The lattice discrepancy of certain three-dimensional bodies (Q540454): Difference between revisions

Given a convex body \(\mathcal{B} \subseteq \mathbb{R}^3\), one might expect the number of integer lattice points inside the scaled body \(x \mathcal{B}\) to grow roughly like \(x^3 \text{volume}(\mathcal{B})\) as \(x\) becomes large. More precisely, one might hope for an asymptotic formula for this number of lattice points, consisting of a main term \(x^3 \text{volume}(\mathcal{B})\), some secondary terms reflecting any peculiar behaviour of the boundary of \(\mathcal{B}\), and some smaller error terms. For certain choices of \(\mathcal{B}\), such as the sphere, this problem has been long-studied, and today a huge amount of work has been done to establish increasingly precise results for different bodies \(\mathcal{B}\). In this paper, the authors prove such an asymptotic formula for the family of bodies \[ \mathcal{B}_{m,k} := \{(u_1,u_2,u_3) \in \mathbb{R}^3 : |u_1|^{mk} + (|u_2|^k + |u_3|^k)^m \leq 1\}, \] where \(k > 2, m > 1\) are any fixed real numbers satisfying \(mk \geq 7/3\). (The numerous ``big Oh'' terms arising in the paper then depend implicitly on \(m\) and \(k\).) This formula consists of the main term \(x^3 \text{volume}(\mathcal{B})\); two secondary terms, of order \(x^{2-2/mk}\) and \(x^{2-1/mk-1/k}\), reflecting the boundary behaviour of \(\mathcal{B}_{m,k}\) near the points \((\pm 1,0,0)\) and \((0,\pm 1, 0)\) and \((0,0,\pm 1)\); and three moderately complicated error terms. This sharpens a previous result of the first author [Commentat. Math. Univ. Carol. 43, No. 4, 755--771 (2002; Zbl 1064.11064)], in the handling of the three error terms. (Also the range of \(m\) and \(k\) treated here is slightly more than previously.) As the authors explain, the proof proceeds by cutting the scaled body \(x \mathcal{B}\) into slices: in other words, for each integer \(|n_1| \leq x\) one tries to count the number of two dimensional lattice points \((n_2,n_3)\) such that \[ (|n_2|^k + |n_3|^k)^m \leq x^{mk} - |n_1|^{mk}, \] and then one sums over all such \(n_1\). In sections 3 and 4 of their paper, the authors import numerous results from previous work, especially the paper [Zbl 1064.11064] of the first author, to carry out this plan. They are left to bound a certain sum of fractional parts, aiming to show that this is a small error term. This is achieved in sections 5 and 6 of the paper. In fact, almost all of the sum can be bounded quite straightforwardly using a two dimensional result of \textit{G. Kuba} [Abh. Math. Semin. Univ. Hamb. 63, 87--95 (1993; Zbl 0799.11037)], (as was already done in the previous work of the first author), but the part corresponding to large values \[ |n_1| > U \gg x^{(104mk - 131)/104mk} (\log x)^{-10307/4160mk} \] receives a more involved treatment using Van der Corput's exponential sum method, which supplies a small saving. The paper involves very elaborate calculations (particularly in the latter sections), and builds heavily on previous similar work, so the reviewer suspects that a non-expert reader might find it tough going. However, the paper is also carefully written, and the reviewer found the technical set-up in sections 5 and 6 to be rather neat and interesting.