Understanding Matrices and Linear Maps

A matrix is a rectangular array of numbers organised into rows and columns. When viewed as a linear transformation, a matrix describes how to move or rotate vectors in space. For instance, a 2×2 matrix can rotate or scale vectors in a plane, a 3×3 matrix operates on three-dimensional space, and a 4×4 matrix works in four-dimensional space.

The relationship between a matrix A and a vector x is central to linear algebra. When we multiply A by x, we obtain Ax—a new vector in the output space. Different input vectors produce different outputs; some map to the zero vector, and those special inputs form the null space.

The Null Space and Kernel

The null space of a matrix A is the set of all vectors x such that Ax = 0. In matrix theory, this is synonymous with the kernel of the matrix when treated as a linear map.

Unlike finite sets, the null space typically contains infinitely many vectors (unless it only contains the zero vector itself). To describe it completely, we determine a basis—a minimal set of linearly independent vectors from which every vector in the null space can be expressed as a linear combination. The basis vectors are the generators of the null space.

  • Trivial null space: Contains only the zero vector (full rank matrices).
  • Non-trivial null space: Contains infinitely many vectors spanning a proper subspace.

Finding the Null Space

To find the null space, we use Gauss-Jordan elimination to reduce the matrix to reduced row echelon form (RREF). This process:

  1. Applies elementary row operations (row swaps, scalar multiplication, and row addition).
  2. Converts the matrix into RREF, where pivot positions clearly identify dependent and free variables.
  3. Expresses free variables as parameters to construct basis vectors.

For an m × n matrix with rank r, the nullity (dimension of null space) is n − r.

Nullity = Number of Columns − Rank of Matrix

  • Nullity — The dimension of the null space—the number of free variables after row reduction.
  • Rank — The number of pivot (non-zero) rows in the reduced row echelon form.
  • Number of Columns — The number of variables (columns) in the original matrix.

Practical Example

Consider a 3×4 matrix:

⎡ 1 2 3 4 ⎤
⎢ 2 4 6 8 ⎥
⎣ 3 6 9 12 ⎦

Notice that rows 2 and 3 are multiples of row 1. After row reduction, we find rank = 1. With 4 columns, the nullity = 4 − 1 = 3. This means the null space has 3 basis vectors, and the system Ax = 0 has infinitely many solutions described by three free parameters.

Common Pitfalls and Caveats

Avoid these mistakes when working with null spaces:

  1. Confusing rank with dimension — The rank tells you the number of independent rows, not the dimension of the null space. Use the rank-nullity theorem: <span style="font-family:monospace">rank + nullity = number of columns</span>. A full-rank square matrix has a trivial null space containing only zero.
  2. Arithmetic errors in row reduction — Gauss-Jordan elimination requires precise arithmetic. A single computational mistake propagates through the remaining steps, producing incorrect basis vectors. Always double-check pivot operations and ensure you're eliminating both above and below each pivot, not just below.
  3. Forgetting to identify free variables — After row reduction, free variables (non-pivot columns) must be parameterised to construct basis vectors. Skipping this step leaves you with partial information. Each free variable corresponds to one dimension of the null space.
  4. Assuming the null space is always non-trivial — Square matrices with full rank (equal to their size) have null spaces containing only the zero vector. Not all matrices have 'interesting' null spaces; check the rank first to determine whether basis vectors exist beyond zero.

Frequently Asked Questions

What does it mean for a vector to be in the null space?

A vector <span style="font-family:monospace">x</span> is in the null space of matrix <span style="font-family:monospace">A</span> if and only if multiplying <span style="font-family:monospace">A</span> by <span style="font-family:monospace">x</span> produces the zero vector. Geometrically, it represents a direction that the matrix completely 'collapses' to zero. For example, if <span style="font-family:monospace">A</span> represents a projection onto a plane, any vector perpendicular to that plane lies in its null space.

How does the null space relate to solving linear systems?

When solving <span style="font-family:monospace">Ax = b</span>, the null space determines the solution structure. If <span style="font-family:monospace">x₀</span> is one particular solution, then all solutions have the form <span style="font-family:monospace">x₀ + v</span>, where <span style="font-family:monospace">v</span> is any vector in the null space. For homogeneous systems (<span style="font-family:monospace">b = 0</span>), the entire solution set is the null space itself, including infinitely many solutions if the null space is non-trivial.

Can a matrix have an empty null space?

No. Every matrix always contains at least the zero vector in its null space, since <span style="font-family:monospace">A · 0 = 0</span> for any matrix <span style="font-family:monospace">A</span>. However, the null space can be <em>trivial</em>, meaning it contains only the zero vector. This occurs when the matrix has full column rank, which happens when all columns are linearly independent.

What is the difference between null space and column space?

The null space (kernel) is the set of input vectors that map to zero. The column space (range) is the set of all possible outputs when the matrix multiplies any vector. These are dual concepts: a vector is in the null space if it's perpendicular (orthogonal in a generalised sense) to every row of the matrix. Together, they characterise how the matrix transforms vectors.

Why does the rank-nullity theorem matter?

The rank-nullity theorem states: <span style="font-family:monospace">rank(A) + nullity(A) = number of columns</span>. This fundamental relationship ensures that you've correctly identified all dimensions of a matrix's behaviour. It provides a consistency check: if you compute rank and nullity separately, they must sum to the column count, confirming your row reduction was accurate.

How do free variables relate to basis vectors?

After Gauss-Jordan elimination, free variables (non-pivot columns) parameterise the null space. Each free variable becomes a parameter in the basis vector construction. If there are <em>k</em> free variables, the null space has dimension <em>k</em>, and you'll construct <em>k</em> linearly independent basis vectors by setting each free variable to 1 (one at a time) and the others to 0.

More math calculators (see all)