Machine learning for iron oxide identification from oxygen K edge in EELS spectra

Abstract

In this work we test machine learning tools such as the Support Vector Machine algorithm and neural network models on the task of Electron Energy-Loss Spectroscopy (EELS) spectra classification. Given many sample spectra of EELS applied on wüstite and magnetite nanocubes, we train both models to determine the oxidation state of iron. We show that SMV exhibits a good performance on classifying clean data, and we demonstrate the capability of neural networks of producing robust results given shifted data