Construction of a multi-wavelength Raman spectral database for 39 hazardous organic peroxides and azo compounds: experimental optimization and DFT validation

Rapid and nondestructive identification of hazardous organic initiators is important for forensic science, industrial safety, and emergency response. In this study, a confocal micro-Raman spectral database was constructed for 39 high-risk compounds, with emphasis on organic peroxides and azo compounds, and the analytical performance of three excitation wavelengths (473, 532, and 785 nm) was systematically compared. Across the representative compounds examined, 532 nm excitation provided the most reliable compromise among signal-to-noise ratio, fluorescence background, and spectral resolution. By contrast, 785 nm excitation reduced autofluorescence in selected samples but generated a reproducible instrumental band near 1777–1780 cm−1, which could be misassigned as a carbonyl vibration without rigorous background control. Density functional theory (DFT) calculations at the B3LYP/6-31G(d,p) level with Grimme D3 dispersion correction were used to support vibrational assignment. A region-dependent frequency-scaling strategy (0.9830 below 1800 cm−1 and 0.9613 above 2000 cm−1) improved agreement between simulated and experimental spectra. The simulated spectra matched the optimized 532 nm data and supported the assignment of diagnostic bands, including C≡N stretching, C = O stretching, and peroxide-associated modes. Homogeneity and repeatability tests confirmed stable peak positions and robust peak intensities, with representative intensity RSD values below 3.0%. The resulting database and validated workflow provide a practical foundation for forensic interpretation, rapid hazard screening, and future AI-assisted Raman identification.