Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd
Upload
37 views

Technische Universit at Berlin: YB Iral CH Afer

This document contains 5 exercises related to functional analysis. Exercise 1 asks to prove that a certain operator T is of weak (1,1) and strong (p,p) type for 1 < p ≤ ∞. Exercise 2 asks to verify that log|x| is in BMO(R). Exercise 3 asks to prove that a dyadic decomposition of unity converges in a certain sense. Exercise 4 asks to show a system of functions is orthonormal but not a basis in L2(0,1). Exercise 5 provides tail estimates for sums of independent random variables.

Uploaded by

Hien Nguyen
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
37 views

Technische Universit at Berlin: YB Iral CH Afer

This document contains 5 exercises related to functional analysis. Exercise 1 asks to prove that a certain operator T is of weak (1,1) and strong (p,p) type for 1 < p ≤ ∞. Exercise 2 asks to verify that log|x| is in BMO(R). Exercise 3 asks to prove that a dyadic decomposition of unity converges in a certain sense. Exercise 4 asks to show a system of functions is orthonormal but not a basis in L2(0,1). Exercise 5 provides tail estimates for sums of independent random variables.

Uploaded by

Hien Nguyen
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

TECHNISCHE UNIVERSITÄT BERLIN Winter Term 2012/13

Institut für Mathematik


J. V YB ÍRAL / M. S CH ÄFER

Exercise Sheet 14
Functional Analysis II

Exercise 1:
Assume that ϕ ∈ L1 (Rn ) satisfies |ϕ(x)| ≤ ψ(|x|) almost everywhere, where ψ
is positive, radial, decreasing (as a function on (0, ∞)) and integrable. Let ϕt (x) =
t−n ϕ(x/t), x ∈ Rn . Prove that the operator T : Lp (Rn ) → Lp (Rn )

(T f )(x) = sup |(ϕt ∗ f )(x)|


t>0

is of weak (1, 1) and strong (p, p) type for 1 < p ≤ ∞.


Hint: Have a look at Theorems 15.2.1 and 15.2.4.

Exercise 2: 4 Points
Verify that log |x| ∈ BMO(R).
Hint: Using log(δx) = log(x) + log(δ) you can restrict to unit-balls.

Exercise 3:
A sequence (ϕj )j∈N0 ⊂ S (Rn ) which satisfies

1. ϕj = ϕ(2−j x) for j ∈ N.

2. supp ϕ0 ⊂ {x ∈ Rn : |x| ≤ 2} and supp ϕ ⊂ {x ∈ Rn : 1/2 ≤ |x| ≤ 2}.


P∞ n
3. j=0 ϕj (x) = 1 for every x ∈ R .

is called a dyadic decomposition of unity. Prove that


N
S′
X
ϕj ⇀ 1. in S ′ (Rn ) for N → ∞.
j=0

For f ∈ S ′ we have
N
S′
(ϕj fˆ)∨ ⇀ f
X
in S ′ (Rn ) for N → ∞.
j=0
Exercise 4: 4 Points
Prove that the system of Rademacher functions

rn (t) = sgn sin(2n πt), t ∈ [0, 1], n ∈ N0 ,

is an orthonormal system in L2 (0, 1), but that it is not a basis.


Hint: Take a look at the function f = χ(0,1/4] − χ(1/4,1/2] .

Exercise 5:
Show the following tail estimates for sums of i.i.d. (independent and identically dis-
tributed) random variables.
1. If {Xi }i=1,...,n is a collection of independent standard Gaussian variables, i.e.
with mean 0 and variance 1, it holds
n
1X  nt2
P Xi ≥ t ≤ exp(− ) (t > 0). (1)
n i=1 2

2. A Rademacher random variable takes values −1 and 1 with probability 1/2


each. For a collection of independent Rademacher variables {Xi }i=1,...,n the
estimate (1) also holds.
Pn  Pn 
Hint: Start with P i=1 Xi ≥ nt = P exp(λ i=1 Xi − λnt) ≥ 1 with some arbitrary
λ > 0. Then proceed as in Remark 15.8.2. Finally you can optimize the parameter λ.

Exercise* :
In the sequel M denotes the Hardy-Littlewood maximal operator. The variations
1
Z
′
M f (x) := sup |f (x − y)| dy, where Qr is a cube centered at x with side-length r,
r>0 |Qr | Qr
1
Z
′′
M f (x) := sup |f (y)| dy, where the Q’s are arbitrary cubes,
Q∋x |Q| Q

are pointwise equivalent to M. For M ′ this means that for all x ∈ Rn it holds

cM ′ f (x) ≤ Mf (x) ≤ CM ′ f (x),

where C, c are constants independent of x and f .


For a non-negative, measurable function ω : Rn → R and 1 ≤ p ≤ ∞ we define the
weighted Lp -spaces by Lp (ω) := {f : Rn → C measurable : kf kLp (ω) < ∞}, where
the norms are given by
Z 1/p
kf kLp (ω) = |f (x)|p ω(x) dx if 1 ≤ p < ∞
Rn Z
n o
and kf kL∞ (ω) = inf C ≥ 0 : ω(x) dx = 0 .
{x:|f (x)|≥C}
1. Show that for f ∈ L∞ (Mω) it holds kMf kL∞ (ω) ≤ kf kL∞ (M ω) .
Hint: Distinguish the cases M ω(x) = 0 a.e. and M ω(x) > 0 for all x.

2. Prove for f ∈ L1 (Mω) and all λ > 0

C1
Z Z
ω(x) dx ≤ |f (x)|Mω(x) dx,
{x:M f (x)>λ} λ Rn

where C1 is independent of λ and f .


Hint: You may assume that f is non-negative and integrable. By means of a Calderon-Zygmund
R
decomposition {Qj } of f at height λ > 0 you can then estimate {x:M ′ f (x)>4n λ} ω(x) dx . For
this estimation you need {x ∈ Rn : M ′ f (x) > 4n λ} ⊂ j 2Qj , and λ|2Qj | < 2n Qj f (x) dx.
S R

3. Finally let 1 < p < ∞ and deduce that there exists a constant Cp such that for
all f ∈ Lp (Mω)
Z Z
p
Mf (x) ω(x) dx ≤ Cp |f (x)|p Mω(x) dx.
Rn Rn

Hint: This follows from the Marcinkiewicz interpolation theorem.

Please, hand in the solutions of the Exercises with points (i.e. 2 and 4) on February 8, 2013
before the lecture.

You might also like