Inspect every deformed structure and raw point
Open the structures at the smallest, central, and largest sampled volumes, then spot-check the rest. Confirm which lattice ratios, internal coordinates, magnetic state, and symmetry constraints were held or relaxed. Maintain a table linking each visible structure to its input, output, volume, energy, stress, convergence evidence, and method identity; plot raw points before fitting. Visual tools and literature sources help identify unintended transitions and defensible sampling ranges.
Optional ledger check: the abstract series is a synthetic traceability example. Use it after inspecting a real set of deformed structures; it validates record completeness, not an energy-volume calculation or equation of state.
Use this guide before fitting an EOS. It filters an abstract A2B2 point ledger so only one structural and electronic branch reaches the fit.
From the repository root, print the full accepted and excluded ledger:
python3 examples/practical-guides/eos_sampling_protocol.pyThe output is a filtered, ordered ledger. It is not an EOS fit or a DFT result.
What this guide verifies
Define one branch before accepting points. In this exercise, retain only neutral four-atom A2B2 cells evaluated with fixture-method-v1, relaxed under the same fixed-volume shape-and-internal-coordinate policy, and ending in state-A. Keep every rejected point and its reason beside the accepted series.
For a production series, also record the source structure, full cell and strain mapping, allowed degrees of freedom, final coordinates, forces and stress, method and potential identity, numerical settings, software version, state diagnostics, and input/output hashes.
A scalar volume is not a complete structure. Fixed-shape scaling, fixed-volume shape relaxation, internal relaxation, and full hydrostatic relaxation trace different branches.
Inspect the series before fitting
The fixture filters for compatibility, then sorts by volume. Its seven accepted points bracket the sampled minimum at per A2B2 cell. Confirm that lower and higher accepted volumes have higher energies; this brackets the sampled minimum but does not establish the fitted continuous minimum.
The report retains v32-state-switch because it ended in state-B, and v48-unfinished because completion and state identity are missing. Both are excluded from the fit rather than relabelled or deleted.
For real points, inspect symmetry, lattice metrics, atom mapping, magnetic state, charge, occupations, forces, stress, and numerical settings across the ordered series. A discontinuity can be a branch change or numerical artifact. It is not repaired by deleting the point.
Decide whether fitting may begin
Proceed only when:
- accepted volumes are unique and bracket the intended minimum or pressure interval;
- every point uses the same declared deformation and relaxation policy;
- the structural and electronic branch remains identifiable;
- energy differences and stress are converged across the changing cells;
- rejected and failed points remain recorded with reasons.
The companion script verifies metadata filtering, ordering, exclusions, and sampled-minimum bracketing. It does not generate structures, run or relax DFT, test pressure convergence, fit an EOS, or establish structural stability.