The size of the sync basin Available to Purchase

We suggest a new line of research that we hope will appeal to the nonlinear dynamics community, especially the readers of this Focus Issue. Consider a network of identical oscillators. Suppose the synchronous state is locally stable but not globally stable; it competes with other attractors for the available phase space. How likely is the system to synchronize, starting from a random initial condition? And how does the probability of synchronization depend on the way the network is connected? On the one hand, such questions are inherently difficult because they require calculation of a global geometric quantity, the size of the “sync basin” (or, more formally, the measure of the basin of attraction for the synchronous state). On the other hand, these questions are wide open, important in many real-world settings, and approachable by numerical experiments on various combinations of dynamical systems and network topologies. To give a case study in this direction, we report results on the sync basin for a ring of $n⪢1$ identical phase oscillators with sinusoidal coupling. Each oscillator interacts equally with its $k$ nearest neighbors on either side. For $k∕n$ greater than a critical value (approximately 0.34, obtained analytically), we show that the sync basin is the whole phase space, except for a set of measure zero. As $k∕n$ passes below this critical value, coexisting attractors are born in a well-defined sequence. These take the form of uniformly twisted waves, each characterized by an integer winding number $q$⁠, the number of complete phase twists in one circuit around the ring. The maximum stable twist is proportional to $n∕k$⁠; the constant of proportionality is also obtained analytically. For large values of $n∕k$⁠, corresponding to large rings or short-range coupling, many different twisted states compete for their share of phase space. Our simulations reveal that their basin sizes obey a tantalizingly simple statistical law: the probability that the final state has $q$ twists follows a Gaussian distribution with respect to $q$⁠. Furthermore, as $n∕k$ increases, the standard deviation of this distribution grows linearly with $n∕k$⁠. We have been unable to explain either of these last two results by anything beyond a hand-waving argument.