Parma, July 14, 2026 – A new study by the University of Parma, published in the prestigious scientific journal *Nature Communications*, opens up new prospects for developing safer, more sustainable, and more efficient chemical processes. The research demonstrates how integrating artificial intelligence with mechanochemistry can help rethink some of the most widely used transformations in organic synthesis, making them easier to execute and more broadly applicable.
The publication in *Nature Communications* underscores the scientific significance of the findings and highlights the University of Parma's role in the international landscape of innovative chemical research. The project was coordinated by Luca Capaldo (Department of Chemical, Life, and Environmental Sustainability Sciences) for the chemistry component, and by Andrea Prati (Department of Engineering and Architecture) for the machine learning component. Mechanistic experiments at the synchrotron were conducted by Paolo P. Mazzeo (Department of Chemical, Life, and Environmental Sustainability Sciences).
At the heart of the study is the Johnson-Corey-Chaykovsky reaction, widely used to produce cyclopropanes and epoxides—molecules employed in the manufacture of bioactive compounds, functional materials, and pharmaceutical intermediates.
Traditionally, this transformation is carried out using procedures developed decades ago; while effective, these methods present significant drawbacks regarding safety and sustainability. To overcome these limitations, the research team combined two innovative approaches: Bayesian optimization—a machine learning technique capable of rapidly identifying the most effective experimental conditions—and mechanochemistry, which harnesses mechanical energy to drive chemical reactions while reducing or eliminating the use of solvents. The developed approach has enabled the establishment of a new solvent-free synthesis protocol based on the use of potassium hydroxide—an inexpensive, readily available reagent that is safer than those traditionally employed. The method proved effective even under mild conditions and without the need for a strictly inert atmosphere, making it particularly promising in terms of operational simplicity and potential scalability.
The methodology was also validated across a wide range of substrates, enabling the preparation of numerous cyclopropanes and epoxides, and demonstrated good scalability—a crucial requirement for the future transfer of results from research to industrial application.
The study also provided insight into the reaction mechanism through time-resolved X-ray powder diffraction experiments conducted during the mechanochemical process. Analyses performed using laser-based experiments highlighted the crucial role of active milling in driving the chemical transformation, yielding valuable information on how the reaction proceeds under solvent-free conditions.
Overall, the work demonstrates how the integration of artificial intelligence, experimental automation, and sustainable technologies can accelerate the development of chemical processes that are safer and aligned with the needs of modern industry. The study thus represents a significant contribution to sustainable chemistry and confirms the growing role of mechanochemistry and machine learning in the future of chemistry.
Link to the publication
