No label defined (Q4944923)

Let \(R\) be a Noetherian local ring, \(M(R)\) denote its maximal ideal. Artin Approximation (hereafter AA) holds in \(R\) if, for any polynomial system \(S\) with coefficients in \(R\), a solution of \(S\) in the \(M(R)\)-adic completion \(\widehat{R}\) of \(R\) can be approximated in \(R\) modulo every power of \(M(R)\). It can be observed that AA holds in \(R\) exactly when \(R\) is existentially closed in \(\widehat{R}\). This remark, and other ones, connect AA with model theory. The paper under review aims to provide a suitable model-theoretic setting where to study AA in a proper way. A more precise tentative program, as stated in the beginning of the paper, is the following: find a first-order theory \(T\) such that, for every Noetherian local ring \(R\), (1) if \(R\) is a model of \(T\), then so is \(\widehat{R}\); (2) if \(R\) admits AA, then \(R\) is an existentially closed model of \(T\). However, it is pointed out that the model-theoretic framework imposes several restrictions to this program, in particular some bounds on both the dimension and the multiplicity of the local ring \(R\). Then the author shows the existence of a first-order theory \(Cm_{d, e}\) whose Noetherian models are exactly the local Cohen-Macaulay rings of dimension \(d\) and multiplicity \(e\). \(Cm_{d, e}\) is explicitly axiomatized. Moreover (1) is proved: the class of models of \(Cm_{d, e}\) is closed under completion; furthermore the completion of a model of \(Cm_{d, e}\) is still Noetherian. With respect to (2), it is shown that, if \(R\) is an equicharacteristic local Gorenstein ring of dimension \(d\) and multiplicity \(e\), \(R\) satisfies AA and the residue field of \(R\) is algebraically closed, then \(R\) is existentially closed among the Noetherian models of \(Cm_{d, e}\). It is also proved that, for equicharacteristic Noetherian local rings \(R\) having an algebraically closed residue field, AA is just equivalent to the solvability property (saying that a polynomial system having a solution modulo every power of \(M(R)\) has a solution in \(R\)). The paper is well organized and provides several historical hints and comments on AA, its links with model theory, and the possible developments of the underlying program. For the entire collection see [Zbl 0932.00038].