UNIGE document Book Chapter
previous document  unige:12439  next document
add to browser collection

Asymptotic expansions and backward analysis for numerical integrators

Lubich, Christian
Published in Rafael de la Llave; Linda R. Petzold & Jens Lorenz. Dynamics of algorithms. New York: Springer. 2000, p. 91-106
Collection The IMA volumes in mathematics and its applications; 118
Abstract For numerical integrators of ordinary differential equations we compare the theory of asymptotic expansions of the global error with backward error analysis. On a formal level both approaches are equivalent. If, however, the arising divergent series are truncated, important features such as the semigroup property, structure perservation and exponentially small estimates over long times are valid only for the backward error analysis. We consider one-step methods as well as multistep methods, and we illustrate the theoretical results on several examples. In particular, we study the preservation of weakly stable limit cycles by symmetric methods.
Keywords Asymptotic expansionsBackward error analysisOne-step methodsMultistep methodsLong-time behavior
Stable URL https://archive-ouverte.unige.ch/unige:12439
Full text
Book chapter (1.1 MB) - public document Free access

215 hits



Deposited on : 2010-11-12

Export document
Format :
Citation style :