Compare the labelled path with the full-zone sample
Open the silicon Brillouin-zone path and band plot, then inspect a separate uniform-mesh view or table of the extrema. Identify where each reported valence maximum and conduction minimum occurs and whether it lies on the selected path; if a three-dimensional viewer is available, use it to locate off-path regions rather than relying on a sorted list alone. Use visual and symmetry tools and electronic-property tools.
Inspect two stored real-output samples: the companion script compares an actual QE path dataset with a 260-point time-reversal-reduced sample of the nominal mesh. Check each extremum’s coordinate and dataset before the scalar separation; neither finite sample establishes a converged fundamental gap.
Use this real Silicon comparison when a band-path extremum appears to support a gap or valley statement. It shows the next required operation: calculate a separate full-zone sample from a declared compatible parent and keep the two datasets distinct. The retained outputs have separately recorded hashes, but the public artifacts do not independently prove one uninterrupted historical save-tree ancestry. Neither dataset is an observable-convergence study.
From the repository root, reconstruct the committed comparison first:
python3 examples/practical-guides/silicon_qe_full_zone.pyInspect the path and mesh extrema, their fractional coordinates, k-point counts, input/output hashes, and the warning that both are finite samples.
Purpose
For a new calculation, first establish the compatible SCF parent inside its prepared QE work directory; the following filenames are work-directory names, not repository-root commands:
pw.x -in scf.in > scf.out
grep -E '^[[:space:]]+convergence has been achieved in[[:space:]]+[0-9]+ iterations[[:space:]]*$' scf.out
grep -F "JOB DONE." scf.outThe first check asks whether QE reported electronic convergence; the second checks program termination only.
Run the SeeK-path line input and its bands.x post-processing as described in the reciprocal-path guide. Then create a separate uniform-zone full-zone.in. Use calculation='nscf', the same cell, structure, pseudopotential, cutoffs, charge, spin/SOC state and prefix, and enough bands for the target energy window. This new NSCF stage reads compatible parent SCF data through the matching prefix and outdir, but it is not an interrupted-run continuation; QE 7.5 explicitly excludes NSCF calculations from restart_mode='restart'.
The retained historical full-zone input used restart_mode='restart' and produced the real warning-bearing output audited later on this page. Preserve that artifact as execution evidence, not as the current recipe. The corrected reusable input below keeps the retained model values but has not been claimed as a rerun:
&CONTROL
calculation = 'nscf', restart_mode = 'from_scratch',
prefix = 'si_cod9013102', outdir = './out', pseudo_dir = './pseudo',
/
&SYSTEM
ibrav = 0, nat = 2, ntyp = 1,
ecutwfc = 40.0, ecutrho = 320.0,
occupations = 'fixed', nbnd = 8,
nosym = .true.,
/
&ELECTRONS
conv_thr = 1.0d-10,
/
ATOMIC_SPECIES
Si 28.0855 Si.pbe-n-rrkjus_psl.1.0.0.UPF
CELL_PARAMETERS angstrom
0.0000000000 2.7152000000 2.7152000000
2.7152000000 0.0000000000 2.7152000000
2.7152000000 2.7152000000 0.0000000000
ATOMIC_POSITIONS crystal
Si 0.0000000000 0.0000000000 0.0000000000
Si 0.2500000000 0.2500000000 0.2500000000
K_POINTS automatic
8 8 8 0 0 0nosym is useful when every retained mesh coordinate is required by the downstream extremum search, but it is not a convergence parameter or universal requirement. Give each refinement its own archived input/output and, when concurrent, its own outdir. Run the full-zone state with separate logs:
pw.x -in full-zone.in > full-zone.out 2> full-zone.err; printf '%s\n' "$?" > full-zone.exit
grep -F "JOB DONE." full-zone.out
grep -Ei "warning|error|stopping|not converged|c_bands" full-zone.out full-zone.err || trueIn the stored full-zone case, the mesh route retained 260 time-reversal-reduced k points with eight bands. The completion marker confirms only that the program ended normally; warning review and eigenvalue completeness must pass separately. For a new run, use a parser compatible with the exact QE output or a documented exporter rather than assuming a bands.x path file is the required full-zone representation.
Inspect the sampled extrema
The two-site structure comes from the CC0 COD 9013102 Silicon record. The 141-point SeeK-path line sample has a 0.574 eV separation between its sampled fourth- and fifth-band extrema. The separate mesh sample gives 0.617 eV. The mesh value need not be lower because both are finite, different samples; neither result is a converged full-zone gap.
The stored full-zone-extrema.json records structure identity, program versions, input/output SHA-256 values, retained coordinates, and the two derived extrema. The conceptual drawing below demonstrates only the possibility of an off-path extremum; it is not Silicon data.
Reconstruct the actual-output comparison
The companion command above does not launch QE. Execution verifies reconstruction of two declared real-output samples: it checks the path-output hash and frozen real-output ledger, k-point and band counts, derived extrema, and plot. This is not a continuous rerun from a fresh public scratch directory.
Decide what remains untested
Before assigning a fundamental gap, repeat the uniform-zone calculation with denser meshes while holding the accepted model fixed. Tabulate the valence maximum and conduction minimum energies, fractional coordinates, band indices, indirect separation, direct separation at each sampled k point, and warnings for every mesh. Refine around candidate extrema or use a validated interpolation/search route, then retest the reported separation. Also test cutoffs, parent sampling, number of bands, occupations, structural and magnetic state, and SOC treatment where they affect the claim.
Success requires warning-free electronic solutions and stable extrema under the declared search refinement; it is not defined by JOB DONE. or one scalar gap. If an extremum moves off the path, the path was incomplete evidence. If its coordinate or energy keeps moving with mesh density, continue the full-zone search. If occupied and unoccupied states overlap anywhere, follow the DOS/Fermi-surface route rather than reporting a path gap. A high-symmetry path cannot prove full-zone metallicity, gap directness, or the absence of an off-path pocket. This example does not establish a converged gap, carrier valley, quasiparticle or optical gap, experimental agreement, or Silicon material conclusion.