PINN Predicts Vibrational Stability of Semiconductors
Analysis
This paper introduces a Physics-informed Neural Network (PINN) to predict the vibrational stability of inorganic semiconductors, a crucial property for high-throughput materials screening. The key innovation is incorporating the Born stability criteria directly into the loss function, ensuring the model adheres to fundamental physics. This approach leads to improved performance, particularly in identifying unstable materials, which is vital for filtering. The work contributes a valuable screening tool and a methodology for integrating domain knowledge to enhance predictive accuracy in materials informatics.
Key Takeaways
- •PINN is used to predict vibrational stability of inorganic semiconductors.
- •Born stability requirements are integrated into the loss function.
- •The model achieves an F1-score of 0.83 and an AUC-ROC of 0.82.
- •The approach improves the identification of unstable materials.
- •Provides a screening tool and a methodology for incorporating domain knowledge.
“The model shows consistent and improved performance, having been trained on a dataset of 2112 inorganic materials with validated phonon spectra, and getting an F1-score of 0.83 for both stable and unstable classes.”