Accelerating Net Zero Manufacturing with Intelligent Optical Reactor Technology

Quickly transitioning to net zero manufacturing requires researchers to use modern methods for process discovery and development. This is especially true for the development of heterogeneously catalysed processes, which are still typically optimised by painstaking trial-and-error methods. As such, the overall objective of this project is to combine pioneering spectroscopy and reactor technology with machine learning and real-time optimisation to achieve self-optimising reactor technology applicable to the synthesis of important chemical products from biomass feedstock using heterogeneous catalysts.

We recently pioneered development of a novel heterogeneous catalyst reactor equipped with fiber optic technology. This reactor allowed us to follow the conversion of glucose to various bio-based chemicals of industrial interest using heterogeneous zeolite catalysts, at true continuous operational conditions and in real time, by performing operando optical spectroscopy.

In particular, unique optical signals related to the active sites of the catalyst and all of the reaction pathways of the chemical process were identified, including the desired reaction pathway and those of undesired side reactions. Moreover, we could relate the intensity of each optical signal to the quantity of products formed, as verified by offline methods (HPLC and 1H-13C HSQC NMR), resulting in real time information of the performance of each catalyst in terms of activity, selectivity, and stability.

As the reaction pathway signals are charge transfer bands associated with activation of the substrate by the catalyst (as opposed to chromophores of reactants and products), they also provide direct insight into the transition states of the various reaction pathways.

These breakthroughs are unprecedented in catalysis research, and represent the starting point for this project. This project will use this breakthrough to target the application of machine learning to heterogeneous catalysis and biomass conversion. In particular, we will combine fast data generation through operando optical technology with online learning algorithms to construct data-driven models in real-time through time series modelling.

This will facilitate the development of ML algorithms that uniquely account for activity, selectivity and stability, at various operational conditions. We will validate this hypothesis, and then use the Intelligent Optical Reactor (IOR) technology to accelerate bio-based chemical manufacture by developing self-optimising reactor technology. In later stages, we will attempt to use the transition state insights provided by the reactor to generate advanced structure activity relationships of relevance to future catalyst design.