Accelerating CALPHAD-based Phase Diagram Predictions in Complex Alloys Using Universal Machine Learning Potentials: Opportunities and Challenges

Abstract

Accurate phase diagram prediction is crucial for understanding alloy thermodynamics and advancing materials design. While traditional CALPHAD methods are robust, they are resource-intensive and limited by experimentally assessed data. This work explores the use of machine learning interatomic potentials (MLIPs) such as M3GNet, CHGNet, MACE, SevenNet, and ORB to significantly accelerate phase diagram calculations by using the Alloy Theoretic Automated Toolkit (ATAT) to map calculations of the energies and free energies of atomistic systems to CALPHAD-compatible thermodynamic descriptions. Using case studies including Cr-Mo, Cu-Au, and Pt-W, we demonstrate that MLIPs, particularly ORB, achieve computational speedups exceeding three orders of magnitude compared to DFT while maintaining phase stability predictions within acceptable accuracy. Extending this approach to liquid phases and ternary systems like Cr-Mo-V highlights its versatility for high-entropy alloys and complex chemical spaces. This work demonstrates that MLIPs, integrated with tools like ATAT within a CALPHAD framework, provide an efficient and accurate framework for high-throughput thermodynamic modeling, enabling rapid exploration of novel alloy systems. While many challenges remain to be addressed, the accuracy of some of these MLIPs (ORB in particular) are on the verge of paving the way toward high-throughput generation of CALPHAD thermodynamic descriptions of multi-component, multi-phase alloy systems.

11 pages, 8 figures

References in corpus (13)

• E(3)-Equivariant Graph Neural Networks for Data-Efficient and Accurate Interatomic Potentials
• The Alloy Theoretic Automated Toolkit: A User Guide
• Multicomponent multisublattice alloys, nonconfigurational entropy and other additions to the Alloy Theoretic Automated Toolkit
• A Universal Graph Deep Learning Interatomic Potential for the Periodic Table
• On-the-fly Closed-loop Autonomous Materials Discovery via Bayesian Active Learning
• MACE: Higher Order Equivariant Message Passing Neural Networks for Fast and Accurate Force Fields
• Scalable Parallel Algorithm for Graph Neural Network Interatomic Potentials in Molecular Dynamics Simulations
• The Design Space of E(3)-Equivariant Atom-Centered Interatomic Potentials
• Autonomous synthesis of metastable materials
• Orb: A Fast, Scalable Neural Network Potential
• Deciphering Chemical Ordering in High Entropy Materials: A Machine Learning-Accelerated High-throughput Cluster Expansion Approach
• Machine-learning potentials enable predictive $\text{and}$ tractable high-throughput screening of random alloys
• Active Learning for Discovering Complex Phase Diagrams with Gaussian Processes

Cited by in corpus (2)

• Machine Learning Potentials for Alloys: A Detailed Workflow to Predict Phase Diagrams and Benchmark Accuracy
• Accelerating Complex Materials Discovery with Universal Machine-Learning Potential-Driven Structure Prediction