Abstract
<jats:p> Nanoparticle catalyst properties are often governed by processing-defined near-surface structures rather than nominal bulk phases. Static idealized models therefore can miss the kinetic pathways that drive subsurface evolution and active-site redistribution during synthesis and activation. Here, we present an activation-mapped kinetic modeling workflow combining density functional theory (DFT), machine-learning-accelerated cluster expansion (ML-CE), Metropolis Monte Carlo (MMC), and kinetic Monte Carlo (KMC) to trace processing-driven structural evolution in L1 <jats:sub>0</jats:sub> and L1 <jats:sub>2</jats:sub> Pt–Fe nanoparticles. We define the working-state ensemble as the statistically sampled Pt-skin configurations generated by the modeled sequence of high-temperature ordering, selective Fe leaching, vacancy-mediated restructuring, and room-temperature activation. This ensemble construction links processing history to subsurface composition, surface morphology, and site-resolved catalytic response. Within the modeled activation sequence, phase-dependent Fe leaching inverts second-layer Fe enrichment relative to ideal ordered references and switches the dominant active-site families, making L1 <jats:sub>0</jats:sub> -derived nanoparticles terrace-controlled, whereas L1 <jats:sub>2</jats:sub> -derived nanoparticles become comparatively edge-controlled. When coupled to steady-state *OH coverage kinetics, these activation-derived ensembles recover the experimental oxygen reduction reaction (ORR) activity ranking and reveal phase-specific size–activity trends missed by ideal models. More broadly, the framework establishes a processing–structure–property linkage for predicting activation-derived structures and highlights how path-dependent reconstruction can overturn static-model rankings in alloy nanoparticle catalysts. </jats:p>