CDS&E: Thin Film Analysis by XPS: Quantitative Modeling of Sputtering and Depth Profile Data, Machine Learning Classifiers, and Novel Applications

点击翻译按钮获取中文摘要

With support from the Chemical Measurement and Imaging program of the Chemistry Division, Lev Gelb and Amy Walker of the University of Texas at Dallas will improve methods for interpreting x-ray photoelectron spectroscopy (XPS) sputter depth profiling data sets and spectra. In XPS, x-rays eject electrons from a sample, which identify the elements present near the sample surface. In depth-profiling, the sample is simultaneously eroded away by blasting (“sputtering”) the sample with a beam of ions, so that the composition at varying depths is also determined. Unfortunately, the x-rays and sputter beam cause unwanted chemical reactions, roughening, interlayer mixing, and other effects which distort the measured composition profiles. By accounting for these effects, this Project improves the quality and reliability of such measurements. Software developed in this project will be freely distributed and promoted as a community resource. XPS is widely used in materials science, nanoscience, semiconductor research, biotechnology and other fields, so improving the performance of this technique will be of significant long-term benefit to society at large.The Project will focus on model-based data analysis. A realistic simulation of the sputter process is used to describe how the sample changes during the experiment, from which XPS spectra are calculated and compared with the collected data. The simulation parameters are then adjusted to give optimal agreement and thus the best estimates of sample properties and sputter rates. These simulations also be leveraged to develop machine-learning data analysis tools. Real XPS data are time-consuming to measure, so assembling a training set of thousands (or more) of such experiments is not practical. Instead, simulations will create training sets of millions of spectra from hypothetical samples. Deep neural network classifiers will then be trained to provide very rapid assignments of sample structure and composition. Finally, these techniques will be used in the analysis of a series of complex samples of both technological and historical interest.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.