1: \documentclass[prl, amsmath, showpacs, preprintnumbers,

2: superscriptaddress, twocolumn, sort&compress]{revtex4}

3: 

4: \usepackage{graphicx}

5: \usepackage{mathptmx}

6: \usepackage[latin1]{inputenc}

7: 

8: \hyphenation{Fesh-bach}

9: 

10: \begin{document}

11: \title{Long-lived Feshbach molecules in a 3D optical lattice}

12: \author{G. Thalhammer}

13: \author{K. Winkler}

14: \author{F. Lang }

15: \author{S. Schmid}

16: \affiliation{Institut f\"ur Experimentalphysik, Universit\"at

17:   Innsbruck, % Technikerstra{\ss}e 25,

18:   6020 Innsbruck, Austria}

19: 

20: \author{R. Grimm}%

21: \affiliation{Institut f\"ur Experimentalphysik, Universit\"at

22:   Innsbruck, % Technikerstra{\ss}e 25,

23:   6020 Innsbruck, Austria}%

24: \affiliation{Institut f\"ur Quantenoptik und Quanteninformation,

25:   \"Osterreichische Akademie der Wissenschaften, 6020 Innsbruck,

26:   Austria}

27: 

28: \author{J. Hecker Denschlag}%

29: \affiliation{Institut f\"ur Experimentalphysik, Universit\"at

30:   Innsbruck, % Technikerstra{\ss}e 25,

31:   6020 Innsbruck, Austria}

32: 

33: \date{\today}

34: 

35: \pacs{32.80.Lg, % Mechanical effects of light on atoms, molecules and ions

36:   34.50.-s, % Scattering of atoms and molecules

37:   03.75.Lm, % Tunneling, Josephson effect, Bose-Einstein condensates

38:             % in periodic potentials, solitons, vortices and

39:             % topological excitations

40:   03.75.Nt % Other Bose-Einstein condensation phenomena

41: }

42: 

43: \begin{abstract}

44:   We have created and trapped a pure sample of $^{87}$Rb$_2$ Feshbach

45:   molecules in a three-dimensional optical lattice.  Compared to

46:   previous experiments without a lattice we find dramatic improvements

47:   such as long lifetimes of up to 700\,ms and a near unit efficiency

48:   for converting tightly confined atom pairs into molecules. The

49:   lattice shields the trapped molecules from collisions and thus

50:   overcomes the problem of inelastic decay by vibrational quenching.

51:   Furthermore, we have developed a novel purification scheme that

52:   removes residual atoms, resulting in a lattice in which individual

53:   sites are either empty or filled with a single molecule in the

54:   vibrational ground state of the lattice.

55: \end{abstract}

56: 

57: \maketitle

58: 

59: Using magnetic Feshbach resonances \cite{feshbach} to create ultracold

60: diatomic molecules in their highest ro-vibrational state has become a

61: key to exciting developments and breakthroughs. Above all, it is the

62: highly versatile control of the interactions and binding energy of

63: these weakly bound molecules which opens the door to various

64: fundamental systems such as strongly-interacting quantum-degenerate

65: Fermi gases \cite{Che05} or Efimov few-body states \cite{Bra03}.

66: 

67: Feshbach molecules made of bosonic atoms behave in a strikingly

68: different way from Feshbach molecules made of fermionic atoms.  For

69: weakly bound dimers of fermionic atoms, vibrational quenching and

70: inelastic decay is strongly suppressed by a Pauli blocking effect in a

71: close encounter of two molecules \cite{Pet04}. This has been vital to

72: the experimental creation of molecular Bose-Einstein condensates (BEC)

73: and investigations of the crossover to a strongly interacting

74: fermionic superfluid \cite{Che05}. For dimers of bosonic atoms

75: \cite{Don02a,Her03,Xu03,Dur04}, however, progress has been hampered by

76: strong inelastic decay due to atom-molecule and molecule-molecule

77: collisions. Therefore the experiments have been focussed on the

78: transient regime, studying e.g. the collision and dissociation

79: dynamics \cite{Dur04,Xu03,Muk04,Dur04b,Hod05,Chi05}.

80: 

81: A three-dimensional optical lattice offers many interesting

82: opportunities for research on ultracold molecules. Lattice sites

83: occupied with exactly two atoms represent a perfectly controlled

84: quantum system which can be rigorously treated theoretically.  Matrix

85: elements for atom-molecule coupling are strongly enhanced with the

86: prospect of efficient atom-molecule conversion. Moreover, it is

87: expected that the lattice can isolate molecules from each other and

88: shield them from detrimental collisions so that a long-lived sample

89: can be created also with dimers of bosonic atoms. Recently, first

90: experiments with molecules in a lattice have studied photoassociation

91: \cite{Rom,Ryu} or demonstrated modifications of the binding energy of

92: tightly confined Feshbach molecules \cite{Stoeferle}.

93: 

94: In this Letter, we report on the creation of a pure sample of

95: ultracold Rb$_2$ Feshbach molecules trapped in a 3D optical lattice.

96: The observed long lifetimes of up to 700\,ms greatly exceed previous

97: values reported for dimers of bosonic atoms

98: \cite{Dur04,Xu03,Muk04,Dur04b}.  Further, we experimentally

99: investigate association and dissociation of the Feshbach molecules and

100: reach efficiencies of 95\% for converting pairs of atoms into

101: molecules. In brief, we adiabatically load a $^{87}$Rb BEC into the

102: vibrational ground state of the lattice. For our experimental

103: conditions, about 20\% of the condensate atoms are grouped in pairs of

104: two into the lattice sites. By ramping adiabatically over a magnetic

105: Feshbach resonance at 1007.4\,G we convert these pairs into molecules.

106: Another 20\% percent of atoms are located in triply and more highly

107: occupied lattice sites. After the Feshbach ramp, however, inelastic

108: collisions between the created molecules and atoms within the high

109: occupancy sites quickly remove these particles from the lattice.

110: Finally, the remaining 60\% of the condensate atoms are found in

111: singly occupied sites and are unaffected by the Feshbach ramp. Using a

112: novel resonant purification scheme we can remove these atoms from the

113: lattice, which results in a pure molecular sample with each molecule

114: being shielded from the others by the lattice potential.

115: 

116: The starting point for our experiments is an almost pure BEC of about

117: $6\times 10^5$ $^{87}$Rb atoms in the spin state $| F = 1, m_F = -1

118: \rangle$ \cite{Tha05}. It is transferred from a quadrupole Ioffe

119: configuration trap (QUIC) into a magnetic trap \cite{movBEC} with trap

120: frequencies $\omega_{x,y,z} = 2\pi\times (7, 19, 20)$ Hz, leading to a

121: peak density of the BEC of about $4\times 10^{13}$\,cm$^{-3}$.  Our 3D

122: lattice is cubic and consists of three retro-reflected

123: intensity-stabilized laser beams which propagate orthogonally to each

124: other. They are derived from a frequency-stable single-mode

125: Ti:Sapphire laser ($\approx$ 500 kHz linewidth) with a wavelength of

126: $\lambda = $ 830.44\,nm. For this wavelength, the laser is detuned by

127: about 100 GHz from the closest transition to an excited molecular

128: level, minimizing light induced losses as a precondition for long

129: molecular lifetimes. The laser beams are polarized perpendicularly to

130: each other and their frequencies differ by several tens of MHz to

131: avoid disturbing interference effects. The waists of all three beams

132: are about 160\,$\mu$m and the maximum obtainable power is about

133: 130\,mW per beam, which results in calculated lattice depths of up to

134: 40 recoil energies ($E_{r}= h^2 / 2m \lambda^{2}$, where $m$ is the

135: atomic mass of $^{87}$Rb and $h$ is Planck's constant). We have

136: verified the lattice depths by measuring the energy gap between bands

137: of the lattice \cite{denschlag}. The relative uncertainty of our

138: lattice depth is $\pm 15\%$.

139: 

140: After the BEC is adiabatically loaded into a 35 $E_r$ deep 3D optical

141: lattice within 100\,ms, we turn off the magnetic trap. By suddenly

142: reversing the bias magnetic field of a few G we flip the spins of our

143: atoms to the high field seeking state $| F = 1, m_F = +1 \rangle$ with

144: an efficiency higher than 99\% (see also \cite{Vol03}). This state

145: features the Feshbach resonance at 1007.4 G. Afterwards we ramp up a

146: homogeneous magnetic field in 3\,ms to about 1015\,G using the QUIC

147: quadrupole coils in Helmholtz configuration. The current through the

148: coils is actively stabilized to a relative accuracy of about

149: $10^{-4}$. The fast diabatic crossing of the Feshbach resonance has

150: basically no effect on the atoms in the lattice. If we slowly ramp in

151: 5\,ms from 1015\,G to 1000\,G (crossing the Feshbach resonance at

152: 1007\,G) molecules are adiabatically produced in the multiply occupied

153: lattice sites. If, however, we cross the Feshbach resonance very

154: quickly e.g. by simply switching off the magnetic field, less than

155: $10\%$ of the atoms are converted into molecules.  Note, that after

156: the first Feshbach ramp, we observe an immediate irretrievable loss of

157: 20\% of the atoms. We attribute this loss to inelastic collisions

158: involving molecules for sites initially occupied by 3 or more atoms.

159: The remaining occupied sites contain each either a single atom or a

160: single molecule.

161: 

162: Atom numbers are measured with absorption imaging at low magnetic

163: fields ($\approx$ 2G) after release from the optical lattice and

164: 11\,ms of ballistic expansion. In order to determine molecule numbers,

165: they are first dissociated into atoms by slowly ramping back across

166: the Feshbach resonance and then quickly switching off the magnetic

167: field.  We also use absorption imaging to map out the band occupation

168: of the lattice.  For this, the lattice is ramped down in 2\,ms and we

169: typically observe a momentum distribution which is fully contained in

170: a cube of width $2\,\hbar k$ corresponding to the first Brillouin zone

171: of the lattice \cite{greiner01}. This demonstrates that atoms and

172: molecules are in the vibrational ground state of the lattice sites.

173: 

174: In order to create a pure molecular sample, we have developed a novel

175: purification scheme to remove all atoms. We apply a combined microwave

176: and light pulse at a magnetic field of 1000\,G for 3\,ms. The

177: microwave drives the transition at a frequency of 9113\,MHz between

178: levels which correlate with $| F = 1, m_F = +1 \rangle$ and $|F=2, m_F

179: = +2 \rangle$. The light pulse drives the closed transition $| F = 2,

180: m_F = +2 \rangle \rightarrow | F = 3, m_F = +3 \rangle$. The optical

181: transition frequency is 1402\,MHz blue detuned compared to the

182: transition at zero magnetic field. After this pulse which heats the

183: atoms out of the lattice and an additional hold time ($\sim$20\,ms),

184: no more atoms can be detected. The direct effect of the microwave and

185: light field pulse on the molecules is negligible because the radiation

186: is off resonance.  As an indirect effect, however, we find that during

187: the first purification pulse we still lose about 40\% of the

188: molecules, probably due to inelastic collisions with the blown away

189: atoms. Further losses are not observed in subsequent purification

190: pulses. We end up with a pure molecular sample formed from about 10\%

191: of the initial atoms, which corresponds to 3$\times 10^4$ molecules.

192: 

193: \begin{figure}

194:   \includegraphics{figure1}

195:   \caption{Decay of molecules in a 3D optical lattice with a potential

196:     depth of $(10\pm 2)\,E_r$. Shown is the remaining fraction of

197:     molecules in purified (filled circles) and unpurified (squares)

198:     samples as a function of hold time.  The continuous lines are

199:     exponential fits to the data indicating a lifetime of 700\,ms and

200:     35\,ms for purified and unpurified molecular samples,

201:     respectively.  In order to determine molecule numbers in the

202:     unpurified sample, purification was performed at the end of the

203:     hold time.  The hold time was limited to 400 ms due to the

204:     heating-up of the coils.  }

205:   \label{fig1}

206: \end{figure}

207: 

208: We have investigated the lifetimes of the Feshbach molecules in the

209: lattice under various conditions (see Figs.~\ref{fig1} and

210: \ref{fig2}). Figure~\ref{fig1} shows the decay of molecules at a

211: lattice depth of 10 $E_r$. The pure molecular sample exhibits a

212: remarkably long lifetime of 700\,ms. For the case of an unpurified

213: sample, where the atoms at singly occupied sites have not been

214: removed, the lifetime of the molecules is considerably reduced to

215: $\approx$35\,ms. This observation suggests that the molecular decay is

216: based on a process where an atom tunnels to a site occupied by a

217: molecule and inelastically collides with it. These inelastic

218: collisions can in principle also happen between two molecules.

219: However, compared to an atom, a molecule has a much lower tunnelling

220: rate, since it experiences twice the dipole potential and has twice

221: the mass than a single atom. Using simple scaling arguments, for a

222: molecule to have the same tunneling rate as an atom, the lattice light

223: intensity needs to be about four times smaller. This explains the

224: comparatively long lifetime of the purified molecular samples. We note

225: that if molecular decay is based on inelastic collisions, its time

226: dependence is intrinsically non-exponential. However, exact modelling

227: of the decay would be quite involved and requires precise knowledge of

228: atom/molecule distributions in the lattice. Since these distributions

229: are not known to us, we simply base our estimates for the molecular

230: lifetimes on an exponential decay law.

231: 

232: \begin{figure}

233:   \includegraphics{figure2}

234:   \caption{Molecular lifetimes for purified molecular samples

235:     (circles) and for unpurified samples (squares) as a function of

236:     the lattice depth. For this measurement the lattice depth was

237:     reduced from $35\,E_r$ to the given value after the creation of

238:     the molecules.  The dashed line is inversely proportional to the

239:     lattice depth.}

240:   \label{fig2}

241: \end{figure}

242: 

243: Figure~\ref{fig2} shows the measured lifetimes of the molecules for

244: various lattice depths. For sufficiently high lattice depths we

245: observe a lifetime for the purified molecular sample inversely

246: proportional to the lattice depth (see dashed line in

247: Fig.~\ref{fig2}). From this we conclude that above a lattice depth of

248: about $12\,E_r$ the tunnelling of the molecules is strongly suppressed

249: and the lifetime is limited by light induced losses.  Below this

250: value, decay is dominated by tunnelling and inelastic collisions. Thus

251: the molecular lifetime is maximized in a tradeoff between tunnelling

252: and light induced losses.  As already shown in Fig.~\ref{fig1}, the

253: presence of atoms considerably reduces the lifetime of the molecules,

254: even at larger lattice depths. In the limit of vanishing lattice

255: depths our experimental lifetimes decrease to values similar to those

256: observed in references \cite{Xu03,Muk04,Hod05,Dur04,Dur04b}.

257: Figure~\ref{fig2} clearly demonstrates that shielding of the molecules

258: against inelastic collisions grows with increasing lattice depth.

259: 

260: We now investigate the dynamics for both association and dissociation

261: of a single Feshbach molecule in a lattice site during Feshbach

262: ramping. This fundamental system is of special interest since it can

263: be theoretically treated exactly and solved analytically \cite{Jul04}.

264: We prepare a purified sample of molecules at 1000 G in a lattice of 35

265: $E_r$ depth. We then ramp the magnetic field in a symmetric way across

266: the Feshbach resonance up to 1015 G and back (see Fig.~\ref{fig3}).

267: Afterwards, purification

268: \begin{figure}

269:   \includegraphics{figure3}

270:   \caption{(a) Scheme for measurement of conversion efficiency, shown

271:     for 7 dissociation/association cycles. The shaded areas indicate

272:     the application of our purification procedure to remove atoms. The

273:     dashed line at 1007\,G shows the position of the Feshbach

274:     resonance.  (b) Conversion efficiency for given number of complete

275:     dissociation/association cycles for two different ramp speeds of

276:     the magnetic field.  We measure a conversion efficiency of 95\%

277:     per cycle for the slow ramp and 89\% per cycle for the fast ramp.

278:     The solid lines are described by exponential fitcurves as

279:     described in the text.  The lattice depth is 35 $E_r$.  }

280:   \label{fig3}

281: \end{figure}

282: is applied to remove atoms which have not recombined to form

283: molecules.  In a last step, the molecule number is measured. If

284: dissociation and association are not fully adiabatic in a conversion

285: cycle, a loss of molecules will result, e.g. during the association

286: ramp a pair of atoms might not be converted into a molecule or during

287: dissociation the molecule might break up into two atoms which, after

288: tunnelling, are located in separate sites.  For slow ramps we observe

289: small loss signals indicating high adiabaticity for the

290: dissociation/association cycle. In order to increase the loss signal,

291: to improve its accuracy and to check for consistency we repeated this

292: experiment with a higher number of cycles (see Fig.~\ref{fig3}). The

293: two data sets in (Fig.~\ref{fig3}b) correspond to two different ramp

294: speeds ($2\times 10^3\,\text{G/s}$, $8\times 10^3\,\text{G/s}$) and

295: can be described by the exponential functions $0.95^n$ and $0.89^n$,

296: respectively, where $n$ is the number of cycles.  Thus, for a slow

297: Feshbach ramp ($2\times 10^3\,\text{G/s}$) we observe an unprecedented

298: high efficiency of up to 95\% for the whole dissociation/association

299: cycle. For a faster ramp ($8\times 10^3\,\text{G/s}$) the efficiency

300: drops to 89\%.  We have taken care that light induced losses have been

301: corrected for in the data (Fig.~\ref{fig3}b). Our high conversion

302: efficiencies in the optical lattice are in strong contrast to the low

303: values of $ \sim $10\% observed previously in a $^{87}$Rb BEC

304: \cite{Dur04b} which were presumably limited by strong inelastic

305: collisions. In our deep lattice, however, inelastic collisions are

306: suppressed.

307: 

308: After having determined the efficiency for the full

309: dissociation/association cycle, we now study dissociation and

310: association individually. Figure~\ref{fig4}a shows the measured

311: conversion efficiency of atom pairs to molecules for different ramp

312: speeds. The atom pairs were prepared by creating a pure molecular

313: sample and then dissociating the molecules by slowly ($2 \times

314: 10^3$G/s) ramping backward over the Feshbach resonance.  Then again

315: the magnetic field was swept across the Feshbach resonance at various

316: speeds and finally after switching off completely the magnetic field,

317: the remaining, non converted atoms are detected. The dashed line in

318: Fig.~\ref{fig4}a is based on a Landau-Zener expression without

319: adjustable parameters \cite{Jul04} and is given by

320: \begin{equation}

321:   p = 1-\exp\left(  -

322:     \frac{2\sqrt{6}\hbar}{m \ a^{3}_\text{ho}}%

323:     \left|\frac{a_\text{bg} \ \Delta B}{\dot{B}}\right|\right),

324:   \label{equ:formel}

325: \end{equation}

326: where $p$ is the probability of creating a molecule, $a_\text{bg} =

327: 100.5\,a_0$ the background scattering length, $\Delta B =

328: 0.21\,\text{G}$ \cite{Dur04b} the width of the Feshbach resonance,

329: $\dot B$ the ramp speed at the Feshbach resonance, and $a_\text{ho} =

330: \sqrt{{\hbar}/{m\omega}}$ the harmonic oscillator length. Using the

331: best estimate for our trapping frequency of $\omega = 2\pi \times (39

332: \pm 3)\,\text{kHz}$ (corresponding to a lattice depth of $35\pm

333: 5\,E_r$) we get good agreement with our data. We note, that even for

334: the slowest ramp speeds the measured conversion efficiency never

335: reaches unity but levels off at 95\%, in agreement with the results in

336: Fig.~\ref{fig3}. This, however, does not exclude a true unit

337: conversion efficiency for atom pairs into molecules, because it is

338: possible that 5\% of the

339: \begin{figure}

340:   \includegraphics{figure4}

341:   \caption{(a) Conversion efficiency of atoms (mostly pairs) into

342:     molecules as a function of the ramp speed. (b) A purified sample

343:     of molecules is dissociated into atom pairs at different ramp

344:     speeds.  We measure the number of atoms which are observed in the

345:     first Brillouin zone of the lattice after release, i.e., atoms

346:     that populate the lowest energy band of the lattice. The data is

347:     normalized to the atom number at the lowest ramp speeds. The

348:     continuous lines in (a) and (b) are calculations as described in

349:     the text.}

350:   \label{fig4}

351: \end{figure}

352: atoms are not grouped in pairs, e.g. due to non-adiabaticity in

353: dissociation and tunnelling.  In order to facilitate the comparison of

354: the data distribution and theory, we have scaled the Landau-Zener

355: curve by a factor of 0.95 (dashed line). The maximum controllable ramp

356: speed ($\sim 10^5$ G/s, see Fig.~\ref{fig4}) is limited by the

357: performance of our current supply for the magnetic field coils. The

358: data point at $3\times 10^6$ G/s was obtained by simply switching off

359: the coil currents with an external switch.  The abrupt switching

360: induces eddy-currents which results in a less controlled ramp with a

361: large error margin. For fast switching we measured an conversion

362: efficiency of 5\%$\pm$5\%.

363: 

364: In Fig.~\ref{fig4}b we study the dissociation of a purified sample of

365: molecules. We determine the number of atoms which populate the lowest

366: band of the lattice after dissociation. At low ramp speeds Feshbach

367: molecules get adiabatically converted to pairs of atoms in the lattice

368: ground state. At higher speeds molecules are energetically lifted

369: above the molecule threshold and can decay into higher lattice bands

370: or into the continuum.  Assuming the reversibility of the Landau-Zener

371: transition, we use the same theory curve as in Fig.~\ref{fig4}a.  For

372: higher ramp speeds, we measure larger atom numbers than expected.

373: This is probably due to imperfections of our data analysis which can

374: overestimate the atom number in the lowest band by adding in some atoms

375: from higher bands.

376: 

377: To summarize, we have demonstrated that ultracold Feshbach molecules

378: can be created with high conversion efficiency in a 3D optical

379: lattice. After purification we observe long molecular lifetimes up to

380: 700\,ms. These strong improvements over previous experiments open

381: promising perspectives for applications, e.g. in high resolution

382: molecular spectroscopy and quantum information processing in optical

383: lattices. They may also represent an important step in the creation of

384: a stable BEC of molecules in their vibrational ground state.

385: 

386: We appreciate the assistance of Matthias Theis and Carlo Sias. We

387: thank Thorsten Köhler and Todd Meyrath for fruitful discussions.  This

388: work was supported by the Austrian Science Fund (FWF) within SFB 15

389: (project part 17) and the European Union in the frame of the Cold

390: Molecules TMR Network under contract No.~HPRN-CT-2002-00290.

391: 

392: \input{bibliography}

393: \end{document}