According to scientists, they have created a “self-driving” lab that provides in-depth studies of catalytic reactions utilised in chemical research and pharmaceutical manufacturing using automated technologies and artificial intelligence (AI). The researchers assert that the new tool, named Fast-Cat, can deliver more information in five days than is feasible in six months of conventional testing, adding that the yield and selectivity of chemical processes in the presence of ligands are at issue.
The research team released their paper in Nature Chemical Engineering titled “Autonomous Reaction Pareto-Front Mapping with a Self-driving Catalysis Lab.”
“Ligands are essential for facilitating difficult chemical transformations mediated by homogenous catalysts that contain transition metals. The researchers state that “ligand development and discovery have proven to be a challenging and resource-intensive undertaking, despite their undeniable role in homogeneous catalysis.”
Self-driving catalysis “In response, we are introducing Fast-Cat, a self-driving catalysis laboratory that enables Pareto-front mapping of high-temperature, high-pressure gas-liquid reactions as well as autonomous and resource-efficient parameter space navigation. With the least amount of human intervention, Fast-Cat facilitates multi-objective catalyst performance measurement and autonomous ligand benchmarking. To be more precise, we use Fast-Cat to quickly identify using Pareto-front the hydroformylation process that occurs between syngas (CO and H2) and olefin (1-octene) when rhodium and other phosphorus-based ligands are present.
We illustrate the scalability of Fast-Cat’s knowledge through reactor benchmarking, which is crucial for the fine and specialty chemical industries. We describe the specifics of Fast-Cat’s modular flow chemistry platform and its autonomous experiment-selection methodology for quickly producing the ideal experimental setups and internal data needed to provide machine-learning methods for reaction and ligand research.
From an industry standpoint, you want selectivity and yield to be as high as feasible. Industrial chemists spend a great deal of time and energy attempting to determine the parameters required to get the most desired reaction outcome since the precise actions you take throughout the catalytic reaction can affect both yield and sensitivity.
The issue lies in the labour-, time-, and material-intensive nature of traditional catalyst discovery and development methods, according to Milad Abolhasani, PhD, the paper’s corresponding author and an associate professor of chemical and biomolecular engineering at North Carolina State University. These methods mostly rely on human intuition and expertise to guide the experimental planning, together with manual sample handling in batch reactors. This human-dependent method to catalyst creation not only results in material inefficiencies but also in significant delays in reaction performance, product characterization, and experiment design decisions.
For this reason, we developed Fast-Cat. It used to take six months to fully understand how a particular ligand functions, but today it just takes five days.
Total independence
According to Abolhasani, “Fast-Cat is fully autonomous, running high-temperature, high-pressure, gas-liquid reactions continuously using AI and automated systems.” Additionally, the autonomous technology examines the results of each of these reactions to ascertain, devoid of human involvement, how various factors influence the results of every trial.
Which experiment Fast-Cat runs next is determined by the outcomes of all the prior experiments it has conducted, both successful and unsuccessful.
According to Abolhasani, “Fast-Cat’s AI is constantly evolving, learning from its previous experiments.”
After providing Fast-Cat with the ligands and precursor chemicals it needs to begin with, users may watch as the system learns over the course of 60 experiments.
Abolhasani adds, “We spent a lot of work optimising Fast-Cat’s AI model to maximise its capacity to offer the most comprehensive understanding of how various parameters affect the selectivity and yield of catalytic events employing a particular ligand. We also took a long time to make sure that the results of Fast-Cat are scalable. Fast-Cat uses incredibly small sample sizes for its tests. However, we needed to be certain that Fast-Cat’s results held true for reactions carried out on the massive sizes that are significant for industrial manufacture if we wanted its conclusions to have practical use.
Abolhasani (abolhasani@ncsu.edu) notes that the researchers have made the hardware and software publicly available in order to use Fast-Cat to support further research. They hope that “other researchers can adopt this technology to accelerate catalysis discovery in academia and industry.”
