6. clopen vs. finitely-determined
problem
- Prove that every clopen subset of is finitely determined.
- Give an example of a clopen set in that is not finitely determined.
-
Suppose is a clopen subset of but is not fd.
finitely-determined means that there exists an so that membership does not depend on coordinates , i.e., for all ,
therefore to say is not fd means that for all , there exist sharing common prefix up to coordinate , but and .
in this way define two sequences of strings, and .
also construct a string inductively as follows. for each , pick so that occurs as a prefix of infinitely many strings in the sequence . each step this is possible by the fact that infinitely-many candidates remain (induction hypothesis), and the pigeonhole principle (among the two choices for the next digit, at least one of them occurs infinitely many times).
at the same time, observe that each and share the same prefix up to index , so if is a prefix of then it also is a prefix of .
by construction, each (finite) prefix of occurs infinitely many times in , so there exists a subsequence of converging to . as in, is a limit point of . by the same reasoning is a limit point of .
the former sequence is in and the latter in . by the hypothesis that is clopen, both and are closed, so , a contradiction.
-
For each let denote the set of strings that begin with (repeated times), and let .
-
is open: observe that is open, and a union of is also open.
-
is closed: we will show that is open. Fix , and let . Consider two cases:
- suppose . by construction doesnβt contain any strings starting with or , so .
- suppose . since we know that there exists such that . furthermore any string satisfying and is not in . so . in both cases we showed that is contained in some open set . thus is open, i.e., is closed.
-
is not fd: fix any . define (with leading copies of ) and observe . next define by setting only the digit at index to zero, and agreeing with everywhere else. observe .
then share a prefix of length but and . since this holds for all we conclude is not f.d.
-