Convergence of the largest singular value of a polynomial in independent Wigner matrices (Q373563)

For noncommutative polynomials in independent Wigner matrices, the author proves the convergence of the largest singular value to the operator norm of the corresponding polynomial in free semicircular variables. Let \(\mathbb C \langle X\rangle\) be the noncommutative polynomial algebra generated over \(\mathbb C\) by a sequence \(X_1,X_2,\dots\) of independent noncommuting algebraic variables. Let \(\text{Mat}_n(\mathbb C\langle X \rangle)\) be the algebra of \(n\times n\)-matrices with entries from \(\mathbb C\langle X \rangle \). This algebra is equipped with a natural involution by declaring each variable \(X_1,X_2,\dots\) to be self-adjoint. Let \(X_1^N,X_2^N,\dots\) be independent Wigner random \(N\times N\)-matrices. It is assumed that the distribution of the entries has a finite fourth moment. Also, let \(\Xi_1,\Xi_2,\dots\) be a sequence of free semicircular noncommutative random variables. The law of the noncommutative random variable \(f(\Xi_1,\Xi_2,\dots)\), for a self-adjoint \(f\in \text{Mat}_n(\mathbb C\langle X \rangle)\), is denoted by \(\mu_f\). It is known that for any self-adjoint \(f\in \text{Mat}_n(\mathbb C\langle X \rangle)\), the empirical distribution of the eigenvalues of \(f(X_1^N/\sqrt N,X_2^N/\sqrt N,\dots)\) converges weakly to \(\mu_f\), almost surely. The main result of the present paper states that for any self-adjoint \(f\in \text{Mat}_n(\mathbb C\langle X \rangle)\) and every \(\varepsilon>0\), the support of the spectrum of \(f(X_1^N/\sqrt N,X_2^N/\sqrt N,\dots)\) is contained in the \(\varepsilon\)-neighborhood of the support of \(\mu_f\) for sufficiently large \(N\), almost surely.