Chapter 12.

Partial Differential Equations (PDEs) 1

§12.1 Basic Concepts
[Basic Concepts]
• Partial differential equations(PDEs)
Ex. When u = u(x, y), ux + yuy + uxy = f (x, y).

Ordinary differential equations(ODEs)

Ex. When y = y(x), y 00 + xy = f (x).

• order of the PDE: order of the highest derivative

Ex. ux + yuy + uxy = f (x, y). order 2
Ex.(ODE) y 000 + xy 00 + xy = f (x). order 3

• linear, nonlinear
Ex. uxy + xuxx = f (x, y). linear

Note. u → c1u1 + c2u2

(c1u1 +c2u2)xy +x(c1u1 +c2u2)xx = c1(u1xy +xu1xx)+c2(u2xy +xu2xx).

Ex. uux + yuxy = 0. nonlinear

• homogeneous, nonhomogeneous
Ex. uxy + xuxx = 0. homogeneous
Ex. uxy + xuxx = f (x, y). nonhomogeneous

• solution of a PDE

• boundary condition, initial condition

Ex. When u = u(x, t), utt = uxx
u(a, t) = 0, u(b, t) = 0 boundary condition
u(x, 0) = f (x) initial condition(initial position),
ut(x, 0) = g(x) initial condition(initial velocity)

Computational Science & Engineering (CSE) C. K. Ko

 Chapter 12. Partial Differential Equations (PDEs) 2

Example 1 Second-Order PDEs

∂ 2u 2
2∂ u
• One-dimensional wave equation =c
∂t2 ∂x2
2
∂u 2∂ u
• One-dimensional heat equation =c
∂t ∂x2

∂ 2u ∂ 2u
• Two-dimensional Laplace equation + =0
∂x2 ∂y 2

∂ 2u ∂ 2u
• Two-dimensional Poisson equation + = f (x, y)
∂x2 ∂y 2
µ ¶
∂ 2u 2 ∂ 2u ∂ 2u
• Two-dimensional wave equation = c +
∂t2 ∂x2 ∂y 2
Theorem 1. (Fundamental Theorem on Superposition)
If u1, u2 are solutions of a homogeneous linear PDE in some region R,
then
u = c1u1 + c2u2
is also a solution of that PDE in the region R, where c1, c2 are constants.

Ex. Consider a homogeneous linear PDE uxy + xuxx = 0 in R.

Suppose that u1 and u2 are solutions in R.
(c1u1 + c2u2)xy + x(c1u1 + c2u2)xx
= c1(u1xy + xu1xx) + c2(u2xy + xu2xx)
= c1 · 0 + c2 · 0
= 0.
Therefore u = c1u1 + c2u2 is also a solution of the PDE in R.

Computational Science & Engineering (CSE) C. K. Ko

 Chapter 12. Partial Differential Equations (PDEs) 3

Example 1. Solving a PDE like a ODE

Find solutions u = u(x, y) of the PDE uxx − u = 0 depending on x and y.

Sol. Since no y-derivatives occur, we can solve this PDE like u00 − u = 0.
We have
u = Aex + Be−x
with constants A and B. Hence the solution of the PDE is
u(x, y) = A(y)ex + B(y)e−x.
Example 2. Solving a PDE like a ODE
Find solutions u = u(x, y) of the PDE uxy = −ux.

Sol. Setting ux = p, the PDE is py = −p and

p = c(x)e−y .
By integration w.r.t. x, we obtain
u(x, y) = f (x)e−y + g(y)
here, f (x) and g(y) are arbitrary.

Computational Science & Engineering (CSE) C. K. Ko

 Chapter 12. Partial Differential Equations (PDEs) 4

§12.2 Modeling: Vibrating String, Wave Equation

• Physical Assumptions
1. The mass of the string per unit length is constant. The string is
perfectly elastic and does not offer any resistance to bending.

2. The action of the gravitational force on the string can be neglected.

3. The string performs small transverse motions in a vertical plane.

Figure 1: The string at fixed time t.

Derivation of the PDE of the model “Wave Equation”

Let u(x, t) be its displacements at any point 0 ≤ x ≤ L and at any time
0 < t. Let
• T = horizontal components of the tension(constant),
• ρ = density of the string(mass of the un-deflected string per unit length).
Then u(x, t) satisfies the wave equation

∂ 2u 2
2∂ u T
=c (c2 = ).
∂t2 ∂x2 ρ

Computational Science & Engineering (CSE) C. K. Ko

 Chapter 12. Partial Differential Equations (PDEs) 5

[Derivation of wave equation]

Figure 2: The string at fixed time t.

Let T1 and T2 be the tensions at P and Q, respectively.

The tension is tangential to the curve of the string at each point.
• Horizontal direction.
There is no motion in the horizontal direction. Hence,
T1 cos α = T2 cos β = T = const.
• Vertical direction.
By Newton’s second law (ρ∆x = mass of ∆x), we have
∂ 2u
T2 sin β − T1 sin α = ρ∆x 2 .
∂t
Thus
T2 sin β T1 sin α ρ∆x ∂ 2u
− = tan β − tan α = .
T2 cos β T1 cos α T ∂t2
Note that tan α and tan β are the slope of the string at x and x + ∆x:
³ ∂u ´¯ ³ ∂u ´¯
¯ ¯
tan α = ¯ and tan β = ¯ .
∂x x ∂x x+∆x
Hence ·µ ¶¯ µ ¶¯ ¸
1 ∂u ¯¯ ∂u ¯¯ ρ ∂ 2u
− = .
∆x ∂x ¯x+∆x ∂x ¯x T ∂t2
If ∆x → 0, then

∂ 2u 2
2∂ u 2 T
= c (c = ).
∂t2 ∂x2 ρ
(One-dimensional wave equation)

Computational Science & Engineering (CSE) C. K. Ko

 Chapter 12. Partial Differential Equations (PDEs) 6

§12.3 Solution by Separating Variables.

Use of Fourier Series.
One Dimensional Wave Equation

∂ 2u 2
2∂ u T
=c (c2 = )
∂t2 ∂x2 ρ
boundary condition
u(0, t) = 0, u(L, t) = 0 for all t.
initial condition
u(x, 0) = f (x), 0 ≤ x ≤ L, initial deflection.
ut(x, 0) = g(x), 0 ≤ x ≤ L, initial velocity.

• Step 1. Two ODEs from the wave equation

By the method of separating variables, set u(x, t) = F (x)G(t). Then
00
00 G̈ F
F G̈ = c2F G and 2 = .
cG F
Since the left side depends on only t and the right side only on x, both
sides must be constant:
00
G̈ F
= = k(= constant),
c2G F
and we have two ODEs
00
F − kF = 0,
G̈ − c2kG = 0.
• Step 2. Satisfying the Boundary Conditions
We determine solutions F and G of two ODEs so that u = F G satisfies
the Boundary Conditions
u(0, t) = F (0)G(t) = 0 for all t,
u(L, t) = F (L)G(t) = 0 for all t.
Computational Science & Engineering (CSE) C. K. Ko
 Chapter 12. Partial Differential Equations (PDEs) 7

If G ≡ 0, then u ≡ 0, which is of no interest.

Hence, G 6= 0 and F (0) = F (L) = 0.

Case k = 0. F 00 = 0 and F (x) = Ax + B.

Since F (0) = F (L) = 0, F ≡ 0 and u ≡ 0, which is of no interest.

Case k > 0, k = µ2. F 00 − µ2F = 0 and F (x) = Aeµx + Be−µx.

Since F (0) = F (L) = 0, F ≡ 0 and u ≡ 0, which is of no interest.

Case k < 0, k = −p2. F 00 +p2F = 0 and F (x) = A cos px+B sin px.
F (0) = A = 0, and F (L) = B sin(pL) = 0.
nπ
Thus p = , n = ±1, · · · and setting B = 1, we obtain
L
nπx
Fn(x) = sin , n = 1, 2, · · · .
L

cnπ
The other ODE G̈ + λ2nG = 0, λn = cp = L , has a solution
Gn(t) = Bn cos λnt + Bn∗ sin λnt.

Hence, solutions of the wave equation are

un(x, t) = Fn(x)Gn(t) n = 1, 2, · · ·
nπx
= (Bn cos λnt + Bn∗ sin λnt) sin .
L
Here, un(x, t) are called the eigenfunctions, λn are called the eigenval-
ues, and the set {λ1, λ2, · · · } is called the spectrum.

Computational Science & Engineering (CSE) C. K. Ko

 Chapter 12. Partial Differential Equations (PDEs) 8

Note: Tuning is done by changing the tension T .

s
λn cn T n
frequency = = = of un
2π 2L ρ 2L
The frequency is proportional to the tension.(Why is a violin smaller
than a double-bass?)
• Step 3. Solution of the Entire Problem. Fourier Series.
∞
X
Since the wave equation is linear and homogeneous, un(x, t) is a
n=1
solution.
∞
X
u(x, t) = un(x, t)
n=1
X∞
nπx
= (Bn cos λnt + Bn∗ sin λnt) sin .
n=1
L
From the Initial Conditions
∞
X nπx
u(x, 0) = Bn sin = f (x),
n=1
L
¯ ∞
X
∂u ¯¯ ∗ nπx
and = B λ sin = g(x).
∂t ¯t=0 n=1 n
n
L
Using Fourier Sine Series,
Z
2 L nπx
Bn = f (x) sin dx, n = 1, 2, · · · ,
L 0 L
Z L
2 nπx
Bn∗ λn = g(x) sin dx
L 0 L
Z L
2 nπx
⇒ Bn∗ = g(x) sin dx, n = 1, 2, · · · .
cnπ 0 L
We have obtained a solution of wave equation that satisfies the boundary
conditions and initial conditions.
Computational Science & Engineering (CSE) C. K. Ko
 Chapter 12. Partial Differential Equations (PDEs) 9

Comment. In case g(x) = 0, then Bn∗ = 0 and,

∞
X nπx cnπ
u(x, t) = Bncos λnt sin , λn = .
n
L L
∞ ∞
1X £ nπ ¤ 1X £ nπ ¤
= Bnsin (x − ct) + Bnsin (x + ct)
2 n=1 L 2 n=1 L
1
= [f ∗(x − ct) + f ∗(x + ct)]
2
where f ∗ is the odd periodic extension of f .

Physical interpretation

f ∗(x − ct)(c > 0): a wave that is traveling to the right as t increases.
f ∗(x + ct)(c > 0): a wave that is traveling to the left as t increases.
=⇒ u(x, t) is the superposition of these two waves.

Computational Science & Engineering (CSE) C. K. Ko

 Chapter 12. Partial Differential Equations (PDEs) 10

Example 1.
Find the solution of the wave equation corresponding to the initial deflection

 2k L

 x if 0 < x <
 L 2
f (x) =


 2k (L − x)

if
L
<x<L
L 2
and initial velocity zero.
Sol. Since g(x) = 0, Bn∗ = 0 and
Z
2 L nπx
Bn = f (x) sin dx
L 0 L
Z Z L
2 h L/2 2k nπx 2k nπx i
= x sin dx + (L − x) sin dx
L 0 L L L/2 L L
8k nπ
= 2 2 sin .
nπ 2
Thus,
∞
X nπx cnπ
u(x, t) = Bncos λnt sin , λn =
n=1
L L
∞
8k X sin nπ
2 cnπt nπx
= 2 cos sin
π n=1 n2 L L
8k ¡ 1 cπt πx 1 3cπt 3πx ¢
= 2 2 cos sin − 2 cos sin + ··· .
π 1 L L 3 L L

Computational Science & Engineering (CSE) C. K. Ko

 Chapter 12. Partial Differential Equations (PDEs) 11

Computational Science & Engineering (CSE) C. K. Ko

 Chapter 12. Partial Differential Equations (PDEs) 12

§12.4 D’Alembert’s Solution of the Wave Equation.

Characteristics
D’Alembert’s Solution of Wave Equation
The wave equation
∂ 2u 2
2∂ u T
=c (c2 = )
∂t2 ∂x2 ρ
has a solution of the form
u(x, t) = Φ(x + ct) + Ψ(x − ct).
This is known as d’Alembert’s solution of the wave equation.

Proof Set v = x + ct, w = x − ct(Why?). By the Chain Rule

ux = uv vx + uw wx = uv + uw ,
ut = uv vt + uw wt = c(uv − uw )
and
uxx = (uv + uw )x = (uv + uw )v vx + (uv + uw )w wx
= uvv + 2uvw + uww ,
utt = c(uv − uw )t = c(uv − uw )v vt + c(uv − uw )w wt
= c2(uvv − 2uvw + uww ).
Inserting these two results in the wave equation, we get
∂ 2u
uvw ≡ = 0.
∂w∂v
This gives
Z
∂u
= h(v) and u = h(v)dv + Ψ(w) = Φ(v) + Ψ(w),
∂v
and so
u(x, t) = Φ(x + ct) + Ψ(x − ct).
=⇒ d’Alembert’s solution of the wave equation.
Computational Science & Engineering (CSE) C. K. Ko
 Chapter 12. Partial Differential Equations (PDEs) 13

D’Alembert’s Solution satisfying Initial Conditions

D’Alembert’s Solution satisfying initial conditions
(a) u(x, 0) = f (x), (b) ut(x, 0) = g(x).
can be represented by the form
Z x+ct
1 1
u(x, t) = [f (x + ct) + f (x − ct)] + g(s)ds.
2 2c x−ct

Proof By differentiating u(x, t) = Φ(x + ct) + Ψ(x − ct), we have

ut(x, t) = cΦ0(x + ct) − cΨ0(x − ct).
By the initial condition,
u(x, 0) = Φ(x) + Ψ(x) = f (x),
ut(x, 0) = cΦ0(x) − cΨ0(x) = g(x)
and by integrating the second equation, we obtain
Z
1 x
Φ(x) − Ψ(x) = k(x0) + g(s)ds,
c x0
where k(x0) = Φ(x0) − Ψ(x0).
Addition and subtraction give
Z
1 1 x 1
Φ(x) = f (x) + g(s)ds + k(x0),
2 2c x0 2
Z x
1 1 1
Ψ(x) = f (x) − g(s)ds − k(x0).
2 2 x0 2
Therefore
Z x+ct
1 1
u(x, t) = [f (x + ct) + f (x − ct)] + g(s)ds.
2 2c x−ct

When g = 0, this solution reduces to that of Section 12.3.

Computational Science & Engineering (CSE) C. K. Ko
 Chapter 12. Partial Differential Equations (PDEs) 14

Types of PDEs
The PDE of the form
Auxx + 2Buxy + Cuyy = F (x, y, u, ux, uy ) (1)
has three Types of PDEs, depending on the discriminant AC −B 2 as follows.
Type Defining Condition Example in Sec.12.1
Hyperbolic AC − B 2 < 0 Wave equation
Parabolic AC − B 2 = 0 Heat equation
Elliptic AC − B 2 > 0 Laplace equation
Note Ax2 + 2Bxy + Cy 2 + Dx + Ey + F = 0.
Ex. x2 − y 2 = 1: hyperbolic, x2 − y = 0: parabolic, x2 + 2y 2 = 1: elliptic

Ex. utt − c2uxx = 0

Set y = ct. utt − c2uxx = c2(uyy − uxx) = 0.
Type of the wave equation is hyperbolic.

Characteristics. Normal Forms of PDEs

The normal forms of (1) and the corresponding transformation are obtained
by solving the characteristic equation
dy
Ay 02 − 2By 0 + C = 0 (y 0 = ).
dx
Its solutions are called the characteristics, and we write them in the form
Φ(x, y) = const, Ψ(x, y) = const. The transforms gives new variables
v, w instead of x, y.

Type New Variables Normal Form

Hyperbolic v = Φ, w=Ψ uvw = F1
Parabolic v = x, w=Φ=Ψ uvv = F2
1 1
Elliptic v = 2 (Φ + Ψ), w = 2i (Φ − Ψ) uvv + uww = F3

Computational Science & Engineering (CSE) C. K. Ko

 Chapter 12. Partial Differential Equations (PDEs) 15

Example 1. D’Alembert’s solution obtained systematically

Solve utt − c2uxx = 0.

Sol. Set y = ct, and then utt − c2uxx = c2(uyy − uxx) = 0.

The characteristic equation (y 0)2 − 1 = 0, y 0 = −1, 1 has two families of
solutions(characteristics):
Φ(x, y) = y + x = const,
Ψ(x, y) = y − x = const.
Notice that the wave equation is a hyperbolic type. This gives new variables
v = Φ = y + x = ct + x,
w = Ψ = y − x = ct − x,
and the normal form
uvw = 0.
Thus d’Alembert’s solution of the wave equation is
u(x, t) = f1(ct + x) + f2(ct − x).

Computational Science & Engineering (CSE) C. K. Ko

 Chapter 12. Partial Differential Equations (PDEs) 16

[Optional]
¯ Ax2 + 2Bxy + Cy 2 + Dx + Ey + F = 0.
µ ¶µ ¶
¡ ¢ A B x
Ax2 + 2Bxy + Cy 2 = x y .
B C y
¯ ¯
¯ A−λ B ¯
¯ ¯ = λ2 − (A + C)λ + (AC − B 2) = 0
¯B C −λ¯
λ = λ1, λ2 (eigenvalues)

Since ∆ = (A+C)2 −4(AC −B 2) = (A−C)2 +4B 2 ≥ 0, λ1 and λ2 are real.

µ 0¶ µ ¶
x x
Take = U for a suitable matrix U (see ch. 8, ex.6), we get
y0 y
Ax2 + 2Bxy + Cy 2 = λ1x02 + λ2y 02.
Therefore the quadratic form is

hyperbolic if λ1 > 0, λ2 < 0 or λ1 < 0, λ2 > 0(⇔ AC − B 2 < 0),

parabolic if λ1 = 0 or λ2 = 0(⇔ AC − B 2 = 0),

elliptic if λ1 > 0, λ2 > 0 or λ1 < 0, λ2 < 0(⇔ AC − B 2 > 0).

¯ Auxx + 2Buxy + Cuyy = F (x, y, u, ux, u) is

hyperbolic if AC − B 2 < 0,

parabolic if AC − B 2 = 0,

elliptic if AC − B 2 > 0.

Note. xuxx + yuxy = 0: mixed type

Computational Science & Engineering (CSE) C. K. Ko

 Chapter 12. Partial Differential Equations (PDEs) 17

¯ Auxx + 2Buxy + Cuyy = 0 (1).

Set v = px + qy and w = rx + sy. Then (1) reduces to

(Ap2 + 2Bpq + Cq 2)uvv + (Ar2 + 2Brs + Cs2)uww

+2[Apr + Csq + B(rq + sp)]uvw = 0. (2)

Consider Aλ2 + 2Bλ + C = 0 and let its roots be λ1 and λ2.

Case 1. AC −B 2 < 0(Hyperbolic) Then λ1 and λ2 are distinct real roots.

Let p = λ1, q = 1, r = λ2, s = 1. That is,

v = y + λ1x, w = y + λ2x.
Then
Ap2 + 2Bpq + Cq 2 = Aλ21 + 2Bλ1 + C = 0,
Ar2 + 2Brs + Cs2 = Aλ22 + 2Bλ2 + C = 0,
Apr + Csq + B(rq + sp) = Aλ1λ2 + C + B(λ1 + λ2)
C 2B AC − B 2
= A + C + B(− ) = 2 <0
A A A
and (1) reduces
uvw = 0. (normal form)
The solution of the PDE is
u(x, y) = Φ(v) + Ψ(w) = Φ(y + λ1x) + Ψ(y + λ1x).
Ex. uxx + 3uxy + 2uyy = 0
Sol. λ2 + 3λ + 2 = 0 λ = −2, − 1.
The PDE is hyperbolic and the normal form is uvw = 0. The solution is
u(x, y) = Φ(y − 2x) + Ψ(y − x).

Computational Science & Engineering (CSE) C. K. Ko

 Chapter 12. Partial Differential Equations (PDEs) 18

Case 2. AC − B 2 > 0(Elliptic) Then λ1 = a + bi, λ2 = a − bi, a, b ∈ R.

Let p = λ1 = a + bi, r = λ2 = a − bi, q = s = 1. That is,

v = y + λ1x = y + (a + bi)x, w = y + λ2x = y + (a − bi)x.
Then (1) reduces to
uvw = 0
and solution of the PDE is
u(x, y) = Φ(v) + Ψ(w) = Φ(y + λ1x) + Ψ(y + λ1x).
Ex. uxx − 2uxy + 2uyy = 0

Sol. λ2 − 2λ + 2 = 0 λ = −1 + i, 1 − i.
The PDE is elliptic and reduces to uvw = 0. The solution is
u(x, y) = Φ(y + (1 + i)x) + Ψ(y + (1 − i)x).

We choose
v+w v−w
v ⇒ = y + ax, w ⇒ = bx.
2 2i
Let v = y + ax, w = bx. That is, p = a, q = 1, r = b, s = 0.

Then (2) reduces to

(Aa2 + 2Ba + C)uvv + (Ab2)uww + 2[Aab + Bb]uvw = 0.

Notice that λ1 + λ2 = 2a = − 2B 2 2 C
A , λ1 λ2 = a + b = A .
Then
Aa2 + 2Ba + C = Aa2 − 2Aa2 + A(a2 + b2) = Ab2,
Aab + Bb = (Aa + B)b = 0.
Thus the normal form is uvv + uww = 0.
Computational Science & Engineering (CSE) C. K. Ko
 Chapter 12. Partial Differential Equations (PDEs) 19

Case 3. AC − B 2 = 0(Hyperbola) Then λ is a double root.

Let p = p, q = q, r = λ, s = 1. That is,

v = qy + px, w = y + λx.
Then (2) reduces to
(Ap2 + 2Bpq + Cq 2)uvv + (Aλ2 + 2Bλ + C)uww

+2[Apλ + Cq + B(λq + p)]uvw = 0

Notice that 2λ = − 2B 2 C
A , λ = A.
Then
Aλ2 + 2Bλ + C = 0,
Apλ + Cq + B(λq + p) = p(Aλ + B) + q(C + Bλ)
= p · 0 + q(Aλ2 − (Aλ)λ) = 0.
We choose p = 1, q = 0, that is,
v = x, w = y + λx.
The the normal form is
uvv = 0
and the solution is
u(x, y) = vΦ(w) + Ψ(w) = xΦ(y + λx) + Ψ(y + λx).

Ex. uxx + 4uxy + 4uyy = 0

Sol. λ2 + 4λ + 4 = 0 λ = −2.
The PDE is parabolic and the normal form is uvv = 0. The solution is
u(x, y) = xΦ(y − 2x) + Ψ(y − 2x).

Computational Science & Engineering (CSE) C. K. Ko

 Chapter 12. Partial Differential Equations (PDEs) 20

§12.5 Heat Equation: Solution by Fourier Series

One Dimensional Heat Equation

.
2
∂u ∂ u K
= c2 2 (c2 = )
∂t ∂x σρ

where c2 = thermal diffusivity, K = thermal conductivity,

ρ = density, σ = specific heat.
boundary condition u(0, t) = 0, u(L, t) = 0 for all t.

initial condition u(x, 0) = f (x), 0 ≤ x ≤ L, (initial temperature)

Notice that f (0) = f (L) = 0 because of the boundary conditions.

• Step 1. Two ODEs from the heat equation

By the method of separating variables, set u(x, t) = F (x)G(t). Then
00
00 Ġ F
F Ġ = c2F G and 2 = .
cG F
Since the left side depends on only t and the right side only on x,
00
Ġ F
= = k,
c2G F
and we have two ODEs
00
F − kF = 0, Ġ − c2kG = 0.

Computational Science & Engineering (CSE) C. K. Ko

 Chapter 12. Partial Differential Equations (PDEs) 21

• Step 2. Satisfying the Boundary Conditions

From the Boundary Conditions
u(0, t) = F (0)G(t) = 0 for all t
u(L, t) = F (L)G(t) = 0 for all t.
If G ≡ 0, then u ≡ 0, which is of no interest.
Hence, G 6= 0 and F (0) = F (L) = 0.

Case k = 0. F 00 = 0 and F (x) = Ax + B. Since F (0) = F (L) = 0,

F ≡ 0 and u ≡ 0, which is of no interest.

Case k > 0, k = µ2. F 00 − µ2F = 0 and F (x) = Aeµx + Be−µx.

Since F (0) = F (L) = 0, F ≡ 0 and u ≡ 0, which is of no interest.

Case k < 0, k = −p2. F 00 +p2F = 0 and F (x) = A cos px+B sin px.
F (0) = A = 0, and F (L) = B sin(pL) = 0.
nπ
Thus p = , n = ±1, · · · and setting B = 1, we obtain
L
nπx
Fn(x) = sin , n = 1, 2, · · · .
L
cnπ
The other ODE Ġ + λ2nG = 0, λn = cp = L , has a solution
2
Gn(t) = Bne−λnt.

Hence, solutions of the heat equation are

un(x, t) = Fn(x)Gn(t)
nπx −λ2nt
= Bn sin e .
L
Here, un(x, t) are called the eigenfunctions, λn are called the eigenval-
ues, and the set {λ1, λ2, · · · } is called the spectrum.

Computational Science & Engineering (CSE) C. K. Ko

 Chapter 12. Partial Differential Equations (PDEs) 22

• Step 3. Solution of the Entire Problem. Fourier Series.

∞
X
Since the heat equation is linear and homogeneous, un(x, t) is a
n=1
solution.
∞
X ∞
X nπx −λ2nt
u(x, t) = un(x, t) = Bn sin e .
n=1 n=1
L

From the Initial Conditions

∞
X nπx
u(x, 0) = Bn sin = f (x),
n=1
L

and using Fourier sine series

R
2 L
Bn = L 0 f (x) sin nπx
L dx, n = 1, 2, ...

We have obtained a solution of heat equation that satisfies the boundary

conditions and initial condition.

Computational Science & Engineering (CSE) C. K. Ko

 Chapter 12. Partial Differential Equations (PDEs) 23

Example 1. (Sinusoidal initial temperature)

(a) Find the temperature u(x, t) in a laterally insulated copper bar 80cm
long if the initial temperature is 100 sin(πx/80)◦C and the ends are
kept at 0◦C. (Use c2 = 1.158[cm2/sec].)
(b) How long will it take for the maximum temperature in the bar to drop
to 50◦C?

Sol. (a)
P∞ nπx −λn t 2
u(x, t) = n=1 Bn sin L e , λn = cnπ/80.
From the Initial Conditions
∞
X nπ πx
u(x, 0) = Bn sin x = 100 sin ,
n=1
80 80
2 2
So B1 = 100, B2 = B3 = · · · = 0 and λ21 = c80π2 = 0.001785[sec−1], the
solution is
πx
u(x, t) = 100 sin e−0.001785t.
80
−0.001785t
(b) When 100e = 50, t = 388sec.

Computational Science & Engineering (CSE) C. K. Ko

 Chapter 12. Partial Differential Equations (PDEs) 24

Example 4. (Bar with insulated ends. Eigenvalue 0.)

Find a solution formula of heat equation with the initial condition u(x, 0) =
f (x), and the boundary condition replaced by the condition that both ends
of the bar are insulated. That is,
2
∂u 2∂ u
=c
∂t ∂x2
boundary condition
ux(0, t) = 0, ux(L, t) = 0 for all t
initial condition
u(x, 0) = f (x), 0 ≤ x ≤ L, initial temperature.

Sol.
• Step 1. Two ODEs from the heat equation
By the method of separating variables, set u(x, t) = F (x)G(t). Then
00
Ġ F
= = k,
c2G F
and we have two ODEs
00
F − kF = 0, Ġ − c2kG = 0.

• Step 2. Satisfying the Boundary Conditions

From the Boundary Conditions
ux(0, t) = F 0(0)G(t) = 0 for all t,
ux(L, t) = F 0(L)G(t) = 0 for all t.
If G ≡ 0, then u ≡ 0, which is of no interest.
Hence, G 6= 0 and F 0(0) = F 0(L) = 0.

Computational Science & Engineering (CSE) C. K. Ko

 Chapter 12. Partial Differential Equations (PDEs) 25

Case k > 0, k = µ2. F 00 − µ2F = 0 and F (x) = Aeµx + Be−µx.

Since F 0(0) = F 0(L) = 0, F ≡ 0 and u ≡ 0, which is of no interest.
Case k = 0. F 00 = 0 ⇒ F (x) = A + Bx, F 0(0) = F 0(L) = −B = 0,
and setting A = 1, F (x) = F0(x) = 1.
Case k < 0, k = −p2(p > 0).
F 00 + p2F = 0 ⇒ F (x) = A cos px + B sin px.
F 0(0) = Bp = 0, and F 0(L) = −Ap sin(pL) = 0.
nπ
p= L, n = 1, 2, · · · and setting A = 1, we obtain
nπ
F (x) = Fn(x) = cos x, n = 1, 2, · · · .
L
The other ODE Ġ + λ2nG = 0, λn = cp = cnπ L , has a solution
2
Gn(t) = Ane−λnt, n = 0, 1, · · · .
Hence, solutions of the heat equation are
nπx −λ2nt
un(x, t) = Fn(x)Gn(t) = An cos e , n = 0, 1, · · · .
L
• Step 3. Solution of the Entire Problem. Fourier Series.
∞
X
Since the heat equation is linear and homogeneous, un(x, t) is a
n=0
solution.
∞
X ∞
X nπx −λ2nt
u(x, t) = un(x, t) = A0 + An cos e .
n=0 n=1
L
From the Initial Conditions
∞
X nπ
u(x, 0) = A0 + An cos x = f (x),
n=1
L
using Fourier cosine series
RL RL
A0 = L1 0 f (x)dx, An = 2
L 0 f (x) cos nπx
L dx.

We have obtained a solution of heat equation that satisfies the boundary

conditions and initial condition.
Computational Science & Engineering (CSE) C. K. Ko
 Chapter 12. Partial Differential Equations (PDEs) 26

[Steady 2-Dim. Heat Problems, Laplace’s Equation]

Consider two-dimensional heat equation
∂u 2 2
¡ 2
2 ∂ u ∂ 2u ¢
=c ∇ u=c +
∂t ∂x2 ∂y 2
for steady (i.e., time independent) problems. Then ∂u/∂t = 0, and heat
equation reduces to Laplace’s equation
2 ∂ 2u ∂ 2u
∇ u = 2 + 2 = 0.
∂x ∂y
[Boundary Value Problem]
2∂ 2u ∂ 2u
∇ u = 2 + 2 = 0.
∂x ∂y
Dirichlet boundary condition

Figure 1: Rectangle R and given boundary values

u(0, y) = u(a, y) = 0 for all 0 ≤ y ≤ b,

u(x, 0) = 0, u(x, b) = f (x) for all 0 ≤ x ≤ a.
Sol. Substituting u(x, y) = F (x)G(y) into Laplace’s Equation,
1 d2 F 2
F dx2 = − G1 ddyG2 = −k.
and we have two ODEs
d2F
+ kF = 0,
dx2
d2G
− kG = 0.
dy 2

Computational Science & Engineering (CSE) C. K. Ko

 Chapter 12. Partial Differential Equations (PDEs) 27

From the Boundary Conditions

u(0, y) = F (0)G(y) = 0 for all 0 ≤ y ≤ b,
u(a, y) = F (a)G(y) = 0 for all 0 ≤ y ≤ b.
If G ≡ 0, then u ≡ 0, which is of no interest.
Hence, G 6= 0 and F (0) = F (a) = 0.

Case k = 0. F 00 = 0 and F (x) = A + Bx. Since F (0) = F (a) = 0,

F ≡ 0 and u ≡ 0, which is of no interest.

Case k < 0, k = −µ2. F 00 − µ2F = 0 and F (x) = Aeµx + Be−µx. Since

F (0) = F (a) = 0, F ≡ 0 and u ≡ 0, which is of no interest.

Case k > 0, k = p2(p > 0).

F 00 + p2F = 0 ⇒ F (x) = A cos px + B sin px.
F (0) = A = 0, and F (a) = B sin(pa) = 0.
p = nπ a , n = 1, 2, · · · and setting B = 1, we obtain
nπx
F (x) = Fn(x) = sin , n = 1, 2, · · · .
a
d2 G
The other ODE dy 2
− ( nπ 2
a ) G = 0 has a solution
nπy
− nπy
Gn(y) = Ane a + Bne a , n = 1, 2, · · · .
From the Boundary Conditions
u(x, 0) = F (x)G(0) = 0 for all 0 ≤ x ≤ a, ⇒ G(0) = 0.
Thus Gn(0) = An + Bn = 0, Bn = −An and
nπy nπy nπy
Gn(y) = An(e a − e− a ) = A∗n sinh .
a

Computational Science & Engineering (CSE) C. K. Ko

 Chapter 12. Partial Differential Equations (PDEs) 28

Hence, solutions of the Laplace’s equation are

nπx nπy
un(x, y) = Fn(x)Gn(y) = A∗n sin sinh .
a a
X ∞
Since the Laplace’s equation is linear and homogeneous, un(x, y) is a
n=1
solution.
∞
X ∞
X nπx nπy
u(x, y) = un(x, y) = A∗n sin sinh .
n=1 n=1
a a
From the boundary condition,
∞
X nπb nπx
u(x, b) = A∗n sinh sin = f (x),
n=1
a a

using the Fourier sine series

Z a
nπb 2 nπx
A∗n sinh = f (x) sin dx.
a a 0 a
So, the solution of our problem u(x, y) is
∞
X nπx nπy
u(x, y) = A∗n sin
sinh
n=1
a a
Z a
2 nπx
where A∗n = f (x) sin dx.
a sinh(nπb/a) 0 a

Computational Science & Engineering (CSE) C. K. Ko

 Chapter 12. Partial Differential Equations (PDEs) 29

§12.7 Modeling : Membrane,

Two-Dimensional Wave Equation

Physical Assumptions
1. The mass of the membrane per unit area is constant. The membrane is
perfectly flexible and offers no resistance to bearing.

2. The tension per unit length T is the same at all points and in all directions
and does not change during the motion.

3. The deflection u(x,y,t) of the membrane is small compared to the size of

the membrane, and all angles of inclination are small.

Computational Science & Engineering (CSE) C. K. Ko

 Chapter 12. Partial Differential Equations (PDEs) 30

Modeling, PDE of the model “Two Dimensional Wave Equation”

Let u(x, y, t) be its displacements at any point (x, y) of the membrane at
time t. Let
T = the tension per unit length,
ρ = the density of the membrane.
Then u satisfies
µ 2 ¶
∂ 2u 2 ∂ u ∂ 2
u 2 T
= c + (c = ).
∂t2 ∂x2 ∂y 2 ρ

[Derivation of the PDE]

• The forces acting on the sides of the portion are approximately T ∆x

and T ∆y.
• These forces are tangent to the moving membrane at every instant.

Horizontal Components of the Forces.

• horizontal components of the forces
= (cosine of of inclination)× forces
• Since the angles of inclination are small(cos α, cos β ≈ 1), the horizontal
components at opposite sides are approximately equal, thus the motion
in a horizontal direction will be negligibly small.

Computational Science & Engineering (CSE) C. K. Ko

 Chapter 12. Partial Differential Equations (PDEs) 31

Vertical Components of the Forces.

T ∆y(sin β − sin α) ≈ T ∆y(tan β − tan α) (∵ angles are small)
= T ∆y[ux(x + ∆x, y1) − ux(x, y2)]
The other sides, similarly,
T ∆x[uy (x1, y + ∆y) − uy (x2, y)]
Using Newton’s Second Law
∂ 2u
ρ∆x∆y 2 = T ∆y[ux(x + ∆x, y1) − ux(x, y2)]
∂t
+T ∆x[uy (x1, y + ∆y) − uy (x2, y)]
and
∂ 2u T £ ux(x + ∆x, y1) − ux(x, y2) uy (x1, y + ∆y) − uy (x2, y) ¤
= + .
∂t2 ρ ∆x ∆y
If ∆x, ∆y −→ 0, then
µ ¶
∂ 2u 2 ∂ 2u ∂ 2u T
= c + c2 = .
∂t2 ∂x2 ∂y 2 ρ
This PDE is called the two-dimensional wave equation.

Computational Science & Engineering (CSE) C. K. Ko