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

Graph neural networks (GNNs) have been used previously for identifying new crystalline materials. However, geometric structure is not usually taken into consideration, or only partially. Here, we develop a geometric-information-enhanced crystal graph neural network (GeoCGNN) to predict the properties of crystalline materials. By considering the distance vector between each node and its neighbors, our model can learn full topological and spatial geometric structure information. Furthermore, we incorporate an effective method based on the mixed basis functions to encode the geometric information into our model, which outperforms other GNN methods in a variety of databases. For example, for predicting formation energy our model is 25.6%, 14.3% and 35.7% more accurate than CGCNN, MEGNet and iCGCNN models, respectively. For band gap, our model outperforms CGCNN by 27.6% and MEGNet by 12.4%.

GeoCGNN: A Geometric-Information-Enhanced Crystal Graph Network for Property Predictions

Graph neural networks (GNNs) have been explored to search for novel crystal materials. But in previous works, geometric structure was not taken into consideration or incompletely. Here, we develop a geometric-information-enhanced crystal graph neural network (GeoCGNN) to predict properties of novel crystal materials. By considering the distance vector between each node and its neighbors, our model can learn full topologic and spatial geometric structure information. Furthermore, we incorporate an effective method based on the mixed basis functions to encode the geometric information into our model, which outperforms other CGNN methods in a variety of databases. As for predicting the formation energy, our model is 30.3%, 14.6% and 13% better than CGCNN, MEGNet and iCGCNN models, respectively. For band gap, our model outperforms respectively 27.6% and 15.2% than CGCNN and MEGNet. Also, we interpret the implied material properties of the learned graph vector in a visible way.

Benchmarking materials property prediction methods: the Matbench test set and Automatminer reference algorithm

We present a benchmark test suite and an automated machine learning procedure for evaluating supervised machine learning (ML) models for predicting properties of inorganic bulk materials. The test suite, Matbench, is a set of 13 ML tasks that range in size from 312 to 132k samples and contain data from 10 density functional theory-derived and experimental sources. Tasks include predicting optical, thermal, electronic, thermodynamic, tensile, and elastic properties given a material’s composition and/or crystal structure. The reference algorithm, Automatminer, is a highly-extensible, fully automated ML pipeline for predicting materials properties from materials primitives (such as composition and crystal structure) without user intervention or hyperparameter tuning. We test Automatminer on the Matbench test suite and compare its predictive power with state-of-the-art crystal graph neural networks and a traditional descriptor-based Random Forest model. We find Automatminer achieves the best performance on 8 of 13 tasks in the benchmark. We also show our test suite is capable of exposing predictive advantages of each algorithm—namely, that crystal graph methods appear to outperform traditional machine learning methods given ~104 or greater data points. We encourage evaluating materials ML algorithms on the Matbench benchmark and comparing them against the latest version of Automatminer.