5
Deterministic Chaos
Period doubling · SDIC · Poincaré · Logistic map
🌀 The 4 Characteristics of Chaos
🗝 All four are required — missing one = wrong answer
| # | Characteristic | What it means | How to check |
|---|---|---|---|
| 1 | Determinism | Everything is determined by the current state (and the equations of the model). Same initial conditions always give the same trajectory. | Does it come from a definite rule/equation? If yes → deterministic. |
| 2 | Boundedness | Does NOT go to infinity. You could draw a box in state space and the trajectory would always stay within it. | Does it stay between a max and min value? If yes → bounded. |
| 3 | Irregularity (aperiodicity) | Does NOT repeat itself. The points in the system never exactly repeat. | Does the time series ever repeat the exact same sequence? If not → aperiodic. |
| 4 | SDIC | Two time series starting very close together will eventually diverge until their behavior is completely uncorrelated. | Compare two runs with nearly identical starting conditions. Do they diverge over time? If yes → SDIC. |
🔥 The most commonly confused point — Chaos is NOT random
"Deterministic" is what separates chaos from randomness. A random process has no underlying rule — outcomes can't be predicted even with perfect knowledge.A chaotic system is perfectly deterministic — if you knew the initial conditions with infinite precision, you could predict the future exactly. But you can't in practice because tiny measurement errors amplify exponentially (SDIC).
⚖️ Chaos vs Random vs Periodic
🗝 Mechanism of chaos: stretching and folding (Class 17)
Chaos arises through stretching and folding — the same mechanism as kneading dough:1. Stretch — nearby points get pulled apart (SDIC)
2. Fold — the stretched region folds back on itself (keeps system bounded)
This is why chaos is bounded AND sensitive: stretching creates SDIC, folding keeps trajectories from escaping to infinity. Bennoun showed this with the Mona Lisa example — after many iterations, the image becomes unrecognizable but stays within the same region.
Side-by-side comparison
| Property | Periodic | Chaotic | Random |
|---|---|---|---|
| Deterministic? | Yes | Yes | No |
| Repeats exactly? | Yes (fixed period) | Never | No |
| Bounded? | Yes | Yes | Not necessarily |
| Short-term predictable? | Yes | Yes (if IC known precisely) | No |
| Long-term predictable? | Yes | No (SDIC) | No |
| Poincaré plot | Finite set of points | Smooth curve (function) | Random scatter |
| Underlying structure? | Yes (regular) | Yes (strange attractor) | No |
🛤️ Period Doubling Route to Chaos
How chaos arises — the cascade
As a parameter $r$ increases, the system goes through successive doublings of its cycle length before becoming fully chaotic:
$$\text{fixed point}\to\text{period 2}\to\text{period 4}\to\text{period 8}\to\cdots\to\text{chaos}$$
Each doubling happens faster and faster (Feigenbaum constant ≈ 4.669). At some critical value, the doublings happen infinitely fast → chaos.
What a bifurcation diagram shows
• x-axis: parameter value (e.g., $r$)• y-axis: the long-term values the variable takes
• Single line: fixed point (period 1)
• Splits to 2 branches: period 2
• 4 branches: period 4
• Dense/filled region: chaos
• Narrow windows of white within chaos: brief periodic windows (period 3, etc.)
🦗 The Logistic Map
The model
$$X_{N+1}=rX_N(1-X_N)$$
Equilibrium: $X^*=0$ or $X^*=\frac{r-1}{r}$ (the non-trivial one).| $r$ range | Long-term behavior |
|---|---|
$0| Population dies to 0 | |
$1| Stable fixed point at $X^*=(r-1)/r$ | |
| $r\approx 3$ | Period-doubling bifurcation → period 2 |
| $r\approx 3.45$ | Period 2 → period 4 |
| $r\approx 3.54$ | Period 4 → period 8, etc. |
| $r>3.57$ | Chaos (with periodic windows) |
Manual calculation ($r=2$, $X_0=0.95$)
$X_1 = 2(0.95)(1-0.95) = 2(0.95)(0.05) = 0.095$$X_2 = 2(0.095)(0.905) = 0.172$
$X_3 = 2(0.172)(0.828) = 0.285$
...continues converging to $X^*=(2-1)/2=0.5$
Poincaré plot: chaos vs random
⚠ Schematic illustration only — not drawn to scale or from simulation.
📍 Poincaré Plots
What is a Poincaré plot and how to read it
Plot $(X_N, X_{N+1})$ — each consecutive pair of values from your time series.| What you see | What it means |
|---|---|
| Single point | Fixed point (period 1) |
| Exactly 2 points | Period 2 cycle |
| $k$ discrete points | Period $k$ cycle |
| Smooth curve (e.g., a parabola) | Chaotic — deterministic rule $X_{N+1}=f(X_N)$ exists |
| Random scatter, no pattern | Random — no deterministic rule |
🗝 Why the smooth curve means chaos, not random
If the system is deterministic, each $X_N$ uniquely determines $X_{N+1}$ via the rule $X_{N+1}=f(X_N)$. This creates a function — and a function graphed as $(X_N, X_{N+1})$ produces a smooth curve.Random processes have no such rule — $X_N$ doesn't predict $X_{N+1}$ at all → the plot is just noise.
📦 Minimum Dimensions for Each Behavior
The Poincaré-Bendixson theorem explains everything
| Behavior | Min. continuous-time dimensions | Min. discrete-time dimensions | Why |
|---|---|---|---|
| Stable EP | 1 | 1 | Any 1D system can have a stable fixed point |
| Stable oscillations (limit cycle) | 2 | 1 | Need a 2D plane for a closed orbit |
| Chaos | 3 | 1 | Poincaré-Bendixson: in 2D continuous, trajectories can't cross → no chaos. 3D removes this constraint (Lorenz attractor). Discrete 1D can be chaotic (logistic map). |
🔥 The most tested fact: chaos requires 3D continuous
Poincaré-Bendixson theorem says: in 2D continuous-time systems, bounded trajectories must tend to either a fixed point or a limit cycle. Chaos is impossible. Add one more dimension → chaos can occur. The Lorenz attractor (chaotic) lives in 3D.
✦ Unit 5 Practice Quiz
✦ Quiz — Deterministic Chaos
Q1. Which is NOT one of the 4 characteristics of deterministic chaos?
Q2. Why does continuous-time chaos require at least 3 differential equations?
Q3. A Poincaré plot of a time series shows a smooth parabolic curve. This means the system is:
Q4. The sequence 1.2, 0.8, 1.35, 0.6, 1.42, 0.55... comes from a deterministic model. It's bounded between 0.5 and 1.5, never repeats exactly. Two nearby starting conditions diverge exponentially. Is it chaotic?
Q5. In the logistic map $X_{N+1}=rX_N(1-X_N)$, chaos first appears at approximately: