Brunn-Minkowski type inequalities for the lattice point enumerator

From MaRDI portal
Publication:2189483



Abstract: Geometric and functional Brunn-Minkowski type inequalities for the lattice point enumerator mathrmGn(cdot) are provided. In particular, we show that mathrm{G}_n((1-lambda)K + lambda L + (-1,1)^n)^{1/n}geq (1-lambda)mathrm{G}_n(K)^{1/n}+lambdamathrm{G}_n(L)^{1/n} for any non-empty bounded sets K,LsubsetmathbbRn and all lambdain(0,1). We also show that these new discrete versions imply the classical results, and discuss some links with other related inequalities.


The paper under review concerns the classical Brunn-Minkowski inequality, \[ \mathrm{Vol}( (1-\lambda)K+\lambda L)^{\frac{1}{n}} \geq (1-\lambda) \mathrm{Vol}(K)^{\frac{1}{n}} + \lambda \mathrm{Vol}(L)^{\frac{1}{n}}, \] where \(K\), \(L\) are non-empty compact subsets in \(\mathbb R^n\), and \(0<\lambda < 1\). Asking for discrete versions of the latter, see [\textit{M. A. Hernandez Cifre} et al., SIAM J. Discrete Math. 32, 1840--1856 (2018; Zbl 1396.52013)], the authors consider the integer lattice point enumerator, \(G_n(M):= \sharp\left( M\cap\mathbb Z^n\right)\), \(M\subset\mathbb R^n\), instead of the volume. As trivial examples demonstrate, a direct analogue of the Brunn-Minkowski inequality for \(G_n\) does not hold. However it turns out to be true, if one slightly modify the Minkowski combination of subsets. Namely, the following sharp inequality for \(G_n\) is proved: \[ G_n( (1-\lambda)K+\lambda L+(-1,1)^n)^{\frac{1}{n}} \geq (1-\lambda) G_n(K)^{\frac{1}{n}} + \lambda G_n(L)^{\frac{1}{n}}, \] where \(K\), \(L\) are non-empty bounded subsets in \(\mathbb R^n\), and \(0<\lambda < 1\). Moreover, one gets \[ G_n( \tfrac{1}{2}K+\tfrac{1}{2} L+[0,1]^n)^{\frac{1}{n}} \geq \tfrac{1}{2} G_n(K)^{\frac{1}{n}} + \tfrac{1}{2} G_n(L)^{\frac{1}{n}} , \] for any bounded subsets \(K\), \(L\subset \mathbb R^n\) satisfying \(G_n(K)G_n(L)>0\). Multiplicative discrete versions of the Brunn-Minkowski inequality are discussed too. It is demonstrated that the following inequality holds true, \[ G_n( (1-\lambda)K+\lambda L+(-1,1)^n) \geq G_n(K)^{1-\lambda} G_n(L)^{\lambda}, \] for any non-empty bounded subsets \(K\), \(L\subset \mathbb R^n\). The proofs are based on the use of a novel fundamental functional inequality extending the Borell-Brascamp-Liemb inequality.











This page was built for publication: Brunn-Minkowski type inequalities for the lattice point enumerator

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2189483)