Relative filtration, the Bernstein Ideal, and its slopes (Q6082478)

Let \(X\) be a complex analytic manifold and let \(f_1,\ldots,f_p:X\to\mathbb{C}\) be some analytic functions defined on certain open subset of \(X\). We will denote by \(\mathcal{D}_X\) the sheaf of rings of differential operators on \(X\) and let \(\mathcal{D}_X[s_1,\ldots,s_p]=\mathbb{C}_X[s_1,\ldots,s_p]\otimes_{\mathbb{C}}\mathcal{D}_X\). We will write \(F=f_1\cdots f_p\). Consider now a section \(m\) of a holonomic \(\mathcal{D}_X\)-module \(\mathcal{M}\). The main object of study of the paper under review is the Bernstein ideal \(\mathcal{B}(m,x_0,f_1,\ldots,f_p)\), consisting of polynomials \(b(s_1,\ldots,s_p)\in\mathbb{C}[s_1,\ldots,s_p]\) such that \[ b(s_1,\ldots,s_p)m\cdot f_1^{s_1}\cdots f_p^{s_p}\in\mathcal{D}_X m[s_1,\ldots,s_p]F\cdot f_1^{s_1}\cdots f_p^{s_p} \] in a neighborhood of the point \(x_0\in X\); the action of \(\mathcal{D}_X\) on the symbol \(f_1^{s_1}\cdots f_p^{s_p}\) is the usual one. Such an ideal is the generalization to several analytic functions of the (more) classical Bernstein-Sato polynomial associated to a single function \(f\). \textit{C. Sabbah} proved in [Compos. Math. 62, 283--328 (1987; Zbl 0622.32012)] that there is a finite set \(\mathcal{H}\) of linear forms \(H(s_1,\ldots,s_p)=\alpha_1s_1+\ldots\alpha_ps_p+\beta\) with rational coefficients such that the \(\alpha_i\) are natural pairwise coprime and \[ \prod_{H\in\mathcal{H}}H(s_1,\ldots,s_p)\in\mathcal{B}(m,x_0,f_1,\ldots,f_p). \] The author of the paper under review provides a minimal set \(\mathcal{H}\) by means of the study of two characteristic varieties of the \(\mathcal{D}_X[s_1,\ldots,s_p]\)-module \[ M:=\frac{\mathcal{D}_X m[s_1,\ldots,s_p]\cdot f_1^{s_1}\cdots f_p^{s_p}}{\mathcal{D}_X m[s_1,\ldots,s_p]F\cdot f_1^{s_1}\cdots f_p^{s_p}}. \] They are defined using the so-called sharp and relative filtrations, which extend the order filtration on \(\mathcal{D}_X\) by assigning, respectively, weight one or zero to the \(s_i\). Namely, using a very detailed study of the grade number of a module, the author shows that the characteristic variety associated to the relative filtration of \(M\) can be decomposed as a finite union of sets of the form \(T_{X_\alpha}^*X\times S_\alpha\), where the \(X_\alpha\) are analytic subspaces of \(X\) and the \(S_\alpha\) are algebraic varieties of dimension \(p-1\). What he proves about the irreducible components of the \(S_\alpha\) is more interesting. Those of maximal dimension are affine hyperplanes given by precisely the linear forms of the minimal set \(\mathcal{H}\) from above, whereas those of dimension less than \(p-1\) can be sent inside of the previous ones by a translation of the form \((s_1,\ldots,s_p)\mapsto(s_1+k,\ldots,s_p+k)\) for some integer \(k\). There are more results in this interesting paper that would deserve a comment, specially those about the case in which \(m\) is a section of a \textit{regular} holonomic module or about the so-called slopes of \((m,f_1,\ldots,f_p)\), but we refer to the article to those interested on them.