Adaptive design of an X-ray magnetic circular dichroism spectroscopy experiment with Gaussian process modelling

Tetsuro Ueno,Yasuo Takeichi,Kanta Ono,Hideitsu Hino,Ai Hashimoto,Masahiro Sawada

doi:10.1038/s41524-017-0057-4

Abstract

Spectroscopy is a widely used experimental technique, and enhancing its efficiency can have a strong impact on materials research. We propose an adaptive design for spectroscopy experiments that uses a machine learning technique to improve efficiency. We examined X-ray magnetic circular dichroism (XMCD) spectroscopy for the applicability of a machine learning technique to spectroscopy. An XMCD spectrum was predicted by Gaussian process modelling with learning of an experimental spectrum using a limited number of observed data points. Adaptive sampling of data points with maximum variance of the predicted spectrum successfully reduced the total data points for the evaluation of magnetic moments while providing the required accuracy. The present method reduces the time and cost for XMCD spectroscopy and has potential applicability to various spectroscopies.

Highlights

Spectroscopy (X-ray, optical, infra-red, electron, etc.) is a popular and important experimental technique for materials analyses and investigations on the fundamental properties of materials.[1,2,3,4,5,6] Large amounts of samples and experimental data need to be measured and treated for materials research and development
We propose an adaptive design for X-ray magnetic circular dichroism (XMCD) spectroscopy experiments that uses a machine learning technique
Magnetic moments are evaluated from the predicted XMCD and XAS spectra

Summary

Introduction

Spectroscopy (X-ray, optical, infra-red, electron, etc.) is a popular and important experimental technique for materials analyses and investigations on the fundamental properties of materials.[1,2,3,4,5,6] Large amounts of samples and experimental data need to be measured and treated for materials research and development. There is a strong demand for high-throughput measurement to reduce the time and cost of spectroscopy experiments. In a conventional spectroscopy experiment, a large amount of data points is usually measured with sufficient measurement time to obtain a spectrum with an adequate signal-to-noise ratio. There are single-shot spectroscopy experiments like wavelength-dispersive X-ray spectroscopy,[7] many kinds of spectroscopy need point-by-point measurement with scanning energy or wavelength. One can not obtain whole spectrum until the end of the experiment in such sequential point-by-point measurement and one can obtain parameters by analysis as a post-process after the measurement. Modern experiments like scanning X-ray microspectroscopy takes relatively long measurement time per energy because it takes scanning image at each energy point.[8]. An efficient measurement by reducing energy data points based on the intelligent design of experiments is needed. Regression models are used to realise precise predictions thanks to advances in machine learning techniques, and such techniques can be applied to the intelligent design of spectroscopy experiments.[9,10]

Methods

Results

Conclusion