Consider two matrices:
A=[[1,2],[3,4]]
B=[[5,−1],[2,0]].
We can add them.
We can multiply either matrix by an ordinary number.
We can also multiply A by B.
But those three operations do not follow the same rule.
Matrix operations become reliable when the learner checks structure before arithmetic: addition needs matching positions, scalar multiplication acts on every element, and matrix multiplication pairs rows with columns.
The quick answer
- Addition/subtraction: matrices must have the same order.
- Scalar multiplication: multiply every element by the scalar.
- Matrix multiplication AB: number of columns of A must equal number of rows of B.
- Order of AB: rows of A × columns of B.
- AB is generally not BA.
Matrix addition is position-by-position
Using A and B:
A+B=[[1+5,2+(−1)],[3+2,4+0]].
So:
A+B=[[6,1],[5,4]].
Each element combines only with the element in the same row and column position.
Why addition requires the same order
A 2×3 matrix has six positions arranged in two rows and three columns.
A 3×2 matrix also has six entries, but they occupy a different shape.
There is no one-to-one positional match between corresponding rows and columns.
Therefore ordinary matrix addition is not defined for those two orders.
Having the same number of entries is not enough. Matrix addition needs the same structural address for every corresponding element.
Matrix subtraction follows the same compatibility rule
A−B=[[1−5,2−(−1)],[3−2,4−0]].
Therefore:
A−B=[[-4,3],[1,4]].
Scalar multiplication
A scalar is an ordinary number multiplying a matrix.
For 3A:
3A=[[3,6],[9,12]].
The scalar multiplies every entry.
Negative scalars reverse signs
−2A=[[-2,−4],[−6,−8]].
Nothing about the order changes.
Only the values of the elements are scaled.
Scalar distribution works as expected
For compatible matrices:
k(A+B)=kA+kB.
This mirrors the ordinary distributive law because the operation is performed entry by entry.
Matrix multiplication is different
To calculate AB, we do not multiply corresponding entries.
Instead, each result entry is built from:
one row of A dotted with one column of B.
Matrix multiplication combines relationships across an inner dimension. The row chooses contributions from the first matrix; the column tells how those contributions enter the second.
The dimension rule for AB
If A is m×n and B is n×p, then AB is defined and has order:
m×p.
The inner dimensions n and n must match.
The outer dimensions m and p survive into the result.
A useful memory structure is:
(m×n)(n×p) → m×p.
Worked multiplication: AB
A=[[1,2],[3,4]]
B=[[5,−1],[2,0]].
Both are 2×2, so AB is defined and will be 2×2.
Entry (1,1):
1×5+2×2=9.
Entry (1,2):
1×(−1)+2×0=−1.
Entry (2,1):
3×5+4×2=23.
Entry (2,2):
3×(−1)+4×0=−3.
Therefore:
AB=[[9,−1],[23,−3]].
Why row-by-column multiplication works
Imagine a matrix B transforming or redistributing the components of an input.
A second matrix A then combines those components in a new way.
Matrix multiplication is built so the composition of those linear relationships can itself be represented by one matrix.
The row-by-column rule is therefore not a strange arithmetic convention. It is the structure needed for transformations and systems to compose consistently.
AB is generally not BA
Now calculate BA.
Entry (1,1):
5×1+(−1)×3=2.
Entry (1,2):
5×2+(−1)×4=6.
Entry (2,1):
2×1+0×3=2.
Entry (2,2):
2×2+0×4=4.
So:
BA=[[2,6],[2,4]].
AB≠BA.
Matrix multiplication is generally non-commutative. The order of operations can change the result because the first transformation changes what the second one receives.
Sometimes AB exists but BA does not
Suppose A is 2×3 and B is 3×4.
AB is defined:
(2×3)(3×4)→2×4.
But BA would require:
(3×4)(2×3).
The inner dimensions 4 and 2 do not match.
So BA is not defined.
Worked rectangular example
Let:
A=[[1,2,3],[4,5,6]]
and:
B=[[1,0],[2,1],[−1,3]].
A is 2×3.
B is 3×2.
AB will be 2×2.
First row, first column:
1(1)+2(2)+3(−1)=2.
First row, second column:
1(0)+2(1)+3(3)=11.
Second row, first column:
4(1)+5(2)+6(−1)=8.
Second row, second column:
4(0)+5(1)+6(3)=23.
Therefore:
AB=[[2,11],[8,23]].
Identity matrix
For a compatible identity matrix I:
AI=A
and:
IA=A.
The identity matrix behaves like 1 under matrix multiplication, with the important condition that the dimensions must fit.
Zero matrix behaves differently in multiplication
For compatible dimensions:
AO=O
and:
OA=O.
But the exact order of the resulting zero matrix follows the multiplication dimensions.
Associativity survives even though commutativity does not
For compatible matrices:
(AB)C=A(BC).
This is important because several transformations can be composed without ambiguity about grouping.
But we still cannot generally swap the order to BAC or CAB.
Distributive law
For compatible matrices:
A(B+C)=AB+AC.
Likewise:
(A+B)C=AC+BC.
Again, dimensions must permit every operation.
Matrices as data transformations
Suppose a data matrix stores quantities of products and a second matrix stores prices or conversion relationships.
Multiplication can combine the data systematically so that each row–column result represents one meaningful total.
This is why row and column labels matter: the inner dimension must refer to the same type of quantity before the multiplication has meaning.
A dimension match is necessary for matrix multiplication to be defined, but a meaningful model also needs the inner row–column categories to correspond conceptually.
Matrices as transformations
A 2×2 matrix can act on a 2×1 coordinate column vector.
The multiplication creates another 2×1 vector.
Several geometric transformations can therefore be represented and composed by matrices.
The fact that AB≠BA becomes geometrically meaningful: performing one transformation and then another can differ from reversing their order.
Common misconception 1: multiply corresponding entries for AB
That operation is not ordinary matrix multiplication. Use row-by-column products.
Common misconception 2: same number of entries means matrices can be added
Addition requires the same order, not merely equal entry count.
Common misconception 3: AB and BA are automatically equal
Matrix multiplication is generally non-commutative.
Common misconception 4: if AB exists, BA must exist
The inner dimensions for the reverse product may fail to match.
Common misconception 5: result order is the matching inner dimensions
The inner dimensions disappear. The result keeps the outer dimensions.
A matrix-operations diagnostic ladder
- Can the learner check order before addition?
- Can the learner add and subtract corresponding elements?
- Can the learner apply a scalar to every entry?
- Can the learner test whether AB is defined?
- Can the learner predict the order of AB?
- Can the learner calculate one row-by-column entry?
- Can the learner construct the full product?
- Can the learner compare AB and BA?
- Can the learner use identity and zero matrices correctly?
- Can the learner connect multiplication order to transformation order?
How this fits Secondary Mathematics
Matrix operations connect symbolic algebra, transformations, data systems and later linear algebra. The crucial habit is structural: check dimensions before arithmetic and interpret the row–column order rather than treating matrices as flat lists.
The deeper lesson: multiplication composes relationships
Addition combines corresponding positions.
Scalar multiplication scales every position.
Matrix multiplication does something more ambitious: it composes one structured relationship with another.
The row-by-column rule becomes less mysterious when we see what matrix multiplication is built to do: preserve the architecture of linked transformations so a sequence of operations can itself be represented as one new matrix.