Asymptotic zero distribution of random orthogonal polynomials (Q2280555)

Let \(\xi_0,\xi_1,\dots\) be a sequence of i.i.d. non-degenerate complex random variables. Consider a measure \(\tau\) supported on a regular compact subset \(K\) of the complex plane and satisfying the Bernstein-Markov condition. The authors study the asymptotic distribution of complex roots of the random polynomials \[ H_n(z):=\sum_{j=0}^n\xi_jq_j(z) \] as \(n\to\infty\), where \(q_0,q_1,\dots\) are deterministic polynomials orthonormal with respect to \(\tau\). They show that if the condition \[ \mathbb P[|\xi_0|>e^n]=o(n^{-1}),\quad n\to\infty, \] holds, then the normalized counting measure of the zeros of \(H_n\) converges weakly in probability to the equilibrium measure of \(K\), and in the case when \(H_n\) is the Kac polynomial (that is, \(q_j(z)=z^j\)) this condition is also necessary for this convergence. From this the authors give a positive answer on the open question asking whether there exists a Kac polynomial with the normalized counting measure of zeros converging weakly in probability and not almost surely. Also the authors consider the random polynomials \[ G_n(z):=\sum_{j=0}^n\xi_j f_{n,j}z^j(z), \] where the complex constants \(f_{n,j}\) satisfy some not very restrictive conditions, and show that if \[ \mathbb E\log(1+|\xi_0)<\infty, \] then the normalized counting measure of the zeros of \(G_n\) converges almost surely to some deterministic measure which is characterized in terms of the limiting shape of \(\{f_{n,j}\}\).