Most Recent Version of GAMESS

by Mike Schmidt

Primary Modification:

The two changes prompting this "new date in the box" at the top of the output are the RHF+CI gradients, which we've had done for about 2 weeks, and the multireference perturbation theory code of Haruyuki, which I've been testing the past week.

There are 7 pages of new input documentation on the MCQDPT (multiconfigurational, quasidegenerate perturbation theory). This compares the method to CASPT2, etc. It is important to note that the method can be used in a multi-state fashion. Besides adding the 2nd order calculation to the standard EXAM09 input file, I have five other files illustrating the input, and comparing the results to the literature:

• ~mike/tests/mp2/lif-mcqdpt.inp (a 2 state run)
• ~mike/tests/mp2/cn-mcqdpt.inp
• ~mike/tests/mp2/ch2-1a1-mcqdpt.inp
• ~mike/tests/mp2/ch2-3b1-mcqdpt.inp
• ~mike/tests/mp2/nh2-mcqdpt.inp

It is not quite true what I say in the documenation about not knowing CPU timings. The NH2 example has 69 AOs and 49 CSFs. It takes 215 seconds to optimize the FULLNR wavefunction (full valence type), and an additional 158 seconds to do the 2nd order energy correction. I do not know how much disk space it uses since Haruyuki closes his files in DELETE mode as the program runs. There are a total of 14 different files used by this method, over and above the standard GAMESS files. I would be interested in hearing from anyone who tries 100 to 200 AOs about their timing results.

there are some smaller changes and bug fixes, too:

• Simon Webb has fixed MCSCF runs for both SCRF and fragment computations, to polarize the ab initio system during the MCSCF iterations.
• The $BASIS group now has a NFFUNC keyword to add Pople's choice of f exponent to atoms Li-Cl.
• The DM1 generation in GUGDM2.SRC is fixed, yet again, and this time I think it is right!
• MCSCF runs will not call the FULLNR's damping routine during SOSCF or FOCAS runs. The damping is handled in another way by SOSCF, and skipping this call has allowed two examples to converge correctly.

Files Changed:

EFINP, EFINTB, GAMESS, GUESS, GUGDM2, INPUTA, INPUTB, IOLIB, MCSCF, MCQDPT (the new one), MP2GRD, SCRF.

Obtaining GAMESS:

Persons not in Ames who would like to take a copy of the source code will find a new tar file in /u2/guest/gamess.tar.Z.

and....

A very big thank you goes to Haruyuki Nakano for coming to Ames this summer to interface his very useful method to GAMESS!