On maximal convex lattice polygons inscribed in a plane convex set (Q2472723)

Let \(K \subset \mathbb R^2\) be a convex compact set with nonempty interior, and for each positive real number \(t\) define a lattice \(\mathbb Z_t = \frac{1}{t} \mathbb Z^2\) in \(\mathbb R^2\). A convex \(\mathbb Z_t\)-lattice \(n\)-gon is a convex polygon with \(n\) vertices, all of which are points of \(\mathbb Z_t\). Define \[ m(K,\mathbb Z_t)= \max\{n: \text{\(K\) contains a convex \(\mathbb Z_t\)-lattice \(n\)-gon}\}, \] that is, \(m(K,\mathbb Z_t)\) is the maximal number of \(\mathbb Z_t\)-lattice points in \(K\) in convex position. The goal of the paper under review is to determine asymptotic behavior of \(m(K,\mathbb Z_t)\) as \(t \to \infty\). In order to state the authors' main result we need additional notation. Let \(S\) be a compact convex set with nonempty interior in \(\mathbb R^2\), and let \[ x_1,\dots,x_n,x_{n+1}=x_1 \] be a subdivision of the boundary \(\partial S\) of \(S\). For each \(1 \leq i \leq n\), let \(l_i\) be the line supporting \(S\) at \(x_i\), and let \(y_i\) be the intersection of \(l_i\) and \(l_{i+1}\). Write \(T_i\) for the area of the triangle \(\text{conv} \{x_i,y_i,x_{i+1}\}\), and define the affine perimeter of \(S\) to be \[ AP(S) = 2 \lim \sum_{i=1}^n \root 3\of{T_i}, \] where the limit is taken over a sequence of subdivisions of \(\partial S\) with \(\max_{1 \leq i \leq n} |x_{i+1}-x_i| \to 0\); this limit exists, and is independent of the sequence of subdivisions. In particular, affine perimeter of a polygon is 0. Now define \[ A(K) = \sup \{ AP(S) : S \subset K,\;S \text{ is convex} \}. \] With this notation, the authors prove that \[ \lim_{t \to \infty} t^{-2/3} m(K,\mathbb Z_t) = \frac{3}{(2\pi)^{2/3}} A(K). \] It is an earlier result of the first author [J. Reine Angew. Math. 484, 71--84 (1997; Zbl 0864.52010)] that supremum in the definition of \(A(K)\) is achieved, i.e. there exists a unique \(K_0 \subset K\) such that \(AP(K_0)=A(K)\). In fact, if \(Q_t\) is any maximizer in the definition of \(m(K,\mathbb Z_t)\), then the authors show that \[ \lim_{t \to \infty} \text{dist}(Q_t,K_0) = 0, \] where \(\text{dist}\) stands for Hausdorff distance between sets. It should be remarked that the problem of estimating \(m(K,\mathbb Z_t)\) has a long history, going back to the classical work of Jarník (1925). More recently, an estimate of the form \[ m(K,\mathbb Z_t) \leq 5t^{2/3} (\text{Area } K)^{1/3} \] follows from the work of Andrews and Rényi and Sulanke [\textit{G. E. Andrews}, Trans. Am. Math. Soc. 106, 270--279 (1963; Zbl 0118.28301); \textit{A. Rényi} and \textit{R. Sulanke}, Z. Wahrscheinlichkeits\-theor. Verw. Geb. 2, 75--84 (1963; Zbl 0118.13701)]. This result is in accordance with the authors' present asymptotic estimate.