Lecture 2: The tangential Cauchy-Riemann

operator and the Levi form

Andy Raich

University of Arkansas

10th Workshop on the Geometric Analysis of PDEs and Several

Complex Variables
Outline of the Talk

I To give several constructions of the ∂¯b -operator

I To relate the ∂¯b operator to more general constructions, e.g.,
Involutive Structures.
I To introduce the Levi form L.
I To discuss the ∂¯b and b problems
Geometry: The tangential Cauchy-Riemann
operator
Recap
Definition
M is an embedded CR manifold if dimR Hp (M) is independent of
p ∈ M.

Definition
For p ∈ U ∩ M, set
∂¯b f = τM {∂¯f˜}.
where τM : Λ0,q T ∗ (Cn )|M → Λ0,q T ∗ (M) is the orthogonal
projection map and f˜ is any smooth extension of f .

The ∂¯b -problem is the the problem of finding a form u so that

∂¯b u = f

for a given form f (which necessarily has ∂¯b f = 0)

Recap
Definition
M is an embedded CR manifold if dimR Hp (M) is independent of
p ∈ M.

Definition
For p ∈ U ∩ M, set
∂¯b f = τM {∂¯f˜}.
where τM : Λ0,q T ∗ (Cn )|M → Λ0,q T ∗ (M) is the orthogonal
projection map and f˜ is any smooth extension of f .

The ∂¯b -problem is the the problem of finding a form u so that

∂¯b u = f

for a given form f (which necessarily has ∂¯b f = 0)

Recap
Definition
M is an embedded CR manifold if dimR Hp (M) is independent of
p ∈ M.

Definition
For p ∈ U ∩ M, set
∂¯b f = τM {∂¯f˜}.
where τM : Λ0,q T ∗ (Cn )|M → Λ0,q T ∗ (M) is the orthogonal
projection map and f˜ is any smooth extension of f .

The ∂¯b -problem is the the problem of finding a form u so that

∂¯b u = f

for a given form f (which necessarily has ∂¯b f = 0)

Say what?

Questions:

I What did we just do?

I Is ∂¯b well-defined?
I How do I compute τM ?
I Does this agree with the calculation on model domains?
I Why should we care about the ∂¯b -problem?

Easy question first: Yes, ∂¯b is well-defined because τM forces

everything to be tangential and hence independent of the extension
(like earlier).
Say what?

Questions:

I What did we just do?

I Is ∂¯b well-defined?
I How do I compute τM ?
I Does this agree with the calculation on model domains?
I Why should we care about the ∂¯b -problem?

Easy question first: Yes, ∂¯b is well-defined because τM forces

everything to be tangential and hence independent of the extension
(like earlier).
More on the ∂¯b -problem

Question: Why should we care about the ∂¯b -problem?

Answer:

1. The ∂¯b -problem is fundamentally tied into the boundary

values of holomorphic functions.
2. The ∂¯b -operator encodes all of the complex structures and
complex geometry on the CR manifold M.
3. Hans Lewy was studying the ∂¯b -equation on the Heisenberg
when he formulated the first nonsolvable PDE.
4. We have seen repeatedly that closed range of ∂¯b is a critical
ingredient to prove embeddability of abstract CR manifolds!!
More on the ∂¯b -problem

Question: Why should we care about the ∂¯b -problem?

Answer:

1. The ∂¯b -problem is fundamentally tied into the boundary

values of holomorphic functions.
2. The ∂¯b -operator encodes all of the complex structures and
complex geometry on the CR manifold M.
3. Hans Lewy was studying the ∂¯b -equation on the Heisenberg
when he formulated the first nonsolvable PDE.
4. We have seen repeatedly that closed range of ∂¯b is a critical
ingredient to prove embeddability of abstract CR manifolds!!
Computing τM if M = bΩ
Recall: If X is a vector and φ is an r -form, then the contraction
X yφ is a (r − 1)-form defined by

hX yφ, L1 ∧ · · · ∧ Lr −1 i = hφ, X ∧ L1 · · · ∧ Lr −1 i

Philosophically, X y removes the dual of X from φ.

Suppose ρ is a defining function for M = bΩ with |dρ| = 1 on M.
Let N P
be the unit outward complex normal (this means
N = 4 nj=1 ∂z∂ρ ∂
j ∂ z̄j
).

Lemma

¯ ∧ φ)
τM {φ} = Ny(∂ρ for φ ∈ Λ0,q T ∗ (Cn )|M
and
∂¯b f = Ny ∂ρ
¯ ∧ ∂¯f˜) for f ∈ E p,q
M
Computing τM if M = bΩ
Recall: If X is a vector and φ is an r -form, then the contraction
X yφ is a (r − 1)-form defined by

hX yφ, L1 ∧ · · · ∧ Lr −1 i = hφ, X ∧ L1 · · · ∧ Lr −1 i

Philosophically, X y removes the dual of X from φ.

Suppose ρ is a defining function for M = bΩ with |dρ| = 1 on M.
Let N P
be the unit outward complex normal (this means
N = 4 nj=1 ∂z∂ρ ∂
j ∂ z̄j
).

Lemma

¯ ∧ φ)
τM {φ} = Ny(∂ρ for φ ∈ Λ0,q T ∗ (Cn )|M
and
∂¯b f = Ny ∂ρ
¯ ∧ ∂¯f˜) for f ∈ E p,q
M
Computing τM if M = bΩ
Recall: If X is a vector and φ is an r -form, then the contraction
X yφ is a (r − 1)-form defined by

hX yφ, L1 ∧ · · · ∧ Lr −1 i = hφ, X ∧ L1 · · · ∧ Lr −1 i

Philosophically, X y removes the dual of X from φ.

Suppose ρ is a defining function for M = bΩ with |dρ| = 1 on M.
Let N P
be the unit outward complex normal (this means
N = 4 nj=1 ∂z∂ρ ∂
j ∂ z̄j
).

Lemma

¯ ∧ φ)
τM {φ} = Ny(∂ρ for φ ∈ Λ0,q T ∗ (Cn )|M
and
∂¯b f = Ny ∂ρ
¯ ∧ ∂¯f˜) for f ∈ E p,q
M
Computing τM if M = bΩ
Recall: If X is a vector and φ is an r -form, then the contraction
X yφ is a (r − 1)-form defined by

hX yφ, L1 ∧ · · · ∧ Lr −1 i = hφ, X ∧ L1 · · · ∧ Lr −1 i

Philosophically, X y removes the dual of X from φ.

Suppose ρ is a defining function for M = bΩ with |dρ| = 1 on M.
Let N P
be the unit outward complex normal (this means
N = 4 nj=1 ∂z∂ρ ∂
j ∂ z̄j
).

Lemma

¯ ∧ φ)
τM {φ} = Ny(∂ρ for φ ∈ Λ0,q T ∗ (Cn )|M
and
∂¯b f = Ny ∂ρ
¯ ∧ ∂¯f˜) for f ∈ E p,q
M
Example 1

Let M = {ρ = 0} ⊂ Cn , p ∈ M, and U ⊂ M a neighborhood of p.

And suppose |dρ| = 1 on M.
Let {L1 , . . . , Ln−1 } be an orthonormal basis for T 1,0 (M ∩ U).
Then {L̄1 , . . . , L̄n−1 } is an orthonormal basis for T 0,1 (M ∩ U).
Let T = i Im N. Then {Lj , L̄k , T : 1 ≤ j, k ≤ n − 1} is a basis of
T (M) ⊗ C. Let {ω1 , . . . , ωn−1 , ω̄1 , . . . , ω̄n−1 , γ} be a dual basis (in
T ∗ (M) ⊗ C).
This means
ωj {Lj } = ω̄k {L̄k } = γ{T } = 1
and all other pairings are 0.
¯
Let ωn = ∂ρ. Then ∂ρ(N) = 1.
Example 1

Let M = {ρ = 0} ⊂ Cn , p ∈ M, and U ⊂ M a neighborhood of p.

And suppose |dρ| = 1 on M.
Let {L1 , . . . , Ln−1 } be an orthonormal basis for T 1,0 (M ∩ U).
Then {L̄1 , . . . , L̄n−1 } is an orthonormal basis for T 0,1 (M ∩ U).
Let T = i Im N. Then {Lj , L̄k , T : 1 ≤ j, k ≤ n − 1} is a basis of
T (M) ⊗ C. Let {ω1 , . . . , ωn−1 , ω̄1 , . . . , ω̄n−1 , γ} be a dual basis (in
T ∗ (M) ⊗ C).
This means
ωj {Lj } = ω̄k {L̄k } = γ{T } = 1
and all other pairings are 0.
¯
Let ωn = ∂ρ. Then ∂ρ(N) = 1.
Example 1

Let M = {ρ = 0} ⊂ Cn , p ∈ M, and U ⊂ M a neighborhood of p.

And suppose |dρ| = 1 on M.
Let {L1 , . . . , Ln−1 } be an orthonormal basis for T 1,0 (M ∩ U).
Then {L̄1 , . . . , L̄n−1 } is an orthonormal basis for T 0,1 (M ∩ U).
Let T = i Im N. Then {Lj , L̄k , T : 1 ≤ j, k ≤ n − 1} is a basis of
T (M) ⊗ C. Let {ω1 , . . . , ωn−1 , ω̄1 , . . . , ω̄n−1 , γ} be a dual basis (in
T ∗ (M) ⊗ C).
This means
ωj {Lj } = ω̄k {L̄k } = γ{T } = 1
and all other pairings are 0.
¯
Let ωn = ∂ρ. Then ∂ρ(N) = 1.
Example 1

Let M = {ρ = 0} ⊂ Cn , p ∈ M, and U ⊂ M a neighborhood of p.

And suppose |dρ| = 1 on M.
Let {L1 , . . . , Ln−1 } be an orthonormal basis for T 1,0 (M ∩ U).
Then {L̄1 , . . . , L̄n−1 } is an orthonormal basis for T 0,1 (M ∩ U).
Let T = i Im N. Then {Lj , L̄k , T : 1 ≤ j, k ≤ n − 1} is a basis of
T (M) ⊗ C. Let {ω1 , . . . , ωn−1 , ω̄1 , . . . , ω̄n−1 , γ} be a dual basis (in
T ∗ (M) ⊗ C).
This means
ωj {Lj } = ω̄k {L̄k } = γ{T } = 1
and all other pairings are 0.
¯
Let ωn = ∂ρ. Then ∂ρ(N) = 1.
Example 1

Let M = {ρ = 0} ⊂ Cn , p ∈ M, and U ⊂ M a neighborhood of p.

And suppose |dρ| = 1 on M.
Let {L1 , . . . , Ln−1 } be an orthonormal basis for T 1,0 (M ∩ U).
Then {L̄1 , . . . , L̄n−1 } is an orthonormal basis for T 0,1 (M ∩ U).
Let T = i Im N. Then {Lj , L̄k , T : 1 ≤ j, k ≤ n − 1} is a basis of
T (M) ⊗ C. Let {ω1 , . . . , ωn−1 , ω̄1 , . . . , ω̄n−1 , γ} be a dual basis (in
T ∗ (M) ⊗ C).
This means
ωj {Lj } = ω̄k {L̄k } = γ{T } = 1
and all other pairings are 0.
¯
Let ωn = ∂ρ. Then ∂ρ(N) = 1.
Example 1

Let M = {ρ = 0} ⊂ Cn , p ∈ M, and U ⊂ M a neighborhood of p.

And suppose |dρ| = 1 on M.
Let {L1 , . . . , Ln−1 } be an orthonormal basis for T 1,0 (M ∩ U).
Then {L̄1 , . . . , L̄n−1 } is an orthonormal basis for T 0,1 (M ∩ U).
Let T = i Im N. Then {Lj , L̄k , T : 1 ≤ j, k ≤ n − 1} is a basis of
T (M) ⊗ C. Let {ω1 , . . . , ωn−1 , ω̄1 , . . . , ω̄n−1 , γ} be a dual basis (in
T ∗ (M) ⊗ C).
This means
ωj {Lj } = ω̄k {L̄k } = γ{T } = 1
and all other pairings are 0.
¯
Let ωn = ∂ρ. Then ∂ρ(N) = 1.
Example 1

Let M = {ρ = 0} ⊂ Cn , p ∈ M, and U ⊂ M a neighborhood of p.

And suppose |dρ| = 1 on M.
Let {L1 , . . . , Ln−1 } be an orthonormal basis for T 1,0 (M ∩ U).
Then {L̄1 , . . . , L̄n−1 } is an orthonormal basis for T 0,1 (M ∩ U).
Let T = i Im N. Then {Lj , L̄k , T : 1 ≤ j, k ≤ n − 1} is a basis of
T (M) ⊗ C. Let {ω1 , . . . , ωn−1 , ω̄1 , . . . , ω̄n−1 , γ} be a dual basis (in
T ∗ (M) ⊗ C).
This means
ωj {Lj } = ω̄k {L̄k } = γ{T } = 1
and all other pairings are 0.
¯
Let ωn = ∂ρ. Then ∂ρ(N) = 1.
Example 1: Sections and extensions

∂¯b f = Ny ∂ρ
¯ ∧ ∂¯f˜)

A section of E 0,q
U is a (0, q)-form
X
f = fJ ω̄ J .
J∈Iq
n6∈J

¯ we can write an extension of f by

Since ω̄n = ∂ρ,
X X
f˜ = f˜J ω̄ J + ∂ρ
¯ ∧ f˜I ω̄ I ,
J∈Iq I ∈Iq−1
n6∈J

where f˜J is an extension of fJ .

Example 1: Sections and extensions

∂¯b f = Ny ∂ρ
¯ ∧ ∂¯f˜)

A section of E 0,q
U is a (0, q)-form
X
f = fJ ω̄ J .
J∈Iq
n6∈J

¯ we can write an extension of f by

Since ω̄n = ∂ρ,
X X
f˜ = f˜J ω̄ J + ∂ρ
¯ ∧ f˜I ω̄ I ,
J∈Iq I ∈Iq−1
n6∈J

where f˜J is an extension of fJ .

Example 1

∂¯b f = Ny ∂ρ
¯ ∧ ∂¯f˜)

Then
X X X
∂¯f˜ = ∂¯f˜J ∧ ω̄ J + f˜J ∂¯ω̄ J − ∂ρ
¯ ∧ ∂¯ f˜I ∧ ω̄ I

J∈Iq J∈Iq I ∈Iq−1
n6∈J n6∈J
n
XX
= L̄k f˜J ω̄ k ∧ ω̄ J + O(f˜) + ∂ρ
¯ ∧ (· · · )
J∈Iq k=1
n−1
XX
= L̄k f˜J ω̄ k ∧ ω̄ J + O(f˜) + ∂ρ
¯ ∧ (· · · )
J∈Iq k=1
n6∈J
Example 1

∂¯b f = Ny ∂ρ
¯ ∧ ∂¯f˜)

Then
X X X
∂¯f˜ = ∂¯f˜J ∧ ω̄ J + f˜J ∂¯ω̄ J − ∂ρ
¯ ∧ ∂¯ f˜I ∧ ω̄ I

J∈Iq J∈Iq I ∈Iq−1
n6∈J n6∈J
n
XX
= L̄k f˜J ω̄ k ∧ ω̄ J + O(f˜) + ∂ρ
¯ ∧ (· · · )
J∈Iq k=1
n−1
XX
= L̄k f˜J ω̄ k ∧ ω̄ J + O(f˜) + ∂ρ
¯ ∧ (· · · )
J∈Iq k=1
n6∈J
Example 1

∂¯b f = Ny ∂ρ
¯ ∧ ∂¯f˜)

Then
X X X
∂¯f˜ = ∂¯f˜J ∧ ω̄ J + f˜J ∂¯ω̄ J − ∂ρ
¯ ∧ ∂¯ f˜I ∧ ω̄ I

J∈Iq J∈Iq I ∈Iq−1
n6∈J n6∈J
n
XX
= L̄k f˜J ω̄ k ∧ ω̄ J + O(f˜) + ∂ρ
¯ ∧ (· · · )
J∈Iq k=1
n−1
XX
= L̄k f˜J ω̄ k ∧ ω̄ J + O(f˜) + ∂ρ
¯ ∧ (· · · )
J∈Iq k=1
n6∈J
Example 1

∂¯b f = Ny ∂ρ
¯ ∧ ∂¯f˜)

And
n−1
XX
¯ ∧ ∂¯f˜ = ∂ρ
∂ρ ¯ ∧ L̄k f˜J ω̄ k ∧ ω̄ J + ∂ρ
¯ ∧ O(f˜).
J∈Iq k=1
n6∈J

¯
Finally, since ∂ρ(N) = 1, it follows that
n−1
XX
∂¯b f = L̄k f˜J ω̄ k ∧ ω̄ J + O(f˜).
J∈Iq k=1
n6∈J

The terms that comprise O(f˜) are important for many applications.
Example 1

∂¯b f = Ny ∂ρ
¯ ∧ ∂¯f˜)

And
n−1
XX
¯ ∧ ∂¯f˜ = ∂ρ
∂ρ ¯ ∧ L̄k f˜J ω̄ k ∧ ω̄ J + ∂ρ
¯ ∧ O(f˜).
J∈Iq k=1
n6∈J

¯
Finally, since ∂ρ(N) = 1, it follows that
n−1
XX
∂¯b f = L̄k f˜J ω̄ k ∧ ω̄ J + O(f˜).
J∈Iq k=1
n6∈J

The terms that comprise O(f˜) are important for many applications.
Example 1

∂¯b f = Ny ∂ρ
¯ ∧ ∂¯f˜)

And
n−1
XX
¯ ∧ ∂¯f˜ = ∂ρ
∂ρ ¯ ∧ L̄k f˜J ω̄ k ∧ ω̄ J + ∂ρ
¯ ∧ O(f˜).
J∈Iq k=1
n6∈J

¯
Finally, since ∂ρ(N) = 1, it follows that
n−1
XX
∂¯b f = L̄k f˜J ω̄ k ∧ ω̄ J + O(f˜).
J∈Iq k=1
n6∈J

The terms that comprise O(f˜) are important for many applications.
The Key Fact

Recall: If M is a CR submanifold of Cn , then

1. H 1,0 (M) ∩ H 1,0 (M) = {0}.

2. H 1,0 (M) and H 0,1 (M) are involutive (i.e., closed under
commutators).

These properties make no mention of a complex structure on Cn !

The Key Fact

Recall: If M is a CR submanifold of Cn , then

1. H 1,0 (M) ∩ H 1,0 (M) = {0}.

2. H 1,0 (M) and H 0,1 (M) are involutive (i.e., closed under
commutators).

These properties make no mention of a complex structure on Cn !

Abstract CR Manifolds

Definition
Let M be a C ∞ manifold and suppose L is a subbundle of
T (M) ⊗ C. The pair (M, L) is called an abstract CR manifold or
CR structure if

1. Lp ∩ Lp = {0} for each p ∈ M

2. L is involutive, that is, [L1 , L2 ] ∈ L whenever L1 , L2 ∈ L.

Remarks

I If M ⊂ Cn , then L = H 1,0 (M) provides the CR structure.

I The number d = dimC (T (M) ⊗ C)/L ⊕ L is the CR
codimension of (M, L).
Abstract CR Manifolds

Definition
Let M be a C ∞ manifold and suppose L is a subbundle of
T (M) ⊗ C. The pair (M, L) is called an abstract CR manifold or
CR structure if

1. Lp ∩ Lp = {0} for each p ∈ M

2. L is involutive, that is, [L1 , L2 ] ∈ L whenever L1 , L2 ∈ L.

Remarks

I If M ⊂ Cn , then L = H 1,0 (M) provides the CR structure.

I The number d = dimC (T (M) ⊗ C)/L ⊕ L is the CR
codimension of (M, L).
Abstract CR Manifolds

Definition
Let M be a C ∞ manifold and suppose L is a subbundle of
T (M) ⊗ C. The pair (M, L) is called an abstract CR manifold or
CR structure if

1. Lp ∩ Lp = {0} for each p ∈ M

2. L is involutive, that is, [L1 , L2 ] ∈ L whenever L1 , L2 ∈ L.

Remarks

I If M ⊂ Cn , then L = H 1,0 (M) provides the CR structure.

I The number d = dimC (T (M) ⊗ C)/L ⊕ L is the CR
codimension of (M, L).
Towards an intrinsic ∂¯b

Key Notation

I For p ∈ M, define Xp to be the orthogonal complement of

Lp ⊕ Lp in Tp (M) ⊗ C. This is not Xp (M).
I The totally
S real part of the tangent bundle is
X (M) = p∈M Xp .
I Define the subbundles

T 0,1 (M) = L and T 1,0 (M) = L ⊕ X (M).

Do not confuse T 1,0 (M) with H 1,0 (M)!!

Towards an intrinsic ∂¯b

Key Notation

I For p ∈ M, define Xp to be the orthogonal complement of

Lp ⊕ Lp in Tp (M) ⊗ C. This is not Xp (M).
I The totally
S real part of the tangent bundle is
X (M) = p∈M Xp .
I Define the subbundles

T 0,1 (M) = L and T 1,0 (M) = L ⊕ X (M).

Do not confuse T 1,0 (M) with H 1,0 (M)!!

Towards an intrinsic ∂¯b

Key Notation

I For p ∈ M, define Xp to be the orthogonal complement of

Lp ⊕ Lp in Tp (M) ⊗ C. This is not Xp (M).
I The totally
S real part of the tangent bundle is
X (M) = p∈M Xp .
I Define the subbundles

T 0,1 (M) = L and T 1,0 (M) = L ⊕ X (M).

Do not confuse T 1,0 (M) with H 1,0 (M)!!

Towards an intrinsic ∂¯b

Key Notation

I For p ∈ M, define Xp to be the orthogonal complement of

Lp ⊕ Lp in Tp (M) ⊗ C. This is not Xp (M).
I The totally
S real part of the tangent bundle is
X (M) = p∈M Xp .
I Define the subbundles

T 0,1 (M) = L and T 1,0 (M) = L ⊕ X (M).

Do not confuse T 1,0 (M) with H 1,0 (M)!!

Towards an intrinsic ∂¯b

Key Notation

I For p ∈ M, define Xp to be the orthogonal complement of

Lp ⊕ Lp in Tp (M) ⊗ C. This is not Xp (M).
I The totally
S real part of the tangent bundle is
X (M) = p∈M Xp .
I Define the subbundles

T 0,1 (M) = L and T 1,0 (M) = L ⊕ X (M).

Do not confuse T 1,0 (M) with H 1,0 (M)!!

Forms on M

Dual to T 0,1 (M) and T 1,0 (M) are T 0,1 (M)∗ and T 1,0 (M)∗ ,
respectively and the bundle of (p, q)-forms is

Λp,q T ∗ (M) = Λp (T 1,0 (M)∗ )⊗Λ

ˆ q (T 0,1 (M)∗ )

ˆ is the anti-symmetric tensor product (e.g., ω I ∧ ω̄ J ).

where ⊗
The bundle of r forms is

Λr (T ∗ (M) ⊗ C) = Λr ,0 T ∗ (M) ⊕ · · · ⊕ Λ0,r T ∗ (M).

Denote the natural projection

p,q
πM : Λr (T ∗ (M) ⊗ C) → Λp,q (T ∗ (M) ⊗ C).
Forms on M

Dual to T 0,1 (M) and T 1,0 (M) are T 0,1 (M)∗ and T 1,0 (M)∗ ,
respectively and the bundle of (p, q)-forms is

Λp,q T ∗ (M) = Λp (T 1,0 (M)∗ )⊗Λ

ˆ q (T 0,1 (M)∗ )

ˆ is the anti-symmetric tensor product (e.g., ω I ∧ ω̄ J ).

where ⊗
The bundle of r forms is

Λr (T ∗ (M) ⊗ C) = Λr ,0 T ∗ (M) ⊕ · · · ⊕ Λ0,r T ∗ (M).

Denote the natural projection

p,q
πM : Λr (T ∗ (M) ⊗ C) → Λp,q (T ∗ (M) ⊗ C).
Forms on M

Dual to T 0,1 (M) and T 1,0 (M) are T 0,1 (M)∗ and T 1,0 (M)∗ ,
respectively and the bundle of (p, q)-forms is

Λp,q T ∗ (M) = Λp (T 1,0 (M)∗ )⊗Λ

ˆ q (T 0,1 (M)∗ )

ˆ is the anti-symmetric tensor product (e.g., ω I ∧ ω̄ J ).

where ⊗
The bundle of r forms is

Λr (T ∗ (M) ⊗ C) = Λr ,0 T ∗ (M) ⊕ · · · ⊕ Λ0,r T ∗ (M).

Denote the natural projection

p,q
πM : Λr (T ∗ (M) ⊗ C) → Λp,q (T ∗ (M) ⊗ C).
Sections on M and an intrinsically defined ∂¯b

Given an open set U ⊂ M, let E p,q (M) be the space of sections of

(p, q)-forms over U. Let dM be the exterior derivative operator on
M.

Definition
The tangential Cauchy-Riemann operator ∂¯b : E p,q p,q
M → E M is
defined by
p,q+1
∂¯b = πM ◦ dM .
Sections on M and an intrinsically defined ∂¯b

Given an open set U ⊂ M, let E p,q (M) be the space of sections of

(p, q)-forms over U. Let dM be the exterior derivative operator on
M.

Definition
The tangential Cauchy-Riemann operator ∂¯b : E p,q p,q
M → E M is
defined by
p,q+1
∂¯b = πM ◦ dM .
Sections on M and an intrinsically defined ∂¯b

Given an open set U ⊂ M, let E p,q (M) be the space of sections of

(p, q)-forms over U. Let dM be the exterior derivative operator on
M.

Definition
The tangential Cauchy-Riemann operator ∂¯b : E p,q p,q
M → E M is
defined by
p,q+1
∂¯b = πM ◦ dM .
Extrinsic vs. intrinsic ∂¯b operators

The intrinsic ∂¯b operator shares many of the properties that the
extrinsically defined ∂¯b operator does.
In fact...

Theorem
Suppose M is a CR submanifold of Cn . The extrinsic and intrinsic
tangential Cauchy-Riemann complexes are isomorphic.
Extrinsic vs. intrinsic ∂¯b operators

The intrinsic ∂¯b operator shares many of the properties that the
extrinsically defined ∂¯b operator does.
In fact...

Theorem
Suppose M is a CR submanifold of Cn . The extrinsic and intrinsic
tangential Cauchy-Riemann complexes are isomorphic.
Comments

I X (M) was defined to be the orthogonal complement to

L ⊗ L. This requires a metric. A different metric will give a
different, but isomorphic, ∂¯b -complex.
I We could avoid metrics altogether using quotients. For
example, if M ⊂ Cn , we could have defined

Λp,q T ∗ (M) = Λp,q T ∗ (C)/I p,q .

There’s a similar construction in the abstract CR manifold

case.
I The pair (M, L) is called a CR structure
I A function f is called a CR function if ∂¯b f = 0.
Comments

I X (M) was defined to be the orthogonal complement to

L ⊗ L. This requires a metric. A different metric will give a
different, but isomorphic, ∂¯b -complex.
I We could avoid metrics altogether using quotients. For
example, if M ⊂ Cn , we could have defined

Λp,q T ∗ (M) = Λp,q T ∗ (C)/I p,q .

There’s a similar construction in the abstract CR manifold

case.
I The pair (M, L) is called a CR structure
I A function f is called a CR function if ∂¯b f = 0.
Comments

I X (M) was defined to be the orthogonal complement to

L ⊗ L. This requires a metric. A different metric will give a
different, but isomorphic, ∂¯b -complex.
I We could avoid metrics altogether using quotients. For
example, if M ⊂ Cn , we could have defined

Λp,q T ∗ (M) = Λp,q T ∗ (C)/I p,q .

There’s a similar construction in the abstract CR manifold

case.
I The pair (M, L) is called a CR structure
I A function f is called a CR function if ∂¯b f = 0.
Comments

I X (M) was defined to be the orthogonal complement to

L ⊗ L. This requires a metric. A different metric will give a
different, but isomorphic, ∂¯b -complex.
I We could avoid metrics altogether using quotients. For
example, if M ⊂ Cn , we could have defined

Λp,q T ∗ (M) = Λp,q T ∗ (C)/I p,q .

There’s a similar construction in the abstract CR manifold

case.
I The pair (M, L) is called a CR structure
I A function f is called a CR function if ∂¯b f = 0.
Comments

I X (M) was defined to be the orthogonal complement to

L ⊗ L. This requires a metric. A different metric will give a
different, but isomorphic, ∂¯b -complex.
I We could avoid metrics altogether using quotients. For
example, if M ⊂ Cn , we could have defined

Λp,q T ∗ (M) = Λp,q T ∗ (C)/I p,q .

There’s a similar construction in the abstract CR manifold

case.
I The pair (M, L) is called a CR structure
I A function f is called a CR function if ∂¯b f = 0.
Comments

I X (M) was defined to be the orthogonal complement to

L ⊗ L. This requires a metric. A different metric will give a
different, but isomorphic, ∂¯b -complex.
I We could avoid metrics altogether using quotients. For
example, if M ⊂ Cn , we could have defined

Λp,q T ∗ (M) = Λp,q T ∗ (C)/I p,q .

There’s a similar construction in the abstract CR manifold

case.
I The pair (M, L) is called a CR structure
I A function f is called a CR function if ∂¯b f = 0.
Comments

I X (M) was defined to be the orthogonal complement to

L ⊗ L. This requires a metric. A different metric will give a
different, but isomorphic, ∂¯b -complex.
I We could avoid metrics altogether using quotients. For
example, if M ⊂ Cn , we could have defined

Λp,q T ∗ (M) = Λp,q T ∗ (C)/I p,q .

There’s a similar construction in the abstract CR manifold

case.
I The pair (M, L) is called a CR structure
I A function f is called a CR function if ∂¯b f = 0.
∂¯b vs. ∂¯ and pullbacks

Warning! CR geometry is subtle! Suppose Ω1 , Ω2 ∈ Cn and

F : Ω1 → Ω2 is a biholomorphism. A basic result in SCV says
¯ ∗ = F ∗ ∂.
∂F ¯

Definition
Suppose (M, LM ) and (N, LN ) are CR structures. A C 1 map
f : M → N is a called a CR map if f∗ LM ⊂ LN .

Lemma
Suppose (M, L) is a CR structure. A C 1 map
f = (f1 , . . . , fm ) : M → Cm is a CR map if and only if each fj is a
CR function.
∂¯b vs. ∂¯ and pullbacks

Warning! CR geometry is subtle! Suppose Ω1 , Ω2 ∈ Cn and

F : Ω1 → Ω2 is a biholomorphism. A basic result in SCV says
¯ ∗ = F ∗ ∂.
∂F ¯

Definition
Suppose (M, LM ) and (N, LN ) are CR structures. A C 1 map
f : M → N is a called a CR map if f∗ LM ⊂ LN .

Lemma
Suppose (M, L) is a CR structure. A C 1 map
f = (f1 , . . . , fm ) : M → Cm is a CR map if and only if each fj is a
CR function.
∂¯b vs. ∂¯ and pullbacks

Warning! CR geometry is subtle! Suppose Ω1 , Ω2 ∈ Cn and

F : Ω1 → Ω2 is a biholomorphism. A basic result in SCV says
¯ ∗ = F ∗ ∂.
∂F ¯

Definition
Suppose (M, LM ) and (N, LN ) are CR structures. A C 1 map
f : M → N is a called a CR map if f∗ LM ⊂ LN .

Lemma
Suppose (M, L) is a CR structure. A C 1 map
f = (f1 , . . . , fm ) : M → Cm is a CR map if and only if each fj is a
CR function.
∂¯b vs. ∂¯ and pullbacks

Warning! CR geometry is subtle! Suppose Ω1 , Ω2 ∈ Cn and

F : Ω1 → Ω2 is a biholomorphism. A basic result in SCV says
¯ ∗ = F ∗ ∂.
∂F ¯

Definition
Suppose (M, LM ) and (N, LN ) are CR structures. A C 1 map
f : M → N is a called a CR map if f∗ LM ⊂ LN .

Lemma
Suppose (M, L) is a CR structure. A C 1 map
f = (f1 , . . . , fm ) : M → Cm is a CR map if and only if each fj is a
CR function.
∂¯b vs. ∂¯

Example
Let
M = {(z, u + iv ) ∈ C2 : v = 0}.
Then L = span{ ∂∂z̄ } and if g is a function

∂g
∂¯b g = d z̄
∂ z̄

Let f : M → M be defined by

f (z, u) = (z + u, u).

and f is a CR map since both functions are holomorphic in z.

∂¯b vs. ∂¯

Example
Let
M = {(z, u + iv ) ∈ C2 : v = 0}.
Then L = span{ ∂∂z̄ } and if g is a function

∂g
∂¯b g = d z̄
∂ z̄

Let f : M → M be defined by

f (z, u) = (z + u, u).

and f is a CR map since both functions are holomorphic in z.

∂¯b vs. ∂¯

Example
Let
M = {(z, u + iv ) ∈ C2 : v = 0}.
Then L = span{ ∂∂z̄ } and if g is a function

∂g
∂¯b g = d z̄
∂ z̄

Let f : M → M be defined by

f (z, u) = (z + u, u).

and f is a CR map since both functions are holomorphic in z.

∂¯b vs. ∂¯

Example
Let
M = {(z, u + iv ) ∈ C2 : v = 0}.
Then L = span{ ∂∂z̄ } and if g is a function

∂g
∂¯b g = d z̄
∂ z̄

Let f : M → M be defined by

f (z, u) = (z + u, u).

and f is a CR map since both functions are holomorphic in z.

∂¯b vs. ∂¯
Example
Let
M = {(z, u + iv ) ∈ C2 : v = 0}.

while
∂¯b f ∗ (z̄) = ∂¯b (z̄ + u) = d z̄.

Theorem
Suppose M and N are CR manifolds and f : M → N is a CR map.
then
∂¯b ◦ π p,q ◦ f ∗ = π p,q+1 ◦ f ∗ ◦ ∂¯b .
M M

Note: It is possible to set up ∂¯b intrinsically so that

∂¯b ◦ f ∗ = f ∗ ◦ ∂¯b .
∂¯b vs. ∂¯
Example
Let
M = {(z, u + iv ) ∈ C2 : v = 0}.

f ∗ ∂¯b z̄ = f ∗ d z̄ = d z̄ + du
while
∂¯b f ∗ (z̄) = ∂¯b (z̄ + u) = d z̄.

Theorem
Suppose M and N are CR manifolds and f : M → N is a CR map.
then
∂¯b ◦ π p,q ◦ f ∗ = π p,q+1 ◦ f ∗ ◦ ∂¯b .
M M

Note: It is possible to set up ∂¯b intrinsically so that

∂¯b ◦ f ∗ = f ∗ ◦ ∂¯b .
∂¯b vs. ∂¯
Example
Let
M = {(z, u + iv ) ∈ C2 : v = 0}.

f ∗ ∂¯b z̄ = f ∗ d z̄ = d z̄ + du
while
∂¯b f ∗ (z̄) = ∂¯b (z̄ + u) = d z̄.

Theorem
Suppose M and N are CR manifolds and f : M → N is a CR map.
then
∂¯b ◦ π p,q ◦ f ∗ = π p,q+1 ◦ f ∗ ◦ ∂¯b .
M M

Note: It is possible to set up ∂¯b intrinsically so that

∂¯b ◦ f ∗ = f ∗ ◦ ∂¯b .
∂¯b vs. ∂¯
Example
Let
M = {(z, u + iv ) ∈ C2 : v = 0}.

f ∗ ∂¯b z̄ = f ∗ d z̄ = d z̄ + du
while
∂¯b f ∗ (z̄) = ∂¯b (z̄ + u) = d z̄.

Theorem
Suppose M and N are CR manifolds and f : M → N is a CR map.
then
∂¯b ◦ π p,q ◦ f ∗ = π p,q+1 ◦ f ∗ ◦ ∂¯b .
M M

Note: It is possible to set up ∂¯b intrinsically so that

∂¯b ◦ f ∗ = f ∗ ◦ ∂¯b .
∂¯b vs. ∂¯
Example
Let
M = {(z, u + iv ) ∈ C2 : v = 0}.

f ∗ ∂¯b z̄ = f ∗ d z̄ = d z̄ + du
while
∂¯b f ∗ (z̄) = ∂¯b (z̄ + u) = d z̄.

Theorem
Suppose M and N are CR manifolds and f : M → N is a CR map.
then
∂¯b ◦ π p,q ◦ f ∗ = π p,q+1 ◦ f ∗ ◦ ∂¯b .
M M

Note: It is possible to set up ∂¯b intrinsically so that

∂¯b ◦ f ∗ = f ∗ ◦ ∂¯b .
∂¯b vs. ∂¯
Example
Let
M = {(z, u + iv ) ∈ C2 : v = 0}.

f ∗ ∂¯b z̄ = f ∗ d z̄ = d z̄ + du
while
∂¯b f ∗ (z̄) = ∂¯b (z̄ + u) = d z̄.

Theorem
Suppose M and N are CR manifolds and f : M → N is a CR map.
then
∂¯b ◦ π p,q ◦ f ∗ = π p,q+1 ◦ f ∗ ◦ ∂¯b .
M M

Note: It is possible to set up ∂¯b intrinsically so that

∂¯b ◦ f ∗ = f ∗ ◦ ∂¯b .
Aside – Involutive Structures

CR structures are only one of a class of structures that are studied

on smooth manifolds M.

Definition
Let M be a C ∞ manifold and suppose L is a subbundle of
T (M) ⊗ C. The pair (M, L) is called an abstract CR manifold or
CR structure if

I L is involutive, that is, [L1 , L2 ] ∈ L whenever L1 , L2 ∈ L.

I Lp ∩ Lp = {0} for each p ∈ M
Aside – Involutive Structures

CR structures are only one of a class of structures that are studied

on smooth manifolds M.

Definition
Let M be a C ∞ manifold and suppose L is a subbundle of
T (M) ⊗ C. The pair (M, L) is called an abstract CR manifold or
CR structure if

I L is involutive, that is, [L1 , L2 ] ∈ L whenever L1 , L2 ∈ L.

I Lp ∩ Lp = {0} for each p ∈ M
Aside – Involutive Structures

CR structures are only one of a class of structures that are studied

on smooth manifolds M.

Definition
Let M be a C ∞ manifold and suppose L is a subbundle of
T (M) ⊗ C. The pair (M, L) is called an formally integrable
structure if

I L is involutive, that is, [L1 , L2 ] ∈ L whenever L1 , L2 ∈ L.

Aside – Involutive Structures

CR structures are only one of a class of structures that are studied

on smooth manifolds M.

Definition
Let M be a C ∞ manifold and suppose L is a subbundle of
T (M) ⊗ C. The pair (M, L) is called an formally integrable
structure if

I L is involutive, that is, [L1 , L2 ] ∈ L whenever L1 , L2 ∈ L.

Examples beyond CR structures:

I A complex structure if Lp ⊗ Lp = Tp M ⊗ C for all p ∈ M

I An essentially real structure if Lp = Lp for all p ∈ M.
Geometry: The Levi Form
The Levi form

Let (M, L) be a CR structure.

Fact: The subbundle L ⊕ L ⊂ T (M) ⊗ C may not be involutive!
Recall:
πp : Tp (M) ⊗ C → Tp (M) ⊗ C/(L ⊕ L)
is the natural projection map.

Definition
The Levi form at p ∈ M is the map
Lp : Lp → Tp (M) ⊗ C/(L ⊕ L) defined by

1
Lp (Lp ) = πp {[L̄, L]p } for Lp ∈ Lp .
2i
The Levi form

Let (M, L) be a CR structure.

Fact: The subbundle L ⊕ L ⊂ T (M) ⊗ C may not be involutive!
Recall:
πp : Tp (M) ⊗ C → Tp (M) ⊗ C/(L ⊕ L)
is the natural projection map.

Definition
The Levi form at p ∈ M is the map
Lp : Lp → Tp (M) ⊗ C/(L ⊕ L) defined by

1
Lp (Lp ) = πp {[L̄, L]p } for Lp ∈ Lp .
2i
The Levi form

Let (M, L) be a CR structure.

Fact: The subbundle L ⊕ L ⊂ T (M) ⊗ C may not be involutive!
Recall:
πp : Tp (M) ⊗ C → Tp (M) ⊗ C/(L ⊕ L)
is the natural projection map.

Definition
The Levi form at p ∈ M is the map
Lp : Lp → Tp (M) ⊗ C/(L ⊕ L) defined by

1
Lp (Lp ) = πp {[L̄, L]p } for Lp ∈ Lp .
2i
The Levi form

Let (M, L) be a CR structure.

Fact: The subbundle L ⊕ L ⊂ T (M) ⊗ C may not be involutive!
Recall:
πp : Tp (M) ⊗ C → Tp (M) ⊗ C/(L ⊕ L)
is the natural projection map.

Definition
The Levi form at p ∈ M is the map
Lp : Lp → Tp (M) ⊗ C/(L ⊕ L) defined by

1
Lp (Lp ) = πp {[L̄, L]p } for Lp ∈ Lp .
2i
Facts about the Levi form

I The Levi form is well-defined, that is, if L, Z ∈ L and

Lp = Zp , then

πp {[L̄, L]p } = πp {[Z , Z ]p }.

I The Levi form is a CR invariant, that is, if (M, LM ), (N, LN )

are CR structures with Levi forms LM and LN , and
F : M → N is a CR diffeomorphism, then

F∗ (p) ◦ LM N
p = LF (p) ◦ F∗ (p)

for all p ∈ M
Facts about the Levi form

I The Levi form is well-defined, that is, if L, Z ∈ L and

Lp = Zp , then

πp {[L̄, L]p } = πp {[Z , Z ]p }.

I The Levi form is a CR invariant, that is, if (M, LM ), (N, LN )

are CR structures with Levi forms LM and LN , and
F : M → N is a CR diffeomorphism, then

F∗ (p) ◦ LM N
p = LF (p) ◦ F∗ (p)

for all p ∈ M
Computing the Levi form

Question: How do we compute the Levi form?

Answer:

Step 1 Choose a metric for T (M) ⊗ C.

Step 2 Identify Tp (M) ⊗ C/(L ⊕ L) with
Xp (M) ⊗ C = Tp (M)/Hp (M) ⊗ C.
Step 3 If M ⊂ Cn and p ∈ M,

Lp : Hp1,0 (M) → Xp (M) ⊗ C

given by
1 
Lp (Lp ) = πp [L̄, L]p , Lp ∈ Hp1,0 (M).
2i
Computing the Levi form

Question: How do we compute the Levi form?

Answer:

Step 1 Choose a metric for T (M) ⊗ C.

Step 2 Identify Tp (M) ⊗ C/(L ⊕ L) with
Xp (M) ⊗ C = Tp (M)/Hp (M) ⊗ C.
Step 3 If M ⊂ Cn and p ∈ M,

Lp : Hp1,0 (M) → Xp (M) ⊗ C

given by
1 
Lp (Lp ) = πp [L̄, L]p , Lp ∈ Hp1,0 (M).
2i
Computing the Levi form

Question: How do we compute the Levi form?

Answer:

Step 1 Choose a metric for T (M) ⊗ C.

Step 2 Identify Tp (M) ⊗ C/(L ⊕ L) with
Xp (M) ⊗ C = Tp (M)/Hp (M) ⊗ C.
Step 3 If M ⊂ Cn and p ∈ M,

Lp : Hp1,0 (M) → Xp (M) ⊗ C

given by
1 
Lp (Lp ) = πp [L̄, L]p , Lp ∈ Hp1,0 (M).
2i
Computing the Levi form

Question: How do we compute the Levi form?

Answer:

Step 1 Choose a metric for T (M) ⊗ C.

Step 2 Identify Tp (M) ⊗ C/(L ⊕ L) with
Xp (M) ⊗ C = Tp (M)/Hp (M) ⊗ C.
Step 3 If M ⊂ Cn and p ∈ M,

Lp : Hp1,0 (M) → Xp (M) ⊗ C

given by
1 
Lp (Lp ) = πp [L̄, L]p , Lp ∈ Hp1,0 (M).
2i
The Heisenberg group

Recall: Hn = {(z, w ) ∈ Cn+1 : Im w = |z|2 } and a basis for

H 1,0 (Hn ) is
∂ ∂
Lj = + 2i z̄j .
∂zj ∂w
∂
We choose a metric so that X (Hn ) is spanned by T = ∂t where
w = t + is. We compute

[L̄k , Lj ] = 2iδjk T

and hence
Lp (Lj ) = 1.
The Heisenberg group

Recall: Hn = {(z, w ) ∈ Cn+1 : Im w = |z|2 } and a basis for

H 1,0 (Hn ) is
∂ ∂
Lj = + 2i z̄j .
∂zj ∂w
∂
We choose a metric so that X (Hn ) is spanned by T = ∂t where
w = t + is. We compute

[L̄k , Lj ] = 2iδjk T

and hence
Lp (Lj ) = 1.
The Heisenberg group

Recall: Hn = {(z, w ) ∈ Cn+1 : Im w = |z|2 } and a basis for

H 1,0 (Hn ) is
∂ ∂
Lj = + 2i z̄j .
∂zj ∂w
∂
We choose a metric so that X (Hn ) is spanned by T = ∂t where
w = t + is. We compute

[L̄k , Lj ] = 2iδjk T

and hence
Lp (Lj ) = 1.
The Levi matrix

Let M ⊂ Cn+1 be the boundary of a domain. Let

{L1 , . . . , Ln } ∈ H 1,0 (M) be an orthonormal basis near a point
p0 ∈ M. We identify the Levi form with the Levi matrix (cjk ) given
by
1
cjk = πp {[L̄k , Lj ]}.
2i
The Levi form/Levi matrix is the most important geometric
structure intrinsic to M.
∂
If M = {(z, t + is) ∈ Cn × C : s = p(z)} then πp picks off the ∂t
component.

1h ∂ ∂p ∂ ∂ ∂p ∂ i ∂2p
cjk = − 2i , + 2i = .
2i ∂ z̄k ∂ z̄k ∂ w̄ ∂zj ∂zj ∂w ∂zj ∂ z̄k
The Levi matrix

Let M ⊂ Cn+1 be the boundary of a domain. Let

{L1 , . . . , Ln } ∈ H 1,0 (M) be an orthonormal basis near a point
p0 ∈ M. We identify the Levi form with the Levi matrix (cjk ) given
by
1
cjk = πp {[L̄k , Lj ]}.
2i
The Levi form/Levi matrix is the most important geometric
structure intrinsic to M.
∂
If M = {(z, t + is) ∈ Cn × C : s = p(z)} then πp picks off the ∂t
component.

1h ∂ ∂p ∂ ∂ ∂p ∂ i ∂2p
cjk = − 2i , + 2i = .
2i ∂ z̄k ∂ z̄k ∂ w̄ ∂zj ∂zj ∂w ∂zj ∂ z̄k
The Levi matrix

Let M ⊂ Cn+1 be the boundary of a domain. Let

{L1 , . . . , Ln } ∈ H 1,0 (M) be an orthonormal basis near a point
p0 ∈ M. We identify the Levi form with the Levi matrix (cjk ) given
by
1
cjk = πp {[L̄k , Lj ]}.
2i
The Levi form/Levi matrix is the most important geometric
structure intrinsic to M.
∂
If M = {(z, t + is) ∈ Cn × C : s = p(z)} then πp picks off the ∂t
component.

1h ∂ ∂p ∂ ∂ ∂p ∂ i ∂2p
cjk = − 2i , + 2i = .
2i ∂ z̄k ∂ z̄k ∂ w̄ ∂zj ∂zj ∂w ∂zj ∂ z̄k
The Levi matrix

Let M ⊂ Cn+1 be the boundary of a domain. Let

{L1 , . . . , Ln } ∈ H 1,0 (M) be an orthonormal basis near a point
p0 ∈ M. We identify the Levi form with the Levi matrix (cjk ) given
by
1
cjk = πp {[L̄k , Lj ]}.
2i
The Levi form/Levi matrix is the most important geometric
structure intrinsic to M.
∂
If M = {(z, t + is) ∈ Cn × C : s = p(z)} then πp picks off the ∂t
component.

1h ∂ ∂p ∂ ∂ ∂p ∂ i ∂2p
cjk = − 2i , + 2i = .
2i ∂ z̄k ∂ z̄k ∂ w̄ ∂zj ∂zj ∂w ∂zj ∂ z̄k
The Levi matrix

Let M ⊂ Cn+1 be the boundary of a domain. Let

{L1 , . . . , Ln } ∈ H 1,0 (M) be an orthonormal basis near a point
p0 ∈ M. We identify the Levi form with the Levi matrix (cjk ) given
by
1
cjk = πp {[L̄k , Lj ]}.
2i
The Levi form/Levi matrix is the most important geometric
structure intrinsic to M.
∂
If M = {(z, t + is) ∈ Cn × C : s = p(z)} then πp picks off the ∂t
component.

1h ∂ ∂p ∂ ∂ ∂p ∂ i ∂2p
cjk = − 2i , + 2i = .
2i ∂ z̄k ∂ z̄k ∂ w̄ ∂zj ∂zj ∂w ∂zj ∂ z̄k
The Levi matrix

Note: Here, M ⊂ Cn+1 so (cjk ) is an n × n matrix!

If the Levi matrix has

I only nonnegative (resp., positive) eigenvalue, then M is called

pseudoconvex, (resp., strictly pseudoconvex)
I at least n − (q + 1) positive or at least q + 1 negative
eigenvalues, then M satisfies Z (q).
I Z (q) and Z (n − q), then M satisfies Y (q).

Geometry Remark: The Levi form is to the complex Hessian what

the Second Fundamental Form is to the real Hessian.
One more remark: The Hodge-* operator induces symmetry
between levels (0, q) and (0, n − q) so symmetric requirements for
∂¯b are necessary.
The Levi matrix

Note: Here, M ⊂ Cn+1 so (cjk ) is an n × n matrix!

If the Levi matrix has

I only nonnegative (resp., positive) eigenvalue, then M is called

pseudoconvex, (resp., strictly pseudoconvex)
I at least n − (q + 1) positive or at least q + 1 negative
eigenvalues, then M satisfies Z (q).
I Z (q) and Z (n − q), then M satisfies Y (q).

Geometry Remark: The Levi form is to the complex Hessian what

the Second Fundamental Form is to the real Hessian.
One more remark: The Hodge-* operator induces symmetry
between levels (0, q) and (0, n − q) so symmetric requirements for
∂¯b are necessary.
The Levi matrix

Note: Here, M ⊂ Cn+1 so (cjk ) is an n × n matrix!

If the Levi matrix has

I only nonnegative (resp., positive) eigenvalue, then M is called

pseudoconvex, (resp., strictly pseudoconvex)
I at least n − (q + 1) positive or at least q + 1 negative
eigenvalues, then M satisfies Z (q).
I Z (q) and Z (n − q), then M satisfies Y (q).

Geometry Remark: The Levi form is to the complex Hessian what

the Second Fundamental Form is to the real Hessian.
One more remark: The Hodge-* operator induces symmetry
between levels (0, q) and (0, n − q) so symmetric requirements for
∂¯b are necessary.
The Levi matrix

Note: Here, M ⊂ Cn+1 so (cjk ) is an n × n matrix!

If the Levi matrix has

I only nonnegative (resp., positive) eigenvalue, then M is called

pseudoconvex, (resp., strictly pseudoconvex)
I at least n − (q + 1) positive or at least q + 1 negative
eigenvalues, then M satisfies Z (q).
I Z (q) and Z (n − q), then M satisfies Y (q).

Geometry Remark: The Levi form is to the complex Hessian what

the Second Fundamental Form is to the real Hessian.
One more remark: The Hodge-* operator induces symmetry
between levels (0, q) and (0, n − q) so symmetric requirements for
∂¯b are necessary.
The Levi matrix

Note: Here, M ⊂ Cn+1 so (cjk ) is an n × n matrix!

If the Levi matrix has

I only nonnegative (resp., positive) eigenvalue, then M is called

pseudoconvex, (resp., strictly pseudoconvex)
I at least n − (q + 1) positive or at least q + 1 negative
eigenvalues, then M satisfies Z (q).
I Z (q) and Z (n − q), then M satisfies Y (q).

Geometry Remark: The Levi form is to the complex Hessian what

the Second Fundamental Form is to the real Hessian.
One more remark: The Hodge-* operator induces symmetry
between levels (0, q) and (0, n − q) so symmetric requirements for
∂¯b are necessary.
The Levi matrix

Note: Here, M ⊂ Cn+1 so (cjk ) is an n × n matrix!

If the Levi matrix has

I only nonnegative (resp., positive) eigenvalue, then M is called

pseudoconvex, (resp., strictly pseudoconvex)
I at least n − (q + 1) positive or at least q + 1 negative
eigenvalues, then M satisfies Z (q).
I Z (q) and Z (n − q), then M satisfies Y (q).

Geometry Remark: The Levi form is to the complex Hessian what

the Second Fundamental Form is to the real Hessian.
One more remark: The Hodge-* operator induces symmetry
between levels (0, q) and (0, n − q) so symmetric requirements for
∂¯b are necessary.
Setting up the problem

The ∂¯b -problem is

∂¯b u = f .
Standard approach: Via the Kohn Laplacian

b = ∂¯b∗ ∂¯b + ∂¯b ∂¯b∗

and the b -equation

b v = f .
Note: b is neither elliptic nor constant coefficient!
Notation: Hp,q (M) = null b on (p, q)-forms is the space of
harmonic forms.
Setting up the problem

The ∂¯b -problem is

∂¯b u = f .
Standard approach: Via the Kohn Laplacian

b = ∂¯b∗ ∂¯b + ∂¯b ∂¯b∗

and the b -equation

b v = f .
Note: b is neither elliptic nor constant coefficient!
Notation: Hp,q (M) = null b on (p, q)-forms is the space of
harmonic forms.
Setting up the problem

The ∂¯b -problem is

∂¯b u = f .
Standard approach: Via the Kohn Laplacian

b = ∂¯b∗ ∂¯b + ∂¯b ∂¯b∗

and the b -equation

b v = f .
Note: b is neither elliptic nor constant coefficient!
Notation: Hp,q (M) = null b on (p, q)-forms is the space of
harmonic forms.
Setting up the problem

The ∂¯b -problem is

∂¯b u = f .
Standard approach: Via the Kohn Laplacian

b = ∂¯b∗ ∂¯b + ∂¯b ∂¯b∗

and the b -equation

b v = f .
Note: b is neither elliptic nor constant coefficient!
Notation: Hp,q (M) = null b on (p, q)-forms is the space of
harmonic forms.
The b -problem
Theorem
Suppose (M, L) is a hypersurface type CR manifold of dimension
2n + 1, 0 ≤ p ≤ n, and 1 ≤ q ≤ n. If b : L2p,q (M) → L2p,q (M) has
closed range, then

I The complex Green operator. b is invertible on Hp,q (M)⊥

with inverse Gp,q .
I Hodge decomposition.

L2p,q (M) = Range ∂¯b ⊕ Range ∂¯b∗ ⊕ Hp,q (M)

I Continuity. Gp,q : L2p,q (M) → L2p,q (M) continuously.

I Solvability of ∂¯b . If f ∈ L2p,q (M), ∂¯b f = 0, and f ⊥ Hp,q (M),
the
u = ∂¯b∗ Gp,q f
is the L2 -minimizing solution to ∂¯b u = f .
The b -problem
Theorem
Suppose (M, L) is a hypersurface type CR manifold of dimension
2n + 1, 0 ≤ p ≤ n, and 1 ≤ q ≤ n. If b : L2p,q (M) → L2p,q (M) has
closed range, then

I The complex Green operator. b is invertible on Hp,q (M)⊥

with inverse Gp,q .
I Hodge decomposition.

L2p,q (M) = Range ∂¯b ⊕ Range ∂¯b∗ ⊕ Hp,q (M)

I Continuity. Gp,q : L2p,q (M) → L2p,q (M) continuously.

I Solvability of ∂¯b . If f ∈ L2p,q (M), ∂¯b f = 0, and f ⊥ Hp,q (M),
the
u = ∂¯b∗ Gp,q f
is the L2 -minimizing solution to ∂¯b u = f .
The b -problem
Theorem
Suppose (M, L) is a hypersurface type CR manifold of dimension
2n + 1, 0 ≤ p ≤ n, and 1 ≤ q ≤ n. If b : L2p,q (M) → L2p,q (M) has
closed range, then

I The complex Green operator. b is invertible on Hp,q (M)⊥

with inverse Gp,q .
I Hodge decomposition.

L2p,q (M) = Range ∂¯b ⊕ Range ∂¯b∗ ⊕ Hp,q (M)

I Continuity. Gp,q : L2p,q (M) → L2p,q (M) continuously.

I Solvability of ∂¯b . If f ∈ L2p,q (M), ∂¯b f = 0, and f ⊥ Hp,q (M),
the
u = ∂¯b∗ Gp,q f
is the L2 -minimizing solution to ∂¯b u = f .
The b -problem
Theorem
Suppose (M, L) is a hypersurface type CR manifold of dimension
2n + 1, 0 ≤ p ≤ n, and 1 ≤ q ≤ n. If b : L2p,q (M) → L2p,q (M) has
closed range, then

I The complex Green operator. b is invertible on Hp,q (M)⊥

with inverse Gp,q .
I Hodge decomposition.

L2p,q (M) = Range ∂¯b ⊕ Range ∂¯b∗ ⊕ Hp,q (M)

I Continuity. Gp,q : L2p,q (M) → L2p,q (M) continuously.

I Solvability of ∂¯b . If f ∈ L2p,q (M), ∂¯b f = 0, and f ⊥ Hp,q (M),
the
u = ∂¯b∗ Gp,q f
is the L2 -minimizing solution to ∂¯b u = f .
The b -problem
Theorem
Suppose (M, L) is a hypersurface type CR manifold of dimension
2n + 1, 0 ≤ p ≤ n, and 1 ≤ q ≤ n. If b : L2p,q (M) → L2p,q (M) has
closed range, then

I The complex Green operator. b is invertible on Hp,q (M)⊥

with inverse Gp,q .
I Hodge decomposition.

L2p,q (M) = Range ∂¯b ⊕ Range ∂¯b∗ ⊕ Hp,q (M)

I Continuity. Gp,q : L2p,q (M) → L2p,q (M) continuously.

I Solvability of ∂¯b . If f ∈ L2p,q (M), ∂¯b f = 0, and f ⊥ Hp,q (M),
the
u = ∂¯b∗ Gp,q f
is the L2 -minimizing solution to ∂¯b u = f .
The Levi form and solvability

I If Ω ⊂ Cn and M = bΩ is bounded and pseudoconvex, then

∂¯b is solvable in L2p,q (Ω), all p, 1 ≤ q ≤ n − 1 and
Hp,q (M) = {0} if 1 ≤ q ≤ n − 2. (Shaw ’85, Boas-Shaw ’86).
I If M = bΩ is bounded and pseudoconvex in a Stein manifold,
then ∂¯b is solvable (same range of p, q) (Kohn ’86)
I M satisfies Y (q) if and only if the solution to ∂¯b gains 1/2
derivatives. (Kohn)
I If M = bΩ in a Stein manifold satisfies weak Y (q), then ∂¯b is
solvable at levels q and q − 1 and Hp,q (M) = {0}
(Harrington-Raich, 2015)
I If M ⊂ Cn and is a CR manifold of hypersurface type and
satisfies weak Y (q), then ∂¯b is solvable at levels q and q − 1
and dim Hp,q (M) < ∞ (Coacalle-Raich, submitted)
The Levi form and solvability

I If Ω ⊂ Cn and M = bΩ is bounded and pseudoconvex, then

∂¯b is solvable in L2p,q (Ω), all p, 1 ≤ q ≤ n − 1 and
Hp,q (M) = {0} if 1 ≤ q ≤ n − 2. (Shaw ’85, Boas-Shaw ’86).
I If M = bΩ is bounded and pseudoconvex in a Stein manifold,
then ∂¯b is solvable (same range of p, q) (Kohn ’86)
I M satisfies Y (q) if and only if the solution to ∂¯b gains 1/2
derivatives. (Kohn)
I If M = bΩ in a Stein manifold satisfies weak Y (q), then ∂¯b is
solvable at levels q and q − 1 and Hp,q (M) = {0}
(Harrington-Raich, 2015)
I If M ⊂ Cn and is a CR manifold of hypersurface type and
satisfies weak Y (q), then ∂¯b is solvable at levels q and q − 1
and dim Hp,q (M) < ∞ (Coacalle-Raich, submitted)
Thank you for your attention!

THANK YOU!!