Despite the various potential applications of dual pH/temperature-responsive microgels, the multiple (and often interacting) physical and chemical factors that influence the volume phase transition temperature (VPTT) in such microgels make it challenging to directly design a microgel with a particular targeted swelling response. Herein, we address this challenge by designing and implementing a data-driven model that can predict a microgel swelling profile and subsequently VPTT based only on the microgel recipe. A clustering-based adaptation of partial least squares modeling is developed and subsequently applied to a data library of pH 4 (fully protonated) and pH 10 (fully ionized) swelling responses of 32 pH/temperature-responsive poly(Nisopropylacrylamide) microgels functionalized with various carboxylic acid-functionalized comonomers. We demonstrate that the best-performing clustering and data arrangement strategies can predict the VPTT of the microgels within 1.0 °C at pH 4 and 2.4 °C at pH 10, an accuracy similar to the uncertainty estimates from the experimental transition temperature data (0.6 °C at pH 4 and 2.2 °C at pH 10). Such an approach thus paves the way for faster customization of a microgel swelling profile as needed for a target application.

No abstract