How to find a column space and decide whether Ax = b has a solution

A matrix-vector product is a combination of the columns of the matrix. That means we can solve Ax = b exactly when those columns can combine to make b. Let’s find a column-space basis and connect it with the null space.

Read the pivot positions from a reduced matrix

A=[21314-28456-2-3]

Row reduction gives the following matrix.

[100-331010-8310011531]

The pivots are in columns 1, 2, and 3. Use those positions to select the corresponding columns from the original matrix A.

Choose the basis from the original columns

A column-space basis is (2,4,5), (1,-2,6), and (3,8,-2). These are three independent vectors in three-dimensional space, so the column space is all of three-dimensional space. For this matrix, Ax=b has a solution for every three-component vector b.

Use the free column to find the null-space dimension

There are four variable columns and three pivots. The fourth variable is free. Letting x4=31t in the reduced system gives

x=t(3,8,-15,31)

So the null space has dimension 1 and the column space has dimension 3. The dimensions add to the number of columns, 4. More generally, a matrix with n columns and r pivots has null-space dimension n minus r.