Firefly and PC GAMESS-related discussion club

 
Learn how to ask questions correctly  
 
 

 

Alex Granovsky
gran@classic.chem.msu.su

for a fixed atomic basis set, CPU requirements of large SCF jobs
scales approximately as the square of the overall number of atoms.

For the fixed number and types atoms, CPU requirements of large
SCF jobs scales approximately as the fourth power of the overall
number of basis functions.

The latter means that SBKJC should be faster but not dramatically
faster as it removes only a small fraction of AOs.  

Kind regards,
Alex Granovsky


On Sat Apr 19 '14 6:32pm, Siddheshwar Chopra wrote
--------------------------------------------------
>Dear Alex,
>Thats an important information. Thanks again. Alex could you comment on their speeds? Logically if they reduce the no. of basis functions, then they should be really fast. I want to be sure about their speeds before using them. It would be good if you could compare their speeds with the 6-31G and its variants.

>Regards,

>On Fri Apr 18 '14 9:10pm, Alex Granovsky wrote
>----------------------------------------------
>>Dear Siddheshwar,

>>for second row elements ECPs are computationally inefficient as
>>they remove only single orbital (i.e. 1s) per atom. If you use
>>SBK, you still need to add polarization function(s) to get
>>reasonable results. SBK basis for Li, Be, B, C, N, O, and F
>>atoms has only two L-type (i.e. combined S and P) shells for
>>valence electrons and thus it is (approximately) a DZV-quality
>>basis set.

>>Kind regards,
>>Alex Granovsky
>>
>>
>>On Tue Apr 15 '14 1:24pm, Siddheshwar Chopra wrote
>>--------------------------------------------------
>>>Dear All,
>>>This is the first time I am using SBKJC ECPs for the same samples for which I used 6-31G basis sets. Could anyone point out their speed and accuracy comparisons (Firefly based)? As per Jensen's book I read that for the second row elements, SBKJC gives almost same error as that of TZP. And I have never used TZPs. I have till now worked with only 6-31G and variants.

>>>Regards,