1: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

2: % Molecule -- Fermionic pair coupling in a Finite System

3: % Since May 9

4: % Last Update Sep 24

5: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

6: \documentclass[aps,pra,twocolumn,showpacs,groupedaddress]{revtex4}

7: \usepackage{graphicx}

8: %----------------------------

9: \def\bra{\langle}

10: \def\ket{\rangle}

11: \def\pd{\partial}

12: \def\su{\uparrow}

13: \def\sd{\downarrow}

14: \def\bs{\bar{s}}

15: %----------------------------

16: \begin{document}

17: 

18: \title{Boson-Fermion coherence in a spherically symmetric harmonic trap}

19: \author{Takahiko Miyakawa and Pierre Meystre}

20: \affiliation{Optical Sciences Center,

21: The University of Arizona, Tucson, AZ 85721}

22: 

23: \date{\today}

24: 

25: \begin{abstract}

26: We consider the photoassociation of a low-density gas of

27: quantum-degenerate trapped fermionic atoms into bosonic molecules

28: in a spherically symmetric harmonic potential. For a dilute system

29: and the photoassociation coupling energy small

30: compared to the level separation of the trap, only those fermions

31: in the single shell with Fermi energy are coupled to the bosonic

32: molecular field.

33: Introducing a collective pseudo-spin operator formalism we show

34: that this system can then be mapped onto the Tavis-Cummings

35: Hamiltonian of quantum optics, with an additional pairing

36: interaction. By exact diagonalization of the Hamiltonian, we

37: examine the ground state and low excitations of the Bose-Fermi

38: system, and study the dynamics of the coherent coupling between

39: atoms and molecules. In a semiclassical description of the system,

40: the pairing interaction between fermions is shown to result in a

41: self-trapping transition in the photoassociation, with a sudden

42: suppression of the coherent oscillations between atoms and

43: molecules. We also show that the full quantum dynamics of the

44: system is dominated by quantum fluctuations in the vicinity of the

45: self-trapping solution.

46: \end{abstract}

47: 

48: \pacs{03.75.Lm, 03.75.Ss, 42.50.Ar} \maketitle

49: %

50: \section{Introduction}

51: 

52: The formation of ultracold diatomic molecules from Feshbach

53: resonances and photoassociation has witnessed spectacular

54: developments in recent years. Early demonstrations of molecule

55: formation using two-photon Raman photoassociation \cite{Wynar00}

56: and a Feshbach resonance \cite{Inouye98} were dominated by the

57: molecular losses due to processes such as inelastic decay to lower

58: energy molecular vibrational states \cite{Yurovsky99}, so that the

59: existence of the molecules could only be inferred from the

60: decrease in the number of atoms. The first unambiguous coherent

61: conversion of atoms into molecules was performed by Donley {\it et

62: al.} \cite{Donley02}, who exploited a Feshbach resonance in a

63: $^{85}$Rb Bose-Einstein Condensate (BEC). In subsequent experiments starting from an

64: atomic condensate of $^{87}$Rb, Rempe and coworkers used adiabatic

65: rapid passage to create the molecules \cite{Durr04}. Because the

66: molecules and atoms have different magnetic moments, they could be

67: spatially separated from each other using a magnetic field

68: gradient via the Stern-Gerlach effect. Similar work has been

69: conducted by Xu {\it et al.} \cite{Xu03}. Starting from a sodium

70: BEC, they used resonant laser light to blast away the remaining

71: atoms in the sample and isolated the molecules. Unfortunately, the

72: conversion efficiency was limited by inelastic losses very close

73: to the resonance so that molecular yields were $\lesssim 10\%$.

74: 

75: For fermionic atoms close to a Feshbach resonance the inelastic

76: collision rate for relaxation  to lower energy vibrational states

77: of the molecules scales like $a(B)^{-2.55}$ whereas for bosons it

78: scales like $a(B)$ where $a(B)$ is the scattering length near the

79: resonance \cite{Petrov04}. This is because close to resonance the

80: effective size of the molecules is of the order $a(B)$, which is

81: comparable to the interparticle spacing. In order for a molecule

82: to decay to a more deeply bound vibrational state with radius

83: $R_e\ll a(B)$, the atoms comprising the molecule along with an

84: additional atom must all collide within a distance $\sim R_e$.

85: Since two of the three atoms are necessarily identical, the

86: collision rate is suppressed for fermions. Taking advantage of

87: this consequence of the Pauli Exclusion Principle, Greiner {\it et

88: al.} \cite{Greiner03} achieved the first molecular BEC by starting from a quantum degenerate spin mixture of

89: $^{40}$K using adiabatic rapid passage through a Feshbach

90: resonance with a conversion efficiency of $\sim 80\%$. At about

91: the same time Zwierlein {\it et al.} \cite{Zwierlein03} and Jochim

92: {\it et al.} \cite{Jochim03} succeeded in producing a BEC of

93: $^{6}$Li$_2$ dimers by evaporatively cooling the atoms at a

94: constant magnetic field just below a resonance where $a(B)$ is

95: large and positive. Two recent experiments have led to the

96: observation of heteronuclear Feshbach resonances in Bose-Fermi

97: mixtures of $^6$Li and $^{23}$Na \cite{Stan04} in one case, and of

98: $^{87}$Rb and $^{40}$K in the other \cite{Inouye04}. We also

99: mention recent experiments by Kerman {\it et al.} \cite{Kerman04},

100: who produced metastable RbCs molecules in their lowest triplet

101: state starting from a laser-cooled mixture of $^{85}$Rb and

102: $^{133}$Cs by photoassociation. Major current experimental and

103: experimental efforts are directed towards the exploration of the

104: crossover between the Bardeen-Cooper-Schrieffer (BCS) state of

105: fermionic atoms and BEC of bosonic

106: molecules as the system crosses a Feshbach resonance

107: \cite{ex_BECBCS}. In the vicinity of these resonances, the system

108: enters a strongly interacting regime that offers a challenge for

109: many-body theories \cite{Timmermans,th_BECBCS}.

110: 

111: It is known that the pairing properties of finite-size systems can

112: be significantly different from those of the bulk material, due to

113: the discrete energy spectrum of the particles involved. The

114: detailed role of the shell structure has been explored both in the

115: nuclei~\cite{Mottelson}  and of superconductor

116: grains~\cite{SCgrain}, in which a collective character of pairs

117: plays an important role. In this paper we consider the

118: photoassociation of a dilute quantum-degenerate gas of fermionic

119: atoms trapped in a spherically symmetric harmonic trap into

120: molecular dimers, including the pairing interaction between

121: fermions. By introducing a pseudo-spin formalism for the time

122: reversal pairing operator~\cite{Anderson,Kerman61}, this problem

123: can be mapped onto an extension of the Tavis-Cummings

124: model~\cite{Tavis,Bogoliubov} that describes the coupling of $N$

125: two-level atoms and a single mode of the electromagnetic field,

126: and has found applications in the study of superradiance in

127: quantum optics~\cite{Dicke,Bonifacio}. This analogy allows us to

128: study in detail the ground state and lower excited states of the

129: system, as well as the coherent dynamics of atom-molecule coupling

130: in the trap. We show that for appropriate conditions, only those

131: fermions on the last energy shell of the trap participate in the

132: photoassociation process, and discuss the impact of the filling of

133: that shell on molecule formation. We find that depending on the

134: detuning $\delta$ of the photoassociation laser from the energy

135: difference between the molecules and atom pairs, the nature of the

136: ground state changes from being predominantly atomic to

137: predominantly molecular in nature. We study the crossover between

138: these two regions in detail, and quantify its property via the

139: joint coherence of the atomic and molecular fields and the

140: entanglement entropy of the system.

141: 

142: The coherent conversion of fermions into bosons has been studied

143: in the homogeneous case by several

144: authors~\cite{Timmermans,dy_BECBCS,Vardi}. Trapped systems, in

145: addition to being closer to the experimental situation, present

146: several unique characteristics: First, the discreteness of the

147: energy levels eliminates many of the difficulties associated with

148: a continuum. In addition, the high degeneracy of spherically

149: symmetric harmonic potentials simplifies significantly the study

150: of coherent quantum dynamics. Indeed, the problem resembles then

151: the dynamics of a bosonic Josephson Junction~\cite{BJJ1,BJJ2},

152: although the nonlinear coupling between fermionic atoms and

153: bosonic molecules leads to considerably richer dynamics. Moreover

154: the additional pairing interaction between fermions is shown to

155: result in a self-trapping transition~\cite{BJJ1,Scott}, with a

156: sudden suppression of the coherent oscillations.

157: 

158: This paper is organized as follows. Section II discusses our model

159: and formulates it in terms of a pseudo-spin formalism. Section III

160: presents results of the static problem where the many-body states

161: are classified by number of unpaired fermions known as {\it

162: seniority}~\cite{Racah} in nuclear physics. We show that for a

163: fixed number of atoms, the ground state of the system is always

164: the state of minimum seniority. We examine the crossover behavior

165: of the ground state as the detuning parameter $\delta$ is varied,

166: as a function of the ratio of the total number of fermionic pairs

167: and molecules to the degenerate number of the Fermi level. The

168: entanglement between fermions and bosons is evaluated and found to

169: be reduced as the pairing interaction becomes stronger. In section

170: IV, we analyze the coherent dynamics of the nonlinear

171: atom-molecule coupling. Using a semiclassical factorization

172: ansatz, we show the appearance of a self-trapping transition in

173: the presence of pairing interaction. An exact quantum solution

174: shows that around that transition point the dynamics is

175: characterized by large quantum fluctuations. Finally, section V is

176: a summary and outlook. Calculational details are relegated to an

177: appendix.

178: 

179: %

180: \section{Model}

181: 

182: We consider a trapped dilute gas of two-component fermionic atoms

183: in hyperfine states of spin $\sigma=\su,\sd$ at zero temperature

184: and coupled to a single-mode gas of bosonic molecules via a

185: two-photon Raman transition. The trap is assumed to be harmonic

186: and spherically symmetric, described by the potential

187:     \begin{equation}

188:     V_f=\frac{1}{2}m_f \omega_{\rm ho}^2 {\bf r}^2

189:     \end{equation}

190: for the atoms, and similarly with $f \rightarrow b$ for the

191: bosonic molecules.

192: 

193: In the absence of interactions between particles, the trap levels

194: have the energies

195:     \begin{equation}

196:     {\cal E}_{n}=\left(n+\frac{3}{2}\right)\hbar\omega_{\rm

197:     ho}.

198:     \end{equation}

199: where the principal quantum number $n$ is positive or zero. In

200: order to deal with the high degree of degeneracy of this

201: potential, it is convenient to introduce the (integer) radial and

202: angular quantum numbers $n_r$ and $l$, which are positive or zero,

203: with \cite{CT}

204:     \begin{equation}

205:     n= 2n_{r} + l.

206:     \end{equation}

207: Each pair $(n_r, l)$ corresponds to a radial wave function

208: $R_{n,l}(r)$ and hence $(2l+1)$ common eigenfunctions of $V_f(r)$,

209: ${\bf L}^2$ and $L_z$,

210:     \begin{equation}

211:     \phi_{n,l,m}({\bf r})= R_{n,l}(r)

212:     Y_{lm}(\theta,\phi).

213:     \end{equation}

214: Taking into account the magnetic quantum number

215:     \begin{equation}

216:     -l \le m \le +l,

217:     \end{equation}

218: the degeneracy of each level ${\cal E}_n$ is therefore

219:     \begin{equation}

220:     \Omega_n = \frac{1}{2} (n + 1)(n +2),

221:     \end{equation}

222: and the total number of states up to the shell $n_F$

223: corresponding to the Fermi energy

224:     $$

225:     {\cal E}_F = \left(n_F + \frac{3}{2}\right) \hbar \omega_{\rm ho}

226:     $$

227: is

228: \[

229:     N_{n_F}=2\times

230:     \sum_{n=0}^{n_F}\Omega_{n}=\frac{1}{3}(n_F+1)

231:     (n_F+2)(n_F+3),

232: \]

233: where the factor of 2 accounts for the two hyperfine spin states

234: of the atoms. In the following, it will be necessary to include

235: the attractive interaction responsible for the pairing between

236: fermions. Since as we show later on this interaction splits the

237: degeneracy of the trap levels, we keep the angular momentum index

238: in the labelling of the atomic energies, ${\cal E}_{n,l}$. The

239: fermionic atoms are therefore described by the annihilation

240: operator $c_{n;l,m,\sigma}$, where $\sigma$ labels the hyperfine

241: spin state of the atom with single-particle wave function

242: $\phi_{n,l,m}({\bf r})$ and eigenenergy ${\cal E}_{n,l}$.

243: 

244: Assuming that the atom-molecule photoassociation energy is smaller

245: than the trap energy spacing, $\hbar\omega_{\rm ho}$, and in

246: addition that the system is sufficiently dilute that the

247: attractive interaction between fermions is likewise less than

248: $\hbar\omega_{\rm ho}$, it is possible to tune the frequency of

249: the photoassociation laser so as to only couple fermions in the

250: shell $n_F$ with Fermi energy ${\cal E}_F$ to molecules in the ground

251: state of the harmonic trap. We can then ignore all shells other than the

252: $n_F$-shell for the fermions, and all trap states above the ground state

253: for the molecules, which are then described in terms of the ground

254: state bosonic annihilation operator $b$ with single particle

255: energy ${\cal E}_0$.

256: 

257: Both atomic pairing and photoassociation involve the creation and

258: annihilation of pairs of atoms, hence it is convenient to

259: introduce pseudo-spin operators $S_l$~\cite{Kerman61}

260: for atoms of angular momentum $l$ in the $n_F$-shell as

261: \begin{eqnarray}

262:     S_l^+&=&\sum^l_{m=-l}(-1)^{l-m} c_{lm\su}^\dagger

263:     c_{l-m\sd}^\dagger,\\

264:     S_l^-&=&\sum^l_{m=-l}(-1)^{l-m} c_{l-m\sd}

265:     c_{lm\su},\\

266:     S_l^z&=&\frac{1}{2}\left[\sum_{m,\sigma}c_{lm\sigma}^\dagger

267:     c_{lm\sigma}-\Omega_l\right]=\frac{1}{2}\left(\hat{n}_l-\Omega_l\right),

268:     \end{eqnarray}

269: where $\Omega_l = 2l+1$ and $\hat{n}_l$ is number operator in each

270: level $l$. (Here and hereafter we have omitted the label $n_F$

271: from the fermionic operator $c_{n_F;l,m,\sigma}$.) They are easily

272: seen to obey the SU(2) algebra

273:     \begin{equation}

274:     [S_l^+,S_{l'}^-]=2S_l^z \delta_{l,l'}, \,\,\,\,\,\,\,

275:     [S_l^z,S_{l'}^{\pm}]=\pm S_l^\pm \delta_{l,l'}.

276:     \end{equation}

277: 

278: Since we need only consider atoms in the Fermi level, the total

279: number of relevant particles in the system is

280:     \begin{equation}

281:     N= n_p + n_b + \nu \equiv M + \nu,

282:     \label{Ntotal}

283:     \end{equation}

284: where

285:     \begin{equation}

286:     M= n_p + n_b

287:     \end{equation}

288: is the number of molecules ($n_b$) and atomic pairs ($n_p$)

289: in the $n_F$-shell, or loosely speaking the number of pairs,

290: and $\nu$ is the number of unpaired atoms in the $n_F$-shell.

291: In that reduced Hilbert space, a complete set of

292: states is given by

293:     \begin{eqnarray}

294:     &&|n_{l},n_{l^\prime},\cdots,n_{l^{\prime\prime}},n_b;\nu\ket\nonumber\\

295:     &&\quad =\frac{1}{\sqrt{\mathcal N}}

296:     (S_l^+)^{n_l}(S_{l^\prime}^+)^{n_{l^\prime}}

297:     \cdots(S_{l^{\prime\prime}}^+)^{n_{l^{\prime\prime}}}

298:     (b^\dagger)^{n_b}|\nu\ket,

299:     \end{eqnarray}

300: where $\mathcal N$ is a normalization constant.

301: 

302: For a given angular momentum $l$, the possible number of atomic

303: pairs $n_l$ in the Fermi level is

304:     $$

305:     0\leq n_l \leq \Omega_l = 2l+1,

306:     $$

307: while the number of molecules is $0\leq n_b \leq N$.

308: 

309: The pseudo-spin operator $S_l^{-}$ annihilates atoms in pairs,

310: hence any state

311: $|\nu\ket\equiv|\nu_l,\nu_{l^\prime},\cdots,\nu_{l^{\prime\prime}}\ket$

312: of unpaired fermions and zero molecules clearly satisfies

313:     \begin{equation}

314:     S_l^{-}|\nu\ket=0,\,\,\,\,\,\, \hat{n}_b|\nu\ket=0,

315:     \label{nu-definition1}

316:     \end{equation}

317: with

318:     \begin{equation}

319:     \hat{n}_l|\nu\ket=\nu_l|\nu\ket,

320:     \label{nu-definition2}

321:     \end{equation}

322: see Eq. (\ref{Ntotal}). The number operator of bosonic molecules has been

323: defined by $\hat{n}_b=b^\dagger b$. $\nu_l$ is the number of unpaired fermions of

324: angular momentum $l$ in each level and is

325: referred to as {\it seniority}~\cite{Racah} in nuclear physics.

326: 

327: With this formal development at hand, we now turn to the

328: discussion of the fermionic pairing and of the photoassociation of

329: atoms into molecules. It is described by the effective Hamiltonian

330: \begin{equation}

331: \label{modelH} H=(\hbar \delta+{\cal E}_0) b^\dagger

332: b+\sum_{l,m,\sigma}{\cal E}_{n_F,l} c_{lm\sigma}^\dagger

333: c_{lm\sigma} +V_{p} +V_{am},

334: \end{equation}

335: where $\delta$ is the two-photon detuning between the Raman lasers

336: and the internal energy difference between atomic pairs and

337: molecules, $V_p$ describes atomic pairing and $V_{am}$ accounts

338: for the photoassociation of atoms into molecules. Before

339: discussing these two interaction Hamiltonians in detail, we first

340: evaluate the mean-field lifting of the single-particle energy

341: degeneracy, ${\cal E}_{n_F} \rightarrow {\cal E}_{n_F,l}$.

342: 

343: In the $s$-wave scattering approximation, valid at $T=0$, atoms of

344: opposite spin interact via the two-body interaction potential

345:     \begin{equation}

346:     V({\bf r}_1 - {\bf r}_2) = \frac{4 \pi \hbar^2 a}{m_f}

347:     \delta({\bf r}_1 - {\bf r}_2),

348:     \end{equation}

349: where $a <0$ is the scattering length, negative for attractive

350: interactions. In the Thomas-Fermi limit, this results in the atoms

351: being subjected to the mean-field potential

352: \begin{equation}

353: V(r)=\frac{2\pi\hbar^2a}{m_f}\rho(r),

354: \end{equation}

355: where the density $\rho(r)$ is given by

356: \begin{equation}

357: \rho(r)\simeq \rho_0(1-r^2/R^2_{TF})^{3/2},

358: \end{equation}

359: for

360:     $$

361:     r\leq R_{\rm TF}=a_{\rm ho}\sqrt{2n_F+3}

362:     $$

363: and is zero otherwise. Here $a_{\rm ho}=\sqrt{\hbar/m_f

364: \omega_{\rm ho}}$ is the oscillator length and

365: $\rho_0=(2n_F+3)^{3/2}/3\pi^2a_{\rm ho}^3$. The resulting mean-field

366: energy splitting of the $l$-states within the $n_F$ manifold is

367: then \cite{Heiselberg}

368: \begin{eqnarray}

369: \label{MFshift}

370: {\cal E}_{n_F,l}&-&{\cal E}_{n_F}\nonumber\\&=&\int\,dr\, r^2V(r)|R_{n_F,l}|^2

371: \\

372: &\simeq&\frac{2}{3\pi}\frac{a}{a_{\rm

373: ho}}(2n_F+3)^{3/2}\hbar\omega_{\rm ho}\left[

374: \frac{4}{3\pi}-\frac{1}{4\pi}\frac{l(l+1)}{n_F^2}\right]\nonumber,

375: \end{eqnarray}

376: where we have used the WKB limit of the radial harmonic oscillator

377: wave function which is valid for $n_F\gg 1$.

378: For an atomic system to be dilute, the mean-field shift Eq.(\ref{MFshift})

379: should be less than the unperturbed energy ${\cal E}_{n}$. This

380: implies that

381: \begin{equation}

382: n_F^{1/2}\frac{|a|}{a_{\rm ho}}\ll 1,

383: \end{equation}

384: which is equivalent to the familiar diluteness condition

385: $\rho_0|a|^3\ll 1$.

386: 

387: It is known that for attractive short range interactions, the potential

388: $V({\bf r}_1-{\bf r}_2)$ favors the creation of a time reversal

389: state and lets pairing take place. The Hamiltonian $V_p$ describes

390: this pairing correlation by including only the correlations for time reversal

391: pair states

392:     \begin{equation}

393:     |L=0,M=0;ll\ket=\sum_{m=-l}^l (l m ,l -m|0 0)|l,m\ket|l,-m\ket,

394:     \end{equation}

395: where

396:     \begin{equation}

397:     (l m,l -m|0 0)=(-1)^{l-m}(2l+1)^{-1/2}

398:     \label{CB-coeff}

399:     \end{equation}

400: is a Clebsch-Gordan coefficient. Assuming for simplicity that the

401: radial part $g$ of the pairing interaction is independent of the

402: angular momenta $l$ and $l'$ of the atomic pairs involved in the

403: interaction, $V_p$ reads explicitly

404:     \begin{equation}

405:     \label{pinteraction}

406:     V_{p}=-g\sum_{l,m}\sum_{l^\prime,m^\prime}

407:     (-1)^{l-m+l^\prime-m^\prime}c_{lm\su}^\dagger c_{l-m\sd}^\dagger

408:     c_{l^\prime m^\prime\sd}c_{l^\prime

409:     -m^\prime\su},

410:     \end{equation}

411: where the terms $(-1)^{l-m}$ and $(-1)^{l^\prime-m^\prime}$ come

412: from the Clebsch-Gordan coefficients (\ref{CB-coeff}) and hence

413: account for the angular part of the wave functions.

414: Strictly speaking, the coupling constant $g_{l,l'}$ should be

415: determined by the spatial integral

416: \begin{equation}

417: g_{l,l^\prime}=\frac{4\pi\hbar^2 |a|}{m_f}\int dr\,r^2\,

418: R_{n_F,l}^2(r)R_{n_F,l^\prime}^2(r). \label{g-spatial}

419: \end{equation}

420: However for our present purpose it is sufficient to estimate $g$ by

421: replacing the spatial integral in Eq. (\ref{g-spatial}) by $\bra

422: r_{n_F}^2 \ket^{-3/2}$, where for a pure harmonic oscillator state

423: in the $n_F$-shell the mean square radius $\bra r_{n_F}^2 \ket$ is

424: given by $(2n_F+3)a_{\rm ho}^2$ from the virial theorem. This

425: gives

426: \begin{equation}

427: g\sim \frac{4\pi\hbar^2 |a|}{m_f} (2n_F+3)^{-3/2}a_{\rm ho}^{-3}

428: \sim \frac{|a|}{a_{\rm ho}}n_F^{-3/2}\hbar\omega_{\rm ho}.

429: \end{equation}

430: 

431: In this model, the strongest pairing occurs for degenerate

432: energies, ${\cal E}_{l}={\cal E}_{l^\prime}$ for all $l,l^\prime$.

433: We will see that all fermions in the $n_F$-shell are then

434: coherently paired, the pairing energy being proportional to the

435: degeneracy factor $\Omega_{n_F}$. In the dilute gas limit, which

436: is equivalent to $g\Omega_{n_F}\sim g n_F^2 \ll \hbar\omega_{\rm

437: ho}$, the pairing takes then place on a single shell, see

438: Ref.~\cite{Heiselberg} for more details.

439: 

440: Turning finally to the photoassociation of atomic pairs into

441: molecules and the reverse process of photodissociation of

442: molecules back into atoms, we note that it is possible to neglect

443: all processes involving atoms other than those in the $n_F$-shell

444: provided that their characteristic frequency, the product $\chi

445: \sqrt{ \langle n_b \rangle}$ of the photoassociation coupling

446: constant, $\chi$, and the square root of the mean number of

447: molecules, remains small compared to the frequency separation

448: between neighboring shells of the trap. This condition is well

449: fulfilled for typical laser strengths and trap depths, in which

450: case $\chi \sqrt{ \langle n_b \rangle} \simeq 10^2 n_F^{-1/2}

451: \,{\rm s}^{-1} \ll \omega_{\rm ho} \simeq 10^3 \,{\rm s}^{-1}$.

452: In this case, the atom-molecule coupling can be approximated by the Hamiltonian

453: \begin{equation}

454: \label{aminteraction}

455: V_{am}=\sum_{l,m}(-1)^{l-m}\left[\chi

456: c_{lm\su}^\dagger c_{l-m\sd}^\dagger b +\chi^* b^\dagger

457: c_{l-m\sd} c_{lm\su}\right],

458: \end{equation}

459: where we only include the coupling between time-reversal atomic

460: pair and bosonic molecules in the ground state of the harmonic

461: trap. The photoassociation coupling constant $\chi$ is

462: proportional to the far off-resonant two-photon Rabi frequency

463: associated with two nearly co-propagating lasers with frequencies

464: $\omega_1$ and $\omega_2$~\cite{Wynar00},

465: \begin{eqnarray}

466: \label{coupling} \chi&=&\frac{\chi_0}{\sqrt{4\pi}} \int\, dr\,

467: r^2\, R^{(b)}_{0,0}(r) R_{n_F,l}^2(r)

468: \nonumber\\

469: &\sim& \frac{\chi_0}{n_F^{3/2} a_{\rm ho}^{3/2}},

470: \end{eqnarray}

471: where $R^{(b)}_{0,0}$ is a ground-state wave function of molecules

472: of mass of $2m_f$ with a spatial width about $a_{ho}$,

473: and in the second equality we

474: have again replaced the radial wave function of the atoms by 

475: $\bra r^2_{n_F} \ket^{-3/2}$.

476: From the coupling constant between $^{87}$Rb$_2$ molecules and $^{87}$Rb

477: atoms of Ref.~\cite{superchemistry} we estimate the

478: photoassociation coupling constant to be of the order of

479: $\chi_0=7.6 \times 10^{-7}$ m$^{3/2}$s$^{-1}$.

480: 

481: In terms of the pseudo-spin operators, and with Eqs.

482: (\ref{pinteraction}) and (\ref{aminteraction}), the total model

483: Hamiltonian Eq.~(\ref{modelH}) finally reads

484: \begin{eqnarray}

485: \label{spinH} H&=&(\hbar \delta+{\cal E}_0) b^\dagger b+\sum_l

486: 2{\cal E}_{n_F,l}(S_l^z+\Omega_l/2)\nonumber\\

487: &+&\chi\sum_l (S_l^+ b+b^\dagger

488: S_l^-)-g\sum_{l,l^\prime}S_l^+S_{l^\prime}^-.

489: \end{eqnarray}

490: This Hamiltonian clearly conserves the spin operators ${\bf

491: S}_l^2$. Applying this operator on the state $|\nu\rangle$ we have

492: \begin{eqnarray*}

493: {\bf S}_l^2|\nu\ket&=&S_l(S_l+1)|\nu\ket \\

494: &=& \{S_l^+S_l^-+S_l^z(S_l^z-1)\}|\nu\ket \\

495: &=& (\Omega_l/2-\nu_l/2)(\Omega_l/2-\nu_l/2+1)|\nu\ket,

496: \end{eqnarray*}

497: where we have used the identity ${\bf

498: S}_l^2=S_l^{+}S_l^{-}+S_l^{z}(S_l^{z}-1)$ and Eqs.

499: (\ref{nu-definition1}) and (\ref{nu-definition2}), so that

500:     $$

501:     S_l=\frac{1}{2}(\Omega_l-\nu_l).

502:     $$

503: This allows us to identify the operator

504:     \begin{equation}

505:     \hat{M}=\sum_l \left(S^z_l+S_l\right)+\hat{n}_b,

506:     \end{equation}

507: as the operator giving the total number of fermionic pairs and

508: molecules, or loosely speaking the ``pair number'' operator, which

509: is easily seen to also be a conserved quantity.

510: 

511: In the limit $g \to 0$ the Hamiltonian~(\ref{spinH}) reduces to

512: the Tavis-Cummings Model~\cite{Tavis,Bogoliubov} of quantum

513: optics, while for $\chi \to 0$, it becomes the Pairing Model

514: ~\cite{Richardson,Dukelsky}, for which Richardson first gave an

515: exact solution in the context of nuclear physics. We also mention

516: a recent family of exactly solvable models of atom-molecule

517: proposed in Ref.~\cite{AMHamiltonian}.

518: 

519: The discussion of the following sections concentrates specifically

520: in that situation where the mean-field energy shift in

521: single-particle energies is smaller than the photoassociation

522: coupling. In this degenerate model, the single-particle energies

523: of all atoms in the $n_F$-shell can then be taken to be equal,

524: ${\cal E}_{n_F,l}= {\cal E}_F$, and the Hamiltonian (\ref{spinH})

525: simplifies to

526:     \begin{equation}

527:     \label{DGmodel}

528:     H=-\hbar \omega\left(S^z+S\right)

529:     +\hbar\chi\left(S^{+}b+b^\dagger S^{-}\right)-\hbar gS^{+}S^{-},

530:     \end{equation}

531: where we have introduced the total spin operator

532:     \begin{equation}

533:     {\bf S}=\sum_l {\bf S}_l,

534:     \end{equation}

535: $\omega=\delta+({\cal E}_0-2{\cal E}_F)/\hbar$, $\Omega_{n_F}=\sum_l \Omega_l$, $\nu=\sum_l \nu_l$, so that the total spin is

536:     \begin{equation}

537:     S= \sum_l S_l=\frac{1}{2}\left (\Omega_{n_F} - \nu \right ).

538:     \label{total spin}

539:     \end{equation}

540: In Eq.~(\ref{DGmodel}), we have neglected constant terms

541: proportional to the conserved quantity $M$.

542: 

543: Assuming an oscillator length $a_{\rm ho} =3.2\times 10^{-6}$ m

544: for the mass of $^6$Li, $a=-114$ nm for its scattering length, and

545: $\omega_{\rm ho}=1000$ s$^{-1}$, then $\chi n_F^{3/2} \simeq 10^{2}$

546: s$^{-1}$.  The validity of the degenerate model, $|{\cal

547: E}_{n_F,n_F}- {\cal E}_{n_F,l=0}| < \chi n_F$ requires $n_F

548: \lesssim 10$, which corresponds to a total number of fermions 

549: $N_{n_F} \lesssim 10^3$.

550: 

551: \section{Ground State}

552: 

553: In this section, we discuss the dependence of the ground state of

554: the model Hamiltonian (\ref{DGmodel}) on the ratio of number of

555: paired fermions and molecules, $M=n_p+n_b,$ to the degeneracy

556: $\Omega$ of the Fermi level (we drop the subscript `$n_F$' for

557: notational clarity from now on). We identify qualitatively

558: different types of ground states, a pair-dominated ground state

559: and a molecular-dominated one, as a function of the parameter

560:     \begin{equation}

561:     \kappa \equiv \frac{\omega}{\chi \sqrt{\Omega}},

562:     \end{equation}

563: where $\omega = \delta +({\cal E}_0-2 {\cal E}_F)/\hbar$ is the

564: photoassociation frequency detuning.

565: 

566: \subsection{Pairing Model}

567: 

568: For a total number of particles $N=2M+\nu \leq \Omega$ , which

569: corresponds to the Fermi energy shell less than half-filled since

570: there are two hyperfine atomic states involved, the seniority

571: $\nu$ can take the values

572: \begin{equation}

573:   \nu = \cases{0,2,4,\dots,N & ({\em N} \, even) \cr

574:              1,3,5,\dots,N & ({\em N} \, odd), \cr}

575: \end{equation}

576: while for $\Omega < N \leq 2\Omega$, corresponding to a shell more

577: than half filled, the permissible values of $\nu$ are

578: \begin{equation}

579:   \nu = \cases{0,2,4,\dots,2\Omega-N & ({\em N} \, even) \cr

580:              1,3,5,\dots,2\Omega-N & ({\em N} \, odd). \cr}

581: \end{equation}

582: 

583: The degenerate Pairing Model of Refs.~\cite{Schuck,Walecka} is

584: obtained by neglecting the photoassociation coupling, $\chi=0$, in

585: the Hamiltonian~(\ref{DGmodel}). In this case, the total energy is

586: given as a function of seniority $\nu$ and the total number $N$ of

587: fermions by

588: \begin{equation}

589: E_{\rm pm}(N,\nu)=-\omega N -\frac{g}{4}(N-\nu)(2\Omega-\nu-N+2),

590: \end{equation}

591: and it is minimized for $\nu=0$ or $1$.

592: 

593: For an even particle number, the ground state corresponds

594: therefore to all pseudo-spins aligned, $S = \Omega/2$,

595:     \begin{equation}

596:     E_{\rm pm}(N,\nu=0)=-\omega N -gN(2\Omega -N + 2).

597:     \end{equation}

598: The first excited state corresponds to $\nu=2$ unpaired atoms, or

599: $S=\Omega/2-1$, and its energy is

600: \begin{equation}

601: \label{gapenergy} E_{\rm pm}(N,\nu=2)-E_{\rm pm}(N,\nu=0)=g\Omega.

602: \end{equation}

603: Thus, the energy needed to break up an atomic pair into two

604: un-paired fermions is independent of the number of fermions in the

605: $n_F$-shell. This energy is consistent to Bogoliubov

606: quasi-particle energy based on BCS variational state except for

607: corrections of relative order $1/\Omega$.

608: 

609: \subsection{Photoassociation}

610: 

611: In the presence of photoassociation, $\chi \neq 0$, the

612: eigenstates of the system consist of a coherent superposition of

613: atoms and molecules. For fixed $N$, we can classify them by the

614: total spin $S=(\Omega-\nu)/2$, each manifold consisting of $(M+1)$

615: eigenstates, where $M=(N-\nu)/2 = n_p + n_b$ is the total number

616: of pairs.

617: 

618: For positive detuning, $\omega>0$, the ground state is a pure

619: fermionic state, i.e., $N=2n_p+\nu$. On the other hand, for a

620: large negative detuning energy, $\omega<0$, and even particle

621: number, the ground state is reached when all particles are

622: molecules, corresponding to a zero seniority, or maximum spin

623: state.

624: 

625: Figure~\ref{gsfig1} shows the energies $E_\nu$ of a few seniority

626: states relative to that of the state $\nu=0$ as a function of

627: $\kappa = \omega/\chi\sqrt{\Omega}$. This example is for $\eta=

628: g\sqrt{\Omega}/\chi=1$ and a half-filled Fermi energy shell,

629: $N=\Omega=120$, corresponding to $n_F=14$, and results from a

630: direct numerical diagonalization of the Hamiltonian. It shows that

631: over the wide range of $\kappa$ considered here, the ground state

632: is always the maximum spin manifold $\nu=0$. For large negative

633: detunings, the energy differences $E_\nu-E_0$ are $\nu/2$ times

634: the single-particle energy difference between a molecule and two

635: fermions, and increase with $|\omega|$. For large positive

636: detunings, on the other hand, the energies of the excited states

637: approach the values $\nu/2$ times $g\Omega$, see

638: Eq.~(\ref{gapenergy}). In the crossover region where the nature of

639: the ground state changes from atomic to molecular, the ground

640: state is a coherent superposition of atoms and molecules that is

641: likewise the maximum spin state. We have verified that in the

642: absence of pairing interaction, $g=0$, the ground state is

643: likewise the state of maximum spin (not shown in figure).

644: 

645: \begin{figure}

646: %  \begin{minipage}{0.45\textwidth}

647:     \begin{center}

648:       \includegraphics[width=7.5cm,clip]{gsfig1.eps}

649:     \end{center}

650:     \caption{Excitation energies, in units of $\chi \sqrt{\Omega}$,

651:      relative to the ground-state energy

652:      for the seniority states

653:      $\nu=2$ (solid line), $\nu=4$ (dashed line), and $\nu=6$

654:      (dash-dotted line), as a function of the dimensionless

655:      parameter $\kappa=\omega/\chi\sqrt{\Omega}$

656:      for $\eta=g\sqrt{\Omega}/\chi=1$ and $N=\Omega=120$.}

657:     \label{gsfig1}

658: \end{figure}

659: %

660: 

661: In the following we concentrate on the state of maximum total

662: spin, $|S=\Omega/2\ket$, for an even number of fermions, $N=2M$.

663: The eigenstates of the atom-molecule system in the spin manifold

664: $S=\Omega/2$ have the general form

665: \begin{eqnarray}

666: |\Phi^\lambda_S\ket&=&\sum_{n_b=0}^{M} C_{\lambda}(n_p)

667: |S=\Omega/2,S^z=-S+n_p\ket_f \nonumber\\&&\otimes |n_b=M-n_p\ket_b,

668: \end{eqnarray}

669: where $n_p$ denotes a number of fermionic pairs and

670: $\lambda=0,1,\cdots M$ represent eigenmodes of the system with

671: eigenenergies $E_\lambda$. We have found these eigenstates and the

672: associated eigenenergies by direct diagonalization of the

673: Hamiltonian~(\ref{DGmodel}).

674: 

675: \begin{figure}

676:   \begin{minipage}{0.45\textwidth}

677:     \begin{center}

678:       \includegraphics[width=7.5cm,clip]{gsfig2.eps}

679:     \end{center}

680:     \caption{Normalized average number of molecules in the ground state

681:     as a function of $\kappa$ and $M$ for $\Omega=120$ and $g=0$.}

682:     \label{gsfig2}

683:   \end{minipage}

684: \hspace{4mm}

685: %

686:   \begin{minipage}{0.45\textwidth}

687:     \begin{center}

688:       \includegraphics[width=7.5cm,clip]{gsfig3.eps}

689:     \end{center}

690:     \caption{Normalized joint coherence function, $|G_{am}|$,

691:       as functions of $\kappa$ and $M$ for $\Omega=120$ and for $g=0$.}

692:     \label{gsfig3}

693:   \end{minipage}

694: \end{figure}

695: 

696: The average number of molecules in the ground state is shown in

697: Fig.~\ref{gsfig2} as a function of

698: $\kappa=\omega/\chi\sqrt{\Omega}$ and $M$ for $g=0$ and

699: $\Omega=120$. As expected, for large positive detuning $\omega>0$,

700: the ground state population consists almost exclusively of atoms,

701: while it is mostly made up of molecules for large negative

702: detunings. In the crossover region around $\omega=0$, the ground

703: state consists of a coherent superposition state between molecules

704: and atomic pairs.

705: 

706: It is possible to characterize this superposition in terms of the

707: normalized joint coherence function

708:     \begin{equation}

709:     G_{am} = \frac{2 \langle S^+ b\rangle}{M\sqrt{\Omega}},

710:     \end{equation}

711: which is shown in Fig.~\ref{gsfig3} as a function of $\kappa$ and

712: the number of pairs $M$. The joint coherence of the atomic and

713: molecular fields shows a remarkable enhancement in the crossover

714: region as well as a change in shape as a function of $M$, due to

715: the nonlinearity of the atom-molecule

716: coupling~(\ref{aminteraction}). This dependence on the filling

717: factor $M/\Omega$ can be understood more quantitatively by

718: considering the two limiting cases $M/\Omega\ll 1$ and $M/\Omega

719: \simeq 1$.

720: 

721: \subsubsection{$M/\Omega \ll 1$ --- mapping on a linear coupled-boson

722: system}

723: 

724: In the limit of small filling factors, it is convenient to

725: describe the system in terms of the Holstein-Primakoff mapping

726: ~\cite{HP40} of the $SU(2)$ generators

727: $S^+=(S^{-})^\dagger$ and $S^{z}$ in terms of bosonic operators.

728: According to this mapping, the Hilbert space of the group $SU(2)$

729: is carried by a subspace of the bosonic Fock space given by the

730: bosonic vacuum $|0\rangle_d$ and the bosonic operators $d$ and

731: $d^\dagger$, with

732:     \begin{equation}

733:     [d,d^\dagger]=1, \qquad d|0\ket_d=0,

734:     \end{equation}

735: The bosonic space being spanned by the $n_d-$boson states

736:     \[

737:     |n_d\ket_d=\frac{1}{\sqrt{n_d !}}(d^\dagger)^{n_d}|0\ket_d \qquad

738:     {\rm for}\,\, n_d=0,1,2,\dots,M.

739:     \]

740: Since we restrict our considerations to the subspace characterized

741: by the angular momentum quantum number $S=\Omega/2$, we can map

742: the operators $S^{\pm}$ and $S^{z}$ as

743: \begin{eqnarray}

744:     \label{HPmapping}

745:     S^{+} &\to& \sqrt{\Omega} d^\dagger

746:     \sqrt{1-\frac{d^\dagger d}{\Omega}},\nonumber \\

747:     S^{-} &\to& \sqrt{\Omega} \sqrt{1-\frac{d^\dagger d}{\Omega}} d,\nonumber \\

748:     S^z &\to& -\frac{\Omega}{2}+d^\dagger d.

749:     \end{eqnarray}

750: 

751: In the limit $M/\Omega\ll 1$, we only need a lowest-order of the

752: operators (\ref{HPmapping}), that is, replace the square roots by

753: 1. The Hamiltonian~(\ref{DGmodel}) reduces then to that of a

754: linearly coupled two-mode boson system,

755: \begin{equation}

756:     \label{linearboson}

757:     H\to H_{\rm linear}=-(\omega+g\Omega)d^\dagger

758:     d +\chi\sqrt{\Omega}(d^\dagger b +b^\dagger d),

759:     \end{equation}

760: with the constraint $M=n_d+n_b$.

761: 

762: This Hamiltonian can be diagonalized by the transformation

763:     \begin{eqnarray}

764:     c_{-}^\dagger&=&\cos{\theta}d^\dagger-\sin{\theta}b^\dagger\nonumber \\

765:     c_{+}^\dagger&=&\sin{\theta}d^\dagger+\cos{\theta}b^\dagger \nonumber,

766:     \end{eqnarray}

767: with $\cot{2\theta}=(\omega+g\Omega)/2\chi\sqrt{\Omega}$,

768: ,$0\le\theta\le 2\pi$, to give

769: \begin{equation}

770: \label{lineardiagonal} H_{\rm linear}=\epsilon_{-}c_-^\dagger

771: c_-+\epsilon_{+}c_+^\dagger c_+,

772: \end{equation}

773: with energies

774:     \begin{eqnarray}

775:     \epsilon_-&=&-(\omega+g\Omega)\cos^2{\theta}

776:     -2\chi\sqrt{\Omega}\cos{\theta}\sin{\theta}, \nonumber \\

777:     \epsilon_+&=&-(\omega+g\Omega)\sin^2{\theta}

778:     +2\chi\sqrt{\Omega}\cos{\theta}\sin{\theta}. \nonumber

779:     \end{eqnarray}

780: Since $\epsilon_-<\epsilon_+$ for $\chi > 0$, the ground state is

781: $(1/\sqrt{M !})(c_-^\dagger)^M|0\ket_d|0\ket_b$.

782: 

783: \subsubsection {$M/\Omega \simeq 1$ --- Mapping on a binary

784: atomic-molecular condensate}

785: 

786: Tikhonenkov and Vardi~\cite{Vardi} showed for in the homogeneous

787: case that when the total number of pairs is equal to the available

788: momentum states of the fermions, the system of fermionic atoms and

789: bosonic molecules can be mapped onto a two-mode atomic-molecular

790: BEC system~\cite{AMBEC1,AMBEC2,Zhou}. The corresponding situation

791: in our case occurs for $M=\Omega$, with an additional

792: atom-molecule two-body collision term required in addition.

793: 

794: To show how this works we introduce the two-mode Hamiltonian of a

795: two-component condensate of atoms and molecules,

796: \begin{equation}

797: \label{amhamiltonian}

798: H_{am}=-\omega b_m^\dagger b_m + \frac{\chi}{2}(b_m^\dagger b_a b_a+h.c.)

799: -\frac{g}{2}b^\dagger_m b_m b^\dagger_a b_a,

800: \end{equation}

801: where $b_a^\dagger$ and $b_m^\dagger$ are the bosonic creation

802: operators for the atomic and molecular modes, respectively.

803: Clearly the total number of particles $N_{am}=n_a+2n_m$ is

804: conserved, where $n_a$ and $n_m$ are the number of atoms and

805: molecules, respectively. For $N_{am}$ even, a general state of the

806: system can be expressed as

807: \begin{eqnarray}

808: |\Phi_{am}\ket&=&\sum_{m_a=0}^{M}C(m_a)|n_a=2m_a\ket\otimes

809: |n_m=M-m_a\ket \nonumber \\

810: &=& \sum_{m_a=0}^{M}C(m_a)|2m_a,n_m \ket,

811: \end{eqnarray}

812: with $M=N_{am}/2$. In this representation, the matrix form of the

813: Hamiltonian~(\ref{amhamiltonian}) is

814:      \begin{eqnarray*}

815:        \label{amMH}

816:        &&\bra 2m_a;n_m |H_{am}| 2m_a;n_m\ket\\

817:        &&\qquad =-\omega(M-m_a)-gm_a(M-m_a)\\

818:        &&\bra 2(m_a-1);n_m+1 |H_{am}| 2m_a;n_m\ket\\

819:        &&\qquad=\bra2m_a;n_m|H_{am}|2(m_a-1);n_m+1\ket \\

820:        &&\qquad=\chi \sqrt{m_a(m_a-1/2)(M-m_a)}.

821:      \end{eqnarray*}

822: Similarly, for our model the matrix form of the

823: Hamiltonian~(\ref{DGmodel}) in the ``pair number'' representation

824: is

825:     \begin{eqnarray*}

826:       \label{fbMH}

827:       &&\bra n_b;n_p|H|n_b;n_p\ket\\

828:       &&\qquad =-\omega(M-n_b)-g (n_b+\Delta M+1)(M-n_p)\\

829:       &&\bra n_b-1;n_p+1 |H| n_p;n_b\ket\\

830:       &&\qquad =\bra n_p;n_b |H|n_b-1;n_p+1\ket\\

831:       &&\qquad =\chi \sqrt{n_b(n_b+\Delta M+1)(M-n_b)}.

832:     \end{eqnarray*}

833: where

834:     \begin{equation}

835:     \Delta M=\Omega-M ,

836:     \end{equation}

837: with $\Delta M>0$. In the limit $M\simeq \Omega$, that is, when we

838: can neglect $\Delta M/M \ll 1$ and other terms of order of

839: $1/\Omega$, these two Hamiltonians are same under the

840: transformations $m_a\leftrightarrow n_b$ and $n_m \leftrightarrow

841: n_p$.

842: 

843: \subsubsection{Intermediate regime}

844: 

845: The behavior of the system apart from these two limiting cases

846: deviates from both models. The most striking difference between

847: these regimes appears in the energy, $\Delta$, of the first

848: excited state. Figure~\ref{gsfig4} shows

849: $\Delta/\chi\sqrt{\Omega}$ in the absence of pairing interaction

850: for $M/\Omega=1$ (solid line), $M/\Omega=0.5$ (dashed line), and

851: $M/\Omega=0.1$ (dash-dotted line) as a function of $\kappa =

852: \omega/\chi\sqrt{\Omega}$ for fixed $M=120$. In the case of

853: $M/\Omega=0.1$, $\Delta$ agrees well with the one-particle energy

854: difference $\epsilon_+-\epsilon_-=\sqrt{\omega^2+ 4\chi^2\Omega}$

855: of the linear coupled-boson model~(\ref{lineardiagonal}) (shown as

856: asterisks in the figure). Increasing $M/\Omega$ shifts the

857: location of the minimum of the energy gap and reduces the value of

858: its minimum. For $M=\Omega$, finally, the minimum gap approaches

859: zero at $\kappa\simeq 2$, consistently with a transition point in

860: the atom-molecule condensate system~\cite{AMBEC1,Zhou}. This

861: result is indicative of the appearance of a quantum phase

862: transition in the limit $M \to \infty$.

863: 

864: \begin{figure}

865:     \begin{center}

866:       \includegraphics[width=7.5cm,clip]{gsfig4.eps}

867:     \end{center}

868:     \caption{Lowest excitation energies as a function of

869:       $\kappa$ for $M/\Omega=1$ (solid line)

870:       and $M/\Omega=0.5$ (dashed line), and $M/\Omega=0.1$ (dash-dotted line),

871:       at fixed pair number $M=120$, where $g=0$. The asterisks correspond

872:       to a linear approximation for small $M/\Omega$ (see text).}

873:     \label{gsfig4}

874: \end{figure}

875: 

876: \subsection{The role of the pairing interaction}

877: 

878: We now examine in more detail the ground-state statistics of the

879: molecular field in the presence of pairing interaction $V_{p}$.

880: Since the cases of $M\ll \Omega$ and $M\simeq \Omega$ can be

881: mapped onto relatively well-known systems, we present results for

882: the situation of a half-filled shell, $M/\Omega=0.5$, only. Figures

883: \ref{gsfig5} and \ref{gsfig6} show the probability $P(n_b)$ of

884: having $n_b$ molecules in the trap, the molecule statistics, for

885: $2M=\Omega=120$ as a function of the dimensionless parameter

886: $\kappa$, with $g=0$ and $\eta=g\sqrt{\Omega}/\chi=10$,

887: respectively. There are several quantitative differences between

888: the two cases: the transition from an atomic to a molecular ground

889: state is shifted by the pairing interaction, and the width of the

890: crossover region in detuning space is significantly broader.

891: \begin{figure}

892:   \begin{minipage}{0.45\textwidth}

893:     \begin{center}

894:       \includegraphics[width=7.5cm,clip]{gsfig5.eps}

895:     \end{center}

896:     \caption{Molecule statistics $P(n_b)$ as a function of

897:       the dimensionless parameter $\kappa$ in the absence of

898:       pairing interaction, $g=0$, for $2M=\Omega=120$.}

899:     \label{gsfig5}

900:   \end{minipage}

901: \hspace{4mm}

902: %

903:   \begin{minipage}{0.45\textwidth}

904:     \begin{center}

905:       \includegraphics[width=7.5cm,clip]{gsfig6.eps}

906:     \end{center}

907:     \caption{Molecule statistics $P(n_b)$ as a function of

908:     the dimensionless parameter $\kappa$ in the presence of

909:       pairing interaction, $\eta=g\sqrt{\Omega}/\chi=10$, for $2M=\Omega=120$.}

910:     \label{gsfig6}

911:   \end{minipage}

912: \end{figure}

913: 

914: This behavior can be understood by noting that the pairing

915: interaction gives rise to an additional detuning effect depending

916: on the number of molecules, as shown by the diagonal terms of the

917: Hamiltonian~(\ref{fbMH}) with $M=\Delta M=\Omega/2$. This

918: interaction leads to number-dependent energy shifts and dephasing

919: between different molecular number states. A similar effect has

920: been studied in the context of a Jaynes-Cummings-like description

921: of photoassociation and has been referred as ``nonlinear detuning''

922: ~\cite{search}.

923: 

924: We can estimate the position of the ``resonance'' point

925: $\omega_{\rm res}(g)$ where the ground state goes from being

926: molecular to atomic in nature by taking $\bra

927: n_b=M/2+1;n_p=M/2-1|H|n_b=M/2+1;n_p=M/2-1\ket =\bra

928: n_b=M/2;n_p=M/2|H|n_b=M/2;n_p=M/2\ket$. This gives

929: \[

930: \omega_{\rm res}(g)=-g\frac{\Omega}{2},

931: \]

932: where we have neglected a term of order $1/\Omega$.

933: 

934: The shift in $\omega_{\rm res}(g)$  due to the pairing interaction

935: is further illustrated in Fig. \ref{gsfig7}, which shows the

936: magnitude of the joint coherence function, $|G_{am}|$, as a

937: function of $\kappa=\omega/\chi\sqrt{\Omega}$ for three values of

938: the pairing coefficient $g$. Finally, Fig. \ref{gsfig8} shows the

939: entanglement entropy $E(\rho_b)$ of the ground state, obtained

940: from the von Neumann entropy of the molecular reduced density

941: operator~\cite{Hines}

942: \[

943: \rho_b=Tr_{f}(\rho),

944: \]

945: as

946: \begin{equation}

947: E(\rho_b)=-\sum_{n_b=0}^{M} |C(n_b)|^2 \log{|C(n_b)|^2},

948: \end{equation}

949: where the logarithm is taken in base $2$, for three values of the

950: pairing interaction strengths, $\eta=0, 5, 10$. The entanglement

951: in Figure~ \ref{gsfig8} is divided by a maximum entanglement,

952: $\log{M}$. Consistently with the results of of Figs. \ref{gsfig5}

953: and \ref{gsfig6} for the molecular statistics, the pairing

954: interaction reduces the entanglement.

955: %

956: \begin{figure}

957:   \begin{minipage}{0.45\textwidth}

958:     \begin{center}

959:       \includegraphics[width=7.5cm,clip]{gsfig7.eps}

960:     \end{center}

961:     \caption{Normalized joint coherence, $|G_{am}|$, of the ground state

962:     as a function of $\kappa$

963:     for three values of the pairing interaction strengths

964:     $\eta=g\sqrt{\Omega}/\chi$,

965:     indicated in the insert.}

966:     \label{gsfig7}

967:   \end{minipage}

968: \hspace{4mm}

969: %

970:   \begin{minipage}{0.45\textwidth}

971:     \begin{center}

972:       \includegraphics[width=7.5cm,clip]{gsfig8.eps}

973:     \end{center}

974:     \caption{Entanglement entropy, $E(\rho_b)$, of the ground

975:     state, in units of the maximum entanglement $\log M$,

976:      as a function of $\kappa$ for three values of the pairing

977:      interaction strengths $\eta=g\sqrt{\Omega}/\chi$ indicated in the insert.}

978:     \label{gsfig8}

979:   \end{minipage}

980: \end{figure}

981: 

982: 

983: \section{Dynamics}

984: 

985: In the absence of  pairing interaction our model is equivalent to

986: the Tavis-Cummings model, whose dynamics has been studied in

987: detail in the context of coherent spontaneous emission from a

988: system of $N$ two-level atoms interacting with a quantized

989: radiation field~\cite{Dicke,Bonifacio}. The main purpose of this

990: section is to extend this work to the study of the system dynamics

991: in the presence of $V_p$. One important result is that the

992: nonlinearity of this interaction produces a self-trapping

993: transition.

994: 

995: We assume that the system consists initially either of atomic

996: pairs or of molecules only, corresponding to a maximal spin state.

997: A photoassociation beam is applied from $t=0$ on, and we examine

998: the subsequent coherent dynamics of the system. Note that although

999: this problem resembles the dynamics of a bosonic Josephson

1000: Junction ~\cite{BJJ1,BJJ2} in an asymmetric trap for the initial

1001: imbalance in populations, the nonlinear coupling term in our model

1002: leads to considerably richer dynamics. Since the total spin is a

1003: constant of motion of the Hamiltonian~(\ref{DGmodel}), we confine

1004: our discussion to the maximal spin state $|S=\Omega/2\ket$.

1005: 

1006: \subsection{Semiclassical approximation}

1007: 

1008: Before proceeding with a full quantum analysis, we first consider

1009: the semiclassical approximation. Our approach is very similar to

1010: that taken in Ref.~\cite{AMBEC2}. Introducing the two operators

1011: \begin{eqnarray}

1012: J_+&\equiv& S^+b,\nonumber \\

1013:  J_-&\equiv&  b^\dagger S^-,

1014: \end{eqnarray}

1015: results in the Heisenberg equations of motion

1016: \begin{eqnarray}

1017: \label{heisenbergeq}

1018: i\frac{d}{dt}J_{+}&=&\omega J_{+}

1019:  +2\chi b^\dagger b S^z+\chi S^+ S^- -2gJ_{+}S^z , \\

1020: i\frac{d}{dt}J_{-}&=&-\omega J_{-}

1021: -2\chi b^\dagger b S^z

1022: -\chi S^+ S^- +2g S^zJ_-, \\

1023: i\frac{d}{dt}S^z&=&\chi (J_+ - J_-).

1024: \end{eqnarray}

1025: Here we note the following relations between operators and conserved

1026: quantities,

1027: \[

1028: S^+ S^- = S(S+1)-S^z(S^z-1), \quad b^\dagger b = M-S-S^z.

1029: \]

1030: 

1031: As usual we introduce a semiclassical approximation by factorizing

1032: the mean values of the various operators that appear in the

1033: Heisenberg equations of motion, such as $\bra S_z J_+\ket=\bra S^z

1034: \ket\bra J_+ \ket$ and $\bra J_- S_z\ket=\bra J_-\ket\bra S_z

1035: \ket$, etc. Introducing the $c-$number functions $s_z\equiv\bra

1036: S_z\ket$, $j_x\equiv \bra (J_{+} + J_{-})\ket/2$, and $j_y\equiv

1037: \bra (J_{+} - J_{-})\ket/2i$, and neglecting corrections of order

1038: $1/\Omega$, that is, setting $\bra S^{+}S^{-}\ket=S^2-s_z^2$,

1039: we obtain the semiclassical equations of motion

1040: \begin{eqnarray}

1041: \label{cdynamics}

1042: \frac{d}{dt}j_x&=&\omega j_y-2gj_ys_z\\

1043: \frac{d}{dt}j_y&=&-\omega j_x+2gj_xs_z-\chi h(s_z)\\

1044: \frac{d}{dt}s_z&=&2\chi j_y,

1045: \end{eqnarray}

1046: where $h(s_z)=-3s_z^2+2(M-S)s_z+S^2$. Noting the additional

1047: conserved quantity

1048: \[

1049: \frac{d}{dt}\left(j_x-\frac{\omega}{2\chi}s_z+\frac{g}{2\chi}s_z^2\right)=0,

1050: \]

1051: we find that the coupled equations~(\ref{cdynamics}) are

1052: equivalent to the classical Newtonian equation of motion

1053: \begin{equation}

1054: \label{Newtoneq} \frac{d^2}{dt^2}s_z=-\frac{d}{ds_z}U(s_z),

1055: \end{equation}

1056: where the potential $U(s_z)$ is determined by the initial conditions

1057: for $j_x(0)$ and $s_0\equiv s_z(0)$. It is sufficient for our

1058: purpose to assume an initial Fock state, so that $j_x(0)=0$ and

1059: $U(s_z)$ is given by

1060: \begin{eqnarray}

1061: \label{scdpotential}

1062: U(s_z)&=&\frac{g^2}{2}s_z^4-(g\omega+2\chi^2)s_z^3\nonumber\\

1063: &+&\left[\frac{\omega^2}{2}+g\omega

1064: s_0-g^2(s_0)^2+2\chi^2(M-S)\right]s_z^2 \nonumber \\

1065: &+&\left[2\chi^2 S^2-\omega^2s_0+\omega gs_0^2 \right]s_z.

1066: \end{eqnarray}

1067: Since the potential has a quartic form of $s_z$,

1068: Eq.~(\ref{Newtoneq}) can be solved analytically in terms of

1069: Jacobian elliptic functions. The derivation of the general

1070: solutions, which is straightforward but lengthy, is given in

1071: Appendix~A.

1072: 

1073: \subsection{Coherent dynamics, $g=0$}

1074: 

1075: We first examine a typical behavior of the semiclassical dynamics

1076: in the absence of pairing interaction, $g=0$, for $M/\Omega\ll 1$,

1077: $M/\Omega \sim 0.5$, and $M/\Omega\simeq 1$.

1078: 

1079: From the semiclassical solutions~(\ref{apg0eq3})

1080: in the case of $g=0$ and on the exact resonance $\omega=0$,

1081: the population imbalance between fermionic pairs and molecules,

1082: $\bra \hat{n}_p\ket - \bra \hat{n}_b \ket=2S-M+2s_z$, and the coherence

1083: function $j_y$ is given by

1084: \begin{eqnarray}

1085: \label{scdresonance1}

1086: \frac{\bra n_p\ket-\bra n_b\ket}{M}&=&2\,{\rm sn}^2(\chi\sqrt{\Omega}t+\phi;k)-1,\\

1087: \label{scdresonance2}

1088: \frac{2j_y}{M\sqrt{\Omega}}&=&2\,{\rm sn}(\chi\sqrt{\Omega}t+\phi;k)

1089: \times{\rm cn}(\chi\sqrt{\Omega}t+\phi;k)\nonumber\\

1090: &\times&{\rm dn}(\chi\sqrt{\Omega}t+\phi;k),

1091: \end{eqnarray}

1092: where {\rm sn}, {\rm cn}, and {\rm dn} are Jacobian elliptic

1093: functions~\cite{elliptic}. Here, $k=\sqrt{M/\Omega}$ is the

1094: elliptic modulus and $\phi$ is a phase factor determined by the

1095: initial conditions. It is equal to $\phi=K$, corresponding to the

1096: complete elliptic integral of the first kind, for an initial

1097: fermionic Fock state $|n_p=M\ket$, and to $\phi=0$ for an initial

1098: the entire population being in the molecular mode. We note that

1099: $j_x(t)$ is zero for all time on the exact resonance.

1100: 

1101: \subsubsection{$M/\Omega \ll 1$ --- Linear coupled-bosons regime}

1102: 

1103: In this case, the elliptic modulus $k\ll 1$ (see Appendix~A) so that

1104: the elliptic functions in Eqs.~(\ref{scdresonance1}) and (\ref{scdresonance2})

1105: can be approximated by ${\rm sn}(u,k)\to \sin{u}$, ${\rm cn}(u,k)\to

1106: \cos{u}$, and ${\rm dn}(u,k)\to 1$, respectively. The imbalance

1107: in atomic and molecular populations which undergoes Rabi

1108: oscillations at the frequency $2\chi\sqrt{\Omega}$ is given by

1109:     \[

1110:         (\bra\hat{n}_p\ket-\bra\hat{n}_b\ket)/M=\cos{(2\chi\sqrt{\Omega}t)}

1111:     \]

1112: for an initial fermionic pair state,

1113:     \[

1114:     (\bra\hat{n}_p\ket-\bra\hat{n}_b\ket)/M=-\cos(2\chi \sqrt{\Omega}t)

1115:     \]

1116: for an initial molecular state. These results are equivalent to

1117: those obtained directly from the Hamiltonian~(\ref{linearboson}).

1118: We have compared these solutions with the exact quantum mechanical

1119: dynamics obtained numerically, and checked that the linear

1120: approximation agrees with the quantum results for

1121: $M/\Omega\lesssim 0.2$ and for times shorter than $t\sim

1122: \pi/\chi\sqrt{\Omega}$.

1123: 

1124: \subsubsection{$M/\Omega \sim 0.5$ --- Intermediate regime}

1125: 

1126: Figure~\ref{dyfig1}a shows the normalized population difference

1127: $(\bra\hat{n}_p\ket-\bra\hat{n}_b\ket)/M$, and Fig.~\ref{dyfig1}b

1128: shows the normalized coherence function $2j_y/M\sqrt{\Omega/2}$,

1129: as a function of the dimensionless time $\tau=\chi\sqrt{\Omega}t$

1130: for a system initially either in a pure atomic state or a pure

1131: molecular state and for $M/\Omega=0.5$. The circles correspond to

1132: the semiclassical description, while the lines are the results of a

1133: full quantum-mechanical analysis. The anharmonicity due to

1134: the nonlinear atom-molecule coupling is clearly apparent, and also

1135: shows that the semiclassical dynamics approximate the quantum

1136: dynamics very well.

1137: 

1138: %

1139: \begin{figure}

1140:     \begin{center}

1141:       \includegraphics[width=7.5cm,clip]{dyfig1.eps}

1142:     \end{center}

1143:     \caption{Comparison of the semiclassical and quantum dynamics

1144:         for $2M=\Omega=120$, $\omega=0$, and $g=0$.

1145:         Figure (a) shows the population difference between fermionic pairs and

1146:         molecules $(\bra\hat{n}_p\ket-\bra\hat{n}_b\ket)/M$. Figure (b) plots

1147:         the coherence function $2j_y/M\sqrt{\Omega}$

1148:         as a function of the dimensionless time $\tau=\chi\sqrt{\Omega}t$.

1149:         The solid and dashed lines give the quantum results for an

1150:         initial fermionic pair sate and a pure molecular state, respectively.

1151:         The corresponding semiclassical results are indicated by filled

1152:         and open circles, respectively.}

1153:     \label{dyfig1}

1154: \end{figure}

1155: 

1156: We note that the atomic pair state in the half-filled shell

1157: corresponds to a Dicke superradiant state~\cite{Dicke}, which is

1158: known from quantum optics to give rise to the strongest collective

1159: enhancement of transition probabilities. This enhancement is

1160: proportional to the product of the number of particle pairs $M$

1161: and the number of hole pairs $\Omega-M$, and is maximum for

1162: $M=\Omega/2$. From Eq.~(\ref{scdresonance1}), we have that for

1163: sufficient short times $\tau=t/\chi\sqrt{\Omega}\ll 1$ the average

1164: number of molecules builds up as

1165: \begin{eqnarray*}

1166: \bra n_b \ket&=&{\rm cn}^2(\chi\sqrt{\Omega}t+\phi)\\

1167: &\rightarrow_{\tau\ll 1}& M\left(\Omega-M\right)(\chi t)^2

1168: =\frac{\chi^2}{4}\Omega^2t^2.

1169: \end{eqnarray*}

1170: %

1171: 

1172: \subsubsection{$M/\Omega \simeq 1$ --- Binary atomic-molecular BEC}

1173: 

1174: As shown in Ref.~\cite{Vardi}, the coherent dynamics in this

1175: regime is qualitatively very similar to that of binary condensate

1176: of atoms and molecules. For $M =\Omega$, corresponding to $k=1$,

1177: the population imbalance between atomic pairs and molecules is

1178: given in the semiclassical approximation by

1179: \begin{equation}

1180: \frac{\bra n_p\ket-\bra

1181: n_b\ket}{M}=2\,\tanh{^2(\chi\sqrt{\Omega}t+\phi)-1},

1182: \end{equation}

1183: indicating that the point $\bra n_p\ket=M$ is stationary. However,

1184: the system is dynamically unstable against small fluctuations ,

1185: see Ref.~\cite{AMBEC2} for a detailed discussion in the context of

1186: binary condensates of atoms and molecules and Ref.~\cite{Walls} in

1187: the context of second-harmonic generation.

1188: 

1189: %

1190: \subsection{\label{selftrap}Self-trapping transition and quantum dynamics}

1191: 

1192: To conclude, we discuss the coherent dynamics of the system in the

1193: presence of pairing interaction, $g\neq 0$, considering only

1194: the case of half-filling for simplicity. We consider specifically

1195: the example $2M=\Omega=120$, which corresponds to $n_F=14$, and

1196: take an initial state as the Fock state $|n_p=M;n_b=0\ket$.

1197: Figure~\ref{appfig1} of the appendix shows the phase diagram of

1198: the semiclassical dynamics in the $\kappa-\eta$ space,

1199: illustrating that a self-trapping transition~\cite{BJJ1,Scott}

1200: takes place when varying the dimensionless detuning frequency

1201: $\kappa=\omega/\chi\sqrt{\Omega}$, provided that the pairing

1202: interaction strength $\eta=g\sqrt{\Omega}/\chi$ exceeds a critical

1203: strength $\eta_c$. For the case of half-filling case, we find

1204: $\eta_c=5.0302$.

1205: 

1206: %

1207: \begin{figure}

1208:     \begin{center}

1209:       \includegraphics[height=4.5cm,width=7.5cm,clip]{dyfig2.eps}

1210:     \end{center}

1211:     \caption{Schematic potential curves, $U(s_z)-U(s_0)$, around

1212:      a self-trapping transition point $\kappa_0$. These curves are

1213:      at (a) above $\kappa_0$, (b) at transition point

1214:      $\kappa=\kappa_0$, and (c) below $\kappa_0$. The filled circle

1215:      indicates an initial position of classical particle.}

1216:     \label{dyfig2}

1217: \end{figure}

1218: %\begin{widetext}

1219: \begin{figure*}

1220:     \begin{center}

1221:       \includegraphics[height=12cm,width=18cm,clip]{dyfig3.eps}

1222:     \end{center}

1223:     \caption{Population imbalance versus the rescaled time

1224:       $\tau=\chi\sqrt{\Omega}t$ for $\eta=6$. Quantum (solid line) and

1225:       semiclassical

1226:       (dashed line) solutions are shown, respectively, for the detuning

1227:       parameters (a) $\kappa=0$ and (b) $\kappa=-3.1$, and

1228:       (c) $\kappa=-3.3$, and (d) $\kappa=-4.0$.

1229:       The dotted line in (c) corresponds to the semiclassical solution

1230:       at the transition point $\kappa_0=-3.2307$.}

1231:     \label{dyfig3}

1232: \end{figure*}

1233: %

1234: \begin{figure*}

1235:   \begin{minipage}{0.45\textwidth}

1236:     \begin{center}

1237:       \includegraphics[height=5cm,width=7.5cm,clip]{dyfig4.eps}

1238:     \end{center}

1239:     \caption{Time-averaged population imbalance as a function of

1240:       $\kappa$ for $\eta=2.0$.

1241:     Quantum result (cross) and semiclassical result (open circle).}

1242:     \label{dyfig4}

1243:   \end{minipage}

1244:   \hspace{4mm}

1245: %

1246:   \begin{minipage}{0.45\textwidth}

1247:     \begin{center}

1248:       \includegraphics[height=5cm,width=7.5cm,clip]{dyfig5.eps}

1249:     \end{center}

1250:     \caption{Time-averaged population imbalance as a function of

1251:       $\kappa$ for $\eta=6.0$.

1252:     Quantum result (cross) and semiclassical result (open circle).}

1253:     \label{dyfig5}

1254:   \end{minipage}

1255: \end{figure*}

1256: 

1257: This transition can be interpreted physically from the motion of

1258: ``classical particle''  in the ``potential'' (\ref{scdpotential}).

1259: Figure~\ref{dyfig2} displays schematic potential curves,

1260: $U(s_z)-U(s_0)$ as a function of $s_z$ in the vicinity of a

1261: self-trapping transition point $\kappa_0$. For our specific

1262: initial conditions the particle ``velocity'' is initially zero,

1263: $ds_z(0)/dt\propto j_y(0)=0$. For $\kappa > \kappa_0$,

1264: Fig.~\ref{dyfig2}-(a), the classical particle oscillates

1265: periodically in the potential. As $\kappa$ approaches $\kappa_0$,

1266: one additional potential barrier appears, and at the transition

1267: point its height equals the initial potential energy of the

1268: particle (see Fig.~\ref{dyfig2}-(b)). At that point, the particle

1269: rests on the potential maximum after reaching it.  Below the

1270: critical point, the barrier confines the particle in a narrow

1271: range, as shown in Fig.~\ref{dyfig2}-(c). The self-trapping effect

1272: provides a sudden suppression of the amplitude of coherent

1273: oscillations. Since the key factor in achieving this transition is

1274: the quartic term in the potential~(\ref{scdpotential}), it

1275: disappears in the absence of pairing interaction.

1276: 

1277: Fig.~\ref{dyfig3} shows the time evolution of the population

1278: difference $(\bra \hat{n}_{p}\ket-\bra \hat{n}_b\ket)/M$ for

1279: $\eta=6$ and for several detuning energies. The semiclassical

1280: solutions (dashed lines) clearly show the self-trapping as a

1281: sudden suppression of coherent oscillations from just above to

1282: just below the transition detuning. The dotted line in

1283: Fig.~\ref{dyfig3}-(c) corresponds to the semiclassical solution

1284: for the threshold detuning $\kappa_0=-3.2307$. The solid lines

1285: show the exact quantum solutions. Apart from the transition point,

1286: the quantum and semiclassical dynamics are similar at least for

1287: short enough times. However, the oscillations of the quantum

1288: solution deviate from those of the semiclassical solution near the

1289: transition point, Fig.~\ref{dyfig3}(b-c). Since in the semiclassical

1290: picture the height of the potential barrier near the transition

1291: point is just below or above the initial potential energy of a

1292: particle, the quantum motion of the particle is very sensitive to

1293: fluctuations and hence deviates significantly from its classical

1294: counterpart. The initial Fock state provides fluctuations of

1295: coherence, and large quantum fluctuations of the population

1296: imbalance arise as a result. We have verified numerically that the

1297: number fluctuations near the transition point are enhanced by an

1298: order of magnitudes as compared to those far away from that point.

1299: 

1300: Figs.~\ref{dyfig4} and \ref{dyfig5} compare the quantum and

1301: semiclassical time-averaged population imbalance as a function of

1302: $\kappa$ for $\eta=2.0$, and for $\eta=6.0$, respectively. In contrast to the

1303: second case, there is no self-trapping transition in the first

1304: case. Hence the semiclassical time-averaged population imbalance

1305: (cross) is a smooth function of $\kappa$, and agrees well

1306: with the quantum results. For the strong pairing coupling of

1307: Fig.~\ref{dyfig5}, in contrast, an abrupt jump of the

1308: semiclassical time-averaged value occurs when varying $\kappa$, a

1309: signature of the self-trapping transition. Due to the large

1310: quantum fluctuations, it differs markedly from the time-averaged

1311: quantum result near a transition point.

1312: 

1313: %

1314: %\begin{figure}

1315: %  \begin{minipage}{0.45\textwidth}

1316: %    \begin{center}

1317: %      \includegraphics[width=7.5cm,clip]{dyfig2a.eps}

1318: %    \end{center}

1319: %    \caption{Dynamics for $g/\chi=1$ and for $\omega/\chi\Omega=0$.}

1320: %    \label{dyfig2a}

1321: %  \end{minipage}

1322: %  \hspace{4mm}

1323: %

1324: %  \begin{minipage}{0.45\textwidth}

1325: %    \begin{center}

1326: %      \includegraphics[width=7.5cm,clip]{dyfig2b.eps}

1327: %    \end{center}

1328: %    \caption{Dynamics for $g/\chi=1$ and for $\omega/\chi\Omega=-0.4$}

1329: %    \label{dyfig2b}

1330: %  \end{minipage}

1331: %\end{figure}

1332: %

1333: %\begin{figure}

1334: %  \begin{minipage}{0.45\textwidth}

1335: %    \begin{center}

1336: %      \includegraphics[width=7.5cm,clip]{dyfig2c.eps}

1337: %    \end{center}

1338: %    \caption{Dynamics for $g/\chi=1$ and for $\omega/\chi\Omega=-0.47$.}

1339: %    \label{dyfig2c}

1340: %  \end{minipage}

1341: %  \hspace{4mm}

1342: %

1343: %  \begin{minipage}{0.45\textwidth}

1344: %    \begin{center}

1345: %      \includegraphics[width=7.5cm,clip]{dyfig2d.eps}

1346: %    \end{center}

1347: %    \caption{Dynamics for $g/\chi=1$ and for $\omega/\chi\Omega=-0.6$}

1348: %    \label{dyfig2d}

1349: %  \end{minipage}

1350: %\end{figure}

1351: 

1352: \section{Summary}

1353: 

1354: In this paper, we have considered the coherent photoassociation of

1355: fermionic atoms into bosonic molecules trapped in a spherically

1356: symmetric harmonic trap. We showed that under a realistic set of

1357: conditions this system can be mapped onto a Tavis-Cummings

1358: Hamiltonian with an additional paring interaction using

1359: pseudo-spin operators. We carried out an exact numerical

1360: diagonalization of the Hamiltonian to determine the ground state

1361: of the system, investigating the crossover from a predominantly

1362: atomic to a predominantly molecular state. We also investigated

1363: the joint coherence and the quantum entanglement between the

1364: atomic and molecular fields, and found that the atomic pairing

1365: interaction suppresses the entanglement between fermions and

1366: bosons. We then analyzed the coherent dynamics of photoassociation

1367: due to the nonlinear atom-molecule coupling. Using a semiclassical

1368: factorization ansatz, we showed the appearance of a self-trapping

1369: transition in the presence of pairing interaction. An exact

1370: quantum solution illustrated the important role of quantum

1371: fluctuations in the neighborhood of that transition point. Future

1372: work will extend this study to a detailed analysis of the

1373: non-degenerate model and to multi-well superradiant systems. For

1374: instance, preparing an atomic Fermi gas in a Josephson-type

1375: configuration and applying a photoassociation beam should lead to

1376: the efficient production of spatially correlated molecules.

1377: 

1378: \acknowledgements This work is supported in part by the US Office

1379: of Naval Research, the National science Foundation, the US Army

1380: Research Office, NASA, and the Joint Services Optics Program.

1381: 

1382: \appendix

1383: \section{Analytic solutions in terms of Jacobian Elliptic

1384: functions}

1385: 

1386: In this appendix, we obtain analytic solutions of semiclassical

1387: dynamics that obey the Newtonian equation of motion

1388: Eq.~(\ref{Newtoneq}), and show the phase diagram of

1389: the semiclassical dynamics based on those solutions.

1390: 

1391: The solution of Eq.~(\ref{Newtoneq}) has the general form

1392: \begin{equation}

1393: \label{integral}

1394: t=\int^{s_z}_{s_0}\frac{ds}{\sqrt{2[U(s_0)-U(s_z)}]},

1395: \end{equation}

1396: with the potential $U(s_z)$ given by Eq.~(\ref{scdpotential}).

1397: 

1398: We analyze the solution for the two cases $g=0$ and $g \neq 0$,

1399: separately, and for simplicity we take the initial state as an

1400: atomic state in the maximum spin manifold $S=\Omega/2$. The

1401: extension to other initial states is straightforward.

1402: 

1403: \subsection{Case $g=0$}

1404: In this case, the ``potential'' $U(s_z)$ is a cubic function of

1405: $s_z$,

1406: \begin{equation} U(s_z)=-2\chi^2

1407: s_z^3+\left[\frac{\omega^2}{2}+2\chi^2 s_0\right]s_z^2

1408: +\left[2\chi^2S^2-\omega^2s_0\right]s_z.

1409: \end{equation}

1410: Here we have used the initial condition $s_0=-S+M$. By introducing

1411: the normalized quantities

1412:    \[

1413:       \bs_z=s_z/2S\quad (-1/2\leq \bs_z\leq 1/2), \qquad

1414:       \bs_0=s_0/2S,

1415:    \]

1416: we obtain the explicit form of the denominator of the right-hand

1417: side of Eq. (\ref{integral}),

1418: \begin{eqnarray}

1419: \varphi(s_z)&\equiv&2[U(s_0)-U(s_z)]\nonumber\\

1420: &=&4\chi^2(2S)^3(\bs_z-\bs_0)(\bs_z-\bs_+)(\bs_z-\bs_-)\\

1421: \bs_{\pm}&=&\frac{\kappa^2}{8}\pm\frac{1}{8}

1422: \sqrt{\kappa^4-16\bs_0\kappa^2+16},

1423: \end{eqnarray}

1424: where $\kappa=\omega/\chi\sqrt{2S}$. We note that the variables

1425: $\bs_\pm$ are always real-valued for any magnitudes of $\kappa^2$,

1426: because $-1/2\le \bs_0 \le 1/2$, and then $\bs_+\ge \bs_0\ge

1427: \bs_z(t) \ge \bs_-$. The solution can be obtained by integrating

1428: the form

1429: \begin{widetext}

1430: \begin{equation}

1431: \label{apg0eq0}

1432: t= \int^{s_z}_{s_0}\frac{ds}{\sqrt{\varphi(s_z)}}

1433: =\frac{1}{\chi\sqrt{2S(\bs_{+} -\bs_{-})}}

1434: \left\{\int^{\phi}_{0}-\int^{\pi/2}_{0} \right\}

1435: \frac{d\theta}{\sqrt{1-k^2 \sin^2{\theta}}},

1436: \end{equation}

1437: \end{widetext}

1438: where,

1439: \[

1440:   \phi(t)=\arcsin{\sqrt{\frac{(\bs_+-\bs_-)(\bs_0-\bs_z)}{(\bs_0-\bs_-)(\bs_+-\bs_z)}}}.

1441: \]

1442: The integral that appears in that equation is an elliptic integral

1443: of the first kind. Noting that the integration within $0\le

1444: \theta\le \pi/2$ gives rise to a complete elliptic integral of the

1445: first kind, $K$, we find

1446: \begin{equation}

1447: \chi\sqrt{2S(\bs_+ -\bs_-)} t+K={\rm sn}^{-1}\left(\sin{\phi(t)};k\right)

1448: \end{equation}

1449: where the function ${\rm sn}^{-1}$ is the inverse of the Jacobian

1450: elliptic function and $k=\sqrt{(\bs_0-\bs_-)/(\bs_+-\bs_-)}$

1451: denotes the elliptic modulus. This gives the evolution of

1452: $\bs_z(t)$

1453: \begin{equation}

1454:     \label{apg0eq1}

1455:     \bs_z(t)=\bs_- +

1456:     (\bs_0-\bs_-){\rm sn}^2\left(\chi\sqrt{2S(\bs_+ - \bs_-)}t+K;k

1457:     \right).

1458: \end{equation}

1459: At the exact resonance, $\omega=0$, we have that  $s_+=1/2$,

1460: $s_-=-1/2$, and $k=\sqrt{M/2S}$ so that this expression reduces to

1461: \begin{equation}

1462: \label{apg0eq3} s_z(t)=-S+M{\rm sn}^2(\chi\sqrt{2S}t+K;k).

1463: \end{equation}

1464: In terms of $s_z(t)$ the coherence functions $j_x(t)$ and $j_y(t)$

1465: are given by

1466: \begin{equation}

1467: \label{apg0eq2}

1468: j_x(t)=\frac{\omega}{2\chi}s_z(t), \quad j_y(t)=\frac{1}{2\chi}

1469: \frac{ds_z(t)}{dt}.

1470: \end{equation}

1471: This results in the expressions for the difference in atomic and

1472: molecular populations $\bra n_{p} \ket-\bra n_b\ket$, and the

1473: coherence function $j_y$ of Eqs.~(\ref{scdresonance1}) and

1474: (\ref{scdresonance2}), respectively.

1475: 

1476: \subsection{Case $g\neq 0$}

1477: The presence of pairing interaction renders the potential

1478: quartic in $s_z$, see Eq.~(\ref{scdpotential}) where we have again

1479: assumed that the initial state is an atomic state of maximum spin,

1480: $|S,-S+M\ket$. It is convenient to introduce the function

1481: $f(\bs_z)$ as

1482: \begin{equation}

1483: \varphi(s_z)=-g^2(2S)^4(\bs_z-\bs_0)\cdot f(\bs_z),

1484: \end{equation}

1485: where

1486: \begin{eqnarray}

1487: f(\bs_z)&=&\bs_z^3+\alpha \bs_z^2+\beta \bs_z+\gamma \nonumber\\

1488: &=&(\bs_z-\bs_1)(\bs_z-\bs_2)(\bs_z-\bs_3),

1489: \end{eqnarray}

1490: $\alpha=\bs_0-2\kappa/\eta-4/\eta^2$,

1491: $\beta=-\bs_0^2+\kappa^2/\eta^2$, and

1492: $\gamma=-\bs_0^3+2\kappa\bs_0^2/\eta-\kappa^2\bs_0/\eta^2+1/\eta$

1493: with $\eta=g\sqrt{\Omega}/\chi$.

1494: 

1495: The variables $\bs_1$, $\bs_2$, and $\bs_3$, which correspond to

1496: the roots of the cubic equation, $f(s_j)=0$, are obtained by

1497: ``Cardano's formula''. With $\xi\equiv e^{i2\pi/3}$,

1498: $p=-\alpha^2/3+\beta$, and $q=2\alpha^3/27-\alpha\beta/3+\gamma$,

1499: and also $D=-(4p^3+27q^2)$, those roots are given by

1500: \begin{eqnarray}

1501: s_j&=&-\frac{\alpha}{3}+\xi^{j-1}\left[-\frac{q}{2}+\frac{1}{6}\sqrt{-\frac{D}{3}}\right]^{1/3}\nonumber\\

1502: &&\qquad+\xi^{1-j}\left[-\frac{q}{2}-\frac{1}{6}\sqrt{-\frac{D}{3}}\right]^{1/3},

1503: \end{eqnarray}

1504: where $j=1,2,3$.

1505: 

1506: The numbers of real and complex roots are determined by the sign

1507: of the polynomial discriminant $D$. If (a) $D > 0$, all three

1508: roots are real and unequal. If (b) $D<0$, one root is real and two

1509: are complex conjugates. If (c) $D=0$, two roots are equal for

1510: $q\neq 0$, and all roots are equal for $q=p=0$.

1511: 

1512: 

1513: \subsubsection{Case $D > 0$}

1514: 

1515: Suppose that $\bs_0\ge \bs_z\ge \bs_a$, and $\bs_a>\bs_b>\bs_c$ or

1516: $\bs_b>\bs_c>\bs_0$, where each $s_{a,b,c}$ corresponds to one of

1517: the roots $s_j$'s. The solution of ~Eq.(\ref{integral}) reads then

1518: \begin{widetext}

1519: \begin{equation}

1520: t=\int^{s_z}_{s_0}\frac{ds}{\sqrt{\varphi(s)}}

1521: =-\frac{1}{gS\sqrt{(\bs_0-\bs_b)(\bs_a-\bs_c)}}\int^\phi_0

1522: \frac{d\theta}{\sqrt{1-k^2\sin^2{\theta}}},

1523: %={\rm sn}^{-1}\left(\sin{\phi(t)}\right),

1524: \end{equation}

1525: where

1526: \begin{equation}

1527: \phi(t)=\arcsin{\sqrt{\frac{(\bs_a-\bs_c)(\bs_0-\bs_z)}{(\bs_0-\bs_a)(\bs_z-\bs_c)}}},

1528: \qquad

1529: k=\sqrt{\frac{(\bs_0-\bs_a)(\bs_b-\bs_c)}{(\bs_0-\bs_b)(\bs_a-\bs_c)}},

1530: \end{equation}

1531: so that

1532: %\begin{widetext}

1533: \begin{equation}

1534: \label{apgeq1}

1535: \bs_z(t)=\frac{\bs_0(\bs_a-\bs_c)+\bs_c(\bs_0-\bs_a){\rm sn}^2\left(-gS\sqrt{(\bs_0-\bs_b)(\bs_a-\bs_c)}t;k\right)}

1536: {(\bs_a-\bs_c)+(\bs_0-\bs_a){\rm sn}^2\left(-gS\sqrt{(\bs_0-\bs_b)(\bs_a-\bs_c)}t;k\right)}.

1537: \end{equation}

1538: \end{widetext}

1539: 

1540: \subsubsection{Case $ D<0$}

1541: 

1542: Letting $\bs_a$ label the real root and with

1543: $\bs_b=\bs_c^*=\bs_R+i \bs_I$, we have that

1544: $\varphi(s_z)=-g^2(2S)^4(\bs_z-\bs_0)(\bs_z-\bs_a)

1545: \left\{(\bs_z-\bs_R)^2+\bs_I^2\right\}$. With the change of

1546: variable from $\bs_z$ ($\bs_0\ge \bs_z \ge \bs_a$) to

1547: \begin{equation}

1548: \frac{\bs_0-\bs_z}{\bs_z-\bs_a}=\frac{A}{B}\frac{1-\cos{\phi}}{1+\cos{\phi}},

1549: \end{equation}

1550: where $A=\sqrt{(\bs_0-\bs_R)^2+\bs_I^2}>0$ and

1551: $B=\sqrt{(\bs_a-\bs_R)^2+\bs_I^2}>0$, and taking the elliptic

1552: modulus as

1553: \begin{equation}

1554: k=\sqrt{\frac{(\bs_0-\bs_a)^2-(A-B)^2}{4AB}},

1555: \end{equation}

1556: the integral Eq.~(\ref{integral}) can then be replaced by

1557: \begin{widetext}

1558: \begin{equation}

1559: t=\int^{s_z}_{s_0}\frac{ds}{\sqrt{\varphi(s)}}

1560: =-\frac{1}{2gS\sqrt{AB}}\int^{\phi}_0\frac{d\theta}{\sqrt{1-k^2\sin^2{\theta}}}.

1561: %={\rm sn}^{-1}(\sin{\phi(t})),\\

1562: \end{equation}

1563: The semiclassical solution is then given by

1564: \begin{equation}

1565: \label{apgeq2}

1566: \bs_z(t)=\frac{\bs_aA+\bs_0B-(\bs_aA-\bs_0B){\rm cn}\left(-2gS\sqrt{AB}t;k\right)}

1567: {A+B-(A-B){\rm cn}\left(-2gS\sqrt{AB}t;k\right)}.

1568: \end{equation}

1569: \end{widetext}

1570: 

1571: \subsubsection{Case $ D=0$}

1572: 

1573: In this case, the solution of Eq.~(\ref{integral}) can be

1574: expressed in terms of elementary functions. For $q\neq 0$ the

1575: solutions are equivalent to Eq.~(\ref{apgeq1}) in which the

1576: elliptic functions are replaced by trigonometric functions for

1577: $k=0$ ($\bs_b=\bs_c$), or hyperbolic functions for $k=1$

1578: ($\bs_0>\bs_a=\bs_b$). If $q=p=0$, the function $f(\bs_z)$ has

1579: triple degenerate roots at a point $\bs_a=\bs_b=\bs_c$, and the

1580: corresponding solution is obtained by

1581: \begin{equation}

1582: \label{apgeq3}

1583: \bs_z(t)=\bs_a+\frac{\bs_0-\bs_a}{1-\left\{-gS(\bs_0-\bs_a)t\right\}^2}.

1584: \end{equation}

1585: 

1586: \subsection{\label{phasediagram}Phase diagram of semiclassical dynamics}

1587: 

1588: In this subsection, we discuss the structure of the semiclassical

1589: dynamics in $\kappa-\eta$ parameter space for the specific case of

1590: a half-filled shell, $2M=\Omega$, by calculating the elliptic

1591: modulus of the semiclassical solution.

1592: 

1593: \begin{figure}

1594:     \begin{center}

1595:       \includegraphics[width=7.5cm,clip]{appfig1.eps}

1596:     \end{center}

1597:     \caption{Elliptic modulus in $\kappa-\eta$ space

1598:       for the half-filled shell: $2M=\Omega$.}

1599:     \label{appfig1}

1600: \end{figure}

1601: 

1602: For $g=0$, Eq.~(\ref{apg0eq1}) describe all possible dynamics for

1603: arbitrary detuning energy $\omega$, while in the presence of a

1604: pairing interaction the dynamics is given by

1605: solutions~(\ref{apgeq1}) and (\ref{apgeq2}), depending on the sign

1606: of $D$. The dynamics is described by Eq.~(\ref{apgeq3}) at a

1607: single singular point in $\kappa-\eta$ space, as we shall see

1608: later.

1609: 

1610: Figure~\ref{appfig1} shows the elliptic modulus of the

1611: semiclassical solution in $\kappa-\eta$ space. The two regions $D>

1612: 0$ and $D<0$ are separated by $D=0$ lines that correspond to the

1613: specific values of elliptic modulus $k=0$ and $k=1$.

1614: Eqs.~(\ref{apgeq1}) and (\ref{apgeq2}) coincide for $k=0$, which

1615: corresponds to the white lines in Fig.~\ref{appfig1}. Hence these

1616: two solutions connect continuously when crossing that line. For

1617: the black line $k=1$, on the other hand, solutions corresponding

1618: to Eq.~(\ref{apgeq1}) differ from Eq.~(\ref{apgeq2}). This

1619: discontinuity gives rise to the self-trapping transition discussed

1620: in subsection~\ref{selftrap}.

1621: 

1622: Figure~\ref{appfig1} shows that the black line and one of the

1623: white lines intersect at the critical point $(\kappa_c, \eta_c)$,

1624: where $q=p=0$, given explicitly by

1625: \begin{eqnarray}

1626: \eta_c&=&2\sqrt{\frac{2}{3}} (45+26\sqrt{3})^{1/4}\simeq 5.0302\\

1627: \kappa_c&=&-\frac{4}{\eta_c}(2+\sqrt{3})\simeq -2.9677.

1628: \end{eqnarray}

1629: Form this result, we conclude that the self-trapping transition

1630: appears by varying the detuning parameter $\kappa$ only for

1631: $\eta>\eta_c$ .

1632: %

1633: %

1634: \begin{references}

1635: \bibitem{Wynar00} R.~Wynar {\it et al.}, Science {\bf 287}, 1016 (2000).

1636: %

1637: \bibitem{Inouye98} S.~Inouye {\it et al.}, Nature (London) {\bf 392}, 151 (1998).

1638: %

1639: \bibitem{Yurovsky99} V.~A.~Yurovsky, A.~Ben-Reuven, P.~S.~Julienne,

1640: and C.~J.~Williams, Phys.~Rev.~A {\bf 60}, R765 (1999).

1641: %

1642: \bibitem{Donley02} E.~A.~Donley, N.~R.~Claussen, S.~T.~Thompson,

1643: and C.~E.~Wieman, Nature (London) {\bf 417}, 529 (2002).

1644: %

1645: \bibitem{Durr04} S.D{\"u}rr, T.~Volz, A.~Marte, and G.Rempe,

1646: Phys.~Rev.~Lett. {\bf 92}, 020406 (2004).

1647: %

1648: \bibitem{Xu03} K.~Xu, T.~Mukaiyama, {\it et al.}, Phys.~Rev.~Lett.

1649: {\bf 91}, 210402 (2003).

1650: %

1651: \bibitem{Petrov04} D.~S.~Petrov, C.~Salomon, and G.~V.~Shlyapnikov,

1652: Phys.~Rev.~Lett. {\bf 93}, 090404 (2004).

1653: %

1654: \bibitem{Greiner03} M.~Greiner, C.~A.~Regal, and D.~S.~Jin, Nature (London)

1655: {\bf 426}, 537 (2003).

1656: %

1657: \bibitem{Zwierlein03} M.~W.~Zwierlein {\it et al.}, Phys.~Rev.~Lett.

1658: {\bf 91}, 250401 (2003).

1659: %

1660: \bibitem{Jochim03} S.~Jochim {\it et al.}, Scince {\bf 301}, 2101 (2003).

1661: %

1662: \bibitem{Stan04} C.~A.~Stan {\it et al.}, cond-mat/0406129.

1663: %

1664: \bibitem{Inouye04} S.~Inouye {\it et al.}, cond-mat/0406208.

1665: %

1666: \bibitem{Kerman04} A.~J.~Kerman, J.~M.~Sage, S.~Sainis,

1667: T.~Bergeman, and D.~DeMille Phys.~Rev.~Lett.

1668: {\bf 92}, 033004 (2004).

1669: %

1670: \bibitem{ex_BECBCS} C.~A.~Regal, M.~Greiner, and D.~S.~Jin,

1671: Phys.~Rev.Lett. {\bf 92}, 040403; M.~Bartenstein {\it et al.},

1672: Phys.~Rev.~Lett. {\bf 92}, 120401 (2004); M.~Zwierlein {\it et al.},

1673: Phys.~Rev.~Lett. {\bf 92}, 120403 (2004).

1674: %

1675: \bibitem{Timmermans} E.~Timmermans, K.~Furuya, P.~W.~Milonni, and A.~K.~Kerman, Phys.~Lett. A {\bf 285}, 288 (2001).

1676: %

1677: \bibitem{th_BECBCS} M.~Holland, S.~J.~J.~M.~F. Kokkelmans, M.~L.~Chiofalo,

1678: and R.~Walser, Phys.~Rev.~Lett. {\bf 87}, 120406 (2001); Y.~Ohashi and

1679: A.~Griffin, Phys.~Rev.~Lett. {\bf 89}, 130402 (2002).

1680: %

1681: \bibitem{Mottelson} A.~Bohr and B.~R.~Mottelson, ``Nuclear Structure''

1682: (Benjamin, New York, 1975), Vols. I and II.

1683: %

1684: \bibitem{SCgrain} C.~T.~Black, D.~C.~Ralph, and M.~Tinkham, Phys.~Rev.~Lett.

1685: {\bf 76}, 688 (1996).

1686: %

1687: \bibitem{Anderson} P.~W.~Anderson, Phys.~Rev.~, {\bf 112}, 1900 (1958).

1688: %

1689: \bibitem{Kerman61} A.~Kerman, Ann.~Phys. (N.Y.), {\bf 12}, 300 (1961).

1690: %

1691: \bibitem{Tavis} M.~Tavis and F.~W.~Cummings, Phys.~Rev. {\bf 170}, 379 (1968).

1692: %

1693: \bibitem{Bogoliubov} N.~M.~Bogoliubov, R.~K.~Bullough, and J.~Timonen,

1694: J.~Phys.~A {\bf 29}, 6305 (1996).

1695: %

1696: \bibitem{Dicke} R.~H.~Dicke, Phys.~Rev. {\bf 93}, 99 (1954).

1697: %

1698: \bibitem{Bonifacio} R.~Bonifatio and G.~Preparata, Phys.~Rev.~A {\bf 2},

1699: 336 (1970).

1700: %

1701: \bibitem{dy_BECBCS} J.~Javanainen {\it et al.}, Phys.~Rev.~Lett. {\bf 92}, 200402 (2004);

1702: A.~V.~Andreev, V.~Gurarie, and L.Radzihovsky, Phys.~Rev.~Lett. {\bf 93},

1703: 130402 (2004); R.~A.~Barankov and L.~S.~Levitov, Phys.~Rev.~Lett. {\bf 93},

1704: 130403 (2004).

1705: %

1706: \bibitem{Vardi} I.~Tikhonenkov and A.~Vardi, cond-mat/0407424.

1707: %

1708: \bibitem{BJJ1} G.~J.~Milburn, J.~Corney, E.~M.~Wright, and, D.~F.~Walls,

1709: Phys.~Rev.~A {\bf 55}, 4318 (1997); A.~Smerzi, S.~Fantoni, S~Giovanazzi,

1710: and S.~R.Shenoy, Phys.~Rev.~Lett. {\bf 79}, 4950 (1997).

1711: %

1712: \bibitem{BJJ2} J.~Javanainen and M.~Y.~Ivanov, Phys.Rev.~A {\bf 60},

1713: 2351 (1999); A.~J.~Leggett, Rev.~Mod.~Phys. {\bf 73}, 307 (2001);

1714: J.~R.~Anglin and A.~Vardi, Phys.~Rev.~A {\bf 64}, 013605 (2001).

1715: %

1716: %

1717: \bibitem{Scott} J.~C.~Eilbeck, P.~S.~Lomdahl, and A.~C~.Scott,Physica D

1718: {\bf 16}, 318 (1985); A.~C.~Scott and J.~C.~Eilbeck, Phys.~Lett. A {\bf 119},

1719: 60 (1986).

1720: %

1721: \bibitem{Racah} G.~Racah, Phys.~Rev. {\bf 63}, 367 (1943).

1722: %

1723: \bibitem{CT} see e.g. C. Cohen-Tannoudji, B. Diu and F. Lalo{\"e},

1724: ``Quantum Mechanics'', Vol.~I Complement $B_{VII}$ (A Wiley-Interscience

1725: Publication, 1977).

1726: %

1727: \bibitem{Heiselberg} H.~Heiselberg and B.~Mottelson, Phys.~Rev.~Lett.

1728: {\bf 88}, 190401 (2002); G.~M.~Bruun and H.~Heiselberg, Phys.~Rev.~A {\bf 65}, 053407 (2002);

1729: H.~Heiselberg, Phys.~Rev.~A {\bf 68}, 053616 (2003).

1730: %

1731: \bibitem{superchemistry} D.~J.~Heinzen, R.~Wynar, P.~D.~Drummond, and

1732: K.~V.~Kheruntsyan, Phys.~Rev.~Lett. {\bf 84}, 5029 (2000).

1733: %

1734: \bibitem{Richardson} R.~W.~Richardson, Phys.~Lett. {\bf 3} 277 (1963);

1735: R.~W.~Richardson and N.~Sherman, Nucl.~Phys. {\bf 52}, 221 (1964).;

1736: R.~W.~Richardson, J.~Math.~Phys. (N.Y.) {\bf 6}, 1034 (1965).

1737: %

1738: \bibitem{Dukelsky}

1739: J.~Dukelsky, C.~Esebbag, and P.~Schuck, Phys.~Rev.~Lett. {\bf 87},

1740: 066403 (2001).

1741: %

1742: \bibitem{AMHamiltonian}

1743: J.~Dukelsky, G.~G.~Dussel, C.~Esebbag, and S.~Pittel, Phys.~Rev.~Lett.

1744: {\bf 93}, 050403 (2004).

1745: %

1746: \bibitem{Schuck} P.~Ring and P.~Schuck, "The Nuclear Many-Body Problem"

1747: (Springer-Verlag, New York, 1980).

1748: %

1749: \bibitem{Walecka} A.~L.~Fetter and J.~D.~Walecka, "Quantum Theory of

1750: Many-Particle Systems" (Dover, Mineola, NY, 2003).

1751: %

1752: \bibitem{HP40} T.~Holstein and Primakoff, Phys.~Rev. {\bf 58}, 1098 (1940).

1753: %

1754: \bibitem{AMBEC1} J.~Javanainen and M.~Mackie, Phys.~Rev.~A {\bf 59},

1755: R3186 (1999).

1756: %

1757: \bibitem{AMBEC2} A.~Vardi, V.~A.~Yurovsky, and J.~R.~Anglin,

1758: Phys.~Rev.~A {\bf 64}, 063611 (2001).

1759: %

1760: \bibitem{Zhou} H.~Q.~Zhou, J.~Links, and R.~H.~McKenzie,

1761: cond-mat/0207540.

1762: %

1763: \bibitem{Hines} A.~P.~Hines, R.~H.~McKenzie, and G.~J.~Milburn,

1764: Phys.~Rev.~A {\bf 67}, 013609 (2003).

1765: %

1766: \bibitem{search} C.~P.~Search, W.~Zhang, and P.~Meystre, Phys.~Rev.~Lett.

1767: {\bf 91}, 190401 (2003); C.~P.~Search, T.~Miyakawa, and P.~Meystre,

1768: Optics Express {\bf 12}, 3318 (2004).

1769: %

1770: \bibitem{elliptic} H.~Hancock: {\it Elliptic Integrals} (Dover Publications,

1771: Inc., New York, 1917); {\it Handbook of Mathematical Functions}

1772: Eddited by M.~Abramowitz and I.~A.~Stegun (Dover Publications, Inc.,

1773: New York, 1972).

1774: %

1775: \bibitem{Walls} D.~F.~Walls, Phys.~Lett. {\bf 32A}, 476, (1970);

1776: D.~F.~Walls and C.~T.~Tindle, J.~Phys.~A {\bf 5}, 534 (1972).

1777: %

1778: \end{references}

1779: \end{document}

1780: