3

Let $p :X \to Y$ be a covering map and $Y$ path connected. Show that the cardinality of $p^{-1}(\{y\})$ is the same for every $y \in Y$.

Let $y_0, y_1 \in Y$, then $\exists \alpha :I \to Y$ such that $\alpha(0)=y_0$ and $\alpha(1)=y_1$. Consider now $x \in p^{-1}(\{y_0\})$. By the lifting criterion we have that $\exists \tilde{\alpha}: I \to X$ such that $\tilde{\alpha}(0)=x$. If I now define $\varphi:p^{-1}(\{y_0\}) \to p^{-1}(\{y_1\})$ and I can show that $\varphi$ is a bijection I think that would be satisfactory. What kind of choices do I have for $\varphi$ or how I should think about it?

2 Answers2

1

An approach that doesn't rely on path-lifting is to define an equivalence relation $y \sim y'$ iff $p^{-1}(\{y\})$ and $p^{-1}(\{y'\})$ have the same cardinality. It follows the idea given in Introduction to Topological Manifolds by John M. Lee.

Now any $y$ in $Y$ has an evenly covered neighborhood $U_y$ in $Y$. As a consequence of the definition of covering maps, we can write $$p^{-1}(U_y) = \bigcup\limits_{\alpha\in A}V_\alpha$$ where the $V_\alpha$ are disjoint (open) subsets of $X$, and $V_\alpha$ is homeomorphic to $U_y$ under the restriction $p|_{V_\alpha}: V_\alpha \to U_y$ for each $\alpha$.

Suppose $y' \in U_y$. We can define a map $f: A \to p^{-1}(\{y'\})$ by $$f(\alpha) = p|_{V_\alpha}^{-1}(y') \in p^{-1}(\{y'\}) \cap V_\alpha$$

Since the $V_\alpha$ are disjoint, $f$ must be injective. Since the $V_\alpha$ cover $p^{-1}(\{y'\})$, $f$ must be surjective. Therefore $f$ is a bijection, which shows that the cardinality of $p^{-1}(\{y'\})$ is the cardinality of $A$ for all $y' \in U_y$. This shows that $U_y$ is contained in the equivalence class of $y$, so the equivalence classes of $\sim$ are open in $Y$.

Now if $y$ is any point in $Y$, let $[y]$ be its equivalence class. As shown above it is open in $Y$. The equivalence classes partition $Y$ so the complement of $[y]$ is a union of equivalence classes, which are also open. This shows that $[y]$ is both open and closed in $Y$.

Path connectedness implies connectedness. Since $Y$ is connected and $[y]$ is nonempty, $[y] = Y$ so there is only one equivalence class.

Tob Ernack
  • 5,319
  • This only works for covering maps of locally path connected spaces as considered by Lee. Evenly covered sets in the sense of Lee have to be open connected sets, and for a general $Y$ it is possible that points do not have evenly covered neigborhoods. – Paul Frost Mar 17 '22 at 14:18
  • @PaulFrost You're right that Lee does define covering maps with the additional requirement that $Y$ is locally path connected, which is not necessarily true when $Y$ is only path connected as the OP assumes. But in this proof, we don't actually need to know that evenly covered neighborhoods are connected. We really only need to know that $p^{-1}(U_y)$ is a union of disjoint subsets of $X$, each of which is homeomorphic to $U_y$. This is enough to show that the cardinality of the fibers of all points in $U_y$ is the same as the cardinality of that particular partition of $p^{-1}(U_y)$. – Tob Ernack Mar 17 '22 at 14:38
  • You are of course right that it is true under weaker assumptions. But this requires a slight change of arguments: "$p^{-1}(U_y)$ is a union of disjoint connected open subsets of $X$" does no longer fit. – Paul Frost Mar 17 '22 at 14:49
  • @PaulFrost I updated my answer, thank you. – Tob Ernack Mar 17 '22 at 15:14
0

Your idea is correct. Given two fibers $F_i = p^{-1}(y_i)$, you can construct a bijection $\varphi : F_0 \to F_1$ as you did:

Choose any path $\alpha : I \to Y$ such that $\alpha(i) = y_i$. For each $x \in F_0$ there exists a unique lift $\alpha_x : I \to X$ such that $\alpha_x(0) = x$. Then define $\varphi(x) = \alpha_x(1) \in F_1$. Clearly $\varphi(x)$ may depend on the choice of $\alpha$, thus one could write more precisely $\varphi_\alpha : F_0 \to F_1$.

Why is $\varphi$ a bijection? Let $\alpha^{-} : I \to Y$ be the inverse path of $\alpha$. As above this yields a function $\varphi^- : F_1 \to F_0$. Then

  1. $\varphi^-\circ \varphi = id$:
    $\varphi^-(\varphi(x))$ is obtained by taking the unique lift $\alpha^{-}_{\varphi(x)} : I \to X$ of $\alpha^{-}$ such that $\alpha^{-}_{\varphi(x)}(0) = \varphi(x)$; then $\varphi^-(\varphi(x)) = \alpha^{-}_{\varphi(x)}(1)$. But it is clear that $\alpha^{-}_{\varphi(x)}$ is nothing else than the inverse path of $\alpha_x$, thus $\varphi^-(\varphi(x)) = x$.
  2. $\varphi \circ \varphi^- = id$:
    This is similar as 1.

Note that all fibers $p^{-1}(y)$ have the same cardinality under the weaker assumption that $Y$ is connected. See If $h : Y \to X$ is a covering map and $Y$ is connected, then the cardinality of the fiber $h^{-1}(x)$ is independent of $x \in X$.

Paul Frost
  • 87,968