Official Resources
- Homepage: https://pages.nist.gov/ThreeBodyTB.jl/
- Repository: https://github.com/usnistgov/ThreeBodyTB.jl
- License: MIT License
Overview
ThreeBodyTB.jl is a high-accuracy tight-binding package developed by NIST. It distinguishes itself from standard Slater-Koster codes by including pre-fit three-body interaction terms, which dramatically improves transferability and accuracy for structures far from equilibrium (e.g., surfaces, defects, high pressure). Implemented in pure Julia, it provides a self-consistent field (SCF) solver that rivals DFT accuracy (specifically PBEsol) for a vast range of elemental and binary systems, but at a fraction of the computational cost.
Scientific domain: Materials Science, High-Throughput Screening
Target user community: Researchers constructing phase diagrams or running large-scale MD
Theoretical Methods
- Tight-Binding with 3-Body terms (TB3): $H = H_{2body} + H_{3body}$. The 3-body terms capture environmental dependence of bonding.
- Self-Consistent Field (SCF): Solves for charge redistribution/transfer, essential for ionic systems and surfaces.
- Parameterization:
- Uses a massive pre-computed database of parameters fitted to DFT data.
- Covers >99% of ICSD prototypes for supported elements.
- Magnetic Moments: Collinear spin-polarized calculations.
Capabilities
- Simulations:
- SCF Ground state energy and forces.
- Band structures and Density of States (DOS).
- Molecular Dynamics (MD) trajectories.
- Phonon spectra (via finite displacement).
- System Support:
- Bulk crystals (Metals, Insulators, Semiconductors).
- Low-dimensional systems (Surfaces, 2D materials).
- Charged defects.
- Ease of Use: "Automatic" mode that guesses initial parameters and symmetry.
Key Strengths
- Accuracy: Benchmarks show energy/force errors comparable to DFT-PBEsol, far superior to traditional non-SCF tight-binding.
- Database: Comes "batteries included" with parameters for most common elements, removing the need for users to perform their own fitting.
- Speed: $O(N)$ sparse matrix operations allow routine simulation of 1000+ atom supercells.
Inputs & Outputs
- Inputs:
- Crystal structure (POSCAR, CIF).
- List of elements.
- Outputs:
- Energies, Forces, Stress tensor.
- Band plots.
Interfaces & Ecosystem
- Input Generation: Uniquely integrated with
CrystalStructure.jl logic (internal) for handling symmetry.
- Visualization: Built-in plotting recipes.
Performance Characteristics
- Efficiency: Exploits Julia's Type system and SIMD; heavily optimized sparse solvers.
- Parallelism: Multi-threaded execution.
Comparison with Other Codes
- vs. DFTB+: Both aim for "DFT quality" tight-binding. ThreeBodyTB.jl's unique selling point is the explicit 3-body term (better for structural relaxation) and its modern, hackable Julia codebase.
- vs. xTB: xTB is semi-empirical and great for molecules/organics. ThreeBodyTB.jl is parameterized specifically for solid-state crystals and materials science.
Application Areas
- Phase Stability: Calculating formation energies of competing crystal polymorphs.
- Defects: Simulating large supercells to study vacancy/interstitial formation energies without finite-size errors.
Community and Support
- Development: Kevin Garrity (NIST).
- Source: GitHub.
Verification & Sources