Peierls bounds from random Toom contours

TL;DR

This paper extends Toom's stability results for deterministic cellular automata to those with intrinsic randomness using a novel random contours method, providing new stability proofs and highlighting method limitations.

Contribution

Develops a new random contours approach to estimate Peierls sums, enabling stability proofs for monotone cellular automata with intrinsic randomness on arbitrary groups.

Findings

Introduces a method for bounding Peierls sums in stochastic cellular automata.

Proves stability results for certain monotone automata with randomness.

Constructs an example where the Peierls sum diverges despite stability expectations.

Abstract

For deterministic monotone cellular automata on the $d$-dimensional integer lattice, Toom (1980) has given necessary and sufficient conditions for the all-one fixed point to be stable against small random perturbations. We are interested in the open problem of extending Toom's result to monotone cellular automata with intrinsic randomness, where the unperturbed evolution is random with i.i.d. update rules attached to the space-time points. For some applications it is also desirable to consider a more general graph structure, so we assume that the underlying lattice is an arbitrary countable group. Toom's proof of stability is based on a Peierls argument. In previous work, we demonstrated that this Peierls argument can also be used to prove stability for cellular automata with intrinsic randomness, but in this case estimating the Peierls sum becomes much harder than in the deterministic case. In the present paper, we develop a method based on random contours to estimate the Peierls sum and apply it to prove new stability results for monotone cellular automata with intrisic randomness. We also demonstrate the limitations of the method by constructing an example where the Peierls sum is infinite for arbitrary small perturbations even though stability is believed to hold.