A Machine Learning Model for Predicting Electron-Molecule Ionization Cross Sections

Electron-molecule collision cross sections play a pivotal role in many areas of applied physics, including plasma physics. Unfortunately, detailed measurements and state-of-the-art theoretical calculations are often too difficult to perform for highly complex molecules or for a wide variety of molecular targets. As a result, the data sets needed for modeling applications are often unavailable or incomplete. Machine learning algorithms may be able to help fill the gap in available cross section data and provide reasonable estimates for molecular targets that are beyond the reach of theoretical models and experimental measurements. We present a feed-forward neural network trained on existing experimental data and show that it provides reasonable estimates of electron-molecule collision cross sections for molecular targets beyond those in its training set. Our model is benchmarked by comparing its predictions with measured cross sections, and our data demonstrate that with training on as a few as 15 molecular targets, the algorithm predicts the measured cross sections to within 10% in many cases. The success of our simple model with relatively few training data sets indicates that machine learning techniques can successfully complement traditional theoretical models and experiment measurements and represent a viable method to provide much needed data for plasma modeling.