Limiting Distribution and Rate of Convergence for GL(3) Fourier Coefficients

Zongqi Yu · 2026

Read Limiting Distribution and Rate of Convergence for GL(3) Fourier Coefficients by Zongqi Yu (2026) 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-2605-20707.txt.

Introduction

Zongqi Yu proves a limiting distribution and convergence rate for GL(3) Fourier coefficients. Heath-Brown showed that the normalized error term in the Pilz divisor problem has a distribution function. Yu proves the analogue for a self-dual GL(3) Hecke-Maass cusp form, showing that x to the power of minus one third times the summed coefficients has a distribution function. The paper also gives a quantitative rate at which the finite distributions approach that limit.

Catalog record

CREATORZongqi Yu
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE60943
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.20707

Browse the full library · About the Tower of Babel

Full text

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

Limiting Distribution and Rate of Convergence for GL(3) Fourier Coefficients Zongqi Yu Abstract . In a work of Heath-Brown, it is proved that in the Pilz divisor problem, the normalized error term Δ 3 ​ ( x ) \Delta_{3}(x) has a distribution function. In this paper, we prove an analogue of this result in the setting of GL ⁡ ( 3 ) \operatorname{GL}(3) . For a given self-dual GL ⁡ ( 3 ) \operatorname{GL}(3) Hecke–Maass cusp form f f with normalized Fourier coefficients A f ​ ( n , m ) A_{f}(n,m) , let Δ f ​ ( x ) = ∑ n ⩽ x A f ​ ( n , 1 ) \Delta_{f}(x)=\sum_{n\leqslant x}A_{f}(n,1) . We show that the function x − 1 / 3 Δ f ( x ) x^{-1/3}\Delta_{f}(x) has a distribution function and we obtain a quantitative rate of convergence for the limiting distribution. † † Date. April, 2026 † † 2020 Mathematics Subject Classification. 11F30, 11N60, 11N56. † † Key words and phrases. Fourier coefficients, distribution function, rate of convergence, mean value estimate. 1. Introduction In analytic number theory, the classical divisor problem concerns the asymptotic behavior of the sum \Delta(x)=\sum_{n\leqslant x}d(n)-x\bigl(\log x+2\gamma-1\bigr), where d(n) is the divisor function and \gamma is Euler’s constant. Wintner [ 24 ] explicitly gives a value distribution of the normalized \Delta(x) . The development and application of probabilistic methods in analytic number theory have greatly enhanced our understanding of the distribution of such error terms. In 1992, Heath-Brown [ 11 ] proved that the normalized error term F(t)=t^{-1/4}\Delta(t) has a limiting distribution. For any interval I\subset\mathbb{R} , he showed that (1.1) \lim_{T\to\infty}\frac{1}{T}\operatorname{meas}\bigl\{t\in[T,2T]:F(t)\in I\bigr\}=\int_{I}g(u)du, where the limiting density g belongs to C^{\infty}(\mathbb{R}) and can be extended to an entire function on \mathbb{C} . Moreover, in that paper, Heath-Brown also proved that similar results hold for the error term \Delta_{3}(x)=\sum_{n\leq x}d_{3}(n)-\underset{s=1}{\operatorname*{\operatorname*{Res}}}\left(\frac{\zeta^{3}(s)x^{s}}{s}\right) , where d_{3}(n) is the triple divisor function, providing that x^{1/4}\Delta(x) is replaced by x^{1/3}\Delta_{3}(x) . The method of Heath-Brown has many applications. In 1993, Bleher et al. [ 4 ] proved that the error term for the number of lattice points inside a shifted circle has a limiting distribution. After that, Bleher also conducted extensive related work; see reference [ 6 , 5 ] . Using similar methods, Cai [ 7 ] proved analogous results on \operatorname{GL}(2) . A profound insight of [ 11 ] is that such a distribution can be obtained by approximating probability models. For a measurable function h on [T,2T] and an event \mathcal{B} , we define \mathbb{P}_{T}\big(h(t)\in\mathcal{B}\big):=\frac{1}{T}\operatorname{meas}\{t\in[T,2T]:h(t)\in\mathcal{B}\}. Let \{\mathbb{X}_{n}\}_{n\text{ square-free}} be a sequence of independent random variables, each uniformly distributed on [0,1] . Consider the random trigonometric series (1.2) F_{\mathbb{X}}:=\frac{1}{\pi\sqrt{2}}\sum_{\begin{subarray}{c}n=1\\ n\text{ square-free}\end{subarray}}^{\infty}\frac{\mu(n)^{2}}{n^{3/4}}\sum_{r=1}^{\infty}\frac{d(nr^{2})}{r^{3/2}}\cos\left(2\pi r\mathbb{X}_{n}-\frac{\pi}{4}\right). Heath-Brown’s result is that the density g in ( 1.1 ) is precisely the probability density function of the random variable F_{\mathbb{X}} . Furthermore, (1.3) \lim_{T\to\infty}\mathbb{P}_{T}\big(F(t)\leq u\big)=\mathbb{P}\big(F_{\mathbb{X}}\leq u\big)\qquad\text{for all }u\in\mathbb{R}. The motivation for the model ( 1.2 ) comes from the analytic structure of \Delta(t) itself. Voronoĭ’s summation formula yields that uniformly for t\in[T,2T] and any fixed \varepsilon>0 , F(t)=\frac{1}{\pi\sqrt{2}}\sum_{n\leq T}\frac{d(n)}{n^{3/4}}\cos\left(4\pi\sqrt{nt}-\frac{\pi}{4}\right)+O(T^{\varepsilon}). Writing n=mr^{2} with m is square-free, we have F(t)=\frac{1}{\pi\sqrt{2}}\sum_{m\leq T}\frac{\mu(m)^{2}}{m^{3/4}}\sum_{r\leq\sqrt{T/m}}\frac{d(mr^{2})}{r^{3/2}}\cos\left(4\pi r\sqrt{mt}-\frac{\pi}{4}\right)+O(T^{\varepsilon}). Lau and Tsang [ 19 ] provided explicit formulas for the moments of the probability density function of a class of functions, and proved that the density function of the remainder term of the divisor function is asymmetric. Lau [ 20 ] established a quantitative bound on the discrepancy between the distribution function of F(t) and that of the random model F_{\mathbb{X}} . More specifically, he proved that D_{F}(T):=\sup_{u\in\mathbb{R}}\left|\mathbb{P}_{T}(F(t)\leqslant u)-\mathbb{P}(F_{\mathbb{X}}\leqslant u)\right|\ll\frac{(\log\log\log T)^{3/4}}{(\log\log T)^{1/8}}. Based on ideas from the work [ 17 ] of Lamzouri, Lester and Radziwiłł, Lamzouri [ 18 ] improved the estimate for D_{F}(T) . He proved that D_{F}(T)\ll\frac{(\log\log\log T)^{9/4}}{(\log\log T)^{3/4}}. Furthermore, by a result of Lau [ 21 ] , Lamzouri also provided the distribution of large values of \Delta(t) in that paper. More specifically, he proved that there exist positive constants b_{1} , b_{2} , b_{3} , b_{4} such that for all real numbers V in the range b_{1}\leqslant V\leqslant b_{2}(\log\log T)^{1/4}(\log\log\log T)^{2^{4/3}-7/4} , one has \exp(-b_3V^4(\log V)^{-3(2^{4/3}-1)})\leqslant\mathbb{P}_{T}(F(t)>V)\leqslant\exp(-b_4V^4(\log V)^{-3(2^{4/3}-1)}). Moreover, the same bounds hold for \mathbb{P}_{T}(F(t)<V) , in the same range of V . In this paper, we are not confined to the classical divisor function, but rather a more general situation: arithmetic functions defined by the Fourier coefficients of automorphic forms on \operatorname{GL}(3) . Specifically, let f be a fixed self-dual Hecke–Maass cusp form for \operatorname{GL}(3) . Let A_{f}(n,m) be its normalized Fourier coefficients and let \Delta_{f}(x):=\sum_{n\leqslant x}A_{f}(n,1). We prove that the normalized function t^{-1/3}\Delta_{f}\left(t\right) has a distribution function g(\alpha) . More precisely, we have Theorem 1.1 . Let f be a self-dual Hecke–Maass cusp form for \operatorname{GL}(3) . The function F_{f}(t)=t^{-1/3}\Delta_{f}\left(t\right) has a distribution function g(\alpha) , in the sense that for any interval I, we have X^{-1}\operatorname{maes}\left\{t\in[1,X]:F_{f}(t)\in I\right\}\rightarrow\int_{I}g(\alpha)d\alpha, as X\rightarrow\infty . The function g(\alpha) and its derivatives satisfy the growth condition \frac{d^{k}}{d\alpha^{k}}g(\alpha)\ll_{A,k}(1+|\alpha|)^{-A}, for k=0,1,2,\ldots and any constant A. Moreover, g(\alpha) can be extended to an entire function on \mathbb{C} . We use the method discussed in Heath-Brown’s work [ 11 ] for estimating the divisor function to prove Theorem 1.1 . In the proof, we do not assume the validity of the Ramanujan conjecture, but rather replace the pointwise estimates in the original method with mean value estimates of Fourier coefficients. In order to prove Theorem 1.1 , we need an estimate for the mean square of the partial sums of the Fourier coefficients. Specifically, we have Theorem 1.2 . Let f be a self-dual Hecke–Maass cusp form for \operatorname{GL}(3) . For any positive \varepsilon , we have \int_{0}^{x}\Delta_{f}\left(y\right)^{2}dy=\frac{1}{10\pi^{2}}\sum_{n=1}^{\infty}\frac{A_{f}\left(1,n\right)^{2}}{n^{\frac{4}{3}}}x^{\frac{5}{3}}+O\left(x^{\frac{18}{11}+\varepsilon}\right). The key step in proving Theorem 1.2 is that we found a \sigma<2/3 such that \int_{1}^{T}\left|L(\sigma+it,f)\right|^{2}dt\ll T^{1+\varepsilon}. which is made possible thanks to the work [ 22 ] of Pal. Next, we use the method used by Lamzouri [ 17 ] in treating the divisor function. However, unlike the case of the divisor function, we use mean value estimates of the Fourier coefficients in place of the pointwise estimates. We establish a quantitative rate of convergence for the limiting distribution in Theorem 1.1 . Theorem 1.3 . Let \{\mathbb{X}_{n}\}_{n\text{cube-free}} be a sequence of independent random variables, each uniformly distributed on [0,1] . For a fixed self-dual Hecke–Maass cusp form f of \operatorname{GL}(3) , we have \sup_{u\in\mathbb{R}}\left|\mathbb{P}_{T}(F_{f}(t)\leqslant u)-\mathbb{P}(F_{f,\mathbb{X}}\leqslant u)\right|\ll\frac{\left(\log\log\log T\right)^{\frac{5}{3}}}{\left(\log\log T\right)^{\frac{1}{3}}}, where (1.4) F_{f,\mathbb{X}}:=\frac{1}{\pi\sqrt{3}}\sum_{\begin{subarray}{c}n=1\end{subarray}}^{\infty}\frac{\varepsilon(n)}{n^{2/3}}\sum_{r=1}^{\infty}\frac{A_{f}(1,nr^{3})}{r^{2}}\cos\left(6\pi r\mathbb{X}_{n}\right), with \varepsilon(n)=1 if n is cube-free and \varepsilon(n)=0 otherwise. Finally, following the approach of Lau [ 21 ] and Lamzouri [ 17 ] we determine the upper and lower bounds of the Laplace transform of F_{f,\mathbb{X}} and obtain the distribution of large values of \Delta_{f}(t) . Due to the lack of strong pointwise estimates for the Fourier coefficients and the inability to apply mean-value estimates, we obtain the following weaker result. Theorem 1.4 . For a fixed self-dual Hecke–Maass cusp form f of \operatorname{GL}(3) and any positive \varepsilon_{1} , \varepsilon_{2} . There exist positive constants b_{f,1} , b_{f,2} , b_{f,3} , b_{f.3} such that for all real numbers V in the range b_{f,1}\leqslant V\leqslant b_{f,2}(\log\log T)^{1/21}(\log\log\log T)^{-5/21} , we have \exp(-b_{f,3}V^{35/2+\varepsilon_1})\leqslant\mathbb{P}_{T}(F_{f}(t)>T)\leqslant\exp(-b_{f,4} V^{5/3-\varepsilon_2}). Moreover, the same bounds hold for \mathbb{P}_{T}(F_{f}(t)<V) , in the same range of V . Remark 1

Read the complete text