VARIATIONS OF HODGE STRUCTURE CONSIDERED

AS AN EXTERIOR DIFFERENTIAL SYSTEM: OLD

AND NEW RESULTS

JAMES CARLSON, MARK GREEN, AND PHILLIP GRIFFITHS

Outline
I. Introduction
II. Preliminaries
II.A. Period domains
II.B. Exterior differential systems (EDS)
III. The exterior differential system associated to a variation of Hodge
structure (VHS)
III.A. Elementary examples
III.B. A Cartan-Kähler example and a brief guide to some of the
literature
III.C. The derived flag of the EDS associated to a VHS
IV. Universal cohomology associated to a homogeneous Pfaffian system
IV.A. EDS aspects of homogeneous Pfaffian systems
IV.B. Characteristic cohomology of the homogeneous Pfaffian system
associated to a VHS
V. “Expected” dimension counts for integral manifolds of an EDS

I. Introduction

Hodge theory provides the basic invariants of a complex algebraic

variety. The two central open problems in the subject, the conjectures
of Hodge and of Bloch-Beilinson, relate Hodge theory to the geome-
try/arithmetic of a complex algebraic variety.
The space of all polarized Hodge structures of weight n and with a
given sequence h = (hn,0 , hn−1,1 , . . . , h0,n ), hp,q = hq,p , of Hodge num-
bers forms naturally a homogeneous complex manifold Dh , and the
moduli space Mh of equivalence classes of polarized Hodge structures
is a quotient of Dh by an arithmetic group acting properly discontinu-
ously. In what we shall call the classical case when the weights n = 1
1
2 J. CARLSON, M. GREEN, AND P. GRIFFITHS 4/19/09

or n = 2 and h2,0 = 1, Dh is a bounded symmetric domain and Mh

is a quasi-projective variety defined over a number field.1 In this case
the relation between the Hodge theory and geometry/arithmetic of a
variety is an extensively developed and deep subject.
In the non-classical, or what we shall refer to as the higher weight,
case2 the subject is relatively less advanced. The fundamental differ-
ence between the classical and higher weight cases is that in the latter
case the Hodge structures associated to a family of algebraic varieties
satisfy a universal system I ⊂ T ∗ D of differential equations. In this
partly expository paper we will discuss the system I from the per-
spective of exterior differential systems (EDS’s) with the three general
objectives:
(i) To summarize some of what is known about I from an EDS
perspective.
(ii) To define and discuss the “universal characteristic cohomology”
associated to a homogeneous Pfaffian system in the special case
of a variation of Hodge structures.
(iii) To discuss and illustrate the question “How must expected di-
mension counts be modified for integral manifolds of the sys-
tem I?”
One overarching objective is this: When one seeks to extend much
of the rich classical theory, including the arithmetic aspects and the
connections with automorphic forms, the various compactifications of
M and the resulting boundary cohomology, the theory of Shimura va-
rieties, etc., the fact that families of Hodge structures arising from
geometry are subject to differential constraints seems to present the
major barrier. Perhaps by better understanding the structure of these
differential constraints, some insight might be gained on how at least
some aspects of the classical theory might be extended. We are espe-
cially interested in properties of variations of Hodge structure that are
not present in the classical case, as these may help to indicate what

1Henceforthwe shall drop reference to the h on D unless it is needed.

2
It being understood that when n = 2 we have h2,0 = 2.
4/19/09 VARIATIONS OF HODGE STRUCTURE 3

needs to be better understood to be able to extend the classical case

to higher weight Hodge structures.
In more detail, in section II.A we will review the definitions and
establish notations for polarized Hodge structures, period domains and
their duals, and the infinitesimal period relation, which is the basic
exterior differential system studied in this paper. In section II.B we
recall some of the basic definitions and concepts from the theory of
exterior differential systems.
In section III we discuss the basic exterior differential system whose
integral manifolds define variations of Hodge structure. The basic gen-
eral observation is that the integral elements are given by abelian sub-
algebras
a ⊂ G−1,1 ⊂ GC

of the complexified Lie algebras of the symmetry group of the period

domain. We then go on in section III.A to discuss in some detail two
important examples where the EDS given by the infinitesimal period
relation may be integrated by elementary methods; both of these have
been discussed in the literature and here we shall summarize, in the
context of this paper those results. Then in section III.B we shall
discuss an example, the first we are aware of in the literature, where
the Cartan-Kähler theoretic aspects of exterior differential systems are
applied to the particular EDS arising from the infinitesimal period
relation. Finally, in section III.C we study the derived flag of the
infinitesimal period relation I. One result is that if all the Hodge
numbers hp,q are non-zero, then the derived flag of I terminates in zero
and has length m where m 5 log n
log 2
, n being the weight of the Hodge
structure.
Over the years there have been a number of studies of the EDS given
by the infinitesimal period relation. Here we mention [A], [C], [CD],
[CKT], [CS], [CT], [CT2] and [M]. Section III should be considered
as an introduction to those works. In particular the paper [M], which
builds on and extends the earlier works, contains a definitive account
of the bounds on the dimension and rigidity properties of maximal
integral elements. At the end of section III.C we shall comment on some
4 J. CARLSON, M. GREEN, AND P. GRIFFITHS 4/19/09

interesting questions that arise from [M] and [A] as well as providing a
brief guide to the earlier works referred to above.
In section IV, we first discuss some general aspects of homogeneous
Pfaffian systems, including expressing the invariant part of their char-
acteristic cohomology in terms of a Lie algebra cohomology construc-
tion.3 We then turn to the group invariant characteristic cohomology
of period domains. Here there is a very nice question
(I.1) Is the invariant part of the characteristic cohmology of a period
domain generated by the Chern forms of the Hodge bundles?
In the classical case the answer is positive and may be deduced from
what is known in the literature; we will carry this out below.
In the non-classical case when the Pfaffian system I associated to a
variation of Hodge structures is non-trivial, new and interesting issues
arise. It seems likely that the question will have an affirmative answer;
this will be the topic of a separate work.4
In this paper we will establish two related results. The first is that
we will show that the invariant forms modulo the algebraic ideal gener-
ated by I are all of type (p, p). A consequence is that on the complex
of invariant forms the Lie algebra cohomology differential δ = 0. This
is analogous to what happens in the Hermitian symmetric case. How-
ever, in the non-classical case there are always more invariant forms
than those generated by the Chern forms of the Hodge bundles, and
the integrability conditions; i.e. the full differential ideal generated by
I must be taken into account. This involves subtle issues in represen-
tation theory and, as mentioned above, this story will be reported on
separately.
What we will prove here is that the integrability conditions imply
topological conditions in the form of new relations among the Chern
classes of a VHS. Denoting by Fp the Hodge filtration bundles we show
3Here, we recall that the characteristic cohomology of an exterior differential
system is the de Rham cohomological construction that leads to cohomology groups
that induce ordinary de Rham cohomology on integral manifolds of the EDS. The
precise definition is recalled in section IV below.
4This question has now been answered in the affirmative and the proof will
appear in a separate publication.
4/19/09 VARIATIONS OF HODGE STRUCTURE 5

that the Chern forms satisfy

ci (Fp )cj (Fn−p ) = 0 if i + j > hp,n−p .

What we see then is that this new algebro-geometric information has

resulted from EDS considerations.
The question (I.1) would have the following algebro-geometric con-
sequence: First, for any discrete group Γ acting properly discontinu-
ously on a period domain D, the invariant characteristic cohomology
HI∗ (D)GR induces characteristic cohomology on the quotient Γ\D. A
global variation of Hodge structure is given by a period mapping

(I.2) f : S → Γ\D

where S is a smooth, quasi-projective algebraic variety. Since (I.2)

is an integral manifold of the canonical EDS on Γ\D, the invariant
characteristic cohomology induces ordinary cohomology

f ∗ HI∗ (D)GR ⊂ H ∗ (S) .

We may think of HI∗ (D)GR as universal characteristic cohomology for

the exterior differential system corresponding to the infinitesimal pe-
riod relation, in that it induces ordinary cohomology on the parameter
space for variations of Hodge structures irrespective of the particular
group Γ. We may thus call it the universal characteristic cohomology.
A positive answer to the question would first of all imply that the uni-
versal characteristic cohomology is generated by the Chern classes of
the Hodge bundles over S. Although, perhaps not surprising to an
algebraic geometer this would be a satisfying result.
In section V we turn to the interesting question

(I.3) How must one correct expected dimension counts in the presence
of differential constraints?

Specifically, given a manifold A and submanifold B ⊂ A, for a “general”

submanifold X ⊂ A where dim X + dim B = dim A, we will have

(I.4) codimB (X ∩ B) = codimA B .

6 J. CARLSON, M. GREEN, AND P. GRIFFITHS 4/19/09

Thus, the RHS of this equation may be thought of as the “expected

codimension” of X ∩ B in B. If X ∩ B is non-empty, the actual codi-
mension is no more than the expected codimension.
Suppose now that there is a distribution W ⊂ T A and X is con-
strained to have T X ⊂ W . Then how does this affect the expected
dimension counts? In case W meets T B transversely, one sees im-
mediately that the “expected codimension” counts decrease. Taking
integrability into account gives a further correction. Rather than try-
ing to develop the general theory, in section V we shall discuss one
particularly interesting special case.
This case concerns Noether-Lefschetz loci. Here we denote by WI ⊂
T D the distribution I ⊥ ; integral manifolds of I have their tangent
spaces lying in WI . Given ζ ∈ HR , there is a homogeneous sub-period-
domain Dζ ⊂ D, defined as the set of polarized Hodge structures of
weight n = 2m where ζ ∈ H m,m . We have

codimD Dζ = h(2m,0) + · · · + h(m+1,m−1) ,

and the distribution WI on D meets T Dζ transversely. For a variation

of Hodge structure, given by an integral manifold

f :S→D

of the canonical system I on D, the Noether-Lefschetz locus Sζ ⊂ S is

given by
Sζ = f −1 (Dζ ) .
In algebro-geometric questions, ζ is usually taken to be a rational vec-
tor, but that will not concern us here. The refined codimension esti-
mate given by WI alone, i.e., without taking integrability conditions
into account, is
codimS Sζ 5 hm−1,m+1 .
In the case m = 1, which algebro-geometrically reflects studying codi-
mension one algebraic cycles, the distribution WI does not enter and
the estimate is classical, especially in the study of curves on an alge-
braic surface. In the case m = 2, it is non-classical and seems only
recently to have been discussed in the literature (cf. [GG], [V], [O1],
4/19/09 VARIATIONS OF HODGE STRUCTURE 7

[O2]). When the integrability conditions are taken into account, the
above codimension estimate is refined to
codimS Sζ 5 hm−1,m+1 − σζ ,
where σζ is a non-negative quantity constructed from ζ and the integral
element of I given by the tangent space to f (S) at the point in question.
Assuming the Hodge conjecture, the above would say that “there are
more algebraic cycles than a naı̈ve dimension count would suggest”.
In addition to establishing the above inequality, we will show that
it is an equality in a significant example, namely, that given by a hy-
persurface X ⊂ P5 of degree d = 6 and which is general among those
containing a 2-plane. This indicates that there is no further general
estimate.
We conclude this section by analyzing the case of Calabi-Yau four-
folds, where the quantity h3,1 −σζ has a particularly nice interpretation,
including an interesting arithmetic consequence of the Hodge conjec-
ture.

II. Preliminaries
II.A. Period domains.5 Let H be a Q-vector space. A Hodge struc-
ture (HS) of weight n is given by any of the following equivalent data:
(II.1) A Hodge decomposition
HC = ⊕ H p,q
(
p+q=n

H q,p = H̄ p,q .
(II.2) A Hodge filtration F n ⊂ F n−1 ⊂ · · · ⊂ F 0 = HC where for
each p
∼
F p ⊕ F̄ n−p+1 → HC ;

(II.3) A homomorphism of real Lie groups

ϕ : C∗ → GL(HR )
of weight n in the sense that for z ∈ C∗ , λ ∈ R∗
ϕ(λz) = λn ϕ(z).
5The general reference for this section is the book [C-MS-P].
8 J. CARLSON, M. GREEN, AND P. GRIFFITHS 4/19/09

The relation between (II.1) and (II.2) is

0 0
F p = ⊕ H p ,n−p
(
p0 =p
p,q
H = F ∩ F̄ q .
p

The relation between (II.1) and (II.3) is

ϕ(z)u = z p z̄ q u, u ∈ H p,q .
This means: The element ϕ(z) ∈ GL(HC ) just given lies in the sub-
group GL(HR ) of GL(HC ) and has weight n.
If we restrict ϕ to the maximal compact subgroup S 1 = {z ∈ C∗ :
|z| = 1} of C∗ , then for z ∈ S 1 , u ∈ H p,q
ϕ(z)u = z p−q u ,
and this shows how to recover the Hodge decomposition as the z p−q
eigenspace of ϕ restricted to S 1 . The Weil operator is defined by
√
C = ϕ( −1) .
We define the Hodge numbers hp,q := dim H p,q , and we set f p :=
p0 ,n−p0
P
p1 =p h .
Since the Hodge filtration point of view will be the dominant one in
this paper, we shall denote a Hodge structure by (H, F ).
Now let
Q:H ⊗H →Q
be a non-degenerate bilinear form satisfying Q(u, v) = (−1)n Q(v, u)
for u, v ∈ H. A Hodge structure (H, F ) is polarized by Q if the Hodge-
Riemann bilinear relations
(
Q(F p , F n−p+1 ) = 0
(II.4)
Q(Cu, ū) > 0 for 0 6= u ∈ HC
are satisfied. The first of these is equivalent to F n−p+1 = (F p )⊥ . A
polarized Hodge structure (PHS) will be denoted by (H, Q, F ).
In the definition (II.3), for a polarized Hodge structure we need to
restrict ϕ to S 1 in order to preserve, and not just scale, the polarization.

Definitions. (i) A period domain D is given by the set of polarized

Hodge structures (H, Q, F ) with given Hodge numbers hp,q . (ii) The
4/19/09 VARIATIONS OF HODGE STRUCTURE 9

compact dual Ď is given by all filtrations F with dim F p = f p and

which satisfy the first bilinear relation F n−p+1 = (F p )⊥ in (II.4).

We shall denote by G the Q-algebraic group Aut(H, Q), and by GR

and GC the corresponding real and complex forms. It is elementary
that GR acts transitively on D, and choosing a reference point F0 ∈ D
we have
D∼= GR /V
where V is the compact subgroup of GR preserving the Hodge decom-
position HC = ⊕ H0p,q corresponding to F0 . In terms of (II.3) we
p+q=n
note that
ϕ(S 1 ) ⊂ V ;
in fact, V is the centralizer of the circle ϕ(S 1 ) in GR .
The complex Lie group GC acts transitively on the compact dual Ď,
and choosing a reference point F0 ∈ D as above we have
Ď ∼
= GC /B
where B ⊂ GC is a parabolic subgroup. We have
V = GR ∩ B .
The compact dual is a projective algebraic variety defined over Q. In
fact we have an obvious inclusion
n+1
[Y2 ]

(II.5) Ď ⊂ Grass(f p , HC )
p=n

and we may embed Ď in a PN by means of the Plücker coordinates of

the flag subspaces F p ⊂ HC .
Hodge structures and polarized Hodge structures are functorial with
respect to the standard operations in linear algebra. In particular, a
Hodge structure (H, F ) induces a Hodge structure of weight zero on
End(H) where
End(HC )r,−r = A ∈ End(HC ) : A(H p,q ) ⊂ H p+r,q−r .

(II.6)
A polarized Hodge structure (H, Q, F0 ) induces a polarized Hodge
structure on the Lie algebra G of G, where Gr,−r is given by (II.6)
10 J. CARLSON, M. GREEN, AND P. GRIFFITHS 4/19/09

and the polarization is induced by the Cartan-Killing form. For later

use, we note that from (II.6) we have

G ,G ⊆ Gr+s,−(r+s) .
 r,−r s,−s 
(II.7)

If we recall the natural identification

TF p Grass(f p , HC ) ∼
= Hom(F p , HC /F p )

it follows that the Lie algebra of B is

b = ⊕ Gr,−r .
r=0

The subalgebra
p = ⊕ G−r,r
r>0

gives a complement to b in GC leading to the natural identification

(II.8) TF • Ď ∼
=p.

By (II.7) the subspace

G−1,1 ⊂ p

is Ad B-invariant and therefore defines a GC -invariant distribution

WI ⊂ T Ď ,

and, by orthogonality, a Pfaffian system

I ⊂ T ∗ Ď .

The sub-bundle I restricts to a GR -invariant sub-bundle I ⊂ T ∗ D.

Definition. The Pfaffian system I is called the infinitesimal period

relation.

It is this exterior differential system that we shall discuss in this

paper.
In the literature the distribution WI ⊂ T D is frequently referred to
as the horizontal sub-bundle.
4/19/09 VARIATIONS OF HODGE STRUCTURE 11

II.B. Exterior differential systems (EDS).6 Although the subject is

usually discussed in the smooth category, here we shall work complex-
analytically. A Pfaffian system is given by a holomorphic sub-bundle
I ⊂ T ∗M
of the cotangent bundle of a complex manifold. Associated to I is the
differential ideal
I ⊂ Ω•M
generated by the holomorphic sections of I together with their exterior
derivatives. We shall assume that the values of the sections of I gener-
ate a sub-bundle of Λ• T ∗ M ; i.e., I is locally free. An integral manifold,
or just an “integral”, of I is given by a complex manifold N and a
holomorphic immersion
f :N →M
such that
(II.9) f ∗ (I) = 0 .
If we denote by
WI = I ⊥ ⊂ T M
the holomorphic distribution associated to I, the condition (II.9) is
equivalent to
f∗ (T N ) ⊂ WI .
An important invariant associated to a Pfaffian system is its derived
flag. The exterior derivative induces a bundle map
δ : I → Λ2 T ∗ M/I ∧ T ∗ M ,
and recalling our assumption that δ has constant rank we set
I[1] = ker δ .
This is again a Pfaffian system, and continuing in this way leads to the
derived flag in T ∗ M
(II.10) I ⊃ I[1] ⊃ I[2] ⊃ · · · ⊃ I[m] = I[m+1] = · · · = I[∞] .
Here, I[∞] is the largest integrable or Frobenius subsystem of I.
6General references for this section are books [IL] and [BCGGG]; especially the
former contains essentially all the background needed for this work.
12 J. CARLSON, M. GREEN, AND P. GRIFFITHS 4/19/09

Dually, for the distribution we denote by

[1]
WI = WI + [WI , WI ]
the distribution generated by WI and the brackets of sections of WI .
Continuing in this way we obtain the flag in T M dual to (II.10)
[1] [m] [m+1] [∞]
WI ⊂ WI ⊂ · · · ⊂ WI = WI = · · · = WI .
[∞]
We say that I is bracket generating in case WI = T M , or equivalently
I [∞] = (0). In this case, by the holomorphic version of Chow’s theorem
we may connect any two points of M by a chain of holomorphic discs
that are integral curves of WI .
Two central aspects of the theory of exterior differential systems
are (i) regular and ordinary integral elements and the Cartan-Kähler
theorem, and (ii) prolongation and involution. For the first, an integral
element for I is given by a linear subspace E ⊂ Tx M such that
θE =: θ |E = 0
for all θ ∈ I. We may think of E as an infinitesimal solution of the
EDS. Denote by
π : Gp (T M ) → M
the bundle whose fibre π −1 (x) = Grp (Tx M ) over x ∈ M is the Grass-
mannian of p-planes in Tx M . In Gp (T M ) there is the complex analytic
subvariety
Gp (I) ⊂ Gp (T M )
of integral elements defined by the Pfaffian system I.
Integral elements are constructed one step at a time by solving linear
equations. For an integral element E ∈ Gp (I), we define the polar space
H(E) = {v ∈ Tx M : span{v, E} is an integral element} .
The equations that define H(E)
hθ(x), v ∧ Ei = 0 for all θ ∈ Ip+1
are linear in v, and we measure their rank by defining
r(E) = codim H(E)
= dim P(H(E)/E) .
4/19/09 VARIATIONS OF HODGE STRUCTURE 13

Given E0 ∈ Gp (I) we choose a p-form Ω such that ΩE0 6= 0. For

θ ∈ Ip and E ∈ Gp (T M ) near E0 , for each ϕ ∈ Ip we write
θE = fθ (E)ΩE .
Then Gp (I) is locally defined by the analytic equations fθ (E) = 0; we
say that E0 is regular if Gp (I) is smoothly defined by these equations.
Given E ∈ Gp (I) we choose a generic flag
0 ⊂ E1 ⊂ · · · ⊂ Ep−1 ⊂ E
and define ci to be the rank of the polar equations of Ei . A central
result is

Cartan’s test. (i) We have

codim Gp (I) = c1 + · · · + cp−1 .
(ii) If equality holds, then for k < p each Ek is regular.

By definition we say that E is ordinary if equality holds in Cartan’s

test. We say that I is involutive near an ordinary integral element.
The Cartan-Kähler existence theorem states that an ordinary integral
element is locally tangent to an integral manifold. It goes further to
say “how many” local integral manifolds there are. This will be briefly
mentioned below.
If Gp (I) is a submanifold near E, but E is not ordinary one needs to
prolong the EDS to be able to use the Cartan-Kähler theorem in order
to construct local integral manifolds. To explain this, we first observe
that there is a canonical Pfaffian system
J ⊂ T ∗ Gp (T M )
defined by writing points of Gp (T M ) as (x, E) and setting
J(x,E) = π ∗ (E ⊥ ) .
For any immersion f : N → M where dim N = p, there is a canonical
lift
Gp (T M )
;
vv
vv
f (1)
v π
vv
vv f 
N / M,
14 J. CARLSON, M. GREEN, AND P. GRIFFITHS 4/19/09

where for y ∈ N we set f (1) (y) = (f (y), f∗ Ty N ). This lift is an integral

manifold of J. The converse is true provided the p-dimensional integral
manifold of J projects to an immersed p-dimensional submanifold in M .
(1) (1)
We define the 1st prolongation (MI , I (1) ) of (M, I) by taking MI
to be the smooth locus of Gp (I) and defining I (1) to be the restriction of
(1) (1)
J to MI . An integral manifold of (M, I) gives one of (MI , I (1) ); the
converse holds in the sense explained above. The Cartan-Kuranishi
theorem states that, with some technical assumptions, after a finite
number of prolongations an exterior differential system either becomes
involutive or else empty (no solutions).
From the above we may infer the following

(II.11) Let f : N → M be an integral manifold of a Pfaffian system. If,

at a general point y ∈ N , f∗ (Ty N ) is not an ordinary integral element
then one of the following holds:
(A) f∗ (T N ) lies in the singular locus of Gp (I); or
(1)
(B) the prolongation f (1) : N → MI is an integral manifold of I (1) .

We may iterate this for the successive prolongations. Since the failure
to be involutive means that f : N → M satisfies additional differential
equations not in I, we have the following

(II.12) Conclusion. If the assumption of (II.11) holds, then f : N → M

satisfies additional differential equations that are canonically associated
to I but are not in the differential ideal I.

III. The exterior differential system associated to a

variation of Hodge structure
In this section, unless mentioned otherwise we shall work locally. Let
D be a period domain and I ⊂ T ∗ D the infinitesimal period relation
with corresponding horizontal distribution WI ⊂ T D. We recall that
both I and WI are the restrictions to D ⊂ Ď of similar structures over
the compact dual of D.

III.A. Elementary examples.

4/19/09 VARIATIONS OF HODGE STRUCTURE 15

Definition. A variation of Hodge structure (VHS) is given by an inte-

gral manifold f : S → D of I.

In the classical case where D is a bounded symmetric domain, it is

well known that holomorphic mappings from quasi-projective varieties
to quotients Γ\D of D by arithmetic groups have strong analytical
properties arising from the negative holomorphic sectional curvatures
in T D. Similar results hold in the general case as a consequence of the
horizontality of a VHS. There are also analogous properties that hold
for the curvatures of the Hodge bundles. These results are classical; cf.
[C-MS-P] for an exposition and references. In this partly expository
paper we will not discuss these curvature properties, but rather we will
focus on variations of Hodge structure from an EDS perspective.
We begin with the
(III.1) Basic observation. Integral elements of I are given by abelian
subalgebras
a ⊂ G−1,1 .
This result is a consequence of the Maurer-Cartan equation; it is
discussed in a more general context at the beginning of section IV below
(cf. the proof of proposition (IV.2). Here we are using the notations in
section II.A above. At a point F ∈ D, the Lie algebra G has a Hodge
structure constructed from F , and (III.1) above identifies the space of
integral elements of I in TF D.
In the theory of EDS, there are two types of exterior differential
systems, which may be informally described as follows:
• Those that may be integrated by ODE methods.
• Those that require PDE techniques.
Although the first type may be said to be “elementary”, it includes
many important EDS’s that arise in practice. Roughly speaking, these
are the exterior differential systems that, by using ODE’s, may be put
in a standard local normal form.
We begin by first recalling the contact system and then giving two
interesting elementary examples that arise in variations of Hodge struc-
tures.
16 J. CARLSON, M. GREEN, AND P. GRIFFITHS 4/19/09

Contact system. Here, dim M = 2k + 1 and I is locally generated

by a 1-form θ with θ ∧ (dθ)k 6= 0. Then, by using ODE methods, it
may be shown that there are local coordinates (x1 , . . . , xk , y, y1 , . . . , yk )
such that θ may be taken to be given by (using summation convention)
θ = dy − yi dxi .
Local integral manifolds are of dimension 5 k, and those of dimension
k on which dx1 ∧ · · · ∧ dxk 6= 0 are graphs
(x1 , . . . , xk ) → (x1 , . . . , xk , y(x), ∂x1 y(x), . . . , ∂xk y(x)) .

Proposition. For weight n = 2 and Hodge numbers h2,0 = 2, h1,1 = k,

I is a contact system ([CT2]).

Proof. We shall use the proof to illustrate and mutually relate the two
main computational methods that have been used to study the differ-
ential system I.
The first is using the Hodge structure on G to identify both the
tangent space at F ∈ D and the fibre WI,F with subspaces of GC .
For weight n = 2 we have for p ∼ = TF D the inclusion
(III.2) p ⊂ Hom(H 2,0 , H 1,1 ) ⊕ Hom(H 2,0 , H 0,2 ) ⊕ Hom(H 1,1 , H 0,2 ) .
Using the polarizing form Q gives
H ∼
 1,1
= Ȟ 1,1
H 0,2 ∼
= Ȟ 2,0 .
Then p is given by A = A1 ⊕ A2 ⊕ A3 in the RHS of (III.2) where

A2 = −tA2
(III.3)
A3 = tA1 .
The transpose notation refers to the identification just preceeding (III.2).
The sub-space G−1,1 is defined by A2 = 0. By (III.1), integral elements
are then given by linear subspaces
(III.4) a ⊂ Hom(H 2,0 , H 1,1 )
satisfying
(III.5) A tB = B tA
for A, B ∈ a.
4/19/09 VARIATIONS OF HODGE STRUCTURE 17

The second method is via moving frames. Over an open set in D we

choose a holomorphic frame field
e , . . . e , f , . . . , fk , e∗1 , . . . , e∗h
|1 {z }h 1
| F 2 {z }
F1
adapted to the Hodge filtrations and to Q in the sense that
Q(eα , e∗α ) = 1
(

Q(fi , fi ) = 1
and all other inner products are zero. Then, using summation conven-
tion,
deα = θαβ eβ + θαi fi + θαβ e∗β
(

dfi = θiα eα + θij fj + θiα e∗α .

Denoting TF D by just T and referring to (III.2), (III.3) we have
θαi ∈ T ∗ ⊗ Hom(H 2,0 , H 1,1 ) ↔ A1
θαβ ∈ T ∗ ⊗ Hom(H 2,0 , H 0,2 ) ↔ A2 .
From 0 = dQ(eα , eβ ) = Q(deα , eβ ) + Q(eα , deβ ) we have
θαβ + θβα = 0
which is the first equation in (III.3). The second equation there is
θiα = θαi .
From the above we see that I is generated by the h(h − 1)/2 1-forms
θαβ for α < β, where h = h2,0 . When h = 1, we are in the classical case
and I is zero. When h = 2, I is generated by a single 1-form θ = θ12
and is thus a candidate to be a contact system.
To calculate dθ we use d(de1 ) = 0, which together with the above
formulas give
dθ ≡ θ1i ∧ θ2i mod θ .
Since the 1-forms θαi are independent, we see that θ ∧ (dθ)k 6= 0 as
desired. 
The simplest case of this example is when k = 1. Then dim D = 3
and I is locally equivalent to the standard contact system
(III.6) dy − y 0 dx = 0
18 J. CARLSON, M. GREEN, AND P. GRIFFITHS 4/19/09

in C3 with coordinates (x, y, y 0 ).

In general, as noted above the contact system is locally equivalent
to the standard one generated by

θ = dy − yi dxi

in C2k+1 with coordinates (x1 , . . . , xk , y, y1 , . . . , yk ). Their integral man-

ifolds are graphs

(x1 , . . . , xk , y(x), ∂x1 y(x), . . . , ∂xk y(x))

where y(x) is an arbitrary function of x1 , . . . , xk . It is of interest to see

explicitly how one may construct integral manifolds depending on one
arbitrary function of k variables.
For this it is convenient to choose a basis for HC relative to which
 
0 0 I }h
Q =  0 −I 0 }k
 

I 0 0 }h
|{z} |{z} |{z}
h k h

Then F 2 will be spanned by the columns in a matrix

 
A }h
F =  B } k
 

C }h
|{z}
h

In an open set we will have det C 6= 0, and it is convenient to normalize

to have C = I. The equations tF QF = 0 are

(III.7) A + tA = tBB .

The infinitesimal period relation tF Q dF = 0 is

(III.8) dA = tB dB .

Note that adding to this relation its transpose gives the differential
of (III.7). Thus

(III.9) If (III.8) is satisfied and (III.7) is satisfied at one point, then

it is satisfied identically.
4/19/09 VARIATIONS OF HODGE STRUCTURE 19

We now specialize to the case h = 2 and write

B1 x1
 
!
A11 A12  .. .. 
A= B= . .
A21 A22
Bk xk
where the Aαβ , Bα , xj are independent variables and eventually the
Aαβ and Bj are to be functions of x1 , . . . , xk . The equations (III.8) are
P 2 
Bj
(i) dA11 = ΣBj dBj = d ;
2
P j 2 
j j (x )
(ii) dA22 = Σx dx = d ;
2
(iii) dA12 = Bj dxj ;
(iv) dA21 = xj dBj .
Now 0 = d(dA12 ) = ΣdBj ∧ dx2 or
Bj (x) = ∂xj B(x)
where B(x) is an arbitrary function of x1 , . . . , xk .
This then leads to the construction of integral manifolds, here con-
structed by ODE methods. As previously noted, in general, PDE meth-
ods — the Cartan-Kähler theorem — are required.
In the above example the 1st derived system I[1] = (0). Below we
shall give an example, usually referred to as the mirror quintic, that
shows a different phenomenon. In general we have
(III.10) dθαβ ≡ θαi ∧ θβi mod span{θαβ ’s} .
We may then think of I as a sort of multi-contact system (cf. [M]).
Note that for each α < β
θαβ ∧ (dθαβ )k 6= 0
and
^
θαβ ∧ (dθαβ )k =


6 0.
α<β

Integral elements are spanned by matrices Aiαλ satisfying (using sum-

mation convention)
Aiαλ Aiβµ = Aiαµ Aiβλ
20 J. CARLSON, M. GREEN, AND P. GRIFFITHS 4/19/09

corresponding to subspaces span{Aiαλ } ⊂ Hom(H 2,0 , H 1,1 ) on which

the RHS of (III.10) vanishes.
When h = h2,0 = 2, the maximal abelian subalgebras (III.4) have
dimension k = h1,1 and correspond to Lagrangian subspaces in the
symplectic vector space G−1,1 . In general, the multi-contact nature of
I suggests a bound of the sort
1
dim a 5 (h2,0 h1,1 ) .
2
This is in fact proved in [C], where it is also shown that the bound is
sharp for h1,1 even. When h1,1 is odd, it is shown there that the sharp
bound is 12 h2,0 (h1,1 − 1) + 1. Because it illustrates yet another method
for estimating the dimension of, and actually constructing, integral
elements we shall give a proof of this result in the cases h2,0 = 2 and
h2,0 = 3. By (III.5), we are looking for a linear subspace
E ⊂ Hom(H 2,0 , H 1,1 )
with a basis A1 , . . . , Ar satisfying
Ai tAj = Aj tAi
for all i, j. A key fact is
Ai (v) = 0 ⇒ Q(Aj (v), Image Ai ) = 0
for all j.

Case h2,0 = 2. We want to show that r = k+1 is impossible. Assuming

that r = k + 1 then some A ∈ E drops rank. With suitable labelling,
for some 0 6= v ∈ H 2,0 we will have Ak+1 (v) = 0. By the above, this
gives
Ai (v) ∈ Im(Ak+1 )⊥
for 1 5 i 5 k. Remembering that h1,1 = k, we have dim Im(Ak+1 )⊥ 5
k − 1 so that A1 (v), . . . , Ak (v) are linearly dependent. Relabelling, we
may assume that Ak (v) = 0. Now
dim(Im Ak + Im Ak+1 ) = 2 ,
since otherwise some linear combination of Ak and Ak+1 is zero. Then
Ai (v) = (Im Ak + Im Ak+1 )⊥
4/19/09 VARIATIONS OF HODGE STRUCTURE 21

for i = 1, . . . , k − 1 forces a linear dependence on A1 (v), . . . , Ak−1 (v).

Proceeding by downward induction gives a contradiction to the as-
sumption r = k + 1. 

Case h2,0 = 3. Since r = k − 1 we have A(v) = 0 for some 0 6= v ∈

H 2,0 , A ∈ E. Choosing a basis A1 , . . . , Am for the kernel of the map
v → A(v), we let

s = dim{Im A1 + · · · + Im Am } .

This gives an injective map

H 2,0 /Cv → Cs ⊂ Ck .

By the previous case, m 5 k. Clearly m 5 2s since dim(Hom H 2,0 /Cv, Ck ) =

2s. Thus m 5 min(2s, k). Since for all A ∈ E

A(v) ∈ (Im A1 + · · · + Im Am )⊥ ∼
= Ck−s
we have m = r − (k − s). Thus min(2s, k) = r − (k − s) which gives

k − s + min(2s, k) = r

or r 5 3k/2 as desired.
When k = 2s this bound is sharp. To see this, take a subspace
U ⊂ H 1,1 ∼= C2s such that Q |U = 0. Let e1 , e2 , e3 ∈ H 2,0 be a basis,
and take linearly independent A1 , . . . , Ak satisfying

e1 → 0 and e2 , e3 → U .

Then take linearly independent B1 , . . . , Bs satisfying

e1 → U and e2 , e3 → 0 .

It follows that E = span{A1 , . . . , Ak , B1 , . . . , Bs } is an integral element

of dimension k + s = 3k/2.

Example. We take weight n = 3 and Hodge numbers h3,0 = h2,1 = 1.

We will show that in this case dim D = 4 and I is locally equivalent to
the Pfaffian system
(
dy − y 0 dx = 0
(III.11)
dy 0 − y 00 dx = 0
22 J. CARLSON, M. GREEN, AND P. GRIFFITHS 4/19/09

in C4 with coordinates (x, y, y 0 , y 00 ). Since d(dy − y 0 dx) = −dy 0 ∧ dx =

−(dy 0 − y 00 dx) ∧ dx, the 1st derived system of (III.11) has rank one. In
fact, (III.11) is just the 1st prolongation of the contact system (III.6).
Anticipating the later discussion of the period domain D for general
weight n = 3 Hodge structures with h3,0 = 1 and h2,1 = h, locally over
an open set U we consider a holomorphic frame field
e0 , e1 , . . . , eh , e∗1 , . . . , e∗h , e∗0
|{z}
3
|F {z }
F2
| {z }
F1

relative to which
(
Q(e0 , e∗0 ) = 1 = −Q(e∗0 , e0 )
(III.12)
Q(eα , e∗α ) = 1 = −Q(e∗α , eα )
and all other pairings are zero. We observe that
The system I is equivalent to each of the following
(
de0 ≡ 0 mod F 2
Q(de0 , F 2 ) = 0 ,
the equivalence resulting from F 2 = (F 2 )⊥ .
We set, again using summation convention,
(
de0 ≡ θα eα + θα e∗α + θe∗0 mod F 3
(III.13)
deα ≡ θαβ e∗β + θβ∗ e∗0 mod F 2 .
Using these equations and the exterior derivatives of (III.12) we have
(
θαβ = θβα
θα∗ = θα .
We may then conclude
(III.14) The 1-forms θ, θα , θα , θαβ for α 5 β are semi-basic for the
fibering GC → Ď, and over U give bases for the cotangent spaces.
In terms of the Lie algebra description of the tangent space we have
 θ , θαβ ↔ G
 α −1,1

θα ↔ G−2,2
↔ G−3,3 .

θ

4/19/09 VARIATIONS OF HODGE STRUCTURE 23

From the exterior derivatives of (III.13) we obtain

(
dθ = 2θα ∧ θα
(III.15)
dθβ = θα ∧ θαβ .
From the Lie algebra description of the tangent space, it follows that
I is generated by the Pfaffian equations
(
θ=0
(III.16)
θα = 0 .

From this we see that the 1st derived system I[1] is generated by θ, and
the 2nd derived system I[2] = 0.
When h = 1 we are on a 4-dimensional manifold with a rank 2
Pfaffian system I and where rank I[1] = 1, I[2] = 0. It is well known
(Engel’s theorem) and elementary [IL] and [BCGGG, Ch. 2]) that such
a system is locally equivalent to (III.11).
We now return to the general case when h3,0 = 1 and h2,1 = h. Be-
cause of their importance in algebraic geometry, we shall be interested
in integral manifolds S ⊂ D where the map

T S → Hom(H 3,0 , H 2,1 )

is an isomorphism. We shall call these integral manifolds of Calabi-Yau

type.

(III.17) Proposition (cf. [BG1]). The EDS for integral manifolds of

Calabi-Yau type is canonically the 1st prolongation of a contact system.

Thus, locally these integral manifolds depend on one arbitrary function

of h variables.

Proof. We set P = PHC ∼ = P2h+1 . Denoting points of P by homoge-

neous coordinate vectors [z] = [z0 , . . . , z2h+1 ], we consider on HC the
1-form
θ = Q(dz, z) .
Rescaling locally by z → f z where f is a non-vanishing holomorphic
function we have
θ → f 2θ .
24 J. CARLSON, M. GREEN, AND P. GRIFFITHS 4/19/09

It follows that θ induces a 1-form, defined up to scaling, on P . Choosing

coordinates so that
 
0 −1
1 0
 

· ·
 
Q=  ,
 
 · 
0 −1
 

1 0
in the standard coordinate system on P where z0 = 1 we have
X m
θ = dz1 + z2j dz2j+1 − z2j+1 dz2j
j=1
m
X
dθ = 2 dz2j ∧ dz2j+1 ,
j=1

from which it follows that θ induces a contact structure on P .

Now let S ⊂ D be an integral manifold of Calabi-Yau type and
consider the diagram
S ⊂ D
π

P
3
where π(F ) = F viewed as a line in HC . Choose local coordinates
s1 , . . . , sm on S and write π |S as
s → [z(s)] .
Letting Fsp ⊂ HC denote the subspace corresponding to s ∈ S, the
condition dFs3 ⊂ Fs2 = (Fs2 )⊥ gives
π ∗ θ |S = Q(dz(s), z(s)) = 0 ;
i.e., π(S) is an integral manifold of the canonical contact system on P .
In terms of z(s), the integral manifold S is given by
(
Fs3 = [z(s)]
(III.18)
Fs2 = span{z(s), ∂s1 z(s), . . . , ∂sn z(s)} ∈ F (1, h)
where F (1, h) is the manifold of flags F 3 ⊂ F 2 ⊂ HC where dim F 3 = 1,
dim F 2 = h + 1, and where we have used the obvious inclusion Ď ⊂
F (1, h).
4/19/09 VARIATIONS OF HODGE STRUCTURE 25

This process may locally be reversed. Given an h-dimensional inte-

gral manifold s → [z(s)] of the canonical Pfaffian system on P , we may
define an integral manifold of I by (III.18).
Finally, referring to (III.16) we see that what we denoted by θ there
corresponds to the θ just above. From equation (III.15) we may infer
that I is in fact locally just the 1st prolongation of the canonical contact
system on P . 

Remark. In terms of an arbitrary function g(s1 , . . . sh ) of h variables,

in the standard coordinate system the above integral manifold of the
contact system is

(s1 , . . . , sn ) → (g(s), ∂s1 g(s), s1 , . . . ∂sn g(s), sn ) .

It follows that S ⊂ D is given parametrically in terms of g(s) and its

first and second derivatives.

III.B. A Cartan-Kähler example and a brief guide to some of

the literature.

Example. We shall now discuss how the Cartan-Kähler theory applies

to the case of weight n = 2, h2,0 = 3 and h1,1 = h. This is the first
“non-elementary” case.

The following is a purely linear algebra discussion dealing with the

data
• Complex vector spaces P, R of dimension 3, h respectively and
where R has a non-degenerate symmetric form giving an iden-
tification R ∼
= Ř;
• In Hom(P, R) ∼ = P̌ ⊗ R we are looking for an abelian subspace
E, that is a linear subspace of P̌ ⊗ R such that
t
AB − tBA = 0

for all A, B ∈ E. Here, the transpose is relative to the identifi-

cation R ∼ = Ř.
We seek h-dimensional abelian subspaces E that are in general position
relative to the tensor-product structure in the sense that for a general
26 J. CARLSON, M. GREEN, AND P. GRIFFITHS 4/19/09

u ∈ P the composite map

E → Hom(u, R) → R
∩
Hom(P, R)
is an isomorphism. With this assumption we shall be able to establish
that general integral elements that satisfy it are ordinary in the EDS
sense.
It is convenient to choose a basis ei for P and an orthonormal basis
eα for R so that
eiα =: ěα ⊗ ei
gives a basis for Hom(P, R). Denote by
θiα ∈ Hom(P, R)∨
the dual basis. The notation has been chosen to line up with the
“moving frame” notations above. Using the summation convention
and setting
Ωij = θiα ∧ θjα = −Ωji ,
abelian subspaces are given by linear subspaces E on which the restric-
tion
(III.19) Ωij |E = 0 .
We may assume without loss of generality that E is of dimension
h and that the condition of general position is satisfied for u = ei ,
i = 1, 2, 3. Setting Θi = ∧α θiα , we have
(III.20) Θi |E 6= 0 .
Then E is defined by linear equations
θ2 = Aαβ θ1β
( α
det kAαβ k =
6 0
θ3α = Bβα θ1β det kBβα k =
6 0.
The equations (III.19) for Ω12 and Ω13 are (Cartan’s lemma)
(
A = tA
B = tB .
The equation (III.19) for Ω23 is
(III.21) [A, B −1 ] = 0 .
4/19/09 VARIATIONS OF HODGE STRUCTURE 27

In fact, if θ1α = Cβα θ3β on E, then A = CB and C = t C gives (III.21).

(III.22) Proposition. The dimension of the space of generic h-

dimensional abelian subspaces is h(h + 3)/2.

Proof. The group operating on Hom(P, R) is GL(P ) × O(R); this

group preserves the set of equations (III.19) that define the abelian
subspaces of Hom(P, R). For elements I × T in this group, the action
on the matrices A and B above is given by

A → t T AT
B → t T BT .
For A generic we may find T so that A = diag(λ1 , . . . , λh ) where the
λα are non-zero and distinct. The condition (III.21) then gives that
B = diag(µ1 , . . . , µh ). We thus have as a basis for E

vα = e1α + λα e2α + µα e3α

where 1 5 α 5 h.

(III.23) Proposition. The equations that define the polar space H(E)
have rank 2h. Thus E is a maximal integral element.

Proof. For v = C α e1α + Dα e2α + E α e3α the equations that express

the condition that span{E, v} be an integral element are

Ωij (v, vβ ) = 0 .

These compute out to be, with no summation on repeated indices,

β β

 D = λβ C

E β = µβ C β

µβ D β = λβ E β .


These are compatible and have rank 2h. For any solution, we replace
v by ṽ = v − C α vα (summation here), and then ṽ = 0 by the above
equations when all C α = 0. 

(III.24) Proposition. A general abelian subspace E that is in gen-

eral position with respect to the tensor product structure is an ordinary
integral element.
28 J. CARLSON, M. GREEN, AND P. GRIFFITHS 4/19/09

Proof. In order to not have the notation obscure the basic idea, we
shall illustrate the argument in the case h = 2. A general line E1 in E
is spanned by w = ρα vα (summation convention). The polar equations
for v as above are Ωij (w, v) = 0 for i < j. These are
ρ1 (D1 − λ1 C 1 ) + ρ2 (D2 − λ2 C 2 ) = 0



ρ1 (E 1 − µ1 C 1 ) + ρ2 (E 2 − µ2 C 2 ) = 0

 1
ρ (λ1 E 1 − µ1 D1 ) + ρ2 (λ2 E 2 − µ2 D2 ) = 0 .
For a general choice of ρα , λα , µα these are independent. Hence, in the
notation of section (II.B) we have

c0 = 0, c1 = 3 .

On the other hand, the 2-dimensional abelian subspaces in Hom(P, R) ∼

=
6
C is a smooth, 5-dimensional subvariety in the 8-dimensional Grass-
mannian Gr(2, 6). Therefore the codimension of E in Gr(2, 6) is 3 =
8 − 5. Consequently, Cartan’s test is satisfied. 

In [CT] it is shown how to integrate the above system, with the

result

(III.25) Integral manifolds of the above system are parametrized by

generating functions f1 , f2 subject to the PDE system

[Hf1 , Hf2 ] = 0

where Hf is the Hessian matrix of f .

A brief guide to some of the literature. Building on earlier works,

in [M] a number of results are proved:
(i) Under rather general assumptions of the type
• the weight n = 2m + 1 is odd, hm+1,m > 2 and all other
hp,q > 1
• the weight n = 2m is even, hm,m > 4, hm+2,m−2 > 2 and
all other hp,q > 1
a bound on the dimension of integral elements is obtained, and
it is shown that there is a unique integral element E that attains
this bound.
4/19/09 VARIATIONS OF HODGE STRUCTURE 29

(ii) There is a unique germ of integral manifold whose tangent space

is E.
The following interesting question arises:
Is E an ordinary integral element?
This is a question of computing the ranks of the polar equations of a
generic flag in E. Setting dim E = p and dim D = N , since E is unique
we have at E

codim Gp (I) = dim Grass(p, N ) = p(N − p)

and Cartan’s test is whether the inequality

p(N − p) = c1 + · · · + cp−1

is an equality. Suppose that equality does hold, so that E is ordi-

nary and the Cartan-Kähler theorem applies. As explained in [IL]
and [BCGGG] there are Cartan characters s0 , s1 , . . . , sp expressed in
terms of the ci such that local integral manifolds of I depend on sp
arbitrary functions of p-variables, sp−1 arbitrary functions of (p − 1)-
variables,. . . , s0 arbitrary functions of 0-variables (i.e., constants).
Mayer’s result would then follow by showing that equality holds in
Cartan’s test and
sp = · · · = s1 = 0, s0 = 1 .
In fact, if E is ordinary then the Cartan characters must be given in
this way. Another interesting question is
Do the Mayer integral manifolds arise from algebraic
geometry?
Since there is a unique one of these through each point of D, one knows
that in the non-classical case a general Mayer integral manifold does
not arise from geometry. In [CS] it is shown that in weight two with
h1,1 even some maximal integral manifolds are realized geometrically.
Among the works that preceeded [M], and in some cases led up to it
are:
[CT2]: In this paper it was recognized that the weight two horizontal
distribution is a generalization of the contact distribution.
30 J. CARLSON, M. GREEN, AND P. GRIFFITHS 4/19/09

[CD]: Here it is shown that most hypersurface variations are maximal,

i.e. their tangent space at a point is not contained in a larger integral
element. These integral elements are not of maximal dimension in the
sense of [M].
[CKT]: This paper gives the results in weight two that were extended
to the general case in [M].
We conclude this section by discussing the recent work [A].
One may think of an integral element E at F ∈ D as giving an action
of Sym E on ⊕ F p /F p+1 that is compatible with Q; i.e., the action
p
preserves the pairings Q : F p /F p+1 ⊗ F n−p /F n−p+1 → C. Especially if
dim E is not small, such an action will have many algebraic invariants.
Experience suggests that those that arise from geometry will have a
special structure.
One such relates to the notion of symmetrizers due to Donagi (cf. [D]
and [CD]) and used in [A]. Given vector spaces A, B, C and a bilinear
map
Φ:A⊗B →C

one defines

Sym Φ = {Ψ ∈ Hom(A, B) : Φ(a, Ψ(a0 )) = Φ(a0 , Ψ(a)) for a, a0 ∈ A} .

One may consider the Sym’s of the various maps

Symk E ⊗ H p,n−p → H p−k,n−p+k

arising above. In [A] these are studied when dim E and the hp,n−p
are the same as for smooth hypersurfaces X ⊂ Pn+1 where n = 3 and
deg X = n+3, and where the special structure arising from Macauley’s
theorem is satisfied. It is shown that a particular Sym is generically
zero, but is non-zero in the geometric case. Thus the integral elements
arising from hypersurface deformations, which are known to be maxi-
mal, satisfy non-trivial algebraic conditions and are thus non-generic.
We shall give a brief discussion of proofs of these results, based on
the paper [Gr], in the case n = 2. For this we use the notations from
the proof of Theorem (V.10) below. For smooth surfaces in P3 the
4/19/09 VARIATIONS OF HODGE STRUCTURE 31

analogous identification to (V.13) is

H 2−p,p ∼
= V (p+1)d−4 /J(p+1)d−4 .
We also have an identification
E∼
= V d /Jd .
Using E ⊆ Hom(H 2,0 , H 1,1 ) there is a map
2
Hom(H 2,0 , E) → Hom(⊗ H 2,0 , H 1,1 ) .
From the above identifications, we have
V 4 /J4 ⊆ Hom(H 2,0 , E) ∼
= Hom(V d−4 , V 2d−4 /J2d−4 )
and under the above map V 4 /J4 lands in Hom(Sym2 H 2,0 , H 1,1 ). Put
another way
V 4 /J4 ⊆ ker Hom(H 2,0 , E) → Hom(∧2 H 2,0 , E) .


It may be seen that for d = 5 and for E a general integral element,

this kernel is non-zero. This is illustrative of the non-genericity results
in [A].
If follows from [Gr] that for d = 5
V 4 → Ȟ 2,0 ⊗ E → ∧2 Ȟ 2,0 ⊗ H 1,1
is exact at the middle term. Thus one has

Integrability. The composition

H 2,0 → Ě ⊗ H 1,1 → ∧2 Ě ⊗ H 0,2
is zero.

Non-generic condition. The kernel

ker Ȟ 2,0 ⊗ E → ∧2 Ȟ 2,0 ⊗ H 1,1


is non-zero, and as a vector space is isomorphic to V 4 .

One may next proceed, for d = 8, to look at
V d−8 → Hom(V 4 , H 2,0 ) → Hom(∧2 V 4 , E) ,
and by [Gr] this will be exact at the middle term of d = 9. This process
continues and leads to a large number of non-linear constraints on the
32 J. CARLSON, M. GREEN, AND P. GRIFFITHS 4/19/09

integral element arising from variations of a smooth surface of degree

d = 5.
Additional constraints, of a different character, arise as follows. Re-
call that polynomials f0 , f1 , f2 , f3 , where fi ∈ V d , form a regular se-
quence if Var(f0 , f1 , f2 , f3 ) = ∅ in P3 . A special case is when

fi (x) = Fxi (x)

where F (x) = 0 defines a smooth surface in P3 . The symmetrizer

constructions and results in [Gr] illustrated above work equally well for
a regular sequence. Thus, for the integral element defined by a surface
variation, there is the additional constraint that the ideal generated by
a regular sequence be a Jacobian ideal.
Although we shall not give the argument (using, as usual, Macaulay’s
theorem), it may be proved that an integral element E defined by a
regular sequence is maximal.
We set dim E = p and denote by

Gp (surfaces) ⊂ Gp (RS) ⊂ Gp (I)

the integral elements corresponding respectively to smooth surfaces, to

regular sequences in P3 , and general integral elements of the differential
ideal generated by I. As just noted, the first inclusion is strict. The
following would seem to be an interesting question:
Is an integral element E ∈ Gp (RS) ordinary (cf. (II.11)
above)?

III.C. The derived flag of the EDS associated to a VHS. We

shall prove the

(III.26) Theorem. Let D be a period domain for polarized Hodge

structures of weight n = 2 and where all hp,q 6= 0. Denote by I ⊂ T ∗ D
the Pfaffian system given by the infinitesimal period relations. Then
for the derived system
log n
I[m] = 0 for m = .
log 2
Corollary. When n = 2, I has no completely integrable subsystem.
4/19/09 VARIATIONS OF HODGE STRUCTURE 33

Remarks. (i) This corollary roughly means that there are no finite
equations satisfied by integral manifolds of I. More precisely, locally
there are no non-constant functions f on open subsets U of D such that,
for a point F ∈ U, all local integral manifolds of I passing through F
lie in the level set f = constant.
(ii) The theorem actually is valid on the dual classifying space Ď.
Over Ď there are Pfaffian systems I(k) defined for k = 1 by

(III.27) dF p ⊆ F p−k .

For the VHS Pfaffian system, I = I(1). We shall establish a stronger

version of the theorem by showing that:
If all hp,q 6= 0, then

(III.28) I[m] ⊆ I(2m ) .

Moreover, if n = 2 and all hp,q = 2 then equality holds in (III.28).

(iii) We shall also see that the above result is sharp. The first case
where some hp,q = 0 but I 6= T ∗ D occurs when n = 4 and h4,0 6= 0,
h3,1 6= 0 but h2,2 = 0. Then I[∞] 6= 0. The finite equations satisfied
by a VHS may be described as follows: Denote by Gr(f 2 , HC ) the
Grassmanian of f 2 -dimensional planes in HC . Then, since dF 3 ≡ 0
mod F 3 , the connected integral manifolds of I lie in the fibres of the
obvious map
Ď → Gr(f 2 , HC ) .

(iv) The simplest interesting case where equality fails to hold in

(III.28) is when n = 4 and h4,0 = 1. We shall see below that this is
reflected in the structure of the derived flag. The conclusion will be:
Among 4-folds with h4,0 6= 0, Calabi-Yau varieties are distinguished by
the behaviour of the derived flag associated to the infinitesimal period
relation.

Proof. In order not to have the notational complexity obscure the basic
idea, we shall give the argument in the case n = 4. From this we hope
that the pattern for the general case will be evident. The assumption
that all hp,q 6= 0 is equivalent to the positive dimensionality of all
34 J. CARLSON, M. GREEN, AND P. GRIFFITHS 4/19/09

graded quotients in the filtrations

F 4 ⊂ F 3 ⊂ F 2 ⊂ F 1 ⊂ F 0 = HC

corresponding to a point of Ď. We consider adapted frame fields

e0α , e1α , e2α , e3α , e4α

|{z}
4
|F {z }
3
| F {z }
F2
| {z }
F1

satisfying
(
1 if i + j = 4, α = β
(III.29) Q(eiα , ejβ ) =
0 otherwise.
As usual we write, using summation convention,
jβ
deiα = θiα ejβ

where
jβ kγ jβ
(III.30) dθiα = θiα ∧ θkγ .
jβ
The forms θiα for j = i + 1 are semi-basic for the fibering GC → Ď; it
is only expressions in these that will have intrinsic meaning in the final
jβ
formulas. We will use the notation ≡ for congruence modulo the θiα
for j 5 i. The system I is given by
jβ
(III.31) I = {θiα : j = i + 2}

where the brackets are used to denote the span.

The exterior derivative of (III.29) gives
4−jβ 4−iα
(III.32) θiα + θjβ =0.

For later reference we write this out in detail

 4β 4α 3β 3α

 θ0α + θ0β =0 θ1α + θ1β =0

 θ3β + θ4α = 0
 2β
θ1α 3α
+ θ2β =0
0α 1β
(III.33) 2β 4α


 θ0α + θ2β =0

 1β 4α
θ0α + θ3β =0.
4/19/09 VARIATIONS OF HODGE STRUCTURE 35

jβ
The 1-forms θiα are all non-zero and are linearly independent modulo
these relations; the first statement uses the assumption that all hp,q 6= 0.
From (III.30), for j = i + 1
jβ
X lγ jβ
(III.34) dθiα ≡ θiα ∧ θlγ
{ i<l<j
γ

which by (III.31) gives the inclusion

jβ
{θiα }j=i+3 ⊆ I[1] .
Indeed, if j = i + 3 then if i < l < j we must have either l − i = 2,
j − l = 2 or both. Now we use (III.31). Further, from (III.34)
i+2β
X i+1,γ i+2β
dθiα ≡ θiα ∧ θi+1γ .
γ
i+2β
Using (III.33) we may verify that no linear combination of the θiα is
in I[1] . Thus we have equality in the above inclusion; i.e.
n o
3β 4β 4β
θ0α , θ0α , θ1α = I[1] .
We next claim that
I[2] = 0 .
Indeed, denoting by = = congruence modulo the algebraic ideal gener-
ated by I[1] , by (III.34) we have
 3β
dθ = θ1γ ∧ θ1γ
3β 2γ
+ θ0α 3β
∧ θ2γ
 0α = 0α


4β = 2γ 4β

dθ0α = θ0α ∧ θ2γ
 dθ4β = θ3γ ∧ θ4β + θ2γ ∧ θ4β ;

1α = 1α 3γ 2α 2γ

no linear combination of the RHS is =

= 0. 

Denoting by θij the matrix kθiα

jβ
k, the general pattern is
I = {θij : j = i + 1}, i.e. j − i > 1 = 20
I[1] ⊆ {θij : j = i + 3}, i.e. j − i > 2 = 21
I[2] ⊆ {θij : j = i + 5}, i.e. j − i > 4 = 22
..
.
I[k] ⊆ {θij : j − i > 2k } .
36 J. CARLSON, M. GREEN, AND P. GRIFFITHS 4/19/09

To show that equality holds, we must analyze in general the relations

such as (III.33) that arise from Q(F p , F n−p+1 ) = 0. The analysis is
an extension of that given above when n = 4, the one difference being
the alteration of signs with the parity of n. We shall not write out the
argument here. Rather we shall look more closely at the n = 4 case
when h4,0 = 1 and when h2,2 = 0.
4β
Example: h4,0 = 1. In this case θ0α = 0 (only β = α occurs and we
4α
have θ0α = 0 by the first equation in (III.33)), and thus

I ∩ I(4) = 0 .

Conversely, if this relation holds then h4,0 = 1.

Example: h2,2 = 0. In this case

2β jβ
θiα = 0, θ2α =0 i = 0, 1 and j = 3, 4 .

Now
3β kγ 3β
dθ1α ≡ θ1α ∧ θkγ
and thus
n o
3β 4β 3β 4β
I[1] = θ0α , θ0α , θ1α , θ1α .
We claim that

(III.35) I[2] = I[1] .

This will imply that I[∞] = I[1] and will verify the claim that leaves of
the foliation given by I[∞] are the fibres of Ď → Gr(f 2 , HC ).
Continuing to denote by = = equivalence modulo the algebraic ideal
generated by I[1] , from (III.34) we have
3β 1γ 3β =
dθ0α ≡ θ0α ∧ θ1γ = 0
4β 1γ 4β 3γ 4β =
dθ0α ≡ θ0α ∧ θ1γ + θ0α ∧ θ3γ = 0,
3β 4β
and similarly for dθ1α and dθ1α . 

We are not aware of a geometric example where h4,0 =

6 0, h3,1 6= 0
but h2,2 = 0. There are of course geometric examples of necessarily
rigid CY 3-folds where h3,0 = 1, h2,1 = 0.
4/19/09 VARIATIONS OF HODGE STRUCTURE 37

IV. Universal cohomology associated to a homogeneous

Pfaffian system
IV.A. EDS aspects of homogeneous Pfaffian systems. In this
section all Lie groups and Lie algebras will be real and connected. Let
G be a Lie group, V ⊂ G a closed subgroup and
M = G/V
the resulting homogeneous space. We denote by G, v the Lie algebras
respecively of G, V .

Definition. A homogeneous Pfaffian system is given by a G-invariant

sub-bundle I ⊂ T ∗ M .

Equivalently, I ⊥ =: W ⊂ T M is a G-invariant distribution. Since V

is connected, this in turn is given by an ad-v invariant subspace
w ⊂ G/v .
If w̃ ⊂ G is the inverse image of w under the projection of G → G/v,
ad-v invariance is equivalent to [v, w̃] ⊆ w̃.
For any subspace a ⊂ w, the condition
(IV.1) [a, a] ⊆ v
is well-defined; the inverse image ã ⊂ G of a should satisfy [ã, ã] ⊆ v.
(IV.2) Proposition. The G-invariant integral elements of a homoge-
neous Pfaffian system are given by the G-translates of subspaces a ⊂ w
that satisfy (IV.1).

π
Proof. For the projection G → M we have that π ∗ I is a G-invariant
sub-bundle of T ∗ G; it is in fact given by the translates of
i = w⊥ ⊂ Ǧ .
For θ ∈ i, the Maurer-Cartan equation gives, for ζ, η ∈ G,
1
(IV.3) hdθ, ζ ∧ ηi = − hθ, [ζ, η]i .
2
For the G-invariant subspaces E ⊂ T M given by the translates of
a ⊂ w, the condition for such a subspace E to be an integral element
38 J. CARLSON, M. GREEN, AND P. GRIFFITHS 4/19/09

is therefore
dθ |E = 0
for all θ ∈ I. By (IV.3) this is equivalent to (IV.1). 
We now suppose that M is a reductive homogeneous space so that
there is an ad-v invariant splitting
G=p⊕v.
Suppose that a ⊂ p is a subspace that gives an abelian Lie-subalgebra;
i.e.,
(IV.4) [a, a] = 0 .
In particular, (IV.1) is satisfied. Let A = exp a be the connected
subgroup of G corresponding to a.
(IV.5) Proposition (cf. Theorem 3.15 in [M]). The orbit A·(eV ) ⊂ M
is an integral manifold of I.
This is clear: The tangents to the A-orbit are the A translates of a,
and by (IV.3) these are all integral elements. 
The space Gp (I) of p-dimensional integral elements is acted on by
GR , and by the proposition every E ∈ Gp (I) satisfying (IV.4) is tangent
to at least one integral manifold. This does not imply either that (i)
Gp (I) is smooth at E, or (ii) that if Gp (I) is smooth at E, then E is
an ordinary integral element; i.e., I may not be involutive at E.

IV.B. Characteristic cohomology of the homogeneous Pfaffian

system associated to a VHS. Associated to a differential ideal I on
a manifold M are its characteristic cohomology groups HI∗ (M ), defined
to be the de Rham cohomology of the complex (Ω• (M )/I, d) where
Ω• (M ) are the complex valued smooth forms on M . One may think of
HI∗ (M ) as the universal cohomology groups induced on integral mani-
folds of I. We refer to [BG2], [BG3] for a general reference as well as
some examples of these groups.
In the case of a homogeneous Pfaffian system one may consider the
complex ((Ω• (M )/I)G , d) of G-invariant forms. By standard arguments
one has the
4/19/09 VARIATIONS OF HODGE STRUCTURE 39

(IV.6) Proposition. H ∗ ((Ω• (M )/I)G , d) ∼

= H ∗ (G, v; w).
Here, H ∗ (G, v; w) is the Lie algebra cohomology of (G, v) relative to w;
everything is taken over the complex numbers. In the reductive case,
which is the one we shall consider, this is defined as follows: First, we
have
i = w⊥ ⊂ p̌ .
To define the complex C • (G, v; w; δ) we use the map
i → i2 ⊂ ∧2 p̌
defined by θ → θ2 where, for ζ, η ∈ p
hθ2 , ζ ∧ ηi = hθ, [ζ, η]i .
By the Maurer-Cartan equation, up to scaling these are just the 2-
forms dθ where θ ∈ I, that together with I generate the differential
ideal I. Then we set
(IV.7) 

 0 when q = 0
C q (G, v; w; δ) = (p̌/i)v when q = 1

 q q−1 q−2 v
(∧ p̌/ ∧ p̌ ∧ i + ∧ p̌ ∧ i2 ) for q = 2 .
These are just the values at the identity coset eV of the forms in
(Ωq (M )/Iq )G . The map δ is just the exterior derivative, which when
written out has the usual form of the differential in Lie algebra coho-
mology.
In applications, one frequently has a discrete subgroup Γ acting prop-
erly discontinuously on M , and one is interested in the global invariants
of maps
f : S → Γ\M .
There is then an induced map
(IV.8) f ∗ : H ∗ (G, v; w) → H ∗ (S) .
We may think of the image of this map as giving the global topological
invariants of maps f : S → Γ\M that may be defined universally for
all Γ’s.
We now turn to the case of a period domain and use the complexified
Lie algebras. I will be the differential ideal generated by I ⊂ T 1,0 D
40 J. CARLSON, M. GREEN, AND P. GRIFFITHS 4/19/09

and I¯ ⊂ T 0,1 D. In the classical case when n = 1, w = 0 and V ⊂ GR

is a maximal compact subgroup, and from what is known it may be
inferred that

(IV.9) H ∗ (G, v; δ) ∼
= (∧∗ p̌)v is the space of invariants, and these are
generated by the Chern forms of the Hodge bundles.

The argument will be given in several steps, some of which will carry
over to the higher weight case.
(i) Since D is a symmetric space, the symmetry about eV is given
by an element in V which acts by −1 in the tangent space
there. It follows that the invariant forms are all of even degree.
In particular δ = 0, which gives the first statement in (IV.9).
(ii) Next, using the unitary trick, and averaging, one infers that the
space of invariant forms is isomorphic to the space of harmonic
forms on the compact dual Ď, and this space is in turn isomor-
phic to H ∗ (Ď). This latter is known to be generated by the
Chern classes of the universal vector bundle over Ď.
We will now show

(IV.10) Theorem. For a period domain, the invariant forms that are
orthogonal to the algebraic ideal generated by I and I¯ are all of type
(p, p).

Here, orthogonal means with respect to the canonical metrics con-

structed from the Hodge metrics in the Hodge bundles. Working in
the orthogonal space is equivalent to working modulo the ideal in the
space of all forms.
The first statement means that for a period domain the forms
h
⊥ iv
∧∗−1 p̌ ∧ iC

are all of type (p, p). This does not yet take into account the forms in
∧∗−2 p̌ ∧ i2 .

Proof. At a reference point F ∈ D, we have a canonical representation

ϕ : S 1 → Aut(HR , Q) = GR
4/19/09 VARIATIONS OF HODGE STRUCTURE 41

obtained from a representation ϕC : C∗ → Aut(HC , Q) defined by

ϕC (z)v = z p z̄ q v v ∈ H p,q .
Restricted to S 1 = {|z| = 1}, ϕC (z) = ϕ(z) is real. We observe that
√
S 1 is a subgroup of V ; the Weil operator is just C = ϕ( −1).
On the Lie algebra G, we have the induced Hodge structure
GC = ⊕ G−i,i .
i

We then have
vC = G0,0
and
pC = p+ ⊕ p− , p− = p̄+
where 
 p+ = ⊕ G−i,i ∼
(1,0)
 = TF D
i>0

 p− = ⊕ G−i,i ∼
(0,1)
 = TF D .
i<0
For the canonical Pfaffian system we have
iC = i+ ⊕ i− , i− = ī+
where
i+ = Ǧ−1,1 ⊂ ŤF1,0 D
(

i− = Ǧ1,−1 ⊂ ŤF0,1 D .
Thus the orthogonal complement to ∧∗−1 p̌C ∧iC are the forms of positive
degree in
∧∗ Ǧ−1,1 ⊕ Ǧ1,−1 ∼= ∧∗ Ǧ−1,1 ⊗ ∧∗ Ǧ1,−1 .

On G−1,1 , Ad ϕ(z) acts by z −1 z̄ = z 2 ; on G1,−1 it acts by z z̄ −1 = z −2 .

Thus on ∧p Ǧ−1,1 ⊗ ∧q Ǧ1,−1 , Ad ϕ(z) acts by z 2(p−q) . Since ϕ(z) ∈ V for
z ∈ S 1 , we see that any ad v-invariant forms must be of type (p, p). 
As mentioned in the introduction, we expect that (IV.10) may be the
first step in showing that the GR -invariant forms modulo the differential
ideal generated by I are generated by the Chern forms of the Hodge
bundles.7 Since, except in the classical case when I = 0, the invariant
7As mentioned in the introduction, this result has now been proved. The inte-
grability conditions play the central role in the argument, which is representation-
theoretic in nature.
42 J. CARLSON, M. GREEN, AND P. GRIFFITHS 4/19/09

forms modulo the algebraic ideal generated by I definitely contain more

than the Chern forms, the integrability conditions will have to enter in
an essential way. In fact, they enter in the proof of the following
(IV.11) Theorem. The Chern forms of the Hodge filtration bundles
satisfy
ci (F p )cj (F n−p ) = 0 if i + j > hp,n−p .

Proof. The argument will proceed in several steps, the key one where
integrability is used being (IV.17) below.
Step one. For a d × d matrix A we set
χA (t) =: det(A − tI) = td + c1 (A)td−1 + · · · + cd (A) .
Then we have
(IV.12) Lemma: If A, B are d × d matrices with AB = 0, then
ci (A)cj (B) = 0 if i + j > d .

Proof. We have
χA (t)χB (u) = det(A − tI) det(B − uI)
= det(−tB − uA + utI) ,
and all terms in det(−tB − uA + utI) have degree at least d in u, t.
From
X
χA (t)χB (u) = ci (A)cj (B)td−i ud−j
i,j

we have ci (A)cj (B) = 0 if (d − i) + (d − j) < d; i.e. if i + j > d. 

Step two. We let E ⊂ G−1,1 be an integral element at a point F ∈ D.

Then we have Ě-valued maps
Ap
F p /F p+1 −−→ F p−1 /F p ,
where we think of Ap as a matrix of size hp,n−p × hp−1,n−p+1 whose
entries are Ě. The integrability conditions are
(IV.13) Ap−1 · Ap = 0 ,
where we are multiplying matrices using the wedge product of their
entries.
4/19/09 VARIATIONS OF HODGE STRUCTURE 43

Using the dualities induced by Q and denoting the transpose of a

matrix A by A∗ we have a commutative diagram
A∗p
P P
p−1 p
(F /F ) −−−−→ (F /F p+1 )
p

ok ok
An−p+1
F n−p+1 /F n−p+2 −−−−→ F n−p /F n−p+1 .
On F p we have an Hermitian metric induced by (v, w) = Q(Cv, w̄).

Lemma. Up to a constant, the curvature matrix of the metric connec-

tion is given by
!
Ā∗p Ap 0
(IV.14) ΘF p = .
0 0
The notation means that we write the orthogonal direct sum decom-
position
F p = H p,n−p ⊕ F p+1 .
A consequence of (IV.14) is
(IV.15) ck (F p ) = 0 for k > hp,n−p .
Proof of the lemma. F p is a sub-bundle of the flat bundle F 0 = HC
with the induced metric. In this situation it is known [G] that the
curvature of the metric connection in F p is up to a constant given by
∗
ΠF p /H · ΠF p /H
where ΠF p /H is the 2nd fundamental form of F p in HC . In the case at
hand the 2nd fundamental form may be identified with Ap . 

Remark. There is one subtlety here. Because of the sign alternation

in the Hodge metrics
√
(u, v) = ( −1)p−q Q(u, v̄) u, v ∈ H p,q
the usual principle that “curvatures decrease on holomorphic sub-bundles”
does not hold for the Hodge bundles. For example, c1 (F n ) > 0 on ΘF p .
However, the signs are not an issue for us here.

Step three. From (IV.13) we have

(IV.16) Ā∗p Ap Ap+1 Ā∗p+1 = 0
44 J. CARLSON, M. GREEN, AND P. GRIFFITHS 4/19/09

where the multiplication of matrix entries is wedge product. For nota-

tional simplicity we omit the blocks of zeroes in the ΘF p ’s, so that

ΘF p = Ā∗p Ap
Θ∗F n−p = −Ap+1 Ā∗p+1 .

Using (IV.16) this gives the remarkable consequence

(IV.17) ΘF p Θ∗F n−p = 0

of integrability. Since χΘ∗ n−p (t) = χΘF n−p (t), by step one we see that
F
ci (F p )cj (F n−p ) = 0 if i + j > hp,n−p . 

V. “Expected” dimension counts for integral manifolds

of an EDS

An important aspect in algebraic geometry is that of “expected”

dimension counts. Informally and in first approximation, this means
counting the number of parameters of solutions to a system of algebraic
equations, where “expected” means assuming some sort of “general
position.” When the solution varieties are also subject to differential
constraints, the problem changes character in an interesting way. In
this section we will discuss this for the EDS arising from VHS’s.
To frame the general issue we assume given a diagram of regular
mappings of complex manifolds
f
X → A
(V.1) ∪ ∪
Y → B
where f is an immersion and Y = f −1 (B). Then

“expected” codimX Y = codimA B

= rank(T A/T B) ,

where it is understood that T A/T B is restricted to B. If Y is non-

empty, then the actual codimension satisfies

(V.2) codimX Y 5 rank(T A/T B) ,

with equality holding when f (X) meets B transversely.

4/19/09 VARIATIONS OF HODGE STRUCTURE 45

Now suppose that I ⊂ T ∗ A is a holomorphic Pfaffian system and

f : X → A is an integral manifold of I. Let W = I ⊥ be the corre-
sponding distribution. Then for the normal bundles we have that
f∗ : T X/T Y → W/W ∩ T B
is injective, so that the above may be refined to
(V.3) codimX Y 5 rank(W/W ∩ T B) 5 rank(T A/T B) .
Informally we may say that: Subjecting f : X → A to a differen-
tial constraint decreases the codimension of Y = f −1 (A) in X. By
“decreases” we mean that codimX Y is less than the “expected” codi-
mension in the absence of differential constraints, as explained above.
However, when we take into account integrability a still further re-
finement of (V.3) may be expected. This is because in general integral
elements of I may be expected to have dimension smaller, frequently
much less, than rank W .
Rather than discuss the general aspects of this, we turn to a special
case that is geometrically motivated. Let D be a period domain for
polarized Hodge structures of even weight n = 2m. At a reference
Hodge structure F ∈ D we let ζ ∈ HR ∩ H m,m be a real vector of type
(m, m).

Definition. The Noether-Lefschetz locus is the subvariety Dζ ⊂ D

where ζ remains of type (m, m).

(V.4) Proposition. Let Gζ ⊂ GR be the subgroup fixing ζ up to

scaling. Then Dζ = Gζ · F is the Gζ -orbit of F . It is a homogeneous
complex sub-manifold of D of codimension given by
codimD Dζ = h(2m,0) + · · · + h(m+1,m−1) .

Proof. This is a matter of routine checking. Setting hk = h(2m−k,k) , we

have first
GR ∼

= O(a, b), a + b = h0 + · · · + h2m
V ∼
= U(h 0 ) × · · · × U(hm−1 ) × O(hm ) .
Next, depending on whether m is even or odd, we have Gζ ∼ = O(a−1, b)
or O(a, b − 1). Finally, the same linear algebra argument that shows
46 J. CARLSON, M. GREEN, AND P. GRIFFITHS 4/19/09

that GR acts transitively on D shows that Gζ acts transitively on Dζ

and
Dζ = Gζ /Vζ
where

Vζ = Gζ ∩ V ∼
= U(h0 ) × · · · × U(hm−1 ) × O(hm − 1) .
From this we may conclude the above codimension count. 

For fixed ζ ∈ HR ∩ H (m,m) and at a variable point in D we write the

Hodge decomposition of ζ as
(
ζ = ζ 2m,0 + · · · + ζ m,m + · · · + ζ 0,2m
ζ 2m−p,p = ζ p,2m−p .
Then Dζ is defined by the equations

ζ m−1,m+1 = · · · = ζ 0,2m = 0 .

The above proposition says that these equations are independent and
define Dζ as a smooth complex submanifold of D.
At a point F ∈ D we let
(
E ⊂ TF D be an integral element of I
Eζ = E ∩ TF Dζ .
For ϕ ∈ G−1,1 ⊂ ⊕ Hom(H 2m−p,p , H 2m−p−1,p+1 ) we write ϕ = ϕ0 + · · · +
ϕ2m−1 where ϕp ∈ Hom(H 2m−p,p , H 2m−p−1,p+1 ) and ϕp and ϕ2m−p−1 are
dual. Then

Eζ = {ϕ ∈ E : ϕm (ζ) = 0 in H m−1,m+1 } .

This is equivalent to

(V.5) Eζ = {ϕ ∈ E : Q(η, ϕ(ζ)) = 0 for all η ∈ H m+1,m−1 } .

Thus, without taking the integrability conditions into account we have

codimE Eζ 5 hm+1,m−1 .

However, due to the integrability conditions the equations (V.5) may

not be independent. In order to illustrate the essential point, we begin
by considering the first non-trivial case m = 2. For any ϕ ∈ Eζ , ψ ∈ E
4/19/09 VARIATIONS OF HODGE STRUCTURE 47

and ω ∈ H 4,0 , using that E is an integral element so that ϕ and ψ

commute,
Q(ϕ(ω), ψ(ζ)) = −Q(ψϕ(ω), ζ)
= −Q(ϕψ(ω), ζ)
= Q(ψ(ω), ϕ(ζ))
= 0;
i.e., for each ψ ∈ E the linear equations
Q(η, ψ(ζ)) = 0 η ∈ H 3,1
that define the condition that ψ ∈ Eζ are decreased in rank by
(V.6) σζ =: dim Im{Eζ ⊗ H 4,0 → H 3,1 } .
We thus have the
(V.7) Proposition. For σζ defined as above
codimE Eζ 5 h1,3 − σζ .
The general case goes as follows: With Eζ defined as above, for each
p with 0 5 p 5 m − 2 we consider the maps
κpζ : E ⊗ Symp−m+1 E ⊗ H 2m−p,p → H m+1,m−1
and we set
σζ = dim span(Images κpζ ) for 0 5 p 5 m − 2

(V.8) .
Then the straightforward extension of the above argument gives the
(V.9) Proposition. For σζ defined by (V.8), we have
codimE Eζ 5 hm−1,m+1 − σζ .
In algebro-geometric terms, this says that the “richer” the multi-
plicative structure in the 1st order variation of the Hodge structure, the
smaller the codimension of Noether-Lefschetz loci.

Remark. The above is predicated on the implicit assumption that, for

a variation of Hodge structure S → Γ\D, we have that the Noether-
Lefschetz locus Sζ ⊂ S is reduced, so that for general points of a
component of Sζ we have codimS (Sζ ) = codimTs S (Ts Sζ ). For far as we
48 J. CARLSON, M. GREEN, AND P. GRIFFITHS 4/19/09

know, there are not yet any examples coming from algebraic geome-
try where this assumption is not satisfied, although we feel that such
examples may be expected.

Example. We consider a smooth hypersurface

X ⊂ P5
of degree d = 6 and which contains a 2-plane P ∼
= P2 .

(V.10) Theorem. For the primitive part of the fundamental class

[P ] ∈ H 4 (X, Z) of P , we have equality in (V.9).

Proof. We will denote by V ∼ = C6 a vector space with coordinates

x1 , . . . , x6 such that X ⊂ PV̌ is given by an equation
F (x) = 0
where F ∈ V d =: Symd V is a homogeneous polynomial of degree d = 6.
It is well-known, and will be proved below, that
(V.11) At a general X containing a 2-plane P , the locus of all degree
d hypersurfaces X 0 close to X and containing a 2-plane P 0 close to P
is smooth and of codimension (d+1)(d+2)
2
− dim Gr(3, 6) in the space of
all degree d hypersurfaces in PV̌ .
Let P be given by x1 = x2 = x3 = 0 so that
3
X
F (x) = xi Gi (x)
i=1

where Gi ∈ V d−1 . Denote by

Ft (x) = F (x) + tḞ (x), Ḟ ∈ V d
a 1st order perturbation of F . The condition that P move to 1st order
to a 2-plane Pt ⊂ Xt = {Ft (x) = 0} is that
X X
Ḟ = xi Ġi + li Gi
i i
d−1 st
where Ġi ∈ V is the tangent to a 1 order variation of Gi and the
li ∈ V are linear forms. We will use the notation (X 0 , P 0 ) for the 1st
order perturbation of (X, P ).
4/19/09 VARIATIONS OF HODGE STRUCTURE 49

We will denote by (H1 , . . . , Hm ) the ideal generated by forms H1 , . . . ,

Hm , and by (H1 , . . . , Hm )k the degree k part of that ideal. Thus, the
subspace of V d that gives the 1st order deformations (X 0 , P 0 ) of (X, P )
is
(x1 , x2 , x3 , G1 , G2 , G3 )d .

The Jacobian ideal is

J = (∂x1 F, . . . , ∂x6 F ) .

For references to Jacobian ideals and the polynomial description of

the cohomology of hypersurfaces we suggest [A], [CD], and [D]. The
tangent space to 1st order deformations of the projective equivalence
class of X is given by
T = V d /Jd .

With this identification, by what was said above the subspace TP ⊂ T

giving 1st order deformations of equivalence classes of pairs (X, P ) is
given by

(V.12) TP = (x1 , x2 , x3 , G1 , G2 , G3 )d /Jd .

A basic identification is

(V.13) H 4−p,p (X)prim = V (p+1)d−6 /J(p+1)d−6 .

Thus, for example

H 4,0 (X) ∼
= V d−6 .
Using the identification T = V d /Jd , the differential of the period map-
ping
T → ⊕ Hom(H 4−p,p (X), H 4−p−1,p+1 (X))
p

is given by multiplication in the ring V • /J• ; i.e. by

(V.14) V d /Jd ⊗ V (p+1)d−6 /J(p+1)d−6 → V (p+2)d−6 /J(p+2)d−6 .

We denote by ζ ∈ H 2,2 (X)prim the primitive part of the fundamental

class [P ] of P .
50 J. CARLSON, M. GREEN, AND P. GRIFFITHS 4/19/09

(V.15) Proposition. If we define

Tζ = {H ∈ V d /Jd : H · ζ = 0 in V 4d−6 /J4d−6 }

then
Tζ = TP .

The inclusion TP ⊆ Tζ is clear geometrically: It means that if (X, P )

deforms to 1st order to (X 0 , P 0 ) as above, then ζ deforms to a Hodge
class ζ 0 ∈ H 2 (X 0 )prim . In fact, ζ 0 = [P 0 ]prim . The proof of the reverse
inclusion will come out indirectly from the argument to be given below.
The assertion (V.11) is a consequence of (V.15).
We set
VP = V |P = V /(x1 , x2 , x3 ) .

The proof of the theorem and the proposition will be based on known
commutative algebra properties of the rings V • /J• and VP• /JP,• ; namely
there are perfect pairings
(i) V k /Jk ⊗ V 6d−12−k /J6d−12−k → V 6d−12 /J6d−12 ∼


 =C

(V.16) (ii) VPk /(G1 , G2 , G3 )P,k ⊗ VP3d−6−k /(G1 , G2 , G3 )P,3d−6−k

→ VP3d−6 /(G1 , G2 , G3 )P,3d−6 ∼

=C.


Here, we denote by (G1 , G2 , G3 )P,k the degree k part of the ideal gener-
ated by the Gi |P . The reason for (V.16) is that (x1 , x2 , x3 , G1 , G2 , G3 )
is a regular sequence on P5 , and this then implies that G1 |P , G2 |P , G3 |P
is a regular sequence on P . Then in general if f1 , . . . , fn is a reg-
ular sequence on Pn where deg fi = di , there is a perfect pairing
V a /(f1 , . . . , fn )a
⊗ V b /(f1 , . . . , fn )b → V a+b /(f1 , . . . , fn ) where a + b = i di − n.
P

The argument will proceed in five steps, the first of which is (V.13)
above.

Step two: We have

TP H 4,0 = (x1 , x2 , x3 , G1 , G2 , G3 )2d−6 /J2d−6 ⊆ V 2d−6 /J2d−6 .

This follows from (V.13) and (V.14) in the case p = 0.

4/19/09 VARIATIONS OF HODGE STRUCTURE 51

Step three. Denoting by [P ] the fundamental class of P , the map

[P ]
H 2,2 (X) → H 4,4 (X) ∼
=C

may, using (ii) in (V.14) which gives an isomorphism V 3d−6 /(G1 , G2 , G3 )P,3d−6
∼
= C, be identified with

V 3d−6 /J3d−6 → VP3d−6 /(G1 , G2 , G3 )P,3d−6 ∼

=C.

Proof. Since X is non-singular the restrictions Gi |P have no common

zeroes and hence form a regular sequence. The corresponding Koszul
resolution of OP then gives

0 → OP (−3(d−1)) → ⊕ OP (−2(d−1)) → ⊕ OP (−(d−1)) → OP → 0 .

3 3

Tensoring with OP (3d − 6) gives, using Ω2P ∼

= OP (−3)

0 → Ω2P → ⊕ OP (d − 4) → ⊕ OP (2d − 5) → OP (3d − 6) → 0 ,

3 3

from which we infer that

VP3d−6 /(G1 , G2 , G3 )P,3d−6 ∼

= H 2 (P, Ω2P ) ∼
=C.

Moreover, under this isomorphism a generator of VP3d−6 /(G1 , G2 , G3 )P,3d−6

[P ]
maps to the fundamental class. Since the map H 2,2 (X) → H 4,4 (X) is
given by evaluating a class in H 2,2 (X) on the fundamental class of P ,
by standard arguments we may infer the assertion in Step 3.

Step four. We first observe that the map

(V.17) ζ ⊗ V d /Jd ⊗ TP ⊗ V d−6 → V 4d−6 /J4d−6 ∼

=C

is zero. Here, we recall that ζ ∈ V 3d−6 /J3d−6 is the primitive part of

[P ] ∈ H 2,2 (X). We are using (V.14) that the action on cohomology of
tangent vectors to deformations of equivalence classes of hypersurfaces
in P5 is given by multiplication in the ring V • /J• . The fact that the
above map is zero results from the definition of Tζ as the kernel of the
map
ζ
T −−→ H 1,3
ok ok
V d /Jd −−→ V 4d−6 /J4d−6 .
52 J. CARLSON, M. GREEN, AND P. GRIFFITHS 4/19/09

Then the claim is that the map (V.17) may be identified with
(V.18) TP · V d−6 |P ∈ VP3d−6 /(G1 , G2 , G3 )P,3d−6 ∼
=C.
This follows from Step three above.
Step five. We now put everything together. To prove the theorem it
will suffice to show that TP is the kernel of
ζ
V d /Jd → V 4d−6 /J4d−6 .
This is because for all S ∈ V d /Jd ,
(V.19) R · ζ = 0 in V 6d−6 /J6d−6 for all R ∈ V 2d−6
is, by (V.16)(i), equivalent to
S · ζ = 0 in V 4d−6 /J4d−6 .
By step four, (V.19) is in turn equivalent to
RS |P ∈ (G1 , G2 , G3 )P,3d−6 ,
which by (V.16)(ii) is the same as the condition
S |P ∈ (G1 , G2 , G3 )P,d .
This last statement is easily seen to be equivalent to

S ∈ TP . 

Reprise. We consider the case that arises in the case of a family of

Calabi-Yau fourfolds. Thus, we assume that h4,0 = 1 and denote by
ω ∈ H 4,0 a generator giving an isomorphism H 4,0 ∼= C. Moreover,
let T ⊂ G −1,1
be an integral element and, as would be the case for
Calabi-Yau’s, we assume that the map given by (V.13)
T → Hom(H 4,0 , H 3,1 ) ∼
= H 3,1
is an isomorphism. Using this map we may identify T with H 3,1 and
denote its dimension by m (for moduli).
Each ζ ∈ H 2,2 defines a quadric
Qζ ∈ Sym2 Ť
given for θ, θ0 ∈ T by
hQζ , θ · θ0 i = Q(θ · θ0 ω, ζ) .
4/19/09 VARIATIONS OF HODGE STRUCTURE 53

The fact that this is symmetric in θ and θ0 is because T is an integral

element.
Denote by Tζ ⊂ T the intersection of T with the tangent space to
the Noether-Lefschetz locus Dζ ⊂ D.
(V.20) Proposition. We have
codim Tζ = rank Qζ .
Moreover, in the case of the family of hypersurfaces
5
X ⊂ P of degree six and where ζ is the primitive part of the fun-
damental class of a plane P ⊂ X, we have TP = Tζ .

Proof. We view Qζ as a map

Qζ : T → Ť .
Then the proof of Propostion (V.7) gives
Tζ = (Im Qζ )⊥ ;
thus
codim Tζ = rank Qζ
as desired. 
For the case of P ⊂ X ⊂ P5 , deg X = 6, we have
T = V 6 /J6 .
Moreover, we have seen in step three in the proof of (V.15) that
Tζ = ker V 6 /J6 → VP6 /(G1 , G2 , G3 )P,6 .


Thus
rank Qζ = dim VP6 /(G1 , G2 , G3 )P,6 .
Now, from the Koszul calculation above
(G1 ,G2 ,G3 )
0 → ⊕ VP −−−−→ VP6
3
is exact. Thus
rank Qζ = dim VP6 − 3 dim VP
 
8
= −3·6
2
= 19 .
54 J. CARLSON, M. GREEN, AND P. GRIFFITHS 4/19/09

On the other hand, the number of conditions for X to contain a 2-plane

is  
6 8
dim VP − dim Gr(3, 6) = − 9 = 19 .
2

Remark. Let X be a Calabi-Yau fourfold and ζ ∈ Hg2 (X)prim a Hodge

class. We then have
(V.21) If the Hodge conjecture is true and rank Qζ = h3,1 is maximal,
then X is defined over a number field.
This is because if Qζ is non-singular, then the Noether-Lefschetz locus
for ζ will be 0-dimensional. If ζ = [Z] is the class of an algebraic cycle,
then X is defined over a field k of transcendence degree = 1, and by
standard arguments we may, after passing to a finite field extension,
assume that Z is also defined over k. The spread of (X, Z) will then
give a positive dimensional component to the Noether-Lefschetz locus
of ζ.
To disprove the consequence of the Hodge conjecture that Hodge
classes are absolute, it would be sufficient to find a Calabi-Yau fourfold
not defined over a number field and a Hodge class ζ such that Qζ is
non-singular.

References
[A] E. Allaud, Nongenericity of variations of Hodge structure for hy-
persurfaces of high degree, Duke Math. J. 129 (2005), 201–217;
arXiv:math.AG0503346; MR 2165541.
[BCGGG] R. Bryant, S. S. Chern, R. Gardner, H. Goldschmidt, and P. Griffiths,
Exterior differential systems, Math. Sci. Res. Inst. Publ. 18, Springer-
Verlag, New York, 1991, viii+475 pp.
[BG1] R. Bryant and P. Griffiths, Some observations on the infinitesimal pe-
riod relations for regular threefolds with trivial canonical bundle, Arith-
metic and Geometry, Vol. II, 77–102, Progr. Math. 36 (1983), Birkhäuser
Boston, Boston, MA.
[BG2] , Characteristic cohomology of differential systems, I. General
theory, J. Amer. Math. Soc. 8 (1995), 507–596.
[BG3] , Characteristic cohomology of differential systems, II. Conser-
vation law for a class of parabolic equations, Duke Math. J. 78 (1995),
531–676.
[C] J. Carlson, Bounds on the dimension of variations of Hodge structure,
Trans. Amer. Math. Soc. 294 (1986), 45–64; Erratum, Trans. Amer.
Math. Soc. 299 (1987), 429.
4/19/09 VARIATIONS OF HODGE STRUCTURE 55

[CD] J. Carlson and R. Donagi, Hypersurface variations are maximal, I, In-

vent. Math. 89 (1987), 371–374.
[CKT] J. Carlson, A. Kasparian, and D. Toledo, Variations of Hodge structure
of maximal dimension, Duke Math. J. 58 (1989), 669–694.
[CS] J. Carlson and C. Simpson, Shimura varieties of weight two Hodge struc-
tures, Lect. Notes Math. 1246, 1–15, Springer-Verlag, Berlin, 1987.
[CT] J. Carlson and D. Toledo, Generic integral manifolds for weight-two
period domains, Trans. Amer. Math. Soc. 356 (2004), 2241–2249.
[CT2] , Variations of Hodge structure, Legendre submanifolds and ac-
cessibility, Trans. Amer. Math. Soc. 311 (1989), 391–411.
[C-MS-P] J. Carlson, S. Mueller-Stoch, and C. Peters, Geometry of the Period
Map, Cambridge Univ. Press, 2003, 559pp.
[D] R. Donagi, Generic Torelli for projective hypersurfaces, Comp. Math.
50 (1983), 325–353.
[Gr] M. Green, Koszul cohomology and the geometry of projective varieties.
II. J. Diff. Geom. 20 (1984), 279–289.
[GG] M. Green and P. Griffiths, Algebraic cycles and singularities of normal
functions, II, Inspired by S. S. Chern, Nankai Tracts Math. 11 (2006),
179–268, World Sci. Publ., Hackensack, NJ.
[G] P. Griffiths, Hermitian differential geometry and the theory of positive
and ample holomorphic vector bundles, J. Math. Mech. 14 (1965), 117–
140.
[IL] T. A. Ivey and J. M. Landsberg, Cartan for Beginners: Differential
Geometry via Moving Frames and Exterior Differential Systems, Grad.
Stud. Math. 61, A.M.S., Providence, RI, 2003.
[M] R. Mayer, Coupled contact systems and rigidity of maximal dimensional
variations of Hodge structure, Trans. Amer. Math. Soc. 352 (2000),
2121–2144.
[O1] A. Otwinowska, Composantes de petite codmension du lieu de Noether-
Lefschetz: un argument asymptotique en faveur de la conjecture de
Hodge pour les hypersurfaces, J. Algebraic Geom. 12 (2003), 307–320.
[O2] , Composantes de dimension maximale d’un analogue du lieu de
Noether-Lefschetz, Compositio Math. 131 (2002), 31–50.
[V] C. Voisin, Hodge loci and absolute Hodge classes, Compos. Math. 143
(2007), 945–958.

Clay Mathematics Institute, UCLA, and the Institute for Advanced

Study,