--- title: Steenbrink Vanishing Overview url: https://www.emergentmind.com/topics/steenbrink-vanishing type: topic --- # Steenbrink Vanishing Overview Steenbrink vanishing is a vanishing theorem for higher direct images of logarithmic differential forms on a resolution of singularities. In its classical form, if \(X\) is a complex variety, \(E\subset X\) is such that \(X\setminus E\) is smooth, and \(T:Y\to X\) is a proper birational morphism with \(Y\) smooth, \(E_Y:=T^{-1}E\) a simple normal crossings divisor, and \(Y\setminus E_Y\cong X\setminus E\), then \[ R^qT_*\Omega_Y^p(\log E_Y)(-E_Y)=0 \qquad \text{for } p+q>\dim X. \] In characteristic \(0\), Steenbrink proved this using mixed Hodge theory. Subsequent work reformulated the theorem in the language of the Deligne–Du Bois complex, sharpened the low-degree cases by introducing the DB index, and developed positive-characteristic and symplectic analogues in which the same vanishing pattern survives under additional hypotheses [1307.7503, 2507.04838, 2410.07515]. ## 1. Classical statement and geometric role The classical theorem concerns a smooth birational model \(Y\) of a singular space \(X\), together with a boundary divisor encoding the exceptional locus and the singular boundary data. The sheaf \(\Omega_Y^p(\log E_Y)(-E_Y)\) consists of logarithmic \(p\)-forms with poles along \(E_Y\), twisted by \(\mathcal O_Y(-E_Y)\), and the theorem asserts that its higher direct images vanish outside the range allowed by the dimension of \(X\) [1307.7503]. A standard reformulation, used repeatedly in later work, is that Steenbrink vanishing controls the local cohomological complexity of singularities by forcing the logarithmic Hodge pieces on a resolution to disappear once the total degree exceeds \(\dim X\). In the formulation recorded by Kovács, Schwede, and Smith, the original theorem is “vacuous” for \(p\le 1\) in the usual smooth/snc setting because \(R^qT_*=0\) for \(q\ge \dim X\); this observation partly explains why later extensions concentrated on the missing low-\(p\) range [1307.7503]. In recent positive-characteristic literature, the same formula is often taken as the definition of “Steenbrink vanishing” for a normal variety \(X\) and a log resolution \(\pi:Y\to X\) with reduced exceptional divisor \(E\): \[ R^j\pi_*\Omega_Y^i(\log E)(-E)=0 \qquad \text{for all } i,j \text{ with } i+j>\dim X. \] That convention makes the characteristic-zero theorem the reference model against which partial positive-characteristic analogues are measured [2507.04838]. ## 2. Deligne–Du Bois reformulation and the DB index A decisive extension replaces logarithmic differential forms on a smooth pair by graded pieces of the Deligne–Du Bois complex of a reduced pair. If \(T:Y\to X\) is a projective morphism of reduced finite type \(\mathbb C\)-schemes, \(E\subset X\) is reduced, \(T\) is an isomorphism over \(X\setminus E\), and \(F=(T^{-1}E)_{\mathrm{red}}\), then \[ \left(R^qT_*\underline{\Omega}^p_{Y,F}\right)_x=0 \qquad \text{for } p+q>\dim_x X, \] and hence globally \(R^qT_*\underline{\Omega}^p_{Y,F}=0\) for \(p+q>\dim X\). For an snc pair \((Y,F)\), one has \(\underline{\Omega}^p_{Y,F}\simeq \Omega_Y^p(\log F)(-F)\), so this recovers the classical theorem in the logarithmic setting [1307.7503]. The same paper introduces the DB index as a numerical measure of how far a reduced pair \((X,E)\) is from being Du Bois. The local DB index is \[ \operatorname{db}_x(X,E) = \min\left\{ q\in\mathbb N \,\middle|\, h^i(\underline{\Omega}_{X,E}^0)_x=0 \text{ for all } i>q \right\}, \] and the global DB index is \(\operatorname{db}(X,E)=\max_{x\in X}\operatorname{db}_x(X,E)\). If \((X,E)\) is a Du Bois pair, then \(\operatorname{db}(X,E)=0\). The paper also proves the bound \[ \operatorname{db}_x(X,E)\le \dim_x X-1 \] when \(E\) contains no irreducible component of \(X\) [1307.7503]. The DB index yields sharper low-degree vanishing. If \(T:Y\to X\) is projective birational, \(E\subset X\) is reduced, \(E_Y=\operatorname{Exc}(T)\), \(F=(T^{-1}E)_{\mathrm{red}}\), and \(Z=T(E_Y)\setminus E\), then \[ \left(R^qT_*\underline{\Omega}^0_{Y,F}\right)_x=0 \qquad \text{for } q>\max\{\operatorname{db}_x(X,E),\ \operatorname{db}_x(Z,Z\cap E)+1\}. \] Under the additional hypotheses \(\operatorname{db}_x(X,E)\le \dim_xX-2\) and \(\operatorname{codim}_x Z\ge 2\), one gets the low-degree extension \[ R^{\dim X-1}T_*\bigl(j_!\mathbb C_{Y\setminus F}\bigr)=0, \] and for a normal \(X\) with resolution \(T:Y\to X\) and \(F\) snc this yields \[ R^{\dim X-1}T_*\Omega_Y^p(\log F)(-F)=0 \qquad \text{for all } p. \] This is the form in which the missing \(p=1\) range is recovered [1307.7503]. ## 3. Positive-characteristic surface and threefold results For normal surfaces over a perfect field of characteristic \(p>0\), a positive-characteristic analogue is available under singularity hypotheses. If \((X,B)\) is a pair consisting of a normal surface and an effective \(\mathbb Q\)-divisor, and either \((X,B)\) is log canonical with \(p>5\) or \((X,B)\) is \(F\)-pure, then for a log resolution \(f:Y\to X\) with reduced exceptional divisor \(E\) and \[ B_Y\coloneqq f^{-1}_{*}\lfloor B\rfloor+E, \] one has \[ R^{1}f_{*}\Omega^i_Y(\log B_Y)(-B_Y)=0 \qquad \text{for all } i\ge 0. \] The essential case is \(i=1\); the paper notes that \(i=0\) was already known for lc surface pairs in all characteristics, while \(i=2\) is Grauert–Riemenschneider vanishing. It also proves that these vanishing statements do not depend on the choice of log resolution [2402.08153]. For threefold pairs in characteristic \(p>5\), a full Steenbrink-type theorem is known for sharply \(F\)-pure pairs. If \((X,B)\) is a three-dimensional sharply \(F\)-pure pair over a perfect field, \(K_X+B\) is \(\mathbb Q\)-Cartier, \(\pi:Y\to X\) is a log resolution whose reduced exceptional divisor \(E\) supports a \(\pi\)-ample divisor, and \[ B_Y\coloneqq \pi^{-1}_*\lfloor B\rfloor + E, \] then \[ R^j\pi_{*}\Omega^i_Y(\log B_Y)(-B_Y)=0 \qquad \text{for } i+j>3. \] The proof factors through Grauert–Riemenschneider vanishing for \(F\)-pure threefolds, \[ R^j\pi_*\omega_Y=0 \qquad (j>0), \] followed by a Cartier-operator argument. The same theorem yields a logarithmic extension theorem for one-forms [2604.14760]. A different positive-characteristic direction isolates the borderline \((d-1,d-1)\)-case for rational singularities. If \(X\) is a normal variety over a perfect field of positive characteristic with \(\dim X=d\ge 3\), \(X\) has rational singularities, and when \(d=3\) is additionally \(F\)-injective, then for any log resolution \(\pi:Y\to X\) whose reduced exceptional divisor \(E\) supports a \(\pi\)-ample divisor, \[ R^{d-1}\pi_*\Omega_Y^{d-1}(\log E)(-E)=0. \] In dimension \(3\), this feeds into full Steenbrink vanishing for strongly \(F\)-regular threefolds and for \(\mathbb Q\)-factorial klt threefolds in characteristic \(p>41\). The paper identifies the essential new case as \[ R^2\pi_*\Omega_Y^2(\log E)(-E)=0. \] Its proof combines a Frobenius–Cartier nilpotence statement with injectivity of inverse Cartier maps derived from rationality and quasi-\(F\)-injectivity [2507.04838]. ## 4. Symplectic refinements For singular symplectic varieties, Steenbrink vanishing interacts with the symplectic form and acquires additional symmetry. If \(X\) is a symplectic variety of dimension \(2n\), the symplectic form on the smooth locus induces an isomorphism \[ \Omega_U^{n-p}\xrightarrow{\sim}\Omega_U^{n+p}, \] and the problem becomes to lift this symmetry to the Du Bois complex. The resulting morphism is \[ L_\sigma^p:\mathbb D_X(\underline\Omega_X^{n+p})\to \underline\Omega_X^{n+p}, \] constructed by combining Grothendieck duality, the Du Bois complex, and wedging with the symplectic form [2410.07515]. The main derived symmetry theorem states that for a symplectic variety of dimension \(2n\), \[ L_\sigma^{n-1}:\mathbb D_X(\underline \Omega_X^{2n-1}) \to \underline \Omega_X^{2n-1} \] is a quasi-isomorphism. In the isolated-singularity case this yields a local vanishing theorem \[ R^j\pi_*\Omega_{\widetilde X}^{n-p}(\log E)=0 \qquad \text{for } n-p0. \] The proof passes to a log resolution and uses the Ambro–Fujino vanishing theorem. The paper explicitly presents this as a Steenbrink-style vanishing statement for log canonical pairs [1311.5545]. Another analogue is a higher-codimensional adjoint vanishing theorem on smooth projective varieties. If \(X\) is smooth projective, \(Y\subset X\) is a smooth subvariety of codimension \(e\), \(Y\) is defined scheme-theoretically by nef divisors \(D_1,\dots,D_r\), and \(L\) is a line bundle such that for every relevant \(e\)-tuple \(\{s_1,\dots,s_e\}\), \[ L\otimes \mathscr O_X\!\bigl(-(m+1)D_{s_1}-D_{s_2}-\cdots-D_{s_e}\bigr) \] is big and nef, then \[ H^i\!\left(X,\ \mathscr J_Y^{\,m+1}\otimes K_X\otimes L\right)=0 \qquad \text{for all } i>0. \] The paper describes this as a higher-codimensional analogue of Kawamata–Viehweg vanishing with a “Steenbrink-like” feature, because the ideal sheaf contribution is absorbed into an adjoint-type vanishing statement [1208.0484]. On toric varieties, the terminology appears in generalized Bott–Danilov–Steenbrink form. If \(X\) is a normal projective toric variety over a perfect field, \(j:U\hookrightarrow X\) is the inclusion of the smooth locus, and \(\mathscr E_1,\dots,\mathscr E_m\) are ample vector bundles, then under the characteristic-\(0\) hypothesis or the stated positive-characteristic rank and lifting hypotheses, \[ H^s\!\left( X,\, j_*\Omega_U^q\otimes \operatorname{Sym}^{a_1}\mathscr E_1\otimes\cdots\otimes \operatorname{Sym}^{a_m}\mathscr E_m\otimes \bigwedge^{b_1}\mathscr E_1\otimes\cdots\otimes \bigwedge^{b_m}\mathscr E_m \right)=0 \] for \[ s>\sum_{i=1}^m (\operatorname{rk}(\mathscr E_i)-b_i). \] Here the classical line-bundle vanishing is generalized to ample vector bundles and reflexive differentials on singular toric varieties [1702.03962]. ## 6. Techniques, boundaries, and adjacent notions The mechanisms behind Steenbrink vanishing vary sharply with context. In characteristic \(0\), the classical proof uses mixed Hodge theory [2507.04838]. The Du Bois reformulation proceeds by distinguished triangles for pairs and by the general vanishing \[ h^q(\underline{\Omega}^p_X)=0 \qquad \text{for } p+q>\dim X, \] while the low-degree improvements use the DB index and a spectral sequence for \(R^qT_*(j_!\mathbb C_{Y\setminus F})\) [1307.7503]. In positive characteristic, current proofs rely instead on Grauert–Riemenschneider vanishing for \(F\)-pure threefolds, dlt modifications, Cartier operators, Frobenius trace, Witt vector sheaves, and inverse Cartier maps [2604.14760, 2507.04838]. The theorem also has clear limits. In positive characteristic it fails in general; one paper gives an explicit cone counterexample for which \[ R^{d-1}\pi_*\Omega_Y^{d-1}(\log E)(-E)\neq 0 \] on the blow-up of the cone vertex [2507.04838]. Even on surfaces, the stronger untwisted statement \(R^{d-1}f_*\Omega_Y^1(\log E)=0\) can fail, and the paper on surfaces stresses that the twisted logarithmic vanishing \[ R^1f_{*}\Omega^i_Y(\log B_Y)(-B_Y)=0 \] is the correct positive-characteristic analogue [2402.08153]. For sharply \(F\)-pure threefolds, the assumption \(p>5\) is used because the proof invokes rationality of three-dimensional klt singularities in characteristic \(p>5\) and the available MMP/dlt-modification framework; the same paper notes that \(p>5\) is not known to be optimal [2604.14760]. The name Steenbrink also appears in adjacent invariants, especially the Steenbrink spectrum \[ \operatorname{Sp}(f)=\sum_{\alpha\in\mathbf Q} n_{f,\alpha}\, t^\alpha, \] used in the study of projective hypersurfaces, pole-order spectral sequences, and Hirzebruch–Milnor classes [1212.1081, 1403.4563, 1312.0392]. This is related to the same Hodge-theoretic milieu but is distinct from Steenbrink vanishing itself. A plausible implication is that “Steenbrink vanishing” should be understood not as a single isolated theorem, but as one central component of a broader package linking resolutions, logarithmic forms, Du Bois theory, vanishing cycles, and singularity classes. Source: https://www.emergentmind.com/topics/steenbrink-vanishing