Agentic Fusion of Large Atomic and Language Models to Accelerate Materials Discovery

2026-04-26

AI Summary

This work addresses the limited autonomy and coordination in traditional materials discovery approaches, which hinder efficient end-to-end integration from design to synthesis. The authors propose ElementsClaw, an intelligent agent framework that, for the first time, synergistically combines a large atomic model (LAM) with a large language model (LLM) within an agent architecture. By leveraging the LLM’s high-level semantic reasoning, a custom Elements-finetuned LAM toolkit, and a dynamic task orchestration mechanism, the framework establishes a closed-loop pipeline that translates natural-language instructions into atomistic computations. Within 28 GPU-hours, ElementsClaw screened 2.4 million crystals, identifying 68,000 high-confidence superconductor candidates and successfully guiding the synthesis of four new superconductors, including Zr₃ScRe₈ with a critical temperature (Tc) of 6.8 K, thereby significantly advancing interactive, human–AI collaborative paradigms in materials discovery.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Search and Optimization: Learning to Search
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Economics, Online Markets and Human Computation: Architectures and workflows that use LLMs for crowd work

Abstract

The discovery of novel materials is critical for global energy and quantum technology transitions. While deep learning has fundamentally reshaped this landscape, existing predictive or generative models typically operate in isolation, lacking the autonomous orchestration required to execute the full discovery process. Here we present ElementsClaw, an agentic framework for materials discovery that synergizes Large Atomic Models (LAMs) with Large Language Models (LLMs). In response to varied human requirements, ElementsClaw dynamically orchestrates a suite of LAM tools finetuned from our proposed model Elements for atomic-scale numerical computation, while leveraging LLMs for high-level semantic reasoning. This shift moves AI-driven materials science from isolated processes toward integrated and human interactive discovery. In the demanding domain of superconductors, our agentic system guides the experimental synthesis of four new superconductors, including Zr3ScRe8 with a transition temperature of 6.8 K and HfZrRe4 at 6.7 K. At scale, ElementsClaw screens more than 2.4 million stable crystals within only 28 GPU hours, identifying 68,000 high-confidence superconducting candidates and vastly expanding the known superconducting space. These results demonstrate how our agent accelerates materials discovery with high physical fidelity.

Problem

• materials discovery
• autonomous orchestration
• superconductors
• AI-driven science
• integrated discovery process

Innovation

• Agentic AI
• Large Atomic Models
• Large Language Models
• Materials Discovery
• Superconductor Prediction