Transpose — definition, properties: Foundations

Before you can trust a single line of the parent note, you need to own every symbol it throws at you. This page builds them one at a time, from the very idea of a grid of numbers, up to the machinery of multiplication and inverses. Nothing here assumes you have seen matrices before.

1. What is a matrix? (the grid)

Look at the figure below. The grid on the left is a matrix. Reading across is a row; reading down is a column.

Figure — Transpose — definition, properties

Why the topic needs this: the transpose is an operation on the whole grid at once. If you don't picture a matrix as a physical grid of pigeonholes, the flip has nothing to happen to. Later you'll meet Matrix Operations which act on these same grids.

2. Size: $m \times n$ (how big is the grid?)

Why the topic needs this: the transpose swaps the size from $m \times n$ to $n \times m$ . The parent note says " $A^{T}$ is an $n \times m$ matrix" — that swap of the two numbers is the whole geometric story, and you can only see it if you know what the two numbers mean.

3. Entries and the index $a_{ij}$ (naming a single box)

Why the topic needs this: the entire definition of transpose is written in index language: $[A^{T}]_{ij} = [A]_{j i}$ That formula says nothing more than "to find the entry at address $(i, j)$ of the flipped grid, go to the swapped address $(j, i)$ of the original grid." If $a_{ij}$ is a mystery to you, this formula is unreadable. Master the address and the formula becomes obvious.

Recall Why does swapping

$i$ and $j$ flip rows into columns? Because $(i, j)$ = (row, column). Swapping them turns "row $i$ , column $j$ " into "row $j$ , column $i$ " ::: so an entry that used to be along a row now lives along a column, and that is exactly a flip across the diagonal.

4. The main diagonal (the mirror line)

Why the topic needs this: this is what makes symmetric and skew-symmetric matrices possible.

A symmetric matrix ( $A^{T} = A$ ) is one where the mirror image looks identical to the original — nothing appears to move.
A skew-symmetric matrix ( $A^{T} = - A$ ) is one where the mirror flips the sign. Since a diagonal entry mirrors to itself, we'd need $a_{ii} = - a_{ii}$ , which forces $a_{ii} = 0$ — that is why the parent note insists the diagonal of a skew-symmetric matrix is all zeros. See Symmetric Matrices.

5. Square vs rectangular

Only square matrices have a main diagonal that runs corner to corner, so symmetry, Determinants, and Matrix Inverse all live in square-matrix land. A rectangular matrix like $2 \times 3$ can still be transposed (it becomes $3 \times 2$ ), but it can never equal its own transpose — the sizes wouldn't even match, let alone the entries.

Recall Can a

$2 \times 3$ matrix be symmetric? No — symmetry needs $A^{T} = A$ , but $A^{T}$ is $3 \times 2$ while $A$ is $2 \times 3$ ; different sizes can't be equal ::: symmetry requires a square matrix.

6. Row and column vectors (the strips that get flipped)

Why the topic needs this: the whole "flip rows into columns" description is made of these single-strip moves. It also connects to Inner Product Spaces, where a row-vector-times-column-vector is how a dot product is written.

7. Matrix addition and scalar multiples (the easy operations)

Why the topic needs this: two of the four "SIRP" properties are about these.

$(A + B)^{T} = A^{T} + B^{T}$ : adding happens entry-by-entry, and flipping happens address-by-address, so the two operations don't interfere — order doesn't matter.
$(k A)^{T} = k A^{T}$ : a scalar touches every box equally, so the mirror can't tell the scaling was done before or after. The reason these are easy (no reversal!) is exactly because addition and scaling treat every box independently.

8. Matrix multiplication and the sum $\sum$ (the hard operation)

This is the one genuinely new machine, and the reversal rule depends entirely on it.

Why the topic needs this — the reversal explained from zero: In $(A B)^{T}$ we ask for address $(i, j)$ , which is address $(j, i)$ of $A B$ , i.e. $\sum_{k} a_{j k} b_{k i}$ . When we transpose each factor separately, every $a_{j k}$ becomes an entry of $A^{T}$ at swapped address, and same for $B$ . The only way to line the shared index $k$ back up between the two matrices is to put $B^{T}$ first and $A^{T}$ second. That's not a trick — it's forced by which finger has to meet which. That is why $(A B)^{T} = B^{T} A^{T}$ and never $A^{T} B^{T}$ . This machinery also drives Eigenvalues and Eigenvectors and Orthogonal Matrices.

9. The identity $I$ and the inverse $A^{- 1}$

Why the topic needs this: the property $(A^{- 1})^{T} = (A^{T})^{- 1}$ is proved by transposing the equation $A A^{- 1} = I$ and using the reversal rule plus the fact that $I^{T} = I$ (the identity is symmetric — its mirror image is itself). It's also the key to Orthogonal Matrices, where $A^{T} = A^{- 1}$ . Full details live in Matrix Inverse.

Recall Why is

$I^{T} = I$ ? Every entry of $I$ satisfies $[I]_{ij} = [I]_{j i}$ (both are $1$ only when $i = j$ , else $0$ ) ::: so $I$ is symmetric — the mirror leaves it unchanged.

How the foundations feed the topic

Read it top-down: the plain grid gives you size and addresses; addresses plus the diagonal give you the transpose; the transpose plus each operation gives you each property; and everything converges on symmetric, inverse, and orthogonal matrices — the payoff of the parent topic Transpose — definition, properties.

Equipment checklist

Test yourself — say each answer out loud before revealing.

What does the size

m \times n

tell you, in order?

m

= number of rows,

n

= number of columns.

What is the entry

a_{ij}

The number in row

i

, column

j

of matrix

A

Rewrite the transpose definition in words.

The

(i, j)

entry of

A^{T}

equals the

(j, i)

entry of

A

— swap the address.

What happens to the size when you transpose?

It swaps from

m \times n

n \times m

Which entries stay put under transpose?

The main-diagonal entries

a_{ii}

, since

(i, i)

swaps to

(i, i)

What does

\sum_{k = 1}^{p} a_{ik} b_{k j}

compute?

Entry

(i, j)

of the product

A B

— the running total of row-

i

-of-

A

times column-

j

-of-

B

Why must the diagonal of a skew-symmetric matrix be zero?

Because

a_{ii} = - a_{ii}

forces

a_{ii} = 0

Why does

(A B)^{T}

reverse the order to

B^{T} A^{T}

Because matrix multiplication is non-commutative and the shared sum-index

k

only lines up when

B^{T}

comes first.

What is the identity matrix

I

and why is

I^{T} = I

The square matrix of

1

s on the diagonal,

0

s elsewhere; it's symmetric, so its mirror image is itself.