On basic semialgebraic sets (Q1182512)

The paper under review intends to be an enlightening explanation of the interplay between the theory of spaces of orderings (developed mainly by \textit{M. Marshall}), the real spectrum (introduced by \textit{M. Coste} and \textit{F. Roy}) and semialgebraic geometry to study the problem of the description of semialgebraic sets by few equations. Namely, let \(R\) be a real closed field and \(V\) a real algebraic variety over \(R\). A subset \(S\) of \(V\) is open semialgebraic if it is of the form: \[ S=\bigcup_{i=1}^ p \bigl\{ x \in V \mid g_{i1}(x)>0,\ldots,g_{ir_ i}(x)>0 \bigr\} \] for some polynomial functions \(g_{ij} \in{\mathcal P}(V)\). That is, \(S\) is the union of basic open pieces \(\bigl\{ g_{i1}(x)>0, \ldots,g_{ir_ i}(x)>0 \bigr\}\). Of course a semialgebraic set admits many different representations and one can ask for the minimal number of polynomials needed for that. This yields to the study of the minimal number of basic pieces needed and to study the number of polynomials needed to describe a basic open semialgebraic set. The first number is called the \(t\)- invariant, and the second the \(s\)-invariant of the variety \(V\). Similarly, a closed semialgebraic set is of the form \[ S=\bigcup_{i=1}^ p \bigl\{ x \in V\mid g_{i1}(x) \geq 0,\ldots,g_{ir-i}(x) \geq 0\bigr\} \] and we introduce the corresponding integers \(\bar t\) and \(\bar s\). Roughly speaking, the game is like that: any semialgebraic subset \(S\) of \(V\) corresponds uniquely to a constructible subset \(\tilde S\) of the real spectrum \(\text{Spec}_ r \bigl( {\mathcal P}(V) \bigr)\), so that the questions on \(S\) are translated to questions on \(\tilde S\). Now, \(\text{Spec}_ r(V)\) is the union of the real spectra of the quotient fields of the factor rings \({\mathcal P}(V){\mathfrak p}\) when \({\mathfrak p}\) ranges over all prime ideals of \({\mathcal P}(V)\), which are nothing but the spaces of orderings of such fields. Thus, the original questions are translated into terms of these spaces of orderings, studied there, and finally all the information is pulled back to the original situation. As a result one can prove that if the dimension of \(V\) is \(n\) then \(s(V)=n\), \(\overline s(V)={1 \over 2}n(n+1)\) and \(t(V)\) and \(\bar t(V)\) are bounded from above by a combinatorial function depending on \(n\). The paper is organized as follows: in the first two sections contain a general presentation of semialgebraic sets and the problems under consideration, with some nice examples; then the technical devices of real spectra and spaces of orderings are presented, specially the latter, to which sections 4 and 5 are devoted. Most of this material is presented without proofs, but stating precisely all the needed results and giving precise references of where to find them. From section 6 on the author applies the preceding results to obtain the geometric information for semialgebraic sets. The author studies also some other questions as the relations between basic open and basic closed set when taking closures and interiors, or the separation of semialgebraic sets. As mentioned at the beginning, the paper should be seen as an enlightening discussion of the problems and many proofs are not included. However it is a very nice table of contents for someone willing to introduce him(her)-self to the topic.