1: \documentclass[prl,twocolumn,amsmath,amssymb,groupedaddress]{revtex4}

2: 

3: %\documentclass[twocolumn,showpacs,preprintnumbers,amsmath,amssymb,aps]{revtex4}

4: %\documentclass[twocolumn,showpacs,preprintnumbers,amsmath,amssymb,aps,eqsecnum]{revtex4}

5: %\documentclass[preprint,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}

6: 

7: % Some other (several out of many) possibilities

8: %\documentclass[preprint,aps]{revtex4}

9: %\documentclass[preprint,aps,draft]{revtex4}

10: %\documentclass[prb]{revtex4}% Physical Review B

11: 

12: %\usepackage{bm}% bold math

13: %\usepackage{dcolumn}% Align table columns on decimal point

14: 

15: \newcommand{\s}{{\sigma}}

16: \newcommand{\rv}{{\bf r}}

17: \newcommand{\xv}{{\bf x}}

18: \newcommand{\kv}{{\bf k}}

19: \newcommand{\eps}{{\varepsilon}}

20: \newcommand{\uv}{{\bf u}}

21: \newcommand{\vv}{{\bf v}}

22: \newcommand{\qv}{{\bf q}}

23: \newcommand{\nv}{{\bf n}}

24: \def\mm#1{\underline{\underline{{#1}}}}

25: \newcommand{\um}{{\mm{u} }}

26: \newcommand{\Lm}{{\mm{\Lambda} }}

27: \newcommand{\pv}{{\bf p}}

28: \newcommand{\Tr}{{\rm Tr}}

29: \newcommand{\px}{{\partial_x}}

30: \newcommand{\py}{{\partial_y}}

31: \newcommand{\ppi}{{\partial_i}}

32: \newcommand{\ppj}{{\partial_j}}

33: \newcommand{\pz}{{\partial_z}}

34: \newcommand{\ppk}{{\partial_k}}

35: \newcommand{\ppl}{{\partial_l}}

36: \usepackage[dvips]{graphicx}

37: 

38: \begin{document}

39: \title[]{Superfluid transitions in bosonic atom-molecule mixtures near

40: Feshbach resonance}

41: %Spontaneously

42: %Orientationally-Ordered Solids}

43: \author{Leo Radzihovsky}

44: \author{Jae Park}

45: %\email[]{radzihov@colorado.edu}

46: %\homepage[]{Your web page}

47: %\thanks{}

48: %\altaffiliation{}

49: \affiliation{Department of Physics, University of Colorado,

50: Boulder, CO 80309}

51: \author{Peter B. Weichman}

52: \affiliation{ALPHATECH, Inc., 6 New England Executive Place,

53: Burlington, MA 01803}

54: 

55: %Collaboration name if desired (requires use of superscriptaddress

56: %option in \documentclass). \noaffiliation is required (may also be

57: %used with the \author command).

58: %\collaboration{}

59: %\noaffiliation

60: 

61: \date{\today}

62: 

63: \begin{abstract}

64: We study bosonic atoms near a Feshbach resonance, and predict that

65: in addition to a standard normal and atomic superfluid phases,

66: this system generically exhibits a distinct phase of matter: a

67: molecular superfluid, where molecules are superfluid while atoms

68: are not. We explore zero- and finite-temperature properties of the

69: molecular superfluid (a bosonic, strong-coupling analog of a BCS

70: superconductor), and study quantum and classical phase transitions

71: between the normal, molecular superfluid and atomic superfluid

72: states.

73: \end{abstract}

74: \pacs{}

75: 

76: \maketitle

77: 

78: Experimental realizations and coherent manipulation of

79: trapped degenerate gases \cite{JILAbec,MITbec} is leading to

80: exciting possibilities for studies of quantum liquids in

81: previously unexplored (e.g., extremely coherent and

82: nonequilibrium) regimes.  Magnetic field-induced Feshbach

83: resonance (FBR) in ultracold atom collisions allows fine

84: tuning of interactions in these quantum fluids, and was recently

85: used to create a degenerate mixture of coherently-coupled alkali

86: atoms and their diatomic molecules \cite{donley}.

87: 

88: \begin{figure}[bth]

89: \centering

90: \setlength{\unitlength}{1mm}

91: \begin{picture}(70,65)(0,0)

92: \put(-30,-62){\begin{picture}(60,60)(0,0)

93: \special{psfile=phase_diagramA.eps vscale=62 hscale=62}

94: \end{picture}}

95: \end{picture}

96: \caption{Mean-field phase diagram for a bosonic atom-molecule mixture,

97: showing molecular (MSF) and atomic (ASF) superfluid phases.}

98: \label{phasediagramA}

99: \end{figure}

100: 

101: In this Letter we study phases and phase transitions that take

102: place in bosonic atom-molecule mixtures. Our main contribution is the

103: prediction of a thermodynamically distinct ``molecular

104: superfluid'' (MSF) phase, that, as illustrated in 

105: Figs.\ \ref{phasediagramA}, \ref{phasediagramB} ubiquitously

106: intervenes between the ``normal'' (N) and ``atomic superfluid''

107: (ASF) phases. Molecular superfluidity [and accompanying

108: off-diagonal long-range molecular order (ODLRO)] distinguishes MSF

109: from the normal state, and the {\em absence} of atomic

110: superfluidity from the ASF, in which both bosonic atoms and

111: molecules display ODLRO.  If atomic and molecular components can

112: be imaged independently \cite{jinBCS}, in a harmonic trap MSF

113: should be easily identifiable by a sharp Bose-Einstein

114: condensation (BEC) peak in the molecular density profile and a

115: broad, seemingly normal, thermal atomic cloud.

116: 

117: As a conventional superfluid, MSF is characterized by a

118: (molecular) acoustic second-sound mode. However, MSF also exhibits

119: a {\em gapped}, Bogoliubov-like mode, derived from unpaired atom

120: excitations.  MSF ground state (bosonic analog of the BCS state)

121: exhibits strong (atom and molecule) pairing correlations that in a

122: trap should be observable in the atomic density-density

123: correlation function. Experimentally, MSF should be accessible by

124: tuning temperature, atomic density (or number), and detuning

125: $\nu$. The MSF-ASF transition is in the $(d+1)$- and

126: $d$-dimensional Ising universality classes for $T=0$ and finite

127: $T$, respectively, and is {\em reentrant} as a function of

128: detuning $\nu$ and density $n$. The tricritical point, where N,

129: MSF and ASF meet, exhibits nontrivial and, to our knowledge,

130: unexplored quantum critical behavior for $d < 4$. We now sketch

131: derivation of these results.

132: 

133: Near a FBR a bosonic atom-molecule system is characterized by the

134: grand-canonical Hamiltonian $\hat{H}_\mu = \hat{H} - \mu \hat{N}$

135: \cite{foot1}

136: \begin{eqnarray}

137: \hat{H}_\mu &=& \int d^d{\bf x} \Bigg[\sum_{\s=1}^2

138: \left(\hat{\psi}^\dagger_\s \hat{h}_\s \hat{\psi}_\s

139: + \frac{g_\s}{2} \hat{\psi}^\dagger_\s

140: \hat{\psi}^\dagger_\s \hat{\psi}_\s \hat{\psi}_\s \right)

141: \nonumber \\

142: &&+\ g_{12} \hat{\psi}^\dagger_1 \hat{\psi}^\dagger_2

143: \hat{\psi}_2 \hat{\psi}_1

144: - \alpha \left(\hat{\psi}^\dagger_1 \hat{\psi}^\dagger_1

145: \hat{\psi}_2 + {\rm h.c.} \right) \Bigg]

146: \label{hamiltonian}

147: \end{eqnarray}

148: where $\hat{\psi}^\dagger_\s({\bf x}),\hat{\psi}_\s({\bf x})$ are

149: bosonic field operators for atoms ($\s=1$) and molecules ($\s=2$),

150: $\hat{h}_{\s} = -(\hbar^2/2m_\s) \nabla^2 - \mu_\s$ are the

151: corresponding single particle Hamiltonians (focusing for

152: concreteness on the case of a homogeneous trap) with effective

153: chemical potentials $\mu_{1} = \mu$ and $\mu_2 = 2\mu - \nu$.

154: Chemical potential $\mu$ tunes the average {\em

155: total} number of atoms (whether free or bound into molecules) to

156: $N$, and detuning $\nu$ is related to the energy of a molecule at rest,

157: that can be experimentally controlled with a magnetic field. In

158: the dilute gas limit $g_1, g_2, g_{12}$ are proportional to the

159: 2-body s-wave atom-atom, atom-molecule and molecule-molecule

160: scattering lengths, respectively, and $\alpha$ characterizes

161: \emph{coherent} atom-molecule interconversion rate,

162: encoding that molecules are composed of two atoms\cite{foot1}.

163: 

164: The mean-field phase diagram as a function of $\mu_{1,2}$ and

165: $\beta = 1/k_B T$ can be worked out by

166: minimizing the imaginary-time ($\tau$)

167: coherent-state action $S = \int_0^{\beta\hbar} d\tau \int d^d{\bf x}

168: \sum_{\s=1}^2 \left[\psi^*_\s\hbar \partial_\tau \psi_\s +

169: H_\mu(\psi^*_\s,\psi_\s) \right]$. Simple analysis leads to three

170: thermodynamically distinct phases (Fig.\ \ref{phasediagramA}): (i)

171: ``normal'' (N): $\Psi_{10} \equiv \langle \hat \psi_1 \rangle =

172: 0$, $\Psi_{20} \equiv \langle \hat \psi_2 \rangle = 0$, (ii)

173: ``molecular superfluid'' (MSF): $\Psi_{10} = 0, \Psi_{20} \neq 0$,

174: (iii) ``atomic superfluid'' (ASF): $\Psi_{10} \neq 0, \Psi_{20}

175: \neq 0$. Condensed atoms cause $\alpha$ to act as an effective

176: field on the molecular order parameter $\Psi_{20}$, so an

177: equilibrium phase in which atoms are condensed, but molecules are

178: not, is forbidden \cite{comment1}.

179: 

180: We now examine in more detail these phases and corresponding phase

181: transitions.  Phase N is stable for $\mu_{1,2} < 0$, with $\mu$

182: determined by the total atom constraint $n = n_1 + 2 n_2$, which

183: in the non-interacting limit, appropriate to a dilute weakly

184: interacting gas, is given by:

185: \begin{equation}

186: n = \frac{1}{\Lambda_T^d} \left[f_{d/2}\left(e^{\beta\mu} \right)

187: + 2^{(d+2)/2} f_{d/2}\left(e^{\beta(2\mu - \nu)} \right) \right],

188: \label{n_conserve}

189: \end{equation}

190: where $\Lambda_T = h/\sqrt{2\pi m_1 k_BT}$ is the thermal de

191: Broglie wavelength and $f_\alpha(z) = \sum_{n=1}^\infty

192: z^n/n^\alpha$ ($|z| < 1$) is the extended zeta-function.

193: 

194: The N-ASF transition line $T_{c1}(n,\nu)$ occurs at $\mu = 0$ for

195: $\nu > 0$, while the N-MSF line $T_{c2}(n,\nu)$ occurs at $\mu -

196: 2\nu = 0$ for $\nu < 0$ (see Fig.\ \ref{phasediagramB}).  Using

197: the appropriate asymptotics of $f_\alpha(z)$, one obtains from

198: (\ref{n_conserve}):

199: \begin{equation}

200: T_{c\sigma}(n,\nu) =

201: \begin{cases}

202: T_{c0} \left[1 + a_\sigma \left(\frac{|\nu|}{k_B T_{c0}}

203: \right)^{\frac{d-2}{2}} \right],

204: & |\nu| \ll k_B T_{c0} \\

205: T_{c\sigma}^\infty = b_\sigma c^{2/d} T_{c0},

206: & |\nu| \gg k_B T_{c0},

207: \end{cases}

208: \label{Tc}

209: \end{equation}

210: with $c = 1 + 2^{(d+2)/2}$, $a_1 = 2^{(d+4)/2}

211: |\Gamma\left(\frac{2-d}{2}\right)|/d c\zeta(d/2)$, $a_2 = 2^{-d}

212: a_1$, $b_1 = 1$, $b_2 = 2^{-(d+2)/d}$  and $T_{c0} = (h^2/2\pi m_1

213: k_B) [n/c\zeta(d/2)]^{2/d}$ the transition temperature at the

214: tricritical point $\nu = 0$.

215: 

216: In the neighborhood of $T_{c1}$ the ``massive'' molecular field

217: $\hat{\psi}_2$ decouples at low energies (can be safely integrated out

218: of the partition function, leading to an effective quartic

219: coupling $g_1 \rightarrow \bar g_1 \equiv g_1 -

220: 2\alpha^2/|\mu_2|$), and the N-ASF transition is identical to that

221: of a single-component system, continuous so long as $\bar g_1 >

222: 0$. At $T=0$, the N-ASF transition takes place at vanishing atom

223: density (the N phase is simply a vacuum of atoms), and although

224: nontrivial, is exactly soluble \cite{uzunov}, corresponding to a

225: build-up of atomic superfluid (with condensate density $n_{10} =

226: |\Psi_{10}|^2 \sim |\mu|^{2\beta}$, with mean-field result $\beta

227: = 1/2$ for $d > 2$, and $\beta = d/4$ for $d < 2$) as the trap is

228: loaded. At $T \neq 0$ the N-ASF transition lies in the usual

229: $d$-dimensional XY-universality class \cite{foot2}.

230: 

231: Similarly, in the neighborhood of $T_{c2}$, $\hat \psi_1$

232: decouples and the resulting N-MSF transitions are in the same

233: universality classes discussed above.  The full phase boundary is

234: illustrated in Fig.\ref{phasediagramB}. In 3d it exhibits a

235: square-root singularity at the tricritical point and for $|\nu|

236: \to \infty$ asymptotes to the single-component BEC temperatures

237: $T_{c\sigma}^\infty$.

238: %with $T_{c1}^\infty/T_{c2}^\infty = 2^{5/3}$.

239: 

240: \begin{figure}[tbh]

241: \centering

242: \setlength{\unitlength}{1mm}

243: \begin{picture}(70,50)(0,0)

244: \put(-22,-66){\begin{picture}(50,50)(0,0)

245: \special{psfile=phase_diagramB.eps vscale=56 hscale=56}

246: \end{picture}}

247: \end{picture}

248: \caption{Phase diagram for a bosonic atom-molecule mixture in

249: $d=3$, expressed in terms of 

250: detuning $\nu$ and temperature $T$. It illustrates a finite-T

251: tricritical point at $T_{c0}$ and  a quantum critical point at

252: $\nu_c(n,0)$. In the weakly-interacting limit appropriate to

253: experiments the ratio $T_{c1}^\infty/T_{c2}^\infty = 2^{5/3}$.}

254: \label{phasediagramB}

255: \end{figure}

256: 

257: To study the MSF phase, we separate $\hat \psi_\sigma =

258: \Psi_{\sigma 0} + \hat{\phi}_\sigma$ into classical condensate

259: fields $\Psi_{\sigma 0}$ (with $\Psi_{10} = 0$ inside MSF) and

260: fluctuations about it. Within MSF it is sufficient to

261: expand $\hat H_\mu$ to second order in fluctuations $\hat

262: \phi_\sigma$, which leads to $\hat H_\mu = E^{(0)}(\Psi_{20}) +

263: \hat H^{(2)}$ with:

264: \begin{eqnarray}

265: E^{(0)} &=& \int d^d{\bf x}

266: \left(\Psi_{20}^* \hat h_2 \Psi_{20}

267: + \frac{g_2}{2} |\Psi_{20}|^4 \right)

268: \\

269: {\hat H}^{(2)} &=& \int d^d{\bf x} \left[\sum_{\sigma = 1}^2

270: \hat \phi_\sigma^\dagger \tilde h_\sigma  \hat \phi_\sigma

271: + \frac{1}{2} \left(\lambda_\sigma \hat \phi_\sigma^\dagger

272: \hat \phi_\sigma^\dagger + {\rm h.c} \right) \right],

273: \nonumber

274: \label{Htr}

275: \end{eqnarray}

276: in which $\tilde h_\sigma = \hat h_\sigma + r_\sigma$, 

277: $r_1 = g_{12} |\Psi_{20}|^2$, $r_2 = 2g_2 |\Psi_{20}|^2$, $\lambda_1 = -2\alpha

278: \Psi_{20}$, and $\lambda_2 = g_2 \Psi_{20}^2$.  The linear term in

279: $\hat \phi_2, \hat \phi_2^\dagger$ vanishes automatically by the

280: self-consistent choice of $\Psi_{20}$ as the true minimum of the

281: free energy. To lowest order this gives

282: \begin{equation}

283: n_{20} \equiv |\Psi_{20}|^2 = (2\mu - \nu)/g_2,

284: \label{n20}

285: \end{equation}

286: which coincides with the minimum of $E^{(0)}$ and allows us to

287: eliminate $\mu$ in favor of $\nu$ and $n_{20}$.

288: 

289: For a homogeneous system, $\hat{H}^{(2)}$ may be diagonalized by

290: Fourier transformation $\hat \phi_\sigma = V^{-1/2} \sum_{\kv}

291: e^{i\kv \cdot \xv} \hat a_{\sigma \kv}$, followed by independent

292: Bogoliubov transformations on atoms and molecules to new boson

293: operators $\gamma_{\s,\kv}, \gamma^\dagger_{\s,\kv}$:

294: \begin{eqnarray}

295: \hat{\gamma}_{\s,\kv} &=& u_{\s,\kv} \hat a_{\s,\kv} +

296: v_{\s,\kv} \hat a^\dagger_{\s,-\kv}

297: \nonumber \\

298: |u_{\s,\kv}|^2 &=& 1 + |v_{\s,\kv}|^2 = \frac{1}{2}

299: \left(\frac{\tilde \eps_{\s,\kv}}{E_{\s,\kv}} + 1 \right)

300: \nonumber \\

301: \hat{H}^{(2)} &=& \sum_{\sigma,\kv} E_{\s,\kv}

302: \left(\hat \gamma^\dagger_{\s,\kv} \hat \gamma_{\s,\kv}

303: - |v_{\s,\kv}|^2 \right),

304: \end{eqnarray}

305: in which $\tilde \eps_{\s,\kv} = \eps_{\s,\kv} - \mu_\s + r_\s$,

306: $\eps_{\s,\kv} = \hbar^2k^2/2m_\sigma$ and $E_{\s,\kv} =

307: \sqrt{\tilde \eps_{\s,\kv}^2 - |\lambda_\sigma|^2}$, and

308: $v_{\s,\kv},u_{\s,\kv}^*$ have the same phase as $\lambda_\sigma$.

309: Using (\ref{n20}) one obtains

310: \begin{eqnarray}

311: E_{1,\kv} &=& \sqrt{\left[\eps_{1,\kv} - \nu/2

312: - (g_2/2 - g_{12}) n_{20} \right]^2 - 4 \alpha^2 n_{20}}\;,

313: \nonumber \\

314: E_{2,\kv} &=& \sqrt{\eps_{2,\kv}^2 + 2g_2 n_{20}\, \eps_{2,\kv}}\;.

315: \end{eqnarray}

316: The MSF ground state is defined as vacuum of atomic and molecular

317: Bogoliubov quasi-particles $\gamma_{\s,\kv}|{\rm MSF}\rangle = 0$.

318: It is easy to show that it is given by:

319: \begin{equation}

320: |{\rm MSF}\rangle=e^{\Psi_{20} \hat a_{2,0}^\dagger}

321: \prod_{\s,\kv} e^{-\chi_{\s,\kv} \hat a_{\s,\kv}^\dagger

322: \hat a_{\s,\kv}^\dagger}|0\rangle,

323: \end{equation}

324: which is a coherent state in the $\kv = 0$ molecular mode and

325: shows BCS-like pairing correlations between time-reversed ($\kv,

326: -\kv$) atomic and molecular single particle states.  The amplitude

327: $\chi_{\s,\kv} =  v_{\s,\kv}/u_{\s,\kv}$ is the Fourier transform

328: of the wavefunction of $\kv \neq 0$ atom ($\sigma=1$) and molecule

329: ($\sigma=2$) pairs.

330: 

331: As in a single-component SF the relation (\ref{n20}) (and more

332: fundamentally, the Goldstone theorem \cite{ZinnJustin}) ensures

333: that MSF exhibits a gapless sound mode $E_{2,\kv} \approx \hbar

334: v_0 k$, for small $k$, with $v_0 = \sqrt{g_2 n_{20}/2m_1}$,

335: corresponding to collective, long wavelength oscillations of the

336: molecular condensate. The resulting $|\kv|$ singularity in

337: $\chi_{2,\kv}$ induces a long ranged power law tail in the Fourier

338: transform $\chi_2(r) \sim 1/r^{d+1}$.  In addition, we find a

339: \emph{gapped} excitation branch $E_{1,\kv}$ describing atomic-like

340: excitations, whose spectrum and eigenmodes ($\hat{\gamma}_{1,\kv},

341: \hat{\gamma}_{1,\kv}^\dagger$ do not even carry a definite atom

342: number) are qualitatively distinct from unpaired atomic

343: excitations, $\hat{a}_{1,\kv}, \hat{a}_{1,\kv}^\dagger$ of the N

344: phase. The condition of being in the MSF is that the atomic gap

345: $E^{(1)}_{\rm gap} = \sqrt{\eps_+ \eps_-}$, where

346: \begin{equation}

347: \eps_{\pm} \equiv -\nu/2-(g_2/2 - g_{12})n_{20}

348: \pm 2\alpha \sqrt{n_{20}},

349: \label{gap}

350: \end{equation}

351: remains positive.  Correspondingly, the atomic pair wavefunction

352: $\chi_1(r) \sim e^{-r/\xi_1}$ decays exponentially at large

353: distance.  The correlation length $\xi_1 = \hbar/\sqrt{2m \eps_-}$

354: characterizes atom-pair size, and diverges as

355: $E^{(1)}_{\rm gap}$ vanishes.  The MSF-ASF transition takes place when

356: the size of atom-pairs becomes comparable to the intermolecular

357: separation, and $\xi_1$ is the diverging size of the coherent

358: exchange loops of atoms between overlapping pairs.

359: 

360: In the present quadratic approximation, the free energy is given

361: by

362: \begin{eqnarray}

363: f &=& -\mu_2|\Psi_{20}|^2 + \frac{g_2}{2}|\Psi_{20}|^4

364: - \frac{1}{2V} \sum_{\s,\kv}(\tilde{\eps}_{\s,\kv} - E_{\s,\kv})

365: \nonumber\\

366: &&+\ \frac{k_B T}{V}\sum_{\s,\kv}\ln(1-e^{-E_{\s,\kv}/k_BT}).

367: \end{eqnarray}

368: The condition of fixed density $n = -(\partial f/\partial

369: \mu)_{n_{20}}$,

370: \begin{equation}

371: n = 2n_{20} + \frac{1}{V} \sum_{\s,\kv} \s

372: \left(v_{\s,\kv}^2 + \frac{u_{\s,\kv}^2 + v_{\s,\kv}^2}

373: {e^{E_{\s,\kv}/{k_BT}}-1} \right),

374: \end{equation}

375: where the first term under the summation represents the $T=0$

376: interaction-induced condensate depletion, may be used to determine

377: the condensate density $n_{20}(n,\nu,T)$.  At $T=0$, in the

378: weakly-interacting limit ($a^3 n\ll 1$), in 3d we find:

379: \begin{eqnarray}

380: n_{20} &=& \frac{1}{2} n - \frac{2^{(d-2)/2}

381: \Gamma\left(\frac{d-1}{2}\right)

382: \Gamma\left(\frac{4-d}{2}\right)}{d \sqrt{\pi}

383: \Gamma\left(\frac{d}{2}\right)} (n a_2)^{d/2}

384: \nonumber \\

385: &&-\ \frac{[(n_\nu - n) \bar a_{12}]^{d/2}}

386: {2^{d+1} \Gamma\left(\frac{d}{2}\right)}

387: {\cal I}_d\left(\frac{\bar \alpha \sqrt{n}}{n_\nu - n} \right)

388: \label{n20n}

389: \end{eqnarray}

390: where $\bar a_{12} \equiv a_{2} - 2 a_{12}$,

391: $a_{2}, a_{12}$ are molecule-molecule and atom-molecule

392: scattering lengths, respectively, 

393: defined by $g_{12,2} = 4\pi\hbar^2 a_{12,2}/m_2$,

394: the ``detuning density'' is $n_\nu = m_1 |\nu|/(\pi \hbar^2 \bar

395: a_{12})$, $\bar \alpha = 2^{3/2} m_1 |\alpha|/(\pi\hbar^2 \bar

396: a_{12})$, and ${\cal I}(y) = y^2 \int_0^\infty dx

397: x^{d/2}[(1+x)^2 - y^2]^{-3/2}$ is a scaling function describing

398: additional molecular-condensate depletion due to atom pairs.  Equation

399: (\ref{n20n}), together with the vanishing of $\eps_-$, determines the

400: MSF-ASF phase boundary. At $T=0$, approximating $n_{20} \approx n/2$,

401: one finds:

402: \begin{equation}

403: \nu_c(n) \approx -(g_2/2-g_{12})n - 2\alpha\sqrt{2n},

404: \label{nuc}

405: \end{equation}

406: which is illustrated for $g_2 > 2g_{12}$ in Fig.\ \ref{phasediagramA}.

407: At finite $T$, in the noninteracting limit, $v_{\s,\kv} = 0$ and

408: $u_{\s,\kv} = 1$, molecular condensate density reduces to

409: \begin{equation}

410: n_{20}(\nu,T) = \frac{n}{2} \left[1 -

411: \left(\frac{T}{T_{c2}^\infty} \right)^{3/2} \right]

412: - \frac{1}{2 \Lambda_T^3} f_{3/2}\left(e^{-|\nu|/2k_B T} \right),

413: \end{equation}

414: with the standard BEC result (first term) corrected by depletion

415: due to thermal atomic excitations (last term), that is exponentially

416: small at large $|\nu|$ and low $T$.

417: 

418: From the structure of the interconversion ($\alpha$) term in $\hat

419: H_\mu$ it is clear that it is a $Z_2$ (Ising) symmetry that is

420: broken at the MSF$\rightarrow$ASF transition \cite{comment2}. We

421: now show that for a homogeneous trap, both the $T=0$ and finite

422: $T$ MSF-ASF transitions lie in the Ising universality class. This

423: can be most easily seen from the coherent state action, which,

424: when expressed in terms of real and imaginary parts of the atomic

425: field $\hat \psi_1 = \hat \psi_R + i \hat \psi_I$, in MSF phase

426: reduces to:

427: \begin{eqnarray}

428: S &=& \int_0^{\beta\hbar} d\tau \int d^d{\bf x}

429: \bigg[-2i \psi_I\hbar \partial_\tau\psi_R

430: + \psi_R(\hat{h}_1 - 2\alpha\Psi_{20})\psi_R

431: \nonumber \\

432: &&\ \ \ \ \ \ \ \ \ \

433: +\ \psi_I(\hat{h}_1 + 2\alpha\Psi_{20})\psi_I\bigg]

434: + S_{\rm int}

435: \end{eqnarray}

436: where $S_{\rm int}$ are terms not essential for our argument. In

437: this form it is clear that in the presence of the molecular

438: condensate, $|\Psi_{20}| > 0$, positive $\alpha$ reduces the

439: $O(2)$ symmetry to $Z_2$, with $\psi_R$ reaching criticality {\em

440: before} $\psi_I$.  Because canonically conjugate field $\psi_I$

441: remains ``massive'' at the critical point [defined by where the

442: coefficient of $\psi_R^2$ vanishes, consistent with (\ref{nuc})],

443: it can be safely integrated out and leads to a $d+1$-dimensional

444: (Lorentz-invariant) action even in the scalar order parameter

445: $\psi_R$. Therefore, as asserted above, the $T=0$ MSF-ASF

446: transition is in the $(d+1)$-dimensional Ising universality class.

447: The Ising transition is well studied, and leads to the following

448: predictions \cite{ZinnJustin}. For $d=3$, up to logarithmic

449: corrections, the mean-field theory derived above will be accurate.

450: On the other hand in 2d, MSF-ASF exponents are nontrivial but are

451: well-known. For example, standard scaling arguments predict:

452: \begin{eqnarray}

453: n_{10} \sim |\nu - \nu_c|^{2\beta_I},\ \

454: E_{gap}^{(1)} \sim |\nu - \nu_c|^{z_I \nu_I},

455: \end{eqnarray}

456: where $\beta_I \approx 0.31$, $z_I = 1$, and $\nu_I \approx 0.63$

457: are 3d Ising exponents. These, together with the

458: relevance of $T$ at this quantum critical point also imply a {\em

459: universal} shape of the MSF-ASF phase boundary $\nu_c(n,T) \sim

460: \nu_c(n,0) + a\, T^{1/\nu_I}$ in Fig.\ \ref{phasediagramB}. One

461: may hope that when long-lived molecular condensates are produced,

462: nontrivial behavior of $E_{gap}^{(1)}(\nu)$ maybe observed in

463: Ramsey fringes experiments \cite{donley}.

464: 

465: Although from symmetry point of view ASF state is quite

466: conventional, it also exhibits a set of interesting features not

467: found in a single-component SF.  For example, because it is a

468: discrete (Ising) symmetry that distinguishes ASF from MSF, it will

469: exhibit a gapped (Ising) mode in addition to the gapless

470: second-sound mode. As discussed above \cite{comment1} for $\alpha

471: \neq 0$, finite $\Psi_{20}$ is always induced within the ASF.

472: However, in the limit of small $\alpha$, we predict a sharp

473: crossover (indicated by a dashed $\mu_2^{\rm eff} \equiv \mu_2 -

474: \mu_1 g_{12}/\bar g_1 = 0$ curve in Figs.\ \ref{phasediagramA},

475: \ref{phasediagramB}, where the $\psi_2$ field would order for

476: $\alpha = 0$\ \cite{comment1}) 

477: between small molecular condensate $n_{20} \approx

478: (\alpha \mu_1/g_1 \mu_2^{\rm eff})^2$ for $\mu_2^{\rm eff} < 0$

479: and large $n_{20} \approx \mu_2^{\rm eff}/g_2$ for $\mu_2^{\rm

480: eff} > 0$, and  vanishes as $n_{20} \sim (\alpha

481: n_{10})^{2/\delta}$ for $\mu_2^{\rm eff} = 0$ and weak $\alpha$,

482: with $\delta$ a universal XY-model magnetic field exponent

483: ($\delta_{d=3} \simeq 5$) \cite{ZinnJustin}.

484: %This is in direct analogy to a crossover

485: %in a weak magnetic field between high

486: %$T$, low magnetization and low $T$, high

487: %magnetization regimes in a ferromagnet.

488: 

489: Another experimentally interesting feature of ASF is that a

490: seemingly standard $2\pi$ vortex in atomic condensate $\Psi_{10}$

491: will generically split into two $\pi$ vortices (see Fig.\

492: \ref{vortex_pi}) confined by a domain wall of length

493: \begin{equation}

494: r_0 = \sqrt{\hbar^2 n_{20}^{3/2}/2m\alpha n_{10}^2},

495: \end{equation}

496: that diverges as ASF$\rightarrow$MSF phase boundary is approached

497: \cite{LRunpublished}.

498: %

499: \begin{figure}[bth]

500: \centering

501: \setlength{\unitlength}{1mm}

502: \begin{picture}(70,22)(0,0)

503: \put(-35,-92){\begin{picture}(50,20)(0,0)

504: \special{psfile=vortex_pi.eps vscale=65 hscale=60}

505: \end{picture}}

506: \end{picture}

507: \caption{$2\pi$ atomic condensate vortex in ASF splits into

508: $\pi+\pi$ vortex pair connected by a ``normal'' domain wall.}

509: \label{vortex_pi}

510: \end{figure}

511: %

512: This arises because in large $\alpha$ limit a $2\pi$ vortex in atomic

513: condensate induces a $4\pi$ vortex in a molecular condensate. Such

514: double molecular vortex is unstable to two fundamental $2\pi$

515: molecular vortices, that, in 2d repel logarithmically, but are confined 

516: linearly inside the ASF phase.

517: 

518: Finally, standard renormalization-group analysis shows that

519: upper-critical dimensions for the tricritical point (dominated by

520: scaling of $\alpha$) are $d_{uc}^{T > 0}=6$ and $d_{uc}^{T=0} =

521: 4$, and therefore should display nontrivial critical properties

522: \cite{LRunpublished}.

523: 

524: We plan to present analysis of these problems as well as

525: generalization to experimentally relevant harmonic trap in a future

526: publication.

527: 

528: To conclude, we have studied a mixture of bosonic atoms and their

529: diatomic molecules and predicted an existence of a molecular

530: superfluid phase, qualitatively distinct from the normal and atomic

531: superfluid states. Ising transition MSF-ASF should be observable in

532: such systems as a function of detuning $\nu$, temperature and/or

533: density.

534: 

535: We acknowledge support by NSF DMR-0321848 and Packard Foundation

536: (LR,JP), and thank Chris Greene for discussions and Steve Girvin for 

537: comments on the manuscript.

538: \vspace{-0.5cm}

539: \begin{thebibliography}{99}

540: \vspace{-0.5cm}

541: 

542: \bibitem{JILAbec} M. H. Anderson, {\em et al.}, {\em Science} {\bf

543: 269}, 198 (1995).

544: 

545: \bibitem{MITbec} K. B. Davis {\em et al.}, \prl {\bf 75}, 3969

546: (1995).

547: 

548: \bibitem{donley} E. A. Donley, {\em et al.}, {\em Nature}

549: (London) {\bf 417}, 529 (2002).

550: 

551: \bibitem{jinBCS} C. A. Regal, {\em et al.}, {\em Nature} {\bf 424}, 47 (2003).

552: 

553: \bibitem{foot1} The effective Hamiltonian (\ref{hamiltonian}) may

554: be derived in various limits from an underlying atomic model, in

555: which molecules are explicitly identified as bound or quasibound

556: states of atom pairs (including atomic spin dynamics, if

557: necessary); J. I. Park, L. Radzihovsky and P. B. Weichman, manuscript 

558: in preparation.

559: 

560: \bibitem{comment1} This contrasts with a mixture of two {\em

561: independent} species of bosons with $\alpha=0$, that in addition

562: admits a $\Psi_{10} \neq 0$, $\Psi_{20} = 0$ distinct phase.

563: 

564: \bibitem{foot2} In the dilute, weakly interacting BEC limit the

565: critical behavior will be that of the ideal gas (i.e., Gaussian),

566: crossing over to XY behavior only very close to $T_{c1}$:  P. B.

567: Weichman, M. Rasolt, M. J. Stephen and M. E. Fisher, Phys.\ Rev.\

568: B {\bf 33}, 4632 (1986);  P. B. Weichman, Phys.\ Rev.\ B {\bf 38},

569: 8739 (1988).

570: 

571: \bibitem{GradshteynRizhik} I. S. Gradshteyn and I. M. Rizhik, {\em

572: Table of Integrals, Series and Products} (Academic Press, New

573: York, 1980).

574: 

575: \bibitem{uzunov} D. I. Uzunov, Phys.\ Lett.\ {\bf 87A}, 11 (1981).

576: 

577: %\bibitem{sachdev} S. Sachdev, {\em Quantum Phase Transitions},

578: %Cambridge University Press, Cambridge, 1999).

579: 

580: \bibitem{ZinnJustin} J. Zinn-Justin, {\em Quantum Field Theory and

581: Critical Phenomena}, (Oxford Science Puplications, Oxford, 1989).

582: 

583: \bibitem{comment2} In the ``hard-spin'' $\psi_\s=e^{i\theta_\s}$

584: description, interconversion term reduces to $\alpha

585: \cos(2\theta_1-\theta_2)$, locking atomic phase to the phase of

586: the molecular condensate up to $\pi$, corresponding to the

587: unbroken $Z_2$ symmetry of the MSF. In terms of a vector order

588: parameter, MSF is characterized by a finite quadrapole moment

589: ($l=2$ angular harmonic, akin to nematic order in liquid

590: crystals), and ASF by a vector order parameter ($l=1$ harmonic).

591: 

592: \bibitem{LRunpublished} L. Radzihovsky, unpublished.

593: 

594: \end{thebibliography}

595: 

596: \end{document}

597: