1: \documentclass[12pt]{iopart}
2: %\documentclass{iopart}
3:
4: \usepackage{iopams}
5: \usepackage{graphicx}
6: \usepackage{pstricks}
7:
8: \newcommand{\mod}{{\rm{mod}}}
9: \newcommand{\ket}[1]{|#1\rangle}
10: \newcommand{\ketf}[1]{\left|#1\right\rangle}
11: \newcommand{\bra}[1]{\langle#1|}
12: \newcommand{\braf}[1]{\left\langle#1\right|}
13: \newcommand{\of}[1]{(#1)}
14: \newcommand{\off}[1]{\left(#1\right)}
15: \renewcommand{\vec}[1]{\mathbf{#1}}
16:
17: \begin{document}
18:
19: \title[Quasi-angular momentum in rotating optical lattices]{Quasi-angular momentum of Bose and
20: Fermi gases in rotating optical lattices}
21: \author{Brandon M. Peden, Rajiv Bhat, Meret Kr\"{a}mer, and Murray J. Holland}
22: \address{JILA, NIST and Department of Physics, University of Colorado, Boulder CO
23: 80309-0440, USA}
24: \ead{pedenb@colorado.edu}
25:
26: \begin{abstract}
27: The notion of quasi-angular momentum is introduced to
28: label the eigenstates of a Hamiltonian with a discrete rotational
29: symmetry. This concept is recast in an operatorial form where the
30: creation and annihilation operators of a Hubbard Hamiltonian carry
31: units of quasi-angular momentum. Using this formalism, the ground states
32: of ultracold gases of non-interacting fermions in rotating optical lattices
33: are studied as a function of rotation, and transitions between states of different
34: quasi-angular momentum are identified. In addition, previous results for
35: strongly-interacting bosons are re-examined and compared to the results for
36: non-interacting fermions. Quasi-angular momentum can be used to distinguish
37: between these two cases. Finally, an experimentally
38: accessible signature of quasi-angular momentum is identified
39: in the momentum distributions of single-particle eigenstates.
40: \end{abstract}
41:
42: \maketitle
43:
44:
45: \section{Introduction}
46:
47: Recent experimental and theoretical studies of quantum gases of ultracold
48: atoms have been successful in replicating the behavior of a wide class of
49: condensed matter systems. In particular, Bose-Einstein condensation (BEC)
50: \cite{anderson1995obe,davis1995bec,bradley1995ebe}
51: and the degenerate Fermi gas \cite{demarco1999ofd} have both been observed, and there have been
52: extensive studies of superfluidity, vortex formation \cite{zwierlein2005vas},
53: and the BCS-BEC crossover \cite{greiner2003emb,milstein2002rtc}.
54: The Mott-insulator to superfluid phase transition in the Bose-Hubbard model has been
55: observed in a system of ultracold bosons in an optical lattice of \cite{fisher1989sit,greiner2002qpt}.
56: In addition, experiments observing novel phases of ultracold matter in atomic mixtures of
57: Bose and Fermi gases in optical lattices have been performed
58: \cite{albus2003mbf,illuminati2004hta,ospelkaus2006lba,ospelkaus2006uhm}.
59:
60: Along these lines, focus has turned to strongly-correlated effects in systems found outside
61: the realm of condensed matter.
62: One promising line of research in the realization of these systems
63: is that of coupled arrays of optical cavities, which allow single-site addressing
64: \cite{angelakis2007pbi,hartmann2006sip,greentree2006qpt}.
65: Another is a mapping between the Hamiltonian of electrons confined in two
66: dimensions in the presence of a constant transverse magnetic field and that of a
67: rotating atomic gas. This has led to various theoretical studies predicting fractional quantum Hall
68: effect (FQHE) behavior in rotating BEC's
69: \cite{wilkin2000ccb,paredes2001asa,fischer2004vlv,baranov2005fqh}. However, experimentally
70: reaching the parameter regimes necessary to observe this behavior is difficult
71: \cite{schweikhard2004rrb}. Optical lattices offer a solution to this problem by enhancing
72: correlations. Optical lattices are formed as standing waves of
73: counter-propagating laser beams and act as a lattice for ultracold atoms. These
74: systems are highly tunable: lattice spacing and depth can be varied by tuning the
75: frequency and intensity of the lasers, and interactions between atoms can be tuned
76: via a Feshbach resonance \cite{bloch2005uqg}.
77: Some recent
78: studies have made direct connections between FQHE physics and strongly-interacting
79: bosons in optical lattices in the presence of an effective magnetic field
80: \cite{jaksch2003cem,palmer2006hff,sorensen2005fqh,bhat2007he,hafezi2007fqh}.
81:
82: Connecting these theoretical studies to experiment requires identifying observables that
83: can act as experimental signatures for the predicted physics. For instance, linear response
84: theory has been applied to current flow in the presence of a ``potential gradient'' in
85: order to observe FQHE physics \cite{bhat2007he}, noise correlation analysis has been
86: applied to bosons
87: in a rotating ring lattice \cite{rey2006llt}, and Bragg spectroscopy has been mentioned as a
88: probe for vortex states \cite{vignolo2007qvo}. A common observable in ultracold atomic
89: gases is the momentum distribution, measured in time-of-flight experiments.
90:
91: In this paper, we present analytical results for strongly-interacting bosons in rotating
92: optical lattices that confirm numerical results in a previous paper \cite{bhat2006bec}
93: and extend them to systems of non-interacting fermions. These results are based
94: on the notion of quasi-angular momentum, which
95: is a quantum number for systems with a discrete rotational symmetry. Quasi-angular momentum
96: is analogous to quasi-momentum for periodic translationally invariant systems
97: \cite{bozovic1984pbs} and has been previously used in the context of rotating ring lattices
98: \cite{buonsante2005aub,ueda1999gsp,buonsante2005aas} and carbon nanotubes
99: \cite{reich2004cnb} to label eigenstates. A formalism for this quantum number for
100: second-quantized systems is presented here and applied to gases of bosons and fermions in
101: rotating optical lattices. The results presented identify transitions between
102: states of different symmetry for the ground state of a rotating system.
103: A possible avenue for experimentally detecting these results via the momentum
104: distribution of the ground state is also presented.
105:
106: The quasi-angular momentum, $m$, of the ground state of an ultracold
107: quantum gas of strongly-interacting bosons or non-interacting fermions in a
108: rotating ring lattice of $N$ sites can be monitored as a function of
109: rotation speed. For the case of strongly-interacting bosons, $m$ cycles through the values
110: $m=nl~\mod~N:l=0,1,2,\dots$, where $n$ is the number of particles. For the
111: case of fermions, $m$ cycles through the values $m=nl~\mod~%
112: N:l=0,1,2,\dots$ for odd numbers of particles and $m=n(l+1/2)
113: ~\mod~N:l=0,1,2,\dots$ for even numbers of particles. A system of non-interacting
114: fermions is thereby distinguishable from a system of hard-core bosons.
115: Similar behavior obains for systems of fermions in two-dimensional square lattices. Signatures
116: of this quantum number in experiment are observable via the momentum
117: distribution of the ground state. For single-particle systems in two-dimensional square
118: lattice geometries, such signatures include the existence of a peak at zero momentum only
119: for the $m=0$ state and peak-spacing differences between $m=2$ and $m=1,3$ state.
120:
121: The paper is structured as follows. In section 2, we draw the analogy
122: between systems with a discrete translational symmetry and systems with a
123: discrete rotational symmetry, thereby generating a Bloch theory for the latter
124: and introducing the notion of quasi-angular momentum. Specific
125: investigations of quantum gases of strongly-interacting bosons and non-interacting
126: fermions in ring and square lattice geometries are carried out in section 3. In section 4,
127: signatures of the quasi-angular momentum in the momentum distribution of a state
128: are identified. Section 5 summarizes the main results of the paper.
129:
130: \section{Quasi-angular momentum}
131:
132: In this section, we briefly review band theory and introduce the language of
133: Bloch functions and quasi-momentum (see for instance \cite{marder2000cmp,kittel1996ssp}).
134: By analogy, the \textit{quasi-angular momentum} of an eigenstate of a system with a
135: discrete rotational symmetry is defined.
136:
137:
138: \subsection{Bloch theory}
139:
140: When a small periodic potential of lattice
141: period $d$ is introduced to a free-particle system, the two wavefunctions
142: $\psi _{1}\left( x\right)=e^{i\pi x/d}$ and $\psi _{2}\left( x\right) =e^{-i\pi x/d}$
143: at the Brillouin zone boundaries ($q=\pm\pi/d$) mix with each other. The
144: potential acts as a perturbation that breaks the degeneracy between these
145: two states, creating new eigenfunctions,
146: \begin{equation}
147: \psi _{\pm }\left( x\right) =\frac{1}{\sqrt{2}}\left( e^{i\pi x/d}\pm
148: e^{-i\pi x/d}\right).
149: \end{equation}
150: This process opens up a band-gap at the Brillouin zone boundaries.
151: In a reduced zone scheme, this energy spectrum becomes
152: the familiar one-dimensional Bloch band diagram (see figure 1).
153: \begin{figure}[tbp]
154: \centerline{
155: \includegraphics[width=2.25in]{figure1.pdf}
156: }
157: \caption{Dispersion relation for free particles (solid) and particles moving in
158: a one-dimensional sinusoidal lattice (dashed) plotted in a reduced zone scheme.
159: Free space dispersion relations are plotted versus momentum, and lattice
160: dispersion relations are plotted versus quasi-momentum. Energy is in units of the
161: recoil energy, $E_R=\hbar^2\pi^2/2md^2$, where $d$ is the lattice spacing. The
162: lattice depth is $V_0=5E_R$.}
163: \end{figure}
164:
165: The introduction of a periodic potential breaks the continuous translational
166: symmetry of the system, and therefore the wavefunctions are no longer momentum
167: eigenstates. Instead, the eigenfunctions of
168: the Hamiltonian are eigenfunctions of a discrete translation
169: operator, $T_{d}$, that translates the entire system by one lattice spacing,
170: \begin{equation}
171: \psi \left( x-d\right) =T_{d}\psi \left( x\right) =e^{-ikd}\psi \left(
172: x\right),
173: \end{equation}%
174: and can be labeled by $k$, the \textit{quasi-momentum}. A quasi-momentum
175: eigenstate of a periodic Hamiltonian can be written as the product of a
176: plane wave and a function periodic in the lattice:
177: \begin{equation}
178: \psi^{(l)} _{k}\left( x\right) =e^{ikx}u^{(l)}_{k}\left( x\right),~~~~u^{(l)}_{k}\left(
179: x-d\right) =u^{(l)}_{k}\left( x\right),
180: \end{equation}
181: where $l$ is a band index.
182:
183: When these Bloch functions are expanded in the
184: momentum basis, the only momenta contributing to the sum are those that
185: differ from $k$ by a reciprocal lattice vector, $G_j=\frac{2\pi j}{d}$. The expansion is
186: then given by
187: \begin{equation}
188: \psi _{k}^{(l)}\left( x\right) =\int_{-\infty }^{\infty }\frac{dq}{\sqrt{2\pi }}%
189: e^{iqx}\psi _{k}^{(l)}\left( q\right) = \sum_{j=-\infty }^{\infty }c_j^{(l)} \frac{%
190: e^{i\left( k+2\pi j/d\right) x}}{\sqrt{2\pi }},
191: \end{equation}%
192: where the expansion coefficients $c_j^{(l)}$ are given by%
193: \begin{equation}
194: c_j^{(l)}=\int_{-\infty }^{\infty }dx\frac{e^{-i\left( k+2\pi j/d\right) x}}{%
195: \sqrt{2\pi }}\psi _{k}^{(l)}\left( x\right).
196: \end{equation}
197: Figure 2(a) is an example of a Bloch function in a sinusoidal lattice, and figure 2(b) is
198: the Fourier transform of this Bloch function -- that is, the Bloch function in momentum space.
199: \begin{figure}
200: \centerline{
201: \includegraphics[width=4.5in]{figure2a_2b.pdf}
202: }
203: \caption{(Color online.) The zero quasi-momentum Bloch function in the first band for an
204: infinite sinusoidal lattice if lattice depth $V_0=5E_R$. (a) The real space wave function is
205: concentrated in the lattice sites. (b) The wave function in momentum space
206: exhibits peaks at reciprocal lattice vectors. The envelope of these peaks is
207: given by a Gaussian for large lattice depths, consistent with a lattice of
208: simple harmonic oscillator wells.}
209: \end{figure}
210:
211: In the rest of the paper, we will consider only lattices that are deep enough that
212: a single-band approximation will be adequate to describe the physics. In the proceeding
213: sections, then, we drop the band index, $l$.
214:
215:
216: \subsection{Lowering the symmetry}
217:
218: If an extra potential, $V\left(x\right)$, satisfying
219: \begin{equation}
220: V\left( x-pd\right) =V\left( x\right),
221: \end{equation}
222: with $p\geq2$ an integer and $d$ the lattice spacing, is introduced to a lattice, then
223: Bloch states $\left\langle x|k\right\rangle=\psi _{k}\left(x\right)$ and
224: $\left\langle x|k^{\prime}\right\rangle=\psi _{k^{\prime}}\left(x\right)$ will mix with
225: each other if $k$ and $k^{\prime }$ differ by $2\pi m/pd$, as shown by the following
226: calculation:
227: \begin{equation}
228: \fl \left\langle k^{\prime }\left\vert V\right\vert k\right\rangle
229: =\int_{-\infty }^{\infty }dx\psi _{k^{\prime }}^{\ast}\left( x\right)
230: V\left( x\right) \psi _{k}\left( x\right) =e^{i\left(k^{\prime }-k\right)
231: pd}\left\langle k^{\prime }\left\vert V\right\vert k\right\rangle.
232: \label{mixeqn}
233: \end{equation}
234: Eigenstates of the new Hamiltonian will therefore be linear combinations of these Bloch
235: states. The potential increases the lattice periodicity and, accordingly,
236: decreases the size of the first Brillouin zone to $2\pi /pd$.
237:
238: In figure 3(a), the lowest band of a one-dimensional lattice with an arbitrary
239: periodic potential is plotted in a reduced zone scheme. Introducing an infinitesimal
240: potential whose periodicity is three times the lattice spacing $d$, the
241: first Brillouin zone is reduced to the region $-\pi /3d\leq k\leq \pi /3d$.
242: Equation (\ref{mixeqn}) shows that the states lying on a vertical
243: line in figure 3(b) mix with each other under the addition of the new
244: potential -- i.e., $\left\langle k^{\prime }\left\vert V\right\vert k\right\rangle$
245: is non-zero for these states. In particular, the degenerate
246: states at $k=0$ mix, breaking the degeneracy and opening up a gap at $k=0$,
247: as shown in figure 3(b). The single band splits into three bands due to
248: the reduction of the symmetry by $V\left(x\right) $.
249:
250: When the addition of a potential reduces the symmetry of a system, as
251: $V\left(x\right) $ did to the generic system described in figure 3, the
252: original eigenstates of the system mix according to their symmetry.
253: Accordingly, gaps open up where there are degeneracies between states
254: of the same symmetry, creating a band structure.
255: \begin{figure}
256: \centerline{
257: \includegraphics[width=4.5in]{figure3a_3b.pdf}
258: }
259: \caption{(Color online.) (a) Lowest band dispersion relation for a one-dimensional
260: sinusoidal lattice plotted in a reduced zone scheme that anticipates
261: the lowering of the symmetry via an additional potential with periodicity $3d$. (b) After the
262: introduction of this potential, states with the same quasi-momentum, $k$ in the
263: reduced zone mix with each other, opening up gaps anywhere that there is a
264: degeneracy. The matrix elements of the added potential between states of the same
265: quasi-momentum are of magnitude $5E_R$.}
266: \end{figure}
267:
268: \subsection{The ground state of a moving lattice}
269:
270: The utility of these symmetry concepts and quasi-momentum can be demonstrated
271: by considering the effect of moving the lattice. In a stationary lattice, a zero
272: quasi-momentum Bloch function is always the ground state. When
273: the lattice is moving, the energetically favored state is one that moves
274: along with the lattice. In order to investigate this phenomenon, we move to
275: a frame co-moving with the lattice. The Hamiltonian in this frame is given by
276: \cite{landau1976m}
277: \begin{equation}
278: H=H_0-\vec{p}\cdot \vec{v},
279: \end{equation}%
280: where $H_0$ is the Hamiltonian in the non-moving frame, and $\vec{v}$ is
281: the velocity of the moving frame with respect to the lab frame. The ground state
282: of $H_0-\vec{p}\cdot \vec{v}$ will be the ground state of the lab frame but
283: written in rotating frame coordinates.
284:
285: Consider an eigenstate of $H_0$, $\psi _{k}\left(
286: x\right) $. In
287: first-order perturbation theory, the energy of this state is%
288: \begin{equation}
289: E_{k}=\left\langle k\left\vert H_{0}-\vec{p}\cdot \vec{v}\right\vert
290: k\right\rangle =E^0_{k}-v\left\langle k\left\vert p\right\vert
291: k\right\rangle.
292: \end{equation}%
293: At zero velocity, the ground state is a zero quasi-momentum state,
294: $\psi_0(x)$. However, the energy of a state that has a positive
295: average momentum, $\left\langle k\left\vert p\right\vert k\right\rangle
296: > 0$, falls below that of $\psi _{0}\left( x\right)$ for large enough $v$.
297: This change is signalled by an exact energy level crossing
298: in the energy spectrum between two quasi-momentum states (see figure 4). As
299: the velocity is increased, the effect of the velocity term is to reorder the
300: energies of the quasi-momentum states, thereby leading to exact energy level
301: crossings.
302: \begin{figure}
303: \centerline{
304: \includegraphics[width=2.5in]{figure4.pdf}
305: }
306: \caption{(Color online.) The four lowest energies in the single-particle energy spectrum as
307: a function of lattice velocity for a four-site sinusoidal lattice with
308: periodic boundary conditions. At zero velocity, the zero quasi-momentum
309: state (solid) is the ground state. At some finite positive velocity, the
310: $k=\pi /2d$ state (dashed) is energetically favorable and becomes the ground state.
311: The ground state eventually cycles through all quasi-momenta.}
312: \end{figure}
313:
314:
315: \subsection{Quasi-angular momentum}
316:
317: A moving $N$-site linear lattice with periodic boundary
318: conditions is analogous to a rotating $N$-site ring lattice. We can make this analogy
319: explicit by considering the Hamiltonian of a moving, one-dimensional,
320: sinusoidal $N$-site lattice in the co-moving frame, given by
321: \begin{equation}\label{H1D}
322: H=-\frac{\hbar ^{2}}{2M}\frac{\partial ^{2}}{\partial x^{2}}+V_{0}\cos
323: ^{2}\left( qx\right)-v\frac{\hbar }{i}\frac{\partial }{\partial x},
324: \end{equation}
325: where $v$ is the velocity of the lattice.
326: The wave-number $q$ can be rewritten as $q=\pi/d$, where $d$ is the lattice
327: spacing. If periodic boundary conditions,
328: \begin{equation}
329: \psi \left( x+Nd\right) =\psi \left( x\right),
330: \end{equation}%
331: are included, then the Hamiltonian describes the system illustrated in figure 5(a)
332: for the explicit case of an 8-site lattice. On the other hand, the Hamiltonian of a
333: rotating, sinusoidal $N$-site ring lattice (figure 5(b)) in the rotating frame is given by
334: \begin{equation}\label{HRing}
335: H=-\frac{\hbar ^{2}}{2M}\frac{1}{R^{2}}\frac{\partial ^{2}}{\partial \phi
336: ^{2}}+V_{0}\cos ^{2}\left( N\frac{\phi }{2}\right)-\Omega\frac{\hbar }{%
337: i}\frac{\partial }{\partial \phi },
338: \end{equation}
339: where $\Omega$ is the rotation frequency, $R$ is the radius of the ring, and
340: $\frac{\hbar}{i}\frac{\partial}{\partial\phi}$ is the angular mometum operator, $L_z$.
341: The inclusion of the term $-\Omega L_{z}$ has the effect of moving to a frame
342: co-rotating with the lattice \cite{landau1976m}. The two Hamiltonians, equations
343: \ref{H1D} and \ref{HRing}, are identical if we perform a transformation
344: $x=\phi Nd/2\pi$ and identify $Nd/2\pi$ with $R$ and $v/R$ with $\Omega$.
345:
346: Since the Hamiltonians are exactly identical, all of the properties of one-dimensional
347: systems with a discrete translational invariance carry over for ring systems with a discrete
348: rotational invariance. The analogy can be carried further, starting with the two-dimensional
349: free-space solution in polar coordinates,
350: \begin{equation}
351: \psi _{j}\left( \phi ,\rho \right)=\e^{ij\phi }R_{j}\left( \rho \right),
352: \end{equation}%
353: where $R_{j}\left(\rho\right)$ is a radial function, irrelevant for our discussion, and
354: $j$ is an integer. In the presence of a potential
355: that breaks the rotational symmetry, the eigenstates are linear combinations
356: of these free-space solutions. If this potential has a discrete $N$-fold
357: rotational symmetry, its eigenstates can be expanded in the free space
358: solutions as
359: \begin{equation}
360: \psi _{m}\left( \phi ,\rho \right) =\sum_{j=-\infty }^{\infty }
361: a_je^{i\left( Nj+m\right) \phi }R_{j}\left( \rho \right).
362: \end{equation}
363:
364: It is evident from this expansion that $\psi _{m}\left( \phi
365: ,\rho \right) $ is an eigenvector with eigenvalue $e^{-i2\pi m/N}$ of the discrete rotation operator
366: $R_{2\pi /N}$ that rotates the system by the angle $2\pi/N$;
367: $R_{2\pi /N}$ takes the place of the discrete translation
368: operator, $T_{d}$. The analogy is complete when we note that
369: the eigenstates are linear combinations of angular momentum eigenstates, in
370: which case we call the number $\hbar m$ the quasi-angular momentum
371: of the state, $\psi _{m}\left( \phi ,\rho \right)$. For the rest of the
372: paper, we will drop the factor of $\hbar$ from the quasi-angular momentum.
373: \begin{figure}[tbp]
374: \centerline{
375: \includegraphics[width=3in]{figure5a_5b.pdf}
376: }
377: \caption{(Color online.) This diagram shows the equivalence between a one-dimensional
378: lattice with periodic boundary conditions and a ring lattice with the same
379: number of sites. The translation operator $T_{d}$ that translates through
380: one lattice site maps onto the rotation operator $R_{2\pi/8}$ that
381: rotates the lattice through one lattice site.}
382: \end{figure}
383:
384: \section{Bosons and fermions in rotating optical lattices}
385:
386: Interacting bosons in one dimension enter the Tonks-Girardeau regime when
387: interactions get very strong \cite{girardeau2004rbs,lieb1963eai}.
388: The gas takes on some of the characteristics
389: of a gas of non-interacting fermions, such as an identical number density. It differs
390: in certain aspects, however, such as the momentum distribution. It then becomes
391: particularly interesting to compare the cases of strongly-interacting bosons and
392: non-interacting fermions in rotating lattices. In this section, gases of non-interacting
393: fermions and hard-core bosons in rotating optical lattices are investigated.
394: A many-particle formalism is generated for quasi-angular momentum states and
395: used to monitor the symmetry of the ground state as a function of rotation.
396:
397: The Hamiltonian for interacting bosons or fermions in a rotating optical
398: lattice takes the second quantized form,
399: \begin{eqnarray}\fl
400: H =\sum_{\sigma }\int d^{D}x\hat{\Psi}_{\sigma }^{\dag }\left( \vec{x}%
401: \right) \left( -\frac{\hbar ^{2}}{2M}\nabla ^{2}+V_{lat}\left(\vec{x}\right)
402: -\Omega L_{z}\right) \hat{\Psi}_{\sigma }\left( \vec{x}%
403: \right) \nonumber \\
404: \fl+\sum_{\lambda \lambda ^{\prime },\mu \mu ^{\prime }}\int d^{D}x\int
405: d^{D}x\hat{\Psi}_{\lambda }^{\prime \dag }\left( \vec{x}\right)\hat{\Psi}_{\lambda
406: ^{\prime }}^{\dag }\left( \vec{x}^{\prime }\right) V\left( \vec{x},\vec{x}%
407: ^{\prime }\right) _{\lambda \lambda ^{\prime },\mu \mu ^{\prime }}
408: \hat{\Psi}_{\mu^{\prime }}\left( \vec{x}\right) \hat{\Psi}_{\mu }\left( \vec{x}^{\prime }\right),
409: \end{eqnarray}%
410: where $D$ is the dimension of the system, $V_{lat}\left(\vec{x}\right)$ is taken to be sinusoidal,
411: $\lambda$, $\mu$, and $\sigma$
412: are spin indices, and $\Psi _{\sigma }\left( \vec{x}\right) $, $\Psi _{\sigma }^{\dag
413: }\left( \vec{x}\right) $ are field operators satisfying the
414: (anti-)commutation relations,
415: \begin{eqnarray}
416: \left[ \hat{\Psi} _{\sigma }\left( \vec{x}\right) ,\hat{\Psi} _{\sigma ^{\prime }}^{\dag
417: }\left( \vec{x}^{\prime }\right) \right] _{\pm } &=&\delta _{\sigma \sigma
418: ^{\prime }}\delta \left( \vec{x}-\vec{x}^{\prime }\right), \nonumber \\
419: \left[ \hat{\Psi} _{\sigma }\left( \vec{x}\right) ,\hat{\Psi}_{\sigma ^{\prime }}\left(
420: \vec{x}^{\prime }\right) \right] _{\pm } &=&0.
421: \end{eqnarray}
422: Assuming spin-independent contact interactions, a Hubbard
423: Hamiltonian can be derived in a lowest-band, tight-binding approximation
424: by expanding the field operators in the lowest-band
425: Wannier functions, $w(\vec{x}-\vec{x}_j)$ \cite{jaksch1998cba}. A basis better
426: suited in the presence of rotation is given by the modified Wannier
427: functions,
428: \begin{equation}\label{modwan}
429: W_{j}\left( \vec{x}\right)
430: =\exp \left(-i\frac{M}{\hbar} \int^{\mathbf{x}}_{\mathbf{x}_j} \vec{A} \left( \vec{x}^{\prime} \right) \cdot d\vec{x}^{\prime} \right) w\left( \vec{x}-\vec{x}_{j}\right),
431: \end{equation}%
432: where $\vec{A}\left( \vec{x}^{\prime }\right) =\vec{\Omega}\times
433: \vec{x}^{\prime }$ \cite{bhat2006qvs}. This modification captures the effect of
434: rotation at low rotation speeds \cite{bhat2006qvs}; however, this approximation
435: needs to be further modified at higher rotation speeds where the density of the
436: Wannier functions is modified along with the phase.
437:
438: The result for spinless bosons is \cite{bhat2007he}
439: \begin{eqnarray}
440: \fl H_{B} =-\sum_{\left\langle i,j\right\rangle }\left( t+\frac{1}{2}M
441: \Omega^{2} A_{1}\right) \left(e^{-i\phi_{ij}}a_{i}^{\dag}%
442: a_{j}+e^{i\phi_{ij}}a_{j}^{\dag}a_{i}\right)\nonumber\\
443: +\sum_{j}\left(
444: \epsilon-\frac{1}{2}M\Omega^{2} \left(
445: r_{j}^{2}+A_{2}\right) \right) a_{j}^{\dag}a_{j}+\frac{U}{2}\sum_{j}%
446: a_{j}^{\dag}a_{j}^{\dag}a_{j}a_{j}.
447: \end{eqnarray}
448: A similar derivation for spin-$1/2$ fermions yields%
449: \begin{eqnarray}
450: \fl H_{F} =-\sum_{\left\langle i,j\right\rangle ,\sigma}\left( t+\frac{1}%
451: {2}M\Omega^{2} A_{1}\right) \left(e^{-i\phi_{ij}}
452: a_{i,\sigma}^{\dag}a_{j,\sigma}+e^{i\phi_{ij}}a_{j,\sigma}^{\dag}a_{i,\sigma
453: }\right)\nonumber\\
454: \fl +\sum_{j,\sigma}\left( \epsilon-\frac{1}{2}M
455: \Omega^{2} \left( r_{j}^{2}+A_{2}\right) \right)
456: a_{j,\sigma}^{\dag}a_{j,\sigma}+\frac{U}{2}\sum_{j}a_{j,\uparrow}^{\dag
457: }a_{j,\downarrow}^{\dag}a_{j,\downarrow}a_{j,\uparrow}.
458: \end{eqnarray}%
459: For both cases, $i,j$ are site indices with $\left\langle i,j\right\rangle $ indicating a
460: sum over only nearest neighbours, and $r_i$ is
461: the radial position of the $i$'th site. The phase $\phi_{ij}$ is given by the expression,
462: \begin{equation}
463: \phi_{ij}=\frac{M}{\hbar} \int^{\mathbf{x}_i}_{\mathbf{x}_j} \vec{A} \left( \vec{x}^{\prime}\right) \cdot d\vec{x}^{\prime}
464: =\frac{M\Omega }{\hbar}(x_i y_j-x_j y_i).
465: \end{equation}
466: The parameters $\epsilon$, $t$, and $U$ are the on-site zero-point energy, the hopping
467: parameter, and the on-site interaction energy, respectively, identical to those of the
468: standard Bose-Hubbard Hamiltonian \cite{jaksch1998cba}. They are given by the expressions,
469: \begin{eqnarray}
470: t \! &=& \! \int d^Dx ~ w^* (\vec{x}-\vec{x}_i) \left(-\frac{\hbar^2}{2M} \nabla^2 + V_{lat} \left(\vec{x}\right) \right) w(\vec{x}-\vec{x}_j), \\
471: \epsilon \! &=& \! \int d^Dx ~ w^* (\vec{x}-\vec{x}_i) \left(-\frac{\hbar^2}{2M} \nabla^2 + V_{lat} \left(\vec{x}\right) \right) w(\vec{x}-\vec{x}_i), \\
472: U&=&g \int d^Dx \left| w(\vec{x}-\vec{x}_i)\right|^4,
473: \end{eqnarray}
474: where $g$ is a two-particle interaction strength. The parameters $A_1$ and $A_2$ arise due to
475: the phase factor in equation \ref{modwan} and are given by
476: \begin{eqnarray}
477: A_1&=& \int dx ~ w^*\left(x-x_i\right) \left(x-x_i\right)^2 w\left(x-x_j\right),\\
478: A_2 &=& 2\int dx ~ w^*(x-x_i) \left( x-x_i \right) ^2 w\left(x-x_i\right),
479: \end{eqnarray}
480: where $w\left(x-x_i\right)$ are one-dimensional Wannier functions. All of these parameters
481: can be numerically evaluated for a lattice of specific shape, period, and depth
482: \cite{bhat2007he}. If a harmonic trap is included,
483: $\Omega ^{2}\rightarrow \Omega^{2}-\Omega _{T}^{2}$, where $\Omega _{T}$
484: is the trap frequency.
485:
486: In the rest of the paper, a lattice depth of $V_0=10E_R$ is assumed, and all
487: parameters in the Hubbard Hamiltonians are numerically computed for this depth.
488:
489:
490: \subsection{Strongly-interacting bosons}
491:
492: In this section, we describe hard-core bosons in a rotating optical lattice. The
493: quasi-angular momentum of an eigenstate for a one-dimensional system is then
494: defined and monitored as a function of rotation speed. The analytic results herein
495: derived are consistent with a previous numerical treatment of a two-dimensional system
496: \cite{bhat2006qvs}.
497:
498: In the limit of very strong interactions, $U\to\infty$, a gas of bosons enters the
499: hard-core boson regime. In an optical lattice, this regime
500: can be characterized by using a number basis where the occupation number of each site
501: is either $l$ or $l+1$, with $l$ an integer. We can encode this
502: fact in the Hamiltonian by formally changing the properties of the creation and
503: annihilation operators. Here we consider the case of filling factors less than one,
504: as the results for higher filling factors are qualitatively identical
505: (see the appendix for the general case). This can be effected by
506: stipulating on-site anti-commutation relations for the operators -- i.e.,
507: \begin{equation}
508: \left[ a_{i},a_{i}^{\dag }\right]_+ =1,~~~~\left[ a_{i},a_{j\neq i}\right]_-=%
509: \left[ a_{i},a_{j\neq i}^{\dag }\right]_- =0.
510: \end{equation}
511: In this case, the interaction term vanishes, although strong-interactions still
512: implicitly exist in the Hamiltonian. For practical purposes, $U=100t$, where
513: $t$ is the tunneling energy, is large enough to enter this regime \cite{bhat2006qvs}.
514:
515: The Hamiltonian for a ring-lattice is given by
516: \begin{eqnarray}\label{1DBHH}\fl
517: H =-\sum_{j=1}^{N}\left( t+\frac{M\Omega^{2}}{2}
518: A_{1}\right)e^{i\Omega\Phi}a_{j+1}^{\dag}a_{j}+h.c.\nonumber\\
519: +\sum_{j=1}^{N}\left( \epsilon-\frac{M\Omega^{2}}{2}M\Omega^{2}
520: \left( R^{2}+A_{2}\right) \right) a_{j}^{\dag}a_{j},
521: \end{eqnarray}
522: where $R$ is the radius of the ring. The parameter $\Phi$ is given by
523: \begin{equation}
524: \Phi=\frac{M}{\hbar}(x_i y_j-x_j y_i),
525: \end{equation}
526: which is a constant on a ring. This Hamiltonian can be diagonalized by first performing the Jordan-Wigner transformation \cite{jordan1928upa,schultz1963ipb},
527: \begin{equation}
528: b_{j}=a_{j}e^{i\pi \sum_{i=1}^{j-1}a_i^{\dag }a_i},
529: \end{equation}%
530: which transforms the mixed $a$ operators into fully fermionic ones. The
531: transformed Hamiltonian takes different forms for even and odd numbers of
532: particles. Performing the canonical transformation,
533: \begin{equation}
534: d_{m}=\frac{1}{\sqrt{N}}\sum_{j=1}^{N}e^{-i\pi f_{m}j/N}b_{j},
535: \end{equation}%
536: where $f_{m}=2m$ for $n$ odd, $f_{m}=2m-1$ for $n$ even, and $n$ is the number
537: of particles, results in
538: \begin{eqnarray}
539: \label{egysp1} \fl H=\sum_{m=1}^{N}E_m d_{m}^{\dag}d_{m},\\
540: \label{egysp} \fl E_m=-2\left( t+\frac{1}{2}M\Omega^{2}
541: A_{1}\right) \cos\left( \frac{\pi f_{m}}{N}-\Omega\Phi\right)
542: +\epsilon-\frac{1}{2}M\Omega^{2} \left( R^{2}+A_{2}\right)
543: \end{eqnarray}%
544: The eigenstates are of the form $d_{m_{1}}^{\dag}\dots d_{m_{n}}^{\dag }
545: \left\vert 0\right\rangle$.
546:
547: A complete picture of this diagonalization requires an interpretation of the
548: eigenstates in terms of quasi-angular momentum. The discrete rotation operator
549: is defined via
550: \begin{equation}
551: Rb_{j\neq N}^{\dag }R^{-1}=b_{j+1}^{\dag },~~~~Rb_{N}^{\dag }R^{-1}=\left(
552: -1\right) ^{n}b_{1}^{\dag },~~~~R\left\vert 0\right\rangle =\left\vert
553: 0\right\rangle .
554: \end{equation}%
555: This operator commutes with the Hamiltonian, and its
556: action on an eigenstate, $\left\vert \psi \right\rangle =d_{m_{1}}^{\dag
557: }\dots d_{m_{n}}^{\dag }\left\vert 0\right\rangle $, is%
558: \begin{equation}
559: R\left\vert \psi \right\rangle =e^{-i\pi \left( f_{m_{1}}+\cdots
560: +f_{m_{n}}\right) /N}\left\vert \psi \right\rangle .
561: \end{equation}%
562: Therefore, the quasi-angular momentum of this state is
563: $\left(\frac{f_{m_{1}}+\cdots +f_{m_{n}}}{2}\right) \mod~N$.
564: A pictorial representation of this discussion is shown in figure 6 for a specific case
565: with one particle.
566: \begin{figure}[tbp]
567: \centerline{
568: \includegraphics[width=4in]{figure6a_6b.pdf}
569: }
570: \caption{(Color online.) (a) The operator $d_{1}^{\dag }$ adds one particle to the system,
571: populating every site with equal probability but a phase that differs by $2%
572: \pi /8$ between lattice sites. The state that is illustrated is $%
573: \left\vert\psi \right\rangle =d_{1}^{\dag }\left\vert
574: 0\right\rangle $; arrows indicate phase. (b) When this state is rotated by
575: one lattice site, the resulting state is $R_{2\pi /8}\left\vert
576: \psi \right\rangle $. Subtracting the phases at a site before and
577: after rotation, we get $\phi=-2\pi /8$. The picture then
578: illustrates the fact that $\left\vert\psi\right\rangle$ is an
579: eigenstate of $R_{2\pi /8}$ with eigenvalue $e^{-i2\protect\pi /8}$. }
580: \end{figure}
581:
582: We are now in a position to use this formalism to investigate the ground state of the system.
583: The ground state is determined by populating the lowest energy levels in the standard
584: way for fermions. Equation \ref{egysp} is the single-particle energy spectrum $E_m$.
585: As $\Omega$ is increased, the minimum value of the cosine moves to the
586: right, as illustrated in figure 7(a). The ground state is then determined by populating
587: the lowest energy $d_{m}^{\dag }d_{m}$ eigenstates.
588: In figures 7(b) and 7(c), this process is illustrated for
589: the case of three particles in an eight-site ring.
590: \begin{figure}
591: \centerline{
592: \includegraphics[width=4.5in]{figure7a_7b.pdf}
593: }
594: \caption{(Color online.) Ground state of three strongly-interacting bosons in a rotating
595: 8-site ring lattice. Solid curves are guides for the eye. (a) At $\Omega=0$, the three
596: lowest single-particle energies carry
597: quasi-angular momentum $m=-1,0,1$. The three-particle ground state therefore
598: has quasi-angular momentum $m=\left( -1+0+1\right)\mod~8=0$. (c) At
599: $\Omega=0.14E_R/\hbar$, the three lowest single-particle energies carry quasi-angular
600: momentum $m=0,1,2$. The three-particle ground state therefore has
601: quasi-angular momentum $m=\left( 0+1+2\right) \mod~8=3$.}
602: \end{figure}
603:
604: In general, if the minimum is near $m=m_0$, we populate states
605: $m_0,~m_0\pm1,\dots,~m_0\pm p$ for $n=2p+1$ particles in the system. The
606: quasi-angular momentum of the state is then
607: \begin{equation}
608: \fl m=\sum_{l=-p}^p \off{m_0+l}~\mod~N=m_0\off{2p+1}~\mod~N=m_0n~\mod~N.
609: \end{equation}
610: A similar argument holds for even numbers of particles. Thus, as rotation is
611: increased from $\Omega=0$, the quasi-angular momentum cycles through the values,
612: \begin{equation}
613: \label{mb} m=jn~\mod~N:j=0,1,2,\dots.
614: \end{equation}
615:
616: One final note for the case of bosons. If there is an infinitesimal
617: potential, $V$, that breaks the symmetry of the original Hamiltonian, $H_0$,
618: down to a four-fold rotational symmetry, the behavior described in section 2.2 obtains.
619: We treat the problem perturbatively. Tiny bandgaps in the single-particle energy
620: spectrum, equation \ref{egysp}, open up at the new Brillouin zone boundaries. The
621: eigenstates of $H_0$ are independent of $\Omega$. Therefore, to first
622: order in $V$, the energy level shifts are independent of $\Omega$, and the
623: eigenstates remain unchanged. The effect of rotation is then exactly the
624: same as in the non-perturbed case, and the quasi-angular momenta have the
625: same behavior as a function of rotation. However, since the old
626: quasi-angular momentum is no longer a good quantum number, we have to take
627: \begin{equation}
628: m\to m~\mod~4.
629: \end{equation}
630: These results are consistent with the treatment in a previous paper \cite{bhat2006qvs}.
631:
632:
633: \subsection{Non-interacting fermions in a ring lattice}
634:
635: Previous results focused on strongly-interacting bosons in rotating optical lattices
636: \cite{bhat2006bec,bhat2006qvs}. Here, we describe non-interacting fermions
637: in ring lattices and compare to the boson case. The quasi-angular
638: momentum of the ground state changes as rotation is increased. The allowed
639: values of this quantity for fermions differ from those of the boson case.
640:
641: For non-interacting fermions, we drop the spin index, and the Hamiltonian for
642: a ring lattice is identical to equation \ref{1DBHH} except that the operators
643: are fermionic. This Hamiltonian can be analytically diagonalized via the
644: canonical transformation
645: \begin{equation}
646: d_{m}=\frac{1}{\sqrt{N}}\sum_{j=1}^{N}e^{-i2\pi mj/N}a_j.
647: \end{equation}
648: The Hamiltonian is identical to that of equations \ref{egysp1} and \ref{egysp} except that
649: $f_m=2m$ for both even and odd numbers of particles.
650: The eigenstates, $d_{m_{1}}^{\dag }\cdots d_{m_{n}}^{\dag
651: }\left\vert 0\right\rangle $, carry quasi-angular momentum $m=\left(
652: m_{1}+\cdots +m_{n}\right) \mathrm{mod}N$, where $n$ is the number of particles.
653: A discussion similar to the one given for bosons yields $m$ as a function of rotation:
654: \begin{eqnarray}
655: m=nj~\mod~N;~~~~j=0,1,2,\dots;~~~~n~odd,\\
656: \label{mfeven} m=n\left( j+\frac{1}{2}\right)~\mod~N;~~~~j=0,1,2,\dots;~~~~n~even.
657: \end{eqnarray}
658:
659: As expected, the single-particle cases for both bosons and fermions are identical.
660: It turns out that the cases of odd numbers of particles also coincide.
661: However, there is an interesting distinction between the two cases when the number
662: of particles in the system is even (compare equations \ref{mb} and \ref{mfeven}).
663: For instance, a system of two fermions in a four-site
664: lattice cycles between quasi-angular momenta $1$ and $3$ whereas a system of two
665: bosons cycles between values $2$ and $4$. This is one way in which non-interacting
666: fermions and strongly-interacting bosons in one dimension differ despite the
667: Jordan-Wigner transformation that maps between the two cases.
668:
669:
670: \subsection{Non-interacting fermions in a two-dimensional square lattice}
671:
672: Differences between fermions and bosons remain when considering two-dimensional
673: systems. Here we describe non-interacting fermions in a two-dimensional $4\times4$
674: square lattice.
675:
676: The Hamiltonian can
677: be written as the sum of three terms, $H=H_{12}+H_{4}+V$, where $H_{12}$
678: is the Hamiltonian of the outer twelve-site ring, $H_{4}$ is the Hamiltonian of
679: the inner four-site ring, and $V$ is the hopping between the rings,
680: \begin{eqnarray}
681: V=-\sum_{\left\langle i,j\right\rangle}\left(t+\frac{M\Omega^{2}}{2}A_{1}\right)
682: a_{i}^{(12)\dag}a_{j}^{(4)}e^{-i\phi_{ij}}+h.c.,
683: \end{eqnarray}
684: where the superscripts $4$ and $12$ indicate whether site $j$ is on the outer
685: $12$-site or inner $4$-site ring.
686: The Hamiltonians $H_{4}$ and $H_{12}$ can be separately diagonalized
687: (apart from the term in $H_{12}$ involving $r_i^2$) via the transformations,
688: \begin{equation}
689: d_{m}^{\left( N\right) }=\frac{1}{\sqrt{N}}\sum_{j=1}^{N}e^{-i2\pi
690: mj/N}a_j^{\left( N\right) }.
691: \end{equation}
692:
693: In figure 8(a) is plotted the single-particle energy spectrum for the
694: Hamiltonian in the absence of hopping between the rings ($V=0$). When the
695: hopping is ``turned on'' (figure 8(b)), level crossings become avoided
696: crossing due to the mixing of states with the same four-fold quasi-angular
697: momentum. In figure 8(c), the case for zero hopping between rings is plotted for
698: $\Omega=0.05E_R/\hbar$, whereas in figure 8(d), the hopping has turned
699: on, and the avoided crossing behavior is observed.
700: \begin{figure}
701: \centerline{
702: \includegraphics[width=4.5in]{figure8a_8b_8c_8d.pdf}
703: }
704: \caption{(Color online.) Single-particle energy spectrum in a $4\times4$ square
705: lattice for (a) zero rotation and zero hopping between inner and outer rings,
706: (b) zero rotation and non-zero
707: hopping, (c) $\Omega=0.05E_R/\hbar$ and zero hopping, and (d)
708: $\Omega=0.05E_R/\hbar$ and non-zero hopping between inner and outer rings.
709: The dashed and solid line are aids to the eye that indicate the band. In (a) and (c),
710: the red dashed line is the spectrum for the inner 4-site lattice.}
711: \end{figure}
712:
713: In figure 9 are plotted the quasi-angular momenta of the ground state as a
714: function of rotation speed for $1$, $2$, $3$, $4$, and $5$ non-interacting
715: fermions in a sixteen-site lattice. Similar to the ring lattice case, the results for
716: odd numbers of particles are identical to those for hard-core bosons, but differ
717: for even numbers of particles. However, there is an extra distinction that does not
718: occur in the ring lattice. The pattern of $m$ values taken on for five particles
719: (see figure 9(e)) differs markedly from that of one particle (see figure 9(a)), whereas
720: for bosons these patterns are essentially identical \cite{bhat2006qvs}.
721: \begin{figure}
722: \centerline{
723: \includegraphics[width=4.2in]{figure9a_9b_9c_9d_9e.pdf}
724: }
725: \caption{(Color online.) Quasi-angular momentum as a function of rotation for (a) 1, (b) 2,
726: (c) 3, (d) 4, and (e) 5 non-interacting fermions in a sixteen site square
727: lattice. These patterns repeat for larger rotation rates.}
728: \end{figure}
729:
730:
731: \section{Signatures of quasi-angular momentum in the momentum distribution}
732:
733: While the quasi-angular momentum $m$ is a good quantum number and a useful
734: tool in investigating the symmetry properties of the ground state, it is not
735: a quantity that can be directly measured in experiments. Instead, experiments are
736: routinely performed in which time-of-flight, far-field, density images of the gas
737: contain directly information about the momentum distribution before expansion. As a first
738: step towards detecting quasi-angular momentum states, we look for a signature of
739: quasi-angular momentum in the momentum distribution of a state.
740:
741: The momentum distribution of the many-body state, $\ket{\psi}$, is given by
742: \begin{equation}
743: n\of{\vec{k}}=\bra{\psi} \hat{\Psi}^{\dag}(\vec{k}) \hat{\Psi}\of{\vec{k}} \ket{\psi},
744: \end{equation}
745: where $\hat{\Psi}(\vec k)$ is the Fourier transform of the field operator $\hat{\Psi}(\vec x)$.
746: Recalling that we expand the field operators as
747: \begin{equation}
748: \hat{\Psi}(\vec{x})=\sum_j a_j W_j(\vec{x})=\sum_j a_j e^{-i\phi_{ij}}w\left( \vec{%
749: x}-\vec{x}_{j}\right) ,
750: \end{equation}
751: it can be shown that the momentum distribution takes the form
752: \begin{equation}\label{momdist}
753: n\of{\vec{k}}=\sum_{ll^{\prime }}w^{\ast }\of{\vec{k}_{l}}
754: w\of{\vec{k}_{l^{\prime }}} e^{i\vec{k}\cdot \left( \vec{x}_{l}-\vec{x}_{l^{\prime }}\right) }
755: \left\langle\psi \right\vert a_{l}^{\dag }a_{l^{\prime }}\left\vert \psi \right\rangle,
756: \end{equation}
757: where
758: \begin{equation}\label{kl}
759: \vec{k}_l=\vec{k}+\Omega\frac{M}{\hbar}\vec{x}_l\times\vec{\hat{z}}.
760: \end{equation}%
761: This distribution is written in the momentum coordinates, $\vec{k}$, corresponding
762: to the rotating frame coordinates, $\vec{x}$. Since the number density rotates in the lab
763: frame coordinates, $\vec{x}_L$, the momentum distribution also rotates in the lab frame
764: momentum coordinates, $\vec{k}_L$. However, far-field pictures of the gas will be
765: snapshots of the momentum distribution at the moment the trap and lattice are turned off,
766: provided the switch-off time is fast enough. Thus, the momentum distributions presented
767: here are accurate representations of what will be measured in time-of-flight measurements,
768: although they may be rotated relative to each other.
769:
770: Since quasi-angular momentum is a reflection of the symmetry of the system, considering
771: single-particle states should be sufficient to capture the basics of how this quantum
772: number affects the momentum distribution. We therefore concentrate on single-particle
773: states only. In this case, $n\of{\vec{k}}$ is merely the square of the Fourier transform of
774: the wavefunction.
775:
776: In order to build up an understanding of the momentum distribution, first drop
777: the Wannier functions. In this case, $n\of{\vec{k}}$ is the Fourier transform of a sum
778: of weighted delta functions. On a four site-lattice, the wavefunction is of the form
779: \begin{equation}
780: \psi \left( \vec{x} \right)=\sum_{j=1}^4 e^{i2\pi mj/4} \delta \left( \vec{x}-\vec{x}_j \right),
781: \end{equation}
782: where $\vec{x}_j$ are the four corners of a square centered at $\vec{x}=0$. In figure
783: 10 are plotted the momentum distributions for $m=0$, $1$, and $2$.
784: The $m=0$ is peaked at $\vec{k}=2\pi n/d$, with $n$ an integer and $d$ the lattice
785: spacing, whereas the other
786: states vanish at these points. The spacing between peaks for both $m=0$ and $m=2$
787: is $\Delta k=2\pi/d$, whereas for $m=1,3$ this spacing is $\Delta k=2\pi/%
788: \sqrt{2}d$; these can be calculated analytically. Thus, since the lattice spacing is known,
789: there are clear measurable distinctions between different generic quasi-angular
790: momentum states. Note that $m=1$ and $m=3$ are identical as they represent similar
791: states with counter-propagating current patterns.
792: \begin{figure}[h]
793: \centerline{
794: \includegraphics[width=5.2in]{figure10a_10b_10c_10d.pdf}
795: }
796: \caption{(Color online.) Fourier transform of generic (a) $m=0$, (b) $m=1,3$, and (c) $m=2$
797: quasi-angular momentum states on four sites. These momentum distributions
798: are plotted over a range $4\pi/d \leq k_x,k_y \leq 4\pi/d$;
799: the first Brillouin zone is given by $\pi/d \leq k_x,k_y \leq
800: \pi/d$. (d) A schematic of the peak structure of the momentum
801: distribution of four delta functions of the $m=0$ (blue), $m=1,3$ (green),
802: and $m=2$ (red) quasi-angular momentum states. The peak spacings for the $m=0
803: $ and $m=2$ states are both $2\pi/d$, where $d$ is the spacing
804: between the delta functions, whereas the peak spacing for $m=1,3$ is $2
805: \pi/\sqrt{2}d$.}
806: \end{figure}
807:
808: Re-including the Wannier functions contributes to the
809: momentum distribution an overall envelope of approximate width $\sqrt{V/E_R}/4$,
810: where $V$ is the depth of the lattice. The modification $\vec{k}\to\vec{k}_l$, equation
811: \ref{kl}, moves the peak of this envelope away from $\vec{k}=0$ as rotation is increased,
812: though the center of the envelope is always located at $\vec{k}=0$. This is consistent with
813: higher momenta being accessed for higher rotation rates.
814:
815: At this point, it is important to note that locating $\vec{k}=0$ in the distribution is necessary
816: for distinguishing the $m=0$ and $m=2$ states. This can be done easily due to the overall
817: envelope in momentum space that is centered at $\vec{k}=0$.
818:
819: There are two additional considerations that come into play for larger lattices,
820: both system-size effects. An overall envelope in position space determines the width
821: of the peaks in momentum space. This envelope determines the size of the system,
822: and if the width is roughly $L$, then the width of the peaks in momentum space scales
823: as $1/L$. If the lattice spacing is increased by a factor of $n$, then the peak-spacing in
824: momentum space is decreased be a factor of $1/n$. The general peak structure is not
825: modified by these considerations.
826:
827: Since the peak structure is not modified by lattice-size effects, we can conclude that
828: the distinctions between quasi-angular momentum states are sustained for more
829: physical systems, which we turn to next. The ground state momentum distribution
830: for a one-hundred-site lattice in the presence of a harmonic trap of frequency
831: $\Omega_T = 0.15E_R/\hbar$ is computed via imaginary time propagation. The
832: trap ensures that the majority of the number density resides on the inner 36 sites.
833: \begin{figure}[h]
834: \centerline{
835: \includegraphics[width=5in]{figure11a_11b_11c_11d_11e_11f_11g_11h_11i.pdf}
836: }
837: \caption{(Color online.) Momentum distributions computed via imaginary time
838: propagation for one particle in a 100-site lattice in the presence of a trap of
839: frequency $\Omega_T=0.15E_R/\hbar$. Plotted
840: are $m=0$ ((a), (b), and (c)), $m=1$ ((d), (e), and (f)), and $m=2$ ((g), (h), and (i))
841: states for rotation speeds of $\Omega=0$ ((a), (d), and (g)), $\Omega=0.1E_R/\hbar$
842: ((b), (e), and (h)), and $\Omega=0.145E_R/\hbar$ ((c), (f), and (i)). At zero rotation, the
843: distinctions between different quasi-angular momenta are sustained. As
844: rotation is increased, it becomes harder to distinguish $m=1$ and $m=2$
845: states. Due to the centrifugal term, higher rotation increases the width of the
846: envelope of the number density, decreasing the width of the peaks in
847: momentum space. Not shown is $m=3$ which has a structure similar to both
848: $m=1$ and $m=2$ for large rotation but has a ring-radius larger than both of them.}
849: \end{figure}
850:
851: The results are summarized in figure 11, in which are plotted the momentum
852: distributions for $m=0$, $1$, and, $2$ states at rotation speeds of $0$, $0.1E_R/\hbar$,
853: and $0.145E_R\hbar$,
854: increasing to the right. We can again make the observation that the $m=0$ state
855: is the only one that is non-zero at exactly the reciprocal lattice vectors, i.e.
856: $\vec{k}=\frac{2\pi n}{d}\vec{\hat{k}}_x+\frac{2\pi m}{d}\vec{\hat{k}}_y$. The peak-spacing is
857: smaller for the $m=1$ state than for the $m=2$ state, distinguishing them from each other.
858: However, the exact peak-spacing is unclear due to complex interference effects between
859: sites at different radii.
860:
861: These differences are washed out at higher rotation speeds. The $m=1$, $m=2$,
862: and $m=3$ (not pictured) states have very similar structures for large rotation. The peaks
863: overlap so that it is very difficult to resolve them; the momentum distributions then all appear
864: to be rings centered at reciprocal lattice vectors. They are distinguishable via the radius
865: of the ring, as the radius is smallest for $m=1$ and largest for $m=3$, but as this is
866: in order of increasing energy, the larger radius may just be an artifact of the larger
867: width in the number density envelope at higher energies. Indeed, figures 11a, 11b, and 11c
868: are consistent with this conclusion, as the momentum peaks are narrowing due to
869: the number density spreading out in real space. These results are consistent with
870: those calculated via direct diagonalization of the Hubbard models previously discussed.
871:
872:
873: \section{Conclusion}
874:
875: By analogy with quasi-momentum in translationally invariant periodic
876: systems, the notion of quasi-angular momentum was introduced to label the
877: eigenstates of a Hamiltonian that has a discrete rotational symmetry. It was
878: shown that quasi-angular momentum is useful in analyzing the
879: ground-state properties of quantum gases of bosons or fermions in rotating
880: optical lattices. In particular, monitoring the quasi-angular momentum of the ground
881: state as a function of rotation allowed us to identify transitions between different
882: circulation values.
883:
884: We also presented a possible avenue by which the quasi-angular
885: momentum of a state can be experimentally determined. We identified
886: characteristics in the momentum distribution distinguishing between different
887: quasi-angular momentum states at low rotation speeds, such as the existence
888: of a peak at reciprocal lattice vectors for an $m=0$ state only, the peak-spacing, and the
889: overall structure of the momentum distribution.
890:
891: There are still open questions as to how the momentum distributions will
892: change when the lattice size or number of particles is increased. The effects of
893: statistics will be of fundamental importance for many-particle systems; there is a
894: question of how well the peaks can be resolved for larger lattices, since peak-%
895: spacing decreases with increasing lattice size; and interference effects
896: between adjacent sites and between rings is not rigorously treatable. However,
897: there will always be signatures of the quasi-angular momentum in
898: the momentum distribution, as it is connected with the phase
899: information of the ground state, which in turn influences the momenta of
900: the system. In addition, a zero quasi-angular momentum
901: state can be distinguished from a non-zero one, which allows the experimenter
902: to verify that vorticity has entered the system.
903:
904: \section*{Acknowledgements}
905:
906: We acknowledge extremely useful discussions with both Brian Seaman and John
907: Cooper. The authors would also like to acknowledge funding support from the
908: Department of Energy, Office of Basic Energy Sciences via the Chemical
909: Sciences, Geosciences, and Biosciences Division, NASA, and Deutsche
910: Forschungsgemeinschaft (MK).
911:
912: \appendix
913:
914: \section*{Appendix}
915:
916: \setcounter{section}{1}
917:
918: The case of fillings between $l$ and $l+1$ can be encoded in the Hamiltonian
919: by formally changing the properties of the creation and annihilation
920: operators. As mentioned in the body of the paper, for fillings less than
921: one, the creation and annihilation operators satisfy the following
922: relations:
923: \begin{equation}
924: \left\{ a_{i},a_{i}^{\dag }\right\} =1,~~~~\left[ a_{i},a_{j}\right] =\left[
925: a_{i},a_{j}^{\dag }\right] =0.
926: \end{equation}
927: If the filling fraction is between $l$ and $l+1$, the following changes are
928: made. First, the site number operator $a_{j}^{\dag }a_{j}$ is replaced by $%
929: a_{j}^{\dag }a_{j}+l$, reflecting the fact that each site is filled with
930: between $l$ and $l+1$ particles. Secondly, recalling that the operators were
931: originally bosonic, $a_{j}^{\dag }\left\vert l\right\rangle =\sqrt{l+1}%
932: \left\vert l+1\right\rangle $ and $a_{j}\left\vert l+1\right\rangle =\sqrt{%
933: l+1}\left\vert l\right\rangle $, one can see that the hopping parameter $t$
934: is scaled by a factor of $l+1$. Finally, the interaction term gets modified
935: according to the replacement $a_{j}^{\dag }a_{j}\rightarrow a_{j}^{\dag
936: }a_{j}+l$:
937: \begin{equation}
938: \sum_{j=1}^{N}a_{j}^{\dag }a_{j}\left( a_{j}^{\dag }a_{j}-1\right)
939: \rightarrow 2l\sum_{j=1}^{N}a_{j}^{\dag }a_{j}+Nl\left( l-1\right)
940: \end{equation}
941: The Hamiltonian in two-state approximation is then given by%
942: \begin{eqnarray}
943: \fl H =\left( Ul-\mu \right) \sum_{j=1}^{N}a_{j}^{\dag }a_{j}-\mu Nl+UN\frac{%
944: l\left( l-1\right) }{2}
945: -\left( l+1\right) \sum_{j=1}^{N}te^{i\phi}a_{j+1}^{\dag }a_{j}+h.c.
946: \end{eqnarray}
947: where the chemical potential $\mu$ has been introduced so that we can work
948: in the grand-canonical ensemble.
949:
950: Upon applying the Jordan-Wigner transformation,
951: \begin{equation}
952: b_{j}=a_{j}e^{i\pi \sum_{k=1}^{j-1}a_{k}^{\dag }a_{k}},
953: \end{equation}
954: the Hamiltonian becomes
955: \begin{eqnarray}
956: \fl H=\left( Ul-\mu \right) \sum_{j=1}^{N}b_{j}^{\dag }b_{j}-\mu Nl+UN\frac{%
957: l\left( l-1\right) }{2}
958: -\left( l+1\right) \sum_{j=1}^{N-1}te^{i\phi}b_{j+1}^{\dag }b_{j}+h.c. \nonumber \\
959: -\left( -1\right) ^{n+1}\left( l+1\right)te^{i\phi}b_{1}^{\dag }b_{N}+h.c.
960: \end{eqnarray}%
961: where $n$ is the number of particles. Now that the Hamiltonian is fully
962: fermionic, we can perform a canonical transformation to diagonalize it,%
963: \begin{equation}
964: d_{m}=\frac{1}{\sqrt{N}}\sum_{k=1}^{N}e^{i\pi f_{m}k/N}b_{k},
965: \end{equation}%
966: where $f_{m}=2m$ for $n$ odd and $f_{m}=2m-1$ for $n$ even. The resulting
967: Hamiltonian is
968: \begin{eqnarray}
969: H =-\mu Nl+\frac{U}{2}Nl\left( l-1\right) +\sum_{m=1}^{N}E_{m}d_{m}^{\dag
970: }d_m, \nonumber \\
971: E_m =-\mu -2\left( l+1\right)t\cos\left(\frac{\pi f_m}{N}-\phi\right)
972: +Ul.
973: \end{eqnarray}%
974: Setting $l$ and $\mu $ to zero, we realize the Hamiltonian derived in Sec. 3.
975:
976: As an aside, this calculation actually yields more than just the description of
977: the quasi-angular momentum of a rotating ring lattice. In the limit of large $N$,
978: this system is indistinguishable from a linear lattice. Taking $t$ real, the
979: Mott-insulator/superfluid phase diagram in two-state
980: approximation can be derived. Allowing the number of particles to
981: vary, it is apparent that the number of particles is determined by the
982: number of $m$'s for which $E_{m}<0$. Varying the chemical potential $\mu$,
983: every time there exists an $m$ such that $E_{m}=0$, there is a phase
984: boundary between ground states of different numbers of particles. Thus, the
985: phase boundaries are given by the expression,
986: \begin{equation}
987: \frac{\mu }{U}=-2\left( l+1\right) \frac{t}{U}\cos \left( \frac{\pi f_{m}}{N}\right) +l.
988: \end{equation}
989: In particular, it is possible to identify regions of integer filling and
990: non-integer filling, the boundaries between which are given by
991: \begin{equation}
992: \frac{\mu }{U}=\pm 2\left( l+1\right) \frac{t}{U}+l
993: \end{equation}
994: for a large number of sites. These boundaries yield an approximation to the
995: Mott-insulator superfluid phase diagram (figure A1). They have been derived
996: in a slightly modified form in \cite{casetti2002vsc} using a semiclassical approach. We note that our
997: method can be easily extended to the case of superlattices, generating for
998: instance Mott phases of half-filling in the phase diagram.
999: \begin{figure}[tbp]
1000: \centerline{
1001: \includegraphics[width=3in]{figureA1.pdf}
1002: }
1003: \caption{(Color online.) Bose-Hubbard phase diagram for an infinite
1004: one-dimensional lattice in two-state approximation.}
1005: \end{figure}
1006:
1007: \section*{References}
1008:
1009: \bibliographystyle{unsrt}
1010: \bibliography{qambib}
1011:
1012: \end{document}
1013: