11
$\begingroup$

On page 241 of William Reinhardt's paper "Ackermann's set theory equals ZF" (Annals of Math. Logic vol. 2, 1970), question 4.15 is the following:

How large is the first ordinal $\gamma$ such that there are ordinals $\alpha_i$, $i\in\omega$ for which $i<j\rightarrow\alpha_i<\alpha_j$ and

3) $(V_\gamma,\in,\alpha_i)_{i\in\omega}\equiv(V_\gamma,\in,\alpha_{i+1})_{i\in\omega}$?

Is the answer to this question now known?

Here $V_\gamma$ is the $\gamma$th level of the von Neumann hierarchy and $\equiv$ denotes elementary equivalence. I am reading the structures as $(V_\gamma,\in,\alpha_0,\alpha_1,\alpha_2,\ldots)$ and $(V_\gamma,\in,\alpha_1,\alpha_2,\alpha_3,\ldots)$.


Some existing work: these ordinals $\gamma$ appear to be such that $V_\gamma$ has a set of indiscernibles of order type $\omega$ (although I can't prove this - in the next section of this post there is a related property 3').) By lemma 1 of F. R. Drake's "A fine hierarchy of partition cardinals" (Fundamenta Mathematicae vol. 81, iss. 3, 1974), if $\gamma$ is such that $\gamma\rightarrow(\omega)^{<\omega}_{2^{\aleph_0}}$, then $\gamma$ has property 3). Reinhardt remarks that via partition properties, if $\gamma$ is measurable then $\gamma$ has property 3). However Reinhardt says "But the condition 3) is probably quite weak". I was unable to produce any sequence $(\alpha_i)_{i\in\omega}$ for some small $\gamma$, e.g. the least inaccessible.


If no complete answer is known, some of my work: I do not know how to deal with $\gamma$ with property 3) that may not be cardinals, but I have been trying to find partition properties implied by 3) as it looks like an indiscernibility property. My current progress is on a similar principle 3'), although I was unable to prove it equivalent to 3):

There are ordinals $\alpha_i$, $i\in\omega$ for which $i<j\rightarrow\alpha_i<\alpha_j$ and

3') For any finite sequences $i_0<i_1<\ldots<i_n$ and $j_0<j_1<\ldots<j_n$ of natural numbers, $(V_\gamma,\in,\alpha_{i_0},\alpha_{i_1},\ldots,\alpha_{i_n})\equiv(V_\gamma,\in,\alpha_{j_0},\alpha_{j_1},\ldots,\alpha_{j_n})$.

I believe I have proved that $\gamma$ is a cardinal and has property 3'), then $\gamma$ satisfies a definable partition principle which might be called "$\gamma\overset{\Sigma_\omega}{\rightarrow}(\omega)^{<\omega}_{2^{\aleph_0}}$", namely for all $f:[\gamma]^{<\omega}\to 2^{\aleph_0}$ such that $f$ is parameter-free definable on $(V_\gamma,\in)$, there is a $Y\subseteq\gamma$ of order type $\omega$ which is homogeneous for $f$.

This is proved by modifying the reverse direction of theorem 2.1 in ch.8, §2 of Drake's Set Theory: An Introduction to Large Cardinals (1974) to use only the structure $\mathfrak A=(V_\gamma,\in)$ in place of structures $\mathfrak A=(\gamma,<,(R_{n,\beta})_{n,\beta<\omega})$, as property 3') does not include extra relation symbols like the book's indiscernibility property does. As $f$ is definable on $(V_\gamma,\in)$, for each relation $R_{n,\beta}$ there is a formula $\rho_{n,\beta}$ in the unaugmented language such that $R_{n,\beta}(x_1,\ldots,x_n)$ iff $(V_\gamma,\in)\vDash\rho_{n,\beta}(x_1,\ldots,x_n)$ for any $x_1,\ldots,x_n\in V_\gamma$. Then once we extract the set $(Y,<)$ of indiscernibles, for all $n,\beta<\omega$ and length-$n$ sequences $x_1<\ldots<x_n$, $x'_1<\ldots<x'_n$ from $Y$, $(V_\gamma,\in)\vDash\rho_{n,\beta}(x_1,\ldots,x_n)$ iff $(V_\gamma,\in)\vDash\rho_{n,\beta}(x'_1,\ldots,x'_n)$ as usual, so $Y$ is homogeneous for $f$.

Definable versions of partition properties have been considered before, e.g. in Bagaria and Bosch's "Generic absoluteness under projective forcing" (Fundamenta Mathematicae 194, pp.95--120, 2007) the property $\kappa\overset{\Sigma_\omega}{\rightarrow}(\kappa)^2$ is considered. Qi Feng's "A hierarchy of Ramsey cardinals" (Annals of Pure and Applied Logic vol. 49, 1990) defines $\Pi_\alpha$-Ramsey cardinals for ordinal $\alpha$, but even for $\alpha\in\omega$ I am not aware of any correspondence to a principle for $\Pi_\alpha$-definable partitions.

$\endgroup$

1 Answer 1

4
$\begingroup$

The consistency strength is above that of $n$-iterable cardinals for finite $n$. Thus, despite the seeming weakness of the statement, the $ω$-Erdős upper bound given by Reinhardt is fairly close to the true consistency strength.

On one hand, $α_i$ can consistently be countable even for arbitrarily large $γ$. For example, given an $ω$-Erdős $δ$, and $γ > δ$, pick indiscernibles $α_i < δ$ for $\text{Theory}(V_γ, ∈, δ)$, and collapse $δ$ to $ω$. Also, if we allow countable ordinals to be encoded by reals (with the theory only having access to the equivalence class for each $α_i$), then under large cardinal axioms ($ω$ Woodin cardinals suffice), $γ=ω+1$ works in $V$: Use the above construction, plus $Σ^1_{ω+1}$ generic absoluteness for posets below the Woodin cardinals. For $γ>ω+1$, the theory of $V_γ$ can be changed by forcing, but countable $α_i$ still exist if we have definable determinacy for games on integers of countable length (not sure if length $ω$ suffices).

On the other hand, the minimum $α_0$ (given $γ$) is $n$-iterable for finite $n$ (and thus quite large) if $V_γ$ has a parameter-free definable (in $V_γ$) well-ordering $≺$. Such $≺$ exists for $L$ (and $V_γ^L$), hence the consistency strength in the question.

Let $M$ be the Skolem hull of $(V_γ, ∈, ≺, α_0, α_1, ...)$, with $≺$ being a well-ordering of $V_γ$, $α_0 < α_1 < ...$ ordinals, and $\text{Theory}(V_γ, ∈, ≺, α_0, α_1, ...) = \text{Theory}(V_γ, ∈, ≺, α_1, α_2, ...)$. Then, $(M, ∈, ≺)$ is nontrivially elementarily self-embeddable with $j(α_i)=α_{i+1}$ (this uniquely determines $j$). Thus, $\operatorname{crit}(j)$ is $n$-iterable for every finite $n$. The implication (i.e. $n$-iterability) holds even though $j$ is not definable in $M$. See my paper Elementarily self-embeddable models of ZFC for many relations between external existence of nontrivial elementary self-embeddings for a model, and large cardinal properties internal to the model.

An interesting variation on the question (that I have not fully worked out) is to use a single predicate $\{α_i : i ∈ ω\}$ instead of infinitely many $α_i$ symbols, with two possibilities depending on whether $α_i$ is cofinal in $γ$.

$\endgroup$

You must log in to answer this question.

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