A Few Notes On The Dirac Delta Function And

The Laplace Transform

David Meyer

dmm613@[Link]

Last Update: June 2, 2024

Initial Version: July 15, 2021

1 Introduction
These notes began life as some thoughts on the Dirac Delta Function and evolved into notes on several
related topics including Laplace Transforms. The Dirac Delta function has all kinds of crazy and interesting
properties. More TBD.

2 The Dirac Delta Function

The Dirac Delta Function is defined as shown in Figure 1. In the limit ( → 0) the Dirac Delta function
is written δa (t) or sometimes δ(t − a). As we will see in a moment, the δa, (t) form of the delta function
is useful when we want to use the Mean Value Theorem for Integrals [2] to evaluate integrals involving the
delta function.

δa, (t)

(
1
 a≤t≤a+
δa, (t) =
0 otherwise

1


a t
a+

Figure 1: The Dirac Delta Function δa, (t)

So δa, (t) is defined to be

(
1
 a≤t≤a+
δa, (t) =
0 otherwise

1
and has the constraint that

Z ∞
δa, (t) = 1
0

That is, δa, (t) is in some sense a probability density.

In the limit the Dirac Delta Function looks like

(
∞ t=a
lim δa, (t) = δa (t) =
→0 6 a
0 t=

or sometimes
(
∞ t=a
δ(t − a) =
0 t 6= a

δa (t) also has the constraint that

Z ∞
δa (t) = 1
0

and so is also a probability density. δa (t) is shown in Figure 2.

δa (t)

(
∞ t=a
δa (t) =
0 otherwise

a t

Figure 2: The Dirac Delta Function δa (t)

2.1 Integrals involving δa, (t)

δa, (t) has all kinds of interesting properties. One of them involves the integral of the product δa, (t) with
some function g(t). Here we would like to evaluate integrals of the form

Z ∞
δa, (t)g(t)dt (1)
0

where g(t) is continuous on the interval [a, a + ].

2
The Mean Value Theorem for Integrals [2] tells us that
Z b
g(t)dt = (b − a)g(c) (2)
a

where the point c lies in the interval [a, a + ]. Now, since we know that δa, (t) is zero everywhere except on
the interval [a, a + ] we can rewrite the improper integral in Equation 1 as the proper integral
Z a+
δa, (t)g(t)dt
a

Here we can notice that δa, (t) = 1 on the interval [a, a + ] so we can rewrite our integral as

Z a+ Z a+
1 1
g(t)dt = g(t)dt
a   a

Now we can use Equation 2, the Mean Value Theorem for Integrals1 , by setting b = a +  and a = a so that
b − a = . Then

Z a+
1 1h i 1
g(t)dt = (a + ) − a g(c) = g(c) = g(c)
 a | {z } 
b−a

where c ∈ [a, a + ].

Finally, if we look at what happens to c as  → 0 we see that lim c = a (sorry about the notation abuse) so
→0
that
Z ∞
δa (t)g(t)dt = g(a) (3)
0

Essentially δa (t) pulls out the value of g at a, that is, g(a).

Another way to look at this [1] is to notice that the integrand of

Z ∞
δ(t − a)g(t)dt
0

R∞ R∞
is zero everywhere except where t = a, so we can rewrite our integral as 0 δ(t−a)g(a)dt = g(a) 0 δ(t−a)dt
R∞
(since g(a) doesn’t depend on t). Then since by definition 0 δ(t − a)dt = 1 we get

Z ∞
δ(t − a)g(t)dt = g(a)
0

1 This is where the δa, (t) form of the delta function comes in handy.

3
3 The Laplace Transform
We start by defining the integral transform of some function f (t).

Definition 3.1. Integral Transform: If a function f (t) is defined on [0, ∞) then we can define an integral
transform to be the improper integral

Z ∞
F (s) = K(s, t)f (t) dt
0

Here t ∈ R≥0 and s ∈ C is a complex frequency domain parameter with s = σ + iω and σ, ω ∈ R.

If the improper integral converges then we say that F (s) is the integral transform of f (t). The function
K(s, t) is called the kernel of the transform. When K(s, t) = e−st the transform is called the Laplace
Transform.

Definition 3.2. Laplace Transform: The Laplace Transform of a function f (t) is defined to be

Z ∞
L{f (t)} = F (s) = e−st f (t) dt (4)
0

The Laplace Transform will turn out to be useful when solving ordinary differential equations (ODEs).
Interestingly, the Laplace Transform of the Dirac Delta Function turns out to be

R∞
L{δa (t)} = e−st δa (t)dt # Equation 4 with f (t) = δa (t)
R0∞
= 0
g(t)δa (t)dt # set g(t) = e−st
= g(a) # by Equation 3
−sa
= e # g(a) = e−sa

3.1 The Linearity Property of the Laplace Transform

Definition 3.3. Linearity Property: L{af (t) + bg(t)} = aL{f (t)} + bL{g(t)}

Interestingly, L is what is called a linear operator in vector space parlance [6].

All good, but why does L have this property? Here’s one way to think about it:

R∞
e−st af (t) + bg(t) dt
 
L{af (t) + bg(t)} = 0
# definition of the Laplace Transform (Definition 3.2)
R ∞ −st R∞
= 0
e af (t)dt + 0 e−st bg(t)dt # by the linearity of improper integrals [3]
R∞ R∞
= a 0 e−st f (t)dt + b 0 e−st g(t)dt # neither a nor b depends on t

= aL{f (t)} + bL{g(t)} # definition of the Laplace Transform (Definition 3.2)

4
3.2 So does every function have a Laplace Transform?
The answer is no (consider a function like f (t) = t−1 ). Ok, then what are the properties that f (t) must have
in order to have a Laplace Transform? First, f must be of ”exponential order”.

Definition 3.4. Exponential Order: A function f is said to be of exponential order c if there exist
constants c, M, and T > 0 such that |f (t)| ≤ M ect for all t > T .

Said another way, in order for f (t) to be of exponential order c we require that

f (t)
lim =0
t→∞ ect

Basically this is saying that in order for f (t) to have a Laplace Transform then in a race between |f (t)| and
ect as t → ∞ ect must approach its limit first. This situation is depicted in Figure 3.

Figure 3: f is of exponential order with constants c, M and T

5
Lemma 3.1. Triangle Inequality for Integrals: The Triangle Inequality for Integrals states that
Z b Z b
f (t) dt ≤ f (t) dt (5)
a a

for f (t) continuous on the interval [a, b].

Proof. There are two cases: one in which t is real and one in which t is complex.

Case 1: a, b ∈ R and f : [a, b] → R

First, recall that

|x| ≤ c ⇔ −c ≤ x ≤ c (6)

We will use Equation (6) below.

Now, we know that −|f (t)| ≤ f (t) ≤ |f (t)| from the definition of absolute value. If we then take the integral
of both (all) sides, we get
Z b Z b Z b
− |f (t)| dt ≤ f (t) dt ≤ |f (t)| dt
a a a

Rb Rb
If we substitute x = f (t) dt and c = |f (t)| dt in Equation (6) we get
a a

Z b Z b
f (t) dt ≤ f (t) dt
a a

That is, Equation (5). 

Case 2: f : [a, b] → C
Rb
First, notice that since a f (t) dt is a complex number we know that we can represent the integral using
polar coordinates. That is
Z b
f (t) dt = reiθ (7)
a
for some constants r, i and θ.

Why does this hold? First, recall that any complex number z can be represented as z = a+bi. Then consider
the complex plane, shown in Figure 4:

z = a + bi = r cos θ + ir sin θ = reiθ

z = a + bi
b
∗
√ zz
| =
|z r sin θ
r=
θ

O r cos θ a
R

Figure 4: The Complex Plane

6
This picture makes it easy to see why z = reiθ . Specifically

z = a + ib # definition of a point z in the complex plane

= r cos θ + ir sin θ # switch to polar coordinates: a = r cos θ and b = r sin θ
= r [cos θ + i sin θ] # factor out r
iθ
= re # Euler’s formula [4]: eiθ = cos θ + i sin θ

Next, if we multiply both sides of Equation (7) by e−iθ we see that

Rb
e−iθ a
f (t) dt = e−iθ reiθ # multiply both sides of Equation (7) by e−iθ
Rb
⇒ e−iθ a f (t) dt = re−iθ eiθ # multiplication is commutative
Rb
⇒ e−iθ a f (t) dt = r # e−iθ eiθ = 1
Rb
⇒ a e−iθ f (t) dt = r # e−iθ is a constant with respect to t

We also know that

Rb
a
f (t) dt = reiθ # take the absolute value of both sides of Equation (7)

= |r||eiθ | # |xy| = |x||y|

= |r| # |eiθ | = 1
= r #r∈R

Rb Rb
Since a
e−iθ f (t) dt = r and a
f (t) dt = r we have

Z b Z b
f (t) dt = e−iθ f (t) dt (8)
a a

Since for |x| ∈ R the LHS of Equation (8) is a real number. Interestingly, the integral on the RHS of Equation
(8) is also real, but the integrand is complex. So if we look at the real part of the RHS we see that

"Z #
b Z b h i
−iθ
Re e−iθ f (t) dt # swap Re and , noting that e−iθ f (t) ∈ C
R  
Re e f (t) dt =
a a
Z b
≤ e−iθ f (t) dt # since Re(z) ≤ |z| for z ∈ C and Re(dt) = dt
a
Z b
= e−iθ f (t) dt # |xy| = |x||y|
a
Z b
= e−iθ f (t) dt # e−iθ doesn’t depend on t
a
Z b
= f (t) dt # |e−iθ | = 1
a

Z b Z b
and so f (t) dt ≤ f (t) dt. 
a a

7
Now we can answer the question of when f has a Laplace Transfrom:

Theorem 3.1. Existence Theorem for Laplace Transforms: If f is s piecewise continuous on the
interval [0, ∞) and is of exponential order c then F (s) = L{f (t)} is defined for all s > c.

Ok, but why? Consider

L{f (t)} = F (s) # definition of the Laplace Transform (Definition 3.2)

Z ∞
= e−st f (t)dt # definition of the Laplace Transform (Definition 3.2)
0
Z ∞
≤ e−st M ect dt # f is of exponential order c (Definition 3.4)
0
Z ∞
= M e−st ect dt # M doesn’t depend on t
0
Z ∞
= M e−st+ct dt # xn · xm = xn+m
0
Z ∞
= M e(c−s)t dt # e−st+ct = ect−st = e(c−s)t
0
Z ∞
= M eu dt # use a u substitution with u = (c − s)t
0
Z ∞
du du
= M eu c−s # u = (c − s)t ⇒ du = (c − s)dt ⇒ dt =
0 c−s
h iZ ∞
M
= c−s eu du # neither c nor s depends on t
0
h i ∞
M
eu eu du = eu + C and the FToC [7]
R
= c−s #
0
h i ∞
M
= c−s e(c−s)t # u = (c − s)t
0
h i
M (c−s)d M (c−s)0
= lim c−s e − c−s e # evaluate at the limits
d→∞
h i
M (c−s)d M M M
= lim c−s e − c−s # e(c−s)0 = e0 = 1 and c−s ·1= c−s
d→∞
h i
= 0− cM
−s # s > c ⇒ c − s < 0 ⇒ lim M (c−s)d
c−s e =0
d→∞
M
= s−c # simplify

h i
So if s = c then s M
−c is not defined and L{f (t)} does not exist. Similarly, if s < c then lim M (c−s)d
e
d→∞ c−s
does not converge and L{f (t)} does not exist. All of this implies that functions that do not satisfy the
conditions of the Existence Theorem do not have Laplace Transforms.

8
3.3 Inverse Laplace Transform
Definition 3.5. Inverse Laplace Transform: If F (s) represents the Laplace Transform of a function f (t)
such that L{f (t)} = F (s) then the Inverse Laplace Transform of F (s) is f (t), i.e. L−1 {F (s)} = f (t).

Here are a few examples:

Laplace Transform Inverse Laplace Transform

f (t) F (s) = L{f (t)} F (s) f (t) = L−1 {F (s)}

1 1
1 s s 1

n! n!
tn sn+1 sn+1 tn

1 1
eat s−a s−a eat

k k
sin(kt) s2 +k2 s2 +k2 sin(kt)

s s
cos(kt) s2 +k2 s2 +k2 cos(kt)

k k
sinh(kt) s2 −k2 s2 −k2 sinh(kt)

s s
cosh(kt) s2 −k2 s2 −k2 cosh(kt)

dg dg
sG(s) − g(0) sG(s) − g(0)
dt dt

9
3.4 Ok, then what is the Laplace Transform of f 0 (t)?
Suppose f (t) is continuous, piecewise smooth and of exponential order, and suppose f 0 (t) is the derivative
of f (t). Then the Laplace Transform of f 0 (t), L{f 0 (t)}, turns out to be

Z ∞
0
L{f (t)} = e−st f 0 (t)dt # definition Laplace Transform (Definition 3.2)
0

Z ∞
= e−st f 0 (t)dt # use integration by parts
0
v du
∞ Z ∞ Z ∞ Z ∞
−st −st
= e f (t) − (−s)e f (t) dt # v du = uv − u dv
0 0 0 0
v u dv u
h i Z ∞
= lim e−sd f (d) − e−s·0 · f (0) − (−s)e−st f (t)dt # evaluate first term at the limits
d→∞ 0

Z ∞ h i
= 0 − e−s·0 · f (0) − (−s)e−st f (t)dt # lim e−sd f (d) = 0
0 d→∞

Z ∞
= 0 − f (0) − (−s)e−st f (t)dt # e−s·0 = e0 = 1 and 1 · f (0) = f (0)
0

Z ∞
= −f (0) + s e−st f (t)dt # s doesn’t depend on t and simplify
0

= −f (0) + sL{f (t)} # definition Laplace Transform (Definition 3.2)

= sL{f (t)} − f (0) # rearrange

= sF (s) − f (0) # definition Laplace Transform (Definition 3.2)

One of the interesting things to note here is that by using the Laplace Transform we’ve taken a statement
about the derivative of f , namely L{f 0 (t)}, and converted it into a statement about f itself: sL{f (t)}−f (0).
That is, we’ve converted a differential equation into an algebraic one. This will come in handy later when
we want to solve differential equations.

10
3.5 What about L{f 00 (t)}?

Once we know how to compute L{f 0 (t)} it is pretty easy to see how to compute L{f 00 (t)}:

L{f 00 (t)} = L{g 0 (t)} # set g(t) = f 0 (t) ⇒ g 0 (t) = f 00 (t)

= sL{g(t)} − g(0) # L{g(t)0 } = sL{g(t)} − g(0) (Section 3.4)
= sL{f 0 (t)} − f 0 (0) # g(t) = f 0 (t)

s sF (s) − f (0) − f 0 (0) # L{f (t)0 } = sF (s) − f (0) (Section 3.4)

 
=

= s2 F (s) − sf (0) − f 0 (0) # simplify

3.6 So what is the general form of L{f (n) (t)}?

The general form of the Laplace Transform of the nth derivative of some function f (t) is

n
X
L{f (n) (t)} = sn F (s) − sn−k f (k−1) (0) (9)
k=1

Here f is assumed to be n-times differentiable and that it’s nth derivative, denoted f (n) , is piecewise contin-
uous and of exponential order. The result then follows by mathematical induction. Note that f (0) , the 0th
derivative of f , is just f .

So for example, for the Laplace Transform of the second derivative (n = 2) of some function f Equation 9
tells us

2
L{f (2) (t)} = s2 F (s) − s2−k f (k−1) (0)
P
# Equation 9 with n = 2
k=1
= s2 F (s) − s 2−1 (1−1)
f (0) − s2−2 f (2−1) (0) # expand terms

= s2 F (s) − s1 f (0) (0) − s0 f (1) (0) # arithmetic

= s2 F (s) − sf (0) − f (1) (0) # s1 = s, f (0) = f and s0 = 1

= s2 F (s) − sf (0) − f 0 (0) # alternate notation (Section 3.5)

11
3.7 What is the derivative of F (s)?

d
F 0 (s) = F (s) # switch to more a convenient notation
ds
Z ∞
d
= e−st f (t)dt # definition of the Laplace Transform (Definition 3.2)
ds 0

Z ∞ Z
∂ −st d
= e f (t)dt # swap with by the Leibniz integral rule
0 ∂s ds
Z ∞
∂ −st
= (−t)e−st f (t)dt # e = −te−st by the chain rule
0 ∂s
Z ∞
= e−st (−t)f (t)dt # rearrange
0

Z ∞
= e−st g(t)dt # set g(t) = −tf (t)
0

= L{g(t)} # definition of the Laplace Transform (Definition 3.2)

= L{−tf (t)} # g(t) = −tf (t)

= −L{tf (t)} # −1 doesn’t depend on t (see Section 3.1)

So F 0 (s) = −L{tf (t)}. We can also pretty easily see that in general F (n) (s) = (−1)n L{tn f (t)}.

3.8 Example: Solving Ordinary Differential Equations (ODEs)

Suppose we have the following Initial Value Problem (IVP) [5]:

y 00 − y 0 − 2y = 0 with y(0) = 1 and y 0 (0) = 0 (10)

We can use the Laplace Transform to solve this ODE. The basic idea is to take the Laplace Transform of
both sides of Equation 10, set Y (s) = L{y}, and then solve for Y (s). Then we can find y(t) by taking the
Inverse Laplace Transform of Y (s).

12
 y 00 − y 0 − 2y = 0 # ODE we want to solve (Equation 10)

⇒ L{y 00 − y 0 − 2y} = L{0} # apply L{·} to both sides

⇒ L{y 00 } − L{y 0 } − L{2y} = L{0} # L{·} is a linear operator (Section 3.1)

⇒ L{y 00 } − L{y 0 } − 2L{y} = 0 # L{af (t)} = aL{f (t)} and L{0} = 0

s L{y} − sy(0) − y 0 (0) − L{y 0 } − 2L{y} = 0 # L{y 00 } = . . . (Section 3.5)

 2 
⇒

s Y (s) − sy(0) − y 0 (0) − L{y 0 } − 2L{y} = 0

 2 
⇒ # set Y (s) = L{y}

s Y (s) − sy(0) − y 0 (0) − sY (s) − y(0) − 2L{y} = 0 # L{y 0 } = sY (s) − y(0) (Section 3.4)
 2   
⇒

s Y (s) − sy(0) − y 0 (0) − sY (s) − y(0) − 2Y (s) = 0

 2   
⇒ # L{y} = Y (s)

⇒ s2 Y (s) − s · 1 − 0 − sY (s) + 1 − 2Y (s) = 0 # IVP: y(0) = 1 and y 0 (0) = 0

⇒ s2 Y (s) − s − sY (s) + 1 − 2Y (s) = 0 # simplify

⇒ s2 Y (s) − sY (s) − 2Y (s) − (s − 1) = 0 # collect terms

 
⇒ Y (s) s2 − s − 2 − (s − 1) = 0 # factor out Y (s)
 
⇒ Y (s) s2 − s − 2 = s − 1 # add s − 1 to both sides

s−1
⇒ Y (s) = # solve for Y (s)
s2 −s−2

s−1
So Y (s) = . Now we can split Y (s) using partial fractions and solve for y(t) using the Inverse
s2 − s − 2
Laplace Transform:

13
 s−1
Y (s) = # see above
s2 −s−2
s−1
= # factor denominator
(s − 2)(s + 1)
s−1 A B
⇒ = + # use partial fractions
(s − 2)(s + 1) (s − 2) (s + 1)

⇒ s − 1 = A(s + 1) + B(s − 2) # multiply both sides by (s − 2)(s + 1)

⇒ −1 = A − 2B # coefficients of 1

⇒ 1=A+B # coefficients of s

⇒ A=1−B # solve for A in previous equation

⇒ −1 = (1 − B) − 2B # plug A = 1 − B into −1 = A − 2B

⇒ −2 = −3B # simplfy

2
⇒ B= 3 # solve for B

1
⇒ A= 3 # plug B into A = 1 − B

1 2
3 3
⇒ Y (s) = +
s−2 s+1

So now we know that

1 2
3 3
Y (s) = +
s−2 s+1

and we also know (Section 3.3) that

 
1
L−1 = eat
s−a

Putting this all together

14
 Y (s) = L{y(t)} # definition of Y (s) (Section 3.8)

⇒ L−1 {Y (s)} = L−1 {L{y(t)}} # take the Inverse Laplace Transform of both sides

⇒ L−1 {Y (s)} = y(t) # L−1 {L{y(t)}} = y(t)

 1 2  1 2
−1 3 3 3 3
⇒ L + = y(t) # Y (s) = + (Section 3.8)
s−2 s+1 s−2 s+1
 1   2 
⇒ L−1 3
+ L−1 3
= y(t) # by the linearity of the Laplace Transform (Section 3.1)
s−2 s+1
   
1 −1 1 2 −1 1 1 2
⇒ 3 L + 3 L = y(t) # neither 3 nor 3 depends on t
s−2 s+1
 
2 −1 1 n o
⇒ 1 2t
3e + 3L = y(t) # L−1 1
s−2 = e2t (Section 3.3)
s+1
n o
⇒ 1 2t
3e + 32 e−t = y(t) # L−1 1
s+1 = e−t (Section 3.3)

⇒ y(t) = 31 e2t + 32 e−t # = is symmetric

So we’ve been able to use the Laplace Transform to solve the ODE in Section 3.8 (Equation 10):

1 2t 2 −t
y(t) = e + e
3 3

4 The Heaviside Function, Convolution, and the Laplace Transform

We saw in Section 3.1 that the Laplace Transform has a linearity property, namely,

L{af (t) + bg(t)} = aL{f (t)} + bL{g(t)}

One natural question is whether there is a ”linearity-like” property for multiplication. Unfortunately

L{f (t)g(t)} =
6 L{f (t)}L{g(t)}

However, convolution behaves better with respect to the Laplace Transform of a product. Specifically

L{f (t) ∗ g(t)} = L{f (t)}L{g(t)} (11)

where ∗ is the convolution operator.

To see why Equation 11 holds we need another piece of machinery, the Heaviside function.

15
4.1 The Heaviside Function
The most basic definition of the Heaviside function is

Definition 4.1. Heaviside Function:

(
0 t<0
u(t) =
1 t≥0

The Heaviside function is shown in Figure 5.

u(t)
(
0 t<0
u(t) =
1 t≥1

t
0

Figure 5: The Heaviside Function u(t)

More frequently we are interested in u(t − a), which is sometimes written as ua (t), is shown in Figure 6.

u(t − a)
(
0 t<a
u(t − a) =
1 t≥a

a t

Figure 6: u(t − a)

Another useful version of the Heaviside function, u(a − t), is shown in Figure 7.

16
 u(a − t)
(
0 t>a
u(a − t) =
1 t≤a

a t

Figure 7: u(a − t)

Now the obvious question is how do we compute L{u(t − a)}? Recall that by the definition of the Laplace
Transform we have
Z ∞
F (s) = e−st u(t − a)dt
0

The first thing to observe is that u(t − a) is zero for t < a and one otherwise (Figure 6), so we can rewrite
our integral with a lower limit of a.

Z ∞
F (s) = e−st u(t − a)dt # definition of L{u(t − a)} (Section 3)
0
Z ∞
= e−st u(t − a)dt # we can change the lower integration limit since u(t − a) = 0 for t < a
a
Z ∞
= e−st · 1dt # u(t − a) = 1 for t ≥ a
a
Z ∞
= e−st dt # simplify
a
∞
e−st
Z
ecx
= # ecx dx = c + C and the Fundamental Theorem of Calculus
−s a

e−sd e−sa
= lim − # evaluate at limits
d→∞ −s −s

e−sa e−sd
= 0− # lim =0
−s d→∞ −s

e−sa e−sa
= # L{u(t − a)} =
s s

4.2 What about L{u(t − a)f (t − a)}?

What can we say about the Laplace Transform of u(t − a)f (t − a), that is, the Heaviside function multiplied
by another function? It turns out that

17
 e−as F (s) = L{u(t − a)f (t − a)}

Ok, but why? First, notice that

(
0 t<a
u(t − a)f (t − a) =
f (t − a) t ≥ a

One way to think about this is that u(t − a) ”turns on” f (t − a) at a. So consider

Z ∞
F (s) = e−sτ f (τ )dτ # definition of Laplace Transform (Section 3)
0
Z ∞
⇒ e−as F (s) = e−as e−sτ f (τ )dτ # multiply both sides by e−as
0
Z ∞
⇒ e−as F (s) = e−as e−sτ f (τ )dτ # move e−as inside integral (e−as doesn’t depend on τ )
0
Z ∞
⇒ e−as F (s) = e−sτ −sa f (τ )dτ # an · am = a(m+n)
0
Z ∞
−as
⇒ e F (s) = e−s(τ +a) f (τ )dτ # factor out s
0
Z ∞
−as
⇒ e F (s) = e−st f (t − a)dt # substitution: t = τ + a so τ = t − a and dτ = dt
a
Z ∞
⇒ e−as F (s) = e−st u(t − a)f (t − a)dt # u(t − a) = 0 for t < a so adjust lower limit to 0
0

⇒ e−as F (s) = L{u(t − a)f (t − a)} # definition of Laplace Transform (Section 3)

4.3 Convolution
Next we need the definition of the convolution of two functions f and g.

Definition 4.2. Convolution:

Z t
f (t) ∗ g(t) = f (τ )g(t − τ )dτ (12)
0

where ∗ is the convolution operator. Note that the righthand side of Equation 12 is some function of t (τ is
the dummy variable for integration). That is

Z t
f (t) ∗ g(t) = f (τ )g(t − τ )dτ
0
some function of t

Convolution has quite a few interesting properties including:

18
 • Commutativity: f ∗ g = g ∗ f
• Associativity: f ∗ (g ∗ h) = (f ∗ g) ∗ h
• Distributivity: f ∗ (g + h) = f ∗ g + f ∗ h

It is pretty easy to see why the commutative property holds:

Z t
f (t) ∗ g(t) = f (τ )g(t − τ )dτ # definition of convolution (Section 4.3)
0
Z t
= − f (t − u)g(u)du # use a u substitution with u = t − τ so τ = t − u and dτ = −du
0
Z 0
= − f (t − u)g(u)du # lower limit: u = t − 0 = t, upper limit: u = t − t = 0
t
Z t
= f (t − u)g(u)du # by the Fundamental Theorem of Calculus
0
Z t
= g(u)f (t − u)du # multiplication is commutative
0

= g(t) ∗ f (t) # definition of convolution (Section 4.3)

19
4.4 Ok, why does L{f (t) ∗ g(t)} = L{f (t)}L{g(t)} hold?

Now we can use the machinery we’ve built up to show why Equation 11 holds:

Z ∞
e−st f (t) ∗ g(t) dt
 
L{f (t) ∗ g(t)} = # definition of Laplace Transform (Definition 3.2)
0
"Z #
Z ∞ t
= e−st f (t − v)g(v)dv dt # definition of convolution (Section 4.3)
0 0
"Z #
Z ∞ ∞
= e−st f (t − v)g(v)u(t − v)dv dt # u(t − v) = 0 for v > t so change upper limit
0 0
"Z #
Z ∞ ∞
= e−st f (t − v)g(v)u(t − v)dv dt # e−st doesn’t depend on v
0 0
"Z #
Z ∞ ∞
= e−st f (t − v)g(v)u(t − v)dt dv # swap order of integration
0 0
"Z #
Z ∞ ∞
= g(v) e−st f (t − v)u(t − v)dt dv # g(v) doesn’t depend on t
0 0
Z ∞
= g(v) e−sv F (s) dv # L{f (t − v)u(t − v)} = e−sv F (s) (Section 4.2)
0
Z ∞
= F (s) g(v) e−sv dv # F (s) doesn’t depend on v
0
Z ∞
= F (s) e−sv g(v)dv # multiplication is commutative
0
Z ∞
= F (s) G(s) # G(s) = e−sv g(v)dv (Definition 3.2)
0

⇒ L{f (t) ∗ g(t)} = L{f (t)}L{g(t)} # the LT of a convolution is the product of LTs

This result also implies that L−1 {F (s)G(s)} = f (t) ∗ g(t). It is pretty easy to see why:

L{f (t) ∗ g(t)} = L{f (t)}L{g(t)} # above result

⇒ L−1 {L{f (t) ∗ g(t)}} = L−1 {L{f (t)}L{g(t)}} # take the inverse LT of both sides

⇒ f (t) ∗ g(t) = L−1 {L{f (t)}L{g(t)}} # L−1 {L{f (t)} = f (t)

⇒ f (t) ∗ g(t) = L−1 {F (s)G(s)} # L{f (t)} = F (s) (Definition 3.2)
⇒ L−1 {F (s)G(s)} = f (t) ∗ g(t) # = is symmetric

5 Conclusions
6 Acknowledgements
Thanks to Dave Neary for catching a typo: was −f (0) + L{f (t)}, should be −f (0) + sL{f (t)}.

20
LATEX Source
References
[1] Leonard Susskind. Lecture 4 — Modern Physics: Quantum Mechanics (Stanford). [Link]
[Link]/watch?v=oWe9brUwO0Q&t=1130s, 2008. [Online; accessed 11-May-2021].

[2] Proof Wiki Contributors. Mean Value Theorem For Integrals. [Link]
Value_Theorem_for_Integrals, 2020. [Online; accessed 11-May-2021].
[3] Tom Lewis. Improper Integrals. [Link]
2014. [Online; accessed 25-May-2021].
[4] Wikipedia Contributors. Euler’s Formula — Wikipedia, The Free Encyclopedia. [Link]
[Link]/w/[Link]?title=Euler%27s_formula&oldid=866429907, 2018. [Online; accessed
11-November-2018].
[5] Wikipedia Contributors. Initial value problem — Wikipedia, The Free Encyclopedia. [Link]
[Link]/w/[Link]?title=Initial_value_problem&oldid=1025065735, 2021. [Online; ac-
cessed 9-June-2021].

[6] Wikipedia Contributors. Linear Map — Wikipedia, The Free Encyclopedia. [Link]
org/w/[Link]?title=Linear_map&oldid=1024503856, 2021. [Online; accessed 6-June-2021].
[7] Wolfram MathWorld. Fundamental Theory of Calculus. [Link]
[Link], 2023. [Online; accessed 8-May-2023].

21