Freddy BOUCHET - ENS de Lyon and CNRS

CLIMATE AND STATISTICAL MECHANICS




STATISTICAL MECHANICS OF SYSTEMS WITH LONG RANGE INTERACTIONS

Keywords

Long range interactions, Self gravitating systems, Plasma physics, Geophysical Flows, Wave particle interactions, Large deviations, Ensemble Inequivalence, Phase transitions, Kinetic theory, Vlasov equation, Stochastic processes.

Context
Research Summary
Main Results
             Classification of phase transitions
             Large deviations for systems with long range interactions
             Kinetic theory and stochastic processes
             Polemic against non extensive statistical mechanics
             Generic ergodicity breaking
             Bicluster dynamics
             Long range interactions in cold atom experiments

Context


In a large number of physical systems, any single particle experiences a force which is dominated by interactions with far away particles. For instance, in a system with algebraic decay of the inter-particle potential V(r) ∼ 1/rα for large r, when α is less than the spatial dimension of the system, the interaction is long range (such interactions are sometimes called ``non-integrable’’). Such long range interacting systems are not-additive, as the interaction of any macroscopic part of the system with the whole is not negligible with respect to the internal energy of the given part.

The main physical examples of non-additive, long range interacting systems are: astrophysical self-gravitating systems, two-dimensional or geophysical fluid dynamics (vortices have a logarithmic interaction), some of the plasma physics models, models for wave-particle interactions, interaction due to multiple light scattering in cold atoms, or due to the dipolar force in optics, etc ... Spin systems and toy models with long range interactions have also been widely studied.

As a consequence of the lack of additivity, peculiar thermodynamic behaviors are likely to be observed in such Hamiltonian systems. For instance, the usual proof of the validity of the canonical ensemble, for a system in contact with a thermostat, or for a part of a bigger isolated system, uses explicitly this additivity property. Hertel and Thirring provided a toy model, mimicking self-gravitating dynamics, which displays inequivalence between canonical and microcanonical solutions, with negative specific heat regions in the microcanonical ensemble. Negative specific heat and ensemble inequivalence, previously known to astrophysicists, were then found in various fields: plasma physics and geophysical fluid dynamics. These examples show that new types of phase transitions are found in long range interacting systems.


Research summary

Recently, a new light was shed on the equilibrium statistical mechanics of these systems. The first reason is that a mathematical characterization (Ellis and col. 2001) of ensemble inequivalence and the study of several simple models have illustrated different types of inequivalence that might exist between the microcanonical and the canonical ensembles. The second is the appearance of a very useful technique, namely the large deviation theory, to compute the microcanonical number of microstates and thus the associated microcanonical entropy (Barré, Bouchet, Dauxois and Ruffo 2005). The third is a full classification of phase transitions and of ensemble inequivalence in such systems (Bouchet and Barré 2005). The last, but not the least reason, is the understanding that the broad spectrum of applications should be considered simultaneously since significant advances were performed independently in the different domains.

The dynamics of all of these systems also share deep analogies between each other. For instance, the analogies between the Vlasov equation and the 2D Euler equation have repeatedly been illustrated in the last forty years. The studies of the equilibrium states linear and nonlinear stabilities, is very similar for each of these models (Arnold’s type theorems) and are deeply related to the thermodynamics stability. Moreover the equilibrium fluctuations and the relaxation towards equilibrium is described by similar kinetic theories.

Recently, we have emphasized the role of Quasi-Stationnary-States in such systems, and explained their long time stability (Yamaguchi and col. 2003) in the framework of usual kinetic theory. We have theoretically explained and predicted, previously observed anomalous diffusion and non exponential relaxation, in these systems (Bouchet and Dauxois 2005). Both of these works, led in the framework of usual statistical mechanics, led to a strong controversy with the defenders of Tsallis non extensive statistics (see for instance Bouchet, Dauxois and Ruffo 2006). We have also proved that the relaxation towards equilibrium of 1D systems with long range interaction is not described by the usual quasi-linear kinetic theory (Lenard-Balescu equation) (Bouchet and Dauxois 2005).


Main results


Classification of phase transitions and ensemble inequivalence

Classification

Equilibrium phase diagram of an ensemble of self graviating stars, illustrating ensemble inequivalence: microcanonical phase transitions (black lines) do not coincide with canonical phase transitions (magenta lines) (see Bouchet-Barre-JStatPhys-2005)

Systems with long range interactions in general are not additive, which can lead to an inequivalence of the microcanonical and canonical ensembles. The microcanonical ensemble may show richer behavior than the canonical one, including negative specific heats and other non-common behaviors.
By analogy with the classical Landau classification of phase transitions, valid for systems with short range interactions ; in this work we proposed a classification of microcanonical phase transitions, of their link to canonical ones, and of the possible situations of ensemble inequivalence, in systems with long range interactions.
This classification is thus an exhaustive list of all possible phase transitions and situations of ensemble inequivalence. We emphasize on impossible situations, due to thermodynamical constraints.
We discuss previously observed phase transitions and inequivalence in self-gravitating, two-dimensional fluid dynamics and non-neutral plasmas. We note a number of generic situations that have not yet been observed in such systems. This opens the quest for observing such phase transitions in real physical systems or in models.

(see Bouchet-Barre-JStatPhys-2005).



Large deviations and systems with long range interactions
 

We have used large deviation estimations for the probability of macrostates in the microcanonical ensemble, in order to solve models with long-range interactions in the microcanonical and canonical ensemble. We show how this can be adapted to obtain the solution of a large class of simple models, which can show ensemble inequivalence. The model Hamiltonian can have both discrete (Ising,Potts) and continuous (HMF, Free Electron Laser) state variables. We treat both infinite range and slowly decreasing interactions and, in particular, we present the solution of the α−Ising model in onedimension with 0 ≤α < 1. This extremely simple model already displays ensemble inequivalence.
Large deviation estimations is a very powerful tool that has been then used in a number of other domains.

(see Barre-Bouchet-Dauxois-Ruffo-JStatPhys-2005).



Kinetic theory and nonlinear dynamics in systems with long range interactions

We have proposed a kinetic description of the Hamiltonian Mean Field model, which is paradigmatic for dynamical systems with long-range interactions, following the classical quasi-linear theory.

(see Yamaguchi-Barre-Bouchet-Dauxois-Ruffo-PhysicaD-2003, Bouchet-PRE-2004, Bouchet-Dauxois-PRE-2004, Yamaguchi-Bouchet-Dauxois-JStatMech-2007 and Jain-Bouchet-Mukamel-JStatMech-2007.).

In this framework, we have shown that generically, the Fokker Planck equation which governs fluctuations, has a non exponential relaxation and leads to anomalous behaviors. This mechanism is not limited to the HMF model. (see Bouchet-Dauxois-PRE-2004)
Studying this Fokker-Planck equation in details, we have predicted algebraic tails for the momentum auto-correlations and anomalous diffusion for the angles. We derive analytically the corresponding laws in the limit of a large number of particles. We argue that the mechanism for such an anomalous transport does not depend on some complex structure of the phase space: indeed, the transport is anomalous for out-of-equilibrium distributions but also for the equilibrium microcanonical distribution. (see Bouchet-Dauxois-PRE-2004 and Yamaguchi-Bouchet-Dauxois-JStatMech-2007)
We have also proved that the relaxation towards equilibrium of 1D systems with long range interaction is not described by the usual quasi-linear kinetic theory (Lenard-Balescu equation). (see Bouchet-Dauxois-PRE-2004)

With K. Jain and D. Mukamel, we have addressed the validity time for the Vlasov description of a N-body system with long range interactions. We have proved that the known lower bound t ∝ log(N) (Braun and Hepp) is actually optimal. We have exemplified this both in systems with continuous state variables (Hamiltonian dynamics) and discrete systems with Monte Carlo type dynamics. (see Jain-Bouchet-Mukamel-JStatMech-2007).



Polemic with C. Tsallis group against the “Non extensive statistical mechanics”

The classical equilibrium statistical mechanics used in all the previous works, is based on the classical and natural hypothesis of averaging with respect to a uniform distribution, on an energy shell, of the N particle phase space (microcanonical distribution). For out of equilibrium problems, our approach deeply rely on the actual dynamics of the system (kinetic theory).
By contrast, an other school uses maximization of a baseless functional, called “non extensive entropy”. Then by logically meaningless considerations, but using analogies, they develop a whole “theory” similar to the classical statistical mechanics.
This point of view concerning the "Non extensive statistical mechanics" may seem extreme, but it is in accordance with the reality, at least as far as I understand it.
Because our works  led to contradictions with previous statements of C. Tsallis’ group, the publication of the latter has been very difficult. This also led to a comment to an article published in Europhysics news. 

(see Yamaguchi-Barre-Bouchet-Dauxois-Ruffo-PhysicaD-2003, Bouchet-PRE-2004, Bouchet-Dauxois-PRE-2004 and Bouchet-Dauxois-Ruffo-EurophysicsNews-2006)



Generic ergodicity breaking in systems with long range interactions

With D. Mukamel, T. Dauxois and S. Ruffo, we have argued that, in systems with long range interactions, in addition with ensemble inequivalence, we may generically observe “strong ergodicity breaking” (the phase space is no more connected). We have exemplified this on very simple models.

(see Bouchet-Dauxois-Mukamel-Ruffo-PRE-2008)



Bicluster dynamics

(see Barre-Bouchet-Dauxois-Ruffo-PRL-2002 and Barre-Bouchet-Dauxois-Ruffo-EPJB-2002)


Long range interactions and cold atom experiments

(see Olivetti-Barre-Marcos-Bouchet-Kaiser-PRL-2009 )