§3 结果（论文体）§3 Results (paper style)

评审更正（2026-10-15，Claude）：本页写于红队核查之前，以下结论已更新，以此为准：（1）L7：墙测到约 55 kpc（39–55 kpc 的超出 B = 2.48 ± 0.43 PW，计入估计量系统误差后 3.8–4.8σ）；55–66 kpc 的延伸改为 CANDIDATE，计入估计量与掩膜系统误差后为 2.0–2.9σ；66 kpc 以外为上限 B < 0.75 PW（95%，仅统计误差）。（2）L9：锥形下缘在约 40 kpc 转为宽度近乎不变的上部结构（MEASURED）；三维“空心烟囱”改为 INFERRED（非轴对称“两片墙”模型同样好），转折高度 1σ 范围 26–43 kpc（联合图），“空心”仅约 2σ。（3）L13：计入晕密度 28% 误差后 P_wall/P_halo = 2.95 ± 0.84（kT_halo = 0.2 keV，∝ 0.2 keV/kT_halo）；“不随高度变化”依赖 z > 14 kpc 截断；最不利的假设组合仍给 1.41 > 1。（4）图 fig_wall_profile 的晕压参考线已改正（原画在 10 倍处）。下一版手册将把这些改动写入正文。

Review correction (2026-10-15, Claude): this page was written before the red-team checks. These updates override the text below: (1) L7: walls are detected to about 55 kpc (39-55 kpc excess B = 2.48 +- 0.43 PW, 3.8-4.8 sigma with the estimator systematic); the 55-66 kpc extension is now CANDIDATE, 2.0-2.9 sigma with the estimator and mask systematics; beyond 66 kpc there is a limit, B < 0.75 PW (95%, statistical only). (2) L9: the cone-like lower edge turns into a near-constant-width upper structure at about 40 kpc (MEASURED); the 3-D "hollow chimney" is now INFERRED (a non-axisymmetric two-sheet model fits as well); the break height is 26-43 kpc at 1 sigma (joint map); "hollow" is only about 2 sigma. (3) L13: with the 28% halo-density error, P_wall/P_halo = 2.95 +- 0.84 (kT_halo = 0.2 keV; scales as 0.2 keV/kT_halo); "flat with height" depends on the z > 14 kpc cut; the most hostile set of assumptions still gives 1.41 > 1. (4) The halo reference line in fig_wall_profile is corrected (it was drawn 10x too high). The next version will put these changes into the text.

来源：cognition-os/REDTEAM_QUEUE.md（RT-01、RT-02、RT-03、RT-06）；I021 L7、L9、L13 的 2026-10-15 注记

Source: cognition-os/REDTEAM_QUEUE.md (RT-01, RT-02, RT-03, RT-06); I021 L7, L9, L13 notes of 2026-10-15

读前须知：单位与状态标签Before you read: units and status tags

距离 D = 19 Mpc，1 角分（′）= 5.53 kpc。面亮度单位 1 PW = 1e-7 个 XSPEC APEC 归一化每平方角分（0.3 keV 等离子体），对应 ∫ne nH dl = 1.49e15 cm⁻⁵。举例：一个 5 PW 的壁面亮度 ≈ 每平方角分 5e-7 个 apec 归一化。热压力 P = 2.3 n_H kT；举例：nH = 5.5e-5 cm⁻³、kT = 0.2 keV 时 P = 4.05e-14 dyn cm⁻²。状态标签（I021）：MEASURED = 直接测量（含场散误差）；INFERRED = 依赖模型；CONDITIONAL = 依赖明示假设；CANDIDATE = 非探测；WITHDRAWN = 不要引用。坐标架：+mnr = ENE（PA 75°），−mnr = WSW（PA 255°），盘方位角 PA 165°。

Distance D = 19 Mpc, 1 arcmin (′) = 5.53 kpc. Surface-brightness unit: 1 PW = 1e-7 XSPEC APEC norm per arcmin² of a 0.3 keV plasma, equivalent to ∫ne nH dl = 1.49e15 cm⁻⁵. Example: a wall brightness of 5 PW ≈ 5e-7 apec norm per arcmin². Thermal pressure P = 2.3 n_H kT; example: with nH = 5.5e-5 cm⁻³ and kT = 0.2 keV, P = 4.05e-14 dyn cm⁻². Status tags (I021): MEASURED = direct, with field-scatter errors; INFERRED = model-dependent; CONDITIONAL = rests on a stated assumption; CANDIDATE = not a detection; WITHDRAWN = do not cite. Frame: +mnr = ENE (PA 75°), −mnr = WSW (PA 255°), disk PA 165°.

来源Source: .agent/jobs/2026-10-15-S09-FACTS.md; I021_s09_paper_plan_v2.md

3.1 数据质量结果（科学的根基）Data-quality results the science rests on

五项数据质量结果决定了后续所有测量的口径（L1–L4）。MEASURED (1) 从 ODF 重新处理恢复了曝光，逐相机做了耀发 GTI 剔除，风区 PN/MOS 亮度比 1.07–1.12（L1）。(2) MOS1 CCD4 的软超出会制造一个假的 ~1 keV 成分，因此 MOS1 退出所有风区光谱（L2）。(3) MOS1 与 MOS2 都带一个随卷帘角变化的望远镜坐标架大尺度梯度（N−S 0.7–1.2 PW/′），PN 干净；因此 XMM 在 ≳5′ 的工作只用 PN，或给每台 MOS 加平面 nuisance 项（L2b）。(4) FXTDAS/CALDB v1.30 的 FXT 点源谱与 XMM 在 0.4–1 keV 一致到 0–10%（v1.20 最多低 37%）（L3）。(5) Chandra 目录中 19 个条目是弥漫气体而非点源，保留在图中（L4）。

Five data-quality results set the conventions of every later measurement (L1–L4). MEASURED (1) Reprocessing from ODFs recovers the exposures; camera-level flare GTIs are applied; the wind-region PN/MOS brightness ratio is 1.07–1.12 (L1). (2) A soft excess on MOS1 CCD4 makes a false ~1 keV component, so MOS1 is excluded from all wind spectra (L2). (3) Both MOS cameras carry a roll-dependent, telescope-frame large-scale gradient (N−S 0.7–1.2 PW/′); PN is clean; XMM work beyond ≳5′ therefore uses PN only, or a plane nuisance per MOS camera (L2b). (4) FXTDAS/CALDB v1.30 FXT point-source spectra match XMM to 0–10% at 0.4–1 keV (v1.20 was up to 37% low) (L3). (5) 19 Chandra-catalogue entries are diffuse gas, not point sources, and stay in the maps (L4).

S09 E085: north minus south pointing-group difference maps for MOS1, MOS2 and PN, with plane fits — S09 E085。XMM 北组（2017-11）与南组（2018-04）地图之差：差异只存在于 MOS1/MOS2（望远镜坐标架平面，~1.1–1.2 PW/′），PN 干净（0.07 ± 0.05 PW/′）。S09 E085. The XMM N-group (2017-11) minus S-group (2018-04) difference maps: the effect sits in MOS1/MOS2 only (a telescope-frame plane, ~1.1–1.2 PW/′); PN is clean (0.07 ± 0.05 PW/′).

来源Source: cognition-os/experiments/E085_s09_xmm_ns_pointing_systematic.md; E083 handover; artifacts/s09_e085_xmm_ns/n3079_s09_e085_xmm_ns.png

3.2 ENE 轴上的背景星系团（L5）The background cluster on the ENE axis (L5)

MEASURED ENE 小轴上 h ≈ 4.3–6.5′（24–36 kpc）的拉长 X 射线条带不是风气体，而是红移 z = 0.593 的背景星系团 WH24 J100242.9+554158（M500 = 1.4e14 M⊙，13 个成员；SDSS 光谱 z = 0.59323/0.59357）。XMM 与 FXT 都看到它：形状延伸（XMM σ 44″ × 17.5″，PSF 5.5″，2ΔlnL = 64），谱上需要 kT ≈ 2.3 keV（z 0.593）的热成分（占 0.4–1.25 keV 超出的 57–84%）。它与小轴方向一致是巧合（chance ~0.2）。所有风分析都用 1.5′ 半径掩罩它（E090）。掩罩后，ENE 轴带亮度从无掩罩时的 4.5–9.3 PW（4.75–6.75′）降到 8.0/8.8 PW（3.75/4.25′）、7′ 以外 1–3 PW；"ENE 锥到 39 kpc 是填满的"这一旧结论是星系团造成的（已撤回）。

MEASURED The elongated X-ray strip on the ENE minor axis at h ≈ 4.3–6.5′ (24–36 kpc) is not wind gas. It is the hot gas of a background galaxy cluster at z = 0.593, WH24 J100242.9+554158 (M500 = 1.4e14 M⊙, 13 members; SDSS spectra z = 0.59323/0.59357). XMM and FXT both see it: extended shape (XMM σ 44″ × 17.5″ against a 5.5″ PSF, 2ΔlnL = 64), and the spectrum needs a hot component at kT ≈ 2.3 keV (z 0.593; 57–84% of the 0.4–1.25 keV excess). Its alignment with the minor axis is a coincidence (chance ~0.2). Every wind analysis masks it with a 1.5′ circle (E090). With the mask, the ENE axis strip falls from 4.5–9.3 PW (unmasked, 4.75–6.75′) to 8.0/8.8 PW at 3.75/4.25′ and 1–3 PW beyond 7′; the old "filled ENE cone to 39 kpc" was the cluster (withdrawn).

ENE minor-axis strip brightness with and without the cluster mask, joint PN plus FXT — ENE 小轴带亮度（联合 PN+FXT），黄色带为 1.5′ 星系团掩罩；灰点为无掩罩的 v9 值。数据：v10_data.json。ENE minor-axis strip brightness (joint PN+FXT); the yellow band is the 1.5′ cluster mask, grey points the unmasked v9 values. Data: v10_data.json.

E090: XMM, FXT and Chandra images of the cluster, optical red-sequence galaxies, spectrum and the axis minus walls test — E090。上：XMM/FXT 光子图与 2σ 形状椭圆；下：Legacy Survey 图像 + 红序星系、点谱、有/无掩罩的轴−壁检验。E090. Top: XMM/FXT photon images with 2σ shape ellipses; bottom: the Legacy Survey image with red-sequence galaxies, the spot spectrum, and the axis−walls test with and without the mask.

来源Source: XMMLineMap cognition-os/experiments/E090_n3079_ene_axis_feature_background_cluster.md; artifacts/s09_web_v10/v10_data.json

3.3 形态：X 形双极风（L6）Morphology: an X-shaped bipolar wind (L6)

MEASURED 0.3 keV 面亮度图显示一对沿小轴的双极风，X 形，逐历元（2017-only、2018-only）与逐仪器（XMM、FXT、联合）都可复现；FXT 复现 XMM 形态到 ~6′（E076）。灰圈为已扣除的点源（40″ 掩罩），黄圈为星系团掩罩。最终图集（C1：全部用 PN 或 PN+FXT 联合图重建）尚未完成，是论文 I 图 1–4 的前置条件。

MEASURED The 0.3 keV surface-brightness maps show a bipolar pair of winds along the minor axis, X-shaped, reproduced per epoch (2017-only, 2018-only) and per instrument (XMM, FXT, joint); FXT reproduces the XMM morphology to ~6′ (E076). Grey circles mark the subtracted point sources (40″ masks), the yellow circle the cluster mask. The final map set (C1: everything rebuilt on PN or the PN+FXT joint map) is not yet done and gates Paper I figures 1–4.

0.3 keV surface-brightness map, joint XMM PN plus FXT, 16 arcmin half-field — 联合 PN+FXT 0.3 keV 图（16′ 半视场）。来源：s09_web_v10/map_jointpn.png。Joint PN+FXT 0.3 keV map (16′ half-field). Source: s09_web_v10/map_jointpn.png.

0.3 keV surface-brightness map, XMM PN only, 16 arcmin half-field — XMM PN-only 0.3 keV 图（16′ 半视场）。来源：s09_web_v10/map_xmmpn.png。XMM PN-only 0.3 keV map (16′ half-field). Source: s09_web_v10/map_xmmpn.png.

来源Source: artifacts/s09_web_v10/map_jointpn.png, map_xmmpn.png; E083 handover (L6, E076)

3.4 壁延伸到 ~66 kpc（L7）The walls extend to ~66 kpc (L7)

MEASURED 用分区剖面（壁楔 θ 15–45° 减去同半径外侧 θ > 52.5°）测壁超出 B。联合 PN+FXT（星系团掩罩后）：39–55 kpc B = 2.48 ± 0.43 PW（5.8σ）；55–66 kpc B = 2.09 ± 0.67 PW（3.1σ）；66 kpc 以外无探测（12–16′ 为 −0.26 ± 0.61）。FXT 单独 3.3σ、XMM PN 单独 1.6σ，符号一致；MOS 单独为负（−0.64 ± 1.13、−2.07 ± 1.92）——正是 MOS 梯度曾把 v9 XMM 壁压低的原因。FXT 60′ 壁发射率可追到 70–80 kpc（WSW，误差大）。

MEASURED The wall excess B is measured from zone profiles (wall wedge θ 15–45° minus the same-radius outside θ > 52.5°). Joint PN+FXT (cluster masked): B = 2.48 ± 0.43 PW at 39–55 kpc (5.8σ); 2.09 ± 0.67 PW at 55–66 kpc (3.1σ); no detection beyond 66 kpc (12–16′: −0.26 ± 0.61). FXT alone gives 3.3σ and XMM PN alone 1.6σ, same sign; MOS alone is negative (−0.64 ± 1.13, −2.07 ± 1.92) — the MOS gradient is what had pulled the v9 XMM walls down. The FXT 60′ wall emissivity is traceable to 70–80 kpc (WSW, large errors).

红队 R1 措辞（已采纳）：55–66 kpc 是一个合并径向 bin 的显著性，不是一个带误差的端点；外延的"终点"需要 C3 的端到端注入-回收与系统学标定。因此本手册写"55–66 kpc 合并 bin 内正对比"，而不写"壁终于 66 kpc ± …"。

Red-team R1 wording (adopted): the 55–66 kpc value is the significance of a pooled radial bin, not an endpoint with an error bar; the outer reach needs the C3 end-to-end injection-recovery and systematics calibration. This handbook therefore says "positive contrast in the pooled 55–66 kpc bin", not "the walls end at 66 kpc ± …".

Wall excess B per 1 arcmin ring for joint PN plus FXT, FXT, XMM PN and XMM MOS only — 壁超出 B 逐 1′ 环（S09 E086，星系团掩罩）。蓝：联合 PN+FXT；橙：FXT；绿：XMM PN；灰：MOS-only（负值即 MOS 梯度效应）。数据：redo_l7_l8.json。Wall excess B per 1′ ring (S09 E086, cluster masked). Blue: joint PN+FXT; orange: FXT; green: XMM PN; grey: MOS-only (negative values = the MOS gradient effect). Data: redo_l7_l8.json.

S09 E086。上：各相机组的壁超出；55–66 kpc 的方位角剖面（超出集中在 θ 10–30°）。下：帽缘上限。S09 E086. Top: wall excess per camera set; the azimuthal profile at 55–66 kpc (the excess sits at θ 10–30°). Bottom: cap-rim limits.

来源Source: cognition-os/experiments/E086_s09_l7_l8_pn_only.md; artifacts/s09_e086_l7_l8_pn/redo_l7_l8.json; R1 redteam REPORT.sol.md (rank 1)

3.5 45–90 kpc 无帽（L8）No cap at 45–90 kpc (L8)

MEASURED 用切圆模板与自由半径扫描（远点 ≥ 45 kpc，34 个旋转空假设，全局 p）搜"帽"。30° 帽扇区（避开壁）内无任何模板显著（p ≥ 0.24，两侧与折叠均如此）。60° 扇区内唯一显著的边（PN 联合，p < 0.03，远点 69 kpc）位于壁楔角内 —— 是壁的顶端，不是帽。95% 帽缘上限（联合 PN+FXT 折叠，60° 扇区）：50–70 kpc K95 ≈ 1.4 PW，70–90 kpc ≈ 2.1 PW（PW，0.3 keV 等离子体口径）；FXT 60′ 折叠到 110 kpc 为 0.9–2.0 PW。0.7–2 keV 硬带（FXT 60′）无热帽探测；在测得的晕密度下，软强击波帽会在软带上限边缘（边缘排除），热（1 keV）帽允许（E093）。红队 R1 措辞（已采纳）：这些是模板依赖的上限，"无帽"应写作"在所述模板与灵敏度内无需帽"，不排除闭式/热泡。

MEASURED A cap search uses tangent-circle templates and a free-radius scan (far point ≥ 45 kpc, 34 rotated nulls, global p). No template is significant in the 30° cap sector (clear of the walls; p ≥ 0.24 on both sides and in the fold). The only significant edge in the 60° sector (PN joint, p < 0.03, far point 69 kpc) lies inside the wall wedge — it is the top of the walls, not a cap. 95% cap-rim limits (joint PN+FXT fold, 60° sector): K95 ≈ 1.4 PW at 50–70 kpc and ≈ 2.1 PW at 70–90 kpc (PW, the 0.3 keV plasma-unit convention); the FXT 60′ fold gives 0.9–2.0 PW to 110 kpc. No hot cap is detected at 0.7–2 keV (FXT 60′); at the measured halo density a soft strong-shock cap would sit at the soft fold limit (marginally excluded) and a hot (1 keV) cap is allowed (E093). Red-team R1 wording (adopted): these are template-dependent limits; "no cap" should read "no cap required within the stated templates and sensitivities", which does not exclude a closed or hot bubble.

E087 cap search: tangent circles on the map with the null scans and rim limits — E087。切圆帽搜索：地图上的切圆、旋转空假设与缘上限。E087. The tangent-circle cap search: circles on the map, rotated nulls, and rim limits.

来源Source: cognition-os/experiments/E086_s09_l7_l8_pn_only.md (L8 tables); XMMLineMap E087, E093; n3079_e087_cap_circles.png

3.6 三维几何（3-D geometry）：WSW 锥→烟囱；ENE 张开（L9）3-D geometry: WSW cone → chimney; ENE flares (L9)

MEASURED（mock 标定；轴对称模型下）对每 0.5′ 横向 bin、1′ 高度 bin（14–80 kpc）用精确 Abel 投影拟合轴对称壳+内部模型。WSW：半张角 44–46° 的厚壁锥（壁厚分数 f 0.45–0.6）在 z_b 36–41 kpc 处变为近烟囱（等半宽 36–41 kpc），28 kpc 以上内部空心（η ≈ 0；mock P < 0.03）；填充模型被拒绝（Δχ² +14.8 FXT 60′ / +18.6 联合 PN）。ENE：~45° 锥，~40 kpc 以上缓慢张开；其填充不受约束（mock P 0.07–0.13）。Kompaneets 剧烈失宠（WSW Δχ² +28.6：底部太宽）。PN+FXT 联合图复核（S09 E087）：WSW 46° 锥 → 36 kpc 近烟囱，与 FXT 60′ 同形；ENE 不变。红队 R1 措辞（已采纳）：轴对称与轴在天空面内是 imposed 的；"烟囱"是模型族内的偏好，不是唯一的 3-D 解。

MEASURED (mock-calibrated; within axisymmetric families) Transverse profiles (0.5′ bins, 1′ height bins, 14–80 kpc) are fitted with axisymmetric shell + interior models projected by an exact Abel integral. WSW: a thick-walled cone of half-opening 44–46° (fractional thickness f 0.45–0.6) becomes a near-chimney (constant half-width 36–41 kpc) at z_b 36–41 kpc, with a hollow interior above 28 kpc (η ≈ 0; mock P < 0.03); a filled model is rejected (Δχ² +14.8 FXT 60′ / +18.6 joint PN). ENE: a ~45° cone flaring slowly above ~40 kpc; its fill is not constrained (mock P 0.07–0.13). Kompaneets blow-out is strongly disfavoured (WSW Δχ² +28.6: its base is too wide). The PN+FXT joint map recheck (S09 E087) gives the same WSW form (46° cone → near-chimney at 36 kpc), in the same shape as FXT 60′; ENE is unchanged. Red-team R1 wording (adopted): axisymmetry and a sky-plane axis are imposed; the "chimney" is a preference within the model families, not a unique 3-D solution.

E094: candidate shapes with delta chi-squared, best 3-d model on the FXT map, eta scan, transverse profiles — E094。候选形状的 Δχ²、最佳 3-D 模型叠图、η 扫描与逐高度横向剖面。E094. Δχ² of the candidate shapes, the best 3-D model on the map, the η scan, and transverse profiles per height.

S09 E087: geometry and over-pressure redone on the PN plus FXT joint map — S09 E087。PN+FXT 联合图上的几何、壁剖面与压力复核：WSW 烟囱与超压都成立。S09 E087. Geometry, wall profile and pressure redone on the PN+FXT joint map: the WSW chimney and the over-pressure both hold.

来源Source: XMMLineMap E094_n3079_wind_3d_geometry.md; E097; S09 E087_s09_joint_products_pn.md; R1 redteam (rank 7)

3.7 温度 kT(z)（L10）与 C2 复核Temperatures kT(z) (L10) and the C2 recheck

MEASURED 1T on-off APEC 谱拟合（Z 固定 0.3）：8–28 kpc kT 0.25–0.31 keV；41–61 kpc 壁 0.19–0.22 keV；每个高度上壁与内部共享一个 kT（E095，dC ≤ 2）。C2 复核（2026-10-15，去 MOS2）：归档门（WWALL:BG，MOS2+PN+FXT）精确复现 kT = 0.187298 keV；PN+FXT 重拟后 WSW 壁 41–61 kpc 0.187 → 0.182 keV、ENE 0.218 → 0.237 keV；所有相机对移动 ≤ 0.036 keV（≤ 1.3 倍归档半宽）；PN-only 更噪（最大移动 +2.53 半宽）。结论 L10 不变。

MEASURED 1T on-off APEC spectral fits (Z fixed at 0.3): kT 0.25–0.31 keV at 8–28 kpc; walls 0.19–0.22 keV at 41–61 kpc; wall and interior share one kT at each height (E095, dC ≤ 2). C2 recheck (2026-10-15, MOS2 removed): the archived gate (WWALL:BG, MOS2+PN+FXT) reproduces kT = 0.187298 keV exactly; with PN+FXT the WSW wall at 41–61 kpc moves 0.187 → 0.182 keV and the ENE wall 0.218 → 0.237 keV; every camera pair moves by ≤ 0.036 keV (≤ 1.3 archived half-widths); PN-only is noisier (largest move +2.53 half-widths). L10 stands.

kT and O/Fe: archived MOS2 plus PN plus FXT versus C2 PN plus FXT refits — 左：1T kT，归档（MOS2+PN+FXT，灰）对 C2 PN+FXT（蓝）。右：1T O/Fe，归档（灰）对 C2（绿）；虚线为太阳比。数值来自 C2 报告 Table 1/2 与其 npy 文件。Left: 1T kT, archived (MOS2+PN+FXT, grey) against C2 PN+FXT (blue). Right: 1T O/Fe, archived (grey) against C2 (green); the dashed line is the solar ratio. Values from C2 report Tables 1/2 and their npy files.

相机对（regspec）Pair (regspec)	归档 kT (keV)Archived kT (keV)	C2 PN+FXT	D (keV)
`WWALL:BG (e092)`	0.187 (0.150–0.226)	0.182 (0.146–0.215)	−0.005
`EWALL:BG (e092)`	0.218 (0.192–0.246)	0.237 (0.211–0.268)	+0.019
`WWLL:BG (e095)`	0.247 (0.231–0.265)	0.248 (0.230–0.270)	+0.001
`EWLL:BG (e095)`	0.239 (0.217–0.265)	0.252 (0.220–0.288)	+0.013

来源Source: .agent/jobs/2026-10-15-c2-outer-spectra-pn-fxt/REPORT.hermes.md Table 1; out_hermes/kT/*.npy; XMMLineMap E092, E095

3.8 O/Fe：归档对 C2，与 Liu et al. 2026 比较（L11）O/Fe: archived against C2, and the Liu et al. 2026 comparison (L11)

MEASURED（相对值）CONDITIONAL（绝对值）1T 轮廓似然 O/Fe（a_O/a_Fe，Asplund 2009 太阳标度）：C2 PN+FXT 使最佳值下降 10–45% —— 壁 17–28 kpc 8.26 → 6.81（dC(O/Fe=1) +8.9 → +5.4）、内部 17–28 kpc 7.47 → 5.88（+11.7 → +6.6）、壁 41–61 kpc 11.2 → 6.6；pooled 壁 28–61 kpc 4.87 → 6.33（上升）。所有区间与归档重叠；上界仍触格点边 20。太阳（O/Fe = 1）仍被排除的：pooled 17–28 kpc 壁与内部、pooled 41–61 kpc 内部（dC +5.4/+6.6/+12.4）；单区 17–28 kpc 与 28–61 kpc 壁不再排除太阳。合并似然注意：pooled 拟合对每个 ON 区重复计一次 BG 谱（e079_fit.py:220，R1 astra 发现、Claude 确认），profile-dC 区间未校准；单区拟合不受影响。与 Liu et al. 2026（arXiv:2609.37646）比较：其 RGS 核心区 O/Fe = 0.47（mCIE+CX，Lodders 2009 标度，与 Asplund 2009 数字比相同，×1.00），EPIC-MOS 1T 为 3.5（1–2′）、5.1（2–3′）、4.1（NE+SW 扇区）—— 绝对 1T 值偏高的直接证据；两者标度换算（Lodders→Asplund）尚未做（C6）。

MEASURED (relative) CONDITIONAL (absolute) 1T profile-likelihood O/Fe (a_O/a_Fe, Asplund 2009 solar scale): C2 PN+FXT lowers the best values by 10–45% — walls 17–28 kpc 8.26 → 6.81 (dC(O/Fe=1) +8.9 → +5.4), interior 17–28 kpc 7.47 → 5.88 (+11.7 → +6.6), walls 41–61 kpc 11.2 → 6.6; pooled walls 28–61 kpc rise 4.87 → 6.33. Every interval overlaps the archive; upper bounds still hit the grid edge at 20. Solar (O/Fe = 1) remains rejected for the pooled 17–28 kpc walls and interior and the pooled 41–61 kpc interior (dC +5.4/+6.6/+12.4); the single 17–28 kpc regions and the 28–61 kpc walls no longer reject solar. Pooled-likelihood caveat: pooled fits count the BG spectrum once per ON region (e079_fit.py:220, found by R1 astra, confirmed by Claude), so their profile-dC intervals are uncalibrated; single-region fits are unaffected. Comparison with Liu et al. 2026 (arXiv:2609.37646): their RGS nucleus O/Fe = 0.47 (mCIE+CX, Lodders 2009 scale, numerically identical ratio to Asplund 2009, ×1.00), EPIC-MOS 1T 3.5 (1–2′), 5.1 (2–3′), 4.1 (NE+SW sectors) — direct evidence that absolute 1T values are biased high; the Lodders→Asplund conversion is not yet done (C6).

谱区（高度）Spec (height)	归档 O/FeArchived O/Fe	C2（PN+FXT）C2 (PN+FXT)	dC(1) 归档→C2archived→C2
`EWLL+WWLL:BG (walls 17–28)`	8.26 (4.09–20.0)	6.81 (3.20–20.0 EDGE)	+8.9 → +5.4
`EAXL+WAXL:BG (interior 17–28)`	7.47 (4.18–17.87)	5.88 (3.19–12.95)	+11.7 → +6.6
`EWLM+WWLM:BG (walls 28–39)`	3.07 (1.23–10.79)	3.96 (1.29–15.76)	+1.5 → +1.5
`EWALL+WWALL:BG (walls 41–61)`	11.23 (1.60–20.0)	6.55 (1.05–20.0 EDGE)	+1.5 → +1.0
`EWLM+WWLM+EWALL+WWALL:BG (walls 28–61 pooled)`	4.87 (2.03–20.0)	6.33 (2.34–20.0 EDGE)	+3.4 → +3.6

E096（归档 MOS2+PN+FXT 版本）。左：FXT 60′ 图上的 O/Fe 谱区；右：按高度与侧的 O/Fe 轮廓似然（所有曲线共享一个值，dC ≤ 0.5）。E096 (archived MOS2+PN+FXT version). Left: the O/Fe regions on the FXT 60′ map; right: O/Fe profile likelihoods by height and side (all curves share one value, dC ≤ 0.5).

来源Source: REPORT.hermes.md Table 2; out_hermes/ofe/*.npy; XMMLineMap E096 §d; P1 REPORT.sonnet.md (Lodders/Asplund ×1.00)

3.9 晕剖面与 47 kpc 密度（L12）The halo profile and the density at 47 kpc (L12)

MEASURED（模型依赖）沿主轴（PA 165/345 ±30° 扇区，1′ 环）的面亮度梯度给出中心热晕：β = 0.5、r_c = 5.53 kpc 固定、kT 0.2 keV 下，nH(47 kpc) ≈ 5.5–5.8e-5 cm⁻³，2σ ≤ 8.6e-5（FXT 60′；rc/β/Rmin 全网格 ≤ 11.2e-5）；P(47 kpc) ≈ 4.1e-14 dyn cm⁻²（2σ ≤ 6e-14）。PN-only XMM 独立给出 2σ ≤ 5.1–12.1e-5。能把壁压平衡掉的晕（nH ≈ 1.4e-4）预测 S(8′)−S(22′) = 4.1–5.3 PW，而 FXT 60′ 实测 −0.1 ± 1.4 PW —— 在所有两侧情形中被排除。稳健性（S09 E084）：背景/前景网格、FXT 60′ 全链注入-回收（转移 1.00–1.18）、FXT A 对 B（各 2σ ≤ 8.0–9.0e-5）、±30% 软粒子形状（±0.3e-5）都通过。开放项：FXT 公共模渐晕曲线未测；0.2 对 0.3 keV 转换（10–20%）；以及一个 ~0.3–0.5 PW 的估值器方差大尺度系统（计入所有微弱大尺度主张的误差）。P2 静力学复核：在固定 n47 下 0.2 keV 接近 HSE 形状，但未observationally建立静力学平衡。

MEASURED (model-dependent) The major-axis brightness gradient (sectors within ±30° of PA 165/345, 1′ rings) gives the centred hot halo: with β = 0.5, r_c = 5.53 kpc fixed and kT 0.2 keV, nH(47 kpc) ≈ 5.5–5.8e-5 cm⁻³, 2σ ≤ 8.6e-5 (FXT 60′; ≤ 11.2e-5 over the whole rc/β/Rmin grid); P(47 kpc) ≈ 4.1e-14 dyn cm⁻² (2σ ≤ 6e-14). PN-only XMM independently gives 2σ ≤ 5.1–12.1e-5. A halo dense enough to balance the walls (nH ≈ 1.4e-4) predicts S(8′)−S(22′) = 4.1–5.3 PW against the measured FXT 60′ value of −0.1 ± 1.4 PW — excluded in every two-sided case. Robustness (S09 E084): the background/foreground grid, the FXT 60′ full-chain injection-recovery (transfer 1.00–1.18), FXT A against B (each 2σ ≤ 8.0–9.0e-5) and a ±30% soft particle-shape error (±0.3e-5) all pass. Open: the common-mode FXT vignetting curve is untested; the 0.2 vs 0.3 keV conversion (10–20%); and a ~0.3–0.5 PW estimator-variance large-scale systematic (carried in the errors of every faint large-scale claim). P2 hydrostatic recheck: at fixed n47, 0.2 keV is close to the HSE shape, but hydrostatic equilibrium is not observationally established.

S09 E084: halo density limit over the background, foreground and beta-model grid, for FXT 60 and PN-only XMM — S09 E084。47 kpc 晕密度上限随 rc、β、内拟半径与误差集的变化；FXT 60′ 与 PN-only XMM。S09 E084. The 47 kpc halo density limit against rc, β, inner fit radius and error set; FXT 60′ and PN-only XMM.

来源Source: cognition-os/experiments/E084_s09_halo_limit_robustness.md; XMMLineMap E091; P2 REPORT.sol.md

3.10 壁超压 P_wall/P_halo（L13）与 C2 重标定Wall over-pressure P_wall/P_halo (L13) and the C2 rescaling

MEASURED（由 L12）壁压力 P_wall = 2.3 nH_wall kT（nH_wall 由发射率、固定波段转换；kT 由 §3.7 表）。在壁的三维半径（3-d radius）r = √(z² + w_c²) 处对局部晕比较：WSW P_wall/P_halo = 2.95 ± 0.16（FXT 60′），随高度持平（d ln q / d ln z = 0.00 ± 0.06，legacy 残差标定误差）；ENE ~2.6（斜率 +0.47 ± 0.11）。联合 PN 图独立分支：WSW 2.7（< 38 kpc）/ 2.3（> 38 kpc）—— 与 FXT 60′ 一致。C2 重标定（固定密度）：kT_new/kT_old = 0.97（WSW 41–61 kpc）与 1.09（ENE），故 WSW 2.95 → 2.87、ENE 2.6 → 2.83。38–40 kpc 处无压力跳变；"壁超压、无跳变"的结论不变。P2 静力学范围：16 个 HSE 晕模型给出 WSW 平均 q = 2.16–3.78（永不接近 1），但斜率 −0.45 到 +0.60 —— "随高度持平"在模型族间脆弱；±0.16 是仅壁归一化误差（halo、几何、共享 kT 未含）。

MEASURED (from L12) The wall pressure is P_wall = 2.3 nH_wall kT (nH_wall from the emissivity at a fixed band conversion; kT from the §3.7 table). Compared against the local halo at the wall's 3-d radius r = √(z² + w_c²): WSW P_wall/P_halo = 2.95 ± 0.16 (FXT 60′), flat with height (d ln q / d ln z = 0.00 ± 0.06, legacy residual-scaled error); ENE ~2.6 (slope +0.47 ± 0.11). The independent joint-PN branch gives WSW 2.7 (below 38 kpc) / 2.3 (above 38 kpc) — consistent with FXT 60′. C2 rescaling (fixed density): kT_new/kT_old = 0.97 (WSW 41–61 kpc) and 1.09 (ENE), so WSW 2.95 → 2.87 and ENE 2.6 → 2.83. No pressure jump at 38–40 kpc; the "walls over-pressured, no jump" reading is unchanged. P2 hydrostatic range: 16 HSE halo models give a WSW mean q = 2.16–3.78 (never near 1), but slopes from −0.45 to +0.60 — "flat with height" is fragile across model families; ±0.16 is a wall-normalisation-only error (halo, geometry and shared kT not included).

WSW wall density and pressure against height, FXT 60 prime, filling factor 1 — WSW 壁密度 nH(z)（左，对数）与壁压力 P_wall(z)（右），FXT 60′、f = 1、归档 kT。虚线：P_halo(47 kpc) = 4.1e-14。数据：e096_wall_profile.json（PN 联合档案）。WSW wall density nH(z) (left, log scale) and wall pressure P_wall(z) (right), FXT 60′, f = 1, archived kT. Dashed: P_halo(47 kpc) = 4.1e-14. Data: e096_wall_profile.json (PN joint archive).

来源Source: XMMLineMap E096 §a, E098a; REPORT.hermes.md Table 3; P2 REPORT.sol.md (HSE grid); artifacts/s09_e087_joint_pn/PN/e096_wall_profile.json

3.11 质量与能量收支（L14、L17；L16 条件性）Mass and energy budgets (L14, L17; L16 conditional)

INFERRED（填充因子 f = 1）现今风（HK2020：每侧 Ṁ < 0.19 M⊙/yr、L_kin < 1.35e40 erg/s、v 475–570 km/s）无法建造或供养该结构。WSW 壁 14–38 kpc 质量 2.5–3.7e8 M⊙、热能 3–4e56 erg；38–80 kpc 质量 5.2–6.4e8 M⊙、E_th 5e56 erg；ENE 相近。按流动时间（74/156 Myr）折算，现今风在质量上差 20–40×、能量上差 10–20×。星暴机械功率（~8e41 erg/s）可在 30–40 Myr 内供上能量。INFERRED L17：L_X(0.5–2 keV, 14–80 kpc) ≈ 2e39 erg/s ≈ SN 功率的 0.25%（快照比值）。CONDITIONAL L16：光滑壁（热支撑）把 47 kpc 总晕压限制在 ≲ 3 P_X；类似 NGC 4594 的隐藏热相（f_X = 5e-3）需要 f_w ≤ 0.04。质量与能量都随标度变化：M_wall ∝ f_v^1/2；E_th ∝ kT；全部能量 ∝ P_halo(47 kpc)/4.1e-14。

INFERRED (filling factor f = 1) The present wind (HK2020: Ṁ < 0.19 M⊙/yr per side, L_kin < 1.35e40 erg/s, v 475–570 km/s) cannot build or feed the structure. WSW walls 14–38 kpc: mass 2.5–3.7e8 M⊙, thermal energy 3–4e56 erg; 38–80 kpc: 5.2–6.4e8 M⊙, E_th 5e56 erg; ENE similar. Per flow time (74/156 Myr) the present wind falls short by 20–40× in mass and 10–20× in energy. The starburst mechanical power (~8e41 erg/s) would supply the energy in 30–40 Myr. INFERRED L17: L_X(0.5–2 keV, 14–80 kpc) ≈ 2e39 erg/s ≈ 0.25% of the SN power (a snapshot ratio). CONDITIONAL L16: smooth, thermally supported walls limit the total halo pressure at 47 kpc to ≲ 3 P_X; an NGC 4594-like hidden hot phase (f_X = 5e-3) needs f_w ≤ 0.04. Scalings: M_wall ∝ f_v^1/2; E_th ∝ kT; all energies ∝ P_halo(47 kpc)/4.1e-14.

WSW 壁WSW walls	14–38 kpc	38–80 kpc
质量（f = 1）Mass (f = 1)	2.5–3.7e8 M⊙	5.2–6.4e8 M⊙
E_th	3–4e56 erg	5e56 erg
现今风供给质量需时Time for the present wind to supply the mass	1.4–2.0 Gyr	2.8–3.4 Gyr
……供给能量需时…to supply the energy	0.7–1.0 Gyr	1.1–1.3 Gyr
500 km/s 流动时间Flow time at 500 km/s	74 Myr	156 Myr

来源Source: XMMLineMap E096 §b; E098a; S09 FACTS (budgets); HK2020 = Hodges-Kluck et al. 2020, ApJ 903, 35

3.12 候选与非探测（L18–L21），伴星系效应与撤回清单Candidates and non-detections (L18–L21), companion effects, and the withdrawn list

MEASURED / INFERRED L18 前向激波鞘：forward-shock sheath: 既未探测也未被排除。双成分上限为预测的 1.36–10×；SHE 孔径被冷的（0.14–0.21 keV）壁裙气体主导；WSW 14–28 kpc 软超出不具全局显著性（p 0.27–0.45）。检验需约 2× 灵敏度。Neither detected nor excluded. Two-component bounds are 1.36–10× the prediction; the SHE aperture is dominated by cool (0.14–0.21 keV) wall-skirt gas; the WSW 14–28 kpc soft excess is not globally significant (p 0.27–0.45). A test needs about 2× more sensitivity.
MEASURED L19： WSW 14–28 kpc 软超出不具全局显著性（相关噪声下全局 p 0.27–0.45）。The WSW 14–28 kpc soft excess is not globally significant (global p 0.27–0.45 with correlated noise).
CANDIDATE L20： FXT 宽场小轴超出到 ~110 kpc 与 92–107 kpc 的三个 blob —— 仅候选，需 C4 的全局空假设重做。The FXT wide-field minor-axis excess to ~110 kpc and three blobs at 92–107 kpc — candidates only; the global nulls must be redone (C4).
CANDIDATE L21： PN-only XMM 在 55–66 kpc 轴上 C = +2.8 ± 1.0，FXT +0.6 ± 0.8（仪器间 ~2σ 不一致）；通过 C12 检验前不入论文。PN-only XMM shows axis emission at 55–66 kpc, C = +2.8 ± 1.0, against FXT +0.6 ± 0.8 (a ~2σ instrument disagreement); it stays out of the paper until the C12 checks pass.
MEASURED 伴星系 MCG 9-17-9（P5）：The companion MCG 9-17-9 (P5): 位置 (maj, mnr) = (−3.83′, −5.36′)，r 6.59′（36.5 kpc），距 WSW 轴 35.5°；是物理伴星（Δv 173–192 km/s，33 kpc HI 桥）。其点源已被扣除（XMM #5，40″ 掩罩）。L7 壁超出 B（7–10′）随 1/1.5/2′ 掩罩变化 −0.03/+0.10/−0.05 PW；烟囱边缘（z 6–7′）不动（w 6.10′ → 6.10/6.20/6.20′）。但 E095/E096 的 WWLM 谱区（WSW 壁 28–39 kpc）有 5%/18%/29% 像素在它 1′/1.5′/2′ 内；其周围 1–3′ 有 +1.4–1.7 PW 的微弱超出（~2σ_null）—— WWLM 的 kT/O/Fe 条件性暴露，需 2′ 掩罩重拟（C2 后续）。Position (maj, mnr) = (−3.83′, −5.36′), r 6.59′ (36.5 kpc), 35.5° from the WSW axis; a physical companion (Δv 173–192 km/s, a 33 kpc HI bridge). Its point source is subtracted (XMM #5, 40″ mask). The L7 wall excess B (7–10′) changes by −0.03/+0.10/−0.05 PW for 1/1.5/2′ masks; the chimney edge (z 6–7′) does not move (w 6.10′ → 6.10/6.20/6.20′). But 5%/18%/29% of the WWLM spectral region (WSW wall 28–39 kpc, E095/E096) lies within 1′/1.5′/2′ of it, and a faint +1.4–1.7 PW excess at 1–3′ sits at ~2σ_null — the WWLM kT/O/Fe are conditionally exposed and need a 2′-mask refit (C2 follow-up).
WITHDRAWN 撤回清单（不要引用）：Withdrawn list (do not cite): 壁到 60–70 kpc（E077/I029 旧表述）；ENE 填满锥到 39 kpc（是星系团）；E095 "epoch 按高度分裂"；E096 用高度 z 处晕算的超压；E098b 鞘排除；E098b-d 的 OUT:BG 温度诊断（孔径重叠）；I030 的 "filled cone to ~39 kpc"。walls to 60–70 kpc (the old E077/I029 wording); the filled ENE cone to 39 kpc (the cluster); E095's "epochs split in height"; E096's over-pressure at height z; the E098b sheath exclusion; the OUT:BG temperature diagnostics of E098b-d (overlapping apertures); I030's "filled cone to ~39 kpc".

来源Source: I021 (L18–L21, withdrawn list); XMMLineMap E098c–e, E098d; E086 (L21); P5 REPORT.sonnet.md