“Gaussian Elimination” solved a system by carrying its augmented matrix to echelon form, and by the end of the reduction you knew how many solutions there were. The determinant answers that question in advance, from the coefficients alone, in a single number.
Definition · Determinant of a 2 by 2 matrixFor A = [ a b ; c d ], the determinant is det A = ad - bc: the product down the main diagonal minus the product up the other one. It is written det A or |A|, and the bar notation is applied directly to the array of entries.
| 3 5 | | 2 6 |
| 1 4 | = (3)(4) - (5)(1) = 7 | 1 3 | = 6 - 6 = 0
the second matrix has a second row that is half the first,
and a determinant of zero is what that always produces
For a 3 by 3 matrix the determinant is built from three 2 by 2 determinants. Take each entry of the first row, cross out the row and column it sits in, and multiply the entry by the determinant of what is left. The three results are then combined with alternating signs: plus, minus, plus.
| 2 1 3 |
| 0 4 -1 |
| 5 2 0 |
| 4 -1 | | 0 -1 | | 0 4 |
= 2· | 2 0 | - 1· | 5 0 | + 3· | 5 2 |
= 2(0 - (-2)) - 1(0 - (-5)) + 3(0 - 20)
= 2(2) - 1(5) + 3(-20)
= 4 - 5 - 60 = -61
The alternating signs are not decoration and they are not optional. The middle term is subtracted, and forgetting that is the single most common way a 3 by 3 determinant comes out wrong. The expansion can be taken along any row or column with the same sign pattern started in the right place; the first row is the usual choice, and a row containing a zero is the smart one, because that term vanishes.
There is a diagonal shortcut for 3 by 3 determinants, sometimes called Sarrus's rule. It gives the same answer, and it works for 3 by 3 matrices and for no other size. The expansion above is the method that keeps working, so it is the one to learn.
Theorem · The determinant decides uniquenessA system of n linear equations in n unknowns has exactly one solution if and only if the determinant D of its coefficient matrix is not zero. When D = 0 the system has either no solution or infinitely many.
Definition · Cramer's ruleWrite the system in standard form and let D be the determinant of the coefficient matrix. Let Dx be the same determinant with the x column replaced by the column of constants, and Dy the same with the y column replaced. If D ≠ 0 then x = Dx/D and y = Dy/D.
The rule extends to any size: for each unknown, replace that unknown's column by the constants and divide by D. A system in three unknowns therefore needs four 3 by 3 determinants.
Example · Two systems with the same Dx + y = 1 with 2x + 2y = 5 has D = (1)(2) - (1)(2) = 0. So does x + y = 1 with 2x + 2y = 2. The first is inconsistent — doubling the first equation gives 2x + 2y = 2, which contradicts 2x + 2y = 5 — and the second is the same equation twice, with infinitely many solutions. One determinant, two different outcomes: D = 0 genuinely cannot tell them apart.
Cramer's rule is a formula, not a strategy. For a 2 by 2 system it is quick, and for coefficients that are letters rather than numbers it is often the clearest thing available. For a 4 by 4 system it asks for five 4 by 4 determinants, where one row reduction would have finished the job. “Gaussian Elimination” remains the method; this lesson is the criterion.