Integer chords and configurations of lattice points (Q266164)

Let \(C\) be a closed convex plane curve, with radius of curvature \(\rho\), arc length \(s\) and tangential direction \(\psi\), so that the curve \(C\) has intrinsic equation \(s=s(\psi)\), where \(\rho=\mathrm{d}s/\mathrm{d}\psi\). The curve \(C\) is said to satisfy the \(C^{\,3}\) \textit{smoothness condition} when \(s\) is twice continuously differentiable, and there exist positive constants \(c_1, c_2\), and \(c_3\), with \[ 0< c_1\leq\rho \leq c_2,\qquad \left |\frac{\mathrm{d}\rho}{\mathrm{d}\psi}\right |\leq c_3, \] so that the radius of curvature is bounded away from 0 and \(\infty\). In this present work the authors estimate the area within a closed convex plane curve \(C\), by enlarging \(C\) by a factor \(R\), translating, counting the set \(J\) of integer points inside the closed curve \(C\), and then scaling back to the original size. They show that if \(C\) satisfies the \(C^{\,3}\) smoothness condition, then the estimate is accurate. More specifically, let \(C(R,u,v)\) be the homothetic image of \(C\), formed by enlarging \(C\) by a factor \(R\), and then translating the centre of the enlarged curve to the point \((u,v)\). The centre of \(C\) is taken to be an arbitrary interior point of \(C\), which is then designated to be the origin of the planar \((x,y)\) coordinate system. Let \(S(R,u,v)\) be the closed convex set bounded by \(C(R,u,v)\), and \(J(R,u,v)\) the set of integer points in \(S(R,u,v)\). Considering the case where \(R\) is fixed and the centre \((u,v)\) varies, gives rise to the domain \(D(J)\), of a configuration \(J\), consisting of the set of centres \((u,v)\), for which \(J(R,u,v)=J\). The authors show that the domains \(D(J)\) are periodic modulo one, forming a repeating pattern, and so are unchanged under translation by an integer vector. Moreover, even though the set \(J\) is very sensitive to translations of the curve, it is shown that as \(R\) tends to infinity, the domains in which each set \(J\) occurs tend to uniform distribution modulo the integer lattice (a result which was only known for the special case of the circle). With \(B\) the area of the difference set of \(C\), and \(T\) the rectangle in the unit square defined by \(\gamma\leq u\leq\alpha\), \(\delta\leq v\leq \beta\), it is shown in Theorem 1 that each of the three properties comprising the number of integer points \((m,n)\) in the difference set of \(C(R,0,0)\); the number of domains \(D(J)\) that meet \(T\), or the number of domains \(D(J)\) that lie entirely inside \(T\), are of the form \[ (\alpha-\gamma)(\beta-\delta)B R^2+O\left (R^{1+\kappa}(\log{R})^\lambda\right ), \] where \[ \kappa=\frac{131}{208}=0.6298\dots,\qquad \lambda=\frac{18627}{8320}=2.2513\dots \] The techniques employed here are intricate, with the approach building on methods pioneered by \textit{J. G. van der Corput} [Over roosterpunten in het platte vlak. (De beteekenis van de methoden van Voronoï en Pfeiffer). Leiden, Groningen: Noordhoff (1919; JFM 47.0158.02)].