Spin-state gaps and self-interaction-corrected density functional approximations: Octahedral Fe(II) complexes as case...

Area of Science:

• Computational Chemistry
• Quantum Chemistry
• Materials Science

Background:

• Accurate prediction of spin-state energy differences is vital for understanding spin crossover phenomena in transition metal complexes.
• Density functional approximations (DFAs), particularly local and semi-local ones, struggle with delocalization errors, leading to challenges in predicting these energy differences.
• Self-interaction error (SIE) is a known limitation of DFAs that affects the accuracy of calculated electronic properties.

Purpose of the Study:

• To investigate the impact of removing self-interaction error from local spin density approximation (LSDA) and Perdew-Burke-Ernzerhof generalized gradient approximation (PBE GGA) on spin-state gaps.
• To evaluate the performance of a recently developed locally scaled self-interaction correction (LSIC) method for predicting spin-state energy differences in Fe(II) complexes.
• To compare the effectiveness of different SIE correction strategies, including Perdew-Zunger self-interaction correction (PZSIC) and LSIC, against high-level computational methods.

Main Methods:

• Application of the locally scaled self-interaction correction (LSIC) method to LSDA and PBE GGA.
• Utilized Perdew-Zunger self-interaction correction (PZSIC) as a specific case of LSIC.
• Employed perturbative and quasi-self-consistent approaches for LSIC implementation.
• Compared results with reference diffusion Monte Carlo (DMC) and coupled-cluster single double and perturbative triple (CCSD(T)) calculations.

Main Results:

• The PZSIC method significantly overestimates spin-state gaps, failing to improve upon standard DFAs.
• Perturbative LSIC-LSDA using PZSIC densities shows significant improvement, with a mean absolute error (MAE) of 0.51 eV, though it slightly overcorrects for strong ligands like CO.
• Quasi-self-consistent LSIC-LSDA yields correct spin-state gap signs for all ligands, achieving an MAE of 0.56 eV, comparable to the benchmark CCSD(T) result (0.49 eV).

Conclusions:

• Self-interaction error correction is crucial for accurate spin-state gap predictions in Fe(II) complexes.
• The LSIC method, particularly in its quasi-self-consistent form, offers a promising approach to improve the accuracy of DFAs for spin crossover systems.
• LSIC provides a computationally feasible and accurate alternative to high-level correlated methods for studying spin crossover phenomena.