Formation energy prediction of material crystal structures using deep learning
Source abstract
Abstract
Abstract Determining the stability of chemical compounds is essential for advancing material discovery. In this study, we introduce a novel deep neural network model designed to predict a crystal’s formation energy, which identifies its stability property. Our model leverages elemental fractions derived from material composition and incorporates the crystallographic symmetry (space group) and stability labels (ground state(stable), metastable, unstable) as an additional input feature. The materials’ symmetry classifications represent the crystal polymorphs and are crucial for understanding phase transitions in materials. The inclusion of symmetry information markedly enhanced the model’s predictive accuracy. At the same time, the addition of stability labels further improved performance, demonstrating that combining compositional, structural, and thermodynamic descriptors yields a more robust prediction framework. These findings underscore the crucial role of crystallographic symmetry and stability information in enhancing the accuracy and interpretability of deep learning models for predicting materials stability.
Source: crossref · Quality: baseline
Research context
OpenAlex topics: Machine Learning in Materials Science · X-ray Diffraction in Crystallography · Crystallography and molecular interactions
SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.
References (26)
Show source-provided references
1. Materials design and discovery with high-throughput density functional theory: the open quantum materials database (OQMD)10.1007/s11837-013-0755-4 ↗
2. The thermodynamic scale of inorganic crystalline metastability10.1126/sciadv.1600225 ↗
3. Commentary: the materials project: a materials genome approach to accelerating materials innovation10.1063/1.4812323 ↗
4. The Open Quantum Materials Database (OQMD): assessing the accuracy of DFT formation energies10.1038/npjcompumats.2015.10 ↗
5. Matminer: an open source toolkit for materials data mining10.1016/j.commatsci.2018.05.018 ↗
6. Deep neural networks for accurate predictions of crystal stability10.1038/s41467-018-06322-x ↗
7. Elemnet: deep learning the chemistry of materials from only elemental composition10.1038/s41598-018-35934-y ↗
8. Predicting materials properties without crystal structure: deep representation learning from stoichiometry10.1038/s41467-020-19964-7 ↗
9. Atomic-position independent descriptor for machine learning of material properties10.1103/PhysRevB.98.214112 ↗
10. Including crystal structure attributes in machine learning models of formation energies via Voronoi tessellations10.1103/PhysRevB.96.024104 ↗
11. Machine-learning-accelerated high-throughput materials screening: discovery of novel quaternary Heusler compounds10.1103/PhysRevMaterials.2.123801 ↗
12. Formation energy prediction of crystalline compounds using deep convolutional network learning on voxel image representation10.1038/s43246-023-00433-9 ↗
13. A geometric-information-enhanced crystal graph network for predicting properties of materials10.1038/s43246-021-00194-3 ↗
14. Materials discovery through machine learning formation energy10.1088/2515-7655/abe425 ↗
15. Pretraining strategies for structure agnostic material property prediction10.1021/acs.jcim.3c00919 ↗
16. Graph neural networks predict energetic and mechanical properties for models of solid solution metal alloy phases10.1016/j.commatsci.2023.112141 ↗
17. Python materials genomics (pymatgen): a robust, open-source python library for materials analysis10.1016/j.commatsci.2012.10.028 ↗
18. Mentel Ł M 2024 Mendeleev—A Python Package with Properties of Chemical Elements, Ions, Isotopes and Methods to Manipulate and Visualize Periodic Table [Software] Zenodo ( https://doi.org/10.5281/zenodo.11237837 ) 10.5281/zenodo.1123783710.5281/zenodo.11237837 ↗
19. A critical examination of compound stability predictions from machine-learned formation energies10.1038/s41524-020-00362-y ↗
20. Aggarwal · 2018
21. Deep sparse rectifier neural networks
22. A method for stochastic optimization
23. Adaptive subgradient methods for online learning and stochastic optimization
24. Lecture 6.5-rmsprop: divide the gradient by a running average of its recent magnitude
25. A comprehensive review of machine-learning approaches for crystal structure/property prediction10.3390/cryst15110925 ↗
26. A framework to evaluate machine learning crystal stability predictions10.1038/s42256-025-01055-1 ↗