Sur l'adhérence d'un ensemble partiellement semi-algébrique. (On the closure of a partially semialgebraic set) (Q2639120)

The author shows in the affirmative the following conjecture of \textit{R. Thom}: ``Given an analytic vector field X over an analytic manifold M, the set of limit directions of X at a point \(a\in M\) is a semialgebraic subset of \({\mathbb{P}}(T_ aM)''\). - To understand this, let X: \(M\to TM\) be an analytic vector field, and let q: \({\mathbb{P}}TM\to M\) be the projective tangent bundle of M. Then if \(Z=\{x\in M| X(x)=0\}\), we have an analytic section s: \(M\setminus Z\to {\mathbb{P}}TM\) defined by \(s(x)={\mathbb{R}}X(x)\), and the set of limit directions of X at a is defined as \(T_ aM\cap \overline{s(M\setminus Z)}.\) On the other hand, the author defines a subset \(S\subset M\times {\mathbb{R}}^ n\) to be semialgebraic if for each point \(x\in M\) there is a neighbourhood U such that S can be described in \(U\times {\mathbb{R}}^ n\) by a boolean formula using functions which are polynomials in \(T_ 1,...,T_ n\) with coefficients analytic functions on U. (In particular if M is a single point, this definition includes the classical one.) Since a subset of the projective space is called semialgebraic if its image in any affine chart is semialgebraic, then the result follows from the main theorem of the paper: The closure of a semialgebraic subset of \(M\times {\mathbb{R}}^ n\) is again semialgebraic. The proof of this result is spread out in different technical lemmata, but it is elementary and easy to read.