Machine Learning Exploration of ORR Catalytic Activity on Metallic (110) Surfaces
Internship made during my second year as an engineering student at Télécom Physique Strasbourg. Using machine learning to study fuel cell catalyst activity on metal surfaces.
Abstract :
Increasing the use of renewable energy is crucial to meeting the growing global demand for energy while also mitigating the impacts of climate change. The two most commonly used forms of renewable energy are wind and solar power. Nevertheless, the wind may cease, and the Sun will eventually sink below the horizon making their respective energies unavailable during that time. The best option existing to solve this problem is the Hydrogen Energy Storage (HES) ; electrical energy is converted to hydrogen (easy to store) that can be later converted back into electrical energy. The reactions involved in HES are using electrocatalysts to speed them up ; their efficiency depends substantially on their chemical composition, crystal system and surface. Finding an electrocatalyst that is not only highly performant but also inexpensive is therefore an important challenge for the world.
The significant advancements in quantum mechanics and particularly the development of the Density Functional Theory, coupled with the recent progress in machine learning have made it elementary to compute and understand more deeply what precisely happens in the electrochemical cells.
Open-source resources like the OC20 database and the EquiformerV2 machine learning model, both developed by the Open Catalyst Project, were used to simulate the process happening in the fuel cell and revealed the best catalysts on alloys made from a list of 14 different transition metals. It has been found that platinum, antimony and their allows are notably efficient for the considered reaction. Further studies showed that local and global electronegativity, atomic radius, first ionization energy and generalized coordination number do not influence the efficiency of a catalyst.
Keywords : Renewable energy, catalyst, fuel cell, adsorption, machine learning.
