EDL-Catalyzed Phase-Field Modeling and Numerically Attentive Interpolation Framework for the Intelligent Design of Cu@Ag Core-Shell Nanoparticles

ABSTRACT Cu@Ag core-shell nanoparticles offer a promising, low-cost alternative to pure nano-Ag for advanced power-device packaging, yet achieving uniform, ultrathin Ag shells remains challenging due to competing nucleation pathways. Here, we systematically investigate the synthesis, thermal reliability, and sintering behavior of Cu@Ag nanoparticles by tuning the Cu-to-Ag molar ratio. We demonstrate that low precursor ratios (1:1, 2:1) induce high supersaturation and homogeneous nucleation, yielding incomplete shells, whereas optimal ratios (3:1, 4:1) promote heterogeneous nucleation, producing continuous 3-5 nm shells that significantly enhance oxidation resistance. Atomistic molecular dynamics simulations reveal a non-linear relationship between shell thickness and thermal properties, identifying an optimal ∼0.8 nm shell that minimizes the melting point and maximizes sintering kinetics. To bridge these multiscale insights with practical synthesis, we develop an electrostatic double layer (EDL)-catalyzed phase-field model, generating a high-fidelity training manifold across a four-dimensional design space. This manifold underpins the CoreShellGPT engine, which integrates a transformer-inspired Numerically Attentive Interpolation Framework (NAIF) with a large language model interface. CoreShellGPT enables natural language-driven inverse design, autonomously mapping desired morphological and thermal targets to optimal synthesis parameters. This work not only establishes Cu@Ag nanoparticles as a high-reliability interconnection material but also provides a scalable, physics-informed AI blueprint for accelerating nanomaterial discovery.