In this answer, Professor Hamkins gives a proof that for models $M$ of ZF, $M$ being a model of $\text{ZFC} + V = \text{HOD}$ is equivalent to there being a definable well-ordering of the universe:


His argument easily extends to an equivalence of these properties to $M$ having a well-ordering of the universe definable with respect to an ordinal parameter. So, if there is well-ordering of the universe with respect to some parameter $p,$ but there is not a well-ordering of $V$ definable without parameters, then necessarily $p \not \in \text{OD}$. Is this situation possible? My intuition is that it shouldn't be possible, since I don't think a non-ordinal parameter should be able to define something so fundamental when an ordinal cannot do the same.

  • $\begingroup$ I'd say welcome to M.SE, Elliot, but it seems you've been here for a while. But it's good to see you utilizing this website! $\endgroup$ – Asaf Karagila Jan 10 '16 at 22:28
  • $\begingroup$ Ay, I've been meaning to for a long time but never got around to it till now. $\endgroup$ – Elliot Glazer Jan 11 '16 at 12:46

Of course.

If you start with a model where there is a definable well-ordering, say $V=L$, and you add a single Cohen real $r$ you have that:

  1. $V=L[r]$, so there is a definable well-ordering with a parameter $r$ (e.g. given two sets in $L[r]$ ask which one has a name appearing first in the order of the ground model, here $L$, that when interpreted with $r$ as the generic give you the two sets).
  2. Since the Cohen forcing is homogeneous, $L=\mathsf{HOD}^L=\mathsf{HOD}^{L[r]}\neq L[r]$.

So while there is a well-ordering definable from a parameter, which in this case is a real number (read: a subset of $\omega$), there is no such well-ordering which is definable without parameters or from ordinals.

(This argument shows that any set forcing over a model with global well ordering from parameters will also have a global well ordering definable from a parameter. You can violate that with a class forcing, though.)


I just noticed this question, which I find quite interesting.

I thought I'd mention the following related result, which some readers may find interesting:

Theorem. The following are equivalent.

  1. The universe is HOD of a set: $\exists b\ (V=\text{HOD}(b))$.
  2. The axiom V=HOD is forceable.
  3. Somewhere in the generic multiverse, the universe is HOD of a set.
  4. Somewhere in the generic multiverse, the axiom V=HOD holds.

The proof is contained in my blog post, Being HOD-of-a-set is invariant throughout the generic multiverse.

In particular, it follows that the axiom V=HOD is a switch, in models for which $V=\text{HOD}(b)$, since it can be forced on and then off again as much as you like. If $V=\text{HOD}$ holds, then you can do the forcing in Asaf's answer, adding a Cohen real, and $V\neq\text{HOD}$ in the extension $V[c]$, but then you can force $V=\text{HOD}$ again in a further forcing extension. And furthermore, whenever $V=\text{HOD}(b)$, then you can force $V=\text{HOD}$, and the assertion $\exists b\ V=\text{HOD}(b)$ is invariant by forcing.

  • $\begingroup$ Actually, I see that I must have noticed it earlier, since I have evidently upvoted both the question and Asaf's answer already. $\endgroup$ – JDH Aug 8 '16 at 1:37
  • 1
    $\begingroup$ I agree that, although now standard, the notation is not great. What is your preferred notation? $\endgroup$ – JDH Aug 8 '16 at 11:17
  • 1
    $\begingroup$ The thing is that HOD and L notations should agree, and they don't. I don't mind which way is what, just that they agree. $\endgroup$ – Asaf Karagila Aug 8 '16 at 11:21
  • 1
    $\begingroup$ This may be obvious, but what exactly is the relationship of $\exists b\,V=\mathrm{HOD}(b)$ to the existence of a well-ordering of the universe definable with a parameter? Are these equivalent? $\endgroup$ – Emil Jeřábek Jul 12 '17 at 12:51
  • 1
    $\begingroup$ Yes, those are equivalent, and with the same parameter $b$. Some authors write $\text{HOD}_{\{b\}}$, to indicate that we are taking the set $b$ as a parameter, not the set of elements of $b$. Given the well-order, any set is the $\alpha^{th}$ with respect to that order, and if $V=\text{HOD}(b)$, then there is a canonical order in terms of the definitions and ordinal parameters that are used. $\endgroup$ – JDH Jul 12 '17 at 13:01

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.