Fourier optimization and pair correlation problems
Mithun Kumar Das, Tolibjon Ismoilov, Antonio Pedro Ramos · 2025
Read Fourier optimization and pair correlation problems by Mithun Kumar Das, Tolibjon Ismoilov, Antonio Pedro Ramos (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-2502-05106.txt.
Introduction
How evenly the zeros of certain number-theory functions spread out is captured by a statistic called the pair correlation, first studied for the Riemann zeta function. Mithun Kumar Das, Tolibjon Ismoilov, and Antonio Pedro Ramos build a general framework for such statistics in a 2025 paper, covering many sequences at once under mild assumptions.
Their route passes through two extreme-value problems in Fourier analysis, which they solve along the way. The payoff is average bounds for the form factor of several sequences of interest, including zeros of Dedekind zeta functions and of functions in the Selberg class. For the real and imaginary parts of the zeta function, their bounds show that a 2018 conjecture by Gonek and Ki cannot hold in some cases.
Catalog record
| CREATOR | Mithun Kumar Das, Tolibjon Ismoilov, Antonio Pedro Ramos |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 90 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2502-05106
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Fourier optimization and pair correlation problems Mithun Kumar Das, Tolibjon Ismoilov, Antonio Pedro Ramos Abstract . We introduce a generic framework to provide bounds related to the pair correlation of sequences belonging to a wide class. We consider analogues of Montgomery’s form factor for zeros of the Riemann zeta function in the case of arbitrary sequences satisfying some basic assumptions, and connect their estimation to two extremal problems in Fourier analysis, which are promptly studied. As applications, we provide average bounds of form factors related to some sequences of number theoretic interest, such as the zeros of primitive elements of the Selberg class, Dedekind zeta functions, and the real and imaginary parts of the Riemann zeta function. In the last case, our results bear an implication to a conjecture of Gonek and Ki (2018), showing it cannot hold in some situations. 2010 Mathematics Subject Classification 41A30, 46E22, 11M06, 11M26 1. Introduction The celebrated work of Montgomery [ 18 ] initiated the study of how non-trivial zeros \rho=\frac{1}{2}+i\gamma of the Riemann zeta function \zeta(s) are distributed along the critical line by studying their pair correlation. His approach consisted of introducing the function F(\alpha) , a Fourier transform of the distribution of the gaps between zeros \gamma-\gamma^{\prime} . In [ 18 ] , Montgomery proposed a conjecture on the limiting distribution of the pair correlation of zeros of zeta, which came to be known as the pair correlation conjecture (PCC). Goldston [ 12 ] subsequently proved an equivalence between PCC and certain asymptotics for average values of Montgomery’s function F(\alpha) , motivating the study of average estimates of F(\alpha) , which was done, for example, in [ 3 , 5 , 11 , 13 ] . The approach of [ 18 ] was extended to many other settings. There are several instances in the literature where analogues of Montgomery’s function were used to study the pair correlation of zeros of L -functions, Dirichlet series, and similar objects. One can see [ 20 , 9 , 2 , 17 , 15 , 6 ] for a few examples. Naturally, there are corresponding conjectures for their asymptotic behavior. In this work, we will look at the problem of average estimates from a much broader perspective, introducing an axiomatic approach which allows us to bound the long-term averages of analogues of Montgomery’s function in a wide variety of settings. We do this by connecting these average bounds to two Fourier optimization problems, which constitute another major object of study of this paper. The investigation of these problems uses techniques from Fourier analysis and the theory of reproducing kernel Hilbert spaces. As a consequence of the bounds we obtain, we address a conjecture of Gonek and Ki [ 15 ] related to the pair correlation of zeros of \operatorname{Re}\zeta , \operatorname{Im}\zeta , and other similar functions along vertical lines in the left-half of the critical strip. We show that the relevant form factor cannot have the conjectured expression in certain regimes. 1.0.1. Notation Throughout this text, we use the following definition of the Fourier transform for a given function f\in L^{1}(\mathbb{R}) : \mathscr{F}[f](\xi)=\widehat{f}(\xi):=\int_{\mathbb{R}}e^{-2\pi i\xi x}f(x)\,\text{\rm d}x. We also adopt this normalization for the Fourier transform of a finite Borel measure \mu , setting \widehat{\mu}(\xi):=\int_{\mathbb{R}}e^{-2\pi i\xi x}\,\text{\rm d}\mu(x). For a real number x , we denote x_{+}=\max\{x,0\} . Finally, the quantity s_{0}:=\min_{x\in\mathbb{R}}\frac{\sin{x}}{x}=-0.217233\ldots will be useful to state many of our results. 1.1. Montgomery’s pair correlation conjecture Assuming the Riemann hypothesis (RH), let \rho=\frac{1}{2}+i\gamma denote a non-trivial zero of \zeta(s) . Montgomery conjectured that, assuming RH, the pair correlation of zeros of zeta has the following limiting distribution: for any fixed \beta>0 , (1.1) \bigg(\frac{T}{2\pi}\log\,{T}\bigg)^{-1}\sum_{\begin{subarray}{c}0<\gamma,\gamma^{\prime}\leq T\\ 0<\gamma-\gamma^{\prime}\leq\frac{2\pi\beta}{\log\,{T}}\end{subarray}}1\sim\int_{0}^{\beta}\bigg(1-\bigg(\frac{\sin\,{\pi u}}{\pi u}\bigg)^{2}\bigg)\text{\rm d}u,\quad\mbox{as }T\rightarrow\infty. This is now known as Montgomery’s pair correlation conjecture (PCC). In order to study the quantity on the left-hand side of ( 1.1 ), Montgomery [ 18 ] introduced the form factor (1.2) F(\alpha)=F(\alpha,T):=\bigg(\frac{T}{2\pi}\log\,{T}\bigg)^{-1}\sum_{0<\gamma,\gamma^{\prime}\leq T}T^{i\alpha(\gamma-\gamma^{\prime})}w(\gamma\!-\!\gamma^{\prime}), where \alpha\in\mathbb{R} , T\geq 2 , w(u)=\frac{4}{4+u^{2}} , and the sums are taken over ordinates of non-trivial zeros of \zeta(s) , counting multiplicity. For any function r\in L^{1}(\mathbb{R}) such that \widehat{r}\in L^{1}(\mathbb{R}) , the Fourier inversion formula gives that \bigg(\frac{T}{2\pi}\log\,{T}\bigg)^{-1}\sum_{0<\gamma,\gamma^{\prime}\leq T}r\bigg((\gamma\!-\!\gamma^{\prime})\frac{\log\,{T}}{2\pi}\bigg)w(\gamma\!-\!\gamma^{\prime})=\int_{\mathbb{R}}\widehat{r}(\alpha)F(\alpha,T)\,\text{\rm d}\alpha. One can see the connection between ( 1.1 ) and ( 1.2 ) from this identity. Indeed, considering an appropriate sequence of functions r that approximate the characteristic function of the interval (0,\beta] in the above identity, we can obtain the asymptotic of the left-hand side of ( 1.1 ). Therefore, if we know the asymptotic behavior of the function F(\alpha,T) as T\to\infty , we know the asymptotic behavior of the pair correlation of zeros. Montgomery [ 18 ] , then Goldston and Montgomery [ 14 , Lemma 8] , obtained partial progress on understanding F(\alpha,T) conditionally on RH. They proved that (1.3) F(\alpha,T)=\left(T^{-2|\alpha|}\log{T}+|\alpha|\right)\left(1+O\bigg(\sqrt{\frac{\log\,\log{T}}{\log{T}}}\bigg)\right),\quad\text{ as }\;\;T\to\infty, uniformly for 0\leq|\alpha|\leq 1 , improving the original work of Montgomery. Not only this, in [ 18 ] Montgomery also conjectured that for |\alpha|\geq 1 and uniformly for any bounded interval, (1.4) F(\alpha,T)\sim 1,\quad\mbox{ as }T\rightarrow\infty, which, together with ( 1.3 ), provides a complete description of the asymptotic behavior of F(\alpha,T) . This statement is what became known as the strong pair correlation conjecture, which, by the reasoning above, implies ( 1.1 ). In 1988, Goldston [ 12 ] proved the equivalence of PCC and the following asymptotic for the averages of F(\alpha,T) : (1.5) \frac{1}{\ell}\int_{b}^{b+\ell}F(\alpha,T)\text{\rm d}\alpha\sim 1,\mbox{ as }T\rightarrow\infty, for any b\geq 1 and \ell>0 . Recently, Carneiro, Milinovich, and Ramos [ 5 ] proved, assuming RH, that \displaystyle 0.9303+o(1)<\frac{1}{\ell}\int_{b}^{b+\ell}F(\alpha,T)\text{\rm d}\alpha<1.3208+o(1),\quad\mbox{ as }T\rightarrow\infty, uniformly on b\geq 1 for the upper bound and for \ell\geq\ell_{0}(b) for the lower bound, improving the previous results in [ 3 ] . They obtained this result by introducing new averaging mechanisms and making full use of the class of test functions considered in [ 7 ] for the sphere packing problem. We now proceed to the axiomatic formulation which allows the generalization of the work of [ 5 ] to address multiple different objects. 1.2. Axiomatic formulation Our main purpose is to obtain a unified approach for various pair correlation functions, capturing the behavior of a wide variety of examples. Our objects of interest will be sequences \Gamma(T)=(\gamma_{n}(T)) of real numbers depending on a parameter T\geq T_{0} . Note that the elements of the sequence are not necessarily distinct. For most of the number-theoretic instances we have in mind, such as when \Gamma consists of the ordinates of non-trivial zeros of a typical L -function, the sequence is independent of T . We introduce the dependence on the parameter because we are also interested in cases such as the ones Gonek and Ki studied in [ 15 ] , where, among other things, they considered the pair correlation of zeros of the real part of \zeta(s) along the lines \operatorname{Re}(s)=\frac{1}{2}-\frac{c}{\log T} for a fixed c>0 , which naturally vary with T . We discuss this and a few other examples in the dedicated Section 1.6 below. Now, sequences of non-trivial zeros of a typical L -function share the behavior that the number of elements with height at most T>0 is asymptotically \frac{\lambda T}{2\pi}\log T and the average spacing of elements up to that height is \frac{\lambda}{2\pi}\log T