We should select them so that xx i i z xi z xi xi 1 i

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: ) i1 Substituting this result into (9.1.15) yields je(x )j j 1 X h5 kg (x )k 2 =1 N xx i j j = 0 1 ::: N i1 i or je(x )j j 1 kg (x )k X h5 = 1 kg (x )k Nh5 : 2 2 =1 N xx j 1 xx i j 1 i Thus, je(x )j C h4 j = 0 1 ::: N j (9.1.28a) where 1 C = 2 (b ; a)kg (x )k : xx j (9.1.28b) 1 The pointwise error has been increased by two orders with the special choice of collocation points dictated by (9.1.26). This is most de nitely an example of nodal superconvergence since the global error is still O(h2). It remains to determine the collocation points that satisfy (9.1.26). Let us begin by transforming these integrals to ;1 1] using the mapping x=x h 12+ ;1 i 2 i; = 1: (9.1.29a) Also let i 1 = x 1 2 ; 1h i; = i i 10 2 = x 1 2 + 2h : i; = i (9.1.29b) Then x; i h ( +2 ) 1 2 1= x; i Conditions (9.1.26a,b) become h )3 Z 1 ( + 2 )( ; 2 )d = 0 (2 1 2 i 2= h ( ; 2 ): 2 2 i (9.1.29c) h )4 Z 1 ( + 2 )( ; 2 ) d = 0: (2 1 2 i i 1 1 ; ; Thus, it su ces to determine 1 and 2 such that Z1 1 ; ( + 2 1)( ; 2 2 )P ( )d = 0 (9.1.30) for all linear polynomials P ( ). Since the integran...
View Full Document

This document was uploaded on 03/16/2014 for the course CSCI 6820 at Rensselaer Polytechnic Institute.

Ask a homework question - tutors are online