Limit laws for products of free and independent random variables (Q4514728)

The free independence as a noncommutative analog of the classical notion of independence in probability theory is considered. In the paper the authors study the parallelism between freenes and independence for products of positive random variables. The results of the paper are written in terms of convolutions. Let \({\mathcal M}_+\) be the Borel probability measures on \([0,\infty]\). Free multiplicative convolution \(\mu\boxtimes\nu\) of \(\mu,\nu\in {\mathcal M}_+\) is the probability distribution of \(X^{{1\over 2}} YX^{{1\over 2}}\), where \(X\) and \(Y\) are free random variables with distributions \(\mu\) and \(\nu\), respectively. Preliminaries concerning the calculation of multiplication convolutions are given. The main result of the paper is Theorem. Let \(\{\mu_n, n\geq 1\}\) be a sequence of measures in \({\mathcal M}_+\), and let \(\{k_n\in{\mathcal N}, n\geq 1\}\), \(\lim_{n\to\infty} k_n=\infty\). The following assertions are equivalent: (i) The sequence \(\mu_n\boxtimes\cdots\boxtimes \mu_n\) (\(k_n\) times) converges weakly to a measure \(\nu\in {\mathcal M}_+\), (ii) The measure \(d\sigma_n(t)= k_n{(t- 1)^2\over t^2+ 1} d\mu_n({1\over t})\) converge weakly in \([0,\infty]\) to a measure \(\sigma\), and the limit \(\gamma= \lim_{n\to\infty} k_n \int_{(0,\infty)} {t^2-1\over t^2+1} d\mu_n({1\over t})\) exists. In this case \(\nu= \nu^{\gamma,\sigma}_{\boxtimes}\). Comparison with the commutative case is given.