Griffiths Bundle in Hodge and Jet Theories

• Griffiths bundle is a construction in complex geometry that appears as a natural line bundle on period domains, a canonical height on curves, and a cocharacter in group-theoretic settings.
• It encapsulates curvature, positivity, and obstruction data, playing a vital role in the study of period maps, quasi-projectivity, and metric properties in both Hodge theory and jet differential geometry.
• Variants such as the Green–Griffiths jet bundle and canonical Griffiths height provide practical tools for addressing deformation, vanishing theorems, and the structure of automorphic forms.

The expression Griffiths bundle is used in several adjacent but non-identical ways in complex and algebraic geometry. In classical Hodge theory it denotes a natural line bundle on a period domain or on the image of a period map, typically extracted from the Hodge filtration; in the theory of variations of Hodge structure over curves it also appears as the canonical line bundle whose degree is the Griffiths height; in the group-theoretic reformulation it becomes a character attached to a reductive group, a cocharacter, and a representation; in hyperbolicity theory it is closely related to the Green–Griffiths and Demailly–Semple jet bundles; and in some extension and cycle-theoretic settings the same expression is used for obstruction or torsion constructions. This suggests that the term is context-dependent rather than canonical across subfields, but in each usage it packages curvature, positivity, or obstruction data in a form that is functorial enough to interact with period maps, flag bundles, or jet towers (Bakker et al., 2018, Goldring, 2018, Chen et al., 2024, Gavrilov, 2020).

1. Terminological scope and standard formulas

Several constructions appearing under the name or immediate orbit of “Griffiths bundle” can be organized as follows.

Context	Object	Formula or description
Period domains and VHS	Griffiths line bundle	$L := \det F^p$ for the highest nonzero piece of the Hodge filtration
VHS over a curve	Canonical Griffiths line bundle	$$GK_C(V,F^\bullet) := \bigotimes_{r=0}^n (\det F^r / F^{r+1})^{\otimes r}$$
Reductive groups	Griffiths character	$\grif(\mathbf G,\mu,r) := \det \Grif(\mathbf G,\mu,r)$
Jet differential theory	Green–Griffiths bundle	$E_{k,n}^{\rm GG}$, a bundle of weighted $k$-jet differentials
Semple tower	Tautological jet line bundle	$\mathscr O_{X_k}(1)$ on $X_k=\mathbb P(V_{k-1})$
Extension theory	Obstruction bundle cohomology	$H^2(X,N_{X/Y}^* \otimes \mathrm{End}(E))$ for the first Griffiths obstruction

The first three rows come from Hodge-theoretic and representation-theoretic constructions, while the last three arise from jet geometry and deformation theory (Bakker et al., 2018, Mordant, 2022, Goldring, 2018, Demailly, 2014, Gavrilov, 2020). A common misconception is to treat these as a single invariant. The literature here indicates instead that they are related by ancestry and analogy, not by a single universal definition.

2. Hodge-theoretic Griffiths line bundles, period maps, and heights

In the period-map setting, one starts with a smooth complex algebraic variety $X$ carrying a pure polarized integral variation of Hodge structure $(V_\mathbb Z,F^\bullet,Q)$, a period domain $$GK_C(V,F^\bullet) := \bigotimes_{r=0}^n (\det F^r / F^{r+1})^{\otimes r}$$0, an arithmetic lattice $$GK_C(V,F^\bullet) := \bigotimes_{r=0}^n (\det F^r / F^{r+1})^{\otimes r}$$1, and a period map

$$GK_C(V,F^\bullet) := \bigotimes_{r=0}^n (\det F^r / F^{r+1})^{\otimes r}$$2

The Griffiths line bundle, also called the Hodge bundle in the cited work, is a natural $$GK_C(V,F^\bullet) := \bigotimes_{r=0}^n (\det F^r / F^{r+1})^{\otimes r}$$3-line bundle $$GK_C(V,F^\bullet) := \bigotimes_{r=0}^n (\det F^r / F^{r+1})^{\otimes r}$$4 on $$GK_C(V,F^\bullet) := \bigotimes_{r=0}^n (\det F^r / F^{r+1})^{\otimes r}$$5, descending to $$GK_C(V,F^\bullet) := \bigotimes_{r=0}^n (\det F^r / F^{r+1})^{\otimes r}$$6, and described there by

$$GK_C(V,F^\bullet) := \bigotimes_{r=0}^n (\det F^r / F^{r+1})^{\otimes r}$$7

for the highest nonzero piece of the universal Hodge filtration. Its Hodge metric has nonnegative curvature in the Griffiths transverse directions, and that positivity is the geometric input behind Griffiths’ quasi-projectivity conjecture for images of period maps (Bakker et al., 2018).

Bakker, Brunebarbe, and Tsimerman prove that the image of a period map is the analytification of an algebraic space $$GK_C(V,F^\bullet) := \bigotimes_{r=0}^n (\det F^r / F^{r+1})^{\otimes r}$$8, that the restriction of the Griffiths bundle to $$GK_C(V,F^\bullet) := \bigotimes_{r=0}^n (\det F^r / F^{r+1})^{\otimes r}$$9 is ample as an algebraic $\grif(\mathbf G,\mu,r) := \det \Grif(\mathbf G,\mu,r)$0-bundle, and in particular that $\grif(\mathbf G,\mu,r) := \det \Grif(\mathbf G,\mu,r)$1 is quasi-projective (Bakker et al., 2018). Their proof combines definable complex analytic spaces, o-minimal GAGA, and algebraization of proper definable images.

For a variation of Hodge structures over a smooth projective complex curve $\grif(\mathbf G,\mu,r) := \det \Grif(\mathbf G,\mu,r)$2, the same tradition produces the canonical Griffiths line bundle

$\grif(\mathbf G,\mu,r) := \det \Grif(\mathbf G,\mu,r)$3

When the variation is defined only on $\grif(\mathbf G,\mu,r) := \det \Grif(\mathbf G,\mu,r)$4, Peters’ construction gives upper and lower Deligne extensions $\grif(\mathbf G,\mu,r) := \det \Grif(\mathbf G,\mu,r)$5 and $\grif(\mathbf G,\mu,r) := \det \Grif(\mathbf G,\mu,r)$6, and the associated Griffiths heights are

$\grif(\mathbf G,\mu,r) := \det \Grif(\mathbf G,\mu,r)$7

For the middle-dimensional cohomology of a pencil of projective complex hypersurfaces, Mordant derives formulas

$\grif(\mathbf G,\mu,r) := \det \Grif(\mathbf G,\mu,r)$8

together with

$\grif(\mathbf G,\mu,r) := \det \Grif(\mathbf G,\mu,r)$9

where $E_{k,n}^{\rm GG}$0 is the number of critical points and $E_{k,n}^{\rm GG}$1 is the intersection-theoretic height defined in the paper (Mordant, 2022). In this form, the Griffiths bundle becomes a height-theoretic object attached to a Hodge filtration with bad reduction allowed.

3. Group-theoretic generalization via Griffiths characters

A systematic generalization replaces Hodge data by a triple $E_{k,n}^{\rm GG}$2, where $E_{k,n}^{\rm GG}$3 is a connected reductive group over an arbitrary field $E_{k,n}^{\rm GG}$4, $E_{k,n}^{\rm GG}$5 is a cocharacter, and $E_{k,n}^{\rm GG}$6 is an $E_{k,n}^{\rm GG}$7-representation with central kernel. Goldring and Koskivirta define the filtration

$E_{k,n}^{\rm GG}$8

where $E_{k,n}^{\rm GG}$9 is the $k$0-weight space for $k$1, then the Griffiths module

$k$2

and finally the Griffiths character

$k$3

This lies in $k$4, with $k$5 (Goldring, 2018).

The classical Hodge-theoretic Griffiths bundle is recovered by taking $k$6, $k$7 the Mumford–Tate group of a $k$8-VHS, and $k$9 the tautological representation; in that case

$\mathscr O_{X_k}(1)$0

A central structural theorem states that when $\mathscr O_{X_k}(1)$1 is $\mathscr O_{X_k}(1)$2-simple, the Griffiths character is, up to positive multiples and suitable identifications, essentially independent of $\mathscr O_{X_k}(1)$3 with central kernel, and the corresponding ray is given by

$\mathscr O_{X_k}(1)$4

This is a strong rigidity statement: the asymptotic positivity direction is controlled by the cocharacter rather than the representation (Goldring, 2018).

The same framework applies in characteristic $\mathscr O_{X_k}(1)$5, where the analogous geometric context is a scheme mapping to a stack of $\mathscr O_{X_k}(1)$6-Zips. Under the orbitally $\mathscr O_{X_k}(1)$7-close condition on $\mathscr O_{X_k}(1)$8, the Griffiths line bundle of a projective $\mathscr O_{X_k}(1)$9-scheme is nef (Goldring, 2018). This provides a direct bridge from Hodge-theoretic positivity to automorphic and mod-$X_k=\mathbb P(V_{k-1})$0 period geometry.

4. Green–Griffiths bundles and jet-differential geometry

In jet geometry, the terminology shifts from the Griffiths line bundle to the Green–Griffiths bundles $X_k=\mathbb P(V_{k-1})$1. Over $X_k=\mathbb P(V_{k-1})$2, the fiber of $X_k=\mathbb P(V_{k-1})$3 consists of polynomials in derivatives of order $X_k=\mathbb P(V_{k-1})$4, with each monomial assigned weight exactly $X_k=\mathbb P(V_{k-1})$5. For $X_k=\mathbb P(V_{k-1})$6 and $X_k=\mathbb P(V_{k-1})$7,

$X_k=\mathbb P(V_{k-1})$8

The same paper gives an explicit basis in terms of determinants built from higher derivatives of affine coordinates, generalizing the Wronskian, and proves that the space is independent of $X_k=\mathbb P(V_{k-1})$9 in the range $H^2(X,N_{X/Y}^* \otimes \mathrm{End}(E))$0. By contrast, the negatively twisted case $H^2(X,N_{X/Y}^* \otimes \mathrm{End}(E))$1, and more generally the construction of explicit global sections over general type submanifolds $H^2(X,N_{X/Y}^* \otimes \mathrm{End}(E))$2, remains completely open (Chen et al., 2024).

Demailly’s directed-geometry formalism organizes jet data through the Semple tower

$H^2(X,N_{X/Y}^* \otimes \mathrm{End}(E))$3

with

$H^2(X,N_{X/Y}^* \otimes \mathrm{End}(E))$4

and tautological line bundle

$H^2(X,N_{X/Y}^* \otimes \mathrm{End}(E))$5

In this jet-space context, the tautological line bundle is the object often described as the Griffiths bundle. Its sections encode jet differentials, and the Green–Griffiths locus is defined from the base loci of

$H^2(X,N_{X/Y}^* \otimes \mathrm{End}(E))$6

A key inclusion is

$H^2(X,N_{X/Y}^* \otimes \mathrm{End}(E))$7

so the existence of enough sections of the tautological jet line bundle forces entire curves into a proper algebraic locus (Demailly, 2014).

A complementary description uses the Demailly–Semple tower and invariant jet differentials $H^2(X,N_{X/Y}^* \otimes \mathrm{End}(E))$8. For sufficiently large $H^2(X,N_{X/Y}^* \otimes \mathrm{End}(E))$9,

$X$0

Equivariant localization on the Demailly–Semple jet differentials bundle yields an affirmative answer to the Green–Griffiths–Lang conjecture for generic projective hypersurfaces $X$1 of degree $X$2 (Berczi, 2015). An alternative compactification motivated by Morin singularities leads to iterated residue formulas for tautological integrals and shows that the polynomial Green–Griffiths–Lang conjecture for a generic projective hypersurface of degree $X$3 follows from a positivity conjecture for Thom polynomials of Morin singularities (Berczi, 2010).

5. Positivity, Hermitian metrics, and differential-geometric theory

The surrounding positivity theory is governed by Griffiths’ conjecture: a holomorphic vector bundle on a projective manifold is ample if and only if it admits a Hermitian metric whose Chern curvature is positive in the sense of Griffiths. Demailly formulates a nonlinear Hermitian–Yang–Mills/Monge–Ampère type elliptic system for a family of Hermitian metrics $X$4,

$X$5

with the property that a solution at $X$6 would be dual Nakano positive and hence Griffiths positive. The cited paper proves essential ellipticity and short-time existence, while long-time existence up to $X$7 remains the central open problem (Demailly, 2020).

A rigidity result due to Pingali shows that the “cushioned” Hermitian–Einstein-type equation in Demailly’s approach has an essentially unique solution on an $X$8-stable bundle: any smooth Hermitian solution is of the form

$X$9

with $(V_\mathbb Z,F^\bullet,Q)$0 Hermitian–Einstein. Since Hermitian–Einstein metrics on stable bundles need not be Griffiths positive, this indicates that the original continuity method must be modified (Pingali, 2021). By contrast, on compact Riemann surfaces, Murakami proves Griffiths’ conjecture analytically using Demailly’s PDE system together with Uhlenbeck–Yau techniques and Pingali’s reduction; in that setting ampleness, Griffiths positivity, and solvability of the Demailly system are equivalent (Murakami, 27 Sep 2025).

Several additional results connect Griffiths positivity to auxiliary bundles and tautological constructions. An alternative definition of singular Hermitian metrics on a vector bundle $(V_\mathbb Z,F^\bullet,Q)$1 is given through $(V_\mathbb Z,F^\bullet,Q)$2-potentials on the tautological line bundle $(V_\mathbb Z,F^\bullet,Q)$3 over $(V_\mathbb Z,F^\bullet,Q)$4, and this framework supports generalized notions of Griffiths and strong Nakano positivity together with a generalized Griffiths vanishing theorem (Wu, 2020). Liu and Xu characterize Griffiths semi-positivity and strict positivity quantitatively through optimal $(V_\mathbb Z,F^\bullet,Q)$5-extension conditions, and identify pluriharmonicity and flatness with the equality case in those extension inequalities (Liu et al., 2022). In another direction, if $(V_\mathbb Z,F^\bullet,Q)$6 is Griffiths negative over a Kähler manifold, then the total space of $(V_\mathbb Z,F^\bullet,Q)$7 carries the Kähler form

$(V_\mathbb Z,F^\bullet,Q)$8

and its curvature is non-positive along the tautological direction (Kim et al., 2022). Finally, pointwise universal Gysin formulas for Hermitian flag bundles show that for a Griffiths semi-positive vector bundle,

$(V_\mathbb Z,F^\bullet,Q)$9

is a strongly positive $$GK_C(V,F^\bullet) := \bigotimes_{r=0}^n (\det F^r / F^{r+1})^{\otimes r}$$00-form and a positive linear combination of Schur forms, yielding new evidence toward a conjecture of Griffiths on positive polynomials in Chern forms (Diverio et al., 2020).

6. Obstruction-theoretic and cycle-theoretic usages

A separate use of the terminology appears in the extension problem for holomorphic vector bundles. Let $$GK_C(V,F^\bullet) := \bigotimes_{r=0}^n (\det F^r / F^{r+1})^{\otimes r}$$01 be a compact submanifold, $$GK_C(V,F^\bullet) := \bigotimes_{r=0}^n (\det F^r / F^{r+1})^{\otimes r}$$02 a holomorphic vector bundle, and

$$GK_C(V,F^\bullet) := \bigotimes_{r=0}^n (\det F^r / F^{r+1})^{\otimes r}$$03

the first-order neighborhood. The first obstruction to extending $$GK_C(V,F^\bullet) := \bigotimes_{r=0}^n (\det F^r / F^{r+1})^{\otimes r}$$04 to $$GK_C(V,F^\bullet) := \bigotimes_{r=0}^n (\det F^r / F^{r+1})^{\otimes r}$$05 is

$$GK_C(V,F^\bullet) := \bigotimes_{r=0}^n (\det F^r / F^{r+1})^{\otimes r}$$06

where $$GK_C(V,F^\bullet) := \bigotimes_{r=0}^n (\det F^r / F^{r+1})^{\otimes r}$$07 is the Atiyah class and $$GK_C(V,F^\bullet) := \bigotimes_{r=0}^n (\det F^r / F^{r+1})^{\otimes r}$$08 is the Kodaira–Spencer invariant of the embedding. Vanishing of this class is necessary and sufficient for the existence of an extension to $$GK_C(V,F^\bullet) := \bigotimes_{r=0}^n (\det F^r / F^{r+1})^{\otimes r}$$09, and the target group is described in the source summary as Griffiths bundle cohomology (Gavrilov, 2020). Here the “Griffiths bundle” is not a line bundle on a period domain or a jet tower, but an obstruction package measuring interaction between intrinsic bundle geometry and extrinsic deformation geometry.

The term also appears, in a more suggestive way, in the study of Griffiths groups. Schreieder proves that for a very general Enriques surface $$GK_C(V,F^\bullet) := \bigotimes_{r=0}^n (\det F^r / F^{r+1})^{\otimes r}$$10 and the Jacobian $$GK_C(V,F^\bullet) := \bigotimes_{r=0}^n (\det F^r / F^{r+1})^{\otimes r}$$11 of a very general quartic curve,

$$GK_C(V,F^\bullet) := \bigotimes_{r=0}^n (\det F^r / F^{r+1})^{\otimes r}$$12

has infinite $$GK_C(V,F^\bullet) := \bigotimes_{r=0}^n (\det F^r / F^{r+1})^{\otimes r}$$13-torsion. The basic explicit torsion classes are

$$GK_C(V,F^\bullet) := \bigotimes_{r=0}^n (\det F^r / F^{r+1})^{\otimes r}$$14

where $$GK_C(V,F^\bullet) := \bigotimes_{r=0}^n (\det F^r / F^{r+1})^{\otimes r}$$15 is the unique nontrivial $$GK_C(V,F^\bullet) := \bigotimes_{r=0}^n (\det F^r / F^{r+1})^{\otimes r}$$16-torsion line bundle and $$GK_C(V,F^\bullet) := \bigotimes_{r=0}^n (\det F^r / F^{r+1})^{\otimes r}$$17 is the Ceresa cycle on $$GK_C(V,F^\bullet) := \bigotimes_{r=0}^n (\det F^r / F^{r+1})^{\otimes r}$$18. The key injective map is

$$GK_C(V,F^\bullet) := \bigotimes_{r=0}^n (\det F^r / F^{r+1})^{\otimes r}$$19

The paper further suggests that the “Griffiths bundle,” understood there as the group of Griffiths group torsion cycles with trivial Abel–Jacobi invariants, can be unexpectedly large (Schreieder, 2020). This is another instance where the phrase is used locally rather than universally.

Across these contexts, the Griffiths bundle is best viewed not as a single object but as a family of constructions attached to Hodge filtrations, reductive groups, jet towers, or deformation data. The recurring structural features are determinant-type functoriality, compatibility with positivity or curvature, and the capacity to convert transcendental or infinitesimal information into algebraic line-bundle or cohomological form.