5
$\begingroup$

This is a follow-up to my question on $\Delta^{1}_{2}$ and degrees of constructibility of real numbers that was answered by the user "William", see here: Can $\Delta^{1}_{2}$ separate degrees of constructibility?

Let us say that the real number $r$ encodes the heriditarily countable set $x$ if and only if there is a bijection $f:\omega\rightarrow\text{tc}(x)\cup\{x\}$ such that $f(0)=x$ and $r=\{p(i,j):f(i)\in f(j)\}$, where $p$ is Cantor's pairing function and $\text{tc}(x)$ denotes the transitive closure of $x$.

Let us say that a $\Delta^{1}_{2}$-formula $\phi$ is "coding invariant" if and only if, for any two codes of the same set $x$, $\phi$ either holds of both or of neither of them; thus, $\phi$ expresses a classification of the heriditarily countable sets. If $x$ is heriditarily countable, we will say that $\phi$ holds of $x$ and write $\phi(x)$ if and only if $\phi$ holds of every real code of $x$.

My question now is whether the following holds: When $\phi$ is a coding invariant $\Delta^{1}_{2}$-formula, $A$ is the set of heriditarily countable sets of which $\phi$ holds and $\bar{A}$ is the set of heriditarily countable sets of which $\phi$ does not hold, does one of $A$ and $\bar{A}$ contain elements of all degrees of constructibility of heriditarily countable sets?

In other words, can $\Delta^{1}_{2}$ separate degrees of constructibility "$\textbf{on the set level}$"?

Note that the answer to Can $\Delta^{1}_{2}$ separate degrees of constructibility? does not immediately yield an answer here, for at least two reasons: (1) not every real number codes a set and (2) codes for the same set can come from very different degrees of constructibility. (3) [added after Douglas Ulrich's comment to Liang Yu's answer]: Not every heriditarily countable set is $L$-equivalent to a real number.

[Here was a wrong example of a heriditarily countable set not $L$-equivalent to a real number, which I deleted after Liang Yu's comment below.]

$\endgroup$
3
  • 1
    $\begingroup$ The degree of $\omega_1^L$ is $0$. $\endgroup$
    – 喻 良
    Commented Nov 1, 2018 at 11:53
  • 1
    $\begingroup$ I think that there should be a negative solution to the question. But the proof is too lengthy to be presented here. $\endgroup$
    – 喻 良
    Commented Nov 13, 2018 at 1:07
  • $\begingroup$ Can you give a sketch? $\endgroup$
    – M Carl
    Commented Nov 14, 2018 at 9:06

1 Answer 1

3
$\begingroup$

Here is a partial positive answer: i.e. either $A$ or $\bar{A}$ contain elements of all degrees of constructibility of reals.

Given an infinite set of numbers $x$, let $f:\omega\to \omega\cup \{x\} $ so that $f(n+1)=n$ and $f(0)=x$. Then the corresponded $r_x\equiv_T x$. Now if $A$ is a coding invariant $\Delta^1_2$-set, then either $A$ or the complement of $A$ contains a nonconstructible $r_x$. W.l.o.g, we assume that $A$ contains a nonconstructible $r_x$. Let $B=\{s\in A\mid s \mbox{ is a coding like }r_x\}=\{s \in A\mid \{p(i,j)\mid 1\leq i\leq j\}\subseteq s \subseteq \{p(i,j)\mid 1\leq i\leq j\}\cup\{p(i,0)\mid i\in \omega\}\}$

Then $B$ is a $\Delta^1_2$ set and $r_x\in B$. So $B$ has a perfect subset $T\in L$. But for each real $s\in [T]$, there is a real $y$ such that $s=r_y\equiv_T y$. So the coded heriditarily countable sets range over all the $L$-degrees.

The partial answer is not quite far away the full one. Since for any heriditarily countable set x, there are two sets of numbers $y$ and $z$ so that $L[z]\cap L[y]=L[\{x\}\cup \mathrm{tc}(x)]$.


Here is a partial negative answer to your question: I.e. there is a $\Pi^1_1$-formula as you required for which $A$ contain elements of all degrees of constructibility of reals but not all degrees of constructibility of heriditarily countable sets.

Given the function as above. Note that for any $x$ set of numbers and $s$ coding $x$, $r_x\leq_h s$. Now let $B=\{s\mid \exists r\leq_h s(r\cong s\wedge (r \mbox{ is }r_x \mbox{ for some }x))\}$ be a set defined by a $\Pi^1_1$-formual $\phi$. Then $\phi$ satisfies your requirements. Let $A$ be the corresponded collection of heriditarily countable sets of which $\phi$ holds. Then $A$ exactly contains all the constructible degrees of reals. If $\omega_1^L$ is countable, the set $\{(\alpha,g_{\alpha})\mid \alpha<\omega_1^L\}$, the sequence produced by a finite supported Cohen forcing of length $\omega_1^L$ over $L$, does not belong to $A$. Actually it has no a constructible degree of reals.

$\endgroup$
2
  • 2
    $\begingroup$ I believe you are misinterpreting the question. We are looking at the constructible degrees of the hereditarily countable sets; note for instance that not every hereditarily countable set need have the same degree as a real. $\endgroup$ Commented Oct 30, 2018 at 19:39
  • $\begingroup$ Ah, you are right. $\endgroup$
    – 喻 良
    Commented Oct 31, 2018 at 5:19

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .