CCL:G: CCSD(T) calculation in MOLPRO and G09

 Sent to CCL by: Herbert Fruchtl [herbert.fruchtl*o*st-andrews.ac.uk]
Without trying it out or looking into the manuals: if there is a significant difference in energy, check if one of the programs uses frozen core (I'm rather sure Gaussian does, don't know about Molpro) or if they correlate all orbitals.
Another source of differences may be the basis set. Molpro uses spherical harmonics (5 d orbitals) for all basis sets by default; Gaussian sticks to the number of d orbitals used in the development of the basis set (but it's always 7f). 
 HTH,
   Herbert
 Gerald Knizia knizia**theochem.uni-stuttgart.de wrote:
 
 Sent to CCL by: Gerald Knizia [knizia%%theochem.uni-stuttgart.de]
 nazanin jamshidi na.jamshidi|a|gmail.com wrote:
 
 Would you please tell me what is the difference between CCSD(T)
 calculations in MOLPRO and G09 packages?
Molpro's CCSD(T) is very efficient, while G09's is rumored to not be terribly fast. 
 
 Is there any difference in accuracy?
Only indirectly. For the same method/basis combination (say, CCSD(T)/AVTZ) you will get exactly the same number from both programs[*]. The indirect difference might arise from possibly using larger and better basis sets in Molpro than in G09. Furthermore Molpro features explicitly correlated coupled cluster methods (CCSD(T)-F12), which will give you basis set limit results already with augmented double- or triple-zeta basis sets, which would require quadruple- or quintuple -zeta basis sets in a non-F12 calculation.
[*] For open-shell calculations there are some subtle issues, like use of UHF vs RHF reference (Molpro uses RHF), and whether frozen core orbitals are frozen before or after semi-canonicalization. Such issues typically have negligible effect on relative energies. 
 --
 Herbert Fruchtl
 Senior Scientific Computing Officer
 School of Chemistry, School of Mathematics and Statistics
 University of St Andrews
 --
 The University of St Andrews is a charity registered in Scotland:
 No SC013532