Evaluation of optical constants in oxide thin films using machine learning

TL;DR

This paper presents a neural network-based inverse analysis method for determining optical constants in oxide thin films from UV-visible spectroscopy, offering an automated alternative to traditional techniques.

Contribution

It introduces a neural network approach that automates optical constant evaluation from spectroscopic data, improving efficiency over conventional methods.

Findings

Neural network accurately predicts optical constants from spectroscopy data.

The method reduces reliance on expert judgment and manual analysis.

Demonstrates effectiveness on oxide thin film samples.

Abstract

This paper describes an inverse analysis method using neural networks on optical spectroscopy, and its application to the quantitative optical constant evaluation. The present method consists of three subprocesses. First, measurable UV-visible spectroscopic quantities were calculated as functions of the optical constants of the solid based on the Tomlin equations [J. Phys. D 1 1667 (1968)] by carefully eliminating the unpractical combinations of optical constants. Second, the back-propagation neural network is trained using the calculated relationships between the measurable quantities and the optical constants. Finally, the trained network is utilized to determine the optical constants from measured responses. The conventional (Newton-Raphson) method tends to require the judgement of a well-experienced analyst, while machine learning shows automatically human-free performance in data conversion.