Considering the Jordan-Brans-Dicke action:

$$S=\int d^4x\sqrt{-g}\left(\phi R+\frac\omega\phi(\partial\phi)^2+\mathfrak{L_{m}}(\psi)\right).$$

I was trying to get the metric field equations by varying the metric and got this:

$$ -\frac{1}{2}g_{\mu\nu}R+R_{\mu\nu}+\frac{\omega}{\phi^2}[-\frac{1}{2}g_{\mu\nu}(\partial\phi)^2+\partial_\mu\phi\partial_\nu\phi]-\frac{1}{2\phi}g_{\mu\nu}\mathfrak{L_{m}}(\psi)=0 $$

I varied the terms $\sqrt{-g}$, $R_{\mu\nu}$ , $g^{\mu\nu}$ and $\partial_\mu \phi \partial_\nu \phi g^{\mu\nu}$. If we are only conserned for the equations of the metric field then this is it right? If I wanted the equations for the gravitational field we would have to vary w.r.t. the metric and the field $\phi$ right?

EDIT: On the 2nd Leibniz rule I considered:

$$ -\nabla^{\alpha}\nabla_{\alpha}(g_{\mu\nu}\phi\delta g^{\mu\nu}) = -g_{\mu\nu}\nabla^{\alpha}\nabla_{\alpha}(\phi) \delta g^{\mu\nu}-g_{\mu\nu}\nabla^{\alpha} (\phi)\nabla_{\alpha}(\delta g^{\mu\nu})-g_{\mu\nu}\nabla_{\alpha} (\phi)\nabla^{\alpha}( \delta g^{\mu\nu})-g_{\mu\nu} \phi \nabla^{\alpha}\nabla_{\alpha}(\delta g^{\mu\nu}) $$

I pulled out the metric so I dont have to deal with 6 terms. The ones we want are only the first and second in the RHS of this equation

  • $\begingroup$ What happened to /phi R term?? $\endgroup$
    – Noone
    Oct 25, 2020 at 20:02
  • $\begingroup$ I divided the whole equation by $\phi$ and that one would be the first term that appears $\endgroup$ Oct 25, 2020 at 20:33
  • $\begingroup$ Τhats wrong. You have to integrate by parts. $\endgroup$
    – Noone
    Oct 25, 2020 at 20:38
  • $\begingroup$ You have already applied the first covariant derivative. You have to do it once again with the other covariant derivative. $\endgroup$
    – Noone
    Oct 26, 2020 at 14:23
  • $\begingroup$ What do you mean? I applied both of them already. What am I confusing where? $\endgroup$ Oct 26, 2020 at 14:48

1 Answer 1


The $\delta(\phi R)$ term will be:

$$\delta(\phi R) = \delta(\phi g^{\mu\nu}R_{\mu\nu}) = \phi\delta g^{\mu\nu}R_{\mu\nu} +\phi\delta R_{\mu\nu}g^{\mu\nu} $$

The term: $\phi\delta g^{\mu\nu}R_{\mu\nu}$ is ready, here the variation of the inverse metric tensor is already a multiplying factor. Now the second term is:

$$\phi\delta R_{\mu\nu}g^{\mu\nu} = \phi (g_{\mu\nu}\Box - \nabla_{\mu}\nabla_{\nu})\delta g^{\mu\nu}$$

where i've used the Palatini Identity. Now we have for example for the box term:

$$\phi g_{\mu\nu}\Box\delta g^{\mu\nu} = \phi g_{\mu\nu}\nabla^{\alpha}\nabla_{\alpha}\delta g^{\mu\nu} =\nabla^{\alpha}(\phi g_{\mu\nu}\nabla_{\alpha}\delta g^{\mu\nu}) -\nabla^{\alpha}\phi g_{\mu\nu}\nabla_{\alpha}\delta g^{\mu\nu} $$

The first term is a total derivative. We will ignore it as a boundary term. Now we use Leibniz rule once again:

$$-\nabla^{\alpha}\phi g_{\mu\nu}\nabla_{\alpha}\delta g^{\mu\nu} = -\nabla^{\alpha}\nabla_{\alpha}(g_{\mu\nu}\phi\delta g^{\mu\nu}) + g_{\mu\nu}\delta g^{\mu\nu}\nabla^{\alpha}\nabla_{\alpha}(\phi)$$

where i've used metric compatibillity. So we have:

$$\phi g_{\mu\nu}\Box\delta g^{\mu\nu} = g_{\mu\nu}\delta g^{\mu\nu}\nabla^{\alpha}\nabla_{\alpha}(\phi) = g_{\mu\nu}\delta g^{\mu\nu} \Box \phi$$ One has to do the same procedure for the two covariant derivatives. The other terms seem correct.

The problem here is that the Ricci Scalar is coupled with $\phi$. When i first came across such coupling terms i had the same problem. In the context of General relativity, the action is:

$$S = \int d^4x \sqrt{-g}R. $$

The variation gives rise to the term $g^{\mu\nu}\delta R_{\mu\nu}$. We can show that this term is a total derivative term and cancel it. In the context of Brans Dicke (or other geometric modifications to Einstein Gravity, $f(R)$ for example, Horndeski, or matter fields non-minimally coupled to gravity) this term is no longer a total divergence. Here, this term is : $\phi\delta R_{\mu\nu}g^{\mu\nu}$. $\phi$ makes things tricky, we cannot now discard this term as it is, it is not a total derivative term. Thus, we follow the procedure i described above.

Regarding the second part of the question, yes you have to vary also with respect to $\phi$. Here $\phi$ is not a matter field, it is a geometric quantity.

  • $\begingroup$ Thank you! In all the action variations I´ve done so far R wasn´t coupled with any other field so I just do it fast by claiming the Palatini identity and Gauss theorem and say it´s zero. $\endgroup$ Oct 25, 2020 at 22:04
  • $\begingroup$ I understand your situation 100%. Been there, done exactly the same thing. $\endgroup$
    – Noone
    Oct 25, 2020 at 22:14
  • $\begingroup$ Just one thing, isn´t there a $g_{\mu\nu}$ missing in the 4th equation you wrote on the 1st term of the RHS? $\endgroup$ Oct 25, 2020 at 22:54
  • $\begingroup$ Of course!! Thank you for pointing it out! $\endgroup$
    – Noone
    Oct 25, 2020 at 22:56
  • $\begingroup$ Sorry could you elaborate in how you did the second Leibniz rule? I get 4 terms when I distribute the two covariant derivatives for this term: $\nabla^{\alpha} \nabla_{\alpha}(g_{\mu\nu} \phi \delta g^{\mu\nu})$ considering the metric compatibility I took the metric out directly and got the other 4 terms, do 2 of them cancel each other out? $\endgroup$ Oct 26, 2020 at 12:16

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.