MIT researchers develop AI tool to monitor material defects

RESEARCHERS at Massachusetts Institute of Technology (MIT) have developed an AI-based tool to help identify and measure atomic-scale defects in materials used in products such as steel and semiconductors.

Defects – small deviations in a material’s atomic structure – are often deliberately introduced to tailor properties such as strength and electrical conductivity. However, identifying exactly what has been created, and in what concentration, remains a longstanding challenge, particularly without damaging the material.

Lead author Mouyang Cheng and his team at the Department of Materials Science and Engineering have developed an AI model that can classify and quantify defects using neutron-scattering data, offering a non-invasive alternative to conventional diagnostic techniques.

Chu-Liang Fu, a postdoc in the research team, said: “Engineers have many ways to introduce defects, like through doping, but they still struggle with basic questions like what kind of defect they’ve created and in what concentration.

“Sometimes they also have unwanted defects, like oxidation. They don’t always know if they introduced some unwanted defects or impurity during synthesis.”

Quantifying defects

To support the approach, the MIT team developed DefectNet, a computational database which includes data on around 2,000 semiconductor materials.

Materials with defects were paired with similar defect-free materials, with neutron-scattering measuring differences in atomic vibrations. The data was then used to train a machine learning model.

Cheng explained: “The model leverages the multihead attention mechanism, just like what ChatGPT is using. It similarly extracts the difference in the data between materials with and without defects and outputs a prediction of what dopants were used and in what concentrations.”

Industry integration

Despite its potential, the team acknowledged that adoption in industrial quality control may be limited by the complexity of neutron-scattering techniques.

Eunbi Rha said: “This method is very powerful, but its availability is limited. Vibrational spectra is a simple idea, but in certain setups it’s very complicated. There are some simpler experimental setups based on other approaches, like Raman spectroscopy, that could be more quickly adopted.”

The team is now looking to expand its research beyond point defects and into the monitoring of grains and dislocations.