I was trying to fit a dihedral parameter using the torsional functional form (with “torsion” keyword) as implemented in gulp. It is not giving a good fit and the sum of squares value is very high, 0.399. Currently, the code is fitting only single torsional constant and the fitting for Phi0 is not available. Is there any way to fit multiple K (torsional constant) and Phi0?
There are many reasons why a fit might give a high sum of squares, but I’ll assume that this is a simple fit of fixed geometries vs energy which should work. The reason why phi0 is not fitted is that you usually know what the periodicity (i.e. n) and equilibrium angle (related to phi0) should be for a torsion based on the type of bond you are dealing with (sp3 vs sp2 hybrisation etc). Therefore there is usually no reason to fit them since they are fixed by the physical problem. If you have a complex torsion (say an sp2 atom bonded to an sp3 atom) then you may need more than one torsion to capture the competing different periodicities of the end atoms. If you do need more torsions with different phi0/n values then you can just add more to the input. GULP has no limit on how many potentials you can overlay for the same interaction (until you hit the memory limit of your computer!).