Incorporating short-range order, magnetism, and advanced techniques beyond density functional theory into melting property simulation schemes

• Applicant: Li-Fang Zhu, Ph.D.
• Subject Area: Computer-Aided Design of Materials and Simulation of Materials Behaviour from Atomic to Microscopic Scale ; Thermodynamics and Kinetics as well as Properties of Phases and Microstructure of Materials
• Term: since 2025
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 568216013

Melting properties are essential for constructing phase diagrams, which in turn guide novel materials design. However, experimental measurements of these properties face severe challenges due to the high melting points of such materials. Ab initio methods based on density functional theory (DFT) can provide accurate melting properties but always come with a high computational cost. To address this issue, the applicant previously developed an efficient ab initio approach by integrating a bespoke machine learning potential. However, this approach is applicable only to metals and alloys, and not yet to more complex systems such as oxides, carbides, and magnetic materials. Extending the present methodology to these materials requires further developments, as outlined below. For oxides and carbides, short-range order (SRO) plays a significant role in both solid and liquid phases due to their stronger bonding than those in alloy systems. It can strongly influence the computed melting properties. Existing methods for studying SRO primarily focus on solid phases and cannot be applied to liquid phases due to the complex dynamic motion in the liquid state. Furthermore, the coupling among vibrations, configurations and electronic excitations must be carefully accounted for. This proposal aims to incorporate SRO investigations into the present methodology and systematically analyze the impact of SRO on melting properties. For magnetic systems, accurately treating magnetic excitations introduces another layer of complexity in melting property studies. At the melting point, the system is typically in a paramagnetic state, where the coupling between magnetic and atomic degrees of freedom plays a significant role. Advanced techniques based on ab initio molecular dynamics simulations can capture the coupling accurately but incur extremely high computational costs. This proposal will develop an efficient ab initio approach for studying the melting properties of magnetic systems by leveraging specially designed magnetic machine learning potentials. Standard DFT has its intrinsic deficiency. For instance, for systems with strong Van der Waals interactions, standard exchange-correlation functionals such as PBE and LDA struggle to capture their physical properties accurately, often requiring DFT+U corrections or approaches beyond standard DFT. Recently, the random phase approximation (RPA), a method beyond DFT, has shown better performance in predicting melting temperatures than standard DFT. However, its extremely high computational cost limits broader applications. This proposal aims to integrate RPA into the present methodology while ensuring high efficiency.

• DFG Programme: Research Grants
• International Connection: United Kingdom, USA
• Cooperation Partners: Dr. Zongrui Pei; Professor Dr. Nils Warnken