Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Machine Learning Potentials for Alloys: A Detailed Workflow to Predict Phase Diagrams and Benchmark Accuracy

Siya Zhu, Doguhan Sariturk, Raymundo Arroyave

TL;DR

The paper addresses the costly challenge of predicting phase diagrams for compositionally complex alloys by integrating machine-learning interatomic potentials (MLIPs) into CALPHAD-style workflows. It introduces PhaseForge, a fully automated pipeline within the ATAT framework via MaterialsFramework, enabling MLIP-driven energy relaxations, vibrational contributions, and liquid-phase molecular dynamics to produce CALPHAD-compatible TDB files. The authors validate the approach on Ni–Re and a Co–Cr–Fe–Ni–V high-entropy alloy, demonstrating accurate topology capture and substantial speedups over full ab initio methods, while also providing a benchmarking pathway for MLIPs. This workflow enables high-throughput phase-diagram predictions across multicomponent spaces and offers a practical platform to assess MLIP performance in thermodynamics, with potential extensions to active learning and visualization.

Abstract

High-entropy alloys (HEAs) have attracted increasing attention due to their unique structural and functional properties. In the study of HEAs, thermodynamic properties and phase stability play a crucial role, making phase diagram calculations significantly important. However, phase diagram calculations with conventional CALPHAD assessments based on experimental or ab-initio data can be expensive. With the emergence of machine-learning interatomic potentials (MLIPs), we have developed a program named PhaseForge, which integrates MLIPs into the Alloy Theoretic Automated Toolkit (ATAT) framework using our MLIP calculation library, MaterialsFramework, to enable efficient exploration of alloy phase diagrams. Moreover, our workflow can also serve as a benchmarking tool for evaluating the quality of different MLIPs.

Machine Learning Potentials for Alloys: A Detailed Workflow to Predict Phase Diagrams and Benchmark Accuracy

TL;DR

The paper addresses the costly challenge of predicting phase diagrams for compositionally complex alloys by integrating machine-learning interatomic potentials (MLIPs) into CALPHAD-style workflows. It introduces PhaseForge, a fully automated pipeline within the ATAT framework via MaterialsFramework, enabling MLIP-driven energy relaxations, vibrational contributions, and liquid-phase molecular dynamics to produce CALPHAD-compatible TDB files. The authors validate the approach on Ni–Re and a Co–Cr–Fe–Ni–V high-entropy alloy, demonstrating accurate topology capture and substantial speedups over full ab initio methods, while also providing a benchmarking pathway for MLIPs. This workflow enables high-throughput phase-diagram predictions across multicomponent spaces and offers a practical platform to assess MLIP performance in thermodynamics, with potential extensions to active learning and visualization.

Abstract

High-entropy alloys (HEAs) have attracted increasing attention due to their unique structural and functional properties. In the study of HEAs, thermodynamic properties and phase stability play a crucial role, making phase diagram calculations significantly important. However, phase diagram calculations with conventional CALPHAD assessments based on experimental or ab-initio data can be expensive. With the emergence of machine-learning interatomic potentials (MLIPs), we have developed a program named PhaseForge, which integrates MLIPs into the Alloy Theoretic Automated Toolkit (ATAT) framework using our MLIP calculation library, MaterialsFramework, to enable efficient exploration of alloy phase diagrams. Moreover, our workflow can also serve as a benchmarking tool for evaluating the quality of different MLIPs.

Paper Structure

This paper contains 25 sections, 5 equations, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Methods
  3. Compound energy formalism
  4. Molecular dynamics for the liquid phase
  5. Phonon and mechanical instabilities
  6. Database for pure elements with ab-initio calculations
  7. Other modeling considerations
  8. Usage
  9. A quick inline command for phase diagram calculation
  10. Incorporate MLIP within MaterialsFramework
  11. Construct Special Quasirandom Structures
  12. Structures and elements selection
  13. Relaxed energies calculations using MLIP
  14. Inflection Detection for mechanically unstable phases
  15. Vibrational Entropy Calculations
  16. ...and 10 more sections

Figures (3)

  • Figure 1: (a) Phase diagram of Ni-Re system with FCC_A1, HCP_A3, D0$_{19}$, D1$_a$, and liquid phases, calculated with VASP and Grace; (b) Phase diagram calculated with SevenNet; (c) Phase diagram calculated with CHGNet.
  • Figure 2: The 10 binary phase diagrams in Co-Cr-Fe-Ni-V system.
  • Figure 3: The 10 ternary phase diagrams in Co-Cr-Fe-Ni-V system at 1400K.