A cardinal is Reinhardt if $\kappa$ is the critical point of a nontrivial elementary embedding of $V$ to itself, where $V$ is the class of all sets. As Reinhardt cardinals are inconsistent with $\mathrm{ZFC}$, work in $\mathrm{ZF}j$, which is $\mathrm{ZF}$ with replacement and separation for formulas including the function symbol $j$, and an axiom schema stating $j$ is a nontrivial elementary embedding $V\to V$.

Let $(V_\alpha)_{\alpha\in\mathrm{Ord}}$ be the usual von Neumann hierarchy whose union is $V$. For two elementary embeddings $j,k:V\to V$, define the embedding $j\cdot k$, the application of $j$ to $k$, to be $\bigcup_{\alpha\in\mathrm{Ord}}j(k\cap V_\alpha)$. The application operation is familiar from the study of rank-into-rank axioms like $\mathrm I3$, a usual reference is Laver's "On the Algebra of Elementary Embeddings of a Rank Into Itself". It is a usual exercise to verify that if $j$ and $k$ are elementary, then $j\cdot k$ is elementary.

If there exists a nontrivial elementary embedding $j:V\to V$ along with a Reinhardt cardinal $\kappa$, then $j\cdot j$ is also elementary, and $j(\kappa)$ is also Reinhardt as it is the critical point of $j\cdot j$. Similarly, $j\cdot (j\cdot j)$ is elementary, and $j(j(\kappa))$ is Reinhardt. More generally, define the critical sequence of $j$ as usual, by $\kappa_0=\kappa$, $\kappa_{n+1}=j(\kappa_n)$, and $\lambda=\mathrm{sup}\{\kappa_n\mid n<\omega\}$. Additionally define $j^0=j$ and $j^{n+1}=j\cdot j^n$. Each $\kappa_n$ in the critical sequence is Reinhardt, as witnessed by the elementary embedding $j^n$. So if there exists a Reinhardt cardinal, there must exist countably many.

Assume there exists a Reinhardt cardinal, and let $\alpha$ be an arbitrary von Neumann ordinal. Does it follow that there is a Reinhardt cardinal $>\alpha$?

Some possible direction: In section 6 of "I0 and rank-into-rank axioms", Dimonte considers direct systems $(M_\beta,j_{\beta,\gamma})_{\beta<\gamma<\alpha}$ of elementary embeddings (although these embeddings are only from $V_\lambda$ to $V_\lambda$), taking direct limits at limit ordinal steps, but due to the different definition $j^{\alpha+1}=j^\alpha\cdot j^\alpha$, if I am correct the critical point of $j^2$ is no longer $j(j(\kappa))$, but only $\kappa$, so no larger Reinhardt cardinals are produced.

  • $\begingroup$ By the way I don’t think you’re correct about the infinite iterations. The reason they don’t produce Reinhardt cardinals is that e.g. $j_{\omega,\omega+1}$ is not an embedding from $V$ to itself, but rather from $M_\omega$ to itself. The fact that each finite iterate $M_n$ is $V$ does not imply the same for $M_\omega$. See Schlutzenberg’s paper on iteratations of embeddings from $V$ to $V$ for more information. $\endgroup$ Apr 25 at 22:12
  • $\begingroup$ arxiv.org/abs/2002.01215 $\endgroup$ Apr 25 at 22:17
  • $\begingroup$ @GabeGoldberg Good point, I now think you are correct. I will also make a small change about what I originally intended with the remark about the embeddings in Dimonte's paper. $\endgroup$
    – C7X
    Apr 26 at 3:59
  • $\begingroup$ I guess the point I was making was unclear. When you iterate an embedding (as in Dimonte's paper), the critical points do increase. In fact, $j^\alpha(j^\alpha) = j^0(j^\alpha)$. The same construction can be done for $j : V\to V$ as well as $j:V_\lambda\to V_\lambda$. Also I think it is bad notation to put the $\alpha$ in the superscript since it makes it look like we are composing $j$ with itself. $\endgroup$ Apr 26 at 15:56

1 Answer 1


No, if the existence of a Reinhardt is consistent, then it is consistent with a Reinhardt cardinal that the class of inaccessible cardinals is bounded in the ordinals. Indeed, if $j : V\to V$ is a nontrivial elementary embedding and $\lambda$ is the least fixed point of $j$ above its critical point, then let $\delta$ be the least inaccessible above $\lambda$ if there is one -- if not, we're done. Note that $j(\delta) = \delta$, so $(V_\delta,V_{\delta+1})$ is a model of full second-order ZF + there is a Reinhardt cardinal + every inaccessible cardinal is less than $\lambda$.

One interesting question here is whether it is consistent that there are exactly $\omega$-many Reinhardt cardinals (in ordertype). It seems like one might have a chance of showing this is true in the model of NBG where all classes are definable from a single elementary $j : V\to V$.


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.