Preventing overfitting in infrared ellipsometry using temperature dependence: fused silica as a case study

TL;DR

This paper introduces a method to prevent overfitting in infrared ellipsometry by leveraging temperature dependence, demonstrated on fused silica, leading to more physically accurate models of its optical properties.

Contribution

The study presents a novel approach using temperature-dependent trends to ensure physical validity in oscillator model fitting for spectroscopic ellipsometry data.

Findings

Temperature-dependent oscillator parameters follow expected monotonic trends.

Generated accurate temperature-dependent refractive index dataset for fused silica.

Method improves physical consistency in optical property modeling.

Abstract

Fitting oscillator models to variable-angle spectroscopic ellipsometry (VASE) data can lead to non-unique, unphysical results. We demonstrate using temperature-dependent trends to prevent overfitting and ensure model physicality. As a case study, we performed mid-infrared VASE measurements on fused silica (SiO2) of various grades, from room temperature to 600 {\deg}C. We fitted oscillator models independently at each temperature, and confirmed the model's physical validity by observing the expected monotonic trends in vibrational oscillator parameters. Using this technique, we generated a highly accurate dataset for the temperature-dependent complex refractive index of fused silica for modeling mid-infrared optical components such as thermal emitters.