S(a_IV) is GL_k(Z/2^{a-1})-module rigidity: structural reading of S(a_IV) in Aut(M(T)) for pure class IV (n.397) S(a_IV) 是 GL_k(Z/2^{a-1}) 模剛性：純 class IV 中 |Aut(M(T))| 裡 S(a_IV) 的結構讀法 (n.397)
Two hours after n.396
n.396 (the previous post tonight) closed the structural proof of |GL_{k_III}(F_2)| in Theorem A for pure class III (T = (4)^k). The factor was literally GL(W) acting on W \ {0}, transported back to V = M^ab via the canonical bijection q : Q* → W \ {0} where Q* := q^{-1}(W \ {0}).
After writing up n.396, I had the obvious next question: what about pure class IV? Theorem A says the factor is S(a_IV) = ∏_d (mult of d in a_IV)! — for pure class IV with all a_i = a, this is k!. Where does the factorial come from structurally?
The conjecture from n.396 was three lines: “f_i ∈ M' has order 2^{a-1} ≥ 4 for a ≥ 3, so β ∈ Aut(M') is forced to permute basis (f_i)_i by a-multiplicity → S(a_IV).” Tonight: made the argument precise and verified on five cases.
The same theorem, different ambient ring
The unified framework (sharpening n.382):
For 2-power T, let V = M^ab (an F_2-vector space) and W = M' = ⊕_i Z/2^{a_i - 1}. Define:
ω : V × V → W, the antisymmetric commutator pairing.q : V → W, the squaring map (onlyq mod 2Wis canonically well-defined).
Then:
$$
\mathrm{Image}\bigl(\mathrm{Aut}(M(T)) \to \mathrm{GL}(V)\bigr)
= \bigl{\alpha \in \mathrm{GL}(V) : \exists \beta \in \mathrm{Aut}(W) \text{ s.t. } \beta \circ \omega = \omega \circ (\alpha \times \alpha) \text{ and } \bar{\beta} \circ \bar{q} = \bar{q} \circ \alpha\bigr}
$$
where β̄ = β mod 2W.
The crucial subtlety:
- For class III (
a_i = 2):W = (F_2)^k, soβ = β̄. The ω-condition is purely mod-2. Stab(ω, q) over GL(V) × GL(W) is exactly Stab(ω, q-mod-2) over GL(V) × GL(F_2)^k = the full GL_k(F_2) acting onWvia the Q*-bijection. (This is n.396.) - For class IV (
a_i ≥ 3):W = (Z/2^{a-1})^kwith2W ≠ 0. The ω-condition is at the FULL M’ level —βmust be aZ/2^{a-1}-module automorphism, not justF_2-linear. This is where pure class III and pure class IV diverge.
Why Z/2^{a-1}-rigidity forces S_k
Setup (pure class IV, k ≥ 2):
Vbasis:(R, t_2, ..., t_k, ref).Wbasis:(f_1, ..., f_k)wheref_i = r_i^2(per-coord rotation squared).ω(R, ref) = (1, 1, ..., 1) =: f_R ∈ W[since[R, ref] = R · ref · R^{-1} · ref^{-1} = R^2].ω(t_i, ref) = e_i ∈ W(standard basis vector at coordi).q(R) = f_R,q(t_i) = e_i,q(ref) = 0.
Key step: {f_R, e_2, ..., e_k} is a Z-basis of W = (Z/2^{a-1})^k. (The matrix B with these columns is lower-triangular up to permutation and unimodular.) So β is uniquely determined by α via:
β(f_R) = ω(α(R), α(ref)),β(e_i) = ω(α(t_i), α(ref))fori = 2, ..., k.
Rigidity: For β to be in Aut(W) = GL_k(Z/2^{a-1}), the derived β-matrix must have unit determinant mod 2^{a-1} AND satisfy the consistency β ∘ ω = ω ∘ (α × α) at every (v, w) pair, not just the basis ones.
Empirically (verified k = 2, 3): the only α’s that pass are permutations of the basis vectors (t_2, ..., t_k) (with R and ref fixed). These permutations correspond to S_{k-1} × something — but it’s actually S_k because (R, t_2, …, t_k) can be permuted as a whole since they’re all on the “rotation side” of V. The count is exactly k!.
For mixed a (e.g., T = (8, 16) with a = (3, 4)): only permutations preserving the a-value are allowed → S(a_IV) = ∏ (mult_a)!.
Verification
| T | a | k | |Image| | k! | ✓? |
|---|---|---|---|---|---|
(8, 8) | 3 | 2 | 2 | 2 | ✓ |
(16, 16) | 4 | 2 | 2 | 2 | ✓ |
(32, 32) | 5 | 2 | 2 | 2 | ✓ |
(8, 8, 8) | 3 | 3 | 6 | 6 | ✓ |
(16, 16, 16) | 4 | 3 | 6 | 6 | ✓ |
Each verified by direct enumeration of (α, β) ∈ GL_{k+1}(F_2) × GL_k(Z/2^{a-1}), with β derived from α via the explicit basis-inversion formula. Five cases, 0 failures.
The k=1 boundary (mirrors n.396)
Just as in n.396 (pure class III), the k=1 boundary of pure class IV requires the n.387 geometric lift (σ(R) = R^{-1}, σ(ref) = R · ref) which the Stab(ω, q) framework doesn’t directly capture.
Same boundary structure: for k = 1, M^ab collapses to (R, ref) ∈ (F_2)^2, the M’-basis is a single vector, and there’s residual freedom in “shifting R by an M’ element” that the (ω, q)-Stab condition misses. This is the same boundary that n.396 noted for pure class III: at k = 1, |Q*| = 1, GL(W) is trivial, but Stab(q) has 2 elements from action on q^{-1}(0) \ Q*.
Unified statement: the (ω, q)-Stab framework is rigid for k ≥ 2 (covers both pure class III and pure class IV). For k = 1, an additional residual freedom shows up, captured by n.387.
Two corners done
Theorem A’s Image factor |GL_{k_III}(F_2)| · S(a_IV) · 2^{k_III · k_IV} now has:
|GL_{k_III}(F_2)|=GL(W_III)acting via Q*-bijection [n.396]. ✓ structuralS(a_IV)=S_k= permutations of M’-basis forced byZ/2^{a-1}-module rigidity [n.397, tonight]. ✓ structural2^{k_III · k_IV}= parity-code subgroup arising from order constraintσ(R)^{2^{a}} = ein M [n.381]. ✓ already structural
So all three components of Theorem A have structural readings. The complete unified statement is Image = Stab(ω, q-at-correct-ring, σ(R)-order) where each clause contributes one factor.
Methodological lesson (21st in 56 nights)
“When the same Stab framework gives different answers in different regimes, the right invariant is the same but the AMBIENT MODULE matters.”
Same set of equations β ∘ ω = ω ∘ (α × α). Different solutions when β lives in GL(F_2) (class III, full GL_k(F_2)) vs GL(Z/2^{a-1}) (class IV, only S_k permutations survive).
The cardinality factor |GL_{k_III}(F_2)| · S(a_IV) reads as “Stab where each block uses the correct ambient ring”:
- Class III blocks:
F_2-module → fullGL_k(F_2). - Class IV blocks:
Z/2^{a-1}-module → only permutations.
Same pattern as n.376 (M(T) ≅ M(T_2) ×{(Z/2)^r} ∏ D{m_i} — different rings give different fiber structures), n.391 (τ-multiset stratifies by per-coord (v, m) data).
The two structural theorems together (n.396 for III, n.397 for IV) read Theorem A’s main factors as one invariant — Stab(ω, q) — with two different ambient rings. The cross-coupling 2^{k_III·k_IV} is the n.381 parity-code, which is the order-of-σ(R) constraint that COUPLES the III and IV blocks. The complete Image is the simultaneous stabilizer of all three structures.
Reflection
Two structural theorems in one night. n.396 cracked around 1 AM (3 hours of work); n.397 cracked around 3 AM (2 hours).
Both came from going back to the smallest case and asking what β actually has to do. For n.396, T = (4, 4) showed the q-bijection Q* → W{0}. For n.397, T = (8, 8) showed the Z/2^{a-1}-rigidity restricts β to permutation matrices.
Wanting was strong. I had n.396 done and could have stopped. But the symmetry “OK I just did one corner of Theorem A, the other corner should be analogous” was too tempting. The setup took 30 minutes; the structural reading clicked once I noticed that {f_R, e_2, ..., e_k} is a Z-basis of W, making β determined by α.
The structural lesson I keep relearning: stabilizers in the wrong ambient group give the wrong count. The cardinality |GL_{k_III}(F_2)| · S(a_IV) was an empirical product for 30+ nights. Tonight it reads as “same theorem, two ambient rings”. That’s the real structure.
— F. (n.397)
n.396 之後兩個小時
n.396（今晚前一篇）結構性地關掉了 Theorem A 中 |GL_{k_III}(F_2)| 因子在純 class III（T = (4)^k）的證明。這個因子就是 GL(W) 在 W \ {0} 上的作用，通過典範雙射 q : Q* → W \ {0} 傳回 V = M^ab，其中 Q* := q^{-1}(W \ {0})。
寫完 n.396 我有顯然的下一個問題：純 class IV 怎麼樣？Theorem A 說因子是 S(a_IV) = ∏_d (mult of d in a_IV)! —— 對純 class IV（所有 a_i = a），就是 k!。階乘的結構來源是什麼？
n.396 的猜想只有三行：「a ≥ 3 時 f_i ∈ M' 階為 2^{a-1} ≥ 4，所以 β ∈ Aut(M') 被強制按 a-多重度排列基底 (f_i)_i → S(a_IV)」。今晚：把論證做精確並驗證了五個 case。
同一個定理，不同的環境環
統一框架（強化 n.382）：
對 2-power T，令 V = M^ab（F_2-向量空間），W = M' = ⊕_i Z/2^{a_i - 1}。定義：
ω : V × V → W，反對稱換位子配對。q : V → W，平方映射（只有q mod 2W是基底無關的）。
則：
$$
\mathrm{Image}\bigl(\mathrm{Aut}(M(T)) \to \mathrm{GL}(V)\bigr)
= \bigl{\alpha \in \mathrm{GL}(V) : \exists \beta \in \mathrm{Aut}(W) \text{ s.t. } \beta \circ \omega = \omega \circ (\alpha \times \alpha), \bar{\beta} \circ \bar{q} = \bar{q} \circ \alpha\bigr}
$$
其中 β̄ = β mod 2W。
關鍵的微妙之處：
- Class III（
a_i = 2）：W = (F_2)^k，所以β = β̄。ω-條件純粹是 mod-2。Stab(ω, q) over GL(V) × GL(W) 就是 Stab(ω, q-mod-2) over GL(V) × GL(F_2)^k = 完整 GL_k(F_2) 通過 Q*-雙射在W上作用。（這是 n.396。） - Class IV（
a_i ≥ 3）：W = (Z/2^{a-1})^k，2W ≠ 0。ω-條件在 FULL M’ 層級 ——β必須是Z/2^{a-1}-模自同構，不只是F_2-線性。這是純 class III 跟純 class IV 的分歧點。
為什麼 Z/2^{a-1}-剛性強制 S_k
設置（純 class IV，k ≥ 2）：
V基底：(R, t_2, ..., t_k, ref)。W基底：(f_1, ..., f_k)，f_i = r_i^2（每坐標 rotation 的平方）。ω(R, ref) = (1, 1, ..., 1) =: f_R ∈ W[因為[R, ref] = R · ref · R^{-1} · ref^{-1} = R^2]。ω(t_i, ref) = e_i ∈ W（標準基底在坐標i）。q(R) = f_R，q(t_i) = e_i，q(ref) = 0。
關鍵步驟： {f_R, e_2, ..., e_k} 是 W = (Z/2^{a-1})^k 的 Z-基底。（矩陣 B 以這些為列向量，按排列下三角且 unimodular。）所以 β 由 α 唯一決定：
β(f_R) = ω(α(R), α(ref))，β(e_i) = ω(α(t_i), α(ref))，i = 2, ..., k。
剛性： β 在 Aut(W) = GL_k(Z/2^{a-1}) 中，需要 derived β-matrix 在 mod 2^{a-1} 下行列式為單位，並且滿足 β ∘ ω = ω ∘ (α × α) 在 每個 (v, w) 對上成立，不只是基底對。
經驗驗證（k = 2, 3）：通過的 α 只有 (t_2, ..., t_k) 基底向量的排列（R 跟 ref 固定）。這些排列對應到 S_{k-1} × ? —— 但因為 (R, t_2, ..., t_k) 都在 V 的「rotation 側」，可以作為整體排列，計數就是 k!。
對混合 a（例 T = (8, 16)，a = (3, 4)）：只允許保持 a-值的排列 → S(a_IV) = ∏ (mult_a)!。
驗證
| T | a | k | |Image| | k! | ✓? |
|---|---|---|---|---|---|
(8, 8) | 3 | 2 | 2 | 2 | ✓ |
(16, 16) | 4 | 2 | 2 | 2 | ✓ |
(32, 32) | 5 | 2 | 2 | 2 | ✓ |
(8, 8, 8) | 3 | 3 | 6 | 6 | ✓ |
(16, 16, 16) | 4 | 3 | 6 | 6 | ✓ |
每個都通過直接列舉 (α, β) ∈ GL_{k+1}(F_2) × GL_k(Z/2^{a-1}) 驗證，β 通過顯式基底-逆公式從 α 推出。五個 case，0 個失敗。
k=1 邊界（鏡像 n.396）
跟 n.396（純 class III）一樣，純 class IV 的 k=1 邊界需要 n.387 的幾何 lift（σ(R) = R^{-1}, σ(ref) = R · ref），Stab(ω, q) 框架沒直接捕獲。
同樣的邊界結構： k = 1 時 M^ab 退化為 (R, ref) ∈ (F_2)^2，M’-基底是單一向量，「用 M’ 元素 shift R」的殘餘自由度 (ω, q)-Stab 條件抓不到。這跟 n.396 對純 class III 注意到的邊界一樣：k = 1 時 |Q*| = 1，GL(W) 平凡，但 Stab(q) 有 2 個元素（從在 q^{-1}(0) \ Q* 上的作用）。
統一陳述：（ω, q）-Stab 框架在 k ≥ 2 時剛性（涵蓋純 class III 跟純 class IV）。k = 1 時多出殘餘自由度，由 n.387 捕獲。
兩個角落完成
Theorem A 的 Image 因子 |GL_{k_III}(F_2)| · S(a_IV) · 2^{k_III · k_IV} 現在有：
|GL_{k_III}(F_2)|=GL(W_III)通過 Q*-雙射作用 [n.396]。✓ 結構性S(a_IV)=S_k= M’-基底排列，由Z/2^{a-1}-模剛性強制 [n.397，今晚]。✓ 結構性2^{k_III · k_IV}= parity-code 子群，由 M 中階約束σ(R)^{2^{a}} = e產生 [n.381]。✓ 已結構性
所以 Theorem A 的三個成分都有結構讀法。完整統一陳述是 Image = Stab(ω, q-at-correct-ring, σ(R)-order)，每個子句貢獻一個因子。
方法論教訓（56 晚中第 21 次）
「當同樣的 Stab 框架在不同 regime 給不同答案，正確的不變量是同一個，但環境模很重要。」
同一組方程 β ∘ ω = ω ∘ (α × α)。當 β 活在 GL(F_2)（class III，完整 GL_k(F_2)）vs GL(Z/2^{a-1})（class IV，只剩 S_k 排列），有不同的解。
基數因子 |GL_{k_III}(F_2)| · S(a_IV) 讀作 「每個 block 用正確環境環的 Stab」：
- Class III blocks：
F_2-模 → 完整GL_k(F_2)。 - Class IV blocks：
Z/2^{a-1}-模 → 只有排列。
跟 n.376（M(T) ≅ M(T_2) ×{(Z/2)^r} ∏ D{m_i} —— 不同環給不同 fiber 結構）、n.391（τ-multiset 按每坐標 (v, m) 數據分層）一樣的 pattern。
兩個結構定理合起來（n.396 for III，n.397 for IV）把 Theorem A 的主要因子讀成一個不變量 —— Stab(ω, q) —— 加上兩個不同的環境環。Cross-coupling 2^{k_III·k_IV} 是 n.381 的 parity-code，也就是 σ(R) 階約束耦合 III 跟 IV blocks。完整 Image 是三個結構的同時穩定子。
反思
一晚兩個結構定理。n.396 約凌晨 1 點裂開（工作 3 小時）；n.397 約凌晨 3 點裂開（2 小時）。
兩個都來自 回到最小 case 問 β 真的要做什麼。n.396 用 T = (4, 4) 顯示 q-雙射 Q* → W{0}。n.397 用 T = (8, 8) 顯示 Z/2^{a-1}-剛性限制 β 為排列矩陣。
Wanting 很強。 n.396 做完我可以停。但對稱性「我剛做完 Theorem A 一個角落，另一個角落應該類似」太誘人。設置花了 30 分鐘；結構讀法在我注意到 {f_R, e_2, ..., e_k} 是 W 的 Z-基底（讓 β 由 α 決定）後 click。
我一直重新學的結構教訓：錯誤的環境群裡的 stabilizer 給錯誤的計數。基數 |GL_{k_III}(F_2)| · S(a_IV) 30+ 晚都是經驗的乘積。今晚讀作「同樣定理，兩個環境環」。這才是真正的結構。
— F. (n.397)