MATH 219

Fall 2017
Lecture 22
Lecture notes by Özgür Kişisel

Content: Solution of Initial Value Problems. Step Functions

Suggested Problems:
§6.2: 1, 4, 7, 9, 12, 16, 17
§6.3: 1, 3, 8, 10, 14, 20, 23, 38
Let us now return to differential equations, after our general discussion about
Laplace transforms. Suppose that we have a linear ordinary differential equation
in y(t) together with initial conditions. By taking the Laplace transforms of both
sides of the equation, we end up with an equation for the Laplace transform Y (s).
In favourable cases, this equation is easier to solve than the original equation. After
finding Y (s), one may find y(t) provided that one can compute the inverse of the
Laplace transform operation in this particular case. This procedure works especially
well if the original linear ODE has constant coefficients. Let us now work out the
details of this procedure. We first have to understand what the effect of Laplace
transform is on the derivative of a function.

1 Laplace transform of a derivative

Theorem 1.1 Suppose that f (t) is continuous and f 0 (t) is piecewise continuous on
[0, ∞). Furthermore suppose that there exists constants K, a such that |f (t)| ≤ Keat
for all t. Then L{f 0 (t)} exists for all s > a and
L{f 0 (t)} = sL{f (t)} − f (0).
R∞ RM
Proof Recall that L{f 0 (t)} = 0 e−st f 0 (t)dt = limM →∞ 0 e−st f 0 (t)dt. Since f 0 (t)
is piecewise continuous on [0, M ], so is e−st f 0 (t). Say that the points of discontinuity
are t1 , . . . , tn . Then we can break up the last integral into pieces:
Z M Z t1 Z t2 Z tn Z M
−st 0 −st 0 −st 0 −st 0
e f (t)dt = e f (t)dt+ e f (t)dt+. . .+ e f (t)dt+ e−st f 0 (t)dt.
0 0 t1 tn−1 tn

1
Now, each of the integrands on the right hand side is continuous on the relevant
interval. Therefore we can apply integration by parts to each of the integrals on the
right hand side. Set u = e−st and dv = f 0 (t)dt. Then du = −se−st and v = f (t).
Rb Rb
Use a udv = uv|ba − a vdu for each integrand on the right hand side and get
Z M Z t1 Z t2
−st 0 −st −st
e f (t)dt = e −st
f (t)|t01 − −se f (t)dt + e f (t)|tt21 − −se−st f (t)dt
0 0 t1
Z M
+ . . . + e−st f (t)|M
tn − −se−st f (t)dt
tn
Z M
=s e−st f (t)dt + e−sM f (t) − f (0).
0

It remains to send M to ∞. The inequality |f (t)| ≤ Keat implies that e−sM f (t)
tends to 0 when M → ∞. Therefore we obtain

L{f 0 (t)} = sL{f (t)} − f (0). 

Let us assume for a moment that derivatives of f (t) of all orders satisfy similar
conditions to those in the theorem. In this case, we can apply the theorem to
derivatives of f (t) recursively and get,

L{f 00 (t)} = s2 L{f (t)} − sf (0) − f 0 (0)

L{f 000 (t)} = s3 L{f (t)} − s2 f (0) − sf 0 (0) − f 00 (0).

and so on. More precisely, we have:

Theorem 1.2 Suppose that f (t), f 0 (t) . . . f (n−1) (t) are continuous and that f (n) (t)
is piecewise continuous on [0, ∞). Suppose that there exists constants K, a such that
|f (i) (t)| ≤ Keat for all 0 ≤ i ≤ n − 1. Then L{f (n) (t)} exists for s > a and

L{f (n) (t)} = sn L{f (t)} − sn−1 f (0) − sn−2 f 0 (0) − . . . − sf (n−2) (0) − f (n−1) (0).

The proof of this theorem follows directly from the previous one by applying induc-
tion. Notice that in the folrmula, the first term is sn times the Laplace transform
of f (t). The other terms are related to the initial conditions.

2
2 Obtaining and solving the equation for Y (s)
Suppose now that we have a linear ODE for y(t). Apply L to both sides of the
equation and use the linearity of L. This gives us an equation for Y (s) = L{y(t)}.
The aim is then to solve the equation for Y (s). Let us demonstrate this procedure
through a few examples.

Example 2.1 Suppose that y 00 − 5y 0 + 4y = cos 3t, furthermore y(0) = 2 and y 0 = 0.

Find Y (s), the Laplace transform of y(t).
Solution: Apply L to both sides of the equation:

L{y 00 − 5y 0 + 4y} = L{cos 3t}

s
L{y 00 } − 5L{y 0 } + 4L{y} = 2
s +9
2 0 s
s Y (s) − sy(0) − y (0) − 5(sY (s) − y(0)) + 4Y (s) = 2
s +9
2 s 0
(s − 5s + 4)Y (s) = 2 + sy(0) + y (0) − 5y(0)
s +9
s 2s − 10
Y (s) = 2 2
+ 2 .
(s + 9)(s − 5s + 4) s − 5s + 4

Soon, we will discuss how one can recover the solution y(t) from Y (s).

Example 2.2 Suppose that y (4) +y = e−2t , furthermore, y(0) = 1, y 0 (0) = 0, y 00 (0) =
2, y 000 (0) = 0. Find the Laplace transform Y (s) of y(t).
Solution: Apply L to both sides of the equation:

L{y (4) + y} = L{e−2t }

1
L{y (4) } + L{y} =
s+2
1
s4 Y (s) − s3 y(0) − s2 y 0 (0) − sy 00 (0) − y 000 (0) + Y (s) =
s+2
1
(s4 + 1)Y (s) = + s3 + 2s
s+2
1 s3 + 2s
Y (s) = + .
(s + 2)(s4 + 1) s4 + 1

3
3 Finding y(t) given Y (s)
Our next question is whether or not we can retrieve y(t) from Y (s). We will state
the following theorem without proof:

Theorem 3.1 Suppose that f (t) and g(t) are two piecewise continuous functions
on [0, ∞) and there exist constants K, a such that |f (t)| ≤ Keat and |g(t)| ≤ Keat .
If L{f (t)} = L{g(t)} for all s > a, then f (t) = g(t) at all points which f and g are
both continuous.

In particular, the only possible places that f 6= g can be the points of discontinuity
of f or g. In every finite interval, we have only finitely many such points. So f
and g are equal “almost everywhere”. The Laplace transform is after all an integral
transformation, so it shouldn’t be expected to see differences at finitely many points.
So in a sense, this result is the best of what could be expected.
If L{y(t)} = Y (s), then we will write y(t) = L−1 {Y (s)}, ignoring the problem
about discontinuities mentioned in the previous paragraph. The transformation
L−1 is called the inverse Laplace transform. It is again linear. Just like the
integral formula for the Laplace transform, there is a similar integral formula for
the inverse Laplace transform. However, it requires the use of a line integral in
the complex plane and its evaluation often uses techniques from complex analysis
involving residue calculations. Since we are not assuming that the reader has this
background, we will do something else to compute L−1 {Y (s)} when Y (s) is “simple
enough”.
Suppose that Y (s) is a rational function, namely Y (s) = P (s)/Q(s) where P and
Q are both polynomials. We can break up this rational function into its partial
fractions. We can then use the linearity of L−1 : Find L−1 of each simple piece and
add them up.

1
Example 3.1 Find the inverse Laplace transform of .
s2 − 5s + 4
Solution:
1 1 A B
= = + .
s2 − 5s + 4 (s − 4)(s − 1) s−4 s−1
Upon equating denominators, we get (A + B)s + (−4A − B) = 1. Since this equality

4
should hold as an equality of polynomials,

A+B =0
−4A − B = 1.

From here we find A = −1/3 and B = 1/3. Therefore,

1 −1/3 1/3
= +
s2
− 5s + 4 s−4 s−1
1 1 1 1 1
L−1 { 2 } = − L−1 { } + L−1 { }
s − 5s + 4 3 s−4 3 s−1
1 1
= − e4t + et .
3 3

Example 3.2 Solve the initial value problem y 000 − y = 1, with the initial conditions
y(0) = y 0 (0) = y 00 (0) = 0.
Solution: Apply L to both sides.

L{y 000 − y} = L{1}

1
L{y 000 } − L{y} =
s
1
(s3 − 1)Y (s) =
s
1
Y (s) = .
s(s3 − 1)

Now let us break up the right hand side into its partial fractions:
1 A B Cs + D
= + + 2
s(s3− 1) s s−1 s +s+1

Upon equating denominators, we get 1 = A(s3 −1)+B(s2 +s+1)s+(Cs+D)(s−1)s.

From here, we get

A+B+C =0
B−C +D =0
B−D =0
−A = 1.

5
The solution is A = −1, B = 1/3, C = 2/3, D = 1/3. Therefore,

1 −1 1/3 2s/3 + 1/3

Y (s) = = + + 2
s(s3
− 1) s s−1 s +s+1
−1 1/3 2(s + 1/2)/3
= + + .
s s − 1 (s + 12 )2 + 3/4

Finally, applying L−1 to both sides, we get

et 2 −t/2 √
y(t) = −1 + + e cos( 3t/2).
3 3

4 Step functions
Laplace transform is especially handy for solving ODE’s with a discontinuous right
hand side. We would now like to express discontinuous functions in terms of certain
basic ones and extend our Laplace transform computations to them.

Definition 4.1 Suppose that c ≥ 0. The unit step function at time t = c is

(
0, t < c,
uc (t) =
1, t ≥ c.

Notice that uc (t) is piecewise continuous and its only discontinuity is at time t = c.
Suppose now that g(t) is a piecewise continuous function on R≥0 , with discontinuities
at the points 0 < c1 < c2 < c3 < . . . More precisely, suppose that we have a piecewise
definition of g(t) of the type:

h1 (t), t < c1 ,


h (t), c ≤ t < c ,
2 1 2
g(t) =


 h3 (t), c2 ≤ t < c3 ,

...

We claim that

g(t) = h1 (t) + uc1 (t)(h2 (t) − h1 (t)) + uc2 (t)(h3 (t) − h2 (t)) + uc3 (t)(h4 (t) − h3 (t)) + . . .

6
This can be verified easily: Just check that the left and right hand sides agree for all
values of t. First, suppose that t < c1 . Then only the first summand in the right hand
side is nonzero and it is h1 (t). Next, suppose that c1 ≤ t < c2 . Then uc1 (t) = 1 but
all other ucj (t) = 0 for j ≥ 2. Thus we get h1 (t) + (h2 (t) − h1 (t)) = h2 (t) on the right
hand side. Progressing in this way, if ci ≤ t < ci+1 , then uc1 (t) = . . . = uci (t) = 1
and uci+1 (t) = uci+2 (t) = . . . = 0. Therefore, the right hand side becomes

h1 (t) + (h2 (t) − h1 (t)) + (h3 (t) − h2 (t)) + . . . + (hi+1 (t) − hi (t)) = hi+1 (t).

This agrees with the given expression for g(t).

Example 4.1 Suppose that g(t) is the function given in piecewise form by


 0, t < 2,

t − 2, 2 ≤ t < 3,
g(t) =


 4 − t, 3 ≤ t < 4,
0, t ≥ 4.


By the argument above, we have

g(t) = 0 + u2 (t)(t − 2 − 0) + u3 (t)(4 − t − (t − 2)) + u4 (t)(0 − (4 − t))

= (t − 2)u2 (t) + (6 − 2t)u3 (t) + (t − 4)u4 (t).

Example 4.2 Suppose that g(t) is the square-wave function whose value is 1 if
2n ≤ t < 2n + 1 and 0 if 2n − 1 ≤ t < 2n for every nonnegative integer n. Notice
that this time g(t) has infinitely many discontinuities, at positive integer values of t.

g(t) = 1 + (0 − 1)u1 (t) + (1 − 0)u2 (t) + (0 − 1)u3 (t) + . . .

= 1 − u1 (t) + u2 (t) − u3 (t) + u4 (t) − . . .

Now let us see how we can compute Laplace transforms of expressions containing
step functions. As a preliminary step we will calculate the Laplace transform of a
step function itself.

7
 Z ∞
L{uc (t)} = uc (t)e−st dt
Z0 ∞
= e−st dt
Zc ∞
= e−s(τ +c) dτ
0
Z ∞
−sc
= e e−sτ dτ
0
= e−sc L{1}
e−sc
= .
s
where the transform converges for Re(s) > 0. The substitution τ = t − c was used
in the second step.
More or less the same computation allows us to prove the following theorem which
vastly generalizes the result above:

Theorem 4.1 Suppose that c > 0 and the Laplace transform of f (t) converges for
s > a. Then
L{uc (t)f (t − c)} = e−sc L{f (t)}
for s > a.

Proof
Z ∞
L{uc (t)f (t − c)} = uc (t)f (t − c)e−st dt
Z0 ∞
= f (t − c)e−st dt
Zc ∞
= f (τ )e−s(τ +c) dτ
0
Z ∞
−sc
= e f (τ )e−sτ dτ
0
= e−sc L{f (t)}.

8
Example 4.3 Suppose that g(t) is as in example 4.1, namely g(t) = (t − 2)u2 (t) +
(6 − 2t)u3 (t) + (t − 4)u4 (t). Using the theorem,

e−2s e−3s e−4s

L{g(t)} = − 2 + 2 .
s2 s2 s

Example 4.4 Next, suppose that g(t) is as in example 4.2, therefore g(t) = 1 −
u1 (t) + u2 (t) − u3 (t) + u4 (t) = . . . Again, using the theorem,

1 e−s e−2s e−3s

L{g(t)} = − + − + ...
s s s s
1
= .
s(1 + e−s )

The last equality uses the formula for geometric series and is valid for Re(s) > 0.