术语表与阅读清单Glossary and reading list

按拼音/字母排序。每一条给一句话定义加一个“在本项目里”的具体含义。Alphabetical. Each entry gives a one-sentence definition plus what it means concretely in this project.

1. 术语表1. Glossary

RGSRGS

Reflection Grating Spectrometer，XMM-Newton 上的反射光栅光谱仪。无狭缝设计：色散方向给波长，交叉色散方向给空间。本项目的全部数据来自 RGS1 第一阶。Reflection Grating Spectrometer, the reflection-grating spectrometer on XMM-Newton. It has no slit: the dispersion direction gives wavelength and the cross-dispersion direction gives position. All of this project's data come from RGS1, first order.

DISP / XDSPDISP / XDSP

RGS 场参考系的两条坐标：DISP 是色散方向（近似波长），XDSP 是交叉色散方向（近似天空位置）。源位置由 SAS 解析成 DELTA_DISP/DELTA_XDSP。The two coordinates of the RGS field-reference frame: DISP along dispersion (roughly wavelength) and XDSP along cross-dispersion (roughly sky position). SAS resolves the source position into DELTA_DISP/DELTA_XDSP.

条带（strip）Strip

沿交叉色散切出的一段矩形抽谱区。本项目每观测用三条 0.75′ 宽的条带，中心在 \(-0.75\)、\(0\)、\(+0.75\)′。A rectangular extraction interval along cross-dispersion. This project uses three strips 0.75 arcmin wide per observation, centred at \(-0.75\), \(0\) and \(+0.75\) arcmin.

响应矩阵 / RMFResponse matrix / RMF

redistribution matrix file，“入射能量到探测器通道”的概率表。由 SAS rgsrmfgen 生成；展源需要每个（源位置、条带）组合一个，9 格加 3 条带共 27 个。The redistribution matrix file, the table of probabilities from incident energy to detector channel. SAS's rgsrmfgen generates it; an extended source needs one per (source position, strip) pair, 27 in all for 9 cells and 3 strips.

rgsrmfgenrgsrmfgen

生成 RMF 的 SAS 任务。它把随 beta 变化的点源 cross-PSF 积分过抽谱多边形，所以孔径透过率随能量变化。项目的探针响应、细网格与标定源审计全部围绕它。The SAS task that generates the RMF. It integrates the beta-dependent point-source cross-PSF over the extraction polygon, so the aperture transmission is energy dependent. The project's probe responses, fine grids and calibrator audits all revolve around it.

线扩散函数 / LSFLine-spread function / LSF

line-spread function：一条无限窄的入射线在探测器上的分布。RGS 的 LSF 核心在 O VII r 处 \(\sigma \approx 437\) km/s（I020）；未建模展宽超过 85 km/s 就会盖过 \(\sigma_v\) 的物理读法。The line-spread function: the detector distribution of an infinitely narrow incident line. The RGS LSF core at O VII r is \(\sigma \approx 437\) km/s (I020); unmodelled broadening beyond 85 km/s would swamp the physical reading of \(\sigma_v\).

O VII 三重线（r、i、f）O VII triplet (r, i, f)

七次电离氧的 He-alpha 三重线：r（resonance，21.60 Å）、i（intercombination，21.81 Å）、f（forbidden，22.10 Å）。G 比 \((f+i)/r\) 对温度敏感，R 比 \(f/i\) 对密度敏感；本项目把三条当独立自由振幅，因为目标是速度场。The He-alpha triplet of seven-times-ionised oxygen: r (resonance, 21.60 Angstrom), i (intercombination, 21.81 Angstrom), f (forbidden, 22.10 Angstrom). The G ratio \((f+i)/r\) is temperature-sensitive and the R ratio \(f/i\) density-sensitive; this project fits the three lines as independent free amplitudes, because the target is the velocity field.

O VIII Ly-alphaO VIII Ly-alpha

类氢氧（八次电离）的 Lyman-alpha 线，18.97 Å。D29 目标的第二条窄带；O VIII 波段需要自己的 Chandra 模板（0.62–0.70 keV，I023）。The Lyman-alpha line of hydrogen-like oxygen (ionised eight times), at 18.97 Angstrom. The second narrow band of the D29 goal; the O VIII band needs its own Chandra template (0.62-0.70 keV, I023).

多普勒频移（Doppler shift）Doppler shift

视向速度 \(v\) 使谱线移动 \(\Delta\lambda/\lambda = v/c\)。在 21.8 Å 处，100 km/s 对应 0.0073 Å——与一个 XDSP 像素（约 3.6″）的几何色散几乎一样大，所以两者必须用讨厌参数分开。A line-of-sight velocity \(v\) shifts a line by \(\Delta\lambda/\lambda = v/c\). At 21.8 Angstrom, 100 km/s corresponds to 0.0073 Angstrom, almost the same as the geometric dispersion of one XDSP pixel (about 3.6 arcsec), so the two must be separated with nuisance parameters.

泊松似然 / C 统计量Poisson likelihood / C-statistic

每个通道的计数服从泊松分布；对数似然的负两倍习惯上叫 C 统计量（Cash 统计量）。项目在全部观测、全部条带上联合最小化它；嵌套模型的 \(\Delta C\) 服从 \(\chi^2\)。Each channel's counts follow a Poisson distribution; minus twice the log-likelihood is conventionally the C-statistic (Cash statistic). The project minimises it jointly over all observations and strips; \(\Delta C\) between nested models follows \(\chi^2\).

前向建模（forward modelling）Forward modelling

从天空模型出发，经仪器算子折叠成期望计数，再与数据比；不做反卷积。项目的仪器算子等于探针响应乘曝光剖面加缩放背景。Start from a sky model, fold it through the instrument operator into expected counts, and compare with the data; no deconvolution. The project's instrument operator = probe responses times the exposure profile, plus the scaled background.

讨厌参数（nuisance parameter）Nuisance parameter

物理上不感兴趣、但不放进拟合就会偏置物理量的量。本项目的 E 组：每观测偏移 \(\delta X_o\)、共同尺度 \(w\)、线翼 \(f\)、条带外接受度 \(a\)。同时放开只花 2.4 % 统计精度（RT15）。A quantity we do not care about physically, but which biases the physical quantities if left out of the fit. The project's set E: one offset \(\delta X_o\) per observation, the common scale \(w\), the wing \(f\), and the out-of-strip acceptance \(a\). Freeing all four at once costs only 2.4 % of statistical precision (RT15).

Fisher 标准误（Fisher SD）Fisher standard deviation

似然面在最优点的 Hessian 逆的对角线平方根；快，但假设抛物面。红队的“0.2 SD”类结论用它。The square root of the diagonal of the inverse Hessian of the likelihood at the optimum; fast, but it assumes a paraboloid. The red teams' "0.2 SD" style results use it.

轮廓似然（profile likelihood）Profile likelihood

固定一个参数、重新优化其余参数，\(\Delta C = 1\) 的范围给 68 % 区间；慢但不依赖抛物面假设。I029 的中括号区间用它。Fix one parameter, re-optimise the rest; the range where \(\Delta C = 1\) gives the 68 % interval; slow but free of the paraboloid assumption. I029's bracketed intervals use it.

Asimov 数据Asimov data

把模型期望值本身当“观测”的无噪声数据。用它检验拟合代码在真值处能否恢复真值（E018c 的 check B）。Noise-free "data" equal to the model expectation itself. It tests whether the fit code recovers the truth when the truth is the input (E018c's check B).

红队（red team）Red team

专职攻击承重结论的独立分析者（D26）。红队只提交发现；主 agent 复跑每条 material 反驳后才改信任状态。红队也会错：RT03 注入两次失败、RT17 公式反号，都是复跑后发现的。Independent analysts whose only job is to attack load-bearing claims (D26). Red teams submit findings; the main agent changes trust only after replaying every material objection. Red teams err too: RT03's injection failed twice and RT17's formula had the sign reversed, both found by replaying.

roll（滚转角）Roll angle

卫星绕视轴的旋转角。不同 roll 让同一天空结构投影到不同的色散几何，从而把空间分布与速度场解耦。三个在用观测近似 0°、90°、180°。The satellite's rotation about the line of sight. Different rolls project the same sky structure onto different dispersion geometries, decoupling the spatial distribution from the velocity field. The three observations in use are near 0, 90 and 180 degrees.

串扰（cross-talk）Cross-talk

条带外的天空通过响应翼进入条带计数的部分。N132D 条带里 14–38 % 的计数来自条带外（I027）。The part of a strip's counts that comes from sky outside the strip, through the response wings. For N132D, 14-38 % of strip counts come from outside (I027).

来源：各条目括号里的记录（03_TRUST_TABLE.md、insights/）；条带几何来自 .agent/jobs/2026-10-19-e018c-ovii-stage2/BRIEF.md；三重线波长来自 agent-memory/procedures.md（经 Suzuki et al. 2020 与 ATOMDB 3.0.9 核对）。Source: the record named in each entry (03_TRUST_TABLE.md, insights/); strip geometry from .agent/jobs/2026-10-19-e018c-ovii-stage2/BRIEF.md; triplet wavelengths from agent-memory/procedures.md (checked against Suzuki et al. 2020 and ATOMDB 3.0.9).

2. 阅读清单2. Reading list

2.1 科学背景Science background

• Suzuki et al. (2020), arXiv:2007.06158 —— N132D 的直接 RGS 等离子体参考：三成分 bvrnei、Wilms 丰度、固定银河 NH。本项目的 NEI 参数起点与谱线表都从这里来。Suzuki et al. (2020), arXiv:2007.06158: the direct RGS plasma reference for N132D, a three-component bvrnei with Wilms abundances and a fixed Galactic column. This project's NEI starting parameters and line table come from it.
• Vogt & Dopita (2011), arXiv:1009.0964 —— N132D 外层马蹄形椭圆的方向（NE-SW）；roll 角选择的依据之一。Vogt & Dopita (2011), arXiv:1009.0964: the orientation of N132D's outer horseshoe ellipse (NE-SW); one basis for the roll-angle choice.
• Xiao & Chen (2008), DOI 10.1016/j.asr.2007.04.019 —— N132D 东北方向的 breakout；同上。Xiao & Chen (2008), DOI 10.1016/j.asr.2007.04.019: N132D's northeast breakout; same use.

2.2 仪器与标定Instrument and calibration

• XMM-Newton Users' Handbook（RGS 章节）—— RGS 几何、坐标与背景的官方描述。The XMM-Newton Users' Handbook (RGS chapters): the official description of RGS geometry, coordinates and background.
• SAS User Guide 第 5.10 节 —— xpsfincl/xpsfexcl 是点源分数，展源填满孔径时会失效；点名 N132D。SAS User Guide section 5.10: xpsfincl/xpsfexcl are point-source fractions that become invalid when an extended source fills the aperture; N132D is named.
• XMM-SOC-CAL-TN-0058 —— 直接用 N132D 测试标准局部背景：背景里含源发射线，线通量会错；由此确立模板背景路线。XMM-SOC-CAL-TN-0058: tests the standard local background directly on N132D; the background contains source emission lines and line fluxes go wrong, which established the template-background route.
• rgsrmfgen 算法文档（本地 SAS 安装 doc/rgsrmfgen/node8.html）—— NarrowLSF/BroadLSF、曝光图乘 cross-PSF 孔径损失、CCD QE 与 PI 香蕉因子；I014 的响应因子分解依据。The rgsrmfgen algorithm page (local SAS install, doc/rgsrmfgen/node8.html): NarrowLSF/BroadLSF, exposure map times cross-PSF aperture loss, CCD QE and the PI banana; the basis of I014's response factorisation.

2.3 统计与写作Statistics and writing

• Cash (1979) —— C 统计量的原始论文（泊松似然比）。Cash (1979): the original paper on the C-statistic (the Poisson likelihood ratio).
• Wilms, Allen & McCray (2000) —— wilm 丰度表，NEI 参考模型使用。Wilms, Allen & McCray (2000): the wilm abundance table used by the NEI reference model.
• Audard et al. (2003) —— Capella 演示的丰度参考（XSPEC angr 转换）。Audard et al. (2003): the abundance reference of the Capella demo (the XSPEC angr conversion).
• Zhang et al. (2019) —— 4 arcmin 窗口在约 5 arcmin RGS 视场里的用法；auto 背景阈值的工程来源（非文献规则）。Zhang et al. (2019): a 4 arcmin window in an about 5 arcmin RGS field; the engineering origin of the auto background threshold (not a literature rule).

来源：agent-memory/environment.md（N132D 参考模型、背景指导与 RGS 响应内部结构各节）；docs/science-contract.md（Audard 丰度）；agent-memory/procedures.md（Zhang et al. 阈值说明）。以上文献均在本项目记录中实际引用；未逐条核对 ADS 条目，标记为待核对。Source: agent-memory/environment.md (the N132D reference model, background guidance and RGS response internals sections); docs/science-contract.md (Audard abundances); agent-memory/procedures.md (the Zhang et al. threshold note). Every reference above is actually cited in the project's records; the ADS entries were not individually re-checked, and are marked as such.