2

I am working on exercises from Hatcher, algebraic topology and in a certain exercise (1.1.5), we identify a loop, which is a path $\gamma:I\to X$ with a map $f:S^1\to X$, which we can do since $\gamma(0)=\gamma(1)$ and project to the quotient which is homeomorphic to $S^1$.

Now, suppose every map $f:S^1\to X$ extends to a map $D^2\to X$. I can show that a map fron $S^1\to X$ is then homotopic to a constant map $S^1\to X$, how can I justify that this implies that every loop is homotopic to a constant loop, I.e. $\pi_1(X,x_0)=0?$

  • If $H:I\times S^1\to X$ is a homotopy then it induces a homotopy $H':I\times I\to X$ such that $H'(t,0)=H'(t,1)$. And vice versa. And $H(0,\cdot)$ is constant if and only if $H'(0,\cdot)$ is. – freakish Feb 18 '19 at 09:43
  • How is the homotopy $H$ induced? –  Feb 18 '19 at 09:45
  • If you start from $H$ then you take the quotient map $q:I\to S^1$ and define $H'(s,t):=H(s, q(t))$. The other way, if you have $H'$ then you define $S^1=I/\sim$ and then $H(s,[t]):=H'(s,t)$. This works fine because $H'(t,0)=H'(t,1)$. – freakish Feb 18 '19 at 09:47
  • You must be careful because it is not always true that $(Z \times A) / ((t,x) = (t,y)$ iff $x \sim y)$ is the same as $Z \times (A/ (x \sim y))$. Although this is the case when $Z$ is locally compact, so it is true here. Essentially, doing the quotient level wise is not necessarily the same as doing the quotient then taking the product. You need that these are the same to argue that a continuous function $I \times I \rightarrow X$ that has $(t,0)=(t,1)$ induces a map on $I \times S^1 $ rather than just on the quotient mentioned before. I believe this is why the smash product is not associative – Connor Malin Feb 18 '19 at 18:40
  • @ConnorMalin Can you elaborate on this please? –  Feb 18 '19 at 18:54
  • See here https://math.stackexchange.com/questions/31697/when-is-the-product-of-two-quotient-maps-a-quotient-map . Essentially what you can say is that if the map $1 \times q : Z \times A \rightarrow Z \times A/\sim$ (where $q: A \rightarrow A/\sim$) were a quotient map, then the two spaces I mentioned are homeomorphic through the map $([(x,y)] \rightarrow (x,[y])$, but the link above discusses when this is and when this isn't true. – Connor Malin Feb 18 '19 at 22:26

2 Answers2

5

Let $p : I \to S^1$ denote the quotient map with $p(0) = p(1) = 1 \in S^1 \subset \mathbb{C}$ (you may take $p(t) = e^{2\pi i t}$). Given a loop $\gamma : I \to X$ at $x_0$, let $\gamma^* : S^1 \to X$ denote the induced map characterized by $\gamma^* \circ p = \gamma$. Let $\phi : D^2 \to X$ be an extension of $\gamma^*$.

Now $q : S^1 \times I \to D^2 \subset \mathbb{C}, q(z,t) = (1-t)z +t$ is a well-defined (note $\lvert q(z,t) \rvert \le 1$) continuous map. We have $q(z,0) = z$, $q(z,1) = 1$ for all $z$ and $q(1,t) = 1$ for all $t$.

Then $$h : I \times I \to X, h = \phi \circ q \circ (p \times id_I)$$ is a homotopy with the properties

1) $h(x,0) = \phi(q(p(x),0)) = \phi(p(x)) = \gamma^*(p(x)) = \gamma(x)$

2) $h(x,1) = \phi(q(p(x),1)) = \phi(1) = \gamma^*(1) = x_0$

3) $h(i,t) = \phi(q(p(i),t)) = \phi(q(1,t)) = \phi(1)= \gamma^*(1) = x_0$ (where $i = 0,1$)

This shows that the loop $\gamma$ is homotopic to the constant loop at $x_0$.

Edited on request:

The following are equivalent:

(1) $\gamma$ is homotopic to the constant loop.

(2) $\gamma^*$ is homotopic to the constant map $c_{x_0} : S^1 \to X, c_{x_0}(z) \equiv x_0$, via a homotopoy $h^* : S^1 \times I \to X$ such that $h^*(1,t) = x_0$ for all $t$ (i.e. via a basepoint-preserving homotopy).

(3) $\gamma^*$ is homotopic to a constant map $c_{x} : S^1 \to X$ for some $x \in X$.

(4) $\gamma^*$ has an extension to $D^2$.

We have already shown $(4) \Rightarrow (1)$.

$(1) \Rightarrow (2)$: Let $h : I \times I \to X$ be homotopy of paths from $\gamma$ to the constant path at $x_0$. The map $p \times id_I : I \times I \to S^1 \times I$ is a quotient map because the factor $I$ is locally compact. Hence $h$ induces a map $h^* : S^1 \times I \to X$ such that $h^* \circ (p \times id_I) = h$. This is the desired basepoint-preserving homotopy.

$(2) \Rightarrow (3)$: Nothing to show.

$(3) \Rightarrow (4)$: Let $H : S^1 \times I \to X$ be a homotopy from $\gamma^*$ to a constant map $c_{x}$. The map $$r : S^1 \times I \to D^2, r(z,t) = (1-t)z$$ is a continuous surjection. Since $S^1 \times I$ and $D^2$ are compact Hausdorff, $r$ is a closed map and therefore a quotient map. Define $$\phi : D^2 \to X, \phi(z) = \begin{cases} H(\frac{z}{\lvert z \rvert },1 - \lvert z \rvert) & z \ne 0 \\ x & z = 0 \end{cases} $$ Obviously $\phi(z) = H(z,0) = \gamma^*(z)$ for $z \in S^1$ and $\phi \circ r = H$ since for $z \in S^1$ we have $\lvert (1-t)z \rvert = 1-t$ and therefore $$\phi(r(z,t)) =\begin{cases} H(\frac{(1-t)z}{\lvert (1-t)z \rvert },1 - \lvert (1-t)z \rvert) = H(z,t) & t < 1 \\ \phi(0) = x = H(z,1) & t = 1 \end{cases} $$ But now $\phi \circ r = H$ implies that $\phi$ is continuous and we are done.

Paul Frost
  • 87,968
  • Thank you, this is very helpful. Does it also work the other way around? That is: Suppose $\pi_1(X,x_0)=0$ for all $x_0\in X$. Given a map $f:S^1\to X$ with $f(1)=x_1$, there exists a loop $\gamma:I\to X$ based at $x_1$ such that $\gamma=f\circ p$. By assumption, $\gamma$ is homotopic to the constant loop. How can I show $f$ is also homotopoic to a constant map? –  Feb 18 '19 at 19:27
  • I edited my answer. – Paul Frost Feb 18 '19 at 23:02
0

It is known that given a function $h:X\to Y$ which is continuous and $h(x_0)=y_0$ we can define a function $h_*:\pi_1(X,x_0)\to\pi_1(Y,y_0)$ given by $h_*([\sigma])=[h\circ\sigma]$. Then $h_*$ is well defined and it is a group homomorphism.

Now to the exercise itself. Take $\gamma:I\to X$ to be a loop around $x_0$. We can define $p:I\to D^2$ by $p(t)=e^{2\pi it}$ and $f:S^1\to X$ by $f(e^{2\pi it})=\gamma(t)$. Then $f$ is well defined because $\gamma(0)=\gamma(1)$ and it is also continuous. Hence it can be extended to $F:D^2\to X$ which is also continuous. And now note that $\gamma=F\circ p$. Hence:

$[\gamma]=[F\circ p]=F_*([p])=[e_{x_0}]$

We use the fact that $D^2$ is simply connected so $[p]$ is the identity element of $\pi_1(D^2)$. As $F_*$ is a homomorphism it sends identity to identity. So we got $\gamma$ is homotopic to the constant loop $e_{x_0}$ as it belongs to the same equivalence class.

Mark
  • 43,582