Matrix multiplication — conditions, process, non-commutativity: Foundations

Before you can trust the parent page Matrix Multiplication, you must be able to read every squiggle it writes without pausing. This page builds each one from nothing, in an order where each idea leans only on the ones before it.

1. A number (scalar) — the smallest brick

Picture: one dot on a number line, or a "how much" label. Nothing has direction yet — a scalar only says how much, never which way.

Why the topic needs it: every entry inside a matrix is a scalar, and the final answers we compute (like $58$ ) are scalars. If you can add and multiply plain numbers, you already own the atoms of everything below.

2. A vector — an arrow with components

Figure — Matrix multiplication — conditions, process, non-commutativity

Look at the figure. The blue arrow is $x$ . Its shadow on the horizontal axis has length $3$ (the first component); its shadow on the vertical axis has length $2$ (the second component). The arrow is the pair of numbers — no more, no less.

Why the topic needs it: matrices act on vectors. The whole story " $B$ transforms $x$ into $B x$ " only means something once you can picture $x$ as an arrow.

3. The dot product — the heartbeat of multiplication

The single computation the parent page repeats over and over is the dot product. Master this one thing and matrix multiplication becomes copy-paste.

Why "position by position, then add"? Because each pair $a_{k} b_{k}$ answers one small question — "how much of ingredient $k$ , at price $k$ ?" — and the sum collects every ingredient's contribution into a single total. This is exactly the smoothie story in the parent's Feynman box.

Look at the figure. The top row (orange) and the side column (blue) are lined up. Each matched pair gets multiplied (the green links), and all products flow into one green box — the answer. Notice the row and column must be the same length, or some entries have no partner. That length-matching rule is the seed of the dimension rule in §6.

4. A matrix — a grid of scalars, read as rows and columns

Two ways to picture it, both useful:

As a stack of rows — each row is a row vector, ready to be the left side of a dot product.
As a line of columns — each column is a column vector, ready to be the right side of a dot product.

The parent page uses rows of $A$ and columns of $B$ , so keep both pictures live.

5. Subscripts and dimensions — the address system

For the $A$ above: $A_{12} = 2$ (row 1, column 2) and $A_{23} = 6$ (row 2, column 3).

Why the topic needs it: the parent's core formula $C_{ij} = \sum_{k} A_{ik} B_{k j}$ (with $C$ the product from §4) is entirely subscript navigation. If $A_{ij}$ is second nature, that formula reads like a sentence.

6. The compatibility condition — when can $A$ and $B$ even multiply?

The dot-product length rule from §3 grows directly into the rule that governs whole matrices.

Why exactly this rule? Every entry $C_{ij}$ is the dot product of a row of $A$ (length $n$ , since $A$ has $n$ columns) with a column of $B$ (length $n$ , since $B$ has $n$ rows). From §3, a dot product needs equal lengths — so the row length $n$ and the column length must be the same $n$ . That is the whole condition, inherited from the dot product's partner-matching.

7. Sigma notation $\sum$ — "add up a pattern"

The parent's compact formula uses $\sum$ . It scares people only because it is unfamiliar; it means nothing more than "add these up."

Concretely, for $n = 3$ : $\sum_{k = 1}^{3} A_{ik} B_{k j} = A_{i 1} B_{1 j} + A_{i 2} B_{2 j} + A_{i 3} B_{3 j}$

Compare that to §3: it is literally the dot product of row $i$ of $A$ with column $j$ of $B$ , and it produces the entry $C_{ij}$ of the product from §4. So $\sum$ is not new machinery — it is shorthand for the adding-up step you already did by hand.

8. The composition idea — why any of this exists

The parent's deepest claim is that $A B$ means "do $B$ first, then $A$ ." That is the concept of composition, borrowed from Linear transformations.

Look at the figure. An arrow enters box $B$ , comes out changed, then enters box $A$ and comes out changed again. The dashed shortcut box labelled $A B$ does both changes at once. Read the boxes right to left — that is why $B$ (written on the right) acts first.

Prerequisite map

The picture below shows which idea feeds which. If your reader's viewer does not render diagrams, here is the same thing in words: scalar is the atom; it builds both the vector and the matrix. Vector and matrix together give the dot product; the matrix also gives subscripts & dimensions, which grow into the compatibility condition and, with the dot product, into sigma notation. Separately, vector and matrix give composition. Both sigma notation and composition then feed the target topic, Matrix Multiplication.

Notice two streams feed the topic: the mechanical stream (dot product → compatibility → sigma) tells you how to compute, and the conceptual stream (composition) tells you why the answer means anything.

Equipment checklist

Self-test: cover the right side and answer aloud.

I can read

x = [32]

as an arrow

Yes — go 3 right, 2 up; the components are its axis-shadows.

The difference between a row vector and a column vector

A row is written flat

[32]

; a column is stacked vertically; the parent multiplies rows of $A$ by columns of $B$ .

How to take a dot product

Multiply the two lists position by position, then add all the products into one scalar; the lists must be the same length.

What

A_{ij}

points to

The scalar in row

i

, column

j

— row first, column second, both starting at 1.

What

A_{2 \times 3}

tells me

A

has 2 rows and 3 columns.

Who

A

B

, and

C

are in

C = A B

A

is the left factor (

m \times n

B

the right factor (

n \times p

) that acts first, and

C

the product (

m \times p

) built one entry at a time.

When can

A

multiply

B

When

A

m \times n

and

B

n \times p

— the inner

n

's match; the product is

m \times p

; otherwise undefined.

How to unpack

\sum_{k = 1}^{n} A_{ik} B_{k j}

Let

k

run

1

n

, form each product

A_{ik} B_{k j}

, and add them — it is exactly the dot product giving

C_{ij}

Why

A B

means "

B

first, then

A

Because

A B

is composition —

A (B x)

— and we read the machines right to left.

Why

A B \neq = B A

in general

Swapping the order swaps which transformation runs first; a different pipeline gives a different result.

Connections

Matrix Multiplication — the parent this page prepares you for
Matrix addition and scalar multiplication — the gentler operations that share the grid picture
Linear transformations — the source of the "matrix as machine" and composition ideas
Identity matrix — the do-nothing machine you meet next
System of linear equations — where matrix-times-vector first pays off