Firefly and PC GAMESS-related discussion club

 
Learn how to ask questions correctly  
 
 

 

Alex Granovsky
gran@classic.chem.msu.su

I was able to converge this job using DLCs. The input and output files are attached.
It seems DLC/GDIIS code is very sensitive to the numerical noise
in computed gradients so I used certain nonstandard optimizer settings
basically disabling redefinition of DLCs and Hessians during optimization

Kind regards,
Alex Granovsky

On Fri Feb 27 '15 7:23pm, Alexander Vasyanin wrote
--------------------------------------------------
>It really conveges! Fantastic!

>I'll try to test how different sets of options, but I see that SCF easily converges on the first geometry step with NCONV=8. Looks like I missed something in my previous attempts to set it.

>Thank you very much, Thom!
>
>
>
>On Fri Feb 27 '15 1:15pm, Thomas Pijper wrote
>---------------------------------------------
>>Dear Alexander,

>>I got your geometry optimization to converge. Input and output are attached. Disabling DLCs seemed to do the trick, though had to increase accuracy a bit further than is normally needed for MP2 gradients with partial linear dependence in the basis. It may be that use of DLCs with IFDMOD=0 (or the finetuning of other DLC-related options) may lead to converge as well. Unfortunately, I currently lack the computational resources to look into this in detail.
>>
>>
>>Kind regards,
>>Thom
>>

This message contains the 1102 kb attachment [ mp2_calcs_dlcs.rar ]