Elementary Row Operations: The Three Operations Explained

Elementary row operations are the three legal operations used to transform the rows of a matrix during row reduction:

• swap two rows;
• multiply a row by a nonzero scalar;
• add a scalar multiple of one row to another row.

Every step in Gaussian elimination and Gauss-Jordan elimination is built from these operations.

They matter because each operation is reversible and preserves row equivalence. When applied correctly to an augmented matrix, they preserve the solution set of the corresponding linear system.

The Three Elementary Row Operations

Use the matrix

[ 1   2 ]
[ 3   4 ]

as a running example.

Its determinant is:

(1)(4) - (2)(3) = -2

That will be useful later when we examine how row operations affect determinants.

Operation 1: Row Swap

Notation: R1 ↔ R2

Interchange two rows.

Applying R1 ↔ R2 gives:

[ 3   4 ]
[ 1   2 ]

The rows have simply exchanged positions.

Inverse: Perform the same swap again.

R1 ↔ R2

A row swap is its own inverse.

Operation 2: Row Scaling

Notation: Ri = k · Ri, where k ≠ 0

Multiply every entry in one row by the same nonzero scalar.

Applying R1 = 3 · R1 gives:

[ 3   6 ]
[ 3   4 ]

Every entry in row 1 has been multiplied by 3.

Why must k ≠ 0?

Because the operation must be reversible.

If you multiply a row by a nonzero number k, you can reverse the operation by multiplying that row by 1/k.

But if k = 0, every possible original row becomes the same zero row:

[ a1   a2   ...   an ]

becomes

[ 0    0    ...    0 ]

Once this happens, you cannot determine what the original row was. Many different rows could have produced the same zero row.

The operation is therefore not invertible.

That is why multiplication by zero is not an elementary row operation.

Inverse: If Ri = k · Ri, reverse it by multiplying the same row by 1/k.

Operation 3: Row Replacement

Notation: Ri = Ri + k · Rj, where i ≠ j

Replace one row by itself plus a scalar multiple of another row.

Applying R2 = R2 - 3R1 to

[ 1   2 ]
[ 3   4 ]

gives:

[ 1    2 ]
[ 0   -2 ]

because:

[3, 4] - 3[1, 2] = [0, -2]

Row 1 remains unchanged.

Inverse: Negate the scalar.

R2 = R2 + 3R1

undoes:

R2 = R2 - 3R1

Why These Operations Matter

Elementary row operations have several important properties.

They are reversible

Each operation has an inverse that is itself an elementary row operation:

• Row swap → swap the same rows again.
• Row scaling by k → scale by 1/k.
• Row replacement by k → use -k.

This reversibility is central to why row operations preserve the information encoded by a matrix.

They generate row equivalence

Two matrices are row-equivalent if one can be obtained from the other by a finite sequence of elementary row operations.

Row equivalence is therefore defined in terms of these three operations.

They can be represented by elementary matrices

Each specific elementary row operation on a matrix with m rows corresponds to an m × m elementary matrix.

Left-multiplying by that elementary matrix performs the row operation.

That connection is developed later in the article.

What Elementary Row Operations Preserve

Elementary row operations preserve several important properties.

They preserve the solution set of a linear system

Suppose a system is written as an augmented matrix:

[A | b]

Applying elementary row operations to the augmented matrix produces a row-equivalent system with exactly the same solution set.

For example:

R2 = R2 - 3R1

replaces one equation by a linear combination of the equations, but it does not change which values of the variables satisfy the system.

This is why row reduction works.

They preserve rank

Every elementary row operation preserves the rank of a matrix.

If matrix B is obtained from matrix A through a sequence of elementary row operations, then:

rank(B) = rank(A)

This is why you can reduce a matrix to REF or RREF and count its pivots there to determine the rank of the original matrix.

They do not preserve everything

Elementary row operations generally change:

• individual matrix entries;
• determinant values;
• eigenvalues;
• column space.

So row equivalence does not mean two matrices are identical in every linear-algebraic sense.

It means they are connected by elementary row operations and therefore share the properties those operations preserve.

Apply Row Operations to the Entire Augmented Row

When solving a linear system, an elementary row operation must be applied to the entire row of the augmented matrix.

Suppose an augmented matrix contains a row such as:

[ 2   3  |  7 ]

and you apply:

R2 = R2 - 2R1

The operation must be applied to every entry in row 2, including the constant on the right side of the augmentation bar.

You cannot modify only the coefficient entries and leave the constant unchanged.

Doing so would produce a different system.

This is one of the most common mechanical errors in row reduction.

How Elementary Row Operations Affect the Determinant

For a square matrix, each elementary row operation has a predictable effect on the determinant.

Row Swap

Swapping two rows multiplies the determinant by -1.

Starting from:

[ 1   2 ]
[ 3   4 ]

the determinant is:

det = -2

Swap the rows:

R1 ↔ R2

[ 3   4 ]
[ 1   2 ]

Then:

det = (3)(2) - (4)(1) = 2

The determinant changes from -2 to 2.

So:

Row swap → determinant changes sign.

Row Scaling

Multiplying one row by k multiplies the determinant by k.

Apply:

R1 = 3R1

The matrix becomes:

[ 3   6 ]
[ 3   4 ]

Its determinant is:

det = (3)(4) - (6)(3)

det = 12 - 18

det = -6

Since the original determinant was -2:

-2 × 3 = -6

So:

Row scaling by k → determinant is multiplied by k.

Row Replacement

Adding a scalar multiple of one row to another leaves the determinant unchanged.

Apply:

R2 = R2 - 3R1

The matrix becomes:

[ 1    2 ]
[ 0   -2 ]

Its determinant is:

det = (1)(-2) - (2)(0)

det = -2

The determinant is unchanged.

So:

Row replacement → determinant stays the same.

Row replacement is especially useful in elimination because it directly creates zeros in selected positions. Its determinant-preserving property is also useful when elimination is being used to calculate a determinant.

Elementary Row Operations and Elementary Matrices

An elementary matrix is obtained by performing one elementary row operation on an identity matrix.

If A is an m × n matrix, the elementary matrix that acts on its rows must be m × m.

If E is that elementary matrix, then:

E · A

produces the same result as applying the corresponding row operation directly to A.

For a 3 × 3 setting:

Row Swap: R1 ↔ R2

Start with the identity matrix and swap rows 1 and 2:

[ 0   1   0 ]
[ 1   0   0 ]
[ 0   0   1 ]

Left-multiplying by this matrix swaps rows 1 and 2 of any compatible matrix.

Row Scaling: R1 = (1/2)R1

The elementary matrix is:

[ 1/2   0   0 ]
[  0    1   0 ]
[  0    0   1 ]

Left-multiplication scales row 1 by 1/2.

Row Replacement: R2 = R2 - (1/2)R1

The corresponding elementary matrix is:

[  1    0   0 ]
[ -1/2  1   0 ]
[  0    0   1 ]

Left-multiplying by this matrix replaces row 2 by:

R2 = R2 - (1/2)R1

Every elementary matrix is invertible, and its inverse is the elementary matrix corresponding to the inverse row operation.

A sequence of row operations therefore corresponds to:

Ek · ... · E2 · E1 · A

Row reduction can be viewed algebraically as successive left-multiplication by invertible elementary matrices.

Putting It Together: Gaussian and Gauss-Jordan Elimination

Gaussian elimination and Gauss-Jordan elimination both use elementary row operations.

Gaussian Elimination

Gaussian elimination reduces a matrix to Row Echelon Form (REF).

The process typically moves from left to right, using row operations as needed to:

• place nonzero entries in pivot positions;
• eliminate entries below the pivots;
• create the staircase structure of REF.

Once REF is reached, a system can usually be solved by back-substitution.

The pivots do not have to equal 1 in REF.

Gauss-Jordan Elimination

Gauss-Jordan elimination continues until the matrix reaches Reduced Row Echelon Form (RREF).

In RREF:

• every pivot equals 1;
• every pivot is the only nonzero entry in its column;
• pivot positions move strictly to the right as you move downward;
• zero rows are at the bottom.

Row swaps, row scalings, and row replacements are used as needed to achieve these conditions.

The exact order of those operations can vary.

The result is RREF, from which the solution structure can be read directly.

Both Gaussian elimination and Gauss-Jordan elimination rely on the same central fact:

Elementary row operations transform a matrix without changing the row-equivalence information needed to solve the original system.

The three operations are simple, but nearly every standard row-reduction procedure in introductory linear algebra is built from them.

Every example in this article can be verified step by step on the row reduce calculator which labels each operation in the same R₂ ← R₂ − 2·R₁ notation used here.