Diffusing up the Hill: Dynamics and Equipartition in Highly Unstable Systems

Description and understanding of stochastic motion of a particle in an unstable potential can be principally limited by a number of diverging trajectories leading to undefined statistic moments of particle's position. Since this breaks down the standard statistical analysis of unstable mechanical processes and their applications, a newly proposed approach takes advantage of the local characteristics of the most-likely particle motion instead of the average motion. We experimentally verify theoretical predictions for a Brownian particle moving near an inflection in a cubic optical potential. The most-likely position of the particle atypically shifts against the force despite the trajectories diverge in opposite direction. The local uncertainty around the most-likely position saturates even for strong diffusion and enables well-resolved position detection. Remarkably, the measured particle distribution quickly converges to the quasi-stationary one with the same atypical shift for different initial particle positions. The demonstrated experimental confirmation of the theoretical predictions approves the utilization of local characteristics for description of unstable stochastic motion. The same approach can be further exploited in thermodynamic processes to uncover energetics in unstable systems.