Suppose $X$ is a contractible space and $A \subset X$ with $A$ being homotopy equivalent to $S^2$. Prove that there is no retraction $X \rightarrow A$

Suppose there is a retraction $r: X \rightarrow A$. Then $ri = Id_A$.

Let's look at the homomorphism induced on the second homology.

$H_2(A)\stackrel{i_*}{\rightarrow}H_2(X) \stackrel{r_*}{\rightarrow} H_2(A)$

Since $A \cong S^2$ we have $H_2(A) = \mathbb{Z}$. Furthermore, $ri_*=Id_{A_*}$.

But $X$ is contractible so $H_2(X)=0$ and thus $ri_*$ factors through zero and thus cannot be the identity on $\mathbb{Z}=H_2(A)$ and thus there is no retraction $r: X \rightarrow A$

  • 1
    $\begingroup$ What you are asking for is clear from the tags, but it wouldn't hurt to include an actual question in the post. $\endgroup$ – qualcuno Jun 7 '19 at 6:55

Spot on. More generally, if $X$ is contractible and $A \subset X$ is not acyclic then $A$ cannot be a retract of $X$. The argument is the same as yours: pick $q \in \mathbb{N}$ such that $0 \neq H_q(A)$. If we were to have a retraction $r : X \to A$, then

$$ 1_{H_q(A)} = H_q(1_A) = H_q(ri) = H_q(r)H_q(i) \tag{1} $$

Since $X$ is contractible, $H_qr$ is the zero map (as it has domain zero) and this together with $(1)$ implies a contradiction: since $H_q(A) \neq 0$, it can't be that $1_{H_n(A)} = 0$. Hence no such retraction exists.

| cite | improve this answer | |
  • $\begingroup$ Even more generally, a retract of a contractible space must be contractible. $\endgroup$ – Connor Malin Jun 7 '19 at 8:02
  • $\begingroup$ What does the notation $H_q(r)$ mean exactly?? The homomorphism induced by $r$, from $(H_q(X)) \rightarrow H_q(A)?$ $\endgroup$ – HaKuNa MaTaTa Jun 8 '19 at 22:28
  • $\begingroup$ Precisely! Recall that homology at degree $q$ can be seen as a functor: to each space $X$ we assign an abelian group $H_q(X)$, and to each continuous function $f : X \to Y$ we assign a group morphism $H_qf : H_q(X) \to H_q(Y)$. Sometimes we note $f_*$ in the case of homology or homotopy groups, but the former notation is standard for any functor. $\endgroup$ – qualcuno Jun 8 '19 at 22:52

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.