cond-mat0309319/bht.tex
1: %\documentstyle[twocolumn,prl,aps,psfig,graphics]{revtex}
2: %\documentstyle[preprint,aps]{revtex4}
3: 
4: \documentclass[twocolumn,aps,showpacs,longtable]{revtex4}
5: %\documentclass[twocolumn,prl,aps,showpacs]{revtex4}
6: %\documentclass[aps]{revtex4}
7: %\documentclass{article}
8: 
9: \usepackage{amsfonts,amsmath,graphicx,latexsym}
10: \usepackage{epsfig}
11: 
12: \newcommand{\vc}[1]{\mathbf{#1}}
13: \newcommand{\g}{\frac{4\pi a}{m}}
14: 
15: \begin{document}
16: %\title{Superfluidity controlled by an Boltzmann-like  H-theorem}
17: %\title{A Boltzmann-like  H-theorem for a superfluid Bose gas}
18: \title{Collisionless dynamics of the condensate 
19: predicted in the random phase approximation}
20: %\title{Superfluidity predicted in the RPA model}
21: \author{Patrick Navez}
22: \affiliation{
23: Ecole Polytechnique, CP 165, Universit\'e Libre de Bruxelles, 1050
24: Brussels, Belgium}
25: 
26: \date{\today}
27: \begin{abstract}
28: From the microscopic theory, we derive a number conserving
29: quantum kinetic equation, valid for a dilute Bose  gas
30: at any temperature,
31: in which the binary collisions between the quasi-particles are
32: mediated by phonon-like
33: excitations (called ``condenson''). This different approach
34: starts from the many-body Hamiltonian of a Boson gas
35: and uses, in an appropriate way, the generalized random phase 
36: approximation.
37: As a result, the collision term of the kinetic equation  
38: contains higher order contributions in the expansion in the 
39: interaction parameter. This different expansion 
40: shows up that a scattering involves the 
41: emission and the absorption of a phonon-like excitation.
42: The major interest of this particular mechanism is 
43: that, in a regime where the condensate is stable,
44: the collision process between condensed and non condensed particles 
45: is totally blocked due to a total annihilation 
46: of the mutual interaction potential induced by the condensate itself. 
47: As a consequence, the condensate is not constrained 
48: to relax and can be superfluid.
49: Furthermore, a Boltzmann-like 
50: H-theorem for the entropy exists for this equation and
51: allows to distinguish between
52: dissipative  and non dissipative phenomena (like vortices).
53: We also illustrate the analogy between this approach and
54: the kinetic theory for a plasma, in which the excitations correspond
55: precisely to a plasmon. Finally, we show the equivalence of this 
56: theory with the non-number conserving Bogoliubov theory 
57: at zero temperature.
58: \end{abstract}
59: 
60: 
61: \pacs{03.75.Hh,03.75.Kk,05.30.-d}
62: \maketitle
63: 
64: \section{Introduction}
65: 
66: \subsection{Superfluidity and H-theorem}
67: 
68: A lot of studies have been devoted to the theoretical 
69: understanding of statistical and dynamical properties 
70: of a weakly interacting Bose condensed gas. In particular, 
71: many works have been accomplished on the derivation of quantum 
72: kinetic equations (QKE) that govern the evolution of the condensate 
73: fraction together with his thermal excitations 
74: \cite{Zaremba,Walser,Kirkpatrick,HM,LevichYakhot,Stoof,
75: Gardiner,Stoof2,Pomeau,Khalatnikov,Proukakis}. 
76: In a simple microscopic model of an homogeneous gas, 
77: one can describe the condensate by the atoms 
78: that populate the lowest ground state of energy and the thermal 
79: excitations by the atoms contained in the 
80: excited energy levels. The population of atoms in each level evolves 
81: according to the probability of scattering between the atoms. In 
82: the case of a uniform gas, the so-called Uehling-Uhlenbeck quantum 
83: kinetic equation (UUQKE) is a 
84: non-linear integral equation which describes 
85: the detailed balance of  
86: the population transfer of atoms for each mode of the wave-vector through 
87: a binary collision term \cite{Balescu}. 
88: This term depends linearly on the scattering differential 
89: cross section and nonlinearly on the mode population. It has also 
90: the remarkable 
91: properties to allow the QKE to obey conservation laws. 
92: These laws guarantee that the total number, the total momentum and the 
93: total kinetic energy 
94: of particles are preserved during the collision processes. More striking 
95: is the law stating that the production of 
96: entropy must be always positive, 
97: guaranteeing that the system obeys the second law of 
98: thermodynamics and, consequently, 
99: that it is always dissipative. 
100: This important requirement is known as the Boltzmann 
101: H-theorem established for a classical gas. 
102: 
103: The UUQKE has been derived in a weak coupling approximation, valid 
104: for a diluted gas for which the kinetic energy is much higher 
105: than the potential energy. The collision term is indeed a second order 
106: expansion in the interaction potential between atoms leading to 
107: an  expression of the differential cross section in the Born 
108: approximation. 
109: 
110: One could use such a QKE in a regime below 
111: the critical point of condensation. In particular, for an 
112: inhomogeneous
113: gas in a trap potential, 
114: the kinetic equation describes the evolution of the 
115: Wigner function and must contain additional terms,
116: taking into account  
117: the free propagation of the atoms, the 
118: influence of the external potential and the Hartree-Fock mean field.
119: The conservation laws are still valid and allow to express the 
120: hydrodynamic equations, including the equation for the local entropy 
121: production. 
122: 
123: Despite these consistencies, the resulting QKE suffers from the 
124: lack of understanding of an important phenomenon: {\it superfluidity} 
125: \cite{Leggett2}. 
126: If this phenomenon really exists for a dilute gas \cite{Cornell}, 
127: then the second order 
128: theory must necessarily be revisited as it does not take into account the 
129: frictionless motion. Indeed according to the H-theorem, the homogeneous 
130: gas with a zero total momentum evolves towards a statistical 
131: equilibrium state characterized by the Bose-Einstein particle number 
132: distribution $n^{eq}_{\vc{k}}=1/[\exp(\beta(\epsilon_{\vc{k}}-\mu))-1]$ 
133: where $\vc{k}$ is the wave-vector, 
134: $\beta=1/k_B T$ the inverse temperature, $\mu$ the chemical potential, 
135: and $\epsilon_{\vc{k}}=\vc{k}^2/2m$ the particle kinetic energy 
136: respectively ($\hbar=1$). 
137: For $\mu \rightarrow 0$ and $\vc{k} \rightarrow 0$, the ground state 
138: has a macroscopic population $n_{\vc{0}}= 1/[\exp(-\beta\mu)-1]$ having no 
139: relative velocity with the non condensed part of the gas.  A non zero 
140: relative velocity corresponds to a non equilibrium situation and collisions 
141: between condensed and non condensed atoms will irrevocably damp 
142: %(attenuated) 
143: this velocity towards zero. 
144: 
145: From this observation, we conclude that the second order theory is 
146: no longer valid as far as superfluidity is concerned. Obviously, 
147: at zero temperature, only the potential energy is the 
148: dominant contribution since 
149: the condensate is at rest and no thermal excitation subsists. 
150: Moreover, other indications confirm that an improved description of 
151: a weakly interacting Bose gas requires a higher order analysis in the 
152: interaction parameter.
153: % (we will choose the scattering length $a$ 
154: %for the interaction parameter). 
155: Among them, let us 
156: mention that the entropy of the condensate must be zero or close to zero 
157: while
158: the H-theorem predicts an entropy $S_0 \sim \log n_{\vc{0}}$ corresponding to 
159: a system in the Grand Canonical ensemble with large statistical 
160: fluctuations of the particle number compared to his average value 
161: $\delta n_{\vc{0}}/ n_{\vc{0}} \sim 1$ \cite{Huang}. This is due to the 
162: Bose enhancement factor which stimulates the collision rate in 
163: a huge manner as long as a condensed particle is involved in the process. 
164: On the other hand, the equilibrium statistical approach based 
165: on the partition 
166: function formalism indicates that  
167: the presence of a small interaction lowers considerably 
168: the fluctuations to an amount irrelevant in the thermodynamic 
169: limit $\delta n_{\vc{0}}/ n_{\vc{0}} \rightarrow 0$ \cite{4eme}. 
170: 
171: Landau was the first to give an explanation of  
172: the superfluidity mechanism. He found 
173: a necessary but not sufficient condition for which this 
174: phenomenon happens \cite{LL,HM}. He showed that a 
175: superfluid interacting with an external body 
176: (for example the wall) cannot release its 
177: energy, unless it evolves with a velocity higher 
178: than a critical one. The argument is based 
179: on the impossibility to satisfy the momentum-energy conservation 
180: requirement because the excitation emitted by the superfluid 
181: has a phonon-like dispersion relation.  
182:   
183: \subsection{Beyond the second order perturbation theory}
184: 
185: Attempts to analyze contribution coming from higher terms have been 
186: carried out with some success ( see \cite{Leggett,Pines} for a review). 
187: At zero temperature, Bogoliubov 
188: has calculated corrections to the ground state energy. 
189: The result is 
190: non analytic in the interaction parameter, due to infrared 
191: divergencies, and is obtained through a re-summation of an infinite 
192: number of contributions. Moreover, the theory 
193: predicts the existence of a phonon-like 
194: excitation, necessary in order to justify the Landau mechanism for 
195: superfluidity. 
196: 
197: Using Green function techniques, Beliaev improved 
198: the description of the phonon like excitation namely 
199: by calculating its damping rate. Later on, Hugenholtz 
200: and Pines (HP) demonstrated that this 
201: excitation must necessarily be gapless.
202: If at zero temperature, the theory seems well 
203: understood and widely accepted, for finite 
204: temperature it seems rather controversial.
205: Popov was one of the first to extent the Bogoliubov 
206: and Beliaev works using the Matsubara formalism 
207: at finite temperature \cite{Popov}. However, this 
208: approach is essentially valid for a weakly depleted 
209: gas and thus for temperatures much below 
210: the critical one $T_c$. To take into account a 
211: strong depletion, Girardeau derived the 
212: Hartree-Fock-Bogoliubov (HFB) mean field equations which 
213: unfortunately, have, the strong inconvenient of 
214: having a gap in the quasi-particle energy  spectrum \cite{Girardeau}. 
215: 
216: Since then, 
217: many attempts, have been made, in order 
218: to suppress this gap and ``rescue'' the HFB equations, with 
219: the purpose 
220: to go beyond the Bogoliubov phonon-like
221: dispersion relation.
222: \cite{Griffin,Gapless,PN}. 
223: A very popular one is the so-called 
224: "Popov approximation" which consists in suppressing 
225: the product of anomalous term in the HFB equations \cite{Griffin}. 
226: Another consists in the renormalization of the HFB equations 
227: but at the price of making the approximation
228: of a weakly depleted Bose gas
229: \cite{Gapless}.
230: Consequently, these approaches are only strictly valid for temperatures 
231: much below $T_c$.
232: {\it A derivation of a well-defined theory, 
233: which is able to describe the weakly interacting Bose  
234: gas for any regime of temperature, is still an open 
235: problem and remains to be established} (see the conclusion of 
236: \cite{NP2}). By well-defined, 
237: we mean that the expansion in the small interaction 
238: parameter must be valid, whether the condensate is strongly 
239: depleted or not. 
240: 
241: 
242: Nevertheless, using these approaches,  
243: QKE have been derived taking into account these higher order 
244: effects in the collision term. 
245: While some authors use the dispersion relation  
246: resulting from the Bogoliubov theory \cite{Gardiner,Kirkpatrick,Imamovic}, 
247: others use the 
248: one  resulting form the renormalized HFB theory \cite{Walser}. 
249: Superfluidity can be achieved in these models. Indeed, 
250: the resulting QKE can preserve detailed balance even 
251: for metastable states, in which a non zero relative 
252: velocity persists. In this situation, the Bogoliubov 
253: dispersion relation becomes asymmetric - due to a Lagrange 
254: multiplier representing the relative velocity - and 
255: remains non negative, as long as this velocity does not exceed 
256: the sound velocity. Otherwise, the dispersion relation 
257: is negative and the gas becomes unstable \cite{Huang}.   
258: Although these models represent a 
259: considerable amount of
260: work, they do not yet provide a full account of 
261: the conservation laws and the H-theorem 
262: that might be deduced from their QKE. 
263: 
264: However, let us mention that a QKE has been proposed 
265: valid in principle for temperatures close to $T_c$ \cite{Zaremba, Pomeau}. 
266: The main problem of this approach is that the dispersion 
267: relation has a gap and thus fails to explain the superfluidity. 
268: 
269: In general, in all these models, 
270: collisions between condensed and non 
271: condensed atoms are possible and their rate is 
272: huge because of the Bose enhancement factor in the same way 
273: as in the UUQKE. If we follow the same reasoning as above and 
274: if an H-theorem really exists, we might expect a  
275: condensate entropy of the same order of magnitude 
276: $S_0 \sim \log n_{\vc{0}}$.  
277: As a consequence, 
278: the particle number fluctuations  of the 
279: condensed mode are also huge, in contradiction with 
280: estimations made from the partition function formalism 
281: \cite{4eme}. 
282: 
283: 
284: \subsection{The random phase approximation approach}
285: 
286: The present paper is devoted to provide a 
287: possible alternative explanation to the problem of superfluidity 
288: in a diluted Bose gas. 
289: For this purpose a different QKE is derived in the 
290: random
291: phase approximation (RPA) in   
292: which higher order terms in the interaction parameter 
293: are retained in the binary collision term (in some 
294: work, the RPA QKE refers to the UUQKE and thus has a different 
295: meaning from this paper \cite{LevichYakhot}).
296: {\it 
297: Furthermore, in comparison with other approaches, 
298: the QKE is valid for any regime of temperature below 
299: and above $T_c$ since, under no circumstances, the 
300: approximation of weak depletion has been used.}   
301: 
302: The RPA is commonly used to describe the 
303: collective excitations in
304: the quantum plasma of an electron liquid \cite{NP}. 
305: The idea behind this approximation is to neglect  
306: contribution containing 
307: average over pair of field operators that are not oscillating 
308: in phase but rather randomly.
309: In 
310: the terminology of optics, we neglect contribution that are not 
311: phase matching. 
312: These contributions come essentially from averages over  
313: pairs of different modes which oscillate with 
314: a relative random phase.
315: The reason for using the RPA in a diluted  
316: plasma is 
317: that, we expect that the excitations propagate over a 
318: sufficiently long time 
319: that non phase-matching terms are destroyed by 
320: interference. By analogy with optics, we make somehow the far field 
321: approximation. The analysis of collective excitations in plasma 
322: using RPA reveals that 
323: the coulombian interaction potential between the electrons 
324: is shielded at long distance, due to the dynamic dielectric 
325: function. In  fact, these results generalize 
326: the Debye theory predicting the screening of the potential 
327: between a charged ion due to the presence of other   
328: surrounding ions.
329: As an extension, the RPA includes also the dynamical aspects of 
330: these charged particles. 
331: 
332: A QKE for the plasma has been 
333: derived \cite{QBalescu,WP}. It predicts that 
334: the coulombian interaction potential 
335: for the cross section is shielded precisely by this dynamic 
336: dielectric factor. 
337: The classical version of this equation is known 
338: as the BGL kinetic equation 
339: (Balescu-Guernsey-Lenard) \cite{Balescu}. For deriving it, Balescu uses 
340: a re-summation of the so-called ring diagrams, which is another 
341: approach equivalent to the RPA \cite{QBalescu}. Later on, 
342: Wyld and Pines 
343: established a connection between the QKE and the plasmon theory \cite{WP}. 
344: In their approach, the  
345: shielded potential results from a more subtile dynamical 
346: mechanism, in 
347: which the two  electrons must emit and reabsorb an 
348: intermediate plasmon during their interaction. 
349: The dispersion relation of the plasmon corresponds 
350: precisely to the collective mode of the plasma and 
351: its decay rate to the Landau damping. 
352: 
353: 
354: In this paper, an identical type of approach will be reproduced in the 
355: case of a dilute Bose gas. It is already 
356: known that the RPA allows to recover the Bogoliubov 
357: results for the ground state energy but with one 
358: important difference \cite{Pines}. Namely, in the RPA,
359: there is no  need of a spontaneous breaking of the symmetry $U(1)$ 
360: but rather, during the derivation, the total particle number 
361: is kept conserved. 
362: Note that a number-conserving formalism has been already 
363: used to derive a QKE \cite{Gardiner}. But this formalism  is 
364: based on a $1/N$ expansion \cite{GCD} and thus differs from 
365: the one in the RPA \cite{Pines}.
366: 
367: These considerations 
368: allow to derive a number conserving QKE in which the 
369: interaction potential is also modified by the presence 
370: of a dynamic dielectric function resulting from collective 
371: excitations. However, in contrast to a plasma,  
372: this function presents a strange behavior 
373: according to the kind of quasi-particles interaction it 
374: mediates. {\it When the interaction involves a particle of the 
375: macroscopic condensate,
376: it has the effect to totally annihilate the potential, forbidding 
377: any binary collision to occur.} 
378:  
379: This surprising result allows to explain why a 
380: metastable condensate cannot decay into a state of 
381: lower energy. Simply, it is forbidden for a 
382: particle of the condensate to scatter with the 
383: excited particle and to become thus excited. 
384: No exchange of particle can occur between the two 
385: fluids. This unexpected phenomenon is a consequence 
386: of the collective behavior induced by the presence 
387: of a macroscopic condensate. In short, collision cannot 
388: happen because an induced 
389: mean field force, generated by the condensate, 
390: compensates exactly the interaction 
391: force felt by the condensed and non condensed particle. 
392: The net result is the absence of an effective interaction 
393: force preventing any scattering.  
394: In other words, the dielectric function used to attenuate the 
395: binary interaction potential becomes simply infinite.
396: 
397: 
398: 
399: %Following Nozi\'eres and Pines, one can derive 
400: %equation of motion for the excitations. The solution 
401: %of this equation provides their energy  spectrum.
402: %There is a scattering solution corresponding 
403: %to an incoherent excitation and  of en 
404: 
405: 
406: This phenomenon of ``collision blockade'' also influences 
407: the mechanism of condensate 
408: formation. Since particle cannot be exchanged with 
409: the macroscopic condensate, other mechanisms 
410: must be found. A closer analysis reveals that the collision 
411: blockade happens only when the Bose gas is stable. By stable we mean 
412: that the collective oscillations are always damped. Precisely, 
413: the Landau damping plays this role in both a plasma and a 
414: condensate as long as we are close to equilibrium \cite{Balescu,SK}. 
415: However, for some non equilibrium situation, it has been 
416: shown for a plasma that the oscillations can grow exponentially 
417: leading to an instability \cite{instB,FR}. Such behavior happens also for 
418: a Bose gas which becomes thus unstable. For example, we 
419: will show that this is the case when the relative velocity between 
420: the normal and superfluid is higher than the speed of sound. 
421: In such instable regime, the picture of a collision blockade 
422: is no longer valid and the QKE gets more 
423: complicated. Such more sophisticated QKE can describe the 
424: exchange of particle with the condensate \cite{instB,FR} and provides an 
425: alternative explanation for the condensate formation 
426: \cite{Keterlee2,BZS,Gardiner2}. 
427: Another possibility is to exchange 
428: them on the edge region where the condensate is not macroscopically 
429: populated so that scattering can occur. But this requires 
430: also a really specific analysis. Therefore, the present 
431: QKE derived here will not address the important issue 
432: of particle exchange with the condensate.  
433: 
434: Assuming that collision processes are local
435: in space, the derivation can be extended for 
436: a weakly inhomogeneous
437: Bose gas. 
438: In this way, we recover the generalized Gross-Pitaevskii
439: equation for the condensate, but with the difference that 
440: the absence of a binary collision term in this equation has 
441: now a clear justification. When the normal cloud is in a local 
442: thermal 
443: equilibrium, no dissipation exists anymore and 
444: we can derive a set of coupled equations 
445: defining 
446: the superfluid regime for any finite temperature below $T_c$. 
447: 
448: 
449: The paper is divided as follows. 
450: We first begin by a heuristic approach of the ``collision blockade'' 
451: mechanism in section 2.
452: In section 3, we derive 
453: the QKE for a homogeneous Bose gas. We review the simple RPA in which 
454: only the Hartree or direct term is considered and 
455: the generalized RPA (GRPA) in which both Hartree and exchange Fock terms are 
456: considered \cite{NP}. An equation of motion for the particle-hole pair operator 
457: is derived from which the collective and scattering excitation 
458: energy spectra are deduced. The solution of the 
459: equation of motion allows to calculate the collision term and 
460: to predict the collision blockade phenomenon. We show 
461: that the QKE conserves the total particle number, the total momentum
462: and the total energy (kinetic and Hartree-Fock) 
463: and that a Boltzmann-like H-theorem exists. 
464: In section 4, the reasoning of the previous section is extended 
465: to the case of a weakly inhomogeneous Bose gas. From the resulting 
466: QKE, we delimitate the superfluid regime i.e. the regime  
467: in which there is no local production of entropy. In section 
468: 5, by analogy with plasmon theory, 
469: we interpret the collision phenomenon as the result 
470: of the exchange of an intermediate boson (condenson). 
471: In section 6, we show how the number conserving RPA theory 
472: allows to recover the Bogoliubov results for the 
473: ground state energy.  Finally, section 7 is devoted 
474: to conclusions and perspectives 
475: 
476: 
477: \section{The ``collision blockade'' phenomenon}
478: 
479: \subsection{Preliminary definitions}
480: 
481: We start from the Hamiltonian:
482: \begin{eqnarray}
483: H=\sum_{\vc{k}}
484: \epsilon_{\vc{k}}
485: %\frac{\vc{k}^2}{2m}
486: c^\dagger_{\vc{k}}
487: c_{\vc{k}}
488: +\sum_{\vc{k},\vc{k'},\vc{q}}
489: \frac{U_\vc{q}}{2V}
490: c^\dagger_{\vc{k}+\vc{q}}c^\dagger_{\vc{k'}-\vc{q}}
491: c_{\vc{k}}c_{\vc{k'}}
492: \end{eqnarray}
493: $c^\dagger_{\vc{k}}$ and $c_\vc{k}$ are the creation and
494: annihilation operators obeying the commutation relations
495: $[c_\vc{k},c^\dagger_\vc{k'}]=\delta_{\vc{k},\vc{k'}}$
496: $[c_\vc{k},c_\vc{k'}]=[c_\vc{k}^\dagger,c^\dagger_\vc{k'}]=0$.
497: $\epsilon_{\vc{k}}=\frac{\vc{k}^2}{2m}$ is the kinetic
498: energy of the particle.
499: $U_{\vc{q}}$ is the interaction potential expressed in
500: Fourier transform.
501: Since we are concerned only with low energy
502: binary collisions in the channel $l=0$,
503: it is common to replace the potential $U_{\vc{q}}$
504: by a pseudo-potential or contact potential
505: that we treat in the Born
506: approximation \cite{LL}:
507: \begin{eqnarray}\label{U0}
508: U_{\vc{q}}=\frac{4\pi a}{m}
509: [1+\frac{4\pi a}{mV}\sum_{\vc{q'}}
510: \frac{1}{\epsilon_{\vc{k}+\vc{q}}
511: +\epsilon_{\vc{k'}-\vc{q}}
512: -\epsilon_{\vc{k}}-\epsilon_{\vc{k'}}}]
513: %\frac{1}{\frac{\vc{q'}^2}{m}+\frac{(\vc{k}-\vc{k'}).\vc{q'}}{m}}]
514: \end{eqnarray}
515: where $a$ is the scattering length.
516: Usually we consider only the first linear term in the
517: scattering length. The second term is ultra-violet divergent and
518: is only present to renormalize the
519: theory by  removing eventual high energy divergencies.
520: In what follows we will concentrate on a repulsive
521: contact interaction $a > 0$.
522: 
523: The so-called annihilation and creation operators of the particle-hole pair
524: are of interest:
525: \begin{eqnarray}
526: \rho_{\vc{k},\vc{q}}=c^\dagger_{\vc{k}}c_{\vc{k}+\vc{q}} 
527: \ \ \ \ \ 
528: \rho^\dagger_{\vc{k},\vc{q}}=c^\dagger_{\vc{k}+\vc{q}}c_{\vc{k}}
529: \end{eqnarray}
530: They represent an excitation of momentum $\vc{q}$ created from
531: a particle which transfers its momentum from $\vc{k}$ to
532: $\vc{k}+\vc{q}$. The kinetic energy transfer for 
533: this excitation is given by 
534: $\omega_{\vc{k},\vc{q}}= \epsilon_{\vc{k}+\vc{q}}
535: -\epsilon_{\vc{k}}$.
536: In particular, we define
537: the density fluctuation operator
538: $\rho_{\vc{q}}=\sum_{\vc{k}} \rho_{\vc{k},\vc{q}}=
539: \rho^\dagger_{\vc{-q}}$,
540: the number operator $\hat n_{\vc{k}}=\rho_{\vc{k},0}$ 
541: and the total number $\hat N = \sum_{\vc{k}}
542: \hat n_{\vc{k}}$.
543: In terms of these operators, the Hamiltonian becomes:
544: \begin{eqnarray}
545: H=\sum_{\vc{k}}
546: \frac{\vc{k}^2}{2m}
547: \hat n_{\vc{k}}
548: +\sum_{\vc{q}}
549: \frac{U_{\vc{q}}}{2V}
550: (\rho^\dagger_{\vc{q}}\rho_{\vc{q}}-
551: \hat N)
552: \end{eqnarray}
553: 
554: 
555: \subsection{Heuristic approach}
556: 
557: Assume a macroscopic condensate containing 
558: $n_{\vc{k_s}}$ particle evolving with momentum $\vc{k_s}$. 
559: In the absence of excitations, there are no fluctuations 
560: of the density operator i.e. 
561: $\langle \rho_\vc{q}\rangle^{eq}=
562: \delta_{\vc{q},\vc{0}}n_{\vc{k_s}}$.
563: Suppose that an external potential is turned 
564: on creating locally fluctuations of 
565: the density of the condensate. Expressing 
566: these perturbations in Fourier space, we can 
567: characterize the external 
568: potential $\phi_{ext}(\vc{q},\omega)$ and 
569: the density fluctuations $\delta n(\vc{q},\omega)=
570: \langle \delta \rho_\vc{q}\rangle=
571: \langle \rho_\vc{q} \rangle-
572: \langle \rho_\vc{q} \rangle^{eq}$ 
573: in terms of its wave-vector $\vc{q}$ and frequencies 
574: $\omega$ components.
575: The Hamiltonian created 
576: by this external potential is given by 
577: $H_{ext}(t)=\sum_{\vc{q}}
578: \rho_{\vc{q}}^\dagger e^{-i\omega t}\phi_{ext}(\vc{q},\omega)
579: + c.c.$
580: The linear response to this potential is 
581: given by the formula:
582: \begin{eqnarray}\label{resp1}
583: \delta n(\vc{q},\omega)
584: =\chi (\vc{q},\omega)
585: \phi_{ext}(\vc{q},\omega)
586: \end{eqnarray} 
587: where $\chi (\vc{q},\omega)$ is the susceptibility 
588: function. This response function can be calculated formally 
589: by treating $H_{ext}(t)$ as a perturbation in the first order. 
590: Without the presence of 
591: a binary interaction potential between the particle, 
592: this function is given by
593: \cite{NP}:
594: \begin{eqnarray}
595: \chi_0 (\vc{q},\omega)=
596: \frac{n_{\vc{k_s}}}{
597: \omega-\omega_{\vc{k_s},\vc{q}}+i0_+}
598: -
599: \frac{n_{\vc{k_s}}}{
600: \omega+\omega_{\vc{k_s-q},\vc{q}}+i0_+}
601: \end{eqnarray} 
602: It represents the transition amplitude
603: for a condensed particle
604: to increase its kinetic energy by an amount
605: $\omega_{\vc{k_s},\vc{q}}=\epsilon_{\vc{k_s}+\vc{q}}
606: -\epsilon_{\vc{k_s}}$  minus the amplitude
607: for an excited particle to be transferred to
608: the condensate releasing the energy
609: $-\omega_{\vc{k_s-q},\vc{q}}=\epsilon_{\vc{k_s}-\vc{q}}
610: -\epsilon_{\vc{k_s}}$.
611: An infinitesimal quantity $i0_+$ has been 
612: added to ensure the convergence and is due to 
613: the adiabatic switching process of external perturbation.
614: In the presence of the interaction, the 
615: condensate acquires a self-interaction 
616: potential energy 
617: given by $U_{\vc{0}}n_{\vc{k_s}}^2/(2V)$. 
618: Moreover, according to the RPA, 
619: the presence of fluctuation 
620: density changes the potential energy by 
621: the amount $\delta H_{int}=
622: \sum_{\vc{q}}
623: \rho_{\vc{q}}^\dagger 
624: e^{-i\omega t}\delta \phi(\vc{q},\omega) + c.c.$
625: where $\delta \phi(\vc{q},\omega)=(U_{\vc{q}}/V)
626: \langle \delta \rho_{\vc{q}} \rangle$ 
627: is the
628: potential induced by the presence of 
629: the density fluctuations.  
630: Consequently, the global response of the 
631: system becomes:
632: \begin{eqnarray}\label{resp2}
633: \delta n(\vc{q},\omega)
634: &=&\chi_0 (\vc{q},\omega)
635: \left(\phi_{ext}(\vc{q},\omega)+\delta \phi(\vc{q},\omega)\right)
636: \nonumber \\
637: &=&\chi_0 (\vc{q},\omega)
638: \left(\phi_{ext}(\vc{q},\omega)+(U_{\vc{q}}/V)\delta n(\vc{q},\omega)\right)
639: \end{eqnarray}
640: Comparison between $(\ref{resp1})$ and 
641: $(\ref{resp2})$ allows to deduce:
642: \begin{eqnarray}\label{susc}
643: \chi (\vc{q},\omega)=\frac{\chi_0 (\vc{q},\omega)}
644: {1- (U_{\vc{q}}/V)\chi_0 (\vc{q},\omega)}
645: \end{eqnarray}
646: The total potential created inside the system 
647: $\phi_{tot}(\vc{q},\omega)=
648: \phi_{ext}(\vc{q},\omega)+\delta \phi(\vc{q},\omega)$
649: is related to the external potential 
650: through the dynamical dielectric function:
651: \begin{eqnarray}\label{heur}
652: {\tilde {\cal K}}(\vc{q},\omega)=
653: \frac{\phi_{ext}(\vc{q},\omega)}{
654: \phi_{tot}(\vc{q},\omega)}=
655: 1 - \frac{U_{\vc{q}}}{V}\chi_0 (\vc{q},\omega)
656: %\frac{n_{\vc{k_s}}/V}{(\omega+ i0_+ -\frac{\vc{k_s}.\vc{q}}{m})^2-
657: %\left(\frac{\vc{q}^2}{2m}\right)^2}
658: \end{eqnarray}
659: In the electromagnetism language, the gradient of 
660: $\phi_{ext}(\vc{q},\omega)$ corresponds to the electrical 
661: displacement, 
662: while the gradient of $\phi_{tot}(\vc{q},\omega)$ corresponds 
663: to the electric field. 
664: These are respectively  the external and 
665: total field force acting on the particle. 
666: The dielectric function usually has the effect of attenuating 
667: the external field force by means of an induced force so that 
668: the total field force is smoothed out. In particular,
669: we notice that this function is infinitely 
670: resonant for frequencies $\omega=\omega_{\vc{k_s},\vc{q}} , 
671: -\omega_{\vc{k_s}-\vc{q},\vc{q}}$ 
672: which means that the induced potential exactly 
673: compensates the external one. For these frequencies, 
674: the external potential affects the density fluctuations but 
675: has no local influence anymore on the condensate particle itself. 
676: This important observation is at the origin of the collision 
677: blockade phenomenon.
678: 
679: 
680: 
681: Indeed, assume that the source of 
682: this finite external potential results from 
683: the transition of an excited particle 
684: of momentum $\vc{k'}$ towards another 
685: excited state with a momentum 
686: $\vc{k'}-\vc{q}$. The  
687: released energy during this process is 
688: $\omega=-\omega_{\vc{k'},-\vc{q}}$.
689: According to the Fermi golden rule, a collision 
690: occurs between the condensed and non 
691: condensed ingoing particles if the transfer energy  
692: are equal $-\omega_{\vc{k'},-\vc{q}}=\omega_{\vc{k_s},\vc{q}}$ 
693: or equivalently the total energy is conserved  
694: $\epsilon_{\vc{k_s}}+\epsilon_{\vc{k'}}=
695: \epsilon_{\vc{k_s}+\vc{q}}+\epsilon_{\vc{k'}-\vc{q}}$. 
696: But then, the dielectric function 
697: becomes infinitely resonant ${\tilde {\cal K}}(\vc{q},\omega) 
698: \rightarrow \infty$ causing  
699: the total potential 
700: $\phi_{tot}(\vc{q},\omega) \rightarrow 0$. Thus 
701: the interaction potential felt by the condensate 
702: particle becomes non-existent, since the induced 
703: mean field cancels the external potential generated 
704: by the excited particle. 
705: That precise 
706: phenomenon is responsible for the 
707: absence of an effective  scattering process since, 
708: without interaction potential, no scattering amplitude 
709: can appear. The same effect happens when the scattering  
710: involves an outgoing particle in the condensate whereas, 
711: in such a case, the transfer energy  
712: $-\omega_{\vc{k_s-q},\vc{q}}$ causes the infinite resonance.
713: Thus, these giant resonances have the effect to protect 
714: the condensate particle from scattering with the others.
715: We must notice that this mechanism works only 
716: if the condensate population is macroscopic, 
717: otherwise, no induced potential is generated  
718: in the thermodynamic limit since, for 
719: $n_{\vc{k_s}}/V \rightarrow 0$, ${\tilde {\cal K}}(\vc{q},\omega)
720: \rightarrow 1$.
721: As we shall see in the next sections, 
722: a more elaborated model confirms this 
723: prediction in the more general case 
724: of a non equilibrium Bose gas, in which 
725: the condensate can be strongly 
726: depleted.  
727:  
728: 
729: 
730: \section{Kinetic theory in the RPA}
731: 
732: Many methods have been developed to derive kinetic equations 
733: for a dilute Bose gas
734: \cite{Walser,Imamovic}.
735: In order to arrive rapidly to the final result , instead of using
736: complicated many body techniques, we base the derivation on an
737: operatorial method developed in \cite{NP} which appeared to be
738: a simpler way to reach results without loss of generality.
739: The approximations are the followings:
740: 1) homogeneous Bose gas, 2) thermodynamic limit, 3) generalized RPA,
741: 4) instantaneous  collisions (Markovian QKE), 5) no fragmentation
742: of the condensate (only one macroscopic population mode).
743: 
744: \subsection{The random phase approximation}
745: 
746: In this subsection, we give a brief overview 
747: of the RPA developed in \cite{NP}.
748: This approximation has been used quite extensively 
749: for the quantum electron liquid for describing 
750: the screening effect. 
751: 
752: The dynamic of $\rho_{\vc{k},\vc{q}}$ is given by 
753: the Heisenberg equation motion. Using the relation: 
754: \begin{eqnarray}
755: [\rho_{\vc{k},\vc{q}},\rho_{\vc{k'},\vc{q'}}]
756: =\delta_{\vc{k}+\vc{q},\vc{k'}}
757: \rho_{\vc{k},\vc{k'}+\vc{q'}-\vc{k}}
758: -\delta_{\vc{k}-\vc{q'},\vc{k'}}
759: \rho_{\vc{k'},\vc{k}+\vc{q}-\vc{k'}}
760: \end{eqnarray}
761: we find
762: \begin{eqnarray}\label{he}
763: i\frac{\partial}{\partial t}\rho_{\vc{k},\vc{q}}
764: =[\rho_{\vc{k},\vc{q}},H]
765: =(\epsilon_{\vc{k}+\vc{q}}
766: -\epsilon_{\vc{k}})
767: \rho_{\vc{k},\vc{q}}
768: \nonumber \\
769: +\sum_{\vc{q'}}\frac{U_{\vc{q'}}}{2V}
770: [\rho_{\vc{k},\vc{q}+\vc{q'}}
771: -\rho_{\vc{k}-\vc{q'},\vc{q}+\vc{q'}},
772: \rho^\dagger_{\vc{q'}}]_+
773: \end{eqnarray}
774: where the brackets refer to 
775: to an anti-commutator.
776: Two cases in the RPA are generally considered 
777: in systems close to equilibrium \cite{NP}. 
778: The simple RPA, in which only the Hartree or direct terms 
779: contribute, and the generalized RPA, in which the Fock 
780: or exchange terms are also retained.
781: We shall concentrate on the second approximation, since for 
782: a contact potential Hartree and Fock terms are identical.
783: Only for the condensate-condensate interaction, 
784: the Fock term does not appear. In the GRPA, 
785: we consider that operator $\rho_{\vc{k},\vc{q}}$
786: with non zero momentum transfer $\vc{q}\not=0$ 
787: gives a negligible contribution in comparison to 
788: $\hat n_{\vc{k}}$, as it involves different modes oscillating 
789: with a random relative phase.  
790: 
791: 
792: The procedure to get the non equilibrium RPA equations is 
793: as follows. For a momentum transfer $\vc{q}=0$ the Eq.(\ref{he}) 
794: is kept unchanged. For $\vc{q}\not=0$, however, we keep among 
795: all terms those combinations of creation and annihilation 
796: operators involving product of $\rho_{\vc{k'},\vc{q}}$ 
797: and $\hat n_{\vc{k''}}$ for all possible values of $\vc{k'}$ 
798: and $\vc{k''}$, 
799: and neglect those combinations that cannot be written in this form.
800: These removed contributions are quadratic in 
801: the operator  
802: $\rho_{\vc{k'},\vc{q'}}$   
803: with $\vc{q'}\not=\vc{q},\vc{0}$. In this approximation, only 
804: contributions conserving the momentum transfer 
805: are relevant, the others coupling the various 
806: $\rho_{\vc{k},\vc{q}}$ with different 
807: momentum transfers are neglected. We expect that 
808: the 
809: excitations of momentum $\vc{q}$ propagate, without 
810: interfering with the others, over an 
811: enough long time determined by the dilution of the gas, 
812: that  (this would
813: correspond to the far field limit). 
814: The result is for $\vc{q}=0$
815: \begin{eqnarray}\label{n}
816: i\frac{\partial}{\partial t}\hat n_{\vc{k}}
817: =\sum_{\vc{q'}}\frac{U_{\vc{q'}}}{2V}
818: [\rho_{\vc{k},\vc{q'}}
819: -\rho_{\vc{k}-\vc{q'},\vc{q'}},
820: \rho^\dagger_{\vc{q'}}]_+
821: \end{eqnarray}
822: and for $\vc{q}\not=0$
823: \begin{eqnarray}\label{rho}
824: i\frac{\partial}{\partial t}\rho_{\vc{k},\vc{q}}
825: &=&
826: (\epsilon_{\vc{k}+\vc{q}} -\epsilon_{\vc{k}})
827: \rho_{\vc{k},\vc{q}}
828: \nonumber \\
829: &+&\sum_{\vc{k'}}[\frac{U_{\vc{k}-\vc{k'}+\vc{q}}}
830: {2V}\hat n_{\vc{k'}}- \frac{U_{\vc{k}-\vc{k'}-\vc{q}}}
831: {2V}\hat n_{\vc{k'+q}},\rho_{\vc{k},\vc{q}}]_+
832: \nonumber \\
833: &+&\sum_{\vc{k'}\not= \vc{k} }\frac{U_{\vc{q}}}{2V}
834: [\hat n_{\vc{k}}- \hat n_{\vc{k+q}},\rho_{\vc{k'},\vc{q}}]_+
835: \nonumber \\
836: &+&\sum_{\vc{k'}\not=\vc{k}-\vc{q}}\frac{U_{\vc{k}-\vc{k'}}}
837: {2V}[\hat n_{\vc{k}},\rho_{\vc{k'},\vc{q}}]_+
838: \nonumber \\
839: &-&\sum_{\vc{k'}\not=\vc{k}+\vc{q}}\frac{U_{\vc{k}-\vc{k'}}}
840: {2V}[\hat n_{\vc{k}+\vc{q}},\rho_{\vc{k'},\vc{q}}]_+
841: \end{eqnarray}
842: Eq.(\ref{rho}) is an integral operatorial equation 
843: linear in $\rho_{\vc{k},\vc{q}}$. We can linearize 
844: this equation by averaging all possible bilinear 
845: contributions. Since the homogeneity of the gas imposes
846: $\langle \rho_{\vc{k},\vc{q}} \rangle =0$ if 
847: $\vc{q}\not= \vc{0}$, we are left with 
848: the average on the number operator 
849: $\langle \hat n_{\vc{k}} \rangle= n_{\vc{k}}$ 
850: and we obtain 
851: an integral equation which possesses the same structure as Eq.(5.183)
852: p. 318 of \cite{NP}:
853: \begin{eqnarray}\label{rho3}
854: \left[i\frac{\partial}{\partial t}
855: -(\epsilon^{HFA}_{\vc{k}+\vc{q}}-\epsilon^{HFA}_{\vc{k}})\right]
856: \rho_{\vc{k},\vc{q}}=
857: (n_{\vc{k}}-n_{\vc{k+q}})
858: \sum_{\vc{k'}\not=\vc{k}}\frac{U_{\vc{q}}}{V}\rho_{\vc{k'},\vc{q}}
859: \nonumber \\
860: +n_{\vc{k}}
861: \sum_{\vc{k'}\not=\vc{k}-\vc{q}}\frac{U_{\vc{k}-\vc{k'}}}
862: {V}
863: \rho_{\vc{k'},\vc{q}}
864: -n_{\vc{k+q}}\sum_{\vc{k'}\not=\vc{k}+\vc{q}}\frac{U_{\vc{k}-\vc{k'}}}
865: {V}\rho_{\vc{k'},\vc{q}}
866: \nonumber \\
867: \end{eqnarray}
868: Indeed, in the term containing the bracket, 
869: we recognize the difference 
870: $\omega_{\vc{k},\vc{q}}=
871: \epsilon^{HFA}_{\vc{k}+\vc{q}}-
872: \epsilon^{HFA}_{\vc{k}}$
873: between the quasi-particle 
874: energy calculated in Hartree-Fock approximation (HFA): 
875: \begin{eqnarray}
876: \epsilon^{HFA}_{\vc{k}}=\frac{\vc{k}^2}{2m} 
877: + \sum_{\vc{k'}}\frac{U_{\vc{0}}}{V}n_{\vc{k'}}+
878: \sum_{\vc{k'}}\frac{U_{\vc{k}-\vc{k'}}}{V}n_{\vc{k'}}
879: \end{eqnarray}
880: On the other hand, in the integral terms, 
881: both Hartree and Fock 
882: terms are present. However, Eq.(\ref{rho3}) shows some differences: 
883: firstly, since we are dealing with bosons, the Fock terms have the opposite 
884: sign compared to fermions; secondly, since a macroscopic condensate 
885: might appear, we should be cautious not to count  
886: twice terms between the same modes. 
887: We have avoided this difficulty by  
888: excluding carefully in the sum over the modes those leading to a 
889: double counting and which appear only in the integral terms.
890: Finally, another difference is that $n_{\vc{k}}$ is 
891: still a function depending on the time, 
892: generalizing in this way the RPA approach for systems in 
893: non equilibrium. 
894: 
895: For comparison, Zaremba et al. \cite{Zaremba} 
896: have derived a QKE in an approximation which would correspond 
897: to a different equation for $\rho_{\vc{k},\vc{q}}$. As we shall 
898: see below, their analysis is equivalent to removing the 
899: integral terms in (\ref{rho3}) and to  
900: approximating the quasi-particle energy by 
901: \begin{eqnarray}\label{ZNG}
902: \epsilon^{ZNG}_{\vc{k}}=\frac{\vc{k}^2}{2m}
903: + \sum_{\vc{k'}}\frac{U_{\vc{0}}}{V}n_{\vc{k'}}+
904: \sum_{\vc{k'}\not=\vc{k}}\frac{U_{\vc{k}-\vc{k'}}}{V}n_{\vc{k'}}
905: \end{eqnarray}
906: We notice immediately that they have avoided a double 
907: counting in the exchange term. As a consequence, 
908: in the particular case of 
909: a contact potential and for $\vc{k_s}=0$, 
910: the energy difference that they obtained 
911: has a gap:
912: \begin{eqnarray}
913: \epsilon^{ZNG}_{\vc{k}}-\epsilon^{ZNG}_{\vc{0}}=
914: \frac{\vc{k}^2}{2m}+\g \frac{n_{\vc{0}}}{V}
915: \stackrel{\vc{k}\rightarrow 0}{\rightarrow}\g \frac{n_{\vc{0}}}{V}
916: \end{eqnarray}
917: Clearly, for $\vc{k} \rightarrow 0$ and if the 
918: mode $\vc{k_s}=0$ is macroscopically populated, 
919: this finite gap cannot be neglected in the 
920: scattering energy spectrum. 
921: In our approach, however, this gap does not appear 
922: anymore in $\epsilon^{HFA}_{\vc{k}}$.
923: 
924: The two operatorial equations (\ref{n}) and 
925: (\ref{rho3}) are the equations leading to the QKE.
926: 
927: \subsection{Collective and scattering excitations}
928: 
929: It is instructive to analyze the frequency 
930: spectrum solution of the eigenvalue problem 
931: given by Eq.(\ref{rho3}). In an electron liquid, 
932: the linear equation possesses two kinds of eigenvectors: 
933: (i) the scattering solutions involving the presence 
934: of only one particle and (ii) the collective 
935: solutions involving the presence of many particles.
936: 
937: For reasons we will explain below, let us concentrate on 
938: the problem of calculating the impulsion response or 
939: dielectric propagator 
940: ${\cal U}_\vc{q}(\vc{k},\vc{k_1},t)$ to an initial 
941: particle having a momentum $\vc{k_1}$. 
942: We replace the operator $\rho_{\vc{k},\vc{q}}$ by this 
943: function and we substitute the interaction potential 
944: by its first order expression in (\ref{U0}). 
945: After simplifications, the integral 
946: equation for this function is: 
947: \begin{eqnarray}\label{U}
948: [i\frac{\partial}{\partial t}
949: -
950: \frac{\vc{k}.\vc{q}}{m}-\frac{\vc{q}^2}{2m}]
951: {\cal U}_\vc{q}(\vc{k},\vc{k_1},t)
952: \nonumber \\
953: =\g \big[\frac{(n_{\vc{k}}-n_{\vc{k+q}})}{V}
954: \sum_{\vc{k'}\not= \vc{k}}
955: {\cal U}_\vc{q}(\vc{k'},\vc{k_1},t)
956: \nonumber \\
957: +\frac{n_{\vc{k}}}{V}
958: \sum_{\vc{k'}\not=\vc{k}-\vc{q}}
959: {\cal U}_\vc{q}(\vc{k'},\vc{k_1},t)
960: -\frac{n_{\vc{k+q}}}{V}\sum_{\vc{k'}\not=\vc{k}+\vc{q}}
961: {\cal U}_\vc{q}(\vc{k'},\vc{k_1},t)\big]
962: \end{eqnarray} 
963: with the initial condition:
964: \begin{eqnarray}
965: {\cal U}_\vc{q}(\vc{k},\vc{k_1},t=0)=
966: \delta_{\vc{k},\vc{k_1}}
967: \end{eqnarray}
968: 
969: If $n_{\vc{k}}$ varies slowly in time and thus 
970: can be considered as constant during the evolution 
971: of the impulsion response, Eq.(\ref{rho3}) can be solved 
972: exactly in the thermodynamic limit. Similar equations have been solved 
973: in the field of plasma physics \cite{Ichimaru}. 
974: For this purpose,  we define 
975: the Laplace transform as: 
976: \begin{eqnarray}
977: {\cal U}_\vc{q}(\vc{k},\vc{k_1},\omega)
978: =\int_0^\infty \!\!\!dt\, e^{i(\omega+i0_+)t}{\cal U}_\vc{q}(\vc{k},\vc{k_1},t)
979: \end{eqnarray}
980: A $0_+$ has been added in order to ensure convergence of 
981: the integral. 
982: If we suppose that $\vc{k_s}$
983: is the wave-vector for the superfluid mode,  
984: then we can make the decomposition between a normal 
985: component and components involving the superfluid mode:
986: \begin{eqnarray}
987: {\cal U}_\vc{q}(\vc{k},\vc{k_1},\omega)=
988: {\tilde {\cal U}}_\vc{q}(\vc{k},\vc{k_1},\omega) 
989: +\delta_{\vc{k},\vc{k_s}}
990: {\cal U}_\vc{q}(\vc{k_s},\vc{k_1},\omega)
991: \nonumber \\
992: +\delta_{\vc{k},\vc{k_s}-\vc{q}}
993: {\cal U}_\vc{q}(\vc{k_s}-\vc{q},\vc{k_1},\omega)
994: \end{eqnarray}
995: Also, we distinguish the normal mode population 
996: $n'_{\vc{k}}=
997: (1-\delta_{\vc{k},\vc{k_s}})n_{\vc{k}}$ from the 
998: condensed mode.
999: Plugging this decomposition into (\ref{rho3}) 
1000: and neglecting some $n'_{\vc{k}}$ by taking the 
1001: thermodynamic limit,
1002: we obtain for the superfluid modes:
1003: \begin{eqnarray}\label{col1}
1004: [\omega+i0_+ -
1005: \frac{\vc{k_s}.\vc{q}}{m}-\frac{\vc{q}^2}{2m}]
1006: {\cal U}_\vc{q}(\vc{k_s},\vc{k_1},\omega)
1007: =i\delta_{\vc{k_s},\vc{k_1}} 
1008: \nonumber \\
1009: +\g \frac{n_{\vc{k_s}}}{V}
1010: (\sum'_{\vc{k'}} 2\tilde{{\cal U}}_\vc{q}(\vc{k'},\vc{k_1},\omega)
1011: +
1012: {\cal U}_\vc{q}(\vc{k_s},\vc{k_1},\omega)
1013: \nonumber\\
1014: +
1015: {\cal U}_\vc{q}(\vc{k_s}-\vc{q},\vc{k_1},\omega))
1016: \end{eqnarray}
1017: \begin{eqnarray}\label{col2}
1018: [\omega+i0_+ -
1019: \frac{\vc{k_s}.\vc{q}}{m}+\frac{\vc{q}^2}{2m}]
1020: {\cal U}_\vc{q}(\vc{k_s}-\vc{q},\vc{k_1},\omega)
1021: =i\delta_{\vc{k_s}-\vc{q},\vc{k_1}}
1022: \nonumber \\
1023: - \g \frac{n_{\vc{k_s}}}{V}
1024: (\sum'_{\vc{k'}} 2\tilde{{\cal U}}_\vc{q}(\vc{k'},\vc{k_1},\omega)
1025: +
1026: {\cal U}_\vc{q}(\vc{k_s},\vc{k_1},\omega)
1027: \nonumber \\
1028: +
1029: {\cal U}_\vc{q}(\vc{k_s}-\vc{q},\vc{k_1},\omega))
1030: \end{eqnarray}
1031: and for the normal component:
1032: \begin{eqnarray}\label{col3}
1033: [\omega+i0_+ -
1034: \frac{\vc{k}.\vc{q}}{m}-\frac{\vc{q}^2}{2m}]
1035: {\cal U}_\vc{q}(\vc{k_s}-\vc{q},\vc{k_1},\omega)
1036: =i\delta_{\vc{k},\vc{k_1}}
1037: \nonumber \\
1038: + \frac{8\pi a}{m} \frac{(n_{\vc{k}}-n_{\vc{k+q}})}{V}
1039: (\sum'_{\vc{k'}} \tilde{{\cal U}}_\vc{q}(\vc{k'},\vc{k_1},\omega)
1040: +
1041: {\cal U}_\vc{q}(\vc{k_s},\vc{k_1},\omega)
1042: \nonumber \\
1043: +
1044: {\cal U}_\vc{q}(\vc{k_s}-\vc{q},\vc{k_1},\omega))
1045: \end{eqnarray}
1046: The prime in the sum excludes terms involving 
1047: the condensate mode. 
1048: This close set of equations is solved in the Appendix A. 
1049: For $\vc{k}\not= \vc{k_s}, \vc{k_s}-\vc{q}$, the scattering 
1050: solutions are $\omega= \epsilon_{\vc{k}+\vc{q}}-\epsilon_{\vc{k}}$.
1051: For excitations involving a superfluid mode 
1052: $\vc{k}= \vc{k_s}, \vc{k_s}-\vc{q}$, the presence of 
1053: interaction with the macroscopic condensate transforms 
1054: the scattering solutions 
1055: into  collective solutions of the 
1056: discriminant equation:
1057: \begin{eqnarray}\label{disprel}
1058: \Delta(\vc{q},\omega)=
1059: {\cal K}_n(\vc{q},\omega)[(\omega+i0_+ -
1060: \frac{\vc{k_s}.\vc{q}}{m})^2 - {\epsilon^B_{\vc{q}}}^2]
1061: \nonumber \\+
1062: ({\cal K}_n(\vc{q},\omega)-1)\frac{8\pi a n_{\vc{k_s}}}{mV}
1063: \frac{\vc{q}^2}{m}=0
1064: \end{eqnarray}
1065: where 
1066: \begin{eqnarray}
1067: \epsilon^B_{\vc{q}}=
1068: \sqrt{c^2 \vc{q}^2 +
1069: (\frac{\vc{q}^2}{2m})^2}
1070: \end{eqnarray}
1071: is the Bogoliubov excitation energy,
1072: \begin{eqnarray}
1073: c=\sqrt{\frac{4\pi a n_{\vc{k_s}}}{m^2 V}} 
1074: \end{eqnarray}
1075: is the sound velocity
1076: and 
1077: \begin{eqnarray}\label{kn}
1078: {\cal K}_n(\vc{q},\omega)=
1079: 1- \frac{8\pi a }{mV}\sum_{\vc{k}}
1080: \frac{n'_{\vc{k}}-n'_{\vc{k+q}}}{
1081: \omega+i0_+ -
1082: \frac{\vc{k}.\vc{q}}{m}-\frac{\vc{q}^2}{2m}}
1083: \end{eqnarray}
1084: is the dynamic dielectric function of the normal 
1085: fluid. 
1086: Eq.(\ref{disprel}) allows to find a 
1087: dispersion relation for any 
1088: function $n_{\vc{k}}$ possibly in non thermodynamic 
1089: equilibrium. 
1090: In this sense,
1091: it generalizes the dispersion relation 
1092: that is obtained equivalently from 
1093: the density fluctuations response formalism for 
1094: a gas at equilibrium \cite{SK,PN,GriffinB,Minguzzi,Giorgini}. 
1095: Note that in \cite{Giorgini} 
1096: the non number conserving 
1097: Beliaev formalism has been used to derive the dispersion 
1098: relation up to the next order beyond the Bogoliubov theory.
1099: The solution 
1100: can be put in the complex form: 
1101: $\omega=\omega_{\vc{q}}-i\gamma_{\vc{q}}$ 
1102: where $\gamma_{\vc{q}}$ corresponds to the Landau damping.
1103: For the particular 
1104: case of a weakly depleted Bose gas, we can solve 
1105: analytically Eq.(\ref{disprel}). In that case, we 
1106: can approximate \cite{SK}:
1107: \begin{eqnarray} \label{Knapp}
1108: {\cal K}_n(\vc{q},\omega) \simeq 
1109: 1+i{\rm Im} {\cal K}_n(\vc{q},\omega)
1110: \end{eqnarray}
1111: Eq.(\ref{disprel}) is obtained considering 
1112: in a first approximation that the imaginary 
1113: term can be neglected. We find that the real 
1114: part corresponds to the Bogoliubov 
1115: spectrum:
1116: \begin{eqnarray}\label{reo}
1117: \omega_{\vc{q}} \simeq \frac{\vc{k_s}.\vc{q}}{m} 
1118: \pm \epsilon^B_{\vc{q}}
1119: \end{eqnarray}
1120: The imaginary part corresponds to the 
1121: Landau damping and can be calculated 
1122: perturbatively assuming 
1123: $|\gamma_{\vc{q}}| \ll \omega_{\vc{q}}$ which is the 
1124: case for a weakly depleted condensate. We find, up to the 
1125: first order, 
1126: \begin{eqnarray}\label{imo}
1127: \gamma_{\vc{q}} \simeq 
1128: {\rm Im}{\cal K}_n(\vc{q},\omega_{\vc{q}})
1129: \frac{\frac{4\pi a n_{\vc{k_s}}}{mV}
1130: \frac{\vc{q}^2}{m}}{\omega_{\vc{q}}-\frac{\vc{k_s}.\vc{q}}{m}}
1131: \end{eqnarray}
1132: By analogy with plasma physics, we say that a Bose gas 
1133: is stable provided that $\gamma_{\vc{q}} \geq 0$ and unstable 
1134: otherwise. In a stable condensate the collective oscillations 
1135: are damped, while in a unstable condensate they are amplified 
1136: exponentially.  
1137: 
1138: For the case of thermal equilibrium (\ref{neq})  
1139: with $\vc{v_n}=0$, $\mu=0$ and a temperature close to zero, 
1140: ${\rm Im}{\cal K}_n(\vc{q},\omega_{\vc{q}})$ 
1141: is calculated in appendix B and is a positive function for 
1142: $\omega_{\vc{q}} \geq 0$ and negative otherwise. Thus, from 
1143: (\ref{imo}) the stability condition could be written as:
1144: \begin{eqnarray}\label{stab}
1145: \epsilon^B_{\vc{q}}-\frac{\vc{k_s}.\vc{q}}{m} \geq 0
1146: \end{eqnarray}
1147: We can check that this inequality is fulfilled for all values of 
1148: the momentum at the condition that 
1149: $|\vc{k_s}/m| \leq c$. In other words, the condensate is 
1150: stable if its velocity relative to the normal fluid is 
1151: much less than the sound velocity. Anticipating the next 
1152: subsections, this condition corresponds to the Landau 
1153: criterion for superfluidity for a weakly depleted Bose gas. 
1154: More generally, the occurrence of instability depends on 
1155: the form of $n_\vc{k}$ which influences the spectrum 
1156: obtained from (\ref{disprel}). In the case of a plasma, 
1157: this problem has been studied long time ago 
1158: \cite{instB,FR}.
1159:   
1160: Finally, in the absence of a macroscopic condensate 
1161: i.e. $n_\vc{k_s}/V \rightarrow 0$, 
1162: we recover 
1163: the two scattering solutions for the superfluid:
1164: \begin{eqnarray}
1165: \omega
1166: \stackrel{n_{\vc{k_s}}\rightarrow 0}{=}
1167: \frac{\vc{k_s}.\vc{q}}{m}\pm\frac{\vc{q}^2}{2m}
1168: \end{eqnarray}
1169: and the collective solution is given by 
1170: \begin{eqnarray}
1171: {\cal K}_n(\vc{q},\omega)=0
1172: \end{eqnarray}
1173: According to \cite{SK}, at equilibrium,
1174: this collective solution contains only an imaginary 
1175: part and thus no collective oscillation can be 
1176: observed in the 
1177: system. Therefore, any collective oscillation results 
1178: specifically from the condensation which transforms 
1179: the scattering solution of the condensed mode 
1180: into a collective solution. 
1181: 
1182: \subsection{Derivation of the kinetic equation}
1183: 
1184: From the results of the previous section, 
1185: we are now ready to derive a generalized Boltzmann like 
1186: equation for the Bose condensed gas. We  
1187: define the spatial correlation function as 
1188: the average $\langle 
1189: \rho_{\vc{k'},-\vc{q}}\rho_{\vc{k},\vc{q}} \rangle$. 
1190: Using Eq.(\ref{rho3}), we derive the 
1191: following equation for the correlation function:
1192: \begin{eqnarray}\label{K1}
1193: [i\frac{\partial}{\partial t}
1194: +
1195: \frac{(\vc{k'}-\vc{k}).\vc{q}}{m}-\frac{\vc{q}^2}{m}]
1196: \langle \rho_{\vc{k'},-\vc{q}} \rho_{\vc{k},\vc{q}}
1197: \rangle
1198: \nonumber \\
1199: =\frac{4\pi a}{mV} \sum_{\vc{k''}}\big[n_{\vc{k}}
1200: (2-\delta_{\vc{k},\vc{k''}}-\delta_{\vc{k}-\vc{q},\vc{k''}})
1201: \nonumber \\
1202: -n_{\vc{k+q}}
1203: (2-\delta_{\vc{k},\vc{k''}}-\delta_{\vc{k}+\vc{q},\vc{k''}})
1204: \big]
1205: \langle \rho_{\vc{k'},-\vc{q}} \rho_{\vc{k''},\vc{q}}
1206: \rangle
1207: \nonumber \\
1208: +\big[
1209: n_{\vc{k'}}
1210: (2-\delta_{\vc{k'},\vc{k''}}-\delta_{\vc{k'}+\vc{q},\vc{k''}})
1211: \nonumber \\
1212: -n_{\vc{k'-q}}
1213: (2-\delta_{\vc{k'},\vc{k''}}-\delta_{\vc{k'}-\vc{q},\vc{k''}})
1214: \big] \langle \rho_{\vc{k''},-\vc{q}}\rho_{\vc{k},\vc{q}}
1215: \rangle
1216: \end{eqnarray}
1217: 
1218: This function 
1219: can be decomposed as a non interacting part and an 
1220: interacting part \cite{Balescu}:
1221: \begin{eqnarray}\label{g}
1222: \langle \rho_{\vc{k'},-\vc{q}}
1223: \rho_{\vc{k},\vc{q}} \rangle=
1224: (n_{\vc{k}}+1)n_{\vc{k'}}
1225: \delta_{\vc{k'}-\vc{k},\vc{q}}
1226: \nonumber\\
1227: +n_{\vc{k}}n_{\vc{k'}}
1228: \delta_{\vc{q},0}(1-\delta_{\vc{k},\vc{k'}})
1229: -n_{\vc{k}}\delta_{\vc{k},\vc{k'}}\delta_{\vc{q},0}
1230: +g_\vc{q}(\vc{k},\vc{k'})
1231: \end{eqnarray}
1232: where $g_\vc{q}(\vc{k},\vc{k'})$ 
1233: represents the correlation function due 
1234: to the interactions. For $\vc{q} \not= 0$ or $\vc{k} \not= \vc{k'}$,
1235: the non interacting part follows the Wick's decomposition and 
1236: since the system is homogeneous $\langle c^\dagger_\vc{k} 
1237: c_\vc{k'} \rangle = \delta_{\vc{k},\vc{k'}} n_\vc{k}$. 
1238: For $\vc{q}= 0$ or $\vc{k}=\vc{k'}$, however, it  
1239: reduces to the quadratic average 
1240: population that we choose to be $\langle {\hat n}_\vc{k}^2 \rangle =
1241: n_\vc{k}^2$. 
1242: 
1243: Would we have used the Wick's decomposition in that 
1244: case then $\langle {\hat n}_\vc{k}^2 \rangle =
1245: 2 n_\vc{k}^2$ which corresponds to non zero particle number 
1246: fluctuations $\langle \delta^2 {\hat n}_\vc{k} \rangle =
1247: n_\vc{k}^2$ and the total particle number would display fluctuations 
1248: as well. Indeed, since 
1249: $\langle {\hat n}_\vc{k} {\hat n}_\vc{k'}\rangle= n_\vc{k}n_\vc{k'}$ 
1250: for $\vc{k}\not=\vc{k'}$, 
1251: we calculate $\langle \delta^2 {\hat N} \rangle=
1252: \sum_{\vc{k}} n_\vc{k}^2$. In the presence of condensation, 
1253: these fluctuations are huge i.e. of the same 
1254: order of the average value \cite{4eme}. 
1255: Consequently, in an isolated system where the 
1256: total particle number is conserved, this unphysical 
1257: situation must be excluded. 
1258: 
1259: Taking the average over each side of
1260: Eq.(\ref{n}), a comparison with (\ref{g}) allows to deduce:
1261: \begin{eqnarray}\label{n2}
1262: i\frac{\partial}{\partial t}n_{\vc{k}}
1263: =\sum_{\vc{q},\vc{k'}}\frac{2\pi a}{mV}
1264: \big(g_\vc{q}(\vc{k},\vc{k'})-
1265: g_\vc{q}(\vc{k}-\vc{q},\vc{k'})
1266: \nonumber \\
1267: +g_\vc{q}(\vc{k'},\vc{k})-
1268: g_\vc{q}(\vc{k'},\vc{k}+\vc{q})\big)
1269: \end{eqnarray}
1270: On the other hand,
1271: inserting this definition into Eq.(\ref{K1}) and using 
1272: (\ref{n2}), 
1273: we obtain 
1274: \begin{eqnarray}\label{K2}
1275: [i\frac{\partial}{\partial t}
1276: +
1277: \frac{(\vc{k'}-\vc{k}).\vc{q}}{m}-\frac{\vc{q}^2}{m}]
1278: g_\vc{q}(\vc{k},\vc{k'})
1279: \nonumber \\
1280: =Q_\vc{q}(\vc{k},\vc{k'})+
1281: \frac{4\pi a}{mV} \sum_{\vc{k''}}
1282:  \big[n_{\vc{k}}
1283: (2-\delta_{\vc{k},\vc{k''}}-\delta_{\vc{k}-\vc{q},\vc{k''}})
1284: \nonumber \\
1285: -n_{\vc{k+q}}
1286: (2-\delta_{\vc{k},\vc{k''}}-\delta_{\vc{k}+\vc{q},\vc{k''}})
1287: \big]g_\vc{q}(\vc{k''},\vc{k'})
1288: \nonumber \\
1289: +\big[n_{\vc{k'}}
1290: (2-\delta_{\vc{k'},\vc{k''}}-\delta_{\vc{k'}+\vc{q},\vc{k''}})
1291: \nonumber \\
1292: -n_{\vc{k'-q}}
1293: (2-\delta_{\vc{k'},\vc{k''}}-\delta_{\vc{k'}-\vc{q},\vc{k''}})
1294: \big]g_\vc{q}(\vc{k},\vc{k''})
1295: \end{eqnarray}
1296: where 
1297: \begin{eqnarray}\label{Q}
1298: Q_\vc{q}(\vc{k},\vc{k'})=
1299: \frac{8\pi a }{mV}
1300: \big[(n_{\vc{k}}-n_{\vc{k+q}})(n_{\vc{k'-q}}+1)n_{\vc{k'}}
1301: \nonumber \\
1302: +(n_{\vc{k'}}-n_{\vc{k'-q}})(n_{\vc{k}}+1)n_{\vc{k+q}}\big]
1303: \nonumber \\
1304: -\frac{4\pi a n_{\vc{k_s}}^2}{mV}(
1305: \delta_{\vc{k},\vc{k_s}}
1306: \delta_{\vc{k'},\vc{k_s}}
1307: -\delta_{\vc{k},\vc{k_s}-\vc{q}}
1308: \delta_{\vc{k'},\vc{k_s}+\vc{q}})
1309: \nonumber \\
1310: \left[(1+n_{\vc{k_s}})n_{\vc{k_s}+\vc{q}}+(1+n_{\vc{k_s}-\vc{q}})n_{\vc{k_s}}
1311: \right]
1312: \end{eqnarray}
1313: is the inhomogeneous term.
1314: To get (\ref{Q}), we eliminate terms involving delta 
1315: functions which will give negligible contribution to the QKE in the 
1316: thermodynamic limit. 
1317: Both  Eq.(\ref{n2}) and Eq.(\ref{K2}) form a close set in which 
1318: $g_\vc{q}(\vc{k},\vc{k'})$ must be eliminated 
1319: in order to get a kinetic equation for $n_{\vc{k}}$. 
1320: This elimination is done following an analog procedure 
1321: to that of Ichimaru \cite{Ichimaru}. 
1322: We assume that the  
1323: correlations due to the interactions are 
1324: non-existent for $t \rightarrow -\infty$. This requirement is usual 
1325: in  kinetic theory and allows to provide the 
1326: following initial condition
1327: $g_\vc{q}(\vc{k},\vc{k'})|_{t \rightarrow -\infty} =0$.
1328: Also, we assume that $n_{\vc{k}}$ evolves on a much 
1329: more long time scale than the duration of a collision 
1330: $g_\vc{q}(\vc{k},\vc{k'})$ and 
1331: so is considered as constant in solving Eq.(\ref{K2}). 
1332: This approximation amounts to claiming that the binary
1333: collision process is instantaneous in comparison
1334: with the time associated to the relaxation of the system.
1335: Inspired by the previous subsections and by \cite{Ichimaru},
1336: we can check that the solution, 
1337: satisfying both Eq.(\ref{K2}) and the initial condition,
1338: is expressed in terms of the dielectric propagator as:
1339: \begin{eqnarray}\label{K3}
1340: g_\vc{q}(\vc{k},\vc{k'},t)=
1341: -i\int_0^\infty \!\!\!\!\!dt' \sum_{\vc{k_1},\vc{k'_1}}
1342: \nonumber \\
1343: {\cal U}_{\vc{q}}(\vc{k},\vc{k_1},t')
1344: {\cal U}_{-\vc{q}}(\vc{k'},\vc{k'_1},t')
1345: Q_\vc{q}(\vc{k_1},\vc{k'_1},t-t')
1346: \end{eqnarray}
1347: We have inserted the explicit time dependence in the functions.
1348: If the creation of such correlations is 
1349: much faster in comparison with the relaxation time for $n_{\vc{k}}$, 
1350: $Q_\vc{q}(\vc{k_1},\vc{k'_1},t-t') \simeq 
1351: Q_\vc{q}(\vc{k_1},\vc{k'_1},t)$ and we obtain a 
1352: Markovian equation. 
1353: We re express the dielectric 
1354: propagator in Fourier transform according to, 
1355: \begin{eqnarray}
1356: {\cal U}_\vc{q}(\vc{k},\vc{k_1},t)
1357: =\frac{1}{2\pi}
1358: \int_{\cal C} d\omega e^{-i\omega t}
1359: {\cal U}_\vc{q}(\vc{k},\vc{k_1},\omega)
1360: \end{eqnarray}
1361: where the contour ${\cal C}$ extends from $-\infty$ to 
1362: $+\infty$ along a path in the upper half of the $\omega$ plane 
1363: in such a way that all the singularities lie below it. 
1364: The substitution allows to carry out successively integrations 
1365: over $t'$ and $\omega'$ by closing the contour in the 
1366: upper half plane  in order to get:
1367: \begin{eqnarray}\label{K4}
1368: g_\vc{q}(\vc{k},\vc{k'})=
1369: \int_{-\infty}^\infty \!\!\!\frac{d\omega}{2\pi i}\sum_{\vc{k_1},\vc{k'_1}}
1370: \nonumber \\
1371: {\cal U}_{\vc{q}}(\vc{k},\vc{k_1},\omega)
1372: {\cal U}_{-\vc{q}}(\vc{k'},\vc{k'_1},-\omega)
1373: Q_\vc{q}(\vc{k_1},\vc{k'_1},t)
1374: \end{eqnarray}
1375: Calculations in appendix C allow to find an explicit 
1376: expression for $g_\vc{q}(\vc{k},\vc{k'})$ in terms of 
1377: the one particle distribution function, provided  
1378: the Landau damping factor is always positive. 
1379: The substitution into 
1380: (\ref{n2}) allows finally to get the GRPA kinetic 
1381: equation for a stable Bose gas:
1382: \begin{eqnarray}\label{K5}
1383: \frac{\partial}{\partial t}n_{\vc{k}}=
1384: {\cal C}^T_{\vc{k}}[n_{\vc{k'}};\vc{k_s}]=
1385: {\cal C}_{\vc{k}}[n_{\vc{k'}};\vc{k_s}]+
1386: {\tilde {\cal C}}_{\vc{k}}[n_{\vc{k'}};\vc{k_s}]
1387: \end{eqnarray}
1388: where we define the collision terms as a functional of $n_{\vc{k'}}$
1389: and a function of $\vc{k_s}$. The first term describes the 
1390: collision rate between particles 
1391: of the normal fluid:
1392: \begin{widetext}
1393: \begin{eqnarray}\label{K6}
1394: {\cal C}_{\vc{k}}[n_{\vc{k'}};\vc{k_s}]
1395: =\sum_{\vc{q},\vc{k'}} \left|
1396: \frac{\frac{8\pi a}{mV}}
1397: {{\cal K}(\vc{q},\epsilon_{\vc{k}+\vc{q}}-\epsilon_{\vc{k}})}
1398: \right|^2 \!\!
1399: (1-\delta_{\vc{k},\vc{k_s}}-\delta_{\vc{k}+\vc{q},\vc{k_s}}-
1400: \delta_{\vc{k'},\vc{k_s}}-\delta_{\vc{k'}-\vc{q},\vc{k_s}})
1401: \pi \delta(\epsilon_{\vc{k}+\vc{q}}+\epsilon_{\vc{k'}-\vc{q}}-
1402: \epsilon_{\vc{k}}-\epsilon_{\vc{k'}})
1403: \nonumber \\
1404: \big[n_{\vc{k}+\vc{q}}n_{\vc{k'}-\vc{q}}
1405: (n_{\vc{k}}+1)(n_{\vc{k'}}+1)-
1406: n_{\vc{k}}n_{\vc{k'}}(n_{\vc{k}+\vc{q}}+1)(n_{\vc{k'}-\vc{q}}+1)\big]
1407: \end{eqnarray}
1408: The second term describes the collision rate between condensed
1409: and non condensed particles, the condensed particle is
1410: either the input or the output state in the scattering
1411: process:
1412: \begin{eqnarray}\label{K7}
1413: {\tilde{\cal C}}_{\vc{k}}[n_{\vc{k'}};\vc{k_s}]
1414: =&\displaystyle \sum_{\vc{q},\vc{k'}}
1415: \frac{\left(\frac{8\pi a}{mV}\right)^2}
1416: {\left|{\cal K}^*(\vc{q},\epsilon_{\vc{k}+\vc{q}}-\epsilon_{\vc{k}})
1417: {\tilde{\cal K}}(\vc{q},\epsilon_{\vc{k}+\vc{q}}-\epsilon_{\vc{k}})\right|}
1418: (\delta_{\vc{k},\vc{k_s}}+\delta_{\vc{k}+\vc{q},\vc{k_s}}+
1419: \delta_{\vc{k'},\vc{k_s}}+\delta_{\vc{k'}-\vc{q},\vc{k_s}})
1420: \nonumber \\
1421: &\pi \delta(\epsilon_{\vc{k}+\vc{q}}+\epsilon_{\vc{k'}-\vc{q}}-
1422: \epsilon_{\vc{k}}-\epsilon_{\vc{k'}})
1423: \big[n_{\vc{k}+\vc{q}}n_{\vc{k'}-\vc{q}}
1424: (n_{\vc{k}}+1)(n_{\vc{k'}}+1)-
1425: n_{\vc{k}}n_{\vc{k'}}(n_{\vc{k}+\vc{q}}+1)(n_{\vc{k'}-\vc{q}}+1)\big]
1426: \end{eqnarray}
1427: \end{widetext}
1428: In the first expression (\ref{K6}), the contact potential has 
1429: been replaced by an effective one depending 
1430: on the transfer particle energy 
1431: $\omega=\epsilon_{\vc{k}+\vc{q}}-\epsilon_{\vc{k}}$:
1432: \begin{eqnarray}\label{ueff}
1433: U^{eff}_{\vc{q}}(\omega)
1434: =\frac{\frac{8\pi a}{mV}}
1435: {{\cal K}(\vc{q},\omega)}
1436: \end{eqnarray}
1437: The correcting term is the dynamic dielectric constant:
1438: \begin{eqnarray}\label{kal}
1439: &&{\cal K}(\vc{q},\omega)
1440: \nonumber \\
1441: &=&
1442: \frac{\frac{8\pi a n_{\vc{k_s}}}{mV}
1443: \frac{\vc{q}^2}{m}}
1444: {(\omega+i0_+ -
1445: \frac{\vc{k_s}.\vc{q}}{m})^2-(\frac{\vc{q}^2}{2m})^2+
1446: \frac{4\pi a n_{\vc{k_s}}}{mV}
1447: \frac{\vc{q}^2}{m}}+{\cal K}_n(\vc{q},\omega)
1448: \nonumber \\
1449: &=&
1450: \frac{\Delta(\vc{q},\omega)}
1451: {(\omega+i0_+ -\frac{\vc{k_s}.\vc{q}}{m})^2-(\frac{\vc{q}^2}{2m})^2+
1452: \frac{4\pi a n_{\vc{k_s}}}{mV}
1453: \frac{\vc{q}^2}{m}}
1454: \end{eqnarray}
1455: In the second expression (\ref{K7}), 
1456: we define another dynamical dielectric function:
1457: \begin{eqnarray}\label{kalc}
1458: {\cal \tilde{K}}(\vc{q},\omega)
1459: =\frac{\Delta(\vc{q},\omega)}
1460: {(\omega+i0_+ -\frac{\vc{k_s}.\vc{q}}{m})^2-(\frac{\vc{q}^2}{2m})^2}
1461: \end{eqnarray}
1462: This result generalizes (\ref{heur}) in the case where 
1463: $n'_{\vc{k}} \not= 0$.
1464: Plugging this expression into Eq.(\ref{K7}), the energy 
1465: conservation imposes the two choices 
1466: $\omega=\frac{\vc{k_s}.\vc{q}}{m} \pm \frac{\vc{q}^2}{2m}$
1467: when a particle of the condensate participates to the collision 
1468: process. But, 
1469: with such an energy transfer and for a macroscopic population 
1470: of the condensate, this dynamic dielectric function gets 
1471: infinite: 
1472: \begin{eqnarray}\label{kalc2}
1473: {\cal \tilde{K}}(\vc{q},\omega)
1474: |_{\omega=\frac{\vc{k_s}.\vc{q}}{m} \pm \frac{\vc{q}^2}{2m}}
1475: \rightarrow \infty
1476: \end{eqnarray}
1477: leading to ${\tilde{\cal C}}_{\vc{k}}[n_{\vc{k'}};\vc{k_s}]
1478: \rightarrow 0$.
1479: In this situation, surprisingly, no collision involving the condensate particle
1480: occurs. As a consequence, in a uniform Bose gas, no transfer
1481: of particle is possible between the condensate and the
1482: normal component. Indeed, due to the
1483: absence of Fock interaction term in the GRPA, the effective
1484: potential has a different expression when a particle
1485: of the condensate is involved in a scattering.
1486: As said in section 2, a potential induced by the 
1487: condensate compensates 
1488: exactly the potential created by the excited particle susceptible 
1489: to scatter. 
1490: As a consequence, the dielectric function suppresses 
1491: the effectiveness of the  
1492: potential which thus becomes  
1493: invisible to the condensate and totally shelters it from collision.
1494: Furthermore, let us note that this 
1495: shielding remains unchanged for any momentum $\vc{k_s}$ 
1496: which preserves the stability of the condensate. If 
1497: the Landau damping factor $\gamma_{\vc{q}}$ becomes negative, 
1498: then an instability occurs and therefore Eq.(\ref{K6}) and Eq.(\ref{K7}) 
1499: are not longer valid, as well as all considerations concerning 
1500: collision blockade. 
1501: In that case, a more elaborated 
1502: derivation must be carried out that 
1503: could allow particle exchange with the condensate.
1504: The expressions (\ref{B2}) and 
1505: (\ref{B3}) must be recalculated taking into account the 
1506: instability \cite{instB,FR}. 
1507: 
1508: 
1509: In this way, the GRPA kinetic equation provides a 
1510: different understanding of
1511: the superfluidity phenomenon as it describes the motion of
1512: two independent fluids that cannot transfer particle through
1513: collisions. Since $\frac{\partial}{\partial t}n_{\vc{k_s}}=0$,
1514: $n_\vc{k_s}$ is really a constant of motion independent of the 
1515: dynamics of the normal fluid.
1516: 
1517:  
1518: The collision blockade is the consequence of a higher 
1519: order expansion in the interaction parameter. Since 
1520: ${\cal \tilde{K}}(\vc{q},\omega)|_{a \rightarrow 0}=1$, 
1521: up to the second order in the interaction,  
1522: we recover the UUQKE
1523: allowing collisions with the condensate.
1524: Only an infinite re-summation of appropriate higher order 
1525: contributions through GRPA generates collectively a perfect collision 
1526: blockade. Also, in the limit where the condensate is not 
1527: macroscopically  populated $n_{\vc{k_s}}/V \rightarrow 0$, 
1528: then ${\cal \tilde{K}}(\vc{q},\omega) \rightarrow 
1529: {\cal K}_n(\vc{q},\omega)$ and 
1530: ${\cal K}(\vc{q},\omega)\rightarrow
1531: {\cal K}_n(\vc{q},\omega)$ and collisions with 
1532: particle of the mode $\vc{k_s}$ become again possible.
1533: The Eq.(\ref{K5}) is a different 
1534: version of the kinetic equations formulated by
1535: Zaremba et al. \cite{Zaremba} in which  
1536: collisions are possible with the condensate particle. 
1537: Indeed, in their approach, the authors have done an 
1538: approximation which is equivalent to 
1539: replacing the energy $\epsilon_{\vc{k}}$ by (\ref{ZNG}) 
1540: and to  
1541: suppressing the 
1542: integral terms of the right hand side in (\ref{rho3}). 
1543: As a consequence, they obtained a QKE 
1544: in which
1545: the dielectric function
1546: does not appear and 
1547: with the 
1548: feature of having a gap in the transition of a 
1549: particle to the condensate.
1550: 
1551: 
1552: Let us notice that the same phenomenon occurs for the 
1553: simple RPA where the exchange terms have been omitted. 
1554: In the SRPA the calculation is similar to those
1555: leading to the quantum kinetic equation for a plasma.
1556: For details of such derivation, see Balescu \cite{QBalescu}.
1557: The only difference is that
1558: the coulombian potential becomes
1559: a contact potential and that the fermion now becomes
1560: a boson, meaning that any factor
1561: $1-n_{\vc{k}}$ now becomes $1+n_{\vc{k}}$ in the
1562: collision term.
1563: In that case, the dielectric function has the 
1564: same form for both the condensed and the normal modes \cite{SK}:
1565: \begin{eqnarray}\label{ksrpa}
1566: {\cal K}^{SRPA}(\vc{q},\omega)
1567: &=&1- \frac{4\pi a }{mV}\sum_{\vc{k}}
1568: \frac{n_{\vc{k}}-n_{\vc{k+q}}}{
1569: \omega+i0_+ -
1570: \frac{\vc{k}.\vc{q}}{m}-\frac{\vc{q}^2}{2m}}
1571: \nonumber \\
1572: &=&
1573: \frac{\frac{4\pi a n_{\vc{k_s}}}{mV}
1574: \frac{\vc{q}^2}{m}}
1575: {(\omega+i0_+ -
1576: \frac{\vc{k_s}.\vc{q}}{m})^2-(\frac{\vc{q}^2}{2m})^2}
1577: \nonumber \\
1578: &&+{\cal K}^{SRPA}_n(\vc{q},\omega)
1579: \end{eqnarray} 
1580: The normal dielectric component 
1581: ${\cal K}^{SRPA}_n(\vc{q},\omega)$ is of the 
1582: same form as Eq.(\ref{kn}), except that the exchange 
1583: interaction term 
1584: has been removed.
1585: Similarly, the particles of the condensate are protected from 
1586: any collision with the surrounding. Nevertheless, in the SRPA, 
1587: collisions involving 
1588: low level excited states are strongly prevented because the energy 
1589: transfer $\omega$ is close to the one corresponding to the 
1590: condensate. On the contrary, in the GRPA, 
1591: a supplementary term coming from the exchange effect 
1592: $\frac{4\pi a n_{\vc{k_s}}}{mV}$ ensures a reasonable value 
1593: for the effective potential guaranteeing an efficient collision rate, 
1594: even with particle in the energy levels close to the condensate level.  
1595: 
1596: 
1597: \subsection{Properties of the RPA collision term}
1598: 
1599: The collision term exhibits a number of remarkable 
1600: properties analog to those encountered by other kinetic 
1601: equations \cite{Ichimaru}. 
1602: These allow to establish the particle number, momentum and
1603: energy conservation laws as well as the Boltzmann H-theorem.
1604: Since the condensed particles 
1605: do not participate to the collision process,
1606: these concerns only the excited particles of the normal fluid. 
1607: Three of them 
1608: can be stated as:
1609: \begin{eqnarray}
1610: \sum'_{\vc{k}}{\cal C}_{\vc{k}}[n_{\vc{k'}};\vc{k_s}] =0
1611: \\
1612: \sum'_{\vc{k}} \vc{k} {\cal C}_{\vc{k}}[n_{\vc{k'}};\vc{k_s}]=0
1613: \\
1614: \sum'_{\vc{k}} \epsilon_{\vc{k}}{\cal C}_{\vc{k}}
1615: [n_{\vc{k'}};\vc{k_s}]=0
1616: \end{eqnarray}
1617: These three integral relations are checked easily by dividing 
1618: the collision term in four 
1619: equal parts.  After carrying out the successive changes of 
1620: the integration variables $\vc{k} \leftrightarrow \vc{k'}, 
1621: \, \vc{q} \leftrightarrow -\vc{q}$ on the second term,
1622: $\vc{k} \leftrightarrow \vc{k'}-\vc{q}, \, 
1623: \vc{k'} \leftrightarrow \vc{k}+ \vc{q}$ on the third term and 
1624: a successive combination of these two variable changes in the 
1625: fourth term, these four terms cancel 
1626: each other if we use the relation 
1627: ${\cal{K}}(\vc{q},\omega)={\cal{K}}^\star(-\vc{q},-\omega)$.
1628: As a consequence, for a uniform Bose gas, the total 
1629: particle number, the total momentum and the total kinetic energy, 
1630: associated 
1631: with the normal component are independent of the time i.e.
1632: $(d/dt)\sum_{\vc{k}} n'_{\vc{k}}=0$, 
1633: $(d/dt)\sum_{\vc{k}} \vc{k}n'_{\vc{k}}=0$ and 
1634: $(d/dt)\sum_{\vc{k}} \epsilon_{\vc{k}} n'_{\vc{k}}=0$.
1635: These properties together with the conservation of 
1636: $n_\vc{k_s}$ imply that the total energy in the 
1637: Hartree-Fock approximation is also conserved \cite{Zaremba} 
1638: i.e.:
1639: \begin{eqnarray}
1640: \frac{d}{dt}
1641: [\sum_{\vc{k}} \epsilon_{\vc{k}} n'_{\vc{k}} +
1642: \frac{4\pi a}{mV}
1643: (N^2 - \frac{1}{2} n^2_\vc{k_s})]=0 
1644: \end{eqnarray}  
1645: 
1646: Another crucial property is the Boltzmann H-theorem.
1647: The entropy, due to thermal excitations, has the expression:
1648: \begin{eqnarray}\label{S}
1649: S=\sum'_{\vc{k}}[(1+n'_{\vc{k}})\ln (1+n'_{\vc{k}})-
1650: n'_{\vc{k}}\ln n'_{\vc{k}}]
1651: \end{eqnarray}
1652: The time evolution of the entropy is always positive. A 
1653: similar derivation using the change of variables allows to 
1654: calculate a positive production of entropy:
1655: \begin{widetext}
1656: \begin{eqnarray}\label{H}
1657: \frac{dS}{dt}
1658: =\!\! \sum_{\vc{q},\vc{k},\vc{k'}}
1659: \left|
1660: \frac{\frac{8\pi a}{mV}}
1661: {{\cal K}(\vc{q},\epsilon_{\vc{k}+\vc{q}}-\epsilon_{\vc{k}})}
1662: \right|^2 \!\!
1663: \frac{\pi}{4}
1664: \delta(\epsilon_{\vc{k}+\vc{q}}+\epsilon_{\vc{k'}-\vc{q}}-
1665: \epsilon_{\vc{k}}-\epsilon_{\vc{k'}})
1666: \ln\left[\frac{n'_{\vc{k}+\vc{q}}n'_{\vc{k'}-\vc{q}}
1667: (n'_{\vc{k}}+1)(n'_{\vc{k'}}+1)}{
1668: n'_{\vc{k}}n'_{\vc{k'}}(n'_{\vc{k}+\vc{q}}+1)(n'_{\vc{k'}-\vc{q}}+1)}\right]
1669: \nonumber \\
1670: \left[n'_{\vc{k}+\vc{q}}n'_{\vc{k'}-\vc{q}}
1671: (n'_{\vc{k}}+1)(n'_{\vc{k'}}+1)-
1672: n'_{\vc{k}}n'_{\vc{k'}}(n'_{\vc{k}+\vc{q}}+1)(n'_{\vc{k'}-\vc{q}}+1)\right]
1673: \geq 0
1674: \end{eqnarray}
1675: \end{widetext}
1676: The positivity is a consequence of the mathematical relation 
1677: $\ln(x/y)(x-y)\geq 0$. The equality is achieved for $x=y$. 
1678: From Eq.(\ref{H}) we deduce that the entropy always increases 
1679: until the system reaches a stationary equilibrium distribution. 
1680: This occurs when the 
1681: production of entropy becomes zero. In that situation, 
1682: \begin{eqnarray}
1683: \frac{n'_{\vc{k}+\vc{q}}n'_{\vc{k'}-\vc{q}}}
1684: {(n'_{\vc{k}+\vc{q}}+1)(n'_{\vc{k'}-\vc{q}}+1)}
1685: =\frac{n'_{\vc{k}}n'_{\vc{k'}}}{(n'_{\vc{k}}+1)(n'_{\vc{k'}}+1)}
1686: \end{eqnarray}
1687: This relation holds only for the Bose-Einstein distribution 
1688: function \cite{Balescu}:
1689: \begin{eqnarray}\label{neq}
1690: n'_{\vc{k}}=n^{eq}_{\vc{k}}=
1691: \frac{1}{\exp{[\beta(\epsilon_\vc{k}-\vc{k}.\vc{v_n}-\mu)]}-1}
1692: \end{eqnarray}
1693: where $\vc{v_n}$ is the average velocity of the normal component. 
1694: In this way, the inverse temperature $\beta$ and the chemical 
1695: potential $\mu$ are defined as the free parameters of the equilibrium 
1696: solution. 
1697: Although similar, the properties of the thermodynamic 
1698: equilibrium predicted by the H-theorem in the RPA are really different 
1699: from the one predicted by the calculation of 
1700: the ensemble partition functions of an ideal Bose gas \cite{Huang}.
1701: The superfluid and normal components can move 
1702: relatively with two different velocities $\vc{v_s}$ and 
1703: $\vc{v_n}$.
1704: The relative difference 
1705: $\vc{v_s}-\vc{v_n}$ and $\mu$ are subject to the only constraint of 
1706: stability which gives a limitation. 
1707: As said in the previous subsection, for $\vc{v_n}=0$, 
1708: and 
1709: the temperature close to zero, and using (\ref{kn4}) in 
1710: (\ref{imo}), this constraint 
1711: is realized provided that $\mu=0$ and that 
1712: the relative velocity does not 
1713: exceed the sound velocity 
1714: (see Eq.(\ref{stab})). Otherwise, the condensate becomes 
1715: instable. Thus, we recover the Landau 
1716: criterion for the weakly interacting Bose gas. 
1717: Note that this result differs from the ZNG approach since 
1718: it corresponds to an equilibrium fugacity of the normal fluid 
1719: equal to one 
1720: (see Eq.(44) in \cite{Zaremba}). 
1721: 
1722: On the contrary, the statistical equilibrium ensemble formalism  
1723: for an ideal Bose gas imposes a zero relative velocity 
1724: and a chemical potential close to zero when condensation 
1725: occurs \cite{Huang}. 
1726: This contradiction might be explained if we remind that 
1727: in this formalism we postulate an equal {\it a priori}
1728: probability of any possible configuration of the gas \cite{Huang, Balescu}. 
1729: This basic assumption of equilibrium statistical 
1730: mechanics originates from the observation that, 
1731: over a long 
1732: time range, the collision process will mix statistically 
1733: these configurations. But, since in the GRPA kinetic equation the 
1734: collision is blocked, the equal {\it a priori}
1735: probability might not work as far as the condensate mode is concerned.
1736: This observation suggests that the condensed  
1737: particle population for an homogeneous systems 
1738: is not distributed anymore but has a well 
1739: defined value. 
1740: 
1741: \subsection{Condensate formation}
1742: 
1743: %In this context, 
1744: The kinetic equation (\ref{K5}) does not 
1745: explain the condensate formation through, for 
1746: example, evaporative cooling \cite{Keterlee2,BZS,Gardiner2}. 
1747: If we start indeed from an initially 
1748: condensed gas with a Bose-Einstein
1749: momentum distribution whose the tail has been cut, then 
1750: according to the RPA model, the irreversible
1751: evolution process towards equilibrium 
1752: will undergo a strange behavior. 
1753: Instead of populating the lowest energy level which is 
1754: forbidden, the excited particles will be transferred to the 
1755: lowest excited levels in which a macroscopic population 
1756: can eventually appear. In that situation, we are led 
1757: to a fragmentation of the condensate into at least two 
1758: energy levels. But such two macroscopic states contradict 
1759: the assumption made at the beginning that only one macroscopic 
1760: state does exist. A more elaborated model can take into account 
1761: more than one level macroscopically populated. Although this 
1762: more general description must not be excluded, 
1763: for energetic reason, this phenomenon is not likely 
1764: to happen \cite{GSS}. A fragmented condensate has a 
1765: much higher potential energy than a non-fragmented one 
1766: due to the presence of a Fock interaction energy between 
1767: two fragmented parts. 
1768: To explain the condensate formation and if we exclude 
1769: fragmentation, we must assume either that the Bose gas 
1770: must be instable due to a far non-equilibrium situation 
1771: or that it must be inhomogeneous with regions of strong depletion. 
1772: Indeed, in those regions - like the edge of the gas - the condensate 
1773: population is not macroscopic anymore and so collisions between 
1774: the two fluids may happen.
1775: 
1776: 
1777: \section{Extension to a weakly inhomogeneous Bose gas}
1778: 
1779: \subsection{Inhomogeneous equations}
1780: 
1781: The extension of the previous equations to 
1782: inhomogeneity 
1783: is important in order to understand how 
1784: the evolution of the condensate and  
1785: the thermal excitations are coupled through 
1786: mean field forces. 
1787: In what follows, we assume that the gas 
1788: is weakly inhomogeneous i.e. most of  
1789: the quasi-particles collide inside a sufficiently small 
1790: volume that can be considered 
1791: as homogeneous. 
1792: To quantify the level of acceptable 
1793: inhomogeneity, we divide the volume $V$ of the gas 
1794: into small cubic volume $\Omega$. The edge $l$ of each  
1795: volume $\Omega=l^3$  is adjusted in such a way to be 
1796: infinitesimal so that the gas is  
1797: homogeneous inside it, but big enough to
1798: contain a large amount of particle 
1799: whose dynamic obeys still locally the homogeneous equation. 
1800: In the literature \cite{Balescu}, 
1801: $l$ is referred as the hydrodynamic  scale 
1802: and is estimated from the formula 
1803: $l(\vc{r},t)= \rho_e(\vc{r},t)/\nabla_{\vc{r}} \rho_e(\vc{r},t)$ 
1804: where $\rho_e(\vc{r},t)$ is the normal fluid density  
1805: and $\vc{r}$ 
1806: is the position in space of the small 
1807: volume. This length scale must be much greater than 
1808: the mean free path 
1809: which can be estimated more less as $1/(\rho_e\sigma)$
1810: where  
1811: $\sigma=8\pi a^2$ is the total cross section \cite{Balescu}.  
1812: 
1813: 
1814: In this infinitesimal volume, we define 
1815: the local condensate wave function
1816: \begin{eqnarray}
1817: \Psi(\vc{r},t)=\sqrt{n_c(\vc{r},t)}
1818: e^{i\theta (\vc{r},t)}
1819: \end{eqnarray}
1820: which corresponds to the eigenfunction of the density matrix
1821: with the highest macroscopic eigenvalue \cite{Leggett}.
1822: The local momentum of the condensate depends also on the position and
1823: the time through the relation
1824: \begin{eqnarray}
1825: \vc{k_s}(\vc{r},t)=\nabla_\vc{r} \theta (\vc{r},t)
1826: \end{eqnarray}
1827: Up to a constant phase, the amplitude and the local momentum
1828: of the condensate characterize fully $\Psi(\vc{r},t)$.
1829: In this volume, we define also the local particle number distribution 
1830: or Wigner function 
1831: $n'_{\vc{k}}(\vc{r},t)$ 
1832: (for clarity, explicit dependence of time has been added). 
1833: Also the particles feel an external local potential $V_{eff}(\vc{r})$.
1834: Up to the first order in the interaction potential, the local 
1835: energy density is divided into the kinetic part, an external 
1836: potential part and 
1837: the Hartree-Fock part \cite{Zaremba,BZS}: 
1838: \begin{eqnarray}
1839: {\cal E}(\vc{r},t)=
1840: \frac{|\nabla_{\vc{r}}\Psi(\vc{r},t)|^2}{2m}
1841: + V_{eff}(\vc{r})|\Psi(\vc{r},t)|^2
1842: \nonumber \\
1843: +\sum_\vc{k} 
1844: \left(\frac{\vc{k}^2}{2m} +V_{eff}(\vc{r})\right)
1845: n'_{\vc{k}}(\vc{r},t)
1846: +\frac{2\pi a}{mV}\big[|\Psi(\vc{r},t)|^4 +
1847: \nonumber \\
1848: 4 |\Psi(\vc{r},t)|^2 \sum_{\vc{k}} n'_{\vc{k}}(\vc{r},t)
1849: +2 (\sum_{\vc{k}} n'_{\vc{k}}(\vc{r},t))^2 \big]
1850: \end{eqnarray}
1851: This energy density varies slowly in space. The gradients 
1852: produced by these variations govern the dynamic of the 
1853: gas particle between each small volume $\Omega$. 
1854: The local energy per excited particle with a momentum $\vc{k}$ 
1855: is given by the derivative of the local energy density 
1856: with respect to the particle number \cite{Zaremba}:
1857: %For the condensate and the excited particle, we obtain 
1858: %respectively
1859: %\begin{eqnarray}
1860: %\epsilon_{\vc{k_s}}(\vc{r},t)=
1861: %\frac{d {\cal E}}{d n_{\vc{k_s}}}=
1862: %\frac{\vc{k_s}^2(\vc{r},t)}{2m}+
1863: %{V}_c(\vc{r},t)
1864: %\end{eqnarray}
1865: \begin{eqnarray}
1866: \epsilon_{\vc{k}}(\vc{r},t)=
1867: \frac{d {\cal E}}{d n'_{\vc{k}}(\vc{r},t)}
1868: =
1869: \frac{\vc{k}^2}{2m}+
1870: {V}_e(\vc{r},t)
1871: \end{eqnarray}
1872: where we define the effective potentials 
1873: felt by the condensate and the particle 
1874: \begin{eqnarray}
1875: {V}_e(\vc{r},t)=
1876: V_{ext}(\vc{r})+
1877: \frac{8\pi a}{mV}
1878: \left[ |\Psi(\vc{r},t)|^2 +
1879:  \sum_{\vc{k}} n'_{\vc{k}}(\vc{r},t) \right]
1880: \end{eqnarray}
1881: Between each infinitesimal volume, particles 
1882: are transferred due to the gradients of the 
1883: particle density and of the potential energy.
1884: As a consequence, the kinetic equations must 
1885: be modified in order to take into account 
1886: locally the modifications inside $\Omega$.
1887: If we consider that the excited particle moves 
1888: classically, then the dynamic of transfer is given 
1889: by the Liouville operator acting on the distribution 
1890: function:
1891: \begin{eqnarray}
1892: \{\epsilon_{\vc{k}}(\vc{r},t),n'_{\vc{k}}(\vc{r},t)\}
1893: \nonumber \\
1894: =\nabla_{\vc{k}}\epsilon_{\vc{k}}(\vc{r},t).
1895: \nabla_{\vc{k}}n'_{\vc{k}}(\vc{r},t)
1896: -\nabla_{\vc{r}}\epsilon_{\vc{k}}(\vc{r},t).
1897: \nabla_{\vc{k}}n'_{\vc{k}}(\vc{r},t)
1898: \end{eqnarray} 
1899: The condensate particles however move locally 
1900: according to the quantum wave function, which 
1901: obeys to the generalized Gross-Pitaevskii 
1902: equation  
1903: with the 
1904: effective potential \cite{Zaremba}: 
1905: \begin{eqnarray}
1906: {V}_c(\vc{r},t)=V_{ext}(\vc{r})+
1907: \frac{4\pi a}{mV}
1908: \left[|\Psi(\vc{r},t)|^2 +
1909: 2 \sum_{\vc{k}} n'_{\vc{k}}(\vc{r},t) \right]
1910: \end{eqnarray}
1911: 
1912: Furthermore, we assume that locally in the
1913: small volume $\Omega$:
1914: 1) the system is sufficiently homogeneous so that the expressions 
1915: (\ref{K6}) and (\ref{K7}) remain valid and the condensate is 
1916: not affected by collision; 2) the condensate is
1917: locally macroscopically populated. 
1918: Combining these results together with the collision 
1919: terms, we can write the kinetic equations for 
1920: weakly inhomogeneous and stable Bose gas in the region of condensation. 
1921: We find two coupled set of equations, one 
1922: for the condensate, the other for the normal fluid: 
1923: \begin{widetext}
1924: \begin{eqnarray}\label{inhomo}
1925: i\frac{\partial}{\partial t} \Psi(\vc{r},t)
1926: &=&\left[-\frac{\nabla_\vc{r}^2}{2m}
1927: +{V}_c(\vc{r},t)
1928: \right]\Psi(\vc{r},t)
1929: \\ \label{inhomo2}
1930: \frac{\partial}{\partial t} n'_{\vc{k}}(\vc{r},t)
1931: &=&
1932: \left[-\frac{\vc{k}}{m}.\nabla_\vc{r}
1933: +\nabla_\vc{r}V_{e}(\vc{r},t)
1934: .\nabla_\vc{k}\right]n'_{\vc{k}}(\vc{r},t)
1935: +{\cal C}_{\vc{k}}[n_{\vc{k'}}(\vc{r},t);\vc{k_s}(\vc{r},t)]
1936: \end{eqnarray}
1937: \end{widetext}
1938: With the exception of the collision term, 
1939: the Eq.(\ref{inhomo}) and (\ref{inhomo2}) are identical 
1940: to the kinetic equations formulated by 
1941: Zaremba et al. \cite{Zaremba}. 
1942: The inhomogeneous kinetic equations satisfy  
1943: locally the conservation laws in these regions of highly 
1944: populated condensate. 
1945: We can derive indeed a 
1946: conservation 
1947: equation for the local particle number 
1948: $N(\vc{r},t)=\sum_\vc{k}n_{\vc{k}}(\vc{r},t)$, the 
1949: local momentum $P(\vc{r},t)=\sum_\vc{k}\vc{k} n_{\vc{k}}(\vc{r},t)$
1950: and the total local energy ${\cal E}(\vc{r},t)$ \cite{Zaremba}. It is also 
1951: easy to verify that the production of the local entropy 
1952: $S(\vc{r},t)$ is always positive \cite{Balescu}. 
1953: 
1954: The local production of entropy stops when we reach 
1955: the local equilibrium:
1956: \begin{eqnarray}\label{neqin}
1957: n'^{eq}_{\vc{k}}(\vc{r},t) =
1958: \frac{1}{\exp{[\beta(\vc{r},t)(\epsilon_\vc{k}-\vc{k}.\vc{v_n}(\vc{r},t)
1959: -\mu(\vc{r},t))]}-1}
1960: \end{eqnarray}
1961: where $\beta(\vc{r},t)$, $\vc{v_n}(\vc{r},t)$ and 
1962: $\mu(\vc{r},t)$ are now local functions of the position and the time.
1963: 
1964: A gap shows up between the energies of the condensed and non condensed 
1965: particles since 
1966: ${V}_c(\vc{r},t) < {V}_e(\vc{r},t)$. This is not a problem 
1967: as long as in the region of the gap the transfer of particle is 
1968: forbidden. The transfer of particle is only possible in the region 
1969: where there is no gap i.e. when the condensate population is not 
1970: macroscopic. 
1971: Therefore, on the 
1972: basis of these non-equilibrium considerations, the Hartree-Fock model 
1973: showing up this forbidden gap does not  
1974: suffer from any inconsistency related to the conservation of energy. 
1975: 
1976: \subsection{The superfluid universe at finite temperature}
1977: 
1978: Using the expression (\ref{neqin}) as the solution of 
1979: the kinetic equation, we can find a set of 
1980: equations describing the non dissipative motion of 
1981: the condensed gas at finite temperature. If we assume 
1982: that $\vc{v_n}(\vc{r},t)=0$ then the substitution of 
1983: (\ref{neqin}) in 
1984: Eq.(\ref{inhomo2}) imposes 
1985: that $n'^{eq}_{\vc{k}}(\vc{r},t)$ must be stationary in time, 
1986: that $\beta(\vc{r},t)=\beta$ is a constant and 
1987: that $\mu(\vc{r},t)=-V_e(\vc{r},t)+\mu_e$, where $\mu_e$ is 
1988: a chemical potential independent of the position controlling the number 
1989: of excited particles.
1990: As a consequence, after carrying out 
1991: the integration over the momentum, the local number 
1992: of excited particle is given by the closed equation:
1993: \begin{eqnarray}\label{super1}
1994: N_e^{eq}(\vc{r})=
1995: \sum_{\vc{k}} n'^{eq}_{\vc{k}}(\vc{r}) 
1996: \nonumber \\
1997: =V
1998: \left(\frac{m}{2\pi \beta}\right)^{3/2}
1999: \!\!\!\!
2000: g_{3/2}\left(e^{\beta[
2001: \mu_e -
2002: V_{ext}(\vc{r})-
2003: \frac{8\pi a}{mV}
2004: (|\Psi(\vc{r},t)|^2 +
2005: N_e^{eq}(\vc{r}) )]} \right)
2006: \nonumber \\
2007: \end{eqnarray}
2008: where $g_{l}(x)=\sum_{j=1}^\infty
2009: j^{-l}x^j$. Since the system is stationary, 
2010: the macroscopic condensate wave function has the form:
2011: \begin{eqnarray}
2012: \Psi(\vc{r},t)=
2013: e^{-i\mu_c t}\Psi_c(\vc{r})
2014: \end{eqnarray}
2015: where $\mu_c$ is the chemical potential of 
2016: controlling the population of the condensate particle.  
2017: The substitution of this form into Eq.(\ref{inhomo})
2018: produces:
2019: \begin{eqnarray}\label{super2}
2020: \left[-\frac{\nabla_\vc{r}^2}{2m}
2021: +
2022: V_{ext}(\vc{r})+
2023: \frac{4\pi a}{mV}
2024: (|\Psi_c(\vc{r})|^2 +
2025: 2N_e^{eq}(\vc{r}) )
2026: \right]\Psi_c(\vc{r})
2027: \nonumber \\
2028: =
2029: \mu_c \Psi_c(\vc{r})
2030: \end{eqnarray}  
2031: The coupled set of equations 
2032: (\ref{super1}) and (\ref{super2}) describe 
2033: locally all the non dissipative or superfluid 
2034: phenomena. These are valid provided the gas is stable. 
2035: As an example, they describe  
2036: the superfluid moving with a different velocity 
2037: than the normal fluid. In that case when 
2038: $V_{ext}(\vc{r})=0$, the 
2039: solution is  
2040: a plane wave function:
2041: \begin{eqnarray}
2042: \Psi_c(\vc{r})=e^{i\vc{k_s}.\vc{r}}
2043: n_{\vc{k_s}}^{1/2}
2044: \end{eqnarray} 
2045: and $N_e^{eq}(\vc{r})$ is a constant. 
2046: More generally, any non dissipative 
2047: complex structure 
2048: at finite temperature, like vortices, 
2049: should be a solution of (\ref{super1}) and (\ref{super2}).
2050: 
2051: In contrast to previous 
2052: studies, non-dissipative phenomena are not a 
2053: requirement or assumption of a model but rather a 
2054: prediction of the GRPA kinetic theory. 
2055: Another difference is that the chemical potential 
2056: for both fluids $\mu_c$ and $\mu_e$ must not 
2057: be necessarily identical, contrary to 
2058: the equilibrium statistical mechanics which 
2059: imposes equality \cite{Huang}. In our example, 
2060: $\mu_e=(8\pi a N)/(mV) \not= 
2061: \mu_c=\vc{k_s}^2/(2m) + 4\pi a (2 N_e+n_{\vc{k_s}})/(mV)$ 
2062: at the difference of \cite{Zaremba} where 
2063: $\mu_e=\mu_c$ for $\vc{k_s}=0$.
2064: 
2065: 
2066: \section{Analogy with plasmon theory}
2067: 
2068: The derivation leading to the QKE for a Bose gas 
2069: is similar to 
2070: the one leading to the QKE equation 
2071: for plasma physics \cite{QBalescu, Ichimaru,WP} except 
2072: that, in these equations, the collective excitations 
2073: are derived in the SRPA. In plasma physics, the 
2074: Coulomb interactions 
2075: potential for the collision process is screened 
2076: beyond the Debye wavelength. The dynamic dielectric 
2077: function  displays this screening and removes the 
2078: singularity in the Coulomb potential in the  
2079: long wavelength limit. 
2080: Static screening is well known since the Debye theory. 
2081: However, for dynamic screening, Wyld and Pines have 
2082: proposed an interpretation in terms of the plasmon 
2083: theory. This theory is an alternative version of the kinetic theory 
2084: of a plasma which emphasizes the role played by the collective 
2085: modes. According to their work, the effective potential 
2086: interaction 
2087: is the result of a plasmon mediating the interaction with 
2088: the quasi particles. In other words, during a collision
2089: process, a quasi-particle emits an intermediate plasmon 
2090: on the energy shell with the momentum transfer $\vc{q}$. 
2091: Later on, this plasmon is 
2092: eventually absorded by another quasiparticle which 
2093: acquires a new momentum and a new energy. 
2094: The energy spectrum for this plasmon corresponds  
2095: precisely to the 
2096: frequency spectrum of the 
2097: collective excitations in a plasma.  
2098: 
2099: In the case of a Bose condensed gas, 
2100: the collective excitations spectrum 
2101: corresponds
2102: to the Bogoliubov frequency spectrum for 
2103: low temperature. This property suggests 
2104: the interpretation that these phonon-like 
2105: excitations play also the role of 
2106: mediators during collisions between 
2107: quasi-particles. Thus, the Bose condensation 
2108: has the effect to transform the scattering modes, 
2109: used for collision involving condensed particle, 
2110: into a collective 
2111: mode used for mediating the interaction between 
2112: non condensed particles.
2113: 
2114: The plasmon theory has been derived using 
2115: the theory of quantum 
2116: electrodynamics. In principle, 
2117: for a Bose gas, a similar 
2118: approach must be carried out taking into account 
2119: that, at a more fundamental level, 
2120: the interaction potential originates 
2121: from processes involving the absorption 
2122: and the emission of photons.
2123: In this paper, we shall not rederive 
2124: a similar theory but rather recover 
2125: heuristically 
2126: the analogy 
2127: with the plasmon theory.  
2128: 
2129: Following this approach, this intermediate
2130: excitation 
2131: behaves like a particle characterized by its 
2132: own distribution function $f_{\vc{q}}$ and that 
2133: we shall call by analogy "condenson".
2134: The energy spectrum of the condenson is given by 
2135: the zeroes of the dynamic dielectric function i.e. 
2136: by the solution of $\Delta(\vc{q},\omega)=0$. 
2137: We consider the simple 
2138: case of an homogenous condensate at rest $\vc{k_s}=0$ 
2139: and weakly depleted.
2140: In that case,  
2141: the solution is the complex number 
2142: $\omega=\omega_{\vc{q}}-i\gamma_{\vc{q}}$, 
2143: where the real part represents the energy
2144: spectrum $\omega_{\vc{q}}=
2145: \epsilon_{\vc{q}}^B$
2146: and the imaginary part represents the decay 
2147: rate of the condenson given by  Eq.(\ref{imo}).
2148: Then, looking at Eq.(7) and Eq.(8) in \cite{WP}, we can 
2149: write two coupled equations, one for the dynamic 
2150: evolution 
2151: of the quasi-particle and the other 
2152: for the evolution of the condenson. 
2153: They describe the time rate change of these quantities 
2154: due to the emission and absorption of a condenson by a 
2155: quasi-particle:
2156: \begin{widetext}
2157: \begin{eqnarray}\label{WP1}
2158: \frac{\partial}{\partial t} n'_{\vc{k}}
2159: &=&\frac{8\pi a}{mV}\!\! \sum_{\vc{q}}
2160: \frac{8\pi a n_\vc{0} \vc{q}^2}{m^2 V\omega_\vc{q}}
2161: \big[
2162: \pi \delta(\omega_{\vc{q}}+
2163: \epsilon_{\vc{k}}-\epsilon_{\vc{k}+\vc{q}})
2164: \left(
2165: (n'_{\vc{k}}+1)n'_{\vc{k}+\vc{q}}
2166: (f_{\vc{q}}+1)
2167: -n'_{\vc{k}}(n'_{\vc{k}+\vc{q}}+1)f_{\vc{q}}
2168: \right)
2169: \nonumber \\
2170: && +\pi\delta(\omega_{\vc{q}}-
2171: \epsilon_{\vc{k}}+\epsilon_{\vc{k}-\vc{q}})
2172: \left(
2173: (n'_{\vc{k}}+1)n'_{\vc{k}-\vc{q}}f_{\vc{q}}
2174: -n'_{\vc{k}}(n'_{\vc{k}-\vc{q}}+1)(f_{\vc{q}}+1)
2175: \right)
2176: \big]
2177: \\ \label{WP2}
2178: \frac{\partial}{\partial t}f_{\vc{q}}&=&
2179: -2\gamma_{\vc{q}}f_{\vc{q}}
2180: +\frac{8\pi a }{mV}
2181: \frac{8\pi a n_\vc{0} \vc{q}^2}{m^2V\omega_\vc{q}}
2182: \sum_{\vc{k}}\pi\delta(\omega_{\vc{q}}-
2183: \epsilon_{\vc{k}}+\epsilon_{\vc{k}-\vc{q}})
2184: n'_{\vc{k}}(n'_{\vc{k}-\vc{q}}+1)
2185: \end{eqnarray}
2186: \end{widetext}
2187: Similarly to the reasoning of Wyld and Pines,
2188: the kinetic equation (\ref{K6}) can be rederived
2189: from Eq.(\ref{WP1}) and Eq.(\ref{WP2}) by eliminating 
2190: adiabatically $f_\vc{q}$. We assume that 
2191: the time
2192: derivative in Eq.(\ref{WP2}) is zero
2193: and the elimination allows to obtain an equation 
2194: for $n'_{\vc{k}}$ only. Eq.(\ref{K6}) is recovered, 
2195: provided  that
2196: the condensate is weakly depleted. Under these circonstances, 
2197: the square of the effective potential has a narrow Lorentzian  
2198: shape whose two peaks and widths correspond to the condenson 
2199: energy and decay rate respectively (see \cite{WP}). It 
2200: plays the role of a propagator for the mediated interaction.
2201: This supposes that the dielectric function can be 
2202: approximated around the resonant frequencies as 
2203: (see Eq.(25) in \cite{WP} for comparison):
2204: \begin{eqnarray}\label{ka}
2205: {\cal K}(\vc{q},\omega)=\left\{
2206: \begin{array}{ccc}
2207: \frac{m^2 V\omega_{\vc{q}}}{4\pi a n_\vc{0} \vc{q}^2}
2208: (\omega -\omega_{\vc{q}}+i\gamma_{\vc{q}}),& 
2209: \omega \sim \omega_{\vc{q}}\\
2210:  -\frac{m^2 V\omega_{-\vc{q}}}{4\pi a n_\vc{0} \vc{q}^2}
2211: (\omega + \omega_{-\vc{q}}-i\gamma_{-\vc{q}}),&
2212: \omega \sim -\omega_{-\vc{q}}
2213: \end{array}
2214: \right.
2215: \end{eqnarray}
2216: leading to the approximation:
2217: \begin{eqnarray}
2218: \frac{1}{|{\cal K}(\vc{q},\omega)|^2}
2219: =
2220: \frac{\pi}{\gamma_{\vc{q}}}
2221: \left(\frac{4\pi a n_\vc{0} \vc{q}^2}{m^2 V\omega_\vc{q}}\right)^2
2222: \!\!\!\!\left(
2223: \delta(\omega-\omega_{\vc{q}})+
2224: \delta(\omega+\omega_{\vc{q}})
2225: \right)
2226: \end{eqnarray}
2227: These coupled equations suggest that a weakly interacting  
2228: Bose 
2229: condensed gas is in reality composed of two kind of 
2230: bosonic particles:
2231: the quasi-particles that become the real particles in 
2232: the limit of an ideal Bose gas and the condensons that 
2233: result from a mediation (see Fig.1).
2234: At thermal equilibrium, the various particles 
2235: obey to the Bose-Einstein statistics: 
2236: \begin{eqnarray}\label{neq2}
2237: n'_{\vc{k}}=n^{eq}_{\vc{k}}=
2238: \frac{1}{\exp{[\beta(\epsilon_\vc{k}-\mu)]}-1}
2239: \end{eqnarray}
2240: and 
2241: \begin{eqnarray}\label{feq}
2242: f_{\vc{q}}=
2243: \frac{1}{\exp{(\beta \omega_{\vc{q}})}-1}
2244: \end{eqnarray}
2245: In this way, the condenson distribution function 
2246: has the same form as the  
2247: excitations distribution function
2248: predicted by Bogoliubov \cite{Pines}. 
2249: But there is an important difference however:
2250: 
2251: {\it 
2252: In the Bogoliubov theory, the Bogoliubov 
2253: excitations 
2254: correspond to the quasi-particle. On the contrary, in the GRPA 
2255: model, the Bogoliubov excitations correspond to the 
2256: condensons and, in this respect, are not the 
2257: quasi-particles.}
2258: 
2259: The QKEs proposed by \cite{Imamovic,Walser,Kirkpatrick,Gardiner}
2260: suggest instead that 
2261: the Bogoliubov excitations correspond to the quasi-particle.  
2262: We must be cautious with this interpretation since 
2263: the GRPA model is not an exact one. It might be that, 
2264: to a next order expansion in the interaction, 
2265: the quasi-particles have a different energy spectrum that is linear 
2266: in the momentum. Without any further investigation, 
2267: we cannot claim that the Goldstone boson, that would be 
2268: normally predicted from the $U(1)$ theory, 
2269: is the condenson or the quasi-particle. 
2270: 
2271: \begin{figure}
2272: \scalebox{1}{
2273: \includegraphics{condenson.eps}} 
2274: \caption{Feymann diagram of the 
2275: interaction of two quasi-particle (full line)
2276: mediated by the condenson (wavy line)}  
2277: \end{figure}
2278: %{\it The Bogoliubov excitations are 
2279: %not the same as an excitation corresponding 
2280: %to a quasi-particle. Rather they are intermediate 
2281: %phonon-like excitation that carry momentum 
2282: %and energy between the particles.}
2283: 
2284: The condenson has a decay rate $2\gamma_{\vc{q}}$. 
2285: Thus, a duration of the collision is limited 
2286: by its lifetime and can be considered as instantaneous, 
2287: if much lower than the inverse relaxation rate. 
2288: Assuming that the particle has a mean velocity 
2289: $1/(\beta m)^{1/2}$, this requirement imposes for the 
2290: typical values:
2291: $\gamma_{\vc{q}} \gg \rho_e \sigma /(\beta m)^{1/2}$.
2292: If it is not fulfilled, then the long lifetime 
2293: of the condenson suggests that its dynamic evolution 
2294: is not instantaneous and that the full set of equations 
2295: (\ref{WP1}) and (\ref{WP2}) must be rather used.
2296: 
2297: 
2298: \section{Equivalence with the Bogoliubov theory}
2299: 
2300: The RPA model presents important  differences compared to
2301: any previous model for finite temperature.
2302: However for zero temperature, it
2303: allows to recover the next order correction for the ground state
2304: energy and the static structure factor 
2305: usually obtained from the Bogoliubov theory.
2306: Such a derivation has already been made by
2307: Nozi\`eres and Pines for an electron gas
2308: using the density response formalism.
2309: They were also able to apply
2310: this formalism to the condensate \cite{Pines}.
2311: The nice result of this section is that,
2312: precisely, the formalism developed in this
2313: paper allows to recover exactly the same results for the 
2314: ground state energy and the structure factor.
2315: 
2316: Indeed, the correlation function defined in (\ref{g}) allows
2317: to calculate the correction to the energy but also to 
2318: the total particle momentum
2319: distribution $N_{\vc{k}}$. This new distribution differs 
2320: from the quasi-particles population 
2321: distribution $n_{\vc{k}}$. We start
2322: initially from the distribution without interaction
2323: $N_{\vc{k}}=\delta_{\vc{k},\vc{0}} n_{\vc{0}}$ where
2324: only the condensed mode $\vc{k_s}=0$ is populated. Then,
2325: we switch on adiabatically the interaction and the
2326: correlations appear in a form given by (\ref{K3}).
2327: As a consequence, the condensate becomes weakly depleted
2328: and the total kinetic and potential energy gets modified.
2329: 
2330: These corrections can be
2331: calculated directly the static structure factor
2332: which, using (\ref{g}) and for $\vc{q} \not= 0$, 
2333: is given by:
2334: \begin{eqnarray}\label{SF}
2335: N S(\vc{q})=\langle \rho^\dagger_{\vc{q}}\rho_{\vc{q}}
2336: \rangle=
2337: \sum_{\vc{k}}(n_{\vc{k}}+1)n_{\vc{k}+\vc{q}}
2338: +
2339: \sum_{\vc{k},\vc{k'}}
2340: g_\vc{q}(\vc{k},\vc{k'}) 
2341: \end{eqnarray}
2342: where $N=\sum_{\vc{k}}N_{\vc{k}}=\sum_{\vc{k}} n_{\vc{k}}$. 
2343: An expression for this integral can be found
2344: in the appendix C. 
2345: In contrast, Nozi\`eres and Pines express the 
2346: structure factor in terms of the susceptibility function 
2347: (\ref{susc}) of the ground state 
2348: as \cite{NP}:
2349: \begin{eqnarray}\label{SFNP}
2350: NS(\vc{q})=\langle \rho^\dagger_{\vc{q}}\rho_{\vc{q}}
2351: \rangle =-
2352: \int_0^\infty
2353: \frac{d\omega}{\pi}
2354: {\rm Im} \chi (\vc{q},\omega)
2355: \end{eqnarray}
2356: At zero temperature, the limit
2357: $n'_{\vc{k}} \rightarrow 0$ must be taken carefully.
2358: It implies that the Landau damping
2359: $\gamma_{\vc{q}} \rightarrow 0$ but this parameter is
2360: different from $0_+$. The limit $0_+ \rightarrow 0$
2361: must be carried out before the limit
2362: $\gamma_{\vc{q}} \rightarrow 0$ and this procedure
2363: can be done by adjusting adequately the infinitesimal value of $0_+$
2364: and $n'_{\vc{k}}$. In particular, for a normal fluid 
2365: in equilibrium,  this amounts to prescribing to 
2366: take the limit $0_+$ to zero and then after taking  the 
2367: limit $\beta \rightarrow \infty$ in Eq.(\ref{neq}).
2368: The physical reason for such
2369: a procedure is that the presence of infinitesimal fraction
2370: of excited particles will force the system to create
2371: the correct equilibrium correlation. 
2372: In these conditions and
2373: using Eq.(\ref{Knapp}) and 
2374: Eq.(\ref{imo}), we can re express for $\vc{k_s}=0$:
2375: \begin{eqnarray}\label{disprel1}
2376: \Delta(\vc{q},\omega) \rightarrow
2377: (\omega+i\gamma_{\vc{q}})^2-
2378: {\epsilon^B_{\vc{q}}}^2
2379: \end{eqnarray}
2380: Also, we note the relation: 
2381: \begin{eqnarray}
2382: {\rm Im}A'(\vc{q},\omega)=
2383: -\frac{mV}{8\pi a}
2384: \frac{1}{\exp(\beta \omega)-1}
2385: {\rm Im}{\cal K}_n (\vc{q},\omega)
2386: \end{eqnarray}
2387: Inserting these results 
2388: and (\ref{imo}) into
2389: (\ref{intg}), setting to zero the real part
2390: of ${\cal K}(\vc{q},\omega)$ and $A'(\vc{q},\omega)$, and 
2391: setting $A_0 (\vc{q},\omega)=N/[\omega+i0^+ +\vc{q}^2/(2m)]$ 
2392: then Eq.(\ref{SF}) becomes:
2393: \begin{eqnarray}
2394: S(\vc{q})
2395: &=&1 - \int_{-\infty}^\infty
2396: \frac{d\omega}{\pi}
2397: \frac{\omega^2-\left(\frac{\vc{q}^2}{2m}\right)^2 +
2398: \frac{4\pi a N}{mV}\frac{\vc{q}^2}{m}}
2399: {|(\omega+i\gamma_{\vc{q}})^2-
2400: {\epsilon^B_{\vc{q}}}^2|^2}
2401: \nonumber \\ 
2402: \nonumber && \times
2403: %\frac{\omega \gamma_{\vc{q}}}{\frac{4\pi a N \vc{q}^2}{m^2V}}
2404: \left(\omega-\frac{\vc{q}^2}{2m}
2405: \frac{e^{\beta \omega}+1}{e^{\beta \omega}-1}\right)  
2406: {\rm Im}{\cal K}_n (\vc{q},\omega)\\
2407: \label{T0'}
2408: &\stackrel{\begin{array}{c} \gamma_{\vc{q}} \rightarrow 0
2409: \\ \beta \rightarrow \infty \end{array}}{=}&
2410: \frac{\vc{q}^2}{2m\epsilon^B_{\vc{q}}}
2411: \end{eqnarray}
2412: In order to get
2413: (\ref{T0'}),
2414: the integration has been carried out transforming 
2415: the Lorentzian factors into delta functions and using 
2416: (\ref{imo}). This result is identical to that
2417: obtained from the Bogoliubov theory.
2418: This result can also been obtained substituting (\ref{susc}) 
2419: into (\ref{SFNP}) and carry out
2420: the integration over $\omega$.
2421: Instead of evaluating the total energy and the 
2422: correction to the particle number distribution directly,
2423: some useful tricks   
2424: allow to 
2425: avoid some complex calculation \cite{NP}. The total ground state 
2426: energy is  
2427: a functional of the interaction $a$ and the dispersion relation 
2428: $\epsilon_{\vc{k}}$ (in what follows, we consider only 
2429: the first term in (\ref{U0})). It can be expressed in terms of 
2430: the matrix element
2431: $E(a,\epsilon_{\vc{k}})=
2432: \langle \psi| H |\psi \rangle$
2433: where $|\psi \rangle$ is 
2434: the ground state depending also on $a$ and $\epsilon_{\vc{k}}$. 
2435: Using the property that $|\psi \rangle$ is a normalized 
2436: eigenfunction,
2437: the first derivative with respect to these parameters 
2438: produces:
2439: \begin{eqnarray}\label{da}
2440: a \frac{\partial E(a,\epsilon_{\vc{k}})}
2441: {\partial a}&=& E_{int}(a,\epsilon_{\vc{k}})
2442: \\ \label{de}
2443: \frac{\partial E(a,\epsilon_{\vc{k}})}
2444: {\partial \epsilon_{\vc{k}}}&=& N_{\vc{k}}
2445: \end{eqnarray}
2446: The total interaction energy can be expressed in terms 
2447: of the structure factor:
2448: \begin{eqnarray}\label{T01}
2449: E_{int}(a,\epsilon_{\vc{k}})=
2450: \frac{2\pi a}{{mV}}\left[ N(N-1)+N\sum_{\vc{q}\not=0}
2451: (S(\vc{q})-1)\right]
2452: \end{eqnarray}
2453: At zero temperature and without interaction, 
2454: all the particles are in the ground state and thus 
2455: $E(0,\epsilon_{\vc{k}})=N\epsilon_{\vc{0}}$.
2456: With this initial condition,  
2457: an integration over $a$ allows to calculate the total energy:
2458: \begin{eqnarray}\label{T03}
2459: E(a,\epsilon_{\vc{k}})=N\epsilon_{\vc{0}}+ 
2460: \int_0^a \frac{da'}{a'} E_{int}(a',\epsilon_{\vc{k}})
2461: \nonumber \\ +
2462: (\frac{2\pi a N}{V})^2\sum_{\vc{q}\not=0}\frac{2m}{\vc{q}^2}
2463: \end{eqnarray}
2464: where a supplementary term coming from (\ref{U0}) 
2465: has been added to 
2466: remove ultra-violet divergencies. The structure factor 
2467: can be recalculated for any arbitrary dispersion relation 
2468: $\epsilon_{\vc{k}}$. Plugging this new expression into 
2469: (\ref{T01}) and carrying out the integral over $a$, we obtain:
2470: \begin{eqnarray}\label{T04}
2471: E(a,\epsilon_{\vc{k}})=
2472: N\epsilon_{\vc{0}}+
2473: \frac{2\pi a N^2}{mV}+
2474: \frac{1}{2}
2475: \sum_{\vc{q}\not=0} \bigg[
2476: (\frac{2\pi a N}{V})^2\frac{m}{\vc{q}^2}+
2477: \nonumber \\
2478: \sqrt{\Delta\epsilon_{\vc{q}}^2+\frac{8\pi a N}{mV}
2479: \Delta\epsilon_{\vc{q}}}-\Delta\epsilon_{\vc{q}}
2480: -\frac{4\pi a N}{mV} \bigg]
2481: \end{eqnarray}
2482: where $\Delta\epsilon_{\vc{q}}=
2483: (\epsilon_{\vc{q}}+\epsilon_{-\vc{q}}-
2484: 2\epsilon_{\vc{0}})/2$. In the usual case, $\Delta\epsilon_{\vc{q}}=
2485: \vc{q}^2/2m$ and an integration over the momentum $\vc{q}$ allows 
2486: to find the ground state energy \cite{Huang}:
2487: \begin{eqnarray}\label{Ecor}
2488: E(a)=
2489: \frac{2\pi a N^2}{mV}
2490: [1+ \frac{128}{15}\sqrt{\frac{a^3 N}{\pi V}}]
2491: \end{eqnarray}
2492: Finally, the derivative with respect 
2493: to $\epsilon_{\vc{k}}$ gives the first 
2494: correction to the momentum particle distribution. We get
2495: for $\vc{k}\not= 0$:
2496: \begin{eqnarray}\label{ncor}
2497: N_{\vc{k}}=\left. \frac{\partial E(a,\epsilon_{\vc{k}})}
2498: {\partial \epsilon_{\vc{k}}}\right|_{
2499: \epsilon_{\vc{k}}=\frac{\vc{k}^2}{2m}} 
2500: =\frac{1}{2}\left(\frac{
2501: \frac{\vc{k}^2}{2m}+\frac{4\pi a N}{mV}}{
2502: \epsilon^B_{\vc{k}}}-1\right)
2503: \end{eqnarray}
2504: and for $\vc{k}=0$:
2505: \begin{eqnarray}
2506: N_{\vc{0}}=N- \sum_{\vc{k}\not=0}N_{\vc{k}}
2507: =N[1-\frac{32}{3}\sqrt{\frac{a^3 N}{\pi V}}]
2508: \end{eqnarray}
2509: From our number conserving approach, we recover the well-known
2510: results usually obtained from the Bogoliubov approach at 
2511: zero temperature, 
2512: i.e. the first order energy correction in (\ref{Ecor}) 
2513: and the first order 
2514: correction to the number of excited particles (\ref{ncor}) showing 
2515: a depletion of the condensate particle. 
2516: In principle, the present method can be extended to 
2517: finite temperature, since the relations (\ref{da}) and 
2518: (\ref{de}) are valid also for any excited state labeled 
2519: by the occupation number $n_{\vc{k}}$. The only problem 
2520: remains the explicit calculation of 
2521: $\langle \rho^\dagger_{\vc{q}}\rho_{\vc{q}}
2522: \rangle$. 
2523:   
2524: 
2525: %\begin{eqnarray}
2526: %N^2&=&\sum_{\vc{q},\vc{k}}
2527: %\langle \rho^\dagger_{\vc{k},\vc{q}}
2528: %\rho_{\vc{k}+\vc{q},\vc{q}} \rangle
2529: %\nonumber \\
2530: %&=&
2531: %\sum_{\vc{k}} n_{\vc{k}}
2532: %\sum_{\vc{k'}} n_{\vc{k'}}
2533: %+\sum_{\vc{q},\vc{k}}
2534: %g_\vc{q}(\vc{k},\vc{k}+\vc{q})
2535: %\end{eqnarray}
2536: 
2537: \section{Conclusions and Perspectives}
2538: 
2539: We attempted to reexamine the kinetic theory of 
2540: the weakly interacting and stable Bose condensed gas with the 
2541: objective to explain superfluidity for any temperature 
2542: and, ultimately, to 
2543: be able to distinguish between its dissipative and 
2544: superfluid behaviors. A different 
2545: QKE, taking into account some higher order terms 
2546: in the interaction parameter, has been derived from the 
2547: microscopic theory 
2548: and has the merit to predict a collision blockade 
2549: between condensed and non condensed particle. 
2550: {\it More precisely, the condensate remains invisible to any
2551: thermal quasi-particle due to an induced force, which
2552: acts as a `` coarse-graining '', removing any
2553: local force necessary for binary scattering.}
2554: In this way a metastable state can be built 
2555: locally with a non-zero relative velocity. 
2556: For the homogeneous gas, a microscopic derivation 
2557: has been carried out using the same expansion procedure as that  
2558: for deriving the QKE for a quantum plasma. The only 
2559: difference is that our approach takes into account the 
2560: exchange term which is of the same order of magnitude 
2561: than the direct term.
2562: For the weakly inhomogeneous 
2563: gas, the derivation needs some supplementary 
2564: justifications from first principles, in particular, for the way 
2565: to obtain the generalized Gross-Pitaevskii 
2566: equation from a number conserving approach, but also for the 
2567: limit of use of the collision terms (\ref{K6}) and (\ref{K7}). 
2568: 
2569: 
2570: The generalized RPA is the basic approximation 
2571: which is at the origin of the prediction of the collision 
2572: blockade phenomenon. 
2573: It is not only a method to achieve 
2574: results but has also the deep meaning that average 
2575: contributions that are not oscillating in phase can 
2576: be neglected for diluted gas. The RPA does not serve 
2577: only to predict the QKE; it allows to build a genuine 
2578: alternative theory that 
2579: can be compared with all previous  approaches. The  
2580: main advantages of this theory are:
2581: 1) it is
2582: number conserving; 2) the statistical
2583: equilibrium is not imposed as a postulate but is deduced 
2584: from the QKE  
2585: and allows superfluidity; 3) it is valid for any 
2586: range of temperature below and above the transition point.
2587: 
2588: The predictions calculated 
2589: from this theory, namely the collective excitations, 
2590: could be compared with results 
2591: predicted 
2592: mainly by the Popov theory 
2593: or by the Hartree-Fock-Bogoliubov theory in the Popov 
2594: approximation \cite{Popov,Griffin}. 
2595: The RPA theory 
2596: allows also to determine the correlation function 
2597: $g_\vc{q}(\vc{k},\vc{k'})$ with which one can calculate 
2598: the next order correction. In particular, we recover the 
2599: Bogoliubov results for the next correction to the ground
2600: state energy and for the particle momentum distribution. 
2601: In principle, these corrections  
2602: can also be determined at finite temperature 
2603: and compared with other previous works. Another interesting 
2604: quantity is the static structure factor $S(\vc{q})$ 
2605: which can be evaluated directly from the correlation function. 
2606: 
2607: The RPA theory for dilute Bose gas has some resemblance 
2608: with the plasma kinetic theory. Both 
2609: predict that scattering proceeds by means of an 
2610: intermediate excitation.
2611: For a plasma,  this excitation is a plasmon responsible 
2612: for the collective plasma oscillation. For a Bose gas, 
2613: it is a phonon-like excitation with a Bogoliubov energy 
2614: spectrum at zero temperature and that, by analogy, we can 
2615: call ``condenson''.
2616: 
2617: Let us also mention the limitation of the present theory.
2618: Firstly, this theory preserves the existence of a gap 
2619: in apparent contradiction with the Hugenholtz-Pines theorem \cite{Pines}.
2620: We consider only the Hartree-Fock approximation for the 
2621: potential energy of
2622: the quasi-particle, excluding in this way higher order 
2623: correlated term.
2624: This contradiction might be solved by claiming 
2625: that the poles of the one particle Green function have a
2626: richer structure; namely one gapless pole which represents 
2627: the Bogoliubov or condenson excitation spectrum and another pole 
2628: which may have a gap and which represents the quasi-particle 
2629: spectrum. In this way, we recover the compatibility with the 
2630: HP theorem.
2631: Secondly, in principle, the calculation must start with the 
2632: real potential and not the effective contact 
2633: potential (\ref{U0}). A resummation of all ladder diagrams 
2634: should allow to reexpress the effective interaction in terms 
2635: of the $T$ matrix for the binary collision. Then the limit 
2636: of low energy allows to deduce the expression (\ref{U0}) \cite{Stoof}.
2637: If rather the $T$ many body matrix is considered, then the 
2638: effective interaction could depend also non linearly on 
2639: the quasi-particle number. 
2640: Thirdly, the assumption that the gas is weakly homogeneous might 
2641: not be true, due to a sharp trap potential realized in real 
2642: experiment. The theory must be improved by decomposing the 
2643: one particle Wigner function in terms of its eigenfunctions  
2644: and by deriving an equation of motion for each of them \cite{Leggett}.
2645:   
2646: Finally, all the predictions, that the RPA theory might suggest, 
2647: must be ultimately confronted to experiments in order to be validated.
2648: In particular, the analysis of the various modes for 
2649: the collective excitations and its damping must be recovered 
2650: \cite{JZ}. 
2651: 
2652: \appendix
2653: 
2654: \section{Solution of the equation of motion for the excitations}
2655: 
2656: The system (\ref{col1}) (\ref{col2}) and (\ref{col3}) can be solved
2657: exactly. We isolate the dielectric propagator on the left 
2658: hand side of each equation. After summing over the superfluid and 
2659: normal modes separately and using (\ref{kn}), 
2660: we obtain after rearranging:
2661: \begin{widetext}
2662: \begin{eqnarray}\label{col12}
2663: [(\omega+i0_+ -
2664: \frac{\vc{k_s}.\vc{q}}{m})^2-{\epsilon^B_{\vc{q}}}^2]
2665: \left({\cal U}_\vc{q}(\vc{k_s},\vc{k_1},\omega) +
2666: {\cal U}_\vc{q}(\vc{k_s}-\vc{q} ,\vc{k_1},\omega)
2667: \right)-\frac{8\pi a n_{\vc{k_s}}}{mV}
2668: \frac{\vc{q}^2}{m}\sum'_{\vc{k'}}
2669: \tilde{{\cal U}}_\vc{q}(\vc{k'},\vc{k_1},\omega)
2670: =
2671: \nonumber \\
2672: i
2673: [\omega+i0_+ -
2674: \frac{\vc{k_s}.\vc{q}}{m}+\frac{\vc{q}^2}{2m}]
2675: \delta_{\vc{k_s},\vc{k_1}}
2676: +i
2677: [\omega+i0_+ -
2678: \frac{\vc{k_s}.\vc{q}}{m}-\frac{\vc{q}^2}{2m}]
2679: \delta_{\vc{k_s}-\vc{q},\vc{k_1}}
2680: \end{eqnarray}
2681: \begin{eqnarray}\label{col32}
2682: \left({\cal K}_n(\vc{q},\omega)-1
2683: \right)
2684: \left({\cal U}_\vc{q}(\vc{k_s},\vc{k_1},\omega) +
2685: {\cal U}_\vc{q}(\vc{k_s}-\vc{q} ,\vc{k_1},\omega)
2686: \right)
2687: +
2688: {\cal K}_n(\vc{q},\omega)
2689: \sum'_{\vc{k'}}
2690: \tilde{{\cal U}}_\vc{q}(\vc{k'},\vc{k_1},\omega)
2691: =
2692: i\frac{1-\Pi_{\vc{k_1}}
2693: }{\omega+i0_+ -
2694: \frac{\vc{k_1}.\vc{q}}{m}-\frac{\vc{q}^2}{2m}}
2695: \end{eqnarray}
2696: where we define the function 
2697: $\Pi_{\vc{k}}=\delta_{\vc{k_s},\vc{k}}+
2698: \delta_{\vc{k_s}-\vc{q},\vc{k}}$. Using the definition 
2699: (\ref{disprel}),
2700: the solution of this coupled set of equations gives:
2701: \begin{eqnarray}\label{col13}
2702: {\cal U}_\vc{q}(\vc{k_s},\vc{k_1},\omega) +
2703: {\cal U}_\vc{q}(\vc{k_s}-\vc{q} ,\vc{k_1},\omega)
2704: =
2705: i\frac{{\cal K}_n(\vc{q},\omega)\Pi_{\vc{k_1}}
2706: \left[(\omega+i0_+ -
2707: \frac{\vc{k_s}.\vc{q}}{m})^2-(\frac{\vc{q}^2}{2m})^2\right]
2708: +\frac{8\pi a n_{\vc{k_s}}}{mV}\frac{\vc{q}^2}{m}
2709: (1-\Pi_{\vc{k_1}})
2710: }{\Delta(\vc{q},\omega)
2711: [\omega+i0_+ -
2712: \frac{\vc{k_1}.\vc{q}}{m}-\frac{\vc{q}^2}{2m}]}
2713: \end{eqnarray}
2714: and 
2715: \begin{eqnarray}\label{col33}
2716: \sum'_{\vc{k'}}
2717: \tilde{{\cal U}}_\vc{q}(\vc{k'},\vc{k_1},\omega)
2718: =
2719: i\frac{(1-{\cal K}_n(\vc{q},\omega))
2720: \left[(\omega+i0_+ -
2721: \frac{\vc{k_s}.\vc{q}}{m})^2-(\frac{\vc{q}^2}{2m})^2\right]
2722: \Pi_{\vc{k_1}}+
2723: [(\omega+i0_+ -
2724: \frac{\vc{k_s}.\vc{q}}{m})^2-{\epsilon^B_{\vc{q}}}^2]
2725: (1-\Pi_{\vc{k_1}})
2726: }{\Delta(\vc{q},\omega)
2727: [\omega+i0_+ -
2728: \frac{\vc{k_1}.\vc{q}}{m}-\frac{\vc{q}^2}{2m}]}
2729: \end{eqnarray}
2730: Consequently, the definitions (\ref{kalc}) and 
2731: (\ref{kalc2}) allow to write
2732: \begin{eqnarray}\label{col34}
2733: \sum_{\vc{k'}}
2734: {\cal U}_\vc{q}(\vc{k'},\vc{k_1},\omega)
2735: =\frac{i}{\omega+i0_+ -
2736: \frac{\vc{k_1}.\vc{q}}{m}-\frac{\vc{q}^2}{2m}}
2737: \left(\frac{1-\Pi_{\vc{k_1}}}{{\cal K}(\vc{q},\omega)}
2738: +\frac{\Pi_{\vc{k_1}}}{{\tilde {\cal K}}(\vc{q},\omega)}\right)
2739: \end{eqnarray}
2740: Plugging these results into the r.h.s. of
2741: (\ref{col1}) (\ref{col2}) and (\ref{col3}), the integral 
2742: terms do not appear anymore and we get the solution:
2743: \begin{eqnarray}\label{Ur}
2744: {\cal U}_\vc{q}(\vc{k},\vc{k_1},\omega)
2745: =\frac{i}{
2746: \omega+i0_+ -
2747: \frac{\vc{k}.\vc{q}}{m}-\frac{\vc{q}^2}{2m}}
2748: \left[\delta_{\vc{k},\vc{k_1}}+
2749: \frac{
2750: \frac{8\pi a }{mV}
2751: (n_{\vc{k}}-n_{\vc{k+q}}){\cal K}^{-1}_{\vc{k},\vc{k_1}}(\vc{q},\omega)}
2752: {
2753: \omega+i0_+ -
2754: \frac{\vc{k_1}.\vc{q}}{m}-\frac{\vc{q}^2}{2m}}
2755: \right]
2756: \end{eqnarray}
2757: where we define the reverse dielectric function
2758: \begin{eqnarray}\label{reverse}
2759: {\cal K}^{-1}_{\vc{k},\vc{k_1}}(\vc{q},\omega)
2760: =
2761: \frac{(1-\Pi_{\vc{k}})(1-\Pi_{\vc{k_1}})}
2762: {{\cal K}(\vc{q},\omega)}
2763: +
2764: \frac{(1-\Pi_{\vc{k}})\Pi_{\vc{k_1}}+
2765: \Pi_{\vc{k}}(1-\Pi_{\vc{k_1}})+
2766: \Pi_{\vc{k}}\Pi_{\vc{k_1}}\left(1-{\cal K}_n(\vc{q},\omega)/2\right)}
2767: {{\tilde {\cal K}}(\vc{q},\omega)}
2768: \end{eqnarray}
2769: This function can be rewritten in matrix notation, 
2770: where we distinguish in the column and the row 
2771: the channel $\vc{k_s}$ and $\vc{k_s}-\vc{q}$ from the 
2772: other channels:
2773: \begin{eqnarray}\label{reversem}
2774: {\cal K}^{-1}_{\vc{k},\vc{k_1}}(\vc{q},\omega)
2775: \equiv
2776: \left(
2777: \begin{array}{cc}
2778: {\cal K}^{-1}(\vc{q},\omega) & {\tilde {\cal K}}^{-1}(\vc{q},\omega) \\
2779: {\tilde {\cal K}}^{-1}(\vc{q},\omega) & 
2780: (1-\frac{{\cal K}_n(\vc{q},\omega)}{2})
2781: {\tilde {\cal K}}^{-1}(\vc{q},\omega) 
2782: \end{array}
2783: \right)
2784: \end{eqnarray}
2785: 
2786: \end{widetext}
2787: 
2788: \section{Calculation of the normal 
2789: dielectric function at equilibrium}
2790: 
2791: The expression (\ref{kn}) can be put in the form 
2792: of an integral over $t$:
2793: \begin{eqnarray}\label{kn1}
2794: {\cal K}_n(\vc{q},\omega)-1= 
2795: \nonumber \\
2796: i\frac{8\pi a }{mV}\sum_{\vc{k}}
2797: \int_0^\infty \!\!\!dt \,
2798: e^{i(\omega+i0_+ -
2799: \frac{\vc{k}.\vc{q}}{m}-\frac{\vc{q}^2}{2m})t}
2800: (n'^{eq}_{\vc{k}}-n'^{eq}_{\vc{k+q}})
2801: \end{eqnarray}
2802: Using the development:
2803: \begin{eqnarray}\label{kn2}
2804: n'^{eq}_{\vc{k}}=
2805: \sum_{j=1}^\infty
2806: e^{-\beta j(\frac{\vc{k}^2}{2m}-\mu)}
2807: \end{eqnarray}
2808: we can carry out the integration over the 
2809: momentum $\vc{k}$ to get:
2810: \begin{eqnarray}\label{kn3}
2811: {\cal K}_n(\vc{q},\omega)=
2812: 1+ \frac{16\pi a }{m}
2813: \int_0^\infty \!\!\!dt\,
2814: e^{i(\omega+i0_+)t}
2815: \sin(\frac{\vc{q}^2}{2m}t)
2816: \nonumber \\
2817: \sum_{j=1}^\infty
2818: \left(\frac{m}{2 \pi \beta j}\right)^{3/2}
2819: e^{\beta j \mu-\frac{\vc{q}^2 t^2}{2\beta j m}}
2820: \end{eqnarray}
2821: As far as the imaginary part is concerned 
2822: the integral over the time can be done 
2823: analytically. The result is:
2824: \begin{eqnarray}\label{kn4}
2825: {\rm Im}{\cal K}_n(\vc{q},\omega)
2826: =
2827: \frac{2a m}{\beta |\vc{q}|}
2828: \ln\left(\frac{1-e^{-\beta[
2829: m\frac{(\omega+\frac{\vc{q}^2}{2m})^2}{2\vc{q}^2}
2830: -\mu]}}
2831: {1-e^{-\beta[
2832: m\frac{(\omega-\frac{\vc{q}^2}{2m})^2}{2\vc{q}^2}
2833: -\mu]}}\right)
2834: %\\
2835: %&\stackrel{\vc{q}\rightarrow 0}{=}&
2836: %8a m c_B\frac{1}{e^{\beta m c_B^2/2}-1}
2837: \end{eqnarray}
2838: For the real part the expression, (\ref{kn3}) 
2839: needs to be developed in series. For  low temperature, 
2840: ${\rm Re}{\cal K}_n(\vc{q},\omega) \simeq 1$.
2841: %Eq.(\ref{kn3}) is approximated as:
2842: %\begin{eqnarray}\label{kn5}
2843: %{\rm Re}{\cal K}_n(\vc{q},\omega)&\simeq&
2844: %1+ \frac{16\pi a }{m}
2845: %\int_0^\infty dt
2846: %e^{i(\omega+i0_+)t}
2847: %\frac{\vc{q}^2}{2m}t
2848: %\nonumber \\
2849: %&&\sum_{j=1}^\infty
2850: %\left(\frac{m}{2 \pi \beta j}\right)^{3/2}
2851: %e^{\beta j \mu}
2852: %\nonumber \\
2853: %&=&1-
2854: %\frac{8\pi a n^{eq}_e}{m^2}
2855: %\frac{\vc{q}^2}{\omega^2}
2856: %\end{eqnarray}
2857: These results are in agreement with \cite{SK}
2858: but with the difference that we have included 
2859: the exchange term which doubles the interaction 
2860: strength.  
2861: 	
2862: \section{Calculation of the collision term}
2863: 
2864: As said in section 3, the calculation in the SRPA is 
2865: easy since it presents strong resemblance with the plasma gas.  
2866: The extension
2867: to GRPA is more complicated  because now the condensed and 
2868: normal modes cannot be treated on the same foot anymore, 
2869: as long as the interaction energy per particle is not 
2870: the same. Nevertheless, the method developed by 
2871: Ichimaru for a classical plasma can be adapted 
2872: straightforwardly. Here we shall give 
2873: the intermediate steps of the calculations. 
2874: 
2875: Before doing so, few symmetry properties are interesting:
2876: \begin{eqnarray}
2877: g_\vc{q}(\vc{k},\vc{k'})&=&g^*_\vc{q}(\vc{k-q},\vc{k'+q})
2878: \\
2879: %g_\vc{q}(\vc{k},\vc{k'})&=&
2880: %g_{\vc{k}+\vc{q}-\vc{k'}}(\vc{k'},\vc{k})
2881: %\nonumber \\ &&
2882: %+\delta_{\vc{k},\vc{k'}-\vc{q}}
2883: %n_{\vc{k}+\vc{q}}-
2884: %\delta_{\vc{q},0}n_{\vc{k}}
2885: %\\
2886: \left({\cal K}^{-1}_{\vc{k},\vc{k_1}}(\vc{q},\omega)\right)^*
2887: &=&{\cal K}^{-1}_{\vc{k},\vc{k_1}}(-\vc{q},-\omega)
2888: \end{eqnarray}
2889: The first  property can be checked from the definition 
2890: (\ref{g}) and the third one is straightforward.
2891: Using the first property and (\ref{n2}),
2892: the collision term defined in Eq.(\ref{K5}) can be expressed in terms 
2893: of the imaginary part of the correlation function:
2894: \begin{eqnarray}\label{B1}
2895: {\cal C}^T_{\vc{k}}[n_{\vc{k'}};\vc{k_s}]=
2896: \sum_{\vc{q},\vc{k'}}\frac{4\pi a}{mV}
2897: {\rm Im}(g_\vc{q}(\vc{k},\vc{k'})
2898: -g_\vc{q}(\vc{k}-\vc{q},\vc{k'}))
2899: \end{eqnarray}
2900: Let us define:
2901: \begin{eqnarray}
2902: A'(\vc{q},\omega)=
2903: \sum_{\vc{k}}(1-\Pi_{\vc{k}})
2904: \frac{(n_{\vc{k}}+1)n_{\vc{k}+\vc{q}}}
2905: {\omega+i0^+ -\frac{\vc{k}.\vc{q}}{m}-\frac{\vc{q}^2}{2m}}
2906: \end{eqnarray}
2907: \begin{eqnarray}
2908: A_0(\vc{q},\omega)
2909: =\sum_{\vc{k}} \Pi_{\vc{k}}
2910: \frac{(n_{\vc{k}}+1)n_{\vc{k}+\vc{q}}}
2911: {\omega+i0^+ -\frac{\vc{k}.\vc{q}}{m}-\frac{\vc{q}^2}{2m}}
2912: \end{eqnarray}
2913: With these definitions together with  
2914: (\ref{col34}), (\ref{Ur}) and (\ref{Q}), carrying out the sum over 
2915: $\vc{k_1}$ and $\vc{k'_1}$ and taking the thermodynamic limit,
2916: after a 
2917: straightforward but lengthly calculation we find the 
2918: following expressions for (\ref{K4}):
2919: \begin{widetext}
2920: \begin{eqnarray}\label{B2}
2921: \sum_{\vc{q},\vc{k'}}(1- \Pi_\vc{k})
2922: g_\vc{q}(\vc{k},\vc{k'})
2923: =
2924: -\sum_{\vc{q}}
2925: \int_{-\infty}^\infty
2926: \frac{d\omega}{2\pi i}
2927: \frac{1}{\omega+i0_+ -
2928: \frac{\vc{k}.\vc{q}}{m}-\frac{\vc{q}^2}{2m}}
2929: \bigg\{
2930: (n_{\vc{k}}+1)n_{\vc{k}+\vc{q}}
2931: \left(\frac{1}{{{\cal K}^*(\vc{q},\omega)}}-1 \right)
2932: \nonumber \\
2933: +
2934: \frac{8\pi a }{mV}(n_{\vc{k}}-n_{\vc{k}+\vc{q}})
2935: \bigg[\frac{(1-{{\cal K}^*(\vc{q},\omega)})A'(\vc{q},\omega)
2936: -A'^*(\vc{q},\omega)}{|{{\cal K}(\vc{q},\omega)}|^2}
2937: +
2938: \frac{A_0(\vc{q},\omega)(1-{\cal K}^*(\vc{q},\omega))-
2939: A^*_0(\vc{q},\omega)}
2940: {|{\tilde{\cal K}}(\vc{q},\omega)|^2}
2941: \bigg]
2942: \bigg\}
2943: \end{eqnarray}
2944: 
2945: \begin{eqnarray}\label{B3}
2946: \sum_{\vc{q},\vc{k'}}
2947: g_\vc{q}(\vc{k_s},\vc{k'})
2948: =
2949: -\sum_{\vc{q}}
2950: \int_{-\infty}^\infty
2951: \frac{d\omega}{2\pi i}
2952: \frac{1}{\omega+i0_+ -
2953: \frac{\vc{k_s}.\vc{q}}{m}-\frac{\vc{q}^2}{2m}}
2954: \bigg\{
2955: (n_{\vc{k_s}}+1)n_{\vc{k_s}+\vc{q}}
2956: \left(\frac{1}{{\tilde{{\cal K}}^*(\vc{q},\omega)}}-1 \right)
2957: \nonumber \\
2958: +
2959: \frac{8\pi a }{mV}(n_{\vc{k_s}}-n_{\vc{k_s+q}})
2960: \bigg[\frac{(1-{{\cal K}^*(\vc{q},\omega)})A'(\vc{q},\omega)
2961: -A'^*(\vc{q},\omega)}{{\tilde{{\cal K}}(\vc{q},\omega)}
2962: {{\cal K}^*(\vc{q},\omega)}}
2963: +
2964: (1-\frac{{\cal K}_n(\vc{q},\omega)}{2})
2965: \frac{A_0(\vc{q},\omega)(1-{\tilde{\cal K}}^*(\vc{q},\omega))-
2966: A^*_0(\vc{q},\omega)}
2967: {|{\tilde{\cal K}}(\vc{q},\omega)|^2}
2968: \bigg]
2969: \bigg\}
2970: \end{eqnarray}
2971: \end{widetext}
2972: Equivalently $\sum_{\vc{q},\vc{k'}}
2973: g_\vc{q}(\vc{k_s}-\vc{q},\vc{k'})$ is obtained by 
2974: substituting $\vc{k_s}$ by $\vc{k_s}-\vc{q}$ in (\ref{B3}). 
2975: The domain of integration over $\omega$ can be extended in the complex plane. 
2976: Since the  various reverse dielectric functions in (\ref{reversem}) 
2977: converge to unity when 
2978: $|\omega| \rightarrow \infty$, the integrand goes to zero in this 
2979: limit faster than $1/|\omega|$. As a consequence,  
2980: the integral over $\omega$ can be carried out by closing the 
2981: contour either in the upper half plane or in the lower half plane. 
2982: Since, we assume that the Bose gas is in a stable regime, the poles of 
2983: the dielectric functions have their imaginary part in the lower 
2984: half of the complex plane, while the complex conjugate of these 
2985: functions have their imaginary part in the upper half plane.  
2986: Thus in (\ref{B2}) and in (\ref{B3}), 
2987: contributions in which poles 
2988: only lie in one of these half planes are canceled since, in that case, 
2989: the contour of integration can be chosen in such a way that 
2990: no pole is surrounded.
2991: On the other hand, in order to 
2992: simplify the Eqs. (\ref{B2}) and (\ref{B3}), we can also add  
2993: arbitrary contributions in which poles
2994: lie only in one of these half planes. Taking into account these 
2995: considerations, we arrive at:
2996: \begin{widetext}
2997: \begin{eqnarray}\label{B5}
2998: \sum_{\vc{q},\vc{k'}}(1- \Pi_\vc{k})
2999: g_\vc{q}(\vc{k},\vc{k'})
3000: =
3001: -\sum_{\vc{q}}
3002: \int_{-\infty}^\infty
3003: \frac{d\omega}{2\pi i}
3004: \frac{1}{\omega+i0_+ -
3005: \frac{\vc{k}.\vc{q}}{m}-\frac{\vc{q}^2}{2m}}
3006: \bigg\{
3007: \frac{(n_{\vc{k}}+1)n_{\vc{k}+\vc{q}} 2i{\rm Im}{{\cal K}(\vc{q},\omega)}}
3008: {|{{\cal K}(\vc{q},\omega)}|^2}+ 
3009: \nonumber \\
3010: +
3011: \frac{8\pi a }{mV}(n_{\vc{k}}-n_{\vc{k}+\vc{q}})
3012: \bigg[\frac{2i {\rm Im}A'(\vc{q},\omega)
3013: }{|{{\cal K}(\vc{q},\omega)}|^2}
3014: +
3015: \frac{2i {\rm Im} A_0(\vc{q},\omega)}
3016: {|{\tilde{\cal K}}(\vc{q},\omega)|^2}
3017: \bigg]
3018: \bigg\}
3019: \end{eqnarray}
3020: \begin{eqnarray}\label{B6}
3021: \sum_{\vc{q},\vc{k'}}
3022: g_\vc{q}(\vc{k_s},\vc{k'})
3023: =
3024: -\sum_{\vc{q}}
3025: \int_{-\infty}^\infty
3026: \frac{d\omega}{2\pi i}
3027: \frac{1}{\omega+i0_+ -
3028: \frac{\vc{k_s}.\vc{q}}{m}-\frac{\vc{q}^2}{2m}}
3029: \bigg\{
3030: (n_{\vc{k_s}}+1)n_{\vc{k_s}+\vc{q}}
3031: 2i{\rm Im}\left(\frac{{\cal K}(\vc{q},\omega)}
3032: {{\tilde{\cal K}}^*(\vc{q},\omega){\cal K}(\vc{q},\omega)}\right)
3033: \nonumber \\
3034: +
3035: \frac{8\pi a}{mV}(n_{\vc{k_s}}-n_{\vc{k_s+q}})
3036: \bigg[\frac{2i {\rm Im} A'(\vc{q},\omega)}
3037: {{\tilde{{\cal K}}(\vc{q},\omega)}
3038: {{\cal K}^*(\vc{q},\omega)}}
3039: +
3040: (1-\frac{{\cal K}_n(\vc{q},\omega)}{2})
3041: 2i{\rm Im}\left(\frac{A_0(\vc{q},\omega)}
3042: {|{\tilde{\cal K}}(\vc{q},\omega)|^2}
3043: \right)
3044: \bigg]
3045: \bigg\}
3046: \end{eqnarray}
3047: \end{widetext}
3048: These expressions can be further simplified if we neglect 
3049: some irrelevant infinitesimal terms proportional to $0_+$. 
3050: Up to these infinitesimal term and using the 
3051: change of variable $\vc{k'}=\vc{k}+\vc{q}$ and the 
3052: formula ${\rm Im}[1/(x+i0_+)]=\pi \delta(x)$,
3053: we notice that for the two different limits 
3054: $n_{\vc{k_s}}/V \rightarrow 0$ or 
3055: $n_{\vc{k_s}}/V$ finite:
3056: \begin{eqnarray}
3057: {\rm Im}{{\cal K}(\vc{q},\omega)}
3058: ={\rm Im}
3059: {{\cal K}_n(\vc{q},\omega)}
3060: \nonumber \\ =
3061: \frac{8\pi a }{mV}\sum_{\vc{k'}}
3062: (n'_{\vc{k'-q}}-n'_{\vc{k'}})
3063: \pi\delta(\omega -
3064: \frac{\vc{k'}.\vc{q}}{m}+\frac{\vc{q}^2}{2m})
3065: \end{eqnarray}
3066: \begin{eqnarray}
3067: {\rm Im}\left(\frac{{\cal K}(\vc{q},\omega)}
3068: {{\tilde{\cal K}^*(\vc{q},\omega)}{{\cal K}(\vc{q},\omega)}}\right)
3069: =
3070: \frac{{\rm Im}{\cal K}(\vc{q},\omega)}
3071: {|{\tilde{\cal K}^*(\vc{q},\omega)}{{\cal K}(\vc{q},\omega)}|}
3072: \end{eqnarray}
3073: \begin{eqnarray}
3074: {\rm Im}\left(\frac{A_0(\vc{q},\omega)}
3075: {|{\tilde{\cal K}}(\vc{q},\omega)|^2}\right)
3076: =\frac{ {\rm Im} A_0(\vc{q},\omega)}
3077: {|{\tilde{\cal K}(\vc{q},\omega)}{{\cal K}^*(\vc{q},\omega)}|}
3078: \end{eqnarray}
3079: Finally, it remains to take the imaginary part 
3080: of these expressions, integrate them over $\omega$ and 
3081: plug the results into (\ref{B1}). 
3082: The integration is easy since it involves only delta 
3083: functions. After rearranging terms we notice that, 
3084: in Eq.(\ref{B6}), the term 
3085: proportional to $1-{\cal K}_n(\vc{q},\omega)/2$ 
3086: will not contribute. 
3087: Note that, concerning the term $g_\vc{q}(\vc{k}-\vc{q},\vc{k'})$, 
3088: we must carry out the  change of variable 
3089: $\vc{q} \rightarrow -\vc{q}$ and $\vc{k'}-\vc{q} 
3090: \rightarrow \vc{k'}$.
3091: In this way, we obtain 
3092: Eq.(\ref{K6}) and Eq.(\ref{K7}).
3093: 
3094: Using (\ref{col34}), a similar reasoning  
3095: allows also to find the expression:
3096: \begin{widetext}
3097: \begin{eqnarray}\label{intg}
3098: \sum_{\vc{k},\vc{k'}} 
3099: g_{\vc{q}}(\vc{k},\vc{k'})=
3100: -
3101: \int_{-\infty}^\infty
3102: \frac{d\omega}{2\pi i}
3103: 2i {\rm Im}
3104: \bigg[\frac{1}{{\cal K}(\vc{q},\omega)}
3105: \left(\frac{1}{{\cal K}^*(\vc{q},\omega)}-1\right)
3106: A'(\vc{q},\omega)
3107: +
3108: \frac{1}{{\tilde{\cal K}}(\vc{q},\omega)}
3109: \left(\frac{1}{{\tilde{\cal K}}^*(\vc{q},\omega)}-1\right)
3110: A_0(\vc{q},\omega)\bigg]
3111: \end{eqnarray}
3112: \end{widetext}
3113: 
3114: 
3115: \bigskip
3116: \centerline{\bf ACKNOWLEDGMENTS}
3117: PN thanks E. Cornell and  W. Ketterle for a useful discussion about 
3118: superfluidity, N. Cerf for encouraging me to do this work and 
3119: V. Belyi for remarks on the manuscript. 
3120: 
3121: PN acknowledges financial support from the Communaut\'e Fran\c caise de
3122: Belgique under grant
3123: ARC 00/05-251, from the IUAP programme of the Belgian
3124: government under grant V-18, from the EU under project RESQ
3125: (IST-2001-35759).
3126: 
3127: \begin{thebibliography}{99}
3128: 
3129: \bibitem{Walser}  R. Walser, J. Williams, J. Cooper, M. Holland,
3130: Phys. Rev. A, {\bf 59}, 3878 (1999);
3131: J. Wachter, R. Walser, J. Cooper \& M. Holland,
3132: Phys. Rev. A {\bf 64}, 053612 (2001).
3133: \bibitem{Zaremba} E. Zaremba, T. Nikuni, and A. Griffin,
3134: J. Low Temp. Phys. {\bf 116}, 227 (1999).
3135: \bibitem{Stoof} N.P. Proukakis, K. Burnett and H.T.C. Stoof,
3136: Phys. Rev. A {\bf 57}, 1230 (1998).
3137: \bibitem{HM}
3138: P.C. Hohenberg and P.C. Martin, Ann. Phys. (N.Y.)
3139: {\bf 34}, 291 (1965).
3140: \bibitem{LevichYakhot} E. Levich and V. Yakhot,
3141: J. Phys. A {\bf 11}, 2237 (1978).
3142: \bibitem{Kirkpatrick} T.R. Kirkpatrick and J.R. Dorfman,
3143: J. Low Temp. Phys. {\bf 58}, 308 (1985); {\bf 58}, 399 (1985).
3144: \bibitem{Gardiner} C.W. Gardiner and P. Zoller,
3145: Phys. Rev. A {\bf 61}, 033601 (2000).
3146: \bibitem{Proukakis}
3147: N.P. Proukakis, J. Phys B {\bf 34}, 4737 (2001).
3148: \bibitem{Stoof2}
3149: H.T.C. Stoof, J. Low Temp. Phys.,{\bf 114}, 11 (1999).
3150: \bibitem{Khalatnikov}
3151: I.M. Khalatnikov, {\it An introduction to the theory of superfluidity}
3152: (Benjamin, New York, 1965).
3153: \bibitem{Pomeau}
3154: Y. Pomeau, M. E. Brachet, S. Metens, and S. Rica,
3155: C. R. Acad. Sci. Paris IIb {\bf 327}, 791 (1999).
3156: \bibitem{Balescu} R. Balescu, {\it Equilibrium and non
3157: equilibrium statistical mechanics} (John Wiley \& Sons, New York, 1975).
3158: \bibitem{Leggett}
3159: A. J. Leggett, Rev. Mod. Phys. {\bf 73}, 307 (2001).
3160: \bibitem{Cornell}
3161: P. Engels, I. Coddington, P. C. Haljan, V. Schweikhard,
3162: and E. A. Cornell,
3163: Phys. Rev. Lett. {\bf 90}, 170405 (2003).
3164: \bibitem{Huang}
3165: K. Huang, {\it Statistical Mechanics}
3166: (John Wiley and Sons, New York, 1987).
3167: \bibitem{4eme} P. Navez, D. Bitouk, M. Gajda, Z. Idziaszek
3168: and K.  Rz\c{a}\.zewski, Phys. Rev. Lett. {\bf 79}, 1789 (1997);
3169: Z. Idziaszek, M. Gajda, P. Navez, M.
3170: Wilkens, and K. Rz\c{a}\.zewski, Phys. Rev. Lett. {\bf 82}, 4376 (1999);
3171: S. Giorgini, L. P. Pitaevskii, and S. Stringari,
3172: Phys. Rev. Lett. {\bf 80}, 5040-5043 (1998).
3173: \bibitem{LL} L.D. Landau  and E.M. Lifshitz,
3174: {Statistical Physics: Part 2}, (Pergamon Press, 1980).
3175: \bibitem{Pines} D. Pines, {\it The Many-Body Problem},
3176: (Benjamin, New York, 1961).
3177: \bibitem{Leggett2}
3178: A. J. Leggett, Rev. Mod. Phys. {\bf 71}, S318 (1999).
3179: \bibitem{Popov} Popov V.N., Sov. Phys.- JETP, {\bf 20} 1185 (1965).
3180: \bibitem{Girardeau} M. Girardeau and R. Arnowitt,
3181: Phys. Rev. {\bf 113}, 755 (1959).
3182: \bibitem{Griffin} A. Griffin, Phys. Rev. B, {\bf 53} 9341 (1996).
3183: \bibitem{Gapless}
3184: D.A.W. Hutchinson, K. Burnett, R.J. Dodd,
3185: S.A. Morgan, M. Rusch, E. Zaremba,
3186: N.P. Proukakis, M. Edwards and C.W. Clark,
3187: J. Phys. B {\bf 33}, 3825 (2000).
3188: \bibitem{PN} P. Navez, Mod. Phys. Lett. {\bf 12}, 705 (1998);
3189: M. Olshanii, L. Pricoupenko,
3190: Phys. Rev. Lett. {\bf 88}, 010402 (2002).
3191: \bibitem{NP2}
3192: P. Nozi\`eres, D. Pines,
3193: {\it The Theory of Quantum Liquids: Superfluid Bose Liquids (Vol. 2},
3194: (Perseus Book Group, 1990).
3195: \bibitem{Imamovic} M. Imamovic-Tomasovic, A. Griffin,
3196: J. Low Temp. Phys. {\bf 122}, 617 (2001). 
3197: %Phys. Rev. A {\bf 60}, 494-503 (1999).
3198: \bibitem{NP} D. Pines and P. Nozi\`eres, 
3199: {\it The theory of Quantum Liquids}, vol 1 (Benjamin, New York,1966).
3200: \bibitem{QBalescu} R. Balescu, Phys. of Fluids {\bf 4}, 94 (1961).
3201: \bibitem{WP} H.W. Wyld and D. Pines, Phys. Rev. {\bf 127},
3202: 1851 (1962).
3203: \bibitem{GCD}
3204: C.W. Gardiner, Phys. Rev. A {\bf 56}, 1414 (1997);
3205: Y. Castin and R. Dum, Phys. Rev. A {\bf 57}, 3008 (1998).
3206: \bibitem{SK}
3207: P. Sze\'pfalusy and I. Kondor,
3208: Ann. Phys. {\bf 82}, 1-53 (1974),
3209: J. Reidl, A. Csord\'as, R. Graham, and P. Sz\'epfalusy,
3210: Phys. Rev. A {\bf 61}, 043606 (2000).
3211: \bibitem{FR}
3212: E. Frieman and P. Rutherford,
3213: Ann. Phys. {\bf 82}, 134 (1964).
3214: \bibitem{instB}
3215: R. Balescu, J. Math. Phys. {\bf 4}, 1009 (1963).
3216: \bibitem{BZS}
3217: M.J. Bijlsma, E. Zaremba and, H.T.C. Stoof,
3218: Phys. Rev. A {\bf 62}, 063609 (2000).
3219: \bibitem{Keterlee2}
3220: H.-J. Miesner, D.M. Stamper-Kurn, M.R. Andrews,
3221: D.S. Durfee, S. Inouye, and W. Keterlee, Science {\bf 279},
3222: 1005 (1998).
3223: \bibitem{Gardiner2}
3224: C. W. Gardiner, M. D. Lee, R. J. Ballagh, M. J. Davis, and P. Zoller
3225: Phys. Rev. Lett. {\bf 81}, 5266-5269 (1998).
3226: \bibitem{Minguzzi}
3227: A. Minguzzi and M.P. Tosi, J. Phys: Cond. Mat., {\bf 9} 10211 (1997).
3228: \bibitem{GriffinB} A. Griffin, {Excitations in a
3229: Bose-Condensed Liquid} (University Press, Cambridge, 1993).
3230: \bibitem{Giorgini}
3231: S. Giorgini, Phys. Rev. A {\bf 61}, 063615 (2000).
3232: \bibitem{Ichimaru} S.Ichimaru, {\it
3233: Frontier in Physics: Statistical Plasma Physics, Vol I: Basic Principle}
3234: (Addison Wesley, 1992).
3235: \bibitem{GSS} see P.Nozi\`eres in
3236: A. Griffin, D.W. Snoke and S. Stringari,
3237: {\it Bose-Einstein Condensation}
3238: (Cambridge University Press, Cambridge, 1995).
3239: \bibitem{JZ}
3240: B. Jackson and E. Zaremba, Phys. Rev. Lett. {\bf 87}, 100404 (2001),
3241: B. Jackson and E.
3242: Zaremba, Phys. Rev. Lett. {\bf 88}, 180402 (2002), 
3243: B. Jackson and E. Zaremba, Phys. Rev. Lett. {\bf 89}, 150402 (2002).
3244: \end{thebibliography}
3245: 
3246: \end{document}
3247: