Scribd
Upload
70 views

Lecture Notes in Measure Theory

The document summarizes two key theorems: 1. The Polar Decomposition Theorem states that the Radon-Nikodym derivative of a positive measure λ is equal to the absolute value of the Radon-Nikodym derivative of λ. 2. The Hahn Decomposition Theorem states that for any real measure μ on a measurable space (X,A), there exists a partition of X into two sets A and B such that μ restricts to a positive measure on A and a negative measure on B. Together these theorems allow the Jordan decomposition of a real measure μ into its positive and negative parts μ+ and μ-.

Uploaded by

primary111
Copyright
© Attribution Non-Commercial (BY-NC)
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
70 views

Lecture Notes in Measure Theory

The document summarizes two key theorems: 1. The Polar Decomposition Theorem states that the Radon-Nikodym derivative of a positive measure λ is equal to the absolute value of the Radon-Nikodym derivative of λ. 2. The Hahn Decomposition Theorem states that for any real measure μ on a measurable space (X,A), there exists a partition of X into two sets A and B such that μ restricts to a positive measure on A and a negative measure on B. Together these theorems allow the Jordan decomposition of a real measure μ into its positive and negative parts μ+ and μ-.

Uploaded by

primary111
Copyright
© Attribution Non-Commercial (BY-NC)
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/ 7

April 11, 2012 -continue consequence of Radon-Nikodym Theorem: Polar Decomposition Theorem (6.

13): Choose (X, A) , a positive measure and set g L1 (X, ). Let


(E ) =
E

gd E A d = gd where g is the Radon-Nikodym derivative of .

Then:
||(E ) =

|g |d,
E

in particular
||(x) = g
L1 (X,)

Proof: Using the polar decomposition of we have


d = hd(||) with |h| = 1.

By assumption set
h(d||) = d = gd

= |h|2 d(||) = h hh

d(||) = hgd, i.e. ||(E ) =


E

hgd, E A.

0. Why ? Hence hg 0 for a.e. Since || and are positive measure's, then hg | = 1, then Since |h = |hg | = |h ||g | = |g | for a.e. hg

d(||) = |g |d.

Hahn Decomposition Theorem (6.14): Choose (X, A) and be a real measure on A.


Then:
A, B A with A B = A B = X such that

E A, + (E ) = (E A)

(E ) = (E B ),

i.e. A carries the positive mass of (+ (E )) and B carries the negative mass of ( (E )) . Rmk: we will use = + . Proof: By the Polar Decomposition for we have
d = hd(||) |h| = 1.

Since is real, we have that h is real h = 1. Let


A = {x X : h(x) = 1} B = {x X : h(x) = 1}.

Rmk: It is clear that A and B partition X . Since


+ =
def

1 (|| + ) , 2

we have, for E A, that


+ (E ) =
def

def

1 (||(E ) + (E )) 2 1 ||(E ) + hd(||) 2 E 1 d(||) + hd(||) , 2 E E (1 + h) d||.


E

1 (E ) = 2
+

Since

1+h = 2 0 x B

1 x A

+ (E ) =
E A

d|| = ||(E A).

Since A, B partition E we have


(E ) = (E A) + (E B )

and
(E ) = + (E ) (E ),

(E ) = (E B ).

If = 1 2 with 1 , 2 positive measure's, Then:

Corollary (Minimum Property of the Jordan Decomposition) : Choose (X, A) and real.

+ 1 2 .

Proof: For all E A


(E ) = 1 (E ) 2 (E ) 1 (E )

where the last inequality is by the assumption on 1 . By the Hahn Decomposition Theorem
+ (E ) = (E A) 1 (E A) 1 (E )

where the rst inequality is by the assumption on 1 and the last inequality is by monotone property since E A E. Hence
+ 1 .

By denition
= + = + 1 + 2

where the last equality is by the Hahn Decomposition,


(E ) = + (E ) 1 (E ) + 2 (E ) 2 (E ).

It is clear that + (E ) 1 (E ) < 0. Hence


(E ) 2 (E ) i.e. 2 .

Rmk: try using (E ) = (E B ). (6.16) Chapter 6 stu for project's (Ahmad and Jorge): Choose (X, A, ) such that is a nite positive measure. p Choose 1 p < + and q = (p 1) , i.e. p and q are conjugates.

Denition: A bounded linear functional is a map


linear

: Lp (X, ) R (or C) where

= sup f
f Lp
f p 1

such that: TFAE (Theorem): 1.) is continuous at 0. 2.) is continuous everywhere. 3.) < +.

Theorem (6.16): Let be a Bounded Linear Functional on Lp (X, ).


Then:
! g Lq (X, ), that represents , i.e. < , f > =
X

f gd, f Lp .

Moreover:
= g
q

Rmk: < , f > is the duality between and f . Rmk: The RHS is well dened by H older's Ineq., i.e. f, g Lp f g L1 . and choose Cc (X ). Then: let be a bounded linear functional on Cc (X ) (or on C0 (X )).
f Cc (C ), < , f >=
X

Riesz Representation Theorem (6.18): Choose X a l.c.h.t.s-locally compact Hausdor topological space

Then: There exists a unique complex Borel measure on X that represents , i.e.
f d.

Moreover:
= ||(X ).

End Chapter 6

Chapter 7Rmk: End of Chapter 7 cover's change of variables in multiple integrals (When Dr R. was a student no one ever covered this part in his classes because his professor's always told him that he would see it in another class. This never happened, so Dr R. had to go over it by himself. We may or may not cover it.

Derivatives of Measures: Theorem (7.1): Choose


a complex Borel measure on R and dene for all x R : f (x) = (, x), which is well dened since 1.) interval is open. 2.) is a complex Borel measure. 3.) ? Let x R and A R. TFAE: 1.) f is dierentiable at x and f (x) = A.

2.) For every > 0, there is a > 0 such that for every open interval I
(I ) A < . m(I )

x with m(I ) < , we have

Rmk: set m(I ) to be the Lebesgue Measure of I . WWTS(we want to show): 1.) 2.) Proof: WLOG assume A = 0. Why ? Instead of we can consider the restriction of ( Am ) on some nite open interval that contains x. Why the restriction ? Since the Lebesgue measure is NOT nite but , the complex Borel measure, is nite. 1.) 2.): Assume f (X ) = 0. Then:
> 0, > 0 t where |t x| < we have

|f (t) f (x)| < |t x|, i.e.

|f (t) f (x)| < . |t x|

Let I = (s, t) x with m(I ) = t s < . Let sn be a sequence such that x > s1 > s2 > > sn

s,

| ([sn , t])

=
def

(, t) (, sn ) f (t) f (sn ) f (t) f (x) + f (x) f (sn ) |t x| + |x sn | |t sn | m(I ).

tri. ineq.

def

<

=
def

<

Since
I = (s, t) =

[sn , t),
n=1

then by Lebesgue's Dominated Convergence (use this since is complex even though sn is monotone), we have
(s, t) < m(I ). (I )

(2) (1) Given > 0, choose > 0 such that for every open interval I
(I ) A < . m(I )

x with m(I ) < , we have

Let s < x < t such that t s < t < s. Let t < s1 < s2 < < sn s. By assumption
(sn t) |t sn |.

Since every demi-set is F and G we can write

[s, t) =

(sn , t).
n=1

Then by Lebesgue's Dominated Convergence


f (t) f (s) = [s, t] t s

where s < x < t < x + ,

|f (t) f (x)| |t x|.

So how do we conclude (i.e. get rid of ? over the ineq.) ? Since f is continuous at x i {x} = 0 (why ?) By 2.) we have that {x} = 0, i.e. A = 0 in the strict inequality
(I ) A < . m(I )

Hence f is dierentiable at x. Rmk: now we see why we can choose A = 0 WLOG.

You might also like