2
$\begingroup$

I believe my question should be simple. The question is more physically oriented and originated from one of Witten's papers, "On Holomorphic Factorization of WZW and Coset Models", where he considered a continuous map, $$\phi: \Sigma\rightarrow M\,,$$ where $\Sigma$ is a two-dimensional closed surface and $M$ is an arbitrary closed manifold. Witten claimed that, if $$\pi_1(M)=\pi_2(M)=0\,,$$ then the map $\phi$ will be automatically nullhomotopic. Denote the set $\,\mathcal{C}=\{\phi:\Sigma\rightarrow M\}$ as all continuous maps from $\Sigma$ to $M$.

Now my question is simply that if either $\pi_1(M)$ or $\pi_2(M)$ are non-trivial. How can one argue that there would be maps in $\mathcal{C}$ NOT homotopic to identity. More generally, if assume $\Sigma$ is a 2-sphere, can one figure out the homotopy of the configuration space $\,\mathcal{C}$, say $\pi_0(\mathcal{C})$, $\pi_1(\mathcal{C})$ and $\pi_2(\mathcal{C})$ etc.

Thanks in advance!

$\endgroup$
4
  • 1
    $\begingroup$ "[...] maps in $\mathcal{C}$ NOT homotopic to identity" did you mean not homotopic to the zero/constant map? Identity doesn't make much sense there. $\endgroup$ Jul 7 '16 at 10:15
  • 1
    $\begingroup$ if $\Sigma=S^2$, then $\mathcal{C}$ is just the double-loopspace $\Omega^2 M$, so $\pi_n(\mathcal{C})=\pi_{n-2}(M)$ $\endgroup$ Jul 7 '16 at 14:46
  • 2
    $\begingroup$ sorry, i meant $\pi_n(\mathcal{C})=\pi_{n+2}(M)$ $\endgroup$ Jul 7 '16 at 19:02
  • $\begingroup$ @BalarkaSen Yes, I should have corrected it as "homotopic to constant map". I have an additional question posted in your answer. Would you please have a check? Thanks! $\endgroup$
    – Loliphilia
    Jul 11 '16 at 11:21
5
$\begingroup$

If we know that either $\pi_1(M)$ or $\pi_2(M)$ is nontrivial, indeed there does exist a homotopically nontrivial map $\Sigma \to M$ for any surface $\Sigma \neq S^2, \Bbb{RP}^2$.

If $\pi_1(M) \neq 0$, then choose a representative $f : S^1 \to M$ of a nontrivial homotopy class $[f]$. Let $\Sigma$ be some surface, and let $\Sigma \to T^2$ be the map given by pinching complement of a punctured torus in the connected sum decomposition of $\Sigma$. Then compose with the projection map $T^2 \to S^1$ to get a map $\Sigma \to S^1$. This is clearly nontrivial on $\pi_1$. Consider then the composition $\Sigma \to S^1 \to M$, which is our desired non-nullhomotopic map, as it is nonzero on $\pi_1$.

(Note that the technique doesn't work if $\Sigma$ doesn't have a torus in the connected sum decomposition, i.e., if $\Sigma \cong S^2$. In that case, indeed, $\pi_1 \neq 0$ is not sufficient for there to exit a non-null map from $S^2$: take, e.g., $M = S^1$. For nonorientable $\Sigma$, same argument applies by finding a Klein bottle component in $\Sigma$ instead of a torus, again with the exception of $\Bbb{RP}^2$ - if there was a homotopically nontrivial map $\Bbb{RP}^2 \to S^1$, there would have been a nontrivial homomorphism $\Bbb Z/2 \to \Bbb Z$ at the level of fundamental groups, which is nonsense)

If $\pi_1(M) = 0$ but $\pi_2(M) \neq 0$, consider a representative of some nonzero homotopy class $[f]$ in $\pi_2(M)$, which gives you a non-nullhomotopic map $f : S^2 \to M$. Moreover, since $\pi_1(M) = 0$, under the Hurewicz isomorphism $\pi_2(M) \to H_2(M)$, $[f]$ is sent to a nontrivial homology class $[f^*(S^2)]$ in $M$, hence $f$ is nontrivial in homology. If $\Sigma$ is some higher genus surface, let $\Sigma \to S^2$ be the degree $1$ map given by pinching a complement of a small ball inside a chart in $\Sigma$. The composition $\Sigma \to S^2 \to M$ is non-nullhomotopic, because it's nonzero in $H_2$ (as so are the two maps in the composition). If $\Sigma$ is nonorientable, exact same argument works, except you have to work in homology mod 2.

$\endgroup$
3
  • 1
    $\begingroup$ Just to add to the picture: If $\Sigma=S^2$ then each map $\Sigma\to M$ is null-homotopic iff $\pi_2(M)=0$. If $\Sigma=RP^2$ and $\pi_1(M)$ has nontrivial 2-torsion and $\pi_2(M)=0$ then there is a homotopically nontrivial map $\Sigma\to M$. The situation when $\pi_2(M)\ne 0$ is more complicated. $\endgroup$ Jul 8 '16 at 16:26
  • $\begingroup$ @BalarkaSen Thanks for your answer and it's very instructive! In addition, if there exists non nullhomotopic maps in $\mathcal{C}$, does it represent a generator in certain homotopy groups of $\mathcal{C}$ $\endgroup$
    – Loliphilia
    Jul 11 '16 at 11:21
  • $\begingroup$ @Loliphilia Sure, it represents a nontrivial element in $\pi_0(\mathcal{C})$. Elements of $\pi_0(\mathcal{C})$ are in one-to-one correspondence with homotopy classes of elements in $\mathcal{C}$. $\endgroup$ Jul 11 '16 at 15:01

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.