Knowledge-Guided Construction of Machine Learning Potentials for Reactive Behavior in Energetic Materials
Abstract
Energetic materials involve multiple interactions, making reliable machine-learning potentials (MLPs) difficult to construct without extensive datasets. Here, we enhance the ability of the neural network interatomic potential (NNAP) to describe molecular systems using sphmulti, a refined spherical-harmonic basis set that combines extra radial basis with element-specific chemical weighting in each radial channel. We further integrate hyperactive learning, farthest-point sampling, and iterative sufficiency criteria with knowledge-guided key events, enabling the NNAP to retain essential information for describing the material system. For 3,3'-diamino-4,4'-azoxyfurazan, the resulting NNAP is trained on only 13,680 configurations yet captures key properties, including the isomerization barrier, pyrolysis activation energy, and shock Hugoniot behavior. NNAP-driven molecular dynamics simulations further reveal that pyrolysis proceeds through furazan-ring opening followed by chain reactions, whereas shock loading induces anisotropic hydrogen transfer driven by interlayer structural deformations. This work enables efficient and reliable MLPs for energetic materials.
Related
- Autonomous Analysis of Thermoelectric Materials Big Data by Multi-Agentic AI and Construction of a Knowledge Base and Specialized AI
- Automated Materials Data and Knowledge Extraction using Machine Learning and Large Language Models
- Load balancing for adaptive-precision interatomic potentials in materials science
- Compiling Chemical Knowledge into Executable Descriptors for Materials Prediction
- Agentic schema-guided extraction of materials process knowledge from scientific literature
BibTeX
@misc{dong2026knowledgeguided,
title = {{Knowledge-Guided Construction of Machine Learning Potentials for Reactive Behavior in Energetic Materials}},
author = {Guanchen Dong and Huanrong Liu and Lin Wang and Yuxuan Chen and Zhengyu Tu and Qing-an Li and Bin Xu and Jian Li and Shan Zhang and Xiaotong Li and Linghua Tan and Guangcheng Yang and Jun Kang and Rui Su and Pengfei Guan},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.17987},
eprint = {2610.17987},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.17987},
}
RIS
TY - UNPB
TI - Knowledge-Guided Construction of Machine Learning Potentials for Reactive Behavior in Energetic Materials
AU - Guanchen Dong
AU - Huanrong Liu
AU - Lin Wang
AU - Yuxuan Chen
AU - Zhengyu Tu
AU - Qing-an Li
AU - Bin Xu
AU - Jian Li
AU - Shan Zhang
AU - Xiaotong Li
AU - Linghua Tan
AU - Guangcheng Yang
AU - Jun Kang
AU - Rui Su
AU - Pengfei Guan
PY - 2026
DA - 2026/10/15
PB - arXiv
DO - 10.48550/arxiv.2610.17987
UR - https://arxiv.org/abs/2610.17987
AB - Energetic materials involve multiple interactions, making reliable machine-learning potentials (MLPs) difficult to construct without extensive datasets. Here, we enhance the ability of the neural network interatomic potential (NNAP) to describe molecular systems using sphmulti, a refined spherical-harmonic basis set that combines extra radial basis with element-specific chemical weighting in each radial channel. We further integrate hyperactive learning, farthest-point sampling, and iterative sufficiency criteria with knowledge-guided key events, enabling the NNAP to retain essential information for describing the material system. For 3,3'-diamino-4,4'-azoxyfurazan, the resulting NNAP is trained on only 13,680 configurations yet captures key properties, including the isomerization barrier, pyrolysis activation energy, and shock Hugoniot behavior. NNAP-driven molecular dynamics simulations further reveal that pyrolysis proceeds through furazan-ring opening followed by chain reactions, whereas shock loading induces anisotropic hydrogen transfer driven by interlayer structural deformations. This work enables efficient and reliable MLPs for energetic materials.
ER -