How efficient is AI chemist? It can complete 6 months of work in 5 days without human intervention

In recent years, to meet the growing needs of industry and science, scientists in the field of chemistry have been seeking to improve the efficiency and selectivity of catalytic reactions.

Catalytic reactions are at the core of many chemical processes, and a key factor is the use of ligands to tune the selectivity and yield of reactions. However, traditional catalyst discovery and optimization methods are usually time-consuming, material-intensive , and heavily dependent on manual operations and experience .

To solve this problem, a research team from North Carolina State University and Eastman Chemical Company developed an automated laboratory called Fast-Cat .

It is reported that by combining artificial intelligence (AI) and automation technology, Fast-Cat achieves fast, efficient and automated catalytic reactions. It can not only run high-temperature, high-pressure, gas-liquid reactions completely autonomously and continuously , but also analyze the output results of each reaction and determine the impact of different variables on each experimental result without human intervention.

The related research paper, titled “Autonomous reaction Pareto-front mapping with a self-driving catalysis laboratory”, has been published in the scientific journal Nature Chemical Engineering.

It is worth noting that it can provide more information in just 5 days than the traditional method of 6 months, providing new possibilities for chemical research and industrial production . The research team said that the emergence of Fast-Cat marks a new era in catalytic reaction research.

How is Fast-Cat faster than humans?

According to the paper, Fast-Cat uses an automated experimental system and AI-driven experimental planning to make the experimental process of catalytic reactions fully automated and intelligent. Its working principle is: based on a deep understanding of the impact of various variables in catalytic reactions, and using AI algorithms to continuously learn and optimize experimental design, it can quickly find the optimal reaction conditions.

Figure | Overview of Fast-Cat workflow. Input: Prior data, constraints, and optimization objective. Loop: Determine current hypervolume (HV), Monte Carlo (MC) sample ML model at new points, rank points based on hypervolume obtained by prediction (green star at point 1 indicates prediction with greatest HV improvement, followed by points 2 and 3 in order), experiment on best prediction and repeat. Output: State-of-the-art surrogate models (surface plots and feature analysis) and experimental data points (Pareto front).

The Fast-Cat operation process mainly includes four steps : preparation, startup, operation and Pareto screening cycle.

In the preparation stage, researchers need to prepare the reagents and ligands required for the experiment in advance, including catalysts, ligands, substrates, etc., and load them into the system. These reagents will be loaded into the automated reagent supply module to ensure that there is always enough reagent supply in the experiment.

The startup phase is the beginning of the experiment. The system automatically adjusts the experimental conditions, including the flow rate, pressure and other parameters of liquid and gas, to achieve the required reaction pressure and composition.

During the operation phase, the system continuously runs high-temperature, high-pressure, gas-liquid phase catalytic reactions, and automatically collects and analyzes reaction products.

When Fast-Cat is started, it automatically runs high-temperature, high-pressure, gas-liquid phase catalytic reactions. These reactions are usually performed under flow conditions to ensure uniform reaction mixing and obtain accurate reaction data. The system continuously performs a series of experiments according to a preset experimental plan, quickly collecting a large amount of reaction data.

After each experiment, Fast-Cat automatically collects the reaction products and analyzes them using online analytical equipment such as gas chromatographs (GC). The analysis results include product type, yield, selectivity, etc.

The experimental data will then be uploaded to Fast-Cat’s data analysis module for real-time processing and analysis. Through machine learning algorithms, Fast-Cat can extract regularities and patterns from a large amount of experimental data and adjust the conditions for the next round of experiments based on this information.

To further optimize the reaction conditions and improve the catalytic efficiency , Fast-Cat will adjust the conditions for the next round of experiments based on the results of each experiment. This cyclic feedback mechanism enables Fast-Cat to gradually find the optimal reaction conditions and achieve rapid optimization of the catalytic reaction, which belongs to the Pareto screening cycle stage.

According to the Pareto optimization principle, Fast-Cat will find the best balance between multiple objectives. For example, increasing product yield may reduce selectivity, so a trade-off needs to be made between these two objectives.

Fast-Cat adjusts the experimental conditions based on the experimental results to find the optimal solution between multiple objectives. This may require multiple rounds of experiments and optimization processes.

During the entire experiment, to ensure the safety and stability of the experiment, Fast-Cat will automatically monitor various parameters, including temperature, pressure, flow, etc. If necessary, the system will automatically replenish reagents and maintain equipment to ensure the continuity and stability of the experiment.
It is reported that Fast-Cat has achieved remarkable results during the research process.

By comprehensively testing and analyzing the catalytic properties of different ligands, the researchers discovered how various experimental conditions affect reaction yield and selectivity.

By optimizing the ligand structure and reaction conditions, Fast-Cat successfully improved the efficiency and selectivity of the catalytic reaction, providing new ideas and methods for research and application in the field of catalysis.

Although Fast-Cat has achieved remarkable results in catalytic reaction research, some limitations and challenges still exist.

For example, the system may be limited by experimental conditions and cannot cover all possible reaction situations; for some complex catalytic reaction systems, the intelligence level of Fast-Cat needs to be further improved.

The researchers said that future research directions include further optimizing system design, developing more advanced artificial intelligence algorithms, and expanding the application of Fast-Cat in a wider range of fields.

AI makes chemical experiments more efficient

In recent years, in addition to Fast-Cat, a series of other important research results have emerged in the field of AI-driven catalysis research, such as the workflow of large-scale optimization of catalyst synthesis by artificial intelligence.

In the field of AI4science, many related studies have also shown that AI can improve research efficiency, such as the self-service chemical synthesis robot RoboChem.

The research team said that as an important technological innovation in the field of catalytic reaction research, Fast-Cat has broad application prospects.

For example, in the chemical and pharmaceutical industries , Fast-Cat is expected to become an important tool for catalyst research and development and optimization, providing support for the development of new drugs and new materials; the intelligent and automated features of Fast-Cat also provide new ways and possibilities for achieving green chemical production and energy conservation and emission reduction .

In the future, with the continuous improvement and promotion of Fast-Cat technology, more research results and innovative applications based on Fast-Cat are expected to emerge.

In addition, with a deeper understanding of the mechanisms and performance of catalytic reactions, AI will hopefully help humans develop more efficient and environmentally friendly catalysts and reaction systems, pushing the development of the catalytic field to new heights.

Reference Links:

https://www.nature.com/articles/s44286-024-00033-5

https://news.ncsu.edu/2024/02/ai-driven-lab-speeds-catalysis-research/

https://pubs.acs.org/doi/full/10.1021/acs.iecr.3c02520