Achieving self-consistent field (SCF) convergence is often the trickiest part of running density functional theory (DFT) calculations, especially for metallic or highly correlated systems. As a regular Quantum ESPRESSO user, I've compiled below some of my favorite strategies and settings for overcoming tricky SCF convergence issues.

1. Choose a Good Initial Guess

• Setup Matters: Quantum ESPRESSO provides several ways to initialize the wavefunctions and electron density. Use

  startingwfc = 'atomic'
  startingpot = 'atomic'

  in your &CONTROL or &SYSTEM inputs to begin with atomic orbitals.
• From Previous Runs: If continuing from a converged calculation, set

  restart_mode = 'restart'

  and ensure your old *.save directories are available.
• Spin Guessing: For magnetic systems, specify the starting_magnetization for each atomic species.

2. Tweak the Mixing Scheme

• Default is Not Always Best: The default mixing (mixing_mode = 'plain', mixing_beta = 0.7) can fail in metals or large systems.
• Broyden Mixing: Try

  mixing_mode = 'local-TF'

  or

  mixing_mode = 'damp'

• Lower Mixing Beta: Reduce mixing_beta (try 0.3 or even 0.1) to damp oscillations in problematic cases.

Sample:

&SYSTEM
    mixing_mode = 'local-TF',
    mixing_beta = 0.2,
/

• Increase Mixing Step: Sometimes, setting

  mixing_ndim = 8

  builds a longer memory of past density changes and helps convergence.

3. Check K-Points and Smearing

• Smearing Is Essential for Metals: Use a Methfessel-Paxton or Marzari-Vanderbilt smearing in occupations:

  occupations = 'smearing'
  smearing = 'mp'
  degauss = 0.02   ! Try varying degauss for best results

• Γ-Only for Insulators: For molecules, try occupations = 'fixed'.
• K-Point Mesh: Start with a coarse mesh to get a feel for the system, then refine.

4. Adjust Electronic Minimization

• Convergence Threshold: Set a more lenient conv_thr at first, then tighten once convergence is achieved, e.g., from 1.0e-5 to 1.0e-8.
• Maximum Iterations: The default (electron_maxstep=100) may not suffice. Set

  electron_maxstep = 200

5. Diagnose Stuck Calculations

If SCF oscillates or diverges:

• Try a different mixing scheme.
• Lower your cutoff energies (ecutwfc, ecutrho) to test for stability.
• Revisit your pseudopotentials; some are less stable than others.
• Check if the structure needs relaxation—distant atoms or unreasonable lattice constants can cause trouble.

6. Useful Input Snippet

Here's a snippet with convergence-aiding settings:

&SYSTEM
  nosym = .true.
  occupations = 'smearing'
  smearing = 'mv'
  degauss = 0.01
  mixing_mode = 'local-TF'
  mixing_beta = 0.3
  mixing_ndim = 10
/
&ELECTRONS
  conv_thr = 1.0e-7
  electron_maxstep = 200
/

Summary Table

Final Thoughts

Don't be afraid to experiment systematically—changing one variable at a time and keeping detailed logs. For persistent problems, the Quantum ESPRESSO forums and documentation are invaluable resources.

What are your go-to tricks for SCF convergence? Share in the comments below!

#QuantumEspresso #DFT #Convergence #MaterialsScience