Algebraic vertices of non-convex polyhedra (Q507217): Difference between revisions

Let \(P\subset\mathbb{R}^d\) be a generalized polyhedron, i.e., the union of convex polyhedra (not necessarily bounded) in \(\mathbb{R}^d\). The authors introduce the notion of algebraic vertex of \(P\). A point \(v\in P\) is called an algebraic vertex of \(P\) if the indicator function of its tangent cone cannot be represented (up to a set of measure zero) as a linear combination of indicator functions of line-cones (i.e., union of parallel lines). The tangent cone of \(P\) at \(v\in P\) is given by \(\{v+x\,:\,v+\epsilon x\in P\text{ for some }\epsilon>0\}\). The notion of algebraic vertex is motivated by the authors in the last section of the paper. If \(P\) is a convex polytope, then the set of algebraic vertices coincides with the set of vertices of \(P\). The authors prove that the set of algebraic vertices can be considered as the smallest set of points needed to define the generalized polyhedron \(P\), in the sense that the indicator function of \(P\) is given (up to a set of measure zero) by a linear combination, with integer coefficients, of the indicator function of the simplices whose vertices lie in the set of algebraic vertices of \(P\). A characterization of the algebraic vertices of \(P\) is given by using the Fourier-Laplace transform of \(P\), which is carefully introduced. The authors prove, in a more general context, that \(v\in P\) is an algebraic vertex of \(P\) if and only if the tangent cone of \(P\) at \(v\) has non-zero Fourier-Laplace transform.