Best Practices for Production Calculations

This section provides guidelines for setting up and running production-quality NEGF-DFT calculations, focusing on accuracy, efficiency, and troubleshooting.

System Preparation

Geometry Optimization

1. Initial Structure

   • Use standard bond lengths and angles

   • Consider symmetry for efficient calculations

   • Optimize geometry without contacts first

2. Contact Placement

   • Choose chemically reasonable contact sites

   • Maintain consistent contact-molecule distances

   • Consider multiple contact configurations

3. Basis Set Selection

   • Start with a compact basis like LANL2DZ for testing

   • Use larger basis sets for production

   • Check basis set superposition error

DFT Setup

Functional Selection

Honestly, there are hundreds of DFT functionals out there. Find one that works for your system and doesn’t make the reviewers too mad. People seem to like pure functionals for metals and hybrid functionals for organics, if you want both you are SOL.

Contact Models

1. Energy-Independent

2. Bethe Lattice

   # Use realistic metallic contacts with extended system
   negf = scfE.NEGFE('molContacts')
   # Assuming triangular contacts on 1,2,3,4 and 5,6,7,8
   inds = setContactBethe([[1,2,3],[6,7,8]], latFile='Au2', eta=1e-5, T=300)
   

3. 1D Chain

   # For molecular wire systems
   negf = scfE.NEGFE('molContacts')
   # Assuming repeating infinite chain extending atoms [1,2] and [3,4]
   inds = setContact1D([[2],[3]], [[1],[4]], eta=1e-5, T=300)
   

Convergence Strategies

SCF Convergence

1. Mixing Parameters

   # Start with default mixing values (values over 0.05 will be unstable!)
   negf.SCF(damping=0.02, maxcycles=200)
   
   # Lower mixing if SCF is unstable
   negf.SCF(damping=0.005, maxcycles=400)
   
   # turn off pulay mixing if cyclical convergence values are observed (convergence will take longer)
   negf.SCF(damping=0.02, maxcycles=1000, pulay=False)
   

2. Pulay Mixing

   # Pulay mixing as implemented works well
   # Increasing nPulay can increase convergence speed for difficult systems
   negf = NEGF(fn='system', nPulay=9)
   # Side effect may be instability, consider setting pulay=False if stability is an issue
   negf.SCF(damping=0.02, maxcycles=1000, pulay=False)
   

3. Change Initial Wavefunction Guess

   negf.setDen(density_guess)
   # Typical values for convergence, don't read checkpoint
   negf.SCF(conv=1e-4, damping=0.02, maxcycles=300, checkpoint=False)
   

Integration Parameters

Note that integration is only used by the NEGFE() class and by default is adaptive:

1. Adaptive Integration used by default*

Change adapative integration tolerance in config.py, which is calculated as the maximum density matrix difference between two consecutive SCF cycles:

# Convergence Tolerances
ADAPTIVE_INTEGRATION_TOL = 1e-3     # Adaptive integration tolerance

2. Manually Set Integration Grid*

   # Set grid size and Emin
   negf.setIntegralLimits(
        N1=100, #Integration from Emin to mu
        N2=50, #Integration from Eminf to Emin
        Emin=-500
   )
   

3. Add Temperature

   # Include finite temperature (300 Kelvin)
   # even for energy-independent contacts
   negf.setSigma([1], [2], sig=-0.05j, T=300)
   

Validation Checks

1. Zero Bias

   • Compare with literature conductance values

   • Check Transmission profile between HOMO-LUMO gap

   • Verify DOS features (molecular orbitals vs contact effects)

2. Finite Bias

   • Check current symmetry with positive and negative bias

   • Check current hysteresis with increasing convergence

   • Monitor charge conservation

3. Spin Systems

   • Check charge and multiplicity

   • Check spin contamination

   • For non-collinear cases check system spin direction