Study material for Week-8 on

Partial Differential Equations

S. Sivaji Ganesh, IIT Bombay.

Distributed to students of MA 515, Autumn 2021.

 Preface

This notes is intended for circulation to students of a course on Partial differential

equations.
Though the notes are proof-read many times, it could still have some misprints.
Please send your comments, corrections etc. to [email protected], where X=msg2sgspde

S. Sivaji Ganesh,
Mumbai.
September 12, 2021.

vii
4.8. IBVP for one dimensional wave equation 239

4.8.2 Method of separation of variables

We were considering the following IBVP at the beginning of this section,

and we discussed how to solve the same starting from the first principles.

u t t − c 2 u x x = 0 for 0 < x < l , t > 0, (4.89a)

u(x, 0) = φ(x) for 0 ≤ x ≤ l , (4.89b)
∂u
(x, 0) = ψ(x) for 0 ≤ x ≤ l , (4.89c)
∂t
u(0, t ) = 0 for t ≥ 0, (4.89d)
u(l , t ) = 0 for t ≥ 0. (4.89e)

In this subsection, we use a general method called separation of variables

method to solve the IBVP (4.89). Solution u := u(x, t ) to the IBVP (4.89) is
a superposition of the solutions u1 (with φ ≡ 0), and u2 (with ψ ≡ 0) of the
IBVP (4.89) as u = u1 + u2 , since the equation (4.89a) is linear. We describe
the method of separation of variables in the special case where ψ ≡ 0, and
the method proceeds similarly when φ ≡ 0.
Method of separation of variables is organized into three main steps.

Step 1: Obtain two families of ODEs from Wave equation.

Separation of variables method looks for solutions to (4.89a) which are

in the separated form

u(x, t ) = X (x)T (t ) for x ∈ (0, l ), t > 0. (4.90)

Substituting the expression for u from (4.90) in the equation (4.89a)

gives
X (x)T 00 (t ) − c 2 X 00 (x)T (t ) = 0. (4.91)
On dividing both sides of the equation (4.91) with X (x)T (t ) and re-arranging
terms yields
T 00 (t ) X 00 (x)
= c2 . (4.92)
T (t ) X (x)
Note that the LHS of the equation (4.92) is a function of t only, while
the RHS is a function of x only. Such an equation can hold if and only if
both the functions are identically equal to a constant function. It means

September 12, 2021 Sivaji

240 Chapter 4. Wave equation I

that for every solution in the separated form there exists a λ ∈ R such that

T 00 (t ) X 00 (x)
= c2 = λ. (4.93)
T (t ) X (x)

One of the tasks is to find all possible λs that arise from separated solutions
satisfying the equation (4.93). The equation (4.93) gives rise to two ODEs
which are given by

λ
X 00 − X = 0, T 00 − λT = 0. (4.94)
c 2

In fact, the equation (4.94) represents two families of ODEs indexed by the
same parameter λ ∈ R. Using the boundary condition (4.89d), we get

u(0, t ) = X (0)T (t ) = 0 for all t > 0.

Since we do not wish to find a trivial solution u ≡ 0, we cannot admit

T (t ) ≡ 0. Thus we get X (0) = 0.
Similarly using the boundary condition (4.89e), we get

u(l , t ) = X (l )T (t ) = 0 for all t > 0.

Since we do not wish to find a trivial solution u ≡ 0, we cannot admit

T (t ) ≡ 0. Thus we get X (l ) = 0.
Thus we are led to the boundary value problem for X given by

λ
X 00 − X = 0, (4.95a)
c2
X (0) = X (l ) = 0. (4.95b)

Similarly, using the initial condition (4.89c) with ψ ≡ 0, we obtain a (par-

tial) initial value problem for T given by

T 00 − λT = 0, (4.96a)
T 0 (0) = 0. (4.96b)

Step 2: Finding non-trivial solutions to the two families of ODEs.

Sivaji IIT Bombay

4.8. IBVP for one dimensional wave equation 241

We are interested in finding non-trivial solutions to the boundary value

problem (4.95), existence of which depends on the value of λ. The λs for
which the BVP (4.95) admits a non-trivial solution is called an eigenvalue
and corresponding non-trivial solutions are called eigenfunctions. Let us
begin the search for the eigenvalues, and corresponding eigenfunctions.
The search for eigenvalues λ ∈ R leads us to the following three cases:
λ = 0, λ > 0, and λ < 0.

(i) (λ = 0) General solution of the ODE (4.95a) is given by X (x) = ax+b .

Applying the boundary conditions (4.95b), we get a = b = 0. Thus
λ = 0 is not an eigenvalue.
(ii) (λ > 0) When λ > 0, we may write λ = µ2 where µ > 0. The ODE
(4.95a) then becomes

µ2
X 00 − X = 0,
c2
µ µ
whose general solution is given by X (x) = ae c x + b e − c x . The bound-
ary conditions (4.95b) lead us to a = b = 0. Thus λ > 0 is not an
eigenvalue.

(iii) (λ < 0) When λ < 0, we may write λ = −µ2 where µ > 0. The ODE
(4.95a) then becomes

µ2
X 00 + X = 0,
c2
µ
µ

whose general solution is given by X (x) = a cos c x + b sin c x .
Applying the boundary conditions (4.95b), we get
µ  µ 
a = 0, a cos l + b sin l = 0. (4.97)
c c
Since we are interested in non-trivial solutions to the BVP (4.95), at
least one of the constants a, b should be non-zero. However, we al-
ready have a = 0 from (4.106). Thus in order to have b 6= 0, we must
have µ 
sin l = 0. (4.98)
c
c nπ
Solving the equation (4.98), we get µn = l for each n ∈ N. Thus
we have the following eigenvalue and corresponding eigenfunction,

September 12, 2021 Sivaji

242 Chapter 4. Wave equation I

indexed by n ∈ N:

c 2 n 2 π2  nπ 
λn = − , Xn (x) = sin x .
l2 l
For each n ∈ N, the solution to the problem (4.96) with λ = λn yields

 nπc 
Tn (t ) = b cos t ,
l
where b ∈ R is arbitrary.

Step 3: Definition of a formal solution to the IBVP

Since for each n ∈ N the function Xn (x)Tn (t ) is a solution to the wave

equation (4.89a) which is linear and homogeneous, we propose a formal
solution to the IBVP as a superposition of Xn (x)Tn (t ) (n ∈ N) by
X
∞  nπ   nπc 
u(x, t ) ≈ bn sin x cos t (4.99)
n=1 l l

where bn ∈ R, n ∈ N. These constants are to be determined using the

initial condition (4.89b). Indeed the function u defined by (4.99) satisfies
the initial condition (4.89b) if and only if
 nπ 
X
∞
φ(x) = u(x, 0) = bn sin x .
n=1 l

Thus bn are the fourier sine coefficients of the function φ. In order to

determine the fourier sine series of the function φ, we extend the function
φ to the interval [−l , l ] as an odd function about x = 0, and we continue
to use the notation φ to denote the extended function. Then the fourier
series of φ takes the form
X
∞  nπ 
φ(x) = bn sin x ,
n=1 l

where bn is given by
Z l  nπ 
2
bn = φ(x) sin x d x. (4.100)
l 0 l

Sivaji IIT Bombay

4.8. IBVP for one dimensional wave equation 243

Thus the formal solution of the IBVP (4.89) is given by

‚ Zl  nπ  Œ  nπ   nπc 
X∞
2
u(x, t ) ≈ φ(x) sin x d x sin x cos t .
n=1 l 0 l l l
(4.101)
A sufficient condition on the function φ which assures that the formal
solution defined by (4.101) is indeed a classical solution to the IBVP is stated
in the following result.

Theorem 4.23. Let φ ∈ C 4 [0, l ] be such that φ(0) = φ(l ) = φ 00 (0) =

φ 00 (l ) = 0, ψ ≡ 0. Then the function defined by (4.101) is a solution to the
IBVP (4.89).

Proof. The formal solution given by (4.101) is indeed a solution to the

IBVP (4.89) once the following properties are verified.
(i) The infinite series in (4.101) converges, and defines a function which
is twice continuously differentiable w.r.t. the variables x and t . Let us
denote the function by u. Once the convergence of the infinite series
in (4.101) is established, the symbol ≈ may be replaced with =
(ii) The function u is a solution to the wave equation (4.89a).
(iii) The function u(x, t ) is continuous up to the boundary of (0, l ) ×
(0, ∞), and the initial-boundary conditions (4.89b)-(4.89e) with ψ = 0
are satisfied.
Step 1: Proof of (i) and (ii) By formal differentiation of the equation
(4.101), using the notation introduced in (4.100), we get

πc X ∞  nπ   nπc 
u t (x, t ) = − nb sin x sin t , (4.102a)
l n=1 n l l
 πc 2 X∞  nπ   nπc 
u t t (x, t ) = − n 2 bn sin x cos t , (4.102b)
l n=1 l l
πX ∞  nπ   nπc 
u x (x, t ) = nb cos x cos t , (4.102c)
l n=1 n l l
 π 2 X
∞  nπ   nπc 
u x x (x, t ) = − n 2 bn sin x cos t . (4.102d)
l n=1 l l

Assuming that the serie in (4.102) are convergent, it follows that u t t =

2
c u x x . Thus it remains to justify that the infinite series in (4.101) defines a
twice differentiable function. By a theorem of real analysis Theorem ??, it

September 12, 2021 Sivaji

244 Chapter 4. Wave equation I

is enough to show that the serie in (4.102) are uniformly convergent, and
the series in (4.101) converges at some point. Clearly all the serie in (4.102)
are uniformly convergent (by comparison test) if there exists a constant
C > 0 such that for each n ∈ N the following inequalities hold.
1
n 2 |bn | ≤ . (4.103)
n2
Let us show that the conditions (4.103) are satisfied.

Z  nπ 
2 l
bn = φ(x) sin x dx
l 0 l
Zl
2 d   nπ 
=− φ(x) cos x dx
nπ 0 dx l
Zl  nπ   nπ  l
2
= 0
φ (x) cos x d x − φ(x) cos x 
nπ 0 l l 0
Zl  
2 nπ
= φ 0 (x) cos x dx
nπ 0 l
Zl
2l 0 d   nπ 
= φ (x) sin x dx
n 2 π2 0 dx l
Zl  nπ   nπ  l
2l
=− 00
φ (x) sin 0
x d x + φ (x) sin x 
n 2 π2 0 l l 0
Zl  
2l nπ
=− φ 00 (x) sin x d x.
n π 0
2 2 l

Continuing the above computations two more times, we get

Z l  nπ 
2l 3 (i v)
bn = φ (x) sin x d x.
n 4 π4 0 l

The last equation gives the following estimate:

2l 4
|bn | ≤ M ,
n 4 π4
where M = max |φ (i v) (x)|, from which the estimate (4.103) follows.
x∈[0,l ]

Step 2: Proof of (ii) Due to assumptions on φ, the fourier sine series

of φ converges to φ, and hence the initial condition (4.89b) is satisfied.

Sivaji IIT Bombay

4.8. IBVP for one dimensional wave equation 245

Other conditions (4.89c)-(4.89e) are checked by substituting appropriately

the values of x and t in (4.101).

Remark 4.24 (Disadvantages of separation of variables method ).

(i) An existence and uniqueness result for the IBVP (4.89), namely Theo-
rem 4.20, was established under milder assumptions on the initial data
φ in Subsection 4.8.1. Under the same hypothesis, not much can be
concluded about the formal solution obtained by separation of vari-
ables method.
(ii) Even when the formal series obtained by separation of variables method
converges, its sum can be found very rarely. This is due to the inherent
disadvantages associated with infinite serie.
(iii) Of course, neither of the results proved for both the methods are use-
ful in some physically relevant situations such as an initially plucked
string, which corresponds to φ that is piecewise linear (graph of φ
is of triangle shape) and hence is not a differentiable function on the
interval [0, l ].
(iv) When φ is only piecewise continuous, the formal solution may be
interepreted as a weak solution. For more details on weak solutions
to wave equation, see Section 5.7 (in Chapter 5).
Remark 4.25 (Advantages of separation of variables method ).
(i) In the previous remark, we noted disadvantages of separation of vari-
ables method. However there are some merits to this method as well
from computation point of view.
(ii) Even if the infinite series converges, it is not possible to compute the
sum except in very rare circumstnces, where exact analytical form of
the sum of the infinite series can be determined.
(iii) Thus a solution can only be computed approximately. One choice
of an approximate solution is to truncate the infinite series to finitely
many terms.
(iv) Note that finite sum always solves homogeneous wave equation ex-
actly. The initial conditions are only approximately satisfied.
A similar procedure may be used to solve other kinds of boundary condi-
tions. We are going to present an example in which the separation of vari-
ables method is used to solve an IBVP with Neumann boundary conditions

September 12, 2021 Sivaji

246 Chapter 4. Wave equation I

for the homogeneous wave equation in Example 4.26. The separation of

variables method may be used to solve IBVPs for nonhomogeneous wave
equation as well, and the method is described in Example ??

Example 4.26. Let us solve the IBVP

u t t − u x x = 0 for 0 < x < 1, t > 0, (4.104a)

u(x, 0) = cos πx for 0 ≤ x ≤ 1, (4.104b)
∂u
(x, 0) = sin πx for 0 ≤ x ≤ 1, (4.104c)
∂t
u x (0, t ) = 0 for t ≥ 0, (4.104d)
u x (1, t ) = 0 for t ≥ 0. (4.104e)

using the method of separation of variables. The method looks for solu-
tions to (4.104a) which are in the separated form u(x, t ) = X (x)T (t ). Sub-
stituting the expression for u in the equation (4.104a) leads to two families
of ODEs indexed by a parameter λ, which are given by

X 00 − λX = 0, T 00 − λT = 0.

Using the boundary condition (4.104d), we get

u x (0, t ) = X 0 (0)T (t ) = 0 for all t > 0.

Since we do not wish to find a trivial solution u ≡ 0, we cannot admit

T (t ) ≡ 0. Thus we get X 0 (0) = 0.
Similarly using the boundary condition (4.104e), we get

u x (1, t ) = X 0 (1)T (t ) = 0 for all t > 0.

Since we do not wish to find a trivial solution u ≡ 0, we cannot admit

T (t ) ≡ 0. Thus we get X 0 (1) = 0.
Thus we are led to the boundary value problem for X given by

X 00 − λX = 0, (4.105a)
0 0
X (0) = X (l ) = 0. (4.105b)

We are interested in finding non-trivial solutions to the boundary value

problem (4.105), existence of which depends on the value of λ. The λs for
which the BVP (4.105) admits a non-trivial solution is called an eigenvalue
and corresponding non-trivial solutions are called eigenfunctions. Let us

Sivaji IIT Bombay

4.8. IBVP for one dimensional wave equation 247

begin the search for the eigenvalues, and corresponding eigenfunctions.

The search for eigenvalues λ ∈ R leads us to the following three cases:
λ = 0, λ > 0, and λ < 0.

(i) (λ = 0) General solution of the ODE (4.105a) is given by X (x) =

a + b x. Applying the boundary conditions (4.105b), we get b = 0.
Thus λ = 0 is an eigenvalue, and X (x) = 1 is an eigenfunction.
(ii) (λ > 0) When λ > 0, we may write λ = µ2 where µ > 0. The ODE
(4.95a) then becomes

X 00 − µ2 X = 0,
µ µ
whose general solution is given by X (x) = ae c x + b e − c x . The bound-
ary conditions (4.95b) lead us to a = b = 0. Thus λ > 0 is not an
eigenvalue.

(iii) (λ < 0) When λ < 0, we may write λ = −µ2 where µ > 0. The ODE
(4.95a) then becomes

X 00 + µ2 X = 0,

whose general solution is given by X (x) = a cos (µx) + b sin (µx).

Applying the boundary conditions (4.95b), we get

b = 0, a sin (µ) = 0. (4.106)

Since we are interested in non-trivial solutions to the BVP (4.105), we

must have
sin (µ) = 0. (4.107)
Solving the equation (4.107), we get µn = nπ for each n ∈ N. Thus
we have the following eigenvalue and corresponding eigenfunction,
indexed by n ∈ N:

λn = −n 2 π2 , Xn (x) = cos (nπx) .

Thus the eigenvalues and eigenfunctions associated to the BVP (4.105) are

λ0 = 0, X0 (x) = 1; (4.108a)
λn = −n π ,
2 2
Xn (x) = cos (nπx) , n ∈ N. (4.108b)

September 12, 2021 Sivaji

248 Chapter 4. Wave equation I

For each n ∈ N ∪ {0}, solution to the equation T 00 − λT = 0 with λ = λn

yields

T0 (t ) = a0 + b0 (t ),
Tn (t ) = an cos (nπt ) + bn sin (nπt ) , n ∈ N.

where an , bn ∈ R (n ∈ N) are arbitrary.

We propose a formal solution to the IBVP as a superposition of Xn (x)Tn (t )

(n ∈ N ∪ {0}) by
X
∞
u(x, t ) ≈ a0 + b0 t + (an cos (nπt ) + bn sin (nπt )) cos (nπx) . (4.109)
n=1

Using the initial condition (4.104b), we get

X
∞
u(x, 0) = cos(πx) = a0 + (an cos (nπt ) + bn sin (nπt ))
n=1

The last equation implies that

a0 = 0, a1 = 1, an = 0 for n ≥ 2.

Therefore, the formula (4.109) becomes

X
∞
u(x, t ) ≈ b0 t + cos(πt ) cos(πx) + bn sin (nπt ) cos (nπx) . (4.110)
n=1

Differentiating the series (4.110) w.r.t. t yields

X
∞
u t (x, t ) ≈ b0 − π sin(πt ) cos(πx) + n bn cos (nπt ) cos (nπx) .
n=1

Using the initial condition (4.104c), we get

X
∞
u t (x, 0) = sin(πx) = b0 + n bn cos (nπx)
n=1

The last equation is a fourier cosine series for sin(πx). Extending the func-
tion sin(πx) (0 ≤ x ≤ 1) to the interval [−1, 1] as an even function. Com-

Sivaji IIT Bombay

4.8. IBVP for one dimensional wave equation 249

puting the coefficients bn , we get

¨ 4
− π(k+1)k(k−1) if k is even,
bk =
0 if k is odd.

Thus the formal solution to the IBVP is given by

2
u(x, t ) ≈ t + cos(πt ) cos(πx)
π
4 X 1
− sin (kπt ) cos (kπx) .
π k even (k + 1)k(k − 1)

Example 4.27. Let us solve the IBVP

u t t − u x x = cos (2πx) cos (2πt ) for 0 < x < 1, t > 0, (4.111a)

u(x, 0) = cos πx for 0 ≤ x ≤ 1,
2
(4.111b)
∂u
(x, 0) = 2 cos (2πx) for 0 ≤ x ≤ 1, (4.111c)
∂t
u x (0, t ) = 0 for t ≥ 0,
u x (1, t ) = 0 for t ≥ 0.

using a modification of the method of separation of variables to suit the

nonhomogeneous equation, and is known as method of eigenfunction expan-
sions. Method of eigenfunction expansions proposes a solution to the IBVP
as a superposition of eigenfunctions corresponding to homogeneous wave
equation with coefficients depending on the variable t . Since the bound-
ary conditions in (4.111) are same as in the Example 4.26, the method to
find eigenfunctions is the same and they are given by (4.108). A formal
solution to the IBVP is proposed as
X
∞
u(x, t ) ≈ T0 (t ) + Tn (t ) cos (nπx) , (4.112)
n=1

where the functions T0 , T1 , · · · are determined using the initial conditions

(4.111b), (4.111c), and the equation (4.111a).

September 12, 2021 Sivaji

250 Chapter 4. Wave equation I

Differentiating the equation (4.112) twice w.r.t. t and x, we get

X
∞
u t t (x, t ) = T000 (t ) + Tn00 (t ) cos (nπx)
n=1
X
∞
u x x (x, t ) = −π2 n 2 Tn (t ) cos (nπx) .
n=1

Substituting u t t and u x x in the equation (4.111a), we get the equations

T000 = 0,
T200 + 4π2 T2 = cos(2πt ),
Tn00 + n 2 π2 Tn = 0, n ∈ N.

Solving for Tn and substituting in (4.112) gives

t
u(x, t ) ≈ a0 + b0 t + sin(2πt ) cos(2πx)
4π
X
∞
+ (an cos(nπt ) + bn sin(nπt )) cos (nπx) .
n=1

It remains to determine the constants ai , bi using the initial conditions

(4.111b) and (4.111c). Using (4.111b), we get

1 1 X∞
+ cos(2πx) = u(x, 0) = a0 + an cos (nπx) .
2 2 n=1

From the last equation, we conclude

1 1
a0 = , a2 = , an = 0 for n 6= 0, 2.
2 2
Using (??), we get
X
∞
2 cos(2πx) = u t (x, 0) = b0 + nπbn cos (nπx) .
n=1

From the last equation, we conclude

1
b2 = , b = 0 for n 6= 2.
π n
Thus the solution to the IBVP is given by

Sivaji IIT Bombay

4.8. IBVP for one dimensional wave equation 251

 ‹
1 1 t +4
u(x, t ) ≈ + cos(2πt ) + sin(2πt ) cos (2πx) .
2 2 4π

Separation of Variables method in higher dimensions

Consider the IBVP for a wave equation in more than one space dimension
given by

u t t − c 2 ƒd u = 0 for x ∈ Ω, t > 0, (4.113a)

u(x, 0) = φ(x) for x ∈ Ω, (4.113b)
u t (x, 0) = 0 for x ∈ Ω, (4.113c)
u(x, t ) = 0 for x ∈ ∂ Ω, t ≥ 0, (4.113d)

where Ω is a bounded domain in Rd (d ≥ 2), and φ, ψ are given functions.

Method of separation of variables assumes a solution of (4.113a) is of
the form
u(x, t ) = X (x)T (t ) for x ∈ Ω, t > 0. (4.114)
Substituting the expression for u from (4.114) in the equation (4.89a)
gives
X (x)T 00 (t ) − c 2 ∆X (x)T (t ) = 0. (4.115)
On dividing both sides of the equation (4.115) with X (x)T (t ) and re-arranging
terms yields
T 00 (t ) ∆X (x)
= c2 . (4.116)
T (t ) X (x)
Note that the LHS of the equation (4.116) is a function of t only, while
the RHS is a function of x only. Thus both of them must be equal to a
constant, and let us denote it by λ. Thus we have

T 00 (t ) ∆X (x)
= c2 = λ. (4.117)
T (t ) X (x)

Thus we get an ODE and a PDE from (4.117) which are given by

λ
∆X − X = 0, T 00 − λT = 0.
c 2

September 12, 2021 Sivaji

252 Chapter 4. Wave equation I

Using the boundary condition (4.113d), we get

u(x, t ) = X (x)T (t ) = 0 for all x ∈ ∂ Ω, t > 0.

Since we do not want to find a trivial solution u ≡ 0, we cannot admit

T (t ) ≡ 0. Thus we get X (x) = 0 for all x ∈ ∂ Ω.
Thus we are led to the boundary value problem for X given by

λ
∆X − X = 0, x ∈ Ω (4.118a)
c2
X (x) = 0, x ∈ ∂ Ω. (4.118b)

Finding solutions to the eigenvalue problem (4.118) is not easy. This is one
of the limitations of the method of separation of variables in more than one
space dimensions. However, when the domain Ω has some symmetries
like invariance under some transformations, the probelm may reduce to
simpler problems which can be solved. For more details, the reader may
consult [26], [55].

Sivaji IIT Bombay

Exercises 253

Exercises
General

4.1. Any solution of ξ t − cξ x = 0 is a function of x + c t . Any solution of

ξ t + cξ x = 0 is a function of x − c t .
4.2. Let u be a twice continuously differentiable function that satisfies
the wave equation for x ∈ R and t ∈ R. That is, u t t − c 2 u x x = 0.
Prove that u is constant on a member of the family of characteristics
x − c t = constant if and only if it is constant along each member of
this family.
4.3. Let u be a classical solution to Wave equation u t t −u x x = 0 for (x, t ) ∈
R × (0, ∞). Show that

(i) for each fixed y ∈ R, the function w(x, t ) := u(x − y, t ) is also a

solution.
∂ ku
(ii) for each k ∈ N, the function w(x, t ) := ∂ x k (x, t ) is also a solu-
tion.
(iii) for each a > 0, the function w(x, t ) = u(ax, at ) is also a solu-
tion.

4.4. For d = 1, 3, find all coordinate change transformations of the x t -

space, namely Rd ×R, that are linear operators under which the wave
operator is invariant.

Cauchy problem in full space

4.5. If the Cauchy data φ, ψ are even (resp. odd or periodic) functions,
then the solution of Cauchy problem for the homogeneous wave
equation is also an even (resp. odd or periodic) function.

September 12, 2021 Sivaji

254 Chapter 4. Wave equation I

1
4.6. Evaluate u(1, 2 ), where u is the solution to the Cauchy problem

u t t − u x x = 0, x ∈ R, t > 0,

 0 if − ∞ < x ≤ −1,


 x + 1 if − 1 < x ≤ 0,
u(x, 0) = ,

 1 − x if 0 < x ≤ 1,


0 if 1 < x < ∞.

 0 if − ∞ < x < −1,

u t (x, 0) = 1 if − 1 ≤ x ≤ 1, .


0 if 1 < x < ∞.

4.7. [44] Let u be the solution of the Cauchy problem

u t t − 9u x x = 0, x ∈ R, t > 0,
¨
1 if |x| ≤ 2,
u(x, 0) = ,
0 if |x| > 2.
¨
1 if |x| ≤ 2,
u t (x, 0) = .
0 if |x| > 2.
1
(i) Evaluate u at the point (0, 6 ).
(ii) Discuss the large time behaviour of the solution, i.e., on lim t →∞ u(x0 , t )
for each fixed x0 ∈ R.

4.8. [44] Solve the following Cauchy problem for a nonhomogeneous

wave equation

u t t − u x x = t 7 , x ∈ R, t > 0,
u(x, 0) = 2x + sin x for x ∈ R,
u t (x, 0) = 0 for x ∈ R.

4.9. Solve the Cauchy problem

u t t − u x x + u = 0, x ∈ R, t > 0,
u(x, 0) = cos x for x ∈ R,
u t (x, 0) = sin x for x ∈ R.

4.10. [31] Let φ ∈ C 2 (R) and ψ ∈ C 1 (R) be such that φ ≡ ψ ≡ 0 outside

Sivaji IIT Bombay

Exercises 255

an interval [a, b ]. Let u(x, t ) be the solution of the Cauchy problem

(4.4).
(i) Show that for each fixed t > 0, the function x 7→ u(x, t ) is iden-
tically zero outside an interval.
(ii) Show that the functions F , G in the decomposition
R (4.12) for u
can be of compact support only when R ψ(s ) d s = 0.
4.11. [53] Find all the spherically symmetric solutions of the three-dimensional
wave equation.


4.12. [53] Solve the three-dimensional wave equation in R3 \ {0} ×(0, ∞)
with zero initial conditions and with the limiting condition

lim 4πr 2 u r (r, t ) = g (t ).

r →0

Assume that g (0) = g 0 (0) = g 00 (0) = 0.

4.13. Isn’t it a surprise that we proved the existence of solutions to Cauchy
problem for d = 1 followed by d = 3, and then for d = 2 via method
of descent? Where does the method of spherical means hit a road-
block when d = 2?
4.14. Show that the Cauchy problem for wave equation for x ∈ R and t > 0
is ill-posed, unlike the corresponding problem posed for x ∈ R and
0 ≤ t ≤ T for a fixed T > 0.
4.15. Let T > 0. Show that function u given by d’Alembert formula is
also a solution for x ∈ R and T < t ≤ 0 and show that the Cauchy
problem is well-posed for x ∈ R and T ≤ t ≤ 0.

Wave equation on proper subdomains

4.16. [31] Let φ, ψ, h ∈ C 2 ([0, ∞)). Solve the following initial boundary
value problem (IBVP).

u t t − u x x = 0, 0 < x < ∞, t > 0,

u(x, 0) = φ(x), u t (x, 0) = ψ(x), 0 ≤ x < ∞,
u(0, t ) = h(t ), t > 0.

from the first principles. Derive the compatibility condition on φ, ψ, h

so that the solution derived above is indeed a classical solution. State
and prove a relevant well-posedness result for the IBVP.
4.17. Let φ, ψ, h ∈ C 2 ([0, ∞)). Solve the following initial boundary value

September 12, 2021 Sivaji

256 Chapter 4. Wave equation I

problem IBVP.

u t t − u x x = 0, 0 < x < ∞, t > 0,

u(x, 0) = φ(x), u t (x, 0) = ψ(x), 0 ≤ x < ∞,
u x (0, t ) = h(t ), t > 0

from the first principles. Derive the compatibility condition on φ, ψ, h

so that the solution derived above is indeed a classical solution. State
and prove a relevant well-posedness result for the IBVP.
4.18. [31] Let α ∈ R be such that α + c 6= 0. Find the solution of the
homogeneous wave equation posed in the domain (x, t ) ∈ (0, ∞) ×
(0, ∞) such that

u(x, 0) = φ(x), u t (x, 0) = ψ(x) for x ≥0

u t (0, t ) = αu x (0, t ) for t ≥ 0,

where φ and ψ belong to the class C 2 (0, ∞) ∩ C [0, ∞) and φ(0) =

ψ(0) = 0. Show that generally no solution exists when α + c = 0.
(Hint: Use the decomposition (4.12)).
4.19. Let u be the solution to the IBVP

u t t − 4u x x = 0, 0 < x < ∞, t > 0,

u(x, 0) = x 2 , u t (x, 0) = 6x, 0 ≤ x < ∞.
u(0, t ) = t 2 , t > 0.

Evaluate u(3, 2) and u(2, 3).

4.20. Solve the IBVP

u t t − u x x = 0, 0 < x < ∞, t > 0,

u(x, 0) = x 2 , u t (x, 0) = 0, 0 ≤ x < ∞.
u(0, t ) = 0, t > 0.

How smooth is your solution?

4.21. Let φ, ψ ∈ C 2 [0, l ]. Solve the following initial boundary value prob-
lem IBVP.

u t t − u x x = 0, 0 < x < l , t > 0,

u(x, 0) = φ(x), u t (x, 0) = ψ(x), 0 ≤ x ≤ l ,
u x (0, t ) = 0, u x (l , t ) = 0, t > 0

Sivaji IIT Bombay

Exercises 257

from the first principles. Derive the compatibility condition on φ, ψ

so that the solution derived above is indeed a classical solution. State
and prove a relevant well-posedness result for the IBVP.
4.22. Let φ, ψ ∈ C 2 [0, l ]. Solve the following initial boundary value prob-
lem IBVP.

u t t − u x x = 0, 0 < x < l , t > 0,

u(x, 0) = φ(x), u t (x, 0) = ψ(x), 0 ≤ x ≤ l ,
u x (0, t ) = 0, u(l , t ) = 0, t > 0

from the first principles. Derive the compatibility condition on φ, ψ

so that the solution derived above is indeed a classical solution. State
and prove a relevant well-posedness result for the IBVP.
4.23. [55] Solve the following initial boundary value problem IBVP.
π
u t t − u x x = 0, 0 < x < , t > 0,
2
π
u(x, 0) = sin x, u t (x, 0) = 0, 0 ≤ x ≤ ,
π  2
u(0, t ) = 0, u x , t = 0, t > 0.
2
4.24. Using separation of variables method solve the following IBVP for
the damped wave equation:

u x x + 2u t − u t t = 0, 0 < x < π, t > 0,

u(x, 0) = φ(x), u t (x, 0) = 0, 0 ≤ x ≤ π,
u(0, t ) = 0, u(π, t ) = 0, t > 0.

4.25. Using separation of variables method solve the following IBVP for
the damped wave equation:

ut t + ut = ux x , 0 < x < π, t > 0,

u(0, t ) = u(π, t ) = 0, 0 ≤ t,
u(x, 0) = sin x, u t (x, 0) = 0, 0 ≤ x ≤ π.
 p p 
−t /2 3 1 3
(Ans: u(x, t ) = e cos( 2 t ) + 3 sin( 2 t ) sin x)
p

4.26. [55] Let c > 0 be such that c 2 < 1, and φ ∈ C 2 [0, 1]. Solve the fol-

September 12, 2021 Sivaji

258 Chapter 4. Wave equation I

lowing initial boundary value problem IBVP.

u t t − c 2 u x x = 0, 0 < x < 1, t > x,

u(x, x) = φ(x), u t (x, x) = 0, 0 ≤ x ≤ 1,
u(0, t ) = 0, u(1, t ) = 0, t > 0.

4.27. For each C ∈ R, verify that

uC (x, t ) = C sin πx sin πt

is a solution to the following boundary value problem for the wave

equation:

ut t = ux x , 0 < x < 1, 0 < t < 1,

u(x, 0) = 0, u(x, 1) = 0 for 0 ≤ x ≤ 1,
u(0, t ) = u(1, t ) = 0 for 0 ≤ t ≤ 1.

What are the maximum and minimum values of the solution uC in

the domain [0, 1] × [0, 1]? Where are they attained? What is your
conclusion regarding a maximum principle for Wave equation?
4.28. [27, 44]
(i) Solve the Darboux problem:

u t t − u x x = 0, t > max{x, −x}, t ≥ 0,


ϕ(t ) if x = t , t ≥ 0,
u(x, t ) =
ψ(t ) if x = −t , t ≥ 0,

where ϕ, ψ ∈ C 2 ([0, ∞)) and satisfy ϕ(0) = ψ(0).

(ii) Explain what is meant by saying that above problem is well-
posed. Is the above problem well-posed? On what domain is
it well-posed? State and prove a precise statement concerning
this.

Sivaji IIT Bombay

 Chapter 5

Wave equation II:

Qualitative properties

In this chapter, we discuss some of the important qualitative properties

of solutions to wave equation. Solutions of wave equation in one space
dimension have a special property called parallelogram identity, which can
be used to find solutions of certain initial-boundary value problems for
wave equation and is discussed in Section 5.1.
In Section 5.2, we introduce two important concepts (which are dual
to each other), namely domain of dependence and domain of influence,
which are meaningful only for hyperbolic equations. We came across these
concepts for a first order quasilinear PDE in Subsection 2.6.2. Section 5.3
discusses a causality principle which is equivalent to concepts of domains of
dependence and influence. Another equivalent formulation of domains of
dependence and influence is in terms of finite speed of propagation which
is discussed in Section 5.4. Huygens principle is studied in Section 5.6.
We study energy of a solution to wave equation in Section 5.5. Concept
of a weak solution is defined in Section 5.7, and propagation of confined
disturbances is studied in Section 5.8 in all space dimensions. Propaga-
tion of singularities in solutions of wave equation is studied in Section 5.9.
Decay of solutions of wave eqution in two and three space dimensions is
studied in Section 5.10.

5.1 Parallelogram identity

Solutions of one dimensional wave equation enjoy a special property called
Parallelogram identity.

Definition 5.1 (Characteristic parallelogram). A parallelogram in the

x t -plane is said to be a characteristic parallelogram if each one of its sides

259
260 Chapter 5. Wave equation II

t = K2 x+
x −c ct
=L
S 2
Q
x+
ct
=L K1
c t=
x−
1

Figure 5.1. characteristic parallelogram.

lies along a characteristic line.

Figure 5.1 depicts a characteristic parallelogram P QRS. Each of its

sides, namely P Q, QR, RS, S P , lies along a member of one of the two
families of characteristic lines for wave equations that are described by the
equations x − c t = constant and x + c t = constant.

Theorem 5.2 (Parallelogram identity). Let u be a function having the

form
u(x, t ) = F (x − c t ) + G(x + c t ) (5.1)
for some functions F , G defined on R. For every characteristic parallelogram
P Q RS with the line segments P R and QS as its diagonals, the values of u
at the vertices P, Q, R, S of the parallelogram satisfy the parallelogram identity

u(P ) + u(R) = u(Q) + u(S). (5.2)

Proof: Without loss of generality, assume that the side P Q lies along the
characteristic line x − c t = K1 for some K1 ∈ R, and that vertices P, Q, R, S
are described in anti-clockwise manner. Since P QRS is a characteristic
parallelogram, there exist L1 , L2 , K2 ∈ R such that the sides S P , QR, and
RS lie along the charactestic lines x + c t = L1 , x + c t = L2 , and x − c t = K2
respectively, as shown in Figure 5.1.
Since the function u is given to have the form (5.1), we have

u(P ) = F (K1 ) + G(L1 ), u(Q) = F (K1 ) + G(L2 ),

u(R) = F (K2 ) + G(L2 ), u(S) = F (K2 ) + G(L1 )

Sivaji IIT Bombay

5.1. Parallelogram identity 261

r +s

R ξ − r + s,τ + c

r
τ −2 x+
−c ct
=ξ

=ξ +c
S ξ − r, τ +
r ct τ+
x−
s

c 2s
Q ξ + s, τ + c
x+
ct τ
=ξ −c
+c ξ
τ c t=
x−
P (ξ , τ)

Figure 5.2. Constructing a characteristic parallelogram.

From the above set of equalities, the equation (5.2) follows.

Remark 5.3. Recall from (4.12) that general solution to the wave equation
is of the form (5.1) for some F , G ∈ C 2 (R). Hence parallelogram identity
is satisfied by any classical solution of the wave equation. A converse of
this result holds, and is the content of the next result.

Theorem 5.4. Let u : R2 → R be a twice continuously differentiable function

such that the parallelogram identity

u(P ) + u(R) = u(Q) + u(S) (5.3)

is satisfied for every characteristic parallelogram P QRS with P R and QS as

its diagonals. Then u solves the homogeneous wave equation u t t − c 2 u x x = 0.

Proof. Note that the parallelogram identity (5.3) is a ‘difference equation’,

and we need to establish a differential equation starting from the difference
equation. This suggests that we need to apply parallelogram identity on
an infinite number of characteristic parallelograms, and obtain the wave
equation as a limiting relation in the parallelogram identities.
Since the parallelogram identity (5.3) holds for every characteristic par-
allelogram, we need to make a clever use of it, by carefully constructing a
family of characteristic parallelograms.
Consider the family of characteristic parallelograms P QRS indexed
by r > 0, s > 0 having vertices the points P, Q, R, S given by
 s  r +s  r
P (ξ , τ), Q ξ + s , τ + , R ξ − r + s, τ + , S ξ − r, τ + ,
c c c
September 12, 2021 Sivaji
262 Chapter 5. Wave equation II

as shown in Figure 5.2.

Parallelogram identity may be written as

u(Q) − u(P ) = u(R) − u(S). (5.4)

Note that the€LHSŠ and RHS of the equation (5.4) represent differences in
1
the direction 1, c in the values of the function u (see Figure 5.2). Dividing
the both sides of the equation (5.4) by s gives

u(Q) − u(P ) u(R) − u(S)

= .
s s
Substituting for P, Q, R, S, we get
s
r +s
r

u ξ + s , τ + c − u(ξ , τ) u ξ − r + s , τ + c − u ξ − r, τ + c
= .
s s
(5.5)
Quantities on the LHS and € the Š RHS of the equation (5.5) are difference
1
quotients in the direction 1, c , at the points P and S respectively. Thus,
passing to the limit as s → 0 in the equation (5.5) yields
 ‹   ‹
1 r 1
∇u(ξ , τ) · 1, = ∇u ξ − r, τ + · 1, (5.6)
c c c
On substituting for the gradient of u on both sides of the equation (5.6),
we get
 ‹  ‹
∂ u 1∂ u ∂ u 1∂ u  r
+ (ξ , τ) − + ξ − r, τ + =0 (5.7)
∂x c∂t ∂x c∂t c
On dividing the equation
€ 1
Š(5.7) by r , the LHS becomes a difference quo-
tient in the direction 1, − c , at the point P (ξ , τ). Passing to the limit as
r → 0, we get
 ‹ ‹
∂ 1 ∂ ∂ u 1∂ u
− + (ξ , τ) = 0.
∂x c∂t ∂x c∂t
On simplification, we get

∂ 2u 1 ∂ 2u
(ξ , τ) − (ξ , τ) = 0.
∂ x2 c2 ∂ t2

Sivaji IIT Bombay

5.2. Domain of dependence, Domain of influence 263

Theorem 5.2 and Theorem 5.4 together assert the equivalence of “a

function being a solution to homogeneous wave equation” and “a func-
tion satisfying the parallelogram identity for every characteristic parallel-
ogram” in the class of twice continuously differentiable functions.

5.2 Domain of dependence, Domain of influence

Let u be a solution of the Cauchy problem (4.2) for homogeneous wave
equation. We are interested in knowing answers to the following two ques-
tions regarding the nature of the relation between solution at a point and
Cauchy data.

Question 1 Let u be a solution to the Cauchy problem (4.2) for homoge-

neous wave equation. Let (x 0 , t0 ) ∈ Rd × (0, ∞). Of course, u(x 0 , t0 )
being a solution to the Cauchy problem, would depend on the Cauchy
data φ, ψ. Does the solution u(x 0 , t0 ) depend on the values of φ(x), ψ(x)
at every x ∈ Rd ?

Question 2 Let u be a solution to the Cauchy problem (4.2) for homoge-

neous wave equation. Let x 0 ∈ Rd . The Cauchy data at x 0 , namely
φ(x 0 ), ψ(x 0 ), is expected to influence the solution u. Does the Cauchy
data at x 0 influence the solution u(x, t ) at every (x, t ) ∈ Rd × (0, ∞)?

The above questions could have been asked in the context of the Cauchy
problem for nonhomogeneous wave equation (4.1). However, as the for-
mulae for solutions to (4.1) suggest, the Cauchy data and the source term
do not interact with each other. Since the questions are concerning the
Cauchy data, the answer still depends on the part of the solution which
corresponds to the Cauchy problem (4.2) for homogeneous wave equation.
The above questions lead us to the following definition.

Definition 5.5 (domains of dependence and influence). Let u be a solu-

tion to the Cauchy problem (4.2) for homogeneous wave equation.
(i) Let (x 0 , t0 ) ∈ Rd × (0, ∞). The set of all x ∈ Rd on which the solution
u(x 0 , t0 ) depends through Cauchy data is called the domain of dependence
for the solution at the point (x 0 , t0 ).
(ii) Let x 1 ∈ Rd . The set of all (x, t ) ∈ Rd × (0, ∞) such that the solution
u(x, t ) is influenced by the Cauchy data at the point x 1 ∈ Rd is called
the domain (region) of influence of the point x 1 .

September 12, 2021 Sivaji

264 Chapter 5. Wave equation II

Remark 5.6 (Questions 1 and 2 re-phrased).

The questions posed at the beginning of this section may be re-phrased
to facilitate their better understanding.
Question 1 (re-phrased) Suppose I am standing at a point x0 ∈ Rd at time
t0 (i.e., at the point (x 0 , t0 ) in the space-time ). What is causing (or
responsible for) the current state u(x 0 , t0 ) from the past situation/data
at time t = 0?
Question 2 (re-phrased) Suppose I am standing at a point x0 ∈ Rd at the
initial time t = 0. What are the points (x, t ) in the space-time at which
the data/situation at x0 (at the current time, initial time) influences
(or) affects the future state u(x, t ) ?
In the next section, we are going to discuss these questions from the point
of view of cause and effect.
Rest of this section is devoted to determination of the domains of de-
pendence and influence in dimensions d = 1, 2, 3 by utilizing the explicit
formulae for the solutions of Cauchy problems that were derived in Sec-
tions 4.2, 4.3, 4.4.

5.2.1 One dimensional wave equation

Domain of Dependence

The d’Alembert formula for the solution of Cauchy problem for homoge-
neous wave equation is
Z
φ(x0 − c t0 ) + φ(x0 + c t0 ) 1 x0 +c t0
u(x0 , t0 ) = + ψ(s) d s .
2 2c x0 −c t0

From the above formula, it is evident that in order to compute the solu-
tion at a point (x0 , t0 ), the values of φ are needed at just two points x0 − c t0
and x0 +c t0 , and the values of ψ are needed in the interval [x0 −c t0 , x0 +c t0 ].
Thus the domain of dependence is the interval [x0 − c t0 , x0 + c t0 ]. See Fig-
ure 5.3 for an illustration.
In other words, if we consider two sets of Cauchy data (φ, ψ) and (φ1 , ψ1 )
such that φ(x) ≡ φ1 (x) and ψ(x) ≡ ψ1 (x) on the interval [x0 −c t0 , x0 +c t0 ],
then solutions of both the Cauchy problems coincide at the point (x0 , t0 ).
In fact, both the solutions coincide in the whole of characteristic triangle,
a triangle with vertices at (x0 , t0 ), (x0 − c t0 , 0), and (x0 + c t0 , 0).
In particular, changing the Cauchy data outside the interval [x0 −c t0 , x0 +
c t0 ] has no effect on the solution at the point (x0 , t0 ). That is the effect of

Sivaji IIT Bombay

5.2. Domain of dependence, Domain of influence 265

t (x0 , t0 )

x+
t0
−c

ct
0
=x

= x0
ct

+c
x−

t0
x
x0 − c t0 x0 + c t0

Figure 5.3. Wave equation in one space dimension: Domain of dependence for solution at (x0 , t0 ).

change in initial data is not felt at the point x0 for all times t ≤ t0 . Thus we
may say that the solution at (x0 , t0 ) has a domain of dependence given by
the interval [x0 − c t0 , x0 + c t0 ].

Domain of Influence

Domain of influence of a point x0 on the x-axis is precisely the set of all

points (x, t ) ∈ R×(0, ∞) whose domain of dependence contains the point
(x0 , 0). Since domain of dependence of solution at (x, t ) is the interval [x −
c t , x + c t ], the domain of influence of x0 is

{(x, t ) ∈ R × (0, ∞) : x − c t ≤ x0 ≤ x + c t } .

Let [a, b ] be an interval on the x-axis. We are interested in the domain

of influence of the set of points in the interval, and it should be union
of domains of influence of each of the points in [a, b ]. Thus domain of
influence of the interval [a, b ] turns out to be the set of all those points
(x, t ) such that the domain of dependence of the solution at (x, t ) has a
non-empty intersection with [a, b ]. That is the domain of influence of the
interval [a, b ] is given by

{(x, t ) ∈ R × (0, ∞) : x − c t ≤ b , a ≤ x + c t } .
Domain of influence of a point (x0 , 0) is illustrated in Figure 5.4, and that
of an interval [a, b ] is illustrated in Figure 5.5 respectively.

September 12, 2021 Sivaji

266 Chapter 5. Wave equation II

x+

0
=x
ct

ct
= x0

x−
t
Domain of influence of
(x0 , 0)

x
x0

Figure 5.4. Wave equation in one space dimension: Domain of influence of the point (x0 , 0).

Domain of influence of the interval [a, b ]

=b t
x+
ct

ct
=a

x−

x
a b

Figure 5.5. Wave equation in one space dimension: Domain of influence of the interval [a, b ].

5.2.2 Case of two dimensional wave equation

The solution of Cauchy problem for two dimensional wave equation is

given by
‚ Z Œ
∂ 1 φ(y1 , y2 )
u(x1 , x2 , t ) = p dy dy
∂ t 2πc D((x1 ,x2 ),c t ) c 2 t 2 − (x1 − y1 )2 − (x2 − y2 )2 1 2
Z
1 ψ(y1 , y2 )
+ p dy dy ,
2πc D((x1 ,x2 ),c t ) c 2 t 2 − (x1 − y1 )2 − (x2 − y2 )2 1 2

where D((x1 , x2 ), c t ) denotes the open disk with center at (x1 , x2 ) having
radius c t .
In this case, domain of dependence for the solution at (x1 , x2 ) is clearly
the open disk D((x1 , x2 ), c t ) having its center at (x1 , x2 ), and having radius

Sivaji IIT Bombay

5.2. Domain of dependence, Domain of influence 267

c t . Domain of influence of a point (y1 , y2 ) is given by

{(x1 , x2 , t ) : k(x1 , x2 ) − (y1 , y2 )k < c t } ,

which is the collection of all those points x which can be reached within
time t from y.

5.2.3 Case of three dimensional wave equation

The solution of Cauchy problem for three dimensional wave equation is
given by
‚ Z Œ Z
∂ 1 1
u(x, t ) = φ(y) d σ + ψ(y) d σ.
∂ t 4πc 2 t S( x ,c t ) 4πc 2 t S( x ,c t )

In this case, domain of dependence for the solution at (x, t ) is the sphere
S(x, c t ). Domain of influence of a point y is given by

{(x, t ) : kx − yk = c t } ,
which is the collection of all those points x which can be reached exactly
at time t from y.

September 12, 2021 Sivaji

268 Chapter 5. Wave equation II

5.3 Causality principle

Causality stands for “cause and effect”. What are the reasons (in the past)
that are responsible for the current state? What will be the future events for
which the current state is responsible for? These questions were answered
in Section 5.2 using the explicit formulae for solutions of Cauchy problem.
In this section, we attempt to answer the same questions without using the
formulae, and thus justifying the use of the word “qualitative properties"
in the title of this chapter. The analysis presented in this section is a typical
illustration of an a priori analysis where conclusions are drawn on solutions
without any knowledge of their existence.
In this discussion we switch-off the nonhomogeneous term in the wave
equation for reasons explained in Section 5.2.

Theorem 5.7 (Causality Principle). Let u : Rd × (0, ∞) → R be a classical

solution to the Cauchy problem for a homogeneous wave equation, i.e., u is a
solution of
€ Š
ƒd u ≡ u t t − c 2 u x1 x1 + u x2 x2 + · · · + u xd xd = 0, x ∈ Rd , t > 0.(5.8a)
u(x, 0) = φ(x), x ∈ Rd , (5.8b)
u t (x, 0) = ψ(x), x ∈R . d
(5.8c)

Let (x 0 , t0 ) ∈ Rd × (0, ∞). The value of u(x 0 , t0 ) depends only on the values
of φ and ψ in the closure of the ball B(x 0 ; c t0 ) with center at x 0 ∈ Rd , and
having a radius of c t0 , lying in Rd × {0}.

Proof. The theorem follows immediately from the formulae for the solu-
tion of (5.8), namely (4.15) for d = 1, (4.34) for d = 3, and (4.12) for d = 2.
A direct proof of this result is presented in the following subsections
without using the explicit formulae for solutions of the Cauchy problems.

5.3.1 Proof of causality principle for d = 1

Proof. [of Causality principle]

Consider a point (x0 , t0 ) ∈ R × (0, ∞). Consider the characteristic tri-
angle, which is a triangle formed by the characteristics

x − c t = x0 − c t 0 , x + c t = x0 + c t0 ,

Sivaji IIT Bombay

5.3. Causality principle 269

and the x-axis.

Fix a T such that 0 < T < t0 , and draw the line t = T . Consider the
trapezium, denoted by F , formed by the lines

x − c t = x0 − c t0 , x + c t = x0 + c t0 , t = T , t = 0.

Multiplying the wave equation u t t − c 2 u x x = 0 with u t , we see that

0 = ut t − c 2 ux x ut
 
1 2 c2 2
= u + u x − c 2 (u t u x ) x
2 t 2 t

Thus,  
1 2 c2 2
(∂ x , ∂ t ). −c u t u x , u t + u x = 0.
2
(5.9)
2 2
Integrating the equation (5.9) on the trapezium F , we get
Z  
1 2 c2 2
(∂ x , ∂ t ). −c u t u x , u t + u x d xd t = 0.
2

F 2 2

Using integration by parts formula (or divergence theorem) in the last

equation, we get

Z   Z  
1 2 c2 2 1 2 c2 2
0 = (∂ x , ∂ t ). −c u t u x , u t + u x =
2
−c u t u x , u t + u x .n d σ,
2

F 2 2 ∂F 2 2

where n is the unit outward normal to the boundary of F .

The boundary of the trapezium F consists of four lines: They are

(i) The base of the trapezium, denoted by B, given by the equation t = 0.

The outward unit normal to B is given by n = (0, −1).
(ii) A part of the characteristic, denoted by K1 , given by the equation
x + c t = x0 + c t0 . The outward unit normal to K1 is given by n =
1
p
1+c 2
(1, c).
(iii) The upper part of the trapezium, denoted by T , given by the equation
t = T . The outward unit normal to T is given by n = (0, 1).
(iv) A part of the characteristic, denoted by K2 , given by the equation
x − c t = x0 − c t0 . The outward unit normal to K2 is given by n =
1
p
1+c 2
(−1, c).

September 23, 2021 Sivaji

270 Chapter 5. Wave equation II

Thus we have
Z   Z  
1 2 c2 2 1 2 c2 2
0= −c u t u x , u t + u x .n d σ =
2
−c u t u x , u t + u x .n d σ.
2

∂F 2 2 B∪K1 ∪T ∪K2 2 2

Thus we get

Z 
1 2 c2 2
0= −c u t u x , u t + u x .n d σ
2

∂F 2 2
Z   Z c c3
1 2 c2 2 −c 2 u t u x + 2 u t2 + 2 u x2
=− ut + ux d σ + p dσ
B 2 2 K1 1 + c2
Z   Z 2 c c3
1 2 c2 2 c u t u x + 2 u t2 + 2 u x2
+ ut + ux d σ + p d σ. (5.10)
T 2 2 K2 1 + c2

Observe that
Z c c3 2 Z
−c 2 u t u x + 2 u t2 + u
2 x c
p dσ = p (u t − c u x )2 d σ ≥ 0, (5.11)
K1 1 + c2 2 1 + c2 K1

and
Z c c3 2 Z
c 2 u t u x + 2 u t2 + u
2 x c
p dσ = p (u t + c u x )2 d σ ≥ 0. (5.12)
K2 1 + c2 2 1 + c2 K2

We now conclude from (5.10), in view of (5.11) and (5.12), that

Z   Z  
1 2 c2 2 1 2 c2 2
ut + ux d σ ≤ u t + u x d σ. (5.13)
T 2 2 B 2 2
The inequality (5.13) is known as domain of dependence inequality.
Let u and v be solutions of the Cauchy problem (5.8) (for d = 1). Let us
denote by w the difference of u and v, i.e., w := u − v. Note that w solves
the wave equation (due to linearity of the wave equation), and the Cauchy
data satisfied by the functioin w are zero functions, as both u and v are
solutions of the same Cauchy problem. Thus w(x, 0) ≡ 0 ≡ w t (x, 0), and
hence on B. As a consequence of the domain of dependence inequality, we
get Z  
1 2 c2 2
w t + w x d σ ≤ 0.
T 2 2

Sivaji IIT Bombay

5.3. Causality principle 271

As a result, w t (x, T ) = w x (x, T ) = 0. Since T < t0 is arbitrary, we conclude

that
w t (x, t ) = w x (x, t ) = 0
for every (x, t ) belonging to the characteristic triangle. This implies that
w is a constant function, and since w = 0 on B, it follows that w is the
zero function inside the characteristic triangle. In particular, u(x0 , t0 ) =
v(x0 , t0 ).
This finishes proof of the theorem.

Remark 5.8 (Consequences of Domain of dependence inequality). The

above discussion shows that u(x0 , t0 ) depends solely on the values of the
Cauchy data on the base of the characteristic triangle. In other words, the
Cauchy data φ, ψ at a spatial point x0 can influence the solution only in
the region enclosed by the two characteristics starting from (x0 , 0) which
is given by

{(x, t ) ∈ R × (0, ∞) : x − c t ≤ x0 ≤ x + c t } .

5.3.2 Proof of causality principle for general d

Characteristic cone a.k.a. Light cone

Definition 5.9 (Characteristic cone, Light cone).

1. The characteristic cone (also called light cone) at a point (x 0 , t0 ) ∈ Rd ×

(0, ∞) is defined as the set

(x, t ) ∈ Rd × (0, ∞) : kx − x 0 k = c|t − t0 | , (5.14)

where kx − x 0 k is the euclidean distance in Rd between x and x 0 .

2. The characteristic cone at a point (x 0 , t0 ) ∈ Rd × (0, ∞) together with its
interior is called the solid light cone. That is, solid light cone is defined as
the set 
(x, t ) ∈ Rd × (0, ∞) : kx − x 0 k ≤ c|t − t0 | ,

Remark 5.10 (On light cone). (i) Note that the definition of a charac-
teristic cone (given in Definition 5.9) is consistent with the definition

September 23, 2021 Sivaji

276 Chapter 5. Wave equation II

Remark 5.11 (Consequences of Domain of dependence inequality).

(i) The solid backward cone is called the past history of the vertex (x 0 , t0 ).
(ii) In other words, the Cauchy data φ, ψ at a spatial point x 0 can influence
the solution only in the future cone with vertex at (x 0 , 0), which is the
forward solid light cone emanating from (x 0 , 0).

See Figure 5.6 for a pictorial presentation of past and future cones located
at a point (x 0 , t0 ).

5.4 Finite speed of propagation

Finite speed of propagation is a common feature for wave equation in all
dimensions. We are going to study the speed of propagation of the Cauchy
data. Hence we switch-off the nonhomogeneous term from the Wave equa-
tion. We illustrate with examples.

Example 5.12 (d = 1). Let us consider initial data φ and ψ that are zero
outside of the interval (−3, −2). Thus u(0, 0) = 0 as u(0, 0) = φ(0) = 0.
Let us now study the behaviour of u(0, t ) for t > 0. For t > 0 such that
2
−c t > −2, i.e., t < c , u(0, t ) = 0. That is the information (Cauchy data)
1
at t = 0 has not reached the point x0 = 0 till t1 = c , from which time the
2
information will be received at x0 = 0. Thus it took a time of t1 = c to
travel a distance of 2, and thus the speed is c. We illustrate this with c = 1
in Figure. 5.7, and u(0, t ) remains zero till t = 2.

Example 5.13 (d = 2, 3). Let us consider the case of d = 2, 3. Suppose that

the Cauchy data φ and ψ is supported in the ball of radius R with center at
origin B(0, R) ⊂ Rd . We know that u(x, t ) depends on the values of initial
data on B[x, c t ]∩B(0, R). If this intersection is empty, then u(x, t ) = 0. In
other words, for each fixed t > 0 the support of the function x 7→ u(x, t )
is contained in ∪y ∈B(0,R) B[y, c t ], which is nothing but B(0, R + c t ). Thus
for each fixed t the support of the solution u(., t ) is a compact set if the
Cauchy data is compactly supported. In other words the support spreads
with finite speed. Let y ∈ / B(0, R). Then not only u(y, 0) = 0 but also
k y k−R
u(y, t ) = 0 for t < c . This is referred to as finite speed of propagation.

Sivaji IIT Bombay

5.5. Conservation of energy 277

t
4

x
−4 −3Support−2 −1 0 1 2 3 4
of φ, ψ

Figure 5.7. Wave equation in one space dimension: finite speed of propagation.

5.5 Conservation of energy

In this section we will prove that the energy associated to the Cauchy prob-
lem for wave equation defined by
Z  
1 2 c2
E(t ) := u t + k∇uk d x2

Rd 2 2

is a constant function. In other words, energy is conserved.

Theorem 5.14. Let the Cauchy data φ and ψ be compactly supported func-
tions defined on Rd . Let u be a solution to the Cauchy problem for the homo-
geneous wave equation. Then
Z  
d 1 2 c2
u + k∇uk d x = 0.
2
(5.23)
d t Rd 2 t 2

Proof. As in the proof of causality principle, on multiplying the homoge-

neous wave equation (5.8a) with u t , and re-arranging the terms, we get

X d
∂ € 2 Š ∂ 1 c2

−c u t u xi + u + k∇uk = 0
2 2

i =1 ∂ xi ∂t 2 t 2

September 23, 2021 Sivaji

278 Chapter 5. Wave equation II

Integrating the last equality over Rd , we get

d Z Z  
X ∂ € 2 Š ∂ 1 2 c2
− c u t u xi d x + u + k∇uk d x = 0. (5.24)
2

i=1 Rd ∂ xi Rd ∂ t 2 t 2

The first term in (5.24) is equal to zero under the condition that for each
fixed t , the function x 7→ u(x, t ) is identically equal to zero for sufficiently
large values of |x| i.e., the function u(., t ) is of compact support for each
fixed t > 0. This is always the case if the initial data φ, ψ are compactly
supported functions.
In such a case, the equation will reduce to
Z  
∂ 1 2 c2
u + k∇uk d x = 0.
2
(5.25)
Rd ∂ t 2 t 2

The last equation (5.25) is nothing but the desired equation (5.23). Thus
the function E(t ) is a constant function. In other words, the energy is
conserved.

5.6 Huygens principle

Huygens principle is concerned with the propagation of information in
space-time. More precisely, Huygens principle is concerned with solutions
of the Cauchy problem for homogeneous wave equation in which the ini-
tial data is assumed to be compactly supported (see Remark ?? for an ex-
planation on this restriction), and how this support propagates with time
under the dynamics of the wave equation.
We present Huygens principle in the two dual points-of-view, namely,
in terms of domains of dependence, and influence, see Definitions 5.15 and
5.15.
There are two forms of Huygens principle in the literature, known as
weak, and strong forms of Huygens principle respectively. Several authors
present the Strong form as the Huygens principle.

5.6.1 Strong form of Huygens principle

Huygens observed that if a wave is sharply localized at some time, then

it will continue to be so for all later times, when the wave propagation is
governed by homogeneous wave equation in space whose dimension is an
odd number greater than or equal to 3. This observation does not hold

Sivaji IIT Bombay

Exercises 303

Exercises
General

5.1. In Example 5.26 we extended the radial functions f , g as continu-

ously differentiable even functions to R. Justify the compatibility
conditions that f , g must satisfy to ensure that such extensions are
possible.

Parallelogram identity

5.2. [31] Let φ, ψ, h ∈ C 2 ([0, ∞)). Solve the following initial boundary
value problem (IBVP).

u t t − u x x = 0, 0 < x < ∞, t > 0,

u(x, 0) = φ(x), u t (x, 0) = ψ(x), 0 ≤ x < ∞,
u(0, t ) = h(t ), t > 0.

using parallelogram identity. (Hint: Consider the two cases x −t ≥ 0

and x − t ≤ 0 separately). Derive the compatibility condition on
φ, ψ, h so that the solution derived above is indeed a classical solu-
tion. State and prove a relevant well-posedness result for the IBVP.
Compare your answer with Exercise 4.16 of Chapter 4.
5.3. Using parallelogram identity, solve the following IBVP

u t t = u x x , 0 < x < l , t > 0,

u(x, 0) = φ(x), u t (x, 0) = ψ(x) for 0 ≤ x < l ,
u(0, t ) = u(l , t ) = 0 for 0<t <∞

for (x, t ) belonging to the region marked 2, 1 in the Figure 4.3 by

deriving an expression for the solution in terms of the given data.
Also explain the reasons for using notation 2, 1 in terms of reflections.

Domains of dependence and influence

5.4. [57] Let u be a solution of the d -dimensional wave equation such that
u(x, 0) = u t (x, 0) = 0 for x ∈ B(v, R). Up to what time t can you be
sure that u(v, t ) = 0?
5.5. [57] Let u be a solution of the two-dimensional wave equation such
that u(x, 0) = u t (x, 0) = 0 for x ∈
/ B(0, 1). Up to what time t can you

September 23, 2021 Sivaji

304 Chapter 5. Wave equation II

be sure that u(x, t ) = 0 for x ∈ { (5, 0), (0, 10), (2, 3) }?

5.6. In the context of the IBVP (4.64), What is the domain of dependence
for a point (x0 , t0 ) ∈ (0, l ) × (0, ∞)? What is the range of influence of
a subinterval [a, b ] of (0, l )?
5.7. Find the domain of dependence for any point (x0 , t0 ) with t0 > |x0 |
for the Darboux problem:

u t t − u x x = 0, t > max{x, −x}, t ≥ 0,


ϕ(t ) if x = t , t ≥ 0,
u(x, t ) =
ψ(t ) if x = −t , t ≥ 0,

where ϕ, ψ ∈ C 2 ([0, ∞)) and satisfy ϕ(0) = ψ(0).

Conservation of energy

5.8. [53] Derive the conservation of energy for the wave equation in a
bounded domain Ω with Dirichlet or Neumann boundary condi-
tions.
5.9. [53] Show that energy defined by the formula
Z  
1 2 c2
E(t ) := u t + k∇uk d x
2

Ω 2 2

decreases if u solves the boundary value problem with the boundary

∂u ∂u
condition ∂ n + b ∂ t = 0 where b > 0.
5.10. [52] Let φ, ψ be twice continuously differentiable functions. Let u
be a solution to the following initial boundary value problem:

u t t − u x x + u t = 0, 0 < x < l , t > 0,

u(x, 0) = φ(x), u t (x, 0) = ψ(x), 0 ≤ x ≤ l ,
u(0, t ) = 0, t ≥ 0, u(l , t ) = 0, t ≥ 0

Prove that the energy defined by

Z
1 l


E(t ) := u t2 + u x2 d x
2 0

is a decreasing function of t .
5.11. Using the energy conservation for the wave equation (Theorem 5.14),
show that the Cauchy problem for wave equation has at most one so-
lution.

Sivaji IIT Bombay

Exercises 305

5.12. Let φ, ψ be twice continuously differentiable functions. Show that

the initial boundary value problem

u t t − u x x = u t − u 3 , 0 < x < l , t > 0,

u(x, 0) = φ(x), u t (x, 0) = ψ(x), 0 ≤ x ≤ l ,
u(0, t ) = 0, u(l , t ) = 0, t ≥ 0

has at most one classical solution.

5.13. Let φ, ψ, a, f be smooth functions defined on the interval [0, l ]. As-
sume that V (x) ≥ 0 for every x ∈ [0, l ]. Show that the initial bound-
ary value problem

u t t − u x x + a(x)u = f (x), 0 < x < l , t > 0,

u(x, 0) = φ(x), u t (x, 0) = ψ(x), 0 ≤ x ≤ l ,
u(0, t ) − u x (0, t ) = 0, u(l , t ) + u x (l , t ) = 0, t ≥ 0

has at most one classical solution.

5.14. Let φ, ψ be twice continuously differentiable functions, and α ∈ R.
Show that the initial boundary value problem

u t t − u x x + αu t = f (x, t ), 0 < x < l , t > 0,

u(x, 0) = φ(x), u t (x, 0) = ψ(x), 0 ≤ x ≤ l ,
u(0, t ) = 0, u(l , t ) = 0, t ≥ 0

has at most one classical solution.

5.15. Let u : R×(0, ∞) → R be a classical solution to the Cauchy problem

u t t − u x x + u = 0, x ∈ R, t > 0,
u(x, 0) = 0, u t (x, 0) = 0, x ∈ R

such that for each fixed t , x 7→ u(x, t ) is a function having compact

support. Show that u ≡ 0. (Hint: Use a suitable energy functional)

Weak solutions

5.16. If F ∈ C (R2 ), then the function u defined by u(x, t ) = F (x − c t ) is a

weak solution to the wave equation u t t = u x x .

September 23, 2021 Sivaji