THE PROPERTIES OF REAL NUMBERS II

BYUNGIK KAHNG

DIVISION OF SCIENCE AND MATHEMATICS

UNIVERSITY OF MINNESOTA, MORRIS, MN 56267, U.S.A.

Abstract. This note is prepared as the follow-up of the lecture note, The Properties of
Real Numbers I. In this note, we will cover the third and the hardest property of R, the
completeness property. These lecture notes are written as a replacement of Section (1.2) of
our textbook [7]. We encourage the readers to consult the reference books [2] and [8], rather
than going through (1.2) of [7].

1. Introduction
The sets of numbers were developed from two very different concepts. One is counting, and
the other is measurement. By drawing the abstract concept from counting, we get the set
of the whole numbers W. Repeated counting yields the addition. Repeated addition yields
the multiplication. By admitting the inverse operations, the subtraction and the division,
we extend W to Z, and then to Q.
Getting the set of the real numbers is not that simple, because it was developed from
completely different concept from the very beginning. The set of real numbers was obtained
from the measurement, the length in particular. The process of assigning a number on each
point on the line, thereby representing the length, is something fundamentally different from
counting. One such process is Dedekind Cut. It is discussed in (1.2) of our testbook [7].
We will not cover Dedekind cut in this course, however. Instead, we will assume the
completeness property as an axiom (Axiom 3) and find an alternative way to fill up the
entire line with the real numbers.

2. Preliminaries
In the lecture note, Properties of Real Numbers I, we studied the algebraic and the order
properties of R. These two properties are shared by Q as well. As far as the algebra and
ordering are concerned, therefore, R and Q are indistinguishable. In this note, we study the
property of R that separates it from Q. We begin with the following elementary example,
as a preparation.

Date: August 26, 2008.

1
2 BYUNGIK KAHNG

Example 2.1. Let I = [0, 1) ⊂ R, that is, I = {x ∈ R|0, x < 1}. Does it have the minimum?
Does it have the maximum?

Answer. [0, 1) has the minimum 0, but the maximum does not exist. (Why?). 

Not every subset in R has the minimum and/or the maximum. In Example 2.1, however,
we see that 1 behaves almost the same as the maximum of [0, 1), except that it not a member
of the set. Such numbers are called the supremum of the set. One of the key difference
between Q and R, is whether a bounded subset has the supremum or not. We will elaborate
this idea through the following definitions and the axiom.

Definition 2.2. A real number u is called an upper bound of a non-empty set S ⊂ R if

it is greater than or equal to every element of S. That is, x ≤ u for any x ∈ S. We denote
it, S ≤ u. The set S is said to be bounded above if there is an upper bound. Moreover,
we define the terms lower bound and bounded below similarly and denote l ≤ S. A set
is said to be bounded if it is bounded above and below at the same time.

3. The Completeness Property of R

Axiom 3 (The Completeness Property of R). Every nonempty subset of S ⊂ R, which is
bounded above, has the least upper bound (supremum) in R. We express it sup S.

The following lemma complements Axiom 3.

Lemma 3.1. Every nonempty subset of S ⊂ R, which is bounded below, has the greatest
lower bound (infimum) in R. We express it inf S.

Proof. It is easy to see that inf S = − sup(−S). We leave the detail to the readers. 

Remark 3.2. Even when the set S is unbounded, we can still consider its supremum and
infimum by allowing them to be ∞ or −∞.

The completeness is the key property of R that separates it from Q. We can exemplify
this as follows.

Example 3.3. Let S = {x ∈ Q|x2 < 2}. Clearly, S ⊂ Q ⊂ R. Find the least upper bound
of S in R. Does S have the least upper bound in Q, too?
√ √
Solution. S has the least upper bound in R. It is 2. However, 2 ∈
/ Q. Therefore, S has
no least upper bound in Q. 

Remark 3.4. From Definition 2.2 and Axiom 3, it is easy to see that the supremum is in
fact the minimum of the set of the upper bounds, while the infimum is the maximum of the
 THE PROPERTIES OF REAL NUMBERS II 3

set of the lower bounds. That is,


sup S = min{u ∈ R : x ≤ u, for all x ∈ S},
(3.1)
inf S = max{l ∈ R : l ≤ x, for all x ∈ S}.

Note that the inequalities in (3.1) already include the equality. Therefore, it makes sense that
the least upper bound and the greatest lower bound must always exist. Recall that [0, 1) =
{x : 0 ≤ x < 1} in Example 2.1 had minimum (≤ side) but no maximum (< side).

4. Properties of Supremum and Infimum

The following theorem completely characterizes the least upper bound (supremum) and
the greatest lower bound (infimum) in terms of inequalities.

Theorem 4.1. Let S be a nonempty subset of R and σ ∈ R. Then σ = sup S if and only if
(1) S ≤ σ (for every x ∈ S, x ≤ σ) and
(2) For every  > 0 there is an x ∈ S such that σ −  < x.
Similarly, τ = inf S if and only if (3) τ ≤ S; and (4) for every  > 0, x < τ +  for some
x ∈ S.

Proof. First, let us assume σ = sup S. Since the least upper bound is an upper bound
nonetheless, we must have S ≤ σ. This proves (1). Given any  > 0, σ −  < σ. Thus, σ − 
is not an upper bound, being smaller than the least upper bound. Hence, there must be
some x ∈ S such that σ −  < x. This proves (2) and thus the forward implication.
Now, let us assume (1) and (2). (1) says that σ is an upper bound. Therefore, we need
only to prove that it is the least upper bound. Any real number smaller than σ can be
expressed as σ −  for some  > 0. But then, (2) says that there must be a certain x ∈ S
such that σ −  < x. Thus, σ −  is not an upper bound. Consequently, any real number
smaller than σ is not an upper bound. Hence, σ must be the least upper bound. This proves
the backward implication.
Similar argument applies to the greatest lower bound. As a result, we get (3) and (4). 

Theorem 4.1 is applied to virtually all problems that are related to suprema and/or infima.
Here is the one such example.

Proposition 4.2. Let A and B be nonempty subsets of R.

(1) If A ⊂ B, then inf B ≤ inf A ≤ sup A ≤ sup B.
(2) If A ≤ B (∀a ∈ A, ∀b ∈ B, a ≤ b), then inf A ≤ inf B and sup A ≤ sup B.
(3) If C = A + B, then inf C = inf A + inf B and sup C = sup A + sup B.
(4) If C = A · B and A, B > 0, then inf C = inf A · inf B and sup C = sup A · sup B.
(5) If C = A/B and A, B > 0, then inf C = inf A/ sup B and sup C = sup A/ inf B.
4 BYUNGIK KAHNG

Proof. We will prove only the first two statements. The other claims are left to the readers
as home work problems.
Let α = sup A and β = sup B. Then, β is an upper bound of B where B ⊃ A. Therefore,
for every x ∈ A ⊂ B, x ≤ β. Thus, β is an upper bound of A, too. Hence, it must be greater
than or equal to the least upper bound, α. That is, α ≤ β. This proves sup A ≤ sup B.
Similar argument applies to infima. This proves (1).
Again, let α = sup A and β = sup B. Then, β is an upper bound of B where B ≥ A.
That is, for every x ∈ A, x ∈ y, where y ∈ B, and consequently,
x ≤ y ≤ β.
Hence, β is an upper bound of A. Thus α ≤ β follows, because α is the least upper bound.
Similar argument applies to infima. This proves (2).
The other claims are harder to prove, because they are equalities. You must show both ≤
and ≥. We leave the proofs to the readers. Theorem 4.1 may turn out handy here. 
It is not easy to prove Proposition 4.2. However, the conclusions of Proposition 4.2 are
nothing out of ordinary. That is, suprema and infima behave just like the maxima
and the minima in many aspects, except that they always exist.

5. The Archimedean Property

The following theorem is called the Archimedean Property. It is named after the
ancient Greek mathematician Archimedes. It turned out, over 2000 years later, that this
property can be proved from the completeness axiom but not vice versa.

Theorem 5.1 (The Archimedian Property). Given any x ∈ R, there is n ∈ N, such that
x < n. That is, N ⊂ R is not bounded above.

Proof. Suppose the conclusion of Theorem 5.1 is false. That is, there is certain x ∈ R such
that n ≤ x for all n ∈ N. In other words, N has an upper bound x ∈ R. By the completeness
property of R, there is a least upper bound l of N. Or,
l = sup N ∈ R.
Since l is the least upper bound, any real number less than l is not an upper bound. In
particular, l − 1 is not an upper bound. Therefore, there is a natural number k ∈ N such
that l − 1 < k. Hence, we must have l < k + 1 ∈ N, and thus l is not an upper bound of N.
This is a contradiction. Consequently, the conclusion of Theorem 5.1 follows. 
The following corollary is an immediate consequence of Theorem 5.1.
1
Corollary 5.2. Given  > 0, there is n ∈ N such that n
< . Consequently, lim 1/n = 0.
n→∞
1 1 1
Proof. 
∈ R. From Theorem 5.1, therefore, 
< n, for some n ∈ N. Thus, n
<  follows. 
 THE PROPERTIES OF REAL NUMBERS II 5

Remark 5.3. Corollary 5.2 finally allows us to do limit calculation of sequences and func-
tions. Without it, we could only show the existence of the given limit and no further. The
standard limit rules are almost useless one implicitly assumes Corollary 5.2.

6. Q is dense in R
It is easy to see that R is densely distributed, and so is Q.

Proposition 6.1. There is at least one real number between any two distinct real numbers.
Also, there is at least one rational number between any two distinct rational numbers.

Proof. Given pair of real (rational) numbers x and y, x < y, we can find a real (rational
number), m = (y − x)/2 in between. 
In this section, we prove far stronger theorem. That is, Q is densely distributed in R.

Theorem 6.2 (Q is dense in R). There is at least one rational number between any two
distinct real numbers.

Proof. First, let us find a rational number between any pair of real numbers. Given two
distinct real numbers x and y, x < y, let us put  = y − x > 0. Now, by Corollary 5.2, there
is a positive integer n ∈ N such that n1 < .
Now, let k be the smallest natural number such that nk > x. (Such k exists because N is
well ordered.) We claim x < nk < y. If not, we must have nk ≥ y, in which case,
k−1 k 1
= −
n n n
1
≥y−
n
> y − (y − x) (∵ y − x =  > 1/n)
=x
k
This contradicts the assumption that k is the smallest natural number such that n
< y − x.
Hence, x < nk < y. Clearly, nk ∈ Q. 
The following corollary supplements Theorem 6.2.

Corollary 6.3. There is at least one irrational number between any two real numbers.

Proof. Given pair of real numbers, x and y, we can find a rational number p between x and
y (Theorem 6.2). Without loss of generality, we can assume x < p < y. Applying Theorem
6.2 again to p and y, we can find q ∈ Q so that x < p < q < y. Finally, let us divide the
√
interval [p, q] by 1 : 2. That is, put
√
p+q 2
s= √ .
1+ 2
6 BYUNGIK KAHNG

Thus, we got an irrational number s ∈ (p, q) ⊂ (x, y). 

Applying Theorem 6.2 and Corollary 6.3 repeatedly, we can find infinitely many rational
numbers and irrational numbers between any pair of distinct real numbers.

7. Acknowledgments
This handout was written in part as a sample LaTex template file for UMM Mathematics
Discipline Senior Seminar (Math 4901). See [4] for more complete reference on LaTex for-
matting. The author thanks Xiaosheng Li for suggesting him to write a LaTex template. On
technical side, the author thanks Barry McQuarrie for sharing some of his LaTex macros,
and Joseph Iverson for informing him the latest LaTex softwares. Finally, the author thanks
to all UMM mathematics faculty members for their generous support.

References
[1] C. Boyer, U. Merzbach and I. Asimov, A History of Mathematics, Wiley, 2nd ed., 1991.
[2] R. Bartle and D. Sherbert, Introduction to Real Analysis, Wiley, 3rd ed., 2000.
[3] R. Buck, Advanced Calculus, Waveland Press, 3rd ed., 2003.
[4] G. Grätzer, Math into LaTex, Birkhäuser, 3rd ed., 2000.
[5] E. Hairer and G. Warner, Analysis by Its History, Springer, 1st ed., 2005.
[6] J. Marsden and M. Hoffman, Elementary Classical Analysis, W. H. Freeman, 3rd ed., 2007.
[7] C. Pugh, Real Mathematical Analysis, Springer, 1st ed., 2003.
[8] M. Spivak, Calculus, Publish or Perish, 3rd ed., 1994.

Division of Science and Mathematics, University of Minnesota, Morris, MN, 56267, U.S.A.
E-mail address: kahng@morris.umn.edu.