Improving the scaling and performance of multiple time stepping‐based molecular dynamics with hybrid density functionals

Abstract

Density functionals at the level of the generalized gradient approximation (GGA) and a plane-wave basis set are widely used today to perform ab initio molecular dynamics (AIMD) simulations. Going up in the ladder of accuracy of density functionals from GGA (second rung) to hybrid density functionals (fourth rung) is much desired pertaining to the accuracy of the latter in describing structure, dynamics, and energetics of molecular and condensed matter systems. On the other hand, hybrid density functional based AIMD simulations are about two orders of magnitude slower than GGA based AIMD for systems containing $100 atoms using $100 compute cores. Two methods, namely MTACE and s-MTACE, based on a multiple time step integrator and adaptively compressed exchange operator formalism are able to provide a speed-up of about 7-9 in performing hybrid density functional based AIMD. In this work, we report an implementation of these methods using a task-group based parallelization within the CPMD program package, with the intention to take advantage of the large number of compute cores available on modern high-performance computing platforms. We present here the boost in performance achieved through this algorithm.

Hydroxide Diffusion in Functionalized Cylindrical Nanopores as Idealized Models of Anion Exchange Membrane Environments: An Ab Initio Molecular Dynamics Study

“…As an alternative to meta-GGA functionals, hybrid functionals could be employed. However, for the system sizes employed in this study, a time-saving feature, such as the combination of adaptively compressed exchange operator schemes and multiple time-stepping, , would be needed. Ultimately, extending the system size in order to examine more sophisticated and representative models will be desirable, at which point, AIMD simulations will become infeasible, and more efficient approaches such as reactive neural network potentials become an attractive option that can retain the accuracy of AIMD if they are trained appropriately.…”

High-Throughput Condensed-Phase Hybrid Density Functional Theory for Large-Scale Finite-Gap Systems: The SeA Approach

“…Here, we would also emphasize that the applicability of extends beyond high-throughput EXX-SCF calculations (i.e., single-point energy and ionic force evaluations) at the hybrid DFT level. For instance, (in its current form) can be used to accelerate Born–Oppenheimer AIMD (BOMD) simulations (to a significantly larger degree than the original implementation of , in the module of due to the additional computational savings from the ACE operator) and also has the potential to increase the length- and time-scales accessible by AIMD (both CPMD and BOMD) when used in conjunction with the multiple time scale approach , based on the ACE operator. ,, Due to the continuous time evolution of the trajectory during AIMD simulations, is not restricted to SCDM orbitals and could also be used in conjunction with other localization schemes (e.g., MLWFs , ) by keeping track of (and continuously refining) the unitary operator connecting the local and canonical representations of the occupied space. When performing hybrid DFT-based CPMD, a gauge-invariant sampling can be achieved using the field-theoretic approach proposed by Tuckerman and co-workers or direct propagation in the canonical orbital representation while evaluating all EXX-related quantities in the localized orbital representation (i.e., akin to what was done for in this work).…”

Hybrid Functional and Plane Waves based Ab Initio Molecular Dynamics Study of the Aqueous Fe2+/Fe3+ Redox Reaction**

“…Recently, we have proposed a few methods to speed up hybrid functional and plane waves (PWs) based AIMD simulations. [61][62][63][64][65] These methods were successfully applied to model chemical reactions in explicit solvent. Here, we use one such method, namely the noise stabilized molecular dynamics (NSMD).…”

“…Thus, AIMD simulations with hybrid functionals are rarely used [48,54] to study redox reactions. Recently, we have proposed a few methods to speed up hybrid functional and PWs based AIMD simulations [62–66] . These methods were successfully applied to model chemical reactions in explicit solvent.…”