Real Numbers

Abstract

We began our rigorous development of number theory with defi-

nitions of N (the natural numbers), Z (the integers), and Q (the
rational numbers).
√ Our task now is to complete the rationals so that
we can include 2, π, and other irrationals. There are at least three dif-
ferent ways to construct the real numbers R, and these constructions
are all the same in the sense of being isomorphic. In particular, they
all satisfy the axioms of R, which are basically the same as those for
Q, namely the field axioms, plus one more, the least upper bound
property. As will be seen, this last axiom ensures that all limits of
rational sequences are contained in R, and in fact we can construct R
in terms of sequences and their limits.
But there is another purpose to R besides ‘including limits.’ In
fact, the way limits arise in the first place is through approximation
of certain quantities, e.g. the length of the circumference of the circle
approximated by inscribed polygons, or the diagonal of a square of
side-length one approximated by rational multiples of the unit side-
length. Real numbers are used in measurement, and for this rea-
son MacLane’s geometric construction, using Hilbert’s update of Eu-
clid’s five postulates, is of utmost importance. It links the original
Pythagorean theory of magnitudes to the modern set-theoretic for-
mulation.

1 Axioms of the Real Numbers

The axioms of R break down into two types: algebraic and analytic. The
algebraic axioms, 1: 1-9 are the same as 4-13 of Q, or any field whatso-
ever. We may also add 2: 1-4 and 3: 1-2 into the algebraic category. The
only analytic axiom is 4, the least upper bound property, is also called
the (order) completeness of R, and it is the axiom which makes limits,
measurement, and calculus possible.

We simply define R to be a structure

(R, ≤, −,−1 , +, ·, 0, 1)

1
which is a complete and totally ordered field. Let us explain this: we define
the real numbers R to be numbers (not yet characterized as a set, or anything
in particular) having two unique elements, 0 and 1, and binary operations +
(addition) and · (multiplication), unary operations − (negation) and −1
(multiplicative inversion), as well as a partial order relation ≤, which
all must satisfy the following axioms:

1. (R, −,−1 , +, ·, 0, 1) is a field, that is for all real numbers x, y, z we

have

1. x + (y + z) = (x + y) + z (associativity of +) 

2. x + y = y + x (commutativity of +)  (R, +) is an


3. ∃0 ∈ R such that x + 0 = x 
 abelian group

4. ∀x ∈ R, ∃ − x ∈ R such that x + (−x) = 0 

(associativity of ·)

5. x(yz) = (xy)z 

6. xy = yx (commutativity of ·)  (R\{0}, ·) is an


7. ∃1 ∈ R such that 1x = x and 0 6= 1 

 abelian group
∀x ∈ R\{0}, ∃x−1 ∈ R such that xx−1 = 1

8. 

9. x(y + z) = xy + xz (distributivity of · over +)

2. (R, ≤) is a totally ordered partially ordered set, that is ≤ must

satisfy, for all real numbers x, y and z,

1. x ≤ y or y ≤ x (totality)
2. x≤x (reflexivity)
3. x ≤ y and y ≤ x =⇒ x = y (antisymmetry)
4. x ≤ y and y ≤ z =⇒ x ≤ z (transitivity)

3. The field operations + and · are compatible with the partial order
≤ in the sense that for all x, y, z ∈ R we have

1. x ≤ y =⇒ x + z ≤ y + z (preservation of order under addition)

2. 0 ≤ x and 0 ≤ y, then 0 ≤ xy (preservation of order under multiplication)

4. ≤ is a complete order relation, meaning it satisfies the least upper

bound property: for any nonempty collection S of real numbers which is
bounded above there exists a real number sup S, the least or smallest upper

2
bound. (The notation sup here stands for supremum, which is a Latinized
form of ‘least upper bound.’)

Definition 1.1 A collection S of real numbers is said to be bounded

above by a real number M if every number x in the collection S satisfies

x≤M

Let us write
def
S u = {all upper bounds of S}
and let us write
def
sup S = min S u
which is the minimum, or smallest of the upper bounds. This number exists
by fiat, by Axiom 4. 

1.1 Motivation for the Least Upper Bound Axiom

Let us here consider the limitations of the rational numbers Q for conducting
measurement.

Theorem 1.2 (Q is Densely Ordered) Between any two distinct ratio-

nal numbers p and q there exists a rational number r, and hence infinitely
many rational numbers.

Proof : If p 6= q are rational numbers, then since Q is totally ordered have

two cases
p < q and p > q
1 1
Case 1: p < q: In this case, the rational number 2 (p + q) = 2p + 12 q lies
between p and q, as you can see directly:

p < q =⇒ 2p < p + q < 2q

1
=⇒ p < (p + q) < q
2
(Note: both implications seem to rely on Axiom 3 of R above, but looking
at properties (10)-(11) of Theorem 3.12 in Numbers, and the subsequent
constructions of Z and Q, it would seem these are in fact derived from

3
Theorem 3.12 in the case of Q: exercise!) Since this is true for each p and
q, in particular it is true for p and 21 (p + q), so that
 
1 1 1 1
p < (p + q) = (p + q) < (p + q) < q
4 2 2 2
Repeating
this process for all n ∈ N we have a sequence of rational numbers
1
2n (p + q) n∈N lying between p and q.

Case 2: p > q: reverse the roles of p and q in the proof of Case 1. 

Theorem 1.3 (Archimedean Property of Q) For all p, q ∈ Q with p >

0, there exists a natural number n ∈ N such that pn > q.

Remark 1.4 If we imagine p and q as masses on a see-saw, or balance,

then supposing p a smaller mass than q, e.g. Archimedes the man versus a
mountain, p < q, then by multiplying the power of Archimedes n-fold times,
e.g. by increasing the length of p’s arm of the lever n-fold, the result is
that Archimedes will lift the mountain. Our assumptions about N, Z, and Q
make this a necessary consequence. 

Proof : If q ≤ 0, then ∃1 ∈ N such that 1p = p > q. Thus, assume that

q > 0. Let p = ab and q = dc , where a, b, c, d ∈ N. Let us consider the options:
(1) p = q
(2) p > q > 0
(3) 0 < p < q
The first two cases are included only out of logical pedantry. In the first
case ∃2 ∈ N such that q = p < 2p. In the second case, p > q > 0, and
∃1 ∈ N such that 1p = p > q.

We move on to the third and only interesting case, when 0 < p < q. We
need to find some n ∈ N such that q < np. I claim that

n = b(c + 1)

will work if we take p = ab and q = dc as above. To see this, note first that
0 < p < q means 0 < ab < dc , and therefore 0 < ad < bc (Theorem 3.12,
Numbers). Since 0 < p < q, all a, b, c, d > 0, and in particular this implies
a, b, c, d ≥ 1. In particular, d ≥ 1 implies dq ≥ 1q ≥ 1, which we insert here:
c
c+1>c=d· = dq ≥ q
d

4
citing again Theorem 3.12 in Numbers for the relevant maneouvers (that
c + 1 > c and d ≥ 1 implying dq ≥ 1q ≥ 1). What do we gain from this?
This:
a
q < c + 1 ≤ a(c + 1) = b(c + 1) = np
b
where n = b(c + 1) ∈ N. 

Proposition 1.5 The rationals Q do not satisfy the least upper bound
property, meaning there does not always exist a least upper bound sup S
in Q for arbitrary bounded subsets S ⊆ Q.

Proof : Though the rational numbers are densely ordered, they do not
satisfy the least upper bound property. If we consider the subset

S = {p ∈ Q | p2 < 2}

then, first of all, S is nonempty because 1 ∈ S, and second of all S is bounded

above by 2 ∈ Q: all p ∈ S satisfy p < 2, for otherwise, if p ≥ 2, we’ll have
p2 ≥ 4 > 2. Thus, S u , the set of upper bounds for S, is nonempty, because
2 ∈ S u . Now comes the crux of the matter: though S u is nonempty, it has
no minimum, no sup S, no least upper bound in Q. For if you suppose to
the contrary,√that S has a least upper bound ` ∈ Q, then first of all `2 6= 2
because ` = 2 ∈ / Q, while both `2 > 2 and `2 < 2 lead to contradictions:

(1) Suppose `2 < 2 (i.e. that ` ∈ S). If ` = sup S and `2 < 2, then
` ∈ S and we conclude that ` = max S. This, however, cannot be:
if we suppose ` = max S, then because 2 − `2 > 0, the Archimedean
property of Q says that ∃n ∈ N such that n(2 − `2 ) > 2` + 1. But then
1
n(2 − `2 ) > 2` + 1 ≥ 2` +
n
or
1 2
 
1
2 1
2>` +2 `+ 2 = `+
n n n
whence ` + n1 ∈ S and ` + n1 > `, so ` 6= max S, a contradiction.
Consequently, if ` ∈ S then ` 6= sup S because ` is not even an upper
bound of S. This contradicts our assumption that ` = sup S.
(2) On the other hand, if we suppose that `2 > 2, then ` is at least an
upper bound of S, but we run into the problem that the set of upper
bounds S u = {r ∈ Q | r > 0, r2 > 2} of S has no minimum in Q: since

5
 r > 0 and r2 − 2 > 0, the Archimedean property of Q guarantees that
∃n, m ∈ N such that nr > 1 and m(r2 − 2) > 2r. Let M = max{n, m}.
Then M ∈ N, and
1
Mr > 1 and M (r2 − 2) > 2r ≥ 2r −
M
or
1 2
 
1 1 1
r>r− >0 and r− = r2 − 2r + 2 >2
M M M M
Thus, for all ` ∈ S u , there exists r ∈ S u with ` > r > 0, so @` =
min S u = sup S. 

The least upper bound property is an abstraction used in the axiomatiza-

tion of R considered separately from any concrete realization, whether using
pure sets or some geometric model of the Euclidean plane (as in MacLane).
Our first construction, via Dedekind cuts, uses it directly, but we’ll come to
see that this construction is isomorphic (i.e. essentially equivalent) to the
one using limits, and also to MacLane’s geometric ‘number line’ construction
out of Hilbert’s re-modeling of Eulidean geometry.

2 Dedekind Cut Construction of R

A Dedekind cut is a set A of the form A = (−∞, p) ⊆ Q, where p ∈ Q,

using interval notation. This is because A 6= ∅ by the Archimedean property
of Q, Theorem 1.3, since there always exists q such that q < p because there
always exists −q ∈ Q+ such that −q > p, and hence ∃q = −(−q) ∈ Q with
q < p whether p is positive or negative. Moreover, for any x ∈ A = (−∞, p)
we have that y ≤ x =⇒ y ∈ A, and since p ∈ / A we have that x <
p =⇒ x ∈ A, which by the order denseness of Q, Theorem 1.2, means that
all y ∈ Q with x < y < p are in A, and there are infinitely many such y, so
that @ max A. Moreover, by the Archimedean property again, since p ∈ Q,
there exists q ∈ Q such that p = 1p < q, whence Q\A 6= ∅. Dedekind cuts
allow us to define the real numbers as follows:

A real number A in this construction is a Dedekind cut, and R is the

set of all Dedekind cuts of Q. To make sure we give R the structure it needs,

6
we must define a partial order ≤, the binary operations of addition + and
multiplication ·, and we must also give R the additive and multiplicative
identities 0 and 1, and then show that these satisfy the usual properties.
First, we define 0 and 1:

0 = {p ∈ Q | p < 00 , 00 ∈ Q}
= (−∞, 00 ) ⊆ Q

1 = {p ∈ Q | p < 10 , 10 ∈ Q}
= (−∞, 10 ) ⊆ Q

partial order Then, we define the partial order ≤ on the reals by requiring

A ≤ B ⇐⇒ A ⊆ B

with A = B iff A = B as sets. We define addition + on R by letting

A + B = {p + q | p ∈ A, q ∈ B}

We define the opposite −A of A ∈ R by

−A = {p ∈ Q | − p ∈
/ A and − p 6= min Q\A}

For example,

−5 = {p ∈ Q | − p ∈
/ 5 and p 6= min Q\5}
= {p ∈ Q | − p > 5}
= {p ∈ Q | p < −5}

This definition is good because Q satisfies the Archimedean property by T

1.3, so (r, ∞) is nonempty. The opposite of a real number A, in turn, allows
us to define the absolute value function | · | : R → R+
0 by
(
A, if A ≥ 0
|A| =
−A, if A < 0

A real number A is said to be positive if A > 0 and is said to be

negative if A < 0. The set of all positive real numbers is denoted R+ ,
the set of all nonnegative numbers is denoted R− −
0 and equals R ∪ {0},
−
the set of all negative real numbers is denoted R , and set of all
nonpositive numbers is denoted R− −
0 and equals R ∪ {0}.

7
 the distance d : R × R → R+
0 between two real numbers A and B is given
by
d(A, B) = |A − B|

Now we can define multiplication ·. We do this in two steps, first for

A, B > 0, then for all A, B ∈ R. If A, B > 0,

AB = Q− + +
0 ∪ {qr | q ∈ A ∩ Q , r ∈ B ∩ Q }

and in general, for all A, B ∈ R we say



 0, if A = 0 or B = 0
AB = |A||B|, if A, B > 0 or A, B < 0

−|A||B|, if A > 0 & B < 0 or if A < 0 & B > 0


For example

5 · 6 = Q− + +
0 ∪ {qr | q ∈ 5 ∩ Q , r ∈ 6 ∩ Q }
= (−∞, 30) ⊆ Q

while

−5 · 6 = −| − 5||6|
= −(5 · 6)
= −(−∞, 30) ⊆ Q
= (−∞, −30) ⊆ Q

The multiplicative inverse of a nonzero real number A is defined by

 
−1 1
A = p ∈ Q | p ≤ 0 or else p > 0 & ∈ / A & p 6= min Q\A
p
For example,
 
−1 1
5 = p ∈ Q | p ≤ 0 or else p > 0 & ∈/ 5 & p 6= min Q\A
p
 
− 1
= Q0 ∪ p ∈ Q | > 5
p
 
− 1
= Q0 ∪ p ∈ Q | p <
5
= (−∞, 1/5)

8
The multiplicative inverse of a negative real A < 0 is given by
A− = −(|A|)−
For example −5− = −(| − 5|)− = −(−∞, 1/5) = (−∞, −1/5). The multi-
plicative inverse is a good definition because, again, Q satisfies the Archimedean
property, so p > 0 always exist such that p < 1q for any q > 0.

The commutativity and associativity of addition and multiplication in

R, as well as distributivity of · over +, follow from the properties of Q and
our definitions of + and · in R above. In fact, all the points taken for axioms
in the previous “construction” except the least upper bound follow from our
definitions. We state this formally in the next theorem:
Theorem 2.1 The set of real numbers R is a structure (R, ≤, −,−1 , +, ·, 0, 1)
which is a totally ordered field:

1. (R, −,−1 , +, ·, 0, 1) is a field, that is for all x, y, z ∈ R we have


1. x + (y + z) = (x + y) + z (associativity of +) 

2. x+y =y+x (commutativity of +)

 (R, +) is an
3. ∃0 ∈ R such that x + 0 = x 
 abelian group

4. ∀x ∈ R, ∃! − x ∈ R such that x + (−x) = 0 
(associativity of ·)

5. x(yz) = (xy)z 

6. xy = yx (commutativity of ·)  (R\{0}, ·) is an


7. ∃1 ∈ R such that 1x = x and 0 6= 1 

 abelian group
∀x ∈ R\{0}, ∃x−1 ∈ R such that xx−1 = 1

8. 
9. x(y + z) = xy + xz (distributivity of · over +)

2. (R, ≤) is a totally ordered set, that is for all x, y, z ∈ R we have

1. x≤x (reflexivity)
2. x ≤ y and y ≤ x =⇒ x = y (antisymmetry)
3. x ≤ y and y ≤ z =⇒ x ≤ z (transitivity)
4. x ≤ y or y ≤ x (totalness)

3. The field operations + and · are compatible with the partial order
≤ in the sense that for all x, y, z ∈ R we have

1. x ≤ y =⇒ x + z ≤ y + z (preservation of order under addition)

2. 0 ≤ x and 0 ≤ y, then 0 ≤ xy (preservation of order under multiplication)

9
Proof : These properties follow from those of Q: 1.1.

x + (y + z) = x + {b + c | b ∈ y, c ∈ z}
= {a + (b + c) | a ∈ x, b ∈ y, c ∈ z}
= {(a + b) + c | a ∈ x, b ∈ y, c ∈ z}
= {a + b | a ∈ x, b ∈ y} + z
= (x + y) + z

1.2. x + y = {a + b | a ∈ x, b ∈ y} = {b + a | a ∈ x, b ∈ y} = y + x.

1.3. We have already defined 0 and 1, so we only need check their

defining properties. First, for any x ∈ R we have x = (−∞, r) ⊆ Q for some
r ∈ Q, so we need to show

x+0 = {a+b|a ∈ x, b ∈ 0} = {a+b|a ∈ (−∞, r), b ∈ (−∞, 00 )} = (−∞, r) = x

which we do as follows: if q ∈ x + 0, then q = a + b for some a ∈ x and

b ∈ 0, i.e. for some a ∈ (−∞, r) and b ∈ (−∞, 00 ). But clearly a + b < a,
so p = a + b ∈ x, which shows that x + 0 ⊆ x. Conversely, if a ∈ x, then
∃a0 ∈ x and b ∈ 0 such that a = a0 + b ∈ x by the order denseness of Q,
since if a ∈ x then a < r and so ∃a0 ∈ x such that a < a0 < r. Letting
b = −(a0 − a) = a − a0 ∈ 0, we have that a = a0 + (a − a0 ) = a0 + b ∈ x + 0.
Thus x ⊆ x + 0, and therefore by the axiom of extensionality x + 0 = x.

1.4. If x = (−∞, r) is a real number, r ∈ Q, then it’s opposite −x =

(−inf ty, −r) is also its additive inverse or negative, which is shown as fol-
lows: first

x + (−x) = {a + b | a ∈ x, b ∈ −x} = {a + b | a < r, b < −r}

so if r = 0 then x = 0, so −x = x = 0 and by the previous point we have

x + (−x) = 0 + 0 = 0. If r > 0 then −r < 0, so if a < r and b < −r, then
a + b < r + (−r) = 0, so a + r ∈ 0. If r < 0 then we apply the same reasoning
to r0 = −r. Thus for all x ∈ R we have that ∃ − x ∈ R with x + (−x). The
uniqueness of (−x) is shown as follows: if there were any other y ∈ R with
x + y = 0, then

−x = −x + 0 = −x + (x + y) = (−x + x) + y = 0 + y = y

1.5. It suffices to prove the case x, y, z > 0, for the other cases follow from

10
the first:
 
x(yz) = x · Q−
0 ∪ {bc | b ∈ y ∩ Q+
, c ∈ z ∩ Q+
}
= Q− + + +
0 ∪ {a(bc) | a ∈ x ∩ Q , b ∈ y ∩ Q , c ∈ z ∩ Q }
= Q− ∪ {(ab)c | a ∈ x ∩ Q+ , b ∈ y ∩ Q+ , c ∈ z ∩ Q+ }
0 
= Q−0 ∪ {ab | a ∈ x ∩ Q +
, b ∈ y ∩ Q+
} ·z
= (xy)z

1.6. xy = {ab | a ∈ x, b ∈ y} = {ba | a ∈ x, b ∈ y} = yx.

1.7. It is enough to prove this for x > 1, for the other cases, x < 0 and
x = 0 follow easily from the first. For any x ∈ R+ we have x = (−∞, r) ⊆ Q
for some r ∈ Q+ , so
x1 = {ab | a ∈ x, b ∈ 1}
We must show that x1 = x: first, if p = ab ∈ x1 for a ∈ x and b ∈ 1, then
since b < 10 we have p = ab < r, so p ∈ x. Thus x1 ⊆ x. Conversely, if
2pr 2pr
p ∈ x, then ∃a = p+r ∈ x ( p+r < r since 2pr = pr + pr < pr + r2 ) and
∃b = p+r p+r 0
2r ∈ 1 ( 2r < 1 since p + r < 2r) such that

2pr p + r
ab = · =p
p+r 2r
Thus x1 ⊃ x, and by extensionality we have x1 = x.

1.8. For any x ∈ R\{0} we have x = (−∞, r) ⊆ Q for some r ∈ Q\{00 },

and so 1r = r−1 ∈ Q\{00 }, whence x−1 = (−∞, r−1 ) is the multiplicative
inverser of x, that is

xx−1 = Q− +
0 ∪ {ab | a ∈ x ∩ Q , b ∈ x
−1
∩ Q+ } = 1 = (−∞, 10 ) ⊆ Q

which we can show as follows: we’ll prove it for x > 0, since the case x < 0
follows from the first by considering |x| = −x: if p = ab ∈ xx−1 for 0 < a < r
and 0 < b < r−1 , then a < r and b < r−1 , so p = ab < rr−1 = 10 , whence
p ∈ 1. This shows that

xx−1 = Q− +
0 ∪ {ab | a ∈ x ∩ Q , b ∈ x
−1
∩ Q+ } ⊆ 1

To show the reverse inclusion, let p ∈ 1. If p < 0 then p ∈ Q− −1

0 ⊆ xx , so
0
suppose 0 < p, so that 0 < p < 1 ∈ Q. There are two cases, x > 0 and
x < 0. We’ll prove x > 0 since x < 0 follows from the x > 0 case by taking

11
the absolute of x. Now, if x > 0, then x = (−∞, r) for some r = hk ∈ Q+ ,
where h, k ∈ N. This produces three cases, r < 10 , r = 10 and r > 10 . a. If
r = 10 , then x = 1, so by 7 above we have that x2 = x1 = x = 1, so x = x−1 ,
which exists because x = 1 exists. b. If r > 1, choose p ∈ 1, i.e. p = m 0
n <1
+
for some m, n ∈ N, and let a ∈ x ∩ Q be given by
 
N −1 h (N − 1)h + k
a=1+ −1 =
N k Nk

for some N ∈ N with N > 1, so that 1 < a < p. Then, note that since
m m
n < 1 we have 1 − n > 0, so by the

Archimedean property of Q there exists
an N ∈ N such that 1 < N 1 − m n , or
 m
1−N 1− <0
n
Choose this N for our expression of a. We now need to find b ∈ r−1 , i.e.
b < hk , such that ab = p. Towards this end, note that by the above inequality
and the fact that 0 < hk iff 0 < hk we have
 m k n − N (n − m) k
1−N 1− <0< ⇐⇒ <
n h n
h
⇐⇒ h n − N (n − m) < nk


⇐⇒ h mN − n(N − 1) < nk
⇐⇒ mhN < nh(N − 1) + nk
mN 1
⇐⇒ <
nh(N − 1) + nk h
mN k k
⇐⇒ < = r−1
nh(N − 1) + nk h

Thus, if we let
mN k mN k
b= =
nh(N − 1) + nk n(h(N − 1) + k)

we have that b < r−1 , so that b ∈ x−1 , and moreover,

(N(−
(1)h
((+ k ((
m
Nk m
ab = ( · ((= =p
Nk n( ((−
(h(N (1) + k) n
 ( (


Thus, if r > 1 we have that for all positive p ∈ 1, ∃a ∈ x and b ∈ x−1 such
that ab = p, so 1 ⊆ xx−1 . c. If r < 1, however, we may repeat the above

12
procedure with r0 = r−1 , that is by switching a and b, so 1 ⊆ xx−1 in this
case too.

Hence by a-c we have that 1 ⊆ xx−1 in all cases where 0 < p < 1, and
since 1 ⊃ xx−1 by the previous arguments, we have by extensionality that
∀x ∈ R+ , ∃x−1 such that xx−1 = 1 whenever 0 < p < 1. This is also true,
trivially, when p < 0, as shown above, so it holds true for all x > 0. Finally,
as promised, we show that the case x < 0 follows from that of x > 0: if
x < 0, then x−1 < 0 also, and so xx−1 = |x||x−1 |, and the above procedure
shows that ∃x−1 making this true, namely x−1 = (−∞, r−1 ), which follows
from the fact that we have |x−1 | = |x|−1 , so that x−1 = −|x|−1 .

1.9. Finally, distributivity follows from distributivity in Q. It suffices to

prove the case x, y, z > 0, for the other cases follow from the first by taking
absolute values:
x(y + z) = x · {b + c | b ∈ y, c ∈ z}
= Q− + + +
0 ∪ {a(b + c) | a ∈ x ∩ Q , b ∈ y ∩ Q , c ∈ z ∩ Q }
= Q− ∪ {ab + ac | a ∈ x ∩ Q+ , b ∈ y ∩ Q+ , c ∈ z ∩ Q+ }
0 
= Q−0 ∪ {ab | a ∈ x ∩ Q +
, b ∈ y ∩ Q +
} +
 
Q−0 ∪ {ac | a ∈ x ∩ Q +
, c ∈ z ∩ Q+
}
= xy + xz

2.1. Clearly, for all x = (−∞, r) we have x = x as sets, so x ≤ x as well.

2.2. If x = (−∞, r) and y = (−∞, s) satisfy x ≤ y and y ≤ x, then

(−∞, r) ⊆ (−∞, s) and (−∞, r) ⊃ (−∞, s), so by extensionality (−∞, r) =
(−∞, s), whence x = y.

2.3. If x = (−∞, r), y = (−∞, s) and z = (−∞, t) satisfy x ≤ y and

y ≤ z, then (−∞, r) ⊆ (−∞, s) and (−∞, s) ⊆ (−∞, t), so by the definition
of set inclusion we have
(−∞, r) ⊆ (−∞, s) ⊆ (−∞, t) =⇒ (−∞, r) ⊆ (−∞, t)
or x ≤ z.

2.4. For all x = (−∞, r), y = (−∞, s) ∈ R we have r ≤ s or s ≤ r by the

totalness of Q, whence (−∞, r) ⊆ (−∞, s) or (−∞, r) ⊃ (−∞, s), so that
x ≤ y or y ≤ x.

13
 3.1. If x = (−∞, r) and y = (−∞, s) satisfy x ≤ y, then for any z =
(−∞, t) we have x+z = {a+b|a ∈ x, b ∈ z} and y+z = {a+b|a ∈ y, b ∈ z}.
But if x ≤ y, then x ⊆ y, so if a ∈ x then a ∈ y, whence if a + b ∈ x + z
then a + b ∈ y + z, so x + z ⊆ y + z as sets, or x + z ≤ y + z.

3.2. If x = (−∞, r) and y = (−∞, s) satisfy 0 ≤ x and 0 ≤ y, then

r, s ≥ 0. If r = s = 0 then x = y = 0 and xy = Q− 0 = 0. If r, s > 0 then
+ +
∃a ∈ x ∩ Q and b ∈ y ∩ Q , so ab > 0 for such a and b, whence

xy = Q− + +
0 ∪ {ab | a ∈ x ∩ Q , b ∈ y ∩ Q }

so that
Q− − + +
0 ( Q0 ∪ {ab | a ∈ x ∩ Q , b ∈ y ∩ Q }

or 0 < xy. Either way, then, 0 ≤ xy. 

We may now state and prove the key property of R:

Theorem 2.2 (R is Complete) R as constructed above satisfies the least

upper bound property: every nonempty subset S of R that is bounded above
has a least upper bound, given by
[
sup S = A
A∈S

Proof : If S ⊆ R and ∃B ∈SR such that A ≤ B for all A ∈ S, then we

must check,Sfirst of all, that A∈S A is a real number, that is a Dedekind
cut. First, A∈S A S is nonempty

because it consists of a union of nonempty
sets, and also Q\ A∈S A is nonempty because, if B ∈ R, then Q\B is
nonempty, so if r ∈ Q\B, then S r 6= A for all A ∈ S, because all A ≤ B
means A ⊆ B.SMoreover, A∈S A has no greatest element, for if it did,
say m = max A∈S A, then m ∈ A for some S A ∈ S, so m = max A,
contradicting the
S fact that @ max A. Finally, A∈S A is a down-set of Q,
because if x ∈ A∈S A and y < x, then,Sx ∈ A for some A ∈ S,Sand A is
a down-set, so y ∈ A, which means y ∈ A∈S A. Thus sup S = A∈S S A is
a real number. Finally, it is an upper bound of S, for clearly A ⊆ A∈S A
for all A ∈ S, and moreover it S is the least upper bound, for clearly for all
upper bounds B of S we have A∈S A ⊆ B because all A ∈ S satisfy A ⊆ B,
whence sup S ≤ B for all upper bounds B. 

14
3 Cauchy Sequence Construction of R

Another explicit construction of the reals arises from an attempt to fix a

differnt problem with the rationals, that not every sequence in Q con-
verges in Q. For example the sequence (1, 2, 32 , 45 , 11 23
8 , 16 , . . . ) in Q con-
√
verges to 2 ∈ / Q (this sequence is constructed recursively using bisection
method and calculus methods, namely the Intermediate Value Theorem–
nevertheless, it is a sequence in Q). We would like to make numbers out of
all such limits of sequences in Q. However, if we go about it directly we will
run into several problems, the biggest of which is that we will make a set
out of strange elements, elements which are constructed inside known sets,
yet which do not belong to any known set. The ‘limits’ of some sequences
in Q do not yet exist in any known set.

Another problem is that there may be several sequences with the same
‘limit,’ though eventually we want that limit to be a single number.

Our solution to the first problem is to make use only of Cauchy se-
quences in Q, that is sequences which satisfy
lim |xn − xm | = 0
n,m→∞

This makes no mention of limits, but it does allow us to introduce the

limits later, once we embed Q in our new set R. The way around the
second problem is to do what we did in the definition of the integers and
the rationals, to invent an equivalence relation, in this case an ‘equivalence’
between Cauchy sequences with the same ‘limit,’ that is with
def
(xn )n∈N and (yn )n∈N ⇐⇒ lim |xn − yn | = 0
n→∞

for Cauchy sequences (xn )n∈N and (yn )n∈N . We have thus hit upon the fol-
lowing construction: we define the real numbers as equivalence classes of
Cauchy sequences whose difference tends to zero (and are in this way ‘equiv-
alent’). Consider again the example of the sequence (1, 2, 32 , 54 , 11 23
8 , 16 , . . . ) in
Q. If we neglect to mention its ‘limit’ and speak only of the fact that it is a
Cauchy sequence in Q, then we can make an equivalence class √ [(xn )n∈N ] out
of the sequence and call this the square root of 2, that is 2 := [x]. More
formally:

A Cauchy sequence in Q is a sequence (xn )n∈N in Q which satisfies

∀ ∈ Q+ , ∃N ∈ N such that m, n ≥ N =⇒ |xn − xm | < 

15
that is
lim |xn − xm | = 0
n,m→∞

The set C consisting of all Cauchy sequences in Q can be adorned with some
algebraic structure: we define addition +, multiplication · and a partial or-
der ≤ on C, as follows:

(xn )n∈N + (yn )n∈N := (xn + yn )n∈N

(xn )n∈N · (yn )n∈N := (xn yn )n∈N
(xn )n∈N ≤ (yn )n∈N iff ∀ ∈ Q+ , ∃N ∈ N s.t. n ≥ N =⇒ xn −  ≤ yn
or
(xn )n∈N ∼ (yn )n∈N

These allow us to establish the equivalence relation on C that we want,

namely that which ‘equates’ Cauchy sequences with the same limit (but
without mentioning the limit):

(xn )n∈N ∼ (yn )n∈N ⇐⇒ lim (xn − yn ) = 0

n→∞

To show that this defines an equivalence relation, let  ∈ Q+ be given.

Then, because (xn )n∈N is a Cauchy sequence, we have that ∃N ∈ N such
that n, m ≥ N =⇒ |xm − xn | < , so

(xn )n∈N ∼ (xn )n∈N

Next, suppose (xn )n∈N , (yn )n∈N ∈ C satisfy (xn )n∈N ∼ (yn )n∈N , that is
limn→∞ (xn −yn ) = 0. Then clearly limn→∞ (xn −yn ) = limn→∞ (yn −xn ) = 0
because |xn − yn | = |yn − xn | for all n ∈ N, so

(xn )n∈N ∼ (yn )n∈N =⇒ (yn )n∈N ∼ (xn )n∈N

Finally, if (xn )n∈N ∼ (xn )n∈N and (yn )n∈N ∼ (zn )n∈N , then ∀ ∈ Q+ ,
∃N1 , N2 ∈ N such that n ≥ N1 =⇒ |xn − yn | < 2 and n ≥ N2 =⇒
|yn − zn | < 2 , and as a result for all n ≥ N = max{N1 |xn − yn |, N2 } =⇒
 
|xn − zn | = |xn − yn + yn − zn | ≤ |xn − yn | + |yn − zn | < + =
2 2
so limn→∞ (xn − zn ) = 0 and (xn )n∈N ∼ (zn )n∈N . This proves that

(xn )n∈N ∼ (yn )n∈N and (yn )n∈N ∼ (zn )n∈N =⇒ (xn )n∈N ∼ (zn )n∈N

16
so ∼ is indeed an equivalence relation. The set of real numbers R is then
defined as the quotient set on C by ∼, that is the set of all equivalence
classes generated by ∼:
R := C/ ∼
We define addition +, multiplication · and a partial order ≤ on R as follows:

x + y = [(xn )n∈N ] + [(yn )n∈N ] := [(xn )n∈N + (yn )n∈N ] = [(xn + yn )n∈N ]

xy = [(xn )n∈N ][(yn )n∈N ] := [(xn )n∈N (yn )n∈N ] = [(xn yn )n∈N ]

x ≤ y or [(xn )n∈N ] ≤ [(yn )n∈N ] ⇐⇒ (xn )n∈N ∼ (yn )n∈N

or
∃N ∈ N s.t. n ≥ N =⇒ xn < yn

After we make sure (Lemma 3.2 below) that for all nonzero real numbers
there exist Cauchy sequence consisting either entirely of positive rational
numbers or entirely of negative rational numbers, we can define the positive
and negative real numbers, as well as the absolute value of a real
number, as follows:

A real number x is said to be positive if it has a representative Cauchy

sequence (xn )n∈N ∈ Q+ consisting entirely of positive rational numbers. We
denote this by x > 0 or x ∈ R+ . The set of all positive real numbers
is denoted R+ . The set of all nonnegative numbers is denoted R+ 0
and equals R+ ∪ {0}. A real number x is said to be negative if it has a
representative Cauchy sequence (xn )n∈N ∈ Q+ consisting entirely of negative
rational numbers. We denote this by x < 0 or x ∈ R− . The set of all
negative real numbers is denoted R− . The set of all nonpositive
numbers is denoted R− −
0 and equals R ∪ {0}. Theabsolute value function
+
is the function | · | : R → R0 is given by
(
x, if x ≥ 0
|x| =
−x, if x < 0

Lastly, the distance d : R × R → R+ 0 between two real numbers x and y is

given by
d(x, y) = |x − y|

17
 We can also embed the rational numbers Q into R by considering the
representative Cauchy sequences (x, x, . . . ) in Q for all x ∈ Q. We define
f : Q ,→ C/ ∼ by
f (x) = [(x, x, . . . )]
Clearly the field operations of Q are preserved by this function, by our
definitions of ≤, + and ·, so f is an order-isomorphism.

We now show formally that this construction of the reals yields a struc-
ture that is equivalent to the first two structures above:

Lemma 3.1 The operations of addition + and multiplication ·, as well as

the relation of partial order ≤ on R, are well defined. That is, if [(xn )n∈N ] =
[(x0n )n∈N ] and [(yn )n∈N ] = [(yn0 )n∈N ], then

[(xn )n∈N ] + [(yn )n∈N ] = [(x0n )n∈N ] + [(yn0 )n∈N ]

[(xn )n∈N ][(yn )n∈N ] = [(x0n )n∈N ][(yn0 )n∈N ]
[(xn )n∈N ] ≤ [(yn )n∈N ] =⇒ [(x0n )n∈N ] ≤ [(yn0 )n∈N ]

Proof : If [(xn )n∈N ] = [(x0n )n∈N ] and [(yn )n∈N ] = [(yn0 )n∈N ], then (xn )n∈N ∼
(x0n )n∈N and (yn )n∈N ∼ (yn0 )n∈N , so that limn→∞ (xn − x0n ) = limn→∞ (yn −
yn0 ) = 0, that is,
 
∀ ∈ Q+ , ∃N ∈ N such that n ≥ N =⇒ |xn − x0n | < and |yn − yn0 | <
2 2
whence
 
|(xn +y+n)−(x0n +yn0 )| = |(xn −x0n )+(yn −yn0 )| ≤ |(xn −x0n )|+|(yn −yn0 )| < + = 
2 2
or limn→∞ (xn +yn )−(x0n +yn0 ) = limn→∞ (xn −x0n )+(yn −yn0 ) = 0+0 = 0.

Consequently, (xn + yn )n∈N ∼ (x0n + yn0 )n∈N , whence

[(xn )n∈N ]+[(yn )n∈N ] := [(xn +yn )n∈N ] = [(x0n +yn0 )n∈N ] =: [(x0n )n∈N ]+[(yn0 )n∈N ]

Likewise, ∀ ∈ Q+ , ∃N ∈ N such that given any n ≥ N , if we define

M = max{1, |yn |, |x0n |}, we have
 
|xn − x0n | < and |xn − x0n | <
2M 2M

18
whence

|xn yn − x0n yn0 | = |xn yn − x0n yn + x0n yn − x0n yn0 |

= |yn (xn − x0n ) + x0n (yn − yn0 )|
≤ |yn (xn − x0n )| + |x0n (yn − yn0 )|
= |yn ||xn − x0n | + |x0n ||yn − yn0 |
≤ M |xn − x0n | + M |yn − yn0 |
 
< M +M
2M 2M
= 

or limn,m→∞ (xn yn − x0n yn0 ) = 0, so that (xn yn )n∈N ∼ (x0n yn0 )n∈N , whence

[(xn )n∈N ][(yn )n∈N ] := [(xn yn )n∈N ] = [(x0n yn0 )n∈N ] =: [(x0n )n∈N ][(yn0 )n∈N ]

Finally, suppose [(xn )n∈N ] ≤ [(yn )n∈N ]. If (xn )n∈N ∼ (yn )n∈N , then we
have by our assumptions (x0n )n∈N ∼ (xn )n∈N and (xn )n∈N ∼ (yn )n∈N and
(yn )n∈N ∼ (yn0 ), so by the transitivity of ∼ we have (x0n )n∈N ∼ (yn0 )n∈N ,
whence
[(x0n )n∈N ] ≤ [(yn0 )n∈N ]
If, on the other hand, ∃N ∈ N such that n ≥ N =⇒ xn < yn , the we
must have that ∃N 0 ∈ N such that n ≥ N 0 =⇒ x0n < yn0 , for otherwise,
if for all N 0 ∈ N there are n ≥ N 0 for which x0n ≥ yn0 , then since since
(xn )n∈N ∼ (x0n )n∈N and (yn )n∈N ∼ (yn0 )n∈N , we have for all  ∈ Q+ we have
|xn − x0n | <  and |yn − yn0 | < , so that xn > x0n −  and yn < yn0 + , and
hence
yn ≤ yn0 ≤ x0n ≤ xn
contradicting xn < yn for all n ≥ N . Thus in this case, too, we must have
[(x0n )n∈N ] ≤ [(yn0 )n∈N ]. 

Lemma 3.2 Let 0 ∈ R be given by [(00 , 00 , · · · )], the equivalence class rep-
resented by (00 , 00 , . . . ) in Q. Then for all x ∈ R\{0} there exists an  ∈ Q+
such that exactly one of the following holds for all representatives (xn )n∈N
in Q of x:

1. xn >  for all but finitely many n ∈ N.

2. xn < − for all but finitely many n ∈ N.

In particular, for any x ∈ R\{0} we can obtain a new sequence (pn )n∈N
from any representative sequence (xn )n∈N of x by discarding finitely many

19
terms of (xn )n∈N , with the property that either pn >  for all n ∈ N or else
pn < − for all n ∈ N.

Proof : Suppose not, that is suppose for all  ∈ Q+ there is a representative

(xn )n∈N of x ∈ R\{0} for which xn ≤  for infinitely many n ∈ N and such
that xn ≥ − for infinitely many n ∈ N. Then, in particular for 2 ∈ Q+
there is a representative (xn )n∈N of x such that |xn | ≤ 2 for infinitely many
terms. As a consequence, for any other representative (yn )n∈N of x, because
(yn )n∈N ∼ (xn )n∈N , we have that ∃N ∈ N such that n ≥ N =⇒ |yn −xn | <

2 . Consequently,

 
|yn | = |yn − xn + xn | ≤ |yn − xn | + |xn | ≤ + = (3.1)
2 2
Now, either we have for all ε ∈ Q+ that there are infinitely many yn in
[−ε, ε] or else there exists some ε0 ∈ Q+ such that [−ε0 , ε0 ] contains only
finitely many yn . The second case is impossible because of (3.1), so we
proceed to the first case: let us denote by (ynk )k∈N the subsequence of the
sequence lying in [−ε, ε]. Because (yn )n∈N is a Cauchy sequence, we have
that ∃N ∈ N such that m, nk ≥ N =⇒

|ym | ≤ |ym − ynk | + |ynk | < 2ε

Since this is true for all ε ∈ Q+ , we have that limn→∞ yn = 0, so that

(yn )n∈N ∈ [(00 , 00 , . . . )] = 0, contrary to assumption. Thus, the assumption
that for all  ∈ Q+ there is a representative (xn )n∈N of x ∈ R\{0} for
which xn ≤  for infinitely many n ∈ N must be wrong, and we must have
only finitely many terms of the sequence (xn )n∈N lying in (−, ). Since
(xn )n∈N is a Cauchy sequence, infinitely many terms satisfy either xn > 
or xn < − for some  ∈ Q+ . Discarding these terms gives us our desired
strictly positive or strictly negative sequence (pn )n∈N in Q, as the case may
be, that represents x. 

20
Theorem 3.3 The real numbers are a structure (R, ≤, −,−1 , +, ·, 0, 1) which
is a field:

1. (R, −,−1 , +, ·, 0, 1) is a field, that is for all x, y, z ∈ R we have


1. x + (y + z) = (x + y) + z (associativity of +) 

2. x+y =y+x (commutativity of +)

 (R, +) is an
3. ∃0 ∈ R such that x + 0 = x 
 abelian group

4. ∀x ∈ R, ∃! − x ∈ R such that x + (−x) = 0 
(associativity of ·)

5. x(yz) = (xy)z 

6. xy = yx (commutativity of ·)  (R\{0}, ·) is an


7. ∃1 ∈ R such that 1x = x and 0 6= 1 

 abelian group
∀x ∈ R\{0}, ∃x−1 ∈ R such that xx−1 = 1

8. 
9. x(y + z) = xy + xz (distributivity of · over +)

2. (R, ≤) is a totally ordered set, that is for all x, y, z ∈ R we have

1. x≤x (reflexivity)
2. x ≤ y and y ≤ x =⇒ x = y (antisymmetry)
3. x ≤ y and y ≤ z =⇒ x ≤ z (transitivity)
4. x ≤ y or y ≤ x (totalness)

3. The field operations + and · are compatible with the partial order
≤ in the sense that for all x, y, z ∈ R we have

1. x ≤ y =⇒ x + z ≤ y + z (preservation of order under addition)

2. 0 ≤ x and 0 ≤ y, then 0 ≤ xy (preservation of order under multiplication)

Proof : 1.1, 2, 5, 6, and 9 follow from the properties of Q, the definition

of Cauchy sequence, and the lemma, which guarantees the well-definedness
of +, · and ≤. We’ll demonstrate 1.1., the others follow similarly: for

21
x, y, z ∈ R, we have

x + (y + z) = [(xn )n∈N ] + [(yn )n∈N + (zn )n∈N ]

= [(xn )n∈N ] + [(xn + yn )n∈N ]
= [(xn )n∈N + (xn + yn )n∈N ]


= xn + (yn + zn ) n∈N


= (xn + yn ) + zn n∈N
= [(xn + yn )n∈N + (zn )n∈N ]
= [(xn + yn )n∈N ] + [(zn )n∈N ]
= [(xn )n∈N + (yn )n∈N ] + [(zn )n∈N ]
= (x + y) + z

1.3. We define 0 ∈ R as the equivalence class of Cauchy sequences in

Q converging to 00 ∈ Q, and represented by 0 = (00 , 00 , . . . ) in Q. This
equivalence class satisfies x + 0 = 0 + x = x for all x = [(xn )n∈N ]:

x + 0 = [(xn )n∈N ] + [0] = [(xn + 00 )n∈N ] = [(xn )n∈N ] = x

and likewise 0 + x = x.

1.4. For each x = [(xn )n∈N ] ∈ R we define its additive inverse −x to be

[(−xn )n∈N ], for then

x + (−x) = [(xn )n∈N ] + [(−xn )n∈N ] = [(xn + (−xn ))n∈N ] = [(00 , 00 , . . . )] = 0

Clearly −x exists whenever x does, so each x ∈ R has an additive inverse.

1.7. We define 1 ∈ R as the equivalence class of Cauchy sequences in

Q converging to 10 ∈ Q, which is represented by 1 = (10 , 10 , . . . ) in Q. This
equivalence class satisfies x1 = 1x = x for all x = [(xn )n∈N ]:

x1 = [(xn )n∈N ] + [1] = [(xn 10 )n∈N ] = [(xn )n∈N ] = x

and likewise 1x = x.

1.8. For each x ∈ R\{0}, let (xn )n∈N be one of the Cauchy sequences
consisting entirely of nonzero rational numbers, guaranteed to exist by L
3.2. We define x’s multiplicative inverse x−1 to be the equivalence class
represented by this sequence
  
−1 1
x :=
xn n∈N

22
This definition indeed gives x’s inverse, for we have
         
−1 1 1 1
xx = [(xn )n∈N ] = (xn )n∈N = xn = [(10 , 10 , . . . )] = 1
xn n∈N xn n∈N xn n∈N

2.1. Clearly, for all x ∈ R, because x is represented by a Cauchy sequence

(xn )n∈N , and because (xn )n∈N ∼ (xn )n∈N , we have that x ≤ x.

2.2. If x ≤ y and y ≤ x, it is not possible that, for representatives

(xn )n∈N and (yn )n∈N of x and y, respectively, ∃N ∈ N such that n ≥ N =⇒
xn < yn and yn < xn simultaneously, so we must have (xn )n∈N ∼ (yn )n∈N ,
in which case x = [(xn )n∈N ] = [(yn )n∈N ] = y.

2.3. Let x ≤ y and y ≤ z, and let (xn )n∈N , (yn )n∈N and (zn )n∈N be
representatives of x, y and z, respectively. There are four cases: 1. If
∃N1 , N2 ∈ N such that n ≥ N1 =⇒ xn < yn and n ≥ N2 =⇒ yn < zn ,
then ∃N = max{N1 , N2 } such that n ≥ N =⇒ xn < yn < zn , whence
xn < zn by the transitivity of < in Q, and therefore x ≤ z. 2. If, however,
∃N ∈ N such that n ≥ N =⇒ xn < yn and (yn )n∈N ∼ (zn )n∈N , then
∀ ∈ Q+ , ∃N 0 ∈ N such that n ≥ N 0 =⇒ |yn −zn | < , so that yn < zn +,
whence for M = max{N, N 0 } we have n ≥ M =⇒ xn < yn < zn +. Since
this is true for all  ∈ Q+ , we have that xn < yn ≤ zn , so by transitivity
we have that ∃M = max{N, N 0 } ∈ N such that n ≥ N =⇒ xn < zn .
Thus, x ≤ z in this case too. 3. Similarly, if (xn )n∈N ∼ (yn )n∈N and
∃N ∈ N such that n ≥ N =⇒ yn < zn so that ∀ ∈ Q+ , ∃N 0 ∈ N such
that n ≥ N 0 =⇒ |xn − yn | < , so that xn < yn + . Consequently,
for M = max{N, N 0 } we have n ≥ N =⇒ xn < yn +  < zn + ,
whence xn < yn +  ≤ zn , so that xn < yn , and hence x ≤ z in this case
too. 4. Finally, if (xn )n∈N ∼ (yn )n∈N and (yn )n∈N ∼ (zn )n∈N , then by the
transitivity of ∼ we have (xn )n∈N ∼ (zn )n∈N , and so x ≤ z. Thus in all cases
we have x ≤ y and y ≤ z =⇒ x ≤ z.

2.4. Let x, y ∈ R be given. If x = [(xn )n∈N ] = [(yn )n∈N ] = y, then

(xn )n∈N ∼ (yn )n∈N , so x ≤ y (and y ≤ x). If, on the other hand, x 6= y, then
(xn )n∈N 6∼ (yn )n∈N , so there must exist an  ∈ Q+ such that ∀N ∈ N we have
n ≥ N =⇒ |xn − yn | ≥ . Consequently, there exists an N ∈ N such that
for all n ≥ N we have xn < yn or yn < xn (we clearly cannot have both,
for (xn )n∈N and (yn )n∈N are Cauchy sequences, and having xn < yn and
ym < xm for arbitrarily large n and m would mean limn→∞ (xn − yn ) = 0).
If xn < yn , then x < y, while if xn > yn , then x > y.

23
 3.1. If x = [(xn )n∈N ] ≤ [(yn )n∈N ] = y, suppose ∃N ∈ N such that
n ≥ N =⇒ xn < yn . Then xn + zn < yn + zn by the properties of <
in Q, so x + z ≤ y + z. If, on the other hand, (xn )n∈N ∼ (yn )n∈N , then
x = [(xn )n∈N ] = [(yn )n∈N ] = y, so clearly x + z = y + z, so in particular
x + z ≤ y + z.

3.2. Suppose 0 ≤ x = [(xn )n∈N ] and 0 ≤ y = [(yn )n∈N ]. If x = 0 and

y = 0, then clearly

xy = [(00 , 00 , . . . )][(00 , 00 , . . . )] = [(00 00 , 00 00 , . . . )] = [(00 , 00 , . . . )] = 0

If x, y > 0, then by L 3.2 there exist 1 , 2 ∈ Q+ and representative Cauchy

sequences (xn )n∈N and (yn )n∈N , respectively, such that xn , yn >  = min{1 , 2 }
for all n ∈ N. Consequently, xn yn >  for all n ∈ N, whence xy =
[(xn yn )n∈N ] > 0. 

Theorem 3.4 (R is Complete) Every nonempty subset S of R that is

bounded above has a least upper bound, sup S = min S u .

Proof : Let u be an upper bound of S, that is x ≤ u for all x ∈ S. If u is not

rational, i.e. if u is not the image of a rational number under the embedding
f : Q → R considered above, then we may choose another, larger u which
is. So assume that u is a rational upper bound of S. Since S is nonempty,
let s ∈ S and let L be a rational number satisfying L < s. Then, define
two rational sequences (ln )n∈N and (un )n∈N as follows: l1 = L, u1 = u,
and mn = (ln + un )/2 for all n ∈ N. If mn ∈ S u , then let un+1 = mn and
ln+1 = ln , while if m ∈/ S u , then let un+1 = un and ln+1 = mn . By induction,
un ∈ S u and ln ∈ / S u for all n ∈ N . Considering the unique rationals
un = f (un ), where f is our embedding, we have that u = [(u0n )n∈N ] is the
0 −1

least upper bound of S, because clearly s ≤ u for all s ∈ S, and moreover

limn→∞ (ln − un ) = 0, so u = [(lu0 )n∈N ] as well, where ln0 = f −1 (ln ), so that
if b < u, then bn < ln for all n ≥ N for some N ∈ N, whence b is not an
upper bound. 

24
4 Equivalence of All Constructions of R

Having made the effort to abstract the properties of the integers, the ratio-
nals and the real numbers, we are now in a position to reap their fruits: we
can show that all constructions of R are isomorphic, that is are in a one-to-
one correspondence which preserves their algebraic properties (addition and
multiplication and their associated properties), their order properties, and
their least upper bound properties. That is if R and R0 are two construc-
tions, then for all a, b ∈ R and a0 , b0 ∈ R0 with a ↔ a0 and b ↔ b0 we have
a + b ↔ a0 + b0 , ab ↔ a0 b0 and a ≤ b ⇐⇒ a0 ≤0 b0 , while for all S ⊆ R and
S 0 ⊆ R0 with S ↔ S 0 we have sup S ↔ sup S 0 . Formally,

Theorem 4.1 Every complete totally ordered field is both isomorphic and
order-isomorphic to R, so in this sense all constructions of R are equivalent.
Proof : In what follows, suppose R is any construction of the reals satisfying
0
the axioms given in Construction 1 of above, and let (F, ≤0 , −0 , −1 , +0 , ·0 , 10 , 00 )
be any other construction of R, by which we here mean any complete totally
ordered field, so that F is an ordered field that satisfies the least upper bound
property. By basic properties of rings, we know there a monomorphism (in-
jective ring homomorphism) f : Q ,→ F which is also an order-embedding.
We can extend this function to an embedding of R into F as follows: for
each r ∈ R let
Dr = {q ∈ Q | q < r}
be the associated Dedekind cut. Since Dr is nonempty and bounded above in
Q, we have that f (Dr ) is nonempty and bounded above in F , so applying the
assumed least upper bound property of F we define the function g : R → F
by
g(r) = sup f (Dr )
Then g is also a monomorphism and order-embedding, since if r, s ∈ R, then
g(r + s) = sup f (Dr+s ) = sup f (Dr ) +0 sup f (Ds ) = g(r) +0 g(s)
g(rs) = sup f (Drs ) = sup f (Dr ) · sup f (Ds ) = g(r)g(s)
r ≤ s =⇒ g(r) = sup f (Dr ) ≤ sup f (Ds ) = g(s)
g(r) = g(s) =⇒ r = s
and clearly g is an extension of f . We’ll prove the second equation, since
the other three follow similarly: if
x ≤0 sup f (Drs )

25
then x ≤0 a for any a ∈ f (Drs )u , so in particular x ≤0 bc for all b ∈ f (Dr )u ,
c ∈ f (Ds )u , whence
x ≤0 sup f (Dr ) ·0 sup f (Ds )
Conversely if
x ≤0 sup f (Dr ) ·0 sup f (Ds )
then x ≤0 bc for all b ∈ f (Dr )u , c ∈ f (Ds )u . Now, for any

a ∈ f (Drs )u \{sup f (Drs )}

there exist such b and c that also satisfy bc ≤ a, as follows: if

a ∈ f (Drs )u \{sup f (Drs )}

then sup f (Drs ) <0 a, so by the order denseness of Q we can always find a
p ∈ Q such that
sup f (Drs ) <0 b = f (p) <0 a
and then we can pick c to be
sup f (Drs ) + a
c=
2b
Then we have sup f (Drs ) <0 bc <0 a. Because we can always find such b and
c we must have
x ≤0 sup f (Drs )

We have just shown that g(R) ⊆ F and R ∼ = g(R). If we can show

that g(R) ⊃ F , we will have proved that g is surjective, and therefore
bijective, and so an isomorphism (both a ring isomorphism and an order-
isomorphism), and we will have thereby finished the proof. Towards this
end, note first that F also satisfies the Archimedean property, which follows
via T ?? from the fact that R is order-embedded into F and R satisfies the
Archimedean property. Consequently, for any k ∈ F there exists an n ∈ N,
and so a g(n) ∈ F , such that −g(n) < k < g(n). Let

Dk = {r ∈ R | g(r) < k}

and note that because g(R) ∼= R we have that we have that Dk and g(Dk )
are both nonempty and bounded above, so that ∃ sup Dk ∈ R, whence
∃g(sup Dk ) ∈ F , and also ∃ sup g(Dk ) ∈ F , and since g(R) ⊆ F we have
that

sup g(Dk ) ≤ g(sup Dk ) (4.1)

26
by the least upper bound property of F . We claim that

sup g(Dk ) = g(sup Dk ) (4.2)

Suppose not, that is suppose sup g(Dk ) < g(sup Dk ). Then, by the Archimedean
property of F , there is some n ∈ N such that

0 = g(sup Dk ) − g(sup Dk ) < g(n)−1 < g(sup Dk ) − sup g(Dk ) (4.3)

This implies that

sup g(Dk ) < g(sup Dk ) − g(n)−1 < g(sup Dk ) (4.4)

But then by (4.1), the second inequality in (4.4), and the definition of
g(sup Dk ) and sup g(Dk ), the first of which implies that g −1 (g(sup Dk ) −
g(n)−1 ) = sup Dk − n ∈ Dk , we have

g(sup Dk ) − g(n)−1 < sup g(Dk )

which by the first inequality of (4.4) implies the contradiction

sup g(Dk ) < sup g(Dk )

Hence we must have sup g(Dk ) = g(sup Dk ), and so g(k) ∈ g(R), whence

g(R) ⊃ F

and consequently g(R) = F . 

27
5 Further Properties of R

5.1 Exponentiation

If a ∈ R+
0 and n ∈ N, then we define the nth power of a as

n times
n z }| {
a := a · a · · · a

This defines a real number, as we show in Theorem 5.5 below. We define a0

as
a0 = 1
for all a ∈ R (including 0), and we define a−n for a ∈ R+ as
1 1
a−n := n
=
a | · a{z· · · a}
a
n times

The number ±n ∈ Z in the above three cases is called the integer exponent
of a. Before we can define nth roots, however, we first need to show that
they are real numbers! Theorem 5.8 below guarantees this fact, and this
fact guarantees that the expression

a1/n
√
also denoted n a, is well defined and represents a unique real number. The
√
notation n a is in keeping with that of the square root of a, which is denoted
√
both as a and as a1/2 . Recall that one of the shortcomings √ of the ratio-
nal numbers was that they did not contain the number 2. √ Theorem 5.8,
therefore, remedies this problem, and ensures not only that 2, but 2 1/n for
all n ∈ N, is a real number. This holds not only for 2, of course, but for all
a ∈ R+0.

The expression a−1/n is defined in the same manner as above,

1
a−1/n :=
a1/n
and is also well defined by Theorem 5.5. Finally, since a1/n is a (nonnegative)
real number, we can define the fractional exponent am/n for m/n ∈ Q,
which we do as follows:
am/n := (am )1/n

28
We will show below that, since am/n ∈ R for all m/n √
∈ Q, we also have
that ab ∈ R for all b ∈ R. Thus, expressions like π e 2 are well defined real
numbers. This will require considerable machinery to develop, in particular
a proper definition of the exponential and logarithmic functions x
√ e and√ ln x.
b b
Then we will define a by a := e b ln a , so that for example π e 2 = e 2 ln π .
e

5.2 Basic Arithmetic and Order Properties of R

Lemma 5.1 The numbers 0 and 1 in R are unique.

Proof : We know there is at least one 0 in R, by Axiom 1.3, and it satisfies

x + 0 = x for all x ∈ R. Let us suppose there were a second zero, 00 ∈ R,
satisfying x + 00 = x for all x ∈ R. Then, letting 00 play the role of x we
have
00 = 0 + 00
while letting 0 play the role of x gives

0 + 00 = 0

Therefore, 00 = 0 + 00 = 0. This shows that there is no second zero, for it

would be equal to the first.

Similarly, we know there is at least one 1 in R, by Axiom 1.7, and it is

characterized by 1x = x for all x ∈ R. If we suppose there were another
10 with this property, then by letting each of 1 and 10 play the role of x we
have
1 = 110 = 10 

Lemma 5.2 (Cancellation Law) For any x, y, z ∈ R we have

x+z =y+z =⇒ x = y

and
x+z ≤y+z =⇒ x ≤ y
Similarly, for z 6= 0,
xz = yz =⇒ x = y

29
Proof : Suppose first x + z = y + z. Then, using Axioms 1.1, 3, and 4,
x = x+0
= x + (z + (−z))
= (x + z) + (−z)
= (y + z) + (−z)
= y + (z + (−z))
= y+0
= y
Suppose next that x + z ≤ y + z, and use Axioms 1.1, 3, 4 and 3.1:
x = x+0
= x + (z + (−z))
= (x + z) + (−z)
≤ (y + z) + (−z)
= y + (z + (−z))
= y+0
= y
Finally, suppose xz = yz and z 6= 0. Then, using Axioms 1.6-8,
x = 1x = x1 = x(zz −1 ) = (xz)z −1 = (yz)z −1 = y(zz −1 ) = y1 = 1y = y 

Lemma 5.3 Axiom 1.9 has a right version,

(x + y)z = xz + yz
Proof : Using Axioms 1.6, 9 we have
(x + y)z = z(x + y) = zx + zy = xz + yz 

Lemma 5.4 0x = 0 and (−1)x = −x for all x ∈ R.

Proof : First, observe that by Axioms 1.3, 7 and the previous Lemma,
x + 0 = x = 1x = (1 + 0)x = 1x + 0x = x + 0x
which by the Cancellation Law means 0 = 0x. Using this result, Axioms
1.3, 7 and the previous Lemma, we get
(−1)x + x = (−1)x + 1x = ((−1) + 1)x = 0x = 0 = (−x) + x
which by the Cancellation Law means (−1)x = −x. 

30
Theorem 5.5 (Arithmetic Properties of R) For all x, y, z, w ∈ R, the
following hold:
z w zw
1. −(−x) = x 9. · = if x, y ∈ R\{0}
x y xy
x z xw + yz
2. x(−y) = −(xy) = (−x)y 10. + = if y, w ∈ R\{0}
y w yw
1
3. (−1)x = −x 11. x 6= 0 =⇒ 6= 0
x
1 y
4. xy = x =⇒ y = 1 if x 6= 0 12. x = if x, y ∈ R\{0}
y x
x x/y xw
5. = 1 if x 6= 0 13. = if x, y, z ∈ R\{0}
x z/w yz
x zx x
6. =x 14. =z if y ∈ R\{0}
1 y y
−x x x

7. xy ∈ R\{0} if x, y ∈ R\{0} 15. = =− if y ∈ R\{0}
y −y y
1 1 1
8. · = if x, y ∈ R\{0}
x y xy
Proof : (1) Every x possesses a negative, −x, by Axiom 1.4, and it satisfies
x + (−x) = 0
What is the negative of −x? Let us use the Cancellation Law:
x + (−x) = 0 = −(−x) + (−x) =⇒ x = −(−x)

(2) x(−y) + xy = x((−1)y) + xy = (x(−1))y + xy = ((−1)x)y + xy =

(−x)y + xy = ((−x) + x)y = 0y = 0 = −(xy) + xy, so the Cancellation Law
implies x(−y) = −(xy) Similarly with (−x)y = −(xy).

(3) This was Lemma 5.4.

(4) This was Lemma 5.1.

(5) This is an Axiom, 1.8, so I don’t know what it’s doing in a theorem.

(6) Follows from the fact that 1−1 = 1 on account of 1 · 1 = 1 = 1 · 1−1

and the Cancellation Law, Lemma 5.2.

(7) If xy = 0, then either x or y must equal 0. Suppose y 6= 0, and

observe that
x = 1x = x1 = x(yy −1 ) = (xy)y −1 = 0y −1 = 0

31
Reversing the roles of x and y shows that if x 6= 0, the y = 0. It is of course
possible that both x and y equal 0. All this shows that if xy = 0 then x = 0
or y = 0, which is the contrapositive of x 6= 0 and y 6= 0 implies xy 6= 0.

(8) This is the statement that x−1 y −1 = (xy)−1 . We already know that
if x, y 6= 0, then xy 6= 0, so all of x, y, and xy have multiplicative inverses.
Now observe that

(xy)(x−1 y −1 ) = y(xx−1 ) = yy −1 = 1 = (xy)(xy)−1

so by the Cancellation Law we must have x−1 y −1 = (xy)−1 . I skipped a

couple of commutations and associations in the above line, FYI.

(9) This is just the associative and commutative laws, Axioms 1.5-6, plus
(8) above:

z w
· = (zx−1 )(wy −1 ) = z(x−1 w)y −1 = z(wx−1 )y −1
x y
zw
= (zw)(x−1 y −1 ) = zw(xy)−1 =
xy

(10) We’ll use Axioms 1.5-9: since y 6= 0 and w 6= 0, y −1 = 1

y and
w−1 = w1 exist, so
x z
+ = (xy −1 ) + (zw−1 )
y w
= (xy −1 )1 + (zw−1 )1
= (xy −1 )(ww−1 ) + (zw−1 )(yy −1 )
= x(y −1 w)w−1 + z(w−1 y)y −1
= x(wy −1 )w−1 + z(yw−1 y −1
= (xw)(y −1 w−1 ) + (zy)(w−1 y −1 )
= (xw)(y −1 w−1 ) + (zy)(y −1 w−1 )
= (xw + zy)(y −1 w−1 )
xw + yz
=
yw

(11) If x 6= 0, then x−1 exists and xx−1 = 1. But 1 6= 0 by Axiom 1.7,

so by the proof of (8) above we must have x−1 6= 0.

32
 (12) This just says that (xy −1 )−1 = x−1 y, which follows from the fact
that (y −1 )−1 = y in the same way that −(−y) = y (as in (1) above).

(13) Combining (9) and (12) above, we get

x/y x w xw
= · =
z/w y z yz

(14) This is nothing but associativity and commutativity, Axioms 1.5-6:

zx x
= (xz)y −1 = (zx)y −1 = z(xy −1 ) = z
y y

(15) This is nothing but (2) applied to −(xy −1 ). 

33
Theorem 5.6 (Order Properties of R) For all x, y, z, w ∈ R, the fol-
lowing hold:

1. x > y and z > w =⇒ x + z > y + w

2. x, y > 0 =⇒ x + y > 0 and xy > 0
3. x > 0 ⇐⇒ −x < 0
4. x > y ⇐⇒ −x < −y
5. x > y and z < 0 =⇒ xz < yz
6. x 6= 0 =⇒ x2 > 0
7. −1 < 0 < 1
8. xy > 0 ⇐⇒ either x, y > 0 or x, y < 0
9. x > 0 =⇒ x1 > 0
10. x > y > 0 =⇒ x1 < y1
x+y
11. x < y =⇒ x < <y
2
Proof : (1) Suppose x > y and z > w. By Axiom 3.1, we have x + z > y + z
and y + z > y + w. The transitivity of > then shows that x + z > y + w.

(2) Let x, y > 0 and observe that Axioms 1.3 and 3.1 give x+y > 0+y =
y, so since y > 0 transitivity of > ensures x + y > 0. Axiom 3.1 also ensures
xy ≥ 0. To see that xy > 0 when x, y > 0, note that xy = 0 only if x = 0 or
y = 0 (as in the proof of Theorem 5.5, (7)).

(3) Suppose x > 0 and use Axioms 1.3 and 3.1:

0 = x + (−x) > 0 + (−x) = −x
Conversely, if 0 > −x, then adding x to both sides gives
x = x + 0 > x + (−x) = 0

(4) Suppose x > y and add −x to both sides and then −y to both sides,
making use of Axiom 3.1:
x > y ⇐⇒ 0 = x + (−x) > y + (−x)


 −y = (−y) + 0

> (−y) + (y + (−x))




⇐⇒ = ((−y) + y) + (−x)

= 0 + (−x)





= −x


34
 (5) Let x > y and 0 > z. Then by (3) above we have 0 < (−z), and
hence Axioms 1.9, 3.1-2 and (1)-(2) of Theorem 5.5 apply to give

x + (−y) > 0 and (−z) > 0 =⇒ −(zx) + zy = (−z)(x + (−y)) > 0

which means, again by Axiom 3.1, zy > zx, and thus by commutativity
(Axiom 1.6) yz > xz.

(6) Let x 6= 0 and consider x2 . If x > 0, then Axiom 3.2 gives x2 >
x · 0 = 0 (the last equality by Lemma 5.4), while if x < 0, then −x > 0 by
(3) above. Now, observe that for any x ∈ R we have by associativity (Axiom
1.5), Lemma 5.4 and Theorem 5.5

(−1)2 x = (−1)((−1)x) = (−1)(−x) = −(−x) = x

so Lemma 5.1 applies to tell us that (−1)2 = 1. We use this in combination

with −x > 0 as follows:

x2 = 1x2 = (−1)2 x2 = ((−1)x)2 = (−x)2 > (−x)0 = 0

the last inequality following from Axiom 3.2 and the last equality from
Lemma 5.4.

(7) We know from Axiom 1.7 that 0 6= 1, while from (6) above we have

1 = 12 > 0

and from (3) above we consequently have 0 > −1.

(8) Suppose xy > 0. We know that x, y 6= 0, else xy = 0 (see proof of (7)

in Theorem 5.5), hence x > 0 or x < 0 and similarly with y. Suppose first
x > 0. If y < 0, then (5) above gives xy < y · 0 = 0, which is in contradiction
with our assumption, so we must have y > 0. Reversing the roles of x and y
shows that if y > 0, then x > 0. If, on the other hand, x < 0, then y < 0, for
otherwise, if y > 0, (5) above gives xy < 0y = 0, a contradiction. Similarly
y < 0 implies x < 0.

(9) Let x > 0 and observe that by (7) above we have x−1 x = 1 > 0, so
(8) applies to give x−1 > 0.

(10) Let x > y > 0. By transitivity of > we have x > 0, and hence by
(9) we conclude that x−1 , y −1 > 0, which means 0 < 1 = x−1 x < x−1 y and
hence 0 < y −1 = y −1 1 < x−1 yy−1 = x−1 1 = x−1 , using Axioms 1.5, 7, 8.

35
 (11) Suppose x < y and consider x+y 1
2 = 2 (x + y) = x/2 + y/2. Now,
0 < 1 implies 1 = 1 + 0 < 1 + 1 = 2, by Axiom 3.1, so that by (8) and
(10) above we get 0 < 1/2 < 2/2 = 1. Multiplying through by x and using
Axiom 3.2 we conclude x/2 < x1 = x if x > 0, and x/2 > x if x < 0 by
(5) above. If x > 0, then by transitivity y > 0 too, and we conclude that
y/2 < y as well, and x < y implies x/2 < y/2. This means

2 1 1 1
x = 1x = x = (1 + 1)x = (1x + 1x) = (x + x)
2 2 2 2
x x x y y y
= + < + < + =y
2 2 2 2 2 2
where the last equality follows from similar reasoning as the LHS for x. If
x < 0, then y < 0, too, and we can apply the above argument to 0 < −x <
−y and conclude that
x+y
−x < − < −y
2
from which it follows (by (3) and (5)) that
x+y
x< <y 
2

Theorem 5.7 For all x, y, x1 , . . . , xn ∈ R and all n ∈ N we have

1. |x + y| ≤ |x| + |y| (Triangle Inequality)

2. |x1 + x2 + · · · + xn | ≤ |x1 | + |x2 | + · · · + |xn |
3. |xy| = |x||y|

Proof : (1) This boils down to cases, which are four, x, y > 0, x, y < 0,
x > 0 and y < 0, and x < 0 and y > 0. If x, y > 0, then this is just an
identity
|x + y| = x + y = |x| + |y| ≤ |x| + |y|
If x, y < 0, then

|x + y| = −(x + y) = (−x) + (−y) = |x| + |y| ≤ |x| + |y|

If, on the other hand, say x < 0 and y > 0, then x < x + y < y, so

|x + y| < max{|x|, |y|} < |x| + |y|

Similarly with x > 0 and y < 0.

36
 (2) Follows inductively: We have |x1 + x2 | ≤ |x1 | + |x2 | by (1) above, so
suppose |x1 + x2 + · · · + xn | ≤ |x1 | + |x2 | + · · · + |xn | for some n ≥ 2 and
consider the (n + 1)st case:

|(x1 + x2 + · · · + xn ) + xn+1 | ≤ |x1 + |x2 + · · · + xn | + |xn+1 |

≤ |x1 | + |x2 | + · · · + |xn | + |xn+1 |

where the first inequality is by (1) and the second by our induction hypoth-
esis.

(3) This again follows by cases. This is clearly true for x, y > 0 because
each side is their positive product. For x, y < 0, we have

|xy| = xy = 1xy = (−1)2 xy = ((−1)x)((−1)y) = (−x)(−y) = |x||y|

The first equality follows from Theorem 5.6, (8), and the fourth equality
from commutativity and associativity, and the fifth from Lemma 5.4. For
x < 0 and y > 0, we have |x| = −x and |y| = y, so xy < 0 and

|xy| = −(xy) = (−x)y = |x||y|

by Theorem 5.5, (2) and Theorem 5.6, (5). Similarly with x > 0 and y < 0.


Theorem 5.8 (Existence-Uniqueness of the n-th Root) For any a ∈

R+ and any n ∈ N there exists a unique n-th root, a real number b > 0 such
that bn = a.

Proof : The proof is totally dependent on the least upper bound property
of R. Let A = {x ∈ R+ | xn ≤ a}. Then A is bounded above (by 1 if a ≤ 1,
and by a if a ≥ 1), so that b = sup A exists. We suppose bn 6= a by way of
supposing bn < a and bn > a in turn, and derive contradictions, to conclude
that bn = a. First, suppose bn < a and let  = a − bn and h ∈ (0, 1] be given
by h = (1+b)n −bn . Then by the binomial theorem,

37
 n(n − 1) n−2 2
(b + h)n = bn + nbn−1 h + b h + · · · + nbhn−1 + hn
 2! 
n n−1 n(n − 1) n−2 n−2 n−1
= b + h nb + b h + · · · + nbh +h
2!
 
n(n − 1) n−2
≤ bn + h nbn−1 + b + · · · + nb + 1 + hbn − hbn
2!
 
n n n−1 n(n − 1) n−2
= b + h b + nb + b + · · · + nb + 1 − hbn
2!
= bn + h(b + 1)n − hbn
= bn + h[(b + 1)n − bn ]

= bn + [(b + 1)n − bn ]
(1 + b)n − bn
= bn + 
= a

But this shows that ∃c = b + h > b such that cn ≤ a, which contradicts the
assumption that c ≤ b, b being an upper bound. Similarly, if we suppose
bn > a, let  = bn − a, and h ∈ (0, 1] given by h = (1+b)n −bn . Then by the
binomial theorem,
n(n − 1) n−2 2
(b − h)n = bn − nbn−1 h + b h − · · · + (−1)n−1 nbhn−1 + (−1)n hn
 2! 
n n−1 n(n − 1) n−2 n−2 n−2 n−1 n−1
= b − h nb − b h + · · · + (−1) nbh + (−1) h
2!
 
n n−1 n(n − 1) n−2 n−2 n−1
≥ b − h nb + b h + · · · + nbh +h
2!
 
n(n − 1) n−2
≥ bn − h nbn−1 − b + · · · + nb + 1 + hbn − hbn
2!
 
n n n−1 n(n − 1) n−2
= b − h b + nb + b + · · · + nb + 1 + hbn
2!
= bn − h(b + 1)n + hbn
= bn − h[(b + 1)n − bn ]

= bn − [(b + 1)n − bn ]
(1 + b)n − bn
= bn − 
= a

But this shows that ∃c = b − h < b such that cn ≤ a, which contradicts the
assumption that b is the least upper bound for A – we have just found a

38
smaller one, c. Consequently, the assumption that b 6= a leads to contradic-
tions, and we must therefore have b = a. 

Theorem 5.9 (Exponent Laws in R) For all a, b ∈ R+ and p, q ∈ Q the

following hold:

1. ap aq = ap+q
2. (ap )q = apq
3. aq bq = (ab)q

If p, q ∈ N, then the theorem holds for all a, b ∈ R.

Proof : We prove this for m, n ∈ Z first, then for 1/m and 1/n, and finally
for p, q ∈ R.

1. a. First we prove this for m, n ∈ N by induction on m. For a fixed

n ∈ N and a ∈ R, we have by definition that an a1 = an a = an+1 , so the
case m = 1 is proved. Now suppose that an am = an+m , and note that
an am+1 = an (am a1 ) = (an am )a1 = am+n a1 = am+n+1 , so it holds for all
m ∈ N. For the case m = 0 note that an a0 = an 1 = an = aa+0 . Finally,
we extend this to m ∈ Z− by noting that a−n = 1/an by definition for any
n ∈ N, so for any n ∈ Z and −m ∈ Z− we have
1
an a−m = an
am
If m = n then clearly they cancel and an a−m = 1 = a0 = an−n = an−m . If
m < n, then
m
z }| { zn−m
n
}| {
n −m a a · aa···a
· ·
= an−m

a a = m =
a a ·
 ·
| {z }·a
m

and if m > n, then by the same argument we have an a−m = 1

am−n
=
a−(m−n) = an−m .

b. For a fixed n ∈ N and a ∈ R, we have (an )1 = an = an·1 , so

the statement holds for m = 1. If we suppose it holds for m ≥ 1, then
(an )m+1 = (an )m (an ) = amn an = amn+n = an(m+1) , so it holds for all
m ∈ N. We can generalize this to Z using the same procedure as in 1 above.

c. For a, b ∈ R we have a1 b1 = ab = (ab)1 by definition, so suppose the

statement holds for some m ≥ 1. Then am+1 bm+1 = am abm b = am bm ab =

39
(ab)m (ab) = (ab)m+1 , so it holds for all m ∈ N. We can generalize this to Z
using the same procedure as in 1 above.

2. a. If a ∈ R+ and m, n ∈ N, then let x = a1/n a1/m . By 1.a. and b.

above we have

n
m
xmn = (a1/m a1/n )n = (a1/m )mn (a1/n )mn = (a1/m )m (a1/n )n = an am = an+m

and therefore x = (xmn )1/mn = (an+m )1/mn = a(n+m)/mn = a1/n+1/m , or

a1/n a1/m = a1/n+1/m

b. If a ∈ R+ and m, n ∈ N, then let x = (a1/n )1/m . By 1.b. above we

have xm = a1/m , and xmn = a, so x = a1/mn .

c. If a, b ∈ R+ and m ∈ N, then let x = a1/m b1/m . Then by 1.a. above

we have xm = (a1/m b1/m )m = (a1/m )m (b1/m )m = ab, so x = (ab)1/m .

Combining parts 1 and 2 gives the result for p, q ∈ R. 

Theorem 5.10 If there is a real number x ≥ 0 which satisfies x < y for all
y > 0, then x = 0.

Proof : If x > 0, then x < x, which is impossible. 

Theorem 5.11 (Archimedean Property of R) For all x, y ∈ R+ there

is an n ∈ N such that y < nx.

Proof : Suppose not, that is suppose for all n ∈ N we have nx ≤ y. Then

the subset
S = {nx | x ∈ R+ and nx ≤ y for all n ∈ N}
of R is bounded above by y. But then by the order-completeness of R the
number s = sup S = min S u exists in R and satisfies s ≤ y. Now, since
x > 0, we have −x < 0 = s − s, or s − x < s, and s − x ∈/ S u , i.e. there
exists an n ∈ N such that nx ∈ S and nx > s − x. But then (n + 1)x > s,
which means s 6= sup S, a contradiction. 

40
Theorem 5.12 (Every Open Interval Contains a Rational Point) If
a < b for some a, b ∈ R, then there is a q ∈ Q such that a < q < b.

Proof : By the Archimedean property of R there exist numbers m, n, p ∈ N

such that n(b − a) > 1, m > na and p > −na. Combining these we get

−p < na < m (5.1)

We can of course choose −p and m to satisfy −p = r − 1 and m = r for

some r ∈ Z, so that
r − 1 ≤ na < r (5.2)
(5.2) (5.2)
Now, this gives na < r ≤ na + 1 < nb, or na < r < nb, which gives
r
a< <b
n
so let q = r/n ∈ Q. 

Corollary 5.13 (Every Open Interval Contains an Irrational Point)

If a < b for some a, b ∈ R, then there is an r ∈ R\Q such that a < r < b.
√ √
Proof : The open interval (a/ 2, b/ 2) contains a rational point q, so
1 1
√ a<q< √ b
2 2
√ √
whence a < q 2 < b, and q 2 is irrational. (We could also have proved this
directly, as above.) 

Theorem 5.14 For every x ∈ R there is a unique n ∈ Z such that x − 1 <

n ≤ x, called the integer part of x and denoted [x].

Proof : By the Archimedean property of R, for any x ∈ R we can find

M, N ∈ N such that M > x and N > −x, or

−N < x < M (5.3)

Now, consider the subset S = {n ∈ N | − N − 1 + n ≤ x} of N. By (5.3) we

know that 1 ∈ S but M + N + 1 ∈ / S, so S is a nonempty subset of N. By
the well-ordering property of N we know that N\S has a smallest element
n + 1. Thus, n ∈ S but n + 1 ∈/ S, i.e.

−N −1+n≤x −N + n > x

41
Subtracting 1 from both sides of the second inequality and combining the
two we get
x − 1 < −N − 1 + n ≤ x
Since −N −1+n ∈ Z, we have established the existence part of the statement.
As to uniqueness, supose there were two integers p, q ∈ Z satisfying x − 1 <
p, q ≤ x. Then, since |p − q| ∈ Z and |p − q| < x − (x − 1) = 1, we see that
p = q. 

Theorem 5.15 For every real x number there is a sequence of rational

numbers converging to x. This sequence (qn )n∈N in Q can always be given
by qn = [nx]
n .

Proof : By the previous theorem, 5.14, we have for all n ∈ N that nx − 1 <
[nx] ≤ nx, so that dividing by n we arrive at
1 [nx] 1
x− < ≤x<x+
n n n
[nx]
from which we get − n1 < n − x < n1 , or
[nx] 1
−x <
n n
1 [nx]
Letting  = n we see that limn→∞ n = x. 

Theorem 5.16 For every real x number there is a sequence of irrational

numbers converging√ to x. This sequence (pn )n∈N in R\Q can always be
[nx]+ 2
given by qn = n .

Proof : As in the previous theorem we have

1 [nx] 1
x− < ≤x<x+
n n n
√ √ √ √
from which we get x − n1 − n2 < x − n1 + n2 < [nx] 2 1
n + n ≤ x < x+ n +
2
n ,
or √ √ √
1+ 2 [nx] + 2 1+ 2
x− < <x+
n n n
√ √ √
Subtracting x from all sides then gives − 1+n 2 < [nx]+
n
2
−x< 1+ 2
n , or
√ √
[nx] + 2 1+ 2
−x <
n n

42
 √ √ √
1+ 2 [nx]+ 2 [nx]+ 2
Letting  = n we see that limn→∞ = x. Of course
[nx] √n √
2
n
is irrational, because n is rational and 2 is irrational, so n is also
√
irrational, and therefore [nx]
n + n
2
is irrational, else the difference of two
rational numbers, √ √
[nx] 2 [nx] 2
+ − =
n n n n
would be rational. 

43