13
$\begingroup$

The following seems like an extremely basic algebraic topology question, but it's not something I ever learned, nor does it look familiar to the algebraic topologists I've asked.

Let $f:X\to Y$ be a map, inducing $f^*:H^*(Y)\to H^*(X)$. Hence the image $R$ of $f^*$ is a subring of $H^*(X)$. Is there a natural way to factor $f$ as $X \to Z \to Y$, such that $f^*$ factors as $H^*(Y) \twoheadrightarrow R \hookrightarrow H^*(X)?$

I have a particular $X,Y$ in mind (the inclusion of one compact complex manifold into another, each with even-degree cohomology) but I'm hoping phrases like "Postnikov tower", "cofibrant replacement", "mapping cone" will serve to give a general answer.

Improve this question
$\endgroup$
3
  • 1
    $\begingroup$ I wonder if you’d have more luck asking for some kind of image factorization for rational spaces, which presumably could be connected to some kind of image factorization for Sullivan models. $\endgroup$
    – Qiaochu Yuan
    Sep 25, 2019 at 2:39
  • 2
    $\begingroup$ I have the suspicion that this is only true for disjoint unions of Eilenberg-MacLane spaces (as the codomain). Or perhaps pairs of spaces where cohomology detects the difference between maps (as is true for any pair with codomain an Eilenberg-MacLane space). $\endgroup$
    – Connor Malin
    Sep 25, 2019 at 2:54
  • 2
    $\begingroup$ I guess you mean $R=f^*(Z)$? $\endgroup$
    – user43326
    Sep 25, 2019 at 5:55

2 Answers 2

Reset to default
25
$\begingroup$

No. Consider the Hopf map $\eta:S^3\to S^2$. If there were such a space $Z$, it would have $\widetilde H^*(Z)=0$, so at the very least $Z$ would be stably trivial, forcing $\eta$ to be stably trivial; but it’s not.

Improve this answer
$\endgroup$
1
  • 2
    $\begingroup$ Well that was straightforward! I'm hoping someone has a more positive answer (i.e. extra conditions to require to make it possible), so I won't accept this quite yet, but it's certainly a definitive answer to the question as stated. $\endgroup$
    – Allen Knutson
    Sep 25, 2019 at 0:05
10
$\begingroup$

One special case of your set up is when $Y=X$ and $f^*$ is idempotent: $f^* \circ f^* = f^*$. In this case, let $Z$ be the mapping telescope of $X \xrightarrow{f} X \xrightarrow{f} X \rightarrow \dots$. This comes with a canonical map $r: X \rightarrow Z$ such that $r^*$ is monic with image equal to the image of $f^*$, and in many cases, one can show that there exists $i: Z \rightarrow X$ such that $r^* \circ i^* = f^*$.

(We are basically looking to lift an idempotent in homology to an idempotent in homotopy. One sufficient condition for $i$ to exist is that $X$ be a $p$--complete CW complex of finite type: see the short paper Atomic spaces and spectra I wrote with J.F. Adams back in the late 1980's.)

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.