CHELPG_HEAD defines as the “head space” 30 (radial extent) of the ChElPG grid. Is that the distance from the atom center or the distance from the vdw surface?
It’s from the vdW surface. (Think about the default value, 3 Angst… wouldn’t make much sense as a distance from the nucleus.)
I get it! Thank you a lot !
Is the current default implementation in qchem 6.1 still the same as your paper (Phys.Chem.Chem.Phys.,2012,14, 7679–7699)? Namely, \delta r = 0.1 bohr, R_off = 3.0 \AA, Bondi radius for vdW surface. We somehow cannot reproduce the results for some molecules. The diff is still significant (> 0.01 a.u.) when \delta x = 0.15 \AA.
I think that paper was still using Cartesian grids. We later switched to Lebedev grids,
Yes. Q-Chem still use Cartesian grids by default. Do you recommend Lebedev grids for all calculations?
eh, not sure that I recommend that for general purposes. We wanted to use CHELPG charges in XPol embedding and for Ewald summation in QM/MM (paper referenced above) and the charge derivatives dQ/P(mu,nu) are a bit expensive, so we tried to pare it down to the smallest number of grid points we could get away with, which is atom-centered and not Cartesian.
Note there can be a fair amount of variation in the numerical CHELPG charges w.r.t. grid, even for small changes like rotating the molecule; it’s one of the things that the community doesn’t generally like about those charges. The fit is to reproduce the ESP, not any kind of physical charges.
Yes. We did observe a significant difference between g09 and Q-Chem. That’s probably due to the different implementation. We are trying to reproduce the implementation in Q-Chem. We still observed a large amount of difference (>0.01 a.u.) between our implementation and Q-Chem 6.1. Based on our tests, the variation, due to the rotation, vdW radii, and \delta r, is ~0.001 a.u.
It would be very helpful if you could print the Cartesian grids info (coords and weights) of a water molecule with Q-Chem default options. I understand if that is too much for you.
O 0.0000000000 -0.0000000000 0.1174000000
H -0.7570000000 -0.0000000000 -0.4696000000
H 0.7570000000 0.0000000000 -0.4696000000
did you implement the switching functions that are described in the PCCP paper? It’s not as simple as “coords and weights”; Q-Chem’s implementation is designed for MD applications, to ensure that the charges are smooth functions of the nuclear coordinates…
As this user pointed out to me privately, the CHELPG head space defaults to 2.8 A whereas the manual says 3.0 A. I have made a correction to the manual for the next release (v. 6.1.1).