Completely Baire spaces, Menger spaces, and projective sets (Q1738925)

In what follows, we assume the reader is familiar with the usual terminology regarding Descriptive Set Theory, Constructibility and Determinacy. A topological space $X$ is said to be \textit{completely Baire} if every closed subspace of $X$ is Baire. A \textit{Menger space} is a topological space which satisfies the selection principle $S_{\text{fin}}(\mathcal{O},\mathcal{O})$, meaning that for every sequence $\{ \mathcal{U}_n: n < \omega \}$ of open covers of $X$ there is a sequence of open families $\{ \mathcal{V}_n: n < \omega\}$ such that $\mathcal{V}_n$ is a finite subset of $\mathcal{U}_n$ for all $n < \omega$ and $\bigcup\limits_{n < \omega} \bigcup \mathcal{V}_n = X$. As usual, the \textit{Axiom of Constructibility} is denoted by $\mathbf{V = L}$ and the \textit{Axiom of Projective Determinacy} is denoted by $\mathbf{PD}$. \par In the 1920's, Hurewicz has shown that analytic Menger subsets of $\mathbb{R}$ are $\sigma$-compact and co-analytic completely Baire subsets of $\mathbb{R}$ are Polish. So, it is natural to ask whether all ``nicely definable''\, Menger subsets of $\mathbb{R}$ are $\sigma$-compact (``Problem 1'') and whether all ``nicely definable''\, completely Baire subsets of $\mathbb{R}$ are Polish (``Problem 2''). \textit{A. W. Miller} and \textit{D. H. Fremlin} have presented, regarding Problem 1, co-analytic counterexamples under $\mathbf{V = L}$ and positive results for projective sets under $\mathbf{PD}$ (see pages 20 and 21 of [Fundam. Math. 129, No. 1, 17--33 (1988; Zbl 0665.54026)]), and it is also known that there are positive results for Problem 2 assuming some appropriate amount of determinacy (see 28.20 of \textit{A. S. Kechris} [Classical descriptive set theory. Berlin: Springer-Verlag (1995; Zbl 0819.04002)]). \textit{A. Medini} and the second author have shown in [Arch. Math. Logic 54, No. 1--2, 231--245 (2015; Zbl 1321.54052)] that there are analytic counterexamples for Problem 2 under $\mathbf{V = L}$. In a paper which is still under preparation, the first author together with Todorčević and Tokgöz has considerably sharpened (with respect to co-analytic sets) the hypotheses concerning Problem 1; in the paper under review, the authors prove results (with respect to analytic sets) regarding Problem 2 which are entirely analogous to those (of Tall, Todorčević and Tokgöz) we have just mentioned, and show that the two problems are essentially equivalent. More precisely, the authors prove that: $(i)$ $\omega_1^{L(a)} < \omega_1$ for all reals $a$ if, and only if, every analytic, completely Baire set of reals is Polish; and $(ii)$ the assertion that every analytic completely Baire set of reals is Polish is equiconsistent with the existence of an inaccessible cardinal. \par The somehow surprising connection between Problems 1 and 2 is given by the following theorem, which is established in the paper under review: suppose $X$ is a Menger set of reals. Then $\mathbb{R} \setminus X$ is completely Baire.