Rank of a matrix — definition, row rank = column rank theorem: Foundations

Before you can trust the parent theorem ("row rank = column rank"), you must be fluent in the small vocabulary it silently assumes. This page builds each word from nothing, in the order that lets the next one make sense.

0. The starting picture: numbers as arrows

The picture. In 2D, $(3, 1)$ means "walk 3 right, 1 up." The arrow is the vector.

Why the topic needs it. The parent note calls rows and columns "vectors" and asks how many directions they cover. You cannot count directions until you see each row and column as an arrow.

Figure — Rank of a matrix — definition, row rank = column rank theorem

1. Matrix — a grid, and two ways to read it

$A = 123246134 is 3 \times 3.$

Two readings of the same grid — this is the heart of everything:

Row reading: three horizontal arrows, each living in $R^{3}$ : $(1, 2, 1)$ , $(2, 4, 3)$ , $(3, 6, 4)$ .
Column reading: three vertical arrows, each also in $R^{3}$ : $(2 1) 3$ , $(4 2) 6$ , $(3 1) 4$ .

Why the topic needs both readings. "Row rank" counts directions in the row arrows; "column rank" counts directions in the column arrows. The whole miracle is that these two totally different bundles of arrows give the same count.

The picture. Transpose = tip the matrix onto its side. The rows of $A$ become the columns of $A^{T}$ .

Why the topic needs it. The proof's Step 4 says "apply the argument to $A^{T}$ ." That is exactly the trick that swaps rows and columns to get equality — so $A^{T}$ must be crystal clear first.

2. Scaling and adding arrows → linear combination

The picture. Give me the arrows $v_{1}, v_{2}$ and any dials $s_{1}, s_{2}$ ; the reachable tips sweep out a whole region. That region is the star of the next section.

Why the topic needs it. Every single move in the parent note is a linear combination: row operations (" $R_{2} \to R_{2} - 2 R_{1}$ "), the factorization $A = C R$ , "column 2 = 2×column 1." If linear combination is fuzzy, the proof is fog.

See Matrix Multiplication as Linear Combination for why $A x$ itself is just a linear combination of the columns of $A$ .

3. Span — everywhere you can reach

The picture, case by case (we cover them all):

Span of one nonzero arrow = the infinite line through it.
Span of two arrows pointing different ways = the whole plane they lie in.
Span of two arrows on the same line (one is a multiple of the other) = still just a line — the second added nothing.
Span of the zero vector alone = just the single point $(0, 0)$ .

Why the topic needs it. "Row space = span of the rows" and "column space = span of the columns." Rank measures how big these spans are — a point, a line, a plane, a volume. Section 5 makes "how big" a number.

4. Linear independence — no arrow is a copy of the others

The pictures (all cases):

$(1, 2)$ and $(2, 4)$ : the second is $2 \times$ the first → dependent (same line). This is exactly the parent's rank-1 trap.
$(1, 0)$ and $(0, 1)$ : neither is a multiple of the other → independent (they open up the whole plane).
Any set that includes the zero vector → automatically dependent ( $1 \cdot 0$ plus zero dials gives $0$ ).

Why the topic needs it. Rank counts only the independent directions. The parent's mistake box ("nonzero rows can still be dependent") is precisely a warning that independence, not "nonzero-ness," is what matters.

(2, 1)

as an arrow — where does it start and end?

Starts at the origin

(0, 0)

, ends at the point 2 right and 1 up.

What does "

A

3 \times 4

" tell you?

A

has 3 rows and 4 columns (rows first, columns second).

Give the two ways to read the same matrix as bundles of arrows.

As row arrows (each horizontal line) and as column arrows (each vertical line).

What is

A^{T}

geometrically?

The matrix with rows and columns swapped — flip across the diagonal.

Write a general linear combination of

v_{1}, v_{2}, v_{3}

s_{1} v_{1} + s_{2} v_{2} + s_{3} v_{3}

for any scalars

s_{1}, s_{2}, s_{3}

What is the span of two arrows that lie on the same line?

Just that one line — the second arrow adds nothing.

State the zero-only test for independence.

They are independent iff

s_{1} v_{1} + \dots + s_{k} v_{k} = 0

forces every

s_{i} = 0

Are

(1, 2)

and

(2, 4)

independent?

No — the second is

2 \times

the first, so they are dependent.

Define a basis in one line.

An independent set that also spans the whole space.

What is dimension?

The number of vectors in any basis — the count of independent directions.

What is a pivot?

The leading nonzero entry of a row once the matrix is in staircase (echelon) form.

What does the kernel of

A

contain?

All input arrows

x

with

A x = 0

— the directions crushed to zero.

Connections

Parent: Rank of a Matrix
Row Space and Column Space
Linear Independence and Basis
Matrix Multiplication as Linear Combination
Gaussian Elimination & Echelon Form
Rank–Nullity Theorem
Invertible Matrix Theorem

Foundations — Rank of a matrix — definition, row rank = column rank theorem

0. The starting picture: numbers as arrows

1. Matrix — a grid, and two ways to read it

2. Scaling and adding arrows → linear combination

3. Span — everywhere you can reach

4. Linear independence — no arrow is a copy of the others

5. Basis and dimension — the honest count

6. Pivots — the computational shortcut

7. Kernel and nullity — the leftovers (for the bound)

How these foundations feed the topic

Equipment checklist

Connections