The separate spectral recurrence of Section 6

Lean contributors (project-level attribution): Zichen Wang.

Source header · Attribution source

The regular one-step formulas below are the exact inverse DFP update in the current eigenbasis. The two controls keep the same cycle scale while the intermediate eigenbasis changes. The signed two-step map and its linearization are those of Unknown identifier `cm`sorry - sorry - sorry - map : ?m.16cm-Unknown identifier `two`two-Unknown identifier `step`step-map, independently of the polar construction.

The source C⁷ graph is constructed from uniform C⁸ zero-start graph-transform iterates and Arzelà–Ascoli. Unknown identifier `centerProposition`centerProposition includes the literal initialization, infinite actual recurrence and inverse-cubic scale law for every sufficiently small positive initial scale. The Section 6.3 interfaces retain the same physical orbit, source frame first crossings, cancellation constants and limiting circle. No polar or earlier counterexample existence theorem is used here.

public sectionnoncomputable sectionopen Filteropen scoped Asymptotics EuclideanSpace Matrix Topologynamespace DFPWolfe.V21.CenterManifold

Helper for cm-center-proposition: the regular secant energy denominator in cm-regular-step.

def regularDenominator (r μ : ℝ) : ℝ := 1 + 2 * μ * r + r ^ 2

Helper for cm-center-proposition: the regular inverse-metric energy in cm-regular-step.

def regularEnergy (p r μ : ℝ) : ℝ := p * r ^ 2 * (r + μ) ^ 2 + (1 + μ * r) ^ 2

Helper for cm-center-proposition: the three source entries (sorry, sorry, sorry) : ?m.1 × ?m.3 × ?m.4(Unknown identifier `a₀`a₀,Unknown identifier `b₀`b₀,Unknown identifier `d₀`d₀) encode the next matrix as !![sorry, sorry * sorry; sorry * sorry, sorry] : Matrix (Fin 2) (Fin 2) ?m.32!![Unknown identifier `a₀`a₀, Unknown identifier `r`r*Unknown identifier `b₀`b₀; Unknown identifier `r`r*Unknown identifier `b₀`b₀, Unknown identifier `d₀`d₀].

def regularEntries (p h r μ : ℝ) : ℝ × ℝ × ℝ := let D := regularDenominator r μ let E := regularEnergy p r μ (h * p * r ^ 2 * (1 + μ * r) ^ 2 / E + r ^ 2 / D, 1 / D - h * p * r * (r + μ) * (1 + μ * r) / E, h * p * r ^ 2 * (r + μ) ^ 2 / E + 1 / D)

Helper for cm-center-proposition: the exact regular updated inverse Hessian in the incoming eigenbasis, before rotating to its new eigenbasis.

def regularMetric (p h r μ : ℝ) : Matrix (Fin 2) (Fin 2) ℝ := let a := regularEntries p h r μ !![a.1, r * a.2.1; r * a.2.1, a.2.2]

Helper for cm-center-proposition: the low diagonal entry of the regular matrix.

theorem regularMetric_low (p h r μ : ℝ) : regularMetric p h r μ 0 0 = (regularEntries p h r μ).1 := by rfl

Helper for cm-center-proposition: the upper off-diagonal entry of the regular matrix.

theorem regularMetric_upper (p h r μ : ℝ) : regularMetric p h r μ 0 1 = r * (regularEntries p h r μ).2.1 := by rfl

Helper for cm-center-proposition: the lower off-diagonal entry of the regular matrix.

theorem regularMetric_lower (p h r μ : ℝ) : regularMetric p h r μ 1 0 = r * (regularEntries p h r μ).2.1 := by rfl

Helper for cm-center-proposition: the high diagonal entry of the regular matrix.

theorem regularMetric_high (p h r μ : ℝ) : regularMetric p h r μ 1 1 = (regularEntries p h r μ).2.2 := by rfl

Helper for cm-center-proposition: the normalized gradient coordinates (sorry, sorry) : ?m.1 × ?m.2(Unknown identifier `q₁`q₁,Unknown identifier `q₂`q₂) with actual incoming-frame vector (sorry, sorry * sorry) : ?m.1 × ?m.2(Unknown identifier `q₁`q₁,Unknown identifier `r`r*Unknown identifier `q₂`q₂).

def regularGradientResiduals (p r μ τ : ℝ) : ℝ × ℝ := let D := regularDenominator r μ (1 - τ * (p + 1) * r * (r + μ) / D, p - τ * (p + 1) * (1 + μ * r) / D)

Helper for cm-center-proposition: the actual normalized updated gradient in the incoming eigenbasis.

def regularGradient (p r μ τ : ℝ) : Fin 2 → ℝ := let q := regularGradientResiduals p r μ τ ![q.1, r * q.2]

Helper for cm-center-proposition: the source scalar denominator is strictly positive whenever failed to synthesize instance of type class AddGroup ℕ Hint: Type class instance resolution failures can be inspected with the `set_option trace.Meta.synthInstance true` command.|sorry| < 1 : Prop|Unknown identifier `μ`μ| < 1.

theorem regularDenominator_pos (r μ : ℝ) (hμ : |μ| < 1) : 0 < regularDenominator r μ := by have hμsq : μ ^ 2 < 1 := by nlinarith [abs_lt.mp hμ, sq_abs μ] unfold regularDenominator nlinarith [sq_nonneg (r + μ)]

Helper for cm-center-proposition: the source inverse-metric energy is strictly positive for positive shape and failed to synthesize instance of type class AddGroup ℕ Hint: Type class instance resolution failures can be inspected with the `set_option trace.Meta.synthInstance true` command.|sorry| < 1 : Prop|Unknown identifier `μ`μ| < 1.

theorem regularEnergy_pos (p r μ : ℝ) (hp : 0 < p) (hμ : |μ| < 1) : 0 < regularEnergy p r μ := by have hnonneg : 0 ≤ p * r ^ 2 * (r + μ) ^ 2 := by positivity by_cases hlinear : 1 + μ * r = 0 · have hr : r ≠ 0 := by intro hr simp [hr] at hlinear have hrμ : r + μ ≠ 0 := by intro hrμ have hμsq : μ ^ 2 < 1 := by nlinarith [abs_lt.mp hμ, sq_abs μ] nlinarith unfold regularEnergy exact add_pos_of_pos_of_nonneg (mul_pos (mul_pos hp (sq_pos_of_ne_zero hr)) (sq_pos_of_ne_zero hrμ)) (sq_nonneg _) · unfold regularEnergy exact add_pos_of_nonneg_of_pos hnonneg (sq_pos_of_ne_zero hlinear)

Helper for cm-center-proposition: the three raw energies in cm-single-update have the source's canceled factors in spectral coordinates.

theorem spectralEnergyIdentities (z : DFP.AbstractSecantStep (Fin 2)) (p h r μ G : ℝ) (hH : z.inverseHessian = Matrix.diagonal ![h * p * r ^ 2, h]) (hg : z.gradient = G • ![(1 : ℝ), p * r]) (hA : z.secantMatrix = !![1, μ; μ, 1]) : z.gradient ⬝ᵥ z.preconditionedGradient = h * p * G ^ 2 * r ^ 2 * (p + 1) ∧ z.preconditionedGradient ⬝ᵥ (z.secantMatrix *ᵥ z.preconditionedGradient) = h ^ 2 * p ^ 2 * G ^ 2 * r ^ 2 * regularDenominator r μ ∧ (z.secantMatrix *ᵥ z.preconditionedGradient) ⬝ᵥ (z.inverseHessian *ᵥ (z.secantMatrix *ᵥ z.preconditionedGradient)) = h ^ 3 * p ^ 2 * G ^ 2 * r ^ 2 * regularEnergy p r μ := by rw [DFP.AbstractSecantStep.preconditionedGradient_def, hH, hg, hA] simp only [regularDenominator, regularEnergy] simp [Matrix.mulVec, dotProduct, Fin.sum_univ_two] constructor · ring · constructor · ring · ring

Helper for cm-center-proposition: positivity of the actual step energies makes every canceled spectral factor nonzero.

theorem spectralParameters_ne_zero (z : DFP.AbstractSecantStep (Fin 2)) (p h r μ G : ℝ) (hH : z.inverseHessian = Matrix.diagonal ![h * p * r ^ 2, h]) (hg : z.gradient = G • ![(1 : ℝ), p * r]) (hA : z.secantMatrix = !![1, μ; μ, 1]) : h ≠ 0 ∧ p ≠ 0 ∧ G ≠ 0 ∧ r ≠ 0 ∧ regularDenominator r μ ≠ 0 ∧ regularEnergy p r μ ≠ 0 := by obtain ⟨_, hβ, hγ⟩ := spectralEnergyIdentities z p h r μ G hH hg hA have hβne := z.stepLengthDenominator_ne_zero have hγne := ne_of_gt z.secantImageEnergy_pos rw [hβ] at hβne rw [hγ] at hγne have hh : h ≠ 0 := by intro hh simp [hh] at hβne have hp : p ≠ 0 := by intro hp simp [hp] at hβne have hG : G ≠ 0 := by intro hG simp [hG] at hβne have hr : r ≠ 0 := by intro hr simp [hr] at hβne exact ⟨hh, hp, hG, hr, right_ne_zero_of_mul hβne, right_ne_zero_of_mul hγne⟩

Helper for cm-center-proposition: cm-single-step has the exact regular spectral displacement .

theorem spectralDisplacement (z : DFP.AbstractSecantStep (Fin 2)) (p h r μ G : ℝ) (hH : z.inverseHessian = Matrix.diagonal ![h * p * r ^ 2, h]) (hg : z.gradient = G • ![(1 : ℝ), p * r]) (hA : z.secantMatrix = !![1, μ; μ, 1]) : z.displacement = (-(G * z.tau * (p + 1) * r / regularDenominator r μ)) • ![r, (1 : ℝ)] := by obtain ⟨hδ, hβ, _⟩ := spectralEnergyIdentities z p h r μ G hH hg hA obtain ⟨hh, hp, hG, hr, hD, _⟩ := spectralParameters_ne_zero z p h r μ G hH hg hA rw [DFP.AbstractSecantStep.preconditionedGradient_def] at hδ hβ rw [z.displacement_def, z.stepLength_def, hδ, hβ, z.preconditionedGradient_def, hH, hg] ext i fin_cases i · simp [Matrix.mulVec, dotProduct, Fin.sum_univ_two] field_simp [hh, hp, hG, hr, hD] · simp [Matrix.mulVec, dotProduct, Fin.sum_univ_two] field_simp [hh, hp, hG, hr, hD]

Helper for cm-center-proposition: the exact next gradient is the source regular (sorry, sorry * sorry) : ?m.1 × ?m.2(Unknown identifier `q₁`q₁,Unknown identifier `r`r*Unknown identifier `q₂`q₂) vector, multiplied by its incoming amplitude.

theorem spectralGradientUpdate (z : DFP.AbstractSecantStep (Fin 2)) (p h r μ G : ℝ) (hH : z.inverseHessian = Matrix.diagonal ![h * p * r ^ 2, h]) (hg : z.gradient = G • ![(1 : ℝ), p * r]) (hA : z.secantMatrix = !![1, μ; μ, 1]) : z.nextGradient = G • regularGradient p r μ z.tau := by rw [z.nextGradient_def, z.gradientChange_def, spectralDisplacement z p h r μ G hH hg hA, hA, hg] unfold regularGradient regularGradientResiduals ext i fin_cases i · simp ring · simp ring

Helper for cm-center-proposition: the source entries of cm-regular-step are the actual inverse DFP matrix, without a remaining singular power of Unknown identifier `r`r.

theorem spectralMetricUpdate (z : DFP.AbstractSecantStep (Fin 2)) (p h r μ G : ℝ) (hH : z.inverseHessian = Matrix.diagonal ![h * p * r ^ 2, h]) (hg : z.gradient = G • ![(1 : ℝ), p * r]) (hA : z.secantMatrix = !![1, μ; μ, 1]) : z.nextInverseHessian = regularMetric p h r μ := by obtain ⟨hh, hp, hG, hr, hD, hE⟩ := spectralParameters_ne_zero z p h r μ G hH hg hA have hgCoord : z.gradient = ![G, G * (p * r)] := by rw [hg] ext i fin_cases i · simp · simp have hβ : (h * p * r ^ 2 * G) * (1 * (h * p * r ^ 2 * G) + μ * (h * (G * (p * r)))) + (h * (G * (p * r))) * (μ * (h * p * r ^ 2 * G) + 1 * (h * (G * (p * r)))) = h ^ 2 * p ^ 2 * G ^ 2 * r ^ 2 * regularDenominator r μ := by simp only [regularDenominator] ring have hγ : (h * p * r ^ 2) * (1 * (h * p * r ^ 2 * G) + μ * (h * (G * (p * r)))) ^ 2 + h * (μ * (h * p * r ^ 2 * G) + 1 * (h * (G * (p * r)))) ^ 2 = h ^ 3 * p ^ 2 * G ^ 2 * r ^ 2 * regularEnergy p r μ := by simp only [regularEnergy] ring rw [z.nextInverseHessian_eigenframe (h * p * r ^ 2) h G (G * (p * r)) 1 μ 1 hH hgCoord hA, hβ, hγ] apply Matrix.ext intro i j have hi : i = 0 ∨ i = 1 := by omega have hj : j = 0 ∨ j = 1 := by omega rcases hi with rfl | rfl · rcases hj with rfl | rfl · simp only [regularMetric_low, regularEntries, Matrix.of_apply, Matrix.cons_val_zero] field_simp [hh, hp, hG, hr, hD, hE] simp only [regularDenominator, regularEnergy] ring · simp only [regularMetric_upper, regularEntries, Matrix.of_apply, Matrix.cons_val_zero, Matrix.cons_val_one, Matrix.cons_val_fin_one] field_simp [hh, hp, hG, hr, hD, hE] ring · rcases hj with rfl | rfl · simp only [regularMetric_lower, regularEntries, Matrix.of_apply, Matrix.cons_val_zero, Matrix.cons_val_one, Matrix.cons_val_fin_one] field_simp [hh, hp, hG, hr, hD, hE] ring · simp only [regularMetric_high, regularEntries, Matrix.of_apply, Matrix.cons_val_one, Matrix.cons_val_fin_one] field_simp [hh, hp, hG, hr, hD, hE] simp only [regularDenominator, regularEnergy] ring

Helper for cm-center-proposition: the determinant of the regular updated matrix is exactly , as in cm-regular-step.

theorem regularMetric_det (p h r μ : ℝ) (hD : regularDenominator r μ ≠ 0) (hE : regularEnergy p r μ ≠ 0) : (regularMetric p h r μ).det = h * p * r ^ 2 * regularDenominator r μ / regularEnergy p r μ := by rw [Matrix.det_fin_two, regularMetric_low, regularMetric_high, regularMetric_upper, regularMetric_lower] dsimp only [regularEntries] field_simp [hD, hE] simp only [regularDenominator, regularEnergy] ring

Helper for cm-center-proposition: the nonsingular source step extends at Unknown identifier `r`sorry = 0 : Propr = 0 to the matrix .

theorem regularMetric_zero (p h μ : ℝ) : regularMetric p h 0 μ = Matrix.diagonal ![(0 : ℝ), 1] := by ext i j fin_cases i · fin_cases j · simp [regularMetric, regularEntries, regularDenominator, regularEnergy] · simp [regularMetric, regularEntries, regularDenominator, regularEnergy] · fin_cases j · simp [regularMetric, regularEntries, regularDenominator, regularEnergy] · simp [regularMetric, regularEntries, regularDenominator, regularEnergy]

Helper for cm-center-proposition: the source high eigenvalue and removable low factor in cm-regular-step.

def regularSpectrum (p h r μ : ℝ) : ℝ × ℝ := let a := regularEntries p h r μ let highValue := RealSymmetric2.high a.1 (r * a.2.1) a.2.2 (h * p * regularDenominator r μ / (regularEnergy p r μ * highValue), highValue)

Helper for cm-center-proposition: the source rotated numerators (sorry, sorry) : ?m.1 × ?m.2(Unknown identifier `X`X,Unknown identifier `Y`Y) and normalization Unknown identifier `N`N, with .

def regularRotatedData (p h r μ τ : ℝ) : (ℝ × ℝ) × ℝ := let a := regularEntries p h r μ let spectrum := regularSpectrum p h r μ let q := regularGradientResiduals p r μ τ let c := a.2.2 - r ^ 2 * spectrum.1 ((c * q.1 - r ^ 2 * a.2.1 * q.2, a.2.1 * q.1 + c * q.2), Real.sqrt (c ^ 2 + r ^ 2 * a.2.1 ^ 2))

Helper for cm-center-proposition: the exact regular map stores (sorry / sorry, sorry, sorry, sorry / sorry) : ?m.1 × ?m.9 × ?m.11 × ?m.12(Unknown identifier `rnew`rnew/Unknown identifier `r`r,Unknown identifier `pnew`pnew,Unknown identifier `hnew`hnew,Unknown identifier `Gnew`Gnew/Unknown identifier `G`G) from cm-regular-step.

def regularStep (p h r μ τ : ℝ) : ℝ × ℝ × ℝ × ℝ := let spectrum := regularSpectrum p h r μ let rotated := regularRotatedData p h r μ τ (spectrum.1 * rotated.1.1 / (spectrum.2 * rotated.1.2), spectrum.2 * rotated.1.2 ^ 2 / (spectrum.1 * rotated.1.1 ^ 2), spectrum.2, rotated.1.1 / rotated.2)

Helper for cm-center-proposition: the source entries at zero radius are (sorry, sorry, sorry) = (0, 1, 1) : Prop(Unknown identifier `a₀`a₀,Unknown identifier `b₀`b₀,Unknown identifier `d₀`d₀) = (0,1,1).

theorem regularEntries_zero (p h μ : ℝ) : regularEntries p h 0 μ = (0, 1, 1) := by simp [regularEntries, regularDenominator, regularEnergy]

Helper for cm-center-proposition: the removable spectrum at zero radius is .

theorem regularSpectrum_zero (p h μ : ℝ) : regularSpectrum p h 0 μ = (h * p, 1) := by simp [regularSpectrum, regularEntries_zero, regularDenominator, regularEnergy, RealSymmetric2.high_apply, RealSymmetric2.gap_apply]

Helper for cm-center-proposition: at zero radius and .

theorem regularRotatedData_zero (p h μ τ : ℝ) : regularRotatedData p h 0 μ τ = ((1, (1 - τ) * (p + 1)), 1) := by simp [regularRotatedData, regularEntries_zero, regularSpectrum_zero, regularGradientResiduals, regularDenominator] ring

Helper for cm-center-proposition: the regular zero-radius step retains its nontrivial radius multiplier and shape coordinate.

theorem regularStep_zero (p h μ τ : ℝ) : regularStep p h 0 μ τ = (h * p / ((1 - τ) * (p + 1)), ((1 - τ) * (p + 1)) ^ 2 / (h * p), 1, 1) := by simp [regularStep, regularSpectrum_zero, regularRotatedData_zero]

Helper for cm-center-proposition: the two exact limiting stages have the source alternating shape cycle and radius factors 2 : ℕ2, 1 / 2 : ℕ1/2.

theorem limitingAlternatingCycle : regularStep 2 1 0 0 (2 / 3) = (2, 1 / 2, 1, 1) ∧ regularStep (1 / 2) 1 0 0 (1 / 3) = (1 / 2, 2, 1, 1) := by rw [regularStep_zero, regularStep_zero] norm_num

Helper for cm-center-proposition: the removable low factor reconstructs the actual lower eigenvalue of the regular updated matrix.

theorem regularSpectrum_low (p h r μ : ℝ) (hD : regularDenominator r μ ≠ 0) (hE : regularEnergy p r μ ≠ 0) (hhigh : RealSymmetric2.high (regularEntries p h r μ).1 (r * (regularEntries p h r μ).2.1) (regularEntries p h r μ).2.2 ≠ 0) : r ^ 2 * (regularSpectrum p h r μ).1 = RealSymmetric2.low (regularEntries p h r μ).1 (r * (regularEntries p h r μ).2.1) (regularEntries p h r μ).2.2 := by have hdet := regularMetric_det p h r μ hD hE rw [Matrix.det_fin_two, regularMetric_low, regularMetric_high, regularMetric_upper, regularMetric_lower] at hdet rw [RealSymmetric2.low_eq_det_div_high_of_ne _ _ _ hhigh] unfold regularSpectrum dsimp only rw [← pow_two] at hdet rw [hdet] field_simp

Helper for cm-center-proposition: the oriented frame in cm-regular-step has columns and (sorry, sorry) / sorry : ?m.6(Unknown identifier `rb₀`rb₀,Unknown identifier `c₀`c₀)/Unknown identifier `N`N.

def regularFrame (p h r μ τ : ℝ) : Matrix (Fin 2) (Fin 2) ℝ := let a := regularEntries p h r μ let c := a.2.2 - r ^ 2 * (regularSpectrum p h r μ).1 (regularRotatedData p h r μ τ).2⁻¹ • !![c, r * a.2.1; -(r * a.2.1), c]

Helper for cm-center-proposition: rotation into the new source frame sends the actual normalized gradient to (sorry, sorry) / sorry : ?m.6(Unknown identifier `X`X,Unknown identifier `rY`rY)/Unknown identifier `N`N.

theorem regularFrame_gradient (p h r μ τ : ℝ) : (regularFrame p h r μ τ).transpose *ᵥ regularGradient p r μ τ = ![(regularRotatedData p h r μ τ).1.1 / (regularRotatedData p h r μ τ).2, r * (regularRotatedData p h r μ τ).1.2 / (regularRotatedData p h r μ τ).2] := by unfold regularFrame regularGradient regularRotatedData ext i fin_cases i · simp [Matrix.mulVec, Matrix.transpose_apply, dotProduct, Fin.sum_univ_two, Matrix.of_apply, div_eq_mul_inv] ring · simp [Matrix.mulVec, Matrix.transpose_apply, dotProduct, Fin.sum_univ_two, Matrix.of_apply, div_eq_mul_inv] ring

Helper for cm-center-proposition: the source normalization Unknown identifier `N`N is the normalization of the actual lower-eigenvector branch.

theorem regularNormalizer_eq_lowDenom (p h r μ τ : ℝ) (hD : regularDenominator r μ ≠ 0) (hE : regularEnergy p r μ ≠ 0) (hhigh : RealSymmetric2.high (regularEntries p h r μ).1 (r * (regularEntries p h r μ).2.1) (regularEntries p h r μ).2.2 ≠ 0) : (regularRotatedData p h r μ τ).2 = RealSymmetric2.lowDenom (regularEntries p h r μ).1 (r * (regularEntries p h r μ).2.1) (regularEntries p h r μ).2.2 := by rw [RealSymmetric2.lowDenom_apply, ← regularSpectrum_low p h r μ hD hE hhigh] unfold regularRotatedData dsimp only congr 1 ring

Helper for cm-center-proposition: the source frame is precisely the canonical oriented eigenframe of the actual regular updated matrix.

theorem regularFrame_eq_lowEigenframe (p h r μ τ : ℝ) (hD : regularDenominator r μ ≠ 0) (hE : regularEnergy p r μ ≠ 0) (hhigh : RealSymmetric2.high (regularEntries p h r μ).1 (r * (regularEntries p h r μ).2.1) (regularEntries p h r μ).2.2 ≠ 0) : regularFrame p h r μ τ = EuclideanPlane.frame (RealSymmetric2.lowVector (regularEntries p h r μ).1 (r * (regularEntries p h r μ).2.1) (regularEntries p h r μ).2.2) := by have hlow := regularSpectrum_low p h r μ hD hE hhigh have hnormal := regularNormalizer_eq_lowDenom p h r μ τ hD hE hhigh rw [RealSymmetric2.lowVector_apply] unfold RealSymmetric2.lowRaw unfold regularFrame dsimp only rw [hnormal] apply Matrix.ext intro i j have hi : i = 0 ∨ i = 1 := by omega have hj : j = 0 ∨ j = 1 := by omega rcases hi with rfl | rfl · rcases hj with rfl | rfl · simp [EuclideanPlane.frame, EuclideanPlane.perp_apply, ← hlow] · simp [EuclideanPlane.frame, EuclideanPlane.perp_apply, ← hlow] · rcases hj with rfl | rfl · simp [EuclideanPlane.frame, EuclideanPlane.perp_apply, ← hlow] · simp [EuclideanPlane.frame, EuclideanPlane.perp_apply, ← hlow]

Helper for cm-center-proposition: wherever Unknown identifier `N`N is nonzero, the source regular frame is positively oriented orthogonal.

theorem regularFrame_oriented (p h r μ τ : ℝ) (hD : regularDenominator r μ ≠ 0) (hE : regularEnergy p r μ ≠ 0) (hhigh : RealSymmetric2.high (regularEntries p h r μ).1 (r * (regularEntries p h r μ).2.1) (regularEntries p h r μ).2.2 ≠ 0) (hN : (regularRotatedData p h r μ τ).2 ≠ 0) : regularFrame p h r μ τ ∈ Matrix.specialOrthogonalGroup (Fin 2) ℝ := by rw [regularFrame_eq_lowEigenframe p h r μ τ hD hE hhigh] rw [regularNormalizer_eq_lowDenom p h r μ τ hD hE hhigh] at hN exact RealSymmetric2.frame_mem_specialOrthogonalGroup_of_lowDenom_ne_zero _ _ _ hN

Helper for cm-center-proposition: the source radius, shape, and amplitude formulas are exactly spectral recovery from and (sorry, sorry) / sorry : ?m.6(Unknown identifier `GX`GX,Unknown identifier `GrY`GrY)/Unknown identifier `N`N.

theorem regularStep_spectralRecovery (p h r μ τ G : ℝ) (hr : r ≠ 0) (hG : G ≠ 0) (hN : (regularRotatedData p h r μ τ).2 ≠ 0) (hX : (regularRotatedData p h r μ τ).1.1 ≠ 0) (hY : (regularRotatedData p h r μ τ).1.2 ≠ 0) (hL : (regularSpectrum p h r μ).1 ≠ 0) (hH : (regularSpectrum p h r μ).2 ≠ 0) : CycleBoundaryState.recoveryRadius (r ^ 2 * (regularSpectrum p h r μ).1) (regularSpectrum p h r μ).2 (G * (regularRotatedData p h r μ τ).1.1 / (regularRotatedData p h r μ τ).2) (G * r * (regularRotatedData p h r μ τ).1.2 / (regularRotatedData p h r μ τ).2) = r * (regularStep p h r μ τ).1 ∧ CycleBoundaryState.recoveryShape (r ^ 2 * (regularSpectrum p h r μ).1) (regularSpectrum p h r μ).2 (G * (regularRotatedData p h r μ τ).1.1 / (regularRotatedData p h r μ τ).2) (G * r * (regularRotatedData p h r μ τ).1.2 / (regularRotatedData p h r μ τ).2) = (regularStep p h r μ τ).2.1 := by unfold CycleBoundaryState.recoveryRadius CycleBoundaryState.recoveryShape dsimp only [regularStep] constructor · field_simp [hr, hG, hN, hX, hY, hL, hH] · field_simp [hr, hG, hN, hX, hY, hL, hH]

Helper for cm-center-proposition: the exact one-step center change is the source vector in cm-center-step.

theorem spectralCenterDisplacement (z : DFP.AbstractSecantStep (Fin 2)) (p h r μ G : ℝ) (hH : z.inverseHessian = Matrix.diagonal ![h * p * r ^ 2, h]) (hg : z.gradient = G • ![(1 : ℝ), p * r]) (hA : z.secantMatrix = !![1, μ; μ, 1]) : z.displacement - z.gradientChange = (G * z.tau * (p + 1) * μ * r / regularDenominator r μ) • ![(1 : ℝ), r] := by rw [z.gradientChange_def, spectralDisplacement z p h r μ G hH hg hA, hA] ext i fin_cases i · simp ring · simp ring

Helper for cm-center-proposition: the source two-step map is analytic at (0, 2, 1) : ℕ × ℕ × ℕ(0,2,1) and fixes that point.

theorem twoStepAnalytic : AnalyticAt ℝ DFP.TwoLeg.stateMap ((0, 2, 1) : ℝ × ℝ × ℝ) ∧ DFP.TwoLeg.stateMap (0, 2, 1) = (0, 2, 1) := by exact ⟨DFP.TwoLeg.stateMapAnalytic, DFP.TwoLeg.stateMap_base⟩

Helper for cm-center-proposition: the zero-scale source cycle has exactly the displayed rational radius, shape, and high eigenvalue factors.

theorem twoStepZeroSlice (p h : ℝ) (hp : 0 < p) (hh : 0 < h) : DFP.TwoLeg.radiusFactor 0 p h = 9 * h * p * (p + 1) / (2 * (9 * h * p + (p + 1) ^ 2)) ∧ (DFP.TwoLeg.stateMap (0, p, h)).2 = (4 * (9 * h * p + (p + 1) ^ 2) ^ 2 / (81 * h * p * (p + 1) ^ 2), 1) := by exact ⟨DFP.TwoLeg.radiusFactor_zero p h hp hh, DFP.TwoLeg.stateMap_zero_transverse p h hp hh⟩

Helper for cm-center-proposition: the linearization part. The derivative of the exact source two-step map is and has the three source eigenvalues 1 : ℕ1, , and 0 : ℕ0. Graph existence is a separate part.

theorem centerMapLinearization : (∀ v : ℝ × ℝ × ℝ, fderiv ℝ DFP.TwoLeg.stateMap (0, 2, 1) v = (v.1, (-(1 : ℝ) / 9) * v.2.1 + (2 / 3) * v.2.2, 0)) ∧ Module.End.HasEigenvalue (fderiv ℝ DFP.TwoLeg.stateMap (0, 2, 1)).toLinearMap (1 : ℝ) ∧ Module.End.HasEigenvalue (fderiv ℝ DFP.TwoLeg.stateMap (0, 2, 1)).toLinearMap (-(1 : ℝ) / 9) ∧ Module.End.HasEigenvalue (fderiv ℝ DFP.TwoLeg.stateMap (0, 2, 1)).toLinearMap (0 : ℝ) := by exact ⟨DFP.TwoLeg.stateMap_fderiv_apply, DFP.TwoLeg.stateMap_centerEigenvalue, DFP.TwoLeg.stateMap_transverseEigenvalue_negNinth, DFP.TwoLeg.stateMap_transverseEigenvalue_zero⟩

Helper for cm-center-proposition: the literal initial point and gradient are and the literal initial matrix is .

theorem prescribedInitialization (p h : ℝ → ℝ) (ε₀ : ℝ) : (DFP.TwoPhaseOrbit.State.initial p h ε₀).point = WithLp.toLp 2 ![(1 : ℝ), p ε₀ * ε₀ ^ 2] ∧ (DFP.TwoPhaseOrbit.State.initial p h ε₀).gradient = WithLp.toLp 2 ![(1 : ℝ), p ε₀ * ε₀ ^ 2] ∧ (DFP.TwoPhaseOrbit.State.initial p h ε₀).metric = Matrix.diagonal ![h ε₀ * p ε₀ * ε₀ ^ 4, h ε₀] := by have hgradient : (DFP.TwoPhaseOrbit.State.initial p h ε₀).gradient = WithLp.toLp 2 ![(1 : ℝ), p ε₀ * ε₀ ^ 2] := by rw [DFP.TwoPhaseOrbit.State.gradient_def, DFP.TwoPhaseOrbit.State.initial_amplitude, DFP.TwoPhaseOrbit.State.initial_frame, DFP.TwoPhaseOrbit.State.initial_p, DFP.TwoPhaseOrbit.State.initial_epsilon] simp refine ⟨?_, hgradient, ?_⟩ · rw [DFP.TwoPhaseOrbit.State.initial_point, hgradient] · rw [DFP.TwoPhaseOrbit.State.metric_def, DFP.TwoPhaseOrbit.State.initial_frame, DFP.TwoPhaseOrbit.State.initial_h, DFP.TwoPhaseOrbit.State.initial_p, DFP.TwoPhaseOrbit.State.initial_epsilon] simp

Helper for cm-center-proposition: the second-stage factor package exposes its spectral, gradient, and canonical pairs.

theorem secondFactors_apply (ε p h : ℝ) : DFP.SecondLeg.factors ε p h = (DFP.SecondLeg.spectralFactors ε p h, DFP.SecondLeg.gradientFactors ε p h, DFP.SecondLeg.canonicalFactors ε p h) := by exact DFP.SecondLeg.factors_apply ε p h

Helper for cm-center-proposition: the ten positive removable factors for both actual stages, in spectral, gradient, and radius order.

def twoStageFactors (x : ℝ × ℝ × ℝ) : Fin 10 → ℝ := let first := DFP.FirstLeg.factors x.1 x.2.1 x.2.2 let second := DFP.SecondLeg.factors x.1 x.2.1 x.2.2 ![first.1.1, first.1.2, first.2.1.1, first.2.1.2, first.2.2.1, second.1.1, second.1.2, second.2.1.1, second.2.1.2, second.2.2.1]

Helper for cm-center-proposition: the ten-factor interface refers to the original stage spectral, gradient, and radius factors.

theorem twoStageFactors_apply (ε p h : ℝ) : twoStageFactors (ε, p, h) = ![(DFP.FirstLeg.spectralFactors ε p h).1, (DFP.FirstLeg.spectralFactors ε p h).2, (DFP.FirstLeg.gradientFactors ε p h).1, (DFP.FirstLeg.gradientFactors ε p h).2, (DFP.FirstLeg.canonicalFactors ε p h).1, (DFP.SecondLeg.spectralFactors ε p h).1, (DFP.SecondLeg.spectralFactors ε p h).2, (DFP.SecondLeg.gradientFactors ε p h).1, (DFP.SecondLeg.gradientFactors ε p h).2, (DFP.SecondLeg.canonicalFactors ε p h).1] := by simp only [twoStageFactors, DFP.FirstLeg.factors_apply, secondFactors_apply]

Helper for cm-center-proposition: both actual stages have continuous removable factor data at the common base state.

theorem twoStageFactors_continuousAt : ContinuousAt twoStageFactors ((0, 2, 1) : ℝ × ℝ × ℝ) := by have hfirst := DFP.FirstLeg.factorsAnalytic.continuousAt have hsecond := DFP.SecondLeg.factorsAnalytic.continuousAt unfold twoStageFactors fun_prop

Helper for cm-center-proposition: the ten actual stage factors at the base state are (2, 1, 1, 1, 2, 2, 1, 1, 2, 1) : ℕ × ℕ × ℕ × ℕ × ℕ × ℕ × ℕ × ℕ × ℕ × ℕ(2,1,1,1,2,2,1,1,2,1).

theorem twoStageFactors_base : twoStageFactors ((0, 2, 1) : ℝ × ℝ × ℝ) = ![(2 : ℝ), 1, 1, 1, 2, 2, 1, 1, 2, 1] := by simp only [twoStageFactors, DFP.FirstLeg.factorsBase, DFP.SecondLeg.factorsBase]

Helper for cm-center-proposition: one fixed neighborhood bounds all ten actual stage factors strictly above 1 / 2 : ℕ1/2.

theorem twoStageFactors_eventually_half_lt : ∀ᶠ x in 𝓝 ((0, 2, 1) : ℝ × ℝ × ℝ), ∀ i, (1 / 2 : ℝ) < twoStageFactors x i := by apply Filter.eventually_all.mpr intro i have hbase : (1 / 2 : ℝ) < twoStageFactors ((0, 2, 1) : ℝ × ℝ × ℝ) i := by rw [twoStageFactors_base] fin_cases i all_goals norm_num have hcontinuous := (continuous_apply i).continuousAt.comp twoStageFactors_continuousAt exact hcontinuous.eventually (Ioi_mem_nhds hbase)

Helper for cm-center-proposition: positive scale, amplitude, and oriented frame give valid actual two-stage data on a fixed neighborhood of (0, 2, 1) : ℕ × ℕ × ℕ(0,2,1).

theorem phaseValidity_near_base : ∀ᶠ x in 𝓝 ((0, 2, 1) : ℝ × ℝ × ℝ), ∀ s : DFP.TwoPhaseOrbit.State, s.coordinates = x → s.frame ∈ Matrix.specialOrthogonalGroup (Fin 2) ℝ → 0 < s.amplitude → 0 < s.ε → s.PhaseValidity := by have hshape : ∀ᶠ x : ℝ × ℝ × ℝ in 𝓝 (0, 2, 1), 0 < x.2.1 := by have hbase : (0 : ℝ) < 2 := by norm_num exact (continuousAt_fst.comp continuousAt_snd).eventually (Ioi_mem_nhds hbase) have hhigh : ∀ᶠ x : ℝ × ℝ × ℝ in 𝓝 (0, 2, 1), 0 < x.2.2 := by have hbase : (0 : ℝ) < 1 := by norm_num exact (continuousAt_snd.comp continuousAt_snd).eventually (Ioi_mem_nhds hbase) have hscale : ∀ᶠ x : ℝ × ℝ × ℝ in 𝓝 (0, 2, 1), x.1 < 1 / 4 := by have hbase : (0 : ℝ) < 1 / 4 := by norm_num exact continuousAt_fst.eventually (Iio_mem_nhds hbase) filter_upwards [twoStageFactors_eventually_half_lt, hshape, hhigh, hscale] with x hfactor hp hh hεlt intro s hcoordinates hframe hG hε have htuple : (s.ε, s.p, s.h) = x := by rw [← DFP.TwoPhaseOrbit.State.coordinates_def] exact hcoordinates rw [← htuple] at hfactor hp hh hεlt have hpositive : ∀ i, 0 < twoStageFactors (s.ε, s.p, s.h) i := by intro i have hhalf : (0 : ℝ) < 1 / 2 := by norm_num exact hhalf.trans (hfactor i) apply DFP.TwoPhaseOrbit.State.PhaseValidity.ofComponents s hframe hε hεlt hp hh hG · simpa [twoStageFactors_apply] using hpositive 0 · simpa [twoStageFactors_apply] using hpositive 1 · simpa [twoStageFactors_apply] using hpositive 2 · simpa [twoStageFactors_apply] using hpositive 3 · simpa [twoStageFactors_apply] using hpositive 4 · simpa [twoStageFactors_apply] using hpositive 5 · simpa [twoStageFactors_apply] using hpositive 6 · simpa [twoStageFactors_apply] using hpositive 7 · simpa [twoStageFactors_apply] using hpositive 8 · simpa [twoStageFactors_apply] using hpositive 9

Helper for cm-center-proposition: a fixed regular neighborhood gives exact agreement with both actual inverse DFP stages, without any invariant-graph premise.

theorem exactCycle_near_base : ∃ δ > 0, ∀ s : DFP.TwoPhaseOrbit.State, dist s.coordinates (0, 2, 1) < δ → s.frame ∈ Matrix.specialOrthogonalGroup (Fin 2) ℝ → 0 < s.amplitude → 0 < s.ε → s.ExactCycle := by obtain ⟨δ, hδ, hvalid⟩ := Metric.eventually_nhds_iff.mp phaseValidity_near_base refine ⟨δ, hδ, ?_⟩ intro s hdist hframe hG hε exact DFP.TwoPhaseOrbit.State.exactCycle_of_phaseValidity s (hvalid hdist s rfl hframe hG hε)

Helper for cm-center-proposition: actual physical successor coordinates and frame are the source signed two-step map and its two ordered eigenframe updates.

theorem physicalTwoStep (s : DFP.TwoPhaseOrbit.State) : s.next.coordinates = DFP.TwoLeg.stateMap s.coordinates ∧ s.next.frame = s.middleFrame * DFP.SecondLeg.frame s.ε s.p s.h ∧ s.next.point = s.point + s.firstDisplacement + s.secondDisplacement := by exact ⟨DFP.TwoPhaseOrbit.State.next_coordinates s, DFP.TwoPhaseOrbit.State.next_frame s, (DFP.TwoPhaseOrbit.State.next_point s).trans (congrArg (fun x ↦ x + s.secondDisplacement) (DFP.TwoPhaseOrbit.State.middlePoint_def s))⟩

Helper for cm-center-proposition: orthogonal congruence preserves positive definiteness of the actual local inverse Hessian.

theorem frameConjugate_posDef (H R : Matrix (Fin 2) (Fin 2) ℝ) (hH : H.PosDef) (hR : R ∈ Matrix.specialOrthogonalGroup (Fin 2) ℝ) : (R * H * R.transpose).PosDef := by have hdet : R.det = 1 := (Matrix.mem_specialOrthogonalGroup_iff.mp hR).2 have hunit : IsUnit R := by apply (Matrix.isUnit_iff_isUnit_det R).mpr rw [hdet] exact isUnit_one rw [← Matrix.conjTranspose_eq_transpose_of_trivial R, ← Matrix.star_eq_conjTranspose] exact (Matrix.IsUnit.posDef_star_right_conjugate_iff hunit).mpr hH

Helper for cm-center-proposition: the first source eigenframe update preserves the orientation of the physical middle frame.

theorem middleFrame_oriented (s : DFP.TwoPhaseOrbit.State) (h : s.PhaseValidity) : s.middleFrame ∈ Matrix.specialOrthogonalGroup (Fin 2) ℝ := by rw [DFP.TwoPhaseOrbit.State.middleFrame_def] exact mul_mem h.frame_specialOrthogonal (DFP.FirstLeg.frame_mem_specialOrthogonalGroup s.ε s.p s.h (ne_of_gt h.firstGradientLow_pos))

Helper for cm-center-proposition: a single fixed neighborhood gives two exact source stages and positive physical matrices, before any initial scale is selected.

theorem localSpectralRecurrence : ∃ δ > 0, ∀ s : DFP.TwoPhaseOrbit.State, dist s.coordinates (0, 2, 1) < δ → s.frame ∈ Matrix.specialOrthogonalGroup (Fin 2) ℝ → 0 < s.amplitude → 0 < s.ε → s.ExactCycle ∧ s.metric.PosDef ∧ s.middleMetric.PosDef ∧ s.next.metric.PosDef := by obtain ⟨δ, hδ, hexact⟩ := exactCycle_near_base refine ⟨δ, hδ, ?_⟩ intro s hdist hframe hG hε have hcycle := hexact s hdist hframe hG hε exact ⟨hcycle, physicalMatrices_posDef s hcycle.valid⟩

Helper for cm-center-proposition: the first-stage source formulas agree with the actual first leg in a valid physical cycle.

theorem firstStageRegularStep (s : DFP.TwoPhaseOrbit.State) (h : s.PhaseValidity) : DFP.FirstLeg.outputMetric s.ε s.p s.h = regularMetric s.p s.h (s.ε ^ 2) s.ε ∧ s.amplitude • DFP.FirstLeg.outputGradient s.ε s.p s.h = s.amplitude • regularGradient s.p (s.ε ^ 2) s.ε (2 / 3) := by have hmetric : (DFP.TwoPhaseOrbit.State.firstStep s h).inverseHessian = Matrix.diagonal ![s.h * s.p * (s.ε ^ 2) ^ 2, s.h] := by rw [DFP.TwoPhaseOrbit.State.firstStep_inverseHessian, ← pow_mul] have hgradient : (DFP.TwoPhaseOrbit.State.firstStep s h).gradient = s.amplitude • ![(1 : ℝ), s.p * s.ε ^ 2] := DFP.TwoPhaseOrbit.State.firstStep_gradient s h have hcontrol : (DFP.TwoPhaseOrbit.State.firstStep s h).secantMatrix = !![1, s.ε; s.ε, 1] := by rw [DFP.TwoPhaseOrbit.State.firstStep_secantMatrix, TwoPhaseControls.first_matrix] have houtput := DFP.TwoPhaseOrbit.State.firstStep_output s h constructor · exact (congrArg Prod.fst houtput).symm.trans (spectralMetricUpdate _ s.p s.h (s.ε ^ 2) s.ε s.amplitude hmetric hgradient hcontrol) · have hregular := spectralGradientUpdate (DFP.TwoPhaseOrbit.State.firstStep s h) s.p s.h (s.ε ^ 2) s.ε s.amplitude hmetric hgradient hcontrol rw [DFP.TwoPhaseOrbit.State.firstStep_tau, TwoPhaseControls.first_tau] at hregular exact (congrArg Prod.snd houtput).symm.trans hregular

Helper for cm-center-proposition: the first canonical factors expose the source recovery quotients for the updated radius and shape.

theorem firstCanonicalFactors_apply (ε p h : ℝ) : DFP.FirstLeg.canonicalFactors ε p h = ((DFP.FirstLeg.spectralFactors ε p h).1 * (DFP.FirstLeg.gradientFactors ε p h).1 / ((DFP.FirstLeg.spectralFactors ε p h).2 * (DFP.FirstLeg.gradientFactors ε p h).2), (DFP.FirstLeg.spectralFactors ε p h).2 * (DFP.FirstLeg.gradientFactors ε p h).2 ^ 2 / ((DFP.FirstLeg.spectralFactors ε p h).1 * (DFP.FirstLeg.gradientFactors ε p h).1 ^ 2)) := by rfl

Helper for cm-center-proposition: after the first rotation the second step's actual inputs have precisely the recovered form.

theorem middleCanonicalInput (s : DFP.TwoPhaseOrbit.State) (h : s.PhaseValidity) : (DFP.TwoPhaseOrbit.State.secondStep s h).inverseHessian = Matrix.diagonal ![(DFP.FirstLeg.spectralFactors s.ε s.p s.h).2 * (DFP.FirstLeg.canonicalFactors s.ε s.p s.h).2 * (s.ε ^ 2 * (DFP.FirstLeg.canonicalFactors s.ε s.p s.h).1) ^ 2, (DFP.FirstLeg.spectralFactors s.ε s.p s.h).2] ∧ (DFP.TwoPhaseOrbit.State.secondStep s h).gradient = (s.amplitude * (DFP.FirstLeg.gradientFactors s.ε s.p s.h).1) • ![(1 : ℝ), (DFP.FirstLeg.canonicalFactors s.ε s.p s.h).2 * (s.ε ^ 2 * (DFP.FirstLeg.canonicalFactors s.ε s.p s.h).1)] := by have hL := ne_of_gt h.firstSpectralLow_pos have hH := ne_of_gt h.firstSpectralHigh_pos have hQ := ne_of_gt h.firstGradientLow_pos have hU := ne_of_gt h.firstGradientHigh_pos constructor · rw [DFP.TwoPhaseOrbit.State.secondStep_inverseHessian, firstCanonicalFactors_apply] congr 1 ext i fin_cases i · norm_num field_simp [hL, hH, hQ, hU] · norm_num · rw [DFP.TwoPhaseOrbit.State.secondStep_gradient, firstCanonicalFactors_apply] ext i fin_cases i · norm_num · norm_num field_simp [hL, hH, hQ, hU]

Helper for cm-center-proposition: the actual second leg is the same regular step applied to the UPDATED spectral coordinates and the ORIGINAL cycle scale.

theorem secondStageRegularStep (s : DFP.TwoPhaseOrbit.State) (h : s.PhaseValidity) : DFP.SecondLeg.outputMetric s.ε s.p s.h = regularMetric (DFP.FirstLeg.canonicalFactors s.ε s.p s.h).2 (DFP.FirstLeg.spectralFactors s.ε s.p s.h).2 (s.ε ^ 2 * (DFP.FirstLeg.canonicalFactors s.ε s.p s.h).1) (-2 * s.ε) ∧ s.amplitude • DFP.SecondLeg.outputGradient s.ε s.p s.h = (s.amplitude * (DFP.FirstLeg.gradientFactors s.ε s.p s.h).1) • regularGradient (DFP.FirstLeg.canonicalFactors s.ε s.p s.h).2 (s.ε ^ 2 * (DFP.FirstLeg.canonicalFactors s.ε s.p s.h).1) (-2 * s.ε) (1 / 3) := by obtain ⟨hmetric, hgradient⟩ := middleCanonicalInput s h have hcontrol : (DFP.TwoPhaseOrbit.State.secondStep s h).secantMatrix = !![1, -2 * s.ε; -2 * s.ε, 1] := by rw [DFP.TwoPhaseOrbit.State.secondStep_secantMatrix, TwoPhaseControls.second_matrix] have houtput := DFP.TwoPhaseOrbit.State.secondStep_output s h constructor · exact (congrArg Prod.fst houtput).symm.trans (spectralMetricUpdate _ _ _ _ _ _ hmetric hgradient hcontrol) · have hregular := spectralGradientUpdate (DFP.TwoPhaseOrbit.State.secondStep s h) _ _ _ _ _ hmetric hgradient hcontrol rw [DFP.TwoPhaseOrbit.State.secondStep_tau, TwoPhaseControls.second_tau] at hregular exact (congrArg Prod.snd houtput).symm.trans hregular

Helper for cm-center-proposition: the parameter choices in the source graph transform make its slope, uniform contraction, and all order-2 through order-8 highest-derivative coefficients strictly contractive simultaneously.

theorem graphTransformParameters (stableRate slope : ℝ) (hstable : stableRate < 1) (hslope : 0 < slope) : ∃ δ > 0, let m := 1 - δ * (1 + slope) let η := δ + (stableRate + δ) * slope 0 < m ∧ η / m < slope ∧ stableRate + δ + η * δ / m < 1 ∧ ∀ r : Fin 7, (stableRate + δ * (1 + slope)) / m ^ (r.val + 2) < 1 := by let m : ℝ → ℝ := fun δ ↦ 1 - δ * (1 + slope) let η : ℝ → ℝ := fun δ ↦ δ + (stableRate + δ) * slope have hm : ContinuousAt m 0 := by dsimp only [m] fun_prop have hη : ContinuousAt η 0 := by dsimp only [η] fun_prop have hmzero : m 0 ≠ 0 := by simp [m] have hmpos : ∀ᶠ δ in 𝓝 (0 : ℝ), 0 < m δ := by have hbase : 0 < m 0 := by simp [m] exact hm.eventually (Ioi_mem_nhds hbase) have hslopeBase : η 0 / m 0 < slope := by simp only [η, m, zero_mul, sub_zero, add_zero, zero_add, div_one] nlinarith have hslopeEvent : ∀ᶠ δ in 𝓝 (0 : ℝ), η δ / m δ < slope := (hη.div hm hmzero).eventually (Iio_mem_nhds hslopeBase) have hcontractionContinuous : ContinuousAt (fun δ ↦ stableRate + δ + η δ * δ / m δ) 0 := by exact (continuousAt_const.add continuousAt_id).add ((hη.mul continuousAt_id).div hm hmzero) have hcontractionBase : stableRate + 0 + η 0 * 0 / m 0 < 1 := by simpa using hstable have hcontraction : ∀ᶠ δ in 𝓝 (0 : ℝ), stableRate + δ + η δ * δ / m δ < 1 := hcontractionContinuous.eventually (Iio_mem_nhds hcontractionBase) have hhighest : ∀ᶠ δ in 𝓝 (0 : ℝ), ∀ r : Fin 7, (stableRate + δ * (1 + slope)) / m δ ^ (r.val + 2) < 1 := by apply Filter.eventually_all.mpr intro r have hcontinuous : ContinuousAt (fun δ ↦ (stableRate + δ * (1 + slope)) / m δ ^ (r.val + 2)) 0 := by have hnum : ContinuousAt (fun δ : ℝ ↦ stableRate + δ * (1 + slope)) 0 := by fun_prop exact hnum.div (hm.pow _) (pow_ne_zero _ hmzero) have hbase : (stableRate + 0 * (1 + slope)) / m 0 ^ (r.val + 2) < 1 := by simpa [m] using hstable exact hcontinuous.eventually (Iio_mem_nhds hbase) have hall := hmpos.and (hslopeEvent.and (hcontraction.and hhighest)) obtain ⟨radius, hradius, hbound⟩ := Metric.eventually_nhds_iff.mp hall have hδ : 0 < radius / 2 := half_pos hradius have hdist : dist (radius / 2) 0 < radius := by rw [Real.dist_eq, sub_zero, abs_of_pos hδ] exact half_lt_self hradius exact ⟨radius / 2, hδ, hbound hdist⟩

Helper for cm-center-proposition: the stable linear block of the source map, in the two transverse coordinates .

def stableDerivative : (ℝ × ℝ) →L[ℝ] (ℝ × ℝ) := (((-(1 : ℝ) / 9) • ContinuousLinearMap.fst ℝ ℝ ℝ) + ((2 : ℝ) / 3) • ContinuousLinearMap.snd ℝ ℝ ℝ).prod (0 : (ℝ × ℝ) →L[ℝ] ℝ)

Helper for cm-center-proposition: the stable block has the exact source triangular action.

theorem stableDerivative_apply (z : ℝ × ℝ) : stableDerivative z = ((-(1 : ℝ) / 9) * z.1 + (2 / 3) * z.2, 0) := by simp [stableDerivative]

Helper for cm-center-proposition: the same source stable block strictly contracts the product norm; the bound 7 / 9 : ℕ7/9 is a coordinate-norm adapter.

theorem stableDerivative_norm_le : ‖stableDerivative‖ ≤ (7 / 9 : ℝ) := by apply ContinuousLinearMap.opNorm_le_bound · norm_num · intro z rw [stableDerivative_apply] apply norm_prod_le_iff.mpr constructor · calc ‖(-(1 : ℝ) / 9) * z.1 + (2 / 3) * z.2‖ ≤ ‖(-(1 : ℝ) / 9) * z.1‖ + ‖(2 / 3) * z.2‖ := norm_add_le _ _ _ = (1 / 9 : ℝ) * ‖z.1‖ + (2 / 3) * ‖z.2‖ := by rw [norm_mul, norm_mul] norm_num _ ≤ (7 / 9 : ℝ) * ‖z‖ := by nlinarith [norm_fst_le z, norm_snd_le z] · simp only [norm_zero] positivity

Helper for cm-center-proposition: the source transverse derivative has operator norm strictly below one.

theorem stableDerivative_norm_lt_one : ‖stableDerivative‖ < 1 := by have hrate : (7 / 9 : ℝ) < 1 := by norm_num exact stableDerivative_norm_le.trans_lt hrate

Helper for cm-center-proposition: translate the actual source map to the center-stable origin, without changing its scale coordinate or parameter map.

def centeredStateMap (x : ℝ × ℝ × ℝ) : ℝ × ℝ × ℝ := DFP.TwoLeg.stateMap (x + (0, 2, 1)) - (0, 2, 1)

Helper for cm-center-proposition: the centered map is the actual source map expressed in coordinates.

theorem centeredStateMap_apply (ε u v : ℝ) : centeredStateMap (ε, u, v) = let y := DFP.TwoLeg.stateMap (ε, u + 2, v + 1) (y.1, y.2.1 - 2, y.2.2 - 1) := by simp only [centeredStateMap, Prod.mk_add_mk, add_zero] apply Prod.ext · simp · rfl

Helper for cm-center-proposition: the centered source map fixes zero.

theorem centeredStateMap_zero : centeredStateMap 0 = 0 := by simp [centeredStateMap, DFP.TwoLeg.stateMap_base]

Helper for cm-center-proposition: the centered source map is analytic at zero, so all finite smoothness orders required by the graph transform are available.

theorem centeredStateMap_analyticAt : AnalyticAt ℝ centeredStateMap 0 := by have hinput : AnalyticAt ℝ (fun x : ℝ × ℝ × ℝ ↦ x + (0, 2, 1)) 0 := analyticAt_id.add analyticAt_const have houter : AnalyticAt ℝ DFP.TwoLeg.stateMap ((0 : ℝ × ℝ × ℝ) + (0, 2, 1)) := by simpa only [zero_add] using DFP.TwoLeg.stateMapAnalytic have hcomposition := houter.comp (f := fun x : ℝ × ℝ × ℝ ↦ x + (0, 2, 1)) (x := 0) hinput have hsub : AnalyticAt ℝ (fun x : ℝ × ℝ × ℝ ↦ DFP.TwoLeg.stateMap (x + (0, 2, 1)) - (0, 2, 1)) 0 := hcomposition.sub analyticAt_const exact hsub

Helper for cm-center-proposition: the derivative of the translated source map is exactly its center identity plus the stable block.

theorem centeredStateMap_hasFDerivAt : HasFDerivAt centeredStateMap (LocalCutoff.centerStable stableDerivative) 0 := by have hinput : HasFDerivAt (fun x : ℝ × ℝ × ℝ ↦ x + (0, 2, 1)) (ContinuousLinearMap.id ℝ (ℝ × ℝ × ℝ)) 0 := (hasFDerivAt_id 0).add_const (0, 2, 1) have houter : HasFDerivAt DFP.TwoLeg.stateMap (fderiv ℝ DFP.TwoLeg.stateMap (0, 2, 1)) ((0 : ℝ × ℝ × ℝ) + (0, 2, 1)) := by simpa only [zero_add] using DFP.TwoLeg.stateMapAnalytic.differentiableAt.hasFDerivAt have hraw := ((houter.comp 0 hinput).sub_const ((0, 2, 1) : ℝ × ℝ × ℝ)) have hcentered : HasFDerivAt centeredStateMap ((fderiv ℝ DFP.TwoLeg.stateMap (0, 2, 1)).comp (ContinuousLinearMap.id ℝ (ℝ × ℝ × ℝ))) 0 := by apply hraw.congr_of_eventuallyEq filter_upwards [] with x rfl apply hcentered.congr_fderiv apply ContinuousLinearMap.ext intro v rcases v with ⟨ε, u, v⟩ simp only [ContinuousLinearMap.comp_apply, ContinuousLinearMap.id_apply, DFP.TwoLeg.stateMap_fderiv_apply, LocalCutoff.centerStable_apply, stableDerivative_apply]

Helper for cm-center-proposition: an invariant C⁷ graph through the base, tangent to the scale axis, has the actual source map's determined Taylor jets. This is coefficient extraction from a supplied graph, not graph existence.

theorem invariantGraphTaylorJets (p h : ℝ → ℝ) (hsmooth : ContDiffAt ℝ 7 (fun ε ↦ (p ε, h ε)) 0) (hbase : (p 0, h 0) = (2, 1)) (htangent : HasFDerivAt (fun ε ↦ (p ε, h ε)) (0 : ℝ →L[ℝ] (ℝ × ℝ)) 0) (hinvariant : (fun ε ↦ DFP.TwoLeg.stateMap (ε, p ε, h ε)) =ᶠ[𝓝 0] (fun ε ↦ let nextScale := (DFP.TwoLeg.stateMap (ε, p ε, h ε)).1 (nextScale, p nextScale, h nextScale))) : (fun ε : ℝ ↦ p ε - (2 + (198 / 5) * ε ^ 3 - (9 / 5) * ε ^ 4)) =O[𝓝 0] (fun ε : ℝ ↦ ε ^ 5) ∧ (fun ε : ℝ ↦ h ε - (1 + 8 * ε ^ 3)) =O[𝓝 0] (fun ε : ℝ ↦ ε ^ 5) ∧ (fun ε : ℝ ↦ (DFP.TwoLeg.stateMap (ε, p ε, h ε)).1 - (ε - (3 / 2) * ε ^ 4 + (5 / 4) * ε ^ 5)) =O[𝓝 0] (fun ε : ℝ ↦ ε ^ 6) := by have htwo : (2 : WithTop ℕ∞) ≤ 7 := by norm_num have hfive : (5 : WithTop ℕ∞) ≤ 7 := by norm_num obtain Tactic `rcases` failed: `x✝ : ?m.416` is not an inductive datatype⟨hpFirst, hhFirst, hpSecond, hhSecond⟩ := Unknown identifier `DFP.TwoLeg.slowCurveInvariantGraph_secondOrderFlat`DFP.TwoLeg.slowCurveInvariantGraph_secondOrderFlat p h (hsmooth.of_le htwo) hbase htangent hinvariant obtain ⟨P₃, H₃, P₄, H₄, hpJet, hhJet⟩ := DFP.TwoLeg.exists_slowCurveTaylorCoefficients p h (hsmooth.of_le hfive) hbase hpFirst hhFirst hpSecond hhSecond obtain ⟨hshapeCubic, hhighCubic, hshapeQuartic, hhighQuartic⟩ := DFP.TwoLeg.slowCurveInvariantGraph_coefficientEquations p h P₃ H₃ P₄ H₄ hinvariant hpJet hhJet have hH₃ : H₃ = 8 := by linarith have hP₃ : P₃ = 198 / 5 := by linarith have hH₄ : H₄ = 0 := by linarith have hP₄ : P₄ = -9 / 5 := by linarith have hpTarget : (fun ε : ℝ ↦ p ε - (2 + (198 / 5) * ε ^ 3 - (9 / 5) * ε ^ 4)) =O[𝓝 0] (fun ε : ℝ ↦ ε ^ 5) := by apply hpJet.congr' · filter_upwards [] with ε rw [hP₃, hP₄] ring · rfl have hhTarget : (fun ε : ℝ ↦ h ε - (1 + 8 * ε ^ 3)) =O[𝓝 0] (fun ε : ℝ ↦ ε ^ 5) := by simpa only [hH₃, hH₄, zero_mul, add_zero] using hhJet have hrecurrence := DFP.TwoLeg.slowGraphSignedRecurrence p h hpTarget hhTarget refine ⟨hpTarget, hhTarget, ?_⟩ simpa only [DFP.TwoLeg.stateMap_fst] using hrecurrence

Helper for cm-center-proposition: coefficient extraction gives precisely the three uniform cubic/quartic expansions displayed in cm-center-expansions. The invariant graph itself must still be constructed by the source graph transform.

theorem invariantGraphSourceExpansions (p h : ℝ → ℝ) (hsmooth : ContDiffAt ℝ 7 (fun ε ↦ (p ε, h ε)) 0) (hbase : (p 0, h 0) = (2, 1)) (htangent : HasFDerivAt (fun ε ↦ (p ε, h ε)) (0 : ℝ →L[ℝ] (ℝ × ℝ)) 0) (hinvariant : (fun ε ↦ DFP.TwoLeg.stateMap (ε, p ε, h ε)) =ᶠ[𝓝 0] (fun ε ↦ let nextScale := (DFP.TwoLeg.stateMap (ε, p ε, h ε)).1 (nextScale, p nextScale, h nextScale))) : (fun ε : ℝ ↦ p ε - (2 + (198 / 5) * ε ^ 3)) =O[𝓝 0] (fun ε : ℝ ↦ ε ^ 4) ∧ (fun ε : ℝ ↦ h ε - (1 + 8 * ε ^ 3)) =O[𝓝 0] (fun ε : ℝ ↦ ε ^ 4) ∧ (fun ε : ℝ ↦ (DFP.TwoLeg.stateMap (ε, p ε, h ε)).1 - (ε - (3 / 2) * ε ^ 4)) =O[𝓝 0] (fun ε : ℝ ↦ ε ^ 5) := by obtain ⟨hpJet, hhJet, hscaleJet⟩ := invariantGraphTaylorJets p h hsmooth hbase htangent hinvariant have hfourFive : 4 < 5 := by norm_num have hfiveSix : 5 < 6 := by norm_num have hpowerFiveFour : (fun ε : ℝ ↦ ε ^ 5) =O[𝓝 0] (fun ε : ℝ ↦ ε ^ 4) := (Asymptotics.isLittleO_pow_pow hfourFive).isBigO have hpowerSixFive : (fun ε : ℝ ↦ ε ^ 6) =O[𝓝 0] (fun ε : ℝ ↦ ε ^ 5) := (Asymptotics.isLittleO_pow_pow hfiveSix).isBigO have hshapeCorrection : (fun ε : ℝ ↦ (9 / 5) * ε ^ 4) =O[𝓝 0] (fun ε : ℝ ↦ ε ^ 4) := (Asymptotics.isBigO_refl (fun ε : ℝ ↦ ε ^ 4) (𝓝 0)).const_mul_left (9 / 5) have hscaleCorrection : (fun ε : ℝ ↦ (5 / 4) * ε ^ 5) =O[𝓝 0] (fun ε : ℝ ↦ ε ^ 5) := (Asymptotics.isBigO_refl (fun ε : ℝ ↦ ε ^ 5) (𝓝 0)).const_mul_left (5 / 4) constructor · apply ((hpJet.trans hpowerFiveFour).sub hshapeCorrection).congr' · filter_upwards [] with ε ring · rfl · constructor · exact hhJet.trans hpowerFiveFour · apply ((hscaleJet.trans hpowerSixFive).add hscaleCorrection).congr' · filter_upwards [] with ε ring · rfl

Helper for cm-center-proposition: the actual two-stage spectral recovery is the source signed square-root scale update and its two transverse coordinates.

theorem twoStepActualRecovery : ∀ᶠ x in 𝓝 ((0, 2, 1) : ℝ × ℝ × ℝ), 0 < x.1 → DFP.TwoLeg.stateMap x = (Real.sqrt (DFP.SecondLeg.recovered x.1 x.2.1 x.2.2).1, (DFP.SecondLeg.recovered x.1 x.2.1 x.2.2).2, (DFP.SecondLeg.eigenvalues x.1 x.2.1 x.2.2).2) := by exact DFP.TwoLeg.stateMap_eventuallyEq_recovered

Helper for cm-center-proposition: the exact scalar curvature ratio of a source secant step is for Unknown identifier `τ`sorry ≤ 1 : Propτ ≤ 1, before specifying an objective function.

theorem oneStepCurvatureRatio {ι : Type*} [Fintype ι] (z : DFP.AbstractSecantStep ι) (hτ : z.tau ≤ 1) : |z.nextGradient ⬝ᵥ z.displacement| / |z.gradient ⬝ᵥ z.displacement| = 1 - z.tau := by have hpair := z.nextSlope_eq rw [DFP.AbstractSecantStep.nextSlope_def, DFP.AbstractSecantStep.slope_def] at hpair have hdecrease := z.predictedDecrease_pos rw [DFP.AbstractSecantStep.predictedDecrease_def] at hdecrease have hnonzero : z.gradient ⬝ᵥ z.displacement ≠ 0 := ne_of_lt (neg_pos.mp hdecrease) rw [hpair, abs_mul, abs_of_nonneg (sub_nonneg.mpr hτ), mul_div_assoc, div_self (abs_ne_zero.mpr hnonzero), mul_one]

Helper for cm-center-proposition: a supplied invariant C⁷ source graph bundles as the existing SlowCurve with its Taylor coefficients DERIVED from invariance.

theorem invariantGraphSlowCurve (p h : ℝ → ℝ) (hsmooth : ContDiffAt ℝ 7 (fun ε ↦ (p ε, h ε)) 0) (hbase : (p 0, h 0) = (2, 1)) (htangent : HasFDerivAt (fun ε ↦ (p ε, h ε)) (0 : ℝ →L[ℝ] (ℝ × ℝ)) 0) (hinvariant : (fun ε ↦ DFP.TwoLeg.stateMap (ε, p ε, h ε)) =ᶠ[𝓝 0] (fun ε ↦ let nextScale := (DFP.TwoLeg.stateMap (ε, p ε, h ε)).1 (nextScale, p nextScale, h nextScale))) : ∃ curve : DFP.TwoLeg.SlowCurve, curve.shape = p ∧ curve.high = h := by obtain ⟨hp, hh, _⟩ := invariantGraphTaylorJets p h hsmooth hbase htangent hinvariant refine ⟨DFP.TwoLeg.SlowCurve.ofAsymptotics p h hp hh hinvariant, ?_, ?_⟩ · exact DFP.TwoLeg.SlowCurve.ofAsymptotics_shape p h hp hh hinvariant · exact DFP.TwoLeg.SlowCurve.ofAsymptotics_high p h hp hh hinvariant

Helper for cm-center-proposition: the actual amplitude multiplier on a supplied source slow curve has coefficient and a uniform sixth-order remainder. This is conditional support for cm-scale-and-angle.

theorem slowCurveAmplitudeExpansion (curve : DFP.TwoLeg.SlowCurve) : (fun ε : ℝ ↦ (DFP.TwoLeg.observableMap (ε, curve.shape ε, curve.high ε)).amplitudeRatio - (1 - (13 / 2) * ε ^ 4)) =O[𝓝 0] (fun ε : ℝ ↦ ε ^ 6) := by exact DFP.TwoLeg.slowCurveAmplitudeDrift curve.shape curve.high curve.isInvariant curve.shapeRemainder curve.highRemainder

Helper for cm-center-proposition: the actual two-stage center cancellation has the source coefficient , uniformly BEFORE the initial scale is chosen. This is conditional support for cm-center-cancellation.

theorem slowCurveCenterCancellation (curve : DFP.TwoLeg.SlowCurve) : ∃ εbar ∈ Set.Ioo (0 : ℝ) (1 / 4), ∃ C > 0, ∀ ε₀ ∈ Set.Ioc 0 εbar, ∀ j : ℕ, let orbit := DFP.TwoPhaseOrbit.ofSlowCurve curve.shape curve.high ε₀ let state := orbit.state j ‖(orbit.state (j + 1)).center - state.center + ((116 / 5) * state.amplitude * state.ε ^ 6) • state.lowVector‖ ≤ C * state.amplitude * state.ε ^ 7 := by exact DFP.TwoPhaseOrbit.slowCurveFullCenterDrift curve.shape curve.high curve.isInvariant curve.shapeRemainder curve.highRemainder

Helper for cm-center-proposition: the actual first-stage center change on a supplied source slow curve is uniformly cubic in the cycle scale.

theorem slowCurveHalfCenterBound (curve : DFP.TwoLeg.SlowCurve) : ∃ εbar ∈ Set.Ioo (0 : ℝ) (1 / 4), ∃ C > 0, ∀ ε₀ ∈ Set.Ioc 0 εbar, ∀ j : ℕ, let state := (DFP.TwoPhaseOrbit.ofSlowCurve curve.shape curve.high ε₀).state j ‖state.halfCenterDisplacement‖ ≤ C * state.amplitude * state.ε ^ 3 := by exact DFP.TwoPhaseOrbit.slowCurveHalfCenterDisplacementBound curve.shape curve.high curve.isInvariant curve.shapeRemainder curve.highRemainder

Helper for cm-center-proposition: the actual unwrapped source frame has increment with one constant for all small initial scales.

theorem slowCurveFrameRotation (curve : DFP.TwoLeg.SlowCurve) : ∃ εbar ∈ Set.Ioo (0 : ℝ) (1 / 4), ∃ C > 0, ∀ ε₀ ∈ Set.Ioc 0 εbar, ∀ j : ℕ, let orbit := DFP.TwoPhaseOrbit.ofSlowCurve curve.shape curve.high ε₀ |orbit.frameAngle (j + 1) - orbit.frameAngle j + 3 * (orbit.state j).ε ^ 2| ≤ C * (orbit.state j).ε ^ 4 := by exact DFP.TwoPhaseOrbit.slowCurveFrameRotation curve.shape curve.high curve.isInvariant curve.shapeRemainder curve.highRemainder

Helper for cm-center-proposition: the scale coordinate of the physical source orbit is exactly the scale of the iterated normalized map.

theorem orbitScale_eq_iterate (p h : ℝ → ℝ) (ε₀ : ℝ) (j : ℕ) : ((DFP.TwoPhaseOrbit.ofSlowCurve p h ε₀).state j).ε = (DFP.TwoLeg.stateMap^[j] (ε₀, p ε₀, h ε₀)).1 := by have hcoordinates := DFP.TwoPhaseOrbit.ofSlowCurve_coordinates p h ε₀ j simpa only [DFP.TwoPhaseOrbit.State.coordinates_def] using congrArg Prod.fst hcoordinates

Helper for cm-center-proposition: the source inverse-cubic comparison scale converges to zero.

theorem inverseCubicScale_tendsto_zero : Tendsto (fun j : ℕ ↦ ((9 / 2 : ℝ) * (j : ℝ)) ^ (-(1 : ℝ) / 3)) atTop (𝓝 0) := by have hcoefficient : (0 : ℝ) < 9 / 2 := by norm_num have hexponent : (0 : ℝ) < 1 / 3 := by norm_num have hbase : Tendsto (fun j : ℕ ↦ (9 / 2 : ℝ) * (j : ℝ)) atTop atTop := (tendsto_natCast_atTop_atTop (R := ℝ)).const_mul_atTop hcoefficient convert (tendsto_rpow_neg_atTop hexponent).comp hbase using 1 funext j simp only [Function.comp_apply] congr 1 ring

Helper for cm-center-proposition: every sufficiently small actual orbit of a supplied source slow curve has the source inverse-cubic scale asymptotic.

theorem slowCurveScaleAsymptotic (curve : DFP.TwoLeg.SlowCurve) : ∃ εbar > 0, ∀ ε₀ ∈ Set.Ioc 0 εbar, (fun j : ℕ ↦ ((DFP.TwoPhaseOrbit.ofSlowCurve curve.shape curve.high ε₀).state j).ε) ~[atTop] (fun j : ℕ ↦ ((9 / 2 : ℝ) * (j : ℝ)) ^ (-(1 : ℝ) / 3)) := by obtain ⟨εbar, hεbar, hscale⟩ := DFP.TwoLeg.slowCurveScaleAsymptotic curve.shape curve.high curve.isInvariant curve.shapeRemainder curve.highRemainder refine ⟨εbar, hεbar, ?_⟩ intro ε₀ hε₀ simpa only [orbitScale_eq_iterate] using hscale ε₀ hε₀

Helper for cm-center-proposition: a supplied source slow curve gives positive limiting amplitude, coefficient 13 / 3 : ℕ13/3, convergent centers, and a uniform cubic center tail. The amplitude statement here is asymptotic equivalence; it does not assert the stronger quadratic amplitude remainder in cm-limiting-scales.

theorem slowCurveLimitingScales (curve : DFP.TwoLeg.SlowCurve) : ∃ εbar ∈ Set.Ioo (0 : ℝ) (1 / 4), ∃ K > 0, ∀ ε₀ ∈ Set.Ioc 0 εbar, let orbit := DFP.TwoPhaseOrbit.ofSlowCurve curve.shape curve.high ε₀ ∃ Glim > 0, ∃ Clim : EuclideanSpace ℝ (Fin 2), Tendsto (fun j : ℕ ↦ (orbit.state j).amplitude) atTop (𝓝 Glim) ∧ (fun j : ℕ ↦ (orbit.state j).amplitude - Glim) ~[atTop] (fun j : ℕ ↦ (13 / 3 : ℝ) * Glim * (orbit.state j).ε) ∧ Tendsto (fun j : ℕ ↦ (orbit.state j).center) atTop (𝓝 Clim) ∧ ∀ j : ℕ, ‖(orbit.state j).center - Clim‖ + ‖(orbit.state j).middleCenter - Clim‖ ≤ K * (orbit.state j).ε ^ 3 := by obtain ⟨ηAmplitude, hηAmplitude, hamplitude⟩ := DFP.TwoPhaseOrbit.slowCurveAmplitudeAsymptotics curve.shape curve.high curve.isInvariant curve.shapeRemainder curve.highRemainder obtain ⟨ηCenter, hηCenter, hcenter⟩ := DFP.TwoPhaseOrbit.slowCurveCenterTendsto curve.shape curve.high curve.isInvariant curve.shapeRemainder curve.highRemainder obtain ⟨ηTail, hηTail, K, hK, htail⟩ := DFP.TwoPhaseOrbit.slowCurveCenterTailUniformBound curve.shape curve.high curve.isInvariant curve.shapeRemainder curve.highRemainder let εbar := min ηAmplitude (min ηCenter ηTail) have hεbar : εbar ∈ Set.Ioo (0 : ℝ) (1 / 4) := by constructor · exact lt_min hηAmplitude (lt_min hηCenter.1 hηTail.1) · exact ((min_le_right _ _).trans (min_le_left _ _)).trans_lt hηCenter.2 refine ⟨εbar, hεbar, K, hK, ?_⟩ intro ε₀ hε₀ have hεAmplitude : ε₀ ∈ Set.Ioc 0 ηAmplitude := ⟨hε₀.1, hε₀.2.trans (min_le_left _ _)⟩ have hεCenter : ε₀ ∈ Set.Ioc 0 ηCenter := ⟨hε₀.1, hε₀.2.trans ((min_le_right _ _).trans (min_le_left _ _))⟩ have hεTail : ε₀ ∈ Set.Ioc 0 ηTail := ⟨hε₀.1, hε₀.2.trans ((min_le_right _ _).trans (min_le_right _ _))⟩ obtain ⟨Glim, hGlim, hGlimLimit, hfirstOrder⟩ := hamplitude ε₀ hεAmplitude obtain ⟨Clim, hClimLimit⟩ := hcenter ε₀ hεCenter exact ⟨Glim, hGlim, Clim, hGlimLimit, hfirstOrder, hClimLimit, htail ε₀ hεTail Clim hClimLimit⟩

Helper for cm-center-proposition: the source frame angle is strictly decreasing, diverges, and has vanishing steps; Nat.find first crossings on successive turns approach every target angle. This is conditional support from a supplied SlowCurve.

theorem slowCurveFrameFirstCrossings (curve : DFP.TwoLeg.SlowCurve) : ∃ εbar ∈ Set.Ioo (0 : ℝ) (1 / 4), ∀ ε₀ ∈ Set.Ioc 0 εbar, let orbit := DFP.TwoPhaseOrbit.ofSlowCurve curve.shape curve.high ε₀ StrictAnti orbit.frameAngle ∧ Tendsto orbit.frameAngle atTop atBot ∧ Tendsto (fun j : ℕ ↦ orbit.frameAngle j - orbit.frameAngle (j + 1)) atTop (𝓝 0) ∧ ∀ θ : ℝ, ∃ j : ℕ → ℕ, ∃ m : ℕ → ℤ, StrictMono j ∧ Tendsto (fun i ↦ orbit.frameAngle (j i) + m i * (2 * Real.pi)) atTop (𝓝 θ) := by obtain ⟨ηRotation, hηRotation, C, hC, hrotation⟩ := slowCurveFrameRotation curve obtain ⟨ηDivergence, hηDivergence, hdivergence⟩ := DFP.TwoPhaseOrbit.slowCurveFrameAngleTendstoAtBot curve.shape curve.high curve.isInvariant curve.shapeRemainder curve.highRemainder obtain ⟨ηGraph, hηGraph, hgraph⟩ := DFP.TwoLeg.slowCurveForwardOrbitOnGraph curve.shape curve.high curve.isInvariant curve.shapeRemainder curve.highRemainder obtain ⟨ηScale, hηScale, hscale⟩ := slowCurveScaleAsymptotic curve let εbar := min ηRotation (min ηDivergence (min ηGraph (min ηScale (min 1 (1 / (C + 1)))))) have hden : 0 < C + 1 := by linarith have hεbarPos : 0 < εbar := by exact lt_min hηRotation.1 (lt_min hηDivergence.1 (lt_min hηGraph.1 (lt_min hηScale (lt_min zero_lt_one (one_div_pos.mpr hden))))) have hεbarLt : εbar < 1 / 4 := (min_le_left _ _).trans_lt hηRotation.2 refine ⟨εbar, ⟨hεbarPos, hεbarLt⟩, ?_⟩ intro ε₀ hε₀ dsimp only let orbit := DFP.TwoPhaseOrbit.ofSlowCurve curve.shape curve.high ε₀ have hsmall : ε₀ ≤ ηRotation ∧ ε₀ ≤ ηDivergence ∧ ε₀ ≤ ηGraph ∧ ε₀ ≤ ηScale ∧ ε₀ ≤ 1 ∧ ε₀ ≤ 1 / (C + 1) := by simpa only [εbar, le_min_iff] using hε₀.2 have hεRotation : ε₀ ∈ Set.Ioc 0 ηRotation := ⟨hε₀.1, hsmall.1⟩ have hεDivergence : ε₀ ∈ Set.Ioc 0 ηDivergence := ⟨hε₀.1, hsmall.2.1⟩ have hεGraph : ε₀ ∈ Set.Ioc 0 ηGraph := ⟨hε₀.1, hsmall.2.2.1⟩ have hεScale : ε₀ ∈ Set.Ioc 0 ηScale := ⟨hε₀.1, hsmall.2.2.2.1⟩ have hscaleRange (j : ℕ) : 0 < (orbit.state j).ε ∧ (orbit.state j).ε ≤ ε₀ := by simpa only [orbit, orbitScale_eq_iterate, Set.mem_Ioc] using (hgraph ε₀ hεGraph j).2 have hscaleZero : Tendsto (fun j : ℕ ↦ (orbit.state j).ε) atTop (𝓝 0) := (hscale ε₀ hεScale).symm.tendsto_nhds inverseCubicScale_tendsto_zero have hstrict : StrictAnti orbit.frameAngle := by apply strictAnti_nat_of_succ_lt intro j have hεPos := (hscaleRange j).1 have hεOne : (orbit.state j).ε ≤ 1 := (hscaleRange j).2.trans hsmall.2.2.2.2.1 have hεDen : (orbit.state j).ε ≤ 1 / (C + 1) := (hscaleRange j).2.trans hsmall.2.2.2.2.2 have hmul : (orbit.state j).ε * (C + 1) ≤ 1 := (le_div_iff₀ hden).mp hεDen have hεSquare : (orbit.state j).ε ^ 2 ≤ (orbit.state j).ε := by nlinarith have hquadratic : C * (orbit.state j).ε ^ 2 ≤ 1 := by nlinarith have hquartic : C * (orbit.state j).ε ^ 4 ≤ (orbit.state j).ε ^ 2 := by calc C * (orbit.state j).ε ^ 4 = (C * (orbit.state j).ε ^ 2) * (orbit.state j).ε ^ 2 := by ring _ ≤ 1 * (orbit.state j).ε ^ 2 := mul_le_mul_of_nonneg_right hquadratic (sq_nonneg _) _ = (orbit.state j).ε ^ 2 := one_mul _ have hbound := (abs_le.mp (hrotation ε₀ hεRotation j)).2 have hsquarePos : 0 < (orbit.state j).ε ^ 2 := pow_pos hεPos 2 nlinarith have hbot : Tendsto orbit.frameAngle atTop atBot := hdivergence ε₀ hεDivergence have hupperZero : Tendsto (fun j : ℕ ↦ 3 * (orbit.state j).ε ^ 2 + C * (orbit.state j).ε ^ 4) atTop (𝓝 0) := by have hraw := ((hscaleZero.pow 2).const_mul 3).add ((hscaleZero.pow 4).const_mul C) simpa using hraw have hdropNonneg : ∀ j : ℕ, 0 ≤ orbit.frameAngle j - orbit.frameAngle (j + 1) := by intro j exact sub_nonneg.mpr (hstrict (Nat.lt_succ_self j)).le have hdropUpper : ∀ j : ℕ, orbit.frameAngle j - orbit.frameAngle (j + 1) ≤ 3 * (orbit.state j).ε ^ 2 + C * (orbit.state j).ε ^ 4 := by intro j have hbound : -(C * (orbit.state j).ε ^ 4) ≤ orbit.frameAngle (j + 1) - orbit.frameAngle j + 3 * (orbit.state j).ε ^ 2 := by simpa only [orbit] using (abs_le.mp (hrotation ε₀ hεRotation j)).1 linarith have hstep : Tendsto (fun j : ℕ ↦ orbit.frameAngle j - orbit.frameAngle (j + 1)) atTop (𝓝 0) := squeeze_zero hdropNonneg hdropUpper hupperZero refine ⟨hstrict, hbot, hstep, ?_⟩ intro θ have hperiod : 0 < 2 * Real.pi := mul_pos zero_lt_two Real.pi_pos exact Real.existsSubseqAddIntMulTendsto (2 * Real.pi) hperiod hstrict hbot hstep θ

Helper for cm-center-proposition: the actual centered source map admits canonical C⁸ cutoff data for its zero-start graph transform. The fixed rational rates are chosen before the cutoff and before any initial orbit scale; no fixed-graph regularity theorem is used.

theorem sourceCutoffData : ∃ d : LocalInvariantGraph.MetricGraphTransformData (ℝ × ℝ), d.nu = 8 ∧ d.L = stableDerivative ∧ d.slope = 1 / 2 ∧ d.epsilon = 1 / 1000 ∧ d.lower = 999 / 1000 ∧ d.linearRate = 8 / 9 ∧ HasFDerivAt d.R (0 : (ℝ × (ℝ × ℝ)) →L[ℝ] (ℝ × (ℝ × ℝ))) (0, 0) ∧ centeredStateMap =ᶠ[𝓝 (0, 0)] LocalInvariantGraph.metricCenterStableMap d ∧ ∀ r : ℕ, 1 ≤ r → r ≤ 8 → ((d.linearRate : ℝ) + (d.epsilon : ℝ) * (1 + (d.slope : ℝ))) * (d.lower : ℝ)⁻¹ ^ r < 1 := by let epsilon : NNReal := 1 / 1000 let lower : NNReal := 999 / 1000 let linearRate : NNReal := 8 / 9 let slope : NNReal := 1 / 2 have hν : 2 ≤ 8 := by norm_num have hνone : 1 ≤ 8 := by norm_num have hepsilon : 0 < epsilon := by norm_num [epsilon] have hlower : 0 < lower := by norm_num [lower] have hadd : lower + epsilon = 1 := by norm_num [lower, epsilon] have hslopeOne : slope ≤ 1 := by norm_num [slope] have hepsilonOne : epsilon < 1 := by norm_num [epsilon] have hlinearRate : linearRate < 1 := by norm_num [linearRate] have hslope : (linearRate * slope + epsilon) * lower⁻¹ ≤ slope := by have hreal : (((linearRate * slope + epsilon) * lower⁻¹ : NNReal) : ℝ) ≤ (slope : ℝ) := by norm_num [linearRate, slope, epsilon, lower] exact_mod_cast hreal have hrate : LocalInvariantGraph.metricGraphTransformRate lower linearRate epsilon slope < 1 := by have hreal : (LocalInvariantGraph.metricGraphTransformRate lower linearRate epsilon slope : ℝ) < 1 := by rw [LocalInvariantGraph.metricGraphTransformRate_coe] norm_num [linearRate, slope, epsilon, lower] exact_mod_cast hreal have hL : ‖stableDerivative‖ ≤ (linearRate : ℝ) := by have hcompare : (7 / 9 : ℝ) ≤ (linearRate : ℝ) := by norm_num [linearRate] exact stableDerivative_norm_le.trans hcompare have hNzero := LocalInvariantGraph.centerStable_remainder_zero centeredStateMap stableDerivative centeredStateMap_zero have hNderiv := LocalInvariantGraph.centerStable_remainder_hasFDerivAt_zero centeredStateMap stableDerivative centeredStateMap_hasFDerivAt have hNsmooth := LocalInvariantGraph.centerStable_remainder_contDiffAt 8 centeredStateMap stableDerivative centeredStateMap_analyticAt.contDiffAt obtain ⟨R, hRsmooth, hRsupport, hRzero, hRLip, hRderiv, hRgerm⟩ := LocalInvariantGraph.cutoffRemainderWithZeroDerivative 8 (fun x ↦ centeredStateMap x - LocalCutoff.centerStable stableDerivative x) hνone hNsmooth hNzero hNderiv hepsilon obtain ⟨stableBound, hstableBound⟩ := LocalInvariantGraph.exists_stableCoordinateBound R hRsmooth.continuous hRsupport have hlinearRateReal : (linearRate : ℝ) < 1 := by exact_mod_cast hlinearRate obtain ⟨radius, hradius⟩ := LocalInvariantGraph.metricRadius_bound hlinearRateReal have hcenter : ∀ ξ : SmallLipschitzGraph (ℝ × ℝ) radius slope, Function.Bijective (fun u ↦ u + (R (u, ξ u)).1) := by intro ξ obtain ⟨homeomorph, hhomeomorph⟩ := LocalInvariantGraph.exists_centerProjectionHomeomorph R hRLip ξ hslopeOne hepsilonOne rw [← hhomeomorph] exact homeomorph.bijective have hinverse : ∀ ξ : SmallLipschitzGraph (ℝ × ℝ) radius slope, LipschitzWith lower⁻¹ (Function.invFun (fun u ↦ u + (R (u, ξ u)).1)) := by intro ξ exact LocalInvariantGraph.lipschitzWith_invFun_centerProjection R hRLip ξ hslopeOne lower hlower hadd have hinverseDistance : ∀ (ξ ζ : SmallLipschitzGraph (ℝ × ℝ) radius slope) (v : ℝ), dist (Function.invFun (fun u ↦ u + (R (u, ξ u)).1) v) (Function.invFun (fun u ↦ u + (R (u, ζ u)).1) v) ≤ (lower⁻¹ : ℝ) * (epsilon : ℝ) * dist ξ ζ := by intro ξ ζ v exact LocalInvariantGraph.dist_invFun_centerProjection_le R hRLip ξ ζ hslopeOne lower hlower hadd v let d := LocalInvariantGraph.metricGraphTransformData_of_cutoff 8 R stableDerivative radius slope epsilon lower linearRate stableBound hν hRsmooth hRsupport hRzero hRLip hslopeOne hlower hadd hcenter hinverse hinverseDistance hL hlinearRate hstableBound hradius hslope hrate have hmodel := LocalInvariantGraph.centerStable_remainder_germ centeredStateMap stableDerivative R hRgerm have hdR : d.R = R := by rfl have hdL : d.L = stableDerivative := by rfl have hmodelGerm : centeredStateMap =ᶠ[𝓝 (0, 0)] LocalInvariantGraph.metricCenterStableMap d := by filter_upwards [hmodel.symm] with x hx rw [LocalInvariantGraph.metricCenterStableMap_eq, hdR, hdL] exact hx refine ⟨d, rfl, hdL, rfl, rfl, rfl, rfl, ?_, hmodelGerm, ?_⟩ · rw [hdR] exact hRderiv · intro r hrOne hrEight have hdSlope : d.slope = 1 / 2 := by rfl have hdEpsilon : d.epsilon = 1 / 1000 := by rfl have hdLower : d.lower = 999 / 1000 := by rfl have hdRate : d.linearRate = 8 / 9 := by rfl rw [hdSlope, hdEpsilon, hdLower, hdRate] interval_cases r all_goals norm_num

Helper for cm-center-proposition: a C⁷ fixed graph of the exact source cutoff transform transfers back to a tangent invariant graph of the original stateMap. This support lemma differentiates invariance only AFTER regularity has been supplied by the source uniform-iterate/Arzelà–Ascoli argument.

theorem invariantGraphOfSmoothFixedGraph (d : LocalInvariantGraph.MetricGraphTransformData (ℝ × ℝ)) (hRderiv : HasFDerivAt d.R (0 : (ℝ × (ℝ × ℝ)) →L[ℝ] (ℝ × (ℝ × ℝ))) (0, 0)) (hgerm : centeredStateMap =ᶠ[𝓝 (0, 0)] LocalInvariantGraph.metricCenterStableMap d) (ζ : SmallLipschitzGraph (ℝ × ℝ) d.radius d.slope) (hfixed : d.transform ζ = ζ) (hsmooth : ContDiffAt ℝ 7 (ζ : ℝ → ℝ × ℝ) 0) : ∃ p h : ℝ → ℝ, ContDiffAt ℝ 7 (fun ε ↦ (p ε, h ε)) 0 ∧ (p 0, h 0) = (2, 1) ∧ HasFDerivAt (fun ε ↦ (p ε, h ε)) (0 : ℝ →L[ℝ] (ℝ × ℝ)) 0 ∧ (fun ε ↦ DFP.TwoLeg.stateMap (ε, p ε, h ε)) =ᶠ[𝓝 0] (fun ε ↦ let nextScale := (DFP.TwoLeg.stateMap (ε, p ε, h ε)).1 (nextScale, p nextScale, h nextScale)) := by have hzero : (ζ : ℝ → ℝ × ℝ) 0 = 0 := SmallLipschitzGraph.zero_apply ζ have hseven : (7 : WithTop ℕ∞) ≠ 0 := by norm_num have hderivative : HasFDerivAt (ζ : ℝ → ℝ × ℝ) (fderiv ℝ (ζ : ℝ → ℝ × ℝ) 0) 0 := (hsmooth.differentiableAt hseven).hasFDerivAt have hmodelZero := LocalInvariantGraph.metricCenterStableMap_zero d have hmodelDerivative := LocalInvariantGraph.metricCenterStableMap_hasFDerivAt_of_zeroDerivative d hRderiv have hmodelInvariant := LocalInvariantGraph.metricFixedGraph_invariant d ζ hfixed have hrate : (d.linearRate : ℝ) < 1 := by exact_mod_cast d.hlinearRate have htangent := LocalInvariantGraph.tangent_zero_of_centerStable_invariant (ζ : ℝ → ℝ × ℝ) (fderiv ℝ (ζ : ℝ → ℝ × ℝ) 0) (LocalInvariantGraph.metricCenterStableMap d) d.L hzero hderivative hmodelZero hmodelDerivative hmodelInvariant (d.hL.trans_lt hrate) have htendsto : Tendsto (ζ : ℝ → ℝ × ℝ) (𝓝 0) (𝓝 0) := by simpa only [hzero] using hsmooth.continuousAt.tendsto have hpathTendsto : Tendsto (fun u : ℝ ↦ (u, ζ u)) (𝓝 0) (𝓝 (0, 0)) := by simpa only [nhds_prod_eq, id_eq] using tendsto_id.prodMk htendsto have hsourceModel := hgerm.comp_tendsto hpathTendsto have hsourceInvariant : (fun u ↦ (centeredStateMap (u, ζ u)).2) =ᶠ[𝓝 0] (fun u ↦ ζ (centeredStateMap (u, ζ u)).1) := by filter_upwards [hmodelInvariant, hsourceModel] with u hu hmap dsimp only [Function.comp_apply] at hmap rw [hmap] exact hu let p : ℝ → ℝ := fun u ↦ (ζ u).1 + 2 let h : ℝ → ℝ := fun u ↦ (ζ u).2 + 1 have hgraphEq : (fun u ↦ (p u, h u)) = fun u ↦ ζ u + ((2, 1) : ℝ × ℝ) := by rfl refine ⟨p, h, ?_, ?_, ?_, ?_⟩ · rw [hgraphEq] exact hsmooth.add contDiffAt_const · simp only [p, h, hzero, Prod.fst_zero, Prod.snd_zero, zero_add] · rw [hgraphEq] exact htangent.add_const (2, 1) · filter_upwards [hsourceInvariant] with u hu rw [centeredStateMap_apply] at hu apply Prod.ext · rfl · apply Prod.ext · have hshape := congrArg Prod.fst hu dsimp only [p] dsimp only at hshape linarith · have hhigh := congrArg Prod.snd hu dsimp only [h] dsimp only at hhigh linarith

Helper for cm-center-proposition: Nat.find first-crossing subsequences of a real angle lift approach every unit direction under actual planar rotation.

theorem rotationSubsequenceOfFirstCrossings (φ : ℕ → ℝ) (hcrossings : ∀ θ : ℝ, ∃ j : ℕ → ℕ, ∃ m : ℕ → ℤ, StrictMono j ∧ Tendsto (fun i ↦ φ (j i) + m i * (2 * Real.pi)) atTop (𝓝 θ)) (v : EuclideanSpace ℝ (Fin 2)) (hv : ‖v‖ = 1) : ∃ j : ℕ → ℕ, StrictMono j ∧ Tendsto (fun i ↦ EuclideanPlane.rotation (φ (j i)) (WithLp.toLp 2 ![(1 : ℝ), 0])) atTop (𝓝 v) := by let e : EuclideanSpace ℝ (Fin 2) := WithLp.toLp 2 ![(1 : ℝ), 0] let θ : ℝ := (EuclideanPlane.orientation.oangle e v).toReal have he : ‖e‖ = 1 := by simp [e, EuclideanSpace.norm_eq, Fin.sum_univ_two] have hangle : (θ : Real.Angle) = EuclideanPlane.orientation.oangle e v := Real.Angle.coe_toReal _ have hnorm : ‖e‖ = ‖v‖ := he.trans hv.symm have hrotation : EuclideanPlane.rotation (θ : Real.Angle) e = v := by rw [hangle] exact (EuclideanPlane.orientation.rotation_oangle_eq_iff_norm_eq e v).mpr hnorm obtain ⟨j, m, hj, hlimit⟩ := hcrossings θ have hcontinuous : Continuous (fun t : Real.Angle ↦ EuclideanPlane.rotation t e) := by simp only [EuclideanPlane.rotation_apply] exact (Real.Angle.continuous_cos.smul continuous_const).add (Real.Angle.continuous_sin.smul continuous_const) have hrotationLimit := ((hcontinuous.comp Real.Angle.continuous_coe).tendsto θ).comp hlimit have hperiod (i : ℕ) : ((φ (j i) + m i * (2 * Real.pi) : ℝ) : Real.Angle) = (φ (j i) : Real.Angle) := by rw [Real.Angle.coe_add, Real.Angle.intCast_mul_eq_zsmul] simp only [Real.Angle.coe_two_pi, smul_zero, add_zero] refine ⟨j, hj, ?_⟩ dsimp only [Function.comp_apply] at hrotationLimit rw [hrotation] at hrotationLimit apply hrotationLimit.congr' filter_upwards [] with i exact congrArg (fun angle : Real.Angle ↦ EuclideanPlane.rotation angle e) (hperiod i)

Helper for cm-center-proposition: a sequence approaching the source rotating circle has EXACTLY that circle as its cluster set when its angle lift has the source first-crossing subsequences. No compactness existence argument is used.

theorem clusterSetOfRotatingCircle (x : ℕ → EuclideanSpace ℝ (Fin 2)) (φ : ℕ → ℝ) (C : EuclideanSpace ℝ (Fin 2)) (G : ℝ) (hG : 0 < G) (hcrossings : ∀ θ : ℝ, ∃ j : ℕ → ℕ, ∃ m : ℕ → ℤ, StrictMono j ∧ Tendsto (fun i ↦ φ (j i) + m i * (2 * Real.pi)) atTop (𝓝 θ)) (herror : Tendsto (fun n ↦ ‖x n - (C + G • EuclideanPlane.rotation (φ n) (WithLp.toLp 2 ![(1 : ℝ), 0]))‖) atTop (𝓝 0)) : {z : EuclideanSpace ℝ (Fin 2) | MapClusterPt z atTop x} = DFP.TwoPhaseOrbit.limitCircle C G := by let model : ℕ → EuclideanSpace ℝ (Fin 2) := fun n ↦ C + G • EuclideanPlane.rotation (φ n) (WithLp.toLp 2 ![(1 : ℝ), 0]) have hnormModel (n : ℕ) : ‖model n - C‖ = G := by simp [model, norm_smul, Real.norm_eq_abs, abs_of_pos hG, EuclideanSpace.norm_eq, Fin.sum_univ_two] have hradialError (n : ℕ) : |‖x n - C‖ - G| ≤ ‖x n - model n‖ := by rw [← hnormModel n] have hnormalized : (x n - C) - (model n - C) = x n - model n := by abel simpa only [hnormalized] using abs_norm_sub_norm_le (x n - C) (model n - C) have hradialZero : Tendsto (fun n ↦ ‖x n - C‖ - G) atTop (𝓝 0) := by apply squeeze_zero_norm · intro n simpa only [Real.norm_eq_abs] using hradialError n · exact herror have hradial : Tendsto (fun n ↦ ‖x n - C‖) atTop (𝓝 G) := by simpa only [sub_add_cancel, zero_add] using hradialZero.add_const G have herrorVector : Tendsto (fun n ↦ x n - model n) atTop (𝓝 0) := tendsto_zero_iff_norm_tendsto_zero.mpr herror ext z constructor · intro hcluster obtain ⟨j, hj, hjlimit⟩ := hcluster.tendsto_subseq have hnormLimit := (hjlimit.sub_const C).norm have hnorm : ‖z - C‖ = G := tendsto_nhds_unique hnormLimit (hradial.comp hj.tendsto_atTop) rw [DFP.TwoPhaseOrbit.limitCircle_eq_sphere C G hG, Metric.mem_sphere, dist_eq_norm] exact hnorm · intro hcircle obtain ⟨v, hv, rfl⟩ := DFP.TwoPhaseOrbit.mem_limitCircle.mp hcircle obtain ⟨j, hj, hrotation⟩ := rotationSubsequenceOfFirstCrossings φ hcrossings v hv have hmodel : Tendsto (fun i ↦ model (j i)) atTop (𝓝 (C + G • v)) := by exact tendsto_const_nhds.add (tendsto_const_nhds.smul hrotation) have hpoint : Tendsto (fun i ↦ x (j i)) atTop (𝓝 (C + G • v)) := by have hsum := (herrorVector.comp hj.tendsto_atTop).add hmodel simpa only [Function.comp_apply, sub_add_cancel, zero_add] using hsum exact hpoint.mapClusterPt.of_comp hj.tendsto_atTop

Helper for cm-center-proposition: the physical low direction is the first column of the actual eigenframe, equivalently its action on (1, 0) : ℕ × ℕ(1,0).

theorem lowVector_eq_frameBasis (s : DFP.TwoPhaseOrbit.State) : s.lowVector = WithLp.toLp 2 (s.frame *ᵥ ![(1 : ℝ), 0]) := by ext i rw [DFP.TwoPhaseOrbit.State.lowVector_apply] simp [Matrix.mulVec, dotProduct, Fin.sum_univ_two]

Helper for cm-center-proposition: an oriented physical eigenframe has a unit low direction.

theorem lowVector_norm (s : DFP.TwoPhaseOrbit.State) (hframe : s.frame ∈ Matrix.specialOrthogonalGroup (Fin 2) ℝ) : ‖s.lowVector‖ = 1 := by rw [lowVector_eq_frameBasis, Matrix.norm_toLp_mulVec_eq_of_mem_specialOrthogonalGroup _ hframe] simp [EuclideanSpace.norm_eq, Fin.sum_univ_two]

Helper for cm-center-proposition: the actual gradient differs from its low-direction component by exactly the high coordinate in norm.

theorem gradientLowVector_error (s : DFP.TwoPhaseOrbit.State) (hframe : s.frame ∈ Matrix.specialOrthogonalGroup (Fin 2) ℝ) : ‖s.gradient - s.amplitude • s.lowVector‖ = |s.amplitude| * |s.p * s.ε ^ 2| := by have hcoordinates : (![(1 : ℝ), s.p * s.ε ^ 2] : Fin 2 → ℝ) - ![(1 : ℝ), 0] = ![(0 : ℝ), s.p * s.ε ^ 2] := by ext i fin_cases i · simp · simp rw [DFP.TwoPhaseOrbit.State.gradient_def, lowVector_eq_frameBasis, ← smul_sub, ← WithLp.toLp_sub, ← Matrix.mulVec_sub, hcoordinates, norm_smul, Real.norm_eq_abs, Matrix.norm_toLp_mulVec_eq_of_mem_specialOrthogonalGroup _ hframe] simp [EuclideanSpace.norm_eq, Fin.sum_univ_two, Real.sqrt_sq_eq_abs]

Helper for cm-center-proposition: physical source states lie on the supplied slow curve and retain positive decreasing cycle scales.

theorem slowCurveStateOnGraph (curve : DFP.TwoLeg.SlowCurve) : ∃ εbar ∈ Set.Ioo (0 : ℝ) (1 / 4), ∀ ε₀ ∈ Set.Ioc 0 εbar, ∀ j : ℕ, let state := (DFP.TwoPhaseOrbit.ofSlowCurve curve.shape curve.high ε₀).state j (state.p, state.h) = (curve.shape state.ε, curve.high state.ε) ∧ 0 < state.ε ∧ state.ε ≤ ε₀ := by obtain ⟨εbar, hεbar, hgraph⟩ := DFP.TwoLeg.slowCurveForwardOrbitOnGraph curve.shape curve.high curve.isInvariant curve.shapeRemainder curve.highRemainder refine ⟨εbar, hεbar, ?_⟩ intro ε₀ hε₀ j have hcoordinates := DFP.TwoPhaseOrbit.ofSlowCurve_coordinates curve.shape curve.high ε₀ j rw [DFP.TwoPhaseOrbit.State.coordinates_def] at hcoordinates have hstateGraph := (hgraph ε₀ hε₀ j).1 rw [← hcoordinates] at hstateGraph have hstateRange := (hgraph ε₀ hε₀ j).2 rw [← hcoordinates] at hstateRange exact ⟨congrArg Prod.snd hstateGraph, hstateRange⟩

Helper for cm-center-proposition: the exact spectral representation approaches its rotating low-direction circle whenever the source scale, amplitude, shape and center have their derived limits.

theorem pointLowVectorApproximation (orbit : DFP.TwoPhaseOrbit) (C : EuclideanSpace ℝ (Fin 2)) (G : ℝ) (hframe : ∀ j, (orbit.state j).frame ∈ Matrix.specialOrthogonalGroup (Fin 2) ℝ) (hscale : Tendsto (fun j ↦ (orbit.state j).ε) atTop (𝓝 0)) (hshape : Tendsto (fun j ↦ (orbit.state j).p) atTop (𝓝 2)) (hamplitude : Tendsto (fun j ↦ (orbit.state j).amplitude) atTop (𝓝 G)) (hcenter : Tendsto (fun j ↦ (orbit.state j).center) atTop (𝓝 C)) : Tendsto (fun j ↦ ‖(orbit.state j).point - (C + G • (orbit.state j).lowVector)‖) atTop (𝓝 0) := by have hcenterNorm : Tendsto (fun j ↦ ‖(orbit.state j).center - C‖) atTop (𝓝 0) := by simpa using (hcenter.sub_const C).norm have hhighNorm : Tendsto (fun j ↦ |(orbit.state j).amplitude| * |(orbit.state j).p * (orbit.state j).ε ^ 2|) atTop (𝓝 0) := by have hraw := (hamplitude.mul (hshape.mul (hscale.pow 2))).abs simpa only [zero_pow (Nat.succ_ne_zero 1), mul_zero, abs_zero, abs_mul] using hraw have hamplitudeNorm : Tendsto (fun j ↦ |(orbit.state j).amplitude - G|) atTop (𝓝 0) := by simpa using (hamplitude.sub_const G).abs have hupper : Tendsto (fun j ↦ ‖(orbit.state j).center - C‖ + |(orbit.state j).amplitude| * |(orbit.state j).p * (orbit.state j).ε ^ 2| + |(orbit.state j).amplitude - G|) atTop (𝓝 0) := by simpa using (hcenterNorm.add hhighNorm).add hamplitudeNorm have hbound (j : ℕ) : ‖(orbit.state j).point - (C + G • (orbit.state j).lowVector)‖ ≤ ‖(orbit.state j).center - C‖ + |(orbit.state j).amplitude| * |(orbit.state j).p * (orbit.state j).ε ^ 2| + |(orbit.state j).amplitude - G| := by let state := orbit.state j have hdecomposition : state.point - (C + G • state.lowVector) = (state.center - C) + (state.gradient - state.amplitude • state.lowVector) + (state.amplitude - G) • state.lowVector := by rw [DFP.TwoPhaseOrbit.State.center_def] module rw [hdecomposition] calc ‖(state.center - C) + (state.gradient - state.amplitude • state.lowVector) + (state.amplitude - G) • state.lowVector‖ ≤ ‖state.center - C‖ + ‖state.gradient - state.amplitude • state.lowVector‖ + ‖(state.amplitude - G) • state.lowVector‖ := (norm_add_le _ _).trans (add_le_add (norm_add_le _ _) le_rfl) _ = ‖state.center - C‖ + |state.amplitude| * |state.p * state.ε ^ 2| + |state.amplitude - G| := by rw [gradientLowVector_error state (hframe j), norm_smul, Real.norm_eq_abs, lowVector_norm state (hframe j), mul_one] exact squeeze_zero (fun _ ↦ norm_nonneg _) hbound hupper

Helper for cm-center-proposition: the actual even and odd endpoints have the full source circle as cluster set when the derived spectral errors vanish and the SOURCE eigenframe has first-crossing subsequences.

theorem endpointCircleFromSourceFrame (orbit : DFP.TwoPhaseOrbit) (φ : ℕ → ℝ) (C : EuclideanSpace ℝ (Fin 2)) (G : ℝ) (hG : 0 < G) (hframe : ∀ j, (orbit.state j).frame ∈ Matrix.specialOrthogonalGroup (Fin 2) ℝ) (hscale : Tendsto (fun j ↦ (orbit.state j).ε) atTop (𝓝 0)) (hshape : Tendsto (fun j ↦ (orbit.state j).p) atTop (𝓝 2)) (hamplitude : Tendsto (fun j ↦ (orbit.state j).amplitude) atTop (𝓝 G)) (hcenter : Tendsto (fun j ↦ (orbit.state j).center) atTop (𝓝 C)) (hsteps : Tendsto (fun j ↦ ‖(orbit.state j).firstDisplacement‖) atTop (𝓝 0)) (hrepresentation : ∀ j, (orbit.state j).lowVector = EuclideanPlane.rotation (φ j) (WithLp.toLp 2 ![(1 : ℝ), 0])) (hcrossings : ∀ θ : ℝ, ∃ j : ℕ → ℕ, ∃ m : ℕ → ℤ, StrictMono j ∧ Tendsto (fun i ↦ φ (j i) + m i * (2 * Real.pi)) atTop (𝓝 θ)) : {z : EuclideanSpace ℝ (Fin 2) | MapClusterPt z atTop orbit.endpoint} = DFP.TwoPhaseOrbit.limitCircle C G := by let model : ℕ → EuclideanSpace ℝ (Fin 2) := fun j ↦ C + G • EuclideanPlane.rotation (φ j) (WithLp.toLp 2 ![(1 : ℝ), 0]) have hevenError : Tendsto (fun j ↦ ‖(orbit.state j).point - model j‖) atTop (𝓝 0) := by have hraw := pointLowVectorApproximation orbit C G hframe hscale hshape hamplitude hcenter apply hraw.congr' filter_upwards [] with j rw [hrepresentation j] have hoddBound (j : ℕ) : ‖orbit.endpoint (2 * j + 1) - model j‖ ≤ ‖(orbit.state j).firstDisplacement‖ + ‖(orbit.state j).point - model j‖ := by rw [DFP.TwoPhaseOrbit.endpoint_odd, DFP.TwoPhaseOrbit.State.middlePoint_def] have hnormalize : (orbit.state j).point + (orbit.state j).firstDisplacement - model j = (orbit.state j).firstDisplacement + ((orbit.state j).point - model j) := by abel rw [hnormalize] exact norm_add_le _ _ have hoddError : Tendsto (fun j ↦ ‖orbit.endpoint (2 * j + 1) - model j‖) atTop (𝓝 0) := by have hupper : Tendsto (fun j ↦ ‖(orbit.state j).firstDisplacement‖ + ‖(orbit.state j).point - model j‖) atTop (𝓝 0) := by simpa using hsteps.add hevenError exact squeeze_zero (fun _ ↦ norm_nonneg _) hoddBound hupper let liftedAngle : ℕ → ℝ := fun k ↦ φ (k / 2) have hfullError : Tendsto (fun k ↦ ‖orbit.endpoint k - (C + G • EuclideanPlane.rotation (liftedAngle k) (WithLp.toLp 2 ![(1 : ℝ), 0]))‖) atTop (𝓝 0) := by apply tendsto_of_tendsto_even_odd · simpa [liftedAngle, model, DFP.TwoPhaseOrbit.endpoint_even] using hevenError · have hdiv (j : ℕ) : (2 * j + 1) / 2 = j := by omega simpa [liftedAngle, model, hdiv] using hoddError have hfullCrossings : ∀ θ : ℝ, ∃ j : ℕ → ℕ, ∃ m : ℕ → ℤ, StrictMono j ∧ Tendsto (fun i ↦ liftedAngle (j i) + m i * (2 * Real.pi)) atTop (𝓝 θ) := by intro θ obtain ⟨j, m, hj, hlimit⟩ := hcrossings θ refine ⟨fun i ↦ 2 * j i, m, ?_, ?_⟩ · intro i k hik exact Nat.mul_lt_mul_of_pos_left (hj hik) Nat.two_pos · simpa [liftedAngle] using hlimit exact clusterSetOfRotatingCircle orbit.endpoint liftedAngle C G hG hfullCrossings hfullError

Helper for cm-center-proposition: every sufficiently small trajectory of a supplied SOURCE slow curve has exactly its limiting circle as the cluster set of ALL actual even and odd endpoints. The decisive subsequences are the source frame's Nat.find first crossings, and the odd steps vanish.

theorem slowCurveCircleGeometry (curve : DFP.TwoLeg.SlowCurve) : ∃ εbar ∈ Set.Ioo (0 : ℝ) (1 / 4), ∀ ε₀ ∈ Set.Ioc 0 εbar, let orbit := DFP.TwoPhaseOrbit.ofSlowCurve curve.shape curve.high ε₀ ∃ Glim > 0, ∃ Clim : EuclideanSpace ℝ (Fin 2), Tendsto (fun j : ℕ ↦ (orbit.state j).amplitude) atTop (𝓝 Glim) ∧ Tendsto (fun j : ℕ ↦ (orbit.state j).center) atTop (𝓝 Clim) ∧ Tendsto orbit.frameAngle atTop atBot ∧ {z : EuclideanSpace ℝ (Fin 2) | MapClusterPt z atTop orbit.endpoint} = DFP.TwoPhaseOrbit.limitCircle Clim Glim := by obtain ⟨ηCrossing, hηCrossing, hcrossing⟩ := slowCurveFrameFirstCrossings curve obtain ⟨ηGraph, hηGraph, hgraph⟩ := slowCurveStateOnGraph curve obtain ⟨ηScale, hηScale, hscale⟩ := slowCurveScaleAsymptotic curve have heighth : (1 / 8 : ℝ) ∈ Set.Ioo 0 (1 / 4) := by norm_num obtain ⟨ηValid, hηValid, hvalid⟩ := DFP.TwoPhaseOrbit.ofSlowCurve_phaseValidity curve.shape curve.high curve.isInvariant curve.shapeRemainder curve.highRemainder (1 / 8) heighth obtain ⟨ηFrame, hηFrame, hframe⟩ := DFP.TwoPhaseOrbit.slowCurveFrameAngleRepresentsLowVector curve.shape curve.high curve.isInvariant curve.shapeRemainder curve.highRemainder (1 / 8) heighth obtain ⟨ηStep, hηStep, hstep⟩ := DFP.TwoPhaseOrbit.slowCurveFirstStepNormRemainder curve.shape curve.high curve.isInvariant curve.shapeRemainder curve.highRemainder obtain ⟨ηLimits, hηLimits, K, hK, hlimits⟩ := slowCurveLimitingScales curve let εbar := min ηCrossing (min ηGraph (min ηScale (min ηValid (min ηFrame (min ηStep ηLimits))))) have hεbarPos : 0 < εbar := by exact lt_min hηCrossing.1 (lt_min hηGraph.1 (lt_min hηScale (lt_min hηValid.1 (lt_min hηFrame.1 (lt_min hηStep hηLimits.1))))) have hεbarLt : εbar < 1 / 4 := (min_le_left _ _).trans_lt hηCrossing.2 refine ⟨εbar, ⟨hεbarPos, hεbarLt⟩, ?_⟩ intro ε₀ hε₀ dsimp only let orbit := DFP.TwoPhaseOrbit.ofSlowCurve curve.shape curve.high ε₀ have hsmall : ε₀ ≤ ηCrossing ∧ ε₀ ≤ ηGraph ∧ ε₀ ≤ ηScale ∧ ε₀ ≤ ηValid ∧ ε₀ ≤ ηFrame ∧ ε₀ ≤ ηStep ∧ ε₀ ≤ ηLimits := by simpa only [εbar, le_min_iff] using hε₀.2 have hεCrossing : ε₀ ∈ Set.Ioc 0 ηCrossing := ⟨hε₀.1, hsmall.1⟩ have hεGraph : ε₀ ∈ Set.Ioc 0 ηGraph := ⟨hε₀.1, hsmall.2.1⟩ have hεScale : ε₀ ∈ Set.Ioc 0 ηScale := ⟨hε₀.1, hsmall.2.2.1⟩ have hεValid : ε₀ ∈ Set.Ioc 0 ηValid := ⟨hε₀.1, hsmall.2.2.2.1⟩ have hεFrame : ε₀ ∈ Set.Ioc 0 ηFrame := ⟨hε₀.1, hsmall.2.2.2.2.1⟩ have hεStep : ε₀ ∈ Set.Ioc 0 ηStep := ⟨hε₀.1, hsmall.2.2.2.2.2.1⟩ have hεLimits : ε₀ ∈ Set.Ioc 0 ηLimits := ⟨hε₀.1, hsmall.2.2.2.2.2.2⟩ obtain ⟨Glim, hGlim, Clim, hamplitude, _, hcenter, _⟩ := hlimits ε₀ hεLimits have hscaleZero : Tendsto (fun j : ℕ ↦ (orbit.state j).ε) atTop (𝓝 0) := (hscale ε₀ hεScale).symm.tendsto_nhds inverseCubicScale_tendsto_zero have hshapeZero : Tendsto curve.shape (𝓝 0) (𝓝 2) := by have hproject : Continuous (fun x : ℝ × ℝ × ℝ ↦ x.2.1) := by fun_prop have hlimit := (hproject.tendsto ((0, 2, 1) : ℝ × ℝ × ℝ)).comp curve.toSlowGraph.path_tendsto simpa only [Function.comp_def, DFP.TwoLeg.SlowGraph.path_apply] using hlimit have hshape : Tendsto (fun j : ℕ ↦ (orbit.state j).p) atTop (𝓝 2) := by apply (hshapeZero.comp hscaleZero).congr' filter_upwards [] with j exact (congrArg Prod.fst (hgraph ε₀ hεGraph j).1).symm have hsquareZero : Tendsto (fun j : ℕ ↦ (orbit.state j).ε ^ 2) atTop (𝓝 0) := by simpa using hscaleZero.pow 2 have hstepRemainderZero : Tendsto (fun j : ℕ ↦ ‖(orbit.state j).firstDisplacement‖ / (orbit.state j).amplitude - 2 * (orbit.state j).ε ^ 2) atTop (𝓝 0) := (hstep ε₀ hεStep).isBigO.trans_tendsto hsquareZero have hquotientZero : Tendsto (fun j : ℕ ↦ ‖(orbit.state j).firstDisplacement‖ / (orbit.state j).amplitude) atTop (𝓝 0) := by have hsum := hstepRemainderZero.add (hsquareZero.const_mul 2) simpa only [sub_add_cancel, mul_zero, zero_add] using hsum have hstepZero : Tendsto (fun j : ℕ ↦ ‖(orbit.state j).firstDisplacement‖) atTop (𝓝 0) := by have hnormalize (j : ℕ) : ‖(orbit.state j).firstDisplacement‖ / (orbit.state j).amplitude * (orbit.state j).amplitude = ‖(orbit.state j).firstDisplacement‖ := div_mul_cancel₀ _ (ne_of_gt (hvalid ε₀ hεValid j).amplitude_pos) have hproduct : Tendsto (fun j : ℕ ↦ ‖(orbit.state j).firstDisplacement‖ / (orbit.state j).amplitude * (orbit.state j).amplitude) atTop (𝓝 (0 * Glim)) := hquotientZero.mul hamplitude simpa only [hnormalize, zero_mul] using hproduct have hinitial : (orbit.state 0).lowVector = WithLp.toLp 2 ![(1 : ℝ), 0] := by rw [DFP.TwoPhaseOrbit.ofSlowCurve_zero, lowVector_eq_frameBasis, DFP.TwoPhaseOrbit.State.initial_frame] simp have hrepresentation (j : ℕ) : (orbit.state j).lowVector = EuclideanPlane.rotation (orbit.frameAngle j) (WithLp.toLp 2 ![(1 : ℝ), 0]) := by have hraw : (orbit.state j).lowVector = EuclideanPlane.rotation (orbit.frameAngle j) (orbit.state 0).lowVector := hframe ε₀ hεFrame j rw [hinitial] at hraw exact hraw have hcircle := endpointCircleFromSourceFrame orbit orbit.frameAngle Clim Glim hGlim (fun j ↦ (hvalid ε₀ hεValid j).frame_specialOrthogonal) hscaleZero hshape hamplitude hcenter hstepZero hrepresentation (hcrossing ε₀ hεCrossing).2.2.2 exact ⟨Glim, hGlim, Clim, hamplitude, hcenter, (hcrossing ε₀ hεCrossing).2.1, hcircle⟩

Helper for cm-center-proposition: compensating the actual amplitude by cancels the leading quartic terms in the source scale and ratio recurrences.

theorem compensatedAmplitudeIdentity (a b x y : ℝ) (ha : a ≠ 0) : b * (1 - (13 / 3) * y) - a * (1 - (13 / 3) * x) = a * ((b / a - 1 + (13 / 2) * x ^ 4) * (1 - (13 / 3) * y) - (13 / 3) * (y - x + (3 / 2) * x ^ 4) + (169 / 6) * x ^ 4 * y) := by field_simp [ha] ring

Helper for cm-center-proposition: uniform fifth-order source ratio and scale residuals give a fifth-order increment bound for the compensated amplitude.

theorem compensatedAmplitudeIncrement (a b x y M KG Kε : ℝ) (ha : 0 < a) (haM : a ≤ M) (hx : 0 < x) (hy : 0 ≤ y) (hyx : y ≤ x) (hsmall : (13 / 3) * x ≤ 1 / 2) (hKG : 0 ≤ KG) (hKε : 0 ≤ Kε) (hratio : |b / a - 1 + (13 / 2) * x ^ 4| ≤ KG * x ^ 5) (hscale : |y - x + (3 / 2) * x ^ 4| ≤ Kε * x ^ 5) : |b * (1 - (13 / 3) * y) - a * (1 - (13 / 3) * x)| ≤ (M * (KG + (13 / 3) * Kε + 169 / 6)) * x ^ 5 := by have hthirteen : (0 : ℝ) ≤ 13 / 3 := by norm_num have hleading : (0 : ℝ) ≤ 169 / 6 := by norm_num have hfactor : 0 ≤ 1 - (13 / 3) * y := by nlinarith have hfactorOne : 1 - (13 / 3) * y ≤ 1 := by nlinarith have hfourth : 0 ≤ x ^ 4 := pow_nonneg hx.le 4 have hfifth : 0 ≤ x ^ 5 := pow_nonneg hx.le 5 have hratioTerm : |(b / a - 1 + (13 / 2) * x ^ 4) * (1 - (13 / 3) * y)| ≤ KG * x ^ 5 := by rw [abs_mul, abs_of_nonneg hfactor] exact (mul_le_mul_of_nonneg_right hratio hfactor).trans (mul_le_of_le_one_right (mul_nonneg hKG hfifth) hfactorOne) have hscaleTerm : |(13 / 3 : ℝ) * (y - x + (3 / 2) * x ^ 4)| ≤ (13 / 3) * Kε * x ^ 5 := by rw [abs_mul] norm_num calc (13 / 3 : ℝ) * |y - x + (3 / 2) * x ^ 4| ≤ (13 / 3) * (Kε * x ^ 5) := mul_le_mul_of_nonneg_left hscale hthirteen _ = _ := by ring have hproductTerm : |(169 / 6 : ℝ) * x ^ 4 * y| ≤ (169 / 6) * x ^ 5 := by rw [abs_of_nonneg] · calc (169 / 6 : ℝ) * x ^ 4 * y ≤ (169 / 6) * x ^ 4 * x := mul_le_mul_of_nonneg_left hyx (mul_nonneg hleading hfourth) _ = (169 / 6) * x ^ 5 := by ring · positivity rw [compensatedAmplitudeIdentity a b x y (ne_of_gt ha), abs_mul, abs_of_pos ha] have hinside : |(b / a - 1 + (13 / 2) * x ^ 4) * (1 - (13 / 3) * y) - (13 / 3) * (y - x + (3 / 2) * x ^ 4) + (169 / 6) * x ^ 4 * y| ≤ (KG + (13 / 3) * Kε + 169 / 6) * x ^ 5 := by calc _ ≤ |(b / a - 1 + (13 / 2) * x ^ 4) * (1 - (13 / 3) * y) - (13 / 3) * (y - x + (3 / 2) * x ^ 4)| + |(169 / 6 : ℝ) * x ^ 4 * y| := abs_add_le _ _ _ ≤ (|(b / a - 1 + (13 / 2) * x ^ 4) * (1 - (13 / 3) * y)| + |(13 / 3 : ℝ) * (y - x + (3 / 2) * x ^ 4)|) + |(169 / 6 : ℝ) * x ^ 4 * y| := add_le_add (abs_sub _ _) le_rfl _ ≤ (KG * x ^ 5 + (13 / 3) * Kε * x ^ 5) + (169 / 6) * x ^ 5 := add_le_add (add_le_add hratioTerm hscaleTerm) hproductTerm _ = _ := by ring have hboundNonneg : 0 ≤ (KG + (13 / 3) * Kε + 169 / 6) * x ^ 5 := by positivity calc a * |(b / a - 1 + (13 / 2) * x ^ 4) * (1 - (13 / 3) * y) - (13 / 3) * (y - x + (3 / 2) * x ^ 4) + (169 / 6) * x ^ 4 * y| ≤ a * ((KG + (13 / 3) * Kε + 169 / 6) * x ^ 5) := mul_le_mul_of_nonneg_left hinside ha.le _ ≤ M * ((KG + (13 / 3) * Kε + 169 / 6) * x ^ 5) := mul_le_mul_of_nonneg_right haM hboundNonneg _ = _ := by ring

Helper for cm-center-proposition: summing the source compensated increments gives the full quadratic amplitude remainder with coefficient 13 / 3 : ℕ13/3. The hypotheses are scalar recurrence residuals and uniform bounds, not the desired tail estimate.

theorem amplitudeQuadraticTail (a ε : ℕ → ℝ) (A M KG Kε : ℝ) (ha : ∀ n, 0 < a n) (haM : ∀ n, a n ≤ M) (hε : ∀ n, 0 < ε n) (hsmall : ∀ n, (13 / 3) * ε n ≤ 1 / 2) (hdecrement : ∀ n, ε n ^ 4 ≤ ε n - ε (n + 1)) (hKG : 0 ≤ KG) (hKε : 0 ≤ Kε) (hratio : ∀ n, |a (n + 1) / a n - 1 + (13 / 2) * ε n ^ 4| ≤ KG * ε n ^ 5) (hscale : ∀ n, |ε (n + 1) - ε n + (3 / 2) * ε n ^ 4| ≤ Kε * ε n ^ 5) (hlimit : Tendsto a atTop (𝓝 A)) (hεzero : Tendsto ε atTop (𝓝 0)) (hAnonneg : 0 ≤ A) (hAM : A ≤ M) (j : ℕ) : |a j - A - (13 / 3) * A * ε j| ≤ 2 * (M * (KG + (13 / 3) * Kε + 169 / 6) + (169 / 9) * M) * ε j ^ 2 := by let b : ℕ → ℝ := fun n ↦ a n * (1 - (13 / 3) * ε n) let L : ℝ := M * (KG + (13 / 3) * Kε + 169 / 6) have hM : 0 ≤ M := (ha 0).le.trans (haM 0) have hL : 0 ≤ L := by dsimp only [L] positivity have hfourFive : 3 + 1 ≤ 5 := by norm_num have htail : ∀ n : ℕ, Summable (fun k : ℕ ↦ ε (n + k) ^ 5) ∧ (∑' k : ℕ, ε (n + k) ^ 5) ≤ 1 * ε n ^ 2 := by intro n have hpositive : ∀ (_ : Unit) k, 0 < ε k := fun _ k ↦ hε k have hdecrease : ∀ (_ : Unit) k, (1 : ℝ) * ε k ^ (3 + 1) ≤ ε k - ε (k + 1) := by intro _ k simpa using hdecrement k have hraw := ParabolicRecurrence.summable_tail_pow_and_tsum_le_of_decrement (ε := fun _ : Unit ↦ ε) (p := 3) (q := 5) zero_lt_one hfourFive hpositive hdecrease () n simpa using hraw have hbound : ∀ n, ‖b (n + 1) - b n‖ ≤ L * ε n ^ 5 := by intro n have hyx : ε (n + 1) ≤ ε n := by have hnonneg : 0 ≤ ε n ^ 4 := pow_nonneg (hε n).le 4 linarith [hdecrement n] simpa only [Real.norm_eq_abs, b, L] using compensatedAmplitudeIncrement (a n) (a (n + 1)) (ε n) (ε (n + 1)) M KG Kε (ha n) (haM n) (hε n) (hε (n + 1)).le hyx (hsmall n) hKG hKε (hratio n) (hscale n) have hsumPower : Summable (fun n ↦ ε n ^ 5) := by simpa using (htail 0).1 have hsumNorm : Summable (fun n ↦ ‖b (n + 1) - b n‖) := (hsumPower.mul_left L).of_nonneg_of_le (fun _ ↦ norm_nonneg _) hbound have hbLimit : Tendsto b atTop (𝓝 A) := by have hconstant : Tendsto (fun _ : ℕ ↦ (1 : ℝ)) atTop (𝓝 1) := tendsto_const_nhds have hfactor : Tendsto (fun n ↦ 1 - (13 / 3) * ε n) atTop (𝓝 1) := by simpa using hconstant.sub (hεzero.const_mul (13 / 3 : ℝ)) simpa only [b, mul_one] using hlimit.mul hfactor have hbError : |b j - A| ≤ L * ε j ^ 2 := by simpa only [Real.norm_eq_abs, mul_one] using SequenceTail.norm_sub_limit_le hL hsumNorm hbLimit hbound htail j have hidentity : (1 - (13 / 3) * ε j) * (a j - A - (13 / 3) * A * ε j) = b j - A + (169 / 9) * A * ε j ^ 2 := by dsimp only [b] ring have hfactorHalf : 1 / 2 ≤ 1 - (13 / 3) * ε j := by linarith [hsmall j] have hfactorNonneg : 0 ≤ 1 - (13 / 3) * ε j := by linarith have hcorrection : |(169 / 9 : ℝ) * A * ε j ^ 2| ≤ (169 / 9) * M * ε j ^ 2 := by rw [abs_of_nonneg] · have hcoefficient : (0 : ℝ) ≤ 169 / 9 := by norm_num exact mul_le_mul_of_nonneg_right (mul_le_mul_of_nonneg_left hAM hcoefficient) (sq_nonneg _) · positivity have hproductBound : (1 - (13 / 3) * ε j) * |a j - A - (13 / 3) * A * ε j| ≤ (L + (169 / 9) * M) * ε j ^ 2 := by calc _ = |(1 - (13 / 3) * ε j) * (a j - A - (13 / 3) * A * ε j)| := by rw [abs_mul, abs_of_nonneg hfactorNonneg] _ = |b j - A + (169 / 9) * A * ε j ^ 2| := congrArg abs hidentity _ ≤ |b j - A| + |(169 / 9 : ℝ) * A * ε j ^ 2| := abs_add_le _ _ _ ≤ L * ε j ^ 2 + (169 / 9) * M * ε j ^ 2 := add_le_add hbError hcorrection _ = _ := by ring have hhalf : (1 / 2) * |a j - A - (13 / 3) * A * ε j| ≤ (1 - (13 / 3) * ε j) * |a j - A - (13 / 3) * A * ε j| := mul_le_mul_of_nonneg_right hfactorHalf (abs_nonneg _) dsimp only [L] at hproductBound linarith

Helper for cm-center-proposition: the ORIGINAL two-step map has a C⁷ invariant tangent graph, constructed by uniform C⁸ ZERO-START graph-transform iterates and Arzelà–Ascoli. No earlier fixed-section regularity theorem is used.

theorem existsInvariantCenterGraph : ∃ p h : ℝ → ℝ, ContDiffAt ℝ 7 (fun ε ↦ (p ε, h ε)) 0 ∧ (p 0, h 0) = (2, 1) ∧ HasFDerivAt (fun ε ↦ (p ε, h ε)) (0 : ℝ →L[ℝ] (ℝ × ℝ)) 0 ∧ (fun ε ↦ DFP.TwoLeg.stateMap (ε, p ε, h ε)) =ᶠ[𝓝 0] (fun ε ↦ let nextScale := (DFP.TwoLeg.stateMap (ε, p ε, h ε)).1 (nextScale, p nextScale, h nextScale)) := by obtain ⟨d, hν, _, _, _, _, _, hR, hgerm, hbunching⟩ := sourceCutoffData have hνEight : 8 ≤ d.nu := by rw [hν] have honeTwo : 1 ≤ 2 := by norm_num have htop : ∀ r : ℕ, 2 ≤ r → r ≤ 8 → ((d.linearRate : ℝ) + (d.epsilon : ℝ) * (1 + (d.slope : ℝ))) * (d.lower : ℝ)⁻¹ ^ r < 1 := by intro r hrTwo hrEight exact hbunching r (honeTwo.trans hrTwo) hrEight let ζ := LocalInvariantGraph.GraphIterateBounds.graphLimit d have hfixed : d.transform ζ = ζ := LocalInvariantGraph.GraphIterateBounds.graphLimit_fixed d have hsmooth : ContDiffAt ℝ 7 (ζ : ℝ → ℝ × ℝ) 0 := (LocalInvariantGraph.GraphIterateBounds.graphLimit_contDiff_seven d hνEight htop).contDiffAt exact invariantGraphOfSmoothFixedGraph d hR hgerm ζ hfixed hsmooth

cm-center-proposition. The source two-step map has eigenvalues and a locally invariant C⁷ graph with coefficients 198 / 5 : ℕ198/5, 8 : ℕ8, and . For EVERY sufficiently small positive initial scale, the LITERAL source data and generate the infinite exact alternating secant recurrence. Both physical matrices in every cycle are positive definite, the spectral coordinates remain on the graph, and .

Unknown identifier `State.ExactCycle`State.ExactCycle records the actual displacement, gradient and inverse DFP updates in the current eigenbasis; Unknown identifier `State.metric_def`State.metric_def and Unknown identifier `gradient_def`gradient_def encode . The graph is obtained from the SOURCE highest-derivative estimates and Arzelà–Ascoli, independently of the polar counterexample construction.

theorem centerProposition : Module.End.HasEigenvalue (fderiv ℝ DFP.TwoLeg.stateMap (0, 2, 1)).toLinearMap (1 : ℝ) ∧ Module.End.HasEigenvalue (fderiv ℝ DFP.TwoLeg.stateMap (0, 2, 1)).toLinearMap (-(1 : ℝ) / 9) ∧ Module.End.HasEigenvalue (fderiv ℝ DFP.TwoLeg.stateMap (0, 2, 1)).toLinearMap (0 : ℝ) ∧ ∃ p h : ℝ → ℝ, ContDiffAt ℝ 7 (fun ε ↦ (p ε, h ε)) 0 ∧ (p 0, h 0) = (2, 1) ∧ HasFDerivAt (fun ε ↦ (p ε, h ε)) (0 : ℝ →L[ℝ] (ℝ × ℝ)) 0 ∧ (fun ε ↦ DFP.TwoLeg.stateMap (ε, p ε, h ε)) =ᶠ[𝓝 0] (fun ε ↦ let nextScale := (DFP.TwoLeg.stateMap (ε, p ε, h ε)).1 (nextScale, p nextScale, h nextScale)) ∧ (fun ε : ℝ ↦ p ε - (2 + (198 / 5) * ε ^ 3)) =O[𝓝 0] (fun ε : ℝ ↦ ε ^ 4) ∧ (fun ε : ℝ ↦ h ε - (1 + 8 * ε ^ 3)) =O[𝓝 0] (fun ε : ℝ ↦ ε ^ 4) ∧ (fun ε : ℝ ↦ (DFP.TwoLeg.stateMap (ε, p ε, h ε)).1 - (ε - (3 / 2) * ε ^ 4)) =O[𝓝 0] (fun ε : ℝ ↦ ε ^ 5) ∧ ∃ εbar ∈ Set.Ioo (0 : ℝ) (1 / 4), ∀ ε₀ ∈ Set.Ioc 0 εbar, let orbit := DFP.TwoPhaseOrbit.ofSlowCurve p h ε₀ ((orbit.state 0).point = WithLp.toLp 2 ![(1 : ℝ), p ε₀ * ε₀ ^ 2] ∧ (orbit.state 0).gradient = WithLp.toLp 2 ![(1 : ℝ), p ε₀ * ε₀ ^ 2] ∧ (orbit.state 0).metric = Matrix.diagonal ![h ε₀ * p ε₀ * ε₀ ^ 4, h ε₀]) ∧ (∀ j : ℕ, (orbit.state j).ExactCycle) ∧ (∀ j : ℕ, (orbit.state j).metric.PosDef ∧ (orbit.state j).middleMetric.PosDef ∧ (orbit.state j).next.metric.PosDef) ∧ (∀ j : ℕ, ((orbit.state j).p, (orbit.state j).h) = (p (orbit.state j).ε, h (orbit.state j).ε) ∧ 0 < (orbit.state j).ε) ∧ (fun j : ℕ ↦ (orbit.state j).ε) ~[atTop] (fun j : ℕ ↦ ((9 / 2 : ℝ) * (j : ℝ)) ^ (-(1 : ℝ) / 3)) := by refine ⟨DFP.TwoLeg.stateMap_centerEigenvalue, DFP.TwoLeg.stateMap_transverseEigenvalue_negNinth, DFP.TwoLeg.stateMap_transverseEigenvalue_zero, ?_⟩ obtain ⟨p, h, hsmooth, hbase, htangent, hinvariant⟩ := existsInvariantCenterGraph obtain ⟨hpJet, hhJet, _⟩ := invariantGraphTaylorJets p h hsmooth hbase htangent hinvariant obtain ⟨hpSource, hhSource, hscaleSource⟩ := invariantGraphSourceExpansions p h hsmooth hbase htangent hinvariant have heighth : (1 / 8 : ℝ) ∈ Set.Ioo 0 (1 / 4) := by norm_num obtain ⟨ηValid, hηValid, hexact⟩ := DFP.TwoPhaseOrbit.ofSlowCurveExact p h hinvariant hpJet hhJet (1 / 8) heighth obtain ⟨ηGraph, hηGraph, hgraph⟩ := DFP.TwoLeg.slowCurveForwardOrbitOnGraph p h hinvariant hpJet hhJet obtain ⟨ηScale, hηScale, hscale⟩ := DFP.TwoLeg.slowCurveScaleAsymptotic p h hinvariant hpJet hhJet let εbar := min ηValid (min ηGraph ηScale) have hεbar : εbar ∈ Set.Ioo (0 : ℝ) (1 / 4) := by constructor · exact lt_min hηValid.1 (lt_min hηGraph.1 hηScale) · exact ((min_le_left _ _).trans hηValid.2).trans_lt heighth.2 refine ⟨p, h, hsmooth, hbase, htangent, hinvariant, hpSource, hhSource, hscaleSource, εbar, hεbar, ?_⟩ intro ε₀ hε₀ dsimp only let orbit := DFP.TwoPhaseOrbit.ofSlowCurve p h ε₀ have hεValid : ε₀ ∈ Set.Ioc 0 ηValid := ⟨hε₀.1, hε₀.2.trans (min_le_left _ _)⟩ have hεGraph : ε₀ ∈ Set.Ioc 0 ηGraph := ⟨hε₀.1, hε₀.2.trans ((min_le_right _ _).trans (min_le_left _ _))⟩ have hεScale : ε₀ ∈ Set.Ioc 0 ηScale := ⟨hε₀.1, hε₀.2.trans ((min_le_right _ _).trans (min_le_right _ _))⟩ have hinitial : (orbit.state 0).point = WithLp.toLp 2 ![(1 : ℝ), p ε₀ * ε₀ ^ 2] ∧ (orbit.state 0).gradient = WithLp.toLp 2 ![(1 : ℝ), p ε₀ * ε₀ ^ 2] ∧ (orbit.state 0).metric = Matrix.diagonal ![h ε₀ * p ε₀ * ε₀ ^ 4, h ε₀] := by simpa only [orbit, DFP.TwoPhaseOrbit.ofSlowCurve_zero] using prescribedInitialization p h ε₀ have hcycles (j : ℕ) : (orbit.state j).ExactCycle := hexact ε₀ hεValid j have hmatrices (j : ℕ) : (orbit.state j).metric.PosDef ∧ (orbit.state j).middleMetric.PosDef ∧ (orbit.state j).next.metric.PosDef := physicalMatrices_posDef (orbit.state j) (hcycles j).valid have hcoordinates (j : ℕ) : ((orbit.state j).p, (orbit.state j).h) = (p (orbit.state j).ε, h (orbit.state j).ε) ∧ 0 < (orbit.state j).ε := by have hraw := DFP.TwoPhaseOrbit.ofSlowCurve_coordinates p h ε₀ j rw [DFP.TwoPhaseOrbit.State.coordinates_def] at hraw have hshape := (hgraph ε₀ hεGraph j).1 rw [← hraw] at hshape exact ⟨congrArg Prod.snd hshape, (hcycles j).valid.ε_pos⟩ have hscaleActual : (fun j : ℕ ↦ (orbit.state j).ε) ~[atTop] (fun j : ℕ ↦ ((9 / 2 : ℝ) * (j : ℝ)) ^ (-(1 : ℝ) / 3)) := by simpa only [orbit, orbitScale_eq_iterate] using hscale ε₀ hεScale exact ⟨hinitial, hcycles, hmatrices, hcoordinates, hscaleActual⟩

Helper for cm-center-proposition: the actual physical scale successor is the first coordinate of the original stateMap at the current spectral state.

theorem orbitScale_succ_eq_stateMap (p h : ℝ → ℝ) (ε₀ : ℝ) (j : ℕ) : ((DFP.TwoPhaseOrbit.ofSlowCurve p h ε₀).state (j + 1)).ε = (DFP.TwoLeg.stateMap (((DFP.TwoPhaseOrbit.ofSlowCurve p h ε₀).state j).ε, ((DFP.TwoPhaseOrbit.ofSlowCurve p h ε₀).state j).p, ((DFP.TwoPhaseOrbit.ofSlowCurve p h ε₀).state j).h)).1 := by rw [DFP.TwoPhaseOrbit.ofSlowCurve_succ] have hraw := congrArg Prod.fst (DFP.TwoPhaseOrbit.State.next_coordinates ((DFP.TwoPhaseOrbit.ofSlowCurve p h ε₀).state j)) simpa only [DFP.TwoPhaseOrbit.State.coordinates_def] using hraw

Helper for cm-center-proposition: the actual source amplitudes have the FULL uniform quadratic remainder in cm-limiting-scales. Constants are selected before every small initial scale.

theorem slowCurveAmplitudeQuadraticTail (curve : DFP.TwoLeg.SlowCurve) : ∃ εbar ∈ Set.Ioo (0 : ℝ) (1 / 4), ∃ K > 0, ∀ ε₀ ∈ Set.Ioc 0 εbar, let orbit := DFP.TwoPhaseOrbit.ofSlowCurve curve.shape curve.high ε₀ ∃ Glim > 0, Tendsto (fun j : ℕ ↦ (orbit.state j).amplitude) atTop (𝓝 Glim) ∧ ∀ j : ℕ, |(orbit.state j).amplitude - Glim - (13 / 3) * Glim * (orbit.state j).ε| ≤ K * (orbit.state j).ε ^ 2 := by have hfiveSix : 5 < 6 := by norm_num have hsixFive : (fun ε : ℝ ↦ ε ^ 6) =O[𝓝 0] (fun ε : ℝ ↦ ε ^ 5) := (Asymptotics.isLittleO_pow_pow hfiveSix).isBigO have hscaleSix : (fun ε : ℝ ↦ (DFP.TwoLeg.stateMap (ε, curve.shape ε, curve.high ε)).1 - (ε - (3 / 2) * ε ^ 4 + (5 / 4) * ε ^ 5)) =O[𝓝 0] (fun ε : ℝ ↦ ε ^ 6) := by simpa only [DFP.TwoLeg.stateMap_fst] using DFP.TwoLeg.slowGraphSignedRecurrence curve.shape curve.high curve.shapeRemainder curve.highRemainder have hcorrection : (fun ε : ℝ ↦ (5 / 4) * ε ^ 5) =O[𝓝 0] (fun ε : ℝ ↦ ε ^ 5) := (Asymptotics.isBigO_refl (fun ε : ℝ ↦ ε ^ 5) (𝓝 0)).const_mul_left (5 / 4) have hscaleFive : (fun ε : ℝ ↦ (DFP.TwoLeg.stateMap (ε, curve.shape ε, curve.high ε)).1 - ε + (3 / 2) * ε ^ 4) =O[𝓝 0] (fun ε : ℝ ↦ ε ^ 5) := by apply ((hscaleSix.trans hsixFive).add hcorrection).congr' · filter_upwards [] with ε ring · rfl have hratioFive : (fun ε : ℝ ↦ (DFP.TwoLeg.observableMap (ε, curve.shape ε, curve.high ε)).amplitudeRatio - 1 + (13 / 2) * ε ^ 4) =O[𝓝 0] (fun ε : ℝ ↦ ε ^ 5) := by apply ((slowCurveAmplitudeExpansion curve).trans hsixFive).congr' · filter_upwards [] with ε ring · rfl obtain ⟨Cε, hCε, δε, hδε, hscaleLocal⟩ := Asymptotics.IsUniformRemainderOn.exists_pos_natPow_bound_of_isBigO hscaleFive obtain ⟨CG, hCG, δG, hδG, hratioLocal⟩ := Asymptotics.IsUniformRemainderOn.exists_pos_natPow_bound_of_isBigO hratioFive obtain ⟨ηGraph, hηGraph, hgraph⟩ := slowCurveStateOnGraph curve obtain ⟨ηScale, hηScale, hscale⟩ := slowCurveScaleAsymptotic curve obtain ⟨ηAmplitude, hηAmplitude, Gmin, hGmin, M, hGminM, hamplitude⟩ := DFP.TwoPhaseOrbit.slowCurveAmplitudeUniformBounds curve.shape curve.high curve.isInvariant curve.shapeRemainder curve.highRemainder let εbar := min ηAmplitude (min ηGraph (min ηScale (min (δε / 2) (min (δG / 2) (min (3 / 26) (1 / (2 * (Cε + 1)))))))) have hden : 0 < 2 * (Cε + 1) := by positivity have hcap : (0 : ℝ) < 3 / 26 := by norm_num have hεbarPos : 0 < εbar := by exact lt_min hηAmplitude.1 (lt_min hηGraph.1 (lt_min hηScale (lt_min (half_pos hδε) (lt_min (half_pos hδG) (lt_min hcap (one_div_pos.mpr hden)))))) have hεbarLt : εbar < 1 / 4 := (min_le_left _ _).trans_lt hηAmplitude.2 let K : ℝ := 2 * (M * (CG + (13 / 3) * Cε + 169 / 6) + (169 / 9) * M) have hM : 0 < M := hGmin.trans_le hGminM have hK : 0 < K := by dsimp only [K] positivity refine ⟨εbar, ⟨hεbarPos, hεbarLt⟩, K, hK, ?_⟩ intro ε₀ hε₀ dsimp only let orbit := DFP.TwoPhaseOrbit.ofSlowCurve curve.shape curve.high ε₀ let a : ℕ → ℝ := fun j ↦ (orbit.state j).amplitude let ε : ℕ → ℝ := fun j ↦ (orbit.state j).ε have hsmall : ε₀ ≤ ηAmplitude ∧ ε₀ ≤ ηGraph ∧ ε₀ ≤ ηScale ∧ ε₀ ≤ δε / 2 ∧ ε₀ ≤ δG / 2 ∧ ε₀ ≤ 3 / 26 ∧ ε₀ ≤ 1 / (2 * (Cε + 1)) := by simpa only [εbar, le_min_iff] using hε₀.2 have hεAmplitude : ε₀ ∈ Set.Ioc 0 ηAmplitude := ⟨hε₀.1, hsmall.1⟩ have hεGraph : ε₀ ∈ Set.Ioc 0 ηGraph := ⟨hε₀.1, hsmall.2.1⟩ have hεScale : ε₀ ∈ Set.Ioc 0 ηScale := ⟨hε₀.1, hsmall.2.2.1⟩ obtain ⟨Glim, hGlimRange, hlimit, hbounds⟩ := hamplitude ε₀ hεAmplitude have hGlim : 0 < Glim := hGmin.trans_le hGlimRange.1 have ha (j : ℕ) : 0 < a j := hGmin.trans_le (hbounds j).1 have haM (j : ℕ) : a j ≤ M := (hbounds j).2 have hεPositive (j : ℕ) : 0 < ε j := (hgraph ε₀ hεGraph j).2.1 have hεBound (j : ℕ) : ε j ≤ ε₀ := (hgraph ε₀ hεGraph j).2.2 have hεZero : Tendsto ε atTop (𝓝 0) := (hscale ε₀ hεScale).symm.tendsto_nhds inverseCubicScale_tendsto_zero have hcoordinates (j : ℕ) : (orbit.state j).coordinates = (ε j, curve.shape (ε j), curve.high (ε j)) := by rw [DFP.TwoPhaseOrbit.State.coordinates_def] exact congrArg (fun z : ℝ × ℝ ↦ (ε j, z)) (hgraph ε₀ hεGraph j).1 have hnextScale (j : ℕ) : ε (j + 1) = (DFP.TwoLeg.stateMap (ε j, curve.shape (ε j), curve.high (ε j))).1 := by have hraw : ε (j + 1) = (DFP.TwoLeg.stateMap ((orbit.state j).ε, (orbit.state j).p, (orbit.state j).h)).1 := orbitScale_succ_eq_stateMap curve.shape curve.high ε₀ j rw [← DFP.TwoPhaseOrbit.State.coordinates_def, hcoordinates] at hraw exact hraw have hnextAmplitude (j : ℕ) : a (j + 1) / a j = (DFP.TwoLeg.observableMap (ε j, curve.shape (ε j), curve.high (ε j))).amplitudeRatio := by have hsuccessor : orbit.state (j + 1) = (orbit.state j).next := DFP.TwoPhaseOrbit.ofSlowCurve_succ curve.shape curve.high ε₀ j dsimp only [a] rw [hsuccessor, DFP.TwoPhaseOrbit.State.nextAmplitudeRatio _ (ne_of_gt (ha j)), hcoordinates] have hscaleBound (j : ℕ) : |ε (j + 1) - ε j + (3 / 2) * ε j ^ 4| ≤ Cε * ε j ^ 5 := by have hdist : |ε j| < δε := by rw [abs_of_pos (hεPositive j)] exact ((hεBound j).trans hsmall.2.2.2.1).trans_lt (half_lt_self hδε) have hraw := hscaleLocal (ε j) hdist rw [← hnextScale j] at hraw simpa only [Real.norm_eq_abs, abs_of_pos (hεPositive j)] using hraw have hratioBound (j : ℕ) : |a (j + 1) / a j - 1 + (13 / 2) * ε j ^ 4| ≤ CG * ε j ^ 5 := by have hdist : |ε j| < δG := by rw [abs_of_pos (hεPositive j)] exact ((hεBound j).trans hsmall.2.2.2.2.1).trans_lt (half_lt_self hδG) have hraw := hratioLocal (ε j) hdist rw [← hnextAmplitude j] at hraw simpa only [Real.norm_eq_abs, abs_of_pos (hεPositive j)] using hraw have hsmallAmplitude (j : ℕ) : (13 / 3) * ε j ≤ 1 / 2 := by have hbound := (hεBound j).trans hsmall.2.2.2.2.2.1 linarith have hdecrement (j : ℕ) : ε j ^ 4 ≤ ε j - ε (j + 1) := by have hcapj := (hεBound j).trans hsmall.2.2.2.2.2.2 have hmul : ε j * (2 * (Cε + 1)) ≤ 1 := (le_div_iff₀ hden).mp hcapj have hlocalSmall : Cε * ε j ≤ 1 / 2 := by nlinarith [hεPositive j] have hresidual : Cε * ε j ^ 5 ≤ (1 / 2) * ε j ^ 4 := by calc Cε * ε j ^ 5 = (Cε * ε j) * ε j ^ 4 := by ring _ ≤ (1 / 2) * ε j ^ 4 := mul_le_mul_of_nonneg_right hlocalSmall (pow_nonneg (hεPositive j).le 4) have hbound := (abs_le.mp (hscaleBound j)).2 linarith refine ⟨Glim, hGlim, hlimit, ?_⟩ intro j exact amplitudeQuadraticTail a ε Glim M CG Cε ha haM hεPositive hsmallAmplitude hdecrement hCG.le hCε.le hratioBound hscaleBound hlimit hεZero hGlim.le hGlimRange.2 j

Helper for cm-center-proposition: the exact source spectral representation forces the boundary gradient norm to have the SAME positive limit as its amplitude.

theorem boundaryGradientNormLimit (orbit : DFP.TwoPhaseOrbit) (G : ℝ) (hG : 0 < G) (hframe : ∀ j, (orbit.state j).frame ∈ Matrix.specialOrthogonalGroup (Fin 2) ℝ) (hscale : Tendsto (fun j ↦ (orbit.state j).ε) atTop (𝓝 0)) (hshape : Tendsto (fun j ↦ (orbit.state j).p) atTop (𝓝 2)) (hamplitude : Tendsto (fun j ↦ (orbit.state j).amplitude) atTop (𝓝 G)) : Tendsto (fun j ↦ ‖(orbit.state j).gradient‖) atTop (𝓝 G) := by have hhighZero : Tendsto (fun j ↦ |(orbit.state j).amplitude| * |(orbit.state j).p * (orbit.state j).ε ^ 2|) atTop (𝓝 0) := by have hraw := (hamplitude.mul (hshape.mul (hscale.pow 2))).abs simpa only [zero_pow (Nat.succ_ne_zero 1), mul_zero, abs_zero, abs_mul] using hraw have hbound (j : ℕ) : |‖(orbit.state j).gradient‖ - (|(orbit.state j).amplitude|)| ≤ |(orbit.state j).amplitude| * |(orbit.state j).p * (orbit.state j).ε ^ 2| := by have hnorm : ‖(orbit.state j).amplitude • (orbit.state j).lowVector‖ = |(orbit.state j).amplitude| := by rw [norm_smul, Real.norm_eq_abs, lowVector_norm _ (hframe j), mul_one] calc |‖(orbit.state j).gradient‖ - (|(orbit.state j).amplitude|)| = |‖(orbit.state j).gradient‖ - ‖(orbit.state j).amplitude • (orbit.state j).lowVector‖| := congrArg (fun z : ℝ ↦ |‖(orbit.state j).gradient‖ - z|) hnorm.symm _ ≤ ‖(orbit.state j).gradient - (orbit.state j).amplitude • (orbit.state j).lowVector‖ := abs_norm_sub_norm_le _ _ _ = _ := gradientLowVector_error _ (hframe j) have herrorZero : Tendsto (fun j ↦ ‖(orbit.state j).gradient‖ - |(orbit.state j).amplitude|) atTop (𝓝 0) := by apply squeeze_zero_norm · intro j simpa only [Real.norm_eq_abs] using hbound j · exact hhighZero have hsum := herrorZero.add hamplitude.abs simpa only [sub_add_cancel, zero_add, abs_of_pos hG] using hsum

Helper for cm-center-proposition: actual prescribed gradient norms along BOTH phases converge to the same positive source amplitude limit.

theorem slowCurveFullGradientLimit (curve : DFP.TwoLeg.SlowCurve) : ∃ εbar ∈ Set.Ioo (0 : ℝ) (1 / 4), ∀ ε₀ ∈ Set.Ioc 0 εbar, let orbit := DFP.TwoPhaseOrbit.ofSlowCurve curve.shape curve.high ε₀ ∀ G > 0, Tendsto (fun j ↦ (orbit.state j).amplitude) atTop (𝓝 G) → Tendsto (fun k ↦ ‖orbit.endpointGradient k‖) atTop (𝓝 G) := by obtain ⟨ηGradient, hηGradient, hgradient⟩ := DFP.TwoPhaseOrbit.slowCurveEndpointGradientNormTendsto curve.shape curve.high curve.isInvariant curve.shapeRemainder curve.highRemainder obtain ⟨ηGraph, hηGraph, hgraph⟩ := slowCurveStateOnGraph curve obtain ⟨ηScale, hηScale, hscale⟩ := slowCurveScaleAsymptotic curve have heighth : (1 / 8 : ℝ) ∈ Set.Ioo 0 (1 / 4) := by norm_num obtain ⟨ηValid, hηValid, hvalid⟩ := DFP.TwoPhaseOrbit.ofSlowCurve_phaseValidity curve.shape curve.high curve.isInvariant curve.shapeRemainder curve.highRemainder (1 / 8) heighth let εbar := min ηGradient (min ηGraph (min ηScale ηValid)) have hεbarPos : 0 < εbar := lt_min hηGradient (lt_min hηGraph.1 (lt_min hηScale hηValid.1)) have hεbarLt : εbar < 1 / 4 := ((min_le_right _ _).trans (min_le_left _ _)).trans_lt hηGraph.2 refine ⟨εbar, ⟨hεbarPos, hεbarLt⟩, ?_⟩ intro ε₀ hε₀ dsimp only let orbit := DFP.TwoPhaseOrbit.ofSlowCurve curve.shape curve.high ε₀ have hsmall : ε₀ ≤ ηGradient ∧ ε₀ ≤ ηGraph ∧ ε₀ ≤ ηScale ∧ ε₀ ≤ ηValid := by simpa only [εbar, le_min_iff] using hε₀.2 have hεGradient : ε₀ ∈ Set.Ioc 0 ηGradient := ⟨hε₀.1, hsmall.1⟩ have hεGraph : ε₀ ∈ Set.Ioc 0 ηGraph := ⟨hε₀.1, hsmall.2.1⟩ have hεScale : ε₀ ∈ Set.Ioc 0 ηScale := ⟨hε₀.1, hsmall.2.2.1⟩ have hεValid : ε₀ ∈ Set.Ioc 0 ηValid := ⟨hε₀.1, hsmall.2.2.2⟩ intro G hG hamplitude have hscaleZero : Tendsto (fun j ↦ (orbit.state j).ε) atTop (𝓝 0) := (hscale ε₀ hεScale).symm.tendsto_nhds inverseCubicScale_tendsto_zero have hshapeZero : Tendsto curve.shape (𝓝 0) (𝓝 2) := by have hproject : Continuous (fun x : ℝ × ℝ × ℝ ↦ x.2.1) := by fun_prop have hlimit := (hproject.tendsto ((0, 2, 1) : ℝ × ℝ × ℝ)).comp curve.toSlowGraph.path_tendsto simpa only [Function.comp_def, DFP.TwoLeg.SlowGraph.path_apply] using hlimit have hshape : Tendsto (fun j ↦ (orbit.state j).p) atTop (𝓝 2) := by apply (hshapeZero.comp hscaleZero).congr' filter_upwards [] with j exact (congrArg Prod.fst (hgraph ε₀ hεGraph j).1).symm have hboundary := boundaryGradientNormLimit orbit G hG (fun j ↦ (hvalid ε₀ hεValid j).frame_specialOrthogonal) hscaleZero hshape hamplitude obtain ⟨G', _, hfull⟩ := hgradient ε₀ hεGradient have hevenMono : StrictMono (fun j : ℕ ↦ 2 * j) := by intro i j hij exact Nat.mul_lt_mul_of_pos_left hij Nat.two_pos have hfullEven : Tendsto (fun j ↦ ‖(orbit.state j).gradient‖) atTop (𝓝 G') := by simpa only [Function.comp_def, DFP.TwoPhaseOrbit.endpointGradient_even, orbit] using hfull.comp hevenMono.tendsto_atTop have hlimitEq : G' = G := tendsto_nhds_unique hfullEven hboundary rwa [hlimitEq] at hfull

Helper for cm-center-proposition: a source SlowCurve with C⁷ regularity, the prescribed base, and zero tangent exists by the ACTUAL source graph transform.

theorem existsSourceSlowCurve : ∃ curve : DFP.TwoLeg.SlowCurve, ContDiffAt ℝ 7 (fun ε ↦ (curve.shape ε, curve.high ε)) 0 ∧ (curve.shape 0, curve.high 0) = (2, 1) ∧ HasFDerivAt (fun ε ↦ (curve.shape ε, curve.high ε)) (0 : ℝ →L[ℝ] (ℝ × ℝ)) 0 := by obtain ⟨p, h, hsmooth, hbase, htangent, hinvariant⟩ := existsInvariantCenterGraph obtain ⟨curve, hshape, hhigh⟩ := invariantGraphSlowCurve p h hsmooth hbase htangent hinvariant refine ⟨curve, ?_, ?_, ?_⟩ · simpa only [hshape, hhigh] using hsmooth · simpa only [hshape, hhigh] using hbase · simpa only [hshape, hhigh] using htangent

Helper for cm-center-proposition: the ACTUAL Section 6 recurrence has positive limiting gradient norm, the FULL 13 / 3 : ℕ13/3 quadratic amplitude expansion, uniform cubic center tails, unbounded source frame rotation, and the EXACT limiting circle for both endpoint phases. The same C⁷ source graph is fixed before the common threshold and every later initial scale.

theorem centerManifoldAsymptotics : ∃ curve : DFP.TwoLeg.SlowCurve, ContDiffAt ℝ 7 (fun ε ↦ (curve.shape ε, curve.high ε)) 0 ∧ (curve.shape 0, curve.high 0) = (2, 1) ∧ ∃ εbar ∈ Set.Ioo (0 : ℝ) (1 / 4), ∃ CG > 0, ∃ CC > 0, ∀ ε₀ ∈ Set.Ioc 0 εbar, let orbit := DFP.TwoPhaseOrbit.ofSlowCurve curve.shape curve.high ε₀ ∃ Glim > 0, ∃ Clim : EuclideanSpace ℝ (Fin 2), Tendsto (fun j : ℕ ↦ (orbit.state j).amplitude) atTop (𝓝 Glim) ∧ Tendsto (fun k : ℕ ↦ ‖orbit.endpointGradient k‖) atTop (𝓝 Glim) ∧ Tendsto (fun j : ℕ ↦ (orbit.state j).center) atTop (𝓝 Clim) ∧ (∀ j : ℕ, |(orbit.state j).amplitude - Glim - (13 / 3) * Glim * (orbit.state j).ε| ≤ CG * (orbit.state j).ε ^ 2) ∧ (∀ j : ℕ, ‖(orbit.state j).center - Clim‖ + ‖(orbit.state j).middleCenter - Clim‖ ≤ CC * (orbit.state j).ε ^ 3) ∧ Tendsto orbit.frameAngle atTop atBot ∧ {z : EuclideanSpace ℝ (Fin 2) | MapClusterPt z atTop orbit.endpoint} = DFP.TwoPhaseOrbit.limitCircle Clim Glim := by obtain ⟨curve, hsmooth, hbase, _⟩ := existsSourceSlowCurve obtain ⟨ηCircle, hηCircle, hcircle⟩ := slowCurveCircleGeometry curve obtain ⟨ηQuadratic, hηQuadratic, CG, hCG, hquadratic⟩ := slowCurveAmplitudeQuadraticTail curve obtain ⟨ηCenter, hηCenter, CC, hCC, hcenterTail⟩ := DFP.TwoPhaseOrbit.slowCurveCenterTailUniformBound curve.shape curve.high curve.isInvariant curve.shapeRemainder curve.highRemainder obtain ⟨ηGradient, hηGradient, hgradient⟩ := slowCurveFullGradientLimit curve let εbar := min ηCircle (min ηQuadratic (min ηCenter ηGradient)) have hεbarPos : 0 < εbar := lt_min hηCircle.1 (lt_min hηQuadratic.1 (lt_min hηCenter.1 hηGradient.1)) have hεbarLt : εbar < 1 / 4 := (min_le_left _ _).trans_lt hηCircle.2 refine ⟨curve, hsmooth, hbase, εbar, ⟨hεbarPos, hεbarLt⟩, CG, hCG, CC, hCC, ?_⟩ intro ε₀ hε₀ dsimp only have hsmall : ε₀ ≤ ηCircle ∧ ε₀ ≤ ηQuadratic ∧ ε₀ ≤ ηCenter ∧ ε₀ ≤ ηGradient := by simpa only [εbar, le_min_iff] using hε₀.2 have hεCircle : ε₀ ∈ Set.Ioc 0 ηCircle := ⟨hε₀.1, hsmall.1⟩ have hεQuadratic : ε₀ ∈ Set.Ioc 0 ηQuadratic := ⟨hε₀.1, hsmall.2.1⟩ have hεCenter : ε₀ ∈ Set.Ioc 0 ηCenter := ⟨hε₀.1, hsmall.2.2.1⟩ have hεGradient : ε₀ ∈ Set.Ioc 0 ηGradient := ⟨hε₀.1, hsmall.2.2.2⟩ obtain ⟨Glim, hGlim, Clim, hamplitude, hcenter, hrotation, hcluster⟩ := hcircle ε₀ hεCircle obtain ⟨G', _, hlimit', htail⟩ := hquadratic ε₀ hεQuadratic have hlimitEq : G' = Glim := tendsto_nhds_unique hlimit' hamplitude rw [hlimitEq] at htail exact ⟨Glim, hGlim, Clim, hamplitude, hgradient ε₀ hεGradient Glim hGlim hamplitude, hcenter, htail, hcenterTail ε₀ hεCenter Clim hcenter, hrotation, hcluster⟩

Helper for cm-center-proposition: actual hatted points have radial excess comparable to the source cycle scale, and their prescribed gradient corrections are uniformly cubic. All constants precede the initial scale. No support-radius cap or separation is assumed here; the manuscript's eventual radius and disjointness assumptions can therefore be supplied separately to the interpolation bridge.

theorem slowCurveRadialCoefficientBounds (curve : DFP.TwoLeg.SlowCurve) : ∃ εbar ∈ Set.Ioo (0 : ℝ) (1 / 4), ∃ cR > 0, ∃ CR > 0, ∃ Ca > 0, ∀ ε₀ ∈ Set.Ioc 0 εbar, let orbit := DFP.TwoPhaseOrbit.ofSlowCurve curve.shape curve.high ε₀ ∃ Glim > 0, ∃ Clim : EuclideanSpace ℝ (Fin 2), Tendsto (fun j ↦ (orbit.state j).amplitude) atTop (𝓝 Glim) ∧ Tendsto (fun j ↦ (orbit.state j).center) atTop (𝓝 Clim) ∧ Tendsto (fun k ↦ (orbit.state (k / 2)).ε) atTop (𝓝 0) ∧ {z : EuclideanSpace ℝ (Fin 2) | MapClusterPt z atTop orbit.endpoint} = DFP.TwoPhaseOrbit.limitCircle Clim Glim ∧ ∀ k : ℕ, 0 < (orbit.state (k / 2)).ε ∧ cR * (orbit.state (k / 2)).ε ≤ ‖orbit.endpoint k - Clim‖ - Glim ∧ ‖orbit.endpoint k - Clim‖ - Glim ≤ CR * (orbit.state (k / 2)).ε ∧ ‖orbit.endpointCorrection Clim k‖ ≤ Ca * (orbit.state (k / 2)).ε ^ 3 := by obtain ⟨ηCircle, hηCircle, hcircle⟩ := slowCurveCircleGeometry curve obtain ⟨ηQuadratic, hηQuadratic, K, hK, hquadratic⟩ := slowCurveAmplitudeQuadraticTail curve obtain ⟨ηGraph, hηGraph, hgraph⟩ := slowCurveStateOnGraph curve obtain ⟨ηScale, hηScale, hscale⟩ := slowCurveScaleAsymptotic curve obtain ⟨ηAmplitude, hηAmplitude, Gmin, hGmin, M, hGminM, hamplitude⟩ := DFP.TwoPhaseOrbit.slowCurveAmplitudeUniformBounds curve.shape curve.high curve.isInvariant curve.shapeRemainder curve.highRemainder obtain ⟨ηRadius, hηRadius, ω, hω, hradius⟩ := DFP.TwoPhaseOrbit.slowCurvePhaseRadiusErrorUniform curve.shape curve.high curve.isInvariant curve.shapeRemainder curve.highRemainder obtain ⟨ηCorrection, hηCorrection, Ca, hCa, hcorrection⟩ := DFP.TwoPhaseOrbit.slowCurveEndpointCorrectionUniformBound curve.shape curve.high curve.isInvariant curve.shapeRemainder curve.highRemainder let ηBase := ηRadius / 2 have hηBase : ηBase ∈ Set.Ioc 0 ηRadius := ⟨half_pos hηRadius.1, half_le_self hηRadius.1.le⟩ have hωBase : 0 ≤ ω ηBase := hω.1 ηBase hηBase let E := K + ω ηBase let cR := (13 / 6) * Gmin let CR := (13 / 3) * M + E have hE : 0 < E := add_pos_of_pos_of_nonneg hK hωBase have hcR : 0 < cR := by dsimp only [cR] positivity have hM : 0 < M := hGmin.trans_le hGminM have hCR : 0 < CR := by dsimp only [CR] positivity let εbar := min ηCircle (min ηQuadratic (min ηGraph (min ηScale (min ηAmplitude (min ηBase (min ηCorrection (min 1 (cR / (E + 1))))))))) have hden : 0 < E + 1 := by positivity have hεbarPos : 0 < εbar := lt_min hηCircle.1 (lt_min hηQuadratic.1 (lt_min hηGraph.1 (lt_min hηScale (lt_min hηAmplitude.1 (lt_min hηBase.1 (lt_min hηCorrection.1 (lt_min zero_lt_one (div_pos hcR hden)))))))) have hεbarLt : εbar < 1 / 4 := (min_le_left _ _).trans_lt hηCircle.2 refine ⟨εbar, ⟨hεbarPos, hεbarLt⟩, cR, hcR, CR, hCR, Ca, hCa, ?_⟩ intro ε₀ hε₀ dsimp only let orbit := DFP.TwoPhaseOrbit.ofSlowCurve curve.shape curve.high ε₀ have hsmall : ε₀ ≤ ηCircle ∧ ε₀ ≤ ηQuadratic ∧ ε₀ ≤ ηGraph ∧ ε₀ ≤ ηScale ∧ ε₀ ≤ ηAmplitude ∧ ε₀ ≤ ηBase ∧ ε₀ ≤ ηCorrection ∧ ε₀ ≤ 1 ∧ ε₀ ≤ cR / (E + 1) := by simpa only [εbar, le_min_iff] using hε₀.2 have hεCircle : ε₀ ∈ Set.Ioc 0 ηCircle := ⟨hε₀.1, hsmall.1⟩ have hεQuadratic : ε₀ ∈ Set.Ioc 0 ηQuadratic := ⟨hε₀.1, hsmall.2.1⟩ have hεGraph : ε₀ ∈ Set.Ioc 0 ηGraph := ⟨hε₀.1, hsmall.2.2.1⟩ have hεScale : ε₀ ∈ Set.Ioc 0 ηScale := ⟨hε₀.1, hsmall.2.2.2.1⟩ have hεAmplitude : ε₀ ∈ Set.Ioc 0 ηAmplitude := ⟨hε₀.1, hsmall.2.2.2.2.1⟩ have hεBase : ε₀ ∈ Set.Ioc 0 ηBase := ⟨hε₀.1, hsmall.2.2.2.2.2.1⟩ have hεCorrection : ε₀ ∈ Set.Ioc 0 ηCorrection := ⟨hε₀.1, hsmall.2.2.2.2.2.2.1⟩ obtain ⟨Glim, hGlim, Clim, hAmp, hCenter, _, hCluster⟩ := hcircle ε₀ hεCircle obtain ⟨G', hGRange, hGLimit, _⟩ := hamplitude ε₀ hεAmplitude have hGident : G' = Glim := tendsto_nhds_unique hGLimit hAmp rw [hGident] at hGRange obtain ⟨G'', _, hG''Limit, hQuadratic⟩ := hquadratic ε₀ hεQuadratic have hG''ident : G'' = Glim := tendsto_nhds_unique hG''Limit hAmp rw [hG''ident] at hQuadratic have hscaleZero : Tendsto (fun j ↦ (orbit.state j).ε) atTop (𝓝 0) := (hscale ε₀ hεScale).symm.tendsto_nhds inverseCubicScale_tendsto_zero have hflatScale : Tendsto (fun k ↦ (orbit.state (k / 2)).ε) atTop (𝓝 0) := hscaleZero.comp (Nat.tendsto_div_const_atTop (ne_of_gt Nat.two_pos)) have hscalePositive (j : ℕ) : 0 < (orbit.state j).ε := (hgraph ε₀ hεGraph j).2.1 have hscaleBound (j : ℕ) : (orbit.state j).ε ≤ ε₀ := (hgraph ε₀ hεGraph j).2.2 have hradial (j : ℕ) (σ : Fin 2) : cR * (orbit.state j).ε ≤ ‖orbit.endpoint (2 * j + σ.val) - Clim‖ - Glim ∧ ‖orbit.endpoint (2 * j + σ.val) - Clim‖ - Glim ≤ CR * (orbit.state j).ε := by have hRadius : |‖orbit.endpoint (2 * j + σ.val) - Clim‖ - (orbit.state j).amplitude| ≤ ω ηBase * (orbit.state j).ε ^ 2 := hradius ηBase hηBase ε₀ hεBase Clim hCenter j σ have hAmpError : |(orbit.state j).amplitude - Glim - (13 / 3) * Glim * (orbit.state j).ε| ≤ K * (orbit.state j).ε ^ 2 := hQuadratic j have hError : |‖orbit.endpoint (2 * j + σ.val) - Clim‖ - Glim - (13 / 3) * Glim * (orbit.state j).ε| ≤ E * (orbit.state j).ε ^ 2 := by calc _ = |(‖orbit.endpoint (2 * j + σ.val) - Clim‖ - (orbit.state j).amplitude) + ((orbit.state j).amplitude - Glim - (13 / 3) * Glim * (orbit.state j).ε)| := by congr 1 ring _ ≤ |‖orbit.endpoint (2 * j + σ.val) - Clim‖ - (orbit.state j).amplitude| + |(orbit.state j).amplitude - Glim - (13 / 3) * Glim * (orbit.state j).ε| := abs_add_le _ _ _ ≤ ω ηBase * (orbit.state j).ε ^ 2 + K * (orbit.state j).ε ^ 2 := add_le_add hRadius hAmpError _ = _ := by dsimp only [E] ring have hOne : (orbit.state j).ε ≤ 1 := (hscaleBound j).trans hsmall.2.2.2.2.2.2.2.1 have hCap : (orbit.state j).ε ≤ cR / (E + 1) := (hscaleBound j).trans hsmall.2.2.2.2.2.2.2.2 have hCapMul : (orbit.state j).ε * (E + 1) ≤ cR := (le_div_iff₀ hden).mp hCap have hSmall : E * (orbit.state j).ε ≤ cR := by nlinarith [hscalePositive j] have hErrorLower : E * (orbit.state j).ε ^ 2 ≤ cR * (orbit.state j).ε := by calc _ = (E * (orbit.state j).ε) * (orbit.state j).ε := by ring _ ≤ _ := mul_le_mul_of_nonneg_right hSmall (hscalePositive j).le have hErrorUpper : E * (orbit.state j).ε ^ 2 ≤ E * (orbit.state j).ε := by have hsquare : (orbit.state j).ε ^ 2 ≤ (orbit.state j).ε := by nlinarith [hscalePositive j] exact mul_le_mul_of_nonneg_left hsquare hE.le have hcoefficient : (0 : ℝ) ≤ 13 / 3 := by norm_num have hLeadingLower : (13 / 3) * Gmin * (orbit.state j).ε ≤ (13 / 3) * Glim * (orbit.state j).ε := mul_le_mul_of_nonneg_right (mul_le_mul_of_nonneg_left hGRange.1 hcoefficient) (hscalePositive j).le have hLeadingUpper : (13 / 3) * Glim * (orbit.state j).ε ≤ (13 / 3) * M * (orbit.state j).ε := mul_le_mul_of_nonneg_right (mul_le_mul_of_nonneg_left hGRange.2 hcoefficient) (hscalePositive j).le have hAbs := abs_le.mp hError constructor · dsimp only [cR] at hErrorLower ⊢ linarith · dsimp only [CR] linarith refine ⟨Glim, hGlim, Clim, hAmp, hCenter, hflatScale, hCluster, ?_⟩ intro k have hRadial : cR * (orbit.state (k / 2)).ε ≤ ‖orbit.endpoint k - Clim‖ - Glim ∧ ‖orbit.endpoint k - Clim‖ - Glim ≤ CR * (orbit.state (k / 2)).ε := by rcases Nat.even_or_odd' k with ⟨j, rfl | rfl⟩ · simpa using hradial j 0 · have hdiv : (2 * j + 1) / 2 = j := by omega simpa [hdiv] using hradial j 1 exact ⟨hscalePositive (k / 2), hRadial.1, hRadial.2, hcorrection ε₀ hεCorrection Clim hCenter k⟩

Helper for cm-center-proposition: removing a nonzero amplitude from a normalized step remainder recovers the source physical step equivalence.

theorem stepEquivalentOfNormalizedRemainder (s a ε : ℕ → ℝ) (c : ℝ) (hc : c ≠ 0) (ha : ∀ j, a j ≠ 0) (hremainder : (fun j ↦ s j / a j - c * ε j ^ 2) =o[atTop] (fun j ↦ ε j ^ 2)) : s ~[atTop] (fun j ↦ c * a j * ε j ^ 2) := by have hproduct := (Asymptotics.isBigO_refl a atTop).mul_isLittleO hremainder have hscaled := Asymptotics.IsLittleO.const_mul_right hc hproduct have hdifference : (fun j ↦ s j - c * a j * ε j ^ 2) =o[atTop] (fun j ↦ c * a j * ε j ^ 2) := by apply hscaled.congr' · filter_upwards [] with j field_simp [ha j] · filter_upwards [] with j ring exact hdifference

Helper for cm-center-proposition: both physical source phase steps have exact leading norms and , uniformly on one initialization neighborhood.

theorem slowCurveStepAsymptotics (curve : DFP.TwoLeg.SlowCurve) : ∃ εbar ∈ Set.Ioo (0 : ℝ) (1 / 4), ∀ ε₀ ∈ Set.Ioc 0 εbar, let orbit := DFP.TwoPhaseOrbit.ofSlowCurve curve.shape curve.high ε₀ (fun j ↦ ‖(orbit.state j).firstDisplacement‖) ~[atTop] (fun j ↦ 2 * (orbit.state j).amplitude * (orbit.state j).ε ^ 2) ∧ (fun j ↦ ‖(orbit.state j).secondDisplacement‖) ~[atTop] (fun j ↦ (orbit.state j).amplitude * (orbit.state j).ε ^ 2) := by obtain ⟨ηFirst, hηFirst, hfirst⟩ := DFP.TwoPhaseOrbit.slowCurveFirstStepNormRemainder curve.shape curve.high curve.isInvariant curve.shapeRemainder curve.highRemainder obtain ⟨ηSecond, hηSecond, hsecond⟩ := DFP.TwoPhaseOrbit.slowCurveSecondStepNormRemainder curve.shape curve.high curve.isInvariant curve.shapeRemainder curve.highRemainder have heighth : (1 / 8 : ℝ) ∈ Set.Ioo 0 (1 / 4) := by norm_num obtain ⟨ηValid, hηValid, hvalid⟩ := DFP.TwoPhaseOrbit.ofSlowCurve_phaseValidity curve.shape curve.high curve.isInvariant curve.shapeRemainder curve.highRemainder (1 / 8) heighth let εbar := min ηValid (min ηFirst ηSecond) have hεbarPos : 0 < εbar := lt_min hηValid.1 (lt_min hηFirst hηSecond) have hεbarLt : εbar < 1 / 4 := ((min_le_left _ _).trans hηValid.2).trans_lt heighth.2 refine ⟨εbar, ⟨hεbarPos, hεbarLt⟩, ?_⟩ intro ε₀ hε₀ dsimp only let orbit := DFP.TwoPhaseOrbit.ofSlowCurve curve.shape curve.high ε₀ have hεValid : ε₀ ∈ Set.Ioc 0 ηValid := ⟨hε₀.1, hε₀.2.trans (min_le_left _ _)⟩ have hεFirst : ε₀ ∈ Set.Ioc 0 ηFirst := ⟨hε₀.1, hε₀.2.trans ((min_le_right _ _).trans (min_le_left _ _))⟩ have hεSecond : ε₀ ∈ Set.Ioc 0 ηSecond := ⟨hε₀.1, hε₀.2.trans ((min_le_right _ _).trans (min_le_right _ _))⟩ have ha (j : ℕ) : (orbit.state j).amplitude ≠ 0 := ne_of_gt (hvalid ε₀ hεValid j).amplitude_pos have htwo : (2 : ℝ) ≠ 0 := by norm_num constructor · exact stepEquivalentOfNormalizedRemainder _ _ _ 2 htwo ha (hfirst ε₀ hεFirst) · have hremainder : (fun j ↦ ‖(orbit.state j).secondDisplacement‖ / (orbit.state j).amplitude - 1 * (orbit.state j).ε ^ 2) =o[atTop] (fun j ↦ (orbit.state j).ε ^ 2) := by simpa only [one_mul] using hsecond ε₀ hεSecond simpa only [one_mul] using stepEquivalentOfNormalizedRemainder _ _ _ 1 one_ne_zero ha hremainder

Helper for cm-center-proposition: actual flattened endpoint differences are precisely the two physical displacements of the original recurrence.

theorem endpointDisplacements (orbit : DFP.TwoPhaseOrbit) (j : ℕ) : orbit.endpoint (2 * j + 1) - orbit.endpoint (2 * j) = (orbit.state j).firstDisplacement ∧ orbit.endpoint (2 * j + 2) - orbit.endpoint (2 * j + 1) = (orbit.state j).secondDisplacement := by constructor · rw [DFP.TwoPhaseOrbit.endpoint_odd, DFP.TwoPhaseOrbit.endpoint_even, DFP.TwoPhaseOrbit.State.middlePoint_def] abel · have hindex : 2 * j + 2 = 2 * (j + 1) := by omega rw [hindex, DFP.TwoPhaseOrbit.endpoint_even, DFP.TwoPhaseOrbit.endpoint_odd, orbit.state_succ, DFP.TwoPhaseOrbit.State.next_point] abel

Helper for cm-center-proposition: a cluster point of centers cannot lie inside ANY one of their pairwise disjoint positive-radius open balls. Early balls are included; tangency at their boundary is allowed.

theorem clusterPoint_outside_disjointBalls {X : Type*} [PseudoMetricSpace X] (c : ℕ → X) (ρ : ℕ → ℝ) (hρ : ∀ i, 0 < ρ i) (hdisjoint : Pairwise (fun i j ↦ Disjoint (Metric.ball (c i) (ρ i)) (Metric.ball (c j) (ρ j)))) (z : X) (hz : MapClusterPt z atTop c) (i : ℕ) : ρ i ≤ dist z (c i) := by by_contra hnot have hzball : z ∈ Metric.ball (c i) (ρ i) := lt_of_not_ge hnot obtain ⟨j, hj, hlimit⟩ := hz.tendsto_subseq have hmem : ∀ᶠ n in atTop, c (j n) ∈ Metric.ball (c i) (ρ i) := hlimit.eventually (Metric.isOpen_ball.mem_nhds hzball) have hafter : ∀ᶠ n in atTop, i < j n := hj.tendsto_atTop.eventually (eventually_gt_atTop i) obtain ⟨n, hmemn, haftern⟩ := (hmem.and hafter).exists have hne : i ≠ j n := ne_of_lt haftern have hself : c (j n) ∈ Metric.ball (c (j n)) (ρ (j n)) := Metric.mem_ball_self (hρ _) exact Set.disjoint_left.mp (hdisjoint hne) hmemn hself

Helper for cm-center-proposition: the actual FULL-circle cluster property and source disjointness exclude every early open interpolation ball from the entire limiting circle, without any extra global support margin.

theorem limitingCircle_outside_disjointBalls (orbit : DFP.TwoPhaseOrbit) (C : EuclideanSpace ℝ (Fin 2)) (G : ℝ) (ρ : ℕ → ℝ) (hρ : ∀ i, 0 < ρ i) (hdisjoint : Pairwise (fun i j ↦ Disjoint (Metric.ball (orbit.endpoint i) (ρ i)) (Metric.ball (orbit.endpoint j) (ρ j)))) (hcluster : {z : EuclideanSpace ℝ (Fin 2) | MapClusterPt z atTop orbit.endpoint} = DFP.TwoPhaseOrbit.limitCircle C G) : ∀ i : ℕ, Disjoint (Metric.ball (orbit.endpoint i) (ρ i)) (DFP.TwoPhaseOrbit.limitCircle C G) := by intro i apply Set.disjoint_left.mpr intro z hzball hzcircle have hz : MapClusterPt z atTop orbit.endpoint := by rw [← hcluster] at hzcircle exact hzcircle have hbound := clusterPoint_outside_disjointBalls orbit.endpoint ρ hρ hdisjoint z hz i exact not_lt_of_ge hbound hzball

Helper for cm-center-proposition: the source graph construction supplies ACTUAL radial, correction, scale-limit and full-circle data for conditional interpolation. Only the manuscript's eventual support-radius orders and explicit disjointness remain for the separate analytic interpolation theorem.

theorem centerManifoldInterpolationGeometry : ∃ curve : DFP.TwoLeg.SlowCurve, ContDiffAt ℝ 7 (fun ε ↦ (curve.shape ε, curve.high ε)) 0 ∧ (curve.shape 0, curve.high 0) = (2, 1) ∧ ∃ εbar ∈ Set.Ioo (0 : ℝ) (1 / 4), ∃ cR > 0, ∃ CR > 0, ∃ Ca > 0, ∀ ε₀ ∈ Set.Ioc 0 εbar, let orbit := DFP.TwoPhaseOrbit.ofSlowCurve curve.shape curve.high ε₀ ∃ Glim > 0, ∃ Clim : EuclideanSpace ℝ (Fin 2), Tendsto (fun j ↦ (orbit.state j).amplitude) atTop (𝓝 Glim) ∧ Tendsto (fun j ↦ (orbit.state j).center) atTop (𝓝 Clim) ∧ Tendsto (fun k ↦ (orbit.state (k / 2)).ε) atTop (𝓝 0) ∧ {z : EuclideanSpace ℝ (Fin 2) | MapClusterPt z atTop orbit.endpoint} = DFP.TwoPhaseOrbit.limitCircle Clim Glim ∧ ∀ k : ℕ, 0 < (orbit.state (k / 2)).ε ∧ cR * (orbit.state (k / 2)).ε ≤ ‖orbit.endpoint k - Clim‖ - Glim ∧ ‖orbit.endpoint k - Clim‖ - Glim ≤ CR * (orbit.state (k / 2)).ε ∧ ‖orbit.endpointCorrection Clim k‖ ≤ Ca * (orbit.state (k / 2)).ε ^ 3 := by obtain ⟨curve, hsmooth, hbase, _⟩ := existsSourceSlowCurve exact ⟨curve, hsmooth, hbase, slowCurveRadialCoefficientBounds curve⟩
end DFPWolfe.V21.CenterManifold