Contents

Column SpaceEdit

Column space is a core idea in linear algebra that describes what a linear transformation can produce. For an m×n matrix A, the column space Col(A) is the subspace of the real space R^m spanned by the columns of A. Equivalently, Col(A) is the set of all vectors b that can be written as b = Ax for some x in R^n; in other words, it is the image (range) of the linear transformation T(x) = Ax. The column space thus encodes the attainable outputs given the inputs, and its dimension—known as the rank of A—measures how much of the output space is reachable.

In the language of linear maps, if A represents a linear transformation T: R^n → R^m with respect to the standard bases, Col(A) corresponds to the image of T. As a subspace, Col(A) sits inside R^m and inherits the structure that comes with a subspace: it is closed under vector addition and scalar multiplication. The rank of A, denoted rank(A), equals dim(Col(A)) and also equals the maximum number of linearly independent columns of A.

Formal definition

Let A be an m×n matrix with columns a1, a2, ..., an in R^m. Then

  • Col(A) = span{ a1, a2, ..., an } ⊆ R^m.
  • Equivalently, Col(A) = { Ax : x ∈ R^n }.

The column space is a subspace of the codomain of the matrix and is intimately connected to the solvability of linear systems Ax = b.

Geometric interpretation

Geometrically, Col(A) is the set of all output vectors you can reach by combining the columns of A with real coefficients. Each x ∈ R^n provides a particular linear combination of the columns, and the resulting vector b = Ax lies in Col(A). If the columns are all multiples of each other, Col(A) is a line; if two columns are independent, Col(A) is a plane in R^m; and if the columns span all of R^m, Col(A) is the entire space.

Basis and dimension

  • A basis for Col(A) is a set of vectors that spans Col(A) and is linearly independent. A standard way to obtain a basis is to row-reduce A to its reduced row-echelon form (RREF). The pivot columns of A (the columns of A that correspond to leading ones in the RREF) form a basis for Col(A) when taken in their original form. The number of pivots equals rank(A).
  • The dimension of Col(A) is called the rank of A, written rank(A) = dim(Col(A)).

Because row operations change the matrix but not the linear relations among columns in a coarser sense, the pivot columns of A (not the pivot columns of the RREF alone) provide a practical basis for Col(A).

Computation

  • To find Col(A) and its basis:
    • Compute the row-echelon form or RREF of A.
    • Identify the pivot columns; the corresponding original columns of A form a basis for Col(A).
    • The number of pivots gives dim(Col(A)).
  • To check if a vector b is in Col(A):
    • Solve the augmented system [A | b] and see if it is consistent. If yes, b ∈ Col(A); if not, b ∉ Col(A). This is equivalent to asking whether there exists x with Ax = b.

Relation to the linear system Ax = b

  • The system Ax = b has a solution if and only if b lies in Col(A). If b is outside Col(A), no choice of x can produce b as an output of the transformation represented by A.
  • The set of all such b forms Col(A) itself, so Col(A) is precisely the collection of all right-hand sides for which the system is solvable.
  • When b is in Col(A), the set of solutions x is an affine subset of R^n, and its structure can be described using the null space (ker A) and the particular solution.

Connections to other subspaces and the rank-nullity relationship

  • Row space: Col(A) is one of several subspaces associated with A; the row space concerns the row vectors of A and is related to dual aspects of solvability. The Column Space and Row Space share a delicate interplay captured in the fundamental theorem of linear algebra for a given A.
  • Null space (kernel): The null space of A, ker(A), consists of all x with Ax = 0. The rank-nullity theorem says dim(ker(A)) + rank(A) = n, tying the dimension of the column space to the degrees of freedom left in the solution vector x.
  • Basis and dimension: The practical basis for Col(A) is obtained from pivot columns, as described above, tying together concepts of span, independence, and dimension.

Examples

  • Example 1: A = [[1, 0], [2, 1], [3, 4]] is a 3×2 matrix with columns a1 = (1, 2, 3) and a2 = (0, 1, 4). Since a1 and a2 are independent, Col(A) = R^3, and rank(A) = 2.
  • Example 2: A = [[1, 2, 3], [2, 4, 6]] is a 2×3 matrix. The columns are a1 = (1, 2), a2 = (2, 4), a3 = (3, 6), all lying on the same line (each is a multiple of a1). Col(A) is a line through the origin in R^2, and rank(A) = 1.
  • Example 3: A = [[1, 0, 0], [0, 1, 0]] is a 2×3 matrix with columns a1 = (1, 0), a2 = (0, 1), a3 = (0, 0). Col(A) = R^2, and rank(A) = 2. The pivot columns are the first two, and they form a basis for Col(A).

Applications

  • In data analysis and numerical methods, Col(A) helps determine the feasibility of linear models: whether a given set of observations can be explained by a linear combination of predictors.
  • In computer graphics, column spaces represent the range of linear transformations applied to shapes and textures, informing how objects can be distorted or projected.
  • In systems theory and control, the column space of a matrix that represents a system captures the space of reachable states under certain inputs.

See also