Prediction of the dopant activity of chemical compounds against ammonia borane with key descriptors: electronegativity and crystal structures
Key descriptors representing the desorption temperature of ammonia borane against hydrides and chlorides are investigated and applied to machine learning. The desorption temperature of ammonia borane with hydrides and chlorides is experimentally measured and a database is constructed. The database indicates a correlation between the desorption temperature and electronegativity. A support vector machine with selected descriptors from the database reveals the optimal amounts of CuCl2 and AgCl needed for decreasing the desorption temperature of ammonia borane. Prediction of the electronegativity that would achieve the desorption temperature is also revealed where the crystal structure is also found to be a key descriptor. Thus, electronegativity and crystal structures are revealed to be key descriptors that predict the desorption temperature of ammonia borane with a chemical compound. This would essentially lead towards finding the optimal amount of dopants and ideal dopants with a minimum number of experiments.