Sharp bounds for maximal sums of odd order Dirichlet characters
Alexander P. Mangerel · 2025
Read Sharp bounds for maximal sums of odd order Dirichlet characters by Alexander P. Mangerel (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-2505-07651.txt.
Introduction
Alexander Mangerel sharpens bounds on sums of Dirichlet characters in a two thousand twenty-five number theory paper. These characters are functions that repeat in a fixed cycle. They sit at the heart of work on prime numbers. The Pólya-Vinogradov inequality sets a cap on how large their part sums can grow.
Mangerel fixes an odd order of at least three. He studies primitive characters of that order. If the generalized Riemann hypothesis holds, he can lift the known upper bounds. A fine factor with iterated logarithms shows up in the gain. He also proves, with no such extra claim, that infinitely many such characters come close to his bound. That match from below shows the bound is sharp, up to a constant factor. The work builds on past work by Granville, Soundararajan, Goldmakher, and Lamzouri.
Catalog record
| CREATOR | Alexander P. Mangerel |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 44 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2505-07651
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Sharp bounds for maximal sums of odd order Dirichlet characters Alexander P. Mangerel Abstract . Let g ≥ 3 g\geq 3 be fixed and odd, and for large q q let χ \chi be a primitive Dirichlet character modulo q q of order g g . Conditionally on GRH we improve the existing upper bounds in the Pólya-Vinogradov inequality for χ \chi , showing that M ( χ ) := max t ≥ 1 | ∑ n ≤ t χ ( n ) | ≪ q ( log log q ) 1 − δ g ( log log log log log q ) δ g ( log log log q ) 1 / 4 , M(\chi):=\max_{t\geq 1}\left|\sum_{n\leq t}\chi(n)\right|\ll\sqrt{q}\frac{(\log\log q)^{1-\delta_{g}}(\log\log\log\log\log q)^{\delta_{g}}}{(\log\log\log q)^{1/4}}, where δ g := 1 − g π sin ( π / g ) \delta_{g}:=1-\tfrac{g}{\pi}\sin(\pi/g) . Furthermore, we show unconditionally that there is an infinite sequence of order g g primitive characters χ j \chi_{j} modulo q j q_{j} for which M ( χ j ) ≫ q j ( log log q j ) 1 − δ g ( log log log log log q j ) δ g ( log log log q j ) 1 / 4 , M(\chi_{j})\gg\sqrt{q_{j}}\frac{(\log\log q_{j})^{1-\delta_{g}}(\log\log\log\log\log q_{j})^{\delta_{g}}}{(\log\log\log q_{j})^{1/4}}, so that our GRH bound is sharp up to the implicit constant. This improves on previous work of Granville and Soundararajan, of Goldmakher, and of Lamzouri and the author. 1. Introduction Let q\geq 3 . Given a non-principal Dirichlet character \chi modulo q we define M(\chi):=\max_{t\geq 1}\left|\sum_{n\leq t}\chi(n)\right|. It is a classical problem to obtain sharp bounds for M(\chi) . The Pólya-Vinogradov inequality, proved over 100 years ago, yields the unconditional bound M(\chi)\ll\sqrt{q}\log q . Montgomery and Vaughan [ 10 ] showed in 1977 that this can be improved, assuming the Generalised Riemann Hypothesis (GRH), to M(\chi)\ll\sqrt{q}\log\log q. The latter bound turns out to be the best possible for general non-principal Dirichlet characters, as Paley [ 11 ] constructed an infinite sequence of quadratic characters \chi_{j} modulo q_{j} such that M(\chi_{j})\gg\sqrt{q_{j}}\log\log q_{j}. Besides the implicit constant therein, no substantial progress has been made towards improving the unconditional bound of Pólya and Vinogradov for general non-principal characters. Almost 20 years ago, Granville and Soundararajan [ 5 ] made a breakthrough when they showed that if \chi is a primitive character of fixed odd order then the unconditional and GRH bounds in the Pólya-Vinogradov inequality may both be improved. More precisely, let g\geq 3 be a fixed odd integer, and let \chi be a primitive Dirichlet character modulo q of order g . In the sequel, write Q:=\begin{cases}\log q&\text{ assuming GRH,}\\ q&\text{ unconditionally.}\end{cases} In [ 5 ] it was shown, using methods from pretentious number theory (that they themselves developed), that as q\rightarrow\infty , M(\chi)\ll\sqrt{q}(\log Q)^{1-\delta_{g}/2+o(1)}, where \delta_{g}:=1-\tfrac{g}{\pi}\sin(\pi/g)>0 . This was subsequently improved by Goldmakher [ 3 ] , who refined their method to replace the exponent 1-\delta_{g}/2 with 1-\delta_{g} . He also constructed, on GRH, an infinite sequence \chi_{j} of primitive characters modulo q_{j} of order g such that for every \varepsilon>0 , M(\chi_{j})\gg_{\varepsilon}\sqrt{q_{j}}(\log\log q_{j})^{1-\delta_{g}-\varepsilon}\text{ as }j\rightarrow\infty. Subsequently, Lamzouri and the author [ 8 ] developed new methods to improve upon Goldmakher’s result, both conditionally and unconditionally. They showed in particular that on GRH there is a constant C>0 , depending at most on g , such that if \chi is a primitive character modulo q of order g then (1) M(\chi)\ll\sqrt{q}(\log\log q)^{1-\delta_{g}}(\log\log\log q)^{-1/4}(\log\log\log\log q)^{C}. Conversely, they showed unconditionally that there is an infinite sequence of primitive characters \chi_{j} modulo q_{j} of order g such that (2) M(\chi_{j})\gg\sqrt{q_{j}}(\log\log q_{j})^{1-\delta_{g}}(\log\log\log q_{j})^{-1/4}(\log\log\log\log q_{j})^{-C}, with the same constant C as above. Evidently, despite the improvements to previous work obtained in [ 8 ] , the sharp rate of growth of M(\chi) , for \chi modulo q of fixed odd order, as a function of q remained unknown. This is in contrast to the situation for even order characters, which appears to be rather better understood. Indeed, assuming GRH Granville and Soundararajan [ 5 ] generalised Paley’s examples to produce, for every even order k an infinite family of primitive characters \chi_{j} modulo q_{j} of order k such that M(\chi_{j})\gg\sqrt{q_{j}}\log\log q_{j}, a result that was then rendered unconditional by Goldmakher and Lamzouri [ 4 ] (see [ 1 ] for an alternative approach, as well as [ 7 ] for a more precise result). In this note, we sharpen ( 1 ) conditionally on GRH, and obtain an unconditional construction of characters \chi_{j} of order g that improve upon ( 2 ). We thereby obtain, for the first time, the sharp rate of growth of M(\chi) as a function of q , for Dirichlet characters \chi of fixed odd order g , conditionally on 1 1 1 Unfortunately, as our method does not circumvent the issues arising from Siegel zeros, we are unable to improve upon the unconditional Theorem 1.1 of [ 8 ] . GRH. Theorem 1.1 . Assume GRH. Let g\geq 3 be a fixed odd integer. Then for any primitive character \chi modulo q of order g we have M(\chi)\ll\sqrt{q}\frac{(\log\log q)^{1-\delta_{g}}(\log\log\log\log\log q)^{\delta_{g}}}{(\log\log\log q)^{1/4}}. Theorem 1.2 . Let g\geq 3 be a fixed odd integer. There are arbitrarily large q and primitive Dirichlet characters \chi modulo q of order g such that M(\chi)\gg\sqrt{q}\frac{(\log\log q)^{1-\delta_{g}}(\log\log\log\log\log q)^{\delta_{g}}}{(\log\log\log q)^{1/4}}. 1.1. Proof Ideas In order to surpass the results in [ 8 ] we refine the methods of the latter paper to gain further insight into what is required of the extremal characters in the problem. With this insight we are able to quickly obtain Theorem 1.1 , and also improve upon the construction of the characters \chi that yield ( 2 ), which leads to Theorem 1.2 . To describe more precisely our key new ideas, we shall recall some of the details of the work on odd order characters in [ 5 ] , and the work that succeeded it. Let \chi be a primitive character of large conductor q and fixed odd order g . Granville and Soundararajan, and later Goldmakher, showed that the estimation of M(\chi) may be (essentially) reduced to obtaining lower bounds for the prime sum \sum_{p\leq y}\frac{1-\text{Re}(\chi(p)\overline{\psi}(p)p^{-it})}{p}, for each (\log q)^{A}\leq y\leq q , each t\in\mathbb{R} satisfying |t|\leq\log y , and every odd 2 2 2 Recall that a Dirichlet character \xi is called odd if \xi(-1)=-1 , and even otherwise. We will use without mention the facts that if \xi is odd then its order is even, and thus if \xi has odd order then \xi must be even. primitive character \psi of conductor m\leq\log y . For the purposes of this explanation we shall assume that t=0 . Letting \mu_{g} denote the set of complex roots of unity of order g , a lower bound for this quantity may naturally be obtained by instead seeking an upper bound for (3) \mathcal{S}(y;\psi,g):=\sum_{p\leq y}\frac{1}{p}\max_{\begin{subarray}{c}z\in\mu_{g}\cup\{0\}\end{subarray}}\text{Re}(z\overline{\psi}(p)). Given \psi modulo m of order k , say, Granville and Soundararajan observed that since the conductor of \psi is at most \log y we may appeal to the Siegel-Walfisz