Fixed-Core Locking and Prime-Core Profile — The Residual Floor Structure of H'
📄 DOI: 10.5281/zenodo.19116625We perform microscopic anatomy of the Ω=2 layer (semi-primes n=pq), discovering that the H' criterion reduces precisely to a prime-core harmonic average problem.
An h₀-free gap formula is established: G(n) = ρ(p) + ρ(q) − ρ(n−1) + 1; j(n)>0 iff G(n)<0. Independent confirmation: γ₂ = 0.453 (R² = 0.996). G is integer-valued, concentrated on {−1, 0, +1, +2} (87%).
A prime penalty is discovered: primes q have systematically higher ρ(q) than the local mean (E[R(p)] ≈ 4.57 vs E[R(n)] ≈ 2.76), because primes are pure additive injection points in the DP recursion. This explains E[diff] = 1.83 for n=2q (far below h₀·ln 2 = 2.48).
Discrete locking of n=2q: diff = ρ(2q−1) − ρ(q) concentrates on {1, 2, 3} (92%). P(j>0) = P(diff≥3) ≈ 0.211, frozen across five decades. In the currently visible range, p=2 is the only prime-core family showing no clear downward trend.
A prime-core profile is established for 95 primes (p=2 to 499): cp = P(j>0|n=pq). cp is primarily determined by margin = E[diff] − threshold (Corr = 0.917). Margin oscillation is driven by the discrete deviation of ρ(p) from h₀·ln p — and ρ(p) deviation is strongly correlated with the smoothness of p−1 (Corr(ln P⁺(p−1), cp) = −0.714). Smooth p−1 → low ρ(p) → high margin → high cp. Rough p−1 → high ρ(p) → low margin → low cp.
Key new finding: except for p=2, nearly all cp(N) for p ≥ 3 decline across decades (2–15%/decade) — cp is not a constant but an N-dependent slowly decaying quantity. At fixed N=10⁸, the partial harmonic average H(P;N) = Σ cp/p ÷ Σ 1/p decreases from 0.35 to 0.27 (P=499) — declining in the P direction but far from zero.
H' criterion for Ω=2 is: does the harmonic Cesàro average c̄h(N) → 0? Current data cannot determine this — whether cp decay is genuine asymptotic behavior or finite-scale effect remains open.
30 fixed-core families in Ω=3 (m=6, 10, 14, …, 95) exhibit the same margin-driven structure. Fixed-core locking is universal across all Ω layers.
Paper XXXI's γk is a mixture effect of fixed-core families. Model A's finite-scale "victory" reflects the decay phase of the mixture.
Keywords: additive complexity, prime penalty, discrete locking, prime-core profile, harmonic Cesàro average, fixed-core family, integer complexity
1. Introduction
1.1 Background
Paper XXXI established the layer-by-layer bottleneck map for H', extracted γk, and proposed the effective index Γ = λ·γk criterion. But γk was found to be a mixture effect. This paper performs microscopic anatomy of the simplest Ω=2 layer, revealing the internal structure of that mixture.
1.2 Contributions
(A) h₀-free gap formula and γ₂ confirmation (§2). (B) Prime penalty (§3). (C) n=2q discrete locking (§4). (D) 95-prime core profile (§5). (E) ρ(p) deviation ↔ p−1 smoothness correlation (§5). (F) cp(N) per-decade decay (§6). (G) Harmonic Cesàro average as H' criterion (§7). (H) Ω=3 fixed-core structure (§8).
2. h₀-Free Gap Formula and γ₂ Confirmation
2.1 Formula
For n = pq (p = P⁻(n), canonical core), define:
$$G(n) = \rho(p) + \rho(q) - \rho(n-1) + 1$$Then j(n) > 0 if and only if G(n) < 0. No h₀ appears.
2.2 Discreteness of G
G is integer-valued. Distribution at [10⁷, 10⁸):
| G | −1 | 0 | +1 | +2 | Other |
|---|---|---|---|---|---|
| Share | 18.5% | 30.1% | 24.2% | 14.4% | 12.8% |
Four values account for 87%. Median = 0, mean = +0.53.
2.3 γ₂ Independent Confirmation
$$q_2(N) \sim 0.877\cdot(\ln N)^{-0.453}, \quad R^2 = 0.996$$
14 points (N = 10³ to 10⁸). Confirms Paper XXXI's γ₂ = 0.449.
2.4 Rapid Decay of Balanced Semi-Primes
The Ω=2 layer is not homogeneous. Restricting to "balanced" semi-primes (q/p < 10) reveals strikingly different behavior.
Theory: for p ≈ q ≈ √n, ρ(p) + ρ(q) ≈ h₀·ln n ≈ ρ(n−1), giving G ≈ +1 (structural positive bias).
Data ([10⁷, 10⁸)): E[G] = +1.23, P(j>0) = 0.107. Fitting:
$$q_2^{\text{bal}}(N) \sim 3.03\cdot(\ln N)^{-1.14}, \quad R^2 = 0.992$$γbal = 1.14 — 2.5× faster than the overall γ₂ = 0.45. This directly proves γ₂ is a mixture of fast-decaying balanced subfamilies and slowly-decaying/frozen unbalanced subfamilies.
3. Prime Penalty
Block 3b data (N = 10⁸):
| Category | E[R(·)] |
|---|---|
| General integers n ∈ [10⁷, 10⁸) | 2.76 |
| Primes p ∈ [10⁷, 10⁸) | 4.57 |
Prime penalty: gap ≈ 1.8. Primes are pure additive injection points: ρ(p) = ρ(p−1) + 1, never compressed by multiplication. Hence ρ(p) is systematically elevated relative to its logarithmic scale.
For n=2q: E[diff₂] ≈ h₀·ln 2 − Δ₂ ≈ 2.48 − 0.65 ≈ 1.83, where Δ₂ ≈ 0.65 is the family-specific prime-conditioned correction for p=2 (distinct from the unconditional gap of ~1.8).
4. Discrete Locking of n=2q
4.1 Threshold
Since ρ(2) = 1: G = 2 + ρ(q) − ρ(2q−1). Thus j>0 iff diff ≥ 3.
4.2 diff Distribution
[10⁷, 10⁸), 2.65 million primes q:
| diff | +1 | +2 | +3 | Other |
|---|---|---|---|---|
| Share | 21.6% | 49.5% | 21.1% | 7.8% |
Three values account for 92%. P(j>0) = P(diff≥3) ≈ 21.1%.
4.3 Five Decades of Confirmed Freezing
| Decade | E[diff] | P(j>0) |
|---|---|---|
| [10³, 10⁴) | 1.843 | 0.195 |
| [10⁵, 10⁶) | 1.840 | 0.209 |
| [10⁷, 10⁸) | 1.829 | 0.211 |
P(j>0) ≈ 0.211 is frozen across five decades. In the currently visible range, p=2 is the only prime-core family in Ω=2 without a clear downward trend.
5. Prime-Core Profile
5.1 Definition
For each prime p, canonical decomposition n = pq (q ≥ p prime, p = P⁻(n)):
$$c_p(N) = P(j>0 \mid n=pq,\; pq \leq N)$$ $$\text{threshold}_p = \rho(p)+1, \quad \text{margin}_p = E[\text{diff}_p] - \text{threshold}_p$$When q=p (prime squares), these are asymptotically negligible (<0.01%) and excluded.
5.2 95-Prime Profile (Selected)
At N = 10⁸:
| p | ρ(p) | thr | E[diff] | margin | cp | P⁺(p−1) | p−1 |
|---|---|---|---|---|---|---|---|
| 2 | 1 | 2 | 1.83 | −0.17 | 0.211 | 1 | 1 |
| 3 | 2 | 3 | 3.52 | +0.52 | 0.543 | 2 | 2 |
| 5 | 4 | 5 | 5.33 | +0.33 | 0.471 | 2 | 4=2² |
| 7 | 6 | 7 | 6.58 | −0.42 | 0.224 | 3 | 6=2·3 |
| 23 | 12 | 13 | 11.16 | −1.84 | 0.035 | 11 | 22=2·11 |
| 59 | 16 | 17 | 14.80 | −2.20 | 0.019 | 29 | 58=2·29 |
| 109 | 16 | 17 | 17.19 | +0.19 | 0.381 | 3 | 108=2²·3³ |
| 179 | 21 | 22 | 19.11 | −2.89 | 0.005 | 89 | 178=2·89 |
| 193 | 18 | 19 | 19.41 | +0.41 | 0.471 | 3 | 192=2⁶·3 |
cp range: [0.005, 0.543]. Full 95-point profile computed.
5.3 Margin Dominance
Corr(margin, cp) = +0.917. Margin is the primary driver; local discrete distribution contributes second-order variation.
5.4 ρ(p) Deviation Driven by p−1 Smoothness
Corr(ln P⁺(p−1), ρ(p)/h₀ ln p) = +0.683.
Corr(ln P⁺(p−1), cp) = −0.714.
Correlation-supported explanation chain: smooth p−1 → low ρ(p−1) → low ρ(p) → low threshold → high margin → high cp. Rough p−1 → reverse.
Extreme cases: p=193 (p−1=192=2⁶·3, P⁺=3) → cp=0.471. p=179 (p−1=178=2·89, P⁺=89) → cp=0.005.
5.5 No Monotone Drift in Current Range
ρ(p)/(h₀·ln p) oscillates between 0.95–1.13 for p ≤ 499. No systematic drift supporting a monotone scissors picture. Since P⁺(p−1) is known to oscillate indefinitely, cp is unlikely to trend to zero via a simple "h₀·ln p outgrows ρ(p)" mechanism — but the harmonic average may still decline through subtler distributional effects.
6. Per-Decade Decay of cp(N)
| p | cp(10³) | cp(10⁵) | cp(10⁷) | Trend |
|---|---|---|---|---|
| 2 | 0.195 | 0.209 | 0.211 | frozen |
| 3 | 0.558 | 0.551 | 0.542 | slow decline |
| 5 | 0.545 | 0.486 | 0.470 | clear decline |
| 7 | 0.267 | 0.239 | 0.223 | clear decline |
| 109 | 0.465 | 0.430 | 0.377 | fast decline |
| 193 | 0.642 | 0.521 | 0.467 | fast decline |
Except p=2, nearly all cp(N) for p ≥ 3 are declining. Rates range from 2%/decade (p=3) to 15%/decade (p=193). cp(N) is not a fixed arithmetic constant — it is an N-dependent slowly decaying quantity. Current data cannot distinguish "cp → constant > 0" from "cp → 0 very slowly."
7. Harmonic Cesàro Average as H' Criterion
7.1 Canonical Decomposition
Each Ω=2 composite n=pq (p=P⁻(n), q=P⁺(n), q≥p) maps uniquely to prime-core p. Prime squares (q=p) contribute <0.01% and are negligible. On squarefree semiprimes:
$$q_2^{(\text{sqf})}(N) = \sum_{p} w_p(N)\cdot c_p(N)$$where wp(N) ≈ (1/p) / Σ(1/ℓ) asymptotically. Since prime squares contribute O(π(√N)/#Ω=2) → 0, q₂ and \(q_2^{(\text{sqf})}\) have the same asymptotics.
7.2 Correct Criterion
q₂ → 0 if and only if the harmonic Cesàro average tends to zero:
$$\bar{c}_h(N) := \frac{\displaystyle\sum_{p \leq \sqrt{N}} c_p(N)/p}{\displaystyle\sum_{p \leq \sqrt{N}} 1/p} \to 0$$
This is not equivalent to "Σ cp/p converges" (which is only sufficient). If cp(N) ~ C/ln ln p, the series diverges yet c̄h → 0.
7.3 Partial Profile at Fixed N
H(P;N) at fixed N=10⁸ as function of truncation P:
| P | H(P; N) |
|---|---|
| 10 | 0.351 |
| 50 | 0.305 |
| 100 | 0.291 |
| 200 | 0.281 |
| 500 | 0.270 |
H declines from 0.35 to 0.27. This is the profile shape at fixed N — not the asymptotic c̄h(N) vs N. The true H' criterion requires understanding the joint profile (p, N) ↦ cp(N).
7.4 H' Implications
(i) If cp(N) are true constants → c̄h stable at ~0.27 → H' fails at Ω=2.
(ii) If cp(N) for p≥3 have genuine asymptotic decay → c̄h(N) → 0 → H' holds at Ω=2.
§6 data favor scenario (ii) but decay rates are too small to confirm. H' status at Ω=2 remains open.
7.5 Partial Validation
Block 7b: Σ wp·cp (p ≤ 23) = 0.174. Actual q₂ = 0.232. Coverage 54%, ratio = 0.75. Weighted sum direction is correct — contributions from p > 23 should close the gap.
8. Ω=3 Fixed-Core Structure
8.1 Framework
For Ω=3 layer, canonical core m (Ω(m)=2, squarefree), n = mq (q = P⁺(n)):
$$G_m(q) = \rho(m) + \rho(q) - \rho(mq-1) + 1, \quad j>0 \;\text{iff}\; \text{diff}_m \geq \rho(m)+2$$Note: current scope covers only squarefree cores m = p₁p₂. Squareful families (n=p²q) not yet included.
8.2 30-Core Profile (Selected)
At N = 10⁸:
| m | Factors | ρ(m) | thr | E[diff] | margin | cm |
|---|---|---|---|---|---|---|
| 6 | 2·3 | 5 | 6 | 6.37 | +0.37 | 0.457 |
| 10 | 2·5 | 7 | 8 | 8.25 | +0.25 | 0.395 |
| 14 | 2·7 | 9 | 10 | 9.41 | −0.59 | 0.166 |
| 15 | 3·5 | 8 | 9 | 9.92 | +0.92 | 0.645 |
| 22 | 2·11 | 11 | 12 | 11.14 | −0.86 | 0.110 |
| 39 | 3·13 | 12 | 13 | 13.51 | +0.51 | 0.508 |
| 77 | 7·11 | 16 | 17 | 15.90 | −1.10 | 0.082 |
| 82 | 2·41 | 15 | 16 | 16.23 | +0.23 | 0.401 |
Same margin-driven structure. cm range [0.063, 0.645]. Fixed-core locking is universal across Ω layers. Paper XXXI's γk is a mixture of fast-decaying balanced subfamilies + slowly-decaying prime-core subfamilies + weight drift.
9. Corrections to Paper XXXI
γk is a mixture effect. Model A's finite-scale "victory" reflects the decay phase of the mixture — fast-decaying subfamilies dominate at current scales. Asymptotically, once these clear, the residual floor emerges — but that floor may itself be slowly decaying (§6), so Model B (qk → constant) is also not guaranteed. True behavior may lie between the two: extremely slow decay.
10. Discussion
10.1 Core Contributions
Nine findings: (1) h₀-free formula + γ₂ confirmation. (2) Prime penalty Δ₂. (3) n=2q discrete locking. (4) 95-prime profile. (5) ρ(p) ↔ P⁺(p−1) smoothness correlation. (6) cp(N) per-decade decay. (7) Harmonic Cesàro criterion. (8) Ω=3 fixed-core universality. (9) γk mixture reinterpretation.
10.2 Open Problems
(1) Is cp(N) decay genuine asymptotic behavior or finite-scale effect? (2) c̄h(N) → 0 or → c̄ > 0? — the ultimate Ω=2 criterion. (3) Higher-Ω fixed-core profiles. (4) Analytic proof of n=2q locking via local limit law. (5) ρ(p) per-prime fluctuations and deep arithmetic structure.
10.3 Acknowledgments
ChatGPT: partition bug fix, family-specific Δp, harmonic Cesàro criterion, canonical decomposition, local limit law target, H(P;N) vs c̄h(N) distinction. Gemini: prime penalty discovery, p−1 smoothness hypothesis, fixed-core generalization. Grok: series consistency. Claude: all scripts (Block 1–10), working notes v1–v3. Final text by author.
11. Data Sources
| Script | Measurement | Section |
|---|---|---|
| p32_block123.py | R_sum + balance + R stats | §2–§3 |
| p32_block45.py | G distribution + γ₂ + n=2q | §2–§4 |
| p32_block5b_fix.py | n=2q threshold fix + per-decade | §4 |
| p32_block67.py | Profile (p≤23) + Ω=3 + weighted sum | §5, §7, §8 |
| p32_block8.py | Extended profile (p≤200, 46 primes) | §5 |
| p32_block9.py | p−1 smoothness (p≤200) | §5.4 |
| p32_block10.py | Harmonic avg (p≤499, 95 primes) + Ω=3 ext | §6–§8 |
Sanity: ρE(10⁷) = 58, ρE(10⁸) = 66.
References
[1] ZFCρ Papers I–XXXI. H. Qin. Paper XXXI DOI: 10.5281/zenodo.19104860.
[2] G. H. Hardy, S. Ramanujan (1917). Quart. J. Math., 48:76–92.
[3] K. Cordwell et al. (2018). J. Number Theory, 189:17–34.
我们对 Ω=2 层(半素数 n=pq)进行微观解剖,发现 H' 的判据可以精确归结为一个 prime-core harmonic 平均问题。
建立 h₀-free gap 公式:G(n) = ρ(p) + ρ(q) − ρ(n−1) + 1。j(n)>0 当且仅当 G(n)<0。独立确认 γ₂ = 0.453(R² = 0.996)。G 是整数值,集中在 {−1, 0, +1, +2}(占 87%)。
发现素数惩罚:素数 q 的 ρ(q) 系统性高于局部均值(E[R(p)] ≈ 4.57 vs E[R(n)] ≈ 2.76),因为素数是 DP 中的纯 additive 注入点。这解释了 n=2q 的 E[diff] = 1.83(远低于 h₀·ln 2 = 2.48)。
发现 n=2q 的离散锁定:diff = ρ(2q−1) − ρ(q) 集中在 {1, 2, 3}(占 92%)。P(j>0) = P(diff≥3) ≈ 0.211,五个 decade 完全冻结。在当前可见尺度上,p=2 是 Ω=2 层中唯一没有显示明确下行趋势的 prime-core family。
建立 prime-core profile:对 95 个素数(p=2 到 499),测量 cp = P(j>0|n=pq)。cp 主要由 margin = E[diff] − threshold 决定(Corr = 0.917)。margin 的振荡由 ρ(p) 相对于 h₀·ln p 的离散偏差驱动——而 ρ(p) 的偏差由 p−1 的平滑度决定(Corr(ln P⁺(p−1), cp) = −0.714)。p−1 平滑 → ρ(p) 低 → margin 高 → cp 高。p−1 粗糙 → ρ(p) 高 → margin 低 → cp 低。
关键的新发现:除 p=2 外,几乎所有 p ≥ 3 的 cp(N) 都在逐 decade 下降(2–15%/decade)——cp 不是常数,是 N-dependent 的缓慢衰减量。固定 N=10⁸ 下的 partial harmonic 平均 H(P;N) = Σ cp/p ÷ Σ 1/p 从 0.35 缓慢下降到 0.27(P=499)——在 P 方向下降但远未趋零。
H' 在 Ω=2 层的判据精确化为:harmonic Cesàro 平均 c̄h(N) → 0?当前数据无法判定——cp 的衰减是真实的渐近行为还是有限尺度效应。
Ω=3 层的 30 个 squarefree fixed-core(m=6, 10, 14, …, 95)展示相同的 margin 驱动结构,强烈支持 fixed-core 锁定是更普遍的现象。
Paper XXXI 的 γk 是 fixed-core families 的混合效应。Model A 在有限尺度上的"胜出"反映的是混合的衰减阶段。
关键词:加法复杂度,素数惩罚,离散锁定,prime-core profile,harmonic Cesàro 平均,fixed-core family,integer complexity
§1 引言
1.1 背景
Paper XXXI 建立了 H' 的逐层瓶颈地图,提取了 γk 并提出了有效指数 Γ = λ·γk 判据。但 γk 被发现是混合效应。本文对最简单的 Ω=2 层进行微观解剖,揭示混合的内部结构。
1.2 本文贡献
(A) h₀-free gap 公式与 γ₂ 独立确认(§2)。(B) 素数惩罚的发现与量化(§3)。(C) n=2q 离散锁定机制(§4)。(D) 95 个素数的 prime-core profile(§5)。(E) ρ(p) 偏差与 p−1 平滑度的强相关(§5)。(F) cp(N) 的逐 decade 衰减——非冻结常数(§6)。(G) Harmonic Cesàro 平均作为 H' 判据(§7)。(H) Ω=3 层的 fixed-core 结构(§8)。
§2 h₀-Free Gap 公式与 γ₂ 确认
2.1 公式
对半素数 n = pq(p ≤ q),约定 p = P⁻(n)(最小素因子)作为规范 core:
$$G(n) = \rho(p) + \rho(q) - \rho(n-1) + 1$$j(n) > 0 当且仅当 G(n) < 0。不涉及 h₀。
2.2 G 的离散性
数值观察 1。 G 是整数值。在 [10⁷, 10⁸):
| G | −1 | 0 | +1 | +2 | 其余 |
|---|---|---|---|---|---|
| 占比 | 18.5% | 30.1% | 24.2% | 14.4% | 12.8% |
四个值占 87%。中位数 = 0,均值 = +0.53。
2.3 γ₂ 独立确认
$$q_2(N) \sim 0.877\cdot(\ln N)^{-0.453}, \quad R^2 = 0.996$$
14 个点(N = 10³ 到 10⁸)。与 Paper XXXI 的 γ₂ = 0.449 吻合。
2.4 平衡半素数的快速衰减
Ω=2 层并非单一群体。限定 q/p < 10 的"平衡半素数"(p 与 q 量级相近)呈现完全不同的行为。
理论:对 p ≈ q ≈ √n,ρ(p) + ρ(q) ≈ 2h₀·ln(√n) = h₀·ln n ≈ ρ(n−1),因此 G(n) ≈ +1——结构性正值。
[10⁷, 10⁸) 的平衡半素数:E[G] = +1.23, P(j>0) = 0.107。
$$q_2^{\text{bal}}(N) \sim 3.03\cdot(\ln N)^{-1.14}, \quad R^2 = 0.992$$γbal = 1.14——比整层的 γ₂ = 0.45 快 2.5 倍。这直接证明 γ₂ 是快速衰减的平衡子群与缓慢衰减/冻结的不平衡子群的混合物。
§3 素数惩罚
Block 3b 数据(N = 10⁸):
| 类别 | E[R(·)] |
|---|---|
| 一般整数 n ∈ [10⁷, 10⁸) | 2.76 |
| 素数 p ∈ [10⁷, 10⁸) | 4.57 |
素数惩罚:差距 ≈ 1.8。素数是 DP 中的纯 additive 注入点:ρ(p) = ρ(p−1) + 1。素数不经过乘法压缩——而合数有多条乘法路径可以压缩 ρ。因此素数的 ρ 相对于其对数尺度偏高。
对 n=2q:E[diff₂] ≈ h₀·ln 2 − Δ₂ ≈ 2.48 − 0.65 ≈ 1.83,其中 Δ₂ ≈ 0.65 是 p=2 family 的 prime-conditioned 修正(进入 diff 公式时的净效应),不同于 §3 的无条件差距 1.8。
§4 n=2q 的离散锁定
4.1 阈值
因为 ρ(2) = 1:G = 2 + ρ(q) − ρ(2q−1)。故 j>0 当且仅当 diff ≥ 3。
4.2 diff 分布
数值观察 2。 [10⁷, 10⁸),265 万个素数 q:
| diff | +1 | +2 | +3 | 其余 |
|---|---|---|---|---|
| 占比 | 21.6% | 49.5% | 21.1% | 7.8% |
三个值占 92%。P(j>0) = P(diff≥3) ≈ 21.1%(其中 diff=3 贡献绝大部分,更高 tail 极小)。
4.3 五个 Decade 确认冻结
| Decade | E[diff] | P(j>0) |
|---|---|---|
| [10³, 10⁴) | 1.843 | 0.195 |
| [10⁵, 10⁶) | 1.840 | 0.209 |
| [10⁷, 10⁸) | 1.829 | 0.211 |
P(j>0) ≈ 0.211 五个 decade 完全冻结。在当前可见尺度上,p=2 是 Ω=2 层中唯一没有显示明确下行趋势的 prime-core family。
§5 Prime-Core Profile
5.1 定义
对每个素数 p,规范分解 n = pq(q ≥ p 素数,p = P⁻(n)),定义:
$$c_p(N) = P(j>0 \mid n=pq,\; pq \leq N)$$ $$\text{threshold}_p = \rho(p)+1, \quad \text{margin}_p = E[\text{diff}_p] - \text{threshold}_p$$当 q=p 时 n=p²——这些 prime square 渐近 negligible(<0.01%),排除。
5.2 95 个素数的 Profile(选录)
数值观察 3。 N = 10⁸:
| p | ρ(p) | thr | E[diff] | margin | cp | P⁺(p−1) | p−1 |
|---|---|---|---|---|---|---|---|
| 2 | 1 | 2 | 1.83 | −0.17 | 0.211 | 1 | 1 |
| 3 | 2 | 3 | 3.52 | +0.52 | 0.543 | 2 | 2 |
| 5 | 4 | 5 | 5.33 | +0.33 | 0.471 | 2 | 4=2² |
| 7 | 6 | 7 | 6.58 | −0.42 | 0.224 | 3 | 6=2·3 |
| 23 | 12 | 13 | 11.16 | −1.84 | 0.035 | 11 | 22=2·11 |
| 59 | 16 | 17 | 14.80 | −2.20 | 0.019 | 29 | 58=2·29 |
| 109 | 16 | 17 | 17.19 | +0.19 | 0.381 | 3 | 108=2²·3³ |
| 179 | 21 | 22 | 19.11 | −2.89 | 0.005 | 89 | 178=2·89 |
| 193 | 18 | 19 | 19.41 | +0.41 | 0.471 | 3 | 192=2⁶·3 |
cp 范围:[0.005, 0.543]。完整 95 点 profile。
5.3 cp 由 Margin 主导
Corr(margin, cp) = +0.917。 margin 是第一主导变量。局部离散分布的宽度和偏斜贡献二阶修正。
5.4 ρ(p) 的偏差由 p−1 的平滑度驱动
数值观察 4。
Corr(ln P⁺(p−1), ρ(p)/h₀ ln p) = +0.683
Corr(ln P⁺(p−1), cp) = −0.714
高相关支持的解释链(非严格因果证明):p−1 平滑 → ρ(p−1) 低 → ρ(p) = ρ(p−1)+1 低 → threshold 低 → margin 高 → cp 高。p−1 粗糙 → 反向。
极端案例:p=193(p−1=192=2⁶·3,P⁺=3)→ cp=0.471。p=179(p−1=178=2·89,P⁺=89)→ cp=0.005。
5.5 当前范围内无 Monotone Drift
在当前可见范围内(p ≤ 499),ρ(p)/(h₀·ln p) 在 0.95–1.13 之间振荡。没有观察到支持 monotone scissors picture 的系统性 drift。由于 p−1 的最大素因子在解析数论中已知不收敛,cp 不太可能因简单的"h₀·ln p 远超 ρ(p)"机制而系统性趋零——但不能排除 cp 的 harmonic 平均因更精细的分布效应而缓慢下降。
§6 cp(N) 的逐 Decade 衰减
数值观察 5。 逐 decade 的 cp:
| p | cp(10³) | cp(10⁵) | cp(10⁷) | 趋势 |
|---|---|---|---|---|
| 2 | 0.195 | 0.209 | 0.211 | 冻结 |
| 3 | 0.558 | 0.551 | 0.542 | 缓降 |
| 5 | 0.545 | 0.486 | 0.470 | 明确下降 |
| 7 | 0.267 | 0.239 | 0.223 | 明确下降 |
| 13 | 0.339 | 0.333 | 0.308 | 缓降 |
| 109 | 0.465 | 0.430 | 0.377 | 快速下降 |
| 193 | 0.642 | 0.521 | 0.467 | 快速下降 |
除 p=2 冻结外,几乎所有 p ≥ 3 的 cp(N) 都在下降。下降率从 2%/decade(p=3)到 15%/decade(p=193)不等。cp(N) 不是精确的算术常数——是 N-dependent 的缓慢衰减量。当前数据无法区分"cp → 常数 > 0"和"cp → 0 极慢"。这是 H' 的核心悬念。
§7 H' 的 Harmonic Cesàro 平均判据
7.1 Ω=2 层的规范分解
每个 Ω=2 合数 n=pq(p=P⁻(n), q=P⁺(n), q≥p)唯一对应一个 prime-core p。Prime squares(n=p²)贡献 <0.01%,渐近 negligible。排除后,在 squarefree semiprimes 上:
$$q_2^{(\text{sqf})}(N) = \sum_{p} w_p(N)\cdot c_p(N)$$其中 wp(N) ≈ (1/p)/Σ(1/ℓ) 渐近。由于 prime squares 的贡献 → 0,q₂(N) 与 \(q_2^{(\text{sqf})}(N)\) 渐近行为相同。
7.2 Harmonic Cesàro 平均
q₂ → 0 的正确判据是 harmonic Cesàro 平均趋零:
$$\bar{c}_h(N) := \frac{\displaystyle\sum_{p \leq \sqrt{N}} c_p(N)/p}{\displaystyle\sum_{p \leq \sqrt{N}} 1/p} \to 0$$
这不等价于"Σ cp/p 收敛"——后者只是充分条件。如果 cp(N) ~ C/ln ln p,Σ cp/p 仍然发散但 c̄h → 0。
7.3 固定 N 下的 Partial Profile
数值观察 6。 H(P;N) 随截断上限 P 的变化(固定 N=10⁸):
| P | H(P; N) |
|---|---|
| 10 | 0.351 |
| 50 | 0.305 |
| 100 | 0.291 |
| 200 | 0.281 |
| 500 | 0.270 |
H 从 0.35 缓慢下降到 0.27。这是在固定 N 下将更多 prime-core 纳入时 profile 的形状——不是 c̄h(N) 随 N 的渐近行为。真正的 H' 判据需要理解联合 profile (p, N) ↦ cp(N) 的全局行为。
7.4 对 H' 的意义——两变量问题
c̄h(N) 同时依赖于(a)每个固定 p 的 cp(N) 是否随 N 衰减,和(b)截断上限 √N 扩大时新纳入的大 p 的 cp 值。
§6 数据:几乎所有 p ≥ 3 的 cp(N) 在逐 decade 下降(N 方向有利)。§7.3 数据:H(P; 10⁸) 随 P 也在下降(P 方向有利)。但衰减率都太小,无法确认是渐近真实还是有限尺度效应。
当前数据无法判定 H' 在 Ω=2 的真伪。这是一个关于 DP 递推深层结构的开放问题。
7.5 加权和部分验证
Block 7b:Σ wp·cp (p ≤ 23) = 0.174。实际 q₂ = 0.232。覆盖 54%,比值 = 0.75。加权和方向正确——p > 23 的贡献补上后应接近闭合。
§8 Ω=3 层的 Fixed-Core 结构
8.1 框架
对 Ω=3 层,规范 core m(Ω(m)=2,squarefree),n = mq(q = P⁺(n)):
$$G_m(q) = \rho(m) + \rho(q) - \rho(mq-1) + 1, \quad j>0 \;\text{当且仅当}\; \text{diff}_m \geq \rho(m)+2$$注意:当前只覆盖 squarefree core m = p₁p₂。squareful families(如 n=p²q)未纳入——渐近上 squarefree 占主导,但完整分解需要进一步工作。
8.2 30 个 Core 的 Profile(选录)
数值观察 7。 N = 10⁸:
| m | 因子 | ρ(m) | thr | E[diff] | margin | cm |
|---|---|---|---|---|---|---|
| 6 | 2·3 | 5 | 6 | 6.37 | +0.37 | 0.457 |
| 10 | 2·5 | 7 | 8 | 8.25 | +0.25 | 0.395 |
| 14 | 2·7 | 9 | 10 | 9.41 | −0.59 | 0.166 |
| 15 | 3·5 | 8 | 9 | 9.92 | +0.92 | 0.645 |
| 22 | 2·11 | 11 | 12 | 11.14 | −0.86 | 0.110 |
| 39 | 3·13 | 12 | 13 | 13.51 | +0.51 | 0.508 |
| 77 | 7·11 | 16 | 17 | 15.90 | −1.10 | 0.082 |
| 82 | 2·41 | 15 | 16 | 16.23 | +0.23 | 0.401 |
同样的 margin 驱动结构。cm 范围 [0.063, 0.645]。Fixed-core 锁定是更普遍的现象。Paper XXXI 的 γk 是快速衰减子群 + 缓慢衰减 prime-core + 权重漂移三种叠加效应的观察。
§9 对 Paper XXXI 的修正
9.1 γk 是混合效应
Paper XXXI 的 γk 是 fixed-core families 的有效叠加指数。真正的本征量是每个 family 的 cm(N) 和权重 wm(N)。
9.2 Model A vs Model B
Paper XXXI 的 Model A(qk ~ (ln N)−γk)在有限尺度上的"胜出"反映了混合的衰减阶段——快衰减子群主导。渐近地,一旦快衰减清零,底板暴露——但底板本身可能也在缓慢衰减(§6),所以 Model B(qk → 常数 > 0)也不一定最终胜出。真实行为可能是介于两者之间的极慢衰减。
§10 讨论
10.1 核心贡献
(1) h₀-free gap 公式 + γ₂ = 0.453 独立确认。(2) 素数惩罚与 family-specific Δ₂ 的发现——解释 E[diff₂] = 1.83。(3) n=2q 离散锁定(c₂ ≈ 0.211 冻结)。(4) 95 个素数的 prime-core profile:cp 由 margin 主导(Corr = 0.917)。(5) ρ(p) 偏差 ↔ P⁺(p−1) 平滑度的强相关(Corr = −0.714)。(6) cp(N) 对 p≥3 的逐 decade 衰减——非冻结常数。(7) H' 判据:harmonic Cesàro 平均 c̄h(N) → 0?(8) Ω=3 的 30 个 core 展示同样的 margin 驱动结构。(9) Paper XXXI 的 γk 是混合效应。
10.2 开放问题
(1) cp(N) 的衰减是否是真实的渐近行为?还是有限尺度效应?(2) c̄h(N) → 0 还是 c̄h(N) → c̄ > 0?——H' 在 Ω=2 的最终判据。(3) 更高 Ω 层的 fixed-core profile 是否有类似结构和衰减?(4) n=2q 的冻结是否可以从 diff 的整数值局部极限定理解析证明?(5) ρ(p) 的逐素数波动——是否与 Artin 猜想或其他深层算术结构有联系?
10.3 致谢
ChatGPT 贡献了 partition bug 修正、Δp 的 family-specific 定义、harmonic Cesàro 平均作为正确判据、规范 core 分解的明确化、以及 local limit law 作为解析目标。Gemini 贡献了素数惩罚的发现、p−1 平滑度 → ρ(p) 偏差的因果链假说、以及 fixed-core family 的通用化框架。Grok 保证了与前 31 篇的一致性。Claude 编写了全部数值计算脚本(Block 1–10)并起草了 working notes v1–v3。最终文本由作者独立完成,所有数学判断由作者负责。
§11 数据来源与可复现性
| 脚本 | 测量 | §引用 |
|---|---|---|
| p32_block123.py | R_sum 分布 + 平衡度 + R 统计 | §2–§3 |
| p32_block45.py | G 分布 + γ₂ + n=2q | §2–§4 |
| p32_block5b_fix.py | n=2q 阈值修正 + per-decade 冻结 | §4 |
| p32_block67.py | Prime-core profile (p≤23) + Ω=3 + 加权和 | §5, §7, §8 |
| p32_block8.py | 扩展 profile (p≤200, 46 primes) | §5 |
| p32_block9.py | p−1 平滑度 vs ρ(p) 偏差 (p≤200) | §5.4 |
| p32_block10.py | Harmonic 平均 (p≤499, 95 primes) + per-decade + Ω=3 扩展 | §6–§8 |
所有脚本 sanity check 通过:ρE(10⁷) = 58, ρE(10⁸) = 66。
References
[1] ZFCρ Papers I–XXXI. H. Qin. Paper XXXI DOI: 10.5281/zenodo.19104860.
[2] G. H. Hardy, S. Ramanujan (1917). Quart. J. Math., 48:76–92.
[3] K. Cordwell et al. (2018). J. Number Theory, 189:17–34.