Interpolating Nonadiabatic Molecular Dynamics Hamiltonian with Inverse Fast Fourier Transform

Abstract

Nonadiabatic (NA) molecular dynamics (MD) allows one to investigate far-from-equilibrium processes in nanoscale and molecular materials at the atomistic level and in the time domain, mimicking time-resolved spectroscopic experiments. Ab initio NAMD is limited to about 100 atoms and a few picoseconds, due to computational cost of excitation energies and NA couplings. We develop a straightforward methodology that can extend ab initio quality NAMD to nanoseconds and thousands of atoms. The ab initio NAMD Hamiltonian is sampled and interpolated along a trajectory using a Fourier transform, and then, it is used to perform NAMD with known algorithms. The methodology relies on the classical path approximation, which holds for many materials and processes. To achieve a complete ab initio quality description, the trajectory can be obtained using an ab initio trained machine learning force field. The method is demonstrated with charge carrier trapping and relaxation in hybrid organic–inorganic and all-inorganic metal halide perovskites that exhibit complex dynamics and are actively studied for optoelectronic applications.

Defect modeling and control in structurally and compositionally complex materials

“…The latter is unavoidably very slow in computation if one employs accurate hybrid functionals in conjunction with spinorbit coupling to reliably capture the electronic structure of complex materials. Computational schemes that significantly accelerate these calculations (for example, constructing the NAMD Hamiltonian with Fourier interpolation 87 or machine learning models 88 ) would be useful.…”

Carrier Transport in 2D Hybrid Organic-Inorganic Perovskites: The Role of Spacer Molecules

“…The fewest switches surface hopping method can also be applied to this work. The DISH method, by realizing the detailed balance and the decoherence effect through the electron wave function collapsing, allows a real-time description of carrier transport and ionic movement. − Moreover, in most solid-state systems where the wave functions are usually delocalized and the ions are mostly heavy atoms, classical path approximation (CPA) can be applied by ignoring the “feedback” of the excited carrier to the ionic motions, which can reduce the computation cost significantly. In this case, the ground-state ionic trajectory is used throughout the DISH simulation, and the NAMD becomes a postprocessing of the ground state ab initio MD.…”

Interpolating Nonadiabatic Molecular Dynamics Hamiltonian with Bidirectional Long Short-Term Memory Networks

“…In this paper, we introduce a novel approach for predicting excitation energies and NACs for NAMD simulations using ML time-series interpolation. Employing Bidirectional Long Short-Term Memory networks (Bi-LSTM), we obtain a significant improvement over the previous methods, including inverse fast Fourier transform (iFFT), kernel ridge regression and neural networks . Well designed for time-interpolation, Bi-LSTM is a multiscale model that allows us to gain a two orders of magnitude savings compared to the direct ab initio calculation, as demonstrated with three metal-halide perovskite (MHP) systems that can undergo slow, large-scale atomic motions.…”