Law B: where the Dumont–Foata ratios stop decreasing near a face of the cube

This record contains the preprint "Law B: where the Dumont–Foata ratios stop decreasing near a face of the cube". Preprint, 18 November 2026, extended 19 November 2026. Not yet independently refereed. The programs, registrations, logs and check scripts are in the companion code record doi:10.5281/zenodo.24410457 STATEMENT Let m_k be the six-variable Dumont–Foata polynomials, written as weighted Motzkin paths, and R_k = m_k / m_(k−1)². The published article proves that R_k decreases in every variable on the cube [1/2, 1]^6 for every k ≥ 3. Below 1/2 that can fail. Law B describes where it fails at a face point, a point where one of the three combined parameters X, Y, Z vanishes linearly in t while the other two and the bottom weight δ_0 stay positive. A direction fails near t = 0 unless it moves the bottom weight while leaving the link fixed to first order; every other direction fails below t*_k ≈ δ_0^(k−1) / (λ_1′(0) · u′_(k−3)), where λ_1′(0) is the slope of the link λ_1 = XYZ at t = 0 and u′_j are the moments of the upper chain, the paths that stay at level 1 or above. Law B is OPEN: it is a registered pass and a measured law, not a theorem. MOTIVATION Near the bottom corner of the cube the limit of monotonicity shrinks very fast with k: about 0.2 at k = 3 and about 0.00001 at k = 8. Law B asks the same question on a face of the cube and answers it with one formula. The idea fits in one sentence. The moment m_k counts weighted paths that either stay at the bottom level or climb into the levels above; when the link between the bottom and the rest is weak, two kinds of path compete, those that stay at the bottom (weight δ_0 per step) and those that make one trip up (weight about λ_1 times the upper levels' own count u′). Monotonicity fails exactly while staying put still wins. The paper is written so that a reader with first-year calculus can follow every step with a pencil. It uses four tools only: counting weighted paths, the quotient rule, a geometric series, and the Gamma integral. Each result is given in the same order: what it says in plain words, the steps by hand at one real point, a case where the method must fail and does, a short "check it yourself" with its answer, and what to take from it. STATUS The paper's first table gives every claim one status word. In summary: Law B at a general face point: OPEN. Law B on 216 fresh point–direction pairs at k = 14 and 18 (test T4): REGISTERED PASS. Law B at k = 13 and 15 (test T12): REGISTERED PASS, and REPLICATED by a second implementation with byte-identical logs. The one-trip expansion of the moments, for any bottom weight (path counting): PROVED. Removing level 0 at a face point leaves the same family at shifted parameters; the corner case (t, t, 1, t, t, 1): PROVED in the companion thresholds paper. The two-term sign function and its root, and Law B as its leading term: DERIVED. The upper chain's moments, closed forms at the three T12 upper chains (Genocchi and Euler numbers), the enclosure Lemma E′, and two-sided bounds on ρ′_j = u′_j / u′_(j−1) for every j: PROVED, given the cited theorem of Gessel and Zeng. The two-term root over Law B lies in an explicit interval for every k ≥ 10 at the four T12 faces: PROVED, given the cited theorem; both ends behave like 1 − π²/k² − c/k³ + …: DERIVED. Which direction sets the threshold at the four T12 faces (the x-type directions, k = 7 to 15): MEASURED. Condition (iv), that no corner pattern fails above the corner's own threshold: PROVED by exact certificate at each k from 3 to 15 and at k = 17; REGISTERED PASS at k = 16, 18, 19 and 20 in a test frozen before the run, with k = 15 as known-answer control. Not yet checked by an independent mathematician. RESULTS Sections 1 to 6 give the law, its derivation by hand, and the evidence. Sections 7 to 11 bound the pieces of the law for every k at the four faces that were tested, find which direction actually sets the threshold, and explain why the bottom corner of the cube wins. Section 12 lists what a proof still needs: a bound on the paths with two or more trips, every face point rather than three, the class of directions that sets the threshold, and the split rule for every k. Section 13 records how each result was found, including two registered tests that failed (T2 and T3) and what each failure taught. The evidence. T4 (14 November 2026): all 144 predicted-failing pairs fail, with exact root over Law B between 0.9736 and 1.0498 at k = 14 and between 0.9846 and 1.0361 at k = 18, inside the registered windows [0.85, 1.15] and [0.90, 1.10]; none of the 72 predicted non-failing pairs fails. T12 (17 November 2026): the x-type directions give 0.9949 to 1.0286 at k = 13 and 1.0009 to 1.0262 at k = 15. APPENDIX: THE MATHEMATICS Setup. A Motzkin path of length k starts and ends at height 0. An up step has weight 1, a level step at height h has weight b_h = 2h² + (X+Y+Z−1)h + δ, and a down step from height h has weight λ_h = h(h+X−1)(h+Y−1)(h+Z−1). The moment m_k is the total weight. Along a direction with derivative D, the quotient rule gives D(m_k / m_(k−1)²) = G_k / m_(k−1)³ with G_k = m_(k−1) Dm_k − 2 m_k Dm_(k−1), so R_k decreases exactly where G_k ≤ 0. One point, by hand. Take P = (x, x̄, y, ȳ, z, z̄) = (1, 1, 1, t, t, 1). Then X = 2, Y = 2, Z = 2t and δ = 1 + 2t, so δ_0 = 1 and λ_1 = XYZ = 8t. The upper chain is the same family at shifted parameters, with b′_0 = 6, b′_1 = 15 and λ′_1 = 18, so its moments are u′_0 = 1, u′_1 = 6, u′_2 = 6² + 18 = 54 and u′_3 = 6³ + 2·6·18 + 18·15 = 216 + 216 + 270 = 702. Step 1, count the paths by trips (PROVED). A path with no trip has weight b_0^n. A path with one trip spends a steps at the bottom, goes up, spends j steps at level 1 or above (total weight u′_j), comes down (weight λ_1), and spends c more steps at the bottom, with a + c = n − 2 − j. So m_n = b_0^n + λ_1 Ŝ_n + O(λ_1²), where Ŝ_n = Σ_(j=0..n−2) (n−1−j) b_0^(n−2−j) u′_j. At P: Ŝ_4 = 3·1 + 2·6 + 1·54 = 69 and Ŝ_5 = 4·1 + 3·6 + 2·54 + 1·702 = 832. Steps 2 and 3, the sign function and its root (DERIVED). Along z the main effect is on the link, Dm_n ≈ Dλ_1 · Ŝ_n. Substituting, G_k / Dλ_1 = b_0^(k−1)(Ŝ_k − 2b_0Ŝ_(k−1)) − λ_1 Ŝ_k Ŝ_(k−1). At P, k = 5, this is 694 − 57408 λ_1, with root λ_1 = 694/57408 = 347/28704, that is t = 347/229632 ≈ 0.0015111. The exact first root of the full G_5 is 0.0014852, so the two-term root is 1.7 per cent high. Step 4, the leading term (DERIVED). The last term u′_(k−3) dominates Ŝ_(k−1), which gives Law B with a correction bracket: t*_k ≈ δ_0^(k−1)/(λ_1′(0) u′_(k−3)) · (1 − 2δ_0/ρ′_(k−3) − 2δ_0/ρ′_(k−2)). At P, k = 5: Law B alone gives 1/(8·54) = 1/432 ≈ 0.0023148, and the bracket is 1 − 2/9 − 2/13 ≈ 0.624, matching the measured ratio 0.642 of the exact root to Law B. As k grows the bracket tends to 1 (0.735 at k = 6, 0.804 at k = 7). Two cases that must fail. Drop the factor δ_0^(k−1): at Q = (1, 1, 2/3, t, t, 2/3), k = 7, the ratio becomes 0.006988, nowhere near 1. This is how the first form of the law failed its registered test. Shift the index to u′_(k−2): at P, k = 7, the ratio becomes 19.2. The index k − 3 is forced, because the longest trip inside m_(k−1) has k − 3 inner steps. Every number above is checked in exact fractions by lawb_derivation_check.py (22 checks, all pass; 12 of 12 deliberate errors caught). Appendices A to D work in the same way: the third-order ratio law (A), a Bernstein certificate of sign (B), pinning the corner's threshold with exact arithmetic (C), and why four magnitude errors all lean the same way (D), each with its own check script. FILES The preprint (PDF, 37 pages). The code record, doi:10.5281/zenodo.24410457, holds lawb_code_record_v4.zip (SHA-256 beginning 4797ce9b8c08a945; SHA256SUMS.txt lists 127 files) under the MIT licence: the registered test T12 (its frozen registration, amendment and freeze record, the program lawb.py, its outputs and their reruns, the result file and the judging script), the check scripts for every section and appendix with their logs and mutation logs, and the registered condition (iv) test. All check scripts use the Python standard library with exact fractions, and each has its own engine, written independently of the code behind the registered tests. AI ASSISTANCE The work was directed by the author. Most of the computational work was done with Claude (Anthropic) and Gemini (Google), with support from Grok (xAI). The text was drafted with AI assistance from the project's records and edited by the author, who takes responsibility for its contents. REFERENCES [1] J. W. Davis, A monotonicity property of the six-variable Dumont–Foata polynomials, version 6, doi:10.5281/zenodo.24158125. [2] J. W. Davis, Monotonicity thresholds for the six-variable Dumont–Foata ratios (the thresholds paper, a companion preprint). [3] I. M. Gessel, J. Zeng, Moments of orthogonal polynomials and exponential generating functions, Ramanuja n J. 61 (2023), no. 2, 675–700, doi:10.1007/s11139-022-00548-6.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-21
DOI
https://doi.org/10.5281/zenodo.24410465
Primary Topic
Advanced Combinatorial Mathematics
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Law B: where the Dumont–Foata ratios stop decreasing near a face of the cube

Joshua W. Davis
Zenodo (CERN European Organization for Nuclear Research)
Advanced Combinatorial Mathematics
preprint

Law B: where the Dumont–Foata ratios stop decreasing near a face of the cube

Joshua W. Davis
preprint en

Abstract

This record contains the preprint "Law B: where the Dumont–Foata ratios stop decreasing near a face of the cube". Preprint, 18 November 2026, extended 19 November 2026. Not yet independently refereed. The programs, registrations, logs and check scripts are in the companion code record doi:10.5281/zenodo.24410457 STATEMENT Let m_k be the six-variable Dumont–Foata polynomials, written as weighted Motzkin paths, and R_k = m_k / m_(k−1)². The published article proves that R_k decreases in every variable on the cube [1/2, 1]^6 for every k ≥ 3. Below 1/2 that can fail. Law B describes where it fails at a face point, a point where one of the three combined parameters X, Y, Z vanishes linearly in t while the other two and the bottom weight δ_0 stay positive. A direction fails near t = 0 unless it moves the bottom weight while leaving the link fixed to first order; every other direction fails below t*_k ≈ δ_0^(k−1) / (λ_1′(0) · u′_(k−3)), where λ_1′(0) is the slope of the link λ_1 = XYZ at t = 0 and u′_j are the moments of the upper chain, the paths that stay at level 1 or above. Law B is OPEN: it is a registered pass and a measured law, not a theorem. MOTIVATION Near the bottom corner of the cube the limit of monotonicity shrinks very fast with k: about 0.2 at k = 3 and about 0.00001 at k = 8. Law B asks the same question on a face of the cube and answers it with one formula. The idea fits in one sentence. The moment m_k counts weighted paths that either stay at the bottom level or climb into the levels above; when the link between the bottom and the rest is weak, two kinds of path compete, those that stay at the bottom (weight δ_0 per step) and those that make one trip up (weight about λ_1 times the upper levels' own count u′). Monotonicity fails exactly while staying put still wins. The paper is written so that a reader with first-year calculus can follow every step with a pencil. It uses four tools only: counting weighted paths, the quotient rule, a geometric series, and the Gamma integral. Each result is given in the same order: what it says in plain words, the steps by hand at one real point, a case where the method must fail and does, a short "check it yourself" with its answer, and what to take from it. STATUS The paper's first table gives every claim one status word. In summary: Law B at a general face point: OPEN. Law B on 216 fresh point–direction pairs at k = 14 and 18 (test T4): REGISTERED PASS. Law B at k = 13 and 15 (test T12): REGISTERED PASS, and REPLICATED by a second implementation with byte-identical logs. The one-trip expansion of the moments, for any bottom weight (path counting): PROVED. Removing level 0 at a face point leaves the same family at shifted parameters; the corner case (t, t, 1, t, t, 1): PROVED in the companion thresholds paper. The two-term sign function and its root, and Law B as its leading term: DERIVED. The upper chain's moments, closed forms at the three T12 upper chains (Genocchi and Euler numbers), the enclosure Lemma E′, and two-sided bounds on ρ′_j = u′_j / u′_(j−1) for every j: PROVED, given the cited theorem of Gessel and Zeng. The two-term root over Law B lies in an explicit interval for every k ≥ 10 at the four T12 faces: PROVED, given the cited theorem; both ends behave like 1 − π²/k² − c/k³ + …: DERIVED. Which direction sets the threshold at the four T12 faces (the x-type directions, k = 7 to 15): MEASURED. Condition (iv), that no corner pattern fails above the corner's own threshold: PROVED by exact certificate at each k from 3 to 15 and at k = 17; REGISTERED PASS at k = 16, 18, 19 and 20 in a test frozen before the run, with k = 15 as known-answer control. Not yet checked by an independent mathematician. RESULTS Sections 1 to 6 give the law, its derivation by hand, and the evidence. Sections 7 to 11 bound the pieces of the law for every k at the four faces that were tested, find which direction actually sets the threshold, and explain why the bottom corner of the cube wins. Section 12 lists what a proof still needs: a bound on the paths with two or more trips, every face point rather than three, the class of directions that sets the threshold, and the split rule for every k. Section 13 records how each result was found, including two registered tests that failed (T2 and T3) and what each failure taught. The evidence. T4 (14 November 2026): all 144 predicted-failing pairs fail, with exact root over Law B between 0.9736 and 1.0498 at k = 14 and between 0.9846 and 1.0361 at k = 18, inside the registered windows [0.85, 1.15] and [0.90, 1.10]; none of the 72 predicted non-failing pairs fails. T12 (17 November 2026): the x-type directions give 0.9949 to 1.0286 at k = 13 and 1.0009 to 1.0262 at k = 15. APPENDIX: THE MATHEMATICS Setup. A Motzkin path of length k starts and ends at height 0. An up step has weight 1, a level step at height h has weight b_h = 2h² + (X+Y+Z−1)h + δ, and a down step from height h has weight λ_h = h(h+X−1)(h+Y−1)(h+Z−1). The moment m_k is the total weight. Along a direction with derivative D, the quotient rule gives D(m_k / m_(k−1)²) = G_k / m_(k−1)³ with G_k = m_(k−1) Dm_k − 2 m_k Dm_(k−1), so R_k decreases exactly where G_k ≤ 0. One point, by hand. Take P = (x, x̄, y, ȳ, z, z̄) = (1, 1, 1, t, t, 1). Then X = 2, Y = 2, Z = 2t and δ = 1 + 2t, so δ_0 = 1 and λ_1 = XYZ = 8t. The upper chain is the same family at shifted parameters, with b′_0 = 6, b′_1 = 15 and λ′_1 = 18, so its moments are u′_0 = 1, u′_1 = 6, u′_2 = 6² + 18 = 54 and u′_3 = 6³ + 2·6·18 + 18·15 = 216 + 216 + 270 = 702. Step 1, count the paths by trips (PROVED). A path with no trip has weight b_0^n. A path with one trip spends a steps at the bottom, goes up, spends j steps at level 1 or above (total weight u′_j), comes down (weight λ_1), and spends c more steps at the bottom, with a + c = n − 2 − j. So m_n = b_0^n + λ_1 Ŝ_n + O(λ_1²), where Ŝ_n = Σ_(j=0..n−2) (n−1−j) b_0^(n−2−j) u′_j. At P: Ŝ_4 = 3·1 + 2·6 + 1·54 = 69 and Ŝ_5 = 4·1 + 3·6 + 2·54 + 1·702 = 832. Steps 2 and 3, the sign function and its root (DERIVED). Along z the main effect is on the link, Dm_n ≈ Dλ_1 · Ŝ_n. Substituting, G_k / Dλ_1 = b_0^(k−1)(Ŝ_k − 2b_0Ŝ_(k−1)) − λ_1 Ŝ_k Ŝ_(k−1). At P, k = 5, this is 694 − 57408 λ_1, with root λ_1 = 694/57408 = 347/28704, that is t = 347/229632 ≈ 0.0015111. The exact first root of the full G_5 is 0.0014852, so the two-term root is 1.7 per cent high. Step 4, the leading term (DERIVED). The last term u′_(k−3) dominates Ŝ_(k−1), which gives Law B with a correction bracket: t*_k ≈ δ_0^(k−1)/(λ_1′(0) u′_(k−3)) · (1 − 2δ_0/ρ′_(k−3) − 2δ_0/ρ′_(k−2)). At P, k = 5: Law B alone gives 1/(8·54) = 1/432 ≈ 0.0023148, and the bracket is 1 − 2/9 − 2/13 ≈ 0.624, matching the measured ratio 0.642 of the exact root to Law B. As k grows the bracket tends to 1 (0.735 at k = 6, 0.804 at k = 7). Two cases that must fail. Drop the factor δ_0^(k−1): at Q = (1, 1, 2/3, t, t, 2/3), k = 7, the ratio becomes 0.006988, nowhere near 1. This is how the first form of the law failed its registered test. Shift the index to u′_(k−2): at P, k = 7, the ratio becomes 19.2. The index k − 3 is forced, because the longest trip inside m_(k−1) has k − 3 inner steps. Every number above is checked in exact fractions by lawb_derivation_check.py (22 checks, all pass; 12 of 12 deliberate errors caught). Appendices A to D work in the same way: the third-order ratio law (A), a Bernstein certificate of sign (B), pinning the corner's threshold with exact arithmetic (C), and why four magnitude errors all lean the same way (D), each with its own check script. FILES The preprint (PDF, 37 pages). The code record, doi:10.5281/zenodo.24410457, holds lawb_code_record_v4.zip (SHA-256 beginning 4797ce9b8c08a945; SHA256SUMS.txt lists 127 files) under the MIT licence: the registered test T12 (its frozen registration, amendment and freeze record, the program lawb.py, its outputs and their reruns, the result file and the judging script), the check scripts for every section and appendix with their logs and mutation logs, and the registered condition (iv) test. All check scripts use the Python standard library with exact fractions, and each has its own engine, written independently of the code behind the registered tests. AI ASSISTANCE The work was directed by the author. Most of the computational work was done with Claude (Anthropic) and Gemini (Google), with support from Grok (xAI). The text was drafted with AI assistance from the project's records and edited by the author, who takes responsibility for its contents. REFERENCES [1] J. W. Davis, A monotonicity property of the six-variable Dumont–Foata polynomials, version 6, doi:10.5281/zenodo.24158125. [2] J. W. Davis, Monotonicity thresholds for the six-variable Dumont–Foata ratios (the thresholds paper, a companion preprint). [3] I. M. Gessel, J. Zeng, Moments of orthogonal polynomials and exponential generating functions, Ramanuja n J. 61 (2023), no. 2, 675–700, doi:10.1007/s11139-022-00548-6.

Zenodo (CERN European Organization for Nuclear Research)
Advanced Combinatorial Mathematics
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon