Predicting Lattice Vibrational Frequencies Using Deep Graph Neural Networks.

Lattice vibrational frequencies are related to many important materials properties such as thermal and electrical conductivity as well as superconductivity. However, computational calculation of vibrational frequencies using density functional theory methods is computationally too demanding for large number of samples in materials screening. Here we propose a deep graph neural network based algorithm for predicting crystal vibrational frequencies from crystal structures. Our algorithm addresses the variable dimension of vibrational frequency spectrum using the zero padding scheme. Benchmark studies on two data sets with 15,000 mixed-structure and 35,552 rhombohedra samples show that the aggregated R 2 scores of the prediction reach 0.554 and 0.724. We also evaluate the structural transferability by predicting the vibration frequencies for 239 individual cubic target structures. The R 2 scores for more than 40% of the targets are greater than 0.8 and can reach as high as 0.98 for the model trained with mixed samples, while the average mean absolute error is 43.69 Thz showing low transferability across structure types. Our work demonstrates the capability of deep graph neural networks to learn to predict lattice vibration frequency when sufficient number of training samples are available.

Accurate Prediction of Voltage of Battery Electrode Materials Using Attention-Based Graph Neural Networks.

Performing first-principles calculations to discover electrodes' properties in the large chemical space is a challenging task. While machine learning (ML) has been applied to effectively accelerate those discoveries, most of the applied methods ignore the materials' spatial information and only use predefined features: based only on chemical compositions. We propose two attention-based graph convolutional neural network techniques to learn the average voltage of electrodes. Our proposed methods, which combine both atomic composition and atomic coordinates in 3D-space, improve the accuracy in voltage prediction significantly when compared to composition-based ML models. The first model directly learns the chemical reaction of electrodes and metal ions to predict their average voltage, whereas the second model combines electrodes' ML predicted formation energy (Eform) to compute their average voltage. Our Eform-based model demonstrates improved accuracy in transferability from our subset of learned Li ions to Na ions. Moreover, we predicted the theoretical voltage of 10 NaxMPO4F (M = Ti, Cr, Fe, Cu, Mn, Co, and Ni) fluorophosphate battery frameworks, which are unavailable in the Material Project database. It could be shown that we can expect average voltages higher than 3.1 V from those Na battery frameworks except from the NaTiPO4F and TiPO4F pair of electrodes, which offer an average voltage of 1.32 V.

Scalable deeper graph neural networks for high-performance materials property prediction.

Machine-learning-based materials property prediction models have emerged as a promising approach for new materials discovery, among which the graph neural networks (GNNs) have shown the best performance due to their capability to learn high-level features from crystal structures. However, existing GNN models suffer from their lack of scalability, high hyperparameter tuning complexity, and constrained performance due to over-smoothing. We propose a scalable global graph attention neural network model DeeperGATGNN with differentiable group normalization (DGN) and skip connections for high-performance materials property prediction. Our systematic benchmark studies show that our model achieves the state-of-the-art prediction results on five out of six datasets, outperforming five existing GNN models by up to 10%. Our model is also the most scalable one in terms of graph convolution layers, which allows us to train very deep networks (e.g., >30 layers) without significant performance degradation. Our implementation is available at https://github.com/usccolumbia/deeperGATGNN.

Graph convolutional neural networks with global attention for improved materials property prediction.

The development of an efficient and powerful machine learning (ML) model for materials property prediction (MPP) remains an important challenge in materials science. While various techniques have been proposed to extract physicochemical features in MPP, graph neural networks (GNN) have also shown very strong capability in capturing effective features for high-performance MPP. Nevertheless, current GNN models do not effectively differentiate the contributions from different atoms. In this paper we develop a novel graph neural network model called GATGNN for predicting properties of inorganic materials. GATGNN is characterized by its composition of augmented graph-attention layers (AGAT) and a global attention layer. The application of AGAT layers and global attention layers respectively learn the local relationship among neighboring atoms and overall contribution of the atoms to the material's property; together making our framework achieve considerably better prediction performance on various tested properties. Through extensive experiments, we show that our method is able to outperform existing state-of-the-art GNN models while it can also provide a measurable insight into the correlation between the atoms and their material property. Our code can found on - https://github.com/superlouis/GATGNN.