In general relativity and tensor calculus, the Palatini identity is

${\displaystyle \delta R_{\sigma \nu }=\nabla _{\rho }\delta \Gamma _{\nu \sigma }^{\rho }-\nabla _{\nu }\delta \Gamma _{\rho \sigma }^{\rho },}$

where ${\displaystyle \delta \Gamma _{\nu \sigma }^{\rho }}$ denotes the variation of Christoffel symbols and ${\displaystyle \nabla _{\rho }}$ indicates covariant differentiation.^[1]

The "same" identity holds for the Lie derivative ${\displaystyle {\mathcal {L}}_{\xi }R_{\sigma \nu }}$. In fact, one has

${\displaystyle {\mathcal {L}}_{\xi }R_{\sigma \nu }=\nabla _{\rho }({\mathcal {L}}_{\xi }\Gamma _{\nu \sigma }^{\rho })-\nabla _{\nu }({\mathcal {L}}_{\xi }\Gamma _{\rho \sigma }^{\rho }),}$

where ${\displaystyle \xi =\xi ^{\rho }\partial _{\rho }}$ denotes any vector field on the spacetime manifold ${\displaystyle M}$.

The Riemann curvature tensor is defined in terms of the Levi-Civita connection ${\displaystyle \Gamma _{\mu \nu }^{\lambda }}$ as

${\displaystyle {R^{\rho }}_{\sigma \mu \nu }=\partial _{\mu }\Gamma _{\nu \sigma }^{\rho }-\partial _{\nu }\Gamma _{\mu \sigma }^{\rho }+\Gamma _{\mu \lambda }^{\rho }\Gamma _{\nu \sigma }^{\lambda }-\Gamma _{\nu \lambda }^{\rho }\Gamma _{\mu \sigma }^{\lambda }}$.

Its variation is

${\displaystyle \delta {R^{\rho }}_{\sigma \mu \nu }=\partial _{\mu }\delta \Gamma _{\nu \sigma }^{\rho }-\partial _{\nu }\delta \Gamma _{\mu \sigma }^{\rho }+\delta \Gamma _{\mu \lambda }^{\rho }\Gamma _{\nu \sigma }^{\lambda }+\Gamma _{\mu \lambda }^{\rho }\delta \Gamma _{\nu \sigma }^{\lambda }-\delta \Gamma _{\nu \lambda }^{\rho }\Gamma _{\mu \sigma }^{\lambda }-\Gamma _{\nu \lambda }^{\rho }\delta \Gamma _{\mu \sigma }^{\lambda }}$.

While the connection ${\displaystyle \Gamma _{\nu \sigma }^{\rho }}$ is not a tensor, the difference ${\displaystyle \delta \Gamma _{\nu \sigma }^{\rho }}$ between two connections is, so we can take its covariant derivative

${\displaystyle \nabla _{\mu }\delta \Gamma _{\nu \sigma }^{\rho }=\partial _{\mu }\delta \Gamma _{\nu \sigma }^{\rho }+\Gamma _{\mu \lambda }^{\rho }\delta \Gamma _{\nu \sigma }^{\lambda }-\Gamma _{\mu \nu }^{\lambda }\delta \Gamma _{\lambda \sigma }^{\rho }-\Gamma _{\mu \sigma }^{\lambda }\delta \Gamma _{\nu \lambda }^{\rho }}$.

Solving this equation for ${\displaystyle \partial _{\mu }\delta \Gamma _{\nu \sigma }^{\rho }}$ and substituting the result in ${\displaystyle \delta {R^{\rho }}_{\sigma \mu \nu }}$, all the ${\displaystyle \Gamma \delta \Gamma }$-like terms cancel, leaving only

${\displaystyle \delta {R^{\rho }}_{\sigma \mu \nu }=\nabla _{\mu }\delta \Gamma _{\nu \sigma }^{\rho }-\nabla _{\nu }\delta \Gamma _{\mu \sigma }^{\rho }}$.

Finally, the variation of the Ricci curvature tensor follows by contracting two indices, proving the identity

${\displaystyle \delta R_{\sigma \nu }=\delta {R^{\rho }}_{\sigma \rho \nu }=\nabla _{\rho }\delta \Gamma _{\nu \sigma }^{\rho }-\nabla _{\nu }\delta \Gamma _{\rho \sigma }^{\rho }}$.

• Einstein–Hilbert action
• Palatini variation
• Ricci calculus
• Tensor calculus
• Christoffel symbols
• Riemann curvature tensor