Data Driven Molecular Modelling for Reactor Design

Chemical reactor design and optimization based on insights and data on a molecular level can become reality in the coming decade. Multiscale simulations including all elementary reaction steps in a full microkinetic model, with information on catalyst structure and adsorption on different sites are within reach. This extensive approach requires software tools to aid the development of accurate and complete microkinetic models for heterogeneously catalyzed processes. These tools include the automatic generation of all possible chemical reactions, and the calculation of the associated data for these reactions, i.e. the thermochemical parameters of all species and all reaction rate coefficients. This includes the automatic calculation of these data using state-of-the-art density functional theory methods. A data-driven approach is envisioned which will enable the use of strong machine learning techniques to maximize the usability of all data. Since all these steps will be automated and user-involvement will be minimized, we expect the set of developed tools to be more exhaustive, more complete and more accurate compared to current manual kinetic model development. The tools will be tested by using the oxidative coupling of methane as case study, in collaboration with several experimental researchers and with computational fluid dynamics researchers to go from the molecular information up to the full reactor scale in multi-scale simulations.