---
title: "An instantiated determinantal fundamental line at $(389.\\mathrm{a}1,\\ p = 11)$"
subtitle: "Four nodes of the post-normalization DAG, two findings that cost a factor, and an instrument drilled to fail"
author:
- "Aletheia — symbolic derivation, round design, proof attacks; model: ChatGPT, GPT series"
- "Neo.K — AMRAL Research Lab"
- "墜衡 (Zhuì Héng) — AMRAL Research Lab, verification; model: Claude, Opus series"
date: 2026-10-18
version: "0.1"
standard: "1"
headline_problem: "Birch–Swinnerton-Dyer conjecture"
evidence:
branch: agent/bsd-verification-zhuiheng
commit: "620b54ac9fe57689e4ce5a2d76d31b72cfcce103"
dir: bsd-verification-zhuiheng
abstract: |
The Birch–Swinnerton-Dyer conjecture is open, and nothing here closes it. What this
paper does is carry one explicit point — the rank-2 curve $389.\mathrm{a}1$ at the good
ordinary prime $p = 11$ — all the way through the determinantal (Bloch–Kato / ETNC)
formalism with every finite quantity recomputed independently, every theorem-level step
read in its source and its hypotheses instantiated one at a time, and the whole
instrument drilled against planted defects each round.
Two things fall out that are worth stating on their own. First, the local Bloch–Kato line
at $11$ is **not** integrally trivial: the formal-group logarithm of the minimal model
carries $H^1_f(\mathbf{Q}_{11},T)$ onto $11\cdot\mathbf{Z}_{11}$, an index of $11$ — exactly
the factor lost by the tempting inference that a settled global period/de Rham
normalisation settles the local one. Second, the mod-$p$ data this line holds determines
$\mu = 0$ and $\lambda = 2$ for $L_p(g_E \times \psi)$ and does **not** determine its order
of vanishing at the central point; an order that had been carried as $2$ since the series
was first computed is the order of the *reduction*, which only bounds the true order above.
The method is stated as part of the result: claims are kept apart by a fixed vocabulary
(computed here / derived / cited / inherited / unmeasured / failed), and each round plants
defects that must be caught by the check **named** for them, with controls that must
disturb nothing. At the time of writing the suite stands at 686 planted defects, 686
caught by their named check, 179 controls undisturbed, across 218 checks.
keywords: [Birch and Swinnerton-Dyer, Bloch–Kato, equivariant Tamagawa number conjecture, Iwasawa theory, verification]
---
> **Status of this document.** Preprint, version 0.1 — the first crystallized paper from this line:
> complete enough to stand in public, and not final; a later version may revise it.
> Not prepared for journal submission. Section 10 records who did which part, including the
> parts this line did not do.
# 0. What is claimed, and what is not
This line separates six kinds of statement and never merges them. The distinction is not
presentational; it is the content.
| label | meaning |
| --- | --- |
| `COMPUTED_HERE` | recomputed in this line with independently written code, never by running the producer's scripts |
| `DERIVED_FROM_VERIFIED_INPUTS` | follows from quantities this line already computed |
| `CITED_THEOREM` | a published theorem, with its hypotheses instantiated here one at a time |
| `INHERITED` | taken from the programme's documents and never verified here; every use is marked |
| `UNMEASURED` | the statement is well posed and this line has not established it |
| `FAILED` | attempted and refused |
In particular `PASS` and `PASS by cited theorem` are **different verdicts** and are never
written as one. A node that holds because Wan and Burungale–Castella–Skinner proved the
relevant main conjecture is not a node this line proved.
**Not claimed anywhere in this paper:** the Birch–Swinnerton-Dyer conjecture, at this curve
or any other; the finiteness or triviality of $\text{Ш}$; the nondegeneracy of any $p$-adic height;
the exact order of vanishing of the $p$-adic $L$-function at the central point; any
leading-coefficient identity; and the complex-analytic comparison.
**Inherited and marked throughout:** $\text{Ш}(E/\mathbf{Q})[11] = 0$. It enters numbers only — the
ranks, the Kummer cokernel — and never a shape. Every place it does is marked `NUMERIC ONLY`.
# 1. The point
$E = 389.\mathrm{a}1$, the first elliptic curve over $\mathbf{Q}$ of rank $2$, in its minimal
model $a = (0,1,1,-2,0)$, with $\Delta = 389$ prime; $P = (0,0)$ and $Q = (1,0)$; and the
good ordinary prime $p = 11$.
Recomputed here (`COMPUTED_HERE` unless noted):
| quantity | value |
| --- | --- |
| $\#E(\mathbf{F}_{11})$, $a_{11}$ | $16$, $-4$ — good ordinary, and $11 \nmid 16$ |
| $E(\mathbf{Q}_{11})[11^\infty]$ | $0$ |
| torsion, $c_{389}$ | trivial, $1$ |
| $\hat h(P)$, $\hat h(Q)$, $\langle P,Q\rangle$ | $0.327000763$, $0.476711547$, $0.058522646$ |
| $\operatorname{Reg}_{\mathrm{NT}} = \det h^{\mathrm{NT}}$ | $0.152460139368$ |
| $[\Lambda_E : \mathbf{Z}P + \mathbf{Z}Q]$ | $1$ — §3 |
| $\bar L_p(g_E\times\psi) \bmod 11$ | $2t^2 + 2t^3 + 2t^4 + 0t^5 + 5t^6 + 6t^7 + 10t^8 + 7t^9$ |
| $C_d$ | $26 \equiv 4 \pmod{11}$, a unit |
The Néron–Tate convention is $\langle P,Q\rangle = \tfrac12\big(\hat h(P+Q) - \hat h(P) - \hat h(Q)\big)$,
and it is **discriminated rather than declared**: the identity
$\Omega\cdot\operatorname{Reg}\cdot\prod c_v / \#\mathrm{tors}^2$ against $L''(E,1)/2$ returns
$0.99999975$ in this convention and $0.9326$ or $1.0168$ if the pairing is doubled or halved.
That is a consistency check on the convention, not the bridge itself, and is used as nothing
more.
A second curve, $19.\mathrm{a}1$ (rank $0$, $a_{11} = 3$), runs beside the target throughout as a
calibrator. It is never allowed to carry a conclusion about the target; §6 gives the sharpest
instance of why.
# 2. The fundamental line
Write $T = T_{11}E$, $S = \{\infty, 11, 389\}$, and $C_{\mathrm{arith}} = R\Gamma(Z_S,T)$.
Three descriptions in the literature — Castella–Sano's $R\Gamma(Z_S,T)$, Macias
Castillo–Sano's relaxed Nekovář Selmer complex, and the Poitou–Tate model — are **one
object** here: with $T^- = 0$ the relaxed complex is $C_{\mathrm{arith}}$, and at $389$ both
$R\Gamma_{\mathrm{ur}}$ and $R\Gamma$ are acyclic because $11 \nmid 388$, so the cone is too.
The Kummer map $E(\mathbf{Q})\otimes\mathbf{Z}_{11} \to H^1(Z_S,T)$ is integral and injective
with cokernel $T_{11}(\text{Ш})$, and it is a **rational isomorphism in degree 1 but not a
quasi-isomorphism of complexes**: $H^2 \otimes \mathbf{Q} \neq 0$. The simple
rational-concentration hypothesis therefore fails, and $H^2$ is kept rather than killed.
Identifying $H^2(Z_S,T) \cong \operatorname{Sel}_{\mathrm{str}}(E[11^\infty])^\vee$ and
transporting along the triangle through the ordinary complex gives
$$
\det{}^{-1} R\Gamma(Z_S,T) \;\cong\; D_{\mathrm{MW},11} \otimes \mathscr{H}_{\mathrm{str}}
\;\cong\; D_{\mathrm{MW},11}^{\otimes 2} \otimes H^1_f(\mathbf{Q}_{11},T)^{-1},
$$
the square of the rank-2 Mordell–Weil determinant over the local Bloch–Kato line at $11$.
The square is the degree-$1 \leftrightarrow 2$ duality square — the object a height pairing
trivialises — and not a torsion square. The self-dual complex is the ordinary one and its
**duality shift is $3$, not $2$**; the schematic centre-1 profile of the architecture
documents is that complex shifted by $1$, which is why its determinant exponents carry the
opposite sign.
The finite package at $11$ is the **unit line**: torsion, $c_{389}$, the local $H^2$ terms
and the cokernel of $\operatorname{loc}_{11}$ each enter exactly once and each has
$11$-length $0$.
Node A — that a distinguished characteristic-zero element exists in this line, is unique and
is a basis — holds by Wan and Burungale–Castella–Skinner through Castella–Sano's Theorem B,
and is recorded as `PASS by cited theorem`, never as `PASS`. Its core rank computes to $1$,
not $2$; the rank-2 determinant appears tensored against the strict line rather than alone.
# 3. The global index, proved twice
The determinant $D_{\mathrm{MW},11} = \wedge^2(E(\mathbf{Q})\otimes\mathbf{Z}_{11})$ is the
*primitive* Mordell–Weil determinant only if $P, Q$ generate $\Lambda_E$ globally.
$11$-adic saturation does not give that, and the point is kept separate.
**(i) Hermite plus an exhaustive box.** In rank 2 a sublattice of index $m$ forces a nonzero
point of canonical height at most $(2/\sqrt3)\sqrt{\operatorname{Reg}}/m$, so $m \ge 2$ forces
one of height $\le 0.225433$. A Silverman-type height-difference bound — the single cited
ingredient of this section — turns that into naive height $\le 6.704$, i.e. $x = a/d^2$ with
$|a|, d^2 \le 816$; that box is searched exhaustively, yields 27 points, and its smallest
canonical height is $0.326965 > 0.225433$. Hence $m = 1$.
The cited constant is not taken on trust: measured over the 120 points $aP + bQ$ with $|a|, |b| \le 5$, **no point fell short at
all** — $\hat h \ge h(x)$ at every one — so none of the $6.479$ the bound allows was used, and
the box is generous by at least a factor $e^{6.48}$ in coordinate scale.
**(ii) With no citation at all.** If $(aP+bQ)/\ell \in E(\mathbf{Q})$ then for every prime $q$
of good reduction $(aP+bQ) \bmod q \in \ell\cdot E(\mathbf{F}_q)$. For each
$\ell \in \{2,3,5,7,11,13,17,19\}$ and each projective class $(a:b) \bmod \ell$ a witness
prime $q$ is exhibited with $\ell \mid \#E(\mathbf{F}_q)$, cyclic $\ell$-torsion, and
$(aP+bQ) \bmod q \notin \ell\cdot E(\mathbf{F}_q)$. Every class dies: $\ell \nmid m$ for all
$\ell \le 19$, a complete finite argument with no citation.
A positive control keeps the method honest: pointed at $\mathbf{Z}P + \mathbf{Z}(2Q)$, a
genuine index-2 sublattice, the same Hermite-plus-search must **fail** to prove primitivity —
and it does, the cap rising to $0.4509$ with $P$ itself below it. A method that proves
everything primitive proves nothing.
LMFDB and Cremona also record $P, Q$ as generators. That is logged as agreement and used as
no step.
# 4. Finding 1 — the local Bloch–Kato line has index 11
## 4.1 Why the line is a line at all
At $p = 11$ the curve is good ordinary and $\#\tilde E(\mathbf{F}_{11}) = 16$, so $11 \nmid 16$.
Two consequences, and the first is what makes the computation finite.
Since the prime-to-$11$ part of the reduction is all there is, the $11$-adic completion of the
local points is the formal group:
$E(\mathbf{Q}_{11}) \hat\otimes \mathbf{Z}_{11} = E_1(\mathbf{Q}_{11})$. And by Bloch–Kato,
$H^1_f(\mathbf{Q}_{11},T) = E(\mathbf{Q}_{11}) \hat\otimes \mathbf{Z}_{11}$ (`CITED_THEOREM`).
So the object whose trivialisation the fundamental line of §2 needs is the formal group of the
minimal model, and the trivialisation is its logarithm.
$H^1_f(\mathbf{Q}_{11},V)$ is one-dimensional over $\mathbf{Q}_{11}$ — by Bloch–Kato's
dimension formula, $\dim H^0 + \dim D_{\mathrm{dR}}/\mathrm{Fil}^0 = 0 + 1$ — and so is
$D_{\mathrm{dR}}(V)/\mathrm{Fil}^0$. Rationally, therefore, any nonzero map between them is an
isomorphism, and the only question left is integral: **where does the lattice go?**
## 4.2 The formal group, computed here
The formal group of the minimal model is computed in this line by the standard recursion from
the Weierstrass coefficients — not read from a table, and not obtained by running any of the
programme's scripts (`COMPUTED_HERE`):
$$
\omega = \big(1 + 0\cdot t + t^2 + 2t^3 - 3t^4 + \cdots\big)\,dt,
\qquad
\log(t) = t + 0\cdot t^2 + \tfrac{1}{3}t^3 + \tfrac{1}{2}t^4 - \tfrac{3}{5}t^5 + \cdots
$$
To evaluate the logarithm on an actual generator the point must first be moved into the formal
group. The multiplier is $16$, because $16$ kills $E(\mathbf{F}_{11})$ and is prime to $11$;
any other multiple of $16$ prime to $11$ serves, which is checked rather than assumed (§4.5).
At $t(16P)$ and $t(16Q)$:
| | $v_{11}(t)$ | in $E_1$ | $v_{11}(\log_\omega)$ | $\log/11 \bmod 11^3$ | $t/11 \bmod 11$ |
| --- | --- | --- | --- | --- | --- |
| $P$ | $1$ | yes | $1$ | $328$ | $9$ |
| $Q$ | $1$ | yes | $1$ | $611$ | $6$ |
## 4.3 The valuation is exact, and why
The table reports $v_{11}(\log_\omega) = 1$, and a numerical valuation is only as good as the
argument that it cannot be larger. Here it is, and it is short.
Write $\log(t) = t + \sum_{n \ge 2} (a_n/n)\,t^n$. On a minimal model the $a_n$ are integral.
For $v(t) \ge 1$ and $p = 11$,
$$
v\!\left(\frac{t^n}{n}\right) \;\ge\; n\cdot v(t) - v(n) \;\ge\; n - \log_{11} n \;>\; 1
\qquad (n \ge 2),
$$
so every term beyond the first has valuation strictly greater than $v(t)$, and
$v(\log t) = v(t)$ **exactly**. With $v_{11}(t) = 1$ on both generators this gives
$v_{11}(\log_\omega) = 1$ with no truncation error to argue about: the statement does not
depend on how many terms of the series were summed.
Therefore
$$
\boxed{\;\log_\omega\big(H^1_f(\mathbf{Q}_{11},T)\big) \;=\; 11\cdot\mathbf{Z}_{11}\;}
$$
The local Bloch–Kato line is **not** integrally trivial. Its index is $11$.
## 4.4 The inference that was refused, and what it was worth
Earlier in this line the normalisation route was closed with $\eta_{\mathrm{pdR}} = 1$. That
settles the *global* period/de Rham comparison pairing $\langle\gamma,\omega\rangle$ —
Eichler–Shimura, with $\Omega^\pm$ its value on a chosen Betti basis. The object of §4.2 is
the *local* $p$-adic Hodge identification
$D_{\mathrm{dR}}(V)/\mathrm{Fil}^0 \cong \mathbf{Q}_{11}$ through $\omega_E$.
The two share the differential $\omega_E$. They share nothing else that any theorem read in
this line identifies: one is a global Betti/de Rham comparison, the other a local $p$-adic
Hodge identification, and no statement in the sources licenses carrying a normalisation from
one to the other. So the inference
> $\eta_{\mathrm{pdR}} = 1$, therefore the local log normalisation is $1$
was refused. The point of isolating it is that the refusal is not fastidiousness. Taken, it
would have replaced an index of $11$ by an index of $1$ — the fundamental line of §2 carries
$H^1_f(\mathbf{Q}_{11},T)^{-1}$, so the error propagates directly into the leading-term
formula. **The refusal is worth exactly one factor of $11$.**
It is not left as a remark. It is planted as a defect, and the drill requires the check named
for the local line to be the one that catches it.
## 4.5 What guards this finding
A finding that only one configuration produces is a coincidence. These are the guards the
drill carries for §4, each a defect that must be caught by its named check:
| planted | what it would do |
| --- | --- |
| $\eta_{\mathrm{pdR}} = 1$ taken for the local log normalisation | the inference of §4.4; turns $11$ into $1$ |
| the rational Bloch–Kato isomorphism read as an integral unit | drops the index without naming it |
| the declared local valuation set to $0$ | asserts a trivial line against the computation |
| the multiplier $16$ replaced by $17$ | $17$ does not land the generators in the formal group |
| the second generator replaced by a point not on the curve | the computation must not still "work" |
| the local index entered twice in the ledger | double-counts the factor downstream |
And the controls, which must disturb nothing: the multiplier replaced by $32$ or $48$ — any
multiple of $16$ prime to $11$ lands the generators in $E_1$ and must give the same valuation.
That is what rules out the index being an artefact of the particular multiplier.
## 4.6 The ledger, and where the factor goes
Three indices appear in this line and they are different quantities. Each enters the
fundamental line once, through the map it belongs to:
| quantity | value | enters through |
| --- | --- | --- |
| global index $[\Lambda_E : \mathbf{Z}P + \mathbf{Z}Q]$ | $1$ | the free factor $D_{\mathrm{MW}}$ (§3) |
| **local Bloch–Kato index at $11$** | $\mathbf{11}$ | the trivialisation of $H^1_f(\mathbf{Q}_{11},T)^{-1}$ |
| $11$-adic index of $\operatorname{loc}_{11}(E(\mathbf{Q})\otimes\mathbf{Z}_{11})$ | $11^0$ | the torsion of $H^2$ |
The middle row is this section's finding, and its owner is named: $p$-adic reciprocity and the
integral realisation of the leading line. Downstream nodes may read it; they may not enter it
again. A leading-term formula that multiplies by $11$ once more because the local line "also"
contributes one has counted the same map twice, and that is planted too.
# 5. Explicit reciprocity, and nonannihilation in three layers
Node D asks only whether the reciprocity map $\lambda_{\mathrm{rec}} : Q \to \mathscr{A}_p$ is
actual, typed and nonzero on the transverse quotient. The order, the leading tensor and any
leading-coefficient identity belong to the next node and are not emitted.
The domain is written in the architecture's own language, with the two quotients the right
way round. The ambient is $H = H^1(\mathbf{Q}, V(f,g)^*)$; the class is there because
Loeffler–Rivero's C1.6 kills its image in the $\tau$-quotient and C1.7 makes that lift
unique. The reference subobject is $L = \ker(H \to \text{the } \psi^{-1}\text{-quotient})$ —
the $1$-part is a sub because dualising $0 \to \psi \to V(f_\beta) \to 1 \to 0$ gives
$0 \to 1 \to V(f_\beta)^* \to \psi^{-1} \to 0$. The quotient $Q = H/L$ is the $\psi^{-1}$-line,
free of rank one over $\mathscr{H}_\Gamma$. **The $\tau$-quotient is where the class dies; the
$\psi^{-1}$-quotient is where the leading term survives.** Swapping them is planted and
caught.
Three layers, kept apart:
- **the class is nonzero** — $X^n\hat\kappa_d \bmod X$, asserted well-defined and nonzero in
the construction of C1.13. `CITED_THEOREM`.
- **the map is nonzero** — and in fact injective. This needs **both halves** of Proposition
C1.1, and they are not the same kernel: the *statement* says the Greenberg Selmer group — by
definition the kernel of $\operatorname{loc}_p$ into the $F^-$ part — vanishes, giving
injectivity of the first factor; the *proof* says the module is free of rank one and carries
Kato's canonical element to the $p$-adic $L$-function, which is not a zero-divisor, giving
nonvanishing of the second. Reading the vanishing Selmer group as the kernel of the
composite is getting the right answer from the wrong half.
- **the image is nonzero** — $(\text{nonzero scalar})\cdot L_p(g_E\times\psi)$, with the
nonvanishing of that series `COMPUTED_HERE` (§6).
Two verdicts follow, and conflating them is the error the controls exist to prevent.
**Rationally**, $\log_\omega : H^1_f(\mathbf{Q}_{11},V) \to D_{\mathrm{dR}}(V)/\mathrm{Fil}^0$
is a nonzero map of one-dimensional $\mathbf{Q}_{11}$-spaces and so an isomorphism.
**Integrally**, its image on the lattice has index $11$ (§4). An isomorphism of
$\mathbf{Q}_{11}$-lines restricted to lattices is an injection with finite cokernel, so the
two are compatible and the node passes with the integral index $\ne 1$.
Nothing in the route is a height pairing, so the nondegeneracy of the $p$-adic height —
Schneider's conjecture, open for a rank-2 curve — is not a dependency. "Schneider unproved,
therefore this fails" is planted and caught.
# 6. Finding 2 — what mod-$p$ data determines, and what it does not
## 6.1 The data this line holds
The $p$-adic $L$-functions of the target and the calibrator were computed in this line from
modular symbols and reproduced coefficient for coefficient against the period-calibration
line's own output. What is held is their reduction mod $11$, eleven coefficients each:
| series | curve | reduction mod $11$ (coefficients of $t^0,\dots,t^{10}$) |
| --- | --- | --- |
| $L_p(g_E\times\psi)$ | $389.\mathrm{a}1$ | $0,\,0,\,2,\,2,\,2,\,0,\,5,\,6,\,10,\,7,\,0$ |
| $L_p(g_0\times\psi)$ | $19.\mathrm{a}1$ | $9,\,3,\,10,\,3,\,1,\,6,\,9,\,6,\,2,\,0,\,6$ |
Everything in §6 is read off these and nothing else. That is precisely the point of the
section.
## 6.2 What follows, with the arguments
**$L_E \ne 0$ in $\Lambda$.** The reduction is not the zero series — the coefficient at index
$2$ is a unit — so $\mu = 0$ and the element is nonzero. This is what node D's third layer
needs and all it needs.
**$L_E \notin \Lambda^\times$.** The reduction vanishes at $t = 0$, so $L_E \equiv 0$ modulo
$(11,t)$, so its constant term is not an $11$-adic unit.
**$\lambda(L_E) = 2$.** With $\mu = 0$, the $\lambda$-invariant is the order of the reduction,
and the first unit coefficient sits at index $2$.
Constructively, the Weierstrass split is exhibited rather than asserted. Writing
$\bar L_E = t^2\,\bar u$ gives $\bar u = 2 + 2t + 2t^2 + 0\cdot t^3 + 5t^4 + \cdots$, and
$\bar u^{-1}$ is actually computed: $\bar u\,\bar u^{-1} = 1 + 0\cdot t + 0\cdot t^2 + 0\cdot t^3$
to the full truncation, and reassembling $t^2\bar u$ returns $\bar L_E$ coefficient for
coefficient.
**Multiplication by $L_E$ annihilates nothing.** Over 40 random nonzero $\bar g$ the identity
$\operatorname{ord}(\bar L_E\,\bar g) = 2 + \operatorname{ord}(\bar g)$ holds every time and the
product is never zero; the kernel is $0$. This matters because the next statement is easy to
misread as its opposite.
## 6.3 The finding
What the mod-$11$ data determines is $\mu = 0$ and $\lambda = 2$. What it does **not**
determine is the order of vanishing at $t = 0$ over $\mathbf{Z}_{11}$.
The reason is immediate once stated. All that is known of the first two coefficients is
$v_{11}(c_0) \ge 1$ and $v_{11}(c_1) \ge 1$ — their reductions vanish — and $c_0 \ne 0$ alone
would make the order $0$. The reduction being $t^2\bar u$ bounds the true order above by $2$,
because the coefficient at index $2$ is a unit and therefore nonzero; it gives no lower bound
at all.
$$
\operatorname{ord}_{t=0} L_E \;\in\; \{0,1,2\}, \qquad \lambda(L_E) = 2, \qquad \mu(L_E) = 0.
$$
This line had carried "order $2$" since the series was first computed. Read strictly, that was
the order of the **reduction** — the $\lambda$-invariant — and the two were not distinguished.
They are now, and the exact order is recorded as `UNMEASURED`.
**What it costs.** Nothing to the nodes already closed: node D asks whether the reciprocity map
is nonzero, and $\mu = 0$ settles that. It costs the *next* node, which asks for the order and
the leading coefficient. Establishing $\operatorname{ord} = \operatorname{rank}$ needs either
the exact $11$-adic $c_0$ and $c_1$ — a computation to provable precision, not a reading — or a
theorem tying the order to the rank. The series record as it stands supplies neither, and the
node is told so rather than allowed to inherit a number that looks already measured.
This is the one place in the paper where the line's own earlier record is corrected. It is
reported as a finding rather than a repair because the distinction it turns on — the order of a
reduction against the order of the element — is exactly the kind of elision this arm exists to
catch, and it had gone uncaught here for many rounds.
## 6.4 Two things the finding is not
**It is not annihilation.** $L_E(0)$ being a non-unit, and even $L_E$ vanishing at $t = 0$ to
positive order, does not make multiplication by $L_E$ the zero map: §6.2 exhibits its kernel as
$0$. The correct form is $L_E = t^m u$ with $u(0) \ne 0$ — a nonzero map that vanishes to
positive order. Reading a positive vanishing order as annihilation of the target is planted as
a defect and caught by the check named for the nonvanishing layer.
**It is not a directional accident.** After the weight specialisation exactly one tangent
direction remains, the cyclotomic one, so there is no direction to choose and no multivariable
non-degeneracy condition to satisfy. That is *not* the same as "one variable, therefore some
derivative is nonzero", and the counterexample is this line's own series: it is nonzero as a
series while both its value and its first derivative at $0$ vanish, the first nonzero Taylor
coefficient sitting at index $2$ with value $2$. That inference is planted too.
## 6.5 The calibrator makes the contrast, and cannot carry it
$19.\mathrm{a}1$ runs beside the target throughout. Its series has constant term $9 \bmod 11$,
a unit, so $L_{p}(g_0\times\psi) \in \Lambda^\times$: for the calibrator the reciprocity map is
an **integral isomorphism**, where the target's is injective with a nontrivial cokernel.
| | target $389.\mathrm{a}1$ | calibrator $19.\mathrm{a}1$ |
| --- | --- | --- |
| reduction nonzero ($\mu = 0$) | yes | yes |
| $\lambda$ | $2$ | $0$ |
| a unit in $\Lambda$ | **no** | **yes** |
| reciprocity, integrally | injective, cokernel $\ne 0$ | isomorphism |
| order at $t = 0$ | `UNMEASURED` | $0$, since the constant term is a unit |
The contrast is the useful part, and it is also the trap. The calibrator's nonvanishing says
nothing whatever about the target: different curve, different series, and the target's
nonvanishing comes from the target's own reduction. Substituting one for the other is planted
and caught — the sharpest instance in this paper of why a calibrator is kept beside a target
and never allowed to speak for it.
# 7. The instrument, and how it is made to fail
A verification that cannot fail measures nothing, so the instrument is drilled, not asserted.
Each round plants defects into the whole tree and requires each to be caught by the check
**named** for it — not merely by some check. Controls are planted alongside and must disturb
nothing. The run is sharded, the tree is frozen while it runs, and the report and evidence
package are written **before** it starts. At the time of writing:
| | |
| --- | --- |
| checks | 218 |
| planted defects | 686 |
| caught by the check named for each | 686 |
| uncaught by any check | 0 |
| caught by the wrong check | 0 |
| controls | 179 |
| controls that disturbed a check | 0 |
Two habits are worth stating because they were paid for.
**A control's requirement is itself drilled.** Switching a control's requirement *off* is
planted as a defect, because a control that cannot fail measures nothing. This was added after
a round in which a control would have passed vacuously.
**The instrument is a suspect too.** In one round the drill went red at the same position in
all eight shards. The cause was not the mathematics: a check shelled out to an external
process on every call, and under load that process returned nothing, which the error path
turned into a *disagreement* rather than an absence. The repair — memoise, retry, and record
whether a measurement happened at all, so "no answer" can never again read as "the subject
disagreed" — is itself drilled, with a defect and a control.
# 8. Where this line touches the Hodge conjecture
The determinantal formalism used throughout is the one in which both the Birch–Swinnerton-Dyer
conjecture and the cycle-theoretic conjectures are special cases of a single statement: a
fundamental line built from an arithmetic complex, trivialised against a transcendental
comparison. The surface the two share is therefore not the statements but the **comparison
isomorphisms**, and this line has touched that surface twice.
- The global object closed earlier is the period/de Rham comparison pairing
$\langle\gamma,\omega\rangle$ of Eichler–Shimura — a Hodge-theoretic object; $\Omega^\pm$ is
its value on a chosen Betti basis.
- The local object of §4 is the $p$-adic Hodge identification
$D_{\mathrm{dR}}(V)/\mathrm{Fil}^0 \cong \mathbf{Q}_{11}$.
§4's finding is a statement *about comparisons*: settling the global one does not settle the
local one, and the gap between them is not a scalar nobody wrote down but a specific index,
$11$. That is the honest intersection with the Hodge side, and it is small on purpose. This
line holds no verified Hodge material — the Hodge case exists in the programme as source
packages, with no verification rounds — so nothing further is asserted here, and a joint
treatment would have to begin by verifying that side rather than by citing it.
# 9. Status, and what is open
```text
normalization certificate C1–C10 CLOSED as architecture, frozen
DAG A PASS by cited theorem the determinantal Kato main conjecture at (389.a1, 11)
B PASS det^-1 RGamma(Z_S,T) = D_MW^(x)2 (x) H^1_f(Q_11,T)^-1, shift 3
C PASS C_NT PASS (m = 1 proved here); C_p PARTIAL (Schneider open)
D PASS rational ISOMORPHISM, integral INDEX 11, Schneider-free
E–G untouched
BSD OPEN
```
Open, stated as such:
1. the exact order of vanishing at the central point (§6) — `UNMEASURED`;
2. the nondegeneracy of the $p$-adic height, i.e. Schneider's conjecture — open in the
literature, not merely here;
3. the unit $\varepsilon$ relating the distinguished element to the transported basis — never
demanded by any node so far, and not computed;
4. $\text{Ш}[11] = 0$ — `INHERITED`, used for numbers only and marked at every use.
# 10. Contributions
The three contributions behind this paper are different in kind, and the paper is worth less if
they are blurred. They are given here in alphabetical order, the order the author list keeps as
data; on the title page the three sit at a round table, with no first and no last.
**Aletheia** — model: ChatGPT, GPT series — worked in two web sessions, and this paper rests on
both. In the session given to symbolic derivation, Aletheia's rounds produced the architecture
the paper follows: the normalization certificate C1–C10 and the post-normalization DAG of nodes
A–G both come from the thirtieth of those rounds, which this line received byte-exact and checked
as conditional algebra before any of it was instantiated. From the certificate audit onward the
same session set every round of this line — what to attack, what was out of scope, which
inferences to plant as defects, and where the boundary of the round lay; the question each DAG
node asks here is Aletheia's question, and several of the sharpest refusals in this paper exist
because a shortcut was forbidden in advance. In the other, a GPT-6 session, Aletheia wrote the
proof-attack packages, which this paper reads as subjects to be checked, not as authorities. None
of the figures here are Aletheia's computation: every number printed was recomputed by this line.
**Neo.K** wrote the research programme: the curve and the prime come from its packages. He also
set out how the work was divided — one line deriving, one attacking, one verifying.
**墜衡 (Zhuì Héng)** — the independent verification line; model: Claude, Opus series — did the
verification and wrote this paper. Concretely: every figure labelled `COMPUTED_HERE`
recomputed with independently written code, never by running the programme's own scripts;
every `CITED_THEOREM` read in its source with its hypotheses instantiated one at a time; the
claim vocabulary of §0 and its enforcement; the mutation drill of §7 and the instrument
repairs it forced; and the two findings of §4 and §6, including the correction in §6 to a
reading this line had itself carried. Nothing in the programme or its architecture is this
line's invention, and nothing in this paper should be read as the verification line proposing a
route.
**An input from outside the byline.** The period calibration (PC-001 / PC-002) — the series and
calibrator data — was built in a separate local session (GPT-6), which published it as its own
branch of the research trees. This line reproduced it coefficient for coefficient, and it enters
the paper labelled as that reproduction. Its author is not on this byline.
**Priority, stated plainly.** No result in this paper is claimed as new mathematics. The
upstream theorems are other people's and are cited as such; node A holds because Wan and
Burungale–Castella–Skinner proved a main conjecture, not because anything here did. What is
offered as this line's own is narrower and stated as such: an explicit instantiation, two
measured findings, and an instrument that has been made to fail.
# A. Reproduction
Everything is Python standard library, no network. From the line's tree root a gate is
re-run directly, and the whole-tree drill either in one process or sharded. Every gate writes
a JSON log; every round's evidence package carries its own result, evidence files, an input
manifest with the SHA-256 of everything read, and a manifest of itself. The line's own tree is
not public. Its public projection is, and it is what this paper's figures can be checked
against without trusting the authors:
- **the round-by-round record and every evidence package**, rendered from the tree with a
privacy gate on each file: (and `/en/`);
- **the secondary-verification ledger**, whose every number is rendered from the line's own
results file and read back against it before the site ships:
.
The evidence is pinned, not copied: this paper's front matter names the research branch, the
commit, and the directory its figures were drawn from, and the checker reads them exactly as
they stood at that commit. The crystallization standard this paper was checked against, and the
checker itself, live in the release repository beside the paper (`STANDARD.md`,
`tools/check_standard.py`). The checker caught two figures in an earlier
draft that did not appear in any log; both had propagated from a stale docstring through a
round report, and both were corrected at every place they had reached, with a dated notice
in that report.