Trial and error got catalysis this far. Rational design will take it further.

For decades, catalysts have been developed by trial and error, and even the successful ones were often poorly understood. Dr. Luca Lucarelli argues that the field is now moving toward rational catalyst design: precise engineering built on measuring the gas-solid interactions that drive performance, with AI helping make sense of the data.

Luca, Senior Application Scientist at Malvern Panalytical, closed our Future Days: Catalysts presentations on September 22 with his talk, ‘From gas-solid interactions to better catalysts: characterization driving the catalysis of tomorrow.’ With more than 31 years in materials characterization, Luca specializes in physisorption and chemisorption, working directly with catalyst researchers and manufacturers on their analytical challenges.

Catalysis underpins everything from ammonia synthesis to petrochemicals and environmental protection, yet catalyst development has long followed the same process: synthesize a formulation, test it, hope it works, and start again if it doesn’t.

Luca’s approach begins with the catalyst surface: characterize its physical and chemical properties, understand how gases interact with it, design the catalyst with precision, and then validate its performance. AI helps interpret the volume of data this approach generates. Here’s a recap of his insights.

Catch up on the full session and more from Future Days: Catalysts.

A higher surface area is not always best

The physical qualities that make a good catalyst go beyond surface area. Take supported metal catalysts, where metal nanoparticles are spread across a porous support. The exposed metal atoms are the active sites, so metal dispersion – the ratio of exposed metal atoms to total metal atoms – is critical, especially for precious metals such as platinum and rhodium.

The support does more than hold the metal. It provides the surface area and pore structure, and it influences how the metal is dispersed and stabilized, which shapes activity and selectivity.

But as Luca pointed out, a high surface area doesn’t automatically make a good catalyst. Molecules still must reach the active sites. Micropores (under 2 nm) offer very high surface area but can limit diffusion. Mesopores (2–50 nm) balance surface area with free movement. Macropores (over 50 nm) act as highways, carrying reactants into smaller pores.

From physical to chemical descriptors

Physisorption measures a catalyst’s surface area and characterizes its pore size and pore volume distributions, while mercury intrusion porosimetry extends pore characterization to larger pores.

Chemisorption takes it a step further, using probe molecules that bond selectively to exposed metal sites to count how many are accessible, which gives the metal dispersion. But as Luca cautioned, accessible doesn’t always mean active.

Temperature-programmed techniques reveal how a catalyst’s surface responds as temperature increases. Temperature-programmed desorption (TPD) examines how strongly molecules bind to the surface and helps characterize acid-base sites. Temperature-programmed reduction (TPR) evaluates metal reducibility and helps identify activation temperatures. Temperature-programmed oxidation (TPO) examines oxidation behavior and can inform regeneration by assessing the removal of carbon deposits. Together, these measurements help connect a catalyst’s surface properties to its performance.

New frontiers in catalyst development

The next generation of catalysts will do more with less metal. Single- and dual-atom catalysts push metal dispersion toward its theoretical 100%, aiming to expose every metal atom on the support. Base metals such as iron, cobalt, nickel, and copper reduce reliance on precious metals like platinum, palladium, and rhodium. Luca also pointed to sustainable chemistry as a major frontier, from CO₂ conversion and green hydrogen to biomass and plastic recycling.

Discovery is speeding up too. AI is accelerating density functional theory (DFT) simulations that can otherwise take days or weeks, alongside high-throughput testing and automated characterization. But no single technique gives the full answer. The value comes from combining surface descriptors with structural and spectroscopic techniques and microscopy to build a complete picture of how a catalyst works.

Key questions from the Q&A

Is trial and error in catalyst design obsolete?

Not according to Luca. AI’s output still has to be filtered by experts, he said, and every catalyst still must be tested in pilot plants under operando conditions, where errors can happen. In his view, rational design and trial and error work in combination.

In 10 years, will AI design better catalysts than humans?

“In 10 months,” Luca countered, pointing to how wary even AI developers are of the technology’s pace. Dr. Matthew Rayner of Ketjen was more direct: anyone not using AI in chemical development risks falling behind. For those who haven’t started, he said, it’s not too late, but there’s catching up to do.

If a young researcher or engineer wanted to have a real impact on this field over the next decade, where should they focus?

Luca’s answer was pragmatic: follow the job possibilities. Prof. Dr.-Ing. Christian Seidel of Wohlers Associates, powered by ASTM International, offered a different perspective: follow your heart, since no one can predict what will be in vogue in 10 years.

Watch the recording of this session to find out how catalysis helps recycle astronauts’ CO₂ on the International Space Station, and more from Future Days: Catalysts.