Let $p\colon E\to B$ be a fibration with fibres simply connected and homotopy equivalent to a compact CW-complex. Must $p_*\colon H_3(E;\mathbb{Q})\to H_3(B;\mathbb{Q})$ be surjective?

COMMENTS. Yes if (EDIT) $B$ is simply connected, even if the fibre is not compact but just finite-dimensional. In general, finite-dimensionality is not enough: consider the homotopy fibre sequence $\mathbb{R}^3\setminus\mathbb{Z}^3\to T^3\setminus\mathrm{point}\to T^3$.

MOTIVATION. If $p_*\colon H_3(E;\mathbb{Q})\to H_3(B;\mathbb{Q})$ is surjective, then any bundle gerbe over $p\colon E\to B$ is rationally trivial, cf. M. Murray, D. Stevenson, A note on bundle gerbes and infinite-dimensionality (http://arxiv.org/abs/1007.4922).

  • $\begingroup$ What exactly do you mean for "fibration"? A locally trivial bundle? A Serre fibration? $\endgroup$ – Daniele Zuddas May 3 '12 at 11:50
  • $\begingroup$ I mean Serre fibration. For locally trivial bundles, the question is equally interesting for me. $\endgroup$ – Semen Podkorytov May 3 '12 at 12:03
  • $\begingroup$ For locally trivial bundles, it is also natural to ask the fibre to be homeomorphic (not just homotopy equivalent) to a compact CW-complex. $\endgroup$ – Semen Podkorytov May 3 '12 at 13:37
  • 1
    $\begingroup$ @Semen: I tried to prove your claim in the special case when $\pi_1(B) $acts trivially. Are you implicitly using a fact from rational homotopy theory? Namely, the result that appears here: mathoverflow.net/questions/81139/homology-of-loop-space ? $\endgroup$ – John Klein May 4 '12 at 21:54
  • $\begingroup$ @John: I found a mistake in my proof for trivial $\pi_1$-action (I used only the Serre spectral sequence, no rational homotopy theory). If $B$ is simply connected, the claim follows from Theorem 2 in Gotay-Lashof-Sniatycki-Weinstein "Closed forms on symplectic fibre bundles" (pims.math.ca/~gotay/GLSW.pdf). $\endgroup$ – Semen Podkorytov May 5 '12 at 13:00

I'm wondering the extent to which the assumptions can be tweaked. Let's assume $B$ is connected and with basepoint. Let $F$ be the fiber over the basepoint.

However, I won't assume $F$ is homotopy finite (i.e., homotopy equivalent to a finite complex). Nor will I assume anything about the action of $\pi_1(B)$. Rather, I will assume

  • $F$ is $1$-connected (just as Semen does), and

  • $H_2(F;\Bbb Q)$ is trivial.

Assertion: With respect to these assumptions, $H_3(E;\Bbb Q) \to H_3(B;\Bbb Q)$ is surjective.

Proof: By slight abuse of notation, let $E/F$ be the the mapping cone of the inclusion $F\to E$.

Then the Blakers-Massey theorem shows that $E/F \to B$ is 3-connected.

We infer that $E\to B$ is $H_3({-};\Bbb Q)$-surjective if $E \to E/F$ is.

But the long exact homology sequence of $F \to E \to E/F$ and the assumption that $H_2(F;\Bbb Q)$ is trivial implies $H_3(E;\Bbb Q) \to H_3(E/F;\Bbb Q)$ is surjective. $\square$

The above leads to the following question: Is there a relationship between the hypotheses

(1) $F$ is homotopy finite, simply connected and $\pi_1(B)$ acts trivially on $H_*(F;\Bbb Q)$;

(2) $F$ is simply connected and $H_2(F;\Bbb Q)$ is trivial


  • 1
    $\begingroup$ Your result seems to be clear if you look at the Serre-Leray spectral sequence of the fibration. $\endgroup$ – Angelo May 3 '12 at 14:45
  • 1
    $\begingroup$ Even with a non-trivial $\pi_1$-action? In any case, I tend to be a luddite in these matters: I don't use spectral sequences when I can see the result without using them. $\endgroup$ – John Klein May 3 '12 at 14:49
  • 1
    $\begingroup$ (I'm not surprised though: since the Blakers-Massey theorem is pretty closely related to the Serre exact sequence of the fibration.) $\endgroup$ – John Klein May 3 '12 at 14:52
  • 1
    $\begingroup$ The spectral sequence gives something better: namely, the requested surjectivity holds if and only if the map from the coinvariants in $H_2(F,\mathbb Q)$ to $H_2(E, \mathbb Q)$ is injective. $\endgroup$ – Angelo May 3 '12 at 16:41
  • 1
    $\begingroup$ Angelo: your observation ("something better") can be also seen without the spectral sequence. It uses the observation that there is a fiber sequence ΣF∧ΩB→E/F→B such that the composite ΣF∧ΩB→E/F→ΣF is the Hopf construction of the action of ΩB on F. I had decided not to post this in my original answer, because I wanted to keep things simple. $\endgroup$ – John Klein May 3 '12 at 18:17

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.