TY - JOUR
T1 - Simple approach to more efficient density functional theory simulations
AU - Benaissa, Mohammed
AU - Ouahrani, Tarik
AU - Hatada, Keisuke
AU - Yoshikawa, Kazuki
AU - Sébilleau, Didier
N1 - Publisher Copyright:
© 2025 Elsevier B.V.
PY - 2025/3
Y1 - 2025/3
N2 - In physics, chemistry and materials science, density functional theory (DFT) is a common method for studying the electronic structure of many-body systems. DFT can address complex materials with relatively low computational costs compared to other physics-based methods; however, it still requires considerable computational power for large-scale simulations. With the aim of reducing the computational footprint of DFT, we suggest the use of the data-efficient Bayesian algorithm to optimize the charge mixing parameters, which reduces the self-consistent field iterations necessary to reach convergence. We show that our algorithm can achieve faster convergence than the default parameters when the VASP code is used as a proof of concept, resulting in significant time savings in DFT simulations and therefore providing a systematic guide for more efficient DFT simulations. We propose adding this procedure to the well-known convergence test procedures, such as cutoff-energy and k-point convergence tests.
AB - In physics, chemistry and materials science, density functional theory (DFT) is a common method for studying the electronic structure of many-body systems. DFT can address complex materials with relatively low computational costs compared to other physics-based methods; however, it still requires considerable computational power for large-scale simulations. With the aim of reducing the computational footprint of DFT, we suggest the use of the data-efficient Bayesian algorithm to optimize the charge mixing parameters, which reduces the self-consistent field iterations necessary to reach convergence. We show that our algorithm can achieve faster convergence than the default parameters when the VASP code is used as a proof of concept, resulting in significant time savings in DFT simulations and therefore providing a systematic guide for more efficient DFT simulations. We propose adding this procedure to the well-known convergence test procedures, such as cutoff-energy and k-point convergence tests.
KW - Bayesian optimization
KW - Charge mixing parameters
KW - Self-consistent field
UR - http://www.scopus.com/inward/record.url?scp=85216959670&partnerID=8YFLogxK
U2 - 10.1016/j.cocom.2025.e01012
DO - 10.1016/j.cocom.2025.e01012
M3 - 学術論文
AN - SCOPUS:85216959670
SN - 2352-2143
VL - 42
JO - Computational Condensed Matter
JF - Computational Condensed Matter
M1 - e01012
ER -