Bartle Intro To Real Analysis Solutions CH 10 PDF

Download as pdf or txt
Download as pdf or txt
You are on page 1of 8
At a glance
Powered by AI
The text discusses concepts related to real analysis including integrals, limits, measurable functions, and measurable sets.

The text discusses concepts like the Fundamental Theorem of Calculus, Hake's Theorem, the Comparison Test, the Dominated Convergence Theorem, and properties of measurable sets and functions.

The text defines a function f to be measurable if the preimage of every open set is measurable. It discusses using measurable functions and sequences of functions with the Dominated Convergence Theorem.

Bartle - Introduction to Real Analysis - Chapter 10 Solutions

Section 10.1
Problem 10.1-11. Let f be Dirichlet’s function on [0, 1] and F (x) := 0 for all x ∈ [0, 1]. Since F 0 (x) = f (x) for all
x ∈ [0, 1]\Q, show that the Fundamental Theorem 10.1.9 implies that f ∈ R∗ [0, 1].

Solution: The function F is continuous on [0, 1]. Let E = [0, 1] ∩ Q. Since E is a subset of the rational numbers, E is
countable by Theorem 1.3.9. Clearly, F 0 (x) 6= f (x) for x ∈ E, but F 0 (x) = f (x) everywhere else on [0, 1]. Applying the
Fundamental Theorem of Calculus (Form I) for the generalized Riemann integral, we see that f ∈ R∗ [0, 1].

Problem 10.1-16. Show that the function g2 (x) := (1/x) sin(1/x) for x ∈ (0, 1] and g2 (0) := 0 belongs to R∗ [0, 1].

Solution: Let C2 (x) = x cos(1/x) for x ∈ (0, 1] and C2 (0) = 0. Clearly C2 is continuous on (0, 1]. Since −x ≤ C2 (x) ≤ x
for all x ∈ [0, 1], by the Squeeze Theorem we have limx→0 C2 (x) = 0 = C2 (0). Therefore, C2 is continuous on [0, 1].
For x 6= 0, we have C20 (x) = cos(1/x) − sin(1/x)/x. Let f : [0, 1] → R be f (x) = cos(1/x) for x ∈ (0, 1] and f (0) = 0.
Observe that C20 (x) = f (x) − g2 (x) for x ∈ (0, 1], so two are unequal only at x = 0. Applying the Fundamental Theorem
of Calculus (Form I), we infer that (f − g2 ) ∈ R∗ [0, 1].
From Exercise 7.2-11 (and for the same reasons as in Exercise 7.2.12), we have f ∈ R[0, 1], so f ∈ R∗ [0, 1]. Applying
Theorems 10.1.5(a) and (b), (f + (−1)(f − g2 )) = g2 ∈ R∗ [0, 1].

Problem 10.1-23. If f : [a, b] → R is continuous and if p ∈ R∗ [a, b] does not change sign on [a, b], and if f p ∈ R∗ [a, b], then
Rb Rb
there exists a ξ ∈ [a, b] such that a f p = f (ξ) a p.

Solution: Since f is continuous on [a, b], it has an absolute minimum m and an absolute maximum M on that interval.
By hypothesis, p never goes negative to positive (or vice-versa) on [a, b]. We may assume without loss of generality that
p(x) ≥ 0 for all x ∈ [a, b], in which case mp(x) ≤ f (x)p(x) ≤ M p(x). (If, on the other hand, p(x) ≤ 0 for all x ∈ [a, b],
simple reverse the roles of m and M and proceed with the same argument below).
It follows that there is a c ∈ R such that f (x)p(x) = cp(x) for all x ∈ [a, b]. It must be that m ≤ c ≤ M . Let
x1 , x2 ∈ [a, b] where f (x1 ) = m and f (x2 ) = M . By Bolzano’s Intermediate Value Theorem, there is a ξ between x1 and
x2 such that f (ξ) = c. Therefore:
Z b Z b
f p = f (ξ) f,
a a

where ξ ∈ [a, b].

Section 10.2
Problem 10.2-3. Apply Hake’s Theorem to g(x) := (1 − x)−1/2 for x ∈ [0, 1) and g(1) := 0.


Solution: Let γ ∈ [0, 1). If G(x) = −2 1 − x for x ∈ [0, 1], then G0 (x) = (1 − x)−1/2 = g(x) for x ∈ [0, γ]. Applying
the Fundamental Theorem of Calculus (Form I):
Z γ
1 √ γ p
√ dx = −2 1 − x = 2(1 − 1 − γ).

0 1−x 0

Taking the left-sided limit at 1:


Z γ √ γ
1
lim √ dx = −2 1 − x = 2.

γ→1− 0 1−x 0
R1
By Hake’s Theorem, f ∈ R∗ [0, 1] and 0
(1 − x)−1/2 dx = 2.

Problem 10.2-5. Show that the function g1 (x) := x−1/2 sin(1/x) for x ∈ (0, 1] and g1 (0) := 0 belongs to L[0, 1]. (This
function was also considered into Exercise 10.1.15.)

Solution: We know from Exercise 10.1.15 that g1 ∈ R∗ [0, 1]. From Example 10.1.10, we know that if h(x) = 1/ x for
x ∈ (0, 1] and h(0) = 0, then h ∈ R∗ [0, 1]. Since h(x) ≥ 0 on [0, 1], it follows that h ∈ L[0, 1]. Because |sin(1/x)| ≤ 1
on (0, 1] and g1 (0) = 0 = h(0), we have |g1 (x)| ≤ h(x) for all x ∈ [0, 1]. By the Comparison Test, f ∈ L[0, 1].

Problem 10.2-7. Determine whether the following integrals are convergent or divergent. (Define the integrands to be 0 where
they are not already defined.)
R 1 xdx R1
(a) 0 sin
x3/2
. (b) 0 cos xdx
x3/2
. ···

Solution: Part (a)


Let f (x) = sin x/x3/2 for x ∈ (0, 1] and f (0) = 0. If γ ∈ (0, 1), wee that f ∈ R∗ [γ, 1] because f is continuous on
[γ, 1]. √
Now let ω(x) = 1/ x for x ∈ (0, 1] and ω(0) = 0. Since |sin x| ≤ x for all x ∈ [0, 1], it follows that f (x) ≤ x/x3/2 =
ω(x) for all x ∈ (0, 1]. From Example 10.1.10(a), we know ω ∈ R∗ [0, 1]. Consequently, we have:
Z 1 Z 1
sin x 1
0 ≤ lim+ ≤ lim+√ = 2.
γ→0 γ x3/2 γ→0γ x
R1 R1
Since |ω| = ω, we have ω ∈ L[0, 1]. Because 0 ω converges, it follows that 0 f converges.
Part (b) Let f (x) = cos x/x3/2 and f (0) = 0, and let ω(x) = (cos 1)/x3/2 for x ∈ (0, 1] and ω(0) = 0. Observe that
f (x) ≥ ω(x) for all x ∈ [0, 1]. If f ∈ R∗ [0, 1], then:
Z ω Z γ
cos x 1
lim ≥ lim+ .
γ→0+ 0 x3/2 γ→0 0 x 3/2
Rγ √ Rγ
But 0 ω is divergent (apply the Fundamental Theorem of Calculus with Ω(x) = −2/ x for x ∈ (0, 1]), so limγ→0+ 0 f
does not exist. By Hake’s Theorem, f ∈ / R∗ [0, 1].

Problem 10.2-11. (a) Give an example of a function f ∈ R∗ [0, 1] such that max{f, 0} does not belong to R∗ [0, 1].
(b) Can you give an example of f ∈ L[0, 1] such that max{f, 0} ∈
/ L[0, 1].

Solution: Part (a) Let (an ) be a sequence where an = (−1)n−1 1/n for all n ∈ N. Clearly,
P
an is convergent. Now
define φ : [0, 1] → R as:
 k
2 ak for 1 − (1/2)k−1 ≤ x ≤ 1 − (1/2)k ,
φ(x) =
0 for x = 0
We then have: Z γ n
X 1
lim φ = lim (−1)k−1 = ln 2.
γ→1− 0 n→∞ k
k=1

Now let ρ(x) = max{φ(x), 0}. This is equivalent to having a2n−1 = 0 and a2n remaining unchanged for all n ∈ N. It
follows that: Z γ n n
X 1 1 X 1
lim ρ = lim = lim .
γ→1 −
0 n→∞ 2k 2 n→∞ k
k=1 k=1

The summation is a harmonic series, which diverges. Therefore, ρ ∈


/ L[0, 1].
Part (b) No, there is no such example. Let f ∈ L[0, 1] be given. Clearly, g(x) = 0 for all x ∈ [0, 1] is a Lebesque
integrable function. Let h(x) = max{f (x), g(x)} = max{f (x), 0}. It follows from Theorem 10.2.6 that h ∈ L[0, 1].

Problem 10.2-13. Write out the details of the proofs of Theorem 10.2.11.

Solution: For each part, let f, g, h ∈ L[a, b] be given.


Part (a) Since dist(f, g) = kf − gk and the seminorm is always greater than zero, it follows that dist f, g ≥ 0.
Rb
Part (b) Suppose f (x) = g(x) for x ∈ [a, b]. It follows that dist(f, g) = kf − gk = a |f − g| = 0.
Part (c) We have:

Page 2
Z b Z b
dist(f, g) = kf − gk = |f − g| = |g − f | = kg − f k = dist(g, f ).
a a

Part (d) From Theorem 10.2.5:


Z b Z b Z b Z b
|f − h| = |(f − g) + (g − h)| ≤ |f − g| + |g − h| .
a a a a

It follows by definition that kf − hk ≤ kf − gk + kg − hk.

Problem 10.2-14. Give an f ∈ L[a, b] with f not identically 0, but such that kf k = 0.

Solution: Let f : [0, 1] → {0, 1} be Dirichlet’s function. Since f (x) = 1 for x ∈ [0, 1] ∩ Q, the function f is not uniformly
0. Let F : [0, 1] → {0} where F (x) = 0 for all x ∈ [0, 1]. It follows that F 0 (x) = f (x) for x ∈ [0, 1]\Q and F 0 (x) 6= f (x)
R1
for x ∈ [0, 1] ∩ Q, which is countable. By the Fundamental Theorem of Calculus, 0 f = F (1) − F (0) = 0. Since f = |f |,
we infer that f ∈ L[a, b] and kf k = 0.

Problem 10.2-15. If f, g ∈ L[a, b], show that |kf k − kgk| ≤ kf ± gk.

Solution: First we will establish a preliminary result that has not yet been proven in Introduction to Analysis (although it
is used without proof in some of the text’s examples).
Theorem 1. (Reverse Triangle Inequality) If x, y ∈ R, then | |x| − |y| | ≤ |x − y|.

Proof. While there is likely a much more elegant solution, I’ll use a brute force approach to prove the result by cases. If
x ≥ 0 and y ≥ 0, then | |x| − |y| | = |x − y|. If x ≥ 0 and (−y) > 0 (i.e., y < 0), then | |x| − |−y| | = |x − y| ≤
|x + y| = |x − (−y)|. If (−x) > 0 and y ≥ 0, then | |−x| − |y| | = |x − y| ≤ |x + y| = |(−x) − y|. If (−x), (−y) > 0,
then | |−x| − |−y| | = |x − y| = |(−1)(−x − (−y)| = |(−x) − (−y)|. As this covers all possible cases, it follows that
| |x| − |y| | ≤ |x − y| for all x, y ∈ R.

With this result in hand, the rest of the problem is straightforward. Observe that by Theorem 10.2.5, since f, g ∈ L[a, b]
it follows that (|f | − |g|), (f ± g) ∈ L[a, b]. By the Reverse Triangle Inequality, | |f (x)| − |g(x)| | ≤ |f (x) − g(x)| for
x ∈ [a, b]. Applying Theorem 10.2.4, we have:
Z Z Z b Z b
b b
(|f | − |g|) = |f | − |g| ≤ |f − g| .

a a a a

By definition, | kf k − kgk | ≤ kf − gk.


By the Triangle Inequality, | |f (x)| − |g(x)| | ≤ |f (x)| + |−g(x)| = |f (x)| + |g(x)| for x ∈ [a, b]. By Theorem 10.2.3:
Z Z Z b Z b
b b Z b
(|f | − |g|) = |f | − |g| ≤ |f | + |g| .

a a a a a

By definition, | kf k − kgk | ≤ kf + gk. It follows that | kf k − kgk | ≤ kf ± gk.

Problem 10.2-16. Establish the easy part of the Completeness Theorem 10.2.12.

Solution: Suppose (fn ) is a sequence where fn ∈ L[a, b] for all n ∈ N and lim fn = f ∈ L[a, b]. By the Cauchy Criterion
for Sequences, for /(b−a) > 0 there is an H ∈ N where if m > n ≥ H, then |fm − fn | < /(b−a). Since fm , fn ∈ L[a, b],
by Theorem 10.2.5 (fm − fn ) ∈ L[a, b]. We can then integrate both sides of the preceding inequality and apply Theorem
10.2.4:
Z b Z b

|fm − fn | < = .
a a b−a
By definition, dist(fm , fn ) = kfm − fn k < .

Page 3
Problem 10.2-18. Let gn (x) := −1 for x ∈ [−1, −1/n), let gn := nx for x ∈ [−1/n, 1/n], and let gn (x) := 1 for x ∈ (1/n, 1].
Show that kgm − gn k → 0 as m, n → ∞, so that the Completeness Theorem 10.2.12 implies that there exists g ∈ L[−1, 1]
such that (gn ) converges to g in L[−1, 1]. Find such a function g.

Solution: Let  > 0 be given, H() = , and m, n ≥ H(). If m = n, then kgm − gn k = 0 < . Now suppose m > n.
Observe that gm (x) = gn (x) for x ∈ [−1, −1/n) ∪ (1/n, 1]. The seminorm then reduces to:
Z 1 Z −1/m Z 1/m Z 1/n
kgm − gn k = |gm − gn | dx = |−1 − nx| dx + |(m − n)x| dx + |1 − nx| dx.
−1 −1/n −1/m 1/m
R −1/m R 1/m
By symmetry, −1/n
|−1 − nx| dx = 1/n
|1 − nx| dx. We then have:
Z 1/m 1/n
1 1 2 n
|1 − nx| dx = 2(x − nx2 )
2 = − + 2.

1/n 2 1/m n m m
R0 R 1/m
Also by symmetry, −1/m |(m − n)x| dx = 0 |(m − n)x| dx. We then have:
Z 1/m 1/m 1 n
2 |(m − n)x| dx = (m − n)x2 = − .

0 0 m m2
Putting everything together, we get:
Z 1
1 1 1
kgm − gn k = |gm − gn | dx = − < = .
−1 n m n
Applying the Completeness Theorem, we infer that lim gn = g ∈ L[−1, 1].
The sequence (gn ) converges to g(x) = −1 for x ∈ [−1, 0) and g(x) = 1 for x ∈ [0, 1].

Section 10.3
Problem 10.3-3. Let f and |f | belong to R∗ [a, γ]R for all γ ≥ a. Show that f ∈ L[a, ∞) if and only if for every  > 0 there
q
exists a K() ≥ a such that if q > p > K(), then p |f | < .


Solution: Suppose f ∈ L[a, ∞). From Hake’s Theorem, it follows that limγ→∞ a
|f | = A for some A ∈ R. Let  > 0
be given. Then there is a K(/2) > a such that if q > p > K(/2), then:
Z p Z p

|f | − A = A − |f | < , (1)

a

a 2
Z q

|f | − A < . (2)

a 2
Adding (1) and (2) and using the Triangle Inequality and the Additivity Theorem, we have:
Z q Z q   Z p  Z q Z p
 
|f | =
|f | − A + A − |f | ≤
|f | − A + A −
|f | < + = ,
p a a a a 2 2
as we sought to show. Rq
Conversely, suppose for every  > 0 there is a K() ≥ a such that if q > p > K(), then p |f | < . Since
− |f | ≤ f ≤ |f |, we have (under the foregoing conditions):
Z q Z p Z q Z q


f− f = f ≤ |f | < .
a a p p

Let (xn ) be a sequence on (a, ∞) where lim xn = ∞. Let  > 0 be given. Because (xn ) is properly divergent, there is
an H(K()) ∈ N such that if n ≥ H(K()), R x thenR x K(). Let m > n > H(K()). Clearly, either xm ≥ xn > K()
n >
x
or xn ≥ xm > K(). If xm = xn , then a m f − a n f = 0 <  in all cases. If xm 6= xn , we may assume without loss of
generality that xm > xn (if this is not true, simply reverse the order of the integrals in (3) below, which will not change
the absolute value). By hypothesis, we then have:

Page 4
Z
xm Z xn


f− f < . (3)
a a
Rx
By the Cauchy Criterion for sequences, ( a n f ) converges. Because this is true for any arbitrary sequence (xn ) with

lim xn = ∞, we infer from the sequential criterion for limits that limγ→∞ a f exists. By Hake’s Theorem, f ∈ R∗ [a, ∞).

Following the same argument, we can show that limγ→∞ a |f | exists, so |f | ∈ R∗ [a, ∞). Therefore, f ∈ L[a, ∞).

Problem 10.3-7. If f is continuous on [1, ∞) and if |f (x)| ≤ K/x2 for x ∈ [x, ∞), show that f ∈ L[a, ∞).

Solution: Since |f | = f and f is continuous on [1, γ] for γ ∈ (1, ∞), it follows that f, |f | ∈ R∗ [1, γ]. By hypothesis and
Theorem 10.1.5(c):
Z γ Z γ  
1 1
|f | ≤ K 2
=K 1− .
1 1 x γ

Taking the limit of the right-hand side, we see 1 |f | is bounded above by K as γ approaches infinity. Let (γn ) be a

sequence where lim γn = ∞. Observe that since |f | ≥ 0, it follows that ( a n |f |) is increasing. Because this sequence is

bounded, we infer from the Monotone Convergence Theorem that ( a n |f |) converges to some A ∈ R. By Hake’s Theorem,
|f | ∈ R∗ [a, ∞). It follows that f ∈ L[a, ∞).

R∞ √
Problem 10.3-14. Show that the integral a
(1/ x) sin xdx converges. [Hint: Integrate by Parts.]


Solution: We will follow pretty much the same argument as with Example 10.3.4(d). Let f (x) = sin x/ x for x ∈ (0, ∞)
and f (0) = 0, so f is continuous on [0, ∞) and f ∈ R∗ [0, γ] for γ ∈ (0, ∞). For 0 < β < γ, it follows that f ∈ R∗ [β, γ].
We may integrate by parts and take the absolute value:
Z γ
sin x cos x γ 1 γ cos x
Z

√ dx = √ + dx
x x β 2 β x3/2

β

1 γ 1 γ 1 1 γ 1 γ
Z
3
≤ √ + dx = √ + √ < √ .
x β 2 β x3/2 x β 2 x β 2 β
where we have relied upon |cos x| ≤ 1.
Let  > 0 be given. If K() = 9/(42 ), then if γ < β < K(), then:
Z γ
sin x
√ dx < .

βx

By exercise 10.3-2, f ∈ R [a, ∞). Therefore, the integral converges.

R∞
Problem 10.3-15. Establish the convergence of Fresnel’s Integral a
sin(x2 )dx. [Hint: Use the Substitution Theorem
10.1.12.]

Solution: With Exercise 10.3-14 under our belts, this exercise is easy. Let ϕ(x) = x2 on [a, ∞), so the integrand is
sin(ϕ(x)). Applying the Substitution Theorem, we have:
Z ∞
1 ∞ sin ϕ
Z
sin(x2 ) = √ dϕ.
a 2 √a ϕ
From Exercise 10.3-14, we infer that the integral convergences.

Section 10.4
1
Problem 10.4-1(c). [See problem prompt] 1+xk
, [0, 1].

Page 5
Solution: Let (fk ) be a sequence where fk (x) = 1/(1 + xk ) for all k ∈ N and all x ∈ [0, 1]. Observe that fk is continuous
on [0, 1] for all k, so fk ∈ R∗ [0, 1]. We can evaluate the limit of the sequence as follows for any x ∈ [0, 1]:
  
1 1 for x ∈ [0, 1)
f (x) = lim = 1
1 + xk 2 for x = 1.
For A1 = [0, 1], we have:

1
= xk = 1 ,

kfk − f kA1 =
1 + xk − 1 1 + xk

A1 A1 2
and for A2 = {1}, we have:

1 1
kfk − f kA2 = −
2 2 = 0.

A2

For A = A1 ∪ A2 , it follows that lim kfk − f kA = 1/2. Consequently from Lemma 8.1.8, (fk ) is not uniformly
convergent on [0, 1].
However, (fk ) is increasing. Let j, k ∈ M with j < k. If x ∈ [0, 1], then xk ≥ xj . Therefore, fk (x) = 1/(1 + xk ) ≤
R1
1/(1 + xj ) = fj (x). In addition, (fk ) is bounded by 1; it follows from Theorem 10.1.5(c) that 0 fk is bounded by
1. Applying the Monotone Convergence Theorem 10.4.4, we infer that f ∈ R∗ [0, 1]. If F (x) = x for x ∈ [0, 1], then
F 0 (x) = f (x) for x ∈ [0, 1]\{1}. Applying the Fundamental Theorem of Calculus, we have:
Z 1 Z 1
lim fk (x)dx = f = F (1) − F (0) = 1.
0 0

Problem 10.4-3. Discuss the following sequences of functions and their integrals on [0, 1]. Evaluate the limits of the integrals,
when possible.
(a) e−kx · · · (c) kxe−kx · · · .

Solution: Part (a) Let (fk ) be a sequence of functions where fk (x) = e−kx . We then have f (0) = lim fk (0) = 1. For
x ∈ (0, 1], we have, f (x) = lim e−kx = lim(1/ex )k = 0. Accordingly,

1 for x = 0,
f (x) =
0 for x ∈ (0, 1].

For A1 = {0}, we have kfk − f kA1 = k1 − 1kA1 = 0. For A2 = (0, 1], we have kfk − f kA2 = e−kx A2 = 1. For
A = [0, 1], we have lim kfk − f kA = 1 6= 0, so (fk ) is not uniformly convergent on [0, 1] by Lemma 8.1.8.
Observe that all fk are continuous on [0, 1] and therefore fk ∈ R∗ [0, 1]. Further, all fk are non-negative and bounded
R1
by 1 on [0, 1], so by Theorem 10.1.5(c), 0 fk is bounded by 1. The sequence (fk ) is decreasing because for any x ∈ [0, 1]
and j, k ∈ N, if j < k, then fk (x) = e−kx ≥ e−jk = fj (x). Applying the Monotone Convergence Theorem 10.4.4, we infer
that f ∈ R∗ [0, 1]. If we let F (x) = 0 for x ∈ [0, 1], then F 0 (x) = f (x) on x ∈ [0, 1]\{0}. By the Fundamental Theorem
of Calculus and the Monotone Convergence Theorem:
Z 1 Z 1
lim fk = f = 0.
0 0

Part (c)
Let (gk ) be a sequence where gk = kxe−kx for all k ∈ N. Since each gk is continuous on [0, 1], it follows that
gk ∈ R∗ [0, 1]. Clearly, g(0) = lim gk (0) = 0. For x ∈ (0, 1], if we apply L’Hopı̂tal’s Rule:
kx x
g(x) = lim = lim kx = 0.
ekx xe
Accordingly, g(x) = 0 for x ∈ [0, 1].
For each k, we see that gk is not monotone but rather has an absolute maximum of 1/e for every term of the sequence.
The first derivative of gk is gk0 (x) = ke−kx (1 − kx), which has a root at x0 = 1/k. For all k ∈ N, we have gk (x0 ) =
√1/e.
In√addition, the sequence (gk ) is not monotone on [0, 1]. Let k ∈ N be given. It follows that gk (1/(2k)) = 1/(2 e) =
1/e · e/2 < 1/e = g2k (1/(2k). On the other hand, gk (1/k) = 1/e > 1/e · 2/e = 2/e2 = g2k (1/e). By definition, (gk ) is
not monotone. Therefore, the Monotone Convergence Theorem does not apply.

Page 6
We can, however, use the Dominated Convergence Theorem 10.4.5. Let α(x) = xe−kx and ω(x) = 1/e for x ∈ [0, 1].
Clearly α(x) ≤ gk (x) ≤ ω(x) for all k ∈ N; further, α, ω ∈ L[0, 1]. By the Dominated Convergence Theorem, f ∈ L[0, 1],
and:
Z 1 Z 1
lim gk = g = 0.
0 0

Problem 10.4-6. Let (fk ) be a sequence on [a, b] such that each fk is differentiable on [a, b] and fk0 (x) → g(x) with |fk0 (x)| ≤ K
for all x ∈ [a, b]. Show that the sequence (fk (x)) either converges for all x ∈ [a, b] or it diverges for all x ∈ [a, b].

Solution: Since every fk is diffentiable on [a, b], it follows that each fk is continuous on [a, b]. By the Fundamental
Theorem of Calculus, fk0 ∈ R∗ [a, b]. Moreover, for any α, β ∈ [a, b], it must be that fk0 ∈ R∗ [α, β]. Let γ(x) = −K and
ω(x) = K, each of which is clearly generalized Riemann integrable. Since γ(x) ≤ fk0 (x) ≤ ω(x) for all x ∈ [α, β], by the
Rβ Rβ
Dominated Convergence Theorem g ∈ R∗ [α, β] and lim α fk0 = α g.
If (fk (ξ)) is divergent for all ξ ∈ [a, b], then clearly (fk ) diverges on all of [a, b]. Now suppose instead that (fk (ξ))
convergesR to f (ξ) for some ξ ∈ [a, b]. Again applying the Fundamental Theorem of Calculus, for any x ∈ [a, b] we have
x
fk (x) = ξ fk0 + fk (ξ). Taking the limit, we have:
Z x  Z x
0
f (x) = lim fk (x) = lim fk + fk (ξ) = g + f (ξ).
ξ ξ

Because x is arbitrary, it follows that (f (x)) converges for all x ∈ [a, b] if it converges at any point on [a, b].

Problem 10.4-7. If fk are the functions in Example 10.4.6(a), show that sup{fk } does not belong to R∗ [0, ∞).

Solution: As in Example 10.4.6, for all k ∈ N, let g : [0, ∞) → R be g(x) = sup{fk (x) : k ∈ N}. We see that g is the
following step function:


 1 for x ∈ [0, 1],



 1/2 for x ∈ (1, 2],
 1/3

 for x ∈ (2, 3],
g(x) = ..

 .
1/k for x ∈ (k − 1, k], k ∈ N,





 ..
.

Let γ ∈ (0, ∞) be given. Because g is a series of step functions, g ∈ R∗ [0, γ]. If n = bγc, then:
γ n
1 γ−n
Z X
g= + .
0 k n+1
k=1

Taking the limit, we have:


Z γ ∞
X 1
lim g= .
γ→∞ 0 k
k=1

Because the geometric series on the right-hand side diverges, the limit does not exist. Applying Hake’s Theorem, we
/ R∗ [0, ∞).
infer that g ∈

Problem 10.14. (a) If (fk ) is a bounded sequence in M[a, b] and fk → f a.e., show that f ∈ M[a, b]. [Hint: Use the
Dominated Convergence Theorem.]
(b) If (gk ) is any sequence in M[a, b] and if fk := arctan ◦gk , show that (fk ) is a bounded sequence in M[a, b].
(c) If (gk ) is any sequence in M[a, b] and if gk → g a.e., show that g ∈ M[a, b].

Solution: Part (a) Suppose there is an M > 0 such that |fk | ≤ M for all k ∈ N. Let α(x) = −K and ω(x) = K for
all x ∈ [a, b]. Clearly α, ω ∈ M[a, b] and α(x) ≤ fk (x) ≤ ω(x) for all k ∈ N. By the Integrability Theorem 10.4.12, each

Page 7
fk ∈ R∗ [a, b]. It follows from the Dominated Convergence Theorem that f ∈ R∗ [a, b]. From the Measurability Theorem,
we infer that f ∈ M[a, b].
Part (b) Since arctan is continuous on R, by Theorem 10.4.9 fk = arctan ◦gk is measurable on any interval [a, b].
Since |arctan x| < π/2 for all x ∈ R, it follows that |fk (x)| < π/2 for all x ∈ R and all k ∈ N. The sequence is therefore
bounded.
Part (c) A straightforward application of Theorem 10.4.9(c) shows that g ∈ M[a, b].

Problem 10.17. [See problem]

Rb
Solution: Part (a) If E = ∅, then 1E (x) = 0 for x ∈ [a, b]. It follows that m(E) = a
0 = 0.
Rb
Given any E ∈ M[a, b], it must be that 0 ≤ 1E (x) ≤ 1 for all x ∈ [a, b]. From Theorem 10.1.5(c), 0 ≤ a 1E =
m(E) ≤ b − a.
Part (b) Let E1 = [c, d], E2 = [c, d), E3 = (c, d], and E4 = (c, d). Let F : [a, b] → R where F (x) = x for all
x ∈ [c, d]. It follows that F 0 (x) = 1E1 on [c, d], F 0 (x) = 1E2 on [c, d]\{d}, F 0 (x) = 1E3 on [c, d]\{c}, and F 0 (x) = 1E4
on [c, d]\{c, d}. In each case, the inequality is over a finite (and therefore countable) set. By the Fundamental Theorem of
Calculus, for each 1 ≤ k ≤ 4:
Z d
m(Ek ) = 1Ek = F (d) − F (c) = d − c.
c

Part (c) Observe that 1E 0 (x) = 1 − 1E (x). Consequently:


Z b Z b Z b
m(E 0 ) = (1 − 1E ) = 1− 1E = (b − a) − m(E).
a a a

Part (d) Observe that 1E∪F = max{1E , 1F } = 1/2(1E + 1F + |1E − 1F |) and 1E∩F = max{1E , 1F } = 1/2(1E +
1F − |1E − 1F |). Accordingly:
Z b 
1 1
m(E ∪ F ) + m(E ∩ F ) = (1E + 1F + |1E − 1F |) + (1E + 1F − |1E − 1F |)
a 2 2
Z b
= (1E + 1F ) = m(E) + m(F ).
a

Part (e) Since E ∩ F = ∅, it follows from part (a) that m(E ∩ F ) = m(∅) = 0. From part (d):

m(E ∪ F ) = m(E) + m(F ) − m(E ∩ F ) = m(E) + m(F ).


Part (f) Suppose (Ek ) in an increasing sequence of sets on M[a, b] (meaning if k ∈ N, then Ek ⊆ Ek+1 ). Since
(Ek ) is bounded
S∞by [a, b], it follows from the Monotone Convergence Theorem for sequences that (Ek ) converges to
E = lim Ek = k=1 Ek . It follows from Exercise 10.4-15(e) that (1Ek ) is an increasing sequence of functions because for
any k ∈ N and any x ∈ [a, b], it follows that 1Ek+1 (x) ≥ 1Ek (x). Since 1Ek is bounded, it follows that (1Ek ) converges to
1E = lim 1Ek .
Rb
We know from Exercise 10.4-15(a) and 10.4-16(a) that each 1Ek ∈ R∗ [a, b] and a 1Ek is bounded. From the Monotone
Convergence Theorem 10.4.4, 1E ∈ R∗ [a, b]. Consequently:

[ Z b Z b
m(E) = m( Ek ) = 1E = lim 1Ek = lim m(Ek ).
k=1 a a
Sk
Part (g) Let (Dk ) be a sequence where Dk = j=1 Ck for all k ∈ N. Clearly (Dk ) is increasing because each
Dk ⊆ Dk+1 .
Pk
We may show by induction that m(Dk ) = j=1 m(Cj ). If k = 1, then m(D1 ) = m(C1 ). Now assume the inductive
hypothesis for some k ∈ N. It follows that Dk+1 = Dk ∪Ck+1 . Since Dk ∩Ck+1 = ∅ (because Ck+1 is disjoint with all other
Pk+1
terms of (Ck )), we infer from part (e) and the inductive hypothesis that m(Dk+1 ) = m(Dk ) + m(Ck+1 ) = j=1 m(Cj ).
Our proposition therefore holds for all k ∈ N. S∞ S∞
Because (Dk ) is increasing, from part(f) we see that D = lim Dk = j=1 Dj = j=1 Cj . Also from part(f), we see
P∞
that m(D) = lim m(Dk ) = j=1 Cj .

Page 8

You might also like