Sie sind auf Seite 1von 7

Math 426: Homework 5 Solutions

Mary Radcliffe
due 14 May 2014

In Bartle:

sh is
ar stu
ed d
vi y re
aC s
o
ou urc
rs e
eH w
er as
o.
co
m

6H. Let X = Z+ , and let be the measure on X which has measure n12 at
the
point n. Show that (X) < . Let f be defined on X by f (n) = n.
Show that f Lp if and only if 1 p < 2.
Solution. We have (X) =
Z

n=1

|f |p d

(n) =

n=1

n=1

2
6 ,

which is finite. Moreover,

|f (n)|p (n)
np/2 n2
np/22 ,

n=1

which converges if and only if p/2 2 < 1, if and only if p < 2. But as
p is assumed to be at least 1, since otherwise we do not have a norm, we
have that the integral is finite if and only if 1 p < 2.

6I. Modify the previous exercise to obtain a function on a finite measure space
which belongs to Lp if and only if 1 p < p0 .

Th

Solution. Let X = Z+ , and let be the measure


with (n) = np0 .
P onpX
0
Note that as p0 > 1, we have that (X) = n=1 n
< .
Take f (n) = n11/p0 . Then we have
Z

X
|f |p d =
|f (n)|p (n)
n=1

n=1

npp/p0 np0

np((p0 1)/p0 )p0 ,

n=1

which converges if and only if p((p0 1)/p0 ) p0 < 1, if and only if


p < (p0 1)(p0 /(p0 1)) = p0 , as desired.

6N. Let (X, F, ) be a measure space, and let f belong to both Lp1 and Lp2 ,
with 1 p1 < p2 < . Prove that f Lp for any value of p such that
p1 p p2 .
https://www.coursehero.com/file/9770580/Homework-5-Spring-2014-Solutions/

Solution. Define E1 = {x X | |f (x)| < 1} and E2 = {{x X | |f (x)|


1}. Note that if x E1 , then |f (x)|p |f (x)|p1 , and if x E2 , then
|f (x)|p |f (x)|p2 . Therefore we have
Z
Z
Z
|f |p d =
|f |p d +
|f |p d
E1
E2
Z
Z
|f |p2 d

|f |p1 d +
E2
E1
Z
Z

|f |p1 d + |f |p2 d <


by hypothesis. Therefore, f Lp .

sh is
ar stu
ed d
vi y re
aC s
o
ou urc
rs e
eH w
er as
o.
co
m

6P. Let f Lp (X, F, ), 1 p < , and let  > 0. Show that there exists
a set E F with (E ) < such that if F F and F E = , then
kf F kp < .
Solution. For n Z+ , define
Gn = R{x X | |f (x)|p > n1 }. Then |f (x)|p
R
1
p
|f | d n1 Gn d = n1 (Gn ), so (Gn ) <
n Gn , and thus >
for all n.

Now, |f |p Gn is an increasing sequence


of functions
limit |f |p , and
R
R with
p
p
+
thus by the
MCT,
we
have
that
|f
|

|f
|
d.
Let
Gn

R
R
R
R N Z
|f |p G d |f |p d < p . Then as |f |p d = |f |p G d+
such
that
N
N
R
R
|f |p X\GN d, we have that |f |p X\GN d < p . Let GN = E .
Now, let F F with F E = , so F X\GN . Then
Z
Z
kf F kpp = |f |p F d |f |p X\GN d < p ,
and the desired result follows.

Th

6Q. Let fn Lp (X, F, ), 1 p < , and let n be defined for E F by


1/p
R
n (E) = E |fn |p d
. Show that |n (E) m (E)| kfn fm kp .
Hence, if {fn } is a Cauchy sequence in Lp , then lim n (E) exists for each
E F.

Solution. Note that n (E) = kfn E kp . Moreover, by the triangle inequality, we have that for any two functions f and g, kf kp kf gkp +kgkp , and
thus kf kp kgkp kf gkp . Wolog, assume that kfn E kp > kfm E kp .
Then
|n (E) m (E)| = |kfn E kp kfm E kp |
= kfn E kp kfm E kp

k(fn fm )E kp

kfn fm kp ,

as desired.
6R. Let fn , n be as in Exercise 6Q. If {fn } is a Cauchy sequence and  > 0,
then there exists a set E F with (E ) < such that if F F and
F E = , then n (F ) <  for all n Z+ .

https://www.coursehero.com/file/9770580/Homework-5-Spring-2014-Solutions/

Solution. By problem 6P, such a set exists for each fn ; for all  > 0, let
(n)
(n)
E be such that if F E = , then kfn E (n) kp < .


Now, fix  > 0. Let N be sufficiently large that if n, m N , then kfn


N
[
(k)
E/2 .
fm kp < 2 . Put E =
k=1
c
Let F F with F E = . If n N , then F E/2
, so kfn F kp =
n (F ) < /2. If n > N , then |n (F ) N (F )| kfn fN kp < 2 , and
thus n (F ) < 2 + N (F ) < .

6T. If f L (X, F, ), then |f (x)| kf k for almost all x. Moreover,


if A < kf k , then there exists a set E F with (E) > 0 such that
|f (x)| > A for all x E.

sh is
ar stu
ed d
vi y re
aC s
o
ou urc
rs e
eH w
er as
o.
co
m

Solution. For all  > 0, there exists a set N such that (N ) = 0 and
S(N ) < kf k + . Take N =
k=1 N1/k . Then (N ) = 0 by countable
additivity, and if x
/ N , then x
/ N1/k for all k. Thus, |f (x)| S(N1/k )
for all k, and thus |f (x)| < kf k + 1/k for all k, so |f (x)| kf k for all
x
/ N . Therefore, |f (x)| kf k for almost all x.

Suppose there exists some A < kf k such that there is no E F with


(E) > 0 and |f (x)| > A for all x E. Let N = {x X | |f (x)| >
A}. Then by hypothesis, (N ) = 0, and thus kf k S(N ) A, a
contradiction. Thus, the result holds.

Additional Exercises:

1. Let L(Rn , Rm ) denote the space of linear transformations from Rn to Rm .


Recall that in 425, we defined the norm of a transformation as follows:
kAk = sup kAxk2 (see Baby Rudin, page 208). We proved that this
kxk2 =1

satisfies the properties of a norm in Theorem 9.7 (again, page 208.) This
is usually called the Euclidean norm, the `2 norm, or the Frobenius norm
for matrices, and is more properly denoted by kAk2 .
(a) Let V be a vector space with norm k k, and let L(V ) be the space
of linear transformations from V to itself. Define N : L(V ) R by
N (A) = sup kAvk. Show that N defines a norm on V . (Usually

Th

kvk=1

we denote this as N (A) = |||A|||). Thus we can define a p-norm for


operators as well.

(b) Show that if |||A||| is defined from a vector norm on V as in part (a),
then we have the inequality |||AB||| |||A||||||B||| for all A, B L(V ).
This property is usually called submultiplicativity.

(c) Let A be an n n real matrix. Define |||A|||p (sometimes we write


just kAkp ) to be the matrix norm derived from the Lp norm on Rn
as in part (a). This is usually called the p-norm on linear operators.
Determine an expression for |||A|||1 and |||A||| in terms of only A.

Solution. (a) Clearly, N (A) 0 for all A L(V ), and N (0) = 0. Now,
suppose that A 6= 0. Then there exists some x V such that Ax 6= 0,
so sup kAvk kAxk/kxk > 0. Thus, N (A) = 0 if and only if A = 0.
kvk=1

Next, if R, we have N (A) = sup kAvk = sup ||kAvk =


kvk=1

||N (A), since k k is a norm on V .


https://www.coursehero.com/file/9770580/Homework-5-Spring-2014-Solutions/

kvk=1

Finally, let A, B L(V ). Then we have


N (A + B)

sup k(A + B)vk


kvk=1

sup (kAvk + kBvk)


kvk=1

sup kAvk + sup kBvk = N (A) + N (B).


kvk=1

kvk=1

Therefore, N is a norm on L(V ).


(b) Consider |||AB||| = sup k(AB)vk. Let y = Bv. Then we have
kvk=1

|||AB||| =

sup k(AB)vk
kvk=1

sup kAyk
kvk=1

kyk
kyk

kAyk
kBvk
kyk

sh is
ar stu
ed d
vi y re
aC s
o
ou urc
rs e
eH w
er as
o.
co
m
=

sup

kvk=1

sup

yV

kAyk
sup kBvk
kyk kvk=1

sup kAyk sup kBvk = |||A||||||B|||.

kyk=1

kvk=1

n
(c) We first
P consider |||A|||1 . Note that if v R has kvk1 = 1, then we
have
|vi | = 1. Thus, if A1 , A2 , . . .P
, An are the columns of A, we
have kAvk = kv1 A1 + + vn An k |vi |kAi k1 max1in kAi k1 .
Moreover, if i attains the maximum, taking v = ei yields that kAei k =
max1in kAi k1 . Therefore, the 1-norm of A is the maximal 1-norm
of the columns of A.
On the other hand, if kvk = 1, then max |vi | = 1. Moreover, if
T
a1 , a2 , . . . , an are the rows of A, then Av
Pn= [a1 v, a2 v, . . . , an v] ,
so kAvk = max1in |ai v|. If S = j=1 |aij | is the maximal row
sum, then, it is clear that kAvk S. Moreover, taking v to be a
vector of all 1 such that aij vj > 0 for all j, we have kAvk = S.
Therefore, |||A||| is the maximal 1-norm of the rows of A.

Th

2. Let (X, F, ) be a measure space.

(a) Prove that the simple functions with finite support are dense in
Lp (X) if 1 p < . That is, for all f Lp and  > 0, there
exists a simple function with finite support and kf kp < .

(b) Prove that the simple functions with finite support are dense in L
if X is of finite measure. Give an example of an L function that
cannot be approximated by simple functions with finite support if
(X) = .

Solution. (a) Let f Lp ,  > 0. Let En = {x X | |f (x)| 1/n}. Note


that Ras |f |p En Rincreases to |f |p , by the MCT there exists n such
p
that |f |p d |f |p En d < 2 . Let g = |f |En , so g has finite
support and g Lp .
By Theorem 2.11, there exists a simple function such that |g | <

for all x, and = 0 whenever g = 0, so has finite
2((En ))1/p
p
p
R
R 
= 2 .
support. Then |g |p d En 2((En ))1/p
https://www.coursehero.com/file/9770580/Homework-5-Spring-2014-Solutions/

Therefore, we have
Z
kf kp

1/p

|f | d

=
Z

|f | d +

|f | d
c
En

En

Z

!1/p

|g | d +
En

  p 1/p
2

   p 1/p

2
2

= 21/p < ,
2
as desired.

sh is
ar stu
ed d
vi y re
aC s
o
ou urc
rs e
eH w
er as
o.
co
m

(b) The first part is immediate by applying Theorem 2.11 to f . For the
second part, if f (x) = 1, where f L (R), it is clear that if g is
any finitely supported function, then kf gk 1, so f cannot be
approximated by finitely supported simple functions.

3. Consider the real space under Lebesgue measure, (R, B, ). Recall that
C(R) is the family of continuous functions on R. Define Cc (R) to be the
family of continuous, compactly supported functions on R.

(a) Show that for 1 p < , Cc (R) is dense in Lp (R). (Hint: Since
we know the simple functions are dense by problem 2, it suffices
to approximate simple functions by compactly supported continuous
functions.)

Th

(b) Does this result hold in L ? Explain.

Solution. (a) First, we claim that if (E) < , then for all  > 0 there
exists a continuous function f such that kf E kp < .
Let En = E [n, n]. As En En+1 and En = E, we have that
(E
 n ) (E). Choose N sufficiently large that (En ) (E)
 p
4 .
p
Let U be an open set such that En U and (U \En ) 4 . By
possibly intersecting with the open set (n , n + ), we can ensure
that U is contained in a compact set. Note, moreover, that as U
is open U can be written as an (at most) countable union of open
intervals. Write U = Ik , where Ik = (ak , bk ).
Define fk as follows. Fix some k <

1
fk (x) =
1+

1
k (x
1
k (x

p (p+1)
.
2kp+2

Then put

ak x bk
bk ) bk < x bk + k
.
ak ) ak k x < ak
else

Then we have fk is supported on (ak k , bk + k ), fk is continuous,

https://www.coursehero.com/file/9770580/Homework-5-Spring-2014-Solutions/

and
kIk

fk kpp

|Ik fk |p d
p
p
Z bk
Z ak
Z bk +k




1
1
p




0 dx +
=
1 + k (x ak ) dx +
1 k (x bk ) dx
ak
ak k
bk
Z 1
Z 0
=
k up du +
k up du
=

k
= 2
p+1
< 2
Let f =

p (p+1)
2kp+2

p+1

  p
2k+1

fk . Then f is continuous, and we have


X
(f Ik )kp

sh is
ar stu
ed d
vi y re
aC s
o
ou urc
rs e
eH w
er as
o.
co
m

kf U kp

kf Ik kp

<

X
k


= /2
2k+1

Therefore, we have
kE f kp

kE En kp + kEn U kp + kU f kp

(E\En )1/p + (U \En )1/p +
2



+ + < .
4 4 2

Now, let be an arbitrary


finitely supported simple function. Then
Pn
we may write = k=1 ak Ek , where (Ek ) < and ak 6= 0 for all
k. Then put fk toP
be a continuous function such that kEk fk kp

.
Taking
f
=
ak fk yields a continuous function satisfying the
|ak |
desired property by the triangle inequality.

Th

(b) No. We have the same problem with compact (aka finite) support
here as in 2b.

Extras:

6C. Let N be a norm on a linear space V and let d be defined for u, v V by


d(u, v) = N (u v). Show that d is a metric on V .

Solution. Clearly d(u, v) 0 for all u, v, and d(u, v) = 0 iff u = v. Also,


d(v, u) = N (vu) = |1|N (uv) = N (uv) for all u, v, and for all u, v, w,
we have d(u, w) = N (u w) = N (u v + v w) N (u v) + N (v w)
by the triangle inequality. Thus, d is a metric.

6D. Whoops, this is a part of additional exercise 2.


6E. If f Lp , 1 p < , and if E = {x X | |f (x)| =
6 0}, then E is -finite.
https://www.coursehero.com/file/9770580/Homework-5-Spring-2014-Solutions/

Solution. Define En = {x XR | |f (x)|p Rn1 }. Notice then that En = E,


and as f Lp , we have > |f |p d En |f |p d n1 (En ), and thus
(En ) < for all n. Therefore, E is -finite.
6F. If f Lp and if En = {x X | |f (x)| n}, then (En ) 0 as n .
Solution. Note that En En+1 , so (En ) (En+1 ). By the same logic
as problem 6E, it is clear that (En ) is finite, and thus {(En )} is a
bounded monotonic sequence in R, and it thus has a limit. Suppose
the
R
p
limit
is
not
0,
so
that
(E
)

C
>
0.
Then
for
all
n,
we
have
|f
|
d

n
R
p
p
|f
|
d

Cn
,
and
thus
the
integral
is
infinite,
a
contradiction.
En
6K. If (X, F, ) is a finite measure space and f Lp , then f Lr for 1 r p.
Apply H
olders Inequality to |f |r in Lp/r and g = 1 to obtain the inequality
kf kr kf kp (X)s , where s = 1/r 1/p.

sh is
ar stu
ed d
vi y re
aC s
o
ou urc
rs e
eH w
er as
o.
co
m

Solution. Let E1 = {x X | |f (x)| < 1}. Then if x E1c , |f (x)|r


|f (x)|p . Therefore,
Z
Z
Z
|f (x)|r d
|f (x)|r d +
|f (x)|r
E1c

E1

1 d +

E1

<

(E1 ) +

|f (x)|p d

E1c
p
kf kp <

Thus, f Lr .

Now, since |f |r Lp/r and 1 Lp/(pr) , which are conjugate indices,


H
olders Inequality yields
k|f |r k1 k|f |r kp/r k1kp/(pr) .

Note that for any x, y, we have k|f |x ky = kf kxxy , and thus we may rewrite
the above inequality as
kf krr kf krp (X)(pr)/p .

Taking 1/r power on both sides yields the given inequality.

Th

6L. Suppose that X = Z+ and is the counting measure on Z+ . If f Lp ,


then f Ls for 1 p s < , and kf ks kf kp .

P
Solution. If f Lp , then we have
|f (n)|p < . Then by comparison
s
p
test, for n sufficiently large
P |f (n)|s< 1,Pand we phave |f (n)| < |f (n)| , and
s
thus f L and kf ks = |f (n)| |f (n)| = kf kp .

6U. If f Lp , 1 p , and g L , then the product f g Lp and


kf gkp kf kp kgk .
Solution. Let M = kgk . If there exists N F with (N ) = 0 and
g(x) > M on N , replace g with a function h that takes value 0 on N
with
Rand agrees
R g elsewhere,
R so g = h a.e. and h is bounded. Then
|f g|p d = |f h|p d M p |f |p d. The result follows.

https://www.coursehero.com/file/9770580/Homework-5-Spring-2014-Solutions/

Powered by TCPDF (www.tcpdf.org)

Das könnte Ihnen auch gefallen