I am new to set theory and have been working through the proof that every measurable cardinal is Mahlo on page 135 of Jech's text. With the help of Asaf's comments (Measurable $\rightarrow$ Mahlo), I have been able to make sense of the first half of the proof.

However, I found the second half (that argues by contradiction that $\{\alpha < \kappa : \alpha \text{ is regular} \} \in D$) quite terse, and cannot quite follow what is going on.

Could someone please provide a detailed version of Jech's proof or perhaps a detailed alternative proof (that mimics the proof that every measurable cardinal is inaccessible, which I believe I understand better).

Thank you in advance for your help.

  • $\begingroup$ I don't have Jech available to me right now. But the simplest proof, in my opinion, is via ultrapowers. Are you familiar with them? $\endgroup$ – Stefan Mesken Mar 22 at 16:39

Here is a proof leveraging ultrapowers. It generalizes well to all sorts of situations which is why I recommend learning it at some point:

In the following let $U$ be a normal ultrafilter on a measurable cardinal $\kappa$ and let $$ \pi \colon V \to \mathrm{Ult}(V;U) $$ be the canonical ultrapower embedding (we regard $\mathrm{Ult}(V;U)$ as transitive).

Claim. Let $X \subseteq \kappa$. Then $\pi(X) \cap \kappa = X$.

Proof. For $\xi < \kappa$ we have $$ \xi \in X \iff \pi(\xi) = \xi \in \pi(X). $$ Q.E.D.

Claim. Let $C \subseteq \kappa$ be a club. Then $\kappa \in \pi(C)$.

Proof. By elementarity $$ \mathrm{Ult}(V;U) \models \pi(C) \text{ is a club in } \pi(\kappa) $$ and $\pi(C) \cap \kappa = C$ is unbounded below $\kappa < \pi(\kappa)$.

Thus $$ \mathrm{Ult}(V;U) \models \kappa \in \pi(C) $$ and (by $\Sigma_0$-absoluteness) hence $\kappa \in \pi(C)$. Q.E.D.

Claim. $\mathrm{Ult}(V;U) \models \kappa \text{ is regular}$.

Proof. $\kappa$ is regular in $V$, $\mathrm{Ult}(V;U) \subseteq V$ and regularity is downward-absolute (a short cofinal sequence in $\mathrm{Ult}(V;U)$ would also witness in $V$ that $\kappa$ is singular). Q.E.D.

Now combine all of this:

Let $C \subseteq \kappa$ be a club. Then $$ \mathrm{Ult}(V;U) \models \kappa \in \pi(C) \text{ and } \kappa \text{ is regular }. $$ In particular $$ \mathrm{Ult}(V;U) \models \pi(C) \text{ contains a regular cardinal}. $$ By the elementarity of $\pi$ we obtain that $$ V \models C \text{ contains a regular cardinal}. $$ Since $C$ was an arbitrary club in $\kappa$, it follows that $\kappa$ is Mahlo.

  • $\begingroup$ Thank you so much for this. I am slowing working my way though this proof. Could you explain what you mean by elementarity (of 𝜋) and how to we show this? To prove the claim that if 𝐶⊆𝜅 is club then 𝜅∈𝜋(𝐶), how does appealing to the elementarity of 𝜋 help? I am not so familiar with ultrapowers (but I have some knowledge of ultraproducts and ultrafilters). $\endgroup$ – E.Green4321 Mar 23 at 1:31
  • $\begingroup$ @E.Green4321 The basics of elementary embedding and ultrapowers are covered in Jech's book. It's not something you can summarize in a few sentences. $\endgroup$ – Stefan Mesken Mar 23 at 13:52
  • $\begingroup$ Thank you so much for pointing me in the right direction and for your explanations. $\endgroup$ – E.Green4321 Mar 23 at 20:13
  • $\begingroup$ @E.Green4321 You're welcome! $\endgroup$ – Stefan Mesken Mar 23 at 21:16

Here is a detailed explanation of Jech's proof.

Let $D$ be a normal measure on $\kappa$. Suppose, towards a contradiction, that $\kappa$ is not Mahlo. Then there is some club $C \subseteq \kappa$ such that $$ C \cap \{ \alpha < \kappa \mid \mathrm{cof}(\alpha) = \alpha \} = \emptyset. $$

Since $D$ is normal, it contains all clubs. In particular $C \in D$. Since $D$ is closed under intersections, we therefore must have that $\{ \alpha < \kappa \mid \mathrm{cof}(\alpha) = \alpha \} \not \in D$ and hence that $$ \{ \alpha < \kappa \mid \mathrm{cof}(\alpha) < \alpha \} = \kappa \setminus \{ \alpha < \kappa \mid \mathrm{cof}(\alpha) = \alpha \} \in D. $$

By normality there is some $\lambda < \kappa$ such that

$$ E_\lambda = \{ \alpha < \kappa \mid \mathrm{cof}(\alpha) = \lambda \} \in D. $$

By replacing $E_\lambda$ with $E_\lambda \setminus \lambda$ we may and shall assume that $E_\lambda \cap \lambda = \emptyset$

For each $\alpha \in E_\lambda$ fix a strictly increasing, cofinal function $$ f_\alpha \colon \lambda \to \alpha $$ Now, for each $\xi < \lambda$, the function $$ g_\xi \colon E_\lambda \to \kappa, \ \alpha \mapsto f_\alpha(\xi) $$ is decreasing. Hence there is some $A_\xi \in D$ and some $y_\xi < \kappa$ such that $f_\alpha(\xi) = y_\xi$ for all $\alpha \in A_\xi$.


$$ A = \bigcap_{\xi < \lambda} A_\xi. $$

Since $\lambda < \kappa$ we have that $A \in D$.

Now let $\alpha \in A$. For all $\xi < \lambda$ we have $f_\alpha(\xi) = y_\xi$ is independent of $\alpha$ (by the construction of $A_\xi$). But $$ \alpha = \sup_{\xi < \lambda} f_\alpha(\xi) = y_\xi $$ is completely determined by the sequence $(y_\xi \mid \xi < \lambda)$.

Hence $A$ contains at most one element. This is a contradiction, since $D$ is non-principal.

It follows that $\kappa$ is Mahlo after all!


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.