Limit laws for sums of independent random products: the lattice case (Q430980)

Let \(\{V_{i,j}; (i,j)\in \mathbb{N}^2\}\) be an array of i.i.d. random variables. Set \( Z_n=\sum_{i=1}^{N_n}\prod_{j=1}^n \exp\{V_{i,j}\}\) for integers \(n\) and \(N_n\). The study of \(Z_n\) is motivated by a number of models in statistical physics, see, e.g., \textit{A. Bovier} et al. [Ann. Probab., 30, No. 2, 605--651 (2002; Zbl 1018.60094)]. For non-lattice \(V_{1,1}\) with finite exponential moments the limit behavior of \((Z_n-A_n)/B_n\) with appropriate choice of normalizing sequences \((A_n)\) and \((B_n)\) (depending on the sequence \((N_n)\) growth) was treated by \textit{M. Cranston} and \textit{S. Molchanov} [Isr. J. Math. 148, 115--136 (2005; Zbl 1083.60013)]. In the present paper the lattice case is investigated. It is proved that \((Z_{n_k}-A_{n_k})/B_{n_k}\) weakly converge to a specified infinitely divisible distribution as \(n\to \infty\) for explicitly indicated \((A_n)\), \((B_n)\) and certain increasing integer sequence \((n_k)\). Thus although the suitably normalized sequence \(Z_n\) fails to converge in law, it is relatively compact in weak topology and its cluster set consists of semi-stable distributions.