Monomeric Neural Network Potential for General Covalent Molecules: Linear Alkanes as an Example.

Area of Science:

• Computational Chemistry
• Materials Science
• Chemical Physics

Background:

• Machine-learning potentials (MLPs) are crucial for molecular simulations.
• Developing accurate and efficient MLPs remains a challenge.

Purpose of the Study:

• Extend the MB-PIPNet framework to general covalently bonded molecular systems.
• Improve computational efficiency and accuracy in MLPs.

Main Methods:

• Combined monomer-based energy decomposition, permutationally invariant polynomial (PIP) descriptors, and neural networks.
• Utilized a fragmentation-based strategy for potential energy representation.
• Applied the framework to linear alkanes (C14H30) for benchmarking.

Main Results:

• MB-PIPNet accurately reproduces ab initio electronic energies.
• The model reliably captures molecular properties like torsional profiles and vibrational frequencies.
• Demonstrated significant computational efficiency advantages over other MLP models for energy and force evaluations.

Conclusions:

• MB-PIPNet is a scalable and efficient framework for constructing MLPs.
• Provides a new pathway for large-scale quantum and classical simulations of complex molecular systems.