Picard error bound
DisprovedMetodosNumericos.picard_error_boundWith a bound for and a Lipschitz constant in on the rectangle, and , the -th Picard iterate satisfies for every and every , where is a solution of the initial value problem. This is the estimate stated in §9.5.
import Mathlib import Definitions.Def_MetodosNumericos_edoDefs
namespace MetodosNumericos
theorem picard_error_bound (f : ℝ → ℝ → ℝ) (phi : ℝ → ℝ) (x0 y0 a b M N h : ℝ)
(ha : 0 < a) (hb : 0 < b) (hM : 0 < M) (hN : 0 ≤ N) (hh : h = min a (b / M))
(hbound : ∀ x ∈ Set.Icc (x0 - a) (x0 + a), ∀ y ∈ Set.Icc (y0 - b) (y0 + b), |f x y| ≤ M)
(hlip : ∀ x ∈ Set.Icc (x0 - a) (x0 + a), ∀ u ∈ Set.Icc (y0 - b) (y0 + b),
∀ v ∈ Set.Icc (y0 - b) (y0 + b), |f x u - f x v| ≤ N * |u - v|)
(hcont : ContinuousOn (fun p : ℝ × ℝ => f p.1 p.2)
(Set.Icc (x0 - a) (x0 + a) ×ˢ Set.Icc (y0 - b) (y0 + b)))
(hphi0 : phi x0 = y0)
(hsol : ∀ x ∈ Set.Icc (x0 - h) (x0 + h), HasDerivAt phi (f x (phi x)) x)
(hrange : ∀ x ∈ Set.Icc (x0 - h) (x0 + h), phi x ∈ Set.Icc (y0 - b) (y0 + b))
(k : ℕ) (hk : 1 ≤ k) (x : ℝ) (hx : x ∈ Set.Icc (x0 - h) (x0 + h)) :
|phi x - picardSeq f x0 y0 k x| ≤ M * N ^ (k - 1) * h ^ k / (Nat.factorial k : ℝ) := by
sorry
end MetodosNumericosRead-back
What the Lean code literally says, in plain math · self-authored-by-drafting-agent (non-blind)
Disclosure: this read-back is not blind. It was written by the same agent that drafted the Lean statement, at the explicit instruction of the mission's human owner, and not by an independent auditor with fresh context.
For a function of two real variables, a function of one real variable, and reals , , , , , , , the hypotheses are: ; ; ; ; ; for all in the rectangle ; for in and in ; continuous on the rectangle; ; has derivative at every ; and for every such .
For a natural number and a point , the conclusion is
where is the -th Picard iterate defined by and . The exponent is truncated natural-number subtraction, which for is the ordinary difference. The bound is stated with , not with .
Unfaithful to the mathematics: the bound |phi(x) - y_k(x)| <= M N^(k-1) h^k / k! is the formula printed in Freitas p. 186, but it is off by one level of the Picard iteration and fails for k = 1 (example: M = 1, N = 4, a = b = 1, x0 = y0 = 0, f(t, y) = clamp((y - t + 1/4) / (1/4), -1, 1) for t >= 0 gives phi(t) = t and y_1(1) = -1/2, error 3/2 > M h = 1). The classical estimate M N^k h^(k+1) / (k+1)! is stated as MetodosNumericos.picard_error_bound_classical, to which the milestone is now linked.
Confirmed by the mission captain (proposal self-audit).