Phantom instabilities in adiabatically driven systems: dynamical sensitivity to computational precision

Abstract

We study the robustness of dynamical phenomena in adiabatically driven nonlinear mappings with skew-product structure. Deviations from true orbits are observed when computations are performed with inadequate numerical precision for monotone, periodic, or quasiperiodic driving. The effect of slow modulation is to “freeze” orbits in long intervals of purely contracting or purely expanding dynamics in the phase space. When computations are carried out with low precision, numerical errors build up phantom instabilities which ultimately force trajectories to depart from the true motion. Thus, the dynamics observed with finite precision computation shows sensitivity to numerical precision: the minimum accuracy required to obtain “true” trajectories is proportional to an internal timescale that can be defined for the adiabatic system.