Approximation is a claim with a limit and an error
Objective: Which hypothesis or scale prevents the statement in “Approximation is a claim with a limit and an error” from becoming an unsupported equality?
Unit 1, chapter 1 begins with the concrete question behind “Approximation is a claim with a limit and an error.” Writing f≈g is incomplete until one says which parameter moves and how the discrepancy is measured. The first job is to identify the quantity that changes, the parameter approaching its limit, and the scale on which an error will be judged. No symbol is allowed to stand in for those three decisions.
The chapter fixes its language through the following definition: An asymptotic approximation on a set D_ε is a chosen simple A_ε together with a stated bound for R_ε=f_ε−A_ε as ε approaches its declared limit. Read every equality, asymptotic sign, derivative, integral, and subscript together with its domain and limiting process. Two formulas that look alike can make different claims when one is pointwise and the other is uniform.
The usable statement is: If |R_ε|≤Cρ(ε) uniformly on D_ε and ρ(ε)→0, then A_ε approximates f_ε with absolute error O(ρ(ε)) on that set. The proof route is not a name to memorize. Subtract the proposed approximation, factor the dominant scale, and bound the remaining dimensionless ratio independently of ε. Each transformation must preserve the declared order of the remainder, and the final line must say where the estimate is valid.
The worked model asks: For f_ε(x)=sin(εx)/ε on |x|≤1, prove that x is an O(ε²) approximation and state whether the claim is uniform. The calculation is organized as scale selection, coefficient or phase calculation, and an independent residual check. Its resulting bounded claim is Taylor’s remainder gives |f_ε(x)−x|≤ε²/6 uniformly for |x|≤1; relative error is not controlled at x=0 by division. The result is evidence only for the stated model and parameter regime.
The failure boundary belongs to the concept: An approximation with small absolute error can have huge relative error near a zero of the target. A strong answer therefore contains the approximation, its remainder scale, the hypotheses, and one explicit test showing what breaks when a hypothesis is removed. A small numerical residual is never silently promoted to a theorem.
If |R_ε|≤Cρ(ε) uniformly on D_ε and ρ(ε)→0, then A_ε approximates f_ε with absolute error O(ρ(ε)) on that set.
Start from the exact definition in this chapter and rewrite the target in the scale named there. For “Approximation is a claim with a limit and an error,” this prevents a formal coefficient comparison from being used before the limiting variable and domain are fixed.
Subtract the proposed approximation, factor the dominant scale, and bound the remaining dimensionless ratio independently of ε. Keep the first omitted contribution visible rather than replacing it by an unexplained ellipsis; it supplies the order against which the retained terms are tested.
Substitute the proposed approximation back into the exact relation and compare the residual with the first neglected scale. This yields Taylor’s remainder gives |f_ε(x)−x|≤ε²/6 uniformly for |x|≤1; relative error is not controlled at x=0 by division. The conclusion stops at the boundary stated in the chapter.
For f_ε(x)=sin(εx)/ε on |x|≤1, prove that x is an O(ε²) approximation and state whether the claim is uniform.
- Declare the limiting parameter, domain, normalization, and target error; then translate An asymptotic approximation on a set D_ε is a chosen simple A_ε together with a stated bound for R_ε=f_ε−A_ε as ε approaches its declared limit. into the notation of the problem.
- Subtract the proposed approximation, factor the dominant scale, and bound the remaining dimensionless ratio independently of ε. Record every retained term and the first omitted scale so that no cancellation or change of order is hidden.
- Evaluate the resulting expression and perform the residual or ratio check. The checked result is Taylor’s remainder gives |f_ε(x)−x|≤ε²/6 uniformly for |x|≤1; relative error is not controlled at x=0 by division.
Result: Taylor’s remainder gives |f_ε(x)−x|≤ε²/6 uniformly for |x|≤1; relative error is not controlled at x=0 by division.