I've been working through Wald's book "General Relativity", and I'm fairly stuck on question 4. a) of chapter 6:

"Let $(M,g_{ab})$ be a stationary space-time with time-like Killing field $\xi^a$. Let $V^2 = -\xi^a \xi_a$. Show that the acceleration $a^b = u^a \nabla_a u^b$ of a stationary observer is given by $a^b = \nabla^b \log (V)$."

Firstly I used that a stationary observer is one whose 4-velocity is the time-like Killing vector $\xi^a$, and I introduced coordinates $t,x,y,z$ so that $\xi^a = (\partial / \partial t)^a$ and the remaining three are arbitrarily defined. Then $\xi^a$ is the vector $(1,0,0,0)$ so \begin{align} a^b &= \xi^a \nabla_b \xi^a \\ &= \xi^a \partial_a \xi^b + \xi^a \Gamma^b \text{}_{ac} \xi^c \\ &= \xi^0 \partial_0 \xi^b + \xi^0 \Gamma^b \text{}_{00} \xi^0 \\ &= 0 + \Gamma^b \text{}_{00} \\ &= (1/2) g^{ab} ( 2 \partial_t g_{0a} - \partial_a g_{00}) \\ &= -(1/2) \nabla^b g_{00}, \end{align} where to reach the last line the independence of $g_{ab}$'s coordinates from the Killing coordinate $t$ and the property $\nabla_a f = \partial_a f$ for any scalar $f$ were both used. Furthermore \begin{align} V^2 = - \xi^a \xi_a = - g_{ab} \xi^a \xi^b = -g_{00}, \end{align} so in fact $a^b = (1/2) \nabla^b (V^2) = V \nabla^b V$. However, $\nabla^b \log (V) = (1/V) \nabla^b V$ by the chain rule, so it appears necessary and sufficient, for $a^b$ to equal $\nabla^b \log(V)$, that $V = \pm 1$; indeed it could only be $+1$ since otherwise $\log(V)$ is undefined. But this would imply the highly unphysical situation that $a^b = 0$.

Can anybody help me resolve my quandary? Many thanks in advance!

  • We require $u^a u_a = -1$ for 4-velocities, but $\xi^a \xi_a = -V^2$ is not generally like this.
  • $g_{00}$ is not a scalar, it's a component of a tensor.

Let's not mess about with specific frames.

We have: $V^2 = \xi^a\xi_a$, $a^b = u^a\nabla_a u^b$, $\nabla_{(a}\xi_{b)}=0$, $u^a = f \xi^a$.

Firstly, $u^a u_a = f^2 \xi^a \xi_a = -V^2 f^2$ so $f=\pm1/V$.

Let us write $$a_b = 1/V\,\xi^a\nabla_a(\xi_b/V) = \frac{1}{V^2}\xi^a\nabla_a\xi_b-\frac{1}{V}\xi^a\xi_b\nabla_a \frac{1}{V}$$ But then using the antisymmetry property $$\xi^a\nabla_a(\xi_b) = -\xi^a\nabla_b(\xi_a) = -\frac{1}{2}\nabla_b(\xi^a\xi_a) = \frac{1}{2}\nabla_b V^2$$ and again $$\xi^a\xi_b\nabla_a V^2 = -\xi^a\xi_b\nabla_a (\xi_c \xi^c) = -2\xi_b\xi^a\xi^c\nabla_a \xi_c = 0$$


$$a_b = \frac{1}{2V^2}\nabla_b V^2 = \nabla_b \log V$$

  • $\begingroup$ Great, that's exactly the pointer I needed. I didn't think a stationary observer need only have 4-velocity that's a scalar multiple of the Killing field. Although I must say I disagree with your statement that $g_{00}$ is not a scalar: it may depend on the choice of coordinates, but it nonetheless assigns a number to each point, making it one of a family of scalar functions on this region of $M$. My calculation now works with the alteration $u^a = \pm \xi^a / V$. Thanks again! $\endgroup$ – Stromael Apr 22 '13 at 22:29
  • $\begingroup$ The definition of a scalar unequivocally requires it to be invariant under coordinate transformations. Put another way, it exists in a coordinate free world too. $\endgroup$ – Sharkos Apr 22 '13 at 22:51
  • $\begingroup$ For example, $\partial_a g_{bc} \neq 0$ in general but $\nabla_a g_{bc} \equiv 0$ always (for the Levi-Civita connection). The only reason your argument works is because you just did two steps in the wrong order/because your assertion works in local inertial coordinates and this is a legal choice to make. You should write $\partial^b g_{00} = \partial^b (-V^2) = \nabla^b (-V^2)$ but you wrote $\partial^b g_{00} = \nabla^b g_{00} = -\nabla^b (-V^2)$. $\endgroup$ – Sharkos Apr 22 '13 at 23:01
  • $\begingroup$ I get that it has to be coordinate-invariant. I guess what I'm trying to say is that the scalar function, say, $h$, defined in one given coordinate system by $h = g_{00}$, is coordinate-invariant. Clearly under another chart $h$ will in general not still equal $g_{00}$. It is still a scalar. Apologies for not being clearer. $\endgroup$ – Stromael Apr 22 '13 at 23:24
  • $\begingroup$ You can define a scalar function by its value in one coordinate system, of course, yes! But the expression $\nabla_a g_{00}$ means something different to $\nabla h$ even if $h=g_{00}$ in this coordinate system; it contains the partial derivative and Christoffel symbols involving the other components of $g$. $\endgroup$ – Sharkos Apr 22 '13 at 23:49

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.