Strategic sampling with stochastic surface walking for machine learning force fields in iron's bcc-hcp phase...

Area of Science:

• Computational Materials Science
• Condensed Matter Physics
• Machine Learning in Materials

Background:

• Simulating phase transitions in materials like iron is crucial for understanding their behavior under extreme conditions.
• Accurate force fields are essential for reliable molecular dynamics simulations.

Purpose of the Study:

• To develop an efficient and accurate machine learning-based force field for simulating the body-centered cubic (bcc) to hexagonal close-packed (hcp) phase transitions of iron.
• To validate the machine learning force field against high-accuracy calculations and experimental data.

Main Methods:

• Machine learning potential construction using Bayesian inference.
• Molecular dynamics (MD) and stochastic surface walking (SSW) sampling methods.
• SOAP (Smooth Overlap of Atomic Positions) descriptors for structural representation.
• Density Functional Theory (DFT) calculations for validation.

Main Results:

• The developed machine learning force field accurately reproduces DFT-calculated energies, forces, and stresses for iron.
• It demonstrates good coverage of the phase transition space.
• Accurate prediction of crystal structure parameters, elastic constants, and bulk modulus for bcc and hcp iron phases.

Conclusions:

• The machine learning force field provides an effective and accurate tool for investigating iron phase transitions.
• This work offers new computational approaches for materials science and solid-state physics research.
• The developed potential enables deeper insights into the complex behavior of iron under varying conditions.