I'm having the following problem: let $T \subset G := SO(2k)$ be the maximal torus acting on $V := \mathbb{R}^{2k}$ by linear transformations on each $2$-dimensional component. Denote by $V_T := (V \times ET) / T$ the homotopy quotient, where $ET \to BT$ is the universal principal bundle associated with $T$. I would like to compute the Euler class of the (oriented) vector bundle (with fiber $V$) \begin{equation} \pi : V_T \to BT, \end{equation} in terms of the cohomology groups $H^*(BT)$.

So far, I have the following, mainly coming from Chern-Weil theory: we first consider the action of the whole group $G$ (we still denote by $\pi$ the vector bundle above for this action). The Chern-Weil homomorphism (which is an isomorphism here since $G$ is compact) provides an algebra isomorphism \begin{equation} \xi : S(g^*)^G \simeq H^*(BG), \end{equation} where $S(g^*)^G$ is the algebra of (say complex-valued if we look at cohomology with complex coefficients) $G$-invariant polynomial functions on the Lie algebra $g = so(2k)$. This isomorphism is given by applying polynomials on $g$ to the curvature $\Omega$ of any given connection form on the universal principal vector bundle $EG \to BG$, and one can show that the classes obtained this way are independent on the choices made. A particular element in $S(g^*)^G$ is called the Pfaffian $Pf$, and its image $Pf(\Omega)$ through $\xi$ is called an Euler form for $EG \to BG$. Now, choose a metric on the bundle $\pi$, and consider the associated principal bundle, that is the bundle $$\pi_F : F(V_G) := (F(V) \times EG) \to BG,$$ where $F(V)$ is the set of oriented orthonormal frames on V. Of course, $G$ acts freely on the total space $F(V_G)$, and one can see that it is also contractible, since it is homeomorphic to $(G \times EG) / G$. Therefore $\pi_F$ can be seen as the universal principal bundle of $G$, and the Euler form $Pf(\Omega)$ as an Euler form for $\pi_F$.

My questions are the following:

  1. How can I prove that this Euler form is indeed a representative of the Euler class of the vector bundle $\pi$ ?
  2. Is there a preferred curvature form for which the computation of $Pf(\Omega)$ would be simple ?
  3. Is there a more direct way (without passing through the Chern-Weil homomorphism) of computing the Euler class of $\pi$ ?
  4. Going back to the action of the torus $T$, how can I compute the Euler class of the vector bundle $\pi : V_T \to BT$ in terms of the cohomology groups $H^*(BT) \simeq \mathbb{C}[u_1,...,u_k]$ (I should find that it is equal to $u_1...u_k$, where $u_i$ are generators of degree $2$ of $H^*(BT)$) ?

My ultimate goal is to answer question 4, but it seems to me that understanding the above is necessary. Of course, if someone has a more straightforward way of treating question 4, it would be great.

I apologize in advance if I don't seem clear (I'm not a specialist of Chern-Weil theory).

Thanks a lot.


1 Answer 1


I write this as an answer because it is too long to post it as a comment.

You need to be careful when using the term connection on a vector bundle over $BG$ since $BG$ is not finite dimensional. Technically, you have to work with finite dimensional approximations for $BG$.

In any case you can find the answer to your question in Chapter 8 of this book.

More precisely, question 1 is answered in Thm. 8.3.17 of Section 8.3.2.

Question 4 is answered in Proposition 8.3.14 in Sect. 8.3.1.

The answer to question 2 is that in general there is no nice curvature that makes the computation of the Pfaffian simpler. On the other hand I refer to Appendix B of this paper for a particularly convenient way of defining the Pfaffian.

The answer to question 3 is yes, but it requires a different definition for the Euler class in terms of the Thom class. See for example chapter 4 of these lecture notes.

  • $\begingroup$ I cannot find the relation between my question 4 and proposition 8.3.14 of your book though. Could you be a bit more specific about how this gives the Euler class in terms of the generators of the cohomology of $BG$ ? $\endgroup$
    – BrianT
    May 17, 2018 at 20:01
  • $\begingroup$ If the bundle reduces to $BT^n$ it splits as a direct sum of oriented plane bundles $L_1, \dots, L_n$. and the Euler class of a sum is the product of the Euler classes of the splitting factors (this is prop 8.3.14 applied to $T^n=\underbrace{SO(2)\times\cdots\times SO(2)}_{n}$. The Euler classes of $L_i$ have degree $2$ and generate the cohomology ring of $BT^n$. $\endgroup$ May 17, 2018 at 20:49
  • $\begingroup$ Oh great, thank you very much for your help. $\endgroup$
    – BrianT
    May 17, 2018 at 22:32
  • $\begingroup$ I'm sorry, I'm going to bother you one last time. You said that if the bundle reduces to $BT^n$, it splits as a direct sum of oriented plane bundles $L_1,...,L_n$. 1- Are we taking about the bundle $V_{T^n} \to BT^n$ (this is the one for which I'm trying to compute the Euler class) ? 2- Could you tell me what result tells me that it splits as $L_1 \oplus ... \oplus L_n$, and what would be each $L_i$ ? Thank you again for all your explanations. $\endgroup$
    – BrianT
    May 17, 2018 at 23:40
  • $\begingroup$ Also, is it clear that the cohomology ring of $BG $ is generated by the Euler classes of the $L_i$ ? $\endgroup$
    – BrianT
    May 17, 2018 at 23:47

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.