For a real vector bundle $E$ of rank 3 on the base manifold $M$, the obstruction to having a $Spin^c$ structure on this vector bundle is given by the integral Stiefel-Whitney classes $W_3 = \beta w_2(E)$, where $\beta$ is the Bockstein homomorphism.

Are there any explicit examples of real vector bundles $E$ of rank 3 on 3-manifolds and on 4-manifolds for which the obstruction class $W_3$ is non-vanishing?

In physics terms, I am asking in what situation an $SO(3)$ gauge field on 3-manifolds or 4-manifolds cannot be lifted to a $U(2) = Spin^c(3)$ gauge field?


If $M$ is an orientable 3-manifold then any real vector bundle $E\rightarrow M$ of rank $\geq 2$ admits a $Spin^c$-structure. This follows since duality shows $H^3(M)$ to be torsion free giving $\beta \omega_2=0$.

Now consider $M=S^1\times \mathbb{R}P^2$. The Kunneth theorem shows that $H^3(S^1\times \mathbb{R}P^2)\cong \mathbb{Z}_2$, generated by an element $\beta (x)$ where $x\in H^2(S^1\times \mathbb{R}P^2;\mathbb{Z}_2)$ and $\beta:H^2(S^1\times \mathbb{R}P^2;\mathbb{Z}_2)\rightarrow H^3(S^1\times \mathbb{R}P^2)$ is the Bockstein connecting map. Moreover obstruction theory show that $x$ classifies some rank 3 real vector bundle $E\rightarrow S^1\times\mathbb{R}P^2$ in that $\omega_2(E)=x$. In particular $\beta(x)\neq 0$ and $E$ does not admit a $Spin^c$-structure.

Similarly if $M$ is an orientable $4$-manifold then it is well known that its tangent bundle admits a $Spin^c$-structure. See for instance the note "All 4-Manifolds Have $Spin^c$-Structures" by Teichner and Vogt (available at http://people.mpim-bonn.mpg.de/teichner/Papers/spin.pdf). It is not true for non-orientable 4-manifolds, however, and the the tangent bundle of of $\mathbb{R}P^2\times\mathbb{R}P^2$, given as an example in the cited note, is easily verified to admit no $Spin^c$-structure.

For an example of an arbitrary rank $3$ real vector bundle over an orientable $4$-manifold that does not admit a $Spin^c$-structure simply note that the vector bundle $E\rightarrow S^1\times \mathbb{R}P^2$ given as my previous example extends to a vector bundle $E'\rightarrow S^1\times \mathbb{R}P^3$ which by the same reasoning admits no $Spin^c$-structure.


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.