Multitask learning with semiempirical orbital charges enables sample-efficient MLIPs

2026-05-22

AI Summary

This work addresses the challenge of high data requirements in machine learning interatomic potentials (MLIPs) for simulating complex materials. The authors propose a multitask learning framework that, for the first time, incorporates linear-scaling GFN1-xTB orbital-resolved charges as an auxiliary task to enhance the model’s physical awareness of electronic structure while preserving computational efficiency. A specially designed equivariant neural network predicts these orbital charges, substantially improving sample efficiency: the approach achieves comparable accuracy with only one-fifth of the training data and reduces the mean absolute error in energy predictions by 46%, outperforming models trained with costly DFT-derived charges. Latent space analysis further reveals that the model naturally clusters metallic elements according to their chemical properties, demonstrating strong physical interpretability.

Abstract

Machine learning interatomic potentials (MLIPs) require generating computationally expensive, large-scale training datasets to accurately simulate materials and molecules. Incorporating electronic structure information using multitask learning improves sample efficiency, however, training on full Hamiltonian matrices, which scale quadratically with the number of atoms, is intractable for large datasets. In this work, we show that multitask learning utilizing orbitally resolved semiempirical charges significantly improves sample efficiency and accuracy in MLIPs. To efficiently predict orbital charges, we implement a specialized equivariant model, reducing charge prediction error compared to an invariant baseline. By augmenting training with computationally inexpensive GFN1-xTB orbital charges, which scale linearly with the number of atoms, our model achieves a 46\% reduction in energy mean absolute error and requires five times less data to match the performance of energy-only models. Furthermore, our approach outperforms models trained on expensive density functional theory (DFT) atomic charges, capturing orbitally resolved electronic complexity and forcing the network to learn a physically accurate latent space that spontaneously clusters metals by shared chemical properties. Because orbital charges are only required during training, this approach preserves inference efficiency, providing a scalable recipe for developing accurate, data-efficient foundation models for complex chemical systems.

Problem

• machine learning interatomic potentials
• sample efficiency
• electronic structure
• orbital charges
• multitask learning

Innovation

• multitask learning
• semiempirical orbital charges
• machine learning interatomic potentials
• equivariant model
• sample efficiency