Asymptotic distribution of the zeros of the Ehrhart polynomial of the cross-polytope (Q387258)

Given a convex polytope \(\mathcal{P} \subset \mathbb{R}^d\) with vertices on the lattice \(\mathbb{Z}^d\), the lattice point counting function is defined as \(\mathcal{L}_\mathcal{P}(m) = \#(\mathbb{Z}^d \cap m \mathcal{P})\). The function \(\mathcal{L}_\mathcal{P}\) is in fact a polynomial in \(m\), called the Ehrhart polynomial of \(\mathcal{P}\). The present article is concerned with the zero distribution of the Ehrhart polynomial of the \(d\)-dimensional cross-polytope, i.e., the convex hull of the set of unit vectors \(\{\pm \vec{e}_j | j = 1,\dots,d \}\) in \(\mathbb{R}^d\), given by the formula \[ \mathcal{L}_d(x) = \sum_{n=0}^d {d \choose n} { d+x-n \choose d}. \] These have some interesting properties, for example, all zeros of \(\mathcal{L}_d\) have real part \(-\frac{1}{2}\). The article studies the distribution of zeros of \(\mathcal{L}_d\) on this line in the limit where \(d \to \infty\). Two results are proved, one regarding the `small zeros' of \(\mathcal{L}_d\), the other regarding the `larger zeros'. For the small zeros, let \(a_d\) and \(b_d\) be sequences of numbers such that \(a_d < b_d\) and \(|a_d|,|b_d| = o(\log d)\). The asymptotic formula \[ \# \{ x : \mathcal{L}_d(x) = 0, \Im \{x\} \in [a_d,b_d] \} = \left[ \frac{\log d}{\pi} + o(\log d) \right] (b_d-a_d) + O(1), \] as \(d \to \infty\), is obtained. For the larger zeros, the author uses the method of steepest descents to evaluate a certain integral representation for the Ehrhart polynomials to obtain a counting function for the zeros, \[ \begin{multlined} \# \{ x : \mathcal{L}_d(x) = 0, \Im \{x\} \in (-d,\tau) \} = \\ \frac{2d+1}{\pi} \arctan \left( \frac{d}{\tau} - \sqrt{\frac{d^2}{\tau^2} - 1} \right) - \frac{\tau}{\pi} \log \Bigg|\frac{d}{\tau} - \sqrt{\frac{d^2}{\tau^2} - 1}\Bigg| + \frac{d}{2} + O(1).\end{multlined} \] Also, the largest zero of of \(\mathcal{L}_d\) is proven to have an imaginary part \(d-\epsilon_d\) with \(\epsilon_d > 0\) and \(\epsilon_d = O(d^{1/3})\).