💻 New AI model creates tailored materials for future technology

• MatterGen creates materials based on specific requirements and desired material properties.
• The model has already led to the creation of a new material, TaCr2O6, which has been successfully tested in laboratory.
• The technology is available to everyone under MIT license, allowing researchers worldwide to use it to develop new materials for batteries, fuel cells and other applications.

How MatterGen works

Microsoft's MatterGen uses a special type of AI called diffusion model that works with materials' 3d geometry. The model has been trained on 608,000 stable materials from the Materials Project and Alexandria databases.

When researchers specify desired properties, MatterGen can generate suggestions for new materials that meet the requirements. This can include specific chemical, mechanical, electronic or magnetic properties.

A concrete example of MatterGen's capability is the creation of the material TaCr2O6. Researchers asked the model to generate a material with a specific compressive strength of 200 GPa. The developed material was then tested in laboratory by a team led by Professor Li Wenjie at Shenzhen Institutes of Advanced Technology. The material showed a compressive strength of 169 GPa - an accuracy considered very good in practical terms.

More efficient than traditional methods

Traditional material search is like looking for a needle in a haystack and requires extensive testing of millions of possible candidates. MatterGen takes a different approach by directly generating materials that meet the desired requirements.

In a test where researchers looked for materials with high compressive strength (above 400 GPa), MatterGen continued to find new candidates long after traditional search methods had reached their limit.

Open source creates global opportunities

MatterGen has been made available as open source under MIT license. This means that researchers and developers worldwide can use and build upon the technology.

The model can be used to develop new materials for many different applications, such as batteries, magnets and fuel cells. Johns Hopkins University Applied Physics Laboratory is already investigating how MatterGen can be used in their materials research.