For the chemical industry to achieve carbon neutrality by 2050, where fossil fuels are replaced as a feedstock and source of energy, many new functional materials will need to be developed. The classical approach to materials discovery has been screening programs where a limited number of materials are synthesised and tested. This approach is resource intensive and slow, and new methodologies for fast discovery of materials have developed over the last 30 years. Two opposite approaches are high throughput synthesis and testing of materials; and secondly, quantum mechanical modelling aimed at understanding processes on a molecular level. First principles quantum mechanical modelling, such as density functional theory (DFT), have made great strides over the last 20 years in its ability to model physical properties and reaction micro-kinetics. However, this modelling is still time consuming.
FUNMAT aims to increase the efficiency of DFT modelling of materials and reactions through validated modelling of properties of key materials, coupled with machine learning, changing the modelling from a descriptive to a predictive mode. The focus of FUNMAT will be the development of functional materials for an electrochemical cell producing green hydrogen with high efficiency. Modelling will be coupled with advanced operando analysis methods on model samples to gain insight into the near surface structure of the materials, providing further validation of the modelling. New materials discovered in FUNMAT will be synthesised and tested in functioning electrochemical cells. The successful outcome from this project, on a practical level will be efficient methods to discover new materials that will help Yara towards its commitment to become a carbon neutral fertilizer producer by 2050. Another important outcome will be to demonstrate the use of advanced modelling and machine learning in the discovery of new functional materials.
The FUNMAT project aims to develop a fast and rigorous methodology for the development of new functional materials that will be necessary to transition the chemical and energy industries to carbon neutral production processes by 2050. The classical experimental screening method, where a limited numbers of materials are synthesized and tested for the target application, is inefficient in terms if resources and critically time. It still relies, to a large extent on intuition to select candidates. ; with advanced computational tools. First principles quantum mechanical calculations, such as Density Functional Theory (DFT) have been proven to be reliable for obtaining a wide range of materials properties. However, even with modern computational power, screening a large matrix of candidate materials is too time consuming to be practical. The approach in this project is to use machine learning, so that the outcome of calculations may be reliably predicted, leading to the so-called High-Throughput Virtual Screening (HTVS). The target test case is a novel electrode for a high temperature electrochemical cell that will co-produce green hydrogen and useful chemicals, when excess renewable electrical power is available; and useful chemicals and electrical power, when intermittent electrical power is absent. This system will reduce the power requirement to produce green hydrogen compared to conventional electrolyzers. Information used in the machine learning process will be validated with experimental data of materials properties; including surface and near surface structure composition and structure, and experimental kinetic data of key reactions. The successful development of this methodology is transferable to many chemical and electrochemical processes, that must to change to meet climate targets; including catalysis, power generation and energy storage. Success may encourage other industrial groups to consider advanced modelling as a tool for materials discovery