Row echelon form (REF) and reduced row echelon form (RREF) both arrange a matrix into a staircase of pivots. RREF goes further: every pivot is 1 and is the only nonzero entry in its column. That extra work makes RREF unique, while a matrix can have many different REFs.
| Property | REF | RREF |
|---|---|---|
| Zero rows are at the bottom | Yes | Yes |
| Each pivot is right of the one above | Yes | Yes |
| Zeros below each pivot | Yes | Yes |
| Pivots equal 1 | Not required (some textbooks require it) | Required |
| Zeros above each pivot | Not required | Required |
| Unique for a given matrix | No | Yes |
| Pivot positions and rank | Same in every REF | Same |
| Method that produces it | Gaussian elimination | Gauss-Jordan elimination |
| How you read a solution | Back substitution | Directly from the rows |
The difference you can see.
| 1 | 2 | 3 |
| 0 | 1 | 2 |
| 1 | 0 | −1 |
| 0 | 1 | 2 |
The REF has a 2 above the second pivot. One more row operation clears it and gives the RREF.
Row echelon form vs. reduced row echelon form.
Row echelon form (REF)
Zero rows sit at the bottom, and each leading entry is to the right of the one above it, with zeros below every leading entry. Textbooks differ on one detail: some also require every leading entry to be 1, others do not. This site's calculator scales REF pivots to 1, which meets both definitions.
Reduced row echelon form (RREF)
RREF adds two conditions: every leading entry is 1, and it is the only nonzero entry in its column. Those extra conditions make the reduced form unique for each matrix.
One matrix, two different REFs, one RREF.
The row echelon form depends on which row operations you choose. Take this 3×4 matrix, which is used in the TI-Basic Developer reference for the TI-84's ref( command:
| 1 | 2 | 5 | 0 |
| 2 | 2 | 1 | 2 |
| 3 | 4 | 6 | 2 |
Reducing it from the top row down, as this site's calculator does, gives the first REF below. The reference reports the second REF for the TI-84's ref( command, which chooses its pivots differently. Both meet every REF rule:
| 1 | 2 | 5 | 0 |
| 0 | 1 | 9/2 | −1 |
| 0 | 0 | 0 | 0 |
| 1 | 4/3 | 2 | 2/3 |
| 0 | 1 | 9/2 | −1 |
| 0 | 0 | 0 | 0 |
Keep reducing either one and you reach the same RREF. The pivots also sit in the same columns, 1 and 2, in both REFs, so the rank is 2 either way. Pivot positions and rank never depend on the route you take, even though the other entries of an REF can.
| 1 | 0 | −4 | 2 |
| 0 | 1 | 9/2 | −1 |
| 0 | 0 | 0 | 0 |
Gaussian vs. Gauss-Jordan elimination.
Gaussian elimination works down the matrix, creating each pivot and clearing the entries below it. It ends at REF, and you then solve a system by back substitution, starting from the last equation. Gauss-Jordan elimination also clears the entries above each pivot, so it ends at RREF and each pivot variable can be read directly. The how to find RREF guide goes through both phases step by step.
Which form should you use?
- Solving a system by hand: REF plus back substitution is usually less arithmetic.
- Reading a solution, free variables or a parametric answer: RREF, because each pivot variable is given directly. Try the augmented system solver.
- Rank, pivot columns and linear independence: either form, since the pivot positions are the same.
- Null space basis: RREF, where each free column gives one basis vector.
- Inverse of a matrix: RREF of [A | I].
- Determinant: REF, by multiplying the diagonal entries and tracking row swaps and scalings.
- Checking an answer or comparing with a classmate: RREF, because it is unique, so two correct answers always match.
Graphing calculators offer both as ref( and rref(; see RREF on a TI-84.
Compare them on your own example.
Use the Form menu in the RREF calculator. Changing the selection clears the previous result so you can calculate the new form explicitly and compare the row operations.
Questions about REF and RREF
Is RREF always unique?
Yes. Every matrix has exactly one reduced row echelon form, whatever row operations you use to reach it.
Can a matrix have more than one REF?
Yes. Different choices of row operations can give different row echelon forms, as the 3×4 example above shows. Their pivot positions are always the same.
Does REF need leading 1s?
It depends on the textbook. Some define REF with every leading entry equal to 1, others only require zeros below each leading entry. RREF always requires leading 1s.
Is every RREF also an REF?
Yes. RREF meets all the REF conditions plus two more, so every matrix in RREF is also in row echelon form.
Do REF and RREF give the same rank?
Yes. The rank is the number of pivots, and the pivot positions are the same in every REF and in the RREF.
References: University of Texas at Austin, Matrix Operations appendix and TI-Basic Developer, ref( command reference. Examples independently calculated with exact arithmetic. Sources checked September 25, 2026.