Selberg’s Central Limit Theorem weighted by Linear Statistics of Zeta Zeros

Alessandro Fazzari, Maxim Gerspach, Paolo Minelli · 2025

Read Selberg’s Central Limit Theorem weighted by Linear Statistics of Zeta Zeros by Alessandro Fazzari, Maxim Gerspach, Paolo Minelli (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-2507-04150.txt.

Introduction

Fazzari, Gerspach, and Minelli extend Selberg's central limit theorem to zeta values weighted by zero statistics in twenty twenty-five. Selberg showed the log of the zeta function on its critical line acts like a bell curve. Here the values are weighted by how the zeros nearby are spaced. With the right scaling, a complex bell curve law still holds.

The test function's frequency range must stay small for the proof as given. For the imaginary part, the range stretches to its natural wall if the Riemann Hypothesis holds. A slow fading link to the zero density is also pinned down. The work merges two classic lines of zeta study.

Catalog record

CREATORAlessandro Fazzari, Maxim Gerspach, Paolo Minelli
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE47 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2507-04150

Browse the full library · About the Tower of Babel

Full text

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

Selberg’s Central Limit Theorem weighted by Linear Statistics of Zeta Zeros Alessandro Fazzari, Maxim Gerspach, Paolo Minelli Abstract . We consider the value distribution of the logarithm of the Riemann zeta function on the critical line, weighted by the local statistics of zeta zeros. We show that, with appropriate normalization, it satisfies a complex Central Limit Theorem, provided that the Fourier support of the test function in the linear statistics is sufficiently small. For the imaginary part, we extend this support condition up to its natural barrier under the Riemann Hypothesis. Finally, we prove that the correlation between log ⁡ ζ \log\zeta and the one-level density, while negligible on the level of Selberg’s Central Limit Theorem, only decays at a rather slow rate if the Riemann Hypothesis is assumed. Our results can be viewed as a combination of Selberg’s Central Limit Theorem with work of Hughes and Rudnick on mock-Gaussian behavior of the local statistics. 2020 Mathematics Subject Classification Primary 11M06; Secondary 11M26. 1. Introduction Selberg’s celebrated Central Limit Theorem states that the logarithm of the Riemann zeta function on the critical line, when suitably normalized, is approximately distributed like a complex Gaussian random variable. Specifically, if \tau is sampled uniformly on an interval [T,2T] , we have convergence in distribution (1.1) \frac{\log\zeta(\tfrac{1}{2}+i\tau)}{\sqrt{\log\log T}}\overset{T\to\infty}{\longrightarrow}\mathcal{N}_{\mathbb{C}}(0,1), where \mathcal{N}_{\mathbb{C}}(0,1) denotes a standard complex Gaussian. The works of Selberg [ 27 , 28 ] and Tsang [ 32 ] provide full proofs of the Central Limit Theorem for the real and imaginary parts of log-zeta individually. A sketch of the proof of ( 1.1 ) in the complex case can be found in [ 17 ] ; see also [ 20 , 5 ] for further discussion and generalizations. The classical approach to proving such a theorem is based on the method of moments. To establish ( 1.1 ), it suffices to show that for any fixed nonnegative integers h,\ell , we have \displaystyle\frac{1}{T}\int_{T}^{2T}\left(\log\zeta(\tfrac{1}{2}+it)\right)^{h}\left(\overline{\log\zeta(\tfrac{1}{2}+it)}\right)^{\ell}dt=\mathbbm{1}(h=\ell)\times h!(\log\log T)^{h}+O((\log\log T)^{(h+\ell-1)/2}). More recent proofs of Selberg’s Central Limit Theorem establish a convergence in distribution for the real part of \log\zeta without requiring the computation of all moments; see [ 5 , 19 ] and the more recent approach proposed by Radziwiłł and Soundararajan [ 24 ] . In modern terminology, a revealing way to understand why Selberg’s Central Limit Theorem holds is via the so-called hybrid model for zeta. In [ 13 ] , Gonek, Hughes, and Keating derive a formula that, under the Riemann Hypothesis, suggests an approximation of the type (1.2) \displaystyle\zeta(\tfrac{1}{2}+it)\approx\prod_{p\leq X}\bigg(1-\frac{1}{p^{1/2+it}}\bigg)^{-1}\times\prod_{|\gamma-t|\leq\frac{1}{\log X}}\bigg(1-\frac{\frac{1}{2}+it}{\frac{1}{2}+i\gamma}\bigg) where \frac{1}{2}+i\gamma runs over the nontrivial zeros of zeta. Thus, ( 1.2 ) expresses zeta as a product of a truncated Euler product and a truncated Hadamard product, with X serving as a parameter that controls the truncation ranges. Upon taking the logarithm, the Gaussian nature emerges: the logarithm of the Euler product in ( 1.2 ) essentially consists of a sum of random variables p^{-i\tau} , weighted by p^{-1/2} . Since p ranges over primes, the variables p^{-i\tau} behave like independent, uniformly distributed points on the unit circle, and thus their sum tends toward a complex Gaussian by the Central Limit Theorem. The bulk of Selberg’s work is then to show that the contribution from the zeros in ( 1.2 ) is negligible on average over t . In light of ( 1.2 ), primes alone do not capture the full nature of zeta; to reflect its complete structure, one also needs an understanding of its nontrivial zeros. The seminal works of Montgomery [ 22 ] and Rudnick–Sarnak [ 26 ] study the level correlations of zeros and reveal a striking connection with random matrix theory. The correlations of the zeros of zeta are conjectured to match those of the eigenvalues of large random unitary matrices. This phenomenon has been proven (partly under the Riemann Hypothesis) for a restricted class of test functions appearing in the definition of these correlations; see [ 22 , 25 ] . In [ 14 ] , Hughes and Rudnick investigate the linear statistics of the nontrivial zeros of zeta. More precisely, denoting the zeros by \frac{1}{2}+i\gamma with \gamma\in\mathbb{C} , they consider the centered moments of the quantity N_{\phi}(t):=\sum_{\gamma}\phi\left(\frac{\log T}{2\pi}(\gamma-t)\right), where \phi is a suitable test function with compactly supported Fourier transform. One can view N_{\phi} as a smooth version of the counting function of zeros in intervals of length comparable to the mean spacing. Hughes and Rudnick prove that the first few moments of N_{\phi} are asymptotic to those of a Gaussian random variable with mean \hat{\phi}(0) and variance (1.3) \displaystyle\sigma_{\phi}^{2}:=\int_{-\infty}^{+\infty}\min(|u|,1)\hat{\phi}(u)^{2}du. Specifically, provided that \hat{\phi} is supported in (-2/m,2/m) , they show that (1.4) \displaystyle\frac{1}{T}\int_{-\infty}^{+\infty}(N_{\phi}(t)-\hat{\phi}(0))^{k}\;\omega\left(\frac{t-T}{T}\right)dt=\mu_{k}\sigma_{\phi}^{k}+o(1), for every k\leq m . Here, \omega is a nonnegative weight function of total mass 1 whose Fourier transform is compactly supported, and (1.5) \displaystyle\mu_{k}:=\begin{cases}(k-1)!!&\text{if }k\text{ even}\\ 0&\text{if }k\text{ odd}\end{cases} is the k -th moment of a standard real Gaussian. Note that the smoothing function \omega in ( 1.4 ) is not strictly necessary; the same result holds with the sharp cutoff \omega=\mathbbm{1}_{[0,1]} , as implied by 1.1 and 4.3 . According to ( 1.4 ), the first m centered moments of N_{\phi} agree with those of a Gaussian. The analogous result has also been proven in the random matrix theory setting by Hughes and Rudnick [ 15 ] , where they further show that higher moments are not Gaussian. This phenomenon, which is also expected for the Riemann zeta function, is referred to as mock-Gaussian behavior. In this paper, we investigate the joint distribution of the logarithm of the Riemann zeta function and the linear statistics of its nontrivial zeros, thereby establishing a joint version of Selberg’s Central Limit Theorem and Hughes-Rudnick’s mock-Gaussian result. In other words, we study how the statistics of zeros near t are influenced by weighting by \log\zeta(\frac{1}{2}+it) . This captures the distribution of zeros around typical points, i.e. those governing the value distribution of log-zeta. To this end, we adopt a setup similar to that in [ 14 ] . Throughout the whole paper, we fix a function \phi that satisfies the following conditions: (1.6) \begin{split}&\phi\text{ is even, real-valued on the real line, and not identically zero;}\\ &\hat{\phi}\text{ is smooth and compactly supported, with }\eta=\eta_{\phi}:=\inf\{a>0:\operatorname{supp}\hat{\phi}\subseteq(-a,a)\}.\end{split} Note that, under these assumptions, \hat{\phi} is real-valued on the real line. Our main result provides an asymptotic formula for the joint moments of \log\zeta , its complex conjugate, and N_{\phi} . Theorem 1.1 . Let T\geq 3 be a (large) parameter and let h,\ell,k be fixed nonnegative integers with k even. For any function \phi that satisfies ( 1.6 ) with \operatorname{supp}\hat{\phi}\subseteq(-\frac{2}{k+2},\frac{2}{k+2}) , we have (1.7) \begin{split}\frac{1}{T}\int_{T}^{2T}&\left(\log\zeta(\tfrac{1}{2}+it)\right)^{h}\left(\overline{\log\zeta(\tfrac{1}{2}+it)}\right)^{\ell}\left(N_{\phi}(t)-\hat{\phi}(0)\right)^{k}dt\\ &=\mathbbm{1}(h=\ell)\times\mu_{k}\sigma_{\phi}^{k}h!\left(\log\log T\right)^{h}+O\left((\log\log T)^{(h+\ell-1)/2}\right),\end{split} where \sigma_{\phi}^{2} and \mu_{k} are defined as in ( 1.3 ) and ( 1.5 ), respectively. The support condition required in 1.1 is not quite optimal in the sense of Hughes and Rudnick [ 14 ] . While they showed that the k -th centered moment of N_{\phi} is Gaussian for \eta<2/k , our result holds only under the slightly more restrictive condition \eta<2/(k+2) . As further discussed in Section 3 , this limitation arises from repeated applications of Hölder’s inequality when approximating \log\zeta and N_{\phi} by suitable Dirichlet polynomials. Assuming the Riemann Hypothesis, we can improve the

Read the complete text