I am studying Elementary Differential Geometry written by Barrett O'Neill.
In page 188, Chapter 4.7, there is an explanation why 2-sphere is simply connected.
The following is from the text :
==================================================================
The 2-sphere $\Sigma$ is simply connected. Consider the following scheme of proof. Let $\alpha$ be a loop in $\Sigma$ at, say, the north pole of $\Sigma$. Pick a point q not on $\alpha$. For simplicity, suppose q is the south pole. Now let x be the homotopy under which each point of $\alpha$ moves due north along a great circle, reaching p in unit time. This x is a homotopy of $\alpha$ to a constant, as required.
But there is a difficulty here: finding the point q. In our usual case, where $\alpha$ is differentiable, techniques from advanced calculus will show that there is always a point q not on $\alpha$. However, if $\alpha$ is merely continuous, it may actually fill the entire sphere. In this case, topological methods can be used to deform $\alpha$ slightly, making it no longer space-filling: then the scheme above is valid.
==================================================================
My questions are:
Why why why why does he pick q ??? for what ???
I cannot even grasp What he wants to say in the 2nd paragraph. Can you explain why the 2nd paragraph is needed?
Thank you.
Well, how would you extend the given homotopy across $q$? The set $S^2\setminus q$ is just $\mathbb{R}^2$ (consider, for example, the Riemann sphere); with that identification, the given homotopy is just the straight-line contraction $h_t(x) = tx$ of $\mathbb{R}^2$ to the origin. The space $S^2$ itself is certainly not contractible.
In brief, the argument above is that any loop $\alpha:S^1 \to S^2$ has to factor through the contractible space $\mathbb{R}^2$ and thus be null-homotopic. It only works if $\alpha$ is not surjective. As the text indicates, it can be eliminated even in the continuous category by, for example, taking a small contractible neighborhood $U$ around a point in $S^2$ and taking a homotopy of $\alpha\cap U$ to something reasonable.