44

Applications to PDE Problems

The use of the Sturm-Liouville theory can be nicely illustrated by solving any of a number of
classical problems. We will consider two: a heat ow problem and a vibrating string problem.
The rst, determining the temperature distribution throughout some heat conducting rod, is
basically the same problem Fourier rst solved using Fourier series. In some ways, this can be
considered the historical starting point of the development of the Sturm-Liouville theory. The
second is the problem of modeling the motion of an elastic string stretched between two points,
a problem undoubtedly of interest to all guitar and banjo players.
44.1 The Heat Flow Problem
Setting Up the Problem
Here is the problem: We have a heat conducting rod of length L , and we want to know how
the temperature at different points in the rod varies with time. To keep our discussion relatively
simple, we assume the rod is one-dimensional, uniform, positioned along the Xaxis with
endpoints at x = 0 and x = L , and with the endpoints being kept at a temperature of 0 degrees
(Fahrenheit, Celsius, Kelvin the actual scale is irrelevant for us). Let
u(x, t ) = temperature of the rods material at horizontal position x and time t .
The endpoint conditions can then be written as
u(0, t ) = 0 and u(L, t ) = 0 for all t .
Lets also assume the rods initial temperature distribution is known; that is, we assume
u(x, 0) = f (x)
where f is some known function on (0, L) . Let us further assume that f is at least piecewise
smooth on the interval.
* Most of this chapter was stolen from chapter 16 of Principles of Fourier Analysis by Howell (he wont mind). Ive
cut out irrelevant stuff and made minor changes to t our discussion.
4/15/2012
Chapter & Page: 442 Applications to PDE Problems
If this were a text on thermodynamics or partial differential equations, we would now derive
the heat equation. But this isnt such a text, so well simply assume that, at every point in the
rod,
u
t
=

2
u
x
2
where is some positive constant describing the thermal properties of the rods material. This
is the famous heat equation derived by Fourier.
1
Our goal is to nd a usable formula for u(x, t ) . Since it only makes sense to talk about the
temperature where the rod exists, x must be between 0 and L . Gathering all the assumptions
from above, we nd that u(x, t ) must satisfy the following system of equations:
u
t

2
u
x
2
= 0 for 0 < x < L (44.1a)
u(0, t ) = 0 and u(L, t ) = 0 (44.1b)
u(x, 0) = f (x) for 0 < x < L (44.1c)
Implicit in this is the requirement that u(x, t ) be a sufciently smooth function of x and t
for the above equations to make sense. Remember, is a positive constant, and f is a known
piecewise smooth function on (0, L) . At this point we have no reason to place any limits on
t other than it must be real valued. So, for now, we will assume no other limits on t . Later,
however, we may need to modify that assumption.
2
A Formal Solution
Solving this problem starts with the rather bold assumption that it has a solution. Supposing this,
let us try to nd a suitable generalized Fourier series representation for this solution u(x, t ) . For
reasons not yet clear, lets use the one generated by the Sturm-Liouville problem

(x) = for 0 < x < L

with
(0) = 0 and (L) = 0 .
Do observe that the boundary conditions at x = 0 and x = L are the same as in the above heat
ow problem. That is one reason we chose this Sturm-Liouvile problem.
From this Sturm-Liouville problem we get the set of sine functions
_
sin
_
k
L
x
_
: k = 1, 2, 3, . . .
_
which, fromour Sturm-Liouville theory, we knowis a complete, orthogonal set of functions with
respect to weight function w(x) = 1 . We also know that, for each xed value of t , we represent
the solution to the above heat ow problem, u(x, t ) , using the corresponding sine series
u(x, t ) =

k=1
b
k
sin
_
k
L
x
_
1
Each person reading this should go through the derivation of the heat equation at least once in their life. Reasonable
derivations can be found in most introductory texts on partial differential equations.
2
Part of solving many a problem is determining just what the problem is, and what can or should be considered
as known at the onset. Here, for example, we know we can nd u(x, t ) for all time t . We are wrong.
The Heat Flow Problem Chapter & Page: 443
where
b
k
=
_
sin
_
k
L
x
_

u(x, t )
_
_
_
_sin
_
k
L
x
__
_
_
2
= =
2
L
_
L
0
u(x, t ) sin
_
k
L
x
_
dx .
Note the that formula for the b
k
s depends on t . So these coefcients are not constants but
functions of t , and we really should rewrite the above as
u(x, t ) =

k=1
b
k
(t ) sin
_
k
L
x
_
(44.2a)
where
b
k
(t ) =
2
L
_
L
0
u(x, t ) sin
_
k
L
x
_
dx . (44.2b)
Note, also, that the above formula for computing each b
k
(t ) is pretty much useless since we
dont have any formula for u(x, t ) that formula is just what we want to nd.
Our goal is to nd a formula for u(x, t ) by nding the formulas for the b
k
s . To do this, we
plug the above series representationfor u into the heat equation. Naively compute the derivatives
in the heat equation by differentiating the terms in the series,
u
t
=

t

k=1
b
k
(t ) sin
_
k
L
x
_
=

k=1

t
_
b
k
(t ) sin
_
k
L
x
__
=

k=1
b
k

(t ) sin
_
k
L
x
_
and

2
u
x
2
=

2
x
2

k=1
b
k
(t ) sin
_
k
L
x
_
=

k=1

2
x
2
_
b
k
(t ) sin
_
k
L
x
__
=

k=1
b
k
(t )
_

_
k
L
_
2
sin
_
k
L
x
_
_
=

k=1
_
k
L
_
2
b
k
(t ) sin
_
k
L
x
_
.
With these expressions for the derivatives, equation (44.1a) becomes

k=1
b
k

(t ) sin
_
k
L
x
_
+

k=1
_
k
L
_
2
b
k
(t ) sin
_
k
L
x
_
= 0 .
Letting
=
_

L
_
2
,
this can be written more concisely as

k=1
_
b
k

(t ) + k
2
b
k
(t )
_
sin
_
k
L
x
_
= 0 .
Chapter & Page: 444 Applications to PDE Problems
Look at this last equation. For each value of t , the left-hand side looks like a sine series
which, according to the equation, equals 0 for all x in (0, L) . Surely, this is only possible if
each coefcient is 0 . Here, though, the coefcients are expressions involving the b
k
s . So each
of these expressions must equal 0 . This gives us a bunch of differential equations,
db
k
dt
+ k
2
b
k
= 0 for k = 1, 2, 3, . . . . (44.3)
These differential equations are easy to solve. Each is nothing more than
dy
dt
+ y = 0
with y = b
k
and = k
2
one of the simplest rst order linear equations around. You
should have no problem conrming that its general solution is y = Be
t
where B is an
arbitrary constant. Hence,
b
k
(t ) = B
k
e
k
2
t
for k = 1, 2, 3, . . .
where the B
k
s are yet unknown constants.
With these formulas for the b
k
s , formula (44.2a) becomes
u(x, t ) =

k=1
B
k
e
k
2
t
sin
_
k
L
x
_
. (44.4)
In deriving this expression for u(x, t ) , we assumed u(x, t ) exists and satises the heat equation
(equation (44.1a)) and the endpoint conditions in equation set (44.1b). All that remains is to
further rene our expression so it also satises the initial condition of equation (44.1c), u(x, 0) =
f (x) . Using the above formula for u(x, t ) in this equation, we get
f (x) = u(x, 0) =

k=1
B
k
e
k
2
0
sin
_
k
L
x
_
=

k=1
B
k
sin
_
k
L
x
_
for 0 < x < L . Cutting out the middle yields
f (x) =

k=1
B
k
sin
_
k
L
x
_
for 0 < x < L ,
which, by an amazing coincidence, looks exactly as if we are representing our known function
f by its sine series. Surely then, each B
k
must be the corresponding sine coefcient for f ,
B
k
=
2
L
_
L
0
f (x) sin
_
k
L
x
_
dx .
That nishes our derivation. If the solution exists and our (occasionally naive) suppositions
are valid, then our heat ow problem (equation set (44.1)) is solved by
u(x, t ) =

k=1
B
k
e
k
2
t
sin
_
k
L
x
_
(44.5a)
where
=
_

L
_
2
(44.5b)
The Heat Flow Problem Chapter & Page: 445
and
B
k
=
2
L
_
L
0
f (x) sin
_
k
L
x
_
dx for k = 1, 2, 3, . . . . (44.5c)
This set of formulas is often called a formal solution to the heat equation problem because we
obtained it through a process of formal manipulations which seemed reasonable, but were not
all rigorously justied.
!

Example 44.1: Consider solving our heat ow problem when L = , = ln 2 , and the
rod is initially a constant temperature throughout, say,
f (x) = 100 .
Here:

/
L
= 1 , formula (44.5b) simplies to
=
_

L
_
2
= ln 2 ,
and
e
k
2
t
= e
k
2
(ln 2)t
=
_
e
ln 2
_
k
2
t
= 2
k
2
t
for k = 1, 2, 3, . . . .
Formula (44.5c) yields
B
k
=
2
L
_
L
0
f (x) sin
_
k
L
x
_
dx
=
2

_

0
100 sin(kx) dx =
200
k
_
1 (1)
k
_
.
Hence, according to formula (44.5a), the formal solution to this heat ow problem is
u(x, t ) =

k=1
B
k
e
k
2
t
sin
_
k
L
x
_
=

k=1
200
k
_
1 (1)
k
_
2
k
2
t
sin(kx)
=
400

_
1
2
_
t
sin(x) +
400
3
_
1
2
_
3
2
t
sin(3x) +
400
5
_
1
2
_
5
2
t
sin(5x)
+
400
7
_
1
2
_
7
2
t
sin(7x) +
400
9
_
1
2
_
9
2
t
sin(9x) + .
Validity and Properties of the Formal Solution
The question remains as to whether formula set (44.5) is a valid solution to our heat owproblem.
There are several parts to this question: Does the series converge for all values of x and t of
interest? If so, is the resulting function suitably smooth for the expressions in equation set (44.1)
to make sense, and if so, does this function satisfy those equations?
Chapter & Page: 446 Applications to PDE Problems
Partial answers to these questions, along with some insight, can be gained by examining the
terms of our series,
B
k
e
k
2
t
sin
_
k
L
x
_
for k = 1, 2, 3, . . . .
Remember > 0 and
|B
k
| =

2
L
_
L
0
f (x) sin
_
k
L
x
_
dx

2
L
_
L
0
| f (x)|

sin
_
k
L
x
_

dx
2
L
_
L
0
| f (x)| dx .
So, letting
A =
2
L
_
L
0
| f (x)| dx ,
we see that

B
k
e
k
2
t
sin
_
k
L
x
_

A e
k
2
t
for k = 1, 2, 3, . . . .
If t is also positive, then each e
k
2
t
shrinks to 0 very rapidly as k . This ensures that
the series formula for u(x, t ) converges absolutely. Consequently, we are assured that u(x, t ) ,
as dened by formula set (44.5), is well dened when 0 x L and 0 < t .
It can be shown that these exponentially decreasing terms ensure that, as long as t > 0 ,
u(x, t ) is an innitely smooth function of both x and t whose partial derivatives can all be
computed by differentiating the series term by term. This will allow us to rigorously conrm
our formal solution to be a valid solution to our heat ow problem (and a very nice one, at that)
when t > 0 .
On the other hand, if t < 0 , then e
k
2
t
= e
k
2
|t |
as k . Thus, unless the B
k
s
shrink to 0 extremely rapidly as k , the terms of our series solution will blow up, and the
series itself diverges whenever t < 0 .
In short:
The series formula given by formula set (44.5) succeeds beautifully as a solution to
our heat ow problem for t > 0 and, typically, fails miserably for t < 0 .
There is something else worth noting about our series solution: Each term in that series,
u(x, t ) =

k=1
B
k
e
k
2
t
sin
_
k
L
x
_
,
rapidly shrinks to 0 as t . From this it can readily be shown that the maximum and
minimum temperatures in the rod must be approaching 0 degrees fairly quickly as t gets large.
?

Exercise 44.1: Let u(x, t ) be the innite series solution found in above example 44.1.
a: Verify that
|u(x, t )| <
400

k=1
_
1
2
_
t k
when t > 0 . (44.6)
The Vibrating String Problem Chapter & Page: 447
b: Using the above and the formula for computing the sum of a geometric series, show that
|u(x, t )| <
800

_
1
2
_
t
when t 1 .
c: Assuming u(x, t ) is the temperature distribution in a rod, what does the above tell you
about the maximum temperature in the rod when t = 1 ? when t = 2 ? when t = 10 ?
?

Exercise 44.2: Again, let u(x, t ) be the innite series solution found in example 44.1,
above. This time, consider this innite series when t = 1 .
a: Write out this series.
b: Verify that this series does not converge to anything when x =

/
2
.
c: Show that this series cannot be the sine series for any piecewise continuous function on
(0, ) . (Remember, if it were the sine series for such a function, then the coefcients would
be bounded.)
44.2 The Vibrating String Problem
Setting Up the Problem
Envision an elastic string (such as you might nd on any guitar or banjo) stretched between two
xed points on the Xaxis, say, from x = 0 to x = L (with L > 0 ). For simplicity, well
assume the string only moves vertically, and we will let
u(x, t ) = vertical position at time t of the portion of string located at horizontal
position x .
Because the ends of the string are xed at x = 0 and x = L , u(x, t ) is only dened for
0 x L , and we have the endpoint conditions
u(0, t ) = 0 and u(L, t ) = 0 .
After making a few idealizations and applying a little physics, it can be shown that

2
u
t
2
= c
2

2
u
x
2
where c is some positive constant (the reason for using c
2
instead of c will be clear later).
3
This is the basic (one-dimensional) wave equation.
4
We will assume the initial shape of the string is given by the graph of some known function
f on (0, L) ,
u(x, 0) = f (x) for 0 < x < L .
3
More precisely, c =
_

where and are, respectively, the tension in and the linear density of the string
when the stretched string is at rest.
4
Another famous equation whose derivation we are skipping. Look it up in any decent introductory book on partial
differential equations.
Chapter & Page: 448 Applications to PDE Problems
For most (unbroken) strings we would expect f to be continuous and piecewise smooth. In
addition, since the string is fastened at the endpoints, we should have f (0) = 0 and f (L) = 0 .
As it turns out, this is not quite enough to completely specify u(x, t ) . An additional initial
condition is necessary. We will take that condition to be
u
t

(x,0)
= 0 for 0 < x < L .
In other words, we assume the string is not moving at time t = 0 . This would be the case, for
example, if we held the string in some xed shape until releasing it at t = 0 .
Gathering all the above equations together, we nd that u(x, t ) must satisfy the following
system of equations:

2
u
t
2
c
2

2
u
x
2
= 0 for 0 < x < L (44.7a)
u(0, t ) = 0 and u(L, t ) = 0 (44.7b)
u(x, 0) = f (x) for 0 < x < L (44.7c)
u
t

(x,0)
= 0 for 0 < x < L (44.7d)
Again, there is an implicit requirement that u(x, t ) be a sufciently smooth function for the above
equations to make sense. Keep in mind that c is a positive constant and f is a known uniformly
continuous and piecewise smooth function on (0, L) satisfying f (0) = 0 = f (L) . (Later we
will realize that f

must also be piecewise smooth.) While it is reasonable to be interested in

the solving this problem just for t > 0 , such a restriction on t turns out to be mathematically
unnecessary. So we will assume the above equations are valid for < t < .
A Formal Solution
The process of nding a solution to our vibrating string problemis very similar to the process we
went through to solve our heat ow problem. As then, we begin by supposing a solution u(x, t )
exists, and, as with our heat ow problem, the end conditions (equation set (44.7b)) suggest that
u(x, t ) should be represented by a Fourier sine series on 0 < x < L with the coefcients being
functions of time,
u(x, t ) =

k=1
b
k
(t ) sin
_
k
L
x
_
. (44.8)
As we noted with the heat ow problem, this formula equals 0 when x = 0 or x = L .
Naively differentiating, we get

2
u
t
2
=

2
t
2

k=1
b
k
(t ) sin
_
k
L
x
_
=

k=1
b
k

(t ) sin
_
k
L
x
_
and

2
u
x
2
=

2
x
2

k=1
b
k
(t ) sin
_
k
L
x
_
=

k=1
b
k
(t )
_
k
L
_
2
sin
_
k
L
x
_
.
The Vibrating String Problem Chapter & Page: 449
With these expressions for the derivatives, equation (44.7a) becomes

k=1
b
k

(t ) sin
_
k
L
x
_
+ c
2

k=1
b
k
(t )
_
k
L
_
2
sin
_
k
L
x
_
= 0 ,
which is written more concisely as

k=1
_
b
k

(t ) + (k)
2
b
k
(t )
_
sin
_
k
L
x
_
= 0
using, for lexicographic convenience,
=
c
L
.
This time each b
k
must satisfy the second order linear differential equation
d
2
b
k
dt
2
+ (k)
2
b
k
= 0 .
Again, we have a simple differential equation that should be familiar to anyone who has had an
elementary course in differential equations. Its solution is
b
k
(t ) = A
k
sin(kt ) + B
k
cos(kt )
where A
k
and B
k
are constants yet to be determined.
With this formula for the b
k
s , equation (44.8) becomes
u(x, t ) =

k=1
[A
k
sin(kt ) + B
k
cos(kt )] sin
_
k
L
x
_
. (44.9)
The A
k
s and B
k
s will be determined by the initial conditions, equations (44.7c) and (44.7d).
For the second initial condition, we will need the partial of u with respect to t , which we might
as well (naively) compute here:
u
t
=

t

k=1
[A
k
sin(kt ) + B
k
cos(kt )] sin
_
k
L
x
_
=

k=1
[ A
k
k cos(kt ) B
k
k sin(kt )] sin
_
k
L
x
_
. (44.10)
Combining formula (44.9) for u(x, t ) with the rst initial condition gives us
f (x) = u(x, 0) =

k=1
[A
k
sin(k0) + B
k
cos(k0)] sin
_
k
L
x
_
=

k=1
[A
k
0 + B
k
1] sin
_
k
L
x
_
for x in (0, L) . Thus, we have
f (x) =

k=1
B
k
sin
_
k
L
x
_
for 0 < x < L ,
Chapter & Page: 4410 Applications to PDE Problems
which looks remarkably like an equation we obtained while solving our heat ow problem. As
before, we are compelled to conclude that the B
k
s are the Fourier sine coefcients for f . That
is,
B
k
=
2
L
_
L
0
f (x) sin
_
k
L
x
_
dx for k = 1, 2, 3, . . . .
The second initial condition, along with formula (44.10), yields
0 =
u
t

(x,0)
=

k=1
[A
k
k cos(k0) B
k
k sin(k0)] sin
_
k
L
x
_
=

k=1
[A
k
k 1 B
k
k 0] sin
_
k
L
x
_
=

k=1
A
k
k sin
_
k
L
x
_
for 0 < x < L ,
strongly suggesting that
A
k
= 0 for k = 1, 2, 3, . . . .
Our derivation is complete. If our vibrating string problem (equation set (44.7)) has a
solution and our (occasionally naive) computations are valid, then that solution is given by
u(x, t ) =

k=1
B
k
cos
_
kc
L
t
_
sin
_
k
L
x
_
(44.11a)
where
B
k
=
2
L
_
L
0
f (x) sin
_
k
L
x
_
dx for k = 1, 2, 3, . . . . (44.11b)
Once again, we have derived a formal solution, that is, a formula obtained through formal
(naive) manipulations which we hope can be rigorously veried later.
!

Example 44.2: Consider solving our vibrating

1 0
1
/
2
1
/
2
X
Figure 44.1: The initial shape of the
string in example 44.2.
string problemassuming L = 1 and c = 3 , and
starting with the middle point of the string pulled
up a distance of
1
/
2
(see gure 44.1). That is,
u(x, 0) = f (x) with
f (x) =
_
x if 0 x
1
/
2
1 x if
1
/
2
x 1
.
The Vibrating String Problem Chapter & Page: 4411
(a) (b) (c)
L L L
Figure 44.2: The (a) rst harmonic, (b) second harmonic, and (c) third harmonic for a
vibrating string of length L sketched at various times as functions of x .
With these choices, equation (44.11b) is
B
k
=
2
L
_
L
0
f (x) sin
_
k
L
x
_
dx
= 2
_ 1
/
2
0
x sin(kx) dx + 2
_
1
1
/
2
(1 x) sin(kx) dx
= = sin
_
k
2
_ _
2
k
_
2
.
Thus, according to formula (44.11a), the solution to this vibrating string problem is
u(x, t ) =

k=1
sin
_
k
2
_ _
2
k
_
2
cos(k3t ) sin(kx)
= 1
_
2

_
2
cos(3t ) sin(x) + 0
_
2
2
_
2
cos(2 3t ) sin(2x)
+ (1)
_
2
3
_
2
cos(3 3t ) sin(3x) + 0
_
2
4
_
2
cos(4 3t ) sin(4x)
+ .
Harmonics of a Vibrating String
One advantage of the Fourier series solution to our vibrating string problem is that it allows us to
analyze the sound produced by such a string by looking at the components of the series solution.
For convenience, lets rewrite that solution as
u(x, t ) =

k=1
B
k
u
k
(x, t )
where
u
k
(x, t ) = sin
_
k
L
x
_
cos(2
k
t ) and
k
=
kc
2L
.
The individual u
k
s are often referred to as the modes of vibration or the harmonics, with u
1
being the rst or fundamental mode/harmonic. The graphs of the rst three harmonics
u
1
(x, t ) , u
2
(x, t ) , and u
3
(x, t ) have been sketched as functions of x for various values of
Chapter & Page: 4412 Applications to PDE Problems
t in gure 44.2. Notice that u
k
(x, t ) is nothing more than a sine function of x being scaled by
a sinusoid function of time with frequency
k
. It is that
k
which determines the pitch of the
sound resulting from that mode of vibration. The magnitude of B
k
, of course, helps determine
the loudness of the sound due to the k
th
harmonic, with the perceived loudness increasing as
B
k
increases. (However, the relation between B
k
and the apparent loudness is not linear and is
strongly inuenced by the ability of the ear to perceive different pitches.)
In theory, one can produce a pure tone corresponding to any one of these frequencies (say

3
) by imposing just the right initial condition (namely, u(x, 0) = u
3
(x, 0) ). In practice, this
is very difcult, and the sound heard is usually a combination of the sounds corresponding to
many of the harmonics. Typically, most of the sound heard is due to the fundamental harmonic,
because, typically, people pluck strings in such a manner that the rst harmonic is the dominant
term in the series solution. For example, whether in a violin or a banjo,
1
is approximately
440 cycles/second for a string tuned to A above middle C. The other harmonics provide the
overtones that modify the sound we hear and help us distinguish between a vibrating violin
string and a vibrating banjo string.
!

Example 44.3: In exercise 44.2 we obtained

u(x, t ) = 1
_
2

_
2
cos(3t ) sin(x) + 0
_
2
2
_
2
cos(2 3t ) sin(2x)
+ (1)
_
2
3
_
2
cos(3 3t ) sin(3x) + 0
_
2
4
_
2
cos(4 3t ) sin(4x)
+
as a solution to a vibrating string problem. From this we see that the rst four harmonics for
this string are
u
1
(x, t ) = sin(x) cos(2
1
t ) , u
2
(x, t ) = sin(2x) cos(2
2
t ) ,
u
3
(x, t ) = sin(3x) cos(2
3
t ) and u
4
(x, t ) = sin(4x) cos(2
4
t )
where

1
=
3
2
,
2
= 3 ,
3
=
9
2
and
4
= 6 .
The fundamental harmonic frequency is
1
=
3
/
2
, and the other harmonic frequencies are
integral multiples of the fundamental. In this case, the rst harmonic is certainly the dominant
component of the above solution. However, if the the units on t are seconds, then the rst
harmonic frequency of
3
/
2
cycles per second is somewhat below what most people can hear,
and so, as far as most people are concerned, the rst harmonic will not contribute signicantly
to the sound heard from this vibrating string.
Additional Exercises
44.3. Consider the series solution to the heat ow problem of exercise 44.1 on page 445.
Using the rst 25 terms of this solution, sketch the temperature distribution throughout
Additional Exercises Chapter & Page: 4413
the rod at t = 0 , t =
1
/
10
, t = 1 , and t = 10 . (Use the computer math package such
as Maple or Mathematica or Mathcad.)
44.4. Using the formal solution derived in the rst section of this chapter, nd the solution to
the heat ow problem described in equation set (44.1) on page 442 assuming L = 2 ,
= 3 , and
a. f (x) = 5 sin(x) b. f (x) = x
Which of these solutions will be valid for all t and which will just be valid for t > 0 ?
44.5. If the endpoints of our heat conducting rod are insulated instead of being kept at 0 de-
grees, then the temperature distribution u(x, t ) satises the following set of equations:
u
t

2
u
x
2
= 0 for 0 < x < L , 0 < t
u
x

(0,t )
= 0 and
u
x

(L,t )
= 0 for 0 < t
u(x, 0) = f (x) for 0 < x < L
where and L are positive constants, and f is piecewise smooth on (0, L) .
a. Why, in this case, would it be better to represent u(x, t ) using a cosine series,
u(x, t ) =
0
(t ) +

k=1

k
(t ) cos
_
k
L
x
_
,
instead of the sine series used for the problemin the rst section of this chapter? What
is the Sturm-Liouville problem that gives rise to this series?
b. Derive the formal series solution for this heat ow problem.
c. Find the solution to this problem assuming = 2 , L = 3 , and
f (x) =
_
1 if 0 < x <
3
/
2
0 if
3
/
2
< x < 3
,
and sketch the temperature distribution (using the rst 25 terms of your series solution)
for t = 0 , t =
1
/
10
, t = 1 , and t = 10 .
d. What happens to the solution found in the last part as t ? Sketch the temperature
distribution at t = .
e. What can be said about the differentiability of the solution derived above in the rst
part of this exercise?
44.6. If our heat conducting rod contains sources of heat, and we start with the rod at 0
degrees and keep the endpoints at 0 degrees, then the temperature distribution u(x, t )
satises the following set of equations:
u
t

2
u
x
2
= f (x) for 0 < x < L , 0 < t
Chapter & Page: 4414 Applications to PDE Problems
u(0, t ) = 0 and u(L, t ) = 0 for 0 < t
u(x, 0) = 0 for 0 < x < L
Again, and L are positive constants, and f is piecewise smooth on (0, L) .
a. Derive the formal series solution to this problemassuming that a solution exists. (Hint:
Start with formula (44.2a).)
b. Find the solution to this problem assuming = 4 , L = 3 , and
f (x) =
_
1 if 1 < x < 2
0 otherwise
,
and sketch the temperature distribution (using the rst 25 terms of your series solution)
for t = 0 , t =
1
/
10
, t = 1 , and t = 10 .
c. What happens to the solution found in the last part as t ? Sketch the temperature
distribution at t = .
d. What can be said about the differentiability of the solution derived above in the rst
part of this exercise?
44.7. Find the formal solution u(x, t ) to the following vibrating string problem:

2
u
t
2
c
2

2
u
x
2
= 0 for 0 < x < L
u(0, t ) = 0 and u(L, t ) = 0
u(x, 0) = 0 for 0 < x < L
u
t

(x,0)
= f (x) for 0 < x < L
where L and c are positive constants and f is piecewise smooth on (0, L) .
44.8. A more realistic model for the vibrating string that takes into account the dampening of
the vibrations due to air resistance is partially given by the equations

2
u
t
2
+ 2
u
t
c
2

2
u
x
2
= 0 for 0 < x < L ,
u(0, t ) = 0 and u(L, t ) = 0
where L , and c are positive constants with being much smaller than c (assume
L < c for the following).
a. Derive, as completely as possible, the formal series solution to the above system
of equations. Because no initial conditions are given, your answer should contain
arbitrary constants.
b. How rapidly do the vibrations die out?
c. How does term modify the frequencies at which the individual terms of the
solution vibrate?
Additional Exercises Chapter & Page: 4415
Chapter & Page: 4416 Applications to PDE Problems
Some Answers to Some of the Exercises
WARNING! Most of the following answers were prepared hastily and late at night. They
have not been properly proofread! Errors are likely!
1c. xmax. temp. < 128 , max. temp. < 64 , max. temp. < 0.25
4a. 5e
3
2
t
sin(x)
4b.

k=1
(1)
k+1
4
k
e
3
k
t
sin
_
k
2
x
_
where
k
=
_
k
2
_
2
5a.

= with

(0) = 0 and

(L) = 0
5b. A
0
+

k=1
a
k
e

k
t
cos
_
k
L
x
_
where
k
=
_
k
L
_
2
,
A
0
=
1
L
_
L
0
f (x) dx and a
k
=
2
L
_
L
0
f (x) cos
_
k
L
x
_
dx
5c.
1
2
+

k=1
2
k
sin
_
k
2
_
e

k
t
cos
_
k
3
x
_
where
k
= 2
_
k
3
_
2
6a.

k=1
c
k

k
_
1 e

k
t
_
sin
_
k
L
x
_
where
k
=
_
k
L
_
2
and c
k
=
2
L
_
L
0
f (x) sin
_
k
L
x
_
dx
7.

k=1
a
k
sin
_
kc
L
t
_
sin
k
L
x where a
k
=
2
kc
_
L
0
f (x) sin
k
L
x dx
8a.

k=1
[A
k
cos(2
k
t ) + B
k
sin(2
k
t )] e
t
sin
_
k
L
x
_
where the A
k
s and B
k
s are arbitrary constants and
k
=
1
L
_
(kc)
2