Dirichlet L L -functions on the critical line and multiplicative chaos
Sami Vihko · 2025
Read Dirichlet L L -functions on the critical line and multiplicative chaos by Sami Vihko (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-2506-16115.txt.
Introduction
Prime-number-like functions in the L family are wildly random on their most important line, and Sami Vihko makes that randomness exact in this 2025 paper. Feed the Dirichlet L-functions a randomly chosen character, and as the modulus grows they settle, he proves, onto a single random limiting object, a wavy distribution tied to Gaussian multiplicative chaos.
The same limit object appeared when others shifted the classic zeta function at random heights, tying two kinds of randomness together. Right of the critical line the convergence is tamer and older, occurring among smooth analytic functions; Vihko's route gives a new proof even there. The conditions are mild: the modulus's smallest prime factor must merely wander off to infinity. The result firms up the statistical picture of L-functions as a chaotic system.
Catalog record
| CREATOR | Sami Vihko |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 56 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-16115
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Dirichlet L L -functions on the critical line and multiplicative chaos Sami Vihko Abstract In this paper we prove that the Dirichlet L L -functions L ( 1 / 2 + i x , χ q ) L(1/2+ix,\chi_{q}) , where χ q \chi_{q} is uniformly random Dirichlet character modulo q q , converges to a random Schwartz distribution ζ rand \zeta_{\mathrm{rand}} , which is related to (complex) Gaussian multiplicative chaos. This is the same limiting object that appeared in [ 34 ] , where the authors proved that the random shifts of the Riemann zeta function on the critical line ζ ( 1 / 2 + i x + i ω T ) \zeta(1/2+ix+i\omega T) , where ω ∼ Unif ( [ 0 , 1 ] ) \omega\sim\mathrm{Unif}([0,1]) and x ∈ ℝ x\in\mathbb{R} , converge as T → ∞ T\to\infty . 1 Introduction The central objects of investigation in this paper are the Dirichlet L -functions L(s,\chi) , where s is a complex variable 1 1 1 We shall use the common convention in analytic number theory that s instead of z denotes a complex variable. and \chi\colon\mathbb{N}\to\mathbb{C} is a function called a Dirichlet character, see Section 1.1 below for proper definitions. Dirichlet characters \chi depend on a positive integer q called the modulus of the Dirichlet characters. For fixed q there are multiple Dirichlet characters and their number is given by the Euler totient function \varphi(q) . In this paper we study L(s,\chi_{q}) , where \chi_{q} is chosen uniformly randomly from the set of Dirichlet characters modulo q . More precisely, we study L(s,\chi_{q}) on the critical line \mathrm{Re}\,(s)=1/2 as q\to\infty . We prove that in certain sense the limit exists and the limiting object is the random generalized function \zeta_{\mathrm{rand}} called the randomized zeta function. It appeared also as a limiting object in [ 34 ] in the study of statistical behaviour of the Riemann zeta function \zeta(s) on the critical line and asymptotically far from the real axis, that is, as the T\to\infty limit of the random shifts \zeta(s+i\omega T) with \omega\sim\mathrm{Unif}([0,1]) . \zeta_{\mathrm{rand}} is related to a very special random generalized function called (complex) Gaussian multiplicative chaos (GMC), see Section 1.2 for more information. GMC has been ubiquitous in many models of mathematical physics and probability theory in roughly the past 15 years, see for example [ 4 , 12 , 10 , 2 , 3 ] . Dirichlet L -functions are in some sense generalizations of the famous Riemann zeta function defined by \zeta(s):=\sum_{n=1}^{\infty}n^{-s} (1.1) for \mathrm{Re}\,(s)>1 and analytically continued to \mathbb{C}\setminus\{1\} . They were originally used to study primes in arithmetical progressions a+nd , where a,d\in\mathbb{N} are relatively prime and n is a positive integer. These functions can also be defined as a series in the same half plane as \zeta , that is, L(s,\chi):=\sum_{n=1}^{\infty}\frac{\chi(n)}{n^{s}}. (1.2) They share many properties with the Riemann zeta function. In particular, the L -functions also have an analytic continuation to \mathbb{C}\setminus\{1\} . Series such as 1.1 and 1.2 are in general called Dirichlet series. The study of the behaviour of such objects on the critical line \mathrm{Re}\,(s)=1/2 is notoriously difficult. This is no surprise since the distribution of the zeros of the Riemann zeta function on the critical strip \mathrm{Re}\,(s)\in(0,1) is related to many open questions about the prime numbers. The Riemann-hypothesis (RH) and its generalization to Dirichlet L -functions (GRH) states that all the non-trivial zeros of the associated functions are located on the critical line. Both the Riemann \zeta -function and Dirichlet L -functions admit an Euler product representation from which the connection to prime numbers is evident \displaystyle\zeta(s) \displaystyle=\prod_{p}(1-p^{-s})^{-1} (1.3) \displaystyle L(s,\chi) \displaystyle=\prod_{p}(1-\chi(p)p^{-s})^{-1}, (1.4) valid in the same region \mathrm{Re}\,(s)>1 as the series representations. Above the product ranges over all primes p . This is easy to verify by using geometric series and the uniqueness of prime factorization. Probabilistic methods have been long used in analytic number theory to study the generic behaviour of the Riemann zeta function. Often in this context the term used is statistical behaviour. The first such result on the critical line is the classical central limit theorem of Selberg from the 1940’s [ 35 ] . An elegant new proof for Selberg’s central limit theorem was provided recently in [ 30 ] . Assuming the GRH Selberg himself generalized this for a certain class of Dirichlet series that also contain a class of Dirichlet L -functions called the primitive L -functions [ 36 ] . A new proof of this result for the primitive L -functions was also provided recently in [ 16 ] in the spirit of [ 30 ] . These results tell us roughly that at a random point with \mathrm{Im}\,(s)\in[T,2T] and large T>0 along the critical line \mathrm{Re}\,(s)=1/2 the logarithm of the \zeta -function and certain L -functions behave like a mean zero Gaussian with a variance that diverges as T\to\infty . The divergence is of order \log\log(T) . Therefore, to properly formulate a central limit theorem the logarithm of the zeta function needs to be multiplied with a vanishing (as T\to\infty ) normalization constant proportional to (\log\log(T))^{-1/2} . From the perspective of this paper, there is another relevant question. Namely, does something similar hold at any point at the critical line if we take a random Dirichlet character \chi modulo q , consider the logarithm of L(\cdot,\chi) at that point and let q tend to infinity? In other words, does the logarithm of L(1/2+ix,\chi) behave like a Gaussian with a variance that diverges as we take q\to\infty for any x\in\mathbb{R} ? This is a hard open question since it is difficult to control the zeros of Dirichlet L -functions as q and the characters modulo q vary. There has been some evidence to this direction, see for example [ 20 , 22 ] . In these papers, the authors conjecture such behaviour based on connections with random matrix theory and numerical computations. See also [ 8 ] , where the authors prove a version of this type of central limit theorem with respect to a weighted measure and assuming the GRH. All of these only consider special cases of x\in\mathbb{R} called central values of the L -function and the class of primitive L -functions. The connection between random matrix theory and the study of Riemann \zeta - and Dirichlet L -functions dates back Montgomery and Dyson in the 1970’s who conjectured that statistically the zeros of Riemann \zeta -function behave as the eigenvalues of a large Hermitian random matrix. See the book review [ 9 ] by Conrey for early developments of this direction of investigation. Since the turn of the millennia the connection with random matrices has provided lots of new conjectures. We do not try to list them here, but only mention the ones that are important from our perspective. Firstly, Keating and Snaith realized that statistically the \zeta -function on the critical line could be modelled by the characteristic polynomial of a random unitary matrix [ 21 ] . Based on this they made predictions about the value distribution and moments of the \zeta -function. As already mentioned the same authors made also conjectures about