02.04.2013 Views

LEFSCHETZ PENCILS Contents 1. Introduction 2 2. Classical ...

LEFSCHETZ PENCILS Contents 1. Introduction 2 2. Classical ...

LEFSCHETZ PENCILS Contents 1. Introduction 2 2. Classical ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

<strong>LEFSCHETZ</strong> <strong>PENCILS</strong><br />

FRANCISCO PRESAS<br />

Abstract. Lectures over the use and applications of Lefschetz<br />

pencils in symplectic and contact geometry. The lectures are divided<br />

in three blocks: projective case, symplectic case and contact<br />

case. First, the classical notion of Lefschetz pencil is introduced<br />

and then the more modern setup in symplectic topology is outlined.<br />

Later, the analogous constructions in the contact category<br />

are briefly sketched.<br />

<strong>Contents</strong><br />

<strong>1.</strong> <strong>Introduction</strong> 2<br />

<strong>2.</strong> <strong>Classical</strong> pencils 2<br />

<strong>2.</strong><strong>1.</strong> Quick review of some basic complex geometry 2<br />

<strong>2.</strong><strong>2.</strong> Positive bundles 4<br />

<strong>2.</strong>3. Lefschetz pencils 6<br />

<strong>2.</strong>4. Symplectic structures and monodromy 8<br />

<strong>2.</strong>5. Monodromy of a pencil 10<br />

<strong>2.</strong>6. Main equivalence theorem I: from pencil to symplectic<br />

structure 14<br />

3. Approximately holomorphic geometry 17<br />

3.<strong>1.</strong> Trivializations 19<br />

3.<strong>2.</strong> Local transversality 20<br />

3.3. Globalizing 20<br />

3.4. Main theorem II: from symplectic structure to pencil 22<br />

4. Approximately holomorphic techniques over contact<br />

manifolds. 23<br />

4.<strong>1.</strong> Almost contact manifolds 23<br />

4.<strong>2.</strong> Giroux’s theorems 23<br />

4.3. Contact pencils 23<br />

Date: May, 201<strong>2.</strong><br />

1991 Mathematics Subject Classification. Primary: 53D10. Secondary: 53D15,<br />

57R17.<br />

Key words and phrases. contact structures, Lefschetz pencils.<br />

The author was supported by the Spanish National Research Project MTM2010-<br />

17389.<br />

1


2 FRANCISCO PRESAS<br />

4.4. Existence of contact structures 23<br />

References 23<br />

<strong>1.</strong> <strong>Introduction</strong><br />

This series of lectures is intended to be an overview of the relation<br />

between Lefchetz pencils and symplectic and contact structures. We<br />

do not try to give precise proofs, but just a feeling of how the whole<br />

theory works. The following readings are recommended to formalize<br />

the outlines of the proofs provided in the lectures:<br />

(i) Complex geometry basics: first chapters of the book [We].<br />

(ii) Basic Picard-Lefschetz theory: [La].<br />

(iii) Symplectic pencils: Donaldson’s articles. The introductory one<br />

is [Do3]. Details provided in [Do2, Do1].<br />

(iv) Basic introduction to open books in contact geometry: introductiry<br />

article [Gi].<br />

(v) Contact pencils: basics [Pr1].<br />

<strong>2.</strong> <strong>Classical</strong> pencils<br />

<strong>2.</strong><strong>1.</strong> Quick review of some basic complex geometry. Let X be<br />

a smooth manifold with a fixed atlas {Vi, φi}i∈I. We say that the<br />

manifold is complex if φi : Vi → φi(Vi) ⊂ R 2n C n can be chosen to<br />

satisfy that φ −1<br />

j ◦ φi are bi-holomorphic diffeomorphisms.<br />

A complex vector bundle over a smooth manifold X is a pair (E, π :<br />

E → X) such that:<br />

(i) E is a smooth manifold,<br />

(ii) π : E → X is a surjective smooth submersion,<br />

(iii) There exists a covering by open subsets {Uj}j∈J, satisfying that<br />

there are diffeomorphisms ψj : π −1 (UJ) → Uj × C r making the<br />

following diagram commutative<br />

π −1 (Uj)<br />

ψj<br />

→ Uj × C r<br />

π ↓ π1 ↓<br />

id<br />

Uj → Uj.<br />

(iv) The map ψ −1 <br />

i ◦ ψj : Ui Uj → Diff(C r ) reduces to ψij =<br />

ψ −1<br />

i ◦ ψj<br />

<br />

: Ui Uj → GL(Cr ).<br />

The last condition implies that the fibers of the map π are canonically<br />

complex vector spaces. The dimension of those vector spaces, the<br />

constant r, is called the rank of the vector bundle.


<strong>LEFSCHETZ</strong> <strong>PENCILS</strong> 3<br />

A morphism of complex vector bundles Φ : E → F is a smooth map<br />

making the diagram commute<br />

E Φ → F<br />

πE ↓ πF ↓<br />

X id → X,<br />

and such that the restriction to any fiber is a linear map. The morphism<br />

is called monomorphism whenever is injective, epimorphim whenever<br />

is surjective and isomorphism if it is bijective.<br />

A vector bundle E is completely determined, up to bundle isomorphism,<br />

by the maps {ψij} for any covering. In the other hand, any such<br />

collection of maps determines a vector bundle if it satisfies the cocycle<br />

condition<br />

ψjk ◦ ψij = ψik.<br />

These maps associated to the bundle are called the transition functions.<br />

With the help of them we can define various operations in the vector<br />

bundle category:<br />

• Let E have transition functions {ψij} and let F have transition<br />

functions {ρij}. We define the direct sum vector bundle as the<br />

bundle E F with transition functions<br />

<br />

ψj 0<br />

0 ρij<br />

• In the same spirit E ⊗ F is defined by the transition functions<br />

ψij ⊗ ρij.<br />

• The dual bundle E ∗ is defined by the transition functions ψ ∗ ij.<br />

• Any other conceivable operation with vector spaces has a translation<br />

to the category of vector bundles.<br />

Let X be a smooth complex manifold. We define a holomorphic<br />

vector bundle E over X to be a complex vector bundle E → X such<br />

that the maps ψij are holomorphic. This, in particular, implies that E<br />

is a complex manifold and π is a holomorphic submersion.<br />

The natural complex vector space structure of the fibers of a complex<br />

vector bundle E induces an isomorphism of the bundle J : E → E<br />

defined by the complex multiplication by i. It satisfies J 2 = −id.<br />

For any complex manifold, the tangent bundle T M → M has a natural<br />

structure of holomorphic vector bundle. Therefore, in particular<br />

a complex structure, i.e. an operator J : T M → T M. We say that a<br />

manifold is almost-complex if the tangent bundle possesses such operator.<br />

Realize that almost-complex manifold is a notion much weaker<br />

than that of complex manifold.<br />

<br />

.


4 FRANCISCO PRESAS<br />

The cotangent bundle T ∗M of an almost-complex manifold, therefore,<br />

has a complex structure as well. Define T ∗MC = T ∗M ⊗C, that it<br />

is naturally identified with the complex valued 1-forms. We have a natural<br />

division of that vector space between T ∗MC = T 1,0M T 0,1M, corresponding to the eigen-values i and −i of the J operator. We define<br />

E p,q = p 1,0 q 0,1 T M T M, its elements are called the (p, q)-forms.<br />

Moreover, for any fix complex vector bundle E we define Ep,q = E⊗E p,q<br />

whose elements are called the E-valued (p, q)-forms. Also, we have<br />

Em = E ⊗ m ∗ T MC. Therefore we obtain<br />

E m = <br />

E p,q .<br />

p+q=m<br />

Let E π → X be a complex vector bundle. We say that a map σ :<br />

X → E is a section of the bundle if π ◦ σ : X → X is the identity. The<br />

space of sections Γ(E) is a non-trivial vector space (it always possesses<br />

the zero section). For instance, the addition of elements is defined as<br />

(σ1 + σ2)(x) = σ1(x) + σ2(x), ∀x ∈ X, taking advantage of the vector<br />

space structure of the fibers. We can define sections over any open set<br />

U ⊂ X and we denote the vector space that they produce as Γ(U, E).<br />

In the same fashion we can define holomorphic sections as sections<br />

that satisfy the additional requirement of being holomorphic. Denote<br />

by H 0 (X, E) the vector space of holomorphic sections of a holomorphic<br />

vector bundle over a complex manifold. Some known facts:<br />

- If X is closed, then H 0 (X, E) is finite dimensional.<br />

- There exists always a choice of morphism ¯ ∂ : Γ(U, E) → Γ(U, E 0,1 ),<br />

such that ker ¯ ∂ = H 0 (U, E). It corresponds to take the anticomplex<br />

part of the differential of the map.<br />

- There is a formula relating its dimension to the topology of the<br />

space and of the bundle. It is called the Riemann-Roch formula.<br />

<strong>2.</strong><strong>2.</strong> Positive bundles. An hermitian metric h for a complex bundle<br />

E is a section of the bundle E ∗ ⊗ E ∗ satisfying that h|Ep : Ep × Ep → C<br />

is a hermitian metric of the complex vector space Ep.<br />

A connection ∇E for the vector bundle is a C-linear mapping ∇E :<br />

Γ(U, E) → Γ(U, E 1 ) satisfying<br />

(1) ∇E(fs) = df ∧ s + f ∧ ∇E(s),<br />

for any function f : U → C and section s ∈ Γ(U, E). We have that<br />

an hermitian metric is compatible with ∇E if the following equation is<br />

satisfied<br />

dh(s1, s2) = h(∇Es1, s2) + h(s1, ∇Es2),<br />

for any pair of sections s1, s<strong>2.</strong>


<strong>LEFSCHETZ</strong> <strong>PENCILS</strong> 5<br />

Lemma <strong>2.</strong><strong>1.</strong> For any complex hermitian vector bundle E, there exists<br />

a compatible connection. Moreover, if the bundle is holomorphic and<br />

we decompose the connection as ∇E = ∂ + ¯ ∂, there exists a unique<br />

compatible connection for which ¯ ∂(s) = 0, for any holomorphic section.<br />

There is a natural way of extending a connection ∇E : Γ(U, E) →<br />

Γ(U, E 1 ) to a C-linear map<br />

∇E : Γ(U, E r ) → Γ(U, E r+1 ),<br />

satisfying equation (1). We can define, then, the operator ΘE := ∇E ◦<br />

∇E : Γ(U, E) → Γ(U, E 2 ) that is called the curvature of the hermitian<br />

complex bundle. We have the following list of facts:<br />

- The operator is C ∞ -linear, i.e. Θ(fs) = fΘ(s), for any function<br />

f ∈ Γ(U, C) and section s ∈ Γ(U, E). Therefore, it is a section<br />

of (E ∗ ⊗ E ∗ ) 2 .<br />

- Moreover it is of type (E ∗ ⊗ E ∗ ) 1,1 .<br />

- dΘ = 0.<br />

- iΘ is a real form. (complexification of a real one)<br />

Let us restrict ourselves to the case of rank 1 holomorphic vector<br />

bundles. They are usually called line bundles. Then the curvature is<br />

a 2-form Θ that by the previous equations satisfies that is closed (i.e<br />

dΘ = 0). We have that<br />

Definition <strong>2.</strong><strong>2.</strong> A hermitian line bundle with connection is called positive<br />

if iΘ is a symplectic form 1 .<br />

Recall that the condition just imposes the non-degeneracy of the<br />

form, since the closedness is guaranteed by the previous results without<br />

any extra hypothesis.<br />

Fix a holomorphic line bundle L. Assume that H 0 (X, L) = 0. Fix<br />

a linear subspace L ⊂ H 0 (X, L), it is usually called a linear system.<br />

Define the base point set of L as<br />

BL = {x ∈ X : s(x) = 0, ∀s ∈ L}.<br />

Then there is a holomorphic map defined as<br />

φL : X \ BL → PL ∗<br />

x → P({φ ∈ L ∗ : φ(s(x)) = 0, ∀s ∈ L}).<br />

The map defined by the whole H 0 (X, L) is called the Kodaira map.<br />

1 I will assume some basic knowledge of symplectic geometry, the level provided<br />

by the first chapters of [MS]


6 FRANCISCO PRESAS<br />

Theorem <strong>2.</strong>3 (Kodaira’s embedding.). Let X be a complex manifold<br />

and L be an hermitian positive line bundle with connection over it. For<br />

any sufficiently large k > 0 positive integer, the Kodaira map associated<br />

to L ⊗k is an embedding.<br />

In fact the dimension of the target projective space is<br />

<br />

H 0 (X, L ⊗k ) = k n<br />

X Θn<br />

n! + O(kn−1 ),<br />

with n = dimCX. So, it roughly, increases with k n times the symplectic<br />

volume of X with respect to the symplectic form Θ.<br />

<strong>2.</strong>3. Lefschetz pencils.<br />

Definition <strong>2.</strong>4. A Lefschetz pencil for the complex manifold X and for<br />

a positive line bundle L is a choice of a 2-dimensional L linear system<br />

in H 0 (X, L).<br />

Recall that it is equivalent to fix 2 non-zero sections s1, s2 (a basis of<br />

the subspace). The map φL with respect to that fixed basis becomes<br />

φL : X \ BL → CP 1<br />

x → [s1(x) : s2(x)].<br />

We call Lefschetz pencil to this map slightly abusing notation. We can<br />

think of the map φL as the composition of the Kodaira map with the<br />

projection<br />

(2)<br />

π : CP n \ CP n−2 → CP 1<br />

[X0 : X1 : · · · : Xn] → [X0 : X1].<br />

In other words, any Lefschetz pencil is the restriction of the standard<br />

projective pencil given by equation (2).<br />

We say that a complex manifold equipped with a bundle L which<br />

provides an embedding into PH 0 (X, L) is a polarized complex variety.<br />

Definition <strong>2.</strong>5. The dual variety of a polarized complex manifold X ⊂<br />

CP N is the set defined as<br />

X ∗ = {H ∈ (CP n ) ∗ : H is tangent to X}.<br />

Theorem <strong>2.</strong>6 (Kodaira dual theorem). For a complex manifold polarized<br />

by a positive line bundle L ⊗k , for any k large enough, the dual<br />

variety is an irreducible algebraic variety of codimension 1 whose singularity<br />

locus is of positive codimension.<br />

A polarization with such a dual variety will be called extra ample.<br />

Thanks to this, we have the following definition


CP n-2<br />

<strong>LEFSCHETZ</strong> <strong>PENCILS</strong> 7<br />

Figure <strong>1.</strong> Pencil in CP N and restriction to X. The red<br />

sub-manifolds are the fibers of the pencil.<br />

Definition <strong>2.</strong>7. A pencil L for an extra ample polarization is called<br />

transverse if it intersects (as a line in PH 0 (X, L)) the dual variety<br />

transversely at smooth points of it.<br />

Then, we have<br />

Corollary <strong>2.</strong>8. There are transverse pencils for any extra ample polarization.<br />

Proof. It is a direct consequence of the Bertini’s theorem since X ∗<br />

and L have complementary dimensions and generically they intersect<br />

transversely (see Figure 2). <br />

The points of intersection of a Lefschetz pencil with the dual variety<br />

are singular values of the projection map φL, if the pencil is transverse,<br />

then the map has isolated critical points {pj} and for each of them<br />

pj ∈ X \ BL, there are holomorphic coordinates (z1, . . . , zn) around<br />

pj ∈ X and around φL(pj) ∈ CP 1 for which the map is written as<br />

φL(z1, . . . , zn) = z 2 1 + · · · z 2 n.<br />

X


8 FRANCISCO PRESAS<br />

L<br />

p 1<br />

Figure <strong>2.</strong> Pencil in (CP N ) ∗ and intersection with X ∗ .<br />

The points of intersection are the critical fibers.<br />

<strong>2.</strong>4. Symplectic structures and monodromy. We are in the position<br />

of giving a purely topological definition of Lefschetz pencil.<br />

Definition <strong>2.</strong>9. Let X be a smooth closed oriented 2n-dimensional<br />

manifold. An oriented smooth pencil on it is a triple (B, C, f) that<br />

conforms the following conditions:<br />

p 2<br />

(i) B is a codimension 4 submanifold,<br />

(ii) C is a set of points C = {pj} ∈ X \ B,<br />

(iii) f : X \ B → CP 1 is a smooth map that is a submersion away<br />

from C,<br />

(iv) for any point ∈ B, there are oriented coordinates (z1, . . . , zn)<br />

around it such that in those coordinates B = {z1 = z2 = 0} and<br />

moreover f(z1, . . . , zn) = z2<br />

z1 ,<br />

(v) for any point p ∈ C, there are oriented coordinates (z1, . . . , zn)<br />

around it such that f(z1, . . . , zn) = z2 1 + · · · + z2 n.<br />

X *


<strong>LEFSCHETZ</strong> <strong>PENCILS</strong> 9<br />

A transverse Lefschetz pencil of a polarized manifold X is a smooth<br />

pencil over it. It has more structure because it also satisfies the following<br />

Definition <strong>2.</strong>10. An oriented Lefschetz pencil over a symplectic manifold<br />

(M, ω) is compatible with the symplectic structure if<br />

(i) The sub-manifold B is symplectic,<br />

(ii) The fibers of the map f are symplectic sub-manifolds away from<br />

the critical points C.<br />

Away from the sets B and C the pencil is a fibration. We have the<br />

following general fact about symplectic fibrations<br />

Lemma <strong>2.</strong>1<strong>1.</strong> Let π : X → B a smooth fibration of a symplectic manifold<br />

(X, ω) with symplectic fibers. There is a natural connection H on<br />

X making the parallel transport 2 by symplectomorphims. Moreover, the<br />

parallel transport along contractible loops can be chosen to be through<br />

Hamiltonian diffeomorphisms.<br />

Proof. The connection is defined by the equation<br />

H(x) = {v ∈ T xX : ω(v, w) = 0, ∀w ∈ T xX such that dπx(w) = 0},<br />

i.e. we are taking the symplectic orthogonal to the symplectic subspace<br />

ker dπx. As usual whenever we have a path γ : [0, 1] → B, there is a<br />

unique lift ˆγx0 provided by the three following conditions<br />

• ˆγx0(0) = x0,<br />

• π(ˆγx0(t)) = γ(t), ∀t ∈ [0, 1],<br />

• ˆγ ′ x0 (t) ∈ H(ˆγx0(t)), ∀t ∈ [0, 1].<br />

This is true by the existence and uniqueness theorem for ODE’s.<br />

Recall that we need a compactness condition to guarantee the existence<br />

of the flow (not just the derivative of it), but we assume it for now. By<br />

using this construction, for any path γ : [0, 1] → B we can produce the<br />

diffeomorphism<br />

p 1 0 : π −1 (γ(0)) → π −1 (γ(1))<br />

x → ˆγx(1).<br />

To check that it preserves the symplectic structure we have just to<br />

check it infinitesimally. Define P 1 0 = π −1 (γ{[0, 1]} that is an immersed<br />

sub-manifold with self-intersections at the self-intersection points of γ.<br />

Denote X(x) = ˆγx(0) ′ . We just have to check that<br />

(3) LXω|P = 0.<br />

2 There is an extra condition to be satisfied related with the completeness of the<br />

connection that we will have to check case by case


10 FRANCISCO PRESAS<br />

Using the Cartan formula we obtain<br />

LXω|P = diXω|P + iXdω|P = diXω|P .<br />

Now recall that X ∈ H and H is symplectically orthogonal to T Ft, so<br />

iXω|P (v) = 0, ∀v ∈ T Ft. Finally we have iXω(X) = 0 and therefore<br />

diXω|P = 0 as we wanted to show. So, being true formula (3), we<br />

obtain that m is a symplectomorphism.<br />

Check the details of the Hamiltonian case in [MS]. <br />

<strong>2.</strong>5. Monodromy of a pencil. The parallel transport of a Lefschetz<br />

pencil is called the monodromy of it. Let us study it. There is a<br />

standard way to transform a pencil into a fibration, it amounts to<br />

a surgery operation on X called a symplectic blow-up (respectively<br />

complex blow-up in the projective case). Let us describe it from a<br />

topological viewpoint skipping the details of the construction.<br />

Start by a codimension 2k (k > 1) symplectic sub-manifold B inside<br />

X. Denote by (ν(B), η) its symplectic normal bundle. We have that<br />

the symplectic structure in a small neighborhood of B is completely<br />

determined by the bundle ν(B). Choose a compatible almost complex<br />

structure J on ν(B) to make it a complex bundle. Now replace the zero<br />

section ZB (that is diffeomorphic to B) by the sub manifold Pν(B) →<br />

B, i.e. the projectivization of the bundle. This can be understood as<br />

reversing the radial coordinate<br />

r → 1<br />

r ,<br />

and adding the infinity hyperplane (by the reversion it shows up at<br />

the origin). Moreover, to have an oriented manifold we place Pν(B)<br />

with the opposite complex orientation (see Figure 3). The new bundle<br />

is denoted ˜ν(B). The surgery takes place close to the zero section,<br />

therefore we can glue back to M \ B to produce a new manifold ˜ M.<br />

It possess a symplectic structure inherited by that of M. The submanifold<br />

E := Pν(B) is called the exceptional divisor and we have<br />

that<br />

˜M \ E → M \ B<br />

is a symplectic diffeomorphism. It is important to observe that there<br />

is a natural projection ˜ M π → M that restricts to the previous symplectomorphism<br />

away from E.<br />

There are multiple choices to be done in the blow-up surgery, in<br />

particular the identification of the normal bundle of B with the neighborhood<br />

of the sub-manifold. In the particular case of a Lefschetz<br />

pencil, we can be careful to ensure the following


<strong>LEFSCHETZ</strong> <strong>PENCILS</strong> 11<br />

ν(B) 0<br />

0<br />

∼<br />

ν(B)<br />

P(ν(Β))<br />

Figure 3. Blow-up scheme. The zero section gets replaced<br />

by P(ν(B))<br />

Lemma <strong>2.</strong>1<strong>2.</strong> Let (f, B, C) a Lefschetz pencil, the blow-up of X along<br />

B has a natural induced Lefschetz pencil structure defined as (π ◦<br />

f, π −1 (B), ∅).<br />

This just means that the Lefschetz pencil becomes a singular fibration,<br />

i.e. the projection map is defined everywhere. Realize that now<br />

the fibers are compact and the parallel transport, as previously defined,<br />

is global. For the following discussion we will be assuming that we have<br />

a Lefschetz fibration, i.e. B = ∅, maybe after blowing-up.<br />

Fix a point z0 ∈ CP 1 and also fix the set of critical values ∆ =<br />

{z1, . . . , zk}. Choose a smooth path γ : S 1 → CP 1 \ ∆, with γ(0) = z0,<br />

this induces a symplectomorphism on the fiber F0 = π −1 (z0) by means<br />

of Lemma <strong>2.</strong>1<strong>1.</strong> Define the symplectic mapping class group of the fiber<br />

as Map(F0) = Symp(F0)/Ham(F0). We have that there is an induced<br />

map<br />

m1 : π1(CP 1 \ ∆, z0) → Map(F0),


12 FRANCISCO PRESAS<br />

called the geometric monodromy representation. Fix a set of disjoint<br />

paths γi : [0, 1] → CP 1 satisfying the following conditions:<br />

(i) γi(0) = z0,<br />

(ii) γi(1) = zi,<br />

(iii) γi is an embedded path.<br />

This is usually called an arc system. We construct an associated set<br />

of paths of π1(CP 1 \ ∆, z0) by choosing small positive embedded circle<br />

paths ci : S1 → CP 1 \ ∆ around each zi. We, then, define li = γ −1<br />

i ◦<br />

ci ◦ γi (see Figure 4). It is clear that {li} are a set of generators of<br />

π1(CP 1 \ ∆, z0). There is a unique relation among them expressed as<br />

(4) l1 ◦ · · · ◦ ln = 1<br />

Therefore, there is a special word in the mapping class group Map(F0)<br />

written as<br />

z 6<br />

z 5<br />

z 1<br />

z 0<br />

z 4<br />

Figure 4. Paths associated to an arc system.<br />

(5) m1(l1) ◦ m1(l2) ◦ · · · ◦ m1(lk) = id.<br />

z 2<br />

z 3


<strong>LEFSCHETZ</strong> <strong>PENCILS</strong> 13<br />

This is just the translation through the morphism m1 of the relation (4).<br />

We have to study in more detail the element m1(li) to geometrically<br />

understand what is happening.<br />

Definition <strong>2.</strong>13. An n-dimensional sub-manifold i : N → M inside a<br />

2n-dimensional symplectic manifold (M, ω) is Lagrangian if i ∗ ω = 0.<br />

It is known (due to Weinstein) that a small tubular neighborhood U<br />

of a Lagrangian sub-manifold L is symplectomorphic to a small tubular<br />

neighborhood of the zero section of the cotangent bundle T ∗ L with its<br />

canonical symplectic structure.<br />

For any diffeomorphism g : M → N, there is a canonical lift g ∗ :<br />

T ∗ N → T ∗ M that is a symplectomorphism.<br />

Recall that a Hamiltonian diffeomorphism φ : M → M is completely<br />

characterized by its Hamiltonian function H : M → R, the Hamiltonian<br />

is recovered out of the function as the time 1 flow associated to the<br />

vector field XH defined by the equation<br />

iXHω = dH.<br />

Denote by e : Sn → Sn the antipodal map over the sphere, i.e. e(v) =<br />

−v. We can give the following<br />

Definition <strong>2.</strong>14. The Dehn twist associated over T ∗ S n is the composition<br />

of the map e ∗ with the Hamiltonian H(v) = ||v|| (choosing an<br />

arbitrary metric and smoothing out at the origin)<br />

This map is compactly supported since the Hamiltonian defines the<br />

geodesic flow at time 1 and the composition e ∗ ◦φH is the identity away<br />

from an arbitrary small neighborhood of the origin. We can, therefore,<br />

use the canonical tubular neighborhood theorem for any Lagrangian<br />

sphere L to define a Dehn twist φL supported on it. Due to the various<br />

choices in the construction, it works up to Hamiltonian isotopy.<br />

Therefore, it defines an element [φL] ∈ Map(M).<br />

Given a path γi as previously defined we can construct the set<br />

Si = {x ∈ F0 : γx0(1) = pi}.<br />

Since, the parallel transport is not correctly defined at the end of the<br />

path, it actually happens that there is not unicity and the set Si is not<br />

a point. It is actually a Lagrangian sphere, that is called the vanishing<br />

sphere (or cycle) for the singular value zi.<br />

Lemma <strong>2.</strong>15. The monodromy for the path li is provided by the Dehn<br />

twist along the vanishing sphere Si associated to the critical value zi.<br />

So, we have that the monodromy of any path is always the composition<br />

of a series of Dehn twists.


14 FRANCISCO PRESAS<br />

We may wonder about the uniqueness of the word provided by equation<br />

(5). There is an exhaustive action of the braid group Bk over the<br />

disk on the arc systems. Recall that the braid group is defined as the<br />

mapping class group of the puntured disk D 2 \{z1, . . . , zk} (the infinity<br />

being z0). The action is the obvious composition. This induces new<br />

arc systems with new associated words. (see Figure 5)<br />

z 1<br />

z 0<br />

=<br />

z 2<br />

z 3<br />

z 1<br />

+<br />

Figure 5. Composition with a braid element of a path system.<br />

<strong>2.</strong>6. Main equivalence theorem I: from pencil to symplectic<br />

structure. We restrict ourselves to the 4-dimensional case in the following<br />

discussion. We want to state that the monodromy representation<br />

provided by the word in formula (5) recovers the symplectic type<br />

of the manifold up to deformation of the symplectic structure. There<br />

are a lot of subtleties that we will skip:<br />

- We consider fibrations instead of pencils. The case of a pencil<br />

amounts to the study of geometric monodromy in symplectic<br />

manifolds with boundary (we remove the base points B on<br />

each fiber producing fibers with boundary). The argument goes<br />

z 0<br />

z 1<br />

z 2<br />

z 3<br />

z 2<br />

z 3<br />

z 1<br />

z 2<br />

z 3


<strong>LEFSCHETZ</strong> <strong>PENCILS</strong> 15<br />

through but it is more technical. For a good summary see<br />

[AMP].<br />

- We restrict ourselves to the dimension 4 case. The statement<br />

is partially true in higher dimensions, but a long discussion is<br />

required. See [Go] for the more general statements.<br />

First an auxiliary construction<br />

Lemma <strong>2.</strong>16. For any genus g ≥ 1 surface Σg and any embedded loop<br />

L ∈ Σg, there exists a fibration f : X → D 2 with a unique singular<br />

point at the origin of the disk and such that:<br />

(i) it conforms the local model (v) in Definition <strong>2.</strong>9,<br />

(ii) the vanishing cycle associated to the singularity is L.<br />

Proof. Start by a holomorphic Lefschetz pencil φ2 : CP 2 \ B4 → CP 1<br />

defined by the line bundle of conics L ⊗2 in CP 2 . Blow it up along the<br />

4 points of the base locus B4 to obtain a singular Lefschetz fibration<br />

ˆφ2 : CP 2 → CP 1 . The map has 3 singularities and the fibers are spheres<br />

(they are conics on CP 2 ). Choose the neighborhood of a singular value<br />

z0 and center a small chart around it U0. Define φ0 = ˆ φ2 restricted to<br />

ˆV0 = ˆ φ −1<br />

2 (U0). Therefore we have a Lefschetz fibration over the disk by<br />

spheres φ0 : ˆ V0 → U0 with vanishing cycle L0.<br />

Since the monodromy is supported at a tubular neighborhood of L0<br />

we can take L 1 0 and L 2 0 two copies of L0 at one fiber and transport<br />

them to produce a pair of 3 dimensional manifolds: ∂V 1<br />

0 the parallel<br />

transport of L0 and ∂V 2<br />

0 the parallel transport of L<strong>2.</strong> We take V0<br />

to be the domain inside ˆ V0 whose boundary is ∂V 1 2<br />

0 ∂V0 and that<br />

contains the singular point of the fibration. We therefore obtain a<br />

fibration φ0 : V0 → U0 by cylinders with monodromy generated by the<br />

Dehn twist around L0. Denote by ¯ φ0 : ∂V0 → U0 its restriction to the<br />

boundary of V0.<br />

Define ˆ Ψ0 : Σg × U0 → U0 to be the trivial Lefschetz fibration over<br />

the disk. Take U(L) a small neighborhood of L in Σg. Set Ψ0 : W0 =<br />

(Σg \ U(L)) × U0 → U0. Restricted to the boundary we get ¯ Ψ0 :<br />

∂U(L) × U0 → U0. Under the natural identification ∂U(L) = ∂V0 we<br />

have that ¯ Ψ0 = ¯ φ0. We can glue the two fibrations along this common<br />

boundary to obtain f = Ψ0#φ0 : W0#V0 → U0 that is a Lefschetz<br />

fibration satisfying all the requirements. <br />

Theorem <strong>2.</strong>17. Let L1, . . . Lk be an ordered set of embedded loops in<br />

Σg such that φLk ◦ · · · ◦ φL1 = id. Assume that g > 1, then there exists<br />

a unique symplectic Lefschetz fibration up to deformation of the symplectic<br />

structure with an arc system providing the given set of vanishing<br />

cycles.


16 FRANCISCO PRESAS<br />

Remark that the symplectic mapping class group Map(Σg) is isomorphic<br />

to the topological mapping class group<br />

Mapt(Σg) = Diff + (Σg)/Diff + 0 (Σg).<br />

This is because the symplectomorphism group in the case of Riemann<br />

surfaces corresponds to the area preserving diffeomorphism group and<br />

it is well-known that the area preserving diffeomorphisms retract to<br />

the orientation preserving ones.<br />

Proof. First we want to recover a differentiable manifold out of the<br />

2 π0 2 combinatorial data. We start with X0 = Σg × D → D . We fix a<br />

genus g Lefschetz fibration as provided by Lemma <strong>2.</strong>16 with vanishing<br />

cycle Li denoted as πk : Vk → Uk.We perform the fiber-connected sum<br />

of the fibrations π0, π1, . . . , πk that gives a fibration over the larger<br />

disk V = D 2 #U1# · · · #Uk. So, we clearly get a fibration Xk → V<br />

over a disk that has an arc system for which the associated word in the<br />

mapping class group is the required one.<br />

Now, it it left to glue another disk to close the infinity. The gluing<br />

morphism is an element λ : S 1 = ∂V → Diff + 0 (Σg). So, the diffeomor-<br />

phim type of the constructed manifold depends only on the homotopy<br />

class of the loop λ inside the space Diff +<br />

0 (Σg). But it is well-known that<br />

for g > 1, the fundamental group of the space Diff + 0 (Σg) is trivial.<br />

Therefore, any choice of λ leads to diffeomorphic smooth manifolds; we<br />

will denote the so built manifold as X. Moreover, there is a section<br />

e : CP 1 → X. This is because we can choose a point in the central<br />

fiber Σg that is away from all the support domains of the Dehn twists<br />

and so the parallel transport of that point provides the section.<br />

As for the symplectic structure we use an adaptation of a Thurston’s<br />

argument due to Gompf (and Donaldson). Represent the Poincaré dual<br />

of the sub-manifold e(CP 1 ) as a closed 2-form τ. Take a covering Ui of<br />

CP 1 by open contractible sets such that each one contains at most one<br />

critical value. We distinguish two cases:<br />

(i) Ui does not contain a critical point. Then we have that fUi :<br />

f −1 (Ui) (Σg × Ui) → Ui, and we can equip Vi = f −1 (Ui) with<br />

the standard symplectic product structure induced out of the<br />

pair of symplectic structures ωUi (restriction of the standard<br />

symplectic structure over CP 1 ) and ωΣg. The choice of ωΣg is<br />

such that the restriction of τ to the fiber is cohomologous to it.<br />

This implies just that<br />

<br />

f −1 <br />

ωΣg =<br />

(pt)<br />

f −1 τ = 〈f<br />

(pt)<br />

−1 (pt), e(CP 1 )〉 = 1,


<strong>LEFSCHETZ</strong> <strong>PENCILS</strong> 17<br />

i.e it has to have volume <strong>1.</strong><br />

(ii) Ui contains a critical point. We place a symplectic structure<br />

on W0 as in the previous case. We have a canonical symplectic<br />

structure on it, because it is a holomorphic Lefschetz fibration<br />

V0. To check that they glue along the boundary we slightly<br />

thicken it, we have that in both cases is a band the thickening<br />

is a band inside T ∗ (S 1 × [−ɛ, ɛ]). Therefore, by Weinstein’s<br />

tubular neighborhood theorem, they are symplectomorphic. To<br />

setup the symplectomorphism we may need to slightly shrink<br />

Ui. We repeat the same trick for making the symplectic form<br />

ωi restricted to the regular fibers cohomologous to τ.<br />

Fix a partition of the unity {χi} subordinated to the covering Ui. We<br />

select 1-forms λi over each Ui such that<br />

Define<br />

ωi − τ = dλi.<br />

ˆτ = τ + Σid(χi ◦ f)λi.<br />

It is a closed 2-form that is symplectic when restricted to any vertical<br />

fiber. We define<br />

ωX,K = ˆτ + Kf ∗ (ω CP 1)<br />

that is symplectic for any large K > 0.<br />

In [Go], you can see why the construction is unique up to deformation<br />

of the sympplectic structure. The intuition is just that in 2dimensions<br />

the space of symplectic structures is contractible and a<br />

Lefschetz fibration is (up to standard singularities) a cartesian product<br />

of 2-dimensional symplectic spaces. <br />

3. Approximately holomorphic geometry<br />

The goal of this Section is to reproduce the results provided in the<br />

first pages of these notes in the projective setting, now for a general<br />

symplectic manifold. We can formally copy the setup just by defining<br />

an ”approximately holomorphic geometry”. Most of the Section is<br />

about creating an approximately holomorphic section. The last Subsection<br />

adapts the construction to create a approximately holomorphic<br />

pencil (complex line of sections).<br />

Let (M, ω) be a symplectic manifold of integer class, i.e. the cohomology<br />

class [ω] ∈ H 2 (M, R) admits an integer lift 3 . Therefore, there<br />

is a complex line bundle with connection L such that the curvature of<br />

3 The form is in the image of the map H 2 (M, Z) → H 2 (M, R).


18 FRANCISCO PRESAS<br />

the connection satisfies curv(∇L) = −iω. We fix a compatible almostcomplex<br />

structure J in the manifold M. We mean by this an endomorphism<br />

of the tangent bundle J : T M → T M such that J 2 = −Id and<br />

g(u, v) = ω(u, Jv) is a Riemannian metric. The space of such adapted<br />

almost-complex structures is non-empty and connected.<br />

We have a sequence of line bundles L k = L ⊗k , k ∈ Z + . They are<br />

naturally hermitian complex line bundles endowed with a connection<br />

∇k. Fix the sequence of Riemannian metrics gk(u, v) = kω(u, Jv). We<br />

may introduce the following definition<br />

Definition 3.<strong>1.</strong> We say that a sequence of sections sk : M → L k is<br />

asymptotically holomorphic if the following set of uniform estimates<br />

hold<br />

|sk| = O(1), | ¯ ∂sk|gk = O(k−1/2 ), |∇ r sk|gk = O(1), |∇r ¯ ∂sk|gk = O(k−1/2 ),<br />

for r = 1, <strong>2.</strong><br />

We also give the following<br />

Definition 3.<strong>2.</strong> A section s : M → L k is ɛ-transverse to zero over the<br />

domain U ⊂ M if at least one of the following conditions hold:<br />

(i) |s(x)| > ɛ,<br />

(ii) |∇ks(x)|ɛ.<br />

This condition sharpens the usual notion of transversality of a function.<br />

It gives a qualitative version of the usual transversality definition.<br />

Our goal is to use the following result:<br />

Proposition 3.3. Let sk : M → L k be an asymptotically holomorphic<br />

sequence of sections. Assume that for k large enough, the sections are<br />

ε-transverse to zero all over M. Then, the zero sets of the sections<br />

Z(sk) are smooth symplectic sub manifolds for k large.<br />

Proof. At x ∈ Z(sk), there is a unitary vector v ∈ TxM such that<br />

|∇vsk| > ɛ. At this point we have that<br />

| ¯ ∂vsk(x)| = O(k −1/2 ), |∂vsk(x)| > 3ɛ<br />

4 , |∂Jvsk(x)| > 3ɛ<br />

4 ,<br />

and so, we obtain that |∇Jvsk(x)| > ɛ/<strong>2.</strong> Thus, we have that ∇sk(x) is<br />

surjective at x ∈ Z(sk) and so the set Z(sk) is a smooth sub manifold<br />

by the implicit function theorem.<br />

By the same argument we have that<br />

|∂vsk(x)| >> | ¯ ∂vsk(x)|,<br />

for all x ∈ Z(sk). A simple linear algebra argument shows that the subspace<br />

TxZ(sk) = ker ∇sk(x) ⊂ (TxM, ω(x), J(x)) is symplectic. The


<strong>LEFSCHETZ</strong> <strong>PENCILS</strong> 19<br />

reason being that it is close to be complex, i.e. TxZ(sk) is approximately<br />

J-invariant. <br />

So it is left to show that such kind of sections do exist.<br />

3.<strong>1.</strong> Trivializations. We want to trivialize in an approximately holomorphic<br />

way the manifold and the bundle L k , in order to compute<br />

things in the euclidean space instead of the manifold. This is our goal<br />

now.<br />

Lemma 3.4. For any point x ∈ (M, kω), there is a chart (Uk :=<br />

Bgk (x, k1/2 ), φk : Uk → (C n , ω0)) that satisfies the following conditions:<br />

a) |(φk)∗J(v) − J0(v)| = ck −1/2 ||v||,<br />

b) φ ∗ k ω0 = kω.<br />

Proof. It is just a matter of selecting a standard Darboux chart ψk :<br />

Uk → C n and compose it with a linear map B ∈ Sp(2n, R), in such a<br />

way that B ◦ dφ(x) is complex-linear. This immediately provides all<br />

the bounds trivially. <br />

Now we go for the bundle<br />

Lemma 3.5. For any point x ∈ M, there is an asymptotically holomorphic<br />

sequence of sections σk,x : M → L k satisfying:<br />

(i) |σk,x(x)| = 1,<br />

(ii) |∇ r σk,x(y)| ≤ ce −d2 k (x,y)/5 ,, for r = 0, 1, 2, 3.<br />

(iii) |∇ r ¯ ∂σk,x(y)| ≤ ck −1/2 e −d2 k (x,y)/5 , for r = 0, 1, <strong>2.</strong><br />

Proof. Fix the Darboux chart φk : Uk → C n , obtained by Lemma 3.4.<br />

Recall that by means of that chart the bundle L k is pushed-forward to<br />

the hermitian complex bundle L0 over C n with curvature −iω0. Trivialize<br />

that bundle by parallel transport along radial directions starting at<br />

0 ∈ C n . We obtain that the connection in that trivialization becomes<br />

∇0 = d + 1<br />

4 (Σzid¯zi − ¯zidzi),<br />

and so we obtain<br />

¯∂0 = ¯ ∂ + 1<br />

4 (Σzid¯zi),<br />

It is an exercise to check that<br />

σ0 = e −|z|2 /4<br />

is a holomorphic section for this bundle. We, therefore, pull-back to<br />

obtain ˆσk,x = φ ∗ k σ0. Cutting-off by a suitable function, we obtain σk,x<br />

that satisfies all the required estimates.


20 FRANCISCO PRESAS<br />

3.<strong>2.</strong> Local transversality. To produce estimated transversality we<br />

need an estimated Sard Lemma. This is the content of the following<br />

result. Define<br />

Fp(δ) = δlog(δ −1 ) −p , δ > 0, p ∈ Z + .<br />

Theorem 3.6 (Theorem 20 in [Do1]). For σ > 0, let Hδ denote the<br />

set of functions f on B(0, 1) such that<br />

(i) ||f|| C 0 ≤ 1,<br />

(ii) ||∂f|| C 1 ≤ σ.<br />

Then there is an integer p. depending only on the dimension n, such<br />

that for any δ with 0 < δ < 1<br />

2 if σ ≤ Fp(δ), then for any f ∈ Hδ there<br />

is a w ∈ C with |w| ≤ δ such that f − w is Fp(δ)-trasnverse to zero<br />

over the region B(0, 1).<br />

3.3. Globalizing. We need to find an asymptotically holomorphic sequence<br />

of sections sk : M → Lk that are ɛ-transverse to zero all over<br />

M. The idea is that there are plenty of those sections and a clever<br />

choice makes the trick. Let us try an informal approach first. Take a<br />

finite number of big Darboux charts (Uj, φj), over each of them take<br />

the lattice L = (k−1/2Z) 2n ⊂ Cn ⊃ φj(Uj), the image of the lattice<br />

Lj = φ −1<br />

j (L) is a set of points {xij} on M. Recall that the number of<br />

points is O(kn ). We define the sequence<br />

sk = wijσk,xij ,<br />

with wij ∈ C such that |wij| ≤ <strong>1.</strong> It is trivial to check that it is asymptotically<br />

holomorphic. The conclusion is that the space of asymptotically<br />

holomorphic sequences is huge. The size of the space, informally<br />

speaking, grows as fast as in the holomorphic setting.<br />

Now, assume that there is a sequence sk fixed. We want to perturb<br />

it to make it ɛ-transverse to zero in a neighborhood of a point xij. We<br />

first trivialize the section by defining<br />

fij = sk<br />

,<br />

σk,xij<br />

that it is well-defined and with bounded derivatives in the ball Bgk (xij, 1).<br />

Now we use Lemma 3.4 to trivialize the manifold, we obtain ˆ<br />

fij =<br />

fij ◦ φk : B(0, 1) → C. It is a function that, for k large, satisfies the<br />

hypothesis of Theorem 3.6. We apply it and we obtain that ˆ fij − wij<br />

is transverse to zero. Going back we have that<br />

sk − wijσk,xij ,<br />

is transverse to zero over Bgk (xij, 1). Now, there are two obvious ways<br />

to proceed:


<strong>LEFSCHETZ</strong> <strong>PENCILS</strong> 21<br />

All together method.<br />

Do the previous process for all the points of the lattices at the same<br />

time, the set of balls Bgk (xij, 1) cover the manifold. We just have<br />

to control the interference between different perturbations. The ɛtransversality<br />

is C 1 -stable, i.e. let f : U ⊂ C n → C be a function<br />

ɛ-transverse to 0 over U, and let g : U ⊂ C n → C be a function satisfying<br />

that |g| C 1 ≤ δ, then we have that f −g is (at least) (ɛ−δ)-transverse<br />

to zero over U. We have that over a fixed point xij the norm of the<br />

rest of the perturbations created by the other points (recall that they<br />

exponentially decay) is around<br />

(6) AδΣ ∞ n=1e (−1/5)n2<br />

,<br />

for a fixed A > 0. We need that number to be smaller than δ and it is<br />

not.<br />

One by one method.<br />

Do the process ball by ball. Call δ1 the allowed amount of the perturbation<br />

in the first ball. We obtain σ1(= Fp(δ1))- transversality in<br />

the neighborhood of that point. To not completely destroy the obtained<br />

transversality we impose the following condition for the second<br />

perturbation<br />

δ2 = 1<br />

2 σ<strong>1.</strong><br />

So, denote σ2 = min{σ1/2, Fp(δ1)}, we obtain σ2 transversally in the<br />

first two balls. We go on to obtain at the end of the day σN-transversality<br />

all over the manifold. The problem is that N = O(k 2n ) and so, the<br />

transversality depends on k. Bad news!<br />

Mixed method.<br />

Mix the two methods. Construct D 2n sublattices (by taking (DZ) 2n ⊂<br />

Z 2n ), D > 0 large but independent of k. We will perturb simultaneously<br />

over each sub lattice. The sub lattice satisfies the property that<br />

any two points x and y verify dk(x, y) ≥ D. Then we have that the<br />

stability equation (equivalent to equation 6) is<br />

(7) σj ≥ AδΣ ∞ n=1e −(D/5)n2<br />

,<br />

for all j = 1, . . . D 2n . So D has to be chosen in such a way that it<br />

holds for all the sublattices (index j moving). It is required a study of<br />

the iteration provided by equation (7). This strongly depends on the<br />

function Fp(δ). Just for fun: we have that<br />

cδ ≥ Fp(δ) ≥ δ 1+α ,<br />

for any small positive constants c > 0 and α > 0 and for every sufficiently<br />

small value of δ. The linear function cδ is not achievable as a


22 FRANCISCO PRESAS<br />

result of a perturbation like te one performed in Theorem 3.6 (there<br />

are counter-examples). The function δ 1+α does not converge for the<br />

global iteration method just described. In other words, the function<br />

Fp(δ) is optimal in many ways.<br />

This concludes the proof of the existence of an asymptotically holomorphic<br />

sequence of sections, just proving that there are “approximately<br />

holomorphic” zero dimensional linear systems.<br />

3.4. Main theorem II: from symplectic structure to pencil.<br />

Now we will prove the converse of the Theorem <strong>2.</strong>17. It is stated<br />

as follows<br />

Theorem 3.7. Let (M, ω) be a closed symplectic manifold of integer<br />

class, for k sufficiently large there exists a symplectic Lefschetz pencil<br />

(f, B, C) over M such that the fibers are Poincaré dual to the class [kω]<br />

The statement about the homology class of the fibers just tells that<br />

the fibers are zeroes of sections of the bundle L k . So we can copy the<br />

arguments of the previous subsection. We need to slightly generalize<br />

the setup. Let us introduce the following definitions<br />

Definition 3.8. A section s : M → E of a hermitian vector bundle is<br />

ɛ-trasnverse to zero over the domain U if at least one of the following<br />

two conditions hold for any point x ∈ U:<br />

(i) |s(x)| ≥ ɛ,<br />

(ii) ∇s(x) is surjective and it admits a right inverse Rx such that<br />

|Rx| ≤ ɛ.<br />

This generalizes the notion of estimated transversality of a section of<br />

a line bundle. Now, we have that given a bundle E, we can construct<br />

the sequence E ⊗ L k , then we have<br />

Definition 3.9. We say that a sequence of sections sk : M → E ⊗ L k<br />

is asymptotically holomorphic if the following set of uniform estimates<br />

hold<br />

|sk| = O(1), | ¯ ∂sk|gk = O(k−1/2 ), |∇ r sk|gk = O(1), |¯ ∂∇sk|gk = O(k−1/2 ),<br />

for r = 1, <strong>2.</strong><br />

We can produce sequences of sections sk : M → E ⊗ L k ɛ-transverse<br />

to zero over M by slightly adapting the arguments of the previous<br />

Subsection.<br />

Let us sketch the proof of Theorem 3.7. Let us take an asymptotically<br />

holomorphic sequence of sections sk,1 ⊕ sk,2 : M → L k L k . We<br />

perturb the sequence to obtain the following transversality conditions:


<strong>LEFSCHETZ</strong> <strong>PENCILS</strong> 23<br />

(i) sk,1 is ɛ-transverse to zero,<br />

(ii) sk,1 ⊕ sk,2 is s ɛ-transverse to zero,<br />

is ɛ-transverse to zero, away from a neighborhood of the<br />

zero set Z(s,1).<br />

(iii) ∂ sk,2<br />

sk,1<br />

The second condition is the one that ensures the good picture around<br />

the base point set B = Z(sk,1 ⊕ sk,2). The first one guarantees that the<br />

zero fiber is symplectic and the critical values of the pencil are away<br />

from that fiber. The third one makes sure that the singularities follow<br />

the “approximately holomorphic” local model, and that the fibers are<br />

symplectic. This last statement is not completely true and in fact the<br />

map fk = [sk,1 ⊕ sk,2] : M − B → CP 1 needs a perturbation in order to<br />

satisfy what we claim. The main reason is that<br />

4. Approximately holomorphic techniques over contact<br />

manifolds.<br />

4.<strong>1.</strong> Almost contact manifolds.<br />

4.<strong>2.</strong> Giroux’s theorems.<br />

4.3. Contact pencils.<br />

4.4. Existence of contact structures.<br />

References<br />

[AMP] D. Auroux, V. Muñoz and F. Presas, Lagrangian submanifolds and Lefschetz<br />

pencils, J. Symplectic Geom. 3(2) (2005), 171–219.<br />

[BW] W. Boothby and H. Wang, On contact manifolds, Ann. of Math. 68 (1958),<br />

721–734.<br />

[CPP] R. Casals, D. Pancholi, F. Presas, Almost contact implies contact in dimension<br />

5, preprint.<br />

[Do1] Donaldson, S. K. Symplectic submanifolds and almost-complex geometry.<br />

J. Differential Geom. 44 (1996), no. 4, 666–705.<br />

[Do2] S. Donaldson, Lefschetz pencils on symplectic manifolds. J. Differential<br />

Geom. 53 (1999), no. 2, 205–236.<br />

[Do3] S. Donaldson, Lefschetz fibrations in symplectic geometry. Proceedings of<br />

the International Congress of Mathematicians, Vol. II (Berlin, 1998). Doc.<br />

Math. 1998, Extra Vol. II, 309–314.<br />

[El] Y. Eliashberg, Classification of overtwisted contact structures on 3–<br />

manifolds, Invent. math. 98 (1989), 623–637.<br />

[Ge] H. Geiges, An <strong>Introduction</strong> to Contact Topology, Cambridge studies in Advanced<br />

Mathematics 109. Cambridge University Press (2008).<br />

[Ge1] H. Geiges, Contact structures on 1–connected 5–manifolds, Mathematica 38<br />

(1991), 303–31<strong>1.</strong>


24 FRANCISCO PRESAS<br />

[Gi] E. Giroux, Gomtrie de contact: de la dimension trois vers les dimensions<br />

suprieures. (French) Proceedings of the International Congress of Mathematicians,<br />

Vol. II (Beijing, 2002), 405414, Higher Ed. Press, Beijing, 200<strong>2.</strong><br />

[Go] R. Gompf, Toward a topological characterization of symplectic manifolds.<br />

(English summary) J. Symplectic Geom. 2 (2004), no. 2, 17–206.<br />

[Ha] A. Hatcher, Algebraic Topology. Cambridge University Press (2002).<br />

[IM2] A. Ibort, D. Martínez–Torres, Lefschetz pencil structures for 2–calibrated<br />

manifolds, C. R. Acad. Sci. Paris, Ser. I. 339 (2004), 215–218.<br />

[IMP] A. Ibort, D. Martínez–Torres, F. Presas, On the construction of contact<br />

submanifolds with prescribed topology, Journal of Differential Geometry 56<br />

(2000), 2, 235–283.<br />

[La] K. Lamotke, The topology of complex projective varieties after S. Lefschetz.<br />

Topology 20 (1981), no. 1, 15–5<strong>1.</strong><br />

[MMP] D. Martínez–Torres, V. Muñoz, F.Presas, Open book decompositions for<br />

almost contact manifolds, Proceedings of the XI Fall Workshop on Geometry<br />

and Physics, Oviedo, 2002, Publicaciones de la RSME.<br />

[Ma] J. Martinet, Formes de contact sur les varieétés de dimension 3, in Proc.<br />

Liverpool Singularity Sympos. II, Lecture notes in Math. 209 (1971), 142–<br />

163.<br />

[MS] D. Mcduff, D. Salamon, <strong>Introduction</strong> to symplectic topology, Oxford Mathematical<br />

Monographs (1998).<br />

[Pr1] F. Presas, Lefschetz type pencils on contact manifolds, Asian J. Math. 6<br />

(2002), 2, 277–30<strong>1.</strong><br />

[We] R. O. Wells, Differential analysis on complex manifolds. Third edition. Graduate<br />

Texts in Mathematics, 65. Springer, New York, 2008. xiv+299 pp<br />

Instituto de Ciencias Matemáticas CSIC-UAM-UC3M-UCM, C. Nicolás<br />

Cabrera, 13-15, 28049, Madrid, Spain<br />

E-mail address: fpresas@icmat.es

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!