Laplace Transform

Instructor G. K. Srinivasan

Textbook: Kreyszig, Advanced Engg. Mathematics, 8th Edition

April 10, 2019 1 / 92

Functions of exponential type

Definition 1
A function f : (0, ∞) −→ R is said to be of exponetial type if
Rt
(i) The integral 0 |f (s)|ds exists for each t > 0 as a proper or improper
Riemann intgegral.
(ii) There are constants a and b such that |f (t)| ≤ exp(at) for all t ≥ b.

April 10, 2019 2 / 92

Functions of exponential type

Definition 1
A function f : (0, ∞) −→ R is said to be of exponetial type if
Rt
(i) The integral 0 |f (s)|ds exists for each t > 0 as a proper or improper
Riemann intgegral.
(ii) There are constants a and b such that |f (t)| ≤ exp(at) for all t ≥ b.

The first condition expresses local integrability of the function and the
second is a statement about growth at infinity.

April 10, 2019 2 / 92

Functions of exponential type

Definition 1
A function f : (0, ∞) −→ R is said to be of exponetial type if
Rt
(i) The integral 0 |f (s)|ds exists for each t > 0 as a proper or improper
Riemann intgegral.
(ii) There are constants a and b such that |f (t)| ≤ exp(at) for all t ≥ b.

The first condition expresses local integrability of the function and the
second is a statement about growth at infinity.
That is to say, f (t) does not grow too rapidly.

April 10, 2019 2 / 92

Functions of exponential type

Definition 1
A function f : (0, ∞) −→ R is said to be of exponetial type if
Rt
(i) The integral 0 |f (s)|ds exists for each t > 0 as a proper or improper
Riemann intgegral.
(ii) There are constants a and b such that |f (t)| ≤ exp(at) for all t ≥ b.

The first condition expresses local integrability of the function and the
second is a statement about growth at infinity.
That is to say, f (t) does not grow too rapidly. Thus we are NOT imposing
any strong regularity restrictions on the function but only Riemann
intgrability.

April 10, 2019 2 / 92

Some exampled of functions of exponential type

Let us illustrate the definition through many examples.

(i) f (t) = sin t and f (t) = cos t
(ii) f (t) = exp(1000t)
(iii) f (t) = t 10 . More generally ALL polynomials.
√
(iv) f (t) = 1/ t.
(v) f (t) = |t|
Note that in the last example the function is unbounded on (0, t) for
EVERY t > 0 but the integral of |f (t) is finite over each (0, t). The
function is bounded on [1, ∞).

April 10, 2019 3 / 92

Some non-examples

Let us look at a couple of examples

(i) f (t) = exp(x 2 )
(ii) f (t) = tan t.
(iii) f (t) = 1/ log t

April 10, 2019 4 / 92

Some non-examples

Let us look at a couple of examples

(i) f (t) = exp(x 2 )
(ii) f (t) = tan t.
(iii) f (t) = 1/ log t
In the first example f (t) = exp(t 2 ) the function is continuous and so the
Riemann integral exists over [0, t] for each t > 0 but the function grows
TOO rapidly at infinity.

April 10, 2019 4 / 92

Some non-examples

Let us look at a couple of examples

(i) f (t) = exp(x 2 )
(ii) f (t) = tan t.
(iii) f (t) = 1/ log t
In the first example f (t) = exp(t 2 ) the function is continuous and so the
Riemann integral exists over [0, t] for each t > 0 but the function grows
TOO rapidly at infinity.
For the case of f (t) = tan t, the function is NOT integrable on (0, t) for
any t > π/2 and the function becomes infinite at π2 , 3π 5π
2 , 2 , . . . and so we
cannot assert
|f (t)| ≤ exp(at), t ≥ b.

April 10, 2019 4 / 92

Exercises:

(1) Which one of the following functions are of exponential type?

(i) f (t) = log t
(ii) f (t) = t cot t.
(iii) f (t) = exp(−x 2 )
Provide a brief justification for each case.
(2) Are the functions sin2 t and cos2 t of exponential type?

April 10, 2019 5 / 92

Definition of the Laplace transform

We now introduce the principal object of these series of lectures:

April 10, 2019 6 / 92

Definition of the Laplace transform

We now introduce the principal object of these series of lectures:

Definition 2
Suppose f (t) is a function of exponential type, the Laplace transform of
f (t) denoted by F (s) or L(f ) is defined as
Z ∞
F (s) = e −st f (t)dt
0

April 10, 2019 6 / 92

Definition of the Laplace transform

We now introduce the principal object of these series of lectures:

Definition 2
Suppose f (t) is a function of exponential type, the Laplace transform of
f (t) denoted by F (s) or L(f ) is defined as
Z ∞
F (s) = e −st f (t)dt
0

Let us observe that the integral makes perfect sense. It ofcourse makes
sense on [0, b] (recall the b in the definition of functions of exponential
type).

April 10, 2019 6 / 92

Comments on the definition

Now we examine what is happening on [b, ∞). Note that since

|f (t)| ≤ exp(at) for t ≥ b,

|f (t)|e −st ≤ exp(−t(s − a))

April 10, 2019 7 / 92

Comments on the definition

Now we examine what is happening on [b, ∞). Note that since

|f (t)| ≤ exp(at) for t ≥ b,

|f (t)|e −st ≤ exp(−t(s − a))

This means the improper integral

Z ∞
f (t)e −st dt
b

would converge absolutely whenever s > a.

April 10, 2019 7 / 92

Comments on the definition

Now we examine what is happening on [b, ∞). Note that since

|f (t)| ≤ exp(at) for t ≥ b,

|f (t)|e −st ≤ exp(−t(s − a))

This means the improper integral

Z ∞
f (t)e −st dt
b

would converge absolutely whenever s > a.

We see that the Laplace transform F (s) is meaningful for s > a.

April 10, 2019 7 / 92

Abscissa of convergence and absolute convergence

The smallest real number ρ such that the integral

Z ∞
f (t)e −st dt
0

converges for all s > ρ is called the the abscissa of convergence and the
smallest real number ρ0 such that the improper integral converges
absolutely is called the abscissa of absolute convergence.

April 10, 2019 8 / 92

Examples
(i) Consider the simplest example f (t) = 1. We compute
Z ∞
1
F (s) = e −st dt =
0 s
and we see that the abscissa of convergence is 0 and is the same as
the abscissa of absolute convergence.
(ii) Let us consider f (t) = sin t. Then
Z ∞
F (s) = e −st sin tdt
0

The integral converges absolutely for all s > 0 and diverges if s ≤ 0.

Thus the abscissa of convergence and absolute convergence are both
0. Also a direct computation shows that
1
F (s) =
1 + s2
April 10, 2019 9 / 92
Exercises:

(3) Calculate the Laplace transforms of cos t, cos kt and sin kt.

April 10, 2019 10 / 92

Exercises:

(3) Calculate the Laplace transforms of cos t, cos kt and sin kt.
Ans: s/(s 2 + 1), s/(s 2 + k 2 ) and k/(s 2 + k 2 ) respectively.
(4) What is the Laplace transform of t, t 2 and t k in general for k ∈ N?

April 10, 2019 10 / 92

Exercises:

(3) Calculate the Laplace transforms of cos t, cos kt and sin kt.
Ans: s/(s 2 + 1), s/(s 2 + k 2 ) and k/(s 2 + k 2 ) respectively.
(4) What is the Laplace transform of t, t 2 and t k in general for k ∈ N?
Ans: 1/s 2 , 2/s 3 and k!/s k+1 respectively.
(5) What is the Laplace transform of te kt , t 2 e kt and t n e kt ?

April 10, 2019 10 / 92

Exercises:

(3) Calculate the Laplace transforms of cos t, cos kt and sin kt.
Ans: s/(s 2 + 1), s/(s 2 + k 2 ) and k/(s 2 + k 2 ) respectively.
(4) What is the Laplace transform of t, t 2 and t k in general for k ∈ N?
Ans: 1/s 2 , 2/s 3 and k!/s k+1 respectively.
(5) What is the Laplace transform of te kt , t 2 e kt and t n e kt ?
Ans:

April 10, 2019 10 / 92

Computation of Laplace transform of t n e kt

Z ∞ Z ∞
n kt −st
F (s) = t e e dt = t n exp(−t(s − k))
0 0

April 10, 2019 11 / 92

Computation of Laplace transform of t n e kt

Z ∞ Z ∞
n kt −st
F (s) = t e e dt = t n exp(−t(s − k))
0 0
The integral converges for s > k and so the abscissa of convergence is k.

April 10, 2019 11 / 92

Computation of Laplace transform of t n e kt

Z ∞ Z ∞
n kt −st
F (s) = t e e dt = t n exp(−t(s − k))
0 0
The integral converges for s > k and so the abscissa of convergence is k.
To compute the integral, Put t(s − k) = u assuming s > k and we get
Z ∞ n −u
u e du n!
F (s) = n+1
=
0 (s − k) (s − k)n+1

(6) Verify that the Laplace transform of e kt sin t is given by

April 10, 2019 11 / 92

Computation of Laplace transform of t n e kt

Z ∞ Z ∞
n kt −st
F (s) = t e e dt = t n exp(−t(s − k))
0 0
The integral converges for s > k and so the abscissa of convergence is k.
To compute the integral, Put t(s − k) = u assuming s > k and we get
Z ∞ n −u
u e du n!
F (s) = n+1
=
0 (s − k) (s − k)n+1

(6) Verify that the Laplace transform of e kt sin t is given by

1
F (s) =
1 + (s − k)2

Can you generalize this to a theorem?

April 10, 2019 11 / 92

First shifting theorem

Theorem 3
Suppose F (s) is the Laplace transform of a function f (t) of exponential
type then for any k ∈ R, the function e kt f (t) is also of exponential type
and its Laplace transform is F (s − k)

April 10, 2019 12 / 92

First shifting theorem

Theorem 3
Suppose F (s) is the Laplace transform of a function f (t) of exponential
type then for any k ∈ R, the function e kt f (t) is also of exponential type
and its Laplace transform is F (s − k)

In short, when the function f (t) is multiplied by an exponential e kt , the

Laplace transform gets shifted on the left by k.
Proof:

April 10, 2019 12 / 92

First shifting theorem

Theorem 3
Suppose F (s) is the Laplace transform of a function f (t) of exponential
type then for any k ∈ R, the function e kt f (t) is also of exponential type
and its Laplace transform is F (s − k)

In short, when the function f (t) is multiplied by an exponential e kt , the

Laplace transform gets shifted on the left by k.
Proof: Quite routine! By inspection !

April 10, 2019 12 / 92

First shifting theorem

Theorem 3
Suppose F (s) is the Laplace transform of a function f (t) of exponential
type then for any k ∈ R, the function e kt f (t) is also of exponential type
and its Laplace transform is F (s − k)

In short, when the function f (t) is multiplied by an exponential e kt , the

Laplace transform gets shifted on the left by k.
Proof: Quite routine! By inspection !
Z ∞ Z ∞
f (t)e kt e −st dt = f (t)e −t(s−k) = F (s − k).
0 0

April 10, 2019 12 / 92

The gamma function

This function has a tendency to pop up at unexpected places as we shall

see.
Definition 4
The gamma function is given by
Z ∞
Γ(a) = e −t t a−1 dt, a > 0.
0

Observe that
Γ(1) = 1, Γ(k + 1) = k!, k ∈ N.

√
(7) Show that Γ(1/2) = π.

April 10, 2019 13 / 92

Some Laplace transform formulas

Let us now compute the Laplace transform of the function f (t) = t a

where a > −1. The condition a > −1 ensures that
Z t
u a du
0

exists for each t > 0

April 10, 2019 14 / 92

Some Laplace transform formulas

Let us now compute the Laplace transform of the function f (t) = t a

where a > −1. The condition a > −1 ensures that
Z t
u a du
0

exists for each t > 0 (the first requirement for it to be of exponential type).
Z ∞ Z ∞  u a du
F (s) = e −st t a dt = e −u .
0 0 s s

April 10, 2019 14 / 92

Some Laplace transform formulas

Let us now compute the Laplace transform of the function f (t) = t a

where a > −1. The condition a > −1 ensures that
Z t
u a du
0

exists for each t > 0 (the first requirement for it to be of exponential type).
Z ∞ Z ∞  u a du
F (s) = e −st t a dt = e −u .
0 0 s s

And hence
Γ(a + 1)
F (s) = .
s a+1
Generalizing a formula we obtained earlier.

April 10, 2019 14 / 92

A special case of the preceding

√ the special case s = −1/2.

We shall have occasion to use
The Laplace transform of 1/ t is given by
√
π
√
s

April 10, 2019 15 / 92

A special case of the preceding

√ the special case s = −1/2.

We shall have occasion to use
The Laplace transform of 1/ t is given by
√
π
√
s

We shall later an application of Laplace transform to a mechanical

problem studied by Abel namely, The Tautochrone problem. In this
connection the above formula would be employed.

April 10, 2019 15 / 92

Differentiating under the integral sign

It would be particularly annoying to compute the Laplace transform of

t 2 sin t by repeated integration by parts.

April 10, 2019 16 / 92

Differentiating under the integral sign

It would be particularly annoying to compute the Laplace transform of

t 2 sin t by repeated integration by parts. Let us see how to circumvent this.
The idea is to introduce a parameter and compute instead the Laplace
transform of t 2 sin at. We begin with
Z ∞
s
I (a) = e −st cos atdt = 2
0 s + a2

April 10, 2019 16 / 92

Differentiating under the integral sign

It would be particularly annoying to compute the Laplace transform of

t 2 sin t by repeated integration by parts. Let us see how to circumvent this.
The idea is to introduce a parameter and compute instead the Laplace
transform of t 2 sin at. We begin with
Z ∞
s
I (a) = e −st cos atdt = 2
0 s + a2

Differentiate both sides with respect to a and we get

Z ∞
d  s 
e −st (t sin at)dt = −
0 da s 2 + a2

April 10, 2019 16 / 92

Differentiating under the integral sign

It would be particularly annoying to compute the Laplace transform of

t 2 sin t by repeated integration by parts. Let us see how to circumvent this.
The idea is to introduce a parameter and compute instead the Laplace
transform of t 2 sin at. We begin with
Z ∞
s
I (a) = e −st cos atdt = 2
0 s + a2

Differentiate both sides with respect to a and we get

Z ∞
d  s  2as
e −st (t sin at)dt = − 2 2
= 2
0 da s + a (s + a2 )2

The Laplace transform of t sin at has been computed. Now an exercise for
you

April 10, 2019 16 / 92

Differentiating under the integral sign

It would be particularly annoying to compute the Laplace transform of

t 2 sin t by repeated integration by parts. Let us see how to circumvent this.
The idea is to introduce a parameter and compute instead the Laplace
transform of t 2 sin at. We begin with
Z ∞
s
I (a) = e −st cos atdt = 2
0 s + a2

Differentiate both sides with respect to a and we get

Z ∞
d  s  2as
e −st (t sin at)dt = − 2 2
= 2
0 da s + a (s + a2 )2

The Laplace transform of t sin at has been computed. Now an exercise for
you
(8) Calculate the Laplace transform of t 2 sin t.

April 10, 2019 16 / 92

Laplace transform and derivatives

The idea in the last slide can be generalized to a theorem

Theorem 5
Suppose that F (s) is the Laplace transform (with abscissa of convergence
ρ) of a function f (t) of exponential type then,
(i) F (s) is differentiable infinitely often on (ρ, ∞)
(ii) tf (t) is also of exponential type and its Laplace transform is given by

d
− (F (s)).
ds
The proof is almost automatic. Simply differentiate under the integral
sign. However some care is needed to justify differentiation under the
integral sign. We shall not discuss the details here.

April 10, 2019 17 / 92

Laplace transform of derivatives
We shall now come to one of the most important result on Laplace
transforms that make it very useful for the study of the initial value
problem for differential equations.
Theorem 6
Suppose f : [0, ∞) −→ R is differentiable and f 0 (t) is of exponential type
then
L(f 0 )(s) = s(Lf )(s) − f (0).
For all s > b where b is as in the definition of functions of exponential
type.

To prove this we integrate by parts once:

Z ∞ Z ∞ ∞
0 −st 0 d −st −st
L(f )(s) = e f (t)dt = − f (t) (e )dt + e f (t)

0 0 dt 0

Since s > b, e −st f (t) vanishes at infinity and we are left with the
boundary term at 0 namely −f (0).
April 10, 2019 18 / 92
Laplace transform of derivatives contd...

The previous result easily generalizes to higher order derivatives. Assume

that the function together with the first k derivatives are of exponential
type. Then applying the formula to f (k−1) in place of f .

(Lf (k) )(s) = s(Lf (k−1) )(s) − f (k−1) (0)

= s 2 (Lf (k−2) )(s) − sf (k−2) (0) − f (k−1) (0)
= s 3 (Lf (k−3) )(s) − s 2 f (k−3) (0) − sf (k−2) (0) − f (k−1) (0)

etc.,

April 10, 2019 19 / 92

The Heaviside unit step function

This is denoted by u and defined as follows


0 t<0
u(t) =
1 t > 0.

The function has a jump discontinuity at the origin.

April 10, 2019 20 / 92

The Heaviside unit step function

This is denoted by u and defined as follows


0 t<0
u(t) =
1 t > 0.

The function has a jump discontinuity at the origin. Now even if f (t) is
defined on the entire real line and is of exponential type as described
earlier, the Laplace transform of f is defined as
Z ∞
Lf (s) = f (t)u(t)e −st dt.
−∞

April 10, 2019 20 / 92

The Heaviside unit step function

This is denoted by u and defined as follows


0 t<0
u(t) =
1 t > 0.

The function has a jump discontinuity at the origin. Now even if f (t) is
defined on the entire real line and is of exponential type as described
earlier, the Laplace transform of f is defined as
Z ∞
Lf (s) = f (t)u(t)e −st dt.
−∞

In other words we clip the signal off or make it zero for t < 0 and integrate
over the real line.

April 10, 2019 20 / 92

The Heaviside unit step function

This is denoted by u and defined as follows


0 t<0
u(t) =
1 t > 0.

The function has a jump discontinuity at the origin. Now even if f (t) is
defined on the entire real line and is of exponential type as described
earlier, the Laplace transform of f is defined as
Z ∞
Lf (s) = f (t)u(t)e −st dt.
−∞

In other words we clip the signal off or make it zero for t < 0 and integrate
over the real line. We shall see later that it is better to think of the
Laplace integral as an integral over the entire real line with the Heaviside
function thrown in.

April 10, 2019 20 / 92

The Riemann Lebesgue Lemma

Let us comparte the Laplace transforms of sin t and cos t:

1
(L sin t)(s) =
1 + s2
s
(L cos t)(s) =
1 + s2

Note that the first decays like 1/s 2 whereas the second line 1/s as
s −→ 0.

April 10, 2019 21 / 92

The Riemann Lebesgue Lemma

Let us comparte the Laplace transforms of sin t and cos t:

1
(L sin t)(s) =
1 + s2
s
(L cos t)(s) =
1 + s2

Note that the first decays like 1/s 2 whereas the second line 1/s as
s −→ 0. At first this would appear strange since sin t and cos t are such
similar looking functions.

April 10, 2019 21 / 92

The Riemann Lebesgue Lemma

Let us comparte the Laplace transforms of sin t and cos t:

1
(L sin t)(s) =
1 + s2
s
(L cos t)(s) =
1 + s2

Note that the first decays like 1/s 2 whereas the second line 1/s as
s −→ 0. At first this would appear strange since sin t and cos t are such
similar looking functions.
To understand the reason why the Laplace transforn of sin t decays faster,
observe that

April 10, 2019 21 / 92

The Riemann Lebesgue Lemma

Let us comparte the Laplace transforms of sin t and cos t:

1
(L sin t)(s) =
1 + s2
s
(L cos t)(s) =
1 + s2

Note that the first decays like 1/s 2 whereas the second line 1/s as
s −→ 0. At first this would appear strange since sin t and cos t are such
similar looking functions.
To understand the reason why the Laplace transforn of sin t decays faster,
observe that u(t) sin t is continuous at the origin and has one sided
derivatives as well. However, u(t) cos t has a jump discontinuity at the
origin.

April 10, 2019 21 / 92

The Riemann Lebesgue Lemma

Let us comparte the Laplace transforms of sin t and cos t:

1
(L sin t)(s) =
1 + s2
s
(L cos t)(s) =
1 + s2

Note that the first decays like 1/s 2 whereas the second line 1/s as
s −→ 0. At first this would appear strange since sin t and cos t are such
similar looking functions.
To understand the reason why the Laplace transforn of sin t decays faster,
observe that u(t) sin t is continuous at the origin and has one sided
derivatives as well. However, u(t) cos t has a jump discontinuity at the
origin. Let us look at more examples in this light.

April 10, 2019 21 / 92

Regularity of the function and decay of the Laplace
transform

1
Lu(t)(s) =
s
1
Ltu(t)(s) = 2
s
2 2
Lt u(t)(s) = 3
s
6
Lt 3 u(t)(s) = 4
s

Notice that the as we go down the list, the function u(t)f (t) gets
increasingly regular and the Laplace transform decays faster at infinity.

April 10, 2019 22 / 92

Riemann Lebesgue Lemma contd...

We now state the Riemann Lebesgue lemma:

Theorem 7
If f : [0, ∞) −→ ∞ is of exponential type, then Lf (s) −→ 0 as s −→ ∞.

April 10, 2019 23 / 92

Riemann Lebesgue Lemma contd...

We now state the Riemann Lebesgue lemma:

Theorem 7
If f : [0, ∞) −→ ∞ is of exponential type, then Lf (s) −→ 0 as s −→ ∞.

As an application, let us find the Laplace transform of the function sin t/t.
Z ∞
(sin t)e −st dt
F (s) =
0 t

April 10, 2019 23 / 92

Riemann Lebesgue Lemma contd...

We now state the Riemann Lebesgue lemma:

Theorem 7
If f : [0, ∞) −→ ∞ is of exponential type, then Lf (s) −→ 0 as s −→ ∞.

As an application, let us find the Laplace transform of the function sin t/t.
Z ∞
(sin t)e −st dt
F (s) =
0 t
Differentiating with respect to s we get

F 0 (s) = −(L sin t)(s) = −1/(1 + s 2 ).

Hence F (s) = C − tan−1 s. To find C use the Riemann Lebesgue lemma

which asserts that F (s) −→ 0 as s → ∞ and this would force C = π2 .

April 10, 2019 23 / 92

Exercises

(9) Find the Laplace transform of (1 − cos t)/t

(10) Is the function
 s +3 
log
2s − 1
the Laplace transform of a function of exponential type?
(11) Calculate the Laplace transform of (1 − cos t)/t 2 .
(12) Calculate the Laplace transform of (1 − e −t )/t.

April 10, 2019 24 / 92

An example

Let us see some examples on how to use this result.

Calculate the Laplace transform of sin2 t/t 2 .
Solution: Denoting by F (s) the Laplace transform of sin2 t/t 2 , we get
∞
sin2 t −st
Z
F (s) = e dt
0 t2

April 10, 2019 25 / 92

An example

Let us see some examples on how to use this result.

Calculate the Laplace transform of sin2 t/t 2 .
Solution: Denoting by F (s) the Laplace transform of sin2 t/t 2 , we get
∞
sin2 t −st
Z
F (s) = e dt
0 t2
Z ∞ 2
0 sin t −st
F (s) = − e dt
0 t

April 10, 2019 25 / 92

An example

Let us see some examples on how to use this result.

Calculate the Laplace transform of sin2 t/t 2 .
Solution: Denoting by F (s) the Laplace transform of sin2 t/t 2 , we get
∞
sin2 t −st
Z
F (s) = e dt
0 t2
Z ∞ 2
0 sin t −st
F (s) = − e dt
0 t
Z ∞
F 00 (s) = sin2 te −st dt
0

April 10, 2019 25 / 92

An example

Let us see some examples on how to use this result.

Calculate the Laplace transform of sin2 t/t 2 .
Solution: Denoting by F (s) the Laplace transform of sin2 t/t 2 , we get
∞
sin2 t −st
Z
F (s) = e dt
0 t2
Z ∞ 2
0 sin t −st
F (s) = − e dt
0 t
Z ∞
F 00 (s) = sin2 te −st dt
0
1 ∞
Z
00
F (s) = (1 − cos 2t)e −st dt
2 0

April 10, 2019 25 / 92

An example

Let us see some examples on how to use this result.

Calculate the Laplace transform of sin2 t/t 2 .
Solution: Denoting by F (s) the Laplace transform of sin2 t/t 2 , we get
∞
sin2 t −st
Z
F (s) = e dt
0 t2
Z ∞ 2
sin t −st
F 0 (s) = − e dt
0 t
Z ∞
F 00 (s) = sin2 te −st dt
0
1 ∞
Z
F 00 (s) = (1 − cos 2t)e −st dt
2 0
1 1 s
F 00 (s) = ( − 2 )
2 s s +4

April 10, 2019 25 / 92

Example contd...

So we have,
1 1 s
F 00 (s) = ( − 2 )
2 s s +4
One integration gives,
1 1
F 0 (s) = (log s − log(s 2 + 4)) + C
2 2

April 10, 2019 26 / 92

Example contd...

So we have,
1 1 s
F 00 (s) = ( − 2 )
2 s s +4
One integration gives,
1 1
F 0 (s) = (log s − log(s 2 + 4)) + C
2 2
The constant of integration must be zero by Riemann Lebesgue Lemma.

April 10, 2019 26 / 92

Example contd...

So we have,
1 1 s
F 00 (s) = ( − 2 )
2 s s +4
One integration gives,
1 1
F 0 (s) = (log s − log(s 2 + 4)) + C
2 2
The constant of integration must be zero by Riemann Lebesgue Lemma.
One more integration gives the desired result:

s 2 ds
Z
1 s 2 1
F (s) = (s log s − s − log(s + 4)) +
2 2 2 s2 + 4
1 s
= (s log s − log(s 2 + 4) − 2 tan−1 (s/2)) + C1
2 2

April 10, 2019 26 / 92

Example contd...

So we have to the stage:

April 10, 2019 27 / 92

Example contd...

So we have to the stage:

1 s
F (s) = (s log s − log(s 2 + 4) − 2 tan−1 (s/2)) + C1
2 2
Appealing to Riemann Lebesgue lemma once again we see that the
constant of integration C1 = π/2 whereby

April 10, 2019 27 / 92

Example contd...

So we have to the stage:

1 s
F (s) = (s log s − log(s 2 + 4) − 2 tan−1 (s/2)) + C1
2 2
Appealing to Riemann Lebesgue lemma once again we see that the
constant of integration C1 = π/2 whereby

1 s
F (s) = (s log s − log(s 2 + 4) − 2 cot−1 (s/2))
2 2
and the job is done!

April 10, 2019 27 / 92

A word of warning:

It is NOT true that every function (even very nice functions - infinitely
differentiable function) decaying to zero as s → ∞ is the Laplace
transform of another function.

April 10, 2019 28 / 92

A word of warning:

It is NOT true that every function (even very nice functions - infinitely
differentiable function) decaying to zero as s → ∞ is the Laplace
transform of another function. For example e −s is NOT the Laplace
transform of any function of exponential type.

April 10, 2019 28 / 92

A word of warning:

It is NOT true that every function (even very nice functions - infinitely
differentiable function) decaying to zero as s → ∞ is the Laplace
transform of another function. For example e −s is NOT the Laplace
transform of any function of exponential type.
The problem of describing precisely the set of functions that are Laplace
transforms of functions of exponential type is a highly non-trivial problem.

April 10, 2019 28 / 92

A word of warning:

It is NOT true that every function (even very nice functions - infinitely
differentiable function) decaying to zero as s → ∞ is the Laplace
transform of another function. For example e −s is NOT the Laplace
transform of any function of exponential type.
The problem of describing precisely the set of functions that are Laplace
transforms of functions of exponential type is a highly non-trivial problem.
We shall not enter into these discussions. The point is to draw your
attention to fact that the theory of Laplace transforms can be quite subtle.

April 10, 2019 28 / 92

Uniqueness theorem of Lerch

Observe that if we alter the value of a Riemann integrable function at a

point or at finitely many points we would again get a Riemann integrable
function and the integrals of the two functions would be identical. Thus
two such functions would have the SAME Laplace transform !

April 10, 2019 29 / 92

Uniqueness theorem of Lerch

Observe that if we alter the value of a Riemann integrable function at a

point or at finitely many points we would again get a Riemann integrable
function and the integrals of the two functions would be identical. Thus
two such functions would have the SAME Laplace transform ! However if
we take a continuous function and alter the value of the function at a
point, the result would be a discontinuous function!

April 10, 2019 29 / 92

Uniqueness theorem of Lerch

Observe that if we alter the value of a Riemann integrable function at a

point or at finitely many points we would again get a Riemann integrable
function and the integrals of the two functions would be identical. Thus
two such functions would have the SAME Laplace transform ! However if
we take a continuous function and alter the value of the function at a
point, the result would be a discontinuous function! This suggests that if
we are looking for a uniqueness result, we must work with the space of
continuous functions.

April 10, 2019 29 / 92

Uniqueness theorem of Lerch

Observe that if we alter the value of a Riemann integrable function at a

point or at finitely many points we would again get a Riemann integrable
function and the integrals of the two functions would be identical. Thus
two such functions would have the SAME Laplace transform ! However if
we take a continuous function and alter the value of the function at a
point, the result would be a discontinuous function! This suggests that if
we are looking for a uniqueness result, we must work with the space of
continuous functions.
We are NOT saying that the uniqueness result CANNOT be formulated for
discontinuous functions.

April 10, 2019 29 / 92

Uniqueness theorem of Lerch

Observe that if we alter the value of a Riemann integrable function at a

point or at finitely many points we would again get a Riemann integrable
function and the integrals of the two functions would be identical. Thus
two such functions would have the SAME Laplace transform ! However if
we take a continuous function and alter the value of the function at a
point, the result would be a discontinuous function! This suggests that if
we are looking for a uniqueness result, we must work with the space of
continuous functions.
We are NOT saying that the uniqueness result CANNOT be formulated for
discontinuous functions. All we are trying is to obtain a statement that is
simple (at this level) and devoid of technicalities.

April 10, 2019 29 / 92

Uniqueness theorem of Lerch

Observe that if we alter the value of a Riemann integrable function at a

point or at finitely many points we would again get a Riemann integrable
function and the integrals of the two functions would be identical. Thus
two such functions would have the SAME Laplace transform ! However if
we take a continuous function and alter the value of the function at a
point, the result would be a discontinuous function! This suggests that if
we are looking for a uniqueness result, we must work with the space of
continuous functions.
We are NOT saying that the uniqueness result CANNOT be formulated for
discontinuous functions. All we are trying is to obtain a statement that is
simple (at this level) and devoid of technicalities. To go beyond
continuous functions and formulate a uniqueness result would demand that
we get into the theory of Lebesgue integrals.

April 10, 2019 29 / 92

Lerch’s Theorem:

Theorem 8
Suppose f (t) and g (t) are CONTINUOUS functions of exponential type
on [0, ∞) such that Lf (s) and Lg (s) are defined on the range c < s < ∞
and
Lf (s) = Lg (s), for all s > c,
then,
f (t) = g (t), for all t ≥ 0.

April 10, 2019 30 / 92

Examples:
Let us find a function whose Laplace transform is
s + 1
log
s +2
Let us denote this function by F (s) and we are looking for a function f (t)
such that
L(f (t)) = F (s)

April 10, 2019 31 / 92

Examples:
Let us find a function whose Laplace transform is
s + 1
log
s +2
Let us denote this function by F (s) and we are looking for a function f (t)
such that
L(f (t)) = F (s)
L(tf (t)) = −F 0 (s)

April 10, 2019 31 / 92

Examples:
Let us find a function whose Laplace transform is
s + 1
log
s +2
Let us denote this function by F (s) and we are looking for a function f (t)
such that
L(f (t)) = F (s)
L(tf (t)) = −F 0 (s)
1 1
L(tf (t)) = −
s +2 s +1

April 10, 2019 31 / 92

Examples:
Let us find a function whose Laplace transform is
s + 1
log
s +2
Let us denote this function by F (s) and we are looking for a function f (t)
such that
L(f (t)) = F (s)
L(tf (t)) = −F 0 (s)
1 1
L(tf (t)) = −
s +2 s +1
L(tf (t)) = L(e −2t − e −t )

By the uniqueness theorem,

April 10, 2019 31 / 92

Examples:
Let us find a function whose Laplace transform is
s + 1
log
s +2
Let us denote this function by F (s) and we are looking for a function f (t)
such that
L(f (t)) = F (s)
L(tf (t)) = −F 0 (s)
1 1
L(tf (t)) = −
s +2 s +1
L(tf (t)) = L(e −2t − e −t )

By the uniqueness theorem,

1
f (t) = (e −2t − e −t ).
t
April 10, 2019 31 / 92
Second Shifting theorem

Theorem 9
Let f (t) be a function of exponential type. Assume c > 0 then

L(f (t − c)u(t − c)) = e −cs (Lf )

Proof:
Z ∞
L(f (t − c)u(t − c)) = f (t − c)u(t − c)e −st dt
Z0 ∞
= f (t − c)e −st du
Zc ∞
= f (u)e −s(c+u) du
0
Z ∞
−cs
= e f (u)e −su du
0

April 10, 2019 32 / 92

Finding the original function

Is there a function of exponential type whose Laplace transform is e −s /s?

April 10, 2019 33 / 92

Finding the original function

Is there a function of exponential type whose Laplace transform is e −s /s?

Ans: f (t) = u(t − 1).
(13) Find a function whose Laplace transform is (e −s − e −2s )/s. Discuss
the continuity of the original function.
(14) Find a function whose Laplace transform is (e −s − e −2s )/s 2 . Discuss
the continuity and differentiability of the original function.
(15) Find a function whose Laplace transform is (e −s − e −2s )/s 3 . Discuss
the continuity and differentiability of the original function.

April 10, 2019 33 / 92

Finding the original function contd...

(16) Find a function whose Laplace transform is

1
s4 −4
(17) Find a function whose Laplace transform is

1
s4 +4

April 10, 2019 34 / 92

Finding the original function contd...

(16) Find a function whose Laplace transform is

1
s4 −4
(17) Find a function whose Laplace transform is

1
s4 +4
Let us take up the first example:
1 1 1 1 
F (s) = = −
(s − 2)(s 2 + 2)
2 4 s2 − 2 s2 + 2
1  1 1  1 1
= √ √ − √ −
8 2 s− 2 s+ 2 4 s2 + 2

April 10, 2019 34 / 92

Discussion of example 16 contd...

1  1 1  1 1
F (s) = √ √ − √ −
8 2 s− 2 s+ 2 4 s2 + 2
This can be rewritten as

1  √2t √   sin √2t 

− 2t
F (s) = √ Le − Le −L √
8 2 4 2
1 √ √
= √ L(sinh 2t − sin 2t)
4 2

The original function is thus

1 √ √
√ (sinh 2t − sin 2t)
4 2
and this happens to be continuous. By Lerch’s theorem this is the only
one which is continuous.
April 10, 2019 35 / 92
Discussion of example 17 contd...

For the example (17) use the Partial Fractions decomposition:

1 1
= 2
s4 +4 (s + 2s + 2)(s 2 − 2s + 2)

April 10, 2019 36 / 92

A matter of notation

Suppose that f (t) has Laplace transform F (s) we shall find it convenient
to write
L−1 F = f

April 10, 2019 37 / 92

A matter of notation

Suppose that f (t) has Laplace transform F (s) we shall find it convenient
to write
L−1 F = f
WARNING: We are NOT asserting that the operator L is invertible and we
are NOT discussing the range of the operator L. We are NOT describing
the image of the Laplace transform operator. Also two functions may have
the same Laplace transform without being equal ! For example tamper the
value of a function f (t) at ONE or say FINITELY many points to get a
new function g (t). Then f and g would have the same Laplace transform.
However if f is continuous then g would certainly become discontinuous!
However, there cannot be TWO continuous function with the SAME
Laplace transform.

April 10, 2019 37 / 92

Solving Initial Value Problems for Ordinary Differential
Equations

It is better to illustrate the technique through a simple example.

y 00 + y = sin t, y (0) = 0, y 0 (0) = 1.

Denoting by Y (s) the Laplace transform of y (t) we get

1
(s 2 + 1)Y − 1 =
s2 +1
which gives
1 1
Y (s) = + 2
s2 + 1 (s + 1)2
The first piece is the Laplace transform of sin t. Let us find a function
whose Laplace transform is the second piece.

April 10, 2019 38 / 92

Example contd...

d s  s2 − 1
− =
ds s 2 + 1 (s 2 + 1)2
s2 + 1 − 2
=
(s 2 + 1)2
1 2
= − 2
s + 1 (s + 1)2
2

Hence
1 1 1
= L sin t − L(t cos t)
(s 2 + 1) 2 2 2

April 10, 2019 39 / 92

Thus,
3 1
Y (s) = L( sin t − t cos t)
2 2
which means
3 1
y (t) = sin t − t cos t
2 2
as the solution of the initial value problem.

April 10, 2019 40 / 92

The General Idea

Note that the procedure is quite clear. One assumes that the differential
equation is of the form
P(D)y = f (t)
d
where P(D) is a polynomial in the derivative dt which means its
coefficients are constants. The function f (t) on the RHS is of exponential
type whose Laplace transform is available in closed form. One then takes
the Laplace transform of both sides and we would get

Y (s) = R(s)

where R(s) is known and in case f (t) is a linear combination of t k e ct ,

t k sin ct and t k cos ct then R(s) is a rational function. One often uses
tables of Laplace transform to recover y (t) from the last displayed
equation.

April 10, 2019 41 / 92

Case of rational functions

We decompose the rational function into partial fractions and we then

have to deal with finding a function f (t) whose Laplace transform is one
of the following:
1 s −c b
k+1
, 2 2 k+1
,
(s − c) ((s − c) + b ) ((s − c) + b 2 )k+1
2

The first is easy:

1
L(t k e ct ) =
(s − c)k+1
The last two can be quite tedious.

April 10, 2019 42 / 92

 b
L(e ct sin bt) =
(s − c)2 + b 2
But,
d b  2b(s − c)
− 2 2
=
ds (s − c) + b ((s − c)2 + b 2 )2
So we see that
s −c 1
ct

= L te sin bt
((s − c)2 + b 2 )2 2b

April 10, 2019 43 / 92

Again,

d s −c  (s − c)2 − b 2
− =
ds (s − c)2 + b 2 ((s − c)2 + b 2 )2
1 2b 2
= −
(s − c)2 + b 2 ((s − c)2 + b 2 )2

Thus,
2b 3
L(bte ct cos bt) = L(te ct sin bt) −
((s − c)2 + b 2 )2
This gives us the function whose Laplace transform is
1
((s − c)2 + b 2 )2

April 10, 2019 44 / 92

Clearly, this is not going to be such a merry exercise with the denominators

((s − c)2 + b 2 )3

and higher powers!

April 10, 2019 45 / 92

A more elegant and perhaps efficient way would be to completely factorize
into linear factors with real or complex roots.
As an example let us take up
1
((s − 1)2 + 1)2

The partial fraction decomposition is (λ = 1 + i)

1 a b α β
2 2
= + 2
+ + (2.1)
((s − 1) + 1) s − λ (s − λ) s − λ (s − λ)2

Since original the rational function has real coefficients, we have

α = a, β = b

April 10, 2019 46 / 92

Multiplying (2.1) by (s − λ)2 and setting s = λ we get the pair

1
b=β=−
4

April 10, 2019 47 / 92

Multiplying (2.1) by (s − λ)2 and setting s = λ we get the pair

1
b=β=−
4
Multipying (2.1) by s − λ and letting s → ∞ we get the equation

a + α = 0.

We need one more equation and for this we set s = 0 in (2.1):

1  a α 1  1 
=− + − Re
4 λ λ 2 λ2
From which we get the pair
i i
a=− , α= .
4 4

April 10, 2019 47 / 92

So (2.1) reads

1 i 1 1  1 1 1 
= − − − +
((s − 1)2 + 1)2 4 s −λ s −λ 4 (s − λ)2 (s − λ)2

We can now use the formula

1 1
= Le ct , = L(te ct )
s −c (s − c)2

which holds for all complex values c.

(18) Complete the job of finding the function f (t) whose Laplace
transform is 1/((s − 1)2 + 1)2 .

April 10, 2019 48 / 92

In general recovering the original function from the Laplace transform is a
HIGHLY non-trivial problem and requires some knowledge of complex
analysis.

April 10, 2019 49 / 92

In general recovering the original function from the Laplace transform is a
HIGHLY non-trivial problem and requires some knowledge of complex
analysis.
For a discussion on Real Inversion theorem see the book G. Fodor, Laplace
transforms in Engineering, Akademiai Kiado, 1965.

April 10, 2019 49 / 92

Solving a system of first order differential equations
Example: Consider the system of ODEs:
dx dy
= −y − t, =x −1
dt dt
with initial conditions x(0) = 0, y (0) = 0. Taking the Laplace transform of
the system
1 1
sX = −Y − 2 , sY = X − .
s s
Solving these simultaneously for X and Y we get
X = 0, Y = −1/s 2
which gives immediately x(t) = 0, y (t) = −t.

Needless to say this example was designed to illustrate the idea and keep
the computational complexity at a minimum. In general we get a pair of
equations for X and Y which we solve and then recover x(t) and y (t)
from these. The difficulty lies in recovering the original function from the
Laplace transforms.
April 10, 2019 50 / 92
Example: Next, let us take up the example
dx dy
= y + sin t, = −x + cos t, x(0) = −1, y (0) = 0.
dt dt
Taking the Laplace transforms of the two equations:

−s 2 s
sX − Y = , X + sY =
s2 + 1 s2 + 1
Solving for X and Y :

−s(s 2 − 1) 2s 2
X = , Y =
(s 2 + 1)2 (s 2 + 1)2

(19) Do the partial fractions decomposition and find x(t) and y (t).

April 10, 2019 51 / 92

Example: Next, let us take up the example
dx dy
= y + sin t, = −x + cos t, x(0) = −1, y (0) = 0.
dt dt
Taking the Laplace transforms of the two equations:

−s 2 s
sX − Y = , X + sY =
s2 + 1 s2 + 1
Solving for X and Y :

−s(s 2 − 1) 2s 2
X = , Y =
(s 2 + 1)2 (s 2 + 1)2

(19) Do the partial fractions decomposition and find x(t) and y (t).
Ans: x(t) = t sin t − cos t and y (t) = t cos t + sin t

April 10, 2019 51 / 92

What are the merits of the Laplace transform technique over other
techniques such as the method of variation of parameters and the method
of undetermined coefficients?

April 10, 2019 52 / 92

What are the merits of the Laplace transform technique over other
techniques such as the method of variation of parameters and the method
of undetermined coefficients?

(1) The Laplace transform method incorporates the initial conditions in

the solution process. One does not have to determine the values of
constants later by solving a system of linear algebraic equations.
(2) It handles linear systems with constant coeffients most efficiently - it
makes no difference whether the matrix of coefficients is
diagonalizable or not.

April 10, 2019 52 / 92

What are the merits of the Laplace transform technique over other
techniques such as the method of variation of parameters and the method
of undetermined coefficients?

(1) The Laplace transform method incorporates the initial conditions in

the solution process. One does not have to determine the values of
constants later by solving a system of linear algebraic equations.
(2) It handles linear systems with constant coeffients most efficiently - it
makes no difference whether the matrix of coefficients is
diagonalizable or not.
Let us take a moment and review the conventional process to understand
comment (2) above.

April 10, 2019 52 / 92

Solving a linear system

We consider the system of ODEs

dx
= Ax (2.2)
dt
where A is an n × n matrix of constants. The system is coupled system of
equations in the sense that all the variables x1 , . . . , xn would appear in all
the equations. In the exceptional case where A is a diagonal matrix with
λ1 , λ2 , . . . , λn down the main diagonal, equation (2.2) reads:

dx1 dxn
= λ1 x1 , . . . = λn xn (2.3)
dt dt
The system (2.3) is said to be a decoupled system.

April 10, 2019 53 / 92

Decoupling a coupled system of equation

Clearly we can solve (2.3) quite easily but as such it is not clear how to
negotiate the coupled system (2.2).

April 10, 2019 54 / 92

Decoupling a coupled system of equation

Clearly we can solve (2.3) quite easily but as such it is not clear how to
negotiate the coupled system (2.2).
Using basic ideas from linear algebra one can decouple the system (2.2) in
certain cases

April 10, 2019 54 / 92

Diagonalizable case

Assume that the matrix is diagonalizable

April 10, 2019 55 / 92

Diagonalizable case

Assume that the matrix is diagonalizable which means there is a basis of

Rn consisting of eigen-vectors of A. Let the basis of eigen vectors be
v1 , . . . , vn . Put
P = [v1 , . . . , vn ]
Then P is an invertible matrix and

P −1 AP = diag(λ1 , . . . , λn )

April 10, 2019 55 / 92

Diagonalizable case

Assume that the matrix is diagonalizable which means there is a basis of

Rn consisting of eigen-vectors of A. Let the basis of eigen vectors be
v1 , . . . , vn . Put
P = [v1 , . . . , vn ]
Then P is an invertible matrix and

P −1 AP = diag(λ1 , . . . , λn )

Put x = Py and the system of ODEs becomes

P ẏ = APy

which means

ẏ = P −1 APy = diag(λ1 y1 , λ2 y2 , . . . , λn yn )

April 10, 2019 55 / 92

So the transformed system is the decoupled system

y˙1 = λ1 y1 , . . . , y˙n = λn yn

After obtaining y (t) we can recover x(t) as

x(t) = Ay (t).

April 10, 2019 56 / 92

So the transformed system is the decoupled system

y˙1 = λ1 y1 , . . . , y˙n = λn yn

After obtaining y (t) we can recover x(t) as

x(t) = Ay (t).

Note that if A is not diagonalizable, the problem becomes much more

complicated and the solution process would get quite involved.

April 10, 2019 56 / 92

So the transformed system is the decoupled system

y˙1 = λ1 y1 , . . . , y˙n = λn yn

After obtaining y (t) we can recover x(t) as

x(t) = Ay (t).

Note that if A is not diagonalizable, the problem becomes much more

complicated and the solution process would get quite involved.
However the method of Laplace transform goes through whether or not A
is diagonalizable!

April 10, 2019 56 / 92

Using the Laplace transform to compute integrals

The Laplace transform can be used to compute certain integrals efficiently.

We shall take up some time to illustrate this VERY practical application!
(20) Compute Z ∞
sin tx
I (t) = , t>0
0 x

April 10, 2019 57 / 92

Using the Laplace transform to compute integrals

The Laplace transform can be used to compute certain integrals efficiently.

We shall take up some time to illustrate this VERY practical application!
(20) Compute Z ∞
sin tx
I (t) = , t>0
0 x
Taking the Laplace transform, we get
Z ∞
sin tx
I = L( ), t>0
0 x

April 10, 2019 57 / 92

Using the Laplace transform to compute integrals

The Laplace transform can be used to compute certain integrals efficiently.

We shall take up some time to illustrate this VERY practical application!
(20) Compute Z ∞
sin tx
I (t) = , t>0
0 x
Taking the Laplace transform, we get
Z ∞
sin tx
I = L( ), t > 0
x
Z0 ∞
dx π
= 2 2
=
0 x +s 2s

Hence I (t) = π2 .

April 10, 2019 57 / 92

Using the Laplace transform to compute integrals

The Laplace transform can be used to compute certain integrals efficiently.

We shall take up some time to illustrate this VERY practical application!
(20) Compute Z ∞
sin tx
I (t) = , t>0
0 x
Taking the Laplace transform, we get
Z ∞
sin tx
I = L( ), t > 0
x
Z0 ∞
dx π
= 2 2
=
0 x +s 2s

Hence I (t) = π2 . This integral plays an extremely important role in

Fourier Analysis and is called the Dirichlet integral.

April 10, 2019 57 / 92

The Frullani Integral

∞
e −ax − e −bx
Z
dx, 0 < a < b.
0 x

April 10, 2019 58 / 92

The Frullani Integral

∞
e −ax − e −bx
Z
dx, 0 < a < b.
0 x
We shall have occasion to use this integral later today! Let us introduce a
parameter and denote

April 10, 2019 58 / 92

The Frullani Integral

∞
e −ax − e −bx
Z
dx, 0 < a < b.
0 x
We shall have occasion to use this integral later today! Let us introduce a
parameter and denote
Z ∞ −tax
e − e −tbx
I (t) = dx
0 x

April 10, 2019 58 / 92

The Frullani Integral

∞
e −ax − e −bx
Z
dx, 0 < a < b.
0 x
We shall have occasion to use this integral later today! Let us introduce a
parameter and denote
Z ∞ −tax
e − e −tbx
I (t) = dx
0 x

Taking the Laplace transform we get

Z ∞ 
1 1 1 
(LI )(s) = − dx
0 x s + ax s + bx
Z ∞
b−a
= dx
0 (s + ax)(s + bx)

April 10, 2019 58 / 92

The Frullani integral contd...

∞
b−a
Z
1
(LI )(s) = s dx
ab 0 (x + a )(x + bs )

(21) Do the partial fraction decomposition and evaluate the integral and
recover I (t). Put t = 1 and complete the evaluation of the Frullani
integral.

April 10, 2019 59 / 92

The Frullani integral contd...

∞
b−a
Z
1
(LI )(s) = s dx
ab 0 (x + a )(x + bs )

(21) Do the partial fraction decomposition and evaluate the integral and
recover I (t). Put t = 1 and complete the evaluation of the Frullani
integral.
Ans: log(b/a)

April 10, 2019 59 / 92

The Cauchy distribution

This appears naturally in statistics. One is interested in the Fourier

transform of the Cauchy distribution and this is the integral

2 ∞ cos xξdx
Z

π 0 1 + x2

(22) Use Laplace transform to compute this integral.

April 10, 2019 60 / 92

The Cauchy distribution

This appears naturally in statistics. One is interested in the Fourier

transform of the Cauchy distribution and this is the integral

2 ∞ cos xξdx
Z

π 0 1 + x2

(22) Use Laplace transform to compute this integral.

(23) Determine the Laplace transform of
Z ∞
sin txdx
I (t) =
0 x2 + 1

Ans for (22): e −ξ

Ans for (23): log s/(s 2 − 1) There is no explicit formula for the function
I (t).

April 10, 2019 60 / 92

Euler’s constant

Theorem 10
The sequence
1 1
1+ + · · · + − log(n + 1) (3.1)
2 n
converges as n → ∞ and its limit is denoted by γ, known as Euler’s
constant.

April 10, 2019 61 / 92

Euler’s constant

Theorem 10
The sequence
1 1
1+ + · · · + − log(n + 1) (3.1)
2 n
converges as n → ∞ and its limit is denoted by γ, known as Euler’s
constant.
As of today it is NOT known whether γ is rational or irrational. The
Euler’s constant has the habit of popping up at unexpected places.

April 10, 2019 61 / 92

Euler’s constant

Theorem 10
The sequence
1 1
1+ + · · · + − log(n + 1) (3.1)
2 n
converges as n → ∞ and its limit is denoted by γ, known as Euler’s
constant.
As of today it is NOT known whether γ is rational or irrational. The
Euler’s constant has the habit of popping up at unexpected places.
Note that log(n + 1) may be replaced by log n in (3.1).

April 10, 2019 61 / 92

Euler’s constant

Theorem 10
The sequence
1 1
1+ + · · · + − log(n + 1) (3.1)
2 n
converges as n → ∞ and its limit is denoted by γ, known as Euler’s
constant.
As of today it is NOT known whether γ is rational or irrational. The
Euler’s constant has the habit of popping up at unexpected places.
Note that log(n + 1) may be replaced by log n in (3.1).
Z ∞
1
e −jt dt = , j = 1, 2, 3, . . .
0 j

Adding over j = 1, 2, . . . , n we get summing the finite geometric series,

April 10, 2019 61 / 92

Euler’s constant contd...

∞
e −t (1 − e −nt )
Z
1 1 1
−t
dt = 1 + + + · · · +
0 1−e 2 3 n
It is useful to write this as
Z ∞ −t
e (1 − e −nt ) t 1 1 1
−t
dt = 1 + + + · · · +
0 t 1 − e 2 3 n

April 10, 2019 62 / 92

Euler’s constant contd...

∞
e −t (1 − e −nt )
Z
1 1 1
−t
dt = 1 + + + · · · +
0 1−e 2 3 n
It is useful to write this as
Z ∞ −t
e (1 − e −nt ) t 1 1 1
−t
dt = 1 + + + · · · +
0 t 1 − e 2 3 n

The trick is to appeal to the Frullani integral:

Z ∞ −t
e (1 − e −nt )
log(n + 1) = dt
0 t

April 10, 2019 62 / 92

Euler’s constant contd...

∞
e −t (1 − e −nt )
Z
1 1 1
−t
dt = 1 + + + · · · +
0 1−e 2 3 n
It is useful to write this as
Z ∞ −t
e (1 − e −nt ) t 1 1 1
−t
dt = 1 + + + · · · +
0 t 1 − e 2 3 n

The trick is to appeal to the Frullani integral:

Z ∞ −t
e (1 − e −nt )
log(n + 1) = dt
0 t

Subtracting we get

April 10, 2019 62 / 92

Euler’s constant contd...

Z ∞
1 1 1 n 1 1 o −t
1+ + +· · ·+ −log(n+1) = (1−e −nt ) − e dt (3.2)
2 3 n 0 1 − e −t t

April 10, 2019 63 / 92

Euler’s constant contd...

Z ∞
1 1 1 n 1 1 o −t
1+ + +· · ·+ −log(n+1) = (1−e −nt ) − e dt (3.2)
2 3 n 0 1 − e −t t

Now using L’Hospital’s rule, the expression within braces tends to 1/2 as
t → 0+.

April 10, 2019 63 / 92

Euler’s constant contd...

Z ∞
1 1 1 n 1 1 o −t
1+ + +· · ·+ −log(n+1) = (1−e −nt ) − e dt (3.2)
2 3 n 0 1 − e −t t

Now using L’Hospital’s rule, the expression within braces tends to 1/2 as
t → 0+. This shows that the integral is not improper at the lower end (it
is obviously so at the upper end).

April 10, 2019 63 / 92

Euler’s constant contd...

Z ∞
1 1 1 n 1 1 o −t
1+ + +· · ·+ −log(n+1) = (1−e −nt ) − e dt (3.2)
2 3 n 0 1 − e −t t

Now using L’Hospital’s rule, the expression within braces tends to 1/2 as
t → 0+. This shows that the integral is not improper at the lower end (it
is obviously so at the upper end). Let us now let n −→ ∞ in (3.2) and
take the limit inside the integral. It is not difficult to justify this but doing
so would be a detour for which we have no time.

April 10, 2019 63 / 92

Euler’s constant contd...

We thus see that the limit in (3.1) exists and IMPORTANTLY, the limit
equals Z ∞
1 1  −t
− e dt
0 1 − e −t t
This integral can be further transformed to a much simpler form as we
shall now see

April 10, 2019 64 / 92

Euler’s constant contd...

It is tempting to separate the two terms in the integrand and then

integrate by parts. There is danger near the lower end!!

April 10, 2019 65 / 92

Euler’s constant contd...

It is tempting to separate the two terms in the integrand and then

integrate by parts. There is danger near the lower end!! Instead, integrate
by parts over [a, ∞) and then we let a → 0+.

April 10, 2019 65 / 92

Euler’s constant contd...

It is tempting to separate the two terms in the integrand and then

integrate by parts. There is danger near the lower end!! Instead, integrate
by parts over [a, ∞) and then we let a → 0+.
(24) Carry this out and check that
Z ∞
γ=− e −t log tdt.
0

April 10, 2019 65 / 92

Euler’s constant contd...

It is tempting to separate the two terms in the integrand and then

integrate by parts. There is danger near the lower end!! Instead, integrate
by parts over [a, ∞) and then we let a → 0+.
(24) Carry this out and check that
Z ∞
γ=− e −t log tdt.
0

(25) Determine the Laplace transform of log t.

April 10, 2019 65 / 92

Euler’s constant contd...

It is tempting to separate the two terms in the integrand and then

integrate by parts. There is danger near the lower end!! Instead, integrate
by parts over [a, ∞) and then we let a → 0+.
(24) Carry this out and check that
Z ∞
γ=− e −t log tdt.
0

(25) Determine the Laplace transform of log t. Ans: −γ/s − (log s)/s.

April 10, 2019 65 / 92

Laplace transforms of Periodic functions
Theorem 11
Assume that f : R −→ R is periodic with period p. Then
Z p
e ps
Lf = ps f (t)e −st dt. (3.3)
e −1 0

We use the second shifting theorem. Assume that p > 0. Observe that
u(t)f (t) − u(t − p)f (t − p)
is zero outside [0, p]. This can be seen analytically but it is easier to see
this graphically.

April 10, 2019 66 / 92

Laplace transforms of Periodic functions
Theorem 11
Assume that f : R −→ R is periodic with period p. Then
Z p
e ps
Lf = ps f (t)e −st dt. (3.3)
e −1 0

We use the second shifting theorem. Assume that p > 0. Observe that
u(t)f (t) − u(t − p)f (t − p)
is zero outside [0, p]. This can be seen analytically but it is easier to see
this graphically. Now over the interval [0, p] the above function equals
f (t) so multiplying by e −st and integrating we get invoking the second
shifting theorem,
Z p
f (t)e −st = F (s) − F (s)e −ps
0

From this the theorem follows.

April 10, 2019 66 / 92
Laplace transform of periodic functions

Applying this result to | sin t| which is periodic with period π we see that
Z π
e πs
L| sin t| = πs e −st sin tdt
e −1 0

(26) Calculate the integral and find the answer.

April 10, 2019 67 / 92

Laplace transform of periodic functions

Applying this result to | sin t| which is periodic with period π we see that
Z π
e πs
L| sin t| = πs e −st sin tdt
e −1 0

(26) Calculate the integral and find the answer. Ans: coth( πs 2 −1
2 )(1 + s )

April 10, 2019 67 / 92

Laplace transform of periodic functions

Applying this result to | sin t| which is periodic with period π we see that
Z π
e πs
L| sin t| = πs e −st sin tdt
e −1 0

(26) Calculate the integral and find the answer. Ans: coth( πs 2 −1
2 )(1 + s )
Next, let us take up the 2π−periodic square wave function

−1 −π < t < 0
f (t) =
1 0 < t < π.

April 10, 2019 67 / 92

Laplace transforms of periodic functions

Using the formula we get

2π
e 2πs
Z
Lf = 2πs e −st dt.
e −1 0

April 10, 2019 68 / 92

Laplace transforms of periodic functions

Using the formula we get

2π
e 2πs
Z
Lf = 2πs e −st dt.
e −1 0

(27) Perform the integration and calculate the Laplace transform.

April 10, 2019 68 / 92

Laplace transforms of periodic functions

Using the formula we get

2π
e 2πs
Z
Lf = 2πs e −st dt.
e −1 0

(27) Perform the integration and calculate the Laplace transform.

 
Ans: 1s tanh πs
2

April 10, 2019 68 / 92

Laplace transforms of periodic functions

Using the formula we get

2π
e 2πs
Z
Lf = 2πs e −st dt.
e −1 0

(27) Perform the integration and calculate the Laplace transform.

 
Ans: 1s tanh πs
2
We shall now take a very different approach to compute this Laplace
transform and compare results.

April 10, 2019 68 / 92

Bringing in the Fourier series! Mittag-Leffler expansions

Let us calculate the Fourier series of the above function:

April 10, 2019 69 / 92

Bringing in the Fourier series! Mittag-Leffler expansions

Let us calculate the Fourier series of the above function:

∞
4 X sin(2k − 1)t
f (t) =
π 2k − 1
k=1

April 10, 2019 69 / 92

Bringing in the Fourier series! Mittag-Leffler expansions

Let us calculate the Fourier series of the above function:

∞
4 X sin(2k − 1)t
f (t) =
π 2k − 1
k=1

Taking the Laplace transform of the above equation we get:

April 10, 2019 69 / 92

Bringing in the Fourier series! Mittag-Leffler expansions

Let us calculate the Fourier series of the above function:

∞
4 X sin(2k − 1)t
f (t) =
π 2k − 1
k=1

Taking the Laplace transform of the above equation we get:

∞
4X 1
Lf =
π (2k − 1)2 + s 2
k=1
∞
4X 1
Lf =
π (2k − 1)2 + s 2
k=1
∞
1  πs  4X 1
tanh = (3.4)
s 2 π (2k − 1)2 + s 2
k=1

April 10, 2019 69 / 92

Special values of zeta function.

The identity (3.4) is very interesting. Letting s → 0+ we infer

1 1 π2
1+ + + · · · =
32 52 8

April 10, 2019 70 / 92

Special values of zeta function.

The identity (3.4) is very interesting. Letting s → 0+ we infer

1 1 π2
1+ + + · · · =
32 52 8

(28) Determine the sum of the series

1 1
1+ 2
+ 2 + ...
2 3
(29) Carry out a similar calculation for | sin t/2| which is periodic with
period 2π.

April 10, 2019 70 / 92

Special values of zeta function.

The identity (3.4) is very interesting. Letting s → 0+ we infer

1 1 π2
1+ + + · · · =
32 52 8

(28) Determine the sum of the series

1 1
1+ 2
+ 2 + ...
2 3
(29) Carry out a similar calculation for | sin t/2| which is periodic with
period 2π.
Ans to problem 28: π 2 /6

April 10, 2019 70 / 92

Convolution of two intergable functions

Assume that f (t) and g (t) are two functions which are absolutely
integrable: Z ∞ Z ∞
|f (t)|dt and |g (t)|dt
−∞ −∞
are both finite.

April 10, 2019 71 / 92

Convolution of two intergable functions

Assume that f (t) and g (t) are two functions which are absolutely
integrable: Z ∞ Z ∞
|f (t)|dt and |g (t)|dt
−∞ −∞
are both finite.
Definition 12
The convolution of f and g denoted by f ∗ g (x) is defined as
Z ∞
(f ∗ g )(x) = f (x − t)g (t)dt
−∞

One immediately sees that

f ∗g =g ∗f

April 10, 2019 71 / 92

Theorem 13
The convolution f ∗ g is absolutely integrable.

Z ∞
|f ∗ g (x)| ≤ |f (x − t)||g (t)|dt
−∞
Integrating with respect to x we get
Z ∞ Z ∞ Z ∞
|f ∗ g (x)|dx ≤ dx |f (x − t)||g (t)|dt
−∞ −∞ −∞

Interchanging the integrals

Z ∞ Z ∞ Z ∞
|f ∗ g (x)|dx ≤ |g (t)|dt |f (x − t)|dx
−∞ −∞ −∞

April 10, 2019 72 / 92

In the inner integral put x − t = y and infer that the right hand side is the
product Z ∞ Z ∞
|f (t)|dt |g (y )|dy
−∞ −∞
so by comparison test we see that f ∗ g is absolutely intgerable.

April 10, 2019 73 / 92

In the inner integral put x − t = y and infer that the right hand side is the
product Z ∞ Z ∞
|f (t)|dt |g (y )|dy
−∞ −∞
so by comparison test we see that f ∗ g is absolutely intgerable.
In particular, the Laplace integral
Z ∞
|f ∗ g (x)|e −xs dx
0

exists for all s > 0.

April 10, 2019 73 / 92

Theorem 14
If f (t) and g (t) are of exponential type then the convolution

(uf ) ∗ (ug )(x)

is also of exponential type.

April 10, 2019 74 / 92

Theorem 14
If f (t) and g (t) are of exponential type then the convolution

(uf ) ∗ (ug )(x)

is also of exponential type.

To see this first note that the last display equals

Z x
f (y )g (x − y )dy .
0

and we must now estimate this. By hypothesis there exist positive

constants A and T such that

|f (t)| < e At , |g (t)| < e At , for all t ≥ T .

April 10, 2019 74 / 92

Thus, if we take x > 2T , then
Z x
|(uf ) ∗ (ug )(x)| ≤ |f (y )||g (x − y )|dy
0
Z T Z x
≤ |f (y )||g (x − y )|dy + |f (y )||g (x − y )|dy
0 T
Z x
≤C+ |f (y )||g (x − y )|dy
T

April 10, 2019 75 / 92

Thus, if we take x > 2T , then
Z x
|(uf ) ∗ (ug )(x)| ≤ |f (y )||g (x − y )|dy
0
Z T Z x
≤ |f (y )||g (x − y )|dy + |f (y )||g (x − y )|dy
0 T
Z x
≤C+ |f (y )||g (x − y )|dy
T

But then x − y > T if T ≤ y ≤ x and so the last integral can be dealt as

follows:

April 10, 2019 75 / 92

Thus, if we take x > 2T , then
Z x
|(uf ) ∗ (ug )(x)| ≤ |f (y )||g (x − y )|dy
0
Z T Z x
≤ |f (y )||g (x − y )|dy + |f (y )||g (x − y )|dy
0 T
Z x
≤C+ |f (y )||g (x − y )|dy
T

But then x − y > T if T ≤ y ≤ x and so the last integral can be dealt as

follows:
Z x Z x
|f (y )||g (x − y )|dy < e Ay e A(x−y ) dy < xe Ax
T T

April 10, 2019 75 / 92

All we need to observe is that

C + xe Ax < e (A+1)x

for all x sufficiently large and the proof is complete.

April 10, 2019 76 / 92

Theorem 15
Suppose f (t) and g (t) are of exponential type then

L((uf ) ∗ (ug )) = (Lf )(Lg )

In other words the Laplace transform converts a convolution into a

product.

April 10, 2019 77 / 92

Theorem 15
Suppose f (t) and g (t) are of exponential type then

L((uf ) ∗ (ug )) = (Lf )(Lg )

In other words the Laplace transform converts a convolution into a

product. This makes it particularly useful for solving integral and
integro-differential equations of convolution type. Such equations arise
naturally in many applications such as for example the theory of birth and
death processes.

April 10, 2019 77 / 92

Theorem 15
Suppose f (t) and g (t) are of exponential type then

L((uf ) ∗ (ug )) = (Lf )(Lg )

In other words the Laplace transform converts a convolution into a

product. This makes it particularly useful for solving integral and
integro-differential equations of convolution type. Such equations arise
naturally in many applications such as for example the theory of birth and
death processes.
For example, The Renewal Equation - See W. Feller, Introduction to the
theory of Probability theory, Vol - II, Chapter 6.

April 10, 2019 77 / 92

Proof of the convolution theorem

Z ∞
L((uf ) ∗ (ug )) = e −st (uf ) ∗ (ug )dt
Z0 ∞ Z ∞
= e −st dt (u(x)f (x)u(t − x)g (t − x))dx
0 −∞
Z ∞ Z ∞
−st
= e dt (f (x)u(t − x)g (t − x))dx
Z0 ∞ Z0 ∞
= f (x)dx u(t − x)g (t − x)e −st dt
0 0

In the inner integral put t − x = y .

Z ∞ Z ∞
L((uf ) ∗ (ug )) = f (x)dx u(y )g (y )e −s(x+y ) dy
0 −x

April 10, 2019 78 / 92

Proof of convolution theorem contd...

Sine x > 0 the presence of u(y ) would imply that for the integral over
[−x, ∞), the range of integration may be replaced by [0, ∞). Thus,
Z ∞ Z ∞
L((uf ) ∗ (ug )) = f (x)dx u(y )g (y )e −s(x+y ) dy
Z0 ∞ 0
Z ∞
−sx
= e f (x)dx g (y )e −sy dy
0 0
= (Lf )(Lg )

This completes the proof of the theorem.

April 10, 2019 79 / 92

An example

Let a and b be positive real numbers. Let us compute the convolution

u(t)t a−1 ∗ u(t)t b−1

Well, the convolution is zero on the negative real axis as first display below
indicates.
Z ∞
a−1 b−1
u(t)t ∗ u(t)t (x) = u(t)t a−1 u(x − t)(x − t)b−1 dt
−∞
Z ∞
= t a−1 u(x − t)(x − t)b−1 dt
Z0 x
= t a−1 (x − t)b−1 dt
0

April 10, 2019 80 / 92

Proof of convolution theorem contd...

Put t = xz and the integral transforms into

Z 1
a+b−1
x z a−1 (1 − z)b−1 dz
0

which is B(a, b)x a+b−1 and zero if x < 0. Thus the convolution is

B(a, b)u(x)x a+b−1

April 10, 2019 81 / 92

The Beta-Gamma relation

Theorem 16
If a and b are positive real numbers then

Γ(a)Γ(b) = Γ(a + b)B(a, b).

To prove this we apply the convolution theorem to the equation

u(t)t a−1 ∗ u(t)t b−1 = B(a, b)u(t)t a+b−1

The result is
Γ(a) Γ(b) Γ(a + b)
a b
= B(a, b) a+b
s s s
Canceling the factor s a+b gives the result.

April 10, 2019 82 / 92

Integral equations of convolution type

These arise in many applications such as Birth and Death processes, The
Renewal Equation in Probability theory and also in the integral
representations of solutions of many initial value problems for ordinary
differential equations.

April 10, 2019 83 / 92

Integral equations of convolution type

These arise in many applications such as Birth and Death processes, The
Renewal Equation in Probability theory and also in the integral
representations of solutions of many initial value problems for ordinary
differential equations.
An integral equation of convolution type is an equation of the form
Z t
y (t) = f (t) + y (t − x)g (x)dx
0

where f and g are given functions.

April 10, 2019 83 / 92

Example of an integro-differential eqn of conv. type

Let us consider the equation

Z t
y (t) = 1 + y (t − x) sin xdx.
0

Taking the Laplace transform we see that

1 Y (s)
Y (s) = + 2
s s +1
which gives
s2 + 1 1 1
Y (s) = 3
= + 3
s s s
The solution y (t) is then
t2
t+
2

April 10, 2019 84 / 92

Integro-differential equations
Solve the following integro-differential equation of convolution type:
Z t
0
y (t) = 1 + y (t − x) cos xdx, y (0) = 0.
0

Take the Laplace transform of both sides and we get

1 sY
sY = + 2
s s +1
Upon simplifying we get

s2 + 1 1 1
Y = 4
= 2+ 4
s s s
The solution is given by
t3
y (t) = t +
6
April 10, 2019 85 / 92
Tautochrone property of a cycloid

Problem: Consider points P = (0, 0) and Q = (x, y ) in the plane with Q

in the first quadrant and a curve C joining Q and P. A bead slides down
along C from an intermediate position R = (u, v ) to the origin namely P.
How to shape the curve C so that the time of descent from R to P is
independent of the point R ?
Note that a generic point on the curve is (u, v ).

In other words, the time of descent must be constant say T irrespective of

the intermediate point R from where the bead was released.

April 10, 2019 86 / 92

The tautochrome property of a cycloid contd...
The energy equation gives:
1  ds 2
mg (y − v ) = m
2 dt
which can be re-written as
p √ ds ds dv
− 2g y − v = =
dt dv dt
Integrating with respect to t we get,
Z y
p ds dv
2g T = √ (1)
0 dv y −v

Let us now assume the curve C to be given by u = f (v ) so that

ds
q
= 1 + (f 0 (v ))2 = φ(v ) say (2)
dv
April 10, 2019 87 / 92
The tautochrome property of a cycloid contd...
Equation (1) now reads:
p  H 
2g T = (φH) ∗ √
v

Taking Laplace transform we get

√ √
2g π
= √ (Lφ)
s s

which gives √ r √
2g π 2g −1/2
Lφ = = Lt
π s π
The function φ(v ) is given by
√
A 2g
φ(v ) = √ , where A =
v π

April 10, 2019 88 / 92

The tautochrome property of a cycloid contd...

Substituting this in (2) we get

r
A2
f 0 (v ) = − −1
v
A2
Put v = A2 sin2 θ = 2 (1 − cos 2θ) and we get immediately

BA2 A2
Z
f (v ) = − A2 2 cos2 θdθ = (B − 2θ − sin 2θ)
2 2

and the curve u = f (v ) is given parametrically by

A2
(B − 2θ − sin 2θ, 1 − cos 2θ).
2
B is a constant of integration. The curve is an inverted cycloid.

April 10, 2019 89 / 92

Additional Problems on Laplace transforms

1 Solve the integral equation

Z t
−t
y (t) = e + sin(t − x)y (x)dx, Ans:2e −t + t − 1.
0

2 Solve y 00 + ty 0 − 2y = 4 with y (0) = 1, y 0 (0) = 0. Note that this is a

variable coefficient ODE. Ans: y (t) = t 2 − 1.
3 An ODE with continuous and non-differentiable RHS:

00 sin t t ∈ [0, π]
y +y =
0 t≥π

Solve the initial value problem with initial conditions

y (0) = 0, y 0 (0) = 0. Ans: 12 (sin t − t cos t) on [0, π] and −(π/2) cos t
on t ≥ π.

April 10, 2019 90 / 92

4 Compute Lf where
Z π
1
f (t) = cos(t sin θ)dθ
π 0
Find a second order ODE satisfied by this function f (t).
5 Find the Laplace transform of the function
Z t
sin udu
, Ans:s −1 cot−1 s.
0 u
6 Apply the Laplace transform to the ODE
t 2 y 00 + 2ty 0 − (t 2 + p(p + 1))y = 0.
Identify the resulting ODE. The original equation is related to the
Bessel equation.
7 Apply the Laplace transform to the Laguerre equation
ty 00 + (1 − t)y 0 + py = 0.
Solve the resulting first order ODE. Show that the Laguerre equation
has polynomial solutions when p ∈ N ∪ {0}. These polynomials are
called Laguerre polynomials.
April 10, 2019 91 / 92
8 Interpret the identity

s −2 e −(a+b)s = (s −1 e −as )(s −1 e −bs )

in the light of the convolution theorem.

9 Calculate the Laplace transform of sin2 t/t 2 . Ans:
(s log s − log(s 2 + 4) − 2 cot−1 (s/2))
10 Calculate the Laplace transform of the greatest integer function
f (t) = [t]. Use of shift theorem may be easier.
11 Suppose f : [0, ∞) −→ R is a non-negative function of exponential
type show that the Laplace transform F (s) satisfies

(−1)n F (n) ≥ 0, for all n = 0, 1, 2, . . . (1)

A function F satisfying the condition (1) is said to be completely

monotone. Show that the gamma function is completely monotone.

April 10, 2019 92 / 92