Let $N(n,k)$ denote the moduli space of stable vector bundles of rank $n$ and degree $k$ over a compact Riemann surface $X$, and let $N_0(n,k)$ denote the moduli space where we fix rank $n$ and some fixed determinant bundle of degree $k$. We know that the determinant map $det: N(n,k)\rightarrow Pic^k(X)$ is a proper submersion with fibers isomorphic to $N_0(n,k)$.

In the paper 'The Yang-Mills equations over Riemann surfaces' by M. F. Atiyah and R. Bott, (Phil. Trans. R. Soc. Lond. A 308, 523-615 (19)) the authors prove the following:

Proposition 9.7. (page 578) For rational cohomology we have $H^*(N(n,k)) \simeq H^*(N_0(n,k))\otimes H^*(Pic^0(X)).$

Immediately after this, the authors mention the following :

"This proposition, which is equivalent to the statement that $\Gamma_n := H^1(X,\mathbb{Z}_n)$ acts trivially on the rational cohomology of $N_0(n,k)$, ..."

( $\Gamma_n$ actually corresponds to the $n$-torsion line bundles on $X$, and a line bundle $L\in \Gamma_n$ acts on $N_0(n,k)$ by sending $E\mapsto E\otimes L$.)

I want to understand the above equivalence , at least the direction why the statement implies the proposition. I believe that it is somehow related to the monodromy action of $\pi_1(Pic^k(X))\simeq H^1(X,\mathbb{Z})$ on the cohomology of fibers via the $det$ map I mentioned above. If this action is trivial, then one can deduce that the map on cohomologies induced by the inclusion of a fiber is surjective, which would imply the proposition thanks to Leray-Hirsch theorem. But the authors state that instead the action of $\Gamma_n=H^1(X,\mathbb{Z}_n)$ is trivial. Of course, applying UCT we have $$H^1(X,\mathbb{Z}_n)=H^1(X,\mathbb{Z})\otimes \mathbb{Z}_n \simeq \pi_1(Pic^k(X))\otimes \mathbb{Z}_n,$$ does this imply that the monodromy action of $\pi_1(Pic^k(X))$ factors through the action of $\Gamma_n$? Is this the reason?

Thanks in advance.


1 Answer 1


Things are actually simpler. View $\Gamma _n=H^1(X,\mathbb{Z}/n)$ as the group of line bundles $L\in \operatorname{Pic}^{0}(X) $ with $L^{{\tiny \otimes }n}=\mathscr{O}_X$. The map $N_0(n,k) \times \operatorname{Pic}^{0}(X) \rightarrow N(n,k)\ $ given by $\ (E,L)\mapsto E\otimes L\ $ identifies $N(n,k)$ to the quotient of $N_0(n,k) \times \operatorname{Pic}^{0}(X) $ by $\Gamma _n$. Thus $H^*(N(n,k))$ is the subgroup of $H^*(N_0(n,k))\otimes H^*(\operatorname{Pic}^{0}(X) )$ invariant by $\Gamma _n$. But since $\Gamma _n$ acts by translation on $\operatorname{Pic}^{0}(X) $, it acts trivially on cohomology; so $H^*(N(n,k))$ is isomorphic to the whole tensor product if and only if $\Gamma _n$ acts trivially on $H^*(N_0(n,k))$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.