A trajectory near a reference trajectory is quadratically close to the linearized solution
ProvedFiniteEnergyPlane.norm_sub_sub_linearized_leanalysisordinary-differential-equations
Let be a real normed space and a vector field. Let and be trajectories of on (, ), let , and let solve the linearized equation . Let be constants with , and suppose:
- on every ball the field is -Lipschitz and satisfies ;
- for ;
- .
Then for all :
Role: over a fixed time window, a trajectory starting near a reference trajectory follows the linearized solution up to an error quadratic in the size of the perturbation, in both time directions.
Formalization Note: trajectories are ∀ t, HasDerivAt x (R (x t)) t; is fderiv ℝ R p.
Preamble
import Mathlib.Analysis.Calculus.FDeriv.Basic import Mathlib.Analysis.Calculus.Deriv.Basic import Mathlib.Analysis.SpecialFunctions.Exp
Formal statement
namespace FiniteEnergyPlane
open Metric Set
/-- Quadratic closeness to the linearization along a trajectory. Let `x` and `y` be trajectories
of a vector field `R` on a real normed space, and let `v` solve the linearized equation
`v'(t) = DR(x(c + t)) v(t)`. Suppose that on every ball `closedBall (x t) δ` the field `R` is
`D`-Lipschitz and satisfies `‖R z - R(x t) - DR(x t)(z - x t)‖ ≤ D ‖z - x t‖²`, that
`‖v‖ ≤ M ≤ δ` on `[-L, L]`, and that `‖y 0 - x c‖ e^{DL} < δ`. Then for `t ∈ [-L, L]`,
`‖y t - x(c + t)‖ ≤ ‖y 0 - x c‖ e^{DL}` and
`‖y t - x(c + t) - v t‖ ≤ (‖y 0 - x c - v 0‖ + D M² L) e^{(D+1)L}`. -/
theorem norm_sub_sub_linearized_le {E : Type*} [NormedAddCommGroup E] [NormedSpace ℝ E]
{R : E → E} {x y v : ℝ → E} {δ D L M c : ℝ} (hD : 0 ≤ D) (hL : 0 ≤ L)
(hlip : ∀ t, ∀ z ∈ closedBall (x t) δ, ∀ z' ∈ closedBall (x t) δ,
‖R z - R z'‖ ≤ D * ‖z - z'‖)
(hquad : ∀ t, ∀ z ∈ closedBall (x t) δ,
‖R z - R (x t) - fderiv ℝ R (x t) (z - x t)‖ ≤ D * ‖z - x t‖ ^ 2)
(hx : ∀ t, HasDerivAt x (R (x t)) t) (hy : ∀ t, HasDerivAt y (R (y t)) t)
(hv : ∀ t, HasDerivAt v (fderiv ℝ R (x (c + t)) (v t)) t)
(hvM : ∀ t ∈ Icc (-L) L, ‖v t‖ ≤ M) (hMδ : M ≤ δ)
(hy0 : ‖y 0 - x c‖ * Real.exp (D * L) < δ) :
∀ t ∈ Icc (-L) L, ‖y t - x (c + t)‖ ≤ ‖y 0 - x c‖ * Real.exp (D * L) ∧
‖y t - x (c + t) - v t‖ ≤
(‖y 0 - x c - v 0‖ + D * M ^ 2 * L) * Real.exp ((D + 1) * L) := by
sorry
end FiniteEnergyPlane
Source
Standard (Grönwall's inequality); used for U. Hryniewicz, Trans. AMS 364 (2012), proof of Lemma 6.9