Learning Piezoelectric Tensors Through Strain-Conditioned Polarization Clusters

367.0 KB of text data describes a physics-informed cluster graph neural network (PCGNN) for predicting piezoelectric tensors. The model, developed by Chunlin Yu, applies controlled strain perturbations and reconstructs macroscopic tensors through symmetry-consistent aggregation of local polarization clusters. On the Materials Project test set, PCGNN achieved a mean absolute error of 0.135 C/m², outperforming baseline models by 32.5% and 52.6%.

Use Cases

• Training or benchmarking machine learning models for piezoelectric tensor prediction based on the described PCGNN architecture.
• Screening unlabeled materials for strong piezoelectric response as mentioned in the description.
• Studying symmetry-consistent aggregation methods for tensorial property prediction based on the strain–polarization response pathway.

Strengths

• The described PCGNN model achieved a mean absolute error of 0.135 C/m² on the Materials Project test set.
• The model outperformed EATGNN and CGCNN baseline models by 32.5% and 52.6%, respectively, as stated in the description.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The dataset is 367.0 KB, indicating a very limited scope, likely containing model description or results rather than a large-scale training corpus.

Provenance

Source

figshare

Collection Method

Likely contains research outputs, model descriptions, or results related to the described PCGNN method.

Freshness

Last updated 2026-05-08 15:36:16; freshness should be verified.