I need help with the following problem:

Let $\kappa$ be a weakly inaccessible cardinal.

Show there exists a closed and unbounded set of $\alpha < \kappa$ such that $L_\alpha\vDash ZFC$.

I know that if $\kappa$ is weakly inaccessible then $L_\kappa$ is a model of ZFC. But other than that I am not really sure now to proceed.


1 Answer 1


Hint: We have $L_\kappa \models \mathrm{ZFC}$ and $\kappa$ is regular.

We will show something stronger, namely

$$ C := \{ \alpha < \kappa \mid L_{\alpha} \prec L_{\kappa} \} $$ is a club.

First show that $C$ is unbounded: Let $\beta < \kappa$. As in the proof of the Reflection Theorem recursively construct a strictly increasing sequence $(\alpha_n \mid n < \omega)$ with $\alpha_0 = \beta$ such that $L_{\sup_{n <\omega} \alpha_n} \prec L_\kappa$. Note that $\beta < \sup_{n <\omega} \alpha_n < \kappa$.

Now show that $C$ is closed. This follows pretty much immediately from the Tarski-Vaught test (/ criterion):

Lemma. (Tarski-Vaught)

Let $M \subseteq N$ be $\mathcal{L}$-models for some first order language $\mathcal{L}$. Then the following are equivalent:

  1. $M \prec N$ and
  2. For every $x_0, \ldots, x_n \in M$ and every $\mathcal{L}$-formula $\phi$. If $$ N \models \exists x \colon \phi(x, x_0, \ldots, x_n), $$ then there is some $x \in M$ (!) such that $$ N \models \phi(x, x_0, \ldots, x_n) $$

(You can find a proof of this here.)

Corollary. Let $N$ be a $\mathcal{L}$ structure and $(M_n \mid n < \omega)$ be a strictly increasing sequence such that $M_n \prec M_{n+1} \prec N$ for all $n < \omega$. Then $M := \bigcup_{n < \omega} M_n \prec N$.

Proof. Let $x_0, \ldots, x_n \in M$ and let $\phi$ be a $\mathcal{L}$-formula such that $$ N \models \exists x \colon \phi(x, x_0, \ldots, x_n). $$

Fix $k$ large enough such that $x_0, \ldots, x_n$ in $M_n$. Since $M_n \prec N$ we have that there is some $x \in M_n$ such that $$ N \models \phi(x, x_0, \ldots, x_n) $$ But $x \in M_n \subseteq M$, hence the Tarski-Vaught criterion is satisfied and we have $$ M \prec N. $$ Q.E.D.

Let us show that $C$ is closed: Fix a strictly increasing sequence $(\alpha_\beta \mid \beta < \gamma)$ of $\alpha_\beta \in C$, for some $\gamma < \kappa$. By definition of $C$ we have $$ L_{\alpha_\beta} \prec L_\kappa $$ for all $\beta < \gamma$.

Claim. We also have $$L_{\alpha_\beta} \prec L_{\alpha_{\beta^*}}$$ for $\beta < \beta^* < \gamma$.

Proof. For all $x_0, \ldots, x_n \in L_{\alpha_\beta}$, we have $$L_{\alpha_\beta} \models \phi(x_0, \ldots, x_n) \iff L_{\kappa} \models \phi(x_0, \ldots, x_n) \iff L_{\alpha_{\beta^*}} \models \phi(x_0, \ldots, x_n)$$ Q.E.D.

It now follows from the corollary that $$ L_{\sup_{\beta < \gamma} \alpha_{\beta}} \prec L_{\kappa}, $$ so that $\sup_{\beta < \gamma} \alpha_{\beta} \in C$, so that $C$ is indeed closed.

Since you don't know the reflection theorem (or rather its proof), you can also use the Tarski-Vaught criterion to show that $C$ is unbounded:

Let $\beta_0 < \kappa$. If $L_{\beta_0} \prec L_{\kappa}$, we are done. Otherwise fix a strict wellorder $<^*$ of $L_{\kappa}$ and select, for all formulae $\phi$ and all $x_0, \ldots, x_n \in L_{\beta_0}$ the $<^*$-least $x_{\phi, x_0, \ldots, x_n} \in L_{\kappa}$ such that $$ L_{\kappa} \models \exists x \phi(x,x_0, \ldots, x_n) \implies L_{\kappa} \models \phi(x_{\phi, x_0, \ldots, x_n},x_0, \ldots, x_n). $$

Let $\beta_0 < \beta_1$ be minimal such that $x_{\phi, x_0, \ldots, x_n} \in L_{\beta_1}$ for all $\phi$ and all $x_0, \ldots, x_n \in L_{\beta_0}$.

Now continue doing this (replacing $\beta_0$ with $\beta_1$ and creating $\beta_2$ and so on) to produce an increasing sequence $(\beta_k \mid k < \omega)$. Let $\alpha := \sup_{k < \omega} \beta_k$ and verify that $L_{\alpha} \prec L_{\kappa}$ with Tarski-Vaught (the proof is basically the same as in the Corollary I've proved above).

  • $\begingroup$ Please excuse my incompetence, I am new to set theory/model theory. I don't follow the proof of unboundedness (as I do not know what the reflection theorem is). Could you also please also explain more on how the Tarski-Vaught test shows closure? $\endgroup$
    – A. Collins
    Mar 8, 2019 at 0:25
  • $\begingroup$ @A.Collins Sure, I'll expand my answer. $\endgroup$ Mar 8, 2019 at 0:31
  • $\begingroup$ @A.Collins Does the recent edit answer your questions? $\endgroup$ Mar 8, 2019 at 0:50
  • $\begingroup$ Thank you for your expanded answer. I am slowing parsing my way thought it. I am not quite done yet, but I think it should. $\endgroup$
    – A. Collins
    Mar 8, 2019 at 1:10
  • $\begingroup$ @A.Collins If you have further questions, just leave a comment. I may not answer quickly (because I'm currently traveling) but I (or maybe someone else) will address it eventually. $\endgroup$ Mar 8, 2019 at 1:17

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.