Search

We study a system of $N$ interacting particles moving on the unit sphere in $d$-dimensional space. The particles are self-propelled and coupled all to all, and their motion is heavily overdamped. For $d=2$, the system reduces to the classic Kuramoto model of coupled oscillators; for $d=3$, it has been proposed to describe the orientation dynamics of swarms of drones or other entities moving about in three-dimensional space. Here we use group theory to explain the recent discovery that the model shows low-dimensional dynamics for all $N \ge 3$, and to clarify why it admits the analog of the Ott-Antonsen ansatz in the continuum limit $N \rightarrow \infty$. The underlying reason is that the system is intimately connected to the natural hyperbolic geometry on the unit ball $B^d$. In this geometry, the isometries form a Lie group consisting of higher-dimensional generalizations of the M\"obius transformations used in complex analysis. Once these connections are realized, the reduced dynamics and the generalized Ott-Antonsen ansatz follow immediately. This framework also reveals the seamless connection between the finite and infinite-$N$ cases. Finally, we show that special forms of coupling yield gradient dynamics with respect to the hyperbolic metric, and use that fact to obtain global stability results about convergence to the synchronized state.

Kuramoto oscillator networks have the special property that their trajectories are constrained to lie on the (at most) 3D orbits of the M\"obius group acting on the state space $T^N$ (the $N$-fold torus). This result has been used to explain the existence of the $N-3$ constants of motion discovered by Watanabe and Strogatz for Kuramoto oscillator networks. In this work we investigate geometric consequences of this M\"obius group action. The dynamics of Kuramoto phase models can be further reduced to 2D reduced group orbits, which have a natural geometry equivalent to the unit disk $\Delta$ with the hyperbolic metric. We show that in this metric the original Kuramoto phase model (with order parameter $Z_1$ equal to the centroid of the oscillator configuration of points on the unit circle) is a gradient flow and the model with order parameter $iZ_1$ (corresponding to cosine phase coupling) is a completely integrable Hamiltonian flow. We give necessary and sufficient conditions for general Kuramoto phase models to be gradient or Hamiltonian flows in this metric. This allows us to identify several new infinite families of hyperbolic gradient or Hamiltonian Kuramoto oscillator networks which therefore have simple dynamics with respect to this geometry. We prove that for the $Z_1$ model, a generic 2D reduced group orbit has a unique fixed point corresponding to the hyperbolic barycenter of the oscillator configuration, and therefore the dynamics are equivalent on different generic reduced group orbits. This is not always the case for more general hyperbolic gradient or Hamiltonian flows; the reduced group orbits may have multiple fixed points, which also may bifurcate as the reduced group orbits vary., Comment: Accepted for publication in Journal of Physics A: Mathematical and Theoretical

We study a network of N identical leaky integrate-and-fire model neurons coupled by α-function pulses, weighted by a coupling parameter K. Studies of the dynamics of this system have mostly focused on the stability of the fully synchronized and the fully asynchronous splay states, which naturally depends on the sign of K, i.e., excitation vs inhibition. We find that there is also a rich set of attractors consisting of clusters of fully synchronized oscillators, such as fixed (N-1,1) states, which have synchronized clusters of sizes N-1 and 1, as well as splay states of clusters with equal sizes greater than 1. Additionally, we find limit cycles that clarify the stability of previously observed quasiperiodic behavior. Our framework exploits the neutrality of the dynamics for K=0 which allows us to implement a dimensional reduction strategy that simplifies the dynamics to a continuous flow on a codimension 3 subspace with the sign of K determining the flow direction. This reduction framework naturally incorporates a hierarchy of partially synchronized subspaces in which the new attracting states lie. Using high-precision numerical simulations, we describe completely the sequence of bifurcations and the stability of all fixed points and limit cycles for N=2-4. The set of possible attracting states can be used to distinguish different classes of neuron models. For instance from our previous work [Chaos 24, 013114 (2014)CHAOEH1054-150010.1063/1.4858458] we know that of the types of partially synchronized states discussed here, only the (N-1,1) states can be stable in systems of identical coupled sinusoidal (i.e., Kuramoto type) oscillators, such as θ-neuron models. Upon introducing a small variation in individual neuron parameters, the attracting fixed points we discuss here generalize to equivalent fixed points in which neurons need not fire coincidently.

We study the dynamics of a generalized version of the famous Kuramoto-Sakaguchi coupled oscillator model. In the classic version of this system, all oscillators are governed by the same ordinary differential equation (ODE), which depends on the order parameter of the oscillator configuration. The order parameter is the arithmetic mean of the configuration of complex oscillator phases, multiplied by some constant complex coupling factor. In the generalized model, we consider that all oscillators are still governed by the same ODE, but the order parameter is allowed to be any complex linear combination of the complex oscillator phases, so the oscillators are no longer necessarily weighted identically in the order parameter. This asymmetric version of the K-S model exhibits a much richer variety of steady-state dynamical behavior than the classic symmetric version; in addition to stable synchronized states, the system may possess multiple stable (N-1,1) states, in which all but one of the oscillators are synchronized, as well as multiple families of neutrally stable states or closed orbits, in which no two oscillators are synchronized. We present an exhaustive description of the possible steady state dynamical behaviors; our classification depends on the complex coefficients that determine the order parameter. We use techniques from group theory and hyperbolic geometry to reduce the dynamic analysis to a 2D flow on the unit disc, which has geometric significance relative to the hyperbolic metric. The geometric-analytic techniques we develop can in turn be applied to study even more general versions of Kuramoto oscillator networks.

We solve a longstanding stability problem for the Kuramoto model of coupled oscillators. This system has attracted mathematical attention, in part because of its applications in fields ranging from neuroscience to condensed-matter physics, and also because it provides a beautiful connection between nonlinear dynamics and statistical mechanics. The model consists of a large population of phase oscillators with all-to-all sinusoidal coupling. The oscillators' intrinsic frequencies are randomly distributed across the population according to a prescribed probability density, here taken to be unimodal and symmetric about its mean. As the coupling between the oscillators is increased, the system spontaneously synchronizes: the oscillators near the center of the frequency distribution lock their phases together and run at the same frequency, while those in the tails remain unlocked and drift at different frequencies. Although this ``partially locked'' state has been observed in simulations for decades, its stability has never been analyzed mathematically. Part of the difficulty is in formulating a reasonable infinite-N limit of the model. Here we describe such a continuum limit, and prove that the corresponding partially locked state is, in fact, neutrally stable, contrary to what one might have expected. The possible implications of this result are discussed.

We analyze the linear stability of the phase-locked state in the Kuramoto model of coupled oscillators. The main result is the first rigorous characterization of the spectrum and its associated eigenvectors, for any finite number of oscillators. All but two of the eigenvalues are negative, and merge into a continuous spectrum as the number of oscillators tends to infinity. One eigenvalue is always zero, by rotational invariance. The final eigenvalue, corresponding to a collective mode, determines the stability of the locked state.

We present a complete classification of attractors for networks of coupled identical Kuramoto oscillators. In such networks, each oscillator is driven by the same first-order trigonometric function, with coefficients given by symmetric functions of the entire oscillator ensemble. For [Formula: see text] oscillators, there are four possible types of attractors: completely synchronized fixed points or limit cycles, and fixed points or limit cycles where all but one of the oscillators are synchronized. The case N = 3 is exceptional; systems of three identical Kuramoto oscillators can also posses attracting fixed points or limit cycles with all three oscillators out of sync, as well as chaotic attractors. Our results rely heavily on the invariance of the flow for such systems under the action of the three-dimensional group of Möbius transformations, which preserve the unit disc, and the analysis of the possible limiting configurations for this group action.

The Kuramoto model, first proposed in 1975, consists of a population of sinusoidally coupled oscillators with random natural frequencies. It has served as an idealized model for coupled oscillator systems in physics, chemistry, and biology. This paper addresses a long-standing problem about the infinite-N Kuramoto model, which is to describe the asymptotic behavior of the order parameter for this system. For coupling below a critical level, Kuramoto predicted that the order parameter would decay to 0. We use Fourier transform methods to prove that for general initial conditions, this decay is not exponential; in fact, exponential decay to 0 can only occur if the initial condition satisfies a fairly strong regularity condition that we describe. Our theorem is a partial converse to the recent results of Ott and Antonsen, who proved that for a special class of initial conditions, the order parameter does converge exponentially to its limiting value. Consequently, our result shows that the Ott-Antonsen ansatz does not completely capture all the possible asymptotic behavior in the full Kuramoto system.

We study the long-term average frequency as a function of the natural frequency for Kuramoto oscillators with periodic coefficients. Unlike the case for more general periodically forced oscillators, this function is never a ``devil's staircase''; it may have plateaus at integer multiples of the forcing frequency, but we prove it is strictly increasing between these plateaus. The proof uses the fact that the flow maps for Kuramoto oscillators extend to M\"obius transformations on the complex plane, and that M\"obius transformations have particularly simple dynamics that rule out $p\ensuremath{\mathbin:}q$ mode locking except in the case of fixed points ($q=1$). We also give a criterion for the degeneration of an integer plateau to a single point and use it to explain the absence of plateaus at even multiples of the collective frequency for a Kuramoto system with a bimodal frequency distribution.

This paper studies dynamical systems on the n-fold torus equivariant under a cyclic permutation of coordinates. It is proved that under a mild condition, these systems have splay-phase solutions. These are periodic orbits in which the n coordinates are given by the same function of time, but equally separated in phase. Applications to systems of equations used to model Josephson junction arrays are discussed.

We explore patterns in the spike timing of neurons receiving periodic inputs, with an emphasis on stable characteristics which are realized in both models and in-vitro whole-cell recordings. We report on whole-cell recordings of pyramidal CA1 cells from rat hippocampus and entorhinal cortex and compare this data to model simulations. Cells were injected with a constant current to induce a steady firing rate and then a modest rhythm was added which altered the spike times and their corresponding phases relative to the rhythm. For both experiment and theory the relationship between consecutive spike phases is characterized by a probability distribution with peaks concentrated near a one-dimensional firing map. As is well-known, stable fixed points of this map correspond to the neuron phase-locking to the rhythm. We show that the interaction between noise and sufficiently steep maps can also cause a new kind of spike-time organization, in which consecutive spike time pairs organize into discrete clusters, with transitions between these clusters proceeding in a fixed sequence. This structure is not just a vestige of the noise-free dynamics. This slow dynamics and temporal organization in the relationship between consecutive spike phases is not evident in either the neuron's voltage traces or single phase or interspike interval histograms. Furthermore, the consecutive spike relationship is also evident in consecutive ISIs, and hence this ordering can be observed without detailed knowledge of the rhythm (e.g. without concurrent LFP recordings).

Systems of N identical phase oscillators with global sinusoidal coupling are known to display low-dimensional dynamics. Although this phenomenon was first observed about 20 years ago, its underlying cause has remained a puzzle. Here we expose the structure working behind the scenes of these systems, by proving that the governing equations are generated by the action of the Mobius group, a three-parameter subgroup of fractional linear transformations that map the unit disc to itself. When there are no auxiliary state variables, the group action partitions the N-dimensional state space into three-dimensional invariant manifolds (the group orbits). The N-3 constants of motion associated with this foliation are the N-3 functionally independent cross ratios of the oscillator phases. No further reduction is possible, in general; numerical experiments on models of Josephson junction arrays suggest that the invariant manifolds often contain three-dimensional regions of neutrally stable chaos.

We consider the interaction of a large number of limit-cycle oscillators with linear, all-to-all coupling and a distribution of natural frequencies. The system exhibits extremely rich dynamics as the coupling strength and the width of the frequency distribution are varied. We find a variety of steady behaviors that can be described by a stationary distribution in phase space: frequency locking, amplitude death, incoherence and partial locking. An unexpected result is that the system can also exhibit unsteady behavior, in which the phase space distribution evolves periodically, quasiperiodically or even chaotically. The simple form of the model allows us to derive several analytical results. The stability boundaries of amplitude death and incoherence are found explicitly. Rigorous results on the existence and stability of frequency locking are also obtained.

We analyze a mean-field model of coupled oscillators with randomly distributed frequencies. This system is known to exhibit a transition to collective oscillations: for small coupling, the system is incoherent, with all the oscillators running at their natural frequencies, but when the coupling exceeds a certain threshold, the system spontaneously synchronizes. We obtain the first rigorous stability results for this model by linearizing the Fokker-Planck equation about the incoherent state. An unexpected result is that the system has pathological stability properties: the incoherent state is unstable above threshold, butneutrally stable below threshold. We also show that the system is singular in the sense that its stability properties are radically altered by infinitesimal noise.

We predict that series arrays of point-contact Josephson junctions, shunted by a resistive load, should be unusually sensitive to noise. This behavior stems from a dynamical symmetry which prohibits the ex- istence of in-phase attractors. Simulations verify that breaking the symmetry can radically improve the phase-locking performance of the array.

We study a set of N globally coupled ordinary differential equations of the form encountered in circuit analysis of superconducting Josephson junction arrays. Particular attention is paid to two kinds of simple time-periodic behavior, known as in-phase and splay phase states. Some results valid for general N , as well as further results for N = 2 and N → ∞, are presented; a recurring theme is the appearance of very weak dynamics near the periodic states. The implications for Josephson junction arrays are discussed.

Transitions between dynamical states in integrate-and-fire neuron models with periodic stimuli result from tangent or discontinuous bifurcations of a return map. We study their characteristic scaling laws and show that discontinuous bifurcations exhibit a kind of phase transition intermediate between continuous and first order. In the model-independent spirit of our analysis we show that a six-dimensional (6D) gating variable model with an attracting limit cycle has similar phase transitions, governed by a 1D return map. This reduction to 1D map dynamics should extend to real neurons in a periodic current clamp setting.

Networks of identical, symmetrically coupled oscillators can spontaneously split into synchronized and desynchronized sub-populations. Such chimera states were discovered in 2002, but are not well understood theoretically. Here we obtain the first exact results about the stability, dynamics, and bifurcations of chimera states by analyzing a minimal model consisting of two interacting populations of oscillators. Along with a completely synchronous state, the system displays stable chimeras, breathing chimeras, and saddle-node, Hopf and homoclinic bifurcations of chimeras., 4 pages, 4 figures. This version corrects a previous error in Figure 3, where the sign of the phase angle psi was inconsistent with Equation 12

A simple model for synchronous firing of biological oscillators based on Peskin's model of the cardiac pacemaker (Mathematical aspects of heart physiology, Courant Institute of Mathematical Sciences, New York University, New York, 1975, pp. 268-278) is studied. The model consists of a population of identical integrate-and-fire oscillators. The coupling between oscillators is pulsatile: when a given oscillator fires, it pulls the others up by a fixed amount, or brings them to the firing threshold, whichever is less. The main result is that for almost all initial conditions, the population evolves to a state in which all the oscillators are firing synchronously. The relationship between the model and real communities of biological oscillators is discussed; examples include populations of synchronously flashing fireflies, crickets that chirp in unison, electrically synchronous pacemaker cells, and groups of women whose menstrual cycles become mutually synchronized.

This paper studies an infinite-dimensional dynamical system that models a collection of coupled spins in a random magnetic field. A state of the system is given by a self-map of the unit circle. Critical states for the dynamical system correspond to equilibrium configurations of the spins. Exact solutions are obtained for all the critical states and their stability is characterized by analyzing the second variation of the system’s potential function at those states. It is proven rigorously that the system exhibits a jump bifurcation and hysteresis as the coupling between spins is varied.

We investigate extensions to the model put forward by Brody and Hopfield [1] for spike-timing based pattern recognition applied to mammalian olfaction. Their model implements a pattern recognition algorithm realized in the dynamics of a network of coupled IF neurons subject to a sine-wave rhythm. Subsets of these neurons can synchronize through the principle of one-to-one mode locking. Their network represents 3 layers of neural activity, the first two of which are inspired by the connectivity of glomeruli and mitral cells in mammalian olfactory circuits and the gamma-rhythm activity observed in the olfactory bulb. Specifically in this model a pattern of glomerular activity representing a given odor causes a particular subset of model mitral cells to synchronize and this synchronous activity can drive a "grandmother" model cortical cell through threshold triggering a recognition event. In this study we quantify the performance of their original model and compare it to our extensions of the model such as using a network-generated rhythm rather than a sine-wave, introducing inhibitive feedback and generalizing to p-q mode locking strategies. We compute the scaling with respect to the number of mitral neurons of a measure of the number of odor patterns the model can recognize. Quite remarkably we find this performance can increase very fast with increasing network size -- consistent with exponential scaling.

We solve a long-standing stability problem for the Kuramoto model of coupled oscillators. This system has attracted mathematical attention, in part because of its applications in fields ranging from neuroscience to condensed-matter physics, and also because it provides a beautiful connection between nonlinear dynamics and statistical mechanics. The model consists of a large population of phase oscillators with all-to-all sinusoidal coupling. The oscillators' intrinsic frequencies are randomly distributed across the population according to a prescribed probability density, here taken to be unimodal and symmetric about its mean. As the coupling between the oscillators is increased, the system spontaneously synchronizes: The oscillators near the center of the frequency distribution lock their phases together and run at the same frequency, while those in the tails remain unlocked and drift at different frequencies. Although this "partially locked" state has been observed in simulations for decades, its stability has never been analyzed mathematically. Part of the difficulty is in formulating a reasonable infinite-N limit of the model. Here we describe such a continuum limit, and prove that the corresponding partially locked state is, in fact, neutrally stable, contrary to what one might have expected. The possible implications of this result are discussed.

In recent numerical experiments on series arrays of overdamped Josephson junctions, Nichols and Wiesenfeld [Phys. Rev. A 45, 8430 (1992)] discovered that the periodic states known as splay states are neutrally stable in all but four directions in phase space. We present a theory that accounts for this enormous degree of neutral stability. The theory also predicts the four non-neutral Floquet multipliers to within 0.1% of their numerically computed values. The analytical approach used here may be appli- cable to other globally coupled systems of oscillators, such as multimode lasers, electronic oscillator circuits, and solid-state laser arrays.

We analyze a model of globally coupled nonlinear oscillators with randomly distributed frequencies. Twenty-five years ago it was conjectured that, for coupling strengths below a certain threshold, this system would always relax to an incoherent state. We prove this conjecture for the system linearized about the incoherent state, for frequency distributions with compact support. The relaxation is exponentially fast at intermediate times but slower than exponential at long times. The decay mechanism is remarkably similar to Landau damping in plasmas, even though the model was originally inspired by biological rhythms.

We analyze a large system of nonlinear oscillators with random pinning, mean-field coupling and external drive. For small coupling and drive strength, the system evolves to an incoherent pinned state, with all the oscillators stuck at random phases. As the coupling or drive strength is increased beyond a depinning treshold, the steady-state solution switches to a coherent moving state, with all the oscillators moving nearly in phase. This depinning transition is discontinuous and hysteretic. We also show analytically that there is a delayed onset of coherence in response to a sudden superthreshold drive. The time delay increases as the threshold is approached from above. The discontinuous, hysteretic transition and the delayed onset of coherence are directly attributable to the form of the coupling, which is periodic in the phase difference between oscillators. The system studied here provides a simple model of charge-density wave transport in certain quasi-one-dimensional metals and semiconductors in the regime where phase-slip is important; however this paper is intended primarily as a study of a model system with analytically tractable collective dynamics.

We study phase-locking in a network of coupled nonlinear oscillators with local interactions and random intrinsic frequencies. The oscillators are located at the vertices of a graph and interact along the edges. They are coupled by sinusoidal functions of the phase differences across the edges, and their intrinsic frequencies are independent and identically distributed with finite mean and variance. We derive an exact expression for the probability of phase-locking in a linear chain of such oscillators and prove that this probability tends to zero as the number of oscillators grows without bound. However, if the coupling strength increases as the square root of the number of oscillators, the probability of phase-locking tends to a limiting distribution, the Kolmogorov-Smirnov distribution. This latter result is obtained by showing that the phase-locking problem is equivalent to a discretization of pinned Brownian motion. The results on chains of oscillators are extended to more general graphs. In particular, for a hypercubic lattice of any dimension, the probability of phase-locking tends to zero exponentially fast as the number of oscillators grows without bound. We also consider a less stringent type of synchronization, characterized by large clusters of oscillators mutually entrained at the same average frequency. It is shown that if such clusters exist, they necessarily have a sponge-like geometry.

© Gauthier-Villars (Editions scientifiques et medicales Elsevier), 1987, tous droits reserves. L’acces aux archives de la revue « Annales scientifiques de l’E.N.S. » (http://www. elsevier.com/locate/ansens), implique l’accord avec les conditions generales d’utilisation (http://www.numdam.org/legal.php). Toute utilisation commerciale ou impression systematique est constitutive d’une infraction penale. Toute copie ou impression de ce fichier doit contenir la presente mention de copyright.

The authors study mutual synchronisation in a model of interacting limit cycle oscillators with random intrinsic frequencies. It is shown rigorously that the model exhibits no long-range order in one dimension, and that in higher-dimensional lattices, large clusters of synchronised oscillators necessarily have a sponge-like structure. Surprisingly, the phase-locking behaviour of the mean-field model is completely different from that of any finite-dimensional lattice, indicating that d= infinity is the upper critical dimension for phase locking.

We present a mean-field analysis of a many-body dynamical system which models charge-density-wave transport in the presence of random pinning impurities. Phase slip between charge-density-wave domains is modeled by a coupling term that is periodic in the phase differences. When driven by an external field, the system exhibits a first-order depinning transition, hysteresis, and switching between pinned and sliding states, and a delayed onset of sliding near threshold.

We develop an analytic framework to investigate the stability of splay states in infinite networks of identical integrate-and-fire neurons coupled through synaptic pulses. More specifically we perform a linear stability analysis of the splay state probability distribution whose dynamics is governed by an appropriate Fokker Planck equation. For exponentially decaying synaptic pulses the splay state is unstable for excitation and stable for inhibition. For excitatory alpha-function pulses the splay state becomes stable for sufficiently large decay times and we find an analytic expression for the boundary of stability. This large decay time stability is analogous to the stability of synchronous states for inhibition studied by van Vreeswijk, Abbott and Ermentrout [1]. For inhibitory alpha-function pulses the splay state is unstable, but for smaller decay times (when there is no stable synchronous state) the splay state exhibits a remarkable attracting meta-stable transient. We complement our analytic framework with numerical simulations on finite networks.