Fields of symmetric tensors on a compact Riemannian manifold (Q1316926)

For an \(n\)-dimensional Riemannian manifold \((M,g)\) an operator \(\delta^* : S^ p M \to S^{p + 1} M\) is considered as the composition of the covariant differentiation \(\nabla : S^ p M \to S^ 1 M \otimes S^ p M\) with a symmetrization \(\sigma : S^ 1 M \otimes S^ p M \to S^{p + 1} M\) where \(S^ p M\) is the vector bundle of symmetric \(p\)-tensor. We have the following orthogonal decomposition \(\nabla \varphi = \text{Pr}_{\text{Im }\delta^*} \nabla \varphi + \text{Pr}_{\text{Ker }\sigma} \nabla \varphi\) which divides the class of fields of symmetric \(p\)-tensor in two; \(\varphi\) is a field of geodesic \(p\)-tensors iff \(\nabla \varphi = \text{Pr}_{\text{Im }\delta^*} \nabla \varphi\) is a field of Codazzi \(p\)-tensors iff \(\nabla \varphi = \text{Pr}_{\text{Ker }\sigma}\nabla \varphi\). For arbitrary vector fields \(X\), \(Y\) and an orthonormal basis \(\{e_ 1, \dots, e_ m\}\) are defined: \(\overset\circ{R}:S^ {2M} S^ 2M\), \(\overset \circ {R} (\varphi) (X,Y) = \sum g (R(X, e_ i) Y,e_ j) \varphi (e_ i, e_ j)\) and a bilinear form \(B : S^ 2 M \otimes S^ 2 M \to S^ 0 M\), \(B(\varphi,\psi) = g(\overset \circ {R} (\varphi), \psi)\) and an associated quadratic form \(Q(\varphi) = B(\varphi,\varphi)\). The main results are: Let \((M,g)\) be a compact oriented Riemannian manifold and \(Q\) take nonnegative values everywhere on \((M,g)\), then each geodesic \(p\)-tensor and each field of Codazzi \(p\)- tensors with zero divergence is parallel. But if \(Q\) is negative-definite (positive-definite), then there exists no field of geodesic \(p\)-tensors (Codazzi \(p\)-tensors with zero divergence) on \((M,g)\). For \(p = 2\) this theorem is a well-known statement on Codazzi-tensor fields.