Foundations — Chaotic systems — sensitivity to initial conditions, Lyapunov exponents

Before you can read a single formula on the parent topic, you need to own every symbol it throws at you. This page builds each one from absolutely nothing — plain words, then a picture, then why the topic can't live without it. Read top to bottom; each block leans on the one above.

0. A "state" and a "trajectory" — the stage everything happens on

Here $x$ is not "a distance on a line". It is a point that may live in many dimensions at once. If the ball needs position and velocity, then $x$ is a pair $(\text{position},\text{velocity})$ — one dot in a 2-D room.

Figure s01 — a state is a dot, its motion is a curve. The picture below plots position (horizontal) against velocity (vertical); the orange dot is the state at $t=0$, the green dot is the same dot later, and the blue curve is the trajectory it sweeps out. Alt text: a spiral blue curve with a labelled starting dot and a later dot.

Why the topic needs this. The whole subject is about two trajectories that start almost on top of each other. If you don't picture a state as a dot moving along a curve, the phrase "nearby trajectories separate" is meaningless. This room-of-states is exactly what Phase space and flows studies.

1. $t$ — time, and what "$t\to \infty$" is asking

You'll see ${\lim }_{t\to \infty }$. In plain words that is: "follow the motion forever and ask what number it settles to." It's not a mystical infinity — it's "wait long enough that one lucky starting spot stops mattering."

Why the topic needs it. The Lyapunov exponent must describe the system, not one specific 5-second clip. Averaging over all of time ($t\to \infty$) washes out the accident of where you started.

2. $\delta$ (delta) — "a tiny gap"

Picture two dots released a hair apart. Draw the little arrow that points from one dot to the other — its length is $|\delta |$ (the vertical bars mean "length of", we'll define them next). At $t=0$ that arrow is ${\delta }_0$; later it's $\delta (t)$. Because a state can be multi-dimensional, $\delta$ is itself an arrow (a vector), not just a single number — it has both a length and a direction in the room of states.

Figure s02 — the gap arrow grows with time. Two runs (blue and orange) start a hair apart; the red double-headed arrows measure the gap $\delta$ at three moments — tiny ${\delta }_0$ at the left, visibly wider $\delta (t)$ later. Alt text: two nearly-overlapping curves pulling apart, with red gap arrows widening left to right.

Why the topic needs it. Chaos is defined by what happens to this one little arrow. If it shrinks → calm. If it grows → chaos. Every formula on the parent page is a statement about $\delta$.

3. $|\cdot |$ — "the size of", ignoring direction

4. $f$ and $\dot{x}$ — the rule that pushes the dot

Figure s03 — the rule is an arrow at every point. The gray arrows show $f(x)$: at each location, the rulebook prescribes exactly one push. The red dot is a balance point where $f=0$ (no push). Alt text: a grid of small gray arrows spiralling toward a central red dot. This arrow-field is the flow.

Why the topic needs it. This equation is what makes the system deterministic: same position $x$ ⇒ same push $f(x)$ ⇒ same future. Determinism is the surprising half of "deterministic yet unpredictable." The field of little arrows is the flow in Phase space and flows, and its balance points are studied in Fixed points and linear stability analysis.

5. $f^{\prime }(x)$ — the derivative, i.e. "the local stretch factor"

Now the most important newcomer. Why does a derivative appear when we ask how a gap grows?

Reading the number:

• $|f^{\prime }|>1$ → nearby points get pushed apart (stretch).
• $|f^{\prime }|<1$ → nearby points get pulled together (squeeze).
• $|f^{\prime }|=1$ → gap unchanged.

Figure s04 — the derivative as a multiplier. A small green "input nudge" produces a red "output change" equal to $f^{\prime }(x)$ times the nudge; the orange line is the local slope of the blue rule. Alt text: a curve with its tangent line, a horizontal green input arrow and a vertical red output arrow.

Why the topic needs it. Step 2 of the parent's derivation, $\dot{\delta }\approx f^{\prime }(x)\delta$ (or $Df(x)\delta$ in higher dimensions), is just "gap's growth rate = local stretch factor × gap." Without owning $f^{\prime }$ as a stretch multiplier, that line is hieroglyphics.

6. $e^{\lambda t}$ — the exponential, and why only it fits

• If $\lambda >0$: $e^{\lambda t}$ explodes upward → gap widens → chaos.
• If $\lambda =0$: $e^0=1$, gap stays put → periodic/marginal.
• If $\lambda <0$: $e^{\lambda t}$ decays to zero → gap dies → calm attractor.

7. $\lambda$ — the Lyapunov exponent, packaged at last

We extract it with two moves you now understand: $\lambda ={\lim }_{t\to \infty }{\lim }_{{\delta }_0\to 0}\frac{1}{t}\ln \frac{|\delta (t)|}{|{\delta }_0|}.$

The two limits, in words:

• ${\delta }_0\to 0$ first: keep the starting gap infinitely tiny so the "only-the-slope-matters" approximation ($f^{\prime }$) stays honest.
• $t\to \infty$ second: average over the whole journey so $\lambda$ describes the system, not one lucky spot.

8. $\Delta$ — the tolerance, and the predictability time

Once you pick $\Delta$, the predictability time is just "how long until the growing gap reaches your threshold": $\Delta ={\delta }_0{e}^{\lambda t_{\text{pred}}}\Rightarrow t_{\text{pred}}\approx \frac{1}{\lambda }\ln \frac{\Delta }{{\delta }_0}.$ This inverts the exponential (using $\ln$) to solve for the time $t_{\text{pred}}$ at which $\delta$ first hits $\Delta$. Because $\ln$ grows so slowly, shrinking your starting error ${\delta }_0$ a thousandfold barely nudges $t_{\text{pred}}$ — the brutal logarithm behind "weather is hopeless past ~2 weeks."

9. Bringing it together — the prerequisite map

Everything downstream — SDIC, the predictability time $t_{\text{pred}}=\frac{1}{\lambda }\ln (\Delta /{\delta }_0)$, the Lyapunov spectrum, strange attractors — is built out of these boxes. The stretch-and-shrink balance feeds directly into Liouville's theorem (phase-space volume) and Strange attractors and fractal dimension; the discrete cousin $|f^{\prime }(x_n)|$ powers The logistic map and period-doubling route to chaos; and the bounded-yet-chaotic picture appears in the Driven damped pendulum and Hamiltonian chaos and the KAM theorem.

Equipment checklist

Test yourself — say the answer out loud before revealing.