2
$\begingroup$

Is there a principle $\sf P$ that $\sf ZFC$ [or some suitable extension of it] proves to be a strict limit on choice principles?

By a choice principle I mean a sentence (or scheme) that is equivalent [over $\sf ZF$] to a selection principle. The latter is a sentence of the form:

$\forall x_1,..,\forall x_n \exists F \forall y: \Omega \to F(x_1,..,x_n)(y) \in y$

Where $\Omega$ is a formula, that may use "$x_1,..,x_n,F,y$" among its free variables.

that is not provable in $\sf ZF$ alone.

By $\sf P$ being a strict limit on choice principles, it means that for every choice principle $\sf c$ we have $\sf ZFP \vdash c$, and $\sf ZFc \not \vdash P$; and such that there does not exist any principle $\sf H$ such that: $\sf ZFP \vdash H$, and $\sf ZFH \not \vdash P$, and for every choice principle $\sf c $ we have $\sf ZFH \vdash c$, and $\sf ZFc \not \vdash H$.

$\endgroup$

1 Answer 1

10
$\begingroup$

The answer is no. Every theorem of ZFC is equivalent to a choice principle. I will prove this is true, but nevertheless also I will be first to agree that the manner in which this is true is trivial and ultimately unsatisfying. I take the real lesson here to be that the property of being a choice principle in this sense is not a useful notion.

Theorem. Every statement provable in ZFC is equivalent over ZF to a choice principle for a definable set family $\Omega$.

Proof. Suppose that $\psi$ is a sentence provable in ZFC. Define the family $\Omega$ so that it is the empty family, in the event that $\psi$ holds, and otherwise it is the family $V_\alpha-\{\emptyset\}$, where $\alpha$ is the least ordinal such that this family has no choice function. (Note that if $\psi$ fails, then AC must fail and so there will indeed be such an ordinal $\alpha$.)

I defined the family $\Omega$ so that ZF proves both that $\psi$ implies $\Omega=\emptyset$, and also that $\neg\psi$ implies that $\Omega=V_\alpha-\{\emptyset\}$, where $\alpha$ is least such that this has no choice function.

Now we argue in ZF. If $\psi$ holds, then the family given by $\Omega$ is empty, and so it vacuously has a choice function. And if $\psi$ does not hold, then AC fails and so indeed there is a first ordinal where the family of nonempty sets of rank below $\alpha$ has no choice function, and $\Omega$ is this family. So in this case, $\Omega$ does not have a choice function.

So the sentence $\psi$ is equivalent over ZF to the existence of a choice function for $\Omega$, as desired. $\Box$

The theorem is meant as a scheme over all $\psi$ that are provable in ZFC. So for each instance of the theorem, $\psi$ is a fixed sentence, and this sentence appears in the definition of the family $\Omega$.

$\endgroup$
16
  • 1
    $\begingroup$ It took me a while to understand what "if $\psi$ holds" and "if $\psi$ fails" mean. It's not about provability in ZF, but rather just a funky construction of $\Omega$ inside ZF, correct? Maybe there's a way to make this more explicit. $\endgroup$ Commented Dec 25, 2023 at 23:05
  • 1
    $\begingroup$ Yes, I am using $\psi$ in my definition of $\Omega$. $\endgroup$ Commented Dec 25, 2023 at 23:06
  • 2
    $\begingroup$ I tried to explain it a little better. Hope that helps... $\endgroup$ Commented Dec 25, 2023 at 23:16
  • 1
    $\begingroup$ And let there be no doubt that I view this family $\Omega$ in a purely instrumental manner---it enables me to prove the theorem and I make no claim about its naturality in any way beyond this. $\endgroup$ Commented Dec 26, 2023 at 0:51
  • 2
    $\begingroup$ But in that case, why bother defining "selection principle" and "choice principle" and so forth, since AC is the strongest one and that would be the only one that matters. You are asking instead: Is there a minimal extension of ZFC? The answer is no, because the Lindenbaum algebra is dense. $\endgroup$ Commented Dec 26, 2023 at 15:57

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .