scientific article; zbMATH DE number 2124927

From MaRDI portal
Publication:4832250





Let \(X\) be a reduced analytic space (over \(\mathbb{C}\)); if \(X\) is singular, then the complex \(D_X^\bullet\) of differential forms on \(X\) in the sense of Grauert-Grothendieck shares with the de Rham complex (on a manifold) all the usual properties, but does not provide a fine resolution for \(\mathbb{C}_X\). Previous work of the authors [C. R. Acad. Sci., Ser I Math (1) 317, 65--70 (1998), Rev. Roum. Math. Pures Appl. 38, 579--594 (1993; Zbl 0810.32009), Bull. Sci. Math. 122, 163--201 (1998; Zbl 0928.32004), Aspects Math. E 26, 1--26 (1994; Zbl 0827.32005)] give (using the desingularization of \(X\)) a different definition of differential forms on \(X\) which posses all the required properties, with the exception of the existence of the pullbak. Thus, in this paper, which has mostly an expository form and the proofs are mainly sketched, the authors give a construction such that also the existence of pullbaks is obtained. They define for \(X\) a family of complexes \(\mathcal{R}(X) = \{\Lambda_X^\bullet\}\) and for every morphism \(f : X \to Y\) a family \(\mathcal{R}(Y, X)\) of morphisms of complexes between the \(\Lambda_Y^\bullet \in \mathcal{R}(Y)\) and one of the \(\Lambda_X^\bullet \in \mathcal{R}(X)\) (called the (admissible)-pullback), \(\Lambda_Y^\bullet \to f_* \Lambda_X^\bullet\) (denoted \(\Lambda_Y^\bullet - \Lambda_X^\bullet\)) such that (I) \(\Lambda_X^\bullet\) is a fine resolution of \(\mathbb{C}_X\) (II) For \(p > 2\) dim \(X\), \(\Lambda_X^p = 0\) (III) If \(X\) is smooth, the usual de Rham complex \(\mathcal{E}_X^\bullet \in \mathcal{R}(X)\) and for every morphism \(F : X \to Y\) between smooth compelx manifolds, the ordinary pullback \(f^* : \mathcal{E}_Y^\bullet \to \mathcal{E}_X^\bullet\) is an admissible pullback and (IV) There is a smooth, open, dense analytic subset \(U \subset X\) such that \(\Lambda_X^\bullet\mid_U\) is the ordinary de Rham complex \(\mathcal{E}_U^\bullet\). The family of pullbacks satisfy the composition property, the existence and unicity of pullbacks, and a natural filtering property. The definition of \(\Lambda_X^\bullet\) and of the pullbacks is given means of a relatively complicated induction procedure on the dimension of the complex space. More precisely, if dim\( X = 0\), \(\mathcal{R}(X)\) contains only the complex \(\mathbb{C}_X\) with \(\mathbb{C}_X^0 = \mathbb{C}_X\) and \(\mathbb{C}_X^p = 0\) otherwise. If \(\mathcal{R}(Y)\) is supposed to be known for complex spaces \(Y\), dim \(Y < n\), then, for \(X\) such that dim \((X) = n\) an element \(\Lambda_X^{\bullet} \in \mathcal{R}(X)\) means the assignement of: i) a nowhere dense closed subspace \(E \subseteq X\), sing \((X) \subseteq E\) and a proper modifaction \[ \begin{matrix}\widetilde E & \mathop{\to}\limits^i & \widetilde X\\ q\downarrow && \downarrow \pi\\ E & \mathop{\to}\limits^j &X \end{matrix} \] where \(j\) is the natural inclusion, \(\widetilde X\) a smooth manifold, \(\widetilde E = \pi^{-1}(E)\) and \(\pi\) induces an isomorphism, \(\tilde X \backslash \tilde E \simeq X \backslash E\) ii) There is a \(\Lambda_E^\bullet \in \mathcal{R} (E)\), \(\Lambda_{\tilde E}^\bullet \in \mathcal{R}(\tilde E)\) and two pullbacks \(\varphi\), resp. \(\psi\), corresponding to \(q\) respectively iii) \(\Lambda_X^\bullet\) is defined as \(\Lambda_X^\bullet = \pi_X^\bullet = \pi_* \mathcal{E}_{\tilde X}^\bullet \oplus j_* \Lambda_E^\bullet \oplus (j \circ q)_* \Lambda_{\tilde E}^\bullet (-1)\) with differential \(d : \Lambda_X^p \to \Lambda_X^{p + 1} \) given by \(d(\omega, \sigma, \theta) = (d\omega, d\sigma, d \theta + (-1)p (\psi (\omega) - \phi(\sigma)))\). The augmentation \(\mathbb{C}_X \to \Lambda_X^\bullet\) \((c \mapsto (c,c, 0))\) make \(\Lambda_X^\bullet\) a resolution of \(\mathbb{C}_X\). iv) There is a uniquely determined family \((X_p, h_l)_{l \in L}\) of smooth manifolds \(X_l\) and proper maps \(h_l : X_l \to X\) such that \(\Lambda_X^p = \oplus_l (h_l)_X \mathcal{E}_{Xl}^{p - q(l)}\) (\(q(p)\) an nonnegative integer) and mappings \(h_{l_m} : X_l \to X_m\), commuting with \(h_l\), \(h_m\) such that the differential previously above is given by \[ d(\omega_l) = (d \omega_l + \sum_m \in_{l_m}^(p) h^*_{l_m} \omega_m) \] where \(\in_{l_m}^{(p)}\) can take the values \(0, \pm 1\). The definition of the pullback is more complicated and cannot be given here. However, in the case \(X, Y\) irreducible the authors give a complete proof of the existence of what they call a primary pullback (corresponding to a morphism \(f : X \to Y\)) using a double induction procedure. (The existence of the pullback in general is proved by the the authors in [Ann. Sci. Norm. Super Pisa, Cl. Sci. 2, 119--150 (2003)]. This is however sufficient to prove (Th.4.1) the following: Let \(X\) be a compact projective variety; for any \(\Lambda_X^\bullet \in \mathcal{R}(X)\) the weight and the Hodge filtration induce the Hodge-Deligne mixed structure on the cohomology of \(X\). Finally the authors show how to modify the previous construction in order to define complex of forms with logaritmic poles at infinity, which describe the Deligne they of mixed Hodge structure on an open algebraic variety. For the entire collection see [Zbl 1047.00016].











This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4832250)