I have a question about an argument used in Szamuely's "Galois Groups and Fundamental Groups" in the excerpt below (or look up at page 38):

enter image description here

In order to show the category equivalence claimed in Thm 2.3.4 to verify the fully flatness of the functor $Y \mapsto Fib_x(Y)$ we have to show that for given two covers $p:Y \to X, q: Z \to X$ each map $\varphi : Fib_x (Y ) → Fib_x (Z)$ of $π_1(X, x)$-sets comes from a unique map $Y → Z$ of covers of $X$.

The author argues as follows: We take a $y \in Fib_x(Y)$ and identifies it with corresponding map $\pi_y: \widetilde{X}_x \to Y$ (remark: $\widetilde{X}_x$ is the universal cover) as claimed in 2.3.5:

enter image description here

For the construction of $\pi_y$ see here:

enter image description here

Indeed this works since by Thm 2.3.5 the functor $Fib_x$ is representable by $\widetilde{X}_x$.

Then it is shown in the proof that the map

$\pi_{\phi(y)}: \widetilde{X}_x \to Z$ factorizes as $$\widetilde{X}_x \xrightarrow{\pi_y} Y \xrightarrow{\psi_y} U_y \backslash \widetilde{X}_x \xrightarrow{p_y} Z$$

The PROBLEM is why is $p_y$ independent of $y$?

Indeed following the proof and using the naturality of diagram from 2.3.5 to show the claim that $\phi$ comes from a map of covers $m: Y \to Z$ it suffice to show that the map $Hom_X(\widetilde{X}_x , Y) \to Hom_X(\widetilde{X}_x , Z)$ is induced by a map $m:Y \to Z$ of covers.

The author showed that by construction $\pi_{\phi(y)}$ is the image of $\pi_y$ by composing with $p_y$.

The goal is to show that for every $y' \in Fib_x(Y)$ the morphism $\pi_{\phi(y')}:\widetilde{X}_x \to Z$ is the composition $\widetilde{X}_x \xrightarrow{\pi_{y'}} Y \xrightarrow{p_y} Z$ (therefore $p_y$ is independent of $y$)) but this isn't clear to me why this holds.


If I understand correctly, you see why there exists a map $f:Y \to Z$ of covers whose restriction to fibres is $\phi$, and your question is why that $f$ is unique.

Assume that $g:Y \to Z$ was also a map of covers whose restriction to fibres was $\phi$, then we can think of $f,g$ as lifts of $\pi_y$ to the covering $Z \to X$. Now as $Y$ is connected, if they agree at a point, $f = g$ (Proposition 2.2.2 I think). But they both agree on the fibres $Fib_x(Y)$, thus they are the same.


We have a map $f:Y \to Z$ of covers, such that for a point $y_0 \in Fib_x(Y)$, $f(y_0) = \phi(y_0) \in Fib_x(Z)$. Now we must show that for every $y \in Fib_x(Y)$, $f(y) = \phi(y)$.

We use that $Y,Z$ can be chosen path-connected. Then $\pi_1(X,x)$ acts transitively on $Fib_x(Y)$. To see this, for any $y \in Fib_x(Y)$, consider a path $\alpha$ from $y_0$ to $y$. Then $[\beta] = [p \circ \alpha] \in \pi_1(X,x)$ takes $y_0$ to $y$. Then $\phi(y) = \phi([\beta] \bullet y_0) = [\beta]\bullet f(y_0)$ where $\bullet$ denotes the action of $\pi_1(X,x)$.

It remains to show $f(y) = [\beta]\bullet f(y_0)$. To see this note that as $qf = p$, we have that $[\beta] = [q \circ \gamma]$, for the path $\gamma = f \circ \alpha$ from $f(y_0)$ to $f(y)$. Thus $[\beta] \bullet f(y_0) = f(y)$.

  • $\begingroup$ No, the problem is the existence. In the construction the author picks a $y \in Fib_x(Y)$ and showed that there exist a $f_y:Y \to Z$ with $f_y(y)= \phi(y)$. The problem is that I don't see why this $f=f_y$ is not dependent of $y$. Namely it's not clear that the whole $\phi$ arises from restricting of $f$ ($=f_y$). The author shows that $f_y$ has only the same image as $\phi$ only for $y$ not for the whole fiber $Fib_x(Y)$. $\endgroup$ – KarlPeter Apr 18 at 10:35
  • $\begingroup$ Ah, I see the issue. This follows from the fact that we can choose $Y,Z$ to be path connected, then the action of the fundamental group is transitive. I have updated the answer to reflect this. $\endgroup$ – Tim The Enchanter Apr 18 at 11:04
  • $\begingroup$ ah yes of couse the $\pi_1(X,x)$ action we have take into account. One remark: Why we can assume that $Y,Z$ are path connected? Or equivlently why we can assume that $Y,Z$ are connected? (indeed here all spaces are locally simply connected and therefore connected and path connected are equivalent) $\endgroup$ – KarlPeter Apr 18 at 16:49
  • $\begingroup$ As you mentioned, all spaces are locally simply connected, hence locally path connected and hence covers $Y$ and $Z$ are homeomorphic to disjoint unions of their path components. Thus to define a map of covers it suffices to define it on each path component of $Y$, and continuity ensures that each path-component of $Y$ maps into a path-component of $Z$. Now the above argument shows that restricted to each path component of $Y$ the map is uniquely defined, from which we recover the uniqueness of the map $Y \to Z$. $\endgroup$ – Tim The Enchanter Apr 18 at 17:51
  • $\begingroup$ For the sake of completeness, the above comment also (somewhat subtly) uses the fact that here, the restriction of a covering map to a connected component is still a covering map. The local homeomorphism condition follows from local simple connectedness of $X$, as we can restrict our evenly covered neighbourhoods to be connected. To see surjectivity use path lifting and (path-)connectivity of $X$. $\endgroup$ – Tim The Enchanter Apr 18 at 18:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.