VC-sets and generic compact domination (Q522325)

This article concerns a purely combinatorial result about topological groups that is motivated by a technical conjecture in model theory. First, fix an infinite set \(G\) and a family \(\mathcal{F}\) of subsets of \(G\). For \(A\subseteq G\), set \(\mathcal{F}\cap A:=\{S\cap A \;: \;S\in \mathcal{F}\}\). One says that \(\mathcal{F}\) has \textit{finite VC-dimension} if there is \(n\) such that, for every \(A\subset G\) of size \(n\), we have \(\mathcal{F}\cap A\not=\mathcal{P}(A)\). The main result of this article is that if \(G\) is a locally compact, second countable group with a fixed Haar measure \(\mu\) and \(X\subseteq G\) is a closed subset of \(X\) such that the family of left translates of \(X\) has finite VC-dimension, then the topological border \(\partial X\) of \(X\) has \(\mu\)-measure \(0\). This result extends to \(X\) that are merely constructible instead of closed if one makes a mild extra technical assumption. This combinatorial result is used to settle the generic compact domination conjecture for NIP groups. Although this conjecture had been claimed to have been proven in an earlier joint work with \textit{E. Hrushovski} and \textit{A. Pillay} [Trans. Am. Math. Soc. 365, No. 5, 2341--2366 (2013; Zbl 1294.03023)], a gap in that proof was found. Thus, the proof appearing in this paper is the first correct proof of this conjecture.