Vous êtes sur la page 1sur 8

Math 131BH: Honors Analysis Problem Set 1 due Wednesday, January 15 1. Prove part (c) of Theorem 6.12 on p.

128 in Rudin. 2. Let f, : [a, b] R be two monotone increasing functions. Suppose that f R (). Show that R (f ) and that
b b

f d +
a a

df = f (b)(b) f (a)(a).

3. Rudin, Exercise 3 on p.138. 4. Rudin, Exercise 8 on p.138. 5. Rudin, parts (a), (b), (c) of Exercise 10 on p.139. 6. Rudin, Exercise 11 on p.140. 7. Rudin, Exercise 13 on p.140.

Homework 1 Solutions
that
1. (Theorem 6.12(c))

Let f R () on [a, b] and a < c < b. Show f R () on [a, c] and [c, b] and

a b

f d =
a

f d +
c

f d

Solution. Let > 0 and let P be a partition of [a, b] such that U (P, f, ) L(P, f, ) < . Let P = P {c}; that is, the partition formed from all the points of P and the point c. Then, U (P , f, )L(P , f, ) U (P, f, ) L(P, f, ) < as P is a renement of P . Write P = Q R where Q = {x P : x c} and R = {x P : x c}. Then, Q is a partition of [a, c], R is a partition of [c, b], and we have

U (P , f, ) L(P , f, ) = (U (Q, f, ) L(Q, f, )) + (U (R, f, ) L(R, f, ))

from which it follows that U (Q, f, ) L(Q, f, ) < and U (R, f, ) L(R, f, ) < as both of these quantities are nonnegative. Hence, f R () on [a, c] and [c, b]. To compute the integral, we note that

a b

f d = inf(U (P, f, ))

where P ranges over all partitions of [a, b]. This can be replaced by inf(U (P , f, )) where P ranges over all partitions of [a, b] containing c as to every P we can associate P = P {c} with U (P , f, ) U (P, f, ). Now, if P is such a partition, let Q, R be as above (so P = Q R) and note U (P , f, ) = U (Q, f, ) + U (R, f, ). Hence,
inf U (P , f, ) = inf (U (Q, f, ) + U (R, f, )) inf (U (Q, f, )) + inf (U (R, f, ))
P P P P

where we have identied P = Q R as above. To prove that


inf (U (Q, f, ) + U (R, f, )) inf (U (Q, f, )) + inf (U (R, f, ))
P P P

let > 0 and let Q1 , R2 be partitions of [a, c], [c, b] arising from P1 and P2 (ie Pj = Qj Rj for j = 1, 2) such that
U (Q1 , f, ) inf (U (Q, f, )) <
P

and
U (R2 , f, ) inf (U (R, f, )) <
P

Then, if P = Q1 R2 , we have
inf U (P , f, ) U (P , f, ) = U (Q1 , f, ) + U (R2 , f, ) < inf (U (Q, f, )) + inf (U (R, f, )) + 2
P P P

so
inf U (P , f, ) inf (U (Q, f, )) + inf (U (R, f, ))
P P P

as > 0 was arbitrary. Lastly,


inf (U (Q, f, )) = inf (U (Q, f, )) =
P Q

f d

and
inf (U (R, f, )) = inf (U (R, f, )) =
P R

f d
c

(which I will not prove here) from which the result follows.

2.

Let f, : [a, b] R be two increasing functions. Suppose f R (). Show that R (f ) and that

a b

f d +
a

df = f (b)(b) f (a)(a)

Solution.

If P is a partition of [a, b] given by a = x0 < x1 < < xn = b, then see


sup
x[xj 1 ,xj ]

f (x) = f (xj ) and

x[xj 1 ,xj ]

inf

f (x) = f (xj 1 )

and similarly
sup
x[xj 1 ,xj ]

(x) = (xj ) and

x[xj 1 ,xj ]

inf

(x) = (xj 1 )

Thus,
n

U (P, f, ) L(P, f, ) =
j =1

(f (xj ) f (xj 1 ))((xj ) (xj 1 )) = U (P, , f ) L(P, , f )

Let > 0. As f R (), we can nd P such that U (P, f, ) L(P, f, ) < . By the above, for this P we also have U (P, , f ) L(P, , f ) < , so R (f ). Now,

a b

f d +
a

df = inf U (P, f, ) + sup L(P, , f )


P P

We see, if P is notated as above, that


n n

U (P, f, ) + L(P, , f )

=
j =1 n

f (xj )((xj ) (xj 1 )) +


j =1

(xj 1 )(f (xj ) f (xj 1 ))

=
j =1

f (xj )(xj ) f (xj 1 )(xj 1 )

= f (b)(b) f (a)(a)

As we may take common renements in evaluating inf P U (P, f, ) + supP L(P, , f ) (ie if Qm is a minimizing sequence of partitions for the inf, Rm a maximizing sequence of partitions for the sup, then we may evaluate this expression by taking the limit as m with the partitions Pm = Qm Rm ), we see this expression is as claimed.
3. (Exercise 6.3) Dene functions j for j = 1, 2, 3 as follows: set j (x) = 0 if x < 0 and j (x) = 1 if 1 . Let f be a bounded function on [1, 1]. x > 0 for all j . Then, set 1 (0) = 0, 2 (0) = 1, and 3 (0) = 2 (a) Claim. f R (1 )

if and only if f (0+) = f (0) and, in that case,

1
1

f d1 = f (0).

0 = xp ). Then,

Proof.

Let P be a partition of [1, 1] given by 1 = x0 < x1 < < xn = 1 and suppose 0 P (say
U (P, f, 1 ) L(P, f, 1 ) = ( sup f (x)
x[xp ,xp+1 ] x[xp ,xp+1 ]

inf

f (x))

as 1 (xj ) 1 (xj 1 ) is 0 if j = p + 1 and 1 if j = p + 1. Suppose f R (1 ). Let > 0. Then, there is a partition P with 0 P (eg take a renement if necessary) and U (P, f, 1 ) L(P, f, 1 ) < . Put = xp+1 xp . Then, if 0 x < , we have
|f (x) f (0)| ( sup
x[xp ,xp+1 ]

f (x)

x[xp ,xp+1 ]

inf

f (x)) = U (P, f, 1 ) L(P, f, 1 ) <

so f (0+) = f (0). Now, suppose f (0+) = f (0). Let > 0 and let (0, 1) be such that 0 x < = |f (x) f (0)| < . , 1, we have Then, for P given by 1, 0, 2
U (P, f, 1 )L(P, f, 1 ) = ( sup f (x)
x[xp ,xp+1 ] x[xp ,xp+1 ]

inf

f (x)) = (

sup

f (x)f (0)+f (0)

x[xp ,xp+1 ]

x[xp ,xp+1 ]

inf

f (x)) 2

so f R (1 ). In this case we see


inf U (P, f, 1 ) = inf (
P P

sup
x[xp ,xp+1 ]

f (x)) = f (0+) = f (0)

(b) Claim. f R (2 )

if and only if f (0) = f (0) and, in that case,

1
1

f d2 = f (0).

Let P be a partition of [1, 1] given by 1 = x0 < x1 < < xn = 1 and suppose 0 P (say 0 = xp ). Then,
U (P, f, 2 ) L(P, f, 2 ) = ( sup
x[xp1 ,xp ]

f (x)

x[xp1 ,xp ]

inf

f (x))

as 1 (xj ) 1 (xj 1 ) is 0 if j = p and 1 if j = p. Suppose f R (2 ). Let > 0. Then, there is a partition P with 0 P (eg take a renement if necessary) and U (P, f, 2 ) L(P, f, 2 ) < . Put = xp xp1 . Then, if < x 0, we have
|f (x) f (0)| ( sup f (x)
x[xp1 ,xp ] x[xp1 ,xp ]

inf

f (x)) = U (P, f, 2 ) L(P, f, 2 ) <

so f (0) = f (0). Now, suppose f (0) = f (0). Let > 0 and let (0, 1) be such that < x 0 = |f (x) f (0)| < . Then, for P given by 1, 2 , 0, 1, we have
U (P, f, 2 )L(P, f, 2 ) = ( sup f (x)
x[xp1 ,xp ] x[xp1 ,xp ]

inf

f (x)) = (

sup

f (x)f (0)+f (0)

x[xp1 ,xp ]

x[xp1 ,xp ]

inf

f (x)) 2

so f R (2 ). In this case we see


inf U (P, f, 2 ) = inf (
P P

sup

f (x)) = f (0) = f (0)

x[xp1 ,xp ]

(c) Claim. f R (3 )

if and only if f is continuous at 0.

0 = xp ). Then,

Proof.

Let P be a partition of [1, 1] given by 1 = x0 < x1 < < xn = 1 and suppose 0 P (say
1 1 ( sup f (x) inf f (x)) + ( sup f ( x) inf f (x)) 2 x[xp1 ,xp ] 2 x[xp ,xp+1 ] x[xp1 ,xp ] x[xp ,xp+1 ]
1 2

U (P, f, 3 ) L(P, f, 3 ) =

as 1 (xj ) 1 (xj 1 ) is 0 if j = p, p + 1 and

if j = p, p + 1.

Suppose f R (3 ). Let > 0. Then, there is a partition P with 0 P (eg take a renement if necessary) and U (P, f, 3 ) L(P, f, 3 ) < . Put = min(xp xp1 , xp+1 xp ). Then, if |x| < , we have
|f (x)f (0)| ( sup f (x)
x[xp1 ,xp ] x[xp1 ,xp ]

inf

f (x))+(

sup

f (x)

x[xp ,xp+1 ]

x[xp ,xp+1 ]

inf

f (x)) = 2(U (P, f, 3 )L(P, f, 3 )) < 2

so limx0 f (x) = f (0). Now, suppose limx0 f (x) = f (0). Let > 0 and let (0, 1) be such that |x| < = |f (x) f (0)| < . Then, for P given by 1, 2 , 0, 2 , 1, we have
U (P, f, 3 )L(P, f, 3 ) = 1 1 1 1 ( sup f (x) inf f (x))+ ( sup f (x) inf f (x)) 2 + 2 = 2 2 x[xp1 ,xp ] 2 x[xp ,xp+1 ] 2 2 x[xp1 ,xp ] x[xp ,xp+1 ]

so f R (3 ). In this case we see


1 1 inf U (P, f, 3 ) = inf ( ( sup f (x)) + ( sup f (x)) = f (0) P P 2 x[xp1 ,xp ] 2 x[xp ,xp+1 ]

(d)

This was proven above. Suppose f : [1, ) [0, ) is monotonically decreasing.


n=1

4. (Exercise 6.8) Claim. Proof.

f (x)dx converges if and only if

f (n) converges.

We note by denition that 1 f (x)dx converges if and only if limA 1A f (x)dx converges. A Now, 1 f (x)dx is an increasing function of A as f 0. Thus, 1 f (x)dx converges if and only if 1A f (x)dx is bounded above uniformly in A.
A f (n) converges. Let A > 1 and suppose m A < m +1 where m Z. Then, 1 f (x)dx m+1 f (x)dx U (P, f, x) for any partition P of [1, m + 1]. Consider P given by xk = k + 1 for k = 0, . . . , m. 1 Then, as f is decreasing,

Suppose

n=1

U (P, f, x) =
k=1

f (xk1 )(xk xk1 ) =


k=1

f ( x k 1 ) =
k=1

f (k )
k=1

f (k )

Thus,

k=1

f (k ) is an upper bound for

A
1

f (x)dx independent of A, so

f (x)dx converges.

m Now, suppose n=1 f (n) diverges. Let M R. Chose m Z such that n=2 f (n) > M . Let P be the partition of [1, m + 1] given by xk = k + 1 for k = 0, . . . , m. Then, as f is decreasing,

m+1

f (x)dx L(P, f, x) =
k=1

f (xk )(xk xk1 ) =


k=2

f (k ) > M

Thus, as M was arbitrary,

A
1

f (x)dx is not bounded above independent of A and so does not converge.


1 p

5. (Exercise 6.10 a-c) (a) Claim. Proof.

Let p, q > 0 be such that


up p

1 q

= 1.

If u 0, v 0, then uv

vq q

. Moveover equality occurs if and only if up = v q .


p q

Fix v 0 and consider the function f (u) = up + vq uv dened on [0, ). For u = 0, we see vq 1 this is q 0. Dierentiating with respect to u, we have f (u) = up1 v . From p +1 q = 1, we see p > 1, so this is increasing in u. Thus, f is decreasing in u until the critical point f (u0 ) = 0 and then increasing for 1 u u0 , so to show the result we evaluate f at u0 where f (u0 ) = 0. This is up = v , which has a unique 0 1 solution for u0 0, namely u0 = v p1 . We have
f (u0 ) =
1 up vq v p1 vq 0 + u0 v = + v p1 v p q p q p

Solving

1 p

1 q

= 1 for q gives q =

p p1

, so
1+p1 vq vq + v p1 = v q v q = 0 p q 1

f (u0 ) =

Thus, f has a unique minimum (as f is strictly increasing) at u = v p1 , which is the same as up = v q , and at this point it attains the value 0.
(b)

Suppose f, g R (), f, g 0, and


b
a

b
a

f p d = 1 =

b
a

g q d.

Claim. Proof.

f gd 1
f (x)p p ) + g (x q . Integrating over
q

For x [a, b], we have 0 f (x), g (x), so by (a) we see f (x)g (x) [a, b] this becomes

f gd
a a

fp gq 1 1 + d = + = 1 p q p q

as desired.
(c)

Suppose f, g R () are complex functions. We have


|
a

Claim.

f gd| (
a

|f | d)

1/p

(
a

|g |q d)1/q b = ( a |g |q d)1/q . First, suppose f |f g |d =


a

b We use the notation f p = ( a |f |p d)1/p and g f| |g | 0. Consider F = | f p and G = g q . Then,

Proof.

, g

>

1 f
p

g |
q a

f gd|

1 f
p

g
q a

F Gd

We see

b p |f | d |f |p a =1 F d = p d = b f p a a |f |p d a b q b and likewise a G d = 1. Moreover, F, G 0, so by (b) we have a F Gd 1. With the above, this gives
b b p

1 f
p

g |
q a

f gd| 1

so |

b
a

f gd| f

as claimed.

b f gd| = 0. Indeed, g is bounded, so |g | is bounded, say 0 Now, suppose f p = 0. We show | a b b b |g | M . Then, | a f gd| | a |f |M d, so it suces to show a |f |d = 0. Now, if f p = 0, then b p f p = a |f |p d = 0, so supP L(P, |f |p , ) = 0. Hence, for any xed P we must have L(P, |f |p , ) = 0 (as |f |p 0), so this shows that for every [xj 1 , xj ] given by P , we either have

inf

[xj 1 ,xj ]

(|f (x)|p ) = 0 or (xj ) (xj 1 ) = 0

If P is a partition of [a, b] given by x0 < < xn , we see


( inf |f (x)|)p = inf
[xj 1 ,xj ]

(|f (x)|p )

[xj 1 ,xj ]

Hence, on [xj 1 , xj ] we have


inf
[xj 1 ,xj ]

|f (x)| = 0 or (xj ) (xj 1 ) = 0

so we have L(P, |f |, ) = 0 for all P . Thus,

a b

|f |d = sup L(P, |f |, )P = 0
p

Likewise, if g

= 0, then |

b
a

f gd| = 0, so the claim holds.


2 2

6. (Exercise 6.11)

For u R (), let u


f h

b 1 = ( a |u|2 d) 2 . Suppose f, g, h R (). Show


2

f g

+ gh

Let F = f h and G = g h, so we wish to show F + G 2 F 2 to show F + G 2 2 ( F 2 + G 2 ) . From the left hand side, we have
Proof.

+ G 2 . For this, it suces

|F + G| d
a 2

|F | + 2|F ||G| + |G| d = F


2 2 2 2

+2
a

|F ||G|d + G

2 2

Now, from exercise 6.10, we have 2


F +G
2 2

b
a

|F ||G|d 2 F
2 2

2 2

G 2 , so we have + G
2 2

+2 F

=( F

+ G 2 )2

as claimed.
7. (Exercise 6.13) (a) Claim. |f (x)| < Proof.

Dene f (x) =
1 x

x+1
x

sin(t2 )dt

for x > 0
sin(t )dt =
x x2 2

Let u = t2 , so du = 2tdt and


x+1 (x+1)2

sin(u)

du 2u 2
1

Integrating by parts, this is


cos(u) 2u
1 2

|u=x2

(x+1)2

(x+1)2

cos(u)
x2

du 4u
3 2

cos(x2 ) cos((x + 1)2 ) 2x 2(x + 1)

(x+1)2

cos(u)
x2

du 4u 2
3

Hence,
1 1 |f (x)| < + + 2x 2(x + 1)
(x+1)2

du 4u
3 2

x2

1 1 1 (x+1)2 1 1 1 1 1 + +( 1 )|u=x2 = + + = 2x 2(x + 1) 2 x 2( x + 1) 2 x 2( x + 1) x 2 2u


1 4u 2
3

Note the inequality is strict here as both cos(u) 1 3 and 4u 2 are not the same (see exercise 6.2).
(b) Claim. 2xf (x) = cos(x2 ) cos((x + 1)2 ) + r (x) Proof.

are continuous on [x, x + 1] and these functions


c x

where |r(x)| <

for some constant c.

Put r(x) = 2xf (x) cos(x2 ) + cos((x + 1)2 ). By the above,

(x+1)2

r ( x) =

2x cos((x + 1)2 ) + cos((x + 1)2 ) 2x 2(x + 1)

cos(u)
x2

du 4u 2
3

cos((x + 1)2 ) 2x (x + 1)

(x+1)2

cos(u)
x2

du 4u 2
3

and so
|r(x)| 1 + 2x x+1
(x+1)2

du 4u
3 2

x2

1 1 1 1 x+1x 2 + 2 x( )= + x( )= x+1 2x 2(x + 1) x+1 x(x + 1) x+1

so c = 2 works.
(c) Claim. Proof.

The upper and lower limits of xf (x) are 1 and 1 respectively.

1 We know x|f (x)| < 1 by (a). From (b), we have xf (x) = 2 (cos(x2 ) cos((x + 1)2 ) + r(x)) where 2 2 r(x) 0 as x , so it suces to show the upper limit of cos(x ) cos((x + 1) lower limit ) is 2 and the 2 2 of cos(x ) cos((x + 1) ) is 2. Consider the sequence formed by taking xn = 2n for n Z+ . We have

2 cos(x2 n ) cos((xn + 1) ) = cos(2n) cos(2n + 2 2n + 1) = 1 cos(2 2n + 1)

where we have used cos(A + B ) = cos(A) cos(B ) sin(A) sin(B ). Hence, we need only show cos(2 2n + 1) has lower limit 1. We know that this has lower limit 1. Let > 0. Let > 0 be such that |x | < = | cos(x) + 1| < . Then, if |x (2m + 1) | < , we have | cos(x) + 1| < by periodicity. Let M > 0 and nd N > M such that 2 2 (n + 1) 2 2n < for n N . Then, as 2 2n as n , we see there is an x M with x = 2 2n for some n N and |x (2m + 1) | < for some m Z. Thus, | cos(2 2n) + 1| < , from which it follows that the lower limit of cos(2 2n + 1) is 1. The proof that the lower limit of xf (x) = 1 is similar with (2n + 1) in place of 2n.
(d) Claim. Proof.

sin(t2 )dt converges.


1 A

It suces to show

sin(t2 )dt converges. As above, we have cos(1) cos(A2 ) 2 2A


1 A2

sin(t2 )dt =
1

cos(u) u2
3

du

and

A2
1

cos(u)
3 u2

du is absolutely convergent,

cos(A2 ) 2A

0 as A , so this converges.

Vous aimerez peut-être aussi