Report for MA502 Presentation:

Lebesgue’s Criterion for Riemann Integrability

Abraham Puthuvana Vinod
March 26, 2015

1 Introduction
We covered Riemann integrals in the first three weeks in MA502 this semester (Chapter 11 in [1]).
This report explores a necessary and sufficient condition for determining Riemann integrability of
f (x) solely from its properties. This condition is known as Lebesgue’s criterion and elucidating the
proof of this condition is the aim of this report.
This report revisits the concepts learnt in MA501/502 in Section 2. Section 2 also covers an exam-
ple showing that functions discontinuous everywhere need not be Riemann integrable. Towards the
end, this section reminds the readers a few results from measure theory useful in proving Lebesgue’s
criterion for Riemann integrability. Section 3 states the Lebesgue’s criterion and provides examples
of functions with countably infinite and uncountably infinite discontinuities which are Riemann in-
tegrable to motivate the usefulness of this criterion. Finally, Section 4 provides the proof for the
Lebesgue’s criterion for Riemann integrability. The references used in this report are attached in the
end.

2 Revisiting what we learnt in MA501/502

2.1 Known conditions on f for Riemann integrability
We define a partition using Definition 11.1.10 in [1] as follows:
Definition 2.1. Let I be a bounded interval. A partition of I is a finite set P of bounded intervals
contained in I, such that every x in I lies in exactly one of the bounded intervals J in P .
Partitions are allowed to contain empty sets. Using Definition 11.3.2, Lemma 11.3.3, Definition
11.3.4 and Proposition 11.3.12 in [1], we define Riemann integrability of a function f (x) as follows
Definition 2.2. Let f : I → R be a bounded function on a bounded interval I. Then, f is Rie-
P if and only if given  > 0, ∃ P
mann integrable P such that 0 ≤ U (f, P ) − L(f, P ) <  where
U (f, P ) = (sup f (x))|J| and L(f, P ) = (inf f (x))|J|.
J∈P :J6=φ x∈J J∈P :J6=φ x∈J

Note that while Definition 2.2 is a necessary and sufficient condition for Riemann integrability, it
requires some sort of computation on f (x) in order to determine if f is Riemann integrable.
Sections 11.5 − 6 in [1] list a few theorems for determining the Riemann integrability of a function
without any computation and just its properties. From Theorem 11.5.1 in [1], we have
Theorem 2.1. Let I be a bounded interval, and let f be a function which is uniformly continuous
on I. Then, f is Riemann integrable.
Proof. Given as proof of Theorem 11.5.1 in pages 284−286 of [1]. It should be noted from Proposition
9.9.15, a uniformly continuous function in a bounded interval is bounded.

1
 From Proposition 11.5.3 in [1], we have

Theorem 2.2. Let I be a bounded interval, and let f : I → R be both continuous and bounded.
Then, f is Riemann integrable.

Proof. Given as proof of Proposition 11.5.3 in pages 286 − 287 of [1].

We use the definition of piecewise continuous functions from Definition 11.5.4 in [1] to give
Definition 2.3 and use Proposition 11.5.6 in [1] to give Theorem 2.3. f |J is defined as the restriction
of a function f to a subset J of its domain.

Definition 2.3. A function f is piecewise continuous on I iff ∃ P of I such that f |J is continuous

on J ∀ J ∈ P .

Theorem 2.3. Let I be a bounded interval, and let f : I → R be both piecewise continuous and
bounded. Then, f is Riemann integrable.

Proof. Consider the function f : [a, b] → R defined in Figure 1 with two discontinuities at points
x = c and x = d. We will illustrate with this example how to proceed with the proof of Riemann
integrability of a bounded piecewise continuous function with 2 discontinuities. The proof of Riemann
integrability of bounded piecewise continuous functions having N discontinuities follows from simple
modifications to the arguments made in this proof.

Figure 1: A piecewise function with two discontinuities

From Definition 2.3, if we consider a partition P cts = {[a, c), [c, d], (d, b]}, we know that the
function f is continuous in every interval of P cts . Let us choose an 1 > 0 small enough to define a
partition P = {[a, c − 1 ], (c − 1 , c + 1 ), [c + 1 , d − 1 ], (d − 1 , d + 1 ), [d + 1 , b]}. Given  > 0. From
Theorem 2.2, we know that ∃ P [a,c−1 ] such that

U (f |[a,c−1 ] , P [a,c−1 ] ) − L(f |[a,c−1 ] , P [a,c−1 ] ) < .
5

2
Similarly, ∃ P [c+1 ,d−1 ] and P [d+1 ,b] satisfying the same requirements for intervals [c + 1 , d − 1 ] and
[d + 1 , b] respectively. We define

Q = P [a,c−1 ] ∪ {(c − 1 , c + 1 )} ∪ P [c+1 ,d−1 ] ∪ {(d − 1 , d + 1 )} ∪ P [d+1 ,b]

such that
  
U (f, Q) − L(f, Q) ≤ + 2B1 + + 2B1 + (1)
5 5 5


≤ Choosing 1 such that 0 ≤ 1 ≤ 5×2B
≤

From Definition 2.2, we conclude that f is a Riemann integrable. The trick used here can be
extended to finitely many discontinuities. Specifically, if there are N discontinuities, replace 5 by

2N +1
in (1).
Remark 2.1. Theorem 2.3 comments only about functions with finite number of discontinuities.
This is because Definition 2.3 uses the notion of partitions which are finite sets.
There are two important observations to note from Theorems 2.1-2.3:
1. All of these theorems have two common conditions:

(a) the candidate function is bounded, and

(b) the interval over which the Riemann integration is to be performed is bounded.

2. All of these theorems give necessary conditions expected from the candidate function to be
Riemann integrable.

3. There seems to be a trend of decreasing strength expected in the continuity of the candidate
function.
The last observation prompts us to ask the question covered in the Subsection 2.3: Can a function
discontinuous everywhere be Riemann integrable? Before answering that question, we revisit some
properties of Q and R \ Q in Subsection 2.2.

2.2 Some properties of Q and R \ Q

Lemma 2.1. The set of rationals, Q, is countable.
Proof. If x ∈ Q, then x = pq where p ∈ Z and q ∈ N \ {0}. Hence, we can enumerate the rational
numbers by listing down all possible combinations of p, q and then sweeping them across diagonals
as shown in Figure 2. This process of sweeping creates a bijection between rationals and natural
numbers, making the set of rationals Q countable.
Lemma 2.2. The set of rationals, Q, is dense. In other words, given a, b ∈ R, a < b, ∃ x ∈ Q such
that a < x < b.
Proof. Assume for contradiction, there is no x. By Archimedian principle, we can find N ∈ N such
1
that b−a < N ⇒ N1 < (b − a). Let A = { N m
: m ∈ N}. Then, for the assumption to hold, if m1 is the
greatest integer such that N < a, then N > b. This means, mN
m1 m1 +1 1 +1
− mN1 > b − a from an extension1
of Proposition 5.4.7(d) in [1]. Hence, N1 > b − a, which is a contradiction.
Lemma 2.3. The set of irrationals, R\Q, is dense. In other words, given a, b ∈ R, a < b, ∃ x ∈ R\Q
such that a < x < b.
1
If a, b, c, d ∈ R and a < b, c < d, then by Proposition 5.4.7(d) in [1], we have a + c < b + c and b + c < b + d. Hence,
a + c < b + d.

3
 Figure 2: Enumeration process used in proving Q is a countable set. (Source: [2])

Proof. Given a < b, consider the reals √a2 and √b .

2
We know √a2 < √b2 . From Lemma 2.2, we know
there is a rational number q such that √a2 < q < √b .
2
In order to get a non-zero q, we apply Lemma 2.2
to √a2 and 0 when a < 0 < b.
√ √ q0
By Proposition 5.4.7(e) in [1], we have a < 2q < b. If 2q = q 0 ∈ Q, then q
is rational2 but
q0
√ √
q
= 2 which is irrational. Hence, 2q = q 0 ∈ R \ Q.

Proofs for the density of Q and R \ Q were inspired from [3].

2.3 Can a function discontinuous everywhere be Riemann integrable?

We demonstrate this by giving a counter-example, the Dirichlet function which is a real-valued
( func-
1 x∈Q
tion which is discontinuous everywhere in R. The Dirichlet function is defined as g(x) = .
0 x∈R\Q
Dirichlet function
R 1 is not Riemann integrable in any subset of R.
Consider 0 g(x)dx. Owing to Lemma 2.2 and Lemma 2.3, we know that, for any partition
P , we have sup g(x) = 1 and inf g(x) = 0. Therefore, L(g, P ) = 0 and U (g, P ) = 1 for any
x∈[0,1] x∈[0,1]
partition P . Hence, g(x) is not Riemann integrable from Definition 2.2. Hence, functions which are
discontinuous everywhere are not necessarily Riemann integrable.

2.4 Set of measure zero

Denoting the length of an interval I as |I|, we have the definition of a set of measure zero from [4],
as follows:

Definition 2.4. A set S is said to have measure zero if given  > 0, ∃ a countable collection of open
intervals {Ik }∞
k=1 such that
[∞ ∞
X
S⊆ Ik and |Ik | < .
k=1 k=1

Lemma 2.4. Any countable set has its measure equal to zero. In particular, the empty set has
measure zero.

Lemma 2.5. Cantor set has uncountably infinite points and has measure zero.

The reader is requested to use Trevor, Angel and Michael’s report on set of measure zero for the
proofs of Lemma 2.4 and Lemma 2.5.
2
Division of rationals give rationals.

4
2.5 Some results from measure theory
We borrow a few results from measure theory which can intuitively understood by appealing to the
notion of measure as “length”.
Remark 2.2. Riemann integral of a function, when it exists, equals the Lebesgue integral of the
function. In other words, Riemann integrable functions are Lebesgue R integrable. The integral of a
characteristic function of an interval X, 1X (x), is its length given by 1X (x)dx = m(X) where m(A)
denotes the Lebesgue measure of the set A.
∞
Remark 2.3. If {Ai }∞
S
i=0 is a sequence of disjoint sets which are Lebesgue-measurable and A = Ai ,
i=0
∞
P
then m(A) = m(Ai ). This result can be used to conclude that for some N ∈ N and disjoint sets
i=0
N
S N
P
Ai , m( Ai ) = m(Ai ).
i=0 i=0

Lemma 2.6. Set of irrationals, R \ Q, form a measurable set of non-zero measure.

Proof. From Lemma 2.1 and Lemma 2.4, we know that the set of rationals in R, Q, is of measure zero.
Since R = Q ∪ (R \ Q) and Q ∩ (R \ Q) = φ, we have, from Remark 2.3, m(R) = m(Q) + m(R \ Q) ⇒
m(R \ Q) = m(R) > 0.
∞ ∞
Remark 2.4. If {Ai }∞
S P
i=0 is a sequence of measurable sets, then 0 ≤ m( Ai ) ≤ m(Ai ). Hence,
i=0 i=0
∞
S ∞
S
if m(Ai ) = 0 ∀ i, then m( Ai ) = 0 and if m( Ai ) > 0, then ∃ i such that m(Ai ) > 0.
i=0 i=0

Remark 2.5. If A ⊆ B, then m(A) ≤ m(B).

3 Lebesgue’s criterion for Riemann integrability and some

motivating examples
From [4], we state Lebesgue’s criterion for Riemann integrability, which will be proven in Section 4,
as follows:
Theorem 3.1. If f (x) is a bounded function defined on [a, b], then f is Riemann integrable iff the
set of points on which f is discontinuous, say S, is a set of measure zero.
From Subsection 2.3, we remarked that g(x) defined as the Dirichlet function is discontinuous
everywhere. Specifically, g(x) is discontinuous on R \ Q. Therefore, from Lemma 2.6, g(x) is discon-
tinuous on a set of measure greater than zero. Hence, g(x) is not Riemann integrable according to
Lebesgue’s criterion of Riemann integrability. This matches with the observations made in Subsec-
tion 2.3.
We now look at two famous examples of Riemann integrable functions having infinite discontinu-
ities. By Definition 2.3, they are not piecewise continuous and hence Theorem 2.3 can not be used
to determine their Riemann integrability. From Theorem 3.1, we can ensure that these examples are
Riemann integrable if their corresponding sets of points of discontinuities are of measure zero.

3.1 Countably infinite discontinuities example: Thomae’s function

Thomae’s function is defined as follows:

 1 x=0

t(x) = 1
q
x ∈ Q ⇒ x = pq , g. c. d(p, q) = 1 .

0 x∈R\Q


5
 We list some important properties of Thomae’s function as lemmas. The proof for these lemmas
were inspired from [5].
Lemma 3.1. t(x) is discontinuous in Q.
Proof. Let r = pq ∈ Q such that g.c.d(p, q) = 1 implying f (r) = 1q > 0. Consider a sequence
xk = r + k√1 2 . Clearly, xk is a sequence in R \ Q and hence, f (xk ) = 0. Even though, xk → r as
k → ∞, we observe that f (xk ) 6→ f (r) 6= 0 as k → ∞. Hence, the function is not continuous in Q
by Proposition 9.4.7 in [1].
Lemma 3.2. t(x) is continuous in R \ Q.
Proof. Let x0 be an irrational number and hence, f (x0 ) = 0. By definition, the function is periodic
with a time period 1. Hence, without loss of generality, choose x0 ∈ [0, 1] \ Q. Given  > 0, we need
to find δ > 0 such that sup f (x) ≤ .
x∈B(x0 ,δ)
Extending interspersing of reals with integers (Exercise 5.4.3 in [1]), if we choose m ∈ N, then
k k+1 k
− x0 |, | k+1


∃ k ∈ Z, x0 ∈ m , m . We define δm = min{| m m
− x0 |} > 0 since x0 6∈ Q. Similarly,
1 1
∃ N ∈ N such that N ≤  < N −1 . If x ∈ [0, 1] is such that f (x) > , then x ∈ Q since irrationals
are mapped to 0 by f . Also, we can say that x = pq is such that 1q >  ≥ N1 ⇒ q < N . Hence, the
number of possible values x can take is only finitely many and is bounded by N 2 (x = pq ≤ 1). So,
we can choose δ = min{δ1 , δ2 , . . . , δm } > 0 since the set of δm is finite, which completes the proof.

π
Figure 3: Thomae’s function (only plotted for prime numbers q ≤ 37) with x0 = 10
and N = 10
π
Figure 3 demonstrates the choice of δ when x0 = 10 ∈ [0, 1] \ Q and  = 0.1 ⇒ N = 10. We find
that δ = 0.012 is sufficient to ensure that ∀ x ∈ B(x0 , δ), 0 ≤ f (x) < .
From Lemma 3.1 and Lemma 3.2, t(x) is known to be discontinuous only in Q. From Lemma 2.1,
Lemma 2.4 and the fact that t(x) is bounded in [0, 1], we can conclude that t(x) is Riemann integrable
in [0, 1] using Theorem 3.1. For any partition P of [0, 1], we have L(t, P ) = 0 since inf t(x) = 0.
x∈[0,1]
R1
Hence, from Definition 2.2, 0 t(x)dx = 0.

6
3.2 Uncountably infinite discontinuities example: 1C (x)
From Lemma 2.5, we know that 1C (x) has uncountably infinite discontinuities and the set of dis-
continuities is the Cantor set. Since 1C (x) is a bounded function in [0, 1], 1C (x) is Riemann inte-
grable
R1 in [0, 1] according to Lebesgue’s criterion (Theorem 3.1). From Remark 2.2 and Lemma 2.5,
1 (x)dx = m(C) = 0.
0 C

4 Proof of Lebesgue’s Criterion

We restate Theorem 3.1: If f (x) is a bounded function defined on [a, b], then f is Riemann integrable
iff the set of points on which f is discontinuous, say S, is a set of measure zero.
We first define Dα as a set containing all discontinuity points who have a jump greater than α
as: ( ! )
Dα = x ∈ [a, b] : sup f (x) − inf f (x) > α, ∀ r > 0
B(x,r) B(x,r)

For the example function discussed in the proof of Theorem 2.3, let us say Figure 4 describes the
jumps of f at c, d ∈ [a, b], then c ∈ Dα but d 6∈ Dα .

Figure 4: Figure illustrating the definition of Dα

We split the proof of Theorem 3.1 into the proof for necessary and sufficient conditions for
Riemann integrability. These proofs were inspired from [4] and [6].

4.1 Some useful lemmas for the proof of Theorem 3.1

Consider the following lemmas:

Lemma 4.1. sup f (x) − inf f (x) = sup |f (x) − f (y)| where E is a subset of the domain of f .
x∈E x∈E x,y∈E

7
Proof. For x, y ∈ E, f (x) − f (y) ≤ |f (x) − f(y)|. From definitions
 of sup and inf, we can
say that sup |f (x) − f (y)| is equivalent to sup sup (f (x) − f (y)) = sup f (x) − inf f (x) since
x,y∈E x∈E y∈E x∈E x∈E
sup(−E) = − inf(E).

Lemma 4.2. Dα is a closed and compact set.

Proof. Let x0 be a limit point of Dα . By definition of a limit point (Definition 6.4.1 in [1]), ∃ a
sequence {xi }∞
i=1 in Dα which has its limit as x0 . Hence, for any r > 0, we have all but finitely
! many

points of the sequence in B(x0 , r). But this means ∀ r > 0, sup f (x) − inf f (x) > α and
B(x0 ,r) B(x0 ,r)

hence, x0 ∈ Dα . Since Dα contains all its limit points, Dα is closed by Definition 9.1.15 in [1].
Since Dα is a subset of a bounded interval, I, it is bounded from Definition 9.1.22 in [1]. From
Heine-Borel theorem, we conclude that since Dα ⊆ R is closed and bounded in R, it is compact in
R.

Lemma 4.3. f is continuous at x0 if and only if given  > 0, ∃ r > 0 such that
sup f (x) − inf f (x) < .
B(x0 ,r) B(x0 ,r)

Proof. The necessary condition for the continuity is proved as follows:

The hypothesis is that given  > 0, ∃ r > 0 such that sup f (x) − inf f (x) < . We know that
B(x0 ,r) B(x0 ,r)
for x, y ∈ B(x0 , r), |f (x) − f (y)| ≤ sup |f (x) − f (y)| <  as given by Lemma 4.1 and hypothesis.
x,y∈B(x0 ,r)
We have met the condition for continuity of f at x (Proposition 9.4.7 in [1]).

The sufficient condition for the continuity is proved by proving its contrapositive: f is not con-
tinuous at x0 if ∃  such that ∀ r > 0, sup f (x) − inf f (x) > .
B(x0 ,r) B(x0 ,r)
The hypothesis is that ∃  such that ∀ r > 0, sup f (x) − inf f (x) > . From Lemma 4.1,
B(x0 ,r) B(x0 ,r)
we know that ∀ r > 0, sup |f (x) − f (y)| > . By definition of sup, ∀ r > 0, ∃ x, y ∈ B(x0 , r),
x,y∈E
|f (x)−f (y)| > . We have met the condition for discontinuity of f at x (Proposition 9.4.7 in [1]).
S
Lemma 4.4. S = Dα .
α>0
S
Proof. If x ∈ Dα , then ∃ α > 0 such that ∀ r > 0, sup f (x) − inf f (x) > α. From Lemma 4.3,
α>0 B(x,r) B(x,r)
S
it is clear that the function is not continuous at x and hence x ∈ S. Therefore, Dα ⊆ S.
α>0
If x ∈ S, then by definition, f is discontinuous
S at x. Hence,
S fromSLemma 4.3, we know that
∃ α > 0 such that x ∈ Dα . Hence, x ∈ Dα implying S ⊆ Dα or Dα = S.
α>0 α>0 α>0

4.2 Proof for necessary condition for Riemann integrability

In this subsection, the necessary condition is proven: If f (x) is a bounded function defined on [a, b]
and if the set of points on which f is discontinuous, say S, is a set of measure zero, then f is
Riemann integrable.
Proof. We would like to show f is Riemann integrable. We know,

1. f is bounded in [a, b].

2. S is of measure zero.

8
 From definition of a set of measure zero, given 1 > 0, ∃ a collection of open intervals {In }∞
n=0
∞
S ∞
P
such that S ⊆ In and |In | < 1 . Given α > 0, Dα ⊂ S from Lemma 4.4. Hence, from
n=0 n=0
Lemma 4.2 and open cover definition of compact sets (Theorem 12.5.8 in [7]), there exists a finite
sub-cover {Ink }N1
k=0 for Dα .
We can extract out from this finite sub-cover, a sub-cover comprising of disjoint open sets Gn .
This new sub-cover is generated using the algorithm of fusing overlapping open sets Ink . Hence,
SN N
S
Dα ⊆ Gn and we define a closed set K = [a, b] \ Gn . By definition,
n=0 n=0
N
X ∞
X
|Gn | ≤ |In | < 1 (2)
n=0 n=0
N
S
and from Remark 2.3, we have m([a, b]) = m(K) + m( Gn ) ≤ m(K) + 1 . Hence, if 1 < m([a, b]),
n=0

m(K) ≥ m([a, b]) − 1 > 0 (3)

Remark 4.1. Note that this approach is very similar to the proof for Theorem 2.3.
Since all points in [a, b] which have a discontinuous jump greater than α have been included
N
S
in Gn , ∃ a partition P K of K such that for all x, y ∈ Ji ∈ P K , |f (x) − f (y)| < α. But,
n=0
f (x) − f (y) ≤ |f (x) − f (y)|, ∀ x, y ∈ Ji . Hence, ∀ Ji ∈ P K , we have, from Lemma 4.1,
sup f (x) − inf f (x) = sup |f (x) − f (y)| ≤ α (4)
x∈Ji x∈Ji x,y∈Ji

 We define a partition
 P for [a, b] defined as P = P K ∪ {{G}N n=0 }. For brevity, we denote

sup f (x) − inf f (x) as (M − m)i for each Ii ∈ P . Since f is bounded, ∃ B ∈ R such that ∀ x ∈ [a, b],
x∈Ii x∈Ii

|f (x)| ≤ B (5)
Given  > 0, (chosen  must be less than3 2B × m([a, b])), consider U (f, P ) − L(f, P ). If f was
not bounded in [a, b], then U (f, P ) − L(f, P ) would not be well-defined. We have
X
U (f, P ) − L(f, P ) = (M − m)i |Ii |
Ii ∈P
h i
N
= By definition, P = P K ∪ {{G}n=0 }
X N
X
= (M − m)i |Ji | + (M − m)i |Gi |
Ji ∈P K i=0
h i
= From (4), (M − m)i ≤ α, ∀ Ji ∈ P K and by (5), (M − m)i ≤ 2B, ∀ {Gi }N
i=0

X N
X
≤α |Ji | + 2B |Gi |
Ji ∈P K i=0
" #
S P
≤ Use (2), and Remark 2.3 and K = Ji gives m(K) = |Ji |
Ji ∈P K Ji ∈P K

≤ αm(K) + 2B1
h i
 
= Choosing 1 = 4B
> 0 and α = 2m(K)
> 0 which is allowed by (3)
≤
Hence, by Definition 2.2, we conclude that f is Riemann integrable.
3
This constraint comes from (3) and the subsequent usage of  in determining 1 .

9
4.3 Proof for sufficient condition for Riemann integrability
We prove the sufficient condition by proving the contrapositive: If f (x) is a bounded function defined
on [a, b] and if the set of points on which f is discontinuous, say S, is not a set of measure zero,
then f is not Riemann integrable.
Proof. We would like to show f is not Riemann integrable. We know,

1. f is bounded in [a, b].

2. S is not of measure zero.

From Lemma 4.4 and Remark 2.4, we know that since m(S) > 0, then ∃ α > 0 such that
m(Dα ) > 0. By density of reals, ∃ α > 0 such that m(Dα ) > α .
Let P = {Ji }N i=0 for some N ∈ N be a partition of [a, b]. We can define an index set A as A =
{i : Ji ∩ Dα 6= φ}. Therefore, A has the indices of those intervals in the partition P which contain at
least one x ∈ Dα . Hence, followingSthe convention in Subsection 4.2, we have (M −Sm)i > α ∀ i ∈ A.
Also, by Remark 2.5, we have m( Ji ) ≥ m(Dα ) ≥ α since by definition, Dα ⊆ Ji .
i∈A i∈A
Given  > 0, consider U (f, P ) − L(f, P ). If f was not bounded in [a, b], then U (f, P ) − L(f, P )
would not be well-defined. We have
X X
U (f, P ) − L(f, P ) = (M − m)i |Ji | + (M − m)i |Ji |
i∈A i∈A
/
 
P
= (M − m)i |Ji | ≥ 0 and using Proposition 5.4.7(d) in [1]
i∈A
/
X
≥ (M − m)i |Ji |
i∈A
≥ αα

 
= Choosing α = α
>0
≥

We have thus produced an  > 0 such that for every partition P , U (f, P ) − L(f, P ) > . Using
Definition 2.2, we conclude that f is not Riemann integrable.

References
[1] T. Tao, Analysis I, 1st ed. Hindustan Book Agency, 2006.

[2] “Ga002: Some historical perspectives - [Link],” March 30, 2015. [Online]. Available:
[Link]

[3] D. Crytser, “Density of the rationals and irrationals in R,” July 23, 2012. [Online]. Available:
[Link]

[4] S. Abbott, Understanding Analysis, 1st ed. Springer, 2010. [Online]. Available:
[Link]

[5] “Thomae’s function,” October 6, 2010. [Online]. Available:

[Link] 10/[Link]

[6] W. F. Trench, Introduction to Real Analysis. Prentice Hall/Pearson Education, 2003.

[7] T. Tao, Analysis II, 2nd ed. Hindustan Book Agency, 2009.

10