Second Partial Test and the Hessian

College Depth 69 in the knowledge graph I know this Set as goal
Unlocks 395 downstream topics
hessian classification

Core Idea

The Hessian H = [[f_xx, f_xy], [f_xy, f_yy]] classifies critical points. If det(H) > 0 and f_xx > 0, it's a local minimum. If det(H) > 0 and f_xx < 0, it's a maximum. If det(H) < 0, it's a saddle point.

Explainer

In single-variable calculus, the second derivative test classifies critical points: if f′(a) = 0 and f′′(a) > 0, the graph is concave up at a, so a is a local minimum; if f′′(a) < 0, it's concave down, so a is a local maximum. For functions of two variables, you already know that critical points satisfy ∇f = 0 (both partial derivatives are zero). The Hessian matrix is the multivariable analogue of the second derivative — it captures the curvature of the surface in all directions simultaneously.

The Hessian at a critical point (a, b) is the 2×2 matrix H = [[f_xx, f_xy], [f_xy, f_yy]], where all second partials are evaluated at (a, b). By Clairaut's theorem (which you have studied), f_xy = f_yx under smoothness conditions, so H is symmetric. The entries measure curvature in specific directions: f_xx is the concavity along the x-axis, f_yy is the concavity along the y-axis, and f_xy measures how the x-slope changes as y varies — a cross-curvature or "twist."

The classification hinges on the determinant D = det(H) = f_xx · f_yy − (f_xy)². Think of D as the product of curvatures in the two "principal" directions of the surface (the directions where the mixed terms vanish). If D > 0, the surface curves the same way in all directions at that point — it's bowl-shaped (either upward or downward), giving a local extremum. The sign of f_xx (or equivalently f_yy when D > 0) tells you which: positive means bowl-opening-upward (local minimum), negative means bowl-opening-downward (local maximum). If D < 0, the surface curves upward in one direction and downward in another — a saddle point, like the surface of a mountain pass that goes up toward ridges but down toward valleys.

Here is the geometric intuition for why D = f_xx · f_yy − (f_xy)² is the right quantity. A quadratic function ax² + 2bxy + cy² (which approximates f near a critical point) curves upward in every direction if and only if a > 0 and ac − b² > 0. The determinant condition is exactly testing whether this quadratic form is positive definite (curves up everywhere), negative definite (curves down everywhere), or indefinite (saddle). When D = 0, the test is inconclusive — you need higher-order information.

The Hessian matrix generalizes naturally beyond two variables: for a function of n variables, H is an n×n symmetric matrix of second partials, and the classification criterion generalizes to checking whether H is positive definite (local min), negative definite (local max), or indefinite (saddle). This is the heart of multivariable optimization, and understanding the Hessian as a curvature object — not just as a formula to memorize — is what makes it usable in more complex settings like constrained optimization and machine learning loss landscapes.

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)Second Partial Test and the Hessian

Longest path: 70 steps · 320 total prerequisite topics

Prerequisites (3)

Leads To (1)