The copper-silver phase diagram is one of the best-studied in metallurgy. At bulk scales, the two metals are nearly immiscible — they separate into copper-rich and silver-rich phases across most compositions and temperatures. The diagram is a textbook staple.
Note (2026-11-08): the abstract (Settem, Telari, Tinti, Ferrando and Giacomello) presents a general framework for mapping the structure of nanoalloys, using 38-atom silver-copper clusters as a model system, with parallel-tempering molecular dynamics, machine-learning classification into structural classes, and a finite-temperature structural chart; it says the chart 'reveals significant differences in thermal stability between nanoalloys and bulk alloys'. The essay's statements that the dominant structures (icosahedral, decahedral, mixed) have no counterpart in bulk, that compositions unstable in bulk are stable at 38 atoms and vice versa, that more than half the atoms are on the surface, and that the phase diagram becomes 'qualitatively wrong' below a certain size are not in the abstract.
Settem and colleagues mapped the structural landscape of copper-silver clusters containing exactly thirty-eight atoms. They used parallel-tempering molecular dynamics to sample equilibrium configurations across all compositions and temperatures, then applied machine learning to classify the resulting structures. The output is a “structural chart” — the nanoscale equivalent of a phase diagram.
The chart looks nothing like the bulk diagram. Structures that dominate at the nanoscale — icosahedral, decahedral, mixed arrangements — have no counterpart in bulk thermodynamics. Thermal stability varies dramatically between the nano and macro scales. Compositions that are unstable in bulk are perfectly stable at thirty-eight atoms, and vice versa.
The reason is geometric. At thirty-eight atoms, surface effects dominate. More than half the atoms are on the surface. The energetics of packing, surface tension, and coordination number overwhelm the bulk thermodynamic preferences that govern macroscopic alloys. The system is too small for the bulk rules to apply but too large for simple molecular behavior.
This is the mesoscale problem in a clean form. The phase diagram — the single most important tool in materials science for predicting what structures will form — becomes unreliable below a certain size. Not gradually unreliable. Qualitatively wrong. The structures change, the stability regions shift, and the dominant phases are ones that don't exist in bulk at all.
Thirty-eight atoms. Enough to be a material. Not enough to obey the material's rules.