%! TEX root = AlgT.tex % vim: tw=50 % 05/02/2024 11AM \phantom{}\\[-2\baselineskip] We now check that $p$ is continuous. It is enough to check that $p^{-1}(U)$ is open for $U \in \mathcal{U}$. If $[\alpha] \in p^{-1}(U)$, then $[\alpha] \in ([\alpha], U) \subset p^{-1}(U)$, so $p^{-1}(U)$ is open. To see that $p$ is a \gls{covmap}, we first show that each \[ p|_{([\alpha], U)} : ([\alpha], U) \to U \] is a homeomorphism. As $U$ is \gls{pathconn}, for any $y \in U$ there is a path $\gamma$ in $U$ from $\alpha(1)$ to $y$, so $p([\alpha \concat \gamma]) = y$, so the map is surjective. If $[\beta], [\beta'] \in ([\alpha], U)$ maps to the same thing under $p|_{([\alpha], U)}$, then $\beta$ and $\beta'$ end at the same point. So there exist \glspl{path} $\gamma$, $\gamma'$ in $U$ such that $[\beta] = [\alpha \concat \gamma]$, $[\beta'] = [\alpha \concat \gamma']$. \begin{center} \includegraphics[width=0.6\linewidth]{images/4581ba67e7c84e3c.png} \end{center} So \[ [\beta'] = [\alpha \concat \gamma \concat \invpath\gamma \concat \gamma'] = [\alpha \concat \gamma] = [\beta] \] where the first equality comes from the fact that $\invpath\gamma \concat \gamma'$ is a loop in $U$, so \gls{homotopic} to a constant loop in $X$. So $p|_{([\alpha], U)}$ is injective. So $p|_{([\alpha], U)}$ is a bijection, and continuous. It is also open as \[ p(([\beta], V)) = V ,\] so $p|_{([\alpha], U)}$ is a homeomorphism. We now claim that $p^{-1}(U)$ is partitioned into $([\alpha], U)$s. When checking that $p$ is continuous, we saw that they cover, so need to show that if they intersect then they are equal. So let $[\gamma] \in ([\alpha], U) \cap ([\beta], U)$, i.e. there are \glspl{path} $\alpha'$, $\beta'$ in $U$ such that $[\gamma] = [\alpha \concat \alpha'] = [\beta \concat \beta']$. Let $[\delta] \in ([\alpha], U)$ So \[ [\delta] = [\alpha \concat \alpha''] = [\alpha \concat \alpha' \concat \invpath{(\alpha')} \concat \alpha''] = [\beta \concat \ub{\beta' \concat \invpath{(\alpha')} \concat \alpha''}_{\subset U}] \] so $[\delta] \in ([\beta], U)$ \begin{center} \includegraphics[width=0.6\linewidth]{images/a1ff9e405d9b4003.png} \end{center} i.e. $([\alpha], U) \subseteq ([\beta], U)$, similarly for the opposite containment. Finally, we need to show that $\tilde{X}$ is \gls{simpconn}. Note: if $\gamma : I \to X$ is a \gls{path}, then its \gls{lift} $\tilde{\gamma} : I \to \tilde{X}$ startng at $[\constpath{x_0}]$ ends at $[\gamma]$, because \[ s \mapsto [t \mapsto \gamma(st)] : I \to \tilde{X} \] is the \gls{lift}. So if a loop $\gamma$ in $X$ \glspl{lift} to a loop in $\tilde{X}$ based at $[\constpath{x_0}]$, so \[ [\gamma] = [\constpath{x_0}] \implies \ps p \pio\bs(\tilde{X}, [\constpath{x_0}]) = \{e\} \subseteq \pio\bs(X, x_0) .\] But $\ps p$ is injective, so \[ \pio\bs(\tilde{X}, [\constpath{x_0}]) = \{e\} \qedhere \] \end{proof} \subsection{The Galois Correspondence} If $p : \tilde{X} \to X$ is a \gls{covmap}, $\tilde{X}$ \gls{pathconn}, $x_0 \in X$ and $\tilde{x}_0 \in p^{-1}(x_0)$, then \[ \ps p : \pio\bs(\tilde{X}, \tilde{x}_0) \to \pio\bs(X, x_0) \] is injective, giving a subgroup of $\pio\bs(X, x_0)$. If $\tilde{x}_0' \in p^{-1}(x_0)$ is another basepoint, let $\gamma$ be a \gls{path} in $\tilde{X}$ from $\tilde{x}_0$ to $\tilde{x}_0'$. Then $p \circ \gamma$ is a loop based at $x_0$, and we have \[ [p \circ \gamma]^{-1} \concat \ps p \pio\bs(\tilde{X}, \tilde{x}_0) \concat [p \circ \gamma] = \ps p \pio\bs(\tilde{X}, \tilde{x}_0') \le \pio\bs(X, x_0) \] So fixing a \gls{pathconn} based space $\bs(X, x_0)$ we get \begin{align*} \left\{\substack{ \text{based \gls{pathconn} \glspl{covmap}} \\ \text{$p : \bs(\tilde{X}, \tilde{x}_0) \to \bs(X, x_0)$} }\right\} &\stackrel{p \mapsto \Im(\ps p)}{\xrightarrow{\hspace*{2cm}}} \left\{\substack{ \text{subgroups of $\pio\bs(X, x_0)$} }\right\} \\ \left\{\substack{ \text{\gls{pathconn} \gls{covmap}} \\ \text{$p : \tilde{X} \to X$} }\right\} &\stackrel{p \mapsto \Im(\ps p)}{\xrightarrow{\hspace*{2cm}}} \left\{\substack{ \text{conjugacy classes of} \\ \text{subgroups of $\pio\bs(X, x_0)$} }\right\} \end{align*} \begin{flashcard}[galois-surj-prop] \begin{proposition*}[Surjectivity of Galois correspondence] \cloze{Let $X$ be \gls{pathconn}, \gls{locpathconn}, and \gls{slsc}. Then for any $H \le \pio\bs(X, x_0)$ there is a $p : \bs(\tilde{X}, \tilde{x}_0) \to \bs(X, x_0)$ with $\ps p \pio\bs(\tilde{X}, \tilde{x}_0) = H$.} \end{proposition*} \begin{proof} Let $\ol{X} \stackrel{q}{\to} X$ be the \gls{unicov} we have constructed in \nameref{unicov_existence}. Define $\sim_H$ on $\ol{X}$ by $[\gamma] \sim_H [\gamma'] \iff \gamma(1) = \gamma'(1)$ and $[\gamma \concat \invpath{(\gamma')}] \in H \le \pio\bs(X, x_0)$. So: \begin{enumerate}[(i)] \item $[\gamma] \sim_H [\gamma]$. \item If $[\gamma] \sim_H [\gamma']$ then $[\gamma \concat \invpath{(\gamma')}] \in H$, so $[\gamma' \concat \invpath{(\gamma')}] \in H$. So $[\gamma'] \sim_H [\gamma]$. \item If $[\gamma] \sim_H [\gamma']$, $[\gamma'] \sim_H [\gamma'']$, then $\gamma(1) = \gamma'(1) = \gamma''(1)$, and \[ [\gamma \concat \invpath{(\gamma'')}] = [\gamma \concat \invpath{(\gamma')} \concat \gamma' \concat \invpath{(\gamma'')}] = \ub{[\gamma \concat \invpath{(\gamma')}]}_{\in H} \ub{[\gamma' \concat \invpath{(\gamma'')}]}_{\in H} \in H .\] So $[\gamma] \sim_H [\gamma'']$. \end{enumerate} So $\sim_H$ is an equivalence relation. Define $\ol{X}_H = \ol{X} / \sim_H$, the quotient space, and $p_H : \ol{X}_H \to X$ to be the induced map. If $[\gamma] \in ([\alpha], U)$, $[\gamma'] \in ([\beta], U)$ satisfy $[\gamma] \sim_H [\gamma']$ then $([\alpha], U)$ and $[\beta], U)$ are identified by $\sim_H$, as $[\gamma \concat \eta] \sim_H [\gamma' \concat \eta]$ for any \gls{path} $\eta$ in $U$. So $p_H$ is a \gls{covmap}. It remains to show that $\ps{(p_H)} \pio\bs(\ol{X}_H, [[\constpath{x_0}]]) = H \le \pio\bs(X, x_0)$. If $[\gamma] \in H$ then the \gls{lift} of $\gamma$ to $\ol{X}$ starting at $[\constpath{x_0}]$ ends at $[\gamma]$, so the \gls{lift} to $\ol{X}_H$ ends at $[[\gamma]] = [[\constpath{x_0}]]$, so is a loop, i.e. \[ H \subseteq \ps{(p_H)} \pio\bs(\ol{X}_H, [[\constpath{x_0}]) .\] If $[\gamma] \in \ps{(p_H)} \pio\bs(\ol{X}_H, [[\constpath{x_0}]])$ then the \gls{lift} $\ol{\gamma}$ of $\gamma$ to $\ol{X}$ starting at $[\constpath{x_0}]$ ends at $[\gamma]$, so $[\gamma] \sim_H [\constpath{x_0}]$, as it becomes a loop in $\ol{X}_H$ by assumption. So $[\gamma] \in H$. \end{proof} \end{flashcard}