MIT researchers use AI to uncover atomic defects in materials
Source: MIT News - AI
Defects: From Problem to Powerful Tool
In biology, defects are generally bad. But in materials science, defects can be intentionally tuned to give materials useful new properties. Today, atomic‑scale defects are carefully introduced during the manufacturing process of products like steel, semiconductors, and solar cells to help improve strength, control electrical conductivity, optimize performance, and more.
Even though defects have become a powerful tool, accurately measuring different types of defects and their concentrations in finished products has been challenging—especially without cutting open or damaging the final material. Without knowing what defects are present, engineers risk making products that perform poorly or have unintended properties.
Now, MIT researchers have built an AI model capable of classifying and quantifying certain defects using data from a non‑invasive neutron‑scattering technique. Trained on 2,000 different semiconductor materials, the model can detect up to six kinds of point defects in a material simultaneously—something that would be impossible using conventional techniques alone.
“Existing techniques can’t accurately characterize defects in a universal and quantitative way without destroying the material,” says lead author Mouyang Cheng, a PhD candidate in the Department of Materials Science and Engineering.
“For conventional techniques without machine learning, detecting six different defects is unthinkable. It’s something you can’t do any other way.”
The researchers say the model is a step toward harnessing defects more precisely in products like semiconductors, microelectronics, solar cells, and battery materials.
“Right now, detecting defects is like the saying about seeing an elephant: each technique can only see part of it. Some see the nose, others the trunk or ears. But it is extremely hard to see the full elephant. We need better ways of getting the full picture of defects, because we have to understand them to make materials more useful.” – Mingda Li, senior author and associate professor of nuclear science and engineering
Joining Cheng and Li on the paper are postdoc Chu‑Liang Fu, undergraduate researcher Bowen Yu, master’s student Eunbi Rha, PhD student Abhijatmedhi Chotrattanapituk ’21, and Oak Ridge National Laboratory staff members Douglas L. Abernathy PhD ’93 and Yongqiang Cheng. The paper appears today in the journal Matter.
Detecting Defects
Manufacturers have gotten good at tuning defects in their materials, but measuring precise quantities of defects in finished products is still largely a guessing game.
“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,” says Fu.
“Sometimes they also have unwanted defects, like oxidation. They don’t always know if they introduced some unwanted defects or impurity during synthesis. It’s a longstanding challenge.”
Because multiple defects often coexist, each characterization method only captures part of the picture:
- X‑ray diffraction and positron annihilation characterize only some defect types.
- Raman spectroscopy can discern defect type but can’t directly infer concentration.
- Transmission electron microscopy (TEM) requires cutting thin slices of the sample for scanning.
In earlier work, Li and collaborators applied machine learning to experimental spectroscopy data to characterize crystalline materials. For the new paper, they extended that approach to defects.
- Database creation – The team built a computational database of 2,000 semiconductor materials. For each material they generated a pair of samples: one doped (defective) and one pristine.
- Neutron‑scattering measurements – They used a neutron‑scattering technique that measures the vibrational frequencies of atoms in solids.
- Model training – A machine‑learning model was trained on the resulting spectra.
“That built a foundational model that covers 56 elements in the periodic table,” Cheng explains. “The model leverages the multi‑head attention mechanism, just like what ChatGPT uses. It extracts the differences between spectra of defective and pristine materials and outputs predictions of which dopants were used and at what concentrations.”
After fine‑tuning and verification on experimental data, the model could:
- Measure defect concentrations in a commonly used electronic alloy.
- Measure defect concentrations in a separate superconductor material.
When the researchers doped materials multiple times to introduce several point defects, the model successfully predicted up to six defects simultaneously, with concentrations as low as 0.2 %.
“We were really surprised it worked that well,” Cheng says. “It’s very challenging to decode the mixed signals from two different types of defects—let alone six.”
A Model Approach
Typically, manufacturers of semiconductors run invasive tests on a small percentage of products as they leave the production line. This slow process limits the ability to detect every defect.
“Right now, people largely estimate the quantities of defects in their materials,” says Yu. “It is a painstaking experience to check the estimates by using each individual technique, which only offers local information in a single grain anyway. It creates misunderstandings about what defects people think they have in their material.”
The researchers acknowledge that neutron‑based vibrational spectroscopy is not yet a quick, on‑line quality‑control tool for most companies.
“This method is very powerful, but its availability is limited,” notes Rha. “Vibrational spectra is a simple …” (the quote continues in the original source)
Takeaway
The MIT team’s AI model demonstrates that non‑destructive neutron‑scattering data combined with modern machine learning can quantitatively identify multiple point defects in semiconductors—a capability that could eventually enable more precise defect engineering across a range of advanced materials.
AI‑Driven Defect Detection: Future Directions
Li says companies have already expressed interest in the approach and asked when it will work with Raman spectroscopy, a widely used technique that measures the scattering of light. The researchers’ next step is to train a similar model based on Raman spectroscopy data. They also plan to expand their approach to detect features that are larger than point defects, such as grains and dislocations.
“To the human eye, these defect signals would look essentially the same,” Li says. “But the pattern recognition of AI is good enough to discern different signals and get to the ground truth. Defects are this double‑edged sword. There are many good defects, but if there are too many, performance can degrade. This opens up a new paradigm in defect science.”
For now, however, the researchers believe their study demonstrates the inherent advantage of AI techniques for interpreting defect data.
Funding
The work was supported, in part, by:
- the Department of Energy
- the National Science Foundation