Synchronization in Complex Systems
Transcription
Synchronization in Complex Systems
Synchronization in Complex Systems M. Biey, Politecnico di Torino, 2016 June 6, 2016 M. Biey - Department. of Electronics and Telecommunications Politecnico di Torino 1 Introduction In 2002, Barahona and Pecora [1] applied the Master Stability Function formalism to analyzing the synchronizability of small-world networks. This marks the beginning of an explosively growing area of research in synchronization in complex networks (see [2]-[9]). [1] M. Barahona, L.M. Pecora, Synchronization in Small-World Systems, Phys. Rev. Lett., Vol. 89, 054101, 2002. June 6, 2016 M. Biey - Dip. di Elettronica e Telecomunicazioni - Politecnico di Torino 2 A measure of synchronizability A relevant contribution in determining the (local) stability of the synchronized states was given in by Pecora and Carroll in 1998 [*] by using the eigenvalues of the Laplacian matrix representing the network. A brief summary: Consider a network of N identical dynamical systems with symmetric linear diffusive coupling. The equations of motion for the system are N xi f (xi ) Gij Hx j , i 1, 2, ,N (1) j 1 where f ( ) governs the dynamics of each isolated node, H( ) is a constant matrix coupling function, ε > 0 is the overall coupling strength, and G is the Laplacian matrix associated to the network. [*] L.M. Pecora, T.L. Carroll, Master Stability Functions for Synchronized Coupled Systems, Phys. Rev. Lett., Vol. 80, pp. 2109-2112, 1998. June 6, 2016 M. Biey - Dip. di Elettronica e Telecomunicazioni - Politecnico di Torino 3 A measure of synchronizability (cont.) The network admits a synchronous state x1 = x2 = = xN = s, whose (local) stability is determined by the corresponding system of variational equations, together with the motion on the synchronous manifold s f (s) (2) This system of variational equations can be diagonalized into N blocks of the form θ ( Df (s) DH (s))θ (3) where θ represents a mode of perturbation from the synchronized state; Df(s) denotes the Jacobian matrix of f evaluated at s and DH(s) denotes the Jacobian of matrix of H evaluated at s; γ = ελi, i = 1, ··· , N; λ1 = 0 >λ2 ≥ ··· ≥ λN are the eigenvalues of G, all of them real as the matrix G is symmetric. Λ(γ ) is defined as the largest Lyapunov exponent of the system defined by Eqs. (2) and (3) as a function of γ. This function, which generally is obtained by numerical The master stability function (MSF) methods, determines the stability of the synchronized state. In particular, the synchronized state is unstable if Λ(ελi) > 0, for at least an index i ∈ {2, . . . , N}. June 6, 2016 M. Biey - Dip. di Elettronica e Telecomunicazioni - Politecnico di Torino 4 A measure of synchronizability (cont.) The region S where (α ) 0 is called synchronization region. In the majority of cases, S may have one of the following form: • S1 = 0 (empty set) • S2 = (m, + ) • S3 = j (jm, jM) αm αM synchronization region In general, j = 1 and m< 0, M < 0 The condition of stable synchronous state are: S2 : ελ2 m ; S3 : a value of ε exists iff λN / λ2 < M / m June 6, 2016 M. Biey - Dip. di Elettronica e Telecomunicazioni - Politecnico di Torino 5 A measure of synchronizability (cont.) In other words, for a large class of oscillators there exist two classes of networks: class-A networks: for which the condition of stable synchronous state is ελ2 m = a; class-B networks: for which this condition reads λN / λ2 < M / m = b; a and b depends on f (·) , the synchronous state s, and the coupling matrix H; a and b do not depend on the Laplacian matrix G. June 6, 2016 M. Biey - Dip. di Elettronica e Telecomunicazioni - Politecnico di Torino 6 A measure of synchronizability (cont.) The ratio Q = λN/λ2 depends only on the network topology, while the ratio αM/αm depends on the dynamics of each node (f ( )) , the synchronous state and the coupling function (H( )). The lower the Q, the wider the interval of possible coupling strength ε such that the corresponding network has a locally stable synchronous state. As a consequence, building networks which are highly synchronizable generally means building networks with a low Q. In practice, the MSF is not always negative only in a finite interval. However, synchronizability in a variety of dynamical processes can be described by similar spectral properties, which tend to improve when the network is more homogeneous. However, simply put, other measures of synchronizability often go ‘‘hand in hand’’ with λN/λ2. June 6, 2016 M. Biey - Dip. di Elettronica e Telecomunicazioni - Politecnico di Torino 7 Classical Random Networks Classical random networks (Erdős - Rényi model (*)) Given N vertices, each edge is independently chosen with probability q : G(N,q) Erdős and A. Rényi, “On random graphs,'' Publ. Math Debrecen, vol. 6, pp. 290-297, 1959. (*) P. June 6, 2016 M. Biey - Dip. di Elettronica e Telecomunicazioni - Politecnico di Torino 8 Classical Random Networks (cont.) For Classical Random Networks G(N,q), for large N, λ2 N, while λN / λ2 approaches 1 [*]. For class-A networks the condition for synchronization (ελ2 a) reads ε > a/N and ε can be chosen arbitrary small. For class-B networks with b > 1, since λN / λ2 approaches 1 for large N, it follows that the network almost surely synchronizes. [*] M. Fidler, Algebraic connectivity of graphs, Czechoslovak Mathematical Journal, Vol. 23, pp. 298-305, 1973. June 6, 2016 M. Biey - Dip. di Elettronica e Telecomunicazioni - Politecnico di Torino 9 Classical Random Networks (cont.) Theorem : Let G(N, q) be a random graph on N vertices. Then the class-A network G(N, q) asymptotically almost surely synchronize for arbitrary small coupling ε; the class-B network G(N, q) asymptotically almost surely synchronize for b > 1. A proof can be found in [9] June 6, 2016 M. Biey - Dip. di Elettronica e Telecomunicazioni - Politecnico di Torino 10 Conclusions For a large class of oscillators there exist two classes of networks with known conditions for the existence of stable synchronous state, namely class-A networks and class-B networks When oscillators belonging to these classes are used to form networks with classic random graph on N vertices, then the synchronization behaviour can be predicted. More detailed results and proofs can be found in Ref. [9]. June 6, 2016 M. Biey - Dip. di Elettronica e Telecomunicazioni - Politecnico di Torino 11 References [1] M. Barahona, L.M. Pecora, Synchronization in Small-World Systems, Phys. Rev. Lett., Vol. 89, 054101, 2002. [2] T. Nishikawa, A.E. Motter, Y.C. Lai,1, and F.C. Hoppensteadt, Heterogeneity in Oscillator Networks: Are Smaller Worlds Easier to Synchronize?, Phys. Rev. Lett., Vol. 91, 014101, 2003. [3] M. Chavez, D.-U. Hwang, A. Amann, H. G. E. Hentschel, and S. Boccaletti, Synchronization is Enhanced in Weighted Complex Networks, Phys. Rev. Lett., Vol. 94, 218701, 2005. [4] L. Donetti, P.I. Hurtado, and M.A. Munoz, Entangled Networks, Synchronization, and Optimal Network Topology, Phys. Rev. Lett., Vol. 95, 188701, 2005. [5] C. Zhou, A.E. Motter, and J. Kurths, Universality in the Synchronization of Weighted Random Networks, Phys. Rev. Lett., Vol. 96, 034101, 2006. [6] C. Zhou, J. Kurths, Dynamical Weights and Enhanced Synchronization in Adaptive Complex Networks, Phys. Rev. Lett., Vol. 96, 164102, 2006. [7] L. Huang, K. Park, Y.-C. Lai, L. Yang, and K. Yang, Abnormal Synchronization in Complex Clustered Networks, Phys. Rev. Lett., Vol. 97, 164101, 2006. [8] D.-H. Kim, A.E. Motter, Ensemble Averageability in Network Spectra, Phys. Rev. Lett., Vol. 98, 248701, 2007. [9] M. Biey, F. Corinto, I. Mishkovski, and M. Righero, Synchronization in complex networks: properties and tools, in L. Kocarev (Editor), Consensus and Synchronization in Complex Networks, Springer-Verlag, Heidelberg, pp. 111-153, 2013. June 6, 2016 M. Biey - Dip. di Elettronica Telecomunicazioni - Politecnico di Torino 12