Neural Network-Driven Exploration in Polymerization-Induced Self-Assembly: From 2D to 3D Pseudo-Phase Diagram

Abstract

We explore the application of machine learning (ML) to predict the morphology of poly(glycerol monomethacrylate)-poly(2-hydroxypropyl methacrylate) (PGMA–PHPMA) diblock copolymer nano-objects prepared via polymerization-induced self-assembly (PISA) in aqueous media. Traditional studies typically map copolymer morphology using two-dimensional (2D) pseudo-phase diagrams, plotting variables such as the mean degree of polymerization (Xn) of the solvophobic block against the copolymer concentration (also known as the solids content). In contrast, our approach utilizes deep neural networks (DNNs) trained on literature data to generate detailed three-dimensional (3D) morphology maps. These maps include the molecular weight of the solvophilic block, providing a comprehensive volumetric view that reveals more complex relationships and transitional morphologies that are difficult to capture in 2D representations. This 3D modeling approach enriches our understanding by highlighting critical thresholds and nuanced transitions within the morphology landscape. Such advanced modeling not only deepens our understanding of how PGMA molecular weight influences copolymer morphology but also significantly reduces the need for extensive experimental trials. Consequently, it simplifies the creation of accurate pseudo-phase diagrams across a broad range of aqueous PISA formulations. Experimental validation confirms the accuracy of our models, demonstrating the potential of ML to make predictive modeling more accessible to chemists and paving the way for future research on other PISA formulations. The data set, along with all codes for model training and evaluation, is publicly accessible via both Zenodo and GitHub platforms.