On the Relevance of the Maximum Entropy Principle in Non–Equilibrium Statistical Mechanics

At first glance, the maximum entropy principle (MEP) apparently allows us to derive, or justify in a simple way, funda- mental results of equilibrium statistical mechanics. Because of this, a school of thought considers the MEP as a powerful and elegant way to make predictions in physics and other disciplines, rather than a use- ful technical tool like others in statistical physics. From this point of view the MEP appears as an alternative and more general predictive method than the traditional ones of statistical physics. Actually, careful inspection shows that such a success is due to a series of fortunate facts that characterize the physics of equilibrium systems, but which are ab- sent in situations not described by Hamiltonian dynamics, or generically in nonequilibrium phenomena. Here we discuss several important examples in non equilibrium statistical mechanics, in which the MEP leads to incorrect predictions, proving that it does not have a predictive nature. We conclude that, in these paradigmatic examples, an approach that uses a detailed analysis of the relevant aspects of the dynamics cannot be avoided.