arXiv subjects
Publications and source records attributed to Patrick Graf.
We prove that a strongly $F$-regular scheme $X$ admits a finite, generically Galois, and étale-in-codimension-one cover $\widetilde X \to X$ such that the étale fundamental groups of $\widetilde X$ and $\widetilde X_{reg}$ agree. Equivalently, every finite étale cover of $\widetilde X_{reg}$ extends to a finite étale cover of $\widetilde X$. This is analogous to a result for complex klt varieties by Greb, Kebekus and Peternell.
Let $(X, H)$ be a normal complex projective polarized variety and $\mathscr E$ an $H$-semistable sheaf on $X$. We prove that the restriction $\mathscr E\big|_C$ to a sufficiently positive general complete intersection curve $C \subset X$ passing through a prescribed finite set of points $S \subset X$ remains semistable, provided that at each $p \in S$, the variety $X$ is smooth and the factors of a Jordan-Hölder filtration of $\mathscr E$ are locally free. As an application, we obtain a generalization of Miyaoka's generic semipositivity theorem.
Let $\mathfrak a \subset \mathscr O_X$ be a coherent ideal sheaf on a normal complex variety $X$, and let $c \ge 0$ be a real number. De Fernex and Hacon associated a multiplier ideal sheaf to the pair $(X, \mathfrak a^c)$ which coincides with the usual notion whenever the canonical divisor $K_X$ is $\mathbb Q$-Cartier. We investigate the properties of the jumping numbers associated to these multiplier ideals. We show that the set of jumping numbers of a pair is unbounded, countable and satisfies a certain periodicity property. We then prove that the jumping numbers form a discrete set of real numbers if the locus where $K_X$ fails to be $\mathbb Q$-Cartier is zero-dimensional. It follows that discreteness holds whenever $X$ is a threefold with rational singularities. Furthermore, we show that the jumping numbers are rational and discrete if one removes from $X$ a closed subset $W \subset X$ of codimension at least three, which does not depend on $\mathfrak a$. We also obtain that outside of $W$, the multiplier ideal reduces to the test ideal modulo sufficiently large primes $p \gg 0$.
We investigate properties of potentially Du Bois singularities, that is, those that occur on the underlying space of a Du Bois pair. We show that a normal variety $X$ with potentially Du Bois singularities and Cartier canonical divisor $K_X$ is necessarily log canonical, and hence Du Bois. As an immediate corollary, we obtain the Lipman-Zariski conjecture for varieties with potentially Du Bois singularities. We also show that for a normal surface singularity, the notions of Du Bois and potentially Du Bois singularities coincide. In contrast, we give an example showing that in dimension at least three, a normal potentially Du Bois singularity $x \in X$ need not be Du Bois even if one assumes the canonical divisor $K_X$ to be $\mathbb{Q}$-Cartier.
We propose and study a generalized version of the Lipman-Zariski conjecture: let $(x \in X)$ be an $n$-dimensional singularity such that for some integer $1 \le p \le n - 1$, the sheaf $Ω_X^{[p]}$ of reflexive differential $p$-forms is free. Does this imply that $(x \in X)$ is smooth? We give an example showing that the answer is no even for $p = 2$ and $X$ a terminal threefold. However, we prove that if $p = n - 1$, then there are only finitely many log canonical counterexamples in each dimension, and all of these are isolated and terminal. As an application, we show that if $X$ is a projective klt variety of dimension $n$ such that the sheaf of $(n-1)$-forms on its smooth locus is flat, then $X$ is a quotient of an Abelian variety. On the other hand, if $(x \in X)$ is a hypersurface singularity with singular locus of codimension at least three, we give an affirmative answer to the above question for any $1 \le p \le n - 1$. The proof of this fact relies on a description of the torsion and cotorsion of the sheaves $Ω_X^p$ of Kähler differentials on a hypersurface in terms of a Koszul complex. As a corollary, we obtain that for a normal hypersurface singularity, the torsion in degree $p$ is isomorphic to the cotorsion in degree $p - 1$ via the residue map.
Let $X$ be a normal variety. Assume that for some reduced divisor $D \subset X$, logarithmic 1-forms defined on the snc locus of $(X, D)$ extend to a log resolution $\tilde X \to X$ as logarithmic differential forms. We prove that then the Lipman-Zariski conjecture holds for $X$. This result applies in particular if $X$ has log canonical singularities. Furthermore, we give an example of a 2-form defined on the smooth locus of a three-dimensional log canonical pair $(X, \emptyset)$ which acquires a logarithmic pole along an exceptional divisor of discrepancy zero, thereby improving on a similar example of Greb, Kebekus, Kovács and Peternell.
Let (X, D) be a projective log canonical pair. We show that for any natural number p, the sheaf (Omega_X^p(log D))^** of reflexive logarithmic p-forms does not contain a Weil divisorial subsheaf whose Kodaira-Iitaka dimension exceeds p. This generalizes a classical theorem of Bogomolov and Sommese. In fact, we prove a more general version of this result which also deals with the orbifoldes géométriques introduced by Campana. The main ingredients to the proof are the extension theorem of Greb-Kebekus-Kovács-Peternell, a new version of the Negativity lemma, the Minimal Model Program, and a residue map for symmetric differentials on dlt pairs. We also give an example showing that the statement cannot be generalized to spaces with Du Bois singularities. As an application, we give a Kodaira-Akizuki-Nakano-type vanishing result for log canonical pairs which holds for reflexive as well as for Kähler differentials.