Dear all,
I wonder about the meaning of the keyword “ESP_CHARGES = 1” and “ESP_CHARGES =2” for computing charges of atoms in molecule’s excited states. On employing these keywords for the pyrazine molecule, the former keyword yields symmetric charges (for chemically equivalent atoms) while the latter gives asymmetric charges for atoms in the pyrazine molecule for excited states.
I could not find the description for this keyword in the Q-Chem 5.3 manual. Help regarding the references behind these keywords or a brief description would be highly appreciated.
Hmmm, indeed this seems not to be documented in the manual. It is not my code but I looked into and this is copied from the documentation block in the code:
It therefore makes some sense to me that ESP_CHARGES=1 might not (quite) satisfy the symmetry of the molecule, because the grid that is used to compute the charges may not (quite) satisfy that symmetry. In principle this could be removed by increasing the number of grid points per atom but it’s not immediately obvious to me (from inspecting the code) how to accomplish that. Let me contact the original developer and see if he can help.
I agree with John that symmetry in atomic charge values is not guaranteed with either grid. But we can add “RESP_CHARGES TRUE” to ensure same charge values for chemically equivalent atoms.
@jherbert thanks for the description. Does this description of grids also hold for other charge methods such as RESP_CHARGES and CM5 charges? @Yihan thanks for the suggestion but not sure if the resp_charges/cm5 charges are available for the excited states as it does not specify the state for which charge is computed? Is it safe to assume that I get resp_charge/cm5 of a specific ith electronic state by setting cis_state_deriv=i?
Additionally, the Hessian and gradient calculations relaxes the CIS density automatically, in this juncture if I use cis_relaxed_density=false would it use the unrelated density for computing charges?
Finally, if I read (in the same input with @@@) the density from the optimized geometry should I still relax the density prior to computing charges?
“ESP_CHARGES 1” and “RESP_CHARGES 1” employ Lebedev grid points for ESP charge fitting. “ESP_CHARGES 2” and “RESP_CHARGES 2” use spherical Harmonics grid points. RESP charge fitting assigns same charge values to chemically equivalent atoms.
When “ESP_CHARGES” and “RESP_CHARGES” are specified in CIS/RPA/TDA/TDDFT singlet calculations, ESP/RESP charges for ALL excited states would be computed. Not sure about the triplet states. Charges are computed using the unrelaxed excited state density, unless “CIS_RELAXED_DENSITY” is set to true.
Here are more details about the specification of Lebedev/spherical-harmonics grid points for ESP charge calculations. Four layers of vdW surfaces are constructed at 1.4, 1.6, 1.8, and 2.0 atomic vdW radii, respectively. For each layer, a Lebedev/spherical-harmonics grid is chosen so that the spacing between grid points is around 0.5 Angstrom. One can use the ESP_SURFACE_DENSITY keyword (whose default value is 500) to set the grid spacing to a non-standard ESP_SURFACE_DENSITY * 0.001 Angstrom. Inner points (whose distance from another atom is shorter than the corresponding scaled vdW radius) are removed.
At least for pyrazine, CM5 charges (and the underlying Hirshfeld charges) for chemically equivalent atoms look the same. This is probably because: (a) the atomic densities (for constructing promolecule density) were spherically averaged; and (b) the numerical integration scheme (use SG-1 grid by default) for obtaining Hirshfeld charges does not noticeably break the chemical symmetry.