Starch Phosphate Carbamate Synthesis in the Age of Machine Learning: Maximization of the Phosphorus Content and the Educt Efficiency via Bayesian Optimization

ABSTRACT Starch phosphate carbamates (SPC) have been a focal point in the research and development of sustainable flame retardants (FRs) for the past two decades. Within this framework, there is a notable emphasis on obtaining SPC with a high phosphorus content. Aligning with the contemporary trend of artificial intelligence and machine learning, the optimization of synthesis conditions for specific product properties can be efficiently achieved through their utilization. Hence, the objective of this study is to optimize the synthesis conditions, including the molar ratio of anhydroglucose units to urea and phosphoric acid, as well as the synthesis temperature and duration. The optimization is conducted via Bayesian optimization, aiming to maximize the phosphorus content and educt efficiency of SPC to obtain a potent, bio‐based FR. As a result, the phosphorus content and educt efficiency are increased to about 20.5% and 11.3%/mol, respectively.