3
$\begingroup$

Suppose that $X$ is a simplicial complex, and $f:X \rightarrow S^k$ a continuous map to a sphere. Is $f$ always homotopic to a simplicial map to the boundary of a $(n+1)$-simplex, $\partial \Delta^k$?

The simplicial approximation theorem implies that there is a barycentric subdivision of $X$, $X'$, and a simplicial map $g:X'\rightarrow \partial \Delta^k$ that is homotopic to $f$. The question is whether we can drop the barycentric subdivision in this particular case.

EDIT: this question originally asked "Does a nonzero cohomology imply existence of a simplicial map to a sphere?" Indeed, as Bertram Arnold pointed out, the complex projective plane has nonzero homology in degree 2, but has no non-null-homotopic maps to the 2-sphere.

$\endgroup$
8
  • 1
    $\begingroup$ Sorry, as you pointed out, $k$ was being used as an index and as a field. Now, k is supposed to be a field, and n is supposed to be the dimension. $\endgroup$
    – Quizzical
    Jun 11 at 5:06
  • 3
    $\begingroup$ Not sure why you're saying that, if $H^n(X,k)$ is nontrivial, then there exists a non-null-homotopic map $X\rightarrow S^n$. Let $C_2$ be the group with 2 elements, and let $X$ be an Eilenberg-Mac Lane $K(C_2,1)$, for example, infinite-dim'l real projective space. Then $H^1(X,k)$ is certainly nonzero for many fields $k$ (after all, it is the group cohomology of the cyclic group $C_2$), but by the (proven) Sullivan conjecture, every pointed map from $X$ to $S^1$ is null-homotopic. Am I missing something? $\endgroup$
    – A.S.
    Jun 11 at 5:40
  • 6
    $\begingroup$ @A.S.: Maps from $X$ to $S^1$ is the first integral cohomology of the infinite dimensional real projective space. Calculating this is somewhat easier than proving the Sullivan conjecture. $\endgroup$ Jun 11 at 7:15
  • 1
    $\begingroup$ @OscarRandal-Williams: Yes, bringing up the Sullivan conjecture in that example is squashing a fly with a truck. I brought it up anyway, because it is an argument that generalizes to other finite groups and higher values of $n$, which I hope might help the original poster to identify what might be going on with the result that the question described. $\endgroup$
    – A.S.
    Jun 11 at 8:15
  • 3
    $\begingroup$ Another source of counterexamples to the claimed result is obtained by noting that the image of $f^*$ inside $H^n(X,k)$ must be annihilated by all cohomology operations (and the resulting secondary and higher cohomology operations). This boils down to lifting the map $X\to K(\mathbb Z,n)$ along $S^n\to K(\mathbb Z^n)$ using obstruction theory. The relevant obstruction groups involve the homotopy groups of $S^n$, so this will be essentially impossible to do in a systematic fashion. $\endgroup$ Jun 11 at 8:44
9
$\begingroup$

Take $X = \partial \Delta^3$. Then homotopy classes of maps from $|X|$ to $S^2$ correspond to elements of $\pi_2(S^2)\cong \mathbb{Z}$. But there are only finitely many maps of simplicial complexes $\partial \Delta^3 \to \partial \Delta^3$. So not all of them can be represented without further subdivision.

$\endgroup$

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.