Optimal stretching for lattice points under convex curves (Q1687816)

Summary: Suppose we count the positive integer lattice points beneath a convex decreasing curve in the first quadrant having equal intercepts. Then stretch in the coordinate directions so as to preserve the area under the curve, and again count lattice points. Which choice of stretch factor will maximize the lattice point count? We show the optimal stretch factor approaches \(1\) as the area approaches infinity. In particular, when \(0 < p < 1\), among \(p\)-ellipses \(|sx|^p+|s^{-1}y|^p=r^p\) with \(s > 0\), the one enclosing the most first-quadrant lattice points approaches a \(p\)-circle (\(s=1\)) as \(r \rightarrow \infty\). The case \(p=2\) was established by \textit{P. R. S. Antunes} and \textit{P. Freitas} [J. Optim. Theory Appl. 154, No. 1, 235--257 (2012; Zbl 1252.90076); Proc. R. Soc. Lond., Ser. A, Math. Phys. Eng. Sci. 469, No. 2150, Article ID 20120492, 15 p. (2013; Zbl 1371.49040); Appl. Math. Optim. 73, No. 2, 313--328 (2016; Zbl 1339.35338)], with generalization to \(1< p <\infty\) by the second author with \textit{S. Liu} [Ark. Mat. 56, No. 1, 111--145 (2018; Zbl 1390.35206)]. The behaviour in the borderline case \(p=1\) (lattice points in right triangles) is quite different, as shown recently by \textit{N. F. Marshall} and \textit{S. Steinerberger} [``Triangles capturing many lattice points'', Mathematika 64, No. 2, 551--582 (2018; \url{doi:10.1112/s0025579318000219})] .