2

Reading "Morse Theory" by J.Milnor I found more than once this way of approach:

Let X and Y be two topological space. Suppose we want to define a continuous map from a space obtained by attaching a n-cell to X along a map $\phi: S^{n-1}\rightarrow X$. Let's call $X\sqcup_{\phi} D^{n}$ this space, $i:X\rightarrow X\sqcup D^{n}$ and $j:D^{n}\rightarrow X\sqcup D^{n}$ the two inclusion maps and $\pi:X\sqcup D^{n}\rightarrow X\sqcup_{\phi} D^{n}$ the natural projection. Additionally, suppose we know two continuous functions $f:X\rightarrow Y$ and $g:D^{n}\rightarrow Y$ such that the function $F:X\sqcup_{\phi} D^{n}\rightarrow Y$ defined by cases:

$F(\pi(i(x)))=f(x)$ if $x\in X$.

$F(\pi(j(x))=g(x)$ if $x\in D^{n}$

is well defined.

My problem concerns the proof of the continuity of the function: The book doesn't say anything about how to prove the continuity of the function defined this way. I tried to give an answer by applying pasting lemma: if I show that both the sets in which we defined F are closed, it follow the desired result. $\pi^{-1}\{\pi(i(x)): x\in X\}=\{(i(x)): x\in X\}\cup S^{n-1}$ is obviously a closed subspace of $X\sqcup D^{n}$, while I'm not able to say much about $\pi^{-1}\{\pi(j(x)):x\in D^{n}\}=j(D^{n})\cup i(\phi(S^{n-1}))$. If X is Hausdorff $\phi(S^{n-1})$ is closed in X and so $j(D^{n})\cup i(\phi(S^{n-1}))$ is closed in $X\sqcup D^{n}$ but in general I don't know how to prove it and in Milnor's book there aren't any hypothesis on X.

Can someone give me an hand? :)

Thanks in advance!

ghiroz
  • 253
  • I think there's something wrong with your equation $\pi^{-1}{\pi(i(x)): x\in X}={\pi(i(x)): x\in X}\cup S^{n-1}$, it should be $\pi^{-1}{\pi(i(x)): x\in X}={i(x): x\in X}\cup S^{n-1}$, right? And at least according to Wikipedia, a CW complex is always Hausdorff. – jplitza Mar 06 '17 at 11:01
  • Thanks, I edited! The fact is that in Milnor's book the author is talking about a general topological space X(and so not necessarily a CW-complex) with a n-cell attached...I don't know if he just implicitly suggests that he has to be Hausdorff. – ghiroz Mar 06 '17 at 11:12

1 Answers1

1

You're basically asking for the universal property of quotient spaces: If $f\colon A\sqcup B\rightarrow C$ is a continuous map with $f(a)=f(b)$ if $a\sim b$, then there is a continuous map $\tilde f\colon A\sqcup B\,/\!\sim\,\rightarrow C$.

To prove this, let $U\subseteq C$ be an open subset. By definition of the quotient topology, $\tilde f^{-1}(U)$ is open in $A\sqcup B\,/\!\sim$ if and only if $\bigcup \tilde f^{-1}(U)=(f|_A)^{-1}(U)\sqcup (f|_B)^{-1}(U)$ is open in $A\sqcup B$, which it is because $f$ is continuous.

In your case, $\bigcup \tilde f^{-1}(U) = \pi^{-1}(F^{-1}(U))=i(f^{-1}(U))\sqcup j(g^{-1}(U))$ is open because $i$ and $j$ are homeomorphisms when restricting the codomain to their corresponding image. In particular, they are open maps.

See also “Proof of the universal property of the quotient topology” and “Adjunction space is a pushout”.

jplitza
  • 223