Machine learning based method for finding adsorption sites

events hall

Mrs. Sean Pachmanov-Dvir MSc candidate


David Wang Auditorium, 3rd floor Dalia Meidan Bldg.


One of the main stages of catalytic reactions research is finding the adsorption site of the molecule or atom adsorbed on the surface. Only then one can calculate the next steps of the reaction. The conventional approach to determine this adsorption site is based on chemical logic and heavy calculations. Since DFT involves high calculation time, the number of initial sites that are examined is limited by the calculation resources availability and cost.

In this work, using DFT single point calculations, a database of adsorption energies has been created. This database contains the adsorption energy of single atoms such as hydrogen or oxygen located on different sites of given surfaces.

By applying machine learning algorithms on this database, our research suggests an improved way of choosing an initial adsorption site, which enables scanning of the entire surface, including the area between the surface atoms.


BSc in Materials Engineering and Chemistry at the Department of Materials Science & Engineering, Technion.

Supervisor: Assoc. Prof. Maytal Caspary Toroker