1: %%%%%%%%%%%%%%%%%%%%%%% file template.tex %%%%%%%%%%%%%%%%%%%%%%%%%
2: %
3: % This is a template file for the global option of the SVJour class
4: %
5: % Copy it to a new file with a new name and use it as the basis
6: % for your article
7: %
8: %%%%%%%%%%%%%%%%%%%%%%%% Springer-Verlag %%%%%%%%%%%%%%%%%%%%%%%%%%
9: %
10: % Choose either the first of the next two \documentclass lines for one
11: % column journals or the second for two column journals.
12: \documentclass[global,twocolumn,final]{svjour}
13: %\documentclass[global,twocolumn,referee]{svjour}
14: % Remove option referee for final version
15: %
16: % Remove any % below to load the required packages
17: \usepackage{latexsym}
18: \usepackage{revsymb}
19: \usepackage{amssymb}
20: \usepackage{graphics}
21: % etc
22: %
23: % Insert the name of "your" journal with the command below:
24: \journalname{Applied Physics B}
25: %
26: \begin{document}
27: %
28: \title{Nanoscale atomic waveguides with suspended carbon nanotubes}
29: \author{V. Peano \inst{1} \and M. Thorwart\inst{1} \and A.\ Kasper\inst{2}
30: \and R. Egger\inst{1}}
31: %
32: %\offprints{V. Peano} % Insert a name or remove this line
33: %
34: \institute{Institut f\"ur Theoretische Physik,
35: Heinrich-Heine-Universit\"at D\"usseldorf, D-40225 D\"usseldorf,
36: Germany \and NTT Basic Research Laboratories, NTT Corporation, Kanagawa 243-0198,
37: Japan}
38: %
39: \date{Received: / Revised version: }
40: % The correct dates will be entered by the editor
41: %
42: \maketitle
43: %
44: \begin{abstract}
45: We propose an experimentally viable setup for the realization of
46: one-dimensional ultracold atom ga\-ses in a nanoscale magnetic
47: waveguide formed by single dou\-bly-clamped suspended carbon
48: nano\-tubes. We show that all common decoherence and atom loss
49: mechanisms are small guaranteeing a stable operation of the trap.
50: Since the extremely large current densities in carbon nanotubes are
51: spatially homogeneous,
52: our proposed architecture allows to
53: create a very regular trapping potential for
54: the atom cloud.
55: Adding a second nanowire allows to create a double-well
56: potential with a moderate tunneling barrier which is desired for
57: tunneling and interference experiments with the advantage of
58: tunneling distances being in the nanometer regime.
59: \end{abstract}
60:
61:
62: %
63: \section{Introduction}
64: \label{intro} The ongoing progress in the fabrication and
65: manipulation of micro- or nanoscale structures has recently allowed
66: for systematic studies of ultracold atom gases, where
67: current-carrying wires and additional magnetic bias fields
68: generate magnetic fields trapping neutral
69: atoms (`atom chips') \cite{Folman02,Reichel02}.
70: For instance, the Bose-Einstein condensation (BEC) of
71: microchip-con\-fined atoms has been successfully demonstrated by
72: several groups \cite{atomBEC}.
73: So far, deco\-her\-ence and atom loss constitute central
74: im\-pe\-di\-ments, since atoms are relatively close to `hot'
75: macroscopic surfaces or current-carrying wires (with typical
76: diameters of several $\mu$m), where the Casimir-Polder potential and
77: Johnson noise can seriously affect stability
78: \cite{Henkel99,chin,Schroll03}. To reduce these effects, further
79: miniaturization to the nanoscale regime would be desirable. In
80: particular, this is
81: promising in the context of integrated atomic matter-wave
82: interferometry and optics \cite{Kasevich}, and combines the
83: strengths of nanotechnology and atomic physics.
84: While at first sight this goal conflicts with the requirement of
85: large currents forming tight trapping potentials,
86: we propose that when
87: using suspended carbon nanotubes (NTs) \cite{tubes} (with diameters
88: of a few nm) as wires, nanoscale atom chip devices with large current
89: densities can be designed. In turn, these devices allow to trap
90: ultracold atom gases basically free
91: of trap-induced decoherence or atom losses, with the gas containing
92: few tens of atoms. Since disorder is generally weak in NTs, the (extremely
93: large) current density distribution is spatially homogeneous, which allows
94: to overcome the problem of fragmentation of the atom cloud.
95: Moreover, they can be built with state-of-the-art
96: technology.
97:
98: With relevant length scales below optical and cold-atom de Broglie
99: wavelengths, this also paves the way for the observation of
100: interesting and largely unexplored many-body physics in one
101: dimension (1D) \cite{Petrov04}. Examples include the interference
102: properties of interacting matter waves \cite{chen}, the 1D analogue
103: of the BEC-BCS crossover \cite{becbcs} and shape resonances in 1D
104: trapping potentials \cite{olshanii}. Previous realizations of 1D
105: cold atoms were reported using optical lattices
106: \cite{esslinger,Paredes04,Weiss} and magnetic traps
107: \cite{Goerlitz01}, but they involve arrays of 1D or elongated 3D
108: systems, where it is difficult to separately
109: manipulate a single 1D atom cloud (the distances between the 1D
110: systems composing the array
111: are few hundred nm). A noteworthy advantage of our proposal against
112: dipole optical traps is that
113: arrays of many NT waveguides can be built, where it is possible to
114: manipulate an individual trap by changing the current through an
115: individual NT. Moreover, our proposal allows to
116: minimize unwanted substrate effects and implies a
117: drastically reduced transverse size (a few nm) of the cloud.
118: We expect that our approach allows to observe new interesting many-body
119: features not accessible by
120: the otherwise very successful atom chip setup. This could provide a
121: fruitful link between atomic and condensed matter physics
122: with a wealth of fascinating effects.
123:
124: \section{The setup}
125:
126: A typical proposed nanoscale waveguide setup to confine ultracold
127: atoms to 1D is sketched in Fig.~\ref{fig.1}. The setup employs a
128: single suspended doubly-clamped NT (left NT in Fig.~\ref{fig.1}, the
129: second suspended NT on the right will be used to create a
130: double-well potential, see below), where nanofabrication techniques
131: routinely allow for trenches with typical depths and lengths of
132: several $\mu$m \cite{tubes}. To minimize decoherence and loss
133: effects \cite{chin}, the substrate should be insulating apart from
134: thin metal strips to electrically contact the NTs. Since strong
135: currents (hundreds of $\mu$A) are necessary, thick
136: multiwall nanotubes (MWNTs) or `ropes' \cite{tubes} are best suited.
137: The suspended geometry largely eliminates the influence of the
138: substrate. A transverse magnetic field $B_x$ is required to create a
139: stable trap while a longitudinal magnetic field $B_z$
140: suppresses Majorana spin flips \cite{sukumar,jones}.
141: With this single-tube setup, neutral atoms in a weak-field seeking
142: state can be trapped. Studying various sources for decoherence,
143: heating or atom loss, and estimating the related time scales, we
144: find that, for reasonable parameters, detrimental effects are small.
145: As a concrete example, we shall consider $^{87}$Rb atoms in the
146: weak-field seeking hyperfine state $|F,m_F\rangle=|2,2\rangle$.
147:
148: We next describe the setup in Fig.~\ref{fig.1}, where the
149: (homogeneous) current $I$ flows through the left NT positioned at
150: $(-x_0,0,z)$. With regard to the decoherence properties of the proposed trap,
151: it is advantageous that the current flows
152: homogeneously through the NT, as disorder effects are usually weak
153: in NTs \cite{tubes}.
154: Neglecting boundary effects due to the finite tube length $L$, the
155: magnetic field at ${\bf x}=(x,y,z)=({\bf x}_\perp,z)$ is given by
156: \begin{equation}
157: \mathbf{B}({\bf x})=\frac{\mu_0 I}{2\pi}
158: \frac{1}{(x+x_0)^2+y^2} \left( \begin{array}{c} -y \\ x+x_0 \\ 0\\
159: \end{array}\right) +\left( \begin{array}{c} B_x\\0\\B_z\\\end{array}\right)
160: \end{equation}
161: with the vacuum permeability $\mu_0$. To create a trapping potential
162: minimum at $y_0$, the transverse magnetic field is
163: $B_x=\mu_0 I / (2\pi y_0)$. Then
164: the transverse confinement potential is
165: $V({\bf x}_\perp)=\mu |{\bf B({\bf x})}|$, where $\mu=m_F g_F \mu_B$
166: with the Land{\'e} factor $g_F$ and the Bohr magneton $\mu_B$.
167: It has a minimum along the line $(-x_0,y_0,z)$, with the distance between
168: the atom cloud and the wire being $y_0$.
169: Under the adiabatic approximation \cite{sukumar},
170: $m_F$ is a constant of motion, and the
171: potential is harmonic very close to the minimum of the trap, i.e.,
172: $V(\mathbf{x})\simeq \mu B_z+\frac{1}{2}m\omega^2
173: [(x+x_0)^2+(y-y_0)^2]$, with
174: frequency $\omega= [\mu/(m B_z)]^{1/2} \mu_0 I / (2\pi y_0^2)$
175: and associated transverse confinement length $l_0=(\hbar/m\omega)^{1/2}
176: \ll y_0$, where $m$ is the atom mass.
177: The adiabatic approximation is valid as long as $\omega\ll \omega_L$
178: with the Larmor frequency $\omega_L= \mu B_z/\hbar$.
179: Non-adiabatic
180: Majorana spin flips to a strong-field seeking state generate atom loss
181: \cite{Folman02,jones} characterized by the rate $\Gamma_{\rm loss} \simeq
182: (\pi\omega/2)\exp(1-1/\chi)$, with $\chi=\hbar\omega/(\mu B_z)$
183: \cite{sukumar}. For convenience, we switch to a dimensionless form
184: of the full potential $V({\bf x}_\perp)$ by measuring energies in
185: units of $\hbar \omega$ and lengths in units of $l_0$,
186: \begin{equation} \label{fullpot}
187: \chi V= \left( 1+\chi \frac { d^2 [ (x+x_0)^2-dy+y^2]^2
188: + d^4(x+x_0)^2 }
189: { [(x+x_0)^2+y^2]^2 } \right)^{1/2},
190: \end{equation}
191: which depends only on $d= y_0/l_0$ and $\chi$. The trap
192: frequency then follows as
193: \begin{equation}
194: \omega=\frac{m\chi\mu^2}{\hbar^3}\Big(\frac{\mu_0I}{2\pi d^2}\Big)^2
195: \, .
196: \end{equation}
197: %
198: Note that a real trap also requires a longitudinal
199: confining potential with frequency $\omega_z \ll \omega$.
200:
201: To obtain an estimate for the design of the nanotrap, we
202: choose realistic parameters: $\chi=0.067$, corresponding to a rate of
203: spin flip transition per oscillation period
204: $\Gamma_{\rm loss}/\omega\sim10^{-6}$.
205: Decreasing $d$ increases the trap frequency. However, $d$ cannot be
206: chosen too small, for otherwise the potential is not confining
207: anymore (and the harmonic approximation becomes invalid).
208: Using $V(\infty)=\chi^{-1}(1+\chi
209: d^2)^{1/2}$ for the potential at $|{\bf x}_\perp|\to \infty$, we now show
210: that for $d\alt 5$, the harmonic approximation
211: breaks down. To see this, note that for $d=10$,
212: the potential provides a confining barrier
213: (in units of the trap frequency $\omega$) of
214: $V(\infty)-V(0,0,z)=23.8$, while for $d=5$, we get only
215: $V(\infty)-V(0,0,z)=9.8$. Thus exceedingly
216: small values of $d$ would lead to unwanted
217: thermal atom escape processes out of the trap.
218: To illustrate the feasibility of the proposed trap design, we show
219: in Table \ref{tab.freq} several parameter combinations with
220: realistic values for the MWNT current together with the resulting
221: trap parameters. In practice, first the maximum possible current
222: should be applied to the NT, with some initial field $B_x$. After
223: loading of the trap, the field $B_x$ should be increased, the cloud
224: thereby approaching the wire with a steepening of the confinement.
225: At the same time, $y_0$ and consequently $d$ decrease.
226: This procedure can be used to load the nanotrap from a larger magnetic
227: trap (ensuring mode matching).
228: For a given current, there is a corresponding lower limit
229: $y_{\rm min}$ for stable values of $y_0$ from the
230: requirement $d\agt 5$, as already mentioned above.
231: To give an example, the confining potential is shown in
232: Fig.~\ref{fig.2}a) for $I=100 \mu$A, representing a reasonable
233: current through thick NTs \cite{tubes}, $d=10$, $x_0= l_0$ and
234: $\Gamma_{\rm loss}/\omega = 10^{-6}$ (where $\chi=0.067$). The
235: resulting trap frequency is $\omega= 2 \pi \times 4.6$ kHz and the associated
236: transverse magnetic field is $B_x=0.14$ G.
237:
238: \section{Influence of destructive effects}
239:
240: For stable operation, it is essential that destructive effects like
241: atom loss, heating or decoherence are small.
242:
243: (i) One loss process
244: is generated by non-adiabatic Majorana spin flips as discussed
245: above.
246:
247: (ii) Atom loss may also originate from noise-induced spin flips,
248: where current fluctuations cause a fluctuating magnetic field
249: generating the Majorana spin flip rate \cite{Henkel99}
250: \begin{equation}
251: \gamma_{\rm sf} \simeq \left(\frac{\mu_0\mu}{2\pi\hbar
252: y_0}\right)^2 \frac{S_I(\omega_L)}{2}, \
253: S_I(\omega)=\int dt e^{-i\omega t}\langle I(t)I(0)\rangle.
254: \end{equation}
255: %
256: At room temperature and for typical voltages $V_0\approx 1$~V, we have
257: $\hbar\omega_L\ll k_BT\ll eV_0$, and $S_I(\omega_L)$ is
258: expected to equal the shot noise $2eI/3$ of a diffusive wire.
259: For the parameters above, a rather small escape rate results,
260: $\gamma_{\rm sf}\approx 0.051$~Hz. If a (proximity-induced)
261: supercurrent is applied to the MWNT, the resulting current fluctuations
262: could be reduced even further.
263:
264: (iii) Thermal NT vibrations might create decoherence and heating,
265: and could even cause a transition to the first excited state of the
266: trap. Using a standard elasticity model for a doubly clamped wire
267: in the limit of small deflections,
268: the maximum mean square displacement is \cite{Sapmaz03}
269: \[
270: \sigma^2=\langle \phi^2(L/2)\rangle=
271: k_B T L^3/(192 Y M_I),
272: \]
273: where $\phi(z)$ is the NT displacement, $L$ the (suspended)
274: NT length, $T$ the temperature, $Y$ the Young modulus, and $M_I$
275: the NT's moment of inertia. For $L=10 \mu$m and typical
276: material parameters from Ref.~\cite{tubes}, we find $\sigma\approx
277: 0.2$~nm at room temperature. This is much smaller than the
278: transverse size $l_0$ of the atomic cloud. Small
279: fluctuations of the trap center could cause transitions to excited
280: transverse trap states. Detailed analysis shows
281: that the related decoherence rate is also negligible, since the
282: transverse fundamental vibration mode of the NT has the frequency
283: %
284: \begin{equation}
285: \omega_{\rm f} =\frac{\beta_1^2}{L^2}\sqrt{\frac{Y M_I}{\rho_L A_c}} \, ,
286: \end{equation}
287: with $\beta_1\simeq 4.73$, the mass density $\rho_L$, and the
288: cross-sectional area $A_c$. For the above parameters,
289: $\omega_{\rm f} = 2 \pi \times 11.9$ MHz is much larger than the trap
290: frequency itself. Due to the strong frequency mismatch, the coupling
291: of the atom gas to the NT vibrations is therefore negligible.
292:
293: (iv) Another decoherence mechanism comes from current fluctuations
294: in the NTs. Following the analysis of Ref.~\cite{Schroll03}, the
295: corresponding decoherence rate is
296: \begin{equation}
297: \frac{\gamma_c}{\omega} = \frac{3 \pi}{4\hbar} k_B T \frac{\sigma_0
298: A}{y_0^3}
299: \left(\frac{\mu_0 \mu_B}{2 \pi}\right)^2 \frac{\chi}{\hbar \omega} \,
300: ,
301: \end{equation}
302: where $\sigma_0$ is the NT conductivity and $A$ the cross-sectional
303: area through which the current runs in the NT.
304: For the corresponding parameters we find $\gamma_{\rm c}/\omega < 10^{-8}$.
305:
306: (v) Another potential source of atom loss could be the attractive
307: Casimir-Polder force between the atoms and the NT surface.
308: The Casimir-Polder interaction potential
309: between an infinite plane and a neutral atom
310: is given by $V_{\rm CP}=-C_4/r^4$ \cite{chin,Casimir48}.
311: For a metallic surface and $^{87}$Rb atoms,
312: $C_4=1.8\times10^{-55}$ Jm$^4$, implying that at a distance of
313: $1\mu$m from the surface, the characteristic frequency associated
314: with the Casimir-Polder interaction is $V_{\rm CP}/\hbar= 2 \pi
315: \times 0.29$ kHz.
316: In our setup, however, we cannot apply this estimate since the assumption of an
317: infinite plane is not realistic for a NT with a diameter of a few tens of
318: nm. Instead,
319: we expect that the distance between the cloud and the NT can be reduced without
320: drastically increasing the Casimir-Polder force.
321: The proposed setup could be an interesting
322: playground to study the Casimir-Polder interaction for our more
323: complicated geometry.
324:
325: (vi) A further possible mechanism modifying the shape of the
326: confining potential is the influence of the electric field
327: between the two contacts of the nanowire and the macroscopic leads
328: which is created by the transport voltage. This field depends
329: strongly on the detailed geometry of the contacts. However, the
330: electric field can in general be
331: reduced if the total length $L_{\rm tot}$
332: of the NT is increased. (Note that $L_{\rm tot}$
333: can be different from the length $L$ over which the NT is
334: suspended). Due to the small intrinsic NT resistivity, the
335: influence of the contact resistance then decreases for longer NTs.
336: Finally, we mention that superconducting leads could be used
337: to reduce the voltage drop.
338:
339:
340: \section{Number of trapped atoms and size of atom cloud}
341:
342: Next we address the important issue of how many atoms
343: can be loaded into such a nanotrap.
344: This question strongly depends on
345: the underlying many-body physics which determines for instance the
346: density profile of the atom cloud. Since the trap frequencies given in Table
347: \ref{tab.freq} exceed typical thermal energies of the cloud,
348: we will consider the 1D situation. Within the framework of
349: two-particle s-wave scattering in a parabolic trap,
350: the effective 1D interaction strength $g_{\rm 1D}=-2\hbar^2/ (ma_{\rm
351: 1D})$ is related to the
352: 3D scattering length $a$ according to \cite{olshanii}
353: %
354: \begin{equation}
355: a_{\rm 1D}=-\frac{l_0^2}{a} \left(1-{\cal
356: C}\frac{a}{\sqrt{2}l_0}\right) \, ,
357: \end{equation}
358: %
359: where ${\cal C}\simeq 1.4603.$ Interestingly,
360: $g_{\rm 1D}$ shows a confinement-induced resonance (CIR) for
361: $a=\sqrt{2} l_0 / {\cal C}$ \cite{olshanii}.
362: For nearly parabolic traps respecting parity symmetry,
363: this CIR is split into three resonances \cite{Peano}.
364: However, for the typical trap frequencies displayed in
365: Table \ref{tab.freq}, corresponding to non-resonant
366: atom-atom scattering, the parabolic confinement
367: represents a very good approximation. For free bosons in 1D,
368: the full many-body problem can be solved analytically \cite{Lieb}. It
369: turns out that the governing parameter is given by
370: $n|a_{\rm 1D}|$, where $n$ is the atom density in the
371: cloud. For weak interactions (large $n|a_{\rm 1D}|$),
372: a Thomas-Fermi (TF) gas results, while in the opposite regime, the
373: Tonks-Girardeau (TG) gas is obtained.
374:
375: For realistic traps with an
376: additional longitudinal confining potential with frequency
377: $\omega_z \ll \omega$, the problem has
378: been addressed in Ref.\ \cite{Dunjko}.
379: The corresponding governing parameter is $\eta=n_{\rm TF}|a_{1D}|$
380: where $n_{\rm TF}=[(9/64)N^2(m\omega_z\hbar)^2|a_{1D}|]^{1/3}$ is the cloud
381: density in the center of the trap in the TF approximation.
382: Small $\eta$ characterizes
383: a TG gas whereas large $\eta$ corresponds to the TF gas.
384: The longitudinal size $\ell$ of the atom cloud in terms of the atom
385: number $N$ and the longitudinal (transversal) trap frequencies
386: $\omega_z$ ($\omega$) has been computed in Ref.\
387: \cite{Dunjko}, with the result
388: \begin{equation}
389: \ell = \left[ \frac{3N (\hbar/m\omega_z)^2}{|a_{\rm 1D}|}
390: \right]^{1/3}
391: \label{length1}
392: \end{equation}
393: in the TF regime and
394: \begin{equation}
395: \ell = \left[
396: 2N (\hbar/m\omega_z)
397: \right]^{1/2}
398: \label{length2}
399: \end{equation}
400: in the TG regime.
401: In order to determine the cloud size $\ell$, we first calculate
402: $\eta$ for fixed $N, \omega_z$ and $\omega$, and then use the
403: respective formula Eq.\ (\ref{length1}) or
404: (\ref{length2}). In the crossover region, both expressions yield
405: similar results that also match the full numerical solution
406: \cite{Dunjko}.
407: Typical results for realistic parameters are listed in Table
408: \ref{tab.number} for $\omega_z=2\pi \times 0.1$ kHz.
409: {}From these results, we conclude that the length of the suspended
410: NT should be in the $\mu$m-regime in order to trap a few tens of
411: $^{87}$Rb atoms.
412:
413: To summarize the discussion of the monostable trap,
414: we emphasize that
415: the proposed nanotrap is realistic, with currents of a few 100 $\mu$A
416: and lengths of few $\mu$m of the suspended parts of NT.
417: No serious fundamental decoherence, heating or loss mechanisms are expected for
418: reasonable parameters of this nanotrap. We note that we did not consider
419: additional specific noise sources from further experimental equipment.
420:
421: \section{Double-well potential with two carbon nanotubes}
422:
423: In order to illustrate the advantages of the miniaturization to the
424: nanoscale, let us consider a setup which allows two stable minima
425: separated by a tunneling barrier. The simplest setup consists of two
426: parallel NTs carrying co-propagating currents $I$, a (small)
427: longitudinal bias field $B_z$ and a transverse bias field $B_x$.
428: Such a double-well potential for 1D ultracold atom gases would
429: permit a rich variety of possible applications. Experiments to study
430: Macroscopic Quantum Tunneling and Macroscopic Quantum
431: Coherence phenomena \cite{weiss}
432: between strongly correlated 1D quantum
433: gases could then be performed.
434: In addition, qubits forming the building blocks for a quantum
435: information processor could be realized.
436: The rich tunability of the potential shape, including
437: tuning the height of the potential barrier as well as the tunneling
438: distance, is a particularly promising feature.
439:
440: To realize this potential, we propose to place a second
441: current-carrying NT at $(+x_0,0,z)$, where the condition
442: $x_0>y_0$ guarantees the existence of two minima
443: located at $y_0(\pm \sqrt{x_0^2/y_0^2-1},1)$. By tuning the
444: transversal magnetic field $B_x$ and the current $I$, $y_0$ and thus
445: the location of the minima can be modified.
446: Around these minima, the
447: potential is parabolic with frequency
448: \begin{equation}
449: \omega=\left[
450: \frac{\mu^2\chi}{m \hbar} \left(\frac{\mu_0I}{2\pi}\right)^2
451: \frac{1}{y_0^2}\left( \frac{1}{y_0^2}-\frac{1}{x_0^2}\right)
452: \right]^{1/3} \, .
453: \end{equation}
454: Similar to the considerations above, we obtain the potential in units of
455: $\hbar \omega$, which depends only on
456: $d_x=x_0/l_0$, $d_y=y_0/l_0$ and $\chi$,
457: \begin{eqnarray} \label{dwpot}
458: \chi V & = & \left( 1+\frac{\chi d_y^4}{1-d_y^2/d_x^2}
459: \left\{
460: \left[
461: \frac{-y}{(x+d_x)^2+y^2} \right.\right.\right. \nonumber \\
462: & & \left. \left.\left.
463: + \frac{-y}{(x-d_x)^2+y^2} +\frac{1}{d_y}
464: \right]^2 \right.\right. \nonumber \\
465: & & \left. \left. +
466: \left[
467: \frac{x+d_x}{(x+d_x)^2+y^2} + \frac{x-d_x}{(x-d_x)^2+y^2}
468: \right]^2
469: \right\}
470: \right)^{1/2}\, . \nonumber \\
471: \end{eqnarray}
472: Figure \ref{fig.2}b) shows
473: the corresponding bistable potential for the particular case of
474: $\chi=0.067$, $I=200 \mu$A, $y_0=100$ nm and $x_0=200$ nm.
475: The two minima are clearly discerned.
476: To see how the frequency in the
477: single well develops if the current in the second wire is turned on,
478: we introduce the reference frequency $\omega_0$ in the single-well case with a
479: fixed current $I$ and a fixed transverse field $B_x$, such that
480: $y_0=x_0/2$. Then we obtain the ratio
481: \begin{equation}
482: \frac{\omega}{\omega_0}
483: =\left[
484: \frac{1}{16} \left(\frac{x_0}{y_0}\right)^4
485: \left( 1-\frac{y_0^2}{x_0^2}\right)
486: \right]^{1/3} \, .
487: \end{equation}
488: For decreasing $B_x$ and keeping $I$ constant, we find that
489: $\omega$ decreases as shown in Fig.\ \ref{fig.3} (black solid line
490: and left scale), while the distance
491: $y_0$ of the atom cloud increases. In the limit $x_0=y_0$, the
492: two minima merge and the potential becomes quartic and monostable,
493: implying that $\omega\rightarrow 0$.
494: For the above parameter set, we find $\omega_0=2 \pi
495: \times 291$ kHz.
496: Since one could obtain the same $\omega_0$ for a larger current
497: $I$ and a correspondingly larger
498: distance $x_0$, one gets the same trap frequency for
499: a fixed ratio of $y_0/x_0$. However, $d_x$ and $d_y$ themselves
500: would change and since the parabolic frequency $\omega$ is fixed,
501: only the non-linear corrections to the parabolic potential
502: will be modified.
503: This in turn influences the height of the potential
504: barrier and
505: the tunneling rate between the two wells. Next we study the influence
506: of the length scale $x_0$ on these two quantities.
507:
508: Taking the full potential into account, we estimate the barrier
509: height and the tunneling rate within a simple single-particle WKB
510: approximation. The barrier
511: height $D$ separating the two stable wells,
512: \begin{equation}
513: \frac{D}{\hbar \omega} = \chi^{-1}
514: \left( 1+ \chi d_y^2 \frac{1-d_y/d_x}{1+d_y/d_x}
515: \right)^{1/2}-\chi^{-1} \, ,
516: \end{equation}
517: is shown as a function of $y_0/x_0$
518: for two values of $I$
519: in the inset of Fig.\ \ref{fig.3}.
520: Note that the barrier height is of the order of a few multiples of
521: the energy gap in the wells, implying that the potential is in the
522: deep quantum regime, favoring quantum-mechanical tunneling between the two
523: wells. The corresponding tunneling rate $\Gamma$ for the
524: lowest-lying pair of energy eigenstates follows in WKB approximation
525: as
526: \begin{equation}\label{gamma1}
527: \frac{\Gamma}{\omega} = e^{-\int_{x_a}^{x_b} dx \sqrt{2 [
528: V(x,d_y)-1]}} \, ,
529: \end{equation}
530: %
531: where $x_{a/b}$ are the (dimensionless) classical turning points in the inverted
532: potential at energy $E=\hbar \omega$, which is approximately the
533: ground-state energy of a single well. The integral in
534: Eq.~(\ref{gamma1}) is calculated along the line connecting the two
535: minima corresponding to $y=y_0$. Results for $\Gamma$ are shown in
536: Fig.\ \ref{fig.3} (red solid lines and right scale) as a function of
537: $y_0/x_0$ for two different values of the current $I$ and the distance
538: $x_0$ yielding the same $\omega_0$. Note that for
539: the smaller current, $I=200$ $\mu$A, $\Gamma$ assumes large
540: values already for large frequencies $\omega$. This also implies that
541: the detrimental effects discussed above are less efficient.
542: On the other hand, for large currents,
543: the tunneling regime is entered only for much smaller trap
544: frequencies. For the above parameters, we find $\omega_0=2 \pi
545: \times 291$ kHz. For the smaller current, the tunneling regime starts at
546: frequencies of around $\omega=0.37 \omega_0 =2 \pi
547: \times 108 $ kHz, corresponding
548: to a temperature of $T=32$ $\mu$K, while for the larger current, the tunneling
549: regime is entered at $\omega=0.18 \omega_0 =2 \pi
550: \times 52 $ kHz corresponding to $T=16$ $\mu$K.
551: This behaviour illustrates qualitatively (in the single-particle picture)
552: one of the benefits of miniaturization. We believe that the features also
553: appear in a more detailed consideration involving the atomic correlations which
554: is not pursued here.
555:
556: A potential drawback of the double wire configuration could be
557: the transverse NT deflection
558: due to their mutual magnetic repulsion. For an estimate, note that
559: the NT displacement field $\phi(z,t)$ obeys
560: the equation of motion $\rho_L \ddot{\phi}=-Y M_I \phi'''' + \mu_0
561: I^2/(4 \pi x_0)$. The static solution under the boundary conditions
562: $\phi(0,L) = \phi'(0,L)=0$
563: is $\phi(z) =\mu_0 [Iz(z-L)]^2/(96 \pi Y M_I x_0)$.
564: Using again parameters from Ref.~\cite{tubes},
565: we find the maximum displacement $\phi(L/2)\approx 0.03$~nm
566: for $L=10\mu$m. Hence the mutual magnetic repulsion of the NTs is very weak.
567: Finally, we note that a potential misalignment of the two NT wires is
568: no serious impediment for the design. Experimentally available techniques could be combined
569: which allow on the one hand to move a NT on a substrate by an atomic
570: force microscope \cite{Henk00}, while on the other hand, the NTs can
571: be suspended and contacted
572: after being positioned \cite{Kim02}.
573:
574: \section{Conclusions}
575:
576: To conclude, we propose a nanoscale waveguide for ultracold atoms
577: based on doubly clamped suspended nanotubes. We have analyzed this
578: scenario from an atom chip point of view.
579: All common sources of imperfection can be made sufficiently
580: small to enable stable operation of the setup.
581: Two suspended NTs can be combined to create a bistable potential in
582: the deep quantum regime.
583: When compared to conventional atom-chip traps employed in
584: present experiments, such nanotraps
585: offer several new and exciting perspectives that hopefully
586: motivate experimentalists to realize this proposal. More refined
587: models to study the interplay between the mechanical motion of the
588: NTs and the coherent dynamics of the atom cloud are imaginable and could
589: establish a link between the field of nanoelectromechanical systems
590: and cold atom physics.
591:
592: First, rather large trap frequencies can be achieved while at
593: the same time using smaller wire
594: currents. This becomes possible here because both the
595: spatial size of the atom cloud and its distance to the current-carrying
596: wire(s) would be reduced to the nanometer scale, and
597: because NTs allow typical current
598: densities of $10 \mu$A$/$nm$^2$, which should be compared to
599: the corresponding densities of $10$ nA$/$nm$^2$ in noble metals.
600: For the case of a single-well trap, the resulting trap
601: frequencies go beyond realized chip traps \cite{Folman02}.
602: Large trap frequencies at low currents are generally
603: desirable, since detrimental effects like decoherence,
604: Majorana spin flips, or atom loss will then be significantly reduced.
605: Moreover, the faster dynamics of the atoms could lead to the construction of
606: fast ''chip circuits''.
607:
608: Second, regarding our proposal of a bistable potential with strong
609: tunneling, the miniaturization towards the nanoscale represents a
610: novel opportunity to study coherent and incoherent tunneling of a
611: macroscopic number of cold atoms. The
612: proposed bistable nanotrap is characterized by
613: considerably reduced tunneling distances, thus allowing for large
614: tunneling rates at large trap frequencies.
615: Note that the energy scale associated with
616: tunneling is larger than thermal energies for realistic temperatures.
617: Such a bistable device could then switch between the two stable
618: states on very short time scales enabling the design of fast
619: switches.
620: Within our proposal the parameters of the bistable potential can be
621: tuned over a wide range by modifying experimentally
622: accessible quantities like the current or magnetic fields.
623:
624: A third advantage of this proposal results from the
625: homogeneity of the currents flowing through the NTs.
626: As NTs are characterized by long mean free paths, they
627: often constitute (quasi-)ballistic conductors, where extremely
628: large yet homogeneous current densities are possible
629: which avoids the fragmentation problem \cite{Folman02}.
630:
631: Detection certainly constitutes an experimental challenge in this
632: truly 1D limit. However, we note that
633: single-atom detection schemes are currently being developed,
634: which would also allow to probe the tight 1D cloud here,
635: e.g., by combining cavity quantum electrodynamics
636: with chip technology \cite{Reichel02}, or by using additional perpendicular
637: wires/tubes `partitioning' the atom cloud \cite{reichel}.
638: This may then allow to study interesting many-body physics in
639: 1D in an unprecedented manner.
640:
641:
642:
643: \section{Acknowledgments}
644:
645: We thank A.\ G\"orlitz, Y.\ Kobayashi, C.\ Mora, H.\ Postma, and J.\ Schmiedmayer
646: for fruitful discussions. V.\ P.\ and M.\ T.\ would like to
647: thank H.\ Takayanagi and K.\ Semba for the kind hospitality at the NTT
648: Basic Research Laboratories, where parts of this work have been
649: accomplished. We acknowledge support by the
650: DFG-SFB TR-12 and by the Japanese CREST/JST.
651:
652: \begin{thebibliography}{99}
653:
654: \bibitem{Folman02} R. Folman, P. Kr\"uger, J. Schmiedmayer, J. Denschlag,
655: and C. Henkel, Adv. At. Mol. Opt. Phys. {\bf 48}, 263 (2002)
656:
657: \bibitem{Reichel02}
658: J. Reichel, Appl. Phys. B {\bf 75}, 469 (2002)
659:
660: \bibitem{atomBEC}
661: H. Ott, J. Fortagh, G. Schlotterbeck, A. Grossmann, and
662: C. Zimmermann, Phys. Rev. Lett. {\bf 87}, 230401 (2001);
663: W. H\"ansel, P. Hommelhoff, T.W. H\"ansch,
664: and J. Reichel, Nature {\bf 413}, 498 (2001);
665: A. Leanhardt, Y. Shin, A. P. Chikkatur, D. Kielpinski,
666: W. Ketterle, and D. E. Pritchard, Phys. Rev. Lett. {\bf 90}, 100404 (2003);
667: S. Schneider, A. Kasper, Ch. vom Hagen, M. Bartenstein,
668: B. Engeser, T. Schumm, I. Bar-Joseph, R. Folman,
669: L. Feenstra, and J. Schmiedmayer, Phys. Rev. A {\bf 67}, 023612 (2003)
670:
671: \bibitem{Henkel99}
672: C. Henkel, S. P{\"o}tting, and M. Wilkens, Appl. Phys. B
673: {\bf 69}, 379 (1999)
674:
675: \bibitem{chin}
676: Yu-ju Lin, I. Teper, C. Chin, and V. Vuleti{\'c},
677: Phys. Rev. Lett. {\bf 92}, 050404 (2004)
678:
679: \bibitem{Schroll03}
680: C. Schroll, W. Belzig, and C. Bruder, Phys. Rev. A
681: {\bf 68}, 043618 (2003)
682:
683: \bibitem{Kasevich}
684: M.A. Kasevich, Science {\bf 298}, 136 (2002)
685:
686:
687: \bibitem{tubes}
688: M.S. Dresselhaus, G. Dresselhaus, and Ph. Avouris (eds.),
689: {\em Carbon Nanotubes} (Berlin, Springer 2001)
690:
691: \bibitem{Petrov04}
692: D.S. Petrov, D.M. Gangardt, and G.V. Shlyapnikov,
693: J. Phys. IV France {\bf 116}, 5 (2004)
694:
695: \bibitem{chen}
696: S. Chen and R. Egger, Phys. Rev. A {\bf 68}, 063605 (2003).
697:
698: %\bibitem{recati}
699: %A. Recati, P.O. Fedichev, W. Zwerger, and P. Zoller,
700: %Phys. Rev. Lett. {\bf 90}, 020401 (2003).
701:
702: %\bibitem{pham}
703: %K.-V. Pham, M. Gabay, and P. Lederer, Phys. Rev. B {\bf 61}, 16397 (2000).
704:
705: \bibitem{becbcs}
706: I.V. Tokatly, Phys. Rev. Lett. {\bf 93}, 090405 (2004);
707: J.N. Fuchs, A. Recati, and W. Zwerger, {\em ibid.} {\bf 93}, 090408 (2004);
708: C. Mora, R. Egger, A.O. Gogolin, and A. Komnik,
709: {\em ibid.} {\bf 93}, 170403 (2004)
710:
711: \bibitem{olshanii} M. Olshanii, Phys. Rev. Lett. {\bf 81}, 938 (1998);
712: T. Bergeman, M.G. Moore, and M. Olshanii,
713: {\em ibid.} {\bf 91}, 163201 (2003)
714:
715: \bibitem{esslinger}
716: T. St\"oferle, H. Moritz, C. Schori, M. K\"ohl, and T. Esslinger,
717: Phys. Rev. Lett. {\bf 92}, 130403 (2004)
718:
719: \bibitem{Paredes04}
720: B.\ Paredes, A.\ Widera, V.\ Murg, O.\ Mandel, S.\ F\"olling, I.\
721: Cirac, G.\ V. Shlyapnikov, T.\ W.\ Hänsch, and I.\ Bloch,
722: Nature {\bf 429}, 277 (2004)
723:
724: \bibitem{Weiss}
725: T. Kinoshita, T. Wenger, and D.S. Weiss, Science {\bf 305}, 112 (2004)
726:
727: \bibitem{Goerlitz01}
728: A.\ G{\"o}rlitz, J. M. Vogels, A. E. Leanhardt, C. Raman,
729: T. L. Gustavson, J. R. Abo-Shaeer, A. P. Chikkatur,
730: S. Gupta, S. Inouye, T. Rosenband, and W. Ketterle, Phys. Rev. Lett. {\bf 87}, 130402 (2001)
731:
732: \bibitem{sukumar} C.V. Sukumar and D.M. Brink, Phys. Rev. A {\bf 56},
733: 2451 (1997)
734:
735: \bibitem{jones}
736: M.\ P.\ A.\ Jones, C.\ J.\ Vale, D.\ Sahagun, B.\ V.\ Hall, and E.A. Hinds,
737: Phys. Rev. Lett. {\bf 91}, 080401 (2003)
738:
739: \bibitem{Sapmaz03}
740: S. Sapmaz, Ya.M. Blanter, L. Gurevich, and H.S.J. van
741: der Zant, Phys. Rev. B {\bf 67}, 235414 (2003)
742:
743: \bibitem{Casimir48}H.\ B.\ G.\ Casimir and D.\ Polder, Phys.\ Rev.\
744: {\bf 73}, 360 (1948)
745:
746: \bibitem{Peano}V.\ Peano, M.\ Thorwart, C.\ Mora, and R.\ Egger,
747: unpublished results, see also cond-mat/0411517
748:
749: \bibitem{Lieb} E.\ H.\ Lieb, W.\ Liniger, Phys.\ Rev.\ {\bf 130}, 1616 (1963)
750:
751: \bibitem{Dunjko}V.\ Dunjko, V.\ Lorent, and M.\ Olshanii,
752: Phys.\ Rev.\ Lett.\ {\bf 86}, 5413 (2001)
753:
754:
755: \bibitem{weiss} U. Weiss, {\em Quantum Dissipative Systems} (World
756: Scientific, Singapore, 1999)
757:
758: \bibitem{Henk00} H.\ W.\ Ch.\ Postma, A.\ Sellmeijer, and C.\ Dekker, Adv.\ Mater.\
759: {\bf 17}, 1299 (2000)
760:
761: \bibitem{Kim02} G.-T. Kim, G. Gu, U. Waizman, and S. Roth, Appl. Phys. Lett. {\bf 80}, 1815 (2002)
762:
763: \bibitem{reichel}
764: J. Reichel and J.H. Thywissen, J. Phys. IV France {\bf 116}, 265 (2004)
765:
766:
767:
768: \end{thebibliography}
769:
770:
771: \begin{table}
772: \caption{Trap frequencies $\omega$, distances $y_0$ of the atomic
773: cloud from the NT wire, and oscillator lengths $l_0$ for
774: $\chi=0.067$ and various $I,d$.
775: \label{tab.freq}}
776: % For LaTeX tables use
777: \begin{tabular}{|r|r|r|r|r|}
778: \hline
779: $I({\rm \mu A})$&$d$&$\omega({\rm kHz})$&$y_0({\rm nm})$&$l_0( {\rm nm})$\\
780: \hline
781: 1000&10&2$\pi\times$460&144&14\\
782: 250&5&2$\pi\times$460&72&14\\
783: 250&10&2$\pi\times$28.7&576&58\\
784: 100&5&2$\pi\times$73.8&180&36\\
785: 100&10&2$\pi\times$4.6&1440&144\\
786: 50&5&2$\pi\times$18.4&360&72\\
787: 25&5&2$\pi\times$4.6&720&144\\
788: \hline
789: \end{tabular}
790: \end{table}
791: %
792:
793: \begin{table}
794: \caption{Typical results for the longitudinal size $\ell$ of the
795: $^{87}$Rb cloud for realistic values of the transversal trap frequency
796: $\omega$ and the atom number $N$, where $\omega_z=2 \pi \times 0.1$ kHz.
797: For $\eta$, see text.
798: \label{tab.number}}
799: % For LaTeX tables use
800: \begin{tabular}{|r|r|r|r|r|}
801: \hline
802: $\omega({\rm kHz})$&$a_{1D}( {\rm
803: nm})$&N&$\eta$&$\ell( \mu{\rm m})$\\
804: \hline
805: 2$\pi\times$460&-26.65&30&0.11&7.7\\
806: 2$\pi\times$460&-26.65&50&0.15&10\\
807: 2$\pi\times$73.8&-223&30&0.67&7.3\\
808: 2$\pi\times$73.8&-223&50&0.94&8.7\\
809: 2$\pi\times$73.8&-223&100&1.49&11\\
810: 2$\pi\times$28.76&-603&30&2.55&5.3\\
811: 2$\pi\times$28.76&-603&50&3.58&6.3\\
812: 2$\pi\times$28.76&-603&100&5.72&7.9\\
813: \hline
814: \end{tabular}
815: \end{table}
816: %
817:
818:
819: \begin{figure}[h!]
820: \begin{center}
821: \resizebox{0.37\textwidth}{!}{%
822: \includegraphics{fig1.eps}
823: }
824: \caption{Sketch of the proposed device. A current-carrying
825: suspended NT is positioned at $(-x_0,0,z)$ and together with the
826: transverse magnetic field $B_x$, a 1D trapping potential is formed.
827: The shaded region indicates the atom gas. A similar two-wire setup
828: allows the creation of a bistable potential. \label{fig.1}}
829: \end{center}
830: \end{figure}
831:
832: \begin{figure}[h!]
833: \begin{center}
834: a1)
835: \resizebox{0.20\textwidth}{!}{%
836: \includegraphics{fig2a1.eps}
837: }
838: \hfill
839: b1)
840: \resizebox{0.20\textwidth}{!}{%
841: \includegraphics{fig2b1.eps}
842: }\\
843: a2)\hspace*{-0.3cm}
844: \resizebox{0.20\textwidth}{!}{%
845: \includegraphics{fig2a2.eps}
846: }
847: \hfill
848: b2)\hspace*{-0.3cm}
849: \resizebox{0.20\textwidth}{!}{%
850: \includegraphics{fig2b2.eps}
851: }\hspace*{0.3cm}
852: \caption{(a) Transverse trapping potential of the nanoscale
853: waveguide for $I=100$ $\mu$A, $d=10$, $\chi=0.067$ and $x_0=l_0$. The
854: resulting trap frequency is $\omega = 2 \pi \times 4.6$ kHz while
855: $y_0=1440$ nm, corresponding to $B_x=0.14$ G. (a1) shows a cut along $y=y_{\rm min}$
856: through the contour plot shown in (a2), see horizontal dashed line.
857: (b) Bistable potential for the double-wire configuration for
858: $\chi=0.067, I=100$ $\mu$A, $x_0=200$ nm and $y_0=100$ nm. (b1) displays a cut
859: along $y=y_{\rm min}$ through the contour shown in (b2), see dashed line.
860: \label{fig.2}}
861: \end{center}
862: \end{figure}
863:
864: %\begin{figure}[h!]
865: %\begin{center}
866: %a)
867: %\resizebox{0.21\textwidth}{!}{%
868: % \includegraphics{fig2a.eps}
869: %}
870: %\hfill
871: %b)
872: %\resizebox{0.21\textwidth}{!}{%
873: % \includegraphics{fig2b.eps}
874: %}
875: %\caption{(Color online). (a) Transverse trapping potential of the nanoscale
876: %waveguide for $I=100$ $\mu$A, $d=10$, $\chi=0.067$ and $x_0=l_0$. The
877: %resulting trap frequency is $\omega = 2 \pi \times 4.6$ kHz while
878: %$y_0=1440$ nm, corresponding to $B_x=0.14$ G.
879: %(b) Bistable potential for the double-wire configuration for
880: %$\chi=0.067, I=100$ $\mu$A, $x_0=200$ nm and $y_0=100$ nm. \label{fig.2}}
881: %\end{center}
882: %\end{figure}
883:
884: \begin{figure}[h!]
885: \begin{center}
886: \resizebox{0.38\textwidth}{!}{%
887: \includegraphics{fig3.eps}
888: }
889: \caption{(Color online). Trap frequency $\omega$ in the bistable potential (left
890: scale) and tunneling rate $\Gamma$ within the WKB-approximation (right
891: scale) as a function of the ratio $y_0/x_0=d_y/d_x$. For the
892: definition of $\omega_0$, see text. The tunneling rate $\Gamma$ is computed
893: for $^{87}$Rb
894: atoms with $\chi=0.067$ and $x_0=200$ nm for two values of the current
895: $I$ given in the figure.
896: \label{fig.3}}
897: \end{center}
898: \end{figure}
899:
900:
901:
902: \end{document}
903:
904: % end of file template.tex
905: