Molecular Dynamics Study of Stability and Melting Transition in MgFe2O4 Using Universal Deep Learning Potentials

Document Type : Regular Article

Authors

1 Department of Mechanical Engineering, Universitas Muhammadiyah Ponorogo

2 Department of Physics, Institut Teknologi Sepuluh Nopember, Surabaya

3 Department of Physics, Universitas Brawijaya, Malang

4 Department of Physics Education, Universitas Muhammadiyah Mataram

5 Department of Informatics Engineering, Universitas Muhammadiyah Ponorogo

6 Malaysia Japan International Institute of Technology, Universiti Teknologi Malaysia

10.22036/pcr.2026.548982.2753
Abstract
Understanding the thermal stability and melting behavior of complex oxides is vital for their processing and high temperature applications in various fields. MgFe2O4, a representative spinel ferrite, has attracted growing interest, but its high temperature properties remain largely unexplored. In this study, we investigated the stability and melting characteristics of MgFe2O4 using molecular dynamics simulations with machine learning based M3GNet universal interatomic potential. The data driven framework of M3GNet offers enhanced accuracy and transferability across diverse chemistries, thereby overcoming the limitations of conventional empirical force fields. Our simulations first examined the structural resilience of MgFe2O4 under extreme thermal conditions, showing that a defect free crystal retains its ordered lattice up to approximately 2100 K before fully transforming into a liquid. This result highlights the remarkable thermal stability of the spinel structure starting from a perfect crystal configuration. The melting temperature was further investigated using the simulated two phase coexistence method, which is regarded as highly reliable in molecular dynamics studies. The analysis revealed that MgFe2O4 undergoes a solid-liquid transition at approximately 1850 K, with the equilibrated system evolving into a fully liquid phase at higher temperatures. These findings serve as a valuable reference for experimental validation and engineering applications.

Keywords

Subjects



Articles in Press, Accepted Manuscript
Available Online from 03 February 2026

  • Receive Date 24 September 2025
  • Revise Date 21 November 2025
  • Accept Date 01 February 2026