Next: 5.3 Existence of Roots
Up: 5. Real Numbers
Previous: 5.1 Sequences and Search
  Index
5.21
Definition (Completeness axiom.)
Let
be an ordered field.
We say that
is
complete if it satisfies
the condition:
Every binary
search sequence in converges to a unique point in .
5.22
Example.
The field
is not complete, since in example
5.16 we found a
binary search sequence in
that does not converge.
5.23
Definition (Real field,
.)
A
real field is a complete
ordered field. We will use the name
to denote
a real field.
5.24
Remark.
It is not at all clear that any real fields exist. If real fields do exist, there is
a question of uniqueness;
i.e., is it the case that any two real fields are ``essentially the same"? I don't want to worry about what mathematical existence
means, so let me formulate the questions: Are the axioms for a real field
consistent; i.e., is it the case that no contradictions can be derived from them?
Note that we are not entirely free to throw axioms together. If I were to make a
definition that a
3-field is an ordered field in which
, I would
immediately get a contradiction:
and
. All I can say about consistency is
that no contradictions have been found to follow from the real field axioms. There
exist proofs that any two real fields are essentially the same, cf. [
35, page
129]. (This source uses a different statement for the completeness axiom
than we have used, but the axiom system is equivalent to ours.) There also exist
constructions of pairs of very different real fields, cf. [
41].
In what follows, I am going to assume that there is a real field
(which I'll
call the real numbers). Any theorems proved will be valid in all real fields.
5.25
Theorem (Archimedean property 1.)
Let
be a real field, and let
. Then there is an integer
such that .
Proof: Let
, and suppose (in order to get a contradiction) that there is no
with . Then for all . Now
is a binary search sequence in
. Since , I have
for all
. We see that
, but clearly
. Since completeness of
implies that a binary search sequence has a unique limit, this yields a
contradiction, and proves the theorem.
5.26
Corollary (Archimedean property 2.)
Let
. Then there is some
such that
.
Proof: By the theorem, there is some
with
. Then
.
Proof: Let
, and let
satisfy
|
(5.29) |
Suppose, in order to get a contradiction, that . Then
by Archimedian property 2 there is some
such that
, i.e.
.
This contradicts (5.29).
5.30
Theorem.
If
, then there is
an integer and a number
such that
.
In order to prove this theorem, I will use the following lemma.
5.31
Lemma.
If
, then the interval contains an integer.
Proof:
- Case 1.
-
: Suppose
and does not
contain an integer. I will show that for all
. This contradicts the
Archimedean property, so no such can exist. For each
, let
". Then is true, since I assumed that
. Let
be a number such that is true; i.e., . If were ,
we'd have , and this cannot happen, since contains no
integers. Hence,
and by induction, for all
. This gives the desired contradiction.
- Case 2.
-
: If
, then by Case 1 there is an integer with
Then
If , then contains . If , then
contains .
Proof of theorem 5.30. Let
. By the lemma, there is an integer
with . Then
and
gives the desired decomposition.
5.32
Theorem.
There is a number
such that .
Proof: Let be the binary search sequence constructed in example
5.16. We know there is a unique
such that
for all . Then
, and by our construction
for all
, so
|
(5.33) |
for all
.
By Archimedean property 3, we conclude that
, i.e., .
5.34
Theorem.
Let
. Then there is a binary search sequence in
such
that
and
for all , and such that
.
Proof: I will suppose . The case where is left to you. By the
Archimedean property of
, there is an integer such that , so
. Now build a binary search sequence as follows:
From the construction, we have
and
.
A simple induction argument shows that
and
for all
,
and an induction proof similar to the one in
example 5.16 shows that
for all
so
.
Next: 5.3 Existence of Roots
Up: 5. Real Numbers
Previous: 5.1 Sequences and Search
  Index