Mixing time of a kinetically constrained spin model on trees: power law scaling at criticality (Q2257122)

The autors study the interacting particle spin system on the state space \(\{0,1\}^{{\mathcal T}_k },\) where \({\mathcal T}_k\) is a regular infinite rooted tree with \(k\) offsprings (= children) of each one of the vertices of the tree \({\mathcal T}_k.\) The dynamics of the model, which are formulated below, mimic those for a class of models in physics, named kinetically constrained spin systems. State transitions of the considered model are governed by the following rule. Each vertex waits, independently of other vertices, an exponential mean \(1\) time and then the value of the spin at the vertex takes the value \(0\) or \(1\) with probabilities \(p\) and \(1-p\), respectively, provided that the spins of all children of the considered vertex are zero. Thus, the product Bernoulli(\(p\)) measure \(\mu\) appears to be the reversible measure of the stochastic process. The aforementioned condition for a state transition is called restriction. It is known that \(p_c= 1/k\) is a critical value of the probability \(p,\) in the sense that if \(p<p_c,\) the process is ergodic, while if \(p>p_c,\) the process is not ergodic. The present paper focuses on the critical case \(p=p_c,\) studying the properties of mixing time and relaxation time in the critical regime. The obtained estimates of these quantities answer a question of \textit{D. Aldous} and \textit{P. Diaconis} posed in [J. Stat. Phys. 107, No. 5--6, 945--975 (2002; Zbl 1006.60095)].