So, in Hobson's general relativity, the following question is asked:

Show that covariant differentiation obeys the usual product rule, e.g. $$\nabla_a(A_{bc}B^{cd})=\nabla_a(A_{bc})B^{cd}+A_{bc}\nabla_a(B^{cd})$$ Hint: Use local cartesian coordinates.

And well, this time I think I got it right by first making $T_b{}^d=A_{bc}B^{cd}$. Then I applied the usual covariant derivative over a tensor:

$$\nabla_a T_b{}^d = \partial_a T_b{}^d - \Gamma^e{}_{ba}T_e{}^d+\Gamma^d{}_{ea}T_b{}^e$$

And here I just reversed the change $A_{bc}B^{cd}=T_b{}^d$ and got:

$$\nabla_a A_{bc}B^{cd} = \partial_a (A_{bc}B^{cd}) - \Gamma^e{}_{ba}A_{ec}B^{cd}+\Gamma^d{}_{ea}A_{bc}B^{ce}=\ ...$$

Now applying the product rule on the first term leads to: $$... \ = \partial_a (A_{bc})B^{cd} + A_{bc}\partial_a (B^{cd}) - \Gamma^e{}_{ba}A_{ec}B^{cd}+\Gamma^d{}_{ea}A_{bc}B^{ce}=\ ...$$

Finally, just have to factor out the $A_{bc}$ and $B^{cd}$ in the corresponding terms and I ended up getting:

$$... \ = (\partial_a A_{bc}- \Gamma^e{}_{ba}A_{ec})B^{cd} + A_{bc}(\partial_a B^{cd}+\Gamma^d{}_{ea}B^{ce}) = \nabla_a(A_{bc})B^{cd}+A_{bc}\nabla_a(B^{cd})$$

Which is the desired result.

Whoever is reading this might probably be wondering why did I write a question about something I believe I got right. If so, I suggest, my reader, that you take a careful look to the exercise again. Yes, at the end of it a hint is given "use local cartesian coordinates", as if it was the key for solving this exercise. Now, my question is: Did I use local cartesian coordinates without even being conscious of it? Is my exercise wrong? I'm so puzzled by it. Any comment on this will be highly appreciated!

  • 1
    $\begingroup$ The covaraint derivative is defined to obey the prduct rule: physics.stackexchange.com/questions/505592/… $\endgroup$
    – mike stone
    Mar 16, 2021 at 16:01
  • $\begingroup$ @mike stone Yes, that is always circular. Asking to prove it obeys the property in the definition. $\endgroup$
    – DanielC
    Mar 16, 2021 at 18:08
  • $\begingroup$ @DaneilC Of course one can start from the other end ---- and I learned $\nabla_\mu$ in a way that requires proving Leibnitz, but I've come to think that defining $\nabla_X$ to be a derivation is the fastest, and easiest, way to understand the concept. Others may disagree :) $\endgroup$
    – mike stone
    Mar 16, 2021 at 20:45
  • $\begingroup$ You may find my answer to an analogous question about the covariant derivative in E&M helpful. $\endgroup$ Mar 16, 2021 at 23:06

1 Answer 1


I guess that local cartesian coordinates are the same thing as local inertial coordinates where $\Gamma^\mu_{\nu\rho} = 0$.
Then (in local intertial coordinates)

$$\nabla_\mu (A_{\nu\rho}B^{\rho\sigma}) = \partial_\mu (A_{\nu\rho}B^{\rho\sigma}) = (\partial_\mu A_{\nu\rho})B^{\rho\sigma} + A_{\nu\rho}\partial_\mu B^{\rho\sigma} = (\nabla_\mu A_{\nu\rho})B^{\rho\sigma} + A_{\nu\rho}\nabla_\mu B^{\rho\sigma}$$

This is a tensor equation and therefore must be valid in any coordinate system, hence

$$\nabla_a (A_{bc}B^{cd}) = (\nabla_a A_{bc})B^{cd} + A_{bc}\nabla_a B^{cd}$$

Your entire proof is independent of any coordinate chart but also correct. Maybe this version is a bit shorter though...


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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