Error message dictionary

Most QE errors arrive as Error in routine <name> (<code>), and the routine name is the strongest clue to the cause.

How to read an error

     %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
     Error in routine  cdiaghg (2):
      problems computing cholesky
     %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

The truly dangerous cases have no error at all; see the section on silent failures.

Input syntax errors

Pseudopotential and density problems

Measured case: in the degauss scan of Example E5, mv smearing with degauss=0.005 on a 12×12×12 grid integrated the charge to 3.003 instead of 3 and stopped with charge is wrong. Smearing width and k-grid must be converged together.

SCF and diagonalization failures

convergence NOT achieved after N iterations

The most common problem. Try these in order.

1. mixing_beta 0.7 → 0.3 → 0.1
2. mixing_mode = 'local-TF' (metals, slabs, magnets)
3. Raise electron_maxstep (200–500)
4. Raise mixing_ndim (8 → 12–16, if memory allows)
5. Temporarily raise degauss, converge, restart with startingpot='file' while lowering it
6. diagonalization = 'cg' or 'ppcg' (slow but robust)
7. Inspect the structure (atoms too close)

For DFT+U with nosym (MD in particular): the SCF stalls because rotations among degenerate orbitals keep the density sloshing. Add mixing_fixed_ns = 30 (measured in E13: stuck at 7×10⁻⁵ after 100 iterations without it, converged in 28 with it).

c_bands: N eigenvalues not converged

• Often ignorable as a warning; repeated occurrences mean diagonalization failure.
• Switch diagonalization, raise nbnd (metals and magnets especially).
• diago_david_ndim 2 → 4 can help (more memory).

cdiaghg: problems computing cholesky / S matrix not positive definite

• The overlap matrix went singular. Usual causes: atoms too close, a bad initial wavefunction, or a linearly dependent basis.
• Try startingwfc = 'random'.
• Recheck the structure: overlapping atoms are more common than you think.

Not enough space allocated for radial FFT

• A very large cell, or atoms straddling the cell boundary.
• Raise cell_factor or move atoms inside the cell.

checkallsym: some of the original symmetry operations not satisfied

• Atomic motion broke the symmetry detected on the initial structure. In MD you will hit this almost immediately (thermal motion destroys symmetric positions in the first step).
• Set nosym = .true. for MD (Chapter 16). If it appears during a relaxation, the initial symmetry only held to numerical noise; refine the structure or use nosym there too.

Parallel, memory, and I/O problems

Silent failures: the most dangerous category

QE prints physically wrong results in a perfectly clean format. Make these checks habitual.

A checklist for magnetic transition-metal oxides

Recurring issues in systems like FeO, Fe₂O₃, Fe₃O₄:

• Is ecutrho at least 10x ecutwfc? (Fe PAW is demanding)
• Are spin-up and spin-down sublattices split into separate labels?
• Is mixing_beta at 0.3 or below with mixing_mode='local-TF'?
• Did you converge from several initial magnetizations and pick the lowest-energy solution?
• Is the HUBBARD card in the new syntax (no lda_plus_u remnants)?
• Is the projector (ortho-atomic etc.) recorded together with the U value?
• Did you steer the orbital occupations with starting_ns_eigenvalue?
• If you use hp.x, did you converge the nq grid?
• Are cutoffs, k-grid, smearing, and U identical across the whole dataset?

When you ask for help

Search the QE users mailing list archive first. Most problems are already answered, and threads answered by the developers (Giannozzi, Timrov, and others) are effectively official documentation.

Always include:

1. The QE version and how it was built
2. The complete input file
3. The error section of the output, with 30 lines of context
4. What you already tried