From root convergence to normalized log convergence
ProvedErdos77.rpow_tendsto_implies_log_div_tendstoLet be a real sequence that is eventually at least , and suppose the normalized roots converge to a positive real limit :
Then the normalized logarithms converge to a real limit (namely ).
This is the standard bridge between exponential-growth-rate limits and their logarithmic form: since is continuous at the positive limit , convergence of gives convergence of , and for large by eventual positivity of the terms. It isolates the pure-analysis content of passing between the root limit in Erd\H{o}s Problem 77 and the logarithmic growth-rate limit.
Formalization Note Lean's real power with exponent at and the division by at are both defined (junk) values; the conclusion only concerns the limit at infinity, so these single terms are irrelevant.
import Mathlib open Filter Topology
namespace Erdos77
theorem rpow_tendsto_implies_log_div_tendsto (a : Nat → Real) (L : Real)
(ha : ∀ᶠ k : Nat in Filter.atTop, 1 <= a k)
(hL : Filter.Tendsto (fun k : Nat => (a k) ^ ((1 : Real) / (k : Real))) Filter.atTop (nhds L))
(hLpos : 0 < L) :
Exists fun l : Real => Filter.Tendsto (fun k : Nat => Real.log (a k) / (k : Real)) Filter.atTop (nhds l) := by sorry
end Erdos77