Nijenhuis geometry of parallel tensors

Abstract

A tensor -- meaning here a tensor field $\Theta$ of any type $(p,q)$ on a manifold -- may be called integrable if it is parallel relative to some torsion-free connection. We provide analytical and geometric characterizations of integrability for differential $q$-forms, $q=0,1,2,n-1,n$ (in dimension $n$), vectors, bivectors, symmetric $(2,0)$ and $(0,2)$ tensors, as well as complex-diagonalizable and nilpotent tensors of type $(1,1)$. In most cases, integrability is equivalent to algebraic constancy of $\Theta$ coupled with the vanishing of one or more suitably defined Nijenhuis-type tensors, depending on $\Theta$ via a quasilinear first-order differential operator. For $(p,q)=(1,1)$, they include the ordinary Nijenhuis tensor.

Further typos corrected, text streamlined in places, some references added

References in corpus (1)

• Bernhard Riemann 1861 revisited: existence of flat coordinates for an arbitrary bilinear form

Cited by in corpus (1)

• Parallel differential forms of codegree two, and three-forms in dimension six