Universal materials model of deep-learning density functional theory Hamiltonian

Yuxiang Wang,Yang Li,Zechen Tang,He Li,Zilong Yuan,Honggeng Tao,Nianlong Zou,Ting Bao,Xinghao Liang,Zezhou Chen,Shanghua Xu,Ce Bian,Zhiming Xu,Chong Wang,Chen Si,Wenhui Duan,Yong Xu

DOI: https://doi.org/10.1016/j.scib.2024.06.011

2024-06-15

Abstract:Realizing large materials models has emerged as a critical endeavor for materials research in the new era of artificial intelligence, but how to achieve this fantastic and challenging objective remains elusive. Here, we propose a feasible pathway to address this paramount pursuit by developing universal materials models of deep-learning density functional theory Hamiltonian (DeepH), enabling computational modeling of the complicated structure-property relationship of materials in general. By constructing a large materials database and substantially improving the DeepH method, we obtain a universal materials model of DeepH capable of handling diverse elemental compositions and material structures, achieving remarkable accuracy in predicting material properties. We further showcase a promising application of fine-tuning universal materials models for enhancing specific materials models. This work not only demonstrates the concept of DeepH's universal materials model but also lays the groundwork for developing large materials models, opening up significant opportunities for advancing artificial intelligence-driven materials discovery.

Computational Physics,Materials Science

What problem does this paper attempt to address?

This paper aims to address the problem of implementing large-scale material models, particularly by developing a general material model (DeepH) for density functional theory (DFT) Hamiltonians through deep learning. Currently, DFT calculations are often limited to small-scale material systems due to high computational costs. The paper proposes a feasible approach to create a universal material model capable of handling various elemental compositions and material structures by constructing a large material database and improving the DeepH method, thereby accurately predicting material properties. In addition, the paper demonstrates the potential of fine-tuning the universal material model to enhance the application of specific material models. By using automated infrastructure and database generation of DFT training datasets, and employing advanced equivariant transformation architectures to train neural networks, they successfully trained a universal DeepH model with low average absolute error, laying the foundation for achieving large-scale material models and advancing AI-driven materials discovery.

Universal materials model of deep-learning density functional theory Hamiltonian

Deep learning density functional theory Hamiltonian in real space

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

Universal Machine Learning Kohn-Sham Hamiltonian for Materials

Improving density matrix electronic structure method by deep learning

Deep learning-based density functionals empower AI for materials

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

Deep Neural Network Representation of Density Functional Theory Hamiltonian

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

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

Neural-network Density Functional Theory Based on Variational Energy Minimization

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

Atomistic Modeling of Lithium Materials from Deep Learning Potential with Ab Initio Accuracy

Deep-Learning Database of Density Functional Theory Hamiltonians for Twisted Materials

DPA-2: Towards a universal large atomic model for molecular and material simulation

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

Large Scale and Quantum Accurate Molecular Dynamics Simulation:Liquid Iron under Extreme Condition

Practicing deep learning in materials science: An evaluation for predicting the formation energies