Second-order linear ODEs — superposition principle, general theory: Visual walkthrough

4.6.9 · D2 · Maths › Ordinary Differential Equations › Second-order linear ODEs — superposition principle, general

Hum kuch symbols use karenge. Pehle, har ek ke baare mein promise:

$y$ ek function hai — ek machine jo ek number $x$ leta hai aur ek number $y (x)$ return karta hai. Ise paper par kheenchi hui ek curve samjho.
$y^{'}$ ka matlab hai "us curve ki slope" (kitni steep hai) aur $y^{''}$ ka matlab hai "slope khud kitni tezi se change ho rahi hai" (bending). Inhe hum Step 1 mein poori tarah earn karte hain.
$L$ ek machine hai jo ek function khaati hai aur ek function return karti hai. Ise hum Step 2 mein banate hain.

Chalo zero se shuru karte hain.

Step 1 — $y^{'}$ aur $y^{''}$ actually kaisi dikhti hain

Hume inki zarurat kyun hai. Ek "ODE" bas ek rule hai jo ek curve ki height, slope, aur bending ko ek saath baandhta hai. Aisa rule padhne ke liye pehle un teen quantities ko ek picture mein dekhna padega. Iske baad ki har cheez unhe pehchanne par depend karti hai.

PICTURE. Figure mein, blue curve $y = cos x$ hai. Marked point par chhota orange segment tangent hai — uska tilt $y^{'}$ hai. Green shading dikhati hai kahan curve neeche ki taraf bend karti hai ( $y^{''} < 0$ ) versus upar ki taraf ( $y^{''} > 0$ ).

Figure — Second-order linear ODEs — superposition principle, general theory

Step 2 — ODE ko ek machine $L$ mein package karna

Har symbol, jahan woh baitha hai:

$y^{''}$ — curve ki bending (Step 1 se).
$p (x)$ — ek given function; ek dial jo weight karta hai ki slope kitna contribute karta hai.
$y^{'}$ — slope.
$q (x)$ — ek given function; ek dial jo height ko weight karta hai.
$y$ — curve ki height.

Box kyun karein. Kyunki phir "ODE solve karna" ek hi sentence ban jaata hai: har woh function $y$ dhundho jise machine target $g$ par bhejengi. Aur machine ki ek magical property hai jo hum draw kar sakte hain.

PICTURE. Figure mein $L$ ko ek box ki tarah dikhaya gaya hai: ek curve $y$ left se andar jaati hai, ek nayi curve $L [y]$ right se bahar aati hai. Alag-alag curves feed karne par alag-alag outputs aate hain.

Step 3 — Magic property: $L$ linear hai

Term by term:

$c_{1}, c_{2}$ — plain numbers (scaling factors, jaise "is curve ko do guna tall karo").
$y_{1}, y_{2}$ — do input curves.
Left side: combined curve ko machine mein feed karo.
Right side: har curve ko alag-alag feed karo, phir outputs combine karo.
Equals sign hi poora claim hai: dono taraf se karo toh same output milta hai.

Yeh sach kyun hai. Kyunki differentiation khud hi yeh maanta hai: ek sum ki slope slopes ka sum hoti hai, aur ek curve ko scale karne se uski slope bhi scale hoti hai. Parent note teen-line algebra dikhata hai; yahan hum bas ise dekhte hain.

PICTURE. Left panel: $y_{1}$ ko $c_{1}$ se stretch karke, $y_{2}$ ko $c_{2}$ se stretch karke, unhe stack karke input banao — phir machine chalao. Right panel: pehle dono ko alag-alag run karo, phir stack karo. Dono red output curves bilkul match karti hain.

Yeh ek identity — "linear map" ki general idea ke liye Linear algebra — vector spaces and bases dekho — neeche ki har cheez ka beej hai.

Step 4 — Superposition: solutions add up hote hain

Do under-braces zero hain kyunki $y_{1}, y_{2}$ solutions maane gaye hain.
Isliye poori cheez zero hai, toh $c_{1} y_{1} + c_{2} y_{2}$ bhi ek solution hai, kisi bhi numbers $c_{1}, c_{2}$ ke liye.

Yeh itna bada kyun hai. Iska matlab hai solutions ko kisi bhi proportion mein mix kar sakte hain aur tum solution hi rahoge. Geometrically: solutions ek flat sheet through the origin par rehte hain — ek plane.

PICTURE. Har solution ko ek abstract "solution space" mein origin se ek arrow ki tarah draw kiya gaya hai. $y_{1}$ ek taraf point karta hai, $y_{2}$ doosri taraf. Shaded plane par har dot $c_{1} y_{1} + c_{2} y_{2}$ ek valid solution hai. Koi bhi plane se bahar nahi jaata.

Step 5 — Independence kyun chahiye (parallel-arrows trap)

$c_{1} + 3 c_{2}$ do knobs ko ek effective knob mein collapse kar deta hai.
Toh tum half solutions miss kar rahe ho.

Yeh kyun matter karta hai. Ek second-order ODE ko do free constants chahiye (Step 8 explain karta hai kyun). Parallel solutions sirf ek dete hain. Hum do non-parallel solutions ko linearly independent kehte hain.

PICTURE. Left: do parallel arrows — saare combinations ek line par rehte hain (bura). Right: do spread-apart arrows — combinations plane bhar dete hain (achha).

Step 6 — Wronskian: ek machine jo parallelness detect karti hai

$W$ ko Wronskian kehte hain.
$W (x_{0}) \neq = 0$ kisi bhi ek point par $\Rightarrow$ koi hidden relation nahi $\Rightarrow$ arrows independent hain.
$W$ measure karta hai ki $y_{1}, y_{2}$ height-and-slope plane mein kitna "spread apart" hain.

Determinant kyun. Quantity $y_{1} y_{2}^{'} - y_{2} y_{1}^{'}$ bilkul signed area hai us parallelogram ka jo do vectors $(y_{1}, y_{1}^{'})$ aur $(y_{2}, y_{2}^{'})$ se bana hai. Zero area = squashed flat = parallel = dependent. Non-zero area = genuine spread = independent. Linear algebra — vector spaces and bases dekho.

PICTURE. Do vectors $(y_{1}, y_{1}^{'})$ aur $(y_{2}, y_{2}^{'})$ tail-to-origin draw kiye gaye hain; unke beech shaded parallelogram; uski area hi $∣ W ∣$ hai. Jab woh line up ho jaate hain toh parallelogram ek segment mein collapse ho jaata hai jis ki area $0$ hoti hai.

Step 7 — Edge case: all-or-nothing (Abel's theorem)

Kisi bhi real number ka exponential strictly positive hota hai — woh kabhi zero nahi hota.
Toh $W (x)$ , $W (x_{0})$ ka strictly positive factor se multiply hota hai: agar $W (x_{0}) \neq = 0$ toh $W$ kabhi bhi zero nahi hoga; agar $W (x_{0}) = 0$ toh $W$ hamesha zero hoga.

Hume kyun parwah hai. Tum Wronskian sirf ek convenient point par check karte ho. Woh chhoti equation kahan se aati hai uske liye Abel's theorem aur First-order linear ODEs dekho.

PICTURE. Exponential envelope curve $exp (- \int p)$ axis ke upar strictly rehti hai; $W (x)$ (blue) woh curve hai jo ek constant se scale ki gayi hai, toh woh ya toh poori tarah axis ke upar ya poori tarah neeche rehti hai — woh cross nahi kar sakti.

Step 8 — Exactly do constants kyun, aur forced case

$y - y_{p}$ homogeneous hai $\Rightarrow$ kisi $c_{1} y_{1} + c_{2} y_{2}$ ke barabar hai.
Toh har forced solution = (poora homogeneous plane) single vector $y_{p}$ se shifted.

Yeh origin se guzarne wala plane kyun nahi hai. Forced solutions same plane banate hain, $y_{p}$ se origin se utha hua — ek affine plane. Do initial conditions $y (x_{0}) = a, y^{'} (x_{0}) = b$ phir ek exact point pin kar deti hain (guaranteed unique by Existence and uniqueness theorems for ODEs).

PICTURE. Blue homogeneous plane origin se guzarta hai; orange forced plane ek parallel copy hai jo green vector $y_{p}$ se shift ki gayi hai. Ek red dot woh unique solution mark karta hai jo initial conditions se select hoti hai.

Practice mein har piece kahan se aati hai: $y_{1}, y_{2}$ Characteristic equation — constant coefficient ODEs se, aur $y_{p}$ Method of undetermined coefficients ya Variation of parameters se.

Ek-picture summary

Poori logical chain: $L$ ki linearity $\Rightarrow$ solutions add hote hain (ek plane) $\Rightarrow$ do independent arrows chahiye $\Rightarrow$ Wronskian independence detect karta hai $\Rightarrow$ Abel ise all-or-nothing banata hai $\Rightarrow$ forced case wahi plane hai $y_{p}$ se shift ki gayi.

Recall Feynman retelling — ise ek story ki tarah bolo

ODE ek machine hai. Isme ek curve daalo, woh tumhe ek curve wapas deti hai. Machine fair hai: input stretch karo, output utna hi stretch hota hai; do inputs add karo, outputs bhi add hote hain. Kyunki woh fair hai, agar do curves dono flat zero ho ke bahar aati hain, toh unka koi bhi mix bhi zero ho ke bahar aayega — toh saare "zero solutions" ek flat sheet par rehte hain. Us poori sheet ko bikhaarne ke liye tumhe do curves chahiye jo same direction mein point na karein; agar ek doosre ki scaled copy hai toh tum sirf ek line paate ho aur half miss kar rahe ho. Yeh check karne ke liye ki woh sach mein alag-alag direction mein point karte hain, vectors $(y_{1}, y_{1}^{'})$ aur $(y_{2}, y_{2}^{'})$ ke beech ki chhoti area compute karo — wahi Wronskian hai. Agar woh area kahin bhi non-zero hai, Abel ka rule kehta hai woh har jagah non-zero hai, toh ek check kaafi hai. Aakhir mein, agar machine se kuch non-zero output maanga jaaye, toh bas ek curve dhundho jo kaam kare aur poori flat sheet ko uske upar shift kar do; do initial conditions phir ek ungli se exact point par point karti hain jo tum chahte ho.

Recall Khud test karo

Tum homogeneous solutions ko scale-and-add kyun kar sakte ho lekin forced ones ko nahi? ::: $L [y] = 0$ ke liye, $L [c_{1} y_{1} + c_{2} y_{2}] = c_{1} \cdot 0 + c_{2} \cdot 0 = 0$ . $L [y] = g$ ke liye, tumhe $(c_{1} + c_{2}) g$ milega, jo $g$ ke barabar sirf tab hoga jab $c_{1} + c_{2} = 1$ — toh free scaling kaam nahi karta. Wronskian geometrically kya measure karta hai? ::: Vectors $(y_{1}, y_{1}^{'})$ aur $(y_{2}, y_{2}^{'})$ se bane parallelogram ka signed area; zero area = parallel = dependent. ODE solutions ke liye $W$ ek point par check karna kyun kaafi hai? ::: Abel: $W (x) = W (x_{0}) exp (- \int p)$ , aur exponential kabhi zero nahi hota, toh $W$ all-or-nothing hai. Forced solutions kis shape mein hote hain? ::: Origin se guzarne wala homogeneous plane, kisi bhi ek particular solution $y_{p}$ se shifted (translated) — ek affine plane.