Dynamics of Complex Systems M.Y. Choi Department of Physics Seoul National University Seoul 151-747, Korea Main Collaborators J. Choi (KU), D.S. Koh (UW), B.J. Kim (AU), H. Hong (JNU), G.S. Jeon (PSU), J. Yi (PNU), M.-S. Choi, M. Lee (KU), H.J. Kim, Y. Shim (CMU), J.S. Lim, H. Kang, J. Jo (SNU) May 2005 PITP Conference Complex System Many-particle system many elements (constituents) a large number of relations among elements Nonlinearity (nonlinear relations) Open and dissipative structure Memory Aging properties Between order and disorder adaptation interactions complicated behavior environment essential information flow critical Large variability ← frustration and randomness Characteristic time-dependence → dynamic approach Potpourri of Complex Systems Electron and superconducting systems: Josephson-junction arrays, Harper’s equation, CDW Glass: glass, spin glass, charge glass, vortex glass, gauge glass Complex fluids: colloids, polymers, liquid crystals, powder, traffic flow, ionic liquids Disordered systems: interface, growth, composites, fracture, coupled oscillators, fiber bundles Biological systems: protein, DNA, metabolism, regulatory and immune systems, neural networks, population and growth, ecosystem and evolution Optimization problems: TSP, graph partitioning, coloring Complex networks: communication/traffic networks, social relations, dynamics on complex networks Socio-economic systems: prisoner’s dilemma, consumer referral, stock market , Zipf’s law similarity out of diversity details irrelevant Dynamics of Driven Systems Relaxation and responses Synchronization and stochastic resonance Mode locking, dynamic transition, and resonance Mesoscopic Systems Quantum coherence and fluctuations (Quantum) Josephson-junction arrays Charge-density waves Biological Systems Insulin secretion and glucose regulation Dynamics of failures Information transfer and criticality Other Systems Complex networks Consumer referral Dynamics of Driven Systems many-particle system time-dependent perturbation (external driving) Ω period τ ≡ 2π/Ω relaxation time τ0 • relaxation time τ0 ≠ 0 response not instantaneous • competition between τ0 and τ rich dynamics dynamic hysteresis, dynamic symmetry breaking, stochastic resonance, mode locking and melting Ubiquitous but equilibrium concepts (free energy) inapplicable No perturbation: equilibrium order parameter m m ≠ 0 → broken symmetry Time-dependent perturbation h(t): dynamics ☜ Langevin equation, Fokker-Planck equation, master equation, etc. equations of motion: symmetric in time order parameter m(t): may not be symmetric in time Q dynamic order parameter 1 dt m Q 0 → dynamic symmetry breaking ordered phase shrinks as ω→0 dynamic divergence of the relaxation time and fluctuations 1D/2D Superconducting Arrays simple complex system superconducting islands weakly coupled by Josephson junctions in magnetic fields driven by applied currents magnetic field/charge → frustration “Fancy” concepts: topological defects, symmetry and breaking, topological order, gauge field, fractional charge, frustration, randomness, gauge glass and algebraic glass order, chaos, Berry’s phase, topological quantization, mode locking and devil’s staircase, dynamic transition, stochastic resonance, anomalous relaxation, aging, complexity, quantum fluctuations and dissipation, quantum phase transition, charge-vortex duality, quantum vortex, QHE, AB/AC effects, persistent current and voltage, exciton Frustrated XY Model H EJ cos( i j Aij ) i, j 2e j Aij A dl, c i A ij 2 0 2f P Symmetry depends on f in a highly discontinuous fashion f = 0 (unfrustrated): U(1), BKT transition T < Tc: critical, power-law decay of phase correlation f = ½ (fully frustrated): U(1)Z2 ground state: doubly degenerate (discrete) → Z2 (Ising) → double transitions (BKT + Ising?) two kinds of coupled degrees of freedom phase (vortex excitation) chirality (domain-wall excitation) Current-driven array of Josephson junctions L L SQ array uniform applied currents I iext I ( x,1 x, L ) resistively shunted junction current conservation → equations of motion d ext A I sin A I j 2eR dt i j ij C i j ij ij i ' 2kT (t t ' )(ik jl il jk ) noise current R I =Id: IV characteristics, current-induced unbinding, CR ij (t )kl (t ' ) I = Ia cos t: dynamics transition, SR I = Id + Ia cos t: mode locking, melting and transition real dynamics (↔ kinetic Ising model) Stochastic Resonance ac driving I = Ia cos t S SNR 10 log 10 N signal S : power spectrum peak at N : background noise level Ia = 0.8; /2 = 0.08: Q > 0 (no osc.) at T = 0 staggered magnetization • SR phenomena peak only at T >Tc ( double peaks around Tc) ☜ τ → ∞ at T <Tc Mode Locking ac + dc driving I = Id + Ia cos t at T = 0 → voltage quantization: giant Shapiro steps (GSS) L IGSS 2e n L f r s: V FGSS s 2e f 0: V n (cf. devil’s staircase) • mode locking ← topological invariance • chaos Dynamic phase diagram melting of voltage steps from the voltage step width w V = 0(□), 1/2(O), 1(∆) Inset: V 1/ 4 Arnold tongue structure dynamic transition ↔ melting of Shapiro steps Biological Systems Paradigm: complex systems displaying life as cooperative phenomena Physics: understanding by means of (simple) models relevant and irrelevant elements • fine-grained modeling: beta cells, protein dynamics • coarse-grained modeling: synchronization, failure, evolution Insulin Secretion and Glucose Regulation β-cells in Islet of Langerhans glucose → bursting behavior → insulin secretion Pancreas Islet of Langerhans Action Potentials Intact β-cells V Isolated β-cells Kinard et al. (1999) Synchronized bursting of β-cells simultaneous recording of the electrical activity from two cells Bursting mechanism Activation and inhibition of GLUT-1 and GLUT-2 transporters by secreted insulin are represented by the solid (+) and dashed (-) arrows. Thick arrows describe physical transport of materials (glucose and ions). glucose ATP ↑ K+ channel closed K+ ↓, depolarized Ca2+ channel open Ca2+ ↑ insulin exocytosis Coupled oscillator model Current equation at each cell i, neighbors of which are linked by gap junctions Noise (thermal fluctuation) increase noise level Noise (stochastic channel gating) Multiplicative or colored noise induces more effectively several consecutive firings than white noise. Coupling (Gap Junction) weak coupling (10 pS) optimal coupling (40 pS) regular bursts induced strong coupling (100 pS) Collective synchronization coherent motion among many coupled cells Josephson junctions, CDW, laser, chemical reactions, pacemaker cells, neurons, circadian rhythm, insulin secretion, Parkinson’s disease, epilepsy, flashing fireflies, swimming rhythms in fish, crickets in unison, menstrual periods, rhythms in applause prototype model: set of N coupled oscillators each described by its phase φi and natural frequency ωi driven with amplitude Ii and frequency Ω N i (t ) i (t ) J ij sin i (t ) j (t ) Aij i Ii cos t i (t ) j natural frequency distribution (e.g. Gaussian with variance σ2 ≡1) phase order parameter 1 N e j i j ei 1 g () ( j ) N j ( 0 : synchronization) Failures in biological systems neurons (Alzheimer) , β cells (diabetes), T cells (AIDS) degenerative disease Time course of HIV infection HIV antibodies CD4+ T cells Virus 2-10 wks Up to 10 yrs Simplest model: system of N cells under stress F = Nf state of each cell: si = ±1 dead/alive state of the system {s1, s2, …, sN } 2N states If cell j becomes dead (sj = 1), stress Vij is transferred to cell i total stress on cell i 1 s j Vi f Vij 2 j death of cell i depends on Vi and its tolerance gi: (Vi gi ) si 0 or si ( Vij s j hi ) 0 j uncertainty due to random variations, environment probabilistic (noise effective temperature T) time delay td in stress redistribution cell regeneration in time t0 → healing parameter a ~ t0-1 a = 0: fiber bundle model rupture, destruction, earthquake, social failure dynamics ← master equation for probability P({si}, t; {si’}, t-td) Time evolution of the average fraction of living cells f fc Phase diagram healthy state Information transfer and evolution Fossil record evolution proceeds not at a steady pace but in an intermittent manner punctuated equilibrium fossil data display power-law behavior critical number of taxa with n sub-taxa: M n ~ n 2 lifetime distribution of genera: M t ~ t number of extinction events of size s: M s ~ s power spectrum of mutation rate: P() ~ 1.5 Basic idea molecular level: random mutation natural selection phenotypic level: power-law behavior evolution dynamics: random mutation and natural selection Evolution dynamics ecosystem consisting of N interacting species configuration x≡{xi} (i = 1,2,…,N) fitness of each species fi(x) total fitness F(x) ≡ ∑i fi(x) (≡ − energy) entropy S ( F ) ln ( F ) ( F ) j dx j ( F i f i ( x)) ecosystem directed to gather information from the environment and to evolve continuously into a new configuration information transfer dynamics entropic sampling environment ecosystem x S (F ) information exchange S0 ln 0 St ( F ) S S0 St (0) F total entropy probability for the ecosystem in state x ( St F |0 ) P( x) 0 e S0 e St S eF ( x ) S ( F ( x )) β → ∞: important sampling β → 0: entropic sampling (St = const., i.e., reversible info exchange) power-law behavior (γ ≈ τ ≈ 2) Mutation Rate and Power Spectrum P( ) 1.5 critical, scale invariant 2D Ising model power spectrum of magnetization and relaxation time P( ) 2 Scale-free behavior emerging from information transfer dynamics L2.6 Other Systems Complex Networks •Regular networks (lattices) highly clustered characteristic path length: O( N ) •Random networks low clustering characteristic path length: O (log N ) •Networks in nature: in between regular and random → complex – Biological networks: neural networks, metabolic reactions, protein networks, food webs – Communication/Transportation networks: WWW, Internet, air route, subway and bus route – Social networks: citations, collaborations, actors, sexual partners Small-world networks Start from regular networks with N sites connected to 2k nearest neighbors Rewire each link (or add a link) to a randomly chosen site with probability p Highly clustered ≈ regular network (p = 0) Average distance between pairs increase slowly with size N ≈ random network (p = 1) Scale-free networks preferential linking hub structure power-law distribution of degrees Coauthorships in network research MEJ Newman & M Girvan Dynamics on small-world networks Phase transition, Synchronization, Resonance: spin (Ising, XY) models and coupled oscillators mean-field behavior for p > pc ( = 0 ?) fast propagation of information for p ≥ 0.5 lower SR peak enhanced system size resonance → cost effective Vibrations: Netons excitation gap → rigidity against low energy deformation Diffusion quantum system: classical system: N2 N N log N fast world Economic Systems: Consumer referral on a network A monopolist having a link with only one out of and N consumers Each consumer considers his/her valuation distributed according to f(v), and decides whether to purchase one at price p. If yes, (s)he decides whether to refer other(s) linked at referral cost δ. Referral fee r is paid if (s)he convinced a linked consumer to buy one. The procedure is continued. ● 1 ● 2 3 ● 4 ● 5 ● 6 ● ● 7 ● 8 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● N branched chain with branching probability P Maximum profit (per consumer) vs N P = 0: maximum profit per consumer ~ 1/N (→ 0 as N → ∞) P≠ 0: maximum profit per consumer saturates (→ finite value as N → ∞) small-world transition Concluding Remarks Physics pursuits universal knowledge (“theory”) “theoretical science” how to understand phenomena and how to interpret nature Physics in 20th century: fundamental principles Reductionism and determinism Simple phenomena (limited, exceptional) Particles and fields Physics in 21st century: interpretation of nature Emergentism, holism, and unpredictability complementary Complex phenomena (diverse, generic) Information Appropriate methods statistical mechanics nonlinear dynamics computational physics Physics of Complex Systems biological physics, econophysics, sociophysics, …