I have been doing some calculation on variation of Ricci's tensor with respect to the metric, that, according with S. Carroll (An Introduction to General Relativity: Spacetime and Geometry, equation 4.62) is \begin{align}\delta R_{\mu \nu}= \nabla_{\rho}(\delta \Gamma_{\mu \nu}^{\rho}) - \nabla_{\nu}(\delta \Gamma_{\lambda \mu}^{\lambda}) \tag{4.62} \end{align} and I got a explicit equation in terms of the metric, which is \begin{equation} g^{\mu \nu} \delta R_{\mu \nu} = -\nabla_{\mu}\nabla_{\nu}\delta g^{\mu\nu} + g_{\mu \nu}\square\delta g^{\mu \nu} \tag{1}\end{equation} that agrees with the result obtained by Prahar here Variation of modified Einstein Hilbert Action. The thing is that, as far as I am concerned, \begin{align} g^{\mu \nu}\delta R_{\mu \nu}= g^{\mu \nu}\frac{\delta R_{\mu \nu}}{\delta g^{\alpha \beta}}\tag{2} \end{align} because it is the variation respect to the metric. So, if we write explicitly the variation $\delta$ as $\frac{\delta}{\delta g^{\alpha \beta}}$ in (1) we would get\begin{align}g^{\mu \nu} \frac{\delta}{\delta g^{\alpha \beta}} R_{\mu \nu} = -\nabla_{\mu}\nabla_{\nu}\frac{\delta}{\delta g^{\alpha \beta}} g^{\mu\nu} + g_{\mu \nu}\square\frac{\delta}{\delta g^{\alpha \beta}} g^{\mu \nu}\end{align} But \begin{align} \frac{\delta g^{\mu \nu}}{\delta g^{\alpha \beta}} = \frac{1}{2} \left(\delta_{\alpha}^{\mu}\delta_{\beta}^{\nu} + \delta_{\alpha}^{\nu}\delta_{\beta}^{\mu} \right) \tag{3} \end{align} and \begin{align}\nabla_{\mu}\delta_{\beta}^{\alpha}=0 \tag{4} \end{align} Therefore, plugging (3) and (4) into (1), we get \begin{align} \frac{\delta R_{\mu \nu}}{\delta g^{\alpha \beta}} = 0. \end{align} Is it true? if not, why does this fall? I think this time notation is messing around with me.



1 Answer 1


This confusion can be settled with a much simpler example. Let $$ F(x) = \frac{d}{dx}f(x)\,, $$ Now $$ \delta F(x) = \frac{d}{dx}\delta f(x)\,. $$ With your reasoning one would say $$ \frac{\delta F}{\delta f} = \frac{d}{dx} \frac{\delta f}{\delta f} = \frac{d}{dx} 1 = 0\,, $$ which is nonsence. But actually $$ \frac{\delta f(x)}{\delta f(y)} = \delta(x-y)\,, $$ so $$ \frac{\delta F(x)}{\delta f(y)} = \frac{d}{dx} \delta(x-y)\,. $$ The derivative of the Dirac distribution is also a distribution. The same goes with your case, you have to put a Dirac delta in $$ \frac{\delta g^{\mu\nu}(x)}{\delta g^{\alpha\beta}(x')} = \frac12\left(\delta^{\mu}_\alpha\delta^\nu_\beta + \delta^{\nu}_\alpha\delta^\mu_\beta\right) \,\delta^d(x-x')\,. $$

  • $\begingroup$ Thank you so much! This resolves my doubt. $\endgroup$ Mar 24, 2020 at 21:55
  • 6
    $\begingroup$ the correct way to say "thank you" is to upvote and/or accept the answer $\endgroup$
    – Yukterez
    Mar 30, 2020 at 6:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.