0% found this document useful (0 votes)
36 views

Tarea Fourier

This document provides solutions to exercises from a homework assignment on maximal functions and the Lebesgue differentiation theorem. It first shows that three different definitions of the maximal function of a function f are equivalent up to constants. It then proves the Lebesgue differentiation theorem, which states that for almost every point x, the average of f over balls centered at x converges to f(x) as the radius goes to 0. Finally, it outlines an alternative proof of the Hardy-Littlewood-Sobolev inequality using maximal functions instead of interpolation.

Uploaded by

hugo mancera
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
36 views

Tarea Fourier

This document provides solutions to exercises from a homework assignment on maximal functions and the Lebesgue differentiation theorem. It first shows that three different definitions of the maximal function of a function f are equivalent up to constants. It then proves the Lebesgue differentiation theorem, which states that for almost every point x, the average of f over balls centered at x converges to f(x) as the radius goes to 0. Finally, it outlines an alternative proof of the Hardy-Littlewood-Sobolev inequality using maximal functions instead of interpolation.

Uploaded by

hugo mancera
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 5

HOMEWORK ASSIGNMENT 3 SOLUTIONS

Exercise 1. (Equivalence between the different notions of a maximal function)


Given f in L1loc (Rn ), we defined M f, M 0 f, M 00 f as follows:
Z
1
M f (x) := sup · |f (x − y)| dy
r>0 µ(Br ) Br
Z
1
M 0 f (x) := sup n
· |f (x − y)| dy
r>0 (2r) Qr
Z
1
M 00 f (x) := sup · |f (y)| dy.
Q3x µ(Q) Q
Q is a cube
Here Br = B(0, r), Qr = [−r, r] and µ is Lebesgue measure on Rn .
n

Show that:
M f ∼n M 0 f ∼n M 00 f.
Solution:

Let us first show that M f ∼n M 0 f .


2r
Let x ∈ Rn and r > 0 be given. We denote by Q1 (x, r) the cube centered at x of sidelength √
n
and let Q2 (x, r) denote the cube centered at x of side length 2r. Then:
(
Q1 (x, r) ⊆ B(x, r) ⊆ Q2 (x, r)
  
µ Q1 (x, r) ∼n µ B(x, r) ∼n µ Q2 (x, r) .
In particular: Z Z Z
− |f | .n − |f | .n − |f |.
Q1 (x,r) B(x,r) Q2 (x,r)
Taking suprema in r > 0, it follows that:
M 0 f .n M f .n M 0 f.
Hence:
M f ∼n M 0 f.
We now need to show that M 0 f ∼n M 00 f . Let us note that, by construction M 0 f ≤ M 00 f , so we
only need to show that M 00 f .n M 0 f . In order to do this, let x ∈ Rn be given and suppose that
x belongs to a cube Q, which is not necessarily centered at x. Let d denote the maximal distance
from x to a side of Q. Let Q
e denote the cube centered at x of sidelength 2d. Then:
(
e⊇Q
Q
e ≤ 2n µ(Q).
µ(Q)
It follows that:

2n 2n
Z Z Z Z Z
1
− |f | = |f | ≤ |f | ≤ |f | = 2n · − |f | ≤ 2n · M 0 f
µ(Q) µ(Q)
e µ(Q)
e
Q Q Q Q
e Q
e

Taking suprema over all cubes Q containing x, it follows that:


M 00 f ≤ 2n · M 0 f.
1
2 HOMEWORK ASSIGNMENT 3 SOLUTIONS

We can now deduce the claim. 


Exercise 2. (Lebesgue Differentiation Theorem)
Let f ∈ L1loc (Rn ) be given. Show the following:

a) For almost every x ∈ Rn ,


Z
1
(1) lim f (x − y) dy = f (x).
r→0+ µ(Br ) Br
n
Here, Br = B(0, r) is the Euclidean ball in R which is centered at the origin and µ is Lebsegue
measure on Rn .

b) Show moreover that, with notation as above, for almost every x ∈ A:


Z
1
f (x − y) − f (x) dy = 0.
(2) lim
r→0+ µ(Br ) B
r

If (2) holds, we say that x is a Lebesgue point of f .


Solution:

a) Let us first consider f ∈ L1 (Rn ). Given r > 0, we define:


Z
1
Tr f (x) := f (x − y) dy.
µ(Br ) Br
Then Tr is a linear operator on L1 (Rn ). Let us consider the associated maximal operator:
T ∗ = sup |Tr |.
r>0
We note that, for all x ∈ Rn , it is the case that:
1 Z 1
Z
T ∗ f (x) = sup f (x − y) dy ≤ sup |f (x − y)| dy ≤ M f (x).

r>0 µ(Br ) Br r>0 µ(Br ) Br

In particular, operator T ∗ is weak (1, 1). By a result proved in class, it follows that the set:
A := {f ∈ L1 (Rn ), Tr f → f a.e. as r → 0+}
is closed in L1 (Rn ). Let us note that C0 (Rn ) is contained in A. Since C0 (Rn ) is dense in L1 (Rn ),
it follows that A = L1 (Rn ). Hence (1) holds when f ∈ L1 (Rn ).

Let us now prove (1) for general f ∈ L1loc (Rn ). For m ∈ N, we consider fm := f · χB(0,m) .
Then, each fm ∈ L1loc . By the previous argument, there exists a set Nm ⊆ Rn with µ(Rn \ Nm ) = 0,
such that for all x ∈ Nm , it is the case that:
(3) lim Tr fm (x) = fm (x).
r→0+

Let us note that, on B(0, m):


(4) lim Tr fm = lim Tr f.
r→0+ r→0+
n
Let N := ∩m NM . Then, µ(R \ N ) = 0. From (3) and (4), it follows that:
lim Tr f (x) = f (x)
r→0+

for all x ∈ N . This proves (1) for general f ∈ L1loc (Rn ).

b) Let us note that it suffices to show that there exists A ⊆ Rn with µ(Rn \ A) = 0 such that,
for all x ∈ A and for all d ∈ C:
Z
1
(5) lim |f (x − y) − d| dy = |f (x) − d|.
r→0 µ(Br ) B
r
HOMEWORK ASSIGNMENT 3 SOLUTIONS 3

Given α ∈ Q + iQ, we can apply part 1) to the function f − α ∈ L1loc (Rn ) and deduce that there
exists a set Aα ⊆ Rn with µ(Rn \ Aα ) = 0 such that, on Aα :
Z
1
lim |f (x − y) − α| dy = |f (x) − α|.
r→0 µ(Br ) B
r

Let us now take: \


A := Aα .
α∈Q+iQ
n
Since Q + iQ is countable, it follows that µ(R \ A) = 0. We will now show that (5) holds on A.

Let us fix x ∈ A. Suppose that d ∈ C is given. Given  > 0, by density, we can find c ∈ Q + iQ such
that |c − d| < . By the triangle inequality, it follows that, for all r > 0:
Z  1 Z Z
1 1 
|f (x − y) − d| dy ∈ |f (x − y) − c| dy − , |f (x − y) − c| dy +  .
µ(Br ) Br µ(Br ) Br µ(Br ) Br
For r sufficiently small, this interval is contained in:
   
|f (x) − c| − 2, |f (x) − c| + 2 ⊆ |f (x) − d| − 3, |f (x) − d| + 3 .
(5) now follows if we let  → 0. 
Exercise 3. (The Hardy-Littlewood-Sobolev inequality revisited)
Let p, q ∈ (1, ∞) with p < q and α ∈ (0, n) be given. Suppose that:
1 1 α
= −
q p n
We recall that the Hardy-Littlewood-Sobolev inequality states that:
1
f ∗ n−α q n ≤ Ckf kLp (Rn )

|x| L (R )

for some C = C(p, q, n) > 0. On homework assignment 2, we showed this inequality by use of
the Marcinkiewicz interpolation theorem. In this exercise, we will outline how to show it by using
maximal functions.
Solution:

Let p, q be as in the assumptions. Suppose that f ∈ Lp (Rn ) is given. We will first obtain a
pointwise bound in x of the quantity:
|f (y)|
Z
Iα |f |(x) := n−α
dy.
Rn |x − y|
Let us choose r > 0 which we will determine later. As in the previous assignment, we write
1 (1) (2)
kα (w) := |w|n−α as kα = kα + kα , where:

kα(1) := kα · χ|w|<r , kα(2) := kα · χ|w|≥r .


We can write:
Iα |f | = kα(1) ∗ |f | + kα(2) ∗ |f |.
For the second term, we recall the bound from Homework 2:
n
kα(2) ∗ |f | (x) . rα− p · kf kLp .

(6)
(1)
We will now show an upper bound for kα ∗ |f | involving the Hardy-Littlewood maximal function.
Namely:

|f (y)| |f (y)|
Z X Z 
kα(1) ∗ |f | (x) =

dy = dy
|x−y|≤r |x − y|n−α 2−k−1 ·r<|x−y|≤2−k ·r |x − y|n−α
k=0
4 HOMEWORK ASSIGNMENT 3 SOLUTIONS

∞  Z
X 
.α,n (2−k r)α · − |f (y)| dy ≤
k=0 |x−y|≤2−k ·r


X
≤ 2−kα rα M f (x)
k=0

(7) .α rα M f (x).
We choose r > 0 such that:
n
rα M f (x) = rα− p · kf kLp .
In other words:
 kf k p  np
L
r= .
M f (x)
For this choice of r, we can combine (6) and (7) and deduce that:
αp αp
n −1+1 αp αp
Iα |f |(x) .α,n rα− p · kf kLp = kf kLnp · M f (x)(− n +1) = kf kLnp · M f (x)(1− n ) .
Putting everything together, it follows that:
|f (y)|
Z αp
(1− αp
n ).
(8) n−α
dy . α,n kf kL
n
p · M f (x)
R n |x − y|
αp
By assumption 1 − n = pq . Hence, we can rewrite (8) as:

|f (y)|
Z αp p

n−α
dy .α,n kf kLnp · M f (x) q .
Rn |x − y|

Taking q-th powers and integrating over Rn , since αp q−p


n = q , it follows that:
Z |f (y)| q Z
q−p
dy . kf k · [M f (x)]p dx.

n−α q n α,n Lp
|x − y|

R d L (R ) Rn

By the strong Lp boundedness of the Hardy-Littlewood maximal operator, this expression is:
.p kf kq−p p q
Lp · kf kLp = kf kLp .

Taking q-th roots, it follows that:


Z |f (y)|
dy q n .α,n,p kf kLp .

|x − y|n−α

Rd L (R )

In particular, by the triangle inequality:


Z f (y)
dy q n .α,n,p kf kLp

n−α
Rd |x − y|

L (R )

for all f ∈ Lp (Rn ). 


Remark: Estimate (8) is known as Hedberg’s inequality.
Exercise 4. (Maximal functions cannot locally oscillate too much)
Let f ∈ L1loc (Rn ) be given. Suppose that there exists a ball B ⊆ Rn and λ > 0 such that M f ≥ λ at
every point of B.
Show that
M f &n λ on 2B
where 2B is the ball with the same center as B of twice the radius.
HOMEWORK ASSIGNMENT 3 SOLUTIONS 5

Solution:

λ
R
We write B = B(c, r). For every x ∈ B, we can find r(x) > 0 such that −
B(x,r(x))
|f | > 2.
By compactness of B, we can find finitely many x1 , x2 , . . . , xm ∈ B such that:
m
[
B⊆ B(xj , r(xj )).
j=1

For simplicity of notation, we let rj := r(xj ) and we let Bj := B(xj , r(xj )). We now consider two
cases.

Case 1: There exists j such that rj ≥ r.

Let y ∈ 2B. Then, by the triangle inequality, it follows that:


|y − xj | ≤ |y − c| + |c − xj | ≤ 2r + r = 3r ≤ 3rj .
Hence, by another application of the triangle inequality, it follows that:
B(xj , rj ) ⊆ B(y, 4rj ).
In particular:
Z Z Z Z
1 1 λ
M f (y) ≥ − |f | ∼n n · |f | ≥ n · |f | ∼n − |f | > .
rj rj 2
B(y,4rj ) B(y,4rj ) B(xj ,rj ) B(xj ,rj )

Case 2: rj < r for all j.

By Wiener’s Vitali Covering Lemma, we can find a subsequence Bi1 , Bi2 , . . . , BiN of B1 , B2 , . . . , Bm
such that the Bi` are disjoint and such that:
N
X m
[
µ(Bi` ) ≥ 3−n µ Bj ≥ 3−n µ(B).


`=1 j=1

Now, for all ` ∈ {1, 2, . . . , m}, we know that ri` < r and so by the triangle inequality Bi` ⊆ 2B.
Moreover, by construction:
Z Z
λ λ
− |f | > ⇒ |f | > · µ(Bi` ).
Bi 2 Bi 2
` `

So, recalling the fact that the Bi` are disjoint and contained in 2B, it follows that:
Z n Z N
X λ X  λ
|f | ≥ |f | > · µ(Bi` ) ≥ · 3−n · µ(B).
2B Bi 2 2
`=1 ` `=1
We can deduce that:
Z
(9) − |f | &n λ.
2B
We claim that (9) implies that M f &n λ on 2B. Namely, if y ∈ 2B, then B(y, 4r) ⊇ 2B by the
triangle inequality. Hence:
Z Z Z Z
1 1
M f (y) ≥ − |f | ∼n n · |f | ≥ n |f | ∼n − |f | &n λ.
r r
B(y,4r) B(y,4r) 2B 2B

The claim now follows. 

You might also like