Scholarlink

<p>Generative artificial intelligence (AI), particularly large language models (LLMs) such as ChatGPT and DeepSeek, is gaining rapid development and providing powerful tools for chemical and material discovery, even for researchers without expertise in programming. However, the complexity of chemical mechanisms and sparse data characteristics impose difficulties in harnessing LLMs in specific materials experimental research. In this study, we propose a chained LLM-human interaction study to utilize GPT-4 in professional chemistry/material research. Specifically, we build an interactive workflow to break down the task into several steps to propose and scrutinize electrolyte materials for four-electron zinc-iodine (Zn-I<sub>2</sub>) batteries, an emerging branch of secondary batteries for energy storage. In the interactive study, GPT-4 retrieves key literature information for human comprehension, recommends promising electrolyte materials for experimental evaluation, and rates the potential materials with humans for battery application. Through the chained LLM-Human interaction, we not only summarize the mechanism and find applicable materials for multi-electron zinc-iodine batteries, enhancing the performance of four-electron Zn-I<sub>2</sub> batteries (more than 1000 cycles with high Coulombic efficiency) but also establish a feasible approach to integrating LLMs in professional experimental research, which could further inspire and accelerate chemistry and materials science.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Chained LLM-human interactive framework for electrolyte design in four-electron Zn-I2 batteries

  • 期刊论文

  • 发表日期：

    2025年7月31日

查看全文
  • Zihui Chen,
  • Jinglang Zhang,
  • Guanhua Zhao,
  • Qilong Wang,
  • Chunpeng Yang,
  • Yu-Guo Guo

摘要

Generative artificial intelligence (AI), particularly large language models (LLMs) such as ChatGPT and DeepSeek, is gaining rapid development and providing powerful tools for chemical and material discovery, even for researchers without expertise in programming. However, the complexity of chemical mechanisms and sparse data characteristics impose difficulties in harnessing LLMs in specific materials experimental research. In this study, we propose a chained LLM-human interaction study to utilize GPT-4 in professional chemistry/material research. Specifically, we build an interactive workflow to break down the task into several steps to propose and scrutinize electrolyte materials for four-electron zinc-iodine (Zn-I2) batteries, an emerging branch of secondary batteries for energy storage. In the interactive study, GPT-4 retrieves key literature information for human comprehension, recommends promising electrolyte materials for experimental evaluation, and rates the potential materials with humans for battery application. Through the chained LLM-Human interaction, we not only summarize the mechanism and find applicable materials for multi-electron zinc-iodine batteries, enhancing the performance of four-electron Zn-I2 batteries (more than 1000 cycles with high Coulombic efficiency) but also establish a feasible approach to integrating LLMs in professional experimental research, which could further inspire and accelerate chemistry and materials science.