A team of U.S. Department of Energy researchers is building an AI system inspired by AlphaFold, Google DeepMind's famous protein-folding tool, to predict how microscopic flaws in computer chips affect their performance and lifespan. Instead of predicting protein structures from genetic sequences, the new framework, called the Materials Discovery Cloud, will connect defect patterns in materials to the electrical and thermal properties that determine whether a microelectronic device works reliably.

Why Are Tiny Defects in Chips Such a Big Problem?

As computer chips become smaller, faster, and more densely packed, they become increasingly vulnerable to microscopic imperfections in their materials and interfaces. These defects can cause overheating, electrical leakage, unreliable switching, and premature device failure. However, not all defects are harmful; some can actually improve electrical and thermal behavior depending on how they are distributed and how they change over time.

The challenge is that no single measurement tool can capture the complete picture of how defects evolve and affect device behavior. Electron microscopes can image tiny features, X-ray methods reveal strain and buried structures, and other techniques measure chemistry and heat flow. Each tool shows only one piece of the puzzle, making it difficult for researchers to understand the full system.

How Is AlphaFold Inspiring a New Approach to Microelectronics?

AlphaFold transformed biology by learning to predict a protein's three-dimensional structure from its amino acid sequence, making structure prediction faster and more accurate than ever before. The Materials Discovery Cloud project, led by Argonne National Laboratory with collaborators at Lawrence Berkeley National Laboratory, Oak Ridge National Laboratory, and Northwestern University, applies a similar principle to a completely different challenge.

"In biology, AlphaFold learned to connect sequence to structure. We want to connect defect distributions in materials and interfaces to the electrical and thermal properties that matter for microelectronics. Such a framework remains elusive," said Subramanian Sankaranarayanan, Argonne scientist and lead principal investigator on the project, as well as a professor at the University of Illinois Chicago.

Subramanian Sankaranarayanan, Argonne scientist and professor at the University of Illinois Chicago

Rather than predicting protein folding, the team aims to predict how networks of defects form, change, and influence whether a device works properly. The framework will combine experimental data from major U.S. Department of Energy facilities, advanced computer simulations, and high-performance computing to learn these relationships.

What Data Sources Will Power the Materials Discovery Cloud?

The project will integrate information from multiple world-class research facilities and computing centers:

• Argonne National Laboratory: The Advanced Photon Source and Center for Nanoscale Materials for X-ray and microscopy measurements, plus the Argonne Leadership Computing Facility for large-scale computing
• Lawrence Berkeley National Laboratory: The Advanced Light Source, the Molecular Foundry, and the National Energy Research Scientific Computing Center
• Oak Ridge National Laboratory: The Center for Nanophase Materials Sciences

By bringing together these diverse data sources, researchers can create a more complete picture of how materials behave. The system will also generate synthetic data from simulations to fill gaps where complete experimental datasets are difficult or time-consuming to collect.

How Will the System Actually Work in Practice?

The Materials Discovery Cloud is designed as a physics-informed AI framework, meaning it will be built around well-established laws of physics rather than operating as a black box that produces answers without explanation. This approach ensures that the AI's predictions are grounded in how materials and devices actually behave in the real world.

The framework will combine several types of AI tools that can process many different data types, learn from both experiments and computer simulations, and identify which new measurements would be most useful to run next. The ultimate goal is to connect atomic-scale features in a material to the larger electrical, thermal, and mechanical behavior of an actual device.

Another key component involves autonomous discovery, which uses AI, machine learning, and robotics to help researchers decide which measurements to conduct next and collect new data more quickly. The team is developing an AI-guided platform that can synthesize samples and perform multiple types of characterization in an integrated, high-throughput workflow, reducing the need to move samples between different facilities.

"What makes the Materials Discovery Cloud powerful is that we can bring together experiments, simulations and AI in one workflow. That gives us a way to learn from limited data today while building a framework that can grow more capable as new data comes in," explained Sankaranarayanan.

Subramanian Sankaranarayanan, Argonne scientist and professor at the University of Illinois Chicago

What Could This Mean for the Future of Chip Design?

If successful, the Materials Discovery Cloud could fundamentally change how scientists design and test new materials and devices. Instead of waiting through lengthy rounds of experiments, researchers may be able to get useful answers from a smaller set of early measurements. This could help them spot problems sooner, avoid spending time on weak candidates, and focus resources on the most promising designs.

The project is part of the U.S. Department of Energy's Genesis Mission, a national initiative aimed at doubling America's research and development productivity within a decade while strengthening U.S. technological leadership and global competitiveness. By combining AI with world-class experimental facilities and computing power, the Materials Discovery Cloud represents a new model for accelerating materials science and microelectronics innovation.