Stochastic resonance

In ordinary settings, noise obscures information. Stochastic resonance describes cases where an intermediate level of noise instead improves a system's response to a weak input. A common model is a bistable system with two stable states separated by a barrier. A weak periodic signal may be unable to move the system between states on its own. Random fluctuations can occasionally push it over the barrier; when the timing of these transitions tends to align with the signal, the output carries a clearer trace of that input. This requires a nonlinear system and appropriate signal and noise conditions; adding noise does not universally improve detection.

With very little noise, threshold crossings are rare and the weak signal may go unnoticed. As noise rises, transitions can occur at times that are more strongly correlated with the signal, improving measures such as signal-to-noise ratio. Beyond an optimal range, fluctuations become so strong that they overwhelm this timing relationship and degrade the response. The characteristic feature is therefore a peak in a measure of signal response as noise intensity changes. The peak's position and even the appropriate measure depend on the system and the signal; the system is not simply amplifying every input.

The term stochastic resonance was introduced in research published in 1981 to describe a proposed mechanism relevant to the recurrence of ice-age cycles. The basic idea joined a weak periodic influence, random climatic fluctuations and a threshold-like response. The climate application remains a model for how the effect can be considered, rather than proof that stochastic resonance alone explains glacial cycles. During the following decades, researchers formulated the phenomenon in nonlinear physics and explored it in other contexts. A 1998 review by Gammaitoni, Hänggi, Jung and Marchesoni surveyed theory and reports in areas including nonlinear optics, solid-state devices and neurophysiology.

Biological systems often involve thresholds, and researchers have investigated whether noise can help weak signals trigger measurable responses. Examples discussed in the literature include sensory neurons and mechanoreceptors, as well as laboratory and modeling studies of neural and other biological systems. Hänggi's 2002 review describes experiments and proposals involving sensory processing, while also noting open questions about when noise enhancement is genuinely functional in living systems. In technology, controlled patterns of noise have been examined as a way to improve detection in particular threshold-based sensors. These findings are system-specific: they do not establish that noise is generally beneficial, nor do experimental effects by themselves amount to clinical treatments.

Despite its name, stochastic resonance is not simply the familiar resonance of an oscillator driven near its natural frequency. Its defining feature is a non-monotonic dependence of signal response on noise intensity: there is an intermediate noise level at which the input is detected or represented most effectively. For periodic signals, synchronization between threshold crossings and the drive can help reveal the effect. For irregular signals, researchers may instead examine information-theoretic or correlation-based measures. Careful experiments therefore compare output quality across noise levels and rule out alternative explanations.