Reaching and violating thermodynamic uncertainty bounds in information engines

TL;DR

This study experimentally investigates a cyclic information engine operating in non-equilibrium steady states, revealing violations of traditional thermodynamic uncertainty relations and supporting newer generalized bounds linked to information theory.

Contribution

It provides the first experimental demonstration of TUR violation in an information engine and validates generalized TUR bounds derived from fluctuation theorems and information measures.

Findings

Engine violates original TUR near maximal efficiency

Engine satisfies generalized TUR bounds involving mutual information

Dissipation is minimized at finite noise where TUR is violated

Abstract

Thermodynamic uncertainty relations (TURs) set fundamental bounds on the fluctuation and dissipation of stochastic systems. Here, we examine these bounds, in experiment and theory, by exploring the entire phase space of a cyclic information engine operating in a non-equilibrium steady state. Close to its maximal efficiency, we find that the engine violates the original TUR. This first experimental demonstration of TUR violation agrees with recently proposed softer bounds: The engine satisfies two generalized TUR bounds derived from the detailed fluctuation theorem with feedback control and another bound linking fluctuation and dissipation to mutual information and Renyi divergence. We examine how the interplay of work fluctuation and dissipation shapes the information conversion efficiency of the engine, and find that dissipation is minimal at a finite noise level, where the original TUR is violated.