Spectral representation of correlation functions for zeros of Gaussian power series with stationary coefficients

Tomoyuki SHIRAI · 2025

Read Spectral representation of correlation functions for zeros of Gaussian power series with stationary coefficients by Tomoyuki SHIRAI (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-03704.txt.

Introduction

Shirai studies zeros of random power series in twenty twenty-five. The series have Gaussian coefficients that follow a stationary process. Such functions are called Gaussian analytic functions. Their zeros form random patterns whose structure is described by correlation functions. Shirai uses spectral measures of the coefficient processes to analyse these functions. He revisits limit theorems and works through examples with numeric computations. The main result is an integral formula for the n point correlation functions of the zero sets, written in terms of those spectral measures.

Catalog record

CREATORTomoyuki SHIRAI
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE41 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-03704

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Spectral representation of correlation functions for zeros of Gaussian power series with stationary coefficients Tomoyuki SHIRAI Abstract . We analyze Gaussian analytic functions (GAFs) defined as power series with coefficients modeled by discrete stationary Gaussian processes, utilizing their spectral measures. We revisit some limit theorems for random analytic functions and examine some examples of GAFs through numerical computations. Furthermore, we provide an integral representation of the n n -point correlation functions of the zero sets of GAFs in terms of the spectral measures of the underlying coefficient Gaussian processes. 2020 Mathematics Subject Classification 30B20; 30C15; 60G15; 60G55 1. Introduction Gaussian analytic functions (GAFs) have attracted significant interest and have been extensively studied by many researchers (cf. [ 17 , 32 , 33 , 15 ] and references therein). One of the striking results in the study of GAFs is that the zeros of random power series \sum_{n=0}^{\infty}\zeta_{n}z^{n} with i.i.d. complex Gaussian coefficients \{\zeta_{n}\}_{n=0}^{\infty} , whose covariance kernel is the Szegő kernel, form the determinantal point process (DPP) on the unit disk associated with the Bergman kernel (1-z\bar{w})^{-2} [ 27 ] (see for details on DPPs (cf. [ 30 , 35 , 31 , 15 ] ). As a natural generalization of this model, various related models have been studied. The singular points of random matrix-valued Gaussian analytic functions are shown to form the DPPs associated with weighted Bergman kernel (1-z\bar{w})^{-\alpha} [ 21 ] . The zeros of random power series \sum_{n=0}^{\infty}\zeta^{\mathbb{R}}_{n}z^{n} with i.i.d. real Gaussian coefficients \{\zeta^{\mathbb{R}}_{n}\}_{n=0}^{\infty} , which is the limiting Kac random polynomial, form the Pfaffian point process on the unit disk [ 11 , 22 ] . The Gaussian Laurant series whose covariance kernel is the weighted Szegő kernel on an annulus was studied in [ 18 ] . We are concerned with the zeros of Gaussian power series \sum_{n=0}^{\infty}\xi_{n}z^{n} with coefficients being stationary, centered, complex Gaussian process \{\xi_{n}\}_{n=0}^{\infty} . In [ 24 , 29 ] , we discussed the expected number of zeros within a disk and the density of zeros in the radial direction, observing their asymptotic behavior as the radius approaches the circle of convergence. Our observations reveal that the spectral measure of the coefficient Gaussian process plays a crucial role in determining the asymptotic behavior of the number of zeros near the radius of convergence. From the viewpoint of the spectral measure, the zeros of Gaussian analytic functions in a strip in the complex plane, with translation-invariant distribution, are studied in [ 9 , 10 ] . This line of study traces back to the work of Paley and Wiener [ 25 ] . In [ 2 , 3 , 1 ] , the authors investigate random power series and their zeros from the perspective of spectral measures (or limiting empirical spectral measures in the case of deterministic coefficients), considering cases where the coefficients are not necessarily Gaussian processes but weakly stationary processes, and even in certain cases where the coefficients are deterministic. In this paper, we also emphasize the spectral perspective to study the zeros of GAFs. The remainder of this paper is organized as follows. In Section 2 , we review the Edelman-Kostlan formula for the 1 -point correlation function and give its spectral representation. We also recall some results of limit theorems for random analytic funcitons established in [ 28 ] . In Section 3 , we explore examples of GAFs associated with three different spectral measures and present some numerical computations. In Section 4 , we provide a spectral representation of the n -point correlation function for the zeros of a GAF whose covariance kernel is a stationary complex Gaussian process. As an example of this represetation result, we compute certain quantities related to the i.i.d. coefficients case to illustrate the Peres-Virág theorem. 2. Gaussian analytic functions and their spectral representation 2.1. Spectral representation of GAF and the intensity of its zeros Let \Xi=\{\xi_{k}\}_{k\in\mathbb{Z}} be a stationary, centered, complex Gaussian process with covariance \gamma(k-l)=\mathbb{E}[\xi_{k}\overline{\xi_{l}}]. Throughout the paper, we assume that \gamma(0)=1 , i.e., the distribution of each marginal \xi_{k} is complex standard normal. We consider the Gaussian power series with dependent coefficients \Xi=\{\xi_{k}\}_{k\in\mathbb{Z}} defined by X_{\Xi}(z):=\sum_{k=0}^{\infty}\xi_{k}z^{k} and its zeros. Since the radius of convergence of X_{\Xi}(z) is equal to 1 a.s. under \gamma(0)=1 , the zeros of X_{\Xi}(z) form a point process \mathcal{Z}_{X_{\Xi}}:=\sum_{z\in\mathbb{D}:X_{\Xi}(z)=0}\delta_{z} on \mathbb{D}:=\{z\in\mathbb{C}:|z|<1\} . We denote the covariance kernel of the GAF \mathbb{X}=\{X_{\Xi}(z)\}_{z\in\mathbb{D}} by K(z,w) . Then, we have the spectral representation of K(z,w) [ 29 ] as K(z,w)=\mathbb{E}[X_{\Xi}(z)\overline{X_{\Xi}(w)}]=\int_{\mathbb{T}}s(z,t)\overline{s(w,t)}F(dt), (2.1) where \mathbb{T}=(-\pi,\pi] , the kernel s(z,t) is given by the Szegő kernel s(z,t)=\frac{1}{1-ze^{-it}}\quad(z\in\mathbb{D},t\in\mathbb{T}) (2.2) and F(dt) is the spectral measure of the coefficient Gaussian process \Xi [ 7 ] , i.e., F(dt) is a probability measure on \mathbb{T} (since \gamma(0)=1 ) such that \gamma(k)=\int_{\mathbb{T}}e^{-ikt}F(dt). We denote the right-hand side of ( 2.1 ) by K_{F}(z,w) . Since |s(z,t)|^{2}=(1-2r\cos(\theta-t)+r^{2})^{-1} for z=re^{i\theta} , the covariance kernel K_{F}(z,z) for z=re^{i\theta} can be expressed as K_{F}(z,z)=\frac{1}{1-r^{2}}\int_{\mathbb{T}}P(r,\theta-t)F(dt), (2.3) where P(r,\theta) is the Poisson kernel given by P(r,\theta)=\frac{1-r^{2}}{1-2r\cos\theta+r^{2}}. (2.4) The first intensity (or 1 -point correlation function) of the zeros of a GAF can be computed from the covariance kernel K(z,w) by the Edelman-Kostlan formula [ 8 ] as follows: Theorem 2.1 (Edelman-Kostlan) . Let X=(X(z))_{z\in D} be a GAF with convariance K(z,w)=\mathbb{E}[X(z)\overline{X(w)}] . Then, the first intensity of zeros of X(z) is given by the formula \rho_{1}(z)=\frac{1}{\pi}\partial_{z}\partial_{\bar{w}}\log K(z,w)\Big|_{w=z}. When K(z,w) is of the form ( 2.1 ), using the fact that \partial_{z}s(z,t)=e^{-it}s(z,t)^{2} and the symmetrization of the variable t , we can see that \rho_{1}(z)=\frac{1}{2\pi K(z,z)^{2}}\int_{\mathbb{T}^{2}}|e^{-it}-e^{-iu}|^{2}|s(z,t)s(z,u)|^{4}F(dt)F(du)\quad(z\in\mathbb{D}). (2.5) In Section 4 , we will provide an integral representation of the n -point correlation function of the zeros. 2.2. Convergence of random analytic functions and their zero processes Let D\subset\mathbb{C} be a domain and \mathcal{H}(D) be the space of complex analytic functions on D equipped with the metric \rho(f,g)=\sum_{j=1}^{\infty}\frac{1}{2^{j}}\frac{\|f-g\|_{K_{j}}}{1+\|f-g\|_{K_{j}}}, where \|f\|_{K} is the supremum norm of f on a compact set K and K_{1}\subset K_{2}\subset\dots\nearrow D is an exhaustion of D by compact sets K_{j} . The metric space (\mathcal{H}(D),\rho) turns out to be complete and separable. We denote the totality of probability measures on \mathcal{H}(D) by \mathcal{P}(\mathcal{H}(D)) . We call an \mathcal{H}(D) -valued random variable a random analytic function . We recall the following fact established in [ 28 , Proposition 2.5 with Remark on page 341] ). Proposition 2.2 . Let D\subset\mathbb{C} be a domain and \{X_{n}\}_{n=1}^{\infty} be a sequence of random analytic functions on D . If \sup_{n}\mathbb{E}[|X_{n}(z)|^{p}] is locally integrable on D for some p>0 , then \{\mu_{X_{n}}\}_{n=1}^{\infty} is tight in \mathcal{P}(\mathcal{H}(D)) . Moreover, if \{X_{n}\}_{n=1}^{\infty} converges to X in the sense of finite-dimensional distributions, then \{\mu_{X_{n}}\} converges weakly to \mu_{X} as n\to\infty , or equivalently, \{X_{n}\}_{n=1}^{\infty} converges to X in law. We also recall the convergence of zero processes [ 28 , Proposition 2.3] . Proposition 2.3 . A sequence of random analytic functions \{X_{n}\}_{n=1}^{\infty} converges in law to X . Then, the zero process \mathcal{Z}_{X_{n}} converges in law to \mathcal{Z}_{X} provided that X\not\equiv 0 almost surely. We apply these results to our current situation. Let X_{F}=\{X_{F}(z)\}_{z\in\mathbb{D}} be the GAF corresponding to the covariance kernel K_{F}(z,w)=\int_{\mathbb{T}}s(z,t)\overline{s(w,t)}F(dt). (2.6) Corollary 2.4 . Let F_{n}(dt),n=1,2,\dots and F(dt) be probability measures on \mathbb{T} and suppose that \{F_{n}(dt)\}_{n=1}^{\infty} converges weakly to F(dt) . Then,

Read the complete text