It is sometimes useful to pursue constructive proofs in set theoretic combinarotics, that is, to avoid the
Axiom of Choice
. Here are some examples where the constructive proofs are particularly simple.
Set families in
The following are three famous problems concerning the power set of the natural numbers. They focus on the countability of and the uncountability of . Relevant remarks are provided in the end of this section.
☆☆☆
A chain is a set of subsets of such that , . Find an uncountable chain .
Solution
Note that it suffices to find any chain with , and then apply the induced bijection on . For our purpose we pick .
For each let
To see that is a chain, it suffices to show for all that , which is easy by the density of .
★☆☆
An almost disjoint family is a set such that , . Find an uncountable almost disjoint family .
Solution
Similar to problem 1, we instead find an almost disjoint family .
For all define
It's not hard to see that for .
To see that is an uncountable almost disjoint family, we show for all that .
Since and since is an increasing limit point of , it's not hard to believe that has only finitely many elements in . Indeed, we have
★☆☆
An independent family is a set such that , . Find an uncountable independent family .
Solution
Similar to problems 1 and 2, we start by constructing a countable set
For each define
Then to see that is an independent family, let be arbitrary. We can pick some that separates all the reals:
Then it's easy to see that . In fact there are infinitely many ways to choose , which shows that
Here are some remarks.
- An interesting alternative proof of problem 2 involves diagonalizing any countably infinite almost disjoint family to add a new member to it, which by transfinite induction gives an almost disjoint family of size . In fact the cardinality of a smallest maximal almost disjoint family is a
cardinal characteristic
, so it's consistent that .
- All three problems above can be generalized to higher cardinals. In problem 3, we used the fact that is a countable dense subset of an uncountable
Hausdorff space
. For an infinite cardinal, one can take to be a Hausdorff space by endowing with the
product topology
. The subset of points of finite
support
is a size- dense subset of , and is a size- topological base of . This then implies the number of ultrafilters on is . See Thm 7.6 of [1].
- Contrary to remark (2), the
lexicographic order
on does not generate as its
order topology
, and so that generalization does not apply to problems 1 and 2. To construct a large linear order with a small dense subset, let be the least such that . Let and let be the set of all of bounded support. Then is a dense subset of of size . The cardinality of thus depends on cardinal arithmetic.
Define the almost equality
equivalence relation
on via ,
For example,
Write for the quotient space of equivalence classes of .
Define the Vitali equivalence relation on via ,
For example, .
Write for the quotient space of equivalence classes of .
From cardinal arithmetic one sees that . The following construction explicitly shows this equality.
For sets and equivalence relations on respectively, we say that iff there exists a function such that ,
This induces an
injection
, and so .
★★☆
Construct explicit functions to show that .
Solution 1/2 ()
For each define
Clearly if then .
It remains to show that if then . Note that
where for infinitely many .
Thus follows from one of the
proofs that is irrational
.
Solution 2/2 ()
This proof is by Itay Neeman.
For every , there exists a unique terminating
factorial base representation
where for all . See also [2].
It's not hard to see that the representation terminates (that is ) iff . Moreover one can show that iff the representations of differ finitely (that is where ).
Thus one can define the function via
In return, , iff . By a simple bijection as in Problems 1,2,3 above, we get as desired.
Here are some remarks.
(1) In general, whenever one can constructively show that , they can then apply the standard
Cantor–Schröder–Bernstein
argument to find a constructive bijection . It follows that constructively.
(2) The function constructed in Solution 1/2 has the property that whenever , the set is in fact -
linearly independent
. One can find an uncountable family of such by using either of Problems 1,2,3 above. Thus constructively we've shown that . Contrast this with the fact that there exists no constructive
-basis of
.
(3) With respect to the
product-discrete topology
on , both are
Borel equivalence relations
, and the functions constructed are Borel reductions between them. In standard notation, can be instead written as .
(4) There exists no constructive bijection . Assuming
, one can in fact show that holds in . See Ch. 4 Sec. 4 of [3].
Miscellaneous Puzzles
I have no idea how to come up with solutions to the following puzzles. However, other people have:
★★★
A 3-term arithmetic progression in is a set of the form for some . Construct a maximal subset without a 3-term arithmetic progression.
That is, has no 3-term arithmetic progressions, but , has a 3-term arithmetic progression.
Link to Solution
★★★
A circle in is any circle of radius . Construct a partition of into circles.
Link to Solution