Size-dependent reliability of empirical potentials for global optimization of Pt-Cu bimetallic clusters

Empirical potentials (EPs) are widely used to pre-screen candidate structures before density functional theory (DFT) refinement, but their transferability to bimetallic clusters must be tested across size, composition, and parameterization. In this work, we compare the Gupta, Sutton-Chen (SC), and Finnis-Sinclair (FS) potentials for Pt_mCu_N-m clusters at N = 13, 23, and 38 using CDE global optimization, geometric and segregation descriptors, and DFT assessment of selected EP minima. At N = 13 the three potentials converge on the same icosahedral motif, with the Gupta and SC candidates nearly degenerate in DFT energy and the FS candidate higher by 0.665 eV. At N = 23, the DFT ranking becomes composition dependent, with SC lowest at the near-equiatomic Pt₁₁Cu₁₂ composition, FS lowest in the Cu-rich case, and Gupta lowest in the Pt-rich case; at N = 38, the SC-derived Pt₁₉Cu₁₉ structure is lowest in the single-point DFT comparison. The FS potential accesses a distinct compact basin at N = 38 with a characteristic 143° dihedral signature, though this motif is not favored in the present single-point DFT screening of EP-optimized N = 38 geometries. These results show that empirical descriptors and DFT energetic competitiveness do not collapse into a single ranking, and that structural diversity from multiple EPs can broaden the Pt-Cu candidate pool when the resulting basins are filtered against first-principles calculations.