High-quality quantum chemical data for spin state determination in transition-metal complexes

Abstract

We have developed a high-level quantum chemical dataset for spin state determination in transition-metal complexes. In this dataset, we have used varied levels of theory with a combined CASPT2/CCSD(T) level as the most accurate one. Due to the presence of results at these different levels of theory, the errors associated with each of these theories can be estimated, and machine learning approaches can be developed to boost the accuracy further using combined approaches. Two different machine learning approaches and models based on these datasets show their efficacy and need for such high-level datasets, especially for transition-metal complexes.