XI. Field Equations
The canonical Levi-Civita connection is characterized by the fact
that the connection is compatible with the metric and has no
torsion. However, a priori, metric and connection
are separate concepts, and the notion of curvature can be defined
for arbitrary connections. To take metric and connection as independent
variables in the action principle, was suggested by work of
Palatini (1919) and given its variational form by Einstein (1923-25)
in his search for the unification of gravity and electrodynamics.
In the Einstein-Palatini framework, the Einstein-Hilbert action
integral (11.06) is written in terms of the Ricci tensor (10.17)
defined by contraction of the Riemann (10.3), so that the Ricci
tensor depends on the connection but not explicitly on the
metric:
\[{S_{{\rm{EP}}}}[g,\Gamma ] :=
\frac{1}{{2\kappa }}\int {{d^4}x} \sqrt {\left| g \right|} \left(
{{g^{\mu \nu }}{R_{\mu \nu }}[\Gamma ] + 2\Lambda } \right)\]
The pair of metric and connection $\left\{ {{g^{\mu
\nu }},\Gamma _{\mu \nu }^\lambda } \right\}$ constitute the basic
gravitational variables, without the restriction that the
connection is torsion free, i.e. symmetric in its lower indices. The
Einstein equations are then obtained by varying the (inverse)
metric, and the metric compatible and torsion-free Levi-Civita
connection by varying the connection. It is called a first-order
formulation as the variables to vary over involve only up to
first derivatives in the action.
In the actual variational calculation it is convenient to introduce
the contortion tensor $C_{\mu \nu }^\lambda $, i.e. the
deviation from the Levi-Civita connection, as the variable to be
varied. Solving the Euler-Lagrange equations of motion, one is led
to the conclusion that the connection is the Levi-Civita one, up to
an arbitrary vector ${A_\mu }$:
\[\Gamma _{\mu \nu }^\lambda =
{\left. {\Gamma _{\mu \nu }^\lambda } \right|_{{\rm{LC}}}} + {A_\mu
}\delta _\nu ^\lambda \qquad{A_\mu }: = C_{\sigma \mu }^\sigma \]
Since ${A_\mu }$ can have any value, one may set this vector equal
to zero. In fact, it was already stated by Einstein in 1925 that, in
order to obtain the correct equations of motion, the trace of
the torsion tensor (1.23) must vanish: $T_{\mu \nu }^\nu = 3
{A_\mu } = 0$.
However, there is altogether no need to impose any condition because
the arbitrariness of ${A_\mu }$ corresponds to a (Noether) gauge
symmetry of the action, as is easily verified. As a
consequence, the equations of motion are invariant. Therefore, one
may conclude that the theory described by the Einstein-Palatini
action (11.7) is indeed Einstein's GRT, without any ambiguity.