24
CZ Martingales A Class of Calder´ on-Zygmund Operators Arising from the Projections of Martingale Transforms Michael Perlmutter Department of Mathematics Purdue University 1 M. Perlmutter(Purdue) CZ Martingales

A Class of Calderón-Zygmund Operators Arising …CZ Martingales A Class of Calder on-Zygmund Operators Arising from the Projections of Martingale Transforms Michael Perlmutter Department

  • Upload
    others

  • View
    1

  • Download
    0

Embed Size (px)

Citation preview

Page 1: A Class of Calderón-Zygmund Operators Arising …CZ Martingales A Class of Calder on-Zygmund Operators Arising from the Projections of Martingale Transforms Michael Perlmutter Department

CZ Martingales

A Class of Calderon-Zygmund Operators Arisingfrom the Projections of Martingale Transforms

Michael Perlmutter

Department of MathematicsPurdue University

1 M. Perlmutter(Purdue) CZ Martingales

Page 2: A Class of Calderón-Zygmund Operators Arising …CZ Martingales A Class of Calder on-Zygmund Operators Arising from the Projections of Martingale Transforms Michael Perlmutter Department

CZ Martingales

Calderon-Zygmund Operators

Let T be an operator on S(Rn), which is bounded on L2(Rn) of the form

Tf (x) = p.v .

∫Rn

K (x , x)f (x)dx .

T is called Calderon-Zygmund if K is C 1 on {x 6= x} and

|K (x , x)| ≤ κ

|x − x |n

|∇xK (x , x)| ≤ κ

|x − x |n+1

|∇xK (x , x)| ≤ κ

|x − x |n+1,

for some universal constant κ whenever x 6= x .

2 M. Perlmutter(Purdue) CZ Martingales

Page 3: A Class of Calderón-Zygmund Operators Arising …CZ Martingales A Class of Calder on-Zygmund Operators Arising from the Projections of Martingale Transforms Michael Perlmutter Department

CZ Martingales

Examples

Hilbert Transform

Hf (x) =1

π

∫R

1

x − xf (x)dx .

Riesz Transform

Ri f (x) = Cn

∫Rn

xi − xi|x − x |n+1

f (x)dx ,

1 ≤ i ≤ n, Cn =Γ( n+1

2 )π(n+1)/2 .

Beurling-Ahlfors Transform

Bf (z) = − 1

π

∫C

1

(z − w)2f (w)dw .

3 M. Perlmutter(Purdue) CZ Martingales

Page 4: A Class of Calderón-Zygmund Operators Arising …CZ Martingales A Class of Calder on-Zygmund Operators Arising from the Projections of Martingale Transforms Michael Perlmutter Department

CZ Martingales

Fourier Multipliers and Pseudo-Differential Operators

Fourier Multipliers

Rj f (ξ) =iξj|ξ|

f (ξ)

Bf (ξ) =ξ2

1 − ξ22 − 2iξ1ξ2

|ξ|2f (ξ)

B = R22 − R2

1 + 2iR1R2.

Pseudo-Differential Operators

Rj f = ∂xj (−4)−1/2f

Bf =∂z∂z

f .

4 M. Perlmutter(Purdue) CZ Martingales

Page 5: A Class of Calderón-Zygmund Operators Arising …CZ Martingales A Class of Calder on-Zygmund Operators Arising from the Projections of Martingale Transforms Michael Perlmutter Department

CZ Martingales

Symmetric α-stable Processes

Characteristic Function

E(e iξ·Xt ) = e−t|ξ|α, 0 < α ≤ 2.

Examples:

α = 2: Xt is Brownian Motion with density ht(x) = 1(4πt)n/2 e

−|x |2/4t .

α = 1: Xt is the Cauchy process with density pt(x) = Cnt

(|x |2+t2)(n+1)/2 .

Properties

Semigroup: ψs ∗ ψt = ψs+t .Scaling: ψt(x) = 1

tn/αψ1( x

t1/α ).

5 M. Perlmutter(Purdue) CZ Martingales

Page 6: A Class of Calderón-Zygmund Operators Arising …CZ Martingales A Class of Calder on-Zygmund Operators Arising from the Projections of Martingale Transforms Michael Perlmutter Department

CZ Martingales

Setup

I (Xt)t>0, rotationally-invariant α-stable process on Rn, 0 < α ≤ 2.

I For y > 0, ϕy (x) = 1ynϕ( xy ), where ϕ denotes the density of X1.

I Let A(x , y) = (ai ,j(x , y)) be an (n + 1)× (n + 1) matrix-valuedfunction

‖A‖ = ‖ sup|v |≤1

(|A(x , y)v |)‖L∞(Rn×[0,∞)) <∞.

I ai ,j(x , y) = ai ,j(y) whenever i or j = n + 1.

6 M. Perlmutter(Purdue) CZ Martingales

Page 7: A Class of Calderón-Zygmund Operators Arising …CZ Martingales A Class of Calder on-Zygmund Operators Arising from the Projections of Martingale Transforms Michael Perlmutter Department

CZ Martingales

Results

Theorem:

Consider the kernel

KA(x , x) =

∫ ∞0

∫Rn

2yA(x , y)∇ϕy (x − x)∇ϕy (x − x)dxdy ,

where ∇ = (∂x1 , . . . , ∂xn , ∂y ). Then the operator

TAf (x) =

∫Rn

K (x , x)f (x)dx

is a CZ operator.

7 M. Perlmutter(Purdue) CZ Martingales

Page 8: A Class of Calderón-Zygmund Operators Arising …CZ Martingales A Class of Calder on-Zygmund Operators Arising from the Projections of Martingale Transforms Michael Perlmutter Department

CZ Martingales

Results cont’d

Remark:

If ai ,j(y) = 0 whenever i or j = n + 1, we may also write our kernel interms of the density of Xt .The scaling relation ψt(x) = 1

tn/αψ1( x

t1/α ) implies ϕt1/α = ψt .Therefore,

KA(x , x) =

∫ ∞0

∫Rn

2

αt

2α−1A(x , t1/α)∇ψt(x − x)∇ψt(x − x)dxdt.

8 M. Perlmutter(Purdue) CZ Martingales

Page 9: A Class of Calderón-Zygmund Operators Arising …CZ Martingales A Class of Calder on-Zygmund Operators Arising from the Projections of Martingale Transforms Michael Perlmutter Department

CZ Martingales

Boundedness of CZ Operators

Strong type

‖Tf ‖p ≤ Cp,n,κ‖f ‖p, 1 < p <∞ (1)

Weak type

|{x : |Tf (x)| > λ}| ≤ C1,n,κ

λ‖f ‖1. (2)

Questions

For which operators can the constant in (1) and (2) be taken independentof n?What are the best possible constants?

9 M. Perlmutter(Purdue) CZ Martingales

Page 10: A Class of Calderón-Zygmund Operators Arising …CZ Martingales A Class of Calder on-Zygmund Operators Arising from the Projections of Martingale Transforms Michael Perlmutter Department

CZ Martingales

Probabilistic Representations of the Riesz Transform

Basic Idea

Lp → Mp → Mp → Lp

Background Radiation

“Time-reversed Brownian motion” in Rn+1+ . B−∞ has Lebesgue

distribution on Rn × {∞}, B0 has Lebesgue distribution on Rn × {0}.

Embedding into Mp

For f ∈ Lp, let uf be it’s Poisson extension to Rn+1+ . Then

(Xt)t≤0 = (uf (Bt))t≤0 is a martingale and

uf (Bt) =

∫ t

−∞∇uf (Bs) · dBs .

10 M. Perlmutter(Purdue) CZ Martingales

Page 11: A Class of Calderón-Zygmund Operators Arising …CZ Martingales A Class of Calder on-Zygmund Operators Arising from the Projections of Martingale Transforms Michael Perlmutter Department

CZ Martingales

Probabilistic Representations of the Riesz Transform cont’d

Martingale Transforms

Let A(x , y) an (n + 1)× (n + 1) matrix-valued function, ‖A‖ <∞.

(A ∗ f )t =

∫ t

−∞A(Bs)∇uf (Bs) · dBs .

Theorem: (Banuelos, Wang, Burkholder)

‖(A ∗ f )‖p ≤ (p∗ − 1)‖A‖‖X‖p 1 < p <∞,

p∗ = max{p, pp−1}.

11 M. Perlmutter(Purdue) CZ Martingales

Page 12: A Class of Calderón-Zygmund Operators Arising …CZ Martingales A Class of Calder on-Zygmund Operators Arising from the Projections of Martingale Transforms Michael Perlmutter Department

CZ Martingales

Probabilistic Representations of the Riesz Transform cont’d

Projection of the Martingale Transform

TAf (x) = E((A ∗ f )0|B0 = (x , 0))

Theorem: (Banuelos and Wang) ‖TAf ‖p ≤ (p∗ − 1)‖A‖‖f ‖p for1 < p <∞.

Orthogonal Martingale Transforms

Theorem: (Banuelos and Wang) Suppose further that A(x , y)v · v = 0 forall (x , y) ∈ Rn+1

+ , v ∈ Rn+1, then

‖TAf ‖p ≤ cot

(2π

p∗

)‖A‖‖f ‖p for 1 < p <∞.

12 M. Perlmutter(Purdue) CZ Martingales

Page 13: A Class of Calderón-Zygmund Operators Arising …CZ Martingales A Class of Calder on-Zygmund Operators Arising from the Projections of Martingale Transforms Michael Perlmutter Department

CZ Martingales

Riesz Transforms

Choose Aj = (ajl ,m),

ajl ,m =

1 l = n + 1, m = j−1 l = j , m = n + 10 otherwise

.

Then TAj f = Rj f and consequently ‖Rj f ‖ ≤ cot(

2πp∗

)‖f ‖p.

Beurling-Ahlfors Transform

Let

B =

2 2i 02i −2 00 0 0

.

Then TB f = Bf and consequently ‖Bf ‖ ≤ 4(p∗ − 1)‖f ‖p.

13 M. Perlmutter(Purdue) CZ Martingales

Page 14: A Class of Calderón-Zygmund Operators Arising …CZ Martingales A Class of Calder on-Zygmund Operators Arising from the Projections of Martingale Transforms Michael Perlmutter Department

CZ Martingales

∫Rn

TAf (x)g(x)dx =

∫Rn

E

(∫ 0

−∞A(Bs)∇uf (Bs) · dBs |X0 = x

)g(x)dx

= E

(∫ 0

−∞A(Bs)∇uf (Bs) · dBsug (B0)

)= E

(∫ 0

−∞A(Bs)∇uf (Bs) · dBs

∫ 0

−∞∇ug (Bs) · dBs

)= E

(∫ 0

−∞A(Bs)∇uf (Bs) · ∇ug (Bs)ds

)=

∫ ∞0

∫Rn

2yA(x , y)∇uf (x , y) · ∇ug (x , y)dxdy .

Using the fact that ∇uf (x , y) = ((∇py ) ∗ f )(x) and applying Fubini’stheorem, we see that we have

KA(x , x) =

∫ ∞0

∫Rn

2yA(x , y)∇py (x − x)∇py (x − x)dxdy .

14 M. Perlmutter(Purdue) CZ Martingales

Page 15: A Class of Calderón-Zygmund Operators Arising …CZ Martingales A Class of Calder on-Zygmund Operators Arising from the Projections of Martingale Transforms Michael Perlmutter Department

CZ Martingales

Space-time Brownian Motion and Heat Martingales

Space-time Brownian Motion

Fix T > 0 and let (Xt)0≤t≤T be Brownian motion with initial distributiongiven by Lebesgue measure on Rn. Let

(Zt)0≤t≤T = (Xt ,T − t)0≤t≤T .

Heat martingales

For f ∈ Lp, let uf (x , t) = (ht ∗ f )(x). By Ito’s formula, (uf (Zt))0≤t≤T is amartingale and

uf (Zt) =

∫ t

0∇xuf (Zs)dXs .

15 M. Perlmutter(Purdue) CZ Martingales

Page 16: A Class of Calderón-Zygmund Operators Arising …CZ Martingales A Class of Calder on-Zygmund Operators Arising from the Projections of Martingale Transforms Michael Perlmutter Department

CZ Martingales

Projections of Heat Martingales

Projection Operator

If A is an n × n matrix valued function, ‖A‖ <∞

STA f (x) = E

(∫ T

0A(Zt)∇xuf (Zs)dXs |XT = x)

),

SAf (x) = limT→∞

STA f (x)

Theorem: (Banuelos and Mendez) ‖SAf (x)‖p ≤ (p∗ − 1)‖A‖‖f ‖p for1 < p <∞.

Analytic Representation

SAf (x) =∫Rn K (x , x)dx where

K (x , x) =

∫ ∞0

∫Rn

A(x , t)∇xht(x − x) · ∇xht(x − x)dxdt.

16 M. Perlmutter(Purdue) CZ Martingales

Page 17: A Class of Calderón-Zygmund Operators Arising …CZ Martingales A Class of Calder on-Zygmund Operators Arising from the Projections of Martingale Transforms Michael Perlmutter Department

CZ Martingales

Projections of Heat Martingales cont’d

Beurling-Ahlfors Transform

Let

B =

(1 ii −1

).

Then SB f = Bf , consequently, ‖Bf ‖p ≤ 2(p∗ − 1)‖f ‖p for 1 < p <∞.

17 M. Perlmutter(Purdue) CZ Martingales

Page 18: A Class of Calderón-Zygmund Operators Arising …CZ Martingales A Class of Calder on-Zygmund Operators Arising from the Projections of Martingale Transforms Michael Perlmutter Department

CZ Martingales

Proof of Theorem 1

K (i ,j)(x , x) =

∫ ∞0

∫Rn

2yai ,j(x , y)∂xiϕy (x − x)∂xjϕy (x − x)dxdy

Case 1: i or j = n + 1:

K (i ,j)(x , x) =

∫ ∞0

∫Rn

2yai ,j(x , y)∂xiϕy (x − x)∂xjϕy (x − x)dxdy

=

∫ ∞0

2yai ,j(y)∂xi∂xjϕ21/αy (x − x)dy

≤ ‖ai ,j‖∞K (x − x)

K (x) =

∫ ∞0

2y |∂xi∂xjϕ21/αy (x)|dy

18 M. Perlmutter(Purdue) CZ Martingales

Page 19: A Class of Calderón-Zygmund Operators Arising …CZ Martingales A Class of Calder on-Zygmund Operators Arising from the Projections of Martingale Transforms Michael Perlmutter Department

CZ Martingales

The Proof of Theorem 1, cont’d

|K (x)| =

∫ ∞0

2y |∂xi∂xjϕ21/αy (x)|dy

=

∫ ∞0

2y |∂xi∂xjϕ21/α|x | y|x|(|x |x ′)|dy

=1

|x |n

∫ ∞0

2y |∂xi∂xjϕ21/αt(x′)|dt.

19 M. Perlmutter(Purdue) CZ Martingales

Page 20: A Class of Calderón-Zygmund Operators Arising …CZ Martingales A Class of Calder on-Zygmund Operators Arising from the Projections of Martingale Transforms Michael Perlmutter Department

CZ Martingales

Proof of Theorem 1, cont’d

Case 2: 1 ≤ i , j ,≤ n

|K (i ,j)(x , x)| ≤ ‖ai ,j‖∞∫ ∞

0

∫Rn

2y |∂xiϕy (w)||∂xjϕy (w − (x − x))|dwdy .

K (x) =

∫Rn

∫ ∞0

2y |∂xiϕy (w)||∂xjϕy (w − x)|dwdy

=1

|x |n

∫Rn

∫ ∞0

2y |∂xiϕt(z)||∂xjϕt(z − x ′)|dzdt.

20 M. Perlmutter(Purdue) CZ Martingales

Page 21: A Class of Calderón-Zygmund Operators Arising …CZ Martingales A Class of Calder on-Zygmund Operators Arising from the Projections of Martingale Transforms Michael Perlmutter Department

CZ Martingales

Estimates on the Derivatives

Lemma

There exists a constant Cn,α, depending only on n and α, such that for allx ∈ Rn, 1 ≤ i , j ≤ n,

|ϕ(x)| ≤ Cn,α

(1 + |x |2)(n+α)/2

|∂xiϕ(x)| ≤ Cn,α|x |(1 + |x |2)(n+2+α)/2

≤ Cn,α

(1 + |x |2)(n+1+α)/2

and

|∂xi∂xjϕ(x)| ≤ Cn,α

(1 + |x |2)(n+2+α)/2.

21 M. Perlmutter(Purdue) CZ Martingales

Page 22: A Class of Calderón-Zygmund Operators Arising …CZ Martingales A Class of Calder on-Zygmund Operators Arising from the Projections of Martingale Transforms Michael Perlmutter Department

CZ Martingales

Subordinated Brownian Motion

Stable Subordinator

Xt = BTt ,

Bt is a standard Brownian motion and Tt is the α/2 stable subordinatorwith density ηα/2(t, ·) .

Estimates

ψt(x) =

∫ ∞0

1

(4πs)n/2e−|x |

2/4sηα/2(t, s)ds,

∂xiϕ(x) =

∫ ∞0

1

(4πs)n/2

xise−|x |

2/4sηα/2 (1, s) ds

ηα/2(t, s) ≤ Cαts−1−α/2,

|∂xiϕ(x)| ≤ Cα|x |n+1+α

∫ ∞0

u(n+α)/2e−udu.

22 M. Perlmutter(Purdue) CZ Martingales

Page 23: A Class of Calderón-Zygmund Operators Arising …CZ Martingales A Class of Calder on-Zygmund Operators Arising from the Projections of Martingale Transforms Michael Perlmutter Department

CZ Martingales

Estimate Near 0

Fourier Inversion

ϕ(x) =

∫Rn

e−ix ·ξe−|ξ|α.

|∂xiϕ(x)| =

∣∣∣∣∫Rn

ξie−ix ·ξe−|ξ|

αdξ

∣∣∣∣=

∣∣∣∣∫Rn

ξi (e−ix ·ξ − 1)e−|ξ|

αdξ

∣∣∣∣≤∫Rn

|ξ| |e−ix ·ξ − 1|e−|ξ|αdξ

≤ 2

∫Rn

|ξ|2|x |e−|ξ|αdξ ≤ Cn,α|x |,

23 M. Perlmutter(Purdue) CZ Martingales

Page 24: A Class of Calderón-Zygmund Operators Arising …CZ Martingales A Class of Calder on-Zygmund Operators Arising from the Projections of Martingale Transforms Michael Perlmutter Department

CZ Martingales

THANK YOU!

24 M. Perlmutter(Purdue) CZ Martingales