M344 - ADVANCED ENGINEERING MATHEMATICS

Lecture 11: Sturm-Liouville Boundary-Value Problems

When solving partial differential equations it will often be necessary to

approximate functions by series of orthogonal functions. One way to obtain
an orthogonal family of functions is by solving a particular type of boundary-
value problem for a second-order linear ordinary differential equation. Up until
now you have dealt only with initial-value problems (IVPs) where x(t0 ) and
x0 (t0 ) are both given at the same value of t.

Def 1 A non-singular Sturm-Liouville boundary-value problem consists

of a second-order linear differential equation of the form
d
(p(t)y 0 (t)) + (λw(t) − q(t))y(t) = 0, (1)
dt
with p, p0 , and w continuous functions, q at least piecewise continuous, on the
interval a ≤ t ≤ b, and p and w both positive functions on [a, b]. The boundary
conditions assumed for y(t) are called homogeneous unmixed boundary
conditions, and are given at two points a and b, in the form:
½
c1 y(a) + c2 y 0 (a) = 0
(2)
c3 y(b) + c4 y 0 (b) = 0

There are other types of Sturm-Liouville problems, where the equation has
one or more singular points, or the boundary conditions are mixed or periodic,
but the above definition describes the type of problem that occurs most often
in the partial differential equations we will be solving. For more information
on Sturm-Liouville problems you should read the relevant chapter in Nagel &
Saff.

The trivial solution y(t) ≡ 0 always satisfies (1) and (2), but we are looking
for non-zero solutions. It can be shown that non-zero solutions of a Sturm-
Liouville problem only exist if the parameter λ belongs to a certain set of
real numbers λn , called eigenvalues of the particular Sturm-Liouville prob-
lem. There is a smallest eigenvalue λ1 , with λ1 < λ2 < · · · < λn < · · · and
limn→∞ λn = ∞. When λ = λn , the corresponding solution y(t) = φn (t) of
equation (1) is called an eigenfunction corresponding to the eigenvalue λn ;
and each φn is unique up to constant multiples.

Theorem 1 The infinite family of eigenfunctions S = {φ1 , φ2 , · · · , φn , · · · } of

a Sturm-Liouville problem (1) with boundary conditions (2)is an orthogonal

1
family of functions on the interval [a, b], with respect to the weight function
w(t) in equation (1); that is
Z b
φn (t)φm (t)w(t)dt = 0, if m 6= n.
a

Example 1 Show that the boundary-value problem

y 00 (t) + λy(t) = 0, y(0) = 0, y 0 (1) = 0 (3)

is a Sturm-Liouville problem, and find all eigenvalues and a corresponding set

of orthogonal eigenfunctions.

We can write y 00 + λy = 0 in the form (1 · y 0 )0 + λ · 1 · y = 0, so this is

a Sturm-Liouville equation with p(t) = w(t) ≡ 1, and q(t) ≡ 0. The given
boundary conditions can be written in the form
½
1 · y(0) + 0 · y 0 (0) = 0
. (4)
0 · y(1) + 1 · y 0 (1) = 0

To find all of the non-trivial solutions, we first determine the general solution
of the differential equation, and then check to see for which values of λ the
given boundary conditions can be satisfied. Since the characteristic equation
of the differential equation in (3) is r2 + λ = 0, there will be three different
cases depending on whether the roots are real and unequal, real and equal, or
complex conjugates. We must consider each of these three cases separately.

Case 1: λ < 0.
Let λ = −K 2 for some non-zero real number K. Then the roots of the
characteristic polynomial r2 − K 2 = 0 are ±K. The general solution in this
case can be written as y(t) = c1 eKt +c2 e−Kt , but we will find it more convenient
to write it in the equivalent form y(t) = A cosh(Kt) + B sinh(Kt). Now,
to satisfy the two boundary conditions, we need to find A and B such that
y(0) = A cosh(K · 0) + B sinh(K · 0) = A = 0 and y 0 (1) = AK sinh(K · 1) +
BK cosh(K · 1) = 0. Since A has to be zero, and K 6= 0, the condition on y 0 (1)
implies that B = 0. This means that the only solution is the trivial solution
y(t) ≡ 0; therefore, there are no negative eigenvalues.

Case 2: λ = 0.
In this case, the characteristic polynomial is r2 = 0, with a double root
r = 0. The general solution is y(t) = c1 e0·t + c2 te0·t = c1 + c2 t, with derivative
y 0 (t) = c2 . To satisfy the two boundary conditions, y(0) = c1 = 0 and y 0 (1) =

2
c2 = 0, so again the only solution is the zero solution; therefore, λ = 0 is not
an eigenvalue.

Case 3: λ > 0.
Assume λ = K 2 for some non-zero K. This is the case where the character-
istic polynomial r2 + K 2 = 0 has complex conjugate roots r = ±Ki. The gen-
eral solution in this case is y(t) = c1 e0·t cos(Kt) + c2 e0·t sin(Kt) = c1 cos(Kt) +
c2 sin(Kt) with derivative y 0 (t) = −Kc1 sin(Kt) + Kc2 cos(Kt). To satisfy the
boundary conditions, we need y(0) = c1 = 0 and y 0 (1) = Kc2 cos(K) = 0.
Since K 6= 0, in order to have c2 6= 0, it is necessary that cos(K) = 0. This is
true for an infinite set of Ks, namely

π 3π (2n + 1)π
K= , ,··· , ,··· .
2 2 2
(2n+1)π
Letting Kn = 2
, the eigenvalue λn is
µ ¶2
(2n + 1)π
λn = Kn2 = , n = 0, 1, 2, · · · .
2
³ ´
(2n+1)πt
The corresponding eigenfunctions are yn (t) = Cn sin 2
, n = 0, 1, · · · .

Note that this gives us a new orthogonal family of functions:

½ µ ¶ µ ¶ µ ¶ ¾
πt 3πt (2n + 1)πt
S̄ = sin , sin , · · · , sin ,··· ;
2 2 2
which is not the family that we used to generate the trigonometric Fourier
Series. The family S̄ is defined on the interval [0, 1], and since the weight
function w(t) in the Sturm-Liouville d.e. x00 + λx = 0 is identically equal to
one, the functions in S̄ satisfy the orthogonality condition
Z 1 µ ¶ µ ¶
(2m + 1)πt (2n + 1)πt
sin sin dt = 0
0 2 2

if m 6= n; and µ µ ¶¶2
Z 1
(2n + 1)πt 1
sin dt = .
0 2 2
Check it!
It can also be shown that any piecewise continuous function z(t), defined
on [0, 1] and satisfying the boundary conditions (2), can be expanded in a

3
convergent series of the form
∞
X ∞
X µ ¶
(2n + 1)πt
z(t) = an φn (t) = an sin ;
n=0 n=0
2

and, as shown in Lecture 9, the coefficients in this orthogonal series are given
by the formula
R1 ³ ´
R1 (2n+1)πt Z 1 µ ¶
z(t)φn (t)dt 0
z(t) sin 2
dt (2n + 1)πt
0
an = R 1 = R ³ ´2 ≡2 z(t) sin dt.
(φn (t))2 dt 1 (2n+1)πt 0 2
0
0
(sin 2
) dt

Example 2 For the piecewise continuous function

½ ¾
0.5t if 0 ≤ t ≤ 0.5
z(t) =
(t − 1)2 if 0.5 < t ≤ 1.0
find the first six non-zero terms in the series approximation
5
X µ ¶
(2n + 1)πt
z(t) ≈ an sin ,
n=0
2
and graph the function z(t) together with the finite series approximation on
the interval 0 ≤ t ≤ 1.
R1 ³ ´
Using the formula an = 2 0 (z(t)) sin (2n+1)πt
2
dt, the coefficients a0 , a1 , · · · , a5
can be found numerically. The graph below shows the function and its approx-
imation on the interval 0 ≤ t ≤ 1.

6-term approximation
0.25

0.2

0.15

0.1

0.05

0 0.2 0.4 0.6 0.8 1

t

It can be seen that the graph of the series approximation lies very close to the
graph of z(t) except at the point where the derivative z 0 (t) is discontinuous.
As you should expect, using more terms in the series will increase the accuracy
of the approximation.

4
Practice Problems:

1. Find all eigenvalues and corresponding eigenfunctions for the Sturm-

Liouville problems below:
a) x00 + λx = 0, x(0) = 0, x(π) = 0.
Ans: λn = n2 , φn (t) = Cn sin(nt), n = 1, 2, · · · .
b) x00 + λx = 0, x0 (0) = 0, x(L) = 0.
³ ´2 ³ ´
(2n+1)π (2n+1)πt
Ans: λn = 2L
, φn (t) = Cn cos 2L
, n = 1, 2, · · · .
c)(*) x00 + λx = 0, x0 (0) = 0, x0 (π) = 0.

2. (*) For the Sturm-Liouville problem x00 + λx = 0, x(0) = 0, x(1) +

x0 (1) = 0,
a) Show that there are no negative eigenvalues.
b) Show that λ = 0 is NOT an eigenvalue.
c) Show that the positive eigenvalues are of the form λn = Kn2 , where
Kn is the nth solution of the equation tan(Kn ) = Kn , and that the
corresponding eigenfunctions are φn (t) = Cn sin(Kn t).
d) Find numerical values for λ1 , λ2 , and λ3 and graph the corresponding
three eigenfunctions φ1 , φ2 and φ3 .
R1
e) Find the numerical value of the integral 0 φ2 (t)φ3 (t)dt. What should
it equal? Why?