This gives the reference wavefunction. It must be an object of type CLSCF for closed-shell molecules and HSOSSCF for open-shell molecules. The is no default.
nfzc
The number of frozen core orbitals. The default is 0. If no atoms have an atomic number greater than 30, then the number of orbitals to be frozen can be automatically determined by specifying nfzc = auto.
nfzv
The number of frozen virtual orbitals. The default is 0.
memory
The amount of memory, in bytes, that each processor may use.
method
This gives a string that must take on one of the values below. The default is mp for closed-shell systems and zapt for open-shell systems.
mp
Use M\o{}ller-Plesset perturbation theory. This is only valid for closed-shell systems. Energies and gradients can be computed with this method.
opt1
Use the OPT1 variant of open-shell perturbation theory. Only energies can be computed for open-shell systems.
opt2
Use the OPT2 variant of open-shell perturbation theory. Only energies can be computed for open-shell systems.
zapt
Use the ZAPT variant of open-shell perturbation theory. Only energies can be computed for open-shell systems.
algorithm
This gives a string that must take on one of the values given below. The default is memgrp for closed-shell systems. For open-shell systems v1 is used for a small number of processors and v2 is used otherwise.
memgrp
Use the distributed shared memory algorithm (which uses a MemoryGrp object). This is only valid for MP2 energies and gradients.
v1
Use algorithm V1. Only energies can be computed. The maximum number of processors that can be utilized is the number of virtual orbitals. This algorithm computes few integrals than the others, but has higher communication requirements.
v2
Use algorithm V2. Only energies can be computed. The maximum number of processors that can be utilized is the number of shells.
v2lb
Use a modified V2 algorithm that may compute more two electron integrals, but may get better load balance on the
part of the calculation. Only energies can be computed. This is recommended only for computations involving large molecules (where the transformation is dominant) on very many processors (approaching the number of shells).
The v1 and v2 algorithms are discussed in Ida M. B. Nielsen and Edward T. Seidl, J. Comp. Chem. 16, 1301 (1995). The memgrp algorithm is discussed in Ida M. B. Nielsen, Chem. Phys. Lett. 255, 210 (1996).
memorygrp
A MemoryGrp object is used by the memgrp algorithm. If this is not given the program will try to find an appropriate default.
Member Function Documentation
void sc::MBPT2::compute
(
)
[protected, virtual]
Recompute at least the results that have compute true and are not already computed.