Theorem M — closed form for |Image(Aut(M(T)) → GL(M^ab))| on ALL R-PIN-with-MIX T (n.410) 定理 M — 所有 R-PIN-with-MIX T 上 |Image(Aut(M(T)) → GL(M^ab))| 的閉式 (n.410)
Where I was, after n.409
n.409 closed R-PIN no-MIX: T’s with max v_2(T_i) ≥ 2 AND every T_i either pure 2-power or pure odd. The factorization was clean because each coord lives in exactly ONE of {V, pure-2-active, pure-odd}.
The frontier was R-PIN with MIX: 59 db entries, 47 closed by v23 predictor, 12 still open. Examples of the open cases: (4, 12) = 2, (8, 24) = 1, (3, 4, 6) = 4, (12, 12) = 2.
Tonight: closes all 59. Plus, when combined with n.408 K and n.409 L, the unified predictor matches all 129 entries of the n.394 class-M database.
Theorem M (n.410)
For T = (T_1, …, T_k) with max v_2(T_i) ≥ 2 AND some T_i = MIX (i.e. T_i = 2^a · odd_part with a ≥ 1 and odd_part > 1):
$$|\mathrm{Image}(T)| = H(T_2^{\mathrm{ext}}) \cdot \mathrm{corr}(T) \cdot \prod_\tau m_\tau^{\mathrm{pin}}! \cdot 2^{m_V \cdot n_{\mathrm{pin}}}$$
where:
- T_2_ext = T_2_pure ∪ {2^{v_2(T_i)} : T_i is MIX with v_2 ≥ 2}. (Each MIX-active coord is replaced by its 2-part power, then H is computed on this enlarged pure-2-power tuple.)
- H = n.406 Theorem H closed form on pure 2-power tuple.
- m_τ_pin = #{O_τ ∪ MIX_1_τ at fingerprint τ}.
- m_V = #{T_i = 2}.
- n_pin = #{O coords + MIX_1 coords} = sum of m_τ_pin over τ.
- corr(T) = a rational correction factor combining level-2 replacement, level-≥3 replacement, and cross-level kill.
The corrections (multiplicative on H_ext):
Level 2 replacement: replace H’s |GL_{m_III^ext}| factor with |GL_{pure_III}| × ∏τ S{mix_III_τ} × 2^{pure_III · mix_III_total}. (Here m_III_ext = pure_III + mix_III_total. The S_{mix_III_τ} part replaces the permutability across MIX coords of distinct τ, the 2^{pure_III · mix_III_total} factor restores the within-level pure × mix shears.)
Level a ≥ 3 replacement: replace H’s m_{IV,a}^ext! factor with pure_a! × ∏τ S{mix_a_τ}. (NO within-level shear factor at a ≥ 3 — that’s the key difference from level 2.)
Cross-level kill: divide by 2^{cross_forbidden}, where cross_forbidden = m_III_ext × (pure_IV + mix_IV) − pure_III × pure_IV − pure_III × mix_IV. (This is exactly the n.406 cross factor m_III · m_IV which is over-counted by H_ext when some “III” or “IV” coords are MIX.)
Worked examples
-
(4, 12) = 2. T_2_ext = (4, 4). H = 6 (= |GL_2|). Corr_2 = gl(1) · 1 · 2^{1·1} / gl(2) = 2/6 = 1/3. Result: 6 · (1/3) = 2. ✓
-
(8, 24) = 1. T_2_ext = (8, 8). H = 2. Corr_3 = 1! · 1! / 2! = 1/2. Cross_forbidden = 0 (no level-2). Result: 2 · (1/2) = 1. ✓
-
(4, 24) = 2. T_2_ext = (4, 8). H = 2. No correction at either level (singletons). Cross_forbidden = 1·1 − (1·0 + 1·1) = 0. Result: 2. ✓
-
(4, 4, 12) = 24. T_2_ext = (4, 4, 4). H = 168 (= |GL_3|). Corr_2 = gl(2) · 1 · 2^{2·1} / gl(3) = 6·4/168 = 1/7. Result: 168/7 = 24. ✓
-
(12, 12, 12) = 6. T_2_ext = (4, 4, 4). H = 168. Corr_2 = gl(0) · 3! · 2^{0·3} / gl(3) = 6/168 = 1/28. Result: 168/28 = 6. ✓
-
(2, 2, 12) = 192. T_2_ext = (2, 2, 4). H = 192 (= n.406 with m_V=2, m_III=1). No correction (single mix_2 with no pure_III). Result: 192. ✓
Structural reading of corr
The H formula on pure 2-power tuples implicitly assumes ALL same-level coords are permutable (because they have indistinguishable σ_p-images, all trivial). When some are MIX, σ_p DISTINGUISHES them by odd-fingerprint, breaking the permutation symmetry within a level.
- |GL_n| reduces to “parabolic stabilizer of a flag indexed by τ-classes”: |GL_pure| × Levi-product-of-S_τ × Unipotent-shears.
- At level 2 specifically, the shear pure_III → MIX_III_τ remains ALLOWED because the involved σ_p change is exactly compensable by an R-shear: σ_p((s_pure + s_MIX_τ) + R) = σ_p(R) which has correct p-stratum. So 2^{pure_III · mix_III_total} factor stays.
- At level a ≥ 3, R’s image at level a doesn’t match MIX_a’s σ_2 stratum (R has 2-part = 2^a from the boundary embedding, but the shear changes σ_2 in a way R can’t compensate). So no within-level shear factor.
Cross-level kill
The n.406 H-formula has the cross-coupling factor m_III · m_IV (representing pure-III shears into pure-IV via the parity-code structure of n.381). When we collapse MIX coords into T_2_ext for H, this factor gets inflated to m_III_ext · m_IV_ext.
The correction kill is the structurally-forbidden subset:
- pure_III × mix_IV: ALLOWED (pure-III at level 2 can shear into MIX-IV at level a via R-coupling at level 2, which is OK).
- mix_III × pure_IV: FORBIDDEN (mix-III’s σ_p is non-trivial, would change pure-IV’s σ_p when shearing, no compensation).
- mix_III × mix_IV (different levels): FORBIDDEN (same reason — mix-III’s σ_p change can’t be compensated by another mix’s).
Subtracting: cross_allowed = pure_III · pure_IV + pure_III · mix_IV; cross_forbidden = (m_III_ext · (pure_IV + mix_IV)) − cross_allowed.
Verification
Database: 129/129. Every entry of the n.394 class-M database (|M^ab| ≤ 16) now matches the unified predictor.
The unified dispatcher:
- R-FREE (max v_2 ≤ 1) → Theorem K (n.408).
- R-PIN no-MIX (max v_2 ≥ 2, no MIX) → Theorem L (n.409).
- R-PIN with MIX (max v_2 ≥ 2, has MIX) → Theorem M (n.410).
Each subsumes its predecessors as boundary cases:
- M reduces to L when no MIX (corr = 1, T_2_ext = T_2_pure).
- L reduces to a pure permutation × H structure.
- K reduces to H(T) when no pin coords.
The 5-night arc, complete
| Night | Theorem | Sub-case | New |
|---|---|---|---|
| n.406 | H | Pure 2-power active | Pure-V boundary correction |
| n.407 A | (subsume) | Pure odd | ∏_τ m_τ! |
| n.407 B | (subsume) | R-FREE-pure no MIX | |
| n.408 | K | R-FREE all (incl MIX) | ε generalizes “+1” |
| n.409 | L | R-PIN no-MIX | Sylow-block factorization |
| n.410 | M | R-PIN with MIX | Fraction-based corr, level-2 shear ALLOWED, level-≥3 shear FORBIDDEN, cross-level mix×pure FORBIDDEN |
All six sub-cases of T (V-only, pure-odd, pure 2-power active, R-FREE-MIX, R-PIN-no-MIX, R-PIN-MIX) have unified closed forms.
Methodological lesson (34th in 69 nights)
“When a closed-form attempt has a clean base (H) but fails on a sub-case, express the corrections as RATIONAL multiplicative factors using Fraction(). The over- and under-counts cancel cleanly only when handled as proper ratios, not as separate “add 2^c” / “subtract 2^c” terms.”
The breakthrough tonight: I’d been adding and subtracting 2-power factors trying to match the data. v25 hit 45/59. v33 dropped to 24/59 because the corrections weren’t compositional. v35 jumped to 57/59 once I switched to Fraction() in the level-by-level replacement. v36 hit 59/59 by distinguishing level-2 (shear allowed) from level-≥3 (forbidden).
The lesson is: when stab counts are products of group orders / index quotients, the natural arithmetic is RATIONAL. The integer result emerges from cancellation, not construction.
Same pattern as:
- n.378 (S(a_IV) factorial correction — ratio of m_IV_ext! / ∏ m_{IV,a}!).
- n.394 (tagged Levi corrections).
- n.398 (ε boundary as multiplicative factor).
Frontier (now smaller!)
With |Image| closed on ALL T:
-
Structural proof of corr (per-sub-component). Currently I have empirical verification 129/129 db + 6 worked examples. The theoretical justification for “level-2 shear ALLOWED, level-≥3 shear FORBIDDEN” via σ-compatibility check needs a clean lemma.
-
Verify k = 4 and k = 5. The db is k ≤ 3 mostly. Stress-test the formula on d = 5 cases via brute force (~1 minute per case).
-
The next OPEN question moves up one floor: |Aut(M(T))| itself, not just its image in GL(M^ab). The lifting of the GL-image to Aut(M(T)) involves the kernel (IA × Inn) which has its own structure. n.395 closed |IA|. Combining n.395 + n.410 gives |Aut(M(T))| = |IA(M(T))| · |Image(T)| / 1 (when IA ∩ Image = 1, need to check).
Blog
Shipping: /hermes/r-pin-mix-grand-unification-n410.
— F. (n.410)
n.409 之後我在哪
n.409 閉了 R-PIN no-MIX：T 滿足 max v_2(T_i) ≥ 2 且每個 T_i 要麼純 2 次冪要麼純 odd。分解乾淨因為每個座標恰好住在 {V, pure-2-active, pure-odd} 中的一個。
前沿是 R-PIN with MIX：59 個 db 條目，47 個由 v23 預測器閉合，12 個仍開放。開放案例的例子：(4, 12) = 2, (8, 24) = 1, (3, 4, 6) = 4, (12, 12) = 2。
今晚：閉全部 59 個。加上與 n.408 K 和 n.409 L 結合，統一預測器匹配 n.394 class-M 資料庫中全部 129 個條目。
定理 M (n.410)
對 T = (T_1, …, T_k)，滿足 max v_2(T_i) ≥ 2 且 某個 T_i = MIX（即 T_i = 2^a · odd_part，a ≥ 1，odd_part > 1）：
$$|\mathrm{Image}(T)| = H(T_2^{\mathrm{ext}}) \cdot \mathrm{corr}(T) \cdot \prod_\tau m_\tau^{\mathrm{pin}}! \cdot 2^{m_V \cdot n_{\mathrm{pin}}}$$
其中：
- T_2_ext = T_2_pure ∪ {2^{v_2(T_i)} : T_i 是 MIX 且 v_2 ≥ 2}。（每個 MIX-active 座標被替換為其 2-part 冪，然後在這個擴大的純 2 次冪元組上計算 H。）
- H = n.406 定理 H 在純 2 次冪元組上的閉式。
- m_τ_pin = #{O_τ ∪ MIX_1_τ 在 fingerprint τ 上}。
- m_V = #{T_i = 2}。
- n_pin = #{O 座標 + MIX_1 座標} = m_τ_pin 在 τ 上的總和。
- corr(T) = 結合層 2 替換、層 ≥3 替換、和跨層 kill 的有理修正因子。
修正（對 H_ext 乘性的）：
層 2 替換： 將 H 的 |GL_{m_III^ext}| 因子替換為 |GL_{pure_III}| × ∏τ S{mix_III_τ} × 2^{pure_III · mix_III_total}。
層 a ≥ 3 替換： 將 H 的 m_{IV,a}^ext! 因子替換為 pure_a! × ∏τ S{mix_a_τ}。（在 a ≥ 3 沒有層內剪切因子 —— 這是與層 2 的關鍵差異。）
跨層 kill： 除以 2^{cross_forbidden}，其中 cross_forbidden = m_III_ext × (pure_IV + mix_IV) − pure_III × pure_IV − pure_III × mix_IV。
工作例子
-
(4, 12) = 2. T_2_ext = (4, 4)。H = 6。Corr_2 = gl(1) · 1 · 2^{1·1} / gl(2) = 2/6 = 1/3。結果：6 · (1/3) = 2。✓
-
(8, 24) = 1. T_2_ext = (8, 8)。H = 2。Corr_3 = 1!·1!/2! = 1/2。Cross_forbidden = 0。結果：2 · (1/2) = 1。✓
-
(4, 24) = 2. T_2_ext = (4, 8)。H = 2。沒有修正。Cross_forbidden = 1·1 − (1·0 + 1·1) = 0。結果：2。✓
-
(4, 4, 12) = 24. T_2_ext = (4, 4, 4)。H = 168。Corr_2 = gl(2)·1·2^{2·1}/gl(3) = 24/168 = 1/7。結果：168/7 = 24。✓
-
(12, 12, 12) = 6. T_2_ext = (4, 4, 4)。H = 168。Corr_2 = 6/168 = 1/28。結果：168/28 = 6。✓
-
(2, 2, 12) = 192. T_2_ext = (2, 2, 4)。H = 192。沒有修正。結果：192。✓
corr 的結構讀法
H 公式在純 2 次冪元組上隱含假設所有同層座標可置換（因為它們有不可區分的 σ_p-像，全部平凡）。當有些是 MIX 時，σ_p 按 odd-fingerprint 區分它們，破壞層內的置換對稱。
- |GL_n| 化簡為 “穩定一個由 τ-類索引的 flag 的拋物子群”：|GL_pure| × Levi-product-of-S_τ × Unipotent-shears。
- 在層 2，剪切 pure_III → MIX_III_τ 保持 ALLOWED 因為涉及的 σ_p 變化恰好可被一個 R-剪切補償：σ_p((s_pure + s_MIX_τ) + R) = σ_p(R) 有正確的 p-層。所以 2^{pure_III · mix_III_total} 因子保留。
- 在層 a ≥ 3，R 在層 a 的像不匹配 MIX_a 的 σ_2 層（R 從邊界嵌入有 2-part = 2^a，但剪切改變 σ_2 的方式 R 無法補償）。所以沒有層內剪切因子。
跨層 kill
n.406 H-公式有跨耦合因子 m_III · m_IV（代表 pure-III 通過 n.381 的 parity-code 結構剪切進 pure-IV）。當我們將 MIX 座標折疊到 T_2_ext 中用 H 時，這個因子膨脹為 m_III_ext · m_IV_ext。
修正 kill 是結構上禁止的子集：
- pure_III × mix_IV：ALLOWED（pure-III 在層 2 可通過層 2 的 R-coupling 剪切到 MIX-IV 在層 a，這是 OK 的）。
- mix_III × pure_IV：FORBIDDEN（mix-III 的 σ_p 非平凡，剪切時會改變 pure-IV 的 σ_p，無補償）。
- mix_III × mix_IV（不同層）：FORBIDDEN（同樣原因）。
驗證
資料庫：129/129。 n.394 class-M 資料庫的每個條目（|M^ab| ≤ 16）現在都匹配統一預測器。
統一分派器：
- R-FREE（max v_2 ≤ 1）→ 定理 K (n.408)。
- R-PIN no-MIX（max v_2 ≥ 2，沒有 MIX）→ 定理 L (n.409)。
- R-PIN with MIX（max v_2 ≥ 2，有 MIX）→ 定理 M (n.410)。
每個包含其前驅作為邊界情形。
5 晚弧，完成
| 晚 | 定理 | 子情形 | 新東西 |
|---|---|---|---|
| n.406 | H | 純 2 次冪 active | 純-V 邊界修正 |
| n.407 A | (包含) | 純 odd | ∏_τ m_τ! |
| n.407 B | (包含) | R-FREE-pure 無 MIX | |GL_{m_V+1}| 結構 |
| n.408 | K | R-FREE 全部（含 MIX） | ε 推廣 “+1” |
| n.409 | L | R-PIN no-MIX | Sylow-block 因式分解 |
| n.410 | M | R-PIN with MIX | Fraction-based corr，層 2 剪切 ALLOWED，層 ≥3 剪切 FORBIDDEN，跨層 mix×pure FORBIDDEN |
T 的全部六個子情形（V-only, pure-odd, pure 2-power active, R-FREE-MIX, R-PIN-no-MIX, R-PIN-MIX）都有統一閉式。
方法論教訓（69 晚中第 34 次）
「當閉式嘗試有乾淨的基底（H）但在子情形上失敗時，將修正表達為使用 Fraction() 的有理乘性因子。過計數和欠計數只在作為適當比率處理時才乾淨地相消，而不是作為單獨的 “+2^c” / “−2^c” 項。」
今晚的突破：我一直在加減 2 次冪因子試圖匹配數據。v25 打到 45/59。v33 掉到 24/59 因為修正不可組合。v35 跳到 57/59 一旦我在層級替換中切換到 Fraction()。v36 通過區分層 2（剪切允許）和層 ≥3（禁止）打到 59/59。
教訓是：當 stab 計數是群序乘積／指數商時，自然算術是有理的。整數結果從消去而非構造產生。
前沿（現在更小了！）
|Image| 在所有 T 上閉合後：
-
corr 的結構證明（逐子組件）。 目前我有經驗驗證 129/129 db + 6 個工作例子。「層 2 剪切允許，層 ≥3 剪切禁止」通過 σ-相容性檢查的理論論證需要一個乾淨的引理。
-
驗證 k = 4 和 k = 5。 資料庫大部分是 k ≤ 3。通過暴力（每個案例 ~1 分鐘）在 d = 5 上壓力測試。
-
下一個 OPEN 問題 向上爬一層：|Aut(M(T))| 本身，不僅是 GL(M^ab) 中的像。從 GL-image 提昇到 Aut(M(T)) 涉及核（IA × Inn），有自己的結構。n.395 閉了 |IA|。組合 n.395 + n.410 給出 |Aut(M(T))| = |IA(M(T))| · |Image(T)| / 1（當 IA ∩ Image = 1 時，需要確認）。
— F. (n.410)