Slater-Koster and EDTB Module#

The Slater-Koster with environmental dependence (EDTB) module gives you tight-binding model construction and fitting tools that go beyond the standard two-center approximation. By including environment-dependent screening corrections, it can capture configuration-dependent effects that matter in surfaces, defects, and strained structures.

Module Layout#

PAOFLOW.defs.models — Model builders:

  • SK_EDTB: The environment-dependent extension. Modulates hopping integrals through a scalar screening factor that depends on the local atomic environment.

  • Slater_Koster: Builds generalized two-center tight-binding models up to third nearest neighbors.

  • Pre-built models for common systems (graphene, cubium variants) to get you started quickly.

PAOFLOW.defs.sk_fitting — Parameter fitting:

  • SKFitter: Single-geometry, eigenvalue-based fitting.

  • SKFitterEDTB: Extends fitting to include the screening parameters (γ values).

  • MultiGeomEDTB: Fits shared parameters across multiple atomic configurations simultaneously — useful when you want a model that transfers across different geometries.

PAOFLOW.defs.edtb_params — Model serialization via the EDTBModel class. Models serialize to JSON, so they’re easy to store, share, and load across different simulation setups.

PAOFLOW.defs.surface_project — Projects bulk band structures onto surface planes, identifying absolute gaps and lens-shaped features.

PAOFLOW.defs.dual_params — Dual-parameter models that give bulk and surface atoms independent parameter sets, which is useful when the bonding at the surface differs significantly from the bulk.

PAOFLOW.defs.band_unfold — Supercell-to-primitive-cell band unfolding with spectral weight analysis.

Both routes feed unfold_bands with a model dict: EDTBModel.to_model_dict() for a tight-binding model and paoflow_to_model_dict(data_controller) for a projected DFT run (after pao_hamiltonian()). The two store the real-space Hamiltonian in opposite Fourier conventions. Model Hamiltonians hold the hopping to tau_B + R, so H(k) = sum_R H(R) exp(+2 pi i k.R); the projected Hamiltonian is HRs = ifftn(Hks), so H(k) = sum_R H(R) exp(-2 pi i k.R) (hopping at tau_B - R). Band energies are blind to the difference (with time reversal the two give H(k) and H(-k)), spectral weights are not: with the sign wrong the weights select the folded branches of the mirror point -k, with the sum rule still satisfied. unfold_bands therefore reads the fourier_sign entry of a _paoflow dict (-1, set by paoflow_to_model_dict) and flips R before the transform; a dict without the entry is classified from the decay of the hoppings with bond length (_detect_fourier_sign, the same test as sparse_hamiltonian.detect_bond_sign). Note that PAOFLOW.bands() along a path uses exp(+2 pi i k.R) on the projected HRs and so evaluates H(-k): identical energies for time-reversal-symmetric scalar systems, mirrored paths otherwise.

The EDTB Idea#

Traditional two-center Slater-Koster models assume that the hopping integral between two atoms depends only on their separation and orbital character. The EDTB formalism relaxes this by introducing a screening factor that accounts for the atoms in between — the mediating environment. This makes it possible to build models that stay accurate across configurations where the local environment changes, without needing a separate parameterization for each one.