Physics news · January 2026

An information engine does better by harnessing sideways thermal fluctuations

Physics 27 January 2026 · Physical Review Letters

Antonio Patrón, John Bechhoefer and David Sivak analysed an information engine consisting of an overdamped Brownian bead in a controllable, multi-dimensional harmonic trap under gravity, with the trap centre updated by a measurement-based feedback protocol.

They showed that performance improves markedly when the engine harnesses thermal fluctuations in directions perpendicular to gravity, effectively feedback-cooling those transverse modes to extract gravitational potential energy and produce directed motion without external work. The study generalises the one-dimensional Szilard-engine picture and shows that a Maxwell-demon-like controller benefits from the full geometry of the fluctuations it measures.

It offers a concrete design principle for squeezing more work out of thermal noise.

Source

Physical Review Letters · further coverage.