Dynamical system Study Guide

📖 Core Concepts Dynamical system – a rule that tells how a state \(x\) in a state space \(X\) evolves with time \(t\). \[ \Phi : T \times X \to X,\qquad \Phi(t0,x0)=x(t0+\Delta t) \] State space – the set of all possible configurations (e.g., \(\mathbb{R}^n\), a function space, a graph). Trajectory / orbit – the ordered set \(\{x(t)\mid t\in T\}\) generated from an initial condition. Continuous vs. discrete – \(T\) is an interval of \(\mathbb{R}\) (flow) or the integers \(\mathbb{Z}\) (map). Periodicity – a trajectory repeats after a period \(P\): \(x(t+P)=x(t)\). Lyapunov stability – nearby trajectories stay close for all forward time. Structural stability – qualitative picture (e.g., number of fixed points) persists under small perturbations of the evolution rule. Bifurcation – a qualitative change in the phase portrait when a parameter \(\mu\) crosses a critical value \(\mu0\). Ergodicity – time average along one long orbit equals the space (measure) average over the whole state space. Chaos – deterministic dynamics with sensitive dependence on initial conditions and topological mixing; long‑term prediction is impossible despite no randomness. --- 📌 Must Remember Existence & uniqueness → guarantees a well‑defined trajectory. Linear flow solution: \(x(t)=e^{At}x0\). Linear map iteration: \(x{n}=A^{n}x0\); with constant term \(b\), shift to \(y{n+1}=Ayn\) using \(y=x+(I-A)^{-1}b\). Eigenvalue test – For a continuous linear system: \(\Re(\lambda)<0\) ⇒ asymptotically stable. \(\Re(\lambda)>0\) ⇒ unstable. For a discrete linear map: \(|\lambda|<1\) ⇒ stable fixed point. \(|\lambda|>1\) ⇒ unstable. Hartman–Grobman theorem – Near a hyperbolic fixed point, the nonlinear flow is topologically conjugate to its linearization. Period‑3 theorem (Sharkovsky) – Existence of a period‑3 orbit implies existence of orbits of all periods. Poincaré recurrence – In a finite‑measure, measure‑preserving system, almost every point returns arbitrarily close to its start infinitely often. Feigenbaum constant – universal scaling factor (4.669) for period‑doubling cascades (logistic map). --- 🔄 Key Processes Constructing a flow from an ODE Write the vector field \( \dot{x}=v(x)\). Solve the IVP \(x(0)=x0\) → obtain \(\Phi(t,x0)\). Linearization near a fixed point Compute Jacobian \(J = D\!F(x^\)\). Write \(F(x)=Jx + O(\|x\|^2)\). Classify via eigenvalues (hyperbolic vs. elliptic). Creating a Poincaré map Choose a transversal surface \(S\). Follow the flow until it returns to \(S\); define \(F:S\to S\). Reduce a continuous‑time problem to a discrete map. Bifurcation analysis (parameter \(\mu\)) Locate equilibria/fixed points as functions of \(\mu\). Track eigenvalues of the Jacobian (continuous) or of \(J\) (discrete). Identify parameter values where eigenvalues cross the imaginary axis (flow) or unit circle (map). Checking ergodicity Verify that the only invariant sets under \(\Phi\) have measure 0 or 1. Use Birkhoff’s ergodic theorem: time averages converge to space averages a.e. --- 🔍 Key Comparisons Continuous flow vs. discrete map Continuous: infinitesimal time, governed by differential equations. Discrete: finite steps, governed by iteration of a function. Lyapunov stability vs. structural stability Lyapunov: concerns individual trajectories staying close. Structural: concerns the whole qualitative picture surviving perturbations. Deterministic vs. stochastic system Deterministic: future state uniquely determined by present state. Stochastic: future state given by a probability distribution (random forces, SDEs). Linear vs. nonlinear dynamics Linear: superposition holds; solutions are combinations of exponentials/eigenvectors. Nonlinear: may exhibit bifurcations, chaos, and no superposition. Hyperbolic vs. elliptic fixed point Hyperbolic: eigenvalues off the unit circle (map) or off the imaginary axis (flow); local dynamics homeomorphic to linear part. Elliptic: eigenvalues on the unit circle/imaginary axis; may support invariant tori (KAM theory). --- ⚠️ Common Misunderstandings Chaos ≠ randomness – chaotic trajectories are deterministic; randomness would require stochastic forcing. Negative real eigenvalues guarantee global stability – only local asymptotic stability; nonlinear terms can create other attractors far away. Period‑3 ⇒ “system is chaotic” – period‑3 forces all periods, but chaos also requires sensitive dependence and mixing. Invariant measure = probability distribution – an invariant measure need not be normalized; only when \(\mu(X)=1\) is it a probability. Lyapunov stable ⇒ structurally stable – a system can be Lyapunov stable yet change its phase portrait under tiny perturbations. --- 🧠 Mental Models / Intuition Flow as a rubber sheet – imagine stretching and folding a sheet of dough; points move continuously, preserving the sheet’s continuity. Map as a “photocopier” – each iteration copies the current picture, possibly distorting it; think of a function that repeatedly stamps the same pattern. Stable/unstable manifolds – picture a saddle: one direction pulls you in (stable), the perpendicular pushes you out (unstable). Bifurcation as a “fork in the road” – as a parameter changes, the trajectory can split into new branches (new equilibria) or merge. Ergodic mixing as shuffling cards – after many shuffles, any given card (state) is equally likely to appear in any position (region). --- 🚩 Exceptions & Edge Cases Linear systems with purely imaginary eigenvalues – not chaotic but can produce closed orbits (center); stability is neutral, not asymptotic. Non‑resonant condition failure – when eigenvalues satisfy resonance relations, higher‑order terms may prevent linearization (small‑divisor problems). Measure‑preserving but not ergodic – e.g., a rotation on the circle with rational angle: invariant measure exists, but trajectories are periodic, not dense. Structural stability in low dimensions – in 2‑D, only certain flows (Morse‑Smale) are structurally stable; most generic systems are not. --- 📍 When to Use Which Continuous‑time problem → write ODE, seek a flow; use linearization, Lyapunov functions, or Poincaré maps if a periodic orbit exists. Discrete‑time problem → treat as iteration of a map; compute Jacobian of the map, check \(|\lambda|=1\) for bifurcations. Parameter‑sweep study → perform bifurcation analysis; locate where eigenvalues cross critical values. Long‑term statistical description → apply ergodic theory, compute invariant measures (e.g., Liouville measure for Hamiltonian flows). Detecting chaos → look for stretching–folding mechanisms (horseshoe, logistic map), compute Lyapunov exponent > 0, verify sensitive dependence. Linear system → solve with matrix exponentials \(e^{At}\) (continuous) or powers \(A^n\) (discrete); no chaos possible if all eigenvalues have \(\Re(\lambda)<0\) or \(|\lambda|<1\). --- 👀 Patterns to Recognize Eigenvalue crossing – \(\Re(\lambda)=0\) (flow) or \(|\lambda|=1\) (map) signals a possible bifurcation. Period‑doubling cascade – successive halving of the distance between bifurcation parameters; indicates route to chaos. Stretch‑fold‑compress – hallmark of chaotic maps (e.g., Arnold cat map, baker’s map). Invariant manifolds intersecting transversely – creates a Smale horseshoe → chaotic dynamics. Time‑average ≈ space‑average – suggests ergodic behavior; check with long simulations. --- 🗂️ Exam Traps Choosing “stable” because eigenvalues are negative – forget to verify that nonlinear terms don’t introduce other attractors. Confusing Lyapunov with structural stability – a question may ask which property survives a model perturbation; answer: structural stability. Assuming any map with \(|\lambda|>1\) is chaotic – chaos also needs mixing and dense periodic orbits; a simple saddle is not chaotic. Misreading the logistic map parameter range – chaos appears for \(r>3.57\); period‑doubling occurs earlier. Over‑applying the period‑3 theorem – it guarantees existence of all periods, not that the system is chaotic in the sense of mixing. Neglecting the constant term \(b\) in linear maps – forgetting to shift coordinates can lead to an incorrect stability conclusion. ---