The Hessian Matrix and Second Derivative Test

College Depth 69 in the knowledge graph I know this Set as goal
Unlocks 397 downstream topics
hessian second-derivative-test eigenvalues

Core Idea

The Hessian matrix H = [[f_xx, f_xy], [f_yx, f_yy]] contains all second partial derivatives. At a critical point, the determinant det(H) and trace tr(H) determine whether it is a local max (det > 0, f_xx > 0), local min (det > 0, f_xx < 0), or saddle point (det < 0).

Explainer

You already know how to find critical points of a multivariable function: set ∇f = 0 and solve. A critical point could be a local minimum, a local maximum, or a saddle point. In single-variable calculus, the second derivative test resolves this cleanly: f''(c) > 0 means the function curves upward, giving a local minimum; f''(c) < 0 means downward curvature, giving a local maximum. The Hessian matrix extends this idea to higher dimensions, packaging all second-order information into a matrix that captures curvature in every direction simultaneously.

For f: R² → R, the Hessian is H = [[f_xx, f_xy], [f_yx, f_yy]]. Since Clairaut's theorem guarantees f_xy = f_yx under continuity, H is symmetric. The question at a critical point is: does f curve upward in every direction (local min), downward in every direction (local max), or upward in some directions and downward in others (saddle)? That question is exactly whether H is positive definite, negative definite, or indefinite — which is determined by the signs of its eigenvalues. Positive definite means all eigenvalues are positive (upward curvature in every direction); negative definite means all eigenvalues are negative; indefinite means mixed signs.

For the 2×2 Hessian, the eigenvalue structure reduces to two computable numbers: det(H) = f_xx · f_yy − (f_xy)² and f_xx. If det(H) > 0 and f_xx > 0, the Hessian is positive definite: both eigenvalues are positive, the function curves upward in every direction from the critical point, and you have a local minimum. If det(H) > 0 and f_xx < 0, the Hessian is negative definite: both eigenvalues are negative, downward curvature everywhere, giving a local maximum. If det(H) < 0, the eigenvalues have opposite signs, giving a saddle point — one direction curves up, another down. When det(H) = 0, an eigenvalue is zero and the test is inconclusive.

The geometric intuition for saddle points is worth dwelling on. At a saddle the function increases along some paths through the critical point and decreases along others — like the center of a mountain pass, which is a local maximum along the ridge but a local minimum along the crossing direction. The Hessian test at such a point has one positive and one negative eigenvalue, hence negative determinant. In optimization applications, saddle points are critical to identify because gradient descent can stall near them, mistaking them for minima. For higher-dimensional functions, the full spectrum of the Hessian — not just its determinant — governs the classification, making eigenvalue analysis the natural tool.

Practice Questions 5 questions

Prerequisite Chain

Counting to 10Counting to 20Understanding ZeroThe Number ZeroCounting to FiveOne-to-One CorrespondenceCombining Small Groups Within 5Addition Within 10Addition Within 20Two-Digit Addition Without RegroupingTwo-Digit Addition with RegroupingAddition Within 100Repeated Addition as MultiplicationMultiplication Facts Within 100Division as Equal SharingDivision as Grouping (Measurement Division)Division: Grouping (Repeated Subtraction) ModelDivision: Fair Sharing ModelDivision as Equal SharingDivision as GroupingBasic Division FactsDivision Facts Within 100Two-Digit by One-Digit DivisionDivision with RemaindersRemainders and Quotients in DivisionDivision Word ProblemsIntroduction to Long DivisionFactors and MultiplesPrime and Composite NumbersEquivalent FractionsRelating Fractions and DecimalsDecimal Place ValueReading and Writing DecimalsComparing and Ordering DecimalsAdding and Subtracting DecimalsMultiplying DecimalsDividing DecimalsDividing FractionsMixed Number ArithmeticOrder of OperationsInteger Order of OperationsVariable ExpressionsCombining Like TermsOne-Step EquationsTwo-Step EquationsSolving Multi-Step EquationsEquations with Variables on Both SidesLiteral EquationsSlope-Intercept FormPoint-Slope FormWriting Linear EquationsParallel and Perpendicular Line SlopesGraphing Linear EquationsPiecewise FunctionsOne-Sided LimitsContinuity DefinitionLimit Definition of the DerivativePower RuleConstant Multiple and Sum/Difference RulesProduct RuleChain RuleHigher-Order DerivativesConcavity and Inflection PointsSecond Derivative TestCurve SketchingOptimization ProblemsCritical Points of Multivariable FunctionsCritical Points and Classification of ExtremaSecond Partial Test for Local Extrema (Hessian)The Hessian Matrix and Second Derivative Test

Longest path: 70 steps · 318 total prerequisite topics

Prerequisites (3)

Leads To (1)