Monotone functions and maps (Q1940343)

This paper is a continuation of the previous work by the authors [\textit{S. Basu, A. Gabrielov} and \textit{N. Vorobjov}, J. Eur. Math. Soc. (JEMS) 15, No. 2, 635--657 (2013; Zbl 1284.14081)], where they introduced the class of \textit{semi-monotone} sets: open, bounded and definable sets in an o-minimal structure over \(\mathbb R\) (for example, real semialgebraic or subanalytic sets). One of the main results of the quoted article is that semi-monotone sets are topologically regular cells, and the authors extend this result in the paper under review to sets of any codimension (notice that open subsets are \(0\)-codimensional). Here, a definable set \(X\subset\mathbb R^n\) is called a \textit{topologically regular \(m\)-cell} if the pair \((\overline{X},X)\) is definable homeomorphic to the pair \(([-1,1]^m,(-1,1)^m)\). The immediate motivation for defining the class of monotone sets was to prove the following conjecture: Conjecture. Let \(K\subset\mathbb R^m\) be a compact definable set and let \(f:K\to\mathbb R\) be a definable function. Then, there exists a definable triangulation of \(K\) such that for each \(n\leq\dim (K)\) and for each open \(n\)-simplex \(\Delta\) of the triangulation, the following conditions hold: {\parindent=6mm \begin{itemize}\item[(1)] The graph \(\Gamma:=\{(x,t)\in\Delta\times\mathbb R:f(x)=t\}\) of the restriction \(f|_{\Delta}\) is a topologically regular \(n\)-cell; \item[(2)] Either \(f|_{\Delta}\) is constant, or each non-empty level set \(\Gamma\cap\{t=k\}\) is a topologically regular \((n-1)\)-cell. \end{itemize}} To approach the proof of this conjecture the authors introduce the notion of \textit{monotone map} defined on a nonempty semi-monotone set \(X\subset\mathbb R^n\) and prove their main result: the graph of a monotone map is a topologically regular cell. As a corollary the authors prove that each \textit{open toric cube} is the graph of a monotone map. The notion of toric cube was introduced in [\textit{A. Engström, P. Hersh} and \textit{B. Sturmfels}, Rend. Circ. Mat. Palermo (2) 62, No. 1, 67--78 (2013; Zbl 1315.52011)]. A subset \(C\) of \([0,1]^n\) is a \textit{toric cube} if there exist a positive integer \(d\) and a subset \({\mathcal A}:=\{a_1,\dots, a_n\}\subset\mathbb N^d\) such that \(C\) is the image of the map \[ f_{\mathcal A}:[0,1]^d\to[0,1]^n,\, t=(t_1,\dots, t_d)\mapsto (t^{a_1},\dots, t^{a_n}), \] where \(t^{a_i}:=t_1^{a_{i,1}}\cdots t_d^{a_{i,d}}\). An open toric cube is the image of \((0,1)^d\) by \(f_{\mathcal A}\). The authors use this result to answer in the affirmative the following conjecture, formulated in the quoted article by Engström, Hersh and Sturmfels: Each CW-complex in which the closure of each cell is a toric cube is a regular cell complex. This article is written in an unusually clear style, closer to very good books than to research papers. In conjunction with the already quoted article by the same authors constitutes an excellent material to introduce graduate students in this exciting field of mathematics.