Raman spectral unmixing of poplar wood through high-dimensional asynchronous representation and two-dimensional correlation spectroscopy.

Accurate characterization of lignocellulosic components is often hampered by strong spectral overlap among cellulose, hemicellulose, and lignin in Raman spectra, particularly obscuring hemicellulose signals. Here, a systematic spectral unmixing framework combining bilinear data construction, asynchronous two-dimensional correlation spectroscopy (2D-COS), and high-dimensional asynchronous representation (nD-Asyn), guided by the systematic absence of cross peaks (SACP) was presented. The method was applied to 100 poplar wood samples differing in chemical composition. Component-associated Raman profiles were recovered in the 1000-1750 cm⁻1 region, which contains major carbohydrate skeletal and lignin aromatic vibrations, and the 2500-3600 cm⁻1 region, which is dominated by C-H and O-H stretching vibrations. Validation against chemically isolated reference components using hash algorithms, cosine similarity, and structural similarity index showed close overall spectral correspondence, with hemicellulose-related features successfully resolved. This framework provides a complementary analytical framework for Raman-based analysis of complex lignocellulosic systems, enhancing interpretability where conventional multivariate or curve resolution methods are limited. The proposed framework uses compositional variation as perturbation information and provides a complementary approach for resolving strongly overlapping vibrational signals and improving component-level interpretation of complex lignocellulosic systems.