1: \documentclass[prl,aps,showpacs,twocolumn,amsmath,superscriptaddress]{revtex4}

2: \usepackage[dvips]{color}

3: \usepackage{graphicx} % Include figure files

4: \usepackage{bm}% bold math

5: 

6: \newcommand{\bra}[1]{\langle #1 |\,}

7: \newcommand{\Bra}[1]{\left\langle #1 \right|}

8: \newcommand{\ket}[1]{\,| #1 \rangle}

9: \newcommand{\Ket}[1]{\left| #1 \right\rangle}

10: \newcommand{\bracket}[2]{\langle #1 | #2 \rangle}

11: \newcommand{\Bracket}[2]{\left.\left\langle #1 \right| #2 \right\rangle}

12: \newcommand{\be}{\begin{equation}}

13: \newcommand{\ee}{\end{equation}}

14: \newcommand{\bea}{\begin{eqnarray}}

15: \newcommand{\eea}{\end{eqnarray}}

16: \newcommand{\bse}{\begin{subequations}}

17: \newcommand{\ese}{\end{subequations}}

18: \renewcommand{\d}{\mathrm{d}}

19: \renewcommand{\a}{a_{\mathrm{1D}}}

20: \newcommand{\chio}{\chi^{(0)}}

21: \newcommand{\pf}{k_{\mathrm F}}

22: \newcommand{\kf}{k_{\mathrm F}}

23: \newcommand{\ef}{\varepsilon_{\mathrm F}}

24: \newcommand{\im}{\mathrm{Im}}

25: \newcommand{\gb}{g_\mathrm{B}}

26: \newcommand{\gf}{g_\mathrm{F}}

27: \newcommand{\q}{q}

28: 

29: %\newcommand{\text}{\mathrm}

30: \newcommand{\eprint}[1]{#1}

31: \renewcommand{\textheight}{25cm}

32: 

33: \begin{document}

34: \title{The Dynamic Structure Factor of the

35: 1D {B}ose Gas near the {T}onks-{G}irardeau Limit}

36: 

37: \author{Joachim Brand}

38: % \email{joachim@mpipks-dresden.mpg.de} 

39: \affiliation{Max Planck Institute for the Physics of Complex

40: Systems, N\"othnitzer Stra{\ss}e 38, 01187 Dresden, Germany}

41: 

42: \author{Alexander Yu.~Cherny}

43: % \email{cherny@thsun1.jinr.ru}

44: \affiliation{Max Planck Institute for the Physics of Complex

45: Systems, N\"othnitzer Stra{\ss}e 38, 01187 Dresden, Germany}

46: \affiliation{Bogoliubov Laboratory of Theoretical Physics, Joint

47: Institute for Nuclear Research, 141980, Dubna, Moscow region, Russia}

48: \date{\today}

49: 

50: \begin{abstract}

51: 

52: While the 1D Bose gas appears to exhibit superfluid response under

53: certain conditions, it fails the Landau criterion according to the

54: elementary excitation spectrum calculated by Lieb.  The apparent

55: riddle is solved by calculating the dynamic structure factor of the

56: Lieb-Liniger 1D Bose gas.  A pseudopotential Hamiltonian in the

57: fermionic representation is used to derive a Hartree-Fock operator,

58: which turns out to be well-behaved and local. The Random-Phase

59: approximation for the dynamic structure factor based on this

60: derivation is calculated analytically and is expected to be valid at

61: least up to first order in $1/\gamma$, where $\gamma$ is the

62: dimensionless interaction strength of the model. The dynamic structure

63: factor in this approximation clearly indicates a crossover behavior

64: from the non-superfluid Tonks to the superfluid weakly-interacting

65: regime, which should be observable by Bragg scattering in current

66: experiments.

67: \end{abstract}

68: \pacs{03.75.Kk, 03.75.Hh, 05.30.Jp}

69: 

70: \maketitle

71: 

72: % \subsubsection*{Introduction}

73: The emergence of superfluidity at low temperatures is one of the most

74: dramatic manifestations of quantum many-body physics in

75: nature. Although the gas of interacting Bosons in 1D is described by

76: the exactly solvable Lieb-Liniger (LL) model \cite{lieb63:1,lieb63:2},

77: the question of superfluidity is not readily answered from the exact

78: solutions. The dimensionless interaction parameter 

79: % $\gamma = -(\a n)^{-1}$ 

80: $\gamma = \gb m /(\hbar^2 n)$ \cite{lieb63:1} governs the crossover

81: from the weakly-interacting quasicondensate for $\gamma \ll 1$ to the

82: strongly-interacting Tonks-Girardeau gas \cite{girardeau60} at

83: $\gamma=\infty$, where $\gb$ is the coupling constant of the 1D Bose

84: gas 

85: %% \footnote{The Boson coupling constant is related to the 1D and 3D

86: %% $s$-wave scattering lengths $\a$ and $a_{\mathrm{3D}}$, respectively,

87: %% by $\gb = -2\hbar^2 / (m\, \a) = 2 a_{\mathrm{3D}}\, \hbar\omega_\rho

88: %% [1-C a_{\mathrm{3D}} \sqrt{m\omega_\rho/(2\hbar)}] $,

89: %% where $\omega_\rho$ is the frequency of strong transverse confinement

90: %% and $C=1.4603\ldots$ \cite{Olshanii1998a}.},

91: \cite{note:coupling},

92: % $\a$ is the 1D scattering length

93: % \cite{Olshanii1998a} and 

94: $n$ is the line density, and $m$ is the particle mass.  The Landau

95: criterion of superfluidity \cite{pitaevskii03:book} predicts

96: a critical velocity $v_{\text{c}}$ for the breakdown of

97: dissipationless flow if no elementary excitations of momentum $p$ are

98: accessible with energy below $p v_{\text{c}}$ to dissipate its

99: energy. The excitation spectrum of the LL model \cite{lieb63:2}

100: contains umklapp excitations at finite momentum $2\pi n$ and energies which

101: tend to zero in a large system predicting a critical velocity of zero

102: by the Landau criterion {\it for any value of $\gamma$} [see the

103: lower thin (blue) line in Fig.~\ref{fig:DSFgamma20}]. On the other hand,

104: %superfluid response appears to be possible in the weakly-interacting

105: %quasi-1D Bose gas as found by approximate methods based on

106: Luttinger-liquid theory and instanton techniques~\cite{kagan00,buchler:100403}

107: predict superfluidity in the LL model for sufficiently small $\gamma$.

108: A resolution of this apparent

109: paradox lies in the probability of excitation by an infinitesimal

110: external perturber which is given by the dynamic structure factor (DSF)

111: $S(q,\omega)$, the Fourier transform of the time-dependent

112: density-density correlation function. 

113: A suppression of transitions in the vicinity of the um\-klapp excitation

114: was found in Refs.~\cite{sonin71,castro_neto94,pitaevskii04}.

115: In this Letter, we

116: report a calculation of the DSF,

117: %%  in the

118: %% thermodynamic limit where $N\to\infty$,

119: valid for large $\gamma$ as

120: shown in Fig.~\ref{fig:DSFgamma20}.

121: %%  shows the limiting

122: %% branches of the elementary excitation spectrum and the dynamic

123: %% structure factor in the thermodynamic limit where $N\to\infty$ as

124: %% calculated in this Letter. 

125: We find that even at fairly large but finite $\gamma$ the low energy

126: umklapp excitations are indeed suppressed while

127: higher-energy excitations are enhanced. This is to be

128: contrasted with the Tonks-Girardeau gas at $\gamma=\infty$ where all

129: excitations are equally well accessible and prevent superfluidity.

130: \begin{figure}

131: \includegraphics[width=\columnwidth]{DSFa015greyscaleD3.eps}

132: \caption{\label{fig:DSFgamma20} Excitation spectrum at $\gamma = 13$.

133: The upper and lower thin (blue) lines show the dispersions

134: $\omega_+(q)$ and $\omega_-(q)$ of Eq.~(\ref{eqn:dispersions}),

135: respectively, limiting the elementary excitations of the LL model. The

136: dynamic structure factor $S(q,\omega)$ in the approximation (\ref{eqn:strdyn})

137: is shown in shades of grey, between zero (white) and 1.0 $

138: N/(\hbar q \ef)$ (black). Here $\kf=\pi n$ and $\ef=\hbar^2 \kf^2 /(2

139: m)$. The the presence of a

140: $\delta$-function contribution is indicated by red dots.}

141: \end{figure}

142: 

143: Experiments have recently made significant progress in confining Bose

144: gases in 1D~\cite{bongs01}

145: %%%\cite{bongs01,goerlitz01,tolra:190401,moritz:250402} and

146: probing the strongly-correlated Tonks-Girardeau regime by increasing

147: interactions up to values of $\gamma\approx 5.5$ \cite{Kinoshita2004}

148: and to effective values of $\gamma_{\text{eff}}\approx 200$ in an

149: optical lattice \cite{paredes04}.

150: %%%The momentum

151: %%%distribution \cite{paredes04}, density profiles \cite{Kinoshita2004},

152: %%%and low-energy compressional excitation modes \cite{moritz:250402}

153: %%%have been the focus of the experimental studies.

154: In a recent

155: experiment~\cite{stoferle:130403}, the zero momentum

156: excitations of a 1D Bose gas in an optical lattice have been measured

157: by Bragg scattering, a technique that could also be used to measure the

158: DSF.

159: 

160: %% Another important correlation

161: %% function is the current-current correlator which gives direct

162: %% information about the superfluid density.

163: Although the LL model is exactly solvable, it is notoriously difficult to

164: calculate the various correlation functions. Many results in limiting

165: cases are summarized in the book~\cite{korepin93}, but the full

166: problem is not yet 

167: solved. Recently, progress has been made on various time-independent

168: correlation functions \cite{Gangardt2003a,olshanii02ep}.

169: While a wealth of information is available for small $\gamma$ where

170: Bogoliubov's perturbation theory can be applied, the

171: strongly-interacting regime was hardly accessible as a systematic

172: expansion in $\gamma^{-1}$ was lacking.

173: 

174: In this Letter we study the repulsively interacting 1D Bose gas in a

175: fermionic representation, introduced as follows.

176: The Bose-Fermi mapping of Girardeau \cite{girardeau60} can be generalized~\cite{cheon99} to

177: map the 1D Bose gas with pair interactions $V(x)=\gb

178: \delta(x)$ to an equivalent {\it interacting} Fermi system with the same

179: excitation energies and the same absolute values of the associated wave functions.

180: Hence,

181: any observable has the same value in both representations if it

182: derives from a local operator in coordinate space, {\it i.e.}~is a

183: function of the density operator $\hat{n}(x)=\sum_{i} \delta(x-x_{i})$.

184: In particular,

185: this is the case for the DSF $S(q,\omega)$, which is the

186: Fourier transform of the density-density 

187: correlation function~\cite{pines89,pitaevskii03:book}, and expresses the

188: probability to excite a particular excited state through a density

189: perturbation 

190: \begin{equation}

191: S(q,\omega)=

192: \sum_{n}|\langle0|\hat{\rho}_{q}|n\rangle|^{2}\delta(\hbar

193: \omega-E_{n}+E_{0}) , 

194: %\frac{1}{N}\int \frac{\d t}{2\pi}\,e^{i\omega t} \langle

195: %\hat{\rho}_{q}(t) \hat{\rho}^{\dag}_{q}(0)\rangle,

196: \label{sqomega}

197: \end{equation}

198: where $\hat{\rho}_{q}=\sum_i \exp(-i q x_i)$ is the Fourier component

199: of the density operator. 

200: In this mapping, large interactions between bosons correspond to

201: small interactions between fermions, which allows

202: for simple variational or perturbative

203: treatment. However, the short-range interaction of Cheon and Shigehara~\cite{cheon99}

204: is highly singular for this purpose.

205: It has been pointed out by Sen~\cite{sen99}

206: %%%\cite{sen99,sen03}

207: that the ground state energy functional can be derived variationally with 

208: the pair-interaction pseudopotential

209: \be \label{eqn:SenPP}

210:   V_{\rm Sen}(x_1,x_2) =

211: %  -a_{\text{1D}}\frac{\hbar^2}{m}\delta''(x_i-x_j),

212:   -\gf \delta''(x_1-x_2),

213: \ee

214: where $\gf = {2 \hbar^4}/({m^2 \gb})$ is the coupling constant in

215: the fermionic representation. This pseudopotential,

216: however, is applicable only  if the 

217: variational functions are continuous and vanish whenever two particle

218: coordinates coincide. This is the case for Slater determinants which

219: we will use later to derive the Hartree-Fock (HF) and Random-Phase

220: approximations (RPA) but not for

221: the exact fermionic wavefunctions \cite{cheon99}.

222: $V_{\rm Sen}$ thus generates the first order

223: correction to the  

224: ground-state energy easily

225: whereas Sen found a renormalizable divergence in second order

226: perturbation theory.

227: 

228: Recently, a more general pseudopotential was proposed by

229: Girardeau and Olshanii  \cite{girardeau03ep},

230: which is

231: also defined for 

232: discontinuous functions and generates the correct energy

233: functional for arbitrary values of the coupling constant. We give the

234: following useful representation of this 

235: pseudopotential in terms of an integral kernel:

236: \begin{align}

237: \nonumber 

238: &V(x_1,x_2;x'_1,x'_2)= \\& -{2}\gf \delta\left(

239: \frac{x_1+x_2-x'_1-x'_2}{2} \right) 

240: \delta'(x_1-x_2)\delta'(x'_1-x'_2). 

241: \label{eqn:our}

242: \end{align}

243: 

244: Based on a pseudopotential Hamiltonian (\ref{eqn:SenPP}) or (\ref{eqn:our}), we

245: can derive the HF approximation and the generalized

246: RPA giving access to the density-density

247: correlation function and the static and dynamic structure factors. The

248: results should be valid at least to first order in $\gamma^{-1}$ and thus be

249: useful near the Tonks-Girardeau limit of $\gamma =\infty$, where the

250: interacting boson problem maps exactly to a free Fermi gas

251: \cite{girardeau60}.

252: The purpose of this work is thus twofold: Besides predicting the behavior of

253: correlation functions of the 1D Bose gas, we also test the validity

254: and usefulness of the fermionic pseudopotential approach within the

255: RPA.

256: 

257: 

258: 

259: %% \subsubsection*{Hartree Fock}

260: 

261: 

262: The HF approximation for the fermionic system is derived 

263: variationally

264: %% from a

265: %% variational principle on the space of Slater determinant wavefunctions

266: in the standard way. We find that both pseudopotentials

267: (\ref{eqn:SenPP}) and (\ref{eqn:our}) lead to exactly the same

268: result: The HF equations for the single-particle orbitals $\varphi_j(x)$

269: take the usual form

270: \begin{equation}\label{eqn:HF}

271: \hat{F}\varphi_j(x)=\varepsilon_j\varphi_j(x).

272: %\label{eigen}

273: \end{equation}

274: Unlike the case of Coulomb interactions where the Fock operator

275: $\hat{F}$ is nonlocal with a local Hartree and a nonlocal exchange

276: term we find a purely local Fock operator

277: \begin{align} \label{eqn:HFoperator}

278: \hat{F}=&-\frac{\hbar^2}{2m}\frac{\partial^2}{\partial x^2}+V_{\rm

279: ext}(x)\nonumber\\

280: &+

281: \gf \left[n(x)\frac{\partial^2}{\partial x^2}+

282: {2}{\cal P}(x)i\frac{\partial}{\partial x}

283: -{\cal T}(x)\right].

284: % \label{onebody}

285: \end{align}

286: The first two terms on the right hand side are just the

287: single-particle part of the Hamiltonian. The mean-field parts in the

288: square bracket involve the single-particle density $n(x) = \sum_j n_j

289: {{\varphi}^{*}}_{j}(x)\varphi_{j}(x)$ and the derivative densities ${\cal

290:   P}(x)=-i\sum_j n_j {{\varphi}^{*}}'_{j}(x)\varphi_{j}(x)$

291: and ${\cal T}(x)=\sum_j n_j

292: [{\varphi}^{*}_{j}(x)\varphi''_{j}(x) +

293:   2{{\varphi}^{*}}'_{j}(x)\varphi'_{j}(x)]$, reminiscent of momentum

294: and energy densities, respectively.

295: Here, $\varphi' = \d \varphi/\d x$ and $n_j$ is the fermionic

296: occupation number

297: $n_j={1}/\{\exp[(\varepsilon_j-\mu)/(k_{\mathrm B} T)]+1\}$.

298: %% \end{equation}

299: At temperature $T=0$, $n_j$ defines the Fermi sea.

300: %%%as usual.

301: 

302: For the homogeneous gas ($V_{\rm ext}\equiv 0$), the quantum number

303: $j$ is associated with the particle momentum, and the 

304: single-particle orbital solutions of Eq.~(\ref{eqn:HF}) are plane waves with

305: energy and effective mass

306: \begin{align} \label{eqn:epsHF}

307: \varepsilon_q=&{\hbar^2 q^2}/{(2m^*)}-\gf {\pi^2} n^3/{3},\\

308: m^* =& m /\left(1-{2 \gf m n}/{\hbar^2}\right) = {m}/({1-4

309:    \gamma^{-1}}), 

310: \end{align}

311: respectively. The HF approximation for the ground-state energy 

312: coincides with the first two terms of the large-$\gamma$

313: expansion of the exact ground-state energy in the LL model

314: \cite{lieb63:1}. 

315: 

316: The HF approximation permits us to calculate the linear response

317: function of time-dependent HF, also known as RPA with exchange or

318: generalized RPA \cite{pines89}. 

319: We calculate the DSF at zero temperature by the relation \cite{pines89}

320: $S(q,\omega)= \im \{\chi(q,-\omega-i\varepsilon) / \pi\}$ for $\omega>0$ and

321: $S(q,\omega)=0$ for $\omega<0$ from the dynamic polarizability

322: $\chi(q,\omega+i\varepsilon)$. The function $\chi(q,\omega+i\varepsilon)$ determines the linear

323: density response to an external field \cite{pines89,pitaevskii03:book}

324: %%%and will be obtained

325: and can be obtained 

326: from the linearized equation of motion of the density operator in the

327: time-dependent HF approximation. In the Fourier representation of

328: momentum and frequency,

329: % ($q$ and $\omega$), 

330: algebraic equations result which can be solved for

331: $\chi(q,\omega)$.

332: In the thermodynamic limit,

333: %($N,L\to\infty$, $n=N/L=$const.), 

334: the sums over momenta

335: %% of Eqs.~(\ref{eqn:HFdens} --

336: %% \ref{eqn:Gamma})

337: are replaced by integrals.

338: We find $\chi(q, z) = {\chio(q, z)} /

339: \{(1-4\gamma^{-1}) [B-D\chio(q,z)]\}$, with $z=\omega+i\varepsilon$ and 

340: $\varepsilon {\to} +0$. 

341: %in Eq.~(ref{eqn:occNum}), 

342: For the DSF this yields for $\omega>0$

343: \begin{align}

344:   S(q,\omega)=&

345:     \frac{- \chi^{(0)}_{2}(q,\omega) B} {\pi (1-4\,\gamma^{-1})

346:     \left[\Big(B-D\chi_{1}^{(0)}\Big)^2 + 

347:     \Big(D\chi_{2}^{(0)}\Big)^2\right]}\nonumber\\

348:   & + \delta[\omega - \omega_0(q)] A(q) ,

349: \label{eqn:strdyn}

350: \end{align}

351: where

352: \begin{align}

353: B=&1-{4\,(3\,\gamma -16)}/{(\gamma-4)^3} ,\\

354:   D=&\frac{4 \ef}{N}\frac{\gamma}{(\gamma -4)^2}\Bigg\{\frac{q^2}{\pf^2}

355:   \frac{2\gamma - 9} {2 \gamma}

356:   - \frac{2}{\gamma}\\& -

357:  \left[\frac{\hbar (\omega+i\varepsilon)\, \pf}{

358:   \ef q} \right]^2 \frac{3\gamma -16}{2( \gamma  -4)^2}\Bigg\} ,

359: \end{align}

360: with the Tonks gas' Fermi wavenumber $\kf=\pi n$ and 

361: energy $\ef =  \hbar^2 \kf^2/(2 m)$. 

362: % frequency $\omega_{\mathrm F} = \hbar \kf^2/(2 m)$. 

363: The polarizability $\chio = \chio_1 +i \chio_2$ of the ideal 1D Fermi

364: gas with renormalized mass.

365: is given by the real and imaginary parts

366: \begin{align}

367:   \chi^{(0)}_{1}(q,\omega)=&\frac{N m^{*}}{2\hbar^2 q \kf}\ln\left|

368:   \frac{\omega^{2} - \omega^{2}_{-}(q)}{\omega^{2}_{+}(q)-\omega^{2}}

369:   \right|, \label{chi1}\\

370:   \chi^{(0)}_{2}(q,\omega)=&- \frac{N \pi m^{*}}{2\hbar^2 q \kf}\left\{

371:   \begin{array}{ll} 

372:   \pm 1,& \omega_{-}\le \pm \omega\le \omega_{+},\\[2mm]

373:   0,& \text{else,}

374: \end{array}\right.

375: \label{chi2}

376: \end{align}

377: where the dispersion relations

378: \be \label{eqn:dispersions}

379:   \omega_\pm(q) = {\hbar |2 \pf q \pm q^2|}/{(2 m^*)}

380: \ee

381: border the 

382: continuum part of the accessible excitation spectrum made up from HF

383: quasiparticle-quasihole excitations~(\ref{eqn:epsHF}), see Fig.~\ref{fig:DSFgamma20}.

384: % energy-momentum regime of accessible excitations of

385: % particle-hole type where the continuum part of the DSF  differs from

386: % zero in the current approximation. 

387: The $\delta$-function part of the DSF (\ref{eqn:strdyn}) relating to

388: discrete excitations of collective character lies outside of the

389: continuum part and comes from possible zeros in the denominator of

390: $\chi(q,z)$. It is determined by the solution $\omega_0(q)$ of

391: the transcendental equation $B-D\chi_{1}^{(0)}=0$. We have solved this

392: equation in various limits and found that at most one solution for

393: $\omega_0(q)$ may exist. The strength $A(q)$ is given by the residue

394: of $\chi$ at the pole $z=\omega_0(q)-i\varepsilon$. For small

395: $\gamma^{-1}$, the strength $A(q) \simeq 2 N \gamma \exp(-\gamma\q/\kf)$

396: is exponentially suppressed and possible solutions are close to the

397: dispersions $|\omega_0-\omega_\pm| \propto

398: \exp(-\gamma\q/\kf)$. Due to this proximity of the discrete and

399: continuous parts and expected smearing of discrete contributions

400: beyond RPA, we may conjecture that the

401: $\delta$-function should be seen as part of the continuum, enhancing

402: contributions near the border. In fact, we know from the exact

403: solutions that energy 

404: spectrum is continuous \cite{lieb63:2}.

405: 

406: %%%Numerical values for $A(q)\omega_0(q)$ are shown in

407: %%%Fig.~\ref{fig:SRCont}.

408: At finite $\gamma$ we find a $\delta$-function

409: contribution below and close to $\omega_-$ for small $q$ whereas for

410: large $q$ there is a discrete contribution at energies larger than

411: $\omega_+$, see Fig.~\ref{fig:DSFgamma20}. In the limit $\q \to \infty$ at finite $\gamma$, the

412: $\delta$ part completely determines the DSF as the continuum part

413: vanishes, and asymptotically  $A\simeq N$, and  $\omega_0 \simeq \hbar

414: q^2/(2 m)$ becomes 

415: the free particle dispersion reminiscent of the DSF for the weakly

416: interacting Bose gas in Bogoliubov theory \cite{pitaevskii03:book}. A

417: more detailed analysis of the $\delta$-function part will be given

418: elsewhere \cite{cherny04tbp}.

419: 

420: 

421: The branches $\omega_\pm$ as shown in Fig.~\ref{fig:DSFgamma20} are

422: large $\gamma$ approximations for the type I and II elementary

423: excitation branches, respectively, introduced by Lieb

424: \cite{lieb63:2}. In accordance with the exact results, both branches

425: share the same slope at the origin and give rise to a single speed of

426: sound given by 

427: $ v_s \equiv {\d \omega_\pm}/{\d k} = {\hbar \kf}/{m^*}$, which is the

428: correct first order expansion of $v_s$ for large $\gamma$

429: \cite{lieb63:2}. Note that the usual Bogoliubov perturbation theory

430: for weakly interacting Bosons gives a similar expansion of $v_s$ for

431: small $\gamma$ and the type I excitation branch. Type II excitations

432: are not described in the Bogoliubov theory.

433: 

434: \begin{figure}

435: \includegraphics[width=\columnwidth]{Fig2.eps}

436: \caption{\label{fig:DSFk2} DSF $S(q,\omega)$ as a function of

437: $\omega$ at $q=0.4 \kf$ [(a) and (b)] and $q=2 \kf$ [(c) and (d)],

438: and at interaction strength $\gamma=13.33$ [(a) and (c)] and

439: $\gamma=40$ [(b) and (d)]. The solid (black) line shows the RPA

440: result~(\ref{eqn:strdyn}), the dashed (red) line shows the first order

441: expansion in $\gamma^{-1}$~(\ref{eqn:first-oder}), and the dotted

442: (blue) line shows the DSF of the Tonks limit ($\gamma=\infty$) for

443: comparison. While the 

444: first-order expansion always has a divergence to $+\infty$ near

445: $\omega_+$, the RPA result shows an enhancement of low-energy

446: excitation near $\omega_-$ at small $\q$. Note the unphysical

447: negative values of the DSF in the first order expansion near

448: $\omega_-$, in particular for the umklapp excitations in panel (c).}

449: \end{figure}

450:  

451: Due to a logarithmic

452: singularity in $\chi_{1}^{(0)}$, the DSF vanishes on the dispersion

453: curves. For the Tonks gas at $\gamma\to\infty$, the value of the

454: DSF within these limits is independent of $\omega$ and takes the value

455: of $N m/(2 \pi \hbar^2 q n)$.

456: %%%as discussed in Ref.~\cite{pitaevskii03:book}. 

457: The energy-dependence in the

458: RPA for finite $\gamma^{-1}$ is shown in

459: Figs.~\ref{fig:DSFgamma20} and \ref{fig:DSFk2}. In particular, we see

460: that the umklapp excitations at $q=2 \kf$ at

461: small $\omega$, prohibiting superfluidity of the Tonks gas, are being

462: suppressed. We find that $S(2\kf,\omega)$ in the RPA approaches zero as

463: $1/\ln^2(\hbar\omega/\ef)$, in contrast to the results~\cite{castro_neto94,pitaevskii04},

464: predicting a power-law dependence on

465: $\omega$ for finite $\gamma$ based on a pseudoparticle-operator approach.

466: 

467: The RPA result may be expanded in $1/\gamma$,

468: which yields

469: \be\label{eqn:first-oder}

470:   S(q,\omega)\frac{\ef}{N} = \kf \frac{1+ 8\gamma^{-1}}{4 q} +

471:    \frac{\ln 

472:   f(q,\omega)}{2\gamma} + {\cal O}(\gamma^{-2})

473: \ee

474: with $f(q,\omega) = |(\omega^2 - \omega_-^2)/(\omega_+^2 - \omega^2)|$.

475: This expansion is supposed to be consistent with first order

476: perturbation theory. However, it can assume negative values as seen in

477: Fig.~\ref{fig:DSFk2} although $S(q,\omega)$ is known to be strictly

478: non-negative, a property that is fulfilled by our RPA result

479: (\ref{eqn:strdyn}). Close to $\omega_+$, the first order expansion has

480: has a logarithmic singularity to $+\infty$ which may be a precursor of

481: the dominance of Bogoliubov-like excitations in the DSF at small

482: $\gamma$. In the RPA this effect is even more pronounced due to a

483: strong and narrow peak of the DSF in the RPA near $\omega_+$ at large

484: momenta as seen in Fig.~\ref{fig:DSFk2}. At finite gamma and for small

485: momenta, however, the RPA predicts a peak near $\omega_-$, contrary

486: to the first-order result. Whether this effect is real or an artefact

487: of the RPA is not obvious and may be decided by more accurate

488: calculations or experiments. Spurious higher order terms in the RPA

489: and an improved approximation scheme have been discussed in

490: Ref.~\cite{brand98}. On the other hand, Roth and Burnett have recently

491: observed a qualitatively similar effect in numerical calculations of

492: the DSF of the Bose-Hubbard model \cite{roth04}.

493: 

494: 

495: We have verified numerically that the $f$-sum rule $m_1\equiv \hbar^2

496: \int \omega 

497: S(q,\omega) \d \omega = N \hbar^2 q^2/(2 m)$ is fulfilled, which

498: should be an exact statement for the RPA from a general theorem

499: \cite{thouless61}. The same follows also from the the large $\omega$

500: asymptotics of $\chi(q,\omega)\simeq 2m_{1}/(\hbar \omega)^2$ when $\chi$ is

501: analytic as a function of $\omega$ in the upper half complex plane

502: \cite{pitaevskii03:book}. Our approximation breaks down when $\gamma\lesssim 8$

503: for small values of momenta $q\lesssim \pf$,

504:  where imaginary poles of $\chi$ appear.

505: 

506: 

507: Our result (\ref{eqn:strdyn}) defines approximations for the static

508: structure factor $S(q) = ({\hbar}/{N}) \int S(q,\omega) \d \omega$ and

509: the pair correlation function $g(x) = 1+\int [S(q) - 1]/(2 \pi n)

510: \exp(iqx) \d x$ that will be discussed in detail elsewhere

511: \cite{cherny04tbp}. Here we only note that the approximation for

512: $S(q)$ is continuous and has the low-momentum expansion $S(q) = q [1+4 \gamma^{-1}-

513: q^2/(\kf^2 \gamma)]/(2 \kf) +{\cal O}(q^5)/\gamma +{\cal

514: O}(\gamma^{-2})$. We also find that $g(x=0)$ vanishes in first order

515: of $\gamma^{-1}$ as consistent with Ref.~\cite{Gangardt2003a},

516: indicating once more the validity of our results.

517: 

518: 

519: Summarizing, we have derived variational approximations that should

520: yield valid expansions of various correlation functions of the 1D Bose

521: gas in $1/\gamma$, making predictions in a so-far unexplored

522: regime. These approximations prove consistent with known limits and

523: sum rules and therefore establish the usefulness of the fermionic

524: pseudopotentials (\ref{eqn:SenPP}) and (\ref{eqn:our}). We have also

525: shown a rare example where a fully analytical calculation of the

526: generalized RPA can be carried out for a non-trivial many-body system.

527: Our results have immediate relevance to current experimental

528: endeavors to explore fermionization in the strongly-interacting 1D

529: Bose gas as they indicate qualitative deviation in the DSF from the

530: Tonks gas limit already for very small $\gamma^{-1}$. The more

531: accessible region of intermediate values of $\gamma$, which would

532: yield more insight into the superfluid properties, is beyond the scope

533: of this study. Answers may come from higher order or

534: density-functional-theory based approximations extending the work of

535: Ref.~\cite{brand04a}, from numerical calculations, or from experimental

536: measurements. Obvious extensions of the present study to finite

537: temperature, inhomogeneous systems, and periodic lattice potentials are under way.

538: 

539: The authors acknowledge discussions with Sandro Stringari who inspired

540: their interest in this problem.

541: 

542: %\bibliographystyle{prsty}

543: %\bibliography{/home/joachim/tex/refs_bec,/home/joachim/tex/refs,/home/joachim/tex/bec,notes}

544: 

545: \begin{thebibliography}{10}

546: 

547: \bibitem{lieb63:1}

548: E.~H. Lieb and W. Liniger, Phys. Rev. {\bf 130},  1605  (1963).

549: 

550: \bibitem{lieb63:2}

551: E.~H. Lieb, Phys. Rev. {\bf 130},  1616  (1963).

552: 

553: \bibitem{girardeau60}

554: M.~D. Girardeau, J. Math. Phys. {\bf 1},  516  (1960).

555: 

556: \bibitem{note:coupling}

557: The Boson coupling constant is related to the 1D and 3D $s$-wave scattering

558:   lengths, $\a$ and $a_{\mathrm{3D}}$, respectively by $\gb = -2\hbar^2 / (m

559:   \a) = 2 a_{\mathrm{3D}} \omega_\rho \hbar [1-C a_{\mathrm{3D}}

560:   \sqrt{m\omega_\rho/(2\hbar)}]$, where $\omega_\rho$ is the frequency of

561:   transverse confinement and $C\approx 1.4603$. M. Olshanii, Phys. Rev. Lett.

562:   {\bf 81}, 938 (1998) and A. Yu. Cherny and J. Brand, Phys. Rev. A, in print

563:   (2004).

564: 

565: \bibitem{pitaevskii03:book}

566: L. Pitaevskii and S. Stringari, {\em {B}ose-{E}instein Condensation}

567:   (Clarendon, Oxford, 2003).

568: 

569: \bibitem{kagan00}

570: Y. Kagan, N.~V. {Prokof'ev}, and B.~V. Svistunov, Phys. Rev. A {\bf 61},

571:   045601  (2000).

572: 

573: \bibitem{buchler:100403}

574: H.~P. B\"uchler, V.~B. Geshkenbein, and G. Blatter, Phys. Rev. Lett. {\bf 87},

575:   100403  (2001).

576: 

577: \bibitem{sonin71}

578: E.~B. Sonin, Sov. Phys. JETP,  {\bf 32}, 773 (1971).

579: 

580: \bibitem{castro_neto94}

581: {A. H. Castro Neto} {\it et~al.}, Phys. Rev. B  {\bf 50},  14032  (1994).

582: 

583: \bibitem{pitaevskii04}

584: {G.~E.~Astrakharchik} and L.~P.~Pitaevskii,  Phys. Rev. A

585:   {\bf 70},  013608  (2004).

586:   

587: \bibitem{bongs01}

588: K. Bongs {\it et~al.}, Phys. Rev. A {\bf 63},  031602(R)  (2001);

589: %\bibitem{goerlitz01}

590: A. G{\"o}rlitz {\it et~al.}, Phys. Rev. Lett. {\bf 87},  130402  (2001);

591: %\bibitem{tolra:190401}

592: B.~L. Tolra {\it et~al.}, {\it ibid.} {\bf 92},  190401  (2004);

593: %\bibitem{moritz:250402}

594: H. Moritz {\it et~al.},  {\it ibid.} {\bf 91},  250402  (2003).

595: 

596: \bibitem{Kinoshita2004}

597: T. Kinoshita, T. Wenger, and D.~S. Weiss, Science {\bf 305},  1125  (2004).

598: 

599: \bibitem{paredes04}

600: B. Paredes {\it et~al.}, {N}ature {\bf 429},  277  (2004).

601: 

602: \bibitem{stoferle:130403}

603: T. St\"oferle {\it et~al.}, Phys. Rev. Lett. {\bf 92},  130403  (2004).

604: 

605: \bibitem{korepin93}

606: V.~E. Korepin, N.~M. Bogoliubov, and A.~G. Izergin, {\em Quantum Inverse

607:   Scattering Method and Correlation Functions} (University Press, Cambridge,

608:   1993).

609: 

610: \bibitem{Gangardt2003a}

611: D.~M. Gangardt and G.~V. Shlyapnikov, Phys. Rev. Lett. {\bf 90},  010401

612:   (2003).

613: 

614: \bibitem{olshanii02ep}

615: M. Olshanii and V. Dunjko, \eprint{cond-mat/0210629}, 2002.

616: 

617: \bibitem{cheon99}

618: T. Cheon and T. Shigehara, Phys. Rev. Lett. {\bf 82},  2536  (1999).

619: 

620: \bibitem{pines89}

621: D. Pines and P. Nozi\`eres, {\em The theory of quantum liquids}

622:   (Addison-Wesley, Redwood City, 1989).

623: 

624: \bibitem{sen99}

625: D. Sen, Int. J. Mod. Phys. {\bf 14},  1789  (1999);

626: %\bibitem{sen03} D. Sen,

627: J. Phys. A {\bf 36},  7517  (2003).

628: 

629: \bibitem{girardeau03ep}

630: M.~D. Girardeau and M. Olshanii, \eprint{cond-mat/0309396}, 2003, and Phys.

631:   Rev. A {\bf 70}, 023608 (2004).

632: 

633: \bibitem{cherny04tbp}

634: A.~Y. Cherny and J. Brand (unpublished).

635: 

636: \bibitem{brand98}

637: J. Brand and L.~S. Cederbaum, Phys. Rev. A {\bf 57},  4311  (1998).

638: 

639: \bibitem{roth04}

640: R. Roth and K. Burnett, J. Phys. B {\bf 37},  3893  (2004).

641: 

642: \bibitem{thouless61}

643: D.~J. Thouless, Nucl. Phys. {\bf 22},  78  (1961).

644: 

645: \bibitem{brand04a}

646: J. Brand, J. Phys. B {\bf 37},  S287  (2004).

647: 

648: \end{thebibliography}

649: 

650: 

651: \end{document}

652: