Hierarchical dynamics in large assemblies of interacting oscillators
PHYSICS LETTERS A
Physics Letters A 160(1991) 227—232 North-Holland
Hierarchical dynamics in large assemblies of interacting oscillators Erik D. Lum...
Hierarchical dynamics in large assemblies of interacting oscillators Erik D. Lumer and Bernardo A. Huberman Department ofApplied Physics, Stanford University, Stanford, CA 94305, USA and Xerox Palo AltoResearch Center, 3333 Coyote HillRoad, Palo Alto, CA 94304, USA Received 9 August 1991; accepted for publication 16 September 1991 Communicated A.R. Bishop
We study a collection of phase-coupled oscillators possessing a hierarchical coupling structure. We establish a necessary condition for the existence ofa phase transition to collective synchrony for finite values ofthe coupling strength in terms of an inequality involving the connectivity between clusters of oscillators, the rate at which coupling strengths decrease with ultrametric distance, and the dispersion of intrinsic frequencies. When the inequality is not satisfied, there is a cascade of discrete transitions to intracluster synchrony as the coupling strength is increased, but no global synchronization is possible in the infinite size limit.
Progress in nonlinear dynamics is leading to an increasing understanding of systems with many degrees of freedom [11. These range from the dynamics of laser arrays [21, and lattices of logistic maps 3], to turbulent fluids [4]. Of particular interest are phenomena of collective synchronization taking place in large assemblies of coupled oscillators [5— 11]. In the biological realm, such regimes characterize situations such as the flickering of swarms of fireflies, and the peristaltic motion of gastrointestinal tracts [5]. Furthermore, recent experimental results in the area of neuroscience lead to the consideration of synchronization in large assemblies of coupled neural oscillators as a possible solution to the so-called binding problem [12—15]. Since most of the models of collective synchronization studied so far considered oscillators interacting via uniform fields, it is of interest to extend them to the case where the coupling strengths vary with distance and are distributed over a hierarchy of values, a situation that seems to be more realistic for heterogeneous systems such as those encountered in the study ofcomplex physical [16,17] and biological structures. This is the purpose of this paper, in which we elucidate the conditions under which collective rythmic activity can exist in a system with a hierarchy of coupling constants. Specifically, we introduce a model of phase-cou-
pled oscillators organized into clusters of coupling strengths, and such that intracluster interactions are weaker than intercluster ones. An ultrametric measure is then introduced that quantifies the interaction distance between any two oscillators in the clustured structure. A similar metric has been used in the past to study the transport properties of hierarchical systems [18]. By varying the shape of the embedding tree of interactions and the functional dependence between coupling strength and ultrametric distance, we are able to consider different classes of interactions in a systematic way. We also show the existence of a phase transition to collectivesynchrony that depends upon the validity of an inequality involving the connectivity between clusters, the rate at which coupling strengths decrease with ultrametric distance, and the dispersion of intrinsic frequencies. By collective synchrony we mean a situation whereby a macroscopic number of oscillators evolve in time with the same instantaneous frequency. These results are supported by extensive numerical simulations. Consider a collection of oscillators with coupling strengths that are weak compared to the forces attracting the oscillators to their limit cycle. Once relaxed to its limit cycle, an oscillator can be described by a single degree offreedom, that is its phase on the cycle. In the limit of weak interactions, a simple
model of the dynamics for the phases consists in the following equation [6,8,9,11], dO
ultrametric distance
(1)
where O~represents the phase of the jth oscillator in a population of N elements and ~ its intrinsic frequency. The second term on the right hand side of eq. (1) describes the interactions ofthejth oscillator with some or all ofthe N— 1 others. Accordingly, the coupling between oscillators i and j is proportional to a coupling strength, K,,, and to a periodic function of their phase difference, h (z~9).A convenient assumption which is often made is that of symmetric interactions, which translates in the symmetry of K~ and the antisymmetry of the coupling function for the permutation of the indices i and j. Furthermore, we will allow the intrinsic frequency of the oscillators, w~,to vary from one element to another and we therefore characterize the distribution of its values by a density function f( a). In the above model, the matrix of coupling strength K: {K,~}defines the pattern of interactions among oscillators. Two special and rather extreme connectivity patterns have been thoroughly studied in the past. The first type, a fully interconnected population of infinite size, corresponds to a matrix K whose nondiagonal components are all equal. Such population was shown by Kuramoto [6] to exhibit a phase transition to global synchronism at a critical value K~of the uniform coupling strengths. The second type, a regular lattice of oscillators with nearest neighbor coupling, was studied by Daido [9], who clarified the lattice dimensionality dependence ofthe onset of collective synchronism. In order to study the case of a hierarchy of coupling strengths, it is useful to introduce a visual representation of the structure of interactions between oscillators as done in fig. 1. The oscillators are represented as the leaves of a tree so that an ultrametric distance, l,~,between any pair (i, I) of oscillators can be defined as the level of their nearest common ancestor node. The ancestor node also defines the cluster of oscillators at the bottom of its branching subtree that are within an ultrametric distance 1~,of each other. The coupling strength between i and f is set to be a decreasing function of l~,and is expressed as 228
I~=3
L=2
L=1
spatial distance Fig. 1. Hierarchy ofcoupling strengths. The oscillators are at the leaves of the tree. The ultrametric distance between two oscillators is given by the level of their nearest common ancestor in the tree. For instance, the distance between i and j in fig. 1 is equal to 2. Their ancestor node also defines the cluster of oscillators branching out (in bold in the figure).
Ku = Kd( lii) (2) The profile of coupling strengths, d(1), defines the range ofmacroscopic behaviors exhibited as the control parameter K ofthe dynamics is driven across its critical value(s). The other relevant parameters of the model are the branching ratio b of the embedding tree, its depth L ~‘, and the specific form of the distribution of intrinsic frequencies f( a,). Concerning the latter, Daido [9] showed that the states of collective Oscillations depend crucially on the asymptotic behavior of f( w) in the limit of large deviations from the average th. Without loss of generality, one may write in this case .