Public Abstract

DE-SC0018331: FLO-SIC: Efficient Density Functional Theory Calculations

Award Status: Active

Institution: Central Michigan University, Mount Pleasant, MI
UEI: JJDYK36PRTL5
DUNS: 624134037

Most Recent Award Date: 08/07/2023
Number of Support Periods: 7
PM: Holder, Aaron

Current Budget Period: 09/01/2023 - 08/31/2024
Current Project Period: 09/01/2021 - 08/31/2025
PI: Jackson, Koblar

Supplement Budget Period: N/A

Public Abstract

FLO-SIC: Efficient density functional theory calculations without self-interaction Koblar A. Jackson (Lead PI) and Juan Peralta (co-PI), Central Michigan University John P. Perdew and Adrienn Ruzsinszky (co-PIs), Temple University Tunna Baruah, Mark R. Pederson and Rajendra R. Zope (co-PIs), University of Texas at El Paso George Christou (co-PI), University of Florida J. Karl Johnson, (co-PI), University of Pittsburgh Density functional theory (DFT) is the foundation of the most widely used computational modeling techniques in chemistry and materials science. DFT methods allow useful predictions of the properties of matter at the atomic scale and are computationally more efficient than alternative methods based on many-electron wave functions. Yet self-interaction errors (SIEs) are present in all DFT calculations and can skew their predictions, especially when bonds between atoms become stretched, but also for some equilibrium geometries when transition metal atoms are involved. This renewal project continues research aimed at eliminating SIEs and enabling efficient DFT calculations without self-interaction that are accurate under all circumstances. In the original project, software based on the Fermi-Löwdin orbital (FLO) self-interaction correction (SIC) method was developed and made public. FLOSIC is a computationally efficient approach to implementing the Perdew-Zunger (PZ) SIC theory. Because of its efficiency, the FLOSIC code allowed extensive testing and benchmarking of PZ-SICthat was not feasible before. Key insights from that work led to a new understanding of the “paradox of SIC”, the fact that PZ-SIC is effective when SIEs are prominent, but degrades the otherwise accurate predictions of DFT methods when they are not. Based on these insights, new alternatives to PZ-SIC were proposed and implemented in the FLOSIC code. Initial tests have yielded remarkably improved results for many properties affected by SIEs. In the renewal project, research on the alternative methods will expand, supported by continued development of the FLOSIC software. The key objective is to create a self-interaction correction that works with the most accurate DFT methods, correcting only when needed. An extension of the FLOSIC software will allow SIC calculations to be used for solids and interfaces, as well as for molecules and molecular complexes. The software will be made more user-friendly, automating technical details unfamiliar to users of other DFT software packages. Workflows that take advantage of large-scale supercomputing resources will also be developed. Finally, FLOSIC methods will be applied to technologically important problems, including studies of molecular magnets and their potential use in quantum computing, investigating routes to generating hydrogen and oxygen gases by catalytic processes based on transition metal chemistry, and converting CO₂ into useful products such as methanol through catalysis in metal-organic framework materials. If fully successful, FLOSIC-based methods will replace conventional DFT methods and realize the goal of “efficient DFT calculations without self-interaction.”

FLO-SIC: Efficient density functional theory calculations without self-interaction

Koblar A. Jackson (Lead PI) and Juan Peralta (co-PI), Central Michigan University

John P. Perdew and Adrienn Ruzsinszky (co-PIs), Temple University

Tunna Baruah, Mark R. Pederson and Rajendra R. Zope (co-PIs), University of Texas at El Paso

George Christou (co-PI), University of Florida

J. Karl Johnson, (co-PI), University of Pittsburgh

Density functional theory (DFT) is the foundation of the most widely used computational modeling techniques in chemistry and materials science. DFT methods allow useful predictions of the properties of matter at the atomic scale and are computationally more efficient than alternative methods based on many-electron wave functions. Yet self-interaction errors (SIEs) are present in all DFT calculations and can skew their predictions, especially when bonds between atoms become stretched, but also for some equilibrium geometries when transition metal atoms are involved. This renewal project continues research aimed at eliminating SIEs and enabling efficient DFT calculations without self-interaction that are accurate under all circumstances. In the original project, software based on the Fermi-Löwdin orbital (FLO) self-interaction correction (SIC) method was developed and made public. FLOSIC is a computationally efficient approach to implementing the Perdew-Zunger (PZ) SIC theory. Because of its efficiency, the FLOSIC code allowed extensive testing and benchmarking of PZ-SICthat was not feasible before. Key insights from that work led to a new understanding of the “paradox of SIC”, the fact that PZ-SIC is effective when SIEs are prominent, but degrades the otherwise accurate predictions of DFT methods when they are not. Based on these insights, new alternatives to PZ-SIC were proposed and implemented in the FLOSIC code. Initial tests have yielded remarkably improved results for many properties affected by SIEs. In the renewal project, research on the alternative methods will expand, supported by continued development of the FLOSIC software. The key objective is to create a self-interaction correction that works with the most accurate DFT methods, correcting only when needed. An extension of the FLOSIC software will allow SIC calculations to be used for solids and interfaces, as well as for molecules and molecular complexes. The software will be made more user-friendly, automating technical details unfamiliar to users of other DFT software packages. Workflows that take advantage of large-scale supercomputing resources will also be developed. Finally, FLOSIC methods will be applied to technologically important problems, including studies of molecular magnets and their potential use in quantum computing, investigating routes to generating hydrogen and oxygen gases by catalytic processes based on transition metal chemistry, and converting CO₂ into useful products such as methanol through catalysis in metal-organic framework materials. If fully successful, FLOSIC-based methods will replace conventional DFT methods and realize the goal of “efficient DFT calculations without self-interaction.”