ZFCρ Paper LXIV: Two-Dimensional Anatomy of the Remainder Measure — Branch-Orthogonality, Free-Boundary Profiles, and Surrogate Layering
ZFCρ 第LXIV篇：余项测度的二维解剪——Branch-Orthogonality、自由边界轮廓与Surrogate分层

§1. Introduction

1.1. From Paper 63 to Paper 64

Paper 62 restated 59.1 as two-river synchronization: |Δ_j| = O(j^{3/2}·√D_j). Paper 63, through the transport pairing negative result and increment second-order orthogonality, located 59.1 as a differential level zero-mode pinning problem.

Paper 64 asks: if 59.1 is zero-mode pinning, in what space does this zero-mode live, and what is its internal composition?

This paper defines the two-dimensional remainder measure M_j(u,g) and anatomizes B_j's sources in (position, gap) coordinates.

1.2. Methodological Note

This paper underwent three experiment-driven reframings. The initial hypothesis "rank-one separation is IC-specific structure" was partially overturned by surrogate tests; the subsequent hypothesis "β-charge pinning reduces 59.1" was shown by EXP G to be a restatement of 59.1. The five surviving findings have all been refined through surrogate testing. This path is recorded faithfully.

1.3. Contributions

(i) Definitive quantitative confirmation of Branch-Orthogonality (§3)

(ii) Asymptotic stability of the free-boundary profile Γ̃(g) (§4)

(iii) Count vs Mean decomposition (§5)

(iv) SVD zero-mode decomposition (§6)

(v) Surrogate three-layer separation (§7)

(vi) λ-Ward gauge check (§8)

(vii) Candidate SAE axiom correspondence (§9)

This paper does not claim to prove or reduce 59.1. Its contribution is defining and analyzing the two-dimensional remainder measure M_j(u,g), showing how B_j's zero-mode is composed of branch-position count, gap marginal profile, mean mass, and low-order SVD modes.

---

§2. Definition of the Two-Dimensional Remainder Measure

2.1. Coordinates

For primes p in dyadic block I_j = [2^j, 2^{j+1}), define two coordinates:

Position coordinate u: u_p = rank_{I_j}(p)/m_j ∈ [0,1], where rank is p's ordinal in the block's prime sequence, m_j is total prime count.

Branch gap coordinate g: g(p-1) = ρ_add(p-1) - ρ_mult(p-1), where ρ_add(p-1) = ρ(p-2)+1 (additive path cost), ρ_mult(p-1) = min_{a·b=p-1}[ρ(a)+ρ(b)+2] (optimal multiplicative cost).

2.2. Three Matrices

Discretize u into n_u bins (default 20), g truncated to [-5, +10] (16 bins):

Note: All observed gaps in the tested range fall within this interval; any out-of-range gap is merged into the endpoint tail bin, ensuring every p ∈ I_j is counted in exactly one (u,g)-cell. The zero-mode identity in §2.3 is therefore exact.

Count matrix: N_j(u,g) = #{p ∈ I_j : u_p ∈ bin(u), g(p-1) = g}

Signed mass matrix: M_j(u,g) = Σ_{p ∈ bin(u,g)} h(p)/p

Mean mass matrix: x̄_j(u,g) = M_j(u,g)/N_j(u,g)

with M = N ⊙ x̄ (Hadamard product).

2.3. Zero-Mode Identity

B_j = Σ_{u,g} M_j(u,g). Defining gap marginal signed mass Γ_j(g) = Σ_u M_j(u,g), we have B_j = Σ_g Γ_j(g). This is an identity (coordinate expansion), not a reduction.

Define near-boundary/tie-side and far-multiplicative-side flux: Γ⁺_j = Σ_{g≤1} Γ_j(g) (near-boundary/tie-side aggregate; dominant positive flux from g = 0, 1; small far-additive negative contributions are included but negligible), Γ⁻_j = Σ_{g≥2} Γ_j(g) (far-multiplicative side, persistently negative). Then B_j = Γ⁺_j + Γ⁻_j.

---

§3. Quantitative Confirmation of Branch-Orthogonality

3.1. Count Matrix SVD

Table 1. Count matrix N_j(u,g) SVD (20 u-bins).

j	s₁/s₂	Rank-1 energy
20	55	99.95%
22	73	99.97%
25	148	99.99%
27	240	99.998%
29	321	99.999%

s₁/s₂ grows monotonically with j; branch-position independence strengthens asymptotically.

3.2. Independence Diagnostics

Table 2. Count independence metrics (20 u-bins).

j	χ²/dof	max TV	I(U;G) normalized
20	0.75	0.024	4.4e-4
22	0.98	0.015	1.7e-4
25	1.44	0.008	3.5e-5
27	1.86	0.005	1.4e-5
29	3.72	0.003	7.2e-6

Max TV distance decreases from 0.024 to 0.003. Mutual information from 4.4 × 10⁻⁴ to 7.2 × 10⁻⁶. Physical deviations shrink while χ²/dof grows due to increasing sample size — the correct signature of asymptotic independence.

3.3. Connection to Paper 63

Paper 63 §3.5 proposed Branch-Orthogonality as an empirical proposition. Tables 1-2 provide its definitive quantitative confirmation: count matrix is nearly perfectly rank-one, mutual information decreases rapidly with j, supporting asymptotic independence of branch type and position.

---

§4. Asymptotic Stability of the Free-Boundary Profile

4.1. Normalized Profile

Define Γ̃_j(g) = Γ_j(g)/Σ_g|Γ_j(g)|.

Table 3. Normalized profile Γ̃_j(g) (selected g values).

g	j=20	j=22	j=25	j=27	j=29
0	+0.309	+0.318	+0.333	+0.336	+0.338
+1	+0.091	+0.125	+0.150	+0.161	+0.166
+2	-0.180	-0.156	-0.132	-0.125	-0.121
+3	-0.220	-0.210	-0.202	-0.198	-0.197
+4	-0.120	-0.116	-0.112	-0.111	-0.111

4.2. Convergence

Table 4. Adjacent-j L1 distances.

j pair	‖Γ̃_j − Γ̃_{j-1}‖₁
24-25	0.023
25-26	0.020
26-27	0.011
27-28	0.010
28-29	0.006

Adjacent-j L1 distances decrease monotonically. Γ̃_j(g) shows strong convergence toward a stable limit profile Γ̃_∞(g) over the tested range.

4.3. Interpretation

Sign flip stable at g ≈ 1.5. Tie primes (g = 0) and near-tie (g = 1) contribute positively; deep multiplicative primes (g ≥ 2) contribute negatively. The min-operation free boundary has an asymptotically stable statistical structure in gap space.

---

§5. Count Versus Mean Decomposition

5.1. Three-Matrix SVD

Table 5. SVD s₁/s₂ of three matrices (20 u-bins).

j	Count N	Signed M	Mean x̄
20	55	13	3.3
22	73	19	3.2
25	148	35	3.3
27	240	41	4.1
28	240	44	5.8
29	321	53	4.0

5.2. Finding

Count is extremely rank-one (s₁/s₂ = 55-321). Mean mass is weakly rank-one (s₁/s₂ = 3-6). Signed mass is intermediate: M = N ⊙ x̄, so M's rank-one = count's extreme rank-one diluted by mean's non-separability.

The initial "99.93% rank-one of signed mass" is primarily count baseline. IC-specific two-dimensional structure lives in the mean mass x̄(u,g).

---

§6. SVD Zero-Mode Decomposition

6.1. Mode Decomposition

If M_j = Σ_r s_r a_r b_r^T, then B_j = Σ_r s_r (Σ_u a_r)(Σ_g b_r). Define B^(r) = s_r (Σ_u a_r)(Σ_g b_r).

Table 6. SVD zero-mode decomposition (20 u-bins).

j	‖B⁽¹⁾‖/√(jD)	‖B⁽²⁾‖/√(jD)	‖B⁽³⁾‖/√(jD)	‖B_j‖/√(jD)
20	4.71	0.02	0.01	4.69
22	5.08	0.14	0.01	4.93
25	3.76	0.17	0.00	3.57
27	1.23	0.39	0.00	0.89
28	1.12	0.34	0.03	1.39
29	2.69	0.22	0.09	2.98

Modes 1-5 reconstruct B_j to residual < 10⁻⁶.

6.2. Interpretation

Mode 1 is the dominant low-mode term of B_j's zero-mode. Mode 2 is non-negligible near zero-crossings, contributing 20-40% of √(jD_j) at j = 27-28. Modes 3+ are negligible in the tested range. Table values are absolute; actual reconstruction uses signed B^(r). Any proof route for 59.1 must control both Mode 1 and Mode 2.

---

§7. Surrogate Three-Layer Separation

7.1. Design

Null 1: Permute h(p)/p within each gap class. Preserves Γ(g) marginals, breaks within-class u-landscape.

Null 2: Randomize gap labels within each u-bin. Preserves position landscape, breaks gap structure.

Null 3: Replace h(p)/p with i.i.d. Gaussian of same variance. Baseline.

7.2. Results

Table 7. Signed mass SVD s₁/s₂ comparison.

j	REAL	Null 1	Null 2	Null 3
22	19	30-43	4-6	1.1-1.4
25	35	69-86	4-6	1.1-1.4
27	41	138-171	5-6	1.2-1.7
28	44	178-207	5-8	1.0-1.3
29	53	223-258	8-9	1.1-1.5

7.3. Three-Layer Interpretation

Null 3 → Null 2 (s₁/s₂ from ~1.3 to ~5-8): Gap distribution non-uniformity.

Null 2 → Null 1 (from ~5-8 to ~30-258): Gap-conditioned mean profile Γ(g) — the dominant source of rank-one.

Null 1 → REAL (from ~30-258 down to 19-53): IC recursion has richer joint structure than marginals, manifesting as rank-2+ components. Real IC data is less rank-one than the marginal baseline.

7.4. Key Finding

Null 1's Γ⁺/Γ⁻ is identical to REAL (exact to last digit, all trials, all j). Reason: within-class permutation does not change class sums. Therefore Γ⁺ + Γ⁻ = B_j is an identity — "β-charge pinning" restates 59.1 in gap coordinates, it does not reduce it.

---

§8. Binning Robustness

Table 8. Signed mass SVD s₁/s₂ vs number of u-bins.

j	10-bins	20-bins	50-bins	100-bins
22	32	19	8	5
25	63	35	13	8
27	67	41	16	10
29	61	53	20	12

s₁/s₂ decreases with more bins (fewer primes per cell → more noise), but remains >> 1 at all binnings. Trends are consistent. Results are robust to binning choice.

---

§9. λ-Ward Perturbation

Perturbing λ → λ + t: dB_j/dt = -0.6931 at all tested j, numerically consistent with -ln 2 (Mertens theorem on dyadic blocks). The 83.3%/16.6% near/far split matches prime counting fractions. λ perturbation is a uniform gauge shift, not a non-trivial Ward identity. Gap-dependent perturbations are deferred to future work.

---

§10. Candidate SAE Axiom Correspondence

This paper gives the first systematic mapping of SAE core concepts to concrete computable mathematical structures in the ZFCρ series, presented as candidate correspondences.

Via Negativa / 非: The IC min operation generates the branch-gap coordinate g. Γ̃(g) is the asymptotically stable statistical profile of this free boundary (§4).

Via Rho / 余: The remainder h(p)/p is not error but signed mass. The core object is M_j(u,g), not B_j. This is "remainder as measure."

Operational pairing: Paper 00's thesis "axiomatic Negativa sole, operational paired" receives strong support at the count level (Branch-Orthogonality, §3) and shows more complex coupling at the mean level (§5).

λ calibration and gauge: λ = 3.857 fixes the global zero-drift gauge. 59.1 is the finite-block residual precision of this gauge fixing.

---

§11. Closure Chain Update

59.1's four-layer evolution:

Layer	Statement	Source
Spectral ratio	R_wt = O(1)	Paper 59
Two-river		Δ_j	= O(j^{3/2}·√D_j)	Paper 62
Zero-mode	Differential level pinning + coboundary	Paper 63
Anatomy	Zero-mode = Γ⁺ + Γ⁻ = free-boundary flux balance Ward defect	Paper 64

Paper 64 does not change the closure chain's logical hypotheses. It provides the two-dimensional anatomy of 59.1's zero-mode.

---

§12. Conclusion

Paper 64 defines the two-dimensional remainder measure M_j(u,g) and anatomizes B_j's zero-mode through SVD decomposition and surrogate testing.

Five surrogate-refined findings: (i) Branch-Orthogonality confirmed to count s₁/s₂ = 321 with mutual information → 7.2 × 10⁻⁶. (ii) Free-boundary profile Γ̃(g) converges (L1 distance 0.006 between j = 28 and 29). (iii) Count vs Mean: signed mass rank-one = count (321) × mean (3-6); IC-specific structure in mean mass. (iv) Zero-mode: Mode 1 dominates, Mode 2 contributes 20-40% of √(jD), Mode 3+ negligible. (v) λ = uniform gauge shift.

This paper does not prove or reduce 59.1. B_j = Γ⁺ + Γ⁻ is a coordinate expansion, not a reduction. The paper provides B_j's natural coordinate system: B_j arises from fine cancellation between tie primes (g ≤ 1, positive flux) and deep multiplicative primes (g ≥ 2, negative return flux). λ calibration guarantees global cancellation; finite-block fluctuation is 59.1. Branch-Orthogonality shows fluctuation does not come from branch-position coupling. The genuine difficulty lies in the mean mass x̄(u,g)'s two-dimensional structure and Mode 2's zero-mode.

---

References

[50] Paper L. Feature stripping, DSF_α, BL_α, SD, closure chain. DOI: 10.5281/zenodo.19381111.

[52] Paper LII. Martingale-coboundary and Gordin decomposition route.

[53] Paper LIII. Screening and correlation truncation. DOI: 10.5281/zenodo.19415440.

[54] Paper LIV. Recursive ancestor inheritance and power-law ACF. DOI: 10.5281/zenodo.19426094.

[59] Paper LIX. Spectral ratio boundedness and two-conjecture closure of H'. DOI: 10.5281/zenodo.19480518.

[61] Paper LXI. Conditional reduction of the spectral derivative bound. DOI: 10.5281/zenodo.19605749.

[62] Paper LXII. Two-River Synchronization. DOI: 10.5281/zenodo.19656365.

[63] Paper LXIII. Increment Second-Order Orthogonality and Zero-Mode Pinning. DOI: 10.5281/zenodo.19659860.

---

AI Contributions

Claude (子路): All numerical experiment design and implementation (paper64_abce.c 2D measure + SVD + sliding window + λ-Ward; paper64_surr.c three surrogate tests + β tracking + zero-mode decomposition; paper64_fghi.c β shape + Null 1 balance + Count/Mean/Mean SVD + mode zero-modes; paper64_final.c binning robustness + independence metrics + L1 profile stability + formal zero-mode table), initial rank-one separation identification, two post-surrogate reframings, working notes v1-v3. ChatGPT (公西华): Original proposal of Remainder-Flux Calculus framework (free-boundary flux balance, calibration-gauge symmetry, Gap-Flux Ward Identity), proposal of R_j(u,g) two-dimensional remainder measure, rank-one SVD test design, λ-Ward perturbation design, Surrogate protocol v1, identification that "β-pinning = B_j restating" (post-EXP G), Rank-One Reduction Lemma theorem structure, normalization issues (SVD scale/sign ambiguity), count vs mean decomposition proposal, Γ profile L1 stability proposal, three rounds of gatekeeper review. Grok (子贡): Positioning of rank-one separation as Paper 64 core framework, free-boundary charge pinning naming, identification of λ-Ward as uniform gauge shift. Gemini (子夏): Physical interpretation of rank-one separation, theoretical prediction of surrogate failure (Marchenko-Pastur), multiplicative cost perturbation as non-trivial Ward identity candidate. Independent Review Claude: Surrogate three-layer protocol (Null 1/2/3), reframing s₁/s₂ j-growth as primary finding, quantitative analysis of rank-2 residual zero-mode pollution, β⁺/β⁻ balance CLT estimation, "IC has richer joint structure than marginals" precise reframing, transfer operator / spectral gap unification framework, "falsification-refined claim is more robust than original claim" methodological observation.

§1. 引言

1.1. 从Paper 63到Paper 64

Paper 62将59.1精确重述为双河同步问题：|Δ_j| = O(j^{3/2}·√D_j)。Paper 63通过transport配对的负结果和差分二阶正交性，将59.1定位为differential level的零模pinning问题。

Paper 64的动机：如果59.1是零模pinning，那这个零模在什么空间里？它的内部组成是什么？

本文定义二维余项测度M_j(u,g)，在(position, gap)坐标系中解剖B_j的来源。

1.2. 方法论说明

本文经历了三次实验驱动的重构。初始假说"rank-one separation是IC-specific structure"被Surrogate tests部分推翻，进一步的假说"β-charge pinning is a reduction of 59.1"被EXP G证明是59.1的restating。最终站稳的五个发现均经过surrogate淬炼。本文如实记录这一路径。

1.3. 本文的贡献

(i) Branch-Orthogonality的definitive定量确认（§3）

(ii) 自由边界轮廓Γ̃(g)的渐近稳定性（§4）

(iii) Count vs Mean分解（§5）

(iv) SVD零模分解（§6）

(v) Surrogate三层分离（§7）

(vi) λ-Ward gauge check（§8）

(vii) 与SAE公理的candidate对应（§9）

本文不声称证明或化简59.1。本文的贡献是定义并分析二维余项测度M_j(u,g)，说明B_j的零模如何由branch-position count、gap marginal profile、mean mass与低阶SVD modes共同组成。

---

§2. 二维余项测度的定义

2.1. 坐标系

对dyadic block I_j = [2^j, 2^{j+1})内的素数p，定义两个坐标：

位置坐标u： u_p = rank_{I_j}(p) / m_j ∈ [0,1]，其中rank是p在block内素数序列中的编号，m_j是block内素数总数。

Branch gap坐标g： g(p-1) = ρ_add(p-1) - ρ_mult(p-1)，其中ρ_add(p-1) = ρ(p-2) + 1（加法路径成本），ρ_mult(p-1) = min_{a·b=p-1}[ρ(a)+ρ(b)+2]（乘法路径最优成本）。g < 0：加法更优。g > 0：乘法更优。g = 0：tie。

2.2. 三个矩阵

将u离散化为n_u个bin（默认n_u = 20），g截断到[G_min, G_max] = [-5, +10]（16个bin），定义：

注： 实验中所有观测到的gap均落入该区间；若出现区间外gap，则合并入端点tail bin，保证每个p ∈ I_j被且仅被计入一个(u,g)-cell。因此§2.3的零模恒等式为精确恒等式。

Count matrix： N_j(u,g) = #{p ∈ I_j : u_p ∈ bin(u), g(p-1) = g}

Signed mass matrix： M_j(u,g) = Σ_{p ∈ bin(u,g)} h(p)/p

Mean mass matrix： x̄_j(u,g) = M_j(u,g) / N_j(u,g)

其中M = N ⊙ x̄（Hadamard积）。

2.3. 零模恒等式

B_j = Σ_{u,g} M_j(u,g)

定义gap marginal signed mass：

Γ_j(g) = Σ_u M_j(u,g)

则B_j = Σ_g Γ_j(g)。这是恒等式（坐标展开），不是reduction。

定义near-boundary/tie侧和far-multiplicative侧通量：

Γ⁺_j = Σ_{g≤1} Γ_j(g)（near-boundary/tie侧aggregate，主正通量来自g=0和g=1；少量far-additive负贡献并入但量级很小）

Γ⁻_j = Σ_{g≥2} Γ_j(g)（far-multiplicative侧aggregate，恒负）

则B_j = Γ⁺_j + Γ⁻_j。

---

§3. Branch-Orthogonality的定量确认

3.1. Count Matrix SVD

表1. Count matrix N_j(u,g)的SVD（20 u-bins）。

j	s₁/s₂	Rank-1 energy
20	55	99.95%
22	73	99.97%
25	148	99.99%
27	240	99.998%
29	321	99.999%

s₁/s₂随j单调增长，branch type与block位置的独立性在大j下渐近增强。

3.2. 独立性统计量

表2. Count independence metrics（20 u-bins）。

j	χ²/dof	max TV	I(U;G) normalized
20	0.75	0.024	4.4e-4
22	0.98	0.015	1.7e-4
25	1.44	0.008	3.5e-5
27	1.86	0.005	1.4e-5
29	3.72	0.003	7.2e-6

max TV distance从0.024递减到0.003。 sup_u |P(g|u) - P(g)|趋向零。

Mutual information从4.4e-4递减到7.2e-6。 Branch type与position的信息依赖趋向零。

χ²/dof从0.75增长到3.72。 物理偏差在缩小（TV, MI均减），但sample size增大使微小偏差变得statistically detectable——这是Branch-Orthogonality的正确表现。

3.3. 与Paper 63的连接

Paper 63 §3.5提出Branch-Orthogonality作为经验命题：branch type χ(p)与block内fractional位置u渐近独立。表1-2给出了这个命题的definitive定量确认：count matrix几乎完美rank-one，mutual information随j快速下降，支持branch type与position的渐近独立。

---

§4. 自由边界轮廓的渐近稳定性

4.1. 归一化轮廓

定义Γ̃_j(g) = Γ_j(g) / Σ_g |Γ_j(g)|。

表3. 归一化轮廓Γ̃_j(g)（选定g值）。

g	j=20	j=22	j=25	j=27	j=29
-2	-0.022	-0.022	-0.022	-0.021	-0.021
-1	-0.007	-0.003	-0.001	+0.000	+0.000
0	+0.309	+0.318	+0.333	+0.336	+0.338
+1	+0.091	+0.125	+0.150	+0.161	+0.166
+2	-0.180	-0.156	-0.132	-0.125	-0.121
+3	-0.220	-0.210	-0.202	-0.198	-0.197
+4	-0.120	-0.116	-0.112	-0.111	-0.111
+5	-0.039	-0.036	-0.035	-0.034	-0.034

4.2. 收敛诊断

表4. 相邻j的L1距离。

j对	‖Γ̃_j − Γ̃_{j-1}‖₁
24 vs 25	0.023
25 vs 26	0.020
26 vs 27	0.011
27 vs 28	0.010
28 vs 29	0.006

相邻j的L1距离单调递减。Γ̃_j(g)在测试区间显示出向稳定limit profile Γ̃_∞(g)收敛的强迹象。

4.3. 结构解读

Sign flip稳定在g ≈ 1.5。 g = 0和g = 1恒正（tie素数和近tie素数），g ≥ 2恒负（深乘法素数）。

g = 0（tie，加法和乘法同样优）的归一化贡献从0.31上升到0.34并趋于稳定。g = 3（深乘法区负峰）从-0.22趋于-0.20。g = +1的正贡献从0.09增长到0.17——近乘法区的正贡献在大j下加强。

这是min操作自由边界的渐近稳定统计轮廓。 IC递推在gap空间的结构在大j下趋于自相似。

---

§5. Count与Mean的分解

5.1. 三个矩阵的SVD

表5. 三个矩阵的SVD s₁/s₂（20 u-bins）。

j	Count N	Signed M	Mean x̄
20	55	13	3.3
22	73	19	3.2
25	148	35	3.3
27	240	41	4.1
28	240	44	5.8
29	321	53	4.0

5.2. 核心发现

Count极端rank-one（s₁/s₂ = 55-321）：branch-position几乎完美独立。

Mean mass弱rank-one（s₁/s₂ = 3-6）：每个(u,g)-cell内的平均h(p)/p有genuine的二维非可分结构。

Signed mass介于两者之间（s₁/s₂ = 13-53）：M = N ⊙ x̄，所以M的rank-one = count的极端rank-one被mean的非可分性稀释。

5.3. IC递推的signature在哪里

initial观察到的"99.93% rank-one of signed mass"主要来自count baseline（branch-position独立性），不是IC递推的特有结构。IC递推的genuine二维signature藏在mean mass x̄(u,g)里——每个gap class内的平均余项值依赖于block位置（s₁/s₂ ≈ 3-6）。

Surrogate tests直接确认了这个分层（见§7）。

---

§6. SVD零模分解

6.1. 零模恒等式

设M_j的SVD为M_j = Σ_r s_r a_r(u) b_r(g)^T。则：

B_j = Σ_{u,g} M_j(u,g) = Σ_r s_r · (Σ_u a_r(u)) · (Σ_g b_r(g))

定义每个mode的零模贡献B^(r) = s_r · (Σ_u a_r) · (Σ_g b_r)。

6.2. 零模数据

表6. SVD零模分解（20 u-bins）。

j	‖B⁽¹⁾‖/√(jD)	‖B⁽²⁾‖/√(jD)	‖B⁽³⁾‖/√(jD)	‖B_j‖/√(jD)
20	4.71	0.02	0.01	4.69
22	5.08	0.14	0.01	4.93
25	3.76	0.17	0.00	3.57
27	1.23	0.39	0.00	0.89
28	1.12	0.34	0.03	1.39
29	2.69	0.22	0.09	2.98

前5个mode的累计和精确重构B_j（residual < 10^{-6}）。

6.3. 解读

Mode 1是B_j零模的主导低模项。 Mode 2在穿零附近不可忽略，贡献达到√(jD_j)的20-40%。Mode 3+在测试区间很小。表中列出绝对值，实际重构使用带符号的B^(r)。

注： Mode 1的Σ_u a₁(u)在所有j上 ≈ ±4.39（稳定），而Σ_g b₁(g)随j振荡且小（从0.45到0.01再到0.02）。B^(1) = s₁ · (Σa₁) · (Σb₁)的振荡主要由Σb₁控制——这就是B_j的振荡/穿零（z_j穿零）在rank-1 mode上的反映。

---

§7. Surrogate三层分离

7.1. 实验设计

三个Null假设：

Null 1： 在每个gap class内permute h(p)/p值。保持β(g)边际，破坏u-landscape内部结构。

Null 2： 在每个u-bin内随机化gap标签。保持位置景观，破坏gap结构。

Null 3： 完全随机高斯替换h(p)/p（same σ）。Baseline。

7.2. 结果

表7. Signed mass M_j的SVD s₁/s₂对比。

j	REAL	Null 1	Null 2	Null 3
22	19	30-43	4-6	1.1-1.4
25	35	69-86	4-6	1.1-1.4
27	41	138-171	5-6	1.2-1.7
28	44	178-207	5-8	1.0-1.3
29	53	223-258	8-9	1.1-1.5

7.3. 三层解读

Null 3 → Null 2（s₁/s₂从~1.3到~5-8）： Gap分布非均匀性的贡献。不同gap class有不同素数比例（83%/17%），这本身产生moderate rank-one。

Null 2 → Null 1（s₁/s₂从~5-8到~30-258）： β(g) profile的贡献。每个gap class有不同的mean h/p——近边界正，深乘法负——这是rank-one的主要来源。

Null 1 → REAL（s₁/s₂从~30-258下降到19-53）： IC递推在marginal rank-one之上添加了结构性rank-2+修正。真实IC数据比纯边际更rich——有gap与position的joint dependence。但这些修正的零模小（§6）。

7.4. 关键发现

Null 1的s₁/s₂高于REAL。 IC递推不是"制造rank-one"，而是在marginal rank-one baseline上引入结构性高阶修正。幸运的是这些修正的零模贡献受控。

Null 1的Γ⁺/Γ⁻与REAL完全一致（精确到最后一位，所有trial，所有j）。原因：gap class内permute不改变class sum。因此Γ⁺ + Γ⁻ = B_j是恒等式——"β-charge pinning"是59.1的gap坐标restating，不是reduction。

---

§8. Binning Robustness

8.1. 不同u-bin数下的SVD

表8. Signed mass M_j的SVD s₁/s₂ vs u-bin数。

j	10-bins	20-bins	50-bins	100-bins
22	32	19	8	5
25	63	35	13	8
27	67	41	16	10
29	61	53	20	12

s₁/s₂随bin数增加而下降（每个cell更少primes → 更多噪声 → rank-one稀释）。但所有binning下s₁/s₂ >> 1，趋势一致。

8.2. 结论

rank-one结构和Γ̃(g) profile不依赖特定binning选择。

---

§9. λ-Ward扰动

9.1. 结果

扰动λ → λ + t后，dB_j/dt在所有j上一致：

j	dB/dt
22	-0.6929
27	-0.6931
29	-0.6931

数值上与-ln 2一致。这是Mertens定理在dyadic block上的标准推论：Σ_{p∈I_j} log(p)/p ~ ln 2。

9.2. 分裂比例

dB_near/dB_total = 83.3%（近边界素数计数比例）

dB_far/dB_total = 16.6%（远乘法素数计数比例）

λ扰动均匀作用于所有gap class。 λ是全局scale calibration（gauge fixing），不区分near和far。λ扰动不是非trivial的Ward identity。真正的Ward identity需要扰动branch normal方向（gap-dependent perturbation），留待后续工作。

---

§10. 与SAE公理的candidate对应

10.1. 定位

本文是ZFCρ系列中首次系统地将SAE核心概念映射到具体可计算数学结构的工作。以下对应作为candidate提出，不作为公理证明。

10.2. Via Negativa / 非

IC的min操作生成branch-gap坐标g = ρ_add - ρ_mult。这是min操作自由边界的normal coordinate。§4的Γ̃(g) profile是这个自由边界的渐近稳定统计轮廓。

10.3. Via Rho / 余

余项h(p)/p不是误差，而是signed mass。Paper 64的核心对象不是B_j，而是二维余项测度M_j(u,g)。这就是"余项作为测度"的具体数学实现。

10.4. 非与余的操作层配对

Paper 00（余之道）的核心命题："公理层上非独一，操作层上非与余成对。"

在IC递推中：公理层只有min操作（非）；运行后自然产生g坐标（free boundary）和h(p)/p值（remainder mass）——两个因子。

Branch-Orthogonality（count matrix rank-one，§3）说明在population层面，branch type与block位置几乎完全独立——两条Via在population层分离。但mean mass x̄(u,g)的弱rank-one（s₁/s₂ ≈ 3-6，§5）说明在individual remainder amplitude层面，gap与position有genuine coupling。

因此Paper 00的"操作层配对"在count层面得到strong support，在mean层面显示出更复杂的耦合结构。

10.5. λ calibration与gauge

λ = 3.857的选择使全局Σh(p)/p = 0——这是Paper 00所说"非给余以存在条件"的gauge-fixing操作。59.1是这个gauge fixing在finite block上的residual精度问题。

---

§11. 闭合链更新

局部 GAO（P62 命题 2.1）+ Tail-k Control（待证）
  → k 一致性 → 59.2 退出独立猜想
59.1 + Guy/Chebyshev → DSF_α ∀α > 0   （Paper 59 §7.1）
59.2 → BL_α ∀α < 1                     （Paper 59 §7.2）
SD（无条件，Paper 50）
  → Hölder → A_h 收敛 → D(s) 边界 → H'

59.1的四层进化：

层级	表述	来源
谱比层	R_wt = O(1)	Paper 59
双河层		Δ_j	= O(j^{3/2}·√D_j)	Paper 62
零模层	differential level pinning + coboundary	Paper 63
解剖层	零模 = Γ⁺ + Γ⁻ = free-boundary flux balance ward defect	Paper 64

Paper 64不改变闭合链的逻辑假设。它给出59.1零模的二维解剖。

---

§12. 结论

Paper 64定义了二维余项测度M_j(u,g)并通过SVD分解和Surrogate tests解剖了B_j零模的内部组成。

经过Surrogate淬炼后的五个发现：

(i) Branch-Orthogonality（§3）： Count matrix s₁/s₂从55增长到321。Branch type与block位置的mutual information从4.4e-4降到7.2e-6。这是Paper 63 §3.5经验命题的definitive定量确认。

(ii) 自由边界轮廓（§4）： 归一化Γ̃(g)跨j收敛。相邻j的L1距离从0.023单调降至0.006。Sign flip稳定在g ≈ 1.5。min操作自由边界有渐近稳定的统计结构。

(iii) Count vs Mean（§5）： Signed mass的rank-one = count的极端rank-one（321）× mean的弱rank-one（3-6）。IC递推的genuine二维结构在mean mass里。

(iv) 零模分解（§6）： Mode 1主导B_j，mode 2贡献√(jD)的20-40%，mode 3+可忽略。B_j不是纯rank-1问题。

(v) λ = gauge shift（§9）： dB/dt = -0.6931 ≈ -ln 2，分裂按素数计数比例。λ是uniform calibration。

本文不证明或化简59.1。 B_j = Γ⁺ + Γ⁻是坐标展开不是reduction——Γ⁺/Γ⁻ balance就是B_j smallness的gap坐标restating。

本文提供的是： B_j零模的natural坐标系。在这个坐标系中，B_j来自tie/near-boundary侧（g ≤ 1，主正通量来自g=0和g=1）和深乘法侧（g ≥ 2，负回流）之间的精细对消。λ calibration保证全局对消；finite block fluctuation是59.1。Branch-Orthogonality说明fluctuation不来自branch-position coupling。IC递推的genuine难点在mean mass x̄(u,g)的二维结构和mode 2零模。

---

参考文献

[50] 第L篇。特征剥离, DSF_α, BL_α, SD, 闭合链。DOI: 10.5281/zenodo.19381111。

[52] 第LII篇。Martingale-coboundary与Gordin分解路线。

[53] 第LIII篇。Screening与相关截断。DOI: 10.5281/zenodo.19415440。

[54] 第LIV篇。递归祖先继承与幂律ACF。DOI: 10.5281/zenodo.19426094。

[59] 第LIX篇。谱比有界性与H'的双猜想闭合。DOI: 10.5281/zenodo.19480518。

[61] 第LXI篇。谱导数界的条件归约。DOI: 10.5281/zenodo.19605749。

[62] 第LXII篇。双河同步。DOI: 10.5281/zenodo.19656365。

[63] 第LXIII篇。余项增量二阶正交性与零模Pinning。DOI: 10.5281/zenodo.19659860。

---

AI贡献

Claude（子路）：全部数值实验设计与实现（paper64_abce.c二维测度+SVD+sliding window+λ-Ward，paper64_surr.c三组Surrogate tests+β tracking+零模分解，paper64_fghi.c β形状+Null 1 balance+Count/Mean/Mean SVD+mode零模，paper64_final.c binning robustness+independence metrics+L1 profile stability+formal零模表），初始rank-one separation的识别，Surrogate后的两次reframe，working notes v1-v3。ChatGPT（公西华）：Remainder-Flux Calculus框架的原始提出（free-boundary flux balance, calibration-gauge symmetry, Gap-Flux Ward Identity β(g) = J(g-½) - J(g+½) + η(g)），R_j(u,g)二维余项测度的提出，rank-one SVD test实验方案，λ-Ward扰动实验设计，Surrogate protocol v1，"β-pinning = B_j restating"的识别（EXP G后），Rank-One Reduction Lemma→Proposition 64.6的theorem结构，normalization问题（SVD尺度/符号任意性），count vs mean分解的提出，Γ profile L1 stability的提出，三轮gatekeeper review（6处blocker+5处hotfix+v3 final check），五个"避免"措辞。Grok（子贡）：rank-one separation作为Paper 64核心框架的定位，free-boundary charge pinning命名，λ-Ward作为uniform gauge shift的识别。Gemini（子夏）：rank-one separation的物理解释（"热力学极限下的宏观态坍缩"），surrogate必败性的理论预测（Marchenko-Pastur），乘法成本扰动作为非trivial Ward identity的建议。独立评审Claude：surrogate三层protocol（Null 1/2/3），s₁/s₂ j-growth作为主发现的reframe（比99.93%更重要），rank-2残差可能污染59.1的定量分析（0.07%×N² >> 2^{-j/2}），β⁺/β⁻ balance CLT估算（|β⁺+β⁻|/(1/√N_j) ≈ O(1)），"IC has richer joint structure than marginals"的精确reframe，transfer operator/spectral gap统一框架提议，"falsification-refined claim比original claim更robust"的方法论observation。