(361b) Automated, Hybrid Model-Driven pH Adjustment of Viscous Liquid Formulations

Liquid-formulated consumer products are adjusted to a target pH physiologically compatible with our bodies. For these complex, viscous mixtures, the titration process is carried out manually in a painstakingly slow trial-and-error process by experienced industrial scientists. Herein, we accelerate the pH adjustment process by i) building a new robotic platform capable of titrating these challenging viscous formulations, ii) developing further the active machine learning (ML)-driven pH adjustment strategy presented by Pomberger et al., 2023. The platform's novel design with a 3d-printed overhead stirrer and pH probe in a single position facilitates the titration of viscous samples. Additionally, a washing station involving a soak & jetting down in isopropyl alcohol enables full automation as the robot moves between the closed-loop pH adjustment of different samples. Consumer products are typically titrated with citric acid, as this weak, polyprotic acid can confer the final product a strong buffering effect to remain at the target pH, resulting in a longer shelf-life. However, the prior literature has only developed an ML strategy compatible with strong-strong acid/base titrant pairs. Here a weak-strong acid/base pair is considered, where the deprotonation state of the weak acid depends on the pH, and there is a possibility of forming a complex buffer solution if the target pH is overshot. Accordingly, a data-driven approach (Gaussian Process regression) is combined with physical chemistry to present a hybrid model capable of efficiently guiding the robotic platform to the target pH. Finally, scale-up studies are underway, with the aim that this technology is adopted into industrial high-throughput formulations workflows.