Topics in Topology (MA73327)

By: Muhammad Said

Department of Mathematics, Pusan National University,
Busan 46241, Rep. of Korea

August 11, 2023

Solution of exercises for the course,

"Topics in Topology"
Problem 1.
Let (𝑉 , Ω) be a symplectic vector space, that is, Ω is a skew-symmetric non-degenerate bilinear
form. For a linear subspace 𝑌 of 𝑉 , the symplectic orthogonal of 𝑌 is the linear subspace defined
by
𝑌 Ω = {𝑣 ∈ 𝑉 | Ω(𝑣, 𝑢) = 0 for all 𝑢 ∈ 𝑌 }.

Prove the following statements.

1. Show that dim 𝑌 + dim 𝑌 Ω = dim 𝑉 .
2. Show that (𝑌 Ω )Ω = 𝑌 .
3. Prove that if 𝑌 and 𝑊 are subspaces of 𝑉 , then 𝑌 ⊂ 𝑊 ⟺ 𝑊 Ω ⊂ 𝑌 Ω .
Solution.
1. Show that dim 𝑌 + dim 𝑌 Ω = dim 𝑉 .
Proof:
Let {𝑦1 , … , 𝑦𝑘 } be a basis for 𝑌 . We want to show that {𝑦1 , … , 𝑦𝑘 , 𝑣1 , … , 𝑣dim 𝑌 Ω } forms a basis for
𝑉 , where {𝑣1 , … , 𝑣dim 𝑌 Ω } is a basis for 𝑌 Ω .
Since Ω is non-degenerate, for each 𝑖 = 1, … , 𝑘, there exists a vector 𝑣𝑖 ∈ 𝑉 such that Ω(𝑣𝑖 , 𝑦𝑗 ) =
𝛿𝑖𝑗 (the Kronecker delta), and Ω(𝑣𝑖 , 𝑦𝑘 ) = 0 for all 𝑘 ≠ 𝑗.
Ω
Now, let 𝑣 ∈ 𝑉 be any vector. We can uniquely write 𝑣 = ∑𝑘𝑖=1 𝛼𝑖 𝑦𝑖 + ∑𝑗=1
dim 𝑌
𝛽𝑗 𝑣𝑗 . Then, for each
𝑖 = 1, … , 𝑘, we have
𝑘 dim 𝑌 Ω
Ω(𝑣, 𝑦𝑖 ) = ∑ 𝛼𝑗 Ω(𝑦𝑗 , 𝑦𝑖 ) + ∑ 𝛽𝑗 Ω(𝑣𝑗 , 𝑦𝑖 ) = 𝛼𝑖 ,
𝑗=1 𝑗=1

and for each 𝑗 = 1, … , dim 𝑌 Ω , we have

𝑘 dim 𝑌 Ω
Ω(𝑣, 𝑣𝑗 ) = ∑ 𝛼𝑖 Ω(𝑦𝑖 , 𝑣𝑗 ) + ∑ 𝛽𝑖 Ω(𝑣𝑖 , 𝑣𝑗 ) = 𝛽𝑗 .
𝑖=1 𝑖=1

1
Thus, the vectors {𝑦1 , … , 𝑦𝑘 , 𝑣1 , … , 𝑣dim 𝑌 Ω } span 𝑉 .
Ω
To show linear independence, suppose we have ∑𝑘𝑖=1 𝛼𝑖 𝑦𝑖 + ∑𝑗=1dim 𝑌
𝛽𝑗 𝑣𝑗 = 0. Taking the sym-
plectic inner product with 𝑦𝑘 for 𝑘 = 1, … , 𝑘 yields 𝛼𝑘 = 0 for all 𝑘. Similarly, taking the
symplectic inner product with 𝑣𝑗 for 𝑗 = 1, … , dim 𝑌 Ω yields 𝛽𝑗 = 0 for all 𝑗. Thus, the set
{𝑦1 , … , 𝑦𝑘 , 𝑣1 , … , 𝑣dim 𝑌 Ω } is linearly independent.
Therefore, dim 𝑌 + dim 𝑌 Ω = dim 𝑉 .
2. Show that (𝑌 Ω )Ω = 𝑌 .
Proof:
Let 𝑣 ∈ (𝑌 Ω )Ω . By definition, 𝑣 ∈ (𝑌 Ω )Ω if and only if Ω(𝑣, 𝑢) = 0 for all 𝑢 ∈ 𝑌 Ω . This implies
that 𝑣 ∈ 𝑌 , as Ω(𝑣, 𝑢) = 0 for all 𝑢 ∈ 𝑌 .
Conversely, let 𝑣 ∈ 𝑌 . We want to show that 𝑣 ∈ (𝑌 Ω )Ω , i.e., Ω(𝑣, 𝑢) = 0 for all 𝑢 ∈ 𝑌 Ω . Since
𝑣 ∈ 𝑌 , we have Ω(𝑣, 𝑢) = 0 for all 𝑢 ∈ 𝑌 , by definition of 𝑌 Ω .
Therefore, (𝑌 Ω )Ω = 𝑌 .
3. Prove that if 𝑌 and 𝑊 are subspaces of 𝑉 , then 𝑌 ⊂ 𝑊 if and only if 𝑊 Ω ⊂ 𝑌 Ω .
Proof:
(⇒) Suppose 𝑌 ⊂ 𝑊 . We want to show that 𝑊 Ω ⊂ 𝑌 Ω .
Let 𝑣 ∈ 𝑊 Ω . By definition, 𝑣 ∈ 𝑊 Ω if and only if Ω(𝑣, 𝑢) = 0 for all 𝑢 ∈ 𝑊 . Since 𝑌 ⊂ 𝑊 , it
follows that Ω(𝑣, 𝑢) = 0 for all 𝑢 ∈ 𝑌 . Therefore, 𝑣 ∈ 𝑌 Ω .
(⇐) Now, suppose 𝑊 Ω ⊂ 𝑌 Ω . We want to show that 𝑌 ⊂ 𝑊 .
Let 𝑦 ∈ 𝑌 . Since 𝑊 Ω ⊂ 𝑌 Ω , we have Ω(𝑣, 𝑦) = 0 for all 𝑣 ∈ 𝑊 Ω . This implies that Ω(𝑤, 𝑦) = 0
for all 𝑤 ∈ 𝑊 , as 𝑊 Ω is a subset of 𝑊 . Therefore, 𝑌 ⊂ 𝑊 .
Problem 2.
Let (𝑉 , Ω) be a symplectic vector space, that is, Ω is a skew-symmetric non-degenerate bilinear
form.
1. For a nonzero vector 𝑒1 ∈ 𝑉 , show that there is a vector 𝑓1 ∈ 𝑉 such that Ω(𝑒1 , 𝑓1 ) = 1.
2. Show that the restriction of the bilinear form Ω to the subspace generated by 𝑒1 and 𝑓1 is
non-degenerate.
Solution.

1. For a nonzero vector 𝑒1 ∈ 𝑉 , show that there is a vector 𝑓1 ∈ 𝑉 such that Ω(𝑒1 , 𝑓1 ) = 1.
Proof:
Since Ω is non-degenerate, for any nonzero vector 𝑒1 ∈ 𝑉 , there exists a vector 𝑓1 ∈ 𝑉
such that Ω(𝑒1 , 𝑓1 ) ≠ 0. Without loss of generality, we can assume that Ω(𝑒1 , 𝑓1 ) = 1 by
scaling 𝑓1 appropriately.
Therefore, for any nonzero vector 𝑒1 ∈ 𝑉 , there exists a vector 𝑓1 ∈ 𝑉 such that Ω(𝑒1 , 𝑓1 ) =
1.

2
2. Show that the restriction of the bilinear form Ω to the subspace generated by 𝑒1 and 𝑓1 is
non-degenerate.
Proof:
Let 𝑈 be the subspace generated by 𝑒1 and 𝑓1 , i.e., 𝑈 = span{𝑒1 , 𝑓1 }.
Suppose, for the sake of contradiction, that the restriction of Ω to 𝑈 is degenerate. Then,
there exists a nonzero vector 𝑣 ∈ 𝑈 such that Ω(𝑢, 𝑣) = 0 for all 𝑢 ∈ 𝑈 .
Since 𝑣 is in the span of {𝑒1 , 𝑓1 }, we can write 𝑣 = 𝛼𝑒1 + 𝛽𝑓1 for some scalars 𝛼 and 𝛽.
Evaluating Ω(𝑢, 𝑣) = 0 for 𝑢 = 𝑒1 and 𝑢 = 𝑓1 , we get

Ω(𝑒1 , 𝑣) = 𝛼Ω(𝑒1 , 𝑒1 ) + 𝛽Ω(𝑒1 , 𝑓1 ) = 𝛼,

Ω(𝑓1 , 𝑣) = 𝛼Ω(𝑓1 , 𝑒1 ) + 𝛽Ω(𝑓1 , 𝑓1 ) = 𝛽.

Since Ω(𝑒1 , 𝑣) = 0 and Ω(𝑓1 , 𝑣) = 0, we have 𝛼 = 0 and 𝛽 = 0, which implies that 𝑣 = 0.

But this contradicts the assumption that 𝑣 is nonzero.
Therefore, the restriction of Ω to the subspace generated by 𝑒1 and 𝑓1 is non-degenerate.

Problem 3.
Let 𝑆 2 be the unit two-sphere in ℝ3 , that is,

𝑆 2 = {(𝑥, 𝑦, 𝑧) ∈ ℝ3 ∣ 𝑥 2 + 𝑦 2 + 𝑧 2 = 1}.

(a) Is the following two-form restricted to 𝑆 2 𝜔 = 𝑥 𝑑𝑦 ∧ 𝑑𝑧 − 𝑦 𝑑𝑥 ∧ 𝑑𝑦 + 𝑧 𝑑𝑥 ∧ 𝑑𝑦

symplectic?
(b) Is the following two-form restricted to 𝑆 2 𝜔 = 𝑥 𝑑𝑦 ∧ 𝑑𝑧 − 𝑦 𝑑𝑥 ∧ 𝑑𝑧 + 𝑧 𝑑𝑥 ∧ 𝑑𝑦
symplectic?
Solution. Let’s check the first two-form:

𝑤 = 𝑥𝑑𝑦 ∧ 𝑑𝑧 − 𝑦𝑑𝑥 ∧ 𝑑𝑦 + 𝑧𝑑𝑥 ∧ 𝑑𝑦

To check if it is non-degenerate, we can calculate the determinant of the matrix of

coefficients:
|0 𝑧 −𝑦 ||
|
|−𝑧 0 𝑥 ||
|
| 𝑦 −𝑥 0 |
| |
We can use the first column as the basis for the exterior product, and we get:

𝑤 ∧ (𝑥𝑑𝑧 ∧ 𝑑𝑦 + 𝑦𝑑𝑥 ∧ 𝑑𝑧 + 𝑧𝑑𝑦 ∧ 𝑑𝑥) = 2𝑥𝑦𝑧(𝑑𝑥 ∧ 𝑑𝑦 ∧ 𝑑𝑧)

So 𝑤 is non-degenerate on 𝑆 2 . Next, let’s check if 𝑤 is closed. We can calculate the

exterior derivative of 𝑤:

𝑑𝑤 = (𝑑𝑥 ∧ 𝑑𝑦 ∧ 𝑑𝑧) − (𝑑𝑦 ∧ 𝑑𝑥 ∧ 𝑑𝑦) − (𝑑𝑥 ∧ 𝑑𝑥 ∧ 𝑑𝑧) = −2𝑑𝑥 ∧ 𝑑𝑥 ∧ 𝑑𝑧

Since 𝑑𝑤 is not zero, 𝑤 is not closed and therefore not symplectic.

3
Problem 4.
Let (𝑀, 𝜔) be a 2𝑛-dimensional symplectic manifold. Let 𝜔𝑛 be the volume form obtained
by wedging the symplectic form 𝜔 with itself 𝑛 times. Show that if 𝑀 is compact without
boundary, then the de Rham cohomology class [𝜔𝑛 ] ∈ 𝐻dR 2𝑛
(𝑀; ℝ) is non-zero.

Solution. To show that the de Rham cohomology class [𝑤𝑛 ] is non-zero, we can use the
following argument: Assume, for the sake of contradiction, that [𝑤𝑛 ] = 0 in the de Rham
cohomology group 𝐻𝑑𝑅2𝑛
(𝑀; 𝑅). This means that there exists a smooth (2𝑛 − 1)-form 𝛼 on
𝑀 such that 𝑑(𝛼) = 𝑤𝑛 . By Stokes’ theorem, we have:

𝑛
∫ 𝑤 = ∫ 𝑑(𝛼) = ∫ 𝛼,
𝑀 𝑀 𝜕𝑀

where ∫𝑀 denotes the integral over the compact manifold 𝑀 and ∫𝜕𝑀 denotes the integral
over its boundary. Since 𝑀 is assumed to be without boundary, the last term on the
right-hand side vanishes, so we have:

𝑛
∫ 𝑤 = 0.
𝑀

However, this contradicts the fact that 𝑤𝑛 is a volume form, i.e., a non-vanishing top-
dimensional form on 𝑀. Therefore, we must conclude that [𝑤𝑛 ] is non-zero in the de
Rham cohomology group 𝐻𝑑𝑅 2𝑛
(𝑀; 𝑅).

Problem 5.

(1). Let 𝜋 ∶ 𝐸 → 𝐵 be a real vector bundle of rank one, that is, each fiber is a one-
dimensional real vector space. If there exists a nowhere-vanishing section 𝑠 of 𝜋,
then prove that 𝜋 is trivial, namely, it is isomorphic to the product bundle 𝐵 ×ℝ → 𝐵.
(2). Let (𝑀, 𝜔) be a 2𝑛-dimensional symplectic manifold. Explain why the top exterior
power ⋀2𝑛 𝑇 ∗ 𝑀 is a trivial bundle over 𝑀.
Solution.
(1). To show that 𝜋 is isomorphic to the product bundle 𝐵 × ℝ → 𝐵, we need to construct
a bundle isomorphism between them. Let’s define such a map: Consider the map
𝐹 ∶ 𝐸 → 𝐵×ℝ that sends each point 𝑒 in 𝐸 to the pair (𝜋(𝑒), 𝑠(𝑒)), where 𝑠 is the given
nowhere-vanishing section of 𝜋. We claim that 𝐹 is a bundle isomorphism. First, we
need to show that 𝐹 is a well-defined map, that is, it sends points in the same fiber of
𝜋 to points in the same fiber of 𝐵 × ℝ. Let 𝑒1 and 𝑒2 be two points in the same fiber of
𝜋, i.e., they have the same base point in 𝐵 and belong to the same one-dimensional
vector space over this base point. Since 𝑠 is nowhere-vanishing, it follows that 𝑠(𝑒1 )
and 𝑠(𝑒2 ) are nonzero multiples of each other, say 𝑠(𝑒1 ) = 𝑐𝑠(𝑒2 ) for some nonzero
constant 𝑐 in ℝ. Then, 𝐹 (𝑒1 ) = (𝜋(𝑒1 ), 𝑠(𝑒1 )) = (𝜋(𝑒2 ), 𝑐𝑠(𝑒2 )) = (𝜋(𝑒2 ), 𝑠(𝑒2 )), which
shows that 𝐹 sends points in the same fiber of 𝜋 to points in the same fiber of
𝐵 × ℝ. Next, we need to show that 𝐹 is a homeomorphism, i.e., it is a continu-
ous bijection with a continuous inverse. Since 𝜋 and 𝑠 are both continuous, it
follows that 𝐹 is continuous as the product of continuous maps. Moreover, 𝐹 is

4
 injective because if 𝐹 (𝑒1 ) = 𝐹 (𝑒2 ), then (𝜋(𝑒1 ), 𝑠(𝑒1 )) = (𝜋(𝑒2 ), 𝑠(𝑒2 )), which im-
plies that 𝜋(𝑒1 ) = 𝜋(𝑒2 ) and 𝑠(𝑒1 ) = 𝑠(𝑒2 ). Since 𝑠 is nowhere-vanishing, we must
have 𝑒1 = 𝑒2 . Finally, 𝐹 is surjective because given any point (𝑏, 𝑟) in 𝐵 × ℝ, we
can find a point 𝑒 in 𝐸 with 𝜋(𝑒) = 𝑏 and 𝑠(𝑒) = 𝑟 by choosing any nonzero
vector in the one-dimensional fiber over 𝑏 and scaling it by 𝑟. It remains to
show that 𝐹 preserves the bundle structure, i.e., it commutes with the projec-
tion maps 𝜋𝐸 and 𝜋𝐵 . For any point 𝑒 in 𝐸, we have 𝜋𝐵 (𝐹 (𝑒)) = 𝜋(𝜋(𝑒), 𝑠(𝑒)) = 𝜋(𝑒),
which shows that 𝜋𝐵 ◦ 𝐹 = 𝜋𝐸 . Moreover, for any point (𝑏, 𝑟) in 𝐵 × ℝ, we have
𝜋𝐸 (𝐹 −1 (𝑏, 𝑟)) = 𝜋𝐸 ({𝑒 ∈ 𝐸 | 𝜋(𝑒) = 𝑏 and 𝑠(𝑒) = 𝑟}) = {𝑣 ∈ ℝ | (𝑏, 𝑣) ∈ 𝐵 × ℝ}, which
shows that 𝜋𝐸 ◦ 𝐹 −1 is the identity map on 𝐵 × ℝ. Therefore, 𝐹 is a bundle isomor-
phism between 𝜋 and 𝐵 × ℝ, which implies that 𝜋 is trivial.

(2). By definition, the top exterior power of the cotangent bundle of 𝑀 is the bundle
whose fiber at each point 𝑝 in 𝑀 is the top exterior power of the cotangent space
𝑇𝑝⋅ 𝑀, i.e., the space of all antisymmetric 𝑛-forms on 𝑇𝑝⋅ 𝑀. Since 𝑀 is symplectic,
it has a non-degenerate 2-form 𝑤, which means that 𝑤𝑛 is a nowhere-vanishing
top-degree form on 𝑀. Therefore, we can define a global section 𝑠 of the top exterior
power bundle by assigning to each point 𝑝 in 𝑀 the element 𝑤𝑝𝑛 in the fiber over 𝑝.
Now, let us show that this section is actually a non-zero section, i.e., that it does
not vanish anywhere. Suppose that there exists a point 𝑝 in 𝑀 such that 𝑠(𝑝) = 0.
Then, the element 𝑤𝑝𝑛 in the fiber over 𝑝 must be the zero element, which implies
that 𝑤𝑝 is degenerate. But this contradicts the assumption that 𝑀 is symplectic and,
therefore, non-degenerate.
Since the section 𝑠 is non-zero, it follows that the top exterior power bundle is a
trivial bundle over 𝑀, i.e., isomorphic to the product bundle 𝑀 × ℝ. This is because
any two non-zero sections of a trivial bundle are homotopic, and hence, can be
continuously deformed into each other.

Problem 6.
Let M be a smooth manifold. (You may assume that M is equipped with a symplectic
form 𝜔.) Show that the tangent bundle TM and cotangent bundle 𝑇 ∗ 𝑀 are isomorphic as
smooth manifolds (or more generally as vector bundles).

Solution. Consider the cotangent bundle 𝑇 ⋅ 𝑀, whose typical fiber is the cotangent space
𝑇𝑝∗ 𝑀 at a point 𝑝 ∈ 𝑀. Each cotangent space 𝑇𝑝∗ 𝑀 is the dual space of the tangent space
𝑇𝑝 𝑀, and the symplectic form 𝜔 gives a natural pairing between these spaces, denoted
by ⟨⋅, ⋅⟩𝜔 ∶ 𝑇𝑝∗ 𝑀 × 𝑇𝑝 𝑀 → ℝ.
Let’s define a smooth map Φ ∶ 𝑇 𝑀 → 𝑇 ⋅ 𝑀 as follows: for each point 𝑝 ∈ 𝑀 and tangent
vector 𝑣 ∈ 𝑇𝑝 𝑀, we define Φ(𝑝, 𝑣) ∈ 𝑇𝑝∗ 𝑀 by Φ(𝑝, 𝑣)(𝑤) = ⟨𝑣, 𝑤⟩𝜔 , where 𝑤 ∈ 𝑇𝑝 𝑀.
It’s clear that Φ is a well-defined smooth map. To show that Φ is a bundle isomorphism,
we need to show that it is a smooth bijection and that it preserves the bundle structure.
Injectivity: Suppose Φ(𝑝, 𝑣) = Φ(𝑞, 𝑤), where 𝑝, 𝑞 ∈ 𝑀 and 𝑣, 𝑤 ∈ 𝑇𝑝 𝑀. Then, for any
tangent vector 𝑢 ∈ 𝑇𝑝 𝑀, we have
⟨𝑣, 𝑢⟩𝜔 = Φ(𝑝, 𝑣)(𝑢) = Φ(𝑞, 𝑤)(𝑢) = ⟨𝑤, 𝑢⟩𝜔 .

5
Since this holds for all 𝑢 ∈ 𝑇𝑝 𝑀, we must have 𝑣 = 𝑤. Therefore, (𝑝, 𝑣) = (𝑞, 𝑤), and Φ is
injective.
Surjectivity: Given any 𝛼 ∈ 𝑇𝑝∗ 𝑀, we can define a tangent vector 𝑣 ∈ 𝑇𝑝 𝑀 as follows:
for any tangent vector 𝑢 ∈ 𝑇𝑝 𝑀, set ⟨𝑣, 𝑢⟩𝜔 = 𝛼(𝑢). This defines a tangent vector 𝑣 that
corresponds to 𝛼 under Φ. Therefore, Φ is surjective.
Smoothness: To show that Φ is smooth, it suffices to show that each component of Φ
is smooth with respect to local charts on 𝑇 𝑀 and 𝑇 ∗ 𝑀. By using local coordinates and
expressions for 𝜔 in terms of the coordinates, it can be shown that Φ is smooth.
Bundle Structure: It can be verified that Φ preserves the bundle structure, i.e., it com-
mutes with the projection maps 𝜋𝑇 𝑀 ∶ 𝑇 𝑀 → 𝑀 and 𝜋𝑇 ∗ 𝑀 ∶ 𝑇 ∗ 𝑀 → 𝑀. For any point
(𝑝, 𝑣) ∈ 𝑇 𝑀, we have 𝜋𝑇 ∗ 𝑀 (Φ(𝑝, 𝑣)) = 𝑝 since Φ(𝑝, 𝑣) = 𝛼 ∈ 𝑇𝑝∗ 𝑀. Similarly, for any point
𝛼 ∈ 𝑇𝑝∗ 𝑀, we have 𝜋𝑇 𝑀 (Φ−1 (𝛼)) = 𝑝.
Since Φ is a smooth bijection that preserves the bundle structure, it is a bundle isomor-
phism between 𝑇 𝑀 and 𝑇 ⋅ 𝑀.
Therefore, the tangent bundle 𝑇 𝑀 and cotangent bundle 𝑇 ∗ 𝑀 are isomorphic as smooth
manifolds (or more generally, as vector bundles).

Problem 7.

(1). Show that the symplectic form 𝜔 on a compact symplectic manifold M without
boundary cannot be exact, that is, there does not exist a differential 1-form 𝛼 such
that 𝑑𝛼 = 𝜔.
(2). Explain why if 𝑛 ≥ 3 there are no symplectic structures on the sphere 𝑆 𝑛 .
Solution.

(1). One way to show that the symplectic form 𝑤 on a compact symplectic manifold 𝑀
without boundary cannot be exact is by using Stokes’ theorem. Assume that there
exists a differential 1-form 𝛼 such that 𝑑𝛼 = 𝑤. Then, by Stokes’ theorem, we have:

∫ 𝑤 = ∫ 𝑑𝛼 = ∫ 𝛼,
𝑀 𝑀 𝜕𝑀

where 𝜕𝑀 is the boundary of 𝑀. However, since 𝑀 is assumed to be without

boundary, 𝜕𝑀 = ∅, and hence ∫𝜕𝑀 𝛼 = 0. Therefore, we obtain:

∫ 𝑤 = 0.
𝑀

This implies that the integral of the symplectic form 𝑤 over the entire manifold
𝑀 is zero. However, this contradicts the nondegeneracy of the symplectic form,
which implies that 𝑤𝑛 (where 𝑛 is the dimension of 𝑀) is a volume form that gives
a positive volume to any nonempty open subset of 𝑀. Hence, ∫𝑀 𝑤𝑛 > 0 for any
nonempty 𝑀. Therefore, we have a contradiction, and thus the symplectic form 𝑤
on a compact symplectic manifold 𝑀 without boundary cannot be exact.

6
(2). A symplectic structure is a non-degenerate, closed 2-form on a smooth manifold.
The sphere 𝑆 𝑛 is a smooth manifold, but it does not admit a symplectic structure
for 𝑛 ≥ 3.
One way to see this is to use the fact that a symplectic structure on a compact
manifold must satisfy certain topological conditions. In particular, the symplectic
form must be integral, which means that its cohomology class in 𝐻 2 (𝑆 𝑛 ; ℤ) must
be an integral class. However, the cohomology group 𝐻 2 (𝑆 𝑛 ; ℤ) is isomorphic to ℤ
for 𝑛 odd and is trivial for 𝑛 even. In either case, there is no non-zero integral class
in 𝐻 2 (𝑆 𝑛 ; ℤ), which means that there can be no symplectic form on 𝑆 𝑛 .
Another way to see this is to use the fact that the sphere 𝑆 𝑛 is simply connected for
𝑛 ≥ 2. By the Darboux theorem, any symplectic structure on a simply connected
manifold must be exact, which means that it can be written as the exterior derivative
of a 1-form. However, it can be shown that any closed 1-form on 𝑆 𝑛 must be exact,
which means that there can be no non-trivial closed 2-form (i.e., a symplectic form)
on 𝑆 𝑛 .
Therefore, for 𝑛 ≥ 3, there is no symplectic structure on the sphere 𝑆 𝑛 .

Problem 8.
Consider the Euclidean space (𝑅2𝑛 ; 𝜔0 = ∑𝑗=1
𝑛
𝑑𝑥𝑗 ∧ 𝑑𝑦𝑗 ). Let 𝑔0 be the standard inner
product on R2n. Describe an almost complex structure J0 such that

𝜔0(𝑢; 𝑣) = 𝑔0 (𝐽0 𝑢; 𝑣)

:
Solution.
An almost complex structure on a real vector space 𝑉 is a linear transformation 𝐽 ∶ 𝑉 → 𝑉
such that 𝐽 2 = −Id, where Id is the identity transformation on 𝑉 . In other words, 𝐽 is a
linear transformation that squares to −1. To construct an almost complex structure 𝐽0
on ℝ2𝑛 , we can define it as follows. Let (𝑒1 , … , 𝑒𝑛 , 𝑓1 , … , 𝑓𝑛 ) be the standard basis for ℝ2𝑛 ,
where each 𝑒𝑗 and 𝑓𝑗 are of dimension 𝑛. Then we define 𝐽0 on this basis by:

𝐽0 (𝑒𝑗 ) = 𝑓𝑗
𝐽0 (𝑓𝑗 ) = −𝑒𝑗
𝐽0 (𝑒𝑘 ) = 𝐽0 (𝑓𝑘 ) = 0 for 𝑘 ≠ 𝑗

We can extend 𝐽0 linearly to all of ℝ2𝑛 by defining 𝐽0 (𝑥 +𝑦) = 𝐽0 (𝑥)+𝐽0 (𝑦) for all 𝑥, 𝑦 ∈ ℝ2𝑛 .
We can now show that 𝑤0 (𝑢, 𝑣) = 𝑔0 (𝐽0 𝑢, 𝑣) for all 𝑢, 𝑣 ∈ ℝ2𝑛 :
𝑛
𝑤0 (𝑢, 𝑣) = ∑(𝑑𝑥𝑗 ∧ 𝑑𝑦𝑗 )(𝑢, 𝑣)
𝑗=1
𝑛
= ∑(𝑑𝑥𝑗 (𝑢)𝑑𝑦𝑗 (𝑣) − 𝑑𝑥𝑗 (𝑣)𝑑𝑦𝑗 (𝑢))
𝑗=1

= 𝑔0 (𝑢, 𝐽0 𝑣) − 𝑔0 (𝑣, 𝐽0 𝑢)
= 𝑔0 (𝐽0 𝑢, 𝑣)

7
Therefore, 𝐽0 is an almost complex structure on ℝ2𝑛 that satisfies the desired condition.

Problem 9.
1. Let (𝑀, 𝜔) be a complex manifold. Suppose that 𝜌0 and 𝜌1 are strictly plurisubharmonic
functions on M. Let 𝜔0 and 𝜔1 be Kahler forms arising from the potential 𝜌0 and 𝜌1 ,
respectively. For 𝑡 ∈ [0, 1] we set

𝜔𝑡 ≔ (1 − 𝑡)𝜔0 + 𝑡𝜔1 .
Is each 𝜔𝑡 symplectic? 2. Let M be a smooth manifold. Suppose that 𝜔0 and 𝜔1 are
symplectic forms on M. For 𝑡 ∈ [0, 1] we set

𝜔𝑡 ≔ (1 − 𝑡)𝜔0 + 𝑡𝜔1 .
Is each 𝜔𝑡 symplectic?
Solution.
Let (𝑀, 𝜔) be a complex manifold. Suppose that 𝜌0 and 𝜌1 are strictly plurisubharmonic
functions on 𝑀. Let 𝜔0 and 𝜔1 be Kähler forms arising from the potentials 𝜌0 and 𝜌1 ,
respectively. For 𝑡 ∈ [0, 1], we set

𝜔𝑡 ≔ (1 − 𝑡)𝜔0 + 𝑡𝜔1 .

Is each 𝜔𝑡 symplectic?
Solution:
A symplectic form is a closed, non-degenerate 2-form. To check whether each 𝜔𝑡 is
symplectic, we need to examine both properties.
First, let’s check the closedness of 𝜔𝑡 :

𝑑𝜔𝑡 = 𝑑((1 − 𝑡)𝜔0 + 𝑡𝜔1 ) = (1 − 𝑡)𝑑𝜔0 + 𝑡𝑑𝜔1 .

Since 𝜔0 and 𝜔1 are Kähler forms, they are closed forms (i.e., 𝑑𝜔0 = 𝑑𝜔1 = 0). Therefore,
𝑑𝜔𝑡 = 0 for all 𝑡 ∈ [0, 1], which means that each 𝜔𝑡 is closed.
Next, let’s check the non-degeneracy of 𝜔𝑡 . The non-degeneracy of a 2-form 𝜔 means that
for any non-zero tangent vector 𝑣 at a point 𝑝 in 𝑀, there exists a tangent vector 𝑤 at 𝑝
such that 𝜔(𝑣, 𝑤) ≠ 0. Since 𝜔0 and 𝜔1 are Kähler forms, they are non-degenerate. Hence,
for any 𝑡 ∈ [0, 1] and any point 𝑝 ∈ 𝑀, there exist tangent vectors 𝑣 and 𝑤 at 𝑝 such that
𝜔𝑡 (𝑣, 𝑤) = (1 − 𝑡)𝜔0 (𝑣, 𝑤) + 𝑡𝜔1 (𝑣, 𝑤) ≠ 0. Therefore, each 𝜔𝑡 is also non-degenerate.
Since each 𝜔𝑡 is both closed and non-degenerate, it is indeed symplectic.
2. Let 𝑀 be a smooth manifold. Suppose that 𝜔0 and 𝜔1 are symplectic forms on 𝑀. For
𝑡 ∈ [0, 1], we set
𝜔𝑡 ≔ (1 − 𝑡)𝜔0 + 𝑡𝜔1 .
Is each 𝜔𝑡 symplectic?
Solution:
Similar to the previous part, let’s check the closedness and non-degeneracy of 𝜔𝑡 .

8
First, the closedness of 𝜔𝑡 :

𝑑𝜔𝑡 = 𝑑((1 − 𝑡)𝜔0 + 𝑡𝜔1 ) = (1 − 𝑡)𝑑𝜔0 + 𝑡𝑑𝜔1 .

Since 𝜔0 and 𝜔1 are symplectic forms, they are closed (i.e., 𝑑𝜔0 = 𝑑𝜔1 = 0). Therefore,
𝑑𝜔𝑡 = 0 for all 𝑡 ∈ [0, 1], and each 𝜔𝑡 is closed.
Next, the non-degeneracy of 𝜔𝑡 . Similar to before, for any 𝑡 ∈ [0, 1] and any point 𝑝 ∈ 𝑀,
there exist tangent vectors 𝑣 and 𝑤 at 𝑝 such that 𝜔𝑡 (𝑣, 𝑤) = (1−𝑡)𝜔0 (𝑣, 𝑤)+𝑡𝜔1 (𝑣, 𝑤) ≠ 0.
Therefore, each 𝜔𝑡 is non-degenerate.
Since each 𝜔𝑡 is both closed and non-degenerate, it is symplectic.

Problem 10.
⎧
⎪
⎪ 𝐺𝐿(2, ℝ) = {𝐴 ∈ 𝑀2×2 (ℝ) ∣ det(𝐴) ≠ 0};
⎪
⎨ 𝑆𝐿(2, ℝ) = {𝐴 ∈ 𝑀2×2 (ℝ) ∣ det(𝐴) = 1}; (0.1)
⎪
⎪ ∗
⎪
⎩ℝ = ℝ ⧵ {0}.
1. Show that 𝐺𝐿(2, ℝ) and 𝑆𝐿(2, ℝ) × ℝ∗ are diffeomorphic as smooth manifolds.
2. Are they isomorphic as Lie groups?

Solution

1. Show that 𝐺𝐿(2, ℝ) and 𝑆𝐿(2, ℝ) × ℝ∗ are diffeomorphic as smooth manifolds.

Solution:
Let’s define a map 𝜙 ∶ 𝐺𝐿(2, ℝ) → 𝑆𝐿(2, ℝ) × ℝ∗ as follows:
1
𝜙(𝐴) = ( 𝐴, det(𝐴)).
det(𝐴)

We need to show that 𝜙 is a diffeomorphism.

First, let’s check that 𝜙 is well-defined. Since det(𝐴) ≠ 0, the inverse exists and is given
by det(𝐴)
1
𝐴. Also, det(𝐴) ∈ ℝ∗ , and hence the second component is well-defined.
Next, let’s show that 𝜙 is smooth. The entries of the matrix 𝐴 and the determinant
function are all smooth, so each component of 𝜙 is smooth. Thus, 𝜙 is a smooth map.
Now, let’s show that 𝜙 is a bijection. The inverse of 𝜙 is given by:

𝜙−1 ((𝐵, 𝑐)) = 𝑐𝐵.

This is well-defined since 𝑐 ∈ ℝ∗ and 𝐵 ∈ 𝑆𝐿(2, ℝ), and the product of a non-singular
matrix with a non-zero scalar is a non-singular matrix.
Finally, let’s show that both 𝜙 and 𝜙−1 are smooth. The inverse is a linear map, and the
entries of 𝐵 and 𝑐 are all smooth functions. Thus, 𝜙−1 is smooth.
Since 𝜙 is a bijection and both 𝜙 and 𝜙−1 are smooth, 𝜙 is a diffeomorphism. Therefore,
𝐺𝐿(2, ℝ) and 𝑆𝐿(2, ℝ) × ℝ∗ are diffeomorphic as smooth manifolds.

9
2. Are they isomorphic as Lie groups?
No, they are not isomorphic as Lie groups. The group operations on 𝐺𝐿(2, ℝ) and
𝑆𝐿(2, ℝ) × ℝ∗ are different. In 𝐺𝐿(2, ℝ), the group operation is matrix multiplication with
non-zero determinant, while in 𝑆𝐿(2, ℝ) × ℝ∗ , the group operation is component-wise
multiplication and matrix multiplication with determinant 1. Therefore, the Lie groups
𝐺𝐿(2, ℝ) and 𝑆𝐿(2, ℝ) × ℝ∗ are not isomorphic.

Problem 11.
Describe the Lie algebra of G. 1. G = SO(n) = fA 2 Mn×n(R) j AT A = AAT = Ig: 2. G =
SL(n) = fA 2 Mn×n(R) j det(A) = 1g:

Solution.

1. Show that 𝐺𝐿(2, ℝ) and 𝑆𝐿(2, ℝ) × ℝ∗ are diffeomorphic as smooth manifolds.

Solution:
Let’s define a map 𝜙 ∶ 𝐺𝐿(2, ℝ) → 𝑆𝐿(2, ℝ) × ℝ∗ as follows:
1
𝜙(𝐴) = ( 𝐴, det(𝐴)).
det(𝐴)

We need to show that 𝜙 is a diffeomorphism.

First, let’s check that 𝜙 is well-defined. Since det(𝐴) ≠ 0, the inverse exists and is given
by det(𝐴)
1
𝐴. Also, det(𝐴) ∈ ℝ∗ , and hence the second component is well-defined.
Next, let’s show that 𝜙 is smooth. The entries of the matrix 𝐴 and the determinant
function are all smooth, so each component of 𝜙 is smooth. Thus, 𝜙 is a smooth map.
Now, let’s show that 𝜙 is a bijection. The inverse of 𝜙 is given by:

𝜙−1 ((𝐵, 𝑐)) = 𝑐𝐵.

This is well-defined since 𝑐 ∈ ℝ∗ and 𝐵 ∈ 𝑆𝐿(2, ℝ), and the product of a non-singular
matrix with a non-zero scalar is a non-singular matrix.
Finally, let’s show that both 𝜙 and 𝜙−1 are smooth. The inverse is a linear map, and the
entries of 𝐵 and 𝑐 are all smooth functions. Thus, 𝜙−1 is smooth.
Since 𝜙 is a bijection and both 𝜙 and 𝜙−1 are smooth, 𝜙 is a diffeomorphism. Therefore,
𝐺𝐿(2, ℝ) and 𝑆𝐿(2, ℝ) × ℝ∗ are diffeomorphic as smooth manifolds.
2. Are they isomorphic as Lie groups?
No, they are not isomorphic as Lie groups. The group operations on 𝐺𝐿(2, ℝ) and
𝑆𝐿(2, ℝ) × ℝ∗ are different. In 𝐺𝐿(2, ℝ), the group operation is matrix multiplication with
non-zero determinant, while in 𝑆𝐿(2, ℝ) × ℝ∗ , the group operation is component-wise
multiplication and matrix multiplication with determinant 1. Therefore, the Lie groups
𝐺𝐿(2, ℝ) and 𝑆𝐿(2, ℝ) × ℝ∗ are not isomorphic.

10
Problem 12.
Let (𝑀, 𝜔) be a symplectic manifold. We want to show that the following two statements
on a vector field 𝑋 on 𝑀 are equivalent:
1. The vector field 𝑋 on 𝑀 preserves 𝜔, that is, 𝐿𝑋 𝜔 = 0.
2. The interior product 𝜄𝑋 𝜔 is closed, that is, 𝑑(𝜄𝑋 𝜔) = 0.
We will prove the equivalence of these statements.
Solution.
(1 ⇒ 2) Suppose 𝑋 preserves 𝜔, i.e., 𝐿𝑋 𝜔 = 0. We want to show that 𝜄𝑋 𝜔 is closed.
Using Cartan’s formula, we have:
𝐿𝑋 𝜔 = 𝑑(𝜄𝑋 𝜔) + 𝜄𝑋 (𝑑𝜔).
Since 𝜔 is closed (i.e., 𝑑𝜔 = 0), this simplifies to:
0 = 𝑑(𝜄𝑋 𝜔).
This shows that 𝜄𝑋 𝜔 is closed.
(2 ⇒ 1) Suppose 𝜄𝑋 𝜔 is closed, i.e., 𝑑(𝜄𝑋 𝜔) = 0. We want to show that 𝑋 preserves 𝜔.
Again, using Cartan’s formula, we have:
𝐿𝑋 𝜔 = 𝑑(𝜄𝑋 𝜔) + 𝜄𝑋 (𝑑𝜔).
Since 𝜄𝑋 𝜔 is closed (i.e., 𝑑(𝜄𝑋 𝜔) = 0) and 𝜔 is symplectic (i.e., 𝑑𝜔 = 0), the equation
becomes:
𝐿𝑋 𝜔 = 0.
This shows that 𝑋 preserves 𝜔.
Therefore, we have shown the equivalence of the two statements.

Problem 13.
Let (𝑀, 𝜔) be a symplectic manifold. The Poisson bracket of two smooth functions 𝑓 , 𝑔
on M is defined by 𝑓 , 𝑔 ≔ 𝜔(𝑋𝑓 , 𝑋𝑔 ) where 𝑋𝑓 (𝑟𝑒𝑠𝑝.𝑋𝑔 ) is a Hamiltonian vector field
generated by f (resp. g). Prove the Jacobi identity, that is, derive
{𝑓 , {𝑔, ℎ}} + {𝑔, {ℎ, 𝑓 }} + {ℎ, {𝑓 , 𝑔}} = 0.

Solution.
Let’s compute each term in the Jacobi identity one by one.
Starting with the left term:
{𝑓 , {𝑔, ℎ}} = {𝑓 , 𝜔(𝑋𝑔 , 𝑋ℎ )}
= 𝜔(𝑋𝑓 , 𝑋𝜔(𝑋𝑔 ,𝑋ℎ ) ).

Now, using the Lie derivative property of the symplectic form 𝜔 along a vector field 𝑋𝑔 ,
we have:
𝑋𝜔(𝑋𝑔 ,𝑋ℎ ) = 𝑋𝑔 (𝜔(𝑋ℎ )) − 𝜔([𝑋𝑔 , 𝑋ℎ ])
= 𝑋𝑔 (𝑑𝜔(𝑋ℎ )) − 𝜔(𝑋[𝑋𝑔 ,𝑋ℎ ] )
= 𝑑(𝑋𝑔 (𝜔(𝑋ℎ ))) − 𝜔(𝑋[𝑋𝑔 ,𝑋ℎ ] ),

11
where 𝑋𝑔 denotes the Lie derivative along the vector field 𝑋𝑔 .
Next, we expand the Lie derivative term:

𝑋𝑔 (𝜔(𝑋ℎ )) = 𝑋𝑔 (𝜔(𝑋ℎ )) − 𝜔([𝑋𝑔 , 𝑋ℎ ])

= 𝑋𝑔 (𝜔(𝑋ℎ )) − 𝜔(𝑋[𝑋𝑔 ,𝑋ℎ ] )
= 𝑋𝑔 (𝑑𝜔(𝑋ℎ )) − 𝜔(𝑋[𝑋𝑔 ,𝑋ℎ ] )
= 𝑑(𝑋𝑔 (𝜔(𝑋ℎ ))) − 𝜔(𝑋[𝑋𝑔 ,𝑋ℎ ] )
= 𝑑(𝑋𝑔 (𝜔)(𝑋ℎ )) − 𝜔(𝑋[𝑋𝑔 ,𝑋ℎ ] ),

where we used the properties of the Lie derivative and the exterior derivative.
Now, substituting this back into the previous expression:

𝑋𝜔(𝑋𝑔 ,𝑋ℎ ) = 𝑑(𝑑(𝑋𝑔 (𝜔)(𝑋ℎ ))) − 𝜔(𝑋[𝑋𝑔 ,𝑋ℎ ] ).

Finally, we plug this into the expression for {𝑓 , {𝑔, ℎ}}:

{𝑓 , {𝑔, ℎ}} = 𝜔(𝑋𝑓 , 𝑑(𝑑(𝑋𝑔 (𝜔)(𝑋ℎ ))) − 𝜔(𝑋[𝑋𝑔 ,𝑋ℎ ] )

= 𝑑(𝜔(𝑋𝑓 , 𝑑(𝑋𝑔 (𝜔)(𝑋ℎ )))) − 𝜔(𝑋𝑓 , 𝑋[𝑋𝑔 ,𝑋ℎ ] ).

Problem 14.
Prove that if 𝑋 and 𝑌 are symplectic vector fields on a symplectic manifold (𝑀, 𝜔), then
the Lie bracket [𝑋 , 𝑌 ] is a Hamiltonian vector field.
Proof:
Recall that a vector field 𝑍 on a symplectic manifold (𝑀, 𝜔) is said to be a Hamiltonian
vector field if there exists a smooth function 𝑓 on 𝑀 such that 𝑍 = 𝜔−1 (𝑑𝑓 ), where 𝜔−1
is the inverse of the symplectic form 𝜔.
We want to show that the Lie bracket [𝑋 , 𝑌 ] is a Hamiltonian vector field, that is, we
need to find a function 𝑓 such that [𝑋 , 𝑌 ] = 𝜔−1 (𝑑𝑓 ).
Using Cartan’s formula for the Lie derivative, we have:

𝐿𝑋 𝜔 = 𝑑(𝜄𝑋 𝜔) + 𝜄𝑋 𝑑𝜔 = 𝑑(𝜄𝑋 𝜔).

Similarly,
𝐿𝑌 𝜔 = 𝑑(𝜄𝑌 𝜔).

Since 𝑋 and 𝑌 are symplectic vector fields, we know that 𝜄𝑋 𝜔 and 𝜄𝑌 𝜔 are closed 1-
forms. Therefore, there exist smooth functions 𝑓𝑋 and 𝑓𝑌 on 𝑀 such that 𝜄𝑋 𝜔 = 𝑑𝑓𝑋 and
𝜄𝑌 𝜔 = 𝑑𝑓𝑌 .
Now, consider the Lie bracket [𝑋 , 𝑌 ]. By properties of the Lie derivative, we have:

𝐿[𝑋 ,𝑌 ] 𝜔 = [𝐿𝑋 , 𝐿𝑌 ]𝜔 = 𝐿𝑋 (𝐿𝑌 𝜔) − 𝐿𝑌 (𝐿𝑋 𝜔) = 𝐿𝑋 (𝑑(𝜄𝑌 𝜔)) − 𝐿𝑌 (𝑑(𝜄𝑋 𝜔)).

12
Using Cartan’s formula again, this simplifies to:
𝐿[𝑋 ,𝑌 ] 𝜔 = 𝑑(𝐿𝑋 (𝜄𝑌 𝜔)) − 𝑑(𝐿𝑌 (𝜄𝑋 𝜔)).

Since 𝑋 and 𝑌 are symplectic vector fields, we know that 𝐿𝑋 (𝜄𝑌 𝜔) and 𝐿𝑌 (𝜄𝑋 𝜔) are
closed 2-forms. Therefore, there exist smooth functions 𝑓𝑋 𝑌 and 𝑓𝑌 𝑋 on 𝑀 such that
𝐿𝑋 (𝜄𝑌 𝜔) = 𝑑𝑓𝑋 𝑌 and 𝐿𝑌 (𝜄𝑋 𝜔) = 𝑑𝑓𝑌 𝑋 .
Now, let’s consider the function 𝑓 = 𝑓𝑋 + 𝑓𝑋 𝑌 − 𝑓𝑌 − 𝑓𝑌 𝑋 . We claim that [𝑋 , 𝑌 ] = 𝜔−1 (𝑑𝑓 ).
Calculating the action of the 1-form 𝜔 on [𝑋 , 𝑌 ], we have:
𝜔([𝑋 , 𝑌 ], ⋅) = 𝜄[𝑋 ,𝑌 ] 𝜔 = 𝜄𝑋 (𝜄𝑌 𝜔) − 𝜄𝑌 (𝜄𝑋 𝜔) = 𝑑𝑓𝑋 𝑌 − 𝑑𝑓𝑌 𝑋 = 𝑑𝑓 .

Therefore, we have shown that the Lie bracket [𝑋 , 𝑌 ] is a Hamiltonian vector field with
the corresponding Hamiltonian function 𝑓 . Thus, the proof is complete.

Problem 15. √
Consider the complex plane C with the standard symplectic form 𝜔0 = 2−1 𝑑𝑧 ∧ 𝑑 𝑧. ̄ Let
the circle 𝑆1 = 𝑡 ∈ ℂ||𝑡| = 1 act on (ℂ, 𝜔0 ) by rotations 𝑡 ⋅ 𝑧 = 𝑡 𝑘 𝑧 for some fixed integer k.
Show that this S1-action is Hamiltonian and compute its moment map.
Solution.
Let’s find the moment map for this 𝑆 1 -action. The moment map 𝜇 ∶ ℂ → ℝ is defined
by ⟨𝑑𝜇, 𝑋 ⟩ = 𝜔0 (𝑋𝜇 , 𝑋 ) for all vector fields 𝑋 on
√ ℂ, where 𝑋𝜇 is the Hamiltonian vector
field generated by the function 𝜇. Let 𝑧 = 𝑥 + −1𝑦.
We have:
⟨𝑑𝜇, 𝑋 ⟩ = ∫ 𝜄𝑋𝜇 𝜔0
𝑋

= ∫ 𝜔0 (𝑋𝜇 , ⋅)
𝑋
√
−1
=∫ (𝑋𝜇 𝑑𝑧 ∧ 𝑑 𝑧̄ − 𝑋𝜇 𝑑 𝑧̄ ∧ 𝑑𝑧)
2
√𝑋
−1
= (𝑋𝜇 𝑑𝑧 ∧ 𝑑 𝑧̄ − (−𝑧̄ 𝑘 𝑋𝜇 )𝑑 𝑧̄ ∧ 𝑑𝑧)
√ 2
−1
= (𝑋𝜇 − (−𝑧̄ 𝑘 𝑋𝜇 ))𝑑𝑧 ∧ 𝑑 𝑧.̄
2
√
Since 𝜔0 = −1
2
𝑑𝑧 ̄ we must have:
∧ 𝑑 𝑧,
𝑑𝜇
𝑋𝜇 − (−𝑧̄ 𝑘 𝑋𝜇 ) = 2 √ .
−1

Solving for 𝑋𝜇 , we get:

√
−1 𝑘 𝑘 𝑑𝜇 1 + (−1)𝑘 𝑧̄ 𝑘
𝑋𝜇 = (1 + (−1) 𝑧
̄ ) √ = 𝑑𝜇.
2 −1 2

So, the moment map is given by:

1 + (−1)𝑘 𝑧̄ 𝑘
𝜇= .
2
13
This shows that the 𝑆 1 -action is Hamiltonian with moment map 𝜇.

Problem 16.
Let G be any compact Lie group and H a closed subgroup of G, with g and h the respective
Lie algebras. The projection 𝑖∗ ∶ 𝑔 ∗ → ℎ∗ is the map dual to the inclusion 𝑖 ∶ ℎ → 𝑔.
Suppose that (𝑀, 𝜔, 𝐺, 𝜇) is a Hamiltonian G-space. Show that the restriction of the
G-action to H is also Hamiltonian with moment map

𝑖∗ ◦ 𝜇 ∶ 𝑀 → 𝑔 ∗ → ℎ∗

Solution.
Consider the restriction of the 𝐺-action to 𝐻 . The action of 𝐻 on 𝑀 is still smooth, as it
is a closed subgroup of 𝐺. We need to show that this restricted action is Hamiltonian
with the moment map 𝑖⋅ ◦ 𝜇.
Let 𝑋 ∈ ℎ be an element of the Lie algebra of 𝐻 . Since 𝐻 is a subgroup of 𝐺, we can view
𝑋 as a left-invariant vector field on 𝑀. The Hamiltonian vector field corresponding to a
function 𝑓 is defined as 𝑋𝑓 = 𝜔♯ (𝑑𝑓 ), where 𝜔♯ ∶ 𝑇 𝑀 → 𝑇 ⋅ 𝑀 is the bundle isomorphism
induced by the symplectic form 𝜔.
Now, let 𝜙𝑡 ∶ 𝑀 → 𝑀 be the flow of 𝑋 on 𝑀. Since 𝑋 is left-invariant, the flow 𝜙𝑡
commutes with the 𝐻 -action, i.e., 𝜙𝑡 (ℎ ⋅ 𝑚) = ℎ ⋅ 𝜙𝑡 (𝑚) for all ℎ ∈ 𝐻 and 𝑚 ∈ 𝑀.
We want to show that the flow 𝜙𝑡 is generated by a moment map for the 𝐻 -action. To do
this, we need to find a function ℎ𝑀 ∶ 𝑀 → ℝ such that 𝑋ℎ𝑀 = 𝑋 .
Since 𝜙𝑡 commutes with the 𝐻 -action, for any ℎ ∈ 𝐻 and 𝑚 ∈ 𝑀, we have:

𝜙𝑡 (ℎ ⋅ 𝑚) = ℎ ⋅ 𝜙𝑡 (𝑚)
𝑒 ⋅ (ℎ ⋅ 𝑚) = ℎ ⋅ 𝑒 𝑡𝑋 ⋅ 𝑚
𝑡𝑋

ℎ ⋅ 𝑒 𝑡𝑋 ⋅ 𝑚 = ℎ ⋅ 𝜙𝑡 (𝑚),

where 𝑒 𝑡𝑋 is the flow of 𝑋 on 𝐻 . This implies that ℎ𝑀 (ℎ ⋅ 𝑚) = ℎ𝑀 (𝑚) for all ℎ ∈ 𝐻 .

Thus, we can define ℎ𝑀 ∶ 𝑀 → ℝ as ℎ𝑀 (𝑚) = 𝑐 for some constant 𝑐, since the value of
ℎ𝑀 is constant along the 𝐻 -orbits.
Now, we need to compute the Hamiltonian vector field 𝑋ℎ𝑀 associated with ℎ𝑀 . Using
the definition of the Hamiltonian vector field and the fact that 𝑑ℎ𝑀 = 0 (since ℎ𝑀 is
constant), we have:
𝑋ℎ𝑀 = 𝜔♯ (𝑑ℎ𝑀 )
= 0.

Therefore, 𝑋ℎ𝑀 = 𝑋 , and the flow 𝜙𝑡 generated by the function ℎ𝑀 coincides with
the flow of the left-invariant vector field 𝑋 on 𝑀. This shows that the restricted 𝐻 -
action is Hamiltonian with the moment map ℎ𝑀 , which corresponds to the composition
𝑖⋅ ◦ 𝜇 ∶ 𝑀 → 𝑔 ⋅ → ℎ⋅ .

Let 𝐺 be any compact Lie group and 𝐻 a closed subgroup of 𝐺, with 𝑔 and ℎ the respective
Lie algebras. The projection 𝑖⋅ ∶ 𝑔 ⋅ → ℎ⋅ is the map dual to the inclusion 𝑖 ∶ ℎ → 𝑔.

14
Suppose that (𝑀, 𝜔, 𝐺, 𝜇) is a Hamiltonian 𝐺-space. We want to show that the restriction
of the 𝐺-action to 𝐻 is also Hamiltonian with the moment map 𝑖⋅ ◦ 𝜇 ∶ 𝑀 → 𝑔 ⋅ → ℎ⋅ .
Solution:
Consider the action of 𝐺 on 𝑀 given by 𝜙 ∶ 𝐺 × 𝑀 → 𝑀, where 𝜙(𝑔, 𝑚) is the action of
𝑔 on 𝑚. This action is Hamiltonian with the moment map 𝜇 ∶ 𝑀 → 𝑔 ⋅ .
Now, let’s consider the restriction of this action to 𝐻 . For ℎ ∈ 𝐻 and 𝑚 ∈ 𝑀, let
𝜓 ∶ 𝐻 × 𝑀 → 𝑀 be the action of ℎ on 𝑚. Since 𝐻 is a closed subgroup of 𝐺, the
restriction of the action of 𝐺 to 𝐻 is well-defined.
We want to show that the restricted action of 𝐻 on 𝑀 is Hamiltonian with the moment
map 𝑖⋅ ◦ 𝜇 ∶ 𝑀 → ℎ⋅ .
First, let’s show that the restricted action of 𝐻 on 𝑀 is still smooth. Since 𝐻 is a closed
subgroup of 𝐺, the action of 𝐻 on 𝑀 is a smooth map.
Next, let 𝑋 ∈ ℎ be an element of the Lie algebra of 𝐻 , and let 𝑋𝑀 be the correspond-
ing infinitesimal generator of the restricted 𝐻 -action. We want to show that 𝑋𝑀 is a
Hamiltonian vector field.
Let 𝑓 ∈ 𝐶 ∞ (𝑀) be any smooth function. Then, the Hamiltonian vector field 𝑋𝑓 associated
with 𝑓 under the 𝐺-action satisfies 𝜄𝑋𝑀 𝜔 = −𝑑𝜇(𝑋𝑓 ). Now, we want to compute the
moment map of the restricted 𝐻 -action, denoted as 𝜈 ∶ 𝑀 → ℎ⋅ .
Using the definition of the moment map, for any 𝑋 ∈ ℎ and 𝑚 ∈ 𝑀, we have ⟨𝑑𝜈(𝑚), 𝑋 ⟩ =
𝜄𝑋𝑀 𝜔𝑚 .
Since 𝑋𝑀 is the infinitesimal generator of the restricted 𝐻 -action, we have 𝜄𝑋𝑀 𝜔𝑚 =
−𝑑𝜇𝑚 (𝑋𝑓 ), where 𝑑𝜇𝑚 ∶ 𝑔 → ℝ is the differential of the 𝐺-moment map at 𝑚.
Now, using the definition of the dual map, we can write 𝑑𝜇𝑚 (𝑋𝑓 ) = ⟨𝜇𝑚 , 𝑋 ⟩. Therefore,
we have:
⟨𝑑𝜈(𝑚), 𝑋 ⟩ = −⟨𝜇𝑚 , 𝑋 ⟩.

This shows that 𝑑𝜈(𝑚) = −𝜇𝑚 , which implies that 𝜈 = −𝑖⋅ ◦ 𝜇, where 𝑖 ∶ ℎ → 𝑔 is the
inclusion map.
Thus, the restricted 𝐻 -action on 𝑀 is Hamiltonian with the moment map 𝑖⋅ ◦ 𝜇 ∶ 𝑀 →
ℎ⋅ .

Problem 17.
Suppose that a compact Lie group 𝐺 acts on a symplectic manifold (𝑀, 𝜔) in a Hamilto-
nian fashion. Let 𝜇 be a moment map for this Hamiltonian 𝐺-action. Assume that 𝐺 acts
freely on 𝜇−1 (0).
Solution.
1. Show that the zero 0 is a regular value of 𝜇.
Proof:
Recall that a point 𝑝 in 𝑀 is a regular value of a smooth function 𝑓 ∶ 𝑀 → ℝ if the
differential 𝑑𝑓𝑝 ∶ 𝑇𝑝 𝑀 → ℝ is surjective.

15
Consider the moment map 𝜇 ∶ 𝑀 → 𝑔 ⋅ , where 𝑔 is the Lie algebra of 𝐺. The differential
𝑑𝜇𝑝 ∶ 𝑇𝑝 𝑀 → 𝑔 ⋅ is given by the action of 𝑔 on 𝑀 induced by the Hamiltonian action.
Since the action of 𝐺 is Hamiltonian and 𝐺 acts freely on 𝜇−1 (0), the differential 𝑑𝜇𝑝 is
surjective for all 𝑝 ∈ 𝜇−1 (0). This implies that 0 is a regular value of 𝜇.
2. Prove that the zero level set 𝜇−1 (0) is a coisotropic submanifold.
Proof:
A submanifold 𝑁 of a symplectic manifold (𝑀, 𝜔) is said to be coisotropic if for every
point 𝑝 in 𝑁 , the annihilator 𝑇𝑝⟂ of the tangent space 𝑇𝑝 𝑁 under 𝜔 is contained in 𝑇𝑝 𝑁 .
That is, 𝜔(𝑣, 𝑤) = 0 for all 𝑣 ∈ 𝑇𝑝 𝑁 and 𝑤 ∈ 𝑇𝑝⟂ .
Let 𝑝 be a point in 𝜇−1 (0). We want to show that 𝑇𝑝⟂ ⊆ 𝑇𝑝 (𝜇−1 (0)). Since 0 is a regular
value of 𝜇, the moment map 𝜇 is a submersion at every point of 𝜇−1 (0). Thus, 𝑇𝑝⟂ is the
kernel of 𝑑𝜇𝑝⋅ , the adjoint of 𝑑𝜇𝑝 .
Since 𝑑𝜇𝑝 is surjective (as shown in part 1), its adjoint 𝑑𝜇𝑝⋅ is injective. Therefore, 𝑇𝑝⟂ is
trivial, which implies that 𝑇𝑝⟂ ⊆ 𝑇𝑝 (𝜇−1 (0)). Thus, 𝜇−1 (0) is a coisotropic submanifold.
Therefore, we have shown that if a compact Lie group 𝐺 acts on a symplectic manifold
(𝑀, 𝜔) in a Hamiltonian fashion and 𝐺 acts freely on 𝜇−1 (0), then the zero level set 𝜇−1 (0)
is a coisotropic submanifold.

Problem 18.
Let (𝑀, 𝜔) be a symplectic manifold. A submanifold 𝑁 is called isotropic if the symplectic
form restricted to the tangent space of 𝑁 vanishes, that is, 𝜔|𝑇 𝑁 = 0.
Suppose that the symplectic manifold (𝑀, 𝜔) has a Hamiltonian 𝑇 𝑘 ∶= (𝑆 1 )𝑘 -action.
Show that every 𝑇 𝑘 -orbit is isotropic.
Solution.
Proof:
Let 𝑥 be a point in 𝑀, and let 𝐺𝑥 be the stabilizer subgroup of 𝑥 under the 𝑇 𝑘 -action.
That is, 𝐺𝑥 = {𝑔 ∈ 𝑇 𝑘 ∣ 𝑔 ⋅ 𝑥 = 𝑥}.
Consider the tangent space 𝑇𝑥 𝑀 at 𝑥. Since 𝑇 𝑘 is a compact Lie group, we can decompose
the tangent space 𝑇𝑥 𝑀 into its weight spaces with respect to the 𝑇 𝑘 -action. That is,
𝑘
𝑇𝑥 𝑀 = ⨁ 𝑉𝜆𝑖 ,
𝑖=1

where 𝑉𝜆𝑖 is the weight space corresponding to the weight 𝜆𝑖 of the 𝑖-th factor 𝑆 1 in 𝑇 𝑘 .
Now, let 𝑣 be a vector in the 𝑇 𝑘 -orbit through 𝑥. That is, 𝑣 = 𝑔 ⋅ 𝑥 for some 𝑔 ∈ 𝑇 𝑘 . Since
the action is Hamiltonian, we have a moment map 𝜇 ∶ 𝑀 → 𝑡 𝑘 where 𝑡 𝑘 is the Lie algebra
of 𝑇 𝑘 . The moment map satisfies 𝜄𝑣 𝜔 = 𝑑⟨𝜇, 𝑣⟩, where ⟨⋅, ⋅⟩ is the dual pairing between 𝑡 𝑘
and 𝑡 𝑘 .
Since 𝑣 is on the 𝑇 𝑘 -orbit through 𝑥, we have 𝜄𝑣 𝜔𝑥 = 0 because the symplectic form 𝜔
vanishes on the tangent space of the orbit. This implies that 𝑑⟨𝜇, 𝑣⟩𝑥 = 0.

16
Recall that the Lie algebra 𝑡 𝑘 of 𝑇 𝑘 can be identified with ℝ𝑘 using the exponential map.
The dual space of 𝑡 𝑘 is also ℝ𝑘 . Thus, 𝑑⟨𝜇, 𝑣⟩𝑥 is a vector in ℝ𝑘 , and since it is zero, each
component of 𝑑⟨𝜇, 𝑣⟩𝑥 is zero.
This means that each component of the Lie derivative 𝑣 𝜇 is zero at 𝑥. In other words,
𝑣 𝜇 vanishes on the entire manifold 𝑀, which implies that the tangent space of the
𝑇 𝑘 -orbit through 𝑥 is isotropic.
Therefore, we have shown that every 𝑇 𝑘 -orbit is isotropic.

Problem 19. √ √
Suppose that the 𝑇 2 -action on ℂ𝑃 2 by (exp −1𝜃1 , exp −1𝜃2 ) ⋅ [𝑧0 ∶ 𝑧1 ∶ 𝑧2 ] = [𝑧0 ∶
√ √ 2 2
exp −1𝜃1 𝑧1 ∶ exp −1𝜃2 𝑧2 ] has a moment map 𝜇[𝑧0 ∶ 𝑧1 ∶ 𝑧2 ] = − 12 ( |𝑧0 |2 +|𝑧|𝑧11 ||2 +|𝑧2 |2 , |𝑧0 |2 +|𝑧|𝑧21 ||2 +|𝑧2 |2 )

1. Find all fixed points of the 𝑇 2 − 𝑎𝑐𝑡𝑖𝑜𝑛.

2. What is the image of ℂ𝑃 2 under the moment map 𝜇?
3. Exhibit explicitly the subset of ℂ𝑃 2 for which the stabilizer under the following
𝑆 1 − 𝑎𝑐𝑡𝑖𝑜𝑛
√ √
−1𝜃1 −1𝜃1
(exp ⋅[𝑧0 ∶ 𝑧1 ∶ 𝑧2 ] = [𝑧0 ∶ exp 𝑧1 ∶ 𝑧2 ].

Solution.

1. Fixed Points of the 𝑇 2 -action

The fixed points of the 𝑇 2 -action are the points [𝑧0 ∶ 𝑧1 ∶ 𝑧2 ] satisfying the equation:
√ √ √
−1𝜃1 −1𝜃1 −1𝜃2
exp ⋅[𝑧0 ∶ 𝑧1 ∶ 𝑧2 ] = [𝑧0 ∶ exp 𝑧1 ∶ exp 𝑧2 ]
√ √
⇒ 𝑧1 = exp −1𝜃1
𝑧1 and 𝑧2 = exp −1𝜃2
𝑧2
The only solution to these equations is 𝑧1 = 0 and 𝑧2 = 0, which corresponds to the fixed
point [1 ∶ 0 ∶ 0].

2. Image of ℂ𝑃 2 under the Moment Map 𝜇

The image of ℂ𝑃 2 under the moment map 𝜇 is given by:
{ }
𝜇(ℂ𝑃 2 ) = 𝜇[𝑧0 ∶ 𝑧1 ∶ 𝑧2 ] ∣ [𝑧0 ∶ 𝑧1 ∶ 𝑧2 ] ∈ ℂ𝑃 2
{ }
1 |𝑧1 |2 1 |𝑧2 |2
= − ,− ∣ [𝑧0 ∶ 𝑧1 ∶ 𝑧2 ] ∈ ℂ𝑃 2 .
( 2 |𝑧0 |2 + |𝑧1 |2 + |𝑧2 |2 2 |𝑧0 |2 + |𝑧1 |2 + |𝑧2 |2 )

3. Subset of ℂ𝑃 2 with Stabilizer under 𝑆 1 -action

The 𝑆 1 -action is given by:
√ √
−1𝜃1 −1𝜃1
exp ⋅[𝑧0 ∶ 𝑧1 ∶ 𝑧2 ] = [𝑧0 ∶ exp 𝑧1 ∶ 𝑧2 ].
The subset of ℂ𝑃 2 for which the stabilizer
√
under this 𝑆 1 -action is√nontrivial consists of
points [𝑧0 ∶ 𝑧1 ∶ 𝑧2 ] such that 𝑧1 = exp −1𝜃1
𝑧1 . This implies that −1𝜃1 = 2𝜋𝑘 for some
integer 𝑘, which means 𝜃1 = −2𝜋𝑘.

17
1. Fixed Points of the 𝑇 2 -action
To find the fixed points of the 𝑇 2 -action, we need to solve the equation
√ √
−1𝜃1 −1𝜃2
[𝑧0 ∶ exp 𝑧1 ∶ exp 𝑧2 ] = [𝑧0 ∶ 𝑧1 ∶ 𝑧2 ].
√ √
This equation implies that exp −1𝜃1 𝑧1 = 𝑧1 and exp −1𝜃2 𝑧2 = 𝑧2 , which gives us 𝜃1 = 2𝜋𝑛1
and 𝜃2 = 2𝜋𝑛2 , where 𝑛1 , 𝑛2 ∈ ℤ. Therefore, the fixed points are the points [𝑧0 ∶ 𝑧1 ∶ 𝑧2 ]
such that 𝑧1 and 𝑧2 are not zero.

2. Image of ℂ𝑃 2 under the Moment Map 𝜇

The image of ℂ𝑃 2 under the moment map 𝜇 is the set of points that satisfy

1 |𝑧1 |2 1 |𝑧2 |2
𝜇[𝑧0 ∶ 𝑧1 ∶ 𝑧2 ] = − ,− .
( 2 |𝑧0 |2 + |𝑧1 |2 + |𝑧2 |2 2 |𝑧0 |2 + |𝑧1 |2 + |𝑧2 |2 )

This describes a map from ℂ𝑃 2 to a subset of ℝ2 .

3. Subset of ℂ𝑃 2 with Stabilizer under the 𝑆 1 -action

The given 𝑆 1 -action is
√ √
−1𝜃1 −1𝜃1
exp ⋅[𝑧0 ∶ 𝑧1 ∶ 𝑧2 ] = [𝑧0 ∶ exp 𝑧1 ∶ 𝑧2 ].

This action stabilizes the points where 𝑧1 is zero, i.e., the points [𝑧0 ∶ 0 ∶ 𝑧2 ]. These
points form a subset of ℂ𝑃 2 .

Problem 20.

(1). Let Δ be the 𝑛-simplex in ℝ𝑛 spanned by the origin and the standard basis vec-
tors (1, 0, … , 0, 0), (0, 1, 0, … , 0, 0), … , (0, 0, … , 0, 1). Show that the corresponding
symplectic toric manifold is a complex projective space ℂℙ𝑛 .
(2). Let Δ be the 𝑛-cubic in ℝ𝑛 whose vertices are (±1, ±1, … , ±1). Show that the
corresponding symplectic toric manifold is a product of complex projective planes
ℂℙ1 × ⋯ × ℂℙ1 .
Solution.

(1). To show that the symplectic toric manifold corresponding to the 𝑛-simplex Δ is
a complex projective space ℂℙ𝑛 , we need to establish the moment map and the
action of a torus.
The vertices of the 𝑛-simplex Δ are the standard basis vectors and the origin in ℝ𝑛 .
The action of the torus 𝑇 𝑛 on ℝ𝑛 is given by

(𝑡1 , 𝑡2 , … , 𝑡𝑛 ) ⋅ (𝑥1 , 𝑥2 , … , 𝑥𝑛 ) = (𝑡1 𝑥1 , 𝑡2 𝑥2 , … , 𝑡𝑛 𝑥𝑛 ),

where 𝑡𝑖 ∈ 𝑆 1 and 𝑥𝑖 ∈ ℝ.

18
 The corresponding moment map is defined as

𝜇 ∶ ℝ𝑛 → ℝ𝑛 ,
1
(𝑥1 , 𝑥2 , … , 𝑥𝑛 ) ↦ (|𝑥1 |2 , |𝑥2 |2 , … , |𝑥𝑛 |2 ).
2
The image of the moment map is the positive orthant in ℝ𝑛 , which is a cone.
Projecting the cone to the quotient space under the 𝑇 𝑛 action yields a complex
projective space ℂℙ𝑛 .
(2). The 𝑛-cubic Δ in ℝ𝑛 has vertices at (±1, ±1, … , ±1). The action of the torus 𝑇 𝑛 on
ℝ𝑛 is the same as in part (1). The corresponding moment map is given by

𝜇 ∶ ℝ𝑛 → ℝ𝑛 ,
1
(𝑥1 , 𝑥2 , … , 𝑥𝑛 ) ↦ (|𝑥1 |2 , |𝑥2 |2 , … , |𝑥𝑛 |2 ).
2
The image of the moment map is again the positive orthant in ℝ𝑛 , which results in
a product of 𝑛 copies of complex projective planes ℂℙ1 × ⋯ × ℂℙ1 .

19