8

It is possible to prove theorems of the form "if $\phi$ is provable in ZFC, then $\phi$ is provable in ZF". For example, let $\phi$ be a statement that is absolute between $V$ and $L$. If $\phi$ were not provable in ZF, then there would be a model $M$ in which it is false, which in turn implies that it is false in $L^M$, hence not provable in ZFC.

I was wondering whether there are natural such statements all known proofs of which involve axiom of choice (or whose proofs become substantially easier when AC is used).

Following Andres Caicedo's answer and comments here, it is possible to "cheat" and turn ZFC proofs into ZF proofs by carrying out the argument in L and then using absoluteness. What I had in mind was to see if there are examples where AC "takes care" of a procedure which in theory we could do without it.

Burak
  • 3,696
  • Someone told me that there is a theorem that if a ring satisfies $x^n = x$ for some integer $n > 0$, then the ring is commutative. The proof for all $n$ apparently used the axiom of choice. But the proof for specific $n$ (e.g. $n = 2, 3, 4, 5$) become progressively more difficult as $n$ gets bigger. I heard about this many years ago, so I cannot attest to either its truth or timeliness. – Stephen Montgomery-Smith May 31 '14 at 00:06
  • I have not heard this about choice. – Rene Schipperus May 31 '14 at 00:40
  • 1
    If $\kappa$ is an infinite (well-ordered) cardinal, $\kappa\to(\kappa,\omega)^2$. See here for a combinatorial proof in $\mathsf{ZF}$ of a consequence of this statement. – Andrés E. Caicedo May 31 '14 at 01:32
  • (Rene (I assume the same Rene from a comment above) proved $\omega_1\to(\alpha)^2$ for all $\alpha<\omega_1$. The argument uses choice, I do not know if this holds in $\mathsf{ZF}$.) – Andrés E. Caicedo May 31 '14 at 01:32

2 Answers2

5

Here is some general theorem that you can prove.$\newcommand{\Ord}{\mathsf{Ord}}$

We write $(\exists^{\Ord}x)\varphi$ and $(\forall^{\Ord}x)\varphi$ to mean that we are bounding the quantifier on $x$ as a set of tuples of ordinals. Meaning $x\subseteq\Ord^{<\omega}$, and $\varphi$ holds. (You can observe that this doesn't increase the complexity of the formula in the Levy hierarchy more than a usual quantifier would.)

If $\varphi(x)$ is a statement which is upwards absolute, then $(\forall^\Ord x)\varphi(x)$ is provable with the axiom of choice if and only if it is provable without the axiom of choice. The proof is almost trivial, given $x\subseteq \Ord^{<\omega}$, we can consider $L[x]$ as a model of choice, $\varphi(x)$ holds there, and it is upwards absolute, so it holds in $V$. $\qquad\square$

This gives an immediate plethora of examples in the sense of finite coloring theorems and properties, e.g. "every coloring of a subset of $[\kappa]^{<\omega}$ in $\lambda$ colors, has a subset of order type/cardinality $\alpha$ which is homogeneous/anti-homogeneous" are all statements of the form $\Pi_2$ whose universal quantifiers can be replaced by $\forall^\Ord$.


You can also take a look in Herrlich's book The Axiom of Choice which includes many examples of less set-theoretical nature.

Asaf Karagila
  • 393,674
  • Is there a name for the sublanguage of the language of ZFC generated by those $\mathsf{Ord}$ bounded quantifiers you mention? I have been pondering the question of what we can and can't express in this sublanguage for the past couple of weeks. – goblin GONE May 31 '14 at 06:03
  • 3
    Not that I know of. I don't even know if the explicit definition of ordinal bounded quantifier was written somewhere before this. – Asaf Karagila May 31 '14 at 06:04
  • Nice, I'm reading it now. – goblin GONE May 31 '14 at 06:06
  • So $(∃^\mathrm{Ord}x)\varphi$ asserts that there exists a set of finite sequences of ordinals satisfying $\varphi$, right? By the way, I misread your notation; the sublanguage I've been pondering is the one generated by quantifiers of the form $\exists x \in \mathrm{Ord}$ and $\forall x \in \mathrm{Ord},$ altogether much less expressive. – goblin GONE May 31 '14 at 06:16
3

If I understand the question, one example of a theorem that becomes substantially easier when the axiom of choice is used is Hindman's theorem.

This is a theorem of infinitary combinatorics:

Hindman's Theorem. If $k \in \mathbb{N}$ and $g\colon \mathbb{N} \to \{1, \ldots, k\}$, there is an infinite set $I \subseteq \mathbb{N}$ and a $c \leq k$ such that for all nonempty finite $F \subseteq I$, $g(\sum_{i \in F} i) = c$. In words, for every finite coloring of $\mathbb{N}$ there is an infinite set $I$ and a color $c$ such that the sum of any positive finite number of distinct elements of $I$ receives color $c$.

There are (essentially) three known proof methods for the theorem. One method uses a special kind of ultrafilter on $\mathbb{N}$, which is constructed by an application of Zorn's lemma. There is an exposition of this proof by Leo Goldmakher here; the construction is theorem 3.1. This method gives the "easiest" proof of the theorem.

The second method, due to Furstenberg and Weiss, uses methods of topological dynamics and ergodic theory. This proof also uses high powered techniques, although it does not directly use ultrafilters.

The third method is based on Hindman's original proof, which is much more difficult but which uses much more elementary methods. Hindman's original proof can be formalized in second-order arithmetic (as shown by Blass, Hirst, and Simpson 1987), so it does not have any essential uses of the axiom of choice. Simplified expositions of this method are given by Baumgartner (1974) and Towsner (2011), online here.

Carl Mummert
  • 81,604
  • 2
    Baumgartner's proof, which you describe as a simplified exposition of Hnidman's, can also be viewed as a forerunner of ultrafilter arguments. Although Baumgartner doesn't explicitly talk about forcing (if I remember correctly), he's working with the natural forcing for adding (what I later called) a stable ordered-union ultrafilter. Also in terms of complexity, his argument is intermediate between Hindman and Galvin-Glazer. Baumgarner uses (again if I remember correctly) $\Pi^1_2$ comprehension. – Andreas Blass May 31 '14 at 14:27
  • 1
    Thanks. For those who are interested, Baumgartner's original paper is at http://www.sciencedirect.com/science/article/pii/0097316574901034 . I believe it is available to download free of charge. – Carl Mummert May 31 '14 at 14:54
  • I do think there are other examples of theorems like this, where there is a "hard" proof that uses only basic methods, and also "easy" proofs that use more complicated methods, which may well involve the axiom of choice (possibly in the form of dependent choice). This was the first concrete one that came to my mind. – Carl Mummert May 31 '14 at 14:56
  • 1
    @Carl: For example Heine-Borel for $\Bbb R^n$; which uses fundamentally the axiom of choice, but can be proved completely without it. – Asaf Karagila May 31 '14 at 15:41
  • The link to Leo Goldmahker's paper 404s, and is not on archive.org. Do you have another link? – Zemyla Mar 11 '24 at 01:32
  • @Zemyla: Try this link: https://web.williams.edu/Mathematics/lg5/Hindman.pdf . I saved it on archive.org now as https://web.archive.org/web/20240312093506/https://web.williams.edu/Mathematics/lg5/Hindman.pdf – Carl Mummert Mar 12 '24 at 09:34