Official Resources
- Homepage: https://docs.pybinding.site/
- Documentation: https://docs.pybinding.site/en/stable/
- Repository: https://github.com/dean0x7d/pybinding
- License: BSD 3-Clause License
Overview
Pybinding is a high-performance Python package for numerical tight-binding calculations. It is engineered to handle large-scale systems (millions of atoms) by combining a user-friendly Python interface with a highly optimized C++ core. It excels at constructing arbitrary lattice geometries, applying external fields and disorder, and computing electronic properties using efficient methods like the Kernel Polynomial Method (KPM).
Scientific domain: Mesoscopic Physics, Graphene/2D Materials
Target user community: Researchers simulating transport and spectra in large disordered systems
Theoretical Methods
- Tight-Binding Formalism: Orthogonal tight-binding models.
- Kernel Polynomial Method (KPM): Order-N expansion of spectral functions (Chebyshev polynomials) to compute Density of States (DOS) and conductivity without full diagonalization.
- Exact Diagonalization: Interfaces to LAPACK and ARPACK (sparse) solvers.
- Recursion Methods: For certain transport properties.
Capabilities
- Model Construction:
- Arbitrary crystal lattices (1D, 2D, 3D).
- Finite and periodic systems.
- Complex shapes (nanoribbons, rings, user-defined polygons).
- Modifiers:
- Strain fields.
- Electric and Magnetic fields (Peierls substitution).
- Disorder (vacancy, onsite energy, hopping).
- Observables:
- Band structures.
- Local Density of States (LDOS).
- Transport (Kubo-Greenwood conductivity).
- Berry curvature (experimentally supported).
Key Strengths
- Performance: The C++ backend ensures that constructing the Hamiltonian and performing KPM calculations is extremely fast, comparable to pure C/Fortran codes.
- Ease of Use: The Python API is modern, intuitive, and designed for rapid prototyping (e.g., using decorators for modifiers).
- Visualization: Built-in plotting helpers for lattices, bands, and LDOS maps using Matplotlib.
Inputs & Outputs
- Inputs: Python scripts defining lattice vectors, sublattices, and hoppings.
- Outputs:
Results objects containing NumPy arrays of energies, DOS, etc.- Plots.
Interfaces & Ecosystem
- Dependencies: NumPy, SciPy, Matplotlib.
- Interoperability: Can export matrices to other solvers if needed.
Performance Characteristics
- Scaling: $O(N)$ for KPM methods, allowing systems with $>10^7$ sites on a desktop.
- Parallelism: OpenMP multi-threading in the C++ core.
Comparison with Other Codes
- vs. Kwant: Kwant is the standard for transport (scattering matrix/Landauer); Pybinding is generally faster for spectral properties (DOS/LDOS) of massive systems due to its specialized KPM implementation.
- vs. TB2J: TB2J is for magnetism parameters; Pybinding is for electronic structure.
Application Areas
- Graphene Nanodevices: simulating strain and edge effects.
- Disordered Systems: Anderson localization studies requiring large statistical ensembles.
- Moiré Lattices: Large unit cells of twisted double-bilayer graphene.
Community and Support
- Development: Dean Moldovan.
- Source: GitHub.
Verification & Sources