Avoiding root flipping and preserving degeneracy in state-averaged MRCI

[Rafalela Eliasquevici]

Dear MOLPRO community,

I am currently calculating potential energy curves (PECs) for several excited states of a diatomic molecule, with the aim of using them in subsequent spin-orbit coupling calculations and for other properties.

Within the C2V framework, I am calculating four states distributed among the four irreducible representations. In symmetry 1, I have a Δ state; in symmetries 2 and 3, the corresponding Π and Φ components; and in symmetry 4, a Δ state together with a Σ⁻ state.

For the subsequent calculations, I need the two components of the Δ state belonging to different symmetries to remain degenerate. However, when I perform the multistate calculation, the corresponding PECs are not degenerate.

I have therefore tried separating the states using REFSTATE, for example:

STATE,1,1 REFSTATE,1,1 STATE,1,2 REFSTATE,1,2

However, when the Σ and Δ reference states change their energetic ordering, the references switch character. This causes problems with state tracking: the degeneracy between the Δ components is lost, and the Σ curve is also affected because it overlaps with the fourth component of the Δ state.

I am using state-averaged MRCI and specifying the appropriate LQUANT values. The expected values of LZLZ are also consistent throughout the calculation.

Is there a recommended way in MOLPRO to avoid root flipping in this situation while maintaining the correct degeneracy between the components of the Δ state? In particular, I would like to reliably track the electronic states by their character rather than simply by their energy ordering, so that the resulting PECs can be used consistently in subsequent spin-orbit calculations.

Any advice on the appropriate use of REFSTATE, LQUANT, SWAP, REF, or other state-tracking options in state-averaged MRCI would be greatly appreciated.

Best regards,

Rafaela Eliasquevici