Cantor-Bendixson theorem and AC

analysisaxiom-of-choicegeneral-topology

For context, the Cantor-Bendixson theorem states that a closed subset $A$ of a Polish space can be written as the union of a perfect subset and a countable set $A=P\cup C$.

Now, I know two proofs of this theorem: one of them uses condensation points, and the other uses Cantor-Bendixson rank; but in the both cases the proof has a step that says "$C$ is a countable union of countable sets, and therefore countable…", which famously relies on AC$_\omega(^\omega\omega)$. But is this really necessary? If this is the case, one reasonable attempt is to construct a Polish space from a countable family of countable sets, but I don't see immediately how that can be carried out.

Best Answer

This is a ZF theorem, and is in fact equivalent to the reverse mathematical system $\Pi^1_1-\mathrm{CA}_0$ over the base theory $\mathrm{RCA}_0$ (this is in Simpson's reverse math textbook). It's worth reading Wikipedia's article "Reverse mathematics" — it has a great summary of such facts.

The Cantor-Bendixson rank argument is essentially already a ZF proof. Define a surjective partial map $f: \omega \cong \mathbb{Q}^2 \rightharpoonup C$ by sending $(q_1, q_2)\mapsto p$ if $p$ is the unique highest rank point in $C \cap (q_1, q_2).$