Reflexive differential forms on singular spaces. Geometry and cohomology (Q479913)

In an important paper, \textit{D. Greb} et al. [Publ. Math., Inst. Hautes Étud. Sci. 114, 87--169 (2011; Zbl 1258.14021)] established an extension theorem for reflexive differential forms on quasi-projective varieties with at most Kawamata log terminal (klt) singularities: any regular holomorphic differential form defined on the smooth part of such a variety defines a regular holomorphic differential form on any log resolution. Actually, [Zbl 1258.14021] contains much more general extension results for logarithmic reflexive differential forms on log canonical spaces. In the present paper, the authors generalize these results to reflexive differential forms on locally algebraic varieties and study a number of problems about existence of differential forms, vanishing theorems, and Hodge-theoretic properties of reflexive differentials. The generalized extension theorem, appearing as Theorem 1.2 in the paper, reads as: { Theorem.} Let \(X\) be a normal complex space and \(D\) be an effective \(\mathbb{Q}\)-divisor such that the pair \((X,D)\) is analytically klt. Suppose \((X,D)\) is locally algebraic. Let \(\pi: \tilde{X}\rightarrow X\) be a log resolution of the pair \((X,D)\). Then, the sheaves \(\pi_*( \Omega^p_{\tilde{X}})\) are reflexive for \(1\leq p\leq n\). Concerning the existence of holomorphic differential forms, the following question is discussed. It is a well-known fact that a rationally connected projective manifold \(M\) does not carry any holomorphic pluri-form: \[ H^0\big(M,(\Omega^1_M)^{\otimes m}\big)=0 \;\;\forall m\geq 1. \] Exploring the singular setup, the authors derive an analogous vanishing result for reflexive pluri-forms on factorial klt spaces (Theorem 1.3): Theorem. Let \(X\) be a rationally chain-connected factorial klt space. Then \[ H^0\big( X, ((\Omega^1_X)^{\otimes m})^{**}\big)=0 \;\;\forall m\geq 1. \] Note that factorial klt spaces automatically have canonical singularities. Moreover, they are rationally connected by a result of Hacon and McKernan, cf. Remark 3.2 of the paper under review. It is interesting that the above theorem fails in the non-factorial canonical case. A corresponding two-dimensional example is provided in Section 3.2 of the paper. The proof of Theorem 1.3 relies on a semistability analysis of sheaves of reflexive differentials with respect to movable curves on klt spaces. The necessary foundational material concerning this semistability notion is developed in an appendix of the paper. Concerning vanishing theorems, the authors prove versions of the Kodaira-Akizuki-Nakano vanishing on klt spaces for reflexive differential forms \(\Omega_X^{[p]}\) in certain extremal ranges, but they also provide an example showing that the corresponding vanishing theorem fails in general. In Section 4.2 they study a projective fourfold \(X\) with a single isolated terminal Gorenstein singularity carrying an ample line bundle \(\mathcal{L}\) such that \[ H^2\big(X,\Omega_X^{[3]}\otimes \mathcal{L}\big) \cong H^2\big( X, \Omega_X^{[1]}\otimes \mathcal{L}^{-1}\big) \cong \mathbb{C}. \] Auxiliary results on Hodge theory are on closedness of reflexive logarithmic \(p\)-forms on log canonical spaces (Proposition 1.4) and a Poincaré-Lemma for reflexive \(1\)-forms on klt spaces. The latter result is then used to show that every globally defined closed reflexive \(1\)-form is represented in a canonical manner by a Kähler differential.