\(h\)-vectors of Gorenstein polytopes (Q857415)

Let \(P \subset {\mathbb R}^n\) be an integral convex \(d\)-polytope, i.e., the \(d\)-dimensional convex hull of finitely many points in \({\mathbb Z}^n\), and denote by \(L_P(t) := \#(P\cap {1 \over t} {\mathbb Z}^n )\) the integer-point counting function associated to \(P\). By Ehrhart's theorem, \(L_P(t)\) is a polynomial of \(t\), and thus its generating function (the Ehrhart series) is a rational function of the form \[ \sum_{ k \geq 0 } L_P(t) \, x^t = {{ h_m x^m + h_{ m-1 } x^{ m-1 } + \cdots + h_0 } \over { (1-x)^{ d+1 } }} \, . \] We call \(h(P) =( h_0, h_1, \dots, h_m)\) the Ehrhart \(h\)-vector of \(P\). The paper under review studies the question when \(h(P)\) is unimodal, i.e., \(h_0 \leq h_1 \leq \cdots \leq h_k \geq h_{ k+1 } \geq \cdots \geq h_m\) for some~\(k\). To state the main result of the paper, we need to define two notions. First, a unimodular triangulation of \(P\) is a triangulation into integral simplices \(\text{conv}( v_0, v_1, \dots, v_d )\) such that \(s_1 - s_0, s_2 - s_0, \dots, s_d - s_0\) generate the lattice \(\text{span} (P) \cap {\mathbb Z}^n\). The regular triangulations are obtained as subdivisions of \(P\) into the domains of linearity of a piecewise affine, concave, and continuous function on~\(P\). For the second notion, consider the cone \(C(P)\) generated by \(P \times \{ 1 \} \subset {\mathbb R}^{ n+1 }\) and the associated monoid \(E(P) = C(P) \cap {\mathbb Z}^{ n+1 }\). The Ehrhart polynomial of \(P\) can then be interpreted as the Hilbert function of the affine monoid algebra \(K[E(P)]\) consisting of monomials \(z_1^{ m_1 } z_2^{ m_2 } \cdots z_{ n+1 }^{ m_{ n+1 } }\), where \((m_1, m_2, \dots, m_{ n+1 } ) \in E(P)\). The algebra \(K[E(P)]\) is Gorenstein if \(E(P) \cap \operatorname{int} C(P) = y + E(P)\) for some \(y \in E(P)\). The main theorem asserts that if an integral polytope \(P\) has a regular unimodular triangulation and \(K[E(P)]\) is Gorenstein, then the Ehrhart \(h\)-vector of \(P\) is unimodal. This theorem generalizes a recent result of Athanasiadis on the Ehrhart \(h\)-vector of the Birkhoff polytope [\textit{C. A. Athanasiadis}, J.~Reine Angew.~Math.~583, 163--174 (2005; Zbl 1077.52011)]. The authors derive their theorem from a more general theorem on Gorenstein affine normal monoids \(M\); one can factor \(K[M]\) by a ``long'' regular sequence in such a way that the quotient is still a normal affine monoid algebra. This technique reduces any question about the Ehrhart \(h\)-vector of \(P\) to the Ehrhart \(h\)-vector of a Gorenstein polytope \(Q\) with exactly one interior lattice point, provided that each lattice point in \(kP\) (for any \(k \in {\mathbb N}\)) can be written as the sum of \(k\) lattice points in \(P\).