Appendix B:

The Laplace Transform

If a function is nonzero only when t > 0, we can replace the Fourier
transform with the Laplace transform. It is particularly useful in solving
initial-value problems involving linear, constant coefficient, ordinary and
partial differential equations. We summarize the main points here.

B.1 DEFINITION AND ELEMENTARY PROPERTIES

Consider a function f (t) such that f (t) = 0 for t < 0. Then the
Laplace integral
 ∞
L[f (t)] = F (s) = f (t)e−st dt (B.1.1)
0

defines the Laplace transform of f (t), which we write L[f (t)] or F (s).
The Laplace transform of f (t) exists, for sufficiently large s, provided
f (t) satisfies the following conditions:
• f (t) = 0 for t < 0,
• f (t) is continuous or piece-wise continuous in every interval,
• tn |f (t)| < ∞ as t → 0 for some number n, where n < 1,
• e−s0 t |f (t)| < ∞ as t → ∞, for some number s0 . The quantity s0
is called the abscissa of convergence.

© 2001 by Chapman & Hall/CRC

• Example B.1.1

Let us find the Laplace transform for the Heaviside step function:

1, t > a,
H(t − a) = (B.1.2)
0, t < a,
where a ≥ 0. The Heaviside step function is essentially a bookkeeping
device that gives us the ability to “switch on” and “switch off” a given
function. For example, if we want a function f (t) to become nonzero at
time t = a, we represent this process by the product f (t)H(t − a).
From the definition of the Laplace transform,
 ∞
e−as
L[H(t − a)] = e−st dt = , s > 0. (B.1.3)
a s

• Example B.1.2

Let us find the Laplace transform of the Dirac delta function or

impulse function. From (1.2.10), the Laplace transform of the delta
function is
 ∞ 
n a+1/n −st
L[δ(t − a)] = δ(t − a)e−st dt = lim e dt (B.1.4)
0 n→∞ 2 a−1/n
 
n
= lim e−as+s/n − e−as−s/n (B.1.5)
n→∞ 2s
 
n e−as s s2 s s2
= lim 1 + + 2 + ··· − 1 + − 2 + ···
n→∞ 2s n 2n n 2n
(B.1.6)
= e−as . (B.1.7)

• Example B.1.3

Although we could compute (B.1.1) for every function that has a

Laplace transform, these results have already been tabulated and are
given in many excellent tables.1 However, there are four basic trans-
forms that the reader should memorize. They are
 ∞  ∞
at −st
L(e ) =
at
e e dt = e−(s−a)t dt (B.1.8)
0 0
∞
e−(s−a)t  1
=−  = , s > a, (B.1.9)
s−a 0 s−a

1 The most complete set is given by Erdélyi, A., W. Magnus, F. Oberhettinger,

and F. G. Tricomi, 1954: Tables of Integral Transforms, Vol I . McGraw-Hill Co.,
391 pp.

© 2001 by Chapman & Hall/CRC

Table B.1.1: Some General Properties of Laplace Transforms with
a>0

function, f (t) Laplace transform, F(s)

1. Linearity c1 f (t) + c2 g(t) c1 F (s) + c2 G(s)

2. Scaling f (t/a)/a F (as)

3. Multiplication ebt f (t) F (s − b)

by ebt

4. Translation f (t − a)H(t − a) e−as F (s)

5. Differentiation f (n) (t) sn F (s) − sn−1 f (0)

−sn−2 f  (0) − · · ·
−f (n−1) (0)
t
6. Integration 0
f (τ ) dτ F (s)/s
t
7. Convolution 0
f (t − τ )g(τ ) dτ F (s)G(s)

 ∞
∞
e−st 
L[sin(at)] = sin(at)e−st dt = − [s sin(at) + a cos(at)]
2
s +a 2 
0 0
(B.1.10)
a
= , s > 0, (B.1.11)
s2 + a2

 ∞
∞
e−st 
L[cos(at)] = cos(at)e −st
dt = 2 2
[−s cos(at) + a sin(at)]
0 s +a 0
(B.1.12)
s
= , s>0 (B.1.13)
s2 + a2

and
 ∞
∞ 
n
tn−m 
L(t ) =
n n −st
t e dt = n!e −st  = n! , s > 0,
(n − m)!sm+1  sn+1
0 m=0 0
(B.1.14)
where n is a positive integer.

© 2001 by Chapman & Hall/CRC

 The Laplace transform inherits two important properties from its
integral definition. First, the transform of a sum equals the sum of the
transforms:

L[c1 f (t) + c2 g(t)] = c1 L[f (t)] + c2 L[g(t)]. (B.1.15)

This linearity property holds with complex numbers and functions as

well.
The second important property deals with derivatives. Suppose
f (t) is continuous and has a piece-wise continuous derivative f  (t). Then
 
∞ ∞ ∞
L[f  (t)] = f  (t)e−st dt = e−st f (t)0 + s f (t)e−st dt (B.1.16)
0 0

by integration by parts. If f (t) is of exponential order,2 e−st f (t) tends

to zero as t → ∞, for large enough s, so that

L[f  (t)] = sF (s) − f (0). (B.1.17)

Similarly, if f (t) and f  (t) are continuous, f  (t) is piece-wise continuous,

and all three functions are of exponential order, then

L[f  (t)] = sL[f  (t)] − f  (0) = s2 F (s) − sf (0) − f  (0). (B.1.18)

In general,

L[f (n) (t)] = sn F (s) − sn−1 f (0) − · · · − sf (n−2) (0) − f (n−1) (0) (B.1.19)

on the assumption that f (t) and its first n−1 derivatives are continuous,
f (n) (t) is piece-wise continuous, and all are of exponential order so that
the Laplace transform exists.

B.2 THE SHIFTING THEOREMS

Consider the transform of the function e−at f (t), where a is any real
number. Then, by definition,
 ∞  ∞

L e−at f (t) = e−st e−at f (t) dt = e−(s+a)t f (t) dt, (B.2.1)

0 0

or

L e−at f (t) = F (s + a). (B.2.2)

2 By exponential order we mean that there exist some constants, M and k, for
which |f (t)| ≤ M ekt for all t > 0.

© 2001 by Chapman & Hall/CRC

Equation (B.2.2) is known as the first shifting theorem and states that
if F (s) is the transform of f (t) and a is a constant, then F (s + a) is the
transform of e−at f (t).

• Example B.2.1

Let us find the Laplace transform of f (t) = e−at sin(bt). Because

the Laplace transform of sin(bt) is b/(s2 + b2 ),

b
L e−at sin(bt) = , (B.2.3)
(s + a)2 + b2

where we have simply replaced s by s + a in the transform for sin(bt).

• Example B.2.2

Let us find the inverse of the Laplace transform

s+2
F (s) = . (B.2.4)
s2 + 6s + 1

Rearranging terms,

s+2 s+2
F (s) = = (B.2.5)
s2 + 6s + 1 (s + 3)2 − 8
√
s+3 1 2 2
= − √ . (B.2.6)
(s + 3)2 − 8 2 2 (s + 3)2 − 8

Immediately, from the first shifting theorem,

√ e−3t √
f (t) = e−3t cosh(2 2t) − √ sinh(2 2t). (B.2.7)
2 2

The second shifting theorem states that if F (s) is the transform

of f (t), then e−bs F (s) is the transform of f (t − b)H(t − b), where b
is real and positive. To show this, consider the Laplace transform of
f (t − b)H(t − b). Then, from the definition,
 ∞
L[f (t − b)H(t − b)] = f (t − b)H(t − b)e−st dt (B.2.8)
0
 ∞  ∞
= f (t − b)e−st dt = e−bs e−sx f (x) dx
b 0
(B.2.9)
 ∞
= e−bs e−sx f (x) dx, (B.2.10)
0

© 2001 by Chapman & Hall/CRC

or
L[f (t − b)H(t − b)] = e−bs F (s), (B.2.11)
where we have set x = t − b. This theorem is of fundamental importance
because it allows us to write down the transforms for “delayed” time
functions. These functions “turn on” b units after the initial time.

• Example B.2.3

Let us find the inverse of the transform e−πs /[s2 (s2 + 1)]. Because

e−πs e−πs e−πs

= − , (B.2.12)
s2 (s2 + 1) s2 s2 + 1

   −πs   −πs 
e−πs −1 e e
L−1 = L − L−1
(B.2.13)
s2 (s2 + 1) s2 s2 + 1
= (t − π)H(t − π) − sin(t − π)H(t − π) (B.2.14)
= (t − π)H(t − π) + sin(t)H(t − π), (B.2.15)

since L−1 (1/s2 ) = t, and L−1 [1/(s2 + 1)] = sin(t).

B.3 CONVOLUTION

In this section, we turn to a fundamental concept in Laplace trans-

forms: convolution. We begin by formally introducing the mathematical
operation of the convolution product:
 t  t
f (t) ∗ g(t) = f (t − x)g(x) dx = f (x)g(t − x) dx. (B.3.1)
0 0

• Example B.3.1

Let us find the convolution between cos(t) and sin(t).

 t
cos(t) ∗ sin(t) = sin(t − x) cos(x) dx (B.3.2)
0
 t
= 1
2 [sin(t) + sin(t − 2x)] dx (B.3.3)
0
 t  t
= 1
2 sin(t) dx + 1
2 sin(t − 2x) dx (B.3.4)
0 0
t t
= 1
2 sin(t) x0 + 1
4 cos(t − 2x)0 = 12 t sin(t). (B.3.5)

© 2001 by Chapman & Hall/CRC

 The reason we have introduced convolution derives from the follow-
ing fundamental theorem (often called Borel’s theorem3 ). If

w(t) = u(t) ∗ v(t) (B.3.6)

then
W (s) = U (s)V (s). (B.3.7)
In other words, we can invert a complicated transform by convoluting
the inverses to two simpler functions.

• Example B.3.2

Let us find the inverse of the transform

 
1 1 a a
= 2 × 2 (B.3.8)
(s2 + a2 )2 a s2 + a2 s + a2
1
= 2 L[sin(at)]L[sin(at)]. (B.3.9)
a

Therefore,
   t
−1 1 1
L = 2 sin[a(t − x)] sin(ax) dx (B.3.10)
(s2 + a2 )2 a 0
 t  t
1 1
= 2 cos[a(t − 2x)] dx − 2 cos(at) dx
2a 0 2a 0
(B.3.11)
t t
1  1 
= − 3 sin[a(t − 2x)] − 2 cos(at) x (B.3.12)
4a 2a
0 0
1
= 3 [sin(at) − at cos(at)]. (B.3.13)
2a

B.4 SOLUTION OF LINEAR ORDINARY DIFFERENTIAL EQUATIONS

WITH CONSTANT COEFFICIENTS

The primary use of Laplace transforms is the solution of ordinary,

constant coefficient, linear differential equations. For example, let us
solve the initial-value problem

dn y dn−1 y dy
n
+ a1 n−1
+ · · · + an−1 + an y = f (t), t>0 (B.4.1)
dt dt dt

3 Borel, É., 1901: Leçons sur les séries divergentes. Gauthier-Villars, p. 104.

© 2001 by Chapman & Hall/CRC

by Laplace transforms, where a1 , a2 , . . . are constants and we know the
value of y, y  , . . . , y (n−1) at t = 0. The procedure is as follows. Ap-
plying the derivative rule (B.1.19) to (B.4.1), we reduce the differential
equation to an algebraic one involving the constants a1 , a2 , . . . , an , the
parameter s, the Laplace transform of f (t), and the values of the ini-
tial conditions. We then solve for the Laplace transform of y(t), Y (s).
Finally, we apply one of the many techniques of inverting a Laplace
transform to find y(t).
Similar considerations hold with systems of ordinary differential
equations. The Laplace transform of the system of ordinary differential
equations results in an algebraic set of equations containing Y1 (s), Y2 (s),
. . . , Yn (s). By some method, we solve this set of equations and in-
vert each transform Y1 (s), Y2 (s), . . . , Yn (s) in turn, to give y1 (t), y2 (t),
. . . , yn (t).
The following example illustrates the details of the process.

• Example B.4.1

Let us solve the ordinary differential equation

y  + 2y  = 8t (B.4.2)

subject to the initial conditions that y  (0) = y(0) = 0. Taking the

Laplace transform of both sides of (B.4.2),

L(y  ) + 2L(y  ) = 8L(t), (B.4.3)

or
8
s2 Y (s) − sy(0) − y  (0) + 2sY (s) − 2y(0) =
, (B.4.4)
s2
where Y (s) = L[y(t)]. Substituting the initial conditions into (B.4.4)
and solving for Y (s),
8 A B C D
Y (s) = = 3+ 2+ + (B.4.5)
s3 (s
+ 2) s s s s+2
8 (s + 2)A + s(s + 2)B + s2 (s + 2)C + s3 D
= 3 = . (B.4.6)
s (s + 2) s3 (s + 2)

Matching powers of s in the numerators of (B.4.6), C +D = 0, B +2C =

0, A + 2B = 0, and 2A = 8 or A = 4, B = −2, C = 1, and D = −1.
Therefore,
4 2 1 1
Y (s) = 3 − 2 + − . (B.4.7)
s s s s+2
Finally, performing term-by-term inversion of (B.4.7), the final solution
equals
y(t) = 2t2 − 2t + 1 − e−2t . (B.4.8)

© 2001 by Chapman & Hall/CRC

B.5 INVERSION BY CONTOUR INTEGRATION

Usually, we can find the inverse of the Laplace transform F (s) by

looking it up in a table. In this section, we show an alternative method
that inverts Laplace transforms through the powerful method of contour
integration.
Consider the piece-wise differentiable function f (x) that vanishes
for x < 0. We can express the function e−cx f (x) by the complex Fourier
representation of
 ∞  ∞ 
−cx 1 iωx −ct −iωt
f (x)e = e e f (t)e dt dω, (B.5.1)
2π −∞ 0

for any value of the real constant c, where the integral

 ∞
I= e−ct |f (t)| dt (B.5.2)
0

exists. By multiplying both sides of (B.5.1) by ecx and bringing it inside

the first integral,
 ∞  ∞ 
1
f (x) = e(c+ωi)x f (t)e−(c+ωi)t dt dω. (B.5.3)
2π −∞ 0

With the substitution z = c + ωi, where z is a new, complex variable of

integration,
 c+∞i  ∞ 
1 zx −zt
f (x) = e f (t)e dt dz. (B.5.4)
2πi c−∞i 0

The quantity inside the square brackets is the Laplace transform F (z).
Therefore, we can express f (t) in terms of its transform by the complex
contour integral:

 c+∞i
1
f (t) = F (z)etz dz. (B.5.5)
2πi c−∞i

This line integral, Bromwich’s integral ,4 runs along the line x = c par-
allel to the imaginary axis and c units to the right of it, the so-called

4 Bromwich, T. J. I’A., 1916: Normal coordinates in dynamical systems. Proc.

London Math. Soc., Ser. 2, 15, 401–448.

© 2001 by Chapman & Hall/CRC

Bromwich contour. We select the value of c sufficiently large so that the
integral (B.5.2) exists; subsequent analysis shows that this occurs when
c is larger than the real part of any of the singularities of F (z).
We must now evaluate the contour integral. Because of the power of
the residue theorem in complex variables, the contour integral is usually
transformed into a closed contour through the use of Jordan’s lemma,
given in Appendix A. The following examples illustrate the proper use
of (B.5.5).

• Example B.5.1

Let us invert
e−3s
F (s) = . (B.5.6)
− 1)
s2 (s
From Bromwich’s integral,
 c+∞i (t−3)z
1 e
f (t) = dz (B.5.7)
2πi c−∞i z 2 (z − 1)

1 e(t−3)z 1 e(t−3)z
= dz − dz, (B.5.8)
2πi C z 2 (z − 1) 2πi CR z 2 (z − 1)
where CR is a semicircle of infinite radius in either the right or left
half of the z-plane and C is the closed contour that includes CR and
Bromwich’s contour. See Figure B.5.1.
Our first task is to choose an appropriate contour so that the inte-
gral along CR vanishes. By Jordan’s lemma, this requires a semicircle
in the right half-plane if t − 3 < 0 and a semicircle in the left half-plane
if t − 3 > 0. Consequently, by considering these two separate cases,
we force the second integral in (B.5.8) to zero and the inversion simply
equals the closed contour.
Consider the case t < 3 first. Because Bromwich’s contour lies to
the right of any singularities, there are no singularities within the closed
contour and f (t) = 0.
Consider now the case t > 3. Within the closed contour in the left
half-plane, there is a second-order pole at z = 0 and a simple pole at
z = 1. Therefore,
 (t−3)z   (t−3)z 
e e
f (t) = Res 2 ; 0 + Res 2 ;1 , (B.5.9)
z (z − 1) z (z − 1)
where
   
e(t−3)z d 2 e(t−3)z
Res 2 ; 0 = lim z 2 (B.5.10)
z (z − 1) z→0 dz z (z − 1)
 
(t − 3)e(t−3)z e(t−3)z
= lim − (B.5.11)
z→0 z−1 (z − 1)2
= 2 − t, (B.5.12)

© 2001 by Chapman & Hall/CRC

 t>3 t<3

(c,0)

Figure B.5.1: Contours used in the inversion of (B.5.6).

and
 
e(t−3)z e(t−3)z
Res 2 ; 1 = lim (z − 1) 2 = et−3 . (B.5.13)
z (z − 1) z→1 z (z − 1)

Taking our earlier results into account, the inverse equals

f (t) = et−3 − (t − 3) − 1 H(t − 3), (B.5.14)

which we would have obtained from the second shifting theorem and
tables.

• Example B.5.2

For our second example of the inversion of Laplace transforms by

complex integration, let us find the inverse of

1
F (s) = , (B.5.15)
s sinh(as)

where a is real. From Bromwich’s integral,

 c+∞i
1 etz
f (t) = dz. (B.5.16)
2πi c−∞i z sinh(az)

© 2001 by Chapman & Hall/CRC

 3 πi/a
2 πi/a
πi/a
(c,0)
- πi/a
- 2 πi/a
- 3 πi/a

Figure B.5.2: Contours used in the inversion of (B.5.15).

Here c is greater than the real part of any of the singularities in (B.5.15).
Using the infinite product for the hyperbolic sine,5
etz etz
= 2 .
z sinh(az) az [1 + a2 z 2 /π 2 ][1 + a2 z 2 /(4π 2 )][1 + a2 z 2 /(9π 2 )] · · ·
(B.5.17)
Thus, we have a second-order pole at z = 0 and simple poles at zn =
±nπi/a, where n = 1, 2, 3, . . .
We convert the line integral (B.5.16), with the Bromwich contour
lying parallel and slightly to the right of the imaginary axis, into a
closed contour using Jordan’s lemma through the addition of an infinite
semicircle joining i∞ to −i∞, as shown in Figure B.5.2. We now apply
the residue theorem. For the second-order pole at z = 0,

   
etz 1 d (z − 0)2 etz
Res ;0 = lim (B.5.18)
z sinh(az) 1! z→0 dz z sinh(az)
 
d zetz
= lim (B.5.19)
z→0 dz sinh(az)
 
etz ztetz az cosh(az)etz
= lim + −
z→0 sinh(az) sinh(az) sinh2 (az)

5 Gradshteyn, I. S. and Ryzhik, I. M., 1965: Table of Integrals, Series and

Products. Academic Press, §1.431, formula 2.

© 2001 by Chapman & Hall/CRC

 (B.5.20)
t
= (B.5.21)
a

after using sinh(az) = az + O(z 3 ). For the simple poles zn = ±nπi/a,

 
etz (z − zn )etz
Res ; zn = lim (B.5.22)
z sinh(az) z→zn z sinh(az)

etz
= lim (0.10.23)
z→zn sinh(az) + az cosh(az)
exp(±nπit/a)
= , (B.5.24)
(−1)n (±nπi)

because cosh(±nπi) = cos(nπ) = (−1)n . Thus, summing up all of the

residues gives
∞ ∞
t  (−1)n exp(nπit/a)  (−1)n exp(−nπit/a)
f (t) = + −
a n=1 nπi n=1
nπi
(B.5.25)
∞

t 2 (−1)n
= + sin(nπt/a). (B.5.26)
a π n=1
n

B.6 SOLUTION OF PARTIAL DIFFERENTIAL EQUATIONS

The solution of linear partial differential equations by transform

methods is the most commonly employed analytic technique after the
method of separation of variables.
Consider the case where we wish to solve a partial differential equa-
tion that depends on x and t. If the solution is denoted by u(x, t), then
its Laplace transform is
 ∞
U (x, s) = u(x, t)e−st dt, (B.6.1)
0

because the transform consists solely of an integration with respect to

time. Partial derivatives involving time have transforms similar to those
that we encounter in the case of functions of a single variable. They
include
L[ut (x, t)] = sU (x, s) − u(x, 0), (B.6.2)
and
L[utt (x, t)] = s2 U (x, s) − su(x, 0) − ut (x, 0). (B.6.3)

© 2001 by Chapman & Hall/CRC

These transforms introduce the initial conditions such as u(x, 0) and
ut (x, 0). On the other hand, derivatives involving x become
d dU (x, s)
L[ux (x, t)] = {L[u(x, t)]} = , (B.6.4)
dx dx
and
d2 d2 U (x, s)
L[uxx (x, t)] = 2
{L[u(x, t)]} = . (B.6.5)
dx dx2
Because the transformation eliminates the time variable t, only
U (x, s) and its derivatives remain in the equation. Consequently, we
have transformed the partial differential equation into a boundary-value
problem for an ordinary differential equation. Because this equation
is often easier to solve than a partial differential equation, the use of
Laplace transforms has considerably simplified the original problem. Of
course, the Laplace transforms must exist for this technique to work.
To summarize this method, we have constructed the following sche-
matic:

We illustrate these concepts by finding the sound waves that arise

when a sphere of radius a begins to pulsate at time t = 0. The symmetric
wave equation in spherical coordinates is
1 ∂ 2 (ru) 1 ∂2u
2
= 2 2, (B.6.6)
r ∂r c ∂t
where c is the speed of sound, u(r, t) is the velocity potential and −∂u/∂r
gives the velocity of the parcel of air. At the surface of the sphere r = a,
the radial velocity must equal the velocity of the pulsating sphere
∂u dξ
− = , (B.6.7)
∂r dt

© 2001 by Chapman & Hall/CRC

where ξ(t), the displacement of the surface of the pulsating sphere,
equals B sin(ωt)H(t). The air is initially at rest.
From (B.6.2)–(B.6.5) the Laplace transform of (B.6.6) is
 
d2 s2
r U (r, s) − r U (r, s) = 0. (B.6.8)
dr2 c2

The solution of (B.6.8) is

r U (r, s) = A exp(−rs/c). (B.6.9)

We have discarded the exp(rs/c) solution because it becomes infinite

in the limit of r → ∞. After substituting (B.6.9) into the Laplace
transformed (B.6.7),
 −sr/c   
d e 
− A  = ωBs = Ae−as/c 1 + s . (B.6.10)
dr r  s2 + ω 2 a2 ac
r=a

Therefore,

ωBa2 cs
r U (r, s) = e−s(r−a)/c (B.6.11)
(s2 + ω 2 )(as + c)
 
ωBa2 c −s(r−a)/c cs + ω 2 a c
= 2 2 e − . (B.6.12)
a ω + c2 s2 + ω 2 s + c/a

Applying the second shifting theorem and tables, the inversion of (B.6.
12) follows directly
      
ωBa2 c2 r−a ωa r−a
ru(r, t) = 2 2 cos ω t − + sin ω t −
a ω + c2 c c c
    
c r−a r−a
− exp − t− H t− . (B.6.13)
a c c
In this example, we eliminated the temporal dependence by using
Laplace transforms. The ordinary differential equation was then solved
using homogeneous solutions. In certain cases, an alternative would be
to solve the ordinary differential equation by Fourier or Hankel trans-
forms. This is particularly true in the case of Green’s functions because
the forcing function is a delta function. For a Green’s function problem
that depends upon a spatial dimension and time, we obtain an alge-
braic equation that we solve to find the joint transform. The inverses
would then be found successively. Whether we invert the Laplace or
the spatial transform first depends upon the problem. This repeated
application of transforms or Fourier series to a linear partial differential
equation to reduce it to an algebraic or ordinary differential equation
can be extended to higher spatial dimensions.

© 2001 by Chapman & Hall/CRC