DISCLAIMER: These lecture notes consist of a compilation of material

taken from numerous sources, and so should not be taken as original text.
Note that there could be some typos in the text.

1
MAT 5122-PARTIAL DIFFERENTIAL EQUATIONS (PDEs)
TOPIC 4: The Wave Equation

0.1 Introduction

The wave equation

utt = c2 O2 u
models the vibrations of a string in one dimension (i.e., u = u(x, t)); vibrations of a thin
membrane in two dimensions (i.e., u = u(x, y, t)); or the pressure vibrations of an acoustic
wave in air (i.e., u = u(x, y, z, t)). The constant c gives the speed of propagation for the
vibrations (or the wave speed). The Laplacian of u,
∂ 2u ∂ 2u ∂ 2u
O2 u = + +
∂x2 ∂y 2 ∂z 2
can be understood as a measure of how much a function u = u(x, y, z) differs at one point
(x, y, z) from its neighbouring points. So, if O2 u is zero at some point (x, y, z), then u(x, y, z)
is equal to the average value of u at the neighbouring points, say in a small disk around
(x, y, z). If O2 u is positive at the point (x, y, z), then u(x, y, z) is smaller than the average
value of u at the neighbouring points. If O2 u is negative, then u(x, y, z) is larger than the
average value of u at the neighbouring points. The one-dimensional wave equation is given
by
utt (x, t) = c2 uxx (x, t) (1)

1
0.2 Methods of solving the wave equation

0.2.1 D’Alembert Approach

By using the change of variables

ξ = x + ct, η = x − ct,

we have
ut = cuξ − cuη utt = c2 uξξ − 2c2 uξη + c2 uηη
ux = uξ + uη uxx = uξξ + 2uξη + uηη .
Substituting
R in (1), we find uξη = 0 and solving this equationR we find uξ = F (ξ) and
u(ξ, η) = F (ξ)dξ + g(η) or u(ξ, η) = f (ξ) + g(η) where f (ξ) = F (ξ)dξ. Now ξ = x + ct
and η = x − ct, hence, we get D’Alembert’s solution to the one-dimensional wave equation

u(x, t) = f (x + ct) + g(x − ct), (2)

where f and g are arbitrary differentiable functions. The function f (x + ct) represents waves
that are moving to the left at a constant speed c and the function g(x − ct) represents waves
that are moving to the right at the same speed c. The function u(x, t) in (2) involves two
arbitrary functions that are determined by initial conditions.
Example 1. Find the solution to the Cauchy problem
utt = c2 uxx , u(x, 0) = k(x), ut (x, 0) = l(x).

Solution: The condition u(x, 0) is the initial position whereas ut (x, 0) is the initial velocity.
We have u(x, 0) = f (x) + g(x) using (2) and ut (x, 0) = cf 0 (x) − cg 0 (x). Therefore,
f (x) + g(x) = k(x) and cf 0 (x) − cg 0 (x) = l(x).

This means that

Z x
0 0 1 1 1
f (x) − g (x) = l(x) and so f (x) − g(x) = L(x) = l(s)ds.
c c c 0

Adding this to f (x) + g(x) = k(x), we get

     
1 1 1 1 1 1
f (x) = k(x) + L(x) and g(x) = k(x) − k(x) + L(x) = k(x) − L(x) .
2 c 2 c 2 c

2
Hence,    
1 1 1 1
u(x, t) = k(x + ct) + L(x + ct) + k(x − ct) − L(x − ct)
2 c 2 c
  
1 1
u(x, t) = k(x + ct) + k(x − ct) + L(x + ct) − L(x − ct)
2 c
Therefore,
1 x+ct
 Z 
1
u(x, t) = k(x + ct) + k(x − ct) + l(s)ds .
2 c x−ct
Example 2. Solve the initial value problem
utt = 9uxx , u(x, 0) = cos x, ut (x, 0) = 0.

Solution: Here, c = 3, k(x) = cos x, and l(x) = 0. Hence,

1 x+3t
 Z 
1 1
u(x, t) = cos(x + 3t) + cos(x − 3t) + 0ds = [cos(x + 3t) + cos(x − 3t)].
2 c x−3t 2
Simplifying yields
u(x, t) = cos x cos 3t.
Example 3. Find the solution of the initial value problem
utt = c2 uxx , x ∈ R, t > 0, u(x, 0) = sin x, ut (x, 0) = cos x.

Solution: The solution of the one dimensional wave equation with initial data u(x, 0) = f (x)
and ut (x, 0) = g(x), is given by
1 x+ct
 Z 
1
u(x, t) = f (x + ct) + f (x − ct) + g(s)ds .
2 c x−ct
Here, f (x) = sin x and g(x) = cos x, and so the solution becomes:
1 x+ct
  Z
1
u(x, t) = f (x + ct) + f (x − ct) + g(s)ds
2 2c x−ct
1 x+ct
  Z
1
= sin(x + ct) + sin(x − ct) + cos sds
2 2c x−ct
    
1 x + ct + x − ct x + ct − (x − ct) 1
= 2 sin cos + (sin(x + ct) − sin(x − ct))
2 2 2 2c
1
= sin x cos ct + cos x sin ct.
c

3
Therefore, the solution is
1
u(x, t) = sin x cos ct + cos x sin ct.
c
Example 4. Solve the following initial value problem:
utt = c2 uxx u(x, 0) = cos x, ut (x, 0) = e−1 .

The solution is
1 x+ct
  Z
1
u(x, t) = f (x + ct) + f (x − ct) + g(s)ds
2 2c x−ct
1 x+ct −1
  Z
1
= cos(x + ct) + cos(x − ct) + e ds
2 2c x−ct
1 1
= (2 cos x cos ct) + (x + ct − (x − ct))e−1
2 2c
= cos x cos ct + e−1 t.
Therefore, u(x, t) = cos x cos ct + e−1 t.
Theorem 5. (Solution of the Wave equation, n = 1) Assume φ ∈ C 2 (R), ϕ ∈ C 1 (R), and
define u by d’Alembert’s formula
1 x+ct
Z
1
u(x, t) = [φ(x + ct) + φ(x − ct)] + ϕ(s)ds, x ∈ R, t ≥ 0.
2 2c x−ct
Then

ˆ u ∈ C 2 (R × [0, ∞))
ˆ utt − c2 uxx = 0inR × (0, ∞)
ˆ lim(x,t)→(x0 ,0) u(x, t) = φ(x0 ), lim(x,t)→(x0 ,0) ut (x, t) = ϕ(x0 ).

0.2.2 Causality

In general we have:
utt − c2 uxx = 0, −∞ < x < ∞; u(x, 0) = f (x), ut (x, 0) = g(x).

4
 Figure 1: Domain of dependence and range of influence

The effect of an initial position f (x) is a pair of waves travelling in either directions at speed
c and at half the original amplitude. The effect of an initial velocity g(x) is a wave spreading
out at speed ≤ c in both directions. So part of the wave may lag behind (if there is an initial
velocity), but no part goes faster than speed c. The last assertion is called the principle of
causality. To see the physical meaning, let us draw in the space-time diagram a triangle
formed by two characteristic lines passing through the observer at (x, t), as shown in figure
1. The base of the triangle along the initial axis t = 0 begins at x − ct and ends at x + ct. The
solution given by d’Alembert’s solution to the homogeneous wave equation subject to general
Cauchy initial conditions, depends on the initial displacement at just the two corners x − ct
and x + ct, and on the initial velocity only along the segment from x − ct to x + ct. Whatever
is outside the triangle does not matter. Therefore, to the observer at (x, t), the domain of
dependence is the base of the characteristic triangle formed by two characteristics passing
through (x, t). On the other hand, the data at any point x on the initial line t = 0 must
influence all observers in the wedge formed by two characteristics drawn from (x, 0) into the
region of t > 0; this characteristic wedge is called the range of influence.

Next we illustrate the physical effect of initial displacement and velocity separately.

1. Initial displacement only: f (x) 6= 0 and g(x) = 0:

In this case the solution is
1 1
u(x, t) = f (x − ct) + f (x + ct),
2 2

5
 which is shown in figure 2 at successive time steps. The initial disturbance is split into
two equal waves propagating in opposite directions at the speed c. The outgoing waves
preserve the initial profile, although their amplitudes are reduced by half.
2. Initial velocity only: f (x) = 0 and g(x) 6= 0:
Consider the simple example where
g(x) = g0 when |x| < b, and 0 when |x| > b.
We divide the x − t plane into six regions by the characteristics with B and C lying on
the x−axis at x = −b and x = b, respectively. This is shown in figure 3. The solution
in various regions is:
u(x, t) = 0 in the wedge ABE
Z x+ct
1 g0
u(x, t) = g0 ds = (x + ct + b) in the strip EBIF
2c−b 2c
Z x+ct
1
u(x, t) = g0 ds = g0 t in the triangle BCI
2c x−ct
1 b
Z
g0 b
u(x, t) = g0 ds = in the wedge FIG
2c −b c
Z b
1 g0
u(x, t) = g0 ds = (b − x + ct) in the strip GICH
2c x−ct 2c
u(x, t) = 0 in the wedge HCD.
In contrast to (1), disturbance persists for all time in the region between the two wave
fronts.

Example 6. Consider the wave equation utt = uxx with initial data
(
F (x), if |x| ≤ 1
u(x, 0) = f (x) = , ut (x, 0) = g(x) = 0.
0, if |x| > 1.

By d’Alembert’s formula, the solution is

 
1
u(x, t) = f (x + t) + f (x − t) .
2
Now we need to identify where |x + t| and |x − t| are less than one; outside these regions the
solution is zero.

6
Figure 2: Waves due to initial displacement

7
Figure 3: Waves due to initial velocity

8
Case 1:|x+1| ≤ 1 and |x−t| ≤ 1. Here the boundaries are −1 ≤ x−t ≤ 1, −1 ≤ x+t ≤ 1.
On the boundaries dx
dt
= ±1, i.e., they are characteristic lines that give a trajectory moving
at the wave speed. We pick out the pertinent boundary lines:
t ≤ x + 1, t ≤ 1 − x.
These are lines with slopes ±1 and a t − axis intercept at 1. The region R1 is below these
lines, and above the x − axis (t = 0).

Case 2:|x − t| ≤ 1 only. In other words, −1 ≤ x − t ≤ 1, and the boundaries are

t ≤ x + 1, t ≥ x − 1,
represented by R2 .
Case 3:|x + t| ≤ 1 only. In other words −1 ≤ x + t ≤ 1. The boundaries of this region are
t ≤ 1 − x, t ≥ −1 − x,
represented by R3 .

We plot the different regions as shown in figure 4. In region

ˆ R1 , |x + t| ≤ 1 and |x − t| ≤ 1.
ˆ R2 , |x − t| ≤ 1 only.
ˆ R3 , |x + t| ≤ 1 only.
ˆ Outside of these regions, both |x + t| and |x − t| exceed 1.

Thus,   
1
F (x + t) + F (x − t) , if (x, t) ∈ R1




 2

1
u(x, t) = 2 F (x − t), if (x, t) ∈ R2
1
F (x + t), if (x, t) ∈ R3




 2
0, otherwise.

Example 7. Consider the wave equation utt = uxx with initial data
(
G(x) = cos2 ( πx
2
), if |x| ≤ 1,
u(x, 0) = 0, ut (x, 0) = g(x) =
0, if |x| > 1.

9
Figure 4: The different regions when |x ± 1| ≤ 1 for initial displacement only

10
We are to solve for the wave. By d’Alemberts’s formula, the solution is
1 x+t
Z
u(x, t) = g(s)ds.
2 x−t

We need to identify where |x + t| and |x − t| are less than one; outside of these regions the
solution is zero.

1. In region R1 , |x + t| ≤ 1 and |x − t| ≤ 1. This implies that −1 ≤ x − t ≤ x + t ≤ 1,

and
Z x+t Z x+t
sin(π(x + t)) − sin(π(x − t)) 1
G(s)ds = cos2 (πs/2)ds = t+ = t+ cos πx sin πt.
x−t x−t 2π π
Hence, in region 1
1 1
u1 (x, t) = t + cos πx sin πt.
2 2π
2. In region R2 , |x − t| ≤ 1 and x + t > 1, so that x − t ≥ −1 and x + t ≥ 1. Hence,
Z x+t Z 1
1 sin(π(x − t))
G(s)ds = cos2 (πs/2)ds = (1 − x + t) − .
x−t x−t 2 2π
Hence, in region 2,
1 sin(π(x − t))
u2 (x, t) = (1 − x + t) −
4 4π
3. In region R3 , |x + t| ≤ 1 and x − t < −1, so that x − t ≤ −1 and x + t ≤ 1. Hence,
Z x+t Z x+t
G(s)ds = cos2 (πs/2)ds
x−t −1
1 sin(π(x + t))
= (x + t + 1) + .
2 2π
Thus, in region 3,
1 sin(π(x + t))
u3 (x, t) = (x + t + 1) + .
4 4π
4. In region R4 , x − t ≤ −1 and x + t ≥ 1.
Z x+t Z 1
G(s)ds = cos2 (πs/2)ds = 1.
x−t −1

Therefore, in region 4, u4 (x, t) = 21 .

11
Figure 5: The different regions when |x ± t| ≤ 1 for initial velocity only

12
 5. In region R5 , x + t < −1, hence G(x) = 0.
6. In region R6 , x − t > 1, thus G(x) = 0.

Therefore, 
1 1

 2
t + 2π cos πx sin πt, if (x, t) ∈ R1 ,
(1 − x + t) − sin(π(x−t))
 1
, if (x, t) ∈ R2 ,



 4 4π
 1 (x + t + 1) + sin(π(x+t)) ,

if (x, t) ∈ R3 ,
u(x, t) = 41 4π

 2
, if (x, t) ∈ R4 ,

0, if (x, t) ∈ R5





0, if (x, t) ∈ R6 .

0.2.3 Reflections of waves

Let us consider the Dirichlet problem on the half-line (0, ∞). Thus the problem is:

vtt − c2 vxx = 0, 0 < x < ∞, 0 ≤ t < ∞ (3)

v(x, 0) = φ(x), vt (x, 0) = ϕ(x), v(0, t) = 0.

Consider the odd extensions of both of the initial functions to the whole line, φodd (x) and
ϕodd (x), i.e., 
φ(x),
 if x > 0
φodd (x) = −φ(−x), if x < 0

0, if x = 0.
and 
ϕ(x),
 if x > 0
ϕodd (x) = −ϕ(−x), if x < 0

0, if x = 0.
Let u(x, t) be the solution of the initial-value problem on (−∞, ∞) with initial data φodd and
ϕodd . Then u(x, t) is once again an odd function of x (Prove this). Therefore, u(0, t) = 0, so
that the boundary condition is satisfied automatically. Define v(x, t) = u(x, t) for 0 < x < ∞
(the restriction of u to the half-line). Then v(x, t) is precisely the solution we are looking
for. From d’Alembert’s formula we have for x ≥ 0,
1 x+ct
Z
1
v(x, t) = u(x, t) = (φodd (x + ct) + φodd (x − ct)) + ϕodd (s)ds.
2 2c x−ct

13
For x > ct only positive arguments occur in the formula so that u(x, t) is given by the usual
formula:
1 x+ct
Z
1
u(x, t) = (φ(x + ct) + φ(x − ct)) + ϕ(s)ds, x > ct.
2 2c x−ct
However, in the region 0 < x < ct, we have
φodd (x − ct) = −φ(−(x − ct)) = φ(ct − x),
so that
Z x+ct Z 0
1 1 1
v(x, t) = (φ(x + ct) − φ(ct − x)) + ϕ(s)ds + ϕ(s)(−ϕ(−s))ds.
2 2c 0 2c x−ct

Switching the variables s → −s in the last integral we have

Z 0 Z 0
(−ϕ(−s))ds = ϕ(s)ds.
x−ct ct−x

Therefore,
Z ct+x
1 1
v(x, t) = (φ(x + ct) − φ(ct − x)) + ϕ(s)ds, 0 < x < ct.
2 2c ct−x

Therefore, the complete solution is:

( R x+ct
1 1
(φ(x + ct) + φ(x − ct)) + ϕ(s)ds, if x > ct
u(x, t) 21 2c
1
Rx−ct
ct+x
2
(φ(x + ct) − φ(ct − x)) + 2c ct−x
ϕ(s)ds, if 0 < x < ct.

Example 8. Solve the problem

utt − 4uxx = 0, 0 < x < ∞, u(0, t) = 0, u(x, 0) = 1, ut (x, 0) = 0.

Solution:
Here, φ(x) = 1, ϕ(x) = 0, c = 2. Hence,
( R 2t+x
1
2
(1 − 1) + 41 2t−x 0ds, if 0 < x < 2t
u(x, t) = 1 R x+2t .
2
(1 + 1) + 14 x−2t 0ds, if x > 2t.

Therefore, (
0, if 0 < x < 2t
u(x, t) =
1, if x > 2t.

14
0.2.4 The method of separation of variables

We consider the homogeneous Dirichlet conditions for the wave equation

utt − c2 uxx = 0, 0 < x < L u(0, t) = 0 = u(L, t) (4)
u(x, 0) = f (x) ut (x, 0) = g(x).
The method consists of building up the general solution as a linear combination of special
ones that are easy to find. A separated solution is a solution of the boundary value problem
utt − c2 uxx = 0, 0 < x < L u(0, t) = 0 = u(L, t)
of the form
u(x, t) = X(x)T (t). (5)
Plugging the form (5) into the wave equation in (4), we get
T 00 X 00
X(x)T 00 (t) − c2 X 00 (x)T (t) = 0 or X(x)T 00 (t) = c2 X 00 (x)T (t) or = .
c2 T X
Since the expression on the left depends only on t and that on the right only depends on x,
we conclude that both sides must be constant, that is,
T 00 X 00
= = k,
c2 T X
where k is a constant. This leads to two linear ODEs
X 00 − kX = 0 and T 00 − c2 kT = 0.

Satisfying the boundary conditions. Now,

u(0, t) = X(0)T (t) = 0 and u(L, t) = X(L)T (t) = 0, for all t.
If T ≡ 0, then u ≡ 0, which is of no interest. Thus, X(0) = 0 and X(L) = 0, and so we solve
the BVP
X 00 − kX = 0, X(0) = 0, X(L) = 0. (6)
There are three case, to consider on k. Either k < 0, k = 0 or k > 0. For k = 0, the general
solution is X(x) = Ax + B. Applying initial conditions yields X ≡ 0, which is of no interest
because it means that u ≡ 0. For positive k = µ2 , the general solution is X(x) = Aeµx +Be−µx
and applying initial conditions we obtain X ≡ 0 as before, which leads to u ≡ 0. Hence we
are only left with the possibility of choosing k negative, say k = −p2 . Hence the ODE in (6)
takes the form X 00 + p2 X = 0, and general solution is X(x) = A cos px + B sin px. Applying

15
BCs we see that 0 = X(0) = A, and so X(x) = B sin px. Also, 0 = X(L) = B sin pL. Hence
either B = 0 or sin pL = 0. If B = 0, then u ≡ 0, which is of no interest. Thus we must have
sin pL = 0, and so pL = nπ ⇒ p = nπ L
, n ∈ Z. Setting B = 1, we obtain infinitely many
solutions X(x) = Xn (x), where
nπ
Xn (x) = sin x, n = 1, 2, 3, · · · . (7)
L
For negative n, we obtain essentially the same solutions, except for a minus sign, because
sin(−α) = − sin(α). Solving T 00 − c2 kT = 0 with k = −p2 = −( nπ
L
)2 we get
cnπ
T 00 + λ2n T = 0, where λn = .
L
A general solution is Tn (x) = An cos λn t+Bn sin λn t. Hence the functions Un (x, t) = Xn (x)Tn (t),
written out as
nπ
un (x, t) = (An cos λn t + Bn sin λn t) sin x, n = 1, 2, 3, · · · (8)
L
are solutions of the BVP. These functions are called eigenfunctions, or characteristic func-
tions, and the values λn = cnπL
are called the eigenvalues, or characteristic values of the
vibrating spring. The set {λ1 , λ2 , · · · } is called the spectrum. The sum of the solutions is
also a solution, hence,
∞ ∞
X X nπ
u(x, t) = un (x, t) = (An cos λn t + Bn sin λn t) sin x (9)
n=1 n=1
L

is also a solution of the BVP. Satisfying the initial conditions we have:

∞ ∞
X nπ X nπ
u(x, 0) = An sin x = f (x), and ut (x, 0) = Bn sin x = g(x).
n=1
L n=1
L

Hence we must choose A0n s and Bn0 s so that u(x, 0) and ut (x, 0) become the Fourier sine
series of f (x) and g(x), respectively. That is,

2 L
Z
nπ
An = f (x) sin xdx, n = 1, 2, 3, · · · (10)
L 0 L

Z L Z L
2 nπ 2 nπ
Bn λn = g(x) sin xdx, n = 1, 2, 3, · · · or Bn = g(x) sin xdx. (11)
L 0 L cnπ 0 L
Thus, for any initial data of this form, the problem (4) has a simple explicit solution.

16
Example 9. Solve the problem
utt = c2 uxx , 0 < x < 1, u(0, t) = u(1, t) = 0, t>0
u(x, 0) = x(1 − x), ut (x, 0) = 0, 0 < x < 1.

Solution: ∞
X nπ
u(x, t) = (An cos λn t + Bn sin λn t) sin x,
n=1
L
cnπ
where λn = L
. Here, L = 1, and so
∞
X
u(x, t) = (An cos cnπt + Bn sin cnπt) sin nπx,
n=1

where, (
Z 1 8
n3 π 3
, if n is odd
An = 2 x(1 − x) sin nπxdx =
0 0, if n is even.
Bn = 0.
Example 10. Solve the problem
utt = c2 uxx , u(0, t) = u(1, t) = 0, 0 < x < 1, t>0
u(x, 0) = sin(5πx) + 2 sin(7πx), ut (x, 0) = 0, 0 < x < 1.

∞
X nπ
u(x, t) = (An cos λn t + Bn sin λn t) sin x,
n+1
L
cnπ
where λn = L
and L = 1. Hence,
∞
X
u(x, t) = (An cos cnπt + Bn sin cnπt) sin nπx,
n=1

where, 
1 Z 1, if n = 5

An = 2 (sin(5πx) + 2 sin(7πx)) sin nπxdx = 2, if n = 7
0 
6 5, 7.
0, if n =

Therefore,
u(x, t) = cos 5cπt sin 5πx + 2 cos 7cπt sin 7πx.

17
0.3 The wave equation and conservation of energy

Consider an infinite string , so that the displacement u(x, t) is governed by

utt = c2 uxx , −∞ < x < ∞.

The kinetic energy K is given by:

Z∞
1
K= u2t dx
2
−∞

To ensure integral convergence, we assume that f (x) and g(x) vanish outside an interval
{|x| ≤ R}. Differentiating the kinetic energy, we can pass the derivative under the integral
 Z∞  Z ∞
dK 1d
= u2t dx = ut utt dx.
dt 2 dt −∞
−∞

Substituting the PDE utt = c2 uxx , and integrating by parts we have:

Z ∞  ∞ Z ∞
dK 2 2 2
=c ut uxx dx = c ut ux −c utx ux dx.
dt −∞ −∞ −∞

This leads to
∞ ∞
d ∞ 1 2 2
Z Z   Z
dK d 1 2
= −c2 utx ux dx = −c 2
u dx = − c u dx.
dt −∞ −∞ dt 2 x dt −∞ 2 x
We can define the potential energy P as:
Z ∞
1 2 2
P = c u dx.
−∞ 2 x
Then the total energy E is given by
Z∞
1
E= (u2t + c2 u2x )dx.
2
−∞

18
Next we prove that the wave equation ensures conservation of energy in the case of Dirichlet
BCs. We consider the problem (4).
 ZL
1 L d 2
 Z
0 d 1 2 2 2
E (t) = (ut + c ux )dx = (ut + c2 u2x )dx
dt 2 2 0 dt
0
1 L
Z Z L
2
= (2ut utt + 2c ux uxt )dx = (ut utt + c2 ux uxt )dx
2 0 0
Z L Z L
2 ∂ 2 ∂
= (ut utt + c ( (ux ut ) − uxx ut ))dx = c (ux ut )dx
0 ∂x 0 ∂x
= c2 (ux ut )|L0 = c2 ux (L, t)ut (L, t) − c2 ux (0, t)ut (0, t).
From the BCs,
u(0, t) = u(L, t) = 0 ⇒ ut (0, t) = ut (L, t) = 0.
Therefore E 0 (t) = 0, hence, E(t) is constant and so the energy is conserved.
Exercise 11. Prove that the energy is conserved even in the case of Neumann BCs.

0.4 Well posedness of the wave equation

Theorem 12. The cauchy problem below is well-posed in the domain −∞ < x < ∞, 0 ≤
t ≤ L, for f ∈ C 2 (R) and g ∈ c1 (R).
utt − c2 uxx = 0 −∞ < x < ∞, t > 0
u(x, 0) = f (x), ut (x, 0) = g(x), −∞ < x < ∞.

Proof
The existence and uniqueness follow directly from the d’Alembert formula. We have already
shown that any solution of the Cauchy problem is necessarily equal to the d’Alembert’s
solution. Note that from our smoothness assumption, it follows that u ∈ C 2 (R × (0, ∞)) ∩
C 1 (R × [0, ∞)), and therefore the d’Alembert’s solution is a classical solution. On the other
hand, for f ∈ C(R) and g that is locally integrable, the d’Alembert’s solution is a generalized
solution. What is left is to prove the stability of the Cauchy problem, i.e., we need to show
that small changes in the initial conditions give rise to a small change in the solution.

Let ui be two solutions of the Cauchy problem with initial conditions fi , gi , where i = 1, 2.
If
|f1 (x) − f( x)| < δ, |g1 (x) − g2 (x)| < δ, for all x ∈ R,

19
then for all x ∈ R and 0 ≤ t ≤ L, we have
 Z x+ct 
1 1 
|u1 (x, t) − u2 (x, t)| =  [f1 (x + ct) + f1 (x − ct) − f2 (x + ct) − f2 (x − ct)] + (g1 (s) − g2 (s))ds
2 2c x−ct
|f1 (x + ct) − f2 (x + ct)| |f1 (x − ct) − f2 (x − ct)| 1 x+ct
Z
≤ + + |g1 (s) − g2 (s)|ds
2 2 2c x−ct
1 x+ct
Z
δ+δ δ
< + δds = δ + (x + ct − (x − ct))
2 2c x−ct 2c
= δ + δt ≤ δ + δL = δ(1 + L).
ε
Therefore, for a given ε > 0, we take δ = 1+L . Then for all x ∈ R and 0 ≤ t ≤ L, we have
|u1 (x, t) − u2 (x, t)| < ε. Therefore, small changes in initial conditions result in small changes
in the solution.

0.5 The Cauchy problem for the nonhomogeneous wave equation

Consider the following problem

utt − c2 uxx = F (x, t) −∞ < x < ∞, t > 0
u(x, 0) = f (x), ut (x, 0) = g(x), −∞ < x < ∞.
This problem models, for example, the vibration of a very long string in the presence of an
external force F. The solution for the problem is given by
1 x+ct
Z Z Z
1 1
u(x, t) = [f (x + ct) + f (x − ct)] + g(s)ds + f (y, z)dydz (12)
2 2c x−ct 2c 4

where 4 is the characteristic triangle defined by the base [a, b] and edges x + ct = a and
x − ct = b.
Proposition 13. The Cauchy problem
utt − c2 uxx = F (x, t) −∞ < x < ∞, t > 0 (13)
u(x, 0) = f (x), ut (x, 0) = g(x), −∞ < x < ∞
admits at most one solution.

Proof:
Assume that u1 , u2 are solutions of problem (13). We prove that u1 = u2 . The function

20
u = u1 − u2 is a solution of the homogeneous problem
utt − c2 uxx = 0 −∞ < x < ∞, t > 0 (14)
u(x, 0) = 0, ut (x, 0) = 0, −∞ < x < ∞.
On the other hand, v(x, t) = 0 is a solution of the same problem (14). By theorem 12,
u = v = 0, hence u1 = u2 .

Next we derive the solution. Recall Green’s formula for a pair of functions P and Q in a
planar domain Ω with a piecewise smooth boundary Γ :
Z Z I
[Q(x, t)x − P (x, t)t ]dxdt = [P (x, t)dx + Q(x, t)dt].
Ω Γ

Let u(x, t) be a solution of problem (13). Integrate the two sides of the PDE in (13) over
the characteristic 4 with a fixed upper vertex (x0 , t0 ) as shown in figure 6. The three sides
of this triangle will be denoted by B, R and L. We have:
Z Z Z Z
− F (x, t)dxdt = (c2 uxx − utt )dxdt.
4 4

Using Green’s formula with Q = c2 ux and P = ut , we obtain

Z Z I Z Z Z
2
− F (x, t)dxdt = (ut dx + c ux dt) = + + (ut dx + c2 ux dt).
4 Γ B R L

On the base B, we have dt = 0 and so using initial conditions, we get

Z Z x0 +ct0 Z x0 +ct0
2
(ut dx + c ux dt) = ut (x, 0)dx = g(x)dx.
B x0 −ct0 x0 −ct0

On the right edge R, x + ct = x0 + ct0 , hence dx = −cdt, and so

Z Z Z
2
(ut dx+c ux dt = −c (ut dt+ux dx) = −c ds = −c[u(x0 , t0 )−u(x0 +ct0 , 0)] = −c[u(x0 , t0 )−f (x0 +ct0 )].
R R R

Similarly, on the left edge L, x − ct = x0 − ct0 , implying dx = cdt, and

Z Z Z
2
(ut dx + c ux dt) = c (ut dt + ux dx) = c du
L L L

= c[u(x0 − ct0 , 0) − u(x0 , t0 )] = c[f (x0 − ct0 ) − u(x0 , t0 )].

Therefore,
Z Z Z x0 −ct0
− F (x, t)dxdt = g(x)dx + c[f (x0 − ct0 ) + f (x0 + ct0 ) − 2u(x0 , t0 )].
4 x0 −ct0

21
Figure 6: Characteristic triangle

22
Solving for u gives
x0 +ct0
f (x0 + ct0 ) + f (x0 − ct0 )
Z Z Z
1 1
u(x0 , t0 ) = + g(x)dx + F (x, t)dxdt.
2 2c x0 −ct0 2c 4

Therefore the explicit formula at an arbitrary point (x, t) is

f (x + ct) + f (x − ct) 1 x+ct
Z Z Z
1
u(x, t) = + g(s)ds + F (ξ, τ )dξdτ,
2 2c x−ct 2c 4

which is also called the d’Alembert’s formula. Finally we prove that the d’Alembert’s formula
we have derived is indeed a solution to the Cauchy problem (I leave this as an exercise).
Theorem 14. The Cauchy problem (13) in the domain −∞ < x < ∞, 0 ≤ t ≤ T, is
well-posed for F, Fx ∈ C(R), f ∈ C 2 (R), g ∈ C 1 (R).

Proof is an exercise.
Corollary 15. Suppose that f, g are even functions, and for every t ≥ 0 the function F (x, t)
is even too. Then for every t ≥ 0 the solution u(x, t) of the Cauchy problem (13) is also even.
Similarly, the solution is an odd function or periodic function with period L (as a function
of x) if the data are odd functions or periodic functions with period L.
Example 16. Solve the following Cauchy problem.
utt − 9uxx = xx − e−x −∞ < x < ∞, t > 0,
u(x, 0) = x, ut (x, 0) = sin x −∞ < x < ∞.

Solution: Using the d’Alembert formula, we have

1 x+3t 1 t x+3(t−τ ) s
Z Z Z
1
u(x, t) = [x + 3t + x − 3t] + sin sds + (e − e−s )dsdτ
2 6 x−3t 6 0 x−3(t−τ )
1 2 2
= x + sin x sin 3t − sinh x + sinh x cosh 3t.
3 9 9
In the next example we prove the uniqueness of solution to the nonhomogeneous wave equa-
tion using the energy method.
Proposition 17. Prove that the Neumann problem
utt − c2 uxx = F (x, t) 0 < x < L, t > 0, (15)
ux (0, t) = a(t), ux (L, t) = b(t), t ≥ 0,
u(x, 0) = f (x), ut (x, 0) = g(x), 0 ≤ x ≤ L,
has at most one solution.

23
Proof:
Assume that u1 , u2 are solutions of problem (15). We prove that u1 = u2 . The function
w = u1 − u2 is a solution of the homogeneous problem
wtt − c2 wxx = 0 0 < x < L, t > 0,
wx (0, t) = 0, wx (L, t) = 0, t ≥ 0,
w(x, 0) = 0, wt (x, 0) = 0, 0 ≤ x ≤ L.

We prove that w ≡ 0. Define the total energy of the solution w at time t as

1 L 2
Z
E(t) = (wt + c2 wx2 )dx,
2 0
and we have already seen that
E 0 (t) = c2 (wx wt )|L0 = c2 wx (L, t)wt (L, t) − c2 wx (0, t)wt (0, t).

Applying the BCs yields that E 0 (t) = 0, hence, E(t) = constant and the energy is conserved.
Now, since w(x, 0) = 0, it follows that wx (x, 0) = 0. Also, we have wt (x, 0) = 0, and so the
energy at time t = 0 is zero. But we have shown that E(t) is constant, thus, E(t) ≡ 0. Since
e(x, t) = wt2 + c2 wx2 ≥ 0, and since its integral over [0, L] is zero, it follows that wt2 + c2 wx2 ≡ 0,
which implies that wt (x, t) = wx (x, t) ≡ 0. Consequently, w(x, t) ≡ constant. By the initial
conditions w(x, 0) = 0, hence w(x, t) ≡ 0. The proof is complete.

0.5.1 Separation of variables for nonhomogeneous wave equation

The method of separation of variables can be extended to the nonhomogeneous equation,

and the technique is called the method of eigenfunction expansion.
Example 18. Solve the problem
utt − uxx = cos 2πx cos 2πt = 0 < x < 1, t > 0, (16)
ux (0, t) = 0, ux (1, t) = 0 = t ≥ 0
u(x, 0) = cos2 πx ut (x, 0) = 2 cos 2πx 0 ≤ x ≤ 1.

Solution: This is left as an exercise to be discussed during a tutorial.

24
Bibliography

[1] A. Salih, Classification of Partial Differential Equations and Canonical forms, De-
partment of Aerospace Engineering, Indian Institute of Space Science and Technology,
Thiruvananthapuram, 22 December 2014.
[2] Yeduda Pnchover and Jacob Rubinstein, An Introduction to Partial Differential Equa-
tions, New York, 2005.
[3] Ravi P. Agarwal and Donald O’Regan, Differential Equations, with Special Functions,
Fourier Series and Boundary Value Problems, Springer Science + Business Media,
2009.
[4] Jeffrey R. Chasnov, Introduction to Differential Equations. Lecture notes for MATH
2351/2352 , The Hong Kong University of Science and Technology, Department of
Mathematics.
[5] Marcel B. Finan, Lecture Notes in Mathematics Arkansas Tech University, Department
of Mathematics, 2018.
[6] StGoChap6.pdf, Partial differential equations, www.cns.gatech.edu .
[7] Ally Learn, First order partial differential equations, Ally Learn-Youtube.

25