MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I'm looking at the cotangent bundle of $CP^{N}$ at the moment in the context of equivariance. For many reasons, it seems to me that this bundle is not $U(1)$-equivariant, or, in other words, cannot be constructed as from a representation of $U(1)$ in the standard manner (see here for example) where we consider $CP^{N}$ as the base space of the principle bundle $S^{2N}$. However, I cannot find a neat convincing argument for why this should be so.

MY ATTEMPT: I would guess that $U(1)$-equivariance would mean that the bundle could be expressed as a direct sum of line bundles, but I cannot see how to show that this is not the case (even though it seems most probable). Again I would guess that some sort of Chern argument comes in, but I don't know how to caclulate Chern classes without a connection. The only connection here I know is the Grassmannian, and at this point the whole thing just becomes a mess ....

share|cite|improve this question
up vote 7 down vote accepted

If there were a representation $V$ of $U(1)$ and an isomorphism $TCP^n \cong S^{2n+1} \times_{U(1)} V$, then the tangent bundle of $S^{2n+1}$ would be the direct sum of the trivial bundle $S^{2n+1} \times V$ plus the trivial real line bundle (the vertical tangent bundle to the $S^1$-bundle. In particular, $S^{2n+1}$ is parallelizable, which implies, by a result by Adams, that $n=0,1$ or $3$. If $n=0$, then the question is void, for $n=1$, $TCP^1$ is obviously a complex line bundle.

That leaves the case of $CP^3$. If $TCP^3$ were a sum of three complex line bundles, then the total Chern class would be $(1+ax)(1+by)(1+cy)$ for some integers $a,b,c$ ($x$ is an appropriate generator of $H^2(CP^n)$. On the other hand, we know the total Chern class very well; it is $1+4x+6x^2+4 x^3$. This is discussed, without any reference to a connection, in many sources, e.g. Milnor-Stasheffs book. So you get

$$a+b+c=4, ab+bc+ac=6, abc=4.$$

It is easy that these equation do not have an integral solution (reduce modulo $2$ to conclude that $a,b,c$ have to be even).

Probably it is possible to do the higher dimensional cases in that arithmetical way, instead of nuking it with Adams' result.

share|cite|improve this answer
Great answer, thanks alot. Is Milnor-Stasheff's book a good place start on learning this stuff. It's not material that's central to my interests at the moment, so I'd like something that's not too heavy going - maybe even fun. – Ago Szekeres Nov 30 '10 at 20:36
Moreover, what is the dimension of the module of vector fields of $S^7$. – Ago Szekeres Nov 30 '10 at 21:08
One can deal with all the cases as you do for $n=3$. Just observe that $T_{P^n}$ is generated by global sections, since by the Euler sequence there is a surjection $\osum_0^n O_{P^n}(1)\to T_{P^n}$.So every line bundle that is direct summand of $T_{P^n}$ is also generated by global sections, hence it has total Chern class $1+ax$ with $a\ge 0$. Then it is easy to see that the analog of the equations given for $n=3$ have no solution. – rita Nov 30 '10 at 21:13
Having taught several seminars on that, I can tell you that Milnor-Stasheff is a rather demanding text compared to others (like Hatchers unfinished book "Vector bundles and K-theory"). I also like to advertise the definition of Chern classes that Matthias Kreck gives here: , page 119 – Johannes Ebert Nov 30 '10 at 21:20
@Ago: $S^7$ is parallelizable, so the dimension is $7$. – Johannes Ebert Nov 30 '10 at 21:22

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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