The main purpose of the new MCQDPT2 code is to speed up calculations in the cases of medium to large basis sets and medium to large active spaces using resolvent fitting (aka table-driven) approach. It is not efficient for small active spaces (e.g., less than 500 CSF) and/or for small number of basis functions (e.g., less than 50). In the latter cases it is recommended to use the default PC GAMESS/Firefly' MCQDPT2 code. It is also not recommended to use new code for calculations which do not use ISA energy denominators shift, as in the cases of intruder states new code can become numerically unstable. On the other hand, we recommend to use ISA technique/nonzero edshft parameter for any MCQDPT2 run.

New MCQDPT2 code is triggered on by the presence of the $mcqfit group in the input file. In most cases, it is sufficient to just specify $mcqfit $end without altering any parameters of the $mcqfit group - the default parameters were selected to introduce errors in the computed energies which is typically less than 10-8 Hartree.

There are three adjustable parameters in the $mcqfit group:

      DELTAE - double precision, default is zero. If nonzero, this value is used to define parameters of the interpolation grid.

      NPOINT - integer, default is 400. Together with deltae this value is used to define parameters of the interpolation grid. Actual grid will be the smaller of npoint value and the number of points computed based on the deltae value. Increasing npoint will result in increase of required CPU time but will reduce errors in computed energy.

      IORDER - integer, default value is 7. Available values are 3, 5, and 7. Defines the order of polynomes used for interpolation.

For better efficiency of the new MCQDPT2 code running in SMP mode, we recommend to manually increase the default value of the length parameter of $mcqdpt group to be equal to the number of CSF generated by your run.

See also:

Last updated: March 18, 2009