Q&A with chemist Max Delferro: Unlocking breakthroughs in catalytic science

Max Delferro is a chemist at the U.S. Department of Energy’s (DOE) Argonne National Laboratory, where he leads the Catalysis group. His research focuses on developing new catalysts for chemical processes with a wide range of applications. Working at the intersection of chemistry, machine learning and robotics, he is exploring how autonomous discovery platforms can accelerate solutions to problems that have challenged scientists for decades. In this Q&A, he discusses the promise of this approach — and how it could unlock breakthroughs in catalytic science.

“Industry has invested billions of dollars in this problem, with some success. Autonomous discovery could change the equation.” — Max Delferro, chemist and group leader

Q: What is catalysis and why does it matter?
Catalysis is the science of helping chemical reactions happen more efficiently — producing more of the product you want and fewer unwanted byproducts. Catalysts underpin much of modern life, from fuels, plastics and fertilizers to medicines and food ingredients. In fact, it’s been calculated that much of the nitrogen in the human body comes from the Haber-Bosch process, a catalytic reaction widely used by industry for over a century.

Catalysis is also an economic powerhouse. It underpins major sectors of the global economy, including petroleum refining, energy production, chemical manufacturing and food processing. Catalysis enables roughly 35% of the global economy. The catalyst market itself is worth roughly $20 billion annually. Discovering catalysts that operate at higher efficiency, lower cost and lower energy demand can reap enormous benefits for society.

Q: What are the challenges with autonomous discovery in your work?
We’re still building toward full autonomy for experimental research purposes. That would mean experiments, data analysis and the design of the next steps all happening automatically, guided by artificial intelligence (AI) and robotics. Three pieces are critical.

First is instrumentation. I’m not a robotics specialist, so I’ve had to learn how to program robotic systems for complex experiments. Second is training AI to solve complex problems in catalysis. We are working on that now with our Data Science and Learning division. Third — and most important — is the science. Every project starts with a question: What problem are we trying to solve? Autonomous discovery is one tool for answering that, and it works best when paired with a scientific goal suitable for an autonomous discovery platform.

Q: How are you using automation in your lab today?
Argonne’s Accelerated Discovery Laboratory has been running high-throughput experiments for more than 15 years. Instead of testing one sample at a time, we can run hundreds — or even thousands of samples — in parallel. It’s like baking every possible variation of a cookie recipe at once rather than making them one by one.

This generates massive datasets and, sometimes, millions of results. Right now, my team analyzes that data by traditional means to find the most promising catalysts. The next step is to have AI analyze the results, spot patterns, identify top candidates and propose the next round of experiments for the humans to assess. That’s the hurdle we’re working on now — creating a true closed-loop system where machines and algorithms work hand in hand with humans to speed discovery.

Q: Can you give an example of a problem where this could make a real difference?
One of the ​“holy grails” of catalysis is converting natural gas directly into methanol. Natural gas is primarily methane, and methane’s carbon-hydrogen bonds are extremely strong. Breaking them selectively, without creating unwanted byproducts, has stumped researchers for decades.

Why does it matter? Methanol is the perfect building block for most industrial catalytic processes. Fuels, plastics, polymers and even pharmaceuticals can all be made from it. The challenge is that methanol is more reactive than methane, so stopping the reaction at methanol — before it continues to carbon dioxide — is extremely difficult.

Industry has invested billions of dollars in this problem, with limited success to date. Autonomous discovery could change the equation. By training AI models on decades of data and linking them to robotic experimentation, we could shrink the development cycle from 15-20 years to potentially fewer than five. That offers real hope for solving a problem that has resisted solution for years.

Q: Why is Argonne a great place to do this work?
Argonne brings together everything needed to push catalysis and autonomous discovery forward: chemists, chemical engineers, materials scientists and computer scientists working side by side. It also houses world-class scientific tools, such as the Advanced Photon Source for atomic-scale studies, the Center for Nanoscale Materials for designing and characterizing new materials and the Argonne Leadership Computing Facility with some of the world’s fastest supercomputers. These DOE Office of Science user facilities combined with the Accelerated Discovery Laboratory let us design a catalyst, test it, analyze the data and plan the next experiment faster than almost anywhere else.

Q: What drew you to chemistry and catalysis?
In high school in Italy, I had a chemistry teacher who completely changed my view of science. Chemistry opened a window into how atoms bond to form molecules, and how you could build something entirely new by controlling those interactions.

Later, as a researcher, I became curious about whether the new materials I was making could serve as catalysts. That curiosity pulled me into catalysis and eventually brought me into the U.S. national laboratory system. Argonne has been the perfect environment to combine fundamental science with use-inspired applications.

Q: Where do you see this heading in the near future?
Our goal is a true closed-loop system where automated machines make and test catalysts, AI analyzes the results instantly, and the system designs and runs the next experiments with minimal human intervention. Looking further ahead, I’d like to see multiple national laboratories collaborate on an AI factory that pools data, expertise and infrastructure. Such a coordinated effort could transform the pace of innovation — not just in catalysis, but across chemistry and materials science.