Machine learning-accelerated path integral molecular dynamics simulations of reactive organic electrolytes.

Area of Science:

• Computational chemistry
• Materials science
• Electrochemistry

Background:

• Hydrogen-bonded electrolytes show promise for clean energy due to accelerated proton transport.
• Accurate modeling requires computationally expensive quantum mechanical simulations of condensed phases.
• Understanding microscopic details is key to designing efficient electrolyte technologies.

Purpose of the Study:

• To demonstrate the efficiency of density functional theory (DFT)-trained machine learning potentials (MLPs) for accelerating path integral molecular dynamics (PIMD) simulations.
• To benchmark PIMD simulations using different DFT functionals for imidazole-levulinic acid mixtures.
• To introduce and validate a ring polymer contraction approach for further PIMD acceleration.

Main Methods:

• Utilized DFT-trained MLPs to accelerate PIMD simulations.
• Performed PIMD simulations on imidazole-levulinic acid mixtures with varying DFT exchange-correlation functionals.
• Introduced and benchmarked a ring polymer contraction method combined with short-range MLPs.

Main Results:

• MLPs significantly accelerate PIMD simulations for studying proton transport in hydrogen-bonded electrolytes.
• PIMD simulations with different DFT functionals were benchmarked for accuracy in predicting electrolyte properties.
• The ring polymer contraction approach achieved an additional fourfold speedup in PIMD simulations.

Conclusions:

• DFT-trained MLPs combined with PIMD offer an efficient pathway for modeling complex electrolyte systems.
• The ring polymer contraction method provides a computationally tractable approach for large-scale PIMD simulations.
• This work facilitates the design and optimization of advanced electrolytes for clean energy applications.