Analysis 2 week 4

Quoc Trong Nguyen

March 5, 2025

Exercise 6
Let F be given by  x
F (x) = f (x)dx.
0

Then F ∈ C 2 ([0, 1]) as f ∈ C 1 ([0, 1]). Therefore, we can apply the Peano form of remainder to F
and obtain
 
i
 i+1       f ′  
n i+1 i i 1 n 1 1
f (x)dx = F −F =f + 2
+o .
i n n n n 2 n n2
n

Using this formula, we have

 1 X
n−1 i+1
n
f (x)dx = f (x)dx
i
0 i=0 n
n−1   n−1    
1 X i 1 X ′ i 1
= f + 2 f +o .
n n 2n n n2
i=0 i=0

Hence
n  1
!   n−1    
1X 1 1 1 X ′ i 1
n f( ) − f (x)dx =n (f (1) − f (0)) − f +o
n n 0 n 2n n n
i=1 i=0

and this gives us

n    1 ! n−1    !
1X 1 1 X ′ i 1
lim n f − f (x)dx = lim f (1) − f (0) − f +o
n→∞ n n 0 n→∞ 2n n n
i=1 i=0
f (1) − f (0)
= f (1) − f (0) −
2
f (1) − f (0)
= . (1)
2
(a) Applying this formula, we get
n  k  1 !
1k + 2k + · · · + nk
 
1 1X i k 1
lim n − = lim n − x dx = .
n→∞ nk+1 k+1 n→∞ n n 0 2
i=1

1
(b) First, note that
n    1 !
1X 1 f (1) − f (0)
lim n f − f (x)dx =
n→∞ n n 0 2
i=1
implies
n    1
1X 1
lim f = f (x)dx.
n→∞ n n 0
i=1
Fix some N ∈ N. Let
n  k n  k n  k
1X 2i − 1 1X 2i 1 1X 2i
An = ; Bn = − ; Cn = .
n n n n N n n
i=1 i=1 i=1
We claim that
lim Bn ≤ lim An ≤ lim Cn .
n→∞ n→∞ n→∞
The second inequality is obvious. To see the first one, let n be sufficiently large so that the sums
of the first N terms of An and Bn are both smaller than some given ε > 0. Then all the remaining
terms of An are larger than those of Bn . Hence lim An ≥ lim Bn − ε, and the inequality follows.
n→∞ n→∞
Next, by applying (1) to Cn and Bn , we get
2k
lim Cn =
n→∞ k+1
"  k+1 #
1 k+1

1 −1
lim Bn = 2− − .
n→∞ 2 (k + 1) N N
Finally, let N → ∞, we obtain
2k
lim Bn = lim An = lim Cn = .
n→∞ n→∞ n→∞ k+1

Exercise 7

Proposition 0.1. Let f be a differentiable function on [0, 1] whose derivative is Riemann

integrable. Then
n    1 !
1X 2i − 1
lim n f − f (x)dx = 0.
n→∞ n 2n 0
i=1

Proof. We have
1        n−1X   2i + 1   
1 2n − 1 2i − 1
f (x)dx = F − F (0) + F (1) − F + F −F
0 2n 2n 2n 2n
i=1
    
2n − 1 2n − 1
f f′
f (0) f ′ (0)
 
2n 2n 
= + 2
+ + 2

2n 4n  2n 4n 

    
2i − 1 ′ 2i − 1
X f
n−1
2n
f
2n 
 
1
+  + 2
 +o .
 n 2n  n2
i=1

2
Therefore
   
2n − 1 2n − 1
 n ! f f ′
1
f ′ (0)

1X 2i − 1 f (0) 2n 2n
n f (x)dx − f = + + +
0 n 2n 2 2 4n 4n
i=1
  n−1    
2n − 1 1 X ′ 2i − 1 1
−f + f +o
2n 2n 2n n
i=1

This gives us
 1 n  !
1X 2i − 1 f (0) f (1) f (1) − f (0)
lim n f (x)dx − f = − + = 0.
n→∞ 0 n 2n 2 2 2
i=1

Now, we have
 1        n−1
X   2i + 1   
1 2n − 1 2i − 1
f (x)dx = F − F (0) + F (1) − F + F −F
0 2n 2n 2n 2n
i=1
   
2n − 1 ′ 2n − 1
 
f f
f (0) f ′ (0)
 
 2n 2n 
= + +  + 
2n 4n2  2n 4n2 

    
2i − 1 ′ 2i − 1
X f
n−1
2n
f
2n 
 
1
+  + 2
 +o
 n 2n  n2
i=1

f (0) f ′ (0) f ′′ (0)

= + +
2n
 8n2 48n
3
  
2n − 1 ′ 2n − 1 ′′ 2n − 1‘
f f f
2n 2n 2n
+ + +
2n 8n2 48n3
n−1
X  1  2i − 1       
1 ′ 2i − 1 1 ′′ 2i − 1 1
+ f + 2f + 3f +o
n 2i 2n 2n 6n 2n n3
i=1

3
Therefore
 1 n  !   n−1  !
2 1X 2i − 1 n 2n − 1 1 X ′ 2i − 1
n f (x)dx − f = f (0) − f + f
0 n 2n 2 2n n 2n
i=1 i=1
   n  
1 ′ ′ 2n − 1 1 X ′′ 2i − 1
+ f (0) + f + f
8 2n 6n 2n
i=1
   
′′ 2n − 1 2n − 1

 f ′′ (0) f f ′′  
2n 2n +o 1

+ 48n + −
48n 6n  n

n  !
n 1 X ′ 2i − 1
= f (0) − f (1) + f
2 n 2n
i=1
   
2n − 1 ′ 2n − 1
f (1) − f f
2n 2n
+ −
1 2
4
2n
   n  
1 ′ ′ 2n − 1 1 X ′′ 2i − 1
+ f (0) + f + f
8 2n 6n 2n
i=1
   
′′ 2n − 1 2n − 1

 f ′′ (0) f f ′′  
2n 2n  1
+ + −  +o
48n 48n 6n  n

Hence, taking limit gives us

 n !
1
f ′ (1) − f ′ (0)
    
1X 2i − 1 1 1 1 1 ′ 1 1
lim n 2
f (x)dx − f = − + + f (1)+ − f ′ (0) = .
n→∞ 0 n 2n 4 2 8 6 8 6 24
i=1

Exercise 8
(a) We claim that
 1
xn
lim dx = 0.
n→∞ 0 1+x
xn 1
Given ε > 0, since 0 ≤ ≤ , there is a δ > 0 such that
1+x 1+x
 1  1
xn 1
0≤ dx ≤ dx < ε.
1−δ 1+x 1−δ 1+x

xn ε
Next, there exists an N such that < for every n > N and x ∈ [0, 1 − δ]. Thus, for
1+x 1−δ
n > N we have
 1  1  1−δ  1−δ
xn xn xn ε
0≤ dx = dx + dx < ε + dx = 2ε.
0 1+x 1−δ 1+x 0 1+x 0 1−δ

4
(b) We claim that
 1
lim f (xn )dx = f (0).
n→∞ 0

Let ε > 0 be given. Since f is continous on [0, 1], f is bounded on [0, 1]. Therefore, there exists a
δ > 0 such that  1
(f (xn ) − f (0)) dx < ε.
1−δ
Concerning the remain part of the integral, since 0 ≤ x < 1 and since f is continous, let n be
sufficiently large we get
ε
|f (xn ) − f (0)| <
1−δ
so that  1−δ
(f (xn ) − f (0)) dx < ε.
0
Now we have
 1  1  1
f (xn )dx − f (0) = f (xn )dx − f (0)dx
0 0 0
 1  1−δ
n
≤ (f (x ) − f (0)) dx + (f (xn ) − f (0)) dx
1−δ 0
< 2ε.

Exercise 13
(a)

Theorem 0.2. If f and g are Riemann integrable on [a, b]; g is nonnegative, non-increasing on
[a, b]. Then there exists ξ ∈ [a, b] such that
 b  ξ
f (x)g(x)dx = g(a) f (x)dx. (1)
a a

Proposition 0.3. Theorem 0.2 holds in case f is a constant function.

Proof. Let f (x) = C. Then (1) reduces to

 b
g(x)dx = g(a) (ξ − a) .
a

Define  b
h(x) = g(a) (x − a) − g(x)dx.
a
Certainly, h is continuous. Since g is nonnegative and non-increasing, we have h (a) ≤ 0 and
h (b) ≥ 0. Whence the existence of ξ satisfying (1) follows.

Proposition 0.4. If Theorem 0.2 holds for intervals [a, c] and [c, b], then it holds for [a, b] as well.

5
  t
Proof. Let F (t) = f (x)dx. By assumption and since g is non-increasing, there exist ξ1 and ξ2
a
such that
 c  ξ1
f (x) (g(x) − g(c)) dx = (g(a) − g(c)) f (x)dx
a a
 b  ξ2
f (x)g(x)dx = g(c) f (x)dx
c c

Adding together gives

 b
f (x)g(x) − g(c)F (c) = (g(a) − g(c))F (ξ1 ) + g(c) (F (ξ2 ) − F (c)) .
a

Divide both sides by g(a)

 b   
1 g(c) g(c)
f (x)g(x) − g(c)F (c) = 1 − F (ξ1 ) + F (ξ2 ) − g(c)F (c).
g(a) a g(a) g(a)

Since F is continuous on a connected set, its image is connected and therefore convex. Hence,
there is ξ3 ∈ [a, b] such that
 
g(c) g(c)
F (ξ3 ) = 1 − F (ξ1 ) + F (ξ2 ).
g(a) g(a)

Thus  b 
1
f (x)g(x) − g(c)F (c) = F (ξ3 ) − g(c)F (c)
g(a) a
leading to our desired result
 b
f (x)g(x) = g(a)F (ξ3 ).
c

Proposition 0.5. Theorem 0.2 holds in case f is a step function.

Proof. This follows from Proposition 0.3 and Proposition 0.4, noting that Proposition 0.3 holds in
case f fails to be constant at one point.

To prove Theorem 0.2 for general f , we will show that the difference
 b  t
f (x)g(x)dx − g(a) f (x)dx
a a

can be made arbitrarily small. It will follow by the continuity of F on [a, b] that a ξ as in the
Theorem exists. Given ε > 0, since f is Riemann integrable, there exists a partition
P = {a = x0 , x1 , x2 , . . . , xn = b} of [a, b] such that
n
X
U (P, f ) − L(P, f ) = (Mi − mi ) (xi − xi−1 ) < ε.
i=1

6
Let h(x) be given by (
h(x) = mi if x ∈ [xi−1 , xi )
h(b) = mn .
Then for arbitrary c ∈ [a, b] we have
 c  c  b  b  b
h(x)dx − f (x)dx ≤ h(x)dx − f (x)dx = U (P, f ) − f (x)dx < ε. (1)
a a a a a

Furthermore, since g is nonnegative, we have

 b  b n
X  xi
f (x)g(x) − h(x)g(x)dx = (f (ξi ) − mi ) g(x)dx
a a i=1 xi−1
n
X
= (f (ξi ) − mi ) (xi − xi−1 ) g(ζi )
i=1
n
X
≤ g(a) (Mi − mi ) (xi − xi−1 ) = g(a)ε. (2)
i=1

By Proposition 0.5, there is an ξ ∈ [a, b] satisfying

 a  ξ
h(x)g(x)dx = g(a) h(x)dx.
b a

Combine this with (1) and (2), we get

 b  ξ  b  b  ξ  ξ
f (x)g(x)dx − g(a) f (x)dx = f (x)g(x) − h(x)g(x)dx + g(a) h(x)dx − g(a) f (x)dx
a a a a a a
 b  b  ξ  ξ
≤ f (x)g(x) − h(x)g(x)dx + g(a) h(x)dx − f (x)dx
a a a a
< g(a)ε + g(a)ε
= 2g(a)ε.

This proves our claim, completing the proof of Theorem 0.2.

(b) The case g being constant holds trivially, so we assume g is strictly monotone. We may also
assume that g is monotonically decreasing since we can replace g by −g in case it is increasing.
Let
g(x) − g(b)
g̃ = .
g(a) − g(b)
Then g̃ is nonnegative and monotonically decreasing. Therefore, we can apply Theorem 0.2 to g̃
and obtain b ξ
f (x)g̃(x)dx = g̃(a) f (x)dx
a a
or  b  ξ
g(x) − g(b)
f (x) dx = f (x)dx.
a g(a) − g(b) a
Multiplying g(a) − g(b) into both sides gives
 b  b  ξ  ξ
f (x)g(x)dx − g(b) f (x)dx = g(a) f (x)dx − g(b) f (x)dx.
a a a a

7
which in turn leads to
 b  ξ  b
f (x)g(x)dx = g(a) f (x) + g(b) f (x)dx
a a ξ

as required.
(c) Define g̃(x) = g(b) − g(x). Then g̃ is nonnegative and non-increasing. We can therefore apply
Theorem 0.2 and obtain b b
f (x)g(x)dx = g(b) f (x)dx
a ξ

for some ξ ∈ [a, b].