Bulk universality holds pointwise in the mean for compactly supported measures (Q715714): Difference between revisions

This is a case of a very informative title, reflecting the main result of this paper. Given a finite positive Borel measure \(\mu\) on \(\mathbb R\), the reproducing kernel is \[ K_n(x,y)=\sum_{j=0}^{n-1} p_j(x) p_j(y), \] where \(p_n\) is the \(n\)-th orthonormal polynomial with respect to \(\mu\). Universality (in the bulk of the spectrum) means that, for any \(\xi\) in the interior of the support of \(\mu\), \[ \lim_n \frac{K_n\left(\xi + \frac{a}{K_n(\xi, \xi)}, \xi + \frac{b}{K_n(\xi, \xi)}\right)}{K_n(\xi, \xi)} = \frac{\sin \pi (a-b)}{\pi (a-b)} \] uniformly for \(a\), \(b\) in compact subsets of the real line. ``Universality'' resides in the independence from both the point \(\xi\) and the orthogonality measure \(\mu\). This property is crucial in the study of the asymptotic statistical quantities for the spectrum of unitary ensembles of random matrices. It is known that the universality property holds under very mild conditions on the measure \(\mu\). This is a hot topic of research, and in the recent years there has been a number of very relevant contributions (many of them, due to the author of this paper) clarifying the subject. For instance, now it is known that the so-called regularity of the measure (a global condition) plus a local Szegő-type condition (integrability of \(\log(\mu')\)) are sufficient for universality a.e. Since the limit above is ``local'', it is natural to ask ourselves what happens if we impose conditions on \(\mu\) only in a neighborhood of \(\xi\). Although we do not know if the pointwise universality still holds, this paper is an important step in this direction: it is shown that if \(\mu'\) is bounded away from \(0\) on an open interval \(J\), then for each \(r>0\), \[ \lim_m \frac{1}{m}\sum_{n=1}^m \sup_{|a|,|b|\leq r} \left| \frac{K_n\left(\xi + \frac{a}{K_n(\xi, \xi)}, \xi + \frac{b}{K_n(\xi, \xi)}\right)}{K_n(\xi, \xi)} - \frac{\sin \pi (a-b)}{\pi (a-b)} \right| = 0 \] a.e.\ on \(J\) (more precisely, at the Lebesgue points of \(\mu\) on \(J\)). The method of proof continues exploiting the fruitful connection with classes of entire functions, found by the author of the paper.