Вы находитесь на странице: 1из 14

LECTURE NOTES: REAL ANALYSIS – MATH 510

Michael J. Johnson

Fall 2008

1. Chapters I and II

The first two chapters of our textbook are mostly review. I will ask the students to read
the following:
Chapter I sections 1,2,3,6
Chapter II sections 1,2,3,4,5
In addition to the above reading, I will discuss section 1.4, 1.6, 2.3, 2.5, 2.7

2.3 Extended Real Numbers

The extended real number system is obtained by appending to the real number
system the two elements +∞ and −∞. If x is any real number, then we agree that
−∞ < x < ∞ and define x + ∞ = ∞, x − ∞ = −∞, x · ∞ = ∞ if x > 0, x · −∞ = −∞ if
x > 0. Additionally, ∞ + ∞ = ∞ and −∞ − ∞ = −∞, but ∞ − ∞ is left undefined. We
do adopt the arbitrary convention that 0 · ∞ = 0.
Further Thoughts: (a) Is the extended real number system a field? (b) How would
you repair question 6 on page 37?

1.6 Countable Sets

We know what finite means, of course (note that this includes the empty set). A set is
called countable if it is empty or it is the range of a sequence. In other words, a nonempty
set E is countable if there exists a function from N onto E. Note that every finite set is
also countable. The phrase countably infinite is used to indicate that a set is countable
but not finite.
Key Results:
1. Every subset of a countable set is countable.
2. A countable union of countable sets is countable.

Typeset by AMS-TEX
1
2 REAL ANALYSIS

2.5 Open and Closed Sets of Real Numbers

Royden takes a different approach to the definition of a closed set than we take in Math
329. In Math 329 a set is closed if its complement is open, and then one proves as a
theorem that a set is closed if and only if it contains all of its limit points (recall that x is
a limit point of E if for every ε > 0 there exists y ∈ E such that |x − y| < ε; note that x
need not belong to E). In Royden’s book, a set is defined to be closed if it contains all of
its limit points and then obtains by theorem that the complement of an open set is closed
and vice-versa.
Regarding unions and intersections, the union of any collection of open sets results in
an open set and also the intersection of finitely many open sets results in an open set.
Similarly, the intersection of any collection of closed sets yields a closed set as does the
union of a finite number of closed sets.
Further Thought: Can you give an example where an infinite intersection of open
sets results in a set which is neither open nor closed?
Key Results:
1. Every open set can be written as the union of a countable collection of disjoint open
intervals.
2. (Lindelöf) If C is an infinite collection of open sets, then there exists a countable
sub-collection {Gi } of C such that

∪G∈C G = ∪∞
i=1 Gi .

3. (Heine-Borel) If F is closed and bounded, then every open covering of F has a finite
subcovering.
4. Let C be an infinite collection of closed sets such that every finite subcollection of
C has a nonempty intersection, and suppose that at least one of the sets in C is bounded.
Then ∩F ∈C F is nonempty.

1.4 Algebras of Sets

A non-empty collection A of subsets of a non-empty set X is called an algebra of sets


e is in A whenever A is.
if (i) A ∪ B is in A whenever A and B are, and (ii) A
Example. Let X be a non-empty set.
(a) Prove that P(X) is an algebra of sets.
(b) Prove that if A is an algebra of sets, then both ∅ and X belong to A.
(c) Prove that if A and B belong to an algebra of sets A, then so do A ∩ B and A ∼ B.
Proposition. Given any collection C of subsets of X, there exists a smallest algebra A
which contains C; that is, there exists an algebra A containing C such that if B is any
algebra containing C, then B contains A.
This smallest algebra containing C is called the algebra generated by C.
Example. Let X = R and C = {[0, 1], {10}}. Find the algebra generated by C.
Give the proof.
MATH 510 3

Proposition. Let A be an algebra of sets and let A1 , A2 , . . . belong to A. Then there exist
disjoint sets B1 , B2 , . . . in A such that

[ ∞
[
Bi = Ai .
i=1 i=1

Proof. Define B1 = A1 and Bn = An ∼ (A1 ∪ · · · ∪ An−1 ) for n = 2, 3, . . . .


Example. Find the algebra A of sets generated by the open subsets of R.
Solution. Let A0 denote the set of all open subsets of R, and define An = {(A ∪ B) ∼
C : A, B, C are subsets of S An−1 } for n = 1, 2, 3.... It’s easy to see that An is contained
S in
A for all n and therefore n An is contained in A. One can then show that n An is an
algebra of sets and therefore it equals A.
S∞
An algebra of sets A is called a σ-algebra if i=1 Ai belongs T∞ to A whenever A1 , A2 , . . .
belong to A. Note that A being a σ-algebra implies that i=1 Ai belongs to A whenever
A1 , A2 , . . . belong to A.
Proposition. Given any collection C of subsets of X, there exists a smallest σ-algebra A
which contains C; that is, there exists a σ-algebra A containing C such that if B is any
σ-algebra containing C, then B contains A.

2.7 Borel Sets

Definition. The collection B of Borel sets is the smallest σ-algebra which contains all of
the open sets.
In order to begin to appreciate the diversity of sets in B, explain the notation Fσ
(countable unions of closed sets), Gδ (countable intersections of open sets), Fσδ and Gδσ .
Does Q belong to B? Yes, Q is an Fσ .

2. Introduction to Lebesgue Measure

The purpose of Lebesgue Measure is to generalize the notion of length on R. When


this is done, we use the term measure instead of length. [You should give an improvised
discussion of what we might expect from a measure]. In order to have precise definitions,
we make the following:
Definition. Let M be a σ-algebra on X. A function m : M → [0, ∞] (Royden calls this
a set function) is called a countably additive measure if
!
[ X
m An = m(An )
n n

whenever {An } is a countable collection of disjoint sets in M.


4 REAL ANALYSIS

Implication. Suppose m is a countably additive measure on a σ-algebra M.


(a) Show that m(A) ≤ m(B) whenever A, B ∈ M with A ⊂ B.
(b) Show that either m(∅) = 0 or m(A) = ∞ for all A ∈ M.
(c) Show that if {An } is any countable subset of M, then
!
[ X
m An ≤ m(An ).
n n

3. Outer Measure

For an interval I ⊂ R, let ℓ(I) ∈ [0, ∞] denote the length ofPI. The outer measure of a

subset A ⊂ R, denoted m (A), is defined to be the infimum of n ℓ(In ), where
S the infimum
is taken over all countable collections of intervals {In } satisfying A ⊂ n In . It follows
immediately from the definition that m∗ (∅) = 0 and that A ⊂ B implies m∗ (A) ≤ m∗ (B).

Proposition. The outer measure of an interval equals its length.

Proof. We first consider the case I = [a, b]. One first shows that m∗ (I) ≤ ℓ(I) by consider-
ing the open inteval I1 = (a − ε, b + ε). If m∗ (I) < ℓ(I), then it follows (using Heine-Borel
Theorem) that there exists a finite collection of intervals I1 , I2 , . . . , Ik which cover I and
Pk
such that ℓ(I) > n=1 Ik , which is clearly a contradiction; therefore m∗ (I) = ℓ(I). One
can then consider any finite interval I by comparing it with a closed sub-interval J having
ℓ(J) > ℓ(I) − ε. And one handles infinite intervals in a similar fashion. 

Proposition. Let {An } be a countable collection of subsets of R. Then


[ X
m∗ ( An ) ≤ m∗ (An ).
n n

Proof. We consider the interesting case when m∗ (An ) < ∞ for all n. Fix ε > 0. For
each n,Pthere exists a countable collection of open intervals {In,i }i which covers An and for
which i m∗ (In,i ) ≤ m∗ (An )+2−n ε.SIt follows that the full
P collection {In,i } isP
a countable

collection of intervals which covers n An and satisfies i,n m (In,i ) ≤ ε + n m∗ (An ).
The desired conclusion readily follows. 

Two immediate consequences of this proposition are that the outer measure of any
countable set is 0, and consequently, any interval having a positive length is necessarily
uncountable.
MATH 510 5

4. Lebesgue Measure

The outer measure m∗ is not, unfortunately, countably additive on P(R); there is,
however, a smaller σ-algebra upon which it is countably additive.
Definition. A set E ⊂ R is measurable if
e for all A ⊂ R.
m∗ (A) = m∗ (A ∩ E) + m∗ (A ∩ E)
e
Note that “≤” is assured since A = (A ∩ E) ∪ (A ∩ E).
Lemma. If m∗ (E) = 0, then E is measurable.
Proposition. The family M of measurable sets is an algebra of sets.
Proof. That M is closed under complimentation follows from the symmetry in the defi-
nition of measurable. Now suppose that D and E are measurable sets and let A be any
subset of R.
e since D is measurable,
m∗ (A) = m∗ (A ∩ D) + m∗ (A ∩ D),
e ∩ E) + m∗ (A ∩ D
= m∗ (A ∩ D) + m∗ (A ∩ D e ∩ E),
e since E is measurable,
≥ m∗ (A ∩ [D ∪ E]) + m∗ (A ∼ [[D ∪ F ]),
e
where the inequality holds since A ∩ (D ∪ E) = [A ∩ D] ∪ [A ∩ E ∩ D]. 
Lemma. Let A be any set and let E1 , E2 , . . . , En be disjoint measurable sets. Then
n
! n
[ X
m∗ (A ∩ Ei ) = m∗ (A ∩ Ei ).
i=1 i=1

Proof. The induction step looks like:


Since En is measurable, we have
n
! n
! n
!
[ [ [
m∗ (A ∩ Ei ) = m∗ (A ∩ Ei ∩ En ) + m∗ (A ∩ Ei en )
∩E
i=1 i=1 i=1
n−1
!
[
= m∗ (A ∩ En ) + m∗ (A ∩ Ei )
i=1
whence follows the conclusion by the induction hypothesis. 
Theorem. The collection M of measurable sets is a σ-algebra.
S∞
Proof. Suppose E = i=1 Ai where A1 , A2 , . . . are measurable sets. Since S∞M is an algebra
of
Sn sets, there exist disjoint measurable sets B 1 , B 2 , . . . such that E = i=1 Bi . Put En =
i=1 Bi . The above lemma yields:
n
X n
X
en )) =
m∗ (A) = m∗ (A∩(B1 ∪B2 ∪· · ·∪Bn ∪E en ) ≥
m∗ (A∩Bi )+m∗ (A∩E e
m∗ (A∩Bi )+m∗ (A∩E).
i=1 i=1
P∞ e
Since the left side is independent of n, it follows that m∗ (A) ≥ i=1 m∗ (A∩Bi )+m∗ (A∩E).
Applying the countable subadditivity of m∗ yields m∗ (A) ≥ m∗ (A ∩ E) + m∗ (A ∩ E). e 
6 REAL ANALYSIS

Lemma. The σ-algebra M of measurable sets contains all intervals E = (a, ∞), a ∈ R.
Proof. Let A be any subset of R and put A1 = A∩(a, ∞), A2 = A∩(−∞, a]. We must show
that m∗ (A) ≥ m∗ (A1 ) + m∗ (A2 ). Given ε > 0, there P exists a countable collection of open
intervals {In } which cover A and which satisfy n ℓ(In ) ≤ m (A) + ε. Put In′ = In ∩ E

and In′′ = In ∩ Ee and note that ℓ(In ) = ℓ(In′ ) + ℓ(In′′ ) = m∗ (In′ ) + m∗ (In′′ ). Since A1 ⊂ ∪n In′
′′
and A2 ⊂ ∪n In , we have
X X
m∗ (A1 ) + m∗ (A2 ) ≤ m∗ (In′ ) + m∗ (In′′ ) ≤ · · · ≤ m∗ (A) + ε.
n n


Theorem. The Borel sets B are contained in the measurable sets M.
Proof. Homework 2.
Definition. The Lebesgue measure of a measurable set E, denoted m(E), is defined to
be the outer measure of E. In other words, the set function m : M → [0, ∞] is defined to
be the restriction of m∗ to M.
At this stage, what we have established is that M is a σ-algebra which is at least as
large as B, and our measure m : M → [0, ∞] is translation invariant, agrees with length
on intervals and is countably subadditive; that is, if {An } be a countable collection of
measurable sets, then
[ X
m( An ) ≤ m(An ).
n n

Furthermore, if E1 , E2 , . . . , En are pairwise disjoint measurable sets and A is any measur-


able set, then !
n
[ Xn
m(A ∩ Ei ) = m(A ∩ Ei ).
i=1 i=1

Our first task (and this was the purpose of defining M) is to show that Lebesgue measure
is in fact countably additive.
Theorem. Lebesgue measure is countably additive.
Proof. It follows from the above equality (with A = R) that Lebesgue measure is finitely
additive. Now, let E1 , E2 , . . . be pairwise disjoint measurable sets. Then for all n ∈ N


[ n
[ n
X
m( Ei ) ≥ m( Ei ) = m(Ei )
i=1 i=1 i=1

S∞ P∞
Taking the supremum over n yields m( i=1 Ei ) ≥ i=1 m(Ei ) and the proof is completed
by using the countable subadditivity of Lebesgue measure. 
MATH 510 7

Proposition. Let {En } be measurable sets.


(a) If E1 ⊂ E2 ⊂ E3 ⊂ · · · , then
[
m( En ) = lim m(En ).
n
n

(b) If E1 ⊃ E2 ⊃ E3 ⊃ · · · and m(E1 ) < ∞, then


\
m( En ) = lim m(En ).
n
n

Proof (a). Assume E1 ⊂ E2 ⊂ E3 ⊂ · · · . With E0 = ∅, we have


[ [ ∞
X
m( En ) = m( (En ∼ En−1 )) = (m(En ) − m(En−1 ))
n n n=1
N
X
= lim (m(En ) − m(En−1 )) = lim m(EN ).
N→∞ N→∞
n=1
T
(b) Assume E1 ⊃ E2 ⊃ E3 ⊃ · · · and m(E1 ) < ∞. Put E = n En and define Bn = E1 ∼
En for n = 1, 2, . . . . Then B1 ⊂ B2 ⊂ · · · and ∪n Bn = E1 ∼ E. By (a), we have

m(E1 )−m(E) = m(E1 ∼ E) = lim m(Bn ) = lim [m(E1 )−m(En )] = m(E1 )− lim m(En ),
n→∞ n→∞ n→∞

whence follows the desired conclusion 


Note that if En = (n, ∞) satisfies the
T hypothesis of the Proposition except that m(E1 ) =
∞, and the conclusion fails since m( n En ) = m(∅) = 0 6= limn m(En ).

5. A non-measurable set

A valid question at this point is whether or not every subset of R is measurable. The
answer to that question is, “It depends”. It depends on whether or not one accepts the
axiom of choice. Using the axiom of choice, we will construct a set P which is not Lebesgue
measurable.
We first define an equivalence relation on [0, 1] by x ≡ y if x − y is rational. This
equivalence relation partitions the set [0, 1] into equivalence classes and using the axiom
of choice we let P ⊂ [0, 1] be a set which contains exactly one element of each equivalence
class. Now, letSq1 , q2 , . . . be an enumeration of the rational numbers contained in [−1, 1],

and put E = n=1 (P + qn ). We first note that E ⊂ [−1, 2] since P ⊂ [0, 1] and {qn } ⊂
[−1, 1]. We next claim that the sets {P + qn } are pairwise disjoint. To see this, suppose
x ∈ (P + qn ) ∩ (P + qm ), say x = p1 + qn = p2 + qm for some p1 , p2 ∈ P . It follows that
p1 ≡ p2 (since p1 − p2 = qm − qn ∈ Q) and hence that p1 = p2 . But this implies that
qn = qm and consequently that n = m. Lastly, we claim that [0, 1] ⊂ E. To see this, let
x ∈ [0, 1]. Then there exists some p ∈ P such that x ≡ p, say x − p = q ∈ Q. Since both x
and p are in [0, 1], it follows that q ∈ [−1, 1], say q = qn . Hence x ∈ P + qn ⊂ E.
8 REAL ANALYSIS

Theorem. P is not Lebesgue measurable.


Proof. Assume that P is measurable. Then E, being Pthe countable union
P of pairwise
disjoint measurable sets, is measurable and m(E) = n m(P + qn ) = n m(P ) (since
Lebesgue measure is translation invariant). If m(P ) = 0, then m(E) = 0 which is a
contradiction since E ⊃ [0, 1]. On the other hand, if m(P ) > 0, then m(E) = ∞ which is
a contradiction since E ⊂ [−1, 2]. 

6. Measurable functions

Proposition. Let f be an extended real-valued function whose domain is measurable.


Then the following are equivalent.
(i) For each α ∈ R, {x : f (x) > α} is measurable.
(ii) For each α ∈ R, {x : f (x) ≥ α} is measurable.
(iii) For each α ∈ R, {x : f (x) < α} is measurable.
(iv) For each α ∈ R, {x : f (x) ≤ α} is measurable.

Definition. An extended real-valued function f is (Lebesgue) measurable if it’s domain


is measurable and it satisfies one of the four above statements.
Exercise. Show that if f is measurable then for all α ∈ [−∞, ∞], {x : f (x) = α} is
measurable.
Example. Show that if f : R → R is continuous, then f is measurable.
Proposition. Let c be a constant and f and g two measurable real-valued functions with
the same domain. Then the functions f + c, cf , f + g, f − g, f g are measurable.
Proof. We’ll use condition (iii) of the Proposition/Definition. Fix α ∈ R and note that
f (x) + g(x) < α if and only if there exists r ∈ Q such that f (x) < r < α − g(x). Hence
[
{x : f (x) + g(x) < α} = [{x : f (x) < r} ∩ {x : g(x) < α − r}]
r∈Q
which is measurable since M is a σ-algebra. This shows that f + g is measurable. In
particular, since the constant function is measurable, we see that f + c is measurable. If
c 6= 0, then 
{x : f (x) < α/c} if c > 0
{x : cf (x) < α} =
{x : f (x) > α/c} if c < 0
which is measurable; hence cf is measurable. In particular, −g = (−1)g is measurable and
consequently f − g is measurable. We next consider the function f 2 . If α > 0, then
√ √
{x : f (x)2 > α} = {x : f (x) < − α} ∪ {x : f (x) > α}
which is measurable and if α ≤ 0, then {x : f (x)2 > α} equals the domain of f which is
assumed to be measurable. Therefore, f 2 is measurable. Writing
1
f g = [(f + g)2 − f 2 − g 2 ]
2
we see that f g is measurable. 
MATH 510 9

Theorem. Let {fn } be a sequence of extended real-valued measurable functions with the
same domain of definition. Then the functions sup{f1 , f2 , . . . , fn }, inf{f1 , f2 , . . . , fn },
supn fn , inf n fn , lim supn fn , lim inf n fn are measurable.
Proof. If h(x) := sup{f1 (x), f2 (x), . . . , fn (x)}, then
n
[
{x : h(x) > α} = {x : fi (x) > α}
i=1
which is measurable. The next three functions are handled in a similar fashion. For the
function lim supn fn , we write
lim sup fn = inf hn , where hn = sup{fn , fn+1 , . . . }
n n

and easily see that it is measurable. The last function is handled in a similar fashion. 
A property is said to hold almost everywhere (a.e.) if the set of points where it fails
to hold is (contained in a) set having measure 0. For example, we say that f = g a.e.
if f and g have the same domain and m{x : f (x) 6= g(x)} = 0. Similarly, we say that
a sequence of functions {fn } converges a.e. to a function g if there exists a set E, with
m(E)e = 0, such that {fn (x)} converges to g(x) for all x ∈ E.
Proposition. If f is a measurable function and f = g a.e., then g is measurable.
Proof. Assume f is a measurable function and f = g a.e. Then there exists a measurable
e = 0 such that f (x) = g(x) for all x ∈ E. For α ∈ R, we have
set E, with m(E)
 
{x : g(x) > α} = (E ∩ {x : f (x) > α}) ∪ Ee ∩ {x : g(x) > α}

which is measurable. 
Question. Does there exist a function which is not measurable?
Answer: It depends on whether one accepts the uncountable axiom of choice.
The characteristic function of a set A, denoted χA , is measurable if and only if A is
measurable. So if we use the axiom of choice to construct the non-measurable set P , then
χP is a non-measurable function.
Reading Assignment. Ask the students to read the beginning of Chapter 6.
Proposition. Let E be a measurable set with finite measure and suppose that {fn } is a
sequence of measurable functions that converge to a real-valued function f a.e. on E. Then
for every ε > 0 and δ > 0, there exists a measurable set A ⊂ E, with m(A) < δ, and an
integer N such that
|f (x) − fn (x)| < ε for all x ∈ E ∼ A.

Proof. Put C = {x ∈ E : {fn (x)} converges to f (x)} and note that m(E ∼ C) = 0
(whence it follows that C is measurable). Fix ε > 0 and define
EN := {x ∈ E : |f (x) − fn (x)| ≥ ε for some n ≥ N }.
T
Then E1 ⊃ E2 ⊃ · · · and N EN ⊂ E ∼ C. By previous results (since m(E1 ) < ∞), it
follows that limN m(EN ) = 0. The desired conclusion now follows easily. 
10 REAL ANALYSIS

Corollary (Egoroff ’s Theorem). Let E be a measurable set with finite measure and
suppose that {fn } is a sequence of measurable functions that converge to a real-valued
function f a.e. on E. Then for every δ > 0 there exists a measurable set A ⊂ E, with
m(A) < δ, such that fn converges to f uniformly on E ∼ A.
Proof is a homework problem.

7. The Riemann Integral

A simple function is a real-valued measurable function, defined on R, whose range is


a finite set. If φ is a simple function with Pnrange {α1 , α2 , . . . , αn } (assumed distinct), then
φ can be written (canonically) as φ = i=1 αi χA , where Ai = {x : φ(x) = αi }. We note
i
that the sum, difference or product of two simple functions is again simple. We mention
that if a1 , a2 , . . . , an are any (not
Pnnecessarily distinct) real values and E1 , E2 , . . . , En are
any measurable sets, then ψ = i=1 ai χE is a simple function. A step function s is a
i Pn
simple function which can be expressed in the form ψ = i=1 ai χI , where I1 , I2 , . . . , In
i
are pairwise disjoint intervals (we’ll admit the singleton a = [a, a] as an interval). It’s also
the case that the sum, difference or product of two step functions is again a step function.
The integral over [a, b] of such a step function is defined (unanimously) as
Z b n
X
ψ= ai ℓ(In ∩ [a, b]).
a i=1

Riemann’s definition of the integral of a bounded function f : [a, b] → R can be formulated


in terms of step functions as follows:
Z b
Rb
Definition. The upper Riemann integral of f , denoted R f , is the infimum of a ψ
a
over all step functions ψ with ψ ≥ f on [a, b], and the lower Riemann integral of f ,
Z b
Rb
denoted R f , is the supremum of a φ over all step functions φ with φ ≤ f on [a, b]. If
a
Z b Z b
R f = R f , then f is said to be Riemann integrable and the Riemann integral
a a
of f is
Z b Z b Z b
R f (x) dx = R f = R f.
a a a

We note that step functions are Riemann integrable.


Example. Let A = Q ∩ [0, 1]. Show that the simple function s = χA is not Riemann
integrable.
MATH 510 11

8. The Lebesgue Integral of a bounded function on a set of finite measure

Whereas step functions serve as the ’building blocks’ of Riemann’s theory of integration,
in Lebesgue’s theory, this role is played by simple functions. We thus
Pn accept the notion that
if φ is a simple function having canonical representation φ = i=1 αi χA ({αi } distinct
i
and {Ai } pairwise disjoint), and if φ vanishes outside a set of finite measure (ie m(Ai ) = ∞
implies ai = 0), then
Z X n
φ= αi m(Ai ).
i=1

Note that it’s possible that for some i we have αi = 0 and m(Ai ) = ∞; but this
R is precisely
where the convention 0 · ∞ = 0R comes into action. We also use the notation E φ, when E
is a measurable set, to denote φχE , where we note that φχE is again a simple function.
Of course simple functions have representations other than canonical and it is a fact that
if our simple function φ above can be written as
m
X
φ= ai χE ,
i
i=1

with m(Ei ) = ∞ only when ai = 0, then


m
X n
X
ai m(Ei ) = αi m(Ai ).
i=1 i=1

This is an elementary consequence of m being finitely additive and M being an algebra


of sets. However, the proof is too tedious and time-consuming for our lecture, so we will
take this as given and the student is refered to page 78 of our textbook.
Proposition.
R R For φ, ψ ∈ Sf , the following hold:
(i) R cφ = c φ forR allRconstants c ∈ R.
(ii) (φ + ψ) = φ R+ ψ.R
(iii) If φ ≥ ψ, then φ ≥ ψ.
Proof. Now that we’ve assumed away the difficulty of dealing with non-canonical repre-
sentations, the proof is straightforward. 
Imitating our definitions of upper and lower Riemann integrals, let us assume that f is
a bounded function defined on a measurable set E of finite measure (we are not assuming
R R
that f is measurable).
R We define L E f toRbe the infimum of E φ over all simple functions
φ ≥ f and L f to be the supremum of E ψ over all simple functions ψ ≤ f . Precisely,
E
Z Z
L f = inf{ φ : φ is a simple function with φ ≥ f on E}
E E
Z Z
L f = sup{ ψ : ψ is a simple function with ψ ≤ f on E}
E E
12 REAL ANALYSIS
R R
We note that L f ≤ L E f because if ψ and φ are simple functions satisfying ψ ≤ f ≤ φ
R E R R R
on E, then E ψ ≤ E φ by (iii) above. Furthermore, we note that we have L f = L E f
E
if and onlyR if for Rall ε > 0, there exist simple functions ψ and φ, with ψ ≤ f ≤ φ on E,
such that E φ − E ψ < ε.
Theorem. Let f be a bounded function defined on a measurable set E of finite measure.
If f is measurable, then
Z Z
L f =L f.
E E

Proof. Assume f is measurable and |f (x)| ≤ M for all x ∈ E, where m(E) < ∞. Let
ε > 0. Then there exists N ∈ N such that M
N
m(E) < ε. We define measurable sets

(n − 1)M nM
En = {x : < f (x) ≤ }, for n = −N, −N + 1, . . . , N,
N N

and note that {En } is a finite collection of measurable, pairwise disjoint sets whose union
PN
equals E. It follows that n=−N m(En ) = m(E). Using these sets, we define simple
functions
XN XN
(n − 1)M nM
ψ(x) = χE and φ(x) = χ ,
N n N En
n=−N n=−N

and note that ψ ≤ f ≤ φ on E. Furthermore,

Z Z N
X   N
X
nM (n − 1)M M M
φ− ψ= − m(En ) = m(En ) = m(E) < ε.
E E N N N N
n=−N n=−N

R R
Therefore, L E
f =L f.
E

Definition. If f is a bounded measurable function defined on a measurable set E of finite
measure, we define the (Lebesgue) integral of f over E by
Z Z Z
f (x) dx = L f =L f.
E E E

Corollary. Let f : [a, b] → R be bounded. If f is Riemann integrable, then f is measurable


and
Z b Z b
R f (x) dx = f (x) dx.
a a
MATH 510 13

Proposition.
R R Let f be a bounded function defined on a measurable set E of finite measure.
If L E f = L f , then f is measurable.
E
R R
Proof. Assume that L E f = L f . Then for each n ∈ N, there exist simple functions
E R R
φn and ψn , with ψn ≤ f ≤ φn , such that E φn − E ψn < n1 . Define φ∗ = inf n φn
and ψ ∗ = supn ψn (which are measurable functions). It follows from the above that
ψ ∗ ≤ f ≤ φ∗ on E. In order to show that f is measurable, it suffices to show that f = φ∗
a.e. This will be accomplished by showing that the set ∆ = {x : φ∗ (x) > ψ ∗ (x)} has
measure 0. We can write ∆ as

[ 1
∆= {x : φ∗ (x) − ψ ∗ (x) > }.
k
k=1
1
∗ ∗
But {x : φ (x) − ψ (x) > ⊂ {x : φn (x) − ψn (x) > k1 } for all n and it’s easy to see that
k}
the measure of the latter set is less than nk . Hence the measure of the former is 0. So ∆
is a countable union of sets having measure 0 and therefore ∆ has measure 0. 
Properties of the Lebesgue integral. If f and g are bounded measurable functions
defined
R on a set E ofR finite measure,
R then:
(i) E (af + bg) = a E fR + b ERg.
(ii) If f = g a.e., then RE f = RE g.
(iii) If f ≤ Rthen E f ≤ E g;
R g a.e.,
Hence E f ≤ e |f |.

R
(iv) If f (x) ∈ [c, d] for all x, then cm(E) ≤ E f ≤ dm(E).R R R
(v) If A and B are disjoint measurable subsets of E, then A∪B f = A f + B f .
R R
Proof. We can express (f + g) in two R ways. First of all it’s the infimum of φ over
φ
R ≥ f .
R But this is ≤ Rthe infimumR of (φ
R 1 + φ2 ) over φ 1 ≥ f and φ 2 ≥ g,R which equals
fR+ g. Therefore,
R R (f + g) ≤ f + g. Similarly, using supremeum of ψ, we arrive
at (f + g) ≥ f + g, which R completes the proof of (i).
Claim: If f ≥ 0 a.e., then f ≥ 0. R
proof. If φ ≥ f , then φ ≥ 0 a.e. and it follows that φ ≥ 0.
(ii) and (iii) follow easily from the claim. And (iv) and (v) are quite easy.
Bounded Convergence Theorem. Let {fn } be a uniformly bounded sequence of mea-
surable functions defined onR a set ER of finite measure. If {fn } converges to a bounded
function f a.e. on E, then E fn → E f as n → ∞.
Proof. We consider first the case when {fn (x)} converges to f (x) for all x ∈ E. There
exists M < ∞ such that |fn (x)| ≤ M for all x and n. It follows that f also satisfies this
ε
bound. Let ε > 0 and put δ = 2M . By previous theorem, there exists N and A ⊂ E, with
m(A) < δ, such that
ε
|fn (x) − f (x)| < for all x ∈ E ∼ A and n ≥ N.
2m(E)
R R
It now follows that E fn − E f < ε for all n ≥ N . The case when {fn (x)} converges a.e.
to f on E then follows from the above and from the homework problem that the integral
of a bounded measurable function over a set of measure 0 is 0. 
14 REAL ANALYSIS

Examples. Below are listed several sequences of functions which converge a.e. to 0.
(a) fn = χ[0,1/n]
(b) gn = nχ[0,1/n]

(c) hn = nχ[0,1/n]
1
(d) kn = n χ[0,n]
1
(e) qn = χ
n3/2 [0,n]
Discuss whether the limit of the integral is 0 and whether or not the Bounded Convergence
Theorem applies. As other theorems are encountered, discuss whether they apply to any
of these examples.

Вам также может понравиться