If I set shape=0 0 1 , in the in.ellipse example, I get a whole bunch of “nan”. So that won’t work. Answered my own question.
Is this something fundamental about the Gay-Berne framework, or is it an artifact of how the integration scheme handles the moment of inertia tensor when it has a zero eigenvalue? If it is the latter, how tough would it be to modify fix_nve_asphere to allow for a zero-moment axis (a rod)? I’m guessing that the numerical scheme would have trouble with large but finite aspect ratios too.