At SINTEF, we are working to create materials for new sustainable technology that the industry needs, such as batteries, solar cells, LED lighting and biomaterials. The materials must have the exact properties, be durable and contain as few substances hazardous to the environment and health as possible.
The design process for new materials has accelerated thanks to powerful computers and advanced calculations. A number of different materials databases that can provide information about the properties of the various materials have been developed. These databases can be used to train AI models to optimize materials.
– In order to create new, correct AI models, we need large amounts of data from all relevant databases. However, the different databases also have different interfaces and ways of representing the data. This has made it difficult to use them together," says Casper Welzel Andersen, a researcher at SINTEF.
Easier access to materials databases
In the last eight years, a new standard has been developed for sharing digital data on materials; OPTIMADE (Open databases integration for materials design). The standard allows users to interact with different materials databases in the same way. It is based on existing web-based protocols and developed in collaboration with large materials databases in Europe and the USA. A large international network of more than 30 institutions worldwide, including SINTEF, is behind it.
With OPTIMADE, researchers at universities or in industry can easily communicate with all these databases and understand the information they receive," says Gian-Marco Rignanese, professor at the Department of Condensed Matter and Nanoscience at UCLouvain in Belgium.
Huge amounts of different data
Today, powerful computers are used to do advanced simulations of how electrons move in materials, which in turn provide information about the various material properties. These calculations provide large amounts of data that can be used to train machine learning models. The AI models can then predict the properties of new materials based on new calculations.
However, huge amounts of data are required to train the models. Over time, many such databases have been emerged from various research groups and projects. Each of these databases has its own way of doing searches, its own way of describing the data, and its own way of describing what the data represents. With OPTIMADE, data from large-scale simulations, and general data on materials, are available through the same protocol. In this way, search queries and data representation will be similar.
– The latest version of OPTIMADE provides a greatly improved ability to accurately describe the properties of different materials using clear, machine-readable definitions. With these, we can understand data from the different databases and use them together," says Casper Welzel Andersen.
Gives researchers more time for scientific work
The OPTIMADE standard is a highly valued tool in the daily work of researchers in material science at SINTEF.
– In one of our AI projects, we faced challenges in aligning experimental results and simulation outcomes for some important metallurgical data. It turns out that even though both communities have been talking about the same things for many years, it can be really difficult to bridge the gap between the two worlds. Solutions that OPTIMADE is bringing to the table can help us spend more time on scientific and engineering challenges and less time dealing with issues that get ‘lost in translation’, says research scientist Daniel Marchand.
The OPTIMADE standard has already been used in the EU-funded project OntoTrans. Here, an Open Translation Environment has been developed, which is a toolbox for understanding and solving industry problems related to different materials. The toobox uses OPTIMADE to give the user access to a wide range of calculation results. OntoTrans is now in its final phase, but the toolbox has already been continued in the industry consortium Semantic Materials, where SINTEF is one of the core contributors.
