Visual walkthrough — Properties — row operations, multiplicativity

4.5.22 · D2 · Maths › Linear Algebra (Full) › Properties — row operations, multiplicativity

Kisi bhi algebra se pehle, hum ek picture par agree karte hain jis par sab kuch tika hua hai.

---

Step 1 — Ek matrix picture ke saath kya kar rahi hai?

KYA. Sabse simple square shape lo: unit square — ek square jo ek unit chauda aur ek unit lamba hai, apne corner ke saath origin par. Uske do edges arrows hain ${\mathbf{e}}_1=(1,0)$ (dayi taraf point karta hua) aur ${\mathbf{e}}_2=(0,1)$ (upar point karta hua).

Ek $2\times 2$ matrix $A=\left(\begin{array}{cc}a& b\\ c& d\end{array}\right)$ ek machine hai: yeh dayi-taraf-wale arrow ${\mathbf{e}}_1$ ko pehle column $(a,c)$ par bhejta hai aur upar-wale arrow ${\mathbf{e}}_2$ ko doosre column $(b,d)$ par.

• $a,c$ — right edge kahan land karti hai (uska naya $x$ aur $y$).
• $b,d$ — up edge kahan land karti hai.

Unit square un do nayi arrows se bana hua ek tilted parallelogram ban jaata hai.

KYUN. Determinant ka poora point ek single number hai: $\det A=\text{us parallelogram ka signed area}.$ "Signed" ka matlab hai: positive agar do arrows apna original anticlockwise turn rakhte hain, negative agar machine ne unhe mirror image mein flip kar diya. Neeche sab kuch bas yahi hai: jab main rows ke saath khelun toh is area ka kya hota hai?

PICTURE.

---

Step 2 — Teen seed facts (draw kiye gaye, assume nahi)

Sab kuch teen pictures se aata hai. Hum har ek ko ek saada naam dete hain.

YEH TEEN KYUN? Yeh "signed area" ka minimum honest description hai. (N) unit fix karta hai. (A) kehta hai ki ek squashed-flat shape ka koi area nahi. (M) kehta hai ki area proportion mein badhta hai is baat se ki tum ek edge ko kitna stretch karte ho — yahi "area" ka matlab hai. Dekho Multilinear Alternating Forms.

PICTURE.

(M) ko term by term annotate karte hue, pehle slot mein: $\underset{\text{tall combined edge ka area}}{\underbrace{D(u+w,v)}}=\underset{u\text{ se area}}{\underbrace{D(u,v)}}+\underset{w\text{ se area}}{\underbrace{D(w,v)}}.$ Har piece same doosra arrow $v$ rakhta hai (shared base), toh areas literally stack karte hain.

---

Step 3 — Do rows swap karo: sign flip ho jaata hai (seeds se derive kiya)

KYA. Do arrows exchange karo: $D(u,v)\to D(v,u)$. Hum dikhate hain ki yeh $-D(u,v)$ ke barabar hona chahiye — koi naya axiom nahi chahiye.

KYUN trick. Machine ko ek aisa arrow do jo ek sum hai, $u+v$, dono slots mein. (A) se yeh zero hai (dono slots equal hain). Phir ise (M) se kholte hain:

$$D(\stackrel{\text{slot 1}}{\overbrace{u+v}},\stackrel{\text{slot 2}}{\overbrace{u+v}})=0.$$

Slot 1 expand karo, phir slot 2:

$$\underset{=0\text{ by (A)}}{\underbrace{D(u,u)}}+D(u,v)+D(v,u)+\underset{=0\text{ by (A)}}{\underbrace{D(v,v)}}=0.$$

Do end terms vanish ho jaate hain kyunki har ek mein ek repeated arrow hai. Jo bacha:

$$D(u,v)+D(v,u)=0⟹==D(v,u)=-D(u,v)==.$$

PICTURE. Do parallelogram same shape ke hain, lekin pehle arrow se doosre tak ka turn reverse ho jaata hai — anticlockwise clockwise ban jaata hai. Woh mirror-flip hi minus sign hai.

---

Step 4 — Ek row scale karo: determinant scale hota hai

KYA. Ek edge-arrow ko $k$ factor se stretch karo: $D(k{R}_1,R_2)$.

KYUN. Yeh exactly (M) ka doosra hissa hai: ek scalar ko ek slot se bahar kheenchna.

$$\underset{\text{ek edge }k\text{ times lambi}}{\underbrace{D(k{R}_1,R_2)}}=\underset{\text{stretch factor}}{\underbrace{k}}\underset{\text{original area}}{\underbrace{D(R_1,R_2)}}.$$

Base $R_2$ untouched hai; $R_1$ ke along height $k$ times badi hai, toh area $k$ times bada hai.

PICTURE.

---

Step 5 — Replacement (shear): determinant NAHI badalta

KYA. Ek arrow ka multiple doosre mein add karo: $R_1\to R_1+k{R}_2$. Dikhao ki area unchanged hai.

KYUN. Naye pehle slot ko (M) se split karo:

$$D(\underset{\text{sheared edge}}{\underbrace{R_1+k{R}_2}},R_2)=\underset{\text{original}}{\underbrace{D(R_1,R_2)}}+\underset{\text{= }k\cdot 0\text{ by (A)}}{\underbrace{kD(R_2,R_2)}}.$$

Extra piece mein $R_2$ dono slots mein hai — ek flat line, area $0$. Toh yeh khatam ho jaata hai: $==D(R_1+k{R}_2,R_2)=D(R_1,R_2).==$

PICTURE. Ek arrow ki tip ko doosre ki direction ke along slide karna ek shear hai. Base same rehta hai, height (base line se perpendicular distance) same rehta hai — same area. Jaise cards ki deck ke top ko sideways push karna: same volume.

---

Step 6 — Degenerate case: flat parallelogram ka matlab $\det =0$

KYA. Kya hoga agar do arrows same direction point karein (ek doosre ka multiple hai), $R_2=m{R}_1$?

KYUN. Scaling (Step 4) phir alternating (Step 2) se:

$$D(R_1,m{R}_1)=m\underset{=0}{\underbrace{D(R_1,R_1)}}=0.$$

Parallelogram ek line mein collapse ho gaya — koi area nahi, toh $\det =0$. Yeh picture hai ek singular matrix ke peeche: machine plane ko ek line par squash kar deta hai, ek dimension kho deta hai.

PICTURE.

---

Step 7 — Row reduction diagonal se determinant padh leta hai

KYA. Steps 3–5 use karke $A$ ko ek upper-triangular matrix $U$ tak grind karo (diagonal ke neeche sab zeros), track karo ki har move ki kya cost hai.

KYUN. Ek triangular matrix ke arrows ek aisa parallelogram banate hain jiska signed area bas diagonal entries $u_{ii}$ ka product hota hai (har cofactor expansion mein ek surviving term hoti hai). Aur hum har move ki price jaante hain:

• har swap $\det$ ko $-1$ se multiply karta hai,
• kisi row ko $c$ se divide karna $\det$ ko $\frac{1}{c}$ se multiply karta hai,
• har replacement free hai.

Sab milake, $s$ = swaps ki number aur $\prod k$ un scalars ka product jisse tumne divide kiya:

$$\det A=(-1{)}^s\cdot \frac{1}{\prod k}\cdot \underset{\text{pivots ka product}}{\underbrace{\prod _i{u}_{ii}}}.$$

PICTURE (worked, parent ka Example 1).

\xrightarrow[\text{(free)}]{R_2-2R_1,\;R_3-3R_1} \begin{pmatrix}2&3&1\\0&1&3\\0&3&6\end{pmatrix} \xrightarrow[\text{(free)}]{R_3-3R_2} \begin{pmatrix}2&3&1\\0&1&3\\0&0&-3\end{pmatrix}$$ Koi swaps nahi ($s=0$), koi divisions nahi ($\prod k=1$), pivots $2,1,-3$: $$\det A=(-1)^0\cdot\frac{1}{1}\cdot(2\cdot 1\cdot(-3))=-6.$$ --- ## Step 8 — Multiplicativity: do machines apne areas multiply karte hain **KYA.** Dikhao ki $\det(AB)=\det A\,\det B$. **KYUN (picture-first).** Machine $B$ apply karo: area $\det B$ se scale hota hai. Phir upar $A$ apply karo: woh already-scaled shape phir $\det A$ se scale hoti hai. Do stretches ek ke baad ek **multiply** karte hain:

\underbrace{\det(AB)}{\text{pehle }B\text{ phir }A}=\underbrace{\det A}{\text{2nd stretch}}\cdot\underbrace{\det B}_{\text{1st stretch}}.

**KYUN (algebra).** Har row operation ek [[Elementary Matrices|elementary matrix]] $E$ se left-multiplication hai, aur Steps 3–5 uska determinant exactly dete hain: swap $\det E=-1$, scale $\det E=k$, replacement $\det E=1$. Har case mein $\det(EA)=\det E\,\det A$. Ek invertible $A$ elementaries ka product hai $A=E_1\cdots E_m$, toh $$\det(AB)=\det(E_1)\cdots\det(E_m)\det(B)=\det A\,\det B.$$ Agar $A$ singular hai ($\det A=0$), Step 6 kehta hai uske arrows already collapse ho jaate hain; $AB$ bhi collapse hoga, toh dono sides $0$ hain. **PICTURE.** ![[deepdives/dd-maths-4.5.22-d2-s08.png]] > [!example] Numbers check > $A=\begin{pmatrix}1&2\\3&4\end{pmatrix}$ ($\det=-2$), $B=\begin{pmatrix}2&0\\1&2\end{pmatrix}$ ($\det=4$), $AB=\begin{pmatrix}4&4\\10&8\end{pmatrix}$, $\det(AB)=32-40=-8=(-2)(4)$. ✓ > [!formula] Free corollaries > $\det(A^{-1})=\dfrac{1}{\det A}$, $\;\det(A^k)=(\det A)^k$, aur $\det(A^T)=\det A$ (toh saari row pictures columns ke liye bhi kaam karti hain). Aakhri wala [[Eigenvalues — det as product of eigenvalues]] se link karta hai. --- ## Ek-picture summary ![[deepdives/dd-maths-4.5.22-d2-s09.png]] Ek parallelogram, chaar fates: **swap** use flip karta hai (sign), **scale** use badhata hai (factor $k$), **shear** use slide karta hai (koi change nahi), **collapse** use flat kar deta hai ($\det=0$). Row reduction bas inhi ki ek sequence hai jab tak shape ek triangular box nahi ban jaati jiska area tum diagonal se padh sakte ho — aur do machines ko stack karna unke areas ko multiply karta hai. > [!recall]- Feynman: ek 12-saal ke bachche ko batao > Ek matrix ek aisi machine hai jo ek paper square ko ek tilted parallelogram mein stretch karti hai. Determinant bas *area kitna bada hua*, plus ek minus sign agar shape mirror mein flip ho gayi. > - **Swap** karo do edge-arrows ko → same shape, mirror-flipped → area ko minus sign milta hai. > - **Stretch** karo ek edge ko $k$ times lambi → area $k$ times bada ho jaata hai. *Dono* edges stretch karo (yahi $2A$ karta hai) → area ko $k$ milta hai *har* edge ke liye, toh ek square ke liye $\times 4$. > - **Lean** karo ek edge ko doosre ke along sideways (ek shear) → base aur height same rehti hai, toh area bilkul nahi badlta. Isliye tumhe matrix simplify karne ke liye rows add karne ki permission hai — yeh area ke baare mein kabhi jhooth nahi bolta. > - Agar do arrows same direction point karte hain, parallelogram ek line mein squash ho jaata hai — koi area nahi — aur determinant zero hai. > - Ek machine karo, phir doosri: areas bas multiply ho jaate hain. Yahi hai $\det(AB)=\det A\cdot\det B$. > [!recall]- Quick self-test > - Swap minus kyun deta hai? ::: Same parallelogram, opposite handedness (orientation flip). > - Shear free kyun hai? ::: Base aur height unchanged; extra term mein do equal rows hain → area $0$. > - $2\times2$ ke liye $\det(2A)$? ::: $4\det A$ (dono edges double ho jaate hain). > - $\det=0$ kab hota hai? ::: Arrows collinear → flat parallelogram → singular. > - $\det(AB)=?$ ::: $\det A\,\det B$ (stretches multiply karte hain).