No label defined (Q4382432)

The author sets out to reveal ``one of the mysteries of the theory of Langlands'', namely the fact that certain automorphic forms (or representations) which ought to be susceptible of a natural arithmetic interpretation seem to escape an interpretation in terms of algebraic geometry and at the present appear to be of a purely analytical nature. He illustrates the problem with a number of interesting examples in his introduction and then concentrates on the following problem. Denote by \(G\) an inner anisotropic form of a unitary group of three variables over \(\mathbb{Q}\) such that \(G_{\mathbb{R}}\cong U(2,1)\). He shows that the mysterious forms occur in the cohomology (in degrees 1 and 2) of a coherent sheaf over an associated Griffiths-Schmid variety. That sheaf is a ``square root'' of the dual sheaf and his result then appears as a natural generalization of the classical case of forms of weight 1. Unfortunately, the arithmetical interpretation looked for does not follow, since the Griffiths-Schmid variety (unlike the Shimura varieties) is not an algebraic variety. He shows that the cohomology classes, whose interpretation is sought, are linear combinations of cup-products of the 1-cohomology classes associated to automorphic representations and are, in particular, rational. He also proves some `non-vanishing' results. The author looks to a further paper in which the algebraic nature of the variety in question is demonstrated in terms of moduli spaces for certain Hodge structures and an explicit correspondence between certain automorphic forms for unitary groups and classes of `automorphic cohomology' which arise (a variation on a theme of Gindikin and Penrose). -- This paper is then essential preliminary reading to further work on that broader theme.