Course contentsShow
← Mathematics
Lesson 2292 of 3,12147. Linear Algebra — Applications and Advanced TopicsFree lesson

Positive Definite, Negative Definite, and Indefinite Matrices

Classify symmetric matrices based on the sign of their quadratic forms for all nonzero vectors.

Positive Definite, Negative Definite, and Indefinite Matrices

What you'll learn: How to classify symmetric matrices by examining whether their quadratic forms are always positive, always negative, or mixed in sign.

The Core Idea

Remember that a quadratic form is an expression like xᵀAx where A is a symmetric matrix and x is any nonzero vector. This scalar quantity changes as you choose different vectors x. The definiteness of a matrix tells us whether this quadratic form consistently produces positive values, negative values, or can be either depending on x.

The Three Main Classifications

Positive Definite

A symmetric matrix A is positive definite if xᵀAx > 0 for all nonzero vectors x. Think of this as a matrix that always "stretches" vectors in a way that produces positive energy—like a bowl that curves upward in every direction.

Equivalent condition: All eigenvalues of A are strictly positive.

Negative Definite

A symmetric matrix A is negative definite if xᵀAx < 0 for all nonzero vectors x. This is the opposite: like an upside-down bowl curving downward everywhere.

Equivalent condition: All eigenvalues of A are strictly negative.

Indefinite

A symmetric matrix A is indefinite if xᵀAx takes both positive and negative values for different choices of x. Picture a saddle surface: curving upward in some directions, downward in others.

Equivalent condition: A has both positive and negative eigenvalues.

Why It Matters

Definiteness tells you about the shape of the surface defined by the quadratic form. In optimization, positive definite matrices indicate a local minimum, while negative definite suggests a local maximum. Indefinite matrices signal a saddle point.

Key Takeaway: A symmetric matrix's definiteness is determined by testing the sign of xᵀAx for all nonzero x, equivalently by checking whether all eigenvalues are positive, negative, or mixed.