Let $A$ be some countable set, and let $A^{\omega}$ be the set of all sequences $\mathbb N \to A$ equipped with the product topology. A subset $X \subseteq A^{\omega}$ is called Suslin set if there is a countable set $B$, a Borel subset $Y \subseteq B^{\omega}$ and a continuous map $f : B^{\omega} \to A^{\omega}$ such that $X = f(Y)$, i.e. it is the continuous image of some Borel set.

Does anybody has a proof of the following result:

If $X \subseteq A^{\omega}$ is a Suslin set, then there exists a closed set $Y \subseteq \mathbb N^{\omega}$ and a continuous map $f : \mathbb N^{\omega} \to A^{\omega}$ such that $X = f(Y)$.

It is mentioned in a book, but its proof is faulty and I cannot find it anywhere else.


As the countable sets are all equipped with discrete topology, might assume they are all $\omega$. So the problem is reduced to the following: if $Y\subset \omega^\omega$ is Borel, then there exists a continuous function $f$ (in fact continuous bijection) whose domain is a closed subset $F$ of $\omega^\omega$ and $f[F]=Y$ (because then you can compose two maps). You only need to verify the following $$\mathcal{A}=\{Y\subset \omega^\omega: \exists\text{ continuous bijection } f\ \exists\text{ closed }F\subset\omega^\omega \operatorname{dom}(f)=F, f''F=Y\}$$ contains a $\sigma$-algebra containing open sets (i.e. the sub algebra that is closed under taking complements). These sets are called Lusin.

More precisely, try to show the following:

  • $\mathcal{A}$ is closed under countable intersection and countable disjoint union
  • $\mathcal{C}=\{A\in \mathcal{A}: A^c\in \mathcal{A}\}$. Show $C$ is a $\sigma$-algebra containing open sets.
  • $\begingroup$ I use “countable” for “bijective with a subset of $\omega$”. If the OP uses this notation the same way as me, you might not replace all countable sets by $\omega$. $\endgroup$ – Pedro Sánchez Terraf Aug 28 '16 at 23:32
  • $\begingroup$ In the case it's finite, the same proof works $\endgroup$ – Jing Zhang Aug 28 '16 at 23:35
  • $\begingroup$ Why you just considered continuous functions $f : F \to \omega^{\omega}$? I asked for functions $f : \omega^{\omega} \to \omega^{\omega}$ with a closed $F \subseteq \omega^{\omega}$ such that $f(F) = Y$? Could every function $f : F \to \omega^{\omega}$ extended to a function on the whole $\omega^{\omega}$? $\endgroup$ – StefanH Aug 29 '16 at 12:09
  • $\begingroup$ Yes. Each closed subset $A$ is a retract of $\omega^\omega$, that means there is a continuous function $f: \omega^\omega \to \omega^\omega$ such that $f''\omega^\omega=A$ and $f\restriction A = id_A$. However, I think the proof strategy should work if you delete $dom(f)=F$ in the definition of $\mathcal{A}$. $\endgroup$ – Jing Zhang Aug 29 '16 at 14:57
  • $\begingroup$ Okay, I tried to prove that $\mathcal A$ is closed under countable intersection and union. So suppose $\{A_i\}_{i\in I} \in \mathcal A$ is some countable collection, then I have to show that $\bigcup_{i\in I} A_i \in \mathcal A, \bigcap_{i\in I} A_i \in \mathcal A$. I know that $A_i = f(F_i)$ for $f : F_i \to \omega^{\omega}$ continuous and $F_i$ closed. Now I have to find $F, F'$ closed and $f$ and $g$ such that $f(F) = \bigcap_{i\in I} A_i, g(F') = \bigcup_{i\in I} A_i$. Is it easy to construct those function, do you mind giving a hint? $\endgroup$ – StefanH Aug 31 '16 at 12:21

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.