Multilinear Taylor remainder estimate on a real normed-space ball
ProvedRealAnalytic.taylor_remainder_boundderivative-estimatesreal-analysistaylor-series
Let E be a real normed space and F a real Banach space. Suppose f:E→F is smooth on B_r(x), where r>0. Fix a displacement h with norm less than r, an integer k≥1, and M≥0. If the operator norm of the kth Frechet derivative is bounded by M everywhere on that ball, then
The zeroth multilinear differential evaluates to f(x). The intended proof restricts f to the segment t↦x+th and applies the vector-valued one-dimensional Taylor remainder estimate. The segment lies in the ball, and each kth directional differential is bounded by M times the kth power of the norm of h. This is a finite Taylor estimate for arbitrary Banach-valued smooth functions; neither a power series nor analyticity is assumed. Its proof remains an independent obligation.
Preamble
import Mathlib.Analysis.Analytic.Basic import Mathlib.Analysis.Calculus.ContDiff.Basic import Mathlib.Analysis.Calculus.ContDiff.FTaylorSeries import Mathlib.Data.Nat.Factorial.Basic open scoped ContDiff set_option autoImplicit false
Formal statement
theorem RealAnalytic.taylor_remainder_bound
{E F : Type*} [NormedAddCommGroup E] [NormedSpace ℝ E]
[NormedAddCommGroup F] [NormedSpace ℝ F] [CompleteSpace F]
{f : E → F} {x h : E} {r : ℝ} (hr : 0 < r)
(hf : ContDiffOn ℝ ∞ f (Metric.ball x r))
(hh : ‖h‖ < r) {k : ℕ} (hk : 1 ≤ k) {M : ℝ} (hM : 0 ≤ M)
(hb : ∀ y ∈ Metric.ball x r, ‖iteratedFDeriv ℝ k f y‖ ≤ M) :
‖f (x + h) - ∑ j ∈ Finset.range k,
((j.factorial : ℝ)⁻¹ • iteratedFDeriv ℝ j f x) (fun _ => h)‖ ≤
M / (k.factorial : ℝ) * ‖h‖ ^ k := by sorrySource
Hunter, Notes on PDEs (revised 6/18/2014), Theorem 1.27 (printed p.10) and Theorem 2.10, equation (2.7) (printed p.24), https://www.math.ucdavis.edu/~hunter/pdes/pde_notes.pdf. Generalized norm form of the finite Taylor remainder, for real normed domains and Banach codomains. Mathlib.Analysis.Calculus.Taylor.exists_taylor_mean_remainder_bound supplies the one-dimensional vector-valued counterpart; this multivariate specialization is not claimed as a verbatim source statement.