Machine learning prediction of a chemical reaction over 8 decades of energy

Recent progress in machine learning has sparked increased interest in utilizing this technology to predict the outcomes of chemical reactions. The ultimate aim of such endeavors is to develop a universal model that can predict products for any chemical reaction given reactants and physical conditions. In pursuit of ever more universal chemical predictors, machine learning models for atom-diatom and diatom-diatom reactions have been developed, yet no such models exist for termolecular reactions. Accordingly, we introduce neural networks trained to predict opacity functions of atom recombination reactions. Our models predict the recombination of <a:math xmlns:a="http://www.w3.org/1998/Math/MathML"> <a:msup> <a:mrow> <a:mi>Sr</a:mi> </a:mrow> <a:mo>+</a:mo> </a:msup> <a:mspace width="4pt"/> <a:mo>+</a:mo> <a:mspace width="4pt"/> <a:mrow> <a:mi>Cs</a:mi> <a:mspace width="0.16em"/> <a:mo>+</a:mo> </a:mrow> <a:mspace width="4pt"/> <a:mrow> <a:mi>Cs</a:mi> <a:mo>→</a:mo> </a:mrow> <a:mspace width="4pt"/> <a:msup> <a:mrow> <a:mi>SrCs</a:mi> </a:mrow> <a:mo>+</a:mo> </a:msup> <a:mspace width="4pt"/> <a:mo>+</a:mo> </a:math> Cs and <h:math xmlns:h="http://www.w3.org/1998/Math/MathML"> <h:msup> <h:mrow> <h:mi>Sr</h:mi> </h:mrow> <h:mo>+</h:mo> </h:msup> <h:mspace width="4pt"/> <h:mo>+</h:mo> <h:mspace width="4pt"/> <h:mrow> <h:mi>Cs</h:mi> <h:mspace width="0.16em"/> <h:mo>+</h:mo> </h:mrow> <h:mspace width="4pt"/> <h:mrow> <h:mi>Cs</h:mi> <h:mo>→</h:mo> </h:mrow> <h:mspace width="4pt"/> <h:msub> <h:mi>Cs</h:mi> <h:mn>2</h:mn> </h:msub> <h:mspace width="4pt"/> <h:mo>+</h:mo> <h:mspace width="4pt"/> <h:msup> <h:mrow> <h:mi>Sr</h:mi> </h:mrow> <h:mo>+</h:mo> </h:msup> </h:math> over multiple orders of magnitude of energy, yielding overall results with a relative error <p:math xmlns:p="http://www.w3.org/1998/Math/MathML"> <p:mrow> <p:mo>≲</p:mo> <p:mn>10</p:mn> <p:mo>%</p:mo> </p:mrow> </p:math> . Even far beyond the range of energies seen during training, our models predict the atom recombination reaction rate accurately. As a result, the machine is capable of learning the physics behind the atom recombination reaction dynamics.