I have noticed that historical or brief introductions of special relativity will discuss it in terms of inertial frames and postulates:

  • Principle of Relativity - (from Einstein's 1905 paper) "the same laws of electrodynamics and optics will be valid for all frames of reference for which the equations of mechanics hold good"
  • Constant speed of light - "light is always propagated in empty space with a definite velocity c which is independent of the state of motion of the emitting body"

While modern descriptions state it instead as the symmetries of the physical laws and space-time. For example, Poincare symmetry of the action.

While these are obviously compatible, does their content differ slightly in precision and reach? Or are they entirely equivalent, differing only in pedagogy?

For example, some thoughts

  • Can we derive angular momentum conservation from the first, or must we just take that as a consequence of what the physical laws happened to be?
  • I could see how the first could be claimed to predict parity symmetry, but not the later.
  • 3
    $\begingroup$ Perhaps this is better suited for History of Science and Mathematics? $\endgroup$
    – Danu
    Commented Jan 8, 2016 at 9:40
  • $\begingroup$ Thank you, I was not aware of that site. I'm not even completely sure if the perceived distinction I'm making is just purely pedagogical or an actual change in precision (the parity point seems to convince me). So with this level of subtlety in the physics content, I still feel it is slightly more appropriate here with many physicists but a history tag, versus there with more historians but less physicists. But I can understand if you guys decide to move it. $\endgroup$
    – BuddyJohn
    Commented Jan 8, 2016 at 9:58
  • $\begingroup$ I'm not sure how anyone could answer the bulleted questions, as it's asking for opinions of deceased persons. $\endgroup$
    – Kyle Kanos
    Commented Jan 8, 2016 at 11:24
  • 1
    $\begingroup$ @KyleKanos The bulletted questions are clearly marked as rhetorical. The point of them was to help highlight how the different presentations may differ in content. The two are obviously compatible, but are they entirely equivalent? $\endgroup$
    – BuddyJohn
    Commented Jan 8, 2016 at 11:30
  • 2
    $\begingroup$ @JohnRennie I removed discussion of history like "how things changed". It now focuses solely now on comparing the physics content of the two. Is that more appropriate now? If not, please revert to the previous wording and move to the history of science site. $\endgroup$
    – BuddyJohn
    Commented Jan 8, 2016 at 11:49

2 Answers 2


The Lorentz group and Poincaré group symmetries are a more general starting point. If you take the historical postulates as a starting point (which most intro special relativity course still do I think), you can, for example, derive time dilation using the light clock argument and the constancy of $c$ (which derives directly from the principle of relativity):

$$t' = \gamma t $$

You can extend this argument to derive length contraction by considering a rod of length $l$ in its rest frame moving relative to clock at speed $v$.

$$ l' = \frac{l}{\gamma} $$

From these you can derive the Lorentz boosts along a single axis by considering the relative position of an event in two different reference frames, and you get the usual:

$$ t' = \gamma\left(t - \beta x\right) \\ x' = \gamma\left(x - \beta ct\right) $$

In matrix form it's clear that this is a hyperbolic rotation:

$$ \begin{pmatrix} \gamma & -\beta\gamma & 0 & 0 \\ -\beta\gamma & \gamma & 0 & 0 \\ 0 & 0 & 1 & 0 \\ 0 & 0 & 0 & 1 \end{pmatrix} \begin{pmatrix} ct \\ x \\ y \\ z \end{pmatrix} = \begin{pmatrix} ct' \\ x' \\ y' \\ z' \end{pmatrix} $$

with $\gamma = \cosh\phi$. Given that this holds true for all time-space coordinate pairs, it suggests that each time-space pair forms a 2D hyperbolic subspace. Hence the $s^2 = (ct)^2 - x^2 - y^2 - z^2$ metric. However, implicit here is the assumption that the 3D spatial subspace is Euclidean and has standard $\mathrm{SO}(3)$ rotational symmetry. So no, you can't derive conservation of angular momentum from the postulates, it has to be taken a priori.

The main problem with this approach is that it limits you to considering only spacetime coordinates and 'real' spacetime vectors and tensors (i.e., things that live in the tangent spaces to the Lorentzian manifold). There's nothing here about spinors, for example. It's useful pedagogy since it motivates the Poincaré group as the correct symmetry group, but once you've reached that point you can use the group theory as the starting point.

By taking the Poincaré group as the fundamental object in physics, you can not only derive the form that spacetime must take, but you also get the unitary representations for quantum mechanics, and the spinor representation for spin $\frac{1}{2}$ fermion fields. Also parity symmetry is included in the Poincaré group. It's also easier to generalise physics to different symmetry groups for higher dimensional spaces if you use Poincaré as a starting point, since your maths is already constructed around building up from group theory.

  • $\begingroup$ Just to add a comment: yes, you could take the Poincare group as the fundamental object in physics (which is how it is done), but if a student accepts this he will have a hard time to understand GR (because then we tell him its not). I believe the historical approach to SR is better if one wants to go beyond SR later. $\endgroup$
    – lalala
    Commented Apr 13, 2017 at 11:44
  • $\begingroup$ The historical perspective is definitely better from a pedagogic point of view (it usually is: e.g., learning QFT after first learning classical FT), but for actually doing theoretical physics I think the group theory is a better starting point. $\endgroup$
    – gautampk
    Commented Apr 13, 2017 at 12:26
  • $\begingroup$ I like the part about spinors, which really brings in the advantage of group theory. I don't agree with the comments on parity though. I would not consider parity in the Poincare group (I'd consider it just boosts, rotations, translations). Considering parity part of Poincare, would suggest special relativity predicts parity symmetry, which I don't think you meant to imply. $\endgroup$
    – BuddyJohn
    Commented Apr 20, 2017 at 10:02
  • $\begingroup$ @BuddyJohn The full Lorentz group does contain the parity transformation, which corresponds to PT symmetry in particle physics speak (you can't just have P or T in the Lorentz group separately because obviously the metric will change form under just P or T). $\endgroup$
    – gautampk
    Commented Apr 20, 2017 at 11:39
  • $\begingroup$ Starting from an inertial frame, a coordinate change like $t' = -t$ would change a sign in a component of a four-vector, but would not affect the metric (nor would a parity transformation) as the diagonal elements would incorporate any sign change twice and so be unaffected. The full Lorentz group $O(1,3)$ does have P, T, rotations, and boosts. However, my understanding is that special relativity is just $SO^+(1,3)$ and translations, which is Poincare symmetry. The standard model has the symmetry expected from Special Relativity, but not full $O(1,3)$ symmetry. $\endgroup$
    – BuddyJohn
    Commented Apr 21, 2017 at 5:09

While the historical postulate approach is entirely compatible with the modern symmetry approach, issues such as the one of parity symmetry you bring up, make it clear that they are not entirely equivalent in their precision.

If you postulate the physical laws are the same in all inertial frames and the speed of light is the same in all inertial frames, to be precise this begs the question of what the precise definition of an inertial frame is. This is one of those concepts we intuitively understand, but which would be very difficult to make precise. Starting from these principles, and mainly drawing from experience from electrodynamics, it would be very tempting to claim special relativity predicts time reversal invariance and parity symmetry.

Then when parity violation is later observed, how would one adjust their understanding to this? Abandon SR, or just absorb it into the definition of inertial frame, thus preserving SR? Again, to avoid these issues, if using the historical postulates approach an inertial frame would need to be defined precisely. This would likely involve a procedural definition, thus depend on the physics, and so one would have to be very careful not to reason circularly. And parts of it would feel very artificial, like demanding an inertial coordinate system was right-handed instead of left-handed.

Postulating a symmetry is much more precise. Does Poincare symmetry include parity transformations? No. Therefore SR doesn't predict one way or the other regarding parity symmetry. Simple, done.

Continuing along the idea of trying to absorb broken symmetries into the definition of an inertial frame, what if one found there was a preferred direction, but Lorentz boosts perpendicular to this axis was still a symmetry? Sure I guess if part of physics involved this "conserved vector", then we could still define an infinite set of frames which are "inertial" frames, move relative to each other, and physics looked the same. In the parity case, it would be, do you want to demand inertial frames using a right-handed coordinate system? Here it would be, do you want to demand that space looks the same in all directions? As you can see, answering such questions in trying to precisely define an inertial frame, is really just inching our way towards the symmetry case.

So while it would be straining credulity: it depends on how you define an inertial frame, whether we could obtain full rotational symmetry of physics from the postulates. Did you include rotational symmetry in the definition or not? And in that sense, it would not even be a derivation, it just has to be taken a priori.


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.