Gleaning Information from Noise
Noise comes in many forms. A microscopic bead twitches in an optical trap; voltage fluctuations flicker through a circuit. But it’s not only a nuisance. Since 1966, physicists have understood that for systems in thermal equilibrium, such randomness also gives valuable information: Spontaneous fluctuations and the system’s response to external perturbations are locked together, frequency by frequency, according to the so-called fluctuation–dissipation theorem (FDT) [1]. That link is the basis of noise-based thermometry, microrheology, and many calibration methods. But thermal equilibrium is rare in the real world. Rather, most physical and biological systems are driven by an external force, fed, or alive, with energy continually flowing through them. That means the equality between fluctuations and response ceases to hold, and the question becomes: What can noise still tell us about the system’s response to perturbations? Andreas Dechant at Kyoto University in Japan has now derived an inequality that describes this response: Even far from equilibrium, the response to any perturbation frequency is constrained by the measured noise and by the properties of the environment [2].
The physical reason why the FDT works for systems in equilibrium is simple. A microscopic bead suspended in a thermal bath of water provides a good example. Water molecules collide constantly with the bead, causing it randomly to jitter. Those same collisions also induce drag on the bead. When an external force pushes the bead, the bombarding water molecules resist its motion. Fluctuation and friction thus share a single microscopic source: the thermal bath. Because both originate from the same mechanism, knowledge of one constrains the other. Measure the noise spectrum of the bead, and you can predict its mechanical response. Conversely, applying a small force and watching how the bead drifts provides direct information about its thermal fluctuations [1]. This reciprocity has made the FDT one of the most useful tools in statistical physics.
Once a system is driven out of equilibrium, additional energy sources inject fluctuations that the thermal bath alone cannot account for (Fig. 1). The tidy link between noise and response breaks. Efforts to repair it have followed several paths. Formal extensions express the nonequilibrium response as a correlation involving variables derived from the stochastic dynamics [3, 4]. But these variables are not directly accessible in experiments. An alternative approach rescues the form of the FDT by defining an effective temperature. But that temperature depends on the type of measurement, so this approach is not broadly applicable [5]. In 2020, Dechant and Shin-ichi Sasa of Kyoto University took a different route. Rather than seeking a new equality, they proved an inequality that established an upper bound on the amplitude of the response [6]. Their bound, however, applied only to static systems—that is, ones subjected to constant, zero-frequency perturbations. A finite-frequency version, based entirely on measurable quantities, has been missing until now.
Dechant’s new finite-frequency fluctuation-response inequality (FRI) fills that gap. For general Markovian systems—those that lack a memory of their past actions—their linear response at each frequency is bounded by a ceiling that depends on the fluctuation spectrum, the bath’s temperature, the damping, and the perturbation direction [2]. All these parameters are accessible in experiments.
It’s worth noting that the new FRI and the long-standing FDT address different questions. The FDT is an equilibrium equality between fluctuations and response [1]. The FRI is an inequality, valid in and out of equilibrium, that defines the largest response compatible with a measured noise level [2]. But the FRI does not converge to the FDT even in equilibrium. The inequality is saturated—it turns into an equality—if the dynamics are linear. In that case, the equality holds whether the system is in equilibrium or not [2].
This saturation gives linear dynamics—a bead in a harmonic trap, coupled oscillators, or a resistor–capacitor circuit—a significance beyond analytical convenience. The degree of saturation is quantified by the response efficiency—a dimensionless number between 0 and 1, specified at each frequency [2]. For linear systems such as a spring network, the response efficiency equals 1 at all frequencies. For nonlinear systems it falls below 1, and the shortfall measures how far the dynamics depart from linearity. Since fluctuation spectra, response functions, temperature, and damping are standard observables in many laboratories, the response efficiency is a diagnostic that can be determined immediately—no model fitting is required.
Dechant illustrates the point with a model of a particle’s periodic motion over a set of repeated barriers—a potential resembling a tilted washboard [2]. Without tilt, the particle rattles within one well. Fluctuation and response both peak near the oscillation frequency. When the landscape is tilted, the particle begins slipping from well to well, adding a low-frequency feature. Driving the system reshuffles where the response lives in the frequency domain but does not cause the ceiling to move. Integrating the bound over all frequencies yields a universal cap on the broadband signal-to-noise ratio, determined by the bath alone. In other words, the total sensing budget is fixed.
This result fits into a longer arc in statistical physics. Observations of Brownian motion first transformed randomness into a window onto invisible molecular agitation. That intuition was then turned into the exact FDT equilibrium relationship [1]. Dechant has now extended the story in a different direction, showing that far-from-equilibrium noise still carries a useful message, not as an equality but as a universal limit [2].
The outlook is broad. Extending the theorem beyond Markovian dynamics would bring memory-rich materials, viscoelastic media, and living matter into the realm of systems for which measured fluctuations offer a valuable probe. Open quantum systems pose another frontier [7]. Just as importantly, response efficiency could become a practical tool for comparing sensors, circuits, and active materials by how closely they approach the thermal ceiling set by their own noise. Noise began as a nuisance. Now it is becoming a calibrated guide to what nonequilibrium devices can and cannot do.
References
- R. Kubo, “The fluctuation-dissipation theorem,” Rep. Prog. Phys. 29, 255 (1966).
- A. Dechant, “Finite-frequency fluctuation-response inequality,” Phys. Rev. Lett. 136, 207101 (2026).
- M. Baiesi et al., “Fluctuations and response of nonequilibrium states,” Phys. Rev. Lett. 103, 010602 (2009).
- U. Seifert and T. Speck, “Fluctuation-dissipation theorem in nonequilibrium steady states,” Europhys. Lett. 89, 10007 (2010).
- L. F. Cugliandolo, “The effective temperature,” J. Phys. A: Math. Theor. 44, 483001 (2011).
- A. Dechant and S.-i. Sasa, “Fluctuation–response inequality out of equilibrium,” Proc. Natl. Acad. Sci. U.S.A. 117, 6430 (2020).
- T. Van Vu, “Fundamental bounds on precision and response for quantum trajectory observables,” PRX Quantum 6, 010343 (2025).