For every vector field we define a map that extends the covariant derivative to general tensor fields by requiring:
(Cov1) For scalar fields,
(Cov2) For 1-forms assume a Leibnitz rule for , ,
(Cov3) for any pair of tensor fields ,
(Cov4) The Leibnitz rule holds with respect to tensor products
These requirements define a unique tensor field for any smooth tensor field . Firstly, let be any 1-form defined on a coordinate chart covering Setting , condition (Cov1) gives , while (Cov2) implies
Hence, using Eq. (18.5) and , we find
where
Verify that Eq. (18.13) implies the coordinate expression for condition (Cov2),
For a general tensor , expand in terms of the basis consisting of tensor products of the and and use (Cov3) and (Cov4). For example, if
a straightforward calculation results in
where
This demonstrates that (Cov1)–(Cov4) can be used to compute the components of at any point with respect to a coordinate chart covering , and thus uniquely define the tensor field throughout .
For every tensor field of type its covariant derivative is the tensor field of type defined by
Show that, with respect to any local coordinates, the tensor field has component defined by Eq. (18.14).
Show that
The covariant derivative commutes with all contractions on a tensor field
This relation is most easily shown in local coordinates, where it reads
the upper index being in the th position, the lower in the th. If we expand the right-hand side according to Eq. (18.14) the terms corresponding to these indices are
and what remains reduces to the expression formed by expanding the left-hand side of Eq. (18.17).
A useful corollary of this property is the following relation:
Problems
Show directly from (Cov1)–(Cov4) that for all vector fields tensor fields and scalar functions
Verify from the coordinate transformation rule Eq. (18.11) for that the components of the covariant derivative of an arbitrary tensor field, defined in Eq. (18.14), transform as components of a tensor field.
Show that the identity Eq. (18.18) follows from Eq. (18.16).