The concept of phase advance is best explained by the above picture. The one-turn map at some arbitrary position is normalized by the map :
There are few points to make from which we deduce the phase advance:
Item 1 is simply the assumption of the normal form. The reader may wonder how is this assumption justified? Is there any basis for it? Actually, thanks to the process of perturbation theory, a package like FPP is equipped with a normal form algorithm on the Taylor map. Therefore the formal existence of the normal form, i.e., its approximate existence closed to the origin is not in doubt.
Item 2 is a little more tricky. If the transformation is symplectic, then one can show that the circles at 1 and 2 are identical. If one allows a non-canonical transformation, it would not be the case. We disallow this for symplectic maps. In any event, the reader can see easily that one could scale the circles at position 2 to insure an identical size if non-canonical transformations were used.
Then it follows trivially that
The phase advance depends critically on the definition chosen for . If is a functional of the one-turn map, then the phase advance is obviously unique between matched position, i.e., between point with identical one-turn map.