A machine learning approach for the automated classification of bulk sp2 and sp3 carbon materials

Abstract Principal component analysis (PCA) and linear discriminant analysis (LDA) were used to classify different types of carbon material based on their Raman spectra. The selected reference materials were highly oriented pyrolytic graphite (HOPG), diamond‐like carbon (DLC), glassy carbon (GC), hydrogenated graphite‐like carbon (GLCH), and hydrogenated polymer‐like carbon (PLCH). These materials vary in crystallinity, predominant carbon hybridization, and hydrogen content. The training dataset was Raman spectra collected from commercial samples (HOPG, DLC, GC) and samples synthesized in our laboratory (GLCH, PLCH). The Raman spectra were collected using 532 nm laser excitation. The classification model revealed that the first principal component (PC1) was the determinant source of information to separate the crystalline from the amorphous carbon samples. PC2 allowed the separation of amorphous material with different levels of hybridization (sp 2 and sp 3 ). Finally, both PC2 and PC3 contributed to separate materials with different levels of hydrogenation. The classification model was tested using a library of Raman spectra of carbon materials reported in the literature, and the results showed a high accuracy prediction (97%). The model presented here provides an avenue for automated classification of carbon materials using Raman spectroscopy and machine learning.