Errors (in ) of the correction evaluated with the aug-cc-pVDZ basis set as virtual orbitals (HF MO or MP2 NO) are removed from the computation. The total corrections for these test cases are −0.281 (ammonia dimer), −0.344 (water dimer), and (methane dimer).
Percent errors for the correction (scaled and unscaled) averaged over the 11 smallest complexes from the S22 test set (Ref. 49) as virtual MP2 NOs are removed from the computation to achieve a certain percentage of deleted virtual orbitals (top panel) or to obtain a specified idealized speedup (bottom panel).
Percent errors for the correction (scaled and unscaled) as virtual orbitals with less than a specified number of electrons are removed from the computation for the 11 smallest complexes from the S22 test set (Ref. 49).
Timings of the evaluation of correction with the aug-cc-pVDZ basis set as various approximations are applied for (from left to right) formic acid dimer, methane-benzene, T-shaped benzene dimer, T-shaped indole-benzene, and hydrogen bonded adenine-thymine.
Average speedup for the correction of water dimer, ammonia dimer, and methane dimer computed with various basis sets when virtual MP2 NOs are removed using a threshold of electrons. An additional factor of approximately two times can be gained by the frozen core approximation.
The effect of the MP2 NO and frozen core approximations on the correction evaluated with the aug-cc-pVDZ basis set for the complexes of the S22 test set (Ref. 49). (MP2 NOs with occupancies less than electrons are removed. Errors are given in .)
Article metrics loading...
Full text loading...