AI predicts the precursor materials necessary for material synthesis

Researchers in Korea have developed a technology that automatically identifies the precursor materials necessary to synthesize specific objective materials.

A joint research team led by the Senior researcher Gyoung S. NA of the Institute of Chemical Technology of Korea (Krict) and Professor Chanyoung Park of Korea Advanced Institute of Science and Technology (Kaist) have developed a retrosis methodology based on The prediction predicts the required precursor materials are based solely on the chemical formula of the target material without descriptors of expensive material and chemical analysis.

Precursor materials refer to all the essential materials required in the synthesis process of the desired objective material.

In recent years, the discovery of materials has become a crucial task in various industries, including batteries and semiconductors. Traditionally, finding the correct intermediate materials for synthesis has required expensive and repetitive experiments. However, there is a growing demand to use AI to identify these materials efficiently.

Existing technologies based on AI to predict material synthesis processes have focused mainly on organic materials, such as drug compounds, while research on inorganic materials has been relatively insufficient. This is because inorganic compounds, such as metals, have complex structures and diverse elementary compositions, which makes it difficult to determine the synthesis pathways.

The research team developed an innovative AI technology that can learn the reverse process of predicting the necessary precursor materials of the target material using only its chemical formula.

Previously, Kritt developed and transferred the “Chicai” platform in 2022, which allows users to predict synthesis information without advanced programming skills and expensive hardware infrastructure.

This recently developed technology exceeds the challenges raised by the complex 3D structures of inorganic materials, such as atomic arrangements and link information. On the other hand, the AI ​​analyzes the types and relationships of elements present in the objective material and calculates the differences in thermodynamic formation energy to identify precursors that facilitate the easiest synthesis reactions.

To improve the accuracy of precursor materials predictions, the equipment used a deep neuronal network specialized in chemical data. The AI ​​model was trained in approximately 20,000 published research papers that detail synthesis processes of precursor materials and materials.

The AI ​​model was tested in around 2,800 synthesis experiments that were not provided in the training data set. The results of the evaluation showed that the necessary precursor materials in more than 80% of the cases in just 0.01 seconds by using the GPUs.

Looking towards the future, the research team plans to expand the set of training data through Krict research projects to achieve 90%prediction precision. By 2026, its objective is to establish a public service based on the web for the discovery of materials based on AI. Future research will focus on the discovery of fully automated materials that predicts precursor materials and synthesis pathways based solely on the chemical formula of the target material.

The research team emphasized the novelty of its approach, stating: “Unlike conventional precursors prediction models that are limited to specific types of materials, our AI can predict precursor materials universally, regardless of the applications of the materials aim”.

The president of Krict, Young-Kuk Lee, added: “This research is expected to improve the efficiency of new materials in various industries.”

The work is published in it Arxiv Preprint server.