Topics in Ultracold Atomic Gases: Strong Interactions and Quantum Hall Physics
DISSERTATION
Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy in the Graduate School of The Ohio State University
By Weiran Li, B.S. Graduate Program in Physics
The Ohio State University 2013
Dissertation Committee:
Professor Tin-Lun Ho, Advisor
Professor Eric Braaten
Professor Jay Gupta
Professor Nandini Trivedi c Copyright by
Weiran Li
2013 Abstract
This thesis discusses two important topics in ultracold atomic gases: strong interactions in quantum gases, and quantum Hall physics in neutral atoms.
First we give a brief introduction on basic scattering models in atomic physics, and an approach to adjust the interactions between atoms. We also include a list of experimental probes in cold atom physics. After these introductions, in Chapter 3, we report a few interesting problems in strongly interacting quantum gases. We introduce the BCS-BEC crossover model and relevant many-body techniques at the beginning, and discuss the details of several specific systems. We find the Fermi gases across narrow Feshbach resonances are strongly interacting at low temperature even when the magnetic field is several widths away from the resonance. We also discuss an approach to describe the metastable repulsive branch of Bose and Fermi gases across the resonance, and find a stable region of repulsive Bose gas close to unitarity. Some studies in two dimensional Fermi gases with spin imbalance are also included, and they are closely related to a number of recent experiments.
In Chapter 4, we discuss quantum Hall physics in the context of neutral atomic gases.
After illustrating how the Berry phase experienced by neutral atoms is equivalent to the magnetic field in electrons, we introduce the newly developed synthetic gauge field scheme in which a gauge potential is coupled to the neutral atoms. We give a detail introduction to this Raman coupling scheme developed by NIST group, and derive the theoretical model of the system. Then we make some predictions on the evolution of quantum Hall states when an extra anisotropy is applied from the external trap. Finally, we propose some experiments to verify our predictions.
ii Acknowledgments
I still remember when I was a kid, on my way to kindergarten every day, my father used to ask me lots of questions about natural phenomena. Most of them ended up being rhetorical questions as expected, but I guess all those lectures gradually generated my passion in maths and sciences. Although seldom expressed, I admire him as my first science teacher, and many more of him. I am extremely lucky to have my mother taking good care of me, and to have my wife taking over this in the past few years. These two important women in my life have always been encouraging me to pursue what I am interested in. It was extremely hard time for my wife in the last year of my PhD study when I stayed in China, and I cannot imagine her being more supportive.
I am extremely grateful to my advisor Dr. Tin-Lun Ho, for supporting me in the past
five years. He has been very generous with his time and energy, trying to help me gain more insights in important physics problems. I really appreciate his effort in educating me in my PhD study. I am also very grateful to all my committee members, Dr. Eric Braaten,
Dr. Jay Gupta, Dr. Nandini Trivedi for all the advice these years. I would like to thank
Dr. Braaten especially, for his help in writing and revising this thesis.
In my PhD study, I really enjoyed the discussions with Dr. Shizhong Zhang, Dr. Zhen- hua Yu, and my collaborator Dr. Xiaoling Cui. They have been very generous with their time, always ready to help me with different kinds of problems. I consider them as my men- tors and wish them the best in their future career. I would also like to thank the hospitality from IASTU, especially from Dr. Hui Zhai’s group, during my visit between 2012-2013.
Last but not least, I thank all my friends in physics department, especially the condensed matter theorists on the second floor, to make my life colorful in Columbus. I will always
iii miss the fun we had.
Finally, I acknowledge the financial support from National Science Foundation and
DARPA.
iv Vita
April 24th, 1985 ...... Born—Harbin, China
July, 2008 ...... B.S., Tsinghua University, Beijing, China
Publications
Bose Gases Near Unitarity Weiran Li and Tin-Lun Ho Physical Review Letters 108, 195301 (2012)
Alternative Route to Strong Interaction: Narrow Feshbach Resonance Tin-Lun Ho, Xiaoling Cui and Weiran Li Physical Review Letters 108, 250401 (2012)
Fields of Study
Major Field: Physics
Studies in Theories on degenerate quantum gases: Professor Tin-Lun Ho
v Table of Contents
Page Abstract ...... ii Acknowledgments ...... iii Vita...... v List of Figures ...... viii
Chapters
1. Introduction ...... 1
2. Basics in ultracold atomic gases ...... 5 2.1 Scattering models in ultracold atoms ...... 5 2.1.1 General scattering theory, T -matrix ...... 6 2.1.2 Low energy scattering, s-wave scattering length and phase shift . . . 8 2.1.3 Zero range model and Fermi’s pseudo potential ...... 14 2.2 Feshbach resonances ...... 18 2.3 Probes in cold atoms experiments ...... 23 2.3.1 Direct Imaging ...... 24 2.3.2 Spectroscopy ...... 27
3. Strongly interacting quantum gases across Feshbach resonances . . . . . 30 3.1 Introductions to strongly interacting quantum gases and the BCS-BEC crossover 32 3.1.1 Superfluidity across BCS-BEC crossover ...... 34 3.1.2 Critical temperatures and ladder approximation in dilute quantum gases ...... 42 3.1.3 The “upper branch” of the quantum gases ...... 51 3.1.4 Summary ...... 54 3.2 Fermi gases across narrow Feshbach resonance ...... 57 3.2.1 Wide resonance and narrow resonance ...... 57 3.2.2 Strong interactions in Fermi gases across narrow resonance . . . . . 62 3.3 Repulsive Bose gases across Feshbach resonance ...... 69 3.3.1 Three body loss in Bose gases close to unitarity, “low recombination” regime ...... 70 3.3.2 Strongly repulsive Bose gases close to unitarity, “shifted resonance” 72 3.3.3 Equation of state and instabilities of Bose gas in a trap ...... 79 3.4 Two dimensional Fermi gases with spin imbalance ...... 82
vi 3.4.1 Fermi gases in two dimensions ...... 83 3.4.2 Thermodynamic quantities of two component Fermi gases in two di- mension ...... 87 3.5 Conclusions ...... 90
4. Rotating gases, synthetic gauge fields, and quantum Hall physics in neutral atoms ...... 93 4.1 Rapidly rotating Bose-Einstein condensates and quantum Hall physics . . . 94 4.1.1 Quantum Hall physics, Laughlin wavefunctions ...... 95 4.1.2 Rotating Bose-Einstein condensates, vortex array and quantum Hall regime ...... 98 4.2 Synthetic gauge field scheme ...... 102 4.2.1 Berry phase in adiabatic states, Abelian gauge field ...... 103 4.2.2 NIST scheme of Abelian synthetic gauge field ...... 106 4.2.3 Non-Abelian gauge fields, spin-orbit coupled gases ...... 112 4.2.4 Summary ...... 117 4.3 Quantum Hall physics of atomic gases with anisotropy ...... 118 4.3.1 Single particle wave functions of particles in rotating anisotropic traps 119 4.3.2 Transition from a condensate to a quantum Hall state ...... 121 4.3.3 Quantum Hall wave functions in broken rotational symmetry . . . . 126 4.3.4 Detection of quantum Hall wave functions in cold gases ...... 130 4.4 Conclusions ...... 134
Bibliography ...... 137
Appendices
A. Path integral formalism of BCS-BEC crossover ...... 145
B. Adiabatic states and their gauge potential in spatially varying magnetic field ...... 149 B.1 General adiabatic states in spatially varying magnetic field ...... 149 B.2 Gauge potentials associated with the adiabatic states ...... 152
vii List of Figures
Figure Page
2.1 A sketch of phase shifts for different scattering lengths...... 13 2.2 A sketch of the s-wave scattering length and the effective range for a square well potential...... 16 2.3 A schematic diagram of the potentials in open and closed channels near a Feshbach resonances...... 19 2.4 Experimental images from time-of-flight (TOF) expansion, in Bose and Fermi gases in optical lattices...... 25
3.1 A sketch of pairing gap and chemical potential for a Fermi gas across unitarity at T =0...... 38 3.2 Class of diagrams included in the ladder approximation. All the legs of the ladder, i.e. the propagators on both ends of the interaction lines, run in the same direction...... 43 3.3 The scattering vertex for the particle-particle channel and particle-hole chan- nel. For dilute gases, particle-hole channel is usually negligible...... 44 3.4 A sketch of contour deformation of Matsubara sum in NSR formalism. . . . 47 3.5 A sketch of the superfluid transition temperature and the chemical potential at Tc for a Fermi gas in BCS-BEC crossover...... 50 3.6 The phase shifts ζ(E) for scattering in the presence of the bound state: the phase shift of the scattering state is modified such that it starts from 0 at the scattering threshold...... 52 3.7 The no-pole approximation in terms of excluding the bound-state contribu- tion, from modifying the integral region after doing the Matsubara sum. . . 54 3.8 A schematic diagram for the Gor’kov-Melik-Barkhudarov (GMB) correction. The scattering matrix Γ is different to the simple ladder approximation, by including one higher order of density fluctuation...... 55 3.9 A comparison of the calculated transition temperature Tc from different ap- proaches: Leggett BCS mean field, GMB mean field, and NSR...... 56 3.10 An illustration of the difference between wide and narrow Feshbach reso- nances in a Fermi gas with Fermi energy EF ...... 60 3.11 Schematic diagram of the relation between wide and narrow resonances: their −1 occupations in the space ((kF r∗) , kF abg)...... 61
viii 3.12 The scattering phase shift δ(k) as a function of incoming wave vettork for wide (A) and narrow (B) resonances...... 63 3.13 A sketch of second virial coefficients b (A) and “interaction energy” (B) − 2 int as a function of magnetic field, across a narrow Feshbach resonance. . . . . 66 3.14 An example of the s-wave scattering length (upper panel) and the interac- tion energy (lower panel) of a Fermi gas near a narrow resonance at low temperatures. The system is quantum degenerate at T = 0.5TF ...... 67 3.15 Experimental data from the Penn State group for fermionic 6Li gases across the 543.25G Feshbach resonance at different temperatures...... 68 3.16 A sketch of how bosonic and fermionic media affect the formation of dimers. The occupation and quantum statistics play important roles in the molecule formations...... 75 3.17 Energy density of an upper branch Bose base across the unitarity, and the critical scattering length for dimer formations in a Bose medium. Both are at T = 4TF ...... 77 3.18 The “phase diagram” of a homogeneous upper-branch Bose gas with fixed density n. The system is divided into three parts...... 78 3.19 “Phase diagram” of an upper branch Bose gas in a trap at fixed temperature T and trap frequency ω. A global view of density profile for any gases can be obtained in this (µ/T )-(λ/as) plane...... 80 3.20 Three typical density profiles for stable and unstable upper-branch Bose gases in a harmonic trap...... 81 3.21 Data of the Cambridge experiment in attractive and repulsive polarons: the interaction energy for both branches, and the lifetime for the repulsive branch. 84 3.22 The interaction energy for the attractive branch (red curve) and the repulsive branch (blue curve) at a high temperature T = 6TF , for equal spin populations. 88 3.23 Spin susceptibility and compressibility for upper branch two dimensional Fermi gas with equal population, at a temperature T = 6TF ...... 90 3.24 A diagram of “stability” of repulsive branches at different temperatures and polarizations...... 91
4.1 A sketch of the origin of integer quantum Hall effects...... 97 4.2 A sketch of the energy levels of Fock Darwin states for particles in rotating traps...... 100 4.3 A sketch of the origin of the gauge field in the presence of the spatial depen- dent magnetic field: the adiabatic states follow the orientation of the external magnetic field, and thus experience the Berry phase...... 105 4.4 Schematic figure of experimental setup in NIST. The Raman coupling is realized by two counter-propagating laser beams...... 107 4.5 The direction of the vector Beff from the effect of Raman coupling plus a linear detuning in magnetic field...... 109 4.6 Profiles of the vector potential and the synthetic magnetic field generated from the NIST setup...... 110 4.7 Schematic diagrams of energy levels in both Abelian and non-Abelian syn- thetic gauge fields...... 113 4.8 A sketch of the energy spectrum of two branches for spin-orbit coupled gases. 116
ix 4.9 Phase diagram of the cloud in rotating anisotropic traps: two phases of the BEC and QH are identified...... 124 4.10 Distortion and transition in vortex lattices of a rapidly rotating BEC in an anisotropic trap...... 125 4.11 Density profiles at cut x = 0 and y = 0 of the Laughlin wavefunctions at different positions in figure 4.9...... 127 4.12 Distribution of zeros in the relative coordinate of the ground state of two body problem in rotating anisotropic traps...... 128 4.13 Variation in the distance of splitting zeros as a function of inverse interaction strength, for fixed anisotropy α = 0.96...... 129 4.14 Density profiles after TOF expansion of the gas in a rotating isotropic trap. 132 4.15 Density profiles after TOF expansion of anisotropy α = 0.94...... 133 4.16 Second order correlation in equation 4.64, in a rotating isotropic (left) and anisotropic (right) trap...... 134 4.17 A sketch of high-resolution in situ imaging technique in neutral atoms. . . . 135 4.18 In situ images of the quantity ψ(r/2, r/2) for an anisotropy α = 0.96. . 136 | − |
x Chapter 1 Introduction
The concept of Bose-Einstein condensation (BEC) can be traced back to 1924-1925, when
Bose and Einstein first predicted that a macroscopic occupation of a single quantum state occurs below a critical temperature in non-interacting Bose systems. After that, superfluid helium was the only example of BEC in nature for decades. However, in helium-4, the density of helium atoms is so high that the system is in the strongly interacting regime.
The fraction of the condensate part is thus less then 10% even when the temperature is far below the critical value. Nevertheless, important theoretical problems have been studied in weakly interacting Bose gases in the condensate phase, even in the absence of a single realistic system of which the models are exactly suitable.
After many years of effort in cooling down the dilute gases of neutral alkali atoms, in
1995, scientists from JILA[1] realized the first weakly interacting dilute Bose system, in which almost 100% of the particles are in the condensate phase. The first gaseous BEC was in a dilute Rubidium gas at a temperature of 170nK, followed by the realizations of
Lithium[2, 3] and Sodium[4] condensates. This breakthrough of cooling down the quantum gases to degenerate and condensation limit brings the area of ultra cold atomic gases, or
“cold atoms” into the front stage of modern physics. The dilute BECs not only can be used to verify so many well developed theories, but also open up new directions of important physical problems[5]. In the past two decades, BECs are realized in a lot more isotopes of alkali, alkali earth[6], as well as rare earth atoms[7, 8]. And also, there are extensive efforts in cooling down the dilute Fermi gases to degenerate limit[9]. These successes have been used
1 to study the superfluidity of paired fermions, analogous to the Cooper pair condensation in conventional charged superconductors.
Implementation of Feshbach resonances in scattering between atoms makes it possible to study the Fermi gases in strongly interacting regime. By applying an external magnetic field, the inverse s-wave scattering length as can be tuned continuously through the scattering resonance, at which the scattering length diverges. The resonance point is also referred to as the unitarity. To the two sides of the resonance, the scattering length has different signs, and a two body bound state appears on the positive scattering length side. The
Feshbach resonance technique gives us opportunities to precisely control the scattering length, or the interaction strengths between atoms. This provides a platform to study some interesting problems, among which is the well addressed problem back in 1970’s, the
BCS-BEC crossover[10]. In the BCS-BEC crossover problem, the properties of superfluid
−1 states change gradually as as increases continuously. It is found that there is a continuous crossover between the superfluidity in the BCS limit ((k a )−1 ) and the BEC limit F s → −∞ ((k a )−1 + ). In BCS, the Cooper pairs form the constituent of the condensate, while F s → ∞ in the BEC limit, each two fermions form a deep bound bosonic molecules such they undergo
Bose condensation as a weakly interacting Bose gas. There is no sharp transition between these two types of superfluids, since the systems in the region connecting these two limits can be described by a single class of wave functions with gradually changing parameters.
The transition temperature of superfluidity close to unitarity has a large ratio of Tc/TF
(almost to 20%) where TF is the Fermi temperature of the system. This system has the
largest Tc/TF value in all the known substances in the universe, and has at least an order
of magnitude higher than other high Tc superfluids, e.g. quark-gluon plasma, and cuprates. Besides the large attractions in the Fermi gases close to unitarity, it is also very in-
teresting to study other important questions in quantum gases with repulsive interactions.
Among them there is an interesting question as the existence of the Stoner ferromagnetism
in uniform systems. In the context of cold atoms, it is known that deep in the BEC
side where the scattering length is small and positive, the “true ground state” is the Bose
gas composed of diatomic molecules. However, there is a metastable branch (the “upper
2 branch”) free of these low-lying bound states. Such atomic gas is effectively weakly repulsive if the underlying tight binding dimers are not populated. The diluteness of the quantum gas assures the system persists in this well defined thermodynamic states on a time scale that long enough for the system to equilibrate by two-body collisions. In this time scale, three body recombinations from which bound states are generated are negligible. While there are well developed theories for weakly interacting Bose and Fermi gases, including the
Bogoliubov theory and Lee-Huang-Yang corrections, for larger scattering lengths, there is no well accepted theory in describing this upper branch. Even the existence of the upper branch at finite temperature is under debate[11, 12, 13, 14].
Also with the capability of precisely controling in the interactions, and the external potential (engineering of overall traps and optical lattices), one may better study some important models in condensed matter physics. In conventional solid state materials, the complications brought by impurity etc makes it difficult to observe the behavior of a “pure model”. This concept of quantum simulation is one of the major worldwide efforts in cold atom laboratories. It aims to build platforms as closely as possible to what a quantum model exactly describes. For instance, it is intriguing to study the low temperature properties of two dimensional Hubbard models and quantum spin models, including their transitions between the normal and superconducting or other magnetic phases. These models are important because they are what many theories in strongly correlated systems are based on. Also, with the tunability of interactions between atoms and the internal degrees of freedom of the atoms, there are potentially some exotic models and quantum phases which are absent in the conventional solid state materials.
On the other hand, apart from studying the broken symmetry phases, cold atoms also enable us to study other nontrivial strongly correlated states, for instance the quantum
Hall states. It has been shown that atoms in rotating quantum gases are analogous to charged particles in the presence of a magnetic field, for the Coriolis force in the rotating frame provides the equivalence of the Lorentz force[15]. In fast rotations, as many as a hundred vortices have been observed in a Bose gas[16], and they form a triangular lattice.
It is conceivable if the rotation frequency closely approaches the trapping frequency, the
3 single particle spectrum is almost flat. The number of vortices will be in the same order as the number of particles, and the lattice will melt in this situation. The vortices finally become invisible as the angular momentum quanta are attached to the particles themselves.
This system with quantum flux bounded to particles is then in a topologically non-trivial quantum Hall (QH) regime. This transition to a quantum Hall state cannot be described by Landau’s phase transition theory, since there is no local order parameter in the quantum
Hall regime. The quantum Hall physics has been mostly studied in two-dimensional electron gases in semiconductor heterostructures, and it is usually very hard to grow a clean sample to verify the predictions of any theoretical models. In cold atoms, one has the advantage of building the sample in a more controllable way. Also, with the ability to implement the internal degree of freedom–the hyperfine states of alkali atoms–it is possible to study some exotic quantum Hall states in cold atoms that are absolutely absent in electrons.
This thesis is organized as following. In Chapter 2, we give a brief introduction to some theoretical and experiment backgrounds in cold atomic gases. We include the basic scattering model widely used in dilute quantum gases, the Feshbach resonance technique to tune the interactions, as well as useful probes in cold atoms experiments. In Chapter 3, we focus on an important aspect of quantum gases: the strong interactions. We give a review of studies in modeling BCS-BEC crossover in Fermi gases across the Feshbach resonance, and present a derivation of the important ladder approximation in dilute gases. Based on these knowledge, we illustrate the strong attractive interactions in narrow Feshbach resonances, and predict some thermodynamic properties for Bose gases near unitarity in both upper and lower branches. We also address some issues in two dimensional Fermi gases with spin imbalance which is related to some recent experiments. In the last Chapter 4, we
first introduce how quantum Hall physics comes into the rotating quantum gases, and then briefly review a recent developed synthetic gauge field scheme with the goal to couple the artificial gauge potential to neutral atoms. Finally we study some interesting properties of quantum Hall states in the presence of geometric distortion of external potentials.
4 Chapter 2 Basics in ultracold atomic gases
In this chapter we give an introduction to some important background knowledge of cold atomic gases. These include the basic scattering problems between the neutral atoms, and the model we use in many-body hamiltonian. It is also necessary to understand the
Feshbash resonance, by which the interactions between particles can be tuned to any value.
Finally in this chapter, some useful experimental probes are listed, and they are compared to experimental approaches in conventional condensed matter physics.
2.1 Scattering models in ultracold atoms
The first step to theoretically study the macroscopic quantum phenomena in quantum gases, is to understand the few-body atomic system. In quantum liquids like liquid helium, the range of interaction is comparable to the interparticle spacing. In contrast, the class of
“quantum gases” we focus on in this thesis have an interparticle spacing much larger than its interaction range. The diluteness of quantum gases ensures that two-body scattering process is the most important element to building up the interaction hamiltonian1. In this section we will focus on the two-body scattering models between neutral atoms.
1In fact in cold atomic gases, especially bosonic gases, there is a family of Efimov states with interesting discrete scaling invariance. They affect the three body recombination process in the atomic gases and sometimes have to be considered more carefully. However we will not go into details of the three body states in this thesis, and they can be found in this review paper [17], and references therein.
5 2.1.1 General scattering theory, T -matrix
Understanding the basic scattering models for two-body collisions is one of the most im- portant, and still active areas in studies on cold atomic gases. A clear and thorough under- standing of two-body physics is a crucial first step to building up a many-body hamiltonian, for any system.
Here we start from introducing the general scattering theory for a single-channel scatter- ing model, formulating the problem to solve and relate them to physical observables later.
For a two-body problem, we separate out the center-of-mass degree of freedom as usual.
In a translation invariant environment2, we are only interested in motions in the relative
coordinates. Further, when we approximate the scattering into single-channel model, the
internal degree of freedom is also frozen out, hence we only consider the elastic scattering
processes. The simplest starting point we have now, is an effective single-particle problem
in the relative motion frame, and in the presence of the scattering potential.
For a short-range potential the hamiltonian is written as H = H0 + V , where H0 is the non-interacting kinetic energy and V is the interaction potential which vanishes at large distances. The asymptotic form of the wave function at infinitely large distance in the relative coordinate is characterized by an “incident wave” ψ , which is a solution of | ini Schr¨odinger’sequation for H . The full wave function ψ with total energy E consists of 0 | i an incident wave part and a scattering wave part ψ , which satisfies the equations below: | sci
ψ = ψ + ψ , (2.1) | i | ini | sci (E H) ψ = 0, (2.2) − | i (E H ) ψ = 0. (2.3) − 0 | ini
Using the Lippmann-Schwinger equation, we can define the scattering T -matrix3 in this
2Even though most cold atom experiments are conducted in harmonic traps, the smoothness of the overall potential enables one to neglect the broken translational symmetry. 3In this thesis we sometimes use the phrase “scattering matrix” to refer to the T -matrix. Actually the terminology of scattering matrix can be referred to as the unitary S-matrix, which relates to T -matrix as S = 1 − 2ikT .
6 model by
ψ = ψ + ψ = (E H )−1V ψ | i | ini | sci − 0 | i = (E H )−1T ψ , (2.4) − 0 E| ini where the scattering T -matrix at energy E is defined as
T ψ V ψ . (2.5) | ini ≡ | i
By left multiplying the interaction operator V on both sides of equation 2.4, we have
V ψ = T ψ = V (E H )−1T ψ , (2.6) | i E| ini − 0 E| ini
TE = VG0(E)TE, (2.7)
where G (E) (E H )−1 is the free Green’s function at energy E. Equation 2.7 above 0 ≡ − 0 gives a relation between the T -matrix and the interacting potential of the system, and
is the starting point of many useful models. As we can see, the scattering T -matrix can
be expressed in a geometric series with all the orders of interaction operator V , which
corresponds to itinerating the scattering process between the two particles:
V T = V + VG (E)V + = . (2.8) E 0 ··· 1 G (E)V − 0 The approximation that keeps only the first order of V in the T -matrix is called the
Born approximation. In the weakly interacting limit, namely when G (E)V 1, the Born 0 approximation provides a good estimate of many scattering quantities. The full expression
above is actually in the matrix form, to the extent that it can be sandwiched by any pair
of states, for instance the scattering between plane-wave states with wave vectors k and
0 0 0 2 2 2 02 k can be written as k TE k TE(k, k ). The condition that E = ~ k /2µ = ~ k /2µ, h | | i ≡ where µ here is the reduced mass of the two particles, is often called the on-shell condition.
Another very useful decomposition of the T -matrix is by partial waves, where the matrix
elements are defined by
0 l, k T l0, k0 T l,l (k, k0), (2.9) h | E| i ≡ E
7 in which l, k is the partial wave with angular momentum l and wave number k. In an | i l,l0 l,l0 isotropic potential with spherical symmetry, T = δl,l0 T is diagonal in angular momen- tum sectors. In this thesis we mostly focus on the s-wave scattering problems, namely when only the l = l0 = 0 sector of the T -matrix is important.
2.1.2 Low energy scattering, s-wave scattering length and phase shift
From the Lippmann-Schwinger equation 2.4, we write it down in the first quantized form by left multiply the bracket r 4, and assume the incident wave as a plane wave with momentum h | k. The full wave function in real space then takes the form
eikr ψ(r) = eikr + f(θ, φ) , (2.10) r where θ, φ are the polar and azimuthal angles with respect to the direction of the incident wave vector. The second term in the full wave function corresponds to outgoing spherical waves. For a spherically symmetric potential, the scattering amplitude f is independent of the azimuthal angle φ, and the differential cross section for scattering is given by
dσ = f(θ) 2. (2.11) dΩ | |
With spherical symmetry, the scattering wave function can also be expanded by partial waves, into the form ∞ ψ = AlPl(cos θ)Rkl(r), (2.12) Xl=0 in which Pl’s are Legendre polynomials, and the radial part Rkl satisfies the equation
00 2 l(l + 1) 2µ R + R0 + k2 V (r) R = 0, (2.13) kl r kl − r2 − 2 kl ~ where V (r) is the first quantization form of the scattering potential. At large distance where
the potential vanishes, the solution of Rkl is a linear combination of spherical Bessel and 4To get the following form with “out-going” wave only, we have to add an infinitesimal imaginary part −1 + −1 in the Green’s function, i.e. substitute (E − H0) by the form (E − H0 + i0 ) .
8 Neuman functions jl and nl. The general form of the solution at large distance reads
sin(kr lπ/2 + δ ) R (r ) − l , (2.14) kl → ∞ ≈ kr in which δl is the phase shift for the l-th partial wave. The ratio between the Neuman
function nl and Bessel function jl at large distances is given by tan δl. In the non-interacting limit where the potential is extremely weak, the asymptotic form of wave function at large
distance is a pure Bessel function, and the phase shift vanishes δ 0. l ≡ By comparing the asymptotic form above and the expression of f(θ) in equation 2.10,
we have
∞ ∞ 1 f(θ) = (2l + 1)(e2iδl 1)P (cos θ), (2l + 1)f P (cos θ) (2.15) 2ik − l ≡ l l Xl=0 Xl=0 with the l-wave scattering amplitude
f (e2iδl 1)/2ik. (2.16) l ≡ −
2 The scattering amplitude relates the scattering T -matrix as fl = Tlµ/2π~ . And the total − cross section of the scattering is given by
∞ ∞ 4π 4π σ = (2l + 1) f 2 = (2l + 1) sin2 δ . (2.17) k2 | l| k2 l Xl=0 Xl=0 Consider a simple finite-range scattering model, with the phase shift δl determined by the boundary condition at small distance. At short distances, the asymptotic form of jl and nl are j (kr) (kr)l, n (kr) (kr)−l−1, at kr 0. (2.18) l ∼ l ∼ →
If the boundary condition is enforced at r = r0 where r0 is the range of potential, the
2l+1 relative ratio between nl and jl will be proportional to (kr0) . In low-energy scattering, namely kr 1 limit, it is natural to see tan δ δ (kr )2l+1, hence the s-wave scattering 0 l ∼ l ∼ 0 (l = 0) is dominant at this energy scale for it is the leading term of kr0 in f. Physically, it corresponds to the fact that a low-energy incident wave, whose wavelength is much larger
than the potential itself, will not be able to probe the detailed structure of the potential.
9 All the information we have here is an almost isotropic cross section of outgoing waves.
In the s-wave channel, since the phase shift is linear in k, we can define the opposite slope as the s-wave scattering length as:
δ = ka , for k 0. (2.19) 0 − s →
The s-wave cross section 4π σ = sin2 δ = 4πa2 (2.20) k2 0 s is the same as if there is a hard sphere with radius as. A more complete expression which relates the s-wave scattering amplitude and the phase shift can be derived from equation
2.16. And in low energy limit, it can also be expanded in powers of k2 as the following5:
1 r∗ f −1(k) = k cot δ(k) ik = + k2 + O(k4) ik (2.21) − −as 2 −
∗ ∗ where r is called the effective range of the scattering potential. The value of as and r are determined by the microscopic properties of the scattering potential.
The scattering length and phase shift are the most important quantities to determine a lot of low-energy physical properties of the system. For example we illustrate here the energy shift of the system as an example. We take a model that the particles are loaded in a hard-wall spherical container with radius R, and the boundary condition is enforced at the surface of the container that ψ(R) 0. In the absence of the interaction potential ≡ between particles, the spectrum takes a set of levels with
nπ k(0) = . (2.22) n R
The range of n is given by the density of the system such that the largest n = N satisfies
Nπ kF = R . First we discuss the “weakly interacting” limit with small scattering length, i.e. k a 1. In this limit when we put the interaction in, a nonzero phase shift emerges and F s 5The series expansion is in powers of k2 in most of the cases, since the scattering matrix is analytical in E. However, for some power-law decaying potentials V ∼ r−n, even when n is large enough such that the phase shift and scattering length are well defined for s-wave (and other partial waves for 2l + 3 < n), the scattering amplitude may have logarithmic dependence on k. See [18, 19] for more details.
10 can be approximated by δ = ka . The corresponding eigen wave vectors have to satisfy − s
k R + δ(k ) = k (R a ) = nπ, (2.23) n n n − s which gives nπ k = . (2.24) n R a − s The corresponding shift in energy levels are given by
2 2 2 (0)2 2 2 ~ kn ~ kn ~ 2 2as 2π~ as (0) 2 ∆En = (nπ) 3 = ψn (0) , (2.25) 2mr − 2mr ≈ 2mr R mr | | where ψ(0) = 1 sin nπr is the wave function of non-interacting systems. The ap- n (2πR)1/2r R proximation δ = ka we use in the previous derivation is based on the assumption that − s the scattering length a Nπ is very small, i.e. the system is in the weakly interacting s R limit with k a 1. In this limit, the effective interaction hamiltonian can be written in F s a delta-function form: 2 2π~ as V (r) = gδ(r), g = , (2.26) mr in which δ(r) is the three dimensional Dirac-delta function, g is the effective coupling con- stant proportional to the s-wave scattering length. This effective hamiltonian is essentially in the perturbative level.
From the discussion in the previous paragraph, we see the physical quantities, such as the energy level shift is determined by the scattering length or the phase shift in the limit k a 1. For the strongly interacting regime, i.e. k a 1 or larger than unity, F s F s ∼ the approximation δ(k) = ka cannot be used in the region 0 < k < k . Instead the − s F phase shift has to be determined by the full expression in 2.21. In this case, one can look at the change of the total energy by virial expansion to the second order of fugacity z = eβµ[20], which gives a flavor of how interacting thermodynamic quantities behaves in the high-temperature and low-fugacity eβµ 1 limit. The grand thermodynamic potential in this expansion to the second order is
Ω = T log 1 + z e−βE1n + z2 e−βE2n , (2.27) − " n n # X X
11 which is approximated to the second power of fugacity. E1n and E2n are the energy levels of
2 −βE2n single-particle and two-body states. We have to calculate Z2 = z n e and subtract the value of a non-interacting system to get the contribution fromP two-body interactions.
As we have the expressions for the eigenstates of the interacting problem in a spherical box (0) (0) kn and the non-interacting ones kn as in 2.23 and 2.22, the difference between Z2 and Z2 is
2 −βE −βE(0) ∆Z = z e 2n e 2n , (2.28) 2 − n X 2 2 (0) 2 (0)2 with E2n = ~ kn/2mr and E2n = ~ kn /2mr. As is shown that kn + δn = nπ, we have
dδ dn = R + dk. (2.29) dk And the sum over n becomes an integral over k, such that
∞ dk dδ 2 2 ∆Z = z2 e−β~ k /2mr z2b , (2.30) 2 π dk ≡ 2 Z0 6 where b2 is defined as the second virial coefficient for the scattering states . We will show
later the magnitude of b2 relies on the phase shift structure, and the more abrupt the
δ changes, the large b2 is. From the expansion of k cot δ in 2.21, we can calculate the
second virial coefficient b2 as a function of the scattering length, even at unitarity when as diverges[21].
We simplify the discussion here by approximating the expansion of the inverse scattering
amplitude in 2.21 to the first order of k, i.e. to use the following expression to determine
the phase shift 1 k cot δ = . (2.31) −as A plot of phase shifts at different negative scattering lengths is shown in figure 2.1. They cover a large range of a from the weakly interacting regime k a = 0.1 to extremely s F s − strongly interacting case k a = 100. The phase shifts for positive scattering lengths are F s − mirror symmetric to the horizontal axis in the plot, namely they only differ by a minus
6Here we neglected all the possible bound states in the two-body channel when we sum over all the eigenstates of the system, since we focus on the scattering state properties at present.
12 Π 2
L Π k H 4 ∆
0 0 0.5 1 1.5 2
kkF
Figure 2.1: A sketch of phase shifts as a function of scattering momentum, for different scat- tering lengths from expression 2.31. Different curves represents different scattering lengths, and from the lowest to highest they correspond to k a = 0.1, 0.5, 1, 10, 100. For F s − − − − − small kF as, the phase shift is almost linear, while for large kF as, it quickly saturates to π/2. The phase shift structures for positive scattering lengths only differ by a sign in δ, and thus are mirror symmetric to the horizontal axis.
sign as in the case for negative scattering length. As we can see, in the weakly interacting case, the phase shift is linear in k up to several Fermi momentum, while in the strongly interacting regime, it quickly saturates to π/2 at small momentum.
From the analysis above, the second virial coefficient b2 can be calculated analytically in two extreme cases: k a 0 and k a . The former is realized by approximating F s → F s → ∞ ∂δ/∂k = a , and for the latter, the approximation is made as ∂δ/∂k = 0 only in a small − s 6 region of momentum k < k∗ k . Simplify the integral in 2.30, we have the following F expressions:
a b(0) = s , (2.32) 2 −√2λ sgn(a ) 2 b∞ = s (1 erf(x))ex , (2.33) 2 − 2 −
(0) ∞ where b2 and b2 are second virial coefficients for weakly and strongly interacting cases respectively. In the expressions, x = λ/(√2πas) with λ = h/√2πmrkBT as the thermal wavelength of the system, and erf(x) is the error function. For the weakly interacting case, b is asymptotically first order in a /λ 1. And the limit that a (x 0) gives 2 s s → ∞ → 13 b = sgn(a )/2 = 1/2, in the order of unity. Thus we conclude the strongly interacting 2 − s ± regimes has a much larger b2 value than the weakly interacting regime. Finally, the contribution from interactions to the thermodynamic quantities are related to the virial coefficients as following:
T P = P + 2√2z2 b , (2.34) 0 λ3 2
3T n 3 b2 √2 T ∂b2 = 0 + (nλ ) + , (2.35) 2 "−√2 3 ∂T # where P and are the pressure and energy density respectively, subindex 0 stands for non- interacting values. By substituting the explicit form of b2, one finds that the magnitude of interacting pressure and energy are monotonic functions of b . The large absolute | 2| value of as close to unitary does give a much stronger interaction than small ones in the weakly interacting limit[21]. In the following chapters, we will find these from many-body calculations as well.
All the discussions above are about the s-wave scattering, which is based on the fact the l = 0 channel is dominant in the energy scale at kr 1. As in the dilute gases, 0 since the energy scale is given by the Fermi wave vector k , and it satisfies k r 1, by F F 0 these discussions we conclude that the physical quantities in such systems are pretty much determined by the s-wave scattering length and the phase shift structure.
2.1.3 Zero range model and Fermi’s pseudo potential
As we have shown in the previous sections that the s-wave scattering length and phase shift are the most important quantities for scattering problems in the dilute limit, in this section we will show how they are related to the microscopics of the interaction potential.
For a real potential between atoms, the details of interactions are usually very compli- cated. The overall profile between neutral atoms is a r−6 decaying van der Waals potential, coming from the virtual dipole process of polarization in neutral atoms. At very short range, there is a repulsive part that overwhelms the attraction at extremely short distances. In the Leonard-Jones picture of interactions between neutral atoms or molecules, the repulsive
14 part comes from the Pauli repulsions and takes a r−12 form. This r−6 attraction plus r−12
repulsion forms a simple description of the interacting potential between neutral atoms,
however it is still very hard to have an analytical form of the solution to this potential
for finite-energy scattering. Fortunately, in dilute gases, we show in the previous section
that the low energy physical properties are determined by the asymptotic form of the wave
functions at large distances. This asymptotic form is governed by the s-wave scattering
length and the phase shift, hence it is possible to approximate the real potential with a
model of simple shaped potential, neglecting the complicated details at short distances. As
long as the scattering length and phase shift are preserved, all the low-energy physics can
be given correctly by the simplified model.
We start from a very simple model with an isotropic square-well interaction, which
resembles the short range r−6 attractive potential, plus a short-range cut off. The square
well potential has a depth V0 and range r0, i.e.
~2κ2 V (r) = 0, r > r0; V (r) = V0 , r < r0. (2.36) − ≡ − 2mr
2 2 Solving the Schr¨odinger’sequation for total energy E = ~ k /2mr state by connecting the wave functions in two regions at r = r0, we find the scattering length rescaled by r0 varies as a function of κ: a tan κr s = 0 1, (2.37) r0 κr0 − and the effective range as
r∗ 1 (κr )2 = 1 0 . (2.38) r − 2κr (κr tan κr ) − 3(κr tan κr )2 0 0 0 − 0 0 − 0
As we plot these in figure 2.2, we can see there are some divergence in as when κr0 reaches some certain values. These values are referred to as resonances, at which a zero energy
bound state appears. The first resonance for square well potential appears at κr0 = π/2. At these resonances, a jumps from to + , but (a )−1 changes continuously. s −∞ ∞ s For this scattering model, at low energy the inverse scattering length 2.21 contains two parts: a momentum independent constant as a−1 and a k-dependent term. In this − s
15 2 0 r * r 0 0 r s a - -2
-4 0 Π 2 Π 3 Π
Κ r0
∗ Figure 2.2: A sketch of the s-wave scattering length the as and the effective range r rescaled by the width of the potential r0 for a square-well potential, as a function of the dimensionless potential depth κr0. The blue curve is the scattering length, and it diverges at the resonances and changes abruptly close to them. The dashed red curve is the effective range: it remains the same order as r0, except for some very narrow region close to where as vanishes.
expression, the leading term of momentum is ik, and the next quadratic term is r∗k2/2. − For low energy scattering k (r∗)−1, such that r∗k2/2 ik , one can simplify the | | | − | inverse scattering length by f −1 = a−1 ik. Now in the simplified model, we have only one − s − relevant parameter: the scattering length. This model is thus universal, to the extent that
the details of the interaction potential are irrelevant, as long as the scattering amplitude is
correctly given at the energy scale we are interested in. Within this spirit, the original van
der Waals potential can be eventually simplified to a “zero-range potential”, in which the
potential vanishes for any finite distance r > 0. This can be understood as the following:
we adjust the potential depth and range together, such that the wave function outside the
original potential remains unchanged. The limit that the range of potential goes to zero
can be expressed by that the wave function is extended to r 0 in this form: → sin(kr ka ) 1 1 ψ(r) − s , r 0. (2.39) ∼ kr ∼ r − as →
The scattering model is thus equivalent to a boundary condition at r 0, which is called →
16 Bethe-Peierls boundary condition. From this boundary condition, Fermi proposed a pseudo potential in the form 2 2π~ as ∂ Vpp = δ(r) r. (2.40) mr ∂r The wave function 2.39 is a solution to a hamiltonian with the interaction term written as
2.40. Although it looks similar to the mean-field interacting model in the weakly-interacting regime 2.26, this form of interaction is beyond the mean-field level and can be used in both weak and strong interactions, namely the value of as can be any value, even much larger than the interparticle spacing.
In the final part of this section we introduce the approach to regularize the zero-range model in momentum space, in the context of the T -matrix formalism. Recall the relation between T -matrix and interaction operator V as 2.7 and 2.8, and for zero-range potential, the Fourier transform of the Dirac-delta function gives a constant interacting potential in momentum space, defined as the bare coupling constantg ¯. The order-by-order expansion
of the T -matrix will have an ultraviolet divergence, since
1 G (E = 0) dpp2 (2.41) 0 ∼ − p2 → −∞ p X Z is linearly divergent in the absence of a high-energy cutoff. This sickness of the bare delta potential should be fixed by the following procedure. The zero-energy scattering matrix is
1 T (k, k0) =g ¯ g¯ T (p, k0), (2.42) − p p X 2 2 where p = ~ p /2mr. In the low-energy limit, we approximate the T -matrix for any energy 2 and equal incoming and outgoing momenta as TE=0(0, 0) = 2π~ as/mr T (0; 0, 0). The ≡ equation above is rewritten as
1 1 1 m 1 = = r . (2.43) g¯ T (0; 0, 0) − 2π 2a − p p ~ s p p X X It gives the relation between the bare coupling constant g and physical observable T -matrix, and this procedure is actually a regularization of the scattering vertex. For a calculation based on the zero-range model on any physical quantities, theg ¯ appearing in final expres-
17 sions should be linked to the physical parameter as according to the formula in 2.43, in order to get a convergent result. In the later chapters, we will show how this regularization works to fix the ultraviolet divergence in many-body calculations starting from this zero-range model in the context of the BCS-BEC crossover.
2.2 Feshbach resonances
Most of the scattering channels between alkali atoms have a scattering length around the order of 100a where a 0.53A˚ is the Bohr radius. Few can have scattering lengths B B ≈ as large as 2000aB. As the most dense stable dilute gases have typical density as large as 1015cm−3, the characteristic scale of the system is given by k a < 1. And as the | F s| discussion above, it is quite far from the most interesting strongly interacting regime.
The Feshbach resonance approach is implemented to tune the interaction strength, usu- ally by making use of a bound state of pairs in different hyperfine channel. This bound state is referred to as the “closed channel”, in contrary to the scattering states which is called the “open channel”. As there is a coupling between the closed and open channel, a divergent scattering length in the open channel appears when the bound-state energy level in the closed channel matches the scattering threshold of the open channel. For most of the Feshbach resonances in alkali atoms, one chooses a closed channel which has a different magnetic moment from the open channel denoted by ∆µ. The energy difference between the two channels can thus be tuned by an external magnetic field.
We illustrate how Feshbach resonances work from a two-channel model with one open channel and one closed channel. We set the zero point of energy level as the scattering threshold of the open channel, i.e. the interaction energy vanishes at large distance in the open channel. Due to different magnetic moment, the scattering threshold between the two channels have a difference ∆µB, where B is the external magnetic field. We denote the energy difference between the bound state in the closed channel and the open channel threshold as the detuning δ, as shown in figure 2.3. As we will see in the following derivations, a large scattering length appears in the open channel when the closed-channel
18 closed