Top Banner
Aging transition and clustering in a large population of coupled oscillators Hiroaki DAIDO Department of Mathematical Sciences Graduate School of Engineering Osaka Prefecture University Sakai 599-8531, JAPAN KIAS conference: NSPCS2008 (Seoul, July 2008) emblem of OPU
28

Aging transition and clustering in a large population of coupled oscillators

Feb 02, 2016

Download

Documents

kerryn

Aging transition and clustering in a large population of coupled oscillators. Hiroaki DAIDO Department of Mathematical Sciences Graduate School of Engineering Osaka Prefecture University Sakai 599-8531, JAPAN. KIAS conference: NSPCS2008 (Seoul, July 2008). - PowerPoint PPT Presentation
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: Aging transition and clustering in a large population of coupled oscillators

Aging transition and clustering in a large population of coupled oscillators

Hiroaki DAIDO

Department of Mathematical Sciences Graduate School of Engineering Osaka Prefecture University Sakai 599-8531, JAPAN

KIAS conference: NSPCS2008 (Seoul, July 2008)

emblem ofOPU

Page 2: Aging transition and clustering in a large population of coupled oscillators

Contents

1. Background2. Aging in globally coupled oscillators#

Aging transition: Examples Universal scaling Clustering of active oscillators Summary3. Aging in locally coupled oscillators System size dependence of pc

Summary

# in collaboration with Kenji Nakanishi.  

Page 3: Aging transition and clustering in a large population of coupled oscillators

1. 1. BackgroundBackground

Coupled oscillators often model biological or physiological systems.

Deterioration due to aging or accidents etc.

self-sustained oscillator damped oscillator

( active oscillator ) ( inactive oscillator )

The problem of aging HD&KN, PRL 93 (2004), 104101.

Page 4: Aging transition and clustering in a large population of coupled oscillators

What happens when the ratio of inactive elements increases ?

( “aging” )active

inactive

How robust is the activity of coupled oscillatorsagainst aging ?

important not only biologically but technologically

effects of bad components

Page 5: Aging transition and clustering in a large population of coupled oscillators

Globally and diffusively coupled oscillators

periodic or chaoticGeneral form

2. Aging in globally coupled oscillators2. Aging in globally coupled oscillators

Examples (laboratory experiments)

Coupled electrochemical reaction systems Kiss et al. Science 296(2002), 176.

Coupled salt-water oscillators Miyakawa et al. Physica D 151(2001), 217.

Page 6: Aging transition and clustering in a large population of coupled oscillators

(1) Coupled Stuart-Landau equations

Active oscillators

Inactive oscillators

Aging transition: examples

a

-b

Page 7: Aging transition and clustering in a large population of coupled oscillators

The behavior of an order parameter

A measure of macroscopic activity

Synchronization withineach group

N=1000vanishes at

( Aging transition )

K=3, p=0.6

Page 8: Aging transition and clustering in a large population of coupled oscillators

Theory

Reduction to a four-dimensional system

0

0.2

0.4

0.6

0.8

1

0 2 4 6 8 10

death

Aging transition

The death stabilized at

q=1-p

Page 9: Aging transition and clustering in a large population of coupled oscillators

(2) Coupled Roessler systems

limit-cycle

fixed point

Parameters (1)

N=100

Page 10: Aging transition and clustering in a large population of coupled oscillators

Parameters (2)

chaos

Reverse period-doubling cascade

Page 11: Aging transition and clustering in a large population of coupled oscillators

Universal Scaling at the aging transition

ExampleCoupled periodic Roessler systemsN=1000

Page 12: Aging transition and clustering in a large population of coupled oscillators

General theory

Assumptions

Reduction

1. The reduction is possible.

2. A fixed point exists.

3. Its becomes Hopf unstable at

active

inactive

Page 13: Aging transition and clustering in a large population of coupled oscillators

Hopf

Universalcrossoverscaling

Page 14: Aging transition and clustering in a large population of coupled oscillators

Clustering of active oscillators

The synchronization within the active group breaks down in a region of the parameter plane.

Coupled SL equations N=4000, p=0

Resonance-like enhancementof inhomogeneity measured by

the bar average over all j

Diffusion-induced inhomogeneitybelow the peak point under thescalar type diffusive coupling:

Dx=Dy=K (x=Re(z), y=Im(z))

< > time average

Page 15: Aging transition and clustering in a large population of coupled oscillators

Analysis of the simplest 2-cluster state

Example of the cluster structure

clusterset of perfectlysynchronized oscillators

fractions of clusters in theorder of the size from aboveexcept the largest

2-cluster states with one cluster much smaller than the otherare abundant near the both ends of the clustering region

Page 16: Aging transition and clustering in a large population of coupled oscillators

Approximation in the large system-size limit

Simplest 2-cluster state

oscillators 1 to N-1oscillator N

main cluster

N-1 1

main cluster is unaffected by oscillator N

periodic oscillation

oscillator N obeys

u=1 perfect sync (stable)

Page 17: Aging transition and clustering in a large population of coupled oscillators

Theory vs. simulation

c2=-3,K=0.51, M=4

c2=-3,K=0.94, M=9

quasiperiodic

periodic

SN: saddle-node bifurcationmiddle: Hopf bifurcationSC: saddle connection

theoretical curves

periodicaperiodic×

(N=1000)

Page 18: Aging transition and clustering in a large population of coupled oscillators

Summary of Part 1

1. The problem of aging : Effects of increasing    inactive elements

2. Aging transition

Strong coupling: favorable for coherence, but less robust against aging !

3. Universal scaling at the aging transition

4. Clustering and Swing-by mechanism of

Diffusion-induced inhomogeneity

Page 19: Aging transition and clustering in a large population of coupled oscillators

3. Aging in locally coupled oscillators3. Aging in locally coupled oscillators

Effects of aging in locally coupled oscillators

a chain under the periodic boundary condition

(i.e. a ring) as a first step

1

Aging proceeds through random inactivation of oscillators

N >> 1

19

1

2

N

N-1

Page 20: Aging transition and clustering in a large population of coupled oscillators

Model & methods

Coupled Stuart-Landau oscillators on a ring

for all active oscillators (a>0)

for all inactive oscillators (b>0)

Number ratio: active: inactive=1-p:p

For K=0

active oscillator → limit-cycle

inactive oscillator   → z=0 20

Page 21: Aging transition and clustering in a large population of coupled oscillators

Aging scheme

Randomly choosing some active sites to inactivate at each step of increasing p .

The chosen oscillators remain inactive for all p after this.

Then, results are averaged over many realizations.

21

Page 22: Aging transition and clustering in a large population of coupled oscillators

(K,p) phase diagram

Example: a=b=1, c1=1, c2=-0.5

Aging transition boundaries

N = 6400 ( 20) 1600 ( 40) 400 ( 50) 100 (100)

active phase

inactive phase

Δp=0.0122

zj=0 for all j

Number of realizationsK

p

Page 23: Aging transition and clustering in a large population of coupled oscillators

Key features of the phase diagram

(1) Existence of Kc insensitive to changes in N

A linear stability analysis of the inactive state for p=(N-1)/N shows that Kc is given, for N → ∞, by

Example

a=b=1, c1=1 Kc=0.648…23

Page 24: Aging transition and clustering in a large population of coupled oscillators

(2) Vanishing of the inactive region for N → ∞

pc(K,N) → 1 for N → ∞ with K fixed

Simulation results suggest

Absence of the aging transition in the thermodynamic limit

Note: This does not imply unimportance of the AT, because (1) convergence of pc is slow, and (2) system sizes of real coupled oscillators are not always huge.

e.g. Lamprey’s CPG N ~ 100 mammalian circadian clocks N ~ 10000

24

Page 25: Aging transition and clustering in a large population of coupled oscillators

Scaling behavior of pc

How does pc approach unity as N grows toward infinity ?

Example a=b=1, c1=1, c2=-0.5

N = 100 ~ 12800K = 1, 1.8, 2.2, 2.6, 3.3, 4

Power laws !

1-pc(N) vs. N

25

Page 26: Aging transition and clustering in a large population of coupled oscillators

K dependence of the power law exponent

fit range N=100 ~ 12800

γ takes small valuesand tends to decreasewith K.

26

Page 27: Aging transition and clustering in a large population of coupled oscillators

Summary of Part 2

Aging in locally coupled oscillators

a chain of Stuart-Landau oscillators with n. n. interactions (as a first step)

(1) Aging transition (AT) in finite-size systems

Existence of Kc

(2) Absence of AT in the thermodynamic limit

27

Page 28: Aging transition and clustering in a large population of coupled oscillators

28

Main referencesMain references

H. D. & K. Nakanishi,

PRL 93(2004), 104101; 96(2006),

054101. PRE 75(2007),

056206; 76(2007), 056206(E).

H. D., to be published.