Problem below (kuratowski) is taken from Kechris's Classical Descriptive Set Theory, page $173.$

Definition: Given any sequence of sets $(A_n),$ $A_n\subseteq X,$ let $$\overline{\lim_n} \, A_n = \bigcap_n\bigcup_{m\geq n} A_m = \{x:x\text{ belongs to infinitely many } A_n\},$$ $$\underline{\lim}_n \, A_n = \bigcup_n\bigcap_{m\geq n} A_m = \{x:x\text{ belongs to all but finitely many } A_n\},$$ It is clear that $\underline{\lim}_n \, A_n \subseteq\overline{\lim_n} \, A_n.$ If they are equal, let $$\lim_n A_n= \overline{\lim_n} \, A_n = \underline{\lim}_n \, A_n .$$

Assume now that $X$ is metrizable, so that every closed set is a $G_\delta$ set. Let $\omega_1$ be the first uncountable ordinal, and for $1\leq \xi<\omega_1,$ define by transfinite recursion the classes $\Sigma_\xi^0,\Pi_\xi^0$ of subsets of $X$ as follows:

$$\Sigma_1^0=\{U\subseteq X: U \text{ is open}\},$$ $$\Pi_\xi^0 =\sim\Sigma_\xi^0,$$ $$\Sigma_\xi^0=\bigg\{ \bigcup_n A_n: A_n\in \Pi_{\xi_n}^0, \xi_n<\xi, n\in\mathbb{N} \bigg\}, \text{ if } \xi>1,$$ where $\sim$ refers to set complement. In addition, let $$\Delta_\xi^0 = \Sigma_\xi^0\cap \Pi_\xi^0.$$

For example, $\Sigma_1^0$ is a topology on $X$ where $\Pi_1^0$ is the collection of closed sets of $X.$ $\Sigma_2^0$ is the collection of all $F_\sigma$ sets while $\Pi_2^0$ is the collection of all $G_\delta$ sets. Note that $\Delta_1^0$ is the collection of all closed and open sets while $\Delta_2^0$ is the collection of all $F_\sigma$ and $G_\delta$ sets.

Question (Kuratowski) : Show that for $\xi>1,$ $$A \text{ is } \Delta_{\xi+1}^0 \iff A=\lim_nA_n\text{ for some sequence } (A_n) \text{ with } A_n\in\Delta_\xi^0.$$

For $(\Leftarrow)$ direction, by definition, $$A = \bigcap_n\bigcup_{m\geq n} A_m = \bigcup_n\bigcap_{m\geq n} A_m.$$ If $A_m\in\Delta_\xi^0 = \Sigma_\xi^0\cap \Pi_\xi^0,$ then $$\bigcup_{m\geq n}A_m\in \Sigma_{\xi}^0\quad \text{and}\quad \bigcap_{m\geq n}A_m\in\Pi_\xi^0.$$ Then by definition of $\Sigma_{\xi+1}^0$ and $\Pi_{\xi+1}^0,$ we have $$ \bigcap_n\bigcup_{m\geq n} A_m\in \Pi_{\xi+1}^0\quad\text{and}\quad \bigcup_n\bigcap_{m\geq n} A_m\in\Sigma_{\xi+1}^0.$$

I have no idea on how to start $(\Rightarrow)$ direction.

  • $\begingroup$ You know, the lim inf and lim sup are very common and pretty easy to recognize, even when you are using a less common symbol for them. But the $\Delta, \sum, \prod$ notations you are using are far less common. I suggest explaining them instead of lim inf and lim sup. $\endgroup$ – Paul Sinclair May 19 '18 at 16:34
  • $\begingroup$ @PaulSinclair: Thanks for mentioning the unusual symbols. I have included their definitions in my question. $\endgroup$ – Idonknow May 19 '18 at 16:43
  • $\begingroup$ @Paul Sinclair Those notations are pretty common as well in descriptive set theory and in logic. $\endgroup$ – Régis Jun 6 '18 at 21:06
  • $\begingroup$ @Régis Do you know how to answer my questions? $\endgroup$ – Idonknow Jun 8 '18 at 6:26
  • $\begingroup$ I used to but I forgot. $\endgroup$ – Régis Jun 8 '18 at 8:03

Note: This only holds if $\xi$ is successor ordinal.

To prove ($\Leftarrow$), first since $A \in \Delta_{\xi+1}^0, A\in\Sigma_{\xi+1}^0$ and $A\in\Pi_{\xi+1}^0$. So $$ A=\bigcup_nB_n=\bigcap_nC_n, \quad B_n\in \Delta_{\xi}^0,\: C_n\in\Delta_{\xi}^0 $$ Where $B_n\subset B_{n+1}\subset \cdots\subset A\subset \cdots\subset C_{n+1}\subset C_{n}$. Thus there is $A_n\in \Delta_{\xi}^0$ that $B_n\subset A_n\subset C_n$. Clearly $$ \overline{\lim_n} \, A_n = \bigcap_n\bigcup_{m\geq n} A_m \subset \bigcap_n\bigcup_{m\geq n} C_m =\bigcap_n C_n=A $$ And $$ \underline{\lim}_n \, A_n = \bigcup_n\bigcap_{m\geq n} A_m \supset \bigcup_n\bigcap_{m\geq n} B_m=\bigcup_n B_n =A $$ Thus $A=\lim_{n}A_n$.

| cite | improve this answer | |

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.