Introduction

The row space of a matrix is one of the fundamental subspaces associated with a matrix. It captures all linear combinations of the matrix’s rows and helps us understand how much “information” the rows contain. Because you already know about Span, the row space will feel like a natural extension of that idea.

This article focuses on:

• What the row space is
• How to compute it
• Why row operations do not change it
• How it relates to solutions of linear systems

What Is the Row Space?

For a matrix $A$ with rows $r_1, r_2, \dots, r_m$:

• The row space of $A$ is $$\text{Row}(A) = \text{Span}\{r_1, r_2, \dots, r_m\}.$$
• Each row is a vector in $\mathbb{R}^n$ (if the matrix has $n$ columns).
• The row space is a subspace of $\mathbb{R}^n$.

Key ideas:

• The row space contains all linear combinations of the rows.
• If some rows are redundant (linear combinations of others), they do not change the row space.
• The dimension of the row space is called the rank of the matrix.

Comparing Row Space and Column Space

The row space and column space describe two different ways a matrix generates vectors, and contrasting them helps clarify what information each one captures. Both are subspaces, both relate to rank, and both tell you something structural about the matrix—but they live in different ambient spaces and answer different geometric questions.

Where Each Space Lives

• Row space is a subspace of $\mathbb{R}^n$ (where $n$ is the number of columns).
  Each row is a vector with $n$ components.
• Column space is a subspace of $\mathbb{R}^m$ (where $m$ is the number of rows).
  Each column is a vector with $m$ components.

This difference in dimension is the first major distinction: the two spaces generally live in different universes.

What Each Space Represents

• Row space collects all linear combinations of the rows.
  It describes the constraints placed on variables in a system $Ax = b$.
• Column space collects all linear combinations of the columns.
  It describes which vectors $b$ can appear on the right-hand side of $Ax = b$.

A useful way to think about it:

• Row space → relationships between variables.
• Column space → relationships between equations.

Rank and the Shared Dimension

Even though the two spaces live in different ambient spaces, they share a key invariant:

• The dimension of the row space
• The dimension of the column space

are always equal. This common dimension is the rank of the matrix.

So while the spaces themselves differ, their “size” is the same.

Why Row Operations Preserve the Row Space

Elementary row operations:

• Swap two rows
• Multiply a row by a nonzero scalar
• Add a multiple of one row to another

These operations do not change the row space. Why?

• Each new row is still a linear combination of the original rows.
• Each original row can be recovered as a linear combination of the new rows.

This means:

• Row-equivalent matrices have the same row space.
• To find a basis for the row space, we can row‑reduce the matrix to row echelon form or reduced row echelon form (RREF).

Finding a Basis for the Row Space

To compute a basis:

1. Write down the matrix.
2. Perform row operations to reach RREF.
3. Take the nonzero rows of the RREF.
4. These rows form a basis for the row space.

Why this works:

• Nonzero rows in RREF are linearly independent.
• They span the same space as the original rows.

Example:

• Suppose $\operatorname{rref}(A)$ has nonzero rows $$[1\ 0\ 2],\quad [0\ 1\ -3].$$ Then these two vectors form a basis for Row$(A)$.

What row space tells us

The row space reveals the internal structure of a matrix as a system of equations, and it often exposes information that the column space alone cannot. A clean way to see this is to look at how row‑equivalent matrices behave. Row‑equivalent matrices always have the same row space, because row operations preserve the linear relationships among the rows, but they can have different column spaces because row operations generally change the columns.

A concrete example

Consider the matrices $$C=\begin{bmatrix} 1 & 2 & 3\\ 0 & 1 & 4 \end{bmatrix}, \qquad D=\begin{bmatrix} 1 & 0 & -5\\ 0 & 1 & 4 \end{bmatrix}.$$ Matrix $D$ is obtained from $C$ by a row operation (replace the first row with the first row minus two times the second). Because row operations preserve the row space, $$\text{Row}(C)=\text{Row}(D).$$ Both matrices impose the same two independent constraints on the variables. Any solution to $Cx=b$ satisfies the same constraints as a solution to $Dx=b$, and the homogeneous systems $Cx=0$ and $Dx=0$ have identical solution sets.

But their column spaces differ. The columns of $C$ and $D$ are different sets of vectors, so the sets of reachable outputs $Cx$ and $Dx$ are not the same. This shows that the column space cannot tell you that the two matrices encode the same equations, while the row space can.

How row operations affect row and column space

Row operations reshape a matrix in ways that do change its column space, but they never change the solutions to the system $Ax=b$.

The key is that when we solve a system, we don’t row‑reduce $A$ alone—we row‑reduce the augmented matrix $[A\mid b]$. Because $b$ is transformed in sync with the rows of $A$, the relationship between $A$ and $b$ is preserved even though the columns of $A$ are altered.

This section explains why the column space changes, why the solution set does not, and how the augmented matrix keeps everything consistent.

How row operations distort the column space

Row operations act on rows, so each operation replaces the columns of $A$ with new linear combinations of the old columns. This means:

• the column vectors change,
• the column space changes,
• and the span of the columns can tilt, rotate, or shrink.

So if we only row‑reduced $A$, we would destroy the original column space and lose the information about which $b$ are reachable.

Why the solution set stays the same anyway

Even though row operations change the columns of $A$, they do not change the equations represented by the augmented matrix. The reason is simple:

• When we row‑reduce $[A\mid b]$,
• we apply the same row operations to both $A$ and $b$,
• so each transformed equation is still logically equivalent to the original one.

Each row of the augmented matrix represents an equation $$r_1x_1 + r_2x_2 + \cdots + r_nx_n = b_i.$$ A row operation transforms both sides of the equation in a way that preserves its truth set. For example:

• Scaling a row multiplies both sides of the equation by the same number.
• Adding a multiple of one row to another adds the same multiple of one equation to another.
• Swapping rows just reorders the equations.

Because the equations remain equivalent, the solution set remains identical.

This is why row reduction is a valid method for solving linear systems.

What this means for solvability

The system $$Ax=b$$ is solvable iff $b$ lies in the original column space of $A$. Row operations do not preserve that column space, but they do preserve the logical equivalence of the system. When we row‑reduce $[A\mid b]$:

• A pivot in the last column signals inconsistency.
• No pivot in the last column means the system is consistent.

This test works because the augmented matrix encodes the equations themselves, not the geometric column space of the reduced matrix.

A compact summary

• Row operations change the column space of $A$.
• Row operations do not change the solution set of $Ax=b$.
• The augmented matrix $[A\mid b]$ ensures that $b$ is transformed consistently with $A$.
• Solvability depends on the original column space, not the column space after row reduction.
• Row reduction is a tool for detecting contradictions, not for preserving column geometry.

Exercises

1. Compute the row space of $$A = \begin{bmatrix}1 & 2 & 3 \\ 2 & 4 & 6\end{bmatrix}.$$

   Solution

   The second row is $2$ times the first, so the row space is $$\text{Span}\{(1,2,3)\}.$$
2. Find a basis for the row space of $$B = \begin{bmatrix}1 & 0 & -1 \\ 3 & 0 & -3 \\ 2 & 5 & 1\end{bmatrix}.$$

   Solution

   Row‑reducing $B$ gives nonzero rows $$(1,0,-1),\ (0,1,4).$$ A basis is $\{(1,0,-1), (0,1,4)\}$.
3. True or false: Row operations can change the row space of a matrix.

   Solution

   False. Row operations do not change the row space.
4. Determine the dimension of the row space of $$C = \begin{bmatrix}1 & 1 \\ 2 & 2 \\ 3 & 3\end{bmatrix}.$$

   Solution

   All rows are multiples of $(1,1)$, so the dimension is $1$.
5. Let $$D = \begin{bmatrix}1 & 2 & 0 \\ 0 & 1 & 3\end{bmatrix}.$$ Write a basis for Row$(D)$.

   Solution

   The rows are already independent. A basis is $$\{(1,2,0),\ (0,1,3)\}.$$
6. Explain why the nonzero rows of RREF$(A)$ always form a basis for the row space.

   Solution

   In RREF, each nonzero row has a pivot in a unique column, so the rows are linearly independent. Because row operations preserve the row space, these rows span the same space as the original rows.
7. Compute the row space of $$E = \begin{bmatrix}0 & 1 & 2 \\ 0 & 2 & 4 \\ 1 & 0 & 0\end{bmatrix}.$$

   Solution

   Row‑reducing $E$ gives nonzero rows $$(1,0,0),\ (0,1,2).$$ So the row space is $$\text{Span}\{(1,0,0),\ (0,1,2)\}.$$