1

Let $p:\tilde{X}\to X$ be a covering map (where $X$ is path conncetd and locally path connected). We call a continuous map $s:X\to\tilde{X}$ a section if $p\circ s=id_X$. Fix a point $x_0\in X$. Prove that there exists a bijective correspondance between the set of points $y\in p^{-1}(x_0)$ such that every loop $f$ with base point $x_0$ lifts to a loop $\tilde{f}$ with basepoint $y$ and the sections of $p$.

Attempt: I defined a map $s_y:X\to\tilde{X}$ with $s_y(x)=\tilde{f}(1)$ where $f$ is a path with $f(0)=x_0,f(1)=x$ and $\tilde{f}(0)=y$. Can we prove this correspondance is onto?

idocomb
  • 339

2 Answers2

3

Let $L(x_0)$ denote the set of points $y\in p^{-1}(x_0)$ such that every loop $f$ with basepoint $x_0$ lifts to a loop $\tilde{f}$ with basepoint $y$. Note that the latter means that the unique lift $\tilde{f}$ of $f$ with $\tilde{f}(0) = y$ has the property $\tilde{f}(1) = y$.

Let $S$ denote the set of sections $s : X \to \tilde X$ and define $\phi : S \to p^{-1}(x_0), \phi(s) = s(x_0)$.

  1. $\phi$ is injective: Each section $s$ is a lift of $id_X$, thus $s$ is uniquely determined by the value $s(x_0)$. Here we use the fact that $X$ is connected.

  2. $\phi(S) \subset L(x_0)$: For any loop $f$ with basepoint $x_0$ the loop $s \circ f$ is a lift of $f$ with basepoint $y$.

It remains to show that each $y \in L(x_0)$ is in the image of $\phi$. We need some preprations.

  1. For each path component $C$ of $\tilde X$, the restriction $p\mid_C : C \to X$ is a covering map.
    i) $p \mid_C$ is surjective: Let $x \in X$. Pick $c_0 \in C$ and choose a path $u : I \to X$ such that $u(0) = p(c_0)$ and $u(1) = x$. There exists a lift $\tilde u$ of $u$ such that $\tilde u (0) = c_0$. Then $\tilde u(1) \in C$ and $p(\tilde u(1)) = u(1) = x$ which shows that $x \in p(C)$.
    ii) Let $x \in X$. Choose a path connected open neigborhood $U$ of $x$ which is evenly covered by $p$. Then $p^{-1}(U) = \bigcup_{\alpha \in A} V_\alpha$ with pairwise disjoint open $V_\alpha \subset \tilde X$ which are mapped by $p$ homeomorphically onto $U$. Thus each $V_\alpha$ is path connected and therefore contained in some path component of $\tilde X$. Hence $(p \mid_C)^{-1}(U) = \bigcup_{\alpha \in A_C} V_\alpha$ for a suitable nonempty subset $A_C \subset A$. This means that $U$ is evenly covered by $p \mid_C$.

  2. The cardinality of $(p \mid_C)^{-1}(x_0)$ is the number of sheets of $p \mid_C$. Let $y \in p^{-1}(x_0)$ and $C$ its path component. We claim that
    (a) If $p \mid_C$ has one sheet, then $y \in \phi(S) \subset L(x_0)$.
    (b) If $p \mid_C$ has more than one sheet, then $y \notin L(x_0)$.
    This clearly proves $\phi(S) = L(x_0)$.
    (a) In this case $p \mid_C$ is an open bijection, i.e. a homeomorphism. Thus $s = (p \mid_C)^{-1} : X \to C \hookrightarrow \tilde X$ is a section such that $s(x_0) = y$ which means $y \in \phi(S)$.
    (b) In this case there exists a point $y' \in C$ other than $y$ such that $p(y') = x_0$. There is a path $\tilde f$ in $C$ from $y$ to $y'$; it is no loop. On the other hand f = $p \circ \tilde f$ is a loop based at $x_0$. But since $\tilde f$ is the unique lift of $f$ with $\tilde f(0) = y$, we see that $y \notin L(x_0)$.

Update:

Instead of steps 3. and 4. we can invoke the following two well-known theorems:

(1) Let $p:\tilde{X}\to X$ be a covering map, $x_0 \in X$ and $y \in p^{-1}(x_0)$. The induced homomorphism $p_* : \pi_1(\tilde X,y) \to \pi_1(X,x_0)$ is injective. The image subgroup $p_*(\pi_1(\tilde X,y)) \subset \pi_1(X,x_0)$ consists of the homotopy classes of loops in $X$ based at $x_0$ whose lifts to $\tilde X$ starting at $y$ are loops.

(2) Lifting theorem. Let $p:\tilde{X}\to X$ be a covering map, $x_0 \in X$ and $y \in p^{-1}(x_0)$ and $g : (Z, z_0) \to (X,x_0)$ be a map with $Z$ path-connected and locally path connected. Then a lift $\tilde g : (Z, z_0) \to (\tilde X, y)$ of $g$ exists if and only if $g_*(\pi_1((Y, y_0)) \subset p_*(\pi_1(\tilde X,y))$.

Now (1) says that $y \in L(x_0)$ if and only if $p_*(\pi_1(\tilde X,y)) = \pi_1(X,x_0)$ and (2) says that the identity $id_X : (X,x_0) \to (X,x_0)$ has a lift $s : (X,x_0) \to (\tilde X,y)$ (which is nothing else than a section of $p$ with $\phi(s) =y$) if and only if $\pi_1(X,x_0) = (id_X)_*(\pi_1((X, x_0)) \subset p_*(\pi_1(\tilde X,y))$. The latter condition means $\pi_1(X,x_0) = p_*(\pi_1(\tilde X,y))$. This shows $L(x_0) = \phi(S)$.

Paul Frost
  • 76,394
  • 12
  • 43
  • 125
  • I thought I settled those cases but I might have made a mistake somewhere. I think 4. only needs case a) since $s$ is an isomorphism of the covering maps $p:\tilde{X}\to X$ and $id_X$ and thus it exists iff $\pi_1(\tilde{X},\tilde{x_0})=\pi_1(X,x_0) \iff p^{-1}(x_0)=1$ – idocomb Feb 09 '21 at 21:45
  • @idocomb You need some kind of proof. My access is completely elementary, but I updated my answer to give an alternative proof. Note that the proof of the lifting theorem includes the steps mentioned in my comment to Sam Freedman's answer: That $s_y(x)$ does not depend on the choice of $f$ and that $s_y$ is continuous. – Paul Frost Feb 10 '21 at 09:47
2

Hints:

To show that the correspondence is onto, start with an arbitrary section $s'$. Your goal is to find a point in $p^{-1}(x_0)$ with correct lift-looping behavior that maps to $s$ under your correspondence $y \mapsto s_y$.

Since $s'(x_0)$ is in the fiber over $x_0$, and $s'(x_0)$ does have correct lift looping (why?), it remains to show that this point maps to $s'$ under the correspondence. That is, you want to show $s_{s(x_0)} = s'$. How can you use properties of covering spaces to show this?

Sam Freedman
  • 3,989
  • I assume correct loop lifting is simply because $f(0)=f(1) \implies s'\circ f(0)=s'\circ f(1)$ ? – idocomb Feb 09 '21 at 15:21
  • Yes that’s right— you’re pushing $f$ upstairs with the section $s’$ – Sam Freedman Feb 09 '21 at 15:25
  • I am thinking of composing a path $f$ with $f(0)=x_0, f(1)=x$ with the identity giving me: $p\circ s_{s(x_0)}\circ f=p\circ s'\circ f$ which will give me then $s_{s(x_0)}\circ f = s'\circ f$ since $s_{s(x_0)}\circ f(0)= s'\circ f(0)=s'(x_0)$. Then $s_{s(x_0)}\circ f (1)= s'\circ f(1) \implies s_{s(x_0)}(x)=s'(x)$. Am I correct? – idocomb Feb 09 '21 at 15:53
  • Yes, also looks good to me—you are using uniqueness of path-lifting to show the two sections agree. – Sam Freedman Feb 09 '21 at 15:56
  • exactly, is this what you had in mind or is there a more direct approach? – idocomb Feb 09 '21 at 15:57
  • This is the approach I thought of, but there certainly could be a better one out there – Sam Freedman Feb 09 '21 at 16:02
  • 1
    There is still of lot of work to be done: You must show that $s_y(x)$ does not depend on the choice of $f$ and that $s_y$ is continuous. The proof must involve the condition that $X$ is locally path connected, – Paul Frost Feb 09 '21 at 17:02