1: \documentclass[aps,prl,twocolumn]{revtex4}
2:
3: %\documentclass[12pt]{article}
4: \usepackage{graphicx}% Include figure files
5: \usepackage{amsmath}
6: \usepackage{amssymb}
7: %\usepackage[pdftex]{graphics}
8: \usepackage{dcolumn}% Align table columns on decimal point
9: %\usepackage{bm}% bold math
10:
11:
12: \begin{document}
13:
14: \newcommand{\kmod}{k_{\mbox{\scriptsize mod}}}
15: \newcommand{\dmsk}{\marginpar{$\clubsuit$}}
16:
17: \title{Spontaneously modulated spin textures in a dipolar
18: spinor Bose-Einstein condensate}
19: \author{M. Vengalattore$^{1}$, S. R. Leslie$^1$, J. Guzman$^1$ and D. M. Stamper-Kurn$^{1,2}$}
20: \affiliation{
21: $^1$Department of Physics, University of California, Berkeley CA 94720 \\
22: $^2$Materials Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA 94720}
23: \date{\today}
24:
25: \begin{abstract}
26: Helical spin textures in a $^{87}$Rb $F=1$ spinor Bose-Einstein
27: condensate are found to decay spontaneously toward a spatially
28: modulated structure of spin domains. This evolution is ascribed to
29: magnetic dipolar interactions that energetically favor the
30: short-wavelength domains over the long-wavelength spin
31: helix. This is confirmed by eliminating the dipolar
32: interactions by a sequence of rf pulses and observing a suppression of the formation of
33: the short-range domains. This study confirms the significance of
34: magnetic dipole interactions in degenerate $^{87}$Rb $F=1$
35: spinor gases.
36: \end{abstract}
37:
38: \maketitle
39:
40: In a wide range of materials, the competition between short- and
41: long-range interactions leads to a rich landscape of spatially
42: modulated phases arising both in equilibrium and as instabilities in
43: non-equilibrated systems \cite{cross,seul}. In classically ordered
44: systems such as magnetic thin films \cite{garel} and ferrofluids
45: \cite{oden02ferrofluid}, short-range ferromagnetic interactions are
46: commonly frustrated by the long-range, anisotropic magnetic dipolar
47: interaction, rendering homogeneously magnetized systems
48: intrinsically unstable to various morphologies of magnetic domains
49: \cite{dabell}. Long-range interactions are also key ingredients in
50: many models of strongly correlated electronic systems \cite{dagotto}
51: and frustrated quantum magnets \cite{scho04qmbook}.
52:
53: In light of their relevance in materials science, strong dipole
54: interactions have been discussed as an important tool for studies of
55: many-body physics using quantum gases of atoms and molecules,
56: offering the means for quantum computation \cite{jaks00},
57: simulations of quantum magnetism \cite{mich06tool} and the
58: realization of supersolid or crystalline quantum phases
59: \cite{goral,buch07polar}. However, in most ultracold atomic gases,
60: the magnetic dipolar interaction is dwarfed by the contact
61: ($s$-wave) interaction. Hence, experimental efforts to attain
62: dipolar quantum gases have focused on non-alkali atoms, notably
63: $^{52}$Cr with its large magnetic moment \cite{laha07ferrofluid},
64: and on polar molecules \cite{doyl04review}.
65:
66: In this Letter, we demonstrate that magnetic dipole interactions
67: play a critical role in the behaviour of a quantum degenerate $F=1$
68: spinor Bose gas of $^{87}$Rb. In this quantum fluid, $s$-wave
69: collisions yield both a spin-independent and a spin-dependent
70: contact interaction \cite{Ho,Ohmi98,sten98spin}, with strengths
71: proportional to $\bar{a} = (2 a_2 + a_0)/3$ and $\Delta a = (a_2 -
72: a_0)/3$, respectively, where the scattering length $a_{F}$ describes
73: collisions between particles of total spin $F$. In $^{87}$Rb, with
74: $a_0 \, (a_2) = 5.39 \, (5.31)$ nm, the spin-dependent contact
75: interaction is far weaker than the spin-independent one;
76: nevertheless, it is a critical determinant of the magnetic
77: properties of degenerate $F=1$ $^{87}$Rb gases
78: \cite{chan05nphys,kron06tune,sadl06symm}.
79: The magnetic dipole interaction strength may be
80: parameterized similarly by a length $a_{d} = \mu_0 g_F^2
81: \mu_B^2 m / (12 \pi \hbar^2)$, where $\mu_0$ is the permeability of
82: vacuum, $g_F = 1/2$ the gyromagnetic ratio, $\mu_B$ the Bohr
83: magneton and $m$ the atomic mass \cite{sant03}. Given $a_{d}/\Delta
84: a = 0.4$, the $F=1$ spinor Bose gas of $^{87}$Rb is an essentially
85: dipolar quantum fluid \cite{pu}.
86:
87: In our experiment, the influence of dipolar interactions on the spinor gas is
88: evidenced by the spontaneous dissolution of deliberately imposed
89: long-wavelength helical spin textures, in favor of a finely modulated pattern of spin domains.
90: We ascribe the emergence of this modulated phase to the
91: magnetic dipole energy that disfavors the homogenously magnetized
92: state and drives the fluid toward short-wavelength spin textures.
93: To test this ascription, we re-examine the behavior of spin
94: helices in condensates in which the dipolar interaction is
95: eliminated using a rapid sequence of rf pulses. The suppression of the
96: modulated phase observed in this case confirms the crucial role
97: of dipolar interactions.
98:
99: For this work, spin-polarized $^{87}$Rb condensates of up to $2.3(1)
100: \times 10^6$ atoms in the $|F=1,m_F=-1\rangle$ hyperfine state and
101: at a kinetic temperature of $T \simeq 50$ nK were confined in a
102: single-beam optical dipole trap characterized by trap frequencies
103: $(\omega_x, \omega_y, \omega_z) = 2 \pi \times (39, 440, 4.2)$
104: s$^{-1}$. The Thomas-Fermi condensate radius in the $\hat{y}$
105: (vertical) direction ($r_y = 1.8\, \mu$m) was less than the spin
106: healing length $\xi_S = (8 \pi \, \Delta a \, n_0)^{-1/2} = 2.4\,
107: \mu$m where $n_0 = 2.3 \times 10^{14}$ cm$^{-3}$ is the peak density
108: of the condensate. This results in a spinor gas that is
109: effectively two-dimensional with regard to spin dynamics.
110:
111: \begin{figure*}[t]
112: \centering
113: \includegraphics[width=0.75\textwidth]{fig1fin}
114: \caption{Spontaneous dissolution of helical textures in a quantum
115: degenerate $^{87}$Rb spinor Bose gas. A transient magnetic field
116: gradient is used to prepare transversely magnetized (b) uniform or
117: (a, c) helical magnetization textures. The transverse magnetization
118: column density after a variable time $T$ of free evolution is shown
119: in the imaged $x-z$ plane, with orientation indicated by hue and
120: amplitude by brightness (color wheel shown). (b) A uniform texture
121: remains homogeneous for long evolution times, while (c) a helical
122: texture with pitch $\lambda = 60 \, \mu\mbox{m}$ dissolves over
123: $\sim$200 ms, evolving into a sharply spatially modulated texture.}
124: \label{fig:dissolutionimage}
125: \end{figure*}
126:
127: The condensate was transversely magnetized by applying a $\pi/2$ rf
128: pulse in the presence of an ambient magnetic field of $B_0 = $
129: 165(5) mG aligned to the $\hat{z}$ axis. Stray magnetic gradients
130: (curvatures) were canceled to less than 0.14 mG/cm (4.3 mG/cm$^2$).
131: A helical spin texture was then prepared by applying a transient
132: magnetic field gradient $dB_z/dz$ for a period $\tau_p =$ 5 -- 8 ms.
133: Larmor precession of the atomic spins in this inhomogeneous field
134: resulted in a spatial spin texture with a local dimensionless spin
135: of ${\mathbf F} = \cos(\kappa z + \omega_L t)\hat{x} + \sin(\kappa z
136: +\omega_L t)\hat{y}$, where $\vec{\kappa} = (g_F \mu_B / \hbar)
137: (dB_z/dz) \tau_p \hat{z}$ is the helix wavevector. The fast
138: time variation at the $\omega_L/2 \pi \simeq 115$ kHz Larmor
139: precession frequency will be henceforth ignored by considering the
140: spin at a particular instant in this rapid evolution. The helix
141: pitch $\lambda = 2 \pi/\kappa$ ranged between 50 and 150 $\mu$m.
142: Given $\lambda \gg \xi_S$, the kinetic energy per atom in this
143: spin texture, $E_\kappa = \hbar^2 \kappa^2 / 4 m$, was always
144: negligible compared to the ferromagnetic contact interaction energy
145: \cite{spinenergyfootnote}.
146:
147: The helical spin texture was then allowed to evolve in a homogenous
148: magnetic field for a variable time before the vector magnetization
149: was measured using a sequence of non-destructive phase contrast
150: images. Because of Larmor precession, a rapid sequence of images
151: taken with circularly polarized light propagating along the
152: $\hat{y}$ direction can be analyzed to determine the
153: column-integrated magnetization perpendicular to the ambient field
154: \cite{jmhshort,veng07mag}, with vector components $\tilde{M}_{x,y} =
155: (g_F \mu_B) \tilde{n} F_{x,y}$ where $\tilde{n}$ the column number
156: density. Subsequent to this imaging sequence, a $\pi/2$ pulse
157: was applied to rotate the longitudinal spin $F_z$ into the
158: transverse spin plane, and a second sequence of images was obtained.
159: A least-squares algorithm comparing data from the two imaging
160: sequences allowed the longitudinal magnetization $\tilde{M}_z$ to be
161: determined \cite{detailsfootnote}.
162:
163: The evolution of helical spin textures is portrayed in Fig.\
164: \ref{fig:dissolutionimage}. While uniform spin textures ($\lambda
165: \gg 2 r_z$) remained homogenous for long times, helical textures
166: ($\lambda < 2 r_z$) spontaneously develop short-wavelength
167: modulations of the magnetization. This modulated phase is
168: characterized by spin domains with typical dimensions of $\simeq
169: 10\, \mu$m, much smaller than the pitch of the imprinted helix, with
170: the magnetization varying sharply between adjacent domains. The
171: spatial modulations nucleated in regions that varied from shot to shot
172: but gradually grew to encompass the entire condensate.
173:
174:
175: To quantify this behaviour, we considered the power spectrum of the
176: spatial Fourier transform of the vector magnetization,
177: $|{\bf{\tilde{M}}}(k_x, k_z)|^2$, where $(k_x, k_z)$ is the spatial
178: wavevector in the image plane. This spectrum was found to
179: consist of two distinct components: a central component that
180: quantifies the long-range order of the helical texture, and a second
181: concentration of spectral power at a discrete set of wavevectors of
182: magnitude $\kmod \simeq 2 \pi / (10 \, \mu\mbox{m})$ representing the short-range
183: order of the final modulated texture. After subtracting out
184: the background representing image noise, we divided spatial Fourier
185: space into regions indicated in Fig.\ \ref{fig:fig2} and defined the
186: integrated spectral power in the central region (annular region) as
187: the parametrization of long-range (short-range) spatial order in the
188: quantum fluid.
189:
190: \begin{figure}[tb]
191: \centering
192: \includegraphics[width=0.37\textwidth]{fig2fin}
193: \caption{Power spectrum of the spatial Fourier transform and the
194: two-point correlation function $G(x,z)$ for the initial spin helix
195: (a, c) and the spontaneously modulated phase (b, d). These data are
196: derived from the same image sequence shown in Fig.\ 1 (d). The
197: images (a, c) correspond to an evolution time $T = 0$ ms while (b,
198: d) correspond to an evolution time $T = 250$ ms. The short-range
199: spatial order is defined as the integrated spectral power in the
200: annular region shown in (b).} \label{fig:fig2}
201: \end{figure}
202:
203: The formation of the spontaneously modulated texture is reflected in
204: the reduction of the long-range order parameter and the concomitant
205: rise of the short-range order parameter (Fig.\ \ref{fig:fig3}).
206: During this process, the total spectral power was found to be
207: roughly constant indicating that the bulk of the quantum fluid
208: remains fully magnetized even as the long-range order is reduced.
209: The growth rate $\gamma$ of the short-range order
210: parameter determined from such data was found to rise monotonically
211: with the wavevector $\kappa$ of the initial helical texture. While
212: the long-range order was found to decrease after sufficiently long
213: evolution times even in condensates prepared with nearly uniform
214: magnetization, we note that stray magnetic field inhomogeneities of
215: $\sim$5 $\mu$G across the axial length of the condensate would by
216: themselves produce a helical winding across the condensate over a
217: period of 300 ms, constraining our ability to test
218: the stability of homogenous spin textures.
219:
220:
221: Another measure of the spontaneous short-range modulation in the
222: condensate is the appearance of polar-core spin vortices throughout
223: the gas. Such vortices were identified as in Ref.\
224: \cite{sadl06symm} by a net winding of the transverse magnetization
225: along a closed two-dimensional path of non-zero magnetization in the
226: imaged gas. The number of identified spin vortices was roughly
227: proportional to the short-range order parameter, with no vortices
228: identified in the initially prepared spin helix and up to 6
229: vortices/image identified in the strongly modulated texture produced
230: after free evolution. In each instance, the number of vortices
231: with positive and negative charge was found to be approximately
232: equal.
233:
234: \begin{figure}[t]
235: \centering
236: \includegraphics[width=0.34\textwidth]{fig3fin}
237: \caption{Growth of the spontaneously modulated phase ($\bullet$)
238: coincides with a reduction in the integrated energy in the low
239: spatial frequency region ($\blacksquare$). The data shown correspond
240: to an initial helical pitch of 60 $\mu$m. Inset: The initial growth rate
241: $\gamma$ of the modulated phase as a function of the helix
242: wavevector. These were extracted from linear fits of the
243: short-range order parameter at short evolution times.} \label{fig:fig3}
244: \end{figure}
245:
246: A striking feature in the evolution of spin textures is the
247: significant rise in the kinetic energy of the condensed atoms,
248: reaching a value of $\hbar^2 \kmod^2/ 8 m = h \times 6$ Hz per atom
249: given that roughly half the spectral weight of the texture's
250: magnetization is at the wavevector $\kmod$. One expects the total
251: energy per atom in the condensate to be constant during this
252: evolution, or even to diminish through the transfer of energy to the
253: non-condensed portion of the gas. Yet, in examining the energy of
254: the initially prepared spin helix, we find the local
255: contact-interaction energy is minimized, the quadratic Zeeman energy
256: is just $q/2 = h \times 1$ Hz at the ambient magnetic field, and the
257: kinetic energy of the spin helix is just $E_\kappa/2 < h \times 0.5$
258: Hz for a helix pitch of $\lambda > 50 \, \mu$m.
259:
260: This apparent energetic deficit of the spin helix state can be
261: accounted for by the magnetic dipole interaction. The on-axis
262: magnetic field produced by a spin helix in an infinite axial column
263: of gas with a gaussian transverse density profile can be simply
264: calculated. From this calculation, we estimate that a gas
265: with uniform transverse magnetization possesses an excess of $E_d =
266: \mu_0 g_F^2 \mu_B^2 n_0/2 \, \sim h \times 5$ Hz compared to the
267: energy of a tightly wound helix, a figure that closely matches the
268: excess kinetic energy of the finely modulated texture.
269:
270: To confirm the role of magnetic dipolar interactions in the
271: evolution of these spin textures, we employed a modification of the
272: NMR technique of spin-flip narrowing \cite{slic78magres} to
273: eliminate effectively the dipolar interactions. The interaction energy
274: of two magnetic dipoles separated by the displacement vector ${\bf r}$
275: is proportional to ${\bf F}_1 \cdot {\bf F}_2 - 3 (\hat{r} \cdot {\bf{F}}_1)
276: (\hat{r}\cdot {\bf{F}}_2)$. If both dipoles experience rapid, common
277: rotations that evenly sample the entire $SO(3)$ group of rotations, the
278: interaction energy will average to zero regardless of the relative orientations of the
279: spin vectors ${\bf F}_{1,2}$ and of the displacement vector ${\bf r}$. We note
280: that such spin rotations also annul the quadratic Zeeman shift. However, since
281: the character of the spontaneously modulated phase was observed to be unchanged
282: as $q$ was varied over a factor of 5 ($0.4 < \frac{1}{h}\, q/2 < 2$ Hz), we ignore this small
283: difference.
284:
285: Experimentally, after the initial spin texture had been prepared as
286: before, we effected such spin rotations by applying a rapid sequence
287: of $\pi/2$ rf pulses to the Larmor precessing atoms at random
288: intervals, and, thus, along random rotation axes, at a mean rate of
289: 1 -- 2 kHz. The spin helix was allowed to evolve under the constant
290: action of these dipole-cancelation pulses until the pulses ceased
291: and the sample was imaged as described earlier. As shown in Fig.\
292: \ref{fig:withandwithout}, in the absence of dipole interactions a
293: spin helix was still prone to spontaneous spatial modulation,
294: consistent with the predictions of dynamic instabilities made in
295: Refs.\ \cite{lama07helix} and \cite{cher07helix}. However, the depth
296: of this modulation was significantly suppressed, as seen from the
297: smaller spectral weight of short-range magnetic order. This suppression was
298: less evident for large helix wavevectors ($\kappa > 110$ mrad$/\mu$m), presumably due to the faster growth
299: of the spontaneously modulated phase. The reduction of the excess kinetic
300: energy of the final spin texture in the spin-rotation-averaged sample supports
301: our identification of the dipole energy as its source.
302:
303: Finally, we note the distinct six-fold structure in the spatial
304: Fourier spectrum of the spontaneously modulated phase (Fig.\
305: \ref{fig:fig2}(b)). The interpretation of this structure is aided by
306: considering the spatial correlation function of the magnetization,
307: which we define as
308: \begin{equation} G(\delta {\bf r}) =
309: \frac{\sum_{\bf r} \tilde{\bf{M}}({\bf r} + \delta {\bf{r}} ) \cdot
310: \tilde{\bf{M}}({\bf r})}{(g_F \mu_B)^2 \sum_{\bf r} \tilde{n} ({\bf
311: r} + \delta {\bf{r}} ) \tilde{n} ({\bf r}) }
312: \end{equation}
313: For the initial spin helix, this correlation function shows the
314: long-range sinusoidal variation of the transverse magnetization at
315: the corresponding pitch. In contrast, the final texture shows
316: significant short-range correlations in both the $\hat{x}$ and
317: $\hat{z}$ directions, with regions of opposite alignment arranged in
318: the form of a checkerboard with a lattice spacing $l_m \simeq 10 \,
319: \mu$m (Fig.\ 2(d)). While these correlations are strongest at short range, they
320: persist, albeit with diminished strength, even for separations
321: $\delta r \gg l_m$. This lattice structure is suppressed in
322: helices evolving under the active cancelation of dipolar
323: interactions.
324:
325: \begin{figure}[tb]
326: \centering
327: \includegraphics[width=0.42\textwidth]{fig4fin}
328: \caption{Spatial power spectra (a) without or (b) with the
329: application of rapid rf pulses during the 200 ms evolution following
330: the preparation of the helical spin texture, are averaged over four
331: experimental repetitions. Eliminating the dipolar interactions
332: suppresses the short-range spatial modulation of the final spin
333: texture. (c) The average number of detected spin vortices and (d)
334: the short-range order parameter are shown vs.\ evolution time.
335: Square symbols indicate data obtained when the dipolar interaction
336: is spatially averaged. In all cases, the initial helix pitch was
337: $\lambda = 80 \, \mu \mbox{m}$.} \label{fig:withandwithout}
338: \end{figure}
339:
340: In conclusion, we observe a magnetic-dipole-mediated instability
341: resulting in the emergence of spontaneously ordered spin domains in
342: a quantum degenerate spinor Bose gas. The demonstration of the
343: significance of magnetic dipolar interactions in a $^{87}$Rb spinor
344: gas presents a new arena for the study of dipolar quantum gases.
345: Of particular interest is the influence of such anisotropic
346: interactions on the true ground state of these spinor gases and the
347: question of whether the self-organized modulated spin texture
348: represents an unforeseen equilibrium phase of this dipolar quantum
349: fluid.
350:
351:
352: We acknowledge insightful discussions with E.\ Demler, C.\ H.\ Greene and
353: A.\ Lamacraft. This work was supported by the NSF, the David and Lucile
354: Packard Foundation, and DARPA's OLE program. Partial personnel and
355: equipment support was provided by the Division of Materials Sciences
356: and Engineering, Office of Basic Energy Sciences. S.\ R.\ L.\
357: acknowledges support from the NSERC.
358:
359: %\bibliographystyle{apsrev}
360: %\bibliography{Helix,allrefs_x2,helixfootnotes}
361: \begin{references}
362: \bibitem{cross} M. C. Cross and P. C. Hohenberg, Rev. Mod. Phys. {\bf 65}, 851 (1993).
363: \bibitem{seul} M. Seul and D. Andelman, Science {\bf 267}, 476 (1995).
364: \bibitem{garel} T. Garel and S. Doniach, Phys. Rev. B {\bf 26}, 325 (1982).
365: \bibitem{oden02ferrofluid} S. Odenbach, ed., {\em Ferrofluids: magnetically controllable
366: fluids and their applications}, vol. 594 of {\em Lecture Notes in Physics} (Springer, New York, 2002).
367: \bibitem{dabell} K. De'Bell, A. B. MacIsaac and J. P. Whitehead, Rev. Mod. Phys. {\bf 72}, 225 (2000).
368: \bibitem{dagotto} E. Dagotto, Science {\bf 309}, 257 (2005).
369: \bibitem{scho04qmbook} U. Schollw\"{o}ck et al., eds., {\em Quantum Magnetism}, vol. 645 of
370: {\em Lecture Notes in Physics} (Springer, Berlin, 2004).
371: \bibitem{jaks00} D. Jaksch et al., Phys. Rev. Lett. {\bf 85}, 2208 (2000).
372: \bibitem{mich06tool} A. Micheli, G. K. Brennan and P. Zoller, Nature Physics {\bf 2}, 341 (2006).
373: \bibitem{goral} K. G\'{o}ral, L. Santos and M. Lewenstein, Phys. Rev. Lett. {\bf 88}, 170406 (2002).
374: \bibitem{buch07polar} H. P. B\"{u}chler et al., Phys. Rev. Lett. {\bf 98}, 060404 (2007).
375: \bibitem{laha07ferrofluid} T. Lahaye et al., Nature {\bf 448}, 672 (2007).
376: \bibitem{doyl04review} J. Doyle et al., Eur. Phys. J D {\bf 31}, 149 (2004).
377: \bibitem{Ho} T. L. Ho, Phys. Rev. Lett. {\bf 81}, 742 (1998).
378: \bibitem{Ohmi98} T. Ohmi and K. Machida, J. Phys. Soc. Jpn. {\bf 67}, 1822 (1998).
379: \bibitem{sten98spin} J. Stenger et al., Nature {\bf 396}, 345 (1998).
380: \bibitem{chan05nphys} M. -S. Chang et al., Nature Physics {\bf 1}, 111 (2005).
381: \bibitem{kron06tune} J. Kronjager et al., Phys. Rev. Lett. {\bf 97}, 110404 (2006).
382: \bibitem{sadl06symm} L. Sadler et al., Nature {\bf 443}, 312 (2006).
383: \bibitem{sant03} L. Santos, G. Shlyapnikov and M. Lewenstein, Phys. Rev. Lett. {\bf 90}, 250403 (2003).
384: \bibitem{pu} S. Yi and H. Pu, Phys. Rev. Lett. {\bf 97}, 020401 (2006).
385: \bibitem{spinenergyfootnote} In a transversely magnetized spin texture in an $F=1$ spinor
386: condensate, only half of the atoms, those in the $m_F = \pm 1$ magnetic
387: sublevels, acquire kinetic energy.
388: \bibitem{veng07mag} M. Vengalattore et al., Phys. Rev. Lett. {\bf 98}, 200801 (2007).
389: \bibitem{jmhshort} J. M. Higbie et al., Phys. Rev. Lett. {\bf 95}, 050401 (2005).
390: \bibitem{detailsfootnote} Further details on the imaging system and signal analysis will
391: be discussed elsewhere.
392: \bibitem{slic78magres} C. Slichter, {\em Principles of Magnetic Resonance} (Springer, New York, 1978).
393: \bibitem{lama07helix} A. Lamacraft, preprint, arXiv:0710.1848.
394: \bibitem{cher07helix} R. W. Cherng et al., preprint, arXiv:0710.2499.
395: \end{references}
396:
397:
398: \end{document}
399: