I wanted to see for which ranks Bernays' Reflection Principle holds; that is, for every class and every property (allowing quantification over all classes) which is true about that class, there is a transitive set $u$ for which $\mathcal{P}(u)$ satisfies that property about that class's intersection with $u$. Like Vopenka did with Vopenka cardinals, I allow "arbitrary $A\subseteq V_\kappa$" to be classes in $V_\kappa$.

It turns out that once analyzed, these ranks bear a striking similarity to those ranks of Indescribable cardinals. I formalized this definition in ZFC as follows:

Let $\kappa$ be $\Pi_m$-Bernays when for every first-order $\Pi_m$ sentence $\phi$ in the language $\{\in,P\}$ where $P$ is an unary predicate symbol:

$$\forall A\in V_{\kappa+1}(\langle V_{\kappa+1};\in,A\rangle\models\phi\rightarrow\exists u\in V_\kappa(u\;\mathrm{is}\;\mathrm{transitive}\land\langle \mathcal{P}(u);\in,A\cap u\rangle\models\phi))$$

A similar definition in ZFC would be that for every first-order $\Pi_m$ unary formula $\phi$ in the language $\mathcal{L}_\in$:

$$\forall A\subseteq V_\kappa(\phi^{V_{\kappa+1}}(A)\rightarrow\exists u\in V_\kappa(u\;\mathrm{is}\;\mathrm{transitive}\land\phi^{\mathcal{P}(u)}(A\cap u)))$$

Of course, the existence of a $\Pi_{<\omega}$-Bernays cardinal is equivalent to the existence of a cardinal rank which satisfies Bernays' Reflection Principle. Thus, $\Pi_{<\omega}$-Bernays cardinals, in consistency strength, are somewhere above Bernays' Reflection Principle, which oddly enough implies the existence of an Inaccessible.

These cardinals are therefore consistency-wise stronger than that of Inaccessible cardinals.

Every cardinal is $\Pi_0$-Bernays. This is pretty simple once one considers that all of the unary first-order formulas in the language $\mathcal{L}_\in$ are either true for every set or true for no set.

This brings me to my question:

Are these cardinals inconsistent with ZFC? Are they any weaker or stronger than Indescribability?

  • $\begingroup$ You write "$c$ is a constant symbol". Do you really mean "$c$ is a relation symbol"? Because you want to interpret $c$ by $A$, which is a relation on the universe $V_\kappa$, not an element. $\endgroup$ – Goldstern Oct 21 '17 at 22:46
  • $\begingroup$ Oh, yes. Sorry. I should have said simply Predicate symbol, in fact. $\endgroup$ – Keith Millar Oct 21 '17 at 22:48
  • $\begingroup$ But if you let $\varphi$ be the formula $\forall x\ P(x)$, then (using $A:=V_\kappa$) $(V_\kappa,\in,V_\kappa)$ satisfies $\varphi$, but $(\mathcal P(u),\in,u)$ does not. $\endgroup$ – Goldstern Oct 21 '17 at 22:55
  • $\begingroup$ True; I will get rid of the first definition then. $\endgroup$ – Keith Millar Oct 21 '17 at 22:59
  • $\begingroup$ Oh, I found the reason my original definition was wrong. I should have defined it using $\langle V_{\kappa+1};\in,A\rangle$, as to accommodate for classes. $\endgroup$ – Keith Millar Oct 21 '17 at 23:10

I would like to add that, for $n>1$, the $\Pi_n-$Bernays cardinals are precisely the $\Pi_n^1-$indescribable cardinals. We can do this really simply by giving a $\Pi_2$ definition of the Axiom of limitation of size:

$$ALZ\leftrightarrow\forall x(\exists W\ni x\leftrightarrow\exists F,y(F\text{ is a bijection } F:x\rightarrow y\land\exists U\ni y))$$

Then $2^u\vDash ALZ$ precisely when $2^u$ satisfies limitation of size. Let $AU$ be the axiom of union (Which is also $\Pi_2$). Then whenever $u$ is infinite, transitive, and $2^u\vDash ALZ\land AU$, then $u$ is a Grothendrick universe and therefore of the form $V_\lambda$ for some inaccessible $\lambda$.

Then, for any $\Pi_2^1$ formula $\phi$ in the language of $\{\in,S\}$ such that $(V_{\kappa+1},\in,S)\vDash\phi$, to get some $(V_{\lambda+1},\in,S\cap V_\lambda)\vDash\phi$, let $\psi$ be the formula $ALZ\land AU\land\phi$, and take some transitive $u$ such that $(2^u,\in,S\cap u)\vDash\psi$. It is easy to see that $\kappa$ is inaccessible.


Every $\Pi_n^1$-Indescribable cardinal is $\Pi_n$-Bernays, and thus the $\Pi_{n+1}^1$-Indescribable cardinals are strictly of greater consistency strength than that of the $\Pi_n$-Bernay cardinals, and also the $1$-Indescribable cardinals are of strictly greater consistency strength than that of Bernay's principle itself.

Here's the proof:

A little Lemma before the proof is to show that if $\exists\alpha<\kappa(\langle V_{\alpha+1};\in,A\cap V_\alpha\rangle\models\phi)$ then $\exists u\in V_\kappa(\langle\mathcal{P}(u);\in,A\cap u\rangle\models\phi)$. Specifically, this becomes quite obvious when one realizes that $V_\alpha$ is such a $u$ described when the first condition holds for $\alpha$.

A cardinal $\kappa$ is $\Pi_n^1$-Indescribable iff for every first-order $\Pi_n$ sentence $\phi$ in the language $\{\in,P\}$ where $P$ is an unary predicate symbol:

$$\forall A\in V_{\kappa+1}(\langle V_{\kappa+1};\in,A\rangle\models\phi\rightarrow\exists\alpha<\kappa(\langle V_{\alpha+1};\in,A\cap V_\alpha\rangle\models\phi))$$

By the previous lemma, this implies that for every first-order $\Pi_n$ sentence $\phi$ in the language $\{\in,P\}$ where $P$ is a unary predicate symbol:

$$\forall A\in V_{\kappa+1}(\langle V_{\kappa+1};\in,A\rangle\models\phi\rightarrow\exists u\in V_\kappa(\langle\mathcal{P}(u);\in,A\cap u\rangle\models\phi))$$

Of course, this is the first definition I gave for a $\Pi_n$-Bernays cardinal. Thus, every $\Pi_n^1$-Indescribable cardinal is $\Pi_n$-Bernays.

  • $\begingroup$ It seems to me that $\Pi^1_n$-Bernays and $\Pi^1_n$-indescribable ought to be equivalent once $n$ is big enough (maybe 3 or so). The idea is that one could build into the formula $\phi$ enough requirements to ensure that any $u$ as in the definition of $\Pi^1_n$-Bernays is necessarily of the form $V_\alpha$. (There might even be something like this in Bernays's paper.) $\endgroup$ – Andreas Blass Jan 16 '18 at 20:25

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.