Q&A with ORNL’s Advincula on autonomous labs in materials research

In a recent webinar organized by the National Academy of Engineering in connection with its forthcoming Fall 2025 Bridge issue, Rigoberto "Gobet" Advincula, Oak Ridge National Laboratory-University of Tennessee (UT-ORNL) Governor's Chair Professor and leader of ORNL's Macromolecular Nanomaterials Group, discussed the transformative role of AI in polymer research. Advincula explained how computational techniques and high-throughput experimentation, enabled by advanced laboratories that integrate both automation and autonomous decision-making, are reshaping sustainable materials science and experimental design in polymer chemistry.

Q&A with Advincula

Q: Could you explain what AI-driven methodologies in polymer research entail and why they are so transformative for the field?

A: Imagine a chemist faced with a vast body of literature on potential new polymers and formulations. Traditional experimental approaches can be slow when testing the myriad hypotheses that arise. AI enables us to rapidly decode existing data and literature, essentially accelerating the scientist's mind. By converting this information into computational models, we can derive more precise experimental hypotheses. In practical terms, AI extends human intuition with the ability to systematically and swiftly optimize experimental planning and execution.

Q: In other words, AI amplifies the scientist's knowledge and skills? Building on that foundational insight, let's explore the inspiration behind integrating AI into experimental polymer science.

A: Precisely. Rather than replacing the human element, these tools refine our focus by targeting specific, nuanced tasks. The fear that AI might displace researchers is unfounded when AI is used wisely. Experienced scientists who leverage these techniques find their work can be more data-driven and efficient, which in turn frees them to pursue innovative and complex lines of inquiry.

Q: What inspired you to integrate AI into experimental polymer science, and how did that shape the core message of your recent webinar?

A: My background is a blend of fundamental chemistry and an affinity for engineering. Early in my career, I recognized the value of applying rigorous simulation methods, such as Taguchi and statistical approaches, to accelerate hypothesis testing. The concept of an autonomous lab, in which the computer constantly refines experimental conditions by closing the feedback loop, caught my imagination. This represents not just automation (preprogrammed steps) but rather true autonomy, in which the system dynamically alters its process based on live data. At ORNL, collaborations on projects like AutoFlowS demonstrated that continuous-flow chemistry combined with AI could accelerate and improve the efficiency of discoveries. This integration of chemistry and automation inspired the core message of my talk: by merging computational prowess with both automation and autonomous control, in which automation executes repeated tasks and autonomy allows adaptive decision-making, we can dramatically transform polymer research.

Q: Your talk introduced the concept of autonomous laboratories. How do these automated labs operate, and what advantages do they offer over traditional experimentation methods?

A: Autonomous labs represent a marriage of mechatronics, real-time data analytics and high-performance computing. In essence, these systems can perform a vast number of experiments by iteratively testing hypotheses in a nearly continuous cycle. Whereas traditional batch experiments might be limited by manual setup and high costs associated with energy and time, autonomous labs distill these trials into a few rapidly optimized tests. This approach minimizes resource use and accelerates data collection, thereby enabling a more dynamic and energy-efficient exploration of chemical space.

Q: The webinar attracted a diverse audience that included prominent professors, policymakers and experts from agencies like the National Institute of Standards and Technology and the Air Force Research Laboratory. How might these AI techniques influence national science and engineering policies?

A: Reflecting on initiatives such as the Materials Genome Initiative, although simulation methods have vastly expanded our theoretical capabilities, the absence of high-throughput experimentation has stalled their full potential. Autonomous labs address this very bottleneck by bridging the gap between hypothesis and production. As universities and industries increasingly integrate these technologies into their workflows, policymakers are likely to see a compelling case for boosted investment in automated experimentation. In essence, integrating AI and machine learning with autonomous labs promises groundbreaking materials and a robust framework for industrial and academic innovation.

Q: Moving from policy implications to the practical realm of experimental breakthroughs, consider the following explanation on how simulation and real-time data integration accelerate innovation. Projects such as AutoFlowS demonstrate the potential of bridging simulation with real-world testing. How does this integration accelerate breakthroughs in next-generation polymer designs and advanced reaction engineering?

A: In the traditional batch approach, experiments are inherently sequential and time-consuming. In these advanced labs, automated systems (such as continuous-flow reactors and robotic manipulators) efficiently execute experiments. Whereas automation handles the execution of routine tasks, autonomy comes into play by enabling the system to dynamically analyze real-time data and adjust experimental parameters. Meanwhile, autonomous algorithms analyze results in real time and adjust experimental parameters without human intervention. This allows simultaneous manipulation of reaction parameters such as temperature, pressure and flow rate. By coupling the rigorous predictions from AI simulations with these rapid, data-rich experiments, we can repeatedly refine the reaction conditions. This synergy ultimately drives higher yields, better reproducibility and a more efficient discovery of novel materials, paving the way for industrial-scale applications.