Bott and Tu introduce without giving much details a theorem relating the Euler class of a vector bundle to the singularities of a section as follows (pp 125):

This theorem can also be phrased in terms of vector bundles. $\pi: E \rightarrow M$ be an oriented rank k vector bundle over a manifold of dimension $k$ and $s$ a section of $E$ with a finite number of zeros. The multiplicity of a zero x of $s$ is defined to be the local degree of $x$ as a singularity of the section $s$ of the unit sphere bundle of E relative to some Riemannian structure on $E$. (This definition of the index is independent of the Riemannian structure because the local degree is a homotopy invariant.) Since the Euler class $e(E)$ of $E$ is a $k$-form on $M$, it is Poincare dual to $nP$, where $n = \int_M e(E)$ and $P$ is a point on $M$. Thus, we have the following:

Theorem 11.17. Let $\pi: E \rightarrow M$ be an oriented rank $k$ vector bundle over $q$ compact oriented manifold of dimension $k$. Let $s$ be a section of E with a finite number of zeros. The Euler class of $E$ is Poincare dual to zeros of $s$ counted with the appropriate multiplicities.


Is it not true that an integral on a point is zero, and the $k$-form should produce a zero integral when wedged product with forms? What is the meaning of "$n$" in $np$ then? I generally honestly do not understand what the paragraph is trying to convey. The "zeros" of $s$ are not even in the domain, what is the meaning of their local degree? I appreciate any clarification on the paragraph and theorem. (I perfectly understand the same theorem for sphere bundles)

  • $\begingroup$ Why do you say the zeroes of $s$ are not in the domain? We have a section $s : M \to E$, so its zeroes are elements of $M$. $\endgroup$ – Michael Albanese Sep 22 '17 at 11:59

By Poincaré duality, a closed $k$-form on a connected, oriented $k$-manifold $M$ corresponds to an element of $H_0(M)\cong\Bbb Z$. You can represent elements of this $\Bbb Z$ as integral multiples of any given point on $M$. In this case, the integer is the Euler number, which is $\int_M e(E)$. On the other hand, they're saying something more interesting: If you choose any smooth section $s$ of $E$ that is transverse to the zero-section, then its zero set (which will consist of points), counted with multiplicities, is homologous to $nP$ and is also Poincaré-dual to the Euler class. Indeed, this is how topologists and geometers think of the Euler class of a bundle (and analogously for Chern classes of complex vector bundles, etc.). (Moreover, the multiplicities are the local intersection numbers of $s$ with the zero-section of $E$.)


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.