Raman Spectroscopy and Neural Network for the Analysis of Berlin Wall Colors

This research begins with the analysis of fragments from Berlin Wall street art to characterize the painting materials used. Spectroscopic results offer an overview of the painting techniques, but more importantly, they highlight a new advantage of applying Raman spectroscopy to the analysis of acrylic colors. The study demonstrates a correlation between peak intensity and compound concentration, leveraging the power of deep learning to quantify pigment mixtures in commercial acrylic products using Raman spectra collected with a handheld device (BRAVO by Bruker). By employing a Convolutional Neural Network (CNN) algorithm1 , this study showcases the ability to analyze spectra and accurately predict the ratios of coloring compounds. Calibration and training reference materials were prepared by diluting commercial acrylic paints (Schmincke brand), mimicking the methods commonly used by street artists. For the first time, Raman spectroscopy offers valuable insights into calibrations for detecting dye dilution in commercial pigment mixtures, opening new possibilities for analytical quantification using handheld Raman spectrometers. This contributes to a deeper understanding of the materials and techniques employed by street artists.