Higgs bundles, branes and Langlands duality (Q1729882)

The authors study the moduli space $\mathcal{M} (G)$ of $G$-Higgs bundles over a compact Riemann surface $X$ in the setting of the geometric Langlands programme. They assume that $X$ is equipped with an anti-holomorphic involution $\tau_X : X \to X$ (i.e., $X$ is a \textit{real Riemann surface}) and that $G$ is endowed with an anti-holomorphic group involution $\tau_G : G \to G$ (i.e., a real form). They use this structure to induce a pair of natural involutions on the moduli space $\mathcal{M} (G)$ whose fixed point sets are an $(A,A,B)$-brane and an $(A,B,A)$-brane. \par Then the question they ask is ``how the various components of $(A, A, B)$ and $(A, B, A)$ branes match up, as one passes from $G$ to ${}^L G$.'' One way this question can be posed is to ask how the real structure interacts with the dual fibres of the moduli spaces $\mathcal{M} (G)$ and $\mathcal{M} ({}^L G)$. The authors give a description of the intersection between a brane and a pair of dual fibres, and prove that these fibres have the same number of real components. \par Another way is to interpret this question globally on the full moduli space, where the situation turns out to be very different in general. Even in the case of $G = \mathrm{SL}(2, \mathbb{C})$ and ${}^L G = \mathrm{PGL}(2, \mathbb{C})$ (the only one considered in the paper), ``the real components are far from matching'': the fixed point locus in $\mathcal{M} ({}^L G)$ has a strictly greater number of connected components. Their technique to obtain this count is first to give a very explicit description of the effect that the involution has on the homology lattice of a given $\mathrm{SL}(2, \mathbb{C})$ spectral curve $S$ over $X$, and then use it to count the number of real components of the Prym variety for $S$.