The problem of defining the derivative of a vector field in the direction
For a vector field near , in the direction, try to calculate the derivative in coordinates
However, the #link(<difference-polynomial>)[difference]
operation is not linearly compatible with coordinate changes of general diffeomorphisms.
But in a metric-manifold, there are special coordinates โ #link(<geodesic-coordinate>)[geodesic coordinates]
. The transformation method for different geodesic coordinates at is , which is linear.
geodesic-derivative
_(tag) Geodesic derivative alias Levi-Civita-derivative
_(tag) Levi-Civita derivative :=
In geodesic coordinates at point , the derivative at point ,
It is also possible to take the derivative of a #link(<tensor>)[]
field . According to the scalar multiplication associated with the tensor structure, calculations can be performed using the #link(<Leibniz-law>)[product-rule]
Example
Prop . Proof In geodesic coordinates
Prop or
Prop The covariant derivative is compatible with the metric-dual e.g. since
Other coordinates may be needed to compute geodesic coordinates, and thus other coordinates may also be needed to express the geodesic derivative.
geodesic-derivative-in-general-coordinate
_(tag)
Compute geodesic coordinates using general coordinates , then in coordinates , the geodesic derivative is
Using #link(<connection-transformation>)[connection transformation]
==>
Use . Substitute into the calculation of
The tangent space linearly transforms to by , but keeps in coordinate , but keep in coordinate
Or written as, in general coordinates, geodesic derivative
For coordinate-frame
Is there a more intuitive explanation, rather than directly using the transformation of connection?
If we only consider linear compatibility, then there are many #link(<principal-bundle-connection>)[linear connections]
, and the one that coincides with the geodesic-derivative is the metric-connection
geodesic-derivative-of-co-vector
_(tag) Prop For co-vector field
Proof
Question Similar to the case of vector fields. Use the transformation and the product-rule
For co-vector coordinate-frame
parallel-transport-metric-connection
_(tag)
Parallel transport as "zero rate of change along the curve" or where
is an ODE
According to calculation (?), covariant derivative can be recovered from parallel transport + difference quotient
orthonormal-frame
_(tag)
Parallel transport of metric-connection preserves the metric
Can be used to construct orthonormal frames
It can be proven that a manifold metric uniquely corresponds to an principal-bundle structure on the manifold
But are there more specific and operational calculation results? Regarding calculating orthonormal frames using parallel transport in geodesic coordinates
A canonical orthonormal frame may be used for simplified some of calculations of spinors on curved manifolds, e.g. Pauli-matrix