Spectral Modeling for the Deconvolution of Doppler-Broadened HR-CS GFAAS Signals for Lithium Isotope Discrimination at Trace Levels

Abstract Lithium isotope analysis by high-resolution continuum source graphite furnace atomic absorption spectrometry (HR-CS GFAAS) remains challenging due to the strong spectral overlap of Doppler-broadened transitions. In this work, a physically constrained spectral deconvolution approach is introduced to resolve the overlapping fine-structure components of 7Li and 6Li without reliance on empirical or data-driven calibration models. The method is based on a physically informed description of the absorption profiles, in which the Gaussian line width (ΔFWHM) and transition parameters are optimized to reproduce the experimental spectra. The approach was evaluated over the concentration range of 1–10 μg L–1 using isotope-specific calibration and validated using synthetic mixtures and certified clinical reference materials. Deconvolution of the D2 (22P3/2 ← 22S1/2) and D1 (22P1/2 ← 22S1/2) transitions enabled isotope-resolved quantification, yielding accurate and consistent results across different matrices, with recoveries ranging from 95.5 to 115.5% when carrying out the determination of Li by isotope dilution using either a single-spike or a double-spike approach. Limits of detection ranged from 0.1 to 0.2 μg L–1, while limits of quantification ranged from 0.4 to 0.7 μg L–1, demonstrating the high sensitivity of the proposed approach. The results also indicate that the analytical performance is primarily controlled by the relative contribution of each isotope to the total absorbance. While the dominant isotope is quantified more easily and with higher reliability, the deconvolution approach enabled the determination of the minor component even under situations where it is increasingly affected by noise propagation and under conditions of strong spectral overlap. Overall, the proposed strategy demonstrates that physically informed spectral modeling enables robust lithium isotope analysis by HR-CS GFAAS and provides a general framework for addressing overlapping spectral features in atomic spectrometry.