Determining electron-biomolecule cross sections using data-driven swarm analysis

Accurate modeling of electron transport through biological media requires the attainment of complete and accurate sets of cross sections for all electron interactions with all relevant biomolecules in the soft-condensed phase. Swarm experiments provide a unique way to assess the self-consistency of such cross section sets. However, when swarm transport measurements are limited in number, the inverse problem of unfolding their underlying cross sections becomes ill-posed. To account for the uncertainty inherent to this "inverse swarm problem", we employ a neural network model that is trained upon sets of cross sections from the LXCat project alongside corresponding transport coefficients found by the numerical solution of Boltzmann's equation. We apply this machine learning approach to measurements from the pulsed-Townsend swarm experiments of de Urquijo and co-workers and subsequently derive plausible neutral dissociation and dissociative electron attachment (DEA) cross sections for the biomolecule analogues tetrahydrofuran (THF) and α-tetrahydrofurfuryl alcohol (THFA).