Logarithmic Sobolev inequality for zero-range dynamics (Q2368842): Difference between revisions

\textit{C. Landim, S. Sethuraman} and \textit{S. R. S. Varadhan} [Ann. Probab. 24, No. 4, 1871--1902 (1996; Zbl 0870.60095)] proved that if the rate function \(c:\mathbb N\to[ 0, +\infty[\), \(c(0)=0\), \(c(n)>0\) for \(n>0\), is Lipschitz and verifies a monotonicity requirement, then a Poincaré inequality holds for the zero-range process associated to \(c(\cdot)\). In this paper, under the same assumptions on the rate function, the following logarithmic Sobolev inequality is proved: for all \(f>0\), \[ \text{Ent}_{\nu_\Lambda^N}(f) \leq C L^2{\mathcal E}_{\nu_\Lambda^N}(\sqrt f,\sqrt f) \] where \(\nu_\Lambda^N\) is the invariant measure and \(\pm {\mathcal E}_{\nu_\Lambda^N}\) is the Dirichlet form associated to the zero-range process in the box \(\Lambda =[ 0, L]^d\cap \mathbb Z^d\). The constant \(C\) may depend on the dimension but not on \(L\) or on the number of particules, \(N\). The proof is inspired by the duplication method of \textit{N. Cancrini, F. Martinelli} and \textit{C. Roberto} [Ann. Inst. Henri Poincaré, Probab. Stat. (2002; Zbl 1174.82310)] and the two blocks-type estimates of \textit{S. L. Lu} and \textit{H.-T. Yau} [Commun. Math. Phys. 156, No. 2, 399--433 (1993; Zbl 0779.60078)].