Interpolating Nonadiabatic Molecular Dynamics Hamiltonian with Inverse Fast Fourier Transform

Area of Science:

• Computational Chemistry
• Materials Science
• Quantum Mechanics

Background:

• Nonadiabatic molecular dynamics (NA MD) is vital for studying far-from-equilibrium processes in materials.
• Current ab initio NA MD methods are computationally expensive, limiting simulations to small systems (approx. 100 atoms) and short timescales (picoseconds).

Purpose of the Study:

• To develop a scalable methodology for extending ab initio quality nonadiabatic molecular dynamics simulations.
• To enable atomistic simulations of nanoscale and molecular materials over nanosecond timescales and with thousands of atoms.

Main Methods:

• A straightforward methodology is developed to sample and interpolate the ab initio NA MD Hamiltonian using Fourier transforms along a trajectory.
• The classical path approximation is employed, ensuring applicability to various materials and processes.
• An ab initio trained machine learning force field can be used for trajectory generation to maintain ab initio quality.

Main Results:

• The developed method significantly extends the accessible length and timescales for ab initio quality NA MD simulations.
• Demonstrated successful application to charge carrier trapping and relaxation in metal halide perovskites.
• The methodology overcomes the computational limitations of traditional ab initio NA MD.

Conclusions:

• The new methodology provides a computationally efficient pathway to achieve ab initio quality NA MD for complex systems.
• This advancement is crucial for understanding and designing advanced optoelectronic materials like perovskites.
• Enables atomistic, time-domain investigations of excited-state dynamics previously inaccessible.