[molpro-user] CCSd(T) with core

Kirk Peterson kipeters at wsu.edu
Mon Mar 1 16:57:31 GMT 2010


Dear Wafaa,

probably the lack of tight functions in your basis set via something like cc-pwCVTZ wouldn't affect your RHF convergence. That is probably something more fundamental and I'm surprised you didn't run into any problems when you optimized your geometry in the frozen core approximation.  In regards to the difference between cc-pCVTZ and cc-pwCVTZ, these were both developed to recover core correlation effects, but the cc-pCVTZ describes both the core-core and core-valence correlation on equal footing while the cc-pwCVTZ is biased towards the intershell core-valence correlation.  They do both converge to the same basis set limit but the latter will give slightly higher energies but generally more accurate core-valence correlation "effects" on bond lengths and relative energetics.

Yes, the explicitly correlated methods,  e.g., UCCSD(T)-F12, are capable of very high accuracy with relatively small basis sets.  I'm quite biased, but you should try the cc-pVnZ-F12 series of basis sets (n=D, T, Q).  For core correlation the cc-pCVnZ-F12 sets have just become available. The latter are in Molpro 2009.2 already and for 2009.1 you can download the sets (including new auxiliary RI fitting sets) from our website, http://tyr0.chem.wsu.edu/~kipeters/basis.html.

best regards,

Kirk

On Feb 28, 2010, at 12:34 PM, Wafaa Fawzy wrote:

> Dear Kirk,
> 
> Thank you so much for your quick response.  May be my problem I that I did not change the basis set, I used aug-cc-pVTZ.  I looked up the cc-pCVTZ or cc-pwCVTZ basis, I found that both of them include 7s,69, 4d, and 3f orbitals for Oxygen.  Do you know what is the difference between the two? Did you try the explicit correlation methods, e.g. UCCSD(T)-F12? What do you think of this method?
> 
> Thanks and Regards, Wafaa
> 
>  
> 
>  
> 
> From: Kirk Peterson [mailto:kipeters at wsu.edu] 
> Sent: Sunday, February 28, 2010 11:37 AM
> To: Wafaa Fawzy
> Cc: molpro-user at molpro.net
> Subject: Re: [molpro-user] CCSd(T) with core
> 
>  
> 
> Dear Wafaa,
> 
>  
> 
> in principle the core directive as you've shown it only affects the uccsd(t) calculation and not the rhf step, which is where your job is failing. So without any other info I would say your optimization has taken a step to a geometry where the rhf is struggling to converge.  Inspection of your log file (not the out file) may be helpful.  Perhaps you could attach your input file?  Note that when correlating the core electrons you should change your basis set to one that is designed for this purpose, e.g., cc-pCVTZ or cc-pwCVTZ.
> 
>  
> 
> regards,
> 
>  
> 
> Kirk
> 
>  
> 
> On Feb 27, 2010, at 10:10 AM, Wafaa Fawzy wrote:
> 
> 
> 
> 
> Hi All,
> 
> I ran optimization calculations that converged without the core electrons being correlated.  I repeated the same job with the option core: “uccsd(t);core”  but the optimization does not converge, I get an error saying : no convergence, the problem occurs in rhfpro”. 
> 
> In my calculations, it is important to get accurate correlation energy.   Does anyone has an explanation or suggestions for solving the problem?
> 
> Thanks in advance,
> 
> Wafaa Fawzy
> 
> Dept. of Chem.
> 
> MSU
> 
>  
> 
> _______________________________________________
> Molpro-user mailing list
> Molpro-user at molpro.net
> http://www.molpro.net/mailman/listinfo/molpro-user
> 
>  
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.molpro.net/pipermail/molpro-user/attachments/20100301/a9ebcbfa/attachment.html>


More information about the Molpro-user mailing list