Week 3 established \(G = U \circ F \circ K \circ C\) as a well-defined composite map. This week asks: what changes on a second pass, \(G^2 = G \circ G\)? Near a fixed point \(x^*\) of \(G\) (Week 4’s Unfolding operator drives trajectories toward exactly such a point, by construction — UnfoldOp.decreases_Phi), the linearisation is governed by the same Lyapunov exponent from the correspondence table: \(\mu_{{\max}} = -2\). Two applications of a linear contraction with rate \(\mu_{{\max}}\) contract at rate \(2\mu_{{\max}}\) in the exponent — deviations shrink twice as fast under \(G^2\) as under \(G\), which is the formal content of “iterating gets you there faster,” stated precisely rather than as a slogan.
Interestingly, \(0.382 = 1/\varphi^2\) exactly (a standard golden-ratio identity: \(1/\varphi^2 = 2-\varphi\)), so the gap \(\tau-\varphi\) and \(\varphi\) itself are related by the same self-referential structure explored in Week 11 — another instance of the fixed-point pattern, not a new independent coincidence.
AXLE_v6.lean — theorem names and Lean identifiers above point at the actual source files.-- dm³ 101 · Week 13 · G² linearisation, μ_max from AXLE_v6.lean
-- Single pass: deviation ξ(t) ~ ξ₀ · e^{μ_max · t}, μ_max = -2
-- Two passes (G²): ξ(t) ~ ξ₀ · e^{2·μ_max · t} -- twice the rate
-- τ - φ = 2 - 1.618... ≈ 0.382 = 1/φ² (standard identity 2-φ=1/φ²)
-- Contrast: 3 - φ ≈ 1.382 (distance to the FOLD threshold c*=3,
-- not the LADDER limit τ=2 — two distinct thresholds, don't merge them)
example : (2:ℝ) - 3 ≠ 2 - ((1+Real.sqrt 5)/2) := by norm_num