MA203: Module 3 - Partial Differential Equations

Debopriya Mukherjee

These notes have largely been taken from the textbook Advanced Engineering Math-
ematics by Erwin Kreyszig (10th Edition, John Wiley & Sons, 2011).

i
1 Wave equation

1.0.1 d’Alembert’s solution of the one-dimensional wave equation

Consider the one-dimensional wave equation
utt = c2 uxx . (1)
Let us consider the change of variables ξ = x − ct and η = x + ct. This implies that
∂ξ ∂ξ ∂η ∂η
= 1, = −c, = 1, = c.
∂x ∂t ∂x ∂t
We assume that all the partial derivatives of u involved are continuous, and apply the
chain rule to obtain
∂ξ ∂η
ux = uξ + uη = uξ + uη ,
∂x ∂x
∂ξ ∂η
ut = uξ + uη = uξ (−c) + uη × c = c(−uξ + uη ),
∂t ∂t
∂(uξ + uη ) ∂ξ ∂(uξ + uη ) ∂η
uxx = + = uξξ + uξη + uηξ + uηη = uξξ + 2uξη + uηη ,
∂ξ ∂x ∂η ∂x
 
∂(−uξ + uη ) ∂ξ ∂(−uξ + uη ) ∂η  
utt = c + = c (−uξξ + uξη )(−c) + (−uηξ + uηη )c
∂ξ ∂t ∂η ∂t
2
= c (uξξ − 2uξη + uηη ).
Substituting these in the one-dimensional wave equation (1), we obtain
c2 (uξξ − 2uξη + uηη ) = c2 (uξξ + 2uξη + uηη ) =⇒ uξη = 0

We shall now integrate both sides of the above equation with respect to ξ and η succes-
sively. On integrating with respect to ξ, we obtain
uη = ψ̄(η),
where ψ̄ is an arbitrary function of η only. Next, integrating this equation with respect
to η, we obtain
Z
u(ξ, η) = ψ̄(η) dη + φ(ξ),
Z
where φ is an arbitrary function of ξ only. Writing ψ̄(η) dη as ψ(η), we obtain the
solution of the wave equation as
u(ξ, η) = φ(ξ) + ψ(η), (2)
where φ and ψ are two arbitrary functions of ξ and η, respectively. Rewriting ξ and η in
terms of x and t, we obtain from (2)

u(x, t) = φ(x − ct) + ψ(x + ct) (3)

This is referred to as the d’Alembert’s solution of the one-dimensional wave equation (1).

1
Example (d’Alembert’s solution satisfying the initial conditions): Determine the solution
of the wave equation
utt = c2 uxx , t > 0, −∞ < x < ∞ (4)
in an infinite one-dimensional medium subject to the initial conditions
u(x, 0) = f (x), ut (x, 0) = g(x) for − ∞ < x < ∞. (5)
Solution: We know that the general solution of the wave equation (4) is given by the
d’Alembert’s solution
u(x, t) = φ(x − ct) + ψ(x + ct). (6)
Therefore,
ut (x, t) = −c φ0 (x − ct) + c ψ 0 (x + ct), (7)
where prime denotes the derivatives with respect to the entire arguments x−ct and x+ct,
respectively.
Let us now apply the given initial conditions. The condition u(x, 0) = f (x) implies
φ(x) + ψ(x) = f (x) (8)
and ut (x, 0) = g(x) implies the condition
−c φ0 (x) + c ψ 0 (x) = g(x). (9)
Dividing the above equation by (−c) and integrating with respect to x, we obtain
1 x
Z
φ(x) − ψ(x) = − g(s) ds + k, (10)
c x0
where k = φ(x0 ) − ψ(x0 ). Adding (8) and (10), and dividing the resulting equation by 2,
we obtain
1 x
Z
1 k
φ(x) = f (x) − g(s) ds + . (11)
2 2c x0 2
Substituting this in (8), we obtain
Z x
1 1 k
ψ(x) = f (x) + g(s) ds − . (12)
2 2c x0 2
From (11) and (12),
1 x−ct
Z
1 k
φ(x − ct) = f (x − ct) − g(s) ds + , (13)
2 2c x0 2
1 x+ct
Z
1 k
ψ(x + ct) = f (x + ct) + g(s) ds − . (14)
2 2c x0 2
From the above equations, the solution of the given problem is
u(x, t) = φ(x − ct) + ψ(x + ct)
Z x0 Z x+ct 
1  1
= f (x − ct) + f (x + ct) + g(s) ds + g(s) ds
2 2c x−ct x0
or
Z x+ct
1  1
u(x, t) = f (x − ct) + f (x + ct) + g(s) ds (15)
2 2c x−ct

2
Remark: Notice that if the initial velocity is zero, solution (15) reduces to

1 
u(x, t) = f (x − ct) + f (x + ct) .
2

Example (d’Alembert’s solution satisfying the initial conditions): Determine the solution
of the wave equation

utt = c2 uxx , t > 0, −∞ < x < ∞ (16)

in an infinite one-dimensional medium subject to the initial conditions

u(x, 0) = 0, ut (x, 0) = sin x for − ∞ < x < ∞. (17)

Solution: From the previous example, the general solution of the wave equation (16)
with the initial conditions u(x, 0) = f (x) and ut (x, 0) = g(x) is given by the d’Alembert’s
solution

1 x+ct
Z
1 
u(x, t) = f (x − ct) + f (x + ct) + g(s) ds (18)
2 2c x−ct

Here, f (x) = 0 and g(x) = sin x. Therefore, the solution for the given problem is
Z x+ct
1 1 x+ct 1h i
u(x, t) = sin s ds = (− cos s) = cos (x − ct) − cos (x + ct) .
2c x−ct 2c x−ct 2c

Example: Let y(x, t) represents transverse displacement in a long stretched string one
end of which is attached to a ring (of negligible diameter and weight) that can slide along
the y-axis. The other end is so far out on the positive x-axis that it may be considered
to be infinitely far from the origin. The ring is initially at the origin and is then moved
along the y-axis (see figure 1) so that y = f (t) when x = 0 and t ≥ 0, where f is a
given continuous function with f (0) = 0. Assume that the string is initially at rest on
the x-axis.

Figure 1: Schematic of the string in the given problem.

(a) Write down the boundary value problem describing the given problem.
Hint: Look at the figure carefully; the boundary value problem should include ‘a’ in
it.

3
(b) The general solution of the partial differential equation identified in part (a) is given
by y(x, t) = φ(x+at)+ψ(x−at), where φ and ψ are two arbitrary functions that need
to be determined using the given conditions. Apply the initial condition(s) identified
in part (a) to show that there is a constant c such that

φ(x) = c and ψ(x) = −c (x ≥ 0).

Then apply the boundary condition(s) identified in part (a) to show that
x
ψ(−x) = f −c (x ≥ 0),
a
where c is the same constant.

(c) With the aid of the results obtained in part (b), show that the solution of given
problem is

0
 x ≥ at,
y(x, t) =
f t − x
 
 x ≤ at.
a

(d) What can you infer from this solution about the displacement in the string due to
the movement of the ring?

Solution: (a) The boundary value problem, which describes the given problem is as
follows.

ytt (x, t) = a2 yxx (x, t) (x > 0, t > 0), (19)

y(x, 0) = 0 and yt (x, 0) = 0 (x ≥ 0), (20)
y(0, t) = f (t) (t ≥ 0), (21)

where y(x, t) is the transverse displacement in the string and a is the wave speed in
the string. The initial conditions are given by (20) and the boundary condition by
(24).

(b) The general solution of (19) is given by (the d’Alembert’s solution)

y(x, t) = φ(x + at) + ψ(x − at). (22)

This implies that

yt (x, t) = a φ0 (x + at) − a ψ 0 (x − at) = a φ0 (x + at) − ψ 0 (x − at) .

 
(23)

Let us now apply the initial conditions (20), which hold for x ≥ 0.

y(x, 0) = 0 =⇒ φ(x) + ψ(x) = 0 =⇒ φ(x) = −ψ(x), (24)

a φ0 (x) − ψ 0 (x) = 0 φ0 (x) = ψ 0 (x).
 
yt (x, 0) = 0 =⇒ =⇒ (25)

Differentiating (24) with respect to x and adding in (25), we obtain

φ0 (x) = 0 =⇒ φ(x) = c,

4
 where c is a constant of integration. With this, eq. (24) implies ψ(x) = −c. Thus,
applying after applying the initial conditions (20), we have

φ(x) = c and ψ(x) = −c (x ≥ 0). (26)

Let us now apply the boundary condition (21), which holds for t ≥ 0.

y(0, t) = f (t) =⇒ φ(at) + ψ(−at) = f (t).

Let us apply the change of variable at = x to obtain

x
φ(x) + ψ(−x) = f (x ≥ 0).
a

After applying the initial conditions, we obtained φ(x) = c. Therefore, the above
equation yields
x
ψ(−x) = f −c (x ≥ 0). (27)
a

(c) The results obtained in part (b) above are (26) and (27) and they hold for x ≥ 0.
Note from the general solution (22) that we need to determine the values of φ(x + at)
and ψ(x − at). From (26) and (27), it is clear that φ(x) = c for all x ≥ 0. Therefore,
φ(x + at) = c for all x ≥ 0 because x + at ≥ x ≥ 0.

Now, we need to determine ψ(x − at) from (26) and (27). From (26), we have

ψ(x − at) = −c (x − at ≥ 0 or x ≥ at) (28)

and from (27), we have

 
at − x  x
ψ(x − at) = f −c=f t− −c (at − x ≥ 0 or x ≤ at). (29)
a a
Consequently, the general solution y(x, t) = φ(x+at)+ψ(x−at) of the above problem
is

y(x, t) = c + (−c) = 0 for x ≥ at

and
h  x i  x
y(x, t) = c + f t − −c =f t− for x ≤ at.
a a
Combining the above two results, the solution of the given problem is

0
 x ≥ at,
y(x, t) =
f t − x
 
 x ≤ at.
a

(d) The solution of the given problems reveals that the part of the string to the right of the
point x = at on the x-axis is unaffected by the movement of the ring prior to time t,
as also shown in figure 1.

5
 [Lecture 10]
1.1 Vibration of a circular membrane
Circular membranes are encountered in many engineering applications, such as in drums,
pumps, microphones, etc. Whenever a circular membrane is plane and its material is
elastic, but offers no resistance to bending (e.g., not a metallic membrane!), its vibrations
are governed by the two-dimensional wave equation. Since the membrane is circular, it is
convenient to use the polar coordinates defined by x = r cos θ and y = r sin θ.
The two-dimensional wave equation in the polar coordinates reads
∂ 2u
 2
1 ∂ 2u

2 ∂ u 1 ∂u
=c + + . (30)
∂t2 ∂r2 r ∂r r2 ∂θ2
We shall consider a membrane of radius R with fixed end (figure 2) and determine solutions
u(r, t) that are radially symmetric (i.e., those solutions which do not depend on θ). In
this case, uθθ = 0 and the two-dimensional wave equation (30) reduces to
∂ 2u
 2 
2 ∂ u 1 ∂u
=c + . (31)
∂t2 ∂r2 r ∂r

Figure 2: Circular membrane

Boundary Condition: Since the membrane is fixed along the boundary r = R, we

have the boundary condition

u(R, t) = 0, for all t ≥ 0. (32)

Initial Conditions: We can obtain radially symmetric solutions only if the initial con-
ditions do not depend on θ. Let us assume that the initial deflection in the membrane is
f (r) and the initial velocity of the membrane is g(r). Therefore, the initial conditions are

u(r, 0) = f (r) and ut (r, 0) = g(r), 0 ≤ r ≤ R. (33)

6
We would like to solve the reduced wave equation (31) along with the boundary condition
(32) and the initial conditions (33) using the method of separation of variables.
Solution: First Step: To find the ordinary differential equations:
Let the solution of this problem be u(r, t) = W (r) T (t). Therefore, it satisfies (31).
Substituting this ansatz in (31), we obtain
 
2 00 1 0
W T̈ = c W T + W T , (34)
r
where the time derivative has been denoted with dots while the spacial derivative has
been denoted by primes. The above equation can be written as
 
1 T̈ 1 00 1 0
= W + W . (35)
c2 T W r
Again, the left-hand side of the above equation is a function of t alone while the right-
hand side is a function of r alone. Therefore, both of them must be equal to a constant,
let us say, it is k. This gives two ordinary differential equations:

d2 T 1
− kc2 T = 0 and W 00 + W 0 − kW = 0. (36)
dt2 r
The equation for T (t) has solutions which grow or decay exponentially for k > 0, are
linear or constant for k = 0, and are periodic for k < 0. Physically, it is expected that
a solution to the problem of a vibrating membrane will be oscillatory in time, and this
leaves only the third case k < 0; let k = −β 2 , β 6= 0. With this, the above ordinary
differential equations become

d2 T
+ λ2 T = 0 and rW 00 + W 0 + β 2 rW = 0, (37)
dt2
where λ = βc. The equation for W can be reduced to the Bessel equation, which is
x2 y 00 + xy 0 + (x2 − ν 2 )y = 0 by a scaling s = βr. With this scaling,
dW dW ds dW
W0 = = =β ,
dr ds 
dr  ds
d2 W 2
 
00 d dW ds d dW 2d W
W = = = β β = β ,
dr2 ds dr dr ds ds ds2
and the equation for W becomes

d2 W dW 2
2 2d W dW
β 2r + β + β 2
rW = 0 =⇒ β r + βr + β 2 r2 W = 0
ds2 ds ds2 ds
d2 W dW
=⇒ s2 2 + s + s2 W = 0, (38)
ds ds
which is the Bessel equation with ν = 0.

Second Step: Satisfying the boundary condition:

The boundary condition u(R, t) = 0 leads to W (R) T (t) = 0 and, hence, to

W (R) = 0 (39)

7
because T (t) = 0 will result into the zero solution which is meaningless.
Solution of the Bessel equation (38) are the Bessel functions J0 (s) and Y0 (s) of the
first and second kind, respectively. It turns out that Y0 (s) becomes infinite at s = 0;
therefore Y0 (s) cannot be a part of the solution because the deflection of the membrane
must always be finite. This leaves us with the solution W (s) = J0 (s) or, in other words,

W (r) = J0 (βr). (40)

The boundary condition (39) implies that

J0 (βR) = 0 (41)

We can satisfy this condition because J0 (s) has infinitely many positive zeros, s =
α1 , α2 , α3 , . . . (see figure 3), with numerical values

α1 = 2.4048, α2 = 5.5201, α3 = 8.6537, α4 = 11.7915, α5 = 14.9309

and so on. These zeros are slightly irregularly spaced, as we can see in the figure.

Figure 3: Bessel function J0 (s)

Equation (41) now implies that

αn
βR = αn =⇒ β = βn = , n = 1, 2, 3, . . . (42)
R
Hence, the functions
α 
n
Wn (r) = J0 (βn r) = J0 r , n = 1, 2, 3, . . . (43)
R
are solutions of (37) that vanish at r = R.

Eigenfunctions and eigenvalues: For Wn in (43), a corresponding general solution

of (37)1 with λ = λn = βn c = αn c/R is

Tn (t) = cn cos λn t + dn sin λn t. (44)

Hence the functions

 
un (x, t) = Wn (r) Tn (t) = cn cos λn t + dn sin λn t J0 (βn r) (45)

with n = 1, 2, 3, . . . are solutions of the wave equation (31) satisfying the boundary con-
dition (32). These are the eigenfunctions of our problem. The corresponding eigenvalues
are λn .

8
Figure 4: Normal modes of the circular membrane in the case of vibrations independent
of the angle θ

Since the zeros of the Bessel function J0 are not regularly spaced on the axis, the
sound of a drum is entirely different from that of a violin. The forms of the normal modes
can be easily obtained from figure 3 and are shown in figure 4.
Third Step: Solution of the entire problem: From (45), the most general solution
of the wave equation (31) that satisfies the given boundary condition (32) is
X∞ X∞   α 
n
u(r, t) = un (r, t) = cn cos λn t + dn sin λn t J0 r . (46)
n=1 n=1
R
Let us now apply the initial conditions (33). The initial condition u(r, 0) = f (r) gives
X∞ α 
n
f (r) = cn J0 r . (47)
n=1
R
To obtain the coefficients cn , we shall use the orthogonality of the Bessel functions, which
is given by

Z 1 0 if a 6= b
x Jp (ax) Jp (bx) dx = 1 (48)
0  J 2 (a) if a = b,
2 p+1
where p is a non-negative integer, and a and b are the zeros of Jp (x).
To use the orthogonality of the Bessel functions for the problem under consideration,
let us replace p with 0, x with r/R, a with αm and b with αn . With this, the above
orthogonality relation changes to

Z R α  α  0 if m 6= n
m n
r J0 r J0 r dr = 1 (49)
0 R R  R2 J 2 (α ) if m = n.
m
1
2

9
Now, to find the coefficients cn in (47), we shall use the

orthogonality relation (49). For
αm
that, let us multiply both sides of (47) with rJ0 R r for some fixed m (m = 1, 2, 3, . . . )
and integrate both sides of the resulting equation with respect to r in (0, R). Using the
orthogonality relation (49), the only term that will be nonzero in the right-hand side will
be for n = m and we would have
Z R α 
m 1
r f (r) J0 r dr = R2 J12 (αm )cm , m = 1, 2, 3, . . . (50)
0 R 2
or
Z R
2 α 
m
cm = 2 2 r f (r) J0 r dr m = 1, 2, 3, . . . (51)
R J1 (αm ) 0 R

Let us now apply the remaining initial condition ut (r, 0) = g(r). For that we first differ-
entiate solution (46) partially with respect to t to obtain
∞
X   α 
n
ut (r, t) = λn − cn sin λn t + dn cos λn t J0 r . (52)
n=1
R

The initial condition ut (r, 0) = g(r) gives

∞
X α 
n
g(r) = λn dn J0 r . (53)
n=1
R

To obtain the coefficients dn , let us multiply both sides of (53) with rJ0 αRm r for some

fixed m (m = 1, 2, 3, . . . ) and integrate both sides of the resulting equation with respect
to r in (0, R). Using the orthogonality relation (49), the only term that will be nonzero
in the right-hand side will be for n = m and we would have
Z R α 
m 1
r g(r) J0 r dr = R2 J12 (αm )λm dm , m = 1, 2, 3, . . . (54)
0 R 2
or
Z R
2 α 
m
dm = 2
r g(r) J0 r dr m = 1, 2, 3, . . . , (55)
αm cRJ1 (αm ) 0 R

where the relation λm = αm c/R has been used. Therefore, the deflection in a (radially
symmetric) vibrating membrane fixed at the boundary and satisfying the initial conditions
(33) is given by
∞ h
X α c   α c i  α 
n n n
u(r, t) = cn cos t + dn sin t J0 r (56)
n=1
R R R

where
Z R
2 α
n

cn = 2 2 r f (r) J0 r dr n = 1, 2, 3, . . . (57)
R J1 (αn ) 0 R
and
Z R
2 α
n

dn = r g(r) J0 r dr n = 1, 2, 3, . . . (58)
αn cRJ12 (αn ) 0 R

10