Performance of Quantum Chemistry Methods for Benchmark Set of Spin–State Energetics Derived from Experimental Data of 17 Transition Metal Complexes (SSE17) DOI Creative Commons
Mariusz Radoń, Gabriela Drabik, Maciej Hodorowicz

et al.

Chemical Science, Journal Year: 2024, Volume and Issue: 15(48), P. 20189 - 20204

Published: Jan. 1, 2024

Accurate prediction of spin-state energetics for transition metal (TM) complexes is a compelling problem in applied quantum chemistry, with enormous implications modeling catalytic reaction mechanisms and computational discovery materials. Computed are strongly method-dependent credible reference data scarce, making it difficult to conduct conclusive studies open-shell TM systems. Here, we present novel benchmark set first-row energetics, which derived from experimental 17 containing FeII, FeIII, CoII, CoIII, MnII, NiII chemically diverse ligands. The estimates adiabatic or vertical splittings, obtained spin crossover enthalpies energies spin-forbidden absorption bands, suitably back-corrected the vibrational environmental effects, employed as values benchmarking density functional theory (DFT) wave function methods. results demonstrate high accuracy coupled-cluster CCSD(T) method, features mean absolute error (MAE) 1.5 kcal mol-1 maximum -3.5 mol-1, outperforms all tested multireference methods: CASPT2, MRCI+Q, CASPT2/CC CASPT2+δMRCI. Switching Hartree-Fock Kohn-Sham orbitals not found consistently improve accuracy. best performing DFT methods double-hybrids (PWPB95-D3(BJ), B2PLYP-D3(BJ)) MAEs below 3 errors within 6 whereas so far recommended states (e.g., B3LYP*-D3(BJ) TPSSh-D3(BJ)) perform much worse 5-7 beyond 10 mol-1. This work first such extensive study chemistry use data. relevant proper choice characterize systems catalysis (bio)inorganic may also stimulate new developments quantum-chemical machine learning approaches.

Language: Английский

Improved modularity and new features in ipie: Toward even larger AFQMC calculations on CPUs and GPUs at zero and finite temperatures DOI
Tong Jiang, Moritz K. A. Baumgarten, Pierre‐François Loos

et al.

The Journal of Chemical Physics, Journal Year: 2024, Volume and Issue: 161(16)

Published: Oct. 25, 2024

ipie is a Python-based auxiliary-field quantum Monte Carlo (AFQMC) package that has undergone substantial improvements since its initial release [Malone et al., J. Chem. Theory Comput. 19(1), 109–121 (2023)]. This paper outlines the improved modularity and new capabilities implemented in ipie. We highlight ease of incorporating different trial walker types seamless integration with external libraries. enable distributed Hamiltonian simulations large systems otherwise would not fit on single central processing unit node or graphics (GPU) card. development enabled us to compute interaction energy benzene dimer 84 electrons 1512 orbitals multi-GPUs. Using CUDA cupy for NVIDIA GPUs, supports GPU-accelerated multi-slater determinant wavefunctions [Huang al. arXiv:2406.08314 (2024)] efficient highly accurate large-scale systems. allows near-exact ground state energies multi-reference clusters, [Cu2O2]2+ [Fe2S2(SCH3)4]2−. also describe implementations free projection AFQMC, finite temperature AFQMC electron–phonon systems, automatic differentiation calculating physical properties. These advancements position as leading platform research chemistry, facilitating more complex ambitious computational method their applications.

Language: Английский

Citations

3
Performance of Quantum Chemistry Methods for Benchmark Set of Spin–State Energetics Derived from Experimental Data of 17 Transition Metal Complexes (SSE17) DOI Creative Commons
Mariusz Radoń, Gabriela Drabik, Maciej Hodorowicz

et al.

Chemical Science, Journal Year: 2024, Volume and Issue: 15(48), P. 20189 - 20204

Published: Jan. 1, 2024

Accurate prediction of spin-state energetics for transition metal (TM) complexes is a compelling problem in applied quantum chemistry, with enormous implications modeling catalytic reaction mechanisms and computational discovery materials. Computed are strongly method-dependent credible reference data scarce, making it difficult to conduct conclusive studies open-shell TM systems. Here, we present novel benchmark set first-row energetics, which derived from experimental 17 containing FeII, FeIII, CoII, CoIII, MnII, NiII chemically diverse ligands. The estimates adiabatic or vertical splittings, obtained spin crossover enthalpies energies spin-forbidden absorption bands, suitably back-corrected the vibrational environmental effects, employed as values benchmarking density functional theory (DFT) wave function methods. results demonstrate high accuracy coupled-cluster CCSD(T) method, features mean absolute error (MAE) 1.5 kcal mol-1 maximum -3.5 mol-1, outperforms all tested multireference methods: CASPT2, MRCI+Q, CASPT2/CC CASPT2+δMRCI. Switching Hartree-Fock Kohn-Sham orbitals not found consistently improve accuracy. best performing DFT methods double-hybrids (PWPB95-D3(BJ), B2PLYP-D3(BJ)) MAEs below 3 errors within 6 whereas so far recommended states (e.g., B3LYP*-D3(BJ) TPSSh-D3(BJ)) perform much worse 5-7 beyond 10 mol-1. This work first such extensive study chemistry use data. relevant proper choice characterize systems catalysis (bio)inorganic may also stimulate new developments quantum-chemical machine learning approaches.

Language: Английский

Citations

1