Reactive machine-learned potentials for fluoropolymer binders: unified physical and chemical property validation

Area of Science:

• Materials Science
• Computational Chemistry
• Polymer Science

Background:

• Accurate atomistic simulation of polymer-bonded explosives (PBXs) requires force fields that capture mechanical deformation, thermal response, and chemical bond scission.
• Existing classical force fields cannot describe bond breaking, and ReaxFF lacks parameterization for F23-series fluoropolymers, hindering simulations of binder degradation and explosive decomposition.

Purpose of the Study:

• To develop a reactive machine learning potential (MLP) for F2311, F2313, and F2314 fluoropolymer binders used in PBXs.
• To achieve *ab initio* accuracy at a reduced computational cost for simulating PBX initiation.

Main Methods:

• Developed a reactive MLP using the deep potential framework with iterative active learning, trained on density functional theory (DFT) datasets.
• Validated the MLP against experimental data and *ab initio* molecular dynamics for physical properties (RDFs, density, Tg, elastic constants).
• Assessed the MLP's accuracy in reproducing DFT atomic forces and bond dissociation energies during bond scission under uniaxial tension.

Main Results:

• The MLP accurately predicts physical properties like radial distribution functions, equilibrium densities, glass transition temperature (Tg), and elastic constants.
• Microscopic analysis revealed a correlation between dihedral activation energy and Tg, offering insight into thermal-mechanical behavior.
• The MLP closely reproduces DFT atomic forces during bond scission, with bond dissociation energies deviating minimally (0.14-0.15 eV) from DFT values.

Conclusions:

• The developed reactive MLP provides an accurate and computationally efficient tool for simulating F23-series fluoropolymer binders in PBXs.
• This work bridges a critical gap in reactive force field coverage for these binders, enabling future atomistic studies of PBX initiation.
• The validated framework spans both physical and reactive property regimes, crucial for understanding explosive performance and safety.