Analytic nonabelian Hodge theory (Q520874)

Simpson's reductive proalgebraic fundamental group is an object that encapsulates the category of semisimple local systems, and thereby captures information about the Betti moduli space of a smooth complex projective variety \(X\). However, it does so only pointwise. For this and other reasons, finer invariants are desirable. This highly technical paper makes a proposal for such invariants, and accumulates properties of them. The basic idea is that instead of looking only at finite-dimensional representations of \(\pi_1\), one should allow representations in Banach algebras. The first part of the paper consists of the verification of the technical properties needed for these constructions to be useful in the same way: namely, there should be moduli functors (Betti, de Rham and Dolbeault) and associated analytic moduli spaces. For further progress one requires notions of pluriharmonic and unitary representation, in order to have counterparts of pluriharmonic bundles and Higgs bundles and hence of the non-abelian Hodge theorem, thought of as a comparison theorem between de Rham and Dolbeault moduli. These requirements can be met if one restricts to \(C^*\)-algebras. The technical detail of this is worked out in the second section of the paper. One has functors \({\mathbf R}^\diamondsuit_{X,x}\), where \(\diamondsuit\) stands for \textbf{B} (framed Betti), \textbf{dR}, \textbf{J} (for pluriharmonic bundles on a compact Kähler manifold) or \textbf{Dol}, on Banach or \(C^*\)-algebras as appropriate: of these, the first two are isomorphic and the third is representable by an object (a pro-\(C^*\)-bialgebra) \(E^J_{X,x}\). Further properties and relations among these functors and objects are established in the third section. The most significant is that \(E^J\) has a continuous \(S^1\)-action, which is what recovers the Hodge structure once one concentrates on the abelian group part (this is illustrated in the paper by an example showing what happens for complex tori). As the author says, rather quietly, ``this suggests that the most natural definition of a pure non-abelian Hodge structure is a continuous circle action on a pro-\(C^*\)-bialgebra''. Then one wants to have a concept of variation of Hodge structure, which is a representation of \(E^J\rtimes S^1\), and leads to the study, in the fourth section, of Hilbert space representations of \(E^J\) and associated local systems. Extending these ideas to the context of twistor structures, as is done in the fifth section, one finds a pure weight~\(n\) twistor structure on the reduced cohomology of these local systems, and in the special case of the fundamental group representation this even gives an analytic Hodge filtration. The final section introduces an enhanced, or globalised, version of these ideas by means of a twistor family of the framed moduli functors \({\mathbf R}^\diamondsuit\), which recover de Rham and Dolbeault functors as general and special fibres respectively.