Visual Tools
Calculators
Tables
Mathematical Keyboard
Converters
Other Tools

How to Check If a Matrix Is Diagonalizable

An n×nn \times n matrix is diagonalizable exactly when it has nn independent eigenvectors. In practice: find the eigenvalues, and for each repeated one check that its eigenspace is as large as its multiplicity.

Recognize the task

The question asks "is AA diagonalizable?", "is AA similar to a diagonal matrix?", "does Rn\mathbb{R}^n have a basis of eigenvectors?", or asks for PDP−1PDP^{-1} and you want to know first whether it exists. Typical inputs: a triangular matrix with a repeated diagonal entry, a 2×22 \times 2 with a double eigenvalue, a symmetric matrix. Test: only repeated eigenvalues need work.
The criterion is in When Is a Matrix Diagonalizable?; the two multiplicities are defined in Algebraic and Geometric Multiplicity.

Ways to do it

Multiplicity comparisonalways decides
mg(λ)=n−rank⁡(A−λI)=ma(λ) for every λm_g(\lambda) = n - \operatorname{rank}(A - \lambda I) = m_a(\lambda) \ \text{for every } \lambda
This is the diagonalizability condition, necessary and sufficient.
Distinct eigenvaluesquick yes
n distinct eigenvalues⇒diagonalizablen \text{ distinct eigenvalues} \Rightarrow \text{diagonalizable}
Distinct eigenvalues give independent eigenvectors. A repeated eigenvalue does not mean no; it means check.
Real symmetricquick yes
A=AT⇒A=QDQTA = A^T \Rightarrow A = QDQ^T
The spectral theorem guarantees an orthonormal basis of eigenvectors.

Steps

StepOn A=(200020103)A = \begin{pmatrix} 2 & 0 & 0 \\ 0 & 2 & 0 \\ 1 & 0 & 3 \end{pmatrix}
1Find the eigenvalues with multiplicities. Use finding eigenvalues; a triangular matrix shows them on the diagonal.λ=2\lambda = 2 (ma=2m_a = 2), λ=3\lambda = 3 (ma=1m_a = 1)
2Apply the quick tests. If all nn eigenvalues are distinct, or AA is real symmetric, the answer is yes: stop.not distinct, not symmetric: continue
3Skip the simple eigenvalues. An eigenvalue with ma=1m_a = 1 always has mg=1m_g = 1.λ=3\lambda = 3 is fine
4Compute the geometric multiplicity of each repeated eigenvalue. Find the rank of A−λIA - \lambda I; then mg=n−rank⁡m_g = n - \operatorname{rank}.A−2IA - 2I has one nonzero row: mg(2)=3−1=2m_g(2) = 3 - 1 = 2
5Compare. Diagonalizable if mg=mam_g = m_a for every eigenvalue, equivalently if the mgm_g add up to nn.2=22 = 2 and 1=11 = 1: diagonalizable ✓

Worked examples

Example 1 · distinct eigenvalues

quick yes
A=(1423)A = \begin{pmatrix} 1 & 4 \\ 2 & 3 \end{pmatrix}: λ2−4λ−5=(λ−5)(λ+1)\lambda^2 - 4\lambda - 5 = (\lambda - 5)(\lambda + 1)
trace 44, determinant −5-5
5≠−15 \ne -1: two distinct eigenvalues, so diagonalizable
no eigenspace check needed

Example 2 · a double root that fails

2×2
A=(3103)A = \begin{pmatrix} 3 & 1 \\ 0 & 3 \end{pmatrix}: λ=3\lambda = 3, ma=2m_a = 2
rank⁡(A−3I)=1\operatorname{rank}(A - 3I) = 1, so mg=2−1=1m_g = 2 - 1 = 1
1<21 < 2: not diagonalizable
only one eigenvector direction

Example 3 · an upper triangular matrix

3×3
A=(231024005)A = \begin{pmatrix} 2 & 3 & 1 \\ 0 & 2 & 4 \\ 0 & 0 & 5 \end{pmatrix}: λ=2\lambda = 2 (ma=2m_a = 2) and 55
read the diagonal
A−2I=(031004003)A - 2I = \begin{pmatrix} 0 & 3 & 1 \\ 0 & 0 & 4 \\ 0 & 0 & 3 \end{pmatrix} has rank 22
two pivots
mg(2)=3−2=1<2m_g(2) = 3 - 2 = 1 < 2: not diagonalizable

See it happen

Diagonalizing means finding coordinate axes along which AA only stretches. Each eigenvalue supplies as many axes as its eigenspace has dimensions. If the axes add up to nn, they fill the space and AA is diagonal in that basis; if a repeated eigenvalue supplies fewer axes than its multiplicity, a shear is hiding in the matrix and no choice of axes removes it. See both outcomes in the matrix diagonalization visualizer.

Common mistakes

Repeated eigenvalue means no

✗ (200020103)\begin{pmatrix} 2 & 0 & 0 \\ 0 & 2 & 0 \\ 1 & 0 & 3 \end{pmatrix} repeats λ=2\lambda = 2, so it is not diagonalizable
✓ mg(2)=2=ma(2)m_g(2) = 2 = m_a(2): it is diagonalizable

Invertible means diagonalizable

✗ det⁡(3103)=9≠0\det\begin{pmatrix} 3 & 1 \\ 0 & 3 \end{pmatrix} = 9 \ne 0, so it is diagonalizable
✓ invertibility and diagonalizability are unrelated; this matrix is invertible and not diagonalizable
Conversely, (1000)\begin{pmatrix} 1 & 0 \\ 0 & 0 \end{pmatrix} is singular and already diagonal.

Using the rank of A

✗ mg(2)=3−rank⁡(A)m_g(2) = 3 - \operatorname{rank}(A)
✓ mg(2)=3−rank⁡(A−2I)m_g(2) = 3 - \operatorname{rank}(A - 2I)

Complex eigenvalues over the reals

✗ (0−110)\begin{pmatrix} 0 & -1 \\ 1 & 0 \end{pmatrix} is diagonalizable over R\mathbb{R}
✓ its eigenvalues are ±i\pm i, so it is diagonalizable over C\mathbb{C} only; over R\mathbb{R} use the real canonical form