Real-Time Monitoring of Copolymerization Kinetics by Infrared Spectroscopy and Partial Least Squares Regression

Abstract Fourier-transform infrared (FT-IR) spectroscopy is a nondestructive, relatively low-cost, and highly responsive analytical method to detect various functional groups of chemical species and their temporal changes in a reaction. The technical barrier to using FT-IR for monitoring copolymerization kinetics lies in the significant band overlap between comonomers, making it difficult to trace individual monomer conversions, especially during online monitoring of polymerizations, where spectral analysis needs to be accurate and fast. In this work, we address this barrier by proposing a chemometric approach to capture the relationship between IR absorption spectra and the comonomer concentrations in a copolymerization system. The developed models are then applied to accurately predict compositional change with time in otherwise unseen spectral data. The proposed workflow first collects copolymerization samples and associated data using inline FT-IR integrated automated setups, then pre-processes the IR spectra, and finally models the experimental data using partial least squares regression (PLS-R). A wavenumber and latent variable (LV) screening approach was applied to improve the robustness of PLS-R models, and external validation with unseen samples were conducted to further evaluate the performance of the trained models, in addition to the conventional model evaluation during training. The versatility of the workflow was demonstrated by its application to different copolymerization systems, comprising acrylates, methacrylates, styrene, acrylamide, and ethyl lipoate. Models with strong predictive performance were obtained (R2 > 0.98 for both calibration and 5-fold cross-validation). When evaluated on independent test data, prediction errors were within 3% for comonomers from different classes and better than 10% for comonomers within the same class, namely, for copolymerization of methyl and ethyl acrylate.