Recent developments in the provision of atomic and molecular data for plasma modelling

Theory has become increasingly important in providing information on key atomic and molecular processes for plasma modelling. First principles electron-molecule scattering calculations can be used to provide complete sets of cross-sections for a given gas including for processes which involve reactive species such as radicals for which there is almost no measured data. Recent developments in our electronic collision calculations include (a) cross sections for collisions from electronically excited states; (b) a general treatment of electron impact vibrational excitation; (c) modelling breakup patterns following electron impact dissociation using ab initio molecular dynamics and, most recently, (d) the use of effective core potentials to allow electron collision calculations from molecules containing heavy atoms. 

Given the large quantities of data required for plasma models, machine learning also has an important role to play in completing modelling data sets. We have explored machine learning of heavy particle (chemical) reactions and break-up patterns following electron collisional ionization. The latter project used libraries of mass spectrocopy data to train a machine learning model to predict fragmentation patterns for a single energy (typically 70 eV). Combining these patterns with total ionization cross sections computed using the BEB (Binary Encounter Bethe) model and computed appearance thresholds for the different ionic fragments, enables us to predict not only the ionization cross section but also the distribution of species that results as a function of energy.