5
$\begingroup$

In $V$, let me call a set theory structure A is a $\omega_1$ model if the $\omega_1$ of $A$ is the same as the $\omega_1$ in $V$ (up to isomorphism). The question I would like to ask is the following: Given any transitive structure $(B,\in)$ containing $\omega_1$ in $V$, is it always possible to find an $\omega_1$ model $A$ such that $B\prec A$ and $B\cap H(\omega_2)$ is in the well-founded part of $A$.

Some remark: Lowenheim-Skolem or Compactness theorem can be employed to construct an extension. But in general this extension is not an $\omega_1$ model. To get one $\omega_1$ model, one can either use some large cardinal or force a generic ultrafilter to construct an ultrapower which is well founded below $\omega_1$. But I am not sure how to get this using ZFC alone. On the other hand, it is known that the existence of $\omega$-model is correct between transitive ZFC models and thus can be constructed under ZFC.

I know very little about model theory of set theory. So it would be very helpful if anyone can provide some reference about this topic.

$\endgroup$
1
  • $\begingroup$ Sorry, I do mean $H(\omega_2)^B$ is in the well-founded part of $H(\omega_2)^A$. It would be better $B$ itself is an element in the well-founded part of $A$. In this case, the $\omega_2$ of $A$ and $B$ are not correct. $\endgroup$
    – user47286
    Feb 22, 2014 at 15:24

1 Answer 1

4
$\begingroup$

In this generality, this is impossible, if you want $B\prec A$ to be a proper extension.

For a counterexample, let $B=V_{\omega_2}$, the rank initial segment of the universe of height $\omega_2$. I claim that there can be no proper extension $\langle B,\in\rangle\prec\langle A,\hat\in\rangle$, where the well-founded part of $A$ includes $H_{\omega_2}$. Suppose that we have such an $A$, and let $\beta$ be any of the new ordinals in $A$, that are not in $B$. (There must be such a new ordinal, since $B$ thinks every set has an ordinal rank, and for the ranks up to $\omega_2$, the original structure $B$ already had all the sets of that rank.) So $\beta$ is larger than every ordinal in $B$. But since $B$ thinks every ordinal has size $\omega_1$, it follows that $A$ thinks it has a bijection of $\beta$ with $\omega_1$. But it is correct about $\omega_1$, and so we can form an actual bijection of the objects that $A$ thinks are below $\beta$ with the actual $\omega_1$. This is impossible, since the former set includes all of $\omega_2$. Contradiction.

In the comments, you inquire whether one can find a counterexample $B$ of size $\omega_1$, and the answer is yes. Let $B=L_{\omega_1+1}$, which is a transitive set that contains $\omega_1$ as an element (I mean to use true $\omega_1$ here, which might be larger than $\omega_1^L$), and indeed $\omega_1$ is the largest ordinal of $B$. Suppose that $\langle B,\in\rangle\prec \langle A,\hat\in\rangle$ is a nontrivial extension, where $A$ is an $\omega_1$-model and $B\cap H_{\omega_2}=B$ is in the well-founded part of $A$. It follows that $A$ is well-founded up to and including $\omega_1$, and by elementarity, since $\omega_1$ is the largest ordinal of $B$, it must also be the largest ordinal of $A$. So the ordinals of $A$ are precisely $\omega_1+1$. Since also $A\models V=L$, it follows by elementarity that $A=L_{\omega_1+1}$, which is to say, that $A=B$, and so this isn't a nontrivial extension after all. Contradiction. Perhaps this example shows that you haven't quite asked the question you intended to ask.

$\endgroup$
4
  • $\begingroup$ Thanks. It's great. And does your answer work for $\omega_1$-size $B$? $\endgroup$
    – user47286
    Feb 22, 2014 at 15:36
  • $\begingroup$ This argument does not seem to work with $B$ of size $\omega_1$, since it used the fact that all ordinals up to $\omega_2$ were in $B$. $\endgroup$ Feb 22, 2014 at 17:18
  • $\begingroup$ But meanwhile, I've found a different counterexample of size $\omega_1$. $\endgroup$ Feb 23, 2014 at 0:31
  • $\begingroup$ I see. There is no such $A$ unless $B$ is closed enough. $\endgroup$
    – user47286
    Feb 23, 2014 at 3:15

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.