MATH2060B Mathematical Analysis II Tutorial 8

Suggested Solution of Exercises on Riemann Integration

Question 1 (2018-19 Final Q2). Define a function g : [0, π/2] → R by
(
cos2 x, if x ∈ Q,
g(x) =
0, otherwise.
Find the upper and lower Riemann integrals of g over [0, π/2]. Is it Riemann integrable?
Solution. Let’s find the lower and upper integrals of g.
• Lower integral: Let P be any partition of [0, π/2]. Note that each subinterval
[xi−1 , xi ] must containing some irrational number, so
mi (g, P ) = 0, ∀i = 1, ..., n.
It follows that the lower sum is given by
n
X
L(g, P ) = mi (g, P )∆xi = 0.
i=1

Taking infimum over all partition P , the lower integral of g is given by

Z π/2
g = 0.
0

• Upper integral: Let P be any partition of [0, π/2]. Note that cos2 x is decreasing on
each subinterval [xi−1 , xi ] and rational numbers are dense, so
Mi (g, P ) = cos2 (xi−1 ), ∀i = 1, ..., n.
Consider f : [0, π/2] → R defined by f (x) = cos2 x. We also have
Mi (f, P ) = cos2 (xi−1 ), ∀i = 1, ..., n.
It follows that
n
X n
X
U (g, P ) = Mi (g, P )∆xi = Mi (f, P )∆xi = U (f, P ).
i=1 i=1

Since f and g have the same upper sum over arbitrary partitions of [0, π/2], they have
the same upper integral, hence
Z π/2 Z π/2 Z π/2 Z π/2 Z π/2
2 1 + cos 2x π
g= f= f= cos xdx = dx = .
0 0 0 0 0 2 4

In summary, the lower and upper intgral of g is given by

Z π/2 Z π/2
π
g = 0 and g= .
0 0 4
Since they are unequal, g is not Riemann integrable.

Prepared by Ernest Fan 1

MATH2060B Mathematical Analysis II Tutorial 8

Question 2 (2016-17 Midterm Q4). Define a funtion f on [0, 1] by

(
1, if x = 1/n for some n ∈ N,
f (x) =
0, otherwise.
Z 1
Show that f is Riemann integrable and find f.
0

Solution. For each natural number N ≥ 2, define the partition PN of [0, 1] by

 
1 1 1 1
PN = 0, ± δ, ± δ, ..., ± δ, 1 − δ, 1 , where 2δ < .
N N −1 2 N (N − 1)
(Visualize this partition!) This ensures that
1 1 1 1 1 1
0< −δ < +δ < −δ < + δ < · · · < − δ < + δ < 1 − δ < 1.
N N N −1 N −1 2 2
Note that each subinterval [xi−1 , xi ] must contain some numbers that cannot be represented
by the reciprocal of some natural number, so
mi (f, PN ) = 0, ∀i = 1, ..., n
On the other hand, note that we have (Why?)
(
1, if i = 1 or even
Mi (f, PN ) = .
0, otherwise

Therefore, the lower sum and upper sum can be calculated by

n
X
L(f, PN ) = mi (f, PN )∆xi = 0
i=1
n   X N
X 1 1
U (f, PN ) = Mi (f, PN )∆xi = −δ + 2δ + δ = + 2(N − 1)δ
i=1
N k=2
N

It follows that
1 1
N −1
Z Z
1 2
0 = L(f, PN ) ≤ f≤ f ≤ U (f, Pn ) < + = .
0 0 N N (N − 1) N
Since n ≥ 2 is arbitrary, letting N → ∞ in the above inequality, we have
Z 1 Z 1
0≤ f≤ f ≤ 0.
0 0

It forces the lower and upper integral of f equal zero. Thus f is Riemann integrable with
Z 1
f = 0.
0

Prepared by Ernest Fan 2

MATH2060B Mathematical Analysis II Tutorial 8

Question 3 (2017-18 Final Q2). .

(i) Define a function f : [0, ∞) → [0, ∞) by
(
1, if x ∈ [n, n + 1/2n ) for some n ∈ N,
f (x) =
0, otherwise.
Z ∞
Show that the improper integral f (x)dx exists but lim f (x) does not exist.
0 x→∞
Z ∞
(ii) Let f be a non-negative R-valued function defined on [0, ∞). Suppose that f (x)dx
0
is convergent and lim f (x) = L. Show that L = 0.
x→∞

Solution. .
(i) The fact that f ∈ R[0, T ] for all T > 0 is left as an exercise. (You need to find a
partition P of [0, T ] for each ε > 0 such that U (f, P ) − L(f, P ) < ε. It is tedious but
the technique used is the same. You usually don’t need to provide the proof for such
questions if you have more to do.)
Notice that for each n ∈ N,
Z n n−1 Z k+1 n−1
X X 1 1
f (x)dx = f (x)dx = k
= 1 − n−1 .
0 k=0 k k=1
2 2

Fix T > 0 and let N to be a natural number such that N ≤ T < N + 1. Since f is
non-negative, we have
Z N Z T Z N +1
1 1
1 − N −1 = f (x)dx ≤ f (x)dx ≤ f (x)dx = 1 − N .
2 0 0 0 2
Z ∞ Z T
Since N → ∞ as T → ∞, by Squeeze Theorem, f (x)dx = lim f (x)dx = 1.
0 T →∞ 0
Now consider the sequences (xn ) and (yn ) defined by
1
xn = n and yn = n + , ∀n ∈ N.
2n
Note the both (xn ) and (yn ) diverges properly to ∞, but f (xn ) = 1 for all n and
f (yn ) = 0 for all n. Hence lim f (x) does not exist.
x→∞

(ii) Since f is non-negative, we must have L ≥ 0. Suppose on a contrary that L > 0. Then
there exist T > 0 such that f (x) ≥ L/2 for all x > T . Then for each M > 0, take
A > max{T, M }, we have
Z A+1 Z A+1
L
f (x)dx = f (x)dx ≥ > 0.
A A 2
Z ∞
It follows by Cauchy Criterion that f (x)dx is divergent. It is a contradiction.
0

Prepared by Ernest Fan 3

MATH2060B Mathematical Analysis II Tutorial 8

Question 4 (2016-17 Final Q2). Let f be a function defined by

sin x
f (x) = , for x ≥ 1.
x
Z ∞
(i) Show that the integral f (x)dx is convergent.
1
Z ∞
(ii) Show that the integral |f (x)|dx is divergent.
1
Solution. .
(i) Fix T > 1. Note that by Integration by Parts,
Z T Z T h cos x iT Z T cos x
sin x
f (x)dx = dx = − − dx.
1 1 x x 1 1 x2
Z ∞
cos x
Hence it suffices to show that the improper integral dx converges.
1 x2
Note that for any A2 > A1 > 1, we have
Z A2 Z A2 Z A2
cos x | cos x| 1 1 1 1
2
dx ≤ 2
dx ≤ 2
dx = − ≤ .
A1 x A1 x A1 x A1 A2 A1
Hence for any ε > 0, we can choose M > 1 such that 1/M < ε. Then whenever
A2 > A1 > M , Z A2
cos x 1 1
2
dx ≤ < < ε.
A1 x A1 M
Z ∞
cos x
It follows by Cauchy Criterion that dx converges.
1 x2
(ii) Note that since |f (x)| is non-negative, we have
Z ∞ Z (N +1)π N Z (k+1)π
| sin x| X | sin x|
|f (x)|dx ≥ dx = dx.
1 π x k=1 kπ x
For each k ∈ N, we can substitute x = t + kπ, then
Z (k+1)π Z π Z π Z π
| sin x| | sin(t + kπ)| sin t 1
dx = dt = dt ≥ sin tdt.
kπ x 0 t + kπ 0 t + kπ (k + 1)π 0
Z π
Write A = sin tdt, we have
0
Z ∞ N N
X A AX 1
|f (x)|dx ≥ = .
1 k=1
(k + 1)π π k=1 k + 1
Since the aboveZ inequality holds for all N ∈ N and the harmonic series diverges to ∞,
∞
it follows that |f (x)dx is also divergent.
1
Remark. Integration by parts is a consequence of the Product Rule and the Funda-
mental Theorem of Calculus. You may try to prove it as an exercise.

Prepared by Ernest Fan 4

MATH2060B Mathematical Analysis II Tutorial 8

Question 5 (2018-19 Final Q3). Let f be a continuous function on [a, b] and ϕ : [α, β] → R
be continuously differentiable such that ϕ(α) = a and ϕ(β) = b. Show that
Z b Z β
f (x)dx = f (ϕ(t))ϕ0 (t)dt
a α
Ru
(Hint: Consider the functions F (u) = a
f (x)dx and H(t) = F (ϕ(t)).)

Solution. Define the functions F : [a, b] → R and H : [α, β] → R by

Z u
F (u) = f (x)dx and H(t) = F (ϕ(t)).
a

Since f is continuous, we have F 0 = f by the Fundamental Theorem of Calculus. Also,

by Chain Rule, we have H 0 (t) = F 0 (ϕ(t))ϕ0 (t) = f (ϕ(t))ϕ0 (t). It follows again by the
Fundamental Theorem of Calculus that
Z b Z β Z β
0
f (x)dx = F (b) − F (a) = H(β) − H(α) = H (t)dt = f (ϕ(t))ϕ0 (t)dt
a α α

Remark. Note that this substitution theorem is a liitle bit different from the lecture notes.
The assumption on ϕ is relaxed, but f is required to be continuous.

Prepared by Ernest Fan 5