The three legal moves
Gaussian elimination works on the augmented matrix [A | b] using only three row operations. Each one changes how the system looks but never changes its solution — that is the whole reason they are allowed.
- Swap two rows (reorder the equations).
- Multiply a row by a nonzero number (scale an equation).
- Add a multiple of one row to another row (combine equations to kill a term).
Climbing to the staircase
The goal is row echelon form: a staircase where every row's leading number sits to the right of the one above it, so the bottom-left corner fills up with zeros. Once you reach it, the last row gives one unknown outright.
Let's take the system from the previous guide and write its augmented matrix, then clear the entry under the first pivot (the leading 2 in row 1).
[A|b] = [ 2 1 | 5 ]
[ 1 3 | 6 ]
R2 -> R2 - (1/2)*R1:
[ 2 1 | 5 ]
[ 0 2.5 | 3.5 ] <- staircase / row echelon formBack-substitution finishes it
Now read the staircase from the bottom up. The last row says 2.5*y = 3.5, so y = 1.4. Substitute that into the first row, 2x + y = 5, and solve for x. This bottom-up unwinding is called back-substitution.
row 2: 2.5*y = 3.5 -> y = 1.4 row 1: 2x + 1.4 = 5 -> 2x = 3.6 -> x = 1.8 check: 1*1.8 + 3*1.4 = 1.8 + 4.2 = 6 (matches b)