Researchers at the National Laboratory of the Rockies are developing "self-driving laboratories" that use artificial intelligence to automate linked experimental tasks, dramatically speeding up materials science research. By combining AI, robotics, and machine learning, these autonomous systems can accomplish routine experiments from sample fabrication to characterization without human intervention, fatigue, or variation.

What Are Self-Driving Laboratories and How Do They Work?

Self-driving laboratories operate on a principle similar to autonomous vehicles, but instead of navigating roads, they navigate experimental workflows. The concept involves automating sequential laboratory tasks using AI-integrated platforms that can be programmed to run experiments unaided by humans. Rather than a chemist manually performing batch chemistry with a stopwatch and glassware, these systems use pumps, robotic arms, and in-line spectroscopy to conduct experiments at scale.

The hardware platforms use control algorithms powered by high-performance computers, allowing researchers to program a selection of experiments that would otherwise require hundreds of manual iterations. For example, studying the effects of different temperatures, curing times, solvents, precursors, and concentrations on semiconductor film growth could result in hundreds of separate experiments. A self-driving lab accomplishes these routine tasks without human oversight.

How Are These Systems Accelerating Materials Discovery?

The speed improvements are remarkable. Researchers at the National Laboratory of the Rockies have already reduced the time required to complete spectroscopic measurements from 60 to 90 minutes down to just 0.1 to 0.3 seconds, a roughly 1,000-times reduction. This dramatic acceleration is enabling faster evaluation and manufacturing of semiconductors and catalytic nanomaterials.

The laboratory is pursuing two primary case studies to build the foundation for an AI-driven research and development framework. The first focuses on automating the manufacturing of semiconductor films, which are useful for light absorption and energy conversion. The second involves reimagining the synthesis of catalytic nanocrystals used in advanced chemical and fuel production processes.

"Fabricating semiconductor films involves both art and skill, and it's also widely considered to be difficult to replicate specific laboratory procedures reported in literature in another lab. We're interested in using automated processing not just because it speeds up lab work but also because it ensures processes are done exactly the same, every time," said Joey Luther, a senior research fellow at the National Laboratory of the Rockies.

Joey Luther, Senior Research Fellow, National Laboratory of the Rockies

Why Is High-Quality Data Critical for AI-Driven Discovery?

One of the biggest challenges in deploying AI for scientific research is the lack of high-quality, reproducible, domain-specific data. Unlike general-purpose AI chatbots trained on billions of internet documents, AI systems designed for scientific discovery have access to far smaller datasets. Self-driving laboratories address this gap by generating large volumes of consistent, high-quality experimental data that can train more effective AI models.

"If AI is going to drive innovation in scientific domains, it has a major hurdle to surmount: the lack of any significant volumes of high-quality, reproducible, domain-specific data. Unlike general purpose chatbots, AI copilots in the scientific domain have an exceedingly small dataset to draw from. We hope that developing rapid, autonomous systems could fill this gap in the short term, generating the data necessary to drive innovative AI applications in the physical sciences," explained Frederick Baddour, a senior chemist in the National Laboratory of the Rockies' bioenergy and bioeconomy program.

Frederick Baddour, Senior Chemist, National Laboratory of the Rockies

Steps to Implementing Autonomous Laboratory Systems in Research

• Design from the Ground Up for AI: Rather than retrofitting existing laboratory procedures with automation, researchers should redesign workflows specifically for AI integration. The National Laboratory of the Rockies took a "ground-up approach" by developing continuous flow chemistry systems and custom control software rather than simply automating traditional benchtop methods.
• Establish Data Generation Pipelines: Build systems capable of producing thousands of high-quality samples daily to create the datasets necessary for training effective AI models. The National Laboratory of the Rockies received funding from the U.S. Department of Energy's Advanced Research Projects Agency-Energy to develop catalysts ten times faster by producing thousands of samples daily and scaling to kilogram-scale production.
• Create Closed-Loop Systems: Work toward platforms where humans set target objectives and the system designs and performs its own experiments for new materials discovery, rather than requiring humans to program each individual experiment.
• Develop Shared Tools and Frameworks: Build databases and standardized frameworks that other researchers can utilize to accelerate their own timelines to autonomy, enabling broader adoption across institutions.

What's the Broader Impact on Materials Science Research?

The implications extend beyond speed. Autonomous laboratories ensure reproducibility, a persistent challenge in materials science where procedures reported in academic literature often cannot be replicated in other laboratories. By standardizing processes through automation, these systems could help address one of science's most vexing problems: the difficulty of reproducing published results.

The National Laboratory of the Rockies' work is part of a larger trend in AI-driven research and development. Across industries, AI is beginning to reshape how scientists approach discovery. In materials science specifically, AI is being used to simulate and predict material properties, accelerating the design of new compounds for batteries, solar cells, and other technologies.

Jovana Milić, an associate professor at the University of Turku in Finland who leads a smart energy materials team, emphasizes the importance of interdisciplinary collaboration in advancing materials science. Her research focuses on developing hybrid supramolecular materials that respond to external stimuli and adapt to operating conditions in emerging energy technologies such as solar cells and neuromorphic systems. This kind of cross-disciplinary work, combined with AI-powered automation, represents the future of materials discovery.

"I am convinced that interdisciplinarity is essential for addressing some of the most pressing societal challenges," noted Jovana Milić, Associate Professor at the University of Turku.

Jovana Milić, Associate Professor, University of Turku

The National Laboratory of the Rockies is now actively seeking opportunities to scale up their autonomous laboratory efforts and expand adoption across the broader research community. As these systems mature, they could fundamentally transform how scientists approach discovery, moving from hypothesis-driven experimentation to data-driven, AI-guided exploration of materials space.