I am reading Algebraic Topology by E.H.Spanier and in the proof of the Thom-Gysin map for disc bundles (on page 260) he says that $p : E \to B $ is a deformation retraction. I do not understand how this is the case. How do we view $B$ as a subspace of $E$ in the first place ? And then how does $p$ become a deformation retraction ?

Also please advise some reference where I could learn basic properties of disc/sphere bundles. Thanks.

Edit : Here is the statement of the assumption part of Theorem 5.7.11 (in whose proof the statement appears) : Let $(\xi,U_\xi)$ be an oriented q-sphere bundle with base B and projection $\dot{p}=p|_\dot{E}:\dot{E} \to B$. Here $(E,\dot{E})$ is a fiberbundle pair with fiber $(D^{n+1},S^n)$ and $p: E \to B $ is the projection map.

  • 2
    $\begingroup$ Isn't $S^n \to \{\ast\}$ a counterexample? $\endgroup$ – Michael Albanese Dec 4 '15 at 17:08
  • 1
    $\begingroup$ Can you include the relevant portions of the text for completeness? I'm not going to go find a copy of Spanier. (This should include both the relevant assertion and the definitions of $E$ and $B$.) $\endgroup$ – user98602 Dec 4 '15 at 17:09
  • $\begingroup$ @MikeMiller Sorry. I am going to edit the question in the light of your comment. Unfortunately Spanier does not give the definition of a sphere bundle, so I assume he is referring to a fiber bundle whose fiber is $S^n$. One can assume that the bundle is orientable. $\endgroup$ – user90041 Dec 4 '15 at 17:12
  • $\begingroup$ @MikeMiller Edited the question now. I think the definition is that $(E, \dot{E})$ is a q-sphere if the fiber pair is $(D^{n+1},S^n)$ . $\endgroup$ – user90041 Dec 4 '15 at 17:20
  • 1
    $\begingroup$ It might be that he demands that the transition functions of your disc bundle are linear, in which case you have a well-defined zero section. If not, to show that the projection is a deformation retraction (well, to make sense of that), you would need to start by showing it has a section, and presumably you'd want one whose image is in the interior of each fiber. $\endgroup$ – user98602 Dec 4 '15 at 17:32

First, I suspect Spanier wants all of his sphere and disc bundles to have linear transition maps, hence there is a canonical zero section $B \hookrightarrow E$. If not, then to make sense of his claim, you need a section; obstruction theory + the fact that $B$ is contractible guarantees that one exists, and indeed you can force it to be in the interior of each fiber.

Once you have a section $s$, note that both $p$ and $s$ are homotopy equivalences by Whitehead + the long exact sequence of homotopy groups of a fibration. Now recall one version of the Whitehead theorem: if $X \hookrightarrow Y$ is a cofibration and also a homotopy equivalence, there is a deformation retraction onto $X$. I would guess that any section of a disc bundle is a cofibration, but if not, pick your section above to be a good one. In any case, once you have this, Whitehead gives you a deformation retraction onto the image of the section, as desired.

| cite | improve this answer | |
  • $\begingroup$ Thank you so much. I understand the first para of your answer and it is really helpful. I too think that Spanier must be assuming the transition maps to be linear (otherwise it is difficult to understand the proof without obstruction theory). $\endgroup$ – user90041 Dec 6 '15 at 17:54
  • $\begingroup$ @Mike Miller I am still learning the area, but the proofs of characteristic classes and other obstructions I have seen all assume the existence of a section to start with (either by assuming linear transition functions or just implicitly assume without proof). How does one get around this bit of $S^1$ logic? $\endgroup$ – wanderingmathematician May 2 '18 at 15:52
  • $\begingroup$ @user334137 I don't really know what you mean. Applying obstruction theory to a fibration definitely does not require the existence of a global section; this is, for instance, a standard proof that a symplectic manifold has a compatible almost complex structure: prove that the space of compatible almost complex structures is the space of sections of a fiber bundle with contractible fibers. A good exposition of obstruction theory is in Davis & Kirk, and the proof involves building a section cell-by-cell. $\endgroup$ – user98602 May 2 '18 at 18:22
  • $\begingroup$ Actually, because $\pi_n D^k = 0$ for all $n > 0$, the obstruction cocycle is even zero - usually in obstruction theory you need to take a step back to take a step forward, and redefine your map over $X_n$ (without changing it over $X_{n-1}$) - but this isn't true here. You can always just extend your map to the next skeleton. You probably want to know the theorem that a fiber bundle over a contractible compact space is trivializable. $\endgroup$ – user98602 May 2 '18 at 18:23

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.