Abstract:Deep-learning electronic structure calculations show great potential for revolutionizing the landscape of computational materials research. However, current neural-network architectures are not deemed suitable for widespread general-purpose application. Here we introduce a framework of equivariant local-coordinate transformer, designed to enhance the deep-learning density functional theory Hamiltonian referred to as DeepH-2. Unlike previous models such as DeepH and DeepH-E3, DeepH-2 seamlessly integrates the simplicity of local-coordinate transformations and the mathematical elegance of equivariant neural networks, effectively overcoming their respective disadvantages. Based on our comprehensive experiments, DeepH-2 demonstrates superiority over its predecessors in both efficiency and accuracy, showcasing state-of-the-art performance. This advancement opens up opportunities for exploring universal neural network models or even large materials models.

What problem does this paper attempt to address?

The paper proposes a deep learning framework called DeepH-2, aiming to improve the calculation of electronic structures, especially the prediction of density functional theory (DFT) Hamiltonian. The current neural network architectures have limitations in wide applications, while DeepH-2 overcomes these issues by introducing an equivariant local coordinate transformation framework. This new framework combines the simplicity of local coordinate transformations and the mathematical advantages of equivariant neural networks, avoiding the drawbacks of previous models such as DeepH and DeepH-E3. DeepH-2 outperforms its predecessors in terms of efficiency and accuracy, demonstrating state-of-the-art performance, which opens up new possibilities for exploring general neural network models and large-scale material models. The paper describes how to leverage equivariance to reduce computational complexity, reducing the cost from O(L^6) to O(L^3), which is particularly important for handling high-angular-momentum channels. Additionally, it adopts the classic architecture of the Transformer model, enhancing the network's ability to describe the relationship between material structures and electronic structures. Experimental results in the paper show that DeepH-2 achieves significantly lower average absolute errors (MAE) of DFT Hamiltonian matrix elements compared to DeepH and DeepH-E3 in testing cases of single-layer and bilayer graphene, as well as MoS2. It achieves sub-millielectron-volt accuracy level, demonstrating its efficiency and accuracy in handling large-scale material systems.

DeepH-2: Enhancing deep-learning electronic structure via an equivariant local-coordinate transformer

A deep equivariant neural network approach for efficient hybrid density functional calculations

Deep-learning density functional theory Hamiltonian for efficient ab initio electronic-structure calculation

Deep learning density functional theory Hamiltonian in real space

Efficient hybrid density functional calculation by deep learning

General framework for E(3)-equivariant neural network representation of density functional theory Hamiltonian

Universal materials model of deep-learning density functional theory Hamiltonian

Improving density matrix electronic structure method by deep learning

Deep-learning electronic-structure calculation of magnetic superstructures

Infusing Self-Consistency into Density Functional Theory Hamiltonian Prediction via Deep Equilibrium Models

Deep Neural Network Representation of Density Functional Theory Hamiltonian

Improving the efficiency of ab initio electronic-structure calculations by deep learning

Deep Density: Circumventing the Kohn-Sham equations via symmetry preserving neural networks

Bridging deep learning force fields and electronic structures with a physics-informed approach

Deepks: A Comprehensive Data-Driven Approach Toward Chemically Accurate Density Functional Theory

Neural-network Density Functional Theory Based on Variational Energy Minimization

Universal Machine Learning Kohn-Sham Hamiltonian for Materials

Generalizing deep learning electronic structure calculation to the plane-wave basis