Different results were obtained with different iscf

[David Danovich]

Hi Fuming and may be other members of the list,

I found some strange behavior using different optimization procedures. This is famous C2 molecules. There are several types of calculations (c2-3group-my*-ONLY.xmo with only two covalent structures in group3. c2-3group-my*.xmo with all 14 structures which belong to group3. c2-3groups-my*.xmo with 64 structures from group 1,2,3. Star(*) may be 2 or 5 which is iscf method). For two covalent structures I found that iscf=2 and iscf=5 produced different energies (-75.26560102 for 2 and -75.20542374 for 5). The same is true also for calculations which invovled all 14 structures of group3 (-75.3504108 for 2 and -75.2986608 for 5). The difference is more than 30 kcal/mol in favor of iscf=2. When I calculated all three group together the order is reverse. Here I also got different energies for iscf=2 and iscf=5 (-75.5755105 and -75.5945355), but now in favor of iscf=5. I checked this with gamess version of xmvb and it is the same. Does it means that iscf=2 is not suitable for this system? How should one know that it is not suitable because for only two covalent structures WF is quite similar. XMO files as well as INT file for preint is attached. 

Thank you   David. 

--

____________________________________________________________________________
Dr. David Danovich,  The Lise Meitner Minerva Center for Computational Quantum Chemistry,

Institute of Chemistry, The Hebrew University, Edmond J. Safra Campus - Givat Ram,

91904 Jerusalem, Israel http://yfaat.ch.huji.ac.il/david.html, david.d...@huji.ac.il

FAX:(+972)-2-6584033, Phone:(+972)-2-6586934(w),  Mobile:(+972)-544-768669

[应富鸣]

Dear David,

If you check the SCF iterations with 2 different ISCF options, you may find that the energy in the first iteration are already different, even with the same initial guess.

But the overlap matrix of structures show that the structures are not orthogonal.

So I checked the overlap between determinants. See the attachments.

The calculations show that a lot of determinants are orthogonal to each other. I think this is the reason.

As we know, the ISCF=2 cannot handle the calculations in which structures or determinants are orthogonal, but ISCF=5 is able to do.

So you should NEVER trust the results provided by ISCF=2 in this case.

You may check the results by ISCF=1/3.

Best regards,

在 2015年9月26日星期六 UTC+8下午7:18:25，David Danovich写道：