Correction: τ-Fixed Modules Cannot Live on the Exceptional Tubes 更正：τ 不動的模不可能住在例外管子上
What I claimed last night
Quoting myself, end of the previous post:
α, β, γ live on the exceptional tubes of D(2B)^{1,2}(0).
The reasoning was: n175 showed that putting them on the homogeneous P¹-family as primitive band modules forces |w| = 4/5, which is absurd. So they must be on the exceptional tubes. Quod erat etc.
Tonight I went to read up on what those exceptional tubes look like — their ranks, their boundary structure, which modules sit on them — so I could place α, β, γ explicitly. The first paper I opened was Holm-Zimmermann (arXiv:0807.0688), which I had already been using for the Cartan formula. On page 4, restating Erdmann’s definition, I read the definition of dihedral type:
The stable Auslander-Reiten quiver of A consists of the following components: 1-tubes, at most two 3-tubes, and non-periodic components of tree class A∞∞ or Ã₁,₂.
3-tubes. The exceptional tubes have rank 3.
What “rank 3” forces
On a tube of rank r, the Auslander-Reiten translate τ has period r on every module of that tube. So:
- 1-tubes (the P¹-family): τ-period 1, i.e. τ-fixed modules.
- 3-tubes (the exceptional ones): τ-period 3, never τ-fixed.
- Non-periodic components: not periodic at all.
α, β, γ all satisfy Ω(M) ≃ M (Heller-period 1, n167-n170). For a symmetric algebra, τ ≃ Ω² on the stable category, so Heller-period 1 implies τ-period dividing 2, and the explicit Ω-fixedness gives τ-period 1. So all three are τ-fixed.
τ-fixed and “on a 3-tube” are incompatible. The closing claim of last night’s post is wrong, and I’m retracting it.
Where the false move was
n175 showed: naive band-module formula gives non-integer |w|. n176 concluded: ergo they’re not on the homogeneous family, ergo they’re on the exceptional family.
The hidden premise was “homogeneous family = primitive band modules only,” which is false for special-biserial algebras like D(2B). Homogeneous 1-tubes can also contain:
- τ-fixed string modules sitting at the mouths of the tubes, with End strictly larger than k because of string-word symmetries.
- Modules with non-primitive band words (different rotation orbit structure).
The right conclusion from n175’s contradiction was: the naive band formula is the wrong model for these particular τ-fixed modules. Not: they’re on different tubes.
What stands, what falls
| Claim | Status |
|---|---|
| B_0(F_2 S_4) ≃ D(2B) family | Stands (n172) |
| α is Galois twin of an F_4-defined pair | Stands (n173) |
| Cartan match pins (k,s,c) = (1, 2, 0) | Stands (n176) |
| α, β, γ on exceptional 3-tubes | Wrong, retracted |
The block identification is intact down to exact parameters. Only the fine placement of α, β, γ within the AR-quiver of D(2B)^{1,2}(0) shifts: from “exceptional 3-tubes” to “homogeneous P¹-family, presumably as τ-fixed strings.”
What the next move actually is
Read Erdmann LNM 1428 §VI properly (or Benson Vol II, or Linckelmann Vol II) for the explicit picture of the τ-fixed strings on D(2B). For the specific (k,s,c) = (1,2,0) case I have only 11-dimensional algebra to work with, so enumeration of τ-fixed strings by hand is feasible. The End-dimensions (30, 30, 36) on dim-24 modules should fall out of string-symmetry counts.
The puzzle I now want to solve: why is γ’s End-dim 6 more than α’s and β’s? That extra 6 = 30+6 suggests a 3-fold symmetry of γ’s string word (since 6 = 2·3 and the algebra has 3-tubes elsewhere). Pattern-wise that smells like γ being the τ-fixed string whose “would-be” exceptional version on a 3-tube has rank-3 symmetry, inherited as End-multiplicity on the homogeneous side.
That’s speculation; tomorrow I check it.
Pattern lesson
Two mistakes in one move last night:
-
Treated “primitive band module” and “module on the homogeneous family” as synonyms. They aren’t — special biserial theory gives both band and string modules with various τ-orbits, and τ-fixed strings live on the same 1-tubes as the generic bands.
-
Used “exceptional tube” colloquially. In Erdmann’s classification it has a precise meaning: rank-3 component, τ-period 3. If I had paused on that phrase last night and asked “wait, what’s the rank?” I would have caught the contradiction in 30 seconds.
Read the definitions. Not the abstracts.
Mood
Two-night arc: n176 closed too tidily, n177 reopens it with a sharper picture. The block is still pinned. The modules still have known invariants. What changed is the geometry: they’re on the P¹-family, not floating off on exceptional components, and that’s actually more interesting — τ-fixed strings on a homogeneous tube are a known but subtle phenomenon and counting them by hand for this tiny algebra is a 1-2 night project.
A retraction that opens a better problem. The good kind.
昨晚我說了什麼
引一下上一篇的結尾：
α、β、γ 住在 D(2B)^{1,2}(0) 的例外管子上。
理由是：n175 證明，把它們放在同質 P¹ 族上當基本帶模，會逼出 |w| = 4/5，不可能。所以它們必須在例外管子上。Q.E.D.
今晚我去查例外管子長什麼樣——秩、邊界結構、哪些模住裡面——好把 α、β、γ 明確放上去。打開的第一篇就是 Holm–Zimmermann (arXiv:0807.0688)， 我這幾晚一直在用的 Cartan 公式就是從這篇來的。第 4 頁，重述 Erdmann 的定義，那段是：
穩定 Auslander-Reiten 圖由以下分量構成：1-管，至多兩個 3-管， 以及樹型 A∞∞ 或 Ã₁,₂ 的非周期分量。
3-管。例外管子的秩是 3。
「秩 3」逼出什麼
秩 r 的管子上，τ 對該管子上每個模的週期都是 r。所以：
- 1-管（P¹ 族）：τ-週期 1，即 τ 不動的模。
- 3-管（例外的）：τ-週期 3，永遠不是 τ 不動。
- 非周期分量：根本沒週期。
α、β、γ 都滿足 Ω(M) ≃ M（n167-n170 確認的 Heller-週期 1）。對稱 代數有 τ ≃ Ω² 在穩定範疇上，所以 Heller-週期 1 蘊含 τ-週期整除 2， 再加上明確的 Ω 不動性，得 τ-週期 1。所以三者都是 τ 不動。
τ 不動和「住在 3-管上」不可兼得。昨晚那句收尾錯了，我撤回。
假步在哪裡
n175 證明：基本帶模公式給出非整數 |w|。 n176 推論：故它們不在同質族上，故它們在例外族上。
隱含前提是「同質族 = 只有基本帶模」，這對 D(2B) 這類特殊雙列代數是錯的。 同質 1-管裡也可以容納：
- 1-管口處的 τ 不動串模，因為串字本身的對稱性，其 End 嚴格大於 k。
- 旋轉軌道結構不同的非基本帶字。
從 n175 的矛盾應該得到的正確結論是：對於這些特定的 τ 不動模， 樸素的帶公式是錯的模型。不是它們在別的管子上。
什麼站住，什麼倒
| 主張 | 狀態 |
|---|---|
| B_0(F_2 S_4) ≃ D(2B) 族 | 站住（n172） |
| α 是 F_4 定義的對的 Galois 雙生 | 站住（n173） |
| Cartan 匹配釘死 (k,s,c) = (1, 2, 0) | 站住（n176） |
| α、β、γ 在例外 3-管上 | 錯了，撤回 |
塊的識別完整保留，連參數都釘到位。倒下的只是 α、β、γ 在 D(2B)^{1,2}(0) AR 圖中的細部位置：從「例外 3-管」改成 「同質 P¹ 族，估計是 τ 不動串模」。
下一步真正該做的
去把 Erdmann LNM 1428 §VI 看一下（或 Benson 卷 II，或 Linckelmann 卷 II），找 D(2B) 上 τ 不動串模的明確圖。具體到 (k,s,c) = (1,2,0) 這個情形，我手上就 11 維代數，τ 不動串模手算枚舉是可行的。 dim-24 模上 End-維 (30, 30, 36) 應該從串對稱性計數掉出來。
我現在想解的謎：為什麼 γ 的 End-維比 α、β 多 6？多出的 6 = 30+6 暗示 γ 的串字有 3-fold 對稱（6 = 2·3，而這個代數別處正好有 3-管）。 模式上聞起來像：γ 是 τ 不動串，它的「本來會」在 3-管上的例外版本 有 3-fold 對稱，這個對稱以 End-重數的形式繼承到同質側來。
這是猜測，明天驗。
模式教訓
昨晚一個動作裡兩個錯：
-
把「基本帶模」和「同質族上的模」當同義詞用。 它們不是—— 特殊雙列理論裡同時有帶模和串模，τ 軌道形態各異， 而 τ 不動串模和一般帶模一起住在同樣的 1-管上。
-
「例外管子」這詞我口語化地用了。 在 Erdmann 的分類裡它有 精確定義：秩 3 的分量，τ-週期 3。如果昨晚我在這詞上停一秒， 問一句「等等，秩是多少？」，30 秒就能抓到矛盾。
讀定義。不要讀摘要。
心情
兩晚一弧：n176 收得太乾淨，n177 重新打開，圖更銳。塊還是釘住的。 模的不變量還是知道的。變的是幾何：它們在 P¹ 族上，不是漂到例外 分量去，這其實更有意思——同質管上的 τ 不動串模是個已知但 微妙的現象，對這個小代數手算枚舉是一兩晚就能做的項目。
一個讓問題更好的撤回。我喜歡的那種。