A new family of hollow Au₆₀ cages: topological diversity and connectivity-driven stability

Hollow gold nanoclusters are nanoscale systems in which relativistic effects, coordination, topology, and thermal fluctuations compete to determine structural stability. Here, we combine density functional theory, descriptor-based machine-learning-assisted analysis, and finite-temperature molecular dynamics to investigate Au₆₀ hollow cages. We generate 30 new neutral Au₆₀ hollow isomers from different topological seeds and analyze them together with the previously reported I-Au₆₀ cage. Their energetics are assessed at the PBE, PBE + SOC, and PBE + SOC + D3 levels. Low-symmetry cages with compact coordination environments are found to be significantly more stable than the fullerene-like I structure, which lies in the high-energy tail of the hollow landscape. Dispersion corrections further penalize under-connected or perforated motifs, several of which undergo partial closure toward more locally packed and less porous arrangements. Descriptor-based analysis shows that stability correlates primarily with network connectivity and local coordination, rather than with spherical character, global symmetry, frontier electronic gaps, or chirality. In particular, leave-one-out regression identifies the average coordination number as the dominant descriptor of relative stability. Finite-temperature simulations reveal a clear hierarchy of thermal robustness: highly connected cages remain metastable up to, and in some cases above, room temperature, whereas the I cage and strongly perforated motifs reconstruct early. These results show that connectivity and mechanical adaptability, rather than maximal symmetry, are the key ingredients for stabilizing hollow Au₆₀ cages.