Analysis 2 week 2
Quoc Trong Nguyen
February 16, 2025
Note that Theorem 5.3.4 still holds if we replace the assumption that a < s < b by a ≤ s < b.
Definition 0.1. A set A ⊆ R is a null set if for every ε > 0, there exists a sequence I1 , I2 , I3 , . . .
of open intervals covering A such that
∞
X
µ (In ) < ε
n=1
in which µ ((an , bn )) = bn − an .
Lemma 0.2 (Lebesgue’s number lemma). Let X be a compact metric space and let U be an
open cover of X. Then, there exists a positive number δ > 0 (called the Lebesgue number) such
that for every subset of X with diameter less than δ, that subset is entirely contained in at least
one set of the open cover.
Proposition 0.3. A countable union of null sets is a null set.
Proof. Obvious.
Exercise 1
8.
Let ε be given. Choose partitions P1 of [a, c] and P2 of [c, d] such that
U[a,c] (P1 , f, α) − L[a,c] (P1 , f, α) < ε
U[c,b] (P2 , f, α) − L[c,b] (P2 , f, α) < ε.
Now, let P be the partition of [a, b] obtained by merging P1 with P2 i.e. P = P1 ∪ P2 . Then, we
have
U[a,b] (P, f, α)−L[a,b] (P, f, α) = U[a,c] (P1 , f, α) − L[a,c] (P1 , f, α) + U[c,b] (P2 , f, α) − L[c,b] (P2 , f, α) < 2ε.
9.
(i) Suppose f is right continuous at 0. Then f is Riemann–Stieltjes integrable on both [−1, 0] and
[0, 1]. Furthermore
0
f dI1 = 0 (obvious)
−1
and 1
f dI1 = f (0) by Theorem 5.3.4 extended.
0
1
Therefore f is Riemann–Stieltjes integrable on [−1, 1] and
1 0 1
f dI1 = f dI1 + f dI1 = f (0).
−1 −1 0
Next, suppose f is not right continuous at 0. Then there exists A > 0 and a sequence {sn }
converging to 0 from the right such that
|f (sn ) − f (0)| > A
for every n. Now, let P be a partition of [−1, 1]; we can assume 0 ∈ P . Explicitly,
P = {−1 = x0 , x1 , x2 , . . . , xk−1 = 0, xk , . . . , xn = 1}. Since [0, xk ] contains some sn , we have
n
X
U (P, f, α) − L (P, f, α) = (Mi − mi ) ∆αi ≥ Mk − mk ≥ A.
i=1
This shows that f is not Riemann–Stieltjes integrable on [−1, 1] if f is not right continuous at 0.
Thus f is Riemann–Stieltjes integrable if and only if f is right continuous at 0.
(ii) Similar to part (i)
(iii) Similar to part (i)
(iv) Similar to part (1), noting that Theorem 5.3.4 holds true for a < s ≤ b if we let
α(x) = I2 (x − s) in place of α(x) = I(x − s).
10.
Suppose, to get a contradiction, that there is an x′ ∈ [a, b] such that f (x′ ) = A > 0. Since f is
continuous, there exists δ > 0 such that f (x) > A/2 whenever |x − x′ | < δ. Choose a partition
P = {x1 , x2 , . . . , xn } satifying |P | < δ, and let [xk−1 , xk ] be the the interval containing x′ . Then
we have [xk−1 , xk ] ⊆ (x′ − δ, x′ + δ), and therefore
X ∆xk A
L (P, f ) = mi ∆xi ≥ mk ∆xk > > 0.
2
i
This implies that
b
f (x)dx > 0
a
contradicting the assumption that
b
f (x)dx = 0.
a
Exercise 2
(a) Let P = {a = x1 , x2 , . . . , xn = b} be a partition of [a, b]. Since both the set of rational
numbers and irrational numbers is dense in [xi , xi+1 ] for every i, we have
X X
U (P, f ) − L (P, f ) = (Mi − mi ) ∆xi = ∆xi = b − a > 0.
i i
Thus, f is not Riemann integrable.
(b) We claim that f is Riemann integrable. Let ε > 0 be given. We may assume ε sufficiently
small. Choose an integer q sufficiently large, so that (b − a)/q < ε/2. Consider the finite set
nr o
| s < p and (r, s) = 1 .
s
2
Since this set is finite, we can cover it by a finite collection of intervals Ii = [ai , bi ] such that
X ε
(bi − ai ) < .
2
i
Now, let P be a partition of [a, b] containing the ends of those intervals. Divide the index of
elements of P into two sets
i ∈ A if [xi−1 , xi ] ⊆ Ik for some k
i ∈ B otherwise.
Then we have
X X
U (P, f ) − L (P, f ) = (Mi − mi ) ∆xi + (Mi − mi ) ∆xi
i∈A i∈B
X X ∆xi
< ∆xi +
q
i∈A i∈B
X b−a
< (bi − ai ) +
q
i
ε ε
< +
2 2
= ε.
Thus f is Riemann integrable.
(c) Using the same technique as in Part (b), we see that f is integrable in this case.
Exercise 3
Suppose Df ([a, b]) is a null set. Let ε > 0 be given. Put K = sup f − inf f and γ = ε/ (b − a).
Cover Df ([a, b]) by a sequence of open intervals I1 , I2 , I3 , . . . such that
∞
X ε
µ (In ) <
K
n=1
and denote
∞
[
I= In .
n=1
For each x ∈ [a, b] such that x ∈ / Df ([a, b]), choose an open interval Ux containing x satifying
f (Ux ) ⊆ (f (x) − γ/2, f (x) + γ/2) (Here we use the continuity of f at x). The collection of these
open intervals including I covers [a, b]. Since [a, b] is compact, this collection has a finite
refinement A = {I, Ux1 , Ux2 , . . . , UxN } that also covers [a, b] i.e.
N
[
[a, b] ⊆ I ∪ Uxn .
n=1
By Lebesgue’s number lemma, there exists δ > 0 such that if [c, d] is an interval satifying
d − c < δ, then either [c, d] ⊆ I or [c, d] ⊆ Uxn for some n. In other words, [c, d] lies entirely in
some element of A. Now, let P = {a = x0 , x1 , x2 , . . . , xk = b} be a partition of [a, b] satifying
3
|P | < δ, so that either [xi−1 , xi ] ⊆ I or [xi−1 , xi ] ⊆ Uxn for some n. Hence, we can divide the
index of elements of P into two set A and B such that i ∈ A implies [xi−1 , xi ] ⊆ I while i ∈ B
implies [xi−1 , xi ] ⊆ Uxn , and obtain
X X
U (P, f ) − L (P, f ) = (Mi − mi ) ∆xi + (Mi − mi ) ∆xi
i∈A i∈B
X X
<K ∆xi + γ ∆xi
i∈A i∈B
ε
< K + γ (b − a)
K
= 2ε.
Thus, if Df ([a, b]) is a null set then f is Riemann integrable on [a, b].
Conversely, suppose that f is Riemann integrable on [a, b]. Recall that
∞
[
Df ([a, b]) = O1
n
n=1
in which O 1 is the set of ponts at which the oscillation of f is larger than 1/n. Hence, we shall
n
prove that O 1 is a null set for every n, from which and Proposition 0.3 it will follow that
n
Df ([a, b]) is a null set as well.
Fix n, and let ε > 0 be given. Since f is Riemann integrable, there exists a partition
P = {a = x0 , x1 , x2 , . . . , xm = b} satisfying
ε
U (P, f ) − L (P, f ) < .
2n
Let A be the set of all i at which the open interval Ui = (xi−1 , xi ) intersects with O 1 . Then, we
n
have
1 X X ε
(xi − xi−1 ) ≤ (Mi − mi ) (xi − xi−1 ) ≤ U (P, f ) − L (P, f ) <
n 2n
i∈A i∈A
leading to X ε
µ(Ui ) < .
2
i∈A
Cover the elements of P by a finite collection I0 , I2 , . . . , Im such that
m
X ε
µ(Ii ) < .
2
i=0
Clearly
[ m
[
O1 ⊆ ( Ui ) ∪ ( Ii )
n
i∈A i=0
and
X m
X
µ(Ui ) + µ(Ii ) < ε.
i∈A i=0
Thus, O 1 is covered by a finite union of open intervals whose sum of lengths is less than the given
n
ε, it is, therefore, a null set.