Simulating point charges
- Date: 2017/06/25 04:59
- Name: Daniil
- Dear developers,
I am trying to calculate a cluster with the embedding potential, which is modeling the periodic crystal surroundings. This potential includes point charges, simulating the electrostatic crystal field. Is there any reliable way to define such entities in Openmx?
I understand from investigating the ADPACK VPS files that there is no direct Coulomb electron-nucleus interaction in the Openmx, as it is included, in a smoothed form, into the local potential. So, my current attempts consist in taking some arbitrary VPS file, extracting a local potential from it, scaling the potential to represent the needed charge, and constructing a new VPS file with local-only potential and a PAO file with zero valence density.
This approach gave rather physical results for a small (positively charged) cluster, surrounded by a small number of negative charges, but for increased number of charges I got "ERROR: Lapack routine DSTEQR failed".
Another problem is that, despite I provide zeros for both valence and total electron numbers in VPS and PAO files, Openmx requires compensating "nuclear" charge by initially arranged electrons, which forces me to populate "negative charges" with a negative number of electrons. After the calculation, Mulliken populations of such charges are ±(1e-4 - 1e-2), so that this problem doesn't seem to be large, but negative populations are still something dangerous.
Also, I haven't yet made a successful calculation with the positive charges, but I already expect the problem, that electrons cannot be stopped from populating their PAOs, while, in the embedding model, all electrons must be concentrated in the central cluster. One way of solving this problem is to use very high-energy PAOs for such centers, but the best way is to provide zero PAOs for them. Is it possible?
Or, as I mentioned in the beginning, maybe, there is a better way to describe point charges in Openmx?