DeePTB: A deep learning-based tight-binding approach with $ab$ $initio$ accuracy

Qiangqiang Gu,Zhanghao Zhouyin,Shishir Kumar Pandey,Peng Zhang,Linfeng Zhang,Weinan E

2023-07-11

Abstract:Simulating electronic behavior in materials and devices with realistic large system sizes remains a formidable task within the $ab$ $initio$ framework. We propose DeePTB, an efficient deep learning-based tight-binding (TB) approach with $ab$ $initio$ accuracy to address this issue. By training with $ab$ $initio$ eigenvalues, our method can efficiently predict TB Hamiltonians for unseen structures. This capability facilitates efficient simulation of large-size systems under external perturbations like strain, which are vital for semiconductor band gap engineering. Moreover, DeePTB, combined with molecular dynamics, can be used to perform efficient and accurate finite temperature simulations of both atomic and electronic behavior simultaneously. This is demonstrated by computing the temperature-dependent properties of a GaP system with $10^6$ atoms.

Materials Science,Computational Physics

What problem does this paper attempt to address?

The paper aims to address the challenges faced in simulating large-scale electronic behavior in materials science using ab initio methods. Specifically, the research team has proposed a deep learning framework called DeePTB, which can efficiently predict tight-binding (TB) Hamiltonians with ab initio level accuracy, thereby enabling effective simulation of large systems.

Key features of the DeePTB method include:
1. **Ab initio level accuracy**: By training on energy eigenvalues obtained from ab initio calculations, DeePTB can efficiently predict TB Hamiltonians for unseen structures, maintaining ab initio level accuracy.
2. **Capability to handle large systems**: DeePTB, combined with molecular dynamics simulations, can efficiently and accurately simulate atomic and electronic behavior at finite temperatures, such as calculating temperature-dependent properties in a GaP system containing millions of atoms.
3. **Addressing issues of traditional TB methods**: Traditional TB methods rely on empirical parameters, whose accuracy and transferability are often questioned; while ab initio TB methods improve accuracy, their transferability to new structural configurations remains limited. DeePTB overcomes these limitations through deep learning, achieving a balance between accuracy and efficiency.
4. **Flexibility**: DeePTB can adapt to different basis function choices and exchange-correlation functionals, and can also handle systems with strong spin-orbit coupling effects.

The paper demonstrates the application of DeePTB on different materials (such as group IV elements and III-V compounds) and validates its performance in molecular dynamics trajectories, large systems, and strain-affected structures. Additionally, DeePTB shows good adaptability to different basis sets, exchange-correlation functionals, and spin-orbit coupling effects. These characteristics collectively indicate that DeePTB is an efficient and accurate tool that can significantly advance computational research in the field of materials science.