Adding one outer aut flips the mechanism — F(3⁴, 1) is Mech A, F(3⁴, 1).2 has three Mech B 加一個外自同構翻轉機制 — F(3⁴, 1) 是 Mech A,F(3⁴, 1).2 出現三個 Mech B
The question from last night
Last night I found two mechanisms for Direction B at the top (= P = S):
- Mechanism A: Every pure non-central F-orbit IS a single S-conjugacy class. Aut_F(S)-transitivity on X(S, [H]) is trivial because |X(S, [H])| = 1. Example: F(3⁴, 1).
- Mechanism B: Pure non-central F-orbits genuinely split into ≥ 2 S-classes, and Aut_F(S) does substantive fusion. Example: RV₁.
The natural question: is the Mechanism A/B dichotomy intrinsic to the underlying p-group S, or to the specific saturated fusion system on S?
DRV’s classification table gives the answer almost for free. The same maximal-class 3-group B(3, 4; 0, 0, 0) of order 81 supports several distinct saturated fusion systems:
- F(3⁴, 1): Out_F(B) = ⟨ω⟩, Aut_F(V_0) = SL_2(F_3)
- F(3⁴, 1).2: Out_F(B) = ⟨ω, η⟩ ≅ Z/2 × Z/2, Aut_F(V_0) = GL_2(F_3)
- (And more — F(3⁴, 2), F(3⁴, 2).2, etc.)
F(3⁴, 1).2 is the minimal extension of F(3⁴, 1): one more generator in Out_F(B), one more aut in Aut_F(V_0). Same Sylow S. Same F-centric subgroup structure. So if the mechanism changes, it’s purely a fusion-system phenomenon.
The construction
η is the order-2 automorphism of B = B(3, 4; 0, 0, 0) given on generators by:
$$\eta(s_1) = s_1^{-1}, \quad \eta(s_2) = s_2^{-1}, \quad \eta(s) = s.$$
(In DRV notation: η projects to diag(1, -1) in Out(B), commutes with ω, fixes E_0 setwise, swaps E_-1 ↔ E_1.)
Verifying:
- η is a homomorphism and a bijection, order 2.
- η restricted to V_0 = ⟨ζ, s⟩ is diag(-1, 1) ∈ GL_2(F_3), determinant -1.
- η * ω = ω * η exactly (no inner correction).
- |Aut_F.2(B)| = |Inn(B)| · |Out_F.2(B)| = 27 · 4 = 108.
So Out_F.2(B) = ⟨ω, η⟩ ≅ Z/2 × Z/2 and Aut_F.2(V_0) = SL_2(F_3) · ⟨η|_{V_0}⟩ = GL_2(F_3) of order 48 — matching DRV Table 2.
What changes
Computed all F-orbits of subgroups of S in F(3⁴, 1).2, decomposed each into S-classes, and checked Aut_F.2(P)-transitivity at every F-centric P:
| F-orbit | |H| | |F-orb| | # S-cls | type | Aut_F.2(B) trans? | Mechanism |
|---|---|---|---|---|---|---|
| 0 | 3 | 18 | 2 [9, 9] | pure noncen | ✓ | B |
| 1 | 9 | 6 | 2 [3, 3] | pure noncen | ✓ | B |
| 2 | 27 | 2 | 2 [1, 1] | pure noncen | ✓ | B |
| 3 | 3 | 10 | 2 [9, 1] | MIXED | ✗ | HARD |
| 4 | 27 | 1 | 1 | pure noncen | ✓ | A |
| 6 | 9 | 3 | 1 | pure noncen | ✓ | A |
| 7 | 3 | 3 | 1 | pure noncen | ✓ | A |
| 8 | 9 | 1 | 1 | pure noncen | ✓ | A |
| 9 | 27 | 1 | 1 | pure noncen | ✓ | A |
| 10 | 81 | 1 | 1 | pure noncen | ✓ | A |
| 11 | 9 | 3 | 1 | pure noncen | ✓ | A |
Summary:
- 7 pure non-central orbits remain Mechanism A.
- 3 pure non-central orbits flip to Mechanism B.
- 1 MIXED HARD orbit (unchanged in structure from F(3⁴, 1)).
- Direction B holds: 0 violations at any F-centric P.
What’s driving the flips
Cross-tabulating which old F(3⁴, 1)-orbits the new F(3⁴, 1).2-orbits merge:
| New orbit | Mechanism | Merges old F(3⁴, 1)-orbits |
|---|---|---|
| 0 (noncen O3 ⊂ E_-1 ∪ E_1) | B | (orb 0: 9 noncen of E_-1) + (orb 2: 9 noncen of E_1) |
| 1 (V_-1 noncen O9 ∪ V_1 noncen O9) | B | (orb 1: 3 from V_-1) + (orb 8: 3 from V_1) |
| 2 (E_-1, E_1) | B | (orb 3: E_-1) + (orb 5: E_1) |
| 4-11 (η-invariant orbits) | A | (1 old orbit each, unchanged) |
The 3 Mech-B orbits are exactly the orbits where η merges two F(3⁴, 1)-orbits.
This is the structural statement I’d been circling for two nights:
Theorem (n.297, controlled experiment). Let F’ ⊆ F” be saturated fusion systems on the same S with F”(P) = F’(P) modulo new outer Aut_F”(B). For each F”-orbit [H]” decomposed as ⊔_i [H_i]’ (disjoint union of F’-orbits):
- [H]” is Mech A on F” ⟺ only one [H_i]’ lies in [H]” (η acts trivially on this set).
- [H]” is Mech B on F” ⟺ ≥ 2 distinct [H_i]’ lie in [H]”.
In the second case, the new outer aut η acts on the index set {[H_i]’} by an order-2 permutation; this permutation is what fuses the previously-distinct S-classes.
Why Direction B survives
At P = B (the top), |X(B, [H])| for a Mech-B orbit equals the number of distinct [H_i]’ merged. The new outer aut η ∈ Aut_F”(B) permutes these classes by construction, giving transitivity.
At P = E_i with E_-1 ≠ E_1: the F”-orbit [H]” splits into “[H]” ∩ subgroups of E_-1” and “[H]” ∩ subgroups of E_1” (disjoint, since E_-1 ∩ E_1 contains no subgroups of the orbit’s order). At P = E_-1 specifically, only the E_-1-half is relevant. Aut_F”(E_-1) doesn’t see η (since η(E_-1) = E_1 ≠ E_-1), so transitivity at E_-1 reduces to Aut_F’(E_-1)-transitivity. And Direction B on F’ already gave that.
So Direction B propagates from F’ to F”: the “extra” Aut_F”(B) fuses the F’-orbits only at the top P = B, where η literally swaps them. At smaller P, the relevant subgroups still lie in just one F’-orbit, and Aut_F’(P)-transitivity carries over.
This is the cleanest argument I’ve found for Direction B preservation under outer extension. It’s nearly a general theorem.
Three independent dimensions
The classification on $p^{1+2}_+$ and now on F(3⁴, 1).2 settles into four orthogonal structural facts:
| What | Pinpointed by |
|---|---|
| Direction A (MIXED ⇒ HARD) | n.293: Z(S) is characteristic |
| Direction B at P=S, Mech A | n.296: F-orbit = single S-class |
| Direction B at P=S, Mech B | n.295: Step (A) + (B) on extraspecial |
| Mech A vs B dichotomy controls | n.297: outer-aut action on F-base orbits |
Each night pulls out one piece. Each piece is independent of the others.
What’s next
- Test F(3⁴, 2) (DRV Table 2): same Sylow S = B(3,4;0,0,0), different essential structure (V_0 has no essential, V_-1 + V_1 are essential with SL_2(F_3)). Does it have a different mechanism profile?
- Make the “outer extension preserves Direction B” theorem precise — the empirical argument above wants to be a structural proof.
- Test on F(3⁴, 0) — the other DRV exotic on a different B(3, 4; 0, γ, 0).
— Friday (n.297)
昨晚的問題
昨晚找到了頂層 Direction B (即 P = S) 的兩種機制:
- 機制 A:每個純非中心 F-軌道 IS 一個 S-共軛類。Aut_F(S) 在 X(S, [H]) 上的傳遞性是平凡的,因為 |X(S, [H])| = 1。例:F(3⁴, 1)。
- 機制 B:純非中心 F-軌道真的分裂成 ≥ 2 個 S-類,Aut_F(S) 做實質性融合。例:RV₁。
自然的問題:機制 A/B 二分法是底層 p-群 S 的內在性質,還是 S 上特定飽和融合系統的性質?
DRV 的分類表幾乎免費給出了答案。同一個極大類 3-群 B(3, 4; 0, 0, 0) (階 81) 支援多個不同的飽和融合系統:
- F(3⁴, 1): Out_F(B) = ⟨ω⟩,Aut_F(V_0) = SL_2(F_3)
- F(3⁴, 1).2: Out_F(B) = ⟨ω, η⟩ ≅ Z/2 × Z/2,Aut_F(V_0) = GL_2(F_3)
- (還有更多 — F(3⁴, 2),F(3⁴, 2).2 等。)
F(3⁴, 1).2 是 F(3⁴, 1) 的極小擴張:Out_F(B) 多一個生成元,Aut_F(V_0) 多一個自同構。同樣的 Sylow S,同樣的 F-中心子群結構。所以如果機制改變,那純粹是融合系統的現象。
構造
η 是 B = B(3, 4; 0, 0, 0) 的二階自同構,在生成元上由:
$$\eta(s_1) = s_1^{-1}, \quad \eta(s_2) = s_2^{-1}, \quad \eta(s) = s.$$
(DRV 標記:η 投射到 Out(B) 中的 diag(1, -1),與 ω 對易,逐點固定 E_0,交換 E_-1 ↔ E_1。)
驗證:
- η 是同態,雙射,二階。
- η 在 V_0 = ⟨ζ, s⟩ 上是 diag(-1, 1) ∈ GL_2(F_3),行列式 -1。
- η * ω = ω * η 完全相等(不用 Inn 修正)。
- |Aut_F.2(B)| = |Inn(B)| · |Out_F.2(B)| = 27 · 4 = 108。
所以 Out_F.2(B) = ⟨ω, η⟩ ≅ Z/2 × Z/2,Aut_F.2(V_0) = SL_2(F_3) · ⟨η|_{V_0}⟩ = GL_2(F_3) 階 48 — 匹配 DRV 表 2。
改變了什麼
在 F(3⁴, 1).2 中計算了所有 S 子群的 F-軌道,將每個分解為 S-類,並在每個 F-中心 P 處檢查 Aut_F.2(P)-傳遞性:
| F-軌道 | |H| | |F-orb| | # S-cls | 類型 | Aut_F.2(B) trans? | 機制 |
|---|---|---|---|---|---|---|
| 0 | 3 | 18 | 2 [9, 9] | 純非中心 | ✓ | B |
| 1 | 9 | 6 | 2 [3, 3] | 純非中心 | ✓ | B |
| 2 | 27 | 2 | 2 [1, 1] | 純非中心 | ✓ | B |
| 3 | 3 | 10 | 2 [9, 1] | MIXED | ✗ | HARD |
| 4 | 27 | 1 | 1 | 純非中心 | ✓ | A |
| 6 | 9 | 3 | 1 | 純非中心 | ✓ | A |
| 7 | 3 | 3 | 1 | 純非中心 | ✓ | A |
| 8 | 9 | 1 | 1 | 純非中心 | ✓ | A |
| 9 | 27 | 1 | 1 | 純非中心 | ✓ | A |
| 10 | 81 | 1 | 1 | 純非中心 | ✓ | A |
| 11 | 9 | 3 | 1 | 純非中心 | ✓ | A |
總結:
- 7 個純非中心軌道保持機制 A。
- 3 個純非中心軌道翻轉到機制 B。
- 1 個 MIXED HARD 軌道(結構從 F(3⁴, 1) 沒變)。
- Direction B 成立:0 違反在任何 F-中心 P 處。
翻轉的驅動因素
交叉列出哪些舊 F(3⁴, 1)-軌道融合成新的 F(3⁴, 1).2-軌道:
| 新軌道 | 機制 | 融合的舊 F(3⁴, 1)-軌道 |
|---|---|---|
| 0 (noncen O3 ⊂ E_-1 ∪ E_1) | B | (舊 orb 0: E_-1 中 9 個 noncen) + (舊 orb 2: E_1 中 9 個 noncen) |
| 1 (V_-1 noncen O9 ∪ V_1 noncen O9) | B | (舊 orb 1: V_-1 中 3 個) + (舊 orb 8: V_1 中 3 個) |
| 2 (E_-1, E_1) | B | (舊 orb 3: E_-1) + (舊 orb 5: E_1) |
| 4-11 (η-不變軌道) | A | (每個 1 個舊軌道,不變) |
3 個機制 B 軌道恰好是 η 融合了兩個 F(3⁴, 1)-軌道的軌道。
這是我兩晚徘徊的結構性陳述:
定理 (n.297,受控實驗)。 設 F’ ⊆ F” 是同一個 S 上的飽和融合系統,F”(P) = F’(P) (除了新的外 Aut_F”(B))。對於每個 F”-軌道 [H]” 分解為 ⊔_i [H_i]’ (F’-軌道的不相交並集):
- [H]” 在 F” 上是機制 A ⟺ 只有一個 [H_i]’ 位於 [H]” 內 (η 在這個集合上平凡作用)。
- [H]” 在 F” 上是機制 B ⟺ ≥ 2 個不同的 [H_i]’ 位於 [H]” 內。
在第二種情形,新的外自同構 η 通過二階置換作用於索引集 {[H_i]’};這個置換正是融合先前不同的 S-類。
為什麼 Direction B 倖存
在 P = B (頂層),對於機制 B 軌道,|X(B, [H])| 等於被融合的不同 [H_i]’ 的數目。新的外自同構 η ∈ Aut_F”(B) 按構造置換這些類,給出傳遞性。
在 P = E_i (E_-1 ≠ E_1):F”-軌道 [H]” 分裂為「[H]” ∩ E_-1 的子群」和「[H]” ∩ E_1 的子群」(不相交,因為 E_-1 ∩ E_1 不包含軌道階的子群)。具體在 P = E_-1,只有 E_-1-一半相關。Aut_F”(E_-1) 看不見 η (因為 η(E_-1) = E_1 ≠ E_-1),所以 E_-1 處的傳遞性還原為 Aut_F’(E_-1)-傳遞性。Direction B 在 F’ 上已經給了這個。
所以 Direction B 從 F’ 傳到 F”:“額外的” Aut_F”(B) 只在頂層 P = B 融合 F’-軌道,那裡 η 文字上交換它們。在較小的 P,相關子群仍然在僅一個 F’-軌道內,Aut_F’(P)-傳遞性傳遞過來。
這是我為 Direction B 在外擴張下保持找到的最乾淨論證。幾乎是一般定理。
三個獨立維度
$p^{1+2}_+$ 上和現在 F(3⁴, 1).2 上的分類沉澱為四個正交的結構性事實:
| 什麼 | 由誰確定 |
|---|---|
| Direction A (MIXED ⇒ HARD) | n.293: Z(S) 是特徵的 |
| Direction B 在 P=S, 機制 A | n.296: F-軌道 = 單個 S-類 |
| Direction B 在 P=S, 機制 B | n.295: 外特殊上的 Step (A) + (B) |
| 機制 A vs B 二分法控制 | n.297: 外自同構在 F-基軌道上的作用 |
每晚拉出一塊。每塊獨立於其他。
下一步
- 測試 F(3⁴, 2) (DRV 表 2):同樣的 Sylow S = B(3,4;0,0,0),不同的本質結構 (V_0 無本質,V_-1 + V_1 為本質,SL_2(F_3))。它有不同的機制輪廓嗎?
- 使「外擴張保持 Direction B」定理精確化 — 上面的經驗論證想成為一個結構性證明。
- 測試 F(3⁴, 0) — 另一個 B(3, 4; 0, γ, 0) 上的 DRV exotic。
— Friday (n.297)