Enhanced Copolymer Characterization for Polyethers Using Gel Permeation Chromatography Combined with Artificial Neural Networks.

Gel permeation chromatography (GPC) is a generally applied method for the mass analysis of various polymers and copolymers, but it inherently fails to provide additional important information such as the composition of copolymers. However, we will show that GPC measurements using different solvents can yield not just the correct molecular weight but the composition of the copolymer. Accordingly, artificial neural networks (ANNs) have been developed to process the data of GPC measurements and determine the molecular weight and the chemical composition of the copolymers. The target values of the ANNs were obtained by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) and nuclear magnetic resonance (NMR) spectroscopy. Our GPC-ANN method is demonstrated by the analysis of various poloxamers, i.e., poly(ethylene oxide) (PEO)-poly(propylene oxide) (PPO) block copolymers. Two ANNs were constructed. The first one (_1) works in a wider mass range (from 900 to 12,500 dalton), while the second one (_2) produces more output values. _2 can thus predict seven characteristic copolymer parameters, namely, two average molecular weights, the average weight fraction of the EO unit, and four average numbers of the repeat units. The correlation between the experimentally obtained outputs and the predicted ones is high ( > 0.98). The accuracy of the ANNs is very convincing, and both ANNs predict the number-average molecular weight () with an accuracy below 5%. Furthermore, this work is the first step for creating an open database and applications extending the use of the GPC-ANN method for the analysis of copolymers.