34 Examples of LLM Applications in Materials Science and Chemistry: Towards Automation, Assistants, Agents, and Accelerated Scientific Discovery

TL;DR

The paper analyzes 34 team submissions from the second LLM Hackathon for Applications in Materials Science and Chemistry to illustrate how large language models can augment the entire research lifecycle. It categorizes projects into seven areas—property prediction, design, automation, education, data management, hypothesis generation, and knowledge extraction—demonstrating techniques such as retrieval-augmented generation, multi-agent reasoning, and tool-calling interfaces. Key contributions include evidence that domain-informed inputs (e.g., orbital bonding descriptors) improve predictions, iterative AI-assisted material design using ReAct and surrogates, and robust data-management and knowledge-graph frameworks that enhance reproducibility and discoverability. The findings underscore the potential for AI-driven accelerated discovery while highlighting challenges in reliability, interpretability, and reproducibility, and emphasize the value of interdisciplinary collaboration fostered by hackathons.

Abstract

Large Language Models (LLMs) are reshaping many aspects of materials science and chemistry research, enabling advances in molecular property prediction, materials design, scientific automation, knowledge extraction, and more. Recent developments demonstrate that the latest class of models are able to integrate structured and unstructured data, assist in hypothesis generation, and streamline research workflows. To explore the frontier of LLM capabilities across the research lifecycle, we review applications of LLMs through 34 total projects developed during the second annual Large Language Model Hackathon for Applications in Materials Science and Chemistry, a global hybrid event. These projects spanned seven key research areas: (1) molecular and material property prediction, (2) molecular and material design, (3) automation and novel interfaces, (4) scientific communication and education, (5) research data management and automation, (6) hypothesis generation and evaluation, and (7) knowledge extraction and reasoning from the scientific literature. Collectively, these applications illustrate how LLMs serve as versatile predictive models, platforms for rapid prototyping of domain-specific tools, and much more. In particular, improvements in both open source and proprietary LLM performance through the addition of reasoning, additional training data, and new techniques have expanded effectiveness, particularly in low-data environments and interdisciplinary research. As LLMs continue to improve, their integration into scientific workflows presents both new opportunities and new challenges, requiring ongoing exploration, continued refinement, and further research to address reliability, interpretability, and reproducibility.

Figures (13)

• Figure 1: The LLM-Powered Research Constellation. At each stage of the research process, from initial ideation through experimental execution and communication of results, LLMs provide a constellation of capabilities spanning hypothesis generation, property prediction, novel interfaces, education, material design, automation, data management, scientific communication, and more. This constellation demonstrates the possibility of LLMs and multimodal models to drive a more efficient, rapid, and creative scientific discovery process through integrations across the research lifecycle.
• Figure 2: Schematic depicting the prompt for fine-tuning the LLM with Alpaca prompt format.
• Figure 3: Workflow overview. The ReAct agent looks up guidelines for designing low band gap MOFs from research papers and suggests a new MOF (likely with a lower band gap). It then checks the validity of the new SMILES candidate and predicts the band gap with epistemic uncertainty estimation using an ensemble of surrogate fine-tuned MOFormers. b. Band gap predictions for new MOF candidates as a function of agent iterations
• Figure 4: LangSim framework for atomistic simulation and inverse design. Custom atomistic modeling tools (such as pyiron, ASE python package functions with underlying EMT and MACE-MP-0 forcefields) are integrated using LangChain @tool decorator. Pydantic model is used to exchange atomic information in a structured format between LLM and tools. The emerging agentic capability for inverse alloy design is demonstrated. LLM agent is able to find the target composition of Cu-Au alloy with the desired bulk modulus.
• Figure 5: Schematic overview of the LLMicroscopilot assistant. The microscope user interface allows the user to input queries, which are then processed by the LLM. The LLM executes appropriate tools to provide domain-specific knowledge, support data analysis, or operate the microscope.