[論文レビュー] El Agente Sólido: A New Age(nt) for Solid State Simulations

Sai Govind Hari Kumar, Yunheng Zou+9 more|arXiv (Cornell University)|Feb 19, 2026

Machine Learning in Materials Science被引用数 0

ひとこと要約

Paper introduces El Agente Sólido, a hierarchical multi-agent framework that autonomously conducts solid-state quantum chemistry workflows with Quantum ESPRESSO, achieving high reproducibility across multiple benchmark iterations.

ABSTRACT

Quantum chemistry calculations are a key component of the materials discovery process. The results from first-principles explorations enable the prediction of material properties prior to experimental validation. Despite their impact, the practical use of first-principles methods remains limited by the expertise required to design, execute, and troubleshoot complex computational workflows. Even when workflows are successfully built, they are sometimes rigid and not adaptable to different use cases. Recent advances in large language models (LLMs) and agentic systems offer a pathway to flexibly automate these processes and lower barriers to entry. Here, we introduce El Agente Sólido, a hierarchical multi-agent framework for automating solid-state quantum chemistry workflows using the open-source Quantum ESPRESSO simulation package. The framework translates high-level scientific objectives expressed in natural language into end-to-end computational pipelines that include structure generation, input file construction, workflow execution, and post-processing analysis. El Agente Sólido integrates density functional theory with phonon calculations and machine-learning interatomic potentials to enable efficient and physically consistent simulations. Extensive benchmarking and case studies demonstrate that El Agente Sólido reliably executes a wide range of solid-state calculations, highlighting its potential to improve reproducibility and accelerate computational materials discovery

研究の動機と目的

• Motivate the need for accessible first-principles solid-state simulations to accelerate materials discovery.
• Develop an autonomous, language-driven framework that translates natural-language objectives into end-to-end QE workflows.
• Demonstrate reproducible, multi-trial performance across diverse benchmarking tasks and case studies.
• Showcase capabilities in structure generation, DFT/phonon calculations, MLIPs, and materials discovery case studies.

提案手法

• Propose a hierarchical multi-agent architecture that orchestrates QE-based workflows.
• Integrate structure generation, input-file creation, DFT/Phonopy calculations, and result analysis into autonomous agents.
• Use pretrained ML interatomic potentials (MLIPs) to pre-relax structures and accelerate ab initio simulations.
• Benchmark across 7 foundational exercises plus 4 case studies with 10 repetitions each.
• Evaluate performance with a rubric focusing on workflow setup and final results, achieving high procedural scores and reproducibility.

実験結果

リサーチクエスチョン

• RQ1 Can El Agente Sólido autonomously plan, execute, and troubleshoot complete solid-state QE workflows from natural-language prompts?
• RQ2 How reproducible and robust are its results across multiple trials and varying prompt difficulties?
• RQ3 What is the framework’s performance in diverse materials tasks, including batteries, catalysis, MOFs/COFs, and phonon analyses?
• RQ4 How effectively does the system integrate MLIPs and Phonopy to enable scalable, physically consistent simulations?

主な発見

• El Agente Sólido achieved an average score of 97.9% across 7 benchmark exercises repeated 10 times each.
• The framework autonomously queries databases, generates structures (including vacancies, supercells, slabs, SQS, MOFs/COFs), and selects computational parameters for each task.
• It performs DFT workflows with structural relaxations, self-consistent and phonon calculations, band structures, DOS, stability analyses, and adsorption studies.
• Four case studies demonstrate applications in Li-ion batteries, catalysis, and porous materials, plus phonon-based thermal properties via Phonopy.
• Phonon and thermodynamic analyses show physically consistent results, including QHA-based thermal properties.