Work

Scaled-Interval Hinge Telescoping, Difference-of-Convex Obstructions, and Smooth Proper-Divisor Matching for Erdős Problem 708

Year: 2026
Type: preprint
Abstract: Preprint manuscript, DualKernel-verified Lean 4 formalization (51 declarations, 0 unproven axioms), and standalone zero-dependency Rust verification suite for Erdős Problem #708 (Erdős–Surányi, 1959). Erdős Problem 708 asks whether every n-element set A ⊆ ℤ_{≥2} with maximum M = max(A) admits a subset B ⊆ I_x = {x + 1, ..., x + M} of cardinality |B| ≤ (2 + o(1))n satisfying ∏_{a ∈ A} a | ∏_{b ∈ B} b across every shifted interval I_x (x ≥ 0). Companion work reduces intervals below Q_n^2 to the shifted-interval hinge inequality (H_c) on Q_n-smooth prime weights w ∈ [0, 1]^{p ≤ Q_n}. This paper develops the scaled-interval integral decomposition of (H_c), establishes four unconditional structural theorems for the Q_n-smooth hinge functional, and isolates three explicit analytical obligations separating current bounds from g(n) ≤ (2 + o(1))n:- **Scaled-interval s-integral telescoping and negative-threshold dominance (Proposition 1 & Theorem 2):** Telescoping over scaled intervals I_{j,r} = {⌊x / p_j^r⌋ + 1, ..., ⌊(x + M) / p_j^r⌋} proves deterministic dominance across all negative thresholds s < 0, eliminates prime powers whenever (r − 1)w_{p_j} ≥ 1, and proves (H_c) pointwise across all thresholds c ≥ 1, M ≥ 2, and x ≥ 0 whenever at most two primes carry positive weight.- **Five-prime difference-of-convex interior counterexample (Proposition 3):** Because both sides of (H_c) are convex in w, their difference can peak strictly inside the unit cube. An explicit Chinese Remainder witness on P = {29, 31, 37, 41, 43} at M = 65,231 and c = 1.06 satisfies max_{w ∈ {0,1}^5} F_{1.06}(w) = 0 across all 32 binary vertices while failing at the interior midpoint F_{1.06}(1/2, ..., 1/2) = +0.90 > 0, ruling out naive binary extreme-point reduction.- **Discrete intermediate-value divisor peeling (Theorem 4):** Bounds high prime powers p_j^r > M / L_0 prime by prime and maps excess prime powers into exact smooth divisor bands inside (s, s + 1 + ε].- **Q_n-smooth proper-divisor matching and binary smooth sieve bounds (Theorem 5):** Proves that a proper-divisor injection π : M_{≥3}(Q_n) ↪ I_x implies (H_2) simultaneously across all continuous weight vectors w ∈ [0, 1]^{p ≤ Q_n}, and establishes (H_{2+ε}) for binary weights concentrated on small primes p ≤ L^{1/2 − θ}. ### Companion Manuscripts & Archives- **Paper I (Exact Small-n Values & Finite Certificate):** https://doi.org/10.5281/zenodo.24366897- **Paper II (Whole-Cofactor Peeling, Smooth LP Rounding & 8.83n Bound):** https://doi.org/10.5281/zenodo.24342717 ### Archive Contents & SHA-256 Digests- `paper/erdos708-smooth-hinge.pdf` (manuscript PDF, 7 pages): `e0d8648a65256f5888af95226a2062750f4e863fb7c3127b9ecea975bc1c14e7`- `paper/erdos708-smooth-hinge.typ` / `paper/erdos708-smooth-hinge.tex` / `paper/references.bib`: Typst, LaTeX, and BibTeX sources- `lean_formalization/`: Five Lean 4 modules (`Prop8.lean`, `HingeInequality.lean`, `Theorem3.lean`, `Corollary4.lean`, `Theorem6.lean`; 51 declarations, 0 `sorry`)- `verify_s_integral_corrected.rs` (scaled-interval s-integral verifier, 3,174 bytes): `05ec9b8d6e30a0d987d3319387955a2d19e5b88a7bd796445944894fda6a2ad3`- `explore_h2_smooth.rs` (smooth Hopcroft–Karp matching & sieve explorer, 27,202 bytes): `48add0a21fd8869f5e66f45114fd48e219bc0f7ebf1b7b527a30362bc3c86fc0`- `verify_three_final_checks.rs` (analytical checks & 5-prime DC counterexample verifier, 17,038 bytes): `84b59b9b7093c0ef1982088eaee9dc5a79140d54a5785586da5bd2c3e2a10e92`
Language: English
Open Access status: green
OpenAlex ID: W7236672485
Cites: 2
Cited by:
Related to: 10