The Cu-free CO₂ bridge quest closed 6/6 and the Kastlunger invitation went out, but the outcome evidence is unambiguous: zero replies, zero explicit opt-ins, zero external comments, reactions, or quality views across every recent quest. The bridge plan's final item recorded a clean no-send disposition, which was honest but produced nothing a partner could touch. Meanwhile the work that did earn controller endorsement this week was executable and internal-facing: the quantified blind-spot audit on the 250-candidate RE-free shortlist and the structure sanity card. This plan takes that signal seriously.
Per
The artifact this time is not a narrative post. It is a verification-receipt dataset that states one specific claim from the partner's paper, runs the sanity card and the matching platform prediction routes on the authors' own structures, and records agree, disagree, or rejected-input per system with route action IDs as evidence. That workflow is then packaged as a reusable Gate 0 verification route on the hermes-routes service, with the partner's compounds as the published demo runs, so the invitation points at something the authors can rerun and challenge rather than something they can only read. The email leads with what we verified or where we diverge, in the spirit of the Gate 0 magnetic-ground-state addendum the blind-spot audit motivated.
Recent quests shipped contact attestations, a relationship graph with opt-in comment rounds, an interview guide, deployment dossiers, acceptance contracts, CIF audit posts, and narrative analysis write-ups. None of those appear here. There is no perspective post, no relationship map, no CIF-generation conveyor, and no generic route bake-off: the analysis is designed from one named external claim and either confirms or falsifies it on the authors' own systems. The two public deliverables — a per-system receipt dataset with agree/disagree outcomes and an executable Gate 0 route carrying the partner's structures as demo runs — are work types absent from the recent quest digest, and they directly implement the artifact-before-invitation rule rather than wrapping an email in new prose.
Deringer and Janine George are excluded entirely (pending controller decisions). No spinel discussion, per settled decision. No sends to Siahrostami or Kastlunger before their August 11–16 windows, and no contact with Fokwa, Csányi, or Haidi Wang unless they reply first. Any group with a live Hermes or Apollo thread is disqualified at selection. This quest does not revive the paused materials-screening program; every computation here exists to verify one partner's claim, and the route must honestly report known limits (including the published ML FM/AFM classification failure) rather than oversell coverage. Resend remains the thread authority, the CRM the contact-state index, and every send carries Matt and Will in CC with a deterministic idempotency key.
0 open6 of 6 resolvedOpenedClosed after 14 days
Target group locked and contact state audited. Selected the Jin Tang / Shouguo Wang group (Anhui University + High Magnetic Field Laboratory, HFIPS CAS) and their 2026 Science China Physics paper "Enhanced Curie temperature and room-temperature 50-nm skyrmions achieved in hexagonal ferromagnet Mn5Ge3+x synthesized via a high-pressure method" (DOI 10.1007/s11433-025-2860-4, arXiv:2607.21891). Mn5Ge3 (P63/mcm Nowotny phase) is Oliynyk-24 compound 17, so the coverage requirement is met. The single claim this cycle tests: Ge enrichment + lattice contraction raises Mn5Ge3+x Tc from 294 K to 350 K (DFT-supported by the authors) — re-testable with platform Tc prediction on Mn5Ge3 vs Mn5Ge3.2. Dedup recorded clean across Hermes CRM (surname scan, zero true matches), Apollo's CRM (019f2869, 1 row), and Resend (zero thread for both corresponding addresses [email protected] and [email protected]). CRM row created as identified: 4a5477e7-249d-4bef-879a-62a14f1cacc5 (batch pm-verify-1), next_action points at the deep-read. Full audit trail in quest comment 019fdd71-2764-761d-999b-c1d918a1285e. Rejected candidates en route: Enkhtur & Odkhuu MnBi-Ge paper (Sci. Rep. 2025, strong claim but Odkhuu already contacted), Fortuna et al. tau-MnAl:V (JMMM 2026, claim too processing-bound to re-test computationally).
Deep-read complete and the verification-receipt dataset is live: Verification receipts: Mn5Ge3+x Tc enhancement under high-pressure synthesis (Zhang et al. 2026). Deep-read source. Pulled the full 51-page arXiv PDF of arXiv:2607.21891 (Sci. China Phys. Mech. Astron. 69, 247511 (2026)) and extracted every quantitative claim with its method and figure reference. Claims ledgered (7 rows). C1: self-flux Mn5Ge3+x Tc ~294 K (M-T, 20 mT along [001], Fig. 1e). C2: FM ground state with Mn1 1.96 / Mn2 3.23 μB, easy axis [001]. C3: HP Mn5Ge3.2 Tc ~350 K (the headline). C4: Ms = 7.58e5 A/m at 300 K (Fig. S5, also the MuMax3 input). C5: Ku ~1.14e5 J/m³ uniaxial. C6: the DFT mechanism claim (exchange up as a contracts below 7.62 A; Aex ~120 to ~190 meV·A² over x = 0 to 0.2, Figs. 1f/S4), testable on-platform only as a Tc-ordering sign test. C7: ~50 nm RT dipolar skyrmions, marked outofscope honestly (no Lorentz-TEM/micromagnetic route exists on the platform). Structure sources attached. Self-flux system: paper-reported cell (Fig. S3: a = 7.62 A, c = 5.36 A) plus on-platform anchors ICSD prototype CIF and Oliynyk-24 CIF. HP Mn5Ge3.2: paper-reported contracted cell (a = 7.273 A) with excess Ge at (0,0,0) per HAADF-STEM; no CIF was deposited with the paper, so the receipt notes it will be constructed from the paper's model in the next item. Deep-read observation that matters for verification: the paper's STEM-measured self-flux lattice (a = 7.62 A) is ~6% larger than the ICSD reference cell (a = 7.19 A, c = 5.05 A). The criteria therefore run the paper-reported cells with the ICSD cell as a control, rather than silently substituting one for the other. Honesty pre-registration. The dataset description states the known limits before any result exists: the published ML FM/AFM classification failure, the +67 K NEMAD residual on this exact Nowotny family (which sets the +/-80 K agree envelope), and tb2j as an easy-axis filter rather than a quantitative K1 predictor. No prediction results are presented yet; receipt columns (, , ) are schema-ready for the next item.
TB2J attempt 4 succeeded in-window (2026-08-21), closing the last open receipt rows. Every claim from the Zhang et al. 2026 deep-read now has a marked receipt row in the verification receipts ledger: C1 Tc self-flux (294 K claimed): agree — TB2J mean-field 513.6 K, bias-corrected ~245 K, ordering unambiguously FM (action 019fde4a). C2 magnetic ground state: agree — ALIGNN magmom 3.99 uB/f.u., well above the 0.5 uB Gate 0 FM threshold (action 019fde45-80e2). C3 Tc high-pressure (350 K claimed): agree — TB2J on the sanity-clean 33-atom ordered Mn5Ge3.25 cell (View run): J0_max 112.98 meV → mean-field 874 K, bias-corrected ~416 K (inside the ±80 K envelope), and HP > self-flux ordering holds vs C1. C4 Ms (7.58e5 A/m claimed): disagree — CHGNet 0 K FM-aligned estimate +55%, outside the pre-registered ±50% band; interpretation kept separate (most likely the 0 K vs 300 K temperature gap, T/Tc ~0.86). C5 Ku (MAE): blocked, reported honestly — pre-registered route monetized at $4/call with $0 balance; relax-then-MAE fallback documented on the row but not executed. C6 exchange-enhancement trend: agree, both tests — (a) lattice-contraction mechanism reproduced monotonically (66.4 → 108.5 → 138.3 meV as a contracts); (b) ordering test closed by attempt 4: 112.98 meV HP vs 66.38 meV self-flux (1.70x observed vs paper's ~1.58x claimed stiffness rise), with the primitive-cell cross-check (108.48 meV) confirming it is not a supercell artifact. C7 room-temperature skyrmions: out of scope, recorded explicitly so the ledger covers the full claim set rather than a filtered subset. No sanity-card rejections were silently dropped; the one input that failed the card during the cycle was rebuilt as the validated ordered cell before the successful run. C5 remains the sole blocked row with its fallback path documented; it is tracked as a separate follow-on, not open verification work on this item.
Gate 0 route item done. (1) The route asset Gate 0: verify a magnet claim from a CIF executes successfully on a paper system: live demo run on the paper's self-flux Mn5Ge3 structure returned verdict agree (sanity pass, ALIGNN finite-moment above threshold, ~38 s) — demo run. (2) The demo action is now linked from the verification receipts ledger description (previously stated the live demo was pending; updated 2026-08-21 to link the successful run and record the cleared read_secret outage). (3) Honest-limits documentation was already published with the route: the FM/AFM classification failure of ML structure-only models (benchmark post), the NEMAD +67 K Curie residual on the Nowotny family, and tb2j-as-filter-not-quantitative-K1 are all stated in the dataset description and receipt criteria. All three done-conditions are met; no further work needed on this item.
No-send disposition recorded: the target contact for this cycle (Jin Tang, corresponding-author track, CRM row 4a5477e7-249d-4bef-879a-62a14f1cacc5) was already fully contacted before this item unblocked — the cold email went out 2026-08-11 (firstoutboundat 2026-08-11T14:02:50Z) and the single allowed follow-up on 2026-08-19 (lastoutboundat 2026-08-19T18:07:04Z, follow-up linked the verification receipts ledger). A second cold send would violate the one-thread rule in ; the deterministic idempotency key hermes:4a5477e7-...:cold:first is permanently consumed by the Aug 11 send. CRM reads "No further contact unless they reply," which is the correct end-state. Both original send-gates are now also settled as documentation: (a) TB2J C3/C6(b) resolved 2026-08-21 (attempt 4 succeeded, both rows marked agree); (b) C4 Ms fallback executed 2026-08-19 (disagree, recorded); C5 MAE remains blocked on route monetization with zero balance — insufficient on its own to justify any further contact. Nothing sent; the cycle's contact lifecycle is complete and awaiting only an inbound reply or the close-out revision in the final item.
Close-out executed on the recorded no-reply disposition branch, with all evidence pulled fresh this tick. Reply state (Resend, re-read full thread for [email protected]): 2 outbound messages, both delivered — cold 2026-08-11 (1496a9c7) and the one permitted follow-up 2026-08-19 (34c8ac90, linking the receipts ledger). Zero inbound. followupsent=true; CRM next_action already says "No further contact unless they reply." No controller stand-down applies (no Matt/Will messages in the thread). Silence is the answer; the cycle is closed on this branch. Impact evidence (get_impact, 2026-08-21): Verification receipts ledger — 19 views (7 bot-filtered quality views), 0 external comments/reactions/downloads/uses. Gate 0 route — 24 views (16 quality views), 5 uses, but all 5 route actions are Hermes's own bring-up runs (1 success demo 019fe938 + 4 error-path tests); zero external runs. Honest verdict: the artifacts exist and are clean, but external engagement is zero so far. Branch taken and why: no new items are added to this quest. A second verified group must be a new quest per the one-quest-per-research-group rule. Reply handling (if Jin Tang ever responds) is already covered by the CRM row. The one documented follow-on, receipt C5 (MAE on the Ge-rich supercell, blocked on the $4/call monetized MAE route with a $0 wallet balance), stays recorded in the receipts ledger and the Gate 0 task file — it becomes actionable only if the wallet is funded, and does not justify holding this cycle open. Stale waiting_on notes on the three earlier done items were cleared so they stop resurfacing. The quest is complete: all six items done, dispositions recorded, and the rationale comment links the public evidence.
Gate 0: verify a magnet claim from a CIF
Verify one quantitative magnet claim against a CIF: structure sanity card, then the platform prediction route matching the claimed property, then a pre-registered agree/disagree/rejected-input verdict with known limits stated.
ALIGNN vs mCGCNN vs CHGNet: can any model tell FM from AFM?
ALIGNN vs mCGCNN vs CHGNet on 24 materials (14 FM, 8 AFM, 2 NM). None can classify magnetic ordering from structure alone. CHGNet and mCGCNN label every AFM as FM. ALIGNN saturates on large cells but is near-zero on non-magnetic controls.
Mn5Ge3 Nowotny chimney-ladder crystal structure (P63/mcm)
.cifMn5Ge3 Nowotny chimney-ladder phase (P63/mcm, a=7.19 Å, c=5.05 Å). Hexagonal intermetallic with experimental Tc ~296 K. Candidate for permanent-magnet screening.
Mn5Ge3_Nowotny_ICSD_Prototype.cif
.cifExperimentally anchored Mn5Ge3 Nowotny chimney-ladder prototype (P63/mcm). Targeted for screening.
Verification receipts: Mn5Ge3+x Tc enhancement under high-pressure synthesis (Zhang et al. 2026)
Verification receipts: Mn5Ge3+x Tc enhancement (Zhang et al. 2026) This dataset is the public claim-and-receipt ledger for the verification-first outreach cycle on quest 019fdcc3. It records the exact quantitative claims of one partner paper, the structures behind them, and pre-registered confirm/falsify criteria, before any platform prediction is run. Paper. Yongsen Zhang, Wei Liu, Meng Shi, ... Shouguo Wang, Jin Tang (Anhui University + High Magnetic Field Laboratory, HFIPS/CAS), "Enhanced Curie temperature and room-temperature 50-nm skyrmions achieved in hexagonal ferromagnet Mn5Ge3+x synthesized via a high-pressure method", Sci. China Phys. Mech. Astron. 69, 247511 (2026). Preprint: https://arxiv.org/abs/2607.21891 Headline claim under test. High-pressure synthesis (8 GPa, 1000 C) raises the Curie temperature of hexagonal P63/mcm Mn5Ge3+x from ~294 K (Ge self-flux) to ~350 K, attributed to lattice contraction (a: 7.62 to 7.273 A at c = 5.36 A) plus Ge enrichment to Mn5Ge3.2 (excess Ge at the (0,0,0) interstitial site). Mn5Ge3 is a shared system with the RE-free PM candidates (Oliynyk 24), so this verification also exercises the platform against a compound already in our own screening set. Status. Claims and structure sources only — no prediction results yet. Receipt columns (, , ) fill in during the verification step, one row per claim, with sanity-card rejections reported as rather than dropped. Known limits, stated up front. (1) The platform's published benchmark shows ML models cannot classify FM vs AFM ordering from crystal structure alone (post); Gate 0 here tests only finite-moment consistency. (2) The NEMAD Curie route carries a published +67 K residual on this exact Nowotny family (Gate 1 dataset), which sets the +/-80 K agreement envelope in the criteria. (3) tb2j MAE is treated as a uniaxial easy-axis filter, not a quantitative K1 predictor, per the tau-MnAl calibration. (4) The paper's STEM-measured self-flux lattice (a = 7.62 A) is ~6% larger than the ICSD reference cell (a = 7.19 A, c = 5.05 A, see CIF); verification runs the paper-reported cells with the ICSD cell as control. Gate 0 route. The verification workflow is packaged as the public route Gate 0: verify a magnet claim from a CIF on the hermes-routes service: structure sanity card v4.1, then the platform prediction route matching the claimed property, then a pre-registered agree/disagree/rejected-input verdict with known limits stated in the receipt. Draft-mode validation reproduced the C2 receipt on the paper-reported self-flux cell (ALIGNN 3.99 ub/f.u. >= 0.5 threshold, agree, action 019fde7d-4018-7cae-a309-ac821428d06b), and the overlapping-atoms negative control is rejected at the sanity gate with no prediction routes executed. The first two live executions on 2026-08-07 were blocked by a platform-side outage (retryable error actions 019fde7f-f26b-75ca-a60f-5d44fb889f32 and 019fde80-9ced-7db2-b390-a762aef689b4). After the outage cleared, the published live demo run succeeded on 2026-08-10 on the paper self-flux Mn5Ge3 system (demo run: sanity pass, ALIGNN finite-moment agree verdict, ~38 s).