3

If we have $\forall x(x\in X\to\exists y(y\in Y\land\varphi(x,y)))$, then we can define a function $f$ from $X$ to $\mathcal P(Y)\setminus\{\emptyset\}$ such that $f(x)=\{y\in Y|\varphi(x,y)\}$, and use axiom of choice to define a choice function $g$ from $\mathcal P(Y)\setminus\{\emptyset\}$ to $Y$ so that the composition $g\circ f$ is a function mapping each $x\in X$ to some $y\in Y$ that satisfies $\varphi(x,y)$.

But if we instead have $\forall x(x\in X\to\exists y(\varphi(x,y)))$, is it still possible to construct a function from each $x\in X$ to some $y$ that satisfies $\varphi(x,y)$, with codomain given by axiom schema of replacement? Is it possible to "make a choice" from the universe of sets instead of from some set $Y$?

RobPratt
  • 45,619
  • The universe is clearly not set-sized, but I do learn from that question about axiom of global choice, although sadly it is not in the language of ZFC. Does the existence of such meta-axiom imply it is indeed impossible to choose from the universe with the tools provided by ZFC? – zap kabosu Jun 11 '22 at 14:44
  • Read the answer at the link, not just the question. – Eric Wofsey Jun 11 '22 at 15:09

1 Answers1

0

If you assume the axiom of foundation, then yes, this is possible. If you do not assume the axiom of foundation, it may be impossible.

The relevant axiom scheme is the axiom scheme of (strong) collection. This scheme can be phrased as follows:

Suppose $\forall x \in X \exists y (\phi(x, y))$. Then there is some indexed family of inhabited sets $\{Y_x\}_{x \in X}$ such that $\forall x \in X \forall y \in Y_x (\phi(x, y))$.

Over the axiom of choice, the collection scheme is equivalent to your proposed choice scheme.

For suppose the collection scheme, and pick some $\{Y_x\}_{x \in X}$ as described. Then the axiom of choice tells us $\prod\limits_{x \in X} Y_x$ has an element $f$, and we see that $\forall x \in X (\phi(x, f(x)))$.

Conversely, suppose there is some function $f$ with domain $X$ such that $\forall x \in X (\phi(x, f(x)))$. Then we can define $Y_x = \{f(x)\}$. As you’d expect, no choice is needed for this direction.

Let’s take a look at the axiom scheme of collection. It is well-known that this scheme follows from $ZF$ set theory using the so-called “Scott’s Trick”. This requires some knowledge of the “cumulative hierarchy” and the axiom of foundation, but if you’ve made a usual introductory study of $ZF$ set theory, it will be within your grasp.

However, it is also known that in $ZFC^{-}$ - that is, $ZFC$ without the axiom of foundation - the axiom scheme of collection can fail. For more details on this, see this question and answer. Be warned that this answer is highly technical and references another technical answer - you should be familiar with permutation models of ZF with atoms and how these can be translated into models with Quine atoms before proceeding.

Notwithstanding all of this, there is a broad class of instances of the axiom scheme of collection, known as “structural collection”, which always holds even in $ZFC^{-}$ due to the presence of the axiom of choice.

For some elaboration on this and a summary of this topic (and some alternatives to foundation), I would recommend my own answer to this question. My answer does use the language of classes, so be warned.

From your comments, you seem like you may be unfamiliar with how to translate statements about classes into statements about sets; I find the perspective in Kunen’s book Set Theory: An Introduction to Independence Proofs to be a useful one if you dislike the “ontology” of classes. He explains how to translate a statement involving classes into a metatheorem about set theory.

Mark Saving
  • 31,855
  • Thanks for the answer. I was trying to prove a version of transfinite recursion without ordinals when I encountered this problem, and now I know it takes Von Neumann universe to resolve, which in term takes transfinite recursion to define, rendering this path a death end. I guess I should go with ordinals, then transfinite recursion, then Von Neumann universe after all. – zap kabosu Jun 11 '22 at 19:32