Visual walkthrough — Chaotic systems — sensitivity to initial conditions, Lyapunov exponents

2.1.25 · D2 · Physics › Analytical Mechanics › Chaotic systems — sensitivity to initial conditions, Lyapuno

Hum sirf yeh assume karte hain ki tum ek graph padh sakte ho aur tumne trajectory ka idea dekha hai (Phase space and flows se) — yaani woh path jo ek system time ke saath trace karta hai.

---

Step 1 — Do dots jo almost saath shuru hote hain

KYA. Ek aisi rule picture karo jo batati hai ki dot aage kahan jayegi. Do dots daalo, almost ek hi jagah. Pehle dot ki position time $t$ par $x(t)$ rakho, aur doosre dot ki position $x^{\prime }(t)$. Unke beech ka chhota arrow-gap hi woh cheez hai jise hum dekhte hain.

KYUN. Chaos kabhi ek trajectory ke baare mein nahi hota — ek akela path kisi cheez ke liye "sensitive" nahi ho sakta. Sensitivity do neighbours ke beech ka statement hai. Isliye sabse pehla object jo hum banate hain woh gap hai, path nahi.

PICTURE. Step Figure 1 dekho. Teal dot $x(t)$ hai, plum dot $x^{\prime }(t)$ hai. $t=0$ par woh ek baal jitne nazdik hain; unke beech ka burnt-orange segment gap hai.

Symbol $\delta$ (Greek "delta") yahan "ek chhota difference" matlab rakhta hai। $|\cdot |$ ka matlab hai "ki length" — hamesha ek positive number, kyunki distance negative nahi ho sakti.

---

Step 2 — Poocho ki gap kaise badal rahi hai, yeh nahi ki woh kahan hai

KYA. Hume dots ki absolute positions se matlab nahi. Hume matlab hai ki jaise time beet ta hai, gap $|\delta |$ badh raha hai ya ghatt raha hai. Isliye hum $\delta$ ke rate of change ke baare mein poochte hain.

KYUN. "Rate of change" exactly wahi hai jo ek derivative batata hai — aur yeh sahi tool hai kyunki hamara poora sawaal "kitni tez?" hai, "kitna?" nahi. $\dot{\delta }$ likhna (kisi symbol ke upar dot ka matlab hai "uska rate of change per unit time") vague phrase "gap badh raha hai" ko kuch compute-karne-layak bana deta hai.

PICTURE. Step Figure 2 un hi do dots ko ek instant baad dikhata hai. Dono apne apne arrows (unki velocities) ke saath aage badhein, aur gap arrow longer ho gaya. Gap mein change dono velocities ka difference hai.

Har dot ek hi rule $\dot{x}=f(x)$ maanta hai — ise padhein "ek dot ki velocity ek fixed function $f$ hai uski current position ki." Toh:

---

Step 3 — Zoom in karo jab tak curve ek straight line na bane

KYA. Dono dots bahut nazdik hain, isliye $\delta$ tiny hai. Ek tiny gap par, rule $f$ kuch fancy nahi kar sakta — yeh ek seedhi ramp jaisi dikhti hai. Hum $x$ ke paas $f$ ki curve ko uski tangent line se replace karte hain.

KYUN. Yeh linear approximation hai (Taylor expansion ka pehla term). Hum ise isliye use karte hain kyunki ek tiny separation sirf $f$ ka local slope feel kar sakti hai — us se door ke bends nahi. $x$ par $f$ ka slope $f^{\prime }(x)$ likha jata hai; prime ${}^{\prime }$ ka matlab hai "$f$ ka derivative", yaani $f$ kitni tezi se badhta hai jaise $x$ daayein jaata hai.

PICTURE. Step Figure 3 $f$ ke graph mein zoom karta hai. Tiny window $[x,x+\delta ]$ par curve (plum) aur uski tangent (burnt orange) mein koi fark nahi dikh ta. Window mein height difference slope × width hai.

---

Step 4 — Constant stretch rate se exponential milta hai

KYA. Ek pal ke liye maano ki slope ek fixed number hai, $f^{\prime }(x)=\lambda$. Tab gap maanta hai $\dot{\delta }=\lambda \delta$: "gap ki growth rate gap ke proportional hai." Woh ek hi function hai jiska rate of change apne aap ka constant multiple hota hai — woh hai exponential $e^{\lambda t}$.

KYUN. Hum $e^{\lambda t}$ isliye laate hain kyunki yeh "apni size ke proportion mein kya badhta hai?" ka unique answer hai — fancy dikhne ke liye nahi. Fixed interest par paisa double hona, bacteria ka divide hona, aur trajectories ka alag hona — sab same equation hai.

PICTURE. Step Figure 4 teen slopes ke liye gap length versus time plot karta hai: $\lambda >0$ (plum, blast hota hai), $\lambda =0$ (teal, flat), $\lambda <0$ (burnt orange, decay hota hai).

---

Step 5 — Rate nikalo: log lo, time se divide karo

KYA. Hamare paas $|\delta (t)|=|{\delta }_0|e^{\lambda t}$ hai lekin hum andar chhupa hua number $\lambda$ chahte hain. $|{\delta }_0|$ se divide karo, exponential ko hatane ke liye natural log lo, phir $t$ se divide karo.

KYUN. Logarithm $\ln$ exactly woh tool hai jo $e$ ko undo karta hai: $\ln (e^{\lambda t})=\lambda t$. Yahi wajah hai ki woh yahan aata hai — yeh sawaal hai "kis exponent ne yeh produce kiya?". $t$ se divide karna phir total stretch ko per-unit-time rate mein convert karta hai.

PICTURE. Step Figure 5 same teen curves ko log vertical axis par dikhata hai. Ab woh straight lines hain, aur $\lambda$ literally har line ka slope hai — "average stretch per unit time" ki picture.

$\frac{|\delta (t)|}{|{\delta }_0|}=e^{\lambda t}\Rightarrow \ln \frac{|\delta (t)|}{|{\delta }_0|}=\lambda t\Rightarrow \boxed{\lambda =\frac{1}{t}\ln \frac{|\delta (t)|}{|{\delta }_0|}}$

• $\frac{|\delta (t)|}{|{\delta }_0|}$ — gap kitni baar bada hua (ek pure ratio, koi units nahi).
• $\ln (\cdots )$ — us "×-factor" ko "+-amount" mein convert karta hai (woh exponent jo ise banaya).
• $\frac{1}{t}$ — us exponent ko beete hue time par evenly spread karta hai → ek rate.

---

Step 6 — Do limits kyun, aur bilkul is order mein

KYA. Real slopes ek wiggly orbit par constant nahi hote, aur $\delta$ hamesha infinitesimal nahi rahega. Isliye final definition boxed result ko do limits mein wrap karti hai: $\lambda ={\lim }_{t\to \infty }{\lim }_{{\delta }_0\to 0}\frac{1}{t}\ln \frac{|\delta (t)|}{|{\delta }_0|}.$

KYUN.

• $\underset{{\delta }_0\to 0}{\lim }$ pehle — Step-3 ki straight-line picture exact rakhta hai. Agar gap kabhi bada ho gaya, toh hamari tangent-line approximation toot jaati aur number bekaar hota.
• $\underset{t\to \infty }{\lim }$ baad mein — hamesha badalta local slope poori orbit par average karta hai, taaki $\lambda$ system describe kare, koi ek lucky starting point nahi.

PICTURE. Step Figure 6: local slope $f^{\prime }(x)$ bahut wildly badlata hai jaise dot orbit par chalta hai (jagged burnt-orange line); us saari jitter ka long-time average ek single flat teal level hai — woh level hi $\lambda$ hai.

---

Step 7 — Woh edge cases jo picture mein bhi cover hone chahiye

KYA & PICTURE. Step Figure 7 degenerate behaviours ko char mini-panels mein collect karta hai, har ek gap-vs-time sketch hai.

Case	Gap kya karta hai	$\lambda$ ka sign	Example
Stable fixed point	dot aur neighbour dono same rest point par girate hain	$\lambda <0$	damped pendulum at rest
Periodic / quasi-periodic	gap oscillate karta hai, kabhi nahi badhta	$\lambda =0$	frictionless pendulum
Chaotic lekin bounded	gap badhta hai phir fold back hota hai	${\lambda }_1>0$, $\sum {\lambda }_i<0$	Driven damped pendulum
Unbounded blow-up	gap hamesha badhta hai, koi folding nahi	$\lambda >0$ lekin chaos nahi	$\dot{x}=x$

---

Ek-picture summary

Step Figure 8 saatein steps ko ek frame mein compress karta hai: do dots alag hote hain (Step 1–2), local slope gap ko stretch karta hai (Step 3), exponential ise badhata hai (Step 4), log-slope $\lambda$ padhta hai (Step 5), orbit-averaging jitter ko flat karta hai (Step 6), aur fold ise bounded rakhta hai (Step 7).

Recall Feynman retelling — poora walkthrough simple shabdon mein

Do toy cars ko ek bumpy track par almost ek hi jagah rakho. Pehle, cars bhulo — gap dekho unke beech (Step 1). Sirf ek cheez poocho: kya gap badh raha hai? Yeh "kitni tez?" wala sawaal hai, isliye hum ek rate use karte hain — ek derivative (Step 2). Kyunki cars itne nazdik hain, gap ke neeche track ek seedhi ramp jaisi dikhti hai, aur ramp ki steepness hi sab kuch matter karti hai (Step 3). "Apni size ke proportion mein badhta hai" exactly exponential rule hai, isliye gap $e^{\lambda t}$ ki tarah fuulta hai (Step 4). Growth rate $\lambda$ padhne ke liye, exponential ko log se undo karo aur time se divide karo — log-paper par $\lambda$ bas ek straight line ka slope hai (Step 5). Lekin ramp ki steepness tab tak badlati rehti hai jab tak cars track par chalte hain, aur straight-line trick sirf tab kaam karti hai jab gap tiny ho — isliye hum pehle gap tiny lete hain, phir ek lambi, lambi ride par average karte hain (Step 6). Finally: ek positive $\lambda$ matlab gap double hota rehta hai, lekin agar track wapis loop back kare toh cars board par rahe bhi jab unka gap blast ho raha ho — stretch locally, fold globally. Woh, aur sirf wahi, chaos hai (Step 7).

Recall Quick self-checks

$\delta$ (aur $x$ nahi) show ka star kyun hai? ::: Chaos neighbours ke alag hone ka statement hai; ek akeli trajectory "sensitive" nahi ho sakti, sirf dono ke beech ka gap ho sakta hai. Exponential kyun aata hai? ::: $\dot{\delta }=\lambda \delta$ kehta hai gap apni size ke proportion mein badhta hai, aur $e^{\lambda t}$ woh unique function hai jo yeh property rakhta hai. $\ln$ kyun lete hain phir $t$ se divide kyun karte hain? ::: $\ln$ exponential ko undo karta hai taaki exponent $\lambda t$ saamne aaye; $t$ se divide karna ise per-unit-time rate bana deta hai. ${\delta }_0\to 0$ pehle kyun, $t\to \infty$ baad mein kyun? ::: Tiny-gap-first straight-line (linear) approximation exact rakhta hai; long-time-second $\lambda$ ko poore attractor par average karta hai. Kya $\lambda >0$ akela chaos prove karta hai? ::: Nahi — $\dot{x}=x$ mein $\lambda >0$ hai lekin woh unbounded hai. Chaos ko boundedness aur folding bhi chahiye.

---

Related builds: Fixed points and linear stability analysis ($\lambda <0$ case), The logistic map and period-doubling route to chaos (is derivation ka discrete version), Hamiltonian chaos and the KAM theorem ($\lambda =0$ regular islands).