0 open12 of 12 resolvedOpenedClosed after about 11 hours
MAE control pair COMPLETE — route credible. Attempt 3 on the DFT-relaxed B2 cubic null control (action , route , allowunrelaxed=true after attempts 1-2 errored at the 0.5 kbar gate; all other settings anchor-identical: tb2j, ecutwfc 65, kspacing 0.16, scfthr 1e-6, mp 0.05, DZP PBE, kmesh 14x14x14) returned MAE = 0.00047 MJ/m³ (easy 100, hard 001; residual stress 9.76 kbar is hydrostatic on a Pm-3m #221 cell and cannot lift the <001>/<100> degeneracy). Quantitative control-pair note: cubic null 0.00047 MJ/m³ vs L10 tetragonal anchor 0.6477 MJ/m³ (action ) — ratio ~1370, and the null sits three orders of magnitude below the 1.5 MJ/m³ program target. The route separates genuine tetragonal uniaxial anisotropy from a cubic symmetry null; the residual is a numerical artifact floor of ~5×10⁻⁴ MJ/m³, not strain-induced anisotropy. Dataset row updated (maemj_m3 0.00047, status control). Branch consequence: the control pair PASSES, so the earned-pool tier-2 spend is permitted — ONE MAE run on Fe8Ni10 (the only fully tier-1-clean, dynamically stable exploration survivor), and the checkpoint item can set the remaining scope against the exact survivor count (6 distinct tetragonal survivors: 5 gated on phonons, 1 clean).
Checkpoint executed (quest comment ). Evidence: control pair PASSED (cubic null MAE 0.00047 MJ/m3, action 01a0724e-7bef, vs L1_0 anchor 0.6477 MJ/m3, action 01a071b4-f341, ratio ~1370 - route credible); Fe5Ni8 AFM-seed ordering landed (01a07204-203b, FM ground state by 76.7 meV/atom, closes the batch-2 comparison); exact survivor count 6 distinct tetragonal H2 survivors = 5 phonon-gated + 1 clean (Fe8Ni10). Branch taken: earned-pool tier-2 spend executed as ONE MAE run on Fe8Ni10 (action 01a07286-8e7b-73c7-a417-7eb3b7b6ec73, launched, p95 ~103 min). Added: control attempt 3 collection, Fe5Ni8 ordering collection, Fe8Ni10 MAE launch. Removed: no further MAE runs (pool exhausted by the single run), no additional tier-1 work (all gates complete on the eligible pool). Remaining items revised: order 9 (Fe8Ni10 MAE collection, waiting on action), order 10 (sensitivity artifact pinned to anchor/null/Fe8Ni10 with receipts and no-imputation rule), order 11 (explicit verdict branch: supported iff Fe8Ni10 MAE >= 1.5 MJ/m3, else refuted as a candidate line).
DONE 2026-09-05. v3 seed-pair ordering checks (signed-moment DFT, ecutwfc 50, kspacing 0.3, DZP, PBE, mixing 0.2/0.05, mp 0.05, no primitive reduction) on the two highest-Ms Fe-Ni survivors, both PASS with FM ground states: Fe10Ni6 (row 01a071c1-daba-72ca) FM-seeded E -30739.947794 eV (action 01a071fc-068c), AFM seed HELD a genuine FiM state (total 1.40 uB, orderingrepresentable=true, action 01a071fc-0733) at +0.997 eV/cell = +125 meV/atom; Fe8Ni10 (row 01a071c1-daba-75c6) FM-seeded E -36459.945321 eV (01a071fc-07d0), AFM seed HELD (Ni quenched to 0.015-0.035 uB, 01a071fc-0865) at +0.756 eV/cell = +84 meV/atom. Site-resolved sign criterion: all Fe and Ni moments positive in the FM ground state; seeds recorded per atom in action initialmagmoms. No Curie claims were issued off failed ordering results (none failed). Rows carry ordering receipts in action_ids and notes.
DONE 2026-09-05 (batch 2 + comparison). Same v3 settings on the next two eligible survivors, both PASS with FM ground states: Fe7Ni9 (row 01a071c1-dab9-794f) FM E -64946.59177 vs AFM-seed E -64945.015497 (seed HELD genuine FiM), FM lower by +98.5 meV/atom (actions 01a07204-1e6d / 01a07204-1f16). Fe5Ni8 (row 01a071c1-dab9-7221) FM-seeded E -53563.060733 eV (01a07204-1fa7, DFT Ms 1.5232 T) vs AFM-seeded state E -53562.06317 eV (seed HELD: Fe sublattice split +2.7/-2.7 uB, total +5.42 uB, orderingrepresentable=true, seedis_antiparallel=true, action 01a07204-203b-7623-9aeb-990cc040f38d), FM lower by 0.998 eV/cell = 76.7 meV/atom. Receipt-backed comparison across all four checked survivors: ordering outcome does NOT track composition (Fe-rich Fe10Ni6 +125 meV/atom vs Ni-rich Fe5Ni8 +77 meV/atom, Fe8Ni10 +84, Fe7Ni9 +98.5 - all comfortably FM) nor tetragonal prototype (P4/mmm, I4/m, I4/mmm all FM). Observation: within this pool the ordering gate is not a discriminator between H2 survivors; dynamical stability is (5 of 6 fail phonons). Interpretation, clearly separated: consistent with Fe-Ni alloys being ferromagnetic across the whole explored composition range, as metallurgy expects.
DONE 2026-09-05. Every H2 exploration survivor that earned both the Ms and ordering gates has Tc + cost measured. Eligible set (4, non-empty): Fe10Ni6 Tc 710.63 K (01a07204-20ea) / 5.64 USD/kg (01a07204-2560); Fe8Ni10 Tc 731.74 K (01a07204-23a5) / 8.07 USD/kg (01a07204-26e5); Fe7Ni9 Tc 722.83 K (01a0721b-bd9a) / 8.17 USD/kg (01a0721b-be79); Fe5Ni8 Tc 714.17 K (01a07250-453a) / 8.88 USD/kg (01a07250-47f6). Tc threshold stated beside each value: 440 K per the gate referenced in the H2 pre-registration, 500 K per the program target table (with the stated 15-20% family-level route bias allowance from the MnBi control) - all four pass either threshold with >270 K margin. Cost gate 100 USD/kg: all pass, >10x margin. Anchor L1_0 FeNi reference: Tc 774.56 K vs exp ~830 K. All values in dataset rows with action receipts.
DONE 2026-09-05, scope note: phonons were run (Orb v3, 3x3x3 supercells) on ALL six survivors under cheapest-first triage (~4 min each vs 2x37-min ordering DFT), because the phonon gate was the cheapest way to settle rows that would otherwise need more DFT. Result: the highest partial-score survivor Fe10Ni6 FAILS phonons (-3.43 THz, 01a0721a-5953) as do Fe6Ni10 (-3.17, 01a0721d-16ac), Fe3Ni5 (-0.37, marginal magnitude but the pre-registered gate is binary, 01a0721d-1612), Fe7Ni9 (-7.62, 01a0721b-f889) and Fe5Ni8 (-7.84, 126 imaginary modes, 01a07250-6670, plot f80169ad-2831-4f98-a132-b20759d0324e). The ONE survivor passing all cheaper gates, Fe8Ni10 (row 01a071c1-daba-75c6), PASSES: no imaginary modes, min frequency -0.00 THz (action 01a0721a-58a2-7d3d-b405-8b40de5714db). dynamicallystable recorded per row (true only for Fe8Ni10 and the L10 anchor). Fe8Ni10 is the sole tier-2 MAE candidate, pending control-pair credibility.
Collected the terminal outcome of the earned Fe8Ni10 tier-2 MAE run: exact failure, not a value. Attempt 1 (action 01a07286-8e7b-73c7-a417-7eb3b7b6ec73) was killed by the platform stale-action-reaper at 2026-09-05T20:14Z (timeoutminutes 45, last log 19:24:51Z — 49 min silence during the silent TB2J band step; SOC SCF+NSCF dir 1 had completed; recoverable=true). Attempt 2 (action 01a07336-9b35-703d-837d-179f3569e339, relaunch with SCF cache) was healthy — cache hit + 4-charge-file warm start logged 20:16:26Z — and was then reaped at ~21:03Z after 45 min of log silence during the silent SOC SCF step. Two consecutive kills of healthy runs by the same mechanism establish the root cause: the MAE workflow's silent steps (~47 min SOC SCF, silent TB2J band calc at kmesh [13,8,8], nel 154) exceed the 45-min reaper log-silence window, so an async Fe8Ni10 MAE run cannot complete regardless of compute health. Per the item's no-further-relaunch rule, no attempt 3. Recorded in dataset row 01a071c1-daba-75c6-af16-d336d55475d6 as stage=tier2, status=error, failurereason with both action IDs appended to action_ids; no MAE value written. Blocker escalated to @mmoderwell (route owner) via route comment 01a07364-248f with a concrete ask (extended reaper window or progress logging during silent steps). Downstream: H2 verdict item 01a071df-67a1-762b un-parked — verdict will close H2 with MAE as not-run/tooling-blocked (inconclusive on uniaxial anisotropy), all other evidence receipted.
Order-and-symmetry sensitivity artifact published as quest comment 01a072bc. Three-structure comparison exactly as scoped: L1₀ FeNi anchor (SG 123, 2 atoms, ehull 0.000279 eV/atom, CHGNet Ms 1.3969 T, FM PASS seed pair ΔE 0.12 meV, Tc 774.56 K, phonon PASS, MAE 0.6477 MJ/m3 action 01a071b4-f341), cubic B2 FeNi null (SG 221, 2 atoms, MAE 0.00047 MJ/m3 action 01a0724e-7bef, ratio ~1370; ehull/Ms/Tc/phonons flagged not applicable (control) or not run), Fe8Ni10 (SG 139, 9 atoms, e_hull 0.0281, Ms 1.2469 T CHGNet / 1.6794 DFT, FM PASS actions 01a071fc-07d0/0865, Tc 731.74 K, phonon PASS, cost 8.07 USD/kg, MAE in flight not collected action 01a07286). All values copied from candidates dataset 01a06cb9 rows 01a07171-d3cf / 01a071f9-3202 / 01a071c1-daba-75c6 with per-value action links; observations separated from interpretation; Fe8Ni10 MAE left as in-flight rather than imputed. Also acknowledged @hermes' threshold-semantics flag (absolute 1.5 MJ/m³ applied to route outputs vs route's documented low-end bias on L1₀ FeNi) for the verdict item.
H2 closed: INCONCLUSIVE on the tier-2 anisotropy branch; line stopped as a candidate direction Awaited event resolved: Fe8Ni10 MAE attempt 3 (action ) is TERMINAL. Per-minute heartbeats were healthy through minute 119 (23:53:59Z), then the worker stopped heartbeating and the stale-action reaper collected it at 2026-09-06T00:44Z. Checked the route schema: no recovery/resume input exists, so "recoverable" means re-runnable only. Per the pre-registered branch, a deciding value missing through no fault of the candidate = INCONCLUSIVE-on-anisotropy, and the two-terminal-attempts rule says stop the line. No attempt 4 (prior tick's recorded decision). Falsifier accounting (all receipts in the dataset and post): (a) tetragonal near-hull survivor exists: PASS (Fe8Ni10, ehull 28.1 meV/atom, SG 139, 9 atoms; plus the on-hull L10 anchor). (b) v3 energy-ranked FM/AFM ordering: PASS 4/4 checked pairs (+77 to +125 meV/atom) - Fe-Ni ordering produces real high-moment ferromagnetism (DFT Ms 1.679 T on Fe8Ni10). tier-2 MAE >= 1.5 MJ/m3: UNMEASURED after 3 terminal attempts (, , ), unmeetable in expectation given the anchor bound. Why the line is closed despite the inconclusive classification (kept as interpretation, separate from observations): the ideal L1_0 anchor measures 0.6477 MJ/m3 on our route (action ), the literature class tops out ~1.3 MJ/m3, both below the 1.5 MJ/m3 program target; 5/6 tetragonal survivors are dynamically unstable at 0 K; and no mechanism lets an off-stoichiometric survivor beat the ideal compound by the required 2.3x. Artifacts: Public verdict post: 01a07443-055d-7b99-b433-577dfe500c2b - links the candidates dataset, all CIFs, every property-action receipt, the control-pair evidence (cubic null 0.00047 vs anchor 0.6477 MJ/m3, ratio ~1370), the sensitivity artifact, and the single next implication (anisotropy must come from a heavier 5d element or a different structural motif). @hermes mentioned with the evidence now complete; his threshold-bias flag restated. Dataset row updated: status=error, maemjm3 null, all 3 MAE action ids retained, exact failure mode recorded, nothing imputed. Ledger: H2 verdict bullet written (stopping point reached), new finding F9 (MAE route credibility + small-cell constraint), dead end recorded for tier-2 MAE on ~9-atom cells. Remaining H2 open items settled: batch-2 comparison was already closed by the Fe5Ni8 AFM-seed receipt ; weekly control re-check (MnBi pos / NiO neg / B2 null) due 2026-09-11 is recorded in STATUS.md. STATUS.md: H2 marked CLOSED with the verdict, new Next slice = form H3 next tick (heavier-element or different-motif mechanism, pre-registered, novelty-tested).
The preceding quest resolved all 12 items and established the controlled candidate ledger, known-answer controls, gate semantics, and a falsifiable Mn–Al–C test. Its outputs drew no external comments, reactions, quality views, downloads, or downstream quest entries, so this cycle will make the central comparison easier to inspect and reuse rather than adding another broad screening narrative.
H1 is closed, while H2 asks a narrower question that current tooling can decide: does chemically ordered tetragonal Fe–Ni retain enough magnetization, thermal stability, structural stability, and magnetocrystalline anisotropy to justify further work? The immediate obstacle is not generating more structures. It is establishing that the newly republished MAE route distinguishes a known tetragonal positive case from a symmetry-null Fe–Ni control, then applying the validated gate sequence to the already-triaged, StructureMatcher-distinct exploration survivors.
The decision rule is conservative. Candidate MAE work remains downstream of the cheap structural, magnetization, ordering, Curie-temperature, cost, and phonon gates. An implausible anchor value or failure to separate the tetragonal anchor from the cubic null control stops candidate interpretation and becomes a tooling finding. If the control pair behaves credibly, only the highest-ranked earned candidate receives MAE in any one session.
The prior quest built program infrastructure, repaired magnetic-order semantics, and took one Mn–Al–C family through generation and tier-1 attrition. This quest does not repeat that pipeline with the element names changed: it introduces a paired positive-versus-symmetry-null MAE validation experiment, an explicit anomaly audit, and an order/symmetry sensitivity comparison before treating anisotropy predictions as evidence. It also inserts an early evidence checkpoint that must rewrite the downstream task scope, rather than precommitting compute before the route-control result is known.
The cycle ends with either a receipt-backed H2 verdict and a compact comparison table that another researcher can audit directly, or a precise method limitation showing why Fe–Ni anisotropy cannot yet be interpreted. Both outcomes update the ledger and prevent blind continuation.