SOLUTIONS MANUAL FOR

ADVANCED TOPICS IN
APPLIED MATHEMATICS

Sudhakar Nair
 c
S. Nair
ii
Contents

1 GREEN’S FUNCTIONS 5

2 INTEGRAL EQUATIONS 59

3 FOURIER TRANSFORMS 93

4 LAPLACE TRANSFORMS 149

1
2 CONTENTS
Preface

This is the solutions manual to accompany the text, “Advanced Topics in

Applied Mathematics.” I have attempted to show all the intermediate steps
in the solutions. For the convenience of posting solutions, a new solution
appears on a new page. Please let me know if anyone notices any errors. Some
figures were created using the MathematicaT M package and I am indebted
to Wolfram Research Incorporated.
S. Nair ( nair@[Link])
Illinois Institute of Technology,
Chicago, Illinois

3
4 CONTENTS
Chapter 1

GREEN’S FUNCTIONS

1.1 The deflection of a beam is governed by the equation

d4 v
EI = −p(x)
dx4
where EI is the bending stiffness and p(x) is the distributed loading
on the beam. If the beam has a length ℓ, and at both the ends the
deflection and slope are zero, obtain expressions for the deflection by
direct integration, using the Macaulay brackets when necessary, if
a) p(x) = p0 ,
b) p(x) = P0 δ(x − ξ),
c) p(x) = M0 δ ′ (x − ξ).
Obtain the Green’s function for the deflection equation from the pre-
ceding calculations.
Solution
Let us make the substitution x → x/ℓ. The derivative of the delta
function in Part (c) will bring 1/ℓ when the non-dimensional x is used.
The beam equation becomes
(a)
p0 ℓ 4 p0 ℓ 4
v ′′′′ = − , v ′′′ = − [x + C1 ],
EI EI
p0 ℓ4 x2
v ′′ = − [ + C1 x + C2 ],
EI 2
p0 ℓ4 x3 x2
v′ = − [ + C1 + C2 x + C3 ],
EI 3 2
p0 ℓ4 x4 x3 x2
v = − [ + C1 + C2 + C3 x + C4 ].
EI 24 6 2
5
6 CHAPTER 1. GREEN’S FUNCTIONS

Using the boundary conditions

v(0) = 0, v ′ (0) = 0, C4 = 0, C3 = 0.
Using
1 1 1 1 1
v(1) = 0, v ′ (1) = 0, C1 + C2 = − , C1 + C2 = − .
2 6 6 2 24

Solving

1 1
C1 = − , C2 = .
2 12

p0 ℓ4 x4 x3 x2
 
v(x) = − − + ,
EI 24 12 24
p0 ℓ 4
= − (x − 1)2 x2 .
24EI

(b)
P0 ℓ4 P0 ℓ4
v ′′′′ = − δ(x − ξ), v ′′′ = − [hx − ξi0 + C1 ],
EI EI
P0 ℓ4
v ′′ = − [hx − ξi1 + C1 x + C2 ],
EI
4
P 0ℓ 1 1
v′ = − [ hx − ξi2 + C1 x2 + C2 x + C3 ],
EI 2 2
P0 ℓ4 1 1 1
v = − [ hx − ξi3 + C1 x3 + C2 x2 + C3 x + C4 ].
EI 6 6 2

Using
v(0) = 0 = v ′ (0), C4 = C3 = 0.
Using
1 1 1 1 1
v(1) = 0 = v ′ (1), C1 +C2 = − (1−ξ)2 , C1 + C2 = − (1−ξ)3,
2 2 6 2 6

Then
C1 = −(1 + 2ξ)(1 − ξ)2 , C2 = ξ(1 − ξ)2.
 7

P0 ℓ4 
hx − ξi3 − (1 + 2ξ)(1 − ξ)2 x3 + 3ξ(1 − ξ)2 x2 .

v(x) = −
6EI

(c)

M0 ℓ3 ′ M0 ℓ3
v ′′′′ = − δ (x − ξ), v ′′′ = − [δ(x − ξ) + C1 ],
EI EI
M0 ℓ3
v ′′ = − [hx − ξi0 + C1 x + C2 ],
EI
3
M0ℓ 1
v′ = − [hx − ξi1 + C1 x2 + C2 x + C3 ],
EI 2
M0 ℓ3 1 1 1
v = − [ hx − ξi2 + C1 x3 + C2 x2 + C3 x + C4 ].
EI 2 6 2

Using
v(0) = 0 = v ′ (0), C3 = C4 = 0,
Using
1 1 1 1
v(1) = 0 = v ′ (1), C1 + C2 = −(1 − ξ), C1 + C2 = − (1 − ξ)2 .
2 6 2 2

Then
C1 = −6ξ(1 − ξ), C2 = (3ξ − 1)(1 − ξ).
M0 ℓ3 
hx − ξi2 − 2ξ(1 − ξ)x3 + (3ξ − 1)(1 − ξ)x2 .

v(x) = −
2EI
The Green’s function in terms of non-dimensional coordinate x/ℓ cor-
responds to P0 = −1;

ℓ4 
hx − ξi3 − (1 − ξ)2x2 (3ξ + x + 2ξx) .

g(x, ξ) =
6EI
(
(1 − ξ)2 x2 (3ξ − x − 2ξx), x < ξ
g(x, ξ) =
(1 − x)2 ξ 2(3x − ξ − 2xξ), x > ξ
8 CHAPTER 1. GREEN’S FUNCTIONS

1.2 Solve the preceding problem when the beam is simply supported. That
is,
v(0) = v(ℓ) = 0, v ′′ (0) = v ′′ (ℓ) = 0.

Solution

Let us make the substitution x → x/ℓ. The derivative of the delta

function will have a 1/ℓ in front when the non-dimensional x is used.
The beam equation becomes

(a)

p0 ℓ 4 p0 ℓ 4
v ′′′′ = − , v ′′′ = − [x + C1 ],
EI EI
p0 ℓ4 x2
v ′′ = − [ + C1 x + C2 ],
EI 2
p0 ℓ4 x3 x2
v′ = − [ + C1 + C2 x + C3 ],
EI 3 2
p0 ℓ4 x4 x3 x2
v = − [ + C1 + C2 + C3 x + C4 ].
EI 24 6 2

Using the boundary conditions

v(0) = 0, v ′′ (0) = 0, C2 = 0, C4 = 0.

Using
1 1
v(1) = 0, v ′′ (1) = 0, C1 = − , C3 = .
2 24

p0 ℓ4 x4 x3
 
x
v(x) = − − + ,
EI 24 12 24
p0 ℓ 4
= − x(x3 − 2x + 1).
24EI
 9

(b)
P0 ℓ4 P0 ℓ4
v ′′′′ = − δ(x − ξ), v ′′′ = − [hx − ξi0 + C1 ],
EI EI
P0 ℓ4
v ′′ = − [hx − ξi1 + C1 x + C2 ],
EI
P0 ℓ4 1 1
v′ = − [ hx − ξi2 + C1 x2 + C2 x + C3 ],
EI 2 2
P0 ℓ4 1 1 1
v = − [ hx − ξi3 + C1 x3 + C2 x2 + C3 x + C4 ].
EI 6 6 2
Using
v(0) = 0 = v ′′ (0), C4 = C2 = 0.
Using
1
v(1) = 0 = v ′′ (1), C1 = ξ − 1, C2 = ξ(1 − ξ)(2 − ξ),
6
P0 ℓ4 
hx − ξi3 − (1 − ξ)(1 − ξ)2 x3 + ξ(1 − ξ)(2 − ξ)x .

v(x) = −
6EI
(c)

′′′′ M0 ℓ3 ′ ′′′ M0 ℓ3
v = − δ (x − ξ), v = − [δ(x − ξ) + C1 ],
EI EI
Note: δ ′ is now the derivative with respect to the non-dimensional variable
x.

′′ M0 ℓ3
v = − [hx − ξi0 + C1 x + C2 ],
EI
3
M0ℓ 1
v′ = − [hx − ξi1 + C1 x2 + C2 x + C3 ],
EI 2
M0 ℓ3 1 1 1
v = − [ hx − ξi2 + C1 x3 + C2 x2 + C3 x + C4 ].
EI 2 6 2
Using
v(0) = 0 = v ′′ (0), C2 = C4 = 0,
Using
1
v(1) = 0 = v ′′ (1), C1 = −1, C3 = [1 − 3(−ξ)2 ].
6
10 CHAPTER 1. GREEN’S FUNCTIONS

M0 ℓ3 
v(x) = − [3hx − ξi2 − x3 + [1 − 3(1 − ξ)2 ]x .
6EI
The Green’s function in terms of the non-dimensional coordinate x/ℓ
corresponds to P0 = −1;

ℓ4 
hx − ξi3 − (1 − ξ)2x3 + ξ(1 − ξ)(2 − ξ)x .

g(x, ξ) =
6EI
(
x(1 − ξ)(2ξ − x2 − ξ 2 ), x < ξ
g(x, ξ) =
ξ(1 − x)(2x − ξ 2 − x2 ), x > ξ
 11

1.3 Obtain the derivative of the function

g(x) = |f (x)|,

in a < x < b, assuming f (x) has a simple zero at the point c inside
the interval (a, b). Use the Signum function and/or delta function to
express the result.

Solution
Let
g(x) = f (x) sgn[f (x)].
(a) Assuming f ′ (c) > 0,

sgn[f (x)] = sgn(x − c).

(b) Assuming f ′ (c) < 0,

sgn[f (x)] = − sgn(x − c).

Then

g(x) = f (x) sgn[f ′ (c)] sgn(x−c), g ′(x) = [f ′ (x) sgn(x−c)+2f (x)δ(x−c)] sgn[f ′ (c)].
12 CHAPTER 1. GREEN’S FUNCTIONS

1.4 Assuming a function f (x) has simple zeros at xi , i = 1, 2, . . . , n, find

an expression for
δ(f (x)).

Solution
Near a zero, x = xi ,

f (x) = f ′ (xi )(x − xi ) + · · · ,

For a simple zero, f ′ (xi ) 6= 0.
Using a test function φ,

Z ∞ n Z
X xi +ǫ
δ(f (x))φ(x)dx = φ(x)δ[f ′ (xi )(x − xi )]dx
−∞ i=1 xi −ǫ

Xn Z ǫ
= φ(xi + ξ)δ[f ′(xi )ξ]dξ,
i=1 −ǫ
n ǫ  
1 η
X Z
= ′
φ xi + δ(η)dη
i=1
f (xi ) −ǫ f ′ (xi )
n
X 1
= φ(xi ).
i=1
f ′ (xi )

Then n
X 1
δ(f ) = δ(x − xi ).
i=1
f ′ (xi )
 13

1.5 Convert the equation

d2 u du
Lu = 2
+ x2 + 2u = f
dx dx
into the Sturm-Liouville form.

Solution
The Sturm-Liouville form is

(pu′ )′ + qu = f ∗ .

Expanding
p′ ′ q f∗
u′′ + u + u= .
p p p
Comparing with the given equation

p′ q f∗
= x2 , = 2, = f.
p p p
3 /3 3 /3 3 /3
log p = x3 /3, p = ex , q = 2ex , f ∗ = ex f.

3 /3 3 /3 3 /3
(ex u′ )′ + 2ex u = ex f.
14 CHAPTER 1. GREEN’S FUNCTIONS

1.6 Find the adjoint system for

xu′′ + u′ + u = 0, u(1) = 0, u(2) + u′ (2) = 0.

Solution

hv, Lui − hu, L∗ vi = 0.

Z 2
hv, Lui = v[xu′′ + u′ + u]dx
1
Z 2
′ ′ 2
= [xvu − (xv) u + vu] + 1 u[(xv)′′ − v ′ + v]dx,
1
L∗ v = (xv)′′ − v ′ + v = xv ′′ + 2v ′ − v ′ + 1.
d2 d
L∗ = x 2 + + 1.
dx dx
The adjoint boundary conditions are found from the bilinear concomi-
tant,
2
[xvu′ − (xv)′ u + vu] 1 = 0.
x = 1, xv = 0, v(1) = 0,
x = 2, u′ = −u, −2v − v − 2v ′ + v = 0, v ′ (2) + v(2) = 0.
 15

1.7 Solve the differential system

u′′ + u′ − 2u = x2 , u(0) = 0, u′(1) = 0.

Solution

For this constant coefficient equation, we try

u = eαx ,

to get
1 3
α2 + α − 2 = 0, α = − ± .
2 2
α1 = 1, α2 = −2.

For a particular solution, we assume

up = Cx2 + Dx + E,

which, upon substituting in the differential equation gives

2C + 2Cx + D − 2Cx2 − 2Dx − 2E = x2 .

Matching the coefficients of equal powers of x,

C = −1/2, D = −1/2, E = −3/4.

The general solution is

1 1 3
u = Aex + Be−2x − x2 − x − .
2 2 4
The boundary conditions give
3
u(0) = 0, A+B− = 0,
4
3
u′ (1) = 0, Ae − 2Be−2 − = 0.
2
Then
3 e2 + 1 3 e3 − 2e2
A= , B= .
2 e3 + 2 4 e3 + 2
16 CHAPTER 1. GREEN’S FUNCTIONS

1.8 Solve the differential system

(x2 u′ )′ − n(n + 1)u = 0, u(0) = 0, u(1) = 1.

Solution
This equation is of variable coefficients of the Euler type. We try

u = xα .

The indicial equation is

α(α + 1) − n(n + 1) = 0, α2 + α − n(n + 1) = 0.

r
1 1
α=− ± + n(n + 1),
2 4
α1 = n, α2 = −(n + 1).

So, our solution is

u = Axn + Bx−(n+1) .
(a) n > 0
u(0) = 0, B = 0, u(1) = 1, A = 1.
u = xn .

(b) n = 0
u = A + B/x, u(0) = 0, A = B = 0,
and u(1) = 1 cannot be satisfied. Thus, there is no solution.
(c) n = −1
u = A/x + B, u(0) = 0, A = B = 0,
again, no solution.
(d) n < −1

u(0) = 0, A = 0, u(1) = 1, B = 1.

u = x−(n+1) .
 17

1.9 Convert the following system to one with homogeneous boundary con-
ditions:

(x2 u′ )′ − n(n + 1)u = 0, u(0) = 0, u(1) = 1.

Solution

Let
u = v + Ax + B.
u(0) = v(0) + B = 0, v(0) = 0, B = 0.
u(1) = 1 = v(1) + A, A = 0, V (1) = 0.
u = v + x, [x2 (v ′ + 1)]′ − n(n + 1)[v + x] = 0.
(x2 v ′ )′ − n(n + 1)v = [n(n + 1) − 2]x, v(0) = v(1) = 0.
18 CHAPTER 1. GREEN’S FUNCTIONS

1.10 Obtain the Green’s function for the equation

9
(xu′ )′ − u = f (x), u(0) = 0, u(1) = 0.
x

Using the Green’s function explicitly, find the solution when f (x) = xn .
Check if there are any values for the integer n for which your solution
does not satisfy the boundary conditions.

Solution
Let u = xα .
α2 − 9 = 0, u = x3 , x−3 .

To satisfy u(0) = 0, we choose

u1 = x3 .

To satisfy u(1) = 0, we choose

u2 = x3 − x−3 .

Then (
x3 (ξ 3 − ξ −3), x<ξ
g=C
ξ 3 (x3 − x−3 ), x>ξ

1 1
[[g ′ ]]x=ξ = , C[ξ 3 (3ξ 2 + 3ξ −4) − 3ξ 2(ξ 3 − ξ −3 )] = .
ξ ξ

1
C= . (1.1)
6
(
1 x3 (ξ 3 − ξ −3), x<ξ
g= .
6 ξ 3 (x3 − x−3 ), x>ξ

For f = xn
 19

Z x 1 
1
Z
3 −3 n+3 3 n+3 n−3
u = (x − x )ξ dξ + x (ξ −ξ )dξ
6 0 x
" #
n+4 n+4 n−2 1
1 x ξ ξ
= (x3 − x−3 ) + x3 ( − )
6 n+4 n+4 n−2 x
xn+4 xn−2
 
1 1 n+7 n+1 3 6
= (x − x ) + x (− − + )
6 n+4 (n + 4)(n − 2) n + 4 n − 2
1  n+1
− x3

= x
(n + 4)(n − 2)

We restrict n 6= 2 for this solution to apply.

When n = 2
x3
 
1 1 9 3 6 3
u = (x − x ) + (1 − x ) + x log x
6 6 6
1 3
= x log x.
6

Note that the boundary conditions cannot be satisfied if n ≤ −1.

20 CHAPTER 1. GREEN’S FUNCTIONS

1.11 Find the Green’s function for

u′′ + ω 2 u = f (x), u(0) = 0, u(π) = 0.

Examine the special cases ω = n, n = 0, 1, · · · .

Solution
The Green’s function satisfies

g ′′ + ω 2 g = δ(x − ξ).

g1 = sin ωx, g2 = sin ω(π − x).

(
g1 (x)g2 (ξ), x < ξ
g=C
g2 (x)g1 (ξ), x > ξ
The jump condition gives

C[g2′ (ξ)g1(ξ) − g1′ (ξ)g2(ξ)] = 1.

That is

Cω[− cos ω(π − ξ) sin ωξ − sin ω(π − ξ) cos ωξ] = 1.

1
C=−
ω sin πω
(
1 sin ωx sin ω(ξ − π), x<ξ
g=
ω sin πω sin ωξ sin ω(x − π), x>ξ
When ω = n, g = sin nx satisfies both the boundary conditions and a
regular Green’s function does not exist. We have to look for a general-
ized Green’s function.
 21

1.12 Express the equation

u′′ − 2u′ + u = f (x), u(0) = 0, u(1) = 0,
in the self-adjoint form. Obtain the solution using the Green’s function
when f (x) = ex .
Solution
Comparing with the Sturm-Liouville form,
p′ /p = −2, p = e−2x .
Our equation becomes
(e−2x u′)′ + e−2x u = e−2x f.

Trying a solution of the form u = eαx ,

α2 − 2α + 1 = 0, α = 1, 1.

To find the Green’s function, let

g1 = xex , g2 = (x − 1)ex .
(
x(ξ − 1)ex+ξ , x < ξ
g=C .
ξ(x − 1)ex+ξ , x > ξ
The jump condition gives
C[ξ + (ξ − 1)ξ − (ξ − 1) − ξ(ξ − 1)]e2ξ = e2ξ .
Then C = 1 and
(
x(ξ − 1), x<ξ
g = ex+ξ .
ξ(x − 1), x>ξ
When f = ex ,
Z x Z 1
x+ξ−ξ
u = ξ(x − 1)e dξ + x(ξ − 1)ex+ξ−ξ dξ
0 x
2 2
x (x − 1)
= (x − 1)ex − xex
2 2
x−1 x 2
= e [x − x(x − 1)]
2
1
= x(x − 1)ex .
2
22 CHAPTER 1. GREEN’S FUNCTIONS

1.13 Transform the equation

xu′′ + 2u′ = f (x); u′ (0) = 0, u(1) = 0,
into the self-adjoint form. Find the Green’s function, and express the
solution in terms of f (x). State the restrictions on f (x) for the solution
to exist.
Solution
Comparing with the Sturm-Liouville form,
p′ /p = 2/x, p = x2 .
The self-adjoint form is
(x2 u′ )′ = xf.
The homogeneous equation is
(x2 u′ )′ = 0,
which can be integrated to get
A A
x2 u′ = A, u′ =2
, u = − + B.
x x
′
To satisfy g1 (0) = 0, we choose A = 0 and B = 1. To satisfy g2 (1) = 0,
we choose A = 1 and B = 1.
(
1 − 1ξ , x<ξ
g=C .
1 − x1 , x>ξ
The jump condition gives
C/ξ 2 = 1/ξ 2 , C = 1.
Then (
1 − 1ξ , x<ξ
g= .
1 − x1 , x>ξ
For a given function f ,
Z x  Z 1 
1 1
u= 1− ξf dξ + 1− ξf dξ ′
0 x x xi
Solution exists if ξf is integrable in (0, x) and (1 − ξ)f is integrable in
(x, 1)
 23

1.14 Find the Green’s function for

x2 u′′ − xu′ + u = f (x), u(0) = 0, u(1) = 0.

Solution

With the solution u = xn ,

n(n − 1) − n + 1 = 0, n = 1, 1.

For this repeated index,

u1 = x, u2 = x log x.
These satisfy the left and right boundary conditions, respectively.
(
xξ log ξ, x<ξ
g=C .
ξx log x, x>ξ

The jump condition gives

C[ξ(log ξ + 1) − ξ log ξ] = 1/ξ 2, C = 1/ξ 3 .

(
xξ −2 log ξ, x < ξ
g= .
ξ −2x log x, x > ξ
24 CHAPTER 1. GREEN’S FUNCTIONS

1.15 Using the self-adjoint form of the differential equation

x2 u′′ + 3xu′ − 3u = f (x), u(0) = 0, u(1) = 0,

find the Green’s function and obtain an explicit solution when f (x) = x.

Solution
The self-adjoint form for this equation is

(x3 u′ )′ − 3xu = xf.

Using u = xn , we find

n(n − 1) + 3n − 3 = 0, n = 1, −3.

u1 = x, and u2 = x − x−3
satisfy the required boundary conditions.
(
x(ξ − ξ −3 ), x<ξ
g=C .
ξ(x − x−3 ), x>ξ

The jump condition gives

1
C[ξ(1 + 3ξ −4) − (ξ − ξ −3 )] = ξ −3 , C= .
4
Then, (
1 x(ξ − ξ −3 ), x<ξ
g= .
4 ξ(x − x−3 ), x>ξ

For f = x,
Z x Z 1 
1 −3 3 −3 2
u = (x − x )ξ dξ + x(ξ − ξ )ξ dξ
4 0 x
4
1 − x4
 
1 −3 x
= (x − x ) + x + x log x
4 4 4
1
= x log x.
4
 25

1.16 Solve the equation

x2 u′′ + 3xu′ = x2 , u(1) = 1, u(2) = 2,

using the Green’s function.

Solution
This equation can be written as

(x3 u′ )′ = x3 .

To obtain homogeneous boundary conditions, let

u = v + x,

where the function x satisfies the non-homogeneous boundary condi-

tions.
Then,
u′ = v ′ + 1
and
(x3 v ′ )′ = x3 − 3x2 .

The homogeneous equation can be integrated to get

A A
(x3 v ′ )′ = 0, x3 v ′ = A, v′ = , v=− + B.
x3 2x2
To satisfy v1 (1) = 0 we choose
1
v1 = 1 −
x2
and to satisfy v2 (2) = 0 we choose
4
v2 = 1 − .
x2
With these the Green’s function can be written as

 
 1 − 12 1 − 42 , x < ξ
x ξ
g=C
 
 1 − 2 1 − 12 , x > ξ
4
x ξ
26 CHAPTER 1. GREEN’S FUNCTIONS

Using the jump condition

    
1 8 4 2 1
C 1− 2 3
− 1− 2 3
= 3.
ξ ξ ξ ξ ξ
1
C= .
6
The solution of the non-homogeneous equation is
 Z x 
1 4 1
v = 1− 2 1 − 2 (ξ 3 − 3ξ 2)dξ
6 x 0 ξ
 Z 1  
1 4 3 2
+ 1− 2 1 − 2 (ξ − 3ξ )dξ
x x ξ
x2 5 41
= −x− 2 + .
8 6x 24
and
x2 5 41
u= − 2+ .
8 6x 24
 27

1.17 For the problem

Lu = u′′ + u′ = 0, u(0) = 0, u′ (1) = 0,

obtain the adjoint system. Solve the eigenvalue problems,

Lu = λu, L∗ v = λv,

and show that their eigenfunctions are bi-orthogonal.

Solution
To obtain the adjoint system, we use integration by parts.
Z 1
hv, Lui = v[u′′ + u′ ]dx
0
Z 1
′ ′ 1
= (vu − v u + vu) 0 + u[v ′′ − v ′ ]dx.
0

Then
L∗ v = v ′′ − v ′ , v(0) = 0, v ′ (1) − v(1) = 0.

The two eigenvalue problems are:

u′′ + u′ − λu = 0, v ′′ − v ′ − λv = 0
u = eαx , v = eβx
α2 + α − λq = 0, β 2 − β −q λ=0
α = − 21 ± 14 + λ, β= 1
2
± 1
4
+λ
p
Let iµ = λ + 1/4

u = e−x/2 [A sin µx + B cos µx], v = ex/2 [C sin µx + D cos µx]

Using u(0) = 0, u′ (1) = 0 and v(0) = 0, v ′ (1) − v(1) = 0 we find

B = 0, tan µ = 2µ, D = 0, tan µ = 2µ.

−x/2
ui = e sin µi x, vj = ex/2 sin µj x,
where µi and µj are solutions of tan µ = 2µ.
28 CHAPTER 1. GREEN’S FUNCTIONS

When µi 6= µj ,
Z 1
I = sin µi x sin µj xdx
0
 1
− cos µi x sin µi x
= sin µj x + cos µj x
µi µ2i 0
µ2j 1
Z
+ 2 sin µi x sin µj xdx
µi 0
µ2j
 
cos µi µj
1− 2 I = − sin µj + 2 sin µi cos µj
µi µi µ
 i 
µj tan µj tan µi
= cos µi cos µj − +
µi µj µi
= 0

Thus hui , vj i = 0.
 29

1.18 By solving the nonhomogeneous problem

u′′ = δǫ (x − ξ), u(0) = 0, u(1) = 0,

where 
0, |x| > ǫ,
δǫ (x) = 1
2ǫ
, |x| < ǫ,
in three parts:
a) 0 < x < ξ − ǫ,
b) ξ − ǫ < x < ξ + ǫ, and
c) ξ + ǫ < x < 1,
show that, in the limit ǫ → 0, we recover the Green’s function.

Solution
Let u1 , u2 , and u3 represent the solutions in the three domains.

u′′1 = 0, x < ξ − ǫ.
u1 = A1 x + B1 , u1 (0) = 0, B = 0.
u1 = A1 x
u′′3 = 0, x > ξ + ǫ.
u3 = A3 (1 − x) + B3 , u3 (1) = 0, B3 = 0.
1
u′′2 = , ξ − ǫ < x < ξ + ǫ.
2ǫ
x2
u2 = + A2 x + B2 .
4ǫ
Using continuity of the solutions

u1 (ξ − ǫ) = u2 (ξ − ǫ), u′1 (ξ − ǫ) = u′2 (ξ − ǫ),

ξ−ǫ (ξ − ǫ)2
A1 = A2 + , A1 (ξ − ǫ) = A2 (ξ − ǫ) + B2 + .
2ǫ 4ǫ
u3 (ξ + ǫ) = u2 (ξ + ǫ), u′3 (ξ + ǫ) = u′2 (ξ + ǫ),
(ξ + ǫ) (ξ + ǫ)2
−A3 = A2 + , A3 (1 − ξ − ǫ) = A2 (ξ + ǫ) + B2 + .
2ǫ 4ǫ
30 CHAPTER 1. GREEN’S FUNCTIONS

(ξ − ǫ)2
B2 = ,
4ǫ
(ξ − ǫ)2 (ξ + ǫ)2 (ξ + ǫ)2
A2 (ξ + ǫ) + + = −A2 (1 − ξ − ǫ) − (1 − ξ − ǫ),
4ǫ 4ǫ 2ǫ
ξ+ǫ (ξ − ǫ)2 + (ξ + ǫ)2
A2 = − (1 − ξ − ǫ) − ,
2ǫ 4ǫ
ξ+ǫ ξ − ǫ (ξ − ǫ)2 + (ξ + ǫ)2
A1 = − (1 − ξ − ǫ) + − ,
2ǫ 2ǫ 4ǫ
ξ+ǫ ξ + ǫ (ξ − ǫ)2 + (ξ + ǫ)2
A3 = − (1 − ξ − ǫ) − + .
2ǫ 2ǫ 4ǫ
As ǫ → 0, u2 (ξ) can be evaluated as
 2 
x
u2 = lim + A2 x + B2
ǫ→0 4ǫ
 2
ξ2 2ξ 3 ξ 2

ξ
= lim + − (1 − ξ) − + + ξ(ξ − 1)
ǫ→0 4ǫ 2ǫ 4ǫ 4ǫ
= ξ(ξ − 1),

where ǫ2 terms have been neglected from the outset.

A1 = (ξ − 1), A3 = −ξ.
u1 = x(ξ − 1), u3 = ξ(x − 1).
Finally, we have (
x(ξ − 1), x<ξ
g= .
ξ(x − 1), x>ξ
 31

1.19 Expanding 
x(ξ − 1), x < ξ,
g(x, ξ) =
ξ(x − 1), x > ξ,
√
in terms of un = 2 sin nπx as a Fourier series, show that
X un (x)un (ξ)
g(x, ξ) = , λn = −π 2 n2 .
n=1
λn

Solution

Let
∞
X Z 1
g(x, ξ) = An sin nπx, An = 2 g(x, ξ) sin nπxdx.
n=1 0

 Z ξ Z 1 
An = 2 (ξ − 1) x sin xdx + ξ (x − 1) sin nπxdx
0 ξ
(  ξ
− cos nπx − sin nπx
= 2 (ξ − 1) x −
nπ (nπ)2 0
  1)
− cos nπx − sin nπx
+ ξ (x − 1) −
nπ (nπ)2 ξ
  
cos nπξ sin nπξ
= 2 (ξ − 1) −ξ +
nπ (nπ)2
 
cos nπξ sin nπξ
+ ξ (ξ − 1) −
nπ (nπ)2
sin nπξ
= −2 2 2 .
nπ
Then
∞ ∞
X 2 X 1
g(x, ξ) = − sin nπx sin nπξ = un (x)un (ξ).
n=1
n2 π 2 n=1
λ n
32 CHAPTER 1. GREEN’S FUNCTIONS

1.20 Obtain the Green’s function for

∂u ∂u
κ∇2 u(x1 , x2 ) = v1 + v2 ,
∂x1 ∂x2
where v1 and v2 are constants, by transforming the dependent variable.

Solution

∂ 2 u v1 ∂ 2 u v2 ∂u f
2
− + 2− =
∂x1 κ ∂x2 κ ∂x2 κ
Let
u = eλ1 x1 +λ2 x2 φ(x1 , x2 ).
Then
∂u ∂φ
= eλ1 x1 +λ2 x2 [λ1 φ + ],
∂x1 ∂x1
∂2u ∂2φ
 
λ1 x1 +λ2 x2 2 ∂φ
=e λ1 φ + 2λ1 + .
∂x21 ∂x1 ∂x21
The given equation becomes

∂2φ ∂2φ
 
2 2 λ1 v1 λ2 v2
+ + λ1 + λ2 − − φ
∂x21 ∂x22 κ κ
 v1  ∂φ  v2  ∂φ f e−(λ1 x1 +λ2 x2 )
+ 2λ1 − + 2λ2 − = .
κ ∂x1 κ ∂x2 κ
We choose
v1 v2
λ1 = , λ2 = ,
2κ 2κ
to get
∇2 φ − k 2 φ = F,
where
v12 + v22 f e−(λ1 x1 +λ2 x2 )
k2 = , F = .
4κ2 κ
The solutions of the homogeneous equation are the modified Bessel
functions, K0 (kr) and I0 (kr), with
q
r = x21 + x22 .
 33

For a bounded solution for r → ∞, we choose

φ = AK0 (kr).

As kr → 0, K0 behaves as − log r and we choose

1
A=− ,
2π
by comparison to the two dimensional Laplace operator.
For the original equation with f on the right hand side,
1 −[v1 (x1 −ξ1 )+v2 (x2 −ξ2 )]/(2κ)
g=− e K0 (kρ)
2π
p
where ρ = (x1 − ξ1 )2 +)x2 − ξ2 )2 .
34 CHAPTER 1. GREEN’S FUNCTIONS

1.21 The anisotropic Laplace equation in a two dimensional infinite domain

is given by
∂2u ∂2u
k12 2 + k22 2 = 0.
∂x1 ∂x2
Find the Green’s function for this equation.

SolutionThe Green’s function satisfies

∂2u 2
2∂ u
k12 + k 2 = δ(x1 , x2 ),
∂x21 ∂x22

if the source is at the origin. Let

y1 = x1 /k1 , y2 = x2 /k2 .

Then, u satisfies the Laplace equation

∂2u ∂2u
+ = 0,
∂y12 ∂y22

away from the origin.

1
q
g(y1, y2 ) = log y12 + y22.
2π
In terms of the original variables, with the source being at (ξ1 , ξ2)
1/2
(x1 − ξ1 )2 (x2 − ξ2 )2

1
g(x1 , x2 , ξ1 ξ2 ) = log + .
2π k12 k12

Note: When g is used in an integral, the original area element dy1 dY2 =
dx1 dx2 /(k1 k2 ).
 35

1.22 In a semi-infinite medium, −∞ < x < ∞, 0 < y < ∞, the pressure

fluctuations satisfy the wave equation

1 ∂2p
2
∇ p= 2 2,
c ∂t

where c is the wave speed and t is time. If the boundary, y = 0, is

subjected to a pressure p = P0 δ(x)h(t), show that

y t
p(x, y, t) = A 2
√ h(t − kr),
r t − k2r2
2

p
where r = x2 + y 2 and k = 1/c, is a solution of the wave equa-
tion. Evaluate the constant A using equilibrium of the medium in the
neighborhood of the applied load. Hint: Use polar coordinates.

Solution

In polar coordinates, the wave equation becomes

∂ 2 p 1 ∂p 1 ∂2p 2
2∂ p
+ + = k .
∂r 2 r ∂r r 2 ∂θ2 ∂t2

Away from r = 0, the given solution,

t sin θ √
p=A , β= t2 − k 2 r 2 ,
rβ
36 CHAPTER 1. GREEN’S FUNCTIONS

has the derivatives

t2
 
∂p sin θ 1
= A −
∂t r β β3
sin θ k 2 r 2
= −A
r β3
∂2p 3k 4 rt
k 2 2 = A sin θ 5
∂t β
2
1∂ p sin θ t
2 2
= −A 3
r ∂θ r β
 2 
∂p k t t
= A sin θ − 2
∂r β3 βr
t sin θ 
= A 2 3 k2r2 − β 2
r β
t sin θ 
= A 2 3 2k 2 r 2 − t2
r β
t2
 
t sin θ 2
= A 3 2k − 2
β r
2
t2 3k 2 r 2t2
  
∂ p 2
= At sin θ 2k − 2 + 3 3
∂r 2 r β5 r β
t sin θ
= A 3 5 6k 4 r 4 − 3k 2 t2 r 2 + 2t2 (t2 − k 2 r 2 )

r β
t sin θ 
= A 3 5 6k 4 r 4 − 5k 2 t2 r 2 + 2t4
r β
Now, the left hand side of the wave equation becomes
 
1 1 1
= At sin θ 3 5
(6k 4 r 4 − 5k 2 t2 r 2 + 2t4 ) + 3 3 (2k 2 r 2 − t2 ) − 3
r β r β r β
t sin θ  4 4
= A 6k r − 5k 2 t2 r 2 + 2t4 + 2k 2 r 2 (t2 − k 2 r 2 )
r3β 5
−t2 (t2 − k 2 r 2 ) − (t4 − 2k 2 r 2 t2 + k 4 r 4 )
t sin θ
= 3A 5 k 4 r.
β

This is identical to the right hand side quantity.

 37

To find the value of A, we draw a semi-circle under the concentrated

load P0 and balance the vertical force. We let r → 0.
Z π/2
P0 = 2 p sin θrdθ
0
Z π/2
= 2A sin2 θdθ = Aπ/2.
0

Then
2P0
A=
π
38 CHAPTER 1. GREEN’S FUNCTIONS

1.23 For the two dimensional wave equation

1 ∂2u
∇2 u = ,
c2 ∂t2
steady state solutions are obtained using u(x, y, t) = v(x, y)e−iΩt where
v satisfies the Helmholtz equation,

Lv = 0, L = ∇2 + k 2 , k = Ω/c.
(1) (2)
Show that the Hankel
p functions H0 (kr) and H0 (kr) satisfy Lg =
δ(x, y) when r = x2 + y 2 6= 0. Examine their asymptotic forms for
kr << 1 and for kr >> 1, using the results shown in Abramowitz
and Stegun (1965) and select multiplication constants A and B to have
(1) (2)
g = AH0 or g = BH0 by comparing the asymptotic form with the
(1)
Green’s function for the Laplace operator (k → 0). Show that H0
(2)
corresponds to an outgoing wave and H0 to an incoming wave.

Solution
The Helmholtz equation,

(∇2 + k 2 )v = δ(x − ξ, y − η)

with a source at the origin, has the polar form

(rv ′ )′ + k 2 rv = δ(r),

where we assume axisymmetry.

Solutions are the Hankel functions as
(1) (2)
v = H0 (kr), H0 (kr).

When kr << 1,

(1) 2
H0 (kr) = J0 + iY0 ∼ log r,
π

(2) 2
H0 (kr) = J0 − iY0 ∼ − log r,
π
 39

Comparing with the 2D Laplace operator

−i (1) i (2)
g= H (kr), or g = H0 (kr).
4 0 4

When kr >> 1, r
(1) 2 i(kr−π/4)
H0 (kr) ∼ e ,
πkr
r
(2) 2 −i(kr−π/4)
H0 (kr) ∼ e ,
πkr
Combining these with exp(−iΩt), we find terms of the form (r − ct) for
(1) (2)
H0 (kr) and (r + ct) for H0 (kr), with former being an outgoing wave
and the latter an incoming wave.
40 CHAPTER 1. GREEN’S FUNCTIONS

1.24 Show that

e±ikr
g=− ,
4πr
satisfies the Helmholtz equation

∇2 g + k 2 g = δ(x − ξ),

in a 3D infinite domain, with r = |x − ξ|. Assuming the Helmholtz

equation is obtained from the wave equation by separating the time
dependence using a factor e−iΩt , show that (±) signs correspond to
outgoing and incoming waves, respectively.

Solution
In spherical coordinates the spherically symmetric Green’s function sat-
isfies
(r 2 g ′)′ + k 2 r 2 g = 0,
away from the source point. Substituting the given solution and ne-
glecting the factor −1/(4π),
   ′
2 1 ik ±ikr
r − 2± e + k 2 re±ikr
r r
= {(−1 ± ikr)′ + (−1 ± ikr)(±ik)} + k 2 r
= ±ik ∓ ik − k 2 r + k 2 r
= 0.

Integrating the equation for the Green’s function over a sphere of radius
r << 1,

4πr 2 g ′ |r = 1
 
1 ik
lim 4πr 2 ∓ e±ikr = 1
r→0 4πr 2 4πr

For the (+) sign, we find the exponential term (r − ct) representing an
outgoing wave and for the (−) sign, (r + ct) representing an incoming
wave. Here c = Ω/k.
 41

1.25 Consider a volume V enclosed by the surface S in 3D. From

∇2 u + k 2 u = f, ∇2 g + k 2 g = δ(x − ξ),

obtain the solution

Z  
∂u ∂g
Z
u(ξ) = g(ξ, x)f (x)dV − g −u dS.
V S ∂n ∂n

Solution
Using the inner products of the first equation with g and the second
equation with u and subtracting

hg, Lui − hu, Lgi = hg, f i − hu, δi.

For a self-adjoint operator, the left hand side can be integrated by parts
(Gauss theorem) to get
Z  
∂u ∂g
Z
u(ξ) = g(ξ, x)f (x)dV − g −u dS.
V S ∂n ∂n
42 CHAPTER 1. GREEN’S FUNCTIONS

1.26 In the previous problem, assuming V is a sphere of radius R centered

at x = 0 and x is a point on its surface, and g is the Green’s function
for the 3D infinite space, show that
Z  ikr
∂ eikr
 
1 e ∂u
u(ξ) = −u dS,
4π S r ∂n ∂n r

if f = 0. If the included angle between x and x − ξ is ψ, show that

dr/dn = cos ψ. Also show that as R → ∞, r → R and R2 (1 − cos ψ) is
finite and for the surface integral to exist
 
∂u
r − iku → 0, and u → 0.
∂r

These are known as the Sommerfeld radiation conditions.

Solution
For an infinite 3D space the Green’s function for the Helmholtz operator
is
1 ikr
g=− e .
4πr
When there is no forcing function
Z  ikr
∂ eikr
 
1 e ∂u
u(ξ) = −u dS
4π S r ∂n ∂n r

From Fig. 1.1,

dr
= cos ψ
dn
Then
Z  
1 1 ∂u ik u
u(ξ) = − u cos ψ + 2 cos ψ eikr dS
4π S r ∂n r r
Z  
1 1 ∂u iku ik u
= − + (1 − cos ψ) + 2 cos ψ eikr dS
4π S r ∂n r r r

From the triangle in Fig. 1.1, with R = |x|, we find

ξ 2 = R2 + r 2 − 2Rr cos ψ.
 43

As R → ∞ and r → ∞, dS ∼ 4πR2 and

2R2 (1 − cos ψ) = ξ 2

is finite and the condition for the existence of the integral is

 
∂u
r − iku → 0, and u → 0.
∂n

dndr

x ψ
r

O ξ

Figure 1.1: A source inside a spherical domain

44 CHAPTER 1. GREEN’S FUNCTIONS

1.27 By reconsidering the previous problem, show that the surface integral
exists if the less restricted condition
 
∂u u
r − iku + → 0.
∂r r

is satisfied.

Solution
In the previous problem, if we set
dr
= cos ψ,
dn
Z  
1 1 ∂u iku u ikr
u(ξ) = − + 2 e cos ψdS
4π S r ∂r r r
2
As dS ∼ 4πr and R and r go to infinity, the integral exists if
 
∂u
r − iku + u → 0.
∂r
 45

1.28 A wedge-shaped 2D domain has the boundaries: y = 0 and y = x.

Obtain the Green’s function for the Poisson’s equation, ∇2 u = f (x, y),
for this domain if u(x, 0) = 0 and u(x, x) = 0. Use the method of
images for your solution.

Solution

(−η, ξ) (η, ξ)

(−ξ, η) (ξ, η)

(−ξ, −η) (ξ, −η)

(−η, −ξ) (η, −ξ)

Figure 1.2: Sources and sinks for the wedge bounded by y = 0 and y = x

As shown in Fig. 1.2, there are 4 sources and 4 sinks to satisfy the
boundary conditions. Using the notation
p p
r1 = (x − ξ)2 + (y − η)2 , r2 = (x + η)2 + (y − ξ)2 ,
p p
r3 = (x + ξ)2 + (y + η)2 , r4 = (x − η)2 + (y + ξ)2 ,
p p
r5 = (x − ξ)2 + (y + η)2 , r6 = (x + η)2 + (y + ξ)2 ,
p p
r7 = (x + ξ)2 + (y − η)2 , r8 = (x − η)2 + (y − ξ)2 ,
the Green’s function is
1 r1 r2 r3 r4
g= log .
2π r5 r6 r7 r8
46 CHAPTER 1. GREEN’S FUNCTIONS

1.29 A unit circle is mapped into a cardioid by the transformation

w = c(1 + z)2 ,
where c is a real number. Obtain the Green’s function, g, for a domain
in the shape of a cardioid with g = 0 on the boundary.
Solution
We have
w = c(1 + z)2 .
For a unit circle,
1 r 2 + ρ2 − 2rρ cos(θ − φ)
g= log 2 2 ,
4π r ρ + 1 − 2rρ cos(θ − φ)
where r = |z|, and ρ = |ζ|, where ζ is the source location.
Let ω be the image of the source point ζ. Then
 w 1/2  ω 1/2
z= − 1, ζ = − 1.
c c
Let
w/c = peiα , ω/c = qeiβ ,
where p and q are real.
√ √
z= peiα/2 , ζ= qeiβ/2 .
From the Green’s function for the unit circle,
1 z−ζ
g= log ,
2π 1 − z ζ̄
the mapping gives
√ iα/2 √ iβ/2
1 pe − qe
g = log √ i(α−β)/2
2π 1 − pqe
√ √
1 p − qeiψ α−β
= log √ iψ , ψ =
2π 1 − pqe 2
√ √ iψ √ √ −iψ
1 ( p − qe )( p − qe )
= log √ √
4π (1 − pqeiψ )(1 − pqe−iψ )
√
1 p + q − 2 pq cos ψ
= log √ .
4π 1 + pq − 2 pq cos ψ
 47

1.30 The steady-state temperature in a semi-infinite plate satisfies

∇2 u(x, y) = 0, −∞ < x < ∞, 0 < y < ∞,

On the boundary, y = 0, the temperature is given as u = f (x). Obtain

the temperature distribution in the domain using the Green’s function.

Solution
From
∇2 u = 0, ∇2 g = δ(x − ξ),
Z
u(ξ) = [u∇2 g − g∇2u]dA.
A
Using a semi-circle of large radius R, the area integral can be trans-
formed into a line integral by using Gauss theorem, to get
I  
∂g ∂u
u(ξ) = u −g ds.
C ∂n ∂n

The Green’s function for this proble is

1 
g= log[(x − ξ)2 + (y − η)2 ] − log[(x − ξ)2 + (y + η)2 ] .
4π
On the semi-circle, u and ∂u/∂n vanish, and the contour integral be-
comes Z ∞ 
∂g ∂u
u(ξ) = f −g dx.
−∞ ∂n ∂n
On the x-axis, u = 0 and
∂g ∂g 1 −2η − 2η
= − =−
∂n ∂y 4π (x − ξ)2 + η 2
1 η
=
π (x − ξ)2 + η 2

The solution is ∞
y f (ξ)dξ
Z
u(x, y) = .
π −∞ (ξ − x)2 + η 2
48 CHAPTER 1. GREEN’S FUNCTIONS

1.31 Show by substitution that the solutions of

h2
[(1 − x2 )u′ ]′ − u = 0,
1 − x2
are  ±h/2
1−x
u= .
1+x
Solution

Let

u = (1 − x)α (1 + x)−α ,
u′ = α[(1 − x)α−1 (1 + x)−α + (1 − x)α (1 + x)−α−1 ],
(1 − x2 )u′ = −α[(1 − x)α (1 + x)−α+1 + (1 − x)α+1 (1 + x)−α ],
((1 − x2 )u′ )′ = −α[−α(1 − x)α−1 (1 + x)−α+1 − (α − 1)(1 − x)α (1 + x)−α
−(α + 1)(1 − x)α (1 + x)−α − α(1 − x)α+1 (1 + x)−α−1 ]
 
α −α 1+x 1−x
= α(1 − x) (1 + x) α + α − +α + 1 + α
1−x 1+x
2 α −α 2
= α (1 − x) (1 + x) (1 − x )
4α2 u
= .
1 − x2
Comparing with the given differential equation

α = ±h/2.
 49

1.32 Obtain the Green’s function for the above operator when h 6= 0 if the
boundary conditions are
u(−1) = finite, u(1) = finite.

Solution
For u(−1) to be finite, we take
 h/2
1+x
u1 = ,
1−x
and for u(1) to be finite, we take
 h/2
1−x
u2 = .
1+x
Then
(
(1 + x)h/2 (1 − x)−h/2 (1 − ξ)h/2 (1 + ξ)−h/2 , x<ξ
g=C .
(1 − x)h/2 (1 + x)−h/2 (1 + ξ)h/2 (1 − ξ)−h/2 , x>ξ

The jump condition

[[g ′]] = 1/p,
gives
( h/2
Ch 1+ξ
[−(1 − ξ)h/2−1 (1 + ξ)−h/2 − (1 − ξ)h/2 (1 + ξ)−h/2−1 ]
2 1−ξ
 h/2 )
1−ξ 1
− [(1 + ξ)h/2−1 (1 − ξ)−h/2 + (1 + ξ)h/2 (1 − ξ)−h/2−1 ] =
1+ξ 1 − ξ2
 
1 1 1 1 2 1
C − − − − = 2
1−ξ 1+ξ 1+ξ 1−ξ h 1 − ξ2
C[−2(1 + ξ) − 2(1 − ξ)] = 2/h
1
C=− .
2h
(
1 (1 + x)h/2 (1 − x)−h/2 (1 − ξ)h/2(1 + ξ)−h/2 , x<ξ
g=− .
2h (1 − x)h/2 (1 + x)−h/2 (1 + ξ)h/2 (1 − ξ)−h/2 , x>ξ
50 CHAPTER 1. GREEN’S FUNCTIONS

1.33 If h = 0 in the preceding problem, show that

1
U(x) = √
2

is a normalized solution of the above homogeneous equation, which

satisfies both the boundary conditions. Obtain the generalized Green’s
function for this case.

Solution
When h = 0, the equation is

((1 − x2 )u′)′ = 0.

Integrating
 
′ A A 1 1
u = = + ,
1 − x2 2 1−x 1+x

A
u= [log |1 − x| + log |1 + x|] + B.
2
For u(−1) and u(1) to be finite, u = B. Normalizing this in (−1, 1),
we get
1
U=√ .
2

The generalized Green’s function satisfies

[(1 − x2 )g ′]′ = δ(x − ξ) − U(x)U(ξ),

which can be written as

1
[(1 − x2 )g1′ ]′ = − , x < ξ,
2
1
[(1 − x2 )g2′ ]′ = − , x > ξ,
2
Integrating these, we get
 51

x
(1 − x2 )g1′ = − + C1 ,
2
2 ′ x
(1 − x )g2 = − + C2 ,
2  
′ 1 x C1 1 1
g1 = − + + ,
2 1 − x2 2 1−x 1−x
 
′ 1 x C2 1 1
g2 = − + + ,
2 1 − x2 2 1−x 1−x
1 C1
g1 = log(1 − x2 ) + [log(1 − x) + log(1 + x)] + C3 ,
4 2
1 C2
g2 = log(1 − x2 ) + [log(1 − x) + log(1 + x)] + C4 ,
4 2
For g1 (−1) to be finite, C1 = −1/2 and for g2 (1) to be finite, C2 = −1/2.
Then
.
1
g1 = log(1 − x) + C3 ,
2
1
g2 = log(1 + x) + C4 ,
2
For continuity at x = ξ, we take

1
C3 = D + log(1 + ξ),
2
1
C4 = D + log(1 − ξ),
2
where D is a new constant. It is found using hU, gi = 0.

ξ 1
1+ξ 1 1−ξ 1
Z Z
2D+ log(1+ξ)+ log(1−x)dx+ log(1−ξ)+ log(1+x)dx = 0.
2 2 −1 2 2 ξ

1+ξ 1−ξ
2D + log(1 + ξ) + log(1 − ξ)
2 2
1 ξ
1 ξ xdx
Z
+ x log(1 − x) +
2 −1 2 −1 1 − x
1 1
1 1 xdx
Z
+ x log(1 + x) − = 0.
2 ξ 2 ξ 1+x
52 CHAPTER 1. GREEN’S FUNCTIONS

1+ξ 1−ξ
2D + log(1 + ξ) + log(1 − ξ)
2 2
1 1 1 1
+ ξ log(1 − ξ) − (−1) log 2 + log 2 − ξ log(1 + ξ)
2 2 2 2
ξ 1
1 1 1 1
− (1 + ξ) − (1 − ξ) − log(1 − x) + log(1 + x) = 0.
2 2 2 −1 2 ξ

1+ξ 1−ξ
2D + log(1 + ξ) + log(1 − ξ)
2 2
1 1
+ ξ log(1 − ξ) + log 2 − ξ log(1 + ξ)
2 2
1 1
−1 − log(1 − ξ) + log 2 − log(1 + ξ) = 0.
2 2
1
D = − log 2
2
(
1 1 log[(1 − x)(1 + ξ)], x < ξ,
g = − log 2 +
2 2 log[(1 + x)(1 − ξ)], x > ξ.
 53

1.34 Find the generalized Green’s function for

u′′ = f (x), u(1) = u(−1), u′ (1) = u′ (−1).

Solution

u′′ = 0, u = Ax + B.
u′ (1) = u′ (−1), A = A.
u(1) = u(−1), A + B = −A + B, A = 0.
Then u = B and normalizing this in (−1, 1), we find
1
U=√ .
2
The generalized Green’s function satisfies
1
g ′′ = δ(x − ξ) − .
2
g1′′ = − 12 , g2′′ = − 21 ,
1
′
g 1 = − 2 x + C1 , g2′ = − 12 x + C2 ,
g1 = − 14 x2 + C1 x + D1 , g2 = − 14 x2 + C2 x + D2 .

The condition, g1 (−1) = g2 (1), gives

1 1
+ C1 = − + C2 .
2 2
This satisfied by using
1 1
C1 = C − , C2 = C + .
2 2
The condition, g1 (−1) = g2 (1), gives
1 1
D1 + − C = D − 2 + + C.
2 2
We choose
D1 = D + C, D2 = D − C
54 CHAPTER 1. GREEN’S FUNCTIONS

to satisfy this. So far

g1 = − 41 x2 + Cx − 12 x + D + C, g2 = − 14 x2 + Cx + 12 x + D − C,
g1 = − 14 x2 + C(x + 1) − 12 x + D, g2 = − 14 x2 + C(x − 1) + 12 x + D.

The continuity condition, g1 (ξ) = g2 (ξ), gives

ξ ξ ξ
C(ξ + 1) − = C(ξ − 1) + , C= .
2 2 2
Then
1 1 1
g1 = − x2 + ξ(x + 1) − x + D,
4 2 2
1 2 1 1
g2 = − x + ξ(x − 1) + x + D.
4 2 2
The remaining constant D is found from hU, gi = 0.
1 ξ
x2 1 1
Z Z
2D − dx + (ξ − 1) xdx + ξ(ξ + 1)
−1 4 2 −1 2
1
1 1
Z
(ξ + 1) xdx − ξ(1 − ξ) = 0.
2 ξ 2
1 1
2D − + ξ 2 + (ξ − 1)(ξ + 1)[ξ − 1 − (ξ + 1)] = 0.
6 4
1 1
D = − − ξ 2.
6 4
(
1 x2 + ξ 2 1 ξ − x, x < ξ,
g=− − xξ + .
6 4 2 x − ξ, x > ξ.
 55

1.35 To solve the self-adjoint problem

Lu = f (x),

with homogeneous boundary conditions at x = a and x = b, we use a

generalized Green’s function g which satisfies

Lg = δ(x − ξ) − U(x)U(ξ),

with homogeneous boundary conditions, where U is the normalized

solution of the homogeneous equation satisfying the same boundary
conditions. The solution is written as
Z b
u(x) = AU(x) + g(x, ξ)f (ξ) dξ.
a

By operating on this equation using L, show that u is the required

solution.

Solution

Z b
Lu = Lgf dξ + ALU
a
Z b
= [δ(x − ξ) − U(x)U(ξ)]f dξ + 0
a
= f (x),

where we have used

LU = 0, and the existence condition hU, f i = 0.

56 CHAPTER 1. GREEN’S FUNCTIONS

1.36 For the equation

Lu = u′′ + u′ = 0, u′(0) = u′ (1) = 0,

obtain the adjoint equation and its boundary conditions. Obtain nor-
malized homogeneous solutions of LU = 0 and L∗ U ∗ = 0. Construct
generalized Green’s functions g and g ∗ .

Solution

Lu = u′′ + u′ = 0, u′(0) = u′ (1) = 0.

The adjoint problem is found by integrating by parts.
Z 1
hv, Lui = v[u′′ + u′ ]dx
0
Z 1
′ ′ 1
= (vu − v u + vu] 0 + u[v ′′ − v ′ ]dx.
0

Thus
L ∗ v = v ′′ − v ′ .
To make the bilinear concomitant 0 at the boundary points,

v ′ (0) − v(0) = 0, v ′ (1) − v(1) = 0.

Next, we find the solutions of the homogeneous equations.

u′′ + u′ = 0, u′ + u = A, u = Be−x + A.

The solution U = 1 satisfies both the boundary conditions.

v ′′ − v ′ = 0, v ′ − v = C, v = Dex − C.

v ′ (0) − v(0) = 0, C = 0.
v ′ (1) − v(1) = 0, v = Dex .
The function U ∗ is the normalized form of v, with normalization con-
dition Z 1
ex 1
UU ∗ = 1, D = , Let β = .
0 e−1 e−1
 57

Then, g and g∗ satisfy

Lg = δ(x − ξ) − U ∗ (ξ)U(x), L∗ g ∗ = δ(x − ξ) − U(ξ)U ∗ (x).

Using the subscripts, 1 and 2, for x < ξ and x > ξ,

g1′′ + g1′ = −βeξ , g2′′ + g2′ = −βeξ ,

g1′ + g1 = −βeξ x + C1 , g2′ + g2 = −βeξ x + C2 ,
g1 = g1h + g1p , g2 = g2h + g2p ,
g1h = B1 e−x + C1 , g1p = βeξ (1 − x), g2h = B2 e−x + C2 , g2p = βeξ (1 − x),
g1 = B1 e−x + βeξ (1 − x) + C1 , g2 = B2 e−x + βeξ (1 − x) + C2 .

The condition g1′ (0) = 0 gives

−B1 − βeξ = 0, B1 = −βeξ .

The condition g2 (1) = 0 gives

B2 e−1 − βeξ = 0, B2 = −βeξ+1 .

Then

g1 = βeξ [−e−x − x + 1] + C1 , g2 = βeξ [−e1−x − x + 1] + C2 .

From g1 (ξ) = g2 (ξ),

C1 = D + βeξ [−e1−ξ − ξ + 1], C2 = D + βeξ [−e−ξ − ξ + 1]

and with these

g1 = D+βeξ [−e−x −e1−ξ −x+1−ξ+1], g2 = D+βeξ [−e1−x −e−ξ −x+1−ξ+1].

The remaining constant D is found form hg, U ∗i = 0.

Z 1
Dβ ex dx + β 2 eξ [(2 − ξ − e1−ξ (eξ − 1) − ξ − ξeξ + eξ − 1]
0

β 2 eξ [(2 − ξ − e−ξ )(e − eξ ) − e(1 − ξ) − e + ξeξ + e − eξ ] = 0.

D + β 2 eξ [(2 − ξ)(e − 1) − e + e1−ξ − e1−ξ + 1
−ξ − ξeξ + eξ − 1 − e + eξ + ξeξ − eξ ] = 0.
 D + β 2 eξ [2e − 2 − eξ + ξ − e − ξ + eξ − e] = 0.
D = 2β 2eξ .
(
e, x < ξ,
g = βeξ [2β + 2 − x − ξ − eξ−x ] − .
1, x > ξ,
From symmetry
(
e, x > ξ,
g ∗ = βex [2β + 2 − ξ − x − ex−ξ ] − .
1, x < ξ,
Chapter 2

INTEGRAL EQUATIONS

2.1 Using differentiation, convert the integral equation

1
1 x2
Z
u(x) = |x − ξ|u(ξ)dξ +
2 0 2

into a differential equation. Obtain the needed boundary conditions

and solve the differential equation.

Solution

1 1 x2
Z
u(x) = |x − ξ|u(ξ)dξ +
2 0 2
Z x Z 1 
1 2
= (x − ξ)u(ξ)dξ + (ξ − x)u(ξ)dξ + x ,
2 0 x
Z x Z 1 
1
u′ = udξ − udξ + 2x
2 0 x
u′′ = u + 1.

Boundary conditions

1 1
1 1
Z Z
′
u(0) = ξudξ, u (0) = − udξ.
2 0 2 0

59
60 CHAPTER 2. INTEGRAL EQUATIONS

General solution is

u = −1 + A cosh(x) + B sinh(x),
u′ = A sinh(x) + B cosh(x),
Z 1
udξ = A sinh(1) + B(cosh(1) − 1)
0
1
1
Z
1 1
ξudξ = − + {ξ[A sinh(ξ) + B cosh(ξ)]} |0 − [A cosh(ξ) + B sinh(ξ)]|0 ,
0 2
1
= − + A[sinh(1) − cosh(1) + 1] + B[cosh(1) − sinh(1)].
2
Using these in the boundary conditions
1
−2 + 2A = − + A[sinh(1) − cosh(1) + 1] + B[cosh(1) − sinh(1)],
2
2B = 1 − A sinh(1) − B(cosh 1 − 1).

Solving for B
1 sinh(1)
B= −A .
1 + cosh 1 1 + cosh(1)
Substituting for A,

3 sinh(1) − cosh(1) − 1
A= +A . (2.1)
2 1 + cosh(1)

Then
3 1 + cosh(1) 3 sinh(1) 1
A= , B=− + .
2 2 + 2 cosh(1) − sinh(1) 2 2 + 2 cosh(1) − sinh(1) 1 + cosh(1)
(2.2)
Finally

sinh x 3 cosh(x) + cosh(1 − x)

u(x) = −1 + + .
1 + cosh(1) 2 2 + 2 cosh(1) − sinh(1)
 61

2.2 Convert Z 1
u(x) = λ e−|x−ξ|u(ξ)dξ
−1

into a differential equation. Obtain the required number of boundary

conditions.

Solution

Z 1
u(x) = λ e−|x−ξ| u(ξ)dξ,
−1
Z x Z 1 
u′ = λ ξ−x
e udξ + x−ξ
e udξ
−1 x
 Z x Z 1 
u′′ = λ −u + ξ−x
e udξ − x−ξ
e udξ
−1 x
′′
u = (1 − 2λ)u,
Z 1 Z 1
u(1) = λ eξ−1 udξ, u(−1) = λ e1−ξ udξ.
−1 −1
62 CHAPTER 2. INTEGRAL EQUATIONS

2.3 Solve the integral equation

Z 1
x
u(x) = e + λ e(x−ξ) u(ξ)dξ,
−1

and discuss the conditions on λ for a unique solution.

Solution
Z 1
x
u(x) = e + λ e(x−ξ) u(ξ)dξ,
−1

Let Z 1
C= e−ξ u(ξ)dξ,
−1

Then

u = ex + λCex = (1 + λC)ex ,
2
C = (1 + λC)2, C =
1 − 2λ
ex
u =
1 − 2λ
For a unique solution λ 6= 1/2.
 63

2.4 Solve Z 1
u(x) = x + (1 − xξ)u(ξ)dξ.
0

Solution
R1 R1
Let C1 = 0
udξ and C2 = 0
ξudξ

u = x + C1 − C2 x,
1 1 1 1 1
C1 = + C1 − C2 , C2 = + C1 − C2 ,
2 2 3 2 3
C1 = 2, C2 = 1,
u(x) = 2.
64 CHAPTER 2. INTEGRAL EQUATIONS

2.5 Solve Z π
u(x) = sin x + cos(x − ξ)u(ξ)dξ.
0

Solution

Z π
u(x) = sin x + [cos x cos ξ + sin x sin ξ]u(ξ)dξ.
0

Let C1 = hu, cosi, and C2 = hu, sini.

u = sin x + C1 cos x + C2 sin x,

Z π π
π
Z
C1 = (1 + C2 ) sin x cos xdx + C1 cos2 xdx = C1
0 0 2
C1 = 0
π
C2 = (1 + C2 ) ,
2
π
C2 = ,
2−π
2
u(x) = sin x
2−π
 65

2.6 Obtain the eigenvalues and eigenfunctions of the equation

Z 2π
u(x) = λ cos(x − ξ)u(ξ)dξ.
0

Solution

Z 2π
u = λ [cos x cos ξ + sin x sin ξ]u(ξ)dξ,
0

Let C1 = hu, cosi, and C2 = hu, sini.

u = λ[C1 cos x + C2 sin x]

C1 = λπC1 , C2 = λπC2 ,
When λ = 1/π, we may obtain two solutions corresponding to C1 = 1, C2 = 0
and C1 = 0, C2 = 1

1 1
u1 = √ cos x, u2 = √ sin x
π π

are the normalized eigenfunctions for the eigenvalue λ = 1/π.

66 CHAPTER 2. INTEGRAL EQUATIONS

2.7 Find the eigenvalues and eigenfunctions of

Z 1
u(x) = λ (1 − xξ)u(ξ)dξ.
0

Solution
Let
C1 = h1, ui, C2 = hξ, ui
u = λ[C1 − C2 x]
1
C1 = λ[C1 − C2 ],
2
1 1
C2 = λ[ C1 − C2 ],
2 3
The determinant of this homogeneous system has to be zero.
1 − λ 12 λ
=0
− 12 λ 1 + 31 λ
Expanding,
2 1 1
1 − λ − λ2 + λ2 = 0
3 3 4
1 2 2
− λ − λ+1 = 0
12 3
2
λ + 8λ − 12 = 0
This has the solutions
√
λ = −4 ± 2 7
√
When λ = −4 + 2 7
√ √
5−2 7 5−2 7
C2 = √ C1 , u1 = 1 − √ x.
2− 7 2− 7
√
When λ = −4 − 2 7
√ √
5+2 7 5+2 7
C2 = √ C1 , u2 = 1 − √ x.
2+ 7 2+ 7
 67

2.8 From the differential equation and the boundary conditions obtained
for Exercise 2.2, find the values of λ for the existence of non-trivial
solutions.
Solution
The differential equation is
u′′ − µ2 u = 0, µ2 ≡ 1 − 2λ

The solution is given by

u = Aeµx + Be−µx ,
From the given integral equation
Z x Z 1 
ξ−x x−ξ
u(x) = λ e udξ + e udξ
−1 x
Z 0 Z 1  Z 1
ξ −ξ
u(0) = λ e udξ + e udξ = λ e−ξ [u(ξ) + u(−ξ)]dξ
−1 0 0
 Z x Z 1 
′ ξ−x x−ξ
u (x) = λ u − u − e udξ + e udξ
−1 x
Z 1 Z 0  Z 1
′ −ξ ξ
u (0) = λ e udξ − e dξ = λ e−ξ [u(ξ) − u(−ξ)]dξ
0 −1 0
Z 1
u(0) + u′ (0) = (1 − µ2 ) e−ξ u(ξ)dξ
0
Z 1
u(0) − u′ (0) = (1 − µ2 ) e−ξ u(−ξ)dξ
0

Using the solution

eµ−1 − 1 e−µ−1 − 1
 
2
(1 + µ)A + (1 − µ)B = (1 − µ ) A +B
µ−1 −µ − 1
 −µ−1
eµ−1 − 1

2 e −1
(1 − µ)A + (1 + µ)B = (1 − µ ) A +B
−µ − 1 µ−1
This can be simplified to get
eµ e−µ
A −B = 0
µ−1 µ+1
e−µ eµ
A −B = 0
µ+1 µ−1
68 CHAPTER 2. INTEGRAL EQUATIONS

The determinant of this system gives

(µ + 1)eµ = (µ − 1)e−µ

or
eµ − e−µ 1
µ −µ
=− (2.3)
e +e µ
Let µ = iν to obtain the characteristic equation

tan ν = 1/ν,

which allows an infinite number of eigenvalues.

 69

2.9 Show that the equation

Z 1
u(x) = λ |x − ξ|u(ξ)dξ
0

has an infinite number of eigenvalues and eigenfunctions.

Solution

Z x Z 1 
u(x) = λ (x − ξ)udξ + (ξ − x)udξ
0 x
Z x Z 1 
′
u (x) = λ udξ − udξ
0 x
′′
u (x) = 2λu(x).

The boundary conditions may be obtained as

Z 1 Z 1
u(0) = λ ξudξ, u(1) = λ (1 − ξ)udξ,
0 0
Z 1 Z 1
′ ′
u (0) = −λ udξ, u (1) = λ udξ.
0 0

From these we get

u(0) + u(1) + u′ (0) = 0, u′(0) + u′ (1) = 0.

Using µ2 = −2λ, the solution of the differential equation is

u(x) = A cos µx + B sin µx. (2.4)
u′(x) = −µA sin µx + µB cos µx.
Then,
u(0) = A, u′ (0) = µB,
u(1) = A cos µ + B sin µ, u′ (1) = −µA sin µ + µB cos µ.

The boundary conditions become

A + A cos µ + B sin µ + µB = 0,
µB − −µA sin µ + µB cos µ = 0.
70 CHAPTER 2. INTEGRAL EQUATIONS

The determinant of this system gives

µ[(1 + cos µ)2 + sin µ(µ + sin µ)] = 0.

This can be simplified to get

1 + cos µ
µ = −2 .
sin µ

This is equivalent to
tan µ/2 = 2/µ. (2.5)
This characteristic equation has an infinite number of solutions.
 71

2.10 Obtain the values of λ and a for the equation

Z 1
u(x) = λ (x − ξ)u(ξ)dξ + a + x2
0
to have (a) a unique solution, (b) a nonunique solution, and (c) no
solution.
Solution
Let Z 1 Z 1
C1 = udξ, C2 = ξudξ,
0 0

Then
u(x) = λ[C1 x − C2 ] + a + x2 ,
1 1
C1 = λ[ C1 − C2 ] + a + ,
2 3
1 1 1 1
C2 = λ[ C1 − C2 ] + a + ,
3 2 2 4
This system of equations can be written as
λ 1
(1 − )C1 + λC2 = a + ,
2 3
λ λ a 1
− C1 + (1 + )C2 = + ,
3 2 2 4
The determinant of this system is
λ2 λ2
1− + = 0, λ2 = −12.
4 3
We have unique solutions for C1 and C2 if λ2 6= −12.
Eliminating C2 ,
λ2 1 5
(1 + )C1 = a + − λ.
12 3 12
When λ2 = −12, the left hand side goes to zero. For the right hand
side to go to zero
1 5
a = − + λ.
3 12
In this case C1 is arbitrary and we have multiple solutions.
For other values of a, there is no solution.
72 CHAPTER 2. INTEGRAL EQUATIONS

2.11 Obtain the resolvent kernel for

Z 2π
u(x) = λ ein(x−ξ) u(ξ)dξ + f (x).
0

Solution

Z 2π
u(x) = λ ein(x−ξ) u(ξ)dξ + f (x).
0
Z 2π
C = e−inπξ udξ,
0
u = λCeinx + f,
C = 2πλC + hf, e−inξ i,
1
C = hf, e−inξ i,
1 − 2πλ
λ
u = f+ hf, e−inξ i,
1 − 2πλ
1
g(x, ξ) = ein(x−ξ) .
1 − 2πλ
 73

2.12 Find the resolvent kernel for

Z π
u(x) = λ cos(x − ξ)u(ξ)dξ + f (x).
0

Solution

Z π
u = λ [cos x cos ξ + sin x sin ξ]udξ + f,
0
C1 = hu, cosi, C2 = hu, sini,
u = f + λ[C − 1 cos x + C − 2 sin x],
λπ
C1 = hf, cosi + C1 ,
2
λπ
C2 = hf, sini + C2 ,
2
Solving for the constants

1
C1 = hf, cosi,
1 − πλ/2
1
C2 = hf, sini,
1 − πλ/2
λ
u = f+ [hf, cosi cos x + hf, sini sin x],
1 − πλ/2
u = f + λhg, f i,
1
g = cos(x − ξ).
1 − πλ/2
74 CHAPTER 2. INTEGRAL EQUATIONS

2.13 Show that for a Fredholm equation

Z b
u(x) = λ k(x, ξ)u(ξ)dξ + f (x),
a

the resolvent kernel g(x, ξ) satisfies

Z b
g(x, ξ) = k(x, ξ) + λ k(x, η)g(η, ξ)dη.
a

Thus, the resolvent kernel satisfies the integral equation for u when the
forcing function f is replaced by k with ξ and λ kept as parameters.

Solution

Z b
u(x) = λ k(x, ξ)u(ξ)dξ + f (x),
a
Z b
u(x) = f (x) + λ g(x, ξ)f (ξ)dξ.
a

Using the second equation in the first

Z b Z b Z b
f (x) + λ g(x, ξ)f (ξ)dξ = λ k(x, ξ)[f (ξ) + λ g(ξ, η)f (η)dη] + f (x),
a a a
Z b Z b Z bZ b
g(x, ξ)f (ξ)dξ = k(x, ξ)f (ξ)dξ + λ k(x, η)g(η, ξ)f (ξ)dηdξ.
a a a a

As f (x) is arbitrary,
Z b
g(x, ξ) = k(x, ξ) + λ k(x, η)g(η, ξ)dη.
a
 75

2.14 Demonstrate the preceding result when k = 1 − 2xξ in the domain

0 < x, ξ < 1.

Solution

Z 1
u(x) = λ (1 − 2xξ)u(ξ)dξ + f (x),
0
R1 R1
Let C1 = 0
udξ and C2 = 0
ξudξ

u(x) = λ[C1 − 2C2 x] + f,

C1 = λ[C1 − C2 ] + hf, 1i,
1 2
C2 = λ[ C1 − C2 ] + hf, ξi,
2 3
(1 − λ)C1 + λC2 = hf, 1i,
λ 2
− C1 + (1 + λ)C2 = hf, ξi,
2 3
The determinant of the system is 1 − λ/3 − λ2 /6

C1 = [(1 + 2λ/3)hf, 1i − λhf, ξi]/(1 − λ/3 − λ2 /6),

C2 = [(λ/2)hf, 1i + (1 − λ)hf, ξi]/(1 − λ/3 − λ2 /6),
(1 + 2λ/3) − λ(x + ξ) − 2(1 − λ)xξ
g(x, ξ) = .
1 − λ/3 − λ2 /6
Let A = λ/(1 − λ/3 − λ2 /6).

Z 1 Z 1
λ (1 − 2xη)g(η, ξ)dη = A (1 − 2xη)[(1 + 2λ/3) − λ(η + ξ)
0 0

−2(1 − λ)ηξ]dη,
= A {(1 + 2λ/3) − λ(1/2 + ξ) − (1 − λ)ξ
− 2x[(1 + 2λ/3)/2 − λ(1/3 + ξ/2) − 2(1 − λ)ξ/3]} ,
= A {1 + λ/6 − ξ − x + xξ(4/3 − λ/6)} .
g(x, ξ) − k(x, ξ) = (A/λ) {(1 + 2λ/3) − λ(x + ξ) − (1 − λ)2xξ
76 CHAPTER 2. INTEGRAL EQUATIONS

−(1 − λ/3 − λ2 /6)(1 − 2xξ) ,

= (A/λ) λ + λ2 /6 − (x + ξ)λ + xξλ(4/3 + λ/6)


= A {1 + λ/6 − x − ξ + (4/3 + λ/6)xξ} .

Thus, Z 1
g(x, ξ) = k(x, ξ) + λ k(x, η)g(η, ξ)dη.
0
 77

2.15 For the Volterra equation,

Z x
u(x) = (1 − xξ)u(ξ)dξ + 1,
0

starting with u(0) = 1, obtain iteratively, u(2) .

Solution

Z x
u(x) = (1 − xξ)u(ξ)dξ + 1.
0
u(0) = 1,
Z x
(1)
u = (1 − 2xξ)dξ + 1 = 1 + x − x2 /2,
Z0 x
u(2) = (1 − xξ)(1 + ξ − ξ 2 /2)dξ
0
= 1 + x + x2 /2 − x4 /8 − x(x2 /2 + x3 /3 − x5 /10)
= 1 + x + x2 /2 − x3 /2 − (11/24)x4 + x6 /10.
78 CHAPTER 2. INTEGRAL EQUATIONS

2.16 For the integral equation of the first kind

Z 2π
sin(x + ξ)u(ξ)dξ = sin x,
0

discuss the consequence of assuming

X X
u(x) = an cos nx + bn sin nx,
n=0 n=1

where the constants, an and bn , are unknown.

Solution
We have Z 2π
[sin x cos ξ + cos x sin ξ]u(ξ)dξ = sin x.
0
Let ∞ ∞
X X
u(x) = an cos nx + bn sin nx.
n=0 n=1

Then
Z 2π ∞ 2π
1X
Z
u(ξ) cos ξdξ = an [cos(n − 1)ξ + cos(n + 1)ξ]dξ
0 2 n=0 0
∞ Z 2π
1X
+ bn [sin(n + 1)ξ + cos(n − 1)ξ]dξ
2 n=1 0
= πa1 .
Z 2π ∞ Z 2π
1X
u(ξ) sin ξdξ = an [sin(n + 1)ξ − sin(n − 1)ξ]dξ
0 2 n=0 0
∞ Z 2π
1X
+ bn [cos(n − 1)ξ − cos(n + 1)ξ]dξ
2 n=1 0

= πb1 .

Comparing with the term on the right hand side of the given equation
1
b1 = , a1 = 0.
π
All other an and bn are arbitrary.
 79

2.17 With the quadratic forms,

Z bZ b Z b
J1 [u] = k(x, ξ)u(x)u(ξ)dξdx, J2 [u] = u2 dx,
a a a
where u is a smooth function in (a, b), show that the functions u, which
extremize
J1 [u]
J[u] = ,
J2 [u]
are the eigenfunctions of [k(x, ξ) + k(ξ, x)]/2. Also show that the ex-
trema of J correspond to the reciprocal of the eigenvalues of k(x, ξ).
Solution
With J = J1 /J2 , δJ = 0 gives
J1
J2 δJ1 − J1 δJ2 = 0, δJ1 − δJ2 = 0.
J2
From Z bZ b
J1 = k(x, ξ)u(x)u(ξ)dξdx,
a a
Z bZ b
δJ1 = k(x, ξ)[u(x)δu(ξ) + u(ξ)δu(x)]dξdx,
a a
which can be written as
Z bZ b
[k(x, ξ) + k(ξ, x)]u(ξ)δu(x)dξdx.
a a
Similarly, Z b
J2 = u2 dx,
a
gives Z b
δJ2 = 2 u(x)δu(x)dx.
a
Using these in the first variation of J, and considering δu(x) as arbi-
trary,
1 J1
[k(x, ξ) + k(ξ, x)]u(ξ)dξ = u(x).
2 J2
Let J1 /J2 = 1/λ. Then λ is the eigenvalue of the kernel [k(x, ξ) +
k(ξ, x)]/2.
As δJ = 0, the eigenvalues correspond to the extremum value of J.
80 CHAPTER 2. INTEGRAL EQUATIONS

2.18 In the preceding problem, assuming k is symmetric, compute J[v] if

X
v(x) = an un (x),
n

where un are normalized eigenfunctions.

Solution
We have X
v= an un .
Then
Z bX X X
2
J2 = kvk = an un am um dx = a2n .
a
Z bZ b X X
J1 = k(x, ξ) an un (x) am um (ξ)dξdx
a a
Z bX
am X
= um (x) an un (x)dx
a λm
X a2
n
= .
λn
P 2
a /λ
J = Pn 2 n .
an
 81

2.19 The generalized Abel equation,

Z x
u(ξ)dξ
α
= f (x),
0 (x − ξ)

is solved by multiplying both sides by (η − x)α−1 and integrating with

respect to x from 0 to η. Implement this procedure, and discuss the
allowable range for the index, α.

Solution
Multiplying the equation by 1/(η − x)1−α on both sides and integrating
form x = 0 to x = η,
Z ηZ x Z η
u(ξ)dξ f (x)dx
β α
= β
.
0 0 (η − x) (x − ξ) 0 (η − x)

where β = 1 − α.

Interchanging the order of integration,

η η η
dx f (x)dx
Z Z Z
u(ξ)dξ = .
0 ξ (η − x)β (x − ξ)α 0 (η − x)β

Let η
dx
Z
I= .
ξ (η − x)β (x − ξ)α
Using
x = (η − ξ)t + ξ,
Z 1
dt
I= α β
,
0 t (1 − t)

which is a constant. Then

x
1 d f (ξ)dξ
Z
u(x) = .
I dx 0 (x − ξ)β

To evaluate the integral I, consider the complex valued function

1
f (z) =
z α (z − 1)β
82 CHAPTER 2. INTEGRAL EQUATIONS

C4 0 C1 1 C2
C3

Figure 2.1: Contour for evaluating the integral I

where β = 1 − α. As shown in Fig. 2.1, we integrate f along the

contour C1 + C2 + C3 + C4. Let z = reiθ and z − 1 = r1 eiθ1 with the
restriction
0 < θ < 2π, 0 < θ1 < 2π.
This creates a branch cut from 0 to 1. On C1 :

θ = 0, θ1 = π, dz = dr

and
1
dr
Z Z
f dz = e−iβπ = e−iβπ I.
C1 0 r α (1− r)β
On C3 :
θ = 2π, θ1 = π, dz = dr
and
0
dr
Z Z
f dz = e−2απi−βπi = e−iαπ I.
C3 1 r α (1 − r)β
The integrals over C2 and C4 vanish when their radii shrink to zero, if
α > 0 and β > 0, respectively. This is equivalent to 0 < α < 1. Thus,
the sum of the four integrals is

(e−απi + e−βπi )I.

Our next step is to stretch the contour to a circle of radius R >> 1.

On this circle r = r1 = R and θ = θ1 . Then, as R → ∞,
I
f (z)dz = −2πi.
 83

Thus,

eαπi − e−απi
I = π.
2i
π
I= .
sin απ
Our solution for the integral equation becomes
Z x
sin απ d f (ξ)dξ
u(x) = .
π dx 0 (x − ξ)1−α
84 CHAPTER 2. INTEGRAL EQUATIONS

2.20 Solve the singular equation

Z x
u(ξ)dξ
= f (x),
0 (x2 − ξ 2 )1/2

using a change of the independent variable.

Solution
Let x2 = t and ξ 2 = τ .
dτ
dξ = √ .
2 τ
The integral equation becomes
√
Z t
u( τ )dτ √
√ √ = f ( t).
0 2 τ t−τ

Let √
u( τ ) √
v(τ ) = √ , g(t) = f ( t).
2 τ
Then, we get the Abel equation
Z t
v(τ )dτ
√ = g(t).
0 t−τ
The solution of this equation is
t
1 d g(τ )dτ
Z
v(t) = √ .
π dt 0 t−τ
Converting to the original variables, we get

u(x) 1 d1 x f (ξ)2ξdξ
Z
= p ,
2x 2πx dx 0 x2 − ξ 2
Z x
2 d f (ξ)ξdξ
u(x) = p .
π dx 0 x2 − ξ 2
 85

2.21 If an elastic, 3D half space (z >p0) is subjected to an axi-symmetric z-

displacement, w(r), where r = x2 + y 2 , we need to solve the integral
equation for the distributed pressure on the horizontal surface, z = 0,
Z r
g(ρ)dρ µ
p = w(r), 0 < r < a,
0
2
r −ρ 2 1−ν

where µ is the shear modulus, ν is the Poisson’s ratio, and a is the con-
tact radius. Obtain the pressure distribution g(r) (see Barber, (2002)).

Solution
From the solution of Exercise 2.20, we have

2µ d r w(ρ)ρdρ
Z
g(r) = p .
π(1 − ν) dr 0 r 2 − ρ2
86 CHAPTER 2. INTEGRAL EQUATIONS

2.22 In the preceding problem, if the contact displacement is due to a rigid

sphere of radius R pressing against the elastic half space, we assume
r2
w(r) = d − ,
2R
where d is the maximum indentation. Obtain the pressure distribution,
the value of the contact radius a, and the total vertical force.

Solution
From the solution of Exercise 2.21, we have
1µ d r (2Rd − ρ2 )ρdρ
Z
g(r) = p .
π(1 − ν)R dr 0 r 2 − ρ2

Substituting for w,
r
2µ d w(ρ)ρdρ
Z
g(r) = p
π(1 − ν) dr 0 r 2 − ρ2
r
 
2µ d 2
p 0 1 2 2 3/2
= (2Rd − ρ ) r 2 − ρ2 r − (r − ρ )
π(1 − ν) dr 3 0
 
µ d 1
= (2Rd − r 2 )r + r 3
π(1 − ν)R dr 3
µ
2Rd − 2r 2 .

=
π(1 − ν)R
The extent of the contact radius a is found using the circle on which
the pressure g is zero.
√
a= Rd.
The total vertical force can be calculated by integrating the pressure.
Z a
P = 2π grdr
0
 
2µ 2 1 4
= Rda − a
(1 − ν)R 2
2
µRd
= .
(1 − ν)
 87

2.23 Consider the integral equation

Z 1
u(x) = k(x, ξ)u(ξ)dξ + x2 ,
0

where (
x(ξ − 1), x < ξ,
k(x, ξ) =
ξ(x − 1), x > ξ.
Obtain an exact solution to this equation. Using the approximate ker-
nel Ax(1 − x), find A using the method of least square error. Obtain
an approximate solution for u. Compare the values of the exact and
approximate solutions at x = 0.5.

Solution
The given k is the Green’s function for the operator L = d2 /dx2 . Then
Z 1
′′
u = Lku(ξ)dξ + Lx2 = u(x) + 2.
0

u′′ − u = 2, u(0) = 0, u(1) = 1.

This equation can be solved in the form

u = A cosh x + B sinh x − 2.

The condition u(0) = 0 gives

A = 2,

and the condition u(1) = 1 gives

3 − 2 cosh 1
B= .
sinh 1
2 sinh(1 − x) + 3 sinh x
uexact = − 2.
sinh 1
To obtain an approximate solution, let
R1
x(1 − x)k(x, ξ)dx
Ax(1 − x) = k(x, ξ), A = 0 R 1 .
x 2 (1 − x)2 dx
0
88 CHAPTER 2. INTEGRAL EQUATIONS

1
1
Z
x2 (1 − x)2 dx = .
0 30
5
A = − ξ(1 − 2ξ 2 + ξ 3 ).
2
The approximate integral equation is
Z 1
5
u(x) = − x(1 − x) ξ(1 − 2ξ 2 + ξ 3 )udξ + x2 .
2 0

Let Z 1
C= ξ(1 − 2ξ 2 + ξ 3)udξ.
0
5C
u=− x(1 − x).
2
2
C= .
37
The approximate solution is
5
u = x2 − x(1 − x).
37
uexact(x = 0.5) = 0.217047, uapprox(x = 0.5) = 0.216216.
 89

2.24 Consider the differential equation

u′′ + u = x, u(0) = 0, u(1) = 0.

Find an exact solution. Obtain a finite difference solution by dividing

the domain into four equal intervals.
Convert the differential equation into an integral equation using the
Green’s function for the operator L = d2 /dx2 . Obtain a numerical so-
lution of the integral equation by dividing the domain into four intervals
and using the trapezoidal rule and the collocation method. Compare
the results with the exact solution.

Solution
The exact solution is

u = A cos x + B sin x + x.

The condition u(0) = 0 gives A = 0 and u(1) = 0 gives

1
B=− .
sin 1
Thus
sin x
u=x− .
sin 1
With step size h = 0.25, the difference equation is

un+2 − 2un + un+1 + h2 un = xn .

With u0 = u4 = 0, the remaining unknowns satisfy

1
    
−2 + 16 1 0  u1  1/4 
 1 1 1 
−2 + 16 1  u2 = 1/2
1  16 
0 1 −2 + 16 u3 3/4
 

Using the Green’s function, the differential equation can be converted

to Z 1 Z 1
u+ g(x, ξ)u(ξ)dξ = g(x, ξ)ξdξ.
0 0
After evaluating the integral on the right hand side, we have
90 CHAPTER 2. INTEGRAL EQUATIONS

1
1
Z
u+ g(x, ξ)u(ξ)dξ = x(x2 − 1).
0 6
With gij = g(xi , ξj ), using the trapezoidal rule, we get the system of
equations
3
1X 1
ui + gij uj = xi (x2i − 1), i = 1, 2, 3.
4 j=1 6

Solving the systems of equations

Exact Finite Diff. Intgral Eq.

u1 −0.044014 −0.044274 −0.044274
u2 −0.069747 −0.070156 −0.070156
u3 −0.060056 −0.060403 −0.060403
 91

2.25 Consider the nonlinear integral equation

Z 1
u(x) − λ u2 (ξ)dξ = 1.
0

Show that for λ < 1/4 this equation has two solutions and for λ > 1/4
there are no real solutions. Also show that one of these solutions is
singular at λ = 0 and two solutions coalesce at λ = 1/4. Sketch the
solutions as functions of λ. (Based on Tricomi (1957).)

Solution
We have Z 1
u(x) − λ u2 (ξ)dξ = 1.
0
Let Z 1
A= u2 dξ.
0
Then
u = 1 + λA, A = (1 + λA)2 .
This gives a quadratic equation for A:
2λ − 1 1
A2 + A + = 0.
λ2 λ2
r
1/2 − λ (1/2 − λ)2 1
A= 2
± 4
− 2.
λ λ λ
Corresponding to (±), there are two solutions:
1 h √ i
u1 = 1 + 1 − 2λ + 1 − 2λ ,
2λ
1 h √ i
u2 = 1 + 1 − 2λ − 1 − 2λ .
2λ
Note:
λ → 0, u1 → −∞, u2 → 1.
λ → 1/4, u1 → 2, u2 → 2.
For λ > 1/4, there is no real solution.
92 CHAPTER 2. INTEGRAL EQUATIONS

10

-1.0 -0.8 -0.6 -0.4 -0.2 0.2 0.4

-5

-10

Figure 2.2: The functions u1 and u2 versus λ.

Chapter 3

FOURIER TRANSFORMS

3.1 From the Fourier transform of

f (x) = h(1 − |x|),

show that ∞
sin x π
Z
dx = .
0 x 2
Solution

Z 1
1
F [h(1 − |x|)] = √ eixξ dx
2π −1
r
1 1  iξ 2 sin ξ
e − e−iξ =

= √ .
2π iξ π ξ

Inverting this
r Z ∞
1 2 sin ξ −ixξ
√ e dξ = h(1 − |x|)
2π π −∞ ξ
Z ∞
sin ξ −ixξ
e dξ = πh(1 − |x|).
−∞ ξ

When x = 0,
∞
sin ξ π
Z
dξ = .
0 ξ 2

93
94 CHAPTER 3. FOURIER TRANSFORMS

3.2 Obtain the cosine and sine transforms of

f = e−ax cos bx,

where a and b are positive constants.

Solution

r
2 a
Fc [e−ax ] =
π a2 + ξ 2
r Z ∞
2 1
Fc [e−ax cos bx] = [cos(ξ + b) + cos(ξ − b)] dx
π 0 2
 
1 a a
= √ + .
2π a2 + (ξ + b)2 a2 + (ξ − b)2
r
2 ξ
Fs [e−ax ] =
π a2 + ξ 2
r Z ∞
2 1
Fs [e−ax cos bx] = [sin(ξ + b) + sin(ξ − b)] dx
π 0 2
 
1 ξ ξ
= √ + .
2π a2 + (ξ + b)2 a2 + (ξ − b)2
 95

3.3 From the Fourier transform of

f (x) = (1 − |x|)h(1 − |x|),

compute the integrals

∞  n
sin x
Z
In = dx,
0 x

for n = 2, 3, 4.

Solution
The given function f is even.
Z 1
2
F (ξ) = √ (1 − x) cos xξdx
2π 0
r " #
2 sin ξx 1 (− cos ξx) 1
= (1 − x) − (−1)
π ξ 0 ξ2 0
r r
2 1 − cos ξ 2 sin2 ξ/2
= = 2 .
π ξ2 π ξ2
Inversion of this gives
√ Z
1 2 2 ∞ sin2 ξ/2
√ √ cos ξxdξ = f (x)
2π π −∞ ξ 2
4 ∞ sin2 ξ/2
Z
cos ξxd(ξ/2) = f (x)
π 0 2(ξ/2)2
2 ∞ sin2 ξ
Z
cos 2ξxdξ = 1 − x, 0≤x<1
π 0 ξ2

When x = 0,
π
I2 = .
2
Integrating the expression for (1 − x) with respect to x,
96 CHAPTER 3. FOURIER TRANSFORMS

∞
2 sin2 ξ x x2
Z Z
cos 2ξxdxdξ = x −
π 0 ξ2 0 2
2 ∞ sin2 ξ sin 2ξx x2
Z
dξ = x −
π 0 ξ2 2ξ 2
When x = 1/2, we get

3π
I3 = .
8
Integrating with x one more time,
∞
1 sin2 ξ x x2 x3
Z Z
sin 2ξxdxdξ = −
π 0 ξ3 0 2 6
Z ∞ 2 2
1 sin ξ 1 − cos 2ξ x x3
dξ = −
π 0 ξ3 2ξ 2 6
When x = 1, we get
π
I4 = .
3
 97

3.4 Obtain the Fourier transform of

2
f (x) = e−ax cos bx,

where a and b are positive constants.

Solution

2 2 1 2
F [e−ax ] = Fc [e−ax ] = √ e−ξ /(4a)
2a
r Z ∞
2 2 1 2
e−ax cos ξxdx = √ e−ξ /(4a)
π 0 2a
r Z ∞
2 2 1 −ax2
F [e−ax cos bx] = e [cos(ξ + b)x + cos(ξ − b)x]dx
π 0 2
1 h −(ξ+b)2 /(4a) 2
i
= √ e + e−(ξ−b) /(4a) .
2 2a
98 CHAPTER 3. FOURIER TRANSFORMS

3.5 Compute the sine transform of

e−ax
f= ,
x
and find its limit as a → 0.

Solution

r Z ∞
2 sin ξx
Fs (ξ) = e−ax dx
π 0 x
r Z ∞Z ξ
2
= e−ax cos ξxdξdx
π
r Z0 ξ 0
2 a
= dξ
π 0 a + ξ2
2
r Z ξ r
2 dξ/a 2
= 2
= tan−1 (ξ/a).
π 0 1 + (ξ/a) π
As a → 0, tan−1 (∞) = π/2,
r
π
Fs [1/x] = .
2
 99
√
3.6 Using f = 1/ x in the cosine and sine forms of the Fourier integral
theorem, show that
Z ∞ Z ∞ r
cos x sin x π
√ dx = √ dx = ,
0 x 0 x 2

and √ √ p
Fc [1/ x] = Fs [1/ x] = 1/ ξ.

Solution
Fourier integral theorem gives
2 ∞
Z ∞
cos ξtdt 1
Z
cos ξx √ dξ = √
π 0 t x
Z ∞ Z 0∞
2 cos τ dτ dξ 1
cos ξx √ √ = √ , τ = ξt
π 0 0 τ ξ x
Z ∞ Z ∞
2 cos ξx cos τ 1
√ dξ √ dτ = √
π 0 ξ 0 τ x
When x = 1,
∞
r
cos x π
Z
√ dx =
0 x 2
Similarly, using the Sine
Z transform,
2 ∞
Z ∞
sin ξtdt 1
sin ξx √ dξ = √
π 0 0 t x
Z ∞ r
sin x π
√ dx =
0 x 2
Using these
 
1 1
Fc √ = √ ,
x ξ
 
1 1
Fs √ = √ .
x ξ
100 CHAPTER 3. FOURIER TRANSFORMS

3.7 Invert the following

(a)
1
F (ξ) = ,
ξ2 + 4ξ + 13
(b)
1
F (ξ) = .
(ξ 2 + 4)2
Solution

(a)
   
−1 1 −1 1
F 2
= F
ξ + 4ξ + 13 (ξ + 2)2 + 9
 
2ix −1 1
= e F
ξ2 + 9
1 −3|x|+2ix
= e .
3
(b) We have
  r
−1 a π −a|x|
F = e .
a + ξ2
2 2
Differentiating this with respect to a,

2a2
  r
−1 1 π
F − = − |x|e−a|x| .
a2 + ξ 2 (a2 + ξ 2 )2 2
When a = 2, we have
  r  
−1 8 π −2|x| 1 −2|x|
F = |x|e − e
(4 + ξ 2 )2 2 2
  r
1 1 π
F −1 = [|x| − 4] e−2|x|
(4 + ξ 2 )2 8 2
 101

3.8 Invert 2
e−ξ
F (ξ) = ,
ξ2 + 4
in terms of complementary error functions.

Solution
Let f (x) be the inverse. We have

2 1 2
F −1 [e−ξ ] = √ e−x /4 ,
2
  r
1 1 π −2|x|
F −1 2 = e .
ξ +4 2 2
Using convolution integral, we have
r Z ∞
1 1 1 π 2
f (x) = √ √ e−2|x−t|−t /4 dt
2π 2 2 2 −∞
Z x Z ∞ 
1 −2x+2t−t2 /4 2x−2t−t2 /4
= √ e + e dt
4 2 −∞ x
 Z x Z ∞ 
1 −2x −t2 /4+2t 2x −t2 /4−2t
= √ e e dt + e e dt
4 2 −∞ x
 Z ∞ Z ∞ 
1 −2x −(t/2+2)2 +4 2x −(t/2+2)2 +4
= √ e e dt + e e dt
4 2 −x x
√
π 
√ e4−2x erfc(2 − x/2) + e4+2x erfc(2 + x/2)

=
4 2
where ∞
2
Z
2
erfc(x) = √ e−y dy.
π x
102 CHAPTER 3. FOURIER TRANSFORMS

3.9 Invert
e−iaξ
F (ξ) = .
(ξ − 1)2 + 4
Solution

Z ∞ −iaξ−ixξ
1 e dξ
f (x) = √ 2
2π −∞ (ξ − 1) + 4
Z ∞
1 e−iyξ dξ
= √ , y = x + a,
2π −∞ (ξ − 1)2 + 4
Z ∞ −iy(η+1)
1 e dη
= √ 2
, η = ξ − 1,
2π −∞ η + 4
r −iy
1 πe
= √ e−2|y|
2π 2 2
1 −i(x+a)−2|x+a|
= e .
4
 103

3.10 For 0 < a < π, compute

   
−1 cosh aξ −1 sinh aξ
f (x) = F , g(x) = F .
sinh πξ sinh πξ
Solution

Let ∞ ∞
cosh aξ −iξx sinh aξ −iξx
Z Z
I= e dξ, J= e dξ.
−∞ sinh πξ −∞ sinh πξ
Consider the complex valued functions
cosh az −izx sinh az −izx
F1 (z) = e , F2 (z) = e .
sinh πz sinh πz

2i C2

i C∞

0 C
1

Figure 3.1: Contour for integration

There are two poles on the contour at z = 0 and at z = 2i. and one
inside at z = i. Using Cauchy principal value, the line integrals of F1
can be evaluated as follows:
Z
F1 (z)dz = I.
C1

The line integrals on C∞ go to zero.

Z −∞
cosh a(ξ + 2i) −ix(ξ+2i)
Z
F1 dz = e dξ
C2 ∞ sinh π(ξ + 2i)
Z ∞
2x cosh aξ cos 2a + i sinh aξ sin 2a −ixξ
= −e e dξ
−∞ sinh πξ
= −e2x [cos 2aI + i sin 2aJ]
104 CHAPTER 3. FOURIER TRANSFORMS

The residues are:

1
Res(0) =
π
1
Res(i) = − cos aex
π
1
Res(0) = cos 2ae2x
π
From the closed contour, we get
(1 − e2x cos 2a)I − ie2x sin 2aJ = i 1 + cos 2ae2x − 2 cos aex
 

Next, we work with F2 ,

Z
F2 (z)dz = J,
C1
Z −∞
sinh a(ξ + 2i) −ix(ξ+2i)
Z
F2 (z)dz = e dξ
C2 ∞ sinh π(ξ + 2i)
Z ∞
2x sinh aξ cos 2a + i cosh aξ sin 2a −ixξ
= −e e dξ
−∞ sinh πξ
= −e2x [cos 2aJ + i sin 2aI]

The residues are:

i
Res(i) = − sin aex
π
i
Res(0) = sin 2ae2x
π
From the closed contour, we get
(1 − e2x cos 2a)J − ie2x sin 2aI = − sin 2ae2x − 2 sin aex
 

The two simultaneous equations for I and J can be solved to get

cos a − cosh x cosh x − cos a
I = 2i sinh x, J =2 sin a.
cosh 2x − cos 2a cosh 2x − cos 2a
These can be simplified to get
sinh x sin a
I = −i , J= .
cosh x + cos a cosh x + cos a
 105

Then
 
−1 cosh aξ 1 sinh x
F = −i √ ,
sinh πξ 2π cosh x + cos a
 
−1 sinh aξ 1 sin a
F = √ .
sinh πξ 2π cosh x + cos a
106 CHAPTER 3. FOURIER TRANSFORMS

3.11 Evaluate the integral

∞
dx
Z
I= ,
−∞ (x2 + a2 )(x2 + b2 )

by (a) using partial fractions and (b) using the convolution theorem.

Solution
(a)  
1 1 1 1
= 2 − .
(x2 + a2 )(x2 + b2 ) b − a2 x2 + a2 x2 + b2
 ∞
1 1 −1 x 1 −1 x π
I= 2 2
tan − tan = .
b −a a a b b −∞ ab(a + b)

(b) To use convolution, let

∞
dt
Z
J(x) = .
−∞ (t2 + a2 )[(x− t)2 + b2 ]

√
   
1 1
F [J] = 2πF 2 F 2
x + a2 x + b2
√ 1 π π −a|ξ| −b|ξ|
r r r
π π −(a+b)|ξ|
= 2π e e = e ,
ab 2 2 2 ab
π a+b
J(x) = ,
ab x2 + (a + b)2
π
I = J(0) = .
ab(a + b)
 107

3.12 For
f (x) = h(1 − |x|),
show that kf k = kF k.

Solution

2
Z 1 √
kf k = 1dx = 2, kf k = 2.
−1
Z 1
1 1 eiξ − e−iξ
F [f ] = F =√ eiξx dx = √ ,
2π −1 2π iξ
r
2 sin ξ
F = .
π ξ
2 ∞ sin2 ξ
Z
kF k2 = dξ = 2.
π −∞ ξ 2
kF k = kf k.
108 CHAPTER 3. FOURIER TRANSFORMS

3.13 Obtain the cosine transform of

x2 π
 
f = cos − ,
2 8

and show that it is self-reciprocal.

Solution

x2 π
 
f = cos −
2 8
2
π x π x2
= cos cos + sin sin
8 2 8 2
x2 ξ2 ξ2
 
1
Fc cos = √ cos + sin ,
2 2 2 2
x2 ξ2 ξ2
   
1
Fc sin = √ cos − sin ,
2 2 2 2
1 ξ2 h π πi
Fc [f ] = √ cos cos + sin
2 2 8 8
2 h
1 ξ π πi
= + √ sin cos − sin .
2 2 8 8
Using

1 h π πi π
√ cos + sin = cos ,
2 8 8 8
1 h π πi π
√ cos − sin = sin ,
2 8 8 8
 2 
ξ π
Fc [f ] = cos − ,
2 8

which shows self-reciprocity.

 109

3.14 Show that

N
1 X sin(2N + 1)θ/2
+ cos nθ = .
2 n=1 2 sin θ/2

Solution
Let " N #
N
1 X X
inθ 1
S= + cos nθ = Re e − .
2 n=1 n=1
2
Using
N
X
n 1 − r n+1
r = ,
n=1
1−r
1 − ei(N +1)θ
 
1
S = Re iθ
−
1−e 2
 i(N +1)θ/2 i(N +1)θ/2 i(N +1)θ/2

e −e e 1
= Re −
e−iθ/2 − eiθ/2 eiθ/2 2
sin(N + 1)θ/2 Nθ 1
= cos −
sin θ/2 2 2
1 sin(2N + 1)θ/2
= .
2 sin θ/2
110 CHAPTER 3. FOURIER TRANSFORMS

3.15 Using the definition

r Z ∞
2
Fc (ξ) = f (x) cos xξdx,
π 0
show that
N Z ∞
1 X 1 sin(2N + 1)xξ/2
Fc (0) + Fc (nξ) = √ f (x) dx.
2 n=1
2π 0 sin xξ/2

Solution
r Z ∞
2
Fc (ξ) = f (x) cos xξdx,
π 0

N r Z ∞ " N
#
1 X 2 1 X
Fc (0) + Fc (nξ) = f (x) + cos nπξ dx
2 n=1
π 0 2 n=1
r Z ∞
2 sin(N + 1)xξ/2
= f (x) dx
π 0 2 sin xξ/2
Z ∞
1 sin(N + 1)xξ/2
= √ f (x) dx.
2π 0 sin xξ/2
 111

3.16 Using the Riemann-Lebesgue localization lemma, show that the pre-
ceding relation yields the Poisson sum formula
" ∞
# " ∞
#
p 1 X √ 1 X
ξ Fc (0) + Fc (nξ) = x f (0) + f (nx) ,
2 n=1
2 n

where xξ = 2π.

Solution
Let
∞
1 X
S = Fc (0) + Fc (nξ)
2 n=1
Z ∞
1 sin(N + 1)xξ/2
= lim √ f (x) dx.
N →∞ 2π 0 sin xξ/2
Using t = xξ/2,
∞
1 sin(2N + 1)t 2dt
Z
S = lim √ f (2t/ξ)
N →∞ 2π 0 sin t ξ
From

sin(2N + 1)t sin 2Nt cos t

= + cos 2Nt
sin t sin t Z ∞
2 1 sin 2Nt cos t
S = lim √ f (2t/ξ) dt
N →∞ ξ 2π 0 sin t
Z ǫ
2 1 sin 2Nt cos t
= lim √ f (2t/ξ) dt
N →∞ ξ 2π 0 t
Z π−ǫ
sin 2Nt cos t
f (2t/ξ) dt
ǫ sin t
Z π+ǫ 
sin 2Nt cos t
= f (2t/ξ) dt + · · ·
π−ǫ sin t
√ " ∞
#
2π 1 X
= f (0) + f (2nπ/ξ) .
ξ 2 n=1

With 2π/ξ = x,
112 CHAPTER 3. FOURIER TRANSFORMS

∞
p √ 1 X
ξS = x[ f (0) + f (nx)].
2 n=1
 113

3.17 From the preceding result, show that the Theta function given by the
sum ∞
X 2 2
θ(x) = e−n x /2 ,
−∞

satisfies √
2π
θ(x) = θ(2π/x).
x

Solution
We have
f (x) = e−x2 /2, Fc (ξ) = e−ξ 2 /2.
" ∞
#
1 X
S(x) = 2 f (0) + f (nx) .
2 n=1

By the Poisson sum formula, using xξ = 2π,

∞
x
X r
2[FC (0) + Fc (nξ)] = S(x).
n=−∞
ξ
" ∞
#
X
−n2 ξ 2 /2
2 1+ e = S(2π/x).
n=−∞

Thus √
2π
θ(x) = θ(2π/x).
x
114 CHAPTER 3. FOURIER TRANSFORMS

3.18 Solve the wave equation for an infinite string

∂2u ∂2u
= ,
∂x2 ∂t2
with the initial conditions
2 ∂u(x, 0
u(x, 0) = U0 e−x , = 0.
∂t

Solution
Taking the Fourier transform

−ξ 2 U(ξ) − Ü,

which has the general solution

U(ξ, t) = Aeiξt + Be−iξt .

The initial conditions give

U0 2
U(ξ, 0) = √ e−ξ /2 , U̇(ξ, 0) = 0.
2
Then
U0 2
A + B = √ e−ξ /2 , A − B = 0.
2
Solving
U0 2
A = B = √ e−ξ /2 .
2 2
U0 h 2 2
i
U = √ e−ξ /2+iξt + e−ξ /2−iξt .
2 2
Inverting
U0 h −(x−t)2 −(x+t)2
i
u(x, t) = e +e .
2
 115

3.19 Solve the wave equation for an infinite string

∂2u ∂2u
= ,
∂x2 ∂t2
with the initial conditions
U0 ∂u(x, 0
u(x, 0) = , = 0.
x2 + 1 ∂t

Solution
Fourier transform gives
r
π
Ü + ξ 2 U = 0, U(ξ, 0) = U0 e|ξ| , U̇(ξ, 0) = 0.
2
Using the solution
U(ξ, t) = Aeiξt + Be−iξt ,
in the initial conditions, we find
r r
π U0 −|ξ|+iξt π U0 −|ξ|−iξt
A= e , B= e .
2 2 2 2
Inverting this, we get
 
U0 1 1
u(x, t) = + .
2 (x + t)2 + 1 (x − t)2 + 1
116 CHAPTER 3. FOURIER TRANSFORMS

3.20 Lateral motion of an infinite beam satisfies

∂4v ∂2v
EI + m = 0,
∂x4 ∂t2
where EI is the bending stiffness and m is the mass per unit length. If
the initial conditions are
2 ∂v(x, 0)
v(x, 0) = V0 e−x , = 0,
∂t
and v → 0 as |x| → ∞, obtain v(x, t).

Solution
The given equation

a2 v ′′′′ + v̈ = 0, a2 = EI/m,

under the Fourier transform, becomes

V̈ + a2 ξ 4 V = 0.

This has the solution

V (ξ, t) = A cos aξ 2 t + B sin aξ 2 t.

At t = 0,
V (ξ, 0) = F (ξ), V̇ (ξ, 0) = 0.
Then
A = F (ξ), B=0
and
V (ξ, t) = F (ξ) cos aξ 2 t.
We have
x2 x2
 
−1 2 1
F [cos aξ t] = √ cos + sin .
2 at 4at 4at
Using convolution,
∞
y2 y2
 
1 1
Z
v(x, t) = √ f (x − y) √ cos + sin dy.
2π −∞ 2 at 4at 4at
 117

Using
y2
z2 = ,
4at
1
Z ∞ √
f (x − 4atz) cos z 2 + sin z 2 dz.
 
v(x, t) = √
2π −∞

For the special case

2 1 2
f (x) = V0 e−x , F (ξ) = √ e−ξ /4 .
2

r Z ∞
2 1 2
v(x, t) = V0 √ e−ξ /4 cos atξ 2 cos ξxdξ
π 0 2
Z ∞
V0 h 2 2
i
= √ e−ξ /4+iatξ + e−ξ /4−iatξ cos ξxdξ
2 π 0
Using
1 1
a21 = − iat, a22 = − iat,
4 4
and
φ = tan−1 4at, r = [1 + 16a2 t2 ]1/4 ,
1 1
a1 = re−iφ/2 , a2 = reiφ/2 .
2 2
Z ∞h
V0 2 2 2 2
i
v(x, t) = √ e−a1 ξ + e−a2 ξ cos ξxdξ
2 π 0
V0 −1 h −a21 ξ2 −a22 ξ 2
i
= √ F e +e
2 2
 
V0 1 −x2 /(4a21 ) 1 −x2 /(4a22 )
= e + e
4 a1 a2
V0 iφ/2 −(x2 /r2 ) exp(iφ)
h
−iφ/2 −(x2 /r 2 ) exp(−iφ)
i
= e e +e e
2r
V0 −(x2 /r2 ) cos φ iφ/2 −i(x2 /r2 ) sin φ
h
−iφ/2 i(x2 /r 2 ) sin φ
i
= e e e +e e
2r  2 
V0 −(x2 /r2 ) cos φ x φ
= e cos 2 sin φ − .
r r 2
118 CHAPTER 3. FOURIER TRANSFORMS

3.21 Consider the motion of a concentrated force on an elastically supported

taut string. The deflection u satisfies

∂2u ∂2u
+ u = − δ(x − ct),
∂x2 ∂t2
with u → 0 as |x| → ∞. Assume the speed of the force, c > 1. We are
interested in the steady state distribution of the deflection. Introducing
the new variables:
ξ = x − ct, τ = t,
rewrite the governing equation. For the steady state, let ∂u/∂τ and
∂ 2 u/∂τ 2 go to zero. Obtain the solution which is continuous at ξ = 0.

Solution
With
ξ = x − ct, τ = t,
∂ ∂ ∂ ∂ ∂
= , = −c .
∂x ∂ξ ∂t ∂τ ∂ξ
The equation becomes

∂2u ∂u ∂2u 2
2∂ u
+ u = − 2c + c − δ(ξ).
∂ξ 2 ∂τ 2 ∂ξ∂τ ∂ξ 2
For steady-state we drop all time derivatives to get

∂2u
(c2 − 1) − u = δ(ξ).
∂ξ 2
The solution for positive and negative values of ξ can be written as
(
Ae−βξ , ξ > 0
u= ,
Beβξ , ξ < 0.

where √
β= c2 − 1.
Continuity at ξ = 0 requires A = B. The jump condition gives

u′ (a+ ) − u′(a− ) = 1/β 2, −2βA = 1/β 2.

 119

Then
1
A=− ,
2β 3
and
1 −β|x−ct||
u(x − ct) = − e .
2β 3
120 CHAPTER 3. FOURIER TRANSFORMS

3.22 The transient heat conduction in a 1D infinite bar satisfies

∂2u ∂u
2
= ,
∂x ∂t
with the initial condition
u(x, 0) = T0 e−|x| .
Obtain the temperature u(x, t).
Solution
Fourier transform gives
r
∂U 2 1
−ξ 2 U = , U(ξ, 0) = T0 .
∂t π ξ2 + 1
The solution of this equation is
r
−ξ 2 t 2 1
U = Ae , A = T0 .
π ξ2 + 1
Then r 2
2 e−ξ t
U(ξ, t) = T0 .
π ξ2 + 1
We invert this using the convolution theorem.
Z ∞
T0 1 2
u(x, t) = √ e|x−y| √ e−y /(4t) dy
2π −∞ 2t
Z x Z ∞ 
T0 −x+y−y 2 /(4t) dy x−y−y 2 /(4t) dy
= √ e √ + e √
2π −∞ 2t x 2t
 Z ∞ Z ∞ 
T0 −x −y−y 2 /(4t) dy x −y−y 2 /(4t) dy
= √ e e √ +e e √
2π −x 2t x 2t
Z ∞ √ Z ∞ √ √
 √

T0 −x −z 2 −2 tz x −z 2 −2 tz
= √ e √
e 2dz + e √
e 2dz
2π −x/(2 t) x/(2 t)
 Z ∞ √
Z ∞ √

T0 −(x−t) −(z− t)2 x+t −(z− t)2
= √ e √
e dz + e √
e dz
π −x/(2 t) x/(2 t)
√ √
     
T0 −(x−t) x x+t x
= e erfc − √ + t + e erfc √ + t .
2 2 t 2 t
 121

3.23 The transient heat conduction in a 1D infinite bar with heat generation
satisfies
∂2u ∂u 1
2
= + 2 ,
∂x ∂t x +1
with the initial condition u(x, 0) = T0 . Obtain u(x, t) in the form of a
convolution integral.
Solution
For the Fourier transform to exist the function should be integrable.
Let
u(x, t) = v(x, t) + T0 .
The v satisfies
∂2v ∂v 1
2
= + 2 ,
∂x ∂t x + 1
with v(x, 0) = 0. Taking the Fourier transform, we have
r
2 ∂V π −|ξ|
−ξ V = + e .
∂t 2
The solution
π e−|ξ|
r
−ξ 2 t
V (ξ, t) = Ae − ,
2 ξ2
with the condition V (ξ, 0) = 0, gives
r −|ξ|
πe
A=
2 ξ2
and
r Z t
π e−|ξ| h −ξ2 t
r i π 2
V (ξ, t) = 2
e −1 =− e−|ξ| e−ξ t dt.
2 ξ 2 0

1 t ∞ −ξ2 t−|ξ|−ixξ
Z Z
v(x, t) = − e dξdt
2 0 −∞
1 t
Z Z 0 Z ∞ 
−ξ 2 t+ξ−ixξ −ξ 2 t−ξ−ixξ
= − e dξ + e dξ dt
2 0 −∞ 0
1 t
Z Z ∞ Z ∞ 
−ξ 2 t−ξ+ixξ −ξ 2 t−ξ−ixξ
= − e dξ + e dξ dt
2 0 0 0

The terms of the form −ξ 2 t − (1 − ix)ξ can be transformed by completing

the square. Let a1 = 1 − ix and a2 = 1 + ix.
122 CHAPTER 3. FOURIER TRANSFORMS

1 t a21 /(4t) ∞ −t(ξ+a1 /(2t))2

Z  Z Z ∞ 
a22 /(4t) −t(ξ+a2 /(2t))2
v(x, t) = − e e dξ + e e dξ dt
2 0 0 0
1 t a21 /(4t)
Z     
a1 a22 /(4t) a2 dt
= − e erfc √ +e erfc √ √ .
2 0 2 t 2 t t
 123

3.24 The transient heat conduction in a 2D infinite plate satisfies

∂2u ∂2u ∂u
2
+ 2 = ,
∂x ∂y ∂t
with
2 +y 2 )
u(x, y, 0) = U0 e−(x .
Obtain u(x, y, t).

Solution
Taking the Fourier transform with x → ξ and y → η,
∂U
−(ξ 2 + η 2 )U = ,
∂t
which has the solution
2 +η 2 )t
U = Ae−(ξ .
The initial condition gives
1 2 2
A = U0 e−(ξ +η )/4 .
2
Then
1 2 2
U = U0 e−(ξ +η )(1/4+t) .
2
Inverting this, we get
U0 −(x2 +y2 )/(1+4t)
u(x, y, t) = e .
1 + 4t
124 CHAPTER 3. FOURIER TRANSFORMS

3.25 The transient heat conduction in a 2D infinite plate satisfies

∂2u ∂2u ∂u
+ = ,
∂x2 ∂y 2 ∂t

with
u(x, y, 0) = U0 h(1 − |x|)h(1 − |y|).

Obtain u(x, y, t).

Solution
We have
∂U
−(ξ 2 + η 2 )U = .
∂t
The initial condition is

U(ξ, η, 0) = U0 F [h(1 − |x|)]F [h(1 − |y|)].

r
2 1
F [h(1 − |x|)] = int (1 − x) cos ξxdx
π 0
r 
sin ξx 1

2 (− cos ξx) 1
= (1 − x) − (−1) 0
π ξ 0 ξ2
r r
2 1 − cos ξ 2 sin2 ξ/2
= = 2 .
π ξ2 π ξ2

The solution
+ η 2 )t
U = Ae−(ξ ,

with the initial condition gives

8 sin2 ξ/2 sin2 η/2

A= .
π ξ2 η2

Then
u(x, y, t) = v(x, t)v(y, t),
 125

where
" r #
2 sin2 ξ/2 −ξ2 t
v(x, t) = F −1 2 e
π ξ2
Z 1
1 2
= √ e−(x−z) /(4t) (1 − |z|)dz
4πt −1
Z 1 Z 0 Z 1 
1 −(x−z)2 /(4t) −(x−z)2 /(4t) −(x−z)2 /(4t)
= √ e dz + e zdz − e zdz
4πt −1 −1 0
√ √
Using x − z = 2 tr, dz = −2 tdr, we find
(Z √ √
1 (x+1)/(2 t) x/(2 t) √
Z
−r 2 2
v(x, t) = √ √
e dr − √
e−r (x − 2 tr)dr
π (x−1)/(2 t) (x+1)/(2 t)
√ )
Z (x−1)/(2 t)
2 √
+ √
e−r (x − 2 tr)dr
x/(2 t)

1 x+1 x−1 x
= erf √ − erf √ − 2xerf √
2 2 t 2 t 2 t
 
x+1 x−1 x
+x erf √ + erf √ − 2erf √
2 t 2 t 2 t
√ h −x2 /(4t) 2
−(x+1) /(4t) −(x−1)2 /(4t) −x2 /(4t)
io
− t e −e +e −e

1 x+1 x−1 x
= (x + 1)erf √ + (x − 1)erf √ − 2xerf √
2 2 t 2 t 2 t
√ h −(x+1)2 /(4t) 2 2
io
+ t e − e−(x−1) /(4t) + e−x /(4t)
126 CHAPTER 3. FOURIER TRANSFORMS

3.26 A semi-infinite rod has an initial temperature distribution

2
u(x, 0) = T0 xe−x ,

and the transient heat conduction is governed by

∂2u ∂u
2
= .
∂x ∂t
If the boundary temperature is u(0, t) = 0, obtain the temperature
u(x, t).

Solution
Taking the Sine transform
∂Us 2
−ξ 2 Us = , Us (ξ, t) = Ae−ξ t .
∂t
2
Us (ξ, 0) = T0 Fs [xe−x ].
Differentiating with respect to ξ, the Cosine transform,
2 x2 1 2 2
Fc [e−a ] = √ e−ξ /4a ,
2a
we get
2 x2 ξ 2 2
−Fs [xe−a ]=− √ e−ξ /4a .
2 2a 3

2 ξ 2
Fs [xe−x ] = √ e−ξ /4 .
2 2
Then
T0 2
A = √ ξe−ξ /4 .
2 2
T0 2
Us = √ ξe−ξ (t+1/4) .
2 2
T0 2
u(x, t) = 3/2
xe−x /(4t+1) .
(1 + 4t)
 127

3.27 In the preceding problem if we replace the condition at x = 0 by

∂u(0, t)
− hu(0, t) = 0,
∂x
where h is a constant, obtain the solution using the mixed trigonometric
transform.

Solution
Taking the mixed transform
r Z ∞
2
FR [u] = U = [ξ cos ξx + h sin ξx]u(x)dx,
π 0

∂U 2
−ξ 2 U = , U = Ae−ξ t .
∂t
From the initial condition
 
∂u
FR [u(x, 0)] = −Fs − hu dx,
∂x
r Z ∞
2 2
A=− T0 [1 − 2x2 − hx]e−x sin ξxdx
π 0
and r ∞
2
Z
2 2
U =− T0 [1 − 2x2 − hx]e−x sin ξxe−ξ t dx.
π 0

For inversion, we use

u′ − hu = −FS−1 [U]
2T0 ∞ ∞
Z Z
2 2
= [1 − 2y 2 − hy]e−y e−ξ t sin ξy sin ξxdydξ
π 0
Z Z 0
T0 ∞ ∞ 2 2
= [1 − 2y 2 − hy]e−y e−ξ t [cos(x − y) − cos(x + y)]dξdy
π 0 0
Z ∞
T0 2
h 2 2
i
= √ (1 − 2y 2 − hy)e−y e(x−y) /4t − e−(x+y) /4t dy
2tπ 0
Next we solve
2 /4t
w ′ − hw = e−(x−u) .
128 CHAPTER 3. FOURIER TRANSFORMS

Using the homogeneous solution, ehx , with variable coefficient

w = Aehx ,

we find
2 2
A′ = e−(x −2xa+a +4htx)/4t
2 2 2
= e−(x−a+2ht) /4t e[(a−2ht) −a ]/4t
√
Letting y = x − a + 2ht)/2 t
,
√
Z (x−a+2ht)/4t
h(ht−a) 2
A = e 2 t e−y dy.
∞
√ x + 2ht − a
w(x, a) = − πteh(x+ht−a) erfc √
2 t
Z ∞
T0 2
u(x, t) = √ (1 − 2y 2 − hy)e−y [g(x, y) − g(x, −y)]dy,
2π 0
where

x + 2ht + y
g(x, y) = eh(x+ht+y) erfc √ .
2 t
 129

3.28 Solve the integral equation

Z ∞
−|x|
u(x) = xe − u(t)e−|x−t| dt.
−∞

Solution
We have r
2 1
F [e−|x| ] = .
π ξ2 + 1
r
−|x| 2 2ξ
F [xe ]= i 2 .
π (ξ + 1)2
The Fourier transform of the integral equation is

√
r r
2 2ξ 2 U
U= i 2 2
− 2π .
π (ξ + 1) π ξ2 + 1
r  
2 1 1
U= i2ξ −
π ξ2 + 1 ξ2 + 3
n √ o
u = 2x e−|x| − e− 3|x| .
130 CHAPTER 3. FOURIER TRANSFORMS

3.29 Find the solution of

Z ∞
−2|x|
u(x) = e − u(t)e−|x−t| dt.
−∞

Solution
r
2 2 U
U= 2
−2 2 .
πξ +4 ξ +1
r r 
2(ξ 2 + 1)

2 2 3 2
U= =2 − .
π (ξ 2 + 4)(ξ 2 + 3) π ξ2 + 4 ξ2 + 3
4 √
u = 3e−2|x| − √ e− 3|x| .
3
 131

3.30 Find the solution of

Z ∞
−p
u(x) = |x| − u(t)e−|x−t| dt,
−∞

where p is a real constant: 0 < p < 1.

Solution

F = F [|x|−p ].
ξ2 + 1 2
U= 2 F =F − 2 F.
ξ +3 ξ +3
r Z ∞
2 √
u = |x|−p − |t|−p e− 3|x−t| dt.
3π −∞
132 CHAPTER 3. FOURIER TRANSFORMS

3.31 Solve ∞
1
Z
−x2 2
u(x) = xe −√ e−(x−t) u(t)dt,
π −∞

in the form of an infinite series.

Solution

2 x2 i 2 2
F [xe−a ]= √ ξe−ξ /4a .
( 2a) 3

i 2 √ 2
U = √ ξe−ξ /4 + πe−ξ /4 U.
( 2) 3

i 2 1
U = √ ξe−ξ /4 √ −ξ2 /4 .
( 2) 3 1 − πe
i −ξ 2 /4
h √ −ξ2 /4 −2ξ 2 /4
i
U= √ ξe 1 + πe + πe +··· .
( 2)3
∞
X π n/2 (n + 1)3/2 2 /(n+1)
u= xe−x .
n=0
23/2
 133

3.32 Solve ∞
x 1
Z
2
u(x) = 2 −√ e−(x−t) u(t)dt,
x +4 π −∞

in the form of a series of convolution integrals.

Solution
  r
−x2 1 2 x π
F [e ] = √ e−ξ /4 , F 2 =F = sgn(ξ)e−2ξ .
2 x +4 2
The integral equation becomes
√
2π 1 2
U = F − √ √ e−ξ /4 U.
π 2

F −ξ 2 /4 2
U= −ξ 2 /4 = F [1 − e + e−2ξ /4 − · · · ].
1+e
X∞
2
U =F + (−1)n F e−nξ /4 .
n=1
∞ ∞
x (−1)n t
X Z
2
u= 2 + √ e−(x−t) /n .
x + 4 n=1 πn −∞ t2 +4
134 CHAPTER 3. FOURIER TRANSFORMS

3.33 Solve the system of integral equations

1 ∞
Z
u1 (x) = u2 (t)e−|x−t| dt + 2e−|x| ,
2 −∞

1 ∞
Z
u2 (x) = u1 (t)e−|x−t| dt − 4e−2|x| .
2 −∞

Solution
Taking the Fourier transform

1√
r r
2 U2 2 1
U1 = 2π 2
+2 ,
2 π1+ξ π 1 + ξ2
1√
r r
2 U1 2 2
U2 = 2π − 4 .
2 π 1 + ξ2 π 1 + ξ2
Rearranging these equations, we find
r
2 2
(1 + ξ )U1 − U2 = 2,
π
r
2 1 + ξ2
−U1 + (1 + ξ 2 )U2 = (−8) .
π 4 + ξ2
Solving these, we get
r 
1 + ξ2

1 2 2
U1 = 2(1 + ξ ) − 8 ,
(1 + ξ 2 )2 − 1 π 4 + ξ2
r 
(1 + ξ 2 )2

1 2
U2 = 2−8 .
(1 + ξ 2 )2 − 1 π 4 + ξ2

These may be simplified to the forms,

r r 
2 ξ2 + 1 2

2 3 1
U1 = = − ,
π ξ2 + 2 ξ2 + 4 π ξ2 + 4 ξ2 + 2
r r 
2 ξ2 + 1 2

2 1 9
U2 = = − ,
π ξ2 + 2 ξ2 + 4 π ξ2 + 4 ξ2 + 2
Inversion gives
 135

3 −2|x| 1 √
u1 = e − √ e− 2|x| ,
2 2
1 −√2|x| 9
u2 = √ e − √ e−2|x| .
2 2
136 CHAPTER 3. FOURIER TRANSFORMS

3.34 Show that

u = 1/2, and u = c − x2 ,
are solutions of
∞
1
Z
2 /2
u(x) ± √ e−t u(x − t)dt = 1,
2π −∞

corresponding to “+” and “−” signs, respectively. Here, c is a constant.

Solution
Let ∞
1
Z
2 /2
J1 = u(x) + √ e−t u(x − t)dt.
2π −∞

When u = 1/2, we have

Z ∞
1 1 2 dt
J1 = +√ e−t /2
2 2π −∞ 2
1 1
= + = 1.
2 2
Let ∞
1
Z
2 /2
J2 = u(x) − √ e−t u(x − t)dt.
2π −∞

When u = c − x2 , we have
Z ∞
2 1 2
J2 = c − x − √ e−t /2 [c − (x − t)2 ]dt
2π −∞
 r   r  Z ∞
1 π 2 1 π 1 2
= c 1− √ 2 −x 1− √ 2 −√ e−t /2 [2x − t]tdt
2π 2 2π 2 2π −∞
Z ∞
1 2
= √ e−t /2 t2 dt
2π −∞
 Z ∞ 
1 −t2 /2 ∞ −t2 /2
= √ e (−t) −∞ + e dt
2π −∞
= 1.
 137

3.35 Obtain all the solutions of the integral equation

Z ∞
1 2
u(x) − √ e−t /2 u(x − t)dt = 1.
2π −∞

Solution
We use the Wiener-Hopf method for this, as 1 does not have a Fourier
transform
Let f (x) = 1.
Z ∞
1 1 1
F+ = √ eiζx 1dx = − √
2π 0 2π iζ
Z 0
1 1 1
F− = √ eiζx 1dx = √ ,
2π −∞ 2π iζ
2 2
k = e−x /2 , K = e−ζ /2 .

Then
F+ − H F− + H
U+ = 2 /2 , U− = ,
1−e −ζ 1 − e−ζ 2 /2
where H is any function, which is analytic between the lines: C+ : η > 0
and C− : η < 0. The solution is given by

1 1 1 e−iζx dζ 1 1 1 e−iζx dζ
Z Z
u(x) = − √ √ 2 + √ √ 2
2π C+ 2π iζ 1 − e−ζ /2 2π C− 2π iζ 1 − e−ζ /2
1 He−iζx dζ
I
+√ 2 .
2π C− −C+ 1 − e−ζ /2

Using the fact the denominator has a second order zero at ζ = 0, we

find
u(x) = −x2 + c0 + c1 x.
138 CHAPTER 3. FOURIER TRANSFORMS

3.36 Solve ∞
1
Z
u(x) − e−2|x−t| u(t)dt = e|x| .
2 −∞

Solution
With
f (x) = e|x| ,
Z ∞
1 1 1
F+ = √ ex(1+iζ) dx = √ ,
2π 0 2π 1 + iζ
Z 0
1 1 1
F− = √ ex(−1+iζ) dx = √ ,
2π −∞ 2π −1 + iζ
Z 
1
Z
−iζx −iζx
u = √ U+ e dζ + U− e dζ .
2π C+ C−
√ √ r
2π −2|x| 2π 2 2 2
k = e , K= 2
= 2 .
2 2 πζ +4 ζ +4
F+ − H F− + H
U+ = 2
, U− = ,
1 − 2/(ζ + 4) 1 − 2/(ζ 2 + 4)
C+ : 1 < η < 2 and C− : −2 < η < −1

1 ζ 2 + 4 (−e−iζx )dζ 1 ζ 2 + 4 (e−iζx )dζ

Z Z
u = 2
+
2π C+ ζ + 2 1 + iζ 2π C− ζ 2 + 2 −1 + iζ
1 ζ2 + 4
Z
+√ 2
H(ζ)e−iζxdζ
2π C− −C+ ζ + 2
√
The last integral has
√ poles at ζ =
√ ±i 2 and using the residue theorem,
it reduces to A cos 2x + B sin 2x. The contributions to the first and
second integrals from these poles are already included in A and B.
Including the contribution from the pole at ±i, we get
√ √
u = 3e|x| + A cos 2x + B sin 2x.
 139

3.37 Obtain all the solutions of the integral equation

Z ∞
λ 2
u(x) − √ e−t /2 u(x − t)dt = x,
2π −∞
with λ being real and positive.

Solution
With
f (x) = x,
Z ∞
1 1 1
F+ = √ xeixζ dx = − √ ,
2π 0 2π ζ 2
Z 0
1 1 1
F− = √ xeixζ dx = √ .
2π −∞ 2π ζ 2
2 2
k = e−x /2 , K = e−ξ /2 .
F+ − H
U+ = ,
1 − λe−ξ2 /2
F− + H
U− = ,
1 − λe−ξ2 /2
The denominator has zeros at
2 /2
p
e−ζ = 1/λ, ζ = ± 2 log λ = ±a.

The solution is written as

1 1 e−iζx dζ 1 1 e−iζx dζ
Z Z
u(x) = − +
2π C+ ζ 2 1 − λe−ξ /2 2π C− ζ 2 1 − λe−ξ2 /2
2

1 H(ζ)e−iζxdζ
Z
√ 2 ,
2π C− −C+ 1 − λe−ξ /2

where C+ is a line above η = 0 and C− is a line below η = 0. Using

the residue theorem, we get
x
u(x) = + A cos ax + B sin ax.
1−λ
140 CHAPTER 3. FOURIER TRANSFORMS

3.38 Solve Z ∞
u(x) − λ e−|x−t| u(t)dt = cos x.
−∞

Solution
Here,
f (x) = cos x,
and
Z ∞ ix −ix
 
1 iζx e + e 1 1 1 1
F+ = √ e dx = − √ + ,
2π 0 2 2π 2i 1 + ζ −1 + ζ
Z 0 ix −ix
 
1 iζx e + e 1 1 1 1
F− = √ e dx = √ + .
2π −∞ 2 2π 2i 1 + ζ −1 + ζ
√ 2
k = 2πe−|x| , K = .
1 + ζ2
Note that K has poles at ζ = ±i, F+ is analytic above η = 0, and
F− is analytic below η = 0. We choose inversion contours, C+ with
0 < η < 1 and C− with −1 < η < 0. The integral equation gives
 
2λ
U+ 1 − = F+ − H,
1 + ζ2
 
2λ
U− 1 − = F− + H.
1 + ζ2

The solution is
1 1 + ζ2 1 1 + ζ2
Z Z
−iζx
u(x) = √ 2 2
F+ e dζ + √ 2 2
F− e−iζx dζ
2π C+ ζ − a 2π C− ζ − a
2
1 H(1 + ζ ) −iζx
Z
√ e dζ,
2π C− −C+ ζ 2 − a2

where a2 = 2λ − 1.
(a) When λ > 1/2, ζ = ±a are real, and the last integral gives
A cos ax + B sin ax. Using the residues
2
u= cos x + A cos ax + B sin ax.
1 − a2
 141

(b) When λ < 1/2, ζ = ±i|a| are imaginary, and the last integral gives
no residues.
2
u= cos x.
1 − a2
(c) When λ = 1/2, ζ = 0, 0 and

u = 2 cos x + A + Bx.
142 CHAPTER 3. FOURIER TRANSFORMS

3.39 Consider the signal

f (t) = 3 cos 3t + 4 sin 2t,

which is modulated to get

g(t) = f (t) cos 100t.

Show that the analytical signal

h(t) = g(t) + iH[g(τ ), τ → t],

allows the factoring

h(t) = f (t)ei100t .

Solution

f (t) = 3 cos 3t + 4 sin 2t,

g(t) = 3 cos 3t cos 100t + 4 sin 2t cos 100t,
3
= [cos 103t + cos 97t] + 2[sin 102t − sin 98t].
2
We have

H[cos ωt] = sin ωt,

H[sin ωt] = − cos ωt.
3
H[g] = [sin 103t + sin 97t] − 2[cos 102t − cos 98t].
2

3 i103t
g + iH[g] = [e + ei97t ] − 2i[ei102t − ei98t ]
2
= ei100t [3 cos 3t + 4 sin 2t] = f (t)ei100t
 143

3.40 Show that the Finite Hilbert transform of (1 − x2 )1/2 is x and that of
(1 − x2 )−1/2 is 0. Use the change of variable:

1 − t2
x= ,
1 + t2

to simplify the integrals.

Solution

√ 1
p
1 1 − ξ2
Z
T1 = T [ 1 − x2 ] = dξ.
π −1 x−ξ
Using
1 − τ2 4τ dτ
ξ= , dξ = − ,
1 + τ2 (1 + τ 2 )2

4τ 2 (x + 1)τ 2 − (1 − x)
1 − ξ2 = , x−ξ = .
(1 + τ 2 )2 1 + τ2
∞
1 8τ 2 dτ
Z
T1 = 2 2
.
π 0 1 + τ (x + 1)τ − (1 − x)
Let
1−x
a2 = , a2 > 0 for − 1 < x < 1.
1+x
∞
8 τ 2 dτ
Z
T1 = .
π(1 + x) 0 (1 + τ 2 )2 (τ 2 − a2 )
Using partial fractions, we have

τ2 A B C
= + + ,
(1 + τ 2 )2 (τ 2 − a2 ) (1 + τ 2 )2 1 + τ 2 τ 2 − a2

where

1 −1 1 a2
A= 2
, B= + 2 , C= .
a +1 a + 1 (a + 1)2
2 (a2 + 1)2
144 CHAPTER 3. FOURIER TRANSFORMS

Using the integrals

Z ∞
dτ π
= ,
0 1 + τ2 2
Z ∞
dτ π
2 2
= ,
0 (1 + τ ) 4
Z ∞
dτ
= 0,
0 τ − a2
2
 
8 A B
T1 = +
π(1 + x) 4 2
 
2 1 2 1
= − +
1 + x a2 + 1 a2 + 1 (a2 + 1)2
2 1 − a2
=
1 + x (a2 + 1)2
2 (1 + x)2 1 + x − 1
= 2
1+x 4 1+x
= x.

Let  
1
T2 = T √ .
1 − x2
Using the change of variable

1 ∞ 1 + τ 2 4τ dτ 1 + τ2
Z
T2 =
π 0 2τ (1 + τ 2 )2 (1 + x)τ 2 − (1 − x)
Z ∞
2 dτ
=
π(1 + x) 0 τ − a2
2

= 0.
 145

3.41 Show that the Finite Hilbert transform of an even function in (-1,1) is
odd and that of an odd function is even.

Solution
Let
1 1 u(ξ)dξ
Z
v(x) =
π −1 x − ξ
Z 0 Z 1 
1 u(ξ)dξ u(ξ)dξ
= +
π −1 x − ξ 0 x−ξ
Z 1 Z 1 
1 u(−ξ) u(ξ)
= + dξ
π 0 x+ξ 0 x−ξ

(a) If u is even, u(−ξ) = u(ξ), and

1 1
Z  
1 1
v(x) = + u(ξ)dξ
π 0 x+ξ x−ξ
2x 1 u(ξ)dξ
Z
= ,
π 0 x2 − ξ 2
which is an odd function.

(b) If u is odd, u(−ξ) = −u(ξ), and 

1 1

1 1
Z
v(x) = − u(ξ)dξ
π 0 x−ξ x+ξ
2 1 ξu(ξ)dξ
Z
= ,
π 0 x2 − ξ 2
which is even.
146 CHAPTER 3. FOURIER TRANSFORMS

3.42 Show that ∗π

1 cos nφdφ sin nθ
Z
=− .
π 0 cos θ − cos φ sin θ
In Eq. (3.295), assuming Fourier series
X X
f (θ) = An cos nθ, g(θ) = Bn sin nθ,
n=0 n=1

show that Bn = −An (n = 1, 2, · · · ). In airfoil theory Bn are known

and An are to be found.

Solution
Let
Z 2π
1 cos nφdφ
I = = Re[J],
2π 0 cos θ − cos φ
Z 2π
1 2einφ dφ
J = , s = eiφ , z = eiθ
2π 0 eiθ + e−iθ − (eiφ + e−iφ )
1 sn ds
I
=
π |s|=1 is[z + z −1 − (s + s−1 )]
1 sn ds
I
= − .
πi (s − z)(s − z −1 )
Using the residue at s = z, we get

1 cos nθ + i sin nθ
J = − .
i sin θ
sin nθ
I = Re[J] = − .
sin θ

For X
f (θ) = An cos nθ,
n=1

π
cos nφ sin θdφ
X Z
g(θ) = An
n=1 0 cos θ − cos φ
X
= − An sin nθ.
n=1
 147

Then,

Bn = An .
148 CHAPTER 3. FOURIER TRANSFORMS
Chapter 4

LAPLACE TRANSFORMS

4.1 Find the Laplace transforms of

f (t) = eat cos bt, g(t) = eat sin bt.

Solution
We have
p p
L[cos bt] = , L[sin bt] = .
p2 + b2 p2 + b2

p−a
L[eat cos bt] = ,
(p − a)2 + b2
b
L[eat sin bt] = .
(p − a)2 + b2

149
150 CHAPTER 4. LAPLACE TRANSFORMS

4.2 Using the relation

∞ √
π −2ab
Z
−(a2 x2 +b2 /x2 )
e dx = e , a, b > 0,
0 2a

obtain the Laplace transforms of

 
1 2 x
f (t) = √ e−x /(4t) , g(t) = erfc √ .
t 2 t

Solution

1 2
f (t) = √ e−x /(4t) .
t
Z ∞
dt
f¯(p) =
2
e−x /(4t)−pt √ .
0 t

Using t = y 2 , dt = 2ydy, we get

Z ∞
f¯ = 2
2 /(4y 2 )−py 2
e−x dy.
0

From
Z ∞ √
2 t2 +b2 /t2 )
e−(a dt = πe−2ab /(2a),
0

we have
√
2 π −2√px/2 π −√px
r
¯
f = √ e = e .
2 p p
Z ∞
2 2
g(t) = √ √
e−y dy
π x/(2 t)
Z ∞Z ∞
2 2
ḡ = √ √
e−y −pt dydt.
π 0 x/(2 t)
 151

We interchange the order of integration in the (t, y)-plane, to have

Z ∞ Z ∞
2 −y 2
ḡ = √ e e−pt dtdy
π 0 x2 /(4y 2 )
Z ∞
2 2 2 2
= √ e−y −px /(4y ) dy
p π 0
√
2 π −√px
= √ e
2 πp
1 −√p x
= e .
p
152 CHAPTER 4. LAPLACE TRANSFORMS

4.3 Find the Laplace transforms of

cos at
f (t) = at , g(t) = √ .
t

Solution

1
f (t) = at = et log a , f¯ = .
p − log a

cos at
g(t) = √ .
t
Z ∞
cos at −pt
ḡ(p) = √ e dt
0 t
Z ∞
2
= cos at2 e−pt dt
Z0 ∞ h i
−(p−ia)t2 −(p+ia)t2
= e +e dt
0
√  
π 1 1
= √ +√ .
2 p − ia p + ia
Let

p + ia p − ia
eiθ = p , e−iθ = p .
2
p +a 2 p2 + a2
√ √
√ cos θ/2 π 1 + cos θ
ḡ(p) = π 2 2 1/4
= 2 2 1/4
√
(p + a ) (p + a ) 2
√
π p
= p [p + p2 + a2 ]1/2 .
2(p + a2 )
2
 153

4.4 Invert the transforms

√
e−2p p
f¯(p) = , ḡ(p) = ,
[(p + 1)2 + 1]2 p−a

Solution

e−2p
f¯(p) = .
[(p + 1)2 + 1]2
1 e−2p+pt dp
Z
f (t) = .
2πi Γ [(p + 1)2 + 1]2
1 epτ dp
Z
= , τ = t − 2,
2πi Γ [(p + 1)2 + 1]2
Z −τ +qτ
1 e dq
= , q = p + 1,
2πi Γ (q + 1)2
2

1 −τ d eqτ dq
Z
= − e
2πi da (q 2 + a) a=1
 √Γ 
d sin aτ
= −e−τ √
da a a=1
√
 
−τ 1 τ √
= −e − 3/2 sin aτ + cos qτ
2a 2a a=1
e−τ
= (sin τ − τ cos τ )
2
1 −(t−2)
= e [sin(t − 2) − (t − 2) cos(t − 2)]h(t − 2).
2

 
1 p 1 a
ḡ = √ = √ 1+
pp−a p p−a
r r
t t
= +a ∗ eat
π π
r Z t
t a √ a(t−τ )
= +√ τe dτ
π π 0
√
r
t 1
= + √ eat erf at.
π 2 a
154 CHAPTER 4. LAPLACE TRANSFORMS

4.5 Show that the Laplace transform of

Z ∞
dτ
f (t) = e−tτ ,
1 τ
is
log(p + 1)
f¯(p) = .
p

Solution

∞
dτ
Z
f (t) = e−tτ ,
τ
Z1 ∞ Z ∞
1 −τ t−pt
f¯(p) = e dτ dt
0 1 τ
Z ∞ ∞
−e−(p+τ )t
= dτ
1 τ (p + τ ) )
Z ∞
dτ
=
τ (p + τ )
Z1 ∞  
1 1 dτ
= −
1 τ p+τ p
1 τ ∞
= log
p p+τ 1
1
= log(p + 1).
p
 155

4.6 Invert the transforms

cosh ap sinh ap
f¯(p) = 3 , ḡ(p) = , 0 < a < 1.
p cosh p p2 sinh p

Solution
1 cosh ap pt
Z
f (t) = e dp.
2πi Γ p3 cosh p
In the complex p-plane, there are poles at p = 0 and at p = ±(n +
1/2)πi = ±n̄πi. For the pole at p = 0, we expand the function in the
form,
cosh ap pt 1
3
e = 3
(1+a2 p2 /2+· · · )(1−p2 /2+· · · )(1+pt+p2 t2 /2+· · · ).
p cosh p p

Res[0] = (a2 + t2 − 1)/2.

The residues at p = n̄πi is

cos n̄πaen̄πti
Res[n̄πi] =
−(n̄π)3 ii sin n̄π
(−1)n
= cos n̄πaen̄πti .
(n̄π)3
The residue at p = −n̄πi is

cos n̄πae−n̄πti
Res[−n̄πi] =
−(n̄π)3 ii sin n̄π
(−1)n
= 3
cos n̄πae−n̄πti .
(n̄π)
Adding the residues
∞
1 2 2
X 2(−1)n
f (t) = (a + t − 1) + 3
cos n̄πa cos π̄t
2 n=1
(n̄π)
1
= [(a + t)2 + (a − t)2 − 2]
4
∞
X (−1)n
+ [cos πn̄(a + t) + cos πn̄(a − t)].
n=1
π 3 n̄3
156 CHAPTER 4. LAPLACE TRANSFORMS

1 sinh ap pt
Z
g(t) = e dp.
2πi Γ p2 sinh p
There are poles at p = 0 and at p = ±nπi. Expanding the function
around p = 0,

sinh ap pt 1
2
e = 2 (1 + a2 p2 /3 + ·)(1 − p2 /3 + · · · )(1 + pt + · · · ).
p sinh p p

Res[0] = t.
1 i sin anπ nπit
Res[nπi] = e .
−n2 π 2 cos nπ
1 −i sin anπ −nπit
Res[−nπi] = e .
−n2 π 2 cos nπ
Adding the residues
∞
X 2(−1)n
g(t) = t + sin anπ sin nπt
n=1
n2 π 2
1
= [(a + t) − (a − t)]
2
∞
X (−1)n
− [cos πn(a + t) − cos πn(a − t)].
n=1
π 2 n2
 157

4.7 Show that the Laplace transform of the Theta function,

∞
X 2 π2 t
θ(t) = e−n ,
n=−∞

can be expressed as √
coth p
θ̄(p) = √ .
p

Solution

∞
X 2 π2 t
θ(t) = e−n ,
n=−∞
∞
X 1
θ̄(p) = .
n=−∞
p + n2 π 2

We may express this as a sum of residues, with

1 cosh z
f (z) = .
p − z 2 sinh z
This function, when integrated along z = x + (N + 1/2)πi with x going
from ∞ to −∞ and returning along z = x − (N + 1/2)πi with two
√
vertical lines at x = ±∞, has poles at z = nπi and at z = ± p. Here,
N will go to infinity. The line integrals are zero. Then
n=∞
X √ √
Res[nπi] + Res[ p] + Res[− p] = 0.
n=−∞

1
Res[nπi] = ,
p2 + n2 π 2
√
√ cosh p 1
Res[ p] = − √ √ ,
sinh p 2 p
√
√ cosh p (−1)
Res[− p] = √ √ ,
sinh p 2 p
√
coth p
θ̄(p) = √ .
p
158 CHAPTER 4. LAPLACE TRANSFORMS

4.8 Using the expansion

1 √
−2 p
√
−2
√ = 1+e
p
+ e−4 p + · · · ,
1−e
invert the transform √
coth p
f¯(p) = √ .
p

Solution

√ √ √
e p + e− p 1 1 + e−2 p
f¯(p) = √ √p √ = √ √
p(e + e− p ) p 1 − e−2 p
1 √ √ √
= √ (1 + e−2 p )(1 + e−2 p + e−4 p + · · · )
p
∞
X √ √ 
= e−2n p + e−2(n+1) p
n=0
∞  
X 1 −4n2 /(4t) 1 −4(n+1)2 /(4t)
f (t) = √ e +√ e
n=0
πt πt
∞
1 X h −n2 /t −(n+1)2 /t
i
= √ e +e
πt n=0
∞
1 X −n2 /t
= √ e .
πt n=−∞
 159

4.9 Assume p = a is a branch point of the transform f¯(p) and near p = a

the transform may be expanded as

f¯(p) = c0 (p − a)ν0 + c1 (p − a)ν1 + c2 (p − a)ν2 + · · · ,

where ν0 < ν1 < ν2 < · · · . Show that, as t → ∞, f (t) has the expansion
 
at c0 c1 c2
f (t) ∼ e + + +··· .
Γ(−ν0 )tν0 +1 Γ(−ν1 )tν1 +1 Γ(−ν2 )tν2 +1

Use the inversion formula

1
L−1 [pν ] = ,
Γ(−ν)tν+1

for this purpose.

Solution
We have
1
L−1 [pν ] = .
Γ(−ν)tν+1

1
Z
f (t) = ept [c0 (p − a)ν0 + c1 (p − a)ν1 + · · · ] dp
2πi Γ
1
Z
= e(pt+at) [c0 pν0 + c1 pν1 + · · · ] dp
2πi Γ
 
at c0 c1
= e + +··· .
Γ(−ν0 )tν0 +1 Γ(−ν1 )tν1 +1
160 CHAPTER 4. LAPLACE TRANSFORMS

4.10 Solve
d2 u du
2
+ 3 + 2u = e2t cos t.
dt dt
with u(0) = du/dt(0) = 0.

Solution
Taking the Laplace transform
p−2
(p2 + 3p + 2)ū = ,
(p− 2)2
+1
p−2 1
ū = .
(p− 2)2
+ 1 (p + 2)(p + 1)
From partial fraction expansion
1 1 1
= − .
(p + 2)(p + 1) p+1 p+2
The solution in convolution form is
Z t
u(t) = e2τ cos τ [e−t+τ − e−2(t−τ ) ]dτ
0
Z t Z t
−t 3τ −2t
= e e cos τ dτ − e e4τ cos τ dτ.
0 0

To evaluate the integrals we use

Z t
I = eaτ cos τ dτ
0
t t
1 aτ 1 1
= e cos τ − 2 eaτ (− sin τ ) − 2 I
a 0 a 0 a
a2 + 1 at
e cos t − 1 e sin t at
I = +
a2 a a2
at
e
I = 2
[a(cos t − 1) + sin t] .
a +1
1 −t 3t 1
u(t) = e e [3(cos t − 1) + sin t] − e−t e4t [4(cos t − 1) + sin t]
10 17
1 2t 1 3t
= e [3(cos t − 1) + sin t] − e [4(cos t − 1) + sin t].
10 17
 161

4.11 Solve
d3 u d2 u du
3
− 6 2
+ 11 − 6u = t,
dt dt dt
with
du d2 u
u(0) = (0) = 2 (0) = 0.
dt dt

Solution
Laplace transform gives
1
(p3 − 6p2 + 11p − 6)ū = .
p2
Noting p = 1 is a root of the characteristic polynomial, the other roots
are p = 2 and p = 3. Then
1 1
ū = 2
p (p − 1)(p − 2)(p − 3)
A B C D E
= 2+ + + + ,
p p p−1 p−2 p−3
where

1 1 1
A = − , C= , D=− ,
6 2 4
1 11
E = , B=− .
18 24
11 1 1 t 1 2t 1
u(t) = − − t + e − e + e3t .
24 6 2 4 18
162 CHAPTER 4. LAPLACE TRANSFORMS

4.12 Solve
d2 u du du
2
+ 2 + 2u = e−t sin t, u(0) = (0) = 0.
dt dt dt

Solution
Laplace transform gives
1
(p2 + 2p + 2)ū = .
(p + 1)2 + 1
1
ū = .
[(p + 1)2 + 1]2
Consider
a
v̄ = , v = e−t sin at.
(p + 1)2 + a2
∂v̄ 1 2a2
= − .
∂a (p + 1)2 + a2 [(p + 1)2 + a2 ]2
 
∂v −t 1 −t 2 −1 1
= ate cos t = e sin at − 2a L .
∂a a [(p + 1)2 + a2 ]2
 
−1 1
u = L
[(p + 1)2 + 12 ]2
1
= [sin t − t cos t]e−t .
2
 163

4.13 Solve the boundary value problem

d2 u du du
2
+ − 2u = e−x , u(0) = 1, (1) = 0.
dx dx dx

Solution
We assume u′ (0) = C and expect to find C using u′ (1) = 0. Then
1
(p2 ū − p − C + pū − 1 − 2ū = .
p+1
1
(p2 + p − 2)ū = 1 + p + C + .
p+1
(1 + p + C)(p + 1) + 1 A B D
ū = = + + .
(p + 1)(p + 2)(p − 1) p+1 p+2 p−1
The coefficients are
1 (C − 1)(−1) + 1 2−C
A=− , B= = ,
2 (−1)(−3) 3

(C + 2)2 + 1 2C + 5
D= = .
(2)(3) 6
1 2 − C −2x 2C + 5 x
u(x) = − e−x + e + e .
2 3 6
1 4 − 2C −2x 2C + 5 x
u′ (x) = e−x − e + e .
2 3 6
As u′ (1) = 0,
1 4 − 2C −2 2C + 5
0 = e−1 − e + e.
2 3 6
8e−2 − 3e−1 − 5e 8 − 3e − 5e3
C= = .
2e + 4e−2 4 + 2e3
164 CHAPTER 4. LAPLACE TRANSFORMS

4.14 Solve the system of differential equations

du dv
− u + v = 0, + v − u = et ,
dt dt
subject to the initial conditions

u(0) = 1, v(0) = 0.

Solution
We have

pū − 1 − ū + v̄ = 1,
1
pv̄ + v̄ − ū = .
p−1
(p − 1)ū + v̄ = 1,
1
−ū + (p + 1)v̄ = .
p−1
Solving for ū and v̄, we get
1 2
ū = , v̄ = 2 .
p−1 p
t
u = e , v = 2t.
 165

4.15 In order to solve

∂ 2 ψ 2 ∂ψ ∂ψ
2
+ = , ψ(a, t) = T0 , ψ(r, 0) = 0,
∂r r ∂r ∂t
we use a substitution
ψ = φ/r.
Obtain the equation for φ. Using the Laplace transform solve for ψ.

Solution
Let
φ ∂ψ φ 1 ∂φ
ψ= , =− 2 + .
r ∂r r r ∂r
Then  
∂ 2 ∂ψ ∂ψ
r = r2
∂r ∂r ∂t
becomes  
∂ ∂φ ∂φ
−φr =r .
∂r ∂r ∂t
∂2φ ∂φ
2
= .
∂r ∂t
Taking Laplace transform,

∂ 2 φ̄ aT0
= pφ̄, φ̄(a, p) = .
∂r 2 p
The solution of this equation vanishing at infinity is
√ aT0 √pa
φ̄ = Ae− pr
, A= e .
p
aT0 −√p(r−a)
φ̄ = e .
p
r−a
φ = aT0 erfc √ .
4t
aT0 r−a
ψ= erfc √ .
r 4t
166 CHAPTER 4. LAPLACE TRANSFORMS

4.16 Obtain the solution of

∂2u ∂2u
= , u(0, t) = U0 sin ωt, u(ℓ, t) = 0,
∂x2 ∂t2
∂u
u(x, 0) = (x, 0) = 0.
∂t

Solution
Laplace transform gives

ū′′ (x, p) = p2 ū(x, p),

ū(x, p) = A sinh px + B sinh p(x − ℓ),
U0 ω
ū(0, p) = −B sinh pℓ = 2 .
p + ω2
ū(ℓ, p) = A sinh pℓ = 0, A = 0.
U0 ω sinh p(ℓ − x)
ū(x, p) = 2 ,
p + ω2 sinh pℓ
U0 ω sinh pa ept dp
Z
u(x, t) = ; a = ℓ − x.
2πi Γ sinh pℓ p2 + ω 2
To the left of the vertical line Γ, there are poles at p = ±iω and at
p = ±inπ/ℓ. Computing the residues,

1 sin ωa eiωt
Res[iω] = ,
ω sin ωℓ 2i
1 sin ωa e−iωt
Res[−iω] = − ,
ω sin ωℓ 2i
1 1 i sin nπa/ℓ inπt/ℓ
Res[inπ/ℓ] = 2 2 2 2
e ,
ω − n π /ell ℓ cos nπ

Adding the residues

( )
sin ωa X (−1)n ωℓ
u(x, t) = U0 sin ωt − 2 ℓ2 − n2 π 2
sin(nπa/ℓ) sin(nπt/ℓ) .
sin ωℓ n=1
ω
 167

4.17 The propagation of elastic stress in a bar: 0 < x < ℓ, is governed by

∂2σ 1 ∂2σ
= ,
∂x2 c2 ∂t2
where c is the wave speed. At time t = 0, the end x = 0 is subjected
to a stress σ0 , while the end x = ℓ is subject to ∂σ/∂x = 0. The initial
conditions are: σ(x, 0) = 0 and ∂σ/∂t(x, 0) = 0. Obtain σ(x, t) by
expanding (1 − e−2pl/c )−1 in a binomial series.

Solution
Laplace transform gives
p2
σ̄ ′′ = σ̄,
c2
which has the solutions

σ̄ = Aepx/c + Be−px/c .

From the boundary conditions

σ0
A+B = ,
p
p  pℓ/c
− Be−pℓ/c = 0.

Ae
c
σ0 e−2pℓ/c σ0 1
A = , B= .
p 1 − e−2pℓ/c p 1 − e−2pℓ/c

σ0 ep(x−2ℓ)/c + e−px/c
σ̄ =
p 1 − e−2pℓ/c
σ0  p(x−2ℓ)/c
+ e−px/c [1 − e−2pℓ/c ]−1

= e
p
σ0  p(x−2ℓ)/c
+ e−px/c 1 + e−2pℓ/c + e−4pℓ/c + · · · .
 
= e
p
      
x 2ℓ − x 2ℓ + x
σ(x, t) = σ0 h t − +h t− +h t− +··· .
c c c
168 CHAPTER 4. LAPLACE TRANSFORMS

4.18 The viscoelastic motion of a semi-infinite bar: 0 < x < ∞, is governed

by
∂2u ∂2u ∂u
c2 2 = 2 + β .
∂x ∂t ∂t
If the boundary and initial conditions are
∂u
u(0, t) = U0 sin ωt, u(x → ∞, t) = 0, u(x, 0) = (x, 0) = 0,
∂x
show that the solution can be written as
Z t
u(x, t) = U0 ω k(x, τ ) cos ω(t − τ )dτ,
0

where, using a branch cut in the p-plane between p = 0 and p = −β,

an expression for k(x, t) can be obtained as
1 β 1 −rt
Z p
k(x, t) = 1 − e sin r(β − r) dr.
π 0 r

Solution
Laplace transform gives
c2 ū′′ = (p2 + βp)ū.
The solution vanishing at infinity has the form
√
2
ū = Ae− p +βpx/c .
From the boundary condition
U0 ω
A= .
p2+ ω2

U0 ω −√p2 +βpx/c
ū = e ,
p2 + ω 2
U0
Z
ω √
− p2 +βpx/c+pt
u(x, t) = e dp,
2πi Γ p2 + ω 2
U0 p ω −√p2 +βpx/c+pt
Z
= e dp,
2πi Γ p2 + ω 2 p
Z t
u(x, t) = U0 ω k(x, τ ) cos ω(t − τ )dτ.
0

where
 169

1 √
p2 +βpx/c+pt dp
Z
−
k(x, t) = e .
2πi Γ p
For manipulating the integral, we create a branch cut along the negative
real axis between (−β) and 0 by setting

p = reiθ , 0 < θ < 2π,

p − β = r1 eiθ1 , 0 < θ1 < 2π.

The inversion integral along Γ can be deformed to go along counter
clockwise on a small circle around the origin, along the upper line along
the branch, along a small circle around p = −β, and, finally, along the
lower line. Then
1 β −rt dr
Z p
k(x, t) = 1 − e sin[ (β − r)rx/c] .
π 0 r
170 CHAPTER 4. LAPLACE TRANSFORMS

4.19 Unsteady heat conduction in a semi-infinite bar is governed by the

equation
∂2u ∂u
κ 2 = ,
∂x ∂t
with the conditions

u(x, 0) = 0, u(0, t) = U0 te−at , u(x → ∞, t) = 0.

Find u(x, t).

Solution
Laplace transform gives
ū′′ = (p/κ)ū.
The solution which vanishes at infinity is
√
− p/κx
ū = Ae .

Let
f (t) = U0 te−at .
Then √
A = f¯, and ū = f¯e− p/κx
.

t
xdτ
Z
2
u(x, t) = U0 (t − τ )e−(t−τ ) e−x /(4κτ ) √
0 4πκτ
Z t
U0 x 2 dτ
= √ (t − τ )e−a(t−τ )−x /(4κτ ) √ .
4πκ 0 τ
 171

4.20 Two semi-infinite bars, A: −∞ < x < 0 and B: 0 < x < ∞, have
thermal diffusivities κA and κB , respectively. Their conductivities are
kA and kB and they are at uniform temperatures, TA0 and TB0 when
t < 0. At time t = 0, their ends are made to contact. Obtain the
transient temperatures TA (x, t) and TB (x, t) for t > 0.

Solution
Taking the Laplace transform with t,

TA0
κA T̄A′′ = pT̄A , T̄A = Aeqx + ,
p
′ T0
κB T̄B′′ = pT̄B , T̄B = Be−q x + B ,
p
where p p
q= p/κA , q′ = p/κB
Matching the temperatures and the heat fluxes at x = 0,

TA0 T0
A+ = B + B,
p p
AkA BkB
√ =√
κA κB
Let
θ = TB0 − TA0 .
√ √
κA kB κB kA
hA = √ √ , hB = √ √ .
κA kB + κB kA κA kB + κB kA
Then
θ T0 θ ′ T0
T̄A = hA eqx + A , T̄B = − hB e−q x + B .
p p p p
Inversion gives
   
x x
TA = TA0 + θhA erfc − √ , TB = TB0 − θhB erfc √ .
2 κA t 2 κB t
172 CHAPTER 4. LAPLACE TRANSFORMS

4.21 A semi-infinite bar is made of two materials: A and B. Material A

occupies 0 < x < 1 and B occupies 1 < x < ∞. Heat conduction in
the two materials is governed by
∂ 2 TA ∂TA ∂ 2 TB ∂TB
κA = , κB = .
∂x2 ∂t ∂x2 ∂t
The boundary and initial conditions are:

TA (0, t) = T0 h(t), TA (x, 0) = TB (x, 0) = 0, TB (x → ∞) = 0,

where h(t) is the Heaviside step function. At the interface x = 1,

∂TA ∂TB
=α , TA = TB ,
∂x ∂x
where a is a constant. Obtain the temperatures in the two sections of
the bar.

Solution
Taking the Laplace transform with t,

κA T̄A′′ = pT̄A , T̄A = A sinh qx + B sinh q(1 − x),

′ ′
κB T̄B′′ = pT̄B , T̄B = Ce−q (x−1) + Deq (x−1)

where p p
q= p/κA , q′ = p/κB
As x → ∞, T̄B = 0. Then, D = 0.
T0 1
T̄A (0, p) = T0 /p, B=
p sinh q
T̄A (1, p) = T̄B (1, p), A sinh q = C
T̄A′ (1, p) = αT̄B′ (1, p), q[A cosh q − B] = −αCq ′
Solving for A, B, and C,
T0 f (q) T0 1 T0 f (q)
A= , B= , C=
p sinh q p sinh q p
where
1
f (q) = ,
cosh q + β sinh q
 173

where p
β=α κA /κB .
Using these
 
T0 f (q) sinh qx sinh q(1 − x)
T̄A = +
p sinh q sinh q
T0 f (q) −q′ (x−1) T0 f (q) −qx̄
T̄B = e = e ,
p p
where r
κA
x̄ = (x − 1).
κB

We may write
2
f (q) =
eq + e−q + β(eq − e−q )
2e−q 1−β
= −2q
, γ= ,
(1 + β) + (1 − β)e 1+β
2 −q
= e [1 − γe−2q + γ 2 e−4q − · · · ],
1+β
eqx − e−qx eq(1−x) − e−q(1−x)
 
T0
T̄A = f (q) q +
p e − e−q eq − e−q
T0  −4q
f (q)(e−q(1−x) − e−q(1+x) ) + e−qx − e−q(2−x) (1 + e−2q+e + · · · )

=
p
 
T0 2 −q(2−x) −qx −2q −qx −q(2−x)
= (e − e )(1 − γe + ···) + e −e (1 + e−2q + · · · )
p 1+β
   
2 2−x x x 2−x
TA (x, t) = T0 erfc √ − erfc √ + · · · + erfc √ − erfc √ ··· ,
1+β 4κA t 4κA 4κA 4κA
 
2T0 x̄ + 1 x̄ + 3
TB (x, t) = erfc √ − γerfc √ +··· .
1+β κA t κA t
174 CHAPTER 4. LAPLACE TRANSFORMS

4.22 If u is the solution of

∂2u ∂u ∂u
2
= , u(0, t) = (0, t) = 0, u(x, 0) = f (x),
∂x ∂t ∂x
and v is the solution of
∂2v ∂2v ∂v ∂v
= 2, v(0, t) = (0, t) = 0, v(x, 0) = 0, (x, 0) = f (x),
∂x2 ∂t ∂x ∂t
show that ∞
1
Z
2 /4t
u(x, t) = √ v(x, τ )e−τ τ dτ.
4πt3 0

(from Snedddon, 1972).

Solution

ū′′ − pū = −f (x), ū(0, p) = ū′ (0, p) = 0,

v̄ ′′ − p2 v̄ = −f (x), v̄(0, p) = v̄ ′ (0, p) = 0,

We may write the complementary solution in the form

√ √
px
ū = Ae + Be− px
.

The particular solution can be found by the method of variable con-

stants. Assuming A and B are functions of x,
√ √
A′ e px + B ′ e− px = 0,
√ √ √
A′ e px − B ′ e− px = −f (x)/ p.
x x
f (ξ) −√pξ f (ξ) √pξ
Z Z
A=− √ e dξ, B= √ e dξ.
0 2 p 0 2 p
The general solution is
Z x
f (ξ)  √p(ξ−x) √
− e− p(ξ−x) dξ

ū = √ e
0 2 p
Z x
f (ξ) √
= √ sinh p(ξ − x)dξ.
0 2 p
 175

Similarly
x
f (ξ)
Z
v̄(x, p) = sinh p(ξ − x)dξ.
0 2p
Note that
√
ū(x, p) = v̄(x, p).
We have
1
Z
u(x, t) = ū(x, p)ept dp
2πi Γ
1 √
Z
= v̄(x, p)ept dp
2πi Γ
Z Z ∞
1 √
= v(x, τ )e− pτ dτ ept dp
2πi
Z ∞ Γ 0
 √ 
= v(x, τ )L−1 e− p dτ
Z0 ∞
τ 2
= v(x, τ ) √ e−τ /(4t) dτ
3
0
Z ∞ 4πt
1 2
u(x, t) = √ v(x, τ )e−τ /(4t) τ dτ.
4πt 03
176 CHAPTER 4. LAPLACE TRANSFORMS

4.23 If u is the solution of

∂2u ∂u ∂u
2
= , u(0, t) = (0, t) = 0, u(x, 0) = f (x),
∂x ∂t ∂x
and v is the solution of
∂2v ∂2v ∂v ∂v
= 2, v(0, t) = (0, t) = 0, v(x, 0) = f (x), (x, 0) = 0,
∂x2 ∂t ∂x ∂t
show that ∞
1
Z
2 /4t
u(x, t) = √ v(x, τ )e−τ τ dτ.
πt 0

(from Snedddon, 1972).

Solution

ū′′ − pū = −f (x), ū(0, p) = ū′ (0, p) = 0,

v̄ ′′ − p2 v̄ = −pf (x), v̄(0, p) = v̄ ′ (0, p) = 0,

We may write the complementary solution in the form

√ √
px
ū = Ae + Be− px
.

The particular solution can be found by the method of variable con-

stants. Assuming A and B are functions of x,
√ √
A′ e px + B ′ e− px = 0,
√ √ √
A′ e px − B ′ e− px = −f (x)/ p.
x x
f (ξ) −√pξ f (ξ) √pξ
Z Z
A=− √ e dξ, B= √ e dξ.
0 2 p 0 2 p
The general solution is
Z x
f (ξ)  √p(ξ−x) √
− e− p(ξ−x) dξ

ū = √ e
0 2 p
Z x
f (ξ) √
= √ sinh p(ξ − x)dξ.
0 2 p
 177

Similarly Z x
v̄(x, p) = f (ξ) sinh p(ξ − x)dξ.
0
Note that
1 √
ū(x, p) = √ v̄(x, p).
p
We have
1
Z
u(x, t) = ū(x, p)ept dp
2πi Γ
1 1 √
Z
= √ v̄(x, p)ept dp
2πi Γ p
Z ∞
1 1
Z
√
= √ v(x, τ )e− pτ dτ ept dp
2πi Γ p 0
Z ∞  −√ p 
−1 e
= v(x, τ )L √ dτ
0 p
Z ∞
1 2
= v(x, τ ) √ e−τ /(4t) τ dτ
0 πt
Z ∞
1 2
u(x, t) = √ v(x, τ )e−τ /(4t) τ dτ.
πt 0
178 CHAPTER 4. LAPLACE TRANSFORMS

4.24 Solve the integral equation

Z t
u(t) − a e−aτ u(t − τ )dτ = e−bt ,
0

where a and b are constants.

Solution
Laplace transform gives
a 1
ū − ū = .
p+a p+b

1p+a 1 a−b
ū = = + .
pp+b p p+b
Inverting
u(t) = 1 + (a − b)e−bt .

4.25 Solve the integral equation

t
√
cos[k t2 − τ 2 ]
Z
√ f (τ )dτ = g(t).
0 t2 − τ 2

Solution
Let
dt′
x2 = x′ , t2 = t′ , dt = √ .
2 t′
f (t) = F (t′ ), g(x) = G(x′ ).
With these, the integral equation becomes
Z x′ √ ′
cos x′ − t′ ′ dt
√ F (t ) √ = G(x′ ).
0 x′ − t′ 2 t′
Taking the Laplace transform, we get
 √   
cos k t F (t)
L √ L √ = L[G].
t 2 t
 179
 √  Z ∞ √ Z ∞
cos k t cos k t −pt 2
L √ = √ e dt = 2 cos kte−pt dt
t 0 t 0
r −k2 /(4p)
πe
= 2 √ .
2 p
 
F (t) 1 √ kr/(4p)
L √ = √ pe L[G].
2 t 2π
Using k → ik,
√ 
√ ek2 /(4p)

cosh k t
L √ = 2π √ .
t p
   √ 
F (t) 2 cosh k t
L √ = pL √ L[G].
2 t π t

Z x′ √
2xd cosh k x′ − t′
F = √ G(t′ )dt′ .
π 2 dx′ 0 ′
x −t ′
Z x √
2d cosh k x2 − t2
f (x) = √ tg(t)dt.
π dx 0 x2 − t2
180 CHAPTER 4. LAPLACE TRANSFORMS

4.26 Solve the integral equation

Z x Z x
u(ξ)dξ f (ξ)dξ
2/3
= 1/3
.
0 (x − ξ) 0 (x − ξ)

Solution
Using

 Γ(α)
L tα−1 = α ,

p
the integral equation becomes

Γ(2/3) f¯
ū = .
Γ(1/3) p1/3

Inversion results in
x
Γ(2/3) f (ξ)dξ
Z
u= .
[Γ(1/3)]2 0 (x − ξ)2/3
 181

4.27 Solve the integral equation

Z x
u(ξ)dξ 1
p =√ .
0 x2 − ξ 2 x

Solution
Let √
2 2 u( η)
x = y, ξ = η, √ = v(η).
2 η
y
v(η) 1
Z
√ dη = 1/4 .
0 y−η y
Taking the Laplace transform, we get

Γ(1/2) Γ(3/4)
√ v̄ =
p p3/4
Γ(3/4) 1
v̄ =
Γ(1/2) p1/4
Γ(3/4) 1
v(y) = 3/4
Γ(1/2)Γ(1/4) y
√ Γ(3/4) 2y 1/2
u( y) =
Γ(1/2)Γ(1/4) y 3/4
2 Γ(3/4) 1
u(x) = √ √ .
π Γ(1/4) x
182 CHAPTER 4. LAPLACE TRANSFORMS

4.28 Find the solution of

un+1 − aun = nbn , u0 = 1.

Solution
We have
1
Z[bn ] = 1 + bz + b2 z 2 + · · · = .
1 − bz
X d 1 bz
Z[nbn ] = b nbn−1 =b = .
n=1
db 1 − bz (1 − bz)2
X b
nbn z n−1 = .
(1 − bz)2
Let
Z = Z[un ].
The given difference equation becomes
1 bz
(Z − 1) − aZ = ,
z (1 − bz)2
bz 2
Z(1 − az) = 1 + ,
(1 − bz)2
1 bz 2
Z = + ,
1 − az (1 − bz)2 (1 − az)
Using partial fractions,

bz 2 A B C
2
= + + ,
(1 − bz) (1 − az) 1 − az 1 − bz (1 − bz)2
b/a2 1/b
A= , C = ,
(1 − b/a)2 1 − a/b
b/a2 1/b
B = − 2
− .
(1 − b/a) 1 − a/b
b b − (a − b) n (n + 1) n
un = an + 2
an − b + b ,
(a − b) (a − b)2 (b − a)b
   
b n a (n + 1) n
un = 1 + a + + b .
(a − b)2 (a − b)2 (b − a)b
 183

4.29 Find the solution of the difference equation

un+2 − 2bun+1 + b2 un = bn , u0 = u1 = 0.

Solution
Let
Z[un ] = Z.
 
1 2b 1
Z 2− + b2 =
z z 1 − bz
z2
Z = ,
(1 − bz)3
1
Zbn = ,
1 − bz
z
Znbn−1 = ,
(1 − bz)2
2z 2
Zn(n − 1)bn−2 = ,
(1 − bz)3
1
un = n(n − 1)bn−2 .
2
184 CHAPTER 4. LAPLACE TRANSFORMS

4.30 Obtain the frequencies ω corresponding to the periodic solutions of

un+2 − (2 − ω 2 )un+1 + un = 0,

subject to the conditions

u0 = u4 = 1, u2 = u5 .

Solution

ω = 2 sin π/N.
When N = 4, √
ω = 2 sin π/4 = 2.
185