Historical behavior of skew products and arcsine laws
Pablo G. Barrientos, Raul R. Chavez
TL;DR
The paper develops a unified probabilistic mechanism for historical behavior in skew products with one-dimensional fibers by linking weak arcsine fluctuations to non-convergence of Birkhoff averages. It shows that when the fiber process is conjugate to a random walk (or when the base measure is ergodic with a fiber fluctuation law), almost every orbit exhibits historical behavior, and the empirical measures accumulate on convex combinations of endpoint deltas. This framework recovers and extends many known examples (e.g., Bonifant–Milnor, Molinek, Thaler-type maps, skew-flows) and applies to both one-step and mild skew products, including skew-translations and skew-flows, via coboundary/cohomology criteria. The results illuminate a probabilistic pathway to Takens’ last problem by demonstrating how arcsine-type fluctuations can underpin comprehensive non-statistical dynamics in broad skew-product settings with one-dimensional fibers. The work thus provides a cohesive, general method to identify, classify, and extend historical behavior in deterministic systems with random-like fiber dynamics and has implications for understanding physical measures and long-term statistical predictions in non-statistical regimes.
Abstract
We study the occurrence of historical behavior for almost every point in the setting of skew products with one-dimensional fiber dynamics. Under suitable ergodic conditions, we establish that a weak form of the arcsine law leads to the non-convergence of Birkhoff averages along almost every orbit. As an application, we show that this phenomenon occurs for one-step skew product maps over a Bernoulli shift, where the stochastic process induced by the iterates of the fiber maps is conjugate to a random walk. Furthermore, we revisit known examples of skew products that exhibit historical behavior almost everywhere, verifying that they fulfill the required ergodic and probabilistic conditions. Consequently, our results provide a unified and generalized framework that connects such behaviors to the arcsine distribution of the orbits.
