Notes on the family S2
Proposition:
is consistent.
Suppose
and
are intervals in
.
We must prove
and
.
Suppose
.
The we have
,
so
,
contradicting
.
We can show in a similar fashion that
is impossible.
Proposition:
is a fine family.
The interval
lies in
since
.
We construct other intervals
in the following way. Suppose
has been constructed (e.g.
).
Let
,
and let
midpoint
of
.
If
then
;
if
,
the
.
It is each to check that
,
defined in this way, belongs to
,
and that
.
Given
,
we can choose
so that
;
for such a
,
.
(For example, if
,
then
.)
Thus,
contains arbitrarily fine intervals.
Proposition: If
belongs to every interval of
,
then
.
We will show that
for every
.
Let
be any interval in
.
Since
belongs to
,
;
since
belongs to
,
.
Thus, both
and
belong to
.
Now
.
Since
belongs to every interval in
,
we can assume that
is one of the intervals constructed by bisection in the previous Proposition.
If this is the case,
and
,
so
,
and we have
.
Since the intervals constructed by bisection are fine, choose
so that
;
then
.
Thus,
and
lie in an interval of length
,
so
for any
.
By the "wiggle" lemma,
,
so
,
and so
.
Proposition: No rational number can have
as its square.
Suppose
has
as its square. By repeated cancellation of
s
from numerator and denominator, we can replace
and
by whole numbers
and
which are not both even; Thus, we have
.
Therefore
and so
is even. Now
can't be odd or its square would be odd (odd times odd = odd), so
must be even;
.
Substituting this into the previous equation gives:
,
so
.
Therefore
is even, so
must be even also. Thus, contrary to assumption,
and
are both even. This contradiction shows that the assumption that a rational
number has square
is untenable.
Proposition: As a real number,
.
We must show that, as families,
.
This amounts to showing that every interval
in
meets
;
i.e. contains
.
A typical interval in
looks like
where
and
belong to
.
We
have
Thus,
and we are done:
.
From the above it follows that
is a real number whose square is
,
so we are justified in actually calling it
.
Furthermore,
is a consistent family of rational intervals, but there is no
single rational number which lies in all of its intervals. (If there
were, it would be a rational with
as its square, by part 3 above. But this is impossible, by part 4.)