Analysis 2 week 5

Quoc Trong Nguyen

March 13, 2025

Exercise 9
Since f is positive on [a, b], using the Cauchy–Schwarz inequality, we obtain the first inequality
 b  b  bp  b  bp !2
dx 2 1 1
f (x)dx = f (x) dx 2 dx ≥ f (x) p dx = (b − a)2 .
f (x)
p
a a a a f (x) a f (x)

To prove the second inequality, consider h : [a, b] → R defined by

h(x) = m(x − a) + M (b − x).

We see that h is continuous and

 b
h(a) = (b − a)M ≥ f (x)dx,
a
b
h(b) = (b − a)m ≤ f (x)dx.
a

Therefore, there is a c ∈ [a, b] such that

 b
f (x)dx = h(c) = m(c − a) + M (b − c).
a

Next, we claim that

 b
dx 1 1
≤ (c − a) + (b − c),
a f (x) m M
or in other words  b   c 
1 1 1 1
− ≤ − .
c f (x) M a m f (x)
To prove this, write
f (x) − m M − f (x)
f (x) = M+ m.
M −m M −m
1
Then by the convexity of g(x) = , we have
x
1 f (x) − m 1 M − f (x) 1
≤ +
f (x) M −m M M −m m

1
which leads to
1 1 f (x) − M 1 M − f (x) 1
− ≤ +
f (x) M M −m M M −m m
f (x) − m
=
Mm
f (x) − m
≤
f (x)m
1 1
= − .
m f (x)
Thus,
 b
dx 1 1
≤ (c − a) + (b − c),
a f (x) m M
as claimed. Now, we have
 b  b  
dx 1 1
f (x)dx ≤ [m(c − a) + M (b − c)] (c − a) + (b − c) .
a a f (x) m M
The right side of this inequality is a quadratic equation of variable c
M m 2 M m
− (2 −
)c − (a + b)(2 − − )c + •.
m M m M
M m a+b
Since 2 − − ≤ 0, this equation attains its maximum at c = . Hence,
m M 2
b b  
dx 1 1
f (x)dx ≤ [m(c − a) + M (b − c)] (c − a) + (b − c)
a a f (x) m M
   
b−a 1 1 b−a
≤ (m + M ) +
2 m M 2
(M + m)2
= (b − a)2 .
4M m

Exercise 10
Put
b 
1
f dx = F,
b−a a
b
1
gdx = G.
b−a a
Then we have
 b  b  b  b
1 1 1
f gdx − f dx gdx = f gdx − F G
b−a a (b − a)2 a a b−a a
b 
1
= f (g − G)
b−a a
b b
1 1
= f (g − G) + F (g − G)dx
b−a a b−a a
b
1
= (f − F ) (g − G) dx.
b−a a

2
Thus,
   
s
b b b  b 2
1 1 1
f gdx − f dx gdx = (f − F ) (g − G) dx
b−a a (b − a)2 a a b−a a
s  b  b
1 2
≤ (f − F ) dx (g − G)2 dx, (1)
(b − a)2 a a

by the Cauchy-Schwarz inequality. Next, consider

 b
1
(f − F )2 dx,
b−a a

and observe that  b

(f − F ) dx = 0.
a
Therefore,
 b  b"   2 # 
1 1 M 1 + m1 M 1 + m 1
(f − F )2 dx = (f − F )2 + 2 F − (f − F ) + F − dx
b−a a b−a a 2 2
 b
M1 + m2 2

1
− F− dx
b−a a 2
 b  b
M1 + m1 2 M1 + m2 2
 
1 1
= f− dx − F− dx
b−a a 2 b−a a 2
 b
M 1 + m1 2

1
≤ M1 − dx
b−a a 2
1 (M1 − m1 )2
= .
b−a 4
Similarly, we also have

1 b
1 (M2 − m2 )2
2
(g − G) dx ≤ .
b−a a b−a 4
Substituting them into (1) gives
   s
1 b
1 b b
1 (M1 − m2 )2 (M2 − m2 )2
f gdx − f dx gdx ≤
b−a a (b − a)2 a a (b − a)2 16
(M1 − m1 ) (M2 − m2 )
= .
4

Exercise 11
Using integration by parts formula, we have
 b  b
1
′
b 1 1
xf (x)f (x)dx = xf (x) − f (x)dx = − .
a 2 a 2 a 2
Applying the Cauchy-Schwarz inequality, we obtain
 b  b  b 2
′

2 2 2 1 ′
f (x) dx x f (x)dx ≥ xf (x)f (x)dx = .
a a a 4

3
Exercise 25
(a) The integral  ∞
x. sin ax
dx
0 k 2 + x2
is not an improper integral of type 2 as
x. sin ax
lim = 0.
x→0+ k 2 + x2
The integral
 b
sin ax dx
a
x
is bounded for every b > 0, and the function g defined by g(x) = has
k2 + x2
lim g(x) = 0.
x→+∞

Furthermore, g is monotonically decreasing for x > k. Thus, the Dirichlet’s test, with
f (x) = sin ax and g(x) defined as above, applies here, meaning that the integral
 ∞
x. sin ax
dx
0 k 2 + x2
is convergent.
(b) Write
 ∞  1  ∞
sin 2x sin 2x sin 2x
I= esin x dx = esin x dx + esin x dx.
0 xλ 0 xλ 1 xλ
Then I is convergent if and only if both of its parts are convergent. If λ = 1 then I is convergent
sin 2x
by Abel’s test with f (x) = and g(x) = esin x . Thus, we may assume that λ ̸= 1.
x
Step 1: We determine the set of values for which the integral
 1
sin 2x
I1 = esin x dx.
0 xλ
converges. Integration by parts gives
 1  1
sin 2x x1−λ sin x 1 x1−λ sin x
esin x λ dx = e sin 2x − 2 e cos x cos 2xdx.
0 x 1−λ 0 0 1−λ
Not that the notations we are using here are just formal. In case λ < 2, we have

x1−λ sin x 1 x1−λ sin x 1

e sin 2x = lim e sin 2x
1−λ 0 δ→0+ 1 − λ δ
1  
= lim esin 1 sin 2 − δ 1−λ esin δ sin 2δ
δ→0+ 1 − λ
 
1 sin 1 2−λ sin δ sin 2δ
= lim e sin 2 − δ e
δ→0+ 1 − λ δ
e sin 1 sin 2
= .
1−λ

4
We also have in this case that the integral
 1
x1−λ sin x
2 e cos x cos 2xdx
0 1−λ
2x1−λ
converges by Abel’s test with f (x) = and g(x) = esin x cos x cos 2x. Thus the integral I1 is
1−λ
convergent if λ < 2. Next, in case λ ≥ 2, we have
x1−λ sin x
 
1 1 sin 1 2−λ sin δ sin 2δ
e sin 2x = lim e sin 2 − δ e = +∞.
1−λ 0 δ→0+ 1 − λ δ
Additionally, in this case, we have  1
x1−λ dx = +∞.
0
x1−λ sin x
For x sufficiently near 0, C1 x1−λ ≤− e cos x cos 2x ≤ C2 x1−λ with some C1 , C2 such that
1−λ
0 < C1 ≤ C2 . Therefore, we have
 1
x1−λ sin x
− e 2 cos x cos 2xdx = +∞
0 1−λ
as well. Thus, we conclude that I1 diverges if λ ≥ 2 and converges if λ < 2.
Step 2: We determine the range of convergence for
∞
sin 2x
I2 = esin x λ dx.
1 x
This is easy. In case λ > 0, I2 converges by Dirichlet’s test with f (x) = esin x sin 2x and
1
g(x) = λ . In case λ ≤ 0, observe that
x
 nπ+δ
sin 2x
An = esin x λ dx ̸= 0
nπ x
for some δ > 0 being sufficiently small. Moreover, |An | is not bounded as n → +∞. Hence, I2
diverges.
In conclusion, the given integral I converges if 0 < λ < 2 and diverges otherwise.
(c) Let u(x) = x2 + x. Then u is a bijection on [0, +∞), and we have
∞  ∞ −λ
sin x2 + x


x
I= λ
dx = sin u(x)du(x).
0 x 0 2x + 1
x−λ
Case 1: λ > −1. Then is eventually monotone. It tends to 0 as u tends to +∞.
2x + 1
Furthermore, we see that  ∞
sin u(x)du(x)
0
is bounded. Hence, I is convergent by Dirichlet’s test.
x−λ
Case 2: λ ≤ −1. In this case, tends to either 1/2, if λ = 1, or +∞, if λ > −1, as u tends
2x + 1
to +∞. Therefore, for n ∈ N be sufficiently large, we have
 
2nπ+π
x−λ 1 2nπ+π 1
sin u(x)du(x) ≥ sin u(x)du(x) =
2nπ 2x + 1 4 2nπ 4
which shows that I does not converges.
In conclusion, I converges if λ > −1 and diverges otherwise.