4.7.9 · HinglishPartial Differential Equations

Solving heat equation — separation of variables

1,701 words8 min readRead in English

4.7.9 · Maths › Partial Differential Equations


HUM KYA SOLVE KAR RAHE HAIN?


KAISE: scratch se derivation

Step 1 — Separation guess. Maano (space-part times time-part). Yeh step kyun? Agar yeh kaam kare, toh derivatives nicely split ho jaate hain; try karke hum kuch nahi kho rahe — agar is form ka koi solution exist nahi karta, toh hum khud jaanenge.

Tab aur . Substitute karo:

Step 2 — Variables ko separate karo. se divide karo: Yeh step kyun? Ab left side sirf par depend karti hai, right side sirf par. Ek function of ka ek function of ke barabar hone ka ek hi tarika hai sabhi ke liye — dono same constant ke barabar hon. Ise bolte hain. Minus sign kyun? Foresight: hume time mein decaying solutions chahiye, jiske liye with chahiye. Hum baaki sign cases verify karenge ki woh fail karte hain.

Step 3 — Space ODE solve karo (ek eigenvalue problem). Yeh BCs kyun? Kyunki sabhi ke liye, aur (warna trivial), toh chahiye. Usi tarah par.

ka sign test karo (Forecast-then-Verify):

  • , maano : . BCs force karte hain . Sirf trivial. ✗
  • : . ; . Trivial. ✗
  • , maano : . , toh . .

Toh eigenvalues aur eigenfunctions hain:

Step 4 — Time ODE solve karo. Yeh step kyun? Yeh first-order linear hai; iska solution pure exponential decay hai — exactly woh "fading knob."

Step 5 — Product solutions banao aur superpose karo. Har PDE + BCs solve karta hai. Equation linear aur homogeneous hai, toh solutions ke sums bhi solutions hain:

Step 6 — Initial condition fit karo Fourier coefficients se. par exponentials hain: Yeh ek Fourier sine series hai. se multiply karo, se tak integrate karo, orthogonality use karo :

Figure — Solving heat equation — separation of variables

Worked examples


Common mistakes


Recall Feynman: 12-saal ke bachche ko explain karo

Socho ek metal ki stick jo beech mein hot hai aur ends par cold hai, jise tum ice mein duba ke rakho taaki ends ° par rahen. Heat hamesha hot spots se cold spots ki taraf creep karti hai, toh hot middle slowly cool hoti hai aur spread out hoti hai. Hum pretend karte hain ki temperature pattern simple "wave shapes" (humps) se bani hai. Har hump apne timer par fade hota hai — skinny humps (bahut saare wiggles wale) super fast fade hote hain, fat humps slowly fade hote hain. Future predict karne ke liye tum sirf har hump ko uske apne timer par fade hone dete ho aur unhe wapas add kar dete ho. Lamba time baad sirf sabse bada mota hump bachta hai, aur phir woh bhi disappear ho jaata hai. Cold stick. Done.


Flashcards

ke liye separation of variables kaunsa guess se start karta hai?
, ek space function aur ek time function ka product.
substitute karne ke baad, dono sides ko constant set karna kyun justify hota hai?
Ek side sirf par depend karti hai, doosri sirf par; sabhi ke liye barabar ⇒ dono constant ke barabar hain.
Separation constant with kyun liya jaata hai?
Decaying time solutions aur zero BCs satisfy karne wale nontrivial sine eigenfunctions pane ke liye.
, ke eigenvalues aur eigenfunctions kya hain?
, ,
Mode ka time factor kya hai?
.
Dirichlet heat problem ka full solution kya hai?
.
Coefficients ka formula kya hai?
.
Kaunsa mode sabse fast decay karta hai aur kyun?
High (wiggly modes); decay rate .
Lamba time baad rod ki profile kaisi dikhti hai?
Ek single half-sine bump ( mode) jo zero ki taraf shrink ho raha hota hai.
Yahan sines kyun hain, cosines kyun nahi?
Zero (Dirichlet) end conditions; cosines par nonzero hain, violate karte hain.
kaunsa orthogonality relation deta hai?
.

Connections

Concept Map

models

has

has

guess

substitute and divide

gives

gives

force

only lambda greater than 0 works

eigenfunctions

solve

product

product

linear superposition

Fourier coefficients

Heat equation u_t = a^2 u_xx

Curvature drives heating

BCs u=0 at ends

Initial temp f of x

Assume u = X x times T t

Both sides equal constant -lambda

Space ODE X'' + lambda X = 0

Time ODE T' = -lambda a^2 T

Eigenvalues n pi over L squared

X_n = sin n pi x over L

T_n = exp -a^2 lambda_n t decay

u_n solutions

u = sum B_n u_n