Second-order Fusion Asymptotics for Sine\b{eta} Correlation Functions

Abstract

Recent work gives an all-$\beta$, all-order stochastic-zeta representation of the correlation functions of the $\Sine_β$ process and determines their leading Vandermonde asymptotics when several variables merge. We compute the first nontrivial correction throughout the regime $m\beta >1$. If $a_1,\dots ,a_m$ are distinct real numbers and \[ V(a)=\sum_{1\le i<j\le m}(a_i-a_j)^2, \] then, as $\epsilon \to 0$, \[ ρ^{(m)}_β(\varepsilon a_1,\ldots,\varepsilon a_m) =C^{(m)}_β|\varepsilon|^{β\binom m2}\prod_{i<j}|a_i-a_j|^β\left[1-\frac{β^2 V(a)}{8(mβ-1)(2m+1)}\varepsilon^2+o(\varepsilon^2)\right]. \] For $m=2$ the normalized second-order coefficient is $-{\beta }^2/[40(2\beta -1)]$ for every $\beta >1/2$. The proof combines a finite-$N$ rotational Ward identity, exact Hua--Pickrell trace moments, compact moment bounds for the stochastic-zeta entire function and its derivatives, and a quantitative multivariate expectation--Taylor lemma. As a by-product we evaluate \[ \E_{\HP_{β,mβ/2}}\sum_x x^{-2}=\frac{mβ}{4(mβ-1)(2m+1)}. \] The pole at $m\beta =1$ marks the boundary of the present second-moment argument and suggests a transition in the form of the next fusion correction.

12 pages. Submitted to Random Matrices: Theory and Applications