Monotonicity thresholds for the six-variable Dumont–Foata ratios

This record contains the preprint "Monotonicity thresholds for the six-variable Dumont–Foata ratios". Preprint, 21 November 2026. Not yet independently refereed. The scripts, registrations, logs and proofs are in the companion code record doi:10.5281/zenodo.24423811. STATEMENT The ratios R_k = m_k / m_(k−1)² of the six-variable Dumont–Foata polynomials decrease in every variable on the cube [1/2, 1]^6 (the published article [D]). This paper asks how far below 1/2 that survives. For k = 3 to 9 the limit is exactly the corner threshold t*_k: R_k decreases strictly on [a, 1]^6 for every a > t*_k and fails at a = t*_k (for k = 3, t*_3 = 0.2124670677...). The corner thresholds fall faster than any geometric rate: 0.2125, 0.05595, 0.01414, 0.001941, 0.0001613, 0.000009334 for k = 3 to 8. From k = 4 to 12 the binding corner is (t,t,1,t,t,1) in direction x. At t = 0 that corner splits the continued fraction into one level and an upper chain, and the upper chain is the same family at shifted parameters. For the first failure point τ_k on that line, 2τ_k u_(k−3) → 1 is proved with an explicit k_0 (Theorem A), which gives τ_k ~ π^(2k−2) / (8 (2k−2)!), and the failure point is proved to be the only one for every k. A law for the threshold at general face points (Law B) passed two registered tests and remains open. MOTIVATION For k = 3 an exact certificate and an explicit failure point pin the limit to a narrow interval, and earlier searches put t*_4 near 0.056 and t*_5 near 0.014, which suggested that the natural cube is fixed by k = 3 alone. The question is how the limit depends on k, and why. The moments m_k are weighted Motzkin paths with level weights b_n = 2n² + (X+Y+Z−1)n + δ and down weights λ_n = n(n+X−1)(n+Y−1)(n+Z−1). The threshold turns out to be a balance between two path weights: staying at the bottom level, and making one trip into the upper chain. That balance explains the super-geometric decay: the upper chain's moments u_j are a binomial transform of the Genocchi numbers. STATUS R_3 decreases on [17/80, 1]^6: PROVED (exact Bernstein certificate in the published article). The whole-cube limit equals the corner threshold, T_k = t*_k, for k = 3 to 9: PROVED (computer-assisted, exact; k = 9 on the author's desktop). Corner thresholds t*_3 to t*_8: REPLICATED (two or three independent implementations). Corner thresholds t*_9 to t*_12: MEASURED (one implementation). The binding corner stays (t,t,1,t,t,1) for k = 9 to 12 (T1): REGISTERED PASS. Geometric decay toward ratio 4 (H4): REGISTERED FAIL (ratios 7.28, 12.03, 17.28). Growth like the upper-chain moments, the alternative stated in advance: REGISTERED PASS. The upper chain is the same family at (2, X+1, Y+1, X+Y+1+δ), and u_j = Σ C(j,i) G_(2i+4): PROVED. Exact formulas for g_0 and g_1, the first two coefficients of the sign function at the binding corner: PROVED (path counting). Theorem A: 2τ_k u_(k−3) → 1 for the first failure point τ_k in direction x, with explicit k_0(ε): PROVED. τ_k ~ π^(2k−2) / (8 (2k−2)!): PROVED. The failure point along x is unique for k = 3 to 400: PROVED (exact polynomials; T5, T6, T7 and T11 registered and passed). The failure point along x is unique for every k: PROVED. At that corner y and z̄ never fail, and ȳ and z fail once and below x, so the corner's own threshold is τ_k: REPLICATED for k = 4 to 60; REGISTERED PASS for k = 61 to 160 (T9, T10); PROVED for k > 160. Law B on 216 fresh face pairs (T4): REGISTERED PASS. Law B at k = 13 and 15 (T12): REGISTERED PASS. Law B at general face points: OPEN. t*_k · u_(k−3) → 1/2 for the corner threshold itself (Conjecture A): OPEN (it follows once no other corner pattern fails later). r_j is nondecreasing for every j, and r_j = ln j + ln(2/π) + γ − 1 + o(1): PROVED. Thresholds away from the faces for k ≥ 10: OPEN. Not yet checked by an independent mathematician. RESULTS Section 2. The continued fraction with four parameters X, Y, Z, δ built from the six variables, the sign function G_k = m_(k−1) Dm_k − 2 m_k Dm_(k−1), whose sign decides monotonicity along a direction, and the corner threshold over all 63 corner patterns and 6 directions. Section 3. The exact corner thresholds for k = 3 to 12, the binding corner and direction at each k, and the four predictions registered on 5 November 2026 before k = 6, 7 and 8 were computed: two failed as stated and are reported as failures. Section 4. The mechanism. At (t,t,1,t,t,1) the parameter Z vanishes at t = 0, removing level 0 leaves the same family at shifted parameters, and paths with one and two trips above the bottom level give g_0 and g_1 exactly. Theorem A and its explicit k_0; Lemma E and the second-order term of u_j / G_(2j+4); the three direction classes at the corner and their closed forms; the localization lemma for every k; why the decay is faster than geometric; the general law at face points, t*_k ≈ δ_0^(k−1) / (λ_1'(0) u'_(k−3)) (Law B), with its registered tests; and the continuous dual Hahn representation of Gessel and Zeng behind the measure used in the proofs. Section 5. The whole-cube closure for k = 3 to 9 (Bernstein certificates just above t*_k, Markov's inequality on every non-vertex coefficient, exact sign counts on every vertex coefficient, controls just below t*_k that fail as they must), and the hypotheses still open, each with what would decide it. Section 6. Methods, every file with its fingerprint, and references. APPENDIX: THE MATHEMATICS Appendix A works the mathematics with pencil in 34 blocks, and every method has a case where it must fail. The Genocchi numbers from the paths. With every variable 1, X = Y = Z = 2 and δ = 3, so b_0, b_1, b_2 = 3, 10, 21 and λ_1 = 8, λ_2 = 54. Listing the paths gives m_2 = 9 + 8 = 17, m_3 = 27 + 48 + 80 = 155 and m_4 = 81 + 216 + 480 + 800 + 64 + 432 = 2073. With every down weight set to 1 instead, the counts are 10 and 43: the weights λ_h carry the variables into the counts. A Bernstein certificate in one variable. 1 − 3t + 3t² has Bernstein coefficients 1, −1/2, 1 on [0, 1], so the test is undecided; on [0, 1/2] they are 1, 1/4, 1/4, all positive, and by symmetry the polynomial is positive on [0, 1]. (1 − 2t)² never passes, however often the interval is split, because it is 0 at t = 1/2. Descartes' rule. G_3/t at the corner along x is 9 − 36t − 118t² − … − 16t⁸, one sign change, so exactly one positive root, t = 0.151902. Two sign changes decide nothing: t² − 3t + 2 has two positive roots and t² − 2t + 2 has none. Richardson extrapolation. For (1 + 1/j)^j, 2a_20 − a_10 = 2.7129, an error of 0.0054 against 0.0650 for a_20; for 1 + 1/√j the same step gains nothing in order. Other blocks work the upper chain, Lemma E at j = 10, one k_0 evaluation at k = 40, the first failure point at k = 6, the closed forms at small k, the closure at k = 3 with Markov's inequality, Theorem A for class C at k = 6, Law B at one face point and its formula derived by hand, the spectral form, the monotonicity of r_j, the shift lemma's second-order term and the fact that δ is a translation that the Hankel determinants do not see. Every number is checked in exact arithmetic by scripts in the code record (for example mustfail_check.py, 29 checks; thresholds_A30_check.py, 114 checks; thresholds_hand_blocks_check_7.py, 23 checks), with mutation tests that catch every deliberate error tried (for example 11 of 11 and 8 of 8). Appendix B records how the results were found. FILES The preprint (PDF, 76 pages). The code record, doi:10.5281/zenodo.24423811, holds thresholds_code_record_v30.zip (SHA-256 beginning e4a0aa64e4425d7a; SHA256SUMS.txt lists 309 files) under the MIT licence: every script as run with its log, every registration written before its test ran (T1 to T7, T9 to T12 and the threshold ladder), and the proofs in full. 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 [D] J. W. Davis, A monotonicity property of the six-variable Dumont–Foata polynomials, Zenodo, version 6, doi:10.5281/zenodo.24158125; certificates for k = 21 to 51 at doi:10.5281/zenodo.23891158. [DF] D. Dumont, D. Foata, Une propriété de symétrie des nombres de Genocchi, Bull. Soc. Math. France 104 (1976), 433–451. [R] A. Randrianarivony, Polynômes de Dumont–Foata généralisés, Sém. Lothar. Combin. 32 (1994), B32d. [Z] J. Zeng, Sur quelques propriétés de symétrie des nombres de Genocchi, Discrete Math. 153 (1996), 319–333. [JV] M. Josuat-Vergès, Generalized Dumont–Foata polynomials and alternative tableaux, Sém. Lothar. Combin. 64 (2010), B64b; arXiv:1005.4007. [GZ] I. M. Gessel, J. Zeng, Moments of orthogonal polynomials and exponential generating functions, Ramanujan J.; arXiv:2107.00255. [BL] H. J. Brascamp, E. H. Lieb, J. Funct. Anal. 22 (1976), 366–389. [KLS] R. Koekoek, P. A. Lesky, R. F. Swarttouw, Hypergeometric Orthogonal Polynomials and Their q-Analogues, Springer, 2010. [DLMF] NIST Digital Library of Mathematical Functions. [M] D. S. Mitrinović, Analytic Inequalities, Springer, 1970. [HLP] G. H. Hardy, J. E. Littlewood, G. Pólya, Inequalities, Cambridge University Press.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-21
DOI
https://doi.org/10.5281/zenodo.24423913
Citations
2
Primary Topic
Advanced Combinatorial Mathematics
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Monotonicity thresholds for the six-variable Dumont–Foata ratios

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

Monotonicity thresholds for the six-variable Dumont–Foata ratios

Joshua W. Davis
preprint en
2 citations

Abstract

This record contains the preprint "Monotonicity thresholds for the six-variable Dumont–Foata ratios". Preprint, 21 November 2026. Not yet independently refereed. The scripts, registrations, logs and proofs are in the companion code record doi:10.5281/zenodo.24423811. STATEMENT The ratios R_k = m_k / m_(k−1)² of the six-variable Dumont–Foata polynomials decrease in every variable on the cube [1/2, 1]^6 (the published article [D]). This paper asks how far below 1/2 that survives. For k = 3 to 9 the limit is exactly the corner threshold t*_k: R_k decreases strictly on [a, 1]^6 for every a > t*_k and fails at a = t*_k (for k = 3, t*_3 = 0.2124670677...). The corner thresholds fall faster than any geometric rate: 0.2125, 0.05595, 0.01414, 0.001941, 0.0001613, 0.000009334 for k = 3 to 8. From k = 4 to 12 the binding corner is (t,t,1,t,t,1) in direction x. At t = 0 that corner splits the continued fraction into one level and an upper chain, and the upper chain is the same family at shifted parameters. For the first failure point τ_k on that line, 2τ_k u_(k−3) → 1 is proved with an explicit k_0 (Theorem A), which gives τ_k ~ π^(2k−2) / (8 (2k−2)!), and the failure point is proved to be the only one for every k. A law for the threshold at general face points (Law B) passed two registered tests and remains open. MOTIVATION For k = 3 an exact certificate and an explicit failure point pin the limit to a narrow interval, and earlier searches put t*_4 near 0.056 and t*_5 near 0.014, which suggested that the natural cube is fixed by k = 3 alone. The question is how the limit depends on k, and why. The moments m_k are weighted Motzkin paths with level weights b_n = 2n² + (X+Y+Z−1)n + δ and down weights λ_n = n(n+X−1)(n+Y−1)(n+Z−1). The threshold turns out to be a balance between two path weights: staying at the bottom level, and making one trip into the upper chain. That balance explains the super-geometric decay: the upper chain's moments u_j are a binomial transform of the Genocchi numbers. STATUS R_3 decreases on [17/80, 1]^6: PROVED (exact Bernstein certificate in the published article). The whole-cube limit equals the corner threshold, T_k = t*_k, for k = 3 to 9: PROVED (computer-assisted, exact; k = 9 on the author's desktop). Corner thresholds t*_3 to t*_8: REPLICATED (two or three independent implementations). Corner thresholds t*_9 to t*_12: MEASURED (one implementation). The binding corner stays (t,t,1,t,t,1) for k = 9 to 12 (T1): REGISTERED PASS. Geometric decay toward ratio 4 (H4): REGISTERED FAIL (ratios 7.28, 12.03, 17.28). Growth like the upper-chain moments, the alternative stated in advance: REGISTERED PASS. The upper chain is the same family at (2, X+1, Y+1, X+Y+1+δ), and u_j = Σ C(j,i) G_(2i+4): PROVED. Exact formulas for g_0 and g_1, the first two coefficients of the sign function at the binding corner: PROVED (path counting). Theorem A: 2τ_k u_(k−3) → 1 for the first failure point τ_k in direction x, with explicit k_0(ε): PROVED. τ_k ~ π^(2k−2) / (8 (2k−2)!): PROVED. The failure point along x is unique for k = 3 to 400: PROVED (exact polynomials; T5, T6, T7 and T11 registered and passed). The failure point along x is unique for every k: PROVED. At that corner y and z̄ never fail, and ȳ and z fail once and below x, so the corner's own threshold is τ_k: REPLICATED for k = 4 to 60; REGISTERED PASS for k = 61 to 160 (T9, T10); PROVED for k > 160. Law B on 216 fresh face pairs (T4): REGISTERED PASS. Law B at k = 13 and 15 (T12): REGISTERED PASS. Law B at general face points: OPEN. t*_k · u_(k−3) → 1/2 for the corner threshold itself (Conjecture A): OPEN (it follows once no other corner pattern fails later). r_j is nondecreasing for every j, and r_j = ln j + ln(2/π) + γ − 1 + o(1): PROVED. Thresholds away from the faces for k ≥ 10: OPEN. Not yet checked by an independent mathematician. RESULTS Section 2. The continued fraction with four parameters X, Y, Z, δ built from the six variables, the sign function G_k = m_(k−1) Dm_k − 2 m_k Dm_(k−1), whose sign decides monotonicity along a direction, and the corner threshold over all 63 corner patterns and 6 directions. Section 3. The exact corner thresholds for k = 3 to 12, the binding corner and direction at each k, and the four predictions registered on 5 November 2026 before k = 6, 7 and 8 were computed: two failed as stated and are reported as failures. Section 4. The mechanism. At (t,t,1,t,t,1) the parameter Z vanishes at t = 0, removing level 0 leaves the same family at shifted parameters, and paths with one and two trips above the bottom level give g_0 and g_1 exactly. Theorem A and its explicit k_0; Lemma E and the second-order term of u_j / G_(2j+4); the three direction classes at the corner and their closed forms; the localization lemma for every k; why the decay is faster than geometric; the general law at face points, t*_k ≈ δ_0^(k−1) / (λ_1'(0) u'_(k−3)) (Law B), with its registered tests; and the continuous dual Hahn representation of Gessel and Zeng behind the measure used in the proofs. Section 5. The whole-cube closure for k = 3 to 9 (Bernstein certificates just above t*_k, Markov's inequality on every non-vertex coefficient, exact sign counts on every vertex coefficient, controls just below t*_k that fail as they must), and the hypotheses still open, each with what would decide it. Section 6. Methods, every file with its fingerprint, and references. APPENDIX: THE MATHEMATICS Appendix A works the mathematics with pencil in 34 blocks, and every method has a case where it must fail. The Genocchi numbers from the paths. With every variable 1, X = Y = Z = 2 and δ = 3, so b_0, b_1, b_2 = 3, 10, 21 and λ_1 = 8, λ_2 = 54. Listing the paths gives m_2 = 9 + 8 = 17, m_3 = 27 + 48 + 80 = 155 and m_4 = 81 + 216 + 480 + 800 + 64 + 432 = 2073. With every down weight set to 1 instead, the counts are 10 and 43: the weights λ_h carry the variables into the counts. A Bernstein certificate in one variable. 1 − 3t + 3t² has Bernstein coefficients 1, −1/2, 1 on [0, 1], so the test is undecided; on [0, 1/2] they are 1, 1/4, 1/4, all positive, and by symmetry the polynomial is positive on [0, 1]. (1 − 2t)² never passes, however often the interval is split, because it is 0 at t = 1/2. Descartes' rule. G_3/t at the corner along x is 9 − 36t − 118t² − … − 16t⁸, one sign change, so exactly one positive root, t = 0.151902. Two sign changes decide nothing: t² − 3t + 2 has two positive roots and t² − 2t + 2 has none. Richardson extrapolation. For (1 + 1/j)^j, 2a_20 − a_10 = 2.7129, an error of 0.0054 against 0.0650 for a_20; for 1 + 1/√j the same step gains nothing in order. Other blocks work the upper chain, Lemma E at j = 10, one k_0 evaluation at k = 40, the first failure point at k = 6, the closed forms at small k, the closure at k = 3 with Markov's inequality, Theorem A for class C at k = 6, Law B at one face point and its formula derived by hand, the spectral form, the monotonicity of r_j, the shift lemma's second-order term and the fact that δ is a translation that the Hankel determinants do not see. Every number is checked in exact arithmetic by scripts in the code record (for example mustfail_check.py, 29 checks; thresholds_A30_check.py, 114 checks; thresholds_hand_blocks_check_7.py, 23 checks), with mutation tests that catch every deliberate error tried (for example 11 of 11 and 8 of 8). Appendix B records how the results were found. FILES The preprint (PDF, 76 pages). The code record, doi:10.5281/zenodo.24423811, holds thresholds_code_record_v30.zip (SHA-256 beginning e4a0aa64e4425d7a; SHA256SUMS.txt lists 309 files) under the MIT licence: every script as run with its log, every registration written before its test ran (T1 to T7, T9 to T12 and the threshold ladder), and the proofs in full. 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 [D] J. W. Davis, A monotonicity property of the six-variable Dumont–Foata polynomials, Zenodo, version 6, doi:10.5281/zenodo.24158125; certificates for k = 21 to 51 at doi:10.5281/zenodo.23891158. [DF] D. Dumont, D. Foata, Une propriété de symétrie des nombres de Genocchi, Bull. Soc. Math. France 104 (1976), 433–451. [R] A. Randrianarivony, Polynômes de Dumont–Foata généralisés, Sém. Lothar. Combin. 32 (1994), B32d. [Z] J. Zeng, Sur quelques propriétés de symétrie des nombres de Genocchi, Discrete Math. 153 (1996), 319–333. [JV] M. Josuat-Vergès, Generalized Dumont–Foata polynomials and alternative tableaux, Sém. Lothar. Combin. 64 (2010), B64b; arXiv:1005.4007. [GZ] I. M. Gessel, J. Zeng, Moments of orthogonal polynomials and exponential generating functions, Ramanujan J.; arXiv:2107.00255. [BL] H. J. Brascamp, E. H. Lieb, J. Funct. Anal. 22 (1976), 366–389. [KLS] R. Koekoek, P. A. Lesky, R. F. Swarttouw, Hypergeometric Orthogonal Polynomials and Their q-Analogues, Springer, 2010. [DLMF] NIST Digital Library of Mathematical Functions. [M] D. S. Mitrinović, Analytic Inequalities, Springer, 1970. [HLP] G. H. Hardy, J. E. Littlewood, G. Pólya, Inequalities, Cambridge University Press.

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