2

Suppose $E\to S^2$ be a complex vector bundle. If $c_1(E)=0$, does it imply that $E$ is a trivial bundle? And why if so? This question is motivated from Audin, Damian: Morse Theory and Floer Homology

enter image description here

Y.H. Chan
  • 2,708
  • Maybe something like this: https://math.stackexchange.com/questions/764429/is-the-hopf-fibration-the-only-fibration-with-total-space-s3 is relevant, by taking the associated sphere bundle? – ThePuix Apr 10 '21 at 09:03

2 Answers2

3

Yes, complex bundles over $S^2$ are completely characterized by their first Chern class. This follows from the fact that $U(1) \rightarrow U(n)$ given by extending by the identity matrix induces an isomorphism on $\pi_1 = \mathbb{Z}$.

To see this implication, $[X,U(n)]$ classifies rank n complex bundles over $SX$, the suspension of $X$. When $X=S^1$ we have from above that $[S^1, U(n)]=\mathbb{Z}$, and it is not hard to see that this is exactly given by evaluating the first Chern class of the associated bundle over $S^2$, evaluated on the fundamental class.

Connor Malin
  • 12,077
  • Can you explain why the vector bundle in $[S^1,U(n)]=\mathbb Z$ corresponds to the first Chern class? I only know that the first Chern class characterises all line bundle over $S^2$. That is because $H^1(S^2, \mathcal C^)=H^2(S^2)$ from the sheaf exact sequence $0\to \mathbb Z\to \mathcal C\to \mathcal C^\to 0$. – Y.H. Chan Apr 10 '21 at 18:01
  • $[S^1 , U(1)] \cong [S^2, BU(1)] \cong H^2(S^2)$ by loops suspension adjunction and your observation about the first Chern class, from which the claim is true for $n=1$. Then you just use the stabilization map I mention in the first paragraph and the fact that Chern classes aren't affected by stabilization. – Connor Malin Apr 10 '21 at 19:58
3

This is basically the same as Connor's answer, but maybe exploiting the fact that $S^2$ is low-dimensional instead of the fact that $S^2$ is a suspension. Interpret the first Chern class $c_1 \in H^2(BU(n); \mathbb{Z})$ as a map $BU(n) \to K(\mathbb{Z},2)$. This map fits into a fiber sequence $$BSU(n) \to BU(n) \xrightarrow{c_1} K(\mathbb{Z},2).$$ So, if a vector bundle $E$ on $X$ has $c_1(E) = 0$, then its structure group can be reduced from $U(n)$ to $SU(n)$. But $BSU(n)$ is $3$-connected and $S^2$ is $2$-dimensional, so every map $S^2 \to BSU(n)$ is nullhomotopic, i.e., every $SU(n)$-bundle $E$ is trivial.

JHF
  • 11,714