Multitask Machine Learning to Predict Polymer-Solvent Miscibility Using Flory-Huggins Interaction Parameters

Predicting and understandingthe phase equilibria or phase separationin polymer-solvent solutions represent unresolved fundamentalproblems in polymer science. The phase behavior and thermodynamicsof polymer miscibility depend on the inter- and intramolecular interactionsof a polymer with a certain molecular weight distribution mixed witha solvent. Here, we develop a machine-learning framework to achievehighly generalized and robust prediction of Flory-Huggins & chi;parameters for polymer-solvent solutions. The model was trainedusing experimentally observed temperature-dependent & chi; parametersfor 1190 samples, comprising 46 unique polymers and 140 solvent species.However, the difficulty was that the data set was quantitatively limitedand qualitatively biased owing to technical issues in determiningthe Flory-Huggins & chi; parameters. To overcome these limitations,we produced an in-house data set of & chi; parameters obtained fromquantum chemical calculations for thousands of polymer-solventpairs and a large list of soluble and insoluble polymer-solventpairs. Using these three data sets, we conducted multitask machinelearning that simultaneously performed the "soluble/insoluble"classification and quantitative evaluation of both experimental andcalculated & chi; parameters. Consequently, we obtained a highlygeneralized model applicable to a wide range of polymer solution spaces.In this paper, the predictive power and physicochemical implicationsof the model are demonstrated, along with quantitative comparisonswith existing methods.