A persistent divisor cover fences every later state
ProvedErdosProblems.Erdos243.CoefficientDivisorFence.stays_below_fenceerdos-243formalizationreciprocal-tail
For a positive integer sequence U with pointwise factors ρ_n≥1, cap ρ_n U_(n+1)≤U_n+B, and exact step ρ_n U_(n+1)=a_n U_n−b_n L_n, suppose U_0<z+B, B≥0, z>B, and every integer in [z,z+B) has a divisor greater than B common to L_n at each n. Then U_n<z+B for every n.
Preamble
import Mathlib.Tactic namespace ErdosProblems.Erdos243.CoefficientDivisorFence end ErdosProblems.Erdos243.CoefficientDivisorFence /-! # Coefficient-independent divisor fences (r5) Finite transition algebra only. This module does not claim the global CRT supply theorem, the analytic positive-numerator transfer, or the Erdős #243 endpoint. The covering hypothesis is explicit: producing such a cover is a separate CRT argument, and producing the required old divisors is a separate record argument. Not imported into `ErdosProblems/Root.lean`. Erdős #243 remains open. -/ open ErdosProblems.Erdos243.CoefficientDivisorFence
Formal statement
theorem ErdosProblems.Erdos243.CoefficientDivisorFence.stays_below_fence
(a b L U rho : ℕ → ℤ) (B z : ℤ)
(hU : ∀ n, 0 < U n)
(hrho : ∀ n, 1 ≤ rho n)
(hB : 0 ≤ B) (hz : B < z)
(hcap : ∀ n, rho n * U (n + 1) ≤ U n + B)
(hstep : ∀ n, rho n * U (n + 1) = a n * U n - b n * L n)
(hstart : U 0 < z + B)
(hcover : ∀ n, ∀ x : ℤ, z ≤ x → x < z + B →
∃ m : ℤ, B < m ∧ m ∣ x ∧ m ∣ L n) :
∀ n, U n < z + B := by sorry
Source
Pinned original Lean theorem and proof: https://github.com/wcook04/plectis-erdos-lean/blob/fbf41fb8b571d217b1648070dcc39d5eda5394ad/ErdosProblems/Erdos243/CoefficientDivisorFence.lean#L59-L77
Paper context and prior-art bibliography: https://github.com/wcook04/plectis-erdos/blob/551bae6dc6e732cf85172d66323c8d2bc77ba962/paper/243/erdos-243-reciprocal-tail-rigidity.tex#L1151-L1235
AI-assisted formalization and editorial review; no human review attested.