cond-mat0612219/prl2.tex
1: %\documentclass[preprint,showpacs,prl]{revtex4}
2: \documentclass[twocolumn,showpacs,prl]{revtex4}
3: \usepackage{amsmath,amssymb,graphicx,color}
4: \usepackage{color}
5: 
6: \begin{document}
7: 
8: \title{Exploring the thermodynamic limit of Hamiltonian models: convergence to the Vlasov equation.}
9: 
10: \author{Andrea Antoniazzi$^{1}$\thanks{andrea.antoniazzi@unifi.it},
11: Francesco Califano$^
12: {2}$\thanks{francesco.califano@df.unipi.it}, 
13: Duccio Fanelli$^{1,3}$\thanks{duccio.fanelli@ki.se}, 
14: Stefano Ruffo$^{1}$\thanks{stefano.ruffo@unifi.it}}
15: 
16: \affiliation{ 1. Dipartimento di Energetica and CSDC, Universit\`a di
17:   Firenze, and INFN, via S. Marta, 3, 50139 Firenze, Italy\\
18:   2. Dipartimento di Fisica "E.Fermi" and CNISM, Universit\`a di
19:   Pisa, Largo Pontecorvo, 3 56127 Pisa, Italy\\
20:   3. Theoretical Physics, School of Physics and Astronomy, University of Manchester, Manchester M13 9PL, United Kingdom} \date{\today}
21: 
22: 
23: \begin{abstract}
24: We here discuss the emergence of Quasi Stationary States (QSS), a universal feature of systems 
25: with long-range interactions. With reference to the Hamiltonian Mean
26: Field (HMF) model, numerical simulations are performed based on both the original $N$-body
27: setting and the continuum Vlasov model which is supposed to hold in the thermodynamic 
28: limit. A detailed comparison unambiguously demonstrates that the Vlasov-wave system provides the correct 
29: framework to address the study of QSS. Further, analytical calculations based on Lynden-Bell's
30: theory of violent relaxation are shown to result in accurate predictions. Finally, in specific regions of parameters space, 
31: Vlasov numerical solutions are shown to be affected by small scale fluctuations, a finding that 
32: points to the need for novel schemes able to account for particles correlations.  
33: \end{abstract}
34: 
35: \pacs{  
36: {52.65.Ff}{ Fokker Planck and Vlasov equations;}
37: {05.45.-a}{ Nonlinear dynamics and nonlinear dynamical systems;}
38: {05.20.-y}{ Classical statistical mechanics;}
39: }
40: 
41: \maketitle
42: 
43: 
44: The Vlasov equation constitutes a universal theoretical framework and plays a role of paramount 
45: importance in many branches of applied and fundamental physics. Structure formation in the 
46: universe is for instance a rich and fascinating problem of classical physics: The fossile radiation 
47: that permeates the cosmos is a relic of microfluctuation in the matter created by the Big Bang, and 
48: such a small perturbation is believed to have evolved via gravitational instability to the pronounced 
49: agglomerations that we see nowdays on the galaxy cluster scale. Within this scenario, gravity is hence 
50: the engine of growth and the Vlasov equation governs the dynamics of the non baryonic ``dark matter" \cite{peebles}. 
51: Furthermore, the continuous Vlasov description is the reference model for several space and laboratory plasma applications, 
52: including many interesting regimes, among which the interpretation of coherent electrostatic structures 
53: observed in plasmas far from thermodynamic equilibrium. The Vlasov equation is obtained as the mean--field 
54: limit of the $N$--body Liouville equation, assuming that each particle interacts with an average field generated by all 
55: plasma particles (i.e. the mean electromagnetic field determined by the 
56: Poisson or Maxwell equations where the charge and current densities are 
57: calculated from the particle distribution function) while inter--particle 
58: correlations are completely neglected.
59: 
60: %Finally, semiclassical Vlasov
61: %equation and the Feynman path integral method are being used as an effective many-body description of nuclear
62: %dynamics below the Coulomb barrier \cite{bonasera}. 
63: 
64: 
65: Numerical simulations are presently one of the most powerful resource 
66: to address the study of the Vlasov equation.
67: %such physical systems and represent a major
68: %challenge in computational physics. 
69: In the plasma context, the Lagrangian
70: Particle-In-Cell approach is by far the most popular, while 
71: Eulerian Vlasov codes are particularly suited for 
72: analyzing specific model problems, due to the associated low 
73: noise level which is secured even in the non--linear 
74: regime \cite{mangeney}. However, any numerical scheme designed to integrate the 
75: continuous Vlasov system involves a 
76: discretization over a finite mesh. This is indeed an  
77: unavoidable step which in turn affects numerical accuracy.  
78: A numerical (diffusive and dispersive) characteristic length is in fact introduced 
79: being at best comparable with the grid mesh size: as soon as the latter matches the 
80: typical length scale relative to the (dynamically generated) fluctuations 
81: a violation of the continuous Hamiltonian character of the equations occurs 
82: (see Refs. \cite{califano}). It is important to emphasize that
83: even if such {\em non Vlasov} effects are strongly localized 
84: (in phase space), the induced large scale topological changes will 
85: eventually affect the system globally. Therefore, aiming at clarifying the 
86: problem of the validity of Vlasov numerical models, it is crucial to compare
87: a continuous Vlasov, but numerically discretized, approach 
88: to a homologous N-body model.  
89: 
90: Vlasov equation has been also invoked as a reference model in many interesting one dimensional problems,
91: and recurrently applied to the study of wave-particles interacting systems. The Hamiltonian Mean Field 
92: (HMF) model~\cite{antoni-95}, describing the coupled motion of $N$ rotators, is in particular assimilated to 
93: a Vlasov dynamics in the thermodynamic limit on the basis of rigorous results~\cite{BraunHepp}. 
94: The HMF model has been historically introduced as representing 
95: gravitational and charged sheet models and is quite extensively analyzed as a paradigmatic 
96: representative of the broader class of systems with long-range interactions~\cite{Houches02}. 
97: %This system displays a gallery of peculiar behaviours ~\cite{Houches02} including {\it inequivalence} between microcanonical 
98: %and canonical ensemble, physically manifested through negative specific heat and temperature jumps,
99: %{\it broken ergodicity} and the existence of  {\it metastable states}.  
100: A peculiar feature of the HMF model, shared also by other long-range interacting systems, is the presence of
101: {\it Quasi Stationary States} (QSS). During time evolution, the system gets trapped in such states, which
102: are characterized by non Gaussian velocity distributions, before relaxing to the final Boltzmann-Gibbs 
103: equilibrium \cite{ruffo_rapisarda}. An attempt has been made \cite{rapisarda_tsallis} to interpret the 
104: emergence of QSSs by invoking Tsallis statistics \cite{Tsallis}. This approach has been later on 
105: criticized in \cite{Yamaguchi}, where QSSs were shown to
106: correspond to stationary stable solutions of the Vlasov equation, for a particular choice of the initial condition.  
107: More recently, an approximate analytical theory, based on the Vlasov equation, which derives the QSSs of the HMF model 
108: using a maximum entropy principle, was developed in ~\cite{antoniazziPRL}. This theory 
109: is inspired by the pioneering work of Lynden-Bell ~\cite{LyndenBell68} and relies on previous work
110: on 2D turbulence by Chavanis \cite{chava2D}. However, the underlying 
111: Vlasov ansatz has not been directly examined and it is recently being debated \cite{EPN}. 
112: 
113: In this Letter, we shall discuss numerical simulations of the continuous Vlasov model, the kinetic counterpart
114: of the discrete HMF model. By comparing these results to both direct N-body simulations and analytical predictions, 
115: we shall reach the following conclusions:
116: (i)  the Vlasov formulation is indeed ruling the dynamics of the QSS; (ii)  the proposed 
117: analytical treatment of the Vlasov equation is surprisingly accurate, despite the approximations involved in the
118: derivation;  (iii) Vlasov simulations are to be handled with extreme caution when exploring 
119: specific regions of the parameters space. 
120: 
121: The HMF model is characterized by the following Hamiltonian
122: \begin{equation}
123: \label{eq:ham}
124: H = \frac{1}{2} \sum_{j=1}^N p_j^2 + \frac{1}{2 N} \sum_{i,j=1}^N 
125: \left[1 -  \cos(\theta_j-\theta_i) \right]
126: \end{equation}
127: where $\theta_j$ represents the orientation of the $j$-th rotor and
128: $p_j$ is its conjugate momentum. To monitor the evolution of the
129: system, it is customary to introduce the magnetization, a macroscopic order
130: parameter defined as $M=|{\mathbf M}|=|\sum {\mathbf m_i}| /N$, where
131: ${\mathbf m_i}=(\cos \theta_i,\sin \theta_i)$ stands for the microscopic
132: magnetization vector. As previously reported \cite{antoni-95}, after an initial transient, 
133: the system gets trapped into Quasi-Stationary States (QSSs), i.e. 
134: non-equilibrium dynamical regimes  whose lifetime diverges when increasing 
135: the number of particles $N$. Importantly, when performing the mean-field limit ($N
136: \rightarrow \infty$) {\it before} the infinite time limit, the system
137: cannot relax towards Boltzmann--Gibbs equilibrium and remains
138: permanently confined in the intermediate QSSs. 
139: %In other words, 
140: %the latter manifest because the infinite time and 
141: %the thermodynamic limits do not commute.
142: As mentioned above, this phenomenology is widely observed for systems with long-range
143: interactions, including galaxy dynamics~\cite{Padmanabhan}, free electron lasers~\cite{Barre},
144: 2D electron plasmas~\cite{kawahara}.
145: 
146: In the $N \to \infty$ limit
147: %i.e. when the system is indefinitely stuck in the QSSs, 
148: the discrete HMF dynamics reduces to the Vlasov equation
149: \begin{equation}
150: \partial f / \partial t + p \, \partial f / \partial \theta \,\, 
151: - (dV / d \theta ) \, \partial f / \partial p = 0 \, , 
152: %\frac{\partial f}{\partial t} + p\frac{\partial f}{\partial \theta} -
153: %\frac{d V}{d \theta} \frac{\partial f}{\partial p}=0\quad ,
154: \label{eq:VlasovHMF}
155: \end{equation}
156: where $f(\theta,p,t)$ is the microscopic one-particle
157: distribution function and
158: \begin{eqnarray}
159: V(\theta)[f] &=& 1 - M_x[f] \cos(\theta) - M_y[f] \sin(\theta) ~, \\
160: M_x[f] &=& \int_{-\pi}^{\pi} \int_{-\infty}^{\infty}  f(\theta,p,t) \, \cos{\theta}  {\mathrm d}\theta
161: {\mathrm d}p\quad , \\
162: M_y[f] &=& \int_{-\pi}^{\pi} \int_{\infty}^{\infty}  f(\theta,p,t) \, \sin{\theta}{\mathrm d}\theta
163: {\mathrm d}p\quad .
164: \label{eq:pot_magn}
165: \end{eqnarray}
166: The specific energy $h[f]=\int \int (p^2/{2}) f(\theta,p,t) {\mathrm d}\theta
167: {\mathrm d}p - ({M_x^2+M_y^2 - 1})/{2}$ and momentum
168: $P[f]=\int \int p f(\theta,p,t) {\mathrm d}\theta
169: {\mathrm d}p$ functionals are conserved quantities. Homogeneous states are characterized by $M=0$, 
170: while non-homogeneous states correspond to $M \ne 0$. 
171: 
172: Rigorous mathematical results \cite{BraunHepp} demonstrate that, indeed, the Vlasov framework applies in the continuum 
173: description of mean-field type models. This observation corroborates the claim that any 
174: theoretical attempt to characterize the QSSs should resort to the above Vlasov based 
175: interpretative picture. Despite this, the QSS non-Gaussian velocity
176: distributions have been {\em fitted}~\cite{rapisarda_tsallis} using Tsallis' $q$--exponentials,
177: and the Vlasov formalism assumed valid {\em only} for the limiting case of homogeneous initial 
178: conditions \cite{EPN}. In a recent paper \cite{antoniazziPRL}, the aforementioned 
179: velocity distribution functions were instead reproduced with an analytical expression derived from the Vlasov scenario, 
180: with no adjustable parameters and for a large class of initial conditions, including inhomogeneous ones. 
181: The key idea dates back to the seminal work by Lynden-Bell \cite{LyndenBell68} (see also \cite{Chavanis06},
182: \cite{Michel94}) and consists in coarse-graining the microscopic one-particle distribution function
183: $f(\theta,p,t)$ by introducing a local average in phase space. It is then possible to
184: associate an entropy to the coarse-grained distribution $\bar{f}$: The corresponding statistical equilibrium
185: is hence determined by maximizing such an entropy, while imposing the conservation of the Vlasov dynamical invariants, 
186: namely energy, momentum and norm of the distribution. We shall here limit our discussion to the case 
187: of an initial single particle distribution which takes only two distinct values: $f_0=1/(4
188: \Delta_{\theta} \Delta_{p})$, if the angles (velocities) lie within an
189: interval centered around zero and of half-width $\Delta_{\theta}$
190: ($\Delta_{p}$), and zero otherwise. This choice corresponds to the
191: so-called ``water-bag" distribution which is fully specified by energy
192: $h[f]=e$, momentum $P[f]=\sigma$ and the initial magnetization ${\mathbf
193: M_0}=(M_{x0}, M_{y0})$.  The maximum entropy calculation is then performed analytically \cite{antoniazziPRL}
194: and results in the following form of the QSS distribution 
195: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
196: \begin{equation}
197: \label{eq:barf} \bar{f}(\theta,p)= f_0\frac{e^{-\beta (p^2/2
198: - M_y[\bar{f}]\sin\theta
199: - M_x[\bar{f}]\cos\theta)-\lambda p-\mu}}
200: {1+e^{-\beta (p^2/2  - M_y[\bar{f}]\sin\theta
201:  - M_x[\bar{f}]\cos\theta)-\lambda p-\mu}}
202: \end{equation}
203: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
204: where $\beta/f_0$, $\lambda/f_0$ and $\mu/f_0$ are rescaled Lagrange multipliers, respectively associated to the
205: energy, momentum and normalization. Inserting expression (\ref{eq:barf}) into the above 
206: constraints and recalling the definition of $M_x[\bar{f}]$, $M_y[\bar{f}]$, one obtains an implicit system 
207: which can be solved numerically to determine the Lagrange multipliers and the expected magnetization in the QSS. Note that the
208: distribution (\ref{eq:barf}) differs from the usual Boltzmann-Gibbs expression because of the
209: ``fermionic'' denominator. Numerically computed velocity distributions have been compared in \cite{antoniazziPRL} 
210: to the above theoretical predictions (where no free parameter is used), obtaining an overall good agreement.
211: %Although not a single free parameter was used, an excellent overall agreement was found in the tails of the
212: %distribution. 
213: However, the central part of the distributions is modulated by the presence of two symmetric bumps, which 
214: are the signature of a collective dynamical phenomenon \cite{antoniazziPRL}. The presence of these bumps is not 
215: explained by our theory. Such discrepancies has been recently claimed to be an indirect proof of the fact 
216: that the Vlasov model holds only approximately true.
217: %with reference 
218: %to the specific HMF case, thus justifying the search for novel theoretical frameworks, alternative to conventional 
219: %statistical mechanics. 
220: We shall here demonstrate that this claim is not correct and that the deviations between theory 
221: and numerical observation are uniquely due to the approximations built in the Lynden-Bell approach. 
222: 
223: A detailed analysis of the Lynden-Bell equilibrium (\ref{eq:barf}) in the parameter plane 
224: $(M_{0},e)$ enabled us to unravel a rich phenomenology, including out of equilibrium phase transitions 
225: between homogeneous ($M_{QSS}=0$)  and non-homogeneous ($M_{QSS} \ne 0$) QSS states. 
226: Second and first order transition lines are found that separate homogeneous and non homogeneous states and
227: merge into a tricritical  point approximately located in $(M_{0},e)=(0.2,0.61)$. When the transition is 
228: second order two extrema of the Lynden-Bell entropy are identified in the inhomogeneous phase: the solution  
229: $M=0$ corresponds to a saddle point, being therefore unstable; the global maximum is 
230: instead associated to $M \neq 0$, which represents the equilibrium predicted by the theory. This argument is
231: important for what will be discussed in the following.
232: 
233: Let us now turn to direct simulations, with the aim of testing the above scenario, and focus first on the kinetic model 
234: (\ref{eq:VlasovHMF})--(\ref{eq:pot_magn}). The algorithm solves the Vlasov equation in phase space and uses the 
235: so-called ``splitting scheme", a widely adopted strategy in numerical fluid dynamics. Such a scheme, pioneered by
236: Cheng and Knorr \cite{Cheng}, was first applied to the study of the Vlasov-Poisson equations in the electrostatic 
237: limit and then employed for a wide spectrum of problems \cite{califano}. For different values of the pair 
238: $(M_{0},e)$, which sets the widths of the initial water-bag profile, we performed a direct integration of 
239: the Vlasov system  (\ref{eq:VlasovHMF})--(\ref{eq:pot_magn}). After a transient, magnetization is 
240: shown to eventually attain a constant value, which corresponds to the QSS value observed in
241: the HMF, discrete, framework. The asymptotic magnetizations are hence recorded when varying the 
242: initial condition. Results (stars) are reported in figure \ref{fig1}(a) where $M_{QSS}$ is plotted as function of $e$. 
243: A comparison is drawn with the predictions of our theory (solid line) and with the outcome of N-body simulation (squares) 
244: based on the Hamiltonian (\ref{eq:ham}), finding an excellent agreement. 
245: This observation enables us to conclude that (i) the Vlasov equation governs the HMF dynamics for 
246: $N \to \infty$ {\it both} in the homogeneous and non homogeneous case; (ii) Lynden-Bell's violent relaxation 
247: theory allows for reliable predictions, including the transition from magnetized to non-magnetized states.  
248: 
249: Deviations from the theory are detected near the transition. 
250: This fact has a natural explaination and raises a number of fundamental questions related to the use of 
251: Vlasov simulations. As confirmed by the inspection of figure \ref{fig1}(b), close to the transition point, 
252: the entropy $S$ of the Lynden-Bell coarse-grained distribution takes almost the same value when evaluated 
253: on the global maximum (solid line) or on the saddle point (dashed line). The entropy 
254: is hence substantially flat in this region,  which in turn implies that there exists an extended basin 
255: of states accessible to the system. This interpretation is further validated by the inset of figure \ref{fig1}(a), 
256: where we show the probability distribution of $M_{QSS}$ computed via N-body simulation. The bell-shaped profile 
257: presents a clear peak, approximately close to the value predicted by our theory. Quite remarkably, 
258: the system can converge to final magnetizations which are sensibly different from the expected value. 
259: Simulations based on the Vlasov code running at different resolutions (grid points) confirmed 
260: this scenario, highlighting a similar degree of variability. These findings point to the fact that in 
261: specific regions of the parameter space, Vlasov numerics needs to be carefully analyzed (see also Ref.~\cite{Elskens}). 
262: Importantly, it is becoming nowadays crucial to step towards an ``extended«« Vlasov theoretical model which enables 
263: to account for discreteness effects, by incorporating at least two particles correlations interaction term. 
264: 
265: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
266: \begin{figure}[htbp]
267:   \centering
268: %  \vspace*{2.5em}
269:   \includegraphics[width=7cm]{fig1.eps}
270:   \caption{Panel (a): The magnetization in the QSS is plotted as function of energy, $e$, at $M_0=0.24$. 
271:   The solid line refers to the Lynden-Bell inspired theory. Stars (resp. squares) stand for Vlasov 
272:   (resp. N-body) simulations. Inset: Probability distribution of  
273:   $M_{QSS}$ computed via N-body simulation (the solid line is a Gaussian fit).    
274:   Panel (b): Entropy $S$ at the stationary points, as function of energy, $e$: 
275:   magnetized solution (solid line) and non--magnetized one (dashed line).}
276:   \label{fig1}
277: \end{figure}
278: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
279: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
280: \begin{figure}[htbp]
281: \centering
282: \includegraphics[width=7cm]{fig2.ps}
283: %  \vspace*{2.5em}
284: \caption{Phase space snapshots for $(M_{0},e)=(0.5,0.69)$.}
285: \label{fig2}
286: \end{figure}
287: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
288: Qualitatively, one can track the evolution of the system in phase space, both for the homogeneous and non 
289: homogeneous cases. Results of the Vlasov integration are displayed in figure \ref{fig2} for  $(M_{0},e)=(0.5,0.69)$,  
290: where the system is shown to evolve towards a non magnetized QSS. The initial water-bag 
291: distribution splits into two large resonances, which persist asymptotically: the latter 
292: acquire constant opposite velocities which are maintained during the subsequent time evolution, in agreement with the findings of 
293: \cite{antoniazziPRL}. The two bumps are therefore an emergent property of the model, which is correctly 
294: reproduced by the Vlasov dynamics. For larger values of the initial magnetization ($M_{0}>0.89$), while keeping $e=0.69$, the
295: system evolves towards an asymptotic magnetized state, in agreement with the theory. In this case 
296: several resonances are rapidly developed which eventually coaelesce giving rise to complex patterns in phase space. 
297: More quantitatively, one can compare the velocity distributions resulting from, 
298: respectively, Vlasov and N-body simulations. The curves are diplayed in figure \ref{fig3} (a),(b),(c) for various 
299: choices of the initial conditions in the magnetized region. The agreement is excellent, thus reinforcing our former 
300: conclusion about the validity of the Vlasov model. 
301: Finally, let us stress that, when $e=0.69$, the two solutions (resp. magnetized and non magnetized)  
302: \cite{antoniazziPRL} are associated to a practically indistinguishible entropy level 
303: (see figure \ref{fig3} (d)). As previously discussed, the system explores an almost flat entropy landscape 
304: and can be therefore be stuck in local traps, because of finite size effects. A pronounced variability 
305: of the measured $M_{QSS}$ is therefore to be expected. 
306: %This observation contributes to 
307: %definitely solve the puzzle about the pecularity of the regime $e=0.69$, a reference value for the energy that has 
308: %attracted the interests of theoreticians during the last decade \cite{Houches02}. 
309: %\vspace*{2.0em}
310: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5
311: \begin{figure}[htbp]
312:   \centering
313: %  \vspace*{2.0em}
314:   \includegraphics[width=7cm]{fig3.eps}
315:   \vspace*{2.0em}
316:   \caption{Symbols: velocity distributions computed via N-body simulations. Solid line: velocity distributions 
317:   obtained through a direct integration of the Vlasov equation. Here $e=0.69$ and $M_0=0.3$ (a), $M_0=0.5$ (b), 
318:   $M_0=0.7$ (c). Panel (d): Entropy at the stationary points as a function of the initial magnetization: 
319:   the solid line refers to the global maximum, while the dotted line to the saddle point.}
320:   \label{fig3}
321: \end{figure}
322: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
323: 
324: In this Letter, we have analyzed the emergence of QSS, a universal feature that occurs in systems with long-range
325: interactions, for the specific case of the HMF model. By comparing numerical simulations and analytical predictions, 
326: we have been able to unambiguously demonstrate that the Vlasov model provides an accurate framework to address 
327: the study of the QSS. Working within the Vlasov context one can develop a fully predictive theoretical approach, which is
328: completely justified from first principles. Finally, and most important, results of conventional Vlasov codes are to be
329: critically scrutinized, especially in specific regions of parameters space close to transitions from homogeneous
330: to non homogeneous states.
331: 
332: We acknowledge financial support from the PRIN05-MIUR project {\it Dynamics and thermodynamics of systems with long-range 
333: interactions}.
334: \vspace*{-0.6cm}
335: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
336: \begin{thebibliography}{99}
337: \vspace*{-0.3cm}
338: \bibitem{peebles} P.J. Peebles, \emph{The Large-scale structure of the Universe}, Princeton, NJ: Princeton
339: Universeity Press (1980).
340: %\bibitem{bonasera} A. Bonasera et al., J. Phys. {\bf 23} 1297-1302 (1997) 
341: \bibitem{mangeney} A. Mangeney et al., J. Comp. Physics {\bf 179}, 495 (2002).
342: \bibitem{califano} L. Galeotti et al., Phys. Rev. Lett. {\bf 95}, 
343: 015002 (2005); F. Califano et al., Phys. Plasmas {\bf 13}, 082102 (2006). 
344: \bibitem{antoni-95} M.~Antoni et al., Phys.~Rev.~E \textbf{52},
345:   2361 (1995).
346: \bibitem{BraunHepp} W. Braun et al., Comm.  Math. Phys. \textbf{56}, 101 (1977).
347:  \bibitem{Houches02} T.~Dauxois et al., Lect. Not. Phys. 
348: {\bf 602}, Springer (2002).
349: \bibitem{ruffo_rapisarda} V. Latora et al. Phys. Rev. Lett. \textbf{80}, 692 (1998).
350: \bibitem{rapisarda_tsallis} V. Latora et al. Phys. Rev. E \textbf{64} 056134 (2001).
351: \bibitem{Tsallis} C. Tsallis, J. Stat. Phys. {\bf 52}, 479 (1988).
352: \bibitem{Yamaguchi} Y.Y. Yamaguchi et al. Physica A, {\bf 337}, 36 (2004).
353: \bibitem{antoniazziPRL} A.~Antoniazzi et al., Phys. Rev. E \textbf{75} 011112 (2007); 
354: P.H. Chavanis Eur. Phys. J. B {\bf 53}, 487 (2006).
355: \bibitem{LyndenBell68} D. Lynden-Bell et al., Mon. Not. R. Astron. Soc.
356: {\bf 138}, 495 (1968).
357: \bibitem{chava2D} P.H. Chavanis, Ph. D Thesis, ENS Lyon (1996).
358: \bibitem{EPN} A. Rapisarda et al., Europhysics News, {\bf 36}, 202 (2005); 
359: F. Bouchet et al., Europhysics News, {\bf 37}, 2, 9-10 (2006).
360: \bibitem{Padmanabhan} T. Padmanabhan, Phys. Rep. {\bf 188}, 285 (1990).
361: \bibitem{Barre} J. Barr{\'e} et al., Phys. Rev E {\bf 69}, 045501(R) (2004).
362: \bibitem{kawahara} R. Kawahara and H. Nakanishi, cond-mat/0611694.
363: \bibitem{Chavanis06}  P.~H.~Chavanis, {Physica A} \textbf{359}, 177 
364: (2006).
365: \bibitem{Michel94} J. Michel et al., Comm. Math. Phys. 
366: \textbf{159}, 195 (1994).
367: \bibitem{Cheng} C.G. Cheng and G. Knorr, J. Comp. Phys. {\bf 22}, 330 (1976).
368: \bibitem{Elskens} M.C. Firpo, Y. Elskens, Phys. Rev. Lett. {\bf 84}, 3318 (2000).
369: \end{thebibliography}
370: 
371: \end{document}
372: 
373: 
374: