cond-mat0506415/mit.tex
1: \documentclass[twocolumn,showpacs,prl,amsmath,amssymb]{revtex4}
2: \usepackage{graphicx}% Include figure files
3: \usepackage{bm}
4: \begin{document}
5: 
6: \title{Cold Atom Optical Lattices as Quantum Analog Simulators for
7: Aperiodic One-Dimensional Localization Without Disorder}
8: \author{V.W. Scarola and S. \surname{Das Sarma}}
9: \affiliation{Condensed Matter Theory Center, 
10: Department of Physics, University of Maryland,
11: College Park, MD 20742-4111}
12: 
13: \begin{abstract}
14: Cold atom optical lattices allow for the study of quantum localization
15: and mobility edges 
16: in a disorder-free environment.  We predict the existence of an 
17: Anderson-like insulator with sharp mobility edges in a 
18: one-dimensional nearly-periodic 
19: optical lattice.  We show that the
20: mobility edge manifests itself as the early onset of pinning in center
21: of mass dipole oscillations in the presence of a magnetic trap which
22: should be observable in optical lattices.  
23: \end{abstract}
24: \pacs{03.75.Lm, 03.75.Kk, 32.80.Pj, 05.60.Gg}
25: \maketitle
26: 
27: \newcommand\degrees[1]{\ensuremath{#1^\circ}}
28: 
29: Optical lattices incorporating ultracold atomic condensates are
30: rapidly becoming ideal quantum systems for studying various model
31: Hamiltonians developed earlier for studying solid-state 
32: phenomena.  This is primarily due to the extraordinary level
33: of precise tunability that experimentalists have achieved in
34: controlling the parameters (e.g. hopping, interaction and lattice
35: periodicity) of the optical lattice, which makes
36: it possible for the cold atom optical lattice to operate as an ideal
37: quantum analog simulator for various many-body condensed matter
38: Hamiltonians.  By contrast, ideal model Hamiltonians 
39: (e.g. Hubbard and Anderson models) often poorly describe
40: solid-state systems since experimental control over
41: complex condensed matter systems is, in general, quite limited.  In
42: addition solid-state systems are invariably contaminated by
43: unknown disorder, defects, and impurities whose effects are not easy to
44: incorporate in model Hamiltonians.  The cold atom optical lattices are
45: therefore becoming increasingly important in advancing our knowledge
46: about the quantum phase diagram and crossover in model many-body
47: Hamiltonians of intrinsic interest.  Examples include: the
48: Bose-Hubbard model \cite{Greiner}, the Tonks-Girardeau gas 
49: \cite{Paredes}, and the BEC-BCS
50: crossover \cite{cross}.  
51: 
52: In addition to studying strong correlation effects (e.g. the
53: superfluid-Mott insulator transition in the Bose-Hubbard model) in
54: many-body Hamiltonians, cold atom optical lattices
55: also offer ideal systems for studying quantum transport phenomena
56: including ballistic quantum transport \cite{Oskay,dArcy,Sadgrove} and quantum
57: localization \cite{Drese,Guidoni,Pezze,Huckans}.  The latter may be
58: more generally classified as metal-insulator transition 
59: phenomena with a direct relationship to the solid-state.  The distinction 
60: between a ``metal'' (i.e. a system with
61: finite resistivity at zero temperature) and an ``insulator'' (i.e. a
62: system with infinite zero temperature resistivity) is purely 
63: quantum.  Broadly speaking, there are four classes of
64: metal-insulator transitions in quantum lattice systems: Metal-band 
65: insulator transition in an ordered periodic lattice 
66: arising from the chemical potential
67: moving into energy band gaps; interaction induced metal-insulator
68: transition as in the Mott transition; disorder induced quantum
69: localization (i.e. Anderson localization \cite{Anderson}); and 
70: quantum localization in
71: aperiodic (but deterministic) potentials in disorder-free lattice
72: systems.  
73:  
74: In this paper, we establish that very
75: general aspects of the metal-insulator transition phenomena (in the
76: disorder-free environment) can be directly experimentally studied
77: in aperiodic cold atom optical lattices with the tuning of experimental
78: parameters leading to the observation of {\em both} band and
79: quantum (Anderson-like) localization in the same system but in different
80: parameter regimes.  Such an experimental study of localization or
81: insulating transitions in deterministic aperiodic systems is 
82: impossible in solid state lattice systems since disorder (which
83: leads to direct Anderson localization) is invariably present in solid
84: state systems overwhelming any subtle localization effects
85: arising from deterministic aperiodic potentials.  In particular,
86: all states are localized in one-dimensional systems in the 
87: presence of any disorder whereas one-dimensional aperiodic potentials
88: allow for the existence of extended quantum eigenstates.  This makes
89: one-dimensional optical lattice systems particularly interesting from
90: the perspective of localization studies in deterministic aperiodic
91: potentials since such studies in the corresponding one-dimensional
92: solid-state systems are essentially impossible due to disorder
93: effects.  We therefore consider aperiodic quantum localization 
94: in one-dimensional optical lattices, conclusively establishing the
95: feasibility of studying this unusual phenomenon in cold atom optical
96: lattices.      
97: 
98: The single-particle quantum localization problem in a deterministic 
99: quasiperiodic potential (i.e. two lattice potentials with mutually
100: incommensurate periods) has a long history \cite{Sokoloff,Last}.  In
101: particular, localization properties have been extensively studied in
102: the Harper (or, equivalently, Aubry) model which has an intriguing
103: self-dual point where the eigenstates form a multifractal Cantor set
104: spectra and are neither localized nor extended.  
105: Away from the dual point conventional wisdom dictates that all
106: states, as a function of the chemical potential, are either 
107: all extended or all localized, depending on the mutual strengths of the 
108: potential and hopping terms.  Such Harper model
109: type quasiperiodic potentials therefore do not allow for the existence
110: of a mobility edge separating extended states (above the mobility
111: edge) from localized states (below the mobility edge) which is the
112: hallmark of the Anderson localization transition in three-dimensional
113: disordered system.  Central to our work is the conclusive theoretical 
114: demonstration of a class of one-dimensional
115: optical lattice systems where the deterministic lattice potential
116: {\em does} allow for the existence of a mobility edge in one dimension
117: \cite{Azbel}, which cannot happen through Anderson localization with
118: disorder.  This class of models distinguishes itself from other models
119: discussed in the context of optical lattices \cite{Diener} 
120: through the formation of a metal-insulator mobility edge rather than a
121: metal-band edge.  We find that: 1) Direct numerical simulation and an  
122: analytic WKB approximation provide conclusive evidence for a 
123: rare metal-insulator mobility edge in a one-dimensional model, the nearly 
124: periodic Harper model. 2)  Transport measurements in suitably designed,  
125: one-dimensional optical lattices can exhibit the mobility edge. 
126: 
127: We consider spinless fermions (or equivalently hardcore bosons sufficiently 
128: near the Tonks-Girardeau regime) in the lowest band 
129: of a one-dimensional, tight binding 
130: lattice with external potentials:
131: \begin{eqnarray}
132: -u_{n+1}-u_{n-1}+(V_n+V_D f_n +\Omega n^2)u_n=Eu_n,
133: \label{H}
134: \end{eqnarray}
135: where the amplitudes $u_n$ multiply the Wannier states at sites $n$ in the
136: real space wavefunction $\Psi(x)=\sum_n u_n w(x-n)$.  
137: We work in units of the hopping matrix element, $t=1$, and 
138: lattice spacing of the primary lattice defining the tight 
139: binding problem, $a=1$, unless otherwise noted.  The statistics 
140: of spinless fermions
141: implicitly allow for an arbitrary on-site interaction in the above
142: single-band model.  In the absence
143: of an external potential the solutions form extended
144: states, $u_n=u_0\exp(in\phi)$, with band energies 
145: $E=2\cos(\phi)$, for $0\leq\phi\leq\pi$.  The band edges lie at
146: $E=\pm 2$.  In the presence of an 
147: oscillatory modulation of strength $V$, much weaker than the 
148: primary lattice, we can ignore
149: modifications to the hopping. In this limit we impose a secondary lattice:
150: $    
151: V_n=V\text{cos}(2\pi\alpha n)-V.
152: $
153:   For $\alpha$ irrational the additional potential establishes an
154:  incommensurate pseudorandom model, the 
155: Harper model (for $\Omega=0$ and $V_D=0$).  The potential $V_D f_n$
156: adds disorder where $f_n$ is a random number satisfying 
157: $0\leq f_n \leq 1$ for each site.  The confinement 
158: potential, $\Omega n^2$, applies to optical lattice systems.      
159: 
160: According to the Aubry-Andre
161: conjecture \cite{Aubry} the Harper model exhibits a 
162: metal-insulator transition at the self-dual point $V=2$.  For $V<2$
163: all states are extended while for $V>2$ all states localize (the
164: states at $V=2$ are critical).  The
165: localized states are characterized by a nonzero Lyapunov 
166: exponent (inverse localization length), $\gamma(E)$, where 
167: $u_n(E)\sim\exp(-\gamma n)$, and gaps in 
168: the energy spectra.  While exceptions to the 
169: Aubry-Andre conjecture have been rigorously proven for specific values of
170: $\alpha$ \cite{Avron}, we discuss here an additional and 
171: experimentally relevant counter example defined by:
172: $\alpha=m\pm\epsilon$, for integer $m$ and  
173: \begin{eqnarray}
174: N^{-1}\ll \epsilon \ll 1,
175: \label{lim}
176: \end{eqnarray}
177: with $N$ sites.  In the limit $N\rightarrow \infty$ the 
178: secondary lattice defines a
179: slowly varying, nearly-periodic potential.  A similar, slowly varying
180: potential has been considered in the context
181: of one-dimensional localization in quasiperiodic systems \cite{DasSarma}.   
182: 
183: In the limit defined by Eq.~(\ref{lim}) the eigenstates 
184: of Eq.~(\ref{H})  with $\Omega=0$ and $V_D=0$ display Anderson-like localization 
185: where we expect to find 
186: {\em only} extended states.  To see this consider $\gamma$ defined
187: in the limit, $N\rightarrow \infty$ \cite{Thouless}:
188: \begin{eqnarray}
189: \gamma(E_j)
190: =\frac{1}{N}\sum_n \ln \left\vert \frac{u_{n+1}}{u_{n}}\right\vert
191: =\frac{1}{(N-1)}\sum_{j\neq l}\ln\left\vert E_j-E_l\right\vert.
192: \label{defg}
193: \end{eqnarray}
194: The first equality allows us to use the transfer matrix method to 
195: calculate $\gamma$ for large system sizes.  The solid
196: line in the top panel of Fig.~\ref{gam} plots the Lyapunov exponent 
197: versus energy for $N=10^7$ and $V=0.5$.  
198: \begin{figure}
199: \includegraphics[clip,width=3.0in]{gam}
200: \caption{Top panel:  Lyapunov exponent versus energy for 
201: the nearly-periodic Harper model, Eq.~(\ref{H}) with $V=0.5$, 
202:   $\epsilon=0.005$, and $\Omega=0$.  The solid (dot-dashed) line
203:   shows the disorder-free, $V_D=0$ (disordered, $V_D=1$) case.
204:   Bottom panel:  Level number versus energy of the disorder-free,
205:   nearly-periodic Harper model for two characteristic values of
206:   $\epsilon$ with $V=0.5$, $N=3000$, and $\Omega=0$.  The lower curve 
207:   is shifted downward by 100 
208:   levels for clarity.   
209: \label{gam}}
210: \end{figure} 
211: The additional potential, $V_n$, shifts 
212: the lower band edge to $E=-2-2V$ while leaving the upper band edge at
213: $E=2$.  We see extended states in the center of 
214: the band, $-2<E<2-2V$, with $\gamma=0$, as 
215: expected from the Aubry-Andre conjecture.
216: However, near the band edges, $-2-2V<E< -2$ and $2-2V<E<2$, the 
217: states localize, $\gamma>0$.  
218: The points $E=-2$ and $2-2V$ define mobility edges which are  
219: unexpected in one dimension but found in three-dimensional models with
220: disorder.  The localization is, in this sense, Anderson-like.  
221: We find that, for $N=10^7$, the mobility edges persist for rational {\em and}
222: irrational values of $\epsilon$ from $10^{-5}$ to $10^{-2}$.  We
223: conjecture that in the limit Eq.~(\ref{lim}) irrational numbers are
224: approximated by rational numbers up to a number much smaller than
225: $N^{-1}$.  For $N\rightarrow \infty$, the spectra can contain an
226: infinite number of infinitely small gaps and therefore localized
227: states eliminating the distinction between an incommensurate and
228: commensurate system \cite{Avron,Sokoloff}. 
229: 
230: The unexpected insulating behavior coincides with a devil's staircase-like
231: structure in part of the energy spectrum \cite{Azbel,Sokoloff}.  The 
232: second equality in Eq.~(\ref{defg}) shows that a degeneracy at $E_j$ 
233: supports non-zero $\gamma(E_j)$.  The lower panel of 
234: Fig.~\ref{gam} plots the 
235: level number as a function of energy as determined by exact
236: diagonalization of Eq.~(\ref{H}) for $N=3000$, $V=0.5$, $V_D=0$, and 
237: $\Omega=0$.  The top line
238: ($\epsilon=0.005$) shows smaller gaps and narrower steps than the bottom line
239: ($\epsilon=0.05$) suggesting that the localized states develop a gapless      
240: insulator in the limits $\epsilon\rightarrow 0$ and $\epsilon \gg N^{-1}$.
241: 
242: We can understand the insulating states in a ``semi-classical''
243: approximation where $\epsilon$ plays the role of $\hbar$.  We analyze 
244: the behavior of each regime as a function of energy.  At low energies,
245: $E<-2$, the 
246: slowly varying potential confines low
247: energy states near the potential minima defined by $V_n$.  Very 
248: little tunneling between minima forces localization.  
249: Intermediate energies,$-2<E<2-2V$, see a smaller 
250: barrier between minima allowing for 
251: extended states and, therefore, the first mobility edge 
252: at $E=-2$.  A second mobility edge forms at   
253: $E=2-2V$ when states localize at the 
254: secondary lattice maxima.  
255: At first this seems counterintuitive but can be understood in a WKB
256: approximation based on the slowly varying nature of $V_n$.  A similar
257: analysis was performed for a different model in Ref.~\cite{DasSarma}.  Our results 
258: show that the high energy
259: states, $E>2-2V$, moving energetically above the lattice slow 
260: when passing secondary lattice maxima to force localization.  We have checked
261: that our analysis based on the WKB approximation reproduces the
262: solid line in the upper panel of Fig.~\ref{gam}.  As an additional
263: check we can, in a continuum approximation \cite{Sokoloff,Wimberger}, 
264: define a position variable, $n\rightarrow \tilde{x}$, and a 
265: difference operator, $u_{n+1}+u_{n-1}\rightarrow 2\text{cos}(p)u(\tilde{x})$
266: (with $p\equiv i\partial/\partial \tilde{x}$), to give the 
267: semi-classical Hamiltonian: $H_{\text{CL}}= -2\text{cos}(p)+V(\tilde{x})$,
268: with the replacement $V_n\rightarrow V(\tilde{x})$.  The phase trajectories of
269: $H_{\text{CL}}$ produce extended and localized states (and therefore
270: mobility edges) in the regimes obtained in Fig.~\ref{gam}. 
271: 
272: We now discuss the possibility of observing this unique type of
273: localization.  In the solid state a necessary correction to
274: the Harper model includes disorder where we add to $V_n$ a potential of the form:
275: $V_{\text D}f_n$.  For $V=0$ (and $\Omega=0$) this defines the one-dimensional Anderson
276: model where we expect all states to localize for arbitrary $V_D$.
277: However, for $V\neq 0$, the states (otherwise extended in the $V_{\text D}=0$
278: case) have a 
279: small localization length which could allow some 
280: remnant of a mobility edge. The
281: dot-dashed line in the upper panel of Fig.~\ref{gam} plots $\gamma$
282: for the same parameters as the solid line but 
283: with $V_{\text D}=1.0$.  We find that a finite amount of
284: disorder obscures the position of the remnant-mobility edges while
285: localizing all states.      
286: 
287: In what follows we consider an essentially disorder-free manifestation of
288: Eq.~(\ref{H}): one-dimensional, cold atom optical lattices.     
289: The interference of appropriately detuned lasers of wavelength
290: $\lambda=2a$ can give rise to our tight binding lattice with
291: a sufficiently strong lattice height $V_L$. To create a 
292: secondary modulating potential, $V_n$,
293: consider an additional pair of lasers at angles $\theta$ and
294: $\pi-\theta$ to the primary lattice with wavelength $\lambda '$ and 
295: amplitude $V_L '$.  
296: The additional lasers interfere
297: to modulate the energy of the $n$th site by: 
298: $V_L '\text{cos}(2\pi n\alpha )
299: \int_{-\infty}^{\infty}\vert w(u) \vert^2
300: \text{cos}(2\pi u\alpha )du-V_L '$,
301: where  $\alpha=(\lambda/\lambda ')\text{cos}(\theta)$.  For small
302: angles we can retrieve, up to an overall constant, our nearly-periodic
303: Harper model with $m=\lambda/\lambda '$ an integer and 
304: $\epsilon\approx-\lambda\theta^2/2\lambda '$.  For realistic parameters:
305: $V_L=5 E_R$, $V_L '=0.1E_R,\theta=\degrees{5}$, and 
306: $\lambda=\lambda'$ (where $E_R$ is the 
307: photon recoil energy), we find $t\approx 0.065 E_R$, $V\approx
308: 0.055 E_R$, and $\epsilon\approx 0.004$ yielding the appropriate parameter 
309: regime.  Furthermore, we find that, in the limit of Eq.~(\ref{lim}),   
310: fluctuations in the relative phase do not alter the position of the mobility
311: edge.  We now include an 
312: important modification to the model which accounts for realistic
313: finite size effects.
314: 
315: A crucial addition to the Harper model in optical lattices 
316: is the parabolic confinement: $\Omega n^2$, which leads to a finite 
317: particle number.  We find
318: that weak confinement leaves the mobility edges intact.  To see this consider
319: the local Lyapunov exponent:      
320: $
321: \gamma^L(E_j)=(2N_{\text{CL}}+1)^{-1}
322: \sum_{n= -N_{\text{CL}}}^{N_{\text{CL}}} 
323: \ln\vert u_{n+1}/u_n\vert,
324: $
325: where the semiclassical limits of the parabolic
326: trap define the number of states participating in 
327: transport, $2N_{\text{CL}}+1$.  The classical turning points give 
328: $N_{\text{CL}}=2\vert x_{\text{CL}}(E) \vert$ and Eq.~(\ref{lim}) 
329: becomes: $(2N_{\text{CL}})^{-1}\ll \epsilon \ll 1$.  To determine 
330: $x_{\text{CL}}$ we set $p=0$ in $H_{\text{CL}}$ with
331: $V(\tilde{x})=V\text{cos}(2\pi \alpha \tilde{x} )-V+\Omega \tilde{x}^2$.  
332: For $\Omega\sim 10^{-5}$ we find $2N_{\text{CL}}\sim 10^3$.  In the
333: limit $\Omega\rightarrow 0$ we retrieve the usual Lyapunov exponent, 
334: $\gamma^L \rightarrow \gamma$.
335: Fig.~\ref{parab} plots the local Lyapunov exponent as a function of
336: energy for $N=10^7$, $\epsilon=0.005$, $V=0.5$, $V_D=0$, and 
337: $\Omega =10^{-5}$.  The 
338: mobility edges remain even with a reduced number of states comprising 
339: the system.    
340: \begin{figure}
341: \includegraphics[clip,width=3.0in]{parab}
342: \caption{The local Lyapunov exponent versus energy for the same
343:   parameters as the upper panel of Fig.~\ref{gam} but with an additional
344:   parabolic confinement $\Omega=10^{-5}$ and no disorder, $V_D=0$.
345:   The inset shows the
346:   normalized density of the same system as a function of 
347:   lattice number for several different chemical potentials.  
348: \label{parab}}
349: \end{figure}
350: The inset shows the normalized density profile as a function of site
351: number for three different chemical potentials, $\mu$.  At zero
352: temperature we include states with $E\leq\mu$.  For $\mu=0.5$
353: (dashed-dotted line) we find extended states with some modulation due
354: to $V_n$.  For
355: $\mu=1.5$ (dashed line) we have crossed the mobility edge and the
356: density pins to unity at some lattice sites.  Here the formation
357: of a mesoscopic version of the Anderson-like gapless insulator 
358: fixes the density.  For $\mu=3.0$ we enter the band
359: insulator regime which fixes a large fraction of the 
360: states at integer density.
361: 
362: Dipole oscillations in harmonically confined atomic gases serve as a 
363: direct probe of localization \cite{Pezze,Huckans}.  A small shift in 
364: the center of mass results in harmonic oscillations 
365: in the absence of an external lattice.
366: The presence of one or more weak lattices allows for weakly localized states 
367: which can suppress oscillations and lead to an effective under-damping of
368: the center of mass motion.  The addition of strongly localized states   
369: can, in the absence of dissipation, eventually pin the center of mass 
370: to effectively over-damp the center of mass oscillations.  Strong experimental
371: and theoretical evidence supports the possibility that band
372: localization has indeed been observed in fermionic, one-dimensional
373: optical lattices \cite{Pezze}.  Similar evidence also suggests such behavior
374: for strongly interacting bosons \cite{Huckans}.    
375: 
376: We now study the onset of
377: the gapless Anderson-like insulator and its effect on center 
378: of mass oscillations.
379: Consider the center of mass to be displaced $\Delta$ lattice sites 
380: at some initial time $T=0$.  For extended states, the center of mass 
381: position, $\bar{X}(T)$, averages to zero for long times  
382: while localized states should pin the center of mass position,  
383: $\bar{X}\sim \Delta$.  The center of mass position can, for some
384: parameters, demonstrate complex, damping-like 
385: behavior as function of
386: time making a damping constant ill-defined.  To extract a simple 
387: quantity to be compared with experiment 
388: we calculate the long time average of the
389: center of mass position, $<\bar{X}>_{\infty}$, as a function of chemical
390: potential by diagonalizing Eq.~(\ref{H}) with a parabolic
391: potential, $\Omega=10^{-5}$, for $N=3000$, $\Delta= -3$,
392: and $V_D=0$.  As an intermediate step we require degenerate 
393: eigenstates (localized
394: at the edges) to simultaneously diagonalize the parity operator since our 
395: system possess reflection symmetry about the origin.  
396: The dashed line in 
397: Fig.~\ref{avx} plots $<\bar{X}>_{\infty}$ as a function of chemical
398: potential in the absence of a secondary lattice, $V=0$.  For $\mu<2$ the 
399: \begin{figure}
400: \includegraphics[clip,width=3.0in]{avx}
401: \caption{  The long time average of the center of mass position 
402:   as a function of chemical potential after an initial 
403:   displacement of three sites, $\Delta=-3$, with no disorder, $V_D=0$.  The dashed 
404:   line is calculated from the bare tight
405:   binding model with no secondary lattice, $V=0$, 
406:   and $\Omega=10^{-5}$.  B.E labels the
407:   upper band edge.  The solid line is calculated for the same
408:   parameters but with $V=0.5$.  The early onset of the gapless insulator occurs at
409:   the upper mobility edge, labeled M.E., while the upper band 
410:   edge remains at $\mu=2$.
411: \label{avx}}
412: \end{figure}
413: extended states perform several oscillations about the trap center but
414: over long times average to zero displacement.     
415: Above the band edge (labeled B.E.), for $\mu>2$, localized states 
416: near the edge pin the center of mass near $\Delta$.  For $\mu \gtrsim 3$
417: the system never leaves its initial position.  
418: 
419: A second weaker
420: lattice causes a mobility edge to form 
421: energetically below the band edge.  The solid
422: line in Fig.~\ref{avx} plots the same as the dashed line but with a
423: second lattice, $V_n$, with $V=0.5$ for chemical potentials near the upper
424: mobility edge (labeled M.E.).  $<\bar{X}>_{\infty}$ remains 
425: zero where we expect
426: extended states but pins near $\Delta$ for $\mu>2-2V$.  The
427: mesoscopic version of the gapless insulator results in the early onset 
428: of pinning in the regime $2-2V<\mu<2$ and Eq.~\ref{lim}.  Furthermore,
429: the localized
430: states with the additional lattice, $V=0.5$, also display weak
431: periodicity in $<\bar{X}>_{\infty}$ as a function of $\mu$.  These 
432: oscillations correspond to the
433: chemical potential passing through peaks and valleys in the
434: corrugated confinement potential. 
435: 
436: Fluctuations in the lattice depth can soften the otherwise sharp
437: mobility edge.  The quantity of
438: interest, $2V/t$, can fluctuate wildly with only moderate changes in $V_L$ 
439: at extremely large lattice depths.  To see this consider 
440: an approximate expression in terms of the hopping 
441: extracted from an analysis of the related Mathieu problem:
442: $V/t\approx (\sqrt{\pi}V/4)(V_L/E_R)^{-3/4}\exp(2\sqrt{V_L/E_R})$.  A relative 
443: error in $V$ and $V_L$, $R_V$ and $R_{V_L}$ respectively,
444: propagates to a relative error in $2V/t$: 
445: $[R_V^2+R_{V_L}^2(3/4-\sqrt{V_L/E_R})^2]^{1/2}$.  
446: We have checked that this formula is
447: quantitatively accurate for $V_L\gtrsim 5 E_R$ by comparing with error
448: derived numerically from the exact tunnelling.  We find that for
449: $R_V=R_{V_L}=5 \%$ the relative error in $2V/t$ 
450: remains below $20 \%$ for $V_L<20E_R$.  
451:   
452: We note that additional time dependence in the model discussed 
453: here possesses other applications.  We take $H_{\text{CL}}$ as a good 
454: approximation to the nearly-periodic Harper model in the limit Eq.~(\ref{lim}).   
455: In the presence of a pulsed secondary lattice:   
456: $V\propto \sum_{j}\delta\left(T-jT_0\right)$, where for integer $j$
457: the secondary lattice oscillates with period $T_0$, we simulate 
458: the kicked Harper model via $H_{\text{CL}}$.  The
459: kicked Harper model exhibits chaotic behavior with the ``classical''
460: to quantum crossover controlled by $\epsilon$.
461:        
462: We have explicitly demonstrated the existence of a mobility edge
463: (and the associated, unusual metal-insulator transition in a
464: deterministic disorder-free environment) in suitably designed
465: aperiodic cold atom optical lattice systems.  The deterministic 
466: aperiodic background potential in these optical lattices leads to
467: exotic and nontrivial energy eigenstates 
468: dependent on the relationship between irrational numbers and their 
469: rational approximations.  The ensuing quantum localization occurs in
470: the absence of disorder and therefore 
471: distinguishes itself from Anderson localization which, in the solid
472: state, masks the presence of mobility edges formed from quasiperiodic
473: potentials in one dimension.
474: 
475: We thank K. Park and G. Pupillo for valuable
476: discussions.  This work is supported by NSA-LPS and ARO-ARDA.
477: 
478: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
479: 
480: \begin{thebibliography}{}
481: 
482: \bibitem{Greiner} M. Greiner {\em et al.},
483: Nature \textbf{415}, 39 (2002).
484: 
485: \bibitem{Paredes} B. Peredes {\em et al.}, 
486: Nature \textbf{429}, 277 (2004).
487: 
488: \bibitem{cross}
489: C. A. Regal {\em et al.}, 
490: Phys. Rev. Lett. \textbf{92}, 040403 (2004);
491: M. Bartenstein {\em et al.}, 
492: Phys. Rev. Lett. \textbf{92}, 120401 (2004);
493: M. Zwierlein {\em et al.}, 
494: Phys. Rev. Lett. \textbf{92}, 120403 (2004);
495: T. Bourdel {\em et al.},
496: Phys. Rev. Lett. \textbf{93}, 050401 (2004).
497: 
498: \bibitem{Oskay} W.H. Oskay {\em et al.},
499: Opts. Commun. \textbf{179}, 137 (2000).
500: 
501: \bibitem{dArcy} M.B. d'Arcy {\em et al.},
502: Phys. Rev. E \textbf{64}, 056233 (2001).
503: 
504: \bibitem{Sadgrove} M. Sadgrove {\em et al.},
505: Phys. Rev. Lett. \textbf{94}, 174103 (2005).
506: 
507: \bibitem{Pezze} L. Pezze {\em et al.}, 
508: Phys. Rev. Lett. \textbf{93}, 120401 (2004).
509: 
510: \bibitem{Huckans} C. D. Fertig {\em et al.},
511: Phys. Rev. Lett. \textbf{94}, 120403 (2005).
512: 
513: \bibitem{Drese} K. Drese and M. Holthaus,
514: Phys. Rev. Lett. \textbf{78}, 2932 (1997).
515: 
516: \bibitem{Guidoni} L. Guidoni {\em et al.},
517: Phys. Rev. Lett. \textbf{79}, 3363 (1997).
518: 
519: \bibitem{Anderson} P.W. Anderson, 
520: Phys. Rev. \textbf{109}, 1492 (1958).
521: 
522: \bibitem{Sokoloff} J.B. Sokoloff, 
523: Phys. Rep. \textbf{126}, 189 (1985).
524: 
525: \bibitem{Last}
526: Y. Last, 
527: {\em Proc. of XIth Int. Congress of Math. Phys.}
528: (Paris, 1994) p. 366;  
529: S. Jitomirskaya, {\em ibid.} p. 373.
530: 
531: \bibitem{Azbel} M. Ya. Azbel, 
532: Phys. Rev. Lett. \textbf{43}, 1954 (1979).
533: 
534: \bibitem{Diener} R. Diener {\em et al.},
535: Phys. Rev. A \textbf{64}, 033416 (2001).
536: 
537: \bibitem{Aubry} S. Aubry and G. Andre,
538: Ann. Israel Phys. Soc. \textbf{3}, 133 (1980).
539: 
540: \bibitem{Avron} J. Avron and B. Simon,
541: Bull. Am. Math. Soc. \textbf{6}, 81 (1982).
542: 
543: \bibitem{DasSarma} S. Das Sarma {\em et al.}, 
544: Phys. Rev. Lett. \textbf{61}, 2144 (1988);  
545: {\em ibid.} Phys. Rev. B \textbf{41}, 5544 (1990).
546: 
547: \bibitem{Thouless} D.J. Thouless, 
548: J. Phys. C \textbf{5}, 77 (1972).
549: 
550: \bibitem{Wimberger} S. Wimberger {\em et al.}, 
551: Phys. Rev. Lett. \textbf{92}, 084102 (2004).
552: 
553: \end{thebibliography}
554: 
555: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
556: 
557: \end{document}
558:  
559: 
560: 
561: 
562: