I conducted a scan in the ground state using M06-HF functional and cc-pVDZ basis set on the system which I am working. Then TDDFT single point energy calculations of the scanned structures were done at the triplet state using cc-pVTZ basis set. The potential energy surfaces corresponding to triplet and ground state were plotted and an anomalous crossing was observed between the surfaces! So, as to find a solution I wanted to check whether it's possible to find an avoided crossing using CASSCF method!
The PES plot of the scan is attached
Just a couple of thoughts (of which I'm sure you're already aware) but comparing using two different bases is very problematic, but easily remedied, especially where you're talking about crossings.. Also, how do you know you are staying on the same potential energy surface during your scans? Did orbital rotations occur as you moved along your coordinate? If so, did you check to make sure you didn't allow for these rotations so that you continued on the same PES?
Without the inclusion of spin orbit coupling CASSCF or any other method will be missing an important (i.e., crucial) component of the interaction between states of different spin. Also, CASSCF is missing dynamical correlation effects and something like CASPT2 will heavily over estimate them (NEVPT2 can give better results but then you have to worry about intruder states which are non-physical, though this has been dealt with by some researchers to a certain extent). All of the CI based approaches are not size-extensive either, which can affect the quality of their description of near degeneracies of states and the Davidson and similar corrections, while assumed to restore this for multireference approaches, has not been proven to actually do this. If you switch to a coupled-cluster based approach which for ground state calculations is size extensive, you run into the issue that they are non-Hermitian theories in general and that can also affect the quality of the crossing/avoidance. Though this has been considered and dealt with as well by certain research groups.
So depending on your system and the problem you are trying to solve, you'll need to figure out which effects are more important than others and decide which method will be able to do what you need. Without benchmarking against a robust many-body method, DFT results will, at best, be unknown if they can be used reliably (e.g., just because you get the correct number to compare with experiment doesn't mean the description provided by your functional is physically meaningful).
Anyway, my two cents which should be taken with a pound of salt. Best of luck with your problem.
See the ORCA program, where CASSCF and state average CASSCF calculations can be done to analyze different excited states.
Just a couple of thoughts (of which I'm sure you're already aware) but comparing using two different bases is very problematic, but easily remedied, especially where you're talking about crossings.. Also, how do you know you are staying on the same potential energy surface during your scans? Did orbital rotations occur as you moved along your coordinate? If so, did you check to make sure you didn't allow for these rotations so that you continued on the same PES?
Without the inclusion of spin orbit coupling CASSCF or any other method will be missing an important (i.e., crucial) component of the interaction between states of different spin. Also, CASSCF is missing dynamical correlation effects and something like CASPT2 will heavily over estimate them (NEVPT2 can give better results but then you have to worry about intruder states which are non-physical, though this has been dealt with by some researchers to a certain extent). All of the CI based approaches are not size-extensive either, which can affect the quality of their description of near degeneracies of states and the Davidson and similar corrections, while assumed to restore this for multireference approaches, has not been proven to actually do this. If you switch to a coupled-cluster based approach which for ground state calculations is size extensive, you run into the issue that they are non-Hermitian theories in general and that can also affect the quality of the crossing/avoidance. Though this has been considered and dealt with as well by certain research groups.
So depending on your system and the problem you are trying to solve, you'll need to figure out which effects are more important than others and decide which method will be able to do what you need. Without benchmarking against a robust many-body method, DFT results will, at best, be unknown if they can be used reliably (e.g., just because you get the correct number to compare with experiment doesn't mean the description provided by your functional is physically meaningful).
Anyway, my two cents which should be taken with a pound of salt. Best of luck with your problem.
- Badges
- Science topic
- Similar topics
- Chemistry
- Theoretical Chemistry