Kinetic model identification for hydrogen borrowing synthesis using a cloud platform for model-based design of experiments

Hydrogen borrowing is an increasingly important catalytic process in the synthesis of pharmaceutical intermediates and active drug compounds. Its mechanism is typically described as a three-step sequence: alcohol oxidation, additive alkylation (or arylation) and hydrogen reduction. While the mechanistic steps are well established, the development of predictive kinetic models is critical to enabling process scalability and automation. In this work, the hydrogen borrowing mechanism is embedded within a model-based design of experiments (MBDoE) framework for controlling automated laboratory experimentation via a cloud service. A case study involving benzyl alcohol and benzylamine reaction over a Ru catalyst was conducted. Candidate kinetic models were developed to describe the dynamics of reactants, intermediates and products based on experimental data. Leveraging MBDoE in combination with a novel sequential parameter estimation technique informed by the reaction network, two statistically adequate and identifiable kinetic models were identified. Although initially indistinguishable based on standard experimental data, in-silico simulations exploiting structural differences between the models show that catalyst amount acts as a key model discrimination factor. This work demonstrates how reaction-informed model discrimination through targeted experimental design can advance understanding and control of hydrogen borrowing synthesis, laying the foundation for more robust and scalable processes in the pharmaceutical industry.