Accession Number:

AD1050897

Title:

Machine Learning Intermolecular Potentials for 1,3,5-Triamino-2,4,6-trinitrobenzene (TATB) Using Symmetry-Adapted Perturbation Theory

Descriptive Note:

[Technical Report, Technical Report]

Corporate Author:

US Army Research Laboratory

Personal Author(s):

• Taylor, DeCarlos E.

Report Date:

2018-04-25

Pagination or Media Count:

22

Abstract:

In this report, intermolecular potentials for 1,3,5-triamino-2,4,6-trinitrobenzene TATB are developed using machine learning techniques. Three potentials based on support vector regression, kernel ridge regression, and a neural network are fit using symmetry-adapted perturbation theory. The potentials are used to explore minima on the TATB dimer potential energy surface. It is demonstrated that the ab initio potential energy surface is accurately characterized by the machine learning potentials and that machine learning methods can accurately describe noncovalent interactions in energetic materials.

Descriptors:

• computational science
• supervised machine learning
• artificial neural networks
• machine learning
• density functional theory
• perturbation theory
• energetic materials
• insensitive explosives
• potential energy
• quantum theory
• artificial intelligence software
• test sets
• artificial intelligence
• materials
• perturbations
• computational chemistry
• computer languages
• energy
• learning
• symmetry

Subject Categories:

• Computer Programming and Software
• Ammunition and Explosives

Distribution Statement:

[A, Approved For Public Release]