**The spectrum of the Berggren transfer operator on locally constant functions is
ProvedBerggrenHarmonic.eigenvalue_eq_zero_or_oneaether-catalogbridges
The spectrum of the Berggren transfer operator on locally constant functions is
{0, 1}. A nonzero locally constant eigenfunction forces the eigenvalue to be 0 or
1; equivalently the Markov operator has spectral gap 1, for every choice of the
weights (p₁, p₂, p₃).
theorem BerggrenHarmonic.eigenvalue_eq_zero_or_one(P : ProbVec) {n : ℕ} {f : Bdry → ℝ} (hf : DependsOn n f)
{c : ℝ} (hc : transfer P f = fun x => c * f x) (hne : ∃ x, f x ≠ 0) :
c = 0 ∨ c = 1 := by sorry
Formalization Note Transplanted verbatim from the Aether Catalog source Bridges/BerggrenTransferSpectrum.lean; the statement is byte-identical to the source declaration, elaborated with autoImplicit disabled in the platform environment.
Preamble
-- Thm stub generated from Bridges/BerggrenTransferSpectrum.lean import Mathlib import Definitions.Def_Bridges_BerggrenHarmonicMeasure import Definitions.Def_Bridges_BerggrenTransferSpectrum /-! # The transfer operator of the Berggren walk and its spectrum The Markov operator of the Berggren random walk acts on functions on the boundary `Bdry = ℕ → Fin 3` by `(L f)(x) = p₁ f(1·x) + p₂ f(2·x) + p₃ f(3·x)`, where `a·x = cons a x` prepends the Berggren move `a`. This is the operator whose fixed points are the harmonic functions of the walk and whose invariant measure is the harmonic measure `bernoulli P` of `Catalog.Bridges.BerggrenHarmonicMeasure`. The conjecture motivating this cycle was that the *spectral gap* of the Berggren walk is governed by the silver ratio `1 + √2` of the tree's metric growth. The results below show that this is false in a strong and clean way: on the natural core of locally constant functions (the union of the finite-dimensional spaces `DependsOn n`, which is dense in every reasonable completion), the transfer operator is **nilpotent modulo constants**. ## Main results * `transfer_dependsOn` : `L` maps functions of the first `n+1` letters to functions of the first `n` letters — the operator strictly decreases the "memory" of a function. * `iterate_transfer_const` : hence `Lⁿ f` is *constant* for every `f` depending on the first `n` letters. This is the nilpotency statement. * `eigenvalue_eq_zero_or_one` : consequently every eigenvalue of `L` on locally constant functions is `0` or `1`, and (`eigenfunction_one_isConst`) the eigenvalue `1` has only the constants as eigenfunctions. The **spectral gap is `1`, independently of `(p₁,p₂,p₃)`**. * `log_silver_not_eigenvalue` : in particular `log(1+√2)` — the growth exponent of the Berggren tree — is *not* an eigenvalue of the Markov operator: the conjectured silver-ratio spectral gap is refuted. (`log_silver_pos_lt_one` isolates the numerical fact `0 < log(1+√2) < 1` on which the refutation rests.) -/ open BerggrenHarmonic open Finset /-! ## The spectrum on locally constant functions -/
Formal statement
theorem BerggrenHarmonic.eigenvalue_eq_zero_or_one(P : ProbVec) {n : ℕ} {f : Bdry → ℝ} (hf : DependsOn n f)
{c : ℝ} (hc : transfer P f = fun x => c * f x) (hne : ∃ x, f x ≠ 0) :
c = 0 ∨ c = 1 := by sorrySource