| 5:27:04 AM PDT - Fri, Jul 27th 2018 |
|
Hi,
I was trying to define specific integration grids using atom tags, but the results I got with both NWChem 6.6 (27746) and 6.8 (v6.8-47-gdf6c956) are not in line with the documentation.
In practice, with atom tags that are more than just the elemental symbol, e.g., "H1" instead of "H", some of the grid's setting that are specified for an atom tag are used also for other atoms.
In addition, NWChem's interpretation of the atom tags is not fully consistent with the matching mechanism described here for the assignation of basis sets. That is, "H1" is not matched by "H".
I've created a trivial example, just to trigger the problem. Try this input file:
START
BASIS
H library lanl2dz_ecp
C library lanl2dz_ecp
O library lanl2dz_ecp
END
GEOMETRY units angstrom
C1 1.08798 -0.05600 0.00496
C2 2.60257 -0.06069 0.01060
O3 3.08280 -1.37216 0.26822
H4 0.70196 0.94473 -0.20816
H5 0.69627 -0.38596 0.97276
H6 0.70338 -0.75169 -0.74806
H7 2.99434 0.61384 0.77794
H8 2.99000 0.25976 -0.96087
H9 2.74153 -1.64382 1.13733
SYMMETRY c1
END
CHARGE 0
dft mult 1
XC pbe0
DISP vdw 3
DIRECT
GRID nodisk lebedev 80 11 H 70 8 O3 90 13
END
task dft energy
My expectation would be something like this:
Tag B.-S. Rad. Rad. Pts. Rad. Cut. Ang. Pts.
--- ---------- --------- --------- ---------
C 0.70 80 7.0 302
O 0.60 90 8.0 434
H 0.35 70 8.0 194
Instead, NWChem's interpretation of the grid directive gives this:
Tag B.-S. Rad. Rad. Pts. Rad. Cut. Ang. Pts.
--- ---------- --------- --------- ---------
C1 0.70 80 7.0 302
C2 0.70 80 7.0 302
O3 0.60 90 8.0 434
H4 0.35 80 8.0 434
H5 0.35 80 8.0 434
H6 0.35 80 8.0 434
H7 0.35 80 8.0 434
H8 0.35 80 8.0 434
H9 0.35 80 8.0 434
The number of angular point specified for O3 is used also for all the hydrogen atoms.
Moreover, the number of radial point for each H atom seems due to the string "lebedev 80 11", which means that "H 70 8" did not match the atom tags H4, H5, ...., H9.
Instead, if I use C,O,and H in the geometry and as well in the definition of the grid, I do get the expected result with respect to the definition of the grid.
In conclusion, the customized definition of integration grids seems incompatible with the use of atom tags that include more than just the elemental symbol.
Cheers,
Marco
|