Worked research case

Aluminium Metallic Electronic-Structure Workflow

Follow an explicit fcc Aluminium teaching model through metallic setup, real QE SCF/full-zone/DOS branches, convergence and E(V) screens, plots, and bounded interpretation.

Scientific objective and starting object

Follow an explicit fcc Aluminium teaching model through metallic numerical choices, SCF and full-zone branches, DOS, real convergence and E(V) screens, and a bounded interpretation.

Current claim boundary: This bounded case does not establish one case-wide historical launcher, a converged DOS, Fermi surface, equation of state, elastic property, carrier density, transport result, universal parameter recommendation, or a material-level conclusion.

Work across the interface that fits the task: use a browser for source records and manuals, VESTA or another structure viewer for visual inspection, a text editor for inputs, and the terminal or scheduler for execution. The Software Bridge points users of other codes to their native artifacts.

Open the starting sources

Automated download is optional. First identify what the source represents, what metadata must return with it, and what still requires human inspection.

Explicit fcc Aluminium teaching model

structure

Human action: Open the record in a browser when one is declared, read its metadata and source publication, download the structure, inspect the text, and visualize the cell before conversion.

No external record is declared: treat this as an explicit teaching model and do not infer database or experimental provenance.

Exact source identity and case files
  • fcc primitive cell
  • celldm(1)=7.653 bohr
  • nat=1
  • explicit teaching input; not a database identity

Boundary: The explicit input is the bounded reference geometry. It is not a relaxed, database-ranked, or experimentally validated structure.

Public Quantum ESPRESSO Aluminium UPF receipt

pseudopotential

Human action: Open the official library or file record, inspect method compatibility and metadata, then acquire and preserve the exact file used.

Exact source identity and case files
  • filename Al.pbe-n-rrkjus_psl.1.0.0.UPF
  • bytes 1500731
  • SHA-256 cc4f5dc6afe09c8f482dc7645e6e7cca546a55f8d907c71c825c62bf85a38d3e
  • verified 2026-08-09
  • UPF header generator atomic v6.3

Boundary: The exact public URL, byte count, and hash establish file identity only, not transferability, cutoff convergence, or scientific suitability.

Follow the calculation

At each stage, open the real object, inspect what a researcher would actually look at, make the bounded decision, and then use a command where the terminal is the natural interface. Return to an earlier stage whenever a model, method, reference state, target calculation, or validation check fails. This case order is not a universal DFT sequence.

  1. Identify the structure source

    Recognize that this case starts from an explicit teaching input with no database or experimental identity, then inspect the fcc structure before calculation.

    Open the declared database, paper, or teaching record. Read its composition, cell, provenance, revision, conditions, and access boundary before downloading or converting anything. Read the source structure as text and visualize it. Check periodicity, cell choice, occupancies, missing or disordered sites, short contacts, orientation, and whether the object matches the intended material.

    Continue when: Accept the source identity for model construction, or return to the database, paper, Supplementary Information, or another source class.

    Run this step: Inspect the explicit Aluminium source record

    Run from: repository-root

    case_root=examples/cases/aluminium-metallic-electronic-structure
    cat "$case_root/source/fixture-metadata.json"
    sed -n '1,160p' "$case_root/input/scf.in"

    Boundary: The case begins from an explicit teaching model. There is no database or experimental source identity to infer.

    Exact files and recorded evidence for this stage
    • input/scf.in — SCF input; SHA-256 660983565543f2bc154fe3d79a7bf480b2e02b9ae705d706457c661a64087ac0

    Boundary: The structure is an explicit one-atom fcc teaching input, not a database or experimental identity.

  2. Inspect the computational model

    Visualize and inspect the one-atom fcc primitive model shared by the electronic routes, including its cell and periodic interpretation.

    Open both the source structure and the exact computational cell. Compare composition, atom count, lattice, orientation, periodicity, symmetry, vacuum where relevant, and the mapping between conventional, primitive, or modified cells.

    Continue when: Proceed only when the transformed model still represents the intended physical system and every deliberate modification is recorded.

    Run this step: Compare the cell and atomic model across route inputs

    Run from: repository-root

    case_root=examples/cases/aluminium-metallic-electronic-structure
    for input in scf.in nscf.in bands.in; do printf '\n### %s\n' "$input"; grep -nE 'ibrav|celldm|nat|ntyp|ATOMIC_SPECIES|ATOMIC_POSITIONS|K_POINTS' "$case_root/input/$input"; done

    Boundary: Matching recorded cell labels support route consistency only. The geometry was not relaxed or independently validated.

    Exact files and recorded evidence for this stage
    • input/scf.in — SCF input; SHA-256 660983565543f2bc154fe3d79a7bf480b2e02b9ae705d706457c661a64087ac0
    • input/nscf.in — NSCF input; SHA-256 1fdef0bc99e977e96b5d7cec1b750fe009170253a2953f197212c01103ff259d
    • input/bands.in — bands input; SHA-256 62a0406cd30f6b85c358e2f8efb76aab1ee79f8d4fbc494329e9d5c906c53141

    Boundary: celldm(1)=7.653 bohr is a recorded teaching geometry, not an independently optimized or universal value.

  3. Obtain and verify the pseudopotential

    Open the public Quantum ESPRESSO potential record, inspect its metadata and method compatibility, then acquire the exact UPF and reject any byte mismatch.

    Open the library or file record before using a download command. Read the family, exchange–correlation compatibility, valence configuration, relativistic treatment, provenance, and recommended cutoff information. Compare those choices with the physical method and elements in the model; then retain the exact file identity.

    Continue when: Choose a compatible family for a controlled convergence study. A matching hash establishes identity, not suitability.

    Run this step: Download and verify the exact Aluminium potential

    Run from: repository-root

    set -euo pipefail
    case_root=examples/cases/aluminium-metallic-electronic-structure
    QE_PSEUDO_DIR=/absolute/path/to/pseudos
    upf="$QE_PSEUDO_DIR/Al.pbe-n-rrkjus_psl.1.0.0.UPF"; partial="$upf.part"
    cat "$case_root/source/pseudopotential-metadata.json"; test ! -e "$upf"; test ! -e "$partial"
    mkdir -p "$QE_PSEUDO_DIR"
    curl --fail --location --proto '=https' --tlsv1.2 --output "$partial" 'https://pseudopotentials.quantum-espresso.org/upf_files/Al.pbe-n-rrkjus_psl.1.0.0.UPF'
    test "$(wc -c < "$partial" | tr -d '[:space:]')" = '1500731'
    printf '%s  %s\n' 'cc4f5dc6afe09c8f482dc7645e6e7cca546a55f8d907c71c825c62bf85a38d3e' "$partial" | sha256sum --check -
    mv "$partial" "$upf"

    Boundary: The byte count and SHA-256 establish the acquired file identity only. Retain the stale qe_plan URL conflict; identity does not establish transferability or numerical suitability.

    Exact files and recorded evidence for this stage

    Boundary: Use the newer exact URL/bytes/hash receipt and retain the stale qe_plan conflict; file identity does not establish scientific suitability.

  4. Prepare the Quantum ESPRESSO inputs

    Inspect the metallic occupations, smearing, cutoff, sampling, prefix, and output ancestry before execution.

    Read every input that will be submitted, including the structure, calculation type, physical method, numerical settings, prefix, scratch path, and parent artifacts. Identify which choices came from the scientific model, which are numerical parameters to test, and which paths or defaults are site-specific.

    Continue when: Submit only an input whose choices and expected outputs you can explain and audit.

    Run this step: Read the metallic inputs and ancestry before execution

    Run from: repository-root

    case_root=examples/cases/aluminium-metallic-electronic-structure
    for input in scf.in nscf.in dos.in bands.in; do printf '\n### %s\n' "$input"; sed -n '1,180p' "$case_root/input/$input"; done

    Boundary: Confirm prefix, outdir, calculation type, occupations, smearing, cutoffs, k sampling, and dos.x window. The displayed settings remain unaccepted teaching values.

    Exact files and recorded evidence for this stage
    • input/scf.in — SCF input; SHA-256 660983565543f2bc154fe3d79a7bf480b2e02b9ae705d706457c661a64087ac0
    • input/nscf.in — NSCF input; SHA-256 1fdef0bc99e977e96b5d7cec1b750fe009170253a2953f197212c01103ff259d
    • input/dos.in — real dos.x input; SHA-256 879806b9ba78e52d0d5fe17d29312bd0a29d592faa527fee8a900f3ec4c50a8f
    • input/bands.in — bands input; SHA-256 62a0406cd30f6b85c358e2f8efb76aab1ee79f8d4fbc494329e9d5c906c53141

    Boundary: PBE, 30/240 Ry, Marzari-Vanderbilt smearing, degauss=0.02 Ry, and 8x8x8 are recorded settings, not converged prescriptions.

  5. Run the calculation

    Execute SCF, full-zone NSCF, dos.x, and path bands in the declared isolated order.

    Read the launcher or job script, working directory, resource request, output paths, and site policy before submission. Watch scheduler state and the complete stdout/stderr without mistaking new text, a queue state, or a termination marker for scientific convergence.

    Continue when: Continue, diagnose, or stop only from the identified job, output, ownership, and failure layer.

    Run this step: Run SCF, full-zone NSCF, dos.x, and path bands

    Run from: repository-root

    set -euo pipefail
    case_root=examples/cases/aluminium-metallic-electronic-structure
    QE_PW=${QE_PW:?Set QE_PW to the absolute path to pw.x}
    QE_DOS=${QE_DOS:?Set QE_DOS to the absolute path to dos.x}
    QE_PSEUDO_DIR=/absolute/path/to/pseudos
    RUN_OUTPUT_ROOT=/absolute/new/path/aluminium-electronic-route
    test -x "$QE_PW"; test -x "$QE_DOS"; test -d "$QE_PSEUDO_DIR"; test ! -e "$RUN_OUTPUT_ROOT"
    QE_PW="$QE_PW" QE_DOS="$QE_DOS" QE_PSEUDO_DIR="$QE_PSEUDO_DIR" RUN_OUTPUT_ROOT="$RUN_OUTPUT_ROOT" bash "$case_root/run.sh"

    Boundary: The runner refuses missing executables, a missing or wrong-hash UPF, an existing destination, or a destination inside committed case evidence. A new run creates new evidence.

    Exact files and recorded evidence for this stage
    • output/recorded-commands.txt — route command and provenance boundary; SHA-256 d0093f79bff0dfd78b82af05c44bea72b4431675c1bd3371cb895df97ea43fc1

    Boundary: Only isolated-dos-route is continuous. The initial captured outputs and later screens are separate routes, and a new run creates new evidence.

  6. Check program and solver outcomes

    Inspect exits, JOB DONE markers, stderr, SCF convergence, retained warnings, and expected artifacts separately.

    Open the complete stage, not only a grep result or final line. Read iteration behaviour, warnings, termination context, forces or stress where applicable, and whether the intended artifacts were actually written.

    Continue when: Classify program completion, solver convergence, optimization convergence, and artifact completeness separately.

    Run this step: Inspect the fresh process records and retained warnings

    Run from: repository-root

    RUN_OUTPUT_ROOT=/absolute/new/path/aluminium-electronic-route
    for output in scf.out nscf-full.out dos.out bands.out; do test -s "$RUN_OUTPUT_ROOT/$output"; printf '\n### %s\n' "$output"; grep -nE 'JOB DONE|convergence has been achieved|c_bands|Error in routine' "$RUN_OUTPUT_ROOT/$output" || true; done
    for err in scf.err nscf-full.err dos.err bands.err; do wc -c "$RUN_OUTPUT_ROOT/$err"; done
    test -s "$RUN_OUTPUT_ROOT/al.dos"; wc -l "$RUN_OUTPUT_ROOT/al.dos"

    Boundary: Normal termination, electronic SCF convergence, retained c_bands warnings, stderr, and expected artifacts are separate observations. None establishes observable convergence.

    Exact files and recorded evidence for this stage

    Boundary: Normal exits and a DOS file do not erase c_bands warnings or establish observable convergence.

  7. Test the required numerical convergence

    Read the real k-mesh/smearing screen as a bounded negative teaching result.

    Open the calculated series and its predeclared comparison quantity before reading the pass/fail summary. Look for a stable tail, oscillation, state changes, outliers, and whether only one parameter varied while the physical model stayed fixed.

    Continue when: Accept only the tested observable and tolerance; otherwise extend or redesign the series without erasing the adverse points.

    Run this step: Read the adverse convergence result before proceeding

    Run from: repository-root

    case_root=examples/cases/aluminium-metallic-electronic-structure
    cat "$case_root/derived/aluminium-convergence-assessment.json"
    sed -n '1,40p' "$case_root/derived/aluminium-convergence-matrix.csv"

    Boundary: The next scientific action is to redesign and run a convergence study suited to the intended observable. Do not convert this FAIL into program failure, observable convergence, or scientific rejection.

    Exact files and recorded evidence for this stage

    Assessment: FAIL

    Observable convergence: not tested

    Boundary: Five SCF programs completed, but the exploratory screen failed its own energy/Fermi thresholds. This is not a program failure, an observable-convergence result, or scientific rejection.

  8. Identify the reference geometry

    State explicitly which geometry feeds the electronic routes when no relaxation is part of the case.

    Open the starting and final structures side by side with the final force, stress, constraint, and optimizer records. Look for unexpected reconstruction, broken symmetry, short contacts, cell distortion, or a visually plausible structure that still fails a numerical criterion.

    Continue when: Accept the geometry only within the declared optimization conditions; visual plausibility alone is not convergence or stability.

    Exact files and recorded evidence for this stage
    • input/scf.in — SCF input; SHA-256 660983565543f2bc154fe3d79a7bf480b2e02b9ae705d706457c661a64087ac0

    Observable convergence: not applicable

    Boundary: No ionic or cell relaxation is included. The reference is the explicit one-atom fcc teaching geometry, not a relaxed, ranked, or experimentally validated state.

  9. Calculate the reference SCF state

    Establish the route-scoped electronic state used by the full-zone NSCF and downstream branches.

    Open the state, occupations, spin treatment, energy, residual history, warnings, and parent geometry together. Look for slow or oscillatory convergence, state changes, inconsistent occupations, and whether the output belongs to the accepted model.

    Continue when: Use the state as a parent only when the electronic criterion and reference-state logic are both adequate.

    Run this step: Inspect the fresh reference SCF

    Run from: repository-root

    RUN_OUTPUT_ROOT=/absolute/new/path/aluminium-electronic-route
    test -s "$RUN_OUTPUT_ROOT/scf.out"; wc -c "$RUN_OUTPUT_ROOT/scf.err"
    grep -nE 'convergence has been achieved|JOB DONE|Error in routine' "$RUN_OUTPUT_ROOT/scf.out"

    Boundary: This checks the fresh SCF process and reported electronic convergence only. The explicit geometry was not relaxed, and target observables remain unconverged.

    Exact files and recorded evidence for this stage
    • input/scf.in — SCF input; SHA-256 660983565543f2bc154fe3d79a7bf480b2e02b9ae705d706457c661a64087ac0
    • output/dos-route/scf.out — rerun SCF stdout; SHA-256 c3e3e16a1fc50fd1d93103b5cebb84da380e462579ad59154388d3a6fe73a741
    • output/dos-route/scf.err — rerun SCF stderr; SHA-256 e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855

    Boundary: The isolated SCF reports electronic convergence, not structural relaxation, target-observable convergence, or scientific acceptance.

  10. Calculate bands and DOS

    Follow the continuous SCF to full-zone NSCF to native dos.x route and the later path-band branch.

    Open the parent SCF/NSCF state, path or full-zone sampling record, energy reference, number of bands, and the resulting bands or DOS data. Check path labels, Fermi reference, crossings or gaps, full-zone limitations, DOS normalization, broadening, warnings, and sampling density.

    Continue when: Use only the observable supported by the sampled domain and its own convergence evidence; a high-symmetry path is not a full-zone proof.

    Run this step: Inspect the fresh native DOS and path outputs

    Run from: repository-root

    RUN_OUTPUT_ROOT=/absolute/new/path/aluminium-electronic-route
    test -s "$RUN_OUTPUT_ROOT/al.dos"; sed -n '1,12p' "$RUN_OUTPUT_ROOT/al.dos"; tail -n 5 "$RUN_OUTPUT_ROOT/al.dos"
    grep -nE 'c_bands|JOB DONE|Error in routine' "$RUN_OUTPUT_ROOT/nscf-full.out" "$RUN_OUTPUT_ROOT/bands.out" || true

    Boundary: The native table and path stdout are fresh artifacts from the declared route. One sampling choice, normal exit, and a nonempty table do not establish DOS, band, or Fermi-surface convergence.

    Exact files and recorded evidence for this stage
    • input/nscf.in — NSCF input; SHA-256 1fdef0bc99e977e96b5d7cec1b750fe009170253a2953f197212c01103ff259d
    • input/dos.in — real dos.x input; SHA-256 879806b9ba78e52d0d5fe17d29312bd0a29d592faa527fee8a900f3ec4c50a8f
    • input/bands.in — bands input; SHA-256 62a0406cd30f6b85c358e2f8efb76aab1ee79f8d4fbc494329e9d5c906c53141
    • output/dos-route/nscf-full.out — rerun NSCF stdout; SHA-256 fd648c510ee3ff4e1e92be2d71fd5e53743ee0b6295f14d8acbbcd60ed94ded1
    • output/dos-route/nscf-full.err — rerun NSCF stderr; SHA-256 e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
    • output/dos-route/dos.out — rerun dos.x stdout; SHA-256 c930a07ec54184fe102d308668a5895b629087fdd977ce36b9a602e705616f6e
    • output/dos-route/dos.err — rerun dos.x stderr; SHA-256 e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
    • output/dos-route/al.dos — real QE dos.x table; SHA-256 95c3342cb9229b0ff2fbc8cb17d48aa13ebd6f3ddea7a5a1c4797de9e64be5f8
    • output/dos-route/bands.out — rerun bands stdout; SHA-256 8affb038e5e4fbf4a2250933fd8a6689c2f4d26c26f87dd8cad84bec163670ba
    • output/dos-route/bands.err — rerun bands stderr; SHA-256 e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855
    • derived/al-dos-x.csv — real dos.x parsed table; SHA-256 0b52929e545395687b05ab12e79920cf0cd9f6b125533afd00d886396a1a1a9a
    • derived/al-dos-x-metadata.json — real dos.x metadata; SHA-256 db44e2f32f8512ff4ae3cd3281af3c06515fbfd947a112cf5851adab05f206a0

    Observable convergence: not tested

    Boundary: The save tree is not published, NSCF and bands warnings remain visible, and one 8x8x8 DOS sample does not establish a converged DOS or Fermi surface.

  11. Rebuild tables and plots

    Open the native dos.x, raw-NSCF, convergence, and bounded E(V) views with their source tables, inspect what each representation shows, then regenerate them from committed evidence if needed.

    Open each real figure beside its underlying calculated values. Check axes, units, reference zeros, calculated points, fits or smoothing, warnings, and features that change with representation.

    Continue when: Retain figures that expose the scientific evidence and its limitations; a regenerated image is not an independently validated result.

    Total density of states parsed from the real QE dos.x Aluminium output and referenced to its printed Fermi energy; no DOS-convergence claim is made.
    Native QE dos.x table plotted from the hash-bound al.dos output.. Read it together with the source table and the stage boundary below. Open the source figure.
    Five real Aluminium SCF energy points versus primitive-cell volume with a bounded quadratic interpolation; not a converged EOS, bulk modulus, or elastic tensor.
    Bounded five-point E(V) teaching fit after the volume-declaration parser correction.. Read it together with the source table and the stage boundary below. Open the source figure.

    Run this step: Inspect the stored tables behind the retained figures

    Run from: repository-root

    case_root=examples/cases/aluminium-metallic-electronic-structure
    sed -n '1,20p' "$case_root/derived/al-dos-x.csv"
    sed -n '1,30p' "$case_root/derived/aluminium-convergence-matrix.csv"
    cat "$case_root/derived/aluminium-eos-fit.json"

    Boundary: These are stored route-scoped tables and a bounded fit record, not plots derived from the new runtime. Inspect the fresh al.dos directly until a reviewed fresh plotting route is supplied.

    Exact files and recorded evidence for this stage
    • figures/al-dos-x.png — real dos.x plot; SHA-256 7e9664006cca6277ddcbea3936125d5f5aa9c1791734dd0dd6c42aadf6215d14
    • figures/aluminium-eos-fit.png — real bounded E(V) fit plot; SHA-256 7a209ca0c4795ca77f5e838161c4b1cf39345dcc54cfdbe039b2f887579e8c3f

    Boundary: The Gaussian representation, native dos.x plot, sampling histogram, and E(V) fit are distinct derived views; none is convergence evidence by itself.

  12. Interpret within the evidence boundary

    Read route-scoped figures and tables together with raw warnings, the exploratory FAIL, relevant literature, and explicit non-claims.

    Compare the calculated object with independent calculations or measurements only after matching structure, state, conditions, and observable definition. Look for contradictions, method sensitivity, missing alternatives, and conclusions controlled by an unconverged or untested stage.

    Continue when: State the narrowest supported claim and the evidence that would change it.

    Run this step: Write the conclusion from the adverse record

    Run from: repository-root

    case_root=examples/cases/aluminium-metallic-electronic-structure
    cat "$case_root/derived/aluminium-convergence-assessment.json"
    cat "$case_root/derived/aluminium-eos-fit.json"
    cat "$case_root/derived/captured-run-summary.json"

    Boundary: Report the exploratory convergence screen as FAIL and observable convergence as not tested. The bounded E(V) fit is mathematical evidence only; the adverse screen is not scientific rejection.

    Exact files and recorded evidence for this stage

    Assessment: Exploratory convergence screen FAIL; bounded E(V) fit mathematical PASS only

    Observable convergence: not tested

    Boundary: The case does not support a converged DOS, Fermi surface, EOS, elastic property, carrier density, transport result, universal parameter recommendation, or material conclusion.

  13. Preserve the reproducible record

    Keep the human decision record, model inspection notes, route-scoped inputs, outputs, figures, hashes, commands, source receipts, parsers, warnings, and non-claims together.

    Collect the human decision history together with sources, inputs, outputs, tables, figures, warnings, failed attempts, and software context. Confirm that another researcher can identify what was done, what was viewed, what was excluded, and which steps can or cannot be reconstructed.

    Continue when: Freeze a versioned study object while keeping hashes and manifests as supporting identity evidence rather than the scientific narrative.

    Run this step: Write a bounded record beside the fresh route

    Run from: repository-root

    set -euo pipefail
    RUN_OUTPUT_ROOT=/absolute/new/path/aluminium-electronic-route
    record="$RUN_OUTPUT_ROOT/STUDY-NOTES.txt"; sums="$RUN_OUTPUT_ROOT/SELECTED-SHA256SUMS"
    test -s "$RUN_OUTPUT_ROOT/scf.in"; test -s "$RUN_OUTPUT_ROOT/scf.out"; test -s "$RUN_OUTPUT_ROOT/nscf-full.out"; test -s "$RUN_OUTPUT_ROOT/al.dos"; test -s "$RUN_OUTPUT_ROOT/bands.out"; test ! -e "$record"; test ! -e "$sums"
    printf '%s\n' 'Question: inspect one fresh fcc-Al SCF to full-zone NSCF to dos.x route plus path bands.' 'Boundary: no relaxation is included; DOS, bands, and Fermi-surface convergence are not tested.' 'Stored adverse result: the separate exploratory convergence screen is FAIL, not program failure or scientific rejection.' > "$record"
    sha256sum -- "$RUN_OUTPUT_ROOT/scf.in" "$RUN_OUTPUT_ROOT/scf.out" "$RUN_OUTPUT_ROOT/nscf-full.out" "$RUN_OUTPUT_ROOT/dos.in" "$RUN_OUTPUT_ROOT/dos.out" "$RUN_OUTPUT_ROOT/al.dos" "$RUN_OUTPUT_ROOT/bands.in" "$RUN_OUTPUT_ROOT/bands.out" > "$sums"
    cat "$record"; cat "$sums"

    Boundary: This preserves selected fresh route inputs, outputs, the human-readable evidence boundary, and byte identities. It does not relabel the stored exploratory screen, create observable convergence, or validate a scientific conclusion.

    Exact files and recorded evidence for this stage
    • output/recorded-commands.txt — route command and provenance boundary; SHA-256 d0093f79bff0dfd78b82af05c44bea72b4431675c1bd3371cb895df97ea43fc1
    • output/compact-source-excerpt.txt — deterministic source-bound excerpt rendered by the Worked Workflow; SHA-256 7535de34523ac74287437773555e1a29eb430634d7c9e07ba8d3ba5d48a08d2d

    Boundary: The public record preserves assembled route evidence and its conflict/non-claim boundaries, not one case-wide historical launcher or unpublished save-tree ancestry.

Reproducibility appendix

Open this only when you need the exact published case record, hashes, historical commands, or replay boundary.

How to use this workflow

Evidence boundary
Program completion, SCF convergence, ionic convergence, expected-artifact presence, target-observable convergence, and scientific support are separate decisions.
Continuity boundary
The public case assembles several real QE 7.5 routes. Only isolated-dos-route is continuous; the case as a whole does not claim one historical launcher, case-wide shell-exit transcript, execution interval, or cross-route save-tree ancestry.
Claim boundary
This bounded case does not establish one case-wide historical launcher, a converged DOS, Fermi surface, equation of state, elastic property, carrier density, transport result, universal parameter recommendation, or a material-level conclusion.

Declared continuous execution routes

Execution contexts

Local interactive execution
Use when: Only when local and site policy permits calculation on the current machine and sufficient resources are available.. Boundary: Never assume that an SSH login shell or cluster login node is a permitted compute context. The published Aluminium run.sh invokes the supplied executables directly and does not claim a multi-rank launcher.
Slurm allocation or batch execution
Use when: Only inside an active compute allocation or a site-approved batch job.. Boundary: Use the site's documented launcher and resource policy. Do not present the current direct-executable run.sh as a reviewed multi-rank Slurm wrapper.

Inspect stored evidence or run a fresh calculation

Inspect stored evidence

This track reads the published, hash-bound case records and does not launch a new DFT calculation.

Reconstruct the stored public evidence

case_root=examples/cases/aluminium-metallic-electronic-structure
(cd "$case_root" && python3 parse.py)
(cd "$case_root" && bash extract.sh)
(cd "$case_root" && bash check.sh)

Output policy: Rebuild case-local tables, figures, summaries, and isolated-copy checks from the committed route-scoped outputs.

Audit

  • Retain the NSCF and bands c_bands warnings.
  • Keep the Gaussian raw-NSCF representation distinct from native dos.x output.
  • Preserve the exploratory convergence FAIL separately from program completion.
  • Preserve the bounded E(V) fit as mathematical evidence only.

Parse and plot boundary: These commands parse and plot the stored public routes; they do not prove that a new DFT calculation ran.

Boundary: Stored reconstruction supports route-scoped parser and artifact lineage only, not case-wide historical continuity, target-observable convergence, or scientific acceptance.

Run a fresh calculation

This track launches software in a caller-controlled runtime directory. Its new outputs do not rewrite the historical records published in the case bundle.

Run the isolated Aluminium electronic route

set -euo pipefail
case_root=examples/cases/aluminium-metallic-electronic-structure
QE_PW=${QE_PW:?Set QE_PW to the absolute path to pw.x}
QE_DOS=${QE_DOS:?Set QE_DOS to the absolute path to dos.x}
QE_PSEUDO_DIR=/absolute/path/to/pseudos
RUN_OUTPUT_ROOT=/absolute/new/path/aluminium-electronic-route
test -x "$QE_PW"; test -x "$QE_DOS"; test -d "$QE_PSEUDO_DIR"; test ! -e "$RUN_OUTPUT_ROOT"
QE_PW="$QE_PW" QE_DOS="$QE_DOS" QE_PSEUDO_DIR="$QE_PSEUDO_DIR" RUN_OUTPUT_ROOT="$RUN_OUTPUT_ROOT" bash "$case_root/run.sh"
for output in scf.out nscf-full.out dos.out bands.out; do test -s "$RUN_OUTPUT_ROOT/$output"; grep -nE 'JOB DONE|convergence has been achieved|c_bands|Error in routine' "$RUN_OUTPUT_ROOT/$output" || true; done
test -s "$RUN_OUTPUT_ROOT/al.dos"; sed -n '1,12p' "$RUN_OUTPUT_ROOT/al.dos"

Output policy: Use a new caller-selected external runtime directory; never overwrite or relabel committed output/, derived/, or figures/ as the new run.

Audit

  • Record the shell exit and inspect every stdout and stderr independently.
  • Check program termination separately from SCF convergence and retained NSCF/bands warnings.
  • Confirm the SCF, full-zone NSCF, native dos.x table, and path outputs before parsing.
  • Do not infer ionic convergence because this bounded case contains no relaxation.
  • Test DOS, band, or Fermi-surface observables with their own numerical convergence protocol.
  • State the scientific claim only after the evidence supports it.

Parse and plot boundary: The fresh route leaves the native al.dos table and solver outputs for direct inspection; no fresh plotting route is supplied. It does not add a relaxation, a bands.x data product, observable convergence, or a material-level claim.

Boundary: This present-day route creates new external SCF, full-zone NSCF, dos.x, and path-bands process records under the exact QE_PW, QE_DOS, QE_PSEUDO_DIR, and RUN_OUTPUT_ROOT contract. No relaxation is part of the case, and no private replay or case-wide historical continuity is claimed.

Retained record conflicts

Record note

Preferred record: source/pseudopotential-metadata.json

Boundary: The pseudopotential metadata and README record the exact public URL receipt verified on 2026-08-09, while qe_plan.json retains an older unverified-URL statement. Do not claim that all case records agree until that plan is reconciled.

Reproduce the exact published evidence

Use this section after understanding the scientific route. It retrieves the exact reviewed checkout and keeps new runtime evidence separate from the published case.

Repository: https://github.com/Maxwell3919/DFT-Research-Workflow.git

Source revision: 2ba561ac0937a938764bc283b25edb3a4700ece0

Working directory: repository-root

Case root: examples/cases/aluminium-metallic-electronic-structure

git clone https://github.com/Maxwell3919/DFT-Research-Workflow.git
cd DFT-Research-Workflow
git checkout --detach 2ba561ac0937a938764bc283b25edb3a4700ece0
test "$(git rev-parse HEAD)" = "2ba561ac0937a938764bc283b25edb3a4700ece0"
case_root=examples/cases/aluminium-metallic-electronic-structure

Inspect the explicit Aluminium model and exact pseudopotential receipt

Run from: repository-root

case_root=examples/cases/aluminium-metallic-electronic-structure
cat "$case_root/source/fixture-metadata.json"
cat "$case_root/source/pseudopotential-metadata.json"

Boundary: The structure is an explicit teaching input rather than a database identity, and the receipt establishes pseudopotential bytes rather than transferability or numerical suitability.

Exact case bundle files

This page is a reader-facing entry point for the case directory. Inspect the files themselves for the exact source record, input, command scope, raw or sanitized output, checks, extraction, parser, derived data, and figure lineage.

Recorded commands and outputs

These records distinguish captured output, replay instructions, and unavailable historical shell details. Records may belong to different declared routes; adjacency on this page does not establish one continuous run. An exit is shown only when the manifest scopes where that value came from, and a QE terminal marker is not silently converted into a shell exit code.

Case-wide timing boundary: No case-wide start, completion, or exit is claimed. The 04:21:54-04:22:02 timestamps belong only to the isolated DOS route.

Initial captured SCF/NSCF/bands outputs

Route ID: initial-captured-electronic-output-set

Route timing: 2026-08-05T00:45:36+08:00 to 2026-08-05T00:47:36+08:00. Earliest and latest QE-printed timestamps across three captured stdout files; gaps and continuous save-tree ancestry are not established.

Exact historical command: Not captured. output/recorded-commands.txt labels the listed shell lines as replay-equivalent only.

Exact historical exit: Not captured; each stdout retains a JOB DONE marker, which is not a shell exit transcript.

  • input/scf.in -> output/scf.out
  • input/nscf.in -> output/nscf-full.out
  • input/bands.in -> output/bands.out
  • No public launcher, scheduler record, or save-tree payload binds these three outputs into one continuous historical invocation.

Boundary: This initial set is real output evidence but its exact historical command/exit/continuity is unknown.

Isolated SCF/NSCF/dos.x/bands rerun

Route ID: isolated-dos-route

Route timing: 2026-08-05T04:21:54+08:00 to 2026-08-05T04:22:02+08:00. QE-printed timestamps for output/dos-route only; not a case-wide completion window.

Exact historical command: The four stage invocations and ordering are preserved in run.sh and output/recorded-commands.txt.

Exact historical exit: The existing route record reports exit 0 for each stage; JOB DONE and empty stderr are independently inspectable, but no separate shell transcript is published.

  • input/scf.in -> output/dos-route/scf.out
  • input/nscf.in with the same runtime prefix/outdir -> output/dos-route/nscf-full.out
  • input/dos.in consumes the NSCF save state -> output/dos-route/dos.out -> output/dos-route/al.dos
  • input/bands.in runs after dos.x -> output/dos-route/bands.out; retained c_bands warnings remain visible.

Boundary: The runtime save tree is not published; route ordering is declared by run.sh and does not establish observable convergence.

Later k/smearing and E(V) screens

Route ID: later-exploratory-screens

Route timing: start not recorded to completion not recorded. Not folded into the isolated DOS-route window; per-output QE timestamps remain in the committed raw stdout.

Exact historical command: Current committed runners are input/run-convergence-matrix.sh and input/run-eos.sh.

Exact historical exit: The existing route record reports runner exit 0; the convergence assessment itself is FAIL and remains distinct from program exit.

  • Generated committed screen inputs -> one SCF stdout/stderr pair per input
  • input/parse_convergence.py -> derived/aluminium-convergence-matrix.csv and aluminium-convergence-assessment.json
  • input/parse_eos.py -> derived/aluminium-eos-samples.csv and aluminium-eos-fit.json

Boundary: A reported earlier blocked parser attempt has no standalone shell transcript; the corrected fit is mathematical evidence only.

isolated DOS-route SCF · exit 0

Route: isolated-dos-route

Record kind: existing-route-execution-record

pw.x -in scf.in > scf.out 2> scf.err

Evidence scope: output/dos-route/scf.out prints JOB DONE; stderr is hash-bound and empty.

Exit scope: Route-scoped existing record; no standalone shell transcript is published.

Inputs: input/scf.in

Outputs: output/dos-route/scf.out -> output/dos-route/scf.err

isolated DOS-route full-zone NSCF · exit 0

Route: isolated-dos-route

Record kind: existing-route-execution-record

pw.x -in nscf.in > nscf-full.out 2> nscf-full.err

Evidence scope: The NSCF stdout prints JOB DONE and retains five c_bands warnings.

Exit scope: Route-scoped existing record; no standalone shell transcript is published.

Inputs: input/nscf.in

Declared parents: output/dos-route/scf.out -> same runtime prefix/outdir; save tree not published

Outputs: output/dos-route/nscf-full.out -> output/dos-route/nscf-full.err

isolated DOS-route dos.x · exit 0

Route: isolated-dos-route

Record kind: existing-route-execution-record

dos.x -in dos.in > dos.out 2> dos.err

Evidence scope: dos.out prints JOB DONE and al.dos is a hash-bound 2,801-row native QE table.

Exit scope: Route-scoped existing record; no standalone shell transcript is published.

Inputs: input/dos.in

Declared parents: output/dos-route/nscf-full.out -> same runtime NSCF save state; payload not published

Outputs: output/dos-route/dos.out -> output/dos-route/dos.err -> output/dos-route/al.dos

isolated DOS-route bands · exit 0

Route: isolated-dos-route

Record kind: existing-route-execution-record

pw.x -in bands.in > bands.out 2> bands.err

Evidence scope: bands.out prints JOB DONE and retains five c_bands warnings.

Exit scope: Route-scoped existing record; no standalone shell transcript is published.

Inputs: input/bands.in

Declared parents: output/dos-route/scf.out -> same runtime prefix/outdir; save tree not published

Outputs: output/dos-route/bands.out -> output/dos-route/bands.err

later exploratory k/smearing screen · exit 0

Route: later-exploratory-screens

Record kind: existing-route-execution-record

bash input/run-convergence-matrix.sh

Evidence scope: Five SCF programs completed; the predeclared numerical screen result is FAIL.

Exit scope: Runner exit and numerical screen status are distinct.

Outputs: output/convergence-screen/ -> derived/aluminium-convergence-matrix.csv -> derived/aluminium-convergence-assessment.json

later bounded E(V) screen · exit 0

Route: later-exploratory-screens

Record kind: existing-route-execution-record

bash input/run-eos.sh

Evidence scope: Five SCF programs completed; the quadratic fit status is mathematical PASS only.

Exit scope: Runner exit does not establish EOS or elastic acceptance.

Outputs: output/eos-screen/ -> derived/aluminium-eos-samples.csv -> derived/aluminium-eos-fit.json

Source-bound excerpt: output/compact-source-excerpt.txt

SCF/NSCF/bands: Quantum ESPRESSO 7.5; 8x8x8 full-zone NSCF; Fermi energy 7.8018 eV; each recorded program reached JOB DONE.
Native DOS: dos.x 7.5 reached JOB DONE; al.dos has 2801 rows and prints E_F=7.802 eV.
Native DOS lineage: dos.out SHA-256 c930a07ec54184fe102d308668a5895b629087fdd977ce36b9a602e705616f6e; al.dos SHA-256 95c3342cb9229b0ff2fbc8cb17d48aa13ebd6f3ddea7a5a1c4797de9e64be5f8.
Exploratory k/smearing screen: FAIL; k10-to-k12 delta E=0.00164525 Ry/cell and delta E_F=0.2449 eV.
Five-point E(V) entry: mathematical fit PASS; c2=2.33027885266e-05 Ry/bohr^6; residual RMS=5.34111987806e-05 Ry.
The k/smearing screen FAIL is retained; the E(V) fit is not a converged EOS, bulk modulus, or elastic tensor. Observable convergence is not established; no scientific conclusion is claimed.

Complete artifact appendix

Stage-scoped lists above identify only the artifacts critical for reading each stage. This appendix retains every artifact declared by the case manifest, including supporting inputs, outputs, checks, tables, and figures.

Show every manifest-bound artifact and checksum

Evidence boundary

Evidence class: real-execution.

  • Quantum ESPRESSO PWSCF and DOS 7.5 (single-process CLI)
  • case-local raw-output parser 2.3 (Python CLI)

Execution host: Talos.

Inputs and identity
Recorded as passed within the stated scope. Declared inputs, route-scoped outputs, screen artifacts, and derived files retain strict hashes and parsing relationships; this edit records no fresh replay.
Program completion
Recorded as passed within the stated scope. The isolated Talos rerun outputs have terminal markers and empty captured stderr; initial historical shell exits remain unrecorded.
Electronic solver and ionic or structural checks
Recorded as passed within the stated scope. The isolated rerun SCF reports electronic convergence in five iterations. NSCF/bands c_bands warnings remain retained.
Artifacts and stage ancestry
Recorded as passed within the stated scope. Real al.dos, real screen outputs, hash-bound stdout/stderr, parser tables, figures, and route ancestry records are present.
Observable convergence
Not established by this case. One 8x8x8 dos.x sample plus exploratory k/smearing and E(V) screens do not establish observable-specific convergence; the convergence screen is explicitly FAIL.
Claim boundary
No material-level claim is made. No physical or material-level scientific conclusion is claimed.

What this case supports

  • Route-scoped real QE 7.5 outputs: an initial captured SCF/NSCF/bands set, an isolated SCF/NSCF/dos.x/bands rerun, and later exploratory screens with raw-output-derived tables and figures.

What this case does not support

  • One continuous historical launcher, case-wide execution interval, or unpublished save-tree ancestry.
  • A converged DOS, Fermi surface, EOS, elastic property, carrier density, transport result, universal parameter prescription, or material-level conclusion.

Program completion, solver convergence, observable convergence, and physical interpretation are separate gates. The case check and parser report their tested scope; an absent gate is not implied by a passing earlier one.

Research-workflow context

Use this case together with Research Workflow for the A–E scientific map. The case illustrates one bounded case and its declared routes; it does not prescribe parameters or establish a general material conclusion.