Impurity Detection and Quantification in Polymer Films by Two-Dimensional Infrared Spectroscopic Imaging and Multivariate Analysis

Abstract

Impurities in polymer films dramatically influence activity and stability, highlighting the necessity of detecting their presence accurately and rapidly. Herein, we report multivariate analytical approaches based on transflection infrared spectroscopic imaging as a proof of concept for the detection and quantification of impurities. The developed methodology enables impurity detection, quantitative estimation of concentration, and spatial mapping within polymer films. When the IR spectra of the impurities were known, classical least-squares (CLS) analysis and supervised machine learning models predicted impurity concentrations with mean absolute errors below 0.1% and 0.8%, respectively, in controlled synthetic benchmarking data sets. When impurities or their spectra were unknown, a principal component analysis (PCA)-based unsupervised approach identified anomalous spectra associated with impurities with a detection accuracy exceeding 90% under synthetic test conditions. The workflow was further evaluated on experimentally collected data sets, demonstrating practical applicability for rapid, high-throughput detection with minimal preprocessing and compatibility with automated sample handling. We finally identify opportunities for further improvement, particularly in mitigating the influence of atmospheric water vapor and carbon dioxide fluctuations during IR measurements.