I've been studying the book "Semi-Riemannian Geometry" by B. O'Neill and doing some of the excersises. Chapter 6 (special relativity) includes the following one:

If $p$ is an event not on the world line of a freely falling observer $\omega$, then $\omega$ meets the nullcone $\Lambda(p)$ of $p$ in exactly two events: $\omega(\tau^{-})$ and $\omega(\tau^{+})$. Furthermore, if $q$ is an event $q=\omega((\tau^{+}+\tau^{-})/2)$, then $\vec{pq} \bot \omega$ and $\vec{pq}=q\omega(\tau^{-})=q\omega(\tau^{+})$

The first part I can see easily: Since the nullcone $\Lambda(p)$ of $p$ stretches to infinity, it must meet $\omega$ in at least two places. It cannot intersect in more than two places, since $\omega$ would then have to move faster than the speed of light. It is the last part that is troubling me.

I've tried drawing the problem as I see it. $q$ must be between $\omega(\tau^{+})$ and $\omega(\tau^{-})$, but showing that the vector $\vec{pq}$ spanning from $p$ to $q$ is perpendicular to the worldline $\omega$ (or lies in the restspace of the worldline, if you will) is giving me some trouble. My guess is that there is some trig identity I don't know. If anyone has a hint, I'd be happy :)


One thing that immediately stands out from your picture is that you seemed to have ignored the "freely falling" condition.

The world line of a freely falling observer is always a geodesic, and in the context of special relativity, is a straight line.

Also pay attention to the fact that "perpendicular" here is used in the Lorentzian sense, so in your picture it will not look "perpendicular".

Finally, the hint. If $\Lambda(p)$ intersects the worldline $\omega$ at the proper times $\tau_\pm$, the position vector relative to an arbitrary origin of $\omega$ can be written as (since it is a straight line)

$$ \omega(\tau) = \omega(\tau_-) + \frac{\tau - \tau_-}{\tau_+-\tau_-} \left[ \omega(\tau_+) - \omega(\tau_-)\right] $$

in slope-intercept form. The term in brackets is a tangent vector of the world line (the slope). You know that $\|\omega(\tau_\pm) - p\| = 0$ by assumption. So next you just need to plug it in and compute

$$ \langle \omega(\frac12(\tau_+ + \tau_-)) - p, \omega(\tau_+)- \omega(\tau_-) \rangle $$

Addendum: let me include the computation here.

$$\begin{align} \langle \omega(\frac12(\tau_+ + \tau_-)) - p, \omega(\tau_+) - \omega(\tau_-) \rangle & = \langle \frac12 (\omega(\tau_+) + \omega(\tau_-)) - p , \omega(\tau_+) - \omega(\tau_-) \rangle \\ & = \frac12 \langle (\omega(\tau_+) - p) + (\omega(\tau_-) - p) , (\omega(\tau_+) - p) - (\omega(\tau_-) - p) \rangle \\ &= \frac12 \left[ \langle \omega(\tau_+) - p, \omega(\tau_+) - p\rangle - \langle \omega(\tau_-) - p, \omega(\tau_-) - p\rangle \right] \end{align} $$ noting that the cross terms cancel each other. The term on the final line is equal to $\frac12 ( \|\omega(\tau_+) - p\|^2 - \|\omega(\tau_-)-p\|^2) = 0$.

  • $\begingroup$ Thanks alot for your answer and the hint! I have a difficulty understanding your slope-intercept form and how it describes the position vector for an arbitrary origin of $\omega$? Doesn't it describe a position vector with origin $\tau_{-}$? The slope is $[\omega (\tau_{+}) - \omega (\tau_{-})]/(\tau_{+}-\tau_{-})$, correct? $\endgroup$ – Snakeyes Oct 26 '15 at 20:06
  • $\begingroup$ (1) the word slope here only indicates the direction, not the magnitude. (The former is independent of parametrisation, the latter is not.) (2) $\tau_-$ is not a point in space, it is a point in $\mathbb{R}$. $\endgroup$ – Willie Wong Oct 26 '15 at 20:11
  • $\begingroup$ (1) I see. I'm probably thinking too much about this in classical terms. Thanks again. Sorry if I'm being a pest, but then I don't see why you choose $\frac{\tau-\tau{-}}{\tau_{+}-\tau_{-}}$? (2) Sorry, I meant "position vector with origin at $\omega(\tau_{-})$"?. $\endgroup$ – Snakeyes Oct 26 '15 at 20:19
  • $\begingroup$ That's so when you plug in $\tau = \tau_\pm$ it gives you the right answers ($\omega(\tau_-) + \omega(\tau_+) - \omega(\tau_-) = \omega(\tau_+)$). // the formula I wrote down for $\omega(\tau)$ is the equation for a line in a vector space satisfying $\omega(\tau_-)$ and $\omega(\tau_+)$ are the requisite values. $\endgroup$ – Willie Wong Oct 26 '15 at 21:09
  • $\begingroup$ Right! Am I correct in assuming, that all I need to show is $\langle \omega (\frac{1}{2}(\tau_{+}-\tau_{-}))-p, \omega(\tau_{+})-\omega(\tau_{-}) \rangle = 0$? $\endgroup$ – Snakeyes Oct 26 '15 at 21:50

Playing really fast and loose with some notation, in units where $c=1$, the triangle $(\tau^{-},p,\tau^{+})$ is a right triangle, with the right angle at $p$. $pq$ bisects $\tau^{-} \tau^{+}$ and produces two new triangles. Geometry will show that these two triangles are symmetric. Geometry will also show that if you bisect the angle $p q \tau^{+}$, the bisector will be parallel to $\tau^{-} p$. That is, it will be parallel to a light ray. Geometry will also show that the angle between $pq$ and this light ray is equal to the angle between the $\omega$ and this same light ray. This means that $pq$ and $\omega$ are normal in the relativistic sense (not "perpendicular" in the Euclidean sense).


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.