Properties — row operations, multiplicativity

---

WHAT are we studying?

---

WHY these rules hold — derivation from scratch

We take ONE foundational fact and derive everything. The determinant is the unique function $D(R_1,\dots,R_n)$ of the rows that is:

• (M) Multilinear: linear in each row separately,
• (A) Alternating: $D=0$ whenever two rows are equal,
• (N) Normalized: $D(I)=1$.

Everything below follows.

Rule 1: Swapping two rows flips the sign

Take rows $u$ and $v$. Consider the matrix with rows $\dots, u+v, \dots, u+v, \dots$ in positions $i$ and $j$. Since two rows are equal: $D(\dots,u+v,\dots,u+v,\dots)=0.$ Expand by multilinearity in both positions: $D(u,u)+D(u,v)+D(v,u)+D(v,v)=0.$ But $D(u,u)=0$ and $D(v,v)=0$ (equal rows). So: $D(u,v)+D(v,u)=0 \;\Rightarrow\; ==D(v,u)=-D(u,v)==.$

Why this step? We only used "two equal rows give 0" plus linearity — that's all it takes. The sign flip is not an extra axiom; it's a consequence.

Rule 2: Scaling a row scales the determinant

By multilinearity (linearity in row $i$): $D(\dots, kR_i,\dots) = ==k\,D(\dots,R_i,\dots)==.$ Why? Linear in that slot means $f(k\mathbf{x})=kf(\mathbf{x})$ directly.

A consequence: $\det(kA)=k^n\det A$ for an $n\times n$ matrix (every one of the $n$ rows gets scaled).

Rule 3: Replacement leaves the determinant unchanged

Replace $R_i \to R_i + kR_j$ (with $i\neq j$). Use multilinearity in slot $i$: $D(\dots, R_i+kR_j, \dots) = D(\dots,R_i,\dots) + k\,D(\dots,R_j,\dots).$ The second term has $R_j$ sitting in both slot $i$ and slot $j$ → two equal rows → it's $0$. $\Rightarrow \;==D(\dots,R_i+kR_j,\dots)=D(\dots,R_i,\dots)==.$

Why this step? This is the workhorse for Gaussian elimination: you can clear out entries without ever changing the determinant.

---

---

Computing $\det$ via row reduction

---

WHY multiplicativity is true

Algebraic derivation

Each row operation on $A$ = multiplying $A$ on the left by an elementary matrix $E$:

• Swap $E$: $\det E = -1$.
• Scale-by-$k$ $E$: $\det E = k$.
• Replacement $E$: $\det E = 1$.

From Rules 1–3, applying $E$ to $A$ gives $\det(EA)=\det(E)\det(A)$ — true for every elementary $E$.

Now factor: if $A$ is invertible, $A = E_1E_2\cdots E_m$ (product of elementaries). Then $\det(AB)=\det(E_1\cdots E_m B)=\det(E_1)\cdots\det(E_m)\det(B)=\det(A)\det(B).$ If $A$ is singular, $\det A=0$ and $AB$ is also singular (rank $\le$ rank $A < n$), so $\det(AB)=0=\det A\det B$. $\blacksquare$

---

Worked examples

---

Common mistakes (steel-manned)

---

Active recall

Recall Self-test (cover the answers)

• Effect of a row swap on $\det$? → multiply by $-1$.
• Effect of $R_i\to R_i+kR_j$? → none.
• $\det(kA)$ for $n\times n$? → $k^n\det A$.
• $\det(AB)=$? → $\det A\,\det B$.
• $\det(A^{-1})=$? → $1/\det A$.

Flashcards

---

Recall Feynman: explain to a 12-year-old

Imagine a stretchy rubber square. A matrix is a machine that stretches and tilts it into a parallelogram, and the determinant is just how many times bigger the area got (with a minus sign if the shape got flipped over like a mirror).

• If you swap two of the machine's instructions, the shape flips over → minus sign.
• If you double one instruction, that direction stretches twice as much → area doubles.
• If you slide one side along another (a shear/lean), the parallelogram leans but keeps the same area → no change. That's why "add a multiple of one row to another" does nothing to the determinant.
• If you run machine $B$ then machine $A$, the total stretch is just the two stretches multiplied — that's $\det(AB)=\det A\cdot\det B$.

---

Connections

• Determinant — Definition (cofactor / Leibniz)
• Elementary Matrices
• Gaussian Elimination & Row Echelon Form
• Invertibility & Singular Matrices
• Volume, Orientation & the Determinant
• Multilinear Alternating Forms
• Eigenvalues — det as product of eigenvalues

Concept Map

Hinglish (regional understanding)

Intuition Hinglish mein samjho

Dekho, determinant basically ek number hai jo batata hai ki tumhara linear map area (ya volume) ko kitna stretch karta hai — sign ke saath, kyunki agar shape "mirror" ki tarah palat jaaye to minus aa jaata hai. Is intuition se saari row operation rules apne aap nikal aati hain. Do rows swap karo to orientation flip hota hai, isliye sign change ($\times -1$). Ek row ko $k$ se scale karo to ek direction $k$ guna khinch jaati hai, isliye determinant bhi $\times k$. Aur sabse important: ek row me doosri row ka multiple add karo ($R_i\to R_i+kR_j$) to parallelogram lean to karta hai par area same rehta hai — isliye determinant bilkul nahi badalta. Yahi reason hai ki Gaussian elimination determinant ke liye "safe" hai.

Multiplicativity ka matlab hai $\det(AB)=\det A\cdot\det B$. Soch ke dekho: pehle map $B$ lagao, area $\det B$ guna ho gaya; phir $A$ lagao, aur $\det A$ guna ho gaya — total scaling $\det A\cdot\det B$. Algebra me har row operation ek elementary matrix se left-multiply karne ke barabar hai, aur har elementary matrix determinant ko ek known factor se multiply karti hai, isliye yeh property natural hai.

Ek bada exam trap: log likh dete hain $\det(2A)=2\det A$ — yeh galat hai. Kyunki $2A$ me har row scale hoti hai, to $n\times n$ matrix me $\det(2A)=2^n\det A$. Yaad rakhna mnemonic: "Swap-Sign, Scale-Same, Shear-Stays". Row reduce karte waqt swaps ginte raho ($(-1)^s$) aur agar kisi row ko divide kiya to wapas multiply karna mat bhoolna — warna final answer galat aayega.