Machine-learning guided prediction of functional organic materials within the Organic Materials Database

SNIC 2018/1-52


SNIC Large Compute

Principal Investigator:

Alexander Balatsky



Start Date:


End Date:


Primary Classification:

10304: Condensed Matter Physics

Secondary Classification:

10201: Computer Sciences



The class of organic crystals comprises carbon-based molecules that crystallize in well-defined lattices. These materials come with wide-ranging applications, such as efficient solar cells, flexible organic LED, carbon sequestration materials or explosives. The theoretical modeling of these materials is bound to characteristic challenges such as distinct energy scales for inter- and intramolecular electron hopping, and strongly localized and correlated electrons. Organics come with the advantage of being soft and conducive to pressure and having an infinite configuration space resulting in a broad variety of options for functionalizing and tuning. With the specific aim of identifying and designing novel functional materials in the class of organic materials, we have initiated the organic materials database OMDB in January 2016. The OMDB is a freely accessible electronic structure database for previously synthesized 3-dimensional organic crystals, available via The contained data has been calculated by us in the framework of the density functional theory using VASP. At the current stage, the database hosts band structures and density of states for about 25,000 materials. The web interface of the OMDB contains non-trivial search tools for the identification of novel functional materials such as band structure pattern matching and density of states similarity search. Currently, we are implementing state-of-the-art machine learning tools, e.g., for the prediction of band gaps and topological classes.