Thermal dynamics and coalescence of Au₁₄₄(SR)₆₀ clusters from a machine-learned potential

Ligand-protected metal nanoclusters have gained significant attention due to their diverse applications in catalysis, bioimaging, and nanomedicine. While their ground-state electronic structure and optical properties have been extensively studied via density functional theory (DFT) methods, theoretical insights into their dynamic behavior, particularly for larger clusters, remain scarce due to the prohibitive numerical cost of long-timescale simulations using forces calculated from DFT. Here we investigate, using molecular dynamics (MD) simulations up to 0.12 μs timescale, thermal dynamics of the well-known Au₁₄₄(SR)₆₀ at 300-550 K using the recently parametrized atomic cluster expansion (ACE) potential, trained from DFT data for thiolate-protected gold clusters. Our findings reveal that thermal effects induce increased mobility in a layer-by-layer fashion, leading to formation of polymeric gold-thiolate units and rings which may fragment from the cluster at high temperatures. The remaining smaller clusters resemble experimentally observed cluster compositions. Close interaction of two Au₁₄₄(SR)₆₀ clusters leads to coalescence, resulting in a cluster composition and structure of the inner metal core close to ones identified in previous experiments. This work reveals mechanisms for thermal effects in ligand-protected gold clusters and larger nanoparticles that are instrumental for understanding their catalytic activity and inter-particle reactions at the atomistic level.