Sumsets and Veronese varieties (Q6157471)

Given a finite subset \(\mathcal A \subset \mathbb Z^n\), the \(t\)-fold \textit{sumset} \(t \mathcal A\) is the set of all sums of \(t\) elements from \(\mathcal A\), not necessarily distinct. \textit{A. G. Khovanskij} proved that there is a unique polynomial \(p_{\mathcal A} (t) \in \mathbb Q [t]\) such that \(|t \mathcal A| = p_{\mathcal A} (t)\) for sufficiently large \(t \in \mathbb N\) [Funct. Anal. Appl. 26, No. 4, 276-281 (1992; Zbl 0809.13012); translation from Funkts. Anal. Prilozh. 26, No. 4, 57--63 (1992)]. The \textit{phase transition} is the minimal \(t_0\) for which \(|t \mathcal A| = p_{\mathcal A} (t)\) for all \(t \geq t_0\), denoted \(t_0 = n_0 (\mathcal A)\). Many authors have studied these invariants of \(\mathcal A\) from the perspective of lattice points and they are well understood for \(n=1\) [\textit{M. B. Nathanson}, Am. Math. Mon. 79, 1010--1012 (1972; Zbl 0251.10002); \textit{J. Wu} et al., Discrete Math. 311, No. 6, 408--412 (2011; Zbl 1225.11016)], but things are more complicated for \(n > 1\). Here the authors study \(p_{\mathcal A} (t)\) and \(n_0 (\mathcal A)\) through the geometry of projective varieties as suggested in work of \textit{Elias} [Mediterranean J. Math., to appear], \textit{V. Jelínek} and \textit{M. Klazar} [Adv. Appl. Math. 41, No. 1, 115--132 (2008; Zbl 1137.13011)] and \textit{E. Miller} and \textit{B. Sturmfels} [Combinatorial commutative algebra. New York, NY: Springer (2005; Zbl 1090.13001)], associating a unique variety \(Y\) to \(\mathcal A\) as follows. There is a unique translation \(\tau: \mathbb Z^n \to \mathbb Z^n\) so that \(\mathcal A \subset \mathbb Z^n_{\geq 0}\) and the monomials in \(\Omega = \{ m_a = x_0^{d_{\mathcal A}-|a|} x_1^{a_1} \dots x_n^{a_n}|a \in \mathcal A\}\) are relatively prime, where \(d_{\mathcal A} = \max\{|\underline a|=a_1+\dots+a_n: \underline a \in \mathcal A\}\). Since \(|t \mathcal A| = |t \tau(\mathcal A)|\), we may replace \(\mathcal A\) with \(\tau (\mathcal A)\) to study the size of the sumsets. Taking \(k\) to be a field and \(S = k[w_1, \dots, w_m]\) the polynomial ring in variables \(w_i\), one for each element in \(\mathcal A\), the map \(w_i \mapsto m_i\) extends to a ring homomorphism \(S \to k[\Omega] \subset k[x_0, \dots, x_n]\) whose kernel is a prime ideal \(I\) defining \(Y\). The Hilbert function for \(Y\) gives the cardinality of sumsets: \(\mathrm{HF}_Y (t) = |t \mathcal A|\) for all \(t \geq 0\), so that \(p_{\mathcal A} (t)\) is the Hilbert polynomial for \(Y\) and satisfies \(\deg p_{\mathcal A} (t) = \dim Y \leq n\), with equality if and only if \(\mathbb Z (\mathcal A - \mathcal A)\) has maximal rank. In case of equality (\(\dim Y = n\)), the authors compute \(p_{\mathcal A} (t)\) when \(|\mathcal A| = n+1\) or \(n+2\) and give a general formula for the leading coefficient of \(p_{\mathcal A} (t)\) in terms of various projections by viewing \(Y\) as a monomial projection of a Veronese variety \(X\). Assuming the convex hull \(\mathrm{conv} (\mathcal A)\) is an \(n\)-simplex, the authors deduce an upper bound on \(n_0 (\mathcal A)\) using recent work of \textit{Curran and Goldmakher} [arXiv:2009.02140v3] and \textit{Granville, Shakan and Walker} [arXiv:2006.01041v2] as well as giving more refined bounds under special conditions, such as knowledge of the reduction number of the semigroup ring \(k[\Omega]\). They also note the bound \(n_0 (\mathcal A) \leq \mathrm{reg} Y + 1\) arising from the Castelnuovo-Mumford regularity of \(Y\). In the last part the authors consider special subsets \(\mathcal A \subset \mathbb A^n\) that arise from solving linear systems of congruences called \(\mathrm{GT}\)-subsets. For these they obtain the stronger bound \(n_0 (\mathcal A) \leq n+1\) and give examples where \(n_0 (\mathcal A) =0\) and where they can compute \(p_{\mathcal A} (t)\) exactly. Finally, they compute the degree and Castelnuovo-Mumford regularity for \(\mathrm{RL}\)-varieties introduced by \textit{Colarte-Gómez and Miró-Roig} [Mediterranean J. Math, to appear].