matrix of the quantum system, is presented in this thesis. In our new formalism,
the lowest order approximation (“two-body approximation”) corresponds to the ex-

act resummation of all binary collision terms, and can be expressed as an integral
equation reminiscent of the thermodynamic Bethe Ansatz (TBA). Two applica-
tions of this formalism are explored: the critical point of a weakly-interacting Bose
gas in two dimensions, and the scaling behavior of quantum gases at the unitary
limit in two and three spatial dimensions. We found that a weakly-interacting 2D
Bose gas undergoes a superfluid transition at Tc ≈ 2πn/[m log(2π/mg)], where n
is the number density, m the mass of a particle, and g the coupling. In the unitary
limit where the coupling g diverges, the two-body kernel of our integral equation
has simple forms in both two and three spatial dimensions, and we were able to
solve the integral equation numerically. Various scaling functions in the unitary
limit are defined (as functions of µ/T ) and computed from the numerical solutions.
For bosons in three spatial dimensions, we present evidence that the gas undergoes
a strongly interacting version of Bose-Einstein condensation at nλ3T ≈ 1.3, where
n is the number density and λT is the thermal wavelength. Finally, we look at
the ratio of shear viscosity to entropy density η/s of the unitary quantum gas,
which has a conjectured lower bound of η/s ≥ ~/4πkb based on the AdS/CFT
correspondence of a strongly coupled Yang-Mills theory.
BIOGRAPHICAL SKETCH
Pye Ton How was born on February 16, 1981 in Pulau Pinang, Malaysia. His
family moved to Taiwan a few months later, and he grew up and went to school
there.
The family moved back to Malaysia once again when he was sixteen. After two
years of high school in Malaysia, he went on to attend the University of Cambridge
in the United Kingdom. He was a member of Selwyn College, elected college scholar
twice, and a fellow of the Cambridge Commonwealth Society.
He obtained the Bachelor of Art (Honor first class) and the Master of Natural
Sciences from University of Cambridge in June 2002, but decided to stay one
more year and attend what is known as the Part III of the Mathematical Tripos.
He received the Certificate of Advanced Studies in Mathematics (CASM) with
distinction in June 2003.

After leaving Cambridge, he became a research associate for one year under
Professor Fun Hoong Kun in the School of Physics, Universiti Sains Malaysia. He
joined the Physics graduate program of Cornell University in August 2004.
iii
ACKNOWLEDGEMENTS
First and foremost, I owe my deepest gratitude to my advisor Prof. André LeClair
for patiently guiding me through all these years of grad school.
I would also like to thank Mr. John Miner and Prof. Jeevak Parpia. If not for
your great patient, understanding, and continual support, I would not have made
it.
I thank Prof. Erich Mueller for his sometimes harsh, but always to-the-point
and extremely useful criticism to my work.
To all of the undergraduate students I have taught as a TA: thank you all for
putting up with my badly designed quizzes and those ill-prepared sections. The
positive feedbacks I occasionally received really made me happy.
And finally, to all Physics graduate students: I really enjoyed my time here,
and it’s all thanks to you guys.
iv
TABLE OF CONTENTS
Biographical Sketch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv
Table of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
List of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
1 Introduction 1
1.1 Non-perturbative method in quantum field theories . . . . . . . . . 1
1.2 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Phase transition of a Bose gas in two dimensions . . . . . . . . . . . 6
1.4 Unitary gas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.5 The road map . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.6 Notations and conventions . . . . . . . . . . . . . . . . . . . . . . . 10
2 Quantum statistical mechanics: the S-matrix approach 12

2.1 Partition function in terms of the S-matrix . . . . . . . . . . . . . . 12
2.2 Cluster decomposition . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3 Diagrammatic description . . . . . . . . . . . . . . . . . . . . . . . 18
3 TBA-like integral equation 24

3.1 Legendre transformation . . . . . . . . . . . . . . . . . . . . . . . . 25
3.2 Integral equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.3 Two-body approximation . . . . . . . . . . . . . . . . . . . . . . . . 29
3.4 Two-body kernel . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.5 Extension to multi-component gases . . . . . . . . . . . . . . . . . . 35
4 Field theories for quantum gases 38

4.1 Actions and Feynman rules . . . . . . . . . . . . . . . . . . . . . . 38
4.2 Two-body scattering in vacuum . . . . . . . . . . . . . . . . . . . . 40
4.2.1 d = 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.2.2 d = 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
5 Critical point of the two-dimensional Bose gas 43
6 Renormalization group and the unitary limit 47

6.1 The 3d case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
6.2 The case of a generic d 6= 2 . . . . . . . . . . . . . . . . . . . . . . . 50
6.3 d = 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
6.4 Thermodynamic scaling functions at the quantum critical point . . 56
7 The two-dimensional quantum gas at unitary limit 61

7.1 Two-body kernal, integral equation, and scaling functions . . . . . . 61
7.2 Attractive fermions . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
7.3 Attractive bosons . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
7.4 Repulsive fermions and bosons . . . . . . . . . . . . . . . . . . . . . 69
v
8 The three-dimensional quantum gas at unitary limit 75
8.1 Two-body kernal and integral equation . . . . . . . . . . . . . . . . 75
8.2 The scaling functions . . . . . . . . . . . . . . . . . . . . . . . . . . 76
8.3 Analysis of fermions . . . . . . . . . . . . . . . . . . . . . . . . . . 78
8.4 Analysis of bosons . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
9 Entropy-to-Viscosity Ratio of Quantum Gases 88

9.1 2d . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
9.2 3d . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
10 Summary and Conclusion 96
A S-matrix and Resolvant 98
Bibliography 103
vi
LIST OF FIGURES
1.1 The Feynman diagrams related to Thomas-Fermi screening. . . . . 3

1.2 The lowest order correction to self energy of a scalar φ4 theory.
The next order correction comes from imposing a self-consistent
condition on this diagram. . . . . . . . . . . . . . . . . . . . . . . . 3
2.1 Examples of contributions to two- and three-particle terms. . . . . 21

2.2 Diagrams contributing to w4 . . . . . . . . . . . . . . . . . . . . . 22
2.3 Ring diagrams. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.1 Graphical representation of the integral equation (3.15). The dou-

ble line represents fe, and the single line represents f0 . . . . . . . . 28
3.2 Foam diagrams. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
4.1 Ladder diagram resummation of the two-body scattering amplitude. 41
5.1 Left and right hand sides of eq. (5.5) as a function of βµc at two
values of the infra-red cut-off ǫ0 , one being the critical value ǫc . . . 46
6.1 Renormalization group flow for space dimension d. The arrow in-
dicates flow to the IR. . . . . . . . . . . . . . . . . . . . . . . . . . 52
7.1 Attractive fermions in 2d: density as a function of µ/T . . . . . . . 63

7.2 Attractive fermion in 2d: filling fraction f at two different values
of µ/T . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
7.3 Attractive fermion in 2d: the scaling function c as a function of µ/T . 64
7.4 Attractive fermion in 2d: the entropy density s divided by mµ as
a function of T /µ. . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
7.5 Attractive boson in 2d: density as a function of µ/T . The origin of
the axes is at (log zc , 0). . . . . . . . . . . . . . . . . . . . . . . . . 67
7.6 Attractive boson in 2d: the filling fraction f for µ/T =
−2, −1.5, −1.08 = log zc . . . . . . . . . . . . . . . . . . . . . . . . 67
7.7 Attractive boson in 2d: the scaling function c as a function of µ/T .
The origin of the axes is at (log .33, 0). . . . . . . . . . . . . . . . . 68
7.8 Attractive bosons in 2d: the energy per particle scaling function ξ˜
as a function of µ/T . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
7.9 Attractive boson in 2d: the energy per particle scaling function ξ
as a function of T /µ. . . . . . . . . . . . . . . . . . . . . . . . . . . 69
7.10 Repulsive fermion in 2d: density as a function of µ/T . . . . . . . . 70
7.11 Repulsive fermion in 2d: filling fraction f for µ/T = −3, −2, −0.55.
The top curve corresponds to the critical density in eq. (7.27). . . 70
7.12 Repulsive fermion in 2d: the scaling function c as a function of µ/T . 71
7.13 Repulsive fermion in 2d: the energy per particle scaling function ξ˜
as a function of µ/T . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
vii
7.14 Density of repulsive bosons (bottom curve) verses fermions as a
function of µ/T . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
7.15 Repulsive boson in 2d: the ratio η/s as a function of µ/T . . . . . . 74
8.1 Attractive fermion in 3d: c(x) and its equivalent for a free theory
as a function of x = µ/T . . . . . . . . . . . . . . . . . . . . . . . . 79
8.2 Attractive fermion in 3d: q(x) and its equivalent for a free theory
as a function of x = µ/T . . . . . . . . . . . . . . . . . . . . . . . . 79
8.3 Attractive fermion in 3d: the occupation numbers as a function of
κ for x = 5, 10, 15. . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
8.4 Attractive fermion in 3d: entropy per particle as a function of x. . 81
8.5 Attractive fermion in 3d: specific heat per particle as a function of x. 81
8.6 Attractive fermion in 3d: energy per particle normalized to ǫF as a
function of T /TF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
8.7 Attractive fermion in 3d: entopy per particle as a function of T /TF . 82
8.8 Attractive fermion in 3d: chemical potential normalized to ǫF as a
function of T /TF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
8.9 Attractive boson in 3d: the free-energy scaling function c as a func-
tion of µ/T compared to the ideal gas case. . . . . . . . . . . . . . 84
8.10 Attractive boson in 3d: the pseudo-energy ǫ at k = 0 as a function
of x = µ/T . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
8.11 Attractive boson in 3d: the occupation number f (κ) for x = −1.275
and xc = −1.2741. . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
8.12 Attractive boson in 3d: the compressibility κ as a function of µ/T . 85
8.13 Attractive boson in 3d: the entropy per particle as a function of
T /Tc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
8.14 Attractive boson in 3d: the specific heat per particle as a function
of T /Tc . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
9.1 The ratio η/s as a function of µ/T . The horizontal line is 1/4π. . . 89
9.2 The ratio η/s as a function of T /TF for attractive fermions. . . . . 90
9.3 The ratio η/s as a function of µ/T as µ/T gets very large.. The
horizontal line is 1/4π. . . . . . . . . . . . . . . . . . . . . . . . . . 91
9.4 The ratio η/s as a function of µ/T in the limit of very low temper-
atures (top curve). The horizontal line is 1/4π. The bottom curve
is the approximation eq. (9.12). . . . . . . . . . . . . . . . . . . . . 92
9.5 The viscosity to entropy-density ratio as a function of T /TF for
fermions. The horizontal line is 1/4π. . . . . . . . . . . . . . . . . 94
9.6 The viscosity to entropy-density ratio as a function of T /Tc for
bosons. The horizontal line is 1/4π. . . . . . . . . . . . . . . . . . 95
viii
CHAPTER 1
INTRODUCTION
1.1 Non-perturbative method in quantum field theories
With the exception of the exactly solvable models, the only known method of quan-
tizing a field theory is through perturbative expansion [37, 11]. Given a classical
filed theory action S, it must first be separated into two parts:
S = S0 + S1 , (1.1)
where S0 is quadratic in the field degrees of freedom and admits (classically) plane-
wave solutions, while S1 is treated as a small perturbation. This way of quantizing
a field theory has a very intuitive interpretation: the plane-wave solutions of S0
describe freely moving particles, and S1 adds to it interaction between particles.
Then the perturbation theory is an expansion in the powers of the interaction

parameter.
The initial application of this perturbative quatization scheme to the quantum

electrodynamics (QED) was a spectacular success [37], thanks to the existence of a
small parameter, the fine structure constant α ≈ 1/137. However, it soon became
evident that not all problems can be treated within this scheme.
It is conceivable that a problem may be without a truly small parameter that
allows for an expansion. There are also cases where the desired result has a non-
analytic dependent in the interaction parameter, rendering any truncated power
series qualitatively incorrect.
To tackle such problems theoretically, one may invent an alternative small
1
parameter and derive an expansion around it. The semi-classical expansion in
powers of ~, the large-N expansion, and the ǫ expansion are a few examples. On
the other hand, one may attack the problem more directly by explicitly resumming
an infinite subset of terms in the perturbation series.
A well-known example of such resummation, though not necessarily known
to all in field-theoretic form, is the Thomas-Fermi screening of the electrostatic

potential in a Fermi liquid [39].
The field theory for a Fermi liquid consists of electron and hole excitations
near the Fermi surface, and the 1/r electrostatic potential as a perturbation. In
momentum space, we may use solid lines to represent electrons and holes, and
dashed line to represent Ũ(k), the Fourier transform of the potential. See Fig.
1.1(a).
The lowest-order correction to the interaction potential is represented by the

Feynman diagram 1.1(b). The loop integral, however, diverges in this case. Thus
it becomes necessary to sum over diagrams of the form Fig. 1.1(c) to all order.
The result is simple:
∞
X Z
Ũef f (k) = Ũ (k) Ũ (k) (loop)
n=0
(1.2)
Ũ(k)
= R ,
1 − Ũ(k) (loop)
and is vanishingly small due to the diverging denominator. As one can see, the
interaction potential Ũ ends up in the denominator of the full expression. This
type of behavior can’t be captured by any truncated expansion.
Another example of non-perturbative phenomena in field theory is the self-
energy of a scalar particle at finite temperature in four dimensional spacetime

[27]. Assuming that the scalar field theory has a quartic interaction as allowed by
2
Fermion propagtor
R
(loop) =
Fourier transform of the potential Ũ (k)
(a) Feynman rule for a Fermi liquid. (b) The one-loop correction to in-
teraction.
P∞
n=0 1 2 ... n
(c) The diagrams that needs to be re-summed.
Figure 1.1: The Feynman diagrams related to Thomas-Fermi screening.
Figure 1.2: The lowest order correction to self energy of a scalar φ4 theory.
The next order correction comes from imposing a self-consistent
condition on this diagram.
symmetry, with coupling strength g, the leading order correction is given by the
simple bubble diagram 1.2, and is linear in g. However, the sub-leading term is
known to be of order g 3/2 . This result is obtained by imposing a self-consistent
condition on the self-energy. Clearly no truncated expansion in powers of g can

yield such a result.
In this thesis, we will develop an alternative representation of the free energy

of an interacting quantum field theory, based on its zero-temperature S-matrix.
Each term in our expansion contains an infinite number of Feynman diagrams.
3
Therefore the lowest order, or “two-body” approximation already corresponds to
an infinite resummation of the Feynman perturbative series. We will then apply our
formalism to the problem of unitary quantum gas, where the interaction parameter
diverges, and a straightforward perturbative expansion is doomed to failure.
1.2 Motivation
The motivation of our new formalism is two-folded. First, we seek a higher-

dimensional generalization of the thermodynamics of integrable models in 1 + 1
dimensions; second, we wish to achieve the separation of thermal and quantum
effects in a given problem, much like the classical kinetic theory of gas.
The Thermodynamic Bethe Ansatz (TBA) was introduced by Yang and

Yang[51, 25]. It builds on the Bethe Ansatz solution of a one-dimensional quantum
spin chain, due to Bethe himself, and allows the exact computation of thermody-
namic properties of some integrable models in 1 + 1 dimensions.
The “prototype” TBA was formulated for a gas of non-relativistic bosons, in-
teracting with a δ-function potential. The solution is contained in the Yang-Yang
equation:
Z ∞
2 1
ǫ(k) = k − µ − K(k, p) log 1 − e−βǫ(p) . (1.3)
2πβ −∞
Here ǫ is the so-called pseudo-energy, and can be considered as the single-
particle energy in the background of all the other particles interacting with it.
The kernel K(k, p) is the derivative of the scattering phase when two particles,
with momenta k and p respectively, scattered in vacuum.
TBA relies on the integrable structure of the theory, and does not generalize to
4
higher dimensions. Thacker showed that the factorizability of the S-matrix directly
leads to the TBA [48, 49]. But, unlike the many models in 1+1 dimensions without
known Lagrangian formulations, the interacting boson gas has a clear generaliza-
tion to higher dimensions. It is then natural to ask: is there an approximation
scheme in the higher dimensions, that is a parallel, or a “generalization” of the

TBA in 1 + 1 dimensions?
Dashen et. al. [8] gives the partition function of a quantum field theory model
in terms of its S-matrix. Based on their result, it should be possible to organize
an expansion that mimics the TBA. However, the S-matrix of a non-integrable
theory does not factorize, and the expansion quickly becomes complicated beyond
the lowest order. Nevertheless, for a dilute gas, the resummation of all two-body
terms, corresponds to an exact treatment of binary collisions, may serve as a useful
approximation intrinsically different from other methods.
Building on the result of Dashen, we derived an integral equation for the pseudo-
energy of a quantum gas:
This integral equation is, of course, only an approximation, since the higher-
dimensional model is not integrable. But we hope that this alternative treatment
would grant us insights otherwise inaccessible via simple perturbation theory.
In the view of the kinetic theory, the microscopic degrees of freedom interact and
evolve according to their microscopic, zero-temperature equations of motion. They
do not directly know about the temperature of the system; rather the temperature
comes in through the distribution of the background particles.
In our integral equation, all “quantumness” of the problem is confined within
the kernel, and all thing thermal is kept in the filling fraction f , reflecting the
5
background distribution. In a sense, the kernel can be understood as the analogy
of the collision integral in the classical kinetic theory. Having a different type of
interaction amounts to plugging in a different kernel to the same integral equation,
with the same temperature dependence. This separation of thermal and quantum
parts is in stark contrast with the usual Matsubara or Keldysh formalism, where
the temperature is directly built into the quantum propagators.
Thus, like the kinetic theory of gas, we also expect the integral equation to
hold in the dilute limit, where it forms a quantum analogy of binary collision
approximation.
1.3 Phase transition of a Bose gas in two dimensions
In two spatial dimensions, the famous Mermin-Wagner-Hohenberg theorem states

that a true long-range order is not possible [31, 17]. As a consequence, a non-
interacting Bose gas does not undergo Bose-Einstein condensation (BEC) in two
dimensions. This result can also be shown directly as followed. A non-interacting
Bose gas undergoes BEC at zero chemical potential. In d spatial dimensions, the
critical density is
Z
dd k 1
nc = d k2
(2π) e 2m − 1
β
d/2 (1.4)
2m
= ζ(d/2)
4πβ
where ζ is the Riemann zeta-function.
When d = 3, ζ(3/2) is finite, and the gas undergoes BEC at a finite critical
density. On the other hand, ζ(2) is divergent when d = 2, meaning the normal
state of a two-dimensional Bose gas can support an arbitrarily large density, and
6
there exists no critical point.
Nevertheless, a two-dimensional system can undergo a Berezinsky-Kosterlitz-

Thouless (BKT) transition, and its correlation function decays as power-law rather
than exponentially [41, 12, 42]. The computation of this critical point in Chapter
5 will be the first application of our integral equation formalism.
1.4 Unitary gas
The method to optically trap neutral atoms and to cool the trapped atomic gas
to a temperature in the nano-Kelvin range was developed in the 1990s [40]. This
development opened up the entire exciting new field of ultra-cold quantum gas.
Ultra-cold quantum gas attracts considerable attention for various reasons.

From the stand point of physics of strongly interacting system, it is an ideal test-
ing ground of new theoretical ideas due to its controllability. Across a Feshbach
resonance, the s-wave scattering length a between two atoms can in principle be
tuned to any arbitrary value between negative infinity and infinity by varying the
external magnetic field [5, 23, 6, 35]. In three spatial dimensions, the atoms in
such a gas can be modeled as integrating via a δ-function pseudo-potential whose
coefficient is the scattering length [11]. The cold quantum gas therefore presents
us an experimental realization of a system, whose interaction strength is can be
arbitrarily and easily tuned to any desired value!
It has long been postulated [10, 34] that, in three spatial dimensions, the
Bardeen-Cooper-Schriefer (BCS) superconductor phase and the Bose-Einstein con-
densation (BEC) phase are connected smoothly as the interaction varies from re-
7
pulsive to attractive, with a crossover at infinite scattering length, the so-called
unitary limit. This crossover has recently been realized experimentally in cold
quantum gas, thanks to the Feshbach resonance technique [1, 9, 14, 35].
With a diverging s-wave scattering length, there is no length scale in the low-
energy theory of a unitary quantum gas. The low-energy properties of such a
unitary gas is thus believed to be scale-invariant and universal [33]. On the other
hand, the absence of a small parameter makes theoretical treatments difficult.
We believe that our integral equation approach is well-suited for exploring the
unitary limit. It will be shown that the two-body kernel in our approach remains
well-behaved as the scattering length diverges– in fact, the structure of the kernel
is drastically simplified in this limit.
The cold quantum gas is a metastable state: the stable equilibrium state is the
crystalline phase. However, in order to form a crystal, bonds must form between
atoms as the first step. Yet the formation of a bounded pair requires at least three
atoms in the process, one to take away the excess energy as dictated by energy
and momentum conservation. Since binary collision between atoms is by far the
most frequent events, the quantum gas can “equilibrate” while staying as a gas.
Collision processes involving more than two atoms contribute to particle loss from
the gas, and the gas would eventually solidify.
Given that the gas reaches its (pseudo) equilibrium state through binary col-
lisions, we believe that our integral equation approach, which treats two-body
processes exactly, can be a useful description for such quantum gas, even in the
unitary limit.
8
1.5 The road map
This thesis consists of works previously published by the author in three separate
papers [19, 20, 21].
The layout of this thesis is as followed. In Chapter 2, we shall developed the
so-called S-matrix formulation of quantum statistical mechanics. A diagrammatic

expansion will be given so that one may write down the partition function or the
free energy density of an interacting system to arbitrary order of the scattering
kernel.
In Chapter 3, the formulation will be developed further, and we shall specialize

in the two-body approximation, which leads to an integral equation similar to the
Yang-Yang equation of TBA (1.3).
In Chapter 4, we turn our attention to the field theoretic model of quantum

gas. In particular, we shall derive the explicit forms of the two-body kernel in
various spatial dimensions.
We take a break from the development of the theoretical framework in Chapter

5, and applies the formalism to the weakly interacting Bose gas in three dimensions.
The correction to the critical density of the Bose-Einstein condensation due to
interaction is computed.
The field-theoretic analysis resumes in Chapter 6. There we will attempt to

interpret the unitary limit, in the context of the interaction kernel introduced in
Chapter 2, in terms of renormalization group flow. The scaling functions at the
unitary limit are also defined in this chapter.
The results of our numerical calculation at the unitary limit, in two and three
9
spatial dimensions, are detailed in Chapter 7 and Chapter 8 respectively.
In Chapter 9, the shear viscosity to entropy ratio of unitary quantum gas is

computed using our integral equation, and compared with the conjectured lower
bound in certain supersymmetric gauge theory. We found that our result is in
good agreement with recent experimental finding by Cao et. al [4].
1.6 Notations and conventions
Throughout this thesis, we will choose our units such that kB = ~ = 1 unless
explicitly stated otherwise.
We assume that any system discussed has a (second-quantized) Hamiltonian

H = H0 + V , where H0 is the “free” part and V is an interaction term. We further
assume that the spectrum of H0 has a free-particle interpretation.
We denote the N-particle eigenstate of H0 , where the individual particles have

momentum k1 , · · · kN respectively, by
|k1 , · · · kN i. (1.5)
Throughout the thesis d shall denote the number of spatial dimensions. Also
we define as a shorthand integration measure in the Fourier space:
dω
(dω) = (1.6)
2π
dd k
(dk) = (1.7)
(2π)d
It will be necessary at times to keep track of the particle statistics. We define
10
the quantity s by: 


 +1 for Bosons
s= (1.8)


 −1 for Fermions
Throughout this thesis we shall primarily consider quantum gases with a single
species of particles. However, a fermionic degree of freedom necessarily has two
components due to its spin, and this may affect the symmetry factor of a Feynman
diagram in the standard perturbation theory. We define the numerical factor α as
a book-keeping tool: 


 1 for Bosons
α= (1.9)


 1
for Fermions.
2
11
CHAPTER 2
QUANTUM STATISTICAL MECHANICS: THE S-MATRIX
APPROACH
2.1 Partition function in terms of the S-matrix
Given a quantum mechanical system with Hamiltonian H, its partition function
is defined as
Z(µ, β) = Tr e−β(H−µN ) , (2.1)
where β = 1/T is the inverse temperature, and µ is the chemical potential.
Following the work of Dashen [8], we shall now recast Z into a different form:
Z = Tr e−β(H−µN )
Z
= dEe−βE Tr δ (E − (H − µN)) (2.2)
Z
−βE 1 1 1
= dEe Tr −
2πi E − (H − µN) − i0+ E − (H − µN) + i0+
Here we have used the identity limη→0+ Im (iπ/(x + iη)) = δ(x) to formally
express the δ-function in a more tractable form. We shall now relate this expression
to the S-matrix in the scattering theory.
We follow the usual prescription of splitting the Hamiltonian H into the “free”
part H0 and the interaction part V . Our convention of the S-matrix is
S(E) = 1 + 2πiδ (E − H0 ) T (E). (2.3)
Also we define:
1
G(E) = lim+
η→0 E − H + iη
(2.4)
1
G0 (E) = lim+ .
η→0 E − H0 + iη
12
The S-matrix is then related to G and G0 by
S = G∗0 (G∗ )−1 GG−1

0 . (2.5)
See appendix for the proof.
Now we claim that Tr S −1 ∂E S = − Tr[(G − G0 ) − (G ∗ −G0 ∗)]. Using the fact

that ∂E G = −G2 and ∂E (G−1 ) = 1, it is easily shown:

Tr S −1 ∂E S = Tr G0 G−1 G∗ (G∗0 )−1 ∂E G∗0 (G∗ )−1 GG−1
0

= Tr (G∗0 )−1 ∂E G∗0 + G∗ ∂E (G∗ )−1 + G−1 ∂E G + G0 ∂E (G0 )−1 (2.6)
= − Tr [(G − G0 ) − (G∗ − G∗0 )] .
Inserting this result into (2.2), we see that the partition function becomes
Z
1
Z= dEe−βE Tr(S −1 (E)∂E S(E)) − Tr(G0 (E) − G∗0 (E))
2πi
Z (2.7)
1 −βE −1
= Z0 + dEe Tr(S (E)∂E S(E)),
2πi
where Z0 = Tr e−βH0 is the partition function of the free theory.
Using the cyclic property of the trace, we may write Tr(S −1 (E)∂E S(E)) =
∂E Tr log S(E). Furthermore we can integrate by part and move the derivative ∂E
onto e−βE . So we arrived at the expression for partition function that we will be
working with:
Z
dE −βE
Z = Z0 − iβ e Tr log S(E) (2.8)
2π
Since log S(E) ∝ δ(E − H0 ), we may factor the energy-conserving δ-function

out, and define the operator W by
Im log S(E) ≡ 2πδ(E − H0 )W (E). (2.9)
13
Insert this definition into (2.8), and integrate by part, one obtains:
Z
Z= dE e−βE Tr [δ(E − H0 ) (W (E)β + 1)] . (2.10)
For simplicity of the notation, we further define
Û = W β + 1 (2.11)
2.2 Cluster decomposition
One can take the principle of cluster decomposition as an axiom of any well-defined
quantum field theory [50]. It is a necessary condition for the field theory being
local.
Take two electrons scattering off two atomic nuclei for example. Formally the
incoming quantum state is a two-electron state, and the S-matrix connects this
state to a two-electron outgoing state.
However, suppose now we are talking about two identical copies of experiments,
each involving a single electron scattering off a single nucleus, being performed at
locations miles away from each other. Certainly we do not expect any correlation
between the two experiments. In this case, the (formal) two-body S-matrix must
be just the tensor product of two one-body S-matrices.
On the other hand, when the two sets of experiment are brought closed to
each other, quantum interference must take place, and the S-matrix is no longer a
product of two independent pieces.
14
Mathematically, we write:
he′1 , e′2 |S|e1 , e2 i = he′1 |S|e1 ihe′2 |S|e2 ic + she′1 |S|e2 ihe′2 |S|e1 ic + he′1 , e′2 |S|e1 , e2 ic .
(2.12)
The subscript c denotes “connected”, meaning the piece cannot be further decom-
posed into a tensor product of smaller matrices. The quantity s arises from particle
statistics: s = +1 for bosons, and s = −1 for fermions.
In our previous example, for the two experiments performed far away to not
affect each other, the connected two-body S-matrix must vanish as the distance
between the two incoming electrons increases. This is in fact a general property of
any connected matrix elements of a physical observable from a local field theory.
It is therefore not surprising that, in the present context, the clustering property
of the S-matrix ensures the extensivity of the free energy, since the later is a
consequence of the locality.
For any operator X, the cluster decomposition can be expressed as follows:
X
hΨ′ |X|Ψi = sp hΨ′1 |X|Ψ1 ic hΨ′2 |X|Ψ2ic hΨ′3 |X|Ψ3ic · · · (2.13)
partitions
where the sum is over partitions of the state |Ψi into clusters |Ψ1 i, |Ψ2i, ..... (The
number of particles in |Ψi i and |Ψ′i i is not necessarily the same.) The above
formula essentially defines what is meant by the connected matrix elements hXic .
In order for the free energy to be extensive, Û must cluster properly, as we

now describe. Introducing the short-hand notation |12 · · ·i = |k1 k2 · · ·i, for the
15
operator Û the above equation reads
h1′ |Û|1i = h1′ |Û|1ic
h1′ 2′ |Û|12i = h1′ 2′ |Û |12ic + h1′ |Û |1ic h2′ |Û|2ic + sh2′ |Û |1ic h1′ |Û|2ic
h1′ 2′ 3′ |Û|123i = h1′ 2′ 3′ |Û|123ic (2.14)

h i
+ h1′ |Û|1ic h2′ 3′ |Û|23ic + permutations
9
h i
′ ′ ′
+ h1 |Û|1ic h2 |Û |2ic h3 |Û|3ic + permutations
6
etc. The subscript [∗]9 denotes the number of terms within the bracket, so that
for three particles there is a total of 1 + 9 + 6 = 16 terms on the right hand side
of the above equation. The connected matrix elements are characterized by being
proportional to a single overall δ-function.
When {k′ } = {k}, the above cluster expansion leads to another specialized type
of cluster expansion that is suited to the computation of the partition function.
From eq. (2.14) we define ŵN (k1 , ..., kN ) as factors that cannot be written as a
product of separate functions of disjoint subsets of the ki :
h1|Û|1i = h1|Û|1ic = ŵ1 (k1 )
h12|Û|12i = ŵ2 (k1 , k2 ) + ŵ1 (k1 )ŵ1 (k2 )
h123|Û|123i = ŵ3 (k1 , k2 , k3 ) + [ŵ2 (k2 , k3 )ŵ1 (k1 ) + ŵ2 (k1 , k3 )ŵ1 (k2 ) + ŵ2 (k1 , k2 )ŵ1 (k3 )]
+ ŵ1 (k1 )ŵ1 (k2 )ŵ1 (k3 )
(2.15)
etc.
The above definition implies that the ŵN are sums of terms involving the con-
16
nected matrix elements of Û . For instance:
ŵ2 (k1 , k2 ) = h12|Û|12ic + sh2|Û|1ic h1|Û|2ic

h i
ŵ3 (k1 , k2 , k3 ) = h123|Û |123ic + sh2|Û|1ic h13|Û|23ic + perm. (2.16)
6
h i
+ h3|Û |1ic h1|Û|2ic h2|Û|3ic + perm.
2
A combinatoric argument then shows:

X∞ Z PN
zN
log Z = (dk) · · · kN e−β i=1 ωi ŵN (k1 , ..., kN ) (2.17)
N =1
N!
where z ≡ eβµ and ωi = ωki .
All of the ŵ are proportional to an overall (2π)d δ (d) (0) = V , so let us factor
out the volume V and define:
ŵN (k1 , ..., kN ) ≡ V wN (k1 , ..., kN ) (2.18)
The free energy density is then completely determined by the functions wN :

∞ Z
1 1 X zN P
F =− log Z = − (dk1 ) · · · (dkN ) e−β i ωi wN (k1 , ..., kN ) (2.19)
βV β N =1 N!
For free particles, the only non-zero connected matrix element is the one-
particle to one-particle one: h2|Û|1i = h2|1i. This still implies ŵN is non-zero
for any N:
ŵN (1, 2, .., N) = sN −1 (N − 1)! hN|1ih1|2ih2|3i · · · hN − 1|Ni (2.20)
Note that although ŵN is a product of N δ-functions, it only has a single δ(0).
All but one k integral is saturated by δ-functions at each N and one obtains the
well-known free-particle result:

Z
s
F0 = (dk) log 1 − sze−βωk (2.21)
β
17
2.3 Diagrammatic description
In the interacting case, terms involving the one-particle factors h2|Û|1ic can be
easily summed over as in the free particle case. We assume that |ki is a stable
one-particle eigenstate of H with energy ωk :
(d)
hk′ |Û |kic = hk′ |ki = (2πi)d δk,k′ (2.22)
Consider for example the ŵ3 contribution coming from the second terms in eq.
(2.16):
Z P3
z3
log Z = ....+ (dk1 )(dk2 )(dk3 ) e−β i=1 ωi
sh2|Û|1ic h13|Û|23ic + perm. +...
3!
(2.23)
Because of the δ-function, one can do one k integral. Combined with the
primary term coming from ŵ2 one finds:
Z
z2
log Z = · · · + (dk1 )(dk2 ) e−β(ω1 +ω2 ) 1 + sz(e−βω1 + e−βω2 ) h12|Û |12ic + · · ·
2
(2.24)
It is easy to see that the terms in the higher ŵN that involve one ŵ2 and N − 1
ŵ1 ’s contribute additional terms in the parentheses of the above equation that sum
Q
up to 2i=1 (1 − sze−βωi )−1 . To summarize this result in the most convenient way,
P
define ′N as the sum over ŵN terms minus any terms involving the 1-particle
factors ŵ1 . Then
Z N
!
1 X′ 1 Y
F = F0 − (dki )f0 (ki ) wN (k1 , ..., kN ) (2.25)
β N ≥2 N! i=1
where F0 is the free contribution in eq. (2.21) and we have defined the free “filling
fraction”:
z
f0 (k) ≡ (2.26)
eβωk − sz
18
The functions wN have many contributions built out of the connected matrix
elements of Û . Define the vertex function V(k′1 , · · · , k′m ; k1 , · · · , kn ) as follows:
hk′1 , ..., k′m |Û |k1 , ..., kn ic = β(2π)d δk,k′ V(k′1 , ..., k′m ; k1 , ..., kn ) (2.27)
The vertices V have no temperature dependence and are essentially on-shell
matrix elements of log S. To lowest order in the scattering matrix T , this vertex
is just the scattering amplitude: V ≈ M.
Given our present interest in non-relativistic systems, only the matrix elements
with n = m are of concern here. We highlight this fact by attaching a subscript n
to V:
Vn (k′1 , ..., k′n ; k1 , ..., kn ) ≡ V(k′1 , ..., k′n ; k1 , ..., kn ) (2.28)
Each term in the corrections F − F0 can be represented graphically. We stress
that this diagrammatic expansion as defined here, while being analogous to the
usual Feynman diagrams, is completely unrelated to the (finite temperature) Feyn-
man diagram approach. This is clear since each vertex in principle represents an
exact zero-temperature quantity summed to all orders in perturbation theory. The
correction F − F0 is the sum of all connected “vacuum diagrams”, made of ori-

ented lines and vertices with n incoming and n outgoing lines attached, n being any
positive integer. Note that there must be at least one vertex; i.e. the circle isn’t
allowed since it is already included as F0 . A diagram is then evaluated according
to the following rules and represents a contribution to F :
1. Each line carries some momentum k. To such a line is associated a factor of

f0 (k).
19
2. To each vertex, with the incoming set and the outgoing set of momenta being
{k1 , . . . , kn } and {k′1 , . . . , k′n } respectively, associate a factor of the vertex
function (2π)d β Vn (k′1 , ..., k′n ; k1 , ..., kn ).
3. Enforce momentum conservation at each vertex.
4. Integrate over all unconstrained momenta with dd k/(2π)d .
5. Divide by the symmetry factor of the diagram, defined as the number of per-
mutations of the internal lines that do not change the topology of the graph,
including relative positions. This factor is identical to the usual symmetry
factor of a Feynman diagram.
6. For fermions, include one statistical factor s = −1 for each loop.
7. Divide the result by the volume of space V = (2π)d δ(0) and by β. Note
that this is equivalent to dividing the space-time volume V β since at finite
temperature, β is the circumference of compactified time.
To illustrate this diagrammatic expansion, we will now describe the first few
terms. There is only one diagram that has two lines, as shown in Figure 2.1. This
diagram corresponds to:

Z
1
F = .... (dk1 )(dk2 )f0 (k1 )f0 (k2 )V2 (k1 , k2 ; k1 , k2 ). (2.29)
2
Comparing with the expression (2.25), we identify w2 (k1 , k2 ) = βV2 (k1 , k2 ; k1 , k2 ).
At the next order, there is again only one diagram consisting of three lines,
shown in Figure 2.1, and this implies w3 (k1 , k2 , k3 ) = sβV3 (k1 , k2 , k3 ; k1 , k2 , k3 ).
There are multiple diagrams with four lines, one of them includes V4 , and the
20
(a) Lowest contribution to w2
(b) Lowest contribution to w3
Figure 2.1: Examples of contributions to two- and three-particle terms.
21
Figure 2.2: Diagrams contributing to w4
others can be built from V2 . They are shown in Figure 2.2. Their sum is
Z "Y 4
#
h1
F = ···+ (dki )f0 (ki ) V4 (k1 , . . . k4 ; k1 , . . . k4 )
i=1
4!
s
+ (2π)d δ (d) (k2 − k3 )βV2 (k1 , k2 ; k1 , k3 )V2 (k2 , k4 ; k3 , k4 ) (2.30)
2
s X 4 i
d
+ (2π) δ( ki )β 2 V2 (k1 , k2 ; k3 , k4 )V2 (k3 , k4 ; k1 , k2 ) .
8 i=1
From this expression, we can identify w4 , which is now a sum of terms built
from the vertices V4 , V2 .
An interesting class of diagrams are the “ring-diagrams” shown in Figure 3,
since they can be summed up in closed form. Let us define
G2 (k, k′ ) = V2 (k, k′ ; k, k′ ) (2.31)
A ring diagram with N external rings attached will have the following value
according to the diagrammatic rules:

Z YN
1 N +1
s (dk0 ) (dki ) [βG2 (k0 , ki )f0 (ki )] f0 (k0 )N if N > 1;
βN
Z i=1 (2.32)
1
(dk0 )(dk1 )βG2 (k0 , k1 )f0 (k1 )f0 (k0 ) if N = 1.
2β
Apart from the N = 1 term which has a different coefficient, all other terms
can be identified with the power series for the logarithmic function. This class of
diagrams are then resummed into:
22
3
.
.
2 .
1 n
Figure 2.3: Ring diagrams.
Z Z
s
Fring =− (dk0 ) log 1 − sf0 (k0 ) (dk1 )βG2 (k0 , k1 )f0 (k1 )
β
Z (2.33)
1
− (dk0 )(dk1 )G2 (k0 , k1 )f0 (k1 )f0 (k0 )
2
23
CHAPTER 3
TBA-LIKE INTEGRAL EQUATION
We have formally developed an expansion for the free energy of a interacting

system in the previous chapter. However, it is still very much a mathematical ex-
ercise at this stage. More work is clearly needed before we can apply the formalism
on real problems.
There are in particular two loose ends: first, the diagrammatic expansion we
developed earlier is still very much a weak-coupling expansion. A resummation

scheme is needed in order to obtain results valid beyond the infinitesimal coupling
regime. Second, the vertex functions V, although formally defined, still lack explicit
expressions.
In this chapter we devise a way to re-sum the terms in our diagrammatic

expansion. The method resembles the resummation of self-energy insertions of the
full propagator in the usual perturbative quantum field theory.
In practice, one of course still has to truncate the resummation at some point,
corresponding to only resumming selected classes of diagrams. We shall introduce
one such approximation that treats two-body interaction exactly but truncates all
other terms.
And then we shall turn our attention to the four-point vertex function involved
in this approximation, and derive its explicit expression in terms of the two-body
scattering amplitude of the underlying filed-theoretic model.
The end result of all these will be an integral equation for the effective single-
particle energy, which turns out to be closely related to the thermodynamic Bethe
24
ansatz of integrable system in one-dimension.
3.1 Legendre transformation
Given F (µ), one can compute the thermally averaged number density n:
Z
∂F
n=− ≡ (dk)f (k) (3.1)
∂µ
The dimensionless quantities f are called the filling fractions or occupation num-
bers. One can express F as a functional of f with a Legendre transformation.

Define
G ≡ F (µ) + µ n (3.2)
Treating f and µ as independent variables, then using eq. (3.1) one has that ∂µ G =
0 which implies it can be expressed only in terms of f and satisfies δG/δf = µ.

Inverting the above construction shows that there exists a functional ̥(f, µ):
Z
̥(f, µ) = G(f ) − µ (dk)f (k) (3.3)
which satisfies eq. (3.1) and is a stationary point with respect to f :
δ̥
=0 (3.4)
δf
The above stationary condition is to be viewed as determining f as a function of
µ. The physical free energy is then F = ̥ evaluated at the solution f to the

above equation. We will refer to eq. (3.4) as the saddle point equation since it is
suggestive of a saddle point approximation to a functional integral.
The function ̥ is also required to satisfy

Z
∂̥
=− (dk) f (k) (3.5)
∂µ
25
In the sequel, it will be convenient to trade the chemical potential µ for the
variable f0 . We will need:
∂f0
= βf0 (1 + sf0 ) (3.6)
∂µ
In the free theory the functional ̥0 is fixed only up to the saddle point equa-
tions, thus its explicit expression is not unique. As we will see, the appropriate
choice that is consistent with the diagrammatic expansion is:

Z
1 f − f0
̥0 (f, f0 ) = − (dk) s log(1 + sf ) − (3.7)
β 1 + sf0
δ̥0
The saddle point equation δf
= 0 implies f = f0 , and inserting this back into
̥0 gives the correct free energy F0 for the free theory. Furthermore, ̥0 satisfies
eq. (3.5) if one uses the saddle point equation f = f0 .
Let us now include interactions by defining
̥ = ̥0 + ̥1 (3.8)
where ̥0 is given in eq. (3.7) and we define U as the “potential” which depends
on f and incorporates interactions:
Z
1
̥1 = − (dk) U(f (k)) (3.9)
β
3.2 Integral equation
The physical filling fraction f can be obtained by differentiating F in (2.25) with
respect to the chemical potential µ, and the result is

X∞ Z "YN
#
1
f (k) = f0 (k) + f0 (k) (sf0 (k) + 1) (dki )f0 (ki ) wN +1 (k, k1 , . . . , kN ).
N =1
N! i=1
(3.10)
26
Since f0 is represented by a line in our graphical expansion, it is tempting to
postulate that f would be a fully dressed line, but this interpretation isn’t exactly
correct.
Graphically, a dressed line is obtained by summing over all possible insertions

due to interactions. The sum can be generated in the following way: take a
(connected) vacuum diagram, cut an internal line into two external legs, and the
resulting diagram contributes to the sum. If we sum over all possible ways of
cutting an internal line from a vacuum diagram, and furthermore sum over all
such diagrams, we have generated all correction terms to the so-called propagator.
Cutting a line and replacing it with two legs corresponds to replacing a factor
of f0 (k) by sf0 (k)2 . We have the extra factor of s because cutting one line removes
one loop integral.
Recall that F − F0 is the sum of all connected vacuum diagrams. The action of
the derivative ∂/∂µ makes sure that we’ve picked out every internal line in every
diagram. However, it replaces f0 by f0 (1 + sf0 ). Consequently, we define the
dressed line fe(k) by

sf0 (k)
fe(k) = f0 (k) + (f (k) − f0 (k)) . (3.11)
sf0 (k) + 1
It will be useful to define ǫk as the single particle pseudo-energy at momentum
k as follows:
1
f (k) = (3.12)
eβǫk − s
Then the eq. (3.11) can be equivalently expressed in the simpler form:
fe(k) = eβ(ǫk −ωk +µ) f (k) (3.13)
27
= + Σ
Figure 3.1: Graphical representation of the integral equation (3.15). The

double line represents fe, and the single line represents f0 .
The following identity will be useful:

δ fe(k) sf0 (k)
′
= (2π)d δ(k − k′ ) (3.14)
δf (k ) 1 + sf0 (k)
This fully dressed line should satisfy an equation analogous to the Dyson equa-
tion for the full propagator:
fe(k) = f0 (k) + sf0 (k)fe(k)Σ(k), (3.15)
where Σ(k) is the sum of all (amputated) one-particle (1PI) insertions. Figure 3.1
represents this equation graphically.
Again we consider the analogy to Feynman diagrams. Take a two-particle

irreducible (2PI) vacuum diagram with each internal line being fully dressed fe,
and remove one internal line to form an amputated 1PI insertion. Summing over
all ways of removing a line of a diagram, and then summing over all 2PI vacuum
diagrams, the result is Σ.
Let −β̥1 be the sum of all 2PI vacuum diagrams, with internal lines being fe
instead of f0 . The operations above can be neatly summarized by the following:
−1
sf0 (k) δ ̥1
Σ(k) = −sβ . (3.16)
1 + sf0 (k) δf (k)
The factor of s is inserted by hand because the removal of one internal line decreases
the loop count by one.
28
One correctly surmises that ̥1 turns out to be the interaction part of ̥.
Indeed, when there is no interaction, all vertex functions are zero and ̥1 vanishes.
The “Dyson” equation now reads:
δ̥1
fe(k) − f0 (k) + β fe(k)(1 + sf0 (k)) = 0. (3.17)
δf (k)
and is actually an integral equation for f .
If we demand that the saddle point equation of the functional ̥ leads to (3.17),
then ̥0 can be uniquely fixed up to an additive constant and an overall scale. If
we further demand that ̥0 reproduces the result of the free theory, then one can
show that the only consistent choice is ̥ = ̥0 + ̥1 , where ̥0 is given in eq. (3.7)
and
̥1 = F 2P I (f0 → fe) (3.18)
i.e. ̥1 is given by the 2PI diagrams with f0 replaced by the fully dressed fe.
3.3 Two-body approximation
We shall now make the following approximations:
1. include only the four-vertex, i.e. V2 , in our diagrammatic expansion;
2. consider only the foam diagrams, shown in Figure (3.2).
In practice, this approximation amounts to setting

Z
1
̥1 = − (dk)(dk′ )f˜(k)f˜(k′ )G2 (k, k′ ), (3.19)
2
and truncating all terms with more f˜.
29
Figure 3.2: Foam diagrams.
Within this approximation, the integral equation (3.17) can be rewritten into a
self-consistent equation for the single-particle pseudo-energy. With the expressions
(3.19) and (3.14), the equation (3.17) reduces to
Z
˜ ˜ ′ ′ ˜ ′
f (k) = f0 (k) + f (k) sβf0 (k) (dk )G2 (k, k )f (k ) . (3.20)
In terms of the pseudo-energy ǫk defined in eq. (3.12), using eq. (3.13) one can
rewrite (3.20) as follows:
Z β(ǫk′ −ωk′ +µ)

1 ′ ′ e
ǫk = ωk − µ − log 1 + β (dk )G2 (k, k ) βǫ ′ (3.21)
β e k −s
One may also want to find the free energy density of the system in terms of the
physical filling fraction f . By using (3.20) and (3.11), it can be shown that
Z
1
̥1 = − (dk)(dk′ )f˜(k)f˜(k′ )G2 (k, k′ )
2
Z
1 f˜(k) − f0 (k)
=− (dk) (3.22)
2β sf0 (k)
Z
1 f (k) − f0 (k)
=− (dk) .
2β 1 + sf0 (k)
The free energy density then takes the simple form:
F = ̥0 + ̥1
Z (3.23)
1 1 f − f0
=− dk s log(1 + sf ) −
β 2 1 + sf0
30
In summary, the two-body approximation consists of the integral equation
(3.21) for the pseudo-energy defined in eq. (3.12), and the above expression (3.23)
for the free energy. What is left is to give an explicit expression for the two-body
kernel G2 .
3.4 Two-body kernel
In this section, we derive the two-body kernel G2 (k1 , k2 ) that enters the approxi-
mation of the last section. From the definition (2.27) and (2.31), we identify the
kernel as the connected part of the diagonal matrix element of log S(E). In other
words,

1 2 2
2πδ E − (k + k2 ) V G2 (k1 , k2 ) ≡ −ihk1 , k2 | log Ŝ(E)|k1 , k2 i. (3.24)
2m 1
where V is the spatial volume.
Following the usual convention, we write the two-body S-matrix as the sum of
the identity operator and the T -matrix, and then formally define log S as a power
series of the T -matrix:
S(E) = 1 + i2πδ(E − H0 )T (E) (3.25)

X∞
1
log S(E) = − (−i2πδ(E − H0 )T (E))n (3.26)
n=1
n
We shall now make two further assumptions that are rather general:
• The Hamiltonian is overall translationally invariant.
• The interaction potential V is isotropic in all direction.
31
The first assumption implies momentum conservation in all collision processes,
while the second assumption is valid if we are in the low-energy, s-wave scattering
regime.
In the graphical language, the T-matrix is the sum of all connected amputated
Feynman diagrams with two incoming and two outgoing external legs.
Allowing the total energy E to go off-shell, the t-matrix element between an

incoming and an outgoing two-particle states can be written as:
hk1 , k2 |T |p1 , p2 i = (2π)d δ (d) (p1 + p2 − k1 − k2 )M(E)p1 ,p2 →k1 ,k2 , (3.27)
where the scattering amplitude M is of course model-dependent.
We won’t explicitly calculate M until the next chapter, but a general result fol-
low immediately from the assumed translational and rotational invariances. When
the energy E is on-shell, the S-matrix describes the actual physical process, and
all the usual symmetries and conservation laws apply:
• Translational invariance means M can only depend on the differences of pair

of momenta.
• Rotational invariance of the total Hamiltonian, as well as the interaction

potential, cut the list of variables down to only |k1 − k2 | and |p1 − p2 |.
• Finally, energy and momentum conservation implies |k1 − k2 | = |p1 − p2 |.
1
Therefore, when E = 2m
(p21 + p22 ) is on-shell, the matrix elements of T can be
32
written as
p2 p2
hk1 , k2 |δ(E − H0 )T (E = 1 + 1 )|p1 , p2 i
2m 2m
1
= (2π)d δ E − (k2 + k22 ) δ (d) (p1 + p2 − k1 − k2 )M(|p1 − p2 |)
2m 1
2m(2π)d
= δ(|p1 − p2 | − |k1 − k2 |) δ (d) (p1 + p2 − k1 − k2 )M(|p1 − p2 |)
|p1 − p2 |
(3.28)
With this, we set out to find the diagonal element of the nth power of δ(E −
H0 )T (E):
hp1 , p2 |(δ(E − H0 )T (E))n |p1 , p2 i

Z
n−11
= hp1 , p2 | [δ(E − H0 )T (E)] × (dk1 )(dk2 )|k1 , k2 ihk1 , k2 |× (3.29)
2α
2 2
p1 p22 p1 p22
δ + − H0 T + |p1 , p2 i.
2m 2m 2m 2m
The left-most energy δ-function ensures that the right-most T (E) is on-shell,
and the on-shell matrix element has the form of (3.28). We have inserted a com-
plete set of two-particle states, with the combinatoric factor α defined in (1.9).
Since the intermediate state |k1 , k2 i is constrained to have the same energy as
the incoming state |p1 , p2 i, the operator T (E) acting on |k1 , k2 i is once again on-
shell. Furthermore, given that |p1 − p2 | = |k1 − k2 |, the amplitude M is constant
for every pair of allowed (k1 , k2 ).
Inserting more complete sets of states in between every factor of δ(E−H0 )T (E),
we see that each matrix element becomes on-shell, and all amplitudes are identi-
cally equal to M(p1 − p2 |).
Let I denotes the phase space volume accessible by each set of intermediate
33
states. It is defined as
Z 2
1 d+1 p1 p22 k21 k22
I= (dk1 )(dk2 )(2π) δ + − − δ d (p1 + p2 − k1 − k2 ).
2α 2m 2m 2m 2m
(3.30)
For definiteness, we give here expressions of I in dimension d = 1, 2, 3:



 m2

 √ 1
d=1

 2 2α |p1 −p2 |

I= m
d=2 (3.31)

 4α




 m |p1 − p2 | d=3
4πα
And we can now express the matrix element (3.29) in a compact form:
hp1 , p2 |(δ(E − H0 )T (E))n |p1 , p2 i

n (3.32)
p21 p21 d d 1 IM
=δ E− − (2π) δ (0) .
2m 2m I (2π)
Each of the (n − 1) intermediate sets of states leads to a factor of I. The

left-most momentum δ-function is saturated, leaving us with (2π)d δ (d) (0) = V , the
volume of space. The left-most energy δ-function stays intact. Now we are in the
position to re-sum the power series in (3.26):

p21 p22 1
hp1 , p2 | log S(E)|p1 , p2 i = V 2πδ E − − log [1 + iIM(|p1 − p2 |)]
2m 2m I
(3.33)
By comparing with (3.24), we obtain a closed-form expression for the two-body
kernel:
i
G2 (p1 , p2 ) = − log (1 + iIM(|p1 − p2 |)) (3.34)
I
Also by looking at (3.24), we see that the kernel G2 can be symbolically con-
sidered as
−i log S(E)
G2 ∼ . (3.35)
(2π)δ(E − H0 )
34
It is therefore sensible to identify the logarithm in (3.34) as the “phase shift” due
to scattering. We make the definition:
Θ(p1 , p2 ) ≡ log (1 + iIM(|p1 − p2 |)) . (3.36)
This quantity Θ will appear again when we discuss the unitary limit of a quantum
gas.
3.5 Extension to multi-component gases
The extension of the above formalism to a multi-component gas is very straight-

forward. Even though our work here primarily focuses on quantum gas with one
species of particles, a fermion with spin-1/2 necessarily has two spin components,
making this extension relevant to our work.
Let a = 1, 2, . . . N labels the N species of particles. In general, they may have

different masses, experience different chemical potentials, or even obey different
statistics. One can nevertheless formally define an SU(N) group that transform
these species into each other. We can then say that the masses, chemical potentials,
and statistical factors s all carries an SU(N) index a and transform as vectors.
By the same token, the filling fractions f and pseudo energy ǫ are also SU(N)
vectors.
Much of the derivation in chapter II remains valid. The only modification

needed is that, in addition to the momentum, each particle is also labeled by its
species. In the diagrammatic expansion, each line now carries an additional species
label, and the vertex function Vn carries n species indices, matching the 2n lines
35
that are joined by the vertex. A summation over the species index is implied for
each closed loop.
The Legendre transformation goes through similarly. The free energy functional
̥ once again consists of the non-interacting part ̥0 , and ̥1 , the part due to
interaction.
The non-interacting part ̥0 is simply a sum over all particle species:
X1Z
fa − fa0
̥0 = − (dk) sa log(1 + sa fa ) − ; (3.37)
a
β 1 + s a fa0
while the interacting part is the sum of all 2PI vacuum diagrams, with the dressed
fea as its internal lines. In particular, the two-body approximation now yields a set
of N coupled integral equations.
A two-particle state now carries two species labels. Since the two particles can
in general be distinguished by their momenta, we do not assume any symmetry
property for these two species indices. The species label for a two-particle state
may hence take N 2 distinct values. The two-particle scattering amplitude can then
be regarded as an N 2 × N 2 matrix.
The two-body kernel follows straightforwardly:

i
G2,ab (p1 , p2 ) = − log (1 + iIM(|p1 − p2 |)) . (3.38)
I ab,ab
Here the logarithm of an N 2 × N 2 matrix is defined by its Taylor series as usual.

The accessible phase space volume I is itself an N 2 × N 2 diagonal matrix, its
components given by:
Z
1 k2 k2
Iab,ab = (dk1 )(dk2 )(2π) d+1
δ E− 1 − 2 δ d (p1 + p2 − k1 − k2 ), (3.39)
2 2ma 2mb
where E denotes the total initial energy.
36
Instead of (3.20), we have
!
XZ
f˜a (k) = fa0 (k) + f˜a (k) sa βfa0 (k) (dk′ )G2,ab (k, k′ )f˜b (k′ ) . (3.40)
b
The pseudo-energy becomes

Z β(ǫb,k′ −ωb,k′ +µb )

1 ′ ′ e
ǫa,k = ωa,k − µa − log 1 + β (dk )G2,ab (k, k ) . (3.41)
β eβǫb,k′ − sb
We now work out the special case of the spin-1/2 fermion gas, with the species
index a = 1, 2. We have s1 = s2 = −1 and m1 = m2 = m, but potentially µ1 6= µ2

in the presence of a magnetic filed. Barring a symmetry-breaking field, the SU(2)
group is an actual symmetry of the action in this case.
The Pauli exclusion principle, along with the assumption of a delta-function

potential (see Chapter IV,) ensures that scattering can only take place between
two fermions of opposite spins. Angular momentum conservation further demands

that the two outgoing particles have opposite spins. We may write
M12,12 = M12,21 = M21,12 = M21,21 = M, (3.42)
while all other components are zero.
The summation over the matrix indices (ab) effectively yields a factor of 2 each
time it is performed. One may absorb this factor into I. The integral equations
and pseudo energy are then

!
XZ
f˜1 (k) = f1,0 (k) + f˜1 (k) −βf1,0 (k) (dk′ )G2 (k, k′ )f˜2 (k′ )
2 (3.43)
Z β(ǫk′ −ωk′ +µ2 )

1 ′ ′ e
ǫ1,k = ωk − µ − log 1 + β (dk )G2 (k, k ); ,
β eβǫk′ + 1
together with equations in which spin indices are exchanged. The kernel G2 is as
given in (3.34).
37
CHAPTER 4
FIELD THEORIES FOR QUANTUM GASES
After developing the S-matrix formulation of quantum statistical mechanics,

and a practical way of carrying out calculation under the frame work, we want to
apply the machinery to the problem of quantum gases.
Before jumping into the problem, however, we wish to present a quick review
of the model we are interested in.
4.1 Actions and Feynman rules
We consider two different models: a single-component boson gas, and a spin-1/2
fermion gas.
The bosonic action is

Z
d † ∇2 g † †
S= dt d x φ i∂t − + µ φ − φ φ φφ (4.1)
2m 4
The chemical potential µ is necessary because we will allow thermal fluctuation

in particle density. However, all vertex functions of our S-matrix formulation are
always associated with scattering in the vacuum.
The field operator φ(x, t) has the usual interpretation of an particle annihilation
operator: it annihilates a boson at position x at time t. Its Hermitian conjugate

φ† is the corresponding creation operator. Due to the Bose statistics, they obey
the equal-time commutation relation:

φ(x, t), φ† (x′ , t) = δ (d) (x − x′ ). (4.2)
38
One can read off the free-particle propagator in momentum space from the
quadratic part of the action (4.1):
1
GR
0 (ω, k) = lim+ . (4.3)
η→0 ω− k2 /2m + µ + iη
The infinitesimal iη prescription is added to ensure the convergence of the path

integral of eiS .
Since we are only interested in scattering processes at T = 0 in the vacuum, we

can take the chemical potential µ to be zero. Note that the advance propagator
is absent as a result. As the “backward propagation” of a particle is really the
forward propagation of a hole in the background of pre-existing particles, such

process cannot take place in the vacuum.
The quartic interaction term in (4.1) implies that these bosons interact with a
δ-function potential. This is a good low-energy description for any finite-ranged
potential. But, being a low-energy approximation, we shall see that an ultraviolet
cut-off will be necessary. This term gives a four-particle vertex in the perturbation
theory.
The fermionic action is:

Z !
X ∇2 g † †
S= dt dd x ψa† i∂t − + µa ψa − ψ↑ ψ↓ ψ↓ ψ↑ (4.4)
a=↑,↓
2m 2
The two spin states of the fermion can be regarded as two distinct species
of particles. Note that the chemical potential µ may be different for each spin,
corresponding to an external magnetic field breaking the spin rotational symmetry.
The field operators ψa and ψa† , a =↑, ↓, are the annihilation and creation op-
erators of the fermion respectively. They obey the equal-time anticommutation
39
relation:
n o
ψa (x, t), ψb† (x′ t, t) = δa,b δ (d) (x − x′ ). (4.5)
For our choice of coefficients, the bosonic four-vertex is proportional to −ig,

while the fermionic version is −ig/2. Recall the definition of α in (1.9), we have
here α = 1 for bosons and α = 1/2 for fermions, and the four-vertex can be
expressed generally as iαg.
4.2 Two-body scattering in vacuum
Central to our integral equation is the two-body scattering process in vacuum, and
we shall look at it in some depth in this section.
The exact two-body scattering amplitude M can be obtained by resumming

the so-called ladder diagrams, as depicted in figure 4.1. Algebraically this is a
geometric series, and can be resummed as such:
Z Z 2
1 2 1 1 3 1
iM = (−iαg) + (−igα) (loop) + (−igα) (loop) + . . .
2 2α 4 2α
Z −1 (4.6)
ig
= −igα 1 + (loop) .
2
Attention is drawn to the extra factor of (2α)−1 in front of each loop integral. A
boson loop carries a combinatoric factor 1/2 since both legs of the loop are identical,
while a fermion loop does not have this factor since the two legs have opposite spins.
The upshot is, regardless of statistics, each additional loop contributes a factor of
R
−igs/2 (loop).
40
= + + +...
iM
Figure 4.1: Ladder diagram resummation of the two-body scattering ampli-

tude.
Assuming that the two incoming particles have a total energy E and total
R
momentum P, the loop integral (loop) is given as
Z Z
i i
(loop) = (dω)(dk)
ω − k 2 /2m + iǫ (E − ω) − (P − k)2 /2m + iǫ
(4.7)
4.2.1 d=2
By simple power counting, it is easy to see that the integral (4.7) is ultraviolet-
divergent is two spatial dimensions. We regulate this divergence by cutting off the
momentum integral at UV cut-off Λ, and retain only the universal terms propor-
tional to a non-negative power of Λ. The result is:
Z !
im Λ2 − iη
(loop) = − log 2
8π − mE − P4 − iη
" 2
! # (4.8)
mi mE − P4
=− log + iπ
8π Λ2
When the total energy of the incoming particles is on-shell, the quantity (mE −
P2
4
) is positive. So it is necessary to keep the infinitesimal iη until the last step to
pick out the correct branch of the logarithm function.
When this expression is plugged into equation (4.6), we get the result:
−iα
iM = 2
. (4.9)
1 m mE− P4
g
+ 8π log Λ2
+ iπ
41
4.2.2 d=3
Once again the integral (4.7) is diverging in the ultraviolet, and the UV momentum
cut-off Λ is needed. The amplitude (4.6) becomes:
−α
iM = q
1 m Λ P2
ig
+ 4π 2 i
+ π2 mE − 4
(4.10)
−iα
= q
1 im P2
gR
+ 4π mE − 4
The cutoff Λ in the last line is absorbed completely into the definition of renor-
malized coupling constant gR :
1 1 m
= + 2 Λ. (4.11)
gR g 4π
In the zero-momentum limit, the amplitude (4.10) reduces to simply
iM → iαgR . (4.12)
We note that the differential cross-section of scattering process is given by:
dσ m2 k d−3
= d−1
|(k)M|2. (4.13)
dΩ 4(2π)
At d = 3, the s-wave scattering length a is defined by σ = πa2 at the zero

momentum limit, where σ is the total cross-section. Since we have M = gR at
zero momentum, the scattering length is related to gR by
mgR
a= (4.14)
2π
42
CHAPTER 5
CRITICAL POINT OF THE TWO-DIMENSIONAL BOSE GAS
As the first application of our new formalism, we shall investigate the critical
point of the two-dimensional weakly-interacting Bose gas. Although we find results
that are similar to the known results [41, 12, 18, 42], we find our treatment to be
considerably simpler and transparent and doesn’t rely on an effective description
of vortices.
The interacting Bose gas in 2 spatial dimensions is defined by the action (4.1)
and taking d = 2.
In terms of the yet-to-be-specified two-body kernel G2 , the integral equation

(3.20) gives the pseudo-energy of the Bose gas ǫ(k). The interacting gas behaves
as if it were non-interacting, with ǫ(k) being the single-particle energy eigenvalues.
The number density n of the gas is then
Z
d2 k 1
n= 2 βǫ(k)
(5.1)
(2π) e −1
where β is the inverse temperature T .
We consider the weakly-interacting limit of the two-body kernel (3.34). As-

suming that the coupling constant g is small, the scattering amplitude is simply
M = g to leading order of g, and is independent of the particle momenta. By

expanding the logarithm in (3.34) to leading order in g, we end up with the simple
result
G2 ≈ −g (5.2)
for small g.
Since when g = 0, ǫ(k) = ωk − µ, at weak coupling an approximate solution
43
to (3.20) is obtained by substituting ǫ(k) = ωk − µ on the right hand side. As in
Bose-Einstein condensation, the critical point is defined to occur at the value of the
chemical potential µc where the occupation number at zero momentum diverges,
i.e. ǫ(k = 0) = 0. This gives the following integral equation for µc :
Z
d2 k 1
µc = − 2
G2 (0, k) β(ω −µc ) (5.3)
(2π) e k −1
Furthermore, we shall make the approximation G2 ≈ −g as reasoned above.

Consider first the case where the interaction is attractive, i.e. g < 0. The equation
(5.3) becomes
mg
βµc = − log 1 − eβµc . (5.4)
2π
For negative g, the above equation has negative µc solutions, which makes
physical sense since energy is released when a particle is added. Here the attractive
interaction makes a critical point possible at finite temperature and density. The
critical density nc ≈ µc /g.
Suppose now the interaction is repulsive, i.e. g > 0. The equation (5.4) now
has no real solutions for µc . This can be traced to the infra-red divergence of the
integral when µc changes sign. Let us therefore introduce a low-momentum cut-off

k0 , i.e. restrict the integral to |k| > k0 . The equation (5.3) now becomes
mg
βµc = βǫ0 − βµc − log eβ(ǫ0 −µc ) − 1 (5.5)
2π
where ǫ0 = k02 /2m. The above equation now has solutions at positive µc , however
for arbitrary k0 there are in general two solutions. If k0 is chosen appropriately to
equal a critical value kc , then there is a unique solution µc . This is shown in Figure
5.1, where the right and left hand sides of equation (5.5) are plotted against µc for
two values of k0 , one being the critical value kc .
44
Since the slopes of the right and left hand sides of eq. (5.5) are equal at
the critical point, taking the derivative with respect to µc of both sides gives the
additional equation:
mg
β(ǫc − µc ) = log 1 + (5.6)
2π
where ǫc = kc2 /2m. The two equations (5.5, 5.6) determine µc and ǫc :

mgT 2π
µc = log 1 +
2π mg
(5.7)
kc2
mg mg mgT 2π
ǫc = =T 1+ log 1 + + log
2m 2π 2π 2π mg
Note that ǫc goes to zero as g goes to zero. Finally, making the approximation
ǫ(k) ≈ ωk − µc in eq. (5.1) and using eq. (5.5) one obtains the critical density
nc ≈ µc /g:

mkB T 2π
nc ≈ log 1 + (5.8)
2π~2 mg
p
The above can be expressed as nc λ2 = log(1+2π/mg) where λ = ~ 2π/mkB T
is the thermal wavelength. As expected, at fixed temperature, the critical density
goes to infinity as g goes to zero.
In obtaining the above result we have only kept the leading term in the kernel
G2 at small coupling, which is momentum-independent. Systematic corrections to

the above results may be obtained by keeping higher order terms in the kernel.
However since the kernel involves log k2 , at small coupling these corrections are
quite small, as we have verified numerically.
In numerical simulations of the critical point, it was found that µc ∼

mgT
2π
log(2C/mg) where C is a constant[42]. This form agrees with our result (5.7)
in the limit of small g, however, whereas we find C = π, numerically one finds a
significantly higher value C ≈ 13.2.
45
RHS, LHS
1.4
1.2 ǫ0 = ǫc βǫ0 = 0.9

1.0
0.8
0.6
0.4
0.2
βµc
0.2 0.4 0.6 0.8 1.0
Figure 5.1: Left and right hand sides of eq. (5.5) as a function of βµc at two
values of the infra-red cut-off ǫ0 , one being the critical value ǫc .
Assuming that C was not overestimated numerically, one possible explanation

for this discrepancy is based on the renormalization group. The beta-function for
g can be found by requiring that the kernel G2 in eq. (7.1) be independent of the
ultra-violet cut-off Λ. To lowest order this gives:
dg mg 2
Λ = (5.9)
dΛ 4π
which implies the coupling decreases at low energies. Integrating this equation
leads to the cut-off dependent coupling g(Λ):
2π 2π mg0
= 1+ log(Λ0 /Λ) (5.10)
mg(Λ) mg0 4π
where g0 = g(Λ0). Thus at low energies where Λ < Λ0 , the combination 2π/mg
grows. However since the dependence on the cut-off is only logarithmic, it is not
clear that this could explain the large C discrepancy. This seems to imply that
higher N-body interactions are important.
46
CHAPTER 6
RENORMALIZATION GROUP AND THE UNITARY LIMIT
With the knowledge of scattering amplitudes, we can now discuss the renor-
malization group flow of our models under scaling transformations.
Let us first consider the free-field case by setting the coupling constant g → 0.
By inspection, the actions (4.1) and (4.4) possess a scale invariance with dynamical
exponent z = 2, i.e. they are invariant under:
t → Λ−2 t, x → Λ−1 x.
We shall now proceed to show that the actions (4.1) and (4.4) are also scale-
invariant in the so-called unitary limit, where the s-wave scattering length of the
model diverges.
It is customary to define the scaling dimension of a field X, dim[X], in unit of

inverse length. In this notation, we have dim[k] = 1, dim[t] = −2, and so on.
Requiring that the action be dimensionless yields dim[φ] = dim[ψ] = d/2 and
dim[g] = 2 − d.
6.1 The 3d case
The renormalization group behavior can be inferred from the coupling constant
dependence of the S-matrix. Consider first the particular case of d = 3 dimensions.
The cut-off dependence of the scattering amplitude is completely absorbed into
the definition of renormalized coupling gR in equation (4.11). Two theories, with

different bare coupling constant g and cut-off Λ, would appear to have the exact
47
same low-energy behavior if they have the same gR .
Define the dimensionless coupling ĝ = gΛ−1. If we require the low-energy

behavior of the model be independent of the cut-off Λ, which represents the high-
energy or short-distance detail of the actual physical system which we don’t want
to deal with, we arrive at the following beta-function:
dĝ m
= −ĝ − 2 ĝ 2 , (6.1)
dl 4π
where l = − log[Λ].
This beta-function goes to zero for two different values of ĝ. The first, trivial
case is ĝ = 0, corresponding to the free-field case. We shall show that the other
zero at ĝ = ĝ∗ ≡ −4π 2 /m corresponds to the unitary limit.
At this renormalization group fixed point, the renormalized coupling gR takes
value
1 Λ m
= + 2 Λ = 0. (6.2)
gR ĝ∗ 4π
As gR is proportional to the scattering length, we have shown that the unitary limit
of the model, characterized by an infinite scattering length, is this scale-invariant
fixed point. This should not come as a surprise, because the scattering length is
the only length scale in the low-energy effective theory, and the only cases where
it can be invariant under a scale transformation are zero and infinity.
The bare coupling ĝ∗ is negative at the fixed point, implying an attractive
potential between each pair of particles. A gas of bosons with mutual attractive
may be thermodynamically unstable and prone to collapse. On the other hand,
the attractive interaction in a fermion gas may be balanced by the Fermi pressure,
and the gas stable against collapse.
48
The bare coupling can approach this non-trivial fixed point from either side of
the ĝ-axis. From (4.11) it is easy to see that gR and (ĝ − ĝ∗ ) have opposite signs.
To the right of the fixed point, the scattering length is positive, indicating
attraction at low energy. For a fermionic gas, it is well-known that an effective
attraction leads to Cooper pairing and the BCS instability at low temperature.
On the other hand, the scattering length is negative on the left side of the fixed
point, indicating an effective repulsion at low energy. The amplitude (4.10) has a
pole at k = 16πi/mgR (which corresponds to a negative center-of-mass energy),

indicating the presence of a bound state. The position of the pole gives the binding
energy of the pair, and it is:
128π 2
Ebind =− 3 2. (6.3)
m gR
This bound state is only present for ĝ < ĝ∗ . The fermion pairs are bosonic, and
would presumably undergo Bose-Einstein condensation at a low enough tempera-
ture.
We therefore identify the fixed point ĝ∗ with the BCS-BEC crossover, at which
the inverse of the scattering length 1/a changes sign. We also note that the binding
energy (6.3) goes to zero as the system approaches the crossover from the BEC
side, and then the bound state ceases to exist on the BCS side. Nevertheless,
the crossover is expected to be smooth if one includes the bound states in the
thermodynamic of the BEC side.
As the model approaches its fixed point gR → 0, the phase shift (3.36) becomes:

i 8π
Θ(p1 , p2 ) = log −1 + (6.4)
gR m|p1 − p2 |
As gR → ±∞, the phase shift becomes independent from the momenta of
49
the particles, and takes value of ±iπ. The interaction is infinitely strong in the
low-energy regime that the phase shift of the two-particle scattering is always
saturated at the largest possible value (any phase shift larger than π is physically
indistinguishable from a smaller negative shift.)
There is a closed parallel in the one-dimensional thermodynamic Bethe Ansatz
of interacting bosons. By simple power counting, the two-boson interaction is a

relevant perturbation in the sense of renormalization group, and would grow in
strength as the model flows to the infrared. And in the Bethe Ansatz solution,
these bosons behave as if they are fermions due to their strong interaction. More
specifically, since the exchange phase and scattering phase can’t be separated in
one-dimension, the phase change due to interaction has the effect of altering par-
ticle statistics.
We shall adopt the saturation of scattering phase shift at ±iπ as an alternative

definition of the unitary limit.
6.2 The case of a generic d 6= 2
The loop integral (4.7) can be evaluated for a general d. The ω integral can always
be done by closing the contour in the upper-half plane and picking up the pole at
ω = k2 /2m − iǫ. The loop integral then become
Z Z
1
(loop) = i (dk) 2
− (p−k)
k 2
E − 2m 2m
− 2iǫ
Z (6.5)
1
= −i m (dk) 2
k2 − (m E − p4 ) − iǫ
Note that, to get to the last line, we have shifted the integration variable (k −
p/2) → k.
50
If 2 < d < 3, the integral is ultraviolet divergent and in need of regularization.
Let us write d = 2 + η, and introduce the UV cut-off Λ:
Z Z Λ
Ωd k 1+η dk
(loop) = −i m
(2π)d 0 k2 − (m E − p2 ) − iǫ
4
Z Λ
Ωd
= −i m dk k η−1 + . . . (6.6)
d
(2π) 0
Ωd Λη
= −i m + ...
(2π)d η
where Ωd = 2π d/2 /Γ(d/2) is the volume of the unit (d − 1)-sphere. The terms that
are omitted are either cutoff-independent or proportional to some inverse power of
Λ, and can be dropped in the low-energy limit.
Define the dimensionless coupling ĝ = gΛη . By requiring the amplitude (4.6)
to be scale-invariant, we see that the combination:
1 Λη Ωd Λη
≡ +m (6.7)
gR ĝ (2π)d η
must be cutoff-independent. This leads to the beta-function:
dĝ Ωd 2
= −(d − 2)ĝ − m ĝ . (6.8)
dl (2π)d
By requiring the beta-function to vanish, we can solve for the non-trivial fixed
point:
m (2π)d
ĝ∗ = (2 − d) (6.9)
Ωd
For the present case, we see that the renormalization group flow is qualitatively
similar to the case of d = 3, which we just worked out explicitly. Apart from the
infrared-stable fixed point ĝ = 0, there exists a ultraviolet-stable fixed point ĝ = ĝ∗

at which the renormalized coupling gR diverges. The renormalization group flow
is depicted in Figure 6.1(a).
51
BEC BCS
g
g∗ 0
(a) d > 2
g
g∗
0
(b) d < 2
g∗
g
0
(c) d = 2
Figure 6.1: Renormalization group flow for space dimension d. The arrow
indicates flow to the IR.
For d < 2, the loop integral is convergent. However, the zero-momentum limit
of the amplitude is not well-defined, and we need an IR cutoff if we wish to speak of
the scattering length in the infrared. The computation leading to the beta-function
is identical to the 2 < d < 3 case and giving identical result as in (6.8).
This time, however, the non-trivial fixed point is infrared stable. The renor-
malization group flow is also depicted in Figure 6.1(b).
6.3 d=2
The loop integral (4.7) is logarithmic divergent at d = 2. Equivalently, the two-

particle interaction is marginal in the sense of renormalization group flow. This
fact makes the d = 2 case more subtle.
The running of the coupling g can be derived from requiring the amplitude
52
(4.9) to be independent of the cutoff Λ. This condition leads to the requirement:
1 m
− log(Λ2 ) = constant, (6.10)
g 8π
and from here we get the beta-function:
dg m
= − g2. (6.11)
dl 4π
This is indeed the same as (6.8) with d = 2. There is only one fixed point at g∗ = 0,
corresponding to the free-field theory, thus there is no analogue of the BCS/BEC
crossover. The renormalization group flow is depicted in Figure 6.1(c).
One can nevertheless formally define the unitary limit as the scattering phase
shift approaching ±iπ, parallel to the 3d case. In this section we shall explore
this possibility, and interpret it using the renormalization group. As we’ll see, this
limit occurs at g = ±∞, and the scattering length diverges.
First let us begin with the beta function above (6.11). Let g = g0 at some
arbitrary scale Λ0 . Integrating the beta function one finds:
g0
g(Λ) = mg0 . (6.12)
1− 4π
log(Λ/Λ0 )
Note that g diverges at the scale:
Λ∗ = Λ0 e4π/mg0 . (6.13)
This is the familiar Landau pole. Whereas in for example the relativistic φ4 theory
in d = 3 where the Landau pole is unphysical due to higher order corrections, here
the beta function (6.11) is exact [32], thus this divergence is physical.
There are two cases to consider:
1. Attractive case: If g0 is negative, Λ∗ < Λ0 , and thus g = −∞ occurs in the

infra-red.
53
2. Repulsive case: If g0 is positive, Λ∗ > Λ0 , and g = +∞ occurs in the
ultra-violet.
Both cases are consistent with the flows depicted in Figure 6.1.
In fact, the scale Λ∗ = Λe4π/mg is an RG invariant:
dΛ∗
=0 (6.14)
d log Λ
and is the natural coupling constant in this problem that physical quantities are
expressed in terms of. The amplitude (4.9) becomes:
−iα
−iM =
m Λ2∗ m
8π
log mE− 41 (p1 +p2 )2
+ iπ 8π
(6.15)
−iα
= .
m 2Λ∗
4π
log |p1 −p2 |
+ i π2 4π
m
where in the last line we have assumed that the momenta p1 and p2 are on-shell.
From here we obtain the scattering phase shift:
 
log |p12Λ ∗
−p2 |
− i π2
Θ(p1 , p2 ) = log   (6.16)
2Λ∗ π
log |p1 −p2 | + i 2
Near the fixed point g(Λ) = 0− , Λ∗ = 0 from (6.13), and the phase shift Θ = 0,
consistent with the free theory. The same applies to the other side of the fixed
point g(Λ) = 0+ , where Λ∗ = +∞.
On the other hand, consider the low energy limit of the attractive case where
g(Λ) = −∞. Here Λ∗ = Λ, and this scale is effectively an infra-red cutoff. Thus
at low energies |p1 − p2 | ' 2Λ∗ and the phase shift Θ = +iπ. One must keep the
positive infinitesimal log(|p1 − p2 |/Λ∗ ) in order to pick out the correct branch of
the logarithmic function.
54
Similar arguments apply to the high energy limit of the repulsive case g(Λ) =
+∞, and the phase shift turns out to be Θ = −iπ. It is clear that, unlike in
3d, this definition of the unitary limit does not correspond to a renormalization
group fixed point in the usual sense of a zero of the beta function and is somewhat
engineered; nevertheless it defines a scale-invariant theory.
Let us turn now to the scattering length. Using the scattering amplitude (6.15),
for |p1 − p2 | = ∆p, the cross section is:
m2 g 2 1
σ(k) = mg
4∆p (1 + 4π log(2Λ/∆p))2 + π 2
(6.17)
4π 2 1
= 2
∆p log (2Λ∗/∆p) + 16π 4/m2 g 2
Near the scale ∆p = 2Λ∗ :
2 mg 2
σ(Λ∗ ) = (6.18)
Λ∗ 4π
Equating σ with a scattering length a, as appropriate to 2d, one sees that a diverges
in the unitary limit g → ±∞.
Because of the logarithmic dependence on |p1 − p2 |, the S-matrix does not
have a pole. However the denominator is zero when |p1 − p2 | = 2iΛ∗ , which is a
remnant of the bound state pole in 3d. The energy of this quasi-bound state is
−2Λ2∗ /m. Note that this quasi-bound state disappears as g0 → 0− , since in this
limit Λ∗ → 0, and this is analogous to the situation in 3d. We thus expect that the
“attractive” case of the unitary limit should be better behaved since this bound
state disappears, in contrast to the repulsive case where Λ∗ → ∞ as g0 → 0+ , and
this will be born out of our subsequent analysis.
55
6.4 Thermodynamic scaling functions at the quantum crit-
ical point
At a quantum critical point, the only length scales of the quantum gas are the
p
thermal wavelength λT = 2π/mT and the length scale n1/d set by the density n.
Equivalently, one can express physical properties in terms of the only two energy
scales, the temperature T and the chemical potential µ, since the density is a
function of T, µ.
In order to fix normalizations in a meaningful way, it is useful to consider the

simplest theories with z = 2 scale invariance: free, non-relativistic bosons and
fermions. The free energy density is given by the well-known expression for each
particle:
Z
s
F= (dd k) log 1 − se−β(ωk −µ) (6.19)
β
where β = 1/T , ωk = k2 /2m and s = 1, −1 corresponds to bosons, fermions
respectively.
The integrals over wave-vectors can be expressed in terms of polylogarithms

R R
Liν (z), where z = eβµ is a fugacity, using dd k = 2π d/2 /Γ(d/2) dkk d−1 and the
integrals
Z ∞
zxν−1
dx = Γ(ν) Liν (z)
0 ex − z
Z ∞ (6.20)
zxν−1
dx x = −Γ(ν) Liν (−z)
0 e +z
valid for ℜ(ν) > 0. The result, as expected, is of the form of T /λdT , multiplied by
a dimensionless scaling function:
d/2
mT
F = −sT Li(d+2)/2 (sz) (6.21)
2π
56
There are two important limits to consider. For Bose gases near Bose-
Einstein condensation, physically the interesting limit is µ/T → 0. Physically
this corresponds to either the high-temperature, or the low-density limit. Since
Liν (1) = ζ(ν), where ζ is Riemann’s zeta function, this leads us to define the
scaling functions cd (µ/T ):
πmT 2
F =− c2 (µ/T ) (d = 2)
12
(6.22)
ζ(5/2)m3/2 T 5/2
F =− c3 (µ/T ) (d = 3)
(2π)3/2
where we have used ζ(2) = π 2 /6. With the above normalizations, cd = 1 for a
single free boson when µ/T = 0.
The above formulas are also well-defined for fermions at zero chemical potential.
Using

1
− Liν (−1) = 1− ζ(ν) (6.23)
2ν−1
one finds as µ/T → 0:
1 1
c2 = , c3 = 1 − √ (free fermions) (6.24)
2 2 2
It should be pointed out that the coefficients cd are analogous to the Virasoro
central charge for relativistic systems in d = 1, as discussed in [28].
The other interesting limit is T /µ → 0, i.e. z → ∞ or z → 0. Physically

this corresponds to the zero-temperature limit, with or without a non-zero particle
density in the ground state, depending on the sign of the chemical potential. (For
bosons, clearly one can only take z → 0, as a non-zero density at zero temperature
requires the presence of a Bose-Einstein condensate. The ground state, and the
Hilbert space on top of it, are qualitatively different from a free gas on which our
analysis is based. But a Fermi gas admits both limits.)
57
Here the scaling forms are naturally based on the zero temperature degenerate
free fermion gas, where µ > 0 is the Fermi energy. In fact, the function Liν (z) has
a branch cut along the real axis from z = 1 to ∞ and the bosonic free energy is
ill-defined at z = ∞, in contrast with fermions. Using the analytic continuation of
the asymptotic behavior

logν (z)
− Liν (−z) ≈ as z → ∞ (6.25)
Γ(ν + 1)
from positive integer ν to half-integer values, we define the scaling functions bd as
follows:
mµ2
F =− b2 (T /µ) (d = 2)
4π
√ 3/2 5/2 (6.26)
2 2m µ
F =− b3 (T /µ) (d = 3)
15π 2
The above normalizations are defined such that bd (0) = 1 for a single free
fermion. One can verify that the above d = 3 expression is the standard result
for a zero temperature, single component, degenerate fermion gas where µ is the
Fermi energy.
Other thermodynamic quantities follow from the free energy. The pressure
p = −F . The density is n = −∂F /∂µ and the entropy density is s = −∂F /∂T .
Using the scaling form in (6.26), one obtains for d = 3:
√
2 2 3/2 5 T ′
n= (mµ) b3 − b3
15π 2 2 µ
√ (6.27)
2 2 3/2 ′
s= (mµ) b3 (T /µ)
15π 2
where b′ is the derivative with respect to its argument T /µ. (Henceforth, g ′ will
always refer to the derivative of the function g with respect to it’s argument µ/T
or T /µ as defined above.) The analogous formulas in 2 dimensions are:

mµ T ′
n= 2b2 − b2
4π µ
(6.28)
mµ ′
s= b
4π 2
58
The energy density ǫ follows from the relation ǫ = T s + µn + F :
d
ǫ=− F (6.29)
2
It is interesting to note that the above result, which in terms of the pressure is
simply ǫ = pd/2, is a consequence of the mechanics of free gases. This shows that
this relation continues to hold for interacting gases at a quantum critical point, as
pointed out by Ho[16].
Also of interest is the energy per particle ǫ1 = ǫ/n:

dF
ǫ1 = − (6.30)
2n
Consider first the limit T /µ → 0, with µ positive. In this limit, if b is a smooth
function of T /µ as T /µ → 0+ , then the b′ terms in the density vanish and one
simply obtains
d
lim + ǫ1 = µ (6.31)
T /µ→0 d+2
which is the same result as for a free gas, where for fermions µ is equal to the
Fermi energy ǫF . The above formula is usually not appropriate to the T → 0 limit
when µ is negative. For the interacting gas µ 6= ǫF , so this leads us to define the
scaling functions ξ:
d + 2 ǫ1
ξd (T /µ) = (6.32)
d ǫF
As T /µ → 0+ , the functions ξd should become universal constants, and for free
fermions ξd (0+ ) = 1. The Fermi energy ǫF can be defined based on its relation
to density in the zero temperature free fermion gas. For bosons, one can formally
use the same definition. For d = 2, ǫF = 2πn/m, whereas for d = 3, ǫF =
√
(3π 2 n/ 2)2/3 /m. This leads to the definitions:
mµ 2
ξ2 = b2 (T /µ)
2πn
√ ! 5/3 (6.33)
2 5/2
ξ3 = (mµ) b3 (T /µ)
3π 2 n
59
Next consider the energy per particle in the limit µ/T → 0. Here it is more
appropriate to use the form in eq. (6.22), which gives
d cd
ǫ1 = T (6.34)
2 c′d
The expression for ǫ1 for free fermions in the limit µ/T → 0 in d = 3 leads us now
˜
to define ξ: √ !
3 2 2 − 1 ζ(5/2) ˜
ǫ1 = √ T ξ3 (µ/T ) (6.35)
2 2 2 − 2 ζ(3/2)
With this normalization, for free fermions, as µ/T → 0, ξ˜3 = 1.
In two dimensions, ǫ1 goes to zero for free bosons as µ/T → 0 since it is

proportional to 1/ζ(1). However ǫ1 is finite for fermions in this limit. Using
limd→2 (2d/2 − 2)ζ(d/2) = 2 log 2, we define
π2T ˜
ǫ1 = ξ2 (µ/T ) (d = 2) (6.36)
12 log 2
With the above normalization, ξ˜2 (0) = 1 for free fermions.
60
CHAPTER 7
THE TWO-DIMENSIONAL QUANTUM GAS AT UNITARY LIMIT
7.1 Two-body kernal, integral equation, and scaling func-
tions
We begin with writing down the two-dimensional version of the kernel introduced
in (3.34), using (4.9) and the appropriate expression from (3.31):
 
mg 2Λ
4iα 1 + 4π
log |k1 −k2
− iπ/2
G2 (k1 , k2 ) = − log  
m mg 2Λ
1 + 4π log |k1 −k2 + iπ/2
  (7.1)
8α 2π
= − arctan   .
m log 2Λ∗
|k1 −k2 |
Here Λ∗ is as defined in (6.13).
In the unitary limit as defined in the previous chapter, g → ±∞ and the theory
is at the scale Λ∗ . The difference in momenta |k1 − k2 | ∼ 2Λ∗ and the logarithmic
function in the kernel (7.1) diverges. We therefore have
4πα
G2 (k1 , k2 ) → ∓ . (7.2)
m
For the attractive case, |k1 − k2 | approaches 2Λ∗ from above as g → −∞.
This thus corresponds to the positive sign in (7.2). Conversely, the repulsive case
corresponds to the negative sign of (7.2).
With the kernel reduced to a constant in the unitary limit, the analysis in two
dimensions is considerably simpler. It is convenient to define
δ(k) ≡ ǫ(k) − (ωk − µ);

(7.3)
y(k) ≡ e−βδ(k) .
61
The free energy (3.23) becomes:
Z
1 −βǫ(k) 1 (1 − y −1 (k))
F =− (dk) −s log(1 − se )− . (7.4)
β 2 eβǫ(k) − s
Since the kernel is independent of k, one can infer from (3.21) that δ must
also be independent of k. With this knowledge, it is now possible to carry out
the integration on the right hand side of (3.21) explicitly. The end result is an
algebraic transcendental equation, instead of an integral equation:

2α
y =1± log(1 − syz). (7.5)
sy
The integrals in (7.4) can also be carried out explicitly. The result agrees with
the scaling form (6.22), with the scaling function

6s 1 −1
c2 = − 2 Li2 (syz) + (1 − y ) log(1 − syz) . (7.6)
π 2
Simple algebra leads to the other scaling function defined in (6.26):
π2
b2 = c2 . (7.7)
3 log2 z
For the remainder of this chapter we shall drop the subscript 2d = in all scaling
functions, as we are specializing in the two-dimensional case.
The number density can either be derived from the thermodynamic relation
n = −∂F /∂µ, or from integrating the filling fraction f (k):
smT
n=− log(1 − syz) (7.8)
2π
˜ as defined in (6.36), is also

The total energy per particle scaling function ξ,
expressed in terms of c:

2 log 2
ξ˜ = c. (7.9)
log(1 − syz)
Finally ξ takes the form:
2
log z
ξ= b. (7.10)
log(1 − syz)
62
7.2 Attractive fermions
For fermions, the parameters s = −1 and α = 1/2. The Kernel (7.2) takes the
positive sign.
The equation for pseudo-energy (7.5) becomes
1
y =1+ log(1 + yz) (7.11)
y
and a solution exists for any positive z, i.e. for −∞ < µ/T < ∞. The density
is plotted in Figure 7.1. It takes on all positive values and approaches ∞ as
µ/T → ∞. As T /µ → 0+ , the density approaches the free field value n = mµ/2π.

However at high temperatures there is a departure from the free field value:
mT
lim n = 0.9546 (7.12)
µ/T →0 2π
Density for attractive fermions

n/mT
1.0
0.8
0.6
0.4
0.2
-4 -2 0 2 4
µ/T
Figure 7.1: Attractive fermions in 2d: density as a function of µ/T .
The filling fraction f is plotted in Figure 7.2. The scaling function c is shown
in Figure 7.3. One finds:
lim c = 0.624816 (7.13)
µ/T →0
63
f for attractive fermions
f
1.0
0.8
µ/T = 6
0.6
µ/T = 2
0.4
0.2
2 4 6 8 10
β k2 /2m
Figure 7.2: Attractive fermion in 2d: filling fraction f at two different values
of µ/T .
which is higher than the free field value c = 1/2. From these results, one obtains
the equation of state:
p = 1.077 nT, (µ/T → 0) (7.14)
c for attractive fermions

c
8
-4 -2 0 2 4
µ/T
Figure 7.3: Attractive fermion in 2d: the scaling function c as a function of

µ/T .
One needs to make sure that all regions of µ/T are physical, and not for instance
metastable. In particular, the entropy must increase with temperature, otherwise

the specific heat is negative. In Figure 7.4 we plot the entropy density s, and one
64
sees that ∂s/∂T < 0 when µ/T > 11.7.
Entropy for attractive fermions

s/mµ
0.4
0.3
0.2
0.1
0.0 0.1 0.2 0.3 0.4

T /µ
Figure 7.4: Attractive fermion in 2d: the entropy density s divided by mµ

as a function of T /µ.
One must bear in mind that our formalism, though non-perturbative in some
respects, is still an approximation, and such a feature could be an artifact that
disappears once the corrections are incorporated. Since it is beyond the scope
of this work to systematically explore these corrections, we will instead try and
interpret our results as they are, for this and subsequent cases.
The above feature of the entropy density could signify a phase transition at
µ/T = 11.7, where the density and temperature are related by:
mT
n = 13.1 ⇐⇒ Tc = 0.076 TF (7.15)
2π
where TF is the Fermi temperature ǫF /kB , where ǫF is defined in section III, and
Tc is the critical temperature for this hypothetical phase transition. The above
value is comparable to the prediction Tc ≈ 0.1 TF in [38] for quasi 2d (trapped)

systems. In all other regions, and for all other cases considered below, ∂s/∂T > 0.
65
7.3 Attractive bosons
With the bosonic values s = +1, α = 1, the integral equation (7.5) becomes
2
y =1± log(1 − zy) (7.16)
y
where the + (−) sign refers to repulsive (attractive) interactions. The scaling
function c is now

6 1 −1
c= 2 Li2 (zy) + (1 − y log(1 − zy) (7.17)
π 2
The scaling function b has the same expression as in eq. (7.7), with the above
c. The density now is
mT
n=− log(1 − zy). (7.18)
2π
The energy per particle scaling functions now take the form:
2 log(2)c
ξ˜ = − (7.19)
log(1 − zy)
and
2
log z
ξ= b (7.20)
log(1 − zy)
For this bosonic case, there is only a solution to eq. (??) for z ≤ zc ≈ .34, or
µ/T ≤ −1.08. The density is shown in Figure 7.5, and note that it has a maximum.
The filling fractions are shown for several µ/T up to log zc in Figure 7.6. From
these plots, it appears that f is diverging at k = 0 as z approaches zc , suggestive
of condensation to a superfluid. Let us refer to the maximum density then as the
critical density
m kB T
nc ≈ 1.24 ⇐⇒ Tc ≈ 0.81 TF (7.21)
2π~2
66
Density for attractive bosons
n/mT
0.20
0.15
0.10
0.05
-4.0 -3.5 -3.0 -2.5 -2.0 -1.5

µ/T
Figure 7.5: Attractive boson in 2d: density as a function of µ/T . The origin
of the axes is at (log zc , 0).
f for attractive bosons

f
5
2 µ/T = log zc
0.0 0.5 1.0 1.5 2.0 2.5 3.0

βk2 /2m
Figure 7.6: Attractive boson in 2d: the filling fraction f for µ/T =
−2, −1.5, −1.08 = log zc .
The scaling function c is shown in Figure 7.7. The limiting value is
c(zc ) ≈ 0.35 (7.22)
which is considerably less than for a free boson with c = 1. The function b decreases
to zero at T /µ → 0, and the limiting value is
b(zc ) ≈ 0.94 (7.23)
67
c for attractive bosons
c
0.4
0.3
0.2
0.1
-4.0 -3.5 -3.0 -2.5 -2.0 -1.5

µ/T
Figure 7.7: Attractive boson in 2d: the scaling function c as a function of

µ/T . The origin of the axes is at (log .33, 0).
The energy per particle scaling functions ξ and ξ˜ are shown in Figures 7.8, 7.9.
The limiting values are
lim ξ˜ = 0.842766, ˜ c ) ≈ 0.39

ξ(z (7.24)
µ/T →−∞
and
lim ξ = ∞, ξ(zc ) ≈ 2.3 (7.25)
T /µ→0−
Attractive bosons
ξ˜
1.0
0.8
0.6
0.4
0.2
-4.0 -3.5 -3.0 -2.5 -2.0 -1.5

µ/T
Figure 7.8: Attractive bosons in 2d: the energy per particle scaling function
ξ˜ as a function of µ/T .
68
Attractive bosons
ξ
100
80
60
40
20
-0.8 -0.6 -0.4 -0.2 0.0

T /µ
Figure 7.9: Attractive boson in 2d: the energy per particle scaling function
ξ as a function of T /µ.
As for the fermionic case, the value ξ˜ = 0.842766 = 12 log 2/π 2 implies the
energy per particle ǫ1 = T , which means the gas is in the classical limit.
7.4 Repulsive fermions and bosons
For the repulsive fermions, there are solutions to (7.5) for all µ/T < 0, and the
density is positive in this range. A plot of the density is shown in Figure 7.10. The
density maximizes at µ/T ≈ −0.56, where n/mT ≈ 0.04. This seems physically
reasonable given the strong repulsion in the unitary limit.
In contrast, for small coupling g, the kernel G ≈ −g, and there are solutions for
positive chemical potential. In the bosonic case, our formalism leads to a critical
mT
density of nc = 2π
log(2π/mg) for the Kosterlitz-Thouless transition. The filling
fractions are shown in Figure 7.11. Note that they are considerably smaller than
in the attractive case, as expected. The scaling function c is shown in Figure 7.12.
69
Density for repulsive fermions
n/mT
nc 0.04
0.03
0.02
0.01
-5 -4 -3 -2 -1 0
µ/T
Figure 7.10: Repulsive fermion in 2d: density as a function of µ/T .
It has the limiting value
lim c = 0.303964, (7.26)

µ/T →0
which is considerably less than the free field value c = 1/2.
f for repulsive fermions

f
0.20
0.15
0.10
0.05
0 1 2 3 4 5
β k2 /2m
Figure 7.11: Repulsive fermion in 2d: filling fraction f for µ/T =

−3, −2, −0.55. The top curve corresponds to the critical density
in eq. (7.27).
The scaling function b has the limiting values limT /µ→0− b = 0, and
limT /µ→−∞ b = ∞.
70
c for repulsive fermions
c
0.14
0.12
0.10
0.08
0.06
0.04
0.02
-6 -5 -4 -3 -2 -1
µ/T
Figure 7.12: Repulsive fermion in 2d: the scaling function c as a function of

µ/T .
Since the pressure is proportional to c, the same density occurs at two different
pressures. One sees that for µ/T < −0.56, the density increases with pressure as it
should. For µ/T > −0.56 the density decreases with increasing pressure, violating
dp/dV < 0, and should thus be viewed as an unphysical region. We interpret this
as a phase transition occurring at zc = e−0.56 = 0.57, where the critical density
mkB T
nc = 0.25 ⇐⇒ Tc ≈ 4.0 TF (7.27)
2π~2
At this critical point c(−0.56) = 0.24. This perhaps corresponds to a transition

to a crystalline phase. We cannot prove this, since we have not calculated the shear
modulus for instance, and it could simply be an artifact of our approximation, but
let us take it as a hypothesis. Whereas a Wigner crystal phase occurs at low
density in three dimensions, it occurs at high density in two dimensions[7]. The
possibility of crystal phases for repulsive bosons was studied in [24]. For Coulomb
repulsion with strength e2 , the thermodynamic state of a classical Coulomb system
√
is determined by Γ = πn e2 /kB T , where n is the density. Experimentally it was
71
found that Γ ≈ 137[15], which gives a critical density
1
nc ≈ 2.15 × 109 T 2 (7.28)
cm2 K 2
On the other hand, since our model has point-like interactions, and is in the
unitary limit, the conditions for a crystal phase are expected to be different. In
particular our formula eq. (7.27) has no e2 dependence, and this leads to a linear
in T dependence. For m equal to the electron mass
1
nc ≈ 4.5 × 109 T (7.29)
cm2 K
It is interesting to note that for T of order 1, which is where the data in [15] was
taken, the two densities (7.28) and (7.29) are comparable.
Repulsive fermion
ξ˜
6
-3.0 -2.5 -2.0 -1.5 -1.0 -0.5 0.0

µ/T
Figure 7.13: Repulsive fermion in 2d: the energy per particle scaling function
ξ˜ as a function of µ/T .
The energy per particle function ξ˜ is shown in Figure 7.13. The limiting values
are:
lim ξ˜ = 0.842766, ˜ c ) = 1.32
ξ(z (7.30)
µ/T →−∞
As in previous cases, the above value of ξ˜ implies ǫ1 = T as µ/T → −∞ and

the gas is thus in the classical limit.
72
Beyond the critical point where the density increases, limµ/T →0 ξ˜ = ∞. Note
that ξ˜ starts to diverge around the proposed phase transition at µ/T = −0.56.
Finally the other energy per particle function ξ is less interesting, as it diverges
at both endpoints of the density range where the density goes to zero. At the
critical point it is still quite large: ξ(zc ) = 12.6.
Density for repulsive fermions verses bosons

n/mT
0.04
0.03
0.02
0.01
-6 -5 -4 -3 -2 -1 0
µ/T
Figure 7.14: Density of repulsive bosons (bottom curve) verses fermions as a

function of µ/T .
The repulsive boson case is similar to the repulsive fermion, except that now
there are only positive density solutions to the eq. (7.5) for µ/T < − log 2. Figure
7.14 compares the density for bosons verses fermions. The maximum density for
bosons, which could again possibly signify a critical point at zc ≈ e−1.45 = 0.235,
is half that of the fermion case:
mkB T
nc = 0.125 ⇐⇒ Tc ≈ 8.0 ǫF (7.31)
2π~2
At this point c(−1.3) ≈ 0.11.
73
Viscosity for repulsive bosons
η/s
120
100
80
60
40
20
-7 -6 -5 -4 -3 -2 -1
µ/T
Figure 7.15: Repulsive boson in 2d: the ratio η/s as a function of µ/T .
74
CHAPTER 8
THE THREE-DIMENSIONAL QUANTUM GAS AT UNITARY
LIMIT
8.1 Two-body kernal and integral equation
Parallel to the d = 2 case, we take the kernel from (3.34), and insert appropriate
expressions from (3.31) and (4.10), to get the two-body kernel:
1 im
!
i4πα gR
− 8π
|k1 − k2 |
G(k1 , k2 ) = − log 1 im
m|k1 − k2 | gR
+ 8π
|k1
− k2 |
! (8.1)
m
−8πα |k
8π 1
− k2 |
= arctan 1
m|k1 − k2 | gR
1
The unitary limit is gR
→ 0± . The plus sign is the repulsive case, where the
scattering length is positive, and can be associated with the BEC side of the BCS-
BEC crossover. The negative sign is associated with the attractive interaction,
negative scattering length, and the BCS side of the crossover.
Recall that on the BEC side of the fixed point, there is a two-particle bound
state in the spectrum of the theory, with binding energy given by (6.3). For the
reminder of this chapter we shall work on the BCS side of the fixed point, so that
the bound state does not need to be incorporated in the thermodynamics.
Taking the limit, the kernel is reduced to
8π 2 α
G2 (k1 , k2 ) → ∓ . (8.2)
m|k1 − k2 |
Unlike the d = 2 case, the kernel remains dependent on the momenta of the
particles. The pseudo-energy shift δk is therefore not a constant. The integral
75
equation is
Z
y(k′ )−1
y(k) = 1 + β (d3 k′ ) G(k, k′ ) (8.3)
eβǫ(k′ ) − s
and does not reduce to an algebraic equation.
The angular part of the integral in (8.3) can be easily performed and lead to
some simplification. Defining the dimensionless variable κ = k2 /2mT , the integral
equation becomes
Z ∞ h p i
′ ′ ′ ′ z
y(κ) = 1 + 4α dκ Θ(κ − κ ) κ /κ + Θ(κ − κ) (8.4)
0 eκ′ − szy(κ′ )
where z = eµ/T is the fugacity and Θ is the Heaviside step function.
The free energy, followed from (3.23), has the same form as (7.4) and is repro-
duced below:
Z
3 −βǫ 1 (1 − y −1 )
F = −T (d k) −s log(1 − se ) − (8.5)
2 eβǫ − s
8.2 The scaling functions
The scaling function c3 , as defined in (6.22), does not admit an analytic close form
expression. It can only be determined numerically from the solution of the integral
equation (8.3) (itself numerical.)
We shall henceforth drop all the subscripts d = 3 of the scaling functions in

this chapter, since we will deal with the three-dimensional case exclusively.
It is also convenient to define the scaling function q, which is a measure of the

quantum degeneracy, in terms of the density as follows:
nλ3T = q, (8.6)
76
p
where λT is the thermal wavelength 2π/mT .
The two scaling functions c and q are of course related since n = −∂F /∂µ,
which leads to
q = ζ(5/2)c′ (8.7)
where c′ is the derivative of c with respect to x ≡ µ/T . Henceforth g ′ will always

denote the derivative of g with respect to x.
Recall from (6.30) that the energy per particle, ǫ1 , is proportional to the free
energy per particle. One may compare this with the result of a free fermion at zero
temperature, ǫ1 = 3ǫF /3, with ǫF the Fermi energy. This leads us to the scaling
function ξ, as defined in the equation (6.32):

1/3
5ζ(5/2) 6 c
ξ(x) = (8.8)
3 π q 5/3
This scaling function is meaningful at the limit x → ∞, or equivalently T → 0.
Another important limit would be x → 0. The relevant scaling function ξ˜ was

defined in (6.35), and it can be rewritten in terms of c and q introduced here:
√
˜ = ζ(3/2)(2 2 − 2) c
ξ(x) √ (8.9)
(2 2 − 1) q
The entropy density is s = −∂F /∂T , and the entropy per particle takes the
form

s 5c/2 − xc′
= ζ(5/2) (8.10)
n q
Next consider the specific heat per particle at constant volume and particle
number, i.e. constant density. One needs ∂x/∂T at constant density. Using the
fact that n ∝ T 3/2 q, at constant density q ∝ T −3/2 . This gives

∂x 3q
T =− ′ (8.11)
∂T n 2q
77
The specific heat per particle is then:

CV 1 ∂E ζ(5/2) c c′
= = 15 − 9 ′ (8.12)
N N ∂T N,V 4 q q
The isothermal compressibility is defined as

1 ∂V
κ=− (8.13)
V ∂p T
where the pressure p = −F . Since n = N/V and N is kept fixed,

3/2
∂n−1 1 q′ 1 mT q′
κ = −n = = (8.14)
∂p T nT q T 2π q2
Finally the equation of state can be expressed parametrically as follows. Given

n and T , one uses (8.6) to find x as a function of n, T . The pressure can then be
written as

ζ(5/2)c(x(n, T ))
p= nT (8.15)
q(x(n, T ))
In order to compare with numerical simulations and experiments, it will be
useful to plot various quantities as a function of q or T /TF :

2/3
T 4
= √ (8.16)
TF 3 πq
8.3 Analysis of fermions
The integral equation (8.4) is then solved numerically by iteration. One first
substitutes y0 = 1 on the right hand side and this gives the approximation y1 for
y. One then substitutes y1 on the right hand side to generate y2 , etc. For regions
of z where there are no critical points, this procedure converges rapidly, and as
little as 5 iterations are needed. For fermions, as one approaches zero temperature,
78
c for fermions
c
free
2
-2 -1 1 2
x = µ/T
Figure 8.1: Attractive fermion in 3d: c(x) and its equivalent for a free theory
as a function of x = µ/T .
q for fermions
q
3.0
2.5 free
2.0
1.5
1.0
0.5
-2 -1 0 1 2
x = µ/T
Figure 8.2: Attractive fermion in 3d: q(x) and its equivalent for a free theory
as a function of x = µ/T .
i.e. x large and positive, more iterations are needed for convergence. The results
presented in this chapter are based on 50 iterations.
When z ≪ 1, y ≈ 1, and the properties of the free ideal gas are recovered, since
the gas is very dilute. There are solutions to equation (8.4) for all −∞ < x < ∞.
(x = µ/T ). The scaling function c, and it’s comparison with a free theory, are
shown in Figure 8.1 as a function of x. The corrections to the free theory become
79
appreciable when x > −2. At x = 0:
c(0) = 0.880, ξ˜ = 0.884, (8.17)
compared to the free gas values of c(0) = 0.646 and ξ˜ = 1.
The scaling function q for the density is shown as function of x in Figure 8.2.
Note that the density in the interacting case is always higher than for a free gas,
due to the attractive interactions. At x = 0, q(0) = 1.18, whereas for a free gas
q = 0.765. At low temperatures and high densities, µ/T ≫ 1, the occupation

numbers resemble that of a degenerate Fermi gas, as shown in Figure 8.3.
Occupation number for fermions
f
1.0
0.8
0.6 x = 15
0.4
0.2
0 5 10 15 20
κ = βk2 /2m
Figure 8.3: Attractive fermion in 3d: the occupation numbers as a function

of κ for x = 5, 10, 15.
Whereas c and q are nearly featureless, other quantities seem to indicate a

critical point, or phase transition, at large density. For instance, the entropy per
particle decreases with decreasing temperature up to x < xc ≈ 11.2, as shown in
Figure 8.4. Beyond this point the entropy per particle has the unphysical behavior
of increasing with temperature.
A further indication that the region x > xc is unphysical is that the specific
heat per particle becomes negative, as shown in Figure 8.5. When x ≪ 0, CV /N
80
entropy per particle for fermions
s/n
1.28
1.26
1.24
1.22
1.20
1.18
1.16
6 8 10 12 14 16 18 20
x = µ/T
Figure 8.4: Attractive fermion in 3d: entropy per particle as a function of x.
approaches the classical value 3/2. This leads us to suggest a phase transition, at
x = xc , corresponding to the critical temperature Tc /TF ≈ 0.1.
Specific heat per particle for fermions

CV /N
0.3
0.2
0.1
6 8 10 12 14
x = µ/T
-0.1
Figure 8.5: Attractive fermion in 3d: specific heat per particle as a function
of x.
As we will show, our analysis of the viscosity to entropy-density ratio suggests

a higher Tc /TF . There have been numerous estimates of Tc /TF based on various
approximation schemes, mainly using Monte Carlo methods on the lattice [36, 29,
2, 22, 3], quoting results for Tc /TF between 0.05 and 0.23. The work [29] puts an
upper bound Tc /TF < 0.14, and the most recent results of Burovski et. al. quote
Tc /TF = 0.152(7). Our result is thus consistent with previous work.
81
The equation of state at this point follows from eq. (8.15):
p = 4.95nT (8.18)
E Energy per particle scaling function for fermions

N ǫF
1.2
1.0
0.8
0.6
0.4
0.2
0.2 0.4 0.6 0.8 1.0

T /TF
Figure 8.6: Attractive fermion in 3d: energy per particle normalized to ǫF as

a function of T /TF .
The energy per particle, normalized to the Fermi energy ǫF , i.e. E/NǫF = 3ξ/5,
and the entropy per particle, are shown in Figures 8.6,8.7 as a function of T /TF ,
where kB TF = ǫF . At high temperatures it matches that of a free Fermi gas, in
agreement with the Monte Carlo simulations in [2, 3]. Note that there is no sign
of pair-breaking at T ∗ /TF = 0.5 predicted in [36], and this also agrees with the
Monte Carlo simulations.

Entropy per particle for fermions
s/n
2.5
2.0
1.5
1.0
0.5
0.2 0.4 0.6 0.8

T /TF
Figure 8.7: Attractive fermion in 3d: entopy per particle as a function of

T /TF .
82
However at low temperatures in the vicinity of Tc , the agreement is not as good.
This suggests our approximation is breaking down for very large z, i.e. the limit of
zero temperature. The same conclusion is reached by examining µ/ǫF , displayed
in Figure 8.8, since the zero temperature epsilon expansion and Monte Carlo give
µ/ǫF ≈ 0.4 − 0.5[47, 3].
Chemical potential for fermions

µ/ǫF
0.8
0.6
0.4
0.2
0.2 0.4 0.6 0.8 1.0

T /TF
-0.2
-0.4
-0.6
Figure 8.8: Attractive fermion in 3d: chemical potential normalized to ǫF as

a function of T /TF .
8.4 Analysis of bosons
For bosons we again solved the integral equation (8.4) by iteration, starting from
y = 1. Since the occupation numbers decay quickly as a function of κ, we intro-
duced a cut-off κ < 10. For x less than approximately −2, the gas behaves nearly
classically.
The main feature of the solution to the integral equation is that for x > xc ≡
−1.2741, there is no solution that is smoothly connected to the classical limit
x → −∞. Numerically, when there is no solution the iterative procedure fails
to converge. The free energy scaling function is plotted in Figure 8.9. Note that
c < 1, where c = 1 is the free field value. We thus take the physical region to be
83
x < xc .
c for bosons
c
0.35
0.30
0.25
0.20 ideal
-1.38 -1.36 -1.34 -1.32 -1.30 -1.28

x = µ/T
Figure 8.9: Attractive boson in 3d: the free-energy scaling function c as a

function of µ/T compared to the ideal gas case.
We find strong evidence that the gas undergoes BEC at x = xc . In Figure

8.10, we plot ǫ(k = 0) as a function of x, and ones sees that it goes to zero at xc .
This implies the occupation number f diverges at k = 0 at this critical point. One
clearly sees this behavior in Figure 8.11.
ǫ(k = 0)/T
0.5
0.4
0.3
0.2
0.1
-1.40 -1.38 -1.36 -1.34 -1.32 -1.30 -1.28

x = µ/T
Figure 8.10: Attractive boson in 3d: the pseudo-energy ǫ at k = 0 as a

function of x = µ/T .
The compressibility is shown in Figure 8.12, and diverges at xc , again consistent

with BEC. We thus conclude that there is a critical point at xc which a strongly
84
Occupation number for bosons
f
500
400
300 xc
200
100
0.000
b
0.005 0.010 0.015 0.020
κ = βk2 /2m
Figure 8.11: Attractive boson in 3d: the occupation number f (κ) for x =
−1.275 and xc = −1.2741.
interacting, scale invariant version of the ideal BEC. In terms of the density, the
critical point is:
nc λ3T = 1.325, (µ/T = xc = −1.2741) (8.19)
Compressibility for bosons

κT (mT )3/2
250
200
150
100
50
-1.40 -1.38 -1.36 -1.34 -1.32 -1.30 -1.28

x = µ/T
Figure 8.12: Attractive boson in 3d: the compressibility κ as a function of

µ/T .
The negative value of the chemical potential is consistent with the effectively
attractive interactions. The above should be compared with the ideal BEC of the
free theory, where xc = 0 and nc λ3T = ζ(3/2) = 2.61, which is higher by a factor of
85
2. At the critical point the equation of state is
p = 0.318nT (8.20)
compared to p = 0.514nT for the free case. (0.514 = ζ(5/2)/ζ(3/2)).
A critical exponent ν characterizing the diverging compressibility can be defined
as
κ ∼ (T − Tc )−ν (8.21)
A log-log plot of the compressibility verses T − Tc shows an approximately straight
line, and we obtain ν ≈ 0.69. This should be compared with BEC in an ideal gas,
where ν ≈ 1.0. Clearly the unitary gas version of BEC is in a different universality
class.
Entropy per particle for bosons
s/n
3.0
2.5
2.0
1.5
1.0
0.5
1.1 1.2 1.3 1.4 1.5 1.6

T /Tc
Figure 8.13: Attractive boson in 3d: the entropy per particle as a function
of T /Tc .
The energy per particle scaling function ξ˜ at the critical point is ξ(x
˜ c ) = 0.281
√ √
compared to 0.453 = (2 2 − 2)/(2 2 − 1) for the free case. The entropy per
particle and specific heat per particle are plotted in Figures 8.13, 8.14 as a function
of T /Tc . At large temperatures, as expected CV /N = 3/2, i.e. the classical value.
It increases as T is lowered, however in contrast to the ideal gas case, it then begins
to decrease as T approaches Tc .
86
Specific heat per particle for bosons
CV /N
1.6
1.5
1.4
1.3
1.2
1.1
1.1 1.2 1.3 1.4 1.5 1.6

T /Tc
Figure 8.14: Attractive boson in 3d: the specific heat per particle as a func-
tion of T /Tc .
87
CHAPTER 9
ENTROPY-TO-VISCOSITY RATIO OF QUANTUM GASES
The ratio of the shear viscosity η to the entropy density s has units of ~/kB in
any dimension. A lower bound was conjectured for 3d relativistic systems[26]:
η ~
≥ (9.1)
s 4πkB
based on the AdS/CFT correspondence. The bound is saturated in certain strongly
coupled supersymmetric gauge theories. It is therefore interesting to study this

ratio for non-relativistic, strongly interacting systems, in 3 and lower dimensions,
since it is unknown whether there really is a lower bound. This ratio was studied
for 3d unitary Fermi gases in [4, 30, 13, 46, 45].
9.1 2d
In two dimensions,the attractive fermion is the most interesting and well-behaved
case in our formalism, as was evident in chapter 6, so we study the issue in this
case in detail.
Consider first a gas with a single species of particle of mass m. For a non-
relativistic system in 2 spatial dimensions:
1
η = nv̄mlfree (9.2)
2
where v̄ is the average speed and lfree the mean free path. The mean free path is
√ √
lfree = 1/( 2nσ) where σ is the total cross-section. The 2 comes from the ratio
of the mean speed to the mean relative speed [43].
Recall that we defined the unitary limit for a repulsive gas as the case where
the scattering phase shift (3.36) taking value of −π. This implies the scattering
88
amplitude M = 2i/I. The accessible phase space volume I is m/4α, as given
in (3.31). Using the formula for differential cross-section (4.13), one can integrate
over all angle and get
16
σ= (9.3)
|k|
in the unitary limit, where k is the difference of momenta of the two scattering
particles. This gives

m 1 2
η= √ mv̄ (9.4)
16 2 2
Since the pressure is determined by the average kinetic energy, and the ideal
gas relation eq. (6.29) still holds for a unitary gas, the average kinetic energy per
particle is just ǫ1 = −F /n, eq. (6.30). Finally we obtain:
−1
η m F ∂F
= √ (9.5)
s 16 2 n ∂T
where all the quantities in the above formula are the single component values.
Attractive fermions
η/s
2.0
1.5
1.0
0.5
0 5 10 15 20 25 30
µ/T
Figure 9.1: The ratio η/s as a function of µ/T . The horizontal line is 1/4π.
In terms of the scaling functions:
η 3 c π b
= √ µ ′ = √ (9.6)
s ′
4 2π c (2c − T c ) 4 2 b′ (2b − Tµ b′ )
89
For two-component fermions the available phase space I is doubled, reflected
by α taking the value 1/2 instead of 1. The cross-section is then halved, since spin
up particles only scatter with spin down. And the entropy is doubled, assuming
symmetry of the two spin components. This gives
η η
=4 (9.7)
s fermi s bose
η/s
2.0
1.5
1.0
0.5
0.0
0.0 0.5 1.0 1.5 2.0
T /TF
Figure 9.2: The ratio η/s as a function of T /TF for attractive fermions.
The above expression is easily evaluated numerically using the expressions from
Chapter 6. The result is displayed for small values of µ/T in Figure 9.1. In this
regime, η/s is well above the conjectured bound, and comparable to the 3d values
extracted from the experimental data[46]. We find
η ~
lim = 7.311 (9.8)
µ/T →0 s 4πkB
This is well-below the values for common substances like water, however it is
comparable to values for liquid helium, which is about 9 times the bound[26]. In
90
the region shown:
η ~
≥ 6.07 (9.9)
s 4πkB
In Figure 9.2 we plot η/s as a function of T /TF , and one observes a behavior
quite similar both qualitatively and quantitatively to the 3d data summarized
in[46], where the minimal value is about 6 times the bound.
Recall that, in Chapter 6, it was argued that the region µ/T > 11.7 is unphysi-
cal, or perhaps metastable, since the entropy increases with decreasing temperature
there. Thus for the regions that are surely physical, the bound (9.9) holds.
Attractive fermions
η/s
0.14
0.12
0.10
0.08
1
4π
0.06
0.04
0.02
0
µ/T
1 ´ 106 2 ´ 106 3 ´ 106 4 ´ 106 5 ´ 106
Figure 9.3: The ratio η/s as a function of µ/T as µ/T gets very large.. The
horizontal line is 1/4π.
It is nevertheless interesting to study the viscosity in this unphysical region.

In the zero temperature limit, i.e. T /µ → 0+ , η/s slowly decreases and seems to
approach the bound. See Figure 9.3. However it eventually dips below it. See
Figure 9.4.
This behavior can be understood analytically as follows. For x = µ/T very

large, the solution to the equation (7.5) is approximately:
√ √
y(x) ≈ x + 1/2 + (log x − 1)/4 x. (9.10)
91
Attractive fermions
η/s
0.10
0.09
0.08
0.07
µ/T
5.0 ´ 106 1.0 ´ 107 1.5 ´ 107 2.0 ´ 107
Figure 9.4: The ratio η/s as a function of µ/T in the limit of very low tem-
peratures (top curve). The horizontal line is 1/4π. The bottom
curve is the approximation eq. (9.12).
This leads to the asymptotic expansions:
c ≈ 3(x2 + x log x − x)/π 2

(9.11)
mT
n≈ (x + log(x)/2)
2π
and
η π
≈ √ (1/(log(µ/T ) − 2) + T /2µ) (9.12)
s 2 2
In terms of the density:
η π
lim= √ (9.13)
µ/T →∞ s 2 2 log(2πn/mT )
It is important to note that although the energy per particle scaling function ξ
approaches the free field value at low temperatures, the above behavior is very dif-
ferent from the free field case. In the latter, the scaling function c = −6 Li2 (−z)/π 2 ,
which gives the diverging behavior at very low temperature:
η 3 µ
≈ √ , (free fermions) (9.14)
s 4 2π T
Finally, as µ/T → −∞, y ≈ 1 + z, and c ≈ 6z/π 2 and n ≈ mT z/2π. This
92
leads to exponential growth:
η πT mT 1
lim = − √ e−µ/T = − √ (9.15)
µ/T →−∞ s 2 2µ 4 2 n log(2πn/mT )
9.2 3d
We shall first extract the viscosity of a single-component quantum gas at the

unitary limit in a way similar to the previous two-dimensional case.
In kinetic theory, the shear viscosity can be expressed as
1
η = nv̄mlfree (9.16)
3
where v̄ is the average speed and lfree is the mean free path. (Note that the
numerical prefactor is different from the two-dimensional case (9.2).)
√
The mean free path is related to the total cross-section σ by lfree = 1/( 2nσ)
[43].
In the unitary limit the phase shift (3.36) is Θ = ±iπ. The scattering amplitude
is then
8πα
iM = − . (9.17)
m|∆k|
This leads to
m2 |M|2 16π
σ= = (9.18)
4π |k|2
where |k| is the momentum of one of the particles in the center of mass frame, i.e.
|k1 − k2 | = 2|k|. This gives
m3 v̄ 3
η= √ (9.19)
48 2π
93
Since the equation (6.30) is the same relation between the pressure and energy
of a free gas, and the pressure is due to the kinetic energy, this implies
1 2 3c
mv̄ = E/N = ′ T. (9.20)
2 2c
Since the entropy density s = −∂F /∂T , one finally has

√
η 3π c 3/2 1
= ′
(9.21)
s 8ζ(5/2) c 5c/2 − xc′
For two-component fermions, the same considerations as those in the two-
dimensional case apply, and the entropy-to-viscosity ratio is four times that of
single-component bosons.
Viscosityentropy for attractive fermions
η/s
0.8
0.6
0.4
0.2
0.2 0.4 0.6 0.8 1.0

T /TF
Figure 9.5: The viscosity to entropy-density ratio as a function of T /TF for

fermions. The horizontal line is 1/4π.
The ratio η/s for fermions as a function of T /TF is shown in Figure 9.5, and
is in good agreement both quantitatively and qualitatively with the experimental
data summarized in [46]. The lowest value occurs at x = 2.33, which corresponds
to T /TF = 0.28, and
η ~
> 4.72 (9.22)
s 4πkB
The experimental data has a minimum that is about 6 times this bound. In the free
fermion theory the minimum occurs at µ/T ≈ 2.3, which gives η/s > 7.2~/4πkB .
94
More recently, Cao et. al. [4] reported that the η/s ratio about five times the
conjectured lower bound, which is in good agreement with our result here.
Viscosity to entropy ratio for bosons
η/s
0.4
0.3
0.2
0.1
1.1 1.2 1.3 1.4 1.5 1.6

T /Tc
Figure 9.6: The viscosity to entropy-density ratio as a function of T /Tc for

bosons. The horizontal line is 1/4π.
For bosons, the ration η/s is plotted in Figure 9.6 as a function of T /Tc . One
sees that it has a minimum at the critical point, where
η ~
> 1.26 (9.23)
s 4πkB
Thus the bosonic gas at the unitary critical point is a more perfect fluid than that
of fermions. On the other hand, the ideal Bose gas at the critical point has a lower
value: p
η 3πζ(5/2) ~
= = 0.53 (9.24)
s 20ζ(3/2)3/2 4πkB
ideal
95
CHAPTER 10
SUMMARY AND CONCLUSION
Building on the original result due to Dashen [8], we have derived the S-matrix
formulation of quantum statistical mechanics. By resumming all two-body terms,
we arrive at the integral equation (3.20), reproduced below:
Z
f˜(k) = f0 (k) + f˜(k) sβf0 (k) (dk )G2 (k, k )f˜(k ) ,
′ ′ ′
(10.1)
which mimics the Yang-Yang equation of the Thermodynamic Bethe Ansatz

(TBA).
This integral equation is first applied to an weakly interacting Bose gas in two
dimensions. In Chapter 5 the critical chemical potential and density at a fixed
temperature is obtained for an attractive Bose gas:

mkB T 2π
nc ≈ log 1 +
2π~2 mg
(10.2)
mgT 2π
µc ≈ log 1 +
2π mg
In the weak coupling limit, the form of critical chemical potential is in agree-
ment with numerical simulation in [42], which is µc = mgT /2π log(2C/mg) where
C ≈ 13.2 numerically. Our result indicated that C = π. The discrepancy may be

attributed to the renormalization group running of g, which causes the effective
1/g to grow in the low-momentum end.
In Chapter 6, we have identified the unitary limit of a quantum gas as the

limit where the two-body scattering phase taking the maximal values of ±π. This
corresponds to the same g → ∞ limit in three dimensions, while for two dimensions
the same limit arises if an infrared cutoff is given, for example coming from a
96
finite system size. Anticipating that the quantum gas is scale-invariant at the
unitary limit, we have also defined various thermodynamics scaling functions. The
numerical solutions to our integral equation at the unitary limit in two and three
dimensions are presented in Chapter 7 and 8, respectively.
For the three dimensional fermion gas, we report a critical temperature Tc /TF ≈
0.1 for the superfluid transition at finite temperature. For the three dimensional
boson gas, we present evidence that the gas undergoes a strongly interacting version
of BEC at critical density nc λ3T ≈ 1.3.
Chapter 9 is devoted to the calculation of the ratio of shear viscosity to entropy

density η/s, which has a conjecture lower bound of
η/s ≥ ~/4πkB (10.3)
based on the relativistic AdS/CFT correspondence in three dimensions [26]. We

calculated this ratio of unitary quantum gas in two and three dimensions using our
integral equation.
In two dimensions, this ratio of attractive bosons drops below the bound by a
factor of 0.4, however this could be an artifact of our approximation. For attractive
fermions, the conjectured lower bound is reached at very large µ/T ≈ 107 , however
this was argued to occur in an unphysical or metastable region; in the physical
regions, η/s ≥ 6.07 ~/4πkB .
In three dimensions, we report ratios that are above the bound for both boson
and fermion gases. The fermionic ratio is 4.72 times the bound, while the bosonic
one is 1.26 times the bound. The recent finding by Cao et. al. [4] put the
experimental value for the fermionic ratio in the range of 0.4-0.5, in good agreement
with our result.
97
APPENDIX A
S-MATRIX AND RESOLVANT
In this section we shall review some elements of scattering theory, with the
ultimate goal of proving the relationship (2.5). For a more thorough introduction,
please see [44].
We restate here the definition of resolvants:

1
G(E) =
E − H + iη
(A.1)
1
G0 (E) = ,
e − H0 + iη
where η is a real positive infinitesimal, and the full Hamiltonian H can be split
into a free part H0 and an interaction part H1 . We have full knowledge of the
eigenstate and eigenvalues of H0 .
In addition, we shall assume that the eigenvalues of H0 are densed, and (without
loss of generality) the ground state energy is zero. So for any given E > 0, there
exists an eigenstate with energy E. Also we assume that the perturbation H1
connects the free and interacting Hilbert spaces adiabatically, so that there is a
one-to-one correspondence of states.
The interaction H1 may shift the ground-state energy. If the exact ground-state
energy of the full Hamiltonian H is ∆E 6= 0, we make the following modifications
so that the ground states of H0 and H have the same energy:
H0 → H0 + ∆E
(A.2)
H1 → H1 − ∆E
We want to solve the full time-independent Schroedinger’s equation for H:
H|Ψai = Ea |Ψa i, (A.3)
98
where a represents all necessary labels. A good starting point is to build |Ψa i from
its free counterpart |Φa i, which satisfies:
H0 |Φa i = Ea |Φa i. (A.4)
Subtracting (A.4) from (A.3), using H = H0 + H1 , and doing some rearrange-

ment, one may formally write:
1
|Ψa i = |Φa i − H1 |Ψa i (A.5)
H0 − Ea
Since the operator (H0 − Ea )−1 has poles along the real axis, we add to the
denominator the infinitesimal ±iη. Depending on our choice of the sign, two sets
of states emerges from two equations:
|Ψ+ +
a i = |Φa i + G0 (Ea )H1 |Ψa i
(A.6)
|Ψ−
ai = |Φa i + G∗0 (Ea )H1 |Ψ−
a i.
These equations (A.6) are known as the Lippmann-Schwinger equations. The state
|Ψ+ i corresponds to the incoming state, while |Ψ− i is the outgoing state. These
equations have formal solutions:
|Ψ+
a i = (1 + G(Ea )H1 ) |Φa i
(A.7)
|Ψ−
ai
∗
= (1 + G (Ea )H1 ) |Φa i
To see that the assignment of incoming and outgoing states make sense, we
consider the case of |Ψ+ i, and define a wave packet localized in time:
X
|Ψ+
a (t)i ≡ cκ e−iEκ t |Ψ+
κ i, (A.8)
κ
where the coefficient cκ is sharply centered around κ = a. It’s non-interacting

counterpart is
X
|Φa (t)i ≡ cκ e−iEκ t |Φκ i (A.9)
κ
99
Using the Lippmann-Schwinger equation (A.6), we can formally write (A.8) in
terms of the free eigenstates:
X
|Ψ+
a (t)i = |Φa (t)i + fn (t)|Φn i ;
n
X (A.10)
e−iEκ t
fn (t) ≡ cκ hΦn |H1 |Ψ+
κ i.
κ
Eκ − En + iη
The label κ contains energy and possibly other quantum numbers. Let us
explicitly display the energy dependences:
cκ → cK (E) ,
(A.11)
Rnκ ≡ hΦn |H1 |Ψ+
κi → RnK (E),
where K denotes all the other labels necessary to specify a quantum state. Then
the coefficient fn (t) is
XZ e−iEt
fn (t) = dE cK (E)RnK (E) (A.12)
K
E − En + iη
Assuming cK (E) and RnK (E) do not contain poles in E, we see that fn (t) = 0
as t → −∞. In the far-past, the wave packet |Ψ+
a (t)i coincides with the non-
interacting |Φa (t)i. We may thus identify the “+”-states with the incoming states
in a scattering process.
Similarly, one identifies the “−”-states with the outgoing states by using the
same argument as above. The only difference is that in (A.12) the sign of η is
reversed, reducing the state to the non-interacting version in the far-future instead.
The scattering matrix is defined to be the overlap of in- and out-states:
Sf i = hΨ− +
f |Ψi i. (A.13)
100
With (A.6) and (A.7), we can formally rewrite the inner product as:
hΨ− + + +
f |Ψi i = hΦf |Ψi i + hΦf |H1 G(Ef )|Ψi i
= hΦf |Φi i + hΦf |G0 (Ei )H1 |Ψ+ +

i i + hΦf |H1 G(Ef )|Ψi i (A.14)

1 1
= δf i + + hΦf |H1 |Ψ+
i i.
Ef − Ei + iη Ei − Ef + iη
To get to the last line, we made use of the fact that the free eigenstates form an
orthonormal set of basis, so that hΦf |Φi i = δf i . Also the operators G and G0 can
be replaced by their respective eigenvalues when acting on their eigenstates |Ψ+
i i
and hΦf |.
Making use of the identity (x + iη)−1 − (x − iη)−1 = 2πiδ(x), we can now recast
(A.14) into the familiar form:
Sf i = δf i + 2πi δ(Ef − Ei ) Tf i ;
(A.15)
Tf i ≡ hΦf |H1 |Ψ+
i i = hΦf | (H1 + H1 G(Ei )H1 ) |Φi i
Now we are free to take the energy of the initial state off-shell by the introduc-
tion of the operators Ŝ(E) and T̂ (E):
Ŝ(E) ≡ 1 + 2πiδ(E − H0 ) T̂ (E) ,
T̂ (E) ≡ H1 G(E)H1 , (A.16)
Sf i = hΦf |Ŝ(Ei )|Φi i.
We are now in the position to derive the equation (2.5), reproduced below:
Ŝ = G∗0 (G∗ )−1 GG−1

0 . (A.17)
To begin with, we make use of the following identities to rewrite (A.17), both
easily proven from the definition (A.1):
(G∗ )−1 G = 1 − 2iη G ,

(A.18)
G∗0 G−1 ∗
0 = 1 + 2iη G0 .
101
The right hand side of (A.17) then becomes:
G∗0 (G∗ )−1 GG−1 ∗ −1

0 = G0 (1 − 2iη G) G0
= (1 + 2iη G∗0 ) − 2iη G∗0 GG−1

0
(A.19)
= 1 + 2iη G∗0 (G0 − G)G−1

0
Next, we use G0 − G = −G0 H1 G to get:
G∗0 (G∗ )−1 GG−1 ∗ −1

0 = 1 − 2iη (G0 G0 )(H1 GG0 ) (A.20)
Finally, we can show that −2iη G∗0 (E)G0 (E) = 2πiδ(E −H0 ), and that GG−1
0 =
1 + GH1 . Substitute these identities into last equation, and we get:
G∗0 (G∗ )−1 GG−1

0 = 1 + 2πi δ(E − H0 )H1 (1 + GH1 ) = Ŝ(E) (A.21)
as required.
102
BIBLIOGRAPHY
[1] T. Bourdel, J. Cubizolles, L. Khaykovich, K. M. F. Magalhães, S. J. J. M. F.

Kokkelmans, G. V. Shlyapnikov, and C. Salomon. Measurement of the inter-
action energy near a feshbach resonance in a 6 li fermi gas. Physical Review
Letters, 91(2):020402, Jul 2003.
[2] A. Bulgac, Joaquı́n E. D., and P. Magierski. Spin 1/2 fermions in the unitary
regime: A superfluid of a new type. Physical Review Letters, 96(9):090404,
Mar 2006.
[3] E. Burovski, N. Prokof’ev, B. Svistunov, and M. Troyer. Critical tempera-

ture and thermodynamics of attractive fermions at unitarity. Physical Review
Letters, 96(16):160402, Apr 2006.
[4] C. Cao, E. Elliott, J. Joseph, H. Wu, J. Petricka, T. Schäfer, and J. E. Thomas.

Universal quantum viscosity in a unitary fermi gas. Science, 331(6013):58–61,
2011.
[5] Cheng Chin, Rudolf Grimm, Paul Julienne, and Eite Tiesinga. Feshbach
resonances in ultracold gases. Reviews of Modern Physics, 82(2):1225–1286,
Apr 2010.
[6] S. L. Cornish, N. R. Claussen, J. L. Roberts, E. A. Cornell, and C. E. Wieman.

Magnetic field dependence of ultracold inelastic collisions near a feshbach
resonance. Physical Review Letters, 85:1795, 2000.
[7] R. S. Crandall and R. Williams. Crystallization of electrons on the surface of

liquid helium. Physics Letters A, 34(7):404–405, 1971.
[8] R. Dashen, S.-K. Ma, and H. Bernstein. S-matrix formulation of statistical

mechanics. Physical Review, 187(1):345, 1969.
[9] K. Dieckmann, C. A. Stan, S. Gupta, Z. Hadzibabic, C. H. Schunck, and

W. Ketterle. Decay of an ultracold fermionic lithium gas near a feshbach
resonance. Physical Review Letters, 89(20):203201, Oct 2002.
[10] D. M. Eagles. Possible pairing without superconductivity at low carrier con-

centrations in bulk and thin-film superconducting semiconductors. Physical
Review, 186(2):456–463, Oct 1969.
103
[11] A. L. Fetter and J. D. Walecka. Quantum Theory of Many-Particle Systems.
Dover Publications, 1971.
[12] D. S. Fisher and P. C. Hohenberg. Dilute bose gas in two dimensions. Physical
Review B, 37(10):4936–4943, Apr 1988.
[13] B. A. Gelman, E. V. Shuryak, and I. Zahed. Ultracold strongly coupled gas:

A near-ideal liquid. Physical Review A, 72(4):043601, Oct 2005.
[14] M. Greiner, C. A. Regal, and D. S. Jin. Emergence of a molecular bose-einstein

condensate from a fermi gas. Nature, 426:537, 2003.
[15] C. C. Grimes and G. Adams. Evidence for a liquid-to-crystal phase transition

in a classical, two-dimensional sheet of electrons. Physical Review Letters,
42(12):795–798, Mar 1979.
[16] T.-L. Ho. Universal thermodynamics of degenerate quantum gases in the

unitarity limit. Physical Review Letters, 92(9):090402, Mar 2004.
[17] P. C. Hohenberg. Existence of long-range order in one and two dimensions.

Physical Review, 158(2):383–386, Jun 1967.
[18] M. Holzmann, G. Baym, J. P. Blaizot, and F. Laloë. The kosterlitz-thouless-

berezinskii transition of homogeneous and trapped bose gases in two dimen-
sions. Proceeding of National Academic of Science USA, 104:1476, 2007.
[19] P.-T. How and A. LeClair. Critical point of the two-dimensional bose gas: an
s-matrix approach. Nuclear Physics B, 824:415, 2010.
[20] P.-T. How and A. LeClair. S-matrix approach to quantum gases in the unitary
limit i: the two-dimensional case. Journal of Statistical Mechanics: Theory
and Experiment, pages –03025, 2010.
[21] P.-T. How and A. LeClair. S-matrix approach to quantum gases in the unitary
limit ii: the three-dimensional case. Journal of Statistical Mechanics: Theory
and Experiment, pages –07001, 2010.
[22] H. Hu, X. J. Liu, and P. D. Drummond. Equation of state of a superfluid fermi

gas in the bcs-bec crossover. EPL (Europhysics Letters), 74(4):574, 2006.
[23] S. Inouye, M. R. Andrews, J. Stenger, H. J. Miesner, D. M. Stamper-Kurn,
104
and W. Ketterle. Observation of feshbach resonances in a bose–einstein con-
densate. Nature, 392:151, 1998.
[24] E. B. Kolomeisky and J. P. Straley. Renormalization-group analysis of the

ground-state properties of dilute bose systems in d spatial dimensions. Phys-
ical Review B, 46(18):11749–11756, Nov 1992.
[25] V. E. Korepin, N. M. Bogoliubov, and A. G. Izergin. Quantum Inverse Scat-

tering Method and Correlation Functions. Cambridge Monographs on Math-
ematical Physics. Cambridge University Press, 1997.
[26] P. K. Kovtun, D. T. Son, and A. O. Starinets. Viscosity in strongly interact-

ing quantum field theories from black hole physics. Physical Review Letters,
94(11):111601, Mar 2005.
[27] M. Le Bellac. Thermal Field Theory. Cambridge University press, 1996.
[28] A. LeClair. Interacting bose and fermi gases in low dimensions and the rie-
mann hypothesis. International Journal of Modern Physics, A23:1371, 2008.
[29] D. Lee and T. Schäfer. Cold dilute neutron matter on the lattice. ii. results
in the unitary limit. Physical Review C, 73(1):015202, Jan 2006.
[30] P. Massignan, G. M. Bruun, and H. Smith. Viscous relaxation and collective

oscillations in a trapped fermi gas near the unitarity limit. Physical Review
A, 71(3):033607, Mar 2005.
[31] N. D. Mermin and H. Wagner. Absence of ferromagnetism or antiferromag-

netism in one- or two-dimensional isotropic heisenberg models. Physical Re-
view Letters, 17(22):1133–1136, Nov 1966.
[32] P. Nikolićand S. Sachdev. Renormalization-group fixed points, universal phase

diagram, and 1/n expansion for quantum liquids with interactions near the
unitarity limit. Physical Review A, 75(3):033608, Mar 2007.
[33] Y. Nishida and D. T. Son. Fermi gas near unitarity around four and two
spatial dimensions. Physical Review A, 75(6):3617, 2007.
[34] P. Nozières and S. Schmitt-Rink. Bose condensation in an attractive fermion

gas: From weak to strong coupling superconductivity. Journal of Low Tem-
perature Physics, 59(195), 1985.
105
[35] K. M. O’Hara, S. L. Hemmer, M. E. Gehm, S. R. Granade, and J. E. Thomas.
Observation of a strongly interacting degenerate fermi gas of atoms. Science,
298(5601):2179–82, 2002.
[36] A. Perali, P. Pieri, L. Pisani, and G. C. Strinati. Bcs-bec crossover at finite

temperature for superfluid trapped fermi atoms. Physical Review Letters,
92(22):220404, Jun 2004.
[37] E. M. Peskin and V. D. Schroeder. An Introduction to Quantum Field Theory.

Westview Press, 1995.
[38] D. S. Petrov, M. A. Baranov, and G. V. Shlyapnikov. Superfluid transition

in quasi-two-dimensional fermi gases. Physical Review A, 67(3):031601, Mar
2003.
[39] P. Phillips. Advanced Solid State Physics. Westview Press, 2002.
[40] W. D. Phillips. Laser cooling and trapping of neutral atoms. Reviews of

Modern Physics, 70:721, 1998.
[41] V. N. Popov. Functional Integrals in Quantum Field Theory and Statistical

Physics. Reidel, Drodrecht, 1983.
[42] N. Prokof’ev, O. Ruebenacker, and B. Svistunov. Critical point of a weakly

interacting two-dimensional bose gas. Physical Review Letters, 87:270402,
2001.
[43] F. Reif. Fundamentals of Statistical and Thermal Physics. McGraw-Hill, 1965.
[44] P. Roman. Advanced Quantum Theory. Addison-Wesley Series in Advanced

Physics. Addison-Wesley, 1965.
[45] G. Rupak and T. Schäfer. Shear viscosity of a superfluid fermi gas in the
unitarity limit. Physical Review A, 76(5):053607, Nov 2007.
[46] T. Schäfer. Ratio of shear viscosity to entropy density for trapped fermions
in the unitarity limit. Physical Review A, 76(6):063618, Dec 2007.
[47] D. T. Son. Toward an ads/cold atoms correspondence: A geometric realization

of the schrödinger symmetry. Physical Review D, 78(4):046003, Aug 2008.
106
[48] H. B. Thacker. Many-body scattering processes in a one-dimensional boson
system. Physical Review D, 14(12):3508–3519, Dec 1976.
[49] H. B. Thacker. Statistical mechanics of a (1 + 1)-dimensional quantum field

theory at finite density and temperature. Physical Review D, 16(8):2515–2525,
Oct 1977.
[50] S. Weinberg. Quantum Theory of Fields, volume 1. Cambridge University

PRess, 1995.
[51] C. N. Yang and C. P. Yang. Thermodynamics of a one-dimensional system

of bosons with repulsive delta-function interaction. Journal of Mathematical
Physics, 10pages, 1969.
107

PH 87

Uploaded by

Document Informationclick to expand document information

Copyright:

Available Formats

PH 87

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

PH 87

Uploaded by

Copyright:

Available Formats

THE S-MATRIX FORMULATION OF QUANTUM

STATISTICAL MECHANICS, WITH APPLICATION TO

in Partial Fulfillment of the Requirements for the Degree of

A novel formalism of quantum statistical mechanics, based on the zero-temperature

the lowest order approximation (“two-body approximation”) corresponds to the ex-

distinction in June 2003.

and it’s all thanks to you guys.

2 Quantum statistical mechanics: the S-matrix approach 12

3 TBA-like integral equation 24

4 Field theories for quantum gases 38

5 Critical point of the two-dimensional Bose gas 43

6 Renormalization group and the unitary limit 47

7 The two-dimensional quantum gas at unitary limit 61

9 Entropy-to-Viscosity Ratio of Quantum Gases 88

10 Summary and Conclusion 96

A S-matrix and Resolvant 98

1.1 The Feynman diagrams related to Thomas-Fermi screening. . . . . 3

2.1 Examples of contributions to two- and three-particle terms. . . . . 21

3.1 Graphical representation of the integral equation (3.15). The dou-

4.1 Ladder diagram resummation of the two-body scattering amplitude. 41

7.1 Attractive fermions in 2d: density as a function of µ/T . . . . . . . 63

1.1 Non-perturbative method in quantum field theories

filed theory action S, it must first be separated into two parts:

Then the perturbation theory is an expansion in the powers of the interaction

The initial application of this perturbative quatization scheme to the quantum

It is conceivable that a problem may be without a truly small parameter that

To tackle such problems theoretically, one may invent an alternative small

A well-known example of such resummation, though not necessarily known

to all in field-theoretic form, is the Thomas-Fermi screening of the electrostatic

The lowest-order correction to the interaction potential is represented by the

type of behavior can’t be captured by any truncated expansion.

Another example of non-perturbative phenomena in field theory is the self-

energy of a scalar particle at finite temperature in four dimensional spacetime

Fourier transform of the potential Ũ (k)

(c) The diagrams that needs to be re-summed.

Figure 1.1: The Feynman diagrams related to Thomas-Fermi screening.

condition on the self-energy. Clearly no truncated expansion in powers of g can

In this thesis, we will develop an alternative representation of the free energy

The motivation of our new formalism is two-folded. First, we seek a higher-

The Thermodynamic Bethe Ansatz (TBA) was introduced by Yang and

Here ǫ is the so-called pseudo-energy, and can be considered as the single-

scheme in the higher dimensions, that is a parallel, or a “generalization” of the

In our integral equation, all “quantumness” of the problem is confined within

1.3 Phase transition of a Bose gas in two dimensions

In two spatial dimensions, the famous Mermin-Wagner-Hohenberg theorem states

Nevertheless, a two-dimensional system can undergo a Berezinsky-Kosterlitz-

1.4 Unitary gas

Ultra-cold quantum gas attracts considerable attention for various reasons.

The layout of this thesis is as followed. In Chapter 2, we shall developed the

so-called S-matrix formulation of quantum statistical mechanics. A diagrammatic

In Chapter 3, the formulation will be developed further, and we shall specialize

Yang-Yang equation of TBA (1.3).

In Chapter 4, we turn our attention to the field theoretic model of quantum

We take a break from the development of the theoretical framework in Chapter

The field-theoretic analysis resumes in Chapter 6. There we will attempt to

In Chapter 9, the shear viscosity to entropy ratio of unitary quantum gas is

1.6 Notations and conventions