DeePTB is an innovative Python package that employs deep learning to construct electronic tight-binding (TB) Hamiltonians with a minimal basis. It is designed to:
- Efficiently predict TB Hamiltonians for large, unseen structures based on training with smaller ones.
- Enable simulations of large systems under structural perturbations, finite temperature simulations integrating molecular dynamics (MD) for comprehensive atomic and electronic behavior.
- Support customizable Slater-Koster parameterization with neural network incorporation for local environmental corrections.
- Operate independently of the choice of bases and exchange-correlation functionals, offering flexibility and adaptability.
- Handle systems with strong spin-orbit coupling (SOC) effects.
DeePTB is a versatile tool adaptable for a wide range of materials and phenomena, providing accurate and efficient simulations. See more details in our DeePTB paper: arXiv:2307.04638
Installing DeePTB is straightforward. We recommend using a virtual environment for dependency management.
- Python 3.8 or later.
- Torch 1.13.0 or later (PyTorch Installation).
- ifermi (optional, for 3D fermi-surface plotting).
- Ensure you have Python 3.8 or later and Torch installed.
- Install DeePTB with pip:
pip install dptb
- Clone the repository:
git clone https://github.com/deepmodeling/DeePTB.git
- Navigate to the root directory and install DeePTB:
cd DeePTB pip install .
For a comprehensive guide and usage tutorials, visit our documentation website.
DeePTB joins the DeepModeling community, a community devoted of AI for science, as an incubating level project. To learn more about the DeepModeling community, see the introduction of community.
We welcome contributions to DeePTB. Please refer to our contributing guidelines for details.
When utilizing the DeePTB package in your research, we request that you cite the following reference:
Gu, Qiangqiang, et al. "DeePTB: A deep learning-based tight-binding approach with ab initio accuracy." arXiv preprint arXiv:2307.04638 (2023).