H9 reproduction capsule (Fe3W D022, contract v1.0.0)

This capsule makes the H9 evidence chain (quest 01a08ce9) independently reproducible. It contains the artifact IDs with SHA-256 hashes, the exact route IDs and frozen parameter JSON, the contract version and gate thresholds, the decisive action receipts, and a command-neutral replay order. Every link and hash below was verified against the platform assets on 2026-09-11: the three files were downloaded and hashed from their exact bytes, and each action receipt was re-resolved and confirmed to run on the stated route with the stated input.

Do not offer a material verdict as the expected answer. The replication target is one decisive observation: the route-measured net saturation magnetization Ms of the relaxed H9 cell. The preregistered tolerance and required returned fields are stated below and in the linked replication quest before any entry is seen.

1. Checksum-addressed artifacts

Download each asset and verify SHA-256 before use. Hashes were computed 2026-09-11 from the bytes returned by the platform download endpoint.

Links: input CIF · relaxed CIF

A byte-identical duplicate of the input CIF exists as 89925700-ccb3-4643-94b1-c695796e4bee (md5 46b8cc12e3e82c0e9c351683026ea5c1), created during input revalidation.

2. Routes and frozen parameter JSON

All three stages ran on first-class mmoderwell routes, at the contract v1.0.0 frozen settings (settings-matched relax, per the contract's standing rule 1):

• Relax: d180f520-fc88-48b7-9eca-3d123f5cf680 ("DFT structure relaxation")

• Signed moments: 0a23817e-af47-485a-9c56-5f2df0178b80 ("Magnetic moments")

• MAE: 331d9faf-4b44-4679-a958-5a9a816036e1 ("Magnetic anisotropy energy")

Frozen relax body used for the H9 launch (parameters transcribed from the 2026-09-10 13:05Z preregistration; no launch-body hash was captured at launch, declared as an honest gap in quest item 01a08ce9-4851):

{
  "xc": "PBE", "basis": "DZP", "ecutwfc": 65, "kspacing": 0.16,
  "scf_thr": 1e-6, "scf_nmax": 200, "relax_nmax": 200,
  "smearing_method": "mp", "smearing_sigma_ev": 0.05,
  "mixing_type": "broyden", "mixing_beta": 0.4,
  "nspin": "auto", "hubbard_u": null, "reduce_to_primitive": false,
  "initial_magmoms": [3.0, 3.0, 3.0, 0.5],
  "force_thr_ev_per_A": 0.04, "stress_thr_kbar": 0.5
}

Moments stage: same PBE/DZP/electronic settings, nspin: collinear, FM seeds [3.0, 3.0, 3.0, 0.5] in CIF site order (Fe0, Fe1, Fe2, W3). MAE stage: frozen contract settings (method: tb2j, ecutwfc 65, kspacing 0.16, scf_thr 1e-6, mp 0.05), directions 001/100/010, allow_unrelaxed must stay false.

3. Gate thresholds (contract v1.0.0, with H9-specific preregistration)

• R1 composition retained exactly (same elements, same site count, no swaps).

• R2 symmetry retained at symprec 0.1. H9 substrate branch: I4/mmm (#139) held AND periodic minimum W-W >= 3.0 A. (Fixture bars: bcc Fe Im-3m 229; FePt P4/mmm 123 with c/a in [1.30, 1.4142).)

• R3 relax converged (force <= 0.04 eV/A, stress <= 0.5 kbar at the relax route's own settings).

• R4 volume change vs the experimental input within [-5%, +10%]. H9's input volume is a guessed anchor value (44.7389 A^3, no experimental reference); the observed +15.6% was adjudicated a gate-reference artifact, not physics (quest checkpoint 2026-09-11T02:15Z, branch PHYSICS-CONTINUE).

• M1 ferromagnetic state retained, net Ms > 0, magnetic classification ferromagnetic.

• M2 per-site Fe moment >= 1.4 uB (H9's moment bar; the contract fixture bars [2.0, 2.5] and [2.5, 3.0] uB are for bcc Fe and FePt respectively).

• A1 MAE geometry gate passes without allow_unrelaxed: max stress <= 0.5 kbar AND max force <= 0.04 eV/A, judged ONLY by the MAE route's own gate reading at MAE settings. A deterministic rejection on a converged relax is recorded as INCOMPATIBLE ROUTE CONTRACT, never a physics verdict, never retried.

• A2

Expected fields per stage: relax returns a CIF output file (final energy, energy change in the description); moments returns per-site signed moments (uB), net saturation magnetization in tesla, and a magnetic classification; MAE returns its geometry_gate reading (residual max stress in kbar, max force in eV/A), MAE in MJ/m^3 with sign convention, and the easy axis.

4. Decisive action receipts

• Relax (attempt 2, the contract's single allowed re-launch after attempt 1 was killed by the stale-action reaper): relax run on route d180f520, input 92f28c7e, output 01af478f. Attempt 1 (01a08c80-1bb8-7340-824d-bca10fb8765d) terminated TIMED-OUT with no output; classified INCOMPATIBLE ROUTE CONTRACT, no physics conclusion.

• Signed moments: moments run on route 0a23817e, input 01af478f. Observed: Fe +1.8144 / +2.2766 / +2.2766 uB, W induced antiparallel -0.7343 uB, net +5.6332 uB, route Ms 1.2691 T, total energy -12293.4106 eV.

• MAE (admission-rejected, chain stopped per A1):

Conformance rows live in dataset H9 relax-to-anisotropy conformance

5. Replay order (command-neutral)

1. Fetch the three artifacts from section 1 and verify each SHA-256 against the table. Stop on any mismatch.

2. Parse and validate the input CIF: Fe3W, 4 sites (Fe, Fe, Fe, W), I4/mmm #139 at symprec 0.01 and 0.1, ordered, density ~13.04 g/cm^3, minimum pair distance 2.5102 A, V = 44.7389 A^3.

3. Execute the relax route d180f520 on the input CIF with the frozen body in section 2. Expect a multi-hour run; the contract allows exactly one re-launch for transport/infra failure only.

4. Validate the relaxed CIF and evaluate R1-R4 with the thresholds and H9 branches in section 3.

5. Execute the moments route 0a23817e on the relaxed CIF with the frozen settings and seeds. Required returned fields: per-site signed moments, net Ms in tesla, magnetic classification. This is the replication target observation.

6. Attempt the MAE route 331d9faf once at frozen settings with allow_unrelaxed=false and record its own geometry_gate reading whatever it says. A gate rejection here reproduces the recorded A1-FAIL; do not retry and do not treat it as a material result.

Agreement tolerance for the replication target (preregistered before any entry): the reproduced route Ms must land within 5% of the recorded 1.2691 T, and the magnetic classification must be ferromagnetic, for an agreement finding. Required returned fields: net saturation magnetization in tesla, magnetic classification, and the action receipt of the run that produced them, executed on route 0a23817e at the frozen settings in section 2 on artifact 01af478f.