\classheader{2012-12-03} \begin{theorem}[Gordon-Luecke, 1980's] Let $K_1,K_2$ be knots in $\R^3$. Then $K_1$ is equivalent to $K_2$, i.e. there is a homeomorphism $h:\R^3\to\R^3$ with $h(K_1)=K_2$, if and only if $\pi_1(\R^3\setminus K_1)\cong\pi_1(\R^3\setminus K_2)$. \end{theorem} (warning: this is an incorrect statement of the theorem) \subsection*{Comparing $H_1$ and $\pi_1$} Let $\Gamma$ be any group. The commutator subgroup of $\Gamma$ is defined to be \[[\Gamma,\Gamma]=\langle \{ghg^{-1}h^{-1}\mid g,h\in\Gamma\}\rangle.\] Then $\Gamma/[\Gamma,\Gamma]$ is abelian, and $\Gamma/[\Gamma,\Gamma]\cong \Gamma^{\ab}$, where $\Gamma^{\ab}$ is by definition the unique group satisfying the following universal mapping property: for any abelian group $A$ and homomorphism $\phi:\Gamma\to A$, there is a unique $\overline{\phi}:\Gamma^{\ab}\to A$ such that \begin{center} \begin{tikzcd} \Gamma\ar{r}{\phi} \ar{d}[swap]{\pi} & A\\ \Gamma^{\ab} \ar[dashed]{ru}[swap]{\overline{\phi}} \end{tikzcd} \end{center} You proved the following theorem on the most recent homework: \begin{theorem} Let $X$ be any path-connected space. Then the map $\phi:\pi_1(X)\to H_1(X,\Z)$ defined by $\phi([\gamma])=\gamma_*([\S^1])$ induces an isomorphism \[\overline{\phi}:\pi_1(X)/[\pi_1(X),\pi_1(X)]\to H_1(X,\Z).\] \end{theorem} \subsection*{Covering Spaces} Throughout, let $X$ and $Y$ be path-connected and locally path-connected spaces. \begin{definition} A map $p:Y\to X$ is a covering space when any $x\in X$ has a neighborhood $U_x$ such that, for each $x_\alpha\in p^{-1}(x)$, there is a neighborhood $V_\alpha$ of $x_\alpha$ such that $p|_{V_\alpha}:V_\alpha\to U_x$ is a homeomorphism, and $V_\alpha\cap V_\beta$ for $\alpha\neq \beta$. \end{definition} \begin{example} A group action on $X$ is just a subgroup $\Gamma\subset \text{Homeo}(X)$. The action is discrete if, for all $x\in X$, there is a neighborhood $U_x$ of $x$ such that for all $g\in \Gamma\setminus\{\id\}$, $gU_x\cap U_x=\varnothing$. You can check that if $\Gamma$ acts discretely on $X$, then the quotient $X\to X/\Gamma$ is a covering map. To give explicit examples of this, we can take $\Gamma\subset\text{Homeo}(\R^2)$ to be \[\Gamma=\langle (x,y)\mapsto (x+1,y),\;(x,y)\mapsto(x,y+1)\rangle,\] in which case we get a covering map $\R^2\to \R^2/\Gamma\cong\T^2$. We could also consider \[\Lambda=\langle (x,y)\mapsto (x,y+1)\rangle,\] in which case we get a covering map $\R^2\to\R^2/\Gamma\cong \S^1\times\R$. Quotienting the rest of the way produces a covering map $\S^1\times \R\to\T^2$. In fact, $\T^2$ covers itself; for any $a,b\in\Z^2$, the map which acts as a degree $a$ map on the first factor of $\T^2=\S^1\times\S^1$ and as a degree $b$ map on the second factor produces a covering map $\T^2\to\T^2$. \end{example} Note that there are lots of different covering maps of $\T^2$, and some of them cover others: \begin{center} \qquad\qquad\begin{tikzpicture} \begin{scope}[scale=0.25] \draw[help lines] (-4,-4) grid (4,4) node[label={[black]right:$\R^2$}] {}; \draw[thick] (-4,0) -- (4,0); \draw[thick] (0,-4) -- (0,4); \end{scope} \draw[thick,->,>=angle 90] (0,-1.5) -- (0,-3); \begin{scope}[scale=0.9,shift={(0,-4.8)}] \node at (1.3,1.1) {$\T^2$}; \draw [,fill=gray!20] (0,0) ellipse (1.2 and 1); \begin{scope} \clip (0,0) ellipse (0.6 and 0.4); \draw[,fill=white] (0,-0.1) ellipse (0.52 and 0.42); \end{scope} \begin{scope} \clip (-1,0) rectangle (1,-1); \draw[,postaction={decorate,decoration={markings,mark=at position 0.65 with {\node (a) {};},mark=at position 0.7 with {\node (b) {};}}}] (0,0) ellipse (0.6 and 0.4); \end{scope} \draw[,postaction={decorate,decoration={markings,mark=at position 0.65 with {\node (c) {};},mark=at position 0.7 with {\node (d) {};}}}] (0,0) ellipse (1.2 and 1); \end{scope} \begin{scope}[scale=0.9,shift={(4.5,-3.8)}] \node at (1.3,1.1) {$\T^2$}; \draw [,fill=gray!20] (0,0) ellipse (1.2 and 1); \begin{scope} \clip (0,0) ellipse (0.6 and 0.4); \draw[,fill=white] (0,-0.1) ellipse (0.52 and 0.42); \end{scope} \begin{scope} \clip (-1,0) rectangle (1,-1); \draw[,postaction={decorate,decoration={markings,mark=at position 0.65 with {\node (a) {};},mark=at position 0.7 with {\node (b) {};}}}] (0,0) ellipse (0.6 and 0.4); \end{scope} \draw[,postaction={decorate,decoration={markings,mark=at position 0.65 with {\node (c) {};},mark=at position 0.7 with {\node (d) {};}}}] (0,0) ellipse (1.2 and 1); \end{scope} \draw[thick,-angle 90] (2.8,-3.8) -- (1.3,-4.1); \begin{scope}[shift={(-0.25,-0.2)}] \begin{scope}[rotate around={-90:(4.2,0)}] \node (u) at (3.9,0.8) {}; \node (w) at (3.9,-0.5) {}; \node (p) at (4.5,0.8) {}; \node (q) at (4.5,-0.5) {}; \fill[gray!20] (p.center) arc (0:180:0.3 and 0.1) -- (w.center) -- (q.center) -- cycle; \draw (p.center) -- (q.center); \draw (u.center) -- (w.center); \draw[dashed] (p.center) arc (0:180:0.3 and 0.1) node[label=above right:{$\S^1\times[0,1]$}] {}; \draw[shade,dashed,top color=gray!20] (4.2,-0.5) ellipse (0.3 and 0.1); \end{scope} \end{scope} \draw[thick,-angle 90] (1.3,0.1) -- (3,-0.1); \draw[thick,-angle 90] (4.1,-0.8) -- (4.1,-2.2); \end{tikzpicture} \end{center} We can build a dictionary between the group theory of $\pi_1(X)$ and the covering space theory of $X$. It is a Galois correspondence. Technically, we need to assume that $X$ is path-connected, locally path-connected, and semi-locally simply connected. \begin{center} \begin{tabular}{c@{$\;\;\longleftrightarrow\;\;$}c} $\pi_1(X)$ & $X$\\\hline subgroups $H\subseteq\pi_1(X)$ & covering spaces $p:Y\to X$\\ $\pi_1(X)$ & $\id:X\to X$\\ $\{e\}$ & universal cover $\widetilde{X}\to X$\\ $H\subset \pi_1(X)$ & $\widetilde{X}/H\to X$\\ normal subgroups & $p:Y\to X$ is regular\\ $[\pi_1(X):H]$ & $\#p^{-1}(x)$, i.e. number of sheets\\ conjugacy classes of elements & free homotopy classes of loops \end{tabular} \end{center} \begin{theorem} If $\Gamma$ acts discretely on $Y$ and $\pi_1(Y)=0$, then $\pi_1(Y/\Gamma)\cong \Gamma$. \end{theorem} \begin{example} We can cover the unit disk model of $\H^2$ with octagons, and act discretely on it by $\Gamma=\langle x,y,z,w\rangle$ for some transformations $x,y,z,w$ which identify the appropriate sides of the octagons, and we get $\H^2/\Gamma\cong\Sigma_2$. \end{example} \begin{example} If $L(p,q)$ is the lens space for $p$ and $q$, we can obtain it as $\S^3/(\Z/p\Z)$. \end{example} \begin{non-example} Let $r\in\R\setminus\Q$, and let $\Z$ act on $\S^1$ by $e^{2\pi i\theta}\mapsto e^{2\pi i (\theta+r)}$. This action is not discrete, and $\S^1/\Z$ is not even Hausdorff. \end{non-example} \begin{non-example} Let $\Z/3\Z$ act on $\D^2$ by rotations by $2\pi/3$. This action is not discrete, though it almost is - it would be discrete if we threw away the origin. This kind of situation is called a branched cover. \end{non-example} \begin{proposition} Let $p:Y\to X$ be a covering map. Then $p_*:\pi_1(Y,y)\to \pi_1(X,p(y))$ is injective. \end{proposition} \begin{proof} The idea is clear: if a loop $p_*([\gamma])$ is trivial in $\pi_1(X,p(y))$, we can lift the homotopy to one demonstrating that $[\gamma]$ is trivial. \end{proof}