Visual Tools
Calculators
Tables
Mathematical Keyboard
Converters
Other Tools

How to Find the Four Fundamental Subspaces

Every m×nm \times n matrix AA of rank rr has four subspaces: the column space and the left null space in Rm\mathbb{R}^m, the row space and the null space in Rn\mathbb{R}^n. Row reduce AA once and read off three of them; the fourth comes from ATA^T.

Recognize the task

Requests such as "find bases for Col⁡(A)\operatorname{Col}(A), Row⁡(A)\operatorname{Row}(A), Nul⁡(A)\operatorname{Nul}(A) and Nul⁡(AT)\operatorname{Nul}(A^T)", "find the null space", "find the rank and nullity", or "describe the image and kernel" of a matrix. Test: is a single matrix given, with questions about the spaces it determines?
The four subspaces are introduced in the fundamental subspaces lesson.

Steps

StepOn A=(123246134)A = \begin{pmatrix} 1 & 2 & 3 \\ 2 & 4 & 6 \\ 1 & 3 & 4 \end{pmatrix}
1Row reduce A to RREF and find the rank. Use reduced row echelon form. The rank rr is the number of pivots.(101011000)\begin{pmatrix} 1 & 0 & 1 \\ 0 & 1 & 1 \\ 0 & 0 & 0 \end{pmatrix},  r=2\ r = 2
2Column space: pivot columns of the original A. Dimension rr, a subspace of Rm\mathbb{R}^m.Col⁡(A)=Span⁡{(1,2,1),(2,4,3)}\operatorname{Col}(A) = \operatorname{Span}\{(1, 2, 1), (2, 4, 3)\}
3Row space: nonzero rows of the RREF. Dimension rr, a subspace of Rn\mathbb{R}^n.Row⁡(A)=Span⁡{(1,0,1),(0,1,1)}\operatorname{Row}(A) = \operatorname{Span}\{(1, 0, 1), (0, 1, 1)\}
4Null space: solve Ax = 0. One vector per free variable (see homogeneous systems). Dimension n−rn - r.x1=−x3x_1 = -x_3, x2=−x3x_2 = -x_3:  Nul⁡(A)=Span⁡{(−1,−1,1)}\ \operatorname{Nul}(A) = \operatorname{Span}\{(-1, -1, 1)\}
5Left null space: solve the transpose system. Row reduce ATA^T and solve ATy=0A^T\mathbf{y} = \mathbf{0}. Dimension m−rm - r.Nul⁡(AT)=Span⁡{(−2,1,0)}\operatorname{Nul}(A^T) = \operatorname{Span}\{(-2, 1, 0)\}
6Check dimensions and right angles. The dimension count: r+(n−r)=nr + (n - r) = n and r+(m−r)=mr + (m - r) = m. The null space is perpendicular to the row space, the left null space to the column space.2+1=32 + 1 = 3, 2+1=32 + 1 = 3; (−1,−1,1)⋅(1,0,1)=0(-1, -1, 1) \cdot (1, 0, 1) = 0 ✓, (−2,1,0)⋅(1,2,1)=0(-2, 1, 0) \cdot (1, 2, 1) = 0 ✓

Worked examples

Example 1 · a rank-one matrix

2×2
A=(1224)A = \begin{pmatrix} 1 & 2 \\ 2 & 4 \end{pmatrix}, RREF (1200)\begin{pmatrix} 1 & 2 \\ 0 & 0 \end{pmatrix}, r=1r = 1
one pivot
Col⁡=Span⁡{(1,2)}\operatorname{Col} = \operatorname{Span}\{(1, 2)\},  Row⁡=Span⁡{(1,2)}\ \operatorname{Row} = \operatorname{Span}\{(1, 2)\}
symmetric matrix: they coincide
Nul⁡(A)=Nul⁡(AT)=Span⁡{(−2,1)}\operatorname{Nul}(A) = \operatorname{Nul}(A^T) = \operatorname{Span}\{(-2, 1)\}
check: −2+2=0-2 + 2 = 0

Example 2 · full row rank

2×3
A=(10201−1)A = \begin{pmatrix} 1 & 0 & 2 \\ 0 & 1 & -1 \end{pmatrix}, already RREF, r=2r = 2
pivots in columns 1 and 2
Col⁡(A)=R2\operatorname{Col}(A) = \mathbb{R}^2,  Row⁡\ \operatorname{Row} basis (1,0,2)(1, 0, 2), (0,1,−1)(0, 1, -1)
pivot in every row
Nul⁡(A)=Span⁡{(−2,1,1)}\operatorname{Nul}(A) = \operatorname{Span}\{(-2, 1, 1)\},  Nul⁡(AT)={0}\ \operatorname{Nul}(A^T) = \{\mathbf{0}\}
m−r=0m - r = 0

Example 3 · dimensions only

no matrix entries
AA is 4×64 \times 6 with rank 33
m=4m = 4, n=6n = 6, r=3r = 3
dim⁡Col⁡=dim⁡Row⁡=3\dim\operatorname{Col} = \dim\operatorname{Row} = 3
both equal the rank
dim⁡Nul⁡(A)=6−3=3\dim\operatorname{Nul}(A) = 6 - 3 = 3,  dim⁡Nul⁡(AT)=4−3=1\ \dim\operatorname{Nul}(A^T) = 4 - 3 = 1
rank–nullity for A and its transpose
See all four spaces of a matrix at once in the four fundamental subspaces visualizer.

See it happen

Picture AA as a map from Rn\mathbb{R}^n to Rm\mathbb{R}^m. On the input side, Rn\mathbb{R}^n splits into the row space and the null space at right angles; the null space is everything AA sends to zero. On the output side, Rm\mathbb{R}^m splits into the column space, everything AA can reach, and the left null space perpendicular to it. Both splits share the same number rr.

Common mistakes

Column space from the RREF columns

✗ Col⁡(A)=Span⁡{(1,0,0),(0,1,0)}\operatorname{Col}(A) = \operatorname{Span}\{(1, 0, 0), (0, 1, 0)\}
✓ use the original pivot columns (1,2,1)(1, 2, 1) and (2,4,3)(2, 4, 3); row operations change the column space

Left null space from A instead of the transpose

✗ Nul⁡(AT)=Span⁡{(−1,−1,1)}\operatorname{Nul}(A^T) = \operatorname{Span}\{(-1, -1, 1)\}
✓ that is Nul⁡(A)\operatorname{Nul}(A); solve ATy=0A^T\mathbf{y} = \mathbf{0} to get (−2,1,0)(-2, 1, 0)

Mixing up the ambient spaces

✗ the null space of a 2×32 \times 3 matrix is a subspace of R2\mathbb{R}^2
✓ it lives in R3\mathbb{R}^3, with the row space; column and left null space live in R2\mathbb{R}^2

Dimensions that do not add up

✗ rank 22 and a two-vector null space basis for a 3×33 \times 3 matrix
✓ rank plus nullity is n=3n = 3; recheck the free variables