Bethe Lattice Contacts

Overview: What is a Bethe Lattice Contact?

A Bethe lattice contact is an infinite, periodic approximation to a bulk metallic electrode. Rather than explicitly model atomic layers, the Bethe lattice uses Slater-Koster hopping parameters to approximate the electronic structure of a 3D bulk material. The approach is particularly well-suited for metallic contacts like Au, where accurate band structure and density-of-states near the Fermi level are essential for transport calculations.

When to Use Bethe Lattice Contacts

Bethe lattice contacts are ideal when you need:

  • Metallic electrodes (Au, Fe, Pt, etc.) with well-characterized Slater-Koster parameters

  • Accurate bulk density-of-states without modeling finite atomic layers

  • Self-consistent SCF + transport calculations with minimized contact artifacts

  • Optional spin-orbit coupling for systems with strong relativistic effects

If you are setting up non-metallic contacts or 1D chain electrodes, see Choosing a Contact Type for alternative approaches.

Reference: gauNEGF.surfGBethe.surfGB (NEGFE integration) and gauNEGF.surfGBethe.surfGBAt (standalone Bethe lattice object).

Warning

Contact-atom basis must be minimal. Atoms participating in the Bethe contact slice must use a minimal basis (typically LANL2DZ for the heavy metal). Polarized or diffuse functions on contact atoms produce a poorly-conditioned overlap matrix that breaks the Bethe lattice surface Green’s function construction. Interior (device) atoms can use any basis. See Best Practices for Production Calculations for the conditioning background.

The setContactBethe Call

Syntax

negf.setContactBethe(contactList, latFile='Au', eta=ETA, T=TEMPERATURE)

Parameters

contactList (nested list)

Specifies which atoms belong to each contact. Format: [[left_atoms], [right_atoms]]. Example: [[1,2,3],[7,8,9]] creates a contact where atoms 1–3 couple to the left electrode and atoms 7–9 couple to the right electrode. Each inner list becomes a separate contact.

latFile (string, default=’Au’)

Selects the Slater-Koster parameter file:

  • 'Au' – Gold FCC [111] surface without spin-orbit coupling

  • 'AuSOC' – Gold FCC [111] surface with spin-orbit coupling enabled

Wrong choice leads to incorrect orbital couplings and transport results.

eta (float, default 1e-5 eV)

Broadening parameter for the surface Green’s function. Controls the numerical stability of the contour integration. Use 1e-6 eV for high precision or when testing convergence; 1e-5 eV is standard for production.

T (float, default 0 K)

Temperature in Kelvin. Set to 300 for room-temperature transport; 0 K gives sharp Fermi-Dirac distribution.

Important Notes on setVoltage

Unlike 1D auto-extract contacts, Bethe lattice contacts provide their own Fermi reference energy via calcFermi. You can pass an explicit Fermi energy (e.g. setVoltage(0.0, 0.0)) as a shortcut, or omit it to let the default fermiMethod='muller' run. See Configuration and Tuning for the full method comparison.

Basic Bethe Contact Workflow (spin=’g’, no SOC)

This is the simplest path for Bethe lattice contacts: generalized spin (‘g’) without spin-orbit coupling.

Minimal Example

# adapted from HemeStudies/AuFe.py
from gauNEGF.scfE import NEGFE

fn = 'AuTipFe'
negf = NEGFE(fn=fn, func='b3lyp', basis='chkbasis', spin='g',
             route='integral=grid=superfine')
negf.setContactBethe([[1,2,3],[7,8,9]], 'Au')

# First pass: constraint device charge
negf.setVoltage(0.0)
negf.SCF(1e-2, 0.02, 100)

# Second pass: remove device charge constraint
negf.setVoltage(0.0, negf.fermi)
negf.SCF(1e-3, 0.02, 100)
negf.SCF(1e-3, 0.05, 1000, pulay=False)

# Save density
negf.saveMAT('AuTipFe_ESCF_Bethe_contFermi.mat')

Key Points

  1. Two setVoltage/SCF calls: First without Fermi to ensure contacts set, second with negf.fermi (refine). This two-pass approach improves convergence.

  2. pulay=False: For the first SCF pass, Pulay mixing helps to quickly get to convergence criteria. In open shell systems, the Pulay mixing can get stuck in a loop. Turning it off during second run can help with convergence if the initial run doesn’t converge

SOC Workflow (spin=’g’, latFile=’AuSOC’)

Adding spin-orbit coupling (SOC) changes both the Hamiltonian and the orbital dimensions. Use this path when magnetism or spin-dependent transport is important.

Changes from Non-SOC Bethe

  1. spin must be ‘g’ – Generalized spinor basis (spin=’g’) is required for SOC calculations. This doubles the orbital dimension (N -> 2N). SOC couples spin and orbital angular momentum, which is not representable in restricted (spin=’r’) or unrestricted (spin=’u’) basis. The generalized spinor basis (‘g’) is the only formalism that captures this coupling. Attempting SOC with spin=’r’ or spin=’u’ will produce incorrect results.

  2. latFile=’AuSOC’ – Reads spin-orbit coupling constants (soc_p, soc_d) from the parameter file.

Warm-Start from Prior Run

For long SCF calculations (especially with SOC), save the density matrix and Fermi energy after an initial convergence, then reload for subsequent bias voltages or refinement passes.

Save After Initial Convergence

# Initial 200 SCF cycles with loose convergence
negf.SCF(1e-2, 0.01, 200)

# Save density and Fermi for warm-start
negf.saveMAT('NEGFSystem_loose.mat')

Warm-Start for Refinement

from scipy import io

A = io.loadmat('NEGFSystem_loose.mat')
negf.setDen(A['den'])
negf.setVoltage(0.0, A['fermi'][0][0])
# Run setVoltage again to turn on fermiSearch if needed:
negf.setVoltage(0.0)

# Refine with tighter convergence and smaller damping
# checkpoint=False to ensure density not overwritten
# by checkpoint file ({fn}_P.mat)
negf.SCF(1e-3, 0.02, 200, checkpoint=False)

# Save refined result
negf.saveMAT(f"{fn}_{extra}_refined.mat")

Why Warm-Start?

  • Speed: Avoids re-converging from scratch; typically 3-5x faster.

  • Stability: Starting from a near-converged density reduces oscillations.

  • IV curves: For bias-dependent transport, warm-start from V=0 state then increment voltage.

For the full IV curve sweep workflow following this setup, see Guide 5: Workflow Recipes.