Development of Comparative and Machine Learning–Based Methodologies for the Identification of Inks Applicable in the Field of Cultural Heritage and Forensic Science

This study proposes the development of comparative and machine learning-based methodologies for the identification of inks and pigments, with potential applications in both cultural heritage diagnostics and forensic science. A preliminary selection of black inks from various pen brands was analyzed using Raman spectroscopy to define a framework for spectral comparison based on peak shifts and area ratios derived from curve fitting. The proposed method introduces a system based on spectral compatibility allowing the classification of inks based on their compositional similarity. In parallel, an automated analysis code was developed to enhance scalability and reproducibility. This system performs baseline removal, peak normalization, first-stage filtering of incompatible spectra, and refined deconvolution through pseudo-Voigt fitting, generating a numerical similarity score for each comparison. Results demonstrate that the approach allows quantitative estimation of ink compatibility and could be extended to broader datasets through the implementation of a spectral database. • Novel Raman-based comparative methodology for identifying and classifying modern black inks. • peak-position analysis, normalized area-ratio metrics, and quantitative compatibility scoring. • Combination of deconvolution-derived observables with an automated, multi-stage spectral processing algorithm. • Nondestructive discrimination of inks sharing a common chromophore (e.g., crystal violet). • Strong applicability in forensic document examination and cultural heritage studies.