Massachusetts Institute of Technology

Department of Mechanical Engineering

2.004 Dynamics and Control II

February 15, 2019

Laplace Transforms

Objectives

• Understand the Concepts

• Convert Time functions into the Laplace domain
• Use Laplace Transforms to convert differential equations into algebraic equations
• Compute Transfer Functions
• Take the Inverse Laplace Transform and find the Time Response of a system
• Use the initial and final theorems to find the initial and steady-state responses of the system
• Relate the time response and the sinusoidal response of the system

The Laplace Transform is an Operational Method used to:

• Convert functions of time into functions that are algebraic functions of a complex variable
• Convert differentiation and integration operations to algebraic operations in the complex plane
• Convert linear differential equations into algebraic equations in a complex variable
• Solve differential equations using Inverse Laplace transform or tables
• Allow the use of graphical methods to predict system performance without solving the differential
equations of the system. These include response, steady-state behavior and transient behavior.
• Analyze general signals
• Provide a spectral representation for signals for which a Fourier Transform does not exist

1
 Laplace Transforms
Let f (t) be a function of time t such that f (t) = 0 for t < 0. Thus, our interest is in signals defined for
t ≥ 0.

• t is a time variable in seconds.

1
• s is a complex frequency variable with units of Hz ( sec ).

• L is the Laplace Transform operator

• F (s) (capital) is the Laplace transform of f (t) (lower case).

One-sided or Unilateral Laplace Transform

Z ∞
L[f (t)] = F (s) = f (t)e−st dt (1)
0

The two-sided or bilateral Laplace transform is obtained by setting the lower limit of the integral to −∞.
In engineering applications, we are concerned with causal systems and thus in general we use the one sided
form.

2
 Region of Convergence
The Laplace Transform F (s) of f (t) exists if the integral converges:
Z ∞
f (t)e−st dt where s = σ + jω (2)
0

All complex values of s for which the integral converges form the region of convergence (ROC).
The ROC is a region in the s-plane.

A function f(t) will have a Laplace transform if it is of exponential order, that is:

lim |f (t)e−σt | = 0 for some real numberσ (3)

t→∞
Z ∞ Z ∞
L[f (t)] = F (s) = f (t)e−st dt = f (t)e−σt e−jωt dt (4)
0 0

The term e−jωt represents sinusoidal functions with frequency ω and therefore is bounded. Thus in
engineering applications, we are concerned with causal systems and thus in general we use the one-sided
form.
Complex frequency combines transient response with sinusoidal steady-state response to get total response
of system to input. A complex frequency: s = σ + jω, where σ is the exponential decay/increase constant
that is related the time constants of the system’s transient response. RC = L/R = σ in circuit analysis.
eαt is an exponentially increasing function over time, while e−αt is an exponentially decreasing function over
time.

3
 Example: Step Function
A step function is written:

f (t) = 0 for t < 0

f (t) = A for t > 0

where A is a real constant.

The Laplace transform of f(t) is:
A
L[f (t)] = (5)
s
The Laplace transform must convergeR 0+ for the transform to exist. Note that the step function is undefined
at t = 0. This does not matter since 0− Ae−st dt = 0.
The unit step function: f (t) = us (t).
Z ∞
1 ∞ 1
Us (s) = 1e−st dt = − e−st 0− = (6)
0− s s
The Laplace transform is valid for all s except at the pole s = 0.

4
 Example: Ramp Function
A ramp function is written:

f (t) = 0 for t < 0

f (t) = At for t ≥ 0

where A is a real constant.

The Laplace transform of f(t) is:
Z ∞
L[f (t)] = A te−st dt
0
−st ∞
e Ae−st
Z
∞
= At 0
− dt
−s 0 −s
Z ∞
A
= e−st dt
s 0
A
= 2
s
e−st
Note: Use integration by parts with u = At, du = Adt, v = −s , and dv = e−st dt.
Z ∞ Z ∞
∞
udv = uv 0
− vdu
0 0

Example: Exponential Function

An exponential function is written:

f (t) = 0 for t<0

−α
f (t) = Ae t for t≥0

where A and α > 0 are real constants.

The Laplace transform of f(t) is:
Z ∞
L[f (t)] = A e−αt e−st dt
0

The Laplace integral must converge for the transform to exist.

The one-sided exponential function f (t) = us (t)e−αt (α > 0).
Z ∞
1
F (s) = e−αt e−st dt =
0− s+a

5
 Example: A Pulse Let f (t) be a very brief pulse over time T with magnitude

1/T and unit area.

The Laplace transform of f(t) is:
T
1 −st
Z
L[f (t)] = e dt
0− T

Because T is small, e−st ≈ 1 over its duration.

T
1
Z
L[f (t)] = 1dt = 1
T 0−

Example: Dirac Delta Function

As T → 0, the amplitude becomes very large. We defined the Dirac delta (or impulse) function δ(t) as
the limiting case, with the following properties:

δ(t − to ) = 0 for t 6= to
δ(t − to ) = ∞ for t = to
Z ∞
δ(t − to )dt = 1
−∞
δ(t) = 0 for all t 6= 0
δ(t) is undefined (infinite) at t=0
Z ∞
δ(t)dt = 1 (unit area)
−∞
L[δ(t)] = 1

Although δ(t) can only be approximated in practical situations, it is a very important theoretical function
in system dynamics and control theory.

δ(t − to ) = 0 fort 6= to
Z to +ǫ
δ(t − to )dt = 1 ǫ>0
to −ǫ

The sifting or sampling property of the unit impulse:

Z t2
f (t)δ(t − to )dt = f (to ) t1 < to < t2
t1
=0 o.w.

6
• Set of delayed Dirac delta functions:
∞
X
s(t, ∆T ) = δ(t − n∆T )
n=−∞

• Discrete sample sequence by Integration across each impulse:

Z n∆T +
fn = f (n∆T ) = f ∗ (t)dt
n∆T −

• Sampled waveform:
∞
X
f ∗ (t) = s(t, ∆T )f (t) = f (t)δ(t − n∆T )
n=−∞

7
 Example: Sine Function
A sine function is written:

f (t) = 0fort < 0

f (t) = Asinωtfort ≥ 0

where A and α > 0 are real constants.

The Laplace transform of f(t) is:
Z ∞
L[f (t)] = A (sinωt)e−st dt
0

Since,

ejωt = cos(ωt) + j sin(ωt)

e−jωt = cos(ωt) − j sin(ωt)

We obtain:
1 jωt
sin ωt = (e − e−jωt )
2j

A ∞ jωt
Z
L[f (t)] = (e − e−jωt )e−st dt
2j 0
A 1 A 1
= −
2j s − jω 2j s + jω
Aω
= 2
s + ω2

Example: Cosinusoid Function

The one-sided Cosinusoid function is written:

f (t) = us (t) cos(ωt)

Using Euler formula:

1 jωt
cos ωt = (e + e−jωt )
2

1
L[f (t)] = (L[ejωt ] + L[e−jωt ])
2
1 1 1 
= +
2 s − jω s + jω
s
= 2
s + ω2

8
Inverse Laplace Transforms

Impulse δ(t) 1

1
Unit Step us (t) s

1
e−αt s+a

1
eαt s−a

ω
sin(ωt) s2 +ω 2

s
cos(ωt) s2 +ω 2

1
Linear ramp t s2

1
te−αt (s+a)2

tn e−αt n!
(s+a)n+1

e−αt sin ωt ω
(s+a)2 +ω 2

e−αt cos ωt s+a

(s+a)2 +ω 2

s sin θ+ω cos θ

sin(ωt + θ) s2 +ω 2

s cos θ−ω sin θ

cos(ωt + θ) s2 +ω 2

f (t − to )u(t − to ), to ≥ 0 e−to s F (s)

f (t)u(t − to ), t≥0 e−to s L[f (t + to )]

9
Properties of Laplace Transforms

Laplace Transform Time Function Comment

F (s) f (t) Transform pair

αF1 (s) + βF2 (s) αf1 (t) + βf2 (t) Superposition

F (s)e−sλ f (t − λ) Time delay(λ ≥ 0)

1 s
|a| F ( a ) f (αt) Time scaling

F (s + a) e−αt) f (t) Frequency shift

sm F (s) − sm−1 f (0) − sm−2 f (0) − · · · − f (m−1) (0) f (m) (t) Differentiation

1
Rt
s F (s) 0
f (ζ)dζ Integration

F1 (s)F2 (s) f1 (t) ∗ f2 (t) Convolution

lims→∞ sF (s) f (0+ ) Initial Value Theorem

lims→0 sF (s) limt→∞ f (t) Final Value Theorem

1
R σc +j∞
2πj σc −j∞
F1 (ζ)F2 (s − ζ)dζ f1 (t)f2 (t) Time Product

1
R +j∞ R∞
2π −j∞
Y (−jω)U (jω)dω 0
y(t)u(t)dt Parseval’s Theorem

d
− ds F (s) tf (t) Multiplication by Time

10
 Inverse Laplace
In principle, we can recover f from F via:
σ+j∞
1
Z
f (t) = F (s)est ds (7)
2πj σ−j∞

where σ is large enough that F (s) is defined for Rs ≥ σ.

In general, functions that increase faster than the exponential function do not have a Laplace Transform.
For example, the follow functions do not have a Laplace transform:
2
f1 (t) = et for0 ≤ t ≤ ∞
t2
f1 (t) = te for0 ≤ t ≤ ∞

Note that f1 (t) can have a Laplace transform if it is for 0 ≤ t ≤ T ≤ ∞ where T is a finite time, and
that f1 (t) = 0 for t < 0 and t > T .

11
 Properties

Uniqueness

f (t) ↔ F (s)
f (t) and F (s) are transform pairs. Each f (t) has a unique F (s) and each F (s) has a unique f (t). The
Laplace transform is one-to-one: if L(f ) = L(g) then f = g.

• F determines f

• Inverse Laplace transform L−1 is well-defined

Linearity

The property of linearity is:

IfF (s) = L[f (t)]andG(s) = L[g(t)]then

L[af (t) + bg(t)] = aF (s) + bG(s)

where a, b are constants. The linearity property is fundamental to our treatment of ODEs and linear
systems.
Example:

L[3δ(t) − 2et ] = 3L(δ(t)) − 2L(et )

2
=3−
s−1
3s − 5
=
s−1
Proof:
Z ∞
L[αf1 (t) + βf2 (t)] = [αf1 (t) + βf2 (t)]e−st dt
0
Z ∞ Z ∞
=α f1 (t)e−st dt + β f2 (t)e−st dt
0 0
= αF1 (s) + βF2 (s)

12
 Final Value Theorem

IfF (s) = L[f (t)]then

lim f (t) = lim sF (s) = f (∞)
t→∞ s→0

provided the limit exists. The F.V. theorem is useful for determining the steady-state response of systems,
i.e. it allows finding f (∞) without computing the inverse of F (s).
Proof: To prove the theorem, we let s approach zero in the equation for the Laplace transform of the
derivative of f (t).
Z ∞h
d i
lim f (t) e−st dt = lim [sF (s) − f (0)]
s→0 0 dt s→0

Since lims→0 e−st = 1, we obtain,

Z ∞h
d i
lim f (t) dt = f (∞) − f (0) = lim [sF (s) − f (0)]
s→0 0 dt s→0

Hence, f (∞) = limt→∞ f (t) = lims→0 sF (s).

Example: Find f (∞) for:
(s + 2)2 − 32
F (s) =
(s + 2)2 + 32
Note that: F −1 (s) = te−2t cos 3t.

(s + 2)2 − 32
f (∞) = lim sF (s) = lim =0
s→0 s→0 (s + 2)2 + 32

Initial Value Theorem

IfF (s) = L[f (t)]then

lim f (t) = lim sF (s) = f (0)
t→0 s→∞

provided the limit exists. The I.V. theorem is useful for determining the initial condition without com-
puting the inverse of F (s).
Example: Find f (0) for:
(s + 2)
F (s) =
(s + 1)2 + 52

s2 + 2s
f (0) = lim sF (s) = lim
s→∞ s→∞ s2 + 2s + 1 + 25
s /s + 2s/s2
2 2
= lim 2 2 =1
s→∞ s /s + 2s/s2 + (26/s2 )

Time Scale

Define signal g(t) = f (at), where a > 0. Then,

13
 1
G(s) = F (s/a) (8)
a
Times are scaled by a, frequencies by 1/a.
Z ∞
1 ∞ 1 s
Z
s
G(s) = f (at)e−st dt = f (τ )e− a τ dτ = ( )F ( )
0 a 0 a a
where τ = at, dτ = adt and dt = d τa .

Example: Find the Laplace transform of f (t) = A sin(t − td ).

1
L[sin(t)] =
s2 +1
A
F (s) = e−std
s2 + 1

14
Exponential Scaling

Let f be a signal and a be a scalar, then define g(t) = eat f (t). Then,

G(s) = F (s − a) (9)
Times are scaled by a, frequencies by 1/a.
Proof: Z ∞ Z ∞
G(s) = e−st eat f (t)dt = e−(s−a)t f (t)dt = F (s − a)
0 0
where τ = at, dτ = adt and dt = d τa .
Example: Find the Laplace Transform of f (t) = A sin(ωt)e−αt) .
ω
L[sin(ωt)] =
s2 + ω 2
Aω
F (s) =
(s + a)2 + ω 2

Time Shift

Let f be a signal and T > 0, then define:

g(t) = 0 0 ≤ t < T
= f (t − T ) t≥T

That is, signal g is f delayed by T seconds and ‘zero-padded’ up to T .

G(s) = e−sT F (s) (10)

Proof: Z ∞ Z ∞
−st
G(s) = e f (t − T )dt = e−s(τ +T ) f (τ )dτ = e−sT F (s)
0 0
where τ = t − T , dτ = dt.

15
Differentiation Property

If F (s) = L[f (t)],

df
L[ = sF (s) − f (0− )]
dt
d2 f
L[ 2 ] = s2 F (s) − sf (0− ) − f˙(0)
dt
n
dn f X
L[ n ] = sn F (s) − sn−k f (k−1) (0)
dt
k=1

If all derivatives of f are zero at t = 0 then,

h dn f (t) i
L = sn F (s) (11)
dtn
Proof: Use Integration by parts:

u = e−st du = −se−st dt
↔
df (t)
v = f (t) ↔ dv = dt = df (t)
dt
Z ∞ Z ∞
∞
udv = uv|0 − vdu
0 0

Z ∞
df
L[ = f (t)e−st |∞
0 − f (t)(−se−st )dt
dt 0
Z ∞
= 0 − f (0) + s f (t)e−st dt
0

Therefore,
h df (t) i
L = sF (s) − f (0)
dt
Examples:
Let f (t) = et , then f ′ (t) = et ,
1
L(f ) = L(f ′ ) =
s−1
Let f (t) = sin ωt = − ω1 dt
d
cos ωt,
1 s  ω
L(sin ωt) = − s 2 2
− 1 = 2
ω s +ω s + ω2
Let f (t) be the unit ramp, then f ′ (t) is the unit step,
1 1
L(f ′ ) = s −0=
s2 s

Derivative of Discontinuous Functions

If signal f is discontinuous at t = 0, then

L(f ′ ) = sF (s) − f (0−) (12)

16
 Integration

hZ ∞ i Z ∞ hZ t i
L f (t)dt = f (x)dx e−st dt (13)
0 0 0
Rt
Use Integration by parts, u = 0
f (x)dx, du = f (t)dt, dv = e−st dt and v = − 1s e−st .
After substitution:
∞ i 1Z ∞
hZ 1
L f (t)dt = f (t)e−st dt = F (s)
0 s 0 s

F (s) f −1 (0)
Z
L[ f (t)dt] = +
s s
where, f −1 (0) = [ f (t)dt]t=0 . If f −1 (0) = 0, then L[ f (t)dt] = F (s)
R R
s .
Let g be the running integral of a signal f ,
Z t
g(t) = f (τ )dτ (14)
0
Then,
1
G(s) = F (s) (15)
s
Rt
Example: Find the Laplace Transform of f (t) = 0 sin ωτ dτ .
The Laplace transform of sin ωt is:
ω
L[sin ωt] = 2
s + ω2
.
ω
F (s) =
s3 + ω2 s

17
 Multiplying by t

Let f be a signal and g(t) = tf (t).

Then,
G(s) = −F ′ (s) (16)
R∞
Proof: Differentiate both sides of F (s) = 0
e−st f (t)dt w.r.t s,
Z ∞
F ′ (s) = (−t)e−st f (t)dt (17)
0
Multiplication by time corresponds to differentiation in the frequency domain. Let us consider:

Z ∞
d d
F (s) = e−st f (t)dt
ds ds 0
Z ∞
= −te−st f (t)dt
0
Z ∞
=− e−st [tf (t)]dt
0
= −L[tf (t)]

d
L[tf (t)] = − F (s)
ds

f (t) = e−t , g(t) = te−t

d 1 1
L(te−t ) = − =
ds s + 1 (s + 1)2
f (t) = te−t , g(t) = t2 e−t
d 1 2
L(t2 e−t ) = − =
ds (s + 1)2 (s + 1)3

In general,
(k − 1)!
L(tk e−t ) = (18)
(s + 1)k+1
Example: Find the Laplace Transform of f (t) = t sin ωt.
The Laplace transform of sin ωt is:
ω
L[sin ωt] = 2
s + ω2
.
dh ω  2ωs
F (s) = − 2 2
= 2
ds s + ω (s + ω 2 )2

18
 Convolution Integral

Z ∞
(f ∗ g)(t) , f (τ )g(t − τ )dτ
Z∞∞
= f (t − τ )g(τ )dτ
∞

The convolution of signals f, g denoted as h = f ∗ g, is the signal:

Z t
h(t) = f (τ )g(t − τ )dτ
0
Z t
= f (t − τ )g(τ )dτ
0

Therefore, f ∗ g = g ∗ f . In terms of Laplace Transforms, H(s) = F (s)G(s). The Laplace transform turns
convolution into multiplication.
Proof: Let us show that L(f ∗ g) = F (s)G(s):

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

where we integrate over the triangle 0 ≤ τ ≤ t.

1. Change the order of integration:
Z ∞ Z ∞
H(s) = e−st f (τ )g(t − τ )dτ dt
τ =0 t=τ

2. Change the variable t to t̄ = t − τ 3. Change the region of integration: dt̄ = dt, τ ≥ 0, t̄ ≥ 0.

Then,

Z ∞ Z ∞
H(s) = e−s(t̄+τ f (τ )g(t̄)dt̄dτ
τ =0 t=τ
Z ∞  Z ∞ 
−sτ −st̄
= e f (τ )dτ e g(t̄)dt̄
τ =0 t̄=0
= F (s)G(s)
Example: Find the ramp response of a first-order system with a pole at +a.
Let f1 (t) = t be the ramp input and f2 (t) = eαt be the impulse response of the first-order system. Then,

h1 1 i
L−1 = f1 (t) ∗ f2 (t)
s2 s − a
Z t
= f1 (τ )f2 (t − τ )dτ
0
Z t
= τ eα(t−τ ) dτ
0
1
= 2 (eαt − αt − 1)
a

19
 Parseval’s Theorem

Parseval’s famous theorem is used to compute the energy in a signal or correlation between two signals.
It tells us that mentioned quantities can be computed either in the time domain or in the frequency domain.
If Z ∞ Z ∞
|y(t)|2 dt < 1 and |u(t)|2 dt < 1
0 0

i.e. y(t), u(t) are square integrable, then

Z ∞ Z ∞
1
y(t)u(t)dt = Y (−jω)U (jω)dω
0 2π −∞
.
Parseval’s result involves only a substitution of the transform for the time functions and an exchange of
integration:

∞ ∞  Z ∞ 
1
Z Z
y(t)u(t)dt = y(t) U (jω)ejωt dω dt
0 0 2π −∞
Z ∞ Z ∞ 
1
= U (jω) y(t)ejωt dt dω
2π ∞ 0
Z ∞
1
= U (jω)Y (−jω)dω
2π −∞

Patterns

While details differ, you can see some interesting symmetric patterns between the time domain (i.e.
signals) and the frequency domains (i.e. their Laplace Transforms).

• Differentiation in one domain corresponds to multiplication by the varaible in the other.

• Multiplication by an exponential in one domain correspnonds to a shift (or delay) in the other

20
Inverse Laplace Transforms
The Inverse Laplace transform is commonly found by using tables and partial fractions expansion.
Write the Laplace transform of a signal f (t) in the following form:

P (s)
F (s) = (19)
Q(s)

• P(s) and Q(s) are polynomials

• s is the Laplace variable

• The order of Q(s) ≥ the order of P(s) (proper form). If this is not the case, P(s) divided by Q(s), using
long division, until the remaining ratio is in proper form.

Partial Fractions

Three cases:

Case 1: Non-repeated roots

k1 k2 kn
F (s) = + + ··· +
s + p1 s + p2 s + pn
Case 1: Complex poles
k1 k1∗
F (s) = +
s + (α + jβ) α − jβ
where k1∗ is the complex conjugate ofk1
Case 1: Repeated poles
P1 (s)
F (s) =
Q1 (s)(s + p1 )r
k11 k12 k1r P1 (s)
= + + ··· + +
s + p1 (s + p1 )2 (s + p1 )r Q1 (s)

21
Case 1:
4(s + 2) A1 A2 A3
F (s) = = + +
(s + 1)(s + 4)(s + 10) s + 1 s + 4 s + 10
The coefficients A1 , A2 , A3 can be found as:

(s + 1)4(s + 2) 4
A1 = |s=−1 =
(s + 1)(s + 4)(s + 10) 27
(s + 4)4(s + 2) 4
A2 = |s=−4 =
(s + 1)(s + 4)(s + 10) 9
(s + 10)4(s + 2) 16
A3 = |s=−10 = −
(s + 1)(s + 4)(s + 10) 27

4 −t 4 −4t 16
f (t) = e + e + − e−10t
27 9 27
Case 2:
P1 (s) K1 K1∗
F (s) = = + + ...
Q1 (s)(s + α − jβ)(s + α + jβ) s + α − jβ s + α + jβ
The coefficient K1 is given by:

(s + α − jβ)P1 (s)
K1 = |s=−α−jβ = |K1 |∠θ = |K1 |ejθ
Q1 (s)(s + α − jβ)(s + α + jβ)

K1 K1∗ |K1 |ejθ |K1 |e−jθ

+ = +
s + α − jβ s + α + jβ s + α − jβ s + α + jβ

h |K |ejθ
1 |K1 |e−jθ i
L−1 + = |K1 |[ejθ e−αt ejβt + e−jβ e−αt e−jβt ]
s + α − jβ s + α + jβ
h ej(βt+θ) ej(βt+θ) i
= 2|K1 |e−αt = 2|K1 |e−αt [cos(βt + θ)]
2

22
Convolution Integral

We will look at how this system is related in the time domain and in the Laplace Transform:

• u(t) is the input

• h(t) is the impulse response

• y(t) is the output

In the time domain, the output y(t) is the convolution of u(t) and h(t).
Z t Z t
y(t) = u(t) ∗ h(t) = u(t − τ )h(τ )dτ = u(τ )h(t − τ )dτ (20)
0 0

In the frequency domain, Y (s) is given by,

Y (s) = L[u(t) ∗ h(t)] = U (s)H(s) (21)

1
Example: Let h(t) = e−4t . Then, H(s) = L[h(t)] = s+4 .

• h(t) is known as the system impulse response because when the input u(t) is the unit impulse δ(t),
L[u(t)] = U (s) = 1.

• Y (s) is also known as the system impulse response, since L−1 [Y (s)] = y(t) = L−1 [H(s)] = h(t).

Example:
Z ∞ Z t
−4(t−τ )
y(t) = e u(τ )dτ = e−4(t−τ ) dτ
∞ 0
Z t
= e−4t e4τ dτ
0
1
= e−4t e4τ |ττ =0
=t
4
h1 1 i
= − e−4t
4 4

1 A B 1/4 1/4
Y (s) = H(s)U (s) = = + = −
s(s + 4) s s+4 s s+4
1
y(t) = [1 − e−4t ]
4

23