As I'm seeing things, there are these choices for a definition of relativistic phase velocity u; a) u = (mc^2)/ p b) u = E / p Now I like choice a) because it leads to a correct-looking index of refraction via n = c / u. It leads to a correct-looking solution to the ray equation. However, references give choice b). Who says b)? Why not a)?