Multi-AI agent framework reveals the "oxide gatekeeper" in aluminum nanoparticle oxidation.

Aluminum nanoparticles (ANPs) are among the most energy-dense solid fuels, yet the atomic mechanisms governing their transition from passivated particles to highly reactive states remain elusive. This stems from a fundamental computational bottleneck: Ab initio methods offer quantum-level fidelity but are restricted to small spatiotemporal scales (<500 atoms, picoseconds), while empirical force fields lack the reactive fidelity required for complex combustion environments. Here, we bridge this gap by employing a human-audited, multiagent framework to make the development and validation of a task-specific machine learning potential (MLP) more traceable and auditable. The workflow coordinates active learning, model evaluation, validation evidence, and human expert review, while leaving scientific interpretation and model-use decisions under researcher control. This collaborative cycle ensures quantum mechanical accuracy while exhibiting near-linear scalability. Strikingly, our simulations reveal a temperature-regulated dual-mode oxidation mechanism: At moderate temperatures, the oxide shell acts as a dynamic "gatekeeper," regulating oxidation through a "breathing mode" of transient nanochannels; above a critical threshold, a "rupture mode" leads to shell failure and rapid combustion. Importantly, we resolve a decades-old controversy by demonstrating that aluminum cation outward diffusion, rather than oxygen transport, dominates mass transfer in studied regimes, with diffusion coefficients consistently exceeding those of oxygen by 2 to 3 orders of magnitude. These results provide an atomistic framework for interpreting ANP reactivity and illustrate how human-audited agentic workflows can make domain-specific reactive MLP development and deployment more reproducible.