4
$\begingroup$

Is the axiom of determinacy (AD) consistent with the following choice principle, and if yes, does it hold in $L(ℝ)$ under AD:

Simultaneous well-orderability: For every function $f:P(Ord)→\text{Wellorderable}$, there is $g$ such that for all $x$, $g(x)$ is a well-ordering of $f(x)$.

Notes:
- $\text{Wellorderable}$ is the class of all well-orderable sets. $P(Ord)$ is the class of all sets of ordinals. Domains of functions are sets.
- Without the $P(Ord)$ restriction, the principle would be equivalent to the axiom of choice for collections of well-orderable sets (i.e. choice for $V→\text{Wellorderable}$), which does not prove AC, but contradicts AD.

The principle looks strong, but I do not see a contradiction with AD. Under AD, AC fails for functions $ω_1→ℝ$, but $ℝ$ cannot be well-ordered. Also, under AD, there is no function that for every Turing degree gives a real of that degree (even though the set of such reals is countable). However, Turing degrees are not sets of ordinals; while Turing degrees are encodable by reals, the encoding is non-unique.

Under $V=\mathrm{HOD}(P(Ord))$, it suffices to consider OD $f$ in the principle. Also, under $V=\mathrm{HOD}(P(Ord))$, a natural strengthening of the principle is that for every $s∈P(Ord)$, every well-orderable non-empty $\mathrm{OD}(s)$ set has an $\mathrm{OD}(s)$ element (equivalently, $\mathrm{OD}(s)$ well-ordering as we can extract elements one by one here). Under $V=\mathrm{HOD}(2^λ)$ for an ordinal $λ$ (and thus for $V=L(ℝ)$), the unstrengthened principle is equivalent to existence of a set of ordinals $t$ such that every $s∈P(Ord)$, every well-orderable non-empty $\mathrm{OD}(s)$ set has an $\mathrm{OD}(s,t)$ element (equivalently, $\mathrm{OD}(s,t)$ well-ordering).

Other choice principles under AD:
- ZF + AD proves DC in $L(ℝ)$.
- Under AD, axiom of choice for sets parameterized by reals (i.e. choice for $f:ℝ→V$) implies, and under $V=L(P(ℝ))$, is equivalent to $\text{AD}_ℝ$ + "$Θ$ is regular".
- It is open whether under AD, every set of reals must be in $\mathrm{HOD}(P(Ord))$ (under $\text{AD}^+$, $P(ℝ)∈L(Ord^{<Θ})$).

$\endgroup$

1 Answer 1

4
$\begingroup$

The following is something quite close. Perhaps with a bit more work one could get it in the exact form above. Assume $V = L(\mathbb{R})$ although much of this will work in $V = L(\mathscr{P}(\mathbb{R}))$ and $\mathsf{AD}^+$.

Woodin's dichotomy theorem states that under $\mathsf{AD}^+$ if $E$ is an equivalence relation on $\mathbb{R}$, then either $\mathbb{R}$ injects into $\mathbb{R} / E$ or $\mathbb{R} / E$ is wellorderable. If you look at the proof of the result, in the case that $\mathbb{R} / E$ is wellorderable, the wellordering is produced uniformly from the $\infty$-Borel code for $E$ (using an ultrapower by the Martin measure).

In $V = L(\mathbb{R})$, if $A \subseteq \mathbb{R}$, $S$ is a set of ordinals, and $A$ is $OD_S$, then $A$ has an $OD_S$ $\infty$-Borel code.

Let $\delta$ be an ordinal, $X$ is a set that $\mathbb{R}$ surjects onto, and $\Phi : \mathscr{P}(\delta) \rightarrow \mathscr{P}(X)$ so that for all $A \in \mathscr{P}(\delta)$, $\Phi(A)$ is a wellorderable subset of $X$. Let $\pi : \mathbb{R} \rightarrow X$ be a surjection. (All sets in $L(\mathbb{R})$ are ordinal definable from a real.) Without loss of generality, say $\pi$ is $OD$ and $\Phi$ is $OD$. For each $A \in \mathscr{P}(\delta)$, let $E_A$ be the equivalence relation on $\mathbb{R}$ defined by $x \ E_A \ y$ if and only if [$\pi(x) = \pi(y)$ and $\pi(x) \in \Phi(A)$] or $(\pi(x) \notin \Phi(A) \wedge \pi(y) \notin \Phi(A))$. Note that $E_A$ is $OD_A$ and $\mathbb{R} / E_A$ is essentially in bijection with $\Phi(A)$, except for one extra class. By the observation above, $E_A$ has an $OD_A$ $\infty$-Borel code. Let $S_A$ be the $OD_A$-least such $\infty$-Borel code using the canonical ordering of $HOD_A$. Apply the uniform procedure from the Woodin dichotomy theorem using this $\infty$-Borel code $S_A$ to uniformly obtain a wellordering of $\mathbb{R} / E_A$ and hence $\Phi(A)$.

If $X$ is not a surjective image of $\mathbb{R}$, then it can be decomposed uniformly into pieces which are images of $\mathbb{R}$ so perhaps one could extends this result subset of such $X$ (now having many equivalence relations to patch together). So possibly the result holds when $\Phi$ is a set and not class function.

The equivalence relation $E_A$ (or many equivalence relations) needs to be obtained uniformly. If $\Phi : \mathscr{P}(Ord) \rightarrow V$ and the image do not sit inside a common set, then it is immediately clear to how to obtain these equivalence relations.


See Theorem 3.2 and Fact 3.3 of "CARDINALITY OF WELLORDERED DISJOINT UNIONS OF QUOTIENTS OF SMOOTH EQUIVALENCE RELATIONS" by Chan and Jackson.

$\endgroup$
1
  • $\begingroup$ The argument works, and thus we get the principle under $\text{AD}^+ V = L(P(ℝ)) + ¬\text{AD}_ℝ$, and the $\mathrm{OD}_S$ strengthening I mentioned if also $V=HOD(ℝ)$ (and thus under $\text{AD} + V=L(ℝ)$). The patching of the equivalence relations works because everything is uniform; $¬\text{AD}_ℝ$ is used for surjection $Ord × ℝ → V$. $\endgroup$ Commented Sep 25 at 18:00

You must log in to answer this question.

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