Computing Anharmonic Free Energies in Solids with Machine-Learning Interatomic Potentials

Area of Science:

• Materials Science
• Computational Chemistry
• Solid State Physics

Background:

• Accurate free energies are vital for phase stability and phase diagram construction.
• Traditional methods like harmonic/quasiharmonic approximations are inaccurate at high temperatures.
• Thermodynamic integration is computationally intensive.

Purpose of the Study:

• To develop an efficient workflow for calculating anharmonic free energies.
• To combine nonequilibrium methods with machine learning interatomic potentials.
• To construct accurate temperature- and composition-dependent phase diagrams.

Main Methods:

• Developed an efficient workflow combining nonequilibrium approaches and machine learning potentials.
• Validated the workflow using fcc Ag, hcp Ru, and bcc Mo.
• Applied the framework to the Fe-C system under syngas conditions.

Main Results:

• The workflow accurately captures anharmonic contributions in tested materials.
• Finite-temperature vibrational free energies qualitatively alter phase stability in Fe-C.
• Anharmonic effects significantly modify the stability window of χ-Fe5C2 compared to harmonic predictions.

Conclusions:

• The developed framework provides a robust method for anharmonic free energy calculations.
• This approach enables accurate phase diagram prediction in solids.
• The findings are consistent with experimental observations for the Fe-C system.