Eric Bertin 


Research interests My research interests mainly deal with nonequilibrium statistical physics. Among other things, I am specifically interested in the statistical physics of nonHamiltonian systems, that is of assemblies of entities having nonHamiltonian (i.e., nonconservative) interactions. This wide class of systems includes for instance active particules (a physicist concept used to model groups of animals, bacteria, or robots), granular matter, foams, fluid turbulence, coupled oscillators or models of social behavior, to quote a few examples. Such apparently very different systems are often described within rather distinct approaches. The aim of my research work is to try to formulate at least the questions, and perhaps some solutions, within a unified framework.One of the key ingredients of this approach is to consider models of nonconservative particles that reduces to conservative ones in a specific limit. In this way, the deviation from standard statistical equilibrium can be identified, and possibly quantified using concepts like the entropy difference with the "closest" equilibrium state. These deviations can have several origins: dissipation in many cases, but also more exotic reasons like in the case of (very simplified) social agents who try to maximize their own welfare rather than the collective one (in contrast to Hamiltonian particles whose evolution is driven by a global energy function). Along this general line of thought, an interesting issue is also to determine relevant macroscopic parameters to describe such systems, for instance by trying to generalize concepts like temperature or chemical potential. Beside this research axis, I'm also interested in probabilistic issues like limit distributions for sums and extreme values of sets of random variables. Though already wellknown in many cases (in particular in the absence of correlations), problems of sums and extreme values have intriguing relations between one another. In addition, they can both be reformulated in an elegant way using renormalization group concepts, leading for instance to straightforward predictions for finitesize effects. But above all this, such probabilistic problems are probably the most simple instances where universal macroscopic properties (the asymptotic distributions) emerge out of microscopic degrees of freedom whose details are irrelevant. As such, they certainly constitute enlighting examples to be understood before dealing with more complicated statistical physics problems.

