Try to imagine a rotation around a timelike axis of an object, that moves together with the observer. The rotation can't be seen, because it is stationary (with the observer). Acceleration can be described as curvature of world lines and we twist this rotation by that. The more we twist it, it would wiggle and seen as radiation. That has two aspects: stability and rotation. The timelike behavior of such a pattern I call the 'mass term', because that is how a mass behaves. The rotation is the 'radiation term', because it is spacelike. If we follow such a pattern, it could be twisted and mass turns to radiation (or back) depending on the movement relative to the observer. The maximum would be, that this rotation is totally flipped to the side and we see only radiation and no more mass.
I have chosen this model, because it followed from the use of quaternions to describe spacetime intervals. This I found useful, because there is a connection of the formula of quaternion rotation to the position operator of quantum mechanics. It is actually the same idea, even though different symbols are used. I have written a kind of
book about this, what is in google.docs presentation form till now. It contains no mathematical model, because I just have started to develop that, but I have found a few websites, where such models are shown. One is that of Daug Sweetser, who is member of this forum.
I just have started to read 'Spinors and Spacetime' by Penrose and Rindler. This method is, what I have in mind.
This is a mexican website called
geom, that covers the topic quite well.