Numerically induced stochasticity (Q751204)

The authors study numerical effects in computing the motion of a particle in a one-dimensional potential well. Though the true motion is regular due to the constraints put on the particle velocity v at the point x by the principle of conservation of energy, stochastic behavior may result if the motion is computed numerically. Considering simple time integration schemes as mappings from a phase space pair \((v,x)^ n\) at a time level \(n\Delta\) to a phase space pair \((v,x)^{n+1}\) at time level \((n+1)\Delta\) the authors demonstrate that particles move on closed orbits if the time step \(\Delta\) is sufficiently small while there is chaos for larger values of \(\Delta\). If the state phase plane includes a separatix chaos may result also for very small time steps. Beside other time stepping schemes a detailed study is devoted to the leapfrog method applied to motion in a particular double-well potential. In this case stochasticity is shown to occur at much smaller time steps than the usual stability threshold predicts. Additionally, a new symmetrical variant of the standard map is introduced.