Bounds for spectral projectors on the three-dimensional torus

Pierre Germain 1, Simon L. Rydin Myerson 2, Daniel Pezzi 3 · 2025

Read Bounds for spectral projectors on the three-dimensional torus by Pierre Germain 1, Simon L. Rydin Myerson 2, Daniel Pezzi 3 (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-2508-05573.txt.

Introduction

Pierre Germain, Simon Rydin Myerson, and Daniel Pezzi bound the size of spectral projectors on the three-dimensional torus, in a 2025 math paper. A spectral projector gathers the wave patterns of a surface whose frequencies fall inside a chosen window. How large such a gathered wave can get, measured in stronger norms, is a classic question with roots in work by Sogge and Bourgain.

For a window of fixed size the answer is classical. The hard case is a shrinking window, where the question turns into one about single eigenfunctions. Working on the three-dimensional torus, the authors prove new cases of a conjecture the first two had posed earlier. They also sharpen older results in every dimension by removing small slack factors. Their tools come from number theory: the geometry of numbers, the circle method, and bounds on exponential sums.

Catalog record

CREATORPierre Germain 1, Simon L. Rydin Myerson 2, Daniel Pezzi 3
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE135 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-05573

Browse the full library · About the Tower of Babel

Full text

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

Bounds for spectral projectors on the three-dimensional torus Pierre Germain 1, Simon L. Rydin Myerson 2, Daniel Pezzi 3 Abstract . We study L 2 L^{2} to L p L^{p} operator norms of spectral projectors for the Euclidean Laplacian on the torus in the case where the spectral window is narrow. With a window of constant size this is a classical result of Sogge; in the small-window limit we are left with L p L^{p} norms of eigenfunctions of the Laplacian, as considered for instance by Bourgain. For the three-dimensional torus we prove new cases of a previous conjecture of the first two authors concerning the size of these norms; we also refine certain prior results to remove ϵ \epsilon -losses in all dimensions. We use methods from number theory: the geometry of numbers, the circle method and exponential sum bounds due to Guo. We complement these techniques with height splitting and a bilinear argument to prove sharp results. We exposit on the various techniques used and their limitations. 1. Introduction Our starting point is the Laplacian \Delta=\frac{\partial^{2}}{\partial x_{1}^{2}}+\frac{\partial^{2}}{\partial x_{2}^{2}}+\frac{\partial^{2}}{\partial x_{3}^{2}} on the torus \mathbb{T}^{3}=\mathbb{R}^{3}/\mathbb{Z}^{3} . For a positive definite quadratic form Q(x)=x^{T}Ax,\qquad\mbox{with $A$ symmetric}, this example can be generalized to operators of the form \Delta_{Q}=Q(\partial_{1},\partial_{2},\partial_{3}) , still defined on \mathbb{T}^{3}=\mathbb{R}^{3}/\mathbb{Z}^{3} (equivalently, we could have kept the Laplacian fixed and considered general tori of the form \mathbb{R}^{3}/\Lambda , where \Lambda is a lattice of \mathbb{R}^{3} of full rank). We refer to the case A=\operatorname{Id} , Q({x})=\|{x}\|^{2} , or \Lambda=\mathbb{Z}^{3} , which are all equivalent, as the square torus . 1.1. The conjectures in dimension 3 We consider \Delta_{Q}=Q(\partial_{1},\partial_{2},\partial_{3}) on \mathbb{T}^{3}=\mathbb{R}^{3}/\mathbb{Z}^{3} , and define by functional calculus the spectral projectors P_{\lambda,\delta}=\mathbf{1}_{(-1,1)}\left(\frac{\sqrt{-\Delta_{Q}}-2\pi\lambda}{{2\pi}\delta}\right), where \delta\in(0,1) is small and \lambda>1 is large (the factors {2\pi} are added in the above definition to make subsequent notations simpler). In other words, P_{\lambda,\delta}=\Phi_{\lambda,\delta}*\cdot\quad\mbox{with}\quad\Phi_{\lambda,\delta}=\sum_{\begin{subarray}{c}k\in\mathbb{Z}^{3}\\ |\sqrt{Q(k)}-\lambda|<\delta\end{subarray}}e^{2\pi ik\cdot x}. By duality and TT^{*} , (1.1) \|P_{\lambda,\delta}\|_{L^{2}\to L^{p}}^{2}=\|P_{\lambda,\delta}\|_{L^{p^{\prime}}\to L^{p}}=\|P_{\lambda,\delta}\|_{L^{p^{\prime}}\to L^{2}}^{2}. The following conjecture appeared in [ 11 ] . Conjecture A . If \lambda^{-1}\leq\delta\leq 1 and p\geq 2 , then we have \|P_{\lambda,\delta}\|_{L^{2}\to L^{p}}\lesssim_{Q,p}(\lambda\delta)^{\frac{1}{2}-\frac{1}{p}}+\lambda^{1-\frac{3}{p}}\delta^{\frac{1}{2}}, except in certain boundary cases (see below). In other words \|P_{\lambda,\delta}\|_{L^{2}\to L^{p}}\lesssim_{Q,p}\begin{cases}\lambda^{1-\frac{3}{p}}\delta^{\frac{1}{2}}&\begin{aligned} &\text{if }4\leq p\leq 6\text{ and }\delta\geq\lambda^{2-\frac{p}{2}};\\ &\text{or }6\leq p\leq\infty\text{ and }\delta\geq\lambda^{-1},\end{aligned}\\ (\lambda\delta)^{\frac{1}{2}-\frac{1}{p}}&\begin{aligned} &\text{if }2\leq p\leq 4\text{ and }\delta\geq\lambda^{-1};\\ &\text{or }4\leq p\leq 6\text{ and }\lambda^{-1}\leq\delta\leq\lambda^{2-\frac{p}{2}}.\end{aligned}\end{cases} For some values of Q,p,\lambda,\delta a subpolynomial loss (“ \epsilon - loss”) should be added to the above conjecture. This is at least the case for p=\infty , \delta=\lambda^{-1} . The statement becomes then \|P_{\lambda,\delta}\|_{L^{2}\to L^{p}}\lesssim_{\epsilon}\lambda^{\epsilon}\left[(\lambda\delta)^{\frac{1}{2}-\frac{1}{p}}+\lambda^{1-\frac{3}{p}}\delta^{\frac{1}{2}}\right]. If this holds, we say that the conjecture holds with \epsilon -loss. The conjecture bounds the operator norm by a sum of two terms, namely (\lambda\delta)^{\frac{1}{2}-\frac{1}{p}} and \lambda^{1-\frac{3}{p}}\delta^{\frac{1}{2}} . This has the following geometric interpretation, see [ 11 ] for more details. The term (\lambda\delta)^{\frac{1}{2}-\frac{1}{p}} corresponds to the L^{p} norm of a quasimode 1 1 1 By definition, a quasimode f is such that P_{\lambda,\delta}f=f , it is thus nearly an eigenfunction. normalized in L^{2} which concentrates at a point; and the term \lambda^{1-\frac{3}{p}}\delta^{\frac{1}{2}} corresponds to the L^{p} norm of a quasimode normalized in L^{2} which concentrates on a closed geodesic. Therefore, we refer to the regions (\lambda\delta)^{\frac{1}{2}-\frac{1}{p}}<\lambda^{1-\frac{3}{p}}\delta^{\frac{1}{2}} and (\lambda\delta)^{\frac{1}{2}-\frac{1}{p}}>\lambda^{1-\frac{3}{p}}\delta^{\frac{1}{2}} as the point focusing regime and geodesic-focusing regime respectively. Let \lambda_{0}>1 . Those eigenvalues \lambda^{2} of the Laplacian -\Delta on the square torus which have size \sim\lambda_{0} have spacing \sim\frac{1}{\lambda_{0}} . Therefore, if \lambda^{2} is an eigenvalue of the Laplacian and \delta=\frac{1}{10}\lambda , the spectral projector P_{\lambda,\delta} is simply the orthogonal projection on the eigenspace associated to the eigenvalue {4\pi^{2}}\lambda^{2} . Then the above conjecture reduces to the earlier conjecture of Bourgain [ 2 , 5 ] . Conjecture B . If u is an eigenfunction of the Laplacian -\Delta on \mathbb{T}^{3} with eigenvalue \lambda^{2} , then \|u\|_{L^{p}}\lesssim_{\epsilon}\begin{cases}\lambda^{\frac{1}{2}-\frac{3}{p}+\epsilon}\|u\|_{L^{2}}&\mbox{if $p\geq 6$}\\ \lambda^{\epsilon}\|u\|_{L^{2}}&\mbox{if $p\leq 6$}\end{cases} (The subpolynomial loss \lambda^{\epsilon} might be superfluous if p<\infty ). 1.2. Known results in dimension 3 Earlier results on Conjecture A are as follows: • For \delta=1 , the conjecture corresponds to the fundamental result of Sogge [ 19 ] , which holds on any compact Riemannian manifold without boundary. • For p=\infty and on the square torus, the conjecture with \epsilon -loss is a consequence of the elementary estimate on the number of lattice points on the sphere \#(\mathbb{Z}^{3}\cap\lambda\mathbb{S}^{2})\lesssim_{\epsilon}\lambda^{1+\epsilon} . If \delta>\lambda^{-\frac{39}{48}} , the conjecture is satisfied without loss by Chamizo-Iwaniec [ 7 ] (Theorem 1.1). • For p=\infty and a general torus, the conjecture is satisfied if \delta>\lambda^{-\frac{85}{158}} by the main theorem of Guo [ 13 ] . • For p=6 and general tori, the conjecture was proved by Bourgain-Shao-Sogge-Yao [ 6 ] in the range \delta>\lambda^{-\frac{85}{252}} , which was then improved by Hickman [ 14 ] to \delta>\lambda^{-\frac{55}{162}} . • If 2\leq p\leq 4 or p\geq 4 and \delta>\max(\min(\lambda^{-\frac{3p-8}{5p-8}},\lambda^{-\frac{8-p}{5p-16}}),\lambda^{-\frac{1}{2}}) and on a general torus, it is shown in a work by the first two authors [ 11 ] that the conjecture holds with \epsilon -loss. • For 2\leq p\leq 4 and \delta>\lambda^{-1+\kappa} , \kappa>0 , and on a general torus, the conjecture is proved by the third author [ 18 ] following partial results by Blair-Huang-Sogge [ 1 ] . Regarding Conjecture B , the cases p=\infty and p=4 can be easily settled: the former case is equivalent to counting the number of lattice points on the sphere, while the latter can be treated via a basic arithmetic argument, as was noted in Bourgain [ 3 ] (an \epsilon -loss is incurred in both cases). It is difficult to do better than interpolating between these two bounds, but this was achieved recently by Mudgal [ 16 ] using methods from additive number theory. Indeed, it follows from Corollary 2.5 in [ 16 ] and an application of Hölder’s inequality that an eigenfunction of the Laplacian with eigenvalue 4\pi^{2}\lambda^{2} satisfies \|u\|_{L^{p}}\lesssim_{\epsilon}\lambda^{\frac{1}{2}-\frac{5}{2p}+\frac{3^{(2-p/2)}}{2p}+\epsilon}\|u\|_{L^{2}}\qquad\mbox{if $p\geq 4$, $p\in 2\mathbb{N}$}. Finally, we refer to [ 10 ] for background on estimating the operator norm of spectral projectors from L^{2} to L^{p} on the torus \mathbb{T}^{d} for d\neq 3 or on general Riemannian manifolds. 1.3. New results Our main result is as follows. Theorem 1.1 (Square torus) . We consider the square torus, and for some \kappa>0 , the range (1.2) \delta>\begin{cases}\lambda^{-1+\kappa}&2\leq p\leq 4,\\ \lambda^{-\frac{316-27p}{104(p-2)}+\kappa}&4<p\leq\frac{235}{52},\\ \lambda^{\frac{p}{2}-3+\kappa}&\frac{235}{52}\leq p<\frac{389}{79},\\ \lambda^{-\frac{9p-224}{444-158p}+\kappa}&\frac{389}{79}<p<5,\\ \lambda^{-\frac{1}{2}+\kappa}&5\leq p\leq 49,\\ \lambda^{-\frac{85p-358}{158p-128}}&49\leq p\leq\infty.\end{cases} Under these conditions, Conjecture A is satisfied, with an implicit constant that depends on \kappa . In the case \kappa=0 , it is satisfied with \epsilon -loss. The proof of this theorem is reached in Section 9 , by combining results obtained in previous sections. Theorem 1.2 (General torus) . The conclusion of Theorem 1.1 also holds for a general torus provided that (1.3) \#\{x\in\mathbb{Z}^{3},\,|\sqrt{Q(x)}-\lambda|<\delta\}\lesssim_{\epsilon}\lambda^{2+\epsilon}\delta for every \epsilon>0 (in other words, the number of lattice points in a thin shell is bounded by the volume of the shell up to a subpolynomial factor). Remark 1.3 . For the square torus the hypothesis ( 1.3 ) holds for \delta>\lambda^{-1} , by counting solutions to x_{1}^{2}+x_{2}^{2}+x_{3}^{2}=n for each integer n=\lambda^{2}+O(\lambda\delta) , and this range for \delta is optimal. For general tori, this bound is known to hold if \delta>\lambda^{-\frac{85}{158}} , as follows from the main result of Guo [ 13 ] referred to above; this implies A in that range. The bound ( 1.3 )

Read the complete text