Machine Learning Interatomic Potentials Enable Molecular Dynamics Simulations of Doped MoS₂.

Area of Science:

• Materials Science
• Computational Chemistry
• Nanotechnology

Background:

• Tuning molybdenum disulfide (MoS2) performance requires understanding dopant effects.
• Current molecular dynamics simulations for doped MoS2 are limited by a lack of accurate interatomic potentials.

Purpose of the Study:

• To evaluate the accuracy of a universal machine learning interatomic potential (MLIP) for 25 different MoS2 dopants.
• To establish a computational workflow for designing doped MoS2 materials using MLIPs.

Main Methods:

• Benchmarking MLIP-predicted formation energies and structural changes against density functional theory (DFT) calculations for 25 MoS2 dopants.
• Performing heating-cooling simulations using the validated MLIP on MoS2 supercells.

Main Results:

• The universal MLIP accurately predicts formation energies and structural changes for various MoS2 dopants.
• MLIP simulations capture complex phenomena like dopant clustering and MoS2 fracturing at significantly reduced computational cost compared to DFT.
• The study provides a validated computational framework for high-throughput screening of doped MoS2.

Conclusions:

• Universal MLIPs are a viable and efficient tool for discovering and designing doped MoS2 materials.
• This work enables accelerated optimization of MoS2 for tribological, electronic, and optoelectronic applications.