Comparative analysis of classical and machine learning models for predicting the mechanical properties of pristine and defective 2D coinage metals

K Kashmarizad and F Alavi and MMS Fakhrabadi, EUROPEAN PHYSICAL JOURNAL PLUS, 140, 1053 (2025).

DOI: 10.1140/epjp/s13360-025-06986-7

The recent synthesis of the hexagonal gold monolayer provides new opportunities to investigate the mechanical properties of free-standing single-layer metals. To overcome the limitations of conventional interatomic potentials such as the Embedded Atom Method (EAM) and Reactive Force Field (ReaxFF) in describing 2D systems due to their bulk-oriented parameterization, this study employs advanced machine learning interatomic potentials (MLIPs), specifically Gaussian approximation potential, to examine the stability, mechanical properties, and fracture behavior of pristine and defective 2D copper, silver, and gold. The excellent accuracy of MLIP-based results is validated by comparing them with DFT calculations, yielding an accuracy of more than 95%. According to the results, the highest value of modulus of elasticity is related to the copper nanosheet with a value of 265/250 GPa in the y/x directions. Also, the highest ultimate tensile strength (UTS) in 300 K is observed in the copper nanosheet, equaling 18/13 GPa along the y/x directions. The ability of pristine and defective copper nanosheets to outperform their counterparts across all investigated properties makes this 2D hexagonal structure a promising candidate for future experimental synthesis.

Return to Publications page