Wednesday, 5 August 2009

peano arithmetic - "Interesting" properties of sets of natural numbers

I'm glad you decided to follow up on my suggestion to look at Descriptive Set Theory! As I mentioned in my earlier answer, there are two flavors of Descriptive Set Theory: the "boldface theory" and the "lightface theory." As you observed, the boldface theory of sets of natural numbers is completely trivial since every subset of mathbbN is clopen, but the lightface theory is not trivial at all.



The lightface theory is completely analogous to the boldface theory except that all operations are required to be computable, or effective in the descriptive lingo. Let's go through the first few levels of the boldface and lightface Borel hierarchies in parallel. Let X be a Polish space and let (Un)in=0nfty be a nice enumeration of a basis for the topology on X.



  • A boldface boldsymbolSigma01 set is simply an open subset of X, or a countable union of basic open sets. Similarly, a lightface Sigma01 set is an effective union of basic open sets, i.e. a union of the form bigcupin=0nftyUf(n) where f is a computable function. Well, not quite, since this informal definition technically excludes the empty set. A better definition is that the lightface Sigma01 sets are the sets Ge=bigcupninWeUn where We is the e-th computably enumerable set. This also gives a better listing of the lightface Sigma01 sets.


  • A boldface boldsymbolPi01 set is simply a closed subset of X, or the complement of a boldface boldsymbolSigma01 set. Similarly, a lightface Pi01 set is a the complement of a lightface Sigma01 set. Note that we have a nice enumeration of all lightface Pi01 sets, namely Fe=XsetminusGe where Ge was defined above.


  • A boldface boldsymbolSigma02 set is a Fsigma subset of X, or a countable union of boldface boldsymbolPi01 sets. Similarly, a lightface Sigma02 set is an effective union of Pi01 sets. As above, the lightface Sigma02 sets are enumerated by Ge=bigcupninWeFn where We is the e-th computably enumerable set and Fe was defined above.


  • A boldface boldsymbolPi02 set is a Gdelta subset of X, or the complement of a boldface boldsymbolSigma02 set. Similarly, a lightface Pi02 set is a the complement of a lightface Sigma02 set. Note that we again have a nice enumeration of all lightface Pi02 sets, namely Fe=XsetminusGe where Ge was defined above.


And so on, lightface Sigma0n+1 sets are effective countable unions of lightface Pi0n sets, which in turn are complements of lightface Sigma0n sets and at each step we have a nice enumeration of all of these sets.



Just as for the boldface Borel hierarchy, there is no reason to stop at omega; this process can be continued well into the transfinite, but there are a few roadblocks along the way. Because we want the hierarchy to be effective, we can only go up to the Church-Kleene ordinal omegaCK1, which is the smallest non-computable ordinal. (Also, the hierarchy is defined using ordinal notations instead of actual ordinals.) In the end, we get the family lightface Borel sets, which is a countable subfamily of the usual Borel sets.



The analytic sets, or boldface boldsymbolSigma11 sets, also have a lightface analogue. The boldface boldsymbolSigma11 sets are precisely the projections onto the X-axis of the boldface boldsymbolPi01 subsets of the Polish space XtimesmathbbNmathbbN, where mathbbNmathbbN denotes Baire space. Similarly, the lightface Sigma11 subsets of X are the projections onto the X-axis of the lightface Pi01 subsets of XtimesmathbbNmathbbN. Note that since we have a nice enumeration of the lightface Pi01 sets, we also have a nice enumeration of the lightface Sigma11 sets. There are similar lightface analogues of the entire projective hierarchy. A lot of the results of the boldface theory transfer to the lightface theory. For example, the Suslin Theorem that the boldface boldsymbolDelta11 sets are precisely the Borel sets. A theorem of Kleene says that the same holds for the lightface Delta11 sets, namely these are precisely the lightface Borel sets described above.



In the case of the discrete (and Polish) space mathbbN, the natural choice of basis are the singletons Un=n. Working through the definitions, we see that the lightface Sigma01 subsets of mathbbN are precisely the computably enumerable subsets of mathbbN and the lightface Pi01 sets are the complements thereof. Hence, the Delta01 subsets of mathbbN are precisely the computable sets. In fact, the lightface Sigma0n, Pi0n, Delta0n, perfectly match the arithmetic hierarchy. The lightface Borel subsets of mathbbN are commonly called hyperarithmetic sets.



Obviously, there are lots and lots of interesting properties of subsets of sets of natural numbers!

No comments:

Post a Comment