From interatomic potentials to phase diagrams: Integrated tools for validation and fitting

• Applicants: Professor Dr. Karsten Albe, since 8/2020; Professor Dr. Jörg Neugebauer
• Subject Area: Computer-Aided Design of Materials and Simulation of Materials Behaviour from Atomic to Microscopic Scale
• Term: from 2018 to 2023
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 405621160

• Final Report Year: 2023

Final Report Abstract

Traditionally, interatomic potentials have been parameterized based on readily available and computationally inexpensive material properties, such as cohesive energies and elastic constants. While these properties provide a useful starting point for potential development, they may not capture the full thermodynamic behavior of materials, especially under extreme conditions such as high temperatures or pressures, or when stoichiometrically complex phases appear. The thermodynamic bulk phase diagrams, which predict the stable phases of a material as a function of composition, temperature, and sometimes pressure, are a much more sensitive measure for the accuracy and transferability of a potential. Unfortunately, constructing phase diagrams using traditional methods to compute free energies as function of these variables is computationally expensive, which limits their use in the fitting approach as it requires thousands to millions of iteration cycles. This project successfully addressed these challenges by introducing innovative algorithms and computational tools that significantly enhance the computation of various free energy contributions. Through the development of novel techniques, including a sampling-free approach to compute anharmonic vibrational free energies, a rescaling approach to include nuclear quantum effects and a various machine learning approaches to describe compositionally as well as structurally complex materials, huge speed-ups in computing free energies and thus phase diagrams could be achieved. These achievements not only enable the rapid validation of potential accuracy, but also opened the door to studying complex materials under a wide range of conditions. Specifically, the boost in numerical performance to compute free energies as function of temperature, pressure, and composition achieved by these tools opens the potential to include phase diagrams in the interatomic potential fitting process. The integrated tools developed in this project have been implemented in the pyiron-framework, providing open access of the developed approaches to the scientific community. The availability of tools for fitting empirical and machine learning potentials represents a significant step forward. The inclusion of thermodynamic phase diagrams as a measure of potential accuracy enhances the predictive power and robustness of these models and broadens their applicability. The availability of these opensource tools provides a more robust basis for computational materials science and will help in designing and exploring advanced materials in real conditions with greater accuracy and efficiency. To disseminate the computational tools and methods, several workshops and conferences with hackathons as a central element have been co-organized by the PIs as part of the DFG-funded Potential initiative.

Publications

• Anharmonic free energy of lattice vibrations in fcc crystals from a mean-field bond. Physical Review B, 102(10).
  Swinburne, Thomas D.; Janssen, Jan; Todorova, Mira; Simpson, Gideon; Plechac, Petr; Luskin, Mitchell & Neugebauer, Jörg
  
• Chemically induced local lattice distortions versus structural phase transformations in compositionally complex alloys. npj Computational Materials, 7(1).
  Ikeda, Yuji; Gubaev, Konstantin; Neugebauer, Jörg; Grabowski, Blazej & Körmann, Fritz
  
• Finite-temperature interplay of structural stability, chemical complexity, and elastic properties of bcc multicomponent alloys from ab initio trained machine-learning potentials. Physical Review Materials, 5(7).
  Gubaev, Konstantin; Ikeda, Yuji; Tasnádi, Ferenc; Neugebauer, Jörg; Shapeev, Alexander V.; Grabowski, Blazej & Körmann, Fritz
  
• Impact of Water Coadsorption on the Electrode Potential of H-Pt(1 1 1)-Liquid Water Interfaces. Physical Review Letters, 126(16).
  Surendralal, Sudarsan; Todorova, Mira & Neugebauer, Jörg
  
• A machine-learned interatomic potential for silica and its relation to empirical models. npj Computational Materials, 8(1).
  Erhard, Linus C.; Rohrer, Jochen; Albe, Karsten & Deringer, Volker L.
  
• Approximating the impact of nuclear quantum effects on thermodynamic properties of crystalline solids by temperature remapping. Physical Review B, 105(18).
  Dsouza, Raynol; Huber, Liam; Grabowski, Blazej & Neugebauer, Jörg
  
• Anharmonicity in bcc refractory elements: A detailed ab initio analysis. Physical Review B, 107(1).
  Srinivasan, Prashanth; Shapeev, Alexander; Neugebauer, Jörg; Körmann, Fritz & Grabowski, Blazej
  
• Quantification of electronic and magnetoelastic mechanisms of first-order magnetic phase transitions from first principles: application to caloric effects in La(Fe x Si 1−x)13. Journal of Physics: Energy, 5(3), 034004.
  Mendive, Tapia Eduardo; Patrick, Christopher E.; Hickel, Tilmann; Neugebauer, Jörg & Staunton, Julie B.
  
• Simulating short-range order in compositionally complex materials. Nature Computational Science, 3(3), 221-229.
  Ferrari, Alberto; Körmann, Fritz; Asta, Mark & Neugebauer, Jörg
  
• Systematic atomic structure datasets for machine learning potentials: Application to defects in magnesium. Physical Review B, 107(10).
  Poul, Marvin; Huber, Liam; Bitzek, Erik & Neugebauer, Jörg