Diagonalizable Matrix: A matrix A is diagonalizable if A = PDP⁻¹ where D is diagonal and P is invertible
Eigenvalue λ: Scalar satisfying Av = λv for some non-zero vector v
Eigenvector v: Non-zero vector satisfying Av = λv
Characteristic Polynomial: det(A - λI) = 0
- Find characteristic polynomial: det(A - λI) = 0
- Solve for eigenvalues: Find roots of characteristic equation
- Find eigenvectors: For each λ, solve (A - λI)v = 0
- Form matrices: P from eigenvectors, D from eigenvalues
- Verify: Check A = PDP⁻¹
• Characteristic equation: det(A - λI) = 0
• Eigenvalue equation: Av = λv
• Diagonalization: A = PDP⁻¹
• Matrix powers: A^n = PD^nP⁻¹
Diagonalization: Express A as A = PDP⁻¹ where D is diagonal matrix of eigenvalues
- Find characteristic polynomial det(A - λI) = 0
- Find eigenvalues λ
- Find eigenvectors v for each λ
- Form P from eigenvectors, D from eigenvalues
- Verify A = PDP⁻¹
A - λI =
det(A - λI) = (4-λ)(1-λ) - (-2)(1) = (4-λ)(1-λ) + 2
= 4 - 4λ - λ + λ² + 2 = λ² - 5λ + 6
λ² - 5λ + 6 = 0 → (λ - 2)(λ - 3) = 0
Therefore: λ₁ = 2, λ₂ = 3
(A - 2I)v = 0 →
This gives: 2x - 2y = 0 and x - y = 0, so x = y
Eigenvector: v₁ =
(A - 3I)v = 0 →
This gives: x - 2y = 0, so x = 2y
Eigenvector: v₂ =
P = [v₁ v₂] =
D =
A = PDP⁻¹ with distinct eigenvalues λ₁ = 2, λ₂ = 3
• Characteristic equation: Set det(A - λI) = 0
• Eigenvector equation: Solve (A - λI)v = 0
• Distinct eigenvalues: Guarantee diagonalizability
3x3 Diagonalization: Same process extended to higher dimensions
det(A - λI) = det
Since this is lower triangular matrix, determinant is product of diagonal: (2-λ)(3-λ)(1-λ)
(2-λ)(3-λ)(1-λ) = 0 → λ₁ = 2, λ₂ = 3, λ₃ = 1
(A - 2I)v = 0 →
This gives: x + y = 0 and -x - z = 0, so y = -x and z = -x
Eigenvector: v₁ =
(A - 3I)v = 0 →
This gives: -x = 0 and -x - 2z = 0, so x = 0 and z = 0
Eigenvector: v₂ =
(A - I)v = 0 →
This gives: x = 0 and x + 2y = 0, so x = 0 and y = 0
Eigenvector: v₃ =
A = PDP⁻¹ with 3 distinct eigenvalues, so A is diagonalizable
• Triangular matrix: Eigenvalues are diagonal entries
• Systematic solving: Solve (A - λI)v = 0 for each eigenvalue
• Linear independence: Verify eigenvectors are independent
Verification: Confirm diagonalization by computing PDP⁻¹ and comparing to A
P =
det(P) = (1)(1) - (2)(1) = 1 - 2 = -1
P⁻¹ = (1/-1)
First compute PD:
Then compute (PD)P⁻¹:
=
PDP⁻¹ = A, confirming our diagonalization is correct
• Matrix inverse: P⁻¹ = (1/det(P)) × adjugate(P)
• Verification: Essential to confirm correctness
• Associativity: (AB)C = A(BC) allows step-by-step computation
Non-diagonalizable Matrix: Matrix that doesn't have enough linearly independent eigenvectors
det(A - λI) = det
(2-λ)² = 0 → λ = 2 (with algebraic multiplicity 2)
(A - 2I)v = 0 →
This gives: y = 0, so x can be any value
Eigenspace is span{
Algebraic multiplicity of λ = 2 is 2
Geometric multiplicity (dimension of eigenspace) is 1
Since algebraic multiplicity ≠ geometric multiplicity, A is not diagonalizable
A cannot be diagonalized because it lacks sufficient linearly independent eigenvectors
• Algebraic multiplicity: Multiplicity of eigenvalue as root of characteristic polynomial
• Geometric multiplicity: Dimension of eigenspace
• Diagonalizability criterion: AM = GM for all eigenvalues
Matrix Powers via Diagonalization: A^n = PD^nP⁻¹ is much more efficient than repeated multiplication
P =
Since D is diagonal, D⁵ =
First compute PD⁵:
Then compute (PD⁵)P⁻¹:
=
A⁵ computed efficiently using diagonalization
• Power efficiency: A^n = PD^nP⁻¹ avoids n-1 matrix multiplications
• Diagonal powers: Just raise diagonal elements to power
• Practical application: Useful for dynamical systems and Markov chains