Visual Tools
Calculators
Tables
Mathematical Keyboard
Converters
Other Tools

How to Find the Kernel and Image of a Linear Transformation

For T(x)=AxT(\mathbf{x}) = A\mathbf{x}, the kernel is the null space of AA (the inputs sent to 0\mathbf{0}) and the image is the column space of AA (every possible output). One row reduction gives a basis for both.

Recognize the task

The question says kernel, null space, image, range, injective (one-to-one) or surjective (onto): "find bases for ker⁡T\ker T and Im⁡T\operatorname{Im} T", "is TT one-to-one?". Typical inputs: a 3×43 \times 4 matrix, a formula like T(x,y)=(x+y, 2x+2y)T(x, y) = (x + y, \ 2x + 2y), a tall matrix from R2\mathbb{R}^2 to R3\mathbb{R}^3. Test: write the standard matrix first; everything is read from its echelon form.
Kernel and image are defined in The Kernel and The Image; their dimensions obey rank–nullity.

Ways to do it

Row reductionany matrix
ker⁡T={x:Ax=0},Im⁡T=span⁡(pivot columns of A)\ker T = \{\mathbf{x} : A\mathbf{x} = \mathbf{0}\}, \quad \operatorname{Im} T = \operatorname{span}(\text{pivot columns of } A)
Reduce with Gaussian elimination and read both answers off one echelon form.
Determinant shortcutsquare A
det⁡(A)≠0⇒ker⁡T={0}, Im⁡T=Rn\det(A) \ne 0 \Rightarrow \ker T = \{\mathbf{0}\}, \ \operatorname{Im} T = \mathbb{R}^n
An invertible matrix is both one-to-one and onto (the equal-dimension case).
Rank–nullity for dimensionsdimensions only
dim⁡ker⁡T+dim⁡Im⁡T=n\dim\ker T + \dim\operatorname{Im} T = n
Find either dimension and the other follows, with nn the number of columns.

Steps

StepOn A=(120−124113610)A = \begin{pmatrix} 1 & 2 & 0 & -1 \\ 2 & 4 & 1 & 1 \\ 3 & 6 & 1 & 0 \end{pmatrix}
1Write the standard matrix. If TT is given by a formula, build AA as in finding the transformation matrix.T:R4→R3T: \mathbb{R}^4 \to \mathbb{R}^3
2Row reduce to reduced echelon form. Mark the pivot columns.(120−100130000)\begin{pmatrix} 1 & 2 & 0 & -1 \\ 0 & 0 & 1 & 3 \\ 0 & 0 & 0 & 0 \end{pmatrix}: pivots in columns 11 and 33
3Kernel: solve with the free variables. Each free variable gives one basis vector of the parametric solution.x2=sx_2 = s, x4=tx_4 = t: ker⁡T=span⁡{(−2,1,0,0),(1,0,−3,1)}\ker T = \operatorname{span}\{(-2, 1, 0, 0), (1, 0, -3, 1)\}
4Image: take the pivot columns of the original AA. Not the columns of the echelon form: row operations change the column space.Im⁡T=span⁡{(1,2,3),(0,1,1)}\operatorname{Im} T = \operatorname{span}\{(1, 2, 3), (0, 1, 1)\}
5Check the dimensions and answer the question. Nullity plus rank equals the number of columns. One-to-one exactly when the kernel is {0}\{\mathbf{0}\} (the kernel criterion); onto exactly when the rank equals the number of rows.2+2=42 + 2 = 4 ✓; not one-to-one, not onto

Worked examples

Example 1 · a line onto a line

2×2
T(x,y)=(x+y, 2x+2y)T(x, y) = (x + y, \ 2x + 2y): A=(1122)→(1100)A = \begin{pmatrix} 1 & 1 \\ 2 & 2 \end{pmatrix} \to \begin{pmatrix} 1 & 1 \\ 0 & 0 \end{pmatrix}
one pivot, in column 1
ker⁡T=span⁡{(−1,1)}\ker T = \operatorname{span}\{(-1, 1)\}, Im⁡T=span⁡{(1,2)}\operatorname{Im} T = \operatorname{span}\{(1, 2)\}
nullity 11 plus rank 11

Example 2 · one-to-one but not onto

3×2
A=(100111)A = \begin{pmatrix} 1 & 0 \\ 0 & 1 \\ 1 & 1 \end{pmatrix}: pivots in both columns
no free variables
ker⁡T={0}\ker T = \{\mathbf{0}\}: one-to-one
Im⁡T=span⁡{(1,0,1),(0,1,1)}\operatorname{Im} T = \operatorname{span}\{(1, 0, 1), (0, 1, 1)\}, a plane in R3\mathbb{R}^3: not onto
rank 2<32 < 3

Example 3 · square but singular

3×3
A=(112011123)A = \begin{pmatrix} 1 & 1 & 2 \\ 0 & 1 & 1 \\ 1 & 2 & 3 \end{pmatrix}, det⁡(A)=0\det(A) = 0
row 3 = row 1 + row 2
RREF (101011000)\begin{pmatrix} 1 & 0 & 1 \\ 0 & 1 & 1 \\ 0 & 0 & 0 \end{pmatrix}: x1=−x3x_1 = -x_3, x2=−x3x_2 = -x_3
ker⁡T=span⁡{(−1,−1,1)}\ker T = \operatorname{span}\{(-1, -1, 1)\}, Im⁡T=span⁡{(1,0,1),(1,1,2)}\operatorname{Im} T = \operatorname{span}\{(1, 0, 1), (1, 1, 2)\}
pivot columns 1, 2 of AA
See both subspaces in the kernel and image 2D visualizer.

See it happen

The kernel lives in the input space and the image in the output space. Every independent direction that AA crushes to 0\mathbf{0} costs the image one dimension: in the trace example a plane of inputs collapses, so four input dimensions leave a two-dimensional image. That trade-off is the rank–nullity count. The four fundamental subspaces visualizer places the null space and the column space side by side.

Common mistakes

Pivot columns of the echelon form

✗ Im⁡T=span⁡{(1,0,0),(0,1,0)}\operatorname{Im} T = \operatorname{span}\{(1, 0, 0), (0, 1, 0)\}, read from the RREF
✓ use columns 11 and 33 of the original AA: (1,2,3)(1, 2, 3) and (0,1,1)(0, 1, 1)

Kernel from the rows

✗ ker⁡T=span⁡{(1,2,0,−1),(0,0,1,3)}\ker T = \operatorname{span}\{(1, 2, 0, -1), (0, 0, 1, 3)\}
✓ those rows span the row space; the kernel is the solution set of Ax=0A\mathbf{x} = \mathbf{0}

Mixing up where they live

✗ ker⁡T⊆R3\ker T \subseteq \mathbb{R}^3 for T:R4→R3T: \mathbb{R}^4 \to \mathbb{R}^3
✓ ker⁡T⊆R4\ker T \subseteq \mathbb{R}^4 (inputs) and Im⁡T⊆R3\operatorname{Im} T \subseteq \mathbb{R}^3 (outputs)

Rank–nullity with the rows

✗ nullity =3−rank⁡=3−2=1= 3 - \operatorname{rank} = 3 - 2 = 1
✓ nullity =n−rank⁡=4−2=2= n - \operatorname{rank} = 4 - 2 = 2, with nn the number of columns