---- > [!theorem] Theorem. ([[constructing universal covers]]) > Let $(X, x_{0})$ be a [[pointed set|based]] [[topological space]]. Does $X$ admit a [[universal cover]]? > *A necessary condition.* If $X$ admits a [[universal cover]] $p: \widetilde{X} \to X$, then it is [[semilocally simply connected]]. Also, there is a [[bijection]] $ \widetilde{X} \leftrightarrow \left\{\begin{align}&\text{path homotopy classes of paths in }X \\ &\text{starting at } x_{0} \text{ and ending anywhere} \end{align} \right\}.$ > *A (more stringent) sufficient condition.* If $X$ is [[path-connected]] and [[locally connected, locally path-connected|locally path-connected]] in addition to being [[semilocally simply connected]], then it admits a [[universal cover]] $\widetilde{X}$ *defined* as the 'path space' $\widetilde{X}:= \left\{\begin{align}&\text{path homotopy classes of paths in }X \\ &\text{starting at } x_{0} \text{ and ending anywhere} \end{align} \right\}$ when this set is endowed with a suitable [[topological space|topology]], with the [[covering space|covering map]] $p: \widetilde{X} \to X$ given by $p([\gamma]):=\gamma(1)$. ^theorem > [!proof]+ Proof. ([[constructing universal covers]]) > *A necessary condition.* Let $U \ni x_{0}$ be a [[neighborhood]] of $x_{0}$ [[evenly covered]] by $p$, so $p ^{-1}(U)= \coprod_{\beta}V_{\beta}$. Fix some $\beta$. Then for any [[parameterized curve|loop]] $\gamma:I \to \textcolor{Skyblue}{U}$ based at $x_{0}$ there is a [[the homotopy lifting lemma|unique lift to a loop]] $\tilde{\gamma}$ *in $V_{\beta}$* starting at $\tilde{x}_{0}=V_{\beta} \cap p ^{-1}(x_{0})$ [^1]. Thus the [[homomorphism of fundamental groups induced by a continuous map|pushforward]] [[group homomorphism|homomorphism]] of [[fundamental group|fundamental groups]]$ p_{*}: \pi_{1}(\widetilde{X}, x_{0}) \to \pi_{1}(X,x_{0})$ maps $[\tilde{\gamma}]$ to $[p \circ \tilde{\gamma}]=[\gamma]$. But since $\widetilde{X}$ is [[simply connected]], $\tilde{\gamma}$ is [[homotopy|nulhomotopic]]. So $\gamma$ is too— hence $X$ is [[semilocally simply connected]]. > Next for deriving the [[bijection]]: since $\widetilde{X}$ is [[simply connected]], for any point $y \in \widetilde{X}$ there is [[in a simply connected space, any two paths having the same initial and final points are path homotopic|exactly one homotopy class]] $[\alpha]$ of [[parameterized curve]] from $\tilde{x}_{0}$ to $y$, and $[p \circ \alpha]$ gives a preferred [[homotopy|homotopy class]] of [[parameterized curve]] from $x_{0}$ to $p(y)$. Thus we can recover $y$ as the endpoint of the [[lifting|lift]] of $[p \circ \alpha]$ starting at $\tilde{x}_{0}$. This defines a [[bijection]] $ \widetilde{X} \leftrightarrow \left\{\begin{align}&\text{homotopy classes of paths in }X \\ &\text{starting at } x_{0} \text{ and ending anywhere} \end{align} \right\}.$ ![[CleanShot 2024-06-10 at [email protected]|500]] > > *A sufficient condition.* Next we wish to *define* a [[topological space]] $\widetilde{X}$ as $\widetilde{X}:= \left\{\begin{align}&\text{homotopy classes of paths in }X \\ &\text{starting at } x_{0} \text{ and ending anywhere} \end{align} \right\}$ and [[covering space|covering map]] $p: \widetilde{X} \to X$ given by $p([\gamma]):=\gamma(1)$. We must (a) construct a [[topological space|topology]] on $\widetilde{X}$, (b) show $p$ is [[continuous]] in this [[topological space|topology]], (c) show that $p$ is a [[covering space|covering map]] in this [[topological space|topology]], and finally (d) show that $\widetilde{X}$ is [[simply connected]]. > **(a).** Consider the collection of subsets [^3] $\mathscr{U}:= \left\{ U \subset X : \begin{align} & U\text{ is open in } X, \text{ is path-connected, and} \\ &\pi_{1}(U,x) \xrightarrow{\iota_{*}} \pi_{1}(X,x) \text{ is trivial for all } x \in U \end{align} \right\}.$ Here, $\pi_{1}(U,x)$ and $\pi_{1}(X,x)$ are the [[fundamental group|fundamental groups]] of $U$ and $X$ based at $x$ respectively, and $\iota_{*}$ denotes the [[inclusion map|inclusion map's]] [[homomorphism of fundamental groups induced by a continuous map|pushforward homomorphism]]. > *Claim:* $\mathscr{U}$ is a [[basis for a topology|basis]] [[topology generated by a basis|generating]] the [[topological space|topology]] on $X$. To show this, we'll invoke: [[condition for obtaining a basis from a topology]]. Let $V$ be an open set in $X$ containing $x \in X$. > >1. As $X$ is [[semilocally simply connected]], for any $x \in X$ there is a [[neighborhood]] [^2] $U' \ni x$ for which $\pi_{1}(U', x) \xrightarrow{\iota'_{*}}\pi_{1}(X,x)$ is [[group homomorphism|trivial]]. >2. Because $X$ is [[locally connected, locally path-connected|locally path-connected]], there is a neighborhood $x \in U \subset V \cap U'$ which is [[path-connected]]. >3. Moreover, $\pi_{1}(U,x) \xrightarrow{\iota_{*}} \pi_{1}(X,x)$ is [[group homomorphism|trivial]] because it factors through the trivial map $\pi_{1}(U', x) \xrightarrow{\iota'_{*}}\pi_{1}(X,x)$. >4. It remains to be shown that the inclusion pushforward is trivial *for all* elements of $U$, and not just $x$. Take $y \in U$ and a [[parameterized curve]] $u:I \to U$ from $x$ to $y$. Then we have a commutative square > ```tikz > \usepackage{tikz-cd} > \usepackage{amsmath} > \begin{document} > % https://tikzcd.yichuanshen.de/#N4Igdg9gJgpgziAXAbVABwnAlgFyxMJZABgBpiBdUkANwEMAbAVxiRAB120sB9ARgAUAVVIBPAJQgAvqXSZc+Qij7kqtRizadu-AQA0xkmXOx4CRFXzX1mrRBy69BBgB5HZIDKcVEyV6jaa9tpOwqRu0mowUADm8ESgAGYAThAAtkhkIDgQSADMARp2IEw8nADE0h4p6Ugq2bmIAEyFtmylFVVJqRmIWTl1xiA1vQUNSC3qbcHsODAuOMA4yVg0WIxSkVJAA > \begin{tikzcd} > {\pi_1(U,y)} \arrow[d, "u_\#"] \arrow[r] & {\pi_1(X,y)} \arrow[d, "u_\#"] \\ > {\pi_1(U,x)} \arrow[r, "\text{trivial}"] & {\pi_1(X,x)} > \end{tikzcd} > \end{document} > ``` > where the vertical maps are the [[fundamental group]] [[group isomorphism|isomorphisms]] outlined in [[change of basepoint isomorphism|this result]], and the bottom horizontal map (inclusion pushforward for $(U,x)$ to $(X,x)$) is [[group homomorphism|trivial]], hence the top horizontal map (inclusion pushforward for $(U,y)$ to $(X,y)$ must be trivial also. This completes the proof of *claim*. > $\mathscr{U}$ is useful because, if $x,y \in U \subset \mathscr{U}$ then there exists [[parameterized curve]] from $x$ to $y$ in $U$, and all such [[parameterized curve|paths]] are [[path homotopy|path homotopic]] *in $X$*. [^4] Now, for a given $[\alpha] \in \widetilde{X}$ and $U \in \mathscr{U}$ such that $\alpha(1) \in U$, define $(\alpha, U):= \left\{ [\beta] \in \widetilde{X} : \begin{align} & [\beta]=[\alpha * \alpha'] \text{ for some path } \\ & \alpha' \text{ in } U \text{ starting at } \alpha(1) \end{align} \right\}.$ ![[CleanShot 2024-06-10 at [email protected] |500]] > We claim that such sets form a [[basis for a topology|basis]] for a [[topological space|topology]] on $\widetilde{X}$. >1. (Covering) This obviously holds because we define one set per element of $\widetilde{X}$ >2. (Nestling) We must show $[\beta] \in (\alpha_{0}, U_{0}) \cap (\alpha_{1}, U_{1})$ has a [[neighborhood]] of the required form. Let $W \subset U_{0} \cap U_{1}$ be a [[neighborhood]] of $\beta(1) \in U_{0} \cap U_{1}$ in the collection $\mathscr{U}$, which exists because $\mathscr{U}$ is a [[basis for a topology|basis for the topology]] on $X$. ![[CleanShot 2024-06-10 at [email protected]]] It will be enough to show that the [[neighborhood]] $(\beta, W)$ of $\beta$ lives in the intersection $(\alpha_{0}, U_{0}) \cap (\alpha_{1}, U_{1})$, i.e., that any [[equivalence class|class]] $[\gamma]$ of [[parameterized curve|paths]] from $x_{0}$ to a point in $W$ that is obtained by 'lengthening $\beta also equals some postconcatenation on $\alpha_{0}$ living in $U_{0}$; likewise for $\alpha_{1}$ and $U_{1}$. Due to [[path-connected|path-connectedness]] of $U_{0}$, $U_{1}$, and (hence) $U_{0} \cap U_{1}$, this is definitely the case, see teal paths below: ![[CleanShot 2024-06-10 at [email protected]]] We give to $\widetilde{X}$ the [[topology generated by a basis|topology generated by this basis]]. > **(b).** We now move to show that our choice of [[topological space|topology]] is a good one. To begin, we'd like to show $[\gamma] \xmapsto{p} \gamma(1)$ [[continuous]]. It will suffice to check using the [[basis characterization of continuity]]. Take $U \in \mathscr{U}$ and consider $p ^{-1}(U)=\{ [\alpha] \in \widetilde{X} : \alpha(1) \in U \}$. If $[\alpha] \in p ^{-1}(U)$, then it is easy to see that $[\alpha] \in (\alpha, U) \subset p ^{-1}(U)$, hence $p ^{-1}(U)= \text{int }p ^{-1}(U)$ is open in $\widetilde{X}$. > > **(c).** Now we show $p$ is a [[covering space|covering map]]. We first claim that each map $p |_{(\alpha, U)}: (\alpha, U) \to U$ is a [[homeomorphism]] by showing it is [[bijection|bijective]] and [[open map|open]]. It is [[surjection|surjective]] because $U$ is [[path-connected]], and hence every point in $U$ may be reached from a [[parameterized curve]] starting at $\alpha(1)$. If $[\beta]$ and $[\beta']$ are two [[homotopy|path homotopy classes]] which end at the same point, and each my be obtained from $\alpha$ by concatenation of a path in $U$, then they differ by a concatenation of a [[parameterized curve|loop]] in $U$. As every loop in $U$ is [[homotopy|nulhomotopic]] in $X$, in follows that $[\beta]=[\beta']$ and [[injection|injectivity]] holds. Finally, $p\big( (\gamma ,V) \big)=V$, so $p$ is an [[open map]], hence $p |_{(\alpha, U)}$ is too. > We next claim that $p ^{-1}(U)$ is [[evenly covered|partitioned]] into [[evenly covered|sets ('slices)']] of the form $(\alpha, U)$. We have already seen it to be [[cover|covered]] by sets of such form, so it suffices now to show that any two such sets must be disjoint or equal. Suppose $[\gamma] \in (\alpha, U) \cap (\beta, U)$. Then $[\gamma]=[\alpha * \alpha']=[\beta * \beta']$ for paths $a'$ and $b'$ in $U$, meaning $[\alpha]=[\beta * \beta' * \overline{\alpha'}]$ and $\beta' * \overline{\alpha'}$ is a [[parameterized curve]] in $U$, so $[\alpha] \in (\beta, U)$. But then if $[\delta] \in (\alpha, U)$, $[\delta]=[\alpha * \alpha'']$, then $[\delta]=[\beta * \beta' * \overline{\alpha'} * \alpha'']$ and so $[\delta] \in (\beta, U)$. Since $[\delta]$ is arbitrary we have $(\alpha, U) \subset (\beta , U)$. The reverse inclusion follows from very similar reasoning. > **(d).** Finally, we show $\widetilde{X}$ is [[simply connected]]. The key observation is that if $\gamma:I \to X$ is a [[parameterized curve]] starting at $x_{0}$, then [[the homotopy lifting lemma|its]] [[lifting|lift]] to $\widetilde{X}$ starting at $[c_{x_{0}}] \in \widetilde{X}$ is the [[parameterized curve]] $s \mapsto [t \mapsto \gamma(st)]: I \to \widetilde{X},$ which ends at the point $[\gamma] \in \widetilde{X}$. Thus if a [[parameterized curve|loop]] $\gamma$ in $X$ based at $x_{0}$ [[lifting|lifts]] to a [[parameterized curve|loop]] in $\widetilde{X}$ based at $[c_{x_{0}}]$, then $[\gamma]=[c_{x_{0}}]$, hence $p_{*}(\pi_{1}(\widetilde{X}, x_{0}))=(e) \leq \pi_{1}(X,x_{0})$. This implies, by [[pushforward morphism of covering map is an embedding]], that $\pi_{1}(\widetilde{X},x_{0})$ is trivial. ---- #### [^1]: It is a [[parameterized curve|loop]] based at $\tilde{x}_{0}$ rather than just a path starting at $\tilde{x}_{0}$ because $\gamma$ is a [[parameterized curve|loop]] *in $U$* and $V_{\beta}$ is mapped [[homeomorphism|homeomorphically]] onto $U$ via $p |_{V_{\beta }}$. [^2]: 'Neighborhood' here is the version of the definition that just means 'open set containing'. [^3]: Maybe one way to motivate this definition would be to recall the definition of [[simply connected]]: [[path-connected]] + [[homotopy|nulhomotopic]] [[parameterized curve|loops]]. And also the (equivalent) definition of [[semilocally simply connected]]: trivial [[homomorphism of fundamental groups induced by a continuous map|pushforward morphisms]] induced by [[inclusion map|inclusion]]. We are trying to show $X$ has a [[basis]] of [[path-connected]] sets (ala the requirement in simple connectedness) with trivial pushforwards induced by inclusion (ala semilocal simple connectedness). [^4]: One should expect this to be true, in analogy to the previous footnote and [[in a simply connected space, any two paths having the same initial and final points are path homotopic]]. Indeed, the reasoning goes just the same: given paths $\alpha,\beta$ between $x$ and $y$, we [[fundamental groupoid|concatenate]] to form the [[parameterized curve|loop]] $[\alpha \circ \overline{\beta}]$ based at $x$ which is [[homotopy|nulhomotopic]] *in $X$* (remember that it need not be nulhomotopic in $U$) and then concatenating $[\beta]$ to the resultant equation $[\alpha \circ \overline{\beta}]=[e_{x}]$ witnesses $\alpha \simeq \beta$ as paths. ----- #### References > [!backlink] > ```dataview TABLE rows.file.link as "Further Reading" FROM [[]] FLATTEN file.tags GROUP BY file.tags as Tag > [!frontlink] > ```dataview > TABLE rows.file.link as "Further Reading" > FROM outgoing([[]]) > FLATTEN file.tags as Tag > WHERE Tag = "#definition" OR Tag = "#theorem" OR Tag = "#MOC" OR Tag = "#proposition" OR Tag = "#axiom" > GROUP BY Tag > ``` #reformatrevisebatch01