6
$\begingroup$

From what I understand in ZF we have that (i) implies (ii) implies (iii) implies (iv)

(i) Infinite and well-orderable

(ii) Infinite and $|A\times A|=|A|$

(iii) Dedekind-Infinite (i.e. $|A|+1=|A|$)

(iv) Infinite (i.e. not bijective with any finite ordinal)

I know that there are models where there exist infinite sets that are Dedekind-finite. I also recently learned there are Dedekind-Infinite sets that do not have the property: $|A\times A|=|A|$. Namely

$$X\sqcup \aleph(X)$$

where $X$ is not well-orderable and $\aleph(X)$ is the Hartogs' ordinal number.

TWO QUESTIONS

(1) I am pretty sure that (ii) does not imply (i), but only because of how Tarski's Theorem is proved (requiring all infinite sets to have property ii). Is this an open problem? I also know for example that assuming CH at an infinite cardinal $\mathfrak{p}$ and its power set $2^\mathfrak{p}$ implies $\mathfrak{p}$ can be well-ordered, but it is not known that if CH only at $\mathfrak{p}$ is sufficient. Since CH($\mathfrak{p}$) implies $\mathfrak{p}^2=\mathfrak{p}$, it seems to me that it is an open question.

(2) Between (iii) and (ii) there is a notion of infinity: Infinite and $|A|\times 2=|A|$. Is this notion strictly in between?

If you have a reference that answers or studies these kinds of questions, then please share it with me. Also thanks for your time and patience in writing a response!

$\endgroup$
4
$\begingroup$

Your intuition is correct that (ii) does not imply (i). There are models of ZF in which $\mathbb{R}$ is not well-orderable, but $\mathbb{R}$ always satisfies (ii) since $|\mathbb{R}\times\mathbb{R}|=2^{\aleph_0}\cdot 2^{\aleph_0}=2^{\aleph_0+\aleph_0}=2^{\aleph_0}$.

The answer to your second question is also yes. To get a set $A$ which satisfies your condition but not (ii), let $X$ be a set that is not well-orderable, let $Y=X\times\mathbb{N}$, and let $A=Y\sqcup\aleph(Y)$. Then $|A|\times 2=|A|$ since the same is true of $Y$ and $\aleph(Y)$. But if $|A\times A|=|A|$, then $Y$ is well-orderable, which implies $X$ is well-orderable since it injects into $Y$.

To get a set $A$ which satisfies (iii) but not your condition, let $X$ be an infinite Dedekind-finite set and let $A=X\sqcup\mathbb{N}$. Clearly $A$ satisfies (iii). If there were an injection $f:A\times\{0,1\}\to A$, then $f$ could map at most finitely many elements of $X\times \{0,1\}$ to $\mathbb{N}$, since otherwise you could invert $f$ on an infinite subset of $\mathbb{N}$ to get an injection from a countably infinite set to $X$. So there is a finite set $F$ such that $f$ restricts to an injection $g:X\times\{0,1\}\setminus F\to X$. Taking finitely many points of $X\times\{1\}$ to replace all the points of $F\cap X\times\{0\}$, we may assume that $g$ is actually defined on all of $X\times\{0\}$, plus at least one point of $X\times \{1\}$. This would mean $|X|+1=|X|$.

| cite | improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.