Infrared Spectroscopy of Chemically Diverse Carbon Clusters: A Data-Driven Approach

Carbon clusters exhibit a broad diversity of topologies and shapes, encompassing fullerene-like cages, graphene-like flakes, more disordered pretzel-like branched structures. Here, we examine computationally their infrared spectra in relation with these structures from statistical perspective. Individual for samples isomers were determined by means the self-consistent charge density functional-based tight-binding method, an interpolation scheme is designed to reproduce spectral features regression on much smaller subset sample. This proceeds encoding using appropriate descriptors selecting them through principal component analysis, Gaussian or inverse distance weighting providing nonlinear functions. Metric learning employed reduce global error preselected testing set. The interpolated satisfactorily specific dependence size shape, enabling quantitative prediction away Finally, classification within four proposed families critically discussed analysis sample based iterative label spreading.