MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

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

Atiyah Patodi and Singer [Spectral asymmetry and Riemannian geometry III] write that if $E$ is a complex flat bundle (non holomorphic, just smooth and complex) on a compact manifold $X$ (more generally a CW-complex) there must be some positive integer $m$ such that the direct sum $E\oplus E\oplus\cdots\oplus E$ ($m$ times) is a trivial bundle. More precisely, they write on top of page 19, "The vector bundle $V_a$ defined by $a$ is flat so its real Chern classes vanish, hence some multiple $kV_a$ is (unitarily) trivial".

Of course since the Chern character is an isomorphism from $K^\bullet(X)\otimes \mathbb{Q}$ to $H^{2\bullet}(X;\mathbb{Q})$, we know the bundle $E$ (or, more precely its K-theory class) is torsion in topological K-theory but this only tells us that $E\oplus E\oplus\cdots\oplus E$ is stably trivial.

I checked an impressive amount of literature on characteristic classes without finding a clue. Several authors cite directly Atiyah who, however, does not prove the claim.

Other authors say that the above statement cannot be true.

Does someone know the answer ?

share|cite|improve this question
Which authors say that this statement is not true? – Johannes Ebert Feb 23 '12 at 9:04

As you already said, if $E$ is flat of rank $r$, then $ch(E)=0$, meaning that the class $[E] \in K^0 (X)$ is torsion and therefore $m[E]=0 \in K^0(X)$ for some $m$. Pick $m$ such that $mr >> dim X$.

Now the stabilization map $BU(n) \to BU$ is roughly $2n$-connected. This means that if $V \to X$ a vector bundle of sufficiently high rank then $V$ is trivial iff it is stably trivial. Apply this observation to $E^{ \oplus m}$. Done.

share|cite|improve this answer
Sure the stable rank Theorem. Thank you. – Paolo Antonini Feb 23 '12 at 9:17
Is there any down-to-earth proof of this fact? It is so innocent-sounding... – diverietti Mar 9 '12 at 16:51
That $BU(n) \to BU$ is highly connected follows from the long exact homotopy sequence of $S^{2n-1} \to BU(n-1) \to BU(n)$. The stable cancellation follows by obstruction theory. Alternatively, if you are talking about smooth vector bundles on smooth manifolds, you can use general position arguments, like the following. If $M^m$ is a manifold, $V$ and $W$ vector bundles on $M$ of rank $r$ and $s$. If $s$ is sufficiently large w.r.t. $m$ and $r$, then any bundle map $f:V \to W$ in general position is injective (use transversality). Now if $V^r \oplus R^i \cong R^{r+i}$, you have ... – Johannes Ebert Mar 9 '12 at 18:58
two bundle monomorphisms $R^i\to R^{r+i}$, one with $V$ as complement, the other one with trivial complement. Taking any homotopy between them and put it into general position. Hence the two bundle mono's are isotopic; hence their complements, i.e. $V$ and $R^r$ are isomorphic. When done carefully, this argument gives the same bound on the dimensions as the obstruction theory argument. – Johannes Ebert Mar 9 '12 at 19:00
Thanks. Indeed is sufficient to take $n$ to be greater tian the stable ranking of the base – Paolo Antonini Jun 25 '14 at 20:15

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.