Autonomous demon exploiting heat and information: stochastic trajectories and current fluctuations

Nanoscale devices can exploit not only heat, like usual engines, but also resources such as information or nonthermal distributions. Since fluctuations are significant at this scale, precision—how much output power noise is suppressed—is a key metric. I present a refrigerator-type device using the thermoelectric effect and powered by a nonthermal resource. It drives heat from cold to hot without any average particle or energy flow from the resource, acting as a “demon”.

The setup consists of three capacitively coupled quantum dots: one connected to two reservoirs at different temperatures (the working substance) and two others linked to hot and cold reservoirs (the resource). It implements an autonomous demon achieving finite cooling power with zero heat extraction on average from the resource. I analyze its thermodynamic performance via stochastic trajectories and full counting statistics, showing how information flow and counter-balancing heat flows shape the trade-off between cooling power, efficiency, and precision [1]. Noise analysis reveals cooling power fluctuations can be much smaller than input noise [2], guiding future energy-conversion designs using nonthermal resources.