THE RIEMANN-STIELTJES INTEGRAL AND SOME OF ITS

APPLICATIONS

BY

OGUNDIMU, BABATUNDE SOLOMON

MATRICULATION NUMBER: 2012/2907

A PROJECT SUBMITTED TO THE DEPARTMENT OF MATHEMATICS,

COLLEGE OF PHYSICAL SCIENCES.
FEDERAL UNIVERSITY OF AGRICULTURE, ABEOKUTA.

IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR

THE AWARD OF BACHELOR OF SCIENCE (B.Sc.) DEGREE

IN MATHEMATICS

OCTOBER, 2015.
 CERTIFICATION

This is to certify that this project was carried out by OGUNDIMU BABATUNDE SOLOMON
and has been read and approved as meeting the requirements of the Department of Mathematics,
Federal University of Agriculture Abeokuta, Ogun State, Nigeria, for the partial fulfillment of the
award of Bachelor of Science degree in Mathematics.

························ ························
OGUNDIMU B. S. Date
(Student)

························ ························
DR. S. A. AKINLEYE Date
(Supervisor)

························ ························
DR. M.O. OMEIKE Date
( Head of Department)

i
 DEDICATION

This project is specially dedicated to ALMIGHTY GOD For raising up the poor out of the dust
and For lifting up the beggar from the dunghill to Set him among princes. And to my beloved father;
CHIEF TIMOTHY OLUSHOLA OGUNDIMU of blessed memory, who out of his encouragement
inspired me in the pursuit of my education. Finally, To All Lovers of Mathematical Analysis.

ii
 ACKNOWLEDGMENT

When a dream or an aspiration becomes a reality, one cannot but thank God who has made it to
be success does not come by chance, but requires dedication and endurance, and if it is out of Gods
knowledge is nothing.
I thank the Lord God Almighty; His love endures forever, nothing I have received that was not
given to me by God. He has surprised me. The Lord is indeed good, for how far he has been
faithful, merciful and daring to me from the beginning of my study in this university to the end of
my degree programme.
I am indeed grateful to my wonderful, hardworking and intelligent project supervisor who at the
same time the NAMSN Staff Adviser, in person of DR. S.A. AKINLEYE for his regular assistance,
constructive criticisms, suggestions and patience during the period of this research works. I pray
that God in his infinite mercies will continually bestow His abundant grace upon him, and crown
all his efforts with success.
Worthy of note also are my indefatigable mentors (lecturers) whom in one way or the other impacted
on me the necessary and sufficient mathematical definitions, theorems, corollaries lemmas and
proofs needed at this level; PROF. J.A. OGUNTUASE, PROF. A.A.A. AGBOOLA, PROF. E.O.
AYOOLA, DR. M.O. OMEIKE, DR. I.A. OSINUGA, DR. S.A. AKINLEYE, DR. E.O. ADELEKE,
DR. I.O. ABIALA, DR. O. MEWOMO, MR. E.O. ILOJIDE, MR. SONUBI and most especially
MR. O. J. OGUNSOLA who serves as a wonderful motivator and encourager to me in all ramifica-
tions. Your morally advice and concern is unforgettable. I would not fail to mention other lecturers
in the department; MR. A.A. ADEYANJU, MR. D.O. ADAMS and MR. A.A. YUSUF who are
always caring about my academic well-being. To you all I say my appreciation is raised to power
infinity(∞).
My sincere appreciation also goes to C.ITP F.A. FATOKUN, MR. D.G. ADEGOKE and MR. S.O.
OGUNRINADE all of department of Mathematics, Federal College of Education Abeokuta for their
mentorship.
My inestimable gratitude goes to someone special to me, who became my mother on May 18, 19xy,∀
x, y ∈ N who enriched my life with quality education, a cause for my passionate commitment to
education. Mum you made me to be the real that you gave birth to. A mother indeed, a mother

iii
like a father. To you, MRS. ESTHER ADEPEJU OGUNDIMU, I say a very big THANK
YOU.
To all my fellows in the department, AdenijiRemi, Oklu Philip, Alade Richard, OdesinaOyebayo,
SanniIbraheem, AdulojuKehinde, Freeborn Blessing, Orija Esther, OdukoyaShina, OdukoyaKay-
ode, Ogunjimi Joel, OtulanaOmoshalewa, AkapnUdoh and lot more. Specifically, OjoOluwatobi
Ruth, BakareFaruqDamilola, and Okusanya Maria Oluwakemi; whom we all entered together as
Direct Entry Students. Im happy we all made it to the end. To you all I say meet you at the TOP.
With all seriousness and sincerity, I acknowledge the efforts of all members of NATIONAL AS-
SOCIATION OF MATHEMATICS STUDENTS OF NIGERIA (NAMSN) FUNAAB Chapter for
their support and constructive criticisms offered me while serving them as the NAMSN President.
Though, the journey was rough but I thank God we landed safely.
My personal appreciation goes to the following personalities for their love, caring and encourage-
ment during the course of my study, Adubifa Mary Oluwaseun (UI), OjeAdedoyin (NYSC), Olubiyi
Moses, Adeniran Samuel, Amosu Matthew, AdebowaleGafar, IdowuOluwatobi, Idowu Esther and
all others that are not listed.

Above all, I say Glory be to God in the highest for making my degree programme
a success and achieving my heart desire amidst good health.

iv
 ABSTRACT

The process of Riemann Integration which is taught in Real Analysis classes is a specific case
of the Riemann-Stieltjes Integration. Thus many of the terms and properties used to describe
Riemann Integration are discussed in this project and they are extended to the Riemann-Stieltjes
integral. This project therefore provides a careful introduction to the theory of Riemann-Stieltjes
integration, and explains the properties of this integral. After doing so, we present some applications
in functional analysis, where we used the fact that continuous functions on a closed interval are
Riemann-Stieltjes Integrable with respect to any function of bounded variation, and this was used
in proving the Riesz Representation Theorem. To show versatility of the Riemann-Stieltjes Integral,
we also present some applications in Probability Theory, where the integral generates a formula
for the Expectation, regardless of its underlying distribution. Other applications considered are
population growth, and Mechanics.

v
 TABLE OF CONTENTS

CERTIFICATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . i
DEDICATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ii
ACKNOWLEDGMENT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii

ABSTRACT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
TABLE OF CONTENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v

1 INTRODUCTION 1

1.1 Riemann-Stieltjes Integration vs Riemann Integration . . . . . . . . . . . . . 3

1.2 Illustrative Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3 Objective of the Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

1.4 Definition of Terms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2 LITERATURE REVIEW 10
2.1 Expectation via the Riemann-Stieltjes integral . . . . . . . . . . . . . . . . . 13

2.2 Riesz and Spectral Theorems . . . . . . . . . . . . . . . . . . . . . . . . . . 14

3 RIEMANN-STIELTJES INTEGRATION 15
3.1 Properties and Theory of Riemann-Stieltjes Integration . . . . . . . . . . . . 15

3.2 Techniques of Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

3.3 Functions of Bounded Variation and Riemann-Stieltjes Integration . . . . . . 26

4 APPLICATIONS OF RIEMANN-STIELTJES INTEGRATION 29

4.1 Application to Probability Theory . . . . . . . . . . . . . . . . . . . . . . . . 29

vi
 4.2 Application to Functional Analysis . . . . . . . . . . . . . . . . . . . . . . . 34
4.3 Application to Population Growth . . . . . . . . . . . . . . . . . . . . . . . 36

4.4 Application to Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

5 CONCLUSION 38
5.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

vii
Chapter 1

INTRODUCTION

The classical Riemann integral (Bernhard Riemann 1826-1866) of ordinary calculus has many
applications including the computation of areas between two curves, volumes of solids of rev-

olution, arc length, work performed by a variable force, hydrostatic pressures, moments and
centers of mass, consumers’ surplus, present value of future income, expected values, and
variance.

In this project we investigate a generalization of the Riemann integral called the Riemann-
Stieltjes integral (Thomas JoannesStieltjes 1856-1894) and its applications. That is, a special
case of the Riemann-Stieltjes integral is the Riemann integral of ordinary calculus.

This project emphasizes on the usefulness of Riemann-Stieltjes Integral in many areas of

Mathematics, Statistics, Mechanics and real-life problems.
For a start, we must recognize that the process of Riemann Integration which is taught in

calculus classes is a specific of Riemann-Stieltjes Integration. Thus many of the terms and
properties used to describe Riemann Integration will be discussed in this project since they
carry over to Riemann-Stieltjes Integration.

In the study of elementary calculus, a good intuitive feel for the idea of Riemann integral of
Rb
the form a f (x)dx is the enclosed area “S” of the region illustrated in Figure 1.1 that lies
between the lines x = a and x = b, the x-axis, and the curve y = f (x).

1
 Figure 1.1: Graph of a Riemann Integral

The definition and existence of Riemann Integral can be given as follows: Let
[a, b] be a given interval on the real line. By a partition ℘ of [a, b] we mean a finite set of

points x0 , x1 , · · · , xn where

a = x0 ≤ x1 ≤ · · · ≤ xn−1 ≤ xn = b

If we write
∆xi = xi − xi−1 (i = 1, . . . n)

and suppose that f is a bounded real function defined on [a, b]. Corresponding to each
partition ℘ of [a, b] we put
Mi = supf (x)(xi−1 ≤ x ≤ xi )

mi = inf f (x)(xi−1 ≤ x ≤ xi )

and define
n
X
U (℘, f ) = Mi ∆xi
i=1
n
X
L(℘, f ) = mi ∆xi
i=1
then,
R b̄
(1) a f dx = Inf U (℘, f )
Rb
(2) a f dx = SupL(℘, f )

2
Where theinf and the supare taken over all partitions ℘of [a, b]. The left members of (1)
and (2) are called the Upper and Lower Riemann Integrals of f over [a, b], respectively.

If the upper and lower integrals are equal, we say that f is Riemann Integrable on [a, b],
written f ∈ < (where< denotes the set of Riemann Integrable functions), and we denote the
common value of (1) and (2) by
Rb
(3) a
f dx,
Or by
Rb
(4) a
f (x)dx.

This is the Riemann Integral of f over [a, b]. Since f is bounded, there exists two numbers;
m and M, such that m ≤ f (x) ≤ M f or a ≤ x ≤ b.
Hence, for every℘,

m(b − a) ≤ L(℘, f ) ≤ U (℘, f ) ≤ M (b − a)

So that the numbers L(℘, f ) and U (℘, f ) form a bounded set. This shows that the upper
and lower integrals are defined for every bounded function f.

1.1 Riemann-Stieltjes Integration vs Riemann Integration

To see how Riemann Integral extends to Riemann-Stieltjes Integral we argue as follows:

Let α be a monotone increasing function on [a, b], and let f be a bounded real-valued func-

tion on [a, b].

For each partition ℘ = x0 , x1 , · · · xn of [a, b] set

∆αi = α(xi ) − α(xi−1 ), i = 1, · · · n

Since α is monotone increasing, ∆αi ≥ 0 for all i.

Let
mi = Inf {f (t) : t ∈ [xi−1 , xi ]}

3
 Mi = Sup{f (t) : t ∈ [xi−1 , xi ]}

The Upper Riemann-Stieltjes sum of f with respect to α and the partition ℘, is defined by
n
X
U (℘, f, α) = Mi ∆αi
i=1

The Lower Riemann-Stieltjes sum of f with respect to α and the partition ℘, is defined by
n
X
L(℘, f, α) = mi ∆αi
i=1

By the given conditions and definitions mi ≤ Mi and ∆αi ≥ 0, we know

L(℘, f, α) ≤ U (℘, f, α).

Let ℘ be any partition of [a, b]. The following sum can be found
n
X
∆αi = [(α(x1 ) − α(x0 )) + (α(x2 ) − α(x1 )) + · · · + (α(xn ) − α(xn−1 ))]
i=1
= α(xn ) − α(x0 )

= α(b) − α(a)

Since Mi ≤ M for all i and∆αi ≥ 0

n
X n
X n
X
Mi ∆αi ≤ M ∆αi = M ∆αi = M [α(b) − α(a)]
i=1 i=1 i=1

and
n
X n
X n
X
mi ∆αi ≥ m∆αi = m ∆αi = m[α(b) − α(a)]
i=1 i=1 i=1

Therefore,

U (℘, f, α) ≤ M [α(a) − α(b)]

and
L(℘, f, α) ≥ m[α(b) − α(a)]

4
Thus, if m ≤ f (x) ≤ M for all x ∈ [a, b], then

m[α(b) − α(a)] ≤ L(℘, f, α) ≤ U (℘, f, α) ≤ M [α(b) − α(a)

for all partitions ℘ of [a, b].

The upper integral of f with respect to α over [a, b], denoted

R b̄ Rb
a
f dα and the lower integral of f with respect to α over [a, b], denoted a
f dα are defined
as:
R b̄
(5) a
f dα = inf {U (℘, f, α) : ℘is a partition of [a, b]},
Rb
(6) a
f dα = sup{L(℘, f, α) : ℘ is a partition of [a, b]}
A function f is said to be Riemann-Stieltjes Integrable with respect to α on [a, b], in the

Riemann sense, if
R b̄ Rb
(7) a
f dα = a
f dα
Rb
We denote the common value of (7) by (8) a
f dα or sometimes by
Rb
(9) a
f (x)dα(x), when f is a bounded real-valued function on [a, b] and α is a monotone
increasing function on [a, b].
This is the Riemann-Stieltjes Integral of f with respect to α, on [a, b], and we write f ∈ <(α).

Remark 1.1 In order for us to truly appreciate the Riemann-Stieltjes Integral, the defini-
tion of the Riemann Integral is reviewed here to allow the Riemann and Riemann-Stieltjes

Integrals to be compared.

Let [a, b], a < b, be a given closed and bounded interval in R. Let ℘ be a partition of

[a, b]. Given

∆xi = xi − xi−1

which is equal to the length of the interval [xi−1 , xi ].

Let

mi = Inf {f (t) : t ∈ [xi−1 , xi ]}

Mi = Sup{f (t) : t ∈ [xi−1 , xi ]}

5
Since f is bounded, by the L.U.B property, the quantities mi and Mi exist in R. If f is a

continuous function on [a, b], then for each i ∃ points ti , si ∈ [xi−1 , xi ] such that Mi = f (ti )
and mi = f (si ).
The upper sum U (℘, f ) for the partition ℘ and function f is defined by
n
X
U (℘, f ) = Mi ∆xi
i=1

The lower sum L(℘, f ) for the partition ℘ and function f is defined by
n
X
L(℘, f ) = mi ∆xi
i=1

Since mi ≤ Mi for all i = 1 · · · n, we know L(℘, f ) ≤ U (℘, f ) for any partition ℘ of [a, b].
Let ℘ be any partition of [a, b]. Since Mi ≤ M and mi ≥ m for all i = 1 · · · n
n
X n
X
U (℘, f ) = Mi ∆xi ≤ M (xi − xi−1 ) = M (b − a)
i=1 i=1

n
X n
X
L(?, f ) = mi ∆xi ≥ m(xi − xi−1 ) = m(b − a)
i=1 i=1

Thus, if m ≤ f (t) ≤ M for all t ∈ [a, b], then

m[b − a] ≤ L(℘, f ) ≤ U (℘, f ) ≤ M [b − a] for all partitions ℘ of [a, b].

The Riemann-Stieltjes Integral is a modification of the Riemann Integral where the function
f is integrated with respect to a function α instead of with respect to x, which means in

the Riemann-Stieltjes Integral ∆α is used while ∆x is used for the Riemann Integral. When
evaluating the Riemann-Stieltjes Integral the upper and lower sums are found by multiplying
Mi and mi , respectively, by ∆αi where

∆αi = α(xi ) − α(xi−1 ).

To evaluate the Riemann Integral the upper and lower sums are found by multiplying Mi

and mi by ∆xi .
When α(x) is defined to equal x(i.e.α(x) = x) then Riemann-Stieltjes Integral and Riemann

6
Integral are equivalent.
Let α(x) = x then α(xi ) = xi and α(xi−1 ) = xi−1

Therefore,
∆α(xi ) = α(xi ) − α(xi−1 ) = xi − xi−1 = ∆xi .

1.2 Illustrative Examples

Let f (x) = −8 + 6x − x2 on [2, 5] and let ℘ = {2, 2.6, 3, 5}.

√
1. If α(x) = x, then
n
X
+
RS (f, α, ℘) = Mk (α(xk ) − α(xk−1 ))
k=1
= f (2.6)(α(2.6) − α(2)) + f (3)(α(3) − α(2.6)) + f (5)(α(5) − α(3))

≈ 0.79013633716.

and
n
X
−
RS (f, α, ℘) = mk (α(xk ) − α(xk−1 ))
k=1
= f (0)(α(2.6) − α(2)) + f (2.6)(α(3) − α(2.6)) + f (5)(α(5) − α(3))

= −11.803892979

2. If 
 2if x < 3.75
α(x)
 Xif x ≥ 3.75

then
n
X
+
RS (f, α, ℘) = Mk (α(xk ) − α(xk−1 ))
k=1
= f (2.6)(α(2.6) − α(2)) + f (3)(α(3) − α(2.6)) + f (3)(α(5) − α(3))

= 3

7
 and
n
X
−
RS (f, α, ℘) = mk (α(xk ) − α(xk−1 ))
k=1
= f (0)(α(2.6) − α(2)) + f (2.6)(α(3) − α(2.6)) + f (5)(α(5) − α(3))

= −9

1.3 Objective of the Study

The objective of this project is

1. to study the Riemann-Stieltjes Integral and its properties and;

2. to apply the integral to some real-life problems.

1.4 Definition of Terms

We recall the following definitions of some mathematical terms which will be used in the
course of this research work. Some of the terms are:

• Partition of an Interval: A partition ℘ of [a, b] = {x ∈ R ≤ x ≤ b} is a finite set of

points ℘ = {x0 , x1 , x2 · · · xn } such that a = x0 < x1 < x2 < x3 · · · xn = b.
Each [xi−1 , xi ], i = 1, 2, · · · n is called a sub-interval of the partition.

We denote the set of all partitions of [a, b] by P[a,b] .

• Mesh or Norm of a Partition:The norm or mesh of ℘ is defined to be length of the

largest sub-interval into which partition divides [a, b]. That is,kP k = max[x1 − x0 , x2 −
x1 , · · · xn − xn−1 ].

• Tagged Partition: Given a point ti selected from each sub-interval [xi−1 , xi ], i =

1, 2, · · · n called the tags of the sub-interval. A set of ordered pairs Ṗ = {[xi−1 , xi ]ti }ni=1

8
 of sub-intervals and corresponding tags is called a Tagged Partition. The mesh of a
tagged partition is the same as that of an ordinary partition.

• Refinement: We say that the partition P* of [a, b] is a refinement of Pif P* ⊃ P (that

is, if every point of Pis a point of P*). Given two partitions, P1 and P2 we say that

P* is their common refinement if P ∗ = P1 ∪ P2 .

• Least Upper Bound or Supremum: Let E be a nonempty subset of R that is bounded

above. An element α ∈ R is called the least upper bound or supremumof E if

1. α is an upper bound of E, and

2. ifβ ∈ R satisfies β < α, then β is not an upper bound of E.

• Greatest Lower Bound or Infimum: Let E be a nonempty subset of R that is bounded

below. An element γ ∈ R is called the greatest lower bound or infimum of E if

1. γ is an lower bound of E, and

2. ifβ ∈ R satisfies β > γ, then β is not a lower bound of E.

9
Chapter 2

LITERATURE REVIEW

Rb
Until 1820’s, the integral a
f denoted the area enclosed by the curve f from a tob and it is
0
defined as F (b) − F (a), where F is the function defined on R such thatF (x) = f (x). Many

authors before Bernhard Riemann (1826 - 1866) and Thomas Joannes Stieltjes (1856 - 1894)
P
such as Cauchy and Euler have considered approximation to the integral by i f (xi )(α(xi )−
α(xi−1 )), in situations where F(x) could not be computed. However, it was Riemann in

1853 who developed the definition of integration and provided that necessary and sufficient
conditions for a bounded function defined on [a, b] to be integrable. Later on, in 1894 Stieltjes
gives a modification of the Riemann Integral which helps give insight into the Riemann-

Stieltjes Integral. It serves as an instructive and useful precursor of the Lebesgue integral.
The researcher approach to Riemann-Stieltjes Integration was summarized below which was
a modification of the Riemann Integral.
Rb
” We defined the Riemann Integral a f (x)dx to be the limit as we refined the tagged partition
x = ((x0 , · · · xn ), (t1 , · · · tn )) of the Riemann Sum f (x) = ni=1 f (ti )(xi − xi−1 ).But why did
P

we multiply by (xi − xi−1 )? Well, that was the width of a rectangular strip we were using

to approximate part of the area under the graph of f . But why should we automatically use
that difference as the ” width”?
Let’s imagine we’re walking past a fence. Sometimes we walk faster, and sometimes we walk

slower, but at time t we can measure the height of the fence right next to us:f(t). So what’s

10
the area of the fence? If we just integrated f(t) we’d get the wrong the answer . The samples
we made when walking fast made fat rectangles, while the samples we made when we were

walking slowly got paired with skinny rectangles, but we gave them same weight if they took
the same time to get through that segment of the partition. We need to reweight our sums to
compensate for how fast we’re walking!

Okay, so how wide should we make the rectangles? Let’s say that at time t we’re at position
α(t) along the fence. Then in the segment of the partition times xi−1 and xi we move from
position α(xi−1 ) to the position α(xi ), so we should make the width come out to (α(xi ) −

α(xi−1 )). I’ll put this into our formalism before and get the “Riemann-Stieltjes Sum”:
n
X
fα,x = f (ti )(α(xi ) − α(xi−1 )).
i=1
And now I can take the limit over tagged partitions as before to get the “Riemann-Stieltjes

Integral”:
Z b
f (x)dα(x).
a

Integration has been an important part of mathematics since the 17th Century, when Got-
tfried Wilhelm Leibniz and Isaac Newton constructed the Fundamental Theorem of Calculus,
which says that the area is essentially the same as taking an anti-derivative. In the 19th

Century, Cauchy investigated integrals of continuous functions. The integral that he used
was later redefined by Riemann, which he used to investigate integrals of discontinuous func-
tions. His definition was far easier to understand and to teach than all of the previous ones,

which resulted in the Riemann Integral becoming the standard integral to be taught to stu-
dents in mathematics class which helps, give an insight to the Riemann-Stieltjes Integral (a
generalization of Riemann Integral).

An important generalization of the Riemann-Stieltjes Integral is the Lesbegue-Stieltjes Inte-

gral which generalizes the Riemann-Stieltjes integral in a way analogous to how the Lebesgue
integral generalizes the Riemann integral. If improper Riemann-Stieltjes integrals are al-

lowed, the Lebesgue integral is not strictly more general than the Riemann-Stieltjes integral.
The Lebesgue integral was based on measure theory and, when mastered, proved a very

11
powerful tool in mathematical applications.
This measure theory made that the Lebesgue Integral was much harder to get familiar with

which is the main reason that teachers choose to educate their students with the specific case
of the Lebesgue-Stieltjes integral which is the Riemann-Stieltjes Integral until they reach the
final stages of their academic career.

In the following years, investigations into even more general forms of integration continued,
mainly because there was lack of good methods to find primitives of functions.
In 1957, the mathematicians Jaroslav Kurzweil and Ralph Henstock independently came up

with a simple expression of an integral.

This integral was based on the Riemann theory and just as easy to work with, but with
powers to match and in certain cases even exceeds those of the Lebesgue integral.

In the research on the calculus there has been a tendency to concentrate on differentiation
and its concepts rather than integration.
Here, a definition of definite integral can be considered whereby a set of n rectangles, which

can be freely constructed, is used to approximate the area. This requires that, for a given
set of well-defined rectangles, the tops of the rectangles intersect with the graph of the func-
tion, the widths of the rectangles partition the horizontal distance involved, and as n → ∞,

the width of the rectangles approaches zero. But a definition of the definite integral based
on these concepts is often quickly forgotten once the fundamental theorem of the calculus
linking areas and anti-derivatives is introduced. This is not unnatural, because the Riemann-

Stieltjes sums used in the definitions are tedious to calculate by hand and require the taking,
and understanding, of a limit, whereas the anti-derivatives involve clear algorithmic pro-
cesses. (Hong, 1996).

The applications of the definite integral are numerous and varied with applications rang-
ing from physics, to engineering, to satellite orbits. Thus, Riemann-Stieltjes sums, as an
approximation to the definite integral, also have many real world applications

12
2.1 Expectation via the Riemann-Stieltjes integral

While the Lebsegue Integral is more than enough for most purposes, there are other types

of integrals that are frequently used in applications. In particular, probability measures are
often induced by distribution functions, and this allows expectation to be expressed as the
Riemann-Stieltjes Integrals.

In additions, when one deals with distribution functions that have densities, all one needs
is the familiar Riemann Integral. Our treatment is fairly comprehensive, but particular em-
phasis is given to Riemann-Stieltjes Integrals within the realm of probability theory.

In abstract term, we defined the expected value of a random variable with respect to a
probability measure as its Lebesgue Integral with respect to that measure. The problem is
that this definition provides us with little ammunition for computing expectations in general

(at least for non-simple random variables). Fortunately, in many situations of particular
interest, we can actually bypass the abstract definition of the Lebesgue integral, and instead
work out the expectation of random variables by means of simpler method; Riemann-Stieltjes

Integral.
Using the Riemann-Stieltjes Integral we can write the expectation in a unified manner
R∞
Eg (X) = −∞ g(x)dFX (x) where g is the cumulative probability distribution function of

a random variable X.
This uses limits of Riemann-Stieltjes sums

n
X
R(g, F ) = g(xi )∆FX (xi )
i=1

For discrete random variables, ∆FX (xi ) = FX (xi+1 )−FX (xi ) = 0 if the i-th interval does not
contain a possible value for the random variable X. Thus, the Riemann-Stieltjes converges
P
to x g(x)fX (x) for X having mass function fX .

For continuous random variables, ∆FX (xi ) ≈ fX (xi )∆x. Thus, the Riemann-Stieltjes

13
 R∞
sum is approximately a Riemann sum for the product g.fX and converges to −∞
g(x)fX (x)dx
for X having mass function fX .

2.2 Riesz and Spectral Theorems

The Riemann-Stieltjes Integral appears in the original formulation of Riesz theorem which
represents the dual space of the Banach space C[a, b] of continuous functions in an interval

[a, b]. In order to prove the Riesz theorem, it was necessary to prove that continuous func-
tions on a closed interval are Riemann-StieltjesIntegrable with respect to any function of
bounded variation.

Haaser and Sullivan use the equivalence of the Riemann-Stieltjes and Darboux-Stieltjes In-
tegrals to prove this integrability result. Later, the theorem was reformulated in terms of
measures.

The Riemann-Stieltjes Integral also appears in the formulation of the spectral theorem for
(non-compact) self-adjoint (or more generally, normal) operators in a Hilbert space. In this
theorem, the integral is considered with respect to a spectral family of projections.

14
Chapter 3

RIEMANN-STIELTJES
INTEGRATION

3.1 Properties and Theory of Riemann-Stieltjes Integration

Definition 3.1.1 Let f be a bounded function on [a, b], a nondecreasing function on [a, b],

and ℘ a partition of [a, b]. If

Mk = lubx∈[xk−1 ,xk ] f (x) (k = 1, 2, 3 · · · , n)

mk = glbx∈[xk−1 ,xk ] f (x) (k = 1, 2, 3 · · · , n)

then the Upper Riemann-Stieltjes sum of f with respect to α on [a, b] is defined by

n
X
RS + (f, α, ℘) = Mk (α(xk ) − α(xk−1 ))
k=1

and the Lower Riemann-Stieltjes sum of f with respect to α on [a, b] is defined by

n
X
−
RS (f, α, ℘) = mk (α(xk ) − α(xk−1 ))
k=1

Since f is bounded function, all Mk and mk are finite. Hence, the upper and lower Riemann-
Stieltjes sums are well-defined real numbers. Observe that f is a nonnegative, continuous
function on [a, b] and α(x) = x, then RS + (f, α, ℘) and RS − (f, α, ℘) are the sum of the areas

15
of the circumscribed and inscribed rectangles induced by the partition ℘ relative to the curve
y = f (x) over [a, b].

Theorem 3.1.1 Suppose f maps [a, b] into [m, M ] and α is nondecreasing on [a, b].

1. If ℘ is a partition of [a, b], then

m(α(b) − α(a)) ≤ RS − (f, α, ℘) ≤ RS + (f, α, ℘) ≤ M (α(b) − α(a))

2. If the partition P2 is a refinement of the partition P1 on [a, b], then

RS − (f, α, P1 ) ≤ RS − (f, α, P2 ) ≤ RS + (f, α, P2 ) ≤ RS + (f, α, P1 )

3. If P1 and P2 are any two partitions of [a, b], then

RS − (f, α, P1 ) ≤ RS + (f, α, P2 )

Proof:

1. Since m ≤ mk ≤ Mk ≤ M for k = 1, 2, 3, · · · , n,

n
X
m(α(b) − α(a)) = m(α(xk ) − α(xk−1 )) (telescopes)
k=1
Xn
≤ mk (α(xk ) − α(xk−1 ))
k=1
n
X
≤ Mk (α(xk ) − α(xk−1 ))
k=1
n
X
≤ M (α(xk ) − α(xk−1 )) (telescopes)
k=1
= M (α(b) − α(a)).

The desired result is established.

16
 2. First, we suppose that P2 contains exactly one more point than P1 . Let this point x∗
be such that xj−1 < x∗ < xj for some particular value of j. Then

n
X
−
RS (f, α, P1 ) = mk (α(xk ) − α(xk−1 ))
k=1
j−1 n
X X
= mk (α(xk ) − α(xk−1 )) + mj (α(xj ) − α(xj−1 )) + mk (α(xk ) − α(xk−1 ))
k=1 k=j+1

j−1 n
X X
∗ ∗
= mk (α(xk ) − α(xk−1 ))+mj (α(x )−α(xj−1 ))+mj (α(xj )−α(x ))+ mk (α(xk ) − α(xk−1 ))
k=1 k=j+1

j−1
X
≤ mk (α(xk ) − α(xk−1 ))+(infx∈[xj−1 ,x∗ ] f (x))(α(x∗ )−α(xj−1 ))+(infx∈[x∗ ,xj ] f (x))(α(xj )−α(x∗ ))
k=1
n
X
+ mk (α(xk ) − α(xk−1 ))
k=j+1

= RS − (f, α, P2 ).

By a mathematical induction argument on the number of extra points in P2 , we have

that
RS − (f, α, P1 ) ≤ RS − (f, α, P2 ).

In a similar fashion, one can prove that

RS + (f, α, P2 ) ≤ RS + (f, α, P1 ).

3. Let T be a common refinement of the partitions P1 and P2 . Then by parts (1) and (2),
we have

RS − (f, α, P1 ) ≤ RS − (f, α, T ) ≤ RS + (f, α, T ) ≤ RS + (f, α, P2 ).

Definition 3.1.2 Let f be a bounded function on [a, b] and let αbe a nondecreasing function
on [a, b]. Then we define the upper Riemann-Stieltjes Integral of f with respect to

α over [a, b] by

17
 Z b
U f dα = glb℘ RS + (f, α, ℘)
a

andthe lower Riemann-Stieltjes Integral of f with respect to α over [a, b] by

Z b
L f dα = lub℘ RS − (f, α, ℘)
a
Rb Rb
If U a
f dα = L a
f dα, then we say that f is Riemann-StieltjesIntegrable over [a, b]

and we denote this common value by

Z b Z b
f dα or f (x)dα(x)
a a

We call the function fthe integrand and the function α the integrator. By convention, if
a < b, then we define
Z a Z b
f dα = − f dα
b a

and
Z a
f dα = 0.
a

Theorem 3.1.2 Let f be a bounded function on [a, b] and let α be a nondecreasing on [a, b].
Then
Z b Z b
m(α(b) − α(a)) ≤ L f dα ≤ U f dα ≤ M (α(b) − α(a))
a a

Proof: We only prove that

Z b Z b
L f dα ≤ U f dα
a a

By part (3) of Theorem 3.1.1, if P1 and P2 are any two partitions of [a, b], then

RS − (f, α, P1 ) ≤ RS + (f, α, P2 ) (3.1)

Fix the partition P1 . Taking the greatest lower bound of the right hand side (3.1) asP2 varies
over all possible partitions P2 of [a, b] yields
Z b
− +
RS (f, α, P1 ) ≤ glbP2 RS (f, α, P2 ) = U f dα (3.2)
a

18
Now, taking the least upper bounds of the left hand side of (3.2) as P1 varies over all possible
partitions of [a, b] we obtain

Z b Z b
L f dα ≤ U f dα
a a

In the next example we show that both strict inequality and equality may occur.

Example 3.1 1. Let 

 1if x ∈ Q
f (x) =
 0if x ∈ R − Q

and let α be any non decreasing function defined on [a, b] with α(a) < α(b). If ℘ is any
partition of [a, b], then since Q and R − Q are dense in [a, b].
n
X
−
RS (f, α, ℘) = mk (α(xk ) − α(xk−1 ))
k=1
n
X
= 0(α(xk ) − α(xk−1 ))
k=1
= 0

and
n
X
+
RS (f, α, ℘) = Mk (α(xk ) − α(xk−1 ))
k=1
Xn
= 1(α(xk ) − α(xk−1 )) (telescopes)
k=1
= α(xn ) − α(x0 ) > 0

Rb Rb
Hence L a
f dα = 0 < α(b) − α(a) = U a
f dα
We conclude that theDirichlet function is not Riemann-StieltjesIntegrable with re-

spect to any nonconstant integrator function α.

Remark 3.1 Dirichlet function also known as indicator function of the rational num-

bers is an example of a nowhere continuous function or everywhere discontinuous func-

tion, a function that is not continuous at any point of its domain.

19
2. Let t ∈ (a, b) and suppose that f is a bounded function on [a, b] that is continuous at
x = t. Suppose also that C1 < C2 and that

 C1 if x < t
α(x) =
 C
2 if x ≥ t

Let ℘ be any partition of [a, b]. Then

Z b Z b
−
RS (f, α, ℘) ≤ L f dα ≤ U f dα ≤ RS + (f, α, ℘)
a a

Let  > 0 be given. Since f is continuous at x = t, there exists a δ > 0 so that if

|x − t| < δ we have |f (x) − f (t)| < 

⇐⇒ f (t) −  < f (x) < f (t) + 

Choosing a partition ℘ = {x0 = a, x1 , x2 , x3 = b}where t − δ < x1 < t < x2 < t + δ.

Then

RS − (f, α, ℘) = m2 (α(x2 ) − α(x1 ))

= m2 (C2 − C1 )

≥ (f (t) − )(C2 − C1 )

= (f (t)(C2 − C1 ) − (C2 − C1 )

and

RS + (f, α, ℘) = M2 (α(x2 ) − α(x1 ))

= M2 (C2 − C1 )

≤ (f (t) + )(C2 − C1 )

= (f (t)(C2 − C1 ) + (C2 − C1 )

It follows that
Z b Z b
(f (t)(C2 − C1 ) − (C2 − C1 ) ≤ L f dα ≤ U f dα ≤ (f (t)(C2 − C1 ) + (C2 − C1 )
a a

20
 The above holds for all  > 0. Taking the limit of the above as  → 0 produces
Z b
f dα = f (t)(C2 − C1 )
a

Or
Z b
f dα = f (t)(α(t+ ) − α(t− )
a

That is, f is Riemann-StieltjesIntegrable with respect to α.

Theorem 3.1.3 ( Riemann’s Condition for Integrability)

Suppose that f is bounded on [a, b] and α is non decreasing function on [a, b]. Then f is

Riemann-Stieltjes Integrable with respect to α on [a, b] iff for each  > 0∃ a partition ? of
[a, b] (depending on ) so that

RS + (f, α, ℘) − RS − (f, α, ℘) < 

Proof:
(⇒) Assume that f is Riemann-Stieltjes Integrable with respect to α on [a, b]. Let  > 0
be given. Since
Z b
L f dα = lub℘ RS − (f, α, ℘)
a

Where ℘ ranges over all possible partitions of [a, b], there exists a partition P1 of [a, b] such
that
Rb Rb
1. L a
f dα − 
2
< RS − (f, α, P1 ) ≤ L a
f dα
Because
Z b
U f dα = glb℘ RS + (f, α, ℘)
a

Where ℘ ranges over all possible partitions of [a, b], there exists a partition P2 of [a, b]

so that
Rb Rb 
2. U a
f dα ≤ RS + (f, α, P2 ) < U a
f dα + 2

Let P3 be a common refinement of P1 and P2 . Then from the inequalities (1) and (2) we
have that

21
 Rb Rb
3. RS + (f, α, P3 ) − RS − (f, α, P3 ) < (U a
f dα + 2 ) − (L a
f dα − 2 )
Z b Z b
= (U f dα − L f dα) + 
a a

Since f is Riemann-Stieltjes Integrable with respect to α on [a, b]

Z b Z b
L f dα = U f dα
a a

Thus, (3) becomes

RS + (f, α, P3 ) − RS − (f, α, P3 ) < .

(⇐) Suppose that for each  > 0 there exists a partition ℘ of [a, b] so that

4. RS + (f, α, ℘) − RS − (f, α, ℘) < . Let  > 0 be given and let ℘ be the partition of [a, b]
corresponding to  so that (4) is satisfied.

By definition 3.1.2

Z b Z b
−
RS (f, α, ℘) ≤ L f dα ≤ U f dα ≤ RS + (f, α, ℘)
a a

and so we have
Z b Z b
U f dα − L f dα ≤ RS + (f, α, ℘) − RS − (f, α, ℘) < 
a a

Because  > 0 was arbitrary, we conclude that

Z b Z b
L f dα = U f dα
a a

Hence, f is Riemann-Stieltjes Integrable with respect to α on [a, b].

Theorem 3.1.4 If f is continuous on [a, b] and α is a nondecreasing function on [a, b] with

α(b) − α(a) > 0, then f is Riemann-Stieltjes Integrable with respect to α on [a, b].

Proof: We use Riemann’s Condition for Integrability to establish the result. Let
 > 0 be given. Since f is continuous on the closed interval [a, b], f is uniformly continuous

22
there. So, there exists a δ > 0 such that


|f (x) − f (y)| < whenever|x − y| < δ
α(b) − α(a)

Let ℘ = {x0 , x1 , x2 · · · xn } be a partition of [a, b] with xk − xk−1 < δ. Then

n
X
+ −
RS (f, α, ℘) − RS (f, α, ℘) = (Mk − mk )(α(xk ) − α(xk−1 ))
k=1
Xn
= (f (sk ) − f (tk )(α(xk ) − α(xk−1 ))(sk , tk ?[xk−1 , xk ])
k=1
n
X 
< (α(xk ) − α(xk−1 ))
k=1
α(b) − α(a)
n
 X
= (α(xk ) − α(xk−1 )) (telescopes)
α(b) − α(a) k=1

= (α(b) − α(a))
α(b) − α(a)
= 

Where we used the Extreme Value Theorem to obtain Mk = f (sk ) and mk = f (tk ) for some
sk , tk ∈ [xk−1 , xk ]. By Riemann’s Condition for Integrability, f is Riemann-Stieltjes

Integrable with respect to α on [a, b].

The previous two theorems provide us with a wealth of functions that are Riemann-Stieltjes
Rb
Integrable although neither result gives us a means to evaluate a f dα beyond that given in

Definition 3.1.2
At this point actually calculating the value of Riemann-Stieltjes Integral is still a highly
nontrivial task as the next example illustrates.

Example 3.2 Let f (t) = t3 , and α(t) = t2 on [1, 5] and let

4 4(2) 4(3) 4(n − 1)

Pn = {x0 = 1, x1 = 1 + , x2 = 1 + , x3 = 1 + , · · · xn−1 = 1 + , xn = 5}
n n n n

be a regular partition of [1, 5]. By theorems 3.1.1 , 3.1.2, 3.1.3 and 3.1.4 we have that
f is Riemann-Stieltjes Integrable with respect to α on [1, 5]. We have
n
X
+
RS (f, α, ℘n ) = Mk (α(xk ) − α(xk−1 ))
k=1

23
 n
X 4k 3 4k 4(k − 1) 2
= (1 + ) ((1 + )2 − (1 + ))
k=1
n n n
n
X 2048 2048 1024 768 768 128 192 8 16
= ( 5 )k 4 + ( 4 − 5 )k 3 + ( 3 − 4 )k 2 + ( 2 − 2 )k + ( − 2 )
k=1
n n n n n n n n n
n n n n n
2048 X 4 2048 1024 X 3 768 768 X 2 128 192 X 8 16 X
=( 5 ) k +( 4 − 5 ) k +( 3 − 4 ) k +( 2 − 2 ) k + ( − 2) 1
n k=1
n n k=1
n n k=1
n n k=1
n n k=1
2048 n(n + 1)(2n + 1)(3n2 + 3n − 1) 2048 1024 n2 (n + 1)2 768 n(n + 1)(2n + 1)
=( 5 ) +( 4 − 5 ) +( 3 )
n 30 n n 4 n 6
128 192 n(n + 1) 8 16
+( 2 − 2 ) + ( − 2 )n
n n 2 n n
4 3 2
8(2343n + 3510n + 620n − 720n − 128) 6248
= 4
=
15n 5
On taking the limit as n → ∞. That is, RS + (f, α, ℘n ) is a sequence of real numbers con-
verging to 6248
5
. Similarly, RS − (f, α, ℘n ) → 6248
5
as n → ∞.
Consequently,
R5
1
f dα = 6248
5
.

The next three results give some of the standard properties for the Riemann-Stieltjes
integral.

Theorem 3.1.5 1. If f and g are both Riemann-Stieltjes integrable with respect to α

on [a, b] and if c and d are any two real numbers, then cf + dg is Riemann-Stieltjes
Integrable with respect to α on [a, b] and
Z b Z b Z b
(cf + dg)dα = c f dα + d gdα
a a a

2. If f is Riemann-Stieltjes integrable with respect to both α and β on [a, b] and if c and

d are nonnegative real numbers, then f is Riemann-Stieltjes Integrable with respect to
γ = cα + dβ and

Z b Z b Z b
f dγ = c f dα + d f dβ
a a a

3. If f and g are both Riemann-Stieltjes Integrable with respect to α on [a, b], then f g is
Riemann-Stieltjes Integrable with respect to α on [a, b].

24
3.2 Techniques of Integration

Theorem 3.2.1 (Integration by parts)

Rb
. If f ∈ <(α) on [a, b], then α ∈ R(f ) on [a, b] and a
f (x)dα(x) = f (b)α(b) − f (a)α(a) −
Rb
a
α(x)df (x).
R 10
Example 3.3 Let f (x) = x, and α(x) = x + [x]. Find 0
f (x)dα(x)

Solution
Z 10 Z 10
f (x)dα(x) = f (10)α(10) − f (0)α(0) − α(x)df (x)
0 0
Z 10
= 10x20 − 0x0 − (x + [x])dx
0
Z 10
= 200 − 50 − [x]dx
0

= 150 − 45 = 105

Theorem 3.2.2 (Change of Variables)

Suppose that f ∈ <(α) on [a, b] and g is α strictly increasing continuous function on [c, d]
with a = g(c), b = g(d).Let h = f ◦ g , β = α ◦ g. Then h ∈ <(β) on [c, d] and

Z b Z d Z d
f (x)dα(x) = f (g(t))dα(g(t)) = h(t)dβ(t)
a c c

.
√
Example 3.4 Let y = x, we have
Z 4 √ √
([ x] + x2 )d x
0
Z 2
= ([y] + y 4 )dy
Z0 2 Z 2
= [y]dy + y 4 dy
0 0
1
= 1 + (y 5 |5y=0
5
37
=
5

25
3.3 Functions of Bounded Variation and Riemann-Stieltjes Inte-
gration

The concept of functions of bounded variation first arose in the 19th Century in connection
with the alignment of curves when studying convergence of Fourier series. It is precisely with
respect to these functions that we may define the Riemann-Stieltjes Integral of continuous

functions.

Definition 3.3.1 A function f : [a, b] → R is said to be of bounded variation on [a, b] if

there exists an M ∈ R such that

n
X
W℘ (f ) = |f (tj ) − f (tj−1 )| ≤ M
j=1

for each partition ℘ = {x0 , x1 , · · · xn } of [a, b] where W℘ (f ) is a sum corresponding to a

partition ℘.

Remark 3.2 1. If f is continuously differentiable on [a, b], then f is of bounded variation

on [a, b].

Proof: Let ℘ = {x0 , x1 , · · · xn } be a partition of [a, b]. by the Extreme Value Theorem,
0
there is an M > 0 such that |f (x)| ≤ M for all x ∈ [a, b]. Therefore it follows from the
Mean Value Theorem that for each k between 1 and n there is a point ck ∈ [xk−1 , xk ]

such that

0
|f (xk ) − f (xk−1 )| = |f (ck )|(xk − xk−1 ) ≤ M (xk − xk−1 )

By telescoping, we obtain W℘ (f ) ≤ M (b − a) for any partition℘ of [a, b]. Therefore,

W℘ (f ) ≤ M (b − a)

However, there exist functions of bounded variation which are not continuously differ-

entiable. An example of this is the indicator function which is of bounded variation but
not continuously differentiable.

26
 2. If f is monotone on [a, b], then f is of bounded variation on [a, b].

Proof: Let f be increasing on [a, b] and ℘ = {x0 , x1 , · · · xn } be a partition of [a, b]. Then

by telescoping
n
X n
X
|f (xj ) − f (xj−1 )| = (f (xj ) − f (xj−1 ))
j=1 j=1
= f (xn ) − f (x0 )

= f (b) − f (a)

= M <∞

Thus, W℘ (f ) ≤ M.
But there exist functions of bounded variation which are not monotone. The function

f (x) = sin x is of bounded variation on [0, 2π] but is not monotone.

3. If f is of bounded variation on [a, b], then f is bounded on [a, b]

Proof: Let x ∈ [a, b] and by definition

|f (x) − f (a)| ≤ |f (x) − f (a)| + |f (b) − f (x)| ≤ W℘ (f )

Hence by the triangle inequality

|f (x)| ≤ |f (a)| + W℘ (f )

However, there exist bounded functions which are not of bounded variation. An example
of this is the function 
 sin 1 x 6= 0
x
y=
 0 x=0
which is bounded but not of bounded variation on the interval [0, b].
Bounded Variation is important to the study of Riemann-Stieltjes Integration due to

the following theorems.

Theorem 3.3.1 Suppose that f is continuous on [a, b] and that α is of bounded variation on
Rb
[a, b]. Then the Riemann-Stieltjes Integral a f dα exists.

27
To prove this theorem, Jordan’s Theorem which is stated below will be used.

Theorem 3.3.2 (Jordan’s Theorem)

: A function f is of bounded variation on [a, b] if and only if it can be written as the difference
of two bounded increasing functions on [a, b].

Proof: We can rewrite the integral with respect to α as the difference of the integrals
respect to α1 and α2 if α1 − α2 = α.

Thus, α is the difference of two monotone increasing functions which can be evaluated
separately and then subtracted. Let ℘ be a partition of [a, b]. Then
Z b
Sup{L(℘, f, α)} ≤ f dα ≤ Inf {U (℘, f, α)}
a

It is enough to show
Sup{L(℘, f, α)} = Inf {U (℘, f, α)}

To prove the Riemann-Stieltjes Integral exists. If α is constant on [a, b] then ∆α = 0 thus

Rb
a
f dα = 0 as do the upper and lower sums. If α is not constant though let  > 0, the

uniform continuity of f implies there exists δ > 0 such that if


|xi − xi−1 | < δ, then Mi − mi < α(b)−α(a)

Therefore, if

|xi − xi−1 | < δ

n
X
0 ≤ Inf {U (℘, f, α)} − Sup{L(℘, f, α)} = (Mi − mi )[f (xi ) − f (xi−1 )] < 
i=1

Thus,
Inf {U (℘, f, α)} − Sup{L(℘, f, α)} < 

So the Riemann-Stieltjes Integral of with respect to α exists.

28
Chapter 4

APPLICATIONS OF
RIEMANN-STIELTJES
INTEGRATION

4.1 Application to Probability Theory

In this section we will use the Riemann-Stieltjes Integral to define the expectation of a
random variable and function of several random variables.

The main reason for defining the expected value with the Riemann-Stieltjes Integral, instead
of with an ordinary Riemann Integral, is that it generates a general formula which holds for
continuous and discrete random variables as well as mixtures of the two. Thus it offers a

more unified approach to the theory of random variables.

In this project work, we consider the real line as our sample space. We begin by defining some
important concepts of probability theory. Our base is a probability space which includes a

set of all possible outcomes, called outcome space Ω, and a probability measure P, which
specifies the likelihood of an event occurring.

Definition 4.1.1 Given a random experiment with an outcome space Ω, we define a random
variable X as a measurable function that assigns exactly one real number X(ω) = x to each

29
element ω ∈ Ω.

A random variable can either be discrete, only assuming a countable number of real values,
or continuous, theoretically assuming an infinite number of real values on finite intervals of

Ω. A discrete random variable can for example be a roll of a pair of dice, and a continuous
random variable could be the height of a person. We will throughout this section assume
that every random variable is defined from the outcome space Ω to a given real dimension.

The probabilities of random variables are measured with the use of distribution function.

Definition 4.1.2 For a random variable X we define the cumulative distribution function
of X as FX (x) = P (X ≤ x) .

P (X ≤ x)represents the probability of the event that the random variable X is equal to

or no larger than a given value of x. Similarly, the probability of the event that a random
variable X belongs to an interval (xi−1 , xi ] is given by

P (xi−1 < X ≤ xi ) = FX (xi ) − FX (xi−1 )

. In the discrete case the distribution function is the area under the probability function.
The probability function is constant on each possible outcome so the distribution function

becomes a sum of step functions with jumps at each probability. In the continuous case the
distribution function, evaluates at a point x, is the area under an integrable function called
the probability density function, from minus infinity up to x.

For random variables that are a mixture of discrete and continuous random variables, the
distribution function will be a combination of two types such that it will be continuous
everywhere except at the points of positive probability, where it will have discontinuous jumps

with a height equal to the corresponding probability. Common to all types of distribution
functions are that they are non-decreasing functions from zero to one.
We often want to estimate the probability of an event connected to a random variable. This

is done by looking at the expectation of a random variable. Recall that the expected value
is a weighted average of all possible values, where the weights are the probabilities of certain

30
events occurring. The expectation is a linear operator which by definition exists whenever
the expression of the distribution function is finite or converges absolutely

If the expectation does exist, then the excepted value of a random variable X is given by
 P
∞
i=1 xi P (X = xi ) whenXisdiscrete

E(X) =
 ∞ xf (x)dx
R
whenXiscontinuous
−∞ X

Here fX (x) is the probability density function of X. This function is related to the cumulative
distribution function by FX0 = fX , if fX is continuous at x.
The integral above is an ordinary Riemann Integral in which integration is performed by

summarizing over
n
X
lim ηi fX (ηi )(xi − xi−1 )
max |xi −xi−1 |→0
i=1

whereηi ∈ (xi , xi−1 ) f or i = 1, 2, · · · n

By recognizing that fX (ηi )(xi − xi−1 ) is an approximation of the probability P (xi−1 < X ≤

xi ), we can rewrite this expression as

ηi fX (ηi )(xi − xi−1 ) ≈ ηi P (xi−1 < X ≤ xi ) = ηi (FX (xi ) − FX (xi−1 ))

In the limit, this approximation becomes exact. Since all distribution functions have the
property of being monotone and bounded on each sub-interval (xi−1 , xi ], our sum satisfies

the conditions for a Riemann-Stieltjes sum and we can write

n
X
E(X) = lim ηi (FX (xi ) − FX (xi−1 ))
max |xi −xi−1 |→0
i=1

We have thus reached a general definition of the expected value of a random variable.

Definition 4.1.3 For a random variable X with distribution function FX , we define the
expectation of X as
Z ∞
E(X) = xdFX (x)
−∞

Expressing the expectation of a random variable by its distribution function thus generates
a general formula, which is valid for both continuous and discrete random variables, as well
as mixtures of the two.

31
 R
Now, let g : R → R be a measurable function such that g(x)dF (x) converges. Let Y
be a random variable with distribution function FY = P (Y ≤ y) and expectation E(Y ) =

∞dFY (y). When Y is a function of another random variable X, we can introduce the following
theorems.

Theorem 4.1.1 Let X be a random variable and let letg : R → R be a measurable function
R
such that g(x)dF (x) < ∞. Then the random variable Y = g(X) has expectation
Z ∞
E(Y ) = g(x)dFX (x) (4.1)
−∞

Proof: The Integral on the RHS in equation (4.1) is a Riemann-Stieltjes Integral. Therefore

we assume that the size of (xi−1 , xi ] is small and locate a Riemann-Stieltjes sum for the LHS
in the expectation Y.
n
X
ηi (FX (yi ) − FX (yi−1 ))
i=1
n
X
= ηi P (Y ∈ (yi−1 , yi ])
i=1
Xn
= ηi P (g(X) ∈ (yi−1 , yi ])
i=1
n
X
= ηi P (X ∈ g −1 (yi−1 , yi ]) where ηi ∈ (yi−1 , yi ]
i=1
−1
We define a variable ξi = g (ηi ) and note that since g is bijective.

ηi ∈ (yi−1 , yi ] ⇔ ξi = g −1 (ηi ) ∈ g −1 ((yi−1 , yi ]) ⇔ g(ξi ) ∈ (yi−1 , yi ]

Thus for ξi ∈ g −1 ((yi−1 , yi ]) we get

n
X
ηi P (X ∈ g −1 ((yi−1 , yi ]).
i=1

If the (yi−1 , yi ] form a partition of an interval in R such that all subintervals are disjoint and
have the entire interval as their union, then each (xi−1 , xi ] = g −1 ((yi−1 , yi ]) will also form a

partition in R. Therefore,
n
X n
X
−1
g(ξi )P (X ∈ g ((yi−1 , yi ])) = g(ξi )P (X ∈ (xi−1 , xi ])
i=1 i=1

32
where ξi ∈ g −1 ((yi−1 , yi ]). The sum on the RHS above is a Riemann-Stieltjes sum for the
integral in equation (4.1.). That is,
Z ∞ n
X
g(x)dFX (x) = g(ξi )P (X ∈ (xi−1 , xi ])
−∞ i=1
Xn
= g(ξi )(FX (xi ) − FX (xi−1 ))
i=1

Theorem 4.1.2 If X and Y are two random variables and a and b are real-valued constants,

then the expectation of X and Y satisfies

E(aX + bY ) = aE(X) + bE(Y )

Proof: By definition,
Z
E(aX + bY ) = (ax + by)dF (x, y)

which is a Riemann-Stieltjes Integral, then this integral is equal to

Z Z
a xdF (x, y) + b ydF (x, y)

Using the definition of the expectation again leaves us with

aE(X) + bE(Y )

When we define the expectation of discrete and continuous random variables separately, as

sums and ordinary Riemann Integral respectively, it is not possible to define the expectation
of X + Y if, for example, X is discrete and Y is continuous. Thus, we have shown that this
can be done by using the Riemann-Stieltjes Integral, and this is why it provides such a good

definition of the expectation of a random variable.

The Riemann-Stieltjes Integral may be very useful when constructing proofs in probability
theory and mathematical statistics. Nevertheless, we must point out that the Riemann-

Stieltjes Integral inprobability theory is strictly a theoretical application in the sense that
one does not evaluate Riemann-Stieltjes Integral when calculating expected values.

33
Furthermore, there is a certain limitation to using Riemann-Stieltjes Integral for the expected
value since they cannot handle so called wild functions, like for example the function

 sin( 1 ) when x 6= 0
x
f (x) =
 0 when x = 0

4.2 Application to Functional Analysis

If g : [a, b] → R is of bounded variation, then there is an associated linear functional Ig on

c[a, b] given by
Z b
Ig (f ) ≡ f dg (4.2)
a

which is the Riemann-Stieltjes integral of f with respect to g. For all functions f ∈ c[a, b]and

g ∈ BV [a, b], f is Riemann-Stieltjes Integrable with respect to g, so Ig is well defined.

Theorem 4.2.1 (The Riesz Representation)

If Ig is a bounded linear functional on C[a, b], then there exists a function g ∈ BV [a, b] such
that L = Ig and kLkC[a.b] = Vab (g).

Proof: By the Hahn-Banach Theorem, the bounded linear functionalL on C[a, b] can be
extended to a bounded linear functional L̄ on B[a, b] with the same norm as L. Let g be

defined by

g(a) = 0, g(x) = L̄(χ[a,b] ) f ora < x < b, and g(b) = L̄(χ[a,b] ) (4.3)

The key property of the function g defined in equation (4.3), is that for any x, y ∈ [a, b] with

x<y

g(y) − g(x) = L̄(χ[x,y] ) if y < b and g(y) − g(x) = χ[x,b] if y = b (4.4)

34
whereL̄ is an extension of L to the space B[a, b] of bounded functions on [a, b]. If P : a =
x0 < · · · < xn = b is any strictly increasing partition, then
n
X n
X
|g(pi ) − g(pi−1 )| = ±(g(pi ) − g(pi−1 ))
i=1 i=1
n−1
X
= L̄(±χ[pn−1 ,b] + ±χ[pi−1 ,pi ] )
i=1
≤ kL̄kB[a,b] = kLkC[a,b]

where the inequality follows from the fact that the argument to L̄ is a function that always
take a value in −1, 1 and therefore has supremum norm equal to 1. This shows that g has a

bounded variation on [a, b] and that

Vab (g) ≤ kLkC[a,b] (4.5)


Given any f ∈ C[a, b] and  > 0, choose δ > 0 such that |x − y| < δ implies |f (x) − f (y)| < 2

for all x, y ∈ [a, b]. This is possible since f is uniformly continuous. Using the fact that f

is Riemann-Stieltjes Integrable with respect to g, choose any partition P of [a, b] with width
less than δ and any sequence x0 , · · · xn−1 satisfies P such that
n
X kLkC[a,b]
| f (xi−1 )(g(pi ) − g(pi−1 )) − Ig (f )| ≤
i=1
2

Define a bounded function s ∈ B[a, b] by

n
X
s = f (xn−1 )χ[pn−1 ,] + f (xi−1 )χ[pi−1 ,pi )
i=1

By equation (4.4) above,

n
X
L̄(s) = f (xi−1 )(g(pi ) − g(pi−1 ))
i=1

It is easy to see from the uniform continuity hypothesis on f that


kf − sk ≤ 2

That is, |f (y) − s(y)| < 2
for all y ∈ [a, b]

35
Thus,
n
X
|L(f ) − Ig (f )| ≤ |L(f ) − L̄(s)| + | f (xi−1 )(g(pi ) − g(pi−1 )) − Ig (f )|
i=1
kLkC[a,b]
≤ |L(f ) − L̄(s)| +
2
kLkC[a,b]
= |L̄(f − s)| +
2
kLkC[a,b]
≤ kf − sksupkLkC[a,b] +
2
≤ kLkC[a,b]

Since f and  were arbitrarily chosen, it follows that L = Ig .

This equation (4.5) implies that
kLkC[a,b] = Vab (g)

Riemann-Stietjes Integral appears to be a useful tool in proving the Riesz representation

theorem because it was necessary to prove that continuous functions on a closed interval are
Riemann-Stieltjes Integrable with respect to any function of bounded variation.

4.3 Application to Population Growth

Suppose the population growth rate at a time of a certain rural locality is given by

1
P (t) = 200 + 6t 2 , t→0

wheret represents years from 1960, (for example, it tell us us that at the first year, the rate
of growth is about 206 people per year). We now use a Riemann-Stieltjes sum to estimate

the population growth from 1960 to 1975 using five equal subintervals of time and using the
left endpoint of each subinterval to estimate the population growth rate for the subinterval.
The interval from 1960 to 1975 corresponds to: 0 ≤ t ≤ 15. Dividing this interval into five
15−0
equal subintervals, each has 5
= 3 width.
The endpoints of the subintervals are 0, 3, 6, 9, 12 and 15. The Riemann -Stieltjes sum to

36
estimate the population growth is as follows:
n
X
s = P (ti )(α(ti ) − α(ti−1 ))
i=1
Xn
s = P (ti )∆αi
i=1
X5
= P (ti ).3
i=1
X 5
= 3 P (ti ) = 3[P (0) + P (3) + P (6) + P (9) + P (12)]
i=1
√ √ √ √
= 3[200 + (200 + 6 3) + (200 + 6 6) + (200 + 6 9) + (200 + 6 12)

= 3192

4.4 Application to Mechanics

In Mechanics the Riemann-Stieltjes Integral is used in the following situation, consider n-

masses each of mass mi , i = 1, 2, · · · n, located along the x-axis at distances ri from the origin

with 0 < r1 < · · · rn . The moment of Inertia I, about an axis through the origin at right
angles to the system of masses is given by
I = ni=1 ri2 mi . This is the discrete case.
P

If instead we have a length of wire l along the x-axis with one end at the origin the moment
of inertia is
Z 1
I= x2 ρ(x)dx,
0

where for each x ∈ [0, l], ρ(x) equals the cross-sectional density at x.

37
Chapter 5

CONCLUSION

5.1 Conclusion

The definition and properties of the Riemann-Stieltjes Integral have been given, with an

explanation of how to calculate the integral. The Riemann-Stieltjes Integral is a useful

mathematical tool when working with discrete and continuous random variables simultane-
ously. It has application in functional analysis, mechanics and in determining accurately the

population growth of an area after a particular time interval.

38
 REFERENCES

[1] Anthony Narkawicz (2011): A Formal Proof of the Riesz Representation Theorem.
Journal of Formalized Reasoning Vol. 4 No. 1 2011.

[2] Apostol T.M. (1975): Mathematical Analysis. Addison Wesley, Reading, MA, 1975.

[3] Ayoola E.O. (2015): Riemann-Stieltjes Theory of Integration. Lecture Notes on

MTS424, 2014/2015 Session, Federal University of Agriculture, Abeokuta.

[4] Candeloro, D. (1996): Riemann-Stieltjes Integration in Riesz spaces. Rend Mat. (Roma).

Ser. VII 16 1996. 563 -585

[5] Colleen Cook McFadden (2011): Riemann-Stieltjes Integration. A Thesis submitted

in partial fulfillment of the requirements for the degree of Master of Mathematics at

Virginia Commonwealth University.

[6] Enderle J.D. (2006): Basis Probability Theory for Biomedical Engineers. Morgan and
Claypool Publishers, San Rafael, CA, 2006

[7] Gray J.D. (1984): The shaping of the Riesz Representation Theorem: A chapter in the
history of analysis. Archive for History of Exact Sciences, 31(2), 1984.

[8] Henstock, R. (1991): The General Theory of Integration, Clarendon Press, Oxford 1991

[9] Hildebrandt, T.H. (1938): Definitions of Stieltjes Integrals of the Riemann Type. The
American Monthly 45(5): 265-278, ISSN 0002-9890

[10] Kar A.F. (1993): Probability. Springer Science + Business Media, New York 1993.

39
[11] KlaraLeffler (2014): The Riemann-Stieltjes Integral and some applications in complex
analysis and probability theory.

[12] http://umu.diva-portal.org/smash/get/diva2:71948/FULLTEXT01.pdf

[13] Mikosch T. (1998): Elementary Stochastic Calculus with Finance in View; World Sci-
entific, Singare, 1998

[14] OK E.A. (2014): Probability theory with economic applications. Lecture Notes.

[15] Robert G. Bartle (2001): A modern theory of integration. American Mathematical

Society. http://www.math.vander.bilt.edu/schectex/ccc/guage

[16] Rudin W. (1976): Principles of Mathematical Analysis. Third Edition. International

Series in Pure and Applied Mathematics. McGraw-Hill Book Co; New York-Auckland-

Dusseldorf, 1976

[17] Simmons G.F. (1963): Introduction to Topology and Modern Analysis. McGraw-Hill
Book Co., 1963

[18] Stroock, Daniel W. (1998): A Concise Introduction to the Theory of Integration (3rd
ed.), Birkhauser, ISBN 0-8176-4073-8

[19] Thomas Hong (1996): The Riemann Integral in Calculus: Students Processes and
Concepts.www.merga.net.au/documents/RPT homasH ong1 996p df

[20] Wheeden R.L., A. Zygmund (1977): Measure and Integral: An introduction to Real

Analysis. Marcel Dekker, Inc., New York 1977

[21] https : //en.m.wikipedia.org/wiki/Riemann − StieltjesI ntegral

[22] http://www.math.unl.edu/ joor1/math826/RiemannStieltjes.pdf

[23] http://www.math.ucdavis.edu./nemsilvia/math127/chapter7.pdf

[24] http://www.maths./th.se/matstat/kurser/mas213/riemannstieltjes.pdf

40
[25] http://ocw.nctu.edu.tw/upload/classbfs

41