diff --git a/notes/issue-636-width4-next-brief-20260912.md b/notes/issue-636-width4-next-brief-20260912.md new file mode 100644 index 00000000..8689a901 --- /dev/null +++ b/notes/issue-636-width4-next-brief-20260912.md @@ -0,0 +1,54 @@ +# Suggested re-scope of existing #636: width-four annular/lifted continuation closure + +This is a proposed next brief for the existing issue, not a new duplicate issue, +production authorization, or claim that a weighted engine has already been built. +It follows the delivered width-three exact analysis. Ordinary CPU only; no GPU, +no width scan, no new Monte Carlo, no free-exponent fit. + +## One decision + +Can a specified local frontier state, carrying appropriate connectivity and +horizontal topological information, preserve the square-site ambient rank under +arbitrary continuation and periodic closure at circumference four? Supply a +sufficient construction with proof, or a counterexample to a precisely stated +candidate quotient. Mere small-size agreement is not the closure proof. + +## Two compulsory witnesses, already solved + +Bit j denotes column j. First compare open prefixes [0,5] and [7,5], followed +by the common suffix [13,0]. Their final 4x4 ranks are 0 and 1. Frontier occupancy +alone is insufficient. + +The stronger pair is [13,5] and [7,5]. Both have five occupied sites, frontier +mask 5={0,2}, ordinary frontier partition {{0,2}}, and no existing transverse +winding. Their lifted 0-to-2 connecting paths have displacements -2 and +2. +Append [13,0] and close through the empty row: ranks 0 and 1, D=-1 and 0. +A state that tracks only the ordinary partition, occupation count and already +completed wrapping flag STILL fails. Both prefix weights are p^5(1-p)^3, so +weighting cannot compensate for identifying the two histories. + +## Deliverable + +Define the state, local row update, equivalence relation and periodic closure +functional; explicitly retain enough annular or lift information to distinguish +these histories. Prove future equivalence for every pair the representation does +identify. Explain how primitive winding, independent windings and rank saturation +are preserved without silently truncating necessary integer offsets. + +Then implement only a bounded width-four reference calculation. Check the existing +4x4 Bernstein polynomial and a small non-square torus by an independent lifted +traversal. Preserve parallel periodic edges. Use integer/rational coefficients; +no asymptotic exponent or p_c claim. If a finite state bound has not been proved, +label it and do not start enumerating more widths. + +If a pTL formulation from Jacobsen is reused, give the actual map between its +states/closure and the black-site rank observable; matching a root numerically +is not an intertwiner. Output a full finite probability closure before discussing +leading eigenvalues; amplitudes and all subleading terms must remain explicit. + +## Stop + +Stop at a demonstrated sufficient state and closure, or one precise obstruction. +Do not claim minimal state count unless a distinguishability argument proves it. +Do not re-prove the width-three formula or pay for another all-honest-torus +five-cell census. A broader engine needs a separate information-gain decision. diff --git a/notes/review-next-frontier-20260912-zh.md b/notes/review-next-frontier-20260912-zh.md new file mode 100644 index 00000000..96142071 --- /dev/null +++ b/notes/review-next-frontier-20260912-zh.md @@ -0,0 +1,101 @@ +# Matching One:新交付复核与下一步实质推进 + +日期:2026-09-12。读取对象:PR #705、#706、其当前文件、评审评论和实际 Actions 结果;本报告不是只复述 PR 开头正文。 + +## 一、总体判断 + +#705 已合并,给出了宽度二的完整有限长度谱公式,而不是只有一个数值根。#706 的有效结论是:N145/290/725 同一方格点渗流观测量的归一化形状仍不对称,且不对称范数沿这三个尺寸减弱;但“新修正场已被识别”或“任意共同非线性坐标都被排除”仍不成立。 + +本次没有再给已经完成的工作开单。新增工作的主要结果是:完成宽度三的全长度精确公式、有限根位移定理,并在宽度四给出必须保留历史连通性和路径绕行信息的显式反例。代码、整数多项式、精确区间证书及测试随包提供,交由项目方提交。本轮未修改远端仓库,也未运行新生产模拟。 + +## 二、#706 应如何接收 + +读取 head:59fff91c6592e8c947ccd1e3df25e15df6a291e5。完整 CI 34685638158 已完成且成功;实际 jobs 中 Tests、smoke、Python 3.9/3.13 编译与 C++17 构建均成功。不能把 CI 成功解释成极端名义显著性的有限样本标定已经完成。 + +PR 开头正文仍有被后续文件修正的旧数字,合并时应改正文和旧 note 的显眼导航,而不是重新生成或覆盖原始结果。 + +1. 宽度标准误差的原公式把 delete-one 样本当成独立观测均值,低估了 99 倍。正确标准误差依次是 3.6730e-6、2.7377e-6、2.3226e-6,已经由非线性 jackknife 复核。 +2. 两个相邻区间共享 N290。比较区间位移范数时必须保留负交叉协方差。修正后 Z 的范数变化约 9.04 个名义标准误差,而不是正文的 10.9;A 的对应变化约 0.897 个标准误差,仍未分辨出来。 +3. “完整逆协方差统计量小于对角和”不能证明逆矩阵不受近零方向影响。单坐标的大信号与完整逆矩阵的校准是不同问题。 + +这些修正已经在 notes/shape-lineage-review-20260912.md 和 results/research-control-20260912/shape-lineage-reviewed-summary.json 中,不应再作为新的待修复发现派发。 + +## 三、比继续拟合一个指数更重要的解释边界 + +修订结果给出 ||A||/W = 0.255963、0.256397、0.256847。这是有用的结构迹象,但完整向量的 A/W 恒定模型并不成立。 + +仅用 N145 中位对称条件确定的二次坐标 phi(p)=p+0.2684067(p-1/2)^2,在 N290/N725 上把不对称残差范数降到原来的约 0.267%/0.068%。这说明大部分范数可以被一个简单坐标形变吸收;但该精确二次模型仍被完整向量诊断拒绝。范数减少不是解释方差、因果归因,也不是已经给出了物理温度坐标。 + +对任意共同递增坐标,现有两个不变量方向——共同对称化反射的交换关系及完整位置-尺度传输的无交叉条件——均未分辨出反例。不能用未拒绝支持“形状纯粹是坐标”,也不应不断提高多项式阶数寻找事后兼容。 + +最后一个传输检验已经实际执行,但原 note 仍写“结果须从 job 读取”。本轮读取 run 34685638146 / job 103532033981 的成功日志并归档:在固定 u=.3,.5,.7 处,组合差分别约 -4.2494e-6、-4.4743e-6、-4.5533e-6,标准误差 1.9009e-6、1.8363e-6、1.8494e-6。三点同时区间均包含零;最大边际标准化值 2.462。没有分辨出符号交叉,因此该必要条件未能排除完整共同坐标塌缩。不是重新跑了一次生产,也不是新的独立证据。 + +本包的 shape-transport-order-finalized-summary.json 是上述已完成日志的字段摘录,明确保留 run、job、head、checkout 和协方差来源。没有把它伪装成本地原始直方图重算。 + +## 四、本次直接完成的数学推进:宽度三 + +对 3×m 正常环面,m≥2,令 F 为“出现全占据行”,V 为“每个相邻行接口都有占据重叠”。我证明了配置级恒等式 + + r = 1_F + 1_V. + +宽度三的每个非空行都连通,非满行则是可缩树;这一特殊几何事实是证明核心。由七个非空行状态的局部转移矩阵 T,得到 + + M_(3,m)(p) = tr(T(p)^m) - (1-p^3)^m. + +这不是穷举外推,而是对所有 m 的事件分类证明。S3 对称性进一步把完整迹分成一个三维块 B 和出现两次的二维块 C: + + tr(T^m)=tr(B^m)+2 tr(C^m). + +圆柱交点由六次方程 + + p^6-3p^5-5p^4-4p^3+p+1=0 + +的唯一 (0,1) 根给出:q3=0.58888069991785299805144269575170493372217…,与 Jacobsen 2015 Table 2 的已发表 n=3 数值一致。这个数不是无限方格点渗流 p_c 的新估计,也不声称发现了新临界常数。 + +保留全部子领先谱后,又得到:每个有限长度根都严格小于 q3;其领先位移为 + + p_(3,m)-q3 ~ -2*rho^m/(m*h'(q3)), + rho=0.22476579734522079…, + h'(q3)=2.69049781437123949…. + +系数 2 来自普通表示重数,不是 Jordan 块。转移矩阵与实对称矩阵相似,因而在 0
= 2. Preserve parallel lifted edges when m=2. Let r be its ambient rational +homology rank. The existing digital-Alexander identity gives D=r-1 and M=E[D]. +Write P_j=P(r=j). Occupation probability is p, not an FK bond fugacity. + +There are seven nonempty occupied subsets S of a three-site row. Define + + u=p(1-p)^2, v=p^2(1-p), w=p^3, + t(S)=p^|S| (1-p)^(3-|S|), + T[S,S'] = 1{S intersects S'} t(S'). + +Thus tr(T^m) is the Bernoulli probability that every row is nonempty and every +successive pair, including the seam, has occupied overlap. T is nonnegative, +irreducible, and aperiodic for 0
0
+and P(1)<0. The rational certificate in the script isolates q and the three B
+roots, showing that lambda_c(q) is the Perron root, not a subleading crossing.
+
+ q=0.58888069991785299805144269575170493372217050345701...
+
+Jacobsen (2015), section 6.1 Table 2, already prints
+0.5888806999178529980514426957517049337221 for n=3. The agreement is with a
+published FINITE-WIDTH estimator. It neither finds a new value of p_c nor proves
+an all-width pTL intertwiner. Equation (5) is called a defining sextic here;
+no unproved assertion of polynomial minimality is needed.
+
+## 5. Every finite root lies below q; exponential displacement with alternating corrections
+
+At q the five distinct eigenvalues are approximately
+
+ beta_1=lambda_c=0.7957876689642490361,
+ beta_2= 0.0362139729093206924,
+ beta_3=-0.1005527942557272624,
+ mu_+ = 0.1788658499322440484, multiplicity 2,
+ mu_- =-0.0793337764501209504, multiplicity 2.
+
+The accompanying Fraction interval calculation proves, rather than infers from
+these decimals,
+
+ beta_2>0, beta_3<0, mu_+>0, mu_-<0,
+ |beta_3|/mu_+ < 3/5, |mu_-|/mu_+ < 1/2,
+ beta_2 < mu_+ < lambda_c.
+
+Consequently M_(3,m)(q)>0 for every m>=2. For even m every remaining term in
+(4) is positive. For odd m, divide by mu_+^m and bound the negative terms by
+(3/5)^m+2(1/2)^m < 2. The finite M is strictly increasing by the monotone-rank
+argument and has endpoint values -1,+1. Its unique root p_(3,m) therefore
+satisfies p_(3,m)0.
+
+The derivative sign is also certified exactly: at q,
+
+ h'(q)=-(q-1)^2 P'(q)/(f_B,L(lambda_c(q),q)*lambda_c(q)).
+
+The numerator derivative is negative and the denominator positive. Analyticity
+of the separated eigenvalues near q and a Taylor expansion of (4)/lambda_c^m
+give
+
+ p_(3,m)-q = -2 rho^m/(m h'(q)) * [
+ 1 + (mu_-/mu_+)^m
+ + (beta_2/mu_+)^m/2 + (beta_3/mu_+)^m/2
+ + O(rho^m)]. (6)
+
+All ratios on the right are evaluated at q. The O(rho^m) relative error accounts
+for evaluation at the displaced root and the nonlinear leading exponential.
+This is a fixed-width asymptotic, with a locally uniform spectral gap; it says
+nothing about a width-uniform bound or the fixed-aspect L^-4 conjecture.
+
+The beta_3 term supplies the largest relative spectral correction and alternates
+with parity. Examples are generated, not fitted:
+
+| m | p_(3,m), diagnostic | actual shift / leading shift |
+|---|---|---|
+| 3 | 0.5865114551126756357 | 0.84206024 |
+| 4 | 0.5883619852843523124 | 1.09362955 |
+| 8 | 0.5888800907157023131 | 1.00649648 |
+| 12 | 0.5888806988874187696 | 1.00055615 |
+| 20 | 0.5888806999178489730 | 1.00000506 |
+
+Exact rational root brackets are stored alongside each diagnostic decimal.
+
+## 6. Why the occupancy-only simplification ends at width four
+
+At width four a proper occupied row may be disconnected. Two short counterexamples
+show that merely counting full rows and successive overlaps can err in either direction.
+Here integer masks use bit j for column j (columns 0 through 3).
+
+* Rows [1,5,4,5] on the 4x4 torus: all interfaces overlap, no row is full,
+ but the ambient rank is zero, not one. The middle masks 5={0,2} do not join
+ the incoming and outgoing paths.
+* Rows [11,14] on the honest 4x2 torus: again no full row, all interfaces
+ overlap, but rank is two, not one. The two complementary paths within the
+ rows create transverse winding even though neither row is full. Parallel
+ length-two edges are retained.
+
+There is a stronger continuation witness, not just failure of one guessed formula.
+Take two open prefixes, ending at the identical frontier 5={0,2}:
+
+ history A: [0,5], history B: [7,5].
+
+Neither prefix has ambient winding. History B connects the two frontier vertices
+via row 7={0,1,2}; history A does not. Append the SAME suffix [13,0], where
+13={0,2,3}, and close longitudinally through the empty row:
+
+ [0,5,13,0] has r=0 and D=-1;
+ [7,5,13,0] has r=1 and D=0.
+
+In the second history the new route via column 3 closes a transverse cycle
+against the old route via column 1. Thus ANY deterministic continuation summary
+that identifies these two prefixes cannot preserve this readout. Frontier
+occupancy and a current wrapping flag alone are insufficient; connectivity of
+frontier vertices is genuine needed memory. This is a scoped lower bound, NOT
+an impossibility theorem for local transfer operators or a proof that all full
+boundary partitions are necessary/minimal.
+
+The ordinary labelled frontier partition is not the end of the story. An even
+stronger pair has the same occupation count, same frontier, same ordinary
+partition, and same current winding flag:
+
+ history A': [13,5], history B': [7,5].
+
+Both contain five of eight sites (identical weight p^5(1-p)^3), both connect
+frontier vertices {0,2}, and neither has a transverse cycle. In A' their lifted
+connecting path has horizontal displacement -2; in B' it has displacement +2.
+Attach the same suffix [13,0]. The new 0-to-2 route has displacement -2, hence
+
+ [13,5,13,0] has r=0, D=-1;
+ [ 7,5,13,0] has r=1, D= 0.
+
+The second closed route has net displacement +2-(-2)=4, one circumference;
+the first has displacement zero. The script independently traverses the OPEN
+prefixes and records all equal ordinary fields and the differing integer lifts,
+then verifies the completed torus ranks. This is genuine topological memory
+that a bare set partition and a flag recording already completed wrapping lose.
+Embedded annular or suitably lifted connectivity states can distinguish it;
+this is NOT a no-go for those representations.
+
+The next useful transfer question, if pursued, is therefore a precise width-four
+connectivity-and-topology closure, tested against both continuation witnesses.
+Plain boundary-partition counts alone do not settle the needed state space.
+It is not another occupancy-mask eigenvalue calculation or a large width scan.
+
+## 7. Executed validation and sources
+
+The script derived unnormalized Bernstein integer coefficients by Newton trace
+recurrences in site fugacity. An independent lifted-edge traversal checked all
+299,584 configurations on 3x2,...,3x6, both the pointwise identity (1) and EVERY
+polynomial coefficient. This generalizes the geometric checking technique in
+#705 and is independent of the row-transfer formula, not claimed as an unrelated
+second research program. The existing 3x3 coefficient vector
+[-1,-9,-36,-78,-90,-36,36,36,9,1] is reproduced exactly.
+
+An independent integer 7x7 matrix-power trace at p=1/2 matches the block recurrence
+for m=2,...,10. Fraction interval signs certify the q bracket, the five eigenvalue
+bands, positivity of h'(q), and the inequalities used in the all-m proof. Seven
+small mathematical tests passed locally. No full-repository CI was run for this
+new patch; no new production, GPU, or fitted exponent was used.
+
+Files:
+- scripts/width3_cylinder_exact.py
+- tests/test_width3_cylinder_exact.py
+- results/research-control-20260912/width3-cylinder-exact.json
+
+Repository inputs read: PR #705 at fc19cc748527249f0ce1c69d7cd2a88a4879427b,
+merged as eb89e9422791d9e3c3a78f0e65d56912b815a7bd; original digital-Alexander
+proof; the 3x3 Bernstein rung via #668/#684.
+
+External primary source, PRIMARY_TEXT_READ on 2026-09-12:
+Jacobsen, Critical points of Potts and O(N) models from eigenvalue identities in
+periodic Temperley-Lieb algebras, arXiv:1507.03027v1, section 6.1 Table 2 and
+sections 4,7.1. https://arxiv.org/html/1507.03027v1
+The table value and limit order are literature facts; the all-width or continuum
+claims not established above remain unclaimed. No broad novelty search was done.
diff --git a/results/research-control-20260912/shape-transport-order-finalized-summary.json b/results/research-control-20260912/shape-transport-order-finalized-summary.json
new file mode 100644
index 00000000..cecaaf5c
--- /dev/null
+++ b/results/research-control-20260912/shape-transport-order-finalized-summary.json
@@ -0,0 +1,82 @@
+{
+ "schema": "matching-one.shape-transport-order-finalized-summary.v1",
+ "provenance": {
+ "head": "59fff91c6592e8c947ccd1e3df25e15df6a291e5",
+ "checkout": "93f44c5dfb52d178046f546e9d644fd84e50dfcd",
+ "run_id": 34685638146,
+ "job_id": 103532033981,
+ "command": "python scripts/shape_transport_order.py",
+ "extraction": "Fields read from completed successful workflow stdout. Not a new local raw-histogram run.",
+ "elapsed_seconds": 17.045987239,
+ "mathematical_tests_in_job": 19
+ },
+ "finalization_date": "2026-09-12",
+ "reference_levels": [
+ 0.3,
+ 0.5,
+ 0.7
+ ],
+ "reference_points": [
+ 0.5638915164071909,
+ 0.5927517136919742,
+ 0.6211934034174541
+ ],
+ "composition_difference_high_precision": [
+ -4.249439184159747e-06,
+ -4.474271110694973e-06,
+ -4.553323687558643e-06
+ ],
+ "standard_errors": [
+ 1.900866250374583e-06,
+ 1.836344494162728e-06,
+ 1.8494040220848198e-06
+ ],
+ "covariance": [
+ [
+ 3.613292501813127e-12,
+ 3.4593671499807256e-12,
+ 3.46768682759206e-12
+ ],
+ [
+ 3.4593671499807256e-12,
+ 3.372161101241766e-12,
+ 3.3685088503157967e-12
+ ],
+ [
+ 3.46768682759206e-12,
+ 3.3685088503157967e-12,
+ 3.4202952369035086e-12
+ ]
+ ],
+ "jackknife_bias_estimate": [
+ -1.0533094393337672e-11,
+ -1.2296963397197432e-11,
+ -2.1912731669465483e-11
+ ],
+ "simultaneous_nominal_alpha": 0.0027,
+ "bonferroni_critical_z": 3.3200541166994784,
+ "lower": [
+ -1.0560418004010983e-05,
+ -1.057103420821836e-05,
+ -1.0693445124521922e-05
+ ],
+ "upper": [
+ 2.0615396356914884e-06,
+ 1.622491986828414e-06,
+ 1.586797749404636e-06
+ ],
+ "max_abs_marginal_z": 2.462049197030357,
+ "resolved_crossing": false,
+ "verdict": "UNRESOLVED_FULL_COMMON_COORDINATE_COLLAPSE",
+ "pooled_float_vs_55dps_max": 1.48427446819855e-15,
+ "composition_quantiles_range_all_deletions": [
+ 0.24880019486659927,
+ 0.7510794674016856
+ ],
+ "limits": [
+ "No resolved crossing is not evidence establishing a common coordinate.",
+ "A nonzero constant-sign commutator would not reject affine conjugacy: Aff(1) is nonabelian.",
+ "Estimated nonlinear jackknife covariance and nominal marginal intervals; no exact finite-sample coverage claim.",
+ "Same effective spin0 laws and fixed central points, not a new independent experiment."
+ ]
+}
diff --git a/results/research-control-20260912/width3-cylinder-exact.json b/results/research-control-20260912/width3-cylinder-exact.json
new file mode 100644
index 00000000..45a5d91b
--- /dev/null
+++ b/results/research-control-20260912/width3-cylinder-exact.json
@@ -0,0 +1,486 @@
+{
+ "schema": "matching-one.width3-cylinder-exact.v1",
+ "scope": "square-site width 3, length m>=2; NOT infinite-lattice pc or an all-width pTL intertwiner",
+ "cylinder_sextic_low_first": [
+ 1,
+ 1,
+ 0,
+ -4,
+ -5,
+ -3,
+ 1
+ ],
+ "cylinder_exact_rational_bracket": [
+ "6566239220600206483872888902707641265663279/11150372599265311570767859136324180752990208",
+ "840478620236826429935729779546578082004899713/1427247692705959881058285969449495136382746624"
+ ],
+ "cylinder_q_diagnostic": "0.58888069991785299805144269575170493372217050345701",
+ "published_n3_value": "0.5888806999178529980514426957517049337221",
+ "source": "Jacobsen 2015 arXiv:1507.03027v1 section 6.1 Table 2, primary HTML read 2026-09-12",
+ "published_match_below_1e_minus_40": true,
+ "block_B_eigenvalues_diagnostic": [
+ "0.79578766896424903605620104266078116604873503565708",
+ "0.036213972909320692438479008027583384373767333677894",
+ "-0.10055279425572726240102376097844011798302112947222"
+ ],
+ "block_C_eigenvalues_diagnostic": [
+ "0.17886584993224404842481657728063602638614720762375",
+ "-0.079333776450120950401600026784588924049863141321157"
+ ],
+ "block_C_eigenvalue_multiplicity": 2,
+ "closed_rate_diagnostic": "0.79578766896424903605620104266078116604873503565708",
+ "rho_diagnostic": "0.22476579734522079198176767579637432574092549867206",
+ "hprime_diagnostic": "2.6904978143712394890696018902702841665091753340637",
+ "all_leading_coefficients": 1,
+ "dominant_subleading_coefficient": 2,
+ "exact_enumeration_checks": {
+ "2": {
+ "configurations": 64,
+ "classification_failures": 0,
+ "bernstein_counts": [
+ -1,
+ -6,
+ -12,
+ -6,
+ 6,
+ 6,
+ 1
+ ],
+ "power_coefficients": [
+ -1,
+ 0,
+ 3,
+ 2,
+ -3
+ ]
+ },
+ "3": {
+ "configurations": 512,
+ "classification_failures": 0,
+ "bernstein_counts": [
+ -1,
+ -9,
+ -36,
+ -78,
+ -90,
+ -36,
+ 36,
+ 36,
+ 9,
+ 1
+ ],
+ "power_coefficients": [
+ -1,
+ 0,
+ 0,
+ 6,
+ 0,
+ 0,
+ 0,
+ -18,
+ 18,
+ -4
+ ]
+ },
+ "4": {
+ "configurations": 4096,
+ "classification_failures": 0,
+ "bernstein_counts": [
+ -1,
+ -12,
+ -66,
+ -216,
+ -456,
+ -624,
+ -498,
+ -84,
+ 222,
+ 192,
+ 66,
+ 12,
+ 1
+ ],
+ "power_coefficients": [
+ -1,
+ 0,
+ 0,
+ 4,
+ 3,
+ 0,
+ 6,
+ 0,
+ -87,
+ 136,
+ -72,
+ 12
+ ]
+ },
+ "5": {
+ "configurations": 32768,
+ "classification_failures": 0,
+ "bernstein_counts": [
+ -1,
+ -15,
+ -105,
+ -450,
+ -1305,
+ -2670,
+ -3885,
+ -3885,
+ -2235,
+ 45,
+ 1245,
+ 1020,
+ 420,
+ 105,
+ 15,
+ 1
+ ],
+ "power_coefficients": [
+ -1,
+ 0,
+ 0,
+ 5,
+ 0,
+ 3,
+ -10,
+ 30,
+ 0,
+ -200,
+ 312,
+ -120,
+ -65,
+ 60,
+ -15,
+ 2
+ ]
+ },
+ "6": {
+ "configurations": 262144,
+ "classification_failures": 0,
+ "bernstein_counts": [
+ -1,
+ -18,
+ -153,
+ -810,
+ -2970,
+ -7938,
+ -15846,
+ -23778,
+ -26478,
+ -20658,
+ -8766,
+ 2232,
+ 6750,
+ 5382,
+ 2520,
+ 774,
+ 153,
+ 18,
+ 1
+ ],
+ "power_coefficients": [
+ -1,
+ 0,
+ 0,
+ 6,
+ 0,
+ 0,
+ -12,
+ 0,
+ 54,
+ 32,
+ -396,
+ 486,
+ 30,
+ -414,
+ 288,
+ -114,
+ 72,
+ -36,
+ 6
+ ]
+ }
+ },
+ "integer_matrix_trace_checks_m_2_to_10": true,
+ "finite_roots": {
+ "2": {
+ "exact_rational_root_bracket": [
+ "400312927426073777010206031846360697872060411/713623846352979940529142984724747568191373312",
+ "800625854852147554020412063692721395744120823/1427247692705959881058285969449495136382746624"
+ ],
+ "root_diagnostic": "0.5609578904515291298145210442389823818645",
+ "root_minus_cylinder_diagnostic": "-0.02792280946632386823692165",
+ "leading_shift_diagnostic": "-0.01877706920495645957168997",
+ "shift_ratio_diagnostic": "1.487069635923441540396485"
+ },
+ "3": {
+ "exact_rational_root_bracket": [
+ "837097121055181458023268930732850654644401453/1427247692705959881058285969449495136382746624",
+ "418548560527590729011634465366425327322200727/713623846352979940529142984724747568191373312"
+ ],
+ "root_diagnostic": "0.5865114551126756356545589766069017348243",
+ "root_minus_cylinder_diagnostic": "-0.002369244805177362396883719",
+ "leading_shift_diagnostic": "-0.002813628621105619791973803",
+ "shift_ratio_diagnostic": "0.8420602446979530384188936"
+ },
+ "4": {
+ "exact_rational_root_bracket": [
+ "839738285972989758802080916395032008952045671/1427247692705959881058285969449495136382746624",
+ "104967285746623719850260114549379001119005709/178405961588244985132285746181186892047843328"
+ ],
+ "root_diagnostic": "0.5883619852843523124446730988582938294302",
+ "root_minus_cylinder_diagnostic": "-0.0005187146335006856067695969",
+ "leading_shift_diagnostic": "-0.0004743056103421040659819628",
+ "shift_ratio_diagnostic": "1.09362955484871977242227"
+ },
+ "5": {
+ "exact_rational_root_bracket": [
+ "420181142003630538273314870547619800015168973/713623846352979940529142984724747568191373312",
+ "840362284007261076546629741095239600030337947/1427247692705959881058285969449495136382746624"
+ ],
+ "root_diagnostic": "0.5887991890279353635088962923004941727878",
+ "root_minus_cylinder_diagnostic": "-0.00008151088991763454254640345",
+ "leading_shift_diagnostic": "-0.00008528614295508369719285866",
+ "shift_ratio_diagnostic": "0.9557342739788642171284672"
+ },
+ "6": {
+ "exact_rational_root_bracket": [
+ "840455282049249677895950177429479812000281615/1427247692705959881058285969449495136382746624",
+ "52528455128078104868496886089342488250017601/89202980794122492566142873090593446023921664"
+ ],
+ "root_diagnostic": "0.5888643480346472844728520094931130823426",
+ "root_minus_cylinder_diagnostic": "-0.00001635188320571357859068626",
+ "leading_shift_diagnostic": "-0.00001597450660316489351440316",
+ "shift_ratio_diagnostic": "1.023623678146899300822827"
+ },
+ "8": {
+ "exact_rational_root_bracket": [
+ "420238875377231236692311921907541377775002945/713623846352979940529142984724747568191373312",
+ "840477750754462473384623843815082755550005891/1427247692705959881058285969449495136382746624"
+ ],
+ "root_diagnostic": "0.5888800907157023131380716130272317396199",
+ "root_minus_cylinder_diagnostic": "-0.0000006092021506849133710827245",
+ "leading_shift_diagnostic": "-0.0000006052700254996210060063077",
+ "shift_ratio_diagnostic": "1.006496480941785588560711"
+ },
+ "12": {
+ "exact_rational_root_bracket": [
+ "840478618766141554937576837740610249769443589/1427247692705959881058285969449495136382746624",
+ "420239309383070777468788418870305124884721795/713623846352979940529142984724747568191373312"
+ ],
+ "root_diagnostic": "0.5888806988874187696311190589824471686986",
+ "root_minus_cylinder_diagnostic": "-1.030434228420323636769258e-9",
+ "leading_shift_diagnostic": "-1.029861467037597738970488e-9",
+ "shift_ratio_diagnostic": "1.000556153813943004503969"
+ },
+ "20": {
+ "exact_rational_root_bracket": [
+ "13132478441200323205557217575903864723271695/22300745198530623141535718272648361505980416",
+ "840478620236820685155661924857847342289388481/1427247692705959881058285969449495136382746624"
+ ],
+ "root_diagnostic": "0.5888806999178489729756992114548407710935",
+ "root_minus_cylinder_diagnostic": "-4.025075743484296864162629e-15",
+ "leading_shift_diagnostic": "-4.025055391908418730765929e-15",
+ "shift_ratio_diagnostic": "1.000005056222560078609715"
+ }
+ },
+ "width4_obstructions": [
+ {
+ "width": 4,
+ "length": 4,
+ "rows_as_bitmasks": [
+ 1,
+ 5,
+ 4,
+ 5
+ ],
+ "occupied_vertices": [
+ 0,
+ 4,
+ 6,
+ 10,
+ 12,
+ 14
+ ],
+ "full_row": false,
+ "every_interface_overlaps": true,
+ "actual_ambient_rank": 0,
+ "naive_rank": 1
+ },
+ {
+ "width": 4,
+ "length": 2,
+ "rows_as_bitmasks": [
+ 11,
+ 14
+ ],
+ "occupied_vertices": [
+ 0,
+ 1,
+ 3,
+ 5,
+ 6,
+ 7
+ ],
+ "full_row": false,
+ "every_interface_overlaps": true,
+ "actual_ambient_rank": 2,
+ "naive_rank": 1
+ }
+ ],
+ "exact_interval_certificate": {
+ "q_rational_interval": [
+ "11777613998357/20000000000000",
+ "29444034995893/50000000000000"
+ ],
+ "q_polynomial_endpoint_signs": [
+ 1,
+ -1
+ ],
+ "eigenvalue_intervals": {
+ "B_negative": {
+ "rational_interval": [
+ "-51/500",
+ "-99/1000"
+ ],
+ "endpoint_signs": [
+ -1,
+ 1
+ ]
+ },
+ "B_small_positive": {
+ "rational_interval": [
+ "7/200",
+ "19/500"
+ ],
+ "endpoint_signs": [
+ 1,
+ -1
+ ]
+ },
+ "B_perron": {
+ "rational_interval": [
+ "397/500",
+ "797/1000"
+ ],
+ "endpoint_signs": [
+ -1,
+ 1
+ ]
+ },
+ "C_negative": {
+ "rational_interval": [
+ "-81/1000",
+ "-39/500"
+ ],
+ "endpoint_signs": [
+ 1,
+ -1
+ ]
+ },
+ "C_positive": {
+ "rational_interval": [
+ "89/500",
+ "9/50"
+ ],
+ "endpoint_signs": [
+ -1,
+ 1
+ ]
+ }
+ },
+ "negative_beta_over_positive_mu_upper": "3/5",
+ "negative_mu_over_positive_mu_upper": "1/2",
+ "P_prime_at_q_sign": -1,
+ "F_lambda_at_perron_sign": 1,
+ "hprime_positive": true,
+ "all_m_at_least_2_defect_positive": true,
+ "arithmetic": "Fraction interval arithmetic; every endpoint sign certified"
+ },
+ "width4_continuation_memory_witness": {
+ "width": 4,
+ "prefixes": [
+ [
+ 0,
+ 5
+ ],
+ [
+ 7,
+ 5
+ ]
+ ],
+ "common_frontier_mask": 5,
+ "common_suffix": [
+ 13,
+ 0
+ ],
+ "completed_torus_ranks": [
+ 0,
+ 1
+ ],
+ "completed_matching_D": [
+ -1,
+ 0
+ ],
+ "conclusion": "Any deterministic continuation state identifying these prefixes cannot preserve the rank readout."
+ },
+ "width4_topology_memory_witness": {
+ "width": 4,
+ "prefixes": [
+ [
+ 13,
+ 5
+ ],
+ [
+ 7,
+ 5
+ ]
+ ],
+ "common_suffix": [
+ 13,
+ 0
+ ],
+ "prefix_signatures": [
+ {
+ "occupied_count": 5,
+ "frontier_mask": 5,
+ "ordinary_frontier_partition": [
+ [
+ 0,
+ 2
+ ]
+ ],
+ "existing_transverse_winding": false,
+ "lifted_frontier_displacements": {
+ "0->2": -2
+ }
+ },
+ {
+ "occupied_count": 5,
+ "frontier_mask": 5,
+ "ordinary_frontier_partition": [
+ [
+ 0,
+ 2
+ ]
+ ],
+ "existing_transverse_winding": false,
+ "lifted_frontier_displacements": {
+ "0->2": 2
+ }
+ }
+ ],
+ "completed_torus_ranks": [
+ 0,
+ 1
+ ],
+ "completed_matching_D": [
+ -1,
+ 0
+ ],
+ "exact_common_prefix_probability": "p^5*(1-p)^3",
+ "conclusion": "Even frontier mask, ordinary set partition, occupation count, and current winding flag do not preserve continuation; lifted/annular path information is needed."
+ },
+ "boundary": "Rows remember occupancy only; width 4 requires within-row/accumulated connectivity. Width-three proof does not justify mask-only transfer at width 4."
+}
diff --git a/scripts/width3_cylinder_exact.py b/scripts/width3_cylinder_exact.py
new file mode 100644
index 00000000..09b766a7
--- /dev/null
+++ b/scripts/width3_cylinder_exact.py
@@ -0,0 +1,393 @@
+#!/usr/bin/env python3
+"""Width-three square-site torus: exact row-transfer spectrum and its scope.
+
+The variable t used by coefficient recurrences is site fugacity. Coefficients
+are the unnormalized Bernstein counts in M(p)=sum c[k] p**k (1-p)**(3*m-k).
+No Monte Carlo, fitted exponent or infinite-lattice threshold estimate is used.
+"""
+from __future__ import annotations
+import argparse
+from fractions import Fraction as F
+from itertools import product
+import json
+from math import comb
+from pathlib import Path
+
+
+def add(*arrays):
+ result = [0] * max(map(len, arrays))
+ for a in arrays:
+ for i, v in enumerate(a):
+ result[i] += v
+ while len(result) > 1 and result[-1] == 0:
+ result.pop()
+ return result
+
+
+def mul(a, b):
+ result = [0] * (len(a) + len(b) - 1)
+ for i, x in enumerate(a):
+ for j, y in enumerate(b):
+ result[i+j] += x*y
+ return add(result)
+
+
+def scale(a, c):
+ return [c*x for x in a]
+
+
+def power(a, n):
+ result = [1]
+ for _ in range(n):
+ result = mul(result, a)
+ return result
+
+
+def evaluate(a, x):
+ result = 0
+ for c in reversed(a):
+ result = result*x + c
+ return result
+
+
+def block_traces(m):
+ """Integer polynomials for tr(B**m), tr(C**m), from Newton identities."""
+ if not isinstance(m, int) or m < 0:
+ raise ValueError('m must be a nonnegative integer')
+ # B eigen-equation L^3 - h L^2 - j L + t^6 = 0.
+ h, j, detneg = [0,1,3,1], [0,0,0,1,2], [0,0,0,0,0,0,1]
+ b = [[3], h, add(mul(h,h),scale(j,2))]
+ for k in range(3, m+1):
+ b.append(add(mul(h,b[k-1]),mul(j,b[k-2]),scale(mul(detneg,b[k-3]),-1)))
+ # C eigen-equation L^2 - t L - t^3 = 0; multiplicity two in the full transfer.
+ c = [[2], [0,1]]
+ for k in range(2,m+1):
+ c.append(add(mul([0,1],c[k-1]),mul([0,0,0,1],c[k-2])))
+ return b[m], c[m]
+
+
+def bernstein_counts(m):
+ if not isinstance(m,int) or m < 2:
+ raise ValueError('the torus requires integer m >= 2')
+ b,c = block_traces(m)
+ out = add(b,scale(c,2),scale(power([1,3,3],m),-1))
+ return out + [0]*(3*m+1-len(out))
+
+
+def power_coefficients(m):
+ n = 3*m
+ out = [0]*(n+1)
+ for k,c in enumerate(bernstein_counts(m)):
+ for j in range(n-k+1):
+ out[k+j] += c*comb(n-k,j)*(-1)**j
+ return add(out)
+
+
+def ambient_rank(mask, width, length):
+ """Independent lift traversal; retains distinct +/- periodic edges.
+
+ Positions are integer lifts of vertices. Every closing edge supplies an
+ ambient period. Rank is computed by integer determinants, not row masks.
+ """
+ if min(width,length)<2:
+ raise ValueError('honest axis periods must both be at least two')
+ positions = {}
+ first = None
+ for root in range(width*length):
+ if not (mask>>root)&1 or root in positions:
+ continue
+ positions[root] = (0,0)
+ stack = [root]
+ while stack:
+ current = stack.pop()
+ x,y = current%width,current//width
+ px,py = positions[current]
+ for dx,dy in ((1,0),(-1,0),(0,1),(0,-1)):
+ vertex = ((y+dy)%length)*width+(x+dx)%width
+ if not (mask>>vertex)&1:
+ continue
+ proposed = (px+dx,py+dy)
+ if vertex not in positions:
+ positions[vertex] = proposed
+ stack.append(vertex)
+ else:
+ wx,wy = proposed[0]-positions[vertex][0],proposed[1]-positions[vertex][1]
+ if wx%width or wy%length:
+ raise AssertionError('cycle displacement is not a torus period')
+ wx,wy=wx//width,wy//length
+ if wx or wy:
+ if first is None:
+ first=(wx,wy)
+ elif first[0]*wy-first[1]*wx:
+ return 2
+ return int(first is not None)
+
+
+def row_events(mask, width, length):
+ rows=[(mask>>(width*j))&((1<