Asymptotic Feynman-Kac formulae for large symmetrised systems of random walks (Q731703)

The authors derive Large Deviation Principles (LDP) under a symmetrized measure of \(N\) random processes for the empirical path measure, for the mean path and for the mean occupation local times, both at finite time horizons. The LD rate functions are given explicitly in terms of variational problems involving an entropy term, due to permutations used in the symmetrization, and a certain Legendre transform depending on the case considered. They thereafter draw a number of conclusions about variants of the principles, laws of large numbers and asymptotic dependence. These LD results may be obtained for random processes, Markovian or not, on any connected graphs with finite or denumerable vertices. One of their motivations comes from the Feynman-Kac formulae used to express thermodynamic functions in quantum statistical mechanics, and they indeed use theie results to get a non-commutative version of Varadhan's lemma with Bose-Einstein statistics. They also apply their results to a system of \(N\) random walks conditioned to stay within a finite set with the counting measure as initial distribution, for which they get the well-known Donsker-Varadhan rate function for the LDP of the mean occupation local times. This rate functions governs also the LDP for the occupation local time of a single random walk, but when time goes to infinity, while their results concern finite time horizons. They interpret this striking result in terms of Bose-Einstein condensation, as a first step towards a rigorous understanding of large bosons systems at positive temperature.