Collaborative Research: NSCI: SI2-SSE: Time Stepping and Exchange-Correlation Modules for Massively Parallel Real-Time Time-Dependent DFT

点击翻译按钮获取中文摘要

Recent advances in high-performance (HPC) computing allow simulations of quantum dynamics of electrons in complex materials, and such simulations are central to advancing various medical and semiconductor technologies, ranging from proton beam cancer therapy to fabricating faster and smaller electronics. At the same time, the increasing scale and complexity of modern high-performance computers exposed a need for development of scientific software that is tailored for computers with large numbers of processors so that simulations can efficiently take advantage of increasing computing power. This project advances scientific software for simulating quantum dynamics of electrons for high-performance computers with tens and hundreds of thousands of processors that are becoming widely available. This work builds the HPC academic research community around the proposed software by extending the existing software available for quantum dynamics simulation with better user-friendly features and analysis techniques. In the process, this project engages graduate students and early-career researchers to use and further develop scientific software for high-performance computers in general. Additionally, a summer school for hands-on training will be conducted. The open source software will be made available to the community on Github (public repository). Real-time propagation in time-dependent density functional theory (RT-TDDFT) is becoming increasingly popular for studying non-equilibrium electronic dynamics both in the linear regime and beyond linear response. RT-TDDFT can be combined to study coupled dynamics of quantum-mechanical electrons with the movement of classical ions within Ehrenfest dynamics. In spite of its great promise, RT-TDDFT is computationally very demanding, especially for studying large condensed-matter systems. The large cost arises from small time steps of numerical integration of the electron dynamics, rendering accurate (hybrid) exchange-correlation (XC) functionals unfeasible, despite their clear benefits. In addition, while modern high-performance computing (HPC) helps tackling great scientific questions, massively parallel, hybrid-paradigm architectures present new challenges. Theoretical and algorithmic methods need to be developed in order to take full advantage of modern massively parallel HPC. This work builds new modules for the RT-TDDFT software component of the Qb@ll code, that enables a large community of researchers to perform advanced first-principles simulations of non-equilibrium electron dynamics in complex condensed-phase systems, using massively parallel HPC. This is done through developing (1) new modules for numerical integration that propagate the underlying non-linear partial differential equations in real time with high efficiency and accuracy, and (2) new modules for improved approximations of the underlying electronic structure, using a modern meta-generalized-gradient XC functional. Furthermore, the work builds the HPC academic research community around RT-TDDFT within the Qb@ll code through (1) development of user-friendly features that interface Qb@ll with other code and analysis techniques and (2) engagement of early-career scientists by incorporating hands-on training on RT-TDDFT using the Qb@ll code in TDDFT summer school.This project is supported by the Office of Advanced Cyberinfrastructure in the Directorate for Computer and Information Science and Engineering, the Materials Research Division and Chemistry Division in the Directorate of Mathematical and Physical Sciences.