Hamiltonian Systems • • • • • • • • • • Introduction Hamilton’s Equations & the Hamiltonian Phase Space Constants of the Motion & Integrable Hamiltonians Non-Integrable Systems, the KAM Theorem & PeriodDoubling The Henon-Heiles Hamiltonian The Chirikov Standard Map The Arnold Cat Map The Dissipative Standard Map Applications Introduction • No dissipation = Conservative – Phase space volume = constant – No transients, no attractors – Isolated (closed) system: Conservative • Open system: Dissipative • Hamiltonian system: Dynamics governed by H(q,p) – Solar system: • Gravitation: conservative • Tidal forces, solar wind: dissipative • Dissipation negligible for short times – Microscopic (quantum) systems • Integrable systems: non-chaotic Hamilton’s Equations & the Hamiltonian • Hamilton formulation: phase space = { qi, pi } • Dynamics ( Hamilton’s equations ): H q, p, t qi pi pi H qi i 1, 2, ,Nf Nf = degrees of freedom E.g., n 3-D particles: q = {qi} = { x1,y1,z1, x2,y2,z2, …, xn,yn,zn } Nf = 3n p = {pi} = { px1,py1,pz1, px2,py2,pz2, …, pxn,pyn,pzn } Equivalent 1st order system of ODEs ( “DoF" = 2Nf ): x x j j 1, , 2N f H x2i 1 f 2i 1 x x2 i x2i 1 qi x2i pi H x2 i f 2 i x x2i 1 i 1, ,Nf Symplectic structure: manifold with a (symplectic) 2-form as metric For the Hamiltonian systems: dH d f X f i d q dpi i X g , X f f , g j f g g f q j p j q j p j (Poisson bracket) Definition in terms of symplectic matrices ( see Goldstein ): ( M is symplectic if MT J M = J ) 0 H xJ x T f g f , g J x x x M x I J I 0 f f df f f H f H f qj pj dt q p t j p j q j t j j j q j p j f f , H t d H H H is conserved ( a constant of the motion ) → if it’s not explicitly time dependent. dt t Phase Space For a conservative system, trajectories in phase space are confined to a constant energy surface. DoF = N → energy surface is (2N-1)-D Volume in state space of autonomous system 1 dV f i V dt i xi xi fi x See §3.13 For a conservative system H(q,p) : 1 dV V dt j q j H pj pj → no attractors H q j 0 → transients persist Liouville’s Theorem Probability of finding a system in dV = q, p,t dV ρ= probability density of systems in phase space. H H d qj pj dt p j t p j q j t j qj j qj pj , H t d = material / hydrodynamic derivative → Lagrangian picture dt = change of ρ at fixed phase point → Eulerian picture t Liouville’s theorem: d 0 dt ~ no-crossing theorem for Hamiltonian systems Integral Invariants I t t dx1 I t d I t dt dxn x, t dx J x, t 0 1 0 0 J det J 0 n dx 0 1 dx d I t 0 dt is an integral invariant if d Jik J dJ xi i k 0 J d t i, k d t J i k i , k xk and Jik is the cofactor of Jik : x10 , xn xn0 xi J J i k 0 xk d dx J dt 0 n I(t) is an integral invariant → x1 , d J d J d J 0 dt dt dt d Jik where dt J i k J i m J m km d xi xi 0 0 d t xk xk J J ik Jik dJ x xi xm i k xi xi ik 0i J i k J J J i m J x J mk 0 d t i , k xk i , k , m xm xk i , k , m xm i , m xm d J d J d d J J x J x x dt dt dt dt t d I t dt 0 dx10 dxn0 J x t I(t) is an integral invariant → t dx1 J x t dxn x t x 0 t Liouville’s theorem: x 0 if ρ is the probability density of a Hamiltonian H(q,p,t). (ρ acts like an incompressible fluid ) Proof: J t det q q0 q p0 p q0 p p0 I J t det q0 t q0 p0 t q0 q0 t p0 I p0 t p0 J t J t det q I 0 t q0 p0 t q0 q0 t p0 I p0 t p0 1 det qi 0 pi 0 i 1 q t 1 p t i0 i0 dJ 0 x J dt → x 0 1 1 q p 1 i 0 i 0 t pi 0 i qi 0 H H 1 t pi 0 qi 0 i qi 0 pi 0 → 1 O t 2 & d x 0 dt t Alternatively: q p H H x i i 0 pi pi qi i qi i qi pi QED In general: 1st order ODEs: x 0 t xF Hamiltonian systems → F 0 t F 0 t Equation is linear in ρ → • Evolution of ρ can’t be chaotic, even when individual trajectories are. • No “divergence” of evolution of nearby ρ’s. • No sensitivity to I.C. • Prototype of chaotic Hamiltonian system: Arnold cat map. Constants of Motion & Integrable Hamiltonians Any quantity that is independent of t is a constant of the motion. For a conservative system, i.e., E H q, p is independent of t. H q t , p t H q 0 , p 0 t Each point on a possible trajectory {q(t),p(t)} has the same energy E. Many different trajectories may have the same energy. E is a constant of the motion. It is also an isolating integral that restricts the motion on surface. Let pj be a constant of the motion, then 0 pj i.e., H is not explicitly dependent on qj. Each trajectory is confined to an (2Nf-2)-D surface. Each trajectory can be characterized by { E, pj0 }. some H q j ( qj is cyclic ) k independent isolating integrals → each trajectory is confined to an (2Nf-k)-D surface If k = Nf, the system is integrable. ( Trajectory Nf-D ) Isolating integrals are also called action variables Ji(q,p). The variable conjugate to Ji(q,p) is called an angle variable. Ji H 0 i i H Ji Θi can always be chosen as dimensionless so that Ji has the dimension of action. { Θi, Ji } can be related to { q, p } by a canonical transformation. If H can be written as H(J), the system is integrable. Ji Ji , H J 0 i can be satisfied iff → System is in involution. J , J 0 i j i,j ( + independence ) Integrability can be examined by expressing the desired canonical transformations in terms of a Birkhoff series. Examples of integrable systems: • All 1-D systems with analytic H → H = ωJ. • All systems with linear equations of motion → normal modes. • All systems that are completely separable. • Solitons Let H = H(J), then Inverse transformation: i H i J Ji → qi t qi θ t , J t i t i t i 0 pi t pi θ t , J t For a bounded system, q and p must be periodic functions of Θ. Canonical perturbation theory: Series diverges → non-integrable q,p as series of Θ, J. Simple Harmonic Oscillator p2 1 2 H kq 2m 2 Nf = 1 → q Phase space: 2-D dp p k q dq q p/m Fixed point: → H p p m 2J q t cos (t ) m p t 2m J sin (t ) H k q q m q k q → Trajectory: 1-D (H = const → ellipse) p d p m k qd q q*, p * 0,0 Switching to (Θ,J): p p2 1 H m 2 q 2 J 2m 2 p 2m k m = elliptic point P Ellipses: periodic 1 2 1 p mk q 2 C 2 2 Q J m q 2 H P2 Q2 PQ 1 pq 2 Trajectory is a circle of radius √J & area πJ in (P,Q) space Q t J cos (t ) P t J sin (t ) H J t 0 t Area of ellipse: pd q J d 2 J E 1 2mL g L 1 cos 2mg L 2 p2 H mg L1 cos 2 2mL p 2 p mg L sin mL p f mgL sin 0 2 mL p p 2mL2 E mg L1 cos p p sin 2 2 2mL g L 1 0.5 → Conservative 4 3 2 2 -0.5 -1 3 4 0 J mgL cos g 2 cos L 1 mL2 0 1 v c 2 mL Trajectories in {θ, p } space for ε = 0.2, 0.6, 1.0, 1.4, 1.8 Elliptic points at Hyperbolic points at pθ = 0, θ = 2nπ pθ = 0, θ = (2n+1)π Separatrices: Stable & unstable manifolds of a hyperbolic point Typical for integrable Hamiltonian systems Elliptic Integrals & Elliptic Functions Ref: M.Abramowitz, I.A.Stegun, “Handbook od Mathematical Functions”. Complete Elliptic integrals: (Incomplete) Elliptic integrals: 1st kind: F \ m 0 2nd kind: K m F \ m 2 d 1 m sin 2 E \ m d E m E \ m 2 1 m sin 2 0 (Jacobian) Elliptic functions: sin sn u | m sn u cos cn u u F \ m am u 1 m sin 2 dn u Mathematica: EllipticK[m] = EllipticF[π/2,m] Sin[φ] = JacobiSN[EllipticF[φ,m],m] Systems with N Degrees of Freedom N Integrable systems: H θ, J i J i i 1 Ji H 0 i ~ N uncoupled oscillators i Q.M.: equally spaced H i Ji ( simple harmonic if ω independent of J ) N constants of motion → trajectories on N-D torus in phase space For N = 2: trajectories are on (invariant) torus 1 H J1 2 ωi incommensurate → q.p. → ergodic → time average = ensemble average Non-integrable systems: tori broken; J(θ) not lines H J2 The Kepler Problem (Integrable) See H.Goldstein, "Classical Mechanics", 2nd ed., §10-7 ( with minor variations ) 2 p pr2 k H 2 2 2 r r m1m2 m1 m2 r, p , , p , J , , J r 1 1 2 2 Solution of the (separable) Hamilton-Jacobi eq gives J 1 2 p H J1, J 2 J 2 J1 k 2 2 k 2 J1 J 2 2 2 E E 4 2 k 2 H 1 2 3 J 1 J1 J 2 → 1 2 k Gm1m2 Nonintegrable Systems Integrable systems: periodic / q.p. → non-chaotic Transition to chaos: integrable → slightly non-integrable Non-integrable systems: Df 2 For Df = 2, integrable → motion on 2-D torus Non-integrable → motion on 3-D constant E surface Poincare sections transverse to 3-D constant E surface is 2-D Integrable system ( sets of nested tori with separatrices ): • Series of discrete points → periodic • Closed paths around point → quasi-periodic orbits around elliptic point • Hyperbolic orbits near hyperbolic points Chaos ? Henon-Heiles, almost integrable KAM Theorem Let H J, θ H0 J H1 J, θ H0 integrable KAM theorem (criteria dropped): Tori that survive perturbation satisfy W g m n n5 g(ε) increases monotonically with ε Implication: • For ε > 0, all tori with rational W break up • (KAM) tori with irrational W persist, then break up 1 by 1 as ε • Last to be destroyed has golden mean ratio ( most irrational ) Qualitative explanation: • W rational → motion sustained by strong resonances between overlapping harmonics • → any perturbation will remove overlappings • → rapid break up of tori increases Sequential break-up of KAM tori Band of width g around m/n n5 Tori within band dissolve Resonance structure Chaos ~ overlap of resonances Df = N → (2N-1)-D const E surface, N-D tori A torus can partition E surface only if N = 2N-1 or (2N-1)-1 → N = 1 or 2 ( KAM tori partition phase space ) For N > 2, tori break up → stochastic web ( no partitioning ) Poincare-Birkhoff Theorem H H ~ twisted map (area preserving) 0 Break up of n/M tori → n/M pairs of 2n/2M elliptic & hyperbolic fixed points → Period-doubling Insets & outsets of hyperbolic point dissolve first → homoclinic & heteroclinc tangles → chaos? Different I.C. may lead to q.p./chaotic motion For Df 2, surviving KAM tori confine chaotic motion near broken tori For Df 3, chaos from any broken torus can roam mostly freely (Arnold diffusion). Lyapunov exponent: • Σλi = 0 for conservative system • Chaos: at least 1 λi > 0 Monodromy matrix M : z t M z 0 Eigenvalues μi of M ~ Floquet multipliers: μi comes in pairs of (μ, μ-1 ) z = periodic orbit Πμi = 1 i e i Period-Doubling Break up of m/n tori • m/n pairs of 2m/2n elliptic & hyperbolic fixed points • Period doubling • ε increases → further period doubling … • δH = 8.721097…, αH = -4.01807… Period-n-tuplings are common in Hamiltonian systems Cause: resonances among constituent nonlinear oscillators Meyer's theorem: 5 types of bifurcations Singularities in H can also cause non-integrability E.g., billiard balls Henon-Heiles Problem A star in axially symmetric galaxy; • Nf = 3 • Known integrals of motion: E, Lz • No known analytic form of 3rd integral • If 3rd integral not exist → σ(vρ) σ(vz) • Observed: σ(vρ) : σ(vz) = 2 : 1 Henon-Heiles model: 1 2 1 3 r r sin 3 2 3 1 1 V x, y x 2 y 2 x 2 y y 3 2 3 V r, Henon-Heiles Hamiltonian 1 2 1 2 1 3 2 2 2 H px p y x y x y y 2 2 3 V V 2 0.4 0.3 1.5 0.2 1 0.1 -2 -1 1 2 y 0.5 -0.1 -0.2 2 x -2 0 1 -1 1 -1 0 -1 -2 2 1 -2 2 -2 7.5 -1 5 2.5 0 0 y 1 -2 -1 0 1 2 2 V 0.4 0.3 Height of potential well around 0 is 1/6 → bound orbits for E < 1/6 1 2 1 2 1 3 2 2 2 H p p x y x y y Hamilton's eqs: x y 2 2 3 -2 H x px px H px x 2x y x H y py py py 0.2 0.1 -1 1 -0.1 -0.2 H y x2 y2 y Nf = 2 → 3-D const E surface → 2-D Poincare section → E 1 2 1 2 1 3 2 p p y y x y 2 2 3 Choice: y-py plane at x = 0 2 y py y 0.3 0.3 0.2 0.2 0.1 0.1 -0.3 -0.2 -0.1 0.1 0.2 0.3 x -0.3 -0.2 -0.1 0.1 0.2 0.3 -0.1 -0.1 -0.2 -0.2 -0.3 -0.3 E = 0.06, x0 = 0, y0 = -0.1475, px0 = 0.3101, py0 = 0. Distorted torus: Quasi-periodic N = 1000 y y py 0.4 0.3 0.3 0.2 0.2 0.1 0.1 -0.3 -0.2 -0.1 0.1 0.2 0.3 x -0.3 -0.2 -0.1 0.1 -0.1 -0.1 -0.2 -0.2 -0.3 -0.3 E = 0.06, x0 = 0, y0 = 0.1563, px0 = 0.18876, py0 = -0.25 Hyperbolic points: • separatrices • Heteroclinic tangles → Stochastic layers (webs) 0.2 0.3 0.4 y N = 2000 y py 0.3 0.2 0.1 -0.3 -0.2 0.1 -0.1 0.2 0.3 0.4 -0.1 x -0.2 -0.3 Outside separtrix → Qualitatively different • x-y orbits wraps y-axis • Bounds allowed region E = 0.06, x0 = 0, y0 = 0, px0 = -0.0428, py0 = -0.3438 N = 200 y 0.4 0.3 0.2 0.2 0.1 -0.3 -0.2 -0.1 0.1 0.2 0.3 0.4 -0.4 -0.2 0.2 -0.1 -0.2 -0.2 -0.3 E = 0.06 -0.4 N = 10000 E = 0.1 N = 40000 • Orbits near separatrices easily disturbed • Breakup of KAM torus → necklace ---- (Birkhoff thm) ( Associated hyperbolic points not shown ) • Remaining KAM tori block chaotic roaming 0.4 0.6 0.4 0.4 0.2 -0.4 -0.2 0.2 0.2 0.4 0.6 -0.4 -0.2 0.2 0.4 0.6 -0.2 -0.2 -0.4 -0.4 -0.6 E = 0.14 N = 40000 E = 0.166 N = 40000 • Single trajectory roams through most places • Lyapunov exponents: +,0,0,- 0.8 The Chirikov Standard Map Aliases: Taylor-Greene-Chirikov map, the standard map. K sin 2 n mod 1 2 n1 n rn1 mod 1 K n rn sin 2 n mod 1 2 rn 1 rn (Moser) Twist map: n1 n n1 n W n mod 1 W = winding number 1 0 J 1 W 1 area preserving 1 K cos 2 J 1 1 1 K cos 2 area preserving Fixed points ( with r → J ): K → Jn sin 2 n J n m 2 K n J n sin 2 n n p → 2 K < 2π → Near fixed point: m = 0 1 and θ* = 0, ½ rn1 1 K cos 2 * rn 1 1 K cos 2 * n n1 Floquet multipliers: 2 2 m m,p = integers sin 2 * K K J* sin 2 * p m p 0, 1 2 1 K cos 2 * 0 1 1 K cos 2 * 1 K cos 2 * K cos 2 * 0 1 1 K cos 2 * K 2 cos 2 2 * 4 K cos 2 * 2 1 1 K cos 2 * K 2 cos2 2 * 4 K cos 2 * 2 1 1 K K 2 4K 2 For (J*,θ*) = (0,0): Reμ |μ| Imμ 1 2 1 1.75 1 2 3 4 5 6 1.5 0.5 1.25 -1 1 3 -2 4 5 6 0.75 -0.5 0.5 -3 0.25 -1 1 2 3 4 5 → (0,0) is a stable spiral for K < 4 1 1 K K 2 4K For (J*,θ*) = (0, ½ ): 2 8 6 μ → (0, ½) is a saddle point for all K 4 2 1 2 3 4 5 6 6 1 1 0.8 0.8 0.6 0.6 0.4 0.4 0.2 0.2 0.54 0.2 0.4 0.6 0.8 0.2 1 0.4 0.6 0.8 1 0.52 K = 0 K = 0.2 0.5 c.f. Henon-Heiles 0.48 θ = 0, J = ½ Period 2 0.46 0.2 0.4 0.6 0.8 1 The Arnold Cat Map xn 1 2 xn yn mod 1 2 1 J 1 1 1 yn 1 xn yn mod 1 2 1 1 0 3 1 0 2 → Area-preserving 1 3 5 2 → All fixed points are saddle points Fixed points: x * y* m x* k xn 1 2 1 xn y 1 1 y mod 1 n n1 2 1 1 2 1 1 0 1 2 1 0 1 1 1 1 1 → y* 0 x* 0 2.618 0.382 → ↓ ← Ex 8.8-1: Fixed points of f(n) have rational coordinates Evolution of 1-D conservative system: d 0 dt → qn , pn , n q0 , p0 ,0 pn 1 2 pn qn mod 1 qn 1 pn qn mod 1 ( n ~ tn ) qn , pn , n Ajnk exp 2 i jqn k pn Fourier analysis: jk qn 1 , pn 1 , n 1 Aj k n 1 Ajnk * A njk exp 2 i jqn 1 k pn 1 jk Ajnk1 exp 2 i j qn pn k 2 pn qn jk Ajnk1 exp 2 i j k qn j 2k pn jk j j k k j 2k j 2 j k k k j qn 1 , pn 1 , n 1 A2 nj1k, k j exp 2 i jqn k pn qn , pn , n j k → n 1 n A2 j k , k j Aj , k n 1 n Aj , k Aj k , j 2 k Spread to new modes, But not IC sensitive The Dissipative Standard Map G.Schmidt, B.H.Wang, PR A 32, 2994 (85) rn 1 J D rn K sin 2 n mod 1 2 n1 n rn1 mod 1 K n J D rn sin 2 n mod 1 2 J D K cos 2 JD 1 1 K cos 2 JD = 0 : Dissipative for JD < 1 K n 1 n sin 2 n mod 1 2 ( Sine circle map with Ω = 0) 1 JD = 0 ~ Circle map 0.8 0.6 0.4 0.2 1 2 3 4 5 JD = 0 , K > 1 → period-doubling route to chaos beyond K∞ Channels overlap for JD > 0 Bifurcated 2n orbits ~ period-doubling 2n : p1, p2 periodic orbits ~ periodic windows: p2', p3' 2n chaotic bands disappear at univeral values of JD Applications • Billiards (elastic collisions, piecewise linear) – – – – Rectangular or circular walls → periodic / q.p. stadium / Sinai billiards ( round obstacle ) → chaotic for some orbits 2 balls + gravity: All kinds of behavior Quantum chaos • Astronomical Dynamics – Orbits of Pluto & some asteroids may be chaotic – Kuiper objects • Particle Accelerators – Avoid possible chaotic trajectories in accelerator design • Superconductivity – Vortex structures under magnetic field (type II superconductors) phaselocking, Arnold tongues, Farey tree, devil's staircase • Optics – small dielectric spheres: whispering gallery lasers – Spheres distorted → chaos