Characterize errors from RAS-reduction mechanisms in GRASP2018

Determine how restricted active space (RAS) reduction mechanisms used in GRASP2018—including the “rmixaccumulate,” “rcsfinteract,” and “cndens” programs, as well as other reduction procedures—affect total energies and what errors result from those reductions, through a full systematic study.

Background

The paper contrasts its RSMBPT-based configuration-selection procedure with RAS-reduction procedures used in regular GRASP2018 calculations. The authors state that GRASP2018 calculations use a restricted active space that is subsequently reduced using the “rmixaccumulate,” “rcsfinteract,” and “cndens” programs, while the RSMBPT method can predict the total-energy error introduced by reducing the RAS by comparing full and reduced spaces.

The unresolved issue is to establish, for the GRASP2018 reduction mechanisms and potentially other RAS-reduction approaches, how the reductions influence total energies and what quantitative errors should be expected. The paper does not provide the full study required to determine these effects.