The March 2026 issue of the Annals of Mathematics contains a paper by Eichinger, Lukić, and Simanek (Eichinger, Lukić, and Simanek 2026) establishing a near-optimal sufficient condition that guarantees the uniform spacing of zeros of orthogonal polynomials. This result resolves a long-standing open problem in the field.
Let be a probability measure on with all moments finite, For any polynomials and with real coefficients, the inner product is therefore well defined and finite.
A sequence of orthogonal polynomials is a family such that for all and whenever . For a given measure , such a sequence is unique up to normalization, that is, multiplication of each by a nonzero constant . For definiteness, we fix the normalization by requiring that each polynomial has positive leading coefficient and satisfies .
With this normalization, , and for the polynomials can be constructed via the Gram–Schmidt process. First we form the unnormalized orthogonal polynomial and then normalize it:
It is well known that each polynomial has exactly real and simple zeros. Understanding the fine-scale distribution of these zeros along the real line is a central topic in the modern theory of orthogonal polynomials.
Fix a point and denote the zeros of by , indexed so that For each , the values are defined for exactly integers . We say that the measure exhibits clock behaviour at if, for every , the zero exists for all sufficiently large , and there is a sequence such that Informally, this means that the zeros of become asymptotically equally spaced near .
A useful tool for studying this phenomenon is the Christoffel–Darboux (CD) kernel, It was shown in (Levin and Lubinsky 2008) that holds at provided that the CD kernel satisfies the scaling limit The right-hand side is interpreted as when .
The limiting kernel in is independent of the measure . For this reason the phenomenon is known as bulk universality. It closely parallels the situation in random matrix theory, where the local statistics of eigenvalues converge to universal limits that depend only weakly on the distribution of matrix entries.
Over the past two decades many authors established under various assumptions on . In 2026, Eichinger, Lukić, and Simanek (Eichinger, Lukić, and Simanek 2026) proved the following theorem, which unifies and significantly extends many earlier results.
Thus Theorem establishes bulk universality , and hence clock behaviour , for a broad class of measures . Moreover, subsequent work (Eichinger, Lukić, and Woracek 2024) shows that the theorem is essentially optimal: the condition is not only sufficient but also almost necessary for bulk universality. In particular, implies while conversely the conclusion of Theorem , together with , implies .