Logarithmic Sobolev constant for the dilute Ising lattice gas dynamics below the percolation threshold. (Q2574524)

In this paper the logarithmic Sobolev constant for a bond dilute Ising lattice gas with Kawasaki dynamics is studied. The bond dilute Ising lattice gas can be described as follows. At each site \(x\) of the \(d\)-dimensional lattice \(\mathbb Z^d\) an occupation number of particle \(\sigma (x)\in \{0,1\}\) is associated. The equilibrium states are described by the Gibbs measure characterized (informally) by the Hamiltonian \(H^{J,\lambda } = -\sum _{[x,y]} J_{xy}(2\sigma (x) - 1)(2\sigma (y) - 1) + \sum _x \sigma (x)\lambda (x)\) where \([x,y]\) denotes a generic bond of the lattice and the couplings \(\{J_{xy}\}\) are i.i.d.\ random variables taking only two values, \({J_{xy}}=0\) (bond closed) and \({J_{xy}}=1\) (bond open) with probability \(1-p\) and \(p\), respectively, and \(\beta >0\) can be interpreted as inverse temperature. Let \(\theta (p)\) be the probability that a cluster of the origin (i.e.\ the set of sites which are connected to the origin by a path of open bonds) is infinite. There exists a percolation threshold \(p_c\) such that \(\theta (p) = 0\) if \(p<p_c\) and \(\theta (p) > 0\) if \(p>p_c\). In Kawasaki dynamics the number of particles is conserved and the rate with which a particle jumps from a site \(x\) to its nearest neighbor \(y\) depends on the particle distribution around and on some external random field (disorder) in such a way that the whole process is reversible with respect to the canonical Gibbs measure described above. The main interest is to analyze the relaxation of the equilibrium of the dynamics in a finite box \(Q_L\) of side \(L\), centered at the origin, as a function of \(L\). Two key quantities give an estimate of the rate of convergence to equilibrium of the system: the inverse to the spectral gap of the generator of the dynamics and the logarithmic Sobolev constant (LSC), where the latter one is connected to the hypercontractivity of the Markov semi-group of the process and the decay of entropy. The main result of the paper is the following: Given \(p<p_c\) and \(0<\varepsilon \ll 1\), there exists a subset of disorder configurations of probability one such that, for any particle density and any \(\beta \), the LSC cannot grow faster than \(L^{2+\varepsilon }\).