Non-vanishing of central values of L-functions with angular restrictions
Filippo Berta and Svenja zur Verth · 2026
Read Non-vanishing of central values of L-functions with angular restrictions by Filippo Berta and Svenja zur Verth (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-2601-21642.txt.
Introduction
Filippo Berta and Svenja zur Verth prove that certain central values of L-functions do not vanish. L-functions encode deep arithmetic patterns, and their value at the center point carries special meaning. Showing the value is not zero, for a positive share of a family, is a classic hard goal.
The authors work with toroidal families under angular restrictions. They use the theory of trace functions plus a smoothing trick called mollification. Classification results pin down the sheaves in play, and a recent bound handles the key sums. Non-vanishing results feed conjectures linking L-functions to the arithmetic of curves and fields.
Catalog record
| CREATOR | Filippo Berta and Svenja zur Verth |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 101 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2601-21642
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Non-vanishing of central values of L-functions with angular restrictions Filippo Berta and Svenja zur Verth Abstract . We study the angular restrictions for the second moment of toroidal families of L L -functions using the general theory of trace functions. With the mollification technique we deduce non-vanishing of a positive proportion. Our two main ingredients are classification results of Katz to determine the sheaves at play and a recent result of Fouvry, Kowalski, Michel and Sawin to bound bilinear sums of their trace functions. 1. Introduction In this paper, we continue the study of the average of L -values at the critical point over toroidal families as introduced by Fouvry, Kowalski and Michel in [ FKM24 ] . In loc. cit. the authors prove that for any a,b\in\mathbb{Z}\smallsetminus\{0\} there exist constants \delta>0 , C(a,b)\neq 0 , so that for any prime number q the following asymptotic formula 1 1 1 See Theorem 7.1 for details. holds. (1.1) \frac{1}{q-1}\sum_{\chi\,(\mathrm{mod}\,{q})}L\left(\chi^{a},\tfrac{1}{2}\right)L\left(\chi^{b},\tfrac{1}{2}\right)=\delta_{a+b=0}\log q+C(a,b)+O(q^{-\delta}). Using Cauchy-Schwarz inequality twice, the evaluation of the above moment yields (see loc. cit. , Remark 1.2(3)) (1.2) |\{\chi\,(\mathrm{mod}\,{q})\ |\ L(\chi^{a},\tfrac{1}{2})L(\chi^{b},\tfrac{1}{2})\neq 0\}|\gg\frac{q}{(\log q)^{4}}, where q goes to infinity along the prime numbers. In the present paper, we improve ( 1.2 ) in two ways: We prove a positive proportion for the simultaneous non-vanishing of L(\chi^{a},\tfrac{1}{2}) and L(\chi^{b},\tfrac{1}{2}) and we consider subfamilies of characters \chi whose angle lies in a given section of the circle. More precisely for a primitive Dirichlet character \chi we define its angle \theta(\chi) as the unique number in (-\pi,\pi] so that \varepsilon(\chi)=e^{i\theta(\chi)} , where \varepsilon(\chi) is the classical normalized Gauss sum of \chi . For the trivial character the angle \theta(\chi) is not defined, we hence restrict our summation to primitive Dirichlet characters. Let I\subset(-\pi,\pi] be an open interval, then we study \frac{1}{q-1}\mathop{{\sum}^{\ast}}\limits_{\begin{subarray}{c}\chi\,(\mathrm{mod}\,{q})\\ \theta(\chi)\in I\end{subarray}}L\left(\chi^{a},\tfrac{1}{2}\right)L\left(\chi^{b},\tfrac{1}{2}\right) or rather the smooth version \frac{1}{q-1}\mathop{{\sum}^{\ast}}\limits_{\begin{subarray}{c}\chi\,(\mathrm{mod}\,{q})\end{subarray}}L\left(\chi^{a},\tfrac{1}{2}\right)L\left(\chi^{b},\tfrac{1}{2}\right)\phi(\theta(\chi)), where \phi\in C_{c}^{\infty}((-\pi,\pi]). In this paper, we show the following positive proportion of non-vanishing Theorem 1.1 . Let I\subset(-\pi,\pi] be a non-empty interval and let a,b be two non-zero integers. For any prime q we define E(a,b;q,I)=|{\{\chi\,(\mathrm{mod}\,{q})\ \text{ non-trivial s.t. }\ L(\chi^{a},\tfrac{1}{2})L(\chi^{b},\tfrac{1}{2})\neq 0,\ \theta(\chi)\in I\}}|. Then there exists a constant C(a,b,I)>0 so that (1.3) \frac{1}{q-1}E(a,b;q,I)\geqslant C(a,b,I) for q\to\infty . For a more precise statement see Theorem 8.1 . To prove Theorem 1.1 , we combine the mollification technique, results of Zacharias [ Zac19 ] on the mollified fourth moments, the evaluation ( 1.1 ) and our main technical result, which is the following. Theorem 1.2 . Let q be a sufficiently large prime. Let a,b non-zero integers such that for |k|=1 we have (a,b) is neither (1,1) nor (-1,-1) . Let 1\leqslant l_{1},l_{2}\leqslant q^{\alpha} for some \alpha>0 , small. There exists an absolute and effective constant \eta=\eta(a,b,\alpha)>0 and a constant c(l_{1},l_{2},b)\geqslant 0 (absolutely bounded in terms of q^{\alpha} ) so that for any integer k\neq 0 the following asymptotic formulae are true: \displaystyle\frac{1}{q-1} \displaystyle\mathop{{\sum}^{\ast}}\limits_{\chi\,(\mathrm{mod}\,{q})}\chi(l_{1}^{a}l_{2}^{b})L\left(\chi^{a},\tfrac{1}{2}\right)L\left(\chi^{b},\tfrac{1}{2}\right)\varepsilon(\chi)^{k} \displaystyle= \displaystyle\begin{cases}\frac{\zeta\left(\frac{|b|+1}{2}\right)}{l_{1}^{a/2}l_{2}^{b/2}}+O_{k}(q^{-\eta})&\text{if }(a,b,k)=(1,b>1,-1)\text{ or }(-1,b<-1,1)\\ c(l_{1},l_{2},b)+O_{k}(q^{-\eta})&\text{if }(a,b,k)=(1,b<0,-1)\text{ or }(-1,b>0,1)\\ O_{k}(q^{-\eta})&\text{else.}\end{cases} Remark 1.3 . This includes all cases except (a,b,k)=(1,1,-1) or (-1,-1,1) . By the functional equation one has \mathop{{\sum}^{\ast}}\limits_{\chi\,(\mathrm{mod}\,{q})}L(\bar{\chi},\tfrac{1}{2})L(\chi^{b},\tfrac{1}{2})\varepsilon(\chi)^{k}=\mathop{{\sum}^{\ast}}\limits_{\chi\,(\mathrm{mod}\,{q})}\chi(-1)i^{-\kappa(\chi)}L(\chi,\tfrac{1}{2})L(\chi^{b},\tfrac{1}{2})\varepsilon(\chi)^{k-1}, where \kappa(\chi)\in{0,1} . Therefore the cases with a=1,\ k=-1 or a=-1,\ k=1 degenerate to k=0 . We then apply an adaptation of ( 1.1 ) to obtain the first two cases of Theorem 1.2 . Thus the main novelty is the general case. Remark 1.4 . The constants C(a,b,I) in Theorem 1.1 and \eta in Theorem 1.2 are explicitly computable. On the other hand, the dependency on k in the term O_{k}(q^{-\eta}) in Theorem 1.2 is not explicit. This is a consequence of using the algebraic geometric arguments in [ Fou+25 ] . Their implicit constant depends upon a certain complexity attached to the geometric data which certainly depends on k . 2 2 2 See [ Saw+23 ] , Definitions 3.2,6,6.6, for the formal Definition of the complexity. See also [ Fou+19 ] , Definition 4.3 for a more gentle first definition. Currently, this dependence is not explicit, but it is expected to be polynomial. Remark 1.5 . The mollification technique has an interesting beneficial side effect. To prove Theorem 1.1 we approximate the interval I with a smooth Fourier series and apply Theorem 1.2 . Since the dependency in k of the error term O_{k}(q^{-\eta}) is not known, we need the mollifier to tame the contribution coming from the tail of the Fourier series. See Remark 8.2 for more details. Similar questions of non-vanishing of central L -values with angular restriction have already been considered in the literature. We describe two selected results qualitatively. Let q be a prime, sufficiently large and \theta an angle. Among other results, in [ Hou16 ] , Theorem 4 , Hough proved that there exists a non-principal \chi\,(\mathrm{mod}\,{q}) such that simultaneously the angle of \chi is close to \theta and the central value of L(\chi,\tfrac{1}{2}) is large. In [ Blo+23 ] (see in particular Theorem 1.8), among many other results, Blomer, Fouvry, Kowalski, Michel, Milićević, and Sawin consider the non-vanishing of L(f\otimes\chi,\tfrac{1}{2}) for f a fixed cusp form on \operatorname{GL}_{2} . Defining the angle \theta(f\otimes\chi) as the argument of the central L -value, they prove a non-zero proportion of non-vanishing for L(f\otimes\chi,\tfrac{1}{2}) , when \chi is ranging over Dirichlet characters so that \theta(f\otimes\chi) lies in a prescribed angular sector. We briefly discuss the strategy for the proof of Theorem 1.2 . After applying standard techniques, like the approximate functional equation, for the study of Dirichlet L -functions, we reduce the proof of Theorem 1.2 to obtaining estimates for bilinear correlation sums of the following shape: (1.4) \frac{1}{(qMN)^{1/2}}\sum_{m\sim M}\sum_{n\sim N}\alpha_{m}K(m^{a}n^{b})\ll_{a,b,k}q^{-\eta}, where K(u)\colon\mathbb{F}_{q}\to\mathbb{C} is depending on a , b and k and of the shape \Kl_{k}(u;q)\coloneqq\frac{1}{q^{(k-1)/2}}\sum_{\begin{subarray}{c}x_{1},\dots,x_{k}\,(\mathrm{mod}\,{q})\\ x_{1}\cdots x_{k}\equiv u\,(\mathrm{mod}\,{q})\end{subarray}}e\left(\frac{x_{1}+\dots+x_{k}}{q}\right) or \K_{k}^{a,b}(u;q)\coloneqq\frac{1}{q^{(k+1)/2}}\sum_{\begin{subarray}{c}x_{1},\dots,x_{k},y_{1},y_{2}\,(\mathrm{mod}\,{q})\\ x_{1}\cdots x_{k}y_{1}^{a}y_{2}^{b}\equiv u\,(\mathrm{mod}\,{q})\end{subarray}}e\left(\frac{x_{1}+\dots+x_{k}+y_{1}+y_{2}}{q}\right). As it is often the case in analytic number theory, this K(u) is the trace function of a complex of \ell -adic sheaves. Although hoping for cancellation as in ( 1.4 ) for every trace function is maybe too optimistic, in [ Fou+25 ] , Fouvry, Kowalski, Michel and Sawin prove cancellation for a broad family of such trace functions, that is, trace functions whose attached complex of \ell -adic sheaves is gallant . Their result plays a key role in our work. The property of being gallant is depending on the monodromy group of the sheaf attached to the trace function. To identify these our second main ingredient is the work of Katz, [ Kat90 ] who developed the theory of how to identify the monodromy groups. Our paper is structured in the following way. • In Section 2 we begin the treatment of \frac{1}{q-1}\mathop{{\sum}^{\ast}}\limits_{\chi\,(\mathrm{mod}\,{q})}\chi(l_{1}^{a}l_{2}^{b})L\left(\chi^{a},\tfrac{1}{2}\right)L\left(\chi^{b},\tfrac{1}{2}\right)\varepsilon(\chi)^{k} by applying the approximate functional equation. • In Section 3 we relate all of our functions K_{k}^{a,b} to hypergeometric sums and in particular to \ell -adic sheaves that are well studied. • In Section 4 we apply a smooth partition of unity and the Pólya-Vinogradov method to reduce the range of summation to dyadic intervals (M,2M] and (N,2N] such that q^{1/2-\delta}\leqslant M,N\leqslant q^{1/2+\delta} . We furthermore reduce the general case in Theorem 1.2 to estimating type I sums. • In Section 5 we prove the estimates on the type I sum required in Section 4 in the case where the trace function is the additive character e_{q}