[molpro-user] 回复: Re: UCCSD-F12 with CASSCF orbitals

[binjiang.nju]

Dear Gerald and Kirk,

Many thanks for your reply. I finally use the procedure hf/multi/hf(maxit=1)/uccsd-f12(orbital,ignore_error), which works fine now. I'm also grateful for Gerald's alert about the ccsd energies with non-rhf orbitals. I would like to do more test to check if the ccsd energies with this strategy are correct.

Best, Bin.
2011-08-10



binjiang.nju



发件人: Gerald Knizia <knizia at theochem.uni-stuttgart.de>
发送时间: 2011-08-11 05:38
主 题: Re: [molpro-user] UCCSD-F12 with CASSCF orbitals
收件人: "binjiang.nju" <binjiang.nju at gmail.com>
抄 送: molpro-user <molpro-user at molpro.net>





binjiang.nju wrote: 
> I am trying to do a uccsd(t) calculation with CASSCF orbitals in  
> Molpro.2010.1, it will be useful when two quasi-degenerate states  
> exist and sometimes improve the convergence, comparing to the  
> uccsd(t)/hf strategy.  
>   
> It works fine when I use only normal uccsd(t) option, but it fails  
> when I use uccsd(t)-f12 option. I have tried to change the basis set,  
> active space, as well as the type of CASSCF orbitals (Natural,  
> Canonical), all of these do not help. Then I find some information for  
> such problem, but still no suitalbe answer. 
>   
> Gerald Knizia has mentioned that the F12 methods do request RHF  
> reference functions, but he would try to fix this. (I cite his words  
> in the following, hopefully he would not mind) 

I have semi-fixed this in the current development version (i.e., it is  
there possible to use such orbitals now, but it requires some persuation  
with rhf,maxit=1). However, this turned out to be more complicated than  
anticipated, which is why I did not backport the changes to the  
production version. In the current production version both CCSD-F12  
programs will give (different) wrong energies when used with non-HF  
orbitals. I would prefer to leave this issue open until the next Molpro  
production version (i.e., the one from the current development version,  
where this will work automatically), unless there is very strong  
interest in having this feature available. 
--  
Gerald Knizia 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.molpro.net/pipermail/molpro-user/attachments/20110810/3759502f/attachment.html>