LLM-Driven Synthesis Planning for Quantum Dot Materials Development.

Area of Science:

• Materials Science
• Quantum Dot Synthesis
• Artificial Intelligence in Chemistry

Background:

• Large language models (LLMs) are increasingly applied in materials science to accelerate discovery and development.
• Optimizing experimental procedures for synthesizing materials with multiple desired properties remains a challenge.

Purpose of the Study:

• To propose a novel framework using LLMs to optimize experimental procedures for synthesizing quantum dot materials with multiple desired properties.
• To integrate synthesis protocol generation and property prediction models fine-tuned on open-source LLMs.

Main Methods:

• Fine-tuning open-source LLMs using parameter-efficient training techniques with in-house synthesis protocol data.
• Integrating a synthesis protocol generation model and a property prediction model.
• Validating generated protocols through property prediction, novelty assessment, and human evaluation.

Main Results:

• Three out of six generated synthesis protocols successfully updated the Pareto front.
• All six protocols improved at least one material property.
• Empirical validation confirmed the framework's effectiveness for synthesis planning.

Conclusions:

• The fine-tuned LLM-driven framework is effective for multitarget optimization in materials synthesis.
• This approach accelerates the development of quantum dot materials with desired properties.
• The framework demonstrates strong performance in optimizing complex synthesis procedures.