1: \documentclass[aps,floatfix,showpacs]{revtex4}
2: %\documentclass[12pt]{iopart}
3: %\documentclass[floatfix,twocolumn,aps,showpacs]{revtex4}
4: \parskip=7mm
5: \def\beq{\begin{equation}}
6: \def\eeq{\end{equation}}
7: \def\bea{\begin{eqnarray}}
8: \def\eea{\end{eqnarray}}
9: \usepackage{graphicx}
10:
11: \begin{document}
12: \title{Superfluid Pairing in the Three Component Fermi Gas}
13: \stepcounter{mpfootnote}
14: \author{S. Y. Chang}
15: \address{Institute for Theoretical Physics,
16: University of Innsbruck, Technikerstr. 25, A-6020 Innsbruck, Austria}
17: \address{Institute for Quantum Optics and Quantum Information,
18: ICT-Geb\"aude, Technikerstr. 21a , A-6020 Innsbruck, Austria}
19: \author{V. R. Pandharipande}
20: \address{Department of Physics, University of Illinois at Urbana-Champaign,
21: 1110 W. Green St., Urbana, IL 61801, USA}
22: \date{\today}
23:
24: \begin{abstract}
25: We present an analysis of the SU(3) symmetric model of the strongly interacting three component Fermi gas in the continuum space using quantum Monte Carlo techniques.
26: Three body effects predominate in the regime of interaction strength beyond that of threshold of the three particle bound state.
27: However, we find that there is an interval of the interaction strength where the
28: SU(2)$\otimes$U(1) broken symmetry superfluidity is possible. For a strong enough
29: interaction, the SU(3) symmetry is restored and the superfluidity is suppressed. Within the interval of the
30: broken symmetry, we also find that on average the particle pairs belonging to the species with superfluid pairing remain further
31: separated than those without the superfluid pairing correlation.
32:
33: \end{abstract}
34: \pacs{ 03.75.Ss, 05.30.Fk, 21.65.-f, 31.15.A-, 31.15.xt}
35: %\submitto{\NJP}
36: \maketitle
37: \section{Introduction}\label{sec_intro}
38:
39: A pair of fermions occupying different internal states can interact in s-wave scattering.
40: Fermi atoms can be loaded into different internal spin projection states to produce
41: interacting degenerate Fermi gas with $T \approx 0$. Comprehensive reviews of the recent theoretical and experimental advances in the
42: study of the dilute Fermi gases can be found in the Ref. \cite{giorgini2007,ketterle2008}.
43: The fermionic alkali atoms can occupy hyperfine states that result from the coupling of the nuclear
44: angular moment $I$ with the electronic spin $S$. In the presence of a strong magnetic field,
45: the electronic degree of freedom is polarized and the relevant internal degrees of freedom are determined by the state
46: of the nuclear angular momentum. In the case of $^6$Li, the nuclear angular momentum $I=1$ allows three different
47: projections. By combining magnetic as well as optical trapping fields,
48: constraints on the trappable hyperfine states are largely removed.
49:
50: For a Fermi gas at low density, an expansion of the ground state energy in terms of a small parameter $ak_F$
51: is known \cite{lenz1929,huang1957,galitskii1958}. In general, when we have a gas of fermions occupying $s$ different states with the same partial densities,
52: we call it the degeneracy $s$ Fermi gas. The ground state energy has an expansion in powers of $ak_F$
53: %\begin{widetext}
54: \beq
55: \frac{E_0}{N} = \epsilon_F\left\{ \frac{3}{5} +
56: (s-1) \left[\frac{2 ak_F}{3\pi} + \frac{4}{35\pi^2}(11-2ln2)(ak_F)^2 \right] + {\cal O}[(ak_F)^3] \right\}
57: \label{eqn_lenz}
58: \eeq
59: %\end{widetext}
60: where we identify the zeroth order term as the free Fermi gas energy $E_{FG} = \frac{3}{5} \epsilon_F$.
61: $\epsilon_F = \frac{\hbar^2 k_F^2}{2m}$ is the Fermi energy, $k_F$ the Fermi momentum and $a$
62: the s-wave scattering length. $E_{FG}$ is independent of $s$ and it is used as the unit of energy
63: throughout this article. The terms that depend on the interaction potential range $R$ are
64: to appear at higher orders \cite{baker1999, bulgac2002}.
65: For $s = 2$ Fermi gas, the dependence on $R$ can be eliminated by taking the limit $R/r_0 \rightarrow 0$, where $r_0$
66: is the average inter-particle distance ($\frac{4}{3}\pi r_0^3 \rho =1$ with $\rho =$ density).
67: The {\it intermediate regime} where $R << r_0 \sim \frac{1}{k_F} << |a|$ is of
68: particular interest for $s=2$ Fermi gas. This is also called the {\it unitarity regime}.
69: In this regime, the mean free path of the atoms $\lambda \equiv \frac{1}{\rho \sigma}$ becomes much
70: shorter than $r_0$ as the cross section $\sigma$ diverges. Examples of the systems in this regime are found in the
71: dilute gases of $^{6}$Li and $^{40}$K atoms at Feshbach resonances close to the zero temperature.
72: The magnitude of the s-wave scattering length $|a|$ can be $\sim 1000 \mathring{A}$ at the Feshbach
73: resonance while the interaction range of the van der Waals forces $R \sim 10-100 \mathring{A}$. Another example is that of the neutron gas.
74: The neutron-neutron interactions by strong force have $a \sim -18.8 fm$ while $ R \sim 1 fm$.
75: At $a \rightarrow \pm\infty$, zero energy two particle bound state appears. However, it is known that the
76: ground state of the many body system has a positive energy per particle
77: $E_0/(N E_{FG}) \approx 0.40 \sim 0.44 $ \cite{carlson2003,carlson2008}.
78: Ground state properties for $s = 2$ Fermi gas were studied\cite{carlson2003,astra2004,chang2004}
79: using the Quantum Monte Carlo for the unitarity as well as other regimes of interaction.
80: The generalization of the low density expansion (Eq. \ref{eqn_lenz}) to $s \ge 3$ becomes troublesome as
81: the limit of $R \rightarrow 0$ cannot be taken. This is expected since for the three particle systems
82: the Efimov effect predicts an effective three particle attractive interaction \cite{bulgac2002} of the form
83: $\sim -\frac{s_0^2 \hbar^2}{2mR^2}$ where $s_0$ is a universal constant. Thus, the minimum set of parameters
84: to describe the $s=3$ system consist of $ak_F$ and $Rk_F$.
85: In general, the mean field treatments of the three component Fermi gas \cite{honerkamp2004,torma2006,torma2007}
86: do not account correctly the three particle physics.
87:
88: Recently, a stable
89: three component degenerate Fermi gas \cite{jochim2008} has been created experimentally.
90: Broad and close lying Feshbach resonances make the strong and
91: attractive interactions among the atoms in the different internal states possible \cite{gupta2003}.
92: For simplicity, we label the atoms occupying these three different states by a coloring scheme: {\it Green} (G), {\it Red} (R) and {\it Blue} (B).
93: The channel dependent s-wave scattering lengths ($a_{GR}$, $a_{GB}$, and $a_{RB}$) have been measured experimentally \cite{barter2005}.
94: It is experimentally difficult to achieve simultaneously strong interactions in all of the channels.
95: However, in the present work we assume a simplified SU(3) symmetric model where $a = a_{GR} = a_{GB} = a_{RB}$.
96: Also, the mass is assumed to be the same for all the components. We leave non-SU(3) symmetric cases to future study. Three component Fermi gas problem
97: is also relevant in relation to the {\it color superconductivity} of the quark matter \cite{alford2001}.
98:
99: In this article, {\it ab initio} Monte Carlo
100: results of the three component Fermi gas are presented. The fermions interact pairwise, by a short but finite range
101: attractive interaction potential. The possibility of the SU(2)$\otimes$U(1) broken symmetry ground state \cite{modawi1997, honerkamp2004}
102: is considered. Here, only two components participate in the superfluid pairing while the third component remains in the normal phase.
103: At weak interaction strengths, the superfluid pairing is exponentially suppressed with the quasiparticle gap $\Delta/T_F \sim e^{\pi/(2ak_F)}$.
104: However, we find that there is an interval of the interaction strength where the broken symmetry pairing has noticeable effects
105: in the ground state energy, the quasiparticle gap, the pair distribution functions, etc. When the strength of the interaction is further increased,
106: the SU(3) symmetry is restored by the predominant three body effects. We present our analysis using the dimensional
107: arguments for the three particle system (Sec. \ref{sec_bound}) and the quantum Monte Carlo method for the three component Fermi gas
108: (Sec. \ref{sec_many},\ref{sec_results}).
109:
110:
111: \section{Three Particle Bound State and Scaling Behavior}\label{sec_bound}
112:
113: L.H. Thomas \cite{thomas1935} noted that in the nucleus of tritium ($^3$H) which has two neutrons and one proton,
114: the binding energy has no lower bound if we assumed finite negative s-wave scattering length for the proton-neutron
115: interaction and took the interaction range to zero. Thus, three particle binding energies depend on the range of the potential
116: and the ground state energy diverges in the limit of zero interaction range. In addition, as the pairwise interaction approaches
117: the resonance ($a \rightarrow -\infty$), infinite number of shallow three particle bound states (called trimers or trions) appear one after
118: another. They are known as the Efimov states \cite{efimov70,efimov71,lim77,esry2005}. This property is
119: strikingly different from the two particle case. Consequently, qualitatively different behavior of the three component
120: gas is expected to emerge in comparison with the two component Fermi gas. For our discussion, we consider a generic
121: three particle Hamiltonian where the particles interact pairwise
122: \beq
123: {\cal H}_3 = - \frac{\hbar^2}{2m} \sum\limits_{i \in \{G,R,B\} } \nabla_i^2 + v_0\sum\limits_{i<j \in\{G,R,B\}} V_R(r_{ij})~.
124: \label{eqn_3body_h}
125: \eeq
126: There are two positive parameters in the Hamiltonian: strength $v_0$ and range $R$ of the interaction.
127: We assume a negative dimensionless function $V_R(r) \le 0$ while $v_0 >0$ has dimension $\sim length^{-2}$. $V_R(r)$ is solely parametrized by $R$.
128: The s-wave scattering length $a$ can be used equivalently instead of the strength $v_0$.
129:
130: In the two particle systems the bound state threshold is at the resonance ($a_2^c = \pm \infty$) independent of the finite $R$.
131: Given a value of $R$, it is easy to see that three particle threshold $a_3^c$ can be such that $a_3^c< 0$ but not equal to $-\infty$.
132: This can be done using an ansatz as the trial wave function; for example, a function of the form $\Psi_{3,trial}({\bf R}) = f(r_{GR}) f(r_{GB}) f(r_{RB})$
133: with the variational $f(r)$. In this case, it suffices to provide a trimer state upper bound in energy to prove the existence of a trimer state with
134: $a_3^c$ in the interval $(-\infty,0)$ for a given $R$. The fact that $a_3^c$ is negative and $a_3^c \ne -\infty$ is the starting point for
135: the analysis on the scaling behavior of the length unit. When we consider
136: rescaling of the length by taking $ R \rightarrow \alpha R$ and $a \rightarrow \alpha a$ (we will always assume $0 < \alpha <1$ from now and on),
137: consistent scaling behavior with $ \langle {\cal H}_3 \rangle \rightarrow \langle {\cal H}_3 \rangle / \alpha^2 $
138: is expected (we will justify in the next paragraph). Here, it can be seen easily from the zero energy scattering solution that the s-wave scattering length $a$ scales analogous
139: to another length quantity $R$. By making $\alpha \rightarrow 0^+$ we expect $R \rightarrow 0^+$, $a_3^c \rightarrow 0^-$ and $\langle {\cal H}_3 \rangle \rightarrow -\infty$
140: for any $a$ that belongs to the interval $(-\infty, a_3^c \rightarrow 0^-)$. This means that, for a zero range interaction,
141: the trimer state is possible for any attractive pairwise interaction of nonzero strength.
142:
143: Now, we consider formally the scaling behavior of the energy for the system of three particles in vacuum.
144: We assume a value of $R$ such that trimer state is allowed with a strength $v_0^c$ corresponding to $a_3^c < 0$.
145: The three particle Schr\"odinger equation with the usual notation ${\bf X} = \{ {\bf x}_G,
146: {\bf x}_R, {\bf x}_B \}$ and $r_{ij} = |{\bf x}_i - {\bf x}_j|$ is
147: \beq
148: {\cal H}_3 \Psi({\bf X}) = E \Psi({\bf X}) ~.
149: \label{eqn_3body_sch}
150: \eeq
151: Let ${\bf X}_s = \alpha {\bf X}$, $R_s = \alpha R$, $v_{0,s} = v_0/\alpha^2$, $r_{ij,s} = \alpha r_{ij}$ and $\Psi_s({\bf X}_s) = \Psi({\bf X})$.
152: Then $\nabla_i^2 \Psi_s({\bf X}_s) =\frac{1}{\alpha^2} \nabla_i^2 \Psi({\bf X})$ and
153: $V_{R_s}(r_{ij,s}) \Psi_s({\bf X}_s) = V_R( r_{ij}) \Psi({\bf X})$.
154: Thus, after scaling all the length quantities by $\alpha$, the Eq. \ref{eqn_3body_sch} becomes
155: \bea
156: & &\left[- \frac{\hbar^2}{2m} \sum_{i \in \{G,R,B\}} \nabla_i^2 +
157: v_{0,s} \sum_{i<j \in \{G,R,B\}} V_{R_s}(r_{ij,s})
158: \right] \Psi_s({\bf X}_s) \nonumber \\
159: & = &\left[- \frac{1}{\alpha^2}\frac{\hbar^2}{2m}\sum_{i \in \{G,R,B\}} \nabla_i^2 + \frac{v_0}{\alpha^2} \sum_{i<j\in \{G,R,B\}} V_R(r_{ij}) \right] \Psi({\bf X}) \nonumber \\
160: & = & \frac{1}{\alpha^2}E \Psi({\bf X})= E_s \Psi_s({\bf X}_s) ~.
161: \eea
162: As a result, after length scaling $\Psi_s({\bf X}_s)$ is the solution with the eigenvalue $E_s = \frac{E}{\alpha^2}$.
163: We arrived at this property by using the dimensional arguments alone.
164: As good illustrative examples of this scaling behavior, the following cases are mentioned:
165: When $a \in (-\infty, a_3^c)$ for a given $R$, there is trimer state with energy $E<0$.
166: Then, we can scale length by an overall factor $\alpha$ but keep $a_s$ constant (that is, increase $|a_s|$ to match $|a|$).
167: It is obvious that taking $\alpha \rightarrow 0^+$ causes $E_s(a)$ to collapse rapidly to $-\infty$.
168: This means that given a trimer state in vacuum, when $R \rightarrow 0$ at a fixed non-zero value of $a$ ($<0$), $E$ has to go to $-\infty$.
169: This is in agreement with the above mentioned Ref. \cite{thomas1935}. As another example, if we had initially
170: $a \in (a^c_3,0)$, the interaction potential is not strong enough to allow trimer state and the total energy of the particles $E = 0$ in the vacuum.
171: In this case, length scaling leaves the particles unbound with $E_s(a_s) = 0$. Thus, unbound particles remain unbound even after length scaling
172: and the collapse does not occur.
173:
174: For the comparison purpose, let's also consider the scaling behavior of a pair of particles. We can see that
175: the scaling of length does not produce the collapse as in the three particle case. For the s-wave scattering length $a$
176: such that $1/a < 0$, there is no bound state and the energy of the pair remains zero ($E = 0$) in the vacuum at any length scale.
177: Then, we consider $1/a > 0$ regime (usually called BEC regime). Using the same scaling analysis, we also have $E_s(a_s) = \frac{E(a)}{\alpha^2}$.
178: In this case, we can solve exactly the contact interaction ($R=0$) problem by replacing the potential by the boundary condition $\frac{u'(0)}{u(0)} = -\frac{1}{a}$ where $u(r)$ is the radial
179: wave function of the pair. The solution for the radial wave function is $u(r) \sim e^{-\frac{r}{a}}$ with
180: $E_{pair}(a) = -\frac{\hbar^2}{m a^2}$. This energy is finite unless $a \rightarrow 0^+$. According to this solution,
181: the scaling behavior of energy $\frac{E_{pair}(\alpha a) }{E_{pair}(a)} = \frac{1}{\alpha^2}$
182: is the same as the result we obtained from the dimensional analysis, $\frac{E_s(\alpha a)}{E(a)} = \frac{1}{\alpha^2}$.
183: Thus, for a pair the energy scaling relation becomes identical to the bound state energy in the strong coupling limit $1/a_s \rightarrow +\infty$.
184: This is qualitatively different from the three body collapse at the weak coupling limit $1/a_s \rightarrow -\infty$ because
185: of the simultaneous scaling of $R$ to zero.
186:
187: \section{Many Particle Ground State}\label{sec_many}
188:
189: For the study of many body systems at finite density, we use {\it ab~initio} stochastic method known as Fixed Node
190: Green's Function Monte Carlo (FN GFMC). In general, we take a trial wave function $\Psi_V$ that obeys antisymmetry upon
191: the exchange of identical fermions and let it evolve in the imaginary time restricted to a definite
192: sign domain given by the nodal surface of the trial wave function itself. If the nodal structure is correct,
193: we get the exact ground state. Otherwise, we get an approximate ground state and an energy upper-bound.
194: The implementation of this method is explained in detail elsewhere \cite{carlson2003, chang2004}.
195:
196: We consider a system with $6 \sim 8$ particles of each color (Green,Red,Blue), so that $18 \le N_{total} \le 24$.
197: The particles are contained in a finite box with the periodic boundary conditions at the walls to simulate
198: the uniform matter. For many particle systems, we cannot rescale the length without changing the density.
199: The scaling behavior analyzed in the previous section is only applicable to the few particle systems in vacuum.
200: In the degeneracy two case, one dimensionless product $a k_F$ uniquely determines the system. The parameter $R$
201: can be pushed in principle to the zero limit and eliminated from the description of the system. In practice,
202: small but finite values of $R/r_0 << 1$ were assumed \cite{carlson2003, chang2004} as long as
203: the results were converged within the statistical errors to the $R k_F \rightarrow 0$ limit. However, from the scaling
204: behavior analysis of the previous section, it becomes clear that for the degeneracy three Fermi gas we need to keep
205: $Rk_F$ finite in order to avoid local trimer instability. Thus, we need both $ak_F$ and $Rk_F$ to fully describe the three component
206: Fermi gas. We keep $R k_F = 0.32 $
207: in the $s = 3$ case which is the same value as in the $s =2$ case ($R/r_0 \sim 0.1$, thus converged to the $R/r_0 \rightarrow 0$ limit).
208: However, we should keep in mind that for $s=3$, this particular value of $R$ is not the limit of $R \rightarrow 0$.
209: We consider the cases with small deviations from the balanced partial densities.
210: We explore the possibility of the superfluid ground state within a certain interval of the interaction strength.
211: We also analyze the qualitative behavior of the pair distribution functions (see Sec.\ref{sec_results}). The many body SU(3) symmetric Hamiltonian is
212: \beq
213: {\cal H} = -\frac{\hbar^2}{2m} \sum\limits_{i} \nabla_i^2 +
214: v_0 \sum\limits_{i<j} V_R(r_{ij})(1-\delta_{c_i,c_j})
215: \eeq
216: where $c_i$ is the color of i-th particle. Only pairs of different color particles interact.
217: $V_R(r)$ is the dimensionless core of the P\"oschl-Teller potential: $V_R(r) = -\frac{1}{ \cosh^2(2 r/R)}$
218: and $v_0 = \frac{8\hbar^2}{mR^2}$. $v_0$ can be adjusted to get the desired scattering lengths.
219: We can impose different nodal restrictions to the solution by using different trial wave functions. We can estimate the energy by using
220: SU(3) symmetric Slater trial wave function
221: \beq
222: \Psi_{FG}= \Psi_{FG,G}\Psi_{FG,R}\Psi_{FG,B}
223: \label{eqn_fg}
224: \eeq
225: where the factors represent the normal
226: states (given by Slater determinants of the plane wave orbitals) of different
227: color species. It was also shown\cite{honerkamp2004} that the pairing fields $(\Delta_{GR}, \Delta_{GB}, \Delta_{RB})$
228: with $\Delta_{\alpha\beta} \sim \sum_{\bf k} \langle c_{{\bf k},\alpha} c_{-{\bf k},\beta}\rangle$ can be mapped
229: into $(\Delta_0, 0,0)$ with the constraint $\sum_{\alpha,\beta} \Delta_{\alpha\beta}^2 = \Delta_0^2$.
230: This is analogous to the analysis of Ref. \cite{modawi1997} by Modawi and Leggett where the ground state should allow
231: one normal phase component. Thus, we consider the SU(2)$\otimes$ U(1) broken symmetry pairing ground state with BCS pairing for two of the Fermi components while
232: the third component remains in the normal phase. The corresponding nodal structure
233: is given by the trial wave function
234: \bea
235: \Psi_{bs-BCS} & = & \Psi_{FG,B} \Psi_{BCS,GR} \nonumber \\
236: &=& \left[ \prod_{|{\bf k}|<k_F}a^\dagger_{{\bf k},B} \right] \left[ \prod_{\bf k} (u_{\bf k} + v_{\bf k} a^{\dagger}_{{\bf
237: k},G} a^{\dagger}_{-{\bf k},R}) \right]|0\rangle \nonumber \\
238: & \rightarrow & \Psi_{FG,B} {{\cal A}}[\phi(r_{11'}) \phi(r_{22'}) ... \phi(r_{MM'})]_{GR}~ .
239: \label{eqn_bs_bcs}
240: \eea
241: Here, we assume that blue species remains in the normal phase (represented by $\Psi_{FG,B}$), while between the green and
242: red species there is pairing correlation (represented by $\Psi_{BCS,GR}$). In the last line of Eq. \ref{eqn_bs_bcs},
243: we assumed fixed number projection of the green and red particles.
244:
245: The complete trial wave function with Jastrow-like factor can be written as
246: \beq
247: \Psi_{V, FG} = \prod_{i<j}\left[ f_S(r_{ij})\delta_{c_i,c_j} + f_D(r_{ij})(1-\delta_{c_i,c_j})\right] \Psi_{FG}
248: \label{eqn_res3_slater}
249: \eeq
250: for the SU(3) symmetric Slater wave function (Eq. \ref{eqn_fg}). Here, we considered separately the correlation that exists
251: between the same color particles $f_S(r)$ ($ = f_{GG}(r) = f_{RR}(r) = f_{BB}(r)$) from the correlation that exists
252: between different color particles $f_D(r)$ ($ = f_{GR}(r) = f_{GB}(r) = f_{RB}(r) $).
253: Usually $f_S(r)$ is qualitatively different from $f_D(r)$. $f_S(r)$ is analogous
254: to $f_{\uparrow \uparrow}(r)$ of the degeneracy two case and includes Pauli exclusion principle ($f_S(0) = 0$).
255: The particular shapes of the correlation functions do not change the GFMC energies. However,
256: they are optimized in order to minimize statistical errors and to have the converged pair
257: distribution functions $g(r)\approx g_{trial}(r) \approx g_{GFMC}(r)$ (see Ref. \cite{chang2005}).
258: We can see that the optimized $f_D(r_{ij})$ deviates largely (more peaked at $r \approx 0$) from the one obtained
259: by using the LOCV equations \cite{carlson2003}. This is due to the strong three body
260: effects even when the pairwise interactions are relatively weak.
261:
262: Analogously, the complete trial wave function with the SU(2)$\otimes$U(1) broken symmetry pairing correlation (Eq. \ref{eqn_bs_bcs}) is
263: \bea
264: \Psi_{V,bs-BCS} & = & \prod_{i<j} [f_S(r_{ij})\delta_{c_i,c_j} + f_{GR}(r_{ij})\delta_{c_i,G}\delta_{c_j,R} +\ldots \nonumber \\
265: & & f_{GB}(r_{ij})(\delta_{c_i,G} + \delta_{c_i,R})\delta_{c_j,B} ] \Psi_{bs-BCS}~.
266: \label{eqn_res3_bcs}
267: \eea
268: This wave function can give better pair distribution functions as the optimization of the correlation functions can be carried out
269: separately for $f_{GR}(r)$ and $f_{GB}(r) = f_{RB}(r)$. Then, we can demonstrate that $f_{GR}(0) < f_{GB}(0)$ as well
270: as $g_{GR}(0) < g_{GB}(0)$ (see the discussion of the Fig. \ref{fig_two} in the Sec. \ref{sec_results}).
271: The optimum nodal structure is tried as $\alpha_I= \{0.2, 0.1, 0.02, 0.01, 0.01\}$ with non zero short range
272: function $\tilde{\beta}(r)$ parametrized by $b=0.44$ (see Ref. \cite{carlson2003,chang2004} for the definitions of these
273: parameters). These parameters are identical
274: to those of the $s=2$ case when $1/ak_F \ge 0$ (called molecular or BEC regime). This trial wave function assumes arbitrarily
275: one of the three possible broken symmetry pairing states
276: \bea
277: |1\rangle & \equiv & \Psi_{FG,B} \Psi_{BCS,GR} \nonumber \\
278: |2\rangle & \equiv & \Psi_{FG,R} \Psi_{BCS,GB} \nonumber \\
279: |3\rangle & \equiv & \Psi_{FG,G} \Psi_{BCS,RB}.
280: \label{eqn_degens}
281: \eea
282: These states are degenerate in energy and the broken symmetry can be assumed by choosing one of these states without
283: loss of generality. The GFMC energies using the trial nodes given by the wave functions of Eq. \ref{eqn_res3_slater} and Eq. \ref{eqn_res3_bcs}
284: are summarized in the Fig. \ref{fig_one}, while the pair correlation functions are presented in the Fig. \ref{fig_two}.
285:
286: \section{Discussion of the Results}\label{sec_results}
287:
288: \begin{figure}
289: \begin{center}
290: \includegraphics[width=8.0cm]{fig1.eps}
291: \caption{(color online) Comparison of the degeneracy two ($s = 2$) and three ($s = 3$) results for $R k_F = 0.32$ in both cases.
292: $E_{Lenz}$ (Eq. \ref{eqn_lenz}) gives a good estimate of the normal state energy for $s = 2$ and $1/ak_F \le -1$.
293: However for $s = 3$, the match is poor with the GFMC results as the
294: $R$ dependent terms and three particle effects predominate.
295: In the region $1/ak_F \lesssim -0.3$, the SU(2)$\otimes$U(1) broken symmetry pairing state is shown to be the ground state for $s = 3$ Fermi
296: gas. However, for $-0.3 \lesssim 1/ak_F$ the SU(3) symmetry is restored(see also Fig. \ref{fig_four}) and the superfluid pairing suppressed.
297: This behavior for $s=3$ is quantitatively different from the $s=2$ case, where with the increasing interaction
298: strength (increasing $1/ak_F$), the superfluid pairing state (crosses) is always the favored ground state compared to the
299: state without superfluid correlation (circles). The size of the symbols approximately correspond to the error bars. }
300: \label{fig_one}
301: \end{center}
302: \end{figure}
303:
304: \begin{figure}
305: \begin{center}
306: \includegraphics[width=8.0cm]{fig2.eps}
307: \caption{(color online) Pair distribution functions at $ak_F = -0.75$(a), $-1.0$(b), $-1.\dot{3}$(c) and $-2.0$(d).
308: We define pair distributions $g_{\alpha \beta}(r) \sim \langle \sum\limits_{i_\alpha,j_{\beta}} \delta(r - r_{i_\alpha,j_{\beta}} ) \rangle$.
309: The angled brackets imply taking thermal average of the possible configurations. $g_{\alpha \beta}(r)$ has boundary condition $\lim\limits_{r \rightarrow \infty} g_{\alpha \beta}(r) =1$.
310: The line corresponds to $g_{GR}(r)$, that is, the pair distribution between the green and red species while the dashed line corresponds
311: to the distribution between the green and blue species $g_{GB}(r)$. Both $g_{GR}(r)$ and $g_{GB}(r)$ assume
312: SU(2)$\otimes$U(1) symmetric state (Eq. \ref{eqn_res3_bcs}).
313: The dotted line represents the pair distribution without
314: superfluidity $g_{Slater}(r)$ calculated with the SU(3) symmetric Slater wave function (Eq. \ref{eqn_res3_slater}).
315: The $g_{Slater}(r)$ is independent of the colors as long as they are different. All $g(r)$'s are calculated with the optimized $f(r)$ functions.
316: In the plots c) and d) where the superfluid pairing is strong, $g_{GR}(0) < g_{GB}(0) = g_{RB}(0) < g_{Slater}(0)$.
317: Thus, on the average the particles that have superfluid pairing correlation remain further apart than the particles that
318: have no superfluid pairing correlations. }
319: \label{fig_two}
320: \end{center}
321: \end{figure}
322:
323: \begin{figure}
324: \begin{center}
325: \includegraphics[width=6.0cm]{fig3.eps}
326: \caption{(color online) Energy excitations with unbalanced particle numbers (allows broken pairs) at $ak_F = -1.\dot{3}$. The circles correspond to the set of particle numbers
327: $\{N_G, N_R, N_B \}=$ $\{6,6,6\}$,$\{7,6,6\}$,$\{7,7,6\}$, $\{7,7,7\}$,$\{8,7,7\}$,$\{8,8,7\}$, and $\{8,8,8\}$.
328: The squares are $\{6,6,7\}$, $\{6,6,8\}$, $\{7,7,8\}$, and $\{7,7,9\}$ cases. The triangles represent the configurations $\{7,6,7\}$ and $\{8,7,8\}$.
329: The size of the symbols approximately correspond to the error bars.}
330: \label{fig_three}
331: \end{center}
332: \end{figure}
333:
334: \begin{figure}
335: \begin{center}
336: \includegraphics[width=8.0cm]{fig4.eps}
337: \caption{(color online) Imaginary time evolution of the overlap of the wave functions defined as $\langle 2|e^{-{\cal H}\tau} |1 \rangle =
338: \langle3 | e^{-{\cal H}\tau} |1 \rangle \sim \langle \Psi_{FG} | e^{-{\cal H}\tau} |1 \rangle$ are shown. Each time step is $\delta\tau = 3.6\times10^{-4} \frac{\hbar}{E_{FG}}$.
339: The states $|1\rangle$, $|2\rangle$ and $|3\rangle$ are as defined in Eq. \ref{eqn_degens} while $|\Psi_{FG}\rangle$ is as defined in Eq.
340: \ref{eqn_res3_slater}. We can see that as the interaction strength is increased, the symmetry is restored with the overlaps
341: approaching 1.}
342: \label{fig_four}
343: \end{center}
344: \end{figure}
345:
346:
347: The results of three component Fermi gas energies are summarized in the Fig. \ref{fig_one}. The pairing correlations produce
348: noticeable effects at $-1 \le 1/ak_F \le -0.5$ (compare the triangles with the circles of the same figure). Close to the $a^c_3$,
349: the pairing effects are small. Here, the energy of the SU(2)$\otimes$U(1) broken symmetry state
350: is not distinguishable within the error bars from that of the SU(3) symmetric state (Slater) .
351: The system is found stable in the regime of interaction considered in this work
352: ($-1.3 \lesssim 1/ak_F \lesssim -0.3$ and $R k_F = 0.32$).
353:
354: We also found that it is possible to see the effects of pairing in the $g(r)$'s.
355: In the plot a) of Fig. \ref{fig_two} ($a k_F = -0.75$), no difference can be seen in the $g(r)$'s because the pairing is weak.
356: However, where the pairing is relevant (in terms of the energy) we notice that $g_{GR}(0) < g_{GB}(0) = g_{RB}(0) < g_{Slater}(0)$
357: (where $g_{Slater}(r)$ is the pair distribution obtained with the non-pairing Slater wave function) and the symmetry is broken
358: (see the plots c) and d) of the Fig \ref{fig_two}). Green and red particles have superfluid pairing correlations,
359: so one may have naively expected $g_{GR}(0) > g_{GB}(0) = g_{RB}(0)$, but the opposite is found to be true.
360: According to this, the relative distances satisfy on average: $r_{GB} = r_{RB} < r_{GR}$.
361: The interpretation we can give is that once a Cooper pair is formed, the third particle feels much stronger attraction
362: toward the center of the mass of the Cooper pair. In fact, we can approximately estimate that the potential strength between the center of mass (CM)
363: of green-red pair and the blue particle is enhanced from $v_0$ to $\frac{4}{3} v_0$, while green and red particles interact with the strength $v_0$.
364: This comes from the observation that the zero energy scattering equation between the CM of green-red and blue is
365: $-\frac{\hbar^2}{2m_r}u''(r) + v_0 V_R(r) u(r) = 0$ with $m_r = \frac{2}{3}m$. From the plot d) of the Fig. \ref{fig_two}, it is still not conclusive whether the symmetry could be restored
366: ($g_{GR}(0) = g_{GB}(0) = g_{RB}(0)$) in the regime of strong three particle correlation $-0.5 < 1/ak_F$. In that regime,
367: the analysis is hindered on the practical ground: it becomes increasingly harder to obtain reliable estimates of $g(r)$
368: keeping the statistical errors small. However, the Fig. \ref{fig_one} and Fig. \ref{fig_four} show clearly the tendency toward the
369: restoration of the SU(3) symmetry entering into a regime completely dominated by three particle physics with negligible
370: paring correlations.
371:
372: Using the GFMC technique, energy gaps can be calculated by allowing variations in the numbers of the green and red particles
373: while keeping the number of blue particles constant. In this way, we allow breaking of the superfluid pairs.
374: We can consider sets of $\{N_G, N_R, N_B \}=$ $\{6,6,6\}$,$\{7,6,6\}$,$\{7,7,6\}$,
375: and $\{7,7,7\}$,$\{8,7,7\}$,$\{8,8,7\}$. We observe the usual odd-even staggering of the ground state energy (Fig. \ref{fig_three}).
376: Furthermore, we can allow the variations $6 \le N_G \le 8$, $6 \le N_R \le 8$, and $6 \le N_B \le 8$ and calculate
377: the excitation energies. The lowest energy excitations at $ak_F = -1$ and $-1.\dot{3}$ are found with the momentum ${\bf k } = {\bf 0}$
378: quasiparticle. In the case of $s = 2$, quasiparticles with zero momentum produce the minimum excitation
379: in the $1/ak_F > 0$ (BEC) regime. Thus, this is consistent with the interpretation that the studied regime ($-1.3 \lesssim 1/ak_F \lesssim -0.3$)
380: is that of strong three particle correlation. The energies for the broken pair states are shown in the Fig. \ref{fig_three} for the $ak_F = -1.\dot{3}$ case.
381: The gap $\Delta$ is estimated from the data sets represented by circles with total $N = 18,19,20$ and $N = 21,22,23$ respectively.
382: The calculated $\Delta/E_{FG} \approx 0.3(3)$. The error bars are large for the pairing gap since the three particle
383: effects predominates rather than two particle pairing. In the Fig. \ref{fig_three}, we notice that the energy dips
384: when total $N$ is a multiple of $3$. We interpret this as an effect analogous to that observed in the $g(r)$'s. This is the evidence
385: that the trimer interaction that brings the green-red pair and the blue particle together is much stronger than simple pairwise interaction.
386: Thus, completing green-red-blue trimer is energetically more favorable than unbalanced excess of one or two species.
387: In fact, the trimer binding energy is so strong that $\Delta < |E_{trimer}/3|$
388: in contrast to $\Delta \approx |E_{pair}/2|$ of $s=2$ Fermi gas in the BEC regime.
389: Consequently, in the quasiparticle spectrum, it is expected that we can observe two distinctive {\it gaps}; one due to the
390: superfluid pairing and another due to the trimer binding. We also notice that at $N = 21$ the dip is as deep as at $N=24$ which
391: indicates possible shell closure effect. As seen in the ground state energy (Fig. \ref{fig_one}) and the overlap of the
392: wave functions (Fig. \ref{fig_four}), the broken symmetry superfluidity is suppressed in the $ak_F > -0.5$ regime where
393: the SU(3) symmetry is restored.
394:
395: For the comparison purpose only, we {\it naively} consider extension of the mean field (BCS-Leggett) method to the three component Fermi gas.
396: For this mean field model, there is no $R$ dependence and the variational ground state is always stable. The two color pairing gap of the degeneracy three superfluid
397: is given by the same relation as that of the degeneracy two Fermi gas. While in the original BCS formalism, the chemical potential
398: $\mu$ is kept constant and $\Delta_{BCS}/\epsilon_F = \frac{8}{e^2}e^{\pi/2ak_F}$ we consider the BCS-Leggett \cite{leggett1980}
399: variational formalism ($\Delta_{BCS-Leggett}$). Here, the chemical potential is changed in order to keep the density constant.
400: The condensation energy of the system is $E_{cond} = - n_0 \frac{\Delta^2}{2}$ where $n_0$ is the state density ($n_0 \equiv \frac{mk_F \Omega}{2 \pi^2 \hbar^2}$).
401: Dependence on the degeneracy $s$ is included in the condensation energy per particle $\frac{E_{cond}}{N} = -\frac{9}{20 s} \frac{\Delta^2}{E_{FG}}$
402: because of the relation $\frac{6\pi^2}{s}\rho = k_F^3$. Thus, at $1/ak_F = -1$, we estimate that $\Delta_{BCS-Leggett}/E_{FG} = 0.33$ and $E_{cond}/(N E_{FG}) \approx -0.016$.
403: At $1/ak_F = -0.75$ (or $ak_F = 1.\dot{3}$) ,we have $\Delta_{BCS-Leggett}/E_{FG} = 0.5$ and $E_{cond}/(N E_{FG}) \approx -0.038$.
404: This estimate is close to the one calculated by GFMC (Fig. \ref{fig_three}, $\Delta/E_{FG} \sim 0.3$ ) at the same interaction strength.
405:
406: Although for $s=3$ Fermi gas, $R$ dependence cannot be removed, clear qualitative differences between the $s = 2$ and $s=3$ Fermi gases emerge.
407: Unlike in the $s=2$ Fermi gas where both the superfluid pair and the bound state are qualitatively similar, in the $s=3$ Fermi gas the paired states
408: decouple from the bound state(trimer) in energy. This can be clearly observed in the quasiparticle excitation spectrum.
409: Realistic interaction potential and channel dependence of the interactions are necessary in order to
410: produce not only qualitative but also quantitatively correct results for a given three component Fermi gas. We found a regime of interaction strength
411: where the broken symmetry pairing is clearly detectable, beyond which three particle effects dominate and the symmetry is restored.
412: Also, non trivial dependence of the $g(r)$'s on the pairing correlations was discussed.
413: This work has been supported in part by the US National Science Foundation via grant PHY 00-98353 and PHY 03-55014. SYC also
414: acknowledges support by the DARPA grant BAA 06-19. The authors acknowledge useful comments from J. Carlson and A. Bulgac. One of the authors(VRP) passed away during the preparation
415: of the present manuscript and the work was posthumously completed.
416:
417:
418:
419:
420: \begin{thebibliography}{16}
421: \expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi
422: \expandafter\ifx\csname bibnamefont\endcsname\relax
423: \def\bibnamefont#1{#1}\fi
424: \expandafter\ifx\csname bibfnamefont\endcsname\relax
425: \def\bibfnamefont#1{#1}\fi
426: \expandafter\ifx\csname citenamefont\endcsname\relax
427: \def\citenamefont#1{#1}\fi
428: \expandafter\ifx\csname url\endcsname\relax
429: \def\url#1{\texttt{#1}}\fi
430: \expandafter\ifx\csname urlprefix\endcsname\relax\def\urlprefix{URL }\fi
431: \providecommand{\bibinfo}[2]{#2}
432: \providecommand{\eprint}[2][]{\url{#2}}
433:
434:
435: \bibitem{giorgini2007} See for example summaries by S.~Giorgini, L.~P.~Pitaevskii and S.~Stringari, Rev. Mod.
436: Phys. {\bf 80}, 1215 (2007), and by R.~Grimm cond-mat/0703091 (2007).
437:
438: \bibitem{ketterle2008} W.~Ketterle, and M.~W.~Zwierlein, arXiv:0801.2500v1 (2008).
439:
440: \bibitem{lenz1929} W.~Lenz, Z. Physik {\bf 56}, 778 (1929).
441:
442: \bibitem{huang1957} K.~Huang, and C.~N.~Yang, Phys. Rev. {\bf 105}, 767 (1957).
443:
444: \bibitem{galitskii1958} V.~M.~Galitskii, Sov. Phys. JETP {\bf 7}, 104 (1958).
445:
446:
447: \bibitem{baker1999} G.~A. Baker, Phys. Rev. C {\bf 60}, 054311 (1999).
448:
449: \bibitem{bulgac2002} A.~Bulgac, Phys. Rev. Lett. {\bf 89}, 050402 (2002).
450:
451: \bibitem{carlson2003} J.~Carlson, S.~Y.~Chang, V.~R.~Pandharipande, and K.~E.~Schmidt,
452: Phys. Rev. Lett. {\bf 91}, 50401 (2003).
453:
454: \bibitem{carlson2008} J. Carlson, and S.~Reddy, Phys. Rev. Lett. Phys. Rev. Lett. {\bf 100}, 150403 (2008).
455:
456: \bibitem{astra2004} G.~E.~Astrakharchik, J.~Boronat, J.~Casulleras, and S.~Giorgini, Phys. Rev. Lett. {\bf 93}, 200404 (2004).
457:
458: \bibitem{chang2004} S.~Y.~Chang, V.~R.~Pandharipande, J.~Carlson, and K.~E.~Schmidt, Phys. Rev. A. {\bf 70}, 043602 (2004).
459:
460: \bibitem{honerkamp2004} C.~Honerkamp, and W.~Hofstetter, Phys. Rev. B {\bf 70}, 094521 (2004).
461:
462: \bibitem{torma2006} T.~Paananen, J.~P. Martikainen, and P.~T\"orm\"a, Phys. Rev. A {\bf 73}, 053606 (2006).
463:
464: \bibitem{torma2007} T.~Paananen, P.~T\"orm\"a, and J.~P. Martikainen, Phys. Rev. A {\bf 75}, 023622 (2007).
465:
466:
467: %\bibitem{bedaque2006} P.~F.~Bedaque, and J.~P.~D'Incao, cond-mat/0602525 (2006).
468:
469: \bibitem{jochim2008} T.~B.~Ottensten,~T.~Lompe, ~M.~Kohnen,~A.~N.~Wenz,~and S.~Jochim, Phys. Rev. Lett.
470: {\bf 101}, 203202 (2008).
471:
472: \bibitem{gupta2003} S.~Gupta, Z.~Hadzibabic, M.~W.~Zwierlein, C.~A.~Stan, K.~Dieckmann,
473: C.~H.~Schunck, E.~G.~M.~van Kempen, B.~J.~Verhaar, and W.~Ketterle,
474: Science {\bf 300}, 1723 (2003).
475:
476: \bibitem{barter2005} M.~Bartenstein, A.~Altmeyer, S.~Riedl, R.~Geursen, S.~Jochim, C.~Chin, J.~H.~Denschlag,
477: R.~Grimm, A.~Simoni, E.~Tiesinga, C.~J.~Williams, and P.~S.~Julienne, Phys. Rev. Lett. {\bf 94}, 103201 (2005).
478:
479: \bibitem{alford2001} M.~Alford, Ann. Rev. Nucl. Part. Sci. {\bf 51}, 131 (2001).
480:
481:
482: \bibitem{modawi1997} A.~G.~K.~Modawi, and A.~J.~Leggett, Journal of Low Temp. Phys. {\bf 109}, 625 (1997).
483:
484: \bibitem{thomas1935} L.~H.~Thomas, Phys. Review {\bf 47}, 903 (1935).
485:
486: \bibitem{efimov70} V.~Efimov, Phys. Lett. {\bf 33B}, 563 (1970).
487:
488: \bibitem{efimov71} V.~Efimov, Sov. J. Nucl. Phys. {\bf 12}, 589 (1971).
489:
490: \bibitem{lim77} T.~K.~Lim, K.~Duffy, and W.~Damert, Phys. Rev. Lett. {\bf 38}, 341 (1977).
491:
492: \bibitem{esry2005} J.~P.~D'Incao, and B.~D.~Esry, Phys. Rev. A {\bf 72}, 032710 (2005).
493:
494: \bibitem{chang2005} S.~Y.~Chang, and V.~R.~Pandharipande, Phys. Rev. Lett. {\bf 95}, 080402 (2005).
495:
496: \bibitem{leggett1980} A.~J.~ Leggett, {\it Modern Trends in the Theory of Condensed Matter},
497: edited by A.~Pekalski, and R.~ Przystawa (Springer-Verlag, Berlin, 1980).
498:
499: \end{thebibliography}
500:
501: \end{document}
502:
503:
504:
505:
506: