Why is (a + b)^(2) different from (a)^(2) + (b)^(2)? Enlarging both sides of a square tile adds two strips, as well as a corner square. Those strips supply the middle term 2ab. We can see the structure, then recover it without a drawing using polynomial multiplication.

Four parts of a squared sum

Take a square of side a + b, with a > 0 and b > 0. Splitting each side separates a square of area (a)^(2), two strips of area ab, and a small square of area (b)^(2). Adding only the squares would leave the strips uncounted.

ababa²ababb²
For a = 3 and b = 2: side 5, areas 9 + 6 + 6 + 4 = 25. The drawing is to scale; a and b are positive lengths in the same units.

Algebra repeats this construction:

(a + b)^(2) = (a + b)(a + b) = (a)^(2) + ab + ba + (b)^(2) = (a)^(2) + 2ab + (b)^(2).

The drawing uses a = 3, b = 2: 9 + 6 + 6 + 4 = 25, just like (5)^(2). The distributive law proves the identity for any real a,b; ordinary areas require positive lengths. Deriving squares in OpenStax §1.4.

A squared difference changes the middle sign

Replace b with − b in the formula. The square ( − b)^(2) = (b)^(2), while the mixed product changes sign:

(a − b)^(2) = (a)^(2) − 2ab + (b)^(2).

Now (99)^(2) = (100 − 1)^(2) = 10000 − 200 + 1 = 9801. Subtracting only (1)^(2) gives 9999: that calculates a different expression, (100)^(2) − (1)^(2).

A letter may stand for an entire expression: (2x − 3)^(2) = 4(x)^(2) − 12x + 9. Here the first square is (2x)^(2), and the middle term is − 2 ⋅ 2x ⋅ 3. Parentheses protect both the coefficient and the sign.

A difference of squares cancels mixed terms

When one binomial has a plus and the other a minus, the mixed contributions are opposites:

(a + b)(a − b) = (a)^(2) − ab + ab − (b)^(2) = (a)^(2) − (b)^(2).

This is a difference of squares, not a squared difference. For example, 103 ⋅ 97 = (100 + 3)(100 − 3) = 10000 − 9 = 9991. Read it backwards too: (x)^(2) − 25 = (x − 5)(x + 5). The product reveals factors hidden in the sum.

Cubes: checking the signs

Multiply (a − b) by (a)^(2) + ab + (b)^(2). Six terms appear: (a)^(3) + (a)^(2)b + a(b)^(2) − (a)^(2)b − a(b)^(2) − (b)^(3). The four middle terms cancel, so:

(a)^(3) − (b)^(3) = (a − b)((a)^(2) + ab + (b)^(2)).

Replacing b with − b gives the sum of cubes:

(a)^(3) + (b)^(3) = (a + b)((a)^(2) − ab + (b)^(2)).

For example, (x)^(3) − 8 = (x − 2)((x)^(2) + 2x + 4), whereas (x)^(3) + 8 = (x + 2)((x)^(2) − 2x + 4). The sign on ab makes the middle contributions cancel. Both cube identities, OpenStax §1.5.

Branch · One step furtherAn extra multiplication; the five main identities have already been derived.

Cubing a binomial

Multiply the squared sum by a + b. The term (a)^(2)b appears once from (a)^(2) ⋅ b and twice from 2ab ⋅ a, giving 3(a)^(2)b. Similarly, we collect 3a(b)^(2):

(a + b)^(3) = (a)^(3) + 3(a)^(2)b + 3a(b)^(2) + (b)^(3).

Substituting − b gives (a − b)^(3) = (a)^(3) − 3(a)^(2)b + 3a(b)^(2) − (b)^(3). This differs from (a)^(3) − (b)^(3): two more contributions lie between the outer cubes.

Special products save writing because every pair of terms has already been counted. If a formula slips your mind, expansion reconstructs it. A small numerical substitution can expose a wrong sign, although it cannot replace a proof.