1 Definition and basic concept
The column space of a matrix is the set of all vectors that can be formed by taking linear combinations of its columns. It is one of the main subspaces associated with a matrix and captures exactly which output vectors the matrix can produce.
For an \(m \times n\) matrix, each column is a vector in \(\mathbb{R}^m\) or, more generally, in the underlying vector space over the chosen field. The column space is therefore a subset of the ambient space of the columns, not of the space of coefficient vectors used to form combinations.
1.1 Matrix columns as vectors
A matrix may be viewed as a list of column vectors placed side by side. Each column contributes a direction or component that may be combined with others. In this interpretation, the matrix is not only an array of numbers but also a collection of vectors with geometric meaning.
This viewpoint is especially useful because it connects matrix multiplication with vector addition and scalar multiplication. Each column acts as a building block for the outputs generated by the matrix.
1.2 Linear combinations of columns
A linear combination of the columns is formed by multiplying each column by a scalar and adding the results. If the columns are \(c_1, c_2, \dots, c_n\), then the column space consists of all vectors of the form
\[ x_1 c_1 + x_2 c_2 + \cdots + x_n c_n. \]
The coefficients \(x_1, x_2, \dots, x_n\) may vary freely. As a result, the column space includes every vector that can be generated by weighting the columns in all possible ways.
1.3 Column space as a subspace
The column space is always a subspace of the ambient vector space. It contains the zero vector, is closed under addition, and is closed under scalar multiplication. These properties follow directly from its definition as the span of a set of vectors.
Because it is a subspace, the column space can be studied using the standard tools of linear algebra, including bases, dimension, and independence. It provides a compact description of the directions accessible from the matrix.
2 Relationship to linear transformations
A matrix defines a linear transformation by acting on vectors through matrix multiplication. The column space describes the set of all possible outputs of that transformation, making it central to the study of linear maps.
2.1 Range of a matrix
The range of a matrix is another name for its column space. When a vector is multiplied by the matrix, the result must lie in the span of the columns. Conversely, every vector in that span can be obtained from some input vector.
This identification is one reason the column space is so important. It translates an algebraic object, the matrix, into a geometric and functional description of what outputs are achievable.
2.2 Image of a linear map
For a linear transformation, the image is the set of all vectors reached from the domain. The column space of a matrix representing that transformation is precisely this image. Thus, the matrix columns encode the transformation’s effect on basis vectors of the domain.
If the input basis is the standard basis, then the columns are the images of those basis vectors. The full image is obtained by combining those column images according to the coordinates of an arbitrary input vector.
2.3 Domain and codomain interpretation
A matrix of size \(m \times n\) maps vectors from an \(n\)-dimensional input space to an \(m\)-dimensional output space. The domain corresponds to the space of input coefficients, while the codomain is the space containing the outputs. The column space sits inside the codomain and consists of the outputs that are actually reachable.
This distinction matters because not every vector in the codomain must be attainable. The matrix may miss some directions entirely, in which case its column space is a proper subspace of the full output space.
3 Finding the column space
The column space can be determined from the matrix columns themselves, but row reduction often provides a more efficient route. The goal is to identify a set of columns that spans the same space with as few redundant vectors as possible.
3.1 Spanning set from the columns
The most direct spanning set for the column space is the full set of columns of the matrix. This set always spans the column space by definition. However, it may contain dependent columns that do not contribute new directions.
In practice, one often seeks a smaller spanning set that still generates the same subspace. Such a reduced set may become a basis if it is also independent.
3.2 Pivot columns and row reduction
Row reduction helps identify which columns are essential. When a matrix is reduced to echelon form, the pivot positions indicate columns that correspond to independent directions in the original matrix. The pivot columns of the original matrix form a basis for the column space.
A key point is that the row operations are performed on the matrix as a whole, but the actual basis vectors must be taken from the original matrix, not from the reduced form. The reduced form is a guide for locating the important columns.
3.3 Basis for the column space
A basis for the column space is a set of columns that spans the space and is linearly independent. Such a basis gives the most economical description of the column space, since no vector in the basis can be written as a combination of the others.
3.3.1 Selecting independent columns
Independent columns are chosen by identifying those that add new directions to the span. Pivot columns are the standard choice because they are guaranteed to be independent and to span the same subspace as the original columns. The number of such columns gives the dimension of the column space.
3.3.2 Expressing dependent columns
Dependent columns can be written as combinations of the independent ones. In row-reduced form, nonpivot columns correspond to relations among the columns of the original matrix. These relations show how some columns are redundant within the spanning set.
4 Properties
The column space has several structural properties that make it a fundamental object in linear algebra. These properties connect it with dimension, independence, and matrix rank.
4.1 Dimension and rank
The dimension of the column space is the rank of the matrix. Rank measures how many independent column directions the matrix has. A higher rank indicates a larger and more expressive column space.
Because rank is invariant under many algebraic manipulations, it serves as a stable numerical and theoretical summary of the matrix’s column structure.
4.2 Subspace structure
As a subspace, the column space inherits the familiar vector space operations. It contains the zero vector, and any sum of two vectors in the column space remains in the column space. Likewise, scaling a vector in the column space by any scalar keeps it inside the space.
This makes the column space suitable for geometric interpretation. In two dimensions it may be a line or the whole plane; in three dimensions it may be a line, plane, or all of space, depending on the matrix.
4.3 Independence and dependence
Independent columns contribute distinct directions to the column space. Dependent columns do not enlarge the span, since they can be reconstructed from other columns. The balance between independent and dependent columns determines the rank.
Understanding dependence is useful when simplifying a matrix or identifying a basis. It also clarifies which parts of a matrix are essential and which are redundant.
4.4 Invariance under elementary row operations
Elementary row operations do not preserve the columns themselves, but they do preserve the relationships needed to identify pivot positions. These operations alter the matrix into a simpler form while keeping track of which original columns belong in a basis for the column space.
For this reason, row reduction is a standard tool for studying the column space. It reveals structure indirectly without changing the final subspace spanned by the selected original columns.
5 Applications
The column space appears throughout linear algebra and its applications. It gives a practical criterion for whether a matrix equation can be solved and provides a way to describe the behavior of a linear system.
5.1 Solving linear systems
A system of linear equations can be written as a matrix equation \(Ax=b\). The system has a solution exactly when the vector \(b\) lies in the column space of \(A\). In other words, \(b\) must be expressible as a linear combination of the columns of \(A\).
This interpretation turns solvability into a geometric question. Instead of asking only whether equations match algebraically, one asks whether the target vector is contained in the span of the columns.
5.2 Determining consistency
The column space provides a direct test for consistency. If the augmented matrix row-reduces to a form showing that the right-hand side is incompatible with the column relations, then the system has no solution. When the right-hand side is in the column space, the system is consistent.
Thus, the column space identifies precisely which right-hand sides are admissible for a given coefficient matrix. It describes the set of all possible outcomes of the corresponding linear system.
5.3 Modeling transformations
In geometry and applied mathematics, matrices are used to model stretching, shearing, rotation, projection, and other linear transformations. The column space shows the set of points that can be reached from the input space under the transformation.
If the column space has lower dimension than the codomain, the transformation compresses the input into a smaller geometric region. This can reveal whether the transformation collapses dimensions or preserves full output freedom.
5.4 Data analysis and matrix factorization
In data analysis, columns often represent variables, measurements, or features. The column space then describes the subspace generated by the observed data patterns. This idea supports methods that reduce complexity by finding lower-dimensional structure.
Matrix factorizations also rely on column-space ideas. Decompositions often isolate a small set of basis columns or latent directions that explain the larger matrix. Such methods are valuable in compression, approximation, and feature extraction.
6 Examples
Concrete examples help show how the column space behaves in specific cases. Small matrices make it easier to see the connection between columns, dimension, and geometric meaning.
6.1 Small matrices
Consider the matrix
\[ \begin{pmatrix} 1 & 2 \\ 3 & 6 \end{pmatrix}. \]
Its second column is twice the first, so the column space is spanned by a single vector, \((1,3)^T\). The column space is therefore a line through the origin in \(\mathbb{R}^2\).
By contrast, the matrix
\[ \begin{pmatrix} 1 & 0 \\ 0 & 1 \end{pmatrix} \]
has columns that span all of \(\mathbb{R}^2\). Its column space is the entire plane.
6.2 Geometric interpretation in two and three dimensions
In two dimensions, the column space of a matrix may be:
- the zero subspace, if all columns are zero;
- a line through the origin, if the columns are dependent but not all zero;
- the whole plane, if two independent directions are present.
In three dimensions, the possibilities expand to include a line, a plane, or all of \(\mathbb{R}^3\). The dimension of the column space tells which of these geometric forms occurs.
6.3 Column space of special matrices
Certain matrices have especially simple column spaces. A zero matrix has only the zero vector in its column space. A diagonal matrix has a column space generated by the coordinate directions corresponding to nonzero diagonal entries.
Projection matrices typically have a column space equal to the subspace onto which they project. In such cases, the matrix acts by sending vectors onto a lower-dimensional target space, and the column space makes that target explicit.
7 Related concepts
The column space is part of a family of subspaces and results that organize linear algebra. These related ideas often appear together in the analysis of matrices and linear maps.
7.1 Row space
The row space is the span of the rows of a matrix. While the column space describes output directions, the row space captures information about the linear conditions imposed by the matrix. Both spaces have the same dimension, which is the rank.
7.2 Null space
The null space consists of all input vectors sent to zero by the matrix. It measures the directions that disappear under the transformation. The null space and column space are complementary in the study of solvability and rank.
7.3 Rank-nullity theorem
The rank-nullity theorem relates the dimension of the column space to the dimension of the null space and the number of columns of the matrix. It expresses a balance between independent output directions and lost input directions. This theorem is one of the central structural results in linear algebra.
7.4 Basis and spanning set
A basis is a linearly independent spanning set. A spanning set may contain redundant vectors, while a basis removes that redundancy without losing coverage of the space. The column space is often analyzed by finding a basis from among the matrix’s columns.