Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
AKPIK: Arbeitskreis Physik, moderne Informationstechnologie und Künstliche Intelligenz
AKPIK 6: AI Methods for Physics and Materials Science
AKPIK 6.7: Vortrag
Donnerstag, 12. März 2026, 18:15–18:30, BEY/0127
Multi-fidelity and -objective optimization of ONCV pseudopotentials — •Austin Zadoks1, Cameron Hargreaves2, Jusong Yu1, Weiguo Jing2, Matteo Giantomassi2, Gian-Marco Rignanese2, and Giovanni Pizzi1 — 1PSI Center for Scientific Computing, Theory and Data, 5232 Villigen PSI, Switzerland — 2Institute of Condensed Matter and Nanosciences, UCLouvain, Louvain-la-Neuve, Belgium
The pseudopotential (PSP) approximation is essential to the tractability of many first-principles methods. However, it requires balancing the number of pseudized states, basis-set convergence, and accuracy w.r.t. all-electron (AE) results. One notable method for constructing soft, faithful, and widely-supported PSPs is the optimized norm-conserving Vanderbilt (ONCV) approach1. Various strategies have been proposed for generating meta-optimal tables of ONCVPSPs such as the SG-152, PseudoDojo3, and SPMS4. Recent efforts to verify DFT codes have highlighted the importance of high-quality PSPs and expanded the necessary AE reference data, notably through Z=965. We present a fully-automated multi-fidelity multi-objective Bayesian optimization of ONCVPSPs targeting these reference data. This approach allows for the efficient mapping of the high-fidelity PW-DFT Pareto frontier by leveraging lower-fidelity intermediate radial- and PW-DFT results. 1D.R. Hamann. PRB, 88 (2013). 2M. Schlipf & F. Gygi. Comp. Phys. Comms., 196 (2015). 3M.J. van Setten, et al. Comp. Phys. Comms., 226 (2018). 4M.F. Shojaei, et al. Comp. Phys. Comms., 283 (2023). 5 E. Bosoni, et al. Nat. Rev. Phys., 6 (2024).
Keywords: pseudopotential; Bayesian optimization