Fundamental limits on nonequilibrium sensing

The performance of equilibrium sensors is restricted by the laws of equilibrium thermodynamics. Here, we investigate the physical limits on nonequilibrium sensing in bipartite systems with both reciprocal and nonreciprocal couplings. We show that one of the subsystems, acting as a Maxwell demon, can significantly suppress the fluctuations of the other subsystem relative to its response to an external perturbation. The importance of nonreciprocal interactions for such negative violations of the fluctuation-dissipation relation to occur is identified. We further demonstrate that these violations can considerably improve the signal-to-noise ratio above its corresponding equilibrium value, allowing the subsystem to operate as an enhanced sensor. In addition, we find that the nonequilibrium signal-to-noise ratio of linear systems may be arbitrarily large at low frequencies after proper parameter optimization, even at a fixed overall amount of dissipation. These results indicate that highly accurate nonreciprocal sensors can be designed at a finite energetic cost.