Semistable and numerically effective principal (Higgs) bundles (Q626124)

The first part of the paper gives a Miyaoka-type criterion for principal \(G\)-Higgs bundles (\(G\) a reductive group) on a complex manifold \(X\) while the second part deals with numerical flatness and numerical effectiveness of \(G\) (Higgs) bundles. Let \(E\) denote a principal \(G\)-Higgs bundle on \(X\) with a Higgs field \(\phi\). Given a parabolic subgroup \(P \subset G\), the authors introduce a subscheme \(R_P(E,\phi) \subset E/P\) consisting of Higgs reductions to \(P\). They show that for every morphism from a smooth curve \(C\) to \(X\), the pull back of \((E,\phi)\) to \(C\) is semistable if and only if for every Higgs reduction to a parabolic subgroup \(P\) and every dominant character of \(P\), the associated line bundle on \(R_P(E,\phi)\) is nef (numerically effective). They also show that these conditions are met for semistable Higgs bundles with \(c_2(ad(E))=0\). An interpretation of these results in the framework of Tannakian categories is given at the end of the paper. The authors define and study nef and nflat (numerically flat) principal (Higgs) bundles in the second part of the paper. They show that a principal (non-Higgs) \(G\)-bundle \(E\) is nflat if and only if there is a reduction of \(E\) to a parabolic subgroup \(P\) with Levi factor \(H\) such that the associated principal \(H\)-bundle is polystable and Hermitian flat (i.e. it has a flat Hermitian-Yang-Mills connection). They prove that consequently the cohomology ring of nflat \(E\), with real coefficients, is trivial.