MATH2065: INTRO TO PDEs

Semester 2, 2009
Tutorial Solutions (Week 4)

1. (a) Notice that the voltage input is zero for 0 < t < a, is V0 between a and b, and is
zero thereafter – so this is a square pulse. Denoting I(s) = L {i(t)}, and applying the
Laplace transform to the differential equation, we get
 −as
e−bs

e
L [sI(s) − i(0)] + R I(s) = V0 − ,
s s

from which (since i(0) = 0), we obtain

V0
e−as − e−bs .


I(s) =
s (Ls + R)
Now, it is necessary to invert the Laplace transform. Do not get confused with all the
parameters – remember that everything except s is a constant. The term we need to
be able to invert is 1/ (s(Ls + R)). So we find its partial fraction decomposition
   
1 1 1 L 1 1 1
= − = − .
s (Ls + R) R s Ls + R R s s + R/L
Hence,  
V0 1 −as −bs

1 −as −bs


I(s) = e −e − e −e .
R s s + R/L
Now, we note that
   
−1 1 −1 1
L =1 and L = e−Rt/L ,
s s + R/L
and hence we can invert each term to get the final solution:
V0 
[H(t − a) − H(t − b)] − H(t − a)e−R(t−a)/L − H(t − b)e−R(t−b)/L .
 
i(t) =
R
(b) If V (t) = V0 δ(t − a) instead, its Laplace transform is V0 e−as . This needs to replace the
right-hand side of the first equation in part (a), leading to
 
V0 −as V0 1
I(s) = e = e−as .
Ls + R L s + R/L

If the exponential term was absent, the inverse Laplace transform would be (V0 /L)e−Rt/L .
The exponential term can be handled using the shifting property of Laplace transforms,
and hence the solution is
V0
i(t) = H(t − a)e−R(t−a)/L .
L
The current therefore remains zero in the circuit until the time of the spike, at which
point it jumps to V0 /L, decaying exponentially thereafter.
Some comments: The advantage in using Laplace transforms rather than attempting
to solve the ODE using our earlier methods is that the solution can be written using
 one expression. Using standard methods for (a), we would have had to separately
solve for the various domains 0 < t < a, a < t < b and t > b, and have to worry
about the ‘initial’ conditions necessary for each such domain. For example, we would
need to determine i(a) as an initial condition for the second domain, based on our
solution for the first domain. Laplace transforms takes care of all this automatically.
Our other standard methods (undetermined coefficients) do not work at all when the
discontinuity is much worse, as occurs in the spike example for part (b). Laplace
transforms are incredibly powerful when we have discontinuous functions in the ODE.

2. Denote the right-hand side (the inhomogeneity) by f (t), and its Laplace transform by F (s).
We can either write it in terms of Heaviside functions (switching the functions 4e−t and 2
“on” and “off” as needed), or can compute F (s) directly using the definition of the Laplace
transform:
Z ∞
F (s) = f (t)e−st dt
Z0 8 Z ∞
−t −st
= 4e e dt + 2e−st dt
0 8
−(s+1)t −st i∞
e e i8
= 4 +2
−(s + 1) 0 −s 8
4   2
= 1 − e−8(s+1) + e−8s .
s+1 s
Therefore, when applying the Laplace transform to the ODE, we get
4   2
sY (s) − y(0) + 3Y (s) = 1 − e−8(s+1) + e−8s ,
s+1 s
where Y (s) is the Laplace transform of the solution y(t). Putting in y(0) = 1 and solving
for Y (s),
4 2 1
1 − e−8 e−8s + e−8s +
 
Y (s) = .
(s + 1)(s + 3) s(s + 3) s+3
Now, using partial fractions, we see that
   
−1 4 −1 2 2
L =L − = 2e−t − 2e−3t
(s + 1)(s + 3) s+1 s+3

and    
−1 2 2 1 1 2
L = L−1 − = (1 − e−3t ) .
s(s + 3) 3 s s+3 3
Therefore, using the t-shift property as well,

2
y(t) = 2e−t − 2e−3t − e−8 H(t − 8) 2e−(t−8) − 2e−3(t−8) + H(t − 8) 1 − e−3(t−8) + e−3t .


3

3. What we are asked to show is the following:

Z t Z t
f (t − t̄) g(t̄) dt̄ = g(t − t̄) f (t̄) dt̄ , (1)
0 0

that is, that it does not matter in which order you choose the two functions. Let us start with
the integral on the left-hand side above. Note that t is a constant as far as this integral
 is concerned, and the integration is over the dummy variable t̄. Change the integration
variable from t̄ to w, where w is defined by
w = t − t̄ .
Now, when t̄ = 0, w = t. When t̄ = t, w = 0. Moreover, dw = −dt̄. Therefore,
Z t Z 0 Z t
f (t − t̄) g(t̄) dt̄ = f (w) g(t − w) (−dw) = g(t − w) f (w) dw ,
0 t 0

where the negative sign has been used to switch the limits of integration. This expression
is exactly the same as that on the right-hand side of (1), since the w (and the t̄) are dummy
variables. That is, they actually are not present in the final expression, they are just
placeholders to indicate that the integrand is to be evaluated at all values ranging from 0
to t.

4. We first note that   

1 1 1
= .
(s2 + 1)2 s2 + 1 s2 + 1
Now, since each of these terms is equal to L {sin t}, it follows from the convolution theorem
that
 
−1 1
L = sin t ? sin t
(s2 + 1)2
Z t
= sin (t − t̄) sin t̄ dt̄
0
1 t
Z
= [cos (2t̄ − t) − cos t] dt̄
2 0
 t Z t
1 sin(2t̄ − t) 1
= − cos t dt̄
2 2 t̄=0 2 0
 
1 sin t sin(−t) 1
= − − t cos t
2 2 2 2
sin t − t cos t
= .
2
In the above, the given trigonometric identity has been used to simplify the first integral,
and the fact that the sine function is odd (that is, sin(−x) = − sin x) has been used towards
the end.

5. Let Y (s) = L {y(t)} and G(s) = L {g(t)}. Taking the Laplace transform of both sides of
the given differential equation, and utilising the initial conditions, gives
s2 Y (s) − s − 1 − Y (s) = G(s) .
Solving for Y (s), we have
   
s+1 1 1 1
Y (s) = 2 + 2
G(s) = + 2
G(s) .
s −1 s −1 s−1 s −1
Hence
   
−1 1 −1 1
y(t) = L +L G(s)
s−1 s2 − 1
 
1
= et + L−1 G(s) .
s2 − 1
 We know (from the standard Laplace transforms table) that
 
−1 1
L = sinh t ,
s2 − 1
and so we can express  
−11
L G(s) = sinh t ? g(t)
s2 − 1
by using the convolution property. Therefore,
Z t
t
y(t) = e + sinh(t − t̄) g(t̄) dt̄ .
0

This is a very powerful method of solution, in that the above indicates the solution for any
given inhomogeneity g(t). Of course, doing the actual integration above may not be easy
for complicated g. Note that we may also have written the integral in the “opposite” way,
since f ? g = g ? f , that is, in the alternative form
Z t
t
y(t) = e + g(t − t̄) sinh(t̄) dt̄ ,
0

which is also correct.

6. Notice that Z t Z t
f (t̄) dt̄ = g(t − t̄) f (t̄) dt̄
0 0
for the choice of the function g(t) = 1. By the convolution property in the Laplace transform
table, the Laplace transform of the above is F (s) G(s). But since g(t) = 1, G(s) = 1/s.
Therefore, the required Laplace transform is
Z t 
F (s)
L f (t̄) dt̄ =
0 s
where F (s) is the Laplace transform of f (t).

7. (a) This corresponds to choosing a = −1, b = 3, and f (t) = t cos 3t in the definition of the
Dirac delta function. Therefore, if b ∈ (−1, 3),
Z 3 

δ(t − b) t cos 2t dt = t cos 2t = b cos 2b .
−1 t=b

On the other hand, if b < −1 or if b > 3, the Dirac delta function contributes zero
to the integrand, since the integral is over (−1, 3). Therefore, the integral yields zero.
The information we have now is that
Z 3 
b cos 2b if b ∈ (−1, 3) ,
δ(t − b) t cos 2t dt =
−1 0 if b < −1 or b > 3 .
Unfortunately, it is not possible to define the value of the integral if b is exactly on
one of the endpoints −1 or 3 (this may not be at all obvious – but the definition of
the Dirac delta function does not cover this case, which turns out to be ill-defined). A
clever way of rewriting the solution above is
Z 3
δ(t − b) t cos 2t dt = [H(b + 1) − H(b − 3)] b cos 2b .
−1
 (convince yourself that this works). This also has the advantage of not giving any
well-defined value at the problematic points b = −1 and b = 3, since the unit-step
functions H(b + 1) and H(b − 3) are respectively ill-defined at those values.
(b) Applying the definition of the Laplace transform,
Z ∞
L {δ(t − b)f (t)} = δ (t − b) f (t) e−st dt
0

−st 

= f (t) e 
t=b
−sb
= f (b) e .

However, we note that the above only works if b is within the inteval (0, ∞), since if
not, the Dirac delta function contributes nothing to the integral, which is just zero.
This can be compactly stated as

L {δ(t − b)f (t)} = H(b) f (b) e−sb ,

since the unit-step function H(b) will be zero if b < 0. Note that once again, it is
not clear what the value would be if b = 0 (this is not defined, just as the unit-step
function H(b) is not defined when b = 0).