Stability of quadratic modules (Q834811)

One of the main questions in real algebraic geometry is the following. Let \(A=\mathbb{R}[X_1,\dots,X_n]\) be the polynomial ring in \(n\) variables and let \(f_1,\dots, f_s\in A\). Consider the basic closed semi-algebraic set \(\mathcal{S}:=\mathcal{S}(f_1,\dots,f_s):=\{x\in\mathbb{R}^n \mid f_1(x)\geq 0,\dots,f_s(x)\geq 0\}\). What can be said about the polynomials that are nonnegative on \(\mathcal{S}\)? Clearly, the elements of the corresponding quadratic module \(QM(f_1,\dots,f_s):=\{\sigma_0+\sigma_1f_1+\dots+\sigma_s f_s\mid \sigma_i\in\sum A^2\}\), finitely generated by \(f_1,\dots,f_s\) (where \(\sum A^2\) denotes the sum of squares in \(A\)), are nonnegative on \(\mathcal{S}\). The interesting question is whether the converse holds. One of the key notions in this context is stability of a quadratic module, introduced by \textit{V. Powers} and \textit{C. Scheiderer} [Adv. Geom. 1, 71--88 (2001; Zbl 0984.44012)]. It can be formulated in the following way. The quadratic module \(M:=QM(f_1,\dots,f_s)\) is called stable, if for every finite dimensional subspace \(U\) of \(A\) there is another finite dimensional subspace \(W\) of \(A\) such that for every \(f\in M\cap U\) there are \(\sigma_i\in\sum W^2\) such that \(f=\sigma_0+\sigma_1f_1+\dots+\sigma_sf_s\). One of the main consequences of stability is that if \(\dim(S)\geq 2\) then \(M\) does not have the so-called Strong Moment property, i.e. there is a linear functional on \(A\) that is nonnegative on \(M\) and does not come from a measure on \(\mathcal{S}\). In particular, \(M\) does not contain all polynomials that are nonnegative on \(\mathcal{S}\) [see \textit{C. Scheiderer}, J. Complex. 21, 823--844 (2005; Zbl 1093.13024)]. The author generalizes the notion of stability in the paper under review and obtains new geometric criterions for the stability of finitely generated quadratic modules. The setting is the following. Let \(\{U_\gamma\}_{\gamma\in\Gamma}\) be a filtration of \(A\) where \(\Gamma\) is an ordered Abelian group. The author calls the quadratic module \(M=QM(f_1,\dots,f_s)\) stable with respect to the filtration, if there is a monotonically increasing map \(\rho:M\to M\) such that for every \(\gamma\in\Gamma\) and every \(f\in U_\gamma\) there are \(\sigma_i\in\sum U_{\rho(\gamma)}^2\) such that \(f=\sigma_0+\sigma_1f_1+\dots+\sigma_sf_s\). Note that this definition does not depend on the finitely many generators \(f_1,\dots,f_s\) of \(M\). Stability of \(M\) with respect to a grading \(A=\bigoplus_{\gamma\in\Gamma}A_\gamma\) of \(A\) is then defined as the stability with respect to the filtration induced by the grading. For example, in the case of the usual degree grading of \(A\) one gets the classical stability. The author introduces another notion. He calls the quadratic module \(M=QM(f_1,\dots,f_s)\) totally stable with respect to a grading of \(A\) if \(\text{deg}(f)\leq \text{deg}(f+g)\) for all \(f,g\in M\) (where \(\text{deg}\) denotes the degree with respect to the grading). Note that total stability implies stability with respect to the grading. The author shows the following. Let \(A=\bigoplus_{\gamma\in\Gamma}A_\gamma\) be a grading and let \(M=QM(f_1,\dots,f_s)\) be a finitely generated quadratic module. If \(\mathcal{S}(f_1^{\max},\dots,f_s^{\max} )\subset\mathbb{R}^n\) is Zariski dense, then \(M\) is totally stable with respect to the grading. Here \(f^{\max}\) denotes the highest degree part of \(f\). He uses this to give a geometric criterion for totally stability via so-called tentacles. Applying these results and covering the usual degree grading by finitely many gradings in a suitable way he obtains new geometric criterions for \(\mathcal{S}=\mathcal{S}(f_1,\dots,f_s)\) such that \(M=QM(f_1,\dots,f_s)\) is stable. With this in hand he gets nice examples in the plane. The paper is written in a very clear way.