Patrick Lei

All notes

Symplectic topology

Lectures by R. İnanç Baykur, UMass, Spring 2020.

Symplectic Topology
Math 705
Notes by Patrick Lei, Spring 2020
Lectures by R. İnanç Baykur

University of Massachusetts Amherst

Disclaimer

These notes were taken during lecture using the vimtex package of the editor neovim. Any errors are mine and not the instructor’s. In addition, my notes are picture-free (but will include commutative diagrams) and are a mix of my mathematical style (omit lengthy computations, use category theory) and that of the instructor. If you find any errors, please contact me at plei@umass.edu.

1 January 21

1.1 Course Description

This is an introductory course on symplectic topology, along with its connections to differential, complex algebraic, and contact geometry and topology.

1.2 Organization

İnanç passed around a syllabus. Prerequisites for this course are smooth manifolds, some algebraic topology (cohomology), and complex analysis. Because this is a second year graduate course, we are expected to read any missing background on our own. Grading will be based on homework and a final presentation. We will cover the first four topics on the syllabus and do some of the last four. Finally, İnanç may add more geometry to the course.

1.2.1 Notational conventions

We will denote finite dimensional real vector spaces over ℝ by V and smooth manifolds by M.

1.3 Basic Notions

Definition 1.1.

A symplectic form (or a symplectic structure) on a vector space V is a nondegenerate alternating bilinear form V⊗V→ℝ.

Definition 1.2.

A symplectic form (or symplectic structure) on a smooth manifold M is a differential form ω∈Ω2⁢M which is closed and everywhere nondegenerate.

Remark 1.3.

A fundamental question to ask is when a manifold admits a symplectic structure. We will see that symplectic structures exist only on even-dimensional manifolds. Saying more is an extremely difficult problem, although we can say that symplectic manifolds admit an almost complex structure and are orientable. Uniqueness up to both symplectomorphism and deformation is also very difficult.

Example 1.4.

Let V=ℝ2⁢n with coordinates x1,…,xn,y1,…,yn. Then we can define a form

where

Checking that ω0 is a nondegenerate alternating bilinear form is easy. Later we will see that this is the only symplectic vector space.

Example 1.5.

Consider M=ℝ2⁢n with coordinates x1,y1,…,xn,yn. Now consider the form

Checking that this form is closed and nondegenerate is easy. We can also define the form on ℂn, where the form becomes

Note that the requirement that symplectic forms are closed is a subtle condition. If we multiply by a general function, the resulting form will not be closed. Recall that smooth manifolds are locally Euclidean. We can require our transitions to lie in the symplectic group, and later we will show that this definition of a symplectic manifold is equivalent to the one we gave today.

For a general manifold M, we want to associate linear spaces to it.

Definition 1.6.

A vector bundle E over M is a symplectic vector bundle if there exists a smooth section ω of E∗∧E∗ such that (Ex,ωx) is a symplectic vector space for all x∈M.

Example 1.7.

For a symplectic manifold M, the tangent bundle T⁢M is a symplectic vector bundle.

Remark 1.8.

Observe that if the tangent bundle is symplectic, then the manifold is not necessarily symplectic.

1.4 Symplectic Linear Algebra

Definition 1.9.

For (Vi,ωi) symplectic vector spaces, a linear symplectomorphism ϕ:(V1,ω1)→(V2,ω2) is an isomorphism of vector spaces such that ϕ∗⁢ω2=ω1. They (V1,ω1) and (V2,ω2) are symplectomorphic.

The usual definition of orthogonal complements carries over and will be denoted Wω for a subspace W⊂V.

Lemma 1.10.

Let (V,ω) be a symplectic vector space.

  1. 1.

    v↦ω⁡(v,−) is an isomorphism V→V∗.

  2. 2.

    dimW+dimWω=dimV.

  3. 3.

    (Wω)ω=W.

  4. 4.

    The following are equivalent:

    1. (a)

      W is a symplectic subspace of V;

    2. (b)

      Wω is symplectic;

    3. (c)

      W∩Wω={0};

    4. (d)

      W⊕Wω=V.

Proof.
  1. 1.

    This part is equivalent to nondegeneracy.

  2. 2.

    Use a rank-nullity argument to note that Wω maps to the annihlator of W under the isomorphism V→V∗.

  3. 3.

    Clearly W⊂(Wω)ω. Then use the dimension result.

  4. 4.

    Clearly a) implies c) and c) and d) are equivalent. Finally, it is easy to see that d) implies a). Finally equivalence of b) to the rest is easy.

□

Theorem 1.11 (Symplectic Basis).

For any symplectic vector space (V,ω), there exists a basis u1,…,un,v1,…,vn such that

called a symplectic basis for (V,ω).

2 January 23

2.1 More Basic Linear Algebra

Last time we stated the following:

Theorem.

For any symplectic vector space (V,ω), there exists a basis u1,…,un,v1,…,vn such that ω⁡(ui,vi)=1 and any other pairing is zero.

Proof.

We will induct on the dimension of the vector space. For n=1, take any nonzero u. Then by nondegeneracy, we can find a desired v. Then we simply use Lemma 1.4.2, note that u1,v1 span a symplectic subspace W of V, and apply the inductive hypothesis to Wω. Because u1,v1 are orthogonal to the symplectic basis for Wω, we are done. □

Corollary 2.1.

Any symplectic vector space is symplectomorphic to (ℝ2⁢n,ω0).

Proof.

Observe that

Define our morphism in the obvious way:

Then it is easy to check that

□

Corollary 2.2.

A skew-symmetric ω on V is symplectic iff ωn≠0.

Proof.

If ω is symplectic, then there exists a symplectic basis {ui,vi}, and ωn⁢(u1,v1,…,un,vn)≠0. In the other direction, if ω is degenerate, there exists u∈0 such that ω⁡(u,v)=0 for all v∈V. Then we can complete u to a basis by u2,…,u2⁢n, and here ωn⁢(u,u2,…,u2⁢n)=0, so ωn=0. □

2.2 Compatible Complex Structures and Inner Products

Tailoring the class to the audience, İnanc will return to the theme of the three geometries: Riemannian, symplectic, and complex. We will see that symplectic geometry lies between the other two geometries (every manifold admits a metric, complex algebraic structures are very rare).

The group of linear symplectomorphisms of (V,ω) is denoted by S⁢p⁢(V,ω).

Definition 2.3.

A real matrix A∈G⁢L2⁢n⁢(ℝ) is symplectic if AT⁢J0⁢A=J0, where J0 was defined in Example 1.3.4. The group of such matrices is S⁢p2⁢n⁢(ℝ).

By Corollary 2.1.1, we see that S⁢p⁢(V,ω)≃S⁢p2⁢n⁢(ℝ).

Definition 2.4.

A complex structure on a real vector space V is an automorphism J:V→V such that J2=i⁢I. Then (V,J) is a complex vector space with the definition (x+i⁢y)⁢v=x⁢v+y⁢J⁢v.

Then Aut⁡(V,J)≃G⁢Ln⁢(ℂ).

Definition 2.5.

If (V,ω) is a symplectic vector space, a complex structure J is called ω-compatible if ω⁡(J⁢u,J⁢v)=ω⁡(u,v) for all u,v∈V and ω⁡(v,J⁢v)>0 for all nonzero v∈V.

Let 𝒥⁡(V,ω) be the space of ω-compatible complex structures on (V,ω) with topology inherited from End⁡V (here endomorphisms are taken in 𝐃𝐢𝐟𝐟). This space will turn out to be contractible, but first we need to show that it is nonempty. This is true because J0 is compatible with the standard form. The taming property is also easy to check.

Recall that an inner product on a real vector space is a nondegenerate symmetric positive-definite bilinear form g. Here, Aut⁡(V,g)≃O⁡(2⁢n,ℝ).

Definition 2.6.

A Hermitian structure on (V,J) is an inner product g on V such that g⁡(J⁢u,J⁢v)=g⁡(u,v) for all u,v∈V.

Remark 2.7.

If g~ is any inner product on V, then g⁡(u,v)=g~⁢(u,v)+g~⁢(J⁢u,J⁢v) is Hermitian.

Remark 2.8.

J∈𝒥⁡(V,ω) if and only if gJ⁢(u,v)≔ω⁡(u,J⁢v) is a Hermitian inner product.

Proof.

Note that ω⁡(J⁢u,J⁢v)=ω⁡(u,v) iff ω⁡(J⁢u,−v)=ω⁡(u,J⁢v) iff gJ⁢(J⁢u,J⁢v)=gJ⁢(u,v). In addition, ω⁡(v,J⁢v)>0 iff gJ⁢(v,v)=0 clearly. □

Example 2.9.

The standard symplectic form, the matrix J0, and the standard inner product on (ℝ2⁢n,ω) is a compatible triple.

Theorem 2.10.

𝒥⁡(V,ω) is contractible.

Proof is left to the next lecture because it will take too long for the rest of this lecture.

Remark 2.11.

The analogous result for almost complex structures on symplectic manifolds will allow us to discuss Chern classes on symplectic manifolds.

Now let 𝒢⁡(V) be the space of inner products on V. We can define rt:𝒢⁡(Vt)→𝒥⁡(Vt,ωt) varying smoothly in t.1

Remark 2.12.

Given a complex vector space (V,J) and a compatible inner product g, we can derive a symplectic structure ω on V for which J is ω-compatible by ω⁡(u,v)=g⁡(J⁢u,v).

Proof of this fact is left to the reader. Exercises will be assigned nect time.

3 January 28

3.1 A Big Theorem

Last time we stated the following:

Theorem.

𝒥⁡(V,ω) is contractible.

Proof.

Let 𝒢⁡(V) be the space of inner products on V with topology given by choosing a unitary basis for V. We identify 𝒢⁡(V) with the set of symmetric positive-definite matrices. Under this identification, 𝒢⁡(V) becomes a smooth manifold. We will show that 𝒢⁡(V) retracts onto 𝒥⁡(V,ω). Because 𝒢⁡(V) is convex, it is contractible.

Consider the automorphism A:V→V given by A=μg−1∘μω Now observe that g⁡(A⁢u,v)=g⁡(u,−A⁢v). Therefore the adjoint of A is −A. Then we write P≔−A2=A⁢AT, which is symmetric and positive definite. We now have

Then we can write Q=P, which is symmetric and positive definite. Now set J=A⁢Q−1.1

We will show that J∈𝒥⁡(V,ω). First note that A and Q commute (because A and P commute). Therefore A preserves the eigenspaces of Q. This gives

Moreover, Q is self-adjoint and A is skew-adjoint, so J is skew-adjoint. Also J commutes with A.

Now we show ω-compatibility. The first condition is just

The taming condition is simply

because all eigenvalues of Q are positive. This gives us a map r:𝒢⁡(V)→𝒥⁡(V,ω) which can be checked to be smooth. We can define the right inverse by J↦ω⁡(−,J−). This can also be checked to be smooth. It is easy to see that if A=J, then the polar decomposition gives Q=−J2=idV, so r is a retraction. Therefore 𝒢⁡(V) is homotopy equivalent to 𝒥⁡(V,ω). □

Lemma 3.1.

Let A be symmetric. Then the following are equivalent:

  1. 1.

    A is positive definite.

  2. 2.

    All eigenvalues of A are positive.

  3. 3.

    A=B⁢D⁢B−1, where B is orthogonal and D is diagonal with positive entries.

  4. 4.

    A=B⁢BT for some nonsingular B.

3.2 More Compatibility

Definition 3.2.

We call ω,J,g on V a compatible triple if g⁡(u,v)=ω⁡(u,J⁢v).

Observe that any two determine the third. Moreover, we have seen that we can complete two of them to a compatible triple if:

  • •

    For ω,J, J is ω-compatible.

  • •

    For g,J, g is J-compatible.

  • •

    There are no conditions on ω,g to obtain a J.

Exercise 3.3.

Let (V,ω,J,g) be a compatible triple on V with dimV=2⁢n. Show that

  1. (a)

    ωn=n!⁢volg.

  2. (b)

    For any subspace W⊂V, J⁢Wω=W⟂.

Note that for standard (ℝ2⁢n,ω0,J0,g0), the automorphism groups overlap as G⁢Ln⁢(ℂ)∩O⁡(2⁢n)∩S⁢p⁢(2⁢n)=U⁡(n).

Theorem 3.4.

Any two of the above groups intersect as U⁡(n).

Proof.

Recall that

  1. 1.

    A∈S⁢p⁢(2⁢n) iff AT⁢J0⁢A=J0;

  2. 2.

    A∈G⁢Ln⁢(ℂ) iff A⁢J0=J0⁢A;

  3. 3.

    A∈O⁡(2⁢n) iff AT⁢A=I.

It is not hard to see that any two imply the third. For example, if we have the first two, then

□

How about the spaces of these structures? We denote the space of symplectic structures by Ω, the space of complex structures by J, and the space of inner products by 𝒢. Note that G=G⁢L2⁢n⁢(ℝ) acts transitively on Ω,J,𝒢 with stabilizers S⁢p⁢(2⁢n),G⁢Ln⁢(ℂ),O⁡(2⁢n). Because G is a Lie group acting transitively on a smooth manifold, every stabilizer is closed. Then a quotient of a Lie group by a closed subgroup is a smooth manifold.

Therefore we may consider the spaces G⁢L2⁢n⁢(ℝ)/S⁢p⁢(2⁢n),G⁢L2⁢n⁢(ℝ)/G⁢Ln⁢(ℂ),G⁢L2⁢n⁢(ℝ)/O⁡(2⁢n). These spaces are exactly Ω,J,𝒢, respectively. Thus they are smooth manifolds.

Exercise 3.5.

Compute their dimensions. Also compute their π0 for n=1,2.

Theorem 3.6.

𝒥⁡(ℝ2⁢n,ω0)=S⁢p⁢(2⁢n)/U⁡(n).

Proof.

If J is ω0-compatible, let gJ⁢(u,v)=ω⁡(u,J⁢v). This is a hermitian structure. For a unitary basis u1,…,un of ℂn, we have a symplectic basis u1,…,un,J⁢u1,…,J⁢un. Then define the map

which is a symplectomorphism. Set 𝒥⁡(ℝ2⁢n,ω0)→S⁢p⁢(2⁢n)/U⁡(n) by J↦[AJ]. This is an isomorphism. □

4 January 30

4.1 Homework Exercises

İnanç will post homework later, and it will be due in approximately two and a half weeks.

Exercise 4.1.

Find J∈J⁡(ℝ2⁢n)∖J⁡(ℝ2⁢n,ω0). In other words, find a complex structure not compatible with the standard symplectic form.

Exercise 4.2.

Find A∈S⁢L2⁢n⁢(ℝ)∖S⁢p2⁢n⁢(ℝ). In other words, find a matrix which is not symplectic.

Here are some facts that will be helpful on the homework.

Proposition 4.3.

Let A∈S⁢p2⁢n. Then

  1. 1.

    A∈S⁢L2⁢n;

  2. 2.

    AT∈S⁢p2⁢n;

  3. 3.

    λ is an eigenvalue with multiplicity m iff 1/λ is as well;

  4. 4.

    If ±1 is an eigenvalue of A, then it has even multiplicity;

  5. 5.

    If v1,v2 are eigenvectors for λ1,λ2 with λ1⁢λ2≠1, then ω0⁢(v1,v2)=0.

  6. 6.

    If A is symmetric and positive-definite, then Aα∈S⁢p2⁢n for any α∈ℝ∖{0}.

Proof.
  1. 1.

    Note that A preserves the symplectic form ω0. Therefore it preserves the volume form and the orientation.

  2. 2.

    We see (AT)T⁢J0⁢AT=A⁢J0⁢J0⁢A−1⁢J0−1=−A⁢A−1⁢(−J0)=J0.

  3. 3.

    Because AT⁢J0⁢A=J0, then AT=J0⁢A−1⁢J0−1, so A,A−1 have the same eigenvalues. Then if λ if an eigenvalue of A with multiplicity m, we have A⁢v=λ⁢v, so λ−1⁢v=A−1⁢v. Thus λ−1 is an eigenvalue of A−1 and therefore of A.

  4. 4.

    −1 must have even multiplicity to ensure that A has determinant 1.

  5. 5.

    Note that ω0⁢(v1,v2)=ω0⁢(A⁢v1,A⁢v2)=ω0⁢(λ1⁢v1,λ2⁢v2)=λ1⁢λ2⁢ω0⁢(v1,v2). Because λ1⁢λ2≠1, we must have ω0⁢(v1,v2)=0.

  6. 6.

    Let Vλ be an eigenspace of A. This is an eigenspace for λα under Aα. By the above, we if λ1⁢λ2≠1, then Vλ1,Vλ2 are orthogonal under ω0. In addition, by the previous, it is easy to see that Aα preserves ω0 on the eigenbasis. □

4.2 Subspaces of Symplectic Vector Spaces

For a given symplectic manifold, there are four kinds of submanifolds we will describe. Two of them are important, and the other two will allow us to reduce to the first two.

Let W be a subspace of the symplectic vector space (V,ω). Recall that W is symplectic if W∩Wω=0.

Definition 4.4.

W is

  • •

    isotropic if W⊂Wω;

  • •

    coisotropic if W⊃Wω;

  • •

    Lagrangian if W=Wω.

Example 4.5.

Consider ℝ2⁢n with the standard form with basis u1,v1,u2,v2. Then

  • •

    The spaces ⟨u1,v1⟩,⟨u2,v2⟩ are symplectic;

  • •

    The spaces ⟨u1⟩,⟨u2⟩,⟨v1⟩,⟨v2⟩ are isotropic;

  • •

    The spaces ⟨u1,u2,v2⟩,⟨u1,v1,v2⟩,⟨u1,u2,v1⟩,⟨u2,v1,v2⟩ are coisotropic;

  • •

    The spaces ⟨u1,u2⟩,⟨v1,v2⟩,⟨u1,v2⟩,⟨u2,v1⟩ are Lagrangian.

Proposition 4.6.

Let (V,ω) be a synplectic vector space of dimension 2⁢n.

  1. 1.

    Any line is isotropic;

  2. 2.

    Any hyperplane is coisotropic;

  3. 3.

    Any isotropic subspace is contained in some Lagrangian subspace;

  4. 4.

    Any coisotropic subspace contains some Lagrabgian subspace;

  5. 5.

    S⁢p⁢(V,ω) preserves the four types of subspaces.

Proof.
  1. 1.

    ω is alternating.

  2. 2.

    Wω is a line, so it is isotropic. Thus Wω⊂(Wω)ω=W.

  3. 3.

    If W is isotropic but not Lagrangian, then there exists 0≠v∈Wω∖W. Then set W1=⟨W,v⟩. This is clearly also isotropic. Repeat until W has dimension n.

  4. 4.

    Wω is isotropic, so use the above to find Wω⊂L. Then W⊃Lω=L.

  5. 5.

    We show that ϕ⁡(Wω)=ϕ⁢(W)ω. If v∈Wω, then ω⁡(−,v)|W=0, so ω⁡(ϕ⁡(−),ϕ⁡(v))|W=0. Therefore, ω⁡(−,ϕ⁡(v))|ϕ⁡(W)=0, so ϕ⁡(v)∈ϕ⁢(W)ω. Thus ϕ⁡(Wω)⊂ϕ⁢(W)ω Because the two spaces have the same dimension, they are equal. □

Proposition 4.7 (Symplectic Reduction).

Let W⊂(V,ω). If W is isotropic (resp. coisotropic), then Wω/W (resp. W/Wω) is symplectic.

Proof.

Let v1,v2∈Wω. Then ω⁡(v1+w1,v2+w2)=ω⁡(v1,v2) for any w1,w2. Therefore ω is defined on equivalence classes modulo W. Therefore we only need to check nondegeneracy. If v1∈Wω is such that ω⁡(v0,v)=0 for all v∈Wω, then v0∈W. Therefore ω is nondegenerate on Wω/W. □

Note that symplectic and Lagrangian subspaces are good for constructing new symplectic spaces and defining symplectic invariants (Floer homology, Gromov-Witten invariants). When we switch to discussing manifolds, we will see that Lagrangian submanifolds are hard to find, while symplectic submanifolds are easy to find. We will only want to consider some symplectic submanifolds.

Define ℒ⁡(V,ω) to be the space of Lagrangian subspaces of (V,ω) and ℒ⁡(n) to be the space of Lagrangian subspaces of ℝω0, the Lagrangian Grassmannian.

Proposition 4.8.

ℒ⁡(n)≅U⁡(n)/O⁡(n).

Sketch.

For Λ∈ℒ⁡(n), choose an orthonormal basis u1,…,un with respect to g0. Set

Therefore, Λ=A⁢Λh, where Λh is the span of the first n standard vectors. Also, note A is unitary. Conversely, for all A∈O⁡(n)⊂S⁢p⁢(2⁢n), clearly Λ=A⁢Λh is also Lagrangian. Set ℒ⁡(n)→U⁡(n)/O⁡(n) by Λ↦[A]. We can check that this is an isomorphism. Next time we will say some things about π1⁢(ℒ⁡(n)). □

5 February 4

5.1 Linear Algebra, Conclusion

Last time we discussed the Lagrangian Grassmannian ℒ⁡(n)=U⁡(n)/O⁡(n).

Remark 5.1.
  1. 1.

    π1⁢(ℒ⁡(n))=ℤ. In addition, the determinant map ℒ⁡(n)→S1 is a fibration with fiber S⁢U⁢(n)/S⁢O⁢(n). Then, the projection onto the first colume S⁢U⁢(n)→S2⁢n−1 is a fibration with fiber S⁢U⁢(n−1), so by the homotopy long exact sequence, we have an exact sequence

    This tells us that π1⁢(S⁢U⁢(n))=1. In particular, π1⁢(S⁢U⁢(n)/S⁢O⁢(n))=1. Now going back to ℒ⁡(n), then we use the homotopy LES to obtain

    so π1⁢(ℒ⁡(n))=ℤ.

  2. 2.

    By the universal coefficient theorem, H1⁢(ℒ⁡(n),ℤ)≃Fr⁡H1⁢(ℒ⁡(n),ℤ)=ℤ. The generator is called the Maslow index Mn. For a loop λ in ℒ⁡(n), we then define the Maslow index of λ as Mn⁢([λ]).

5.2 Symplectic Vector Bundles

Recall that a symplectic vector bundle over a manifold M is a real vector bundle E→M with a C∞ section ω of E∗∧E∗ such that Ex,ωx is a symplectic vector space for all x∈M. Then two symplectic vector bundles (Ei,ωi) are isomorphic if there exists an isomorphism ϕ:E1→E2 such that ϕ∗⁢ω2=ω1.

Example 5.2.

If (M,ω) is symplectic, then (T⁢M,ω) is a symplectic vector bundle. In addition, the pushforward of a symplectomorphism is an isomorphism of symplectic vector bundles.

Remark 5.3.

Given any vector bundle E→M, we can define a symplectic vector bundle on the Whitney sum E⊕E∗→M by

This is clearly bilinear, antisymmetric, and non-degenerate.

Definition 5.4.

A complex vector bundle over a manifold M is a real vector bundle E→M with a C∞ section J of End⁡(E) such that (Ex,Jx) is a complex vector space for all x∈M.

We say two complex vector bundles (Ei,Ji) are isomorphic if there exists a vector bundle isomorphism ϕ:E1→E2 such that ϕ⁡(J1⁢v)=J2⁢ϕ⁢(V).

Example 5.5.

Let M be a complex manifold of dimension n. Then define the usual multiplication by i on Tp⁢M. Now we need to see if this is globally defined. To do this, we simply compute on two charts and then use the Cauchy-Riemann equations. This is left to the reader.

Remark 5.6.

Given any vector bundle E→M, we can define a complex vector bundle on E⊗ℂ.

Definition 5.7.

If (E,o⁢m⁢e⁢g⁢a) is a symplectic vector bundle over M, a complex structure J on E is called ω-compatible if Jx∈𝒥⁡(Ex,ωx) for all x∈M.

Denote by 𝒥⁡(E,ω) the space of ω-compatible complex structures on (E,ω) with the topology inherited from End⁡E.

Definition 5.8.

A Hermitian structure on (E,J) is a J-compatible inner product g (at every x∈M).

Remark 5.9.
  1. 1.

    We can construct a Hermitian inner product from any inner product as before.

  2. 2.

    The space of Hermitian structures is convex.

  3. 3.

    J∈𝒥⁡(E,ω) if and only if gJ⁢(u,v)≔ω⁡(u,J⁢v) is Hermitian.

Theorem 5.10.

𝒥⁡(E,ω) is nonempty and contractible.

Proof.

Define the retract by using the map in Theorem 2.10 pointwise. In a local trivialization of E, we can see that r:G⁡(E)→𝒥⁡(E,ω) is smooth. □

Now we will consider chart transitions for (E,ω),(E,J),(E,g). Denote the chart for the local trivializations by {Ui} with trivializations ϕi.

Type Group
E G⁢L⁢(2⁢n)
(E,ω) S⁢p⁢(2⁢n)
(E,J) G⁢L⁢(n,ℂ)
(E,g) O⁡(2⁢n)

In each special case, the structure of the bundle can be reduced. Note here that we also have the two out of three property for reduction to U⁡(n) as before. We will illustrate this by checking the (E,ω) case. We simply note that E|Ui≃Ui×ℝ2⁢n, which admits a symplectic basis. Therefore we have a symplectic trivialization of (E,ω). Therefore, locally we have an isomorphicm to Ui×ℝ2⁢n,ω0. Therefore we have reduction of the structure to S⁢p⁢(2⁢n).

Now suppose the structure of (E,ω) can be reduced to U⁡(n). Then the corresponding J is ω-compatible.

Theorem 5.11.
  1. 1.

    Let (E,ω) be a symplectic vector bundle and J1,J2∈𝒥⁡(E,ω). Then (E,J1)≃(E,J2).

  2. 2.

    Let (Ei,ωi) be symplectic vector bundles and Ji∈𝒥⁡(Ei,ωi). Then (E1,ω1)≃(E2,ω2) if and only if (E1,J1)≃(E2,J2).

6 February 6

6.1 Proof of Theorem 5.11

İnanç will post the homework tonight. Last time we stated Theorem 5.11, which is reproduced below.

Theorem.
  1. 1.

    Let (E,ω) be a symplectic vector bundle and J1,J2∈𝒥⁡(E,ω). Then (E,J1)≃(E,J2).

  2. 2.

    Let (Ei,ωi) be symplectic vector bundles and Ji∈𝒥⁡(Ei,ωi). Then (E1,ω1)≃(E2,ω2) if and only if (E1,J1)≃(E2,J2).

Proof.
  1. 1.

    First, note that for any structure group G of a real vector bundle, there exists a classifying space B⁢G with a contractible universal G-bundle E⁢G. In particular, G-bundles E/M are classified by homotopy classes of maps M→B⁢G. In other words, B⁢G represents the functor M↦{principle ⁢G⁢-bundles on ⁢M} on the homotopy category.

    Now use the homotopy LES of the fibration U⁡(n)→S⁢p⁢(2⁢n)→S⁢p⁢(2⁢n)/U⁡(n)≃pt to obtain isomorphisms πk⁢(U⁡(n))≃πk⁢(S⁢p⁢(2⁢n)). In particular, the inclusion U⁡(n)→S⁢p⁢(2⁢n) induces an isomorphism on all πk. By Whitehead, this is a homotopy equivalence. Therefore B⁢U⁢(n)≃B⁢S⁢p⁢(2⁢n). Therefore for any (E,ω), this determines f:M→B⁢S⁢p⁢(2⁢n), which can be homotoped to f′:M→B⁢U⁢(n), which reduces (E,ω) uniquely to a U⁡(n)-bundle.

  2. 2.

    By the previous, isomorphism as S⁢p⁢(2⁢n)-bundles implies isomorphism as U⁡(n)-bundles. Now consider the inclusion U⁡(n)→G⁢Ln⁢(ℂ). This is a homotopy equivalence (use Gram-Schmidt to deform any loop G⁢Ln⁢(ℂ) into U⁡(n)). By similar arguments as above, two U⁡(n)-bundles are isomorphic iff they are isomorphic as G⁢Ln⁢(ℂ)-bundles. This concludes the proof. □

Remark 6.1.

Note that this allows us to determine whether two symplectic vector bundles are isomorphic by comparing their Chern classes.

6.2 Vector Bundles, Continued

Let F be a subbundle of the symplectic vector bundle (E,ω).

Definition 6.2.

The symplectic complement of F is

Definition 6.3.

We define the vector subbundle F to be

  • •

    symplectic if F∩Fω is the zero section of E;

  • •

    isotropic if F⊂Fω;

  • •

    coisotropic if F⊃Fω;

  • •

    Lagrangian if F=Fω.

Many previous results on subspaces carry over to subbundles.

Proposition 6.4.

Let F be a subbundle of the symplectic vector bundle (E,ω).

  1. 1.

    If F is symplectic, J1∈𝒥⁡(F,ω|F), then there exists J∈𝒥⁡(E,ω) extending J1.

  2. 2.

    If F is Lagrangian, J∈𝒥⁡(E,ω), then (E,J)≅F⊗ℂ.

Proof.
  1. 1.

    If F is symplectic, then so if Fω. Therefore there exists J2∈𝒥⁡(Fω,ω|Fω). Then we have E=F⊕Fω, so we write J=J1⊕J2. All properties of an ω-compatible complex structure follow from the orthogonal decomposition of E.

  2. 2.

    Let g be a compatible inner product for ω,J. Then E=F⊕F⟂g=F⊕J⁢Fω=F⊕J⁢F≅F⊗ℂ. To see this, note that J⁢(u+J⁢v)=−v+J⁢u, which is the same complex structure as F⊗ℂ. □

6.3 Compatible Triples on Manifolds

Let E=T⁢M with symplectic structure ω, complex structure J, and inner product g. When only defined on T⁢M, then ω is an almost symplectic structure and J is an almost complex structure. In this course, we will focus on the case when ω is a true symplectic structure on M. We define the following spaces:

  • •

    Ω⁡(M) the space of symplectic structures on M;

  • •

    𝒥⁡(M) the space of almost complex structures on M;

  • •

    𝒢⁡(M) the space of almost complex structures on M;

  • •

    𝒥⁡(M,ω) the space of compatible almost complex structures.

Definition 6.5.

Now let S be a submanifold of (M,ω). Then T⁢S is a subbundle of (T⁢M|S,ω), and we call S

  • •

    symplectic if T⁢S⊂T⁢M|S is symplectic;

  • •

    isotropic if T⁢S⊂T⁢M|S is isotropic;

  • •

    coisotropic if T⁢S⊂T⁢M|S is coisotropic;

  • •

    Lagrangian if T⁢S⊂T⁢M|S is Lagrangian;

Note that S is symplectic iff (S,ω|S) is a symplectic manifold. In addition, note that S must be even dimensional and the nomal bundle ν⁢S=(T⁢S)ω. Also, note that S is Lagrangian iff ω|S=0.

Definition 6.6.

For S a submanifold of (M,J), we call S, we call S J-holomorphic if J⁡(T⁢S)⊂T⁢S, which happens iff J|S is an almost complex structure on S.

Remark 6.7.

Our results about symplectic vector bundles carry over to (T⁢M,ω). In particular, 𝒥⁡(M,ω) is nonempty and contractible.

Proposition 6.8.

Let S be a submanifold of (M,ω).

  1. 1.

    If S is symplectic, then S is J-holomorphic for some J∈𝒥⁡(M,ω);

  2. 2.

    If S is J-holomorphic for any J∈𝒥⁡(M,ω), then S is symplectic.

Proof.
  1. 1.

    This is (mostly) just the first part of Proposition 6.4. However, we must extend J from T⁢M|S to all of M. First we take a compatible metric, extend it by a partition of unity, and then use the retract to find a compatible J′.

  2. 2.

    Note J⁡(T⁢S∩(T⁢S)ω)⊂J⁢T⁢S∩J⁢(T⁢S)ω=T⁢S∩(T⁢S)⟂g={0}. Therefore T⁢S∩(T⁢S)ω={0}, so S is symplectic.

□

Exercise 6.9.

Find a symplectic submanifold of ℝ2⁢n that is not J0-holomorphic.

7 February 11

7.1 Obtaining Compatible Triples

Input Condition Output Integrable?
ω,J J is ω-compatible g⁡(u,v)=ω⁡(u,J⁢v) metric g flat?
J,g g is Hermitian ω⁡(u,v)=g⁡(J⁢u,v) almost symplectic ω closed?
ω,g none r⁡(J)∼μg−1⁢μω almost complex J integrable?
  1. 1.

    This is the same as local flatness. By a theorem of Bieberbach, all compact flat manifolds are finitely covered by tori.

  2. 2.

    This is the same as 𝑑ω=0. If J is a complex structure and g is Hermitian, then ω is a fundamental 2-form. The Goldberg conjecture states that if g is Einstein, then 𝑑ω=0.

  3. 3.

    By a theorem of Newlander and Nirenberg, there exists a complex structure on M inducing J if and only if the Nijenhuis tensor NJ=0.

Definition 7.1.

The Nijenhuis tensor of J is defined by

Exercise 7.2.
  1. 1.

    If the map −↦[−,v] is J-holomorphic for all v, then NJ=0.

  2. 2.

    Check that NJ is a tensor.

  3. 3.

    If M has dimension 2, then NJ=0.

7.2 Complex Structures

Definition 7.3.

(M,ω,J,g) is a Kähler manifold when J is integrable.

Example 7.4.

Any M=Σh a closed oriented surface of genus h is a Kähler manifold. To see this, take ω=volg. Then it is easy to see that ω is bilinear, skew-symmetric, nondegenerate, and closed. Now let J=r⁡(g), so (ω,J,g) is a compatible triple on Σ. But any almost complex structure on a surface is integrable by Exercise 7.2, so we are done.

Now let J be an almost complex structure on M. Take the complexified vector bundle T⁢M⊗ℂ. Then J extends linearly to T⁢M⊗ℂ as

Then we can see that J2=−I, so the eigenvalues are ±i. Then we can define T1,0,T0,1 as the eigenspaces of ±i. Therefore we obtain an isomorphism

In addition, we can write T∗⁢M⊗ℂ≅T0,1⊕T1,0. Then we have

Therefore we have

We want to define a differential that describes this structure. We can define ∂:Ωℓ,m⁢(M)→Ωℓ+1,m by

and then define ∂¯ analogously.

Now if d=∂+∂¯, then we must have ∂2=0,∂¯2=0,∂∂¯+∂¯⁡∂=0. This allows us to define Dolbeaut cohomology.

Suppose J is a complex structure on M. Then in a local chart U, we have coordinates zj and we can write T1,0=⟨∂∂zj⟩,T0,1=⟨∂∂z¯j⟩. In addition, T1,0=⟨𝑑zj⟩,T0,1=⟨d⁢z¯j⟩. Writing a form β locally, we will see that d=∂+∂¯.

Remark 7.5.

For any almost complex structure J on M, if ∂¯2=0, then J is integrable.

Theorem 7.6 (Newlander-Nirenberg).

Let J be an almost complex structure on M. Then the following are equivalent:

  1. 1.

    J is induced by a complex structure on M;

  2. 2.

    NJ=0;

  3. 3.

    d=∂+∂¯;

  4. 4.

    ∂¯2=0.

Now suppose (M,ω,J,g) be Kähler. Then what can we say about ω?

  • •

    ω is a closed form over ℂ;

  • •

    If ω=∑aj⁢k⁢𝑑zj∧𝑑zk+∑bj⁢k⁢𝑑zj∧d⁢z¯k+∑cj⁢k⁢d⁢z¯j∧d⁢z¯k, then aj⁢k=cj⁢k and bj⁢k=−b¯k⁢j;

  • •

    Computing J∗⁢ω, we will see that ω is a (1,1)-form.

  • •

    ∂¯⁡ω=0.

8 February 13

Recall that there will be no class next Tuesday. Last time we began the discussion of Kähler forms.

8.1 Kähler Forms Continued

Recall from last time that Kähler forms are of type (1,1). Also recall that locally we have

In order to express the coefficients as a metric, we can rewrite

Then the matrix (hj⁢k) is a Hermitian matrix. In particular, the volume form is locally

Proposition 8.1.

Now let M be a compact complex manifold of dimension n with complex structure J. Then the following are equivalent:

  1. 1.

    ω is a Kähler form on M;

  2. 2.

    ω is a ∂ and ∂¯-closed (1,1)-form locally given by (8.1) with coefficient matrix given by a Hermitian positive-definite matrix.

Proof.

We did one of the directions above. In the other direction, observe that g⁡(u,v)≔ω⁡(u,J⁢v) is locally given by g⁡(u,v)=uT⁢H⁢v for H=(hj⁢k) a Hermitian positive definite matrix. Thus ω is nondegenerate. In addition, ω is clearly J-compatible because it has type (1,1). □

Example 8.2.

(ℂn,ω0) is Kähler with complex structure J0.

Example 8.3.

Consider ℂ⁢ℙn=ℂn+1/{0}/ℂ∗. Recall the standard complex atlas for ℂ⁢ℙn. Let J0 be the induced complex structure on ℂ⁢ℙn. Then there exists a symplectic form ωF⁢S, the Fubini-Study metric, on ℂ⁢ℙn compatible with J0.

Proposition 8.4.

Any complex submanifold of a Kähler manifold is Kähler.

First Proof.

Let M be a complex manifold of dimension n and S be a complex manifold of dimension m. Then locally at p∈S⊂M, we have charts for M such that S={zm+1=⋯=zn=0}. If ω is a Kähler form on M, then locally we have

In particular, it is easy to see that the coefficient matrix is Hermitian and positive definite. Finally, ω is a ∂,∂¯-closed (1,1)-form, so ι∗⁢ω is as well. □

Second Proof.

Note that S is a J-holomorphic submanifold of M. By Proposition 6.8, S is also a symplectic submanifold of M. Therefore ω|S is symplectic and compatible with J|S, which is already integrable. □

8.2 Some Algebraic Geometry

Thus we have proven the following corollary, which tells us that any smooth affine or projective complex algebraic variety is Kähler.

Theorem 8.5.

Any complex submanifold of ℂn or ℂ⁢ℙn is Kähler.

Remark 8.6.

We will state some facts from complex algebraic geometry in context for culture.

  1. 1.

    Many complex submanifolds of ℂn are not affine, e.g. ℂ2∖{0}.

  2. 2.

    By Chow’s theorem (1949), any closed complex submanifold of ℂ⁢ℙn is algebraic.

  3. 3.

    The Kodaira embedding theorem (1960s) states that a closed Kähler manifold (M,ω) is algebraic if and only if ω is integral.

Proposition 8.7.

In particular, any compact complex curve Σ is projective.

Proof.

On Σ equipped with complex structure J, take any J-compatible metric g. Then set ω⁡(u,v)≔g⁡(J⁢u,v). Then ω is symplectic. Therefore ω is a Kähler form on Σ compatible with J. Now suppose that

Because any nonzero real multiple of a Kähler form is still Kähler, then 1α⁢ω is the desired integral form. □

Remark 8.8.

Not all compact Kähler surfaces are projective. In complex dimension 2, Kähler manifolds become projective after deforming the complex structure. In dimension at least 4, there exist compact Kähler manifolds (due to Voisin) that are not even homotopy equivalent to a projective variety.

Theorem 8.9 (Gromov’s Embedding Theorem).

Let (M,ω) be a closed symplectic manifold. If ω is integral, then there is a symplectic embedding of M into (ℂ⁢ℙN,ωF⁢S) for some high enough N.

Remark 8.10.

İnanç says this theorem is conceptually nice, but he does not know any use for the result.

Example 8.11.

Let Cd=Z⁡(∑zid)⊂ℂ⁢ℙ2. Then C1≅C2≅ℂ⁢ℙ1,C3≅T2,C4≅Σ6.

Now let Sd be a hypersurface of degree d in ℂ⁢ℙ3. Then S1=ℂ⁢ℙ2,S1=ℙ1×ℙ1,S3=Bl6⁡ℙ2, and S4 is a K3 surface (use the adjunction formula).

8.3 Stein Manifolds

Let M be a complex manifold of dimension n and let ρ∈C∞⁢(M,ℝ) be a proper, strictly plurisubharmonic function on M.

Definition 8.12.

A function ρ is structly plurisubhamonic if on each complex chart, H∂∂¯⁢ρ is positive-definite.

In this case,

is a Kähler form on M, and we say (M,ω) is a Stein manifold, where ρ is the Kähler potential.

9 February 20

9.1 Stein Manifolds Continued

Today’s lecture will be slower than usual. Last time we defined Stein manifolds (M,ρ), where ρ is strictly plurisubharmonic. It is easy to see that ω=i2⁢∂∂¯⁡ρ is closed and real. Also, J∗⁢ω=ω.

Example 9.1.

Choose M=ℂn and ρ=∑|zj|2. Then we can check that H∂∂¯ is the identity matrix. Then we can also see that

is the standard form.

Thus every complex submanifold of ℂn is a Stein manifold.

Theorem 9.2 (Remmert Embedding Theorem (1956)).

Let (M,ω) be a Stein manifold. Then there exists a proper holomorphic embedding of M into ℂN for some N.

Corollary 9.3.

Stein manifolds do not have compact complex submanifolds of positive dimension. In particular, they are never compact.

Theorem 9.4 (Behnke-Stein (1958)).

Every connected non-compact complex curve is Stein.

9.2 Topological Properties of Kähler Manifolds

Theorem 9.5 (Hodge).

On a compact Kähler manifold (M,ω), the Dolbeaut cohomology groups satisfy the Hodge decomposition

In addition, we have an isomorphism

Theorem 9.6 (Serre Duality).

For any complex manifold M, we have Hp,q⁢(M)≃Hn−p,n−q⁢(M).

Remark 9.7.

We obtain the Hodge numbers hp,q=dimHp,q, which are traditionally arranged in the Hodge diamond, which has symmetry given by the Hodge symmetry and Serre duality.

Example 9.8.

Consider a curve of genus g. Then the Hodge diamond is

Example 9.9.

The Hodge diamond of ℙ2 is

Example 9.10.

The Hodge diamond of ℙ1×ℙ1 is

Example 9.11.

The Hodge diamond of ℙ1×Σg is

Example 9.12.

Consider the Hopf surface, which is the quotient of ℂ2∖{0} by the action of ℤ acting by powers of 2. This is a biholomorphic, free, and properly discontinuous action. Up to diffeomorphism, our surface is M=ℂ2∖{0}/(z1,z2)∼(2⁢z1,2⁢z2)≃S3×S1. Note that the second cohomology vanishes, so the surface is not Kähler. The Hodge diamond of the Hopf surface is

Remark 9.13.

The odd Betti numbers of compact Kähler manifolds are even. The even Betti numbers are positive because ωk is closed but not exact. In fact, any symplectic manifold has nonvanishing even cohomology. In particular, hp,p is always positive.

9.3 Complex and Symplectic Structures on 4-Manifolds

Let X be a closed, connected, oriented, smooth 4-manifold. The equivalences will be orientation-preserving diffeomorphisms. We will see that for complex surfaces, Hodge numbers depend only on the topological type of the oriented manifold.

Recall the Euler characteristic e=∑(−1)i+j⁢hi,j and the signature σ=∑(−1)j⁢hi,j. The signature of the manifold is the signature of the quadratic form QX, called the intersection form.

Remark 9.14.

If X1≃X2 are homotopy equivalent, then they have the same intersection form. Also, if X¯ is X with the opposite orientation, then QX¯=−QX. Third, if X=X1⁢#⁢X2, then QX=QX1⊕QX2.

Theorem 9.15 (Whitehead).

If X1,X2 are simply-connected and QX1≅QX2, then X1≃X2.

Theorem 9.16 (Freedman).

If X1,X2 are simply connected, are either both smoothable or both not smoothable, and QX1≅QX2, then X1≅X2.1

Example 9.17.

The intersection form of ℙ2 is Qℙ2=(1) and the intersection form of ℙ1×ℙ1 is

10 February 25

10.1 Complex Structures in Dimension 4

Recall that for Kähler surfaces, the Hodge numbers depend only on the top type of the oriented manifold.

Theorem 10.1 (Kodaira-Siu (1981)).

Let X be a compact complex surface. Then X is Kähler if and only if b1⁢X is even.

Remark 10.2.

The existence of an almost complex structure is a purely homotopy-theoretic problem.

Theorem 10.3 (Wu’s criterion).

Let X be a closed oriented 4-manifold. Then X admits an almost complex structure J if and only if c∈H2⁢(X,ℤ) such that c2=2⁢e+3⁢σ and c⋅α≡α⋅αmod2.1

Example 10.4.

On S4, the second cohomology vanishes, so c=0. This implies c2=0, but 2⁢e+3⁢σ=4. Therefore S4 admits no almost complex structure.

Example 10.5.

On ℙ2, note that 2⁢e+3⁢σ=9, so c=±3⁢H. Clearly c satisfies the second part of Wu’s criterion, so ℙ2 admits a complex structure. However, the two almost complex structures are not equivalent.

Example 10.6.

Consider ℙ¯2. Here, if c=m⁢H, then c2=−m2. However, 2⁢e+3⁢σ=3, so there is no almost complex structure.

Example 10.7.

Now consider the manifold ℙ2⁢#⁢ℙ2. Then we have H2=ℤ2, so if we write c=(m,n), then c2=m2+n2. However, 2⁢e+3⁢σ=14, which is not a sum of two squares. Therefore there is no almost complex structure.

Remark 10.8.

We see that orientation reversal and connected sum are not almost complex operations. Therefore, they are also not symplectic operations.

Example 10.9.

Consider the connected sum ℙ2⁢#⁢ℙ2⁢#⁢ℙ2. Here we see that 2⁢e+3⁢σ=19=32+32+12. In addition, 3,3,1 are all odd, so this manifold does admit an almost complex structure.

10.2 Symplectic Structures in Dimension 4

10.2.1 Seiberg-Witten Invariants

These come from a certain system of differential equations on X which come from physics. The solutions yield a very nice moduli space ℳX,c under facorable conditions for each c∈H2⁢(X): If b+⁢(X)>0 and X admits an almost complex structure, then ℳX,c=∅ for all but finitely many c. Then ℳX,c is a compact oriented manifold with dimension

For c with dimℳX,c=0, then we can assign the signed count of points in ℳX,c. If b+⁢(X)=1, we need an additional choice. Then this defines the Seiberg-Witten invariant S⁢WX:H2⁢(X,ℤ)→ℤ. Then for any α with S⁢WX⁢(α)≠0 is called a Seiberg-Witten basic class.

Remark 10.10.

Here are some fundamental results:

  1. 1.

    The Seiberg-Witten invariant is invariant under orientation preserving diffeomorphisms.

  2. 2.

    (Vanishing Theorem). If X=X1⁢#⁢X2, with b+⁢Xi>0 for i=1,2, then S⁢WX≡0.

  3. 3.

    (Taubes). If X is symplectic with b+⁢X>0, then there exists a particular c∈H2⁢(X,ℤ) such that S⁢WX⁢(c)=1. In fact, c=c1⁢(X,J) for any ω-compatible J.

Theorem 10.11 (Diversity).

There exist infinitely many closed, connected, oriented maniflds of dimension 2⁢n≥4. Smooth manifolds in each subclass are depicted below.

Refer to caption
Figure 10.1: Classes of manifolds.
Proof.

First, Freedman’s E8-manifold is not smoothable. Then S4 has no almost complex structure. Third, ℙ2⁢#⁢ℙ2⁢#⁢ℙ2 is not symplectic by Taubes and not complex by Kodaira-Siu. Fourth, S1×S3 admits a complex structure, but is not symplectic or Kähler. Fifth, the Kodaira-Thurston manifold is complex and symplectic, but not Kähler. □

Remark 10.12.

Infinitely many examples in each subclass can be obtained by blowups X⁢#⁢ℙ¯2.

Example 10.13 (Kodaira-Thurston Manifold).

Let G=⟨γ1,γ2,γ3,γ4⟩ act in ℝ4 with coordinates x1,y1,x2,y2, where γi increases the ith component by 1 for i=1,3,4, while γ2=(x1,y1+1,x2+y2,y2). This is a free and properly dicontinuous action, so X=ℝ4/G is a smooth manifold. Then q:ℝ4→X is the universal cover of X. Then we see that all γi commute exept for [γ2,γ3]. Therefore H1⁢X=G/[G,G], so b1⁢X=3. Therefore X cannot be Kähler.

11 February 27

Exercise 11.1.

Let f:X→Y be a finite unbranched cover. Show that if Y is symplectic, almost complex, complex, Kähler, projective, or Stein, then so is X.

11.1 Kodaira-Thurston Manifold

Recall the definition of the Kodaira-Thurston manifold X as a quotient of ℝ4 by a discrete group G from last time.

Remark 11.2.

We may obtain X as a T2-bundle over the torus.

To see this, note that we can project X onto the first two coordinates. This is a map f:X→ℝ2/ℤ2=T2, where ℤ2 acts by shifts. To compute the fiber, note that f−1(0,0)=ℝ2/∼, where ∼ identifies (x2,y2)∼(x2+1,y2) and (x2,y2)∼(x2,y2+1). Thus the fiber is a torus. The action of γ2 becomes nontrivial monodromy.

To see the monodromy, note that if we move in the x1-direction in the base, γ1 identifies (0,0,x2,y2) with (1,0,x2,y2). However, in the y1-direction, γ2 identifies (0,0,x2,y2) with (0,1,x2+y2,y2).

This gives us a description of the monodromy by the matrices A=I2,B=(1011)∈S⁢L2⁢(ℤ).1

Theorem 11.3 (Thurston (1976)).

The manifold X is symplectic.

Proof.

Choose the standard symplectic form ω0=𝑑x1∧𝑑y1+𝑑x2∧𝑑y2. Next we can see that γi∗⁢ω0=ω0 for all i=1,…,4. The only one we need to check is γ2, and we see that

because 𝑑y2∧𝑑y2=0. Therefore each γi is a symplectomorphism, so G is a symplectomorphism of (ℝ4,ω0). Thus ℝ4→X is a covering with symplectic deck transformations. Thus there is an induced symplectic form ω on X. □

Theorem 11.4 (Kodaira (1969)).

The Kodaira-Thurston manifold is complex.

Proof.

Let α=𝑑x1,β=𝑑y1,γ=𝑑x2−y1⁢𝑑y2,δ=𝑑y2. Then this gives a G-invariant basis for Ω1⁢(ℝ4). The only one we need to check is γ with γ2, and we see that

Then note that α,β,δ are all closed, but 𝑑γ=−𝑑y1∧𝑑y2=−β∧δ=δ∧β. Then the dual basis

is a G-invariant basis for the Lie algebra of vector fields on ℝ4. To see this, we only need to check γ2,d, and we see

as desired. Now define J⁢a=c,J⁢c=−a,J⁢b=d,J⁢d=−b. Then we see that

Similarly, note that

Here we use the fact that all commutators vanish besides [b,d]=−c. Therefore J is a complex structure on ℝ4. In addition, the G-invariant basis for vector fields on ℝ4 descends to a basis for vector fields on X. Also, J descends to an integrable almost complex structure on X. □

Corollary 11.5.

The Kodaira-Thurston manifold is complex and symplectic but not Kähler.

Remark 11.6.

The G-invariant basis α,β,γ,δ descends to Ω1⁢(X). In addition, the symplectic form ω0 descends to the form

11.2 Algebraic Topology of X

For simplicity, we will drop the bars. Recall that 𝑑α=𝑑β=𝑑δ=0, while 𝑑γ=δ∧β. Also, note that α∧α=β∧β=δ∧δ=0. Then recall that dimH1⁢(X,ℝ)=3 and note that e⁡(X)=0. Therefore, we obtain b2=4.

We will write a basis for each cohomology and write the intersection form. We have

and

Next, we see that H3⁢(X,ℝ)=⟨[β∧γ∧δ],[α∧γ∧δ],[α∧β∧γ]⟩ and H4⁢(X,ℝ)=⟨[α∧β∧γ∧δ]⟩.

Finally, we can see the intersection form is given by

From this, we see that b+⁢X=B−⁢X=2, so σ⁡(X)=0.

12 March 3

Note: I was not here on this day. Notes were provided by Arthur Wang.

12.1 Topology of Complex Manifolds

We will answer the following question:

Question 12.1.

Are there any differential topological constraints for a symplectic 4-manifold to admit a complex structure?

Recall that all symplectic manifolds have an almost complex structure. Also, they have positive b+.

Theorem 12.2 (Kodaira-Siu).

If X is a compact complex surface, X is Kähler if and only if b1⁢X is even.

We can consider X with even first Betti number. Then if X is complex, then it is Kähler. However, there are no constraints that come from Hodge theory. There are, however, more constraints from differential topology.

Theorem 12.3 (Hard Lefschetz).

For ω the Kähler form, Lωk:Hn−k⁢(M,ℂ)→Hn+k⁢(M,ℂ) is an isomorphism for all k. Here Lωk sends α to α∪ωk.

This gives us Serre duality on Dolbeault cohomology.

Remark 12.4.

All triple Massey products on Kähler M are zero: Let a1,a2,a3∈H∗⁢(M,ℝ) with a1∪a2=0=a2∪a3. Then

is defined as follows:

Take αi∈Ω∗⁢M with ai=[αi], so α1∧α2=d⁢η12 and α2∧α3=d⁢η23. Then

There is a constraint from algebraic topology. Not all finitely presented groups are fundamental groups of Kähler manifolds. For example, if the abelianization of π1 has odd rank, then M is not Kähler. Also, π1 that are nontrivial free products cannot be realized.

Remark 12.5.

In dimension 4, we can also use the Enriques-Kodaira classification of complex surfaces.

12.2 Chern Classes

We will define the canonical class of a complex or symplectic manifold. First, however, we need to define Chern classes. The k-th Chern class of a complex vector bundle E is a cohomology class ck⁢(E)∈H2⁢k⁢(M,ℤ) and the total Chern class of E is the sum c⁡(E)=c0⁢(E)+c1⁢(E)+⋯.

These uniquely satisfy the following axioms:

  1. 1.

    c0⁢(E)=1;

  2. 2.

    (Naturality) For all f:N→M, ck⁢(f∗⁢E)=f∗⁢ck⁢(E).

  3. 3.

    (Additivity) c⁡(E⊕F)=c⁡(E)∪c⁡(F).

  4. 4.

    (Normalization) For the tautological line bundle over ℙk, we have c=1−h, where h is the hyperplane class.

Remark 12.6.

The existence and uniqueness of the Chern classes follows from the theory of classifying spaces. Also, ck⁢(E)=0 for all k>dimE. The top Chern class is the Euler class, and for E=T⁢M, e⁡(E)⁢[M]=e⁡(M), the Euler characteristic. Finally, Chern classes are invariant under isomorphism.

Note that if M is symplectic, then the space of compatible J is contractible and nonempty, so we can define Chern classes uniquely.

Definition 12.7.

The canonical class of a (symplectic, complex) manifold M is K≔−c1⁢(M).1

Now specialize again to the 4-dimensional case. Then we have two cohomological invariants: The class of the symplectic form and the canonical class.

Definition 12.8.

We define X to be minimal if X is not a connected sum of another closed manifold and ℙ¯2.

For minimal X, the we can define the symplectic Kodaira dimension of (X,ω) as:

For non-minimal manifolds, then we define the Kodaira dimension to be the Kodaira dimension of a minimal model.

Remark 12.9.

There is a minimal madel program for symplectic 4-manifolds. One can always find a minimal symplectic X such that X′=X⁢#⁢m⁢ℙ¯2.

Remark 12.10.

By Taube’s work on Seiberg-Witten invariants of symplectic manifolds, no other combinations for K⋅[ω] and K2 can occur.

This implies that κ⁡(X) is well-defined.

Remark 12.11.

If κ⁡(X)=−∞, then X is either rational or ruled.

Remark 12.12.

When X is Kähler, then its algebriac and symplectic Kodaira dimensions agree.

The general problem we want to solve is:

Question 12.13.

For a given class of manifolds, what are the constraints on their algebraic topology? Which values can be realized as invariants of such manifolds?

13 March 5

13.1 Geography Problem

Today we will discuss problems of realizing certain algebraic invariants with a manifold admitting a certain structure. Let X be a closed connected oriented smooth 4-manifold. Which pairs of integers (x,y)∈ℤ2 can be realized as

  1. 1.

    (e⁡(X),σ⁡(X))?

  2. 2.

    (c12⁢(X),c2⁢(X))?

  3. 3.

    (c12⁢(X),χh⁢(X))?

Note that c12=2⁢e+3⁢σ and χh=e+σ4 (Noether’s formula). We will ask that X is a minimal (irreducible smooth, almost complex, complex, symplectic, Kähler, projective) manifold. We will also fix G=π1⁢X, which we will usually take to be trivial. We can also fix the type for QX.

Remark 13.1.

Any pair determines the other pairs. Moreover, if G=1, then by results of Serre, …, Donaldson, then e,σ,t determine QX. Then Freedman tells us that QX determines the homeomorphism of X.

13.2 Geography of Compact Complex Surfaces

Let X4=S be a minimal complex surface. Then we have the following:

Kodaira

If S is not Class VII, rational, or ruled, then c12≥0 and c2≥0.

Noether

If S is minimal and has Kodaira dimension 2, then c12≥2⁢χh−6.

Bogolomov-Minyaoka-Yau

If S is not rational, ruled, or Class VII, then c12≥9⁢χh.

Yau

If c12=9⁢χh, then X is a complex ball quotient (if X is not ℙ2).

Moreover, we have a restricted complete list for:

  • •

    If κ=−∞, then X is rational or ruled.

  • •

    If κ=0, then X is Enriques, K3, T4, or bielliptic.

  • •

    If κ=1, then X is elliptic.

Theorem 13.2 (Hirokawa, Xiao, Persson-Peters,…).

All lattice points with 9⁢χh≥c12≥2⁢χh−2 and c12>0 are realized by minimal Kähler surfaces (mostly with π1 trivial).

Example 13.3.

Note that T2 and bielliptic surfaces lie at the origin. The ruled surfaces lie on y=8⁢x in the third quadrant. ℙ2 lies at the point (1,9). The Class VII and Hopf surfaces lie on the negative y-axis. The elliptic surfaces lie on the positive x-axis, with the first two being Enriques and K3.

Theorem 13.4 (Kodaira, Hirzebruch, Sommest, Persson, Moishezon-Teicher, Roulleau-Urzua).

Several lattice points with 9⁢χh>c12>8⁢χh can be realized arbitrary close to the BMY line with π1=1.

Theorem 13.5 (Mumford, Ishida-Cato, Keum, Prasad-Yeung, Cartwright-Steger).

There are exactly 1+50+1 surfaces the point (1,9). These are fake ℙ2 when b1=0.

13.2.1 Geography of Kähler Surfaces

We take the compact complex picture and simply remove the Class VII and Hopf surfaces.

13.3 Geography of Simply-Connected Symplectic 4-Manifolds

Theorem 13.6 (Taubes).

If X is not rational or ruled, then c12≥0.

Remark 13.7.

The Noether inequality fails in the symplectic case. This is due to Gompf, Fintushel-Stern, J. Park, who showed that all lattice points with 2⁢χh−6>c12>0 are realized by minimal simply-connected symplectic X.

Remark 13.8.

There are some restricted cases.

  • •

    If κ=−∞, then X is rational or ruled.

  • •

    If κ=0, then the known examples are Enriques, K3, T4, bielliptic, or T2-bundles over T2. We also know that any manifold with κ=0 has the same rational homology as one of these (Li-Bauer).

  • •

    For κ=1, Gompf and Fintuschel-Stern showed that there are infinitely more minimal symplectic 4-manifolds than elliptic surfaces. In addition, by a result of Gompf, J. Park, Stipsicz, Akhmedov-Boldridge-Baykur-Kirk-Park, there are infinitely many sumplectic manifolds on almost all lattice points with 8⁢χh≥c12≥0.

13.3.1 Five Fundamental Problems

  1. 1.

    Does every symplectic 4-manifold which is not a ruled surface have c2=e≥0?

  2. 2.

    (Symplectic BMY) Are there any symplectic 4-manifolds that are not ruled that violate BMY?

  3. 3.

    (Symplectic Yau) Are there any symplectic 4-manifolds with c12=9⁢c2 Kähler?

  4. 4.

    (Symplectic Poincare Conjecture) Every symplectic 4-manifold homeomorphic to ℙ2 is diffeomorphic to it.

  5. 5.

    (Symplectic Calabi-Yau Conjecture) Every symplectic 4-manifold with c1=0 is diffeomorphic to a K3 or a T2-bundle over T2.

Solution to any of these problems will give us an A for the course regardless of what else happens during the semester. İnanç have us five problems so we wouldn’t be fighting over them.

14 March 10

Note: I was away on this day; notes were provided by Arthur Wang.

14.1 Darboux-Moser-Weinstein Local Theory

Let ψ:M×ℝ→M be a smooth isotopy. This means that ψ⁡(−,t) is a diffeomorphism for all t and ψ⁡(−,0) is the identity on M. From this flow we may obtain a time-dependent vector field Vt such that

Equivalently, we have

or

From any time-dependent vector field Vt we can find a ψt solving the above ODE locally. If Vt is compactly supported on M (for example if M is compact), then we can obtain ψt globally.

14.1.1 Moser’s Method

We will construct an isotopy to match symplectic forms by constructing a time-dependent flow in an analogous fashion. Let ωt∈Ω2⁢M be a family of symplectic forms. Assume that

for some σt∈Ω1⁢M. We want to conclude that there exists an isotopy ψt of M such that

This will imply that (M,ωt) is symplectomorphic to (M,ω0). If M is compact, then it suffices to construct a flow satisfying (14.1).

Differentiating and integrating with respect to t, we see that (14.1) is equivalent to

by Cartan’s magic formula, which means we must have

We need this equation to have a solution for all t, but it does because ωt is nondegenerate. Therefore, we can take Vt≔μωt−1⁢(−σt).

Lemma 14.1 (Moser’s Isotopy).

Let M be a 2⁢n-dimensional manifold and S⊂M be a compact submanifold. Let ω0,ω1∈Ω2⁢M be closed forms such that their restrictions to TS⁢M are equal and nondegenerate on S. Then there exist open neighborhoods Ni⊃S and a diffeomorphism ψ:N0→N1 such that ψ∗⁢ω1=ω0 and ψ|S=idS.

Proof.

Due to Moser’s method, it suffices to show that there exists an open neighborhood N0⊃S and σ∈Ω1⁢N0 such that ω1−ω0=𝑑σ and σ|TS⁢M≡0.

Using this, we can take ωt=(1−t)⁢ω0+t⁢ω1 on N0. It is easy to see that ωt is closed for all t. Then because nondegeneracy is an open condition, we can shrink N0 to a smaller open neighborhood of S to ensure nondegeneracy of ωt. Therefore we have a family of symplectic forms ωt on N0 such that ωt=ω0+t⁢𝑑σ. Therefore we have a vector field Vt whose flow is an isotopy ψt of N0, where we shrink N0 further if needed so that ψt∗⁢ωt=ω0. Because σ is identically zero on σ|TS⁢M, we have Vt|S≡0 and thus ψt|S=idS.

To show the existence of N0, fix a Riemannian metric g on M and identify the normal bundle νS with T⁢S⟂. Consider the restriction of the exponential map T⁢S⟂→M around the open neighborhood of the zero-section:

Because S is compact, exp|Uε is a diffeomorphism for small enough ε. Set N0=exp⁡(Uε). Define ϕt:N0→N0 by ϕt⁢(exp⁡(s,v))=exp⁡(s,t⁢v) so it is a diffeomorphism for t>0 and ϕ0⁢(N0)⊂S. We also have ϕ1=idN0 and ϕt|S=idS.

Therefore, for τ=ω1−ω0∗, we have ϕ0∗⁢τ=0 and ϕ1∗⁢τ=τ. Because ϕt is a diffeomorphism for t>0, there exists a vector field

whose flow is ϕt. Therefore, for δ>0, we have

Therefore, as δ→0+, σδ→σ, so ω1−ω0=τ=ϕ1∗⁢τ=ϕ1∗⁢τ−ϕ0∗⁢τ=𝑑σ. Thus ω1−ω0=𝑑σ and

□

Theorem 14.2 (Darboux).

Let (M,ω) be a symplectic manifold. Around any point p∈M, there exists a local coordinate chart (U,{xi,yi}) such that

It follows that chart transitions for M lie in S⁢p⁢(2⁢n).

Proof.

Using any symplectic basis for Tp⁢M,ω|p, construct coordinates centered at p and defined in some neighborhood U′ of p such that ω=∑𝑑xi′∧𝑑yi′. Then apply Moser’s lemma for S={p},ω0=ω, and ω1=𝑑xi∧𝑑yi. Then there exist neighborhoods U0,U1 of p and a diffeomorphism ψ such that ψ∗⁢ω1=ω0 and ψ⁡(p)=p. Then

□

15 March 12

Note: I was away on this day; notes were provided by Arthur Wang.

Exercise 15.1.

Let X be a closed, connected, oriented almost complex 4-manifold.

  1. 1.

    Show that X⁢#⁢ℙ¯2 is also almost complex;

  2. 2.

    Show that any lattice point can be realized by a non-minimal simply-connected almost-complex manifold.

Exercise 15.2.

Let Σ be a closed symplectic surface in a closed symplectic 4-manifold (X,ω).

  1. 1.

    Show that any symplectic form on Σ is determined up to isotopy by ∫σω;

  2. 2.

    Any symplectic neighborhood of Σ is determined by Σ⋅Σ and ∫Σω.

We know there always exists a standard symplectic neighborhood of S=pt. Generally, the goal is when S⊂(M,ω) is symplectic, Lagrangian, or (co)isotropic given some data on ν⁢S, the normal bundle of S. We will obtain a standard form for ω on a small tubular neighborhood of S.

15.1 S is Symplectic

Recall that there exists J∈𝒥⁡(M,ω) such that S is J-holomorphic. Then ν⁢S=(T⁢S)ω=(T⁢S)⟂, so ν⁢S is a symplectic vector bundle on S. Then recall that the isomorphism class is determined by the isomorphism class of the complex vector bundle (ν⁢S,J). We will show that a neighborhood of S is completely determined by ω|S and the isomorphism class of (ν⁢S,ω).

Theorem 15.3 (Weinstein, Symplectic neighborhood theorem).

For j=0,1, let (Mj,ωj) be symplectic manifolds with compact symplectic submanifolds Sj⊂Mj such that there exists a vector bundle isomorphism Φ:(ν⁢S0,ω0)→(ν⁢S1,ω1) commuting with a symplectomorphism ϕ:S0→S1. Then ϕ extends to a symplectomorphism ψ of neighborhoods Ni⊃Si with d⁢ψ=Φ.

Proof.

There exists Jj∈𝒥⁡(Mj,ωj) and compativle gj for which Sj is Jj-holomorphic and ν⁢Sj=T⁢Sj⟂. Under these identifications, let φj:ν⁢Sj→Mj be the exponential maps. Then

is a diffeomorphism from a neighborhood of S0 to a neighborhood of S1 where

Then use the Moser isotopy lemma for S0,φ∗⁢ω1,ω0. □

15.2 L is Lagrangian

In this case, dimL=n=dimM/2, and we will see that the symplectomorphism class of a tubular neighborhood of L is completely determined by the diffeomorphism class of L. Observe that if L⊂(V,ω) is a Lagrangian subspace of a symplectic vector space, then ω gives a canonical identification of V/L with L∗.

In the manifold case, if L⊂(M,ω) is Lagrangian, then ν⁢L=T∗⁢L. Therefore a neighborhood of L in M is diffeomorphic to a neighborhood of the zero-section of the cotangent bundle T∗⁢L.

Example 15.4 (Canonical Symplectic Structure on the Cotangent Bundle).

Let L be any n-dimensional smooth manifold and M=T∗⁢L. Then we will define the tautological 1-form λ on M. We have the projection π:M→L. For any v∈M=T∗⁢L, v is pulled back by π to π∗⁢v∈Tv∗⁢M. Then we define λ∈Ω1⁢M by

The canonical symplectic structure ω on M is ω=−𝑑λ, an exact form.

In local coordinates, if pj are the coordinates on L and qj are the cotangent coordinates, then we have

In fact λ is characterized by the property that for all σ∈Ω1⁢L, σ∗⁢λ=σ. By the local characterization, ω is symplectic. Finally, it is relatively easy to check that L is Lagrangian in T∗⁢L (just use the local description).

Theorem 15.5 (Weinstein, Lagrangian neighborhood theorem).

Let (M,ω) be a symplectic manifold and L⊂(M,ω) be a compact Lagrangian submanifold. Then there exist neighborhoods U of the zero section in T∗⁢L and V of L in M and a diffeomorphism ϕ:U→V such that ϕ∗⁢ω=−𝑑λ that is the identity on L.

Example 15.6.

Let L be a closed Lagrangian surface in a closed symplectic 4-manifold (X,ω). Since ν⁢L≃T∗⁢L≃T⁢L as bundles over L, then they have the same Euler class. Then we see that eT⁢L⁢[L]=eν⁢L⁢[L], so L⋅L=e⁡(L). Thus the isomorphism class of ν⁢L is determined by the diffeomorphism type of L.

By Theorem 15.5 there exists a Weinstein neighborhood N⁡(L)⊂(X,ω) such that

Observe that any radial push-off of L in N⁡(L) is also Lagrangian.

For L=T2, note that L⋅L=e⁡(L)=0 and T∗⁢L is trivial, so N⁡(L)≃T2×Bε⁢(0).

Exercise 15.7.

Let X=ℙ1×ℙ1 and ω=π1∗⁢ωF⁢S+π2∗⁢ωF⁢S. Then recall that ℝ⁢ℙ1 is the fixed locus of complex conjugation on ℙ1. Then L=ℝ⁢ℙ1×ℝ⁢ℙ1 is Lagrangian. How about a Lagrangian Klein bottle?

16 March 24

İnanç decided on the following format for online class:

  • •

    He sends us lecture notes in a private Dropbox folder for us to read.

  • •

    There will be a Zoom discussion every Thursday at the usual time.

From now on, these notes will simply be a transcription of the notes we were sent.

16.1 Proof of Lagrangian Neighborhood Theorem

Theorem.

Let (M,ω) be a symplectic manifold and L⊂(M,ω) be a compact Lagrangian submanifold. Then there exist neighborhoods U of the zero section in T∗⁢L and V of L in M and a diffeomorphism ϕ:U→V such that ϕ∗⁢ω=−𝑑λ that is the identity on L.

Proof.

First note that if (V,ω) is a symplectic vector space with L Lagrangian, then J⁢L is also Lagrangian for any almost complex structure J. Finally, L⟂J⁢L under a compatible metric g.

For (M,ω) a symplectic manifold with L⊂(M,ω) Lagrangian and J∈𝒥⁡(M,ω), the above applies to J⁢T⁢L. In addition, J⁢T⁢L=T⁢L⟂. Identifying this with ν⁢L, then we have an isomorphism with T∗⁢L. Then for g compatible with ω,J, take the exponential map

defined by

Fixing the decomposition T(q,0)⁢T∗⁢L=Tq⁢L⊕Tq∗⁢L for q∈L, write any v∈T(q,0)⁢T∗⁢L as v=(v0,v1∗), then 𝑑ω(q,0)⁡(v)=v0+β⁡(v1∗). Therefore

On the other hand, at q∈L, {∂∂qj},{𝑑qj} form a basis for Tq⁢L and Tq∗⁢L, respectively. Therefore, for u=∑aj⁢∂∂qj+∑bj⁢𝑑qj and v=∑xj⁢∂∂qj+∑yj⁢𝑑qj, then

Therefore ψ∗⁢ω=−𝑑λ on T(1,0)⁢T∗⁢L for all q∈L. The desired result then follows from Moser’s isotopy lemma. □

Example 16.1.

Let L be a closed Lagrangian surface in a closed symplectic 4-manifold (X,ω). Because ν⁢L≅T∗⁢L≅T⁢L as bundles over L, then eν⁢L⁢[L]=eT⁢L⁢[L], which implies that L⋅L=e⁡(L)=2−2⁢g.

For L≅Σg, then the diffeomorphism type of L determines the isomorphism class of ν⁢L. By the Lagrangian neighborhood theorem, there exists a Weinstein neighborhood N⁡(L)⊂(X,ω) such that

Ovserve that any “radial push-off” of L in N⁡(L) is also Lagrangian. For L=T2, then L⋅L=e⁡(L)=0 and T∗⁢L=L×ℝ2, so N⁡(L)=T2×Bε⁢(0).

17 March 26

17.1 More Local Theory

Example 17.1.

Let X=ℙ1×ℙ1=S2×S2 with ω=p1∗⁢(ωF⁢S)+p2∗⁢(ωF⁢S). Then any S2×y0 or x0×S2 are symplectic submanifolds because it is easy to see that i∗⁢ω=ωF⁢S.

Now let L=ℝ⁢ℙ1×ℝ⁢ℙ1 which is the fixed locus of complex conjugation. Then N⁡(L)≅T2×ℝ2. Here, for u,v∈T⁡(S1×S1)≅T⁢S1×T⁢S1 with u=(u1,u2) and v=(v1,v2), then we have

Observe that for all S1→a⊂S2 and S1→b⊂S2, a×b is Lagrangian, so all La,b are Lagrangian tori. All of these are boundaries, so they are topologically trivial.

Exercise 17.2.

Construct a Lagrangian Klein bottle in the same symplectic manifold. Also construct a homologically essential Lagrangian torus in Σg×Σg.

Remark 17.3.

There are similar theorems for isotropic manifolds (due to Weinstein) and coisotropic manifolds (due to Gotny) with more information on ω in a neighborhood.

17.2 Another Application of Isotopy

Another application of Moser’s isotopy conncerns equivalence of symplectic forms on a given manifold. Let ω0,ω1 be two symplectic forms on M. They they are

Symplectomorphic

if there exists a diffeomorphism ϕ of M such that ϕ∗⁢ω1=ω0.

Deformation equivalent

if there exists a smooth family ω∗ of symplectic forms joining ω0 to ω1.

Isotopic

if there exists a deformation equivalence ω∗ where [ω∗] is fixed.

Strongly isotopic

if there exists an isotopy ϕ∗ of M such that ϕ1∗⁢ω1=ω0.

Observe that strongly isotopic implies isotopic which implies deformation equivalent. Also strongly isotopic implies symplectomorphic. If M is compact, then a result of Moser says that isotopic implies strongly isotopic.

Theorem 17.4 (Moser Stability).

Let M be a closed symplectic manifold and ω∗ be a smooth family of symplectic forms all in the same cohomology class. Then there exists an isotopy ψ∗ of M such that ψt⁢ωt=ω0.

Sketch of Proof.

To apply Moser’s argument, we need a smooth family σt∈Ω1⁢M such that d𝑑t⁢ωt=𝑑σt. Because the ωt are in the same cohomology class, then τt=ddt⁡ωt must be exact. For each t, there exists σt∈Ω1⁢M such that dσt=τt. Such a smooth family is constructed using the Poincare lemma and an inductive argument on the number of sets good covers of M. □

Corollary 17.5.

Let Sa={ω∈Ω2⁢M∣ω⁢is symplectic and⁢[ω]=a}. Then for a closed symplectic manifold M, any two forms on the same path component of Sa are symplectomorphic.

Example 17.6.

For M=Σ with ω0,ω1 symplectic forms on Σ, then if [ω0]=ω1, there exists an isotopy ψt of Σ such that ψ1∗⁢ω1=ω0.

18 March 31

18.1 Constructions of Symplectic Manifolds Through Surgery

Here we will study important examples of symplectic surgeries based on the local theory.

18.1.1 As Smooth Operations

Blowup

Here M is replaced by

Blowdown

Consider ℂ⁢ℙn−1≅S⊂M with ν⁢S≅𝒪⁡(−1). For N⁡(S)≅ν⁢S, we have ∂N⁡(S)≅S2⁢n−1. Then we replace M with

Fiber Connected Sum

Consider Si2⁢n−2⊂Mi2⁢n such that there exists an orientation reversing bundle isomorphism ϕ¯:ν⁢S1→ν⁢S2. This induces ψ:∂N⁡(S1)→∂N⁡(S2). Then we replace M with

Torus Surgery

Let L⊂X4 be a Lagrangian torus with trivial normal bundle. Then we replace X by

where N⁡(L) is twisted by the automorphism ψ=η−1∘ϕ∘η for some diffeomorphism ϕ of T3.

Rational Blowdown

Let S1,…,Sp−1 be embedded S2 in X4 such that the space Cp formed by plumbing disc bundles in the following sequence:1

∙∙...∙−(p+2)−2−2up−1up−2u1

embeds in X4. Then ∂Cp=∂Bp=L⁡(p2,p−1). Here Bp is a rational ball (meaning it has the same rational homology as the disc) with fundamental group ℤ/p⁢ℤ. Then rational blowdown is the process of removing the interior of Cp and replacing it with Bp. Note taker: I only understood this part after looking at the original paper by Fintushel and Stern.

18.1.2 As Symplectic Operations

Symplectic Blowup

First recall the construction in the complex category. Here, the local model replaces ℂn with the total space of 𝒪⁡(−1), the tautological line bundle over ℙn−1.2

Recall that there are two holomorphic projections π:C~n→ℂn and p:C~n→ℙn−1, which correspond to the blowup map and the projection from 𝒪⁡(−1). The fiber E=π−1⁢(0) is called the exceptional divisor of the blowup. The normal bundle of E in C~n is simply 𝒪⁡(−1), where E is the zero section. Note that c1⁢(L)=−H, where H is the hyperplane class.

On the other hand, the normal bundle ν⁢ℙn−1 of ℙn−1 in ℙn is simply 𝒪⁡(1). A complex line bundle on projective space is entirely determined by its first Chern class, so

and therefore

19 April 2

19.1 Complex Blowups

Note taker: A good reference for this lecture from the complex algebraic perspective is Griffiths and Harris, Principles of Algebraic Geometry.

Example 19.1.

For n=2, any exceptional divisor on a surface has self-intersection number −1. To see this, note that c1⁢(L)=−h.1

It is possible to show that any local biholomorphic map ℂn→ℂn fixing the origin lifts to a local biholomorphic map on the blowup. Therefore, we can blow up any complex manifold M. From the topological point of view, this simply replaces M with

In addition, it is easy to see that

In particular, for surfaces, we see that the canonical class of the blowup is

Example 19.2.

Let X=Bl1⁡ℙ2. Then we have a projection to ℙ1 with fiber ℙ1. In addition, the exceptional divisor is the zero section, so this is a Hirzebruch surface. In addition, we have F⋅F=0 and E⋅F=1.

Example 19.3.

Let C0,C1 be two transverse, nonsingular cubics in ℙ2. Then, we can consider the linear system generated by C0,C1 and resolve it. This will be a blowup of ℙ2 in 9 points. In addition, the linear system gives us a map to ℙ1 which has generic fiber an elliptic curve. Thus X≅E⁡(1), an elliptic surface.

Example 19.4.

The local model for the blowups above can be seen in the standard example of “resolution of singularities.” Let C=(z1⁢z2=0)⊂ℂ2. Then the proper transform of C is simply a disjoint union of two lines.

Remark 19.5.

In general, one can blow up along any complex subvariety to resolve singularities.2

20 April 7

20.1 Symplectic Operations

In a symplectic manifold, we can blow up following a local J-holomorphic model to obtain a new symplectic manifold (M⁢#⁢ℙ¯n,ω~), where ω~ depends on some additional parameters.

Let ψ:B2⁢n⁢(x)→M be a symplectic embedding of a ball of radius λ. Then extend this by some ε>0 and replace ψ:B2⁢n⁢(λ2+ε2) by the standard ε-neighborhood of the zero section in L. Then recall that L is the tautological bundle over ℙn−1, so we can write

This is a Kähler form. Let L⁡(ε) be the ε-neighborhood of the zero section in L. Then

so we can glue symplectically. Moreover, we can normalize the construction so that:

Proposition 20.1 (McDuff).

The deformation class of ω~ is unique, and the isotopy class is unique if we fix λ.

Theorem 20.2 (McDuff, Symplectic Blowup).

Given a symplectic manifold (M,ω), a compatible J, and a point p0∈M, we can symplectically blow up M at p0 and obtain a new symplectic manifold (M~,ω~) such that π:M~→M is holomorphic, the exceptional divisor is J~-holomorphic, and

Any two symplectic blowups of (M,ω) at points p0,p1 are equivalent up to symplectomorphism and deformation equivalence.

Remarks 20.3.
  1. 1.

    Complex blowup is infinitesimal, while symplectic blowup is local;

  2. 2.

    Complex blowup is intrinsic, while symplectic blowup depends on λ.

  3. 3.

    If M is Kähler complex blowup is a symplectic blowup, but not the other way around.

  4. 4.

    Blowup can simplify the topology of (singular) submanifolds and thei configurations, while complicating the topology of the ambient manifold. Also, blowing up decreases the symplectic volume.

  5. 5.

    Complex curves in a complex surface intersect positively if if they are distinct. By a result of Gromov, the same holds for almost complex J.

    However, this is not true for symplectic surfaces in a symplectic 4-manifold. For example, in the smooth category, we can move the exceptional divisor E to E′, and then E⋅E′=−1. Thus resolution of singularities does not behave well in the symplectic category.

21 April 9

21.1 Symplectic Blowdown

Suppose there exists a symplectic submanifold ℙn−1≅E⊂(M,ω) and ν⁢E≅L, then by the symplectic neighborhood theorem, there exists a small neighborhood N⁡(E) isomorphic to L⁡(ε). Therefore, we can symplectically blow-down E to obtain a new symplectic manifold

By a result of McDuff, only the choice of E matters.

Theorem 21.1 (McDuff, Symplectic Blowdown).

Given a symplectic manifold (M,ω), compatible J, and symplectic exceptional divisor E, we can symplectcially blow down M along E and obtain a new symplectic manifold (M′,ω′) sith compatible J′ so that that the projection M→M′ is holomorphic. Any two blowdowns along the same E are symplectomorphic.

Example 21.2.

Let X=ℙ1×ℙ1 with the product Kähler structure. If we blowup at a point p0=(x0,y0), then denote F=x0×ℙ1, S=ℙ1×y1 with y1≠y0. These are complex submanifolds of X. Then let [F~],[S~] be the proper transforms of F,S in the blowup, we see that [F~]=π∗⁢[F]−E,[S~]=π∗⁢[S]−E.1

Now we blow down F~ and obtain (S′,ω′). Then note that S is another symplectic exceptional sphere, which can be blown down again to obtain ℙ2. Therefore X~ has two different minimal models.

If we perform only complex blowdowns, then X′′=ℙ2 because there exists a unique symply-connected complex surface on the BMW line. Thus X′=Bl⁡ℙ2. If we perform symplectic blowdowns without worrying about complex structures, then E→E′′, and there exists a symplectic E′′≅ℙ1 with self-intersection +1.

Remark 21.3.

Note that the local model of the blowup is always the same, so we must always have [F~]2=F2−1.

Proposition 21.4 (McDuff).

If there exists a homologically essential degree of non-negative self-intersection in (X,ω), then X is rational or ruled.

Remark 21.5.

This implies that (X′′,ω′′)≅(ℙ2,ωF⁢S) in our example.

22 April 14

22.1 Symplectic Fiber Connected Sum

Let (Mi,ωi) be symplectic manifolds of dimension 2⁢n and let Si be codimension 2 symplectic submanifolds. Here ν⁢Si is determined by the Euler number of Si. Then there exists an orientation-reversing bundle isomorphism ν⁢S1→ν⁢S2 if there is an orientation preserving isomorphism S1→S2 and e1=−e2.

For the symplectic construction, we want the orientation-reversing bundle isomorphism to be a symplectomorphism. For simplicity, let’s suppose the normal bundle is trivial.

Then by the symplectic neighborhood theorem, there exists a neighborhood N⁡(Si) such that N⁡(Si)≅B2⁢(r0)×Si. For any 0<r1<r0, we have a symplectic submanifold, the annulus A2⁢(r1,r0)⊂B2⁢(r0). Then given a symplectomorphism between the bundles, we obtain a symplectomorphism A2⁢(r1,r0)×S1→A2⁢(r1,r0)×S2 by switching the two boundary components of the annulus. This gives us a symplectomorphism

where ξi is the isomorphism N⁡(Si)→B2⁢(r0)×Si. Therefore, we can set

Remarks 22.1.
  1. 1.

    The same construction extends to the case 0≠e1=−e2, where instead of the product, we take an annular fiber bundle isomorphism.

  2. 2.

    M is determined up to orientation preserving diffeomorphism by the pairs (Mi,Si), trivializations ξi, and the symplectomorphism ϕ:S1→S2. Often the trivializations are understood from the context, for example when the Si are fibers of a fibration.

  3. 3.

    When 2⁢n=4, there exists a symplectomorphism ϕ:S1→S2 if they are diffeomorphic and have the same volume. The latter can always be achieved by multiplying one of the ωi by a constant. Thus it is enough to have g⁡(S1)=g⁡(S2) and S12=−S22.

  4. 4.

    When 2⁢n=2, then the Si are each a single point, and the fiber connected sum is the same as the regular connected sum.

  5. 5.

    Note that one cannot find a symplectomorphism of B2⁢n⁢(r0)∖B2⁢n⁢(r1) when n>0. Otherwise, we could glue two symplectic discs to form a symplectic sphere.

Theorem 22.2 (Symplectic Fiber Connected Sum).

For j=1,2 let (Mj,ωj) be 2⁢n-dimensional symplectic manifolds with (2⁢n−2)-dimensional submanifolds Sj. If there exists a symplectomorphism ϕ:Si→S2 and the Euler numbers ej of ν⁢Sj satisfy e1=−e2, then there exists a new symplectic manifold (M,ω)=(M1⁢ω1)⁢#ϕ⁢(M2,ω2).

For 2⁢n=4, it is enough to have g⁡(S1)=g⁡(S2) and S12=−S22.

Theorem 22.3 (Usher).

If the Xj are minimal and g⁡(Sj)≥1, then X=X1⁢#S1=S2⁢X2 is minimal.

23 April 16

Example 23.1.

Consider E⁡(1)=Bl9⁡ℙ1 and F a smooth fiber. Then we know F2=0. If we let (E⁡(2),ω2)=(E⁡(1),ω1)⁢#⁢(E⁡(1),ω1), and more generally,

then the elliptic fibration on E⁡(1) extends to E⁡(n). In fact, all E⁡(n) are Kähler. In fact, the pullback of E⁡(1) along z↦zn is birational to E⁡(n).

Now we will study the algebraic topology of X=E⁡(2). We know that

Here, π1⁢(E⁡(1))≅π1⁢(E⁡(1)∖N⁡(F))∗π1⁢(N⁡(F))≅π1⁢(E⁡(1)∖N⁡(F))∗ℤ2. Recall that F is a 9 times blown-up smooth cubic, so π1⁢(E⁡(1))≅π1⁢(E⁡(1)∖N⁡(F))=F, and thus π1⁢(X)=1. Next, we know that

and e⁡(E⁡(1))=e⁡(E⁡(1)∖N⁡(F)), so e⁡(E⁡(2))=24. In addition, we can show that

In general, E⁡(n) is simply connected, e⁡(E⁡(n))=12⁢n, and σ⁡(E⁡(n))=−8⁢n. Then, we see that K2=2⁢e+3⁢σ=0. Therefore, all E⁡(n) have Kodaira dimension.

Proposition 23.2 (Adjunction Formula).

If Σ⊂(X,ω) is a symplectic surface, then −e⁡(Σ)=[Σ]2+KX⋅Σ. In particular, it holds for any J-holomorphic Σ.

Proof.

Note that T⁢X|Σ is a symplectic vector bundle. In particular, T⁢Σ is a symplectic subbundle. Therefore, we can write

so we have

Applying this to intersection with Σ gives us the desired formula. □

From this, we see that any section of E⁡(2) has self-intersection −2. In general, there exists a symplectic sphere of self-intersection −n in E⁡(n). Therefore K⋅S=n−2, so K is not torsion for n≠2. Thus E⁡(n) has Kodaira dimension 1 for n≥3, while E⁡(2) is a K3 surface.

24 April 21

24.1 Luttinger Surgery

This is a symplectic surgery along a Lagrangian torus. When 2⁢n=4, L is an embedded Lagrangian torus inside a symplectic 4-manifold (X,ω). Let γ be a loop inside L which is co-oriented.

Here N⁡(L)≅ν⁢L≅T∗⁢L≅T2×ℝ2 is the trivial bundle. Therefore we can perform a torus surgery along L in X. To construct an explicit model, note that by Weinstein, there exists a neighborhood N⁡(L) of L diffeomorphic to T∗⁢L, where L corresponds to the zero-section. Moreover, T∗⁢L≅T2×ℝ2, where we can identify T2=ℝ2/ℤ2=q⁡(ℝ2) with coordinates x1,x2 and γ=ℝ/ℤ=q⁡(ℝ) with co-orientation ∂∂x2.

Thus for (y1,y2) the dual coordinates in the cotangent fibers, we have

where L⟷T2×0. In fact, any T2×pt is Lagrangian. We call ξ a Lagrangian framing.

Let r>0 such that Ur=ℝ2×ℤ2×[−r,r]2⊂ξ⁡(N⁡(L)). Then let 𝒳:[−r,r]→[0,1) be a C∞ step function such that 𝒳⁡(t)=0 for all t≤−r2 and 𝒳⁡(t)=1 for t≥r3 and 𝒳′⁢(t)>0 in between. Finally, we want

For any k∈ℤ, define ϕi:(Ur∖Ur/2)→(Ur∖Ur/2) by

We can check that ϕk is a symplectomorphism.

Luttinger surgery replaces (X,ω) with (X′,ω′), where

and ω′=ω on X and ω0 on Ur. We will denote X′=X⁡(L,γ,k). Note that (L,γ,k) determine X up to orientation-preserving diffeomorphism:

  • •

    L and the Lagrangian framing ξ determine N⁡(L);

  • •

    The co-oriented γ determines the x1 direction and the choice of k determines ϕk;

  • •

    Different 𝒳 determine isotopic ϕk.

Under the Lagrangian framing, γ can be pushed off to a curve γ′ in a neighborhood of L. The image of γ′ under regluing determines the diffeomorphism typd of X′ by handle theory.

Theorem 24.1 (Luttinger, Auroux-Donaldson-Katzarkov).

(X′,ω′)=(X⁡(L,γ,k),ω′) is uniquely determined up to symplectomorphism.

Example 24.2.

Let ωΣ be a symplectic form on Σ and ϕ be a symplectomorphism of (Σ,ωΣ).

and

where G is generated by g1⁢(s,t,x)=(s+1,t,x) and g2⁢(s,t,x)=(s,t+1,ϕ⁡(x)). Because both of these are symplectomorphisms, X is a symplectic manifold. Observe that X is a Σ-bundle over T2 with has monodromy the identity in the s-direction and ϕ in the t-direction.

Now let γ be a loop in a fiber F≅Σg of this bundle. Then L≔S1×T0×γ is a Lagrangian torus inside (X,ω). Therefore, we have an isomorphism

where Tγ0 is a positive Dehn twist along γ for the opposite co-orientation. This gives a new Σ-bundle over 𝕋2 with trivial monodromy s in the s-direction and monodromy Tγ0k∘ϕ in the t-direction.

In particular, for X=T4 and γ=(s0,t0)×S1×pt, we get X⁡(L,γ,1) is the Kodaira-Thurston manifold.

Remark 24.3.

By a result of Ho-Li, there exists an embedded surface S⊂X∖N1⁢(L)=X0 such that for ιX:X0→X, (ιX)∗⁢[S]=Kω and for ιX′:X0→X′, (ιX′)∗⁢[S]=Kω′.1

When X,X′ are minimal, we can compute the Kodaira dimensions:

Therefore, the Kodaira dimensions are the same!.

Theorem 24.4 (Enriques-Kodaira).

There exist finitely many Kähler manifolds C⊃T4 with Kodaira dimension 0.

Now, any Σg-bundle over Σn with g,h≥1 is acyclic (π2=0). Therefore, it is minimal. Then for all k∈ℤ, if we set Xk≔T4⁢(L,γ,k) as above, then every T2 bundle over T2 is a minimal symplectic manifold with the same Kodaira dimension as T4. We can show that H1⁢(Xk)=ℤ3⊕ℤ/k⁢ℤ, so we get only symplectic manifolds, not Kähler manifolds.

25 April 23 and 28

Note that İnanç posted both of these lectures in the same file.

25.1 Fundamental Groups of Symplectic Manifolds

Theorem 25.1 (Gompf).

Any finitely presentable group G is the fundamental group of a closed symplectic 4-manifold.

Remark 25.2.

The analogous result is true in any larger dimension by taking products with ℂ⁢ℙ1.

Our proof of Theorem 25.1 will use the following trick due to Gompf of turning homologically essential Lagrangian submanifolds into symplectic submanifolds in dimension 4.

Proposition 25.3.

Let (X4,ω) be a closed symplectic 4-manifold and F1,…,Fr be closed, connected, oriented, disjoint embedded Lagrangian submanifolds of (X,ω). Suppose that [F1],…,[Fr]∈H2⁢(X,ℝ) lie in an affine subspace that does not contain 0. Then there exists an arbitrarily small perturbation ω′ of ω such that (X,ω′) is symplectic and the Fi are symplectic submanifolds (respecting the given orientations and of equal area).

Proof.

The affine subspace hypothesis implies there exists a linear function on H2⁢(X,ℝ) evaluating to 1 on each [Fi]. Therefore, there exists a closed 2-form η such that

for all i. Then let ωi be a symplectic form on Fi with ∫Fiωi=1. Thus for ji:Fi→X the inclusion map, we have

so ωi−j∗⁢η=𝑑αi for some 1-form αi on Fi. Now extend αi to a form on X by first pulling it back over a small tubular neighborhood N⁡(Fi). Then we smoothly taper it to zero outside of N⁡(Fi). Therefore

is a closed 2-form such that ji∗⁢η′=ωi for all i. If we set ω′=ω+t⁢η′ for a fixed t>0, ω′ is closed. Because nondegeneracy is an open condition, ω′ is also nondegenerate for small enough t. Moreover, ji∗⁢ω′=0+t⁢ωi is a symplectic form on Fi with area equal to t for all i. □

Remark 25.4.

Now suppose ω′=ω+t⁢η is degenerate at x∈X for some t>0. Then there exists some 0≠u∈Tx⁢X such that ω⁡(u,v)+t⁢η⁢(u,v)=0 for all v∈Tx⁢X. If this is also true for t/2, then ω⁡(u,v)=0 for all v∈Tx⁢X. Thus for each x∈X, there exists tx>0 for which ω′ is nondegenerate. Because X is compact, we can find a global t.

Proof of Theorem 25.1.

Fix a finite presentation ⟨g1,…,gk|r1,…,rℓ⟩≅G. Then let F=Σk and let a1,b1,…,ak,bk be a standard collection of oriented embedded circles on F. Here, ai⋅bj=δi⁢j.When suitably attached to a base point, a1,b1,…,ak,bk are generators of π1⁢F. Here,

Therefore, π1⁢F/N⁡(b1,…,bk)≅⟨a1,…,ak⟩, the free group in the ai. Under the above isomorphism, each condition ri is a word in ai±1.

Let γi be a smooth, immersed, oriented circle in F with only double points representing this word. We can then add handles at self-intersection points of each γi, so each γi is embedded in some Σk⁢#⁢M⁢T2. We also add handles so each γi is non-separating.

Let F=Σg, where g=k+M be this bigger surface, where we have additional ai,bi for i=k+1,…,g. For S={b1,…,bk,γ1,…,γk,ak+1,bk+1,…,ag,bg}, we have π1⁢F/N⁡(S)≅G.

For F≅Σg and S a collection of circles as above, we can assume g≥1 by adding trivial relations if needed. To simplify the notation, label the circles in S as ci for i=1,…,m.

Take X=T2×F with the product symplectic form. For S1=[0,1]/0∼1, let 0<t1<⋯<tm<1. Then the embedded tori Li≔S1×ti×ci are all Lagrangian in (X,ω). Moreover, since each ci is non-separating in G, there exists di⊂F such that |ci∩di|=1. Therefore, the torus Di≔pt×S1×di is dual to Li.

Applying Proposition 25.3 repeatedly, we can perturb ω to ω′ so that all Li are now symplectic. Clearly Li2=0, so each is a symplectic torus with trivial neighborhood.

Therefore, π1⁢X≅π1⁢T2⊕π1⁢F. Let p0∈F∖S. Then L0=T2×p0 is also a symplectic torus with trivial neighborhood and disjoint from the Li.

Let (XG,ωG) be the symplectic manifold we obtain by symplectic fiber sum along L0,…,Lm with m+1 copies of E⁡(1) along a smooth fiber F. Because π1⁢(E⁡(1)∖F)=1, we apply Seifert-van Kampen to obtain

Remarks 25.5.
  1. 1.

    We can get infinitely many many symplectic manifolds with the same fundamental group G by connected sum with E⁡(n) instead of E⁡(1).

  2. 2.

    By a result of Usher, XG is minimal.

  3. 3.

    κ⁡(XG)=1.

  4. 4.

    There are infinitely many non-Kähler symplectic manifolds, all on the “x”-line,

Example 25.6.

For example, there exists a closed symplectic 4-manifold X with π1⁢X=ℤ∗ℤ.