Temperature-dependent Raman spectra of 2H-MoS2 from Machine Learning-driven statistical sampling

TL;DR

This study uses machine learning-driven statistical sampling to accurately model the temperature-dependent Raman spectra of 2H-MoS2, aligning well with experimental data and enhancing theoretical understanding.

Contribution

It introduces a computational framework that incorporates disorder, doping, and temperature effects to reliably predict Raman spectra of Molybdenum sulfides.

Findings

Calculated Raman frequencies and linewidths match experimental temperature trends.

The method accounts for anharmonic and thermal broadening effects.

Framework can be extended to amorphous Molybdenum sulfides.

Abstract

Molybdenum sulfides are in the spotlight of materials science thanks to their interesting properties for applications in optoelectronics, nanocomposites, lubricants, and catalysis. The structural characterization of Molybdenum sulfides is a crucial step to understand and tune their properties. Vibrational techniques, such as infrared and Raman spectroscopy, can directly link to structural features, but the experimental literature suffers from large variability. Theoretical calculations are a powerful tool complementing and explaining empirical measurements. The reliability of first-principles calculation depends on the level of approximation made, taking into account disorder, doping, or temperature to yield a good description of the phonon statistics and related measurable quantities, such as the infrared and Raman peaks. In this study we calculate the Raman spectrum of crystalline 2H-MoS2, including broadening and shifts due to thermal and anharmonic effects. Our results demonstrate excellent agreement with experimental measurements; notably, the calculated temperature trends in frequencies and linewidths align with empirical observations. These findings establish a robust computational framework, paving the way for similar studies on amorphous Molybdenum sulfides.