Finding the right homogeneous catalyst is one of the slowest jobs in chemistry. For a reaction like rhodium-catalyzed hydroformylation, the result depends on choosing the right ligand and dialing in temperature, pressure, gas composition, and concentrations all at the same time. That search space is huge, the gases involved are toxic and flammable under pressure, and the usual trial-and-error approach eats up time, money, and material. Research teams at North Carolina State University, UNC Chapel Hill, and Eastman Chemical Company decided to hand that job to a machine, built on Chemspeed automation and driven by Chemspeed software that runs the show end to end, sequencing every step and capturing every result in real time.
- The system is a closed-loop autonomous platform that prepares catalysts, runs pressurized reactions, analyzes the products, and picks its own next experiment, with no hands in the loop.
- Across 680 experiments in three campaigns, it steered propylene hydroformylation toward linear, branched, or deliberately tunable aldehyde products.
- Turnover frequency climbed more than 2.5 times, from roughly 2,900 to 7,500 per hour, while the reachable range of product selectivity widened at both ends.
- It uncovered "flexible" ligands that act like chemical dimmer switches, flipping between product types just by changing the conditions on the same catalyst system.
- Winning conditions transferred cleanly from 2 mL discovery vials to a 20 mL format, a tenfold scale step, inside one automated workflow.
This is what catalyst discovery looks like when the lab does the deciding, not just the pipetting.
Authors and organizations: J. A. Bennett, N. Orouji, A. Velayati, and M. Abolhasani (North Carolina State University); Eamon F. Reynolds, Matthew E. Porowski, and Alexander J. M. Miller (University of North Carolina at Chapel Hill); M. Saha and D. Billodeaux (Eastman Chemical Company).
Published in
Nature Communications, 20 June 2026.
https://www.nature.com/articles/s41467-026-74425-x
This post summarizes openly published research and is shared for informational purposes. All rights to the original article remain with its authors and publisher. The work described was carried out by the research teams named above; Chemspeed is proud to see its automation and software support science like this. All referenced trademarks belong to their respective owners.
