Substitutional alloying using crystal graph neural networks

Dario Massa,Daniel Cieśliński,Amirhossein Naghdi,Stefanos Papanikolaou

DOI: https://doi.org/10.1063/5.0163765

IF: 1.697

2024-01-01

AIP Advances

Abstract:Materials discovery, especially for applications that require extreme operating conditions, requires extensive testing that naturally limits the ability to inquire the wealth of possible compositions. Machine Learning (ML) has nowadays a well-established role in facilitating this effort in systematic ways. The increasing amount of available accurate Density Functional Theory (DFT) data represents a solid basis upon which new ML models can be trained and tested. While conventional models rely on static descriptors, generally suitable for a limited class of systems, the flexibility of Graph Neural Networks (GNNs) allows for direct learning representations on graphs, such as the ones formed by crystals. We utilize crystal graph neural networks (CGNNs) known to predict crystal properties with DFT level accuracy through graphs by encoding the atomic (node/vertex), bond (edge), and global state attributes. In this work, we aim at testing the ability of the CGNN MegNet framework in predicting a number of properties of systems previously unseen in the model, which are obtained by adding a substitutional defect to bulk crystals that are included in the training set. We perform DFT validation to assess the accuracy in the prediction of formation energies and structural features (such as elastic moduli). Using CGNNs, one may identify promising paths in alloy discovery.

materials science, multidisciplinary,physics, applied,nanoscience & nanotechnology

What problem does this paper attempt to address?

This paper discusses the use of Crystal Graph Neural Networks (CGNNs) for substitutional alloying. In materials science, the discovery of new materials, especially those suitable for extreme conditions, often requires a large number of experimental tests, which limits the research on possible compositions. Machine learning (ML) methods have been used to accelerate this process in a systematic way. CGNNs are a flexible learning tool that can directly learn representations on graphs such as crystal structures. The authors of the paper used a CGNN framework called MEGNet to predict the properties of single-atom substitutional defect crystals that have not appeared in the model training set, such as formation energy and structural features (e.g. elastic modulus). They evaluated the predicted accuracy through Density Functional Theory (DFT) verification. In the study, the authors first introduced the MEGNet model, which uses atomic indices as node attributes and spatial distances as edge attributes, while considering global state attributes. They collected data from the Materials Project database and made predictions using the pre-trained MEGNet model, which were then validated using DFT calculations. The paper focuses mainly on two issues: 1) the impact of the same defect in multiple host crystals; 2) the impact of different elements as substituting atoms in the same host crystal. The results show that CGNNs have potential in predicting defect crystal properties, but there are also limitations. This provides a new pathway for alloy discovery, but also reveals the predictive capability boundaries of the model in handling newly arranged atomic species.

Substitutional alloying using crystal graph neural networks

Substitutional Alloying Using Crystal Graph Neural Networks

Graph Neural Networks for an Accurate and Interpretable Prediction of the Properties of Polycrystalline Materials

A machine learning‐based crystal graph network and its application in development of functional materials

Graph neural network framework for energy mapping of hybrid monte-carlo molecular dynamics simulations of Medium Entropy Alloys

Reinforce Crystal Material Property Prediction with Comprehensive Message Passing Via Deep Graph Networks

Dynamic Graph Attention Network-Based Crystal Space-Enriched Representation for Improving Material Property Prediction

Atomistic Graph Neural Networks for metals: Application to bcc iron

Adaptative Scaler-Moment Crystal Graph Attention Neural Network for Material Property Prediction

Rapid Assessment of Stable Crystal Structures in Single Phase High Entropy Alloys Via Graph Neural Network Based Surrogate Modelling

Enhancing material property prediction with ensemble deep graph convolutional networks

General Graph Neural Network-Based Model To Accurately Predict Cocrystal Density and Insight from Data Quality and Feature Representation

Rapid and Automated Alloy Design with Graph Neural Network-Powered LLM-Driven Multi-Agent Systems

3D-Structure-attention Graph Neural Network for Crystals and Materials

A Novel Neural Network-based Alloy Design Strategy: Gated Recurrent Unit Machine Learning Modeling Integrated with Orthogonal Experiment Design and Data Augmentation

Defect graph neural networks for materials discovery in high-temperature clean-energy applications

Accelerating Defect Predictions in Semiconductors Using Graph Neural Networks

ADA-GNN: Atom-Distance-Angle Graph Neural Network for Crystal Material Property Prediction

Graph Neural Network Modeling of Grain-scale Anisotropic Elastic Behavior using Simulated and Measured Microscale Data

A geometric-information-enhanced crystal graph network for predicting properties of materials

Crystal Graph Convolutional Neural Networks for an Accurate and Interpretable Prediction of Material Properties