A pivot is the leading non-zero entry in a row after applying Gaussian elimination. It marks a position where a variable is constrained or a row is linearly independent.

• Each pivot corresponds to a linearly independent row or column
• Pivot positions determine the rank of a matrix

---

📐 Structural Limits

Let $A \in \mathbb{R}^{m \times n}$. Then:

• Max number of pivots = $\min(m, n)$
• You can’t have more than one pivot per row or column
• If $m > n$, at least $m - n$ rows will be pivotless (i.e. dependent)

---

🧮 Pivot ↔ Rank ↔ Independence

Structure	Pivot Present	Interpretation
Column	✅ Yes	Linearly independent
Column	❌ No	Linearly dependent
Row	✅ Yes	Adds a new constraint
Row	❌ No	Redundant (dependent)

---

📊 Example: Row Reduction

Let:

$$ A = \begin{bmatrix} 1 & 2 & 3 \\ 0 & 1 & 4 \\ 0 & 0 & 0 \\ 0 & 0 & 0 \end{bmatrix} $$

• Pivot positions: (1,1), (2,2)
• Rank = 2
• Rows 3 and 4 are dependent (no new constraints)

---

📐 Deriving the Rouché–Capelli Theorem from Pivot Logic

Let $Ax = b$ be a system of $m$ equations in $n$ unknowns.
Form the augmented matrix $[A \mid b]$.

Step-by-step:

1. Row reduce both $A$ and $[A \mid b]$
2. Count pivot rows in each
3. Compare:

$$ Ax = b \text{ has a solution} \iff \operatorname{rank}(A) = \operatorname{rank}([A \mid b]) $$

Why?

• Each pivot row in $A$ corresponds to an independent constraint
• If $b$ introduces a new pivot in $[A \mid b]$, it means $b$ is not in the column space of $A$
• That makes the system inconsistent

✅ If ranks match → $b$ lies in the column space → solution exists
❌ If ranks differ → $b$ is outside the column space → no solution

---

🧭 Geometric Interpretation

Pivot rows define the dimension of the row space
Pivot columns define the dimension of the column space

Imagine each pivot as a new direction in space. Non-pivot rows/columns lie within the span of earlier ones—they don’t expand the space.

---

💡 Callouts