1: \section{Quantitative analysis of density distributions}

2: \label{c:analysis}

3: 

4: The purpose of imaging and image processing is to record density

5: distributions of the atomic cloud, either trapped or during

6: ballistic expansion.  All our knowledge about the properties of

7: cold atom systems  comes from the analysis of such images.  They

8: are usually compared to the results of models of the atomic gas.

9: Some models are exact (for the ideal gas), others are

10: phenomenological or approximations.  Many important models for

11: bosonic atoms have been presented in our 1999 Varenna notes. Here

12: we discuss important models for fermions, which allow us to infer

13: properties of the system from recorded (column) density

14: distributions.

15: 

16: 

17: \subsection{Trapped atomic gases}

18: 

19: \subsubsection{Ideal Bose and Fermi gases in a harmonic trap}

20: 

21: The particles in an atom trap are isolated from the surroundings,

22: thus the atom number $N$ and total energy content $E_{\rm tot}$ of

23: the atomic cloud is fixed. However, it is convenient to consider

24: the system to be in contact with a reservoir, with which it can

25: exchange particles and energy (grand canonical ensemble). For

26: non-interacting particles with single-particle energies $E_i$, the

27: average occupation of state $i$ is

28: \begin{eqnarray}

29:     \left<n_i\right> =  \frac{1}{e^{(E_i - \mu)/k_B T} \mp 1}

30:     \label{e:BoseFermidist}

31: \end{eqnarray}

32: with the upper sign for bosons, the lower sign for fermions. These

33: are the Bose-Einstein and Fermi-Dirac distributions, respectively.

34: For a fixed number of particles $N$ one chooses the chemical potential $\mu$ such that $N

35: = \left<N\right> = \sum_i \left<n_i\right>$.

36: 

37: Let us now apply these distributions to particles confined in a

38: harmonic trap, with trapping potential

39: \begin{equation}

40: V(\vect{r}) = \frac{1}{2} m (\omega_x^2 x^2 + \omega_y^2 y^2 +

41: \omega_z^2 z^2) \label{e:potential}

42: \end{equation}

43: We assume that the thermal energy $k_B T \equiv 1/\beta$ is much

44: larger than the quantum mechanical level spacings $\hbar

45: \omega_{x,y,z}$ (Thomas-Fermi approximation). In this case, the

46: occupation of a phase space cell

47: $\left\{\vect{r},\vect{p}\right\}$ (which is the phase-space

48: density times $h^3$) is given by Eq.~\ref{e:BoseFermidist}

49: \begin{eqnarray}

50: f(\vect{r},\vect{p}) = \frac{1}{e^{(\frac{\vect{p}^2}{2m} +

51: V(\vect{r}) - \mu)/k_B T} \mp 1} \label{e:FermiBose}

52: \end{eqnarray}

53: The density distribution of the thermal gas is

54: \begin{eqnarray}

55: n_{th}(\vect{r}) &=& \Intp{p}\; f(\vect{r},\vect{p})\nonumber\\

56: &=& \pm \frac{1}{\lambda_{ dB}^3}\, {\rm Li}_{3/2}\left(\pm

57: e^{\beta\left(\mu - V(\vect{r})\right)}\right) \label{e:density}

58: \end{eqnarray}

59: where $\sqrt{\frac{2\pi \hbar^2}{m k_B T}}$ is the de Broglie

60: wavelength. ${\rm Li}_n(z)$ is the $n^{th}$-order

61: Polylogarithm, defined as

62: \begin{eqnarray}

63:     {\rm Li}_n(z)\; \equiv\; \frac{1}{\pi^n} \int {\rm d}^{2n}r \frac{1}{e^{\vect{r}^2}/z - 1}\; \stackrel{n\ne 0}{=}\; \frac{1}{\Gamma(n)}\int_0^\infty {\rm d}q \frac{q^{n-1}}{e^q/z - 1}

64: \label{e:polylog}

65: \end{eqnarray}

66: where the first integral is over $2n$ dimensions, $\vect{r}$ is

67: the radius vector in $2n$ dimensions, $n$ is any positive

68: half-integer or zero and $\Gamma(n)$ is the Gamma-function

69: \footnote{The Polylogarithm appears naturally in integrals over

70: Bose-Einstein or Fermi-Dirac distributions.  Some

71: authors~\cite{huan87} use different functions for bosons

72: $g_n(z)={\rm Li}_n(z)$ and for fermions

73: $f_n(z)=-{\rm Li}_n(-z)$.  The Polylogarithm can be

74: expressed as a sum ${\rm Li}_n(z) = \sum_{k=1}^\infty

75: \frac{z^k}{k^n}$ which is often used as the definition of

76: the Polylogarithm.  This expression is valid for all complex

77: numbers $n$ and $z$ where $|z|\le 1$. The definition

78: given in the text is valid for all $z\le l$.

79: 

80: Special cases: ${\rm Li}_0(z) = \frac{1}{1/z - 1}$,

81: ${\rm Li_1}(z) = -\ln(1-z)$. $f(\vect{r},\vect{p})$ can

82: be written as $\pm{\rm Li}_0(\pm

83: \exp[\beta(\mu-\frac{\vect{p}^2}{2m} - V(\vect{r}))])$. When integrating density distributions to obtain column densities, a useful formula is:

84: \begin{equation}

85:     \int_{-\infty}^\infty dx \;{\rm Li}_n(z\,e^{- x^2}) = \sqrt{\pi}\; {\rm Li}_{n+1/2}(z)

86:     \label{e:polyintegral}.

87: \end{equation}

88:  Limiting values: ${\rm Li}_n(z) \stackrel{z \ll 1}{\rightarrow} z$ and $-{\rm Li}_n(-z) \stackrel{z\rightarrow\infty}{\rightarrow} \frac{1}{\Gamma(n+1)}\; \ln^n(z)$.}.

89: Note that expression~\ref{e:density} is correct for any potential

90: $V(\vect{r})$. The constraint on the number of thermal particles

91: is

92: \begin{equation}

93: N_{th} = \Int{r} \; n_{th}(\vect{r})

94: \end{equation}

95: For a harmonic potential (~\ref{e:potential}), we obtain

96: \begin{equation}

97: N_{th} = \pm \left(\frac{k_B T}{\hbar \bar{\omega}}\right)^3 {\rm

98: Li}_3(\pm\,e^{\beta\mu}) \label{e:numberofatoms}

99: \end{equation}

100: with $\bar{\omega} = (\omega_x \omega_y \omega_z)^{1/3}$ the

101: geometric mean of the trapping frequencies.

102: 

103: In the classical limit at high temperature, we recover the

104: Maxwell-Boltzmann result of a gaussian distribution,

105: \begin{equation}

106:     n_{cl}(\vect{r}) = \frac{N}{\pi^{3/2} \sigma_x \sigma_y \sigma_z} e^{- \sum_i x_i^2/\sigma_{x_i}^2} \qquad {\rm with} \; \sigma_{x,y,z}^2 = \frac{2 k_B T}{m \omega_{x,y,z}^2}

107: \end{equation}

108: 

109: The regime of quantum degeneracy is reached when $\lambda_{dB}

110: \approx n ^{-1/3}$, or when the temperature $T \approx T_{\rm

111: deg}$.  The  degeneracy temperature $T_{\rm deg} =

112: \frac{\hbar^2}{2m k_B} n^{2/3}$ is around or below one $\mu \rm K$

113: for typical experimental conditions.

114: 

115: For {\bf bosons}, it is at this point that the ground state

116: becomes macroscopically occupied and the condensate forms. The

117: density profile of the ideal gas condensate is given by the square

118: of the harmonic oscillator ground state wave function:

119: \begin{equation}

120: n_c(\vect{r}) = \frac{N_0}{\pi^{3/2} d_x d_y d_z} e^{-\sum_i

121: x_i^2/d_{x_i}^2}

122: \end{equation}

123: where $d_{x_i} = \sqrt{\frac{\hbar}{m \omega_{x_i}}}$ are the

124: harmonic oscillator lengths. The density profile of the thermal,

125: non-condensed component can be obtained from Eq.~\ref{e:density}

126: if the chemical potential $\mu$ is known. As the number of

127: condensed bosons $N_0$ grows to be significantly larger than 1,

128: the chemical potential $\mu \approx - \frac{k_B T}{N_0}$ (from

129: Eq.~\ref{e:BoseFermidist} for $E_0 = 0$) will be much closer to

130: the ground state energy than the first excited harmonic oscillator

131: state. Thus we set $\mu = 0$ in the expression for the

132: non-condensed density $n_{th}$ and number $N_{th}$ and obtain

133: \begin{eqnarray}

134: n_{th}(\vect{r}) &=& \frac{1}{\lambda_{dB}^3} {\rm Li}_{3/2}(e^{-V(\vect{r})/k_B T})\\

135: N_{th} &=& N (T/T_C)^3\qquad \mbox{for $T<T_C$}

136: \end{eqnarray}

137: with the critical temperature for Bose-Einstein condensation in a

138: harmonic trap

139: \begin{equation}

140: T_C \equiv \hbar \bar{\omega}\; (N / \zeta(3))^{1/3} = 0.94 \;

141: \hbar \bar{\omega} N^{1/3}

142: \end{equation}

143: where $\zeta(3) = {\rm Li}_3(1) \approx 1.202$. At $T=T_C$, the

144: condition for Bose condensation is fulfilled in the center of the

145: trap, $n = {\rm Li}_{3/2}(1)/\lambda_{dB}^3 =

146: 2.612/\lambda_{dB}^3$. For lower temperatures, the maximum density

147: of the thermal cloud is ``quantum saturated'' at the critical value

148: $n_{th} = 2.612/\lambda_{dB}^3 \propto T^{3/2}$. The condensate

149: fraction in a harmonic trap is given by

150: \begin{equation}

151: N_0/N = 1 - (T/T_C)^3

152: \end{equation}

153: For $T/T_C = 0.5$ the condensate fraction is already about 90\%.

154: 

155: For {\bf fermions}, the occupation of available phase space cells

156: smoothly approaches unity without any sudden transition:

157: \begin{equation}

158:   f(\vect{r},\vect{p}) = \frac{1}{e^{(\frac{\vect{p}^2}{2m} + V(\vect{r}) - \mu)/k_B T} + 1} \stackrel{T \rightarrow 0} \rightarrow \left\{%

159: \begin{array}{ll}

160:     1, & \hbox{$\frac{\vect{p}^2}{2m} + V(\vect{r}) < \mu$} \\

161:     0, & \hbox{$\frac{\vect{p}^2}{2m} + V(\vect{r}) > \mu$} \\

162: \end{array}%

163: \right. \label{e:fermiphasespace}

164: \end{equation}

165: Accordingly, also the density profile changes smoothly from its

166: gaussian form at high temperatures to its zero temperature shape:

167: \begin{eqnarray}

168: n_F(\vect{r}) &=& \Intp{p} \, f(\vect{r},\vect{p}) \stackrel{T\rightarrow 0}{\rightarrow} \int_{\left|\vect{p}\right|< \sqrt{2m(\mu-V(\vect{r}))}} \frac{{\rm d}^3\vect{p}}{(2\pi\hbar)^3}\nonumber\\

169: &=& \frac{1}{6\pi^2} \left(\frac{2m}{\hbar^2}\right)^{3/2}

170: \left(\mu - V(\vect{r})\right)^{3/2}.

171: \end{eqnarray}

172: 

173: From Eq.~\ref{e:fermiphasespace} we observe that at zero

174: temperature, $\mu$ is the energy of the highest occupied state of

175: the non-interacting Fermi gas, also called the Fermi energy $E_F$.

176: The (globally) largest momentum is $p_F \equiv \hbar k_F \equiv

177: \sqrt{2 m E_F}$, the Fermi momentum. {\it Locally}, at position

178: $\vect{r}$ in the trap, it is $p_F(\vect{r}) \equiv \hbar

179: k_F(\vect{r}) \equiv \sqrt{2 m \epsilon_F(\vect{r})} \equiv \hbar

180: (6\pi^2 n_F(\vect{r}))^{1/3}$ with the local Fermi energy

181: $\epsilon_F(\vect{r})$ which equals $\mu(\vect{r},T=0) = E_F -

182: V(\vect{r})$. The value of $E_F$ is fixed by the number of

183: fermions $N$, occupying the $N$ lowest energy states of the trap.

184: For a harmonic trap we obtain

185: \begin{eqnarray}

186:     N &=& \Int{r} \; n_F(\vect{r}) = \frac{1}{6} \left(\frac{E_F}{\hbar \bar{\omega}}\right)^3\nonumber\\

187: \Rightarrow    E_F &=& \hbar \bar{\omega} (6 N)^{1/3}

188:     \label{e:Ferminumber}

189: \end{eqnarray}

190: and for the zero-temperature profile

191: \begin{eqnarray}

192: n_F(\vect{r}) &=& \frac{8}{\pi^2} \frac{N}{R_{Fx} R_{Fy} R_{Fz}}

193: \; \left[\max \left(1 - \sum_i

194: \frac{x_i^2}{R_{Fi}^2},0\right)\right]^{3/2}

195: \label{e:Fermidensity}

196: \end{eqnarray}

197: with the Fermi radii $R_{F{x,y,z}} = \sqrt{\frac{2 E_F}{m

198: \omega_{x,y,z}^2}}$. The profile of the degenerate Fermi gas has a

199: rather flat top compared to the gaussian profile of a thermal

200: cloud, as the occupancy of available phase space cells saturates

201: at unity.

202: 

203: At finite $T \lesssim T_F$, we can understand the shape of the

204: cloud by comparing $k_B T$ with the local Fermi energy

205: $\epsilon_F(\vect{r})$. For the outer regions in the trap where $k_B

206: T \gg \epsilon_F(\vect{r})$, the gas shows a classical (Boltzmann)

207: density distribution $n(\vect{r}) \propto e^{-\beta V(\vect{r})}$. In

208: the inner part of the cloud where $k_B T \ll

209: \epsilon_F(\vect{r})$, the density is of the zero-temperature form $n(\vect{r})

210: \propto (E_F - V(\vect{r}))^{3/2}$. The Polylogarithm smoothly

211: interpolates between the two regimes. We notice here the

212: difficulty of thermometry for very cold Fermi clouds: Temperature

213: only affects the far wings of the density distribution. While for

214: thermal clouds above $T_F$, the size of the cloud is a direct

215: measure of temperature, for cold Fermi clouds one needs to extract

216: the temperature from the shape of the distribution's wings.

217: 

218: Note that the validity of the above derivation required the Fermi

219: energy $E_F$ to be much larger than the level spacing $\hbar

220: \omega_{x,y,z}$. For example, in very elongated traps and for low

221: atom numbers one can have a situation where this condition is

222: violated in the tightly confining radial dimensions.

223: 

224: \subsubsection{Trapped, interacting Fermi mixtures at zero temperature}

225: \label{s:trappedmixtures}

226: \begin{figure}

227:     \centering

228:     \includegraphics[width=4in]{figs_quantanalysis/phasediagram.eps}

229:     \caption{Phase diagram of interacting Fermi mixtures in a

230:     harmonic trap, as a function of temperature and interaction

231:     strength $1/k_F a$. Shown is the critical temperature $T_C$

232:     for the formation of a superfluid as a function of $1/k_F a$

233:     (full line) as well as the characteristic temperature $T^*$ at

234:     which fermion pairs start to form (dashed line),

235:     after~\cite{pera04temp}. The shading indicates that pair

236:     formation is a smooth process, not a phase transition.}

237:     \label{f3:phasediagram}

238: \end{figure}

239: 

240: We now consider the case of $N$ fermionic atoms equally populating

241: two hyperfine states (``spin up'' and ``spin down''). Atoms in

242: different spin states interact via $s$-wave collisions

243: characterized by the scattering length $a$. A dimensionless

244: parameter measuring the strength and sign of the interaction

245: strength is $1/k_F a$, essentially the ratio of the interparticle

246: spacing to the scattering length. For weak attractive

247: interactions, $1/k_F a \rightarrow -\infty$, the ground state of

248: the system is a BCS superfluid (see chapter~\ref{c:BECBCStheory}).

249: As the magnitude of the scattering length increases to a point

250: where $a \rightarrow \mp \infty$ diverges (thus $1/k_F a

251: \rightarrow 0$), a  two-body molecular bound state enters the

252: interparticle potential. For weak repulsive interactions, $1/k_F a

253: \rightarrow +\infty$, the ground state of the system is then a

254: Bose-Einstein condensate of weakly-interacting molecules of mass

255: $M = 2m$, in which two fermions of opposite spin are tightly

256: bound.

257: 

258: Fig.~\ref{f3:phasediagram} summarizes the different regimes within

259: this BEC-BCS crossover. We see that the character of the

260: Fermi mixture drastically changes as a function of temperature and

261: interaction strength. For temperatures $T \gg T^*$ fermions are

262: unpaired, and a free Fermi mixture exists on the BEC- and the

263: BCS-side of the phase diagram. On resonance, the mixture might

264: still be strongly interacting even at high temperatures, thus

265: possibly requiring an effective mass description of the

266: interacting gas. The density distribution will have the same shape

267: as a free Fermi gas at all interaction strengths. Below $T^*$,

268: fermion pairs start to form. On the BEC-side, where fermions are

269: tightly bound, the thermal distribution should now be that of a

270: gas of bosons with mass $M = 2m$. As a consequence, the cloud will

271: shrink. Below $T_C$, we will finally observe a superfluid,

272: condensed core, surrounded by a thermal cloud of molecules in the

273: BEC-limit, or of unpaired fermions in the BCS-limit.

274: 

275: In general, the calculation of density

276: distributions in the strongly interacting regime is a difficult

277: affair. Simple expressions for the densities can be derived for

278: superfluid gases at zero temperature, for molecular gases on the

279: ``BEC''-side at large and positive $1/k_F a$, for weakly interacting

280: Fermi gases on the ``BCS''-side for large and negative $1/k_F a$,

281: and in the classical limit at high temperatures.

282: 

283: \paragraph{BEC limit}

284: 

285: The molecular Bose-Einstein condensate is described by a many-body

286: wave function $\psi(\vect{r})$ which obeys the {\rm

287: Gross-Pitaevskii equation}~\cite{peth02bec}

288: \begin{equation}

289: \left(-\frac{\hbar^2 \nabla^2}{2M} + V_M(\vect{r}) + g

290: \left|\psi(\vect{r},t)\right|^2\right) \psi(\vect{r},t) = i \hbar

291: \frac{\partial}{\partial t} \psi(\vect{r},t)

292: \label{e:grosspitaevskii}

293: \end{equation}

294: where $V_M(\vect{r})$ is the trapping potential experienced by the

295: molecules, and $g = \frac{4\pi \hbar^2 a_M}{M}$ describes the

296: intermolecular interactions. We can identify $\left|\psi\right|^2$

297: with the condensate density $n_c$, which for weak interactions and

298: at zero temperature equals the density of molecules $n_M$. The

299: validity of Eq.~\ref{e:grosspitaevskii} is limited to {\it weakly

300: interacting} gases of molecules, for which the gas parameter $n_M

301: a_M^3 \approx (\frac{k_F a}{6.5})^3 \ll 1$. In typical experiments

302: on BECs of bosonic atoms, the corresponding condition is very well

303: fulfilled. For a sodium BEC with $n \approx 10^{14} \, \rm

304: cm^{-3}$ and $a = 3.3\,\rm nm$, we have $n a^3 \approx 4 \cdot

305: 10^{-6}$. However, for molecular condensates near a Feshbach

306: resonance,  this condition can be easily violated (see

307: chapter~\ref{c:expobservation}).

308: 

309: In equilibrium, the ground-state wave function is $\psi(\vect{r},t)

310: =  e^{-i \mu_M t/\hbar} \psi(\vect{r})$, where $\mu_M$ is the

311: ground state energy and is identified with the molecular chemical

312: potential, and $\psi(\vect{r})$ is a solution of the stationary

313: Gross-Pitaevskii equation

314: \begin{equation}

315: \left(-\frac{\hbar^2 \nabla^2}{2M} + V_M(\vect{r}) + g

316: \left|\psi(\vect{r})\right|^2\right) \psi(\vect{r}) = \mu_M

317: \psi(\vect{r}) \label{e:statgrosspitaevskii}

318: \end{equation}

319: In the {\it ideal gas limit}, $g n_c \ll \hbar \omega_{x, y, z}$,

320: we recover the harmonic oscillator result for the condensate's

321: density distribution $n_c(\vect{r})$. In the {\it Thomas-Fermi

322: limit}, on the other hand, interactions dominate over the kinetic

323: energy of the condensate wave function, $g n_c \gg \hbar

324: \omega_{x,y,z}$. Already for weakly interacting alkali gases, this

325: condition is very well fulfilled, with typical interaction

326: energies of $g n_c \sim k_B \times 150 \,\rm nK$ and

327: $\hbar\omega_r \approx k_B \times 5 \,\rm nK$. In this

328: approximation we obtain the condensate density $n_c(\vect{r}) =

329: \left|\psi(\vect{r})\right|^2$:

330: \begin{equation}

331:     n_c(\vect{r}) = \max \left(\frac{\mu_M - V_M(\vect{r})}{g},0\right)

332: \end{equation}

333: Thus, a condensate in the Thomas-Fermi approximation ``fills in''

334: the bottom of the trapping potential up to an energy $\mu_M$,

335: which is determined by the total number of molecules, $N_M = N/2 =

336: \Int{r} \; n_c(\vect{r})$. Taking $V_M(\vect{r}) = 2 V(\vect{r})$

337: with the harmonic trapping potential for single atoms in

338: Eq.~\ref{e:potential}, one obtains a parabolic density profile,

339: \begin{equation}

340:     n_c(\vect{r}) = \frac{15}{8\pi} \frac{N_M}{R_x R_y R_z} \max \left(1 - \sum_i \frac{x_i^2}{R_i^2},0\right)

341:     \label{e:BECparabola}

342: \end{equation}

343: where the {\it Thomas-Fermi radii} $R_i = \sqrt{\frac{2 \mu_M}{M

344: \omega_i^2}}$ give the half-lengths of the trapped condensate

345: where the density vanishes. The chemical potential is given by

346: \begin{equation}

347:     \mu_M = \frac{1}{2} \hbar \bar{\omega} \left(\frac{15 N_M a_M}{\bar{d}_{h.o.}}\right)^{2/5}

348:     \label{e:chempot}

349: \end{equation}

350: where $\bar{d}_{h.o.} = (d_x d_y d_z)^{1/3} = \sqrt{\hbar/M

351: \bar{\omega}}$ is the geometric mean of the harmonic oscillator

352: lengths for molecules.

353: 

354: Interactions thus have a major effect on the shape of the

355: Bose-Einstein condensate, changing the density profile from the

356: gaussian harmonic oscillator ground state wave function to a broad

357: parabola, as a result of the interparticle repulsion. The

358: characteristic size of the condensate is no longer given by the

359: harmonic oscillator length but by the generally much larger

360: Thomas-Fermi radius $R_{x,y,z} = d_{x,y,z}

361: \sqrt{\frac{\bar{\omega}}{\omega_{x,y,z}}} \left(\frac{15 N_M

362: a_M}{\bar{d}_{h.o.}}\right)^{1/5}$. Also the aspect ratio changes,

363: for example in the $x$-$y$ plane from $\frac{d_x}{d_y} =

364: \sqrt{\frac{\omega_y}{\omega_x}}$ to $\frac{R_x}{R_y} =

365: \frac{\omega_y}{\omega_x}$. Nevertheless, weakly interacting

366: condensates are still considerably smaller in size than a thermal

367: cloud at $k_B T > \mu_M$, and more dense. This leads to the clear

368: separation between the dense condensate in the center of the cloud

369: and the large surrounding thermal cloud, the ``smoking gun'' for

370: Bose-Einstein condensation (both in the trapped and in the

371: expanding cloud, see section~\ref{s:expansion} below). In the case

372: of strong interactions, when the chemical potential $\mu_M$

373: becomes comparable to $k_B T_C$, this direct signature of

374: condensation will be considerably weaker. In this regime we also

375: have to account for the mutual repulsion between the thermal cloud

376: and the condensate (see section~\ref{s:molecularclouds} below).

377: 

378: \paragraph{BCS limit}

379: 

380: In the weakly interacting BCS limit ($1/k_F a \rightarrow

381: -\infty$), pairing of fermions and superfluidity have very small

382: effects on the density profile of the gas. The sharp Fermi surface

383: in $k$-space at $k_F$ is modified only in an exponentially narrow

384: region of width $\sim k_F \exp(-\frac{\pi}{2 k_F |a|})$. The

385: density, i.e. the integral over occupied $k$-states, is thus

386: essentially identical to that of a non-interacting Fermi gas. The

387: result is Eq.~\ref{e:Fermidensity} with the number of spin-up

388: (spin-down) atoms $N_{\uparrow,\downarrow} = N/2$ and Fermi energy

389: $E_{F} = \hbar \bar{\omega} (6

390: N_{\uparrow,\downarrow})^{1/3} = \hbar \bar{\omega} (3 N)^{1/3}$.

391: As one approaches the strongly interacting regime $1/k_F a \approx

392: -1$, it is conceivable that the formation of the superfluid leaves

393: a distinct trace in the density profile of the gas, as this is the

394: situation in the BEC-limit, and the crossover between the two

395: regimes is smooth. Indeed, several theoretical studies have

396: predicted kinks in the density profiles signalling the onset of

397: superfluidity~\cite{holl01,pera04temp,ho04uni,staj05dens}. We were able to observe

398: such a direct signature of condensation on resonance ($1/k_F a =

399: 0$) and on the BCS-side ($1/k_F a < 0$) in unequal Fermi mixtures

400: (see section~\ref{s:imbalance}). In equal mixtures, we detected a

401: faint but distinct deviation from the Thomas-Fermi profile on

402: resonance (see~\ref{s:directunitarity}).

403: 

404: \paragraph{Unitarity}

405: \label{s:unitarity}

406: 

407: The regime on resonance ($1/k_F a = 0$) deserves special

408: attention. The scattering length diverges and leaves the

409: interparticle distance $n^{-1/3} \sim 1/k_F$ as the only relevant

410: length scale. Correspondingly, the only relevant energy scale is

411: the Fermi energy $\epsilon_F = \hbar^2 k_F^2 / 2 m$. The regime is

412: thus said to be universal. The chemical potential $\mu$ can then

413: be written as a universal constant times the Fermi energy: $\mu =

414: \xi \epsilon_F$. In the trapped case, we can use this relation

415: locally  (local density approximation) and relate the local

416: chemical potential $\mu(\vect{r}) = \mu - V(\vect{r})$ to the

417: local Fermi energy $\epsilon_F(\vect{r}) \equiv \hbar^2

418: k_F(\vect{r})^2 / 2 m \equiv \frac{\hbar^2}{2 m} (6\pi^2

419: n_{\uparrow}(\vect{r}))^{2/3}$, where $n_{\uparrow}(\vect{r})$ is the density of atoms in one spin state. We then directly obtain a relation for the density profile $n_{\uparrow U}(\vect{r})$ of the unitary Fermi gas:

420: $$n_{\uparrow U}(\vect{r}) = \frac{1}{6\pi^2} \left(\frac{2m}{\xi \hbar^2}\right)^{3/2} \left(\mu - V(\vect{r})\right)^{3/2}.$$

421: The constraint from the number of particles in spin up, $N_\uparrow = N/2$, determines $\mu$:

422: \begin{eqnarray}

423:     N_\uparrow &=& \Int{r} \; n_{\uparrow U}(\vect{r}) = \frac{1}{6} \left(\frac{\mu}{\sqrt{\xi} \hbar \bar{\omega}}\right)^3\nonumber\\

424: \Rightarrow    \mu &=& \sqrt{\xi} E_F.

425: \end{eqnarray}

426: The density profile becomes

427: \begin{eqnarray}

428: n_{\uparrow U}(\vect{r}) &=& \frac{8}{\pi^2} \frac{N_\uparrow}{R_{Ux} R_{Uy} R_{Uz}}

429: \; \left[\max \left(1 - \sum_i

430: \frac{x_i^2}{R_{Ui}^2},0\right)\right]^{3/2}

431: \label{e:Unitaritydensity}

432: \end{eqnarray}

433: with the radii $R_{U{x,y,z}} = \xi^{1/4} R_{F{x,y,z}}$.

434: Table~\ref{tab:zerotemptrappeddensity} summarizes the various

435: density profiles of interacting Fermi mixtures.

436: 

437: Remarkably, the functional form $n_{\uparrow U}(\vect{r}) \propto

438: (\mu-V(\vect{r}))^{3/2}$ is identical to that of a non-interacting

439: Fermi gas.  The underlying reason is that the equation of state

440: $\mu \propto n^{2/3}$ has the same power-law form as for

441: non-interacting fermions. The universal constant $\xi$ simply

442: rescales the radii (by a factor $\xi^{1/4}$) and the central

443: density (by a factor $\xi^{-3/4}$).  One thus has direct

444: experimental access to the universal constant $\xi$ by measuring

445: the size of the cloud at unitarity (see section~\ref{s:energymeasurements}).

446: 

447: \begin{table}

448:   \centering

449:   \begin{tabular}{p{0.02\linewidth}c|c|c|cp{0.3\linewidth}}

450:     % after \\: \hline or \cline{col1-col2} \cline{col3-col4} ...

451: &     &  BEC-limit  &   Unitarity  & BCS-limit \\

452: &$\frac{1}{k_F a}$ & $\infty$ & 0& $-\infty$\\ \hline

453: &    $\gamma$ (in $\mu \propto n^\gamma$) & 1 & 2/3 & 2/3 \\

454: &    $n_\uparrow(\vect{r})/n_\uparrow(\vect{0})$ & $1 - \sum_i \frac{x_i^2}{R_{i}^2}$ & $(1 - \sum_i \frac{x_i^2}{R_{Ui}^2})^{3/2}$ & $(1 - \sum_i \frac{x_i^2}{R_{Fi}^2})^{3/2}$ \\

455: &$n_\uparrow(\vect{0})$ & $\frac{15}{8\pi} \frac{N_\uparrow}{R_x R_y R_z}$ &

456: $\frac{8}{\pi^2}  \frac{N_\uparrow}{R_{Ux} R_{Uy} R_{Uz}}$ &

457: $\frac{8}{\pi^2}  \frac{N_\uparrow}{R_{Fx} R_{Fy} R_{Fz}}$ \\

458: &Radii & $R_i = \sqrt{\frac{2 \mu_M}{M \omega_i^2}}$ & $R_{Ui} =

459: \xi^{1/4} R_{Fi}$ & $R_{Fi} = \sqrt{\frac{2 E_F}{m \omega_i^2}}$

460: \end{tabular}

461:   \caption{Zero-temperature density profiles of a trapped, interacting Fermi mixture in the BEC-BCS crossover. The density is zero when the expressions are not positive. For definitions see the text.}

462:   \label{tab:zerotemptrappeddensity}

463: \end{table}

464: 

465: 

466: \subsection{Expansion of strongly interacting Fermi mixtures}

467: \label{s:expansion}

468: 

469: Intriguingly rich physics can be uncovered by the simple release

470: of ultracold gases from their confining trap. From the size of the

471: expanded cloud and the known time-of-flight one directly obtains

472: the energy content of the gas: the temperature in the case of

473: thermal clouds, the Fermi energy for non-interacting degenerate

474: Fermi gases, the mean-field energy for Bose-Einstein condensates.

475: In the case of free ballistic expansion, where no collisions occur

476: during expansion, the density distribution of the expanded cloud

477: directly reveals the original momentum distribution of the

478: particles in the trap. Thermal clouds will become spherical after

479: ballistic expansion, reflecting their isotropic momentum

480: distribution in the trap. The expansion of Bose-Einstein

481: condensates is not ballistic but mean-field driven, leading to

482: superfluid hydrodynamic expansion. As mean-field energy is

483: preferentially released in the direction(s) of tight confinement,

484: this allows for the famous ``smoking gun'' signature of

485: Bose-Einstein condensation: inversion of the condensate's aspect

486: ratio after expansion out of an anisotropic trap. In strongly

487: interacting gases the normal, uncondensed cloud can be

488: collisionally dense, and will expand according to classical

489: hydrodynamics. As particles will preferentially leave the cloud

490: along the narrower dimensions, where they undergo fewer

491: collisions, this also leads to an inversion of the cloud's initial

492: aspect ratio. It is thus no longer a ``smoking gun'' for

493: condensation, but merely for strong interactions. Expansion is

494: also useful to measure correlations in momentum

495: space~\cite{grei05corr}. Finally, in the case of harmonic

496: trapping, expansion of a superfluid cloud can often be described

497: as a ``magnifying glass'', a mere scaling of the density

498: distribution in the trap. This allows for example to observe

499: quantized vortices~\cite{zwie05vort}, which are too small to be

500: observable in the trap.  In this section, we show how quantitative

501: information can be derived from images of expanding clouds.

502: 

503: 

504: \subsubsection{Free ballistic expansion}

505: \label{s:ballistic}

506: 

507: Let us consider the expansion of a non-condensed thermal cloud. If

508: the mean free path $\lambda_c$ between collisions is longer than

509: the size of the trapped cloud $R$, we can neglect collisions

510: during expansion, which is hence ballistic. The collision rate is

511: $\Gamma = n \sigma v$, with density $n$, collisional cross section

512: $\sigma$ and thermal (root mean square) velocity $v$, which gives

513: $\lambda_c = v/\Gamma = 1/n \sigma$. As $R = v/\omega$ for a

514: harmonic trap, the condition $\lambda_c \gg R$ is equivalent to

515: having $\Gamma \ll \omega$, that is, the mean time interval

516: between collisions should be larger than a period of oscillation

517: in the trap.

518: 

519: This condition can be fulfilled for the cloud of uncondensed

520: molecules in the BEC limit where $1/k_F a \gg 1$ and collisions

521: are negligible (this has been the case also for atomic BECs with

522: the exception of very large thermal clouds,

523: see~\cite{stam98coll,shva03hydro}), and for the cloud of unpaired

524: fermions in the BEC- and in the BCS-limit for $k_F |a| \ll 1$ (the

525: exact criterion is still $\Gamma \ll \omega$). For molecules with

526: mass $M$, we need to replace $m \rightarrow M$ in the following

527: discussion.

528: 

529: In the ballistic case, a particle initially at point $\vect{r}_0$

530: in the trap, will reach point $\vect{r} = \vect{r}_0 +

531: \frac{\vect{p}_0}{m} t$ after expansion time $t$. We obtain the

532: density at point $\vect{r}$ at time $t$ by adding the

533: contributions from particles at all points $\vect{r}_0$ that had

534: the correct initial momentum $\vect{p}_0 = m(\vect{r} -

535: \vect{r}_0)/t$. In terms of the semi-classical distribution

536: $f(\vect{r},\vect{p})$, Eq.~\ref{e:FermiBose}, this is

537: \begin{eqnarray}

538: n(\vect{r},t) &=& \Int{r}_0 \Intp{p_0} \; f(\vect{r_0},\vect{p_0})\; \delta\left(\vect{r} - \vect{r_0} - \frac{\vect{p}_0}{m}t\right) \nonumber\\

539: &=& \Intp{p_0}\; f\left(\vect{r}-\frac{\vect{p}_0}{m}t,\vect{p}_0\right) \nonumber\\

540: &=& \Intp{p_0}\; \left\{\exp\left[\beta\frac{\vect{p}_0^2}{2m} +

541: \beta V\left(\vect{r}-\frac{\vect{p}_0}{m}t\right) - \beta\mu\right] \mp

542: 1\right\}^{-1} \label{e:expandeddensity}

543: \end{eqnarray}

544: The integral can be carried out analytically in the case of a

545: harmonic potential (Eq.~\ref{e:potential}):

546: \begin{eqnarray}

547: n(\vect{r},t) &=& \Intp{p_0}\; \left\{\exp\left[\beta\sum_i \left(1+\omega_i^2 t^2\right) \frac{p_{0i}^2}{2m} +\beta\sum_i\frac{1}{2}m\frac{\omega_i^2 x_i^2}{1+\omega_i^2 t^2} - \beta\mu\right] \mp 1\right\}^{-1} \nonumber \\

548: &=& \pm \frac{1}{\lambda_{dB}^3}\prod_i

549: \frac{1}{\sqrt{1+\omega_i^2 t^2}}\; {\rm Li}_{3/2}\left[\pm

550: \exp\left(\beta\mu - \beta\sum_i\frac{1}{2}m\frac{\omega_i^2

551: x_i^2}{1+\omega_i^2 t^2}\right)\right] \nonumber

552: \end{eqnarray}

553: Note that this has the same form as the density distribution in

554: the trap, but with spatial dimension $i={x,y,z}$ rescaled by the

555: factor $b_i(t) = \sqrt{1 + \omega_i^2 t^2}$. Ballistic expansion

556: of a thermal (bosonic or fermionic) cloud from a harmonic trap is

557: thus a scaling transformation:

558: \begin{eqnarray}

559:     n(\vect{r},t) = \frac{1}{\mathcal{V}(t)}\;n\left(\frac{x}{b_x(t)},\frac{y}{b_y(t)},\frac{z}{b_z(t)},t =0\right)

560: \label{e:scalingtrafo}

561: \end{eqnarray}

562: where the unit volume scales as $\mathcal{V}(t) = b_x b_y b_z$.

563:  After an expansion time long compared to the trapping periods ($t \gg 1/\omega_i$), we have

564: \begin{eqnarray}

565: n(\vect{r},t \gg 1/\omega_i) &=& \pm \frac{1}{\lambda_{

566: dB}^3}\frac{1}{(\bar{\omega} t)^3}\, {\rm Li}_{3/2}\left[\pm

567: \exp\left(\beta\mu -

568: \beta\frac{1}{2}m\frac{\vect{r}^2}{t^2}\right)\right]

569: \label{e:expdensitylongtimes}

570: \end{eqnarray}

571: As expected, we obtain an isotropic density profile, reflecting

572: the original isotropic momentum distribution of the trapped gas.

573: Importantly, the {\it shape} of the density profile, i.e. its

574: variation with $\vect{r}$, becomes insensitive to the trapping

575: potential. Eq.~\ref{e:expdensitylongtimes} thus holds for a

576: general trapping geometry, for expansion times long compared to

577: the longest trapping period. Even if the trapping potential is not

578: known in detail, one can still determine the cloud's temperature

579: and even decide whether the gas is degenerate. Note that the {\it

580: momentum distribution} at point $\vect{r}$ after long expansion

581: times $t \gg 1/\omega_i$ has become {\it anisotropic}:

582: \begin{eqnarray}

583: f(\vect{r},\vect{p_0},t) &=& \Int{r}_0 \; f(\vect{r_0},\vect{p_0})\; \delta\left(\vect{r} - \vect{r_0} - \frac{\vect{p}_0}{m}t\right) \nonumber\\

584: &=&  f\left(\vect{r}-\frac{\vect{p}_0}{m}t,\vect{p}_0\right) \nonumber\\

585: \stackrel{t \gg 1/\omega_i}{\rightarrow} &=&

586: \left(\exp\left[\beta\left(\sum_i \omega_i^2 t^2

587: \frac{\left(p_{0i} - m \frac{x_i}{t}\right)^2}{2m}

588: +\frac{1}{2}\,m\frac{\vect{r}^2}{t^2} - \mu\right)\right] \mp

589: 1\right)^{-1} \label{e:ballisticmomentum}

590: \end{eqnarray}

591: The momentum distribution at point $\vect{r}$ is ellipsoidal,

592: centered at $\bar{\vect{p}} = m\frac{\vect{r}}{t}$, and with

593: characteristic widths $\Delta p_i \propto m\frac{\Delta x_i}{t}

594: \propto \frac{1}{\omega_i}$ directly mirroring the ellipsoidal

595: atomic distribution in the trap.

596: 

597: \paragraph{Ballistic expansion into a saddle potential}

598: \label{s:saddleballistic}

599: 

600: In many experiments, atoms are released from an optical trap, but

601: magnetic fields (Feshbach fields) are still left on.  In general,

602: these magnetic fields are inhomogeneous, either due to technical

603: limitations, or deliberately, e.g. in case of the optical-magnetic

604: hybrid trap discussed in section~\ref{s:opticaltrap}.  We focus here on the

605: important case of a magnetic field created by pair of coils which

606: generates a saddle point potential.

607: 

608: So we assume that at $t>0$, the gas is not released into free

609: space, but into a new potential.  We define $V(\vect{r},t>0) =

610: \frac{1}{2} m \left(\omega_{Sx}^2 x^2 + \omega_{Sz}^2 y^2 +

611: \omega_{Sz}^2 z^2\right)$, and can describe expansion into

612: anticonfining potentials with imaginary frequencies. For

613: example, for the magnetic saddle potentials relevant for the MIT

614: experiments, the radial dimension is anticonfining and

615: $\omega_{Sx,y} = {\rm i} \frac{1}{\sqrt{2}}\omega_{Sz}$. In the

616: potential $V(\vect{r},t>0)$, particles with initial position

617: $\vect{r_0}$ and momentum $\vect{p_0}$ will reach the point

618: $\vect{r}$ with $x_i = \cos(\omega_{Si} t) x_{0i} +

619: \frac{1}{\omega_{Si}} \sin(\omega_{Si} t) \frac{p_{0i}}{m}$ after

620: expansion time $t$. The calculation of the density profile is

621: fully analogous to the case of free expansion, after the change of

622: variables $x_{0i} \rightarrow \tilde{x}_{0i}/\cos(\omega_{Si}t)$

623: and the substitution $t \rightarrow

624: \sin(\omega_{Si}t)/\omega_{Si}$. We again obtain a scaling

625: transformation, Eq.~\ref{e:scalingtrafo}, but for this case with

626: scaling parameters $b_i(t) = \sqrt{\cos^2(\omega_{Si}t) +

627: \frac{\omega_i^2}{\omega_{Si}^2} \sin^2(\omega_{Si}t)}$. For

628: expansion into the magnetic saddle potential, this gives

629: $b_\perp(t) = \sqrt{\cosh^2(\frac{1}{\sqrt{2}}\omega_{Sz}t) +

630: \frac{2\omega_\perp^2}{\omega_{Sz}^2}

631: \sinh^2(\frac{1}{\sqrt{2}}\omega_{Sz}t)}$ and $b_z(t) =

632: \sqrt{\cos^2(\omega_{Sz}t) + \frac{\omega_z^2}{\omega_{Sz}^2}

633: \sin^2(\omega_{Sz}t)}$. For the MIT trap, the initial axial

634: trapping potential is dominated by the magnetic field curvature,

635: while the initial radial potential is almost entirely due to the

636: optical trap. After switching off the optical trap, we have

637: $\omega_{Sz} = \omega_z$ and $\omega_{Sx} = {\rm

638: i}\frac{1}{\sqrt{2}}\omega_z$. In this case, $b_z(t) = 1$ and the

639: cloud expands only into the radial direction.

640: 

641: \subsubsection{Collisionally hydrodynamic expansion}

642: \label{s:collhydro}

643: 

644: If the mean free path $\lambda_c$ is short compared to the cloud

645: size, the gas is in the hydrodynamic regime, and collisions during

646: expansion can no longer be neglected.  Collisions will tend to

647: reestablish local thermal equilibrium, in particular an isotropic

648: momentum distribution. For anisotropic traps, this directly leads

649: to anisotropic expansion, in strong contrast to the ballistic

650: case: Particles trying to escape in one direction suffer

651: collisions that redistribute their momenta equally in all

652: directions. The escape is hindered more for the weakly confined

653: directions where the cloud is long initially and particles can

654: undergo more collisions. For cylindrically symmetric clouds, this

655: leads to an inversion of the aspect ratio of the cloud during

656: expansion.

657: 

658: Hydrodynamic expansion can take place for $1/k_F |a| < 1$, which

659: includes (for $a>0$) strongly interacting clouds of

660: uncondensed molecules, and (for $a<0$) a strongly interacting,

661: normal Fermi mixture.  There is no sharp boundary between

662: molecular hydrodynamics and fermionic hydrodynamics, since $1/k_F

663: |a| < 1$ is the strongly interacting regime where many-body

664: physics dominates and the single-particle description (molecules

665: in one limit, unbound fermions in the other) is no longer valid.

666: 

667: In the hydrodynamic regime, the evolution of the gas is governed

668: by the continuity equation for the density $n(\vect{r},t)$ and,

669: neglecting friction (viscosity), the Euler equation for the

670: velocity field $\vect{v}(\vect{r},t)$:

671: \begin{eqnarray}

672: \frac{\partial n}{\partial t} + \nabla \cdot (n \vect{v}) &=& 0 \\

673: m \frac{{\rm d}\vect{v}}{{\rm d}t} = m \frac{\partial

674: \vect{v}}{\partial t} + m (\vect{v} \cdot \nabla) \vect{v} &=& -

675: \nabla V(\vect{r},t) - \frac{1}{n} \nabla P(\vect{r},t)

676: \end{eqnarray}

677: where $P$ is the pressure. Friction is

678: negligible deep in the hydrodynamic regime, when the mean free

679: path approaches zero.  The Euler equation is simply Newton's law

680: for the collection of gas particles at $\vect{r}$. In steady

681: state, we recover the equilibrium solution

682: \begin{equation}

683: \nabla P_0(\vect{r}) = n_0(\vect{r}) \nabla \mu_0(\vect{r}) = -

684: n_0(\vect{r}) \nabla V(\vect{r},0) \label{e:initialpressure}

685: \end{equation}

686: where we have used the expression for the local chemical potential

687: $\mu_0(\vect{r}) = \mu - V(\vect{r})$.

688: \paragraph{Scaling solution for harmonic potentials}

689: In the case of free expansion, the potential $V(\vect{r},t)$ is

690: the initial harmonic trapping potential for $t<0$, with radial and

691: axial trapping frequencies $\omega_{\perp}(0)$ and

692: $\omega_{z}(0)$, and zero for $t>0$. We can more generally

693: consider here an arbitrary time variation $\omega_\perp(t)$ and

694: $\omega_z(t)$ of the trapping frequencies. For this case, the

695: Euler equation allows a simple scaling solution for the

696: coordinates and velocities~\cite{kaga97bose}

697: \begin{eqnarray}

698: x_i(t) &=& b_i(t)\, x_{0i} \nonumber\\

699: v_i(t) &=& \frac{\dot{b}_i}{b_i}\, x_i(t)

700: \end{eqnarray}

701: with initial conditions $b_i(0) = 1$ and $\dot{b}_i(0) = 0$. The

702: unit volume scales as $\mathcal{V}(t) = b_x\,b_y\,b_z$, the

703: density varies as $n(\vect{r},t) = n_0(\vect{r}_0)/\mathcal{V}$,

704: where the fluid element at initial position $\vect{r}_0$ has

705: propagated to $\vect{r}$ at time $t$.

706: 

707: \paragraph{Pressure}

708: The thermodynamic properties of a simple fluid or gas only depend

709: on three variables, that are, in the grand canonical description,

710: the temperature $T$, the chemical potential $\mu$ and the volume

711: $V$. From the grand canonical partition function Z, one obtains in

712: this case the pressure $P = k_B T \frac{\ln Z}{V}$. For a

713: non-interacting, ideal gas of bosons or fermions, the average

714: energy is $E = \frac{3}{2} k_B T \ln Z$, leading to the relation

715: $PV = \frac{2}{3} E$. This equation is no longer true for an

716: interacting gas, for example the van der Waals gas. It is very

717: remarkable, then, that this relation nevertheless holds also for

718: the strongly interacting, unitary gas on resonance, for all

719: temperatures~\cite{ho04uni,thom05virial}~\footnote{On resonance,

720: universality requires that the energy $E = N \epsilon_F f(T/T_F)$

721: with a universal function $f$. Entropy can only be a function of

722: $T/T_F$, so adiabaticity requires this ratio to be constant. The

723: pressure is then $P = -\partial E/\partial V|_{S,N} = -N f(T/T_F)

724: \partial \epsilon_F/\partial V = \frac{2}{3} E/V$.}. Under an

725: adiabatic expansion, the energy $E$ changes according to ${\rm d}E

726: = - P {\rm d}V$. Hence $\frac{3}{2}{\rm d} (PV) = \frac{3}{2}

727: (V{\rm d} P + P{\rm d} V) = - P {\rm d}V$, which leads to the law

728: $P V^{5/3} = {\rm const}$ for adiabatic expansion. The pressure

729: thus scales as $\mathcal{V}^{-5/3}$, and the force, using

730: Eq.~\ref{e:initialpressure},

731: \begin{eqnarray}

732:     -\frac{1}{n} \frac{\partial}{\partial x_i} P(\vect{r},t) &=& -\frac{\mathcal{V}}{n_0} \frac{1}{b_i} \frac{\partial}{\partial x_{0i}} \frac{P_0(\vect{r}_0)}{\mathcal{V}^{5/3}} = \frac{1}{b_i\mathcal{V}^{2/3}} \frac{\partial}{\partial x_{i0}} V(\vect{r}_0,0)\nonumber \\

733: &=& \frac{1}{b_i\mathcal{V}^{2/3}} m \omega_{i}^2(0) x_{i0}

734: \end{eqnarray}

735: 

736: The Euler equations then reduce to equations for the scaling

737: parameters $b_i(t)$, which can be solved numerically:

738: \begin{eqnarray}

739:     \label{e:eulerclass}

740:     \ddot{b}_i = -\omega_i^2(t)\, b_i + \frac{\omega_{i}^2(0)}{b_i\mathcal{V}^{2/3}}

741: \end{eqnarray}

742: 

743: In the following section we will see that superfluid hydrodynamics

744: leads to very similar scaling equations, with the exponent $2/3$

745: for the volume scaling parameter $\mathcal{V}$ replaced by the

746: parameter $\gamma$ in the equation of state of the superfluid

747: $\mu(n) = n^\gamma$. The discussion of free expansion, the

748: long-time behavior, inversion of the aspect ratio etc. will be

749: identical for superfluid hydrodynamics, so we defer the topic until

750: the next section.

751: 

752: \paragraph{From ballistic to hydrodynamic expansion}

753: 

754: The regime in between ballistic, collisionless expansion and pure

755: hydrodynamic, collisional expansion can be treated approximately.

756: For the effects of interactions on a classical gas,

757: see~\cite{guer99osc,pedr03}, for the case of Fermi gases with

758: attractive interactions, see~\cite{meno02}.

759: 

760: \subsubsection{Superfluid hydrodynamic expansion}

761: 

762: In the simplest (scalar) case, a superfluid is described by a

763: macroscopic, complex order parameter $\psi(\vect{r},t) =

764: \sqrt{n(\vect{r},t)} e^{{\rm i} \phi(\vect{r},t)}$ parameterized by

765: the {\it superfluid density} $n(\vect{r},t)$ and a phase

766: $\phi(\vect{r},t)$. The dynamics of the order parameter are well

767: described by a time-dependent Schr\"odinger equation of the type

768: \begin{equation}

769:     {\rm i} \hbar \frac{\partial}{\partial t} \psi(\vect{r},t) = \left(-\frac{\hbar^2}{2m}\nabla^2 + V(\vect{r},t) + \mu(n(\vect{r},t))\right)\psi(\vect{r},t)

770: \label{e:GPequation}

771: \end{equation}

772: where $\mu(n)$ is the chemical potential given by the equation of

773: state of the superfluid. In the case of weakly interacting BECs,

774: this is the Gross-Pitaevskii equation for the condensate

775: wave function from section~\ref{s:trappedmixtures}. For fermionic superfluids, a formally similar equation is the Ginzburg-Landau equation, which is however valid only close to $T_C$.

776: Rewriting Eq.~\ref{e:GPequation}

777: in terms of the superfluid density $n$ and velocity $\vect{v}$,

778: neglecting the curvature $\nabla^2 \sqrt{n}$ of the magnitude of

779: $\psi$ and using the fact that the superfluid is

780: irrotational $\nabla \times \vect{r} = 0$, we arrive again at the

781: continuity equation and the Euler equation for classical inviscous

782: flow:

783: \begin{eqnarray}

784: \frac{\partial n}{\partial t} + \nabla \cdot (n \vect{v}) &=& 0\\

785: m \frac{\partial \vect{v}}{\partial t} + m (\vect{v} \cdot \nabla)

786: \vect{v} &=& - \nabla \left(V+ \mu(n)\right)

787: \end{eqnarray}

788: The validity of these hydrodynamic equations is restricted to superfluids whose

789: healing length is much smaller than the sample size and thus, for fermionic superfluids in

790: a harmonic trap, for a superfluid gap larger than

791: the harmonic oscillator energies $\hbar

792: \omega_{x,y,z}$~\cite{meno02}.

793: For a power-law equation of state $\mu(n) \propto n^\gamma$, the

794: equations allow a scaling solution for (possibly time-varying)

795: harmonic potentials. The scaling parameters $b_i(t)$ are

796: given by the differential equations~\cite{cast96,kaga97bose,meno02,hu04coll}

797: \begin{eqnarray}

798:     \ddot{b}_i = -\omega_i^2(t)\, b_i + \frac{\omega_{i}^2(0)}{b_i\mathcal{V}^{\gamma}}

799:     \label{e:euler}

800: \end{eqnarray}

801: Important limiting cases in the BEC-BCS crossover are:

802: \begin{itemize}

803: \item[-] BEC-limit ($1/k_F a \gg 1$): Here, the mean-field repulsion between molecules leads to a chemical potential per fermion $\mu(n) = \frac{\pi\hbar^2 a_M n}{m}$, so $\gamma = 1$.

804: \item[-] BCS-limit ($1/k_F a \ll -1$): In the BCS-limit, the dominant contribution to the chemical potential comes from the kinetic energy of the constituent fermions, given by the Fermi energy. So here $\mu(n) = \epsilon_F \propto n^{2/3}$ and $\gamma = 2/3$.

805: \item[-] Unitarity limit ($1/k_F a = 0$): In the unitarity limit, the only remaining energy scale is the Fermi energy. One necessarily has $\mu(n) \propto \epsilon_F \propto n^{2/3}$ and $\gamma = 2/3$, just as in the BCS-limit.

806: \end{itemize}

807: 

808: Note that the scaling laws for the BCS- and the unitarity limit~\cite{cast04scal}

809: are identical to those found for a collisionally hydrodynamic gas

810: in section~\ref{s:collhydro}.

811: For a derivation of superfluid hydrodynamics in the BCS-limit, we refer the reader to the contribution of Y. Castin to these lecture notes.

812: 

813: \begin{figure}

814:     \centering

815:     \includegraphics[width=3in]{figs_quantanalysis/gamma.eps}

816:     \caption{The exponent $\gamma$ as a function of the interaction parameter $1/k_F a$. $\gamma$ approximately describes the superfluid equation of state $\mu(n) \sim n^\gamma$ in the BEC-BCS crossover. A similar figure can be found in~\cite{hu04coll}.}

817:     \label{f:gamma}

818: \end{figure}

819: 

820: The Leggett ansatz (see section~\ref{s:crossoverwavefunction}) allows to interpolate

821: between the BEC- and the BCS-regime and gives a chemical potential

822: $\mu(n)$ that correctly captures the physics in the two limits.

823: With its help, we can define an effective exponent $\gamma =

824: \frac{n}{\mu}\frac{\partial \mu}{\partial n}$ and write $\mu(n)

825: \simeq n^\gamma$, assuming that $\gamma$ varies slowly with the

826: interaction parameter $1/k_F a$. This exponent, shown in

827: Fig.~\ref{f:gamma}, attains the correct limiting values in the

828: BEC- and the BCS-limit, as well as on resonance, so we may use it

829: for the present purpose as an approximate description of the gas'

830: equation of state throughout the crossover.

831: 

832: \paragraph{In-trap density profile}

833: The in-trap density profile of the superfluid at zero temperature can be deduced from

834: the Euler equations in steady state. Neglecting kinetic energy

835: $\frac{1}{2}\, m\, \vect{v}^2$ (Thomas-Fermi approximation), the

836: equation simply reads $V(\vect{r}) + \mu(n(\vect{r})) = {\rm

837: const.} = \mu(n(\vect{0}))$. For the power-law equation of state

838: $\mu(n)\propto n^\gamma$, we directly obtain

839: \begin{equation}

840:     n(\vect{r}) \propto \left(\mu(n(\vect{0})) - V(\vect{r})\right)^{1/\gamma}

841: \end{equation}

842: for $\mu(n(\vect{0})) > V(\vect{r})$ and zero otherwise. For a BEC

843: and harmonic trapping, we recover the inverted parabola,

844: Eq.~\ref{e:BECparabola}, for a BCS superfluid in the limit of weak

845: interactions the density distribution of an ideal Fermi gas,

846: Eq.~\ref{e:Fermidensity}. Note that in the crossover $1/k_F |a|

847: \lesssim 1$, the correct calculation of the density profile is

848: less straightforward, as the parameter $1/k_F(\vect{r}) a$ depends

849: on position, and the equation of state varies across the cloud.

850: The power-law approach, using a fixed $\gamma =

851: \gamma(1/k_F(\vect{0}) a)$, will only provide an approximate

852: description. Fortunately, on resonance evidently $1/k_F(\vect{r})a

853: = 0$ across the entire cloud, and the power-law equation of state

854: becomes exact at $T=0$.

855: 

856: \paragraph{Free expansion out of a cylindrically symmetric trap}

857: 

858: In this case $\omega_i(t>0) = 0$, and $\omega_x(0) = \omega_y(0)

859: \equiv \omega_\perp$. We have

860: \begin{eqnarray}

861: \label{e:cylindricscaling1}

862: \ddot{b}_\perp &=& \frac{\omega_\perp^2}{b_\perp^{2\gamma+1}

863: b_z^{\gamma}}\\ \ddot{b}_z &=&

864: \frac{\omega_z^2}{b_\perp^{2\gamma} b_z^{\gamma+1}}

865: \label{e:cylindricscaling2}

866: \end{eqnarray}

867: The MIT trap is cigar-shaped, with an aspect ratio of short to

868: long axes $\epsilon = \omega_z/\omega_\perp \ll 1$. In such a

869: case, expansion is fast in the radial, initially tightly confined

870: dimensions, whereas it is slow in the $z$-direction. For times

871: short compared to $\tau_\epsilon = \frac{1}{\omega_z \epsilon}$,

872: many axial trapping periods, we can set $b_z \approx 1$ on the

873: right side of Eqs.~\ref{e:cylindricscaling1} and \ref{e:cylindricscaling2}, decoupling the

874: transverse from the axial expansion. For $\gamma = 1$, the case of

875: a Bose-Einstein condensate of tightly bound molecules, the

876: simplified equations for $t\ll\tau_\epsilon$ have an analytic

877: solution~\cite{cast96,kaga97bose}: $b_\perp(t) = \sqrt{1 +

878: \omega_\perp^2 t^2}$ and $b_z(t) = 1 + \epsilon^2

879: \left(\omega_\perp t \arctan(\omega_\perp t) - \ln\sqrt{1 +

880: \omega_\perp^2 t^2}\right)$.  For long times $t$, the

881: expansion is linear in time: $b_\perp(t) = \omega_\perp t$ for $t \gg 1/\omega_\perp$ and

882: $b_z(t) = (\pi/2) \epsilon^2 \omega_\perp t$ for $t \gg \tau_\epsilon$. Note that the radial

883: expansion accidentally follows the same scaling law as that of a

884: ballistically expanding normal cloud.

885: 

886: The general behavior of the expanding gas is the same for all

887: relevant $\gamma$. Driven either by repulsive interactions

888: (BEC-case) or by kinetic energy (BCS-case), the gas first expands

889: radially at constant acceleration $\ddot{R}_\perp(t\ll

890: 1/\omega_\perp) = R_\perp(0) \omega_\perp^2$, and over a radial

891: trapping period reaches a final expansion velocity $\dot{R}(t\gg

892: 1/\omega_\perp) \approx \omega_\perp R_\perp(0)$. The axial size

893: grows as $b_z(t) -1 \approx \epsilon^2 \omega_\perp t$, leading to

894: an inversion of the cloud's aspect ratio from initially $\epsilon$

895: to $\sim 1/\epsilon$. This inversion is in contrast to the

896: isotropic aspect ratio of a ballistically expanding gas, and is

897: thus  characteristic for hydrodynamic expansion, which can be of

898: collisional {\it or} of superfluid origin. Fig.~\ref{f:aspect} and

899: table~\ref{t:expansionsummary} summarize the time evolution of the

900: cloud's radii and aspect ratios for $\gamma = 1$ (BEC) and $\gamma

901: = 2/3$ (BCS and Unitarity), while Fig.~\ref{f:velocityplot}

902: compares the long-time behavior of the velocities and aspect

903: ratios across the BEC-BCS crossover. For expansion out of an

904: elongated cigar-shaped trap and $\gamma = 2/3$, which holds in the

905: BCS-limit, at unitarity, but also for a collisionally hydrodynamic

906: gas, the asymptotic expansion velocity is $v_\perp =

907: \sqrt{\frac{3}{2}} \omega_\perp R_\perp(0) \approx 1.22\;

908: \omega_\perp R_\perp(0)$. This can be understood by noting that

909: the cloud's kinetic energy, initially distributed isotropically,

910: is released only into the radial direction during hydrodynamic

911: expansion, so $\frac{1}{2} m v_\perp^2 = \frac{3}{2} \mu =

912: \frac{3}{4} m \omega_\perp^2 R_\perp(0)^2$.

913: 

914: \begin{figure}

915:     \centering

916:     \includegraphics[width=3in]{figs_quantanalysis/aspect.eps}

917:     \caption{Aspect ratio $\epsilon(t) = R_x(t)/R_z(t)$ as a function of time for the MIT trap ($\epsilon = 1/6$) in ballistic, collisional or superfluid hydrodynamic expansion ($\gamma = 2/3$) and superfluid hydrodynamic expansion of a molecular BEC ($\gamma = 1$). }

918:     \label{f:aspect}

919: \end{figure}

920: 

921: 

922: \begin{figure}

923:     \centering

924:     \includegraphics[width=3in]{figs_quantanalysis/velplot.eps}

925:     \caption{Asymptotic velocities and aspect ratio for hydrodynamic expansion out of a very elongated cigar-shaped trap ($\epsilon = \omega_z/\omega_x \ll 1$), as a function of the interaction parameter $1/k_F a$. The dashed lines show the asymptotic values in the BCS-limit. A similar figure can be found in~\cite{hu04coll}.}

926:     \label{f:velocityplot}

927: \end{figure}

928: 

929: \begin{table}

930:   \centering

931:   \begin{tabular}{p{0.03\linewidth}r|c|c|cp{0.25\linewidth}}

932:   % after \\: \hline or \cline{col1-col2} \cline{col3-col4} ...

933: &   & Ballistic & Hydrodynamic (BEC) & (BCS, unitarity, collisional) \\ \hline

934: &  $b_\perp(t)$ & $\sqrt{1+\omega_\perp^2 t^2}$ & $\sqrt{1+\omega_\perp^2 t^2}$ & $\sim 1.22\; \omega_\perp t$ \\

935: &  $b_z(t)$ & $\sqrt{1+\omega_z^2 t^2}$ & $1 + \epsilon^2 (\omega_\perp t \arctan(\omega_\perp t)$ & $\sim 2.05 \;\frac{\pi}{2}\epsilon^2 \omega_\perp t$ \\

936: && & $\qquad \quad- \ln\sqrt{1 + \omega_\perp^2 t^2})$ & \\

937: &  AR & 1 & $\frac{2}{\pi} \frac{1}{\epsilon}$ & $\sim 0.60\; \frac{2}{\pi} \frac{1}{\epsilon}$

938: \end{tabular}

939:   \caption{Comparison between ballistic and hydrodynamic expansion.

940: Formulas for hydrodynamic expansion assume a long cigar-shaped

941: trap ($\epsilon = \omega_z/\omega_x \ll 1$), formulas for the

942: aspect ratio (AR) and for the BCS, Unitarity, collisional limit

943: give the asymptotic behavior. The formula for BEC-expansion is valid at short times $t \ll \omega_\perp/\omega_z^2$, but also captures the correct long time limit.}\label{t:expansionsummary}

944: \end{table}

945: 

946: 

947: \paragraph{Hydrodynamic expansion into a saddle potential}

948: 

949: As discussed in section~\ref{s:saddleballistic}, expansion may not occur into free

950: space, but into an inhomogeneous magnetic field which is often

951: described by a saddle potential.

952: 

953: The Euler equations (\ref{e:euler}) now read for $t>0$

954: \begin{equation}

955:     \ddot{b}_i = -\omega_{S,i}^2\, b_i + \frac{\omega_{i}^2(0)}{b_i\mathcal{V}^{\gamma}}

956: \end{equation}

957: Here, $\omega_{S,i}$ are the real or imaginary frequencies

958: characterizing the saddle point potential. These equations

959: typically need to be solved numerically. For a Bose-Einstein

960: condensate of molecules expanding from long cigar-shaped traps

961: ($\epsilon \ll 1$), the {\it radial} equation again allows for an

962: analytic solution identical to that for a ballistically expanding,

963: non-interacting gas. One obtains

964: \begin{equation}

965: b_\perp(t) = \sqrt{\cosh^2\left(\frac{1}{\sqrt{2}}\omega_{Sz}t\right) +

966: \frac{2\omega_\perp^2}{\omega_{Sz}^2}

967: \sinh^2\left(\frac{1}{\sqrt{2}}\omega_{Sz}t\right)}.

968: \end{equation}

969: However, the {\it axial} cloud size behaves drastically different

970: from a non-interacting cloud. For $\omega_{Sz} = \omega_z$, the

971: axial cloud size of a non-interacting gas would never change

972: ($b_z(t) = 1$), whereas a hydrodynamic gas, released into the

973: radial dimensions, will start to {\it shrink} axially under the

974: influence of the confining axial potential. The cloud's energy

975: (interaction energy for a BEC, kinetic energy for a BCS

976: superfluid) escapes radially, hence there is not sufficient

977: pressure to maintain the axial cloud size.

978: 

979: Further discussions of superfluid hydrodynamics and scaling transformations can be found in the contributions of Y. Castin and S. Stringari to these proceedings.

980: 

981: \subsection{Fitting functions for trapped and expanded Fermi gases}

982: 

983: In the preceding sections we derived the 3D density distribution

984: of a Fermi mixture in various regimes. However, all imaging

985: techniques record column densities, density profiles integrated

986: along the line of sight (the $z$-axis in the following).

987: 

988: For condensed gases, where $n(\vect{r}) \approx n(\vect{0})\, (1 -

989: V(\vect{r})/\mu)^{1/\gamma}$, one obtains the column density

990: \begin{equation}

991:     n_{\rm 2D, c}(x,y) =  n_c\left(1 - \frac{x^2}{R_x^2} - \frac{y^2}{R_y^2}\right)^{\frac{1}{\gamma}+\frac{1}{2}}

992: \end{equation}

993: For thermal Bose (molecular) and Fermi clouds, we have

994: \begin{equation}

995: \label{e:coldens}

996:     n_{\rm 2D}(x,y) = n_{\rm 2D,0}\, {\rm Li}_2\left(\pm \exp\left[\beta\mu - \beta\frac{1}{2}m(\omega_x^2 x^2 + \omega_y^2 y^2)  \right]\right)/\,{\rm Li}_2\left(\pm\, e^{\beta\mu}\right)

997: \end{equation}

998: 

999: In the following, we will discuss the fitting functions valid in the different regimes of interaction, and the derived quantities.

1000: 

1001: \subsubsection{Non-interacting Fermi gases}

1002: \paragraph{Cloud size}

1003: In the classical regime at $T/T_F \gg 1$, the characteristic cloud

1004: size is given by the gaussian radius $\sigma_i = \sqrt{\frac{2 k_B

1005: T}{m\omega_i^2}}$. In the degenerate regime, however, the cloud

1006: size saturates at the Fermi radius $R_{Fi} = \sqrt{\frac{2

1007: E_F}{m\omega_i^2}}$. It is thus convenient to define a fit

1008: parameter that interpolates between the two limits:

1009: \begin{eqnarray}

1010:     R_i^2 &=& \frac{2 k_B T}{m\omega_i^2} \; f(e^{\mu\beta}) \rightarrow \left\{%

1011: \begin{array}{ll}

1012:     \sigma_i, & T/T_F \gg 1 \\

1013:     R_{Fi}, & T/T_F \ll 1 \\

1014: \end{array}%

1015: \right.    \nonumber \\

1016:     \mbox{where } f(x) &=& \frac{{\rm Li}_1(-x)}{{\rm Li}_0(-x)} = \frac{1 + x}{x} \ln(1+x)

1017: \end{eqnarray}

1018: For all temperatures, $R_i$ is thus directly related to the

1019: physical size of the cloud, and thus a better choice as a fit

1020: parameter than $\sigma_i$, which goes to zero at $T=0$, or

1021: $\sqrt{\frac{2\mu}{m\omega_i^2}}$, which goes to zero around

1022: $T/T_F = 0.57$. Numerically, using $R_i$ is easier to implement

1023: than using the root mean square radius of the cloud

1024: \begin{equation}

1025:     \left<x_i^2\right> = \frac{k_B T}{m \omega_i^2}\; \frac{{\rm Li}_4(-e^{\mu\beta})}{{\rm Li}_3(-e^{\mu\beta})}

1026: \end{equation}

1027: 

1028: \paragraph{Fitting function}

1029: 

1030: The fit function used for the density profiles of Fermi clouds is

1031: then in 2D

1032: \begin{equation}

1033:     n_{\rm 2D}(x,y) = n_{\rm 2D,0}\, \frac{{\rm Li}_2\left(\pm \exp\left[q - \left(\frac{x^2}{R_x^2} + \frac{y^2}{R_y^2}\right)\,f(e^q)\right]\right) }{{\rm Li}_2\left(\pm e^q\right)}

1034: \end{equation} and for 1D

1035: \begin{equation}

1036:     n_{\rm 1D}(x) = n_{\rm 1D,0}\, \frac{{\rm Li}_{5/2}\left(\pm \exp\left[q -\frac{x^2}{R_x^2} \,f(e^q)\right]\right) }{{\rm Li}_{5/2}\left(\pm e^q\right)}.

1037: \end{equation}

1038: The parameter $q = \mu\beta$, the logarithm of the fugacity,

1039: determines the {\it shape} of the cloud. For a small fugacity

1040: (large and negative $q$), the above functions reduce to the simple

1041: gaussian distribution of thermal clouds. For high fugacity (large

1042: and positive $q$), they tend to the

1043: zero-temperature distribution $n_{\rm 2D,0}(1 -

1044: \frac{x^2}{R_{Fx}^2})^2$ (in 2D) and $n_{\rm 1D,0}(1 -

1045: \frac{x^2}{R_{Fx}^2})^{5/2}$ (in 1D).

1046: 

1047: \paragraph{Derived quantities}

1048: 

1049: {\it Degeneracy}

1050: The degeneracy parameter $T/T_F$ can be calculated by combining

1051: Eq.~\ref{e:numberofatoms} with Eq.~\ref{e:Ferminumber}:

1052: \begin{equation}

1053:     \frac{T}{T_F} = \left[-6\, {\rm Li}_3(-e^{q})\right]^{-1/3}

1054:     \label{e:Degeneracy}

1055: \end{equation}

1056: This parameter depends only on the {\it shape} of the cloud. A

1057: characteristic point where shape deviations due to quantum

1058: statistics start to play a role is the point where $\mu$ changes

1059: sign, and we see from Eq.~\ref{e:Degeneracy} that this occurs at

1060: $T/T_F \approx 0.57$. Many non-ideal aspects of imaging, such as

1061: finite resolution, out of focus imaging, saturation, heating of

1062: the cloud by the probe pulse etc., tend to wash out the

1063: non-gaussian features of a highly degenerate Fermi cloud and hence

1064: lead to a larger value of $T/T_F$. However, dispersive effects due

1065: to non-resonant imaging light can potentially mimic sharp edges of

1066: the cloud, which the fitting routine would then falsely interpret

1067: to result from a very low $T/T_F$. It is clear that care has to be

1068: taken when determining the degeneracy parameter from the shape of

1069: the cloud alone.

1070: 

1071: \paragraph{Temperature}

1072: \label{s:temperature}

1073: The size of the cloud and the shape parameter $q$ give the

1074: temperature as

1075: \begin{equation}

1076:     k_B T = \frac{1}{2} m \omega_i^2 \frac{R_i^2}{b_i(t)^2} \frac{1}{f(e^q)}

1077:     \label{e:thermometry}

1078: \end{equation}

1079: where we have used the expansion factor $b_i(t)$ from

1080: section~\ref{s:expansion}. We recall that $b_i(t) =

1081: \sqrt{1+\omega_i^2 t^2}$ for the free expansion of a

1082: non-interacting Fermi gas. For low temperatures $T \ll T_F$,

1083: $f(e^{\mu\beta}) \rightarrow \mu\beta = \mu/k_B T$ and $R_i =

1084: b_i(t)\,R_{Fi}$. In this case, temperature only affects the wings

1085: of the density distribution, where the {\it local}

1086: $T/T_F(\vect{r})$ is still large. In fact,

1087: \begin{equation}

1088:     n_{\rm 1D}(x) \propto \left\{%

1089: \begin{array}{ll}

1090:     (1 - \frac{x^2}{R_{Fx}^2})^{5/2} & \hbox{for } x \ll R_{Fx} \\

1091:     e^{-\frac{x^2}{\sigma_x^2}} & \hbox{for } x \gg R_{Fx}, \\

1092: \end{array}%

1093: \right.

1094: \end{equation}

1095: and we see that temperature only affects the cloud's wings beyond

1096: the zero-temperature Fermi radius. Thermometry of very low

1097: temperature Fermi clouds is thus difficult, limited by the

1098: signal-to-noise ratio in the low-density wings of the

1099: distribution. This is different from thermometry of thermal clouds

1100: at high temperature $T \gg T_F$, where the entire size of the

1101: cloud $\sigma_i$ directly gives the temperature.

1102: 

1103: Because of the sensitivity to the cloud's wings, thermometry is

1104: more robust when the full 2D distribution is used for the fit.

1105: Alternatively, one can rely on the known trapping geometry plus

1106: the local density approximation and perform an average over the

1107: elliptical equipotential lines in the $x$-$y$ plane (line of sight

1108: integration necessarily mixes points at different values of the

1109: potential energy.) As the number of points included in the average

1110: grows with the distance from the cloud's center, the

1111: signal-to-noise will actually be best in the wings. Such an

1112: average is superior to a simple integration along the $x$-axis,

1113: for example, as this will more strongly mix regions that have

1114: different local $T/T_F$.

1115: 

1116: The ideal gases (Fermi, Bose, Boltzmann) are the only systems for

1117: which we have an exact description.  Therefore, they are

1118: attractive as a thermometer, when brought in contact with strongly

1119: interacting systems.  This concept has been recently carried out

1120: by determining absolute temperatures for imbalanced Fermi gases at

1121: unitarity~\cite{shin07phasediagram}.  In these systems, for sufficiently

1122: high imbalance, the majority cloud extends beyond the minority

1123: cloud, and is (locally) an ideal gas.  Therefore, in Ref.~\cite{shin07phasediagram} the spatial

1124: wings of these clouds could be fitted with the functions for the

1125: ideal Fermi gas discussed in this section, and absolute

1126: temperatures for the superfluid phase transition could be

1127: determined. The fitting of the majority wings had to be done with in-trap

1128: profiles, which required to address the effect of anharmonicities of the

1129: optical trapping potential.  Usually, for thermometry, ballistic

1130: expansion is preferable since velocity distributions are

1131: independent of the shape of the trapping potential.  However, in

1132: the case of imbalanced Fermi gases, the atoms in the wings

1133: can collide with the strongly interacting core during expansion,

1134: modifying their velocity distribution.

1135: 

1136: Another way to perform ideal gas thermometry is done by converting

1137: the sample to a non-interacting system by sweeping sufficiently

1138: far away from the Feshbach resonance.  If such magnetic field

1139: sweeps are adiabatic, they conserve entropy (but not temperature).

1140: By fitting the spatial profiles of the non-interacting cloud, the

1141: entropy $S$ of the strongly interacting system can be determined.

1142: If it is possible to vary the energy of the strongly interacting

1143: system in a controlled way (e.g. by using the virial theorem at

1144: unitarity~\cite{ho04uni,thom05virial} or by providing controlled heating~\cite{kina05heat}), one

1145: can determine the derivative $dS/dE$ which is equal to the inverse

1146: absolute temperature.  So far this method could be implemented

1147: only for a balanced Fermi system at unitarity~\cite{luo07entropy} and, due to the need

1148: of determining a derivative, could only provide temperatures

1149: averaged over a range of energies.

1150: 

1151: \paragraph{Number of atoms and Fermi energy}

1152: 

1153: The number of atoms in the observed spin state can be obtained

1154: from the total absorption recorded in the cloud's CCD image. The

1155: transmission of resonant light at pixel $(x,y)$ is given by

1156: $\tilde{T}(x,y) = e^{-\sigma_0 \int n_{\rm 3D}(x,y,z)\,{\rm d}z}$,

1157: where $\sigma_0$ is the resonant atom-photon cross section for

1158: light absorption. Thus, the number of atoms is

1159: \begin{equation}

1160:     N_\uparrow = \frac{A}{M \sigma_0} \sum_{pixels} - \ln(\tilde{T}(x,y))

1161: \end{equation}

1162: where $A$ is the area per pixel and $M$ the optical magnification.

1163: 

1164: Typically, the fitting functions are applied to the optical

1165: density $\sigma_0 \int_z n_{\rm 3D}(x,y,z) = -

1166: \ln\left(\tilde{T}(x,y)\right)$. The fit parameter $n_{\rm 2D,0}$

1167: thus measures the peak optical density of the cloud, while the

1168: radii $\tilde{R}_x$ and $\tilde{R}_y$ have units of camera pixels.

1169: The number of atoms described by the fitting function is thus

1170: given by

1171: \begin{eqnarray}

1172:     N_{\rm fit} &=& \frac{A}{M \sigma_0}\pi\, n_{\rm 2D,0}\, \tilde{R}_x \tilde{R}_y \frac{{\rm Li}_3(-e^{q})}{{\rm Li}_2(-e^{q})}\frac{{\rm Li}_0(-e^{q})}{{\rm Li}_1-e^{q})} \nonumber \\

1173: &\rightarrow& \frac{A}{\sigma_0}\left\{%

1174: \begin{array}{ll}

1175:     \frac{\pi}{3} \, n_{\rm 2D,0} \, \tilde{R}_{Fx} \tilde{R}_{Fy}, & T \ll T_F \\

1176:     \pi\, n_{\rm 2D,0} \, \tilde{\sigma}_{x} \tilde{\sigma}_{y}, & T \gg T_F. \\

1177: \end{array}%

1178: \right.

1179: \end{eqnarray}

1180: 

1181: From the number of atoms and the trapping frequencies, one can

1182: calculate the Fermi energy $k_B T_F$:

1183: \begin{equation}

1184: k_B T_F = \hbar \bar{\omega} (6 N_{\rm fit})^{1/3}

1185: \end{equation}

1186: 

1187: An independent determination of the Fermi energy is provided

1188: by the measured (physical) size of the cloud $R_i$ for highly

1189: degenerate clouds. For $T\rightarrow 0$, $R_i \approx b_i(t)

1190: R_{Fi}$ and thus $ k_B T_F = \frac{1}{2} m \omega_i^2

1191: \frac{R_i^2}{b_i(t)^2} $.  As only the trapping frequencies and

1192: the magnification of the imaging system enter into this equation,

1193: this relation allows a calibration of the light absorption cross

1194: section which may be reduced from the resonant cross section by

1195: detuning, non-ideal polarization of the probe light, and

1196: saturation.

1197: 

1198: For arbitrary temperature, the shape parameter $q$ enters the

1199: relation for the Fermi energy:

1200: \begin{equation}

1201: k_B T_F = k_B T \frac{T_F}{T} = \frac{1}{2} m \omega_i^2

1202: \frac{R_i^2}{b_i(t)^2} \frac{\left(-6\, {\rm

1203: Li}_3(-e^{q})\right)^{1/3}}{f(e^q)}

1204: \end{equation}

1205: 

1206: \subsubsection{Resonantly interacting Fermi gases}

1207: \label{s:resonantgases}

1208: 

1209: The calculation of density profiles of interacting gases is

1210: delicate. Already above the superfluid transition temperature,

1211: attractive interactions lead to a shrinking of the cloud. Since

1212: interactions (parameterized by the local $k_F a$) vary across the

1213: cloud, there is a priori no simple analytical function describing

1214: interacting Fermi gases. Experimentally, it turns out that the

1215: difference in the {\it shape} of a (balanced) interacting and a

1216: non-interacting Fermi mixture is minute around resonance and on

1217: the BCS-side. Especially for the resonant case ($1/k_F a = 0$),

1218: this has led to the common practice of using the shape of the

1219: non-interacting Fermi gas as  fitting function, and quote an

1220: effective temperature $\tilde{T}$ and effective degeneracy

1221: $\frac{\tilde{T}}{T_F}$ of resonantly interacting

1222: clouds~\cite{ohar02science,kina05heat}. In fact, universality on

1223: resonance implies that the gas' chemical potential must be

1224: $\mu(\vect{r}) = \xi(T/T_F) \epsilon_F(\vect{r})$, with a

1225: universal function $\xi(T/T_F)$ which only depends on the reduced

1226: temperature $T/T_F$~\cite{ho04uni}. The zero-temperature limit of

1227: $\xi \equiv \xi_0$ has been subject of extensive experimental and

1228: theoretical studies (see section~\ref{s:energymeasurements}), and its

1229: value is $\xi(0) \approx 0.42$. At $T=0$, we have for a trapped

1230: gas $\mu(\vect{r}) = \mu_0 - V(\vect{r}) = \xi_0

1231: \epsilon_F(\vect{r}) \propto n^{2/3}(\vect{r})$. The density

1232: profile will then have the exact same shape as a non-interacting

1233: Fermi gas, with a renormalized Fermi temperature. However, for

1234: finite temperature, $\xi(T/T_F)$ differs from the temperature

1235: dependence of a non-interacting gas~\cite{bulg06TC}, and there is

1236: no a priori reason that the shape of the cloud at unitarity should

1237: be similar to that of a non-interacting Fermi gas. It turns out

1238: that the difference is very small.

1239: 

1240: \begin{figure}[h]

1241: \begin{center}

1242: \includegraphics[width=5.3in]{figs_quantanalysis/dilemma.eps}

1243: \caption[The absence of a clear signature of condensation in the

1244: spatial profile of strongly interacting Fermi gases]{The absence

1245: of a signature of condensation in the spatial profile of strongly

1246: interacting Fermi gases.  Shown are high-resolution images of spin

1247: up atoms in a resonantly interacting, equal mixture of spin up and

1248: spin down for different temperatures. The lower graphs show

1249: azimuthally averaged radial profiles (noise level well below 1\%

1250: of the maximum optical density). All three clouds are very well

1251: fit using a finite-temperature Thomas-Fermi distribution (with

1252: fugacity $e^{\mu / k_B \widetilde{T}}$, central density $n_0$ and

1253: mean square radius $<r^2>$ as free parameters, see

1254: Eq.~\ref{e:coldens}).   However, the

1255: empirical temperatures of $\widetilde{T}/T_F = 0.22$ (a), $0.13$

1256: (b) and $0.075$ (c) determined from the profiles' wings indicate

1257: that at least clouds $b$ and $c$ should be in the superfluid

1258: regime. Trap parameters $\nu_r = 162$ Hz, $\nu_z = 22.8$ Hz, 10 ms

1259: time of flight, expansion factor 13.9,  atom numbers $N$ per spin

1260: state were $10.2$ (a), 9.5 (b) and 7.5 $\times

1261: 10^6$.}\label{f:dilemma}

1262: \end{center}

1263: \end{figure}

1264: 

1265: The shape similarity was an important issue in the quest for

1266: superfluidity in Fermi gases. In the case of weakly interacting

1267: BECs, condensation is apparent from the sudden appearance of a

1268: dense, central core in midst of a large thermal cloud. In contrast

1269: to that, Fermi gases do not show such a  signature, at least at

1270: first sight (see Fig.~\ref{f:dilemma}), and different detection

1271: methods for superfluidity were explored.

1272: 

1273: The only loophole that may allow seeing a signature of superfluidity in the

1274: spatial profile of balanced Fermi gases would be a rapid variation of $\xi(T/T_F)$ around the critical temperature $T_C$. This would

1275: translate into a sudden variation of the density at the interface

1276: between the normal and superfluid region, e.g. where the gas is

1277: locally critical, $T = T_C(\vect{r})$.

1278: 

1279: 

1280: We have indeed found a faint signature of condensation in density

1281: profiles of the unitary gas on resonance after expansion. These

1282: results will be presented in section~\ref{s:directunitarity}. Note

1283: that the observation of such a feature in the density profiles

1284: draws into question the common practice of determining an

1285: ``effective temperature'' from density profiles at unitarity using

1286: the ideal gas fitting function. In contrast to balanced Fermi

1287: mixtures, a striking signature of condensation can be observed in

1288: the density profiles of mixtures with {\it imbalanced} populations

1289: of spin up and spin down fermions. This will be discussed in

1290: section~\ref{s:directimbalance}.

1291: 

1292: \subsubsection{Molecular clouds}

1293: \label{s:molecularclouds}

1294: In partially condensed molecular gases that are weakly

1295: interacting, one can neglect the mutual repulsion between the

1296: condensate and the surrounding thermal cloud of molecules. The

1297: density distribution is typically well-fit by a bimodal sum of an

1298: inverted parabola for the condensate

1299: \begin{equation}

1300:     n_c(x,y) = n_{c0}\left(1 - \frac{x^2}{R_{cx}^2} - \frac{y^2}{R_{cy}^2}\right)

1301:     \label{e:condensatefit}

1302: \end{equation}

1303: and a Bose-function for the thermal cloud, as in

1304: Eq.~\ref{e:coldens}, with the parameter $q = \mu\beta$ often left as an adjustable parameter (instead of fixing it via the condensate's chemical potential $\mu = g n_{c0}$):

1305: \begin{equation}

1306:     n_{th}(x,y) = n_{th0} \, {\rm Li}_2\left(\exp\left[q-\frac{x^2}{R_{th,x}^2} - \frac{y^2}{R_{th,y}^2}\right]\right)/{\rm Li}_2(\exp\left[q\right])

1307: \label{e:bimodalfit}

1308: \end{equation}

1309: For practical purposes, this is often simplified by a gaussian, as if $q \ll 0$. Then $n_{th} \approx n_{th0} \, \exp\left(-\frac{x^2}{R_{th,x}^2} - \frac{y^2}{R_{th,y}^2}\right)$.

1310: 

1311: Once the condensate mean-field $2 g n_c$ (with $g =

1312: \frac{4\pi\hbar^2 a_M}{M}$) experienced by thermal molecules is no

1313: longer small compared to $k_B T$, the mutual repulsion can no

1314: longer be neglected. The thermal molecules will then experience a

1315: ``Mexican-hat'' potential, the sum of the confining harmonic

1316: potential $V_M(\vect{r})$, and the repulsion from the condensate $2

1317: g n_c(\vect{r})$ and from the surrounding thermal cloud, $2 g

1318: n_{th}(\vect{r})$. The thermal molecules themselves will in turn

1319: repel the condensate. The situation can be captured by two coupled

1320: equations for the condensate (in Thomas-Fermi approximation) and

1321: the thermal cloud:

1322: \begin{eqnarray}

1323:     g n_c(\vect{r}) &=& {\rm Max}\left(\mu - V_M(\vect{r}) - 2 g n_{th}(\vect{r}),0\right) \nonumber \\

1324:     n_{th}(\vect{r}) &=& \frac{1}{\lambda_M^3}\,{\rm Li}_{3/2}(e^{\beta\left(\mu - V_M(\vect{r}) - 2 g n_c(\vect{r}) - 2 g n_{th}(\vect{r})\right)}) \nonumber \\

1325: &=& \frac{1}{\lambda_M^3}\,{\rm

1326: Li}_{3/2}(e^{-\beta\left|\mu - V_M(\vect{r}) - 2 g

1327: n_{th}(\vect{r})\right|})

1328: \end{eqnarray}

1329: where $\lambda_M$ is the thermal de Broglie wavelength for molecules. In the case of weakly interacting Bose gases, one can neglect the mean field term $2 g n_{th}(\vect{r})$~\cite{nara98semi}.

1330: 

1331: Note that these coupled equations are only an approximative way to

1332: describe the strongly correlated gas. The mean field approximation

1333: for the thermal molecules neglects phonons and other collective

1334: excitations.  The above equations can be solved numerically. In

1335: the limit of strong interactions, the condensate almost fully

1336: expels the thermal molecules from the trap center, so that the

1337: thermal cloud forms a shell around the condensate.

1338: 

1339: The practical implication of this discussion is that there is no

1340: simple analytic expression for the density distribution of partially condensed

1341: clouds in the strongly interacting regime. For fitting purposes

1342: one may still choose for example the bimodal fit of

1343: Eq.~\ref{e:bimodalfit}, but one must be aware that quantities like

1344: the ``condensate fraction'' thus obtained depend on the model

1345: assumed in the fit.   For tests of many-body calculations, the

1346: full density distributions should be compared to those predicted

1347: by theory.

1348: 

1349: \paragraph{Derived quantities}

1350: {\it Temperature.}

1351: 

1352: For weakly interacting Bose gases, Eq.~\ref{e:bimodalfit} holds

1353: and the temperature is given by

1354: \begin{equation}

1355:     k_B T = \frac{1}{2} m \omega_i^2 \frac{R_{th,i}^2}{b_i(t)^2}

1356: \end{equation}

1357: where $b_i(t) = \sqrt{1 + \omega_i^2 t^2}$ is the expansion factor

1358: of the thermal gas. To ensure model-independent results, only the

1359: thermal gas should be included in the fit, not the condensed core.

1360: For strongly interacting clouds, temperature can in principle

1361: still be obtained from the thermal wings of the {\it trapped}

1362: molecular distribution, which is gaussian at distances $\vect{r}$

1363: for which $V_M(\vect{r}) \gg \mu$.  However, a

1364: possible systematic correction can occur in expansion due to

1365: interactions of molecules in the wings with the core, which may be

1366: either condensed or strongly interacting.

1367: 

1368: Note that unless the whole cloud is deep in the hydrodynamic

1369: regime, there is no simple scaling law for the expansion of such

1370: strongly interacting molecular gases.  Absolute thermometry of

1371: strongly interacting, balanced gases is still a challenging

1372: problem.

1373: 

1374: 

1375: {\it Chemical potential}

1376: 

1377: In a confining potential, and at zero temperature, the chemical

1378: potential is given by the size of the condensate, as $V(\vect{r})

1379: = \mu$. It can be expressed by the fit parameters according to

1380: Eq.~\ref{e:condensatefit} as

1381: \begin{equation}

1382:     \mu = \frac{1}{2} m \omega_i^2 \frac{R_{c,i}^2}{b_i(t)^2}

1383: \end{equation}

1384: with $b_i(t)$ the expansion factor for superfluid hydrodynamic

1385: expansion into direction $i$. At finite temperatures and for

1386: strong interactions, the thermal cloud will mostly reside outside

1387: the condensate and can affect the actual or fitted condensate

1388: size.

1389: 

1390: {\it Condensate fraction}

1391: 

1392: In the field of dilute atomic gases, the condensate fraction is a

1393: key quantity to characterize the superfluid regime.  In contrast

1394: to superfluid helium and superconductors, gaseous condensates can

1395: be directly observed in a dramatic way. However, unless

1396: interactions are negligible,  the determination of the condensate

1397: fraction is  model dependent. For weakly interacting gases (or

1398: those obtained after a rapid ramp into the weakly interacting

1399: regime), the density distribution can typically be well fit with

1400: the bimodal distribution of Eqs.~\ref{e:condensatefit} and

1401: \ref{e:bimodalfit}. A robust way to define a ``condensate

1402: fraction'' is then to ascribe the total number of molecules in the

1403: narrower distribution to the condensate. For strong interactions

1404: however, the mean-field repulsion of thermal and condensed

1405: molecules (see above) will lead to the expulsion of a large part

1406: of thermal molecules from the condensate. In addition, low-energy

1407: excitations such as phonons, as well as quantum depletion will

1408: modify the non-condensed fraction at the position of the

1409: condensate, and the fitted condensate fraction depends on the form

1410: of the fitting function for the bimodal fit.  In these cases, it

1411: is better to directly compare density distributions with

1412: theoretical predictions.

1413: