Short-Distance Correlation Properties of the Lieb-Liniger System and Momentum Distributions of...

4
Short-Distance Correlation Properties of the Lieb-Liniger System and Momentum Distributions of Trapped One-Dimensional Atomic Gases Maxim Olshanii * and Vanja Dunjko Department of Physics & Astronomy, University of Southern California, Los Angeles, California 90089-0484, USA and Institute for Theoretical Atomic and Molecular Physics, Harvard-Smithsonian Center for Astrophysics, Cambridge, Massachusetts 02138, USA (Received 29 October 2002; published 26 August 2003) We derive exact closed-form expressions for the first few terms of the short-distance Taylor expansion of the one-body correlation function of the Lieb-Liniger gas. As an intermediate result, we obtain the high-p asymptotics of the momentum distribution of both free and harmonically trapped atoms and show that it obeys a universal 1=p 4 law for all values of the interaction strength. We discuss the ways to observe the predicted momentum distributions experimentally, regarding them as a sensitive identifier for the Tonks-Girardeau regime of strong correlations. DOI: 10.1103/PhysRevLett.91.090401 PACS numbers: 05.30.Jp, 02.30.Ik, 03.75.–b Introduction.—Even though the correlation functions for the Lieb-Liniger gas of -interacting one-dimensional bosons [1] have been an object of intense research in the integrable systems community since the late 1970s [2], the full closed-form expressions are known only in the Tonks-Girardeau limit of infinitely strong interactions [3]. While the scaling properties of the long-range [4] asymptotics of the correlation functions can be derived from Haldane’s theory of quantum liquids [5], conformal field theory [6], and quantum inverse scattering method [2,7], virtually nothing is known about short-range one- body correlations at finite coupling strength [8]. One of the goals of this paper is to extend the existing knowledge in this direction. The interest to the correlation functions of the Lieb- Liniger gas has been significantly revived over the past two years. The few-body local correlation functions for both weak and strong interactions were computed via new methods [9–11]. Relevant to the current experiments, research on correlation functions of a harmonically trapped gas has been performed in [9,12–14]. It is known that, while for weak interactions the Lieb- Liniger system is well described by the mean-field theory, the opposite, Tonks-Girardeau regime of infinitely strong interactions [15,16] constitutes a strongly corre- lated system dual to a free Fermi gas. In experiments with one-dimensional atomic gases [17,18], the one-body mo- mentum distribution of the gas, along with the density profiles [19] and phase fluctuations [9,20,21], can readily help to distinguish between the two quantum regimes. In the Tonks-Girardeau limit, the momentum distribution for both free and harmonically confined gases was inves- tigated by several authors [3,22–24]. In this paper, we address the question of the momentum distribution in the intermediate, in between mean-field and Tonks- Girardeau, regime, as more realistic from the experimen- tal point of view. System of interest.—Consider a one-dimensional gas of N-interacting bosons confined in a length L box with periodic boundary conditions. The Hamiltonian of the system reads ^ H h 2 2m X N j1 @ 2 @z 2 j g 1D X N i1 X N ji1 z i z j (1) Z L=2 L=2 h 2 2m @ z ^ y @ z ^ g 1D 2 ^ y ^ y ^ ^ ; (2) where m is the atomic mass, and g 1D is the one- dimensional coupling constant, whose expression for real atomic traps is given in [22]. This Hamiltonian can be diagonalized via Bethe ansatz [1]. At zero tem- perature, the energy of the system is given through E=N h 2 =2mn 2 e, where the dimensionless pa- rameter 2=nja 1D j is inversely proportional to the one-dimensional gas parameter nja 1D j, n is the one- dimensional number density of particles, a 1D 2 h 2 =mg 1D is the one-dimensional scattering length introduced in [22], and the function e is given by the solution of Lieb-Liniger system of equations [1]: It is tabulated in [25]. Note the asymptotic behavior of e (first computed in [1]): e !0 and e !1 1 3 2 2 2 , where ! 0 corresponds to the mean-field or Thomas-Fermi regime, whereas !1 corresponds to the Tonks-Girardeau regime. High-p momentum distribution.—Our first object of interest is the high-p asymptotics of the one-body mo- mentum distribution in the ground state. To evaluate it, we need two mathematical facts, (a) and (b). (a) The presence of the delta-function interactions in the Hamiltonian (1) implies that its eigenfunctions undergo, at the point of contact of any two particles i and j, a kink in the derivative proportional to the value of the eigenfunction at this point [26]: z 1 ; ... ;z i ; ... ;z j ; ... ;z N z 1 ; ... ;Z ji ; ... ;Z ji ; ... ;z N f1 jz ji j=a 1D "jz ji j; fZ ji gg "jz ji j; fZ ji gOjz ji j 2 ; (3) PHYSICAL REVIEW LETTERS week ending 29 AUGUST 2003 VOLUME 91, NUMBER 9 090401-1 0031-9007= 03=91(9)=090401(4)$20.00 2003 The American Physical Society 090401-1

Transcript of Short-Distance Correlation Properties of the Lieb-Liniger System and Momentum Distributions of...

P H Y S I C A L R E V I E W L E T T E R S week ending29 AUGUST 2003VOLUME 91, NUMBER 9

Short-Distance Correlation Properties of the Lieb-Liniger System and MomentumDistributions of Trapped One-Dimensional Atomic Gases

Maxim Olshanii* and Vanja DunjkoDepartment of Physics & Astronomy, University of Southern California, Los Angeles, California 90089-0484, USA

and Institute for Theoretical Atomic and Molecular Physics, Harvard-Smithsonian Center for Astrophysics,Cambridge, Massachusetts 02138, USA

(Received 29 October 2002; published 26 August 2003)

090401-1

We derive exact closed-form expressions for the first few terms of the short-distance Taylor expansionof the one-body correlation function of the Lieb-Liniger gas. As an intermediate result, we obtain thehigh-p asymptotics of the momentum distribution of both free and harmonically trapped atoms andshow that it obeys a universal 1=p4 law for all values of the interaction strength. We discuss the ways toobserve the predicted momentum distributions experimentally, regarding them as a sensitive identifierfor the Tonks-Girardeau regime of strong correlations.

DOI: 10.1103/PhysRevLett.91.090401 PACS numbers: 05.30.Jp, 02.30.Ik, 03.75.–b

N -interacting bosons confined in a length L box with ji ji ji

Introduction.—Even though the correlation functionsfor the Lieb-Liniger gas of -interacting one-dimensionalbosons [1] have been an object of intense research in theintegrable systems community since the late 1970s [2],the full closed-form expressions are known only in theTonks-Girardeau limit of infinitely strong interactions[3]. While the scaling properties of the long-range [4]asymptotics of the correlation functions can be derivedfrom Haldane’s theory of quantum liquids [5], conformalfield theory [6], and quantum inverse scattering method[2,7], virtually nothing is known about short-range one-body correlations at finite coupling strength [8]. One ofthe goals of this paper is to extend the existing knowledgein this direction.

The interest to the correlation functions of the Lieb-Liniger gas has been significantly revived over the pasttwo years. The few-body local correlation functions forboth weak and strong interactions were computed via newmethods [9–11]. Relevant to the current experiments,research on correlation functions of a harmonicallytrapped gas has been performed in [9,12–14].

It is known that, while for weak interactions the Lieb-Liniger system is well described by the mean-fieldtheory, the opposite, Tonks-Girardeau regime of infinitelystrong interactions [15,16] constitutes a strongly corre-lated system dual to a free Fermi gas. In experiments withone-dimensional atomic gases [17,18], the one-body mo-mentum distribution of the gas, along with the densityprofiles [19] and phase fluctuations [9,20,21], can readilyhelp to distinguish between the two quantum regimes. Inthe Tonks-Girardeau limit, the momentum distributionfor both free and harmonically confined gases was inves-tigated by several authors [3,22–24]. In this paper, weaddress the question of the momentum distribution inthe intermediate, in between mean-field and Tonks-Girardeau, regime, as more realistic from the experimen-tal point of view.

System of interest.—Consider a one-dimensional gas of

0031-9007=03=91(9)=090401(4)$20.00

periodic boundary conditions. The Hamiltonian of thesystem reads

HH h2

2m

XNj1

@2

@z2j g1D

XNi1

XNji1

zi zj (1)

Z L=2

L=2

h2

2m@z

y@zg1D2

yy ; (2)

where m is the atomic mass, and g1D is the one-dimensional coupling constant, whose expression forreal atomic traps is given in [22]. This Hamiltoniancan be diagonalized via Bethe ansatz [1]. At zero tem-perature, the energy of the system is given throughE=N h2=2mn2e, where the dimensionless pa-rameter 2=nja1Dj is inversely proportional to theone-dimensional gas parameter nja1Dj, n is the one-dimensional number density of particles, a1D 2 h2=mg1D is the one-dimensional scattering lengthintroduced in [22], and the function e is given bythe solution of Lieb-Liniger system of equations [1]:It is tabulated in [25]. Note the asymptotic behaviorof e (first computed in [1]): e

!0 and

e !1 1

32 2

2, where ! 0 corresponds to themean-field or Thomas-Fermi regime, whereas ! 1corresponds to the Tonks-Girardeau regime.

High-p momentum distribution.—Our first object ofinterest is the high-p asymptotics of the one-body mo-mentum distribution in the ground state. To evaluate it, weneed two mathematical facts, (a) and (b).

(a) The presence of the delta-function interactions inthe Hamiltonian (1) implies that its eigenfunctionsundergo, at the point of contact of any two particles iand j, a kink in the derivative proportional to the value ofthe eigenfunction at this point [26]:

z1;...;zi;...;zj;...;zNz1;...;Zji;...;Zji;...;zN

f1jzjij=a1D"jzjij;fZjigg

"jz j;fZ gOjz j2; (3)

2003 The American Physical Society 090401-1

P H Y S I C A L R E V I E W L E T T E R S week ending29 AUGUST 2003VOLUME 91, NUMBER 9

where Zji zi zj=2 and zji zj zi are the center-of-mass and relative coordinates of the ij pair ofparticles, respectively, and fZjig fZji; z1; . . . ; zi1;zi1; . . . ; zj1; zj1; . . . ; zNg denotes a set consisting ofthe center-of-mass coordinate of the ith and jth particlesand the coordinates of all the other particles.

090401-2

(b) Imagine that a periodic function fz, defined onthe interval L=2;L=2, has a singularity of theform fz jz z0jFz, where Fz is a regular func-tion, > 1 and 0; 2; 4; . . . ; then the leadingterm in the asymptotics of the Fourier transform of freads [27]

Z L=2

L=2dz eikzfz

jkj!12 cos

2 1

1eikz0Fz0

1

jkj1 O

1

jkj2

; (4)

where k 2=Ls and s is an integer. For multiple singular points of the same order, the full asymptotics is the sum ofthe corresponding partial asymptotics of the form (4).

Let us evaluate, using (3) and (4), the momentum representation of the ground state wave function of the Hamiltonian(1) with respect to the first particle:

p1; z2; . . . ; zN L1=2Z L=2

L=2dz1 ep1z1= hz1; z2; . . . ; zN

8i: 2iNL1=2

Z L=2

L=2dz1 eip1z1= h

z1 Z1i; . . . ; zi Z1i; . . . ; zN f1 jz1ij=a1D . . .g

jp1j!1XN

i2

2L1=2=a1Deip1zi= h z1 zi; . . . ; zi; . . . ; zN

1

p1= h2 : (5)

Here p1 2 h=Ls, where s is an integer.Let us now turn to the one-body momentum distribution per se. After a lengthy but straightforward calculation, it

takes the form

wp Z L=2

L=2dz2 . . .

Z L=2

L=2dzN jp; z2; . . . ; zNj

2 jpj!1 4N 1 20; 0; 0; 0

a21D

1

p= h4; (6)

where 2z1; z2; z01; z02 is the two-body density matrix,

normalized asRL=2L=2 dz1

RL=2L=2 dz2 2z1; z2; z1; z2 1,

and wp is the momentum distribution, normalized asP1s1 w2 hs=L 1.The expression (6) involves the two-body density ma-

trix whose form is unknown for a finite system. However,an elegant thermodynamic limit formula for 20; 0; 0; 0does exist due to Gangardt and Shlyapnikov [9], whoderived it using the Hellmann-Feynman theorem[28]: L2 20; 0; 0; 0 e0. We are now ready to givea closed-form thermodynamic limit expression for thehigh-p asymptotics of the one-body momentum distribu-tion for one-dimensional -interacting bosons in a boxwith periodic boundary conditions:

Wp jpj!1 1

hn2e02

hnp

4; (7)

where Wp L=2 hwp is normalized asR11 dpWp 1. Notice that this asymptotics is univer-

sally described by a 1=p4 law for all values of the cou-pling strength . (Note that, for ! 1, this law waspredicted in [24].) Formula (7) is the first of the twoprincipal results of our paper.

Harmonically trapped 1D gas: momentum distribu-tion.—To evaluate the high-p asymptotics of the momen-tum distribution of atoms confined in a harmonicoscillator potential, we are going to employ the localdensity approximation (LDA). It is based on an intui-tive, but hard to justify rigorously, assumption that in

the thermodynamic limit the short-range correlationproperties of a trapped gas are indistinct from theones of a uniform gas of the same local density:"z; z0

zz0l"uz z0 j nz z0=2, where "z; z0

and "uz z0 j n are one-body density matrix of thetrapped gas and one-body density matrix of a densityn uniform gas, respectively, both normalized to the re-spective number of particles, nz is the density profileof the trapped gas, and l is the typical length on whichthe density changes. [Tested against the exact results oncorrelation function of the trapped Tonks-Giradeau gas[13], this assumption can be shown to lead to an exactprediction for the value of the coefficient in front ofthe z z02 term in the Taylor expansion around z0 z 0.] From this ansatz, it immediately follows thatthe high-p asymptotics of the momentum distribu-tion (sensitive to the short-range correlations only) isgiven by the spatial average of the uniform case expres-sion (7) over the density profile of the atomic cloud. Thedensity profiles themselves can also be obtained usingLDA [ [19], Eqs. 21 and 22, where the governing pa-rameter % should be replaced by 2=0

TF, see below], andthis is the method we used. The final result is presented inFig. 1. There the dimensionless coefficient HO limjpj!1Wpp4= hn03 in front of the high-p asymp-totics of the momentum distribution is plotted as a func-tion of the interaction strength parameter 0 in the centerof the cloud; 0 in turn depends on of the experimental

090401-2

1e-06

0.0001

0.01

1

100

10000

1e+06

0.0001 0.01 1 100

ΩH

O(γ

0 )

Interaction strength, γ0

W(p) -> (/hn0)-1ΩHO(γ0)(/hn0/p)4

γ0TF ΩHO

exactThomas-Fermi

Tonks-Girardeau

FIG. 1. Dimensionless coefficient HO0 in front of the

leading term of the high-p asymptotics of the momentumdistribution of harmonically trapped atoms, as a function ofthe interaction strength 0 in the center of the cloud. Alsoshown is the directly related to the experimental parametersThomas-Fermi estimate 0

TF 8=32=3Nma21D!= h2=3 for

the interaction strength in the center, as a function of 0.Both momentum and momentum distribution are measuredin units related to the density in the center n0, that can beexpressed through 0 using 0 2=n0a1D.

P H Y S I C A L R E V I E W L E T T E R S week ending29 AUGUST 2003VOLUME 91, NUMBER 9

parameters, such as the number of particles N, the cou-pling constant g1D, and the longitudinal trap frequency!,through a system of implicit equations [19]. Here n0 is thedensity in the center of the trap, and HO stands for the‘‘harmonic oscillator.’’ To establish a link to the experi-mental parameters, we also present a plot for the Thomas-Fermi (weak interactions) prediction for in the centerof the atomic cloud, 0

TF 8=32=3Nma2!= h2=3, as afunction of 0.

In the limiting, Thomas-Fermi and Tonks-Girardeauregimes, the momentum distribution is given by

Wp jpj!1;!0 1

pHO

2 32=3

5N2=3

aHOja1Dj

5=2

pHO

p

4;

Wp jpj!1;!1 1

pHO

2

p 128

453 N3=2

pHO

p

4; (8)

where aHO h=m!1=2 and pHO h=aHO.Short-range expansion for the correlation function.—

Let us now redirect our attention to the ground state one-body correlation function,

g1z hyz0i; (9)

and, in particular, to its Taylor expansion around zero:

g1z=n 1 X1i1

cijnzji: (10)

In the limit of infinitely strong interactions ! 1, thisexpansion is known to all orders [3]: cTG1 0; cTG2 2=6; cTG3 2=9; cTG4 4=120; . . . . Our goal

090401-3

now is to obtain the first few (through the order jzj3)coefficients of the expansion (10) for an arbitrary inter-action strength .

The knowledge of the momentum distribution (7) iscrucial for determining the c1 and c3 coefficients. Let uslook at the relation between the momentum distribu-tion and the correlation function, where the former issimply the Fourier transform of the latter: Wp 2 hn1

R11 dz e

ipz= hg1z. Since the leading term inthe asymptotics of Wp is 1=p4, we may conclude, usingthe Fourier analysis theorem (4), that the lowest oddpower in the short-range expansion of the correlationfunction g1z is jzj3, and therefore the jzj term is absentfrom the expansion:

c1 0: (11)

Furthermore, the theorem (4) allows one to deduce thecoefficient c3 from the momentum distribution (7):

c3 1

122e0: (12)

To obtain the coefficient c2, we employ the Hellmann-Feynman theorem [28] again. Let a HamiltonianHHw depend on a parameter w. Let Ew be an eigen-value of this Hamiltonian. Then the mean value of thederivative of the Hamiltonian with respect to the parame-ter can be expressed through the derivative of theeigenvalue: hEwj

ddw HHwjEwi

ddw Ew. Let us

now denote the fraction h2=m as ( and differen-tiate the Hamiltonian (2) with respect to (. Accord-ing to the Hellmann-Feynman theorem, we get12

RL=2L=2 dz@

2=@z @z0hyzz0ijz0z dE=d(. Now,using hyz z0i hyz z0 0i, we obtain 1

2Ld2=dz2g1zjz0 dE=d(, and, finally,

c2 12fe e0g; (13)

where we have used the known expression for the energygiven above.

Note that, as expected, our expressions for the coeffi-cients c13 converge, in the limit ! 1, to the known[3] listed above results for the impenetrable bosons. Thiscan be easily verified using the ! 1 expansion for thefunction e given above.

Expressions (11)–(13) constitute the second principalresult obtained in our paper.

Concluding remarks.—Below, we present a discussionon empirical observation of and application for the 1=p4

momentum distribution tails, in experiments with har-monically trapped atomic gases.

We would like to note that the 1=p4 law predicted inour paper corresponds to the values of momentum largeas compared to the inverse correlation length pm)= h21=2, where ) is the chemical potential. The lawshould persist at finite temperatures, but the prefactorgiven above will be valid only for small temperature:

090401-3

P H Y S I C A L R E V I E W L E T T E R S week ending29 AUGUST 2003VOLUME 91, NUMBER 9

For higher temperatures, the value of prefactor can becorrected via methods developed in [10].

We believe that the coefficient in front of the high-ptail of the momentum distribution (Fig. 1) may provide arobust experimental identifier of the quantum regime ofthe gas of interest, and, in particular, serve to detect theTonks-Girardeau regime. (i) The high-p tail is not sensi-tive to the finite temperature corrections to the correlationfunction, which appear predominantly in the low-p(long-range) domain. (ii) In experiments with 2D opticallattices, where a single cigar-shaped trap is replaced by anarray of traps [17], the effect of the residual 3D mean-field pressure acting during the expansion becomes rele-vant: The high-p part of the momentum distribution is farless sensitive to this effect as compared to the low-p part.(iii) The theoretical interpretation of the experimentalresults is simpler in the high-p case thanks to the appli-cability of the LDA. In the opposite low-p case, the LDAleads to entirely wrong predictions [29].

Summary.—In this paper, we present a short-range Taylor expansion (up to the order jzj3) for thezero-temperature correlation function g1z of a one-dimensional -interacting Bose gas [see Eqs. (10)–(13)].We compute the leading term in the high-p asymptotics ofthe momentum distribution for both free [Eq. (7)] andharmonically trapped (Fig. 1) atoms.We regard the high-ptail of the momentum distribution as an efficient tool foridentification of the Tonks-Girardeau regime in experi-ments with dilute trapped atomic gases.

The authors are grateful to M. Girardeau, D. M.Gangardt, and C. Menotti for enlightening discussionson the subject. M. O. would like to acknowledge thehospitality of the European Centre for TheoreticalStudies in Nuclear Physics and Related areas (ECT)during the 2002 BEC Summer Program, where the pre-sented work was initiated. This work was supported bythe NSF Grant No. PHY-0070333 and ONR GrantNo. N000140310427. The authors appreciate financialsupport by NSF through the grant for Institute forTheoretical Atomic and Molecular Physics, HarvardSmithsonian Center for Astrophysics.

0904

*Electronic address: [email protected][1] Elliott H. Lieb and Werner Liniger, Phys. Rev. 130, 1605

(1963).[2] V. E. Korepin, N. M. Bogoliubov, and A. G. Izergin,

Quantum Inverse Scattering Method and CorrelationFunctions (Cambridge University Press, Cambridge,England, 1993).

[3] T. D. Schultz, J. Math. Phys. (N.Y.) 1, 666 (1963);A. Lenard, J. Math. Phys. (N.Y.) 5, 930 (1964); H. G.Vaidya and C. A. Tracy, Phys. Rev. Lett., 42, 3 (1979); acorrected version of the Vaidya-Tracy expansions can be

01-4

found in M. Jimbo, T. Miwa, Y. Mori, and M. Sato,Physica (Amsterdam) 1D, 80 (1980).

[4] The boarder between short and long range is given by thecorrelation length lc h=

m)

p, ) being the chemical

potential.[5] F. D. M. Haldane, Phys. Rev. Lett. 47, 1840 (1981);

F. D. M. Haldane, Phys. Lett. 81A, 153 (1981).[6] A. Berkovich and G. Murthy, Phys. Lett. A 142, 121

(1989).[7] V. Korepin and N. Slavnov, Phys. Lett. A 236, 201 (1997).[8] The first few terms of the 1= expansion of the short-

range correlation function can be deduced from theresults of Dennis B. Creamer, H. B. Thacker, and DavidWilkinson, Phys. Rev. D 23, 3081 (1981); Michio Jimboand Tetsuji Miwa, Phys. Rev. D 24, 3169 (1981).

[9] D. M. Gangardt and G.V. Shlyapnikov, Phys. Rev. Lett.90, 010401 (2003).

[10] K.V. Kheruntsyan, D. M. Gangardt, P. D. Drummond,and G.V. Shlyapnikov, cond-mat/0212153.

[11] D. M. Gangardt and G.V. Shlyapnikov, New J. Phys. 5, 79(2003).

[12] N. M. Bogoliubov, R. K. Bullough, C. Malyshev, andJ. Timonen, cond-mat/0207338.

[13] T. Papenbrock, Phys. Rev. A 67, 041601(R) (2003).[14] P. J. Forrester, N. E. Frankel, T. M. Garoni, and N. S.

Witte, Phys. Rev. A 67, 043607 (2003).[15] L. Tonks, Phys. Rev. 50, 955 (1936).[16] M. Girardeau, J. Math. Phys. (N.Y.) 1, 516 (1960).[17] M. Greiner, O. Mandel, T. Esslinger, T.W. Hansch, and

I. Bloch, Nature (London) 415, 39 (2002).[18] A. Gorlitz, J. M. Vogels, A. E. Leanhardt, C. Raman, T. L.

Gustavson, J. R. Abo-Shaeer, A. P. Chikkatur, S. Gupta,S. Inouye, T. Rosenband, and W. Ketterle, Phys. Rev. Lett.87, 130402 (2001); F. Schreck, L. Khaykovich, K. L.Corwin, G. Ferrari, T. Bourdel, J. Cubizolles, andC. Salomon, Phys. Rev. Lett. 87, 080403 (2001); KevinE. Strecker, Guthrie B. Partridge, Andrew G. Truscott,and Randall G. Hulet, Nature (London) 417, 150(2002); S. Burger, K. Bongs, S. Dettmer, W. Ertmer,K. Sengstock, A. Sanpera, G.V. Shlyapnikov, andM. Lewenstein, Phys. Rev. Lett. 83, 5198 (1999).

[19] Vanja Dunjko, Vincent Lorent, and Maxim Olshanii,Phys. Rev. Lett. 86, 5413 (2001).

[20] D. S. Petrov, G.V. Shlyapnikov, and J. T. M. Walraven,Phys. Rev. Lett. 85, 3745 (2000).

[21] J. O. Andersen, U. Al Khawaja, and H. T. C. Stoof, Phys.Rev. Lett. 88, 070407 (2002).

[22] M. Olshanii, Phys. Rev. Lett. 81, 938 (1998).[23] M. D. Girardeau, E. M. Wright, and J. M. Triscari, Phys.

Rev. A 63, 033601 (2001).[24] A. Minguzzi, P. Vignolo, and M. P. Tosi, Phys. Lett. A

294, 222 (2002).[25] The numerically tabulated function e can be down-

loaded from http://physics.usc.edu/~olshanii/DIST/.[26] See, for example, [2], p. 7.[27] N. Bleistein and R. Handelsman, Asymptotic Expansions

of Integrals (Dover, New York, 1986).[28] H. Hellmann, Z. Phys. 85, 180 (1933); R. P. Feynman,

Phys. Rev. 56, 340 (1939).[29] C. Menotti (private communication).

090401-4