By design, this potential is not expected to be accurate for geometries with extremely low coordination -- such as small molecules -- which were not included in the training set.

By design, this potential is not expected to be accurate for geometries with extremely low coordination -- such as small molecules -- which were not included in the input set.

The Al part of this potential leads to very low free surface energy. The potential may easily produce spontaneous cavity nucleation in MD which will be an artifact of this potential.

This potential was developed to accurately reproduce properties of a single phase of the Ni-Al system, namely, B2-NiAl. Properties of other phases of this system are reproduced less accurately. Application of this particular potential to simulations of pure Ni or pure Al is not recommended.

The paper reporting on this potential (G.P. Purja Pun and Y. Mishin, "Development of an interatomic potential for the Ni-Al system," Phil. Mag. 89, 3245 (2009)) contains an error in Table 6. The generalized stacking fault energies in Ni3Al listed in this table are incorrect. The correct energies computed with the new potential are (in J/m2): APB (111) 180 CSF (111) 229 SISF (111) 21 APB (100) 20

This model was developed as a prototype case for the Gibbs-Duhem protocol for optimizing melting points. As such, the melting point was the primary metric for its development. We have shown that the modifications to the original Mei-Davenport potential that we used to optimize T_m did not change the quality of the potential with respect to elastic constants, density, etc., but no systematic study of other properties was performed.

Model was fit to room temperature properties and extrapolated, classical (i.e. non-quantum) low temperature properties. The intended application is for shock compression simulations, but validation against shock pressure conditions is limited. The model appears to provide accurate results in the elastic shock regime (i.e. pressure ~ 5 to 10 GPa), as documented in the paper by Zimmerman, Winey and Gupta (Phys. Rev. B, 2011). Behavior at higher pressures and in the elastic-plastic regime is unknown.

All of the cross interactions are determined through a universal mixing function and not every combination of species has been tested. The database is not suitable for modeling metal compounds.

By design, this potential is not expected to be accurate for geometries with extremely low coordination -- such as small molecules -- which were not included in the input set.

By design, this potential is not expected to be accurate for geometries with extremely low coordination -- such as small molecules -- which were not included in the input set.

This model was automatically fit using Lorentz-Berthelot mixing rules. It reproduces the dimer equilibrium separation (covalent radii) and the bond dissociation energies. It has not been fitted to other physical properties and its ability to model structures other than dimers is unknown.

This potential is designed for the structural properties of High Entropy Alloys (HEA)s and Complex Concentrated Alloys (CCAs). The fitting procedure involved developing all included unary, binary and ternary systems so it can be used for any alloy subset. This potential focuses on the structural analysis of alloys including shear strength and elastic constants, dislocation dynamics and their impact on alloy strength, and the analysis of defect effects, such as voids, on material properties. However, the potential was not optimized for temperature-dependent properties and was not fit to density, thermal expansion coefficients, or thermal conductivity data.

Several models from the PolyMLP repository are available for AlAu, all laying on a pareto front of computational speed vs. RMS accuracy. They are labeled p1, p2, p3, p4, p5 in their KIM IDs. Higher numbers mean slower models. The author-recommended model for this system is p1 (this model).

Several models from the PolyMLP repository are available for AlGe, all laying on a pareto front of computational speed vs. RMS accuracy. They are labeled p1, p2, p3, p4, p5 in their KIM IDs. Higher numbers mean slower models. The author-recommended model for this system is p1 (this model).

Several models from the PolyMLP repository are available for AlSc, all laying on a pareto front of computational speed vs. RMS accuracy. They are labeled p1, p2, p3 in their KIM IDs. Higher numbers mean slower models. The author-recommended model for this system is p1 (this model).

Several models from the PolyMLP repository are available for AlSi, all laying on a pareto front of computational speed vs. RMS accuracy. They are labeled p1, p2, p3, p4, p5 in their KIM IDs. Higher numbers mean slower models. The author-recommended model for this system is p1 (this model).

Several models from the PolyMLP repository are available for AlZr, all laying on a pareto front of computational speed vs. RMS accuracy. They are labeled p1, p2, p3, p4, p5 in their KIM IDs. Higher numbers mean slower models. The author-recommended model for this system is p1 (this model).

Several models from the PolyMLP repository are available for TiAl, all laying on a pareto front of computational speed vs. RMS accuracy. They are labeled p1, p2, p3, p4, p5 in their KIM IDs. Higher numbers mean slower models. The author-recommended model for this system is p1 (this model).

Several models from the PolyMLP repository are available for AlAg, all laying on a pareto front of computational speed vs. RMS accuracy. They are labeled p1, p2, p3, p4, p5 in their KIM IDs. Higher numbers mean slower models. The author-recommended model for this system is p2 (this model).

Several models from the PolyMLP repository are available for AlCu, all laying on a pareto front of computational speed vs. RMS accuracy. They are labeled p1, p2, p3, p4, p5 in their KIM IDs. Higher numbers mean slower models. The author-recommended model for this system is p2 (this model).

Several models from the PolyMLP repository are available for AlPt, all laying on a pareto front of computational speed vs. RMS accuracy. They are labeled p1, p2, p3, p4, p5 in their KIM IDs. Higher numbers mean slower models. The author-recommended model for this system is p2 (this model).

Several models from the PolyMLP repository are available for AlSr, all laying on a pareto front of computational speed vs. RMS accuracy. They are labeled p1, p2, p3, p4, p5 in their KIM IDs. Higher numbers mean slower models. The author-recommended model for this system is p2 (this model).

Several models from the PolyMLP repository are available for AlZn, all laying on a pareto front of computational speed vs. RMS accuracy. They are labeled p1, p2, p3, p4, p5 in their KIM IDs. Higher numbers mean slower models. The author-recommended model for this system is p2 (this model).

By design, this potential is not expected to be accurate for geometries with extremely low coordination -- such as small molecules -- which were not included in the input set.