Un critère pour reconnaître les fonctions algébriquement constructibles (Q4513072)

Let \(V\) be an algebraic subset of \(\mathbb{R}^n\). A function \(\phi: V\to \mathbb{Z}\) is said to be constructible if \(\phi\) is constant on each member of some finite partition of \(V\) in semialgebraic subsets. A typical constructible function on \(V\) is a sum of signs of finitely many polynomials on \(V\). In fact, these constructible functions are called algebraically constructible [\textit{C. McCrory} and \textit{A. Parusiński}, Ann. Sci. Éc. Norm. Supér., IV. Sér. 30, No. 4, 527-552 (1997; Zbl 0913.14018)]. The question is to characterize those constructible functions that are algebraically constructible. Due to general results on real spaces [see the book by \textit{C. Andradas, L. Broecker} and \textit{J. M. Ruiz}: ``Constructible sets in real geometry'' (1996; Zbl 0873.14044)], one has a theoretical criterion for a constructible function to be algebraically constructible. This is a criterion using fans, and seems to be hard to handle in practice. In the first part of the paper under review, the author establishes a nice inductive criterion for a constructible function on a nonsingular compact algebraic subset to be algebraically constructible (theorems 2 and 3). With this criterion at hand, it becomes considerably easier to decide whether or not a constructible function is algebraically constructible. In the second part, the author establishes an explicit upper bound on the number of polynomials that are needed to write a given algebraically constructible function as a sum of signs of polynomials. She also shows that her bound is sharp in case \(V=\mathbb{R}^n\).