JOVANA
Library Glossary Getting Started Three Levels Fields How it works Mission
Join the mission
All guides

The Four Fundamental Subspaces

Every matrix carries four subspaces — column space, row space, null space, and left null space. Rank-nullity stitches them together, and two of them meet at a right angle.

Four spaces from one matrix

An m-by-n matrix A hands you four subspaces. Two live on the input side (in R^n): the row space (span of the rows) and the null space (everything A sends to zero). Two live on the output side (in R^m): the column space (span of the columns, where outputs actually land) and the left null space (the null space of the transpose A^T).

Rank-nullity ties them together

All four sizes come from one number, the rank r. Column space and row space both have dimension r — that is the theorem that row rank equals column rank wearing a new hat. On the input side, rank-nullity gives nullity = n - r. On the output side, the left null space has dimension m - r. Know r, m, and n, and you know all four.

A is m x n with rank r:
  row space      dim = r       (in R^n)
  null space     dim = n - r   (in R^n)
  column space   dim = r       (in R^m)
  left null space dim = m - r  (in R^m)

example: 3 x 4, r = 2
  row=2, null=2 (sum 4 = n)
  col=2, left null=1 (sum 3 = m)
All four dimensions, read off from r, m, and n alone.

A preview: right angles

The four spaces are not arranged at random. On the input side, the row space and the null space meet at a perfect right angle — they are orthogonal complements. Every row of A is perpendicular to every solution of A*x = 0, which is just a restatement of A*x = 0 row by row: each row dotted with x is zero. Together they fill all of R^n with nothing left over.