Lecture 11

Weston Barger

July 28, 2016

1 Wave equation on infinite domain

Consider the wave equation on the infinite domain



 x ∈ (–∞, ∞), t ∈ (0, ∞),

utt = c 2u xx , u(x , 0) = f (x ),


ut (x , 0) = 0.



We consider the case that ut (x , 0) = g(x ) = 0 for simplicity. Taking the Fourier transform
of the wave equation gives

∂2
F {utt } = c 2 F {uxx } ⇒ U (ω, t ) = –c 2 ω 2 U (ω, t ), (1.1)
∂t
With the initial conditions
∂
U (ω, 0) = F {f (x )} , U (ω, 0) = 0.
∂t
Solving the ODE (1.1) yields

U (ω, t ) = A(ω) cos (cωt ) + B (ω) sin (cωt ) .

Applying the initial condition

∂U (ω, 0)
= cωB (ω) = 0 ⇒ B (ω) = 0,
∂t
and

U (x , 0) = A(ω) = F {f (x )} .

1
Therefore,

U (ω, t ) = A(ω) cos(cωt ), A(ω) = F{f (x )}.

Using the inverse Fourier transform, the solution of the one-dimensional wave equation
is
Z ∞
u(x , t ) = U (ω, 0) cos (cωt ) e –i ωx dω.
–∞

Using Euler’s formula

e iy + e –iy
cos(y) = ,
2
we get
Z ∞
u(x , t ) = U (ω, 0) cos (cωt ) e –i ωx dω
–∞
1Z ∞  
= U (ω, 0) e –i ω(x –ct ) + e –i ω(x +ct ) dω.
2 –∞
Since U (ω, 0) is the Fourier transform of f (x ), we have
Z ∞
f (x ) = U (ω, 0)e –i ωx dω.
–∞

Thus,
1Z ∞  
u(x , t ) = U (ω, 0) e –i ω(x –ct ) + e –i ω(x +ct ) dω
2 –∞
1
= (f (x – ct ) + f (x + ct )) ,
2
which is d’Alembert’s solution.

2 Heat equation on a semi-infinite interval

Suppose we would like to solve the problem



 x ∈ (0, ∞), t ∈ (0, ∞)


u(x , 0) = f (x )


ut = σuxx (2.1)



 u(0, t ) = 0,



limx →∞ u(x , t ) = 0.

2
This corresponds to an infinitly long rod whose temperature we are holding constant at
the left endpoint. Using seperation of variables, we set u(x , t ) = X (x )T (t ) and get the
ODEs

X 00 = λX T 0 = σλT

Let us consider the ODE X 00 = λX . We put the boundary conditions on X (x ) as

X (0) = 0 lim |X (x )| ≤ 0.
x →∞

As we have seen before, the only time a solution of X 00 – λX = 0 stays bounded as x → ∞

is when λ < 0. Letting –ω 2 = λ for ω > 0, we get

X (x ) = c1 cos(ωx ) + c2 sin(ωx ).

Applying the boundary condition at x = 0 gives

X (0) = c1 = 0.

Rename c2 = c. Now, solving T 0 = σλT = –σω 2 T yields

2
T (t ) = T0 e –σω t .

By superposition,
Z ∞
2
u(x , t ) = c(ω) sin(ωx )e –σω t dω.
0

Now, we apply our intial condition

Z ∞
u(x , 0) = c(ω) sin(ωx ) dω = f (x ).
0

Now, since f (0) = 0, let us consider the odd extension of f , called fˆ(x ). Note that the
Fourier transforms can be written as
γ Z∞ˆ
F̂ (ω) = f (x )e i ωx dx
2π –∞
1Z∞
fˆ(x ) = F̂ (ω)e –i ωx dω,
γ –∞
for γ 6= 0. We can insert this γ because
γ
g(x ) = g(x )
γ

3
 γ –1
= F {F {g(x )}}
γ
1
= F–1 {γF {g(x )}} .
γ

Now, sicne fˆ is odd, the Fourier transform of fˆ can be simplified

n o γ Z∞ˆ
F fˆ(x ) = f (x )e i ωx dx
2π Z–∞
γ ∞ˆ
= f (x )(cos(ωx ) + i sin(ωx )) dx
2π –∞
γ Z∞ˆ γi Z ∞ ˆ
= f (x ) cos(ωx ) dx + f (x ) sin(ωx ) dx
2π –∞ 2π –∞
γi Z ∞ ˆ
= f (x ) sin(ωx ) dx
2π –∞
Z ∞
γi
= fˆ(x ) sin(ωx ) dx .
π 0
Additionally,
1Z∞ˆ
F̂ (ω) = f (x )(cos(ωx ) – i sin(ωx )) dω
γ –∞
–2i Z ∞ ˆ
= f (x ) sin(ωx ) dω.
γ 0

We can choose γ = 2/i to get that

ˆ
n o 2 Z ∞ˆ
F f (x ) = f (x ) sin(ωx ) dx .
π 0
This motivates the following definitions

Definition 2.1. For a funtion f (x ) defined on x ∈ (0, ∞), we call the fucntion
2Z∞
F (ω) := f (x ) sin(ωx ) dx
π 0
the Fourier sine transform of f (x ). We denote F (ω) = Fs {f (x )}.

Definition 2.2. For a function F (ω) defined on ω ∈ (0, ∞), we call the function
Z ∞
f (x ) := F (ω) sin(ωx ) dω
0

the inverse Fourier sine transform of F (ω). We denote f (x ) = Fs–1 {F (ω)}.

4
 We can see that
Z ∞
Fs–1 {c(ω)} = c(ω) sin(ωx ) dω = f (x ).
0
So, the full solution to (2.1) is
 
2
u(x , t ) = Fs–1 c(ω)e –σω t c(ω) = Fs {f (x )} .

Suppose we would like to solve the problem




 x ∈ (0, ∞), t ∈ (0, ∞)

ut = σuxx u(x , 0) = f (x )


ux (0, t ) = 0, limx →∞ u(x , t ) = 0.



Using seperation of variables, we set u(x , t ) = X (x )T (t ) and get the ODEs

X 00 = λX T 0 = σλT

Let us consider the ODE X 00 = λX . We put the boundary conditions on X (x ) as

Xx (0) = 0 lim |X (x )| ≤ 0.
x →∞

As we have seen before, the only time a solution of X 00 – λX = 0 stays bounded as x → ∞

is when λ < 0. Letting –ω 2 = λ for ω > 0, we get

X (x ) = c1 cos(ωx ) + c2 sin(ωx ).

Applying the boundary condition at x = 0 gives

X 0 (0) = ωc2 = 0.

Rename c1 = c. Now, solving T 0 = σλT = –σω 2 T yields

2
T (t ) = T0 e –σω t .

By superposition,
Z ∞
2
u(x , t ) = c(ω) cos(ωx )e –σω t dω.
0
Now, we apply our intial condition
Z ∞
u(x , 0) = c(ω) sin(ωx ) dω = f (x ).
0
Following similar arguments as before, motivate the following definitions

5
Definition 2.3. For a funtion f (x ) defined on x ∈ (0, ∞), we call the fucntion
2Z∞
F (ω) := f (x ) cos(ωx ) dx
π 0
the Fourier cosine transform of f (x ). We denote F (ω) = Fc {f (x )}.

Definition 2.4. For a function F (ω) defined on ω ∈ (0, ∞), we call the function
Z ∞
f (x ) := F (ω) cos(ωx ) dω
0

the inverse Fourier cosine transform of F (ω). We denote f (x ) = Fc–1 {F (ω)}.

2.1 Sine and cosine transforms on derivatives

Note that as before,
( )
∂ 2Z∞ ∂
Fs u(x , t ) = u(x , t ) sin(ωx ) dx
∂t π 0 ∂t
∂ 2Z∞
= u(x , t sin(ωx ) dx
∂t π 0
∂
= Fs {u(x , t )} .
∂t
Similarly,
( )
∂ ∂
Fc u(x , t ) = Fc {u(x , t )} .
∂t ∂t

Now let us consider derivatives with respect to x . Recall that we are assuming that
limx →∞ u(x , t ) = 0. Using intergration by parts, we see that
( )
∂ 2Z∞ ∂
Fs u(x , t ) = u(x , t ) sin(ωx ) dx
∂x π 0 ∂x
∞
2  2Z∞
= f (x ) sin(ωx ) – ω
 f (x ) cos(ωx ) dx
π 0 π 0
= –ωFc {u(x , t )} .

Additionally, using integration by parts again gives

( )
∂ 2Z∞ ∂
Fc u(x , t ) = u(x , t ) cos(ωx ) dx
∂x π 0 ∂x

6
 2 ∞ 2Z∞

= u(x , t ) cos(ωx ) + ω
 u(x , t ) sin(ωx ) dx
π 0 π 0
2
= – u(0, t ) + ωFs {u(x , t )} .
π
We can now compute the second sine and cosine transforms of second derivatives with
respect to x . Computing,

∂2
( ) ( )
∂
Fs u(x , t ) = –ωFc u(x , t )
∂x 2 ∂x
2
 
= –ω – u(x , t ) + ωFs {u(x , t )}
π
2
= ωu(0, t ) – ω 2 Fs {u(x , t )} .
π
Similarly, we compute

∂2
( ) ( )
2 ∂u(0, t ) ∂
Fc 2 u(x , t ) = – + ωFs u(x , t )
∂x π ∂x ∂x
2 ∂u(0, t )
=– – ω 2 Fc {u(x , t )} .
π ∂x
We are now ready to solve (2.1). We take the Fourier sine transform of both sides of
(2.1) to get

Fs {ut (x , t )} = Fs {σuxx (x , t )} ⇒ U t (ω, t ) = –σω 2 U (ω, t ).

Solving the resulting ODE above gives

2
U (ω, t ) = c(ω)e –σω t .

Applying the intitial condition gives

U (ω, 0) = c(ω) = Fs {f (x )} .

Therefore, the solution is

2 2Z∞
U (ω, t ) = c(ω)e –σω t , c(ω) = f (x ) sin(ωx ) dx .
π 0
It should be noted that c(ω) is an odd function, since c(–ω) = –c(ω). Thus,
1Z ∞ 2
u(x , t ) = c(ω)e –σω t sin(ωx ) dω
2 –∞

7
 e i ωx – e –i ωx
!
1Z ∞ 2
= c(ω)e –σω t dω
2 –∞ 2i
1Z ∞ –σω 2 t e i ωx 1Z ∞ 2 e –i ωx
= c(ω)e – c(ω)e –σω t dω
2 –∞ 2i 2 –∞ 2i
1Z ∞ –σω 2 t e i ωx 1Z ∞ –σω 2 t e i ωx
= c(ω)e + c(ω)e dω
2 –∞ 2i 2 –∞ 2i
Z ∞
c(ω) –σω2 t i ωx
= e e dω. (2.2)
–∞ 2i

If we introduce the odd extension of f (x ) called fˆ(x ), we see that

Z ∞
c(ω) 2 sin(ωx )
= f (x ) dx
2i π 0 2i
Z ∞
1 sin(ωx )
= fˆ(x ) dx
π –∞ 2i
i ωx – e –i ωx
Z ∞ !
1 1 e
= fˆ(x ) dx
π –∞ 2i 2i
1 Z∞ˆ i ωx 1 Z∞ˆ
=– f (x )e dx + f (x )e –i ωx dx
4π –∞ 4π –∞
1 Z∞ˆ –i ωx 1 Z∞ˆ
= f (x )e dx + f (x )e –i ωx dx
4π –∞ 4π –∞
1 Z∞ˆ
= f (x )e –i ωx dx . (2.3)
2π –∞
Now, we know from before that (2.2) and (2.3) are the results from the heat equation on
the real line. Therefore,
Z ∞
1 2
u(x , t ) = fˆ(y) √ e –(x –y) /(4σt ) dy.
–∞ 4πσt
Now, our problem (2.1) only uses f (x ), not it’s odd extension. So, we write
Z ∞
1 2
fˆ(y) √ e –(x –y) /(4σt ) dy
–∞ 4πσt
"Z #
1 0 2 /(4σt )
Z 0
2 /(4σt )
=√ –f (–y)e –(x –y) dy + f (y)e –(x –y) dy
4πσt –∞ –∞
"Z #
1 ∞ 2 /(4σt )
Z 0
2 /(4σt )
=√ –f (z )e –(x +z ) dz + f (y)e –(x –y) dy (sub z = –y)
4πσt 0 –∞
Z ∞
1
 
–(x –y) 2 /(4σt ) –(x +y) 2 /(4σt )
=√ f (y) e –e dy.
4πσt 0

8
3 Laplace’s Equation in the half-plane
Suppose that we want to solve the problem



 x ∈ (–∞, ∞), y ∈ (0, ∞),


u(x , 0) = f (x )


uxx + uyy = 0 (3.1)



 limx →±∞ u(x , y) = 0, y ∈ (0, ∞)

x ∈ (–∞, ∞).


limy→∞ u(x , y) = 0,
This is Laplace’s equation in the upper-half plane. We can think of this problem as
modeling the steady-state heat behavior of an infinite wall, whose temperature is specified.
We see that u has homogenous “boundary” conditions in x . So, we think perhaps we
should use a Fourier transform in x to solve this problem. Note that

F {uxx (x , y)} = (–i ω)2 F{u(x , y)} = –ω 2 U (ω, y).

Furthermore,
n o 1 Z ∞ ∂2
F uyy (x , y) = u(x , y)e i ωx dx
2π –∞ ∂y 2
∂2 1 Z∞
 
= u(x , y)e i ωx dx
∂y 2 2π –∞
∂2
= F {u(x , y)}
∂y 2
= U yy (ω, y).

Therefore, our transformed problem is the ODE

U yy (ω, y) – ω 2 U (ω, y) = 0. (3.2)

This ODE has the solution

U (ω, y) = a(ω)e ωy + b(ω)e –ωy , (3.3)

along with the two boundary conditions

lim U (ω, y) = 0, U (ω, 0) = F {f (x )} . (3.4)

y→∞

To clarify, the boundary condition as y → ∞ comes from the fact that

1 Z∞
lim F {u(x , y)} = lim u(x , y)e i ωx dx
y→∞ y→∞ 2π –∞

9
 Z ∞
1
= e i ωx lim u(x , y) dx
2π –∞ y→∞
Z ∞
1
= e i ωx · 0 dx
2π –∞
= 0.

So, how do we make our boundary conditions (3.4) match our solution (3.3)? These
boundary condition are suppose to hold for all ω ∈ (–∞, ∞). Okay, it is tempting to say
that

lim a(ω)e ωy + b(ω)e –ωy = 0 ⇒ a(ω) = 0.

y→∞

However, this is false because ω can be negative. What we do to match the boundary
condition as y → ∞ is the following: we set

 a(ω)e ωy ω≤0
U (ω, y) = (3.5)
 b(ω)e –ωy ω > 0.

Note that (3.5) still satisfies the ODE (3.2). It is more convienient to note that we can
rewrite

U (ω, y) = c(ω)e –|ω|y .

From this form, we can apply the nonhomogeneous boundary condition at y = 0 to see
that

U (ω, 0) = c(ω) = F{f (x )}.

Therefore, the solution to (3.2) is

U (ω, y) = c(ω)e –|ω|y , where c(ω) = F {f (x )} .

Furthermore, the solution to (3.1) is u(x , y) = F–1 {U (ω, y)}. In order to compute
n o
F–1 {U (ω, y)}, we first need to compute F–1 e –|ω|y . We can actually compute this di-
rectly:
n o Z ∞
F–1 e –|ω|y = e –|ω|y e –i ωx dω
–∞
Z 0 Z ∞
= e ωy e –i ωx dω + e –ωy e –i ωx dω
–∞ 0

10
 Z 0 Z ∞
= e ω(y–ix ) dω + e –ω(y+ix ) dω
–∞ 0
e ω(y–ix ) 0 e –ω(y+ix ) ∞
 
= –
y – ix –∞ y + ix 0
1 1
= +
y – ix y + ix
2y
= 2 .
x + y2
Let
2y
g(x , y) = , G(ω, y) = e –|ω|y .
x 2 + y2
Note that F {g(x , y)} = G(ω, y) and recall that F {f (x )} = c(ω). Therefore, we can use
the convolution theorem to take the inverse Fourier transform of U (ω, y). Note,
n o
u(x , y) = F–1 U (ω, y)
= F–1 {c(ω)G(ω, y)}
= f (x ) ∗ g(x , y)
1 Z∞ 2y
= f (z ) · dz . (3.6)
2π –∞ (x – z )2 + y 2
Example 3.1. Suppose that we wish to solve (3.1) where

 0 x <0
f (x ) =
 1 x ≥ 0.

Then, by (3.6), we have the solution

1 Z∞ 2y
u(x , y) = f (z ) · dz
2π –∞ (x – z )2 + y 2
1 Z∞ 2y
= dz
2π 0 (x – z )2 + y 2
yZ∞ 1
= dz .
π 0 (x – z )2 + y
Letting u = x – z and –du = dz , we get
yZ∞ 1 y Z –∞ 1
2 dz = – 2 du
π 0 (x – z ) + y π x u +y
yZx 1
= 2 du
π –∞ u + y

11
 x
!! 
y 1 u
= arctan 
π y y 
–∞
!
1 x 1
= arctan + .
π y 2

Let θ be the angle from the y-axis. Then

θ 1
u(x , y) = + .
π 2

12