sorry to bother you again. As the implementation of pGFN-FF is not in GULP manual yet, I should ask you some information about the cutoffs to ensure I am setting my MD simulations properly. I am simulating MOFs with adsorbate and solvent molecules.
When using classical FFs, we always pay attention about the cutoff for the nonbonded interactions (using at least twice its value in each dimension of the box).
However, GFN-FF is different in many ways, and in Spicher and Grimme’s original paper they don’t address the cutoffs subject. On the other hand, in your pGFN-FF paper, there are a few cutoff values, such as 50 u.a. for the dispersion term.
By analyzing a GULP output, we see: “Maximum range for interatomic potentials = 100000.000000 Angstroms”.
Can you provide any information/advice concerning the size of the system and the cutoffs?
If there is a minimum size for the system (related to some cutoff, as in the classic FFs), and if we can/need to change some cutoff.
The situation in pGFN-FF is complex as there are lots of cut-offs depending on the terms & many are set via accuracy factors (see gfnff_accuracy in help.txt) that determine the cut-off, rather than by giving a distance directly. The bottom line is that the defaults should all be fine such that you don’t have to worry about them.
A few other general comments:
The 100000 Angstrom cutoff in the output is the default for the overall value, which is just an arbitrary large value. This means the cutoffs will be specified by those of the individual potentials. You can change the overall value using “cutp”.
The idea of choosing cutoffs to be less than half the cell lengths is the minimum image convention which is often used in MD codes. In GULP there is no need to follow this convention since multiple images of the same pair of atoms are allowed. The idea of the minimum image convention is that it makes things faster (just look for the nearest pair rather than lots) and suppresses periodicity by not seeing multiple images. However, if you have charges present then in reality you always see multiple images anyway through the Ewald sum & so it doesn’t make much difference whether you use minimum image or not.
Hope that helps,