You are on page 1of 24

Moduli spaces of Lie algebras and foliations

Sebastián Velazquez*

Abstract
arXiv:2209.01752v1 [math.AG] 5 Sep 2022

Let X be a smooth projective variety over the complex numbers and S(d) the
scheme parametrizing d-dimensional Lie subalgebras of H 0 (X, T X). This article
is dedicated to the study of the geometry of the moduli space Inv of involutive
distributions on X around the points F ∈ Inv which are induced by Lie group
actions. For every g ∈ S(d) one can consider the corresponding element F (g) ∈ Inv,
whose generic leaf coincides with an orbit of the action of exp(g) on X. We show that
`
under mild hypotheses, after taking a stratification i S(d)i → S(d) this assignment
`
yields an isomorphism φ : i S(d)i → Inv locally around g and F (g). This gives
a common explanation for many results appearing independently in the literature.
We also construct new stable families of foliations which are induced by Lie group
actions.

1 Introduction
Let X be a smooth projective variety of dimension n. A singular holomorphic foliation
F on X is said to be an L-foliation if there exists a complex Lie group G acting on X
with orbits of constant dimension in codimension 1 whose generic orbit coincides with
the generic leaf of F. Observe that such a foliation is uniquely determined by the finite-
dimensional Lie algebra H 0 (X, T F) of global vector fields tangent to it. These foliations
have played a central role in the study of singular foliations and their moduli spaces.
For instance, in the literature there are many results regarding the stability of different
families of L-foliations, i.e., showing irreducible components of moduli spaces of foliations
whose generic point is an L-foliation, see for example [2], [4], [5] and [10] among others.
In this work we present a common framework for these phenomenons and generalize
many of them to other classes of varieties and group actions. We will do so by means of
comparing the moduli problem associated to the subalgebra H 0 (X, T F) ⊆ H 0 (X, T X)
and the one corresponding to T F ⊆ T X as an involutive coherent sheaf.
We will now give an overview of our approach. Let us use the notation L =
0
H (X, T X) for the algebra of global vector fields on X. First, observe that every L-
foliation F is induced by the action of the connected Lie group exp(g) ⊆ Aut(X) for
g = H 0 (X, T F). Conversely, for an arbitrary Lie subalgebra g ⊆ H 0 (X, T X) acting in-
finitesimally on X with orbits of constant dimension in codimension 1 we will denote by

1
2

F(g) the corresponding foliation on X. We will say that a subalgebra g ⊆ L is maximal


if the inclusion g ⊆ H 0 (X, T F(g)) is an equality.

Definition. A family of codimension q foliations on X over a scheme S is an exact


sequence
F : 0 → T F → TS (X × S) → NF → 0
such that T F is closed under the Lie bracket of relative vector fields and its normal
sheaf NF is torsion-free on the fibers and of rank q. The family is said to be flat if NF
is flat over S.

Let Inv ⊆ Quot(T X) be the locally closed subscheme representing the functor assign-
ing to each scheme S the set of flat families of foliations on X over S (for more details see
Definition 23). This space was introduced in [13] and [14] independently. The problem of
classifying foliations and their flat families corresponds to understanding the geometry
of this scheme, which is our main object of study. Let us also denote Inv ⊆ Quot(T X)
the subscheme representing the functor of flat families of coherent involutive subsheaves
of T X (this is, families whose normal sheaf may not be torsion-free).

Definition. A family of d-dimensional Lie subalgebras of vector fields with base S is a


locally free subsheaf G ⊆ L⊗C OS such that [G, G] ⊆ G, where [ , ] stands for the OS -linear
extension of the Lie bracket on L.

Proposition 13 shows that the corresponding functor of d-dimensional Lie subalgebras


of L is represented by a subvariety S(d) of the Grassmanian Gr(L, d). A study of some
local aspects of its geometry can be found in [15] and the references within. Of course,
S(d) is equipped with a universal family Gd of d-dimensional subalgebras of L. Now let

Fd : 0 → T Fd → TS (X × S) → NFd → 0

be the family over S(d) such that the restriction (Fd )g to the fiber Xg over a generic
point g ∈ S(d) satisfies (Fd )g = F(g). Since this family may not be flat, we shall consider
the flattening stratification a
S(d)i → S(d)
i

associated to the sheaf Nd . We thus get over each stratum S(d)i a flat family Fi :=
F|S(d)i and therefore a canonical morphism φi : S(d)i → Inv. It is worth mentioning
that these schemes carry natural actions of Aut(X) and φi turns out to be Aut(X)-
equivariant. An element is said to be rigid if its orbit contains a non-empty Zariski
open. We will denote by a
φ: S(d)i → Inv
i
3

the morphism induced by the φi ’s. In these terms, the stability of L-foliations corre-
sponds to the local surjectivity of φ. Our main result states that under mild hypotheses
this is actually a local isomorphism. Observe that in order to understand the geometry
of the moduli space around F(g) one can always replace g by the maximal subalgebra
H 0 (X, T F(g)) ⊆ L.

Theorem A. Let g ⊆ L be a maximal subalgebra of dimension d. Suppose further that


`
h1 (X, T F(g)) = 0. Then, there exist neighborhoods U ⊆ i S(d)i and V ⊆ Inv of g and
F(g) respectively such that the restriction φ : U → V is an isomorphism. In particular,
if g ⊆ L is rigid then F(g) is rigid in Inv.

As an immediate application, we get

Corollary B. Let g ⊆ L be a maximal semisimple subalgebra with h1 (X, T F(g)) = 0.


Then F(g) is rigid in Inv.

To our knowledge, the tangent sheaf of a generic element of every known stable family
of L-foliations is trivial, i.e., the map g ⊗C OX → T F(g) is an isomorphism. When
X ⊆ Pr is arithmetically Cohen-Macaulay, Theorem A can be stated in the following
way.

Corollary C. Let X ⊆ Pr be smooth arithmetically Cohen-Macaulay of dimension n


and g ⊆ L a Lie subalgebra such that the morphism of sheaves g ⊗C OX → T F(g) admits
a resolution of length n − 3 of the form
rn−3 r1
M M
0→ OX (ein−3 ) → · · · → OX (e1i ) → g ⊗C OX → T F(g) → 0.
i=1 i=1
`
Then φ : i S(d)i → Inv is an isomorphism locally around g and F(g). In particular, if
g ⊆ L is rigid then F(g) is rigid in Inv.

Remark. The above Corollary opens up the possibility of constructing stable families of
L-foliations whose generic element has non-trivial tangent sheaf. We do so in Example
48, where we show that the adjoint representation of sln is rigid in Inv.

Remark. In the case of codimension 1 foliations we also prove Theorem 50 and Corollary
51, which are the analogous statements corresponding to Theorem A and Corollary C
for the classical spaces

F 1 (X, L) = {[ω] ∈ PH 0 (X, Ω1X ⊗ L) | ω ∧ dω = 0}

parameterizing classes of integrable twisted 1-forms with coefficients in some line bundle
L on X.
4

Acknowledgements. The content of this article is part of the author’s Ph.D thesis
under the advice of Fernando Cukierman, to whom the author is very grateful for his
support and guidance. Gratitude is also due to Carolina Araujo, César Massri, Jorge
Vitório Pereira and Federico Quallbrunn for their useful comments and suggestions.

2 Preliminaries
Let us first recall some facts on (families of) coherent sheaves that will be needed in the
rest of the article.

Definition 1. Let F be a coherent sheaf on X. The singular scheme of F is the closed


subscheme !
M n
i
Sing(F) = Supp Ext (F, OX ) .
i=1

Definition 2. Let f : Y → S be a morphism of noetherian schemes. A family of


coherent sheaves on the fibers of f is a coherent sheaf F on Y . The family is flat if F is
flat over S. A family of coherent sheaves on X with base S is just a family of coherent
sheaves on the fibers of the projection X × S → S. We will use the notation Fs for the
restriction of F to the fiber Ys . In the case Y = X × S we will just write Xs instead.

For the sake of clarity we will now state the so called “Cohomology and Base Change”
Theorem.

Theorem 3 (Theorem 12.11, Chapter III, [8]). Let f : Y → S be a projective morphism


of noetherian schemes and F a flat family of coherent sheaves on its fibers. Let s be a
point in S. Then:

1. if the natural map


ϕi (s) : Ri f∗ (F) ⊗ k(s) → H i (Ys , Fs )
is surjective, then it is an isomorphism and the same holds for every s′ in a suffi-
ciently small neighborhood of s.

2. if ϕi (s) is surjective, then the following are equivalent:

• ϕi−1 (s) is also surjective.


• Ri f∗ (F) is locally free in a neighborhood of s.

Lemma 4. Let F be a family of coherent sheaves on X with base S and s ∈ S be a point


such that h1 (Xs , Fs ) = 0. Then for every global section X ∈ H 0 (Xs , Fs ) there exists a
neighborhood U ⊆ S of s and an element XU ∈ H 0 (X × U , F) whose restriction to Xs
coincides with X.
5

Proof. By the semicontinuity theorem we know that h1 (Xs′ , Fs′ ) = 0 for every s′ in a
neighborhood U of s. Shrinking U if necessary, by the first item in Theorem 3 it must
be R1 p∗ (F) ⊗ k(s′ ) = 0 for every s′ ∈ U . But then s ∈
/ Supp(R1 p∗ (F)) and therefore

ϕ0 (s) : p∗ (F) ⊗ k(s) → H 0 (Xs , Fs )

is an isomorphism. In particular, every global section of Fs extends to a section defined


over some neighborhood U of s.

In the case X ⊆ Pr is arithmetically Cohen-Macaulay of dimension n we can replace


the hypothesis on the first cohomology group of F by a condition on the (graded)
homological dimension of F. For every coherent sheaf F on X one can construct a
graded free resolution
kn
M k0
M
d d d
··· → OX (−dni ) −→
n 1
· · · −→ OX (−d0i ) −→
0
F0 → 0.
i=1 i=1

We say that the resolution is of length l if kl > 0 and kl′ = 0 for l′ > l. In the case
X = Pn for instance Hilbert’s Syzygy Theorem tells us that we can always construct a
resolution of length n.

Lemma 5. Let X ⊆ Pr be arithmetically Cohen-Macaulay of dimension n and F0 a


coherent sheaf on X. If F admits a free resolution of length n − 3 then h1 (X, F) = 0.

Proof. Recall that for an arithmetically Cohen-Macaulay X the cohomology groups


H i (X, OX (d)) vanish for every d ∈ Z and 1 ≤ i ≤ n − 1. Now suppose a coherent sheaf
F admits a resolution of length n−3. For every 0 ≤ r ≤ n−3 let us denote Kr = ker(dr ).
The first of these kernels fits in the exact sequence
k0
M d
0 → K0 → OX (−d0i ) −→
0
F0 → 0,
i=1

which implies that h1 (X, F) = h2 (X, K0 ). In the same manner, for every r the sequence
kr+1
M dr+1
0 → Kr+1 → O(−dr+1
i ) −−−→ Kr → 0
i=1

is exact and therefore we have hr+2 (X, Kr ) = hr+3 (X, Kr+1 ). But then since Kn−3 ≃
Lkn−3 n−3
i=1 OX (−di ) we can conclude

h1 (X, F) = h2 (X, K0 ) = · · · = hn−1 (X, Kn−3 ) = 0.


2.1 Families of Lie algebras of vector fields 6

Lemma 6. Let X ⊆ Pr be a smooth arithmetically Cohen-Macaulay variety of dimension


n, F a coherent sheaf on X and V ⊗C OX → F a surjection for some vector space V .
If the morphism V ⊗C OX → F admits a resolution of the form
rn−3 r1
M M
0→ OX (ein−3 ) → ··· → OX (e1i ) → V ⊗C OX → F → 0,
i=1 i=1

then h1 (X, F) = 0 and the corresponding map V → H 0 (X, F) is surjective.

Proof. Let V ⊗C OX → F be as above and K its kernel. Since both F and K admit a
resolution of length n − 3, by the previous Lemma we have h1 (X, F) = h1 (X, K) = 0.
Then looking at the cohomology of

0 → K → V ⊗C OX → T F → 0

we can conclude that the map H 0 (X, V ⊗ OX ) = V → H 0 (X, F) is surjective.

We end this section by stating a functorial characterization of the zero-locus of a


global section s ∈ H 0 (X, H) for a locally free sheaf H.

Definition 7. Let H be a locally free sheaf on a scheme X and s : OX → H a global


section. The zero-locus Z(s) of s is the closed subscheme whose ideal sheaf is the image
of the dual map s∨ : H∨ → OX .

Lemma 8 (Proposition 3.10, [14]). Let H be a locally free sheaf on X and s ∈ H 0 (X, H)
a global section. Then the scheme Z(s) represents the subfunctor Zs ֒→ Hom(X, −)
defined by
Zs (S) = {f ∈ Hom(S, X) | f ∗ s = 0 ∈ H 0 (S, f ∗ H)}.

2.1 Families of Lie algebras of vector fields


In this section we will establish the notions regarding the deformation theory of Lie
algebras that will be necessary for our purposes. Let us now first briefly recall the
definition of Lie algebra cohomology. For further details on this subject the reader is
referred to [Section 7.7, [18]].
Let g be a finite dimensional complex Lie algebra and M a g-module. The co-
homology groups H • (g, M ) are defined as the homology of the Chevalley-Elilenberg
V
complex (C • (g, M ), δ) of M , where C • (g, M ) = HomC ( • g, M ) and the differential
δk : C k (g, M ) → C k+1 (g, M ) is given by the equations
k+1
X X
δ(f )(x1 ∧ · · · ∧ xk+1 ) = (−1)i xi · f (b
xi ) + (−1)i+j f ([xi , xj ] ∧ x
bij ).
i=1 i<j

We will use the notation Z k (g, M ) = ker(δk ) and B k (g, M ) = Im(δk−1 ).


2.1 Families of Lie algebras of vector fields 7

Definition 9. Let S be a C-scheme. A family of Lie algebras of dimension r with base S


V
is a locally free sheaf OS -module of rank r together with a morphism [−, −] : 2 L → L
that satisfies the Jacobi identity at the level of local sections.

Now let L = H 0 (X, T X) be the complex Lie algebra of global vector fields on X.
For a base S, let us consider the trivial family over S with fiber L, which consists of the
sheaf LS := L ⊗C OS together with the OS -linear extension of the Lie bracket on L.

Definition 10. A family of subalgebras of dimension d of L with base S is a locally free


subsheaf G ֒→ LS such that rg(G) = d and [G, G] ⊆ G. We will denote Subd : SchC →
Sets the functor assigning to each scheme S the set Subd (S) of families of subalgebras
of L of dimension d over S.

Example 11. For every n ≥ 5 consider the family G of subalgebras of vector fields in
Pn of dimension 3 with base S = Spec(C[t]) generated by the elements
∂ ∂ ∂ ∂
X1 = (x1 + tx2 ) + x2 + (x4 + tx5 ) + x5 ,
∂x1 ∂x2 ∂x4 ∂x5
∂ ∂
X2 = −x0 − x3 and
∂x1 ∂x4
∂ ∂ ∂ ∂
X3 = −tx0 − x0 − tx3 − x3
∂x1 ∂x2 ∂x4 ∂x5
As a family of Lie algebras, its structure is given by

[X1 , X2 ] = X2 , [X1 , X3 ] = tX2 + X3 .

For t 6= 0 the algebra G(t) is isomorphic to G(1) (via the map defined by X2 (t) 7→
tX2 (t)). For t = 0 on the other hand the algebra G(0) is not isomoprhic to G(1) (the
complete classification of complex Lie algebras of dimension less or equal than 4 and
their degenerations can be found in [1]).

Remark 12. The functor Subd is a subfunctor of the Grassmann functor: every family
of subalgebras of dimension d is in particular a family of subspaces of L. More precisely,
let
q
0 → K → LGr(L,d) − →Q→0
be the tautological exact sequence over the grassmanian of d-dimensional subspaces of
V
L and b : 2 LGr(L,d) → LGr(L,d) the Lie bracket on LGr(L,d) . We will also denote by
V
b its restriction b : 2 K → LGr(L,d) . Then for a scheme S every family of subalgebras
G ⊆ LS ∈ Subd (S) is of the form f ∗ (K ⊆ LGr(L,d) ) for some unique f : S → Gr(L, d).
The condition [G, G] ⊆ G is equivalent to the vanishing of the map of vector bundles
2
^
f ∗ (q ◦ b) : G → f ∗ Q ≃ LS /G.
2.1 Families of Lie algebras of vector fields 8

Proposition 13. The functor Subd is represented by a closed subscheme S(d) ⊆ Gr(L, d).

Proof. Following Remark 12, let us denote by s the global section of the locally free
sheaf Hom(K, Q) corresponding to the morphism
2
^
q◦b: K → Q.

Let S be a scheme. The discussion above implies that the functor Subd can actually be
described as

Subd (S) = {f ∈ Hom(S, Gr(L, d)) | f ∗ s = 0 ∈ H 0 (S, f ∗ Hom(K, Q))}.

By Lemma 8 this implies that Subd is represented by the closed subscheme Z(s) =:
S(d) ⊆ Gr(L, d) of the section s.

Remark 14. The automorphism group Aut(X) acts on T X and therefore on L. This
yields an action on G(L, d), which of course stabilizes S(d).

Definition 15. A subalgebra g ⊆ L is rigid if its orbit Aut(X) · g ⊆ S(d) contains a


non-empty Zariski open.

The following proposition allows us to check rigidity in terms of the Chevalley Eilen-
berg complex of the g-module L/g.

Proposition 16. Let g ⊆ L be a subalgebra of dimension d. Then

1. the tangent space to S(d) at g is isomorphic to Z 1 (g, L/g) and

2. under the above isomorphism, the tangent space to the orbit Aut(X) · g at g corre-
sponds to B 1 (g, L/g).

Proof. This is [Proposition 6.1, [15]] and [Proposition 7.1, [15]].

Example 17. Let g be a semisimple subalgebra of L. By a theorem of Whitehead we


know that H 1 (g, M ) = 0 for every finitely generated g-module M . In particular, g is
rigid in L.

Example 18. Let us consider the Lie subalgebra g ⊆ H 0 (Pn , T Pn ) of dimension 3


corresponding to the fiber over t = 1 in Example 11. Using the software GAP (see [17])
we can compute the corresponding tangent spaces and obtain dimC (Z 1 (g, L/g)) = 32
and dimC (B 1 (g, L/g)) = 28. This is, g is rigid as an abstract Lie algebra but admits
non-trivial deformations as a subalgebra of L.
2.2 Foliations and moduli spaces 9

2.2 Foliations and moduli spaces


We will now give the general definitions and properties on foliations that we will be used
afterwards.

Definition 19. A singular foliation F of codimension q on X is a short exact sequence

0 → T F → T X → NF → 0,

such that the tangent sheaf T F of F is closed under the Lie bracket of vector fields
and the normal sheaf NF is torsion-free of rank q. The singular scheme Sing(F) of the
foliation is the singular scheme of its normal sheaf.

Definition 20. A family of foliations of codimension q on X with base S is a short


exact sequence
0 → T F → T XS → NF → 0
such that T F is closed under the Lie bracket of relative vector fields and its normal
sheaf NF is torsion-free on the fibers and of rank q. Its singular scheme Sing(F) is the
singular scheme of its normal sheaf. The family is said to be flat if NF is flat over S.

Definition 21. For every polynomial P ∈ Q[t] let InvP : Sch → Sets be the functor
that assigns to every scheme S the set of flat families F of foliations on X with base S
such that the restriction of NF to every fiber has Hilbert polynomial equal to P .

With a similar argument to the proof of Proposition 13 one can show the following

Proposition 22 (Proposition 6.3, [14]). The functor InvP is represented by a locally


closed subscheme InvP ⊆ QuotP (T X).

The following definition is in order to mantain a clearer exposition.

Definition 23. We will use the notation


a
Inv = InvP .
P ∈Q[t]

We will also denote by Inv ⊆ Quot(T X) the scheme representing the functor of flat
families of coherent involutive subsheaves of T X (this is, families whose normal sheaf
may not be torsion-free). Of course, these spaces are equipped with a universal family
which will be denoted by FInv .

Remark 24. Just as in Remark 14, the natural action of Aut(X) on Quot(T X) restricts
to an action on Inv. We will say that F is rigid if its orbit is Zariski open.

Let us now move on to describing the tangent space to Inv at a point F.


2.2 Foliations and moduli spaces 10

Definition 25. Let F be a foliation on X. The leaf complex of F is the complex


2
^
d0 d1 d2
LF : T X −→ Hom(T F, NF ) −→ Hom( T F, NF ) −→ ··· ,

where the morphisms are defined at the level of local sections by

d0 (v)(w) = [v, w] and


r+1
X X
dr (α)(w1 , . . . , wr+1 ) = (−1)i [wi , α(c
wi )] + d
(−1)i+j α([wi , wj ], wij ).
i=1 i<j

This object is very relevant for the theory of deformations of F. Recall that the
tangent space to Quot(T X) at the point F is isomorphic to Hom(T F, NF ). Let ψ :
Hom(T F, NF ) → TF Quot(T X) be the canonical isomorphism. Using the ideas pre-
sented in [7] we can easily relate the leaf complex of F to the tangent space to Inv at
the point F.

Theorem 26. Let F be a foliation on X and O(F) its orbit under the natural action
of Aut(X) on Inv. The following diagram of vector spaces is commutative and has exact
rows:

0 Sym(F) ⊕ H 0 (X, T F) L TF O(F) 0


H 0 (d0 )
ψ
0 ker(H 0 (d1 )) TF Inv 0.

Proof. The deformation problem considered in [7] includes possible deformations of X.


There, the author proves that the vector space consisting of first order deformations of
F coincides with the first hypercohomology group H1 (X, LF ) of the leaf complex (see [7,
Theorem 2.4]. This is, if U is a sufficiently fine cover of X then giving a first order defor-
mation X of X and an involutive subsheaf F ′ ⊆ T XSpec(D) is equivalent to specifying a
1-cocycle θ ∈ C 1 (U , T X) (which defines X ) and an element η ∈ C 0 (U , Hom(T F, NF ))
satisfying some equations of compatibility and integrability.
In order to apply this correspondence to our deformation problem we have to consider
elements of the form

(0, η) ∈ C 1 (U , T X) ⊕ C 0 (U , Hom(T F, NF )).

In this case, the compatibility equations indicate that the element η defines a global
section η ∈ H 0 (X, Hom(T F, NF )). The integrability condition on the other hand is
exactly d1 (η) = 0.
Alternatively, one can apply [7, Theorem 1.4] to the case where X is the trivial
deformation of X.
2.2 Foliations and moduli spaces 11

Following [6] every foliation F on X can be defined by an integrable and locally


decomposable away from its singular points differential q-form as follows: for every fo-
liation F of codimension q on X we can consider the corresponding monomorphism
det(NF )∨ ֒→ ΩqX , which is induced by a unique element

ωF ∈ H 0 (X, ΩqX ⊗ det(NF ))

up to scalar multiplication. Thus ωF satisfies the following properties:

1. around every point p such that ωF (p) 6= 0 there exist local 1-forms ωi such that
ωF = ω1 ∧ · · · ∧ ωq ( i.e., ωF is decomposable around p),

2. ker(ωF ) = T F is involutive (ωF is integrable) and

3. Z(ωF ) = Sing(F) is of codimension at least 2.

The first condition is equivalent to requiring that ωF satisfies the Plücker equations

(P) ıv (ωD ) ∧ ωD = 0
V
for every local section v of q−1 T X. Integrability on the other hand corresponds to the
Frobenius integrability condition

(F) ıv (ωD ) ∧ dωD = 0


V
for every local section v of q−1 T X. Also, the zero locus of ωF coincides with the
singular scheme of F.
We will be specially interested in the following type of singularities.

Definition 27. Let F be a codimension 1 foliation on X. We say that p ∈ X is a Kupka


singularity if ωF (p) = 0 and dωF (p) 6= 0. We will denote by K(F) the set of Kupka
points of F.

Remark 28. Around a point p ∈ K(F) the foliation F can be described as a pullback
of a germ of foliation at (C2 , 0) with an isolated singularity at the origin. In particular,
K(F) is pure of codimension 2. For further details on the local structure of F around a
Kupka point see [Capitulo 1.4, [9]].

Definition 29. Let F be a codimension 1 foliation. We say that a singularity p ∈ X is


of Reeb-type if there exists an anallytic neighborhood U of p such that F|U is the foliation
P
induced by some ω = ni=1 fi dzi , where fi (p) = 0 for every i and (df1 )p , . . . , (dfn )p are
linearly independent.
2.2 Foliations and moduli spaces 12

Conversely, for a line bundle L on X every element ω ∈ H 0 (X, ΩqX ⊗ L) satisfying


conditions (P), (F) and codim(Z(ω)) ≥ 2 defines a singular foliation Fω of codimension
q with tangent sheaf
T Fω = ker(ω)
and such that det(NF ) = L. Moreover, two such forms induce the same foliation if and
only if they differ by a multiplicative constant λ ∈ H 0 (X, OX
∗ ) = C. This motivates the

following:

Definition 30. Let L be a line bundle on X. The space of foliations of codimension q


and degree L on X is defined as

F q (X, L) = {[ω] ∈ PH 0 (X, ΩqX ⊗ L) | ω satisfies codim(Z(ω)) ≥ 2, (P), (F)}.

This scheme also admits a natural action of Aut(X). A point F ∈ F q (X, L) is said
to be rigid if its orbit contains a non-empty Zariski open.

Remark 31. Using the restriction of the tautological sequence on PH 0 (X, ΩqX ⊗ L) we
can construct the tautological family of foliations FLq with base F q (X, L), i.e., the family
whose restriction to the fiber over each [ω] is the foliation Fω .

We have now defined two spaces paremetrizing foliations on X (with different notions
of continuity). In the case P ic(X) is discrete, we can indeed compare the geometry of
these spaces.

Proposition 32. Suppose that P ic(X) is discrete and let F0 be a codimension q foliation
on X. Let L = det(NF0 ) and C ⊆ F q (X, L) be the stratum of the flattening stratification
of NF q passing through F0 . Then there exists an neighborhood U ⊆ Inv of F such that
L
the morphism ω : U → C defined as ω(F ′ ) = [ωF
′ ] is an Aut(X)-equivariant isomorphism

around F0 and [ωF0 ].

Proof. Let F be a flat family of codimension q foliations with affine base S = Spec(A)
and s ∈ S some point such that Fs = F0 . The relative version of the argument above
yields a unique (up to multipllication by units of A) global section

ωF ∈ H 0 (X × S, ΩqX×S|S ⊗ det(NF ))

such that the kernel of the contraction


q−1
ωF : T XS → ΩX×S|S ⊗ det(NF )

coincides with T F. Its image is canonically isomorphic to NF (which is flat), which


implies that for every s ∈ S we have

(⋆) ker(ωF (s)) = ker(ωF )s = T Fs .


13

Also, being P ic(X) discrete by the Seesaw Theorem [Corollary 6, p.54, [11]] we know
that det(NF ) = p∗1 det(NF0 ) and therefore

ΩqX×S|S ⊗ det(NF ) = p∗1 (ΩqX ⊗ L).

This is, the section ωF defines an element of H 0 (S, H 0 (X, ΩqX ⊗L)⊗C OS ) or equivalently
a morphism S → H 0 (ΩqX ⊗ L). By (⋆) (and the fact that ωF is integrable and locally
decomposable away from its singular points), this descends to a morphism

ωF : S → F q (X, L)

such that the family F is the pullback under ω of the tautological family over F q (X, L).
Since F is indeed a flat family, ω must factor through the flattening stratification asso-
ciated to NF q .
L
Applying the above to the case S = Spec(OInv,F0 ) one checks that the universal
family FInv is the pullback under ωFInv of the tautological family over C, which easily
implies that ω : S → C induces an isomorphism locally around F0 and [ω0 ].

3 L-foliations
Let G ⊆ Aut(X) be a connected Lie group with Lie algebra g. Let us suppose that
the natural action of G on X satisfies that dim G · x is constant in codimension 1. The
leaves of the foliation F(g) induced by this action are exactly the orbits of maximal
dimension. The derivative of G ֒→ Aut(X) at the identity corresponds to the inclusion
g ֒→ H 0 (X, T X) = L. The restriction of global sections yields a morphism

ρg : g ⊗C OX → T X

whose image is the tangent sheaf T F(g). As for the twisted differential form defining
F(g), we will use the notation ω(g) = ωF (g) .

Definition 33. Let F be a foliation on X. We say that F is an L-foliation if F = F(g)


for some subalgebra g ⊆ L.

As we will now see, these foliations admit some additional structures.

Definition 34. Let F be a foliation on X. We say that a global vector field X ∈


H 0 (X, T X) is a symmetry of F if the flow associated to X satisfies exp(tX)∗ F = F and
X is not tangent to F.

A general foliation does not admit neither symmetries nor algebraic leaves. The
following is an adaptation of [Théoreme 1.8, [3]].

Theorem 35. Let F be an L-foliation on X. Then


14

1. F admits a symmetry, or

2. every leaf of F is algebraic.

Proof. If F is an L-foliation, then F = F(g) for g := H 0 (X, T F) ֒→ L. Let G1 be the


connected subgroup
G1 := hexp(tY )iY ∈g < Aut(X)
generated by the flows of g. On the other hand, let us consider the group Aut(F) =
{φ ∈ Aut(X)|φ∗ F = F} of automorphisms of F. Observe that Aut(F) is the fiber over
ωF of the map ω : Aut(X) → PH 0 (X, det(NF )) satisfying ω(φ) = [ωφ·F ]. In particular,
Aut(F) is an algebraic subgroup of Aut(X). Also, G1 < Aut(F). Let G1 be the Zariski
closure of G1 and Aut(F)0 the connected component of the identity in Aut(F). Then

G1 < G1 < Aut(F)0 .

If G1 = G1 , then the orbit G1 · x is algebraic for every x ∈ X. But then every leaf of F
is algebraic.
If on the other hand the above inclusion were strict, there must be an element
X ∈ Te Aut(F)0 \ g, i.e. a symmetry of F.

Example 36. For g = sl2 , let us consider the foliation F(g) in X = PSym4 (C2 ) ≃ P4
associated to the action of PGL2 (C) by changes of coordinates, which was studied in [2].
This foliation satisfies condition 2 in the above Theorem, since the closure of a generic
orbit coincides with a fiber of the j-invariant.
The restriction of this foliation to an appropiate hyperplane yields an important
construction in the theory of foliations in projective spaces, namely the exceptional
component: this restriction is induced by an action of the affine Lie algebra aff(C) and
is rigid in F 1 (P3 , 4).

Example 37. More generally, by [Théoreme 1.22, [3]] every L-foliation F(g) on Pn of
codimension 1 such that [g, g] = g satisfies the second condition in Theorem 35.
In [5] the authors study foliations on Pn induced by infinitesimal actions of Lie
subalgebras g ⊆ H 0 (Pn , T Pn ) which are locally free in codimension one, i.e., such that
the morphism g ⊗C OX → T F(g) is an isomorphism. Regarding their stability, in
[Corollary 6.1, [5]] they show that under some hypotheses on the singular scheme of
F(g), the elements g and F(g) have isomorphic neighborhoods in S(d) and F q (Pn , n + 1)
respectively. As a consequence, they are able to construct irreducible components of this
space consisting generically of L-foliations.
3.1 Families of subalgebras and foliations 15

3.1 Families of subalgebras and foliations


The last example shows that under certain hypotheses it is possible to define a morphism
between S(d) and some moduli space of foliations. We will now explain the formalisms
that lead to a similar constrution around every point in the moduli space of subalgebras.

Definition 38. Let G ⊆ LS be a family of Lie subalgebras parametrized by some scheme


S. The associated family F(G) of involutive distributions is the family whose tangent
sheaf is the image of the morphism

ρG : G ⊗ OX×S → T XS .

Since the scheme S(d) is equipped with a universal family GS(d) ⊆ LS(d) of subalge-
bras of L of dimension d, it is only natural to consider the case S = S(d). We will use
the notation Fd := F(GS(d) ) for the induced family

Fd : 0 → T Fd → T XS(d) → Nd → 0.

Being this family not necessarily flat, we have to consider the flattening stratification
a
S(d)i → S(d)
i

associated to the sheaf Nd . We thus get over each stratum S(d)i a flat family Fi :=
F(GS(d)i ) and therefore a morphism φi : S(d)i → Inv such that

0 → T Fi → T PnS(d)i → Nd → 0

is the pullback under φi of the universal family over Inv. Observe that if g ∈ S(d)i then
its orbit O(g) under the action of Aut(X) is contained in S(d)i . It follows from the
above that φi is Aut(X)-equivariant. In particular, we have φi (O(g)) = O(F(g)).
`
Definition 39. We will denote φ : i S(d)i → Inv the morphism induced by the φi ’s.

For a family G ⊆ LS of subalgebras over some affine scheme S we have G ⊆


H 0 (S, p2,∗ T F(G)). In general, this may be a strict inclusion. In what follows, we want
to be able to recover the family of subalgebras from its associated family of foliations.
The following definition was first introduced in [3]:

Definition 40. Let g ⊆ L be a subalgebra. We say that g is maximal if g = H 0 (X, T F(g)).

Remark 41. Maximality defines an open subscheme of S(d). Indeed, a point g ∈ S(d)
is maximal if and only if the fiber (p2,∗ T Fd )(g) has dimension less than d + 1. These
are also the points around which the quotient sheaf p2,∗ (T Fd )/GSd is zero. Observe
also that in order to understand the geometry of Inv around F(g) we can replace g by
H 0 (X, T F(g)).
3.2 Stability 16

Example 42. Let us consider the family G on Pn with base S = Spec(C[t]) introduced
in Example 11. The family F = F(G) is of dimension 3 and generated by the elements
X1 , X2 , X3 ∈ H 0 (Pn × S, T PnS ), which induce the morphism

X : OP⊕3
n ×S −
→ T F.

Since X(t) : OP⊕3n → T Ft is an isomorphism for every t ∈ C, the family F(G) is automat-

ically flat over S.

3.2 Stability
We will now prove the results stated in the introduction. Recall that according to Remark
41, in order to study the geometry of Inv around F(g) there is no loss of generality in
assuming that g is maximal.

Theorem 43. Let g ⊆ L be a maximal subalgebra of dimension d. Suppose further that


`
h1 (X, T F(g)) = 0. Then, there exist neighborhoods U ⊆ i S(d)i and V ⊆ Inv of g
and F(g) of respectively such that the restriction φ : U → V is an isomorphism. In
particular, if g ⊆ L is rigid then F(g) is rigid in Inv.

Proof. Let 0 → T F → T XInv → NF → 0 be the universal family over Inv and


⊕d
X0 : OX ։ T F(g) the morphism induced by the action of g. By Lemma 4 we can
extend this morphism, i.e., there exists some neighborhood V of F(g) and a morphism
XV : OV⊕d → T F|V restricting to X0 . Without loss of generality we can assume that XV
is also surjective.
We will denote by π1 and π2 the projections of X × V onto X and V respectively.
By Theorem 3 (shrinking V if necessary) we know that π2,∗ T F|V is locally free. The
pushforward of T F ֒→ T XV therefore becomes a monomorphism of vector bundles

π2,∗ T F|V ֒→ π2,∗ T XV = H 0 (X, T X) ⊗ OV = LV ,

whose image is an element of SubL d (V). By the representability of this functor there
exists a morphism f : V → S(d) such that π2,∗ T F|V is the pullback under f of the
universal subalgebra. Let us now consider the commutative diagram

1×f
X ×V / X × S(d)

π2 p2
 f 
V / S(d),

where p2 : X × S(d) → S(d) is the usual projection. Being T Fd globally generated,


as a subsheaf of T XS(d) it must coincide with the image of the restriction of sections
3.2 Stability 17

p∗2 p2,∗ T Fd → T XS(d) . Since g is maximal, by Remark 41 we can restrict ourselves to a


neighborhood Ue ⊆ S(d) of g and redefine V = f −1 (Ue) in order to guarantee the equality

p2,∗ (T FUe ֒→ T XUe ) = GUe ֒→ LUe ,

where GUe = GS(d) |Ue . But then the tangent sheaf of the family f ∗ (Fd |Ue ) is
 
Im (1 × f )∗ (p∗2 p2,∗ (T Fd |Ue ) → T XUe ) = Im (f ◦ π2 )∗ p2,∗ T FUe → T XV

= Im (f ◦ π2 )∗ GUe → T XV
= Im (π2∗ π2,∗ T F|V → T XV ) .

Being T F|V also globally generated, we have

Im(π2∗ π2,∗ T F|V → T XV ) = T F|V .

This is, the pullback of Fd |Ue is

0 → T F|V → T XV → (1 × f )∗ N (d) → 0

and therefore f ∗ N (d) ≃ NF is flat over V. Then, f factorizes over the flattening strati-
fication a
V→ S(d)k → S(d)
k

and therefore (shrinking V if necessary) over the stratum S(d)i passing through g. Let
us also observe that we have proven the equality

f ∗ (Fd |Ue ) = FInv |V .

Let U = φ−1 ∗
i (V) ⊆ S(d)i . The composition φi ◦ f : V → Inv satisfies (f ◦ φi ) F|V =
F|V . It follows from the representability of Inv that this composition coincides with the
inclusion of V in Inv. Analogously, the composition f ◦ φi : U → S(d)i coincides with
the inclusion of U in S(d)i and therefore φi : U → V is an isomorphism.

As a consequence, we can state a version of [Corollary 6.4, [5]] for more general
varieties and actions admitting positive-dimensional generic stabilizers.

Corollary 44. Let g ⊆ L be a maximal semisimple subalgebra with h1 (X, T F(g)) = 0.


Then F(g) is rigid in Inv.

Proof. Let S(d)i be the stratum passing through g in S(d). Being g semisimple, its orbit
is open. By Theorem 43 it follows that F(g) is rigid in Inv.

For arithmetically Cohen-Macaulay subvarieties X ⊆ Pr , we also state the above


taking into consideration the (graded) homological dimension of T F(g).
3.2 Stability 18

Corollary 45. Let X ⊆ Pr be smooth arithmetically Cohen-Macaulay of dimension n


and g ⊆ L a Lie subalgebra such that the morphism of sheaves g ⊗C OX → T F(g) admits
a resolution of the form
rn−3 r1
M M
0→ OX (ein−3 ) → ··· → OX (e1i ) → g ⊗C OX → T F(g) → 0.
i=1 i=1
`
Then φ : i S(d)i → Inv is an isomorphism locally around g and F(g). In particular, if
g ⊆ L is rigid then F(g) is rigid in Inv.
Proof. It follows directly from Lemma 6 and Theorem 43.

Example 46. The j-invariant in PSym4 (C2 ) and the exceptional foliation. The in-
finitesimal action of sl2 on PSym4 (C2 ) is locally free outisde a set of codimension 2.
In particular, the morphism sl2 ⊗C OP4 → T F(sl2 ) is an isomorphism. Since this is a
semisimple Lie algebra, by the Theorem above the element F(sl2 ) is rigid in Inv1 .
Since aff(C) ⊆ sl4 is also rigid, the same holds for the exceptional foliation F(g)
in P3 for g = aff(C). This is, the foliation induced by the restriction of F(sl2 ) to the
properly embedded P3 ⊆ PSym4 (C2 ) is rigid in the corresponding moduli space Inv1 of
foliations in P3 .
Example 47. An action of aff(C) on the smooth quadric Q ⊆ P4 . In [10] the authors
classify the irreducible components of F q (X, ωX ) for Fano 3-folds with P ic(X) ≃ Z. In
the case X is the smooth quadric in P4 , this space has 3 irreducible components and only
one of them consists generically of L-foliations, which are associated to a rigid action of
aff(C). We will now see that we can recover its rigidity in Inv as an application of the
above results.
Let us identify the space PSym4 (C2 ) with the set of effective divisors in P1 . The
closure of the point q = 1 + (−1) + i + (−i) under the action of sl2 in the example
above is a smooth quadric Q. Let us consider the foliation induced by the restriction
of this action to the affine Lie algebra aff(C) fixing the point ∞. Since its orbits are of
dimension 2 outside the set {p + 3∞} ∪ {3p + ∞} we have

T F(aff(C)) ≃ aff(C) ⊗C OQ .

In particular, we are under the hypotheses of Corollary 45. Recall that in this case
we have L = so5 . Using the GAP software we can easily compute the dimensions
corresponding to the inclusion aff(C) ⊆ sl2 ⊆ so5 :

dim Z 1 (aff(C), so5 /aff(C)) = dim B 1 (aff(C), so5 /aff(C)) = 8.

It follows that this is a rigid subalgebra and therefore the foliation F(aff(C)) is rigid in
the space Inv1 of foliations on Q.
3.2 Stability 19

Example 48. The adjoint representation of sln . Let g = sln y V be the adjoint repre-
sentation. Let us consider the foliation F(g) associated to the corresponding action on
PV . Its orbits coincide with the conjugacy classes of sln under the action of SLn . Its
algebra of invariants is the polynomial algebra on the n − 2 coefficients of the character-
istic polynomial χ. This is, the generic leaf of this foliation coincides with the generic
fiber of the map
n−3
χ : PV 99K P(2,...,n) .
The inclusion sln ֒→ H 0 (PV, T PV ) can be interpreted as follows: for every element
X ∈ Sln , the evaluation at a point p ∈ PV is

X(p) = [X, p] ∈ V /hpi ≃ Tp PV.

With this in mind, we can compute the kernel of the morphism

ρ : sln ⊗ OPV → T F(g).

Let us consider the Zariski open U of PV

U = {[p] ∈ PV | {Id, p, . . . , pn−1 } is linearly independent}.

For a point p ∈ U and an element X ∈ sln , we have X(p) = 0 if and only if [X, p] = 0.
Indeed, if this were not the case then there would exist an element A ∈ sln such that

[A, p] = p.

By induction one can see that this easily implies that [A, pk ] = nk pk for some {nk }k∈N ⊆
N. But then tr(pk ) = 0 for every k ∈ N and therefore p must be a nilpotent element,
which leads to a contradiction. This means that the fiber ker(ρ)(p) is the centralizer of
p, which is of dimension n − 1 (see for instance [16]). For every 1 ≤ k ≤ n − 1 consider
the matrix Pk ∈ C[V ]kn×n such that for every p ∈ V we have

Pk (p) = pk − tr(pk )Id.

Each of these elements defines a section Zk ∈ H 0 (PV, sln ⊗OPV (k)) that trivially satisfies
Zk ∈ ker(ρ). Observe that these sections are linearly independent around every p ∈ U .
We have now constructed a complex
n−1
M
0→ OPV (−k) → sln ⊗ OPV → T F(g) → 0
k=1

that is exact on U . Since codim(PV \ U ) ≥ 2, this means that this is in fact a resolution
of T F(g).
Being sln semisimple, if n ≥ 5 we can apply Corollary 45 in order to conclude that
the foliation F(g) is rigid in Inv.
3.2 Stability 20

Example 49. Let us consider the foliation of codimension 2 on P5 induced by the vector
fields
∂ ∂ ∂ ∂
X1 = (x1 + x2 ) + x2 + (x4 + x5 ) + x5 ,
∂x1 ∂x2 ∂x4 ∂x5
∂ ∂
X2 = −x0 − x3 and
∂x1 ∂x4
∂ ∂ ∂ ∂
X3 = −x0 − x0 − x3 − x3
∂x1 ∂x2 ∂x4 ∂x5
(this is, the foliation induced by the subalgebra g ⊆ L of Example 18 and Example 42).
Since g ⊗C OP5 → T F(g) is an isomorphism, this subalgebra satisfies the hypotheses
of Corollary 45. Recall from Example 18 that the space tangent to S(3) at the point g
is of dimension 32 and that the dimension of the tangent space of its orbit under the
action of Aut(P5 ) is 28. Applying Corollary 45, we can conclude that every irreducible
component of Inv passing through g is of dimension less or equal than 32 (and greater or
equal than 28). Since g is rigid as a Lie algebra, every such component must correspond
to an irreducible component of the space of Lie algebra morphisms HomLie (g, sl6 ) ⊆
HomC (g, sl6 ) passing through the point g ֒→ sl6 defining the foliation.
In the case of foliations of codimension 1 we can translate the above results to the
spaces F 1 (X, L), provided that P ic(X) is discrete:
Theorem 50. Suppose P ic(X) is discrete and let g ⊆ L be a maximal subalgebra acting
with orbits of dimension dim(X) − 1 in codimension 1 and such that h1 (X, T F(g)) = 0.
If Sing(F(g)) is without embedded components and Sing(F(g)) = K(F(g))∪{p1 , . . . , pr }
`
where the pi ’s are Reeb-type singularities, then there exist neighborhoods U ⊆ i S(d)i
and V ⊆ F 1 (X, NF∨∨ (g) ) of g and F(g) respectively such that the morphism ω : U → V
defined by g 7→ [ω(g)] is an isomorphism. In particular, if g is a rigid subalgebra then
F(g) is rigid in F 1 (X, NF∨∨ (g) ).

Proof. Let g ⊆ L be a subalgebra satisfying the hypotheses of the Theorem and L =


det(NF (g) ) = NF∨∨ 1 1
(g) . Let FL be the tautological family of foliations with base F (X, L)
constructed in Remark 31. The hypothesis on Sing(F(g)) allow us to apply [Theorem
8.13, [14]] in order to ensure that the singular scheme Sing(FL1 ) of the family is flat in
a neighborhood of [ω(g)] ∈ F 1 (X, L). Then, by [Theorem 7.8, [14]] the normal sheaf
NF 1 is also flat in a neighborhood of [ω(g)]. This is, the stratum of the correponding
L
flattening stratification passing through [ω(g)] is open in F 1 (X, L). Using Theorem 43
and Proposition 32 we can conclude that the composition
a
ω: S(d)i → Inv1 → F 1 (X, L)
i

is an isomorphism locally around g and F(g).


3.2 Stability 21

Corollary 51. Let X be arithmetically Cohen-Macaulay of dimension n and g ⊆ L a


subalgebra acting with orbits of dimension n − 1 in codimension 1. Suppose further that
the morphism g ⊗C OX → T F(g) admits a resolution of length n − 3 of the form
rn−3 r1
M M
0→ OX (ein−2 ) → ··· → OX (e1i ) → g ⊗C OX → T F(g) → 0.
i=1 i=1

If Sing(F(g)) is without embedded components and Sing(F(g)) = K(F(g))∪{p1 , . . . , pr }


`
where the pi ’s are Reeb singularities, then there exist neighborhoods U ⊆ i S(d)i and
V ⊆ F 1 (X, NF∨∨(g) ) of g and F(g) respectively such that the morphism ω : U → V defined
by g 7→ ω(g) is an isomorphism. In particular, if g is a rigid subalgebra then F(g) is
rigid in F 1 (X, NF∨∨(g) ).

Proof. It follows from the Theorem above and Lemma 6.

As was already mentioned, the deformation theory of g and F(g) are governed by
the leaf complex LF (g) and the Chevalley-Eilenberg complex C • (g, L/g) respectively. We
will end this section by giving a first order intepretation of Theorem 43 in terms of these
complexes. We will denote the scheme of dual numbers by D = Spec(C[ε]/(ε2 )).
Let g ⊆ L be a maximal subalgebra such that h1 (X, T F(g)) = 0 and S(d)i the
stratum passing through g. The derivative of φ at a point g is the map

dg φ : Tg S(d)i ⊆ SubL
d (Spec(D)) → Inv(Spec(D)) = TF (g) Inv
(G ⊆ LD ) 7→ F(G),

where Tg S(d)i is the set of families of subalgebras G ⊆ LD over D such that G(0) = g
and whose normal sheaf NF (G) is flat over D. By Remark 41 every such family satisfies
π2,∗ T F(G) = G. In particular this implies that dg φ is injective. Just as in the proof
of Theorem 43, we can construct a section for this map (but this time in terms of the
corresponding complexes). By Proposition 16 and Theorem 26 we can realize dg φ as a
map
Vi ⊆ Z 1 (g, L/g) → ker(H 0 (d1 ))
where V is the subspace corresponding to Tg S(d)i ⊆ Tg S(d) and d1 is the differential of
LF . Observe that the hypotesis on the first cohomology group of T F(g) implies that
the cohomology of F(g) is of the form

0 → g → L → H 0 (X, NF (g) ) → 0 → · · ·

and therefore H 0 (X, NF (g) ) ≃g−mod L/g.


On the other hand, the exterior powers of the morphism g ⊗ OX → T F(g) are
epimorphisms

^ •
^
f• : g ⊗C OX → T F(g).
3.2 Stability 22

In order to simplify the notation we will use the symbol f for any of these maps. Their
duals

^ •
^ •
^

f : Hom( T F(g), NF (g) ) → Hom( g ⊗C OX , NF (g) ) ⊆ HomC ( g, L/g)

induce a monomorphism of complexes f ∗ : H 0 (LF ) → C(g, L/g) of the form

0 0

H 0 (d) V H 0 (d)
L Hom(T F(g), NF (g) ) Hom( 2 T F(g), NF (g) ) ···
id f∗ f∗
V
L HomC (g, L/g) δ
HomC ( 2 g, L/g) δ
···

Being f ∗ a morphism of complexes we must have f ∗ ker(H 0 (d)) ⊆ ker(δ). This is, f ∗ is
a transformation
f ∗ : ker(H 0 (d1 )) → Z 1 (g, L/g).
Now let FD be a first order deformation of F(g) and η ∈ ker(H 0 (d1 )) the corresponding
element according to Theorem 26. Now f ∗ η : g → L/g defines the family of subalgebras
of L over D given by
G = H 0 (XD , T FD ) ֒→ LD .
The differential of φ applied to this element is the family with tangent sheaf Im(ρG :
G ⊗ OX×D → T XD ) ⊆ T FD . Being T F(g) globally generated, this inclusion is in fact
an equality. We can then conclude that f ∗ η lies in Vi (this is, it is tangent to the stratum
S(d)i ) and that f ∗ is in fact a section of the differential of φ at the point g, which is
therefore bijective.

Remark 52. The above discussion would actually prove that φ is an isomorphism locally
`
around g and F(g), provided that g or F(g) are smooth points of the schemes S(d)i
and Inv.
References 23

References
[1] D. Burde and C. Steinhoff. Classification of orbit closures of 4-dimensional complex
Lie algebras, Journal of Algebra, 1999, vol. 214, no 2, p. 729-739.

[2] O. Calvo-Andrade and F. Cukierman. A note on the j invariant and foliations,


Proceedings of the XVIth Latin American Algebra Colloquium, Bibl. Rev. Mat.
Iberoamericana (Rev. Mat. Iberoamericana, Madrid, 2007), pp. 99– 108.

[3] D. Cerveau and J. Déserti. Feuilletages et actions de groupes sur les espaces projec-
tifs, Société Mathématique de France, 2005.

[4] D. Cerveau and A. Lins Neto. Irreducible components of the space of holomorphic
foliations of degree two in CP(n), n ≥ 3. Annals of Mathematics, 1996, p. 577-612.

[5] F. Cukierman and J. V. Pereira. Stability of holomorphic foliations with split tangent
sheaf, American journal of mathematics, 2008, vol. 130, no 2, p. 413-439.

[6] A. S. de Medeiros. Singular foliations and differential p-forms, Annales de la Faculté


des Sciences de Toulouse: Mathématiques, 2000, p. 451-466.

[7] X. Gómez-Mont. The transverse dynamics of a holomorphic flow, Annals of Math-


ematics, 1988, vol. 127, no 1, p. 49-92.

[8] R. Hartshorne. Algebraic geometry, Springer Science and Business Media, 2013.

[9] A. Lins-Neto. Componentes irredutı́veis dos espaços de folheaçoes. Publicações


Matemáticas do IMPA, 2007.

[10] F. Loray, J. V. Pereira and F. Touzet. Foliations with trivial canonical bundle on
Fano 3-folds. Mathematische Nachrichten, 2013, vol. 286. 8-9, p. 921-940.

[11] D. Mumford, C.P. Ramanujam and I. Manin. Abelian varieties, Oxford: Oxford
University Press, 1974.

[12] A. N. Parshin and I. R. Shafarevich. Algebraic geometry IV: linear algebraic groups,
invariant theory, Springer-Verlag, 1994.

[13] G. Pourcin. Deformations of coherent foliations on a compact normal space, An-


nales de l’institut Fourier, 1987, p. 33-48.

[14] F. Quallbrunn. Families of distributions and Pfaff systems under duality, Journal
of Singularities, 2015, vol. 11, p. 164-189.

[15] J. R. Richardson. A rigidity theorem for subalgebras of Lie and associative algebras,
Illinois Journal of Mathematics, 1967, vol. 11, no 1, p. 92-110.
References 24

[16] J. Rotman. Advanced modern algebra, Vol. 114. American Mathematical Soc., 2010.

[17] The GAP Group, GAP – Groups, Algorithms, and Programming, Version 4.11.1;
2021. (https://www.gap-system.org).

[18] C. A. Weibel. An introduction to homological algebra. Cambridge University Press,


1995.

* Instituto de Matemática Pura e Aplicada, Rio de Janeiro, Brazil. The author was
supported by CONICET-Argentina and CNPq-Brazil.
E-mail adress: sebastian.velazquez@impa.br

You might also like