Accelerating materials discovery

Funded by the Science and Technology Facilities Council’s (STFC) Commercial Pump Priming Fund, the Hartree Centre developed Hartree-MaDE, a new computational tool for materials discovery.

Challenge

Computational materials discovery involves searching vast material spaces for potential candidates that have the optimal combination of suitable properties for target applications. High fidelity predictions are required to make meaningful predictions, so navigating a material space containing tens or hundreds of thousands of materials is a significant challenge. A tool that can help navigate through a maze of candidate materials will help accelerate the discovery process.

Approach

The Hartree Centre’s Materials Discovery Engine (Hartree-MaDE) is a tool that simplifies and automates the process of materials discovery. Its main focus is on alloy discovery – particularly substitutional high entropy alloys – but the tool can model all hard materials that exhibit crystalline order, including solid ordered phases of polymers. It relies on so-called descriptors to predict the stability of alloys from the mechanical properties of the component materials, which are predicted at high fidelity using state-of-the-art density functional theory. Observable properties of the alloys are similarly predicted from those of the component materials, via carefully chosen averaging schemes. The only user input required is the structure and composition of the component materials. Hartree-MaDE then automatically builds the database of every possible substitutional alloy that can be engineered from combinations of the base materials, allowing users to go forward and quickly identify the best candidate materials for their target application.

Benefits

Using Hartree-MaDE to screen an initial set of materials de-risks and accelerates materials discovery while significantly reducing development costs. Businesses can use the software to supply them with candidate materials without the need for expensive trial and error procedures in the lab. By predicting a broad range of both thermal and mechanical properties, Hartree-MaDE can help businesses identify and develop new hard materials. Future versions of the tool will include an active learning AI algorithm that will further enhance fidelity and continue to accelerate materials discovery.

Working with The Hartree Centre enabled us to efficiently explore an extremely complex area of ceramic material discovery for a niche application where currently available options are far from ideal.

Richard White, Lucideon