Assuming $ZF$ itself is consistent, it is consistent that there are sets $D$ which are infinite but cannot be placed in bijection with any of their proper subsets; such sets are called "strictly Dedekind-finite." Consistently, there is even a Dedekind-finite set of reals.

My question is, is it consistent to be able to partition $\mathbb{R}$ into strictly Dedekind-finite sets?

The simplest way to produce strictly Dedekind-finite sets of reals is to use Cohen forcing and take a symmetric submodel of the resulting forcing extension. We can also create $\kappa$-many disjoint strictly Dedekind-finite sets in a similar fashion, for any $\kappa$, without collapsing $\kappa$ (of course, the continuum is bumped above $\kappa$). However, as far as I can see, there is no simple way to adapt this to provide a partition of $\mathbb{R}$ into such sets.

One simple thing I've been able to figure out: suppose $\kappa$ is a (well-ordered) cardinal, $\{D_i: i\in I\}$ is a partition of $\mathbb{R}$ into strictly Dedekind-finite sets, and $\kappa^+$ injects into $\mathbb{R}$. Then there cannot be an injection $I\rightarrow\kappa$. However, letting $\Psi$ be the least ordinal not injectible into $\mathbb{R}$, it is not even clear to me that we cannot partition $\mathbb{R}$ into $\Psi$-many strictly Dedekind-finite pieces; and of course this says nothing about the case when the index set $I$ itself is non-well-orderable.

  • $\begingroup$ I'm pretty sure what I'm calling "$\Psi$" actually has a name, but I can't seem to remember/find it right now. $\endgroup$ – Noah Schweber Feb 9 '14 at 4:36
  • 1
    $\begingroup$ Nice question! Your cardinal $\Psi$ is called the Hartog number $\aleph(\mathbb{R})$, see en.wikipedia.org/wiki/Hartogs_number. $\endgroup$ – Joel David Hamkins Feb 9 '14 at 5:01
  • 1
    $\begingroup$ @Joel, Hartogs, he was French. $\endgroup$ – Asaf Karagila Feb 9 '14 at 5:17
  • 1
    $\begingroup$ @Joel: Yeah, although I prefer "Hartogs' number". $\endgroup$ – Asaf Karagila Feb 9 '14 at 5:39
  • 2
    $\begingroup$ @AsafKaragila I think he was German (though born in Brusseles). $\endgroup$ – user9072 Feb 9 '14 at 10:49


Suppose that there is an infinite Dedekind-finite set of real numbers $A$ (e.g. Cohen's first model). Simple cardinal arithmetic shows that, $$|\Bbb R|\leq|\Bbb R\times A|\leq|\Bbb{R\times R}|=|\Bbb R|.$$

Clearly $\Bbb R\times A$ can be partitioned into infinite Dedekind-finite sets, simply consider $\{\{r\}\times A\mid r\in\Bbb R\}$. Now use a bijection of $\Bbb R\times A$ with $\Bbb R$ to transport this partition to a partition of $\Bbb R$ without changing the cardinality of its parts.

  • $\begingroup$ Nice! A followup question: can we partition $\mathbb{R}$ into fewer than continuum-many amorphous sets? $\endgroup$ – Noah Schweber Feb 9 '14 at 9:35
  • 1
    $\begingroup$ No, not even more than continuum many. Any map from an amorphous set to a linearly ordered set has a finite range. So we cannot partition $\Bbb R$ in a way that there is any part which is amorphous. (Well, I'm not counting finite sets, of course.) $\endgroup$ – Asaf Karagila Feb 9 '14 at 10:18
  • $\begingroup$ Sorry, I meant "strictly Dedekind-finite," not amorphous. :P $\endgroup$ – Noah Schweber Feb 9 '14 at 18:27
  • $\begingroup$ Noah, that's a much more difficult task. I'll have to think about it for a bit. $\endgroup$ – Asaf Karagila Feb 9 '14 at 21:09
  • $\begingroup$ I've turned my followup into a real question (mathoverflow.net/questions/157204/…). $\endgroup$ – Noah Schweber Feb 10 '14 at 9:19

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.