Lattice polytopes, Hecke operators, and the Ehrhart polynomial (Q2474165)

From the text: Let \(L\) be a rank \(n\) lattice, embedded in a real \(n\)-dimensional vector space \(V\). Let \(\mathcal P(L)\) be the set of \(n\)-dimensional convex polytopes in \(V\) with vertices in \(L\). For any \(P\in \mathcal P(L)\), and for any nonnegative integer \(t\), let \(tP\) be \(P\) scaled by the factor \(t\). Then by a result of \textit{E. Ehrhart} [J. Reine Angew. Math. 226, 1--29 (1967); 227, 25--49 (1967; Zbl 0155.37503)], the function \(t \to \#(tP \cap L)\) is a degree \(n\) polynomial with rational coefficients, called the Ehrhart polynomial of \(P\). Hence one can think of the Ehrhart polynomial as giving a map \(E\) from \(\mathcal P(L)\) to the polynomial ring \(\mathbb Q[t]\). The Ehrhart polynomial depends not just on the combinatorial type of \(P\), but rather on the pair \((P,L)\). Hence it is natural to consider how \(E(P)\) changes as \(L\) is varied. The theory of automorphic forms provides a powerful machine to accomplish this, namely the technique of Hecke operators. Thus let \(p\) be a prime, and let \(k\leq n\) be a positive integer. Given a lattice polytope \(P\) with Ehrhart polynomial \(E(P)\), the authors define a new polynomial \(T(p, k)E(P)\) as follows. Let \(p^{-1}L\) be the canonical superlattice of \(L\) of coindex \(p^n\). Then \(p^{-1}L/L \simeq\mathbb F^n_p\), and any lattice \(M\) satisfying \(p^{-1}L\supset M\supset L\) determines a subspace \(M\subset \mathbb F^n_p\). Let \(\mathcal L_k\) be the set of such lattices with \(\dim M = k\). Then we define \[ T(p, k)E(P) =\sum_{M\in \mathcal L_k}E(P_M),\tag{1} \] where \(P_M\in \mathcal P(M)\) denotes the lattice polytope with vertices in \(M\) canonically determined by \(P\). In this paper the authors consider the relationship between \(T(p, k)E(P)\) and \(E(P)\). For any nonnegative integer \(l\leq n\), fix an \(l\)-dimensional subspace \(U\) of \(\mathbb F^n_p\), and define \[ \nu_{n,k,l}(p) = \sum_{{W\subset\mathbb F^n_p}\atop{\dim W=k}}p^{\dim W\cap U}. \] This value is independent of the choice of \(U\). Finally, for any polynomial \(f\in \mathbb Q[t]\) let \(c_l(f)\) be the coefficient of \(t^l\) in \(f\). Then their first main result states: Theorem 1.4. For each triple \((n, k, l)\), there is a polynomial with positive coefficients \(\Phi_{n,k,l}(t)\in \mathbb Z[t]\), independent of \(p\), such that \[ \Phi_{n,k,l}(p) = \nu_{n,k,l}(p). \] Moreover, \[ c_l(T(p, k)E(P)) = \nu_{n,k,l}(p)c_l(E(P)),\tag{2} \] independently of \(P\). The ratios \(\nu\) satisfy \[ \nu_{n,k,l}(p)/\nu_{n,n-k,n-l}(p) = p^{k+l-n}. \] Their second main result describes how the Brion-Vergne formula [\textit{M. Brion} and \textit{M. Vergne}, J. Am. Math. Soc. 10, 797--833 (1997; Zbl 0926.52016)] for \(E(P)\) transforms under the Hecke operators for nonsingular lattice polytopes \(P\) (see Theorem 1.8). The proof of Theorem 1.4 is a counting argument. The new lattice points appearing in \(P\) in the sum (1) all lie in the superlattice \(p^{-1}L\), and to compute \(T(p, k)E(P)\) one keeps track of which lattice points appear in a given Hecke image. This gives an expression for \(T(p, k)E(P)\) in terms of \(E(P)(t)\), \(E(P)(p^t)\), and the cardinalities of some finite Grassmannians. An additional argument shows that this expression implies (2). The proof of Theorem 1.8 is more complicated. At the heart of main identity are certain ``distribution relations'' of Dedekind sums, essentially coming from a distribution relation satisfied by the Hurwitz zeta function. In the proof of Theorem 1.8, these relations appear in identities involving Dedekind sums and the cardinalities of strata in certain stratifications of finite Grassmannians. Rather than proving these identities directly, it is shown that they occur in the computation of the constant term of \(T(p, j)E(P)\) for lower-dimensional polytopes \(P\) and for \(j\leq k\). Since these constant terms are always 1, by appropriately choosing \(P\) the authors show that their identities hold. Then they use induction to complete the argument. The paper is organized as follows. Section 2 recalls background about lattice polytopes and their normal fans, and discusses the connection between Hecke operators and finite Grassmannians. Section 3 gives the proof of Theorem 1.4. Section 4 discusses the computation of the Ehrhart polynomial using the Todd operator, and Section 5 gives the proof of Theorem 1.8. Section 6 discusses explicit examples of Theorem 1.8 for three-dimensional polytopes, and relates the identities occurring in the proof of Theorem 1.8 to Dedekind sums and the Hurwitz zeta function. Finally, Section 7 addresses the problem of computing the average Ehrhart polynomial as one varies over a family of superlattices.