Rapid discrimination of porous bio-carbon derived from nitrogen rich biomass using Raman spectroscopy and artificial intelligence methods

Granular porous bio-carbons were prepared by pyrolysis of different biomass precursors such as mung bean, black urad bean, and black grape seed, using ZnCl2 as an activating agent at different activation temperatures of 450–750 °C. The derived bio-carbons samples were extensively characterized using electron microscopy, surface area analysis, and CO2 adsorption capacity studies. As the activation temperature increased, the surface area, pore volume, nitrogen content, and the CO2 removal efficiency of the bio-carbons varied from 254 to 937 m2/g, 0.1241–0.4212 mL/g, 1.51–6.23%, and 2.11–5.48 mmol/g (at 25 °C under 3 bar pressure) respectively. Furthermore, Raman spectroscopic technique was used as a tool to understand the structural development that occurred in the biomasses during pyrolysis. Additionally, multivariate analysis such as combined Principal Component Analysis, partial least square-discriminant analysis (PLS-DA) was employed for the Raman data to discriminate the biomass based on their source and activation temperature. In addition, deep learning methods such as LeNET, ResNet, CAE were evaluated to classify the bio-carbon samples with respect to temperature and precursor material. All the models gave 100% accuracy of classification with respect to the temperature of activation. An overall classification accuracy of >92 ± 0.0665% was obtained for LeNET model. [Display omitted] •Facile one pot synthesis of porous bio-carbon from different biomasses by pyrolysis at different temperature.•Correlation of the Physiochemical properties of bio-carbon with Raman spectroscopic signatures with experimental conditions.•Raman spectroscopy with deep learning techniques for rapid classification of bio-carbons from different biomasses.