Machine-Learning Assisted Screening of Energetic Materials.

Area of Science:

• Computational Chemistry
• Materials Science
• Chemical Engineering

Background:

• Energetic materials are crucial for various applications.
• Discovering new energetic materials with enhanced performance is an ongoing challenge.
• Traditional screening methods are often time-consuming and resource-intensive.

Purpose of the Study:

• To develop and apply a machine learning (ML) model for efficient screening of potential energetic materials.
• To identify novel energetic material candidates with high heat of explosion (ΔH) and power index (P).

Main Methods:

• Integration of machine learning (ML), materials informatics (MI), and thermochemical data.
• Utilizing cohesive energy and oxygen balance as critical descriptors.
• Training a surrogate ML model on a theoretically labeled ΔH dataset.
• Screening large databases (ICSD, PubChem) using the trained ML model.

Main Results:

• A ML model was trained to predict the heat of explosion (ΔH).
• Initial screening identified 2732 CHNO-based molecular candidates.
• Further thermochemical screening yielded 262 candidates with P > 1.5.
• 29 novel energetic material candidates with P > 1.8 were discovered.

Conclusions:

• The ML-driven approach effectively screens for high-performance energetic materials.
• The identified 29 candidates represent new additions to the known energetic materials.
• This methodology accelerates the discovery of advanced energetic materials.