Notes 8

Laplace Transform

8.1 Laplace Transform

Under some suitable condition on a function f (t) defined for all t ≥ 0 1 , the following
improper integral with parameter s will converge to a finite value for all large enough s,
giving rise to a function of s which is called the Laplace transform of f :
Z ∞
def
L{f (t)} = F (s) = e−st f (t)dt.
0

1 1
Example 8.1.1 (i) L{1} = , (ii) L{eat } = .
s s−a
Just by direct computation of improper integrals: for any s > 0,
Z ∞
1 ∞ 1 1 1
L{1} = e−st · 1dt = − e−st = − lim e−st + =
0 s 0 t→∞ s s s

which is well-defined for any s > 0. On the other hand, for any s > a,
Z ∞ Z ∞
at −st at 1 −st ∞ 1
L{e } = e · e dt = e−(s−a)t dt = − e =
0 0 s−a 0 s−a

b s
Exercise (i) L{sin bt} = , (ii) L{cos bt} = .
s2 + b2 s2 + b2
Just recall that
Z
eau
eau sin budu = (a sin bu − b cos bu) + C
a2 + b2
Z
eau
eau cos budu = (a cos bu + b sin bu) + C
a2 + b2

The following basic properties of Laplace transform are easy to verify:

1
Main Existence Theorem If f is piecewise continuous on the interval 0 ≤ t ≤ ∞ such that
|f (t)| ≤ Keat when t ≥ M for some real numbers K > 0, M > 0, and a, then the Laplace transform
L{f (t)} = F (s) is well-defined for all s > a.

49
 50

Basic Properties of Laplace Transform

1. L is a linear transform; i.e. for any constants c1 , c2 ,

L{c1 f1 (t) + c2 f2 (t)} = c1 L{f1 (t)} + c2 L{f2 (t)}

2. L{f ′ (t)} = sL{f (t)} − f (0)

(Integrating by parts, under the assumption that e−st f (t) → 0 as t → ∞)
L{f ′′ (t)} = s2 L{f (t)} − sf (0) − f ′ (0)
L{f (n) (t)} = sn L{f (t)} − sn−1 f (0) − · · · − sf (n−2)(0) − f (n−1) (0)
(Under suitable sub-exponential growth conditions on the derivatives!)

With these properties, initial value problems for linear ODEs with constant coefficients
can be transformed into problems in algebraic equations.

Main Application: Solving Initial Value Problems Via L, L−1

Initial Value Problem

Laplace Transform ✲
ay ′′ + by ′ + cy = g(t) an equation for F (s)
L
y(0) = α, y ′ (0) = β
solution ✻ find

L−1 ❄
y(t) ✛ F (s)
Inverse Laplace Transform

Example 8.1.2 Solve the initial value problem

4y ′′ + 5y ′ + y = e2t , y(0) = 1, y ′ (0) = −1.

Taking the Laplace transform of both sides,

4L{y ′′ } + 5L{y ′ } + L{y} = L{e2t }

1
4(s2 F (s) − sy(0) − y ′ (0)) + 5(sF (s) − y(0)) + F (s) = s−2
1
(4s2 + 5s + 1)F (s) + (−4s + 4 − 5) = s−2
1 4s2 −7s−1
(s + 1)(4s + 1)F (s) = s−2 + 4s + 1 = s−2
1 10
4s2 −7s−1 1 2t − 16
F (s) = (s−2)(s+1)(4s+1) = 27
s−2 + = L{ 27
9
s+1 +e } + L{ 10
27
4s+1
−t 4 −t/4
9 e } − L{ 27 e }
n o n o n o
1 −1 1 10 −1 1 4 −1 1
y(t) = 27 L s−2 + 9 L s+1 − 27 L s+1/4
1 2t 10 −t 4 −t/4
y(t) = 27 e + 9 e − 27 e

Partial fraction decompositions are quite often used when looking for inverse Laplace
transform.
 51

p(s)
Recall that a proper rational rational function of the form , where p(s), q(s) are
q(s)
real polynomials with deg p(s) < deg q(s), can be expressed in terms of simple partial
fractions of the following types:
A A A
Type 1: , , ··· ;
s−a (s − a)2 (s − a)n
Cs + D Cs + D Cs + D
Type 2: 2 2
, 2 2 2
, ··· ,
(s − a) + b ((s − a) + b ) ((s − a)2 + b2 )m
depending on how q(s) is factored as a product of linear factors, or as a product of of
linear factors and quadratic factors without real root. Theoretically speaking, if you use
complex numbers, then q(s) can be completely factored as the product of linear factors,
and the rational function can be decomposed into the sum of the first type of (complex)
simple partial factions.

Additional properties of Laplace transform and basic Laplace transform formulas can
be used to identify the inverse Laplace transform.

Basic Laplace Transform Formulas and Properties

f (t) L{f (t)} −→ f (t) L{f (t)}

1 n!
1 tn
s L{tf (t)} = −F ′ (s) sn+1
1 n!
eat dn tn eat
s−a L{tn f (t)} = (−1)n F (s) (s − a)n+1
dsn
a b
sin at eat sin bt
s + a2
2 L{ect f (t)} = F (s − c) (s − a)2 + b2
s s−a
cos at eat cos bt
s + a2
2 −→ (s − a)2 + b2
e−cs
L{uc (t)} = L{uc (t)f (t − c)} = e−cs L{f (t)}
s
Z t
L{δ(t − c)} = e−cs L{ f (t − τ )g(τ )dτ } = L{f (t)}L{g(t)}
0
(convolution integral: for finding L−1 {F (s)G(s)})

Example 8.1.3 Find the inverse Laplace transform of the given rational function. L−1 .)
3s − 5 2 1 L−1
(i) = + , −→ 2et + e2t .
(s − 1)(s − 2) s−1 s−2
3s2 − 5s 1 2 2 L−1
(ii) = + + , −→ et + 2tet + 2e2t .
(s − 1)2 (s − 2)2 s − 1 (s − 1)2 s−2
4s2 − 10s + 9 4s2 − 10s + 9 1 3s − 2
(iii) 2 = 2 2 = + 2 ,
2s + 5})
(s − 2)(s| − {z (s − 2)[ (s − 1) + 2 ] s − 2 s − 2s + 5
| {z }
no real root completing squares

1 3(s − 1) + 1 1 3(s − 1) 1 L−1

= + 2 = + + −→ e2t +3et cos 2t+et sin 2t.
s − 2 s − 2s + 5 s − 2 (s − 1) + 2 (s − 1)2 + 22
2 2
 52

3 i 3 i
3s − 2 2 − 4 2 + 4
(iv) Note that if complex numbers are used, = +
(s − 1)2 + 22 s − 1 − 2i s − 1 + 2i
s4 + 4s2 − 10s + 9 A Bs + C Ds + E
(v) 2 2
= + 2 2
+ ,
(s − 2)(s − 2s + 5) s − 2 (s − 1) + 2 [(s − 1)2 + 22 ]2
where the constants A, B, C, D, E can be found by comparing the coefficients on
both sides of:

s4 + 4s2 − 10s + 9 = A(s2 − 2s + 5)2 + (Bs + C)(s − 2)(s2 − 2s + 5) + (Ds + E)(s − 2).

Exercise Work out the details of the last two examples (iv) and (v).

8.2 Initial Value Problems with Piecewise Continuous In-

put Functions
The method of Laplace transform can be applied to solve nonhomogeneous linear differ-
ential equations with constant coefficients with given initial values:
(
ay ′′ + by ′ + cy = g(t)
y(0) = α, y ′ (0) = β

where the “input function” g(t) is only piecewise continuous, or is certain impulse func-
tion (which causes sudden change in y ′ at a certain instant).

Unit Step Functions (Heaviside Functions)

Like sending out a unit signal at time t = c:

(
0 if 0 ≤ t < c
uc (t) =
1 if t ≥ c 1

c t

Others:

( 
 0 if 0 ≤ t < a
1 if 0 ≤ t < c 
1 − uc (t) = , ua (t) − ub (t) = 1 if a ≤ t < b ,
0 if t ≥ c 


0 if t ≥ b
“unit signal stopped at t = c” “unit signal between t = a and t = b”

1 1

c t a b t
 53

The effect of multiplying uc (t) to f (t − c) is like translating horizontally f (t) by c

units; or delaying the signal f (t).

 0
 if 0 ≤ t < c
uc (t)f (t − c) = y = f (t)

 f (t − c) y = uc (t)f (t − c)
if t ≥ c
t c t

The effect of multiplying ua (t) − ub (t) to f (t) is like cutting the left (t < a) and right
(t ≥ b) tails of f (t):

 0 if 0 ≤ t < a

 y = f (t) y = [ua (t) − ub (t)]f (t)
[ua (t) − ub (t)]f (t) = f (t) if a ≤ t < b



0 if t ≥ b
t a b t

Basic Laplace Transform Formulas for Unit Step Functions

The following properties of Laplace transform are just exercises in computing improper
integrals!
e−cs e−πs
L{uc (t)} = e.g. L{uπ (t)} =
s s
L{uc (t)f (t − c)} = e−cs L{f (t)} e.g. L{uπ (t)t} = L{uπ (t)(t − π) + πuπ (t)}
= L{uπ (t)(t − π)} + πL{uπ (t)}
e−πs
= e−πs L{t} + π
h i s
= e−πs s12 + πs

The unit step functions are often used when dealing with initial value problems

ay ′′ (t) + by ′ (t) + cy(t) = g(t), y(0) = α, y ′ (0) = β,

where the “input function” g(t) is defined piecewise, i.e., defined by different expressions
over some disjoint intervals.

Example 8.2.1 Solve the following initial value problem for a spring-mass system with
piecewise defined forcing function:


 0 if 0 ≤ t < 2

′′
y + 4y = f (t) = t − 2 if 2 ≤ t < 8 , y(0) = 0, y ′ (0) = 0.



6 if 8 ≤ t

(Roughly speaking, there is no external force applied to the system initially at rest until
t = 2, when an external force of magnitude t − 2 is applied to the system during the time
interval 2 ≤ t < 8. The external force is kept at a constant 6 units starting at t = 8.)
 54

Using unit step functions, the forcing term f (t) can be expressed as:

f (t) = u2 (t)(t − 2) − u8 (t)(t − 2) + 6u8 (t) = u2 (t)(t − 2) − u8 (t)(t − 8),

hence, by applying L,
1 −2s 1 −8s
(s2 +4)Y (s) = L{u2 (t)(t−2)}−L{u8 (t)(t−8)} = e−2s L{t}−e−8s L{t} = e − 2e
s2 s
 
1 1 1 1 1
Y (s) = (e−2s − e−8s ) 2 2 = (e−2s − e−8s ) · 2− · 2
s (s + 4) 4 s 4 (s + 4)
1 1 1 1
Y (s) = e−2s L{t − sin 2t} − e−8s L{t − sin 2t}
4 2 4 2
Take the inverse Laplace transform:
h1 1 i 1 1 
y(t) = u2 (t) (t − 2) − sin 2(t − 2) − u8 (t) (t − 8) − sin 2(t − 8) .
4 8 4 8
In piecewise defined form:


 0 if 0 ≤ t < 2

1 1
y(t) = 4 (t − 2) − 8 sin 2(t − 2) if 2 ≤ t < 8


 3 1 1
2 − 8 sin 2(t − 2) + 8 sin 2(t − 8) if 8 ≤ t

This is actually equivalent to solving three initial value problems:

 ′′  ′′  ′′

 y + 4y = 0 
 y + 4y = t − 2 
 y + 4y = 6
  
y(0) = 0, y ′ (0) = 0 → initial values: y(2), y ′ (2) → initial values: y(8), y ′ (8)

 
 

  
0≤t<2 2≤t<8 8≤t
Exercise Solve these three initial value problems and check that they agree with the
solution obtained by the Laplace transform method.

8.3 Delta Function and Impulse

The unit impulse function δ(t) is a “generalized function” defined by some kind of
limiting process:
δ(t) = lim dτ (t)
τ →0
1
where τ > 0, and dτ (t) = 2τ when −τ < t < τ , and dτ (t) = 0 otherwise. Note that
R∞
−∞ dτ (t)dt = 1.
( ( 1
1
if − τ < t < τ τ if − τ /2 < t < τ /2
2τ dτ /2 (t) =
dτ (t) =
0 otherwise 0 otherwise
Area = 1

−τ τ − τ2 τ
2
 55

It is not hard to see then for any t 6= 0, δ(t) = lim dτ (t) = 0, since when τ < |t|,
τ →0
1
dτ (t) = 0. However, δ(0) = lim = ∞.
τ →0 2τ

δ(t) is not a function in usual sense. The main properties we are going to use are as
follows.
Basic Properties of δ(t)

(i) δ(t) = 0 if t 6= 0;
Z ∞ Z ∞
means
(ii) The “integral” of δ(t) is defined by: δ(t)dt = lim dτ (t)dt = 1
−∞ τ →0 −∞

(iii) For any continuous function f ,

Z ∞ Z ∞ Z c+τ
means 1
δ(t − c)f (t)dt = lim dτ −c (t)f (t)dt = lim f (t)dt = f (c)
−∞ τ →c −∞ τ →c 2τ c−τ

Example 8.3.1 Taking f (t) = 1 or f (t) = e−st , we have

Z ∞ Z ∞
δ(t − c)dt = 1 and thus L{δ(t − c)} = e−st δ(t − c) dt = e−cs (c ≥ 0).
−∞ 0

Remark Recall that if mv is the momentum of a mass in certain rectilinear motion

dv
of a particle, then F = ma = m , and
dt
Z b
F (t)dt = mv(b) − mv(a)
a

is the change in momentum caused by the force acting during the time interval a ≤ t ≤
Z b
b. In mechanics, the integral F (t)dt of time-dependent force function is called the
a
impulse imparted by the force during the time interval a ≤ t ≤ b.
So, the delta function δ(t) can be thought of as a mathematical
Z ∞ gadget to describe
a sudden force applied at the instant t = 0, with unit impulse δ(t)dt = 1 which in
−∞
effect results in a sudden change of momentum by 1 N·s.
In other physical problems modelled by second order linear equations, “impulse” may
have different physical meanings.

Example 8.3.2 Consider a mechanical vibration problem

2y ′′ + y ′ + 2y = δ(t − 5), y(0) = 0, y ′ (0) = 0.

Here the delta function δ(t − 5) can be interpreted as a forcing term of impulse 1 applied
to a damped spring-mass system at the instant t = 5.
By Laplace transform, we have
(2s2 + s + 2)F (s) = L{δ(t − 5)} = e−5s
√
e−5s 1 e−5s 15
F (s) = 2s2 +s+2
= 2 (s+ 1 )2 + 15 = 12 e−5s L{ √415 e−t/4 sin 4 t}
4 16
 56

Using the Laplace transform property e−cs L{f (t)} = L{uc (t)f (t − c)}, the inverse Laplace
transform of F (s) is

√  0
 if t < 5
2 − t−5 15
y(t) = √ u5 (t)e 4 sin (t − 5) = √
15 4  √2 e− t−5
 15
4 sin
15 4 (t − 5) if t ≥ 5

If you calculate the derivative y ′ (t) at t = 5 by the two formulas for t < 5 and
′ (5) = 0, and the right-sided derivative
t > 5 respectively, the left-sided derivative is y−
′ 1
is y+ (5) = 2 . Roughly speaking, if you punch the mass with a unit impulse force at the
instant t = 5, you will cause a jump in its velocity by 12 m/s at that moment, i.e. a
change of momentum at t = 5 by 1 Ns.

In fact, by Newton’s 2nd Law: ma = F

2y ′′ = −y ′ − 2y + δ(t − 5)

where 2 is the mass, y ′′ = v ′ is the acceleration, i.e., rate of change of the velocity function
v = y ′ . If you compute the impulse over a short time interval containing t = 5,
Z 5+ Z 5+
mv ′ dt = −y ′ − 2y + δ(t − 5)dt
5− 5−

 Z 5+ 
mv(5+ ) − mv(5− ) = −y(5+ ) + y(5− ) − 2 ydt + 1 = 1
5−
| {z }
= 0, as y(t) is continuous

i.e., the effect of the impulse force δ(t − 5) is causing a jump in the velocity at the instant
t = 5:
1 1
v(5+ ) − v(5− ) = =
m 2

8.4 Convolution Integrals - Dealing with “Inverse Laplace

Transform of a Product”

Theorem If F (s) = L{f (t)} and G(s) = L{g(t)} both exist for s > a ≥ 0, then the
inverse Laplace transform of the product F (s)G(s) is given by the convolution integral
(f ∗ g)(t) of f and g defined by
Z t Z t
(f ∗ g)(t) = f (t − τ )g(τ )dτ = f (τ )g(t − τ )dτ = L−1 {F (s)G(s)}
0 0

The result above can be derived by a direct calculation as follows:

Z ∞ Z ∞Z t
−st
L{(f ∗ g)(t)} = e (f ∗ g)(t)dt = e−st f (t − τ )g(τ )dτ dt
0 0 0
Z ∞Z ∞
= e−st f (t − τ )g(τ )uτ (t)dτ dt
0 0
 57

Z ∞ Z ∞ 
−st
= g(τ ) e uτ (t)f (t − τ )dt dτ
0 0
Z ∞ Z ∞

= g(τ ) (L{uτ (t)f (t − τ )}) dτ = g(τ ) e−sτ F (s) dτ = F (s)G(s)
0 0
Remark Note that
Z t Z ∞
f (t − τ )g(τ )dτ = f (t − τ )g(τ )uτ (t)dτ
0 0

since uτ (t) = 1 if 0 ≤ τ ≤ t, and uτ (t) = 0 if τ > t.

Example 8.4.1 H(s) = s12 = 1s · 1s = L{1}L{1}. Therefore its inverse Laplace transform
is given by the convolution integral of the constant functions f (t) = 1, g(t) = 1:
Z t
−1 1
L { 2 } = (1 ∗ 1)(t) = 1 · 1dt = t
s 0

1
Example 8.4.2 (s−a)(s−b) = L{eat }L{ebt }, (a < b) hence by takeing the convolution
integral of f (t) = eat and g(t) = ebt :
  Z t
1
L −1
= ea(t−τ ) ebt dτ
(s − a)(s − b)) 0
Z t h 1 it 1
= eat e(b−a)τ dτ = eat e(b−a)τ = (ebt − eat )
0 b−a 0 b−a
which agrees with the result obtained by the method of partial fractions:
( )
  1 1
1 − b−a 1 at 1 bt
L−1 = L−1 + b−a =− e + e
(s − a)(s − b)) s−a s−b b−a b−a