Row-Echelon and Reduced Row-Echelon Form

The Central Idea

Consider the matrix

The leading entries move to the right as we move downward through the rows.

The entries below each leading entry are zero.

This staircase pattern is the essential idea of row-echelon form.

If we continue reducing until each leading entry is \(1\) and is the only nonzero entry in its column, we obtain reduced row-echelon form.

Leading Entries

In a nonzero row, the leading entry is the first nonzero entry from the left.

For example, in

the leading entry is

Pivot Positions

A leading entry used to organize row reduction is called a pivot.

The location of a pivot is called a pivot position.

Pivots help determine:

• which variables are directly constrained;
• which columns are pivot columns;
• which variables may remain free.

Pivot Columns

A pivot column is a column containing a pivot position.

For example, in

all three columns are pivot columns.

Zero Rows

A zero row contains only zeros.

For example,

In echelon form, all zero rows must appear below all nonzero rows.

Definition of Row-Echelon Form

A matrix is in row-echelon form if:

1. all nonzero rows are above any zero rows;
2. the leading entry of each lower nonzero row lies to the right of the leading entry above it;
3. all entries below each leading entry are zero.

The Staircase Pattern

Row-echelon form often looks like a staircase:

The boxed entries represent pivots.

Example of Row-Echelon Form

The matrix

is in row-echelon form.

The pivots are

Pivots Need Not Be \(1\) in REF

In row-echelon form, pivot entries do not have to equal \(1\).

For example,

is still in row-echelon form.

Zeros Above Pivots Are Not Required in REF

Consider

This is in row-echelon form even though entries above the second and third pivots are nonzero.

REF requires zeros below pivots, not above them.

Definition of Reduced Row-Echelon Form

A matrix is in reduced row-echelon form if it is in row-echelon form and also:

1. every pivot equals \(1\);
2. each pivot is the only nonzero entry in its column.

Example of RREF

The matrix

is in reduced row-echelon form.

The first two columns are pivot columns.

The third column is not.

REF Versus RREF

Why REF Is Useful

Row-echelon form creates a triangular system.

That structure supports back substitution.

For example,

can be solved from the bottom row upward.

Back Substitution from REF

The last row gives

so

The second row gives

Thus,

Finish the REF Example

The first row gives

Substitute

Then

So

Therefore,

Why RREF Is Useful

Reduced row-echelon form often lets us read the solution directly.

For example,

immediately gives

RREF with a Free Variable

Consider

The first and second columns are pivot columns.

The third column has no pivot.

Therefore, \(z\) is free.

Parameterize the Free Variable

Let

Then the first row gives

so

The second row gives

so

Thus,

Pivot Variables

A variable corresponding to a pivot column is called a pivot variable.

In the previous example:

and

are pivot variables.

The variable

is free.

Free Variables

A free variable corresponds to a nonpivot variable column.

Free variables are not fixed to one unique value by the system.

They are assigned parameters so that the entire solution family can be described.

Unique Solution Criterion

For a consistent system with \(n\) variables, a unique solution occurs when every variable column contains a pivot.

Then no free variables remain.

Infinite-Solution Criterion

If a system is consistent but at least one variable column lacks a pivot, then at least one free variable remains.

The system has infinitely many solutions.

Inconsistency Criterion

An augmented matrix is inconsistent if row reduction produces a row of the form

where

This represents

which is impossible.

Example of Inconsistency

Consider

The second row represents

Therefore,

Zero Rows and Redundancy

A zero row such as

represents

It adds no new constraint.

Such rows usually arise from dependent or redundant equations.

Zero Rows Belong at the Bottom

A matrix such as

is not in row-echelon form because a nonzero row appears below a zero row.

Leading Entries Must Move Right

Consider

The second and third rows both begin in column \(2\).

Therefore, this is not in row-echelon form.

Zeros Must Appear Below Each Pivot

The matrix

is not in row-echelon form because the entry below the first leading entry is nonzero.

RREF Requires Pivot \(1\)s

The matrix

is in row-echelon form but not reduced row-echelon form because the first pivot is

rather than

RREF Requires Zeros Above Pivots

The matrix

is in row-echelon form.

It is not in reduced row-echelon form because the second pivot column contains a nonzero entry above the pivot.

Reducing REF to RREF

Start with

First scale the second row:

Then

Clear Above the Pivot

Now use

This gives

The matrix is now in reduced row-echelon form.

REF Is Not Unique

A matrix can often be reduced to more than one valid row-echelon form depending on the row operations chosen.

For example, different nonzero scalar multiples of pivot rows may still satisfy the REF conditions.

RREF Is Unique

Every matrix is row equivalent to exactly one reduced row-echelon form.

This uniqueness makes RREF especially useful for classification and comparison.

Different correct row-reduction paths must eventually produce the same RREF.

Why RREF Uniqueness Matters

If two students use different legal row operations on the same matrix, their intermediate work may look different.

If both reduce completely and correctly, their final RREF must agree.

This provides a strong consistency check.

Row-Echelon Form and Gaussian Elimination

Gaussian elimination typically reduces a matrix to row-echelon form.

The system is then solved by back substitution.

Thus, REF is the natural target of forward elimination.

Reduced Row-Echelon Form and Gauss-Jordan Elimination

Gauss-Jordan elimination continues beyond REF until the matrix reaches RREF.

This eliminates both below and above each pivot.

Forward Elimination

During forward elimination:

1. choose a pivot;
2. move it into position if needed;
3. create zeros below it;
4. move down and right to the next pivot position.

The result is row-echelon form.

Backward Reduction

During backward reduction:

1. scale pivots to \(1\);
2. create zeros above each pivot;
3. continue upward until every pivot column is clean.

The result is reduced row-echelon form.

Example: Reduce to REF

Start with

Use

and

This gives

Finish the REF Reduction

Now use

Then

This matrix is in row-echelon form.

Solve from the REF Matrix

The last row gives

so

The second row gives

so

The first row gives

so

Thus,

Continue the Example to RREF

From

scale row \(3\):

Then

Clear Above the Third Pivot

Use

and

Then

Clear Above the Second Pivot

Use

Then

This is RREF.

RREF Makes the Solution Immediate

The RREF matrix gives

No back substitution is needed.

REF and RREF Encode the Same Solution Set

The original matrix, its REF, and its RREF are row equivalent.

They may look very different, but they represent equivalent systems with exactly the same solutions.

Classifying a System from REF

Suppose row reduction gives

There are three variable columns but only two pivots.

The system is consistent with one free variable.

Therefore, it has infinitely many solutions.

Classifying a System from a Contradiction Row

Suppose row reduction gives

The last row says

Therefore,

Rank Preview

The number of pivot positions in a matrix is closely related to its rank.

Rank measures how much independent row-column structure survives after reduction.

This concept will later help classify systems and understand linear dependence.

Rank and Free Variables

For a consistent system with \(n\) variables:

Later, this becomes part of the rank-nullity relationship.

Homogeneous Systems in RREF

A homogeneous system has augmented column zero:

Its RREF can reveal whether the only solution is the zero vector or whether free variables generate nontrivial solutions.

Homogeneous Unique Solution

If every variable column has a pivot, then the homogeneous system has only

This is the trivial solution.

Homogeneous Infinite Solutions

If at least one variable is free, then the homogeneous system has infinitely many solutions.

Because the system is homogeneous, those solutions form a structured set passing through the origin.

RREF and Solution Geometry

RREF reveals the dimension of the solution family:

• no free variables: a point;
• one free variable: typically a line;
• two free variables: typically a plane;
• more free variables: higher-dimensional solution sets.

One Free Variable Means One Degree of Freedom

Suppose

represents two independent equations in three variables.

One variable is free.

Thus, the solution set has one degree of freedom.

Two Free Variables Mean Two Degrees of Freedom

One independent equation in three variables leaves two free variables.

Geometrically, the solution set is typically a plane.

Pivot Structure Does Not Depend on Row-Reduction Path

Different valid row operations may produce different intermediate matrices.

But the final RREF has the same pivot positions.

This means pivot structure is an intrinsic feature of the matrix, not an accident of one elimination path.

REF Can Be Faster by Hand

If the goal is only to solve one system, stopping at REF and using back substitution may require fewer steps than reducing fully to RREF.

This is why Gaussian elimination is often computationally efficient.

RREF Can Be Better for Interpretation

If the goal is to:

• identify free variables;
• describe the entire solution set;
• compare equivalent systems;
• study rank or dependence;

RREF is often more transparent.

Do Not Confuse REF with RREF

A matrix may be in REF without being in RREF.

For example,

is REF.

It is not RREF because the second pivot is not \(1\), and the entry above it is nonzero.

Do Not Assume Every Staircase Matrix Is REF

Check all conditions.

A matrix may look triangular but fail REF if:

• a zero row appears above a nonzero row;
• a lower leading entry does not move right;
• an entry below a pivot is nonzero.

RREF of the Identity Matrix

The identity matrix

is already in reduced row-echelon form.

Each pivot is \(1\), and every pivot column contains zeros everywhere else.

RREF of a Zero Matrix

A zero matrix is also in reduced row-echelon form.

It contains no nonzero rows and therefore violates none of the echelon conditions.

Augmented RREF and Direct Reading

Consider

The system is simply

This is the ideal direct-reading structure.

RREF with a Nonpivot Column

Consider

The third variable column is nonpivot.

Thus, the third variable is free.

A Reliable REF-Recognition Workflow

1. Move all zero rows mentally to the bottom and check whether they are already there.
2. Locate the leading entry of each nonzero row.
3. Check that each leading entry moves right as rows descend.
4. Check that every entry below each leading entry is zero.

A Reliable RREF-Recognition Workflow

1. First verify all REF conditions.
2. Check that every pivot equals \(1\).
3. Check that each pivot is the only nonzero entry in its column.

A Reliable REF-Solving Workflow

1. Reduce the augmented matrix to REF.
2. Check for contradiction rows.
3. Identify pivot and free variables.
4. Solve the bottom nonzero row first.
5. Back-substitute upward.
6. Parameterize free variables if necessary.

A Reliable RREF-Solving Workflow

1. Reduce the augmented matrix fully to RREF.
2. Check for inconsistency.
3. Identify pivot and nonpivot variable columns.
4. Assign parameters to free variables.
5. Read pivot variables directly from the rows.
6. Write the solution set clearly.

A Reliable Classification Workflow

1. Row reduce to REF or RREF.
2. If a contradiction row appears, classify the system as inconsistent.
3. If the system is consistent, count pivot columns among variable columns.
4. If every variable column has a pivot, the solution is unique.
5. If at least one variable column is nonpivot, free variables remain and infinitely many solutions exist.

Common Errors

Requiring Pivots to Equal \(1\) in REF

That condition belongs to RREF, not ordinary REF.

Requiring Zeros Above Pivots in REF

REF requires zeros below pivots. RREF requires zeros both below and above.

Leaving a Zero Row Above a Nonzero Row

All zero rows must be at the bottom in echelon form.

Failing to Move Leading Entries Right

Each lower pivot must lie strictly to the right of the pivot above it.

Assuming a Zero Row Means No Solution

A zero row represents \(0=0\), not a contradiction.

Missing a Contradiction Row

A row ending in a nonzero augmented value after all variable entries are zero proves inconsistency.

Calling a Nonpivot Variable “Missing”

It is usually a free variable, not an error.

Failing to Parameterize Free Variables

Infinite solution sets should be described explicitly.

Assuming REF Is Unique

Many different REF matrices can be row equivalent to the same original matrix.

Assuming RREF Depends on the Reduction Path

RREF is unique for each matrix.

Practice

1. Identify REF

Is

in row-echelon form?

2. Identify RREF

Is

in reduced row-echelon form?

3. REF but Not RREF

Classify

4. Not REF

Is

in row-echelon form?

5. Find Pivot Columns

For

which columns are pivot columns?

6. Identify the Free Variable

For the previous matrix, if the first three columns correspond to \(x,y,z\), which variable is free?

7. Classify a Unique System

What does

tell you?

8. Classify an Infinite System

What does

tell you?

9. Classify an Inconsistent System

What does

tell you?

10. Solve from REF

Solve

11. Parameterize from RREF

Describe the solutions of

12. RREF Uniqueness

Can two correct complete row reductions of the same matrix produce different RREF matrices?

Key Ideas

• The leading entry of a nonzero row is its first nonzero entry from the left.
• Pivots organize the structure of row reduction.
• In row-echelon form, zero rows are at the bottom, pivots move right as rows descend, and entries below pivots are zero.
• REF does not require pivots to equal \(1\).
• REF does not require zeros above pivots.
• Reduced row-echelon form requires every pivot to equal \(1\) and be the only nonzero entry in its column.
• REF supports back substitution.
• RREF often allows the solution structure to be read directly.
• Pivot columns correspond to pivot variables.
• Nonpivot variable columns correspond to free variables.
• A consistent system with a pivot in every variable column has a unique solution.
• A consistent system with at least one free variable has infinitely many solutions.
• A contradiction row proves inconsistency.
• REF is generally not unique, while RREF is unique.
• Gaussian elimination targets REF; Gauss-Jordan elimination targets RREF.

Where This Leads

Row-echelon and reduced row-echelon form give row reduction a precise destination. The staircase of pivots reveals which variables are constrained, which remain free, whether contradictions occur, and whether the system has one solution, infinitely many solutions, or none.

From here, the section can formalize Gaussian elimination and Gauss-Jordan elimination as complete algorithms, then connect pivots and reduced forms to rank, matrix equations, inverses, determinants, homogeneous systems, and the broader structure of linear algebra.