CCSD(T) - frozen core dependency

[YuraM]

Good morning to everyone,

for reactions A + B -> C + D I have detected strong dependency of the reaction enthalpy on the number of electrons included in the frozen core in CCSD(T) calculations. Sometimes the results with more electrons included in correlation treatment are closer to experiment, sometimes the opposite trend is observed.

Just to illustrate, with the default (noble gas configuration is frozen) settings I get +20 kcal/mol, when I include more electrons (let say not the previous but one before previous noble gas configuration is frozen) then I can easily get +5 kcal/mol.

The basis sets are cc-pvtz, cc-pvqz. Any personal experience? Comments? Suggestions? Also, one of the latest articles of Truhlar is interesting wrt "accuracy" of CCSD(T).

http://pubs.acs.org/doi/abs/10.1021/acs.jctc.5b00081

Thank you in advance!

[VTur]

А без замороженных остовов не обойтись? Там есть проблема сшивки остова и валентной области и проблема ортогонализации этих орбиталей этих частемй. Иногда используются дополнительные функции для решения ортогонолизации, а они плавающие. Для полной электронной энергии это плохо. Лучше бы полноэлектронный базис

[ Post made via Android ]

[YuraM]

Thank you,
do you mean something like ANO-RCC basis sets?

[YuraM]

ANO-RCC did not really help...