The unchanging speed of light in a vacuum is a foundational fact of relativity. This constant speed has been tested to unprecedented accuracy, but there are some that argue that isn’t enough. In special relativity, it is assumed that the speed of light doesn’t depend upon what direction the light is traveling in, or where it is in space. Physical processes might affect the speed of light, but mere location and direction doesn’t. This is actually part of a broader metaphysical idea that the universe is homogeneous and isotropic. Basically, it’s the assumption that the laws of physics (whatever they might be) are the same everywhere in the universe. This is in contrast to ideas such as geocentrism, which assumes that Earth holds a special place in the cosmos. It’s been an assumption as far back as Newton, though it has been tested in several ways, and has held up so far. But what if the assumption about light is wrong? What if the speed of light is actually anisotropic?
The initial verification of an invariant speed of light comes from the Michelson-Morley experiment in 1887, which showed that the speed of light didn’t depend upon the motion of the Earth. This implied there wasn’t an absolute reference frame, or a luminiferous aether through which light propagated. Over the years the speed of light has been measured with ever greater precision, and it’s always appeared to be a physical constant. But most of these experiments rely upon light to make a round trip in two directions, so technically it’s a measure of the “two-way” speed of light. What hasn’t been done is a direct “one way” speed of light measurement. You might think that’s easy enough to do, simply measure the start and finish time of a photon, for example. But to do that you’d need to synchronize your clocks, which means you’d have to set them at the same time when they are side by side, then move one clock to the finish line. Of course, when you do that, the motion of the clock would affect its measure of time, and you can’t be sure they they are still in sync without assuming some model like special relativity.
Suppose then that the speed of depended upon its direction of motion? Suppose it travelled almost instantly when heading toward us, but at half the “speed of light” when traveling away from us. The round trip time would be the same as relativity predicts with a constant speed of light. Most physicists don’t worry about this kind of thing since relativity keeps passing all the tests, but philosophers love to explore these kinds of metaphysical weaknesses. So the “one-way light problem” appears every now and then in the literature.
So what if the speed of light isn’t the same when moving toward or away from us? Are there any observable consequences? Not to the limits of observation so far. We know, for example, that any one-way speed of light is independent of the motion of the light source to 2 parts in a billion. We know it has no effect on the color of the light emitted to a few parts in 1020. Aspects such as polarization and interference are also indistinguishable from standard relativity. But that’s not surprising, because you don’t need to assume isotropy for relativity to work. In the 1970s, John Winnie and others showed that all the results of relativity could be modeled with anisotropic light so long as the two-way speed was a constant. The “extra” assumption that the speed of light is a uniform constant doesn’t change the physics, but it does make the mathematics much simpler. Since Einstein’s relativity is the simpler of two equivalent models, it’s the model we use. You could argue that it’s the right one citing Occam’s razor, or you could take Newton’s position that anything untestable isn’t worth arguing over.
Models such as anisotropic light are useful and interesting as a way of exploring the limits of what our scientific theories can tell us, but unfortunately they’re also used in a range of pseudoscientific models. In this case, the idea of a young Earth. One of the basic challenges for young Earth models is the starlight problem. If the universe is only a few thousand years old, how can we see light from the edge of our galaxy, much less other galaxies. One way to address this issue was to propose that the speed of light was much faster in the past, allowing distant starlight to reach us in a short time. But observations of line spectra from distant nebula shows that speed of light has changed no more than one part in a billion over the past 7 billion years. Then in 2010 Jason Lisle revived the idea of anisotropic light. If light moving toward us travelled at infinite speed, and away from us at half the traditional speed of light, then it would allow the most distant light in the young universe to reach us while still agreeing with relativity.
As crazy as that might sound, Lisle is right in claiming that such an effect would be indistinguishable from relativity, and this has made the work popular with young Earth supporters. However agreement with relativity isn’t enough. If light did actually reach us from distant galaxies instantly, we would expect galaxies at all distances (or more formally redshifts) to all look the same age. In fact, what we see is that more distant galaxies are younger than closer ones. If Lisle’s idea was correct, we wouldn’t see the magnification of distant galaxies due to cosmic expansion, nor fluctuations in a cosmic background, nor galaxy clustering in agreement with dark energy, nor a host of other observational results.
On its own, relativity doesn’t require isotropy and homogeneity, even though we generally assume it to be true. But when we combine relativity with the confluence of evidence we have in astronomy, we find that assumption is not only justified, but valid to the limits of observation so far.
Paper: Md. Farid Ahmed, et al. Results of a one-way experiment to test the isotropy of the speed of light. arXiv:1310.1171 [gr-qc]
Paper: John Winnie. Special relativity without one way velocity assumptions. Philosophy of Science, Vol. 37, No. 2 (1970)
Paper: Jason P. Lisle. Anisotropic Synchrony Convention—A Solution to the Distant Starlight Problem. Answers Research Journal 3 191–207 (2010)