Parallel transport time-dependent density functional theory calculations with hybrid functional on summit

Abstract

Real-time time-dependent density functional theory (rt-TDDFT) with hybrid exchange-correlation functional has wide-ranging applications in chemistry and material science simulations. However, it can be thousands of times more expensive than a conventional ground state DFT simulation, hence is limited to small systems. In this paper, we accelerate hybrid functional rt-TDDFT calculations using the parallel transport gauge formalism, and the GPU implementation on Summit. Our implementation can efficiently scale to 786 GPUs for a large system with 1536 silicon atoms, and the wall clock time is only 1.5 hours per femtosecond. This unprecedented speed enables the simulation of large systems with more than 1000 atoms using rt-TDDFT and hybrid functional.

Parallel transport dynamics for mixed quantum states with applications to time-dependent density functional theory

“…In this section, we provide the numerical results of the parallel transport dynamics. Extensive numerical results and implementations on real chemical systems have been demonstrated for the PT dynamics with pure initial states [1,16,17,18]. Hence, we focus on the case of a mixed initial state in this section.…”

“…(1.1). When combined with implicit integrators (such as the Crank-Nicolson method or the implicit midpoint rule), the PT dynamics has been applied to rt-TDDFT simulations for real materials with thousands of atoms at the level of generalized gradient approximation exchange-correlation functionals (GGA, such as the Perdew-Burke-Ernzerhof [31] functional) [16] and hybrid exchange-correlation functionals (such as the Heyd-Scuseria-Ernzerhof [15] functional) [17,18].…”

Large-scale ab initio simulation of light–matter interaction at the atomic scale in Fugaku

“…Last year, a large-scale real-time TDDFT calculation with the hybrid functional was reported using Summit at ORNL (Jia et al, 2019). Using the hybrid functional, the dominant calculation becomes the operation of the nonlocal exchange operator on electron orbitals that scales as O ( N 3 ).…”

On the Efficient Evaluation of the Exchange Correlation Potential on Graphics Processing Unit Clusters

“…Recently, there has been significant research effort afforded to porting electronic structure software to the GPU (Gordon et al, 2020 ). In the case of large-scale calculations, much work has gone into the development of massively parallel GPU implementations of methods based on plane wave (Maintz et al, 2011 ; Wang et al, 2011 ; Jia et al, 2019 ), real space (Andrade and Aspuru-Guzik, 2013 ; Hakala et al, 2013 ), finite element (Das et al, 2019 ; Motamarri et al, 2020 ), and various other discretizations (Genovese et al, 2009 ; van Schoot and Visscher, 2016 ; Yoshikawa et al, 2019 ; Huhn et al, 2020 ) of the Kohn–Sham equations. In this work, we consider the Gaussian basis set discretization of the Kohn–Sham equations (Pople et al, 1992 ), which poses a number of challenges for GPU implementations.…”

“…Recently, there has been significant research effort afforded to porting electronic structure software to the GPU [28]. In the case of large scale calculations, much work has gone into the development of massively parallel GPU implementations of methods based on plane wave [38,52,79], real-space [3,31], finite-element [16,55], and various other discretizations [36,25,84,78] of the Kohn-Sham equations. In this work, we consider the Gaussian basis set discretization of the Kohn-Sham equations [68], which poses a number of challenges for GPU implementations.…”