0% found this document useful (0 votes)
65 views11 pages

Math 248 Homework 1: Edward Burkard

This document contains the solutions to 8 exercises related to Fourier transforms and Schwartz space functions. The exercises involve proving properties of Fourier transforms, computing Fourier transforms of specific functions, and showing that the Schwartz space is a vector space and certain quantities define seminorms on it.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
65 views11 pages

Math 248 Homework 1: Edward Burkard

This document contains the solutions to 8 exercises related to Fourier transforms and Schwartz space functions. The exercises involve proving properties of Fourier transforms, computing Fourier transforms of specific functions, and showing that the Schwartz space is a vector space and certain quantities define seminorms on it.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 11

Math 248 Homework 1

Edward Burkard

Exercise 1. Prove the following Fourier Transforms where a > 0 and c R:


a. fd
(x) = fb()
\
b. f (x
c) = e2ic fb()
2ixc
c. [e
f (x)]b = fb( c)
\
d. f (ax) = 1 fb()
a

0 (x) = 2i fb()
e. f[
1 d b
\
f. xf
(x) = 2i
d f ()
g. [(f g)(x)]b = fb()b
g ()

Proof.
a.
def
fd
(x) =

e2ix f (x) dx = fb()

b.
Z
\
f (x
c)

e2ix f (x c) dx

=
R

u=xc du=dx

e2i(u+c) f (u) du
Z
e2ic
e2iu f (u) du

2ic

fb()

c.
Z

[e2ixc f (x)]b =

e2ix e2ixc f (x) dx

e2i(c)x f (x) dx

=
R

= fb( c)
d.
Z
f\
(ax)

e2iax f (ax) dx
R
Z
1
e2iu f (u) dx
a R
1b
f ()
a

=
u=ax,

=a

du=dx

=
e. Suppose f C0 (R).
0 (x) =
f[

e2ix f 0 (x)dx

Let u = e2ix and dv = f 0 (x)dx, then du = 2ie2ix dx and v = f . So


Z
Z

e2ix f 0 (x) dx = e2ix f (x) + 2ie2ix f (x) dx
R
R

2ix

= e
f (x) + 2i fb()
=

2i fb()
1

f.
1 d b
f ()
2i d

=
=
=
=
=

Z
1 d
e2ix f (x) dx
2i d R
Z
1
d 2ix
e
f (x) dx
2i R d
Z
1
(2ix)e2ix f (x) dx
2i R
Z
1
(2i) e2ix [xf (x)] dx
2i
R
Z
2ix
e
[xf (x)] dx
R

\
= xf
(x)
g.
Z
[(f g)(x)]b

=
F ubini0 s T heorem

2ix

e
ZR Z

Z Z
(f g)(x) dx =

f (x y)g(y) dy dx
R

e2i(xy+y) f (x y)g(y) dx dy
ZR R
Z
e2iy g(y) e2i(xy) f (x y) dx dy
R
ZR
e2iy g(y)fb() dy
R
Z
fb() e2iy g(y) dy

fb()b
g ()

=
=


Exercise 2. Compute the Fourier Transforms of the following functions where a > 0:
2

a. f (x) = eax
1
b. f (x) = x2 +a
2
(Hint: Use contour integration and the Residue Theorem)
c. f (x) = ea|x|
Proof.
2

a. Since f 0 (x) = 2axeax , by Exercise 1.f above we have:


2
d b
id
i
2 b
0 )() = (2i)fb() =
f () = [(2axeax )]b() = (f
f ()
d
a
a
a
So we are left with the differential equation:

2
d b
f () = fb()
d
a
Solving for fb(), we see the general solution is:

fb() = ce a

Consider the function f1 () = e

2
a

f(). So we have

2 d
d
2 2 b
2 2 b
2 2 b
f1 () =
e a f () + e a
e a f ()
e a f () = 0
fb() =
d
a
d
a
a
so that f1 () is constant. To find the constant, let = 0 in order to obtain
Z
2
b
f1 (0) = f (0) =
eax .

To compute this integral, consider the following


(fb)2 () =

Z

eax

2

Z
=

ea(x

+y 2 )

R2

which follows from Fubinis theorem. By changing to polar coordinates this becomes
(fb)2 ()

Z
=

rear ddr

rear dr

2
0

Let u = ar2 , then du = 2ardr

=
=

1
a
1
a

eu du

hence,
f1 () = e

2
a

1
fb() = fb(0) =
a

or

e a
fb() =
a
b. Lets try to take the transform of f :
Z
fb() =
R

e2ix
dx
x2 + a2

To do this integral, use contour integration:


e2iz
dz
2
2
Cr z + a
where Cr (without loss of generality say r > a) is the semicircle in the upper complex plane (Im(z) 0).
Then we have that
!
Z
Z
e2iz
e2iz
b
f () = lim
dz
dz .
2
2
2
2
r
Cr z + a
Cupper z + a
Z

First consider the second integral. Since r > a we have


|z 2 + a2 | ||z|2 |a|2 | = ||a|2 |z|2 | |a2 r2 | = r2 a2
so

Z
Z

e2iz
|e2iz |

dz

dz


2
2
2
2
Cupper z + a

Cupper |z + a |
Z
1

dz
2 a2
r
Cupper
r
=
2(r2 a2 )
Since this tends to 0 as r tends to infinity, by the Residue theorem:
e2iz
1
dz =
lim Res(f ; ia)
2
2
r C z + a
2i r
r
To compute the residue, consider the function
Z

fb() = lim

f1 (z) =
then

e2iz
z + ia

1
e2a
fb() =
lim f1 (ia) =
2i r
4a

c.
fd
(x)

=
=
=
=

a|x|
e\
Z
e2ix ea|x| dx
R
Z
Z
e2ix eax dx +
e2ix eax dx
(,0]
[0,)
Z
Z
e(2ixax) dx +
e(2ix+ax) dx
(,0]

=
=
=

[0,)



ex(2ia)
e

+
2i + a
2i a 0

 

1
1
0 + 0
2i + a
2i a
1
1
+
a 2i
a + 2i
x(2ia) 0

Proof. Note that ex


Now consider

ex dx = 1. (Hint:
R
2
> 0 so R ex dx > 0.

Exercise 3. Prove that

Z
e

e(x

R
R2

x2

+y 2 )

2 Z
dx =

= )

e(x

+y 2 )

dxdy

R2

By switching to polar coordinates we have:

rer ddr

let u = r2 , then
1
2

eu ddu =

eu du

10=1

=
So

R
R

ex dx = 1.

Exercise 4. Let f L1 (R). Define F (x, ) := e2ix f (x). Then F is continuous for each x. Show that
Z
G() :=
F (x, )dx
R

is continuous.
Proof. Let R be an arbitrary point. Now choose any sequence {n } converging to , and define Fn (x) = F (x, n ).
Then since F is continuous we have:
Z
lim Fn (x) = lim
F (x, n ) dx = F (x, )
n

and |Fn | |F |. Now since F L , by the dominated convergence theorem we have:


Z
Z
lim G(n ) = lim
Fn (x) dx =
F (x, ) dx = G()
n

So G is continuous.
Exercise 5. If f (x), xf (x) L1 (R), then fb D(R) (differentiable functions on R) and


dfb
d

\ ().
= 2ixf

Proof.

 2ihx
Z
fb( + h) fb()
e
1
=
dx
e2ix
h
h
R
The integrand in the above equation is bounded by xf (x) which is an L1 function and tends to 2ixf (x)e2ix
pointwise. So by the Lebesgue Dominated Convergence Theorem
Z

e2ix

e2ihx 1
h


dx [2ixf ]b()

in the L1 norm. So letting h 0, we see that the first equation becomes:


d b
f ()
d

fb( + h) fb()
lim
h0
h

 2ihx
Z
e
1
=
dx
e2ix
h
R
= [2ixf ]b()
=

So clearly fb is differentiable.

Exercise 6. xn f (m) S (R) n, m 0.


Proof.
Lemma 1. f S (R) if and only if p, q 0 C such that |xp f (q) | C(1 + |x|)p
Let n, m 0 be arbitrary.
|xp (xn f (m) )(q) | =

since n n p + 1

|xp (nxn1 f (m) + xn f (m+q) )|

|nxn+p1 | + |xn f (m+q) |

c1 n(1 + |x|)(n+p1) + c2 (1 + |x|)n

c1 n(1 + |x|)np+1 + c2 (1 + |x|)n

C(1 + |x|)n

Therefore xn f (m) S (R) as desired.

Exercise 7. Find an example of an f L1x (R) such that fb


/ L1 (R).
Proof. Consider the function

f (x) = rect(x) :=

0
1

|x| >
|x|

1
2
1
2

Clearly this function is in L1x (R) as:


Z

Z
|rect(x)| dx =

rect(x) dx = 1
R

The Fourier transform of rect(x) is:


Z
R

e2ix rect(x) dx =

1
2
1
2

e2ix dx =

sin
:= sinc(x)

Claim: sinc
/ L1 (R).
proof of claim:
Z

Z
|sinc(x)| dx

|sinc(x)| dx

2
0

| sin(x)|
dx
x
0


X
sin 2n+1

Z
=

n=1

4X1
3 n=1 n

2
2n+1
2

2n+1
2
n=1

1
2n + 1
n=1

X
1
3n
n=1

which diverges. Thus sinc


/ L1 (R).
rect(x) is a function that satisfies the requirements.

Exercise 8.
a) Show that S (Rd ) is a R vector space.
b) Show that for every N, , , ||f ||(N,) = supxRd (1 + |x|)N | f (x)| is a seminorm on S (Rd ).
Proof.
a) Let f, g S (Rd ), let c R.
Clearly f + g S (Rd ) and cf S (Rd ) as limits break up over sums and we can factor constants out of
limits. So S (Rd ) is closed under addition and scalar multiplication.
Now check the 8 axioms of a vector space:
Let f, g, h S (Rd ) and r, s R.
(a) Clearly (f + g) + h = f + (g + h).
(b) Clearly f + g = g + f .
(c) Clearly 0 S (Rd ).
(d) Clearly f S (Rd ).
(e) Clearly r(f + g) = rf + rg.
(f) Clearly (r + s)f = rf + sf .
(g) Clearly r(sf ) = (rs)f .
(h) Clearly 1f = f .
Therefore it is (obviously) a vector space, as all of the above properties follow from the definition of a limit.
Give S (Rd ) the topology defined by the seminorms on it, i.e.
||f ||(N,) = sup (1 + |x|)N | f (x)|
xRd

The operations of addition and scalar multiplication are continuous with respect to this topology, so it remains to check that S (Rd ) is Hausdorff and complete to show that it is a Frechet space.
Lemma 2. Let X be a vector space with the topology defined by a family {p }A of seminorms. Then X
is Hausdorff iff for each x 6= 0 there is an A such that p (x) 6= 0.
Since S (Rd ) satisfies the above lemma, it is Hausdorff.
Now to show that S (Rd ) is complete let {fk } be a Cauchy sequence in S (Rd ), then ||fj fk ||(N,) 0 for
all N, . In particular, for each the sequence { fk } converges uniformly to a function, say g . Let ej be

the j th basis vector of Rd , observe:


t

Z
fk (x + tej ) fk (x) =

j fk (x + sej ) ds
0

Let k then the above gives:


Z

g0 (x + tej ) g0 (x) =

gej (x + sej ) ds
0

By the fundamental theorem of calculus we have that gej = j g0 , and using induction on || we get g = g0
for all . Then it follows that ||fk g0 ||(N,) 0 for all .
Hence S (Rd ) is a Frechet space.
(I showed this since showing it was a vector space seemed too easy...)
b) Let f, g S (Rd ) and let c R.
||f + g||(N,)

sup (1 + |x|)N | (f + g)|


xRd

sup (1 + |x|)N | f + g|
xRd

sup (1 + |x|)N (| f | + | g|)


xRd

sup (1 + |x|)N | (f )| + sup (1 + |x|)N | (g)|


xRd

xRd

= ||f ||(N,) + ||g||(N,)


||cf ||(N,)

sup (1 + |x|)N | (cf )|


xRd

sup (1 + |x|)N |c f |
xRd

sup (1 + |x|)N |c|| f |


xRd

= |c| sup (1 + |x|)N | f |


xRd

= |c| ||f ||(N,)



Exercise 9. Show that there is no constant 0 < Cd, < such that |x||| Cd, |xd |, x Rd .
Proof. Let x = ej (the j th standard basis vector of Rd ). Then |x| = 1 and so |x||| = 1 for all . However xd = 0
and hence, since any constant times 0 is 0 we have that there is no constant such that |x||| Cd, |xd |.

Exercise 10. Convolution is well-defined in L1 (Rd ) and
||f g||1 ||f ||1 ||g||1 .
1

Proof. Let f, g L (R ).


Z Z


f (x y)g(y)dy dx


Z Z
|f (x y)||g(y)|dydx
Z Z
|f (x y)||g(y)|dxdy
Z
Z
|g(y)| |f (x y)|dxdy
Z
|g(y)|||f ||1 dy
Z
||f ||1 |g(y)|dy

||f ||1 ||g||1 <

Z
||f g||1 =

|f g|dx

T onelli0 s T hm

=
=

Thus since the L1 norm of f g is finite we get that |f g| is finite a.e. and thus f g is finite a.e.
f g L1 (Rd ).


Exercise 11. Show that the Fourier Transform


F : S (Rd )
7

S (Rd )
fb

and the Inverse Fourier Transform


F 1 : S (Rd )
f

S (Rd )
f

are continuous (as linear) maps with respect to the natural topology of the Schwartz space.
Proof.
Lemma 3.
(1) If x f L1 for || k, then fb C k and fb = [(2ix) f ]b.
f ) = (2i) fb().
\
(2) If f C k , f L1 for || k, and f C0 for || k 1, then (
Lemma 4. If f C , then f S iff x f is bounded for all multi-indicies , iff (x f ) is bounded for all
multi-indicies , .
Let f S , then x L1 C0 for all , . So by Lemma 3, we have that fb is C and
f ) = (1)|| (2i)|| || ( fb)
(x\

Hence ( fb) is bounded for all ,, so fb S by Lemma 4. Moreover, since

(1 + |x|)n1 dx < we have:

f )|| ||x f || C||(1 + |x|)n+1 x f ||


||(x\
u
1
u

where ||f (x)||u = sup{|f (x)| | x X}. So we have that


||fb||(N,) CN,

||f ||(N +n+1,)

||||

by the proof of Lemma 4, and hence F is continuous.


Moreover, since f(x) = fb(x) we have that F 1 is continuous as well.


Exercise 12. Show f = f.


Proof. Take the Fourier transform of both sides then:

f = f
then sinceandare inverses of each other we get:

f = f



Exercise 13. Show f = f = f .

Proof.
f(x)

= f (x)
Z

f() =
e2ix f (x) dx
R
Z
Z

2ix

f (x) =
e
e2ix f (x) dx d
R

= f (x)
= f(x)

f(x)

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


Exercise 14. Calculate the Fourier Series of {x} := x [x] where [x] is the integer part of x.
Proof. Note first that {x} := x [x] can be represented by f (x) = x, x [0, 1) with f (x) = f (x + 1) and L = 21 .
Then
Z 21
a0 = 2
x dx = 0
12

and
Z
an

1
2

x cos(2nx) dx
12

21
Z 21

1
1
x sin(2nx)
sin(2nx) dx
1 2n
2n
1
2
2
21

1
00+
cos(2nx)
2
2
4n
1
2

and
Z
bn

1
2

x sin(2nx) dx
12

21
Z 12

1
1
x cos(2nx) +
cos(2nx) dx
1 2n
2n
1
2
2
21

1
1
cos(n) +
sin(2nx)
2
2n
(2n)
1
2

=
=

1
cos(n) + 0
2n
(1)n+1
2n

So
d=
{x}

X
(1)n+1
sin 2n
2n
n=1



Exercise 15. Calculate the Fourier Series of f (x) =

1
2 [n,n+1)
12 [n,n+1)

n even
n odd

10

Proof. Note that the period is 2 and hence L = 1.


Z 1
f (x) cos(nx) dx
an =
1

1
2

1
2

cos(nx) dx +

cos(nx) dx
0

0
1
1 sin(nx)
1 sin(nx)

+
2
n 1
2
n 0
0

=
=
This holds for all n 0.

Z
bn

f (x) sin(nx) dx
1

1
2

sin(nx) dx +
1

1
2

sin(nx) dx
0

0
1
1 cos(nx)
1 cos(nx)


2
n 1
2
n
0


1
1 cos(n)
1 cos(n)
1

2n 2 n
2 n
2n
1
(1)n

n
n
1 + (1)n+1
n

=
=
=
=
This holds for all n 1.
And so the transform of f is

fb() =

X
1 + (1)n+1
sin(n)
n
n=1



Exercise 16. Calculate the Fourier Series of f (x) =

x + 21
x + 21

12 x < 0
0 x 21

Proof. Note that the period is 1 and hence L = 12 .


Z

1
2

a0 = 2

21

Z
an

f (x) dx =

Z
(2x + 1) dx +

21

1
2

(2x + 1) dx =
0

1
2

1
2

f (x) cos(2nx) dx

2
12

=
12

Z
(2x + 1) cos(2nx) dx +

1
2

(2x + 1) cos(2nx) dx
0

0
21


1
+ 1 (cos(2nx) n sin(2nx) + 2nx sin(2nx))
(cos(2nx)
+
n
sin(2nx)
+
2nx
sin(2nx))


2
2
2
2
2n
2n
0
1
2

=
=
=

1
1
1
1
2 2 cos(n) + 2 2 2 2 cos(n)
2
2
2n
2n
2n
2n
1
1
2 2 cos(n)
n2 2
n
1 + (1)n+1
n2 2

11

Z
bn

1
2

f (x) sin(2nx) dx
12

=
12

1
2

(2x + 1) sin(2nx) dx

(2x + 1) sin(2nx) dx +

0
12


1
+ 1 (sin(2nx) + n cos(2nx) 2nx cos(2nx))
(n
cos(2nx)

sin(2nx)
+
2nx
cos(2nx))


2
2
2
2
2n
2n
1
0
2

1
1
1
1
=
sin(n)
+
2 2 sin(n)
2
2
2n
2n 2n 2n
= 0
Thus we have:

1 X 1 + (1)n+1
fb() = +
cos(2n)
2 n=1
n2 2

e2ia e2ib
Exercise 17. Show fb() =
is defined for = 0 (fb(0) = b a).
2i
Proof.
lim fb()

=
L0 H o
pital

=
=
as desired.

e2ia e2ib
0
2i
2iae2ia (2ib)e2ib
lim
0
2i
2ia + 2ib
2i
ba
lim

You might also like