IKEBANA: Data-Driven Neural-Network Predictor of Electron-Impact K-Shell Ionization Cross Sections

Abstract

A fully connected neural network was trained to model the K-shell ionization cross sections based on two input features: the atomic number and the incoming electron overvoltage. The training utilized a recent, updated compilation of experimental data covering elements from H to U, and incident electron energies ranging from the threshold to relativistic values. The neural network demonstrated excellent predictive performance, compared with the experimental data, when available, and with full theoretical predictions. The developed model is provided in the ikebana code, which is openly available and requires only the user-selected atomic number and electron energy range as inputs.