Lecture 29 Convolution

11/21/2011

Review.
• Solution procedure of Laplace transform method:
1. Transform the equation: left hand side and the right hand side;
2. Obtain Y (s);
3. Take inverse transform y(t) = L−1{Y }.
• What are involved in transforming the right hand side:
1. Table of Laplace transform of basic functions;
2. L{f (t) u(t − a)} = e−as L{f (t + a)} when the right hand side is piecewise defined;
3. L{f (t) δ(t − a)} = e−as f (a) when the right hand side involves the impulse function.
• What are involved in the inverse transform:
1. Table of Laplace transforms;
2. L−1{e−as F (s)} = u(t − a) f (t − a).
• A remark about the smoothness of solution:
1. When the right hand side is smooth, the solution is smooth;
2. When the right hand side is piecewise defined, the solution is at least C 1 (meaning: y and y ′
are continuous functions);
3. When the right hand side involves δ, the solution is continuous – y ′ is not continuous anymore.
Convolution.
• A “binary” operation: Two inputs and one output. Other examples of binary operations include +, −,
×, /. Each of them takes two functions as inputs and give out one single function as the output.
• Convolution is much harder to compute than the four basic operations. In particular, don’t confuse
convolution with multiplication!
• Given two functions f (t) and g(t), their convolution is defined as

6
Z t
h(t) {f ∗g }(t) = f (t − τ ) g(τ ) dτ . (1)
0

• Basic properties of convolution:

◦ f ∗g = g∗f ;
◦ f ∗(g1 + g2) = f ∗g1 + f ∗g2;
◦ (f ∗g)∗h = f ∗(g∗h);
◦ f ∗0 = 0∗f = 0.
Proofs of the above are left as homework.
Note: Don’t confuse convolution with multiplication! In particular, f ∗1  f !

Example 1. Compute et∗1 and cos t∗1.

Solution. By definition we have Z t
et∗1 = eτ 1 dτ = et − 1; (2)
0
Z t
(cos t)∗1 = cos (τ ) dτ = sin t. (3)
0

1
2 Lecture 29 Convolution

• Why bother with convolution:

Theorem 2. If L{f } = F , L{g } = G, then L −1{F (s) G(s)} = f ∗g.

Proof. All we need to show is

Z ∞
e−st (f ∗g)(t) dt = F (s) G(s). (4)
0
We have
t
Z ∞ Z ∞
Z 
−st −st
e (f ∗g)(t) dt = e f (t − τ ) g(τ ) dτ dt
0
Z0 ∞ Z t
0

= [e−st f (t − τ ) g(τ )] dτ dt. (5)

0 0

We now switch the order of the integration. To do this we have to first understand what the region
of integration look like. We have t running from 0 to ∞, and τ from 0 to t. This is the triangular
region 0 < τ < t < ∞ in the t − τ plane:
τ

To switch the order of integration, we first check what values can τ take? The answer is any value
from 0 to ∞. Once τ is given, t can only run from τ to ∞. So we have
Z ∞Z t Z ∞Z ∞
[e−st f (t − τ ) g(τ )] dτ dt = [e−st f (t − τ ) g(τ )] dt dτ . (6)
0 0 0 τ

Remark. Another way of doing this is to first observe: 0 < t < ∞ together with 0 < τ < t is the same
as 0 < τ < t < ∞. Now ignore t we have 0 < τ < ∞, once τ is given, the range for t is τ < t < ∞.

Further calculate:
Z ∞Z ∞ Z ∞ Z ∞ 
[e−st f (t − τ ) g(τ )] dt dτ = g(τ ) e−st f (t − τ ) dt dτ
0 τ
Z0 ∞  Zτ ∞ 
−sτ −s(t−τ )
= g(τ ) e e f (t − τ ) dt dτ
Z0 ∞ τ 
Z ∞ 
= e−sτ g(τ ) e−s(t−τ ) f (t − τ ) dt dτ . (7)
0 τ

Setting t ′ = t − τ in the t-integral, we have

Z ∞ Z ∞
e−s(t−τ ) f (t − τ ) dt =
′
e−st f (t ′) dt ′ = F (s). (8)
τ 0
Therefore we reach
Z ∞ Z ∞
 Z ∞
e−sτ g(τ ) e−s(t−τ ) f (t − τ ) dt dτ = e−sτ g(τ ) F (s) dτ = F (s) G(s) (9)
0 τ 0

and the proof ends. 

 11/21/2011 3

• Applications of convolution (From least to most important):

◦ Find Laplace transforms for integrals:
R t
Example 3. Find the Laplace transform of f (t) = 0 (t − τ )2 cos 2 τ dτ .
Solution. First recognize:
Z t
(t − τ )2 cos 2 τ dτ = t2∗(cos 2 t). (10)
0
Thus we have
F (s) = L{t2∗(cos 2 t)}
= L{t2} L{cos 2 t}
2 s 2
= 3 2 = 2 2 . (11)
s s + 4 s (s + 4)

◦ Alternative way of computing inverse Laplace transform:

2
Example 4. Find the inverse Laplace transform of s2 (s2 + 4) .
Solution. Write
2 1 2
= . (12)
s2 (s2 + 4) s2 s2 + 4
Thus we have
     
2 1 2
L −1 =L −1 ∗L −1 = t∗(sin 2 t). (13)
s2 (s2 + 4) s2 s2 + 4
Note: It’s not OK to stop here!
Compute
Z t
t∗(sin 2 t) = (t − τ ) sin (2 τ ) dτ
0 Z
1 t
= − (t − τ ) dcos (2 τ )
20
N
Z t 
1 τ =t
= − (t − τ ) cos (2 τ ) τ =0 − cos (2 τ ) d(t − τ )
2
 Z t  0
1
= − −t + cos (2 τ ) dτ
2 0
t 1
= − sin (2 t). (14)
2 4
Check  
t 1 1 1 2 2
L − sin (2 t) = 2 − 2
= 2 2 . (15)
2 4 2s 4 s + 4 s (s + 4)

◦ Obtain solution formula for general right hand sides.

Example 5. Express the solution of the following initial value problem in terms of a convo-
lution integral:
y ′′ + 4 y ′ + 4 y = g(t); y(0) = 0, y ′(0) = 0. (16)
Solution.
First transform the equation:
L{y ′′} = s2 Y − s y(0) − y ′(0) = s2 Y ; (17)
L{y ′} = s Y − y(0) = s Y (18)
Denoting L{g } = G(s), we have the transformed equation as
(s2 + 4 s + 4) Y = G(s). (19)
4 Lecture 29 Convolution

So
G(s)
Y= . (20)
s2 + 4 s + 4
Now take inverse transform:
    Z t
G(s) 1
y = L −1
= L −1 −1
∗L {G} = (e −2t
t)∗g = e−2(t−τ ) (t −
(s + 2)2 (s + 2)2 0
τ ) g(τ ) dτ . (21)

Remark 6. Once we obtain this formula, we can get some information of the solution without
really compute them. For example, if we are further told that g is bounded, that is there is a
constant M such that |g | 6 M , then we can immediately compute
Z t   
2 t + 1 −2t
e−2(t−τ ) (t − τ ) dτ  = M 1 −
 
|y | 6 M  e . (22)
0 4
In particular we can conclude |y | share the same bound M as g without requiring detailed
information of g.