I was trying to fit Ab initio MD energy surfaces using GULP. However, the fitting parameters are way different form the expected values. In fact, in some cases values are opposite in sign as well. Is there any way to define a range of values for the fitting parameters?
You should take a look at some of the previous posts on this topic on the forum. The long and short of it is that you almost certainly haven’t included enough data from appropriate configurations in the fit. Many people make the mistake of only using near equilibrium configurations but this doesn’t provide any discrimination between the source of the binding and so you can only really fit a set of harmonic springs. You need to sample the asymptotes to dissociation if you want to distinguish say electrostatics from dispersion, so you almost certainly need to rethink your dataset.