Long-time behaviour of discretizations of the simple pendulum equation

TL;DR

This paper evaluates various numerical discretizations of the simple pendulum, emphasizing their long-term behavior, and introduces a new scheme that accurately preserves the period of small oscillations.

Contribution

It compares existing schemes and proposes a modified discrete gradient method that better preserves oscillation periods over long times.

Findings

New scheme preserves period of small oscillations accurately

Symplectic and integrable schemes show improved long-term behavior

Systematic errors in period and amplitude are analyzed

Abstract

We compare the performance of several discretizations of the simple pendulum equation in a series of numerical experiments. The stress is put on the long-time behaviour. We choose for the comparison numerical schemes which preserve the qualitative features of solutions (like periodicity). All these schemes are either symplectic maps or integrable (preserving the energy integral) maps, or both. We describe and explain systematic errors (produced by any method) in numerical computations of the period and the amplitude of oscillations. We propose a new numerical scheme which is a modification of the discrete gradient method. This discretization preserves (almost exactly) the period of small oscillations for any time step.