Hello,
I run open-shell DLPNO-CCSD(T) calculations using ORCA package for some molecule-radical systems. Some systems show the T1 diagnostic value of ~0.025 which as far as I am aware, means that the systems show multiconfigurational character.
I have run stability check on the UHF reference wave function by including the
%scf stabperform true end
line in the input file
The results show that the wave function is stable. Indeed, the CCSD iteration converges slowly, taking about 75 iterations using default ORCA setting.
I use the quasi-restricted orbital for the DLPNO-CCSD(T) calculations
Is there any way to get a reliable reference wave function so that the T1 diagnostic is below 0.025?
You are right about the multiconfigurational character upon an elevated T1 value.
That is, however, in most cases not a problem of the reference wave function but rather of the applicability of CC to your system. You can of course try around with larger basis sets which may reduce the value a little, but if a system has multiconfiguration character, and with radicals you tend to have that, you will probably have to shift to an actual multireference method.
Thank you very much for the clarification. Is it allowed to calculate the reaction energy using different method? For example for the product using the CCSD(T) method and the reactant using the multireference method? Any advice?
No, that wouldn't make sense. You have to stay within the same method (and basis set for the sake of completion), otherwise errors that normally self-compensate will pile up, so if you switch e.g. to CASSCF or CASPT2 for the educt, the same needs to be done for the product.
- Badges
- Science topic
- Similar topics
- Books