Coursework 4: Fourier transforms

(1) Using Fourier transforms, solve the heat equation on the infinite line (−∞ < x < ∞) subject to the
initial conditions 2
(a) u(x, 0) = e−x /4
x 2
(b) u(x, 0) = − e−x /4
2
(c) u(x, 0) = aδ(x) + bδ ′ (x),
(d) u(x, 0) = sin κx,
where δ(x) is Dirac’s delta function, and a, b and κ are constants.

(2) Reduce the solution to

utt + u + uxxxx = 0, −∞ < x < ∞, u(x, 0) = f (x), ut (x, 0) = g(x).

to an integral over k.

(3) Prove that the Fourier transform of the convolution,

Z ∞
f (x) ◦ g(x) ≡ f (x − u)g(u)du,
−∞

is the product, fˆ(k)ĝ(k). Using Fourier transforms, solve the integral equation for f (x),
Z ∞
2
e−x /2
= e−|x−u| f (u)du.
−∞

(4) The Fourier sine transform and its inverse are

∞ ∞
2
Z Z
FS {f (x)} = f (x) sin(kx)dx, FS−1 {f (x)} = f (x) sin(kx)dk.
0 π 0

If u → 0 and ux → 0 for x → ∞, show that

FS {uxx (x, t)} = −k 2 FS {u(x, t)} + ku(0, t)

and
k
FS {e−ax } = .
a2 + k2
Use the Fourier sine transform to write the solution to

ut = uxx + e−x , 0 ≤ x < ∞, u(0, t) = 0, u(x, 0) = 0,

in terms of an inverse sine transform.

1
 Coursework 4: Fourier transforms; warm-ups

(1) Solve the heat equation on the infinite line (−∞ < x < ∞) subject to the initial conditions
2 x 2
(a) u(x, 0) = e−x /4
, (b) u(x, 0) = − e−x /4 ,
2
(c) u(x, 0) = δ(x), (d) u(x, 0) = sin κx,
where δ(x) is Dirac’s delta function and κ is a constant.
Fourier transforming the heat equation and integrating implies that û(k, t) = fˆ(k)e−k t , or
2

∞
1
Z
2
u(x, t) = √ f (y)e−(x−y) /4t
dy.
4πt −∞

For (a), the latter gives

2
e−x /4(1+t)
u(x, t) = √ .
1+t
The solution to (b) is the x−derivative of the solution to (a)! For (c), using the properties of the delta-
function, we find
2
e−x /4t
u(x, t) = √ .
4πt
For (d), we have fˆ(k) = −iπ[δ(k − κ) − δ(k + κ)], and so u = e−κ t sin κx.
2

(2) Solve the heat equation on the infinite line (−∞ < x < ∞) subject to the initial conditions
2
u(x, 0) = e−x /2
sin γx

for some parameter γ.

We know, or can compute
2 √ −k2 /4
F{e−x } = πe
The shifting theorems, F{f (ax)} = fˆ(k/a)/|a| and F{eiax f (x)} = fˆ(k − a) therefore imply that
2 √ 2 2 √ 2
F{e−x /2
}= 2πe−k /2 F{e±iγx−x /2
}= 2πe−(k±γ) /2 .

Hence 2 √ 2 2
F{e−x /2
sin γx} = −i 2πe−(k +γ )/2 sinh γk.
Thus,
∞
1
Z
2
+γ 2 )/2−k2 t
u(x, t) = √ eikx−(k sinh(γk) dk
i 2π −∞

Finally, we use the fact that

∞ ∞−iX/2a
r
π
Z Z
−ak2 +ikX+kY (iX+Y )2 /4a −az 2 (iX+Y )2 /4a
e dk = e e dz = e ,
−∞ −∞−iX/2a a

equivalent to what was established in class, to obtain

 2
(γ − x2 ) γ 2
  
1 γx
u(x, t) = √ exp − sin .
1 + 2t 2(1 + 2t) 2 1 + 2t

2
 (3) Solve
ut = −tux , −∞ < x < ∞, u(x, 0) = f (x).
Verify your solution by direct substitution into the PDE.
Fourier transforming:
2
ût = −iktû, → û(k, t) = fˆ(k)e−ikt /2

in view of the transformed initial condition. Hence

Z ∞
1 2
u(x, t) = fˆ(k)eikx−ikt /2 dk
2π −∞

≡ f (x − t2 /2)
We have ut = −tf ′ (x − t2 /2) and ux = f ′ (x − t2 /2) by the chain rule; hence ut = −tux .

(4) For prescribed g(x) and K(x),

Z ∞
g(x) = K(x − y)f (y)dy
−∞

defines an integral equation for f (x). Solve this equation by first taking the Fourier transform, and finding
an expression for fˆ(k), and then undoing the Fourier transform. If K(x) = ag(x − b), for some constants a
and b, what is f (x)? Find the solution if K(x) = g − gxx .
The Fourier transform indicates that

ĝ(k) = K̂(k)fˆ(k).

Hence ∞
1 ĝ(k)
Z
f (x) = eikx dk.
2π −∞ K̂(k)
If K(x) = ag(x − b), then K̂(k) = ae−ikb ĝ(k) (using a shifting theorem) and so f (x) = δ(x − b)/a, using
the definition of Dirac’s delta function.
If K(x) = g − gxx , then K̂(k) = (1 + k 2 )ĝ(k), and so f = F −1 {(1 + k 2 )−1 }. This inverse transform
can be determined by either noting that F{e−|x| } = 2/(1 + k 2 ), or by direct computation of the integral
(achieved by extending it to an infinite semi-circular arc on the complex plane and evaluating the residues
of the poles at k = ±i, depending on which is enclosed). Thence, f (x) = e−|x| /2.

(5) The Fourier cosine transform and its inverse are

Z ∞
2 ∞
Z
FC {f (x)} = f (x) cos(kx)dx, FC−1 {f (x)} = f (x) cos(kx)dk.
0 π 0

If u → 0 and ux → 0 for x → ∞, show that

FC {uxx (x, t)} = −k 2 FC {u(x, t)} − ux (0, t)

and
a
FC {e−ax } = .
a2 + k2
Use the Fourier cosine transform to write the solution to

ut = uxx , 0 ≤ x < ∞, ux (0, t) = 0, u(x, 0) = e−x ,

in terms of an inverse cosine transform.

3
 Using the definitions and integration by parts,
Z ∞ Z ∞
FC {uxx } = uxx cos(kx)dx = −ux (0, t) + k ux sin(kx)dx
0 0
Z ∞
= −ux (0, t) − k 2 u cos(kx)dx = −ux (0, t) − k 2 FC {u}.
0

Then,
∞ ∞ ∞
1 k 1 k2
Z Z Z
−ax −ax −ax
FC {e }= e cos(kx)dx = − e sin(kx)dx = − 2 e−ax cos(kx)dx,
0 a a 0 a a 0

giving the needed result.

Applying the cosine transform, we find
2
∂ e−k t
ûC = −k 2 ûC → ûC (k, t) = ,
∂t 1 + k2

where ûC (k, t) = FC {u(x, t)}. Hence

( 2
)
e−k t
u(x, t) = FC−1 .
1 + k2

4
 More from 2018

(1) Reduce the solution to the beam equation,

σutt − N uxx + Buxxxx = 0, −∞ < x < ∞, u(x, 0) = f (x), ut (x, 0) = g(x),

to an integral over k, where σ, N and B are constants. For B = 0, recover d’Alembert’s solution to the wave
equation.

(2) Using Fourier transforms, solve the integral equation for f (x),
Z ∞
2 2
e−(x/α) = e−(x−u) f (u)du,
−∞

for (a) α > 1 and (b) α = 1.

(3) The Fourier sine transform and its inverse are

∞ ∞
2
Z Z
FS {f (x)} = f (x) sin(kx)dx, FS−1 {f (x)} = f (x) sin(kx)dk.
0 π 0

If u → 0 and ux → 0 for x → ∞, show that

FS {uxx (x, t)} = −k 2 FS {u(x, t)} + ku(0, t)

and
k
FS {e−ax } = .
a2 + k2
Use the Fourier sine transform to write the solution to

ut = uxx + e−x , 0 ≤ x < ∞, u(0, t) = 0, u(x, 0) = 0,

in terms of an inverse sine transform.

5
 Solutions

(1) Transforming the PDE and initial conditions:

r
k2
ûtt = −Ω û, 2
Ω(k) = (Bk 2 + N ) û(k, 0) = fˆ(k), ût (k, 0) = ĝ(k).
σ

Hence,
∞ 
1 ĝ(k)
Z
u(x, t) = fˆ(k) cos[Ω(k)t] + sin[Ω(k)t] eikx dk.
2π −∞ Ω(k)
If B = 0 and N k/σ = Ω = kc, then
∞   Z x+ct
1 ĝ(k) 1 1
Z
u(x, t) = fˆ(k) cos(kct) + sin(kct) eikx dk = [f (x + ct) + f (x − ct) + g(z)dz.
2π −∞ kc 2 2c x−ct

(2) The Fourier Transform of the integral equation is

2
/α2 2 2
(α2 −1)/4
F{e−x } = fˆ(k)F{e−x } or αe−k = fˆ(k).

Inverting the transform for (a) gives

α 2
√
/ α2 −1
f (x) = p e−x .
π(α2 − 1)

For (b), we have fˆ = 1, and so f (x) = δ(x).

(3) Using the definitions and integration by parts,

Z ∞ Z ∞
FS {uxx } = uxx sin(kx)dx = −k ux cos(kx)dx
0 0
Z ∞
2
= ku(0, t) − k u sin(kx)dx.
0

Then,
∞ ∞ ∞
k k k2
Z Z Z
FC {e−ax } = e−ax sin(kx)dx = e−ax cos(kx)dx = 2
− 2 e−ax sin(kx)dx,
0 a 0 a a 0

giving the needed result.

Applying the sine transform, we find

∂ k 1 2
ûS = −k 2 ûS + → ûS (k, t) = (1 − e−k t ),
∂t 1 + k2 k(1 + k 2 )

where ûS (k, t) = FS {u(x, t)}. Hence

( 2
)
(1 − e−k t )
u(x, t) = FS−1 .
k(1 + k 2 )

6
 More 2019

(1) Using Fourier transforms, find fˆ(k) = F{u(x, 0)} and solve the heat equation on the infinite line
(−∞ < x < ∞) subject to the initial conditions
2
(a) u(x, 0) = ex−x

(b) u(x, 0) = H(x + 1) − H(x − 1)

(c) u(x, 0) = δ ′ (x − 1),
(d) u(x, 0) = cos2 x,
Rz 2
where H(x) is the Heaviside step function, δ(x) is Dirac’s delta function. Note that Erf(z) = √2 e−z dz.
π 0

(2) Consider the heat equation with a moving source:

ut = uxx + eγt G(x − ct), u(x, 0) = 0, (u, G) → for x → ±∞,

where γ and c are parameters. Find û(k, t) = F{u(x, t)}. Then, if g(z) satisfies G(z) = γg − cg ′ − g ′′ ,
establish that Z ∞
2 dy
u(x, t) = eγt g(x − ct) − g(y)e−(x−y) /4t √ .
−∞ 4πt

(3) Using Fourier transforms, solve the integral equation for f (x),
Z ∞
−x2
e = e−u f (x − u)du.
0

(4) The Fourier sine transform and its inverse are

∞ ∞
2
Z Z
FS {f (x)} = f (x) sin(kx)dx, FS−1 {f (x)} = f (x) sin(kx)dk.
0 π 0

If u → 0 and ux → 0 for x → ∞, show that

FS {uxx (x, t)} = −k 2 FS {u(x, t)} + ku(0, t)

and
k
FS {e−ax } = .
a2 + k 2
Use the Fourier sine transform to write the solution to

ut = uxx + xe−x−t , 0 ≤ x < ∞, u(0, t) = 0, u(x, 0) = 0,

in terms of an inverse sine transform.

7
 Solutions

(1) Fourier transforming the heat equation and integrating implies that û(k, t) = fˆ(k)e−k
2
t
or
Z ∞
1 2
u(x, t) = √ f (y)e−(x−y) /4t dy.
4πt −∞
For (a), the latter gives  2
(x + 2t)2

1 x
u(x, t) = √ exp − + .
1 + 4t 4t 4t(1 + 4t)
In (b), we have fˆ(k) = 2k −1 sin k, and
Z (1−x)/√4t     
1 −z 2 1 1−x 1+x
u(x, t) = √ e dz = Erf √ − Erf − √ .
π −(1+x)/√4t 2 2 t 2 t
For (c), fˆ(k) = 1, and using the properties of the delta-function and an integration by parts, we find
(1 − x) 2
u(x, t) = √ 3/2 e−(x−1) /4t .
4 πt
For (d), we have f (x) = cos2 x = 1
2 (cos 2x + 1) and so fˆ(k) = π
2 [δ(k − 2) + δ(k + 2) + 2δ(k)], giving
u = (1 + e−4t cos 2x)/2.

(2) Fourier transforming the heat equation, and using F{G(x − ct)} = Ĝ(k)e−ikct , we obtain
2 Ĝ(k) 2
û(k, t) = (eγt−ikct − e−k t ) = (eγt−ikct − e−k t )ĝ(k)
γ + k 2 − ikc
2
Inverting the transform now provides the needed result, given that the e−k t
term corresponds to a known
convolution integral.

(3) The Fourier Transform of the integral equation is

2 2 √
F{e−x } = fˆ(k)F{e−x H(x)} or (1 + ik)e−k /4
π = fˆ(k).
Inverting the transform gives
2
f (x) = (1 − 2x)e−x .

(4) Using the definitions and integration by parts,

Z ∞ Z ∞
FS {uxx } = uxx sin(kx)dx = −k ux cos(kx)dx
0 0
Z ∞
= ku(0, t) − k 2 u sin(kx)dx.
0
Then,
∞ ∞ ∞
k k k2
Z Z Z
FS {e−ax } = e−ax sin(kx)dx = e−ax cos(kx)dx = 2
− 2 e−ax sin(kx)dx,
0 a 0 a a 0
giving the needed result. Note that
d 2ak
FS {xe−ax } ≡ − FS {e−ax } = 2 .
da (a + k 2 )2
Applying the sine transform, we find
2
∂ 2ke−t 2k(e−t − e−k t )
ûS = −k 2 ûS + → ûS (k, t) = 2 ,
∂t (1 + k 2 )2 (k − 1)(1 + k 2 )2
where ûS (k, t) = FS {u(x, t)}, and then u(x, t) = FS−1 {ûS (k, t)}.