Asymptotic properties of random matrices and pseudomatrices (Q643442)

This paper treats the asymptotic properties of sums of random pseudomatrices (matricially free random variables), and the author tries to compare them to those of random matrices with block-identical variances. First the author obtains the limit joint distributions of blocks for objects of both types. In fact, Theorem. (cf. Theorem 3.1, p. 2415) Under the assumptions: (A1) for an \(n\)-dimensional square array of self-adjoint variables \(( X_{i,j}(n))\) in a unital \(\ast\)-algebra \({\mathcal A}(n)\) equipped with an array of states \(( \phi_{i,j}(n))\) defined by a family of states \(( \phi_j(n))\), \(1 \leqslant j \leqslant n\), \(( X_{i,j}(n))\) is matricially free with respect to \(( \phi_{i,j}(n))\) for any \(n \in {\mathbb N}\); (A2) the variables have zero expectations \(\phi_{i,j}(n)( X_{i,j}(n))=0\) for all \(i,j \in [n]\) and \(n \in {\mathbb N}\); (A3) their variances are block-identical and are of order \(1/n\), namely, \(\phi_{i,j}(n)\) \((X_{i,j}^2(n))\) \(=\) \(u_{p,q}/n\) for any \(i \in N_p\), \(j \in N_q\), where each \(u_{p,q}\) is a non-negative real number; (A4) their moments are uniformly bounded, i.e., \(\forall m\), \(\exists M_m \geq 0\) such that \[ | \phi_{i,j}(n)( X_{i,j}^m (n) ) | \leqslant \frac{M_m}{n^{m/2} } \tag{1} \] for all \(i,j \in [n]\) and \(n \in {\mathbb N}\); the limit distributions of random pseudomatrices under partial traces have the form \[ \lim_{n \to \infty} \psi_k(n) ( S^m(n)) = \sum_{ \pi \in {\mathcal N} C_m^2 } b_k(\pi), \tag{2} \] where \(k \in [r]\) and \(m \in {\mathbb N}\). Consequently, \(\psi(n)( S^m(n))\) converges to \[ \sum_{\pi \in {\mathcal N}{\mathcal C}_m^2} b(\pi) \qquad \text{as} \quad n\to \infty, \tag{3} \] where \(b(\pi) = \sum_{k=1}^r d_k b_k(\pi)\). Next the author derives Hilbert space realizations of the limit joint distributions of random pseudomatrices \(S(n)\) and their blocks under \(\phi(n)\), \(\psi_k(n)\) and \(\psi(n)\). Interestingly enough, all these realizations are given on the same type of Hilbert spaces which is a matricially free product of Fock spaces. In particular, if the variance matrices are symmetric, the asymptotics of symmetric blocks of random pseudomatrices agree with those of symmetric random blocks. It is also shown that blocks of random pseudomatrices are asymptotically matricially free, and that the corresponding symmetric random blocks are asymptotically symmetrically matricially free. Finally, the author shows that row blocks of square, block-lower-triangular and block-diagonal pseudomatrices are asymptotically free, monotone independent and boolean independent, respectively. For other related works, see, e.g., [\textit{R. Lenczewski}, J.\ Funct.\ Anal.\ 258, No. 12, 4075--4121 (2010; Zbl 1194.46089)].