Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
These are notes for Math 324, an introductory course in Measure and Integra-
tion. Students are advised to go through all sections in detail and attempt all
problems. These notes will be modified and expanded as the course proceeds.
i
ii
Contents
2 Lebesgue Integration 15
2.1 Review of the Riemann integral . . . . . . . . . . . . . . . . . 15
2.2 Lebesgue integral for bounded functions on sets of finite measure 26
2.3 Lebesgue integrals of nonnegative measurable functions . . . . 35
iii
Chapter 1
E(f ≥ α) = {x ∈ E : f (x) ≥ α}
E(f ≤ α) = {x ∈ E : f (x) ≤ α}
E(f = α) = {x ∈ E : f (x) = α}
1
2 CHAPTER 1. LEBESGUE MEASURABLE FUNCTIONS
Proof. This theorem follows from Exercise 1.2.2 and the fact that for any
α ∈ R, [ [
(E1 E2 )(f < α) = E1 (f < α) E2 (f < α).
is measurable. [Hint: For each rational number r, the set {x ∈ E : f (x) <
r < g(x)} is measurable.]
(a) f ± c
(b) f + g
(c) cf
(d) f − g
4 CHAPTER 1. LEBESGUE MEASURABLE FUNCTIONS
(e) |f |
(f ) f 2
(g) f g
(h) f /g, provided g(x) 6= 0 on E.
Proof. Parts (a) and (b) are left as easy exercises. To prove (c), for any
α ∈ R,
E(f + g > α) = {x ∈ E : f (x) > α − g(x)}.
Since g is measurable, so is α − g. So,
[ \
{x ∈ E : α − g(x) < f (x)} = (E(α − g < r) E(f > r)),
r∈Q
and \
E(f∗ > α) = E(f1 > α) E(f2 > α) ∈ M.
inf fn := − sup(−fn ),
n n
F (g > α), being a subset of a set with measure zero, is measurable. Since F
is measurable, so is F C . Thus,
\
F C (f > α) = E(f > α) F C ∈ M.
Thus, E(g > α), being a union of two measurable sets, is measurable. This
proves that g is measurable.
Example 1.3.3 Let f (x) = 0 for x ∈ [0, 1]. Define, for each n ≥ 1,
(
xn if 0 ≤ x < 1,
fn (x) =
0 if x = 1.
We can check that fn (x) −→ f (x) for each x ∈ [0, 1]. However, for each
n ≥ 1,
sup |fn (x) − f (x)| = 1.
x∈[0,1]
Thus,
lim sup |fn (x) − f (x)| =
6 0.
n→∞ x∈[0,1]
This implies that the sequence {fn } does not converge uniformly to f on
[0, 1]. On the other hand, it is interesting to observe that if we chose an
with 0 < < 1 and consider the set E = [1 − , 1], then, for every n ≥ 1,
sup |fn (x) − f (x)| = (1 − )n .
x∈[0,1−]
Thus,
lim sup |fn (x) − f (x)| = 0.
n→∞ x∈E
Proof. For each k ≥ 1, we partition the interval [0, ∞] into disjoint interval
as follows: [ [ [
[0, ∞] = [0, 1) [1, 2) · · · [k − 1, k) [k, ∞]
1.4. APPROXIMATION OF MEASURABLE FUNCTIONS BY SIMPLE FUNCTIONS9
k2k
[ i−1 i
[
= , [k, ∞].
i=1
2k 2k
k
Define, for each 1 ≤ i ≤ k2 ,
−1 i−1 i
Eki = f ,
2k 2k
k2 k
X i−1
χEk,i + kχEk .
i=1
2k
If f (x) < ∞, then there exists a positive integer k such that f (x) < k. Since
x ∈ Ek,i for some 1 ≤ i ≤ k2k , we have
i i−1 1
0 ≤ f (x) − φk (x) ≤ k
− k = k.
2 2 2
If f (x) = ∞, then φk (x) = k for every k. Thus, for each x ∈ E,
i−1
φk (x) = .
2k
On the other hand,
2i − 2 2i − 1
φk+1 (x) = or ,
2k+1 2k+1
10 CHAPTER 1. LEBESGUE MEASURABLE FUNCTIONS
(k+1)2k+1
X i−1 k2k+1
φk+1 (x) = χE ≥ = φk (x).
2k+1 k,i 2k+1
i=k.2k+1 +1
Proposition 1.5.1 Let E be a measurable set with finite measure and {fn }
be a sequence of measurable functions defined on E. Let f be a real valued
measurable function such that fn (x) −→ f (x) for each x ∈ E. Given > 0
and δ > 0, we can find a measurable subset A of E with m(A) < δ and an
integer N such that |fn (x) − f (x)| < for all x ∈ E − A and all n ≥ N.
1.5. ALMOST UNIFORM CONVERGENCE OF SEQUENCES OF FUNCTIONS11
That is,
Ek = {x ∈ E : |fn (x) − f (x)| ≥ for some n ≥ k}.
Clearly, Ek+1 ⊆ Ek . Moreover, since {fn } converges to f for each x ∈ E,
there must be some positive integer k such that x 6∈ Ek . Thus,
Ek k Ek+1 k · · ·
and ∞
\
Ek = ϕ.
n=1
Thus,
lim m(En ) = 0.
n→∞
[Note that it is here that we use the fact that m(Ek ) < m(E) < ∞.] That
is, given δ > 0, there exists a natural number N such that m(Ek ) < δ for all
k ≥ N. In particular, m(EN ) < δ. Moreover, for each x ∈ E − EN , we have
|fn (x) − f (x)| < for all n ≥ N.
Exercise 1.5.2 Prove the following modification of the above theorem: Let E
be a measurable set with finite measure and {fn } be a sequence of measurable
functions defined on E. Let f be a real valued measurable function such that
fn (x) −→ f (x) almost everywhere in E. Given > 0 and δ > 0, we can
find a measurable subset A of E with m(A) < δ and an integer N such that
|fn (x) − f (x)| < for all x ∈ E − A and all n ≥ N.
Note that the condition m(E) < ∞ cannot be relaxed. Define, for each
n ≥ 1, (
0 if 0 ≤ x ≤ n
fn (x) =
1 if x > n.
One can check that the sequence {fn } converges to f on E, but the theorem
does not hold for = 1 and some 0 < δ < 1.
12 CHAPTER 1. LEBESGUE MEASURABLE FUNCTIONS
The converse of the statement in the above exercise forms the content of a
very important theorem due to Egoroff.
Theorem 1.5.5 [Egoroff ’s theorem] Let E be a measurable set with m(E) <
∞ and let {fn } be a sequence of functions which converge to f almost ev-
erywhere on E. Then, the sequence {fn } converges almost uniformly to f on
E.
Proof. Let η > 0 and for each k ≥ 1, choose (k) = k1 and δk = 2ηk . Taking
E = E0 and repeatedly applying Exercise 1.5.2, we can find a measurable
set Ak ⊂ Ek−1 and a positive integer Nk such that m(Ak ) < δk and |fn (x) −
f (x)| < k1 for all n ≥ Nk and for all x ∈ Ek = Ek−1 − Ak . Set
∞
[
A= Ak .
k=1
Note that ∞ ∞
X X η
m(A) ≤ m(Ak ) < k
= η.
k=1 k=1
2
One can easily check that
∞
\
E−A= Ek .
k=0
Is the converse of the above theorem true? If not, how close is it to being
true?
Exercise 1.6.2 Show that the following functions are measurable on their
respective domains.
(a)
2
x
if x < 1,
f (x) = 2 if x = 1,
2−x if x > 1.
(b)
(
1, if x is a rational number in [0, 1]
f (x) =
0, if x is an irrational number in[0, 1].
14 CHAPTER 1. LEBESGUE MEASURABLE FUNCTIONS
Notice that in the above exercise, the function defined in (a) is continu-
ous all except finitely many points, where as the function defined in (b) is
continuous nowhere. Nonetheless, both are measurable.
In the next exercise, we prove that every measurable function is almost
continuous.
Lebesgue Integration
Remark 2.1.3 The value of the Riemann integral of a step function is in-
dependent of the choice of the partition [a, b] as long as the step function is
constant on the open subintervals of the partition. The, the Riemann integral
of a step function is well defined.
15
16 CHAPTER 2. LEBESGUE INTEGRATION
Definition 2.1.5 Let f (x) be a bounded function on [a, b], that is, let α, β be
real numbers such that α ≤ f (x) ≤ β for all x ∈ [a, b]. The lower Riemann
integral of f on [a, b] is defined as
Z b Z b
f (x)dx := sup φ(x)dx : φ(x) ≤ f (x), φ is a step function (2.1)
a a
Exercise 2.1.9 Check that the definition for Riemann integral for a step
function in Definition 2.1.8 coincides agrees with Definition 2.1.1.
Conversely, suppose > 0 and there are step functions φ ≤ f ≤ ψ such that
Z b
0≤ [ψ(x) − φ(x)]dx < .
a
Thus,
Z b Z b
0≤ f (x)dx − f (x)dx < .
a a
Since this is true for any > 0,
Z b Z b
f (x)dx = f (x)dx.
a a
Exercise 2.1.11 The aim of the following exercises is to prove that every
continuous function on [a, b] is Riemann integrable.
(b) Prove that f is uniformly continuous on [a, b]. Thus, given > 0, there
exists a δ > 0 such that
|f (x) − f (y)| <
b−a
for all x, y ∈ [a, b] for which |x − y| < δ.
such that xk − xk−1 < δ for all 1 ≤ k ≤ n. Show that for each such k,
sup f − inf f≤ .
(xk−1 ,xk ) (xk−1 ,xk ) b−a
and
ψ(x) = sup f (x), x ∈ (xk−1 , xk ).
(xk−1 ,xk )
Show that Z b
0≤ [ψ(x) − φ(x)]dx ≤ .
a
Definition 2.1.12 Let f be a bounded function on an interval [a, b]. For any
interval J, the oscillation of f on J is defined to be the number
Thus,
ωf ((x0 − δ, x0 + δ) ∩ [a, b]) ≤ 2.
We have shown that for any > 0,
Thus, ωf (x0 ) = 0.
Conversely, suppose ωf (x0 ) = 0 and let > 0. Then here exists an open
interval I containing x0 such that 0 ≤ ωf (I) < . Since I is open, choose
δ > 0 such that
(x0 − δ, x0 + δ) ⊂ I.
Thus, there exists a δ > 0 such that if x ∈ (x0 − δ, x0 + δ) ∩ [a, b], then
− < f (x) − f (x0 ) < .
2.1. REVIEW OF THE RIEMANN INTEGRAL 21
Let
z ∈ (x − δ, x + δ) ∩ [a, b].
We have
1
ωf (z) ≤ ωf ((x − δ, x + δ)) ≤ ωf (I) < .
n
That is, z ∈ DnC . This proves that DnC is open.
1
ωf (x) < , x ∈ J1 ∪ J2 ∪ · · · ∪ JL .
n
Thus, there exists an open interval Ix containing x such that ωf (Ix ) < n1 . So,
{Ix }x∈Ji
is an open cover of the closed and bounded and therefore compact set Ji .
Thus, a finite subcover must cover Ji . Thus, we have a partition of Ji and
1
sup f − inf f <
Ix Ix n
1
ωf (Ji0 ) < .
n
Define
φ = inf f on Ik1 , Ik2 , · · · Ikm , J1 , J2 , · · · JL0 ,
ψ = sup on Ik1 , Ik2 , · · · Ikm , J1 , J2 , · · · JL .
f
Then,
Z b
1X 0 X
[ψ(x) − φ(x)]dx ≤ l(Ji ) + 2B l(Iki )
a n i i
1
≤ (b − a) + 2B
n 4B
< ,
φ≤f ≤ψ
with Z b
[ψ(x) − φ(x)]dx < .
a N
Let
a = x0 < x1 < · · · < xn = b
be a partition associated with φ and ψ and split the collection
So,
Z b X X
[ψ(x)−φ(x)]dx = (ψ−φ)(x)(xk −xk−1 )+ (ψ−φ)(x)(xk −xk−1 ) < .
a D D
N
1 2
P
For D1 , some point of DN is in each subinterval. That is, each subinterval
in D1 contains a point x with ωf (x) ≥ N1 . Thus, in these subintervals
1
ψ − φ ≥ sup − inf ≥ .
f f N
So,
X X
> (xk − xk−1 ).
N D 1 (xk−1 ,xk )∈D1
That is, X
(xk − xk−1 ) < .
(xk−1 ,xk )∈D1
(d) If a < c < b, then f is Riemann integrable on [a, c] and [c, b] and
Z b Z c Z b
f (x)dx = f (x)dx + f (x)dx.
a a c
(a) Prove that fk (x) converges to f (x) on [0, 1] as k → ∞, but not uni-
formly.
R1
(b) Evaluate 0 fk (x)dx for each k ≥ 1.
Proof. Since f 0 is Riemann integrable on [a, b], we have, for every > 0,
step functions φ, ψ such that
Z b Z b Z b
0
φ(x)dx ≤ f (x)dx ≤ ψ(x)dx
a a a
and Z b
(ψ(x) − φ(x))dx < .
a
and n
X
ψ(x) = di χ(xi−1 ,xi ) .
i=1
for some
xi−1 ≤ θi ≤ xi .
26 CHAPTER 2. LEBESGUE INTEGRATION
that is, Z b Z b
φ(x)dx ≤ f (b) − f (a) ≤ ψ(x)dx.
a a
This implies, Z b
f 0 (x)dx = f (b) − f (a).
a
and n
X
φ(x) = ck χEk (x).
k=1
2.2. LEBESGUE INTEGRAL FOR BOUNDED FUNCTIONS ON SETS OF FINITE MEASURE27
where Ei ’s are measurable and mutually disjoint and so are Fj ’s. Suppose
φ(x) is a simple function such that φ(x) = ci for every x ∈ Ei and φ(x) = dj
for every x ∈ Fj . Prove that
n
X m
X
ci m(Ei ) = dj m(Fj ).
i=1 j=1
This shows that the Lebesgue integral of a simple function is well defined.
R
Exercise 2.2.4 Let f (x) be as in Exercise 2.1.17. Evaluate [0,1] f.
Exercise 2.2.5 Let φ and ψ be simple functions on E and m(E) < ∞. Let
k ∈ R. Prove the following statements:
(a) kφ is a simple function on E and
Z Z
k φ= kφ.
E E
(c) If φ ≤ ψ on E, then Z Z
φ≤ ψ.
E E
Exercise 2.2.7 • Show that the above notions are well defined.
• Show that if φ and ψ are simple functions on E, then
Z Z Z Z
φ≤ f ≤ f ≤ ψ.
E E E E
Proof.
Z b Z b
f (x)dx = sup{ φ(x)dx : φ ≤ f, φ a step function}
a a
Z
= sup{ φ : φ ≤ f, φ a step function}
[a,b]
Z
≤ sup{ φ : φ ≤ f, φ a simple function}
[a,b]
Z Z
= f≤ f
[a,b] [a,b]
Z
= inf{ ψ : f ≤ ψ, ψ a simple function}
[a,b]
Z
≤ inf{ ψ : f ≤ ψ, ψ a step function}
[a,b]
Z b
= inf{ ψ(x)dx : f ≤ ψ, ψ a simple function}
a
Z b
= f (x)dx.
a
Thus, Z Z
f= f.
[a,b] [a,b]
φn ≤ φ ≤ f ≤ ψ ≤ ψn . (2.3)
Thus, Z Z
| f| ≤ |f |.
E E
Proof. Parts (a), (b), (d), (e), (f) are left as exercises. To prove (c), let
ψ be a simple function on E such that ψ ≥ f − g. Since f − g = 0 almost
everywhere on E, we get that ψ ≥ 0 almost everywhere on E. Let
E1 = {x ∈ E : ψ(x) ≥ 0}.
Thus, Z Z
(f − g) = inf{ ψ : ψ simple andψ ≥ f − g} ≥ 0.
E E
R R
Similarly,
R E
(g − f ) ≥ 0, that is, by (a), E
(f − g) ≤ 0. We conclude that
E
(f − g) = 0.
Exercise
R 2.2.14 With f and E as above, if m(E) = 0, then show that
E
f = 0.
R
Exercise 2.2.15 If f (x) = r almost everywhere on E, Rshow that E f =
km(E). Thus, if f (x) = 0 almost everywhere on E, then E f = 0.
Remark 2.2.16 Converse of the statement in the above theoremR is not true.
There exist functions which are not zero anywhere on E, but E f = 0. For
example,
R for −1 ≤ x ≤ 1, let f (x) = 1 if x ≥ 0 and −1 otherwise. Then
[−1,1]
f = 0.
R
Theorem 2.2.17 If E
f = 0 and f (x) ≥ 0 on E, then f = 0 almost
everywhere on E.
Let
1
E1 (n) = x ∈ E : f (x) > .
n
If possible, suppose there is a positive integer N such that m(E1 (N )) > 0.
Then Z Z
1
f≥ f ≥ m(E1 (N )) > 0,
E E1 (N ) N
R
which contradicts the fact that E f = 0. Thus,m(E1 (n)) = 0 for all n ≥ 1.
This proves the theorem.
Thus, Z Z Z Z
fn − f = f n − f ≤ |fn − f |
E E E E
Z Z
= |fn − f | + |fn − f |
E−A A
< m(E − A) + 2M <
2m(E) + 1 4M
for all n ≥ N. Thus, Z Z
lim fn = f.
n→∞ E E
2.3. LEBESGUE INTEGRALS OF NONNEGATIVE MEASURABLE FUNCTIONS35
Exercise 2.2.19 Let fn and f satisfy the same conditions as in the theorem
above and fn → f almost everywhere on E. Show that f is Lebesgue integrable
on E and Z Z
f = lim f.
E n→∞ En
Remark 2.3.2 One can readily verify that the Lebesgue integral for simple
functions satisfies all properties listed out in Exercises 2.2.3 and 2.2.5 also
hold when m(E) = ∞.
36 CHAPTER 2. LEBESGUE INTEGRATION
such that
lim φn (x) = f (x), x ∈ E.
n→∞
This shows that the above definition for the Lebesgue integral of a nonneg-
ative, measurable function is independent of the choice of the sequence of
simple functions that converge to the function.
0 ≤ φn ≤ f = lim ψm .
m→∞
Suppose we have mutually disjoint measurable sets {Ei }Ni=1 such that E =
∪N E
i=1 i and nonnegative real numbers ci , 1 ≤ i ≤ N such that
Z N
X
φn = ci m(Ei ).
E i=1
By Exercise 2.2.5(d), which shows the additivity of the integral over disjoint
measurable components of the domain,
Z N Z
X
ψm = ψm .
E i=1 Ei
2.3. LEBESGUE INTEGRALS OF NONNEGATIVE MEASURABLE FUNCTIONS37
Bm = {x ∈ Ei : ψm (x) ≥ αci }.
Bm ⊆ Bm+1 .
lim ψm ≥ φn (x) = ci .
m→∞
Thus,
∞
[
Ei = Bm
m=1
and
lim m(Bm ) = m(Ei ).
m→∞
Thus, Z
lim ψm ≥ lim αci m(Bm ) = αci m(Ei ).
m→∞ Ei m→∞
This holds for any 0 < α < 1. We get Equation (2.4). From this, we deduce
Z N
X Z
lim ψm ≥ lim ( (ci m(Ei )) = φn .
m→∞ E m→∞ E
i=1
Then
lim fn (x) = f (x).
n→∞
Then, Z Z Z
f= lim fk = lim fk .
E E k→∞ k→∞ E
Thus, Z Z
lim fk ≤ f.
k→∞ E E
We now show that Z Z
f ≤ lim fk .
E k→∞ E
let φ be any simple function such that 0 ≤ φ ≤ f. If we are able to show that
Z Z
φ ≤ lim fk ,
E E
40 CHAPTER 2. LEBESGUE INTEGRATION
R
this would imply that lim E fk is an upper bound of the set
Z
{ φ : φ a nonnegative simple function, φ ≤ f }.
E
Thus, Z Z
f ≤ lim fk .
E k→∞ E
In order to show that Z Z
φ ≤ lim fk ,
E E
we observe that
Z N
X [
φ= ci m(Ei ), Ei = E, ci ≥ 0.
E i=1
R
Since E
fk is additive on E, it is sufficient to show that
N Z
X N
X
lim fk ≥ ci m(Ei ).
k Ei
i=1 i=1
This can be done by repeating the same argument used to prove equation
(2.4), by simply replacing φm by fk . This proves the theorem.