7 Wave Equation in Higher Dimensions

We now consider the initial-value problem for the wave equation in n dimensions,
_
_
_
u
tt
c
2
u = 0 x R
n
u(x, 0) = (x)
u
t
(x, 0) = (x)
(7.1)
where u

n
i=1
u
x
i
x
i
.
7.1 Method of Spherical Means
Ref: Evans, Sec. 2.4.1; Strauss, Sec. 9.2
We begin by introducing a method to solve (7.1) in odd dimensions. First, we introduce
some notation. For x R
n
, let
B(x, r) = Ball of radius r about x
B(x, r) = Boundary of ball of radius r about x
(n) = Volume of unit ball in R
n
n(n) = Surface Area of unit ball in R
n
.
With this notation, the volume of the ball of radius r about x R
n
, written as Vol(B(x, r)),
is given by (n)r
n
and the surface area of the ball of radius r about x R
n
, written as
S.A.(B(x, r)), is given by n(n)r
n1
.
For f : R
n
R, we dene the average of f over B(x, r) as

_
B(x,r)
f(y) dy
1
Vol(B(x, r))
_
B(x,r)
f(y) dy =
1
(n)r
n
_
B(x,r)
f(y) dy.
We dene the average of f over B(x, r) as

_
B(x,r)
f(y) dS(y)
1
S.A.(B(x, r))
_
B(x,r)
f(y) dS(y) =
1
n(n)r
n1
_
B(x,r)
f(y) dS(y),
where dS(y) denotes the surface measure of B(x, r) in R
n
.
Example 1. For n = 3, Vol(B(x, r)) =
4
3
r
3
. Therefore, for f : R
3
R, the average of f
over B(0, r) is given by

_
B(0,r)
f(y) dy =
3
4r
3
_

0
_
2
0
_
r
0
f(, , )
2
sin d d d.
For n = 3, S.A.(B(x, r)) = 4r
2
. Therefore, for f : R
3
R, the average of f over B(0, r)
is given by

_
B(0,r)
f(y) dS(y) =
1
4r
2
_

0
_
2
0
f(r, , )r
2
sin d d =
1
4
_

0
_
2
0
f(r, , ) sin d d.

1
Our plan to solve (7.1) is the following. Fix a point x R
n
. For r > 0, we dene
u(x; r, t)
_
B(x,r)
u(y, t) dS(y),
the average of u(, t) over B(x, r). For r = 0, we dene u(x; 0, t) = u(x, t). For r < 0, we
dene u(x; r, t) = u(x; r, t). We claim that for u smooth, u is a continuous function of r,
and, therefore,
lim
r0+
u(x; r, t) = u(x, t).
In order to solve (7.1), we will assume u is a solution of (7.1) and look for an equation u
solves. Note: We will assume c = 1. For c ,= 1, we can make a change of variables to derive
the solution from the solution in the case c = 1.
Lemma 2. If u solves
_

_
u
tt
u = 0, x R
n
, t 0
u(x, 0) = (x)
u
t
(x, 0) = (x),
then u(x; r, t) solves
_

_
u
tt
u
rr

(n 1)
r
u
r
= 0, 0 < r < , t 0
u(x; r, 0) = (x; r)
_
B(x,r)
(y) dS(y)
u
t
(x; r, 0) = (x; r)
_
B(x,r)
(y) dS(y)
for every x R
n
.
Proof.
u(x; r, t) =
_
B(x,r)
u(y, t) dS(y)
=
_
B(0,1)
u(x + rz, t) dS(z).
2
Therefore,
u
r
(x; r, t) =
_
B(0,1)
u(x + rz, t) z dS(z)
=
_
B(x,r)
u(y, t)
y x
r
dS(y)
=
_
B(x,r)
u

(y, t) dS(y)
=
1
n(n)r
n1
_
B(x,r)
u

(y, t) dS(y)
=
1
n(n)r
n1
_
B(x,r)
u(y, t) dy
=
1
n(n)r
n1
_
B(x,r)
u
tt
(y, t) dy
by the Divergence Theorem, and using the fact that u solves the wave equation, u
tt
u = 0.
Therefore,
u
r
(x; r, t) =
1
n(n)r
n1
_
B(x,r)
u
tt
(y, t) dy
which implies
r
n1
u
r
(x; r, t) =
1
n(n)
_
B(x,r)
u
tt
(y, t) dy.
Therefore,
(r
n1
u
r
(x; r, t))
r
=
1
n(n)
_
B(x,r)
u
tt
(y, t) dS(y)
=
r
n1
n(n)r
n1
_
B(x,r)
u
tt
(y, t) dS
= r
n1

_
B(x,r)
u
tt
(y, t) dS(y)
= r
n1
u
tt
(x; r, t).
Therefore,
(r
n1
u
r
(x; r, t))
r
= r
n1
u
tt
(x; r, t),
which implies
(n 1)r
n2
u
r
+ r
n1
u
rr
= r
n1
u
tt
.
Therefore,
u
tt
u
rr

(n 1)
r
u
r
= 0
3
and
u(x; r, 0) =
_
B(x,r)
u(y, 0) dS =
_
B(x,r)
(y) dS = (x; r).
Similarly,
u
t
(x; r, 0) = (x; r)
as claimed.
Solution for n = 3.
We now consider the case of the wave equation in three dimensions. Assume u is a
solution of (7.1) for n = 3. As before dene the function u(x; r, t) such that
u(x; r, t) =
_
B(x,r)
u(y, t) dS(y).
Next introduce a function v(x; r, t) such that
v(x; r, t) = ru(x; r, t)
and new functions g(x; r) and h(x; r) such that
g(x; r) = r(x; r) = r
_
B(x,r)
(r) dS(y)
h(x; r) = r(x; r) = r
_
B(x,r)
(r) dS(y).
Lemma 3. For each x R
n
, the function v(x; r, t) solves the one-dimensional wave equation
on the half-line with Dirichlet boundary conditions,
_

_
v
tt
v
rr
= 0 0 < r < , t 0
v(x; r, 0) = g(x; r) 0 < r <
v
t
(x; r, 0) = h(x; r) 0 < r <
v(x; 0, t) = 0 t 0.
Proof.
v
tt
= ru
tt
= r
_
u
rr
+
2
r
u
r
_
= ru
rr
+ 2u
r
= (ru
r
+ u)
r
= (ru)
rr
= v
rr
.
4
Next,
v(x; r, 0) = ru(x; r, 0)
= r
_
B(x,r)
u(y, 0) dS(y)
= r
_
B(x,r)
(y) dS(y)
= r(x, r)
= g(x; r)
Similarly,
v
t
(x; r, 0) = h(x; r).
Now,
v(x; 0, t) = 0 u(x; 0, t) = 0.
Therefore, v(x; r, t) solves the one-dimensional wave equation on a half-line with Dirichlet
boundary conditions, as claimed.
Now we use this fact to construct the solution of (7.1). By dAlemberts formula, we
know that for 0 r t, the solution v(x; r, t) is given by
v(x; r, t) =
1
2
[g(x; r + t) g(x; t r)] +
1
2
_
r+t
r+t
h(x; y) dy.
Now
u(x, t) = lim
r0
+
u(x; r, t)
and
v(x; r, t) = ru(x; r, t).
Therefore,
u(x, t) = lim
r0
+
v(x; r, t)
r
= lim
r0
+
_
1
2r
[g(x; t + r) g(x; t r)] +
1
2r
_
r+t
r+t
h(x; y) dy
_
=
d
dt
g(x; t) + h(x; t).
Now
g(x; r) = r(x; r)
implies
g(x; t) = t(x; t) = t
_
B(x,t)
(y) dS(y).
Similarly,
h(x; t) = t(x; t) = t
_
B(x,t)
(y) dS(y).
5
Therefore, the solution of the wave equation in R
3
(with c = 1) is given by
u(x, t) =

t
_
t
_
B(x,t)
(y) dS(y)
_
+ t
_
B(x,t)
(y) dS(y).
If is smooth, the solution can be simplied further. In particular, for smooth, we
have
d
dt
g(x; t) =
d
dt
_
t
_
B(x,t)
(y) dS(y)
_
=
d
dt
_
t
_
B(0,1)
(x + tz) dS(z)
_
=
_
B(0,1)
(x + tz) dS(z) + t
_
B(0,1)
(x + tz) z dS(z)
=
_
B(x,t)
(y) dS(y) + t
_
B(x,t)
(y)
_
y x
t
_
dS(y)
=
_
B(x,t)
(y) dS(y) +
_
B(x,t)
(y) (y x) dS(y).
And,
h(x; t) = t(x; t) = t
_
B(x,t)
(y) dS(y).
Therefore, we have
u(x, t) =
_
B(x,t)
[(y) +(y) (y x) + t(y)] dS(y).
We note that in R
3
,

_
B(x,t)
=
1
n(n)t
n1
_
B(x,t)
=
1
4t
2
_
B(x,t)
.
Therefore, the solution of the IVP for the wave equation in R
3
(with c = 1 and smooth)
is given by
u(x, t) =
1
4t
2
_
B(x,t)
[(y) +(y) (y x) + t(y)] dS(y). (7.2)
This is known as Kirchos formula for the solution of the initial value problem for the
wave equation in R
3
.
Remark. Above we found the solution for the wave equation in R
3
in the case when c = 1.
If c ,= 1, we can simply use the above formula making a change of variables. In particular,
consider the initial-value problem
_

_
v
tt
c
2
v = 0 x R
n
v(x, 0) = (x)
v
t
(x, 0) = (x).
(7.3)
6
Suppose v is a solution of (7.3). Then dene u(x, t) v(x,
1
c
t). Then
u
tt
u =
1
c
2
v
tt
v = 0
implies u is a solution of
_

_
u
tt
u
xx
= 0 x R
n
u(x, 0) = (x)
u
t
(x, 0) =
1
c
(x).
Therefore, u is given by Kirchos formula above. Now by making the change of variables

t =
1
c
t, we see that
v(x,

t) = u(x, c

t),
and we arrive at the solution for (7.3),
v(x, t) =
1
4c
2
t
2
_
B(x,ct)
[(y) +(y) (y x) + t(y)] dS(y).
7.2 Method of Descent
In this section, we use Kirchos formula for the solution of the wave equation in three
dimensions to derive the solution of the wave equation in two dimensions. This technique
is known as the method of descent. This technique can be used in general to nd the
solution of the wave equation in even dimensions, using the solution of the wave equation in
odd dimensions.
Solution for n = 2.
Suppose u is a solution of the initial value problem for the wave equation in two dimen-
sions,
_

_
u
tt
u = 0, x R
2
, t 0
u(x, 0) = (x)
u
t
(x, 0) = (x).
We will nd a solution in the 2-D case, by using the solution to the 3-D problem. Let
u(x
1
, x
2
, t) be the solution to the 2-D problem. Dene
u(x
1
, x
2
, x
3
, t) u(x
1
, x
2
, t).
Therefore,
u(x
1
, x
2
, x
3
, 0) u(x
1
, x
2
, 0) = (x
1
, x
2
)
u
t
(x
1
, x
2
, x
3
, 0) u(x
1
, x
2
, 0) = (x
1
, x
2
).
Clearly, u(x
1
, x
2
, x
3
, t) is a solution of the 3D wave equation with initial data (x
1
, x
2
) and
(x
1
, x
2
),
_

_
u
tt
u
x
1
x
1
u
x
2
x
2
u
x
3
x
3
= 0
u(x
1
, x
2
, x
3
, 0) =

(x
1
, x
2
, x
3
) = (x
1
, x
2
)
u
t
(x
1
, x
2
, x
3
, 0) =

(x
1
, x
2
, x
3
) = (x
1
, x
2
).
7
Now we can solve the 3D wave equation using Kirchos formula. In particular, our
solution is given by
u(x
1
, x
2
, 0, t) =
_
B(x,t)
[

(y) +

(y) (y x) + t

(y)] dS(y)
where B(x, t) is the ball of radius t in R
3
about the point x = (x
1
, x
2
, 0). Now we note that

_
B(x,t)

(y) dS(y) =
1
4t
2
_
B(x,t)

(y) dS(y)
=
1
2t
2
_
B(x,t)
(y)(1 +[(y)[
2
)
1/2
dy
where B(x, t) is the ball in R
2
of radius t about the point x = (x
1
, x
2
) and (y) = (t
2
[y
x[
2
)
1/2
. Therefore,
(y) =
y x
(t
2
[y x[
2
)
1/2
which implies
(1 +[(y)[
2
)
1/2
=
_
t
2
t
2
[y x[
2
_
1/2
.
Therefore,

_
B(x,t)

(y) dS(y) =
1
2t
2
_
B(x,t)
t(y)
(t
2
[y x[
2
)
1/2
dy.
Similarly,

_
B(x,t)
t

(y) dS(y) =
1
2t
2
_
B(x,t)
t
2
(y)
(t
2
[y x[
2
)
1/2
dy
and

_
B(x,t)

(y) (y x) dS(y) =
1
2t
2
_
B(x,t)
t(y) (y x)
(t
2
[y x[
2
)
1/2
dy.
Therefore, the solution of the initial-value problem for the wave equation in R
2
(with c = 1)
is given by
u(x, t) =
1
2t
2
_
B(x,t)
t(y) + t
2
(y) + t(y) (y x)
(t
2
[y x[
2
)
1/2
dy. (7.4)
Again, by making a change of variables, we see that the solution of the wave equation in
two dimensions is given by
u(x, t) =
1
2c
2
t
2
_
B(x,ct)
ct(y) + ct
2
(y) + ct(y) (y x)
(c
2
t
2
[y x[
2
)
1/2
dy.
8
7.3 Huygens Principle
Note that for the initial-value problem for the wave equation in three dimensions, the value
of the solution at any point (x, t) R
3
(0, ) depends only on the values of the initial data
on the surface of the ball of radius ct about the point x R
3
; that is, on B(x, ct). That is
to say, disturbances all travel at exactly speed c. This is known as Huygenss principle.
In contrast, in two dimensions, the value of the solution u at the point (x, t) depends on the
initial data within the ball of radius ct about the point x R
2
. Signals dont all travel at
speed c. In fact, as we will see, for n 3 and odd, Huygenss principle holds. That is, all
signals travel at exactly speed c. In even dimensions, however, that is not the case.
7.4 Wave Equation in R
n
, n > 3
Ref: Evans, Sec. 2.4.1
Note: In this section, we assume c = 1. For c ,= 1, we can make a change of variables to
nd the solution.
Odd dimensions.
For the case of odd dimensions, we use the method of spherical means as we did for the
case of n = 3. Let n = 2k + 1. Let x R
n
. Dene
v(x; r, t)
_
1
r

r
_
k1
(r
2k1
u(x; r, t))
g(x; r)
_
1
r

r
_
k1
(r
2k1
(x; r))
h(x; r)
_
1
r

r
_
k1
(r
2k1
(x; r)).
Notice that for k = 1, these denitions reduce to those functions introduced in the case
n = 3.
First, we will show that v(x; r, t) solves the wave equation on the half-line with Dirichlet
boundary conditions.
Lemma 4. For each integer k 1, for each x R
n
, the function v(x; r, t) dened above
solves
_

_
v
tt
v
rr
= 0 r > 0
v(x; r, 0) = g(x; r)
v
t
(x; r, 0) = h(x; r)
v(x; 0, t) = 0.
The proof relies on the following lemma.
Lemma 5. Let : R R be C
k+1
. Then for k = 1, 2, . . .
1.
_
d
2
dr
2
__
1
r
d
dr
_
k1
_
r
2k1
(r)
_
=
_
1
r
d
dr
_
k
_
r
2k
d
dr
(r)
_
9
2.
_
1
r
d
dr
_
k1
_
r
2k1
(r)
_
=
k1

j=0

k
j
r
j+1
d
j

dr
j
(r)
where each
k
j
is independent of .
3.

k
0
= 1 3 5 (2k 1).
Proof. Use induction.
Proof of Lemma 4.
v
rr
=
2
r
_
_
1
r
d
dr
_
k1
(r
2k1
u(x; r, t))
_
=
_
1
r
d
dr
_
k
(r
2k
u
r
(x; r, t)) by Lemma 5
=
_
1
r
d
dr
_
k1
_
1
r
d
dr
_
(r
2k
u
r
(x; r, t))
=
_
1
r
d
dr
_
k1
_
1
r
[2kr
2k1
u
r
+ r
2k
u
rr
]
_
=
_
1
r
d
dr
_
k1
_
r
2k1
_
2k
r
u
r
+ u
rr
__
=
_
1
r
d
dr
_
k1
_
r
2k1
_
n 1
r
u
r
+ u
rr
__
=
_
1
r
d
dr
_
k1
_
r
2k1
u
tt
_
=
2
t
_
1
r
d
dr
_
k1
_
r
2k1
u
tt
_
= v
tt
Clearly, v(x; r, 0) = g(x; r), v
t
(x; r, 0) = h(x; r) and v(x; 0, t) = 0. Therefore, the lemma is
proved.
Now v(x; r, t) is a solution of the one-dimensional wave equation on the half-line with
Dirichlet boundary condition implies for 0 r t, the solution is given by
v(x; r, t) =
1
2
[g(x; r + t) g(x; t r)] +
1
2
_
t+r
tr
h(x; y) dy.
Recall:
u(x, t) = lim
r0
u(x; r, t).
10
Now
v(x; r, t) =
_
1
r
d
dr
_
k1
_
r
2k1
u(x; r, t)
_
=
k1

j=0

k
j
r
j+1

j
r
j
u(x; r, t)
=
k
0
ru(x; r, t) +
k
1
r
2
u
r
(x; r, t) + . . . +
k
k1
r
k

k1
r
k1
u(x; r, t).
Therefore,

k
0
ru(x; r, t) = v(x; r, t)
k
1
r
2
u
r
(x; r, t) . . .
k
k1
r
k

k1
r
k1
u(x; r, t),
which implies
u(x; r, t) =
v(x; r, t)

k
0
r


k
1

k
0
ru
r
(x; r, t) . . .

k
k1

k
0
r
k1

k1
r
k1
u(x; r, t).
Therefore,
u(x, t) = lim
r0
_
v(x; r, t)

k
0
r


k
1

k
0
ru
r
(x; r, t) . . .

k
k1

k
0
r
k1

k1
r
k1
u(x; r, t)
_
= lim
r0
v(x; r, t)

k
0
r
= lim
r0
1

k
0
_
g(x; t + r) g(x; t r)
2r
+
1
2r
_
t+r
tr
h(x; y) dy
_
=
1

k
0
[
t
g(x; t) + h(x; t)]
where
k
0
= 1 3 5 (2k 1). Recall
g(x; r) =
_
1
r

r
_
k1
(r
2k1
(x; r)).
Now n = 2k + 1 implies k = (n 1)/2, and, therefore,
g(x; t) =
_
1
t

t
_n3
2
_
t
n2

_
B(x,t)
(y) dS(y)
_
.
And,
h(x; r) =
_
1
r

r
_
k1
(r
2k1
(x; r)).
Therefore,
h(x; t) =
_
1
t

t
_n3
2
_
t
n2

_
B(x,t)
(y) dS(y)
_
.
11
Therefore,
u(x, t) =
1

n
[
t
g(x; t) + h(x; t)]
implies
u(x, t) =
1

n
_

t
__
1
t

t
_n3
2
_
t
n2

_
B(x,t)
(y) dS(y)
_
+
1

n
_
1
t

t
_n3
2
_
t
n2

_
B(x,t)
(y) dS(y)
_
where
n
= 1 3 5 (n 2).
Even dimensions.
As in the case of n = 2 dimensions, we use the method of descent. In particular, suppose
u(x
1
, . . . , x
n
, t) is a solution of the wave equation in R
n
with initial data u(x
1
, . . . , x
n
, 0) =
(x
1
, . . . , x
n
) and u
t
(x
1
, . . . , x
n
, 0) = (x
1
, . . . , x
n
). Then dene
u(x
1
, . . . , x
n+1
, t) u(x
1
, . . . , x
n
, t)

(x
1
, . . . , x
n+1
) (x
1
, . . . , x
n
)

(x
1
, . . . , x
n+1
) (x
1
, . . . , x
n
).
Therefore, u is a solution of the wave equation in R
n+1
, where now n + 1 is odd. Therefore,
from the formula above for the case when the dimension is odd, our solution at the point
(x, t) = (x
1
, . . . , x
n
, 0, t) is given by
u(x, t) =
1

n+1
_

t
__
1
t

t
_n2
2
_
t
n1

_
B(x,t)

(y) dS(y)
_
+
1

n+1
_
1
t

t
_n2
2
_
t
n1
_
B(x,t)

(y) dS(y)
_
where
n+1
= 1 3 5 (n 1), and where B(x, t) is the ball in R
n+1
of radius t about the
point x = (x
1
, . . . , x
n
, 0).
Now,

_
B(x,t)

(y) dS(y) =
1
(n + 1)(n + 1)t
n
_
B(x,t)

(y) dS(y).
But, notice B(x, t) y
n+1
0 is the graph of the function (y) (t
2
[y x[
2
)
1/2
. And,
similarly, B(x, t) y
n+1
0 is the graph of . Therefore,
1
(n + 1)(n + 1)t
n
_
B(x,t)

(y) dS(y) =
2
(n + 1)(n + 1)t
n
_
B(x,t)
(y)(1 +[(y)[
2
)
1/2
dy
Now
(1 +[(y)[
2
)
1/2
= t(t
2
[y x[
2
)
1/2
.
12
Therefore,

_
B(x,t)

(y) dS(y) =
2
(n + 1)(n + 1)t
n
_
B(x,t)
t(y)
(t
2
[y x[
2
)
1/2
dy
=
2t(n)
(n + 1)(n + 1)(n)t
n
_
B(x,t)
(y)
(t
2
[y x[
2
)
1/2
dy
=
2t(n)
(n + 1)(n + 1)

_
B(x,t)
(y)
(t
2
[y x[
2
)
1/2
dy.
Therefore, our solution formula is given by
u(x, t) =
1

n+1
_

t
__
1
t

t
_n2
2
_
t
n1

_
B(x,t)

(y) dS(y)
_
+
1

n+1
_
1
t

t
_n2
2
_
t
n1

_
B(x,t)

(y) dS(y)
_
=
1

n+1

2(n)
(n + 1)(n + 1)
_
_

t
__
1
t

t
_n2
2
_
t
n

_
B(x,t)
(y)
(t
2
[y x[
2
)
1/2
dy
_
+
_
1
t

t
_n2
2
_
t
n

_
B(x,t)
(y)
(t
2
[y x[
2
)
1/2
dy
_
_
.
Now
n+1
= 1 3 5 (n 1) and
(n) =

n/2

_
n+2
2
_,
where (n) is the gamma function,
(n) =
_

0
e
x
x
n1
dx.
Therefore,
1

n+1

2(n)
(n + 1)(n + 1)
=
1
1 3 5 (n 1)

2

n/2
(
n+2
2
)
(n + 1)

(n+1)/2
(
n+3
2
)
=
1
1 3 5 (n + 1)

1

1/2

(
n+3
2
)
(
n+2
2
)
.
Using properties of the gamma function, namely that
(m + 1) = m(m)
and
(1/2) =
1/2
,
13
we can conclude that

_
n + 3
2
_
=
_
n + 1
2
_

_
n 1
2
_

_
1
2
_

_
1
2
_
and

_
n + 2
2
_
=
_
n
2
_

_
n 2
2
_

_
2
2
_
.
And, therefore,
1

n+1

2(n)
(n + 1)(n + 1)
=
1
2 4 (n 2) n
Therefore, the solution of the wave equation in even dimensions is given by
u(x, t) =
1

n
_
_

t
__
1
t

t
_n2
2
_
t
n

_
B(x,t)
(y)
(t
2
[y x[
2
)
1/2
dy
_
+
_
1
t

t
_n2
2
_
t
n

_
B(x,t)
(y)
(t
2
[y x[
2
)
1/2
dy
_
_
where
n
2 4 (n 2) n.
7.5 Wave Equation in R
n
with a source.
In this section, we consider the inhomogeneous wave equation in R
n
. First, recall Duhamels
Principle. If S(t) is the solution operator for the rst-order initial-value problem
_
U
t
+ AU = 0
U(0) = ,
then the solution of the inhomogeneous problem
_
U
t
+ AU = F
U(0) =
should be given by
U(t) = S(t) +
_
t
0
S(t s)F(s) ds.
Now consider the initial-value problem for the wave equation in R
n
,
_

_
u
tt
u = f(x, t) x R
n
u(x, 0) = (x)
u
t
(x, 0) = (x).
(7.5)
14
Introducing a new function v = u
t
, we can rewrite this equation as
_

_
_
u
v
_
t
=
_
0 1
0
_ _
u
v
_
+
_
0
f
_
x R
n
_
u(x, 0)
v(x, 0)
_
=
_
(x)
(x)
_
.
(7.6)
or
_
U
t
+ AU = F
U(x, 0) = (x)
where
U =
_
u
v
_
A =
_
0 1
0
_
F =
_
0
f
_
=
_

_
.
Now in order to solve (7.5), we look for the solution operator S(t) associated with the
rst-order system (7.6).
First, consider the case n = 3. In three dimensions, we can nd the solution operator
S(t) by using Kirchos formula. Recall that the solution of the initial-value problem for the
homogeneous wave equation in three dimensions (with c = 1) is given by
u(x, t) =
1
4t
2
_
B(x,t)
[(y) +(y) (y x) + t(y)] dS(y),
which implies the solution operator S(t) associated with (7.6) is given by
S(t) = S(t)
_

_
=
_
1
4t
2
_
B(x,t)
[(y) +(y) (y x) + t(y)] dS(y)

t
_
1
4t
2
_
B(x,t)
[(y) +(y) (y x) + t(y)] dS(y)
_
_
.
Therefore,
S(t s)F(s) = S(t s)
_
0
f(s)
_
=
_
1
4(ts)
_
B(x,ts)
f(y, s) dS(y)

t
_
1
4(ts)
_
B(x,ts)
f(y, s) dS(y)
_
_
.
Now using the fact that the solution of (7.6) is given by
_
u(x, t)
v(x, t)
_
= U(x, t) = S(t)(x) +
_
t
0
S(t s)F(x, s) ds,
we see that the solution of (7.5) is given by the rst component of U. Therefore, the solution
of the initial-value problem for the inhomogeneous wave equation in three dimensions (with
c = 1) (7.5) is given by
u(x, t) =
1
4t
2
_
B(x,t)
[(y) +(y) (y x) + t(y)] dS(y)
+
_
t
0
1
4(t s)
_
B(x,ts)
f(y, s) dS(y) ds.
15
Similarly, in two dimensions, the rst component of the solution operator is given by
S
1
(t) = S
1
(t)
_

_
=
1
2t
2
_
B(x,t)
t(y) + t
2
(y) + t(y) (y x)
(t
2
[y x[
2
)
1/2
dy.
Therefore, the solution of the initial-value problem for the inhomogeneous wave equation in
two dimensions (with c = 1) (7.5) is given by
u(x, t) =
1
2t
2
_
B(x,t)
t(y) + t
2
(y) + t(y) (y x)
(t
2
[y x[
2
)
1/2
dy
+
_
t
0
1
2(t s)
2
_
B(x,ts)
(t s)
2
f(y, s)
((t s)
2
[y x[
2
)
1/2
dy ds.
7.6 Wave Equation on a Bounded Domain in R
n
.
In this section, we consider the initial-value problem for the wave equation on a bounded
domain R
n
,
_

_
u
tt
c
2
u = 0, x
u(x, 0) = (x)
u
t
(x, 0) = (x)
u satises certain boundary conditions on ,
As before, we look for a solution using separation of variables. In particular, we look for a
solution of the form u(x, t) = X(x)T(t). Substituting a function of this form into our PDE,
we arrive at the equation
T

X c
2
TX = 0.
This equation implies the functions T and X satisfy the following equation for some scalar
,

c
2
T
=
X
X
= .
Consequently, we are lead to the following eigenvalue problem
_
X = X, x
X satises certain boundary conditions on .
Suppose we nd eigenvalues
n
with corresponding eigenfunctions X
n
(x). Then for each
n, we just need to solve
T

n
(t) + c
2

n
T
n
(t) = 0.
If
n
is positive, this means
T
n
(t) = A
n
cos(
_

n
ct) + B
n
sin(
_

n
ct).
If
n
= 0, this means
T
n
(t) = A
n
+ B
n
t.
16
If
n
is negative, this means
T
n
(t) = A
n
cosh(
_

n
ct) + B
n
sinh(
_

n
ct).
Then dening the function
u(x, t) =

n
T
n
(t)X
n
(x),
for X
n
, T
n
as dened above for any choice of constants A
n
, B
n
, we have found a solution of
the wave equation on the bounded domain R
n
, which satises our boundary conditions.
Now in order for our initial conditions to be satised, that is, u(x, 0) = (x) and u
t
(x, 0) =
(x), we need to choose constants A
n
, B
n
such that
u(x, 0) =

n
A
n
X
n
(x) = (x)
and
u
t
(x, 0) =

n=0
c
_
[
n
[B
n
X
n
(x) +

n=0
B
n
X
n
(x) = (x).
If our eigenfunctions are orthogonal, then we can nd coecients A
n
, B
n
satisfying the above
equations, by multiplying these equations by X
m
for a xed m and integrating over . Doing
so, we see that our coecients A
n
are given by
A
n
=
X
n
, )
X
n
, X
n
)
=
_

X
n
(x)(x) dx
_

X
2
n
(x) dx
,
and
c
_
[
n
[B
n
=
X
n
, )
X
n
, X
n
)
=
_

X
n
(x)(x) dx
_

X
2
n
(x) dx
for
n
,= 0
B
n
=
X
n
, )
X
n
, X
n
)
=
_

X
n
(x)(x) dx
_

X
2
n
(x) dx
for
n
= 0.
17