NUST develops full process machine chemist

On October 17, news came from theUniversity of Science and Technology of China (USTC) that a team of ProfessorsLuo Yi and Jiang Jun from the University's School of Chemistry and MaterialsScience has developed a data-intelligence-driven full-process machine chemistby developing and integrating a mobile robot, a chemistry workstation, anintelligent operating system, and a scientific database.

The international reviewers commented thatthis work "will have a huge impact on chemical science". The workbreaks the limits of the traditional trial-and-error research paradigm anddemonstrates the great advantages of the new paradigm of intelligence guided bythe "strongest chemical brain", leading chemical research toward thefuture trend of digital knowledge understanding, operational command andcreation of templates, and establishing China's position as a global leader inthe field of intelligent chemical innovation.

"The objects of chemical research areincreasingly complex and high-dimensional, and the traditional researchparadigm relies mainly on exhaustive and trial-and-error means. Faced with ahuge chemical space, the search for formulations and processes often stops atlocal optimality, and global exploration is not possible." Jiang Jun saidthat the machine chemist platform they developed is loaded with an intelligentplatform of computational brain, theoretical model and open operating system,which can realize the whole process development of chemicalsynthesis-characterization-testing driven by both big data and intelligentmodel, and can be applied to photocatalytic and electrocatalytic materials,luminescent molecules, optical thin film materials and other fields.

Specifically, the platform can use machineintelligence to find and read literature, draw expert experience from massiveresearch data, propose scientific hypotheses and develop experimental protocols.2 mobile robots and 15 intelligent chemistry workstations can complete thewhole process of high-throughput synthesis, characterization and testing ofchemical experiments. Through the back-end operating system, the platform canrealize automatic data acquisition, processing, analysis and visualization;after loading the cloud database, it can also call and update the databaseinformation in real time. The platform's computational brain can integrateintelligent models into the underlying theoretical laws and complex chemicalexperimental evolution by invoking physical models, theoretical calculations,machine learning and Bayesian optimization, making machine scientists moreunderstanding of chemistry and better at chemical creation.

Take the example of catalysts for highentropy compounds with great potential, whose highly disordered mixing ofmultiple elements brings high stability and poses great challenges for manualexperiments to find the optimal ratios. Obtaining the optimal formulationrequires traversing and testing extremely large combinations of chemicalratios, which are currently limited to optimizing up to three metalcombinations. By intelligently reading 16,000 papers and independentlyselecting 5 non-precious elements, the machine chemist combines 20,000 sets oftheoretical calculations and 207 sets of full-process machine experimental datato find the optimal high-entropy catalyst from 550,000 possible metal ratios,reducing the traditional "frying pan" traversal search from 1400years to 5 weeks.

There is still a lot of room to improve theintelligence of the platform, and the scope of application will continue toexpand as the platform is upgraded and expanded," said Jun Jiang. We willadd infrared vision and spectral inversion algorithms to the robot to give itchemical-grade perceptual intelligence and judgment; develop multiple robotcollaboration algorithms and parallel experimental optimization software formultiple smart labs; and build multiple smart labs and connect them to multiplelarge scientific devices in China to form a data-sharing smart chemistry cloud,relying on big data to form more advanced chemical intelligence. "