XARTATOMS · M104 热晕 X 射线分析手册XARTATOMS · M104 hot-halo X-ray analysis handbook

1. 区域方案（region schemes）

1. Region schemes

• 环（annuli）：R2-4、R4-6、R6-8、R8-10、R9-11…R12-14，纯径向控制（Scheme C）。
• Annuli: R2-4, R4-6, R6-8, R8-10, R9-11…R12-14 — the radial-only control (Scheme C).
• 半环（Angle1/Angle2）：Angle1 是东西侧（天球 PA 45–135 与 225–315°），Angle2 是南北侧补集。这是 v18 的主数据集（67 个区域）。
• Half annuli (Angle1/Angle2): Angle1 is the east+west side (celestial PA 45–135 and 225–315°), Angle2 the north+south complement. This is the main v18 dataset (67 regions).
• 扇区（sectors）：R4–7 四扇区（Disk W / Wind N / Disk E / Wind S，90°）与 22.5° 偏置的八扇区（45°）。扇区必须用验证过的 DS9 fk5 pie() 文件（D30 例外合同）。
• Sectors: R4–7 four sectors (Disk W / Wind N / Disk E / Wind S, 90°) and the 22.5°-offset eight sectors (45°). Sectors must use verified DS9 fk5 pie() files (the D30 qualified exception).
• WVT 自适应分区：31 区（SNR≈25，最小面积 1 arcmin²），MOS2 31/31 完整，MOS1 26/31（R04/R13/R14/R15/R23 探测器不可用）。
• WVT adaptive binning: 31 regions (SNR≈25, min area 1 arcmin²); MOS2 31/31 complete, MOS1 26/31 (R04/R13/R14/R15/R23 detector-unusable).

chandra_xmm_joint_20261001/config/common_regions.json（Angle1/2 定义）；cognition-os/03_TRUST_TABLE.md T86–T88

2. v18 拟合模型

2. The v18 fitting model

源:    1T APEC (kT 自由, Z=0.3 Z☉ 冻结; 0.4–3.2 keV)
       — 脚本名里的 "2T" 是历史命名; v18 的源模型是
         单个 APEC, kT 自由 (v16 起取代了 v5–v15 的固定 T=0.75)
背景:  CXB (xspegpwrlw, norm 0.002 冻结) + QPB 模板 + SP 幂律
       + LHB/MW halo APEC + 仪器线经验高斯 (Al 1.49, Si 1.75, ~1.3 keV)
       + 背景缩放因子 (v11+)
统计:  chi2gehrels 分组; conf() 区间
联合:  多 ObsID × MOS1/MOS2 同时拟合, 源参数共享
归一:  所有天空成分乘 skyarea=BACKSCAL·(1/20/60)² arcmin²,
       保存的 flux_* 与 SrcApec.norm 均为每 arcmin² (B020)

关键约定：背景参数在 66/67 区域冻结（T30）→ 温度梯度不来自背景系统差；缺 SP 成分时外区 kT 会失控到 4–12 keV（T43）；参数边界不是误差棒；不要把保存的 flux 再除以 area_arcmin2（B020：双重归一，Deprecated）。

Key conventions: background parameters frozen in 66/67 regions (T30) → the temperature gradient is not a background systematic; without the SP component outer kT runs away to 4–12 keV (T43); parameter bounds are not error bars; do not divide the saved flux by area_arcmin2 again (B020: double normalisation, Deprecated).

cognition-os/PITFALLS.md P09；cognition-os/03_TRUST_TABLE.md T30/T43/T148；cognition-os/bug_impacts/B020_*.md

2a. GTI 与点源遮蔽（2026-10-15 新结果）

2a. GTIs and source masking (new results, 2026-10-15)

• GTI：保留 2.filter（I066/T150）。标准 ESAS 2.filter 已使 MOS2/PN 的 FWC 归一化 IN/OUT ≈ 0.95–1.06；更严格的多尺度切割只保留 40–80% 曝光、换来 ≤0.04（~1σ）的改善，最严格档保留 22–70% 且无显著收益 → 不采用。MOS1 在每个指向都持续偏高（中心指向 1.07±0.02 [6–9.5 keV]；1.20±0.02 [6–9.5+10–12 keV, 2–15′]；0900170701 为 1.16；0900170501 为 1.06），而同指向的 MOS2 是 0.97–1.06 → 器件/静态 SP，不是天空；转为 Q-A 阶段 2 的拟合模型问题（MOS1-free vs MOS1 加自有 SP 项）。真实的 2.1.filter GTI/clean 产品只存在于 0900170601；所有 v18 拟合用的都是 A。
• GTI: keep 2.filter (I066/T150). The standard ESAS 2.filter already gives FWC-normalised IN/OUT ≈ 0.95–1.06 for MOS2/PN; a stricter multiscale cut keeps only 40–80% of the MOS exposure for a ≤0.04 (~1σ) gain, and the strictest keeps 22–70% with no significant gain → not adopted. MOS1 stays high in every pointing (central 1.07±0.02 [6–9.5 keV]; 1.20±0.02 [6–9.5+10–12 keV, 2–15′]; 1.16 for 0900170701; 1.06 for 0900170501) while MOS2 of the same pointings is 0.97–1.06 → instrumental/quiescent SP, not sky; it becomes a Q-A stage-2 fitting-model question (MOS1-free vs MOS1 with its own SP term). Real 2.1.filter GTI/clean products exist only for 0900170601; every v18 fit used A.
• mask 规则 v2（E035）：点源半径取 F·PSF(R) = 1× 晕面亮度的位置（x=1；中位 30″，最大 360″；更保守的 x=10 档中位 20″）；星系团 WARP J1240.4−1147 遮 242″（z = 0.193，12.3′，R500 = 3.05′），1eRASS J124032.2−120437 遮 644″。45′ 内共 958 个 X 射线源（CSC 2.1 + 4XMM-DR14/DR14s + 自建 stack；4XMM 不含偏置指向，0501 需自建 stack data/edetect_stack/srclist_sum.fits）。遮蔽无法把残余压到晕面亮度的 5% 以下（R11-13 Angle2 的残余/晕比：当前 mask 3.6 → v2 x=1 后仍 0.95）→ PSF 翼与星系团发射必须建模（星系团作为 APEC 成分 + 空间加权残余模型），且固定 CXB 归一须与 mask 流量阈值匹配。
• Mask rule v2 (E035): point-source radius where F·PSF(R) = 1× the halo SB (x=1; median 30″, max 360″; the more conservative x=10 level gives a median of 20″); the cluster WARP J1240.4−1147 masked at 242″ (z = 0.193, 12.3′, R500 = 3.05′), and 1eRASS J124032.2−120437 at 644″. 958 X-ray sources within 45′ (CSC 2.1 + 4XMM-DR14/DR14s + the own stack; 4XMM lacks the offset pointings, so 0501 needs the own stack data/edetect_stack/srclist_sum.fits). Masking cannot push the residual below 5% of the halo SB (residual/halo in R11-13 Angle2: 3.6 with the current mask → still 0.95 after v2 x=1) → the PSF wings and cluster emission must be modelled (the cluster as an APEC component + a spatially weighted residual model), and the fixed CXB normalisation must match the mask flux cut.

cognition-os/insights/I066_*.md（T150/T151）；cognition-os/experiments/E035_*.md；.agent/jobs/2026-10-14-source-mask-design/out/sonnet/mask_*_v2_*；D47

3. 方法矩阵（Chandra+XMM 联合）

3. The method matrix (Chandra+XMM joint)

analysis_plans/M104_Chandra_XMM_temperature_comparison_20261001.md（方法矩阵表）

20 个单元里 8 个 PASS（2026-10-05 状态）。R6–8 的非物理值（0.109 keV 开区间、双分支、2.04 keV）与 v18 ≈ 0.75 keV 形成鲜明对比。 8 of 20 cells PASS (state of 2026-10-05). The unphysical R6–8 values (0.109 keV open interval, bimodal, 2.04 keV) contrast with v18 ≈ 0.75 keV.

chandra_xmm_joint_20261001/products/full_temperature_comparison.csv

4. E031 温度图交付设计与 20% 验收规则

4. The E031 map delivery design and the 20% acceptance rule

E031 设计（XMM 主图 + Chandra 辅助）：XMM 有效覆盖提供大尺度主图；Chandra 先做点源/污染控制，只在有可比支持的局部区域提供联合测温。不要求每个像元包含全部相机。分区不是按计数，而是按“预计温度精度”：用完整谱信息矩阵（含共享背景协方差），lnT 近似对称时两侧 ≤20% 对应信息量 I ≥ 30.08（设计裕量建议 I ≥ 36）。区域增长后必须用真实谱验收：名义 68% 单一连续区间、两侧相对误差 ≤20%、95% 无竞争分支；不达标就灰标，不插值。

The E031 design (XMM main map + Chandra assist): XMM coverage provides the large-scale primary map; Chandra first controls point sources/contamination and adds joint temperature measurements only where comparable support exists. No pixel is required to contain every camera. Binning is driven by predicted temperature precision, not counts: using the full-spectrum information matrix (including shared background covariance), both sides within 20% with a symmetric lnT interval corresponds to information I ≥ 30.08 (a design margin of I ≥ 36 is suggested). Grown regions must pass real-spectrum acceptance: a single continuous nominal 68% interval, both sides within 20%, no competing branch at 95%; otherwise the bin is greyed out, never interpolated.

analysis_plans/m104_temperature_area_20261005/m104_temperature_map_delivery_strategy_zh.md；temperature_map_binning_recipe_zh.md §2

E030 的 8 个真实区域：绿=通过 20% 统计检验，红=不通过。北侧单扇区（north）是唯一不通过者。 E030's 8 real regions: green = passes the 20% statistical test, red = fails. The north single sector is the only failure.

analysis_plans/m104_temperature_area_20261005/actual_summary_v1/actual_combinations.csv

5. C1/X1/J1/X2/J2 之外的本地策略家族

5. Local strategy families beyond C1/X1/J1/X2/J2

cognition-os/02_CURRENT_QUESTION.md（Q-A 列出的策略）；analysis_plans/m104_temperature_area_20261005/actual_src_bkg_results_zh.md