Rapid ellipsometric determination and mapping of alloy stoichiometry with a neural network

Abstract

Due to their tunable physical and chemical properties, alloys are of fundamental importance in material science. The determination of stoichiometry is crucial for alloy engineering. Classical characterization tools such as energy-dispersive x-ray spectroscopy (EDX) are time consuming and cannot be performed in an ambient atmosphere. In this context, we introduce a new methodology to determine the stoichiometry of alloys from ellipsometric measurements. This approach, based on the analysis of ellipsometric spectra by an artificial neural network (ANN), is applied to electrum alloys. We demonstrate that the accuracy of this approach is of the same order of magnitude as that of EDX. In addition, the ANN analysis is sufficiently robust that it can be used to characterize rough alloys. Finally, we demonstrate that the exploitation of ellipsometric maps with the ANN is a powerful tool to determine composition gradients in alloys.

Monitoring MBE Substrate Deoxidation via RHEED Image-Sequence Analysis by Deep Learning

“…Methods from artificial intelligence, including deep learning, are being increasingly applied to nano and materials science. − Recently, first attempts have been reported to use statistical methods and machine learning for RHEED image interpretation. − Inspired by these pioneering works, we propose to use a deep-learning (DL) approach for classification of oxidized and deoxidized substrates via their RHEED patterns, to resolve the problems described above. As mentioned above, due to the sample rotation, the RHEED signal can confidently indicate deoxidation only during short moments, when the electron beam is aligned with a lattice direction of the crystal.…”