[molpro-user] CCSD(T)-F12 for complexes containing He atom

[Kirk Peterson]

Dear Yulia,

glad it now works ok.

best regards,

Kirk


On Jul 11, 2012, at 2:36 PM, kaluginy wrote:

> I'm very appreciated for your help! All works fine.
> 
> Best regards, Yulia
> 
> Le 11/07/2012 21:22, Kirk Peterson a écrit :
>> Dear Yulia,
>> 
>> I did some tests and it is certainly a fitting basis set issue.  I changed the df_basis and df_basis_exch both to av5z/mp2fit and the interaction energy at 100 bohr is now:
>> 
>>        RI     RR    TH      ECPA            ECPB
>>     100.0   2.28   0.0   -0.0002907513   -0.0002840335
>> 
>> You might also want to increase the global energy threshold to 1e-11 and add a "accu,14" to the hf directives.  In fact, making these changes only results in small changes:
>> 
>>        RI     RR    TH      ECPA            ECPB
>>     100.0   2.28   0.0   -0.0000308016   -0.0000296355
>>  
>> best,
>> 
>> -Kirk
>> 
>> On Jul 11, 2012, at 10:07 AM, kaluginy wrote:
>> 
>>> Thank you for your response. Here is the output with correct basis set. It gives positive energy for R=100 a0, but it should be negative.
>>> 
>>> best regards, Yulia
>>> 
>>> 
>>> Le 11/07/2012 18:29, Kirk Peterson a écrit :
>>>> Dear Kaluginy,
>>>> 
>>>> perhaps there is something else to blame, but your basis set input in the attached output file is very inconsistent.  In particular, the OptRI sets must only be used with the matching orbital basis sets - the former only include the complementary functions, so they are not standalone auxiliary basis sets.  You also specify the same set as the df_basis, but this is also not correct. These should generally be the MP2FIT basis sets.  Also the df_basis_exch sets should be the JKFIT sets of Weigend, probably extended with additional diffuse functions for your particular application.  Lastly, you haven't specified an orbital basis set, so the program defaults to cc-pVDZ, which of course is really terrible for any F12 calculation and in particular for your case.  Can you send an example output where the basis sets are correctly specified and you've set singles=0 ?
>>>> 
>>>> best regards,
>>>> 
>>>> Kirk
>>>> 
>>>> 
>>>> On Jul 11, 2012, at 4:46 AM, kaluginy wrote:
>>>> 
>>>>> Dear Dr. Werner and Dr. Peterson,
>>>>> 
>>>>> I'm a post-doc at the University of Le Havre (France) and Tomsk State University (Russia) and I'm studing complexes for astrophysical applications. I've faced a problem with the calculation of PES at the CCSD(T)-F12 level for complexes containing the He atom. For other complexes this method works perfectly (but only for VTZ and AVTZ basis set). For larger basis sets there is always a problem at long-range separations ( for example, CCSD(T)-F12/AVQZ gives Eint=0.16 cm-1 at R=30 a0 for CH4-N2 complex, but it should be 0.06 cm-1: both CCSD(T)-F12/AVTZ and CCSD(T)/AVXZ give correct answer) .
>>>>> There are some works where authors say that CCSD(T)-F12 (even for AVTZ basis set) fails for long-range separations: I've attached the papers for simplicity.
>>>>> Here are the complexes for which we are sure that (R)CCSD(T)-F12/AVTZ doesn't work well for large intermolecular separations: NH3-He, C4-He, O2-He, CN-He, HCl-He, CH4-He. For interactions with other molecules (not He atom), like H2,N2,CH4, this method works well.
>>>>> When I was in Stuttgart at the AMACQ, Dr. Werner proposed me to put singles=0. I've tried, hovewer, it didn't help.
>>>>> May be it is because of the basis set for the He atom? I've tried diffrent basis sets...
>>>>> I would be appriciated if you could try to look at this problem, as this method is a really powerful tool.
>>>>> 
>>>>> For example, I attached the output for O2-He complex for CCSD(T) and CCSD(T)-F12 calculations with the same basis set.
>>>>> 
>>>>> Best regards,
>>>>> 
>>>>> <O2HE_ccsdt.out><O2HE_ccsdt-F12.out><NH3-He.pdf><c4-He_.pdf>
>>> 
>>> <O2HE_F12_singles=0.out>
> 
>