Finding the needle in a haystack using automated computational quantum chemistry

Functional materials are at the center of most technologies of our everyday life: cell phones, solar cells, batteries, thermoelectric devices, digital storage, catalysis, and many other applications need at least one material fulfilling a task through its intrinsic physical properties. Finding a material well-suited for a given application is far from trivial. Indeed, even though close to 300 000 crystal structures have been reported today, less than 1 percent of those materials have been characterized for any measurable property. 

 

The mapping between crystal structures and physical properties is very far from completely known. Such mapping has to be constructed by characterizing known (or even unknown) materials. The traditional approach is to try and synthetize materials in lab conditions and measure their properties. When looking for a new material for a given application, the next step is to assess whether it is worth pursuing with this material or if another one should be tested. This trial-and-error approach is tedious, long, and demands a lot of resources and researchers. Alternatively, using computational quantum chemistry, it is possible to determine many physical properties of a material using solely its crystal structure. This allows the community to save time and resources by assessing the capabilities of a material before synthesizing it. 

 

However, such computations can sometimes be intricate with many steps involved before reaching the desired property. Even in situations where the computation is not very difficult to perform for a given material, it becomes a daunting task when considering the number of materials for which the property should be assessed. For this reason, in recent years, tools have been developed to automate quantum chemistry computations relying on density-functional theory. These tools can be used to perform thousands of computations and generate faster the mapping between crystal structures and their physical properties, thereby helping materials scientists find suitable materials for their applications.

 

At Matgenix, we are experts at developing new workflows for automated computations. We also use these workflows to generate data that can be stored in databases. Artificial intelligence can then be used to go even faster for materials properties predictions. In the NICKEFFECT project, we aim at developing new workflows for computing magnetic properties of Ni-based coatings. This will help our partners find the most suitable material for digital storage devices. The catalytic activity of various geometries of Ni-based catalysts will also be explored using automated tools.