Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

This is quite a detailed question: I'm struggling to understand a few parts of a proof of the following Lemma. I've placed stars ($\bigstar$) where I'd like to draw your attention.


Lemma: Let $P:Y\to X $ be a covering projection and $ f: Z \to X $ a continuous map where $Z$ is a simply connected and locally path connected space. Suppose given base points $x_0, y_0, z_0$ of $X,Y,Z$ with $p(y_0) = x_0 = f(z_0) $. Then there is a unique continuous $ \tilde{f} : Z \to Y$ with $ p \tilde{f} = f$ and $ \tilde{f}(z_0) = y_0$.


The proof, summarised/paraphrased, is as follows:

Given $ z \in Z$, choose a path $u$ from $z_0$ to $z$ in $Z$. Let $ \tilde{u} $ be the unique lifting of $fu$ to a path in $Y$ starting at $y_0$, and define $ \tilde{f}(z) = \tilde{u}(1) $.

Can then show $ \tilde{f} $ is well-defined, and that it's the only possible mapping that could work. It remains to be shown that it is continuous.

To do so, let $ z \in Z$ and let $V$ be an open neighbourhood of $ \tilde{f}(z)$. Without loss of generality ($\bigstar$), we can assume $V$ is of the form $ h^{-1}(U \times {d}) $, where $U$ us an evenly covered neighbourhood of $ p\tilde{f}(z) = f(z)$ and $ h: p^{-1}(U) \to U \times D$ is a homeomorphism with $D$ discrete.

$f^{-1}(U)$ is an open neighbourhood of $z$, so contains a path connected neighbourhood $W$. For any $ z' \in W$, we can choose a path from $z_0$ to $z'$ of the form $u.v$ where $u$ is a path from $z_0$ to $z$ and $v$ takes values in $W$. Now $fv$ takes values in $U$, which is evenly covered by $p$. So its lifting to a path in $Y$ starting at $\tilde{f}(z)$ must take values in $V \ (\bigstar \bigstar)$, and in particular $ \tilde{f}(z') = \tilde{(u.v)}(1) \in V$. So $W \subseteq \tilde{f}^{-1}(V) $ and hence $\tilde{f}$ is continuous.


My confusions are as follows:

$ \bigstar $: Is the following reasoning as to why this does not lose generality correct? Given any $y = \tilde{f}(z)$, consider $ p(y) $. We can find an evenly covered neighbourhood $U$ containing $p(y)$. Then there is a homeomorphism $ h : p^{-1} (U) \to U \times D$ where $D$ is a discrete space. Then $ (p(y), d) \in U \times \{d\} $ for any $ d \in D$ and $ h^{-1} (p(y),d) \subseteq h^{-1} (U \times \{d\})$. But $ h^{-1}(p(y),d) = p^{-1} p(y) $ which contains $y$.

$ \bigstar \bigstar $: I'm confused with this bit. Intuitively, $ p^{-1}(U)$ 'looks like' lots of copies of $U$, and $V$ is one of those copies. But I can't turn this into a definitive reason why $ \tilde{v} $ (i.e. the lifting of $fv$) must be entirely contained within $V$. On a different note, am I correct in saying $ V \cong U $? I'm convinced there's a simple set-theoretic explanation, but it is eluding me.

Any help will be greatly appreciated - I understand that this is a long post.

Thank you.

share|improve this question
9  
+1 for the details and pointing out your question clearly. :) –  Srivatsan Nov 2 '11 at 14:18
add comment

2 Answers

The explanation you give for $\bigstar$ seems correct to me.

As for $\bigstar \bigstar$:
Since the trace of $\tilde v$ (that is the set of points $\tilde v ([0,1] )$ in $Y$ through which the path $\tilde v$ runs) is a connected set and since it is contained in the disjoint union of the open sets $h^{-1} (U\times\lbrace \delta \rbrace) \; (\delta \in D)$ it must be contained in exactly one of them, namely $V=h^{-1} (U\times\lbrace d \rbrace) $ since this is the one that contains $\tilde v(0)=\tilde f(z)$.

Finally, you are absolutely correct that $U$ and $V$ are homeomorphic.
The natural homeomorphism is the restriction $p|V:V\to U$, whose inverse is the continuous map $h^{-1}\circ j:U\to V $, where $j:U\to U\times D: u\mapsto (u,d)$ is the insertion of $U$ in the relevant slice of the product.

share|improve this answer
add comment

I've not had time to think about this for long, so I'd welcome any corrections to the following.

$\bigstar $: Yes, that's fine. Covering projections can seem a strange concept at first, but I find that it helps a lot to bear in mind the typical example $ p : \mathbb{R} \to S^1 $ with $ p(t) = e^{2 \pi i t} $, which I'm sure you'll have seen (imagine bending the real line into a helix, which projects down to cover the unit circle: our discrete space is just $ \mathbb{Z}$). Given any point on the helix, it's clear you can find an open interval around that point which is one of the 'copies' of the interval on $S^1$ that it projects down onto under $p$.

$ \bigstar \bigstar $: You have that $fv(t) \in U $ for all $t$. This means that $ (fv(t),d) \in U \times \{d\}$ for all $t$. Now, you should know that by the definition of 'evenly covered', you have that $ ph^{-1} (x,d) = x $ for all $x \in X$ and $d \in D$ (see here: Query about proof of homotopy lifting property). This means that $ ph^{-1}(fv(t),d) = fv(t) $ for all $t$. But, by uniqueness of lifting, this exactly gives us the lifting of $fv$: we must have that $ \tilde{v} = h^{-1}(fv(t),d) $. By the form of $V$, we have that $ \tilde{v}(t) \in V $ for all $t$.

share|improve this answer
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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