Researchers at Pacific Northwest National Laboratory have created an AI-powered system that reduces the time needed to extract critical minerals from industrial waste from months or years to just days. The breakthrough could reshape how the United States recovers valuable materials from discarded magnets, petroleum wastewater, and other industrial byproducts, offering a faster path to domestic mineral sourcing.

How Does This AI System Work?

The research team, led by materials scientist Elias Nakouzi, built a semi-autonomous laboratory connected to specially designed AI agents called Computer Intelligence for Critical Elements Recovery and Optimization, or CICERO. The system combines a liquid-handling robot, sample handling equipment, and analytical instruments with AI software to automate the mineral recovery process.

Here's how the workflow operates in practice:

• Initial Assessment: Scientists feed the AI agents a description of what chemicals are present in the industrial waste, and the agents evaluate the value, concentration, and potential purity of recovered materials.
• Experiment Design: Within a single day, the AI uses published scientific literature to design 96 simultaneous experiments, including precise recipes for chemical separation, the order chemicals should be added, and exact timing steps.
• Automated Execution: A liquid-handling robot executes all 96 experiments automatically, dramatically accelerating what would normally require months of manual preparation and testing.
• Iterative Refinement: AI automatically evaluates the results and can recommend a second round of 96 experiments to optimize purity and yield if needed.

What Materials Can CICERO Recover?

To demonstrate the system's capabilities, the research team tested three different types of industrial waste. The AI agents recommended recovery of magnesium from wastewater produced during oil and gas extraction, neodymium and praseodymium from spent magnet waste, and samarium, a rare-earth element critical to high-performance aerospace magnets and nuclear reactors.

"We connected a liquid-handling robot, a sample handling device, and two analytical instruments and created an AI-aided workflow that quickly isolated critical minerals from industrial samples," said Elias Nakouzi, materials scientist at Pacific Northwest National Laboratory. "These industrial feedstocks are a complex soup of chemicals. Developing an effective method to isolate one element from the soup can take months or years. We have reduced that time to days with CICERO."

Elias Nakouzi, Materials Scientist at Pacific Northwest National Laboratory

Traditionally, evaluating which elements could be recovered from a given feedstock and determining the best separation method required months of analysis and preliminary laboratory protocol preparation. CICERO collapses that timeline dramatically.

Why Does This Matter for U.S. Manufacturing?

The United States faces growing demand for critical minerals needed in everything from renewable energy systems to defense applications. Currently, much of the world's supply comes from overseas, creating supply chain vulnerabilities. CICERO offers a pathway to extract valuable materials from what has traditionally been treated as waste, potentially creating new economic incentives for domestic mineral production.

The researchers emphasized that the chemistry behind CICERO is already proven at industrial scale. The commodity chemicals used in the experiments are already deployed in other chemical separation processes across the industry, meaning the pathway from laboratory demonstration to real-world manufacturing is more straightforward than it might appear.

"The agentic AI allows us to get more mileage out of existing industry practices for critical mineral recovery," noted Maxim Ziatdinov, a physical scientist at Pacific Northwest National Laboratory whose research has merged AI, data science, and instrument controls. "It may be possible to target additional critical materials in a broader range of feedstocks as more and more experimental results are processed."

Maxim Ziatdinov, Physical Scientist at Pacific Northwest National Laboratory

CICERO is powered by SciLink, an agentic AI platform developed at PNNL and supported by the Department of Energy Office of Science. Agentic AI refers to AI systems that can reason independently, plan multi-step workflows, and make decisions without constant human intervention.

What's Next for This Technology?

The research team is rapidly exploring additional opportunities for CICERO, asking the system to reason beyond initial ideas and incorporate data from early experiments to generate even better recovery strategies. As the system processes more experimental results, researchers believe it may be possible to target additional critical materials in a broader range of feedstocks.

"We are on the cusp of something exciting here, not just for optimization and efficiency, as we've shown here, but also potentially for new chemistry and new materials science that we could discover with these platforms," said Nakouzi.

Elias Nakouzi, Materials Scientist at Pacific Northwest National Laboratory

The research was published in the journal Materials Horizons and represents a significant step forward in applying AI to real-world materials science challenges. While recycling magnets and petroleum wastewater has not yet been done on an industrial scale, CICERO demonstrates that the pathway is economically and technically feasible.