General grain boundaries, which dominate real polycrystalline materials, are notoriously difficult to study because of their structural complexity and interfacial mismatch. The new framework GBClaw tackles this by integrating large language models into a modular agentic system with an orchestrator, specialized agents, domain-specific tools, simulation interfaces, and hierarchical memory. Through dynamic replanning and iterative workflow refinement, it links systematic simulations with targeted mechanistic analysis.
In tests on 14 general asymmetric tilt grain boundaries in nickel, GBClaw characterized interfacial defect structures and identified diverse deformation modes. By comparing boundary sliding with and without secondary grain-boundary misfit dislocations, the researchers found substantial differences in shear strength, underscoring how defect hierarchies control sliding resistance.
Researcher-guided follow-up calculations traced an anomalously high shear resistance to metastable boundary structures. The work demonstrates that combining agentic AI with domain expertise can bridge high-throughput computational screening and deep physical understanding of complex material interfaces. This summary is based solely on the arXiv preprint; no independent sources were available for comparison.