Suppose we a working with an axiomatic set theory such as $\bf ZFC$. Formulae with 2 free variables define class relations. "Small" class relations can be turned into sets of ordered pairs, the others are proper classes. Some formulae define relations that are well-orders. A "small" such relation can be turned into a well-ordered set that is order-isomorphic to some ordinal $\alpha\in\bf Ord$. Proper class relations also be can well-orders, e.g. a formula $\phi(x,y)$ saying that $x$ and $y$ are both ordinals and $x\in y$ defines the usual well-order on the class of ordinals $\bf Ord$.

Now, we can also write a formula that well-orders all ordinals except $\bf 0$ in the usual way, but considers $\bf 0$ as larger than all of them. Apparently, it gives us an order type that is "longer" than $\bf Ord$, and can be denoted as $\bf Ord+\bf1$. In a similar way, formulae having even "longer" order types such as $\bf Ord+\bf Ord$, etc. can be written. These order types are not sets existing in the universe, but rather, informal equivalence classes of formulae defining well-ordered class relations under (provable) order isomorphism. But, still, we can think of them as very long "ordinals" that occur when we count beyond all set-sized ordinals and even beyond the proper-class-sized "ordinal" $\bf Ord$. I'm interested in how far those "ordinals" can extend, and whether there is some systematic or formal way to study them, or develop ordinal arithmetic on them. Perhaps, a different axiomatic theory, rather than $\bf ZFC$, would be more suitable for that?


A good way in my opinion to study this phenomenon is to switch from $V$ to transitive set-sized models of $\mathsf{ZFC}$ (making appropriate existence assumptions of course). Specifically, suppose $M$ is a transitive model of $\mathsf{ZFC}$. Then we can talk unproblematically about the supremum $S_M$ of the order-types of $M$-definable well-orderings; this supremum will always be $>\mathsf{Ord}^M$ (which is just $\mathsf{Ord}\cap M$ of course) by precisely the example you give.

Especially when $M$ is countable (if $M$ is uncountable we need to perform some circumlocutions) there is a natural computability-theoretic ordinal to compare with $S_M$, namely the supremum $O_M$ of the ordinals computable from an isomorphic copy of $M$ with domain $\omega$. This $O_M$ "sees" that $M$ is countable, so on the face of things doesn't seem to be connected to $M$ very well at all. However, it turns out that this is a rather important thing to consider, so it's natural to compare it to $S_M$. We always have $O_M\ge S_M$, and for $M$ "sufficiently closed" this inequality is strict; however, equality can occur.

If we really want to talk about class relations on $V$ rather than "set-ifying" things, we need to look at some class theory. The standard examples are $\mathsf{NBG}$ (which is a conservative extension of $\mathsf{ZFC}$) and $\mathsf{MK}$ (which isn't). The latter is more appropriate for things related to class-length recursions, which suggests that it's the right framework here. There is definitely some work done on class theories and their relation to "long" ordinals; I think Joel David Hamkins has some good papers on the topic.


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.