On the extension of Nash functions (Q913038)

Let S be a real (resp. complex) Nash subvariety of \({\mathbb{R}}^ n\) (resp. \({\mathbb{C}}^ n)\). In this paper the authors deal with the problem of finding open sets U in \({\mathbb{R}}^ n\) (resp. \({\mathbb{C}}^ n)\) such that the restrictions to \(S\cap U\) of Nash functions on S extend to U. If U is semialgebraic and S is a real smooth subvariety that is the zero locus of a Nash function on U, a result of G. Efroymson [cf. \textit{J. Bochnak, M. Coste}, and \textit{M. F. Roy}, Géométrie algébrique réelle (1987; Zbl 0633.14016)] says that every Nash function on S extends to U. This result is founded both upon the existence of tubular neighborhoods for Nash submanifolds and Efroymson's approximation theorem. In the complex case there is not such a global approximation theorem, and the results that are given in the present paper are more local, in a neighborhood of a compact set, since they depend on Runge's approximation theorem, also in its relative version established by the authors [Boll. Unione Mat. Ital., VII. Ser., A 3, No.3, 343-350 (1989; Zbl 0691.32006)]. A local tubular neighborhood theorem for complex Nash submanifolds is stated and so the existence of a Nash retraction \(U\to U\cap S\) solves trivially, the problem of extending every Nash function on \(U\cap S\) to U. When \(S\cap U\) is singular such a retraction cannot exist. However if S is normal an extension theorem with some control on the extended function is given, even if U depends on the function to be extended. This is quite useful since, as is well-known, results like Cartan's theorems A and B do not hold in the Nash setting. From the complex case the authors deduce some new results in the real one. In particular it is proved that a compact, normal and coherent Nash variety in \({\mathbb{R}}^ n\) has global Nash equations in a semialgebraic neighborhood, thus extending a result of \textit{M. Shiota} [`Nash manifolds' (1987; Zbl 0629.58002)] stated in the smooth case. For the above varieties it is also possible to improve Efroymson's result on the extension of real Nash functions.