跳至正文

「第四种中微子」的三十年生灭账

目录

机制裁决第 144 篇 · 对称双向第 139 篇 · section B 物理被神化的大叙事 · 全库第 202 篇 本篇审的是「第四种中微子」这个名号从被声称到被宣布死亡，中间经历了什么。先读三句红线：

  1. 本篇不裁决任何粒子物理理论对错、不评价任何实验组与任何在世研究者、不给出「该不该继续寻找惰性中微子」的意见——只审「第四种中微子曾被发现」「这些异常全是系统误差闹剧」「这个领域已经结案」这三类句子从文件里读出来时承不承得住。
  2. 「一个解释被排除」与「一份测量被推翻」是两件不同的事：MicroBooNE 2025 年排除的是「LSND 与 MiniBooNE 异常的单一轻惰性中微子解释」这个带限定词的对象，不是 LSND 那 87.9 个多出来的事件本身。本篇对称审计「第四种中微子已被发现」与「异常全是错觉、领域空无一物」两个方向。
  3. 全篇承重句均给出可点击来源；英文逐字引用一律标注层级（一手逐字＝全文取回；摘要逐字＝仅摘要取回；需亲核＝仅元数据）。引用保留出处排版原貌（arXiv 摘要页的 LaTeX 写法、PDF 文本层的上下标平铺），不逐字处均为转述并写明。本篇出现三套统计语言不混用：单实验内部显著度（σ）、排除置信度（CL/CLs）、全球拟合参数拟合优度（PG p 值）——各自出处单列，互不换算。证据截止日 2026-08-12。

零、一句话裁决

LSND、MiniBooNE、反应堆、镓这四条异常全是真实的测量记录——每条都经过多轮复核与独立检验活到今天——但「第四种中微子」这个名号把三件不同的事混在了一起：一个实验内部相对自己零假设的 σ 偏离（不跨实验转移）、一个随理论分母选择而生死的「异常」标签（Huber–Müller 换尺生出反应堆异常、Kopeikin 换尺几乎消解它、2023 年法国新模型又让它「复活」至 2.2σ）、以及一个在全球拟合里被自己的两条腿互相否决的 3+1 模型（Diaz 2019 同文：相对三味模型偏好 5σ、内部张力 4.5σ）。真的不是「第四种中微子曾被发现然后被排除」（截至 2026-08-12，没有任何一个共同体级判据宣布过「发现」；被排除的始终只是带限定词的解释通道）和「异常全是系统误差闹剧、领域空无一物」（LSND 的 87.9 个事件至今无逐字解释；镓异常在 ⁷¹Ge 半衰期逃生门被两把独立高统计测量关闭后「survives with high statistical significance」；5 MeV bump 在所有通量模型下未解；Snowmass 白皮书自陈异常「still persist at the level of 1-5σ」）这两层被声称的定论。把「一个实验相对它自己的零假设偏了多少」读成「宇宙里有（或没有）一种新粒子」，与把「一种解释死了」读成「一份测量死了」，是同一个动作的两个方向。

一、本篇测什么：异常跳 × 换尺跳 × 守零跳 ＋ 反向红跳与六件可判定的事

本篇审的不是「惰性中微子存在吗」（这是物理学家的事），而是「『第四种中微子』这个名号从被声称到被宣布死亡，中间经历了哪些跳」。拆成六件可判定的事：

  1. 守真锚：三味振荡框架真不真、异常家族有哪几条——本篇全部裁决建立在「两端都是真」上。
  2. 异常跳：appearance 家族（LSND→MiniBooNE）的 σ 怎么从「单实验内部偏离」被读成「新粒子证据」——以及通行三十年的「LSND 3.8σ」究竟出自哪份文件（取证发现：不出自 LSND 自己）。
  3. 换尺跳：反应堆异常 2011 年生于分母换尺、2021 年几乎死于分母再换尺、2026 年又随新分母「复活」——异常的分母是理论汇编不是测量，这是 g-2 篇的镜像。
  4. 守零跳（反例侧）：镓异常如何在两轮「解释路线被关闭」后仍活着——它是「领域已结案」叙事的最大反例。
  5. 判据账：为什么单实验 σ 凑不成「第四种中微子成立」——appearance 与 disappearance 两个家族在全球拟合里互相否决的定量账。
  6. 消费账与反向红跳：媒体三十年「判死→复活→再判死」的话术化石，与「被错杀的真测量」登记。

结构胎记：异常跳 × 换尺跳 × 守零跳，附反向红跳。

  • 异常跳：把「一个实验内部相对零假设的 σ 偏离」读成「自然界有第四种中微子的证据」——σ 度量的是该实验自己的零假设，不跨实验转移；且 appearance 家族的正信号与 disappearance 家族的 null 在全球拟合里互相否决（Dentler 2018 的 PG 检验 p<2.6×10⁻⁶）。
  • 换尺跳：把「预期通量/截面的理论分母被重算」读成「异常被实验否定」或「异常复活」——反应堆异常的分母（Huber–Müller 通量汇编）是理论不是测量：Mention 2011 用新分母算出 0.943(0.023)「生出」异常，Kopeikin 2021 重测 β 比把分母压回、Giunti 等宣布「demise」，同一批作者 2026 年换用 2023 年法国新分母又让异常「复活」至 2.2σ——反应堆数据本身一次没动过。
  • 守零跳：把「某一解释通道被排除」读成「第四种中微子不存在、异常全部解决」——MicroBooNE 2025 排除的是「single light sterile neutrino interpretation」；KATRIN 排除的是镓异常的最佳拟合点（96.6%）；而 LSND/MiniBooNE 的过量、镓的亏缺、5 MeV bump 这些测量本身无一被逐字解释。
  • 反向红跳：反过来把「3+1 全球拟合被否决」读成「LSND/MiniBooNE/BEST 测错了」同样不立——KARMEN 写的是「does not confirm」不是「LSND 错了」；⁷¹Ge 半衰期逃生门被关闭两次后镓异常「survives」；keV 暗物质与 see-saw 标尺上的惰性中微子是另一本账（本篇不审，只登记）。

共用动作：把「一个实验相对它自己的零假设偏了多少」读成「宇宙里有（或没有）一种新粒子」；以及反过来，把「一种解释死了」读成「一份测量死了」。

与库内相关篇的分界：

  • 与物质-反物质篇（2026-07-25）分界写死：那篇管中微子质量起源、0νββ、leptogenesis（重惰性中微子作为宇宙学玩家，其 sterile 命中是 Davidson-Nardi-Nir 综述语境）；本篇管 eV² 尺度「第四种中微子」假说在实验室异常里的生灭。那篇已有的中微子振荡确证链本篇只称重不重做，补的是那篇没有的异常家族三十年账。
  • 与暗物质篇（2026-07-08）分界写死：那篇审 keV 惰性中微子作为暗物质候选（Angus/MOND 语境）；本篇不审暗物质候选身份（第九章仅登记分界）。
  • 与 g-2 篇（2026-08-04）镜像分界：那篇审单条张力死于理论预言汇编换尺（HVP 数据驱动→格点，「张力归零」）；本篇审一整族异常，其中反应堆一条生于换尺、几乎死于换尺、再随新通量汇编「复活」（「异常生灭」）。两篇合读是「分母账」的正反两面：那边动的尺在理论预言侧，这边动的尺也在理论预言侧——数据都没动。
  • 与系外行星生物特征篇（2026-08-08）分界：那篇审「一个团队的初步证据→共同体宣布发现」的判据跳；本篇共享宣布学手法（σ 语义与话术化石），但对象是地面实验室异常家族，判据机制不同（粒子物理靠独立实验互搏与全球拟合，无 CoLD 式标尺）。
  • 与质子衰变篇（2026-08-05）分界：那篇的守零是「五十年没看到衰变≠理论死了」；本篇的守零跳是「解释被排除≠测量错了≠领域结案」。同名不同物，分界不重叠。

编号落位：机制裁决第 144 篇·对称双向第 139 篇·section B·全库第 202 篇（2026-08-12 全库扫描确认未被占用：index.md 与正篇 md 双源最大值均为 143/138/201）。去重实测（python 全库 201 个 md 文件）：LSND/MiniBooNE/MicroBooNE/STEREO/PROSPECT/DANSS/NEOS/镓异常/Huber-Mueller/反应堆异常 全零命中；sterile neutrino 1 处＝物质-反物质篇 leptogenesis 语境；惰性中微子 1 处＝暗物质篇 MOND 语境；BEST 3 处在表观时钟篇（别义）、3+1 4 处为时空维度（别义）——专篇零命中，处女地，邻近但不重叠。

二、守真锚：三味振荡是教科书级真，异常家族有四条

本库原则：审「声称」必须回到声称的原件；审「异常」必须先确认「正常」是真的。 「第四种中微子」这个主张的全部前提是「三种已确立」——先把这一端钉死，再点名异常家族。

2.1 三味框架：五条确证链，全部一手逐字

  • 大气端（Super-Kamiokande 1998，PRL 81, 1562）：33.0 kt·yr（535 天）曝光，μ 中微子天顶角依赖缺失，与双味 νμ↔ντ 振荡一致。摘要 [一手逐字]：The data exhibit a zenith angle dependent deficit of muon neutrinos which is inconsistent with expectations based on calculations of the atmospheric neutrino flux. ... The data are consistent, however, with two-flavor nu_mu <-> nu_tau oscillations with sin^2(2theta)>0.82 and 5x10^-4 < delta m^2 < 6x10^-3 eV^2 at 90% confidence level.（arXiv:hep-ex/9807003）
  • 太阳端（SNO 2001，PRL 87, 071301；2002，PRL 89, 011301）：2001 年 CC 率对 Super-K ES 通量给出 [摘要逐字]：a 3.3 sigma difference, providing evidence that there is a non-electron flavor active neutrino component in the solar flux（PDF 正文版措辞为 a 3.3σ difference, assuming the systematic uncertainties are normally distributed，两版同源）；2002 年 NC 直接测量 [一手逐字]：The non-electron neutrino component is 3.41+/-0.45(stat.)+0.48,-0.45(syst.) x10^6/(cm^2 s), 5.3 standard deviations greater than zero, providing strong evidence for solar electron neutrino flavor transformation. The total flux measured with the NC reaction is 5.09 +0.44,-0.43(stat.)+0.46,-0.43(syst.)x10^6/(cm^2 s), consistent with solar models.（arXiv:nucl-ex/0106015、arXiv:nucl-ex/0204008）——太阳中微子没有消失，是换了味；总活性通量与太阳模型一致。
  • 反应堆端（KamLAND 2002/2003，PRL 90, 021802）：162 吨·年曝光，观测/无消失预期之比 [一手逐字]：the ratio of the number of observed inverse $\beta$-decay events to the expected number of events without disappearance is $0.611\pm 0.085 {\rm (stat)} \pm 0.041 {\rm (syst)} $ for $\bar{\nu}_e$ energies $>$ 3.4 MeV. The deficit of events is inconsistent with the expected rate for standard $\bar{\nu}_e$ propagation at the 99.95% confidence level.（arXiv:hep-ex/0212021）
  • θ13 端（Daya Bay 2012，PRL 108, 171803；RENO 2012，PRL 108, 191802）：大亚湾 [一手逐字]：has measured a non-zero value for the neutrino mixing angle $\theta_{13}$ with a significance of 5.2 standard deviations ... $\sin^22\theta_{13}=0.092\pm 0.016({\rm stat})\pm0.005({\rm syst})$ in a three-neutrino framework.（arXiv:1203.1669）；RENO 独立 4.9σ（arXiv:1204.0626，摘要逐字 with a significance of 4.9 standard deviations）。
  • 官方封顶（2015 诺贝尔物理学奖）：授予梶田隆章（Super-K）与 McDonald（SNO），引文 [一手逐字]：for the discovery of neutrino oscillations, which shows that neutrinos have mass（nobelprize.org 2015-10-06 官方新闻稿，新闻稿引文格与 meta 区措辞一致）。

现行权威表述（PDG 2025《Neutrino Masses, Mixing, and Oscillations》，Gonzalez-Garcia & Wendell，2025-08 修订） [一手逐字]：The minimum scenario to describe these results requires the mixing between the three flavour neutrinos of the standard model in three distinct mass eigenstates.（rpp2025-rev-neutrino-mixing.pdf，§14.7「The 3ν Paradigm」）同文 §14.8 把「第四种」的入场条件写得同样清楚：whatever the extension of the SM we want to consider, it must contain only three light active neutrinos. Therefore if we need more than three light massive neutrinos, we must add sterile neutrinos to the particle content of the model.

2.2 判据底座：5σ 是约定不是自然律

本篇全部「σ」都按领域自己的约定读，不神化：

  • PDG 2025《Statistics》（Cowan & Lista，2025-06 修订） [一手逐字]：Often in particle physics the level of significance where an effect is said to qualify as a discovery is Z = 5, i.e., a 5σ effect, corresponding to a p-value of 2.87 × 10−7. One's actual degree of belief that a new process is present, however, will depend in general on other factors as well, such as the plausibility of the new signal hypothesis and the degree to which it can describe the data, one's confidence in the model that led to the observed p-value, and possible corrections for multiple observations out of which one focuses on the smallest p-value obtained (the "look-elsewhere effect"...).（rpp2025-rev-statistics.pdf）同文还有一句对本篇至关紧要：It must then be communicated carefully that in excluding s = 0 at, say, 90% or 95% confidence level from the interval, one is not necessarily claiming to have discovered the effect, for which one would usually require a higher level of significance (e.g., 5 σ).——90/95% CL 的排除零 ≠ 发现，这把尺子对「排除」与「发现」两个方向一视同仁。
  • Lyons 2013《Discovering the Significance of 5 sigma》 [一手逐字]：It has become the convention in Particle Physics that in order to claim a discovery of some form of New Physics, the chance of just the background having a statistical fluctuation at least as large as the observed effect is equivalent to the area beyond 5σ in one tail of a normalised Gaussian distribution, or smaller; i.e. the p-value is no larger than 3 × 10−7. Indeed, journals are very reluctant to allow the word 'discovery' to appear in the title, abstract or conclusions of a paper, unless the p-value is that small. 以及历史论据：In the past there have been many 'phenomena' that corresponded to 3 or 4σ effects that have gone away when more data were collected. The 5σ criterion is designed to reduce the number of such false claims.（arXiv:1310.1284）Lyons 本人对一刀切持保留：This is far too blunt a tool for dealing with issues such as the Look Elsewhere Effect, the plausibility of the searched-for effect, the role of systematics, etc.

守真锚含义一：本篇引用的每一个「3.8σ」「4.8σ」「6.0σ」都是单实验内部的零假设偏离度，按约定连「discovery」的字眼都不配用——它们从一开始就是「evidence 级」不是「发现级」，媒体与后来的叙事把它们读高或读死，都发生在这把尺之外。

2.3 异常家族点名：共同体自己的编组

「异常家族有几条」不是本篇的发明，是共同体的制度编组：

  • P5 2023 报告（美国粒子物理项目优先排序委员会，2023-12-08 HEPAP 批准）[一手逐字]：MeV/m have found some anomalous results: The liquid scintillator neutrino detector (LSND) anomaly, the reactor antineutrino anomaly, the MiniBooNE low-energy excess, and the gallium anomaly.（arXiv:2407.19176，§1）
  • Snowmass 2021 NF02 专题组的正式名称就是「Understanding Experimental Neutrino Anomalies」——异常家族作为一个独立编组被处理（arXiv:2209.05352 摘要逐字 based on the NF02 Topical Group-contributed White Paper on Light Sterile Neutrino Searches and Related Phenomenology）。其白皮书（2022，J. Phys. G 51, 120501 (2024)）按源（source）／反应堆（reactor）／π 静止衰变（decay-at-rest，LSND 属此）／π 飞行衰变（decay-in-flight，MiniBooNE/MicroBooNE 属此）分 sector——本篇章节划分沿用这个官方编组。
  • PDG 2025 混合综述 §14.8 的历史句 [一手逐字]：Historically, models with extended light neutrino sectors were invoked to explain a set of anomalies observed at short baselines and which could not be explained within the 3ν framework but could be interpreted as hints for the existence of additional neutrino states with masses at the eV scale.

守真锚含义二：四条异常（LSND／MiniBooNE 低能过量／反应堆／镓）每条都是真实的、被共同体制度性承认的测量记录；「第四种中微子」是被请来解释它们的一个候选解释。本篇的全部裁决都在「测量记录」与「解释候选」这条分界线上展开——两边后来经历了完全不同的命运。

三、异常跳账：appearance 家族三十年——一个实验的 σ 怎么被读成「新粒子证据」

异常跳的第一现场是 ν̄μ→ν̄e（及 νμ→νe）出现通道：LSND 看到过量、KARMEN 没看到、MiniBooNE 又看到、MicroBooNE 逐通道否决、MicroBooNE 双束流收口。本账全部按原件记，σ 的归属与语义逐条钉住。

3.1 LSND 1995：异常的原点，措辞是「候选事件」

LSND（Los Alamos 液体闪烁体探测器，μ⁺ 静止衰变束流）1995 年首报 [一手逐字]：Using strict cuts to identify γ's correlated with positrons results in a signal of 9 events, with an expected background of 2.1 ± 0.3. A likelihood fit to the entire e+ sample yields a total excess of 16.4+9.7 −8.9 ±3.3 events, where the second uncertainty is systematic. If this excess is attributed to neutrino oscillations of the type ν̄µ → ν̄e , it corresponds to an oscillation probability of (0.34+0.20 −0.18 ± 0.07)%.（arXiv:nucl-ex/9504002，PRL 75, 2650 (1995)，卷页经 arXiv citation_doi ＋ INSPIRE 双源一致，APS 页被拦标需亲核）

注意原件的限定结构：超出是「似然拟合总超出 16.4⁺⁹·⁷₋₈.₉±3.3 事件」；「振荡概率」是在「If this excess is attributed to neutrino oscillations」条件下的换算。从第一行起，「测量到的过量」与「归因于振荡」就是两层东西。

3.2 LSND 2001 终版：原文全文没有一个 σ

终版（六年数据）报总超出 87.9±22.4±6.0 事件，折合振荡概率 (0.264±0.067±0.045)% [一手逐字]：A total excess of 87.9 ± 22.4 ± 6.0 events consistent with ν̄e p → e+ n scattering was observed above the expected background. This excess corresponds to an oscillation probability of (0.264 ± 0.067 ± 0.045)%, which is consistent with an earlier analysis.（arXiv:hep-ex/0104049，PRD 64, 112007 (2001)）

取证发现（本篇最要紧的 σ 来源学之一）：通行三十年的「LSND 3.8σ」不出自 LSND 自己。对 LSND 2001 全文 136,233 字符做 python 全文检索，3.8／standard deviation／significance 均零命中——原文的显著度口径是 Table X 涨落概率 [一手逐字]：Rγ > 1 205 106.8 ± 2.5 39.2 ± 3.1 59.0 ± 14.5 ± 3.1 7.8 × 10−6（Rγ>1 样本涨落概率 7.8×10⁻⁶；Rγ>10 样本 1.1×10⁻⁴）。「LSND (3.8σ)」最早可核的出处是 MiniBooNE 2018/2020 论文正文对 LSND 的转引（见 3.4）。这不是吹毛求疵：σ 是一种零假设偏离度，出自谁家的零假设、按谁家的误差模型换算，直接决定它能不能跨实验相加。

3.3 KARMEN 2002：「不确认」的原文措辞

英国 RAL 的 KARMEN 2（同为 μ⁺ 静止衰变束流）2002 年结果 [一手逐字]：In total, 15 candidates fulfill all conditions for the ν̄e signature, in agreement with the background expectation of 15.8±0.5 events, yielding no indication for oscillations. ... Thus, KARMEN does not confirm the LSND experiment and restricts significantly its favored parameter region for ν̄µ → ν̄e .（arXiv:hep-ex/0203021，PRD 65, 112001 (2002)）

15 个候选 vs 预期背景 15.8±0.5——干净 null。但措辞是「does not confirm the LSND experiment」，不是「LSND 测错了」。两家的联合频率派分析把这句话定量 [一手逐字]：At a combined confidence level of 36 %, there is no area of oscillation parameters compatible with both experiments. For the complementary confidence of 1 − 0.36 = 64 %, there are two well defined regions of oscillation parameters (sin2 (2Θ),∆m2 ) compatible with both experiments. 正文总结：the two experiments are either incompatible or lead to just one single oscillation solution at ∆m2 < 1 eV2（arXiv:hep-ex/0203023，PRD 66, 013001 (2002)）。从 2002 年起，「appearance 信号 vs 其它实验 null」的结构性张力就已经在场——它只是后来换了几代实验。

3.4 MiniBooNE 五版：σ 的迭代与版本账

MiniBooNE（Fermilab 增强器束流，L/E 设计与 LSND 相似）是 LSND 的官方复核实验，共五版口径：

  • 2007 首版（>475 MeV null） [一手逐字]：With two largely independent analyses, we observe no significant excess of events above background for reconstructed neutrino energies above 475 MeV. The data are consistent with no oscillations within a two-neutrino appearance-only oscillation model.（arXiv:0704.1500，PRL 98, 231801 (2007)）——注意限定词「above 475 MeV」与「two-neutrino appearance-only」：这次「判死」（见第八章消费账）从一开始就带着能量窗。
  • 2010 反中微子模式：475<E<1250 MeV 超出 20.9±14.0 事件，has a probability for consistency with the background-only hypothesis of 0.5%（arXiv:1007.1150，PRL 105, 181801 (2010)，[一手逐字]）。调研 brief 给的 arXiv:1006.5685 经原件核实是凝聚态论文——错 ID 已在捆笔记存证，以 1007.1150 为准。
  • 2012 ν+ν̄ 联合（arXiv-only，未期刊发表） [一手逐字]：A total excess of 240.3 ± 34.5 ± 52.6 events (3.8σ) is observed from combining the two data sets in the energy range 200 < EνQE < 1250 MeV. ... the background-only fit has a χ2 -probability of 0.03% relative to the best oscillation fit.（arXiv:1207.4809，INSPIRE 无期刊记录；曾有人以为它占了 PRD 86, 052001，经 Crossref 核实该卷页属 COUPP 暗物质论文）——MiniBooNE 自己的 3.8σ 首次出现于这份未发表预印本。
  • 2018 关键版（PRL 121, 221801）——版本账，两版均全文在手：v1（2018-05-30）摘要作 460.5 ± 95.8 events (4.8σ)＋the significance of the combined LSND and MiniBooNE excesses is 6.1σ；v2（2018-10-26，即 PRL 正式版）改作 [一手逐字]：a total νe plus ν̄e charged-current quasielastic event excess of 460.5 ± 99.0 events (4.7σ) is observed. ... the significance of the combined LSND and MiniBooNE excesses is 6.0σ.（arXiv:1805.12028）正文写明算法：The significance of the combined LSND (3.8σ) [1] and MiniBooNE (4.7σ) excesses is 6.0σ, which is obtained by adding the significances in quadrature——「联合 σ」是两个单实验内部 σ 的平方和开方，而 LSND 的 3.8σ 正是前文所述的转引数字。BBC 2018 报道的「exceeds six sigma」用的还是 v1 口径（6.1σ），PRL 正式版口径是 6.0σ。
  • 2020 终版（PRD 103, 052002） [一手逐字]：The MiniBooNE experiment at Fermilab reports a total excess of 638.0±52.1(stat.)±122.2(syst.) electronlike events from a data sample corresponding to 18.75 × 1020 protons-on-target in neutrino mode ... The overall significance of the excess, 4.8σ, is limited by systematic uncertainties, assumed to be Gaussian, as the statistical significance of the excess is 12.2σ.（arXiv:2006.16883）正文又回到 v1 口径：The significance of the combined LSND (3.8σ) [1] and MiniBooNE (4.8σ) excesses is 6.1σ——同一家合作组，PRL 版用 6.0σ，自己的终版用 6.1σ，两个数字都在流通。终版同时排除了一条替代路线：光子模型被 disfavors models that explain the event excess due to entering or exiting photons 拒绝（摘要末句；arXiv 元数据摘要措辞不同、含义同向）。

异常跳的现场定格：4.8σ 是 MiniBooNE 相对「纯背景」假设的偏离度；6.0/6.1σ 是两个这种偏离度的平方和。它们度量的是「数据不像背景」，不是「存在第四种中微子」——后者还需要 disappearance 家族和全球拟合投票，而那两票是反对的（见第六章）。

3.5 MicroBooNE：逐通道否决到双束流收口

MicroBooNE（液氩 TPC，同一增强器束流下游）用五年时间把「MiniBooNE 过量是什么」拆成通道逐一检验：

  • 2021 四篇 eLEE（低能电子样过量）同日投出：
  • 包含道（arXiv:2110.13978，PRD 105, 112005）[一手逐字]：we select 56 fully contained νe CC candidates while expecting 69.6 ± 8.0 (stat.) ± 5.0 (sys.) ... candidates after constraints for the absence (eLEEx=0 ) ... Under a simple-vs-simple hypotheses test, the eLEEx=1 hypothesis is rejected at 3.75σ, while the eLEEx=0 hypothesis is shown to be consistent with the observation at 0.45σ. In the context of the eLEE model, the estimated 68.3% confidence interval of the νe CC hypothesis to explain the LEE observed in the MiniBooNE experiment is disfavored at a significance level of more than 2.6σ (3.0σ) considering MiniBooNE's full (statistical) uncertainties.
  • 多末态拓扑（arXiv:2110.14054，PRL 128, 241801）三条独立寻找 are found to be either consistent with or modestly lower than the nominal νe rate expectations ... and no excess of νe events is observed（v3 即 PRL 版措辞）。
  • 无 π 道（arXiv:2110.14065，PRD 105, 112004）：合并 p 值 0.098；两通道各自 >90% CL disfavor 低能超出模型；最灵敏 1eNp0π 道观测低于预期；正文 The eLEE signal model is disfavored by the 1eNp0π channel at the 97.9% CL.
  • CCQE 深度学习（arXiv:2110.14080，PRD 105, 112003）：25 个 νe 候选 vs 预期 29.0±1.9±5.4；数据排除 0.25（0.38）倍 MiniBooNE LEE 中位信号强度 @90%（2σ）CL。
  • 口径提醒：「MicroBooNE 排除了 MiniBooNE 的 3+1 最佳拟合点」这种通行转述在 2021 原件中不存在；四篇全部以 eLEE 信号强度/假设检验口径表述。
  • 2021 Δ→Nγ 测量（光子解释线的定量检验，PRL 128, 111801） [一手逐字]：The data lead to a bound on an anomalous enhancement of the normalization of NC ∆ radiative decay of less than 2.3 times the predicted nominal rate for this process at the 90% confidence level (CL). The measurement disfavors a candidate photon interpretation of the MiniBooNE low-energy excess as a factor of 3.18 times the nominal NC ∆ radiative decay rate at the 94.8% CL（arXiv:2110.00409）。
  • 2022 3+1 联合（PRL 130, 011801） [一手逐字]：We observe no evidence of light sterile neutrino oscillations and derive exclusion contours at the 95% confidence level in the plane of the mass-squared splitting ∆m241 and the sterile neutrino mixing angles θµe and θee , excluding part of the parameter space allowed by experimental anomalies.（arXiv:2210.10216）
  • 2024 五年全数据（PRL 135, 081802，2025 见刊） [一手逐字]：the first measurement to use data from all five years ... 1.11 × 1021 protons on target ... p-values ≥ 26.7% when the two νe samples are combined ... This measurement excludes an electron-like interpretation of the MiniBooNE excess based on these models at > 99% CLs in all kinematic variables.（arXiv:2412.14407）
  • 2025 双束流（Nature 648, 64）——appearance 账的现行收口 [一手逐字]：Here, we use data obtained from the MicroBooNE liquid-argon time projection chamber [10] in two accelerator neutrino beams to exclude the single light sterile neutrino interpretation of the LSND and MiniBooNE anomalies at the 95% confidence level (CL). Additionally, we rule out a significant portion of the parameter space that could explain the gallium anomaly [6–8]. ... We find no evidence for either νµ → νe flavour transitions or νe disappearance that would indicate non-standard flavour oscillations.（arXiv:2512.07159，卷页经 INSPIRE 核实）正文零假设检验 p 值：a p-value of 0.96 evaluated using the Feldman-Cousins procedure；The region allowed at 99% CL by the LSND measurement and the vast majority of the region allowed at the 95% CL by the MiniBooNE experiment are excluded.

3.6 旁证与替代解释登记

  • ICARUS LNGS 2013 两件（CNGS 高能束流，730 km）：早期结果 Two clear νe events have been found, compared with the expectation of 3.7 ± 0.6 events from conventional sources. ... this result is compatible with the absence of a LSND anomaly.（arXiv:1209.0122，EPJC 73, 2345）；加倍样本更新确认（arXiv:1307.4699，EPJC 73, 2599）。注意其原文同时写明 LSND 窗口被压缩而非关死：strongly limits the window of open options for the LSND anomaly to a narrow region around ∆m2 , sin2 (2θ) new = (0.5 eV2 , 0.005), where there is an overall agreement (90% CL) between the present ICARUS limit, the published limits of KARMEN and the published positive signals of LSND and MiniBooNE Collaborations.
  • ICARUS 2026 BNB 首篇振荡分析（arXiv:2603.22557）：νμ disappearance 道，we find no statistically significant muon neutrino disappearance at the ICARUS baseline of 600 meters from the BNB target，自陈 systematics limited。调研 brief 记忆中的「ICARUS 2024 SBN 首结果论文」经 INSPIRE 三路＋OpenAlex 检索查无此文（仅有 Neutrino 2024 会议录）——诚实空位见第十章，记账以原件为准。
  • SBND 时间线（仅登记）：2024-07 开始运行、2024-12 稳定取数，∼7, 000 neutrino events per day（arXiv:2504.00245）；FNAL 新闻稿 2024-09-10 首中微子（news.fnal.gov）。
  • 替代解释登记（不审，登记其存在与命运）：Gninenko 2009 重中微子辐射衰变模型（arXiv:0902.3802，PRL 103, 241802）；MiniBooNE 2020 终版「进入/逸出光子」模型 disfavored；MicroBooNE Δ→Nγ 把最强光子解释压到 94.8% CL——振荡解释与替代解释是两条线，否决后者不等于证实前者。

3.7 本章账

appearance 家族三十年：LSND 过量（真实记录，原文无 σ）→KARMEN「不确认」（不是「错杀」）→MiniBooNE 过量复现（4.8σ，单实验内部偏离）→MicroBooNE 逐通道否决过量来源→双束流 95% CL 排除「single light sterile neutrino interpretation」。LSND 与 MiniBooNE 的过量测量本身至今无人逐字解释——MicroBooNE 排除的是「单一惰性中微子解释」这个模型，不是那 638 个电子样事件。这是第七章守零跳的伏笔。

四、换尺跳账：反应堆异常生于分母、死于分母、复活于分母

反应堆反中微子异常（RAA）是换尺跳的教科书现场：它的「异常」定义在「观测率 / 预言率」的比值上，而那个预言分母是理论汇编不是测量。分母换三次，异常生灭三次；反应堆实验的数据一次没动过。

4.1 2011：分母换尺，异常诞生

  • Mueller et al. 2011（PRC 83, 054615）重分析 1980 年代 ILL β 谱，新谱归一化上移 [一手逐字]：the normalization is shifted by about +3% on average（arXiv:1101.2663）。
  • Huber 2011（PRC 84, 024617）独立转换确认 [一手逐字]：a net 3% upward shift in energy averaged anti-neutrino fluxes；正文定量：The upward shift in fluxes is significant at the (3.7 − 4.0) σ level. 且 Huber 自己加了警告：若系统误差取大，the new found support for sterile neutrinos ... would be weakened considerably（arXiv:1106.0687）。
  • Mention et al. 2011（PRD 83, 073006）——命名与定标 [一手逐字]：The synthesis of published experiments at reactor-detector distances <100 m leads to a ratio of observed event rate to predicted rate of 0.976(0.024). With our new flux evaluation, this ratio shifts to 0.943(0.023), leading to a deviation from unity at 98.6% C.L. which we call the reactor antineutrino anomaly.（arXiv:1101.2755）

异常的定义动作在最后一行：同一个 19 个短基线实验的综合观测率没动，分母换上新通量后，比值从 0.976±0.024 移到 0.943±0.023——5.7% 的平均亏缺、98.6% C.L.（正文另一口径 −2.2σ）的「异常」就此被命名。同文还把反应堆＋镓＋MiniBooNE 联合，给出 disfavors the no-oscillation hypothesis at 99.8% C.L.——「第四种中微子」的入场券里有它一份。

4.2 实测对照：数据指着同一把有问题的尺

  • Daya Bay 2016/2017（CPC 41, 013002） [一手逐字]：the ratio of measured to predicted flux was found to be 0.946±0.020 (0.992±0.021) for the Huber+Mueller (ILL+Vogel) model；正文：there is a 5% normalization deficit when compared with the Huber+Mueller model prediction（arXiv:1607.05378）。
  • Daya Bay 燃料演化 2017（PRL 118, 251801）：A 7.8% discrepancy between the observed and predicted 235U yield suggests that this isotope may be the primary contributor to the reactor antineutrino anomaly.（arXiv:1704.01082，摘要逐字）——矛头从「振荡」转向「²³⁵U 谱算错了」。
  • RENO 2019（PRL 122, 232501）同向：Reevaluation of the 235U IBD yield per fission may mostly solve the Reactor Antineutrino Anomaly (RAA)（arXiv:1806.00574，摘要逐字）。
  • STEREO 2020 产额（PRL 125, 201801）：亏缺 (5.2 ± 0.8 [stat] ± 2.3 [sys] ± 2.3 [model])% 与世界平均一致（arXiv:2004.04075，摘要逐字）。

4.3 短基线直接检验：振荡解释被逐台排除

  • NEOS 2017（PRL 118, 121802）：It is found to be consistent with no oscillation. ... the optimum point for the previously reported reactor antineutrino anomaly are excluded with a confidence level higher than 90%.（arXiv:1610.05134，摘要逐字）
  • DANSS 2018（PLB 787, 56）：The excluded area covers a wide range of the sterile neutrino parameters up to sin²2θ14<0.01 in the most sensitive region.（arXiv:1804.04046，摘要逐字）
  • STEREO 三连：2018（66 天）the best fit of the reactor antineutrino anomaly is excluded at 97.5% C.L.（arXiv:1806.02096）→ 2020（179 天）The best-fit point of the reactor antineutrino anomaly is rejected at more than 99.9% C.L.（arXiv:1912.06582）→ 2023 终版（Nature 613, 257） [一手逐字]：Our analysis excludes the parameter space favored by the RAA up to a few eV2 at 95% CL or greater；图 2 说明：The initial RAA best fit point, marked by a star, is rejected with very high confidence level (p-value < 10−4).；零假设检验：our data are compatible with the no oscillation hypothesis H(0, 0) with a p-value of 0.52.；终版给出的病因诊断：Such a direct measurement of the antineutrino energy spectrum suggests instead that biases in the nuclear experimental data used for the predictions are at the origin of the anomalies.（arXiv:2210.07664）
  • PROSPECT：2018 disfavors the best fit of the Reactor Antineutrino Anomaly at 2.2σ confidence level（arXiv:1806.02784）→ 2020 disfavor the Reactor Antineutrino Anomaly best-fit point at the 2.5σ confidence level（arXiv:2006.11210）。

4.4 2021–2022：分母再换尺，「demise」宣告

  • Kopeikin et al. 2021（PRD 104, L071301）在库尔恰托夫研究所重测 ²³⁵U/²³⁹Pu 累积 β 谱比值 [一手逐字]：A discrepancy with the β spectra measured at Institut Laue-Langevin (ILL) was observed, indicating a steady excess of the ILL ratio by the factor of 1.054±0.002. ... We conclude that the new predictions are consistent with the results of Daya Bay and STEREO experiments.（arXiv:2103.01684）——1980 年代的 ILL 基准测量本身被指有 5.4% 系统偏高，HM 分母的源头被撼动。
  • Giunti et al. 2022（PLB 829, 137054）——「demise」宣告 [一手逐字]：We show that the reactor rates and the fuel evolution data are consistent with the predictions of the Kurchatov Institute (KI) conversion model and with those of the Estienne-Fallot (EF) summation model, leading to a plausible robust demise of the reactor antineutrino anomaly. 正文数值：EF 口径 differs from unity by only 1.2 σ、KI 口径 only 1.1 σ、HM 口径 2.5σ（arXiv:2110.06820）。
  • Letourneau et al. 2023（PRL 130, 021801）求和法独立复现异常形状，病因同指 ILL β 谱：suggesting that biases on the reference fission-electron spectra could be at the origin of the anomalies.（arXiv:2205.14954）

4.5 2023–2026：同一批数据，四把分母尺，两个世界

  • Perissé et al. 2023（arXiv:2304.14992）首个带完整误差预算的求和模型（即下文「CEA 模型」）：与最新 IBD 测量相符，但 unable to properly describe the reference aggregate β spectra measured at the Institut Laue-Langevin High-Flux reactor in the 80s，preferentially points toward a misprediction of the 235U ν̄e spectrum（摘要逐字）。
  • Daya Bay 2025 终版（PRL 134, 201802，470 万 IBD 全数据集） [一手逐字]：the Daya Bay flux shows 5.0% deficit with respect to the HM model, while it is consistent with the KI and SM2023 models.；The Daya Bay spectrum, however, disagrees with all model predictions.；²³⁵U 通量对 HM 亏缺 7.0%（约 3σ），但对 KI 仅低 1.8%；Daya Bay has no strong preference for 235U as the sole offending isotope in the HM prediction.（arXiv:2501.00746）——2017 年「²³⁵U 是主犯」的指认在全数据集口径下后退。
  • Revival 2026（Giunti, Li, Zhang，arXiv:2605.10353v2）——「复活」 [一手逐字]：This anomaly was first identified in 2011 (2.5σ) as a consequence of the Huber-Muller reactor antineutrino flux calculation. It was thought to be resolved in 2021 ... In this work, we present the RAA obtained with the latest reactor antineutrino flux calculation published in 2023 by a French research group ... The result indicates a revival of the RAA at the level of 2.2σ. 表 I 四把尺并排：HM 2.6σ／EF 1.2σ／KI 0.6σ／CEA 2.2σ（arXiv:2605.10353）。同一批作者 2021 年用 KI/EF 尺宣布 demise，2026 年换 CEA 尺宣布 revival——变的只有分母。
  • ROPP 2026 综述（Reports on Progress in Physics，2026-07-28 刊）定调 [一手逐字]：more advanced models—including the Kurchatov Institute (KI) model ... and the Estienne-Fallot (EF) model ... can largely account for the RAA, potentially resolving the discrepancy within SM uncertainties.（arXiv:2506.15306）

截至证据截止日（2026-08-12），「复活 2.2σ」与「大体消解」两种定调并存于最新一手文献，分歧点在误差预算与相关性处理，不在任何测量数据。

4.6 5 MeV bump：与归一化异常是两件事，至今未解

别把 bump 和 RAA 混为一谈——权威分列（Letourneau 2023，[一手逐字]）：They refer to an observed ∼6% deficit in the detected rate, known as "Reactor Antineutrino Anomaly" (RAA), and a ∼10% excess of events in the 4-6 MeV range, known as the "5-MeV bump"。证据链：Daya Bay 2016 an excess of events in the region of 4-6 MeV was found in the measured spectrum, with a local significance of 4.4 σ（同文同时给 5% 归一化亏缺，两量分列）；Daya Bay 2025 终版 a clear excess around 5 MeV with respect to the HM model at more than 5σ significance and the SM2023 model at 3σ significance；ROPP 2026 两连：Theoretical flux models that significantly alleviate the RAA paradoxically exacerbate the bump anomaly.＋STEREO [14] and PROSPECT [16], have excluded the no-bump scenario with significances of 3.5σ and 2σ, respectively.——消解 RAA 的模型反而加重 bump；「bump 存在」本身被 STEREO/PROSPECT 以排除「无 bump」场景确认。这是一条真实的、所有现行模型都解释不了的测量异常，但它不是「第四种中微子」的证据——它恰恰是「分母（理论谱）有问题」这条诊断线上最硬的一颗钉子。

4.7 本章账

RAA 生灭全程：测量数据零改动，分母换尺三次。2011 年生于 HM 换尺（0.976→0.943）；2021 年「死」于 KI/EF 换尺（1.1–1.2σ）；2026 年「复活」于 CEA 换尺（2.2σ）——同一批作者。短基线实验排除的是「eV 级惰性中微子振荡」这个解释通道；分母侧的病因诊断（ILL β 谱偏置）有 Kopeikin 重测与 Letourneau 求和法互证；而 5 MeV bump 作为独立的形状异常至今无解。换尺跳在这里拿到全貌：「异常有/无」这个问题的答案，随分母的选择在 0.6σ 到 2.6σ 之间滑动，且滑动方向由理论家手里的尺决定。

五、镓异常账：两轮「解释路线被关闭」后仍然活着的那条异常

镓异常是守零跳的最大反例：它用人工中微源（⁵¹Cr、³⁷Ar）照射镓靶做探测器标定，测到的 ⁷¹Ge 产率系统性低于预言——三十年里每一次「测量错了」或「解释死了」的宣判，都被下一份原件顶回来。

5.1 历史源实验：四次照射的账本

  • GALLEX 双 ⁵¹Cr 源终篇（PLB 420, 114-126 (1998)） [一手逐字，Wayback 取回 CDS 扫描件并目视复核]：The ratio, R, of the the neutrino source strength derived from the measured rate of 71Ge production, divided by the directly determined source strength is R = 1.01 +0.12 −0.11 for the first source and R = 0.84 +0.12 −0.11 for the second one. The combined value of R for the two source experiments is R = 0.93 ± 0.08. It shows that the > 40 % deficit of solar neutrino flux observed by GALLEX cannot be attributed to experimental artifacts（INSPIRE 记录；DOI 10.1016/S0370-2693%2897%2901562-1；”of the the” 系印刷原文）。调研 brief 记的「PLB 447, 127」已证伪（447/127 是 GALLEX IV 太阳结果篇）。
  • SAGE ⁵¹Cr 首报（PRL 77, 4708 (1996)）：The ratio of the measured production rate to that predicted from the source activity is 0.95±0.11(stat)+0.05/−0.08(syst). This agreement verifies that the experimental efficiency is measured correctly, establishes that there are no unknown systematic errors at the 10% level（INSPIRE＋APS 摘要页双源，摘要逐字；全文订阅墙）。
  • SAGE ⁵¹Cr 全文（PRC 59, 2246 (1999)）——换尺跳的天然标本 [一手逐字]：the cross section ... was measured to be [ 5.55 ± 0.60 ( stat ) ± 0.32 ( syst ) ] × 10 − 45 cm 2 . The ratio of this cross section to the theoretical cross section of Bahcall for this reaction is 0.95 ±0.12 ... and to the cross section of Haxton is 0.87±0.11（arXiv:hep-ph/9803418）。同一组数据、两把理论尺、0.95 与 0.87 两个世界——镓异常从第一天起就带着分母依赖。
  • SAGE ³⁷Ar 源（PRC 73, 045805 (2006)） [一手逐字]：The measured production rate was 11.0−0.9 +1.0 (stat) ± 0.6 (syst) atoms of 71 Ge/d, which is 0.79−0.10 +0.09 of the theoretically calculated production rate. When all neutrino source experiments with gallium are considered together, there is an indication the theoretical cross section has been overestimated（arXiv:nucl-ex/0512041）。同文 Table I 四次源实验齐全：GALLEX Cr1 1.00、Cr2 0.81、SAGE Cr 0.95、SAGE Ar 0.79（GALLEX 两值与 PLB420 摘要的 1.01/0.84 微差，系 Kaether 2010 复核下修所致，Gavrin 2024 综述有逐字说明）。
  • Kaether et al. 2010（PLB 685, 47） [一手逐字]：GALLEX 用新刻度数据复核，73.4+7.1 ... SNU. This is 5.3% below the old value of 77.5——实验方自己重算分母，异常值同向下调（arXiv:1001.2731）。

5.2 截面换尺与 BEST 双件：异常跳到 4σ

  • Kostensalo et al. 2019（PLB 795, 542-547） [一手逐字]：新壳模型截面把 the significance from 3.0σ to 2.3σ（arXiv:1906.10980）——又一次分母换尺、异常缩水，但没消失。
  • BEST 2022 PRL 128, 232501 [一手逐字]：3.414 MCi ⁵¹Cr 源置双嵌套镓靶中心，R in = 0.79 ± 0.05 and R out = 0.77 ± 0.05 . The ratio of the outer to the inner result is 0.97 ± 0.07 , which is consistent with unity within uncertainty. The rates at each distance were found to be similar, but 20%–24% lower than expected, thus reaffirming the anomaly. These results are consistent with ν e → ν s oscillations with a relatively large Δ m 2 ( > 0.5 eV 2 ) and mixing sin 2 2 θ ( ≈ 0.4 ).（arXiv:2109.11482）
  • BEST 2022 PRC 105, 065502（姊妹篇·细节篇） [一手逐字]：Comparable deficits in the measured ratios relative to predicted values are found for both zones, with the 4 σ deviations from unity consistent with the previously reported gallium anomaly. If interpreted in the context of neutrino oscillations, the deficits give best-fit oscillation parameters of Δ m 2 = 3 . 3 − 2.3 + ∞ eV 2 and sin 2 2 θ = 0 . 42 − 0.17 + 0.15 , consistent with ν e → ν s oscillations governed by a surprisingly large mixing angle.（arXiv:2201.07364；期刊为 PRC 非 PRD，brief 线索已勘正）——「surprisingly large mixing angle」是作者自己写的：拟合参数大到反常，本身就是解释吃力的信号。
  • 另记：调研 brief 的「BEST 2023 双区实验论文（arXiv:2305.01826）」经三路检索证伪（该 ID 是量子计算论文；BEST 的两区数据即 2022 双件）；2023 年承重产出是 Elliott/Gavrin/Haxton 截面篇（PRC 108, 035502）：四次老标定合并 R = 0.866 ± 0.054、BEST 两区 R = 0.791 ± 0.05 and 0.766 ± 0.05（arXiv:2303.13623，[一手逐字]）。

5.3 联合约束：3+1 解释这一格被压死

  • Giunti et al. 2022（JHEP 10 (2022) 164） [一手逐字]：In the combined scenario with all available data, the parameter goodness of fit is below 0.042%, corresponding to a severe tension of 4–5σ, or stronger. Therefore, we conclude that one should pursue other possible solutions beyond short-baseline oscil- lations for the Gallium Anomaly.（arXiv:2209.00916）
  • Berryman et al. 2022（JHEP 02 (2022) 055）：Monte Carlo 替代 Wilks 定理普遍砍掉约 1σ 后，gallium data (dominated by the BEST result) show more than 5 σ of evidence supporting the sterile-neutrino hypothesis ... This explanation is, however, in significant tension (∼ 3 σ) with solar neutrino experiments.（arXiv:2111.12530，摘要逐字）
  • Gonzalez-Garcia, Maltoni, Pinheiro 2024（太阳模型无关） [一手逐字]：in most cases the compatibility between Gallium and solar+KamLAND data only occur at the 3σ level or higher；太阳＋KamLAND 的 2σ 界 sin2 θ ≲ 0.025–0.045 ... clearly disfavouring the sterile oscillation interpretation of the Gallium anomaly——BEST 需要 sin²2θ≈0.4，太阳侧允许值小了近一个数量级（arXiv:2411.16840）。同文两套合并口径：GALLEX+SAGE 0.88±0.05（2.4σ）、含 BEST 0.80±0.05（>4σ）。
  • Brdar, Gehrlein, Kopp 2023（JHEP 05 (2023) 143） [一手逐字]：现状 R = 0.803 ± 0.035；逐一排查截面、半衰期、未知激发态、提取效率后——at face value, none of these potential single points of failure can be responsible for the anomaly（arXiv:2303.05528）。

5.4 「测量错了」路线连关两道门：半衰期双排除

镓异常最像样的「测量侧逃生门」是 ⁷¹Ge 半衰期——若它比公认值长约 1 天，异常即消。这条路被两把独立高统计测量钉死：

  • Collar & Yoon 2023（PRC 108, L021602） [一手逐字]：A dedicated high-statistics measurement of the 71 Ge half-life is found to be in accurate agreement with an accepted value of 11.43±0.03 d, eliminating a recently proposed route to bypass the "gallium anomaly" ... Following the incorporation of this new information, the gallium anomaly survives with high statistical significance.（arXiv:2307.05353）
  • Norman et al. 2024（PRC 109, 055501） [一手逐字]：Our results yield a half-life of 11.468 ± 0.008 days ... the potential explanation of the Gallium Anomaly being due to an unexpectedly long 71 Ge half-life has been ruled out, leaving the anomaly's origin as an open question.（arXiv:2401.15286）

5.5 2024–2026 活口：结案证据为零

  • Gavrin et al. 2024 综述（PPNP 104082，账房口径） [一手逐字]：The gallium anomaly found in the four original calibration experiments represents about a 2.5σ deviation from expectations. With the completion of BEST, the deviation has risen to 6σ.（保守假设口径约 4σ）；But BEST found no evidence of distance dependence and thus no explicit indication of new physics.（arXiv:2306.03299）——异常升、新物理证据无，两句同文。
  • Cadeddu et al. 2025/2026 两件：自洽电子波函数重算截面，best fit of the data is found to be R = 0.881 ± 0.031, which corre- sponds to a 3.8σ anomaly（arXiv:2507.13103）；因子化之外再算（v3, 2026-06-17）the revised cross- section can be significantly reduced, potentially resolving the gallium anomaly without invoking new physics（arXiv:2512.20560）——最新「声称解决」件，尚待共同体复核，且同组 2026 年在 Nuovo Cim. C 49, 13 (2026) 的标题就叫《The Gallium Anomaly: An unsolvable puzzle?》（需亲核，仅元数据）。
  • Parini & Stefanovich 2026（核反冲振荡新机制，2026-06-29） [一手逐字]：反冲驱动振荡 makes a substantial contribution to the observed deficit，但自陈 nuclear recoil by itself cannot fully account for the observation（arXiv:2606.30758）。
  • 未来检验三路：BEST-2（⁵⁸Co 源，arXiv:2501.08127）；LIQUID（100 吨 ¹¹⁵In 靶，正式发表 PRD 113, 115058 (2026)，can probe the complete parameter space of the gallium anomaly）；νe 散射同源性检验（换探测器不换源，arXiv:2504.16590）。
  • KATRIN 2025 一锤的正确读法：PDG 2025 §14.8 在镓异常振荡解读的语境下逐字写明 This intepretation, however is in tension with the analysis of solar neutrino experiments [215], and most recently has been excluded with high significance by KATRIN [216]（intepretation 系原文拼写如此），CERN Courier 2026-01 定量 the best fit of the sterile neutrino from the gallium anomaly is excluded at 96.6% confidence——但 Huber 同文立刻补：That is not to say the gallium anomaly went away（CERN Courier）。排除的是「镓异常的单一惰性中微子最佳拟合点」，不是异常本身。

5.6 本章账

镓异常三十年：四次源照射＋BEST 两区，全部低于预言；统计显著性随截面换尺在 2.3σ–6σ 之间滑动；3+1 惰性解释被反应堆＋太阳＋KATRIN 联合压死（PG<0.042%、太阳 2σ 界差一个数量级、最佳拟合点 96.6% 排除）；「测量错了」路线在半衰期这一格被两把独立测量关闭；2026 年最新机制主张（核反冲）自陈「cannot fully account」；账房定调「unsolvable puzzle?」。这是「解释被排除≠异常解决≠测量错了」三重区分的活体标本——守零跳与反向红跳在这里双不立。

六、判据账：为什么单实验 σ 凑不成「第四种中微子成立」

异常家族的另一侧是 disappearance 家族的一排 null。两排数据放进同一个 3+1 全球拟合里互相否决——这是「第四种中微子」三十年立不起来的定量原因，判据工具有名字：Parameter Goodness-of-fit（PG）检验。

6.1 判据工具与 disappearance 家族

  • PG 检验（Maltoni & Schwetz 2003，PRD 68, 033020） [一手逐字]：We discuss a goodness-of-fit method which tests the compatibility between statistically independent data sets. ... it avoids the problem that a possible disagreement between data sets becomes diluted by data points which are insensitive to the crucial parameters.（arXiv:hep-ph/0304176）——它测的不是「拟合好不好」，而是「几组数据在同一个模型下能不能共存」。
  • NOMAD 2003（PLB 570, 19）：No evidence for oscillations was found. ... This result excludes the LSND allowed region of oscillation parameters with delta m^2 > 10 eV^2.（arXiv:hep-ex/0306037，摘要逐字）
  • OPERA 2013/2018：Data are compatible with the non-oscillation hypothesis（arXiv:1303.3953）；终版 sin^2 2θ_{μe} < 0.021 at 90% C.L.（arXiv:1803.11400，摘要逐字）
  • MINOS/MINOS+ 2019（PRL 122, 091803）：yields no evidence for sterile neutrino mixing using a 3+1 model. The most stringent limit to date is set on the mixing parameter sin^2θ_{24}（arXiv:1710.06488，摘要逐字）
  • MINOS+/Daya Bay/Bugey-3 2020（PRL 125, 071801）：excluding the 90% C.L. sterile-neutrino parameter space allowed by the LSND and MiniBooNE observations at 90% CL_s for Δm^2_{41}<5 eV^2（arXiv:2002.00301，摘要逐字）
  • IceCube 2020（PRL 125, 141801）：最拟点 sin²(2θ₂₄)=0.10、Δm²₄₁=4.5 eV²，which is consistent with the no sterile neutrino hypothesis with a p-value of 8.0%（arXiv:2005.12942，摘要逐字；调研 brief 的 2007.09514 经核实是椭圆曲线数学论文）——disappearance 端「看着有点、统计不立」的范本。IceCube 2024（PRL 133, 201804）：10.7 年数据，consistent with the null hypothesis of no sterile neutrinos with a p-value of 3.1%（arXiv:2405.08070）
  • DeepCore 2024：compatible with the absence of mixing ... |Uμ4|² < 0.0534 ... at 90% CL；These null results add to the growing tension between anomalous appearance results and constraints from disappearance searches（arXiv:2407.01314，摘要逐字）
  • NOvA 2024→PRL 134, 081804 (2025)：NOvA finds no evidence for active-to-sterile neutrino oscillations under the 3+1 model at 90% confidence level. ... excluding some regions currently allowed by IceCube at 90% confidence level.（arXiv:2409.04553，摘要逐字）
  • JSNS2 2026 首结果（直测 LSND 型）：two events are observed, consistent with the expected background of 2.3±0.4 events. ... the expected number of events due to LSND anomaly is 1.1±0.5, thus this result is not yet conclusive.（arXiv:2602.06274，摘要逐字；调研 brief 的「JSNS2 2021/2024 结果」经检索证伪不存在）

6.2 全球拟合三十年：张力定量账

  • Gariazzo/Giunti 2017：an unacceptable appearance-disappearance tension which becomes tolerable only in a pragmatic fit which neglects the MiniBooNE low-energy anomaly（arXiv:1703.00860，摘要逐字）——「不可接受」这个词 2017 年就出现在摘要里。
  • Dentler et al. 2018（JHEP 08 (2018) 010）——4.7σ 排除 [一手逐字，全文在手]：Under the sterile neutrino oscillation hypothesis, the p-value for those data sets being consistent is less than 2.6×10−6. Therefore, an explanation of the LSND anomaly in terms of sterile neutrino oscillations in the 3+1 scenario is excluded at the 4.7σ level. This result is robust with respect to variations in the analysis and used data, in particular it depends neither on the theoretically predicted reactor neutrino fluxes, nor on constraints from any single experiment. 正文加固：去掉全部反应堆数据后 p 值仍是 3.8×10⁻⁵；张力由 LSND 主导。同文留出活口：Our results rule out the sterile neutrino oscillation hypothesis as an explanation of the LSND and MiniBooNE anomalies, but it remains a viable option for the reactor and gallium anomalies.（arXiv:1803.10661）——2018 年，LSND 的 3+1 解释在全球拟合层面已被 4.7σ 排除；同年 MiniBooNE 发表 4.8σ 单实验结果。两个「4.x σ」方向相反，这正是判据账的核心一幕。
  • Diaz et al. 2019（Phys. Rept. 884 (2020) 1-59）——「5σ 偏好」与「4.5σ 内伤」同文 [一手逐字]：we find an improvement of Δχ2=35 for 3 additional parameters compared to a model with no sterile neutrino. This is a 5σ improvement, indicating that an effect that is like that of a sterile neutrino is highly preferred by the data. However we note that separate fits to the appearance and disappearance oscillation data sets within a 3+1 model do not show the expected overlapping allowed regions in parameter space. 正文定量：Our conclusion on the 3+1 model is that, although the 3+1 model is favored over the null model at about 5σ, there is a clear problem of internal consistency at the 4.5σ level.（arXiv:1906.00045）同文方法学自曝（判据学直接可用，原文语法如此）：Lastly, we point out that analyses presenting limits often receive less scrutiny that signals.——给出限值的分析常常不如给出信号的分析受审视，这句写在领域旗舰综述里。
  • Böser et al. 2020（PPNP 111, 103736）：as yet are not able to provide a definite conclusion on the existence of eV-scale sterile neutrinos（arXiv:1906.01739，摘要逐字）
  • Hardin et al. 2022：In the commonly-used 3+1 plane wave model, there is a well-known 4.9σ tension between data sets sensitive to appearance versus disappearance of neutrinos.（arXiv:2211.02610，摘要逐字）

6.3 共同体定调：Snowmass 2022 白皮书（头号承重件）

《Light Sterile Neutrino Searches and Related Phenomenology》（Snowmass 2021 NF02 专题组白皮书，J. Phys. G 51, 120501 (2024)，201 页全文在手）：

  • Executive Summary [一手逐字]：Historically, these anomalies were first interpreted as oscillations due to the existence of light sterile neutrinos ... However, this purely oscillatory interpretation is disfavored by several other direct and indirect experimental tests.
  • 全球拟合节转引定量：this tension is found to yield a p-value of 3.7×10−7 ... Moreover, removing any individual null experiment from the fit does not lead to significant improvements in the p-value, evidencing that the tension is robust.
  • §8 结论 [一手逐字]：The experimental data collected over the past decade are undeniably challenging the simplest theoretical interpretation of experimental anomalies in neutrino physics—namely, that of a light sterile neutrino within the context of a 3+1 model—as a single underlying source of the outstanding short-baseline anomalies. 以及同一节：experimental anomalies in the radioactive source, reactor, pion decay-at-rest, and pion decay-in-flight neutrino sectors still persist at the level of 1-5σ, and therefore remain as outstanding questions.（arXiv:2203.07323）

同一份官方文件，「解释受挑战」与「异常未决」并立——这就是第七章守零跳的制度底本：共同体自己的口径从来是双轨的，单轨的「结案」只出现在消费层。

6.4 2025–2026 现行数值

  • MicroBooNE 2025 双束流：LSND 99% CL 允许区＋MiniBooNE 95% CL 允许区绝大部分被排除，零假设 p=0.96（见 3.5，arXiv:2512.07159）。
  • Woodward et al. 2026-03（SBI 全球拟合，最新定量件）：despite an overall > 5σ improvement over the 3ν Standard Model (SM) based on a χ2 fit，these data sets favor 3+1 at 3.6σ and 1.8σ respectively, while the tension between the two is 3.3σ ... After correcting for normalization differences ... the tension relaxes to 2.2σ, indicating reduced but non-negligible disagreement.（arXiv:2603.15322，摘要逐字）——截至 2026-03，「整体 >5σ 偏好 3+1」与「内部张力犹在」依旧同文并存。
  • Giunti/Li/Zhang 2026-05：镓 vs 其余数据 3.8σ 张力；放大镓误差后降至 1.3σ（arXiv:2605.10353，见 4.5）。
  • 方法学新线（登记）：SBI 三部曲（2501.08988／2507.01153／2512.05784）把全球拟合推进到「不假设 Wilks 定理」；Rodrigues et al. 2025（arXiv:2503.13594）论证 MicroBooNE 单实验已覆盖 MiniBooNE 2σ 偏好区于 2–3σ，SBND＋ICARUS 可推过 4σ。

6.5 本章账

判据账的核心一句话：σ 不跨实验转移。LSND/MiniBooNE 的 4.8σ 度量「数据不像背景」；PG 检验度量「appearance 与 disappearance 能不能共存在一个 3+1 模型里」——答案是 2017「unacceptable」、2018 p<2.6×10⁻⁶（4.7σ 排除 LSND 的 3+1 解释）、2019「4.5σ 内伤」、2022「well-known 4.9σ」、2026「3.3σ→2.2σ 仍在」。媒体与科普层把单实验 σ 当「发现度」用，是异常跳的消费端放大器（第八章）；而三套统计语言——σ（单实验内部偏离）、CL/CLs（排除置信度）、PG p 值（全球兼容性）——在原件里各有各的定义域，互不换算。

七、守零跳账：「被排除」的限定词账——排除的是解释，不是测量

守零跳的动作：把「某一解释通道在某置信度被排除」读成「第四种中微子不存在、异常全部解决、领域结案」。本篇把每次「排除」的限定词逐字摊开。

7.1 MicroBooNE 2025 的三重限定词

Nature 648, 64 的原文（见 3.5）排除的是 the single light sterile neutrino interpretation of the LSND and MiniBooNE anomalies——single（一个，不是 3+2 或更多）＋light（eV 级轻，不碰 see-saw 重区）＋interpretation（解释，不是异常测量本身）。维基现行版对这个限定结构转述得意外准确：effectively excludes the entire single light sterile neutrino model parameter space to a 95% confidence interval. Even though the single light sterile neutrino model cannot explain the LSND and MiniBooNE anomalies, there are still several other sterile neutrino models that have not been ruled out, such as multiple light sterile neutrinos and sterile neutrinos with non-standard interactions.（en.wikipedia.org/wiki/Sterile_neutrino，2026-08-11 抓取快照，维基＝消费层样本）

7.2 死亡宣告史：「已死」被宣布过多少次

逐字登记（全部一手存盘，详见第八章消费账）：

  • 2007 Fermilab 通稿：MiniBooNE researchers showed conclusively that the LSND results could not be due to simple neutrino oscillation——第一次「判死」（带着 >475 MeV 能量窗限定）。
  • 2018 Neutrino 海德堡会场，Maltoni 基于 disappearance＋宇宙学宣布（Quanta 2021 转引 Argüelles-Delgado 回忆逐字）：'If you didn't know that it was over, you should now know that it's over'——第二次「判死」；同一年 MiniBooNE 发表 4.8σ，媒体写「复活」。
  • 2021 Fermilab 通稿：deal a blow to a theoretical particle known as the sterile neutrino ... no sign of the sterile neutrino——第三次。
  • 2025-12 Nature N&V（Huber）标题《Still no sign of hypothetical sterile-neutrino particle》；2026-01 CERN Courier《Two strikes for the light sterile neutrino》；2026-04-08 Quanta《Experiments Ring the ‘Death Knell’ for Sterile Neutrinos》（Ross-Lonergan 引语逐字：'This is, in my opinion, the death knell for sterile neutrinos,'——注意引语自带 in my opinion）——第四次，现行结案口径成型。

7.3 同一时刻的原件双轨

「结案」口径只在「3+1 单一轻惰性中微子解释」这条通道成立；同一证据截止日的原件证明多轨并行：

  • Snowmass 白皮书（官方）：解释「disfavored」与异常「still persist at the level of 1-5σ」同文（见 6.3）。
  • ROPP 2026 综述：RAA「potentially resolving」与 bump「paradoxically exacerbate」同文（见 4.6）。
  • Revival 2026：RAA 换分母「复活 2.2σ」与镓 3.8σ 张力同文（见 4.5）。
  • Huber 在 CERN Courier 的双句：the best fit of the sterile neutrino from the gallium anomaly is excluded at 96.6% confidence ＋ That is not to say the gallium anomaly went away；以及 The MicroBooNE result is just confirming what we knew from global fits for a long time——领域内一号口径人物自己的表述就是「排除解释≠异常消失」。
  • KATRIN/MicroBooNE 2025 同日双论文（Nature 648）排除的是「single light sterile」参数空间；P5 2023 的制度口径（见 8.3）是「anomalies have not been confirmed … will be conclusively tested」——「异常未被确认」与「解释被排除」在制度文本里是两个独立命题。

7.4 本章账

守零跳的完整机制：限定词在传播中被吃掉。原件的「single light sterile neutrino interpretation … at the 95% confidence level」到标题层变成「Still no sign」「Two strikes」「Death Knell」；到中文层变成《最新实验结果未能支持验证惰性中微子存在》。每一步都有原文出处，每一步限定词都更少一级。而测量侧的三条硬事实——LSND/MiniBooNE 过量、镓亏缺、5 MeV bump——在全部「结案」文献里无一被逐字解释。结案的是解释，活着的是账。

八、消费账与制度账：三十年的「判死→复活→再判死」话术化石

8.1 媒体宣布史（逐字抽样，全部存盘）

  • 1995-01-31 NYT 头版（《Cosmos’s Missing Mass: Wispy Particle Weighs In》，Wayback 取回全文）——LSND 首报当日直接被读成质量证据 [一手逐字]：'this is a very big discovery.' ... 'It's the golden evidence for neutrino mass. It's the discovery of more matter in the universe than we've known up to now.'（Primack）；LSND 发言人 White 自己的口径其实克制得多：We're 95 percent sure of our conclusion. Oscillation seems to us to be the most likely explanation.（nytimes.com 原文，经 Wayback 快照 2021-10-28 取回）
  • 2001-12-04 LANL 通稿（EurekAlert 承载）：标题《LSND strengthens evidence for neutrino oscillations》；It has been hypothesized that some unknown new phenomenon--such as a fourth 'sterile' neutrino ... might explain the data.（eurekalert.org）
  • 2007-04-11 Fermilab 通稿（《MiniBooNE opens the box》，Wayback 取回）——判死轨 [一手逐字]：MiniBooNE researchers showed conclusively that the LSND results could not be due to simple neutrino oscillation；Louis：'We live in interesting times.'；同一份通稿已埋下低能伏笔：They include detailed investigation of data we observe at low energy that do not match what we expected to see...＋At this time, the source of the apparent low-energy discrepancy is unknown. 配套四件：CERN Courier《MiniBooNE solves neutrino mystery》；Physics Today 记录 'closed one door ... it may have opened another'＋Conrad 'It may be just a prosaic problem with our analysis,' says Conrad, 'or it could be a harbinger of important new physics.'；symmetry 杂志标题《No-show neutrino, and a new mystery》。
  • 2018-05/06 MiniBooNE 4.8σ——双轨样本：
  • 兴奋轨：Quanta《Evidence Found for a New Fundamental Particle》a finding that is most easily explained by the existence of a new elementary particle（quantamagazine.org）；BBC the significance exceeds six sigma（v1 口径）；SciAm the findings would be nothing short of revolutionary。
  • 怀疑轨（同日）：Science/AAAS（Adrian Cho）《Reports of sterile neutrino’s resurrection may be greatly exaggerated》Don't toss out your particle physics textbooks just yet. ... the situation is more ambiguous than some reports suggest.（science.org 原文，经 Wayback 取回）——兴奋件与怀疑件同一周并排存在，是 2018 年消费层的真实结构。
  • 2021-10-27 Fermilab 通稿（《MicroBooNE experiment’s first results show no hint of a sterile neutrino》） [一手逐字]：deal a blow to a theoretical particle known as the sterile neutrino ... no sign of the sterile neutrino；ruled out the most likely source of photons as the cause of MiniBooNE's excess events with 95% confidence and ruled out electrons as the sole source with greater than 99% confidence；Zeller 守住测量侧：'But that earlier data from MiniBooNE doesn't lie. There's something really interesting happening that we still need to explain.'（news.fnal.gov）
  • 2022-06-16 BEST：phys.org/LANL 通稿《Experiment results confirm anomaly, could point to new elementary particle》；Elliott：'This definitely reaffirms the anomaly we've seen in previous experiments. But what this means is not obvious. There are now conflicting results about sterile neutrinos.'；DOE 2024-01-08 科普页沿用 finds evidence of the sterile neutrino 口径（energy.gov）。
  • 2025-12→2026-04 结案口径成型：Nature N&V 2025-12-03（Huber）《Still no sign of hypothetical sterile-neutrino particle》（nature.com，经 Wayback 取回付费墙外部分）；CERN Courier 2026-01《Two strikes for the light sterile neutrino》；Quanta 2026-04-08《Experiments Ring the ‘Death Knell’ for Sterile Neutrinos》。

8.2 话术化石与缺席模板

标题级模板的演化（全部逐字存盘）：1995 情态证据级（「golden evidence」）→ 2007 宣判级（「solves neutrino mystery」）→ 2018 双轨（「Evidence Found」/「greatly exaggerated」）→ 2021 否决级（「no hint」）→ 2022 再暗示（「could point to new elementary particle」）→ 2026 宣判级（「Death Knell」）。缺席的模板同样记账：「smoking gun」在全部 30+ 存盘消费件中出现 0 次（python 计数）；最接近「超越标准模型的首个证据」模板的是 Weiner 2018 引语 the first discovered particle beyond the Standard Model。

8.3 制度账：P5 与 Snowmass 的原文

  • P5 2014《Building for Discovery》Recommendation 15 [一手逐字]：Select and perform in the short term a set of small-scale short-baseline experiments that can conclusively address experimental hints of physics beyond the three-neutrino paradigm.（P5 报告 PDF，p.12）——SBN 计划的制度出生证；Fermilab 2015 立项叙事《The short-baseline detectives and the mysterious case of the sterile neutrino》，Schmitz：'Either we will rule out the earlier hints, or we make, frankly, the most exciting discovery in particle physics in some time.'
  • P5 2023《Exploring the Quantum Universe》 [一手逐字]：These anomalies have not been confirmed, and the reactor antineutrino anomaly has been recently resolved. The remaining phase space will be conclusively tested by the current short-baseline neutrino program at Fermilab.（arXiv:2407.19176，§1）——制度层的现行口径：异常家族点名（LSND／RAA／MiniBooNE 低能过量／镓）、「未被确认」、RAA「已解决」（注意：此句写在 2023-12，早于 2026 年 Revival 的 2.2σ——制度的「已解决」与理论家的「复活」并存）。
  • Snowmass 2021 NF02：专题组正式名称《Understanding Experimental Neutrino Anomalies》——「异常家族」作为独立编组被制度承认（arXiv:2209.05352）。

8.4 中文消费层（四件，全部存盘）

  • 腾讯新闻转载「原理」（中科院物理所）2021-11-03《寻找第四种中微子》：新的结果与粒子物理学的标准模型一致，即只存在三种中微子，不多也不少，并没有惰性中微子存在的迹象。（news.qq.com）
  • 新华网转科技日报 2025-12-05《最新实验结果未能支持验证惰性中微子存在》：这一结果以95%的置信度排除了包含单一惰性中微子的理论模型，为此前的'异常'提供了重要检验。（news.cn）——注意这条保留了「单一惰性中微子」限定词，是中文层少见的准确转述。
  • 《物理》（中国物理学会）2018 编译 Kopp 评述《难以捉摸的惰性中微子》：因为这些信号可能是惰性中微子的信标，惰性中微子只通过引力产生相互作用，而且没有包含在标准模型中。；怀疑轨同文：但是并没有发现这种µ中微子的消失。（wuli.iphy.ac.cn，PDF 全文）
  • 领研网转环球科学 2022-06-24《BEST实验结果表明惰性中微子可能存在》：实验表明锗-71的产生速率比理论模型低了20～24%……但它也可能表明我们需要更新对中微子散射截面的解释。（linkresearcher.com，meta 层逐字）

8.5 消费层失真登记

英文维基 “Sterile neutrino” 词条（2026-08-11 快照）对 RAA 的表述：Two separate detectors near a nuclear reactor in France found 3% of anti-neutrinos missing. They suggested the existence of a fourth neutrino with a mass of 1.2 eV/c2.——与 Mention 2011 原件（5.7% 平均亏缺、|Δm²|>1.5 eV² @95%）不符，「3%」是通量上移幅度不是亏缺幅度；且英文维基无 RAA 独立词条（API 返回 missing）。登记备查，不作承重。

8.6 本章账

消费账的主结构是同一事件、双轨发表、限定词逐级脱落：1995「golden evidence」（LSND 首报被直接读成质量证据）→ 2007「solves mystery」（>475 MeV 限定消失）→ 2018「Evidence Found」与「greatly exaggerated」同周 → 2021「no hint」→ 2022「could point to」→ 2026「Death Knell」。制度账（P5 2014/2023、Snowmass NF02）反而全程保持「异常家族＝独立编组、未确认、待决定性检验」的中性口径——消费层的「判死/复活」震荡频率远高于制度层与原件层，这是本篇登记的第三种跳：语义在传播链上脱水。

九、反向红跳账：「解释死了」不等于「测量错了」

反向红跳是异常跳的镜像：把「3+1 全球拟合被否决／单惰性解释被排除」读成「LSND/MiniBooNE/BEST 测错了、异常不存在」。这条路线同样不立，原件四处钉死：

  1. KARMEN 的措辞是「不确认」不是「证伪」（见 3.3）：15 vs 15.8±0.5 的 null 与 LSND 的过量在 36% 联合置信水平内无共同参数区——两实验「要么不兼容、要么只剩 Δm²<1 eV² 单一解」。不兼容是两种实验之间的张力，不是任何一方「错了」的证据。
  2. MicroBooNE 的排除对象是解释模型：2021 四篇以 eLEE 信号强度口径否决「MiniBooNE 过量＝νe 或光子」的特定解释；2025 双束流排除「single light sterile neutrino interpretation」。Zeller（MicroBooNE 共同发言人）在官方通稿里明说：'But that earlier data from MiniBooNE doesn't lie.'（见 8.1）——做排除的实验方自己拒绝「MiniBooNE 测错了」的读法。
  3. 镓异常的「测量错了」路线被两把独立测量钉死：⁷¹Ge 半衰期逃生门 2023/2024 连关两道后，Collar 原文写 the gallium anomaly survives with high statistical significance，Norman 原文写 leaving the anomaly's origin as an open question（见 5.4）。
  4. 5 MeV bump 在所有模型下未解：消解 RAA 的 KI/SM2023 模型下 bump 仍在（3σ），「无 bump」场景被 STEREO/PROSPECT 排除（3.5σ/2σ）（见 4.6）——这条测量异常连「候选解释」都没有，谈不上「解释死了所以测量错了」。

平行生灭线登记（不审，仅登记分界）：keV 尺度惰性中微子作为暗物质候选是另一本账（代表件：Bulbul et al. 2014 星系团 3.55 keV 未认证线，An intriguing possibility is the decay of sterile neutrino, a long-sought dark matter particle candidate，arXiv:1402.2301，ApJ 789, 13）——归暗物质篇管；see-saw/HNL 重中性轻子对撞机寻找是又一本账（Abdullahi et al. 2022 Snowmass 白皮书《The Present and Future Status of Heavy Neutral Leptons》，arXiv:2203.08039，J. Phys. G 50, 020501）——质量起源侧归物质-反物质篇管。三条线共用「sterile neutrino」一词，质量尺度与证据结构完全不同，消费层常混为一谈，本篇按库内分界各归各位。

十、裁决层

10.1 三句话裁决

  1. 「第四种中微子（3+1，eV 级，单一轻惰性）作为 LSND/MiniBooNE 异常的解释」：成立不了——Dentler 2018 PG 检验 4.7σ 排除、MicroBooNE 2025 双束流 95% CL 排除、P5 2023「not been confirmed」；这是本篇证据最硬的一格裁决。
  2. 「四条异常测量本身」：全部活着——LSND/MiniBooNE 过量无人逐字解释、镓异常 3.8–6σ 悬而未决（「unsolvable puzzle?」）、RAA 随分母换尺在 0.6–2.6σ 间滑动（2026 年「复活」与「消解」并存）、5 MeV bump 所有模型下未解。「异常家族」是 Snowmass NF02 与 P5 2023 制度性承认的未决账，不是被媒体叙事创造出来的。
  3. 「第四种中微子这个名号本身」：死于限定词脱落——从「single light sterile neutrino interpretation excluded at 95% CL」到「Death Knell for Sterile Neutrinos」，三级传播各掉一级限定词；而 keV 暗物质线与 see-saw/HNL 线上的「sterile」是同名不同物，继续各自活着。

10.2 三套统计语言不混

本篇全部数字按三个定义域归档，互不换算：σ＝单实验相对其自身零假设的偏离度（LSND 涨落概率、MiniBooNE 4.8σ、BEST 4σ）；CL/CLs＝排除置信度（MicroBooNE 95% CL、KATRIN 96.6%、STEREO p<10⁻⁴）；PG p 值＝全球拟合中数据子集的兼容性（Dentler 2.6×10⁻⁶、白皮书 3.7×10⁻⁷、Woodward 3.3σ→2.2σ）。PDG Statistics 综述的原文界限适用于每一格：90/95% CL 排除零 ≠ 发现（见 2.2）。媒体层的一切误读，几乎都可以还原为把其中一套语言的数字读进另一套语言的语义。

10.3 诚实空位（本篇取证未闭合处）

  1. APS 摘要页被 Cloudflare 拦：全部期刊卷页经 arXiv citation_doi ＋ INSPIRE 双源核实，APS 页本身未亲核（存盘挑战页可为凭）。
  2. ICARUS 2024「SBN 首结果论文」查无此文：INSPIRE 三路＋OpenAlex 检索仅得 Neutrino 2024 会议录（其 PDF 被拦）；论文级首篇为 2026-03 νμ disappearance（3.6 节已按原件记账）。
  3. LSND「3.8σ」出处：LSND 2001 全文无 σ（python 全文检索实证）；「3.8σ」最早可核出处为 MiniBooNE 论文转引。本篇行文凡写「LSND 3.8σ」处均已注明系转引口径。
  4. Snowmass 白皮书 §8 的「ESM」缩写：pdftotext 逐字转录，未做 pdftoppm 目视复核；本篇引用 §8 时避开该词。
  5. SAGE 1996 PRL 77 全文订阅墙：仅摘要逐字（INSPIRE＋APS 页双源一致）。
  6. Berryman 2022、Kostensalo 2023 Corrigendum、Cadeddu 2026 Nuovo Cim. C 三件：摘要层/仅元数据，已逐处标注。
  7. Shibah 2026-07 Zenodo 单人存档：403 未取回，仅登记存在，不作承重。
  8. MicroBooNE 2025 Nature 卷页（648, 64）：经 INSPIRE 核实，Nature 页本身未亲核。
  9. 简报线索证伪清单（全部以原件为准）：arXiv:1006.5685→MiniBooNE 2010 实为 1007.1150；1309.7893→ICARUS 实为 1209.0122/1307.4699；「BEST 2023」2305.01826 不存在；2007.09514→IceCube 实为 2005.12942/2005.12943；「Daya Bay Nature 604（2003.11911）」→同位素谱实为 PRL 123,111801／128,081801／130,211801；GALLEX「PLB 447,127」→实为 PLB 420,114-126；SAGE/BEST 姊妹篇在 PRC 非 PRD；JSNS2 2021/2024 结果不存在（首结果 2602.06274）；MINOS+「2018」不存在（实体结果 1710.06488／2002.00301）。

10.4 方法学自限

本篇裁决建立在六捆 521 个存盘原始件（211 MB）之上：核心承重句全部一手逐字（全文 PDF 取回并 python 比对），摘要层件与需亲核件逐处标注；消费层（维基/媒体/通稿）全部标注其层级，不与学术一手混用。本篇不审：3.5 keV 线真伪、HNL 对撞机物理、0νββ、宇宙学 Σmν 约束（各归邻篇或超出范围）。本篇的裁决对象是「声称与解释的生灭」，不是「惰性中微子是否存在于某个质量尺度」——后者是物理问题，前者是证据学问题，本篇只审后者。

来源清单

全部链接已于写作时逐条核对（arXiv abs 页标题比对／Crossref DOI 标题比对），层级标注：【一手逐字】=全文取回；【摘要逐字】=仅摘要取回；【消费层】=媒体/维基/通稿样本。

守真锚（第二章）

appearance 账（第三章）

反应堆账（第四章）

镓账（第五章）

判据账（第六章）

消费层与制度账（第七、八章）

  • NYT 1995-01-31《Cosmos’s Missing Mass》【消费层，Wayback】：nytimes.com
  • LANL 通稿 2001-12-04（EurekAlert 717771）【消费层】：eurekalert.org
  • Fermilab 通稿 2007-04-11《MiniBooNE opens the box》【消费层，Wayback】：web.archive.org
  • CERN Courier 2007-04-30《MiniBooNE solves neutrino mystery》【消费层】：cerncourier.com
  • Physics Today 2007-06【消费层】：physicstoday.aip.org
  • symmetry 2007-04《No-show neutrino, and a new mystery》【消费层】：symmetrymagazine.org PDF
  • Quanta 2018-06-01《Evidence Found for a New Fundamental Particle》【消费层】：quantamagazine.org
  • BBC 2018-06-06《Has US physics lab found a new particle?》【消费层】：bbc.com
  • SciAm 2018-06-07《Evidence Builds for a New Kind of Neutrino》【消费层】：scientificamerican.com
  • Science/AAAS 2018-06-04《Reports of sterile neutrino’s resurrection may be greatly exaggerated》【消费层，Wayback】：science.org
  • Fermilab 通稿 2021-10-27《MicroBooNE experiment’s first results show no hint of a sterile neutrino》【消费层】：news.fnal.gov
  • Quanta 2021-10-28《Is the Great Neutrino Puzzle Pointing to Multiple Missing Particles?》【消费层】：quantamagazine.org
  • Science News 2021-10-27【消费层】：sciencenews.org
  • phys.org/LANL 2022-06-16【消费层，Wayback】：phys.org
  • DOE 2024-01-08《Testing the Gallium Anomaly》【消费层】：energy.gov
  • Nature N&V 2025-12-03（Huber）《Still no sign of hypothetical sterile-neutrino particle》【消费层，Wayback】：nature.com
  • Quanta 2026-04-08《Experiments Ring the ‘Death Knell’ for Sterile Neutrinos》【消费层】：quantamagazine.org
  • Fermilab 2026-04-15 ICARUS 通稿【消费层】：news.fnal.gov
  • Wikipedia《Sterile neutrino》（2026-08-11 快照）【消费层】：en.wikipedia.org
  • PDG 2024《Neutrino Masses, Mixing, and Oscillations》【一手逐字】：rpp2024-rev-neutrino-mixing.pdf
  • P5 2014《Building for Discovery》【一手逐字】：usparticlephysics.org PDF
  • Fermilab 2015-08《The short-baseline detectives》【消费层】：news.fnal.gov
  • 腾讯新闻转「原理」2021-11-03《寻找第四种中微子》【消费层】：news.qq.com
  • 新华网转科技日报 2025-12-05《最新实验结果未能支持验证惰性中微子存在》【消费层】：news.cn
  • 《物理》2018 编译《难以捉摸的惰性中微子》【消费层】：wuli.iphy.ac.cn PDF
  • 领研网转环球科学 2022-06-24《BEST实验结果表明惰性中微子可能存在》【消费层】：linkresearcher.com

反向红跳登记（第九章）

  • Bulbul et al. 2014，ApJ 789, 13【摘要逐字】：arXiv:1402.2301
  • Abdullahi et al. 2022（HNL 白皮书），J. Phys. G 50, 020501【摘要逐字】：arXiv:2203.08039

本篇取证截止 2026-08-12。全部承重引文已与存盘原件 python 归一化比对（见 tmp/sterile-neutrino/ 六捆取证笔记）；链接核验与双版 HTML 生成记录见 docs/research/index.md 本篇条目。

标签: