The Einstein equations can be written as (1): $$R_{ab}-\frac{1}{2}Rg_{ab} = -8\pi GT_{ab}$$

or by contracting the above equation with the metric tensor and resubstituting: (2) $$R_{ab}=8\pi G(\frac{1}{2}Tg_{ab}-T_{ab}).$$

In a vacuum, equation (1) reduces to $R_{ab}-\frac{1}{2}Rg_{ab}=0$ and equation (2) reduces to $R_{ab}=0$, which implies that in a vacuum, $R=0$.

However, if I explicitly calculate $R$ for a plane wave of the form $$h_{ab} = A_{ab}\exp(ikx)$$ (the Minkowski metric $\eta_{ab}$ perturbed by $h_{ab}$), I obtain: $$R=k^ak^bh_{ab}-k^\lambda k_\lambda h\ \ ,$$ where $h=\eta_{ab}h^{ab}$, which looks like some sort of wave equation, but is nonzero. It's supposed to be $0$ but is not. Why?

  • $\begingroup$ The Ricci scalar is zero for a gravitational wave, as is any curvature scalar. See Hans-Juergen Schmidt, "Why do all the curvature invariants of a gravitational wave vanish?" arxiv.org/abs/gr-qc/9404037 . The fact that you get a nonvanishing Ricci scalar tells us that your metric is not a gravitational wave in vacuum, or that you made a mistake in your calculation. $\endgroup$
    – user4552
    Commented Mar 6, 2018 at 1:28

1 Answer 1


Gravitational waves are transverse and travel along null rays. Thus, you have to have $k^{a} h_{ab} = 0$ and $k^a k_a = 0$. (More accurately, the transversality condition can be viewed as a gauge condition: we can always apply a local diffeomorphism such that this first condition is true.) Under these conditions, the Ricci scalar does indeed vanish—as should the Ricci tensor, if you've done everything correctly. But a completely arbitrary wave-like metric with an arbitrary polarization $h_{ab}$ and propagation vector $k_a$ will not, as you've found, satisfy Einstein's equation in vacuum.

  • $\begingroup$ Hmm but $R$ is gauge invariant, so you shouldn't need to impose any transversality condition to ensure it vanishes, right? What am I missing here? $\endgroup$ Commented Mar 5, 2018 at 22:31
  • $\begingroup$ It's just like how Maxwell's equations imply that the photon polarizations are transverse. $\endgroup$ Commented Mar 5, 2018 at 23:41
  • 1
    $\begingroup$ @AccidentalFourierTransform The plane wave ansatz given in the question is only a solution to the (linearised) vacuum Einstein equations in transverse gauge. If we move out of this gauge, the form of the metric changes, and so $R$ picks up extra terms that cancel the now non-vanishing $k^a h_{ab}$. $\endgroup$
    – gj255
    Commented Mar 6, 2018 at 0:47

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.