Activated switching between coexisting limit cycles

In the natural world, many systems settle into a state of rest. A ball in a valley, a molecule folded into a specific shape, or a chemical reaction that has run its course: these are states of equilibrium where the system stays put until a rare, large jolt pushes it over a hill and into a new valley. Scientists have long understood how this "noise-activated" switching works for stationary states, using the familiar image of a ball rolling over an energy barrier. However, many important systems do not rest at all. Instead, they settle into a continuous, repeating motion, like a heartbeat or the rhythmic firing of neurons. These self-sustaining loops are called limit cycles. Because they are constantly moving, they cannot be described by a static landscape of hills and valleys. For decades, it remained a mystery how a system trapped in one of these moving loops could be nudged by random fluctuations to jump to a different, coexisting loop. Without a static hill to climb, the rules for how such a switch happens were unknown.

A team of researchers at ETH Zurich and the University of Konstanz has now brought this elusive phenomenon into the light. They built a physical system using two coupled electrical resonators—essentially circuits that vibrate at a specific frequency—and carefully tuned them to create two distinct, stable loops of motion that could exist at the same time. By introducing controlled electrical noise, they were able to watch these loops switch back and forth, a rare event that happens only when a fluctuation is strong enough to push the system out of its current path. Their experiments revealed that while the switching still follows a pattern similar to the old "ball over a hill" idea, the mechanism is fundamentally different. Instead of climbing a single, fixed peak, the system travels along a specific, narrow path through its state space. The likelihood of the switch depends on the length and shape of this path, which changes depending on how hard the system is being driven.

The researchers constructed their experiment using two nearly identical electrical circuits, each containing a resistor, an inductor, and a capacitor. They made these circuits non-linear by adding a special diode, which allowed the circuits to interact in complex ways. They drove one of the circuits with a specific electrical signal, which caused the system to settle into one of two possible rhythmic states. These states were not static; the voltages in the circuits were constantly oscillating, tracing out closed loops in a mathematical space that describes their motion. Under normal conditions, the system would stay in one loop forever. To test how it might switch, the team injected random electrical noise into both circuits. This noise acted as a series of tiny, unpredictable kicks. Over time, these kicks occasionally provided enough energy to push the system out of its current loop and into the other one.

By measuring the voltage in the circuits over time, the team observed these rare jumps clearly. They found that the rate at which the system switched depended exponentially on the strength of the noise, a relationship that mirrors the behavior of stationary systems. However, the way the switching rate changed with the strength of the driving signal was surprising. Intuitively, one might expect that driving the system harder would make the loops larger and more stable, making it harder to switch. Instead, the researchers found that increasing the drive actually made switching more frequent. This counterintuitive result occurred because the stronger drive distorted the shape of the loops, pulling them closer together in the mathematical space. This deformation shortened the narrow channel through which the system had to pass to switch, making the jump easier despite the larger size of the loops.

To understand this behavior, the team turned to a mathematical framework known as large-deviation theory. This approach allows scientists to calculate the probability of rare events by finding the "most probable path" a system takes when it escapes a stable state. For stationary systems, this path is a straight line over a barrier. For these moving limit cycles, the path is an extended route through the phase space. The researchers calculated this path and found that the "cost" of the switch, which determines how likely it is, is not a fixed height but an action value accumulated along this specific trajectory. Their theoretical calculations matched the experimental data perfectly, confirming that the switching is governed by the geometry of the path rather than a simple energy barrier.

The study also revealed that the system's behavior is highly sensitive to the precise conditions of the drive. As the researchers adjusted the driving strength, they observed the loops changing shape and, in some cases, undergoing subtle structural changes that appeared as small jumps in the theoretical predictions. While these fine details were too small to see clearly in the current experiment due to noise, the agreement between the broad trends of the data and the theory provides a solid foundation for understanding these dynamics. The work demonstrates that the concept of activated switching, previously limited to systems at rest, can be successfully extended to systems in constant motion.

This discovery has implications for how we model complex systems in nature and technology. Many biological and engineered systems, from neural networks to chemical reactors, operate in rhythmic, non-stationary states. Understanding how these systems switch between different modes of operation could lead to better control strategies for complex machines. The researchers suggest that their findings could eventually help in designing new types of computing devices, such as machines that solve optimization problems by switching between different rhythmic states rather than static ones. By treating these moving loops as addressable states, similar to the static states used in current computing models, it may be possible to create more efficient and flexible systems. The ability to control these transitions with external drives offers a new way to tune the behavior of such systems, potentially leading to improved performance in tasks that require rapid and reliable switching.

The experiment was conducted with high precision, using electrical resonators with a frequency of approximately 3.2 million cycles per second. The team measured the switching rates over periods of 30 seconds, collecting tens of thousands of data points for each setting of noise and drive strength. They verified their results by comparing the experimental data with numerical simulations that used the exact parameters of their physical setup. The simulations reproduced the observed spectral features and the dependence of the switching rate on the drive strength, confirming that the underlying physics was correctly captured. The researchers also noted that the noise they introduced was independent for each resonator, ensuring that the switching was truly a result of the system's internal dynamics interacting with external fluctuations.

In summary, the researchers have provided the first clear experimental demonstration of noise-activated switching between two coexisting limit cycles. They showed that while the process shares some similarities with the switching of stationary states, it is governed by a different set of rules that depend on the geometry of the moving paths. The findings extend our understanding of stochastic dynamics to a broader class of systems and open the door to new ways of controlling and utilizing rhythmic behaviors in driven-dissipative systems. The work stands as a bridge between the well-understood world of static equilibria and the complex, dynamic world of self-sustained oscillations, offering a new framework for analyzing how systems transition between different modes of operation.