Some applications of microlocalization for local complete intersection subvarieties (Q6943733)

Hypersurface singularities are intimately related to Bernstein-Sato polynomials \(b_f(s)\), which in turn are related to the \(V\)-filtration of Kashiwara and Malgrange on the \(\mathcal{D}\)-module \(\mathcal{B}_f\) by Sabbah's work. It turns out that \((s+1)\) is always a root of \(b_f(s)\), and the \textit{reduced} Bernstein-Sato polynomial associated to \(f\) is defined by \N\[\N\tilde{b}_f(s):=b_f(s)/(s+1).\N\]\NIn [\textit{M. Saito}, Bull. Soc. Math. Fr. 122, No. 2, 163--184 (1994; Zbl 0810.32004)], Saito defines \textit{microlocal \(V\)-filtration} and relates it to the reduced Bernstein-Sato polynomial. He also defines the \textit{minimal exponent} \N\[\N\tilde{\alpha}(f):=\min\left\{\gamma\mid \tilde{b}_f(-\gamma)=0\right\}\N\]\Nwhich has since found many applications in the study of singularities.\N\NFor subvarieties defined by \(f_1,\cdots,f_r\), there is an analogous notion of Bernstein-Sato polynomial [\textit{N. Budur} et al., Compos. Math. 142, No. 3, 779--797 (2006; Zbl 1112.32014)]. There is also the \(\mathcal{D}\)-module \(\mathcal{B}_f\) admitting \(V\)-filtration, which, as in the hypersurface case, relates to the Bernstein-Sato polynomial. Moreover, in the local complete intersection (LCI) case, \((s+r)\) is a root of the Bernstein-Sato polynomial, whence one can similarly define its \textit{reduced} version. However, in the \(r>1\) case, it is unclear how to define the \textit{microlocal} variant of the \(V\)-filtration. The article under review presents a microlocalization construction for LCI subvarieties, and explains why it is reasonable to expect that this is the ``correct'' notion when \(r>1\).\N\NThe author's construction is through the microlocalization functor \(\mu(-):=\mathrm{FL}\circ\mathrm{Sp}(-)\) following [\textit{M. Kashiwara} and \textit{P. Schapira}, Sheaves on manifolds. With a short history ``Les débuts de la théorie des faisceaux'' by Christian Houzel. Berlin etc.: Springer-Verlag (1990; Zbl 0709.18001)] that sends a mixed Hodge module on \(X\times \mathbf{A}^r\) to a monodromic mixed Hodge module. Applying this functor on \(B_f^H\) (Hodge module associated to \(\mathcal{B}_f\)) and applying the inverse Fourier transform, the author obtains a map \N\[\N\Psi:L\to \mathrm{Sp}(B_f^H)\] where \(L\) is isomorphic to \(i_*\mathbf{Q}_{Z\times\mathbf{A}^r}^H[\dim X]\). Let \(Q\) be the cokernel of the above, and \((\mathcal{Q},F)\) be its underlying \(\mathcal{D}\)-module. The author explains why \(\mu(\mathcal{B}_f)\) and \(\mathcal{Q}\) are convincing generalizations of Saito's microlocalization functor by proving results about the singularities of \(Z\) through this construction.\N\NTo begin with, Lemma 3.2 relates \(\mathcal{Q}\) to Saito's microlocalization construction in the special case when \(r=1\). Next, Proposition 3.4 in the article shows an isomorphism of \(\mu(\mathcal{B}_f)\) with the vanishing cycles along the hypersurface \(g:=\sum_{i=1}^r y_if_i\). This hypersurface was used in [\textit{Q. Chen} et al., J. Reine Angew. Math. 811, 219--256 (2024; Zbl 1546.14039)] to define the minimal exponent \(\tilde{\alpha}(Z)\) of \(Z\). Moreover, The author uses the morphism \(\Psi\) also to study the spectrum \(\mathrm{Sp}(Z,x)\) defined as in [\textit{A. Dimca} et al., Ann. Inst. Fourier 61, No. 4, 1633--1653 (2011; Zbl 1241.32025)], which is a generalization of the definition of Steenbrink for isolated hypersurface singularities. In Theorem 1.1, the author proves the following result: Let \(Z\subseteq X\) be an LCI of pure codimension \(r\), and \(x\in Z\). If \(\tilde{\alpha}_x(Z)>r-1\), then \(\tilde{\alpha}_x(Z)-r+1\leq\mathrm{Sp}_{\min}(Z,x)\), with equality if \(Z_{\mathrm{sing}}=\left\{x\right\}\). This result generalizes the well-known result that in the isolated hypersurface singularity case, minimal exponent agrees with the minimal spectral number.\N\NThe article also considers deformation aspects of minimal exponent. It follows from the works of Varchenko and Steenbrink that a \(\mu\)-constant family of isolated hypersurface singularities has constant spectrum where \(\mu\) is the Milnor number. The author proves in Corollary 1.4 an analog of this statement where it is shown that under certain conditions on a \textit{reduced family of isolated complete intersection singularities} \(X\subseteq Y\times T\) (we refer to the article for details of this notion), \(\tilde{\alpha}(X_t)\) is independent of \(t\).\N\NFinally, recall that for an LCI subvariety \(Z\) defined by \(f_1,\cdots,f_r\), the reduced Bernstein-Sato polynomial is defined as \N\[\N\tilde{b}_f(s)=b_f(s)/(s+r).\N\]\NIt was asked in [\textit{Q. Chen} et al., J. Reine Angew. Math. 811, 219--256 (2024; Zbl 1546.14039)] if \N\[\N\tilde{\gamma}(Z):=\min\left\{\gamma\mid \tilde{b}_f(-\gamma)=0\right\}\N\]\Ncoincides with \(\tilde{\alpha}(Z)\) (as defined in [loc. cit.]). The author answers the question in the affirmative, which in turn implies Conjecture 9.11 of [\textit{M. Mustaţă} and \textit{M. Popa}, Forum Math. Pi 10, Paper No. e22, 58 p. (2022; Zbl 1511.14010)].\N\NThe author ends by discussing the constructibility of the roots of the local Bernstein-Sato polynomials in Theorem 1.7, which is analogous to Remark 2.11 of [\textit{M. Saito}, Bull. Soc. Math. Fr. 122, No. 2, 163--184 (1994; Zbl 0810.32004)].