This article was downloaded by: [The University of Manchester Library]

On: 04 March 2013, At: 03:35

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954
Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH,
UK
PRIMUS: Problems, Resources,
and Issues in Mathematics
Undergraduate Studies
Publication details, including instructions for
authors and subscription information:
http://www.tandfonline.com/loi/upri20
Discovering the Laplace
Transform in Undergraduate
Differential Equations
Terrance J. Quinn & Sanjay Rai
Version of record first published: 05 Aug 2008.
To cite this article: Terrance J. Quinn & Sanjay Rai (2008): Discovering the Laplace
Transform in Undergraduate Differential Equations, PRIMUS: Problems, Resources, and
Issues in Mathematics Undergraduate Studies, 18:4, 309-324
To link to this article: http://dx.doi.org/10.1080/10511970601131613
PLEASE SCROLL DOWN FOR ARTICLE
Full terms and conditions of use: http://www.tandfonline.com/page/terms-
and-conditions
This article may be used for research, teaching, and private study purposes.
Any substantial or systematic reproduction, redistribution, reselling, loan,
sub-licensing, systematic supply, or distribution in any form to anyone is
expressly forbidden.
The publisher does not give any warranty express or implied or make any
representation that the contents will be complete or accurate or up to
date. The accuracy of any instructions, formulae, and drug doses should be
independently verified with primary sources. The publisher shall not be liable
for any loss, actions, claims, proceedings, demand, or costs or damages
whatsoever or howsoever caused arising directly or indirectly in connection
with or arising out of the use of this material.
P
R
I
M
U
S

2
0
0
8
.
1
8
:
3
0
9
-
3
2
4
.

d
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
t
a
n
d
f
o
n
l
i
n
e
.
c
o
m
PRIMUS, XVIII(4): 309324, 2008
Copyright Taylor & Francis Group, LLC
ISSN: 1051-1970 print / 1935-4053 online
DOI: 10.1080/10511970601131613
Discovering the Laplace Transform in
Undergraduate Differential Equations
Terrance J. Quinn and Sanjay Rai
Abstract: The Laplace Transform is an object of fundamental importance in pure
and applied mathematics. In addition, it has special pedagogical value in that it can
provide a natural and concrete setting for a student to begin thinking about the modern
concepts of operator and functional. Most undergraduate textbooks, however, merely
define the transform, and go on to properties and applications. The present article is
intended to help students discover the Laplace Transform for themselves. The empha-
sis, then, is neither elaboration of technique nor axiomatic proof, but initial discovery
of the basic idea.
Keywords: Laplace Transform, discovery, differential equations, undergraduate
mathematics.
1. INTRODUCTION
The Laplace Transform is an object of fundamental importance in the under-
graduate mathematics curriculum. It goes back at least to the work of Pierre-
Simon Laplace (17491827), who made extensive use of the formula in his
probability theory [3, p. 419]. Since the middle of the twentieth century, it
has become a hallmark of most textbooks on differential equations, and its
usefulness in applications is well known. In addition, it has special pedago-
gical value in that it can provide a natural and concrete setting for a student to
begin thinking about the modern concepts of operator and functional.
Despite its lineage, what seems to be lacking in the literature and the
textbooks is a pedagogical introduction. Instead, the common approach starts
with the formula, and follows with some combination of techniques, verified
properties and applications. Advanced treatises, of course, give rigorous
proofs of the analytic foundations of the formula. Certainly, seeking grounds
Address correspondence to Terrance J. Quinn, Department of Mathematical Sciences,
Middle Tennessee State University, P.O. Box 34, Murfreesboro TN 37132 USA.
E-mail: [email protected]
P
R
I
M
U
S

2
0
0
8
.
1
8
:
3
0
9
-
3
2
4
.

d
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
t
a
n
d
f
o
n
l
i
n
e
.
c
o
m
for a result and going on to axiomatic proofs are creative and eminently
worthwhile goals that help characterize the very nature of mathematical
endeavor. Every endeavor, though, has a beginning. The question, therefore,
remains: How might one discover the Laplace transform for one-self? For the
teacher, how might one lead a student to consider such a formula in the first
place? The present article is intended to complement existing undergraduate
textbooks, by helping the undergraduate differential equations student get
started with the basic idea. The emphasis is neither elaboration of techniques
nor axiomatic verification, but discovery of a focal idea.
Certain works should be mentioned to help contextualize the present
article. In 1945 Widder wrote the article What is the Laplace Transform?
[4]. Widders article makes no claim to pedagogical purpose. It takes the
transform as a starting place, and gives some answer to what the transform is,
in the sense of relating the formula to other transform equations. More
recently, there is the 1992 article called Teaching the Laplace Transform
Using Diagrams [3]. That article also takes the transform as its starting
place, and goes on to a different pedagogical purpose regarding ways to
evaluate the transform. As the authors of [3] state, they present an approach
to evaluation of the Laplace Transform and its inverse using commutative
diagrams [3, p. 309].
Unfortunately now out-of-print, Braun [1] gives a useful but brief intro-
ductory discussion of how the Laplace Transform can resemble the natural
logarithm. This, in fact, ties in nicely with the present approach. The present
article though is intended to be more complete. Of course there can be many
ways to discovery, so there is no claim that the present approach offers a
unique discovery path to the Laplace Transform, or that it is even indicative
of the understanding that Laplace had. The primary intention is to help the
student discover and appreciate the formula for themselves, in the context of
undergraduate differential equations.
The article is intentionally informal and conversational. It is directed
mainly to student readers who are already familiar with topics usually treated
in the first weeks of a first undergraduate course in differential equations. In
particular, it will be assumed that the reader knows something about how
solutions to a second-order homogeneous equation with constant coefficients
can be generated by combinations of exponential functions. We pose various
questions along the way. For the student reader, it will be helpful for your
learning if you spend at least a little time on each question before reading on.
The second section of the article reviews some ideas from algebra that
can be helpful in understanding the transform. The third section explores the
Key Hypothesis that combinations of exponential functions generate solu-
tions. Fourth is the section that leads to initial discovery of the transform, and
the fifth is on using the transform. The section concludes with a short
application to Newtons Law of Cooling, followed by a summary of the
Laplace Transform approach. The sixth section raises further questions, and
310 Quinn and Rai
P
R
I
M
U
S

2
0
0
8
.
1
8
:
3
0
9
-
3
2
4
.

d
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
t
a
n
d
f
o
n
l
i
n
e
.
c
o
m
briefly points to well known results found in textbooks, and the seventh
consists of a few concluding remarks.
2. AN ALGEBRAIC PROBLEM
As we will find out, certain algebraic calculations can help one appreciate
both the strategy and the techniques of the Laplace Transform. Lets begin
then by reviewing what for many readers will be a familiar problem in basic
algebra.
Find a quadratic function y = ax
2
+ bx + c that solves the equation y = ax
2
+
bx + c = 4x
2
+ 7x + 12.
Now, I imagine that more than one student reader has jumped to the
(correct) solution a = 4, b = 7, c = 12. It can be worthwhile though to pause,
to review what that solution means.
Why is that the solution? What is it that we are asked to find?
We are looking for a quadratic function. In other words, we need to solve
the equality ax
2
+ bx + c = 4x
2
+ 7x + 12 for all values of x for which the
formula makes sense. So, finding the quadratic function that solves the
problem happens to reduce to finding only three numbers a, b, c.
Can you see how one of the coefficients can be identified immediately?
Algebraically, what is a value for x that simplifies things considerably?
Equality is to hold for all x, so how about trying x = 0, to see what that
gives us? We get a(0)
2
+ b(0) + c = 4(0)
2
+ 7(0) + 12. It follows that c = 12.
What about the other coefficients a and b? We have already made use of x = 0.
Now that we know that c = 12, we seek the function equality ax
2
+ bx + 12 =
4x
2
+ 7x + 12.
A traditional approach is to continue in this way, by taking it one step at
a time. At each stage, start by substituting x = 0; find the degree zero
coefficient at that stage; subtract that constant to obtain a new equation
where all terms have x degree at least one. (In the example just given, we
would get ax
2
+ bx = 4x
2
+ 7x.) Then, assuming x ,= 0, divide by x ,= 0 to
reduce the degree of the equation by one, and thereby isolate the next
constant. (In the example, this would give ax
2
+ b = 4x
2
+ 7.) Repeat.
Note, however, that when we re-substitute x = 0 at a subsequent stage, it
might appear that there is a difficulty with the logic - and indeed at first
glance there is. For in order to isolate the next constant, we first divide by x ,=
0. In what sense can we then go on to set x = 0 again? In fact, it is possible to
make this approach rigorous by continuity and by taking appropriate limits.
While this approach is valid, it is ad hoc and somewhat inefficient.
There is a more systematic and purely algebraic approach that exploits the
fact that an equality of functions means equality for all x in a common domain.
Substituting three distinct values x
1
, x
2
, x
3
into the equation ax
2
+ bx + c = 4x
2
+
7x + 12 produces three linear equations for the three unknowns a, b, c:
Discovering the Laplace Transform 311
P
R
I
M
U
S

2
0
0
8
.
1
8
:
3
0
9
-
3
2
4
.

d
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
t
a
n
d
f
o
n
l
i
n
e
.
c
o
m
ax
2
1
bx
1
c = 4x
2
1
7x
1
12
ax
2
2
bx
2
c = 4x
2
2
7x
2
12
ax
2
3
bx
3
c = 4x
2
3
7x
3
12
As the reader may recall (or investigate), there exists a unique solution to a
linear system exactly when the determinant (or volume element) is non-zero.
In the present case, a straightforward calculation shows that the determinant
is the Vandermonde determinant,
det
1 x
1
x
2
1
1 x
2
x
2
2
1 x
3
x
2
3
_
_
_
_
= x
3
x
2
( ) x
3
x
1
( ) x
2
x
1
( );
which evidently is not zero under the hypothesis that the values x
1
, x
2
, x
3
are
distinct. So, this approach provides a linear system of equations that can be
solved using any one of the familiar methods based on substitution and
elimination.
The approach that is perhaps most efficient as well as most appropriate
for a course in differential equations is to use calculus operations from the
beginning. If ax
2
+ bx + c = 4x
2
+ 7x + 12 for all x, then as before, we can
substitute x = 0 to get c = 12. But, if two functions are identical, then what
does that mean about their slopes? In other words, the derivatives must also
be equal. Note that there is no change of domain, and this evidently produces
linear functions. So we can again substitute x = 0, this time into the derivative
equation 2ax + b = 8x + 7. And so on.
There are three main intended purposes to this pre-amble around solving
a polynomial function equation. A first is to help the reader recall that the
meaning of a function equation is that the equality is to hold for all x for which
the functions are defined. A second purpose is to illustrate that even though a
function can be defined for all real numbers, under the additional constraint
that the function be a polynomial, solving the equation reduces to finding just
a few constants (namely, the coefficients). A third purpose is to recall how
calculus can play a role in helping solve a function equation. For again, if two
functions are identical, then so are their first derivatives, their second deriva-
tives, and so on.
3. THE KEY HYPOTHESIS
Consider the initial-value problem
y
//
3y
/
2y = e
3t
; y 0 ( ) = 1; y
/
0 ( ) = 0
312 Quinn and Rai
P
R
I
M
U
S

2
0
0
8
.
1
8
:
3
0
9
-
3
2
4
.

d
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
t
a
n
d
f
o
n
l
i
n
e
.
c
o
m
As mentioned in Section 1, we assume that the reader has some familiarity
with the results from the first few weeks of a first course in differential
equations. You will therefore need to recall and/or review how solutions to
such equations typically are linear combinations of exponential functions.
We go then directly to the
KEY HYPOTHESIS: We take as a basic hypothesis that solutions are
combinations of exponential functions.
Note the similarity with the algebraic problem discussed above, where
the form of the unknown function is pre-assigned. For the purposes of
illustration, suppose in fact that a solution of the equation y 3y + 2y =
e
3t
, y(0) = 1, y(0) = 0 is of the simple form y = ae
At
+ be
Bt
, a, A, b, B complex
numbers. The problem then is to find these four complex numbers a, A, b, B.
We can certainly substitute y = ae
At
+ be
Bt
into the differential equation
to get
(aA
2
e
At
+ bB
2
e
Bt
) 3(aAe
At
+ bBe
Bt
) + 2(ae
At
+ be
Bt
) = e
3t
. Collecting
terms, we get
a A
2
3A 2
_ _
e
At
b B
2
3B 2
_ _
e
Bt
= e
3t
: (1)
By including the initial values y(0) = 1, y (0) = 0, we get two additional
equations
a b = 1
aA bB = 0
The left side of Equation 1 is a combination of two possibly distinct
powers of the function e
t
, namely, (e
t
)
A
and (e
t
)
B
; while the right hand
side is (e
t
)
3
.
If this were a polynomial equation, then, as we reviewed in Section 2, we
could simply compare coefficients. But, the exponential terms make things
somewhat more subtle, for not only do we have powers of e
t
, but the
unknowns are partly in the exponents. Can you see one way at least, to get
rid of the exponential terms, in a way that gives a relationship involving the
unknown constants a, A, b, B? Remember that any non-zero number raised to
the power zero gives unity. So, as we did with the polynomial equation, we
could try t = 0 to get a(A
2
3A + 2) + b(B
2
3B + 2) = 1.
This last equation has many solutions, and thus does not uniquely
determine the four unknowns a, A, b, B. Typically we would need as many
as four equations to uniquely determine four unknowns. One idea might be to
imitate the systematic algebraic approach that works for finding coefficients
to polynomial equations. In other words, we might substitute three more
Discovering the Laplace Transform 313
P
R
I
M
U
S

2
0
0
8
.
1
8
:
3
0
9
-
3
2
4
.

d
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
t
a
n
d
f
o
n
l
i
n
e
.
c
o
m
values for t to produce three more equations. But, for the present equation, it
turns out that there is a difficulty with that approach. Can you see what that
is? Try it and see what happens. Unknowns remain in the exponents.
Evidently, it might well be helpful to use more than traditional algebra to
solve this problem.
4. DISCOVERING THE TRANSFORM
To solve an arithmetic problem, one uses the operations of arithmetic,
namely, addition, subtraction, multiplication and division. To solve an alge-
braic equation, one uses the operations of algebrawhich in fact are the
operations of arithmetic but more precisely defined, generalized, and with a
broader application.
A differential equation is an equation that involves not only some
unknown function, but also includes information on rates of change of the
unknown function (first derivative, second derivative, and so on.) In other
words, a differential equation is basically a calculus equation involving
derivativeshence the name differential. So, in order to solve a differential
equation we have at our disposal all of the operations and results of calculus.
What are these operations and results?
There are, of course, the algebraic operations for functions (addition,
subtraction, multiplication and division); there is also evaluation of a func-
tion at a particular element in its domain; certain limits can exist - in
particular a limit of ratios can give an exact ratio called the derivative;
and when the derivative is applied to an area function, it becomes
evident from the geometry that the rate of change of an area is the length
of the moving front-line - that is,
d
dx
Area [ [ = length of front-line [ [.
Using Riemanns 19
th
century notation for area under a curve, the
Fundamental Theorem can be written as the familiar looking formula
d
dx
_
x
0
f t ( )dt
_
_
_
_
= f x ( ).
Now that we have highlighted the operations and basic results
from calculus, lets take another look at the differential equation y
3y + 2y = e
3t
, y(0) = 1, y(0) = 0. Recall the Key Hypothesis. We are
hoping to find a solution that is a combination of exponential functions,
and for the moment at least, we are looking for a solution of the form
y = ae
At
+ be
Bt
.
The solution to the Algebraic Problem in Section 2 was a polynomial
function determined by the finite collection of coefficients, one coefficient
for each power of x. In trying to solve the differential equation y 3y + 2y =
e
3t
, y(0) = 1, y(0) = 0, is there perhaps something similar going on?
314 Quinn and Rai
P
R
I
M
U
S

2
0
0
8
.
1
8
:
3
0
9
-
3
2
4
.

d
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
t
a
n
d
f
o
n
l
i
n
e
.
c
o
m
Again, a real function can be defined across a large domain. In the present
setting, however, we have the Key Hypothesis. It follows that a solution will
be completely and uniquely determined by simply four real numbers,
namely, a, A, b, B. Unlike the algebraic problem, we do not have a right
hand side function with which to directly compare coefficients of y(t). We
do, however, have information on how the growth rates (derivatives) are
related to each other. And when we combine that information with the Key
Hypothesis, we get Equation 1, that is, a(A
2
3A + 2) e
At
+ b(B
2
3B + 2)
e
Bt
= e
3t
, together with the initial value equations a + b = 1, aA + bB = 0.
This may seem like we are back where we were at the end of the third
section, but there is a difference. For, we are on the move with a strategy -
namely, the strategy of trying to exploit the full potential of the now-
identified operations of calculus. Recalling the economy of using derivatives
to solve the Algebraic Problem of the second section, one might be inspired
try that same approach with Equation 1. But, unlike what happens for
polynomials, differentiating an exponential function does not reduce its
degree, but instead actually brings additional factors into play. Not all is
lost. For, we still have two main calculus operations to try, that is, integration
and evaluating limits. So, our problem is now more focused: Is there some
strategic way to use integration and/or limits to obtain information about a,
A, b, B?
Remembering the Key Hypothesis, lets try to get some clues from a
numerical example, say y = e
7t
. In this example, we are given up front that
the exponent is 7. But the issue now is to see whether or not integration and/
or limit techniques might be used to extract that exponent from the given
exponential function. Trying integration, we get
_
x
0
e
7t
dt =
1
7
e
7x
1
_ _
.
Does this get us anywhere? We have used integration. There is one operation
left. Is there a limit here that might be useful? Evidently, as x , the
integral approaches
1
7
, the negative reciprocal of the exponent. Now, this is
starting to look promising. For, having extracted the reciprocal of the expo-
nent (in a systematic way) means that we can now easily get the exponent
itself.
Lets not rush on though. Not all improper integrals converge. What was
it about the exponent that provided the existence of this limit as x ?
Probably you see that the exponent being negative is what did the trick. But,
surely, not all differential equations will be solvable in terms of exponential
functions with strictly negative exponents. Does this mean that investigating
the limit of an integral may not be generally strategic for our present
purposes?
To see what might be done, lets look to an example with a positive
exponent. In fact, to make such an example, lets just change the sign of the
Discovering the Laplace Transform 315
P
R
I
M
U
S

2
0
0
8
.
1
8
:
3
0
9
-
3
2
4
.

d
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
t
a
n
d
f
o
n
l
i
n
e
.
c
o
m
previous exponent to get y = e
7t
. Then
_
x
0
e
7t
dt =
1
7
e
7x
1
_ _
. Clearly, in this
case, as x gets arbitrarily large, the integral diverges to infinity. So, if there is
some hope that limits of integrals will be of use, we will need to get around
this difficulty.
Is there some way to adjust the equation so that convergent improper
integrals are obtained? The problem at the moment is with the positive
exponent. But, recall two things from basic algebra: When q ,= 0,
p
q
=
r
q
if
and only if p = r; and
e
Ax
e
sx
= e
As ( )x
. Two things follow: For any s, Equation 1
is equivalent to
a A
2
3A 2
_ _
e
sA ( )t
b B
2
3B 2
_ _
e
sB ( )t
= e
s3 ( )t
: (2)
The freedom in choosing s allows us to require that s be large enough in order
to ensure that all of the corresponding improper integrals converge as x
. In the present paragraph, how large is large enough? That is, how large
must s be in order that each of the exponents (s + A), (s + B), (s + 3) is
negative? Clearly, we need s > 3, s > A and s > B. As we did for the
numerical example given by y = e
7t
, integrating Equation 2 across the
interval [0, x] and taking the limit of all terms as x , we get
a A
2
3A 2 ( )
s A

b(B
2
3B 2)
s B
=
1
s 3
: (3)
Remember though, that A, B are two of four unknowns, so even if we assume
that s is sufficiently large so that all integrals converge as x , we do not
get a specific value for how large that must be.
Still, the clues are beginning to gather, and we are in fact getting closer
to being able to define the Laplace Transform. If we make the assumption
that s is sufficiently large so that all integrals converge, we get rational
functions in s on both sides of the integrated equation. This is a good thing
for our purposes. For, by clearing denominators, the equation in rational
functions of one real variable can be converted to an equivalent polynomial
equation in the new variable s. And from our earlier work on the Algebraic
Problem, this then gives the sort thing we might be able to use to pin down
the unknown constants. In particular, clearing the denominators from
Equation 3, we get
a A
2
3A 2
_ _
s B ( ) s 3 ( ) b B
2
3B 2
_ _
s A ( ) s 3 ( ) = s A ( ) s B ( ):
316 Quinn and Rai
P
R
I
M
U
S

2
0
0
8
.
1
8
:
3
0
9
-
3
2
4
.

d
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
t
a
n
d
f
o
n
l
i
n
e
.
c
o
m
Collecting powers of s gives the quadratic equation
s
2
a A
2
3A 2
_ _
b B
2
2B 2
_ _
1
_
s a A
2
3A 2
_ _
B 3 ( ) b B
2
3B 2
_ _
A 3 ( ) A B ( )
_
3a A
2
3A 2
_ _
B 3b B
2
3B 2
_ _
A AB
_
= 0
Using the algebraic method of the section second gives three equations
a A
2
3A 2
_ _
b B
2
3B 2
_ _
1 = 0
a A
2
3A 2
_ _
B 3 ( ) b B
2
3B 2
_ _
A 3 ( ) A B ( ) = 0
3a A
2
3A 2
_ _
B 3b B
2
3B 2
_ _
A AB = 0
Recall that we also have the equations coming from the initial-values
y (0) = 1, y(0) = 0, namely,
a b = 1
aA bB = 0
This certainly seems to be closing in on the unknowns a, A, b, B.
However, while the calculations are beginning to show some promise,
several questions may have occurred to the reader. In the example under
discussion, there are five equations and only four unknowns. Since there
are more equations than unknowns, does this system actually have a
solution?
Note also that for the present discussion it was assumed that our
unknown function had only two exponential terms, and so was of the special
form y = ae
At
+ be
Bt
. What if, still in keeping with the Key Hypothesis, the
correct solution is actually of the form y = ae
At
+ be
Bt
+ ce
Ct
or some even
larger sum of exponential terms? (The solution is given in the fifth section.)
Evidently, in order to explicitly allow for all possibilities indicated by the
Key Hypothesis, we need to assume that y(t) is of the general form y(t) =

a
i
e
Ait
(with the number of terms left open). Assuming that there is suitable
convergence of sums, we get y(t) =

a
i
A
i
e
Ait
and y
//
t ( ) =

a
i
A
2
i
e
A
i
t
. We
would then substitute these terms into the second-order differential equation
and proceed as above to obtain finally a polynomial in s and therefore a
system of algebraic equations.
Are we now at an impasse with this approach? How can we solve an
unknown number of simultaneous algebraic equations for an unspecified
sequence of unknowns a
1
, A
1
, a
2
, A
2
, a
3
, A
3
, . . . .? Lets take a fresh look
at what we have so far. As we have found, if we make explicit use of the
Key Hypothesis, algebraic difficulties become formidable.
Discovering the Laplace Transform 317
P
R
I
M
U
S

2
0
0
8
.
1
8
:
3
0
9
-
3
2
4
.

d
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
t
a
n
d
f
o
n
l
i
n
e
.
c
o
m
Nevertheless, using the factor
1
e
st
and improper integration to identify
exponential terms seems promising.
So, lets try again, keeping this idea in play, but now suppressing explicit
reference to the unknowns a
1
, A
1
, a
2
, A
2
, a
3
, A
3
, . . . . In other words, instead
of writing the explicit form of the unknown function, look to the pattern of
(calculus) operations. The derivation of Equation 3 then can be expressed by
the one calculus equation
_

0
y
//
3y
/
2y
e
st
dt =
_

0
e
3t
e
st
dt; y 0 ( ) = 1; y
/
0 ( ) = 0
This, of course, is the same as
_

0
e
st
y
//
dt 3
_

0
e
st
y
/
dt 2
_

0
e
st
y dt
=
_

0
e
st
e
3t
dt; y 0 ( ) = 1; y
/
0 ( ) = 0
(4)
In keeping with our results so far, we note the recurring form of integral in
Equation 4 and obtain the following standard definition:
Definition: Let f(t) be defined for 0 _ t < . The Laplace Transform of
f(t), denoted F(s), or sometimes {f(t)} (s), is given by the formula
F s ( ) = f t ( ) s ( ) =
_

0
e
st
f t ( )dt
Note that the definition includes the hypothesis that the initial value f(0) is known.
5. USING THE TRANSFORM
In the transform notation, the differential equation
y
//
3y
/
2y = e
3t
; y(0) = 1; y
/
(0) = 0
is written
y
//
t ( ) s ( ) 3 y
/
t ( ) s ( ) 2 y t ( ) s ( ) = e
3t
_ _
s ( )
318 Quinn and Rai
P
R
I
M
U
S

2
0
0
8
.
1
8
:
3
0
9
-
3
2
4
.

d
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
t
a
n
d
f
o
n
l
i
n
e
.
c
o
m
Or, in terms of standard integral notation,
_

0
e
st
y
//
dt 3
_

0
e
st
y
/
dt 2
_

0
e
st
ydt =
_

0
e
st
e
3t
dt;
y(0) = 1; y
/
(0) = 0
(5)
Do any of the integrals look familiar, at least in form? Take the middle
integral say, that is,
_

0
e
st
y
/
dt. To solve the initial-value problem is to find
the unknown function y(t). Some of the terms of Equation 5 involve first
and second derivatives of y(t), which, in principle at least, could add
unavoidable complexity. In the present case, however, do you see a way
to evaluate so that the integrals are given in terms of the one unknown
function y(t)?
Even if you dont see it yet, keep trying, for that effort has its own
value. If you are stuck, let us re-express the integral in a way that should
clinch it for you: For the moment, we take s as a fixed sufficiently large real
number, so we can keep notation somewhat simplified and not mention s at
all (at least for the time being). Give e
st
a new name, by letting u(t) = e
st
.
In this slightly simplified notation, the integral
_

0
e
st
y
/
dt becomes
_

0
u(t)y
/
dt.
Is the integration-by-parts formula now coming to mind? Check to see
for yourself that as long as all terms converge appropriately, we get
_

0
u t ( )y
/
dt = s
_

0
u t ( )ydt y 0 ( ):
Using the transform notation, this is
y
/
(t) s y(t) y(0) (6)
Obviously, we can apply this same pattern to the second-order derivative as
well, to get
_

0
u(t)y
//
dt = s
_

0
u(t)y
/
dt y
/
(0)
Discovering the Laplace Transform 319
P
R
I
M
U
S

2
0
0
8
.
1
8
:
3
0
9
-
3
2
4
.

d
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
t
a
n
d
f
o
n
l
i
n
e
.
c
o
m
Again, using the transform notation, we get
y
//
(t) = s y
/
t ( ) y
/
0 ( )
It is then left as a simple but useful exercise to combine these two equations
to get
y
//
t ( ) = s
2
y t ( ) sy 0 ( ) y
/
0 ( ): (7)
Substituting these results into our initial-value problem
y 3y + 2y = e
3t
, y(0) = 1, y(0) = 0, and recallingthe notation {y(t)} = Y(s),
we get
s
2
Y(s) s 3 sY s ( ) 1 [ [ 2Y s ( ) =
1
s 3
:
This means that we can avoid the complexities that emerge with explicit use
of the Key Hypothesis, and instead take the higher viewpoint of identifying
the operations of the transform equation, in terms of the transform of the as
yet unknown function y(t). The background Key Hypothesis, though,
remains in place. And so we are trying to find y(t), where the information
about the exponents and constants of y(t) =

a
i
e
Ait
is encapsulated by
s
2
Y s ( ) s 3 sY s ( ) 1 [ [ 2Y s ( ) =
1
s 3
Some algebra then reveals that the transform Y(s) satisfies
Y s ( ) =
5
2
1
s 1
_ _
2
1
s 2
_ _

1
2
1
s 3
_ _
:
What, then, is the original function y (t)? Recall that the calculations that got
this approach going in the first place. What kind of exponential functions
produce each of these summands in s? See the fourth section, and, in
particular, Equation 3. What is the transform of a single exponential function
ae
At
? A straightforward calculation gives that
ae
At
_ _
=
a
s A
: (8)
Evidently, the terms in Y (s) can be obtained as the transforms of
5
2
e
t
, 2e
2t
and e
3t
respectively. Hence, there is an obvious candidate for the solu-
tion y(t), namely y t ( ) =
5
2
e
t
2e
2t

1
2
e
3t
. A straightforward calculation
320 Quinn and Rai
P
R
I
M
U
S

2
0
0
8
.
1
8
:
3
0
9
-
3
2
4
.

d
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
t
a
n
d
f
o
n
l
i
n
e
.
c
o
m
verifies that indeed this is a solution to the initial-value problem. Note that by
the existence and uniqueness theorem, among continuous solutions we also
get that the solution just given is unique.
5.1. ApplicationNewtons Law of Cooling
The Laplace Transform is not needed to solve Newtons Law of Cooling
(a first-order homogenous ordinary differential equation). Applying the trans-
form to the Law of Cooling, however, can be instructive. For, while the
equation is familiar, it is also involved enough to call on the main features of
the transform approach.
As can be found in many calculus books, the Law of Cooling is given
by the equation
dT
dt
= k T T
room
( ), where T is temperature of a cooling
body and k > 0 is a positive constant. Suppose that you have a cup of tea
that starts out at 200

F, and that the conference room you are in is at a

temperature of 72

F. The conference session is to go on for some time,

and you wonder how your tea will cool. According to the Law of Cooling,
the rate of change of the tea temperature is given by
dT
dt
= k T 72 ( ),
T (0) = 200.
Let L(s) be the Laplace Transform of the unknown temperature function
T(t). The transform of the left side can be calculated using Equation 6
(integration by parts). The transform of the constant term kT
room
= 72k can
be easily calculated directly. We therefore get sL 200 = kL
72k
s
Solving
for L gives L s ( ) =
200
sk

72k
s sk ( )
. Note that, so far, we have been able to use the
transform by isolating terms of the form
a
sA
. (See Equation 8.) Can we use this
approach here? As the reader may recall from the algebra of partial fractions,
1
s sk ( )
=
1
k
1
s

1
sk
_ _
. Substituting this into the last equation,
(s) =
200
s k
72k
1
k
1
s

1
s k
_ _
=
128
s k

72
s
=
128
s (k)

72
s 0
:
Appealing now to what we already know about the Laplace Transform, we
get T(t) = 128e
kt
+ 72. In other words, the temperature drops off exponen-
tially in time, and levels off toward the 72

F room temperature.
5.2. Summary of the Approach
Motivated by classical results frequently taught in a first course in differential
equations, one can make the Key Hypothesis that solutions of certain
differential equations are of the form y (t) = Sa
i
e
Ait
. The problem is to then
Discovering the Laplace Transform 321
P
R
I
M
U
S

2
0
0
8
.
1
8
:
3
0
9
-
3
2
4
.

d
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
t
a
n
d
f
o
n
l
i
n
e
.
c
o
m
identify the coefficients a
i
and the exponents A
i
. In order to obtain an
equation that, in particular, involves the exponents A
i
, we look to the impro-
per integral of the (initial-value) differential equation. In order to assure that
the improper integrals suitably converge, we divide all terms by e
st
where s is
taken to be sufficiently large. The improper integral of
ae
t
e
st
= ae
st
e
At
is
a
sA
. Next, observe that by appealing to the integration by parts formula,
integrals of terms e
st
y
(n)
(t) that involve higher order derivatives can all be
expressed in terms of the integral of e
st
y(t). We can then avoid explicit
reference to the constants a
i
, A
i
. Instead, the problem of solving a differential
equation is transformed into the purely algebraic problem of solving for
Y(s) = y(t) (s) =
_

0
e
st
y t ( ) dt as a rational function in s. We then use
algebra to decouple the rational function into its component summands of the
form Y(s) =

a
i
sA
i
. Questions of uniqueness aside, we obtain a solution to
the original differential equation, namely, y(t) =

a
i
e
Ait
.
Note that the transform is not only elegant, but has numerous applica-
tions in science and technology. For instance, in some cases a direct approach
toward finding solutions of a differential equation can pose significant
computational challenges, while at the same time the corresponding trans-
formed algebraic equation can be more tractable.
6. FURTHER QUESTIONS ABOUT THE LAPLACE TRANSFORM
The s variable: Is there more that we can say about the s - variable? The
Key Hypothesis is that a solution is of the form y(t) =

a
i
e
Ait
. As already
described, the formal transform of y(t) = Sa
i
e
Ait
is Y(s) =

a
i
sA
i
. Evidently,
the singularities of the formal transform Y(s) are identical with the real
exponents of the solution function y(t). The s - variable, therefore, provides
an ambient set for all possible real exponents {A
i
} that occur in the solution
function y(t) = Sa
i
e
Ait
.
Stability: As just mentioned, the singularities of the transform
Y(s) =

a
i
s A
i
correspond with exponents of solutions of the differential equation. The
Laplace transform therefore gives information on asymptotic behavior of
solutions. But, in a first course in differential equations, asymptotic stability
is more commonly determined by looking to the real parts of the roots of the
characteristic equation. It is therefore natural to enquire into a possible
relationship between the two approaches, at least with regard to asymptotic
stability of solutions.
322 Quinn and Rai
P
R
I
M
U
S

2
0
0
8
.
1
8
:
3
0
9
-
3
2
4
.

d
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
t
a
n
d
f
o
n
l
i
n
e
.
c
o
m
In fact, we leave part of this as a useful exercise for the student reader.
Start with the initial-value problem ay + by + cy = f (t), y(0) = y
0
, y(0) = y
0

and set Y(s) = {y (t)} (s) and F(s) = {f (t)} (s). Using the reductions given
by Equations 6 and 7, we obtain
Y s ( ) =
(as b)y
0
as
2
bs c

ay
/
0
as
2
bs c

F s ( )
as
2
bs c
:
Does the denominator look familiar?
The denominator IS the characteristic equation of the original differen-
tial equation. Therefore, with regard to asymptotic stability of solutions, the
two approaches give comparable results. In both cases it is a question of
identifying the real roots of a polynomial function, which in one case is
called the characteristic equation, and in the setting of Laplace transforms is
the common denominator of Y(s). This equivalence of the two approaches
for the question of asymptotic stability need not be too surprising, since
as the reader may recall from their first course in differential equations,
the development of the characteristic equation also begins with the
Key Hypothesis.
Properties of the Laplace Transform: Suppose there is a physical
process modeled by a non-homogenous initial-value problem ay + by +
cy = f(t), y(0) = y
0
, y(0) = y
0
, where f(t) is not necessarily explicitly
exponentia. For instance, the right hand could be a periodic forcing function
for a spring system. In that case the right hand side could involve the sine or
cosine functions. Or, suppose that we are modeling an electric circuit where
power is suddenly turned on at time t = c. This application leads to the
definition of the Heaviside functions H
c
(t), which are defined to be zero for
0 _ t < c and unity for t _ c. So, while the Laplace transform can be
discovered from exploring the implications of the Key Hypothesis, it may
have occurred to the reader that it will be useful, and perhaps even necessary,
to investigate possible application of the transform to functions besides real
exponential functions. The improper integral
_

0
e
st
f t ( )dt obviously can exist
for a class of functions that includes, but is not limited to, exponential
functions. One need only require that the growth rate of f(t) be eventually
(asymptotically) dominated by the exponential function e
st
.
In general, then, there arises the question of the full range of applicability
of the transform. Linearity of the transform becomes useful. For example, since
e
ivt
= cos vt + i sin vt, linearity makes it possible to easily calculate the
transforms of the sine and cosine functions. Naturally, since the setting is
initially differential equations, one also should investigate how the transform
behaves with respect to differentiation, both with respect to t before the trans-
form is applied, and with respect to s after transform is applied.
Discovering the Laplace Transform 323
P
R
I
M
U
S

2
0
0
8
.
1
8
:
3
0
9
-
3
2
4
.

d
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
t
a
n
d
f
o
n
l
i
n
e
.
c
o
m
7. CONCLUDING REMARKS
As mentioned at the beginning of this article, the emphasis has been neither
elaboration of technique nor axiomatic verification, for both of these are
regularly taught in courses in differential equations. Instead, it is hoped that
this article has helped the reader toward discovery of initial reasons for
defining the Laplace Transform in the first place. Such basic understanding
can, of course, lead to the pleasure of mathematical understanding. Even if
ones main interest is applications, such basic understanding is the foundation
of control and versatility.
REFERENCES
1. Braun, M. 1993. Differential Equations and Their Applications, 4
th
Ed. New York:
Springer-Verlag.
2. Halmos, P. and V. S. Sunder. 1978. Bounded Integral Operators on L
2
-Spaces.
Berlin and New York: Springer-Verlag.
3. Ngo, V. and S. Ouzomgi. 1992. Teaching the Laplace Transform using diagrams.
College Math. J., 23(4): 309312.
4. Widder, D. V. What is the Laplace Transform? Amer. Math. Monthly. 52: 419425.
BIOGRAPHICAL SKETCHES
Terrance J. Quinn received his PhD in operator theory, at Dalhousie
University, Nova Scotia, 1992. Since then he has held positions in Ireland,
Texas, and Ohio and is presently Professor and Chair of the Department of
Mathematical Sciences at Middle Tennessee State University. He has done
research in various areas of mathematics and has an increasing appreciation
for the critical role pedagogical studies will play in the global community of
mathematical sciences.
Sanjay Rai received his PhD in differential equations at the University of
Arkansas, Fayetteville, Arkansas, 1994. He is presently Instructional Dean of
Science, Engineering, and Mathematics, Rockville Campus, Montgomery
College, Maryland. Previously, he was Chair and Professor of
Mathematics, Jacksonville University in Florida. He has published articles
in differential equations and their applications, has authored scholarly work
on pedagogy in mathematics, and has won several university awards for
excellence in teaching.
324 Quinn and Rai
P
R
I
M
U
S

2
0
0
8
.
1
8
:
3
0
9
-
3
2
4
.

d
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
t
a
n
d
f
o
n
l
i
n
e
.
c
o
m