Eigenvalue Criterion for Definiteness
What You'll Learn
You'll learn how to use eigenvalues as a quick, reliable test to classify whether a symmetric matrix is positive definite, negative definite, or indefinite.
The Core Idea
In the previous lesson, you learned that matrices can be classified by how their associated quadratic forms behave—whether they're always positive, always negative, or mixed. But checking all possible vectors is impractical. Eigenvalues give us a shortcut.
For a symmetric matrix A, the eigenvalues encode all the information about definiteness:
- Positive Definite: All eigenvalues are positive (λ > 0)
- Negative Definite: All eigenvalues are negative (λ < 0)
- Positive Semidefinite: All eigenvalues are non-negative (λ ≥ 0), at least one is zero
- Negative Semidefinite: All eigenvalues are non-positive (λ ≤ 0), at least one is zero
- Indefinite: Eigenvalues have mixed signs (some positive, some negative)
Why This Works
Remember the Spectral Theorem: every symmetric matrix can be orthogonally diagonalized. In the right coordinate system (using eigenvectors as basis), your matrix becomes diagonal with eigenvalues on the diagonal. The quadratic form xᵀA****x becomes a simple weighted sum of squares: λ₁y₁² + λ₂y₂² + ... + λₙyₙ².
If all λᵢ > 0, this sum is always positive (positive definite). If signs are mixed, the sum can be positive or negative depending on y (indefinite).
Practical Use
Instead of testing infinitely many vectors, just:
- Find the eigenvalues of your symmetric matrix
- Check their signs
- Classify immediately
This is especially powerful for optimization problems, where definiteness of the Hessian matrix tells you whether you have a minimum, maximum, or saddle point.
Key Takeaway: For symmetric matrices, eigenvalue signs directly reveal definiteness—all positive means positive definite, mixed signs mean indefinite, making classification computational and exact.