Speaker: Michael Römelt, Humboldt-Universität zu Berlin 

The exploration of energy landscapes becomes complicated when multiple electronic states are close in energy, as for example observed in transition metal complexes with multiple open shells. In such cases, multireference electronic structure methods provide a sound approach to deal with strong correlation effects.^[1,2] Recently, we have reported a configuration-based variant of heatbath-CI (HCI) that allows for SCF calculations with large active spaces while retaining spin symmetry.^[3] Our implementation utilizes the HCI logic^[4], extensively exploits symmetry and introduces a novel parallelization scheme based on configuration prefixes to achieve high efficiency. In addition to energy calculations, state averaged nuclear gradients allow for the exploration of energy landscapes for the ground and excited states even in the presence of (near-) degeneracies.^[5] Reliable results can, however, only be obtained when dynamic electron correlation effects are considered.^[6] Therefore, we present two variants of second order N-electron valence state perturbation theory (NEVPT2) to capture these effects on top of our HCISCF method. A “canonical” variant incorporates the important residual terms^[7] to avoid the appearance of false intruder states in the infamous S_i and S_a perturber spaces. In contrast, a combination of Epstein-Nesbet PT2 and NEVPT2 avoids the critical terms altogether. Both variants make use of the heatbath-CI logic to dramatically reduce the associated computational cost and furthermore use prefix-based parallelization to extend their applicability. Embedded in the open-source HUMboldt MultiReference (HUMMR) program^[8] this allows for accurate calculations of chemically relevant systems with many strongly correlated electrons.

 

[1]      D. D. Malik et al.,  J. Am. Chem. Soc. 2024, 146, 20, 13817-13835
[2]      O. S. Ablyasova, M. Ugandi, E. B. Boydas, M. da Silva Santos, M. Flach, V. Zamudio-Bayer, M. Roemelt, J. T. Lau, K Hirsch, J. Am. Chem. Soc. 2025, 147, 9, 7336-7344
[3]      M. Ugandi, M. Roemelt, J. Comput. Chem. 2023, 44, 2374-2390
[4]      A. A. Holmes, H. J. Changlani, C. J. Umrigar, J. Chem. Theory Comput., 2016, 12, 1561-1571
[5]      M. Ugandi, M. Roemelt, J. Chem. Theory Comput. 2025, 21, 3930-3944
[6]      A. Khedkar, M. Roemelt, Phys. Chem. Chem. Phys 2021, 23, 17097-17112
[7]      Y. Guo, K. Sivalingam, C. Kollmar, F. Neese, J. Chem. Phys. 2021, 154, 214113
[8]      https://scm.cms.hu-berlin.de/hummr-dev-team/hummr