1: \documentclass[showpacs,twocolumn,amssymb]{revtex4}
2: \usepackage{graphicx}
3:
4: \begin{document}
5:
6: \title{Phase Diagram of Cold Polarized Fermi Gas in Two Dimensions}
7: \author{Lianyi He and Pengfei Zhuang}
8:
9: \affiliation{Physics Department, Tsinghua University, Beijing
10: 100084, China}
11:
12: \begin{abstract}
13: The superfluid phase diagrams of a two-dimensional cold polarized
14: Fermi gas in the BCS-BEC crossover are systematically and
15: analytically investigated. In the BCS-Leggett mean field theory,
16: the transition from unpolarized superfluid phase to normal phase
17: is always of first order. For a homogeneous system, the two
18: critical Zeeman fields and the critical population imbalance are
19: analytically determined in the whole coupling parameter region,
20: and the superfluid-normal mixed phase is shown to be the ground
21: state between the two critical fields. The density profile in the
22: presence of a harmonic trap calculated in the local density
23: approximation exhibits a shell structure, a superfluid core at the
24: center and a normal shell outside. For weak interaction, the
25: normal shell contains a partially polarized cloud with constant
26: density difference surrounded by a fully polarized state. For
27: strong interaction, the normal shell is totally in fully polarized
28: state with a density profile depending only on the global
29: population imbalance. The di-fermion bound states can survive in
30: the whole highly imbalanced normal phase.
31: \end{abstract}
32:
33: \pacs{03.75.Ss, 05.30.Fk, 74.20.Fg, 34.90.+q} \maketitle
34:
35:
36: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
37: \section {Introduction}
38: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
39: The effect of Zeeman energy splitting $h$ induced by a strong
40: magnetic field between spin up and down electrons on
41: Bardeen-Cooper-Shriffer(BCS) superconductivity, which has been
42: investigated many years ago~\cite{CC,Sarma,FFLO}, promoted new
43: interest in recent years due to the progress in the experiments of
44: ultracold Fermi gases\cite{exp1,exp2,exp3,exp4,exp5,exp6}. The
45: well-known result for weak-coupling s-wave superconductivity is
46: that, at a critical Zeeman field or the so-called
47: Chandrasekhar-Clogston(CC) limit $h_c=0.707\Delta_0$ where
48: $\Delta_0$ is the zero temperature gap, the Cooper pairs are
49: destroyed and a first order quantum phase transition from the
50: gapped BCS state to the normal state occurs\cite{CC}. Further
51: studies showed that the inhomogeneous
52: Fulde-Ferrell-Larkin-Ovchinnikov(FFLO) state\cite{FFLO} where
53: Cooper pairs have nonzero momentum can survive above the CC limit
54: up to $h_{\text{FFLO}}=0.754\Delta_0$. However, since the
55: thermodynamic critical field is much smaller than the CC limit due
56: to strong orbit effect\cite{CC}, it is hard to observe the CC
57: limit and the FFLO state in ordinary superconductors.
58:
59: Recent experiments on ultracold Fermi gas trapped in an external
60: harmonic potential, serve as an alternative way to study the pure
61: Zeeman effect on Fermi
62: superfluidity~\cite{exp1,exp2,exp3,exp4,exp5,exp6}. The atom
63: numbers of the two lowest hyperfine states of $^6$Li atom,
64: $N_\uparrow$ and $N_\downarrow$, are adjusted to create a
65: population imbalance or polarization
66: $P=(N_{\uparrow}-N_{\downarrow})/(N_{\uparrow}+N_{\downarrow})$,
67: which simulates the Zeeman field $h$ in a superconductor. The
68: s-wave attraction between the two hyperfine states is tuned around
69: the Feshbach resonance to realize a strongly rather than weakly
70: interacting Fermi gas. In three-dimensional case, the density
71: profiles observed in experiments exhibit an unpolarized superfluid
72: core in the center of the trap and a polarized normal gas shell
73: outside\cite{exp1,exp2,exp3,exp4}, which justifies that the ground
74: state around unitary is a phase separation state(although there
75: may be some dispute on the shell structure), predicted by early
76: theoretical works\cite{BCR,Cohen,Carlson,Sheehy,MFwork}.
77:
78: One of the theoretical interests in the study of polarized Fermi
79: superfluidity is to determine its phase structure in the whole
80: interaction strength region, namely in the BCS-BEC (Bose-Einstein
81: Condensation) crossover~\cite{Sheehy,MFwork}. The complete mean
82: field phase diagram in coupling-imbalance plane as well as the
83: critical Zeeman field $h_c$ and critical polarization $P_c$ are
84: theoretically predicted in three dimensional case
85: (3D)~\cite{Sheehy}. Since the s-wave mean field equations can not
86: be solved analytically in the whole coupling region, it is hard to
87: determine a precise phase diagram in 3D even in mean field
88: approximation. Recently, the quantitatively correct phase diagram
89: in 3D has been obtained in quantum Monte Carlo
90: calculations\cite{MC1,MC2}. However, the theoretical prediction of
91: the phase diagram for a homogeneous system can not be directly
92: examined in ultracold Fermi gas experiments, due to the effect of
93: the external harmonic trap. To have a comparison with the
94: experimental data, one should investigate the phase diagram and
95: the density profile in the presence of an external trap potential,
96: using the same equation of state. In the case of 3D, the density
97: profile can only be treated numerically\cite{trap}.
98:
99: While we have well understood the 3D phase diagram, the phase
100: structure of polarized Fermi gases in low dimensions promoted
101: recently experimental and theoretical interests. In one
102: dimensional case the phase diagram is determined via exact
103: solvable models~\cite{1D}. In the two-dimensional case, while
104: exact solvable models are lacked, the s-wave mean field equations
105: can be solved analytically in the whole coupling parameter
106: region~\cite{2Da,2Db,2Dc,2Dreview}, and the Fermi surface topology
107: and stability condition are not trivial and different from those
108: in 3D~\cite{guba}. In this paper, we will determine the phase
109: diagrams of a polarized Fermi gas in two dimensions, for both
110: homogeneous and trapped systems, and calculate the density profile
111: of a trapped imbalance Fermi gas. Our results are totally
112: analytical in the whole coupling parameter region, including the
113: phase diagrams and the density profile. In the final part of this
114: paper, we also discuss the existence of di-fermion bound states in
115: the polarized normal phase.
116:
117: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
118: \section {BCS-BEC Crossover in Two Dimensions}
119: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
120: The BCS-BEC crossover problem in two dimensions has been widely
121: discussed in the literatures\cite{2Da,2Db,2Dc,2Dreview}. In this
122: paper, we employ an effective 2D Hamiltonian where the
123: renormalized atom-atom interaction can be characterized by an
124: effective binding energy\cite{2Da}. For a wide Feshbach resonance,
125: the effective grand canonical Hamiltonian can be written as
126: \begin{eqnarray}
127: H&=&\sum_{\sigma=\uparrow,\downarrow}\int d^2{\bf
128: r}\psi_{\sigma}^\dagger({\bf
129: r})\left(-\frac{\hbar^2}{2M}\nabla^2-\mu-\sigma_zh\right)\psi_{\sigma}^{\phantom{\dag}}({\bf r})\nonumber\\
130: &-&U\int d^2{\bf r}\psi_{\uparrow}^\dagger({\bf
131: r})\psi_{\downarrow}^\dagger({\bf
132: r})\psi_{\downarrow}^{\phantom{\dag}}({\bf
133: r})\psi_{\uparrow}^{\phantom{\dag}}({\bf r}),
134: \label{Hamiltonian.2D}
135: \end{eqnarray}
136: where $M$ is the fermion mass, $\mu$ is the chemical potential,
137: $U>0$ is the contact attractive interaction, and $\sigma_z=\pm1$
138: correspond to $\sigma=\uparrow,\downarrow$. We choose the unit
139: $\hbar=1$ through the paper. The Zeeman field $h$ can be created
140: by either an external field\cite{CC,Sarma,FFLO} or a population
141: imbalance\cite{exp1,exp2}. In the former case, the total particle
142: number $N$ is conserved, but the particles $N_\uparrow$ and
143: $N_\downarrow$ in the states $\uparrow$ and $\downarrow$ can
144: transfer to each other\cite{liu}, i.e., the chemical potentials
145: for the two components are always the same, but the external field
146: $h$ induces an effective chemical potential difference. In the
147: latter case, $N_\uparrow$ and $N_\downarrow$ are both conserved,
148: and the two chemical potentials can be expressed as
149: $\mu_\uparrow=\mu+h$ and $\mu_\downarrow=\mu-h$.
150:
151: At finite temperature in 2D, the long range order is absent and no
152: phase transition can happen. At zero temperature, however, there
153: do exist long range order~\cite{2Dreview} and one can safely
154: consider phase transitions among different states. In this paper,
155: we will study the phase diagrams at zero temperature in the
156: BCS-Leggett mean field approximation which is accepted to
157: adequately describe the BCS-BEC crossover at $T=0$~\cite{BCSBEC}.
158:
159: In the balanced case with $h=0$, the thermodynamic potential
160: density of a uniform Fermi gas can be evaluated as\cite{2Dreview}
161: \begin{eqnarray}
162: \Omega(\mu;\Delta)=\frac{\Delta^2}{U}+\int\frac{d^2{\bf
163: k}}{(2\pi)^2}\left(\xi_{\bf k}-E_{\bf k}\right)
164: \end{eqnarray}
165: with the definition of particle energies $E_{\bf k}=\sqrt{\xi_{\bf
166: k}^2+\Delta^2}$ and $\xi_{\bf k}={\bf k}^2/(2M)-\mu$ and the
167: superfluid order parameter
168: $\Delta=-U\langle\psi_{\downarrow}^{\phantom{\dag}}\psi_{\uparrow}^{\phantom{\dag}}\rangle$.
169: In the dilute limit, the UV divergence in the expression of
170: $\Omega$ can be eliminated via introducing the two body scattering
171: length.
172:
173: While the two-body bound state in 3D forms only at sufficiently
174: strong attraction where the s-wave scattering length diverges and
175: changes sign, the bound state in 2D can form at any arbitrarily
176: small attraction\cite{QM}. For an inter-atomic potential described
177: by a 2D circularly symmetric well of radius $R_0$ and depth $V_0$,
178: the bound-state energy $\epsilon_{\text B}$ is given by
179: $\epsilon_{\text B} \simeq 1/(2MR_0^2) \exp{[-2/(MV_0R_0^2)]}$
180: with $V_0 R_0^2 \to 0$. As a consequence, the solution of the
181: BCS-BEC problem in 2D is much simpler than that in the case of 3D
182: in terms of special functions\cite{2Db}. It is shown that the
183: existence of the two-body bound state in vacuum is a necessary
184: (and sufficient) condition for the Cooper instability\cite{2Da}.
185: To regulate the UV divergence in $\Omega$, we introduce a high
186: energy cutoff $\Lambda={\bf k}_\Lambda^2/(2M)$ in the integral.
187: The momentum cutoff ${\text k}_\Lambda$ corresponds to the inverse
188: of the range $r_0$ of the interaction potential. Due to the energy
189: independence of the density of states in 2D, after performing the
190: integration over ${\bf k}$ one obtains
191: \begin{eqnarray}
192: \Omega&=&\frac{\Delta^2}{U}-\frac{M\Delta^2}{4 \pi} \bigg[ \ln
193: \frac{\Lambda-\mu + \sqrt{(\Lambda-\mu)^{2}+\Delta^{2}}}
194: {\sqrt{\mu^{2}+\Delta^{2}}-\mu}\nonumber\\
195: &+&\frac{\Lambda-\mu}{\Lambda- \mu + \sqrt{(\Lambda- \mu)^{2} +
196: \Delta^{2}}} + \frac{\mu}{\sqrt{\mu^{2} + \Delta^{2}} - \mu}
197: \bigg].
198: \end{eqnarray}
199:
200: In this paper we consider a dilute Fermi gas with effective
201: interaction range $r_0\rightarrow 0$. Taking large enough cutoff
202: $\Lambda$ and small enough attraction $U$, we can introduce a 2D
203: two-body binding energy\cite{2Dreview} to replace the cutoff in
204: this limit,
205: \begin{equation}
206: \epsilon_{\text B} = 2\Lambda \exp \left(-\frac{4 \pi}{M U}
207: \right),
208: \end{equation}
209: which does not include any many-particle effect. With the binding
210: energy, the cutoff dependence can be eliminated in the dilute
211: limit with $\Lambda \to \infty$ and $U \to 0$ but finite
212: $\epsilon_{\text B}$. We obtain in this limit
213: \begin{equation}
214: \Omega=\frac{M\Delta^2}{4\pi} \left( \ln \frac{\sqrt{\mu^{2} +
215: \Delta^{2}} - \mu}{\epsilon_{\text B}} - \frac{\mu}{\sqrt{\mu^{2}
216: + \Delta^{2}} - \mu} - \frac{1}{2} \right) .
217: \end{equation}
218: The above procedure is equivalent to directly substituting the
219: coupling constant $U$ by the 2D bound state equation
220: \begin{equation}
221: \frac{1}{U}=\int\frac{d^2{\bf k}}{(2\pi)^2}\frac{1}{{\bf
222: k}^2/M+\epsilon_{\text{B}}}.
223: \end{equation}
224:
225: The BCS-BEC crossover phenomenon in 2D can be observed by solving
226: the coupled gap and number equations, namely
227: $\partial\Omega/\partial\Delta=0$ and
228: $n=-\partial\Omega/\partial\mu$. Defining the Fermi energy
229: $\epsilon_{\text F}=\pi n/M$ in 2D, the gap and number equation
230: can be analytically expressed as
231: \begin{equation}
232: \sqrt{\mu^{2} + \Delta^{2}} - \mu = \epsilon_{\text B}, \qquad
233: \sqrt{\mu^{2} + \Delta^{2}} + \mu = 2\epsilon_{\text F},
234: \end{equation}
235: respectively. Their solution takes a very simple form\cite{2Da}
236: \begin{equation}
237: \Delta_0 = \sqrt{2 \epsilon_{\text B} \epsilon_{\text F}} \,,
238: \qquad \mu_0 = \epsilon_{\text F}- \frac{\epsilon_{\text B}}{2} ,
239: \end{equation}
240: or rewrite it in terms of a dimensionless quantity
241: $\eta=\epsilon_{\text B}/\epsilon_{\text F}$,
242: \begin{equation}
243: \frac{\Delta_0}{\epsilon_{\text F}}=\sqrt{2\eta} \,, \qquad
244: \frac{\mu_0}{\epsilon_{\text F}}=1-\frac{\eta}{2} .
245: \end{equation}
246: One sees very clear that the chemical potential decreases with
247: increasing coupling or decreasing density, which indicates a
248: BCS-BEC crossover. The Chemical potential changes sign at
249: $\epsilon_{\text B}=2\epsilon_{\text F}$. To understand the
250: physical significance of these simple results, we consider two
251: limits. For very weak attraction (or high density), the
252: two-particle binding energy is extremely small, i.e.
253: $\epsilon_{\text B} \ll \epsilon_{\text F}$, and we recover the
254: well-known BCS result with strongly overlapping Cooper pairs. In
255: this limit we have the chemical potential
256: $\mu_0\simeq\epsilon_{\text F}$ and the gap function
257: $\Delta_0\ll\epsilon_{\text F}$. For the opposite limit of very
258: strong attraction (or low particle density), we have a deep
259: two-body bound state with $\epsilon_{\text B} \gg \epsilon_{\text
260: F}$, and the system is in the BEC region with composite bosons. In
261: this limit the chemical potential takes $\mu_0 \simeq
262: -\epsilon_{\text B}/2$. It should be kept in mind that in the
263: local pair regime ($\mu_0<0$) the fermion excitation gap
264: $E_{\text{gap}}$ in the quasi-particle excitation spectrum is not
265: $\Delta_0$ (as in the case $\mu_0> 0$) but rather
266: $\sqrt{\mu_0^{2}+\Delta_0^{2}}$.
267:
268: In ultracold Fermi gas experiments, a quasi-2D Fermi gas can be
269: realized by arranging a one-dimensional optical lattice along the
270: axial ($z$) direction and a weak harmonic trapping potential in
271: the radial ($x$-$y$) plane, such that fermions are strongly
272: confined along the $z$ direction and form a series of
273: pancake-shaped clouds~\cite{2Dexp,zhang}. Each such cloud can be
274: considered as a quasi-2D Fermi gas when the axial confinement is
275: strong enough to turn off inter-cloud tunnelling. The strong
276: anisotropy of the trapping potentials, namely $\omega_{z}\gg
277: \omega$ where $\omega_z$($\omega$) is the axial(radial) frequency,
278: allows us to use an effective 2D Hamiltonian to deal with the
279: radial degrees of freedom\cite{zhang}. The effective binding
280: energy $\epsilon_{\text B}=\hbar\omega_z\exp{[4\pi
281: a_z^2/U_{p}^{\text{eff}}(a_s,a_z)]}$ is related to the energy
282: scale $\hbar\omega_z$ and the 3D s-wave scattering length $a_s$,
283: where $a_z$ is defined as $a_z=\sqrt{\hbar/(m\omega_z)}$ and the
284: quantity $U_{p}^{\text{eff}}(a_s,a_z)$ defined in \cite{zhang}
285: carries the dependence on 3D scattering length. By adjusting the
286: 3D scattering length $a_s$ and/or the axial frequency $\omega_z$,
287: a quasi-2D BCS-BEC crossover can be realized.
288:
289: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
290: \section {Equation of State}
291: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
292: We now turn on the Zeeman splitting $h\neq 0$. To determine the
293: superfluid phase diagrams and calculate the density profile for
294: imbalanced Fermi gas in a harmonic trap, we first establish the
295: equations of state~(EOS) for various phases in grand canonical
296: ensemble\cite{Sheehy}. In the BCS-Leggett mean field theory, the
297: pressure ${\cal P}=-\Omega$ as a function of $\mu$ and $h$ can be
298: evaluated as\cite{Sheehy}
299: \begin{eqnarray}
300: {\cal P}(\mu,h)&=&c\int_0^\infty dz\bigg[E_z-z+\mu-\frac{\Delta^2}{2z+\epsilon_{\text{B}}}\nonumber\\
301: &-&(E_z-h)\Theta(h-E_z)\bigg]
302: \end{eqnarray}
303: with $ E_z=\sqrt{(z-\mu)^2+\Delta^2}$ and $c=M/(2\pi)$. We have
304: set $h>0$ without loss of generality. The superfluid order
305: parameter
306: $\Delta(\mu,h)=-U\langle\psi_{\downarrow}^{\phantom{\dag}}\psi_{\uparrow}^{\phantom{\dag}}\rangle$
307: is determined self-consistently from the gap equation
308: \begin{equation}
309: \Delta\int_0^\infty
310: dz\left[\frac{1}{2z+\epsilon_{\text{B}}}-\frac{\Theta(E_z-h)}{2E_z}\right]=0.
311: \end{equation}
312: The step function $\Theta(x)$ in this paper is defined as
313: $\Theta(x)=0$ for $x<0$ and $\Theta(x)=1$ for $x>0$.
314:
315: Unlike the 3D case\cite{Sheehy}, the EOS in 2D can be analytically
316: obtained. At fixed $\mu$ and $h$, we have three possible phases:
317: the unpolarized superfluid phase (SF), the polarized normal phase
318: (N) and the polarized superfluid phase or Sarma phase
319: (S)~\cite{Sarma}. The phase SF corresponds to the solution
320: $\Delta(\mu)=\sqrt{\epsilon_{\text B}(\epsilon_{\text B}+2\mu)}$
321: in the region $h<E_{\text
322: g}=\sqrt{\mu^2\Theta(-\mu)+\Delta^2(\mu)}$, and the pressure can
323: be evaluated as
324: \begin{eqnarray}
325: {\cal P}_{\text{SF}}(\mu)=c\left(\mu+\frac{\epsilon_{\text
326: B}}{2}\right)^2
327: \end{eqnarray}
328: which does not depend explicitly on $h$. The total number density
329: $n=n_\uparrow+n_\downarrow$ and the magnetization
330: $m=n_\uparrow-n_\downarrow$ can be expressed as
331: \begin{eqnarray}
332: n_{\text{SF}}(\mu) &=& 2c\left(\mu+\frac{\epsilon_{\text
333: B}}{2}\right),\ \ \ m_{\text{SF}}(\mu)=0. \label{EOSSF}
334: \end{eqnarray}
335: The polarized superfluid phase or Sarma phase~(S) corresponds to
336: the solution in the region $h>E_{\text g}$. This phase can be
337: ruled out from the positive secondary derivative~\cite{guba}
338: \begin{equation}
339: \frac{\partial^2{\cal P}}{\partial \Delta^2}\Big|_{\text
340: S}=c\left[\frac{h(1+\Theta(\mu))}{\sqrt{h^2-\Delta^2}}-\frac{\mu\Theta(\mu)}{\sqrt{\mu^2+\Delta^2}}-1\right]>0,\label{Sarma}
341: \end{equation}
342: which means that the Sarma phase is always unstable for any
343: coupling in 2D.
344:
345: The polarized normal phase corresponds to the solution $\Delta=0$.
346: The pressure takes the form of non-interacting Fermi gas,
347: \begin{eqnarray}
348: {\cal P}_{\text
349: N}(\mu,h)=\frac{c}{2}\left[(\mu-h)^2\Theta(\mu-h)+(\mu+h)^2\Theta(\mu+h)\right],\nonumber\\
350: \end{eqnarray}
351: where the case $\mu+h<0$ corresponds to the vacuum without atoms.
352: For $\mu+h>0$, the total number density and the magnetization read
353: \begin{eqnarray}
354: n_{\text N}(\mu,h)&=&2c\mu\Theta\left(\mu-h\right)
355: +c(\mu+h)\Theta\left(h-\mu\right),\nonumber\\
356: m_{\text N}(\mu,h)&=&2ch\Theta\left(\mu-h\right)
357: +c(\mu+h)\Theta\left(h-\mu\right).\label{EOSN}
358: \end{eqnarray}
359: The cases $\mu>h$ and $\mu<h$ correspond to the partially
360: polarized (N$_{\text{PP}}$) and fully polarized (N$_{\text{FP}}$)
361: normal phases respectively. Since we treat the superfluid and
362: normal phase in mean field approximation, the normal phase is
363: considered as a non-interacting gas. In fully polarized case, this
364: is correct since only s-wave interaction is considered. However,
365: in partially polarized case, the interaction may be important in
366: some coupling parameter region, like the finding around the
367: unitary region in 3D~\cite{unitary,MC1,MC2}. Including
368: fluctuations, which can not be treated analytically even in 2D, is
369: necessary for a more realistic study.
370:
371: Since the polarized superfluid phase is always located at the
372: maximum of the thermodynamic potential, there exists at fixed
373: $\mu$ a first order quantum phase transition from the SF phase to
374: the normal phase when the Zeeman field $h$ increases. The critical
375: value $h_c$ is determined by the condition ${\cal
376: P}_{\text{SF}}(\mu)={\cal P}_{\text{N}}(\mu, h_c)$. The analytical
377: expression for $h_c$ can be written as
378: \begin{eqnarray}
379: h_c(\mu)&=&\sqrt{\epsilon_{\text B}\left(\mu+\frac{\epsilon_{\text
380: B}}{4}\right)}\Theta(\mu-h_0)\nonumber\\
381: &+&\left[(\sqrt{2}-1)\mu+\frac{\epsilon_{\text
382: B}}{\sqrt{2}}\right]\Theta(h_0-\mu).
383: \end{eqnarray}
384: Equivalently, for a given $h$, SF-N phase transition happens when
385: the chemical potential $\mu$ becomes less than the critical value
386: \begin{eqnarray}
387: \mu_c(h)&=&\left(\frac{h^2}{\epsilon_{\text
388: B}}-\frac{\epsilon_{\text
389: B}}{4}\right)\Theta\left(h-h_0\right)\nonumber\\
390: &+&\frac{\sqrt{2}h-\epsilon_{\text
391: B}}{2-\sqrt{2}}\Theta\left(h_0-h\right), \label{UC}
392: \end{eqnarray}
393: where $h_0=(\sqrt{2}+1)\epsilon_{\text B}/2$ is determined by the
394: equation $h_0=\mu_c(h_0)$. We can easily prove that
395: $h>h_0$($h<h_0$) is equivalent to the condition
396: $\mu_c>h$($\mu_c<h$).
397:
398: The grand canonical phase diagram in the $\mu-h$ plane is shown in
399: Fig.\ref{fig1}. The analytical expressions for the phase
400: boundaries can be obtained from the above expression for $h_c$.
401: The SF phase, N$_{\text{FP}}$ phase and the vacuum meet at the
402: point $(\mu,h)=(-\epsilon_{\text B}/2,\epsilon_{\text B}/2)$,
403: while the three phases SF, N$_{\text{PP}}$ and N$_{\text{FP}}$
404: meet at $(\mu,h)=((\sqrt{2}+1)\epsilon_{\text
405: B}/2,(\sqrt{2}+1)\epsilon_{\text B}/2)$. The grand canonical phase
406: diagram is of great help for us to understand the density profile
407: in a harmonic trap.
408: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
409: \begin{figure}[!htb]
410: \begin{center}
411: \includegraphics[width=8cm]{Gphase.eps}
412: \caption{The grand canonical phase diagram in the $\mu-h$ plane.
413: $\mu$ and $h$ are scaled by the binding energy $\epsilon_{\text
414: B}$. \label{fig1}}
415: \end{center}
416: \end{figure}
417: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
418:
419: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
420: \section {Homogeneous Fermi Gas}
421: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
422: In this section we determine the phase diagram of the homogeneous
423: system. Since the total atom number $N=N_\uparrow+N_\downarrow$ or
424: equivalently the total atom density $n$ is fixed, the chemical
425: potential $\mu$ is not a free variable in the canonical ensemble
426: and should be determined by the number conservation. One may
427: distinguish two different cases: (1)The Zeeman field $h$ can be
428: experimentally adjusted by using Raman detuning\cite{liu}; (2)The
429: atom number for each species, $N_\uparrow$ and $N_\downarrow$, can
430: be adjusted\cite{exp1,exp2}. Since the phase structure should be
431: essentially independent of the ensemble we choose, we firstly
432: discuss the phase diagram using $h$ as tunable parameter, and then
433: translate it into the case where the global polarization
434: $P=(N_{\uparrow}-N_{\downarrow})/(N_{\uparrow}+N_{\downarrow})$ is
435: directly adjusted.
436:
437: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
438: \subsection {Critical Zeeman Fields}
439: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
440: We now consider the problem: When does the superfluidity disappear
441: when a Zeeman splitting $h$ is turned on? The total density $n$
442: and the Zeeman field $h$ are thermodynamic variables, and the free
443: energy of the system should be defined as ${\cal F}(n,h)=\mu
444: n-{\cal P}$. Since $n=M\epsilon_{\text F}/\pi$ is fixed, we will
445: write ${\cal F}(n,h)={\cal F}(h)$. At nonzero Zeeman field $h$,
446: the solutions of the coupled gap and number equations
447: corresponding to the above three homogeneous bulk phases can be
448: analytically solved:
449:
450: \noindent {\bf I}. $\Delta_{\text{SF}}(h)=\Delta_0$ and
451: $\mu_{\text{SF}}(h)=\mu_0$ in the phase SF. The solution exists in
452: the region $0<h<\Delta_0$ for $\eta<2$ or $0<h<\epsilon_{\text
453: F}+\epsilon_{\text B}/2$ for $\eta>2$.
454:
455: \noindent {\bf II}. $\Delta_{\text N}(h)=0$ and $\mu_{\text
456: N}(h)=\epsilon_{\text F}\Theta(\epsilon_{\text
457: F}-h)+(2\epsilon_{\text F}-h)\Theta(h-\epsilon_{\text F})$ in the
458: phase N. The first and second term correspond, respectively, to
459: partially and fully polarized normal phase.
460:
461: \noindent {\bf III}.
462: $\Delta_{\text{S}}(h)=\sqrt{\Delta_0(2h-\Delta_0)}$ and
463: $\mu_{\text S}(h)=\epsilon_{\text F}-
464: \Delta_{\text{S}}^2(h)/(4\epsilon_{\text F})$ in the region of
465: $\Delta_0/2<h<\Delta_0$ and $\eta < 2$ and
466: $\Delta_{\text{S}}(h)=\sqrt{\epsilon_{\text B}(2h-\epsilon_{\text
467: B})}$ and $\mu_{\text S}(h)=2\epsilon_{\text F}- h$ in the region
468: of $\epsilon_{\text B}/2<h<\epsilon_{\text F}+\epsilon_{\text
469: B}/2$ and $\eta > 2$ in the phase S. There are two gapless Fermi
470: surfaces at $\eta < 2$ and only one gapless Fermi surface at $\eta
471: > 2$.
472:
473: The polarized superfluid phase or the Sarma phase, which is a
474: gapless superfluid, is again an unstable state at any coupling,
475: directly from the reentrance phenomenon (three solutions of
476: $\Delta$ at fixed $h$), in contrast to the case in 3D where it
477: becomes the stable ground state in the strong coupling BEC
478: region~\cite{Sheehy,MFwork}. This is an important difference of
479: the Fermi surface topology and the stability condition between 3D
480: and 2D cases\cite{guba}. Explicitly, the free energy (density)
481: ${\cal F}(h)=\mu(h) n-{\cal P}(\mu(h),h)$ in the three homogeneous
482: bulk phases reads
483: \begin{eqnarray}
484: {\cal F}_{\text{SF}}(h)&=&c(\epsilon_{\text F}^2-\epsilon_{\text
485: F}\epsilon_{\text B}),\nonumber\\
486: {\cal F}_{\text{N}}(h)&=&c\big[(\epsilon_{\text
487: F}^2-h^2)\Theta(\epsilon_{\text
488: F}-h)\nonumber\\
489: &+&2(\epsilon_{\text F}^2-\epsilon_{\text
490: F}h)\Theta(h-\epsilon_{\text
491: F})\big],\nonumber\\
492: {\cal F}_{\text{S}}(h)&=&c[2(\epsilon_{\text F}^2-\epsilon_{\text
493: F}h)+h^2+(h-\epsilon_{\text B}/2)^2]\Theta(\eta-2)\nonumber\\
494: &+&c[(\epsilon_{\text F}^2-\epsilon_{\text F}\epsilon_{\text
495: B})+(\Delta_0-h)^2]\Theta(2-\eta).
496: \end{eqnarray}
497: It is easy to see that the polarized superfluid phase has always
498: higher free energy. If there exist no other possible phases, a
499: first order quantum phase transition from the phase SF to the
500: phase N will occur at a critical Zeeman field $h_{c}$ determined
501: by ${\cal F}_{\text{SF}}(h_c)={\cal F}_{\text{N}}(h_c)$. We find
502: $h_c=\sqrt{\eta}\epsilon_{\text F}=\Delta_0/\sqrt{2}$ for $\eta<1$
503: and $h_c=\frac{1}{2}(1+\eta)\epsilon_{\text F}$ for $\eta>1$. It
504: is interesting to note that the relation $h_c=\Delta_0/\sqrt{2}$
505: at $\eta<1$ is only an approximate result at weak coupling in
506: 3D~\cite{CC,Sarma}.
507:
508: If there are only the two bulk phases SF and N, we have only one
509: CC limit at which the first order phase transition occurs, and the
510: experimentally observed phase separation (PS) will be hidden in
511: the $\eta$-$h$ phase diagram. However, since the total atom
512: density $n$ is fixed, unlike the grand canonical ensemble, we
513: should consider possible mixed phases constructed via the Gibbs
514: phase equilibrium condition. Here we will neglect the interfacial
515: energy\cite{surface}, since for a macroscopic phase separation,
516: this energy contribution is subdominant in the thermodynamic
517: limit. From equation (\ref{Sarma}), the only possibility is the
518: SF-N mixed phase. When the phase separation is favored in a region
519: $h_{c1}<h<h_{c2}$, the chemical potential $\mu_{\text{PS}}$ is
520: different from $\mu_{\text{SF}}$ and $\mu_{\text{N}}$, it should
521: be determined by the phase equilibrium condition ${\cal
522: P}_{\text{SF}}(\mu)={\cal P}_{\text N}(\mu,h)$, which leads to
523: \begin{eqnarray}
524: \mu_{\text{PS}}(h)&=&\left(\frac{h^2}{\epsilon_{\text
525: B}}-\frac{\epsilon_{\text
526: B}}{4}\right)\Theta\left(h-h_0\right)\nonumber\\
527: &+&\frac{\sqrt{2}h-\epsilon_{\text
528: B}}{2-\sqrt{2}}\Theta\left(h_0-h\right), \label{UPS}
529: \end{eqnarray}
530: where $h_0=(\sqrt{2}+1)\epsilon_{\text B}/2$ satisfies the
531: equation $h_0=\mu_{\text{PS}}(h_0)$. Since the chemical potential
532: $\mu_{\text PS}$ is determined by the condition ${\cal
533: P}_{\text{SF}}(\mu)={\cal P}_{\text N}(\mu,h)$, it is equivalent
534: to the critical chemical potential $\mu_c(h)$ in the grand
535: canonical ensemble. The cases $h>h_0$ and $h<h_0$ indicate,
536: respectively, the mixed phases with partially polarized normal
537: bubbles (SF-N$_{\text{PP}}$) and fully polarized normal bubbles
538: (SF-N$_{\text{FP}}$). The volume fractions of the phases SF and N
539: in the phase separation, denoted by $x$ and $1-x$ respectively,
540: are determined by the number conservation,
541: $n=x(h)n_{\text{SF}}(\mu_{\text{PS}},h)+[1-x(h)]n_{\text
542: N}(\mu_{\text{PS}},h)$. Using the expressions (\ref{EOSSF}),
543: (\ref{EOSN}) and (\ref{UPS}) for $\mu_{\text{PS}}, n_{\text N}$
544: and $n_{\text{SF}}$, we find
545: \begin{eqnarray}
546: x(h)&=&2\left(\frac{\epsilon_{\text{F}}}{\epsilon_{\text{B}}}+\frac{1}{4}-\frac{h^2}{\epsilon_{\text{B}}^2}\right)\Theta\left(h-h_0\right)\nonumber\\
547: &+&\left(\frac{2\sqrt{2}\epsilon_{\text F}}{2h-\epsilon_{\text
548: B}}-\sqrt{2}-1\right)\Theta\left(h_0-h\right).
549: \end{eqnarray}
550:
551: We now determine the region of the mixed phase, i.e., the lower
552: and upper critical fields $h_{c1}$ and $h_{c2}$~\cite{Sheehy}. In
553: the grand canonical ensemble with fixed chemical potential $\mu$,
554: we have only one critical field $h_c(\mu)$ determined by the
555: condition ${\cal P}_{\text{SF}}(\mu,h)={\cal P}_{\text N}(\mu,h)$,
556: and the signal of SF-N phase separation is denoted by the first
557: order phase transition line in the $\mu-h$ phase diagram. In the
558: standard BCS-BEC crossover problem, the total atom number $N$
559: rather than the chemical potential $\mu$ is fixed, and the CC
560: limit splits into two values $h_{c1}=h_c(\mu_{\text{SF}})$ and
561: $h_{c2}=h_c(\mu_{\text N})$~\cite{Sheehy}. The mixed phase links
562: continuously the phases SF and N with
563: $\mu_{\text{PS}}=\mu_{\text{SF}}$ at $h=h_{c1}$ and
564: $\mu_{\text{PS}}=\mu_{\text{N}}$ at $h=h_{c2}$ and ensures $0\leq
565: x\leq 1$ with $x(h_{c1})=1$ and $x(h_{c2})=0$. The critical fields
566: $h_{c1}$ and $h_{c2}$ are explicitly given by
567: \begin{eqnarray}
568: h_{c1}&=&\epsilon_{\text{F}}\sqrt{\eta\left(1-\frac{\eta}{4}\right)}\Theta(\eta_1-\eta),\nonumber\\
569: &+&\epsilon_{\text{F}}\left(\sqrt{2}-1+\frac{\eta}{2}\right)\Theta(\eta-\eta_1),\nonumber\\
570: h_{c2}&=&\epsilon_{\text{F}}\sqrt{\eta\left(1+\frac{\eta}{4}\right)}\Theta(\eta_2-\eta),\nonumber\\
571: &+&\epsilon_{\text{F}}\left(2-\sqrt{2}+\frac{\eta}{2}\right)\Theta(\eta-\eta_2),
572: \end{eqnarray}
573: where $\eta_1=2-\sqrt{2}\simeq0.586$ and
574: $\eta_2=2(\sqrt{2}-1)\simeq0.828$ are determined by
575: $\mu_{\text{SF}}(h_0)=h_0$ and $\mu_{\text N}(h_0)=h_0$,
576: respectively. There is always the relation $h_{c1}<h_{c}<h_{c2}$,
577: and the splitting disappears in the weak coupling limit
578: $\eta\rightarrow 0$ which recovers the well known result shown in
579: \cite{CC,Sarma}. On the other hand, the splitting keeps as a
580: constant $(3-2\sqrt{2})\epsilon_{\text F}\simeq
581: 0.172\epsilon_{\text F}$ at strong coupling $\eta>\eta_2$.
582:
583: The final step is to prove that the SF-N mixed phase has the
584: lowest free energy in the region $h_{c1}<h<h_{c2}$. Using the
585: analytical expressions for $x(h)$ and $\mu_{\text{PS}}(h)$ as well
586: as the EOS for the phases SF and N, we can evaluate the free
587: energy in the mixed phase defined by ${\cal
588: F}_{\text{PS}}(h)=\mu_{\text{PS}}n-x(h){\cal
589: P}_{\text{SF}}(\mu_{\text{PS}},h)-[1-x(h)]{\cal
590: P}_{\text{N}}(\mu_{\text{PS}},h)$. The difference between ${\cal
591: F}_{\text{PS}}$ and ${\cal F}_{\text{SF}}$ and between ${\cal
592: F}_{\text{PS}}$ and ${\cal F}_{\text{N}}$ can be explicitly
593: expressed as
594: \begin{eqnarray}
595: &&{\cal F}_{\text{PS}}(h)-{\cal F}_{\text{SF}}(h)=-\epsilon_{\text
596: B}^{-2}c(h^2-h_{c1}^2)^2\Theta(h-h_0)\nonumber\\
597: &&\ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ -(\sqrt{2}+1)^2c(h-h_{c1})^2\Theta(h_0-h),\nonumber\\
598: &&{\cal F}_{\text{PS}}(h)-{\cal F}_{\text{N}}(h)=-\epsilon_{\text
599: B}^{-2}c(h^2-h_{c2}^2)^2\Theta(h-h_0)\nonumber\\
600: &&\ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \
601: -(\sqrt{2}+1)^2c(h-h_{c2})^2\Theta(h_0-h).
602: \end{eqnarray}
603: The above expressions show explicitly that the mixed phase has
604: really the lowest free energy in the region $h_{c1}<h<h_{c2}$.
605: While in 3D the conclusion that the mixed phase corresponds to the
606: lowest free energy is analytically proven in the weak coupling
607: limit~\cite{BCR}, our result here in 2D is for any coupling.
608:
609: The SF-N mixed phase has a nonzero global polarization $P$ since
610: the normal bubble is polarized. From the definition
611: $P=(N_{\uparrow}-N_{\downarrow})/(N_{\uparrow}+N_{\downarrow})$ we
612: find
613: \begin{eqnarray}
614: P(h)=[1-x(h)]\frac{m_{\text N}(\mu_{\text{PS}},h)}{n}.
615: \end{eqnarray}
616: Using the expression for $x(h)$ and $\mu_{\text{PS}}$, we have
617: \begin{eqnarray}
618: P(h)&=&\frac{2h(h^2-h_{c1}^2)}{\epsilon_{\text F}\epsilon_{\text
619: B}^2}\Theta(h-h_0)\nonumber\\
620: &+&\frac{(\sqrt{2}+1)^2(h-h_{c1})}{\epsilon_{\text
621: F}}\Theta(h_0-h).
622: \end{eqnarray}
623: The global polarization is zero at $h=h_{c1}$ and then increases
624: with $h$.
625:
626: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
627: \subsection {Critical Polarization}
628: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
629: Finally, we convert the above result into the one where both
630: $N_\uparrow$ and $N_\downarrow$ are fixed and the exchange between
631: particles in states $\uparrow$ and $\downarrow$ is forbidden,
632: corresponding to recent experiments on ultracold Fermi gas with
633: population imbalance\cite{exp1,exp2}. The free energy density in
634: this case should be defined as ${\cal
635: F}(n_\uparrow,n_\downarrow)=\mu_\uparrow n_\uparrow+\mu_\downarrow
636: n_\downarrow-{\cal P}$ or ${\cal F}(n,m)=\mu n+ mh-{\cal P}$. The
637: possible phases with nonzero global polarization $P$ are the
638: normal, Sarma and SF-N mixed phases. Since the phase structure
639: should be essentially independent of the ensemble we
640: choose\cite{Sheehy}, we do not need to compare again the free
641: energies of the three phases\cite{BCR}. Since $P(h_{c1})=0$ and
642: $P(h)$ increases with $h$, we conclude that the ground state is
643: the unpolarized superfluid at $P=0$, and SF-N phase separation
644: becomes energetically favored for $0<P<P_c$. The critical
645: polarization $P_c$ where the superfluid bubble disappears
646: completely is the global polarization at $h_{c2}$,
647: \begin{eqnarray}
648: P_c&=&P(h_{c2})=\frac{h_{c2}}{\epsilon_{\text{F}}}\Theta(\eta_2-\eta)+\Theta(\eta-\eta_2)\nonumber\\
649: &=&\sqrt{\eta(1+\frac{\eta}{4})}\Theta(\eta_2-\eta)+\Theta(\eta-\eta_2).
650: \end{eqnarray}
651: The critical polarization increases from $P_c=0$ at $\eta=0$ to
652: $P_c=1$ at $\eta=\eta_2$ and then keeps as a constant $P_c=1$ at
653: strong coupling $\eta>\eta_2$
654:
655: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
656: \begin{figure}[!htb]
657: \begin{center}
658: \includegraphics[width=8cm]{hom01.eps}
659: \includegraphics[width=8cm]{hom02.eps}
660: \caption{The phase diagrams in the planes $\eta-h$ (upper panel)
661: and $\eta-P$ (lower panel). $h$ is scaled by the Fermi energy. SF
662: means unpolarized superfluid, $N_{\text{PP}}$ and $N_{\text{FP}}$
663: indicate the partially and fully polarized normal phases, and
664: SF-N$_{\text {PP}}$ and SF-N$_{\text {FP}}$ are the mixed phases
665: of superfluid and normal gas with $N_{\text{PP}}$ and
666: $N_{\text{FP}}$. \label{fig2}}
667: \end{center}
668: \end{figure}
669: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
670:
671:
672: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
673: \subsection {Phase Diagrams}
674: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
675: Fig.\ref{fig2} summarizes the above analytical results. The phase
676: diagram in the $\eta-h$ plane is shown in the upper panel. The
677: partially and fully polarized normal phases $N_{\text{PP}}$ and
678: $N_{\text{FP}}$ are separated by the dashed line
679: $h/\epsilon_{\text F}=1$ which ends at $\eta=\eta_2$. The two
680: solid lines indicate the lower and upper critical Zeeman fields
681: $h_{c1}$ and $h_{c2}$ with the two phase separations PS-I and
682: PS-II in between. PS-I (SF-N$_{\text {PP}}$)and PS-II
683: (SF-N$_{\text {FP}}$) are the mixed phases of superfluid and
684: normal gas with $N_{\text{PP}}$ and $N_{\text{FP}}$, and they are
685: separated by the dotted line $h/\epsilon_{\text
686: F}=(\sqrt{2}+1)\eta/2$ starting at $\eta=\eta_1$ and ending at
687: $\eta=\eta_2$. The phase diagram in the $\eta-h$ plane can be
688: easily converted into the one in the $\eta-P$ plane shown in the
689: lower panel, by taking the fact $P(h_{c1})=0$ and $P(h_{c2})=P_c$.
690: The critical polarization $P_c=\sqrt{\eta(1+\eta/4)}$ (solid line)
691: increases from $P_c=0$ at $\eta=0$ to $P_c=1$ at $\eta=\eta_2$ and
692: then keeps as a constant $P_c=1$ for $\eta>\eta_2$. The phases SF
693: and $N_{\text{FP}}$ are now located at $P=0$ and $P=1$
694: respectively. The dotted line which separates PS-I from PS-II
695: becomes $P=(4+3\sqrt{2})\eta/2-(\sqrt{2}+1)$ in the $\eta-P$
696: plane.
697:
698: The above analytical results show that, to correctly calculate the
699: critical polarization $P_c$ and the phase diagrams, one should
700: treat the mixed phase carefully~\cite{Sheehy}. Some other methods
701: taken in literatures may lead to quantitatively incorrect results.
702: For instance, the method of stability analysis will result in an
703: incorrect critical polarization (see also the comments in
704: \cite{comment}). With this method, one first solve the mean field
705: gap and number equations for the Sarma phase and then analyze the
706: stability of this phase. If it is applied to the 2D system, the
707: critical polarization becomes~\cite{tem}
708: \begin{eqnarray}
709: \label{eqq}
710: P_c=\frac{\Delta_0}{2\epsilon_{\text
711: F}}=\sqrt{\frac{\eta}{2}},
712: \end{eqnarray}
713: which is the maximum polarization of the unstable Sarma phase and
714: deviates significantly from our result
715: $P_c=\sqrt{\eta(1+\eta/4)}$. Especially, our critical polarization
716: grows up to unity at $\eta\simeq0.828$, but the result (\ref{eqq})
717: becomes unity at $\eta=2$. On the other hand, if one takes only
718: the phases SF and N into account but neglect the phase separation,
719: there will be only one critical field $h_c$ where the polarization
720: jumps from $0$ to $h_c/\epsilon_{\text F}=\sqrt{\eta}$.
721:
722: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
723: \section {Bound State in Polarized Normal Phase}
724: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
725: In recent experiment on highly polarized normal phase in 3D
726: unitary Fermi gas\cite{exp5}, it is found that while the
727: superfluidity disappears completely, full pairing of minority
728: atoms always exists, which indicates that the fermion pairing may
729: be easy to occur in the presence of polarization. It is well known
730: that the bound state in 2D can form at arbitrary small attractive
731: interaction\cite{QM}, which is quite different to the 3D case. It
732: is natural to ask: Do the di-fermion bound states exist above the
733: upper critical field $h_{c2}$ or critical polarization $P_c$?
734:
735: In this section, we study the spectrum of bound states in the
736: highly polarized normal phase. In the Green function method, the
737: energy $\omega$ of the bound states with zero total momentum in
738: this case is determined through the equation~\cite{2Dreview}
739: \begin{equation}
740: \int_{0}^{\infty} dz \left[\frac{1}{2z+\epsilon_{\text B}} -
741: \frac{1-\Theta(\mu_\uparrow-z)-\Theta(\mu_\downarrow-z)}{2z-2\mu-\omega}\right]=0.
742: \end{equation}
743: In the vacuum with $\mu=h=0$, it self-consistently gives the
744: solution $\omega=-\epsilon_{\text B}$. In general case with medium
745: effect, the bound states can survive when the above equation has
746: real solution of $\omega$.
747:
748: The integration in the above equation can be analytically worked
749: out, and finally we obtain
750: \begin{equation}
751: \ln\frac{\omega+2\mu}{-\epsilon_{\text
752: B}}+\Theta(\mu-h)\ln\frac{\omega+2h}{\omega+2\mu}+\Theta(\mu+h)\ln\frac{\omega-2h}{\omega+2\mu}=0.
753: \end{equation}
754: In the partially polarized normal phase, we have
755: $\mu=\epsilon_{\text F}$, the spectrum equation becomes
756: \begin{equation}
757: \omega^2+\epsilon_{\text B}\omega+2\epsilon_{\text
758: F}\epsilon_{\text B}-4h^2=0
759: \end{equation}
760: which has real solutions
761: \begin{equation}
762: \omega=-\frac{1}{2}\left(\epsilon_{\text
763: B}\pm\sqrt{J(h)}\right)\label{ph}
764: \end{equation}
765: for
766: \begin{equation}
767: J(h)=\epsilon_{\text B}^2+16h^2-8\epsilon_{\text B}\epsilon_{\text
768: F}>0.
769: \end{equation}
770: In the fully polarized normal phase, one finds that the spectrum
771: equation directly gives a real solution $\omega=2h-\epsilon_{\text
772: B}$. However, this solution is unphysical since we always have
773: $\omega>0$. This can be well understood when we consider the fact
774: that there exist only $\uparrow$ particles in this phase.
775:
776: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
777: \begin{figure}[!htb]
778: \begin{center}
779: \includegraphics[width=8cm]{difgap.eps}
780: \caption{The excitation gaps for the di-fermions and di-holes as a
781: function of $\eta$ at fixed polarization $P=0.8$. The curves are
782: meaningful only for $\eta<0.56$, since the ground state is not a
783: normal phase for $\eta>0.56$. \label{fig3}}
784: \end{center}
785: \end{figure}
786: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
787:
788: In the balanced normal phase with $h=0$ (note that the true ground
789: state in this case is the superfluid phase), the bound states
790: remain stable only at strong enough coupling $\eta>8$ or
791: equivalently low enough density $\epsilon_{\text
792: F}<\epsilon_{\text B}/8$~\cite{2Dreview}, which indicates that the
793: Fermi sea or medium effect disfavors the formation of bound
794: states. One may simply think that, the presence of a Zeeman
795: splitting will further destroy the bound states. However, this is
796: not true. Since the condition $J(h)>0$ is easier to be satisfied
797: at $h\neq 0$, the bound states in highly polarized normal phase
798: are easier to survive than in the balanced Fermi sea. In the whole
799: partially polarized normal phase which exists in the region
800: $0<\eta<\eta_2$ in Fig.\ref{fig2}, we have $\epsilon_{\text
801: F}\sqrt{\eta(1-\eta/4)}<h<\epsilon_{\text F}$. Analyzing the
802: condition $J(h)>0$, we conclude the bound states can exist in the
803: whole N$_{\text{PP}}$ phase in Fig.\ref{fig2}. Especially, they
804: can survive at high polarization even in the weak coupling limit,
805: as pointed out in~\cite{fum} in the case of 3D.
806:
807: There exist two real solutions for $\omega$ in the N$_{\text{PP}}$
808: phase. The negative~($\omega<0$) and positive~($\omega>0$)
809: solutions in (\ref{ph}) correspond to the excitation gaps
810: $E_{\text{df}}$ and $E_{\text{dh}}$ for the di-fermions(df) and
811: di-holes(dh) respectively\cite{fum},
812: \begin{eqnarray}
813: E_{\text{df}}&=&-\frac{\epsilon_{\text
814: F}}{2}\left(\eta+\sqrt{\eta^2-8\eta+16P^2}\right),\nonumber\\
815: E_{\text{dh}}&=&\frac{\epsilon_{\text
816: F}}{2}\left(\sqrt{\eta^2-8\eta+16P^2}-\eta\right).
817: \end{eqnarray}
818: In Fig.\ref{fig3}, we plot the excitation gaps for di-fermions and
819: di-holes at a fixed polarization $P=0.8$. We found that the
820: symmetry in the spectrum($E_{\text{df}}=-E_{\text{dh}}$) holds
821: only at weak coupling.
822:
823: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
824: \section {Density Profile in a Harmonic Trap}
825: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
826: We have determined the phase diagram for homogeneous system.
827: However, the phase structure can not be directly examined in
828: ultracold Fermi gas experiments, due to the effect of the external
829: harmonic trap. To justify the theoretical prediction for
830: homogeneous system, one should calculate the corresponding phase
831: diagram and the density profile in the presence of an external
832: trap potential, using the same equation of state. In this section,
833: we will calculate analytically the density profile of an imbalance
834: Fermi gas in a 2D isotropic harmonic trap potential
835: $V(r)=\frac{1}{2}M\omega^2r^2$. The frequency $\omega$ here is
836: different from the energy of the bound state defined in Section V.
837:
838: The effect of a harmonic trap can be treated in the local density
839: approximation(LDA). In the frame of LDA, the system is
840: approximately taken to be uniform but with a local chemical
841: potential given by
842: \begin{eqnarray}
843: \mu(r)=\mu_0-\frac{1}{2}M\omega^2r^2,
844: \end{eqnarray}
845: where $\mu_0$ is the chemical potential at the center of the trap
846: and is the true chemical potential(a Lagrangian multiplier) still
847: enforcing the total atom number $N=N_\uparrow+N_\downarrow$. Since
848: $N_\uparrow$ and $N_\downarrow$ are both conserved, the
849: spatially-varying spin-up and spin-down local chemical potentials
850: can be expressed as $\mu_\uparrow(r)=\mu(r)+h$ and
851: $\mu_\downarrow(r)=\mu(r)-h$ in terms of the averaged chemical
852: potential $\mu(r)$ and Zeeman field $h$.
853:
854: To calculate the density profile, namely the atom density of the
855: spin-up and spin-down states as a function of the radius $r$,
856: $n_\uparrow(r)$ and $n_\downarrow(r)$, or equivalently the total
857: density $n(r)=n_\uparrow(r)+n_\downarrow(r)$ and the magnetization
858: $m(r)=n_\uparrow(r)-n_\downarrow(r)$, one should know the equation
859: of state, $n_\sigma(r)=n_\sigma(\mu(r),h)$ with
860: $\sigma=\uparrow,\downarrow$. Using the EOS calculated in Section
861: III, we can determine $\mu_0$ and $h$ from the known total
862: particle number $N$ and the global polarization
863: $P=(N_\uparrow-N_\downarrow)/(N_\uparrow+N_\downarrow)$,
864: \begin{eqnarray}
865: N=2\pi\int rdr n(r),\ \ \ PN=2\pi\int rdr m(r).\label{number}
866: \end{eqnarray}
867:
868: Let us firstly consider a non-interacting system with balanced
869: populations, $N_\uparrow=N_\downarrow$, which can help us to
870: define the Fermi energy $\epsilon_{\text F}$ for trapped 2D
871: system. $\epsilon_{\text F}$ is defined as the chemical potential
872: $\mu_0$ at the center of the trap for non-interacting gas. The
873: density profile is also balanced,
874: $n_{\uparrow}(r)=n_{\downarrow}(r)$, and is given by
875: \begin{eqnarray}
876: n_\uparrow(r)=n_\downarrow(r)=c\left(\epsilon_{\text
877: F}-\frac{1}{2}M\omega^2r^2\right),
878: \end{eqnarray}
879: which vanishes at the so called Thomas-Fermi radius
880: \begin{eqnarray}
881: R_{\text{T}}=\sqrt{\frac{2\epsilon_{\text
882: F}}{M\omega^2}}.\label{TF}
883: \end{eqnarray}
884: The total density $N$ is then given by the integral
885: \begin{eqnarray}
886: N=2\pi \int_0^{R_{\text T}}rdr2c\left(\epsilon_{\text
887: F}-\frac{1}{2}M\omega^2r^2\right)=\left(\frac{\epsilon_{\text
888: F}}{\hbar\omega}\right)^2.\label{FermiE}
889: \end{eqnarray}
890: We find that the Fermi energy in 2D is $\epsilon_{\text
891: F}=\sqrt{N}\hbar\omega$, in contrast to the result
892: $\epsilon_{\text F}=(6N)^{1/3}\hbar\omega$ in 3D (note that we
893: have recovered $\hbar$ in these expressions).
894:
895: We then turn to an attractive Fermi gas. For balanced populations,
896: the ground state is a superfluid state, and the density profile
897: can be obtained from equation (\ref{EOSSF}),
898: \begin{eqnarray}
899: n_\uparrow(r)=n_\downarrow(r)=c\left(\mu_0+\frac{\epsilon_{\text
900: B}}{2}-\frac{1}{2}M\omega^2r^2\right).
901: \end{eqnarray}
902: Comparing with the non-interacting gas, we find no difference
903: between the normal and the superfluid states. The chemical
904: potential at the center of the trap reads
905: \begin{eqnarray}
906: \mu_0=\sqrt{N}\hbar\omega-\frac{\epsilon_{\text
907: B}}{2}=\epsilon_{\text F}-\frac{\epsilon_{\text B}}{2}.
908: \end{eqnarray}
909: This relation is exactly the same as in the homogeneous
910: case\cite{2Da}. The order parameter profile $\Delta(r)$ is given
911: by
912: \begin{eqnarray}
913: \Delta(r)=\Delta_0\sqrt{1-r^2/R^2_{\text T}},\ \ \
914: \Delta_0=\sqrt{2\epsilon_{\text B}\epsilon_{\text F}}.
915: \end{eqnarray}
916:
917: For a system with population imbalance, $N_\uparrow\neq
918: N_\downarrow$, we should have $h\neq0$. By comparing ${\cal
919: P}_{\text{SF}}$ with ${\cal P}_{\text{N}}$, a first order phase
920: transition from the phase SF to the phase N occurs for a given
921: $\mu$ when the Zeeman field $h$ becomes larger than the critical
922: value $h_c(\mu)$, or equivalently speaking, for a given $h$ the
923: SF-N phase transition happens when the chemical potential $\mu$
924: becomes less than the critical value $\mu_c(h)$. In LDA, the phase
925: behavior as a function of chemical potential $\mu$ is translated
926: into a spatial cloud profile through $\mu(r)$. The critical phase
927: boundary $\mu_c$ corresponding to the critical radius $r_c$ is
928: defined by
929: \begin{eqnarray}
930: \mu_c=\mu(r_c)=\mu_0-\frac{1}{2}M\omega^2r_c^2
931: \end{eqnarray}
932: at which the states SF and N have the same pressure. Thus, at
933: fixed $h$, any region of the system which satisfies $\mu(r)>\mu_c$
934: is in the state SF, while a region which satisfies $\mu(r)<\mu_c$
935: will be in the state N. Since $\mu(r)$ decreases with increasing
936: $r$, it is clear that the high density superfluid region will be
937: confined in the center of the trap, and the low density polarized
938: state N is expelled to the outside. The shell structure with
939: radius $r_c$ of the SF-N interface is a striking signature of
940: phase separation in a trap. The superfluid core will disappear
941: when the population imbalance $P$ becomes larger than the critical
942: value $P_c$ which is determined by the equation $r_c=0^+$ or
943: $\mu_0=\mu_c$.
944:
945: We should have two types of shell structure corresponding to the
946: cases $h>h_0$ and $h<h_0$. For the case $h>h_0$, we have
947: $\mu(r_c)=\mu_c=h^2/\epsilon_{\text B}-\epsilon_{\text B}/4>h$,
948: which means that there exists a shell of partially polarized
949: normal gas in the region $r_c<r<r_0$, with $r_0$ given by
950: $\mu(r_0)=h$. We call it the phase PS-I. Thus we have the
951: following density profile
952: \begin{equation}
953: n(r)=
954: \left\{ \begin{array}
955: {r@{\quad,\quad}l}
956: 2c\left(\mu_0+\frac{\epsilon_{\text B}}{2}-\frac{1}{2}M\omega^2r^2\right)&
957: 0<r<r_c \\
958: 2c\left(\mu_0-\frac{1}{2}M\omega^2r^2\right) & r_c<r<r_0\\
959: c\left(\mu_0+h-\frac{1}{2}M\omega^2r^2\right) &
960: r_0<r<R
961: \end{array}
962: \right.
963: \end{equation}
964: and
965: \begin{equation}
966: m(r)=
967: \left\{ \begin{array}
968: {r@{\quad,\quad}l}
969: 0&
970: 0<r<r_c \\
971: 2ch & r_c<r<r_0\\
972: c\left(\mu_0+h-\frac{1}{2}M\omega^2r^2\right) &
973: r_0<r<R
974: \end{array}
975: \right.
976: \end{equation}
977: where $R=\sqrt{2(\mu_0+h)/M\omega^2}$ is the edge of the cloud.
978: After some algebra according to the equation (\ref{number}),
979: $\mu_0$ is simply given by $\mu_0=\epsilon_{\text
980: F}-\epsilon_{\text B}/2$ as in the balanced case and $h$ is
981: solved from the cubic equation
982: \begin{eqnarray}
983: 2h\left(\frac{h^2}{\epsilon_{\text B}}-\frac{\epsilon_{\text
984: B}}{4}\right)=P\epsilon_{\text F}^2,
985: \end{eqnarray}
986: where $\epsilon_{\text F}=\sqrt{N}\hbar\omega$ is the Fermi energy
987: defined in (\ref{FermiE}).
988:
989: From the condition $h>h_0$ which ensures $r_0>r_c$, we have
990: $P\epsilon_{\text F}^2>2h_0^2$, which leads to the relation
991: \begin{eqnarray}
992: P>P_0=\frac{3+2\sqrt{2}}{2}\eta^2.
993: \end{eqnarray}
994: The critical polarization $P_c$ is determined by the condition
995: $\mu_0=\mu_c$. A simple algebra gives
996: \begin{eqnarray}
997: P_c=(2-\eta)\sqrt{\eta-\frac{\eta^2}{4}}, \ \ \ \ \ 0<\eta<\eta_1
998: \end{eqnarray}
999: with $\eta_1=2-\sqrt{2}\simeq0.586$. Note that both $P_0$ and
1000: $P_c$ reach unity at $\eta=\eta_1$, they are the two boundaries of
1001: the phase PS-I in the $\eta-P$ plane.
1002:
1003: For the case $h<h_0$ or $P<P_0$, we have
1004: $\mu(r_c)=\mu_c=(\sqrt{2}h-\epsilon_{\text B})/(2-\sqrt{2})<h$,
1005: which means that the normal gas shell outside the superfluid core
1006: is fully polarized. The density profile reads
1007: \begin{equation}
1008: n(r)=
1009: \left\{ \begin{array}
1010: {r@{\quad,\quad}l}
1011: 2c\left(\mu_0+\frac{\epsilon_{\text B}}{2}-\frac{1}{2}M\omega^2r^2\right)&
1012: 0<r<r_c \\
1013: c\left(\mu_0+h-\frac{1}{2}M\omega^2r^2\right) &
1014: r_c<r<R
1015: \end{array}
1016: \right.
1017: \end{equation}
1018: and
1019: \begin{equation} m(r)=
1020: \left\{ \begin{array}
1021: {r@{\quad,\quad}l}
1022: 0&
1023: 0<r<r_c \\
1024: c\left(\mu_0+h-\frac{1}{2}M\omega^2r^2\right) &
1025: r_c<r<R
1026: \end{array}
1027: \right.
1028: \end{equation}
1029: After the integration in equation (\ref{number}), we still have
1030: $\mu_0=\epsilon_{\text F}-\epsilon_{\text B}/2$ and $h$ is
1031: explicitly given by
1032: \begin{eqnarray}
1033: h=(\sqrt{2}-1)\sqrt{P}\epsilon_{\text F}+\frac{\epsilon_{\text
1034: B}}{2}.
1035: \end{eqnarray}
1036: One can easily check that the condition $h<h_0$ is equivalent to
1037: $P>P_0$, and we have $P_c=1$ for $\eta>\eta_1$.
1038:
1039: Fig.\ref{fig4} summarizes the the coupling-imbalance phase diagram
1040: for two-dimensional imbalanced Fermi gas in a harmonic trap. The
1041: critical polarization $P_c=(2-\eta)\sqrt{\eta-\eta^2/4}$ (solid
1042: line) increases from $P_c=0$ at $\eta=0$ to $P_c=1$ at
1043: $\eta=\eta_1\simeq0.586$ and then keeps as a constant $P_c=1$ for
1044: $\eta>\eta_1$. The dashed line, analytically given by
1045: $P=(3+2\sqrt{2})\eta^2/2$, separates the two types of phase
1046: separation, PS-I and PS-II with different shell structure. In the
1047: phase PS-I, the density profile exhibits a
1048: SF-N$_{\text{PP}}$-N$_{\text{FP}}$ shell structure, while in the
1049: phase PS-II, the shell is in the form of SF-N$_{\text{FP}}$.
1050: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1051: \begin{figure}[!htb]
1052: \begin{center}
1053: \includegraphics[width=8cm]{trap01.eps}
1054: \caption{Global phase diagram for trapped 2D Fermi gas in the
1055: $\eta-P$ plane. \label{fig4}}
1056: \end{center}
1057: \end{figure}
1058: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1059:
1060: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1061: \begin{figure}[!htb]
1062: \begin{center}
1063: \includegraphics[width=8cm]{trap02.eps}
1064: \includegraphics[width=8cm]{trap03.eps}
1065: \caption{The profiles for the total density $n(r) $(solid line)
1066: and magnetization $m(r)$ (dashed line) in two cases, $\eta=0.2$
1067: and $P=0.5$ in the region PS-I and $\eta=0.7$ and $P=0.5$ in the
1068: region PS-II. \label{fig5}}
1069: \end{center}
1070: \end{figure}
1071: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1072:
1073: The analytical result of the density profile can be summarized as
1074: follows. In the region PS-I, we have
1075: \begin{equation}
1076: \frac{n(r)}{n_0}=
1077: \left\{ \begin{array}
1078: {r@{\quad,\quad}l}
1079: 2(1-x^2)&
1080: 0<x<x_c \\
1081: 2\left(1-\frac{\eta}{2}-x^2\right) & x_c<x<x_0\\
1082: 1-\frac{\eta}{2}+\delta-x^2 &
1083: x_0<x<X
1084: \end{array}
1085: \right.
1086: \end{equation}
1087: and
1088: \begin{equation} \frac{m(r)}{n_0}=
1089: \left\{ \begin{array}
1090: {r@{\quad,\quad}l}
1091: 0&
1092: 0<x<x_c \\
1093: 2\delta & x_c<x<x_0\\
1094: 1-\frac{\eta}{2}+\delta-x^2 &
1095: x_0<x<X
1096: \end{array}
1097: \right.
1098: \end{equation}
1099: with $n_0=c\epsilon_{\text F}$, $x=r/R_{\text T}$, $R_{\text T}$
1100: being the Thomas-Fermi radius of non-interacting gas defined in
1101: ({\ref{TF}), and $\delta=h/\epsilon_{\text F}$ being the real
1102: solution of the cubic equation
1103: $\delta^3-\eta^2\delta/4-P\eta/2=0$,
1104: \begin{eqnarray}
1105: \delta&=&\left(\frac{P\eta}{4}\right)^{1/3}\left[\left(1+\gamma\right)^{1/3}+\left(1-\gamma\right)^{1/3}\right]
1106: \end{eqnarray}
1107: with $\gamma=\sqrt{1-\eta^4/(108P^2)}$. The scaled radii
1108: $x_c=r_c/R_{\text T},x_0=r_0/R_{\text T}$ and $X=R/R_{\text T}$
1109: are given by
1110: \begin{eqnarray}
1111: x_c&=&\sqrt{1-\frac{\eta}{2}-\frac{P}{2\delta}},\nonumber\\
1112: x_0&=&\sqrt{1-\frac{\eta}{2}-\delta},\nonumber\\
1113: X&=&\sqrt{1-\frac{\eta}{2}+\delta}.
1114: \end{eqnarray}
1115: A numerical sample for $\eta=0.2,P=0.5$ is shown in
1116: Fig.\ref{fig5}(a). There is an interesting phenomenon which is
1117: different from that found in 3D: The magnetization profile $m(r)$
1118: exhibits a visible platform structure in the partially polarized
1119: normal shell in the region $r_c<r<r_0$. For partially polarized
1120: gas, the interaction may be important, like the finding around the
1121: unitary region in 3D~\cite{unitary,MC1,MC2}. However, for the 2D
1122: system, since partially polarized normal shell appears only at
1123: small coupling where the effect of interaction is not important,
1124: our conclusion will not be qualitatively changed. We also observe
1125: a density jump $\Delta n$ at the critical radius $r_c$. In the
1126: region PS-I, $\Delta n$ is independent of the global polarization
1127: and depends only on the coupling strength,
1128: \begin{eqnarray}
1129: \Delta n=\eta n_0=\frac{M}{2\pi}\epsilon_{\text B}.
1130: \end{eqnarray}
1131: Thus the experimental data for $\Delta n$ can be used to extract
1132: the effective two-body binding energy $\epsilon_{\text B}$.
1133:
1134: In the region PS-II, the density profile reads
1135: \begin{equation}
1136: \frac{n(r)}{n_0}=
1137: \left\{ \begin{array}
1138: {r@{\quad,\quad}l}
1139: 2(1-x^2)&
1140: 0<x<x_c \\
1141: 1+(\sqrt{2}-1)\sqrt{P}-x^2 &
1142: x_c<x<X
1143: \end{array}
1144: \right.
1145: \end{equation}
1146: and
1147: \begin{equation} \frac{m(r)}{n_0}=
1148: \left\{ \begin{array}
1149: {r@{\quad,\quad}l}
1150: 0&
1151: 0<x<x_c \\
1152: 1+(\sqrt{2}-1)\sqrt{P}-x^2 &
1153: x_c<x<X
1154: \end{array}
1155: \right.
1156: \end{equation}
1157: The scaled radii $x_c$ and $X$ now takes very simple form
1158: \begin{eqnarray}
1159: x_c&=&\sqrt{1-\sqrt{P}},\nonumber\\
1160: X&=&\sqrt{1+(\sqrt{2}-1)\sqrt{P}}.
1161: \end{eqnarray}
1162: A numerical sample of the density profile for $\eta=0.7,P=0.5$ is
1163: shown in Fig.\ref{fig5}(b). It is very surprising that the density
1164: profile does not depend on the coupling parameter $\eta$, but only
1165: on the global polarization $P$. As a result, the ratio $r_c/R$
1166: exhibits a universal behavior when $\eta>\eta_1=0.586$, as shown
1167: in Fig.\ref{fig6}.
1168: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1169: \begin{figure}[!htb]
1170: \begin{center}
1171: \includegraphics[width=8cm]{trap04.eps}
1172: \caption{The ratio of the superfluid radius to the cloud radius as
1173: a function of the global polarization $P$ for $\eta>\eta_1=0.586$.
1174: \label{fig6}}
1175: \end{center}
1176: \end{figure}
1177: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1178:
1179: Finally, two comments on our results should be made. The first is
1180: on the BCS-Leggett mean field theory. In this theory, the quantum
1181: fluctuation in the superfluid phase and the interaction in the
1182: partially polarized phase are totally neglected. The effect of
1183: interaction in the partially polarized phase may change the
1184: platform structure in the region $r_c<r<r_0$. However, since the
1185: three-shell structure appears in the weak coupling region, we
1186: expect this effect to be small. At very strong coupling, the
1187: correction in the superfluid phase due to quantum fluctuation
1188: should be important, and the universal behavior in the region
1189: PS-II may be destroyed. Since the BEC region is reached at
1190: $\eta>2$, we expect that our conclusion holds at the BCS side
1191: $\eta_1<\eta<2$. The second comment is on the model we used.
1192: Recently, it is argued that the model we used is not sufficient to
1193: discuss BCS-BEC crossover in quasi-2D Fermi gas due to the
1194: importance of dressed molecules\cite{zhang}. However, from the
1195: study in \cite{zhang}, this effect is important only at strong
1196: coupling (may be for $\eta>2$). Obviously, the comparison of our
1197: prediction with the experimental data can tell us whether the
1198: quantum fluctuation, dressed molecules and other possible effects
1199: are important.
1200:
1201: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1202: \section {Summary}
1203: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1204: In summary, mean field phase structure of polarized Fermi gas in
1205: 2D is analytically investigated. In the normal phase, the
1206: di-fermion bound states at high polarization are easier to survive
1207: than in the balanced Fermi sea. In the BCS-Leggett mean field
1208: theory, the transition from the unpolarized superfluid phase to
1209: the normal phase is always of first order, and there exists no
1210: stable gapless superfluid phase. In the homogeneous system, we
1211: analytically determined the critical Zeeman fields and the
1212: critical population imbalance in the whole coupling parameter
1213: region. We found two critical Zeeman fields in the BCS-BEC
1214: crossover, and proved that the mixed superfluid-normal phase is
1215: the energetically favored ground state. However, from recent
1216: Monte-Carlo simulations~\cite{MC1,MC2}, our mean field results may
1217: be only qualitatively correct in some parameter region, due to the
1218: importance of interactions in the normal phase.
1219:
1220: To compare our theoretical results with future experimental data,
1221: we have also calculated analytically the density profile for an
1222: imbalanced 2D Fermi gas confined in a harmonic trap. For balanced
1223: populations, the density profiles for normal and superfluid matter
1224: are the same and can not be used as a signature of superfluidity.
1225: For imbalanced populations, the density profile exhibits a shell
1226: structure, a superfluid core in the center and a normal shell
1227: outside. At small coupling, there exists a partially polarized
1228: normal shell and the density difference shows a platform
1229: structure. For large attraction, however, the normal shell is
1230: fully polarized, and the density profile depends only on the
1231: global population imbalance. Our theoretical prediction can be
1232: examined in the future experiments on 2D ultracold Fermi gases,
1233: which can help us to see whether quantum fluctuations and other
1234: possible effects are important in determining the phase
1235: structure\cite{zhang}.
1236:
1237: {\bf Acknowledgments:}\ The work is supported by the NSFC Grants
1238: 10575058 and 10735040 and the National Research Program Grant
1239: No.2006CB921404.
1240:
1241: \begin{thebibliography}{99}
1242: \bibitem{CC}
1243: B.S.Chandrasekhar, Appl. Phys. Lett.{\bf 1},7(1962); A.M.Clogston,
1244: Phys. Rev. Lett.{\bf 9}, 266(1962)
1245: \bibitem{Sarma}
1246: G.Sarma, J. Phys. Chem. Solid {\bf 24},1029(1963)
1247: \bibitem{FFLO}
1248: P.Fulde and R.A.Ferrell, Phys. Rev {\bf A135}, 550(1964);
1249: A.I.Larkin and Yu.N.Ovchinnikov, Sov. Phys. JETP {\bf 20},
1250: 762(1965)
1251: \bibitem{exp1}
1252: M.W.Zwierlein, et al., Science {\bf 311}, 492(2006)
1253: \bibitem{exp2}
1254: G.B.Partridge, et al., Science {\bf 311}, 503(2006)
1255: \bibitem{exp3}
1256: M.W.Zwierlein, et.al., Nature {\bf 442}, 54(2006)
1257: \bibitem{exp4}
1258: Y.Shin, et.al., Phys. Rev. Lett.{\bf 97}, 030401(2006)
1259: \bibitem{exp5}
1260: C.H.Schunck, et.al., Science {\bf 316}, 867(2007)
1261: \bibitem{exp6}
1262: Y.Shin, et.al., arXiv:0709.3027
1263: \bibitem{BCR}
1264: P.F.Bedaque, et al., Phys. Rev. Lett.{\bf 91}, 247002(2003)
1265: \bibitem{Cohen}
1266: T.D.Cohen, Phys. Rev. Lett.{\bf 95}, 120403(2005)
1267: \bibitem{Carlson}
1268: J.Carlson and S.Reddy, Phys. Rev. Lett.{\bf 95}, 060401(2005)
1269: \bibitem{Sheehy}
1270: D.E.Sheehy and L.Radzihovsky, Phys. Rev. Lett. {\bf 96},
1271: 060401(2006); Ann. Phys. (N.Y.){\bf 322}, 1790(2007);
1272: \bibitem{MFwork}
1273: C.H.Pao, et al., Phys. Rev. {\bf B73}, 132506(2006); Z.C.Gu,
1274: et.al., arXiv:cond-mat/0603091; H.Hu and X.Liu, Phys. Rev. {\bf
1275: A73}, 051603(R)(2006); M.Iskin and C.A.R.Sa de Melo, Phys. Rev.
1276: Lett. 97, 100404 (2006); L.He, M.Jin and P.Zhuang, Phys.Rev.{\bf
1277: B73}, 214527(2006); Phys.Rev.{\bf B74}, 214516(2006)
1278: \bibitem{MC1}
1279: C.Lobo, et.al., Phys. Rev. Lett. {\bf 97}, 200403(2006)
1280: \bibitem{MC2}
1281: S.Pilati and S.Giorgini, arXiv:0710.1549
1282: \bibitem{trap}
1283: J.Kinnunen, et.al., Phys. Rev. Lett.{\bf 96}, 110403(2006);
1284: M.Haque and H.T.C.Stoof, Phys. Rev. {\bf A74}, 011602(2006); W.Yi
1285: and L.-M. Duan, Phys. Rev. {\bf A73}, 031604(R)(2006); K.Machida,
1286: et.al., Phys. Rev. Lett. {\bf 97}, 120407(2006); T.N.De Silva and
1287: E.J. Mueller, Phys. Rev. {\bf A73}, 051602(R)(2006); Phys. Rev.
1288: Lett. {\bf 97}, 070402(2006);
1289: \bibitem{1D}
1290: Hui Hu, et.al., Phys. Rev. Lett.{\bf 98}, 070403(2007); G.Orso,
1291: Phys. Rev. Lett.{\bf 98}, 070402(2007)
1292: \bibitem{2Da}
1293: M.Randeria, J.-M. Duan and L.-Y. Shieh, Phys. Rev. Lett.{\bf 62},
1294: 981(1989); Phys.Rev.{\bf B41}, 327(1990)
1295: \bibitem{2Db}
1296: M.Marini, F.Pistolesi and G.C.Strinati, Eur.Phys.J.{\bf 1}, 151
1297: (1998)[arXiv:cond-mat/9703160]
1298: \bibitem{2Dc}
1299: E.Babaev and H.Kleinert, Phys.Rev.{\bf B59}, 12083(1999);
1300: E.Babaev, Phys.Rev.{\bf B63}, 184514(2001)
1301: \bibitem{2Dreview}
1302: V.M.Loktev, et al., Phys. Rept. {\bf 349}, 1(2001)
1303: \bibitem{guba}
1304: E.Gubankova, et.al., Phys. Rev. {\bf B74}, 064505(2006)
1305: \bibitem{liu}
1306: For cold atom gas, the transfer can be realized in optical
1307: lattices via Raman detuning, see W.V.Liu, et.al., Phys. Rev. {\bf
1308: A70}, 033603(2004)
1309: \bibitem{BCSBEC}
1310: A.J.Leggett, in \emph{Modern trends in the theory of condensed
1311: matter}, Springer-Verlag, Berlin, 1980, pp.13-27
1312: \bibitem{QM}
1313: L.D. Landau and E.M. Lifshitz, {\it Quantum Mechanics. Non
1314: Relativistic Theory. Course of Theoretical Physics}, Vol. 3
1315: (Pergamon Press, New York, 1989).
1316: \bibitem{2Dexp}
1317: S.Stock, et.al., Phys.Rev.Lett.{\bf 95}, 190403(2005); Z.
1318: Hadzibabic, et.al., Nature{\bf 441}, 1118(2006)
1319: \bibitem{zhang}
1320: W.Zhang, G.-D.Lin and L.-M. Duan, arXiv:0803.2488
1321: \bibitem{unitary}
1322: F.Chevy, Phys. Rev. Lett.{\bf 96}, 130401(2006); Phys. Rev. {\bf
1323: A74}, 063628(2006); A.Bulgac and M.M.Forbes, Phys. Rev. {\bf A75},
1324: 031605(2007)
1325: \bibitem{surface}
1326: T.N.De Silva and E.J. Mueller, Phys. Rev. Lett. {\bf 97},
1327: 070402(2006); H.Caldas, J.Stat.Mech. P11012(2007)
1328: \bibitem{tem}
1329: J.Tempere, et.al., Phys. Rev. {\bf B75}, 184526(2007)
1330: \bibitem{comment}
1331: D.E.Sheehy and L.Radzihovsky, Phys. Rev. {\bf B75}, 136501(2007)
1332: \bibitem{fum}
1333: F.Fumarola, et.al., arXiv:cond-mat/0703003
1334: \end{thebibliography}
1335: \end{document}
1336: