equiv

# equiv - 2 ( G-H ) H + . . . + ( G-H ) H n-1 Every term has...

This preview shows pages 1–2. Sign up to view the full content.

Equivalence Theorem (Lax-Richtmyer) The Fundamental Theorem of Numerical Analysis .F o r consistent numerical approximations, stability and convergence are equivalent. Lax proved for IVPs. Applies as well to BVPs, approximations to func- tions and integrals, and PDEs. Approximate Lu = f by L n u n = f n .W ew i l la s s um et h ep r o b l em sa r e close, and prove that the solutions are close. Consistency implies f n f and L n u Lu as n →∞ t 0, with n Δ t = t ±xed. Stability implies L - 1 n remains uniformly bounded. Then as n →∞ , u n converges to u : u - u n L - 1 n ( L n u - Lu )+ L - 1 n ( f - f n ) 0 Let’s look at the details for the IVP du/dt = au , u ( t 0 )= u 0 (proof due to Strang). The exact solution is u ( t )= e at u 0 H n u 0 ,where H =exp { a Δ t } is the exact growth factor. The approximate solution is u n = G n u 0 . Stability implies | G n |≤ e Kn Δ t = e Kt where K is a positive constant independent of n . Consistency implies | G - H |≤ C Δ t p +1 ,p> 0 Then we’ll prove convergence : | G n u 0 - H n u 0 |→ 0 as n →∞ . We use a telescoping identity ( ± ) G n - H n = G n - G n - 1 H + G n - 1 H - G n - 2 H 2 + ... +

This preview has intentionally blurred sections. Sign up to view the full version.

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

Unformatted text preview: 2 ( G-H ) H + . . . + ( G-H ) H n-1 Every term has a factor G-H , and | G-H | C t p +1 by consistency. Every term has a power of G (possibly G ) which is bounded by stability. Every term has a power of H which is bounded since the continuous problem is well-posed. There are n = t/ t terms in Eq. ( ). Therefore as n , | G n-H n | t t e Kt C t p +1 = O ( t p ) The telescoping series ( ) is exactly how error accumulates in a di f erence equation. ( ) G n-H n = n j =1 G n-j ( G-H ) H j-1 H j-1 propagates the exact solution to timelevel j-1; ( G-H ) is the local truncation error going From timelevel j-1 to j ; and G n-j propagates this error Forward with the diference method to timelevel n . 2...
View Full Document

## This note was uploaded on 03/11/2012 for the course MAT 421 taught by Professor Staff during the Fall '11 term at ASU.

### Page1 / 2

equiv - 2 ( G-H ) H + . . . + ( G-H ) H n-1 Every term has...

This preview shows document pages 1 - 2. Sign up to view the full document.

View Full Document
Ask a homework question - tutors are online