Professor Li Weixue from the University of Science and Technology of China has a folder on her computer that contains 329 different versions of the same paper. Each version is named by date, with the earliest version being September 5, 2018 and the last version being October 20, 2024. Recently, this paper that solves the nearly 50 year old problem in the field of heterogeneous catalysis was published online in Science. Li Weixue's team utilized artificial intelligence (AI) technology to reveal the essence of the "metal support interaction" in supported metal catalysts. Li Yadong, an academician of the CAS Member and professor of Tsinghua University, believes that this achievement "solves a major basic scientific problem in the research of heterogeneous catalysis, and has great guiding value for the rational design of highly efficient loaded catalysts". More than 90% of chemicals are synthesized and prepared with the help of catalysts in a very "elegant" way, and oxide supported metal catalysts are one of the most widely used industrial catalysts. For a long time, scientists have been striving to develop catalysts with high activity, selectivity, and stability. There are two cornerstones in a supported catalyst system - the relationship between the catalyst and reactants and the relationship between the catalyst and the support. The former is the key to improving catalytic activity and selectivity, while the latter plays the role of a 'stabilizing sea needle' The corresponding author of the paper, Li Weixue, said. Early research mostly focused on the interaction between metal catalysts and reactants, while neglecting the interaction between metal catalysts and supports. With the continuous deepening of research, it has been found that oxide supports not only play a role in stabilizing metal catalysts, but also further affect the charge transfer, morphology changes, formation of new interface sites, chemical composition, and interface coating in the catalyst system, thereby significantly affecting the activity and selectivity of the catalyst. In fact, as early as 1978, scientists discovered that oxide carriers exhibited the phenomenon of encapsulating metal catalysts in high-temperature reduction environments. This phenomenon is attributed to the strong metal carrier interaction. Later, people used this concept to explain all experimental phenomena that exhibited significant interface effects. There are too many and complex factors that affect the interaction between metals and supports, involving the composition, structure, size, morphology, etc. of catalysts and supports. This effect is also sensitive to the preparation process and reaction conditions of the catalyst Li Weixue said that currently, there are nearly 7000 papers on the concept of metal carrier interaction every year, and it is steadily increasing. However, the research on this interaction has not been quantified, especially the lack of clear structure-activity relationships. For this reason, starting from 2017, he led students to use AI technology to "tackle" the problem, and after 8 years, finally solved this nearly 50 year old problem in the field of heterogeneous catalysis. What surprised us was that among many complex factors, it was the metal metal bond formed between the metal catalyst and the metal in the support that had a decisive impact, rather than the metal oxygen bond that everyone had long imagined Li Weixue said that this provides a new perspective for understanding the interaction between metal and carrier. The reviewer of Science highly praised this work: "This study is very important for improving industrial catalysts, and I congratulate the author for starting from the atomic level and solving this problem in a 'beautiful' way! The research results are highly creative, thoughtful, and profound." In this study, researchers first collected 178 experimental data of 25 metals and 27 oxides, and then used an interpretable AI algorithm to construct a candidate space composed of 30 billion expressions, and based on this, established a machine learning formula with clear physical significance. Interpretable AI algorithms start from the basic properties of materials, iteratively combine known physical and chemical parameters of materials through mathematical combinations, and obtain 30 billion expressions Li Weixue said that they further derived an equation that can reproduce all experimental data from 30 billion expressions based on the principle of compressive sensing. It is relatively simple to derive a machine learning formula, but the difficulty lies in how to combine specific problems to make the formula highly interpretable and extract physical meanings and scientific principles from it Li Weixue said that this depends on the researcher's understanding and judgment of the problem. In the end, they combined domain knowledge and theoretical deduction to establish a physically clear and numerically accurate concise equation, which for the first time fully revealed the two key physical quantities that affect the metal carrier interaction, namely "metal oxygen interaction" and "metal metal interaction". After obtaining the formula, they repeatedly confirmed and verified it, analyzed 675 metal oxide systems, and found that "metal oxygen interaction" was the main "contributor" to the formation of interfaces, while "metal metal interaction" was the key factor in distinguishing the influence of different carriers. Simply put, all oxide carriers contain oxygen, but they contain different metals. Therefore, it is evident that the 'metal metal interaction' has a decisive impact on the carrier effect Li Weixue sighed, it took nearly 50 years for people to pierce through this layer of "window paper". It is worth mentioning that they also proposed the principle criterion of "strong metal metal interaction", which means that when the interaction between two metals is stronger than the interaction between the metals themselves in the oxide, the oxide support will encapsulate the metal catalyst. For example, the metal catalyst is platinum and the carrier is titanium oxide. When the platinum titanium bond is stronger than the titanium titanium bond, the phenomenon of titanium oxide coating platinum catalyst occurs. This concise criterion effectively explains almost all the encapsulation phenomena observed in such systems currently. A relay race spanning time and space, Li Weixue conducted postdoctoral research at the Fritz Haber Institute of the Max Planck Society in Germany. After returning to China, she established the Sino German Max Planck Partner Research Group with the institute's director Matthias Scheffler. In 2016, Scheffler was invited to attend an academic conference at the University of Science and Technology of China. When meeting with Li Weixue, Scheffler praised his postdoctoral fellow, Ouyang Runhai, who was previously a doctoral student of Li Weixue. He developed an interpretable AI algorithm called SISSO, which has broad prospects in the field of materials research. This inspired Li Weixue. In 2017, he arranged for Wang Tairan, who had just joined the research group as a graduate student, to use SISSO to study the interaction between metal and support. Wang Tairan collected a large amount of high-quality interface interaction data from hundreds of literature and established corresponding formulas using SISSO. However, the physical significance and significant value behind the formula still need to be further explored. In 2021, this paper had already taken shape, but Li Weixue was not in a hurry to publish it. Firstly, the physical image of the equation obtained at that time was not very clear, and secondly, the importance of the new metal metal interaction variables in the catalytic process was not yet clear. Li Weixue led a team to conduct extensive exploration around the above issues, using AI based equations to predict and calculate various possible physical quantities, and attempting to link them with important catalytic problems. The article also repeatedly started over, but the results were still unsatisfactory. The turnaround will occur in 2023. PhD student Hu Jianyu successfully reproduced the encapsulation phenomenon discovered in experiments in 1978 using molecular dynamics simulations based on neural network potential functions. This time, they found a breakthrough in the application of theory. With repeated revisions to the innovation, importance, logic, and clarity of the article, their understanding of the physical meaning of formulas has become increasingly clear, profound, and comprehensive. The research team officially submitted their paper to Science on July 30, 2024, in just 85 days from submission to acceptance. This is the 320th version of the paper. On September 7th, they received the review comments. To Li Weixue's surprise, the editor has thoroughly revised the paper. This indicates that the paper has been basically accepted. According to the review comments, the research team completed the corresponding work and returned the manuscript on October 2nd. Soon, on October 22nd, the paper was officially accepted. The research took 8 years and the paper was revised 329 times. It only took 85 days from submission to acceptance for publication... Behind the slow and fast progress is Li Weixue's dedication to scientific research and the relay of three students crossing time and space. Once this theoretical work was published, it immediately attracted the attention of researchers in the same field. Currently, two experimental research groups have utilized the theory proposed this time to synthesize a series of new coated catalytic material systems, and the subsequent research on new catalytic reactions is also rapidly advancing, "said Li Weixue. At present, we have only described the interaction between metal and oxide interfaces. Next, we will continue to study the interactions between metals and various metal compound carriers Li Weixue said that their long-term goal is to establish a general theory describing material interface interactions, accelerate the discovery of new catalytic materials and reactions, and promote green upgrading and sustainable development in the fields of energy, environment, and materials. Li Weixue stated that this scientific breakthrough also demonstrates the enormous potential of interpretable AI algorithms, providing a new perspective and solution for solving major scientific problems. (New Society)