What is Row Echelon Form (REF)?
A matrix is in Row Echelon Form (REF) when it represents the output of standard Gaussian elimination. REF is a simplified, upper triangular structure that makes back-substitution straightforward. A matrix is defined to be in REF if it meets three primary criteria:
- All rows containing only zeros are positioned at the bottom of the matrix.
- The leading entry (the first non-zero number from the left, also called a pivot) of any row is strictly to the right of the leading entry of the row above it.
- All entries in a column below a leading entry are zero.
Note that in REF, the leading entry does not have to be 1 (though many solvers scale it to 1 for simplicity), and there can be non-zero numbers above the leading entries. This is the main difference between REF and RREF.
REF vs RREF: Key Differences
It is common for students of linear algebra to confuse Row Echelon Form with Reduced Row Echelon Form. Here is a side-by-side comparison:
| Property | Row Echelon Form (REF) | Reduced Row Echelon Form (RREF) |
|---|---|---|
| Entries below pivots | Must be zero | Must be zero |
| Entries above pivots | Can be any number | Must be zero |
| Pivot values | Can be any non-zero value | Must be exactly 1 |
| Uniqueness | Not unique (multiple forms possible) | Unique (only one RREF exists per matrix) |
How is REF Used to Solve Equations?
REF is highly efficient for computing determinants and solving systems of equations by hand. Once a system's augmented matrix is reduced to REF, the corresponding system of equations is upper triangular. You can solve it using a process called back-substitution:
- Solve the last equation for the last variable (since it is the only variable in that equation).
- Substitute that value back into the equation above to solve for the second-to-last variable.
- Continue this process moving upwards until all variables are solved.
Because REF requires fewer row operations than RREF (since we do not need to eliminate values above the pivots), it is computationally cheaper and represents the first phase of the Gauss-Jordan elimination algorithm.