Académique Documents
Professionnel Documents
Culture Documents
12.1 -Algebras
The Axiom of Choice implies that there is no way to create a function
defined on every subset of R so that all of the following hold:
(i) 0 (E) for every E R,
(ii) ([a, b]) = b a,
(iii) if E1 , E2 , . . P
. are finitely or countably many disjoint subsets of R, then
(k Ek ) = k (Ek ),
(iv) (E + h) = (E) for all E R and h R.
c
2012
Christopher Heil
425
426
E = E C = R\E .
427
Note that if 1 , 2 are -algebras, then 1 2 is not formed by intersecting the elements of 1 with those of 2 . Rather, it is the collection
of all sets that are common to both 1 and 2 . Thus, if is a -algebra
that contains E then (E) , which explains why (E) is the smallest
-algebra that contains E.
The next definition introduces the particular -algebra that will concern
us in this chapter.
Definition 12.1.3 (Borel -algebra). The Borel -algebra B on R is the
smallest -algebra that contains all the open subsets of R. That is,
B = (U)
where U = {U R : U is open}.
Ek =
k
428
If |(K)| < for every compact set K R, then we say that is a locally
finite measure.
If (E + h) = (E) for all h E then we say that is a translationinvariant measure.
Note that many texts define a measure to be what we are calling a
positive measure.
Notation 12.2.2. For simplicity of notation, we will use abbreviations such
as (a, b) = ((a, b)), [a, b] = ([a, b]), {x} = ({x}).
Example 12.2.3. The -algebra L of Lebesgue measurable subsets of R contains all the open subsets of Rd , and hence contains the Borel sets. Therefore
L B . There do exist Lebesgue measurable sets that are not Borel sets. For
example, all sets with exterior Lebesgue measure zero are Lebesgue measurable, but need not be Borel sets. Therefore we can consider Lebesgue measure
restricted to the Borel sets (which we simply refer to as Lebesgue measure),
and this gives us an example of a positive, unbounded Borel measure. We
usually denote this measure by dx. That is, we write Lebesgue measure dx
is a positive Borel measure. However, we will continue to write |E| for the
Lebesgue measure of E instead of dx(E).
Our next example modifies Lebesgue measure by weighting the real line.
Exercise 12.2.4.
R
RLet f : R [, ] be Lebesgue measurable, with at
least one of f + , f finite. Show that
Z
f (x) dx, E B ,
(E) =
E
429
430
(E) = (E P ) =
f + (x) dx,
(E) = (E N ) =
f (x) dx.
Show that ,
Although we will not prove it, the next result states that this same kind
of decomposition holds for arbitrary signed measures.
Theorem 12.2.12 (Jordan Decomposition Theorem). If is a signed
Borel measure on R, then there exist unique positive Borel measures + ,
such that = + and + .
Consequently, by definition of mutually singular measures, given a signed
measure there exist disjoint Borel sets P, N R with P N = R such that
= + , is null on P, and + is null on N. We call = + the
Jordan decomposition of (since it is unique), and R = P N an associated
Hahn decomposition of R (it is not unique).
Definition 12.2.13 (Positive, Negative, and Total Variation Measures). Given a signed Borel measure , let = + be its Jordan
decomposition.
(a) We call + the positive variation of , and the negative variation of .
(b) The total variation of is the positive measure
|| = + + .
E B .
(12.1)
E B .
E B .
Of course, ||(R) could be infinite, but if it should be finite then it follows that
|(E)| is bounded by the finite quantity ||(R) for every E B , and hence
is a bounded measure. In fact, the next exercise shows that the converse is
also true, which explains the terminology bounded measure instead of just
finite measure: If (E) is finite for all Borel sets E, then there is a finite
upper bound to the values of |(E)|.
431
432
Here are some useful equivalent formulations of the positive, negative, and
total variation measures.
Exercise 12.2.15. Let be a signed Borel measure. Show that if E B ,
then
+ (E) = sup (A) : A B , A E ,
(E) = inf (A) : A B , A E ,
X
n
n
S
||(E) = sup
Ek disjointly .
|(Ek )| : n N, Ek B , E =
k=1
k=1
Now that we have defined the total variation measure, we can define finite measures.
Definition 12.2.16 (-Finite Measures). Let be a signed Borel measure. If we can write R = Ek using at most countably many sets Ek B
each with ||(Ek ) < , then we say that is a -finite measure.
All bounded measures are -finite. Lebesgue measure is an example of an
unbounded measure that is -finite, while counting measure is an example of
a measure that is not -finite.
Exercise 12.2.17. Show that every locally finite signed Borel measure is
-finite, but a -finite signed Borel measure need not be locally finite.
Problems
12.2.18. Show that E B is a null set for a signed measure if and only
if ||(E) = 0.
12.2.19. Let = + be the Jordan decomposition of a signed measure ,
and let R = P N be an associated Hahn decomposition. Show that + (E) =
(E P ), (E) = (E N ), and ||(E) = (E P )(E N ) for E B .
12.2.20. Supposethat is a signed P
measure and Ek are disjoint Borel sets
|(Ek )| < .
such that | Ek | < . Show that
433
S
(Ek ).
Ek
k=1
k=1
k=1
434
d =
435
(x) d(x) =
N
X
ak (Ek ).
k=1
R
R
We write E f d to mean f E d.
If is a positive measure and a certain property holds except for a set E
with (E) = 0, then we say that this property holds -a.e.
We have the following convergence theorem for positive measures, analogous to Theorem 6.4.13 for Lebesgue measure.
Theorem 12.3.5 (Monotone Convergence Theorem). Let be a positive Borel measure on R, and let {fn }nN be a sequence of Borel measurable, nonnegative, monotone increasing functions on R. If we define
f (x) = limn fn (x), then
Z
Z
lim
fn d =
f d.
n=1
Z
X
fn d.
fn d =
n=1
436
R
R
(b) If f : R C is Borel measurable and Re(f ) d and Im(f ) d both
exist and are finite, then we define
Z
Z
Z
f d =
Re(f ) d + i Im(f ) d.
R
In all other cases, f d is undefined. Note in particular that if f is
R
complex-valued, then f d, if it exists, is a complex scalar. On the other
R
hand, if f is real-valued, then f d, if it exists, can be either a finite real
scalar or .
R
Lemma 12.3.11. If is a positive Borel measure on R and f d exists,
then
Z
Z
f d
|f | d.
When defining the space L1 () of functions that are integrable with respect
to , we have a choice between letting our functions be extended real-valued
or complex-valued. In this volume, we will take L1 () to consist of complexvalued -integrable functions.
Definition 12.3.12. If is a positive Borel measure on R, then L1 () consists of -integrable functions f : R C. The L1 -norm of f L1 () is
Z
kf k1 =
|f | d.
437
There are many other notations commonly used to denote L1 (), including
L (d), L1 (R; ), or L1 (R; d).
Note the implicit dependence of the notation kf k1 on .
R When we need
to emphasize the dependence on , we will write kf k1, = |f | d.
1
As for Lebesgue measure, the following result is one of the most useful
convergence theorems.
Theorem 12.3.14 (Dominated Convergence Theorem). Let be a positive Borel measure on R. Let {fn }nN be a sequence of Borel measurable
functions on R such that:
(a) fn (x) f (x) for -a.e. x, and
(b) there exists g L1 () such that |fn (x)| g(x) -a.e. for every n.
Then fn converges to f in L1 -norm, i.e.,
Z
lim kf fn k1 = lim
|f fn | d = 0.
n
Consequently, limn
fn d =
f d.
Problems
12.3.15. Show that if f is Borel measurable and a R, then
Characterize L1 (a ).
f da = f (a).
12.3.17. GivenPa positive Borel measure on R, let S be the set of all simple
N
functions = k=1 ck Ek such that (Ek ) < for each k. Show that S is
1
dense in L ().
438
|f | d , |f | d <
|f | d|| < ,
and, in this case,
Z
Z
f d
|f | d||.
Note that L1 () = L1 ( + ) L1 ( ).
The following exercise gives an important class of signed measures related
to a given positive Borel measure .
Exercise 12.4.4. Let be a positive Borel measure on R.
(a) Show that if f is a real, extended -integrable function, then
Z
(E) =
f d,
E B ,
(12.2)
E B ,
439
(E) = 0 = (E) = 0.
440
f (0) =
f d =
Although there
is no such function (x), it is common to abuse notation
R
and write f (x) (x) dx = f (0), even though what is really meant with
these symbols is integration of Rf with respect
to the measure, which in our
R
notation should be written as f d or f (x) d(x).
Problems
12.4.14. Let = + be the Jordan decomposition of a signed Borel
measure , and let R = P N be an associated Hahn decomposition. Show
that d + = P d, d = N d, and d|| = (P N ) d.
12.4.15. Show that if is a signed Borel measure, then
Z
||(E) = sup f d : |f | 1 .
E
X
N
f (xj ) : N N, xj R .
f d = sup
j=1
441
Ek =
k
Note that for a complex measure we have |(E)| < for every Borel
set E. The following exercise shows that a complex measure is bounded.
Exercise 12.5.2. Let be a complex Borel measure. For E B , define
r (E) = Re((E)) and i (E) = Im((E)). Show that r , i are bounded
signed measures, and for any E B we have
|(E)| |r (E)| + |i (E)| |r |(R) + |i |(R).
Conclude that is bounded in the sense that supEB |(E)| < .
(E) = 0 = (E) = 0.
r i ,
i r ,
i i .
442
Next we give an important structure theorem for complex measures (compare Theorem 12.4.8 for the case of signed measures).
Theorem 12.5.7 (LebesgueRadonNikodym Theorem). Let be a
complex Borel measure on R, and let be a -finite positive measure on R.
Then there exists a function f L1 () and a complex Borel measure such
that
= f d + , .
(12.3)
e where fe L1 (d) and
e , then
e = and
If we also have = fed +
fe = f -a.e.
We will need the following exercise in order to define the total variation
of a complex measure.
Exercise 12.5.8. Let be a complex Borel measure on R, and define =
|r | + |i |, so is a positive bounded Borel measure. Show there exists a
function f L1 () such that d = f d.
The total variation of a complex measure is a little more awkward to define
than it is for a signed measure. By Exercise 12.5.8, we know that if is a
complex measure, then there exists at least one positive measure and one
function f L1 () such that d = f d. We will define the total variation of
to be the measure d|| = |f | d, but of course we need to know that this is
well-defined. The following theorem shows that this definition is independent
of the choice of and f.
Theorem 12.5.9. Let be a complex Borel measure on R. If 1 , 2 are
bounded positive measures and f1 L1 (1 ), f2 L1 (2 ) are such that
f1 d1 = d = f2 d2 , then |f1 | d1 = |f2 | d2 .
Proof. Define = 1 + 2 . Then since 1 , there exists a function
g1 L1 () such that d1 = g1 d. Likewise, there exists some g2 L1 ()
such that d2 = g2 d. Because 1 , 2 , 0, we have g1 , g2 0 -a.e.
Thus, we have d = f1 d1 and d1 = g1 d. Exercise: Show that Exercise
12.4.11 generalizes to complex measures, and use this to show that d =
f1 g1 d an d = f2 g2 d (see also Problem 12.3.16).
The uniqueness statement in the LebesgueRadonNikodym Theorem
therefore implies that f1 g1 = f2 g2 -a.e. Consequently,
|f1 | g1 = |f1 g1 | = |f2 g2 | = |f2 | g2 -a.e.,
and hence
|f1 | d1 = |f1 | g1 d = |f2 | g2 d = |f2 | d2 .
443
k=1
X
S
(b) ||(E) = sup
Ek disjointly .
|(Ek )| : Ek B , E =
k=1
k=1
Z
(c) ||(E) = sup f d : |f | 1 ||-a.e. .
(12.4)
444
Consequently, if is a positive measure, then L1 () Mb (R). More precisely, if we define dg = g d for each g L1 (), then Exercise 12.5.16
shows that T : g 7 g is an isometric embedding of L1 () into Mb (R). If
is -finite, then range(T ) = { Mb (R) : }. In particular, Lebesgue
measure is a positive Borel measure, and hence if we identify f L1 (R) with
f dx Mb (R), then L1 (R) Mb (R).
Definition 12.5.17. The space of bounded discrete Borel measures is
X
X
Md (R) = Mb (R) : =
ck ak , distinct ak R,
|ck | < .
k=1
k=1
Exercise 12.5.18.
P Show that Md (R) is a closed subspace of Mb (R), and the
norm of = k=1 ck ak Md (R) is
kk =
|ck |.
k=1
Problems
12.5.19. Let be a positive Borel measure on R,
that g L1 ().
R and suppose
R
1
Define = g d. Show that if f L () then f d = f g d.
|r |, |i | || |r | + |i | = r+ + r + i+ + i ,
and L1 () = L1 (||).
12.5.21. Show that if is a complex Borel measure and (R) = ||(R), then
= ||.
12.5.22. Let be a complex Borel measure and a positive Borel measure.
Show that if and only if || .
12.5.23. Let be a complex Borel measure. Define the complex conjugate
of by (E) = (E) for E B . Show that is a complex measure,
R
R
L1 () = L1 (
), and f (x) d
(x) = f (x) d(x) for f L1 ().
445
446
(c) g(x) =
447
so is inner regular on E.
The fact that the real line is -compact, i.e., is a union of countably many
compact sets, is clearly an important ingredient of the preceding proof. On
a general space, a Radon measure will be inner regular on any subset that is
-finite.
We state the following useful property of Radon measures without proof.
Theorem 12.7.3 (Luzins Theorem). Let be a Radon measure on R. If
f : R C is Borel measurable and {f 6= 0} < , then for every > 0,
there exists a function Cc (R) such that
{f 6= } < .
xR
448
Problems
12.7.4. Show that if is a Radon measure, then Cc (R) is a dense subspace
of L1 ().
12.7.5. Since every subset of N is open, every positive measure on N is
a Borel measure. Show that is locally finite if and only if {k} < for
all k N, and the Radon measures on N are precisely the locally finite
positive measures on N.
449
all f Cc (R).
(12.8)
Since Cc (R) is a dense subspace of the Banach space C0 (R), Exercise 9.1.18
implies that such a has a unique extension to a bounded linear functional
on all of C0 (R), and we also refer to this extension as .
Example 12.8.2 (The Inductive Limit Topology). For each compact set K
R, define
C(K) = f Cc (R) : supp(f ) K .
Each of the spaces C(K) is a Banach space with respect to the uniform norm.
Further, as a set,
S
C(K) : K R, K compact .
Cc (R) =
We can define a topology on Cc (R) by declaring that a function f is continuous on Cc (R) if and only if for each compact K the restriction of f to C(K)
is continuous with respect to the L -norm on C(K).
In particular, a linear functional : Cc (R) C is continuous with respect
to this topology if and only if |C(K) : C(K) C is continuous for each
compact set K. Since C(K) is a normed space, this happens if and only if
each restriction |C(K) is bounded with respect to the norm on C(K), which
means that there exists a constant CK > 0 such that
|hf, i| CK kf k ,
all f C(K).
(12.9)
450
(12.10)
However, the converse implication does not hold in general, so these are
two distinct topologies on Cc (R). Equation (12.10) implies that the uniform
topology on Cc (R) is weaker than the inductive limit topology.
In this section we focus on Radon measures (which by definition are positive but possibly unbounded) and positive linear functionals on Cc (R). For
these results it is the inductive limit topology on Cc (R) that will be important.
In contrast, in Section 12.10 we will consider complex Radon measures (which
are necessarily bounded) and corresponding linear functionals on Cc (R), and
there it will be the L -topology on Cc (R) that will be important.
(12.11)
451
x R.
Hence kf k K f 0, so
0 kf k K f, = kf k hK , i hf, i.
Consequently,
|hf, i| hK , i kf k .
Now let f C(K) be arbitrary. Then
|hf, i| |hRe(f ), i| + |hIm(f ), i| 2 hK , i kf k ,
so the result follows with CK = 2 hK , i.
Although we will not prove it, the Riesz Representation Theorem completes the characterization of positive linear functionals on Cc (R): Every
positive linear functional on Cc (R) is induced from a Radon measure.
Theorem 12.8.7 (Riesz Representation Theorem I). If : Cc (R)
C is a positive linear functional, then there exists a unique positive Radon
measure on R such that
Z
hf, i =
f d, f Cc (R).
Moreover, if U R is open, then
(U ) = sup hf, i : f Cc (R), 0 f 1, supp(f ) U ,
and if K R is compact then
(K) = inf hf, i : f Cc (R), f K .
Thus, Radon measures and positive linear functionals on Cc (R) are equivalent. Therefore, we often use the same symbol
R to represent a Radon measure
and the positive functional f 7 hf, i = f d that it induces.
452
Problems
12.8.8. This problem will show that the locally finite positive measures on
N (which by Problem 12.7.5 are precisely the Radon measures on N) are in
1-1 correspondence with the positive linear functionals on c00 .
(a) Give the convergence criterion corresponding to the inductive limit
topology on c00 .
P (b) Show that if is a positive locally finite measure on N, then hf, i =
f (k) {k} defines a positive linear functional on c00 that is continuous with
respect to the inductive limit topology on c00 .
(c) Show that if is a positive linear functional on c00 , then there exists
a
unique
sequence of nonnegative scalars w = (wk )kN such that hf, i =
P
f (k) wk for f c00 . Show there is a unique locally finite positive measure
on N such that wk = {k} for every k.
and
(U \F ) < .
Proof. Since is -finite, there exist disjoint sets Ek B with (Ek ) <
such that E = Ek . Since is outer regular, there exist open sets Uk Ek
such that (Uk \Ek ) < 2k1 . Then U = Uk is open, U E, and
(U \E)
(Uk \Ek )
k=1
.
2
+
= .
2 2
453
j=1
454
Problems
12.9.4. Show that if is a Radon measure and f L1 () with f 0, then
f d is a Radon measure.
455
f Cc (R),
(12.13)
shows that Mb (N) = . Combining those two problems we see that, in the
discrete setting, the dual of c0 is the space of complex Radon measures on N:
c0
= Mb (N).
456
Although we will not prove it, the Riesz Representation Theorem states that
an analogous characterization holds on the real line.
Theorem 12.10.4 (Riesz Representation Theorem II). Given a measure Mb (R), define : C0 (R) C by
Z
hf, i =
f d
,
f C0 (R).
Then T : 7 is an antilinear isometry of Mb (R) onto C0 (R) .
Thus, C0 (R)
= Mb (R). We often write C0 (R) = Mb (R), meaning equality in the sense of the identification given in Theorem 12.10.4. Since Cc (R)
is dense in C0 (R) with respect to the uniform topology, this implies that
Cc (R)
= Mb (R). A complementary development of complex Radon measures could have started by declaring a complex Radon measure to be an
element of the dual space of Cc (R) or C0 (R) with respect to the uniform
topology. We could go further in this direction and declare the space of unbounded complex Radon measures to be the dual space of Cc (R) with respect
to the inductive limit topology. Indeed, Theorem 12.8.7 shows that the positive Radon measures correspond exactly to the positive linear functionals on
Cc (R) that are continuous with respect to the inductive limit topology.
Problems
12.10.5. Show directly that if is an unbounded Radon measure on R, then
there exist functions fn Cc (R) with fn 0 and kfn k 1 such that
hfn , i as n .
w
12.10.6. Given fn , f C0 (R), show that fn f if and only if sup kfn k <
and fn (x) f (x) pointwise for each x R (weak convergence of sequences
is discussed in Section 11.7).