Inspect calculation records before sorting energies
Open the input, output, accepted structure, and state record for every candidate. In a spreadsheet, mark composition, cell size, charge, magnetic identity, evaluator, energy convention, completion, and relevant convergence evidence; reject incomparable rows before ranking anything. Open structural outliers in a viewer and return to the original paper or Methods section when state identity is unclear. See visual tools and literature sources for common inspection routes.
Optional arithmetic check: the abstract A2B3 ledger is a synthetic teaching object. Use it only after inspecting the real calculation records above; it demonstrates a comparability gate and supplies no material ranking.
Use this guide before subtracting total energies from different calculation records. The fixture contains two accepted abstract A2B3 candidates and three deliberately incompatible rows.
From the repository root, print the complete report:
python3 examples/practical-guides/formation_energy_ledger.pyThe report is the output to inspect. It records accepted rows, rejected rows and reasons, normalized energies, and the bounded comparison. It does not run an electronic-structure program.
What this guide verifies
Filter the rows before ranking. In this exercise, accept only neutral A2B3 rows evaluated with fixture-method-v1, carrying electronic_energy, and marked both complete and final-state verified. A row that fails any one of those comparisons stays visible with its exclusion reason.
For production data, add structure and state identifiers, method and potential identity, numerical settings, geometry status, correction scheme, software version, artifact hashes, and target-convergence evidence. A common unit is not enough to make two rows comparable.
Normalize before ranking
The accepted alpha cell contains two formula units and the accepted beta cell contains one. The script converts both totals to eV per A2B3 formula unit before selecting a reference. It also reports eV per atom from the explicit five-atom formula unit.
Inspect that cell contents and denominators are present in every accepted row. Never rank raw cell totals from different cell sizes.
Read exclusions as evidence
The report rejects:
charged-a2b3for a different total charge;mixed-method-a2b3for a different evaluator;unfinished-a2b3because completion and final-state identity are unverified.
Do not delete these rows. Their exclusion reasons define the boundary of the candidate comparison and expose whether a lower-looking number was omitted for a scientific reason.
Within the accepted fixture, alpha-a2b3 defines zero and beta-a2b3 is higher by eV per formula unit, or eV per atom. These invented values validate filtering and arithmetic only.
Decide whether the ledger can continue
Continue to a reaction, formation energy, EOS, or hull only when every accepted row shares the required composition or balance, charge, evaluator, energy field, state identity, correction convention, and numerical adequacy. Retain the raw value and normalized value together.
This guide does not test DFT execution, numerical convergence, method accuracy, global structural search, formation energy, or phase stability. A deterministic ledger result is software evidence, not a material conclusion.