Machine Learning Models for High Explosive Crystal Density and Performance.

Area of Science:

• Materials Science
• Computational Chemistry
• Machine Learning

Background:

• The discovery of new explosives with enhanced energy density and performance has stagnated.
• Machine learning offers a pathway to accelerate the identification of novel energetic molecules through rapid property prediction.
• A large database of synthesized energetic molecules is crucial for training accurate predictive models.

Purpose of the Study:

• To develop accurate machine learning models for predicting the density, detonation velocity, and detonation pressure of energetic materials.
• To accelerate the discovery of new explosives with superior performance characteristics.
• To establish a foundation for AI-driven screening of a vast number of potential energetic compounds.

Main Methods:

• Assembled a database of 21,000 experimentally synthesized energetic molecules.
• Calculated detonation velocities and pressures using electronic structure and atomistic simulations.
• Trained machine learning models using experimental density and calculated performance metrics.
• Introduced a novel molecular descriptor, MolDensity, and analyzed descriptor importance.

Main Results:

• Developed machine learning models capable of predicting crystal density, detonation velocity, and detonation pressure.
• Achieved a 20% reduction in root-mean square error for crystal density prediction compared to previous state-of-the-art models.
• Demonstrated the effectiveness of chiral-specified SMILES strings and the MolDensity descriptor in improving model accuracy.
• Identified key descriptors that influence material density and performance.

Conclusions:

• The developed machine learning models provide inexpensive and highly accurate predictions for key explosive properties.
• These models can significantly accelerate the screening of large chemical spaces (>10^6 compounds) for high-performing energetic materials.
• This work lays the groundwork for future AI-driven discovery of advanced energetic materials.