I'm working on PolyDiMethylSiloxanes and am often having difficulty with geometry optimization.
I build 5 or so ensembles of around 5000 atoms using Amorphous Cell, then perform a Forcite Geometry Optimization on all ensembles at Medium quality to find the lowest energy one. Some converge - but some don't:
Frame Converged Total Non-bond H-bond van der Waals Electrostatic 3-Body Diagonal Bond Angle Torsion Inversion Cross terms
1 Yes -42157.816 -41508.598 0.000 -3035.796 -38472.802 0.000 -782.951 36.017 926.978 -1745.947 0.000 133.733
2 Yes *-42164.971* -41513.701 0.000 -3039.935 -38473.766 0.000 -783.755 36.307 935.235 -1755.297 0.000 132.485
WARNING No progress. Optimization with current method stopped.
3 No -42100.228 -41461.093 0.000 -2996.761 -38464.332 0.000 -770.522 38.476 930.447 -1739.445 0.000 131.386
WARNING No progress. Optimization with current method stopped.
4 No -42135.182 -41485.222 0.000 -3017.112 -38468.110 0.000 -782.590 37.061 935.223 -1754.875 0.000 132.630
5 No -42102.349 -41457.202 0.000 -2991.882 -38465.320 0.000 -776.907 36.954 930.748 -1744.609 0.000 131.760
I pick the lowest energy ensemble that has converged, perform an NPT dynamics run (at Coarse quality) to optimize the density, then again run a Forcite Geometry Optimization on the last 10 frames
at Medium quality, and again pick the lowest energy ensemble that has converged. This often has similar behaviour to above.
I then pick the lowest energy ensemble, and re-run the Geometry Optimization it at Fine quality. If I'm lucky I get a nice gradual decline in energy and it converges. If not, I get:
or worse. Any ideas ???
