...
Validation set 3
Here we performed the same validation as above but this time starting from a trajectory extracted from an MD simulation with the k values for all rotatable torsions set to 0. This should allow the simulation to cross high energies barriers easily without any basis to the starting conformation.
Here we again see a large energy range for all of the molecules similar to set 1
We also see a general increase in accuracy with each version of the openff forcefields
All of the forcefields disagree with QM on the lowest energy structure in the validation sets
Using conformer specific charges set 2
...