Mathematics RU

Practice · Chapter 36

Gaussian elimination

Gaussian elimination: solve a system, find the rank of a matrix, tell how many solutions a system has, and bring a matrix to reduced row echelon form.

How to solve it

Gaussian elimination simplifies a system without changing its solutions: rows may be swapped, multiplied by a non-zero number, and one row may get another row times a number added to it. The goal is echelon form, with zeros under every pivot. From there the solution is read from the bottom up.

Step by step

  1. Write the augmented matrix: the coefficients and, past the bar, the right-hand sides.
  2. Choose a pivot in the first column ($\pm 1$ is handy; if it is zero, swap rows) and clear everything below it.
  3. Move to the next column and the next row and repeat. You reach echelon form.
  4. The rank is the number of non-zero rows. A row $0 = c$ with $c \ne 0$ means no solutions; a rank below the number of unknowns means infinitely many; otherwise exactly one.
  5. To solve, go from the bottom up, substituting the unknowns found. For the reduced form, clear the entries above the pivots too.
The coefficient matrix. The augmented matrix. If its rank exceeds that of $A$, there are no solutions. The number of unknowns. Equal ranks below $n$ mean infinitely many solutions. Example: a matrix with two equal rows loses one row to zero after a subtraction, so its rank is one less than the number of rows.

Common mistakes

  • Changing a row but not its right-hand side. Operations act on the whole row of the augmented matrix.
  • Adding a multiple of a row to itself — that is scaling, not elimination.
  • Deciding how many solutions there are before reaching echelon form.
  • Taking the number of rows of the original matrix as the rank instead of the non-zero rows of the echelon form.

Example