Review : Power Series — Question 4

PDF ↗

Question 4

For an integer N≥0N\ge 0, define SN(x)=∑n=0Nxn,f(x)=11−x,|x|<1.S_N(x)=\sum_{n=0}^{N}x^n,\qquad f(x)=\frac 1{1-x},\qquad |x|<1. A student says that choosing a sufficiently large NN makes this approximation uniformly accurate throughout the entire convergence interval.

Tasks

  1. Derive the exact remainder f−SNf-S_N from a finite geometric identity. State its sign for positive and negative xx.

  2. Find the least NN guaranteeing absolute error at most 10−310^{-3} for every x∈[−1/2,1/2]x\in[-1/2,1/2]. State both the largest exponent and the number of retained terms.

  3. Express the relative error |f−SN|/|f||f-S_N|/|f|. For any fixed NN, find a point in (0,1)(0,1) where this error is exactly 1/21/2.

  4. Decide whether convergence is uniform on (−1,1)(-1,1), for either absolute or relative error, and justify the distinction from convergence at each fixed point. Plot S2,S5,S10S_2,S_5,S_{10} and ff on [0,0.85][0,0.85].

Original worksheet page 1: question and worked solution for 6-1-004
Show solutionHide solution

Question 4 – Solution

Strategy. An exact remainder supplies a sharp interval bound and exposes what fails when points approach the boundary as the degree changes.

Step 1: Retain the exact geometric remainder. Multiplication gives (1−x)SN=1−xN+1(1-x)S_N=1-x^{N+1}. Therefore f(x)−SN(x)=xN+11−x.\boxed{f(x)-S_N(x)=\frac{x^{N+1}}{1-x}.} For 0<x<10<x<1 it is positive. For −1<x<0-1<x<0, its sign is (−1)N+1(-1)^{N+1}; at x=0x=0 it is zero. The denominator is positive throughout the interval.

Step 2: Find the least degree on the compact interval. For |x|≤1/2|x|\le 1/2, |f−SN|≤(1/2)N+11−1/2=2−N.|f-S_N|\le\frac{(1/2)^{N+1}}{1-1/2}=2^{-N}. Equality holds at x=1/2x=1/2, so this is the exact maximum, not merely a convenient estimate. Since 2−9>10−32^{-9}>10^{-3} and 2−10<10−32^{-10}<10^{-3}, N=10\boxed{N=10} is least. This retains 11 terms\boxed{11\text{ terms}}, including the constant.

Step 3: Select a moving point with fixed relative error. The relative error simplifies to |x|N+1\boxed{|x|^{N+1}}. Taking xN=2−1/(N+1)x_N=2^{-1/(N+1)} gives a point in (0,1)(0,1) with relative error exactly 1/21/2 for every N≥0N\ge 0. These points approach 11 as NN increases.

Step 4: Distinguish fixed-point from uniform accuracy. At each fixed |x|<1|x|<1, both errors tend to zero. Yet for every fixed NN, xN+1/(1−x)→+∞x^{N+1}/(1-x)\to+\infty as x→1−x\to 1^-, so the supremum of the absolute error on (−1,1)(-1,1) is infinite. The supremum of relative error is 11. Neither error converges uniformly to zero on that whole interval. The plot shows slower approximation near the right boundary; on any fixed [−r,r][-r,r], r<1r<1, the bound rN+1/(1−r)→0r^{N+1}/(1-r)\to 0 does give uniform convergence.

See the diagram in the original worksheet below.

Original worksheet page 2: question and worked solution for 6-1-004

Original worksheet layout. Use Enlarge or open the PDF for a closer view.