@Shuler:  "I'm well aware GR people fall back on "it can't be visualized," but I do not believe that."

I do not say that, and I understand the thing as completely as is ever possible short of doing the calculation. However, you will admit that it is difficult for us to imagine a curved 3D space. Whatever you need in terms of diagrams is there in the papers by Comer and Lathrop and by Ehlers and Rindler that I cited.

Your citation of Carlip: ""We have another way of "weighing" kinetic energy: we can send a beam of particles past a large mass (the Sun, say) and see how it is deflected. It is well known that the deflection of light is twice that predicted by Newtonian theory; in this sense, at least, light falls with twice the acceleration of ordinary "slow" matter."

This is precisely what I meant: It is *only* in this sense that light falls twice as fast than particles. If you divide off the prefactor that goes as 1/v2 and is therefore larger for particles than for light, then you end up with a bigger factor over Newtonian theory for light than for particles. For light that factor is 2, for particles with v

I do not have an answer to this question, though I do believe it might be a very important one. I read some Julian Barbour hoping he would have something interesting to say about it, but could not find what I was looking for.

Perhaps I can offer a relevant observation: the most general group of spacetime transformations that leave light cones (and, in 4 dimensions, Maxwell's equations) invariant is the conformal group, which contains the Lorentz-Poincaré group. So the statements, "Maxwell's equations are valid" vs. "the laws of physics are conformally invariant" are in some sense equivalent. A good question is why we restrict the invariance to the Lorentz-Poincaré group, but it is not hard to see that the conformal part can be absorbed into a variable gravitational constant, which is constrained by observational limits. (On the other hand, as early as the 1940s, Hill (PRD 68, 232, 1945) proposed that the apparent expansion of the universe may, in fact, be an artifact of conformal geometry.)

Is there an advantage to absorbing a conformal factor into G, rather than leaving it 'exposed' in the metric, as in FRW Cosmologies?

In any case, I take the point that the conformal group of transformations might be a better place to start in the search for an alternative epistemology for special relativity. It struck me as interesting for example that just as one can 'derive' (the possibility of) an 'omega-cubed' spectrum characteristic of the EM ZPF based on the requirement of Lorentz Invariance, invariance under Cosmological expansion achieves the same end in that regard.

I'm not sure if I'd call it an advantage, but conformally transforming the metric amounts to physically reinterpreting observations: e.g., what appears as a variable-G in one frame may appear as a scalar field coupled to matter in another. Having said that, I believe that even Maxwell's theory ceases to be conformally invariant in the presence of charges, so we don't necessarily have the freedom to pick any conformal frame we want without observational consequences. (On the other hand, we are free to perform an arbitrary Poincaré transformation.)

It's interesting how quickly Maxwell's equations came up in your conversation. It's hard to distinguish them as a postulate different from constant lightspeed in modern thinking (i.e. without ether thinking). I started with isotropically uniform lightspeed in only one reference frame, and a rule for electromagnetic coupling to moving objects. I don't know if the rule can be derived from Maxwell's equations. But if there happens to be any non-electromagnetic energy, it gets dragged into the derivation by conservation of momentum and energy.

I'd like to raise an important point. Given a thrice differentiable vector field and a metric, Maxwell's equations are simply identities that are always satisfied (i.e., they don't need to be postulated or stated as a rule.) Or rather, either Maxwell's or Proca's equations are satisfied depending on the definition of current. Now if the current is defined such that the vector field remains massless, this definition survives a conformal transformation. A conformal transformation is, of course, the most general group of transformations that leaves light cones invariant. So perhaps these concepts really are inseparable, just different ways of stating more or less the same thing?

The necessary conditions that I found were that the speed of electromagnetic energy propagation be isometric in at least one reference frame, and that the interaction of two things with moving objects vary as a geometric mean. These two things are applied force along the direction of motion, and the length of interatomic bonds whenever the phase velocity of propagation is not the same in both directions (in which case it is the geometric mean of the two velocities). By conservation of momentum alone the first assumption extends to all angles and all types of interaction that produce distinguishable and measurable momentum or energy. The bond length, which I figure likely is a wave phenomenon, (they both likely are) remains true only in the direction of motion. See draft writeup linked below question if this summary is too high level. Anyway, these two assumptions say nothing really about Maxwell's equations. I would be curious if they can be derived from Maxwell's equations. Originally, of course, Maxwell envisioned an ether. He died 8 years before the Michelson-Morley experiment.

@Mozafar, since the Einstein version of the relativity principle, in my understanding, includes the velocity of light as well as other laws of physics, that's why. Did you have a different view of the relativity principle?

I disagree Mozafar, and it is my question. : )

I don't expect to change your opinion. However, the principle of relativity is sweeping, and we have not and cannot ever validate it by direct observation, because to do so we'd have to make measurements in every possible reference frame. It is much preferable to derive it from a smaller and more easily verified set of assumptions. I have in fact put forward, linked in the original question, a strawman derivation from assumptions that are for only one particular reference frame, and therefore much easier to validate. From this essentially I derive the relativity principle as a result.

I am asking if anyone knows of other attempts to do this, or lacking that, of any proofs that it is not possible. I am not interested in philosophical justification for an assumption that must be applied to all reference frames because that is already the status quo for 108 years and is not new. Thanks.

The constancy of the speed of light is outside Relativity, in fact Einstein was obliged to introduce the second postulate generating like this a different theory (SR). Also General Relativity has little or nothing in common with Relativity. The question is if these theories are useful, of course for more than a century greater part of scientists believed them also because it seemed those theories were able to explain physical events that other theories were unable to explain. Now we know other more satisfactory theories are possible.

Relativity is beautiful and simple and it claims only laws of physics are invariant for inertial reference frames. Consequently also many quantities are invariant (for instance applied forces, mass of mechanical systems, many constants) but other quantities are non-invariant. I refer to speed in general and to speed of light in particular, I refer to frequency and wavelength of the Doppler effect, I refer to electrodynamic mass, etc. We live a time with little certainties and also physics is subjected to this guideline, but it is sure physics will be able to solve those questions. It is only a matter of time, meantime we contribute to clarification.

Good Holidays and Happy New Year!

Dear Robert

I agree with you that sometimes "It's hard to distinguish Maxwell's equations" as a postulate different from constant lightspeed in modern thinking. Nevertheless, it is perhaps interesting to note that the so called Lorentz transformations had been encountered before Lorentz by Voigt in 1887 in a somehow different context. Voigt simply starts with a (scalar) wave equation and finds some linear and homogeneous transformations which keep its form intact. Thus, in Voigt's sense, Lorentz transformations emerge as a group of transformations linked with the "form conservation" of a homogeneous, linear, second order, hyperbolic equation (nothing more than this!). I think it is clearer to start the discussion with the wave equation itself rather than with Maxwell's equation (which reduce to the wave equation only in some particular instance). There is a nice discussion about this issue in the classic book by Rene Dugas, History of Mechanics.

The "constancy" of light velocity cannot be derived, it is assumed.

@ V. Toth

Your 'important point' is remarkable and it surprises me. Can you explain the precise statement? As I understand your statement, it can't be valid: With respect to an assumed global chart (thus effectively on R^4) I specify any 4 functions R^4->R as components of a vector field and diag(1,-1,-1,-1) as a metric. By stipulating the transformation properties of a vector field and of a symmetric covariant tensor-field everything is defined on a manifold. No chance that Maxwells equations are automatically satisfied. What has to be added to my scenario to make your claim valid?

@E. Goulart, yes I found the history about Voigt and he is referenced in the paper I linked with my question above. Thanks for the reference, I'll try to find it. So far I have only discovered that electromagnetic interactions (forces) with moving particles obey a geometric effectiveness law relative to plus and minus the velocity. More detail on how this comes about would be interesting, especially from odd angles. But I imagine the math is pretty complex.

@RL, an interesting proposition. But why not just invert any possible answer to match the inverted question? For example, if I, or anyone, proposes that a b and c conditions imply the principle of relativity, and the logic steps are reversible (I suspect they will be), then the condition for producing a non-relativistic physics would be (not a or not b or not c).

Ulrich: I just saw your request for an explanation. I explain this stuff about Maxwell's equations on my Web site; see https://www.vttoth.com/CMS/physics-notes/75. But the idea is not mine, I am just repeating other people's wisdom. The gist of it is that the exterior derivative operator (let me denote it with d) is nilpotent, so given any twice differentiable 4-vector field A, d^2A = 0. There go two of Maxwell's equations, automatically true. Now, if you have a metric, you can form the Hodge dual (denoted with a star, *). Defining F = dA, we can define the 4-current J = *d*F, which is conserved, so long as A is three times differentiable: *d*J = *d**d*F = *d^2*F = 0. These are your other two Maxwell equations.

One way to absorb this is to realize that Maxwell's equations do not constrain the 4-vector field A, they only constrain the components of the Maxwell tensor F. However, if F is the exterior derivative of a vector field A, its components automatically satisfy Maxwell's equations.

Lastly, a different definition of the current can lead to a different theory, e.g., J - m^2A = *d*F yields Proca's massive electromagnetism.

Hello Robert (RL), thank you for your thorough analysis and comments!

I assumed you were serious, and I was just asking for a little more explanation and justification because I was not at all sure I saw, from a purely logic and math point of view, what was changed. If one proved only sufficient conditions, for example, then the reverse implication could not be made and the question could not be turned around. Or I suppose certain types of proofs might not go in both directions, about which you would know more than I would.

If I came across with a bit of an undercut, please blame my trainers. I was indoctrinated in Godel and Turing from my undergrad days. I remember sitting at lunch with Dr. Trammel - probably no other students ate lunch with this intimidating gentleman but I found him fascinating - and he was educating me about Godel. I know that you don't think so much of Godel's theorem, you have said so, but it is my training to always take the reverse of a proposition and feed it back into the proposition just to see if anything peculiar happens. ; )

Now you seem to be arguing more philosophically. For that I have an answer. I doubt you will agree. The mathematical structure of Minkowski space does not seem to me to correspond to a physical thing. I have cited papers by Swann, for example, that clocks do not automatically assume the right time. Swann believed a quantum system would synchronize itself. I believe that it does and that is what produces length contraction. 'I doubt that Minkowski space actually exists. I just read a paper this evening by Stefanovich regarding Minkowski space as only a useful model which in certain cases conflicts with Relativistic Quantum Dynamics (a term he defines, I don't know the QM field terribly well.)

http://link.springer.com/article/10.1023%2FA%3A1016052825257

http://www.geocities.ws/meopemuk/FOPpaper.html (free version)

I have done quite a bit of analysis using the Equivalence Principle, developing a 2011 paper demonstrating the isotropy of inertia by that method. Earlier this year I published a more complete, but not fully complete, theory of inertia based on quantum measurement-like interactions. It was not complete because I could not really explain the relativistic inertial mass increase described by SR. I had to use Length Contraction at a minimum. That stuck in my craw, so to speak. I wanted a complete theory. I thought perhaps I'd find an interaction basis for length contraction. That did not pan out. I resorted to trying to find an interaction basis for Lorentzian inertia directly.

That is very easy to do. Probably someone has done it if one looked hard enough. But it always ends with a preferred frame. My instinct said there was probably not a preferred frame, at least as far as the speed of light was concerned. So I posted this question thread to get a feel for the lay of the land, what has been done, what people are thinking about this. A lot think like you, but for some reason it intrigues to post on the subject and I hope you will continue.

Addendum to RL regarding momentum conservation under extreme circumstances. This comes from almost any gravity theory that involves time dilation, and also from Special Relativity. So yes it is using a bit of what comes out of relativity theory, I won't deny that. On the one hand I've tried to avoid being blatantly circular. One cannot start with length contraction, mass increase, or time dilation. But just kind of arbitrarily I'm viewing anything else as fair game if there is some rational for it. I'm only trying to peel back one or two layers, along the lines you indicated of simpler assumptions.

@ALL, Here is an interesting link to a condensed matter paper on Lorentz transforms at the speed of sound in a fishbowl world that "can't" detect the existence of an outside world with faster energy transfer: http://arxiv.org/abs/0705.4652 (I've only just skimmed it)

http://arxiv.org/abs/0705.4652

@ V.Toth

Thank you for your lucid explanation.

If your first statement had been

'Given a thrice differentiable vector field A and a metric g, Maxwell's equations are always satisfied for A playing the role of the 4-vector potential and a 4-current constructed from A and g in a natural manner.'

then it would not have suggested a (actually non-existing) conflict between the predictive character of Maxwells equations (you can compute the radio waves from the antenna currents!) and your statement.

@Robert Low - it doesn't sound like you use Minkowski space as a physical thing. I suppose such [the physical] would be a point of view more likely a physicist would take. Since it wouldn't occur to you, of course it would puzzle you why anyone cared, so that is what you were asking, right? Did I answer satisfactorily?

@Mozafar & RL, are you guys trying to (A) derive a particular value for the speed of light, (B) determine why the speed of light, whatever value it is, should be the speed limit, or (C) determine that there is a speed limit which is empirically and coincidentally found to be the speed of light (in a vacuum)?

@V. Toth, since you seem to be quite handy with EM, may I ask you a question? I have always assumed that all EM effects and fields propagate at the speed of light. Is that still viewed as correct? I have run across two papers that speak as if there were an instantaneous component of the Coulomb effect which could not be viewed as retarded and apparently some controversy. For example it is a driver for the discussion in a paper on relativistic quantum mechanics which I'll link below.

http://www.geocities.ws/meopemuk/FOPpaper.html

@RL, very good answer on speed of light, are you sure you are really a mathematician and not a closet physicist? ; )

Hi Robert, I ran across what looks to be a paper of yours from 1990, from an era in which perhaps you were more aggressively mathematical? I can only see the abstract, though. It appears to be an unusual and speculative particular case of the twins paradox. Is it anything you'd like to tell us about, either here or on Ivan's QA?

http://iopscience.iop.org/0143-0807/11/1/003

Robert S: In classical EM, I don't think there are any serious doubts that there are no instantaneous or superluminal effects. It is of course known that action-at-a-distance is not compatible with relativity, but classical EM is a field theory, so action-at-a-distance does not apply.

In the context of QFT, I think Peskin and Schroeder in their 1995 book provide a very lucid explanation of why QFT (not just QED, QFT in general) preserves causality. As a matter of fact, I think their book directly answers the concerns raised by Stefanovich, but I have not studied Stefanovich's paper in detail so I don't feel like I am in a position to critique it.

@V. Toth, thanks for comments. I don't think Stefanovich (or anyone) thinks the supposed instantaneous effects actually violate causality, but like you I haven't studied enough to be sure. Don't know if you've heard of [the late] Van Flandern, an astronomer who sort of went over to the dark side, but he goes on and on about how GR does not really absolve the long-known instantaneous aspects of gravity (gravity vector points to current position of sun and other planets, not retarded position). However, it's possible to argue that since gravity affects everything (universal potential, aka the equivalence principle) it can't actually be used for FTL communication. This is [supposedly] a separate matter from gravitational waves propagating at the speed of light, which are undetected even though sensors are supposedly sensitive enough.

@Mozafar re: lapse, no problem, you just own me one free pass. ; )

Thought for the day . . . Einstein's 1918 solution to the Twin Clock Paradox using a uniform gravitational field does not correspond to the stated problem, because it substitutes an absolute (gravitational) time effect for a relative one, thereby de-synchronizing all of A's reference frame clocks (assuming A bothered to keep any, and I think we can just add them to the problem). This is an effect that A can easily detect by just re-synchronizing and finding whoops! My clocks are all in error in proportion to distance from origin along B's line of flight! Obviously this wouldn't happen in the original rocket setup. How can such a large mistake go unnoticed for 95 years! Even if we exempt Einstein from critical scrutiny from relativists, there are enough semi-literate radicals out there that one of them should have discovered this, but I see no mention of it.

@RL, thanks for link to paper above. It made me realize that I need not classify the question of age in a one-way twins' trip meaningless. And further that the new method I've worked out gives the correct answer in this case. Thanks very much.

Robert S: Yes, I've heard of Van Flandern's argument about gravity propagation. I don't think it can be taken seriously. See, e.g., http://arxiv.org/abs/gr-qc/9909087.

@V. Toth, thanks for the link to the excellent paper. Answered a lot of questions I have been wondering about. Unfortunately I'm not able to develop the math for myself at that level, but the paper is very well written at multiple levels. I sorely need someone to help me with my inertia theory (which is beyond the scope of this thread).

I noticed our friend Robert Low among the references. He keeps popping up everywhere. : ) It is on a subject of interest to me in connection with NASA work, and I think I'll ask him for a copy.

It seems, from the paper you linked, that binary pulsar decay from gravitational radiation is verified to higher accuracy than I might have guessed. Though the paper didn't directly address that. Only that one parameter was verified to within 1%. I wonder if the energy could be lost by other means, such as by near field coupling to nearby objects? Maybe that's not the right terminology, but I'm trained as an EE not as an astrophysicist.

@V. Toth, to be sure I understand, I made a powerpoint animation of the field-charge relationship, only at one distance and very crudely due to limitations of powerpoint animations (or my skill with them). There is a naive incorrect one, followed by what I understand to be correct. I simply had never thought about it in exactly those terms before, and realize now that in the past I have gravitated to one or the other of them on different occasions inconsistently. Fortunately not in print. :D I'm a little surprised a reviewer would not have asked van Flandern to address this. He was held to a different standard than I, for sure.

Does anyone here know if gravity has been experimentally shown to be Lorentz invariant, or if not what a test might look like? I'm thinking it may be difficult, if it has not already been done. But maybe some astronomical effects provide data from which it can be deduced?

Robert S: I think it'd be very difficult for a binary pulsar to mimic GR. Indeed, even for modified gravity theories, binary pulsar data can represent a serious headache. That is because in GR, gravitational radiation is quadrupole and higher order (there are no gravitational dipoles, and monopoles don't radiate thanks to Birkhoff's theorem.) If a theory predicts, e.g., dipole radiation, it would likely be completely out of whack with binary pulsar data. This can badly affect scalar-tensor or vector-tensor theories.

@V. Toth, have you ever heard of Wheeler-Feynman Electrodynamics? Apparently it was an alternative to Maxwell with action at a distance, intended to solve some normalization problems or something, but turned out to be hard to quantize and was not used. Noticed it in this old paper: http://www.rand.org/content/dam/rand/pubs/research_memoranda/2006/RM2820.pdf

@V. Toth, thanks for the analysis of gravitational radiation problems at lower orders. My trouble is rather different, though. I don't yet have a well quantified field, but it works through uncertainty of momentum and position rather than time and energy like a normal quantum field. By near field I mean not really radiation. I find it hard to get radiation, but I may just not know how. But on the count of explaining absence of detection of gravitational waves, I can do wonders. Obviously Feynman and Wheeler did for their electrodynamics theory. Most theories of gravity will not have the lower order radiative terms, I would guess, because of various peculiarities of gravity.

Uh, this question is way out of scope, but I could not resist.

Robert S: Interesting paper by Coleman. May I recommend the book, Classical Charged Particles, by Fritz Rohrlich, originally published in 1965. It discusses the theory of Feynman and Wheeler, among others. The most recent edition was in 2007, I believe. Personally, I always felt that this was an overblown issue (there is no such thing as a classical point particle so why should I worry about its properties?) but that's perhaps just my own ignorance. (Then again, I did buy Rohrlich's book, so maybe I was more interested than I am willing to admit.) Anyhow, Rohrlich also explains why Feynman's and Wheeler's approach was basically superseded by subsequent research.

V. Toth, thanks for recommendations!

If from some other fundamental theory or mechanism, one can derive the constancy of the speed of light, surely there should be new predictions which can be tested in experiments. Are there any novel predictions from this new theory?

@Patrick Das Gupta, excellent question, you make a perfect "straight man." Do you know the expression?

@Robert Low, the answer I will give to Patrick also constitutes a "new and improved" answer to your question about "why look at this?"

Please click the length below as the answer, still a short note, became awkward to edit in this text box. The general title of the response is:

LORENTZIAN STIFFNESS VS. COORDINATE TIME

https://www.researchgate.net/publication/259444207_LORENTZIAN_STIFFNESS_VS_COORDINATE_TIME?ev=prf_pub

Thanks @Robert Shuler.

Hello Robert,

The Lorentz transformation is a 'natural' symmetry of the Lorentz-Maxwell equations. It is hard to imagine a mechanism without knowing/understanding what the electrical charge is.

BTW, Einstein's 1st postulate is a trivia, actually novel is his 2nd postulate. As it is a kinematic statement, it applies to kinematics only. Its application to dynamics is an additional postulate.

Alternatively, one can generalize Euler's ansatz for the change of the velocity, v,

dv = F/m dt

to

d[f(v) v] = F/m dt

Here, f(v) describes the dependence of dv on v. This leads to f = Lorentz factor.

Hi Peter, thanks for your comments. In my lifespan, I've seen the estimation of the first postulate go from "enormous" to "trivial," the latter of which you suggest. I suggest that this is only because it is introduced to physicists educated since about 1970'ish in the same manner as the postulates of geometry or algebra, as an obvious and unquestionable truth without which we won't get far. The odd thing is, Einstein himself attempted to undo the earlier postulates of geometry. And the first postulate, once the second is known, turns out to be a self-fulfilling prophecy. If one had postulated differently one would get a different answer. This is explained further in the link I posted above on LORENTZ STIFFNESS VS. COORDINATE TIME.

This is even more true with the second postulate, about light. If one utilizes any other measurement technique than clocks synchronized by ... light! ... then one gets a very different answer. There is a lot of detail about this in a draft paper I wrote a few weeks ago, a critical history of the Twin Clock Paradox, with new visualizations and methods, and the errors exposed in Einstein's and many other proposed solutions that arrive at the right answer for the two clocks about which the question was posed, but the wrong answer about any other clock one cares to introduce. It is linked below.

@Charles, Special Relativity holds quite well in the presence of a much faster force, if the faster force is not selective (e.g. like gravity). We have no way of determining the speed of gravity at present due to the relative weakness of its effect and the universality of its effect and the consequent large expenditures of energy we'd need to produce effects that register on our instruments.

However, two things immediately need to be pointed out:

1. I have for several years been working on a proof that if gravity propagated at a higher speed, we'd still only be able to perceive the locus of action at a speed-of-light information rate, because we are made out of the "light stuff," i.e. the 3 unified forces.

2. Stellar events would seem to produce sufficient gravitational disturbance for our instruments to detect it, and they lose energy at the correct rate to suggest they are radiating, but we cannot detect it even though our instruments have long been sufficiently sensitive. However, if my guess is correct, that gravity propagates faster, then the wavelength is correspondingly longer, many magnitudes longer than the instruments, and this completely explains the lack of detection.

Charles, yes that is what I'm referring to. So far attempts to validate that part of the theory are unsuccessful. There is no measurement. If we could detect gravitational waves and correlate them with a simultaneous optical event, then that would greatly narrow the possibilities. But the current instruments should be sensitive enough and have not detected them. A negative result is harder to confirm, so I imagine either we'll find them, or keep looking for quite some time.

Meanwhile, on the theoretical side, any faster acting force is difficult to detect with instruments built out of the 3 unified forces with demonstrated Lorentz stiffness "c". But a non-selective one is especially difficult. One with a random component is also difficult to detect, and until the Bell inequality and other versions of it were developed there was not even hope, but in the last couple of decades we seem to be closing in on confirmation of "spooky action," whatever force that is. There is another thread on RG about the mechanism of entanglement, and it has degenerated into a lot of arguing. So has the one about the Twins in which you also participate, even though for that one I can give definitive answers entirely within classical SR. It is very frustrating.

Interesting citation. In a quick search I did not see the original paper because it was hidden by pages and pages of "comments on" papers, most of them opposed in one way or another, e.g. http://www.sciencedirect.com/science/article/pii/S1384107606000479

I have no personal opinion.

Hi Charles, so that paper does not take into account spiral structure, you are saying? I did a little looking into this a while back. My tendency was to presume that the "experts" have taken all such things into account in determining galactic rotation curves. But while I can find explanations of the very odd orbits that "probably" explain stable spirals (which are not particularly Keplerian and not entirely dominated by a central mass), I have not seen a good accounting for the non-astrophysicist of what the rotation curve "should be" taking into account the odd orbits, vs. what is measured. Also the measurements themselves would wind up with data from various parts of orbits, and I've seen no explanation of how that is sorted out.

Charles, thank you for this information. It is the clearest I have seen. Is there in it somewhere a clear summary of the "current status" of discrepancy in galactic rotation curves, or is the computation still not complete?

IMPORTANT REFERENCES - I have just become aware of the literature on "constructivist" or "dynamical" special relativity, with which there is a good deal of overlap with this question. Two of these are online and two others are a bit harder to find:

Swann 1941 Rev. Mod. Phys. 13, 197–202 (1941)

Relativity, the Fitzgerald-Lorentz Contraction, and Quantum Theory

http://rmp.aps.org/abstract/RMP/v13/i3/p197_1

Bell 1976 How to teach special relativity

http://philpapers.org/rec/BELHTT

Brown & Pooley 2001 The origins of the spacetime metric: Bell's lorentzian pedagogy and its significance in general relativity

http://philpapers.org/rec/BROTOO-3

http://philsci-archive.pitt.edu/1385/1/9908048.pdf

Miller 2013 A constructive approach to the special theory of relativity

https://www.researchgate.net/publication/45860648_A_constructive_approach_to_the_special_theory_of_relativity?ev=pub_cit

Article A constructive approach to the special theory of relativity

Необходимо помнить, что эйнштейн на протяжении всей своей жизни подчеркивал, что опыт майкельсона-морли никак на его теории не сказался. На момент создания своей теории от жил и работал в берне. А это город всемирно известных часовщиков. И чтобы у них у всех было одинаковое время, они придумали синхронизацию - за точное время брали среднее время между уходом и приходом звукового сигнала. Эйнштейн здесь просто скорость звука заменил на скорость света. Его гениальность прявилось в том, что скорость звука зависит от скорости источника звука. Эйнштейн же сказал, что скорость света не завистит от скорости источника. Таким образом он получил всю современную физику - синхронизировал измерение времени и сказал свету, чтобы его скорость не зависела от скорости источника. Можно сказать и по другому - он объединил физику и математику. Физически он ввел метод измерения, а математически он указал на существование инварианта.

@Vasiliy, Thanks for interesting story about synchronization of watchmakers in Bern. Спасибо за интересный рассказ о синхронизации часовщиков в Берне.

Historically, there are two approaches that were taken to explain the Michelson-Morley experimental result. First, there was one group of people, particularly in Great Britain such as A.S. Eddington, who tried to physically explain the result through the Lorentz-Fitzgerald contraction. The contraction was supposed to be the result of the response of the physical space that surrounds the electrons of the material making up the experimental apparatus to its motion. This contraction, essentially given by the Lorentz transformation equations, was such that the measured speed of light is always the same. Another group of people, particularly the Germans, were content to simply say that the constant speed of light is a fundamental law of physics. I tend to agree with the latter approach - after all, when Maxwell wrote out his equations, the theory did not a priori assume a particular frame of reference. (Implicitly, the observer was stationary to the frame of reference to which the electric and magnetic fields were to be measured.) The fundamental constant of the speed of light fell out of the theory such as, e.g., Planck's constant falls out of the theory of radiant heat. The requirement (i.e., the law) that the speed of light is the same for all (inertial) observers brings a symmetry into the scenario by unconditionally connecting the state of the observer with the event being observed. By approaching the issue from this path, one is ultimately lead to the fact that the structure of physics at the observational level is governed by the associative Clifford algebra C(4) along with its various implications. Eddington started to arrive at this result in his very hard to read "Fundamental Theory". For a more straightforward development, see the work by E. W. Bastin and C. W. Kilmister originally done in the 1950's. More recently, a paper by N. Salingaros (Foundations of Physics, vol. 15, #6, pp. 683-690 (1985)) reviews these results.

@Robert Manning, thanks for your discussion. Einstein in 1918 in an article in the NY Times made an analogy with the theory of gases. There is a constructive theory, the kinetic theory of gases, and a principled theory which is much less detailed, the 2nd law of thermodynamics. He stated that we feel more like we have explained something when we have a constructive theory, but that at the time (1918) no one knew what the constructive theory of rigid bodies should be. (EM obviously did not explain why electrons did not decay into the nuclei, and QM had not been developed.) So he in desperation chose the principle approach to relativity. He did not say a constructive theory would never be found. Planck thought inquiry into a constructive theory was necessary. Later J. S. Bell and several others advocated for teaching SR using an incomplete or merely analogous constructive approach. (I do not necessarily agree with that since the approach used is usually EM which is complicated and known to be flawed.) But none of this implies a constructive theory is not possible or shouldn't be looked into. When I was phrasing my question, I had not yet learned this terminology. (So the Q&A has been very useful, and thanks to everyone contributing!)

I think when people say they want to develop a quantum theory of gravity, they are basically saying they want a constructivist theory ... how does one construct the GR space time. What I am saying is, GR is based on SR which is much simpler. So, what is a modern constructive theory of SR look like?

Я прочитал выше написанные сообщения. Я не могу согласиться с тем, что в настоящее время можно создать кватовую теорию гравитации, нет хороших экспериментальных данных. Эйнштейн основывался на хорошо усвоенные за несколько сот лет принципы - бернские часовщики научились синхронизировать время, а со времен ньютона физики привыкли пользоваться инвариантами - энергией, импульсом, вращением и другими. Видим, что история учит - нельзя использовать только одни инварианты, как сейчас пытаются делать. Необходим еще один или несколько других принципов, никак не связанные с инвариантами. Так что, надо искать эти принципы. И здесь необъятное поле деятельности.

Конечно, согласен обо гравитации. Я только сказал что люди хотят это когдя они искать квантовая гравитация.

Robert M:

I'm not sure how much time the Germans spent worrying about how to explain the M&M result. Einstein was supposed to have been asked twice about the extent to which he was motivated by M&M, and is supposed to have replied (1) that he hadn't heard of the result at the time he was developing SR, and (2), that, okay, maybe it was possible that he //might// have heard of it by 1905, but that if so, it hadn't made any significant impression on him (or enough for him to remember it).

Consider how proud of their status many Germans were of their own research institutions, and how much they'd have tended to look down on experimental research that was not only not being conducted at a major German lab, but - gasp - wasn't even being conducted in Europe!

Also consider that even Michelson's own colleagues at Case were reported to have considered that he was "to be pitied" for not only having produced two failed experiments in a row, but having then been stupid enough to publish them!

Also consider that while Lorentz did review Michelson's first aether drift experiment's null result, Lorentz politely pointed out that Michelson had screwed up his calculations in such a way as to make the results worthless, and his first reported null result irrelevant.

Michelson went away and redid the experiment more carefully (this time with a collaborator) and reported another null result ... but since the first null result could be explained away by Michelson having screwed up, it'd have been comparatively easy to choose to interpret the second null result as being evidence of a probable second screwup, if one wanted to.

Also consider how many years passed before anyone was able to reliably replicate Michelson's null result. Einstein attributed Michelson's ability to get a result that eluded others to his being an "artist" with exceptionally keen eyesight - Einstein seems to have meant it as a genuine compliment, but if especially keen eyesight was a factor, then it also suggested that other people using the same hardware might not necessarily be able to get the same results out of it, which is a slightly troubling concept.

When one looks back at a chronological sequence of events, it's easy to assume that they correspond to a causal chain (and it's convenient for physics teachers to teach the sequence //as// a causal chain), but ordering doesn't always equate to causality.

@Eric Baird, thanks for the historical insight. As I understand it, there is still a tremendous bias against publishing a null or negative result, but it's hard to decide which claims are rational. In the light of history, I suppose MM is now transformed into a positive result, confirming Special Relativity, though actually it confirms Lorentzian Relativity just as well.

I think that a positive answer to all of your questions is possible. To be adequate, however, such an answer should be unavoidably long and multifaceted. I have recently published some papers on this topic and more in general on the link between quantum mechanics and relativity, e.g. S. Tosto, “Quantum Uncertainty and Relativity”, Progress in Physics, 2012, vol 2, pp 58-81. This paper is included in my (partial) list of publications. Regards. Sebastiano

Hi Robert S,

I think we can derive pretty much all of special relativity without presupposing the PoR, if we’re allowed to insist that a model's lightspeed geometry has to appear flat (=free from intrinsic curvature) for all observers.

@Eric, just flat, not constant?

@Sebastiano, hello I looked through several of your papers. We share the idea of using the uncertainty principle as a derivational basis, though there is only a little overlap in what we've done (light bending for example) and the derivations appear at least superficially very different. If interested to compare further see link below.

http://www.scirp.org/journal/PaperInformation.aspx?PaperID=27250

@RobertS, I //think// so. The full thing only occurred to me today (when I was thinking through your question again), so some of the reasoning might be too weak to count as a full derivation. Not sure. I've tried typing it up - sorry about the length. Needs diagrams, really.

I suppose that the condition of flat spacetime for everybody might be seen as a //sort// of principle of relativity.

@RobertS: Okay, here goes:

Imagine a stationary hollow mirrored sphere with a central point-reflector or transponder, which emits a fresh pulse of light every time one is received (or, we could try to set up a standing wave within the sphere). Imagine also that the individual rays within the sphere at specific angles can be identified (perhaps using some sort of screen or mesh that isolates beams at given angles). We’ll imagine that the apparatus has been constructed and calibrated according to the assumption (rightly or wrongly) that the speed of light has a constant value wrt the sphere – it’s constructed to reflect all the rays back to their origin, in-phase, and if it //looks// spherical when it views itself, and that’s our functional definition of its apparent “sphericality”.

We now fly past the sphere at high speed, and redescribe the propagation of a single pulse moving within the sphere.

If we’re at a sufficient distance to be able to safely assume that our own motion isn’t affecting or altering the sphere’s internal physics, then for us, the pulses are being emitted at one spacetime location and being refocused at another, which means that the geometry of the sphere’s reflecting surface, for us, has instead to have the form of an elongated ellipsoid, with rays being emitted at an event at one focus and being received at an event at the other.

In order to be able to move at the same speed as the sphere, the sphere’s transverse-aimed rays, for us, need to be described as advancing at the same speed as the sphere, which means that for us, those rays are tilted forward in the sphere’s direction of travel. Calculating the exact amount of this forward deflection, for each ray, at any given velocity, then gives us the aberration formula used by old-fashioned emission-theory, which is also, conveniently , the formula used by special relativity. So far, we haven’t assumed fixed c in our vicinity or frame, merely the assumption that our own motion hasn’t altered the propagation of light inside the distant sphere.

If we take a lengthwise cross-section of the elongated ellipsoid, it turns out that in the resulting ellipse is Lorentz-proportioned – it has to have a major diameter that is greater than its minor diameter by the Lorentz factor. Also, if we measure off the ray-distances for each beam of light between its emission point and its perimeter reflection point, that measurement should then give us the corresponding wavelength seen by us for light emitted at that angle.

----

However, at this point, there’s a complication: although we know that the geometry of the original sphere is distorted, and we know that to be Lorentz-proportioned, //at least one// of its dimensions must have changed … but we don’t yet know the rules for how the reflector array’s internal dimensions ought to be scaled to produce those Lorentz proportions – we have the new outline, but not the scaling.

Deciding on the correct scaling isn’t trivial - for instance, if we’d carried out the “ellipse” exercise assuming Newtonian emission theory, the altered wavelengths inside the sphere would have generated an ellipse whose internal width and length would both increase with relative velocity – with Newtonian Optics, the ellipse’s internal width increases with the Lorentz factor, and the internal length increases with the square of the Lorentz factor, giving the same overall Lorentz proportions previously mentioned, so different theories produce different scalings. We seem to have a continuous range of possible sets of Doppler equations that would fit inside the ellipse and match the criteria that we’ve set up so far, with each possible equation’s predictions differing from those of its siblings by a Lorentzlike factor [1-vv/cc]^(?) .

As an aside, this behaviour suggests that Newtonian optics isn’t compatible with flat spacetime – in order to compact the NO ellipse’s internal dimensions back into its original volume, when we look at the shape’s elliptical cross-section and crush it by different amounts on its x and y/z axes, the ray distances end up protruding off the page to form a shape that looks something like a gravitational well with its throat tilted in the direction of motion. That’s presumably why attempts to model old-fashioned emission theory in flat spacetime didn’t work, NO’s wavelength-distances only seem to work in curved spacetime.

Anyway … our range of potential solutions that fit the ellipse can be broken down into two ranges – one in which increasing the relative velocity seems to increase the amount of positive curvature required to fit everything in, and another in which increasing v seems to increase the amount of //negative// curvature required in order to get everything to fit.

At the transition between these two ranges, there’s a solution that doesn’t have an apparent variation in curvature with velocity. In this solution, the width of the Lorentz-proportioned ellipse is constant, and only the length changes … since this elongation can be removed with a simple coordinate contraction of the x-axis to restore the original spherical outline, it’s not creating any intrinsic curvature.

If we draw up ellipses for that one solution, and measure off the corresponding wavelength-distances, they turn out to give the "relativistic Doppler" relationships of special relativity, so if we want those Doppler relationships, we can insist that basic geometrical rules (such as the law of reflection) are maintained for the sphere regardless of our velocity, and insist that the resulting lightbeam geometry shouldn’t involve curvature, and that eliminates all the other ellipse solutions.

----

Once we have flat spacetime, and the SR aberration formula, and the SR relativistic Doppler equations, I think we probably have the whole theory. It’s enough to give us the energetics of the sphere’s rays in all directions, for all velocities, and the energy than gives us the corresponding momentum calculations.

The last stage is to prove that the equations work in both directions – that the same rules for how we see the spheres also hold for how the spheres see us. To do that, we can set up an array of spheres, all exchanging signals, and fly though the array. If we fly directly along a signal path linking two spheres, then the condition of flat spacetime says that our motion shouldn’t affect the signal that passes through our position. If we carry a transponder with zero reaction time, or a zig-zag reflector with an infinitesimally-small distance between the mirrors, then the energy of signals passed through our location should be exactly the same as if we weren’t moving, or weren’t there. So given that the spheres regard themselves as stationary, if the Doppler shift that we see for “SphereA” directly ahead of us is “freq’/freq=D”, then for a second sphere behind us to see zero shift in the retransmitted signal (freq’/freq=1), the Doppler shift that “Sphere B” behind us sees on our signal must be 1/D.

This sort of cancellation is a feature of SR, so if our incoming signals agree with the SR relationships, and we want the outgoing effects to cancel out (to support flat spacetime), then our outgoing shifts are forced to agree with those of special relativity, too.

This combination of symmetry and cancellation doesn’t work with most systems, but it’s built into the SR relationships.

At this point, we haven’t //proved// that the speed of light is fixed in our own frame … we originally assumed that c was fixed in the sphere frame … but we’ve shown that the physical effects due to relative motion have to be symmetrical for both parties, so either party can use the same assumption about the speed of light being fixed in the other guy’s frame, and get the same results. The “real” frame for the speed of light in this exercise turns out to be unknowable and unphysical.

If we then want to introduce another inertial observer (or set of inertial observers) with an intermediate velocity, and demand that //their// motion as reflectors or transponders mustn’t affect any signals that pass through them (flat spacetime again), then we can work backwards from the need for complete cancellation to derive the SR velocity-addition theorem. At this point, the speed of light could be fixed to any inertial frame whatsoever without affecting the results, which disposes of any "aether drift" issues.

So it seems that (unless I’ve screwed up somewhere), we can probably start with the assumption that spacetime is flat for all valid inertial observers, and end up deriving the SR relationships, and then the mutuality of the equations, and then the associated result that the SoL could be assumed to be globally fixed for any valid inertial observer without affecting the outcome.

It’s 1am here – apologies in advance if I’ve screwed anything up, but the gist of it seems to be correct.

@Eric, thanks for the very interesting sphere thought experiment. I wish you had included a diagram, but I think I get it.

You speak of it being elongated. You mean in the direction of travel? I will assume this for my analysis. If you meant something else, well then the analysis is irrelevant. ; )

The elongated sphere, now an ellipsoid, has two focal points. That is kind of a problem. Because in the original sphere the light comes back to the emitter. I don't know however if you really meant that the way it came across, because what we actually have if I may assume the full Lorentz transform in order to analyze the experiment, is three things: time dilation, length contraction, and simultaneity skewing. I have cooked up the shorthand for the last item as "leading clocks lag" for the moving frame. I show this in the figure below.

So, while clock synchronous measurement by observer A shows the squashed sphere, the photons (shown as arrows) do not all make contact with the interior surface at that time because simultaneity is altered. Leading clocks lag, so the leading photons are late contacting the surface.

Now here is the interesting part of your thought experiment, to me. If one traces out the physical points of contact of photons with the sphere in A's frame, it will be slightly elongated. For very high velocity perhaps even elongated into an ellipsoid. The two focal points in that geometry are valid because the source has moved to (approximately, I'm not sure what I'm saying is exact) to the second focal point. Voila, it transforms dynamically into the geometry of your intuition. The aberration has to be included in this distortion also.

Did you get the white paper I originally posted with the question? I have taken it down now, because of several reasons. I don't want it available while reviewers are looking at a different paper that I just submitted on the twins paradox. I can send it if desired. It required the speed of light to be isotropic in only one reference frame, and with two additional assumptions derived the Lorentz transform. It's different than the constructive treatments of Bell and Lorentz because it is still quasi-principled, not fully constructive, but it is complete.

Hi Robert!

Yep, the ellipse, constructed around aberration effects and wavelength-change effects, is elongated. If the sphere is moving, then the events at which a lightpulse is created and reabsorbed at the centre of the sphere take place at different locations, because the sphere has moved while the light is in flight. I checked the relationships some years back, and and it all checks out.

William Moreau did a short study of the SR version of the shape within special relativity in 1994:

"Wave Front Relativity", Am.J.Phys [62] 426-429 (1994)

As far as I recall, Moreau used the "moving reflective sphere" idea and then applied the SR principle of the relativity of simultaneity. He argued that although we can imagine that the pulse hits all points on a “stationary” sphere’s surface at the same moment, if we’re moving wrt the sphere then our sense of SR simultaneity will be different and we’ll assign SR coordinates to those events that have nominal time-offsets that depend on an event’s location along the x axis.

So, in an SR description of the situation for an observer for whom the sphere is moving, the whole sphere supposedly doesn’t illuminate at once – the pulse is assumed to hit the rear of the sphere first (which is moving towards the point of emission), it hits the front of the sphere last (which is moving away from the point of emission), and the rest of the surface illuminates in sequence from the front to the back, as a moving illuminated ring.

Since the sphere is moving forwards while the ring advances over its surface, the spacial coordiates corresponding to the reflection/illumination events forms a stretched version of the sphere, giving an ellipse.

----

You should also be able to derive the thing using Minkowski spacetime.

Reduce the sphere to a representative cross-section taken through the x-y axes. If the sphere is nominally stationary, we expect all points on the circle to illuminate at the same moment, and the illumination events then represent the intersection of the pulse’s light-cone with a horizontal plane.

If we now apply a Lorentz transform, and tilt-and-skew all the coordinates accordingly, then the illumination events still have to lie on the surface of the lightcone, and still have to lie on a common plane, but that plane is now tilted wrt to the plane of simultaneity used by the new observer, so our circle becomes a conic section, which for speeds less than lightspeed is … our elongated ellipse.

----

The confusing aspect of the construction is that when we construct the ellipse around the SR aberration angles and wavelength distances, the distance between the foci isn’t the nominal spatial distance travelled by the sphere’s centre.

In the SR case, we can explain this discrepancy by pointing out that when the ellipse diagram is embedded in Minkowski spacetime, it's intersecting the light-cone at an angle, so the distance between the foci isn’t just a spatial distance, it also contains a time component – in order to turn the distance back into a purely spatial distance, we have to project it’s “shadow” onto a spatial axis, which shrinks the distance by a Lorentz factor, and brings us back to nominal distance moved, as measured by the selected reference-observer.

----

All the interpretation in the paragraphs above is specific to the SR version of the ellipse. However, the ellipse method itself works just fine as-is, and I used to find it useful as a way of getting straight to a theory’s physical predictions without worrying about any associated metaphysical infrastructure.

Article Wave front relativity

Oops. lost the graphic:

Dear All,

I read the following in this regards and found it excellent.

http://urania.udea.edu.co/sitios/astronomia-2.0/pages/descargas.rs/files/descargasdt5vi/Cursos/CursosRegulares/Relatividad/2010-1/Papers/lee_ajp_43_434_75.pdf

http://www.feynmanlectures.caltech.edu/I_15.html

@Eric, I'm skeptical about the distance between the foci not being the spatial distance. I believe SR requires it to be. The math of any detailed constructive analysis of SR is often so tedious that even brilliant people cannot do it all correctly. I formed this opinion after reading a book analyzing dozens of different clocks and concluding that a few of them didn't dilate properly with motion.

@Mohammad, your first link was hard to follow because it did not restate for clarity exactly what you take the POR to be and identify which part of it plays a key roll in the argument. In other words, I'm suspicious of circular reasoning without this information, and not inclined to want to wade through a bunch of math to try and find it. : )

Your second link to one of the Feynman lectures begins with the interesting claim that it is all about the mass increase. I once thought this. But it is not true. The length contraction is a separate issue with a separate cause.

Without any postulates -- it follows from Maxwell.

1. For restricted domain the homogeneous Maxwell equations have the plain wave solution, therewith the wave group velocity (and the same for the phase one) is constant, which can not contain any information from the wave source speed.

2. Metrics with time-coordinate relation based on this constant speed gives Lorentz transformation.

Я уже ответил на этот запрос. Ничего нового добавить к сказанному не могу.

@RobertS:

The distance between the foci obviously //does// represent the distance moved by the moving object … the problem is that different relativistic models tend to assign different apparent distance values to the same spatial separation for differently-moving observers, and SR is no exception. If we're trying to derive a theory without presupposing how these distances transform, we can't use the interfocal distance as our starting point.

Which I agree is incredibly annoying, but that's the geometry.

This is why the ellipse construction using wavelengths is useful for cross-theory comparisons – it works with agreed physically-uninterpreted values that can't be arbitrarily rescaled. If we decide to declare that the ellipse outline //must// transform back into the original sphere with no intrinsic curvature, then we //have// to make the ellipse minor diameter constant, that then defines the ellipse scaling so that we can't avoid obtaining the SR “relativistic Doppler” relationships for all the ray-distances, and because the focus-focus distance is now too large (by a Lorentz ratio), we can then go on to derive the SR result that a moving observer's coordinates have to be Lorentz-contracted to make them fit our measurement grid.

If you like, you can sanity-check this by constructing the ellipse differently, for instance by setting the interfocal distance to be the nominal distance moved by the object according to the onlooker. However, //that// exercise doesn't yield special relativity, but an unfamiliar (and slightly freakish) relativistic model that combines relativistic aberration with the “classical theory” Doppler shift relationships – its relationships would be “bluer” than the SR version, by a Lorentz factor. This would not be a desired result.

@RobertS: Yep, brilliant people often misconstruct SR – part of the problem is that many of them will have been mistaught what the theory's physical predictions really are, and in order to be able to arrive at what their uni lecturers //told// them was a correct answer, they've had to learn to fudge their derivations, and to consider the fudging as legitimate.

If anyone here's into the idea of constructing relativistic derivations using ruler-change arguments, please do bear in mind that the traditional 1950s description of how moving objects are seen under SR (with spheres compacted to contracted ellipsoids) was wrong. The 1959 Terrell and Penrose papers pointed out that there was an important distinction between the description given by SR coordinate systems, and what the theory predicted an SR observer would actually see.

I visited a few university physics departments in 1994 and found that they still seemed to be teaching the faulty pre-1959 version.

@Eric, yes I'm familiar with Terrell & Penrose. There are (at least) three classes of observation and all "see" different things. I have attempted to clarify this in regard to the Twins Paradox (or "problem" or "puzzle" as professor Low prefers) but am not posting the paper here until reviewed. In brief summary:

1. A point observer with a clock will measure lengths as contracted by traversing and timing them. Such an observer measures a time speed up or anti-dilation for reasons given below.

2. A signal observer is a point observer with radar, optical, Doppler or other signal transmission or reception capability, and sees time distorted by Doppler and shapes rotated as in Terrell & Penrose.

3. A multi-point or "Einstein" observer uses a grid of point observers each with a clock, and the clocks synchronized by the Einstein method, but otherwise uses signaling only to communicate observations not to make them. The Einstein observer can only "see" something by making what are to him simultaneous observations at multiple points and assembling a picture from those points. As any two of the grid clocks are passed by a single clock in a moving frame, they observe time dilation.

Since both frames agree on any one observation made at a coincident point where there is no argument about simultaneity, then a table of readings taken when the single clock passes either of the two grid clocks is identical in both frames. Therefore time observations are consistent between the single and the multi-clock observers, with the single clock observer noticing a speedup of time in the Einstein frame. If the single clock observer then decides to use also an Einstein frame, he would explain this by saying the other clocks were not in sync and so it was like crossing time zones. Some of the speedup is due to passing clocks with different time skew.

@Michael Chernogubovsky, Yes I understand about Maxwell's equations. Fresnel, FitzGerald and others could not figure out how this could be true. But there is a logical flaw in extending this to "everything." In the time of Lorentz, there was speculation about an "electromagnetic theory of everything." In which case Maxwell's equations can be made logically consistent by using Lorentz contraction and time dilation. As time went on at first it didn't seem like everything was going to turn out to be electromagnetic, and at first Special Relativity was adopted over Lorentzian Relativity, and then broader interpretations were given, particularly Minkowski space-time. Swann and others pointed out later that reference frames do not "automatically" adjust after acceleration, but it was too late to slow the pace of philosophical expansion. Then 3-force unification became credible, so maybe all or most things are electromagnetic.

However, that again raises the possibility of constructive relativity, derived from dynamics. But if even one thing is not electromagnetic (or some unifiable variant) the broad philosophical implications are missing. And it seems that gravity may not be unifiable. And ... if a photon is orbiting a black hole, you can move the black hole around (using the gravity of galaxies, for example) and the photon will go with it. Einstein actually tried to re-introduce nomenclature of the ether, but it was already out of favor. Anyway, gravity is certainly not a universal ether as it varies from place to place.