TBPLaS

Official Resources

• Homepage: https://www.tbplas.net/
• Repository: https://github.com/deepmodeling/tbplas
• License: BSD 3-Clause License

Overview

TBPLaS (Tight-Binding Package for Large-scale Simulation) is a high-performance Python package designed for the simulation of electronic structure and quantum transport in macroscopic tight-binding models. Developed by the DeepModeling community, it leverages efficient numerical algorithms—such as the Tight-Binding Propagation Method (TBPM) and Kernel Polynomial Method (KPM)—to perform calculations on systems with millions of atomic orbitals, scaling linearly with system size.

Scientific domain: Large-scale Tight-Binding, Quantum Transport, 2D Materials Target user community: Researchers studying disordered systems, Moiré superlattices, and quasicrystals

Theoretical Methods

• Tight-Binding Propagation Method (TBPM): An $O(N)$ method that calculates time-correlation functions of random states to extract spectral and transport properties without diagonalization.
• Kernel Polynomial Method (KPM): Chebyshev expansion of the density of states and spectral functions.
• Green's Functions: Recursive Green's Function (RGF) for exact transport in smaller systems.
• Exact Diagonalization: Classic solvers for small systems or band structures.

Capabilities

• Observables:
  • Density of States (DOS) and Local DOS.
  • Optical Conductivity $\sigma(\omega)$.
  • DC Conductivity (Kubo formula).
  • Hall Conductivity ($\sigma_{xy}$) and Chern numbers.
  • Polarization and Dielectric function.
  • Quasieigenstates.
• Systems:
  • Graphene, TMDs, and Twisted Bilayers (Moiré).
  • Disordered alloys (Anderson localization).
  • Fractal lattices and quasicrystals.

Key Strengths

• Scalability: Capable of simulating $>10^7$ atoms on a single workstation, thanks to the linear scaling ($O(N)$) of TBPM/KPM.
• Performance: Critical kernels are optimized in Cython/Fortran and parallelized with OpenMP/MPI.
• Ease of Use: Object-oriented Python API allows for intuitive system construction and analysis.
• Integration: Part of the DeepModeling ecosystem (DeepMD-kit), facilitating ML-driven potentials/Hamiltonians.

Inputs & Outputs

• Inputs:
  • Lattice and Hamiltonian definitions (Python objects).
  • Simulation parameters (time steps, energy resolution).
• Outputs:
  • Spectral data (DOS, conductivity) as NumPy arrays.
  • Visualization files.

Interfaces & Ecosystem

• ASE: Interface to Atomic Simulation Environment for structure manipulation.
• DeepModeling: Potential for future integration with Deep Wannier/Deep Hamiltonian.

Performance Characteristics

• Speed: TBPM is orders of magnitude faster than Exact Diagonalization for large $N$.
• Efficiency: Sparse matrix operations reduce memory footprint.

Comparison with Other Codes

• vs. KITE: Both use linear scaling methods (KPM/TBPM). KITE emphasizes "disorder on the fly" and C++ backend; TBPLaS offers a pure Python-centric workflow and includes the powerful propagation method (TBPM).
• vs. Kwant: Kwant is better for open systems with leads (scattering); TBPLaS excels at bulk properties of massive disordered systems using spectral methods.

Application Areas

• Twistronics: Electronic properties of twisted bilayer graphene at magic angles (thousands of atoms per cell).
• Anderson Localization: Scaling theory of localization in 2D/3D disordered lattices.
• Quantum Hall Effect: Calculating Chern numbers in large topological systems.

Community and Support

• Development: DeepModeling Community (Yuan Ping Feng group / Contributors).
• Source: GitHub.

Verification & Sources

• Website: https://www.tbplas.net/
• Primary Publication: Y. Li et al., arXiv:2209.00806 (2022).
• Verification status: ✅ VERIFIED
  • Active and modern research tool.