Nanyang Technological University and researchers of their collaborators successfully used the power of Chat-GPT to simplify text analysis for solid synthesis, focusing on the triad pepper agent. This innovative method is designed to optimize the synthesis of high -quality crystal materials, which is the key to advancing thermal power equipment. This study was led by Dr. Kedar Hippalgaonkar of Nanyang Technology University, Dr. Maung Thway, Mr. Andre Low, Dr. Jose Rocatala-Gomez and Dr. Andy Cheny. ANG Technogical University, Samyak at the Indian Institute of Technology in Mumbai Mr. Khetan was published in the “Digital Discovery” magazine.
Solid -state synthesis is the key method of discovering new inorganic materials, especially the inorganic materials used in thermoelectric applications to convert heat into electricity. The traditional data -driven comprehensive method requires a trace of manual extraction and synthetic recipes in the huge text. This process is not only time -consuming, but also proposed a high obstacle, especially for the material of literary sparseness.
In order to cope with these challenges, the team recommends using large language models (LLM) (such as GPT-3.5) in Chat-GPT to analyze comprehensive recipes, and intuitively capture basic synthetic information based on primary and secondary heat-ups. By developing a data set (gold standard) planned by experts, they designed a prompt for Chat-GPT to accurately copy this data set (silver standard).
The focus of the study is the composition of the three yuan grapes, such as CUINTE/SE, which is known for its thermoelectric characteristics at the temperature. From the database of the research papers, Chat-GPT successfully parsed a part, and then used it to develop a classifier to predict the purity of the phase. This method proves the universality of LLMS to text analysis, providing a potential change paradigm in the synthesis and representation of new materials.
Dr. Hippalgaonkar emphasized the importance of their work and pointed out: “Our method provides a roadmap for future efforts to seek the combination of mergers with materials scientific research, and provides potential potential in the synthesis and representation of novel materials. Transformation paradigm.
The researchers carefully extracted the data from the papers published between 2000 and 2023. The focus is on Cuinte/SE, which excluded methods such as solution synthesis and Bridgman methods. They have determined the key aspects of the most important aspect of obtaining a purified compound: primary heating, secondary heating, annealing and density. The prompt is optimized to ensure that the relevant synthetic details are extracted in a structured format.
The extraction data allows a comprehensive analysis of the synthetic conditions, revealing the sub -heating, annealing and primary heating significantly affect the purity of the phase. Their decision -making tree classifier proves the potential of the use of machine learning based on the comprehensive results of the text share data.
Dr. Hippalgaonkar said: “The data in solid -state synthesis may be biased towards a positive formula, and the balanced data set is necessary for pushing the scene to advance.” The generality of text analysis, especially for literary sparse materials. ” Their work also proves the potential of Chat-GPT insertion and inferring the synthetic conditions of similar materials, which shows the practical method of synthetic new compounds.
This study emphasizes the importance of combining advanced AI tools with traditional materials science methods, which pave the way for more efficient and more accurate synthesis. Dr. Hippalgaonkar and his team successfully opened a new way on Chat-GPT to use LLM in scientific research, especially in the fields of limited literature and complex data extraction needs.
Journal reference
Maung Thwey, Andre Ky L Ow, Samyak Khetan, Haiwen Dai, Jose Recatala-Gomez, Andy Paul Chen, and Kedar Hippalgaonkar. “Using GPT-3.5 for text analysis in solid-state synthesis-case research on the ternary gene generated by the ternary gene.” Digital found that 2024. Doi: https: //doi.org/10.1039/d3dd00202k
About the author
Associate Professor of Kedar HippalgaOnkar It is the joint appointment of the Department of Materials Science and Engineering of NRF Researcher (2021) and the Materials Science and Engineering Department of NANYANG Technological University (NTU), and is a senior scientist research (*star) of the Institute of Materials Research and Engineering (IMRE) of science, technology and science and technology institutions (*Star). Essence Since 2018-2023, he has led the Accelerated Materials Development (AMDM) program, focusing on introducing new materials, process and optimization development of machine learning, AI and high -throughput computing, and electronics and plasma materials and polymers. He also led the Hybrid (inorganic organic) thermoelectric program in 2016-2020. He published more than 70 research papers and co -founded a startup (Xinterra, INC.). He won the starting award of the Ministry of Education in 2021 and was nominated for material chemistry magazines in 2019. Young leaders with social technology in Kyoto in 2015. His outstanding graduate studies won the silver medal of the Materials Research Association in 2014. Through A*Star National Science scholarship funding, he received a bachelor’s degree from Divellive in 2003. The Department of Mechanical Engineering of Purdue University obtained a PhD in philosophy from the Department of Mechanical Engineering at the University of California Berkeley in 2014. During a doctorate research, he studied the basic principles of thermal, charging and light of solid materials.
Dr. Maung Thweay It is a teaching analysis and application researcher at the student (Atlas) students (ATLAS). His research involves the impact of Gen-AI application in college learning. Earlier, he was a researcher at the School of Materials Science and Engineering, the School of Materials Science and Engineering, a associate professor, where he developed the method of accelerating materials. He obtained a Ph.D. in Electrical Engineering, Singapore, Singapore, Singapore in 2020. His research during his doctoral degree includes the manufacture and integration of Perovskite/Si and IIII-V/SI TANDEM solar cells.
Andre Ky low It is a graduate student at the Department of Materials Science and Engineering of Nanyang Technological University, Singapore, and is supervised by Associate Professor Kedar Hippalgaonkar. His argument is to develop and apply a multi -target optimization algorithm to accelerate the discovery of materials. Andre has won the A*Star Graduate Scholarship, which is affiliated to the Institute of Materials Research and Engineering. Andre has previously obtained a bachelor’s degree in materials science and engineering from Nanyang Technology University, becoming a speaker of graduation class in 2021.
Joserecatalàgómez He is a researcher at the Department of Materials Science and Engineering of Nanyang Technology University in Singapore, and works at Associate Professor of Kedar Hippalgaonkar. He specifically integrated AI and machine learning with high -throughput solids to discover inorganic materials for energy and environmental applications. Jose obtained a bachelor’s degree in chemistry at the Spanish University in 2015. In 2016, he was a master’s degree from the University of Madrid, Spain, and in 2021, he received a doctorate degree from the University of Southampton, England. *Star -level Research Appendix Plan (ARAP) scholarships, and spent two years at the Institute of Materials Research and Engineering (IMRE) at Singapore.
