Professional Documents
Culture Documents
Richard F. Bass
February 2, 2010
-algebras
Example 1.6 Let X = [0, 1], and B1 , . . . , B8 subsets of X which are pairwise
disjoint and whose union is all of X. Let A be the collection of all finite unions
of the Bi s as well as the empty set. (So A consists of 28 elements.)
Note that if we take an intersection of -algebras, we get a -algebra; this is
just a matter of checking the definition.If we have a collection C of subsets of
X, there is at least one -algebra containing C, namely, the one consisting of
all subsets of X. We can take the intersection of all -algebras that contain
C; we denote this intersection by (C). If A is any -algebra containing C,
then A (C).
If X has some additional structure, say, it is a metric space, then we can
talk about open sets. If G is the collection of open subsets of X, then we
call (G) the Borel -algebra on X, and this is often denoted B. We will see
later that when X is the real line, that B is not equal to the collection of all
subsets of X.
We end this section with the following proposition.
2
(4) The proof of this comes from (3), using that (a, b] = (a, ) (b, )
and (a, ) =
n=1 (a, a + n].
Measures
(Ai ).
i=1
Example 2.2 X is any set, A is the collection of all subsets, and (A) is
the number of elements in A. This is called counting measure.
A
.
So
(A)
=
(B
)
(Ai ).
i
i=1 i
i=1 i
(c) Define the Bi as in (b). Since ni=1 Bi = ni=1 Ai , then
(A) =
(
i=1 Ai )
(
i=1 Bi )
(Bi )
i=1
= lim
n
X
i=1
Example 2.6 To see that (A1 ) < is necessary, let X be the positive
integers, counting measure, and Ai = {i, i + 1, . . .}. Then the Ai decrease,
(Ai ) = for all i, but (i Ai ) = () = 0.
Definition 2.7 A probability or probability measure is a measure such that
(X) = 1. In this case we usually write (, F, P) instead of (X, A, ).
a
).
If
A
R,
define
i
i=1 i
m (A) = inf{m(G) : G open, A G}.
(3.1)
We will prove these two facts (and a bit more) in a moment, but lets first
make some remarks.
A set N is a null set with respect to m if m (N ) = 0. Let L be the
smallest -algebra containing B and all the null sets. More precisely, let N
be the collection of all sets that are null sets with respect to m and let
L = (B N ). L is called the Lebesgue -algebra, and sets in L are called
Lebesgue measurable.
As part of our proof of (B) we will show that m is a measure on L.
Lebesgue measure is the measure m on L. (A) shows that L is strictly
smaller than the collection of all subsets of R.
It is easy to get lost in the construction of Lebesgue measure, so let us
summarize our steps.
First we prove (A), which Proposition 3.1.
We then turn to the construction of Lebesgue measure. It is more convenient for technical reasons to define
X
m (A) = inf{ (bi ai ) : A
i=1 (ai , bi ]}.
(3.2)
i=1
There is no real difference between this and (3.1) since it is clear that the
difference between an open set and a set of the form
i=1 (ai , bi ] is countable,
q[2,2]
m (A + q), and
I
and
i,j ij
i=1 i
X
X
X
X
m(Iij )
m (Ai ) +
/2i =
m (Ai ) + .
i,j
Since is arbitrary, m (
i=1 Ai )
i=1
m (Ai ).
n
X
i=1
m (E Ai ).
Since Bn A, then
m (E) = m (E Bn ) + m (E
Bnc )
n
X
m (E Ai ) + m (E B c ).
i=1
m (E Ai ) + m (E B c )
i=1
c
m (
i=1 (E Ai )) + m (E B )
= m (E B) + m(E B c )
m (E).
This shows B A.
If we set E = B in this last equation, we obtain
m (B) =
m (Ai ),
i=1
or m is countably additive on A.
If m (A) = 0 and E X, then
m (E A) + m (E Ac ) = m (E Ac ) m (E),
which shows A contains all null sets.
P
Define m(ni=1 (ai , bi ]) = ni=1 (bi ai ) if the (ai , bi ] are disjoint. Note m is
well-defined (a set might be expressible as a union of such intervals in more
than one way).
The last step in the construction is the following.
Proposition 3.6 Every set in the Borel -algebra is m - measurable.
Proof. Since the collection of m -measurable sets is a -algebra, it suffices
to show that every interval J of the form (a, b] is m -measurable. Let E be
any set with m (E) < ; we need to show
m (E) m (E J) + m (E J c ).
9
(3.4)
Since
E
I
,
we
have
m
(E
J)
m (Ii J) and m (E J c )
i
P
m (Ii J c ). Hence we have
X
m (E J) + m (E J c )
[m (Ii J) + m (Ii J c )].
i
m (Ii ) m (E) + .
Example 4.1 Recall the Cantor set is constructed by taking the interval
[0, 1], removing the middle third, removing the middle thirds of each of the
two remaining subintervals, and continuing. The Cantor set is what remains;
it is closed, uncountable, and every point is a limit point. Moreover, it
contains no intervals.
After one stage, the measure of the two intervals is 2( 31 ), after two stages
4(1/9), and after n stages, (2/3)n . Since the Cantor set C is the intersection
of all these sets, the Lebesgue measure of C is 0.
Suppose we define f0 to be 1/2 on the interval (1/3, 2/3), 1/4 on the interval (1/9, 2/9), 3/4 on the interval (7/9, 8/9), and so on. Define f (x) =
inf{f0 (y) : y x} for x < 1. Define f (1) = 1. Notice f = f0 on the complement of the Cantor set. f is monotone, so it has only jump discontinuities.
But if it has a jump continuity, there is a rational of the form k/2n with
10
Proof. (a) For each i, there is an open set Gi that contains A and such that
m(Gi A) < 2i . This follows from the fact that m(A) = m (A) and the
definition of m . Then Hi = ji Gj will contain A, is open, and since it is
contained in Gi , then m(Hi A) < 2i . Let H =
i=1 Hi . H need not be
open, but it is the intersection of countably many open sets. The set H is a
Borel set, contains A, and m(H A) m(Hi A) < 2i for each i, hence
m(H A) = 0.
(b) If A [0, 1], let Fi = [0, 1] Hi , where Hi is a decreasing sequence of
open sets containing Ac such that m(Hi Ac ) < 2i . (The Hi are constructed
as in the proof of (a), but in terms of Ac instead of A.) Then Fi is an
increasing sequence of closed sets, Fi A for each i, and m(A Fi ) < 2i
for each i. Our result follows from letting F = i Fi since m(A F )
m(A Fi ) < 2i for each i, hence m(A F ) = 0..
The countable intersections of open sets are sometimes called G sets; the
G is for geoffnet, the German word for open and the for Durchschnitt,
the German word for intersection. The countable unions of closed sets are
called F sets, the F coming from ferme, the French word for closed, and
the coming from Summe, the German word for union.
Therefore, when trying to understand Lebesgue measure, we can look at
G or F sets, which are not so bad, and at null sets, which can be quite bad
but dont have positive measure.
Next we prove the Caratheodory extension theorem. We say that a measure is -finite if there exist E1 , E2 , . . . , such that (Ei ) < for all i and
X
i=1 Ei .
Theorem 4.5 Suppose A0 is an algebra and m restricted to A0 is a measure.
Define
nX
o
m (E) = inf
m(Ai ) : Ai A0 , E
A
i=1 i .
i=1
Then
(a) m (A) = m(A) if A A0 ;
(b) every set in A0 is m -measurable;
12
m(Bi )
i=1
m(Ai ).
i=1
Bi A0 such that E
B
and
i=1 i
i m(Bi ) m (E) + . Then
m (E) +
m(Bi ) =
m(Bi A) +
m(Bi Ac )
i=1
i=1
i=1
m (E A) + m (E A ).
Since is arbitrary, m (E) m (EA)+m (EAc ). So A is m -measurable.
Finally, suppose we have two extensions to the smallest -algebra containing A0 ; let the other extension be called n. We will show that if E is in this
smallest -algebra, then m (E) = n(E).
Since m is -finite, we can reduce to the case where m is a finite measure:
if X = i Ki with m(Ki ) < and we prove uniqueness for the measure mi
defined by mi (A) = m(AKi ), then uniqueness for m follows. So we suppose
m(X) < .
Since E must be m -measurable,
m (E) = inf{
m(Ai ) : E
i=1 Ai , Ai A0 }.
i=1
But m = n on A0 , so i m(Ai ) =
which implies n(E) m (E).
n(Ai ),
P
Ai A0 such that m (E) + i m(Ai ) and E i Ai . Let A = i Ai and
Bk = ki=1 Ai . Observe m (E) + m (A), hence m (A E) < . We have
m (A) = lim m (Bk ) = lim n(Bk ) = n(A).
k
Then
m (E) m (A) = n(A) = n(E) + n(A E) n(E) + m(A E) n(E) + .
Since is arbitrary, this completes the proof.
Remarks: (1) Uniqueness implies there is only one possible Lebesgue measure.
(2) We will use the Caratheodory extension theorem in the study of product measures. It is also used in the Riesz representation theorem and in the
Daniell-Kolmogorov extension theorem.
We now define Lebesgue-Stieltjes measures. Let : R R be nondecreasing and right continuous (i.e., (x+) = (x) for all x, where (x+) =
limyx,y>x (y)). Suppose we define m ((a, b]) = (b) (a), define
X
m (
((bi ) (ai ))
i=1 (ai , bi ]) =
i
lim (x).
yx,y<x
Measurable functions
Proposition 5.6 If X is a metric space, A contains all the open sets, and
f is continuous, then f is measurable.
15
2
2
> a} {x : f (x) <
f is measurable since {x : f (x) > a) = {x : f (x)
a}. The measurability of f g follows since f g = 12 [(f + g)2 f 2 g 2 ].
{x : max(f (x), g(x)) > a} = {x : f (x) > a} {x : g(x) > a}, and the
argument for min(f, g) is similar.
i=1 {x : fi (x) > a}, and the proof for inf fi is similar.
Integration
17
s(x) =
ai Ei (x)
i=1
sn =
n2
X
i1
i=1
2n
Eni + nFn .
P
Definition 6.3 If s = ni=1 ai Ei is a nonnegative measurable simple function, define the Lebesgue integral of s to be
Z
s d =
n
X
ai (Ei ).
(6.1)
i=1
18
(6.2)
R
R
If f is measurable and at least one of the integrals f + d, f d is finite,
where f + = max(f, 0) and f = min(f, 0), define
Z
Z
Z
+
f d = f d f d.
(6.3)
Finally, if f = u + iv and
Z
(6.4)
R
R
Proof.
R ForRthe real case, this is easy. f |f |, so f |f |. Also f |f |,
so f |f |. Now combine these two facts.
R
For the complex case, Rf is a complex number. If it is 0, the inequality
is trivial. If it is not, then f = rei for some r and . Then
Z
Z
Z
i
f = ei f.
f = r = e
R
R
R
From the
definition
of
f
when
f
is
complex,
we
have
Re
(
f
)
=
Re (f ).
R
Since | f | is real, we have
Z
Z
Z
Z
i
i
e f = Re (e f ) |f |.
f = Re
R
R
R
We do not yet have that (f + g) = f + g.
Limit theorems
20
fn
fn c
An
m
X
Z
=c
=c
sn
An
ai Ei
An i=1
m
X
ai (Ei An ).
i=1
m
X
Z
ai (Ei ) = c
s.
i=1
s.
R
Example 7.2 Let X = [0, ) and
f
(x)
=
1/n
for
all
x.
Then
fn =
n
R
, but fn f where f = 0 and f = 0. The problem here is that the fn
are not nonnegative.
Example
7.3 Suppose fn = nR(0,1/n) . Then fn 0, fn 0 for each x, but
R
fn = 1 does not converge to 0 = 0. The trouble here is that the fn do
not increase for each x.
Once we have the monotone convergence theorem, we can prove that the
Lebesgue integral is linear.
Theorem 7.4 If f1 and f2 are integrable, then
Z
Z
Z
(f1 + f2 ) = f1 + f2 .
Proof. First suppose f1 and f2 are nonnegative and simple. Then it is clear
from the definition that the theorem holds in this case. Next suppose f1
21
fn =
n=1
Proof. Let FN =
PN
n=1
Z
X
fn .
n=1
fn and write
Z X
N
X
Z
fn =
lim
n=1
fn
n=1
Z
=
lim FN = lim
= lim
N Z
X
fn =
FN
Z
X
(7.1)
fn ,
n=1
n=1
using the monotone convergence theorem and the linearity of the integral.
22
R
If we take the limit as n , on the left hand side we obtain lim inf fn by
the monotone
convergence theorem, while on the right hand side we obtain
R
lim inf n fn .
A typical
R use of Fatous lemma is the following. Suppose we have fRn f
and supn |fn | K < . Then |fn | |f |, and by Fatous lemma, |f |
K.
Another very important theorem is the dominated convergence theorem.
Theorem 7.7 Suppose fn are measurable functions and fn (x) f (x). Suppose there
exists Ran integrable function g such that |fn (x)| g(x) for all x.
R
Then fn d f d.
Proof. Since fn + g 0, by Fatous lemma,
Z
Z
(f + g) lim inf (fn + g).
Since g is integrable,
Z
Z
f lim inf
Similarly, g fn 0, so
Z
fn .
Z
(g f ) lim inf
and hence
Therefore
(g fn ),
Z
f lim inf
Z
(fn ) = lim sup
fn .
Z
f lim sup
fn ,
23
If in the monotone convergence theorem or dominated convergence theorem we have only fn (x) f (x) almost everywhere, the conclusion still
holds. For if the fn and f are measurable and A = {x : fn (x) f (x)}, then
f A f ARfor each Rx. And since Ac has measure 0, we see from Proposition
6.5(d) that f A = f , and similarly with f replaced by fn .
f d = 0.
f
and
s
/2.
m
R
Rm
If we find g such that |sm g| < /2, then |f g| < . So it suffices to
consider the case where f is a simple function.
R
P
If fP= pi=1 ai Ai and we find gi continuous such that |ai Ai gi | < /p,
then pi=1 gi will be the desired function. So we may assume f is a constant
times a characteristic function, and by linearity, we may assume f is equal
to A for some A contained in a bounded interval [n, n].
We can choose G open and F closed such that F A G and m(GF ) <
. We can replace G by G (n 1, n + 1). Gc [n 1, n + 1] and F
are compact sets, so there is a minimum distance between them, say, . Let
g(x) = max(0, 1 dist (x, F )/). Then g is continuous, 0 g 1, g is 1 on
F , g is 0 on Gc , and g is 0 outside of [n 1, n + 1]. Therefore
|g A | G F ,
so
Z
|g A |
(G F ) = m(G F ) < .
The method of proof, where one proves a result for characteristic functions, then simple functions, then non-negative functions, and then finally
integrable functions is very common.
25
i=1
and
L(P, f ) =
n
X
inf
xi=1 xx
i=1
Set R(f ) = inf{U (P, f ) : P is a partition} similarly R(f ). Then the Riemann integral exists if R(f ) = R(f ), and the common value is the Riemann
integral, which we denote R(f ).
Theorem 8.4 A bounded measurable function f on [a, b] is Riemann integrable if and only if the set of points at which f is discontinuous has Lebesgue
measure 0, and in that case, the Riemann integral is equal in value to the
Lebesgue integral.
Proof. If P is a partition, define
TP (x) =
n
X
i=1
and
SP (x) =
n
X
i=1
We see that
sup
xi1 yxi
inf
xi1 yxi
TP = U (P, f ) and
SP = L(P, f ).
We conclude T = S = f a.e. If x is not in the null set where T (x) 6= S(x) nor
in i Pi , which is countable and hence of Lebesgue measure 0, then TPi (x)
f (x) and SPi (x) f (x). This implies that f is continuous at such f . Since
Z
Z
R(f ) = lim U (Pi , f ) = lim TPi = f,
i
Z
L(Pi , f ) =
SPi
f.
Modes of convergence
A =
k=1 j=k Aj
has measure less than j=k Aj for every k, hence less than 2k+1 for every
k. Therefore A has measure 0. If x
/ A, then x
/
j=k Aj for some k, so
|fnj (x) f (x)| 1/j for j k, which means fnj f a.e. on Ac .
Example 9.3 Part (a) of the above proposition is not true if (X) = .
Let X = R and let fn = (n,n+1) .
Example 9.4 For an example where fn f in measure but not almost
everywhere, let X = [0, 1],
and let fn (x) =
Pn+1
Pnlet be Lebesgue measure,
Fn (x), where Fn = {y : ( j=1 1/j)( mod 1) y ( j=1 1/j)( mod 1)}.
z( mod 1) is defined as the fractional part of z (where the largest integer less
than z is subtracted from z). Let f (x) = 0 for all x.
Then (Fn ) 1/n 0, so fn f in measure. But any x will be in
infinitely many Fn s, so fn does not converge to f (x) at any point.
The following is known as Egoroffs theorem.
28
10
Product measures
If A1 A2 and A =
i=1 Ai , we write Ai A. If A1 A2 and
A =
A
,
we
write
A
A.
i
i=1 i
Definition 10.1 M is a monotone class is M is a collection of subsets of
X such that
(a) if Ai A and each Ai M, then A M;
(b) if Ai A and each Ai M, then A M.
The intersection of monotone classes is a monotone class, and the intersection of all monotone classes containing a given collection of sets is the
smallest monotone class containing that collection.
The next theorem, the monotone class lemma, is rather technical, but very
useful.
Theorem 10.2 Suppose A0 is a algebra, A is the smallest -algebra containing A0 , and M is the smallest monotone class containing A0 . Then
M = A.
29
(
i=1 Ai ) B = i=1 (Ai B), and similarly for intersections. Therefore
N2 = M; in other words, if B A0 and A M, then A B M.
Let N3 = {A M : A B M for all B M}. As in the preceding
paragraph, N3 is a monotone class contained in M. By the last sentence of
the preceding paragraph, N3 contains A0 . Hence N3 = M.
We thus have that M is a monotone class closed under the operations of
taking complements and taking intersections. This shows M is a -algebra,
and so A M.
Suppose (X, A, ) and (Y, B, ) are two measure spaces, i.e., A and B are
-algebras on X and Y , resp., and and are measures on A and B, resp.
A rectangle is a set of the form A B, where A A and B B. Define a
set function on rectangles by
(A B) = (A)(B).
Lemma 10.3 Suppose AB =
i=1 Ai Bi , where A, Ai A and B, Bi B
and the Ai Bi are disjoint. Then
(A B) =
(Ai Bi ).
i=1
Proof. We have
AB (x, y) =
Ai Bi (x, y),
i=1
and so
A (x)B (y) =
X
i=1
30
Ai (x)Bi (y).
Holding x fixed and integrating over y with respect to , we have, using (7.1),
A (x)(B) =
Ai (x)(Bi ).
i=1
Now use (7.1) again and integrate over x with respect to to obtain the
result.
Let C0 = {finite unions of rectangles}. It is clear that C0 is an algebra. By
Lemma 10.3 and linearity, we see that is a measure on C0 . Let A B
be the smallest -algebra containing C0 ; this is called the product -algebra.
By the Caratheodory extension theorem, can be extended to a measure
on A B.
We will need the following observation. Suppose a measure is -finite.
So there exist Ei which have finite measure and whose union is X. If we
let Fn = ni=1 Ei , then Fi X and (Fn ) is finite for each n.
If and are both -finite, say with Fi X and Gi Y , then will
be -finite, using the sets Fi Gi .
The main result of this section is Fubinis theorem, which allows one to
interchange the order of integration.
Theorem 10.4 Suppose
R f : X Y R is measurable with respect to AB.
If f is nonnegative or |f (x, y)| d( )(x, y) < , then
(a) the function g(x) =
(c) we have
Z
Z Z
f (x, y) d(x) d(y)
Z Z
=
f (x, y) d(y) (dx).
f (x, y) d( )(x, y) =
31
11
then is a measure. The only part that needs thought is the countable additivity, and this follows from (7.1) applied to the functions f Ai . Moreover,
(A) is zero whenever (A) is. We sometimes write f = d/d for (11.1).
Definition 11.1 A measure is called absolutely continuous with respect
to a measure if (A) = 0 whenever (A) = 0. This is frequently written
.
32
2k = 0,
(F ) = lim (k=n Ek ) lim
n
k=n
but
(F ) = lim (
k=n Ek ) ;
n
Proof. If E is a negative set, we are done. If not, there exists a subset with
positive measure. Let n1 be the smallest positive integer such that there
exists E1 E with (E1 ) 1/n. Let k 2. If Fk = E (E1 Ek1 )
is negative, we are done. If not, let nk be the smallest positive integer such
that there exists Ek Fk with (Ek ) 1/nk . We continue.
If the construction stops after a finite number of sets, we are done. If not,
let F = k Fk = E (k Ek ). Since 0 > (E) > and (Ek ) 0, then
(E) = (F ) +
(Ek ).
k=1
(ni=1 Bi ))
= lim
n
X
(C Bi ) 0.
i=1
So E is a negative set.
Since E is negative,
(E) = (An ) + (E An ) (An ).
Letting n , we obtain (E) = L.
34
If = + is another such decomposition with + , mutually singular, let E 0 and F 0 be the sets in the definition of mutually singular. Then
X = E 0 F 0 gives another Hahn decomposition, hence EE 0 is a null set
with respect to . Then for any A A,
+ (A) = (A E 0 ) = (A E) = + (A),
and similarly for , .
The measure + + is called the total variation measure and is written
||.
We now are ready for the Radon-Nikodym theorem.
Theorem 11.11 Suppose is a -finite measure and is a finite measure
such that is absolutely continuous with respect toR. There exists a integrable nonnegative function f such that (A) = A f d for all A A.
Moreover, if g is another such function, then f = g almost everywhere with
respect to .
Proof. Let us first prove the uniqueness assertion. For every set A we have
Z
(f g) d = (A) (A) = 0.
A
Z
n
o
F = g : 0 g, g d (A) for all A A .
A
R
F is not empty because 0
R F. Let L = sup{ g d : g F}, and let gn be
a sequence in F such that gn d L. Let hn = max(g1 , . . . , gn ).
36
AB
(A B) + (A B c )
= (A).
By an induction argument, hn is in F.
The hn increase, say to f . By monotone convergence
Z
f d (A)
f d = L and
(11.2)
for all A.
Let A be a set where there is strict inequality in (11.2); let be chosen
sufficiently small so that if is defined by
Z
(B) = (B)
f d (B),
B
then (A) > 0. is a signed measure; let F be the positive set as constructed
in Theorem 11.7. In particular, (F ) > 0. So for every B
Z
f d + (B F ) (B F ).
BF
BF
(B F c ) + (B F ) = (B).
This says that f + F F. However,
Z
Z
L (f + F ) d = f d + (F ) = L + (F ),
37
which implies (F ) = 0. But then (F ) = 0, and hence (F ) = 0, contradicting the fact that F is a positive set for F with (F ) > 0.
(A E) (A)
Z
= (A)
f d.
A
12
contained
in O, then
P
i=N +1 m(Ii ) < /5.
i=1
Let R = E N
i=1 Ii ; we will show m(R) < . Let Jn be the interval
with the same center as In but five times the length. Let x R. There
exists
an interval I PO containing
P
P x with I disjoint from I1 , . . . , IN . Since
m(In ) < , then
an 2 m(In ) < , and an 0. So I must either
be one of the In for some n > N or at least intersect it, for otherwise we
would have chosen I at some stage. Let n be the smallest integer such that
I intersects In ; note n > N . We have m(I) an1 2m(In ). Since x is in
I and I intersects In , the distance from x to the midpoint of In is at most
m(I) + m(In )/2 (5/2)m(In ). Therefore x Jn .
P
P
Then R
i=N +1 Jn , so m(R)
i=N +1 m(Jn ) = 5
i=N +1 m(In ) < .
f (x + h) f (x)
,
h
h0+
f (x + h) f (x)
D+ f (x) = lim inf
,
h0+
h
If all the derivates are equal, we say that f is differentiable at x and define
f 0 (x) to be the common value.
Theorem 12.2 Suppose f is nondecreasing on [a, b]. Then f is differentiable
Rb
almost everywhere, f 0 is integrable, and a f 0 (x) dx f (b) f (a).
Proof. We will show that the set where any two derivates are unequal has
measure zero. We consider the set E where D+ f (x) > D f (X), the other
sets being similar. Let Eu,v = {x : D+ f (x) > u > v > D f (x)}. If we show
m(Eu,v ) = 0, then taking the union of all pairs of rationals with u > v shows
m(E) = 0.
Let s = m(Eu,v ), let > 0, and choose an open set O such that Eu,v O
and m(O) < s + . For each x Eu,v there exists an arbitrarily small interval
[x h, x] contained in O such that f (x) f (x h) < vh. Use Lemma 12.1
to choose I1 , . . . , In which are disjoint and whose interiors cover a subset A
40
n=1
n
X
n=1
i=1
[f (xn ) f (xn hn )]
n=1
M
X
i=1
and so v(s + ) > u(s 2). This is true for each , so vs us. Since u > v,
this implies s = 0.
This shows that
f (x + h) f (x)
h0
h
is defined almost everywhere and that f is differentiable wherever g is finite.
Define f (x) = f (b) if x b. Let gn (x) = n[f (x + 1/n) f (x)]. Then
gn (x) g(x) for almost all x, and so g is measurable. Since f is increasing,
gn 0. By Fatous lemma
Z b
Z b
Z b
[f (x + 1/n) f (x)]dx
g lim inf
gn = lim inf n
a
a
a
Z a+1/n i
Z a+1/n i
h
h Z b+1/n
= lim inf n
f n
f = lim inf f (b) n
f
g(x) = lim
f (b) f (a).
41
For the last inequality, we use the fact that f is increasing. This shows that
g is integrable and hence finite almost everywhere.
P
A function is of bounded variation if sup{ ki=1 |f (xi ) f (xi1 )|} is finite,
where the supremum is over all partitions a = x0 < x1 < < xk = b of
[a, b].
Lemma 12.3 If f is of bounded variation on [a, b], then f can be written as
the difference of two nondecreasing functions on [a, b].
Proof. Define
P (y) = sup
k
nX
o
[f (xi )f (xi1 )] ,
+
N (y) = sup
k
nX
i=1
o
[f (xi )f (xi1 )] ,
i=1
where the supremum is over all partitions a = x0 < x1 < < xk = y for
y [a, b]. P and N are measurable since they are both increasing. Since
k
X
i=1
[f (xi ) f (xi1 )] =
k
X
i=1
42
i=1
xi1
i=1
xi1
On
F (x + 1/n) F (x)
,
1/n
then
Z
x+1/n
f (t) dt,
fn (x) = n
x
43
Rc
using the fact that F is continuous. So a [F 0 (x) f (x)] dx = 0 for all c,
which implies F 0 = f a.e. by Lemma 12.5.
Rb
Rb
a.e. Since n is arbitrary, F 0 f a.e. So a F 0 a f = F (b) F (a). On
Rb
Rb
the other hand, by Theorem 12.2, a F 0 (x) dx F (b) F (a) = a f . We
Rb
conclude that a [F 0 f ] = 0; since F 0 f 0, this tells us that F 0 = f a.e.
Rb
a
A function
is absolutely continuous on [a, b] if given there exists such
Pk
0
that
whenever {xi , x0i )} is a finite collection of
i=1 |f (xi ) f (xi )| <
Pk
nonoverlapping intervals with i=1 |x0i xi | < .
It is easy to see that absolutely continuous functions are continuous and
that the Cantor-Lebesgue function is not absolutely continuous.
44
Rx
a
Rb
Proof. Let > 0. Choose a simple function s such that a |f s| < /2.
Let K be a bound for |s| and let = /2K. If {(xi , x0i )} is a collection of
nonoverlapping intervals, the
is less than , then set
R sum of whose lengths
R
A = ki=1 (xi , x0i ) and note A |f s| < /2 and A s < K = /2.
Lemma 12.9 If f is absolutely continuous, then it is of bounded variation.
Proof. Let correspond to = 1 in the definition of absolute continuity.
Given a partition, add points if necessary so that each subinterval has length
at most . We can then group the subintervals into at most K collections,
each of total length less than , where K is an integer larger than (1+ba)/.
So the total variation is then less than K.
45
13
Lp spaces
We assume throughout this section that the measure is -finite. For 1 p <
, define the Lp norm of f by
Z
1/p
kf kp =
|f (x)|p d
.
For p = , define the L norm of f by
kf k = inf{M : ({x : |f (x)| M }) = 0}.
For 1 p the space Lp is the set {f : kf kp < }.
The L norm of a function f is the supremum of f provided we disregard
sets of measure 0.
It is clear that kf kp = 0 if and only if f = 0 a.e.
Proposition 13.1 (Holders inequality) If 1 < p, q < and p1 + q 1 = 1,
then
Z
|f (x)g(x)| d kf kp kgkq .
This also holds if p = and g = 1.
46
R
R
Proof. If M = kf k , then f g M |g| and the case p = and q = 1
follows.
So let us assume 1 < p, q < . If kf kp = 0, then f = 0 a.e and
R
f g = 0, so the result is clear if kf kp = 0 and similarly if kgkq = 0. Let
F (x) = |f (x)|/kf kp and G(x)
R = |g(x)|/kgkq . Note kF kp = 1 and kGkq = 1,
and it suffices to show that F G 1.
The second derivative of the function ex is again ex , which is positive, and
so ex is convex. Therefore if 0 1, we have
ea+(1)b ea + (1 )eb .
If F (x), G(x) 6= 0, let a = p log F (x), b = q log G(x), = 1/p, and 1 =
1/q. We then obtain
F (x)G(x)
F (x)p G(x)q
+
.
p
q
x 1.
(13.1)
x 1.
kf kpp
.
ap
kfn f kpp
0.
p
49
m=j
Am ) = lim
(
m=j Am )
lim
(Am ) = 0.
m=j
So except for a set of measure 0, for each x there is a last j for which
x
m=j Am , hence a last j for which x Aj . So for each x (except
for the null set) there is a j0 (depending on x) such that if j j0 , then
|fnj (x) fnj1 (x)| 2j .
Set
gj (x) =
j
X
m=1
gj (x) increases for each x, and the limit is finite for almost every x by the
preceding paragraph. Let us call the limit g(x). We have
kgj kp
j
X
2j + kfn1 kp 2 + kfn1 kp
m=1
Suppose x is not in the null set where g(x) is infinite. Since |fnj (x)fnk (x)|
|gnj (x) gnk (x)| 0 as j, k , then fnj (x) is a Cauchy series (in R), and
hence converges, say to f (x). We have kf fnj kp = limm kfnm fnj kp ;
this follows by dominated convergence with the function g defined above as
the dominating function.
We have thus shown that kf fnj kp 0. Given = 2(j+1) , if m nj ,
then kf fm kp kf fnj kp + kfm fnj kp . This shows that fm converges
to f in Lp norm.
Next we show:
Proposition 13.7 The set of continuous functions with compact support is
dense in Lp (Rd ).
50
R
Proof. Suppose f Lp . By dominated convergence |f f [n,n] |p 0
as n , the dominating function being |f |p . So we may suppose f has
compact support. By writing f = f + f we Rmay suppose f 0. By taking
simple functions sm increasing to f , we have |f sm |p 0 by dominated
convergence, so it suffices to consider simple functions. By linearity, it suffices
to consider characteristic functions with compact support. Given such a E
and > 0 we showed in Proposition 8.3 that there exists
R g continuous with
compact support and
with
values
in
[0,
1]
such
that
|g E | < . Since
R
R
p
|g E | 1, then |g E | |g E | < .
The following is very useful.
Proposition 13.8 For 1 < p < and p1 + q 1 = 1,
nZ
o
kf kp = sup
f g : kgkq 1 .
(13.2)
R
R
convergence. If |f |p = , then |sn |p by monotone convergence.
For n sufficiently large, ksn kp > 0.
Let
gn (x) = (sgn f (x))
|sn (x)|p1
p/q
ksn kp
R
( |sn |(p1)q )1/q
p/q
ksn kp
p/q
ksn kp
p/q
ksn kp
= 1.
= ksn kp(p/q)
.
f gn =
p
p/q
p/q
ksn kp
ksn kp
R
Since p (p/q) = 1, then f gn ksn kp , which tends to kf kp .
The above proof also establishes
Corollary 13.9 For 1 < p < and p1 + q 1 = 1,
nZ
o
kf kp = sup
f g : kgkq 1, g simple .
The space Lp is a normed linear space. We can thus talk about its dual,
namely, the set of bounded linear functionals on Lp . The dual of a space Y
is denoted Y . If H is a bounded linear functional on Lp , we define the norm
of H to be kHk = sup{H(f ) : kf kp 1}.
Theorem 13.10 If 1 < p < and p1 + q 1 = 1, then (Lp ) = Lq .
What this means is that if H is a bounded
linear functional on Lp , then
R
q
q
Rthere exists g L such that H(f ) = fpg and that if g L , then H(f ) =
f g is a bounded linear functional on L .
R
Proof. If g Lq , then setting H(f ) = f g for f Lp yields a bounded linear functional; the boundedness follows from Holders inequality. Moreover,
from Holders inequality and Proposition 13.8 we see that kHk = kgkq .
52
53
14
Fourier transforms
eiux f (x)dx,
u Rn .
(14.1)
Rn
We are using u x for the standard inner product in Rn . Various books have
slightly different definitions. Some put a negative sign before the iu x, some
have a 2 either in front of the integral or in the exponent. The basic theory
is the same in any case.
Some basic properties of the Fourier transform are given by
Proposition 14.1 Suppose f and g are in L1 . Then
(a) fb is bounded and continuous;
d)(u) = afb(u);
(b) (f\
+ g)(u) = fb(u) + gb(u); (af
(c) if fa (x) = f (x + a), then fba (u) = eiua fb(u);
(d) if ga (x) = eiax g(x), then gba (u) = fb(u + a);
(e) if ha (x) = f (ax), then b
ha (u) = an fb(u/a).
Proof. (a) fb is bounded because f L1 and |eiux | = 1. We have
Z
i(u+h)x
iux
b
b
f (u + h) f (u) =
e
e
f (x)dx.
So
|fb(u + h) fb(u)|
Z
iux ihx
1 |f (x)|dx.
e e
One reason for the usefulness of Fourier transforms is that they relate
derivatives and multiplication.
Proposition 14.2 Suppose f L1 and xj f (x) L1 , where xj is the j th
coordinate of x. Then
Z
fb
(u) = i eiux xj f (x)dx.
uj
Proof. Let ej be the unit vector in the j th direction. Then
Z
fb(u + hej ) fb(u)
1 i(u+hej )x
=
e
eiux f (x)dx
h
h
Z
eihxj 1
= eiux
f (x)dx.
h
Since
1
eihxj 1 |xj |
h
R
and xj f (x) L1 , the right hand side converges to eiux ixj f (x)dx by dominated convergence. Therefore the left hand side converges. Of course, the
limit is fb/uj .
55
f (y)g(xy)dy, so f g = g f .
e
Z Z
=
Z
=
iux
Z
f (x y)g(y)dy dx
We applied Fubini in the first equality; this is valid because as we saw in (a),
the absolute value of the integrand is integrable.
We want to give a formula for recovering f from fb. First we need to
calculate the Fourier transform of a particular function.
56
1
2
e|x| /2 .
n/2
(2)
2
Then fbn (u) = e|u| /2 .
g 0 (u)
= u,
g(u)
g(u) = c2 eu .
(14.2)
R
2
Since g(0) = ex /2 dx = 2, c
2. Substituting this value of c2 in
2 =
(14.2) and dividing both sides by 2 proves (a).
57
(a) Then kf A f k1 0 as A 0.
(b) If f is continuous with compact support, then f A converges to f
pointwise.
Proof. (a) Let > 0. Choose g continuous with compact support so that
kf gk1 < . Let h = f g. A change of variables shows that kA k1 = kk1 .
Observe
kf A f k1 kg A gk1 + kh A hk1
and
kh A hk1 khk1 + kh A k1 khk1 + khk1 kA k1 < (1 + kk1 ).
So since is arbitrary, it suffices to show that g A g in L1 .
We start by writing
Z
Z
g A (x) g(x) = g(x y)A (y)dy g(x) = g(x Ay)(y)dy g(x)
Z
= [g(x Ay) g(x)](y)dy.
R
We used a change of variables and the fact that (y)dy = 1. Because g
is continuous with compact support, then g is bounded, and the integral on
the right goes to 0 by dominated convergence, the dominating function being
kgk |(y)|. Therefore g A (x) converges to g(x) pointwise.
58
Z
GA (y)|(y)|dy
a.e.
2 u2 /2
. Therefore if we let
Ha (x) =
1 |x|2 /2a2
e
,
(2)n
59
we have
b a (u) = (2)n/2 an ea2 |u|2 /2 .
H
We have
Z
fb(u)eiuy Ha (u)du
Z Z
=
eiux f (x)eiuy Ha (u)dx du
Z Z
=
eiu(xy) Ha (u)du f (x) dx
Z
b a (x y)f (x)dx.
= H
(14.3)
60
iux
Z
f (x)dx =
eiux f (x)dx
Z
=
f[
g(u)Ha (u)du.
(14.6)
Since f[
g(u) = fb(u)b
g (u) = |fb(u)|2 , the right hand side of (14.6) converges
R
by monotone convergence to (2)n |fb(u)|2 du as a . Since f and g
are continuous with compact support, then it is easy to see Rthat f g is also,
and
left hand side of (14.6) converges to f g(0) = f (y)g(y)dy =
R so the
|f (y)|2 dy by Proposition 14.5(b).
(b) The set of continuous functions with compact support is dense in L2 .
Given a function f in L2 , choose a sequence of continuous functions with
compact support {fm } such that fm f in L2 . By the result in (a), {fbm } is
a Cauchy sequence in L2 , and therefore converges to a function in L2 , which
0
we call fb. If {fm
} is another sequence of continuous functions with compact
0
support converging to f in L2 , then {fm fm
} is a sequence of continuous
functions with compact support converging to 0 in L2 ; by the result in (a),
0
fbm fbm
converges to 0 in L2 , and therefore fb is defined uniquely up to almost
everywhere equivalence. By passing to the limit in L2 on both sides of (14.5),
we see that (14.5) holds for f L2 .
Richard F. Bass
Department of Mathematics
University of Connecticut
Storrs, CT 06269-3009, USA
bass@math.uconn.edu
61