Laplacenotes

SECTION 6.
1 317
CHAPTER 6 LAPLACE TRANSFORMS
The Laplace transform is one of many integral transforms in applied mathematics.

Through an improper integral, the Laplace transform creates an association between
a class of functions denoted by f (t) and a class of functions denoted by F (s).
The advantage of this association as far as our discussions are concerned is that
solving a differential equation for f (t) is replaced by solving an algebraic equation
for F (s). The fact that the Laplace transform is a linear operator (in the sense of
equation 4.12) makes it particularly useful for solving the linear differential equations
encountered in Chapters 4 and 5. Furthermore, you will recall that in Chapter 4
we assumed continuity of nonhomogeneous terms in linear differential equations.
This was a matter of convenience rather than necessity. In Exercises 32 and 33
of Section 4.5, we hinted at the awkwardness of incorporating discontinuities into
the techniques of Chapter 4. But discontinuous nonhomogeneities occur frequently
in applications. For instance, in Example 3.11 of Section 3.4, the nonhomogeneity
1/10 in the linear differential equation
dS 5S 1
+ 6 =
dt 10 10
is a result of brine with concentration 2 kilograms per 100 litres being added to the
tank. But suppose after 10 minutes the concentration is doubled to 4 kilograms per
100 litres, and after another 10 minutes it is increased to 5 kilograms per 100 litres.
The nonhomogenity would now be

 1/10, 0 < t < 600
f (t) = 1/5, 600 < t < 1200

1/4, t > 1200.
It is possible to solve this problem with techniques from Chapter 4, but it isn’t very
convenient to do so. An even more awkward situation for Chapter 4 is an LCR-
circuit where the applied voltage is continually turned on for a second, turned off for
another second, turned on again, turned off again, ad infinitum. Even more devas-
tating for Chapter 4 is imparting an instantaneous force to a vibrating mass-spring
system say by striking the mass with a hammer. Chapter 4 just cannot handle this
situation, but Laplace transforms can. These are prominent advantages of Laplace
transforms over the methods of Chapter 4. Discontinuous rates in mixing problems,
discontinuous forcing functions in vibrating mass-spring systems, and discontinuous
driving voltages in LCR-circuits are easily handled by Laplace transforms. Further-
more, nonhomogeneities that represent “point” concepts in space or time cannot be
handled with the techniques of Chapter 4, but they present no problem for Laplace
transforms. Such occurrences include bulk additions of ingredients in mixing prob-
lems, instantaneously applied forces in vibration problems, voltage spikes in electric
circuits, and point loads on beams.
We make one last point before defining the Laplace transform of a function.
Currents in an electrical network satisfy a system of interrelated differential equa-
tions, a topic that we take up in Chapter 7. Laplace transforms of these equations
reduce them to a system of algebraic equations. The functions in these equations
can determine whether the network is stable or whether it leads to unacceptably
high currents, and this can be done without actually solving for the currents.
318 SECTION 6.1
§6.1 The Laplace Transform and its Inverse
The Laplace transform is one of many useful integral transforms in mathematics.

They all map a function of one variable to a function of a different variable. In
general, an integral transform maps a function f (t) of t to a function F (s) of s
according to the definition
Z b
F (s) = K(t, s)f (t) dt. (6.1)
a
The function K(t, s) is called the kernel of the transformation (and it is given).
The function F (s) is called the transform of f (t), and to find F (s), we must multiply
f (t) by K(t, s) and integrate the product from a to b. According to the following
definition, the kernel of the Laplace transform is K(t, s) = e−st , and limits are a = 0
and b = ∞.
Definition 6.1 When f is a function of t, its Laplace transform denoted by F = L{f } is a
function with values defined by
Z ∞
F (s) = L{f }(s) = e−st f (t) dt, (6.2)
0
provided the improper integral converges.

The Laplace transform of a function f (t) never exists for all values of s; there is
always a restriction on the values that can be used. In other words, the real problem
is to find values of s for which improper integral 6.2 converges, and for these values,
and only these values, F (s) is the Laplace transform of f (t).
It is customary to choose t as the independent variable of functions that are to
be transformed because the transform is so often associated with problems in which
t represents time. Do not get the impression, however, that Laplace transforms are
only associated with time. In Section 6.6, we take Laplace transform with respect
to x in order to calculate deflections of beams under various loads. It is customary
to use a lower case letter f to represent a function that is to be transformed, and
its capital counterpart F to represent the transform of f .
When t is time, with units of seconds, then s must have units of one divided
by seconds. If this were not the case, then −st in e−st would not be dimensionless,
and what then would be the units of the exponential. Hence, the Laplace transform
is a mapping from the time domain to the frequency domain.
The improper integral in Definition 6.1 is defined as
Z ∞ Z T
−st
F (s) = e f (t) dt = lim e−st f (t) dt,
0 T →∞ 0
provided the limit exists (and also provided f (t) has acceptable discontinuities,
further discussion coming). If g(s, t) is an antiderivative of e−st f (t) with respect to
t, we could write
T
F (s) = lim {g(s, t)}0 = lim [g(s, T ) − g(s, 0)] = lim g(s, T ) − g(s, 0).
T →∞ T →∞ T →∞
To shorten the notation, we customarily write

SECTION 6.1 319
Z ∞
∞
F (s) = e−st f (t) dt = {g(s, t)}0 ,
0
understanding that the limit should be taken as t → ∞. Appendix E contains a

brief discussion of improper integrals of this type for readers who are meeting them
for the first time, and for readers who would like a quick review.
For our purposes, s is a real variable, in which case F is a real-valued function of
a real variable s. The reader should be aware, however, that in advanced applications
of Laplace transforms, especially for solving partial differential equations and in
many areas of electrical engineering, s is complex, in which case F (s) is a complex-
valued function of a complex variable.
We should determine properties of functions that guarantee existence of their
Laplace transforms. The following three examples point us in the correct direction.
Example 6.1 Find the Laplace transform of f (t) = eat where a 6= 0 is a constant.
Solution According to equation 6.2, the Laplace transform has values defined by
Z ∞ Z ∞ ∞
−st at (a−s)t 1 (a−s)t 1 h i
F (s) = e e dt = e dt = e = lim e(a−s)t − 1 .
0 0 a−s 0 a − s t→∞
This limit exists, and has value 0, only when s > a. Hence, the Laplace transform
of f (t) = eat is 1/(s − a), but only for s > a. We write
1
F (s) = , s > a.•
s−a
Example 6.2 Find the Laplace transform of f (t) = t.
Solution According to equation 6.2, the Laplace transform has values defined by
Z ∞
F (s) = t e−st dt.
0
Integration by parts leads to

∞
t −st 1 −st t −st 1 −st 1
F (s) = − e − 2e = lim − e − 2e + 2.
s s 0
t→∞ s s s
This limit exists, and has value 0, only when s > 0. In other words, the Laplace
transform of f (t) = t is F (s) = 1/s2 , but the function is only defined for s > 0.•
2
2t , 0 ≤ t ≤ 1
Example 6.3 Find the Laplace transform of the discontinuous function f (t) =
1, t > 1.
It is shown in Figure 6.1.
Solution According to equation 6.2,
the Laplace transform has values
defined by
Z ∞ 2
F (s) = e−st f (t) dt
0
Z 1 Z ∞ 1
2 −st
= 2t e dt + e−st dt.
0 1
Two integrations by parts on the first 1 t
integral lead to Figure 6.1
320 SECTION 6.1
2 1 −st ∞
t 2t 2 −st −e 1 4 4 4
F (s) = 2 − − 2 − 3 e + =− + 2 + 3 e−s + 3 ,
s s s 0 s 1 s s s s
provided s > 0 •
What have we learned from Definition 6.1 and these three examples? First,
when f (t) is discontinuous, we subdivide the interval 0 < t < ∞ into subintervals
in which f (t) is continuous. To avoid an infinite number of such subintervals, we
could demand that f (t) have a finite number of discontinuities. It turns out that
this is not entirely necessary, although
it is often the case. Instead, we demand
that f (t) have a finite number of dis-
continuities on every interval 0 ≤ t ≤ T
of finite length. This would allow a
periodic function like that in Figure 6.2
into our discussions. Such a function 1 2 3 t
could represent an applied voltage in an Figure 6.2
electrical circuit that was periodically
turned off. In addition, to guarantee existence of the integral of e−st f (t) on each
subinterval in which f (t) is continuous, we demand that right- and left-hand limits
of f (t) exist at every discontinuity. When a function has a finite number of discon-
tinuities on an interval and right- and left-hand limits exist at all discontinuities in
the interval, the function is said to be piecewise-continuous on that interval. We
shall assume therefore that f (t) is piecewise continuous on every interval 0 ≤ t ≤ T
of finite length.
The second thing that we saw in Examples 6.1–6.3 is that there is always a
restriction on values of s. The function F (s) is not defined for all s; it is defined
only for s larger than some number (a in Example 6.1 and 0 in Examples 6.2 and
6.3). This is due to the fact that for improper integral 6.2 to converge, the integrand
must approach 0 as t → ∞, and must do so sufficiently quickly. This means that
f (t) must not increase so rapidly that it cannot be suppressed by e−st for some
value of s. A sufficient restriction on the growth of f (t) for large t is contained in
the following definition.
Definition 6.2 A function f (t) is said to be of exponential order α, written O(eαt ), if there exist
positive constants T and M such that |f (t)| < M eαt for all t > T .
What this says algebraically is that for sufficiently large t (t > T ), |f (t)| must
grow no faster than a constant M times eαt . Geometrically, the graph of |f (t)| must
be below that of M eαt for t > T . It is important to realize that the exponential
order of a function f (t), if it has one, is concerned with function behaviour for large
t, not for small t. The absolute value |f (t)| must eventually be less than M eαt , and
stay less, but it need not be so for all t. This is shown in Figure 6.3. For example,
the exponential function e4t is O(e4t ) since M can be chosen as 2 and T as zero.
Constant functions are of exponential order zero. The trigonometric functions sin at
and cos at are O(e0t ) since both are less than 2 = 2e0t for all t. The exponential
order of tn is discussed in the following example.
SECTION 6.1 321
(n /e)ne - n
M e at
f (t)
T t n/e t
Figure 6.3 Figure 6.4

Example 6.4 Show that the function tn , where n is a positive integer, is O(et ) for arbitrarily
small, positive .
Solution Consider the function f (t) = tn e−t for arbitrary > 0. To draw its
graph we first calculate that
f 0 (t) = ntn−1 e−t − tn e−t = tn−1 e−t (n − t).
There is a relative maximum at t = n/ and when this is combined with the fact
that limt→∞ tn e−t = 0, the graph in Figure 6.4 results. It shows that the function
tn e−t is bounded by M = (n/)n e−n for all t ≥ 0. In other words, tn e−t < 2M
for all t > 0; that is, tn < 2M et for t > 0, and tn is O(et ).•
If a function f (t) is of exponential order α1 , then it is of exponential order for
any α ≥ α1 . It is customary to choose the smallest value of α when stating that
f (t) is of exponential order α.
We now show that piecewise-continuous functions of exponential order always
have Laplace transforms.
Theorem 6.1 If f (t) is piecewise-continuous on every finite interval 0 ≤ t ≤ T , and is of exponen-
tial order α, then its Laplace transform exists for s > α.
Proof The improper integral in equation 6.2 can be divided into integrals over
the intervals 0 ≤ t ≤ T and T ≤ t < ∞, for any T ,
Z ∞ Z T Z ∞
F (s) = e−st f (t) dt = e−st f (t) dt + e−st f (t) dt.
0 0 T
Since f (t) is piecewise-continuous on 0 ≤ t ≤ T , there is no question that the first

of these integrals exists, and does so for all values of s. Furthermore, since f (t) is
O(eαt ), there exist constants M and T such that |f (t)| < M eαt for t > T . Hence,
Z ∞ Z ∞ Z ∞ Z ∞
e−st f (t) dt ≤ e−st |f (t)| dt < M e−st eαt dt = M e(α−s)t dt
T T T T
∞
M (α−s)t M (α−s)T
= e = e ,
α−s T s−α
provided s > α. In other words, the improper integral over the interval T ≤ t ≤ ∞
converges when s > α. Thus, the Laplace transform of f (t) is defined for s > α.
Theorem 6.1 provides sufficient conditions for existence of Laplace transforms.
Functions that are not piecewise continuous or not of exponential
√ order may or
may not have transforms. For example, the function f (t) = 1/ t is not piecewise
322 SECTION 6.1
continuous due to the infinite discontinuity at t = 0. It does, however, have a

Laplace transform (see Exercise 35).
In calculating Laplace transforms of known functions by means of Definition
6.1, it is not necessary to determine whether the function is of exponential order
prior to use of the integral; evaluation of the integral will yield the interval on which
the transform is defined. When using techniques other than the defining integral
to find Laplace transforms, however, it may be necessary to know that the function
is of exponential order and piecewise-continuous on every finite interval. We shall
develop other techniques in the next section. In this section we concentrate on the
integral definition for the transform.
Example 6.5 Find the Laplace transform for f (t) = tn , where n is a positive integer.
Solution Integration by parts gives
Z ∞ n −st ∞ Z ∞ Z
t e n n ∞ n−1 −st
F (s) = tn e−st dt = − − tn−1 e−st dt = t e dt,
0 −s 0 0 s s 0
provided s > 0. A second integration by parts yields
Z ∞ Z
n ∞ n−1 −st n tn−1 e−st n ∞ n − 1 n−2 −st
F (s) = t e dt = − − t e dt
s 0 s −s 0 s 0 s
Z
n(n − 1) ∞ n−2 −st
= t e dt.
s2 0
Further intergations by parts lead to

Z ∞ ∞
n(n − 1)(n − 2) · · · (1) −st n! e−st n!
F (s) = e dt = n = ,
sn 0 s −s 0 sn+1
provided again that s > 0. This is consistent with Theorem 6.1 and Example
6.4. According to Example 6.4, tn is O(et ) for arbitrarily small, positive , and
therefore its Laplace transform should exist for s > for arbitrarily small > 0.
This is tantamount to s > 0.•
Example 6.6 Find the Laplace transform for f (t) = cos at, where a > 0 is a constant.
Solution In Exercise 33, you are asked to perform two integrations by parts on
the integral
Z ∞
e−st cos at dt
0
in order to find the Laplace transform. We provide an alternative using complex

exponentials, and we do so for two reasons. First, the use of complex exponentials
provides a much easier derivation; in particular, it avoids integration by parts.
Secondly, complex numbers can facilitate many calculations in this chapter, and the
sooner you are exposed to them the better. Instead of evaluating the above integral,
we consider the integral
Z ∞
e−st eati dt.
0
We have simply replaced cos at with eati , remembering that by Euler’s identity,
eati = cos at + i sin at. Now,
SECTION 6.1 323
Z ∞ Z ∞ ∞
−st ati (−s+ai)t e(−s+ai)t
e e dt = e dt = .
0 0 −s + ai 0
The limit of the antiderivative as t → ∞ is

−st ati −st
e e e (cos at + i sin at)
lim = lim = 0,
t→∞ −s + ai t→∞ −s + ai
provided s > 0. Thus,
Z ∞
1
e−st eati dt = .
0 s − ai
We now display real and imaginary parts of both sides of this equation,
Z ∞
1 s + ai s + ai
e−st (cos at + i sin at) dt == = 2 .
0 s − ai s + ai s + a2
When we take real and imaginary parts,
Z ∞ Z ∞
s a
e−st cos at dt = 2 , e−st sin at dt = .
0 s + a2 0 s2 + a2
In other words,
s a
L{cos at} = , L{sin at} = .
s2 + a2 s2 + a2
Thus, by using complex exponentials, we not only found the Laplace transform of
cos at, we also found the transform of sin at, all without integration by parts.•
The following table contains Laplace transforms of functions that occur very
frequently in differential equations. They can be verified with equation 6.2.
f (t) F (s) f (t) F (s)

n! 1
tn eat
sn+1 s−a
a s
sin at cos at
s2 + a2 s2 + a2
2as s2 − a2
t sin at t cos at
(s + a2 )2
2 (s2 + a2 )2
2a3 2as2
sin at − at cos at at cos at + sin at
(s + a2 )2
2 (s + a2 )2
2
a s
sinh at cosh at
s − a2
2 s − a2
2
Table 6.1
We have included transforms of the hyperbolic sine and cosine functions, but
we make no use of them in this chapter. We do this for the sake of those who have
not studied hyperbolic functions. Those familiar with these functions will be able
to provide simpler solutions to some of the examples and exercises.
324 SECTION 6.1
The Inverse Laplace Transform

Definition 6.3 When F is the Laplace transform of f , we call f the inverse Laplace transform of
F , and write
f = L−1 {F }. (6.3)
For instance, Table 6.1 yields

−1 1 s √
L = e−2t and L −1
2
= cos 3t.
s+2 s +3
The Laplace transform F (s) of a function f (t) is unique, every function has
exactly one Laplace transform. On the other hand, many functions have the same
transform. For example, the functions
( 0, t = 1
f (t) = t2 and g(t) = t2 , t 6= 1, 2
0, t = 2,
which are identical except for their values at t = 1 and t = 2 both have the same
transform 2/s3 . The fact that F (s) = 2/s3 follows from Table 6.1; G(s) = 2/s3
follows from integration, (or by noting that because f (t) and g(t) differ only at
isolated points, this makes no difference to integral 6.2). What we are saying is that
the inverse transform f = L−1 {F } in Definition 6.3 is not an inverse in the true
sense of inverse; there are many possibilities for f for given F . In advanced work, a
formula for calculating inverse transforms is derived, and this formula always yields a
continuous function f (t), when this is possible. In the event that this is not possible,
the formula gives a piecewise-continuous function whose value is the average of
right- and left-limits at discontinuities, namely lim→0 [f (t + ) + f (t − )]/2. The
importance of this formula is that it defines f = L−1 {F } in a unique way. Other
functions which have the same transform F differ from f only in their values at
isolated points; they cannot differ from f over an entire interval a ≤ t ≤ b. When
f is a continuous function with transform F , there cannot be another continuous
function with the same transform. With this in mind, we adopt the procedure of
choosing a continuous function L−1 {F } for given F whenever this is possible. In the
confines of differential equations, this is always possible since solutions of differential
equations are always continuous functions. Thus, when solving differential equations
by Laplace transforms, there will always be only one inverse transform for a given
function F (s). In other words, we can talk about the inverse Laplace transform of
F (s).
According to the following theorem, the Laplace transform and its inverse are
linear operators in the sense of equation 4.12. That the transform is linear is a
direct result of the fact that integration is a linear operation; once the transform is
linear, so also is the inverse.
Theorem 6.2 The Laplace transform and its inverse are linear operators; that is, for arbitrary
functions f and g, arbitrary transforms F and G, and an arbitrary constant c,
L{f + g} = L{f } + L{g}, L{c f } = c [L{f }], (6.4a)
−1 −1 −1 −1 −1
L {F + G} = L {F } + L {G}, L {c F } = c [L {F }]. (6.4b)
For instance, using linearity and Table 6.1,
SECTION 6.1 325

2 4
L{2e−t + 3 sin 4t} = 2L{e−t } + 3L{sin 4t} = +3 ,
s+1 s2 + 16
and

−1 2 4s −1 1 −1 s t3 √
L 4
− 2 = 2L − 4L =2 − 4 cos 5t.
s s +5 s4 2
s +5 6
The following result can serve as a partial check on calculations of Laplace
transforms.
Theorem 6.3 If f (t) is is piecewise-continuous on every finite interval 0 ≤ t ≤ T , and is of
exponential order α, its Laplace transform has limit zero as s → ∞; that is,
lim F (s) = 0. (6.5)

s→∞
Proof: The definition of F (s) gives

Z ∞ Z ∞ Z T Z ∞
|F (s)| = e−st f (t) dt ≤ e−st |f (t)| dt = e−st |f (t)|dt + e−st |f (t)|dt.
0 0 0 T
Since f (t) is piecewise continuous on 0 ≤ t ≤ T , it is bounded thereon, and there

exists a number M such that |f (t)| < M in this interval. Furthermore, since f (t) is
of exponential order α, there exist constants T > 0 and M > 0, such that for t > T ,
|f (t)| < M eαt . We can therefore write that
Z T Z ∞ −st T (α−s)t ∞
−st −st αt e e
|F (s)| < Me dt + e M e dt = M +M
0 T −s 0 α−s T
M M e(α−s)T
= (1 − e−sT ) + ,
s s−α
provided that s > 0 and s > α. The limit of this is zero as s → ∞.

We said that this theorem could serve as a check on calculations. For instance,
if we calculated the transform of a function f (t) (piecewise continuous on every
finite interval, and of exponential order) to be
s2 + 2s − 5
F (s) = ,
3s2 + 10s + 15
we would know that we had made an error since the limit of this function as s → ∞
is not equal to zero; it is equal to 1/3. In Section 6.5, we will encounter a function
that is an exception to this rule, but it will not be a function that is piecewise
continuous. Another check on the validity of a Laplace transform can be found in
Exercise 53 of Section 6.3.
We now give a preview of what is to come. Consider solving the initial-value
problem
d2 y dy
2
−4 − 5y = 6 − 5t, y(0) = 1, y 0 (0) = −1. (6.6)
dt dt
We can certainly solve this with the techniques of Chapter 4. Use the auxiliary
equation to find a general solution yh (t) of the associated homogeneous equation;
use undetermined coefficients to find a particular solution yp (t) of the differential
326 SECTION 6.1
equation; add these together and evaluate constants with the initial conditions. The
Laplace transform takes an entirely different approach; it reduces the initial-value
problem to an algebraic equation. To do this, we need to know how to take Laplace
transforms of derivatives. We quote two results that will be verified in Section 6.3.
If F (s) is the Laplace transform of f (t), then Laplace transforms of f 0 (t) and f 00 (t)
can be written in terms of F (s) as follows:
L{f 0 (t)} = sF (s) − f (0), (6.7a)

00 2 0
L{f (t)} = s F (s) − sf (0) − f (0). (6.7b)
If we take Laplace transforms of both sides of differential equation 6.6, and use the
fact that the transform is a linear operator, we obtain
L{y 00 } − 4L{y 0 } − 5L{y} = 6L{1} − 5L{t}.
When we denote the Laplace transform of y(t) by Y (s), and use formulas 6.7 on the
left, we get
6 5
[s2 Y (s) − s(1) − (−1)] − 4[sY (s) − 1] − 5Y (s) = − 2.
s s
This is an algebraic equation for Y (s) which is easily solved,
6 5
− 2 +s−5
Y (s) = s s .
s2 − 4s − 5
When this rational expression is simplified and written in its partial fraction de-
composition, the result is
17/6 1/6 1 2
Y (s) = + + 2− .
s+1 s−5 s s
We can take the inverse Laplace transform of each of these terms to find the solution
of the initial-value problem,
17 −t 1 5t
y(t) = e + e + t − 2.
6 6
This example is typical of Laplace transforms at work on initial-value problems
like those in Chapters 4 and 5. The transform reduces the differential equation
in y(t) to an algebraic equation in its transform Y (s). The algebraic equation is
solved for Y (s) and the inverse transform then yields the solution y(t) of the initial-
value problem. In order to solve other initial-value problems, we need to expand
the catalogue of functions with known Laplace transforms beyond those in Table
6.1. Furthermore, in Chapter 4 we assumed continuity of nonhomogeneous terms in
linear differential equations. This was a matter of convenience rather than necessity.
However, in Exercises 32 and 33 of Section 4.5, we hinted at the awkwardness of
incorporating discontinuities into the techniques of Chapter 4. We shall give other
examples of discontinuous nonhomogeneities in this chapter, and see how easily they
are handled by Laplace transforms. Section 6.2 concentrates on efficient ways to
calculate transforms and inverse transforms, and Sections 6.3 and 6.4 then return
to full discussions of differential equations.
SECTION 6.1 327
EXERCISES 6.1
In Exercises 1–10 use linearity and Table 6.1 to find the Laplace transform of the
function.
1. f (t) = t3 − 2t2 + 1 2. f (t) = t + et
3. f (t) = 5e4t 4. f (t) = e−2t + 2et
5. f (t) = sin 4t + 3 cos 4t 6. f (t) = cos 2t − 3 sin 4t
7. f (t) = 5t cos 2t 8. f (t) = 3t sin 4t
9. f (t) = 5t cos t − 2t sin t 10. f (t) = 3t sin t − cos t
In Exercises 11–20 use linearity and Table 6.1 to find the inverse Laplace transform of
the function.
7 2 3
11. F (s) = 3 12. F (s) = − 4
s s s
1 4 3
13. F (s) = + 14. F (s) =
s + 5 s2 s−1
s 3 2s 5
15. F (s) = 2 − 2 16. F (s) = 2 − 2
s +4 s +4 s +2 s +9
2s s2
17. F (s) = 2 2
18. F (s) = 2
(s + 2) (s + 9)2
3s − s2 s2 − 2
19. F (s) = 2 20. F (s) =
(s + 4)2 (s2 + 3)2
In Exercises 21–32 use Definition 6.1 to find the Laplace transform of the function.

0, 0 < t < 3 1, 0 < t < 4
21. f (t) = 22. f (t) =
1, t > 3 2, t > 4

t, 0 < t < 2 t2 , 0 < t < 1
23. f (t) = 24. f (t) =
2, t > 2 0, t > 1

0, 0 < t < 1 0, 0<t<1
25. f (t) = 2 26. f (t) = 2
t , t>1 (t − 1) , t > 1
( 0, 0 < t < 1 ( t, 0<t<1
27. f (t) = 1, 1 < t < 2 28. f (t) = 2 − t, 1<t<2
0, t > 2 0, t>2

2t, 0 < t < 1 1 + t2 , 0<t<1
29. f (t) = 30. f (t) =
t, t>1 2t, t>1
( 0, 0 < t < a
0, 0 < t < a
31. f (t) = (a constant) 32. f (t) = 1, a < t < b (a, b constants)
1, t > a
0, t > b
33. Use integration by parts to find Laplace transforms for sin at and cos at.
34. Derive the Laplace transforms for t cos at and t sin at in Table 6.1. Hint: Use complex expo-
nentials.
√
35. The function 1/ t is not piecewise p continuous because of √ its infinite discontinuity at t = 0.
Show that it has Laplace transform π/s. Hint: Set u = t in the definition of the Laplace
√ R∞ 2 √
transform of 1/ t in terms of a definite integral and use the fact that 0 e−u du = π/2.
328 SECTION 6.1
36. Are all bounded functions (functions that satisfy |f (t)| < M for all t > 0) of exponential order?
37. Are all continuous functions of exponential order?
38. Are all polynomial functions of exponential order?
39. In Example 6.6 we used complex exponentials to find the Laplace transforms for cos at and
sin at. In Exercise 33, we used integration by parts. Another method that also works for other
functions is to use Maclaurin series. The Maclaurin series for sin at is
X∞ X∞
(−1)n 2n+1 (−1)n a2n+1 2n+1
sin at = (at) = t .
n=0
(2n + 1)! n=0
(2n + 1)!
Assuming that the operation of taking the Laplace transform can be interchanged with the
summation operation, derive the Laplace transform of sin at.
40. Use the technique of Exercise 39 to derive the Laplace transform of cos at.
41. The Bessel function of the first kind of order zero has Maclaurin series
X∞
(−1)n 2n
J0 (t) = t .
n=0
22n (n!)2
1
Show that its Laplace transform is L{J0 (t)} = √ .
1 + s2
a
sin at
42. Verify that L = Tan−1 .
t s
bt
e − eat s−a
43. Verify that L = ln when s > b > a.
t s−b
44. In Example 6.5, we developed the formula for the Laplace transform of tn when n is a positive
integer. In this exercise, we extend the result to the case that n is positive, but not an integer.
First, we need to extend the definition of factorials by defining the gamma function,
Z ∞
Γ(t) = e−u ut−1 du, t > 0.
0
This function is discussed in more detail in Appendix A, but for our present purposes, only the
property in part (a) is required.
(a) Verify that the gamma function satisfies the recursive formula
Γ(t + 1) = t Γ(t), and hence that Γ(n + 1) = n! when n is positive integer.
(b) Now prove that for r > 0,
Γ(r + 1)
L{tr } = .
sr+1
2
45. (a) Prove that the function et is not of exponential order.
2
46. (a) Prove that the function f (t) = sin (et ) is of exponential order.
(b) Prove that the derivative f 0 (t) is not of exponential order? In spite of this, f 0 (t) has a
Laplace transform. (See Exercise 55 in Section 6.3.)
SECTION 6.2 329
6.2 Algebraic Properties of the Laplace Transform and its Inverse

Early in your calculus studies you were required to use the limit-definition of the
derivative to differentiate various functions. You were quickly brought to the realiza-
tion that rules could be developed so that use of the definition could be eliminated,
and you certainly appreciated these rules (power, product, and quotient to name a
few). The same was true when it came to the definition of the definite integral as
the limit of a summation. It is impossible to use the definition to find the definite
integral of all but a handful of functions, and therefore using anti-derivatives to
calculate definite integrals is essential. Likewise, seldom is it necessary to evaluate
the improper integral in Definition 6.1 to find the Laplace transform for a function;
other techniques prove more efficient. The purpose of this section is to develop some
of these shortcuts. In addition, recall that our intention is to use Laplace transforms
to provide another method for solving linear differential equations and extensions
which are difficult or impossible to solve with the techniques of Chapter 4. With
this in mind, note how many of the algebraic properties of the Laplace transform
uncovered in this section are directed toward the functions so prevalent in solving
linear differential equations, namely, tn , eat , sin at, cos at, and sums and products of
these functions. In Section 6.3, we derive formulas 6.7 for taking Laplace transforms
of derivatives of functions. With these formulas and the results of this section, we
will be well prepared to solve differential equations.
One of two shifting properties is contained in the following theorem.
Theorem 6.4 When F is the Laplace transform of f ,
L{eat f (t)} = F (s − a), (6.8a)

−1 at
L {F (s − a)} = e f (t). (6.8b)
Proof By Definition 6.1,

Z ∞ Z ∞
at at −st
L{e f (t)} = e e f (t) dt = e−(s−a)t f (t) dt.
0 0
But this is equation 6.2 with s replaced by s − a; that is,

L{eat f (t)} = F (s − a).
Equation 6.8b is equation 6.8a written in terms of inverse transforms rather than
transforms.
The notation in equation 6.8 is not quite as described earlier. The Laplace
transform and its inverse operate on functions, not on function values as is suggested
by 6.8. These equations would be more properly stated in the form
L{eat f }(s) = F (s − a), (6.9a)

−1 at
L {F (s − a)}(t) = e f (t). (6.9b)
We feel that the shifting property is more clearly conveyed for most readers by
6.8a,b, and we apologize to readers who are offended by the notation. It may
be convenient to repeat this practice in describing other properties of the Laplace
transform, but we shall attempt to minimize its use.
330 SECTION 6.2
When calculating transforms and inverse transforms, we often write properties

6.8 in the form
L{eat f (t)} = L{f (t)}|s→s−a , (6.10a)

−1 at −1
L {F (s − a)} = e L {F (s)}. (6.10b)
On the right of property 6.10a, L{f (t)} is a function of s. The subscript |s → s − a

means that each s is to be replaced by s − a.
Equation 6.8a, or its alternatives, states that multiplication by an exponential
eat in the t-domain is equivalent to a translation or shift by a in the s-domain. It
provides a quick way to find the Laplace transform of any function f (t) multiplied
by an exponential, provided the Laplace transform of f (t) is known. For example,
property 6.10a implies that

2 −5t 2 2 2
L t e = L t |s→s+5 = 3
= .
s |s→s+5 (s + 5)3
Property 6.10b yields

−1 1 6t −1 1 6t t4
L =e L =e .
(s − 6)5 s5 4!
Example 6.7 In Exercise 34 of Section 6.1, we found Laplace transforms for t cos at and t sin at, by
using Euler’s identity and integration. Property 6.8a is superior if we again replace
trigonometric functions with complex exponentials. This is an ongoing theme; al-
ways consider replacing sines and cosines with complex exponentials. We can write
that

ati 1 1
L{te } = L{t}|s→s−ai = = .
s |s→s−ai (s − ai)2
2
We now display the real and imaginary parts of both sides of the equation,
(s + ai)2 (s2 − a2 ) + 2asi
L{t(cos at + i sin at)} = = .
(s − ai)2 (s + ai)2 (s2 + a2 )2
When we take real and imaginary parts,
s2 − a2 2as
L{t cos at} = , L{t sin at} = .•
(s2 + a2 )2 (s2 + a2 )2
Property 6.10b is particularly useful in finding inverse Laplace transforms of

rational functions of s that contain irreducible quadratic factors in denominators.
We encounter them constantly. Here is an example.
Example 6.8 Find the inverse Laplace transform for F (s) = (s + 1)/(s2 − 6s + 14).
Solution First, by completing the square on the quadratic, we can express F (s)
in the form
s+1 (s − 3) + 4
F (s) = 2
= .
(s − 3) + 5 (s − 3)2 + 5
We can now use property 6.10b to find the inverse transform,
SECTION 6.2 331

s+4 √ 4 √
f (t) = e3t L−1 = e3t
cos 5t + √ sin 5t .•
s2 + 5 5
In physical applications, we often encounter quantities that are turned on and

off, or quantities that change abruptly. For example, in mixing problems, such
as Example 3.11 of Section 3.4, the concentration of salt added to the tank could
suddenly be changed at any time; the applied voltage in an LCR-circuit could
be turned on and off any number of times, and the forcing function in a mass-
spring system could be turned on or off, or sharply changed. Such functions are
conveniently described by Heaviside unit step functions introduced in Section
5.5. The fundamental unit step function is

0, t < 0
h(t) = (6.11)
1, t ≥ 0.
(Some authors replace t ≥ 0 in this definition with t > 0 so that the function is
undefined at t = 0. The rest of this chapter can be developed with either convention
with minor adjustments in results.) A graph of this function is shown in Figure 6.5;
there is a discontinuity of magnitude unity at t = 0, hence the name unit step
function.
1 1
a
t t
When the discontinuity occurs at t = a, the function is denoted by

0, t < a
h(t − a) = (6.12)
1, t ≥ a.
Its graph is shown in Figure 6.6. This notation is consistent with that in elementary
calculus where replacing the variable t in a function f (t) by t−a translates the graph
of the function a units to the right.
Heaviside unit step functions provide
compact representations for functions
whose descriptions vary from one interval
to another; such functions may have, or
a b t
may not have, discontinuities at points
that separate these intervals. Such a Figure 6.7
function is shown in Figure 6.7. It is
defined differently on the intervals 0 < t ≤ a, a < t < b, and t > b. It is continuous
at t = a, but not at t = b.
An important function in our discussions is shown in Figure 6.8. It is called a
pulse function. It can be expressed algebraically in the form h(t − a) − h(t − b),
except at t = a and t = b. In the event that the height of the nonzero portion is
332 SECTION 6.2
c rather than unity (Figure 6.9), we obtain c[h(t − a) − h(t − b)], again except at
t = a and t = b.
1
c
a b t a b t
Pulse functions can be combined algebraically to describe other step functions
(besides h(t − a)), such as that in Figure 6.10. It is the sum of two pulse functions,
4[h(t) − h(t − 3)] + 2[h(t − 3) − h(t − 6)] = 4h(t) − 2h(t − 3) − 2h(t − 6),
except at t = 0, 3, and 6. The step function in Figure 6.11 is the sum of three pulses,
3[h(t − a) − h(t − b)] + 4[h(t − b) − h(t − c)] + h(t − c)
= 3h(t − a) + h(t − b) − 3h(t − c),
except at x = a, b, and c. In future representations of piecewise defined functions
in terms of Heaviside functions, we will omit mentioning the exceptions.
4 4
2 2
3 6 t a b c t

A convenient representation for the function in Figure 6.12 is t2 [h(t) − h(t − a)],
and for the function in Figure 6.13, [2 − (t − a)/(b − a)][h(t − a) − h(t − b)].
a2 2
Parabola
1
a a b
t t
What these examples illustrate is that to “turn a function on” for t ≥ a,
multiply it by h(t − a). It will be zero for t < a. To turn it on between t = a and
t = b, multiply it by h(t − a) − h(t − b). It will be zero for t < a and t ≥ b. The
parabola in Figure 6.14 has equation a2 +(t−a)2 for t > a. To turn it on, we multiply
by h(t − a); that is, the function can be expressed in the form [a2 + (t − a)2 ]h(t − a).
For the function in Figure 6.15, we turn on the straight line y = a − a(t − a)/(b − a)
for a < t < b, and then the horizontal line y = c for t > b,
[a − a(t − a)/(b − a)][h(t − a) − h(t − b)] + c h(t − b).
SECTION 6.2 333
2 a2
a
a2
a 2a a b
t t

You may have noticed that in Figures 6.8–6.15, functions do not have values
at discontinuities. Because of this, representations of these functions in terms of
Heaviside functions are not valid at discontinuities. If a function has a value at a
discontinuity, its representation in terms of Heaviside functions may or may not be
valid at the discontinuity. For instance, the Heaviside representation of the functions
in Figures 6.16a,b is

b−c
f (t) = c + (t − a) h(t − a).
a
It is valid at the discontinuity t = a in Figure 6.16a, but not in 6.16b. None of
this really matters when it comes to Laplace transforms. Because the transform
of a function is defined as a definite integral, the transform is the same whether
the function has a value at a discontinuity or not. We will therefore continue the
practice of leaving a function undefined at discontinuities (when the purpose is to
take the transform of the function).
b (2 a, b) b (2 a, b)
c c
a 2a t a 2a t
Figure 6.16a Figure 6.16b
The Laplace transform of the Heaviside unit step function is
Z ∞ Z ∞ −st ∞
−st −st e e−as
L{h(t − a)} = e h(t − a) dt = e dt = = , (6.13)
0 a −s a s
provided s > 0.
In Section 6.1, we used integration to find the Laplace transform of piece-
wise defined functions. In the above discussions, we represented such functions as
products of functions multiplied by Heaviside functions, and we did so in order to
circumvent integrations. The following theorem enables us to do this.
Theorem 6.5 When f (t) has a Laplace transform,
L{f (t)h(t − a)} = e−as L{f (t + a)}. (6.14a)
Proof According to Definition 6.1,

Z ∞ Z ∞
L{f (t)h(t − a)} = e−st f (t)h(t − a) dt = e−st f (t) dt.
0 a
334 SECTION 6.2
If we change variables of integration with u = t − a, then

Z ∞
L{f (t)h(t − a)} = e−s(u+a) f (u + a) du
0
Z ∞
−as
=e e−su f (u + a) du = e−as L{f (t + a)}.
0
We illustrate how to use this result in the following examples.

0, 0≤t≤2
Example 6.9 Find the Laplace transform for the function f (t) = 2 shown
(t − 2) , t > 2,
in Figure 6.17.
Solution Since f (t) can be expressed in the form f (t) = (t−2)2 h(t−2), equation
6.14a gives
2e−2s
F (s) = L{(t − 2)2 h(t − 2)} = e−2s L{t2 } = .•
s3
1 2 3 t
-1
1 2 3 t

0, 0≤t<2
Example 6.10 Find the Laplace transform for the function f (t) = shown in
t − 3, t > 2,
Figure 6.18.
Solution Since f (t) can be expressed in the form f (t) = (t−3)h(t−2), its Laplace
transform is

−2s −2s −2s 1 1
F (s) = L{(t − 3)h(t − 2)} = e L{(t + 2) − 3} = e L{t − 1} = e − .•
s2 s
Example 6.11 Find the Laplace transform for the function in Figure 6.19.
Solution The function is con-
tinuous, but because it is defined 3
differently on the intervals 0 ≤ t ≤ 1,
1 < t ≤ 2, and t > 2, it can be rep-
resented efficiently in terms of Heavi-
side functions,
1 2 t
f (t) = 3(t − 1)[h(t − 1) − h(t − 2)] + 3h(t − 2) Figure 6.19
= 3(t − 1) h(t − 1) + (6 − 3t) h(t − 2).
We can now use equation 6.14a to find its transform,

−s −2s 3e−s −2s 3
F (s) = e L{3t} + e L{6 − 3(t + 2)} = 2 − e .•
s s2
SECTION 6.2 335
Example 6.12 Find the Laplace transform for e−3t sin 2t h(t − 1).
Solution We can find the transform by either attacking the exponential e−3t
first, or the Heaviside function h(t − 1) first. We show both solutions. If we attack
the Heaviside function first, we use property 6.14a,
L{e−3t sin 2t h(t − 1)} = e−s L{e−3(t+1) sin 2(t + 1)}

= e−s−3 L{e−3t sin 2(t + 1)}
= e−(s+3) L{sin 2(t + 1)}|s→s+3 (using equation 6.10a)
= e−(s+3) L{cos 2 sin 2t + sin 2 cos 2t}|s→s+3

−(s+3) (cos 2)2 (sin 2)s
=e + 2
s2 + 4 s + 4 |s→s+3

2 cos 2 (sin 2)(s + 3)
= e−(s+3) + .
(s + 3)2 + 4 (s + 3)2 + 4
If we attack the exponential first, we use property 6.10a,
L{e−3t sin 2t h(t − 1)} = L {sin 2t h(t − 1)}s→s+3

= e−s L{sin 2(t + 1) s→s+3 (using equation 6.14a)
= e−(s+3) [L{(sin 2t cos 2 + cos 2t sin 2}]s→s+3

−(s+3) 2 cos 2 (sin 2)s
=e + 2
s2 + 4 s + 4 s→s+3

2 cos 2 (sin 2)(s + 3)
= e−(s+3) + •
(s + 3)2 + 4 (s + 3)2 + 4
The equivalent of property 6.14a in terms of inverse transforms is equally as

important as 6.14a itself, and from the inverse statement it gets its name the second
shifting property of Laplace transforms. We state it as a corollary to Theorem 6.5.
Corollary 6.5.1 If f = L−1 {F }, then
L−1 {e−as F (s)} = f (t − a)h(t − a) = L−1 {F (s)}|t→t−ah(t − a). (6.14b)
The graph of f (t − a)h(t − a) is that of f (t) (Figure 6.20a) shifted a units to

the right and turned on for t > a (Figure 6.20b).
t a t
Thus, to find the inverse transform of a function in the form e−as F (s), we find the
inverse transform of F (s), translate it a units to the right, and turn it on for t > a.
336 SECTION 6.2

−1 5e−4s
Example 6.13 Find L .
s3
Solution With property 6.14b,
−4s 2
−1 5e −1 1 t
L 3
= 5L 3
h(t − 4) = 5 h(t − 4)
s s |t→t−4 2 |t→t−4
5
= (t − 4)2 h(t − 4).•
2
A graph of (5/2)(t − 4)2 h(t − 4) is shown in Figure 6.21.

1/2
1 2 3 4 5 t
5/2 -1/2
4 5 t
e−s − e−2s
Example 6.14 Find the inverse transform for F (s) = .
s2 + 5
Solution Property 6.14b gives
−s −2s
−1 e −1 e
f (t) = L −L
s2 + 5 s2 + 5

−1 1 −1 1
=L h(t − 1) − L h(t − 2)
s2 + 5 |t→t−1 s2 + 5 |t→t−2

1 √ 1 √
= √ sin 5t h(t − 1) − √ sin 5t h(t − 2)
5 |t→t−1 5 |t→t−2
1 h √ √ i
= √ sin 5(t − 1) h(t − 1) − sin 5(t − 2) h(t − 2) .
5
The function is shown in Figure 6.22. We can write it without the Heaviside func-
tions as follows:

 0, √ √
0≤t≤1
f (t) = (1/ 5) sin 5(t − 1), 1<t≤2
 √ √ √
(1/ 5)[sin 5(t − 1) − sin 5(t − 2)], t > 2.•
Finding inverse transforms is often a matter of finding the partial fraction

decomposition of a rational function, together with the above properties and a set
of tables. We illustrate in the following example. For readers who have never studied
partial fractions, or need a refresher on the topic, we have provided coverage of the
topic in Appendix D.
Example 6.15 Find inverse Laplace transforms for the following functions:
s2 − 9s + 9 1 e−s
(a) F (s) = (b) F (s) = (c) F (s) =
s3 (s2 + 9) s2 (s2 − 4) s2 − s
SECTION 6.2 337
Solution (a) The partial fraction decomposition of F (s) gives

2
−1 s − 9s + 9 −1 1 1 1 t2 1
f (t) = L 3 2
= L 3
− 2
+ 2
= − t + sin 3t.
s (s + 9) s s s +9 2 3
(b) Once again partial fractions give

1 1/16 1/16 1/4 1 2t 1 t
f (t) = L−1 = L−1
− − = e − e−2t − .
s2 (s2 − 4) s−2 s+2 s2 16 16 4
(c) With partial fractions and property 6.14b,
−s
e 1 1
f (t) = L−1 = L−1
e−s
−
s2 − s s−1 s

1 1
= L−1 − h(t − 1)
s − 1 s |t→t−1
= (et − 1)|t→t−1 h(t − 1) = (et−1 − 1)h(t − 1).•
Periodic Functions
The sine and cosine functions are periodic and there was no difficulty in finding
their transforms. The function in Figure 6.23a is also periodic, but it is not obvious
how to find its transform. To use the definition of the transform as an improper
integral requires the addition of an infinite number of definite integrals, one over
each period of the function. Alternatively, we can write the periodic function as the
sum of an infinite number of functions, turned on and off by Heaviside functions,
and add all their transforms. Although both techniques work, neither is necessary.
Two simpler derivations lead to a procedure for finding the Laplace transform of
a periodic function that involves one integration over one period, and a procedure
that involves no integrations at all. Suppose f (t) is an unspecified periodic function
with period p, such as that in Figure 6.23a.
p 2p 3p t p 2p 3p t
We divide the range of integration in Definition 6.1 into two parts
Z p Z ∞
−st
F (s) = e f (t) dt + e−st f (t) dt.
0 p
We substitute u = t − p in the second integral,

Z p Z ∞
−st
F (s) = e f (t) dt + e−s(u+p) f (u + p) du.
0 0
But f (t) has period p, so that f (u + p) = f (u), and therefore

Z p Z ∞ Z p
F (s) = e−st f (t) dt + e−ps e−su f (u) du = e−st f (t) dt + e−ps F (s).
0 0 0
338 SECTION 6.2
We can solve this for

Z p
1
F (s) = e−st f (t) dt. (6.15)
1 − e−ps 0
The is the first of the results that we sought, a method for finding the transform of a
periodic function that involves only integration over one period of the function. The
second method is to not integrate at all. Suppose that f1 (t) denotes the function
that is equal to f (t) over the first period 0 ≤ t ≤ p of the function, and is otherwise
equal to zero (Figure 6.23b). This function is sometimes called the windowed
version of f (t). The Laplace transform of f1 (t) is defined by the definite integral
in equation 6.15. In other words, we can write that
1
F (s) = L{f1 (t)}. (6.16)
1 − e−ps
This is the second result that we were looking for, a method to calculate the Laplace
transform of a periodic function that does not require integration. Formula 6.16 does
this, provided we are willing, and able, to find L{f1 (t)} without integration.
The same results can be obtained in another way. We write the function f1 (t)
in terms of f (t),
f1 (t) = f (t)[h(t) − h(t − p)] = f (t)[1 − h(t − p)].
We now take Laplace transforms, and use property 6.14a,
F1 (s) = F (s) − e−ps L{f (t + p)}.
But f (t + p) = f (t), so that
F1 (s) = F (s) − e−ps L{f (t)} = F (s) − e−ps F (s) = (1 − e−ps )F (s).
When we solve this equation for F (s), we get
1 1
F (s) = −ps
F1 (s) = L{f1 (t)},
1−e 1 − e−ps
equation 6.16.
Example 6.16 Find the Laplace transform for the periodic function in Figure 6.24a.
Solution Since the function has period 2, formula 6.15 yields
Z 2
1
F (s) = (1 − t)e−st dt.
1 − e−2s 0
Integration by parts gives
2
1 (t − 1) −st 1 1 + e−2s 1
F (s) = e + 2 e−st = − 2.
1 − e−2s s s 0
−2s
s(1 − e ) s
Alternatively, formula 6.16 gives
1
F (s) = L{f1 (t)},
1 − e−ps
where f1 (t) is the windowed version of f (t) in Figure 6.24b. Its Laplace transform
is
SECTION 6.2 339
L{(1 − t)[h(t) − h(t − 2)]} = L{(1 − t) h(t)} + L{(t − 1) h(t − 2)}

1 1
= − 2 + e−2s L{t + 1}
s s
1 1 −2s 1 1
= − 2 +e + .
s s s2 s
Hence,

1 1 1 −2s 1 1 1 + e−2s 1
F (s) = −2s
− 2
+ e 2
+ = −2s
− 2 .•
1−e s s s s s(1 − e ) s
1 1
1 2 3 t 1 2 t
-1 -1

Example 6.17 Find the Laplace transform for | sin 2t|.
Solution Since | sin 2t| has period π/2 (see Figure 6.25), formula 6.15 gives
Z π/2
1
L{| sin 2t|} = e−st sin 2t dt,
1 − e−πs/2 0
and we could use integration by parts to evaluate this integral. Alternatively, we
can use formula 6.16,
1
L{| sin 2t|} = L{sin 2t[h(t) − h(t − π/2)]}
1 − e−πs/2
1
= −πs/2
L{sin 2t} − L{sin 2t h(t − π/2)}
1−e

1 2 −πs/2
= − e L{sin 2(t + π/2)}
1 − e−πs/2 s2 + 4

1 2 −πs/2
= +e L{sin 2t}
1 − e−πs/2 s2 + 4

1 2 2e−πs/2
= + 2
1 − e−πs/2 s2 + 4 s +4 1
2(1 + e−πs/2 )
= .•
(s2 + 4)(1 − e−πs/2 )
p p 3p t
2 2
Figure 6.25
We have just seen that Laplace transforms of periodic functions that are not
sinusoids contain factors 1/(1 − e−ps ). When we solve differential equations that
340 SECTION 6.2
have periodic inputs in Section 6.4, we will have to invert transforms with such
factors. The following example illustrates how to do this.
2
Example 6.18 Find the inverse Laplace transform for F (s) = .
s3 (1 − e−2s )
Solution Property 6.14b enables us to find the inverse transform of any function
multiplied by e−as . We can write the above F (s) as a sum of terms in this form if
we expand 1/(1 − e−2s ) in a geometric series
2 2
F (s) = = 1 + e−2s + e−4s + e−6s + · · · .
s3 (1 − e−2s ) s3
We now invert each term,

∞
X
f (t) = t2 + (t − 2)2 h(t − 2) + (t − 4)2 h(t − 4) + · · · = (t − 2n)2 h(t − 2n).•
n=0
Properties of Laplace transforms and their inverses that we have discussed in

this section have been gathered together for quick reference in Table 6.2. Included
also are the transform pairs in Table 6.1. There are also properties that have yet to
be considered, namely lines 12, 17, and 18. These will be developed in subsequent
sections. Notice the arrows in the middle column. A double arrow ↔ indicates that
this line is useful in taking Laplace transforms and their inverses; a right arrow →
indicates that the property is most useful in taking transforms; and a left arrow ←
indicates that the property is most useful in taking inverse transforms.
EXERCISES 6.2
In Exercises 1–12 represent the functions in Exercises 21–32 of Section 6.1 in terms of
Heaviside unit step functions. Find the Laplace transform of each function.
In Exercises 13–20 represent the function algebraically in terms of Heaviside unit step
functions. Find the Laplace transform of each function.
13. 14.
2 2
1 t 1 2 3 t
15. 16.
4
Parabola 1
Parabola
1 t 1 2 t
SECTION 6.2 341
f (t) F (s) = L{f }(s)
n!
tn (n = 0, 1, 2, . . .) ↔
sn+1
1
eat ↔
s−a
a
sin at ↔
s2 + a2
s
cos at ↔
s2 + a2
2as
t sin at ↔
(s + a2 )2
2
s2 − a2
t cos at ↔
(s2 + a2 )2
2a3
sin at − at cos at ↔
(s2 + a2 )2
2as2
sin at + at cos at ↔
(s2 + a2 )2
a
sinh at ↔
s − a2
2
s
cosh at ↔
s − a2
2
e−as
h(t − a) ↔
s
−as
δ(t − a) ↔ e
at
e f (t) ↔ F (s − a)
f (t)h(t − a) → e−as L{f (t + a)}
f (t − a)h(t − a) ← e−as F (s)
Z p
1
p − periodic f (t) → e−st f (t) dt
1 − e−ps 0
Z t
f (u)g(t − u) du ← F (s)G(s)
0
dn F
tn f (t) (n = 1, 2, 3, . . .) ↔ (−1)n
dsn
f 0 (t) → sF (s) − f (0)
f 00 (t) → s2 F (s) − sf (0) − f 0 (0)
f (n) (t) → sn F (s) − sn−1 f (0) − sn−2 f 0 (0) − · · · − f (n−1) (0)
Table 6.2
342 SECTION 6.2
17. 18.
1 1
Sine function
Sine function
4p
p 2p t 2p t
-1 -1
19. 20.
1
p 4p 2 2e - t
t
1
-1 ln 2 t
In Exercises 21–32 use property 6.8a to find the Laplace transform for the function.
21. f (t) = t3 e−5t 22. f (t) = t2 e3t
23. f (t) = 4te−t − 2e−3t 24. f (t) = 5eat − 5e−at
25. f (t) = et sin 2t + e−t cos t 26. f (t) = 2e−3t sin 3t + 4e3t cos 3t
27. f (t) = tet cos 2t 28. f (t) = te−2t sin t
29. f (t) = 2et (cos t + sin t) 30. f (t) = (t − 1)e2−3t sin 4t
31. f (t) = t2 cos at 32. f (t) = t2 sin at
In Exercises 33–42 use property 6.14a to find the Laplace transform of the function.
33. f (t) = (t − 2)2 h(t − 2) 34. f (t) = sin 3(t − 4) h(t − 4)
35. f (t) = t h(t − 1) 36. f (t) = (t + 5) h(t − 3)
2
37. f (t) = (t + 2) h(t − 1) 38. f (t) = cos t h(t − π)
39. f (t) = cos t h(t − 2) 40. f (t) = et h(t − 4)
41. f (t) = t2 et h(t − 3) 42. f (t) = et cos 2t h(t − 1)
In Exercises 43–47 find the Laplace transform of the periodic function.
43. f (t) = t, 0 < t < a, f (t + a) = f (t)

1, 0<t<a
44. f (t) = f (t + 2a) = f (t)
−1, a < t < 2a
45. f (t) = | sin at|

t, 0<t<a
46. f (t) = f (t + 2a) = f (t)
2a − t, a < t < 2a

1, 0 < t < a
47. f (t) = f (t + 2a) = f (t)
0, a < t < 2a
Find the inverse Laplace transform in Exercises 48–69.
1 s
48. F (s) = 2 49. F (s) = 2
s − 2s + 5 s + 4s + 1
e−2s e−3s
50. F (s) = 2 51. F (s) = 2
s s +1
SECTION 6.2 343
se−5s se−s
52. F (s) = 53. F (s) =
s2 + 2 (s2 + 4)2
1 s
54. F (s) = 2
55. F (s) = 2
4s − 6s − 5 s − 3s + 2
4s + 1 e−3s
56. F (s) = 57. F (s) =
(s + s)(4s2 − 1)
2 s+5
e−2s 1
58. F (s) = 59. F (s) =
s2 + 3s + 2 s3 + 1
5s − 2 e−s (1 − e−s )
60. F (s) = 2
61. F (s) =
3s + 4s + 8 s(s2 + 1)
s s2 + 2s + 3
62. F (s) = 63. F (s) =
(s + 1)5 (s2 + 2s + 2)(s2 + 2s + 5)
s2 1
64. F (s) = 65. F (s) =
(s2 − 4)2 s(1 − e−s )
1 1
66. F (s) = 67. F (s) =
s(1 + e−s ) (s + 4)(1 − e−3s )
2
1 (s2 + 1)e−2s
68. F (s) = 69. F (s) =
(s + 5s)(1 − e−2s )
3 (s4 + 2s2 )(1 + e−s )
70. To find the inverse transform of a rational function with irreducible quadratic factors in de-
nominators, we have used property 6.8b. Example 6.8 contained such a situation, and some
of the above exercises. If you love to work with complex numbers, you might be pleased to
know that you can always replace irreducible real factors with complex linear factors. It is not
a method that we recommend, but it is at least comforting to know that it can be done. We
illustrate with a simple example, hoping that it convinces you not to persue complex linear
factors in the future.
(a) Use property property 6.8b to find the inverse transform of
s+2
F (s) = .
s2 + 2s + 5
(b) Find the complex roots of s2 + 2s + 5 = 0, and use them to show that the partial fraction
decomposition of F (s) with complex linear factors is

1 2−i 2+i
F (s) = + .
4 s + 1 − 2i s + 1 + 2i
(c) Use the decomposition in part (b) to find L−1 {F (s)}.
71. The following two formulas, called reduction of order formulas, can be useful in taking
inverse transforms,

−1 s t −1 1
L = L ,
(s2 + a2 )n+1 2n (s2 + a2 )n

−1 1 −t −1 s 2n − 1 −1 1
L = L + L .
(s2 + a2 )n+1 2na2 (s2 + a2 )n 2na2 (s2 + a2 )n
Verify these formulas using

d
L{tf (t)} = − L{f (t)}.
ds
344 SECTION 6.2
This result will be verified and extended in Section 6.7.

72. Use the reduction of order formulas in Exercise 71 to verify lines 5–8 in Table 6.2.
Use the reduction of order formulas in Exercise 71 to find the inverse Laplace transform
in Exercises 73–76.
1 s
73. F (s) = 2 74. F (s) = 2
(s + a2 )3 (s + a2 )3
1 s+2
75. F (s) = 2 3
76. F (s) = 2
(s − 2s + 5) (s − 4s + 13)3
77. If F (s) = L{f (t)} for s > α, for what values of s is F (s − a) the Laplace transform of eat f (t)?
78. Find the Laplace transform of the function
 2
 t /4, 0≤t<1
2
f (t) = −(t − 4t + 2)/4, 1 ≤ t < 3 f (t + 4) = f (t).

(t − 4)2 /4, 3≤t≤4
79. Verify the change of scale property: If F (s) = L{f (t} for s > α, then for a > 0,
1 s
L{f (at)} = F , s > αa.
a a
SECTION 6.3 345
6.3 Laplace Transforms and Differential Equations

The Laplace transform is a powerful technique for solving linear, ordinary and par-
tial differential equations. It replaces differentiations with algebraic operations. Like
the techniques of Chapter 4, the transform cannot be used on nonlinear problems.
A simple example such as the following nonlinear equation illustrates why,
yy 00 + 2y 0 + 3y = t2 .
You may have noticed that we have not developed a general formula for the Laplace
transform of the product of two functions, and the reason is that there just isn’t
one. There are special cases such as when eat multiplies another function, but not
for a product such as yy 00 . This is why the transform is not applied to nonlinear
problems.
The following theorem and its corollary simplify the process of applying the
Laplace transform to linear differential equations.
Theorem 6.6 Suppose f is continuous for t ≥ 0 with a piecewise-continuous first derivative on
every finite interval 0 ≤ t ≤ T . If f is O(eαt ), then L{f 0 } exists for s > α, and
L{f 0 (t)} = sF (s) − f (0). (6.17)
(A more precise representation of the left side of this equation is L{f 0 }(s).)
Proof We prove the result when f 0 (t) has a single disontinuity at t0 . A proof for
any number of discontinuities can be found in Exercise 56. If T > t0 ,
Z T Z t0 Z T
−st 0 −st 0
e f (t) dt = e f (t) dt + e−st f 0 (t) dt.
0 0 t0
Since f 0 is continuous on each subinterval, we may integrate by parts on both

subintervals,
Z T Z t0 Z T
−st 0
−st t0 −st
−st T
e f (t) dt = e f (t) 0 + s e f (t) dt + e f (t) t + s e−st f (t) dt.
0
0 0 t0
Because f is continuous, f (t0 + ) = f (t0 − ), and therefore

Z T Z T
−st 0 −sT
e f (t) dt = −f (0) + e f (T ) + s e−st f (t) dt.
0 0
Thus,
Z ∞ Z T
0 −st 0
L{f } = e f (t) dt = lim e−st f 0 (t) dt
0 T →∞ 0
" Z #
T
−sT −st
= lim −f (0) + e f (T ) + s e f (t) dt
T →∞ 0
= sF (s) − f (0) + lim e−sT f (T ),

T →∞
provided the limit on the right exists. Since f is O(eαt), there exists M and T such
that for t > T , |f (t)| < M eαt . Thus, for T > T ,
e−sT |f (T )| < e−sT M eαT = M e(α−s)T
346 SECTION 6.3
which approaches 0 as T → ∞ (provided s > α). Consequently,

L{f 0 } = sF (s) − f (0).
This result is easily extended to second and higher order derivatives. For ex-
tensions when f is only piecewise-continuous, see Exercise 50.
Corollary 6.6.1 Suppose f and f 0 are continuous for t ≥ 0, and f 00 is piecewise-continuous on every
finite interval 0 ≤ t ≤ T . If f and f 0 are O(eαt ), then L{f 00 } exists for s > α, and
L{f 00 } = s2 F (s) − sf (0) − f 0 (0). (6.18)
Proof Since f 0 is continuous, f 00 is piecewise-continuous, and f 0 is O(eαt ), equa-

tion 6.17 gives
L{f 00 } = sL{f 0 } − f 0 (0).
We can apply equation 6.17 once again to obtain
L{f 00 } = s[sF (s) − f (0)] − f 0 (0) = s2 F (s) − sf (0) − f 0 (0).
The extension to nth -order derivatives is contained in the next corollary.
Corollary 6.6.2 Suppose f and its first n − 1 derivatives are continuous for t ≥ 0, and f (n) (t) is
piecewise-continuous on every finite interval 0 ≤ t ≤ T . If f and its first n − 1
derivatives are O(eαt ), then L{f (n) (t)} exists for s > α, and
L{f (n) (t)} = sn F (s) − sn−1 f (0) − sn−2 f 0 (0) − · · · − f (n−1) (0). (6.19)
In Section 6.1, we demonstrated how to use Laplace transforms to solve an

initial-value problem. We now consider further examples.
Example 6.19 Solve the initial-value problem
y 00 − 2y 0 + y = 2et , y(0) = y 0 (0) = 0.
Solution First we assume that the solution of the problem is a function satisfying
the conditions of Corollary 6.6.1. We can then take Laplace transforms of both sides
of the differential equation,
L{y 00 } − 2L{y 0 } + L{y} = 2L{et }.
Properties 6.17 and 6.18 yield
2
[s2 Y (s) − sy(0) − y 0 (0)] − 2[sY (s) − y(0)] + Y (s) = .
s−1
We now substitute from the initial conditions y(0) = y 0 (0) = 0,
2
s2 Y (s) − 2sY (s) + Y (s) = ,
s−1
and solve this equation for Y (s),
2
Y (s) = .
(s − 1)3
The required function y(t) can now be obtained by taking the inverse transform of
Y (s),
SECTION 6.3 347

−1 2 −1 1
y(t) = L = 2L (by linearity)
(s − 1)3 (s − 1)3

1
= 2et L−1 (by property 6.10b)
s3
2
t t
= 2e (from Table 6.1)
2
= t2 et .•
This example is typical of Laplace transforms at work on initial-value problems.

We begin by assuming that the solution of the problem satisfies whatever conditions
are necessary to apply the transform to the differential equation. In the case of Ex-
ample 6.19, this meant assuming that y(t) satisfies the conditions of Corollary 6.6.1.
In actual fact, we need only assume that y(t) and y 0 (t) are of exponential order.
Since the nonhomogeneity 2et is continuous, our theory in Chapter 4 indicates that
the solution has a continuous second derivative. In applying the Laplace transform
to a third-order differential equation, we would assume that the solution satisfies
the conditions of Corollary 6.6.2 for n = 3. The Laplace transform reduces the
differential equation in y(t) to an algebraic equation in its transform Y (s). Notice
how initial conditions for the solution of the initial-value problem are incorporated
by the Laplace transform at a very early stage, unlike the techniques of Chapter 4
where they are used to determine arbitrary constants in a general solution. The al-
gebraic equation is solved for Y (s) and the inverse transform then yields a function
y(t). That y(t) is a solution of the initial-value problem can be verified in two ways.
First, we can check that y(t) and y 0 (t) are of exponential order, thus vindicating the
initial assumption. Alternatively, we can verify that y(t) satisfies the differential
equation and initial conditions. We will omit these formal verifications, although
the problem is not truly solved until one of these actions has been taken.
In each occurrence of the Laplace transform of y(t) in the above example, we
wrote Y (s). In order to keep notation as simple as possible in further examples,
we will write Y in place of Y (s) when taking Laplace transforms of a differential
equation.
y 00 + 4y = 3 cos 2t, y(0) = 1, y 0 (0) = 0.
Solution Assuming that the solution and its first derivative are of exponential
order, we take Laplace transforms of both sides of the differential equation and use
the initial conditions,
3s
[s2 Y − s(1) − 0] + 4Y = .
s2 +4
The solution of this equation for Y (s) is
3s s
Y (s) = + 2 ,
(s2 + 4) 2 s +4
and Table 6.1 gives
348 SECTION 6.3

t
y(t) = 3 sin 2t + cos 2t.•
4
Laplace transforms thrive on initial-value problems; they use the initial condi-
tions of the problem when the Laplace transform is applied to the derivative terms
in the differential equation. They can also be adapted to boundary-value problems,
as the following example illustrates.
Example 6.21 Solve the following boundary-value problem on the interval 0 ≤ t ≤ π/2,
y 00 + 9y = cos 2t, y(0) = 1, y(π/2) = −1.
Solution The solution of the problem is only desired on the interval 0 ≤ t ≤ π/2.
What we do is solve the problem on the interval t ≥ 0, and then restrict the solution
to the interval 0 ≤ t ≤ π/2. When we apply formula 6.18 to the second derivative
in the differential equation, the derivative y 0 (0) is needed. Since it is not one of the
given pieces of information in the problem, we assign a letter to represent it; that
is, we let y 0 (0) = A. If we assume that the solution and its first derivative are of
exponential order and apply the Laplace transform to the differential equation, we
obtain
s
[s2 Y − s(1) − A] + 9Y = 2 .
s +4
We now solve for Y (s),
s+A s
Y (s) = + 2 .
s + 9 (s + 4)(s2 + 9)
2
Partial fractions on the second term gives

s+A s/5 −s/5 4s/5 + A s/5
Y (s) = 2
+ 2 + 2 = 2
+ 2 .
s +9 s +4 s +9 s +9 s +4
Inverse transforms yield
4 A 1
y(t) = cos 3t + sin 3t + cos 2t.
5 3 5
The boundary condition y(π/2) = −1 can now be used to find A,
A 1 12
−1 = − − =⇒ A= .
3 5 5
The solution of the boundary-value problem is
4 4 1
y(t) = cos 3t + sin 3t + cos 2t.•
5 5 5
Laplace transforms are particularly adept at handling initial conditions, and
as we have just seen, they can be adapted to boundary conditions. They can
also provide general solutions to linear differential equations, as shown in the next
example.
Example 6.22 Find a general solution of the differential equation y 00 + 2y 0 − 3y = t2 .
Solution We denote initial values of the solution and its first derivative by y(0) =
A and y 0 (0) = B. If we assume that the solution and its first derivative are of
SECTION 6.3 349
exponential order, and take Laplace transforms of both sides of the differential
equation,
2
[s2 Y − s(A) − B] + 2[sY − A] − 3Y = .
s3
The solution of this equation for Y (s) is
2 As + (B + 2A)
Y (s) = + .
s3 (s2 + 2s − 3) s2 + 2s − 3
The partial fraction decomposition of the first term is
2 −2/3 4/9 14/27 1/2 1/54
= − 2 − + + .
s3 (s2 + 2s − 3) s3 s s s−1 s+3
Hence,
−2/3 4/9 14/27 1/2 1/54 As + (B + 2A)
Y (s) = 3
− 2 − + + + .
s s s s−1 s+3 (s − 1)(s + 3)
If we are not concerned with preserving the fact that A and B represent initial
values for y(t) and its first derivative, we can write that Y (s) is of the form
−2/3 4/9 14/27 C D
Y (s) = 3
− 2 − + + ,
s s s s−1 s+3
where C and D are constants. Inverse transforms now give a general solution
t2 4t 14
y(t) = − − − + Cet + De−3t .•
3 9 27
Did you notice that the denominator of the transform in each of the above
examples is the function φ(m) in the auxiliary equation of Chapter 4 with m replaced
by s; that is, it is φ(s). This is always the case for linear differential equations with
constant coefficients; and it can serve as a partial check on calculations.
Example 6.23 A 2-kilogram mass is suspended from a spring with constant 128 newtons per metre.
It is pulled 4 centimetres above its equilibrium position and released. An external
force 3 sin ωt newtons acts vertically on the mass during its motion. If damping is
negligible, find the position of the mass as a function of time.
Solution The initial-value problem describing oscillations of the mass is
d2 x
2 + 128x = 3 sin ωt, x(0) = 1/25, x0 (0) = 0.
dt2
If we take Laplace transforms of both sides of the differential equation,
3ω 3ω s
2[s2 X − s/25] + 128X = =⇒ X(s) = + .
s2 +ω 2 2 2 2 2
2(s + 64)(s + ω ) 25(s + 64)
When ω 6= 8, partial fractions on the first term on the left leads to
3ω 3ω s
X(s) = 2 2 2
− 2 2
+ 2
.
2(64 − ω )(s + ω ) 2(64 − ω )(s + 64) 25(s + 64)
Hence, displacement in the absence of resonance is
350 SECTION 6.3
3 3ω 1
x(t) = 2
sin ωt − 2
sin 8t + cos 8t.
2(64 − ω ) 16(64 − ω ) 25
When ω = 8, the Laplace transform X(s) takes the form
12 s
X(s) = + ,
(s2 + 64) 2 2
25(s + 64)
in which case Table 6.1 gives the resonant solution
12 1 3 3t 1
x(t) = 3
(sin 8t − 8t cos 8t) + cos 8t = sin 8t − cos 8t + cos 8t.•
2(8) 25 256 32 25
Convolutions
As a linear operator, the Laplace transform efficiently handles sums and dif-
ferences of functions. What it does not handle is the product of functions; that is,
we do not have a formula for the Laplace transform of the product of two functions
f (t) and g(t). We can take the transform of certain products such as an exponential
multiplying another function (formula 6.8). In Section 6.7, we also find out how
to take the Laplace transform of the product tn f (t), where n is a positive integer.
But, in general, there is no formula for the Laplace transform of the product of two
arbitrary functions. One would expect that there would therefore be no formula for
the inverse Laplace transform of the product of two functions. Surprisingly, there is
a formula for L−1 {F (s)G(s)} when inverse transforms of F (s) and G(s) are known.
We shall see shortly that the inverse of F (s)G(s) is what is called the convolution
of f (t) and g(t).
Definition 6.4 The convolution of two functions f and g is a function denoted by f ∗g with values
defined by
Z t
(f ∗ g)(t) = f (u)g(t − u) du. (6.20)
0
The following properties of convolutions are easily verified using Definition 6.4:
f ∗ g = g ∗ f, (6.21a)
f ∗ (kg) = (kf ) ∗ g = k(f ∗ g), k a constant (6.21b)
(f ∗ g) ∗ h = f ∗ (g ∗ h), (6.21c)
f ∗ (g + h) = f ∗ g + f ∗ h. (6.21d)
Example 6.24 Find the convolution of f (t) = sin t and g(t) = cos 4t.
Solution According to equation 6.20,
Z t
(f ∗ g)(t) = sin u cos 4(t − u) du.
0
With the trigonometric identity sin A cos B = (1/2)[sin (A + B) + sin (A − B)], we

obtain
Z
1 t
(f ∗ g)(t) = [sin (4t − 3u) + sin (5u − 4t)] du
2 0
t
1 1 1
= cos (4t − 3u) − cos (5u − 4t)
2 3 5 0
1
= (cos t − cos 4t).•
15
SECTION 6.3 351
The importance of convolutions lies in the corollary to the following theorem.

Theorem 6.7 If f and g are O(eαt ) and piecewise-continuous on every finite interval 0 ≤ t ≤ T ,
then
L{f ∗ g} = L{f }L{g}, s > α. (6.22a)
Proof If F = L{f } and G = L{g}, then

Z ∞ Z ∞ Z ∞Z ∞
−su −sτ
F (s)G(s) = e f (u) du e g(τ ) dτ = e−s(u+τ ) f (u)g(τ ) dτ du.
0 0 0 0
Suppose we change variables of integration in the inner integral with respect to τ

by setting t = u + τ . Then
Z ∞Z ∞ Z TZ ∞
−st
F (s)G(s) = e f (u)g(t − u) dt du = lim e−st f (u)g(t − u) dt du.
0 u T →∞ 0 u
We would like to interchange orders of integration, but to do so requires that the

inner integral converge uniformly with respect to u. To verify that this is indeed
the case we note that since f and g are O(eαt ) and piecewise-continuous on every
finite interval 0 ≤ t ≤ T , there exists a constant M such that for all t ≥ 0, |f (t)| <
M eαt and |g(t)| < M eαt . For each u ≥ 0, we therefore have |e−st f (u)g(t − u)| <
M 2 e−st eαu eα(t−u) = M 2 e−t(s−α) . Thus,
Z ∞ Z ∞ −t(s−α) ∞
−st 2 −t(s−α) 2 e
e f (u)g(t − u) dt < M e dt = M
u u α−s u
M 2 e−u(s−α) M2
= < ,
s−α s−α
u
provided s > α, and the improper integral
is uniformly convergent with respect
to u. The order of integration in u=t
the expression for F (s)G(s) may therefore

be interchanged (Figure 6.26),
T t
and we obtain
Figure 6.26
"Z Z
T t
F (s)G(s) = lim e−st f (u)g(t − u) du dt
T →∞ 0 0
Z Z #
∞ T
+ e−st f (u)g(t − u) du dt .
T 0
Since
Z ∞ Z T Z ∞Z T
−st
e f (u)g(t − u) du dt < M 2 e−t(s−α) du dt
T 0 T 0
∞
2 e−t(s−α) M 2 T e−T (s−α)
=M T =
α−s T s−α
352 SECTION 6.3
provided s > α, it follows that

Z ∞ Z T
lim e−st f (u)g(t − u) du dt = 0.
T →∞ T 0
Thus,
Z T Z t Z T
−st
F (s)G(s) = lim e f (u)g(t − u) du dt = lim e−st f ∗ g dt = L{f ∗ g}.
T →∞ 0 0 T →∞ 0
More important in practice is the inverse of property 6.22a.

Corollary 6.7.1 If L−1 {F } = f and L−1 {G} = g, where f and g are O(eαt ) and piecewise-continuous
on every finite interval, then
L−1 {F G} = f ∗ g. (6.22b)
This is line 17 in Table 6.2. The following example illustrates how to use this
corollary.
2
Example 6.25 Find the inverse transform of F (s) = .
s2 (s2 + 4)
Solution Since L−1 {2/(s2 +4)} = sin 2t and L−1 {1/s2 } = t, convolution property
6.22b gives
Z t
−1 2
L = u sin 2(t − u) du
s2 (s2 + 4) 0
t
u 1 t 1
= cos 2(t − u) + sin 2(t − u) = − sin 2t.
2 4 0 2 4
The alternative is to use partial fractions.•

Convolutions are particularly useful when solving differential equations that
contain unspecified nonhomogeneities.
Example 6.26 Find the solution of the initial-value problem
y 00 + 2y 0 + 3y = f (t), y(0) = 1, y 0 (0) = 0,
where f (t) is of exponential order and piecewise-continuous for t ≥ 0.
Solution The only technique from Chapter 4 that can handle this problem is
variation of parameters; the other techniques require that the form of the nonhomo-
geneity be known. To show that Laplace transforms can also be used, we assume
that the solution and its first derivative are of exponential order, and take Laplace
transforms of both sides of the differential equation,
[s2 Y − s] + 2[sY − 1] + 3Y = F (s).
We solve for Y (s),
F (s) s+2
Y (s) = + .
s2 + 2s + 3 s2 + 2s + 3
To find the inverse transform of this function, we first note that
SECTION 6.3 353

1 1 1 1 −t √
L−1 = L−1
= e−t −1
L = √ e sin 2t.
s2 + 2s + 3 (s + 1)2 + 2 s2 + 2 2
Convolution property 6.22b on the first term of Y (s) now yields
Z t
1 √ (s + 1) + 1
y(t) = f (u) √ e−(t−u) sin 2(t − u) du + L−1
0 2 (s + 1)2 + 2
Z t
1 √ s+1
=√ f (u)e−(t−u) sin 2(t − u) du + e−t L−1
2 0 s2 + 2
Z t
1 −(t−u)
√ −t
√ 1 √
=√ f (u)e sin 2(t − u) du + e cos 2t + √ sin 2t .•
2 0 2
The following nontrivial problem makes clever use of convolutions.

The Tautochrone
A bead, with zero initial velocity is to
slide frictionlessly down a wire from a
point P (x, y) to the origin (Figure 6.27). y
P (x,y)
Tautochrone is the name attached to the
shape of the curve for which the time of x = x ( y)
descent is independent of the height, y,
(h, z )
on the curve from which the bead starts.
Suppose we let the equation of the curve
be x = x(y). Since the kinetic energy of the x
bead at any point (η, ζ) along the curve Figure 6.27
is equal to the loss in gravitational potential
energy of the bead, we can write that
1
mv 2 = mg(y − ζ),
2
where m is the mass of the bead and v is its velocity. It follows that the velocity of
the bead at (η, ζ) is
p p
v = 2g y − ζ.
p
Since the time for the bead to traverse an element of arc length dη 2 + dζ 2 at point
(η, ζ) is this length divided by v, the total time to travel from P to the origin is
s 2
dη
Z yp 2 2
Z y 1+
dη + dζ 1 dζ
T = √ √ =√ √ dζ.
0 2g y − ζ 2g 0 y−ζ
p
If we set f (ζ) = 1 + (dη/dζ)2 , then
Z y
1 f (ζ)
T =√ √ dζ.
2g 0 y−ζ
√
This integral can be interpreted as the convolution of the functions f (y) and 1/ y.
If we take Laplace transforms of both sides of the equation with respect to y, and
note that T is a constant, we obtain
354 SECTION 6.3
√
2gT 1
= L f (y) ∗ √ .
s y
√ p
According to Exercise 35 in Section 6.1, the Laplace transform of 1/ y is π/s.
Thus,
√ r √ r
2gT π 2gT π
= F (s) =⇒ F (s) = .
s s π s
The inverse transform now gives
s 2 √
dx 2gT 1
f (y) = 1+ = √ ,
dy π y
a differential equation for x(y). It can be rewritten in the form
s
dx k2 − y
=− ,
dy y
√
where we have set k = 2gT /π. We now integrate and set y = k 2 sin2 θ,
Z p 2 Z
k −y k cos θ 2
−x + C = √ dy = 2k sin θ cos θ dθ
y k sin θ
Z Z
2 2 2 1 + cos 2θ 2 1
= 2k cos θ dθ = 2k dθ = k θ + sin 2θ
2 2
√ √ r
y y y
= k 2 Sin−1 + 1− 2
k k k
√ p
y
= k 2 Sin−1 + k2 y − y2 .
k
To pass through the origin, C must be zero, so that the equation of the tautochrone
is
√ p
2 −1 y
x = −k Sin − k2 y − y2 .
k
By setting φ = −2θ, parametric equations for the tautochrone are
k2 k2
x=− (2θ + sin 2θ) = (φ + sin φ),
2 2
2 2 2 1 − cos 2θ k2
y = k sin θ = k = (1 − cos φ).
2 2
These are parametric equations for a cycloid (see Exercise 61 in Section 2.2).
EXERCISES 6.3
In Exercises 1–16 use Laplace transforms to solve the initial-value problem.

1. y 00 + 3y 0 − 4y = t + 3, y(0) = 1, y 0 (0) = 0
2. y 00 + 2y 0 − y = et , y(0) = 1, y 0 (0) = 2
SECTION 6.3 355
3. y 00 + y = 2e−t , y(0) = y 0 (0) = 0

4. y 00 + 2y 0 + y = t, y(0) = 0, y 0 (0) = 1
5. y 00 − 2y 0 + y = t2 et , y(0) = 1, y 0 (0) = 0
6. y 00 + y = t, y(0) = 1, y 0 (0) = −2
7. y 00 + 2y 0 + 5y = e−t sin t, y(0) = 0, y 0 (0) = 1
8. y 00 + 6y 0 + y = sin 3t, y(0) = 2, y 0 (0) = 1
9. y 00 + y 0 − 6y = t + cos t, y(0) = 1, y 0 (0) = −2
10. y 00 − 4y 0 + 5y = te−3t , y(0) = −1, y 0 (0) = 2

00 0 1, 0 < t < 1
11. y + 4y = f (t), y(0) = 0, y (0) = 1, where f (t) =
0, t > 1
12. y 00 + 2y 0 − 4y = cos2 t, y(0) = 0, y 0 (0) = 0
13. y 00 − 3y 0 + 2y = 8t2 + 12e−t , y(0) = 0, y 0 (0) = 2
14. y 00 + 4y 0 − 2y = sin 4t, y(0) = 0, y 0 (0) = 0
15. y 00 + 8y 0 + 41y = e−2t sin t, y(0) = 0, y 0 (0) = 1

00 0 0 t, 0<t<1
16. y + 2y + y = f (t), y(0) = 0, y (0) = 0, where f (t) =
0, t>1
In Exercises 17–19 use Laplace transforms to solve the boundary-value problem.
17. y 00 + 9y = cos 2t, y(0) = 1, y(π/2) = −1
18. y 00 + 3y 0 − 4y = 2e−4t , y(0) = 1, y(1) = 1
19. y 00 + 2y 0 + 5y = e−t sin t, y(0) = 0, y(π/4) = 1
20. There are two ways to use Laplace transforms to solve the initial-value problem
y 00 + 2y 0 − 3y = sin 2t, y(π/2) = 3, y 0 (π/2) = −1.
Do both:
(a) Find a general solution and then use the initial conditions to find constants.
(b) Translate the initial conditions to u = 0 by setting u = t − π/2. Solve the problem for y(u),
and then return to y(t).
In Exercises 21–24 use Laplace transforms to find an integral representation for the
solution to the problem.
21. y 00 − 4y 0 + 3y = f (t), y(0) = 1, y 0 (0) = 0
22. y 00 + 4y 0 + 6y = f (t), y(0) = 0, y 0 (0) = 0
23. y 00 + 16y = f (t)
24. y 00 + 3y 0 + 2y = et f (t)
In Exercises 25–28 use convolutions to find the inverse Laplace transform for the func-
tion.
1 1
25. F (s) = 26. F (s) = 2
s(s + 1) (s + 1)(s2 + 4)
356 SECTION 6.3
s s
27. F (s) = 28. F (s) =
(s + 4)(s2 − 2) (s2 − 4)(s2 − 9)
In Exercises 29–34 use Laplace transforms to find a general solution of the differential
equation.
29. y 00 − 2y 0 + 4y = t2 30. y 00 − 2y 0 + y = t2 et
31. y 00 + y = f (t) 32. y 00 + 2y 0 + 5y = e−t sin t
33. y 00 + 4y 0 + y = t + 2 34. y 00 − 4y = f (t)
35. To find a general solution for y 00 + 9y = t sin t, replace t sin t by teti , solve the equation, and
then take imaginary parts.
36. To find a general solution for y 00 −2y 0 +3y = t cos 2t, replace t cos 2t by te2ti , solve the equation,
and then take real parts.
Solve the problem in Exercises 37–38.
37. y 000 − 3y 00 + 3y 0 − y = t2 et , y(0) = 1, y 0 (0) = 0, y 00 (0) = −2
38. y 000 − 3y 00 + 3y 0 − y = t2 et
One end of a spring with constant k newtons per metre is attached to a mass of M
kilograms and the other end is attached to a wall (figure below).
Dashpot
k
Wall
M
x= 0 x
Attached to the mass is a dashpot that provides, or represents, a resistive force on the
mass directly proportional to the velocity of the mass. If all other forces are grouped
into a function denoted by f (t), the differential equation governing motion of the mass
is
d2 x dx
M 2
+β + kx = f (t),
dt dt
where β > 0 is a constant. The position of M when the spring is unstretched corresponds
to x = 0. Accompanying the differential equation will be two initial conditions x(0) = x0
and x0 (0) = v0 representing the initial position and velocity of M . In Exercises 39–45,
solve the initial-value problem with the given information.
39. M = 1/5, β = 0, k = 10, f (t) = 0, x(0) = −0.03, x0 (0) = 0
40. M = 2, β = 0, k = 16, f (t) = 0, x(0) = 0.1, x0 (0) = 0
41. M = 1/5, β = 3/2, k = 10, f (t) = 0, x(0) = −0.03, x0 (0) = 0
42. M = 1/5, β = 3/2, k = 10, f (t) = 4 sin 10t, x(0) = 0, x0 (0) = 0
43. M = 1/10, β = 1/20, k = 5, f (t) = 0, x(0) = −1/20, x0 (0) = 2
44. M = 1/10, β = 0, k = 4000, f (t) = 3 cos 200t, x(0) = 0, x0 (0) = 10
45. M = 1, β = 0, k = 64, f (t) = 2 sin 8t, x(0) = 0, x0 (0) = 0
SECTION 6.3 357
If the first or second derivative of a function f (t) yields functions with known Laplace
transforms and/or f (t), then equations 6.17 and 6.18 can be used to find the Laplace
transform of f (t). We illustrate this in Exercises 46–49.
46. Find the Laplace transform of sin2 at by:
(a) using a trigonometric identity;
(b) equation 6.17 and the fact that L{sin at} = a/(s2 + a2 ).
47. Find the Laplace transform of t cos at by calculating its second derivative and using equation
6.18.
48. In Example 6.5, we used multiple integrations by parts to find the Laplace transform of tn
when n is a nonnegative integer. Use equation 6.17 and mathematical induction to verify the
transform.
√
49. (a) Use equation 6.17 and Exercise 35 in Section 6.1 to find the Laplace transform of t.
(b) Extend the result in part (a) to find the Laplace transform of t(2n+1)/2 when n ≥ 0 is an
integer.
50. (a) Let f be O(eαt) and be continuous for t ≥ 0 except for a finite discontinuity at t = t0 > 0;
and let f 0 be piecewise continuous on every finite interval 0 ≤ t ≤ T . Show that
L{f 0 } = sF (s) − f (0) − e−st0 [f (t0 +) − f (t0 −)].
(b) What is the result in part (a) if t0 = 0?
(c) Extend the result in part (a) when f (t) has discontinuities at the points tn , n ≥ 1 an integer?
51. Use the result of Exercise 50 to find the Laplace transform of the floor function btc, t ≥ 0.
52. (a) Calculate the convolution of tm and tn when m and n are positive integers.
(b) Calculate the convolution of tm and tn when m > 0 and n > 0 are not positive integers.
53. Verify that when f (t) satisfies the conditions of Theorem 6.6, then its Laplace transform satisfies
the equation
lim [sF (s)] = f (0).
s→∞
This is called the initial-value theorem. Like Theorem 6.3, it can serve as a partial check on
taking the Laplace transform of a given function f (t).
54. Suppose that f (t) satisfies the conditions of Theorem 6.6 and that f 0 (t) is also of exponential
order. Verify that if limt→∞ f (t) exists, then
lim s F (s) = lim f (t).
s→0 t→∞
This is called the final-value theorem.

2
55. Prove that the derivative of the function sin (et ) in Exercise 46 of Section 6.1 has a Laplace
transform.
56. Verify Theorem 6.6 when f 0 (t) has n discontinuities.
358 SECTION 6.4
6.4 Piecewise-defined and Discontinuous Nonhomogeneities

Nonhomogeneities for the linear differential equations in Section 6.3 were all contin-
uous. As a result, Laplace transforms did not prove overly advantageous compared
to methods of Chapter 4. In this section we show that Laplace transforms are ex-
ceptional for handling discontinuous and piecewise-defined nonhomogeneities. But,
before doing so, we examine the expectations of solutions to linear differential equa-
tions that contain piecewise-continuous nonhomogeneities.
We begin with the initial-value problem associated with a linear first-order
differential equation,
dy
+ P (t)y = Q(t), y(0) = y0 , (6.23)
dt
on the interval t > 0, where P (t) is continuous. When Q(t) is also continuous,
the solution y(t) of the initial-value problem is unique. It is continuous and has
a continuous first derivative. But what can we say about the solution if Q(t) is
piecewise-continuous? Suppose, for example, that Q(t) has a single, finite-jump
discontinuity at some value t0 > 0. The initial-value problem has a continuous
solution with continuous derivative on the interval 0 < t < t0 , call it y1 (t). The
differential equation also has a general solution, call it y2 (t) on the interval t > t0 .
We can match these solutions at t0 by demanding that the solution be continuous
at t0 . This requires
lim y1 (t) = lim+ y2 (t).
t→t−
0 t→t0
This determines the arbitrary constant in the general solution y2 (t). It creates a
continuous function that satisfies the differential equation at every value of t except
t0 . The differential equation requires dy/dt = Q(t)−P (t)y(t). Because y(t) and P (t)
are continuous, but Q(t) is discontinuous at t0 , it follows that y(t) is differentiable
at every value of t except t0 . Since we can match solutions at every discontinuity
of Q(t), we have the following theorem.
Theorem 6.8 When Q(t) is piecewise-continuous on the interval t > 0, there exists a unique so-
lution of initial-value problem 6.23 that is continuous and has a continuous first
derivative except at points of discontinuity of Q(t). It therefore satisfies the differ-
ential equation except at discontinuities of Q(t).
Here is an example to illustrate this matching.
Example 6.27 Find the solution of the following initial-value problem with a piecewise-continuous
nonhomogeneity

dy t, 0 < t < 1
+ 3y = f (t), where f (t) =
dt 2, t > 1,
subject to y(0) = 1.
Solution First we solve the differential equation on the interval 0 < t < 1, in
which case it is
dy
+ 3y = t.
dt
SECTION 6.4 359
An integrating factor is e3t , so that multiplication of the differential equation by e3t

results in
dy d
e3t + 3ye3t = te3t or (ye3t ) = te3t .
dt dt
Antidifferentiation gives
Z
3t t 3t 1 3t
ye = te3t dt = e − e + C,
3 9
and therefore
t 1
y(t) = − + Ce−3t .
3 9
The initial condition y(0) = 1 requires 1 = −1/9 + C, and therefore C = 10/9. The
solution on the interval 0 < t < 1 is
t 1 10
y(t) = − + e−3t .
3 9 9
We now consider the differential equation on the interval t > 1,
dy
+ 3y = 2.
dt
Once again e3t is an integrating factor, and this leads to the solution
2
y(t) = + De−3t , for t > 1.
3
What remains is to evaluate constant D. According to Theorem 6.8, there is a
solution that is continuous for all t, and in particular at t = 1. This requires

t 1 10 −3t 2 −3t
lim y(t) = lim y(t) =⇒ lim − + e = lim + De .
t→1− y→1+ t→1− 3 9 9 t→1+ 3
Evaluating the limits gives

1 1 10 −3 2 1
− + e = + De−3 , from which D = (10 − 4e3 ).
3 9 9 3 9
Thus, the solution of the initial-value problem is the function

 t 1 10
 − + e−3t , 0≤t≤1
y(t) = 3 9 9

 2 + 1 (10 − 4e3 )e−3t , t > 1.•
3 9
It is graphed in Figure 6.28. As predicted 1 y
by Theorem 6.8, it is continuous, even
at the discontinuity t = 1 of f (t), but 1/2
it does not have a derivative there.•
1 2 t
Figure 6.28
We now give a similar discussion for initial-value problems associated with
second-order, linear differential equations,
360 SECTION 6.4
d2 y dy
a2 (t) + a1 (t) + a0 (t)y = f (t), y(0) = y0 , y 0 (0) = y00 . (6.24)
dt2 dt
We assume as usual that a2 (t), a1 (t), and a0 (t) are continuous for t ≥ 0, and
a2 (t) 6= 0 for any value of t ≥ 0. When f (t) is also continuous, the solution
y(t) of the initial-value problem is unique. It is continuous and has continuous
first and second derivatives. Suppose, however, that f (t) has a single, finite-jump
discontinuity at some value t0 > 0. The initial-value problem has a continuous
solution with continuous first and second derivatives on the interval 0 < t < t0 , call
it y1 (t). The differential equation also has a general solution, call it y2 (t) on the
interval t > t0 . We can match these solutions at t0 by demanding that the solution
and its first derivative be continuous at t0 ,
lim y1 (t) = lim y2 (t), lim y10 (t) = lim y20 (t).
t→t−
0 t→t+
0 t→t−
0 t→t+
0
This determines the arbitrary constants in the general solution y2 (t). It creates
a continuous function with a continuous first derivative that satisfies the differ-
ential equation at every value of t except t0 . Because d2 y/dt2 = f (t)/a2 (t) −
[a1 (t)/a2 (t)]dy/dt − [a0 (t)/a2 (t)]y, y(t) has a second derivative at every value of t
except t0 . Since we can match solutions at every discontinuity of Q(t), we have the
following theorem.
Theorem 6.9 When f (t) is piecewise-continuous on the interval t > 0, there exists a unique
solution of initial-problem 6.24 that is continuous, with a continous first derivative,
and has a continuous second derivative except at points of discontinuity of f (t). It
therefore satisfies the differential equation except at discontinuities of f (t).
The following example illustrates this matching.
Example 6.28 Solve the following initial-value problem with a piecewise-continuous nonhomogene-
ity,
y 00 + 2y 0 + y = f (t), y(0) = 1, y 0 (0) = 0,

t, 0 < t < 1
where f (t) =
0, t > 1.
Solution The auxiliary equation m2 + 2m + 1 = 0 has double root m = −1. On
the interval 0 < t < 1, a particular solution of the differential equation is yp = t − 2,
and hence a general solution on this interval is y1 (t) = (C1 + C2 t)e−t + t − 2. The
initial conditions require
1 = y(0) = C1 − 2, 0 = y 0 (0) = C2 − C1 + 1,
the solution of which is C1 = 3 and C2 = 2. On the interval 0 < t < 1, then,
y1 (t) = (3 + 2t)e−t + t − 2.
For t > 1, a general solution of the differential equation is y2 (t) = (D1 + D2 t)e−t .
For the solution to be continuous and have a continuous first derivative at t = 1,
we must have
lim y1 (t) = lim+ y2 (t), lim y10 (t) = lim+ y20 (t).
t→1− t→1 t→1− t→1
Substitution for y1 (t) and y2 (t) gives

SECTION 6.4 361
5e−1 − 1 = (D1 + D2 )e−1 , −3e−1 + 1 = −D1 e−1 .

These can be solved for D1 = 3 − e and D2 = 2, and therefore the solution of the
initial-value problem is

(3 + 2t)e−t + t − 2, 0 ≤ t ≤ 1
y(t) =
(3 − e + 2t)e−t , t > 1.•
It is graphed in Figure 6.29. As predicted y
1
by Theorem 6.9, it is continuous, and
appears to have a continuous first deriva-
1/2
tive at the discontinuity t = 1 of f (t).
The second derivative is discontinuous
at t = 1, but we cannot see this graphically.• 1 2 t
Figure 6.29
This procedure of matching solutions at finite discontinuities of nonhomo-
geneities can be extended to include initial-value problems associated with nth -order,
linear differential equations
dn y dn−1 y dy
an (t) n
+ an−1 (t) n−1
+ · · · + a1 (t) + a0 (t)y = f (t), (6.25a)
dt dt dt
subject to initial conditions
(n−1)
y1 (0) = y0 , y 0 (0) = y00 , ··· , y (n−1) (0) = y0 . (6.25b)
At each discontinuity of f (t), the function and its first n − 1 derivatives are matched
to produce a solution that has continuous derivatives of orders up to and including
n − 1, but a discontinuity in the nth derivative results.
What is most important to realize from the above discussion is that matching
a solution and its derivatives at discontinuities of a piecewise-continuous nonhomo-
geneity is tedious, especially as the number of discontinuities increases. Laplace
transforms provide an excellent alternative. If we apply Laplace transforms to the
initial-value problem of Example 6.27, we get
sY − 1 + 3Y = L{f (t)},
where
L{f (t)} = L{t[h(t) − h(t − 1)] + 2 h(t − 1)} = L{t + (2 − t)h(t − 1)}

1 −s 1 −s 1 1
= 2 + e L{2 − (t + 1)} = 2 + e − .
s s s s2
Thus,

1 1 −s 1 1 1 + s2 e−s (s − 1)
Y (s) = 1+ 2 +e − 2 = 2 + 2 .
s+3 s s s s (s + 3) s (s + 3)
Partial fractions and inverse transforms give

−1 −1/9 1/3 10/9 −s 4/9 1/3 4/9
y(t) = L + 2 + +e − 2 −
s s s+3 s s s+3

1 t 10 4 1 4
= − + + e−3t + − (t − 1) − e−3(t−1) h(t − 1).
9 3 9 9 3 9
362 SECTION 6.4
This is the solution obtained in Example 6.27.

We now solve the initial-value problem in Example 6.28 using Laplace trans-
forms. When we take transforms of the differential equation, we get
[s2 Y − s] + 2[sY − 1] + Y = L{f (t)},
where
L{f (t)} = L{t [h(t) − h(t − 1)]} = L{t − L{t h(t − 1)}

1 1 1 1
= 2 − e−s L{t + 1} = 2 − e−s + .
s s s2 s
Thus,

1 1 −s 1 1
Y (s) = s+2+ 2 −e +
(s + 1)2 s s2 s
−s
(s + 1) + 1 1 e
= 2
+ 2 2
− 2 .
(s + 1) s (s + 1) s (s + 1)
Partial fractions on the second and third terms leads to

1 1 2 1 2 1 −s 1 1 1
Y (s) = + + − + 2+ + +e − −
s + 1 (s + 1)2 s s s + 1 (s + 1)2 s s2 s+1

2 1 3 2 1 1 1
=− + 2 + + 2
+ e−s − 2− .
s s s + 1 (s + 1) s s s+1
Consequently,
y(t) = −2 + t + 3e−t + 2te−t + [1 − (t − 1) − e−(t−1) ] h(t − 1)

= (3 + 2t)e−t + t − 2 + (2 − t − e1−t ) h(t − 1).
This solution is identical to that in Example 6.28.

We now use Laplace transforms to solve other initial-value problems with
piecewise-defined and/or discontinuous nonhomogeneities. We invite the reader to
make comparisons to solutions obtained by matching solutions and their derivatives
at discontinuities.
Example 6.29 Find the amount of salt in the tank of Example 3.11 if pure water is added for the
first ten minutes at 5 millilitres per second, and then the brine mixture is added
thereafter.
Solution If we change from the letter S in Example 3.11 to represent the number
of grams of salt in the tank to x, the initial-value problem for x(t) is
dx 5x
= 0.1h(t − 600) − 6 , x(0) = 5000.
dt 10
When we take Laplace transforms on both sides of the differential equation,
e−600s 5X e−600s /(10s) + 5000
sX − 5000 = − 6 =⇒ X(s) = .
10s 10 s + 5/106
Inverse transforms now give
SECTION 6.4 363
6
−5t/106 1 −1 10 /5 106 /5 −600s
x(t) = 5000e + L − e
10 s s + 5/106
6
h 6
i
= 5000e−5t/10 + 20 000 1 − e−5(t−600)/10 h(t − 600).•
A graph of this function is shown in Figure 6.30a. Since pure water is added for the
first 10 minutes, the amount of salt in the tank decreases. This is evidenced by the
negative slope of the graph for 0 < t < 600. At t = 600, the amount of salt in the
tank is
6
lim x(t) = lim (5000e−5t/10 ) = 5000e−600/20 000 ≈ 4852 grams.
t→600− t→600−
Because the concentration of salt in the tank at this time is 4852/106 = 4.852 ×
10−3 g/mL which is less than the concentration 0.02 g/mL of incoming salt, the
amount of salt now begins to increase (the slope of the graph is positive). For
large t, the amount of salt in the tank approaches 20 000 grams, with concentration
20 000/106 = 0.02 g/mL, the concentration of incoming brine. The asymptote is
shown in Figure 6.30b.
In agreement with Theorem 6.8, there is a discontinuity in the slope of the
graph at t = 600 when the input rate of salt is discontinuous. More importantly,
the graph is continuous. The Laplace transform has matched the solutions on the
intervals t < 600 and t > 600 to produce a continuous solution at t = 600 (as
Theorem 6.8 says can be done).•
x x
20 000
5000
2500 10 000
1000 2000 t 20 000 40 000 t

It should be noted that Laplace transforms cannot be applied to mixing prob-
lems when the amount of liquid in the tank is not constant. For instance, the
differential equation for the amount of salt in the tank of Example 3.12 is
dS 1 5S
= − 6 .
dt 5 10 + 5t
We have no formula for the Laplace transform of the second term on the right side
of this equation.
It is set into motion by lifting it 10 centimetres above its equilibrium position and
then releasing it. A sinusoidal force A sin 8t acts on the mass but only for t > 1.
Find the position of the mass as a function of time if damping is negligible.
Solution The initial-value problem for displacement is
d2 x 1
2 + 512x = A sin 8t h(t − 1), x(0) = , x0 (0) = 0.
dt2 10
364 SECTION 6.4
If we take Laplace transforms,

s
2 s2 X − + 512X = Ae−s L{sin 8(t + 1)}
10
= Ae−s L{cos 8 sin 8t + sin 8 cos 8t}

−s 8 cos 8 (sin 8)s
= Ae + .
s2 + 64 s2 + 64
Hence,
s Ae−s [8 cos 8 + (sin 8)s]
X(s) = + .
10(s2 + 256) 2(s2 + 256)(s2 + 64)
Partial fractions on the second term gives

s Ae−s 8 cos 8 + (sin 8)s 8 cos 8 + (sin 8)s
X(s) = + − ,
10(s2 + 256) 384 s2 + 64 s2 + 256
and therefore

1 A −1 8 cos 8 + (sin 8)s 8 cos 8 + (sin 8)s
x(t) = cos 16t + L − h(t − 1)
10 384 s2 + 64 s2 + 256 |t→t−1

1 A 1
= cos 16t + cos 8 sin 8t + sin 8 cos 8t − cos 8 sin 16t
10 384 2

− sin 8 cos 16t h(t − 1)
|t→t−1
1 A
= cos 16t + cos 8 sin 8(t − 1) + sin 8 cos 8(t − 1)
10 384
1
− cos 8 sin 16(t − 1) − sin 8 cos 16(t − 1) h(t − 1).
2
The form of the solution changes at t = 1 when the force A sin 8t is applied. For
0 ≤ t ≤ 1,
1
x(t) = cos 16t,
and for t > 1, 10
1 A
x(t) = cos 16t + cos 8 sin 8(t − 1) + sin 8 cos 8(t − 1)
10 384
1
− cos 8 sin 16(t − 1) − sin 8 cos 16(t − 1) .
2
For 0 ≤ t ≤ 1, motion is simple harmonic with ampitude 1/10 and period π/8. For
t > 1, motion is periodic with period π/4, but it is not simple harmonic. The
function is graphed in Figure 6.31 for x
A = 100. Even though the force is 0.4
discontinuous at t = 1, the graph is 0.2
continuous and so also is the first 1 2 3 t
derivative. In other words, displace- -0.2
ment and velocity of the mass are con- -0.4
tinuous at t = 1. This could be shown
algebraically, but it is a direct result Figure 6.31
SECTION 6.4 365
of Theorem 6.9. There is a discontinuity in the second derivative, but this cannot
be seen graphically.•
Example 6.31 Repeat Example 6.30 when the applied force is A sin 16t.
Solution The initial-value problem for displacement is
2
d x 1
22
+ 512x = A sin 16t h(t − 1), x(0) = , x0 (0) = 0.
dt 10
If we take Laplace transforms,
s
2 s2 X − + 512X = Ae−s L{sin 16(t + 1)}
10
= Ae−s L{cos 16 sin 16t + sin 16 cos 16t}

16 cos 16 (sin 16)s
= Ae−s 2 + 2 .
s + 256 s + 256
Hence,
s Ae−s [16 cos 16 + (sin 16)s]
X(s) = + .
10(s2 + 256) 2(s2 + 256)2
First we use Table 6.2 to calculate that

−1 16 cos 16 (sin 16)s 16 cos 16 sin 16
L 2 2
+ 2 2
= 3
(sin 16t − 16t cos 16t) + t sin 16t.
(s + 256) (s + 256) 2(16 ) 32
Hence,

1 A cos 16
x(t) = cos 16t + [sin 16(t − 1) − 16(t − 1) cos 16(t − 1)]
10 2 512

sin 16
+ (t − 1) sin 16(t − 1) h(t − 1).
32
Because of the (t − 1)-factors, we have undamped resonance. It is interesting to

plot a graph of this function. Figure 6.32 is a plot on the interval 0 ≤ t ≤ 4 (with
A = 2). It looks like damped oscillations as opposed to resonance. Figure 6.33,
with a longer time interval, shows the resonance.•
0.1
y y
0.1
2 4 x 4 x
-0.1
-0.1

The delayed sinusoidal nonhomogeneity presented no problem in Examples 6.30
and 6.31. When the nonhomogeneity is periodic, but not sinusoidal, additional
difficulties arise. Compared to a solution by methods of Chapter 4, however, Laplace
transforms are vastly superior. We illustrate in the following two examples.
366 SECTION 6.4

x00 + 4x = f (t), x(0) = 0, x0 (0) = 0,
where f (t) is the periodic function

1, 0 < t < π/2
f (t) = f (t + π) = f (t).
0, π/2 < t < π
Solution When we take Laplace transforms of both sides of the differential equa-
tion, and use property 6.16 for the transform of a periodic function, we obtain
Z π/2
2 1 1
s X + 4X = L{f (t)} = e−st dt = L{h(t) − h(t − π/2)}
1 − e−πs 0 1 − e−πs

1 1 e−πs/2 1
= −πs/2 −πs/2
− = .
(1 + e )(1 − e ) s s s(1 + e−πs/2 )
Thus,
1
X(s) = .
s(s2 + 4)(1 + e−πs/2 )
We now partial fraction 1/[s(s2 + 4)], and expand 1/(1 + e−πs/2 ) in a geometric
series

1 1 s
X(s) = − 2 1 − e−πs/2 + e−πs − e−3πs/2 + · · · .
4 s s +4
Each term in the series has an easily calculated inverse transform,
1 1 1
x(t) = (1 − cos 2t) − [1 − cos 2(t − π/2)]h(t − π/2) + [1 − cos 2(t − π)]h(t − π) − · · · .
4 4 4
In sigma notation,
∞
1X
x(t) = (−1)n [1 − cos 2(t − nπ/2)] h(t − nπ/2).
4 n=0
To evaluate x(t) for any given t, it is necessary to include only those terms in the
series for which nπ/2 < t. For example, the solution at t = 3.4 is given by
1 1 1
x(3.4) = [1 − cos 2(3.4)] − [1 − cos 2(3.4 − π/2)] + [1 − cos 2(3.4 − π)] = −0.402.•
4 4 4
Two points are noteworthy in this example. First, consider using the techniques
of Chapter 4 to find x(3.4). We would solve the differential equation on the intervals
0 < t < π/2, π/2 < t < π, π < t < 3π/2, match at t = π/2 and t = π, and then
find x(3.4) from the solution for π < t < 3π/2. Try it. You will be convinced that
Laplace transforms are superior. Secondly, recall that after Example 5.7 in Section
5.3, we questioned whether non-sinusoidal, periodic forces could produce resonance
in vibrating mass-spring systems. If we interpret this problem as describing oscilla-
tions x(t) of a 1 kilogram mass on the end of a spring with constant 4 newtons per
metre, there is resonance, but it is not obvious. If we write out the first few terms
of the series for x(t), we obtain
SECTION 6.4 367
1
x(t) = (1 − cos 2t)h(t) − [1 − cos 2(t − π/2)]h(t − π/2) + [1 − cos 2(t − π)]h(t − π)
4
+ [1 − cos 2(t − 3π/2)]h(t − 3π/2) + [1 − cos 2(t − 2π)]h(t − 2π)
+ [1 − cos 2(t − 5π/2)]h(t − 5π/2) + [1 − cos 2(t − 3π)]h(t − 3π) + · · ·
1
= 1 − cos 2t − (1 + cos 2t)h(t − π/2) + (1 − cos 2t)h(t − π)
4
− (1 + cos 2t)h(t − 3π/2) + (1 − cos 2t)h(t − 2π)

− (1 + cos 2t)h(t − 5π/2) + (1 − cos 2t)h(t − 3π) · · ·

 1 − cos 2t, 0 ≤ t < π/2



 −2 cos 2t, π/2 ≤ t < π

 x
1  1 − 3 cos 2t, π ≤ t < 3π/2 5
= −4 cos 2t, 3π/2 ≤ t < 2π
4

 1 − 5 cos 2t, 2π ≤ t < 5π/2



 −6 cos 2t, 5π/2 ≤ t < 3π 2 4 6 8 t
etc.
-5
Amplitudes of these oscillations
increase with time. We have shown a Figure 6.34
graph of the function in Figure 6.34.
Resonance has occurred because the natural frequencey of the mass-spring system
is 1/π, and this is the frequency of the applied force f (t).
The following example is fascinating. It may defy your intuition at first, but
when we reason it out, it makes perfect sense.
Example 6.33 A mass M hangs motionless from a spring
with constant k. At time t = 0, the mass
is acted upon by the periodic force in Figure
6.35. The force is a constant value F for p F
seconds, then it is turned off for p seconds,
back on for p seconds, and so on. During
p 2p 3p 4p t
its subsequent motion, the mass experiences
no damping. Find displacements of the mass p Figure 6.35
Discuss these displacements when p = 2π M/k.
Solution The initial-value problem for displacements of the mass is
d2 x
M + kx = f (t), x(0) = 0. x0 (0) = 0,
dt2
where f (t) is the function in Figure 6.35. Its representation in terms of Heaviside
functions is
f (t) = F [h(t) − h(t − p)], 0 < t < 2p, f (t + 2p) = f (t).
When we take Laplace transforms,
1 1
M s2 X + kX = L{f (t)} = −2ps
L{f1 (t)]} = L{F [1 − h(t − p)]}
1−e 1 − e−2ps

F 1 e−ps F (1 − e−ps ) F
= −2ps
− = −ps −ps
= .
1−e s s s(1 + e )(1 − e ) s(1 + e−ps )
368 SECTION 6.4
Hence,
F
X(s) = .
s(M s2 + k)(1 + e−ps )
With the partial fraction decomposition of 1/[s(M s2 + k)], and geometric series for
1/(1 + e−ps ), the inverse transform of X(s) is

−1 1/k M s/k 1
x(t) = F L −
s M s2 + k 1 + e−ps
( X ∞
)
F −1 1 s n −pns
= L − (−1) e
k s s2 + k/M n=0
∞
" r #
F X k
= (−1)n 1 − cos (t − pn) h(t − pn).
k n=0 M
p
When p = 2π M/k, displacements become
∞
F X n 2π
x(t) = (−1) 1 − cos (t − pn) h(t − pn)
k n=0 p
∞
F X n 2πt
= (−1) 1 − cos h(t − np)
k n=0 p
∞
F 2πt X
= 1 − cos (−1)n h(t − np).
k p n=0
To graph this function, we write the summation out,

F 2πt
x(t) = 1 − cos [h(t) − h(t − p) + h(t − 2p) − · · ·],
k p
in which case we see that


 F 2πt

 1 − cos , 0≤t<p

 k p


 0, p ≤ t < 2p
x(t) = F 2πt


 1 − cos , 2p ≤ t < 3p

 k p




 0, 3p ≤ t < 4p
etc.
A graph of this function is shown in x
Figure 6.36. Here is an explanation 2 F/k
of why this happens. During the inter-

val 0 ≤ t ≤ p/2, the force F moves the
mass upward against the spring, and
for p/2 < t ≤ p, the spring returns the
p 2p 3p t
mass against F to the equilibrium posi-
tion. When the mass reaches equilibrium, Figure 6.36
its velocity is zero. Since the force now
becomes zero, the mass remains at equilibrium for time p. This sequence of motion
then repeats itself in each interval of length 2p thereafter.•
SECTION 6.4 369
EXERCISES 6.4
1. Solve the initial-value problem
dy
+ 3y = f (t), y(0) = 1,
dt

t, 0 < t < 1
where by:
1, t > 1.
(a) the techniques of Chapter 4,
(b) Laplace transforms.
In Exercises 2–12 solve the initial-value problem.

00 0 0, 0 < t < 4
2. y + 9y = f (t), y(0) = 1, y (0) = 2, where f (t) =
1, t > 4

00 0 2, 0 < t < 4
3. y + 9y = f (t), y(0) = 1, y (0) = 2, where f (t) =
0, t > 4

t, 0 < t < 1
4. y 00 + 4y 0 + 4y = f (t), y(0) = 0, y 0 (0) = −1, where f (t) =
1, t > 1

2 − t, 0 < t < 2
5. y 00 + 4y 0 + 4y = f (t), y(0) = −1, y 0 (0) = 0, where f (t) =
t − 2, t > 2

0, 0<t<π
6. y 00 + 4y 0 + 3y = f (t), y(0) = 1, y 0 (0) = 2, where f (t) =
sin t, t > π

sin t, 0 < t < π
7. y 00 + 4y 0 + 3y = f (t), y(0) = 1, y 0 (0) = 2, where f (t) =
0, t>π

3, 0<t<1
8. y 00 + 2y 0 + 5y = f (t), y(0) = 0, y 0 (0) = 0, where f (t) =
−3, t > 1
( 4, 0<t<1
9. y 00 + 2y 0 + 5y = f (t), y(0) = 0, y 0 (0) = 0, where f (t) = −4, 1 < t < 2
0, t>2

00 0 t, 0 < t < 1
10. y +16y = f (t), y(0) = 2, y (0) = 0, where f (t) = f (t+2) = f (t)
0, 1 < t < 2

t, 0<t<1
11. y 00 + 16y = f (t), y(0) = 2, y 0 (0) = 0, where f (t) = f (t + 2) = f (t)
2 − t, 1<t<2

00 0 0 1, 0<t<1
12. y + y = f (t), y(0) = 0, y (0) = 0, where f (t) = f (t + 2) = f (t)
−1, 1<t<2
13. Use Laplace transforms to solve the mixing problem of Example 3.11 in Section 3.4.
14. Can you use Laplace transforms to solve the mixing problem in Example 3.12 of Section 3.4?
15. Solve Example 3.11 in Section 3.4 if after 10 minutes the concentration of the brine being added
to the tank changes to 1 kilogram per 100 litres.
16. Use Laplace transforms to solve Exercise 15 in Section 3.4.
370 SECTION 6.4
17. Find the amount of salt in the tank of Example 3.11 in Section 3.4 if the brine mixture is
added for 2 minutes, replaced by pure water for 2 minutes, replaced by the brine mixture for
2 minutes, etc.
18. When a patient is admitted to the hospital, the amount of glucose in his bloodstream is g0
grams. He is immediately put on intravenous which transfers glucose to his bloodstream at a
rate of R grams per minute. At any given time, his body uses the glucose up a rate prportional
to how much is present in the bloodstream at that time. If he remains on intravenous for 4
hours and then the intravenous is discontinued, how much glucose is in the bloodstream 6 hours
after he is admitted?
19. Find the amount of glucose in the bloodstream of the patient in Exercise 18 as a function of
time (in minutes) after he is admitted to the hospital if glucose is administered for an hour,
turned off for an hour, turned on for an hour, turned off for an hour, etc.
20. The initial mass of a certain species of fish in a lake is estimated as m0 kilograms. Suppose
that left alone, the fish would increase their mass at a rate described by the Malthusian model
3.12. Commercial fishing harvests (removes) H kilograms each year, at a uniform rate, but
only in the first month of the year.
(a) Find the mass m(t) of fish in the lake as a function of time t.
(b) Determine the value of H in order that the mass of fish in the lake return to m0 after one
year.
21. A 100-gram mass is suspended from a spring with constant 40 newtons per metre. The mass
is pulled 10 centimetres above its equilibrium position and given velocity 2 metres per second
downward. If a force of 100 newtons acts vertically upward for the first 4 seconds, find the
position of the mass as a function of time. Ignore all damping.
22. Repeat Exercise 21 if the force is turned on after 4 seconds.
23. Repeat Exercise 21 if a damping force with constant β = 5 also acts on the mass.
p
27. Repeat Example 6.33 if p = 4π M/k.
p
28. Repeat Example 6.33 if p = 3π M/k.
29. A 1 kilogram mass is motionless at the
end of a spring with constant 16 new-
tons per metre. When the ramp force
in the figure to the right (units of new- 20
tons) acts the mass, the mass moves
vertically without damping. Find its
subsequent displacements. Plot a graph
of the displacement function. 1 2 t
SECTION 6.5 371
6.5 The Dirac Delta Function and its Applications

There are many common situations that cannot be represented mathematically by
functions as we know them. For instance, consider:
1. suddenly adding a sizeable quantity of dissolving substance in a mixing problem
in Section 3.4
2. striking a mass on the end of a spring with a hammer in Section 5.3
3. accommodating a voltage spike in an electric circuit in Section 5.4
4. representing a concentrated load on a beam in Section 5.5
We introduce the “Dirac delta” function
in this section in order to model these
situations mathematically. We begin with
1
what are called unit pulse functions. The a
unit pulse at time t = t0 of duration a is
the function in Figure 6.37. It can be rep-
resented in terms of Heaviside unit step
t0 t0 + a t
functions as
Figure 6.37
1
p(t0 , a, t) = [h(t − t0 ) − h(t − t0 − a)]. (6.26)
a
Value t0 identifies the time at which the pulse begins, and a is the duration of the
pulse. What is important to notice is that the area under the curve is one; hence
the name unit pulse. The units of p(t0 , a, t) are inverses of the units of t. If t is
time in seconds (s), then units of p(t0 , a, t) are s−1 . The Laplace transform of this
function is
1 h −t0 s i
L{p(t0 , a, t)} = e − e−(t0 +a)s . (6.27)
as
The Unit Impulse
Even more important than the 80
unit pulse is the unit impulse. It is
70
the limit of the unit pulse p(t0 , a, t)
as the time interval t0 < t < t0 + a 60
becomes indefinitely short. As a gets
smaller and smaller in Figure 6.37, 50
the area under the curve remains 40
unity; the function is nonzero over
shorter and shorter time intervals, 30
but the height of the function gets
20
larger and larger. We have shown
the situation for a = 1/10, 1/20, 10
1/40, and 1/80 in Figure 6.38.
t0 t0 +0.05 t0 +0.1 t
t0 +0.0125 t0 +0.025
Figure 6.38
372 SECTION 6.5
The limit of this function as a → 0 is not a function in the normal sense of function.
It has value 0 for all t except t = t0 where its value is “infinite”. Such functions
are discussed in advanced mathematics; they are known as generalized functions or
distributions. This particular one is called the unit impulse or the Dirac delta
function. It is denoted by
1
δ(t − t0 ) = lim [h(t − t0 ) − h(t − t0 − a)]. (6.28)
a→0 a
The Dirac delta function can be defined as the limit of other sequences of
functions besides those in Figure 6.38; all lead to identical properties. Two such
sequences are shown in Figures 6.39 and 6.40. In both cases, the area under each
curve is unity. Like the functions in Figure 6.38, those in Figure 6.39 are discontinu-
ous, but they are symmetric around t0 . The functions in Figure 6.40 are continuous
and symmetric around t0 .
80
80
70
70
60
60
50
50
40
40
30
30
20
20
10
10
t0 -0.05 t0 t0 +0.05 t
t0 -0.025 t0 +0.025
t0 -0.025 t0 t0 +0.025 t t0 -0.0125 t0 +0.0125
t0 -0.0125 t0 +0.0125
t0 -0.05625 t0 +0.05625
The Dirac delta function does not conform to the conditions of Theorem 6.1;
it is of exponential order, but it is not piecewise-continuous on every finite interval.
It does, however, have a Laplace transform, and we could write
n o
L{δ(t − t0 )} = L lim p(t0 , a, t) .
a→0
We define the Laplace transform of δ(t − t0 ) by interchanging the operations of

taking the limit and taking the Laplace transform; that is, we define
L{δ(t − t0 )} = lim L{p(t0 , a, t)}. (6.29)
a→0
Substituting from equation 6.27 and using L’Hopital’s rule on the limit gives
−t0 s −(t0 +a)s
e − e−(t0 +a)s se
L{δ(t − t0 )} = lim = lim = e−t0 s . (6.30)
a→0 as a→0 s
For readers who find this definition unsatisfactory, an alternative is contained

in the Appendix at the end of this section. It briefly discusses how “generalized
functions” such as the delta function should be manipulated.
SECTION 6.5 373
Notice that when t0 = 0, the Laplace transform of δ(t) is

L{δ(t)} = 1.
In Section 6.1, we proved that the limit as s → ∞ of the Laplace transform of every
piecewise continuous function of exponential order is equal to zero. The limit of the
Laplace transform of δ(t) is not zero as s → ∞, but it is not a piecewise-continuous
function.
We will give arguments to demonstrate that the delta function can be used
to model the situations described at the beginning of this section, situations that
cannot be modelled by ordinary functions. Important to realize is the units of
δ(t − t0 ). Since the delta function is the limit of the unit pulse function, and taking
limits does not alter units, units of the delta function are those of the pulse function,
namely, 1 over the units of t. Hence, if t is time in seconds (s), then units of δ(t−t0 )
are s−1 .
We begin with a mixing problem.
Mixing Problems and the Delta Function
A tank contains 1000 litres of water in which 5 kilograms of salt have been
dissolved. A brine mixture with concentration 2 kilograms of salt for each 100 litres
of water is added to the tank at 5 millilitres per second. At the same time, mixture
is being drawn from the bottom of the tank at 5 millilitres per second. Suppose
that 3 kilograms of salt are suddenly added to the tank at the 5 minute mark, and
we wish to find the amount of salt in the tank as a function of time. As usual, we
assume that the mixture in the tank is always sufficiently well-stirred that at any
given time, concentration of salt is the same at all points in the tank, even when
the 3 kilograms is suddenly added to the tank. This amounts to assuming that the
3 kilograms of salt dissolve instantaneously.
If we let x(t) represent the number of grams of salt in the tank at any given
time, then x(0) = 5000. As in Section 3.4, we write symbolically that

dx rate at which rate at which
= − .
dt salt is added salt is removed
Since 5 mL of mixture enter the tank each second, and each millilitre contains 0.02
g of salt, it follows that salt is being added to the tank at a constant rate of 0.1 g/s.
The rate at which salt is removed from the tank is not constant; it changes as the
concentration of salt in the tank changes. Since the tank always contains 106 mL of
solution, the concentration of salt in the solution at time t, in grams per millilitre,
is x(t)/106 . With solution being drawn off at 5 mL/s, the rate at which salt leaves
the tank in grams per second is 5x/106 . The difficulty is how to model the sudden
insertion of 3 kilograms of salt into the tank at the 5 minute mark. We will show
that the delta function is the key. First, suppose we assume that the 3 kilograms of
salt are added to the tank uniformly over a time interval of a seconds beginning at
time t = 5. This means adding it a rate of 3000/a g/s over the a seconds. This can
be represented by
3000
[h(t − 300) − h(t − 300 − a)].
a
If we now insert these rates into the equation for dx/dt, we obtain
374 SECTION 6.5
dx 1 5x 3000
= − 6+ [h(t − 300) − h(t − 300 − a)].
dt 10 10 a
1 5X 3000 −300s
sX − 5000 = − + [e − e−(300+a)s ],
10s 106 as
from which
1 3000 −300s
5000 + + [e − e−(300+a)s ]
X(s) = 10s as
5
s+ 6
10
5000 105 1 1
= + −
s + 5/106 5 s s + 5/106

6 × 108 1 1
+ − [e−300s − e−(300+a)s ].
a s s + 5/106
Inverse transforms give

6 105 6
x(t) = 5000e−5t/10 + (1 − e−5t/10 )
5
h 6
i h 6
i 
 1 − e−5(t−300)/10 h(t − 300) − 1 − e−5(t−300−a)/10 h(t − 300 − a) 
+ (6 × 108 ) .
 a 
This is the number of grams of salt in the tank if the 3 kilograms of salt are added
at a constant rate over a seconds beginning at t = 300. In order to find the result
when the 3 kilograms of salt is added instantaneously, we use L’Hopital’s rule to
take the limit of this function as a → 0,
6 6
x(t) = 20 000 − 15, 000e−5t/10 + (6 × 108 ) lim [(5/106 )e−5(t−300−a)/10 h(t − 300 − a)]
a→0
−5t/106 −5(t−300)/106
= 20 000 − 15, 000e + 3000e h(t − 300).
We now show that when the delta function is used to model the instantaneous
addition of the 3 kilograms of salt, we get the same result with a fraction of the
work. Since the units of δ(t − t0 ) are seconds to the negative 1, 3000δ(t − 300) has
units of grams per second, consistent with other input and output rates. Suppose
we use this expression to represent the addition of the 3000 grams of salt at t = 300
seconds. The differential equation for x(t) becomes
dx 1 5x
= − 6 + 3000δ(t − 300).
dt 10 10
1 5X
sX − 5000 = − 6 + 3000e−300s ,
10s 10
from which
5000 1/(10s) + 3000e−300s
X(s) = + .
s + 5/106 s + 5/106
SECTION 6.5 375
Inverse transforms give

−5t/106 105 −1 1 1 6
x(t) = 5000e + L − + 3000e−5(t−300)/10 h(t − 300)
5 s s + 5/106
6 6
= 20 000 − 15 000e−5t/10 + 3000e−5(t−300)/10 h(t − 300).
We have shown therefore that the delta function representation of the instantaneous
addition of 3 kilograms of salt yields the same result as adding the salt over a small
interval of time and then taking the limit as this interval approaches zero, and it
does so with far less work. It is worthwhile noting that
6
lim x(t) = 20 000 − 15 000e−5(300)/10 ,
t→300−
6
lim x(t) = 20 000 − 15 000e−5(300)/10 + 3000.
t→300+
In other words, the amount of salt jumps by 3000 grams at t = 300 seconds, as we
should expect.
Displacements in Mass-Spring Systems and the Delta Function
We now consider the situation in which a pulse represents an external force
applied to the mass in a vibrating mass-spring system such as that in Figure 6.41.
When damping and surface friction are negligible, the differential equation describ-
ing the position of the mass relative to its equilibrium position is M d2 x/dt2 + kx =
f (t) where f (t) represents all forces on M other than the spring.
k
F
a
M
x t0 t0 + a t
Suppose the mass is at rest at its equilibrium position for t0 seconds starting at
time t = 0, and at time t0 , the mass is subjected to the pulse in Figure 6.42. The
magnitude of the force is F/a, but the area under the curve is F , and is so for any
length of duration a of the pulse. We call this a pulse of size F . The initial-value
problem for displacements of the mass due to this pulse is
d2 x
M + kx = F p(t0 , a, t), x(0) = 0, x0 (0) = 0. (6.31)
dt2
If we take Laplace transforms of both sides of the differential equation, and use
formula 6.27,
F h −t0 s i F [e−t0 s − e−(t0 +a)s ]
M s2 X + kX = e − e−(t0 +a)s =⇒ X(s) = .
as as(M s2 + k)
Partial fractions give
376 SECTION 6.5

F 1 s
X(s) = − 2 (e−t0 s − e−(t0 +a)s ),
ka s s + k/M
from which
" r #
F k
x(t) = 1 − cos (t − t0 ) h(t − t0 )
ka M
" r #
F k
− 1 − cos (t − t0 − a) h(t − t0 − a). (6.32)
ka M
A typical graph of this function is

shown in Figure 6.43. The mass is 2F/ (ka)
at equilibrium for 0 ≤ t < t0 . At

time t0 , the force causes the mass
to move. While the force acts
(t0 < t < t0 + a), the mass exper-
iences simple harmonic motion with t0 t 0 +a t
amplitude F/(ka),
Figure 6.43
" r #
F k
x(t) = 1 − cos (t − t0 ) . (6.33a)
ka M
After the force is removed at t = t0 + a, the displacement of the mass is

" r r #
F k k
x(t) = cos (t − t0 − a) − cos (t − t0 ) . (6.33b)
ka M M
This is once again simple harmonic motion, but with a different amplitude. For some
values of the parameters, the amplitude may be enhanced and for other values, it
may be diminished (see Exercise 15). The derivative of this function is the velocity
of the mass. Since the derivative of the Heaviside function is zero, except at t0
where it is undefined, the velocity is
"r r #
F k k
v(t) = sin (t − t0 ) h(t − t0 )
ka M M
"r r #
F k k
− sin (t − t0 − a) h(t − t0 − a)
ka M M
" r r #
F k k
= √ sin (t − t0 )h(t − t0 ) − sin (t − t0 − a)h(t − t0 − a) .
a kM M M
Left and right limits of this function are the same at t0 and t0 + a so that there is
no abrupt change in velocity at t0 and t0 + a where the pulse is discontinuous. This
is reflected by the smoothness of the graph in Figure 6.43; the slope of the tangent
line is continuous at t0 and t0 + a. This is consistent with the fact that the solution
of the differential equation, and its first derivative, must be continuous even with a
piecewise-continuous nonhomogeneity (Theorem 6.9).
SECTION 6.5 377
Suppose we use L’Hopital’s rule to take the limit of this displacement function
as the duration time of the pulse becomes indefinitely small,
q q
k k
1 − cos M (t − t0 ) h(t − t0 ) − 1 − cos M (t − t0 − a) h(t − t0 − a)
F
x(t) = lim
k a→0+ a
q q
k k
M
sin M (t − t0 − a) h(t − t0 − a)
F
= lim
k a→0+ 1
r
F k
= √ sin (t − t0 )h(t − t0 ). (6.34)
kM M
This is the displacement of the mass at time t due to a pulse of size F as the duration
of the pulse becomes indefinitely short. We regard this displacement as that due to
an instantaneously applied force at
time t0 . A typical graph of this
,
function is shown in Figure 6.44. F kM
The mass is at equilibrium for
0 ≤ t < t0 . At time t0 , the
instantaneous force imparts a t0 t
velocity to the mass. To find
this velocity, we differentiate
the displacement with respect
to t, Figure 6.44
r
F k
v(t) = cos (t − t0 )h(t − t0 ).
M M
This function is discontinuous at t0 , and its limit as t → t+ 0 is F/M . In other
words, the instantaneous force has given the mass an initial velocity F/M metres
per second. This is reflected in the corner of the curve in Figure 6.44 at t = t0 , the
slope of the graph jumps from zero to F/M .
Our hope is that the Dirac delta function will be the mathematical represen-
tation of an instantaneously applied force (such as that in the above mass-spring
system, perhaps the result of hitting the mass with a hammer). Consider again,
then, the mass-spring system in Figure 6.41. We showed that the position of the
mass due to a pulse of magnitude F applied to the mass at time t0 over a time
interval of length a is given by equation 6.32. We took the limit of this function as
a → 0 to obtain displacement 6.34 due to an instantaneously applied force.
Suppose we replace the pulse function in equation 6.31 with F δ(t − t0 ), so that
the initial-value problem for displacements is
d2 x
M + kx = F δ(t − t0 ), x(0) = 0, x0 (0) = 0. (6.35)
dt2
When we take Laplace transforms, we get
F e−t0 s F e−t0 s
M [s2 X] + kX = F e−t0 s =⇒ X(s) = = .
M s2 + k M s2 + k/M
The inverse transform gives
378 SECTION 6.5
r r r
F M k F k
x(t) = sin (t − t0 )h(t − t0 ) = √ sin (t − t0 )h(t − t0 ).
M k M kM M
This is solution 6.34. We have shown then that the problem of finding displacements
for a vibrating mass on the end of a spring, subjected to a pulse of magnitude F ,
and taking the limit as the interval of application of the pulse approaches zero, can
be handled much more efficiently with the Dirac delta function. This is a partial
justification for the use of the Dirac delta function to represent instantaneously
applied forces to masses in mass-spring systems.
We call F δ(t − t0 ) an impulse force of size F at t0 . Although an impulse
force of size F can be thought of as the limit of a pulse of size F as the duration of
the pulse approaches zero, there is a better way to view it. As a force, F δ(t − t0 )
has units of newtons or kilogram-metres per second squared. Since the units of the
delta function are seconds to the power negative one, it follows that F itself is not a
force; it does not have units of newtons. It has units of kilogram-metres per second;,
units of momentum. In other words, when an instantaneously applied force in the
form F δ(t − t0 ) acts on a mass, it imparts F units of momentum to the mass. This
is consistent with what we saw earlier. Application of an impulse force F δ(t − t0 )
to a mass M results in a velocity change of F/M . This means a change of F in
momentum. Thus, instead of saying that a vibrating mass is struck with an impulse
force of size F , it is more informative to say that it is struck with a force that gives
the mass F units of momentum.
Here is another example, one that includes damping.
Example 6.34 A 100-gram mass is suspended from a spring with constant 50 newtons per metre.
It is set into motion by raising it 10 centimetres above its equilibrium position and
giving it a velocity of 1 metre per second downward. During the subsequent motion
a damping force acts on the mass and the magnitude of this force is one-fifth the
velocity of the mass. If an impulse force that imparts two units of momentum to
the mass is applied vertically upward to the mass at t = 3 seconds, find the position
of the mass for all time.
Solution The initial-value problem for the position of the mass is
1 d2 x 1 dx 1
2
+ + 50x = 2δ(t − 3), x(0) = , x0 (0) = −1.
10 dt 5 dt 10
If we multiply the differential equation by 10, and take Laplace transforms,

2 s 1
s X− + 1 + 2 sX − + 500X = 20e−3s .
10 10
Thus,
s/10 − 4/5 20e−3s
X(s) = +
s2 + 2s + 500 s2 + 2s + 500

1 (s + 1) − 9 20e−3s
= + .
10 (s + 1)2 + 499 (s + 1)2 + 499
The inverse transform is

1 −t −1 s−9 −1 20e−3s
x(t) = e L 2
+L .
10 s + 499 (s + 1)2 + 499
SECTION 6.5 379
Since

−1 20 −t −1 1 20 −t √
L = 20e L = √ e sin 499t,
(s + 1)2 + 499 2
s + 499 499
it follows that

1 −t √ 9 √ 20 −(t−3) √
x(t) = e cos 499t − √ sin 499t + √ e sin 499(t − 3) h(t − 3).
10 499 499
It is straightforward to show that this
solution satisfies the initial conditions
x(0) = 1/10 and x0 (0) = −1. A graph x
0.8
of the function is shown in Figure 6.45.
Due to excessive damping, oscillations 0.4
essentially disappear after 3 seconds,

1 2 3 6 t
but the impulse force restores them at
-0.4
t = 3 seconds. Notice the abrupt change
in slope (velocity) at t = 3 due to the -0.8
impulse force. Damping again brings the Figure 6.45

mass essentially to rest after a few seconds.•
We have already seen that when the nonhomogeneity in an nth -order, constant
coefficient, linear differential equation is continuous, the solution and its first n
derivatives are all continuous. When the nonhomogeneity is piecewise-continuous,
the solution and its first n − 1 derivatives are continuous, but the nth derivative has
discontinuities at the discontinuities of the nonhomogeneity. Based upon the above
examples, we can expect that when a nonhomogeneity is a Dirac delta function,
both the nth and the (n − 1)th derivatives of the solution will have discontinuities.
Transfer Functions and Impulse Response Functions

Consider the initial-value problem consisting of the nth -order constant coefficient
differential equation
dn y dn−1 y dy
an + an−1 + · · · + a1 + a0 y = f (t), (6.36a)
dtn dtn−1 dt
with initial conditions
(n−1)
y(0) = y0 , y 0 (0) = y00 , ... , y (n−1) (0) = y0 . (6.36b)
If we take Laplace transforms, we get

(n−1)
an [sn Y − sn−1 y0 − sn−2 y00 − · · · − y0 ]
+ an−1 [s n−1
Y −s n−2
y0 − · · · − y0n−2 ] + · · · + a0 Y = F (s).
When we solve this for Y (s), we can write the result in the form
Q(s) F (s)
Y (s) = + , (6.37)
P (s) P (s)
where P (s) = an sn + an−1 sn−1 + · · · + a0 , and Q(s) is a polynomial of order less

than or equal to n − 1 determined by the coefficients ai and the initial conditions.
380 SECTION 6.5
The thing to notice about this transform is that if f (t) ≡ 0, then so also is F (s),
and Y (s) = Q(s)/P (s). The inverse transform of this Y (s) can be thought of in
two ways. First, if we regard the n initial values as unspecified, then the inverse is
a general solution yh (t) of the homogeneous initial-value problem associated with
6.36. The initial values are the arbitrary constants in the solution. Alternatively, if
the initial values are regarded as specified constants, then the inverse of Y (s) is the
solution of the homogeneous equation subject to these conditions. Suppose instead
that all initial conditions are equal to zero. Then Q(s) ≡ 0, and Y (s) = F (s)/P (s).
The inverse transform of this is a particular solution yp (t) of differential equation
(n−1)
6.36 that satisfies yp (0) = yp0 (0) = · · · yp = 0. Expression 6.37 has separated
effects of the initial conditions and the nonhomogeneity into separate terms. Let us
illustrate these ideas with an example before proceeding with the discussion.
It is put into motion by lifting it 10 centimetres above its equilibrium position and
giving it velocity 2 metres per second downward. During its motion, it is subject to
a damping force equal to twice the velocity of the mass, and a vertical force 3 sin 4t.
Find expression 6.37 for the problem, and inverse transforms of each term.
Solution The initial-value problem for the motion of the mass is
d2 x dx
+2 + 65x = 3 sin 4t, x(0) = 1/10, x0 (0) = −2.
dt2 dt
If we take Lapace transforms,

2 s 1 12
s X− + 2 + 2 sX − + 65X = 2 .
10 10 s + 16
When we solve for X(s),
s 9
− 12
X(s) = 2 10 5 + 2 .
s + 2s + 65 (s + 16)(s2 + 2s + 65)
The first term contains the effect of the initial conditions, and the second term
contains the transform of the nonhomogeneity. Inverse transforms of these terms
are

−1 s/10 − 9/5 1 −1 (s + 1) − 19
xh (t) = L = L
s2 + 2s + 65 10 (s + 1)2 + 64

1 −t −1 s − 19 1 −t 19
= e L = e cos 8t − sin 8t ,
10 s2 + 64 10 8

12
xp (t) = L−1
(s + 16)(s2 + 2s + 65)
2

12 −1 −2s + 49 2s − 45
= L + 2
2465 s2 + 16 s + 2s + 65

12 49 −1 2(s + 1) − 47
= −2 cos 4t + sin 4t + L
2465 4 (s + 1)2 + 64

12 49 −t 47
= −2 cos 4t + sin 4t + e 2 cos 8t − sin 8t .
2465 4 8
SECTION 6.5 381
It is straightforward to check that xh (t) satisfies the homogeneous differential equa-

tion x00 + 2x0 + 65x = 0 and the initial conditions xh (0) = 1/10 and x0h (0) = −2.
Function xp (t) satisfies the nonhomogeneous equation x00 + 2x0 + 65x = 3 sin 4t, and
the initial conditions xp (0) = x0p (0) = 0.•
We now return to discussion of expression 6.37. If we write differential equation
6.36a in operator notation φ(D)y = f (t), then the function P (s) in expression 6.37 is
φ(s). The function H(s) = 1/P (s) is called the transfer function for system 6.36.
Its inverse transform h̄(t) is called the unit impulse response function, because
it describes the solution of the initial-value problem when all initial conditions are
zero and the nonhomogeneity is the delta function δ(t). This is easily seen by noting
that when all initial conditions are zero and f (t) = δ(t), then in expression 6.37,
Q(s) = 0, F (s) = 1, and therefore
1
Y (s) = = H(s).
P (s)
Consequently, y(t) = h̄(t). We can write the solution of initial-value problem 6.36
in terms of the unit impulse response function. We have already noted that the
inverse transform of the first term Q(s)/P (s) in expression 6.37 is yh (t), the general
solution of the associated homogeneous problem. The inverse of the second term
F (s)/P (s) can be written as the convolution of f (t) and h̄(t),
Z t
−1 −1
L {F (s)/P (s)} = L {F (s)H(s)} = f (u)h̄(t − u) du.
0
Thus, the solution of initial-value problem 6.36 is

Z t
y(t) = yh (t) + f (u)h̄(t − u) du. (6.38)
0
The first term contains contributions to the solution of the initial-value probem
due to the initial conditions, and the second term is the contribution due to the
nonhomogeneity in the differential equation.
Example 6.36 Use formula 6.38 to find the solution to the initial-value problem in Example 6.35.
Solution Since the auxiliary equation m2 +2m+65 = 0 has solutions m = −1±8i,
a general solution of the associated homogeneous equation is
xh (t) = e−t (C1 cos 8t + C2 sin 8t).
The initial conditions require 1/10 = x(0) = C1 , and −2 = x0 (0) = −C1 + 8C2 ,
and therefore the solution of the associated homogeneous differential equation that
satisfies the initial conditions is

1 19 1 −t
e−t cos 8t − sin 8t = e (8 cos 8t − 19 sin 8t).
10 80 80
The unit impulse response function is

−1 1 −1 1 1
L 2
=L 2
= e−t sin 8t.
s + 2s + 65 (s + 1) + 64 8
If we denote the second term in formula 6.38 by xp (t), then
382 SECTION 6.5
Z t
1 −(t−u)
xp (t) = 3 sin 4u e sin 8(t − u) du.
0 8
Lengthy integration leads to
147 24 24 −t 141 −t
xp (t) = sin 4t − cos 4t + e cos 8t − e sin 8t.
2465 2465 2465 4930
Thus, the solution of the initial-value problem is
1 −t
x(t) = e (8 cos 8t − 19 sin 8t)
80
147 24 24 −t 141 −t
+ sin 4t − cos 4t + e cos 8t − e sin 8t.•
2465 2465 2465 4930
EXERCISES 6.5
1. A tank contains 1000 litres of water in which 5 kilograms of salt have been dissolved. Starting
at time t = 0, pure water is added to the tank at 5 millilitres per second. At the same time,
mixture starts being drawn from the bottom of the tank at 5 millilitres per second. After 1
minute, 500 grams of salt are suddenly added to the tank, and an additional 500 grams every
minute thereafter. Assume that the mixture in the tank is always sufficiently well-stirred that
at any given time, concentration of salt is the same at all points in the tank, even when the 500
grams are added to the tank. Find the amount of salt in the tank as a function of time.
2. Repeat Exercise 1 if a brime mixture containing 2 kilograms of salt per 100 litres of water is
added to the tank instead of pure water.
3. A 2-kilogram mass is suspended from a spring with constant 512 newtons per metre. If it is set
into motion at time t = 0 by a unit impulse force, find its subsequent displacement. Assume
negligble damping.
4. Repeat Exercise 3 if damping with constant β = 80 is taken into account.
5. Repeat Exercise 4 if β = 8.
6. A 2-kilogram mass is suspended from a spring with constant 512 newtons per metre. It is set
into motion at time t = 0 by moving it to position x0 and then releasing it. If a unit impulse
force is applied at t0 > 0, find the position of the mass for all time.
7. Repeat Exercise 6 if motion is initiated by giving the mass velocity v0 at position x = 0.
8. Repeat Exercise 6 if motion is initiated by giving the mass velocity v0 from position x0 .
9. A 1-kilogram mass is suspended from a spring with constant 100 newtons per metre. It is
subjected to a unit impulse force at t = 0 and again at t = 1. Find the position of the mass as
a function of time.
10. A mass of M kilograms hangs at equilibrium on the end of a spring with constant k newtons
per metre. At time t = 0, the mass is subjected to a unit impulse force. Show that if the
initial-value problem
d2 x
M + kx = δ(t), x(0) = 0, x0 (0) = 0,
dt2
is solved, the function does not satisfy the initial velocity condition. Can you explain why?
SECTION 6.5 383
11. Repeat Exercise 9 if unit impulse forces are applied one each second beginning at time t = 0.
Express the solution in sigma notation.
12. Repeat Exercise 11 if unit impulse forces are π/5 seconds apart, the first at time t = 0. Is there
resonance?
13. A mass M , suspended from a spring with constant k, is set into undamped motion by giving
it displacement x0 from its equilibrium position and velocity v0 . At time t0 , it is struck with
an impulse force F . Find an expression for F if the impulse force brings the mass to an
instantaneous stop.
14. An unstretched spring (with constant k), in the horizontal position, is attached on the left to
a wall and on the right to a mass M . At time t = 0, the mass is struck with an impulse force
F to the right which causes the mass to move. The coefficient of kinetic friction between the
mass and the table along which it slides is µ. Assume that damping can be ignored.
(a) What is the initial velocity of the mass as a result of the hit?
(b) When does the mass come to a stop for the first time?
(c) If the coefficient of static friction between the mass and the table is µs , show that the mass
moves to the left if
r
g 2 µs M 3 (µs + 2µ)
F > .
k
15. Show that the amplitude of displacement function 6.33b is
r
2F k a
sin .
ka M2
Convince yourself that for certain values of a, this could be greater than F/(ka) and for other
values of a, it could be less than F/(ka).
16. (a) Suppose the equation of the speed bump in Exercise 23 of Section 5.2 is f (x) = 3x(1 − x)/5.
Find displacement of the front end of the car when M = 200 and k = 1000. Assume β = 0
so that shock absorbers are not working, and that the car has slowed down to 20 kilometres
per hour.
(b) Show that motion of the car after the speed bump is simple harmonic and find its amplitude.
In Exercises 17–18 find the transfer and unit impulse response functions for the initial-
value problem. Express the solution in form 6.38.
d2 y dy
17. −3 − 4y = f (t), y(0) = 1, y 0 (0) = −2
dt2 dt
d2 y dy
18. 2
+2 + 3y = f (t), y(0) = A, y 0 (0) = B
dt dt
19. A system modelled by differential equation 6.36a is said to be stable if its unit impulse response
function h̄(t) is bounded as t → ∞. Show that this is the case if, and only if, real roots of
P (s) = 0 are less than or equal to zero, as are real parts of complex roots.
20. A system modelled by differential equation 6.36a is said to be asymptotically stable if its
unit impulse response function h̄(t) approaches zero as t → ∞. Show that this is the case if,
and only if, real roots of P (s) = 0 are negative, as are real parts of complex roots.
384 SECTION 6.5
Appendix
In the theory of generalized functions, or distributions, functions such as the delta
function δ(t − t0 ) are never stand-alone functions; that is, they are not assigned
values at specific values of t. It is easy to see why the delta function δ(t − t0 )
should not be considered in a pointwise sense. If it has value zero for t 6= t0 , and
value “infinity” at t0 , then what is the difference between δ(t − t0 ) and 4δ(t − t0 )?
Generalized functions are considered to be mappings, or transformations, that map
ordinary functions to numbers. For instance, suppose that g(t) is a fixed function
that is continuous on the interval a ≤ t ≤ b. If f (t) is any other function that is
integrable on a ≤ t ≤ b, then the integral
Z b
f (t)g(t) dt
a
is a real number. We can say that through this integration, g(t) defines a mapping
from the set of integrable functions on a ≤ t ≤ b to the reals. If we denote this
mapping by G, then we can write that
g(t)
Z b
f (t) −−−−−→ G{f } = g(t)f (t) dt. (6.39)
a
For instance, if g(t) = t2 , and the interval is 0 ≤ t ≤ 2, then

Z 2 2
t2
2 t4
t −−−−−→ G{t} = t(t ) dt = = 4,
0 4 0
Z 2
t
t2
t
2
e −−−−−→ G{e } = t2 et dt = t2 et − 2tet + 2et 0
= 2(e2 − 1).
0
It is this view of an ordinary function as a mapping, or operator, that is adopted

to define δ(t − t0 ). The “generalized” function δ(t − t0 ) is the operator that maps
a function f (t), continuous at t = t0 , onto its value at t = t0 ,
δ(t−t0 )
f (t) −−−−−→ f (t0 ). (6.40)
For example,
δ(t−2) δ(t)
t2 + 2t − 3 −−−−−→ 5, and (t + 1)2 cos t −−−−→ 1.
In order that the delta function have an integral representation, we write
δ(t−t0 )
Z ∞
f (t) −−−−−→ f (t0 ) = f (t)δ(t − t0 ) dt. (6.41)
−∞
Because δ(t − t0 ) cannot be regarded pointwise, the multiplication in this integral,

and the integral itself, are symbolic. When we encounter an integral such as that
in equation 6.41, we interpret it as the action of the function δ(t − t0 ) operating on
f (t) and immediately write f (t0 ). For example,
Z ∞ Z ∞
2 2
t + 2 δ(t) dt = 2, and δ(t + 2) dt = 1
−∞ t +1 −∞
SECTION 6.5 385
(since the left side of the latter integral is interpreted as the delta function δ(t + 2)
operating on the function f (t) ≡ 1).
Because δ(t − t0 ) picks out the value of a function at t = t0 , we also write
Z b
f (t)δ(t − t0 ) dt = f (t0 ) (6.42a)
a
whenever a < t0 < b; that is, the limits on the integral need not be ±∞. Further-
more, if t = t0 is not between a and b, we set
Z b
f (t)δ(t − t0 ) dt = 0. (6.42b)
a
For instance,
Z 6 √ Z 3
√
t + 5 δ(t) dt = 5, and (t2 + 2t − 4)δ(t + 1) dt = 0.
−2 2
With this interpretation for the delta function, its Laplace transform is
Z ∞
L{δ(t − t0 )} = e−st δ(t − t0 ) dt = e−t0 s .
0
What is important to remember from this discussion is that δ(t − t0 ) should never
be regarded in a pointwise sense. We see it in initial-value problems such as
d2 y dy
+2 = 3y = δ(t − 3), y(0) = 1, y 0 (0) = −2,
dt2 dt
but we never consider the differential equation at a specific value of t. The first step
is always to take the Laplace transform of both sides of the equation, in which case
the operational property of the delta function is invoked,
[s2 Y − s + 2] + 2[sY − 1] + 3Y = L{δ(t − 3)} = e−3s .

Laplacenotes

Uploaded by

Copyright:

Available Formats

Laplacenotes

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Laplacenotes

Uploaded by

Copyright:

Available Formats

SECTION 6.

CHAPTER 6 LAPLACE TRANSFORMS

The Laplace transform is one of many integral transforms in applied mathematics.

§6.1 The Laplace Transform and its Inverse

The Laplace transform is one of many useful integral transforms in mathematics.

provided the improper integral converges.

To shorten the notation, we customarily write

understanding that the limit should be taken as t → ∞. Appendix E contains a

Integration by parts leads to

Figure 6.3 Figure 6.4

Since f (t) is piecewise-continuous on 0 ≤ t ≤ T , there is no question that the first

continuous due to the infinite discontinuity at t = 0. It does, however, have a

Further intergations by parts lead to

in order to find the Laplace transform. We provide an alternative using complex

The limit of the antiderivative as t → ∞ is

f (t) F (s) f (t) F (s)

The Inverse Laplace Transform

lim F (s) = 0. (6.5)

Proof: The definition of F (s) gives

Since f (t) is piecewise continuous on 0 ≤ t ≤ T , it is bounded thereon, and there

provided that s > 0 and s > α. The limit of this is zero as s → ∞.

L{f 0 (t)} = sF (s) − f (0), (6.7a)

6.2 Algebraic Properties of the Laplace Transform and its Inverse

L{eat f (t)} = F (s − a), (6.8a)

Proof By Definition 6.1,

But this is equation 6.2 with s replaced by s − a; that is,

L{eat f }(s) = F (s − a), (6.9a)

When calculating transforms and inverse transforms, we often write properties

L{eat f (t)} = L{f (t)}|s→s−a , (6.10a)

On the right of property 6.10a, L{f (t)} is a function of s. The subscript |s → s − a

Property 6.10b yields

Property 6.10b is particularly useful in finding inverse Laplace transforms of

In physical applications, we often encounter quantities that are turned on and

Figure 6.10 Figure 6.11

Figure 6.14 Figure 6.15

L{f (t)h(t − a)} = e−as L{f (t + a)}. (6.14a)

Proof According to Definition 6.1,

If we change variables of integration with u = t − a, then

We illustrate how to use this result in the following examples.

Figure 6.17 Figure 6.18

L{e−3t sin 2t h(t − 1)} = e−s L{e−3(t+1) sin 2(t + 1)}

If we attack the exponential first, we use property 6.10a,

L{e−3t sin 2t h(t − 1)} = L {sin 2t h(t − 1)}s→s+3

The equivalent of property 6.14a in terms of inverse transforms is equally as

L−1 {e−as F (s)} = f (t − a)h(t − a) = L−1 {F (s)}|t→t−ah(t − a). (6.14b)

The graph of f (t − a)h(t − a) is that of f (t) (Figure 6.20a) shifted a units to

A graph of (5/2)(t − 4)2 h(t − 4) is shown in Figure 6.21.

Finding inverse transforms is often a matter of finding the partial fraction

Solution (a) The partial fraction decomposition of F (s) gives

We substitute u = t − p in the second integral,

But f (t) has period p, so that f (u + p) = f (u), and therefore

We can solve this for

L{(1 − t)[h(t) − h(t − 2)]} = L{(1 − t) h(t)} + L{(t − 1) h(t − 2)}

Figure 6.24a Figure 6.24b

We now invert each term,

Properties of Laplace transforms and their inverses that we have discussed in

f (t) F (s) = L{f }(s)

Verify these formulas using

This result will be verified and extended in Section 6.7.

6.3 Laplace Transforms and Differential Equations