REAL ANALYSIS AND MEASURE

THEORY

BY

DR. PRATULANANDA DAS

DEPARTMENT OF MATHEMATICS
JADAVPUR UNIVERSITY
KOLKATA- 700032
WEST BENGAL, INDIA
E-mail : [email protected]
 Chapter 10

Module 1

Riemann-Stieltjes Integral

1
 10.1. Introduction

In the last chapter of this course we focus on a generalization of Riemann

integral which is known as Darbaux-Stieltjes or Riemann-Stieltjes integral. The
generalization is essentially based on changing the parameter ∆(xk ) i.e. xk −
xk−1 in the Riemann or Darbaux upper or lower sums of the integrand f corre-
sponding to a given partition P = {a = x0 < x1 < x2 < · · · < xn = b} of [a, b]
and replacing it with α(xk ) − α(xk−1 ) for some function α defined on [a, b]. It is
easy to note that the definitions of Riemann or Darbaux integrals appear when
we take α as the identity function on [a, b] i.e. when α(x) = x ∀x ∈ [a, b].

We will observe that when the function α has a continuous derivative, the
generalized definition is such that the Riemann-Stieltjes integral of the function
f with respect to the function α actually becomes the Riemann integral of the
function f (x)α0 (x) over [a, b]. But one should remember that for the basic
development of this integral we do not even need α to be continuous. Therefore
this approach is much more general and actually its importance lies in the fact
that this integral has found important applications in other areas like Physics
and Probability Theory. A simple example is when we consider the problem
of finding the moment with respect to the y-axis of a distribution of mass over
[a, x]. If m(x) is the amount of mass over [a, x], then it is known that the
moment with respect to the y-axis is described by the following sum
n
X
xk (m(xk ) − m(xk−1 ))
k=1

where m(xk ) − m(xk−1 ) stands for the mass between xk−1 and xk . In particular
this integral is more used to deal with situations where the function α is appro-
priately chosen to express a finite or infinite sum as a Riemann-Stieltjes integral.

Now coming to the basic development of the theory of this integral, it is

almost similar to the development of Riemann integral and many proofs are
just minor modifications of proofs of known results for Riemann integration.
We give the formal definitions and then establish mainly those properties whose
treatment are somewhat different.

10.2. Riemann-Stieltjes Integral

Let P = {a = x0 < x1 < · · · < xn = b} be a partition of [a, b]. As usual,

the norm (or mesh) of the partition P is defined to be the length of the largest
subintervals of P and is denoted by ||P ||. It is easy to understand that if a
partition P 0 of [a, b] is finer than P (or is a refinement of P ) i.e. P ⊆ P 0 then
||P 0 || ≤ ||P ||. Therefore refinement of a partition decreases its norm, but the
converse does not necessarily hold which can be easily verified.

We assume that, unless otherwise stated, all functions denoted by f, g, α, β

2
etc. are real valued bounded functions defined on [a, b].

The symbol ∆αk denotes the difference ∆αk = α(xk ) − α(xk−1 ), so that
n
X
∆αk = α(b) − α(a).
k=1

The set of all possible partitions of [a, b] is denoted by π[a, b].

We now introduce the definition of Riemann-Stieltjes integral as follows

which is quite similar to the definition of Riemann integral.

Definition 10.1. Let P = {a = x0 < x1 < · · · < xn = b} be a partition of [a, b]

and let tk be a point in the subinterval [xk−1 , xk ]. A sum of the form
n
X
S(P, f, α) = f (tk )∆αk
k=1

is called a Riemann-Stieltjes sum of f with respect to the function α. We say

f is Riemann integrable with respect to α on [a, b] if there exists a number A
such that for every ε > 0 there exists a partition Pε of [a, b] such that for every
partition P finer than Pε and for every choice tk in [xk−1 , xk ] we have

|S(P, f, α) − A| < ε.

In this case we write f ∈ R(α) on [a, b].

When such a number A exists, it is uniquely determined and is denoted by

Z b
f dα. f is called the integrant and α is referred as integrator respectively. If
a
α(x) = x ∀x ∈ [a, b] then we say that f ∈ R which means that f is Riemann
integrable.

The proofs of following theorems are easy modifications of the corresponding

proofs for Riemann integration and so are left to the reader.

Theorem 10.2. (Linearity property) If f, g ∈ R(α) on [a, b] then for all c, d ∈ R,

cf + dg ∈ R(α) on [a, b] and
Z b Z b Z b
(cf + dg)dα = c f dα + d gdα.
a a a

From the above result it follows that R(α) forms a vector space over R.

3
Theorem 10.3. If f ∈ R(α) and f ∈ R(β) on [a, b] then f ∈ R(c1 α + c2 β) on
[a, b] for all c1 , c2 ∈ R and
Z b Z b Z b
f d(c1 α + c2 β) = c1 f dα + c2 gdβ.
a a a

Z b Z c
Theorem 10.4. If c ∈ (a, b) and if any two of the integrals f dα, f dα
Z b a a

and f dα exist then the third integral also exists and we have
c
Z c Z b Z b
f dα + f dα = f dα.
a c a

Z a Z b Z b
Definition 10.5. If a < b, we define f dα = − f dα, provided f dα
Z a b a a

exists. Further we define f dα = 0.

a

Theorem 10.6. Let f ∈ R(α) on [a, b] and α has a continuous derivative α0 on

Z b
[a, b]. Then the Riemann integral f (x)α0 (x)dx exists and
a
Z b Z b
f (x)dα(x) = f (x)α0 (x)dx.
a a

Proof: For a given partition P = {a = x0 < x1 < · · · < xn = b} of [a, b] and a

point tk in the subinterval [xk−1 , xk ] for k = 1, 2, · · · n we consider the Riemann
sum
Xn
S(P, f α0 ) = f (tk )α0 (tk )∆xk .
k=1

Now for the same partition P and points tk we consider the Riemann-Stieltjes
sum
n
X
S(P, f, α) = f (tk )∆αk .
k=1

Now applying Mean Value Theorem on the function α in the interval [xk−1 , xk ]
we get
∆αk = α0 (ck )∆xk

4
for some ck ∈ (xk−1 , xk ). Now
n
X n
X
S(P, f, α) − S(P, f α0 ) = f (tk )∆αk − f (tk )α0 (tk )∆xk
k=1 k=1
Xn
= f (tk ) [α0 (ck ) − α0 (tk )] ∆xk .
k=1

Since f is bounded so there exists M > 0 such that |f (x)| ≤ M . Again note
that since α0 is continuous on [a, b], so it is also uniformly continuous on [a, b]
and consequently for every ε > 0 there exists a δ > 0 such that
ε
|α0 (x) − α0 (y)| <
2M (b − a)

whenever x, y ∈ [a, b] and |x − y| < δ.

Now choose a partition Pε0 with norm ||Pε0 || < δ. Then for any refinement P
of Pε0 we have
ε
|α0 (ck ) − α0 (tk )| < .
2M (b − a)
Therefore
ε
|S(P, f, α) − S(P, f α0 )| <
.
2
Again since f ∈ R(α) on [a, b], there exists a partition Pε00 such that for any
partition P finer than Pε00
Z b
ε
S(P, f, α) − f dα < .
a 2

Subsequently for any partition P which is finer than Pε = Pε0 ∪ Pε00 , we obtain
Z b
S(P, f α0 ) − f dα < ε.
a

This completes the proof of the theorem.

Theorem 10.7. Let a < c < b. Define

α(x) = α(a) if a ≤ x < c

= α(b) if c ≤ x < b.

Assume that at least one of the functions f and α is continuous from the left
at c and at least one is continuous from right at c. Then f ∈ R(α) on [a, b] and
Z b
f dα = f (c) [α(c+) − α(c−)] .
a

5
Proof: Let P = {a = x0 < x1 < · · · < xn = b} be a given partition of [a, b] and
a point tk is chosen from the subinterval [xk−1 , xk ] for k = 1, 2, · · · n. If c ∈ P ,
say c = xk , then we can write

S(P, f, α) = f (tk−1 ) [α(c) − α(c−)] + f (tk ) [α(c+) − α(c)] .

Set Ω = S(P, f, α) − f (c) [α(c+) − α(c−)]. Then

Ω = f (tk−1 ) [α(c) − α(c−)] + f (tk ) [α(c+) − α(c)] − f (c) [α(c+) − α(c−)]

= f (tk−1 ) [α(c) − α(c−)] + f (tk ) [α(c+) − α(c)]
−f (c) [α(c+) − α(c) + α(c) − α(c−)]
= [f (tk−1 ) − f (c)] [α(c) − α(c−)] + [f (tk ) − f (c)] [α(c+) − α(c)] .

Hence

|Ω| ≤ |f (tk−1 ) − f (c)| |α(c) − α(c−)| + |f (tk ) − f (c)| |α(c+) − α(c)| .

Let ε > 0 be given.

Case I: First let f be continuous at c. Then there exists a δ > 0 such that
||P || < δ implies |f (tk−1 ) − f (c)| < ε as well as |f (tk ) − f (c)| < ε. In that case
we have
|Ω| ≤ ε |α(c) − α(c−)| + ε |α(c+) − α(c)| .
Case II: If f is discontinuous both from left and right at c then from the given
conditions we must have α(c) = α(c−) and α(c) = α(c+) which implies that
Ω = 0.

Case III: Now suppose that f is discontinuous from left at c but is continuous
from right at c. Then from the given condition α(c) = α(c−). Hence

|Ω| ≤ |f (tk ) − f (c)| |α(c+) − α(c)| < ε |α(c+) − α(c)|

provided ||P || < δ.

Case IV: Finally suppose that f is discontinuous from right at c but is continuous
from left at c so that we have α(c) = α(c+). Therefore

|Ω| ≤ |f (tk−1 ) − f (c)| |α(c) − α(c−)| < ε |α(c) − α(c−)|

provided ||P || < δ.

Thus combining all the above cases we observe that if P is a partition of

[a, b] finer than Pδ where ||Pδ || < δ then we also have that ||P || ≤ ||Pδ || < δ and
consequently
|S(P, f, α) − f (c) [α(c+) − α(c−)]|

6
 = |Ω| ≤ ε |α(c) − α(c−)| + ε |α(c+) − α(c)| .
Hence we have f ∈ R(α) on [a, b] and
Z b
f dα = f (c) [α(c+) − α(c−)] .
a