Row echelon form and reduced row echelon form: Visual walkthrough

4.5.10 · D2 · Maths › Linear Algebra (Full) › Row echelon form and reduced row echelon form

Neeche sab kuch ek running example use karta hai, parent ke Worked Example 1 wala same system: $x + 2 y + z = 6, 2 x + 5 y + 3 z = 15, x + 3 y + 3 z = 11.$

Step 1 — Linear equation actually hai kya? (ek picture, formula nahi)

KYA HAI. Sirf pehli equation lo, $x + 2 y + z = 6$ . Letters $x, y, z$ teen unknown numbers hain jinhe hum dhundh rahe hain. Equation ek rule hai: chahe teen numbers kuch bhi hon, jab tum $x$ ki $1$ copy, $y$ ki $2$ copies, aur $z$ ki $1$ copy lo aur add karo, to exactly $6$ milna chahiye.

YEH YAHAN SE KYU SHURU KARTE HAIN. Koi bhi row shuffle karne se pehle, hume dekhna chahiye ki ek equation kya hoti hai. Teen unknowns wali linear equation koi curve nahi hoti — yeh ek flat sheet (plane) hoti hai jo room mein float karti hai. Us sheet ka har point ek valid triple $(x, y, z)$ hai.

PICTURE. Neeche, room ki teen directions $x$ , $y$ , $z$ axes hain. Single red sheet woh saari triples hain jo $x + 2 y + z = 6$ obey karti hain. Gaur karo: ek akeli equation ke liye infinitely many solutions hain — freedom tabhi ghatti hai jab hum aur rules add karte hain.

Step 2 — Teen equations = teen sheets; ek solution = jahan woh sabhi milti hain

KYA HAI. Hamare system mein teen equations hain, to teen planes hain. Ek triple $(x, y, z)$ poore system ko solve karta hai tabhi jab woh teeno sheets par ek saath ho.

KYU. Yahi "solve the system" ka geometric matlab hai: common point(s) dhundhna. Teen generic planes ke liye overlap ek single point hota hai — ek $x$ , ek $y$ , ek $z$ . Wahi point hai jise hum dhundh rahe hain.

PICTURE. Teen sheets room mein slice kar rahi hain. Red dot unka unique meeting point hai. Ab se hum jo bhi karenge woh sirf sheets ko idhar-udhar tilate aur slide karte hain bina us red dot ko hilaye — yahi poora trick hai.

Step 3 — Sheets ko numbers ke grid mein package karo (augmented matrix)

KYA HAI. $x, y, z, =$ baar baar likhna wasteful hai — positions pehle se hi batati hain ki koi number kis unknown ka hai. Letters hata do; coefficients aur constants ko ek grid mein rakh do.

$12125313361511$

YEH TOOL KYU AUR WORDS KE SAATH KYUN NAHI RAHEIN. Ek matrix "teen equations ko manipulate karo" ko "numbers ki teen rows ko manipulate karo" mein badal deta hai. Vertical bar ek fence hai: iske baaye coefficients hain (sheet ka tilt), daaye constants hain (sheet origin se kitni door float karti hai).

PICTURE. Har row apne plane ke saath colour-label ki gayi hai. Red column, bar ke daaye, constants hai — woh part jo bata raha hai ki har sheet kahan float karti hai.

Step 4 — Ek legal move: ek row subtract karo (kyun dot wahi rehti hai)

KYA HAI. Row 2 lo aur usme se $2 \times$ Row 1 subtract karo: $R_{2} \to R_{2} - 2 R_{1}$ . Equation language mein, equation 2 ban jaati hai (equation 2) minus twice (equation 1).

YEH LEGAL KYU HAI — HAR CHEEZ KI JAR. Agar ek point pehle se dono sheet 1 aur sheet 2 par hai, to woh equation 1 aur equation 2 satisfy karta hai, to woh unka koi bhi combination satisfy karta hai — jisme "equation 2 minus twice equation 1" bhi shamil hai. Aur yeh move reversible hai: $2 R_{1}$ wapas add karo to original mil jaata hai. Reversible ka matlab hai koi solution na banta hai aur na hi koi destroy hota hai — red dot move nahi kar sakta.

PICTURE. Blue sheet (equation 2) ek naye orientation mein tilt hoti hai, lekin phir bhi red dot se guzarti hai. Yahi elimination ka poora justification hai: humein sheets ko unke common pin ke around swing karne ki permission hai.

Row 2 par subtraction term by term kya karta hai: $old R_{2} (2, 5, 3 ∣ 15) - 2 R_{1} (1, 2, 1 ∣ 6) = new R_{2} (0, 1, 1 ∣ 3) .$

Leading $2$ bana $0$ — yahi goal tha: equation 2 se $x$ erase karo.
$5 - 2 (2) = 1$ , $3 - 2 (1) = 1$ , $15 - 2 (6) = 3$ — baaki har entry lockstep mein update hoti hai taaki sheet red dot ko phir bhi hold kare.

Step 5 — Staircase banao (REF tak pahuncho)

KYA HAI. "Subtract to make zeros below" move ko repeat karo, left to right, top to bottom. Step 4 se hum $R_{3} \to R_{3} - R_{1}$ bhi karte hain, phir $R_{3} \to R_{3} - R_{2}$ :

\;\xrightarrow{R_3\to R_3-R_2}\; \left[\begin{array}{ccc|c} \mathbf{1}&2&1&6\\ 0&\mathbf{1}&1&3\\ 0&0&\mathbf{1}&2\end{array}\right]\ \textbf{(REF)}$$ **KYU.** Har column ki pehli non-zero entry (bold **pivot**) us row se neeche apne variable ko anchor karne wali *ek hi cheez* ban jaati hai. Row 3 ab sirf $z$ contain karta hai: $z = 2$. Staircase ka yahi point hai — bottom step ek single unknown ko trap karta hai. **PICTURE.** Dekho pivots (red) **neeche aur daaye** march karte hue, steps bana rahe hain. Har pivot ke neeche: sab zeros. Yahi shape *hai* Row Echelon Form. - Row 3 = $(0,0,1\mid 2)$ ka matlab hai $0x+0y+1z = 2$, yaani $z=2$. **Ek variable, solved.** - Row 2 = $(0,1,1\mid 3)$ ka matlab hai $y + z = 3$; $z=2$ jaante hue $y=1$ milta hai — yahi ==back-substitution== hai. - Row 1 deta hai $x + 2y + z = 6 \Rightarrow x = 2$. Stairs chadhte jao, fill karte jao. > [!definition] Pivot aur REF > ==Pivot== kisi row ki pehli non-zero entry hoti hai. Ek matrix **REF** mein hai jab zero rows bottom par hoon, har pivot strictly usse upar wale ke daaye ho (staircase), aur har pivot ke **neeche** sab kuch $0$ ho. REF *unique* nahi hoti — kai tarah ki staircases (alag pivot values) same dot tak pahunch sakti hain. Rank $=$ pivots ki sankhya; [[Rank of a Matrix]] dekho. --- ## Step 6 — Pivots ke upar clean karo (RREF tak pahuncho, answer padho) **KYA HAI.** Ab **bottom-up** jao: har pivot ko $1$ pe scale karo (yahan pehle se $1$ hai) aur har pivot ke **upar** entries kill karne ke liye subtract karo. $$\left[\begin{array}{ccc|c} 1&2&1&6\\ 0&1&1&3\\ 0&0&1&2\end{array}\right] \xrightarrow[R_1\to R_1-R_3]{R_2\to R_2-R_3} \left[\begin{array}{ccc|c} 1&2&0&4\\ 0&1&0&1\\ 0&0&1&2\end{array}\right] \xrightarrow{R_1\to R_1-2R_2} \left[\begin{array}{ccc|c} \mathbf 1&0&0&2\\ 0&\mathbf 1&0&1\\ 0&0&\mathbf 1&2\end{array}\right]\ \textbf{(RREF)}$$ **KYU.** Jab har pivot apne column mein **ek hi** non-zero hota hai, to har row degenerate ho jaati hai $x = \text{number}$ mein. Koi climbing nahi, koi back-substitution nahi — constants column *hi* answer hai. Yahi [[Gauss-Jordan Elimination]] hai. **PICTURE.** Sheets tilt ki gayi hain jab tak woh coordinate walls ke saath align na ho jaayein: ek sheet hai "$x=2$", ek "$y=1$", ek "$z=2$". Teen perpendicular walls red dot par cross karti hain — answer literally axes se readable hai. Last grid ko row by row padhte hain: $x=2,\ y=1,\ z=2.$ ✅ (Step 2 ka wahi red dot — woh kabhi nahi hila.) --- ## Step 7 — RREF *unique* kyun hoti hai (missing piece) **KYA HAI.** REF unique nahi thi; RREF hai. Kyun? **KYU.** RREF pivot ki **positions** aur pivot ke **column shapes** dono pin down karta hai. Pivot positions forced hain: woh leftmost column jo solution set mein fully zero nahi hai *usme pivot hona hi chahiye* — steps *kahan* jaayenge isme koi choice nahi hai, sirf numbers mein hai, aur RREF numbers bhi fix kar deta hai (pivot $=1$, upar aur neeche zeros). Jis freedom ne REF ko non-unique banaya tha (koi bhi non-zero pivot value, upar koi bhi leftover entries) woh *exactly* wahi hai jo conditions 4–5 hata dete hain. Do log, do alag move-sequences, ek destination. **PICTURE.** *Same* matrix ki do alag staircases (do REFs) — alag pivot values, alag upper entries — lekin jab dono fully clean ho jaati hain, dono *identical* RREF grid par collapse ho jaati hain. > [!definition] RREF (aur uski uniqueness) > **RREF** = REF plus: har pivot $1$ hai, aur har pivot apne column mein *ek hi* non-zero entry hai. Kisi bhi matrix ke liye exactly ==ek== RREF hota hai. Isliye RREF rank, free variables ([[Free and Basic Variables]]), aur [[Matrix Inverse via RREF|inverse]] ko bhi unambiguously *define* kar sakta hai. --- ## Step 8 — Degenerate cases (inhe kabhi skip mat karo) **KYA HAI AUR KYU.** Har system mein ek dot nahi hota. Do cheezein "galat" ho sakti hain, aur staircase *batata hai kaun sa*: **Case A — ek free variable (solutions ki poori line).** Agar pivots unknowns se kam hain, to ek column mein koi pivot nahi hai; woh variable **free** hai. Parent se example: $$\left[\begin{array}{ccc|c} 1&2&0&1\\ 0&0&1&1\end{array}\right]\Rightarrow x=1-2y,\ z=1,\ y\text{ free.}$$ Do planes ek dot mein nahi balki ek *line* mein milti hain — ek degree of freedom bachti hai. **Case B — inconsistency (koi solution nahi).** Agar ek pivot **constants** column mein land kare: $$\left[\begin{array}{cc|c} 1&1&2\\ 0&0&1\end{array}\right]\Rightarrow 0=1.$$ Row 2 kehta hai $0x+0y = 1$: ek impossible sheet, planes parallel hain aur kabhi ek point share nahi karti. Empty solution set. **PICTURE.** Baaye: do sheets ek red line mein cross karti hain (free variable). Daaye: do parallel sheets ek gap ke saath — kaheen koi red dot nahi (inconsistent). > [!mistake] Pivot-in-the-last-column trap > Constants column mein ek pivot ek ordinary pivot jaisa *lagta* hai, lekin yeh decode hota hai $0 = \text{nonzero}$ mein. Last column ke pivots ko hamesha **contradiction alarm** ki tarah padho, variable ki tarah nahi. --- ## Ek-picture summary Ek frame, poora safar: messy sheets (Step 2) → below zero-out karne ke liye subtract karo (Steps 4–5, red staircase banta hai) → upar clean karo (Step 6, sheets axis-walls se snap karti hain) → red dot padho. Bottom strip mein do escape hatches hain: ek leftover line (free variable) ya parallel gap (inconsistent). > [!recall]- Feynman: plain words mein walkthrough > Socho teen badi glass sheets ek room mein float kar rahi hain; jis ek spot par teeno overlap karti hain wahi secret answer hai. Solving ko *tilt* karne ki permission hai sheets ko, lekin sirf aisi tarah se jo unhe us same overlap spot se pinned rakhe — kyunki "ek equation dusre se subtract karo" hamesha undo kiya ja sakta hai, to kuch lost nahi hota. Hum unhe ek neat staircase mein tilt karte hain jahan bottom sheet sirf ek letter mention karti hai (to woh letter humein free mein de deta hai), phir hum upar chadhte hain baaki fill karte hue. Agar hum polish karte rahe jab tak har sheet ek wall ke saath flat align na ho jaaye — ek sheet kehti hai "$x$ yahan hai", ek "$y$ yahan hai", ek "$z$ yahan hai" — to hum chadhte bhi nahi; answer walls par likha hai. Woh perfectly-polished picture hamesha same hoti hai chahe tum kaisa bhi wahan pahunche — isliye RREF unique hai. Aur agar sheets ek dot ki jagah poori line mein milti hain, to koi letter ghoomne ke liye free hai; agar do sheets parallel hain aur kabhi touch nahi karti, to simply koi answer hi nahi hai. --- ## Connections - [[Gaussian Elimination]] — staircase banata hai (Steps 4–5). - [[Gauss-Jordan Elimination]] — pivots ke upar clean karta hai (Step 6). - [[Rank of a Matrix]] — tumhare banaye pivots ki sankhya. - [[Solving Systems of Linear Equations]] — Steps 1–2 ka geometric matlab. - [[Free and Basic Variables]] — Step 8, Case A. - [[Matrix Inverse via RREF]] — RREF ki uniqueness ko kaam mein lagana. - [[Linear Independence]] — pivot columns independent directions hain. - [[Elementary Matrices]] — har legal move ek matrix ke roop mein baaye se.