Convergence of Madelung-like lattice sums (Q4211095)

The authors investigate convergence properties of sums of the form \[ A_\nu(s):= \sum_{(x_1,\dots, x_k)\in C_\nu} \frac{(-1)^{x_1+\cdots+ x_k}} {Q(x_1,\dots, x_k)^s} \] where \(Q\) is a positive definite symmetric quadratic form and \(C_\nu= \nu C\cap (\mathbb{Z}^k \setminus 0)\) with \(C\subseteq \mathbb{R}^k\) bounded. The prototype is Madelung's constant for NaCl: \[ \sum_{-\infty}^\infty \frac{(-1)^{n+m+p}} {\sqrt{n^2+m^2+p^2}}= -1.74756459\dots, \] which depends on the summation process. The main result is that \(\lim_{\nu\to\infty} A_\nu(s)\) exists and is analytic down to \(\text{Re } s>(k-1)/2\) for all reasonably shaped regions \(C\) (e.g. convex in one of the variables), and hence that the limit is independent of such \(C\). Moreover, they prove that for any specific regions \(C\) (e.g., certain ellipses or polygons) the area of convergence can be extended; for rectangles even down to \(\text{Re } s>0\).