High-Throughput Voigt Deconvolution of Luminescence Spectral Maps with Curvature-Informed Peak Initialization

Abstract

Accurate deconvolution of overlapping emission bands remains a key challenge in luminescence spectroscopy, particularly for large-scale data sets generated by cathodoluminescence (CL) and photoluminescence (PL) mapping. This work presents a scalable Python-based workflow for high-throughput Voigt deconvolution of CL spectra. The approach integrates curvature-informed peak initialization with nonlinear least-squares optimization to improve fit stability and consistency across large data sets and extract quantitative spectral parameters, including emission energy, amplitude, and Gaussian and Lorentzian broadening contributions. The resulting parameters enable direct statistical analysis and spatial mapping across data sets. Synthetic validation and full-map uncertainty analysis show that peak center and total line width are more robustly recovered than the individual Gaussian and Lorentzian contributions when spectral features strongly overlap. Application to CdMgSeTe thin films demonstrates spatial variation in emission behavior and distinguishes differences in spectral heterogeneity between the analyzed films. This workflow enables high-throughput conversion of hyperspectral CL maps into spatially resolved spectral fingerprints with peak-level detail suitable for quantitative comparison across samples and processing conditions.