Null Spaces, Column Spaces, and Linear Transformations

Null Space of a Matrix

  • Definition of Null Space: The null space of an m×nm \times n matrix AA, denoted as Nul A\text{Nul } A, is the set of all solutions to the homogeneous equation Ax=0A\mathbf{x} = \mathbf{0}. Formally, Nul A={x:x∈Rn and Ax=0}\text{Nul } A = \{ \mathbf{x} : \mathbf{x} \in \mathbb{R}^n \text{ and } A\mathbf{x} = \mathbf{0} \}.
    • Geometric Context: Because the matrix AA has nn columns, the vectors x\mathbf{x} in the equation Ax=0A\mathbf{x} = \mathbf{0} have nn entries. Therefore, Nul A\text{Nul } A is a subset of Rn\mathbb{R}^n.
    • Theorem 2: The null space of an m×nm \times n matrix AA is a subspace of Rn\mathbb{R}^n. Equivalently, the set of all solutions to a system Ax=0A\mathbf{x} = \mathbf{0} of mm homogeneous linear equations in nn unknowns is a subspace of Rn\mathbb{R}^n.
    • Proof of Theorem 2: To confirm Nul A\text{Nul } A is a subspace, it must satisfy three specific properties:
      • Zero Vector: The zero vector 0\mathbf{0} is in Nul A\text{Nul } A because A0=0A\mathbf{0} = \mathbf{0}.
      • Closure Under Addition: If u\mathbf{u} and v\mathbf{v} are any two vectors in Nul A\text{Nul } A, then Au=0A\mathbf{u} = \mathbf{0} and Av=0A\mathbf{v} = \mathbf{0}. It follows that A(u+v)=Au+Av=0+0=0A(\mathbf{u} + \mathbf{v}) = A\mathbf{u} + A\mathbf{v} = \mathbf{0} + \mathbf{0} = \mathbf{0}. Thus, u+v\mathbf{u} + \mathbf{v} is in Nul A\text{Nul } A.
      • Closure Under Scalar Multiplication: If u∈Nul A\mathbf{u} \in \text{Nul } A and cc is any scalar, then A(cu)=c(Au)=c(0)=0A(c\mathbf{u}) = c(A\mathbf{u}) = c(\mathbf{0}) = \mathbf{0}. This shows that cuc\mathbf{u} is in Nul A\text{Nul } A.
    • Nature of the Null Space:
      • There is no obvious or direct relationship between the vectors in Nul A\text{Nul } A and the specific entries within the matrix AA.
      • The null space is defined implicitly. It is defined by a condition (Ax=0A\mathbf{x} = \mathbf{0}) that must be tested or checked for any given vector.
      • To produce an explicit description or list of the elements in Nul A\text{Nul } A, one must solve the equation Ax=0A\mathbf{x} = \mathbf{0}.

Finding a Spanning Set for Nul A

  • General Solution Process: Finding a spanning set for Nul A\text{Nul } A involves describing the general solution of the homogeneous equation in terms of free variables.
    • Step 1: Row Reduction: Transform the augmented matrix [A0][A \quad \mathbf{0}] into reduced echelon form. This identifies the basic variables and the free variables.
    • Step 2: Basic Variables: Express the basic variables in terms of the free variables.
    • Step 3: Parametric Vector Form: Write the general solution as a linear combination of vectors, using the free variables as the weights (parameters).
    • Characteristics of the Spanning Set:
      • Every linear combination of the resulting vectors (u,v,w\mathbf{u}, \mathbf{v}, \mathbf{w}, etc.) is an element of Nul A\text{Nul } A.
      • The produced set is automatically linearly independent because the free variables serve as the weights on these vectors.
      • The number of vectors in the spanning set for Nul A\text{Nul } A is exactly equal to the number of free variables in the equation Ax=0A\mathbf{x} = \mathbf{0}.

Column Space of a Matrix

  • Definition of Column Space: The column space of an m×nm \times n matrix AA, denoted as Col A\text{Col } A, is the set of all linear combinations of the columns of AA. If A=[a1…an]A = [\mathbf{a}_1 \dots \mathbf{a}_n], then Col A=Span{a1,…,an}\text{Col } A = \text{Span} \{ \mathbf{a}_1, \dots, \mathbf{a}_n \}.
    • Notation and Range: A typical vector in Col A\text{Col } A can be written as AxA\mathbf{x} for some vector x∈Rn\mathbf{x} \in \mathbb{R}^n. This is because the matrix-vector product AxA\mathbf{x} is inherently a linear combination of the columns of AA.
    • Theorem 3: The column space of an m×nm \times n matrix AA is a subspace of Rm\mathbb{R}^m.
    • Relationship to Linear Transformations: Col A\text{Col } A is the range of the linear transformation T(x)=AxT(\mathbf{x}) = A\mathbf{x}.
    • Full Space Condition: Col A\text{Col } A is equal to the entire space Rm\mathbb{R}^m if and only if the equation Ax=bA\mathbf{x} = \mathbf{b} has a solution for every possible vector b∈Rm\mathbf{b} \in \mathbb{R}^m.

Row Space of a Matrix

  • Definition of Row Space: If AA is an m×nm \times n matrix, each of its rows contains nn entries. These rows can be identified as vectors in Rn\mathbb{R}^n. The set of all linear combinations of these row vectors is the row space of AA, denoted as Row A\text{Row } A.
    • Subspace Location: Because each row has nn entries, Row A\text{Row } A is a subspace of Rn\mathbb{R}^n.
    • Relationship to Transpose: The rows of matrix AA are the columns of its transpose ATA^T. Consequently, the row space of AA is identical to the column space of ATA^T: Row A=Col AT\text{Row } A = \text{Col } A^T.

Comparison Between Nul A and Col A

For an m×nm \times n matrix AA:

  • Location and Subspace Membership:

    • Nul A: Subspace of Rn\mathbb{R}^n.
    • Col A: Subspace of Rm\mathbb{R}^m.
  • Definition and Clarity:

    • Nul A: Defined implicitly. You are only given the condition (Ax=0A\mathbf{x} = \mathbf{0}) that vectors must satisfy.
    • Col A: Defined explicitly. You are given the building blocks (the columns) and told how to construct the vectors.
  • Ease of Accessing Vectors:

    • Nul A: It takes time to find vectors; row operations on [A0][A \quad \mathbf{0}] are required.
    • Col A: It is easy to find vectors; the columns of AA are already displayed, and others are simple combinations of them.
  • Relation to Matrix Entries:

    • Nul A: No obvious relation between the null space vectors and the entries in matrix AA.
    • Col A: Clear relationship to matrix entries; each column of AA is directly an element of Col A\text{Col } A.
  • Testing Vector Membership:

    • Nul A: Given a vector v\mathbf{v}, it is easy to check if v∈Nul A\mathbf{v} \in \text{Nul } A by computing the product AvA\mathbf{v} and checking if it equals 0\mathbf{0}.
    • Col A: Given a vector v\mathbf{v}, it may take time to determine if v∈Col A\mathbf{v} \in \text{Col } A; row operations on the augmented matrix [Av][A \quad \mathbf{v}] are required to check consistency.
  • Properties and Mapping Connections:

    • Nul A: Nul A={0}\text{Nul } A = \{ \mathbf{0} \} if and only if the equation Ax=0A\mathbf{x} = \mathbf{0} has only the trivial solution. This is also equivalent to the linear transformation x↦Ax\mathbf{x} \mapsto A\mathbf{x} being one-to-one.
    • Col A: Col A=Rm\text{Col } A = \mathbb{R}^m if and only if Ax=bA\mathbf{x} = \mathbf{b} is consistent for every b∈Rm\mathbf{b} \in \mathbb{R}^m. This is equivalent to the linear transformation x↦Ax\mathbf{x} \mapsto A\mathbf{x} mapping Rn\mathbb{R}^n onto Rm\mathbb{R}^m.

Kernel and Range of a Linear Transformation

  • Generalization Beyond Rn\mathbb{R}^n: Subspaces in vector spaces other than Rn\mathbb{R}^n are often described using the properties of a linear transformation rather than a specific matrix.
    • Definition of Linear Transformation: A rule TT from a vector space VV into a vector space WW (T:V→WT: V \rightarrow W) such that:
      • T(u+v)=T(u)+T(v)T(\mathbf{u} + \mathbf{v}) = T(\mathbf{u}) + T(\mathbf{v}) for all u,v∈V\mathbf{u}, \mathbf{v} \in V.
      • T(cu)=cT(u)T(c\mathbf{u}) = cT(\mathbf{u}) for all u∈V\mathbf{u} \in V and all scalars cc.
    • Kernel (Null Space): The kernel (or null space) of TT is the set of all u∈V\mathbf{u} \in V such that T(u)=0T(\mathbf{u}) = \mathbf{0} (where 0\mathbf{0} is the zero vector in WW). The kernel is a subspace of the domain VV.
    • Range: The range of TT is the set of all vectors in WW that are of the form T(x)T(\mathbf{x}) for at least one x∈V\mathbf{x} \in V. The range is a subspace of the codomain WW.