What is RREF (reduced row echelon form)?

Reduced row echelon form (RREF) is a unique canonical form of a matrix achieved through Gaussian–Jordan elimination. A matrix is in RREF when: (1) every leading entry in a non-zero row is 1, called a pivot; (2) each pivot is the only non-zero entry in its column; (3) pivot columns are to the left of pivot columns in lower rows; and (4) all-zero rows are at the bottom. RREF is used to solve systems of linear equations, determine the rank of a matrix, and find the null space.

How does row reduction work?

Row reduction uses three elementary row operations: (1) swap two rows, (2) multiply a row by a non-zero scalar, and (3) add a multiple of one row to another. The Gauss-Jordan algorithm works column by column, finding the pivot (using partial pivoting for numerical stability), scaling the pivot row so the pivot equals 1, and then eliminating every other entry in that column above and below. Repeating this for each pivot column produces the RREF.

What is the difference between REF and RREF?

Row echelon form (REF) requires only that pivots move strictly right as you go down rows, and that all zeros are below each pivot. RREF adds the additional requirements that every pivot equals 1 and that every entry above each pivot is also zero. REF is sufficient for back-substitution, but RREF gives the solution directly with no additional substitution needed. Gaussian elimination produces REF; Gauss-Jordan elimination produces RREF.

When is a matrix in RREF?

A matrix is in RREF when all four conditions hold: (1) any all-zero rows are at the bottom; (2) the first non-zero entry in each non-zero row (the leading entry or pivot) is 1; (3) each pivot is to the right of the pivot in the row above it; and (4) every entry in a pivot column, other than the pivot itself, is zero. The identity matrix is always in RREF.

What are the elementary row operations?

There are exactly three elementary row operations: (1) Row swap — interchange two rows (written R₁ ↔ R₂); (2) Row scaling — multiply all entries in a row by a non-zero constant (written Rᵢ ← k·Rᵢ); (3) Row addition — add a multiple of one row to another (written Rᵢ ← Rᵢ + k·Rⱼ). These operations do not change the solution set of the linear system represented by the matrix and together are sufficient to achieve any RREF.

How do you use RREF to solve a system of equations?

Write the system as an augmented matrix [A|b] where A contains the coefficients and b is the right-hand side column. Compute the RREF of the augmented matrix. If a row has all zeros except in the last column, the system is inconsistent (no solution). Otherwise, each pivot column corresponds to a basic variable that can be read directly from the RREF: if the RREF row for variable x is [1 0 0 | 5], then x = 5. Columns without pivots correspond to free variables, indicating infinitely many solutions.

What does matrix rank tell you?

The rank of a matrix is the number of pivot columns in its RREF — equivalently, the number of linearly independent rows or columns. For an m×n matrix A representing a system Ax = b: if rank(A) = n (full column rank), the system has at most one solution. If rank(A) < n, there are free variables and potentially infinitely many solutions. If rank(A) < rank([A|b]), the system is inconsistent (no solution). Rank also tells you the dimension of the column space — how many independent directions the transformation spans.

How is Gaussian elimination used in practice?

Gaussian elimination (row reduction) is one of the most important numerical algorithms in use. It solves systems of equations in finite element analysis (structural engineering, fluid dynamics), linear programming (operations research), computer graphics (3D transformations), and machine learning (solving the normal equations for linear regression). Virtually every scientific simulation involving multiple simultaneous constraints uses Gaussian elimination at its core. Modern implementations use partial pivoting for numerical stability.

RREF Calculator — Row Reduce Any Matrix

Row reduces any 2×2, 2×3, 3×3, or 3×4 matrix to reduced row echelon form with step-by-step operations.

How to Use the RREF Calculator

This RREF calculator row reduces any matrix to reduced row echelon form — select the matrix size (2×2, 2×3, 3×3, or 3×4) then fill in every cell. For systems of equations, choose a size where the number of columns equals the number of variables plus one (the right-hand side): use 2×3 for 2 equations in 2 unknowns, or 3×4 for 3 equations in 3 unknowns. Click Compute RREF to see the result matrix and the list of row operations. Toggle Show row operations to follow each step. For matrix inverses, see our inverse matrix calculator. For simpler 2×2 and 3×3 systems without matrix notation, our systems of equations calculator accepts coefficients directly and shows step-by-step Gaussian elimination.

The calculator uses partial pivoting (choosing the row with the largest absolute value in each column) to maximize numerical stability — the same strategy used by professional linear algebra libraries.

What Is Reduced Row Echelon Form?

A matrix is in reduced row echelon form (RREF) when all four conditions are met:

Any rows consisting entirely of zeros are at the bottom.
The first non-zero entry in each non-zero row is 1 — this is called the pivot or leading 1.
Each pivot is strictly to the right of the pivot in the row above it (staircase pattern).
Every entry in a pivot column other than the pivot itself is zero — both above and below.

RREF is unique: every matrix has exactly one RREF, regardless of the sequence of row operations used to reach it. This makes it a powerful canonical form for analysis and computation.

AdvertisementResponsive Ad

The Gauss-Jordan Elimination Algorithm

Gauss-Jordan elimination extends Gaussian elimination by eliminating entries above each pivot as well as below, producing RREF directly. The algorithm proceeds column by column:

Select the pivot column — move to the next column from left to right.
Partial pivoting — find the row with the largest absolute value in this column at or below the current pivot row. Swap it into the pivot position. This reduces floating-point error.
Scale the pivot row — divide every entry in the pivot row by the pivot value so the pivot becomes 1.
Eliminate above and below — for every other row, subtract the appropriate multiple of the pivot row to zero out the entry in the pivot column.
Advance the pivot row — move to the next row and repeat.

The three allowed operations — row swap, row scaling, and row addition — preserve the solution set of the linear system at every step.

Reading the Solution from RREF

When you row-reduce an augmented matrix [A|b], the RREF reveals the solution immediately:

Unique solution: each variable column has exactly one pivot, and the last column gives the variable values directly. For a 3×4 matrix, the solution is x = last column row 1, y = last column row 2, z = last column row 3.
No solution (inconsistent): a row appears that has all zeros in the variable columns but a non-zero value in the last column — this is a contradiction (0 = non-zero).
Infinitely many solutions: the number of pivots is less than the number of variables. Columns without pivots are free variables; set them to parameters and express the other variables in terms of them.

For matrix rank and invertibility, our determinant calculator is complementary — the determinant is non-zero exactly when the matrix has full rank and a unique solution.

RREF Example: 3×4 Augmented Matrix

Solve the system: x + y + z = 6, 2x + y − z = 1, x − y + 2z = 5.

Write as the augmented matrix and apply row operations:

Initial: rows [1, 1, 1 | 6], [2, 1, −1 | 1], [1, −1, 2 | 5]
R2 ← R2 − 2·R1 → [0, −1, −3 | −11]
R3 ← R3 − R1 → [0, −2, 1 | −1]
R2 ← −R2 → [0, 1, 3 | 11]; R3 ← R3 + 2·R2 → [0, 0, 7 | 21]
R3 ← R3 / 7 → [0, 0, 1 | 3]; back-substitute: z = 3, y = 11 − 9 = 2, x = 6 − 2 − 3 = 1
RREF: [1, 0, 0 | 1], [0, 1, 0 | 2], [0, 0, 1 | 3] → x = 1, y = 2, z = 3

AdvertisementResponsive Ad

Applications of RREF

Solving linear systems — the most common use; works for any number of equations and variables, regardless of whether the system has a unique, no, or infinite solutions.
Finding matrix rank — the rank equals the number of non-zero rows in the RREF. A 3×3 matrix with 3 pivot rows has rank 3 (full rank) and is invertible.
Computing the null space — the null space (kernel) of a matrix can be read from the RREF by expressing free variables as parameters.
Testing linear independence — a set of vectors is linearly independent if and only if the matrix formed by those vectors has full column rank, visible from its RREF.
Computing matrix inverses — by augmenting with the identity matrix [A|I] and row reducing, if A is invertible you obtain [I|A⁻¹].

Common Mistakes

Skipping partial pivoting — always choose the largest pivot to avoid amplifying rounding errors. Never use a zero pivot.
Sign errors in row addition — when eliminating, double-check whether you are adding or subtracting a multiple of the pivot row.
Forgetting to eliminate above the pivot — RREF requires zeros above each pivot, not just below. REF only requires zeros below; RREF requires both.
Confusing rows and columns — row operations act on entire rows, not columns. Each operation affects every entry across that row.

Sources & References

Row Echelon Form — Khan Academy — Khan Academy
Gaussian Elimination — Math is Fun — Math is Fun

RREF Calculator