Visual Tools
Calculators
Tables
Mathematical Keyboard
Converters
Other Tools

Which Method Should I Use to Solve a Linear System?

Gaussian elimination solves every linear system, so it is the default. The other methods are shortcuts for special situations: a question about the number of solutions only, a zero right-hand side, a system that is already triangular, or a small square system with a nonzero determinant.

When to use this

Look at three things in the problem: what it asks for (the solution, or only how many there are), whether the constants are all zero, and the shape of the coefficient matrix (triangular, small and square, or anything else).

Decision tree

only how many solutionsthe solutionsyes, Ax = 0noalready triangularsmall, square, det ≠ 0same A, many banything elseWhat does the question askfor?Solvability check →Are all the constants zero?Homogeneous system →What does the matrix looklike?Back substitution →Cramer's rule →Inverse matrix →Gaussianelimination →

The same tree, written out

  1. 1The question asks only whether a solution exists, whether it is unique, or for which parameter values the system is consistent. Use the solvability check: reduce the augmented matrix and compare ranks (rank criterion).
  2. 2Every equation has zero on the right-hand side. Use the homogeneous systems method: reduce AA only and write the null space with one vector per free variable.
  3. 3The system is already in echelon or upper triangular form. Use back substitution directly, from the last row up.
  4. 4The system is square and small (2 × 2 or 3 × 3), its determinant is not zero, and you want a formula or only one unknown. Use Cramer's rule: each unknown is a ratio of two determinants.
  5. 5Several systems share the same invertible coefficient matrix, or the inverse is given or asked for. Use the inverse matrix method: find the inverse once, then multiply it by each right-hand side.
  6. 6Any other system: larger, not square, with a zero determinant, or with no obvious structure. Use Gaussian elimination. If free variables remain, finish with parametric solutions.