4
$\begingroup$

Let $E$ be a complex vector bundle whose structure group $G$ is a product: $G=G_1\times G_2$. If the fiber of $E$ is $n$-dimensional, then are the Chern classes of $E$ related to the Chern classes of vectors bundles with structure groups $G_1$ and $G_2$, also with $n$-dimensional fibers and over the same base?

In case it makes a difference, I'm especially interested in the case $G_1=U(1)$ and $G_2=SU(N)$, but the more general the better.

I'm aware of the Whitney sum formula (which expresses the Chern classes of a Whitney sum $E_1\oplus E_2$ of two vector bundles in terms of the Chern classes of the individual vector bundles $E_1$ and $E_2$), and I'm aware that the Whitney sum of $G_1$ and $G_2$ principal bundles is a $G_1\times G_2$ principal bundle [1], but I don't know if these observations have anything to say about my question.


Reference:

  1. Page 22 in Cohen (1998), The Topology of Fiber Bundles Lecture Notes (http://math.stanford.edu/~ralph/fiber.pdf)
$\endgroup$

1 Answer 1

1
$\begingroup$

If $E \to X$ is a complex vector bundle with structure group $G=U(1)\times SU(n-1) \subset U(n)$ there exists complex line bundle $E_1$ (with structure group $U(1)$) and a complex vector bundle $E_2$ with structure group $SU(n-1)$ such that $E$ is isomorphic to $E_1 \oplus E_2$.

This makes the Chern classes of $E$ computable in the terms of chern classes of $E_1$ and $E_2$.

For the statement, consider the $G$ principle bundle $P\to X$ associated to $E$. The proection \begin{align*} pr_1:G \to U(1) \end{align*} induces a $U(1)$ principle bundle $P_1 := (P \times U(1))/G$. Where the right action of $G$ on $P\times U(1)$ is given by \begin{align*} (p,z)\cdot g:= (p\cdot g, pr_1(g)^{-1}\cdot z). \end{align*}

Analogously $pr_2:G \to SU(n-1)$ defines a $SU(n-1)$ principle bundle $P_2$. There is an ismorphism of $G$ principle bundles $P \overset{\sim}{\to} P_1 \times P_2$. Take $E_1,E_2$ to be the associated vector bundles respectively.

This corresponds to a map of classifying spaces $B(G_1\times G_2) \to BG_1 \times BG_2$ which is homotopy inverse to $BG_1\times BG_2 \to B(G_1 \times G_2)$.

$\endgroup$
4
  • $\begingroup$ Thank you for the answer! I'll need to spend some time processing it to make sure I understand everything. $\endgroup$ Commented Oct 13, 2021 at 0:11
  • $\begingroup$ If I understand correctly, you've demonstrated the existence of vector bundles $E_1$ and $E_2$ which satisfy the conditions I described, and whose Chern classes can be used to compute the Chern classes of $E$ though the Whitney sum formula, correct? And pardon my ignorance, but are the associated bundles $E_1$ and $E_2$ uniquely determined by the principal-bundle part of what you described? Its fine if they're not, just want to be sure I understand. $\endgroup$ Commented Oct 14, 2021 at 1:26
  • 1
    $\begingroup$ @ChiralAnomaly Yes that is correct. I am not sure if I understand your second question. Starting with your vector bundle $E$ with structure group $G$ one has explicit constructions for $P$, $P_1$,$P_2$, $E_1$, $E_2$, in that sense they are uniquely determined by these constructions, but I am not sure if that is what you mean. Are you more familiar with transition functions (cocycles) for vector bundles than with principal bundles? $\endgroup$
    – Jonas
    Commented Oct 14, 2021 at 7:57
  • $\begingroup$ Yes, that's what I meant by my second question in the comment, and your reply clears it up for me. My background is physics, and my understanding of fiber bundles is still pretty shallow -- enough to appreciate the definitions, but not enough to stand on my own yet, so everything you wrote is helpful. Thanks again! $\endgroup$ Commented Oct 14, 2021 at 14:18

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .