7
$\begingroup$

EDIT: Joel's answer shows that no $\Sigma_2$ large cardinal property will do the job - however, $\Pi_2$ properties (such as unfoldability and its relatives) may still be useful.


Throughout this question, I'm working in "$V=L$". So when I say "large cardinal," I mean "small large cardinal".

Informally, I want to ask: How much of the "$L$-theory" of a cardinal does a large cardinal property imply? That is, if I know that $\alpha$ and $\beta$ both have some large cardinal property $(*)$, what sorts of (not necessarily first-order) sentences do I know that $L_\alpha$ and $L_\beta$ agree about?

Formally, fix an ordinal $\gamma$. Say that a large cardinal property $(*)$ is "$\gamma$-decisive" if whenever $\alpha, \beta$ have property $(*)$, we have $$L_{\alpha+\gamma}\equiv L_{\beta+\gamma}.$$ My question is:

What are some examples of $\gamma$-decisive large cardinal properties?

(CAVEAT: Obviously any large cardinal property can be arbitrarily decisive: suppose it has no instances, or exactly one! So really I want to know about large cardinal properties whose decisiveness can be proved in ZFC+V=L, or some strengthening thereof that does not limit the number of relevant large cardinals.)

Note that decisiveness looks a lot (at least to me) like indescribability; however, indescribability doesn't quite seem to do the job, unless I'm missing something.


I'm also interested in decisiveness with parameters - e.g. for a finite tuple of sets $\overline{c}\in L$, say that a large cardinal property $(*)$ is $\gamma$-decisive over $\overline{c}$ if whenever $\alpha,\beta$ have property $(*)$ with $\overline{c}\in L_\alpha\cap L_\beta$, we have $$(L_\alpha, \overline{c})\equiv(L_\beta,\overline{c}).$$ But understanding the parameter-free version should come first, and besides, I imagine the situations aren't too different.

By contrast, something potentially interesting happens if we allow, not tuples of elements of $L$, but tuples $\overline{C}$ of subsets of $L$ which are in $V$ but not necessarily $L$ - coded as unary predicates, of course. A really overkill example of this would be the $0^\#$ as a unary predicate, but already tamer examples could be interesting. For instance, suppose $\overline{C}$ is a finite sequence of mutually Cohen reals over $L$ - then what are some $\gamma$-decisive over $\overline{C}$ large cardinal properties? But this question seems much broader, and much harder, so I'll merely mention it as a curiosity. (Note that in this context, detrivializing things becomes a bit trickier, so this is a more informal question.)

$\endgroup$

1 Answer 1

6
$\begingroup$

Updated answer. I claim that none of the familiar large cardinal notions consistent with $V=L$ are provably $\gamma$-decisive for any $\gamma$. This includes the cases of wordly cardinals, inaccessible cardinals, uplifting cardinals, Mahlo cardinals weakly compact cardinals, $\Pi^n_m$-indescribable cardinals, totally indescribable cardinals, unfoldable cardinals and all the others.

First, as a warm-up, let's handle the case of $\Sigma_2$-definable large cardinal notions.

Theorem. If a large cardinal notion LC is $\Sigma_2$-definable and there are a proper class of them in $L$, then there is no $\gamma$ for which they are $\gamma$-decisive.

Proof. To see this, fix any $\gamma$, and let $\kappa$ be the least LC cardinal above $\gamma$. Note that both $\kappa$ and $\gamma$ are definable in $L_{\kappa+\gamma}$, since $\kappa$ is the largest cardinal of this structure, and $\gamma$ is least such that $\kappa+\gamma$ does not exist. Further, since $\kappa$ was the least LC cardinal above $\gamma$, it follows that $L_{\kappa+\gamma}$ will agree that there are no such large cardinals of that type in the interval $(\gamma,\kappa)$, since $\Sigma_2$ definitions express locally verifiable properties (see my blog post about Local properties in set theory). But the corresponding fact will not be true in $L_{\delta+\gamma}$ for some other much larger LC cardinal $\delta$, since once $\delta$ is large enough, then $L_{\delta+\gamma}$ will think that $\kappa$ is an LC cardinal. So $L_{\kappa+\gamma}$ is not elementarily equivalent to all $L_{\delta+\gamma}$, and so the LC large cardinal notion is not $\gamma$-decisive in this situation. QED

Next, with a slightly stronger assumption on consistency, we can handle the $\Pi_2$-definable notions.

Theorem. Assume that a large cardinal notion LC is $\Pi_2$-definable and there are a sufficiently stationary proper class of them in $L$, then there is no $\gamma$ for which they are $\gamma$-decisive.

Proof. Fix any $\gamma$, and assume that the class of LC cardinals in $L$ meets every $\Pi_2$-definable class club. Let $C$ be the club of $\Sigma_2$-correct cardinals $\delta$, those for which $V_\delta\prec_{\Sigma_2} V$. This is $\Pi_2$-definable. Let $\kappa$ be the least LC cardinal in $C$ above $\gamma$. Note that both $\kappa$ and $\gamma$ are definable in $L_{\kappa+\gamma}$. Further, since $\kappa$ is $\Sigma_2$-correct, it must be a beth-fixed-point, and so $L_\kappa=(V_\kappa)^L$. Since $L_\kappa$ is $\Sigma_2$-correct and hence also $\Pi_2$-correct, it is correct about the LC cardinals below $\kappa$. And it is also correct about the $\Sigma_2$-cardinals below $\kappa$. So by the minimality of $\kappa$, the structure $L_\kappa$ can see that there is no $\Sigma_2$-correct LC cardinal below $\kappa$ and above $\gamma$, and this is part of the theory of $L_{\kappa+\gamma}$. But now, if $\delta$ is a larger LC cardinal in $C$, then $L_{\delta+\gamma}$ will be able to see that there is a $\Sigma_2$-correct LC cardinal below $\delta$. And so $L_{\kappa+\gamma}$ and $L_{\delta+\gamma}$ do not have the same theory. So the LC cardinals are not $\gamma$-decisive. QED

Finally, we can use this idea to push through an argument for any first-order definable large cardinal notion.

Theorem. No first-order definable large cardinal notion LC is provably $\gamma$-decisive for any $\gamma$, if it is consistent that there is a stationary proper class of LC cardinals in $L$.

Proof. Suppose that the LC large cardinal notion is $\Sigma_n$-definable, and the class of LC cardinals in $L$ is stationary with respect to all $\Pi_n$-definable class clubs. Fix any $\gamma$, and let $\kappa$ be the least $\Sigma_n$-correct LC cardinal above $\gamma$. There is such a $\kappa$ because the class of $\Sigma_n$-correct ordinals is a $\Pi_n$-definable class club. Both $\kappa$ and $\gamma$ are definable in $L_{\kappa+\gamma}$, and $L_\kappa$ is correct about LC cardinals and about $\Sigma_n$-correctness. So $L_\kappa$ will see no $\Sigma_n$-correct LC cardinals above $\gamma$. But if we use a much larger $\Sigma_n$-correct LC cardinal $\delta$, then $L_\delta$ will see that $\kappa$ is a $\Sigma_n$-correct LC cardinal above $\gamma$. So $L_{\kappa+\gamma}$ and $L_{\delta+\gamma}$ will have different theories. QED

$\endgroup$
16
  • $\begingroup$ Sorry, I'm being slow: why will $L_{\kappa+\gamma}$ realize that $\kappa$ is indeed a large cardinal of the relevant type? (That is, I don't see why $L_{\kappa+\gamma}$ is correct about $\Sigma_2$ facts - what if the witness lives in $L_\mu$ for some $\mu>\kappa+\gamma$?) $\endgroup$ Nov 5, 2016 at 0:49
  • 1
    $\begingroup$ I see, I didn't explain it correctly. The point is that $L_{\kappa+\gamma}$ won't have any of those cardinals below $\kappa$, but larger $L_{\delta+\gamma}$ will have such cardinals below $\delta$. I'll edit to explain it correctly. $\endgroup$ Nov 5, 2016 at 0:51
  • 1
    $\begingroup$ Every $L_\theta$ is $\Sigma_1$ elementary in $L$, for $\theta$ a cardinal, and so $\Sigma_2$-properties are upward absolute from $L_\kappa$. $\endgroup$ Nov 5, 2016 at 0:55
  • 1
    $\begingroup$ There are many $\Pi_2$ large cardinal notions, consistent with $L$, such as: unfoldable, strongly unfoldable, and others. Those notions are defined by properties that involve a universal $\forall\theta$, and then having an embedding jumping beyond $\theta$. Violations of the property are locally verifiable, hence $\Sigma_2$, and so the notion is $\Pi_2$. My argument does not work for that case, so unfoldable might be a natural case. $\endgroup$ Nov 5, 2016 at 1:01
  • 1
    $\begingroup$ Incidentally, I have a blog post on $\Sigma_2$ properties and the idea of local verification. See jdh.hamkins.org/local-properties-in-set-theory. $\endgroup$ Nov 5, 2016 at 1:02

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.