Chapter 7

Functions of a Real Variable

The purpose of this chapter is to consider the relationship between Lebesgue

integration and differentiation for functions of a real variable. Thus this
chapter is concerned with the adaptation to the Lebesgue Integral of what is
called the Fundamental Theorem of Calculus for a single real variable.

7.1 Functions of Bounded Variation

Definition 7.1.1. A function f : I → R is said to be of bounded variation
on an interval I, written as f ∈ BV(I), provided that there exists a real
number M > 0 such that the variation
n
X
v(∆) = |f (ti ) − f (ti−1 )| ≤ M
i=1

for every partition

∆ = {t0 ≤ t1 ≤ . . . ≤ tn } ⊂ I

of I by finitely many points t0 , . . . , tn .

Exercise 7.1.1. † Prove that BV(I) is a vector space. That is, prove that
if both f and g are in BV(I) then the same is true for af + g for each a ∈ R.

Theorem 7.1.1. A function f ∈ BV[a, b] if and only if f can be represented

as f = f1 − f2 where f1 and f2 are both monotonically increasing functions.

103
Proof. Sufficiency is easy to establish. If f is any monotone function on [a, b],
whether it is increasing or decreasing, then v(∆) ≤ M = |f (b) − f (a)| for all
∆. Thus f1 and f2 have bounded variation, and we apply Exercise 7.1.1.
We turn next to the proof of necessity. So we suppose that f ∈ BV[a, b].
Define the positive variation and the negative variation by
n
X
p(∆) = (f (ti ) − f (ti−1 ))+
1
n
X
n(∆) = (f (ti ) − f (ti−1 ))−
1

where the superscript + means the positive part, and serves only to replace
the value of the parenthetic expression by zero if it happens to be negative.
The superscript − means the negative part, and it serves to replace the value
of the parenthetic expression by zero if it happens to be positive and by its
absolute value if it happens to be negative.
If x ∈ [a, b], then f is still of bounded variation on the subinterval [a, x].
Let ∆ denote an arbitrary partition of [a, x] and define

p(x) = sup p(∆)

∆
n(x) = sup n(∆)
∆
v(x) = sup v(∆)
∆

each of which is bounded above by M, which is defined in Definition 7.1.1.

It is easy to see that
(i)
v(x) = sup (p(∆) + n(∆)) ≤ sup p(∆) + sup n(∆) = p(x) + n(x)
∆ ∆ ∆

from the definitions. But we need to prove equality in (i). To prove this,
suppose that ǫ > 0. Then
ǫ ǫ
p(x) − p(∆1 ) < and n(x) − n(∆2 ) <
2 2
for some suitable partitions ∆1 and ∆2 . Let ∆ = ∆1 ∪ ∆2 . We claim that

p(∆) ≥ p(∆1 ) and n(∆) ≥ n(∆2 ). (7.1)

104
This can be done by demonstrating that partitioning an interval [xi−1 , xi ] by
means of a point x of that interval forces
(f (xi ) − f (x))+ + (f (x) − f (xi−1 ))+ ≥ ((f (xi ) − f (xi−1 )))+
and
(f (xi ) − f (x))− + (f (x) − f (xi−1 ))− ≥ ((f (xi ) − f (xi−1 )))− .
It follows from Equation 7.1 that
ǫ ǫ
v(x) ≥ v(∆) = p(∆) + n(∆) > p(x) − + n(x) − = p(x) + n(x) − ǫ
2 2
for each ǫ > 0. This establishes that
v(x) = p(x) + n(x). (7.2)
It is immediate also that the three functions p, n, and v are all monotone
increasing, since any partition ∆ of [a, x] can be extended to a partition
∆ ∪ {x′ } of [a, x′ ] for x′ > x. By reasoning as we did in the proof of Equation
7.1, we can show that p(x) − n(x) = f (x) − f (a), since this is true for every
∆ that is a partition of [a, x]. Hence
f (x) = (f (a) + p(x)) − n(x)
which is a difference of two monotone increasing functions. Equivalently, this
can be written as
f (x) − f (a) = p(x) − n(x).

Remark 7.1.1. We note that the representation f (x)−f (a) = p(x)−n(x) in

Theorem 7.1.1 is not unique. In fact, we could take any monotone increasing
function h(x) and write f (x) − f (a) = [p(x) + h(x)] − [n(x) + h(x)].
Exercise 7.1.2. Suppose for f as in Theorem 7.1.1 we had monotone in-
creasing functions f1 and f2 such that
f (x) − f (a) = p(x) − n(x) = f1 (x) − f2 (x).
Show that f1 (x) − p(x) = f2 (x) − n(x) and prove that
p(x) + n(x) ≤ (f1 (x) − f1 (a)) + (f2 (x) − f2 (a)).
This establishes that the decomposition in the proof of Theorem 7.1.1 is
minimal in the sense that f1 and f2 must increase faster than p and n do,
and that the difference cancels out as in Remark 7.1.1.

105
Exercise 7.1.3. If f is continuous on an interval [a, b] and has a bounded
derivative in (a, b), show that f is of bounded variation on [a, b]. Is the
boundedness of f ′ necessary for f to be of bounded variation? Justify your
answer.

1.0

0.5

x
0.2 0.4 0.6 0.8 1.0

-0.5

-1.0

π


Figure 7.1: f (x) = x sin x
, with envelope u(x) = x, l(x) = −x

Exercise 7.1.4. † Let

(

π
xn sin x
if x 6= 0
fn (x) =
0 if x = 0.
We claim that fn ∈ BV[0, 1] if n ≥ 2 but it is not in this space if n = 1. See
Figures 7.1 and 7.2.

Exercise 7.1.5. If both f and g are in BV[a, b], prove that f g ∈ BV[a, b].
(Caution: the product of two monotone functions need not be monotone.)
Remark 7.1.2. Let f : [a, b] → R be Lebesgue integrable, which is equiva-
lent to f + and f − being Lebesgue integrable. The indefinite integral of f is
given by
Z x Z x Z x
F (x) = f (t) dl(t) = +
f (t) dl(t) − f − (t) dl(t)
a a a

106
 y

1.0

0.5

x
0.2 0.4 0.6 0.8 1.0

-0.5

-1.0

π


Figure 7.2: f (x) = x2 sin x
, with envelope u(x) = x2 , l(x) = −x2

which is a difference of two monotone increasing functions, so that F lies in

BV[a, b]. We can see that

Z x
v(x) = |f (t)| dl(t) = p(x) + n(x)
a

Rx Rx
where p(x) = a
f + (t) dl(t) and n(x) = a
f − (t) dl(t).
Thus, although f ∈ L1 (R, L, l) need not be Riemann integrable, the in-
definite integral of f has bounded variation, and is therefore Riemann inte-
grable, being a difference of two monotone functions. These observations are
significant for the theory of functions of a real variable, and they lead us to
the Fundamental Theorem of Calculus for the Lebesgue integral in the next
section. First, in preparation, the reader should solve the following exercise.

Exercise 7.1.6. Give an example of a Lebesgue integrable function f on

[0, 1] which there is no function F (x) with the property that F ′ (x) = f (x)
for all x ∈ [0, 1].

107