The following question is somewhat similar to a previous one on MathOverflow, except that my application does not directly involve Eilenberg-MacLane spaces $K(G,n)$, and so I don't see the immediate need for $n$-simple maps as in the theory of Postnikov $k$-invariants.

Suppose that $\Omega T \longrightarrow F \longrightarrow E$ is a homotopy fibration sequence, where $T$ and $F$ are connected finite CW complexes. Is there a quantification in some sort of obstruction theory for extending this sequence further to the right in the form $F \longrightarrow E \longrightarrow T$ ? This is a 1950s-era topology question.

Of course, the original sequence extends further to the right as $F \longrightarrow E \longrightarrow B hAut(\Omega T)$. First, this is by Stasheff's paper for finite CW complexes as the fiber and any CW base, working with products in the full topological category [Sta63]. Later, this is by May's simplicial upgrade to infinite CW complexes as the fiber, working in the compactly generated category [9.5, 9.8][May75].

In my application, $T$ happens to have the structure of a Lie group, but this is a happy accident of its dimension, and I'm hoping for an answer that can work independent of this fact.

[Sta63]: James D Stasheff, A classification theorem for fibre spaces, Topology 2, 1963

[May75]: J Peter May, Classifying spaces and fibrations, Memoirs AMS 1:155, 1975

EDIT: As suggested, I renamed the base space $B$ to $T$, in order to not confuse it with the classifying space functor. Also, I added two citations.

  • $\begingroup$ (I rename $B$ to $T$ to avoid confusion) I think the only obstruction is factorizing action map $E \to BAut(\Omega T)$ through delooping of tautological action $T \to BAut(\Omega T)$ which is done by usual obstruction theory. $\endgroup$ – Denis T. Oct 22 '19 at 21:50
  • $\begingroup$ @Denis T. : Любезно, what is the explicit formula for this tautological action? Does it it assume that $T$ is a topological group, say by pointwise-conjugating a loop by an element of $T$? If so, isn't this instead a map $T \longrightarrow Aut(\Omega T)$? $\endgroup$ – Qayum Khan Oct 22 '19 at 22:49
  • $\begingroup$ @QayumKhan The tautological action is the one corresponding to the pathspace fibration $\Omega T\to P T\to T$ (morally it is the action of $\Omega T$ on itself by left(?) multiplication) $\endgroup$ – Denis Nardin Oct 23 '19 at 7:48
  • $\begingroup$ @QayumKhan DenisNardin is right, I mean just an application of classifying space functor to map $\Omega T \to Aut(\Omega T)$ representing left (more precisely, the side $\pi_1$ of base acts on fiber in your preferrable conventions) multiplication. Also possibly you want to use Moore loops for that to avoid some nuances with non-strictly associative actions etc. $\endgroup$ – Denis T. Oct 23 '19 at 16:55

This question is addressed in the paper

Ganea, T., Induced fibrations and cofibrations, Trans. Am. Math. Soc. 127, 442-459 (1967). ZBL0149.40901.

A first observation is that $\Omega T\to F\to E$ extends to the right if and only if it is induced from the based path fibration $\Omega T\to PT\to T$ by a map $p:E\to T$. In Section 2, various sufficient conditions for such a fibration to be induced are given.

Two sample results:

Corollary 2.5: Suppose that $\pi_q(E)\neq 0$ only if $m\le q\le n + m-1$ and that $\pi_q(\Omega T) \neq 0$ only if $n\le q\le n + m-1$, where $n\ge m\ge 2$. If the Whitehead product pairing $W:\pi_n(\Omega T) \otimes \pi_m(F)\to \pi_{n + m-1}(F)$ vanishes, then $\Omega T\to F\to E$ is induced.

Theorem 2.10: Suppose $\Omega T$, $F$ and $E$ all have the homotopy type of aspherical CW complexes. Then $\Omega T\to F\to E$ is induced if and only if the image of the induced map $\pi_1(\Omega T)\to \pi_1(F)$ lies in the center.

| cite | improve this answer | |
  • $\begingroup$ I would be very interested in knowing the level of generality in which these results hold. Do they work, for instance, in any $\infty$-topos? $\endgroup$ – skd Oct 28 '19 at 0:26
  • $\begingroup$ @MarkGrant : I appreciate the partial answer; however, my specific application (which I didn't specify in the question) fails to satisfy the connectivity hypotheses of your answer here and also the previously linked MathOverflow question. Specifically, all of my finite CW complexes $T, F, E$ are non-simply connected and have homotopy groups in infinitely many dimensions. My opinion is that the results in Ganea's paper are more-or-less a 'thickening' of the preexisting result on Eilenberg--MacLane spaces that he advances. $\endgroup$ – Qayum Khan Oct 28 '19 at 17:35

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.