sisl

Official Resources

• Homepage: https://zerothi.github.io/sisl/
• Documentation: https://zerothi.github.io/sisl/docs/latest/
• Source Repository: https://github.com/zerothi/sisl
• License: Mozilla Public License 2.0 (MPL-2.0)

Overview

sisl is a high-performance, modern Python framework for electronic structure calculations and large-scale tight-binding modeling. Developed as a successor to the utility scripts of TranSIESTA, it has evolved into a general-purpose API that interfaces with multiple DFT codes (SIESTA, VASP, OpenMX, Wannier90, BigDFT) to manipulate Hamiltonians, geometries, and real-space grids. It is particularly renowned for its ability to handle extremely large sparse matrices, enabling tight-binding calculations on systems with millions of orbitals which are inaccessible to standard dense-matrix tools.

Scientific domain: Quantum Transport, Large-scale Tight-Binding, DFT Post-processing Target user community: Users of SIESTA/TranSIESTA, researchers modeling large nanostructures, and developers needing a robust localized-basis API

Theoretical Methods

• Tight-Binding (TB) / LCAO: Core data structure is the sparse Hamiltonian matrix $H_{ij}$ in a localized basis.
• Green's Functions: Interfaces with TBtrans to compute transport properties using Non-Equilibrium Green's Functions (NEGF).
• Real-Space Grid Algebra: Efficient manipulation of charge densities and potentials on 3D grids.
• Geometry Operations: Algorithms for constructing supercells, adding strain, creating vacancies, and building complex heterostructures.

Capabilities

• Input/Output:
  • Reads/Writes: SIESTA (.fdf, .XV, .TSHS), VASP (POSCAR, KPOINTS), OpenMX, Wannier90 (_hr.dat), BigDFT, XYZ, PDB.
  • Binary file support for high-speed I/O (NetCDF).
• Analysis:
  • Electronic: Band structure, Density of States (DOS), Projected DOS (PDOS), COOP/COHP.
  • Wavefunctions: Real-space plotting of orbitals and wavefunctions.
  • Transport: Analysis of transmission spectra and currents (with TBtrans).
• Model Construction:
  • Create Graphene/TMD nanoribbons, nanotubes, and flakes.
  • Twist bilayers, apply strain, and create defects.
  • Construct tight-binding models from scratch with nearest-neighbor rules.

Key Strengths

• Scalability: Optimized C/Cython backend allows handling of sparse matrices with millions of rows/columns, far exceeding the capacity of numpy or scipy.sparse for physics workflows.
• Interoperability: Acts as a "Rosetta Stone" converting between different DFT formats (e.g., VASP structure to SIESTA input).
• Transport Workflow: Tightly integrated with TBtrans, providing a seamless pythonic interface for setting up and analyzing complex transport calculations.

Inputs & Outputs

• Inputs: Almost any standard structure or Hamiltonian file from supported DFT codes.
• Outputs:
  • Normalized files for other codes.
  • Data arrays (Bands, DOS) for plotting.
  • Visualization files (VMD, XSF, Cube).

Interfaces & Ecosystem

• Upstream:
  • SIESTA: Deep integration (reads dense/sparse matrices, grids).
  • Wannier90: Can read Hamiltonians for large-scale transport setup.
• Downstream:
  • TBtrans: The primary transport solver associated with sisl.
  • KITE: Interface available for quantum transport.

Performance Characteristics

• Speed: Critical sections (Hamiltonian construction, neighbor searching) are written in Cython/C.
• Memory: Sparse matrix storage (CSR/CSC) ensures minimal memory footprint for large empty systems.
• Parallelism: OpenMP threading for computationally intensive matrix operations.

Limitations & Known Constraints

• Basis Limitation: Strictly localized basis sets (LCAO, Wannier, TB); not suitable for plane-wave data (except for grid operations).
• Solver: sisl itself is primarily a manipulator/analyzer; it diagonalizes small/medium matrices but relies on external codes (TBtrans) for heavy inversion/transport tasks on large systems.

Comparison with Other Codes

• vs. ASE (Atomic Simulation Environment): ASE focuses on geometry and calculators; sisl focuses on the Hamiltonian and Electronic Structure matrices themselves, offering much deeper access to the physics of the model.
• vs. PythTB: PythTB is excellent for small toy models; sisl is built for "production" large-scale DFT-derived tight-binding models.

Application Areas

• Nanodevices: Modeling FETs, molecular junctions, and interconnects.
• 2D Materials: Twisted bilayers (Moiré physics) and defect engineering.
• Topology: Constructing large hamiltonians for topological invariant analysis (e.g., with Kite).

Community and Support

• Development: Maintained by Nick Papior (DTU/eScience).
• Documentation: Extensive documentation and tutorials.
• Forum: Active Discord channel and GitHub discussions.

Verification & Sources

• Official Website: https://zerothi.github.io/sisl/
• Primary Publication: N. Papior et al., Comp. Phys. Comm. 212, 8 (2017) (TBtrans/sisl paper).
• Verification status: ✅ VERIFIED
  • widely used in the TranSIESTA community.

  • 400 citations.