5
$\begingroup$

I posted this question on Math StackExchange but did not get a full answer. I hope it's not a problem if I ask again here.

Is there a way to compute explicitly the $L$−rank $\rho(\bigcup x)$ of $\bigcup x$ in terms of $\rho(x)$? I know it's necessarily $\rho(\bigcup x)\leq\rho(x)$ and that both $\rho(\bigcup x)<\rho(x)$ and $\rho(\bigcup x)=\rho(x)$ can hold, but what else can be said, maybe distinguishing between $\rho(x)$ limit ordinal and successor ordinal? For example, is it possible that equality holds if $\rho(x)$ is a successor ordinal? If so, could you give an example?

$\endgroup$
1

1 Answer 1

9
$\begingroup$

There is no way to compute explicitly $\rho(\bigcup x)$ in terms of $\rho(x)$, in any meaningful fashion: e.g. (working in $L$) for arbitrarily large countable $\alpha$ there are reals $x\subseteq \omega$ with $\rho(x)$ a limit or successor greater than $\alpha$ but with $\rho(\bigcup x) =\omega$ of course. (Because for arbitrarily large such $\alpha$ there is a new real $x$ definable over $L_\alpha$. If $\alpha$ is not a successor, then there will always be a new real definable over $L_{\alpha +1 } $ too; for example the real that codes $L_\alpha$; similarly if $\alpha$ is not a limit there will in any case be a new real definable over $L_{\alpha + \omega}$.) Thus the gap between the two ranks under either assumption can be as wide as it could conceivably be.

If $\rho(x)=\gamma+1$ then $\rho(\bigcup x)= \rho(x)$ can happen. Work in $L$: Suppose $x\subseteq \omega$ has $\rho(x)=\gamma +1$. Let $y = \{ \{u\} \mid u \in x \}$. Then $\rho(y)\leq\gamma +1 $; $\rho(\bigcup y) =\rho(x)$, but $\rho(y)\leq\gamma$ would imply that $\rho(x)\leq\gamma$.

$\endgroup$
7
  • $\begingroup$ Actually this is not totally correct. It's true that new reals are constructed arbitrarily high, but surprisingly, there are also arbitrarily long gaps where no new reals appear. Let $\alpha < \omega_1$ be arbitrary, and let $\omega_1 > \beta > \alpha$ be such that there is an elementary $j : L_\beta \to L_{\omega_2}$ with $cr(j) = \gamma \geq \alpha$. Then for all $\delta < \gamma$, $L_{\omega_2} \models$ "There are no new reals constructed between $\omega_1$ and $\omega_1 + \delta$." By elementarity and absoluteness, there are no new reals added between $\gamma$ and $\gamma + \delta$. $\endgroup$ Nov 15, 2013 at 18:29
  • $\begingroup$ Thank you for your answer. Could you give me an explanation for your last statement? I can't see why equality can't hold if $\rho(x)=\gamma+1$. $\endgroup$
    – phil
    Nov 15, 2013 at 18:33
  • $\begingroup$ @Monroe: which part of my answer is incorrect? $\endgroup$ Nov 15, 2013 at 19:10
  • $\begingroup$ It seems like you said if $\alpha$ is a countable limit, then a new real appears in $L_{\alpha+1}$. $\endgroup$ Nov 16, 2013 at 0:58
  • 2
    $\begingroup$ @Monroe: I was discussing in the parenthetical remark ``such $\alpha$'' with a new real appearing definably over $L_\alpha$. Then one will also appear over $L_{\alpha +1}$, ... , $L_{\alpha+\omega}$... :) $\endgroup$ Nov 16, 2013 at 9:13

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.