Processing slightly resolved ro-vibrational spectra during chemical vapor deposition of carbon materials: machine learning approach for plasma thermometry

TL;DR

This paper introduces a machine learning-based optical spectroscopic method to rapidly determine rotational and vibrational temperatures in plasma during chemical vapor deposition of carbon materials, using ro-vibrational spectra of C2 molecules.

Contribution

It presents a novel application of Random Forest machine learning to estimate local plasma temperatures from optical emission spectra in CVD processes.

Findings

Accurate temperature maps for $T_{rot}$ and $T_{vib}$ were generated.

The method provides detailed electron temperature and emission intensity distribution maps.

The approach enables fast, non-invasive plasma thermometry during material synthesis.

Abstract

A fast optical spectroscopic method for determination rotational ($T_{rot}$) and vibrational ($T_{vib}$) temperatures in two-temperature Boltzmann distribution of the excited state by using machine learning approach is presented. The method is applied to estimate molecular gas temperatures in a direct current glow discharge in hydrogen-methane gas mixture during plasma-enhanced chemical vapor deposition of carbon film materials. Slightly resolved ro-vibrational optical emission spectrum of the $C_2$ ($\nu'=0 \to \nu''=0$) Swan band system was used for local temperature measurements in plasma ball. Random Forest algorithm of machine learning was explored for determination of temperature distribution maps. In addition to the $T_{rot}$ , $T_{vib}$ maps, distribution maps and their gradients for electron temperature ($T_e$) and for the emission intensity of the spectral line 516,5$nm$ corresponding to $C_2$ species is presented and is discussed in detail.