JC: did you need to reduce timestep at all from 2 fs for the steeper repulsion in the condensed phase benchmarks? AH: not really. Molecules spend most of their time in the bottom of the potential well. I used 2 fs. r^14 is incredibly steep, there winds up being a singularity that you don’t get with the exponential. Possibly if you did GCMC where you could overlap 2 atoms, you would need to reject superposition immediately or you might get fusion. I had to make a change to Evaluator to fix error handling, but otherwise these are all the base settings.
LW: you mentioned that you implemented polarisation in smee, have you done fits with it? AH: the smee work was earlier this year. Descent gave me lots of warnings in the documentation that it’s experimental, so I never ran any fits with it. I did a proof of concept directly using smee, not with descent. Right now I have a big software dependency loop. I can’t use smee with the AmoebaMultipoleForce, so I have to write my own force.
JC: back to RMSDs with density – difference between green and blue bars?
|