This article presents a new “Tripartite Interaction Representation Algorithm” for crystal graph neural networks (CGNNs). This algorithm enhances existing models by explicitly incorporating atoms, bond lengths, and bond angles into the crystal structure representation. It improves the prediction of material properties, particularly the formation energy of crystalline compounds, by capturing higher-order interactions and updating edge vectors. The framework demonstrates superior accuracy and generalization compared to traditional models, offering better insights into atomic environments for material science research.
Please continue reading the full article under the link below:
https://www.nature.com/articles/s41598-024-76309-w
Please consult also the Quantum Server Marketplace platform for the outsourcing of computational science R&D projects to external expert consultants through remote collaborations:
#materials #materialsscience #materialsengineering #computationalchemistry #modelling #chemistry #researchanddevelopment #research #MaterialsSquare #ComputationalChemistry #Tutorial #DFT #simulationsoftware #simulation