Change in atom positions after convergence

pmetz

Apologies if this has been raised elsewhere in the forum.

I have a refinement setup with several rigid bodies. Everything works nicely during a refinement cycle (note converged starting point)...

... but goes haywire after the refinement finishes...

I can't tell if this is an issue of the rigid body definitions or of parameter limits or... ? Any insight would be appreciated, as the upset atom coordinates wind up being written to subsequent .cif exports, etc.

(I can't tell from the preview if these images will render. They can also be viewed here https://github.com/PetMetz/PlaceHolder)

Edit: TA-64 v6 | Windows 10

johnsoevans

My experience with rigid bodies is that they are (of course) really powerful, but it's easy to make mistakes!

I'm only guessing, but is it something like refining a rotation angle that takes a group off a special position thereby effectively changing cell contents? e.g. if a benzene ring has all atoms on a mirror plane, you can only rotate about an angle perpendicular to the ring.

The calculated density before/after refinement would give you an indication. Or the appearance of short bond lengths.

pmetz

Indeed the molecule centroids are sitting on special sites, so I've written them as split sites so that I can test relaxation (of the centroid) away from the ideal position. The density is staying around what I would expect.

In this case, the problem occurs when I'm refining the distance parameters for one of the rigid body definitions. Perhaps this is a clue that I need to change my rigid body definition? I'm hung up on this, however, because if I don't refine the distance parameters of the rigid body, the structure I'm expecting is calculated...

Thanks for your thoughts. I'll post here again if I can pin down the problem.

pmetz

Added test example to https://github.com/PetMetz/PlaceHolder/tree/master/test_rgd_bugs.

pmetz

If I eliminate the dummy sites I had used to specify the coordinate system of the molecule, I can stably refine positions and orientations.

This is somewhat aggravating, as I'm now stuck using non-orthogonal rotational axes...

Working example uploaded here https://github.com/PetMetz/PlaceHolder/tree/master/test_rgd_bugs

johnsoevans

I didn't quite follow the end of this thread. Is everything now working ok?