You are on page 1of 10

18.

014 Calculus with theory, Fall 2012


What are the real numbers, really?
0. A bit about sets and functions
A fundamental notion in mathematics, perhaps the fundamental notion, is that of a set. Please read the
(very short) chapter on set theory in Apostols book and do the exercises (Introduction, Part 2).
Let us also recall what a function is.
Denition 0.1. Suppose A and B are sets (they dont have to be distinct). A function from A to B is a
rule which assigns to each element a of A a unique element f(a) of B. It is denoted f : A B. The set A
is called the domain, and B the codomain of f.
Given an element b B, if there exists an a A such that f(a) = b, we say that b is in the image of f.
Note that the image of f and its codomain need not coincide, although the image is necessarily a subset of
the codomain.
Example 1. Let A = B be the set of all people (say, all people that ever lived); the rule which
assigns to each person their mother is a function (because everybody has exactly one mother).
The rule that assigns to every person their DNA is a function. What is its codomain?
Is the rule which assigns to a person their car a function? Why, or why not? Make it a function.
Denition 0.2. Let f : A B be a function.
(1) If f(x) = f(y) implies that x = y, f is called injective, or one to one.
(2) If for every y B, there is an x A such that f(x) = y, f is called surjective or onto. (It is only
in case f is surjective that the image of f equals its codomain.)
(3) If f is both injective and surjective, then f is called bijective.
Are the functions above injective or surjective?
1. The natural numbers
We assume the existence of a set N, whose elements we call the natural numbers, and which satises the
following axioms (called Peanos axioms).
Axiom 1. There exists an element one of N denoted 1.
Axiom 2. For each n N, there exists a unique element (n) N, called the successor of n. In
other words, there is a function : N N.
Axiom 3. The element 1 is not a successor of any element of N (i.e. 1 is not in the image of ).
Axiom 4. For any two n, m N, if (n) = (m) then n = m (i.e. is injective).
Axiom 5 aka Principle of Mathematical Induction. Let S be a subset of N, with the following
properties:
(a) 1 S, and
(b) If n S then (n) S.
Then S = N.
There is a standard way to denote the elements of N (Arabic numerals) as 1, 2 = (1), 3 = (2)... Of
course, this is not the only possible representation. The Romans used, I, II, III, IV, V, . . . to denote the
natural numbers. A set theoretical representation of N is as {{}, {, {}}, {, {, {}}}, . . . }.
From these axioms we go on to prove lots of properties of the natural numbers, and construct more and
more numbers.
Theorem 1.1. For any n N, n = (n).
Proof. This will be a proof by mathematical induction, which means, a proof using Axiom 5. Let S denote
the set of elements n of N, such that n = (n).
(a) By Axiom 3, 1 S.
(b) Suppose n S, i.e. n = (n). We need to show that (n) is also in S, i.e. we need to how that
(n) = ((n)). Suppose otherwise, i.e. suppose (n) = ((n)). Set m = (n) on the right hand
side to get (n) = (m). By Axiom 4, it follows that n = m, i.e. n = (n), which contradicts our
assumption that n is an element of S. Therefore, (n) S as well.
1
2
Therefore, by the principle of mathematical induction, the set S is the set of all natural numbers.
Remark 1.2. For part (b) we used something called proof by contradiction, which goes as follows. Say you
want to prove that statement P is true. If you can show that the negative of P implies a contradiction,
it means that P is true. (Note, a contradiction is when both a statement and its negative are true at the
same time, which is excluded by the so-called law of bivalence.) Warning: be very careful when negating
statements!
Theorem 1.3. If n = 1, then there exists an element m N such that n = (m).
Proof. Let S be the set which is the union of {1} and the subset of N consisting of all those n which are in
the image of . Symbolically, S = {1} {n N|m N : n = (m)}.
(a) Clearly, 1 S, by denition.
(b) Suppose n S. We need to show that (n) S. But (n) is in the image of , so it is in S by
denition.
By the principle of mathematical induction, S = N.
Remark 1.4. The number m such that n = (m) is unique, because of Axiom 4, and it is called a predecessor
of n.
1.1. Addition.
Theorem 1.5 (Addition law for N). Let n, m be natural numbers. There is a unique way to assign to n, m
another natural number, denoted n + m, such that the following properties are satised:
(1) n + 1 = (n)
(2) n + (m) = (n + m).
Proof. Fix a natural number n. Then the following lemmas will tell us how to dene n + m for all m N,
and since we can x n to be any natural number, it means (after the lemmas) we can dene n + m for all
n, m N.
Lemma 1.6 (Uniqueness). There is at most one way of dening n + m for all m N, such that the above
properties hold.
Proof of the Lemma. Suppose there are two ways to dene n + m, as a(m) and as b(m). This means, there
are numbers a(m) such that
a(1) = (1)
a((m)) = (a(m))
and likewise for b(m).
Let S denote the set of those natural numbers for which a(m) = b(m). By Property 1 together with
Axiom 2, we know that 1 S. Now suppose some m is in S, i.e. a(m) = b(m). Then we have
a((m)) = (a(m)) = (b(m)) = b((m)),
so (m) S. By Axiom 5, S = N, so in fact for all natural numbers m there is at most one unique way of
dening n + m.
Now we need to show that at least one possible way exists.
Lemma 1.7 (Existence). For each n, it is possible to dene n + m for all m N, in a way satisfying the
properties.
Proof of the Lemma. We will use mathematical induction again. Suppose S is the set of natural numbers
m for which n + m can be dened (uniquely, by the previous lemma). We know how to dene n + 1, it is
required to be (n); so 1 S. Suppose we can dene n +m for some m. We can use Property (2) to dene
n + (m) = (n + m),
so (m) S as well. Hence S = N and we are done.

Theorem 1.8 (Properties of addition). Let n, m, k N.


(1) Associative Law for Addition. (n + m) + k = n + (m + k)
3
(2) Commutative Law for Addition. n + m = m + n
(3) n + m = m
(4) Cancellation Law. If m = k then n + m = n + k. Equivalently, if n + m = n + k then m = k.
Theorem 1.9 (Ordering on N). (1) Trichotomy. For any n, m N, exactly one of the following holds:
(a) n = m
(b) There is u N, such that m = n + u. In this case we say m > n or n < m.
(c) There is v N, such that n = m + v. In this case we say m < n or n > m.
(2) n + m > m
(3) Transitivity. If n < m and m < k, then n < k
(4) If m < k then m + n < k + n
(5) If m k then m < k + 1
Proof. Homework.
Remark 1.10. By Trichotomy, the negation (i.e. opposite) of n < m is n m.
Theorem 1.11 (The well-ordering principle). Every nonempty subset of N has a smallest element.
Proof. Let S N be a subset such that S does not have a smallest element. Then 1 cannot be in S, because
1 is smaller than every natural number. Let
P = {n N|s S, n < s}.
We just argued that 1 P. Now suppose n P, so n < s for all s S. Is (n) = n +1 also in P? Suppose
not. This would mean that for some t S, t (n) = n + 1. Since S has no smallest element, t is not
the smallest element of S, so there exists an element p S such that p < t. This implies, p < n + 1, or
equivalently, p n. But we assumed that every element in P is strictly smaller than every element in S.
This is a contradiction, so our assumption that (n) = n + 1 is not in P is wrong, so (n) must be in P.
By the principle of mathematical induction, it follows that P = N, from where it follows that S must be
empty.
1.2. Multiplication.
Theorem 1.12 (Multiplication law). Let n, m be natural numbers. There is a unique way to assign to n, m
another natural number, denoted n m or nm, such that the following properties are satised:
(1) n 1 = n
(2) n (m) = n m + n.
Proof. As for addition, we rst show there is at most one way to assign nm, and then show that it exists.
The details are left as an exercise.
Theorem 1.13 (Properties of multiplication). Let n, m, k N.
(1) Associative Law for Multiplication. (nm)k = n(mk)
(2) Commutative Law for Multiplication. nm = mn
(3) Distributive Law. n(m + k) = nm + nk
(4) Order-preserving:
(a) Cancellation Law. m = k if and only if mn = kn
(b) m < k if and only if mn < kn
(5) If n > k, m > l then nm > kl.
Proof. Homework and/or exercise, except, lets say, order-preserving.
Suppose m = k. Let S be the set of those n N, such that mn = kn. Certainly 1 S, since by
construction 1m = m, 1k = k, and by assumption m = k, so 1m = 1k. Now suppose n S, so that
mn = kn. Then
m(n) = mn + m = kn + k = k(n),
so that (n) S. Note that we have used Property (2) of multiplication, as well as the cancellation
law for addition. Therefore S = N.
4
Suppose m < k. Let S be the set of those n N, such that mn < kn. Certainly 1 S by
construction. Suppose n S, so mn < kn. Then
m(n) = mn + m < kn + k = k(n),
so that (n) S. Note that we have used Property (2) of multiplication, as well as the property of
addition that if x < y then x + z < y + z. Therefore S = N.
Same goes for m > k.
In the other direction:
Suppose mn = kn. By the Trichotomy law, there are exactly three, mutually exclusive options for
m and k: either m = k, m < k, or m > k. If m < k then mn < kn by the above argument, but it
cannot be that mn < kn and mn = kn at the same time (also by Trichotomy). Likewise for m > k.
Therefore it must be that m = k.
The proofs for the other two are exactly the same.

2. Integers
We dened addition in N by saying that given some elements m, k N, there is a unique n = m + k
satisfying certain properties. Suppose now we ask the question, given m and n, is there a k N such that
n = m + k? The answer to this question in general is no; we must have n > m for such a k to exist. The
integers arise when we try to enforce that the answer to this question be yes. What we want to do is subtract;
to do so, we need to extend N to include negative numbers.
Denition 2.1. Let Z be the set composed of the following elements:
The natural numbers N,
For each natural number n, a (unique) element n,
An element 0 (zero).
Theorem 2.2. There is a unique way to dene a commutative and associative addition law + on the integers
Z such that
if m, n N Z, then m + n in Z is the same as m + n N Z,
for any a Z, a + 0 = a,
n + (n) = 0
if n, m N, (n + m) = (n) + (m).
Proof. First lets dene the law. Let m, n N
m + 0 = 0 + m = m
(m) + 0 = 0 + (m) = (m)
m + n = m + n N Z
(m) + (n) = (m + n)

m + (n) = (n) + m =

if m > n, u such that m = u + n m + (n) = u


if m < n, u such that n = u + m m + (n) = u
Now we need to show this is well-dened (meaning, for each a, b Z there is a unique a+b), is commutative
(i.e. a + b = b + a), and is associative (i.e. (a + b) + c = a + (b + c)).
Well-dened-ness is clear except possibly for m + (n). For this it suces to show that if m < n, then
there is a unique u such that n = u + m. This follows from the cancellation law for addition on N: if
u + m = u

+ m then u = u

Commutativity is clear by construction.


Associativity is left as homework. You need to consider several cases.
Theorem 2.3 (Ordering on Z). The relation < on Z dened as follows:
For m, n N, m < n in Z if and only m < n in N
0 < n for all n N Z
(m) < n for all m, n N
(m) < (n) (in Z) if and only if n < m (in N)
5
satises the following properties, for all a, b, c Z:
(1) If a < b and b < c then a < c
(2) Exactly one of a = b, a < b, or b < a holds
(3) a < b if and only if a + c < b + c for all c Z.
Proof. Homework.
Theorem 2.4 (Multiplication in Z). There is a unique way to dene a commutative and associative multi-
plication in Z which extends the multiplication on N, satises the distributive law with respect to addition,
and such that
0a = 0 for all integers a.
n(m) = (nm)
(n)(m) = nm.
Proof. Homework.
Theorem 2.5. There is a bijection : N N, such that (n) = n, (n) = n, and 0 = 0.
Denition 2.6. For a, b Z, a b is dened to be a + (b).
Denition 2.7. An integer n is said to be positive if n > 0 and negative if n < 0. The element zero is
neither positive nor negative.
Theorem 2.8. Suppose x, y, z Z. If x < y, then
(1) xz < yz if and only if z > 0.
(2) xz = yz if and only if z = 0.
(3) xz > yz if and only if z < 0.
Proof. Homework.
3. Rational numbers
Just like the negative numbers are additive inverses of the positive, we would like to have multiplicative
inverses. For that, we need to extend the integers even further.
Denition 3.1. The set Q of rational numbers consist of fractions
a
b
, where b is a natural number and a is
an integer, subject to the relation
a
b
=
c
d
ad = bc.
We think of the integers Z as the subset of Q consisting of fractions of the form
a
1
, so we denote those
fractions as a.
Corollary 3.2.
0
n
=
0
1
for any n N.
Theorem 3.3. The set Q has well-dened operations + and , such that
a
b
+
c
d
=
ad + bc
bd
,
a
b

c
d
=
ac
bd
.
Proof. We need to show that if
a
b
=
x
y
and
c
d
=
z
w
, then
a
b
+
c
d
=
x
y
+
z
w
and
a
b

c
d
=
x
y

z
y
.
By denition, if
a
b
=
x
y
and
c
d
=
z
w
, then ay = bx and cw = dz.
For addition, the question is whether
ad + bc
bd
=
xw + yz
yw
,
which is equivalent to asking whether
(ad + bc)(yw) = (bd)(xw + yz).
By distributivity, this is equivalent to whether
(ad)(yw) + (bc)(yw) = (bd)(xw) + (bd)(yz),
which by commutativity and associativity of multiplication is equivalent to
(ay)(dw) + (by)(cw) = (bx)(dw) + (by)(dz).
Now we use that ay = bx and cw = dz to deduce that indeed,
a
b
+
c
d
=
x
y
+
z
w
.
6
For multiplication, we need to show that
ac
bd
=
xz
yw
, or equivalently, that (ac)(yw) = (bd)(xz). But
(ac)(yw) = (ay)(cw) = (bx)(dz) = (bd)(xz).

Theorem 3.4. For any x, y, z, w Q, the following are satised:


(1) Commutativity. x + y = y + x, xy = yx
(2) Associativity. (x + y) + z = x + (y + z), (xy)z = x(yz)
(3) Distributivity. x(y + z) = xy + xz
(4) Identity elements. x + 0 = x and x1 = x.
(5) Negatives. For every x Q, there is a y Q such that x + y = 0. We will denote such a number y
as x.
(6) Reciprocals. For every non-zero x Q, there is a y Q such that xy = 1. We will denote such y as
1
x
. We will denote by
z
y
the rational number z
1
y
.
(7) (x) = x
(8)
1
1/x
= x
FYI: A set with two operations satisfying the above conditions is called a eld.
Proof. Homework.
Theorem 3.5. The rational numbers are ordered with respect to the relation <, where
a
b
<
c
d
i ad < bc.
Let x, y, z, w Q. Then
(1) x = y if and only if x + z = y + z for all z Q,
(2) x < y if and only if x + z < y + z for all z Q,
(3)
x
y
<
z
w
if and only if xw < yz.
(4) Let n, m, k, l N. Then x =
n
m
< 0, y =
k
l
> 0, and x < y. In other words, a negative rational is
smaller than a positive rational.
(5) Suppose x < y. Then
(a) xz < yz if and only if z > 0.
(b) xz = yz if and only if z = 0.
(c) xz > yz if and only if z < 0.
Proof. Homework.
Theorem 3.6. If x < y are arbitrary rational numbers, there exists a rational number z between them, i.e.
such that x < z < y.
Proof. For example, z =
x+y
2
is a rational number such that x < z < y.
Theorem 3.7. Let r be a rational number, r > 1. Then the set R = {r
n
|n N} is not bounded above, i.e.
for every integer m N, there exists an element of R larger than m.
Proof. We will use the lemma proven below, that for any rational number a, (1+a)
n
1+na. In particular,
for a = r 1, we get that r
n
1 + n(r 1).
Suppose we are given m N, and we try to nd n N such that r
n
> m. Let n be any natural number
greater than or equal to
m
r1
. (Such a number exists; for any rational number
p
q
Q, where p, q Z, there
is a natural number larger than
p
q
; for example, p if p is positive, or p if p is negative.) For such n we have
that
r
n
1 + n(r 1) > n(r 1) m.

Lemma 3.8. Let a be a rational number, and n a natural number. Then (1 + a)


n
1 + na.
Proof. When n = 1, the inequality reduces to 1 + a 1 + a which is trivially true. Suppose the inequality
holds for some n; we want to show that it holds for n + 1. We have
(1 + a)
n+1
= (1 + a)
n
(1 + a) (1 + na)(1 + a) = 1 + na + a + na
2
1 + (n + 1)a,
since na
2
0. Therefore, the inequality holds for all n, by the principle of mathematical induction.
Remark 3.9. Show that na
2
0 for all a Q and n N!

7
4. Real numbers
We postulated the natural numbers, and from them we constructed the integers by adjoining additive
inverses and zero; from the integers, we constructed the rational numbers by adjoining multiplicative inverses.
The construction of the real numbers from the rationals follows a somewhat dierent pattern. What we want
to adjoin now are so-called limits of rational numbers. We will come back to this point later.
4.1. Dedekind cuts.
Denition 4.1. A set of rational numbers is called a Dedekind cut if
(1) is nonempty,
(2) The complement Q\ is nonempty,
(3) If x then every y Q such that y < x is also in , and
(4) does not contain a greatest element.
Exercise 1. Show that any rational number x Q denes a Dedekind cut, denoted x

, by x

=
{y Q|y < x}.
Show that the set {y Q|y
2
< 2} is a Dedekind cut.
Show that the set Q\ is innite.
Remark 4.2. To aid your intuition, think of a Dedekind cut as representing the least upper bound of the
set , whatever that means. (We will learn what it means in just a little bit).
Denition 4.3 (Ordering). Let and be cuts. We say < if is a strict subset of , and = if they
are equal as sets.
Theorem 4.4 (Trichotomy). Given two cuts , , exactly one of the following holds:
= , < , or > .
Proof. Suppose there exists an element x \ (i.e. x and x / ). Let y be any element of . If y > x,
then x < y, so x would need to be an element of by the denition of a cut. But x / , from which we
deduce that every element y of is strictly less than x, which in turn implies that y . In other words,
, i.e. < .
In case there exists no element x \, either = or y \. If the latter is true, the above argument
shows (by symmetry) that < .
Theorem 4.5. If < and < , then < .
Proof. Follows from transitivity for the subset relation.
Theorem 4.6 (Addition of Dedekind cuts). Let and be cuts, and let + denote the set of rational
numbers of the form x +y, where x and y . Then + is itself a cut. Moreover, no element of +
can be written as z + w, where z / and w / .
Proof. (1) Since is a cut, it is nonempty, so there is an element x . Likewise, there is a y .
Therefore, + is nonempty, as it contains the sum x + y.
(2) Suppose z / and w / . Then z > x for all x , and w > y for all y . Therefore z +w > x+y
for all x and y , so it cannot be that z +w = x +y for some such x, y. In particular, z +w is
an element of the complement of + .
(3) Suppose x +y +, with x and y . Let z < x +y be an arbitrary rational number smaller
than x+y. Then z+(y) < x, so z+(y) . But then, by construction, (z+(y))+y = z +.
(4) Let x+y +, with x and y , be an arbitrary element. Since is a cut, it doesnt contain a
greatest element, so in particular, x is not greatest, and there exists z , z > x. Then z +y > x+y
is also an element of +. For every element of + , we showed there is a greater element also in
+ , so it doesnt contain a greatest element.
Therefore, + satises the properties for being a Dedekind cut. (The proof of (2) contains the proof
that no element of + can be written as z + w, where z / and w / .)
Theorem 4.7 (Properties of addition of Dedekind cuts).
(1) Commutativity. + = +
(2) Associativity. ( + ) + = + ( + )
8
(3) < if and only if + < +
(4) If , are cuts, there is a unique cut such that + = .
(5) 0

+ = .
Proof. Homework.
Denition 4.8. The negative of a cut is dened to be the unique cut such that +() = 0

. We will
denote by the cut + ().
Theorem 4.9. We can nd the negative of a cut by
= {x Q|y x, y 0

, x / }.
Moreover, () = , and for a rational number x, (x)

= (x

).
Proof. Homework.
Denition 4.10. A cut is positive if > 0

and negative if < 0

. The cut 0

is neither positive nor


negative.
Theorem 4.11. A cut is positive if and only if it contains a positive rational number.
Proof. Suppose is positive, i.e. > 0

. Then there exists an element x \0

, which has to be either zero


or positive, as 0

contains all negative rationals. If x > 0, we are done. Otherwise, x = 0. But is a cut, so
it has no greatest element, and in particular, 0 is not greatest. Therefore, there is an element x

, with
x

> x = 0, i.e. positive.


For the other direction, suppose contains a positive rational number x. Then contains all rationals
less than x. But we saw that all negative rational numbers are smaller than any positive rational number.
Hence 0

< .
Theorem 4.12 (Multiplication of Dedekind cuts). Let and be cuts. Dene a set as follows:
If , are both positive, then = {xy|0 x , 0 y } 0

,
If at least one of , equals 0

, then = 0

,
If is negative and positive, then = (()()),
If is positive and negative, then = (()()),
If , are both negative, then = ()().
Then is a Dedekind cut itself.
Proof. Homework.
Theorem 4.13 (Properties of multiplication of Dedekind cuts).
(1) Commutativity. =
(2) Associativity. () = ()
(3) Distributivity. ( + ) = +
(4) = 0

if and only if at least one of , equals 0

(5) Division. If , are cuts, = 0

, then there is a unique cut such that = .


(6) 1

= for all = 0

.
Proof. (4) By denition, if at least one of , equals 0

, then their product = 0

. For the other


direction, suppose , are both non-zero. We need to consider a couple of cases:
(a) , > 0

. There exist positive elements x of and y of . The product xy is positive by


Theorem 3.5 (5), and is an element of by denition. Therefore > 0

.
(b) For all other cases, it suces to show that if = 0

, then = 0

. (Show that this is sucient!)


Suppose < 0

. Then there exists a negative rational number y 0

\. Since 0

is a cut, y
is not a greatest element of 0

, so there exists z > y, z 0

. Then z y is an element of
by Theorem 4.9, which is positive since z > y. Hence, by Theorem 4.11, > 0

. Show that if
> 0

then < 0

.
(5) First we will show that there is at most one such , and then we will show that at least one exists
to complete the proof. Suppose
1
and
2
were two cuts such that
1
= =
2
. Then
0

=
1

2
= (
1

2
).
9
By (4), at least one of or (
1

2
) must be 0

. We assumed = 0

, so
1

2
= 0

. Therefore,

1
=
2
, i.e. we established uniqueness.
Now we need to show that such a exists. First, suppose is positive. Let
=

x
y
x , y /

.
This set is a Dedekind cut:
is nonempty: there exists an x , and y Q\ since and are cuts. For those x, y,
x
y
, so is nonempty.
Q\ is nonempty: Proof left as an exercise.
is closed under <: Suppose
x
y
, where x and y Q\. Let z Q be an
arbitrary rational number such that z <
x
y
. This means that zy < x, so zy , by the
corresponding cut property of . Therefore, z =
zy
y
by our denition of .
does not contain a maximal element: Let
x
y
, with x and y Q\ be an arbitrary
element of . We need to nd a larger element in . Now because is a cut, it has no
maximal element, so there is an element x

such that x

> x. Then
x

y
is an element
of which is larger than
x
y
.
Now we need to show that = , which amounts to showing that and . Suppose,
for now, that is also positive.
: Then , as we constructed it, is positive, as it contains a positive rational number.
For positive cuts, we dened the product to be the set
{wz|0 w , 0 z } 0

.
The elements of are all of the form
x
y
, where x and y / . Hence, the positive
elements of are of the form w
x
y
, where y , w / , and x . In particular, y < w,
so that
y
w
< 1, which implies that
wx
y
< x, so that w
x
y
.
: Let x be arbitrary. We need to show that x is also an element of . For this,
it suces to nd rational numbers x

, y, w, such that x < w


x

y
and x

, y / , w .
Let x

be any element larger than our given x (it exists, because is a cut). Let
r =
x

x
. We use the following lemma, proved below, to nd suitable y, w.
Lemma 4.14. Let r > 1 be a rational number, and let be a Dedekind cut. Then there
exist w and y / , such that
y
w
r.
Complete the proof for the cases where and are not positive!

Proof of the lemma. Suppose otherwise, i.e. suppose for all w and y / , we have that
y
w
> r. We ask,
is rw an element of or not? If rw / , then we can take y = rw and get that
y
w
=
rw
w
= r > r, which
is impossible. Therefore, it must be that rw is not an element of for every w . Consequently, the set
R = {r
n
w|n N} is a subset of . But by Theorem 3.7, this set is unbounded, which implies that is
unbounded. This is impossible since is a Dedekind cut. Hence our assumption must be wrong, so there
must exist w and y / such that
y
w
r.
Denition 4.15. A real number is a Dedekind cut. R is the set of all real numbers. By the above theorems
it follows that R is a eld, i.e. it has two operations + and , satisfying the conditions (eld axioms) of
Theorem 3.4. Moreover, R has an ordering < which is preserved by addition (Theorem 4.7 (3)).
Remark 4.16. Note that the way we dened things, R is a set whose elements are sets of rational numbers.
Let this not confuse you: remember that even the integers can be constructed as sets just by starting with the
empty set.
5. Dedekinds main theorem
Theorem 5.1. Let A, B be subsets of R, such that
(1) R = A B,
(2) Neither A nor B is empty, and
(3) Every element of A is strictly less than every element of B.
10
Then, there exists a unique real number , such that every element of A is , and every element of B is
.
Remark 5.2. Conversely, any real number gives such sets A, B, by letting A = { R| < }, and
B = { R | }.
Proof. First, let us show that there can be at most one such . For if there were two, say
1
<
2
, then

1
<
1+2
1+1
<
2
(Why?), which means that
1+2
1+1
is both in A and in B, which is impossible because (3)
implies that A B = .
Now we need to show that a exists. Let =

A
. We will show that is a real number (i.e. a
Dedekind cut) satisfying the requirement.
(1) = : A is not empty, so there is an element A, which is a Dedekind cut by denition, so is
not empty. Hence there exists a rational number y which is an element of . By the denition of
union of sets, it follows that y .
(2) Q\ = : Since B is not empty, there is a cut B. By property (3), > , for every A. This
means that there exists a rational number z such that z / for all A. Therefore, z cannot
be an element of the union

A
= , so z Q\.
(3) is closed under <: Let x be arbitrary, and let y Q be such that y < x. We need to show
that y . Since x , it follows that x for some cut in A (again, by the denition of union
of sets). But then y , which in turn implies that y

A
= .
(4) has no maximal element: Let x ; we need to nd x

> x, x

. Since x , there must be a


cut A such that x . Since is a cut, it has no maximal element, so there is a x

> x, x

.
But then x

.
Therefore, as dened is a Dedekind cut. It remains to show that every element of A is less than or equal
to , and every element of B is greater than or equal to . The rst is clear since is the union of A.
For the second, let B be a cut. Suppose < =

A
. Then there exists a rational number x
which is not in . But x must be an element of some A, so x \, which contradicts our assumption
that < . Therefore .
We end this theoretical fun with one last important fact, that the real numbers constructed this way
satisfy the so-called completeness axiom (Axiom 10, I 3.9 in Apostols book). First we need a denition.
Denition 5.3. A number b is called a least upper bound or a supremum for a set of numbers S if
(1) b is an upper bound for S, i.e. for all x S, x S, and
(2) if a < b, then a is not an upper bound for S.
Remark 5.4. If S has a maximal element, than that element is also a supremum for S. However, S does
not need to contain its supremum: if it does, then it is a maximal element.
Exercise 2. Show that least upper bounds are unique.
Theorem 5.5. Every non-empty set S of real numbers which has an upper bound has a supremum.
Sketch of proof. Complete the details by yourself ! We will use Dedekinds main theorem. Let A be the subset
of real numbers which contains all x R such that x y for some element y S. In other words, A contains
all elements of S as well as those less then some element of A. Let B = R\A. Show that these A and
B satisfy the assumptions for Dedekinds main theorem. Then there exists a unique number such that
a b for all a A and b B. The claim is that this is a supremum for S.
(1) is an upper bound for S directly from Dedekinds main theorem.
(2) Suppose

is another upper bound for A. Show that if

< , then there exists an element x of S


such that

< x < . This is a contradiction, so it must be that

You might also like