PC GAMESS-related discussion club

Alex Granovsky
gran@classic.chem.msu.su

actually, PC GAMESS uses only two criteria during geometry optimization - namely, maximum gradient and RMS gradient. The reason of this is that maximum component of displacement and RMS displacement are primary the property of geometry optimization routine used, not the PES itself (while gradients are the property of PES only).

It is always hard to compare two different programs, especially PC GAMESS with Gaussian as we are banned by Gaussian and do not have access to it. However, according to multiple reports of our users, PC GAMESS by default indeed has more strict criteria for geometry optimization. Basically, this is because in most cases the allowed RMS force is the strongest component of two criteria used by PC GAMESS, and they differ considerably between PC GAMESS and Gaussian.

The reason why PC GAMESS uses stronger criteria than Gaussian is that in the case of flat PES, you may find significant difference between geometries optimized at OPTTOL=1d-4 and, say, OPTTOL=5d-4. In general, the default set of parameters used by PC GAMESS seems to be reasonable for almost all cases. However, for very flexible molecules you may need to decreaseOPTTOL to, e.g., 3d-5.

My personal opinion is that many default criteria in Gaussian are not very reasonable and were chosen based on compromise between speed and accuracy with emphasis on calculation's speed.

Best regards,
Alex Granovsky


On Mon Feb 18 '08, Tae-Yun Park wrote
-------------------------------------
>Dear PCGAMESS users,

>While I¡¯ve been enjoying the performance of PCGAMESS recently, I found another factor that the performance of PCGAMESS could possibly be compared unfairly if one just uses default settings for geometry optimization. The default settings in PCGAMESS are,
>OPTTOL = 0.00010
>RMS Force = OPTTOL/3 = 0.00003333¡¦
>RMAX = 0.1
>RMIN = 0.0015

>In the case of G03, the default settings are,
>Max. Force = 0.00045
>RMS Force = 0.00030
>Max. Displacement = 0.00180
>RMS Displacement = 0.00120

>If OPTTOL and Max. Force have got the same meaning, it looks to me that PCGAMESS settings are more severe, which may result in more iterations and calculation time but more accurate geometry. Same for the case of RMS Force setting. I wonder if this is a correct view.

>If indeed the default settings are not really equivalent, I wonder if the PCGAMESS¡¯ OPTTOL can be set to 0.00045 to make the similar conditions compared to the case of G03. This makes PCGAMESS¡¯ RMS Force = OPTTOL/3 = 0.00015, which is still smaller than G03. If it is so, what about the RMAX and RMIN? What values could to be used to have the similar conditions to G03 default settings?

>Finally, if indeed the PCGAMESS default settings are more sever compared to those of G03, is there any specific reason for that?

>Any comments would be highly appreciated.

>Many thanks,

>Tae-Yun
>