Semi-monotone sets (Q1940824)

For a bounded, open subset \(U\subseteq{\mathbb R}^n\), the notion that \(U\) is a semi-monotone set can be defined in an elementary way to mean that \(U\) has a connected intersection with any (possibly degenerate) box with sides parallel to the coordinate hyperplanes. This is an analogue of Euclidean convexity but depends on the coordinate system. The main result is that if \(U\) is non-empty, semi-monotone, and definable with respect to some o-minimal structure over \({\mathbb R}\), with closure \(\overline{U}\), then \((\overline{U},U)\) is definably homeomorphic to \(([-1,1]^n,(-1,1)^n)\). The steps in the proof are given as Lemmas, some of which are elementary, others depend on results of PL topology, stated in an Appendix, that require treating the \(n\leq5\) and \(n>5\) cases differently. The semialgebraic case is also considered over an arbitrary real closed field. The study of the semi-monotone property is motivated by a conjecture on the existence of a triangulation of a graph of a definable function from a subset of \({\mathbb R}^m\) to \(\mathbb R\), compatible with the coordinate system of \({\mathbb R}^{m+1}\).