Machine-Learning Assisted Screening of Energetic Materials

Abstract

In this work, machine learning (ML), materials informatics (MI), and thermochemical data are combined to screen potential candidates of energetic materials. To directly characterize energetic performance, the heat of explosion ΔH e is used as the target property. The critical descriptors of cohesive energy, averaged over all constituent elements and the oxygen balance, are found by forward stepwise selection from a large number of possible descriptors. With them and a theoretically labeled ΔH e training data set, a satisfactory surrogate ML model is trained. The ML model is applied to large databases ICSD and PubChem to predict ΔH e . At the gross-level filtering by the ML model, 2732 molecular candidates based on carbon, hydrogen, nitrogen, and oxygen (CHNO) with high ΔH e values are predicted. Afterward, a fine-level thermochemical screening is carried out on the 2732 materials, resulting in 262 candidates with TNT equivalent power index P e(TNT) greater than 1.5. Raising P e(TNT) further to larger than 1.8, 29 potential candidates are found from the 2732 materials, all are new to the current reservoir of well-known energetic materials.

Materials-Genome Approach to Energetic Materials

“…Despite research having shown that the results of machine learning and screening can be comparable to theoretical calculations with sufficient data, the problem is that the existing molecular screening procedures cannot yet accurately identify all interest features due to the diversity of screening conditions and learning methods. A database containing macro molecules can effectively screen potential energetic compounds with general standards such as oxygen balance, heat of formation, heat of explosion, volume of gaseous products and explosive power etc. ,, For the design of energetic materials for specific applications, the construction of database allows the computer to identify and learn the generality of such materials effectively, thus accelerating the process of screening and design, and thereby accelerating the research and development rate of new explosives.…”

“…A database containing macro molecules can effectively screen potential energetic compounds with general standards such as oxygen balance, heat of formation, heat of explosion, volume of gaseous products and explosive power etc. 20,22,23 For the design of energetic materials for specific applications, the construction of database allows the computer to identify and learn the generality of such materials effectively, thus accelerating the process of screening and design, and thereby accelerating the research and development rate of new explosives. The second key step is to set screening criteria (like "fishing nets") for capturing target compounds from the "molecular oceans", while concomitantly identifying the key functional groups of energetic materials.…”

“…These data training strategies were proved to be effective in avoiding the above problems and screening in PubChem library with over 100 million compounds and substances. 20 Not all training methods need build a comprehensive database for providing guidance in machine learning. The learning curve obtained from plots of cross-validated test error vs the number of training examples will reduce the dependence on modeling data.…”

“…Guo and co-workers combined machine learning (ML), materials informatics (MI), and thermochemical data for high throughput screening of potential energetic molecules. 20 By training the basic parameter and analyzing the heat of explosion for classic energetic materials, the ML model effectively selected 2732 potential energetic molecules from 140 million compounds. Combining the requirements of highperformance explosives in oxygen balance, volume of gaseous products, and explosive power, 29 compounds were identified as the most promising candidates.…”

Machine Learning Enabled Quickly Predicting of Detonation Properties of N‐Containing Molecules for Discovering New Energetic Materials

“…[ 25 ] Machine learning (ML), materials informatics (MI), and thermochemical data were combined to screen molecular candidates based on high Δ H e values. [ 26 ] Property prediction and molecular screening strategies based on machine learning have high potential on discovering new high‐energy density materials. [ 27 ]…”