No label defined (Q4550339)

The authors discuss (mostly at the physical level) some concepts connected with the problem of shadowability of (finite and infinite) pseudotrajectories in dynamical models of physical systems. They mainly consider pseudotrajectories generated by numerical methods and pay the most attention to continuous shadowability (i.e., the possibility of constructing a ``homotopy'' between a pseudotrajectory and the shadowing exact trajectory). They describe various ``glitches'' (i.e., points at which the system fails to have the property of shadowability). Let \(\{p_i\}\) be a pseudotrajectory with one-step error \(\delta\) and let \(\{x_i\}\) be the shadowing trajectory. The authors define the brittleness of the pseudotrajectory as the value \(\max_i|p_i-x_i|/\delta\). Since the evaluation of the brittleness requires the knowledge of the exact shadowing trajectory, a more practical value is the test brittleness defined as follows. Let \(\{p_i\}\) be a pseudotrajectory with one-step error \(\delta\) and let \(\{y_i\}\) be a less noisy pseudotrajectory produced by one step of an appropriate Newton method. Then the test brittleness equals \(\max_i|p_i-y_i|/\delta\). The authors analyze in detail the test brittleness for a forced damped pendulum and for a double rotor system. They study the dependence of the expected shadowing distance and time on the size of the one-step error. Special attention is paid to the influence of fluctuating Lyapunov exponents. The main conclusion of the authors is as follows: for certain multidimensional chaotic systems, no model will produce trajectories that are close to trajectories of the natural system. For the entire collection see [Zbl 0982.37002].