Ordinary Differential Equations

for Engineers and Scientists

Gregg Waterman
Oregon Institute of Technology
 c 2017 Gregg Waterman

This work is licensed under the Creative Commons Attribution 4.0 International license. The essence of the license
is that

You are free to:

• Share copy and redistribute the material in any medium or format

• Adapt remix, transform, and build upon the material for any purpose, even commercially.

The licensor cannot revoke these freedoms as long as you follow the license terms.

Under the following terms:

• Attribution You must give appropriate credit, provide a link to the license, and indicate if changes
were made. You may do so in any reasonable manner, but not in any way that suggests the licensor
endorses you or your use.

No additional restrictions You may not apply legal terms or technological measures that legally restrict
others from doing anything the license permits.

Notices:

You do not have to comply with the license for elements of the material in the public domain or where your
use is permitted by an applicable exception or limitation.

No warranties are given. The license may not give you all of the permissions necessary for your intended
use. For example, other rights such as publicity, privacy, or moral rights may limit how you use the material.

For any reuse or distribution, you must make clear to others the license terms of this work. The best way
to do this is with a link to the web page below.

To view a full copy of this license, visit https://creativecommons.org/licenses/by/4.0/legalcode.

Contents
5 Boundary Value Problems 137
5.1 Deflection of Horizontal Beams . . . . . . . . . . . . . . . . . . . . . . . . . . 139
5.2 Second-Order Boundary Value Problems, Eigenfunctions and Eigenvalues . . . . 144
5.3 Eigenvalue Problems, Deflection of Vertical Columns . . . . . . . . . . . . . . . 151
5.4 The Heat Equation in One Dimension . . . . . . . . . . . . . . . . . . . . . . . 159
5.5 Chapter 5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163

D Solutions to Exercises 208

D.5 Chapter 5 Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208

i
5 Boundary Value Problems

Learning Outcome:

5. Set up and solve boundary value problems.

Performance Criteria:

(a) Solve a boundary value problem for the deflection of a horizontal beam.
(b) Give the boundary conditions for a horizontal beam.
(c) Predict the shape of the deflection curve for a horizontal beam that is
supported in a given manner.
(d) Determine whether a function is an eigenfunction of a differential oper-
ator. If it is, give the corresponding eigenvalue.
(e) Give eigenfunctions of the first or second derivative, for a given eigen-
value.
(f) Solve a boundary value problem for eigenvalues and the corresponding
eigenfunctions.
(g) Give the boundary conditions for a vertical column.
(h) Find the buckling modes (non-trivial solutions) for a vertical column.
(i) Find the critical loads for a vertical column.
(j) Give the pinning conditions resulting in each of the buckling modes of a
vertical column.

All of the applications that we have studied so far have involved some quantity that is a function of
time; that is, time has been the independent variable. Arbitrary constants have arisen in the process of
solving the associated ODEs, and we have used given initial conditions to determine the values of the
constants. In this chapter we look at deflection (bending) of horizontal beams and vertical columns.
For horizontal beams the deflection is a function of the distance along the beam or column. The
independent variable is then a one dimensional position variable, as discussed in Section 1.4. As seen in
Section 1.7, we use boundary conditions, rather than initial conditions, to determine the values of the
arbitrary constants.
We will designate the variable x to denote the distance along the beam (or column) from one end
or the other. Due to the weight of the beam there will be some deflection y off of the horizontal line
the beam would follow if it had no weight. The deflection will be different at different points along the
beam, so y = y(x). That is, the amount of deflection depends on where one is looking along the
length of the beam. The solution function is obtained from a fourth order ODE having four boundary
conditions. Solving such problems is relatively straightforward.
The situation will be significantly different for vertical columns, in a way that might be somewhat
surprising. We’ll see that such a column will remain straight as more and more weight is added to
it until, at some weight (called the first critical load), it suddenly deflects (“buckles”). It will then
either deform or break as a the load is increased. However, if we prevent the middle of the column from
deflecting, each half will deflect at a load (called the second critical load) that is four times the first
critical load.
Solving the boundary value problems associated with vertical columns requires solving what we call
an eigenvalue problem, which is more nuanced that the boundary value problems associated with

137
horizontal beams. We will devote two sectoins of this chapter to eigenvalue problems and vertical
columns. We then conclude the chapter with a look at perhaps the simplest applicatoin of partial
differential equation, heat distribution in a rod. The method of solution leads us two two types of
ODEs, one of which is an eigenvlalue problem.

138
5.1 Deflection of Horizontal Beams

Performance Criteria:

5. (a) Solve a boundary value problem for the deflection of a horizontal beam.
(b) Give the boundary conditions for a horizontal beam.
(c) Predict the shape of the deflection curve for a horizontal beam that is
supported in a given manner.

In this section we will take a look at the differential equations associated with beams that are
suspended horizontally in some way. The beams themselves will not be horizontal over their entire
lengths, because the force of gravity will cause some bending. The first thing to understand is the
mathematical setup. Suppose that we have a beam of length 10 feet. We put the cross-sectional center
of its left end at the origin of an x-y coordinate plane, and the cross-sectional center of its right end
at the point (10, 0). The longitudinal axis of symmetry of the beam then runs along the x-axis from
x = 0 to x = 10; see the figure below and to the left.

y y

x x

10 10

Now the beam will deflect (a fancy term for “sag”) in some way, due to any weight it is supporting,
including its own weight. The shape it takes will depend on the manner in which it is supported (we
will get into that soon), but one possibility is shown in the figure above and to the right. The points
along what was the original axis of symmetry of the beam now follow the graph of a function, which
we will call y(x). Note that the domain of the function is just the interval [0, 10]. Our goal will be
to find the mathematical equation of the function.
Let us still consider a 10 foot beam, but we will represent it
with just the curve described by the deflection of the longitudi- y
nal axis of symmetry. (From now on, when we talk about the
beam, we really mean the deflected original longitudinal axis of
x
symmetry of the beam.) Suppose also that the ends of the beam
are what we call embedded. This means that they are not only 10
supported at both ends, but the ends are also held horizontal by
being “clamped” somehow. A good image to keep in mind is a
beam that is stuck into two opposing walls of a structure. See Figure 5.1(a)
the diagram to the right.
The theory behind obtaining a differential equation to model a horizontal beam is beyond the scope
of this class. Suffice it to say that it involves ideas from the area of statics, like the “bending moment”
of the beam, and the properties of the material from which the beam is built. The differential equation
itself is fourth order:
d4 y
EI 4 = w(x) (1)
dx

139
Here E is Young’s modulus of elasticity for the material from which the beam is made, I is the
moment of inertia of a cross-section of the beam and w(x) is the load per unit length of the beam.
If the beam has uniform cross-section and the only weight that it is supporting is its own weight, then
w(x) is a constant. We will consider only that situation.
Let’s consider the situation shown in Figure 5.1(a), where both ends are embedded. Because the
ODE (1) is fourth order, we will need four boundary conditions to determine all of the constants that
will arise in solving it. We first recognize that because the two ends are supported, there will be no
deflection at either end. Therefore y(0) = y(10) = 0. This will be the case for any horizontal beam
that is supported at both ends. Next we consider the fact that the ends of the beam are embedded
horizontally into a wall. The embedding causes both ends to be horizontal right at the points where they
leave the walls they are embedded in, so the slope of the beam is zero at those points. Mathematically
we express this by y ′ (0) = y ′ (10) = 0. When we put the ODE together with these boundary conditions
we get a boundary value problem. Suppose that for our ten foot beam beam E = 10, I = 5 and
w(x) = 100. The boundary value problem that we have is then

d4 y
50 = 100, y(0) = 0, y ′ (0) = 0, y(10) = 0, y ′ (10) = 0 (2)
dx4
We solve this by simply taking a succession of antiderivatives and finding constants along the way, when
we are able to.

⋄ Example 5.1(a): Solve the boundary value problem (2) above.

d4 y
Solution: We begin by dividing both sides by 50 to get = 2. Our task now is to keep
dx4
d3 y
integrating both sides until we find y = y(x). Integrating once gives = 2x + C1 , and
dx3
2
d y dy
integrating again gives 2
= x2 +C1 x+C2 . Next we find that = 13 x3 + 21 C1 x2 +C2 x+C3 ,
dx dx
and applying the initial condition y ′ (0) = 0 gives C3 = 0. Substituting this value and integrating
1 4
one more time we get y = 12 x + 61 C1 x3 + 21 C2 x2 + C4 , and applying the boundary condition
y(0) = 0 results in
1 4
y = 12 x + 61 C1 x3 + 21 C2 x2 . (3)
We now apply the initial condition y(10) = 0 to get 0 = 10,000 1000 100
12 + 6 C1 + 2 C2 , and the
′ 1000 100
initial condition y (10) = 0 to get 0 = 3 + 2 C1 + 10C2 . To solve this system we multiply
the first equation by 12 and the second by 6 to get the system to the left below, which can be
solved in the manner shown in the other steps:

2000C1 + 600C2 = −10, 000 20C1 + 6C2 = −100

=⇒
300C1 + 60C2 = −2000 −30C1 − 6C2 = 200

− 10C1 = 100
C1 = −10
Substituting this value for C1 and solving for C2 gives us C2 = − 40
3 . Putting these values
1 4
into (3), the solution to the IVP is y = 12 x − 53 x3 − 20
3 x 2.

Use your calculator or an online grapher like Desmos to graph the solution from x = 0 to x = 10,
using a y scale that allows you to actually see the deflection of the beam. Does the result surprise
you? (It should!) One annoying feature of the differential equation is that it is based on taking down

140
to be the positive direction. To see what the actual shape of the beam will be, multiply your solution
by −1, then graph it. Now the result should look something like Figure 5.1(a).
Suppose again that we have a 10 foot beam, but now it
is supported by a fulcrum at each end, and each end is free
to pivot around the fulcrum. We will call the beam simply
supported in this case. Civil engineers might call this “pinned-
pinned.” See the diagram to the right. In this case, we will have x=0 x = 10
y(0) = y(10) = 0, just like the embedded case. However, we
can see that we will not have y ′ (0) = 0 or y ′ (10) = 0.
So how do we get two more boundary conditions? Note that the downward force of gravity in the
interior and the upward force of the supports at the ends bend the beam into a concave up shape.
(Remember concavity from differential calculus?) The upward concavity here means that it must be
the case that y ′′ (x) > 0 for values of x between, but not equal to 0 and 10. However, there are no
opposing forces to bend the beam right at its ends. Thus there is no concavity right at the ends of the
beam, resulting in the conditions y ′′ (0) = y ′′ (10) = 0.
The final situation we will consider for now is a beam that
is embedded at the left end and free at the right end, as shown.
The left end of the beam is embedded, so we know the values
y(0) = y ′ (0) = 0. We know neither the displacement nor
the slope of the right end, but what we do know there is that
there is no concavity there, so y ′′ (10) = 0. This gives us
three boundary conditions, but of course we need four. The last
condition comes from some theory we won’t go into here, but it x=0 x = 10
is y ′′′ (10) = 0.
Let us now summarize the the possible boundary conditions for a horizontal beam:

• At an embedded end both y and y ′ are zero.

• At a simply supported end y and y ′′ are zero.

• At a free end y ′′ and y ′′′ are zero.

I expect you be able to give any of those conditions - you should be able to figure all of them out
each time you need them, without memorization, with the possible exception of the third derivative just
discussed.

Section 5.1 Exercises To Solutions

1. For each beam pictured below, list the boundary conditions. Assume that the height of the left
end of each is zero.

(a) (b)

12 ft

8 ft

141
 (c) (d)

20 ft x=0 x = 15

2. Suppose that you have a beam that is 8 feet long. For each of the scenarios given, determine first
whether it would make any sense physically to have the beam supported in the manner given. If
not, explain why. If it does make sense, give the four boundary conditions.

(a) Left end embedded horizontally, right end simply supported.

(b) Left end simply supported, right end free.

(c) Both ends free.

3. Find the deflection function y = y(x) for an eight foot beam that is embedded at both ends,
carrying a constant load of w(x) = 150 pounds per foot. Suppose also that E = 30 and
I = 80, in the appropriate units.

(a) Give the appropriate boundary value problem (differential equation plus boundary conditions).

(b) Solve the differential equation and apply the boundary conditions in order to determine the
constants. What is the final solution?

(c) Graph your solution on the appropriate x interval, using a y scale that allows you to
actually see the deflection of the beam. Remember to multiply the right side of your solution
from (b) by negative one so that it appears the same way that the beam will.

(d) Where do you believe the maximum deflection should occur? Find the deflection there - you
need not give units with your answer, since I have been somewhat vague about the units of
the constants E and I.

4. (a) Sketch a graph of the deflection of a beam that is embedded at its left end and free at its
right end.

(b) Suppose that the beam is 10 feet long, with values of w0 , E and I of 100, 10 and 5,
respectively. Solve the boundary value problem.

(c) Graph your solution and compare with your sketch in part (a). Of course they should be the
same.

(d) What is the maximum deflection of the beam, and where does it occur?

5. Repeat parts (a)-(d) of Exercise 4 for a 10 foot beam that is simply supported on both ends.

142
6. Repeat steps (a)-(c) of Exercise 4 for an eight foot beam, with the same parameters as in Exercise
3, that is embedded on the left end and simply supported on the right end. Then do the following:

(d) It was intuitively clear where the maximum deflection occurred for the two previous situations,
but it is not so clear in this case. Take a guess as to about where you think it should occur for
this case. Then use the graph on your calculator, along with the trace function, to determine
where the maximum deflection occurs, and how much it is.

7. The graphs below are those of some fourth degree polynomials. The points labeled A, D, E, H, I
and L are maxima for their respective functions, and the points labeled B, C, F, G, J and K are
inflection points. For each of the following boundary situations, give the endpoints of a section
of graph that has the shape the deflection curve would take. Assume that both ends of the beam
are supported at the same level, and that the dashed lines are horizontal.

(a) Simply supported at the left end, embedded at the right end.

(b) Embedded at both ends.

(c) Simply supported at both ends.

E H I
A D L
B C J K
F G

143
5.2 Second-Order Boundary Value Problems, Eigenfunctions and Eigenvalues

Performance Criteria:

5. (d) Determine whether a function is an eigenfunction of a differential oper-

ator. If it is, give the corresponding eigenvalue.
(e) Give eigenfunctions of the first or second derivative, for a given eigen-
value.

We begin by returning to a boundary value problem that we saw in Section 1.7. It is similar to a sort
of problem that comes up often in applications. The main thing that distinguishes this from an initial
value problem is that the idependent variable is postion, x, rather than time. Another difference we
will usually see in boundary value problems is that we are given values of the function at two different
values of the independent variable, in this case at zero and π.

⋄ Example 5.2(a): Solve the boundary value problem

y ′′ + 41 y = 0, y(0) = 3, y(π) = −4.

1
Solution: The auxiliary equation for the differential equation is r 2 + 4 = 0, which has the
solution r = ± 12 i. This gives us the solution

y = C1 sin 12 x + C2 cos 12 x (1)

to the differential equation. To find the values of the constants we apply the boundary conditions
y(0) = 3, y(π) = −4. For the boundary condition y(0) = 3 we substitute x = 0 and
y = 3 into (3) to get
3 = C1 sin 21 (0) + C2 cos 12 (0).
This gives us C2 = 3. Substituting x = π and y = −4 into (1) gives us C1 = −4. Therefore
the solution to the boundary value problem is y = −4 sin 21 x + 3 cos 12 x.

Differential Operators, Again

Recall from Section 3.5 that a mathematical object that “works on” a function to produce another
function is called an operator, and the derivative is probably the simplest example of an operator. Of
course the second derivative is an operator as well. In that section we also showed that we can combine
derivatives to get other operators.

d2
⋄ Example 5.2(b): The second derivative is an operator. You should be quite familiar with
dx2
its action:
d2
5x3 + 7x2 − 2x + 4 = 30x + 14


dx 2

144
 ⋄ Example 5.2(c): We can create new operators by forming something called a linear combination
of derivatives. As an example, we can define an operator

d2 d
D=3 +5 −4
dt2 dt
by its action on a function y = y(t):

d2 y dy
D(y) = 3 2
+ 5 − 4y.
dt dt
So, for example, if y = e−2t ,

d2
d −2t

D e−2t = 3 2 e−2t + 5 e − 4e−2t = 12e−2t − 10e−2t − 4e−2t = −2e−2t
dt dt
You saw such operators when we studied second order linear ODEs.

 
−4 −6
⋄ Example 5.2(d): You may know that when we multiply the matrix A = times
3 5
 
⇀ 1
the vector u = we get
3
      
⇀ −4 −6 1 (−4)(1) + (−6)(3) −22
Au = = =
3 5 3 (3)(1) + (5)(3) 18
 
⇀ 2
Similarly, for the vector v = ,
−1
    
⇀ −4 −6 2 −2
Av = =
3 5 −1 1

We can think of A as an operator that acts on vectors with two components to create other
vectors with two components.

Recall that the derivative operator is what we call a linear operator. What this means is that if
f and g are functions, and c is a constant, then
d df dg d df
[f (x) + g(x)] = (x) + (x) and [cf (x)] = c (x)
dx dx dx dx dx
⇀ ⇀
This behavior is not unique. If A is a matrix, u and v are vectors, and c is a scalar (constant),
⇀ ⇀ ⇀ ⇀ ⇀ ⇀
A(u + v) = A u +A v and A(c u) = cA u

Linear operators have these two properties, of “distributing over addition” and “passing through con-
stants.” (This is where the the language “linear” in linear algebra comes from.) Many operators used
in applications are linear operators.

145
Eigenfunctions and Eigenvalues
Let’s go back to the differential equation y ′′ + 14 y = 0 from Example 5.2(a). Note that we can
arrange the differential equation as
d2 y
= − 14 y. (2)
dx2
When we seek a solution to this differential equation, the equation tells us that we are looking for
a function y = y(x) whose second derivative is one-fourth the function itself. We looked at such
equations in Section 1.2, and established by guessing and checking that a function of the form

y = C1 sin 12 x + C2 cos 21 x (3)

is a solution for any values of C1 and C2 . Of course we now know how to solve (2) using its
auxiliary equation, and we know also that every solution to (2) must have the form (3). The fact that
the action of the second derivative operator the function (3) is to simply multiply the the function by
− 41 is something fairly special. That’s not the case for most other functions when the second derivative
operator “works on” them. Here’s an example of a more complicated operator and two functions, one
of which has this property that the operator acting on it is the same as multiplying by a number, and
the another function for which this is not the case.

⋄ Example 5.2(e): Let L be the differential operator defined on a function y = y(x) by

d2 y dy
L(y) = (x2 − 1) 2
+ 2x .
dx dx
Apply this operator to the functions p(x) = x2 − 5x + 2 and q(x) = 5x3 − 3x.

Solution: We see that

d2 d 2
L x2 − 5x + 2 = (x2 − 1) 2 x2 − 5x + 2 + 2x




x − 5x + 2
dx dx
= (x2 − 1)(2) + 2x(2x − 5)
= 6x2 − 10x − 2

and
d2 d
L 5x3 − 3x = (x2 − 1) 5x3 − 3x + 2x 5x3 − 3x




dx 2 dx
= (x2 − 1)(30x) + 2x(15x2 − 3)
= 60x3 − 36x

There is nothing special about the result when the operator L of the previous example is applied
to p(x) = x2 − 5x + 2, but we see that

L q(x) = L 5x3 − 3x = 60x3 − 36x = 12(5x3 − 3x) = 12q(x).

 

Note that the ultimate effect of L on q is to multiply it by twelve. When an operator operates
on a function and the result is to simply multiply the function by a constant, we call the function an
eigenfunction of the operator:

146
 Eigenfunctions and Eigenvalues
Let A be an operator that operates on functions and let y be a nonzero function
for which there is a constant λ such that

Ay = λy.

Then y is an eigenfunction of the operator A, with corresponding eigenvalue

λ. Note that λ = 0 is allowable, but y = 0 is not.

⋄ Example 5.2(f): For the operator L of Example 5.2(e), give an eigenfunction and the corre-
sponding eigenvalue.

Solution: Because L q(x) = 12q(x), q(x) = 5x3 − 3x is an eigenfunction of L with

 

eigenvalue 12.

d2
⋄ Example 5.2(g): Consider again the second derivative , and note that
dx2
d2
(sin 12 x) = − 14 sin 21 x.
dx2
The effect of the derivative on sin 12 x is to simply multiply the function by − 41 , so sin 12 x is
d2
an eigenfunction for the operator , with corresponding eigenvalue − 41 .
dx2

⇀
⋄ Example 5.2(h): Note that in Example 5.2(d), the result of A times v was simply −1 times
 
⇀ ⇀ 2
v. We say that v = is an eigenvector (instead of eigenfunction) for the matrix
−1
 
−4 −6
A= , with eigenvalue −1.
3 5

⋄ Example 5.2(i): Because the derivative of a constant is zero, which is also zero times the
d
function, every nonzero constant function is an eigenfunction of the first derivative operator ,
dx
with corresponding eigenvalue zero. This emphasizes that even though the zero function isn’t
allowed as an eigenfunction, eigenfunctions are allowed to have eigenvalues of zero.

Eigenfunction Problems
Here we see how eigenfunctions are important to us in our study of differential equations. The
differential equation  
d dy
(1 − x2 ) + n(n + 1)y = 0 (4)
dx dx

147
is called Legendre’s Differential Equation and arises when modeling steady-state heat distribution in
a solid medium using polar coordinates. If we let n(n + 1) = λ, move the term n(n + 1)y = λy to
the right side, apply the derivative outside the brackets on the left (product rule!) and negate both
sides, (4) becomes
d2 y dy
(x2 − 1) 2 + 2x = λy. (5)
dx dx
If we then let L be the operator of Example 5.2(e) defined by

d2 y dy
L(y) = (x2 − 1) 2
+ 2x .
dx dx
then (5) becomes the eigenfunction/eigenvalue equation

L(y) = λy. (6)

Solving (4) then is equivalent to finding eigenfunctions and eigenvalues of the operator L. This is
what we mean by solving the eigenvalue problem (5).
We now make some observations related to Example 5.2(g), where we saw that y = sin 12 x is
an eigenfunction for the second derivative with eigenvalue − 41 , and Example 5.2(a). First, for any
constant C,
d2
(C sin 12 x) = − 41 C sin 12 x = − 41 (C sin 12 x). (7)
dx2
This indicates that any constant multiple of an eigenfunction is also an eigenfunction, with the same
eigenvalue. This holds for eigenvectors as well; we don’t really think of such multiples as new eigen-
functions or eigenvectors. We also see that

d2
(cos 21 x) = − 14 cos 12 x, (8)
dx2
showing that an operator can have more than one eigenfunction (beyond just constant multiples) with
the same eigenvalue. This also holds for matrices and eigenvectors.
Finally, combining (7) and (8) gives us that any function of the form

y = C1 sin 12 x + C2 cos 21 x

is an eigenfunction of the second derivative with eigenvalue − 14 . This indicates that solving the
differential equation y ′′ + 14 y = 0 amounts to finding eigenfunctions of the second derivative with
eigenvalue − 14 .

Section 5.2 Exercises To Solutions

d
1. In this exercise you will be considering the first derivative operator .
dx

(a) The function y = e3x is an eigenfunction for the operator. What is the corresponding
eigenvalue?

(b) Give the eigenfunction of the operator that has eigenvalue −5.

(c) Based on your answers to parts (a) and (b), what is the general form of any eigenfunction
of the first derivative operator and what is the corresponding general eigenvalue?

148
 d2
2. Now consider the second derivative operator . There are three general forms of the eigen-
dx2
functions for this operator, depending on whether the eigenvalues are positive, negative or zero.

(a) Give a specific function that has eigenvalue zero; that is, the second derivative of the function
is zero.

(b) Give the most general form of function whose eigenvalue is zero.

(c) Give two different functions (neither of them being a multiple of the other) that are eigen-
functions of the second derivative with eigenvalue −4.

(d) Give two different functions, neither of them being a multiple of the other, that are eigen-
functions of the second derivative with eigenvalue −3.

(e) Give two different functions, neither of them being a multiple of the other, that are eigen-
functions of the second derivative with eigenvalue −λ2 , where λ is a positive real number.

(f) Give the eigenfunctions that will have eigenvalue nine; there are two of them!

(g) Give the general form of the eigenfunctions of the operator that have positive eigenvalues,
and give the general eigenvalue.

d2 d
3. Let D be the operator D = + 2 − 3, whose action on a function y = y(t) is defined
dt2 dt
d2 y dy
by Dy = 2
+ 2 − 3y.
dt dt

(a) Show that y = e−2t is an eigenfunction for this operator, and determine the corresponding
eigenvalue.

(b) In general, any function of the form ekt is an eigenfunction for D. Determine the general
eigenvalue.

(c) Give two values of k for which ekt is an eigenfunction of D with eigenvalue zero.

(d) Give two values of k for which ekt is an eigenfunction of D with eigenvalue five.

4. A very important ODE in many applications is y ′′ + λ2 y = 0. Note that this can be rearranged
to get y ′′ = −λ2 y, which says that any y that is a solution to the differential equation is an
eigenfunction with eigenvalue −λ2 .

(a) Give the eigenfunctions of the second derivative with eigenvalue −λ2 .

(b) (Challenge) Let S be the set of functions y = f (x) that have continuous second derivatives
on the interval [0, 2π] and for which f (0) = f (2π) = 0. Determine ALL eigenfunctions
of the second derivative with eigenvalue −λ2 that are in S.

149
5. Let L be the differential operator defined on a function y = y(x) by

d2 y dy
L(y) = x 2
+ (1 − x) .
dx dx
Determine which of the following are eigenfunctions of L. For those that are, give the corre-
sponding eigenvalue.

(a) y = x2 + 3x (b) y = 1 − x

(c) y = x3 − 9x2 + 18x − 6 (d) y = 2x + 3

(e) y = x2 − 4x + 2

6. Example 5.2(d) showed that the function P3 (x) = 5x3 − 3x is an eigenfunction of the operator
L defined by
d2 y dy
L(y) = (x2 − 1) 2 + 2x ,
dx dx
with eigenvalue 12. The function P3 (x) is called a Legendre polynomial. There are more
Legendre polynomials, each of which is an eigenfunciton for L - here are a few of them:

P1 (x) = x, P2 (x) = 3x2 − 1, P4 (x) = 35x4 − 30x2 + 3, P5 (x) = 63x5 − 70x3 + 15x.

(a) Determine the corresponding eigenvalue for each of the eigenfunctions given above.

(b) Make a table of n values for n = 1, 2, 3, 4, 5 and the corresponding eigenvalues.

(c) There should be a pattern to the eigenvalues, but you may find it difficult to figure out. Give
it a try, and take a guess as to what the eigenvalue is for n = 7. The eigenfunction is
P7 = 429x7 − 693x5 + 315x3 − 35x - apply L to it to check your guess for the eigenvalue.

(d) What is the eigenvalue for the nth Legendre polynomial Pn (x)?

150
5.3 Eigenvalue Problems, Deflection of Vertical Columns

Performance Criteria:

5. (f) Solve a boundary value problem for eigenvalues and the corresponding
eigenfunctions.
(g) Give the boundary conditions for a vertical column.
(h) Find the buckling modes (non-trivial solutions) for a vertical column.
(i) Find the critical loads for a vertical column.
(j) Give the pinning conditions resulting in each of the buckling modes of a
vertical column.

When solving the equation

d2 y
= − 14 y
dx2
we see that we are looking for eigenfunctions of the second derivative with eigenvalue − 14 . In many
applications we are looking for eigenfunctions without knowing what the eigenvalue is. This seems like
an impossible problem to solve, but it isn’t, as we see in the following example.

⋄ Example 5.3(a): Solve the boundary value problem

y ′′ + λ2 y = 0, y(0) = 0, y ′ (2π) = 0, (1)

where λ is a positive value to be determined.

Solution: The auxiliary equation for the differential equation is r 2 + λ2 = 0, which leads to
r 2 = −λ2 , so r = ±λi. The solution to the ODE is then
y = C1 sin λx + C2 cos λx.
Applying the first boundary condition y(0) = 0 gives us C2 = 0, so the solution is
y = C1 sin λx.
From this we can compute y ′ = C1 λ cos λx, and applying the second boundary condition gives
us
C1 λ cos 2πλ = 0.
There are three possibilities here: C1 = 0, λ = 0, or cos 2πλ = 0. The first two result in a
solution of y = 0 for the boundary value problem. This is a valid solution, but quite uninteresting!
For this reason we will refer to it as the trivial solution. To get a non-trivial solution it must be
the case that cos 2πλ = 0. Now cos θ = 0 when θ = π2 , 3π 5π
2 , 2 , .... Thus we have
π 3π 5π
2πλ = 2 , 2 , 2 , ...
1 3 5
(2)
λ = 4 , 4 , 4 , ...

Changing C1 to just C, the non-trivial solutions to the boundary value problem (1) are then
y = C sin 41 x, y = C sin 34 x, y = C sin 54 x, ..., (3)
each corresponding to one of the values of λ determined in (2).

151
 As stated, each of the solutions (3) to the BVP (1) corresponds to a particular value of λ. Let’s
verify one of those solutions.

⋄ Example 5.3(b): Verify that y = C sin 54 x is a solution to the boundary value problem

y ′′ + λ2 y = 0, y(0) = 0, y ′ (2π) = 0 (1)

when λ = 45 .
25
Solution: The differential equation in this case is y ′′ + 16 y = 0. We see that

y = C sin 45 x =⇒ y ′ = 54 C cos 45 x =⇒ y ′′ = − 16
25
C sin 54 x, (4)

so
25
y ′′ + 16 y = − 25 5
16 C sin 4 x +
25
16 (C sin 54 x) = 0,
showing that y = C sin 54 x is a solution to the differential equation.
We must now show that y = C sin 54 x satisfies the boundary conditions. Note that we found
y′ in (4).

y(0) = C sin 45 (0) = 0 and y ′ (2π) = 54 C cos 45 (2π) = 45 C cos 5π

2 = 0.

Because y = C sin 54 x satisifies both the differential equation and boundary conditions, it is a
solution to the BVP (1) when λ = 45 .

We note two ways our solution to the boundary value problem from Example 5.3(a) differs from the
initial value problems we’ve solved, and the particular boundary value problems that we have solved up
to now:

• There are infinitely many solutions to this boundary value problem, each corresponding to a specific
choice of λ.

• There is an arbitray constant whose value we cannot determine from the information given.

For the particular application that we will look at in this section, we consider only one of the solutions

y = C sin 41 x, y = C sin 34 x, y = C sin 54 x, ..., (3)

at a time. For other applications we need at some point to consider instead the arbitray linear combi-
nation of solutions
y = C1 sin 41 x + C2 sin 43 x + C3 sin 54 x + · · · (5)
When these applications occur (in the solving of partial differential equations), there is additional
information that can be used to determine the values of all these constants. Some of you may recognize
(5) as a Fourier series.
We now examine Example 5.3(a) in the context of eigenvalues and eigenfunctions. Note that the
differential equation from the BVP can be written

d2
(y) = −λ2 x,
dx2
which is saying we are looking for functions that are eigenfunctions of the second derivative with
eigenvalues −λ2 . This is called an eigenvalue problem, and it is typical that we need to find both

152
the eigenvalues and eigenfuctions. (Any of you who have had linear algebra may recall that in that
course you had to find both eigenvalues and eigenvectors.) The result of Example 5.3(a) gives us the
eigenvalues (using the results of (2))

−λ2 = − 16
1
, − 9
16 , − 25
16 , ...,

with corresponding eigenfunctions

y = C sin 41 x, y = C sin 34 x, y = C sin 54 x, ...

We now look at one application of these ideas that some of you may have encountered in a strengths
of materials course.

Deflection of Vertical Columns

Now we will examine the behavior of a vertical column Load
Force
when a load is applied directly downward on its top, as shown ⇓ ⇓
in the first picture to the right. We will begin by considering
columns that are pinned; this means we will allow the top
and bottom of the column to be at angles other than ver-
tical. We will require the top and bottom of the column to
be vertically aligned with each other - later we will consider
a situation where we will relax this condition. So, for exam-
ple, when enough force is applied downward the column will
deflect horizontally as shown in the second picture to the
right.
We will set up a coordinate system as shown below and to the right, with x indicating the distance
upward from the bottom of the column and y = y(x) representing the horizontal deflection of the
column at any point x. The differential equation governing the deflection of the column is

d2 y
EI = −P y, (1)
dx2
where the parameters E and I are again the modulus of elasticity and cross-sectional moment of
inertia of the column. They are properties that could vary along the column (with the variable x), but
this would be unusual. We’ll only consider columns where they do not
x
change. P is the (positive) force exerted downward on the top of the
x=L
column, and we will look at the effects of different values of P , but y(L) = 0
for purposes of solving the ODE it is a constant. (It is a parameter
rather than a variable, but its value is to be determined when solving
the ODE.) Because the ODE is second order we will need two conditions
to determine the solution. If the length of the column is denoted by L,
we have the boundary conditions y(0) = 0 and y(L) = 0. If we y(0) = 0
rearrange the equation and combine it with the boundary conditions we y (pos)
get the boundary value problem (BVP)

d2 y P
2
+ y = 0, y(0) = 0, y(L) = 0. (2)
dx EI
We will see that solving this boundary value problem is somewhat different than solving the kind of
BVPs we saw for horizontal beams, in the previous section. This is because we can rearrange the ODE
(1) to get the eigenvalue problem
d2 y P
2
=− y. (3)
dx EI

153
 P
Note that y is an eigenfunction of the second derivative, with eigenvalue − EI , where P , E and
I are all positive. As we now know very well, y must be of the form

y = C1 sin ax + C2 cos ax

for some constant a yet to be determined. Let’s now go through the details for a specific case:

⋄ Example 5.3(c): Solve the boundary value problem

d2 y P
+ y = 0, y(0) = 0, y(20) = 0, (4)
dx2 EI
with E = 800, I = 150 and P > 0.

Solution: Substituting the values for E and I into the ODE gives us

d2 y P
2
+ y=0 (5)
dx 120, 000
P
Using our methods from Chapter 3, this equation has auxiliary equation r 2 + 120,000 = 0 and,
q
P
because P > 0, its roots are r = ±i 120,000 . The solution to (4) is then
q q
P P
y = C1 sin 120,000 x + C2 cos 120,000 x

Applying the boundary condition y(0) = 0 gives us C2 = 0, so the solution is then

q
P
y = C sin 120,000 x.

(I’ve omitted the subscript for simplicity.) Now here is where things start to get interesting! The
other boundary condition tells us that
q
P
0 = C sin 120,000 (20) (6)
q
P
which, in turn, tells us that either C = 0 or sin 120,000 (20) = 0. The first possibility gives
us the “trivial” solution y = 0 - it satisfies the differential equation and boundary conditions,
but it isn’t particularly interesting! Considering the second possibility, sin θ = 0 for θ =
0, π, 2π, 3π, ..., nπ, ... so, for C 6= 0, (6) will be true for
q
P
120,000 (20) = 0, π, 2π, 3π, . . . , nπ, . . . , (7)

the first of which also gives us the trivial solution y = 0. Therefore, in theory at least (we’ll
talk later about what all this means from a practical point of view), we can only have nonzero
solutions to the BVP if q
P π 2π 3π nπ
120,000 = 20 , 20 , 20 , . . . , 20 , . . .

This gives us the nonzero solutions

π nπ
y = C sin 20 x, y = C sin 2π 3π
20 x, y = C sin 20 x, . . . , y = C sin 20 x, . . .

to the boundary value problem.

154
 This isn’t really the end of the story, but we need to pause to catch our breath and develop some
terminology before resuming. At this point what we know about the boundary value problem

d2 y P
+ y = 0, y(0) = 0, y(20) = 0, (4)
dx2 EI
is that y = 0 is a solution, called the trivial solution (becauseq it is mathematically uninteresting),
P
and we only have nontrivial solutions for discrete values of 120,000 ; those solutions are the ones the
example concluded with. The non-trivial solutions are called buckling modes (for reasons you’ll soon
see). The first non-trivial solution is called the first buckling mode, the second is the second buckling
mode, and so on. The only values of P for which we can have nontrivial solutions are those that
satisfy q
P
120,000 (20) = π, 2π, 3π, . . . , nπ, . . . . (7)
Solving for P gives us
 2  2
 π 2 2π 3π  nπ 2
P = 120000 , 120000 , 120000 , . . . , 120000 , ...
20 20 20 20

= 300π 2 , 300(2π)2 , 300(3π)2 , . . . , 300(nπ)2 , . . .

= 300π 2 , 4(300π 2 ), 9(300π 2 ), . . . , n2 (300π 2 ), . . .

These values of P are called critical loads. Like the buckling modes, they are numbered, so 300π 2 is
the first critical load, 4(300π 2 ) is the second critical load, etc. The word “load” refers to the load
held up by the column. Note that the second critical load is four (two squared) times the first critical
load, the third critical load is nine (three squared) times the first critical load, and so on.
What happens physically is this: When there is no load on the column it is perfectly straight (the
solution y = 0), and it remains that way as we increase the load, until the first critical load is reached.
π
At that point the column will deflect sideways, taking the shape of the curve y = C sin 20 x, shown at
the left below. In reality, as the load increases beyond the first critical load, the deflection will remain
the same shape, but with increasing amplitude, until the column fails.
If we were able to prevent the middle point of the column, at x = 10, from deflecting, the column
would be able to support the second critical load. Because the column is held with y(10) = 0, the
deflection of the column will take the shape of a full period of the sine function, as shown in the middle
picture below - this is the second buckling mode. The act of preventing deflection is sometimes called
“pinning.” If we pin the column at points one-third and two-thirds of the way along its length, the
column would be able to support the third critical load, and the shape of the deflection would be given
by the third buckling mode, shown to the right below.

20 20 20

2
3 (20)
10
1
3 (20)

1st buckling mode 2nd buckling mode 3rd buckling mode

155
 Let’s revisit the boundary conditions y(0) = y(L) = 0, where L is the length of the column. We
should first note that the only requirement we really had was that the top and bottom of the column
were vertically aligned. We could just as well have put y(0) = y(L) = 2, as shown in the diagram to
the left below; however, the mathematics involved are a bit simpler if we instead use y(0) = y(L) = 0,
as we did. Physically we must still insist that the top and bottom be aligned. Without this restriction,
any horizontal shifting of the “ceiling” would result in hinging and a collapse. The beginning of such a
collapse is indicated by the three diagrams to the right below.

x Load
⇓ ⇓ ⇓
L

y (pos)
y(0) = y(L) = 2

Now suppose the top and bottom were embedded, rather than being pinned. If the column still has
length L, we then have the boundary conditions

y ′ (0) = y ′ (L) = 0. (5)

This gives us two boundary conditions, which is what we should need in order to solve the second order
ODE
d2 y P
2
+ y = 0,
dx EI
and leaves us without any conditions on y(0) and y(L). In this situation, we could conceivably allow
the “ceiling” to “drift” laterally without collapse, because the column being held in a vertical alignment
at the bottom and top would provide enough rigidity to prevent collapse. The first three buckling
modes for this situation are shown below; if the ceiling were prevented from drifting, the second mode
would then become the first, the fourth would become the second, and so on. You will investigate this
situation, along with the corresponding critical loads, in the exercises.

1st buckling mode 2nd buckling mode 3rd buckling mode

156
Section 5.3 Exercises To Solutions

Give all answers in exact form. When asked for the first ... buckling modes or critical loads,
give only nonzero modes or values.

1. For each of the following, boundary values are given to go with the ODE y ′′ + λ2 y = 0, which
has the solution
y = C1 sin λx + C2 cos λx.
Use the method of Example 5.3(a) to determine the first four nonzero values of λ for which the
boundary value problem has a solution, and give the corresponding four solutions.

(a) y(0) = 0, y(5) = 0 (b) y ′ (0) = 0, y ′ (3π) = 0

(c) y(0) = 0, y ′ (π) = 0 (d) y ′ (0) = 0, y(7) = 0

(e) y ′ (0) = 0, y ′ (10) = 0 (f) y(0) = 0, y(5π) = 0

2. A 12 foot vertical column is pinned at both ends. For the material it is made of we have
E = 500 and, from its design, we have I = 200, both in the appropriate units.

(a) Give the first four nonzero buckling modes, showing all steps of solving the IVP to get them.
(b) Give the first four nonzero critical loads.
(c) How many times larger is the third nonzero critical load than the first nonzero critical load?

3. Now suppose we have a 6 foot vertical column with E = 500, I = 200, and both ends pinned.

(a) Give the first four nonzero buckling modes, and find the corresponding critical loads.
(b) Compare your critical loads with those from the 12 foot length (Exercise 1). How do the
corresponding critical loads for the six foot column compare with those for the 12 foot
column? Does that make sense intuitively?
(c) The first nonzero buckling mode for the six foot column is the same as a six foot section of
which buckling mode for the 12 foot column? Draw a picture showing what is going on
here.

4. Consider now a 12 foot vertical column with E and I values of 500 and 200 again, but
this time with both ends embedded. Suppose also that the “ceiling” is not allowed to drift, so
the top and bottom of the column are vertically aligned.

(a) Give the first three nonzero buckling modes.

(b) Give the first three nonzero critical loads.
(c) How does the third nonzero critical load compare to the first? (That is, how many times
larger is it?)
(d) How does the first nonzero critical load compare with the first nonzero critical load for the
12 foot column with pinned ends? (See Exercise 1(a).) How about the other critical loads?

157
5. Repeat the previous exercise, but with the assumption that the ceiling is allowed to drift. (Hint:
You should be able to use your computations from Exercise 4, rather than re-doing all of them.)

6. Repeat parts (a) and (b) of Exercise 2 for a vertical column of length L that is pinned at both
ends, with modulus of elasticity E and moment of inertia I. This will give a general form of
the buckling modes and critical loads for pinned ends.

7. Repeat parts (a) and (b) of Exercise 5 for a vertical column of length L that is embedded at
both ends, with the ceiling allowed to drift. Again use a modulus of elasticity E and moment of
inertia I. This will give a general form of the buckling modes and critical loads for embedded
ends.

158
5.4 The Heat Equation in One Dimension

In this section we will consider the physically impossible but mathematically convenient situation: We
have a metal rod of length L (see picture below) that is perfectly insulated along its length, so that
no heat can enter or escape along its length, but for which
heat can enter or leave the ends. At any time t greater
than zero and any position x along the rod, the function
u(x, t) gives the temperature at that point x and time t . x=0 x=L
(We will think of the rod as being “infinitely thin,” so that
the rod has only one point at each x position. If you are not happy with this, an alternative is to think
that if the rod had some thickness the temperature at every point in a cross-sectional slice at some x is
the same, so we need not consider the other two space dimensions.) Suppose that at time zero there
is some distribution of temperatures along the rod, given by a the function f (x) for 0 ≤ x ≤ L, and
suppose also that the ends of the rod are held at temperature zero at all times t ≥ 0. The function
f gives an initial condition for each point x along the length of the rod, and the conditions that the
ends are held at temperature zero are boundary conditions.
What we would like to know is whether, and how, we can determine the temperature at any point
x with 0 < x < L (we know the temperatures at x = 0 and x = L are always zero), at any
time t > 0. Here the dependent variable u depends on the two independent variables x and t.
Some physical principles concerning heat give us a differential equation for this situation and, due to
there being two independent variables, it is a partial differential equation. The equation (called the heat
equation), and the conditions given above can all be stated as

∂u ∂2u
= k 2, u(0, t) = u(L, t) = 0, u(x, 0) = f (x) (1)
∂t ∂x
Here the conditions u(0, t) = u(L, t) = 0 are boundary conditions and u(x, 0) = f (x) is essentially an
initial condition (for every point along the rod). Thus we have a problem that is a sort of combination
initial value/boundary value problem. But we can really think of it as a boundary value problem for this
reason: If we were to think of the Cartesian plane as representing position
t
x along the horizontal axis and time t along the vertical axis we get a
picture like the one to the right, where each point in the shaded region
represents a point x in the rod and some time t. Our goal is then to find
the temperature at each of those points; in this way we can think of trying
to find function values in a region that is bounded by the line from zero to
L on the x-axis and the two “half-infinite” lines from zero to infinity in the
t direction at x = 0 and x = L. The condition u(x, 0) = f (x) can be
thought of as a boundary condition along the bottom, and the conditions L x
u(0, t) = u(L, t) = 0 are boundary conditions along the two sides.
∂f
Recall that if we have the function f (x, y) = x2 y 3 , to get the partial derivative we simply
∂x
∂f
take the derivative of x2 y 3 , treating y (and therefore y 3 ) as a constant. Similarly, we get by
∂y
treating x as a constant, so we have
∂f ∂f
= 2xy 3 and = 3x2 y 2
∂x ∂y
Let’s try another.

∂u ∂u
⋄ Example 5.4(a): Find and for u(x, t) = e−2t sin 3x.
∂x ∂t

159
 ∂u
Solution: When finding we consider t to be a constant, so e−2t is as well. The derivative
∂x
is then
∂u
= 3e−2t cos 3x.
∂x
∂u
When finding , sin 3t is essentially a constant, so
∂t
∂u
= −2e−2t sin 3x.
∂t

Now we’ll see that we have a solution to the heat equation!

∂u ∂2u
⋄ Example 5.4(b): Show that u(x, t) = e−2t sin 3x satisfies the heat equation = k 2.
∂t ∂x
Solution: We already have
∂u ∂u
= 3e−2t cos 3x and = −2e−2t sin 3x.
∂x ∂t
from the previous example. Taking the partial derivative of the first of these with respect to
x again gives us
∂2u
= −9e−2t sin 3x
∂x2
so
∂u 2 ∂2u
k = 29


= −2e−2t sin 3x = −9e−2t sin 3x = k 2 ,
∂t 9 ∂x

We will soon see that u(x, t) = e−2t sin 3x is not the most general solution to the equation.
Those who’ve had a multivariable calculus course will perhaps recall that the computation of partial
derivatives can be significantly more complicated (and therefore difficult) than the ones done above,
but if we understand the two examples just given we are ready to understand how
∂u ∂2u
= k 2, u(0, t) = u(L, t) = 0, u(x, 0) = f (x) (1)
∂t ∂x
is solved. The method for doing it is called separation of variables, which is similar in execution, up
to a point, to the method of the same name we used to solve separable first order ODEs up to a point.
After that point we must proceed differently.
To begin, we assume that the function u(x, t) is actually a product of a function of x alone and
another function of t alone. There is no practical reason to think this might be the case, but the
method works, so we’ll use it! (This method was invented/discovered in the 1700s by Daniel Bernoulli,
part of a family of a number of accomplished mathematicians and scientists. Daniel was also involved in
the derivation of the ODE we used for horizontal beams.) If we let X be the function of x and T be
the function of t, then u(x, t) = X(x)T (t). (This use of a capital letter for a function and the lower
case of the same letter for the independent variable is common practice in the study partial differential
equation solution methods.) Now remember that if we are taking the derivative of X(x)T (t) with
respect to x, T (t) is treated as a constant, and when taking the derivative with respect to t, X(x) is
treated as a constant, so
∂u ∂2u
= X(x)T ′ (t) and = X ′′ (x)T (t).
∂t ∂x2

160
The differential equation in (1) then becomes X(x)T ′ (t) = kX ′′ (x)T (t). If we divide both sides by
kX(x)T (t) we get
T ′ (t) X ′′ (x)
= . (2)
kT (t) X(x)
Here is where we deviate from the procedure for solving first order separable equations. (2) needs to be
true for all values of x and t, and this is likely only the case if both sides of (2) are equal to some
constant (again, we’ll see that it works!) that we will call −λ2 . For reasons we won’t go into here,
λ is positive. Setting each side equal to −λ2 and multiplying by the denominators we get
dT d2 X
(t) = −kλ2 T (t) and (x) = −λ2 X(x). (3)
dt dx2
In addition to this, we also have the boundary conditions X(0) = X(L) = 0 for the second equation.
The first equation in (3) tells us that T (t) is an eigenfunction for the first derivative operator, with
2
eigenvalue −kλ2 , and we know that T (t) is then any constant multiple of e−kλ t . That is,
2
T (t) = C1 e−kλ t .

The second equation says that X(x) is an eigenfunction of the second derivative operator with
eigenvalue −λ2 . Because λ is positive, the eigenfunctions are constant multiples of sin λx and
cos λx, as determined in the previous section, and we have

X(x) = C2 sin λx + C2 cos λx.

The general solution to the heat equation then looks like

2
u(x, t) = X(x)T (t) = e−kλ t (A sin λx + B cos λx) (4)

where A = C1 C2 and B = C1 C3 .
Let’s focus a bit more on the second ODE in (3) and its boundary values X(0) = X(L) = 0. The
general solution to the ODE is X(x) = A sin λx + B cos λx. Applying the condition X(0) = 0 gives
us B = 0, so the solution is X(x) = A sin λt. (At this point this story should be starting to feel
familiar!) We now consider the boundary condition X(L) = 0, which gives us 0 = A sin λL. As
before, when considering vertical columns, we don’t want to let A = 0, so we must have sin λL = 0.
This implies that
π 2π 3π
λL = 0, π, 2π, 3π, ... =⇒ λ = 0, , , , ...
L L L
and the solutions to the boundary value problem (disregarding constants and the zero solution arising
from λ = 0) are
π 2π 3π
sin x, sin x, sin , ...
L L Lx
kπ 2 4kπ 2 9kπ 2
The solution T then becomes T (t) = e− L2 t , e− L2
t
, e− L2
t
, ... depending on λ, so we get a
sequence of solutions u(x, t) = X(x)T (t):
kπ 2 4kπ 2 9kπ 2
u(x, t) = e− L2 t sin L
π
x, e− L2
t
sin 2π
L x, e
−
L2
t
sin 3π
L x, ... (5)

Recall that when solving an ODE like y ′′ + 3y ′ + 2y = 0 we assumed y = ert for some constant
r. From this we obtain y = e−t or y = e−2t , but we saw that the sum of constant multiples of
these two, y = C1 e−t + C2 e−2t is the most general solution. By the same reasoning, the most general
solution to the PDE we’re looking at is an infinite sum of the solutions in (5):
kπ 2 4kπ 2 9kπ 2 n2 kπ 2
u(x, t) = A1 e− L2 t sin Lπ x+A2 e− L2
t
sin 2π
L x+A3 e
−
L2
t
sin 3π
L x+· · ·+An e
−
L2
t
sin nπ
L x+· · · (6)

161
This story goes on quite a bit longer, but let’s end it with the following. In order to try to meet the
condition u(x, 0) = f (x) we must have

f (x) = A1 sin Lπ x + A2 sin 2π 3π nπ

L x + A3 sin L x + · · · An sin L x + · · · (7)

The right hand side of (7) is something called a Fourier series. This brings up the question

In what way (or ways) do we interpret the equal sign in (7), and for what functions
f can such interpretation(s) be made?

Attempts to answer this question gave birth to a large amount of mathematics over many years, starting
with Joseph Fourier’s work in the early 1800s, and with a major result proved as late as 1966. Perhaps
some of you will investigate this subject more in later coursework.

162
5.5 Chapter 5 Summary
• Boundary value problems arise when the independent variable of an ODE is length. Applications
include the deflection of horizontal beams and vertical columns along their lengths.

• The differential equation for a horizontal beam is fourth order, and the solution is a fourth order
polynomial with four arbitrary constants.

• There are two boundary conditions at each of the two ends of a horizontal beam, giving four
conditions used to determine the values of the constants.

• There are three possible end conditions for each end of a horizontal beam:

– Embedded: This is when the end of the beam is “clamped” horizontally. The mathematical
conditions for such an end are y = 0 and y ′ = 0.
– Simply Supported (Pinned): This is when the beam is held up but allowed to pivot.
Mathematically, y = 0 and y ′′ = 0.
– Free: This is when an end is completely unsupported, and the other end must be embedded.
The mathematical conditions for such an end are y ′′ = 0 and y ′′′ = 0.

• Let A be an operator that operates on functions and y a nonzero function. If there is a constant
λ such that
Ay = λy
then y is an eigenfunction of the operator A, with corresponding eigenvalue λ.

• The ODE for a vertical column is second order, and the solution is either a sine function or a
cosine function, depending on the end conditions:

– When the ends are pinned (hinged) the solution is a sine function.
– When the ends are embedded the solution is a cosine function.

• Mathematically, there are infinitely many solutions for a vertical column that is pinned at its ends.

– Each is some multiple of a half period of a sine function beginning at x = 0.

– The first solution, called the first buckling mode, is a single half-period of the sine function.
This occurs physically when the column is allowed to deflect over its entire length.
– Each additional solution (buckling mode) consists of n2 periods of the sine function for
n = 2, 3, 4, 5, .... Physically, the solution consisting of n2 periods of the sine function
occurs when the the column is pinned along its length at n − 1 equally spaced points.

• Mathematically, there are infinitely many solutions for a vertical column that is embedded at its
ends.

– Each is some multiple of a half period of a cosine function beginning at x = 0.

– In the case that the ends of the column are embedded, it is physically possible that the
ceiling can float (move laterally).
– If the ceiling is allowed to float the first buckling mode is a single half-period of the cosine
function. Each additional buckling mode consists of n2 periods of the cosine function for
n = 2, 3, 4, 5, ....

163
 – If the ceiling is NOT allowed to float the first buckling mode is a single period of the cosine
function. Each additional buckling mode consists of n periods of the cosine function for
n = 2, 3, 4, 5, ....

• The load that causes the first buckling mode is called the first buckling load, and the nth
buckling load leads to the nth buckling load.

• The nth buckling load is n2 times the first buckling load.

164
D Solutions to Exercises
D.5 Chapter 5 Solutions
Section 5.1 Solutions Back to 5.1 Exercises

1. (a) y(0) = 0, y ′′ (0) = 0, y(12) = 0, y ′ (12) = 0

(b) y(0) = 0, y ′ (0) = 0, y ′′ (8) = 0, y ′′′ (8) = 0
(c) y(0) = 0, y ′ (0) = 0, y(20) = 0, y ′′ (20) = 0
(d) y(0) = 0, y ′′ (0) = 0, y(15) = 0, y ′′ (15) = 0

2. Only (a) is possible, y(0) = 0, y ′ (0) = 0, y(8) = 0, y ′′ (8) = 0

d4 y
3. (a) (30)(80) = 150, y(0) = y ′ (0) = y(8) = y ′ (8) = 0
dx4
1 4 1 3
(b) y = 384 x − 24 x + 61 x2
(d) The maximum deflection should appear at the middle of the beam (x = 4). The deflection
there is 23 .
1 4 10 3
4. (b) y = 12 x − 3 x + 50x2
(d) The maximum deflection is 2500 at x = 10, the right hand end of the beam.
1 4
5. (b) y = 12 x − 35 x3 + 250
3 x
3125
(d) The maximum deflection is 12 at x = 5

d4 y
6. (a) (30)(80) = 150, y(0) = y ′ (0) = y(8) = y ′′ = 0
dx4
1 4 5 3
(b) y = 384 x − 96 x + 41 x2
(d) The maximum deflection is about 1.39 at x = 4.6

7. (a) J and L (b) E and H (c) F and G

Section 5.2 Solutions Back to 5.2 Exercises

1. (a) The eigenvalue is λ = 3. (b) y = e−5x

(c) Eigenfunction: y = ekx Eigenvalue: k

2. (a) y = 3x, y = 5, y = 2x − 1, etc. (b) y = Ax + B

√ √
(c) y = sin 2x, y = cos 2x (d) y = sin 3x, y = cos 3x

3. (a) D(e−2t ) = 4e−2t − 4e−2t − 3e−2t = −3e−2t , the eigenvalue is −3

(b) D(ekt ) = k2 ekt + 2kekt − 3ekt = (k2 + 2k − 3)ekt , the eigenvalue is k2 + 2k − 3
(c) k = −3, 1 (d) k = −4, 2

208
Section 5.3 Solutions Back to 5.3 Exercises

1. (a) λ = π5 , 2π 3π 4π
5 , 5 , 5 , ..., y = C sin π5 x, C sin 2π 3π 4π
5 x, C sin 5 x, C sin 5 x, ...

(b) λ = 13 , 2 3 4
3 , 3 , 3 , ..., y = C cos 13 x, C cos 23 x, C cos 33 x, C cos 43 x, ...
(c) λ = 12 , 3 5 7
2 , 2 , 2 , ..., y = C sin π5 x, C sin 2π 3π 4π
5 x, C sin 5 x, C sin 5 x, ...
π 3π 5π 7π π
(d) λ = 14 , 14 , 14 , 14 , ..., y = C cos 14 x, C cos 3π 5π 7π
14 x, C cos 14 x, C cos 14 x, ...
π 2π 3π 4π π
(e) λ = 10 , 10 , 10 , 10 , ..., y = C cos 10 x, C cos 2π 3π 4π
10 x, C cos 10 x, C cos 10 x, ...

(f) λ = 51 , 2 3 4
5 , 5 , 5 , ..., y = C sin 15 x, C sin 25 x, C sin 53 x, C sin 45 x, ...
π
2. (a) y = C sin 12 x, y = C sin 2π 3π 4π
12 x, y = C sin 12 x, y = C sin 12 x, . . .
6250 2
(b) P = 9 π , 4( 6250 2 6250 2 6250 2
9 π ), 9( 9 π ), 16( 9 π ), . . .

(c) The third critical load is nine times the first critical load.

3. (a) y = C sin π6 x, y = C sin 2π 3π 4π

6 x, y = C sin 6 x, y = C sin 6 x, . . .
P = 25000 2 25000 2 25000 2 25000 2
9 π , 4( 9 π ), 9( 9 π ), 16( 9 π ), . . .

(b) Each critical load is four times the corresponding critical load for the twelve foot column.
(c) The first buckling mode for the six foot column is the same as the second buckling mode
for the twelve foot column, but is half as high, so it only includes half a period of the sine
function, whereas the twelve foot column’s second buckling mode has a full period of the
sine function.

4. (a) y = C cos 2π 4π 6π
12 x, y = C cos 12 x, y = C cos 12 x, . . .

(b) P = 4( 6250 2 6250 2 6250 2

9 π ), 16( 9 π ), 36( 9 π ), . . .

(c) The third critical load is nine times the first critical load.
(d) Each critical load is four times the corresponding critical load for the column with pinned
ends.
π
5. (a) y = C cos 12 x, y = C cos 2π 3π 4π
12 x, y = C cos 12 x, y = C cos 12 x, . . .
6250 2
(b) P = 9 π , 4( 6250 2 6250 2 6250 2
9 π ), 9( 9 π ), 16( 9 π ), . . .

(c) The third critical load is (again!) nine times the first critical load.
(d) The critical loads are the same as those for the pinned ends.

6. (a) y = C sin Lπ x, y = C sin 2π 3π 4π

L x, y = C sin L x, y = C sin L x, . . .
EI 2
(b) P = L2
π , 4( EI
L2
π 2 ), 9( EI
L2
π 2 ), 16( EI
L2
π 2 ), . . .

7. (a) y = C cos Lπ x, y = C cos 2π 3π 4π

L x, y = C cos L x, y = C cos L x, . . .
EI 2
(b) P = L2
π , 4( EI
L2
π 2 ), 9( EI
L2
π 2 ), 16( EI
L2
π 2 ), . . .

209