CONSTRUCTION OF NUMBER SYSTEMS

N. MOHAN KUMAR
1. Peanos Axioms and Natural Numbers
We start with the axioms of Peano.
Peanos Axioms. N is a set with the following properties.
(1) N has a distinguished element which we call 1.
(2) There exists a distinguished set map : N N.
(3) is one-to-one (injective).
(4) There does not exist an element n N such that (n) = 1. (So,
in particular is not surjective).
(5) (Principle of Induction) Let S N such that a) 1 S and b)
if n S, then (n) S. Then S = N.
We call such a set N to be the set of natural numbers and elements
of this set to be natural numbers.
Lemma 1.1. If n N and n = 1, then there exists m N such that
(m) = n.
Proof. Consider the subset S of N dened as,
S = {n N | n = 1 or n = (m), for some m N}.
By denition, 1 S. If n S, clearly (n) S, again by denition
of S. Thus by the Principle of Induction, we see that S = N. This
proves the lemma.
We dene the operation of addition (denoted by +) by the following
two recursive rules.
(1) For all n N, n + 1 = (n).
(2) For any n, m N, n + (m) = (n + m).
Notice that by lemma 1.1, any natural number is either 1 or of the
form (m) for some m N and thus the dention of addition above
does dene it for any two natural numbers n, m.
Similarly we dene multiplication on N (denoted by , or sometimes
by just writing letters adjacent to each other, as usual) by the following
two recursive rules.
(1) For all n N, n 1 = n.
1
2 N. MOHAN KUMAR
(2) For any n, m N, n (m) = n m + n.
Agian, lemma 1.1 assures that this denes multiplication of any two
natural numbers. To get a feel for how we identify this set N as our
usual number system, let me prove some of the properties we are fa-
miliar with. Remember, we may use only the axioms, denitions and
whatever we have proved before to prove the successive statements.
This principle should be rigidly adhered to follow our rules of logic.
Lemma 1.2 (Associativity). If x, y, z N, then x+(y+z) = (x+y)+z.
Proof. As before let us dene a subset of N as follows.
S = {z N | x, y N, x + (y + z) = (x + y) + z}
To prove the lemma, we must show that S = N and again we plan to
use the Principle of Induction. To apply the Principle, we must check
two things and we will check them below.
Step 1: 1 S.
For any x, y N, we have,
x + (y + 1) = x + (y) (by denition of addition)
= (x + y) (by denition of addition)
= (x + y) + 1 (by denition of addition)
Thus we get 1 S.
Step 2: If z S, then (z) S.
For any x, y N, we have
x + (y + (z)) = x + (y + z) (by denition of addition)
= (x + (y + z)) (by denition of addition)
= ((x + y) + z) (since z S)
= (x + y) + (z) (by denition of addition)
This proves the lemma.
Lemma 1.3 (commutativity of addition). For any x, y N, x + y =
y + x.
Proof. As always we start with a subset S of N.
S = {y N | x N, x + y = y + x}
To use induction, we need to check two things. Of course, if we show
S = N, we would have proved the lemma.
Step 1: 1 S.
For this we dene a new subset T of N as follows.
T = {x N | x + 1 = 1 + x}
We apply induction to this set T.
CONSTRUCTION OF NUMBER SYSTEMS 3
Step a): Clearly 1 T, since 1 + 1 = 1 + 1.
Step b): Assume x T. Then
1 + (x) = (1 + x) (by denition of addition)
= (x + 1) (since x T)
= ((x)) (by denition of addition)
= (x) + 1 (by denition of addition)
Thus we see that T = N. Going back, we see that this implies 1 S.
Step 2: If y S, then (y) S.
Let x N.
x + (y) = x + (y + 1) (by denition of addition)
= (x + y) + 1 (by associativity, proved before)
= (y + x) + 1 (since y S)
= 1 + (y + x) (since 1 S)
= (1 + y) + x (by associativity)
= (y + 1) + x (since 1 S)
= (y) + x (by denition of addition)

The correct order to prove some of the remaining properties of N

after the above is the following. (There may be other possibilities, but
at least this order will work).
(1) (Cancellative law) For any x, y, z N, if x + z = y + z, then
x = y.
(2) (Distributive law) If x, y, z N then x (y + z) = x y + x z
and (y + z) x = y x + z x.
(3) (Associative law for multiplication) For any x, y, z N, x(yz) =
(xy)z.
(4) For any x N, 1 x = x.
(5) (commutative law for multiplication) For any x, y N, xy = yx.
Let me prove the distributive law now. We will only prove one of them.
Lemma 1.4 (Distributive law). For all x, y, z N, x(y+z) = xy+xz.
Proof. Again, let
S = {z N | x(y + z) = xy + xz, x, y N}
Step 1: 1 S.
x(y + 1) = x(y) (by denition of addition)
= xy + x (by denition of multiplication)
= xy + x 1 (by denition of multiplication)
4 N. MOHAN KUMAR
Step 2: If z S, then (z) S.
x(y + (z)) = x(y + z) (by denition of addition)
= x(y + z) + x (by denition of multiplication)
= (xy + xz) + x (since z S)
= xy + (xz + x) (by associativity of addition)
= xy + x(z) (by denition of multiplication)
Thus by Induction, S = N and we have proved the lemma.
Exercise 1. Prove the remaining properties stated above. Remember,
you may use anything proved earlier for a proof, but no later property
may be used in the proof.
1.1. Ordering on N. Next we introduce the ordering on N.
Denition 1. If n, m N, we say that n is less than m, written n < m,
if there exists a k N such that m = n+k. We also write n m, read
n is less than or equal to m, to mean that either n = m or n < m.
Lemma 1.5. If x, y, z N and x < y and y < z then x < z.
Proof. The assumption x < y means y = x + k for some k N.
Similarly we get z = y + l for some l N. Thus, we get,
z = y + l
= (x + k) + l
= x + (k + l) (by associativity)
Thus by denition x < z since k + l N.
The same argument can be used to show the following.
Lemma 1.6. Let x, y, z N.
(1) If x y and y < z, then x < z.
(2) If x < y and y z, then x < z.
(3) If x y and y z, then x z.
Exercise 2. If x, y, z N and x < y show that x + z < y + z and
xz < yz.
Lemma 1.7. For any m N, m = m + 1.
Proof. As usual dene a subset S N as follows:
S = {n N | n = n + 1}
Clearly, 1 S, since if not, 1 = 1+1 = (1), by denition of addition
and 1 = (k) for any k N by Peanos axioms (number 4).
CONSTRUCTION OF NUMBER SYSTEMS 5
Assume n S. If (n) is not in S, then (n) = (n) + 1. Then
by denition of addition, (n) = ((n)). By the third axiom, this
means, n = (n) which in turn is n +1 by denition of addition. This
is impossible since n S.
Lemma 1.8. For any m, k N, m = m + k.
Proof. Again dene a subset S N as follows:
S = {k N | m N, m = m + k}
From the previous lemma, we see that 1 S. If k S, we want to
show that (k) S and then by induction we would be done. That
is, we want to show that m = m + (k) for any m. Notice that
m + (k) = (m + k), by denition of addition.
Let us dene a subset T as follows:
T = {m N | m = (m + k)}
Clearly 1 T by axiom 4. Assume m T. Want to show that
(m) T. If (m) = ((m) + k), by axiom 3, we see that m =
(m) + k. Thus we have,
m = (m) + k
= (m + 1) + k (by denition of addition)
= m + (1 + k) (by associativity)
= m + (k + 1) (by commutativity)
= m + (k) (by denition of addition)
= (m + k) (by denition of addition)
But this contradicts our assumption that m T.

Lemma 1.9 (Well ordering of N). If n, m N, then exactly one of the

following is true. Either n < m or n = m or m < n.
Proof. Let us rst prove only one of these can hold. If n < m, then by
denition, m = n+k for some element k N. By the previous lemma,
we see that m = n. If m < n, then there exists an l N such that
n = m + l which implies m = (m + l) + k = m + (l + k) which again
is not possible by the previous lemma. The other cases are similar and
left as an exercise.
To nish the proof we consider the set S, xing an n.
S = {m N | n < m, n = mor m < n}
Step 1: 1 S.
6 N. MOHAN KUMAR
If n = 1, clearly 1 S. If n = 1, then by lemma 1.1, n = (k) =
k + 1 = 1 + k and thus by denition of our ordering, 1 < n.
Step 2: If m S then (m) S.
Assume m S. This means we have three possibilities, namely
n < m or n = m or m < n. First, let us look at the case n < m.
Then m = n + k for some k. Thus (m) = (n + k) = n + (k) and
so n < (m) and hence (m) S. Next possibility is n = m. Then
(m) = m+1 = n+1 and thus n < (m) and again (m) S. Finally,
we have the possibility of m < n. Thus n = m + k for some element
k N. If k = 1, then n = (m) and thus (m) S. If not, by lemma
1.1, k = (l) and thus n = m + (l) and thus,
n = m + (l)
m + (l + 1) (by denition of addition)
= m + (1 + l) (by commutativity)
= (m + 1) + l (by associativity)
= (m) + l (by denition of addition).
Therefore (m) < n by denition of our ordering and thus (m) S.
Thus in any case (m) S. Therefore by induction, S = N and we are
done.

Remark 1. As usual, we will write m > n, read m is greater than n to

mean n < m. Similarl meaning is assigned to m n.
Next I want to prove some alternate forms of induction which are
frequently used. We start with a denition.
Denition 2. Let S N. Then an element n S is called a least
element if for any m S, n m.
Lemma 1.10. Let S N. If S has a least element, then it is unique.
So it makes sense to use the denite article the instead of a if a least
element exists and call it the least element.
Proof. Assume S has two least elements, say n and m. Thus we see
that n m and m n. From the well ordering lemma, it is clear that
n = m.

Theorem 1.1 (First alternate form of Induction). If S N and S =

then S has a least element.
Proof. As usual let
T = {n N | if n S N, then S has a least element}
CONSTRUCTION OF NUMBER SYSTEMS 7
Caution: Here T is a xed set dened as above. But, S is a variable
subset of N.
I will leave it as an exercise to check that 1 T. Next assume
that n T and we want to show that (n) T. So let S N with
(n) S. If n S, then by hypothesis, we know that S has a least
element. So assume that n is not in S and consider A = S {n}. Then
n A and thus A has a least element by hypothesis. Let us call this
least element a.
There are two possibilities. Either a = n or a = n. If a = n, then
a S. If m S, then clearly m A and thus a m. So we see that
a S is a least element.
Next assume a = n. Then I claim that (n) is the least elemnt of
S. If m S, we know that a = n m. But n S and thus n = m.
Thus by denition of less than or equal to, we see that n < m. Thus
we may write m = n + k for some k N by dention. If k = 1, then
m = n+1 = (n). If k = 1, then by the rst lemma, k = (l) for some
l N. Thus
m = n + (l + 1) = n + (1 + l) = (n + 1) + l = (n) + l
and thus (n) < m. Thus we see that (n) is a least element of S and
we are done.

Some of you may be more familiar with the following form of induc-
tion, though all the three are equivalent.
Theorem 1.2 (Second alternate form of Induction). Let P(n) be math-
ematical statements for n N. Assume
(1) P(1) is true.
(2) If P(n) is true, then P(n + 1) is true.
Then P(n) is true for all n N.
Proof. Dene a set S = {n N|P(n) is false}. We wish to show that
S = . If non-empty, by the previous form of induction,Theorem 1.1,
we have a least element m S. By the rst hypothesis of the theorem,
m = 1. Then m = p + 1 for some p N. Since p < m and m being
the least elelement of S, we know that p S and thus P(p) is true by
denition of the set S. Now, by our second hypothesis, P(p+1) = P(m)
is true and hence m S, a contradiction, proving the result.
Theorem 1.3 (Third alternate form of Induction). Let P(n) be math-
ematical statements fos n N. Asuume,
(1) P(1) is true.
(2) If n > 1 and P(k) is true for all k < n, then P(n) is true.
8 N. MOHAN KUMAR
Then P(n) is true for all n.
Proof. As before, let S = {n N|P(n) is false} and we wish to show
that S = . So assume that it is non-empty and let m S be the
least element assured by Theroem 1.1. Again, as before, by the rst
hypothesis, 1 S and thus m > 1. By minimality of m, if k < m, then
k S and hence P(k) is true. Thus by second hypothesis P(m) is true
and thus m S, which is a contradiction, proving the theorem.
We will use these forms in the next section on Number Theory to
prove results familiar to you. We state some more properties of natural
numbers, which can be proved using the above ordering properties of
N.
(1) (cancellative law for multiplication) For any x, y, z N, if xz =
yz then x = y.
(2) (uniqueness of identity) For some x, y N, if xy = x, then
y = 1.
At this point, we will use our usual nomenclature for natural
number. We already have called a special number 1 and then
we call 2 = 1 + 1, 3 = 2 + 1 etc. in the usual fashion.
2. Integers
We will briey desribe the construction of integers and rational num-
bers below and state various properties in the correct order and prove
just a few to give a avour.
Consider the set S = N N and put a relation on it as follows:
(a, b) (c, d) if and only if a + d = b + c. (As usual, we denote a
typical element in S by an ordered pair of natural numbers)
Check that this is an equivalence relation on S. Let Z be the set of
equivalence classes under this relation. Dene an operation (addition)
on Z as follows: If A, B Z, then recall that A, B are non-empty
subsets of S and thus we may pick elements (a, b) A and (c, d)
B. With our notation for equivalence classes, this means A = [(a, b)]
for example. Dene an operation tentatively denoted by , to avoid
confusion, as follows:
A B = [(a + c, b + d)]
There is a priori a problem with this denition. To make sure that
the operation is well-dened (a term we will see several times in the
sequel), we must make sure that the right hand side above has only
value. This operation is supposed to be a function from Z Z Z,
given two integers, we must get a well-dened integer as its sum. But,
let us look at our denition. Here we picked some (a, b) A and
CONSTRUCTION OF NUMBER SYSTEMS 9
(c, d) B and declared that A B = [(a + c, b + d)]. We could have
easily picked another (a

, b

) A and (c

, d

) B. Then we would have

got AB to be [(a

+c

, b

+d

)], which may very well be dierent from

[(a+c, b +d)] and then we really do not have function from ZZ Z
and no real denition. So, in such situations, the rst order of business
is always making sure that it is well-dened. In other words, with the
above notation, we must check that [(a + c, b + d)] = [(a

+ c

, b

+ d

)].
For once let us check this.
Lemma 2.1. The addition dened above is well-dened.
Proof. As discussed, we must check that if A = [(a, b)] = [(a

)] and
B = [(c, d)] = [(c

, d

)], then we must check that [(a +c, b +d)] = [(a

+
c

, b

+d

)]. Unwinding the dentions, this means that if (a, b) (a

, b

)
and (c, d) (c

, d

), then we must check (a+c, b +d) (a

+c

, b

+d

).
Again looking at our relation, we get that a + b

= a

+ b and c + d

=
c

+d. Adding them we get a+b

+c+d

= a

+b+c

+d, which implies

that (a + c, b + d) (a

+ c

, b

+ d

).
(Intuitively, the element [(a, b)] should be thought of as a b in
our familiar settings though we are yet dene subtraction.) Dene
multiplication tentatively denoted by as follows:
A B = [(ac + bd, ad + bc)]
and as before make sure that this is well-dened. Now proofs of all
the familiar properties of addition and multiplication of integers can
be carried out, by using the denitions and corresponding properties
of natural numbers.
(1) Associativity of addition.
(2) Commutativity of addition.
(3) Cancellative property of addition.
(4) For any two natural number a, b N, (a, a) (b, b) and thus
[(a, a)] = [(b, b)] whic we denote by the symbol 0. Then for any
A Z, A 0 = A = 0 A.
(5) Additive inverse: If A = [(a, b)], then we denote by A =
[(b, a)], the additive inverse of A. Then A (A) = 0.
(6) If AB = 0 then B = A and in particular the additive inverse
is unique.
(7) Distributivity.
(8) Associative law for multiplication.
(9) Commutative law for multiplication.
(10) For any a, b N, ((a), a) ((b), b) and thus we denote the
equivalence class [((a), a)] for any a N, by the symbol 1.
Then for any A Z, A 1 = A = 1 A.
10 N. MOHAN KUMAR
(11) Cancellative law for multiplication: If A, B, C Z and AC =
B C with C = 0, then A = B.
(12) Uniqueness of identitiy: If A B = A and A = 0, then B = 1.
(13) A 0 = 0 for all A Z.
(14) If A B = 0, then either A = 0 or B = 0.
(15) For any a, b, k N, (a+k, a) (b +k, b) and thus [(a+k, a)] =
[(b + k, b)]. So dene a map f : N Z by f(k) = [(a + k, a)]
for some a N. Then f is one-one and f(a + b) = f(a) f(b)
and f(ab) = f(a) f(b).
The last property ensures that N Z via f and the addition and
multiplication are respected by f. Thus we may now drop , and
write just +, .
As before we dene an ordering on Z by saying that A < B if A =
[(a, b)] and B = [(c, d)], then a +d < b +c. Make sure that this is well
dened and we dene A B if either A = B or A < B. Also show
that if a, b N, then a < b if and only if f(a) < f(b) and thus the
ordering is also respected by f. We say that A Z is positive if 0 < A
and non-negative if 0 A. If A < 0, we say that A is negative. As
usual, we write A > B to mean B < A etc.
Lemma 2.2. Let A = [(a, b)]. Then A is positive if and only if b < a.
Proof. If 0 < A, then since 0 = [(a, a)], we see that a + b < a + a by
denition of the ordering. Using the denition of ordering in N and
cancellation, the result follows. Converse is equally easy.
Lemma 2.3. A is positive if and only if A = f(k) for some natural
number k.
Proof. Let A = [(a, b)]. First assume that it is positive. Then b < a
from the previous lemma and thus a = b + k for some natural number
k and thus A = [(b + k, b)] = f(k). Converse is equally easy.
Lemma 2.4. A Z is positive if and only if A is negative.
Proof. Left as an exercise.
Lemma 2.5. If a, b, c N with a < b then ac < bc.
Proof. If a < b, then b = a + k for some natural number k. Thus
bc = ac + kc. Since kc N, by denition, ac < bc.
Corollary 2.1. If a < b for a, b N, then a
2
= a a < b
2
= b b.
Proof. From the previous lemma, since a < b, we get a
2
< ab. Applying
the lemma again, we get ab < b
2
. Putting them together, we get
a
2
< b
2
.
CONSTRUCTION OF NUMBER SYSTEMS 11
Lemma 2.6. Let a, b N and 1 < a. Then there exists a natural
number N such that for all n N, b < a
n
. (Recall, a
n
is just a
convenient way of writing the product of a, n times).
Proof. Consider the set
S = {b N | there exisits N such that n with n N, b < a
n
}
Then 1 S. For this take N = 1 and apply the previous lemma as
follows. Let
T = {n N|1 < a
n
}.
Then 1 T since 1 < a. If n T, then we have, 1 < a
n
. Multiplying
by a, from the previous lemma, we have a < a
n+1
. Putting these
together, we have 1 < a
n+1
and thus n + 1 T and thus by induction
we see that T = N, proving 1 S.
Assume that b S. So, there exists N such that b < a
n
for all
n N. I claim that for (b), we can take instead of N, N + 1. Let
n N + 1. Since n = 1, we may write n = m + 1 and m N.
Thus by induction hypothesis, b < a
m
and thus by the lemma above,
ab < a
n
. Since 1 < a, we can write a = k + 1. Thus ab = bk + b So,
ab = bk + b b + 1. Thus we get b + 1 < a
n
. So b + 1 S and we are
done by induction.

Let us denote by N

= N {0} Z, where N is identied with the

subset f(N). We have a map Z N

called the absolute value, denoted

by | |. |a| = a if a is non-negative and |a| = a if a is negative.
Lemma 2.7. For any a Z, we have |a| a |a|. Converseley, if
c 0 is an integer and c a c, then |a| c
Proof. First assume that a 0. Then |a| = a by dention. So, the
inequlaity we need is a a a. Since a a by denition, we only
need to verify that a a. But this is same as 2a 0, by adding a
to both sides of the inequality and this is clear. The case of a < 0 is
equally easy.
To prove the second statement, we work exactly as before by looking
at the two cases of a 0 and a < 0. If a 0, then a = |a| and since
a c, we get that |a| c. If a < 0, then |a| = a. Since c a, we
get that c a = |a|, proving the result.
Lemma 2.8 (Triangle inequality). If a, b Z then |a + b| |a| +|b|.
Proof. From the previous lemma, we have |a| a |a| and |b|
b |b|. Adding these, we get,
(|a| + b|) a + b |a| +|b|.
12 N. MOHAN KUMAR
Since |a| +|b| 0, by the previous lemma. we are done.
Finally now we can dene a new operation subtraction denoted by
in Z as follows. If a, b Z, then a b = a + (b). I will leave the
standard properties of subtraction to be veried by the reader. Let me
close this section by yet another form of induction.
Theorem 2.1 (Fourth alternate form of Induction). Let S Z be
a non-empty subset of the integers with the property that there is an
m Z such that for all a S, a > m. (That is to say that the set S
is bounded below). Then S has a least element.
Proof. Consider the set T = {a m|a S}. Since S = nor is T.
Since a > m, a m > 0 and hence T N. Thus, by Theorem 1.1,
T has a minimal element, say p. Then q = p + m S and we claim
that q is the least element of S. If a S, then a m T and hence
am p and thus a p+m = q. Thus by denition of least element,
q is the least element of S.
Exercise 3. Show that if a, b Z, then a (b) = (ab).
3. Some Number Theory
We will not use most of what we prove in this section in the sequel.
I have included it only to connect our discussions with facts familiar
to you and thus giving you some bearing in this abstract jungle. We
start with one of the earliest results you might have studied in school.
Theorem 3.1 (Division Algorithm). Let a, d Z with d = 0. Then
there exists unique integers q, r Z such that a = qd + r and 0 r <
|d|.
Proof. Since d = 0, d is either positive or negative. We rst treat the
case d > 0. Consider the set S = {a qd|a qd 0, q Z}. First I
claim that a+|a|d 0, showing that S = . If a = 0, then a+|a|d = 0,
if a > 0, then a + |a|d = a + ad = a(1 + d) > 0 and if a < 0, then
a + |a|d = a ad = a(1 d) 0, since a < 0 and 1 d 0. Since
S is bounded below by zero, we can apply Theorem 2.1 to conclude
that S has a minimal element, say 0 r = a qd for some q Z.
So, a = qd + r and if we show that r < d, we would have proved the
existence part of the theorem. If r d, then a (q +1)d = r d 0.
Thus r d S and since r d < r, this contradicts the minimality of
r. So, r < d.
Next, we prove uniqueness. If a = qd + r = q

d + r

with 0
r, r

< d, we get (q q

)d = r

r. If q = q

, then this implies

r = r

, proving uniqueness. If q = q

, then taking absolute values, we

CONSTRUCTION OF NUMBER SYSTEMS 13
get, d |(q q

)d| = |r r

|. But since 0 r, r

< d, we see that

|r r

| < d, which leads to a contradiction. Thus uniqueness is proved.

Finally, if d < 0, let e = d = |d| > 0. Thus by the previous part,
we have a = qe+r with q, r Z and 0 r < e = |d|. So, a = (q)d+r
as desired.
As usual if 0 = d Z, and a Z, we say that d divides a if a = md
for some m Z. Symbolically, this is written as d | a.
Denition 3. Let a, b Z with at least one of them non-zero. Then the
greatest common divisor of a, b, written gcd(a, b) is a number d N
satisfying the following two properties.
(1) d | a and d | b.
(2) If e N divides both a and b, then e | d.
Notice that gcd is dened only for two numbers with at least one
of them non-zero. Also, notice that it is a natural number. What is
not clear from the denition is whether such a number exists and if it
exists whether it is unique and these we proceed to prove.
Lemma 3.1. Let a, b Z with at least one of them non-zero. If
gcd(a, b) exists, then it is unique.
Proof. Let d = gcd(a, b) and e = gcd(a, b). We wish to show that d = e.
By rst property in the denition applied to e we get that e | a, e | b.
Now applying the second property, we see that e | d. Reversing the
roles, we see that d | e. It is easy to see then d = e, though let me give
an explicit proof.
e | d means we can write d = pe with p Z. Since d, e N, this
forces p > 0 and hence p N. Similalry we get e = qd with q N.
Substituting, we get, d = pqd. Cancelling d > 0 (which we have shown
earlier), we get 1 = pq. If we show that this implies p = q = 1, we
would be done. If p = 1, we could write p = s + 1, s N from earlier
properties of natural numbers. Thus we get qs + q = 1. If q = 1, we
have d = e, so we may also assume that q > 1 and thus we can write
q = t +1. Then qs+t +1 = (qs+t) = 1, but Peanos axioms say that
1 is not in the image of , a contradiction, proving what we need.
Theorem 3.2 (Existence of gcd). Let a, b Z with at least one of
them non-zero. Then gcd(a, b) exists.
Proof. Consider the set, S = {ma + nb > 0|m, n Z}. As usual, we
claim that this set is non-empty. Since at least one of a, b is non-zero,
we may assume that a = 0, possibly after renaming them. Then let
m = a, n = 0. So, a a + 0 b = a
2
> 0 and hence a
2
S. So S is
14 N. MOHAN KUMAR
non-empty. Since S N, by Theorem 1.1, S has a minimal element,
say d. By denition of the set S, we have m, n Z so that d = ma+nb.
I claim that d = gcd(a, b).
For this, we need to check the two conditions in the denition of
gcd. The second one is easy. If e divides both a, b, then we can write
a = pe, b = qe for some p, q Z and substituting, we get, d = ma+nb =
mpe + nqe = (mp + nq)e and since mp + nq Z, we see that e | d.
For the rst condition, we proceed as follows. First, we show that d |
a, the other case being similar, we shall omit it. By division algorithm,
we can write a = qd + r with q, r Z and 0 r < d. If r = 0, d | a,
so let us assume that r > 0 and then we will arrive at a contradiction.
Then r = a qd = a q(ma + nb) = (1 qm)a + (qn)b and by
denition of the set S, we see that r S. But, r < d, contradicting
the minimality of d.
The above proof gives something stronger and it is in this form it is
often used, so let us state this explicitly.
Corollary 3.1. Let a, b Z with at least one of them non-zero. Then
gcd(a, b) exists and is unique. Further, it is the smallest natural number
of the form ma + nb with m, n Z.
Here is an immediate application.
Corollary 3.2. Let p, a, b Z and assume that p = 0. If p | ab and
gcd(p, a) = 1, then p | b.
Proof. gcd(p, a) = 1 implies there exists integers m, n such that 1 =
mp +na. Multiplying by b we get, b = mpb +nab. Since p | ab, we can
write ab = sp for some s Z. Thus we get, b = mpb+nsp = (mb+ns)p
and since mb + ns Z, we get that p | b.
Denition 4. A natural number p is a prime number if p = 1 and the
only natural numbers dividing it are 1 or p. An integer n is called a
composite number if |n| = 0, 1 and |n| is not a prime.
I will not give examples of primes, since I am sure most of you are
familiar with them.
Exercise 4. (1) Show that if p is a prime then for any a N,
gcd(p, a) = 1 or gcd(p, a) = p.
(2) Show that if p is a prime and p | ab for a, b Z, then p | a or
p | b.
(3) Show that if a prime p divides a prime q, then p = q.
(4) Show that if a prime p divides q
n
for a prime q and n N, then
p = q.
CONSTRUCTION OF NUMBER SYSTEMS 15
Theorem 3.3 (Fundamental Theorem of Arithmetic, Part 1). Let
1 = n N. Then there exists primes, p
1
, p
2
, . . . , p
k
such that n =
p
1
p
2
p
k
. That is, any n N, n > 1 is a product of primes.
Proof. As usual we start with the set
S = {n N|n > 1, nis not a product of primes}.
We wish to show that S is empty and if it is non-empty, we let n S
be the least element, assured by Theorem 1.1. Now, n can not be a
prime, since then n is the product of one prime. Thus by denition of
a prime, there exists a natural number d which divides n and d = 1, n.
Then by properties of natural numbers, we get 1 < d < n. Since n = de
for some natural number e, we get that 1 < e < n. By minimality of
n, we get d, e S and since they are not 1, they are product of primes.
So, d = p
1
p
r
, e = q
1
q
s
for primes p
i
, q
j
. Thus,
n = de = p
1
p
r
q
1
q
s
and hence n is a product of primes. So, n S, leading to the desired
contradiction.
Thus, if n N and n > 1, we can write n = p
1
p
k
for primes p
i
s.
Collecting the primes and ordering them, we may assume that there
exists primes p
1
< < p
m
and natural numbers a
1
, . . . , a
m
such that
n = p
a
1
1
p
a
m
m
.
Theorem 3.4 (Fundamental Theorem of Arithmetic, Part 2). Let
n > 1 be a natural number and let n = p
a
1
1
p
a
m
m
as in the previous
paragraph. Then this expression is unique. That is, if n = q
b
1
1
q
b
l
l
with q
1
< < q
l
primes and b
i
N for all i, then m = l, p
i
= q
i
and
a
i
= b
i
for all i.
Proof. Again, let S be the set of natural numbers not equal to 1 and
which have two dierent such decompositions into primes. We wish to
show that S is empty and if not pick n, the least element assured by
Theorem 1.1. So, we have,
n = p
a
1
1
p
a
k
k
= q
b
1
1
q
b
l
l
with p
i
, q
j
primes, p
1
< p
2
< < p
k
, q
1
< q
2
< < q
l
, a
i
, b
j
N.
Now, p
1
divides n and thus p
1
divides q
b
1
1
q
b
l
l
. By exercise 4, we then
have p
1
= q
i
for some i. Similarly, q
1
= p
j
for some j.
We rst look at the case i > 1. Then p
1
p
j
(j may be 1) and since
p
j
= q
1
< q
i
= p
1
, we get p
1
< p
1
, a contradiction. So, we see that
i = 1 and p
1
= q
1
. Next we claim that a
1
= b
1
. If not we may assume
16 N. MOHAN KUMAR
by well ordering, that a
1
> b
1
(or the other way around, but we can
interchange p, q. Cancelling p
b
1
1
, we get,
p
a
1
b
1
1
p
a
2
2
p
a
k
k
= q
b
2
2
q
b
l
l
.
But then p
1
= q
i
for some i > 1 as before, but all these are greater
than q
1
= p
1
, which is impossible. This proves that a
1
= b
1
. Then we
have,
m = p
a
2
2
p
a
k
k
= q
b
2
2
q
b
l
l
.
Since m < n, m S and thus, by minimality of n, the theorem is true
for m. This implies, k = l, p
i
= q
i
for i > 1 and a
i
= b
i
for i > 1. This
says, since p
1
= q
1
and a
1
= b
1
, that the theorem is true for n and hence
n S, which is the desired contradiction, proving the theorem.
Theorem 3.5 (Innitude of primes). There are innitely many primes.
Proof. We know that there at least some primes, for example 2, 3. As-
sume there are only nitley many primes, say p
1
, . . . , p
k
. Consider the
natural number N = p
1
p
2
p
k
+1. Then clearly N > 1 and hence by
the fundamental theorem, there exists a prime number q which divides
N. But, since the p
i
s are all the primes, q = p
i
for some i. Thus p
i
divides N. Then p
i
divides Np
1
p
i
p
k
= 1 and no prime number
can divide 1, leading to a contradiction. This proves the theorem.
4. Rational numbers
The idea is the same. So, I will briey sketch the construction. Now
consider the set S = Z (Z {0}) and put a relation as follows.
(a, b) (c, d) if ad = bc. One easily checks that this is indeed an
equivalence relation on S and the set of equivalence classes is called
the rational numbers and denoted by Q. As usual, addition is dened
by [(a, b)] [(c, d)] = [(ad + bc, bd)] and multiplication by [(a, b)]
[(c, d)] = [(ac, bd)], where now we have addition and multiplication
of integers inside the brackets. As usuaul, we check that this is well
dened and all the standard properties. We also have a one-one map
f : Z Q given by f(a) = [(a, 1)] and then f(a + b) = f(a) f(b)
and f(ab) = f(a) f(b). Intuitively, we are thinking of [(a, b)] as a/b.
Using f, we can identify Z as a subset of Q in the usual way. Thus
again, we can drop , and write the usual symbols for addition and
multiplication.
In Q, we have a new operation, division, as usual denoted by a/b
for a, b Q and b = 0. This is dened as follows: If A = [(a, b)] and
B = [(c, d)], with a, b, c, d Z, b = 0 = d, then B = 0 implies that
c = 0. Dene A/B = [(ad, bc)] and make sure that this is well dened.
Also notice that since both b and c are not zero, bc = 0.
CONSTRUCTION OF NUMBER SYSTEMS 17
Lemma 4.1. If A is any rational number, then A = [(a, b)], for some
a, b Z with b positive.
Proof. By dention, A = [(a, b)] with b = 0, a, b Z. If b is positive,
then we are done. If not, we know that b is positive. One easily sees
that A = [(a, b)] and thus we are done.
We introduce an order on Qas follows. If A, B Q, write A = [(a, b)]
and B = [(c, d)] with b, d both positive. Then we dene, A < B if
ad < bc. One checks that this is well dened and has all the usual
properties. As always, write A B to mean either A = B or A < B.
Absolute value of a rational number can be dened as before, after
dening what is positive, negative etc. Again, I will assume that we
can prove all the usual properties of ordering on Q.
Exercise 5. (1) Show that if a < b are two rational numbers, then
there exists c Q with a < c < b.
(2) Prove triangle inequality (see lemma 2.8) for Q.
5. Real Numbers
Let me start with some observations. We started with natural num-
bers which had this very important property of Induction. In other
words, any non-empty subset had a least element. But the system
lacked operations like subtraction and thus we were forced to enlarge
the system to integers, which had subtraction, and at least a suitable
form of induction; namely, any non-empty subset which is bounded be-
low had a minimal element. But integers still lacked division and thus
we enlarged our system to the rational numbers to rectify this lacuna.
But, now, alas we have lost any semblance of the minimal element
property which was so important and desirable. In other words, there
are now subsets of Q which are bounded below with no minimal ele-
ment in sight. (Not just in the set, but not even in Q). So, the plan of
constructing real numbers is to rectify this important problem with Q.
Let me elaborate the above statement. Let
S = {x Q | x
2
> 2 and x is positive}.
Clearly this set is non-empty and bounded below. For example, 2 S
and if x S, then x > 1. But I claim that this set has no minimal
element even in Q. Assume a Q is such a minimal element. Then a
has two properties. First, a x for all x S. Second, if b Q, with
b x for all x S, then b a. I claim no such rational number exists.
Since 1 works as a b, we see immediately that a 1. In particular a
must be positive. I claim that a
2
2. If not, a
2
> 2 and a S. So
18 N. MOHAN KUMAR
a
2
= 2 + r for some positive rational number. We can choose a large
natural number N so that 2a/N < r. Consider c = a (1/N) < a.
Then
c
2
= a
2
2
a
N
+
1
N
2
= 2 + (r 2
a
N
) +
1
N
2
> 2.
Thus c S and c < a. This is a contradiction. Thus a
2
2.
Next I leave it as a (non-trivial) exercise that there is no rational
number with a
2
= 2 and thus we must have a
2
< 2. Then again
write 2 = a
2
+ r with r a positive rational number. As before we can
choose a large natural number N such that rN 2a 1 and thus
2a/N +1/N
2
< r. Now consider b = a +(1/N) > a. One easily checks
that b
2
< 2 and thus b x for x S. This again is a contradiction.
Thus, though rational numbers had several of the arithmetic prop-
erties for numbers that we desire, it lacks a very important property
necessary for Mathematics. In real life, we rarely have to deal with
a real number which is not rational in some strict sense, considering,
any non-rational number is usually approximated to a rational num-
ber, like all the numbers you may get out of a calculator or computer.
For Mathematics, approximation of this kind is never sucient, if you
want to be rigorous and precise. So, this is some justication for the
construction of real numbers.
5.1. Cauchy Sequences. Now we plunge into this construction. As
you would expect, real numbers are got by approximating rational num-
bers. So we make a couple of denitions.
Denition 5. A sequence of rational numbers is a set map f : N Q.
In other words, we are given rational numbers x
n
for every natural
number n. This is usually abbreviated by the notation {x
n
}.
Denition 6. A sequence {x
n
} is a Cauchy Sequence (we will abbreviate
it by writing CS), if given any 0 < Q, there exists a natural number
N (which is allowed to depend on ) such that for all n, m N, natural
numbers, we have |x
n
x
m
| < .
I suggest that you mull over this important denition. We will give
a few examples below. Typically, given a sequence of rational numbers
{x
n
}, to show that it is not a CS, we will have to exhibit one positive
rational number such that for any N N, there exists n, m N
with |x
n
x
m
| . On the other hand, if we wished to show that the
sequence is a CS, we must show that for any positive rational number
, there exists an N N and for any n, m N, |x
n
x
m
| < .
To warm yourself to the concept of CS, here is are some easy exer-
cises.
CONSTRUCTION OF NUMBER SYSTEMS 19
Exercise 6. (1) If a is any rational number, show that the sequence
dened as x
n
= a for all n N is a CS.
(2) Let g : N N be an increasing function. That is, if n > m
then g(n) > g(m). If {x
n
} is a Cauchy sequence, show that
{y
n
} is a CS, where y
n
= x
g(n)
. ({y
n
} is called a sub-sequence
of {x
n
}).
(3) If {x
n
} is a CS of rational numbers, show that for any rational
number a, {ax
n
} is a CS.
Example 1. (1) Let x
n
= n for all n N. Then this is not a CS.
As I said in the previous paragraph, we need to nd just one
positive rational number which violates the CS condition. For
this take = 1. If an N existed, then we must have |x
n
x
m
| <
= 1 for all n, m N. But if we take n = N and m = N + 1,
clearly we get a contradiction. (How did I decide to take to be
1? Usually, one works backwards and analyzes what one needs.
Sometimes this can be tricky.)
(2) Let x
n
= 2
n
for all n N. Then {x
n
} is a CS.
Here, we are not allowed to pick an . We must take any
positive rational number and gure out an appropriate N
guaranteeing the CS condition. So, let > 0 be given. (Here
again, to gure out the N, one may have to work backwards
and can be tricky.) Let us do the analysis for once.
We need an N so that for any n, m N, we must have,
|x
n
x
m
| = |2
n
2
m
| < .
Without loss of generality, we may assume that n m N.
Then |2
n
2
m
| = 2
m
|2
mn
1|. Since 0 < 2
mn
1,
because n m, we see that |2
mn
1| < 1 and thus we see
that |2
n
2
m
| < 2
m
. But, since m N, we also have
2
m
2
N
. So, if we had 2
N
< , then we would be done.
Notice that we nally ended up with a condition on N and not
on which n, m N we need to look at. This is usually the
essence of most such arguments.
So, we see that we only need to nd an N such that 2
N
< ,
for the given positive . Since = 0, we have =
1
Q and
since is positive so is . So, we want an N so that 2
N
> . If
we write = a/b with a, b positive integers, then it is clear that
a . So, if we found an N so that 2
N
> a, we would be done.
But this is essentially the content of lemma 2.6.
So if you want to write a proof, we invert the steps, so that
it is easy to follow the arguments. Let me do this now for
illustration.
20 N. MOHAN KUMAR
Proof. We wish to show that the sequence {x
n
} = {2
n
} is a
CS.
So, let > 0 be a positive rational number. Then =
1
Q
is a positive rational number. Writing = a/b for positive
integers a, b, we see immediatley that a . By lemma 2.6,
since 2 > 1, there exists an N such that
2
N
a (1)
Then we claim that for any n, m N, (for the N from the equa-
tion above) |x
n
x
m
| < , which will prove that our sequence
is a CS.
We may assume that n m N. Then
|x
n
x
m
| = |2
n
2
m
| = 2
m
|2
mn
1|.
Since n m, we have 0 < 2
mn
1 and thus, |2
mn
1| < 1.
So, we get |x
n
x
m
| < 2
m
if n m N. Since m N, we
have 2
m
2
N
and thus we get,
|x
n
x
m
| < 2
m
2
N

1
= ,
where the last inequlaity follows from equation 1 and this is
valid for n m N. This is what we set out to prove and
thus the sequence {x
n
} is a CS.
(3) For any natural number n, we have 2n
2
< 2n
2
+ 1. Thus by
exercise (5), we can choose a rational number y
n
such that 2n
2
<
y
n
< 2n
2
+ 1. Then the sequence {x
n
= y
n
n
2
} is a CS.
Given > 0, easy to see that there exists an N N such that
for all n N, we have 0 < n
2
< . If n m N, then we
have,
|x
n
x
m
| < 2 + m
2
2 = m
2
<
(4) Let
x
n
= 1 +
1
2
+
1
3
+ +
1
n
Then {x
n
} is not a CS. This is left as a not too easy exercise.
Or may be look it up in your calculus text.
5.2. Construction of Real Numbers. Next let S be the set of all
Cauchy Sequences in Q. We put a relation on S as follows: {x
n
}
{y
n
}, if given any 0 < Q, there exists an N N such that for all
n, m N, we have |x
n
y
m
| < .
This is indeed an equivalence relation. Reexivity is just the fact that
the sequences are Cauchy. Symmetry is obvious since |xy| = |y x|.
Transitivity follows from triangle inequality (see Exercise 5).
CONSTRUCTION OF NUMBER SYSTEMS 21
Thus we consider the set R as the set of equivalence classes and call
it the real numbers.
Before we introduce the operations on R, let us prove a few things
about CS.
Lemma 5.1. Let {x
n
} be a CS. Then there exists a rational number
M Q, such that |x
n
| < M for all n. That is to say, the sequence is
bounded.
Proof. Pick = 1. So we have an N N satisfying the properties of
the denition of a CS. Let
M = max{|x
1
|, |x
2
|, . . . , |x
N1
|, |x
N
|} + 1.
By choice, |x
n
| < M if n N. If n > N, then by choice, we have,
|x
n
x
N
| < 1 and then using triangle inequality (see Exercise 5), we
are done.
|x
n
| = |x
n
x
N
+ x
N
| |x
n
x
N
| +|x
N
| < 1 +|x
N
| M

As usual we dene the operations as follows. If A = [{x

n
}] and
B = [{y
n
}], dene AB = [{x
n
+ y
n
}] and AB = [{x
n
y
n
}]. Let us
check that these make sense.
Lemma 5.2. If {x
n
} and {y
n
} are CS, then so are {x
n
+ y
n
} and
{x
n
y
n
}.
Proof. Given , we can nd N
1
, N
2
N such that for all n, m N
1
we
have |x
n
x
m
| < /2 and for all p.q N
2
, y
p
y
q
| < /2. Take N =
max{N
1
, N
2
} and use the triangle inequality to show that {x
n
+y
n
} is
Cauchy.
For the multiplication, we use lemma 5.1 and thus there exists M
1
, M
2

Q such that |x
n
| < M
1
for all n and |y
n
| < M
2
for all n. Clearly,
we may replace M
1
, M
2
by M = max{M
1
, M
2
}. Given > 0, con-
sider = /2M > 0. As before, there exists N
1
, N
2
such that for all
n, m N
1
, we have |x
n
x
m
| < etc. Again we may repalce N
1
, N
2
by N = max{N
1
, N
2
}. Thus we get,
|x
n
y
n
x
m
y
m
| = |(x
n
x
m
)y
n
+ x
m
(y
n
y
m
)|
|(x
n
x
m
)y
n
| +|x
m
(y
n
y
m
)|
by triangle inequality. If n, m N, this in turn gives
|x
n
x
m
||y
n
| +|x
m
||y
n
y
m
| < M + M =
proving that our sequence is Cauchy.

22 N. MOHAN KUMAR
Now, as usual, I will leave it to you to check that this denition of
addition and multiplication are well dened. (Please try to check this
it is a good exercise). I will check that addition is well dened below,
just for illustration.
Lemma 5.3. Addition of real numbers as dened above is well dened.
Proof. Our denition was if A = [{x
n
}] and B = [{y
n
}], two real
numbers, then AB = [{x
n
+y
n
}] and we have checked in the previous
lemma that {x
n
+y
n
} is a CS. If A = [{t
n
}] and B = [{u
n
}], then AB
is dened as [{t
n
+u
n
}]. So, we must check that [{x
n
+y
n
}] = [{t
n
+u
n
}].
That is {x
n
+y
n
} {t
n
+u
n
}. So, we must check that given > 0 there
exists an N N such that for all n, m N, |x
n
+ y
n
t
m
u
m
| < .
Since [{x
n
}] = [{t
n
}], by our denition {x
n
} {t
n
} and hence we
can nd an N
1
N so that for all n, m N
1
, |x
n
t
m
| < /2. Similarly
we can nd an N
2
N so that for all n, m N
2
, |y
n
u
m
| < /2. Let
N = max{N
1
, N
2
} and then for any n, m N,
|x
n
+ y
n
t
m
u
m
| |x
n
t
m
| +|y
n
u
m
| <

2
+

2
= ,
the middle inequlity being the triangle inequality. This proves what
we set out to prove.
We also have a map f : Q R, sending a Q to the equivalence
class of the CS, x
n
= a for all n (That this is indeed a CS was checked
in Exercise 6). The function f is one-one and thus we can identify Q
as a subset of R, f(a + b) = f(a) f(b) and f(ab) = f(a) f(b).
Thus now, we may drop our complicated notation and write the usual
symbols for addition and multiplication. Also, we can check all the
standard properties of this operation. We also dene an ordering on R
as follows. If {x
n
} and {y
n
} are CS, and A = [{x
n
}] and B = [{y
n
}],
then A < B, if there exists an > 0, rational number and an N N
such that for all n, m N, y
m
x
n
> . Again make sure that this
is well dened. If we dene A B as usual, by saying that A < B or
A = B, then one can easily check the following. A = {x
n
} B = {y
n
}
if and only if for any positive rational number , there exists an N such
that for all n, m N, y
n
x
m
> .
Exercise 7. Show that if {x
n
} is CS of rational numbers and {y
n
} is
a subsequence of {x
n
} (which I have noted also must be a CS), then
show that {x
n
} {y
n
}.
5.3. Supremum and Inmum. Now we have all the machinery re-
quired to prove the following important properties of the real numbers.
The proofs, while elementary, are fairly subtle. You may try it for fun.
CONSTRUCTION OF NUMBER SYSTEMS 23
If you do not succed, do not be discouraged. Again, as a warm up
exercise, let us prove an easy lemma.
Lemma 5.4. If A < B are two real numbers, there is a rational number
q, such that A < q < B. Remember that a rational number q as an
element of R is just the equivalence class of the Cauchy sequence with
all terms equal to q.
Proof. Write A = [{x
n
}] and B = [{y
n
}]. Then by denition, A < B
means, there exists a positive rational number and an N
1
N such
that for all n, m N
1
, y
n
x
m
> . Since {x
n
} is a CS, there exists an
N
2
so that for all n, m N
2
, we have |x
n
x
m
| < /2. Similarly, there
exists N
3
N so that for all n, m N
3
, |y
n
y
m
| < /2. If we choose
N = max{N
1
, N
2
, N
3
}, then for all n, m N, all the above three
inequlaities hold. Now, let q = (x
N
+ y
N
)/2. I claim, this rational
number is between A and B. We will show that q < B, and the
inequality A < q will be similar.
For n N, we should compute y
n
q.
y
n
q = y
n

x
N
+ y
N
2
=
y
n
x
N
2
+
y
n
y
N
2
Since n, N N N
1
, we have y
n
x
N
> . On the other hand, since
n, N N N
3
, we have |y
n
y
N
| < /2. This implies by lemma 2.7,
y
n
y
N
> /2. Substituting these in the above, we get,
y
n
q >

2


4
=

4
This shows that q < B. Similarly one shows that A < q.
Denition 7. Let S R and a R. Then a is called a lower bound for
S if a s for all s S. Similarly an element b R is called an upper
bound for S if s b for all s S.
Now, we prove the most important property of real numbers, from
which all the other subtle properties can be deduced. The proof is long
and so take your time mulling over the steps.
Theorem 5.1. Let S R be a non-empty subset and assume that it
has a lower bound M. Then there exists a real number (called the
inmum of S) such that for any s S, s and if x R is such that
s x for all s S, then x.
Proof. First, notice that we may assume the M that we have can be
assumed to be rational, by choosing a smaller number by the previous
24 N. MOHAN KUMAR
lemma. Our aim is to construct as in the theorem and by now we
should have the feeling that in general it is going to be a real number,
but not a rational number. Thus to get there, we must construct an
approriate Cauchy sequence, which will be a real number using our
relation. Let us during the proof, call a number a R a lower bound
for S if a s for all s S. So M is a lower bound. Now pick any
s S. (This is where we use the fact that S is not empty). Again, easy
to see that we can pick a rational number N > s. Let q = N M Q.
Call M
1
= M. Consider the rational number M + (q/2). Then there
are two possibilities. Either this number is a lower bound for S or not.
If it is, call M
2
= M +(q/2). If it is not call M
2
= M
1
. Let us see what
we have.
By choice, M
2
is still a lower bound for S, M
1
M
2
, M
2
M
1
q/2
and there exists an s S such that s M
2
< q/2.
Now we repeat the process. That is consider M
2
+(q/4). Then again
there are two possibilities. Either it is a lower bound for S or not. If
it is call M
3
= M
2
+(q/4) or else call M
3
= M
2
. Again notice that M
3
is a lower bound for S, M
2
M
3
, M
3
M
2
q/4 and there exists an
s S with s M
3
< q/4. We continue this process by replacing q/4
with q/8, q/16 etc. to get a sequence {M
n
}.
Next we check that this is indeed a Cauchy sequence and = [{M
n
}]
is an inmum for S. Let us repeat the basic properties of this sequence
of rational numbers.
M
n
is a lower bound for S n N
M
1
M
2
M
n
M
n+1

M
n+1
M
n

q
2
n
n N (2)
s
n
S such that s
n
M
n
<
q
2
n1
, n N
If one wants to be very precise, this is how the above should be
phrased. We would like to construct recursively M
n
for n N satisfy-
ing the above properties. We are given a lower bound M which we call
M
1
. Since S = , pick an element s S and let N be a rational number
such that N > s. Let q = N M a positive rational number. Then we
construct M
2
as described above, satisfying the properties in equation
2 for n = 1, 2. So assume that we have constructed M
1
, M
2
, . . . , M
n
satisfying the above properties. We wish to construct an M
n+1
satis-
fying the above. So, we consider M
n
+
q
2
n
. If this number is a lower
bound for S, we call this M
n+1
. Otherwise we let M
n+1
to be the same
as M
n
. So, by choice, we still have M
n+1
a lower bound for S and we
CONSTRUCTION OF NUMBER SYSTEMS 25
also have M
n
M
n+1
. Since M
n+1
M
n
= 0 or q/2
n
, we also have
M
n+1
M
n
q/2
n
. So, we only need to verify the last requirement.
For this again we look at the two cases when M
n+1
= M
n
+
q
2
n
or
M
n
. In the rst case, we take s
n+1
= s
n
. Then
s
n+1
M
n+1
= s
n
M
n

q
2
n
<
q
2
n1

q
2
n
=
q
2
n
,
which is what we want. In the case, M
n+1
= M
n
, we know that M
n
+
q
2
n
is not a lower bound for S and thus there exists an s
n+1
S such that
s
n+1
< M
n
+
q
2
n
. Since M
n
is a lower bound, we have,
M
n
s
n+1
< M
n
+
q
2
n
.
Subtracting M
n
, we get s
n+1
M
n
<
q
2
n
, which is what we want. Thus
recursively we can dene the sequence M
n
, satisfying the properties
stated above.
If n m, we have,
|M
n
M
m
| = M
n
M
m
= (M
n
M
n1
) + (M
n1
M
n2
) + + (M
m+1
M
m
)

q
2
n1
+
q
2
n2
+ +
q
2
m
(3)
=
q
2
m

1 +
1
2
+ +
1
2
nm2
+
1
2
nm1

<
q
2
m
2 =
q
2
m1
Given > 0, choose an N N so that q/2
N1
< . This can be done
by lemma 2.6. If n m N, by equation 3, we get,
|M
n
M
m
| <
q
2
m1

q
2
N1
< .
This proves that {M
n
} is a Cauchy sequence and so = [{M
n
}] is a
real number.
Next, we check that is a lower bound for S. For this, let s S.
and let > 0 be given. Then we may choose an N
1
N so that
q/2
N
1
1
< /2 as before. Since M
N
1
is lower bound for S, we have
M
N
1
s. If we write s = [{u
n
}] for a Cauchy sequence {u
n
}, there
exists an N
2
N so that for all n N
2
, u
n
M
N
1
> /2, by
denition of inequality discussed earlier. Now, let N = max{N
1
, N
2
}.
If n, m N, we get, using equation 3,
u
n
M
m
= u
n
M
N
1
+ M
N
1
M
m
>

2

q
2
N
1
1
> .
26 N. MOHAN KUMAR
This proves that s.
Finally we check that is an inmum for S. So, let x be a lower
bound for S. We must show that x . We prove this by contradiction.
If this is not true, then x > and let us write x = [{x
n
}] as usual. So,
there exists an > 0 and N
1
N so that for all n, m N
1
,
x
n
M
m
> . (4)
Choose as before an N
2
so that q/2
N
2
1
< /3. Then by equation 2, we
have an s S so that s M
N
2
< q/2
N
2
1
, /3. Writing this s = [{u
n
}],
there exists an N
3
N so that for all n N
3
, u
n
M
N
2
/3 < /3.
We rewrite this as,
M
N
2
u
n
>
2
3
(5)
Now, let N = max{N
1
, N
2
, N
3
}. Then for all n, m N, we have, using
equations 2, 4 and 5,
x
n
u
m
= (x
n
M
N
) + (M
N
M
N
2
) + (M
N
2
u
m
)
> + 0
2
3
=

3
.
Thus by denition, we get that x > s. This is a contradiction, since x
was a lower bound and hence x s for all s S.

An identical argument can be used to prove the following,

Theorem 5.2. Let S R be a non-empty subset and assume it has an
upper bound. Then there exists a real number (called the supremum
of S) such that for any s S, s and if x R is such that s x
for all s S, then x.
Now, I repeat the properties of inmum and supremum explicitly, so
that we can use it again. These properties are easy, but the existence
as proved above is not.
Lemma 5.5. Let S be a non-empty set bounded below and let be its
inmum. Then is a lower bound for S. Further, given any positive
number , there exists an s S such that s < + .
Similarly, if S is non-empty set bounded above, then its supremum
is an upper bound for S. Further given any positive number there
exists an s S so that < s .
Proof. The proof is easy. We already know that the inmum is a lower
bound. Now, assume that we are given > 0. Then by the property
of inmum, + can not be a lower bound for S. But this means
CONSTRUCTION OF NUMBER SYSTEMS 27
precisely that there is an s S with s < +. The proof for supremum
is identical.
The phrase greatest lower bound (glb for short) is used interchageably
with inmum and least upper bound (lub for short) for supremum.
Next we dene Cauchy Sequences exactly as before, but with real
numbers instead of rational numbers.
Denition 8. A sequence of real numbers {A
n
} is a CS, if for any
positive real number there exists an N N so that for all n, m N,
|A
n
A
m
| < .
Of course, we could have dened a CS of real numbers with a
rational number.
Denition 9. A sequence of real numbers {A
n
} is a CS, if for any
positive rational number there exists an N N so that for all n, m
N, |A
n
A
m
| < .
Lemma 5.6. The two dentions above are equivalent. That is, a se-
quence {A
n
} of real numbers is a CS in the rst denition if and only
if it is so in the second denition.
The proof is an easy application of lemma 5.4.
Theorem 5.3. Let {x
n
} be a Cauchy sequence of real numbers. Then
there exits a unique real number x such that given any > 0, real
number, there exists an N N, such that for all n N, |x
n
x| < .
The above theorem states that if we repeated the construction of
reals from rationals, then we get nothing new. This is usually termed
as the completeness of the real numbers.
Proof. We rst look at non-empty sets, S
i
for i = 0, 1, . . . as follows.
Let S
0
= {x
1
, . . . , x
n
, . . .}, S
1
= {x
2
, x
3
, . . . , x
n
, . . .} etc. These are sets
and not sequences (though they look like sequences). So, S
n
would be
the set of elements of the sequence starting from x
n+1
. These are all
non-empty and as we have seen in lemma 5.1, these are bounded, since
{x
n
} is a CS. Thus by theorem 5.1, we have
n
the inmum of S
n
.
First notice that
n

n+1
for all n. This is because, since
n
is a
lower bound for the set S
n
and S
n+1
S
n
, it is clearly a lower bound
for S
n+1
and thus by property of inmum, we must have
n

n+1
.
Now, since all the s are bounded above, since {x
n
} is CS, we have a
supremum for the non-empty set, {
0
, . . . ,
n
, . . .}. Let this supremum
be x. I claim that this x has the property stated in the theorem.
28 N. MOHAN KUMAR
So, assume that we are given an > 0 and let = /3. Then there
exists an N
1
N so that for all n, m N
1
,
|x
n
x
m
| < . (6)
Since x is the supremum of the set {
n
}, by lemma 5.5, there exists an
element say
N
2
so that, x <
N
2
x. Since
n

n+1
for all n, we
may replace N
1
, N
2
by N = max{N
1
, N
2
}. Thus, we have
x <
N
x. (7)
Finally, since
N
is the inmum of the set S
N
, by lemma 5.5 again,
there exists an element x
l
S
N
so that

N
x
l
<
N
+ (8)
Notice that since x
l
S
N
, l > N. Putting these three equations
together, we get, for n N,
|x x
n
| = |x
N
+
N
x
l
+ x
l
x
n
|
|x
N
| +|
N
x
l
| +|x
l
x
n
|
< + + =
Now, we only need to prove that this x is unique. I leave it as an
exercise.

Denition 10. If {x
n
} is a Cauchy sequence of real numbers and x is
the unique real number we found in the above theorem, we call x the
limit of the sequence {x
n
} and write x = limx
n
. For a set S, if it has
an inmum, we denote it by inf S and similarly, if it has a supremum
denote it by sup S.
Exercise 8. (1) Let x
1
x
2
x
3
be a sequence (of real
numbers) which is bounded above. Show that {x
n
} is a CS and
limx
n
= sup{x
1
, x
2
, x
3
, . . .}.
(2) Let {x
n
} be a CS and assume that m is a lower bound for the set
{x
n
} and M an upper bound. Then show that m limx
n
M.
Theorem 5.4. Let S be an innite bounded set. That is, S is innite
and there exists an M > 0 such that for all s S, |s| M. Then
there exists an x R such that, for any > 0, there are innitely
many elements s S such that |x s| < .
The proof is on similar lines as above and I will not prove it.
By now, it should be clear to you, that while the arguments are not
dicult, they can be rather tedious. These arguments repeat them-
selves in several guises.
CONSTRUCTION OF NUMBER SYSTEMS 29
6. Continuous functions
In this section we study continuous functions from R to R. You
should do these for functions dened in an open interval, but I will not
go into it in any detail since it is mostly routine. We start with the
denition of a continuous function at a point x R.
Denition 11. A function f : R R is called continuous at a point
x R if for any CS, {x
n
}, with limx
n
= x, the sequence {f(x
n
)}
is a CS. A function is continuous on R (or any open interval) if it is
continuous at every point of R (respectively, at every point of the open
interval).
As usual, we will write [a, b] for a < b, the closed interval, which is
the set of all x R such that a x b. Similarly, open and half-closed
intervals can be dened.
Before we give the usual examples of continuous functions, let me
prove some easy lemmas.
Lemma 6.1. If f, g are continuous functions on R, so is f +g and fg,
where (f + g)(x) = f(x) + g(x) and (fg)(x) = f(x)g(x) for all x R.
Proof is just an application of lemma 5.2.
Lemma 6.2. If f is continuous at x and {x
n
} is any CS with limx
n
=
x, then limf(x
n
) = f(x).
Proof. Consider a new sequence {y
n
} dened as follows. For n = 2m
1, dene y
n
= x
m
and for n = 2m dene y
n
= x. It is easy to check
that then {y
n
} is a CS and limy
n
= x. So, by continuity, {f(y
n
)} is
a CS. Let y = limf(y
n
). So, given any > 0, there exists an N such
that for n N, we have |f(y
n
) y| < . But, if n N and n is
even, since y
n
= x, we get |f(x) y| < . But, since was arbitrary,
this can happen only if y = f(x). Now, taking 2m 1 = n N, one
has y
n
= x
m
and so we get |f(x
m
) f(x)| < , which implies that
limf(x
n
) = f(x).

Example 2. (1) The identity function Id : R R is continuous.

That is, the function Id(x) = x is continuous.
(2) By the lemma above, we see that the function f(x) = x
n
is
continuous for any n N.
(3) Polynomial functions are continuous. These are functions of the
form, f(x) = a
0
x
n
+a
1
x
n1
+ +a
n
, where n N and a
i
R.
Next let me prove a fundamental theorem about continuous func-
tions, which is frequently used as the denition, often called the -
denition of continuity.
30 N. MOHAN KUMAR
Theorem 6.1. Let f be a function (from R or an open interval to
R). Then f is continuous at a point x if and only if, given any > 0,
there exists a > 0 such that for any t such that |t x| < , we have
|f(t) f(x)| < .
Proof. First let us assume that f is continuous. The proof is by contra-
diction. So, assume that we are give an > 0 and assume that there is
no > 0 satisfying the above condition. Then for any n N, there ex-
ists an x
n
such that |x
n
x| < 1/n and |f(x
n
)f(x)| . Consider the
sequence {x
n
}. It is immediate that {x
n
} is a CS and limx
n
= x. Since
f is continuous, we must have {f(x
n
)} a CS and limf(x
n
) = f(x). But,
this means that for all n >> 0, |f(x
n
) f(x)| < , which contradicts
the previous inequality.
To prove the converse, assume we are given a CS {x
n
} with limx
n
=
x. Given an > 0, we have a > 0 so that for all t with |t x| < ,
|f(t) f(x)| < /2. Since limx
n
= x, there exists an N so that for all
n N, |x
n
x| < . So, if n, m N, we get,
|f(x
n
) f(x
m
)| = |f(x
n
) f(x) + f(x) f(x
m
)|
|f(x
n
) f(x)| +|f(x
m
) f(x)|
<

2
+

2
=
This shows that {f(x
n
)} is a CS, proving continuity.
We next prove the last theorem for this topic, used often by all of
you in your calculus classes.
Theorem 6.2 (Intermediate Value Theorem). Let f be a continuos
function (on R or in the closed interval [a, b]) where a < b. Then given
any real number u between f(a) and f(b), there exists a real number
a c b such that f(c) = u.
Proof. If f(a) = f(b), then u = f(a) and there is nothing to prove. So,
we may assume f(a) = f(b). So, we will assume that f(a) < f(b), the
case f(a) > f(b) being similar. We may also assume that f(a) < u <
f(b).
We consider the following set.
S = {x [a, b] | z [a, x], f(z) u}.
Notice that S is bounded (below by a and above by b) and it contains
a and hence non-empty. Hence by theorem 5.2, it has a supremum,
which we call c. Since a c b, if we show that f(c) = u, we would
have proved the theorem.
CONSTRUCTION OF NUMBER SYSTEMS 31
First we show that f(c) u. The proof is by contradiction. So,
assume that f(c) > u. Notice that c > a, since otherwise, c = a and
then f(c) = f(a) < u. Let 0 < = c a. Then for all n >> 0,
natural numbers we have a c 1/n < c. Since c was the supremum
of the set S, there must exist an x
n
S with c 1/n < x
n
c. It
is easy to see that {x
n
} is a CS and limx
n
= c. So, by continuity we
have limf(x
n
) = f(c). But, since f(x
n
) u becuase x
n
S, we see
that u is an upper bound for the set {f(x
n
)} and hence by exercise 8,
f(c) u. This is a contradiction.
Next we show that the assumption f(c) < u also leads to a contra-
diction and then we would have proved the result. So, assume that
f(c) < u. Choose an > 0 such that f(c) + < u. (For example,
you could choose = (u f(c))/2). Then there exists a > 0, by
continuity of f such that for all t with |t c| < , |f(t) f(c)| < .
We will then show that, if t c + /2, then f(t) u. If t < c, there
exists an x S such that t < x c and we know that for any such t,
f(t) u by dention of our set S. If t = c, this is our assumption. If
c < t c+/2, we know that |f(t)f(c)| < and thus f(t)f(c) < .
This means f(t) < f(c) + < u. Thus, we see that c +/2 S by our
dention of the set S. But, then c < c + /2 can not be a supremum
for S, since it is not an upper bound for S. This contradcition proves
the theorem.