Beginning with a metric with small perturbations \begin{eqnarray} g_{00} &=& 1-2\frac{U}{c^2} + \mathcal{O}(c^{-4}) \\ g_{0i} &=& \mathcal{O}(c^{-3}) \\ g_{ij} &=& -\delta_{ij}\left(1+2\frac{U}{c^2} + \mathcal{O}(c^{-4})\right) \end{eqnarray}

I'm trying to show that the components of the 4-velocity vector $u^0$ and $u^i$ can be written as \begin{eqnarray} u^0 &=& 1 + \frac{1}{2}\frac{v^2}{c^2} + \frac{U}{c^2} + \mathcal{O}(c^{-4}) \\ u^i &=& \frac{v^i}{c}\left(1 + \frac{1}{2}\frac{v^2}{c^2} + \frac{U}{c^2}\right) + \mathcal{O}(c^{-5}) \end{eqnarray}

I want to explicitly keep powers of $c$ in the calculation to help learn about different orders of smallness in PPN. The method I am attempting to use is to write $u^\mu u_\mu = u^0 u_0 + u^i u_i = +1$. This may be rearranged as \begin{eqnarray} u^0 u^0 g_{00} &=& 1 - u^i u^j g_{ij} - 2u^0 u^i g_{0i} \\ \left(u^0\right)^2 \left[ 1-2\frac{U}{c^2}\right] &=& 1 + \left(\frac{dx^i}{ds}\right)^2\left[ 1+2\frac{U}{c^2} \right] \end{eqnarray} I dropped the $g_{0i}$ term because it is $\mathcal{O}(c^{-3})$. I found that if I write $\frac{dx^i}{ds} = \frac{1}{c}\frac{dx^i}{dt}\frac{dt}{ds} = \frac{v}{c}$, solve the equation above for $u^0 = \sqrt\frac{...}{...}$, and then taylor expand about the small terms $\left(\frac{v}{c}\right)^2$ and $\frac{U}{c^2}$, I get the correct form of $u^0$.

My questions are:

  1. Is this the best approach to solve for $u^0$ and $u^i$, or is there a better/more highly recommended method to use?

  2. I don't understand if or why it is correct to write $\left(\frac{dx^i}{ds}\right)^2 = \left(\frac{v}{c}\right)^2$. The term $\frac{dx^i}{ds}$ should have no units (like $\frac{v}{c}$), but when I use the chain rule to expand $\frac{dx^i}{ds}$, shouldn't there be a factor of c next to dt in the numerator above as well? Could it have something to do with thinking of ds as proper time, and then writing $dt\approx ds$ for slow moving objects?

  3. Also, I read that metric components of $g_{0i}$ will have odd powers of $c^{-1}$, while metric components of $g_{00}$ and $g_{ij}$ will have even powers of $c^{-1}$, but have never seen a justification for that claim.

  4. Ultimately my goal is to understand the PPN formalism. Are there any textbooks or papers where the author works through expanding a perturbed metric while keeping factors of $c^{-1}$ explicitly in the calculation to show different orders of the small terms (e.g. does not set $c=1$)? So far, the report I linked below is the only paper I have ever seen that does not work in units where $c=1$.

My question is about how to arrive at eq. (15) in this link (the link will automatically download a pdf of the paper).

If you don't want to click on a link that automatically downloads a paper, the title of the paper is "The post-Newtonian formalism" by Rene Michelsen.

  • $\begingroup$ Is it helpful to express the four velocity as $u^\beta = g^{\beta \alpha} u_{\alpha}$? $\endgroup$ Jul 1, 2018 at 22:43
  • $\begingroup$ I have been using that relation to make all 4-velocities contravariant (e.g. they all have upper indices). I haven't yet tried making 4-velocities covariant (have lower indices), which I may attempt next, but it's not obvious to me whether that will help or not. Somewhat related, this method hasn't yet worked for $u^i$, since Taylor expanding a term $\propto\sqrt\frac{v}{c}$ gives infinity at lowest (non-zero) order. $\endgroup$
    – Bob
    Jul 2, 2018 at 3:52
  • $\begingroup$ Sept. 2021: Link now dead. $\endgroup$
    – Qmechanic
    Sep 4, 2021 at 12:57

1 Answer 1


Express four-velocities $u^\alpha$ and coordinate velocities $v^i$ according to

$$ u^\alpha = \frac{dx^\alpha}{d\tau}, \quad v^i = \frac{dx^i}{dt}, $$

where proper time and coordinate time is given by $\tau$ and $t$ respectively. Hereafter, Latin ($i,j = 1,2,3$) and Greek ($\alpha,\beta = 0,1,2,3$) indices account for spatial and spacetime variables respectively.

I will answer your question $(1)$ in terms of arbitrary metric tensor components, where I will outline the procedure you need to follow in order to get the desired results.

A timelike particle obeys the normalisation condition given by

$$ g_{\alpha \beta} u^\alpha u^\beta = -c^2, \tag{1} \label{eqn1}$$

where $c$ is the speed of light. Eq. (\ref{eqn1}) is given explicitly by

$$ \left( u^0 \right)^2 \left( g_{00} + 2 g_{0i} \frac{u^i}{u^0} + g_{ij} \frac{u^i}{u^0} \frac{u^j}{u^0} \right) = -c^2. \tag{2} \label{eqn2} $$

Now, we know $x^0 = ct$. Hence $d x^0 = cdt. $ Using this we can express $u^i/u^0$ as the following

$$ \frac{u^i}{u^0} = \frac{dx^i}{d\tau} \left(\frac{d x^0}{d\tau} \right)^{-1} = \frac{dx^i}{dt} \frac{dt}{d\tau} \left( \frac{dx^0}{dt} \frac{dt}{d\tau} \right)^{-1} = \frac{v^i}{c}. \tag{3} \label{eqn3}$$

(Note: I hope that addresses your confusion regarding question $2$. ) Returning to Eq. (\ref{eqn2}) we have

$$ u^0 = c \left[ - g_{00} - 2g_{0i} \left( \frac{v^i}{c} \right) - g_{ij} \left(\frac{v^i}{c}\right) \left( \frac{v^j}{c} \right) \right]^{-1/2}.$$

Substituting in the appropriate metric tensor components and expanding the square root to PN order will give the desired $u^0$. Consulting Eq. (\ref{eqn3}) gives the appropriate expression for the spatial part of the four-velocity namely

$$ u^i = \frac{v^i}{c} u^0. $$

So to answer your question $(1)$, I would consult the normalisation condition to easily obtain the components of $u^\alpha$. By following the method outlined above you will obtain the expressions you have quoted for $u^0$ and $u^i$.

I should note that there is a big difference between the post-Newtonian and parametrised post-Newtonian formalisms which I would advise you familiarise yourself with. I assume you mean you want to understand the PN formalism in your question $(4)$. Works by Chandrasekhar on the PN formalism generally retain the factor of $1/c$ for convenience. However, you should know that this is simply a matter of correctly determining the appropriate dimensions and reinstating. Being able to do this is more beneficial in the longrun.

  • $\begingroup$ Thanks for the reply. In regard to question (4), I mean I want to work with the PPN formalism. I have a equation of motion that reduces to the trace-free EFE at lowest order and has a scalar field, but want to know if it is ruled out at first order in PPN parameters. $\endgroup$
    – Bob
    Jul 13, 2018 at 14:22
  • $\begingroup$ Should your eq. (1) read $g_{\mu\nu}u^\mu u^\nu = +1$ since (i) the spacetime metric signature is (+,-,-,-) and (ii) we are normalising the 4-velocities? $\endgroup$
    – Bob
    Jul 13, 2018 at 20:15
  • $\begingroup$ @Bob if you are concerned with the PPN formalism, shouldn’t there be parameters $\alpha,\beta$ in your metric components? Again I stress the PN and PPN formalisms are different animals. $\endgroup$ Jul 13, 2018 at 23:56
  • $\begingroup$ @Bob the sign will change according to your signature. However, I thought you wanted to retain $c$ explicitly? $\endgroup$ Jul 13, 2018 at 23:58

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.