Artificial intelligence boosts research and development of new materials

 The development of artificial intelligence is booming. With the continuous extension of its application fields, other disciplines are also gaining unexpected gains from the combination of artificial intelligence, and new materials is one of them.

At present, there have been cases of artificial intelligence assisting the research and development of new materials reported abroad. Researchers at the University of Liverpool in the UK have developed a robot that engineered chemical reactions on its own in eight days, completing 688 experiments and finding a highly efficient catalyst to improve the photocatalytic properties of polymers, an experiment that would have taken months to complete manually. Recently, a professor at Osaka University in Japan used 1,200 kinds of photovoltaic cell materials as a training database to study the relationship between polymer material structure and photoelectroinduction through machine learning algorithm. He successfully selected compounds with potential application value in one minute, while traditional methods take five to six years.

Such a successful application opens up endless possibilities for exploring new materials and technological advances. Throughout human history, every scientific and technological revolution is closely related to the development of materials. Before the Industrial Revolution, the development of stone, bronze and iron tools gradually separated handicraft from hunting and farming. After the first Industrial Revolution, steel and composite materials gradually took over People's Daily life. After the third industrial Revolution, semiconductor, high crystal silicon and polymer materials have developed rapidly and become new materials in great demand. Since the 21st century, with the further improvement of high-end manufacturing industry, new materials have been deeply integrated with emerging industries such as nanotechnology, biotechnology and information technology on the development path of functionalization, intelligence and integration, and become an important means of scientific and technological progress.

The development of new materials is a process of mutual integration of basic research and applied basic research, which often needs to undergo chemical property improvement and physical processing improvement, which is not easy. Take the emerging smart fiber as an example in recent years. This new material can change in volume or form with external environmental stimulation, and can be used to build wearable smart devices. When it is developed, it is necessary to understand its stimulus response mechanism and establish a suitable physical model to explain it. Secondly, appropriate materials should be selected as the research object, the function and properties of its functional units should be improved by chemical means, the stimulus response conditions should be explored through repeated experiments, and the performance of structural units should be improved. Finally, production and processing, after spinning, dyeing and finishing, weaving and other different processing processes, continuous process optimization and technical improvement. Thus, new material research and development is a typical trial and error research and development, experience cycle is often long.

In order to shorten the development cycle, AI can serve as a powerful auxiliary tool to predict and screen the physical and chemical properties of advanced materials through data sharing, thus speeding up the synthesis and production of new materials. In the past, material design was based on theoretical calculations to establish the relationship between structure and properties. But because there are so many different ways atoms can combine, designing a new molecular structure is like a game of building blocks, in which the properties of the molecules are unpredictable. As a branch of artificial intelligence, machine learning algorithm is particularly "effective" in assisting the design of new materials, and its working process mainly includes "descriptor" generation, model construction and verification, material prediction, and experimental verification. The so-called "descriptor" is to describe some special properties of materials according to existing data, and then build a training model in the form of nonlinear, so as to predict the properties of new materials. This process no longer depends on physical knowledge.

Artificial intelligence still faces some challenges in creating more sparks with new materials. For example, it is difficult for AI algorithm to accurately predict crystal structure, and the reliability of training data still needs the development of theoretical methods. In order to give full play to the multiplier effect of interdisciplinary integration, in addition to the continuous improvement of algorithms, the development of theoretical computational chemistry and the research and development of material characterization methods should go hand in hand. In the future, we believe that through the efforts of scientists from all sides, innovative achievements of new materials will continue to emerge.