Invariants and calculus for projective geometries (Q2563412): Difference between revisions

In principle, it is already known how to write down all local scalar invariants on a projective manifold. However, this is the first article which actually does it. Moreover, the method is much refined from the `in principle' procedure. The `in principle' procedure involves two steps, one of which is also due to \textit{A. R. Gover} [J. Am. Math. Soc. 7, 145-158 (1994; Zbl 0845.53010)]. This step is purely algebraic, solving a linearised problem arising from a geometric normal form essentially due to T. Y. Thomas. In the Riemannian case, the corresponding steps are taken using Weyl's classical invariant theory and Riemann's normal coordinates. In this article, Gover extends the tractor calculus of Thomas so that one can write down, in terms of it, a series of formulae for projective invariants. He calls them Weyl invariants. The main result is that, apart from a small class of exceptions which may be listed separately and explicitly, all projective invariants are linear combinations of Weyl invariants. This is like a first main theorem of invariant theory but it is formulated and proved directly on the manifold. This is an extremely significant step. Though its proof follows a linearised and purely algebraic model (included in this article for purposes of motivation), it really does combine the usual two steps into a more coherent whole. A deeper motivation for this new approach comes from other so-called `parabolic' geometries (of which conformal and CR are probably the best known) where the `in principle' construction is known to fail for invariants of sufficiently high order. Gover's most recent work is in the conformal case where the analogous approach effectively finds the form of all invariants no matter how high their order.