Eigenvalue & Eigenvector Calculator

Anatomy of Eigenvalues & The Characteristic Equation

The German prefix eigen translates roughly to "own," "inherent," "characteristic," or "peculiar." In linear algebra and matrix theory, an eigenvector v of a square matrix A represents an invariant directional axis in vector space: when the linear transformation A acts upon v, the vector is not rotated, but simply stretched, compressed, or flipped by a scalar factor λ.

To solve this equation for unknown values of λ and v, rearrange terms into a homogeneous linear system:

For a non-trivial solution v ≠ 0 to exist, the matrix operator (A - λI) must have a non-trivial kernel (nullspace). By matrix inversion theory and Cramer's rule, a matrix has a non-trivial nullspace if and only if its determinant equals zero. Evaluating this determinant yields an n-th degree polynomial in λ whose roots are precisely the eigenvalues of matrix A. Once eigenvalues are computed, finding the corresponding directional nullspace vectors is carried out using our dedicated Eigenvector Calculator.

Computing 2x2 Matrix Eigenvalues & Trace-Determinant Formula

For any general 2×2 matrix A = [[a, b], [c, d]], expanding the determinant yields a streamlined quadratic polynomial:

Applying the quadratic formula yields the direct closed-form expression for both eigenvalues (solve any second-degree polynomial directly with our Quadratic Formula Calculator):

The determinant det(A) dictates area scaling and invertibility (calculate arbitrary determinants with our Matrix Determinant Calculator). The discriminant of this quadratic is Δ = tr(A)² - 4det(A). By testing this discriminant using our Discriminant Calculator, you can immediately determine whether eigenvalues are real and distinct (Δ > 0), repeated (Δ = 0), or complex conjugate pairs (Δ < 0):

Computing 3x3 Matrix Eigenvalues & Invariant Expansion

For any 3×3 square matrix, the characteristic equation expands into a cubic polynomial:

The coefficients of this cubic polynomial represent fundamental geometric invariants that remain unchanged under any similarity transformation P⁻¹AP:

• Trace tr(A): The sum of the diagonal entries: tr(A) = a₁₁ + a₂₂ + a₃₃. This matches the sum of the three eigenvalues: λ₁ + λ₂ + λ₃.
• Sum of Principal Minors (M): The sum of the three 2×2 principal diagonal sub-determinants:
  M = M₁₁ + M₂₂ + M₃₃ = (a₂₂a₃₃ - a₂₃a₃₂) + (a₁₁a₃₃ - a₁₃a₃₁) + (a₁₁a₂₂ - a₁₂a₂₁) This equals the sum of pairwise products of eigenvalues: λ₁λ₂ + λ₂λ₃ + λ₁λ₃.
• Determinant det(A): The full 3×3 determinant evaluated via cofactor expansion. This equals the product of the three eigenvalues: λ₁ · λ₂ · λ₃.

Once these coefficients are determined, the roots of the cubic characteristic polynomial can be resolved through the rational root theorem, synthetic division, or our automated Cubic Equation Solver.

Shortcut Theorem for Triangular & Diagonal Matrices

When a matrix is diagonal, upper-triangular, or lower-triangular, calculating eigenvalues requires no polynomial solving or matrix inversion whatsoever. Because all entries either above or below the main diagonal are zero, the determinant of (A - λI) is simply the product of its diagonal entries:

Consequently, the eigenvalues of any triangular or diagonal matrix are precisely the entries situated on its main diagonal:

This fundamental theorem is the primary reason why advanced numerical linear algebra libraries (such as LAPACK and SciPy) utilize orthogonal similarity transformations (the QR algorithm) to reduce dense matrices to Schur triangular form.

Finding Eigenvectors via the Nullspace Kernel (A - λI)v = 0

Once an eigenvalue λ is known, its associated eigenvectors belong to the nullspace (kernel) of the matrix (A - λI).

Because det(A - λI) = 0, the rows of (A - λI) are linearly dependent, creating at least one free variable:

For larger matrices, determining the nullspace kernel requires Gaussian row-reduction of the homogeneous system; you can automate these steps with our Row Echelon Form Calculator and System of Equations Calculator.

Algebraic vs Geometric Multiplicity & Defective Matrices

When dealing with matrices that possess repeated eigenvalues, linear algebra distinguishes between two fundamental types of multiplicity:

It is a universal mathematical theorem that 1 ≤ GM ≤ AM. When GM < AM, the matrix fails to produce enough eigenvectors to span the space and is termed a defective matrix. Defective matrices cannot be diagonalized and require generalized eigenvectors and Jordan Canonical Form.

The Spectral Theorem for Symmetric Matrices

One of the crown jewels of linear algebra is the Spectral Theorem for real symmetric matrices (Aᵀ = A). It guarantees three remarkable properties:

• All Real Eigenvalues: Every eigenvalue of a real symmetric matrix is guaranteed to be a real number; complex conjugate eigenvalues are impossible.
• Orthogonal Eigenvectors: Eigenvectors corresponding to distinct eigenvalues are mutually perpendicular (orthogonal): v_i · v_j = 0, easily verified using our Dot Product Calculator.
• Orthogonal Diagonalizability: The matrix can always be factored as A = Q D Qᵀ, where Q is an orthonormal rotation/reflection matrix and D is a pure diagonal matrix of eigenvalues.

Matrix Diagonalization and Power Computation (A^k = P D^k P⁻¹)

Multiplying a large matrix by itself dozens or thousands of times (e.g., computing A¹⁰⁰ in Markov chains or economic simulations) is computationally brutal. Eigenvalue decomposition makes this virtually instantaneous:

To invert the eigenvector modal matrix P and perform automated matrix multiplications, use our Matrix Inverse Calculator and Matrix Multiplication Calculator.

The Cayley-Hamilton Theorem & Matrix Inverses

The Cayley-Hamilton Theorem asserts that every square matrix satisfies its own characteristic polynomial equation. That is, if the characteristic equation of A is p(λ) = det(A - λI) = 0, then substituting the matrix A for λ yields the zero matrix:

This provides an extraordinarily fast algebraic method to compute matrix inverses without cofactor expansion or row reduction:

High-Dimensional Numerical Methods (QR Algorithm & Power Iteration)

In industrial engineering, machine learning, and quantum mechanics, matrices commonly possess dimensions from 100 × 100 to millions of entries (such as Google's PageRank transition matrix or quantum mechanical Hamiltonians). Computing characteristic polynomials for such systems is impossible because:

• Abel-Ruffini Theorem: No algebraic formula with radicals exists for general polynomials of degree 5 or higher.
• Wilkinson's Condition Problem: The roots of high-degree polynomials are notoriously ill-conditioned; infinitesimal floating-point perturbations in matrix entries can cause catastrophic errors in polynomial root calculations.

Consequently, modern computational linear algebra libraries employ robust iterative algorithms:

Dynamical Systems Stability & S-Plane State-Space

In control systems, electrical circuit analysis, and aerospace flight stability, linear systems of differential equations are represented as dx/dt = Ax. The eigenvalues of A dictate the physical behavior of the system:

Step-by-Step Worked Examples (Symmetric, Rotation, 3x3)

Common Operational Errors & Diagnostic Pitfalls Matrix