%! TEX root = TA.tex % vim: tw=50 % 29/02/2024 12PM \subsubsection*{* Non-examinable material} The similarity between \nameref{baire_category} on zero measure is immediately apparent. However they give very different kinds of genericity. ** This is the end of the non-examinable comments. Now let us prove the existence of subsets of $[0, 1]$ which are closed, have empty interior and have no isolated points. (Using standard Euclidean metric). We prove that for Hausdorff metric \[ d(E, F) = \sup_{e \in E} \inf_{f \in F} |e - f| + \sup_{f \in F} \inf_{e \in E} |e - f| ,\] except for a collection of category 1, all sets have this property. Sufficient to prove: \begin{enumerate}[(1)] \item The collection of non-empty compact sets with no isolated points is category $1$. \item The collection of non-empty compact sets with empty interior is of category $1$. \end{enumerate} To show collection with no isolated points is of first category, consider \[ \mathcal{E}_n = \{\text{$E$ compact} : \exists x \in E, E \cap B(x, 1 / n) = \{x\}\} .\] Then: \begin{enumerate}[(1)] \item $\mathcal{E}_n$ is closed in Hausdorff metric: Suppose $E_j$ such that there exists $x_j \in E$ with $B(x, 1/n) = \{x_j\}$. Then there exists $j(m) \to \infty$ such that $x_{j(m)} \to x^*$. By extracting to a subsequence, may suppose $x_j \to x^*$. If $\eta > 0$ and $n$ large enough, $|x_j - x^*| < \eta$ for all $j \le n$, and if $m \ge n$ then $|x^* - y| \ge \frac{1}{n} - 2\eta$ for all $y \in E_j$, $y \neq x_j$. Thus $|x^* - y| \ge \frac{1}{n} - 2\eta$ for all $y \in E$, $y \neq x_n$. But $\eta$ arbitrary, so $|x^* - y| \ge \frac{1}{n} - 2 \eta$ for all $\eta > 0$, $y \neq x$. So $E \in \mathcal{E}$. On the other hand if $E \neq \emptyset$ is closed then writing \[ F_m = \left\{ \frac{r}{m} : d \left( \frac{r}{m}, E \right) \le \frac{1}{m} \right\} \] we know $d(F_n, E) \le \frac{2}{n}$ and provided $n \ge 4\delta^{-1} + 1$, then there does not exist $y \in F_n$ such that $B(y, \delta) \cap E = \{y\}$. \end{enumerate} So (1) is true. To prove (2), let \[ \mathcal{E}_{r, n} = \left\{\text{$E$ closed such that $E \supseteq \left[ \frac{r}{n}, \frac{r + 1}{n} \right]$}\right\} .\] Then $\mathcal{E}_{r, n}$ is closed: $E_m \supseteq \left[ \frac{r}{n}, \frac{r + 1}{n} \right]$, $E_m \dto E$, $E \supseteq \left[ \frac{r}{n}, \frac{r + 1}{n} \right]$ but as before, if $E$ closed and non-empty, and $\delta > 0$, then there exists $F$ finite such that $d(E, F) < \delta$ and $F \subseteq \mathcal{E}_{r, n}$. If $E \notin \bigcup_{n = 1}^\infty \bigcup_{r = 0}^{n - 1} \mathcal{E}_{r, n}$ then $E$ contains non non-trivial interval. Finally we give a positive use of \nameref{baire_category}.