Phase transitions in probabilistic cellular automata

We investigate the low-noise regime of a large class of probabilistic cellular automata, including the North-East-Center model of Toom. They are defined as stochastic perturbations of cellular automata with a binary state space and a monotonic transition function and possessing a property of erosion. These models were studied by Toom, who gave both a criterion for erosion and a proof of the stability of homogeneous space-time configurations. Basing ourselves on these major findings, we prove, for a set of initial conditions, exponential convergence of the induced processes toward the extremal invariant measure with a highly predominant state. We also show that this invariant measure presents exponential decay of correlations in space and in time and is therefore strongly mixing. This result is due to joint work with Augustin de Maere. For the two-dimensional probabilistic cellular automata in the same class and for the same extremal invariant measure, we give an upper bound to the probability of a block of cells with the opposite state. The upper bound decreases exponentially fast as the diameter of the block increases. This upper bound complements, for dimension 2, a lower bound of the same form obtained for any dimension greater than 1 by Fernández and Toom. In order to prove these results, we use graphical objects that were introduced by Toom and we give a review of their construction.