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

[Gerald Knizia]

binjiang.nju wrote:
> Dear Kirk,
>  
> Thanks again for your suggestion. I am not sure what is the intrinsic 
> reason for the incorrect anwsers in such case, but in my case it seems 
> to be good. I did some test for hoco system and the results obtained 
> from hf/uccsd(t)-f12 and hf/multi/hf/uccsd(t)-f12 are almost the same. 
> Here I mean the energies for H+CO2, trans-hoco and cis-hoco minimum 
> relative to OH+CO, the total energies for each geometry are of course 
> different in such two cases. Note that the results are also close to 
> experimental measurements.  So are these results meaningless or 
> coincidence? Please let me know if I was wrong. 

The problem with the uccsd(t)-f12 program was that it would effectively 
count the RHF orbital relaxation within the MO basis twice when used 
with non-HF orbitals (once as part of the normal CCSD(T) singles energy 
and once as part of EF12_RHFRELAX). You should be able to get a good 
approximation to the correct energy by using (ENERGY - EMP2_SINGLES 
instead of just ENERGY as a work-around (EF12_RHFRELAX would be the CABS 
relaxation energy and EMP2_SINGLES the RMP2 singles energy within the MO 
basis) . This has been fixed in the development version, but as far as I 
know not in the production version.
-- 
Gerald Knizia