2
$\begingroup$

What is the cofinality of this ordinal (say in ZFC)? Is it countable or not?

Edit: Based on reason for close, I have added some informal motivation for question. Based on the computability-theoretic definition, whenever $\omega^{CK}_\alpha$ is countable then so is $\omega^{CK}_{\alpha+1}$. But, formally speaking, it doesn't apply directly when $\alpha$ isn't countable.

Denoting $\beta=\omega^{CK}_{\omega_1+1}$, intuitively there are several "interesting" and large enough points $p$ (such that $p<\beta$) so that $p$ will have countable finality (as one example when $p$ is the first fixed point of a normal function such as $x \mapsto (\omega_1)^x$).

But this informal intuition doesn't help in giving a formal answer to cofinality of $\omega^{CK}_{\omega_1+1}$. Hence the reason for asking the question (I want to know how well the informal intuition does or doesn't transfer into formal reasoning).

$\endgroup$
  • 1
    $\begingroup$ What is $\omega_{\omega_1 + 1}^{CK}$? I can make sense of $\omega_\alpha^{CK}$ for countable $\alpha$ but for uncountable ones, I'm not sure what the intended meaning is. If $\omega_\alpha^{CK}$ is the $\alpha$-th admissible ordinal, then $\omega^{CK}_{\omega_1 + 1}$ has countable cofinality. $\endgroup$ – Stefan Mesken Mar 4 at 9:20
  • $\begingroup$ @StefanMesken I would be out of depth in trying to explain $\omega_\alpha^{CK}$ (for uncountable $\alpha$) myself. But you might look at the answer of a question I asked number of months ago: mathoverflow.net/questions/310244. So given that answer, yes it seems that "$\omega_\alpha^{CK}$ as the $\alpha$-th admissible ordinal" should be the intended meaning. $\endgroup$ – SSequence Mar 4 at 11:06
  • $\begingroup$ Given your comment (as I took it), $\omega^{CK}_{\omega_1 + 1}$ has (provably) countable cofinality. In that case, you might convert your comment as an answer. $\endgroup$ – SSequence Mar 4 at 11:09
  • $\begingroup$ The accepted answer is incorrect - see my answer and Stefan's comment. $\endgroup$ – Noah Schweber Mar 4 at 14:23
  • $\begingroup$ @NoahSchweber Thanks for the effort on the answer. I was definitely not expecting $cf(\theta)>\omega$. I certainly can't make heads or tails out of this. I will try to explain the brief reasoning (in a paragraph or two) why I am feeling very confused by this by bumping the relevant MO question (which was the reason for this question) a bit later. $\endgroup$ – SSequence Mar 4 at 15:10
2
$\begingroup$

I disagree with Stefan Mesken's answer. First, let me explain why I think the general claim made there is false. Consider the theory $T$ = KP + V=L + "There is some ordinal $\alpha$ such that $\alpha+\alpha$ doesn't exist but $\alpha+\beta$ exists for all $\beta<\alpha$." We have $L_{\omega_1+\omega_1}\models T$, but no $\alpha\in (\omega_1,\omega_1+\omega_1)$ models $T$; so $L_{\omega_1+\omega_1}$ is the first level of $L$ satisfying $T$ above $\omega_1$, yet $cf(\omega_1+\omega_1)\not=\omega$.


Now here's an argument showing that the cofinality of $\omega_{\omega_1+1}^{CK}$ - which latter ordinal I'll call "$\theta$" for simplicity - is $\omega_1$.

The key is projecta. If $\alpha$ is the next admissible above $\beta$, then there is an injection $\pi$ of $\alpha$ into $\beta$ which is $\Sigma_1$-definable (in $L_\alpha$). This doesn't contradict admissibility of $\alpha$ since in order to turn this into a $\Sigma_1$ cofinal map from $\beta$ to $\alpha$ we would need $ran(\pi)$ to be $\Delta_1$, and it's only $\Sigma_1$.

Indeed, $ran(\pi)$ has the following strong property: for any cofinal set $A\subseteq \alpha$, $\pi[A]$ is not $\Delta_1$. Otherwise we could indeed get a $\Sigma_1$ map from $\beta$ to $\alpha$ with cofinal image - namely, send $x\in A$ to $\pi^{-1}(x)$ and $x\not\in A$ to $17$. This is in fact $\Sigma_1$, since if we know ahead of time that $x\in ran(\pi)$ we can search for $x$'s preimage without risk of looping forever.

OK, now suppose $cf(\theta)=\omega$. Fix$^1$ a constructible cofinal $\omega$-sequence $(\eta_i)_{\eta<\omega}$, and let $\gamma_i=\pi(\eta_i)$. Then $S=(\gamma_i)_{i\in\omega}$ is an $\omega$-sequence of countable ordinals, so by condensation$^1$ we get $S\in L_{\omega_1}$. But now take $S$ - or rather, the range of $S$ - to be our $A$.

(OK fine, strictly speaking all I've shown is that $cf(\theta)>\omega$. But since $\vert\theta\vert=\omega_1$, this does mean $cf(\theta)=\omega_1$.)


$^1$Why does this exist? On the face of things, after all, I seem to be using V=L (or at least that $L$ correctly computes cofinalities) here. This isn't necessary, but at present I'm forgetting how the argument goes; I'll add it when I have time to recall it.

$\endgroup$
  • $\begingroup$ I agree that my argument doesn't work -- I incorrectly assumed that closing a given level under Skolem terms would land me strictly below the model of $T$ of uncountable cofinality in the hierarchy. $\endgroup$ – Stefan Mesken Mar 4 at 14:21
  • $\begingroup$ Btw, you should mention that you work in $L$, i.e. $\omega_1 = \omega_1^L$ in your post. $\endgroup$ – Stefan Mesken Mar 4 at 14:27
  • $\begingroup$ @StefanMesken Whoops, good point - addressed. $\endgroup$ – Noah Schweber Mar 4 at 14:48

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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