April 11, 2013 BVP-1

Boundary Value and Eigenvalue Problems

Up to now, we have seen that solutions of second order ordinary differential
equations of the form

y 00 = f (t, y, y 0 ) (1)
exist under rather general conditions, and are unique if we specify initial
values y(t0 ), y 0 (t0 ). Let us use the notation IVP for the words initial value
problem.
In many applications, one wants solutions to (1) in which one specifies
the values of the solution y(t) at two separate points t0 < t1 rather than
specifying the value of y(t) and its derivative at a single point.
This leads to the subject of Boundary Value Problems, a very large and
important area of mathematics. The subject is studied for both ordinary and
partial differential equations. In the case of partial differential equations, one
deals with solutions which are defined on subsets of various Euclidean spaces,
and, hence there are many interesting regions for which to specify boundary
condtions.
In this course, we will only study two-point boundary value problems
for scalar linear second order ordinary differential equations. In most ap-
plications, the independent variable of the differential equation represents a
spatial condition along a real interval rather than time, so we use x for the
independent variable of our functions instead of t.
The general linear second order boundary value problem has the form

y 00 + p(x)y 0 + q(x)y = h(x), BC (2)

Here x is in some interval I = (a, b) ⊂ R, p(x), q(x), h(x) are continuous
real valued functions on I, α < β are two fixed real numbers in I, and BC
refers to specific boundary condtions.
Let us use the letters BVP to denote boundary value problem.
We wish to study all solutions of such a problem. In the cases considered
here, we can replace x by x + α if necessary and assume that α = 0. We will
denote the right boundary point by L.
We will consider four types of boundary conditions, which we denote by
the expressions 00, 01, 10, 11. These are defined by

type 00: y(0) = 0, y(L) = 0

April 11, 2013 BVP-2

type 01: y(0) = 0, y 0 (L) = 0

type 10: y 0 (0) = 0, y(L) = 0
type 11: y 0 (0) = 0, y 0 (L) = 0
where L > 0.
The BVP

y 00 + p(x)y 0 + q(x)y = 0, y(0) = 0, y(L) = 0 (3)

is called a homogeneous boundary value problem and will be denoted by
HBVP. Any BVP which is not homogeneous will be called a non-homogeneous
BVP.
Given a BVP of the form (2) of type 00, 10,01, or 10, there is an associ-
ated HBVP of type 00 obtained by replacing h(x) by the zero-function and
replacing the boundary conditions by y(0) = 0, y(L) = 0.
From our experience with IVP’s (initial value problems), we might expect
that the solutions to a general NBVP are related to those of its associated
HBVP. It turns out that BVP’s behave very differently than IVP’s. For
instance, a BVP may have no solution at all, infinitely many solutions, or
it may have a unique solution. In a certain sense, BVP’s behave more like
systems of linear algebraic equations than IVP’s.
For comparison, let us recall some general properties of linear algebraic
equations.
Consider the matrix equation

Ax = b

where A is an n × n matrix, and x are n × 1 matrices (which we think of

as column vectors). Here, A and b are known, and we wish to find x.
We have the following facts.

1. if det(A) 6= 0, then

(a) Ax = 0 has only the trivial solution, and

(b) Ax = b has a unique solution for every b.

2. if det(A) = 0,then

(a) Ax = 0 has infinitely many solutions, and

April 11, 2013 BVP-3

(b) Ax = b has either no solutions at all or infintely many solutions.

We will restrict our study of BVP’s to the case in which p and q are
constants and the boundary conditions are of the types 00, 01, 10, and 11.
Let us begin with the HBVP

y 00 + py 0 + qy = 0, y(0) = 0 y(L) = 0 (4)

Proposition. If (4) has a non-trivial solution, then p2 − 4q < 0. That
is, the characteristic polynomial z(r) = r2 + pr + q has no real roots.
Proof. Assume, by way of contradiction that z(r) has real roots.
Case 1: z(r) = (r − r1 )(r − r2 ) where r1 6= r2 .
Then the general solution to the differential equation has the form

y(x) = c1 er1 x + c2 er2 x

Assume that it is a non-trivial solution, so that either c1 6= 0 or c2 6= 0.

The first BC y(0) = 0 gives c1 + c2 = 0, so that c1 = −c2 .
The second BC y(L) = 0 gives

er1 L = er2 L

But, the function x → ex is strictly increasing, so this implies that r1 = r2 ,

a contradiction.
Case 2: z(r) = (r − r1 )2 .
Here the general solution is y(x) = c1 er1 x + c2 xer1 x .
The first BC, y(0) = 0 gives c1 = 0.
Then, the second BC, y(L) = 0, gives c2 Ler1 L = 0. But, since L > 0,
we have that Ler1 L > 0, so c2 = 0 also. This is a contradiction and the
Proposition is proved.
Remark. In a similar manner, one can prove that the BVP’s y 00 + py 0 +
qy = 0 of types 01, 10, and 11 only have non-trivial solutions if z(r) has no
real roots. We leave the proof as an exercise.
Thus, in considering constant coefficient BVP’s of type 00, 01, 10, 11, we
might as well assume that p2 − 4q < 0. The quadratic formula gives that the
roots have the form
q
r = −p/2 ± p2 − 4q/2
April 11, 2013 BVP-4

√ 2
Letting a = −p/2 and b = p − 4q/2, we get the general solution to the
differential equation as

y(x) = eax (c1 cos(bx) + c2 sin(bx))

It turns out that the methods and ideas in the study of this expression
with BC’s of types 00, 01, 10, 11 are not much different when a 6= 0 or a = 0.
So, for simplicity, we only consider the case when a = 0. That is, the case in
which z(r) has purely imaginary roots.
This means that our differential equation has the form

y 00 + q y = 0

where q > 0. It will simplify things if we write q = λ2 where λ > 0.

Thus, we consider the BVP’s of the form

y 00 + λ2 y = 0, BC (5)
where BC is one of the four types 00, 01, 10, 11.
Remark. We will see below that even for equations of the type (5),
non-trivial solutions only occur if there are special relations between λ and
L.
Example: An equation of the form (5) with no non-trivial solutions.
Consider

y 00 + 2y = 0, y(0) = 0, y(1) = 0 ( so L = 1).

The general solution to the ODE is

√ √
y(x) = c1 cos( 2x) + c2 sin( 2x)

The
√ condition y(0) = 0, gives c1 cos(0) = 0 or c1 = 0. So, y(x) =
c2 sin( 2x). √ √
The condition y(1) = 0 gives c2 sin( 2) = 0, or 2 = nπ for some integer
n. But there is no such n.
Conclusion: There are no non-trivial solutions.
April 11, 2013 BVP-5

Eigenvalue Problems
A real number λ2 such that the BVP (5) has a non-trivial solution yλ (x) is
called an eigenvalue of the BVP and the function yλ (x) is called an eigen-
function associated to (or corresponding to) λ2n . It turns out that if yλ (x)
is an eigenfunction, then so is any non-zero multiple Cyλ (x), so we usually
just take the constant C = 1.
Let us give some examples.
Example 1
Consider the BVP

y 00 + λ2 y = 0, y(0) = 0, y(L) = 0 (6)

We seek all solutions.
The general solution has the form

y(x) = c1 cos(λx) + c2 sin(λx)

The first BC gives c1 = 0 so we are dealing with the function c2 sin(λx).

The second BC (if c1 6= 0) gives that sin(λL) = 0.
Thus, λL = nπ for some positive integer n.
It follows that, for each positive integer n, if
nπ
λn =
L
then the function
nπx
yn (x) = sin( )
L
is a non-trivial solution of (6).
Thus, the eigenvalues of (6) are the numbers
nπ 2
λ2n = ( )
L
and the associated eigenfunctions are
nπx
yn (x) = sin( )
L
Example 2 Find the eigenvalues and associated eigenfunctions for the
BVP
April 11, 2013 BVP-6

y 00 + λ2 y = 0, y(0) = 0, y 0 (L) = 0

Again, we begin with the general solution.

y(x) = c1 cos(λx) + c2 sin(λx)

BC-1: y(0) = 0 =⇒ c1 = 0.
BC-2: y 0 (L) = 0 =⇒ cos(λL) = 0
The cosine function assumes the value 0 at the odd multiples of π/2.
We are interested in the positive ones, so we can write them as

(2n − 1)π
, n = 1, 2, 3, 4, . . .
2
Thus, BC-2 requires that

(2n − 1)π
λ=
2L
for some positive integer n.
Our eigenvalues are

(2n − 1)π 2
λ2n = ( )
2L
and our associated eigenfunctions are

(2n − 1)πx
sin( )
2L
for n = 1, 2, 3, . . ..
We have found the eigenvalues and associated eigenfunctions for BVP’s
of types 00 and 01 and the equation

y 00 + λ2 y = 0

Similar techniques work for the types 10 and 01.

We list a type, eigenvalue, eigenfunction table for the equation y 00 +λ2 y =
0 on the next page.
April 11, 2013 BVP-7

The results are in the following table. These are all for the differential
equation y 00 + λ2 y = 0.

type 00 01 10 11

eval ( nπ
L
)2 ( (2n−1)π
2L
)2 ( (2n−1)π
2L
)2 ( nπ
L
)2

efun sin( nπx

L
) sin( (2n−1)πx
2L
) cos( (2n−1)πx
2L
) cos( nπx
L
)