MA 201, Mathematics III, July-November 2022,

Fourier Transform (Contd.)

Lecture 19

Lecture 19 MA 201, PDE (2022) 1 / 24

Fourier Transform
For a given function f , the improper integral
Z 
1 1 1
Z ∞ ∞ Z ∞
√ eiσt √ f (τ )e−iστ dτ dσ = √ eiσt g(σ)dσ
−∞ 2π −∞ 2π −∞ 2π
is called Fourier integral of f .

If f is continuous at t, we have
1
Z ∞
f (t) = √ eiσt g(σ)dσ.
−∞ 2π

The function g is called Fourier transform of a function f and it is denoted by

F {f (t)}
1
Z ∞
F{f (t)} = g(σ) = √ e−iστ f (τ ) dτ. (1)
2π −∞

The inverse Fourier transform is defined as

1
Z ∞
f (t) = F −1 {g(σ)} = √ g(σ)eiσt dσ. (2)
2π −∞

Lecture 19 MA 201, PDE (2022) 2 / 24

Fourier cosine transform
Suppose f is defined in (0, ∞).

For an even extension fe defined in (−∞, ∞), the Fourier integral is

Z 
1 1
Z ∞ ∞
√ eiσt √ fe (τ )e−iστ dτ dσ (Basic definition)
−∞ 2π −∞ 2π
2
Z ∞Z ∞
= fe (τ ) cos στ cos σt dτ dσ (Special case)
π 0 0
r "Z r #
2 2
Z ∞ ∞
= cos σt f (τ ) cos στ dτ dσ
0 π 0 π
r
2
Z ∞
= cos σtgc (σ)dσ
0 π

We can define the Fourier cosine transform of f as

1
Z ∞
Fc {f (t)} = gc (σ) = √ e−iστ fe (τ ) dτ (If we use extension)
2π −∞
r Z
2 ∞
= f (τ ) cos στ dτ. (From the special case)
π 0

The inverse Fourier cosine transform is defined as

r
2
Z ∞
Fc−1 {gc (σ)} = f (t) = gc (σ) cos σt dσ. (From the special case)
π 0
1
Z ∞
= √ eiσt gc (σ)dσ. (If we use extension)
−∞ 2π

Lecture 19 MA 201, PDE (2022) 3 / 24

Fourier sine transform
For an odd extension fo defined in (−∞, ∞), the Fourier integral is
Z 
1 1
Z ∞ ∞
√ eiσt √ fo (τ )e−iστ dτ dσ (Basic definition)
−∞ 2π −∞ 2π
2
Z ∞Z ∞
= fo (τ ) sin στ sin σt dτ dσ (From the special case)
π 0 0
r "Z r #
2 2
Z ∞ ∞
= sin σt f (τ ) sin στ dτ dσ
0 π 0 π
r
2
Z ∞
= sin σtgs (σ)dσ
0 π

We can define the Fourier sine transform of f as

1
Z ∞
Fs {f (t)} = gs (σ) = √ e−iστ fo (τ ) dτ (If we use extension)
2π −∞
r Z
2 ∞
= f (τ ) sin στ dτ. (From the special case)
π 0

The inverse Fourier sine transform is defined as

r Z ∞
2
Fs−1 {gs (σ)} = f (t) = gs (σ) sin σt dσ. (From the special case)
π
Z ∞ 0
1 iσt
= √ e gs (σ)dσ. (If we use extension)
−∞ 2π

Lecture 19 MA 201, PDE (2022) 4 / 24

Example
Consider f (t) = e−t , t > 0.
Justify that
(a) f has Fourier sine and cosine transforms.
(b) Find Fourier cosine transform of f .
(c) Use part (b) to evaluate the following improper integral:

cos σt
Z ∞
dσ, t > 0.
0 σ2 + 1

(a) Use even and odd extensions of f in (−∞, ∞).

(b) Suppose fe is the even extension of f in (−∞, ∞). Then

1
Z ∞
Fc {fe (t)} = gc (σ) = √ e−iστ fe (τ ) dτ (From Basic definition)
2π −∞
1 2
= √
2π 1 + σ 2
Lecture 19 MA 201, PDE (2022) 5 / 24
Example (Contd.)
(c) Finally, use inverse Fourier cosine transform to have
r Z
2 ∞
fe (t) = Fc−1 {gc (σ)} = gc (σ) cos σt dσ. (From Special case)
π
Z ∞ 0
2 cos σt
= dσ. (From part (b))
0 π σ2 + 1

The required integral has the form

cos σt
Z ∞
e−t π
dσ =
0 σ2 + 1 2

What about the following integrals?

σ 3 sin σt
Z ∞
(a) dσ, t > 0.
σ4 + 4
Z 0∞
sin σt
(b) dσ, t > 0.
0 σ

Here, f is not given explicitly − for (a), we can take f (t) = e−t cos t, t > 0.

Lecture 19 MA 201, PDE (2022) 6 / 24

Application of Fourier Transform to PDEs
Fourier transform:
The Fourier transform of a function U (x, t) with respect to x is defined as

1
Z ∞
F{U (x, t)} = √ e−iσx U (x, t) dx = U (σ, t). (3)
2π −∞

Inverse Fourier transform:

The inverse Fourier transform U (x, t) of U (σ, t) is defined as

1
Z ∞
U (x, t) = F −1
{U (σ, t)} = √ eiσx U (σ, t) dσ.
2π −∞

Lecture 19 MA 201, PDE (2022) 7 / 24

Application of Fourier Transform to PDEs

Under the assumption that U and ∂U ∂x

vanish as x → ±∞, we obtain the
following results of transforms of the partial derivatives:
 
∂
F U (x, t) = iσU (σ, t),
∂x
 2 
∂
F U (x, t) = (iσ)2 U (σ, t),
∂x2
 
∂ d
F U (x, t) = U (σ, t),
∂t dt
 2
d2

∂
F 2
U (x, t) = 2 U (σ, t).
∂t dt

Lecture 19 MA 201, PDE (2022) 8 / 24

Application of Fourier Transform to PDEs
Result I:
 
∂
F U (x, t) = iσU (σ, t). (4)
∂x

Proof:
 
1
Z ∞
∂ ∂U
F U (x, t) =√ e−iσx dx
∂x 2π −∞ ∂x
 
1
Z ∞
= √ [e−iσx U (x, t)]∞
−∞ + iσ e−iσx
U (x, t)dx
2π −∞
= iσU (σ, t).

Lecture 19 MA 201, PDE (2022) 9 / 24

Application of Fourier Transform to PDEs
Result II:
∂2
 
F U (x, t) = (iσ)2 U (σ, t). (5)
∂x2

Proof:

∂2 ∂2U
  Z ∞ 
1
F U (x, t) =√ e−iσx 2 dx
∂x2 2π ∂x
 −∞ 
1
Z ∞
∂U ∂U
= √ [e−iσx ]∞
−∞ + iσ e−iσx dx
2π ∂x −∞ ∂x
 
1
Z ∞
= √ iσ([e−iσx U (x, t)]∞
−∞ + iσ e−iσx
U (x, t)dx)
2π −∞
= (iσ)2 U (σ, t).

Lecture 19 MA 201, PDE (2022) 10 / 24

Application of Fourier Transform to PDEs
Result III:
 
∂ d
F U (x, t) = U (σ, t). (6)
∂t dt

Proof:
 
1
Z ∞
∂ ∂U
F U (x, t) = √ e−iσx dx
∂t 2π −∞ ∂t
1 ∂
Z ∞
= √ e−iσx U (x, t)dx
2π ∂t −∞
1 d
Z ∞
= √ e−iσx U (x, t)dx
2π dt −∞
d
= U (σ, t).
dt


Proceeding in a similar manner as above, Result 4 can be obtained as follows:

∂2 d2
 
F U (x, t) = 2 U (σ, t). (7)
∂t2 dt


Lecture 19 MA 201, PDE (2022) 11 / 24

Application of Fourier Transform to PDEs
The use of Fourier transform to solve partial differential equations is best described by
examples.

Example A:
An infinitely long string extending in −∞ < x < ∞ under uniform tension is displaced
into the curve y = f (x) and let go from rest with velocity g(x). To find the
displacement U (x, t) at any point at any subsequent time.

Solution:
The initial value problem is

∂2U ∂2U
2
= c2 2 , −∞ < x < ∞, t > 0, (8)
∂t ∂x
U (x, 0) = f (x) (initial displacement), (9)
∂U
(x, 0) = g(x) (initial velocity). (10)
∂t
Lecture 19 MA 201, PDE (2022) 12 / 24
Application of Fourier Transform to PDEs
Taking Fourier transform on both sides of PDE (8),
d2
U (σ, t) = −c2 σ 2 U (σ, t). (11)
dt2

(11) can be written in standard form as

d2
U (σ, t) + c2 σ 2 U (σ, t) = 0. (12)
dt2

On solving, we get

U (σ, t) = A(σ) cos(cσt) + B(σ) sin(cσt). (13)

Taking Fourier transforms on the initial conditions (9) and (10),

U (σ, 0) = f (σ), (14)

d
U (σ, 0) = g(σ), (15)
dt
where f (σ) and g(σ), are, respectively the Fourier transform of f (x) and g(x).

Lecture 19 MA 201, PDE (2022) 13 / 24

Application of Fourier Transform to PDEs
Using the initial conditions, A(σ) and B(σ) can be obtained as:

U (σ, 0) = A(σ) = f (σ),

d
U (σ, 0) = cσB(σ) = g(σ).
dt

Now
1
U (σ, t) = f (σ) cos(cσt) + g(σ) sin(cσt). (16)
cσ

To get the solution, we use the inverse Fourier transform to obtain

 
1 1
Z ∞
U (x, t) = √ f (σ) cos(cσt) + g(σ) sin(cσt) eiσx dσ. (17)
2π −∞ cσ

Formula (17) gives us the solution of the initial value problem for one-dimensional
wave equation (infinite string) in the form of an integral involving the Fourier
transform of the initial displacement and velocity.

Lecture 19 MA 201, PDE (2022) 14 / 24

Application of Fourier Transform to PDEs
Example B:
Consider the heat conduction in an infinite rod with thermal diffusivity α with initial
temperature distribution f (x). To find the temperature distribution U (x, t) at any
point at any subsequent time.

Solution:
The initial value problem is

∂U ∂2U
= α 2 , −∞ < x < ∞, t > 0, (18)
∂t ∂x
U (x, 0) = f (x) (initial temperature distribution). (19)

Taking Fourier transform on both sides of PDE (18), we observe

2
U (σ, t) = f (σ)e−ασ t , (20)

where f (σ) is the Fourier transform of f (x).

We use the inverse Fourier transform to obtain

1
Z ∞
2
U (x, t) = √ f (σ)e−ασ t eiσx dσ. (21)
2π −∞

Lecture 19 MA 201, PDE (2022) 15 / 24

Application of Fourier Transform to PDEs
We summarize the Fourier transform method as follows:

Step 1: Take Fourier transform with respect to x on the given equation in U (x, t)
when −∞ < x < ∞ and get an ordinary differential equation in U (σ, t) in the variable
t.

Step 2: Solve the ordinary differential equation and find U (σ, t).

Step 3: Take Fourier transform with respect to x on the given initial condition(s) and
use them in the transformed equation in U (σ, t) to find the coefficients.

Step 4: Take inverse Fourier transform to get U (x, t).

Lecture 19 MA 201, PDE (2022) 16 / 24

Application of Fourier Transform to PDEs
Example C: Semi-infinite rod with Dirichlet condition

Mathematical Model
The initial boundary value problem is the following:

∂U ∂2U
= α 2 , x > 0, t > 0, (22)
∂t ∂x
U (x, 0) = f (x), t > 0, U (0, t) = U0 , x > 0. (23)

Example D: Semi-infinite rod with Neumann condition

Consider the same equation as in the previous example subject to the boundary and
initial conditions
∂U
(0, t) = 0, U (x, 0) = f (x).
∂x

The Fourier sine and cosine transforms can be employed to solve a partial differential
equation when the range of the spatial variable extends from 0 to ∞, not in
−∞ < x < ∞.
Lecture 19 MA 201, PDE (2022) 17 / 24
Application of Fourier Transform to PDEs

If the boundary condition is in terms of some value of U (0, t) (i.e., Dirichlet boundary
condition), then sine transform is to be used.

∂U
When the boundary condition is in terms of some value of ∂x (0, t) (i.e., Neumann
boundary condition), then cosine transform is to be used.

The Fourier sine transform of a function U(x, t) with respect to x is defined as

r Z
2 ∞
Fs {U (x, t)} = U (x, t) sin σx dx = U s (σ, t). (24)
π 0

Lecture 19 MA 201, PDE (2022) 18 / 24

Application of Fourier Transform to PDEs
∂U
Under the assumption that U and ∂x
vanish as x → ∞,

 
∂U (x, t) d
Fs = U s (σ, t), (25)
∂t dt
 2
d2

∂ U (x, t)
Fs = U s (σ, t), (26)
∂t2 dt2
 2  r
∂ U (x, t) 2
Fs = σU (0, t) − σ 2 U s (σ, t). (27)
∂x2 π

Lecture 19 MA 201, PDE (2022) 19 / 24

Application of Fourier Transform to PDEs

The Fourier cosine transform of a function U(x, t) with respect to x is defined as

r Z ∞
2
Fc {U (x, t)} = U (x, t) cos σx dx = U c (σ, t). (28)
π 0

∂U
Under the assumption that U and ∂x
vanish as x → ∞,

 
∂U (x, t) d
Fc = U c (σ, t), (29)
∂t dt
 2
d2

∂ U (x, t)
Fc = U c (σ, t), (30)
∂t2 dt2
 2  r
∂ U (x, t) 2
Fc =− Ux (0, t) − σ 2 U c (σ, t). (31)
∂x2 π

Lecture 19 MA 201, PDE (2022) 20 / 24

Application of Fourier Transform to PDEs
Example C:
If U (x, t) is the temperature at time t and α the thermal diffusivity of a semi-infinite
metal bar, find the temperature distribution in the bar at any point at any subsequent
time if the initial temperature distribution is given as f (x) and the boundary is kept at
U0 degrees.

Solution:
The initial boundary value problem is the following:

∂U ∂2U
= α 2 , x > 0, t > 0, (32)
∂t ∂x
U (x, 0) = f (x), t > 0 U (0, t) = U0 , x > 0. (33)

The boundary condition suggests that we need to use Fourier sine transform.
Taking the transform on (32),
r
d 2
U s (σ, t) = α[ σU (0, t) − σ 2 U s (σ, t)]. (34)
dt π
Lecture 19 MA 201, PDE (2022) 21 / 24
Application of Fourier Transform to PDEs
Using the boundary condition,
r
d 2 2
U s (σ, t) + ασ U s (σ, t) = α σU0 . (35)
dt π

On solving,
r
−ασ2 t 1 2
U s (σ, t) = A(σ)e + U0 . (36)
σ π

Using the initial condition,

r
2 U0
fs (σ) = A(σ) + ,
π σ
where fs (σ) is the Fourier sine transform of f (x), that is,
r
2 U0
A(σ) = fs (σ) − .
π σ
Lecture 19 MA 201, PDE (2022) 22 / 24
Application of Fourier Transform to PDEs

Now U s (σ, t) is
r r
2 U0 −ασ2 t 2 U0
U s (σ, t) = (fs (σ) − )e +
π σ π σ
r
2 2 U 0 2
= fs e−ασ t + (1 − e−ασ t ). (37)
π σ

The inversion gives

r Z " r #
2 ∞ −ασ2 t 2 U0 −ασ2 t
U (x, t) = fs e + (1 − e ) sin σx dσ. (38)
π 0 π σ

Lecture 19 MA 201, PDE (2022) 23 / 24

Application of Fourier Transform to PDEs

Similarly, we can try the same heat conduction problem with a Neumann condition.

Like-wise, we can try the vibration problem in a semi-infinite string 0 < x < ∞.

In this case, there will be a second-order ordinary differential equation upon using the
2
transform due to the presence of the term ∂∂tU2 .

The two initial conditions U (x, 0) and ∂U ∂t (x, 0) will help us in determining the
coefficients of Uc (σ, t) or Us (σ, t) selection of which will depend on the type of
boundary condition at x = 0.

Lecture 19 MA 201, PDE (2022) 24 / 24