General-Purpose Machine-Learned Potential for CrCoNi Alloys Enabling Large-Scale Atomistic Simulations with First-Principles Accuracy

Abstract

CrCoNi medium-entropy alloys exhibit exceptional mechanical properties arising from pronounced chemical complexity, including short-range order (SRO), and low stacking fault energy, posing challenges for large-scale atomistic simulations. While most models focus on equimolar compositions, deviations from equimolarity provide an effective route to tuning properties, requiring transferable interatomic potentials that capture composition-dependent behavior. Here we develop a general-purpose machine-learned interatomic potential for the CrCoNi system within the neuroevolution potential (NEP) framework, achieving near first-principles accuracy with high computational efficiency. Trained on a comprehensive dataset spanning pure elements, binary and ternary alloys across a wide compositional range, diverse crystal structures and thermodynamic conditions, and based on spin-polarized \textit{ab initio} data, the model accurately reproduces equations of state, phonons, elastic constants, dislocation dissociation, surface and defect energies, melting temperatures and strain-induced phase transformations. It further captures SRO and its effect on stacking fault energies across both equimolar and non-equimolar compositions, in agreement with first-principles and experiments. In contrast to existing potentials, typically limited to equimolar alloys and less accurate for pure elements, the present model delivers consistent accuracy across the full compositional space while retaining superior efficiency. These results enable reliable atomistic simulations of composition-dependent behaviour and provide a framework for the design of non-equimolar CrCoNi alloys.

Figures (9)

• Figure 1: Pairwise comparisons of energies, forces, and stresses for the (a) NEP training dataset, (b) MTP dataset cao2025capturing, and (c) validation dataset, as predicted by NEP, EAM1 li2019strengthening, EAM2 farkas2020model, MEAM choi2018understanding, and MTP cao2025capturing. (d) Principal component analysis (PCA) of the dataset based on NEP descriptors, where each point represents an individual atomic configuration.
• Figure 2: Energy–atomic volume relations for (a) BCC Cr, (b) HCP Co, (c) FCC Ni, and (d) CrCoNi alloy, as obtained from DFT, NEP, EAM1 li2019strengthening, EAM2 farkas2020model, MEAM choi2018understanding, and MTP cao2025capturing calculations.
• Figure 3: Phonon dispersions of (a) BCC Cr, (b) HCP Co, and (c) FCC Ni calculated using DFT, NEP, and experimental data. The corresponding comparisons with EAM1 li2019strengthening, EAM2 farkas2020model, MEAM choi2018understanding, and MTP cao2025capturing are shown in (d)–(f) for clarity. Experimental data for BCC Cr shaw1971investigation, HCP Co wakaba1982lattice, and FCC Ni birgeneau1964normal, obtained from room-temperature neutron diffraction measurements, are included for comparison.
• Figure 4: Comparison of Warren–Cowley short-range order parameters at (a) 500 K, (b) 800 K, and (c) 1200 K as predicted by NEP, EAM1 li2019strengthening, EAM2 farkas2020model, MEAM choi2018understanding, and MTP cao2025capturing. Error bars represent the standard deviation obtained from 20 independent simulations with different initial configurations. At 500 K, the shaded gray region denotes the range of DFT MC results reported in three literature sources cao2025capturingding2018tunabletamm2015atomic. For 800 and 1200 K, where only single DFT reference values are available tamm2015atomic, the results are shown as discrete points. (d) Cumulative relative error ($\varepsilon^{\mathrm{SRO}}$) with respect to DFT.
• Figure 5: Intrinsic stacking fault energies $\gamma_{\mathrm{isf}}$ for (a) random solid solutions and (b) ordered solid solutions, predicted by NEP, EAM1 li2019strengthening, EAM2 farkas2020model, MEAM choi2018understanding, and MTP cao2025capturing. DFT results for random solid solutions in (a) are taken from Ref. zhu2023effects, while experimental data for ordered solid solutions in (b) are from Refs. laplanche2017reasonsliu2018stacking. Stacking fault widths associated with edge and screw dislocations in (c–d) Ni and (e–f) CrCoNi. The experimental data in (c) are taken from Ref. carter1977the, while the DFT results in (d) are from Ref. tan2019dislocation.