Raman analysis of black carbon using artificial neural networks for emission source classification

Black Carbon (BC), a major component of Particulate Matter (PM), plays a crucial role in air pollution, climate change, and public health impacts, due to its light-absorbing nature and ability to penetrate deeply into the respiratory system. Despite being the subject of scientific research for many years, a significant gap remains in BC source apportionment, consisting in the identification of the specific combustion sources. However, the identification and relative quantification of BC contributions by source type (e.g., domestic heating, biomass combustion, fossil fuels) provide critical information for the development and implementation of effective pollution mitigation strategies. In this paper, we demonstrate the potential of Raman spectroscopy for the source apportionment of BC samples originating from gasoline, diesel, and biomass combustion. Recorded Raman spectra were used to train a multilayer perceptron (MLP) classifier, which achieved an identification accuracy of 96.9% on testing data, confirming the method's reliability and paving the way for the development of a routine source apportionment technique based on Raman spectroscopy. Finally, the classification was performed on BC collected in two distinct sampling stations showing different contribution from fossil fuel (diesel and gasoline) and biomass burning, consistent with their respective emission contexts.