In the
theory of general relativity (GRT), spacetime is modeled as a smooth curved
Lorentzian (pseudo-Riemannian) 4d manifold $\mathcal{M}_4 = \{ {\mathbf{x}}\}
$, parametrized by coordinates $\{ {x^\mu };\mu = 0,1,2,3\} $ so that
${\mathbf{x}} = {\mathbf{x}}({x^0},{x^1},{x^2},{x^3}): = {\mathbf{x}}(x)$. The
chosen metric signature is $( + , - , - , - )$, one dimension of time and
three of space; units such that the velocity of light and the gravitational
constant both have the value one: $c = 1$, $G = 1$; where appropriate these
constants will be written explicitly.
A set of
four linearly independent local coordinate basis vectors are defined
by:
At each point ${\mathbf{x}}$ of the spacetime manifold these tangent
vectors $\{ {{\mathbf{g}}_\mu }({\mathbf{x}})\} $ provide a holonomic
basis for the vector space ${V_4}({\mathbf{x}}) =
{T_{\mathbf{x}}}\mathcal{M}$ called the tangent space to
${\mathcal{M}_4}$ at ${\mathbf{x}}$. The tangent space is by definition a Minkowski
space. The tangent bundle of spacetime and tangent space is said to be soldered,
because they are both four-dimensional spacetimes. This geometrical structure
is intrinsic to the coordinatization of the manifold, independent of any prior
assumptions.
Physical
quantities of interest in GRT are represented by tensor fields, or other
geometric objects like spinors, that live in the tangent vector space attached
to each point in spacetime, including the higher-order tensor spaces
constructed therefrom. These geometric objects can be specified in terms of
arbitrary coordinates, but are themselves coordinate independent. It is a
fundamental principle in physics that equations between such geometric objects
are generally covariant, that is, they are invariant under any change
of coordinate system.
In the
subject of differential geometry, it is now common practice to identify
directional derivatives as tangent vectors. However, there is a complete
isomorphism between coordinate vectors $\{ {{\mathbf{g}}_\mu }\} $ as defined
in (1.1) and the corresponding directional derivatives $\{ {\partial _\mu }\}
$. For all practical purposes, the two viewpoints are equivalent, and
calculations performed in either scheme will return the same results.